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>
75 * This structure really needs to be cleaned up.
76 * Most of it is for TCP, and not used by any of
77 * the other protocols.
80 /* Define this to get the SOCK_DBG debugging facility. */
81 #define SOCK_DEBUGGING
83 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
84 printk(KERN_DEBUG msg); } while (0)
86 /* Validate arguments and do nothing */
87 static inline __printf(2, 3)
88 void SOCK_DEBUG(const struct sock
*sk
, const char *msg
, ...)
93 /* This is the per-socket lock. The spinlock provides a synchronization
94 * between user contexts and software interrupt processing, whereas the
95 * mini-semaphore synchronizes multiple users amongst themselves.
100 wait_queue_head_t wq
;
102 * We express the mutex-alike socket_lock semantics
103 * to the lock validator by explicitly managing
104 * the slock as a lock variant (in addition to
107 #ifdef CONFIG_DEBUG_LOCK_ALLOC
108 struct lockdep_map dep_map
;
116 typedef __u32 __bitwise __portpair
;
117 typedef __u64 __bitwise __addrpair
;
120 * struct sock_common - minimal network layer representation of sockets
121 * @skc_daddr: Foreign IPv4 addr
122 * @skc_rcv_saddr: Bound local IPv4 addr
123 * @skc_hash: hash value used with various protocol lookup tables
124 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
125 * @skc_dport: placeholder for inet_dport/tw_dport
126 * @skc_num: placeholder for inet_num/tw_num
127 * @skc_family: network address family
128 * @skc_state: Connection state
129 * @skc_reuse: %SO_REUSEADDR setting
130 * @skc_reuseport: %SO_REUSEPORT setting
131 * @skc_bound_dev_if: bound device index if != 0
132 * @skc_bind_node: bind hash linkage for various protocol lookup tables
133 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
134 * @skc_prot: protocol handlers inside a network family
135 * @skc_net: reference to the network namespace of this socket
136 * @skc_node: main hash linkage for various protocol lookup tables
137 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
138 * @skc_tx_queue_mapping: tx queue number for this connection
139 * @skc_flags: place holder for sk_flags
140 * %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
141 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
142 * @skc_incoming_cpu: record/match cpu processing incoming packets
143 * @skc_refcnt: reference count
145 * This is the minimal network layer representation of sockets, the header
146 * for struct sock and struct inet_timewait_sock.
149 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
150 * address on 64bit arches : cf INET_MATCH()
153 __addrpair skc_addrpair
;
156 __be32 skc_rcv_saddr
;
160 unsigned int skc_hash
;
161 __u16 skc_u16hashes
[2];
163 /* skc_dport && skc_num must be grouped as well */
165 __portpair skc_portpair
;
172 unsigned short skc_family
;
173 volatile unsigned char skc_state
;
174 unsigned char skc_reuse
:4;
175 unsigned char skc_reuseport
:1;
176 unsigned char skc_ipv6only
:1;
177 unsigned char skc_net_refcnt
:1;
178 int skc_bound_dev_if
;
180 struct hlist_node skc_bind_node
;
181 struct hlist_node skc_portaddr_node
;
183 struct proto
*skc_prot
;
184 possible_net_t skc_net
;
186 #if IS_ENABLED(CONFIG_IPV6)
187 struct in6_addr skc_v6_daddr
;
188 struct in6_addr skc_v6_rcv_saddr
;
191 atomic64_t skc_cookie
;
193 /* following fields are padding to force
194 * offset(struct sock, sk_refcnt) == 128 on 64bit arches
195 * assuming IPV6 is enabled. We use this padding differently
196 * for different kind of 'sockets'
199 unsigned long skc_flags
;
200 struct sock
*skc_listener
; /* request_sock */
201 struct inet_timewait_death_row
*skc_tw_dr
; /* inet_timewait_sock */
204 * fields between dontcopy_begin/dontcopy_end
205 * are not copied in sock_copy()
208 int skc_dontcopy_begin
[0];
211 struct hlist_node skc_node
;
212 struct hlist_nulls_node skc_nulls_node
;
214 int skc_tx_queue_mapping
;
216 int skc_incoming_cpu
;
218 u32 skc_tw_rcv_nxt
; /* struct tcp_timewait_sock */
223 int skc_dontcopy_end
[0];
226 u32 skc_window_clamp
;
227 u32 skc_tw_snd_nxt
; /* struct tcp_timewait_sock */
233 * struct sock - network layer representation of sockets
234 * @__sk_common: shared layout with inet_timewait_sock
235 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
236 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
237 * @sk_lock: synchronizer
238 * @sk_rcvbuf: size of receive buffer in bytes
239 * @sk_wq: sock wait queue and async head
240 * @sk_rx_dst: receive input route used by early demux
241 * @sk_dst_cache: destination cache
242 * @sk_dst_pending_confirm: need to confirm neighbour
243 * @sk_policy: flow policy
244 * @sk_receive_queue: incoming packets
245 * @sk_wmem_alloc: transmit queue bytes committed
246 * @sk_write_queue: Packet sending queue
247 * @sk_omem_alloc: "o" is "option" or "other"
248 * @sk_wmem_queued: persistent queue size
249 * @sk_forward_alloc: space allocated forward
250 * @sk_napi_id: id of the last napi context to receive data for sk
251 * @sk_ll_usec: usecs to busypoll when there is no data
252 * @sk_allocation: allocation mode
253 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
254 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
255 * @sk_sndbuf: size of send buffer in bytes
256 * @sk_padding: unused element for alignment
257 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
258 * @sk_no_check_rx: allow zero checksum in RX packets
259 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
260 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
261 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
262 * @sk_gso_max_size: Maximum GSO segment size to build
263 * @sk_gso_max_segs: Maximum number of GSO segments
264 * @sk_lingertime: %SO_LINGER l_linger setting
265 * @sk_backlog: always used with the per-socket spinlock held
266 * @sk_callback_lock: used with the callbacks in the end of this struct
267 * @sk_error_queue: rarely used
268 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
269 * IPV6_ADDRFORM for instance)
270 * @sk_err: last error
271 * @sk_err_soft: errors that don't cause failure but are the cause of a
272 * persistent failure not just 'timed out'
273 * @sk_drops: raw/udp drops counter
274 * @sk_ack_backlog: current listen backlog
275 * @sk_max_ack_backlog: listen backlog set in listen()
276 * @sk_priority: %SO_PRIORITY setting
277 * @sk_type: socket type (%SOCK_STREAM, etc)
278 * @sk_protocol: which protocol this socket belongs in this network family
279 * @sk_peer_pid: &struct pid for this socket's peer
280 * @sk_peer_cred: %SO_PEERCRED setting
281 * @sk_rcvlowat: %SO_RCVLOWAT setting
282 * @sk_rcvtimeo: %SO_RCVTIMEO setting
283 * @sk_sndtimeo: %SO_SNDTIMEO setting
284 * @sk_txhash: computed flow hash for use on transmit
285 * @sk_filter: socket filtering instructions
286 * @sk_timer: sock cleanup timer
287 * @sk_stamp: time stamp of last packet received
288 * @sk_tsflags: SO_TIMESTAMPING socket options
289 * @sk_tskey: counter to disambiguate concurrent tstamp requests
290 * @sk_socket: Identd and reporting IO signals
291 * @sk_user_data: RPC layer private data
292 * @sk_frag: cached page frag
293 * @sk_peek_off: current peek_offset value
294 * @sk_send_head: front of stuff to transmit
295 * @sk_security: used by security modules
296 * @sk_mark: generic packet mark
297 * @sk_cgrp_data: cgroup data for this cgroup
298 * @sk_memcg: this socket's memory cgroup association
299 * @sk_write_pending: a write to stream socket waits to start
300 * @sk_state_change: callback to indicate change in the state of the sock
301 * @sk_data_ready: callback to indicate there is data to be processed
302 * @sk_write_space: callback to indicate there is bf sending space available
303 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
304 * @sk_backlog_rcv: callback to process the backlog
305 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
306 * @sk_reuseport_cb: reuseport group container
307 * @sk_rcu: used during RCU grace period
311 * Now struct inet_timewait_sock also uses sock_common, so please just
312 * don't add nothing before this first member (__sk_common) --acme
314 struct sock_common __sk_common
;
315 #define sk_node __sk_common.skc_node
316 #define sk_nulls_node __sk_common.skc_nulls_node
317 #define sk_refcnt __sk_common.skc_refcnt
318 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
320 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
321 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
322 #define sk_hash __sk_common.skc_hash
323 #define sk_portpair __sk_common.skc_portpair
324 #define sk_num __sk_common.skc_num
325 #define sk_dport __sk_common.skc_dport
326 #define sk_addrpair __sk_common.skc_addrpair
327 #define sk_daddr __sk_common.skc_daddr
328 #define sk_rcv_saddr __sk_common.skc_rcv_saddr
329 #define sk_family __sk_common.skc_family
330 #define sk_state __sk_common.skc_state
331 #define sk_reuse __sk_common.skc_reuse
332 #define sk_reuseport __sk_common.skc_reuseport
333 #define sk_ipv6only __sk_common.skc_ipv6only
334 #define sk_net_refcnt __sk_common.skc_net_refcnt
335 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
336 #define sk_bind_node __sk_common.skc_bind_node
337 #define sk_prot __sk_common.skc_prot
338 #define sk_net __sk_common.skc_net
339 #define sk_v6_daddr __sk_common.skc_v6_daddr
340 #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
341 #define sk_cookie __sk_common.skc_cookie
342 #define sk_incoming_cpu __sk_common.skc_incoming_cpu
343 #define sk_flags __sk_common.skc_flags
344 #define sk_rxhash __sk_common.skc_rxhash
346 socket_lock_t sk_lock
;
349 struct sk_buff_head sk_error_queue
;
350 struct sk_buff_head sk_receive_queue
;
352 * The backlog queue is special, it is always used with
353 * the per-socket spinlock held and requires low latency
354 * access. Therefore we special case it's implementation.
355 * Note : rmem_alloc is in this structure to fill a hole
356 * on 64bit arches, not because its logically part of
362 struct sk_buff
*head
;
363 struct sk_buff
*tail
;
365 #define sk_rmem_alloc sk_backlog.rmem_alloc
367 int sk_forward_alloc
;
368 #ifdef CONFIG_NET_RX_BUSY_POLL
369 unsigned int sk_ll_usec
;
370 /* ===== mostly read cache line ===== */
371 unsigned int sk_napi_id
;
375 struct sk_filter __rcu
*sk_filter
;
377 struct socket_wq __rcu
*sk_wq
;
378 struct socket_wq
*sk_wq_raw
;
381 struct xfrm_policy __rcu
*sk_policy
[2];
383 struct dst_entry
*sk_rx_dst
;
384 struct dst_entry __rcu
*sk_dst_cache
;
385 atomic_t sk_omem_alloc
;
388 /* ===== cache line for TX ===== */
390 atomic_t sk_wmem_alloc
;
391 unsigned long sk_tsq_flags
;
392 struct sk_buff
*sk_send_head
;
393 struct sk_buff_head sk_write_queue
;
395 int sk_write_pending
;
396 __u32 sk_dst_pending_confirm
;
397 /* Note: 32bit hole on 64bit arches */
399 struct timer_list sk_timer
;
402 u32 sk_pacing_rate
; /* bytes per second */
403 u32 sk_max_pacing_rate
;
404 struct page_frag sk_frag
;
405 netdev_features_t sk_route_caps
;
406 netdev_features_t sk_route_nocaps
;
408 unsigned int sk_gso_max_size
;
413 * Because of non atomicity rules, all
414 * changes are protected by socket lock.
416 unsigned int __sk_flags_offset
[0];
417 #ifdef __BIG_ENDIAN_BITFIELD
418 #define SK_FL_PROTO_SHIFT 16
419 #define SK_FL_PROTO_MASK 0x00ff0000
421 #define SK_FL_TYPE_SHIFT 0
422 #define SK_FL_TYPE_MASK 0x0000ffff
424 #define SK_FL_PROTO_SHIFT 8
425 #define SK_FL_PROTO_MASK 0x0000ff00
427 #define SK_FL_TYPE_SHIFT 16
428 #define SK_FL_TYPE_MASK 0xffff0000
431 kmemcheck_bitfield_begin(flags
);
432 unsigned int sk_padding
: 2,
438 #define SK_PROTOCOL_MAX U8_MAX
439 kmemcheck_bitfield_end(flags
);
442 unsigned long sk_lingertime
;
443 struct proto
*sk_prot_creator
;
444 rwlock_t sk_callback_lock
;
448 u32 sk_max_ack_backlog
;
450 struct pid
*sk_peer_pid
;
451 const struct cred
*sk_peer_cred
;
457 struct socket
*sk_socket
;
459 #ifdef CONFIG_SECURITY
462 struct sock_cgroup_data sk_cgrp_data
;
463 struct mem_cgroup
*sk_memcg
;
464 void (*sk_state_change
)(struct sock
*sk
);
465 void (*sk_data_ready
)(struct sock
*sk
);
466 void (*sk_write_space
)(struct sock
*sk
);
467 void (*sk_error_report
)(struct sock
*sk
);
468 int (*sk_backlog_rcv
)(struct sock
*sk
,
469 struct sk_buff
*skb
);
470 void (*sk_destruct
)(struct sock
*sk
);
471 struct sock_reuseport __rcu
*sk_reuseport_cb
;
472 struct rcu_head sk_rcu
;
475 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
477 #define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
478 #define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
481 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
482 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
483 * on a socket means that the socket will reuse everybody else's port
484 * without looking at the other's sk_reuse value.
487 #define SK_NO_REUSE 0
488 #define SK_CAN_REUSE 1
489 #define SK_FORCE_REUSE 2
491 int sk_set_peek_off(struct sock
*sk
, int val
);
493 static inline int sk_peek_offset(struct sock
*sk
, int flags
)
495 if (unlikely(flags
& MSG_PEEK
)) {
496 s32 off
= READ_ONCE(sk
->sk_peek_off
);
504 static inline void sk_peek_offset_bwd(struct sock
*sk
, int val
)
506 s32 off
= READ_ONCE(sk
->sk_peek_off
);
508 if (unlikely(off
>= 0)) {
509 off
= max_t(s32
, off
- val
, 0);
510 WRITE_ONCE(sk
->sk_peek_off
, off
);
514 static inline void sk_peek_offset_fwd(struct sock
*sk
, int val
)
516 sk_peek_offset_bwd(sk
, -val
);
520 * Hashed lists helper routines
522 static inline struct sock
*sk_entry(const struct hlist_node
*node
)
524 return hlist_entry(node
, struct sock
, sk_node
);
527 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
529 return hlist_entry(head
->first
, struct sock
, sk_node
);
532 static inline struct sock
*sk_head(const struct hlist_head
*head
)
534 return hlist_empty(head
) ? NULL
: __sk_head(head
);
537 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
539 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
542 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
544 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
547 static inline struct sock
*sk_next(const struct sock
*sk
)
549 return sk
->sk_node
.next
?
550 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
553 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
555 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
556 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
557 struct sock
, sk_nulls_node
) :
561 static inline bool sk_unhashed(const struct sock
*sk
)
563 return hlist_unhashed(&sk
->sk_node
);
566 static inline bool sk_hashed(const struct sock
*sk
)
568 return !sk_unhashed(sk
);
571 static inline void sk_node_init(struct hlist_node
*node
)
576 static inline void sk_nulls_node_init(struct hlist_nulls_node
*node
)
581 static inline void __sk_del_node(struct sock
*sk
)
583 __hlist_del(&sk
->sk_node
);
586 /* NB: equivalent to hlist_del_init_rcu */
587 static inline bool __sk_del_node_init(struct sock
*sk
)
591 sk_node_init(&sk
->sk_node
);
597 /* Grab socket reference count. This operation is valid only
598 when sk is ALREADY grabbed f.e. it is found in hash table
599 or a list and the lookup is made under lock preventing hash table
603 static __always_inline
void sock_hold(struct sock
*sk
)
605 atomic_inc(&sk
->sk_refcnt
);
608 /* Ungrab socket in the context, which assumes that socket refcnt
609 cannot hit zero, f.e. it is true in context of any socketcall.
611 static __always_inline
void __sock_put(struct sock
*sk
)
613 atomic_dec(&sk
->sk_refcnt
);
616 static inline bool sk_del_node_init(struct sock
*sk
)
618 bool rc
= __sk_del_node_init(sk
);
621 /* paranoid for a while -acme */
622 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
627 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
629 static inline bool __sk_nulls_del_node_init_rcu(struct sock
*sk
)
632 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
638 static inline bool sk_nulls_del_node_init_rcu(struct sock
*sk
)
640 bool rc
= __sk_nulls_del_node_init_rcu(sk
);
643 /* paranoid for a while -acme */
644 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
650 static inline void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
652 hlist_add_head(&sk
->sk_node
, list
);
655 static inline void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
658 __sk_add_node(sk
, list
);
661 static inline void sk_add_node_rcu(struct sock
*sk
, struct hlist_head
*list
)
664 if (IS_ENABLED(CONFIG_IPV6
) && sk
->sk_reuseport
&&
665 sk
->sk_family
== AF_INET6
)
666 hlist_add_tail_rcu(&sk
->sk_node
, list
);
668 hlist_add_head_rcu(&sk
->sk_node
, list
);
671 static inline void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
673 if (IS_ENABLED(CONFIG_IPV6
) && sk
->sk_reuseport
&&
674 sk
->sk_family
== AF_INET6
)
675 hlist_nulls_add_tail_rcu(&sk
->sk_nulls_node
, list
);
677 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
680 static inline void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
683 __sk_nulls_add_node_rcu(sk
, list
);
686 static inline void __sk_del_bind_node(struct sock
*sk
)
688 __hlist_del(&sk
->sk_bind_node
);
691 static inline void sk_add_bind_node(struct sock
*sk
,
692 struct hlist_head
*list
)
694 hlist_add_head(&sk
->sk_bind_node
, list
);
697 #define sk_for_each(__sk, list) \
698 hlist_for_each_entry(__sk, list, sk_node)
699 #define sk_for_each_rcu(__sk, list) \
700 hlist_for_each_entry_rcu(__sk, list, sk_node)
701 #define sk_nulls_for_each(__sk, node, list) \
702 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
703 #define sk_nulls_for_each_rcu(__sk, node, list) \
704 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
705 #define sk_for_each_from(__sk) \
706 hlist_for_each_entry_from(__sk, sk_node)
707 #define sk_nulls_for_each_from(__sk, node) \
708 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
709 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
710 #define sk_for_each_safe(__sk, tmp, list) \
711 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
712 #define sk_for_each_bound(__sk, list) \
713 hlist_for_each_entry(__sk, list, sk_bind_node)
716 * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset
717 * @tpos: the type * to use as a loop cursor.
718 * @pos: the &struct hlist_node to use as a loop cursor.
719 * @head: the head for your list.
720 * @offset: offset of hlist_node within the struct.
723 #define sk_for_each_entry_offset_rcu(tpos, pos, head, offset) \
724 for (pos = rcu_dereference((head)->first); \
726 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
727 pos = rcu_dereference(pos->next))
729 static inline struct user_namespace
*sk_user_ns(struct sock
*sk
)
731 /* Careful only use this in a context where these parameters
732 * can not change and must all be valid, such as recvmsg from
735 return sk
->sk_socket
->file
->f_cred
->user_ns
;
749 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
750 SOCK_DBG
, /* %SO_DEBUG setting */
751 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
752 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
753 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
754 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
755 SOCK_MEMALLOC
, /* VM depends on this socket for swapping */
756 SOCK_TIMESTAMPING_RX_SOFTWARE
, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
757 SOCK_FASYNC
, /* fasync() active */
759 SOCK_ZEROCOPY
, /* buffers from userspace */
760 SOCK_WIFI_STATUS
, /* push wifi status to userspace */
761 SOCK_NOFCS
, /* Tell NIC not to do the Ethernet FCS.
762 * Will use last 4 bytes of packet sent from
763 * user-space instead.
765 SOCK_FILTER_LOCKED
, /* Filter cannot be changed anymore */
766 SOCK_SELECT_ERR_QUEUE
, /* Wake select on error queue */
767 SOCK_RCU_FREE
, /* wait rcu grace period in sk_destruct() */
770 #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
772 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
774 nsk
->sk_flags
= osk
->sk_flags
;
777 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
779 __set_bit(flag
, &sk
->sk_flags
);
782 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
784 __clear_bit(flag
, &sk
->sk_flags
);
787 static inline bool sock_flag(const struct sock
*sk
, enum sock_flags flag
)
789 return test_bit(flag
, &sk
->sk_flags
);
793 extern struct static_key memalloc_socks
;
794 static inline int sk_memalloc_socks(void)
796 return static_key_false(&memalloc_socks
);
800 static inline int sk_memalloc_socks(void)
807 static inline gfp_t
sk_gfp_mask(const struct sock
*sk
, gfp_t gfp_mask
)
809 return gfp_mask
| (sk
->sk_allocation
& __GFP_MEMALLOC
);
812 static inline void sk_acceptq_removed(struct sock
*sk
)
814 sk
->sk_ack_backlog
--;
817 static inline void sk_acceptq_added(struct sock
*sk
)
819 sk
->sk_ack_backlog
++;
822 static inline bool sk_acceptq_is_full(const struct sock
*sk
)
824 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
828 * Compute minimal free write space needed to queue new packets.
830 static inline int sk_stream_min_wspace(const struct sock
*sk
)
832 return sk
->sk_wmem_queued
>> 1;
835 static inline int sk_stream_wspace(const struct sock
*sk
)
837 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
840 void sk_stream_write_space(struct sock
*sk
);
842 /* OOB backlog add */
843 static inline void __sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
845 /* dont let skb dst not refcounted, we are going to leave rcu lock */
846 skb_dst_force_safe(skb
);
848 if (!sk
->sk_backlog
.tail
)
849 sk
->sk_backlog
.head
= skb
;
851 sk
->sk_backlog
.tail
->next
= skb
;
853 sk
->sk_backlog
.tail
= skb
;
858 * Take into account size of receive queue and backlog queue
859 * Do not take into account this skb truesize,
860 * to allow even a single big packet to come.
862 static inline bool sk_rcvqueues_full(const struct sock
*sk
, unsigned int limit
)
864 unsigned int qsize
= sk
->sk_backlog
.len
+ atomic_read(&sk
->sk_rmem_alloc
);
866 return qsize
> limit
;
869 /* The per-socket spinlock must be held here. */
870 static inline __must_check
int sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
,
873 if (sk_rcvqueues_full(sk
, limit
))
877 * If the skb was allocated from pfmemalloc reserves, only
878 * allow SOCK_MEMALLOC sockets to use it as this socket is
879 * helping free memory
881 if (skb_pfmemalloc(skb
) && !sock_flag(sk
, SOCK_MEMALLOC
))
884 __sk_add_backlog(sk
, skb
);
885 sk
->sk_backlog
.len
+= skb
->truesize
;
889 int __sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
891 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
893 if (sk_memalloc_socks() && skb_pfmemalloc(skb
))
894 return __sk_backlog_rcv(sk
, skb
);
896 return sk
->sk_backlog_rcv(sk
, skb
);
899 static inline void sk_incoming_cpu_update(struct sock
*sk
)
901 sk
->sk_incoming_cpu
= raw_smp_processor_id();
904 static inline void sock_rps_record_flow_hash(__u32 hash
)
907 struct rps_sock_flow_table
*sock_flow_table
;
910 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
911 rps_record_sock_flow(sock_flow_table
, hash
);
916 static inline void sock_rps_record_flow(const struct sock
*sk
)
919 if (static_key_false(&rfs_needed
)) {
920 /* Reading sk->sk_rxhash might incur an expensive cache line
923 * TCP_ESTABLISHED does cover almost all states where RFS
924 * might be useful, and is cheaper [1] than testing :
925 * IPv4: inet_sk(sk)->inet_daddr
926 * IPv6: ipv6_addr_any(&sk->sk_v6_daddr)
927 * OR an additional socket flag
928 * [1] : sk_state and sk_prot are in the same cache line.
930 if (sk
->sk_state
== TCP_ESTABLISHED
)
931 sock_rps_record_flow_hash(sk
->sk_rxhash
);
936 static inline void sock_rps_save_rxhash(struct sock
*sk
,
937 const struct sk_buff
*skb
)
940 if (unlikely(sk
->sk_rxhash
!= skb
->hash
))
941 sk
->sk_rxhash
= skb
->hash
;
945 static inline void sock_rps_reset_rxhash(struct sock
*sk
)
952 #define sk_wait_event(__sk, __timeo, __condition, __wait) \
954 release_sock(__sk); \
955 __rc = __condition; \
957 *(__timeo) = wait_woken(__wait, \
958 TASK_INTERRUPTIBLE, \
961 sched_annotate_sleep(); \
963 __rc = __condition; \
967 int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
968 int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
969 void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
970 int sk_stream_error(struct sock
*sk
, int flags
, int err
);
971 void sk_stream_kill_queues(struct sock
*sk
);
972 void sk_set_memalloc(struct sock
*sk
);
973 void sk_clear_memalloc(struct sock
*sk
);
975 void __sk_flush_backlog(struct sock
*sk
);
977 static inline bool sk_flush_backlog(struct sock
*sk
)
979 if (unlikely(READ_ONCE(sk
->sk_backlog
.tail
))) {
980 __sk_flush_backlog(sk
);
986 int sk_wait_data(struct sock
*sk
, long *timeo
, const struct sk_buff
*skb
);
988 struct request_sock_ops
;
989 struct timewait_sock_ops
;
990 struct inet_hashinfo
;
995 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
996 * un-modified. Special care is taken when initializing object to zero.
998 static inline void sk_prot_clear_nulls(struct sock
*sk
, int size
)
1000 if (offsetof(struct sock
, sk_node
.next
) != 0)
1001 memset(sk
, 0, offsetof(struct sock
, sk_node
.next
));
1002 memset(&sk
->sk_node
.pprev
, 0,
1003 size
- offsetof(struct sock
, sk_node
.pprev
));
1006 /* Networking protocol blocks we attach to sockets.
1007 * socket layer -> transport layer interface
1010 void (*close
)(struct sock
*sk
,
1012 int (*connect
)(struct sock
*sk
,
1013 struct sockaddr
*uaddr
,
1015 int (*disconnect
)(struct sock
*sk
, int flags
);
1017 struct sock
* (*accept
)(struct sock
*sk
, int flags
, int *err
);
1019 int (*ioctl
)(struct sock
*sk
, int cmd
,
1021 int (*init
)(struct sock
*sk
);
1022 void (*destroy
)(struct sock
*sk
);
1023 void (*shutdown
)(struct sock
*sk
, int how
);
1024 int (*setsockopt
)(struct sock
*sk
, int level
,
1025 int optname
, char __user
*optval
,
1026 unsigned int optlen
);
1027 int (*getsockopt
)(struct sock
*sk
, int level
,
1028 int optname
, char __user
*optval
,
1029 int __user
*option
);
1030 #ifdef CONFIG_COMPAT
1031 int (*compat_setsockopt
)(struct sock
*sk
,
1033 int optname
, char __user
*optval
,
1034 unsigned int optlen
);
1035 int (*compat_getsockopt
)(struct sock
*sk
,
1037 int optname
, char __user
*optval
,
1038 int __user
*option
);
1039 int (*compat_ioctl
)(struct sock
*sk
,
1040 unsigned int cmd
, unsigned long arg
);
1042 int (*sendmsg
)(struct sock
*sk
, struct msghdr
*msg
,
1044 int (*recvmsg
)(struct sock
*sk
, struct msghdr
*msg
,
1045 size_t len
, int noblock
, int flags
,
1047 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
1048 int offset
, size_t size
, int flags
);
1049 int (*bind
)(struct sock
*sk
,
1050 struct sockaddr
*uaddr
, int addr_len
);
1052 int (*backlog_rcv
) (struct sock
*sk
,
1053 struct sk_buff
*skb
);
1055 void (*release_cb
)(struct sock
*sk
);
1057 /* Keeping track of sk's, looking them up, and port selection methods. */
1058 int (*hash
)(struct sock
*sk
);
1059 void (*unhash
)(struct sock
*sk
);
1060 void (*rehash
)(struct sock
*sk
);
1061 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
1063 /* Keeping track of sockets in use */
1064 #ifdef CONFIG_PROC_FS
1065 unsigned int inuse_idx
;
1068 bool (*stream_memory_free
)(const struct sock
*sk
);
1069 /* Memory pressure */
1070 void (*enter_memory_pressure
)(struct sock
*sk
);
1071 atomic_long_t
*memory_allocated
; /* Current allocated memory. */
1072 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
1074 * Pressure flag: try to collapse.
1075 * Technical note: it is used by multiple contexts non atomically.
1076 * All the __sk_mem_schedule() is of this nature: accounting
1077 * is strict, actions are advisory and have some latency.
1079 int *memory_pressure
;
1086 struct kmem_cache
*slab
;
1087 unsigned int obj_size
;
1090 struct percpu_counter
*orphan_count
;
1092 struct request_sock_ops
*rsk_prot
;
1093 struct timewait_sock_ops
*twsk_prot
;
1096 struct inet_hashinfo
*hashinfo
;
1097 struct udp_table
*udp_table
;
1098 struct raw_hashinfo
*raw_hash
;
1101 struct module
*owner
;
1105 struct list_head node
;
1106 #ifdef SOCK_REFCNT_DEBUG
1109 int (*diag_destroy
)(struct sock
*sk
, int err
);
1112 int proto_register(struct proto
*prot
, int alloc_slab
);
1113 void proto_unregister(struct proto
*prot
);
1115 #ifdef SOCK_REFCNT_DEBUG
1116 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
1118 atomic_inc(&sk
->sk_prot
->socks
);
1121 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
1123 atomic_dec(&sk
->sk_prot
->socks
);
1124 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
1125 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
1128 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
1130 if (atomic_read(&sk
->sk_refcnt
) != 1)
1131 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
1132 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
1134 #else /* SOCK_REFCNT_DEBUG */
1135 #define sk_refcnt_debug_inc(sk) do { } while (0)
1136 #define sk_refcnt_debug_dec(sk) do { } while (0)
1137 #define sk_refcnt_debug_release(sk) do { } while (0)
1138 #endif /* SOCK_REFCNT_DEBUG */
1140 static inline bool sk_stream_memory_free(const struct sock
*sk
)
1142 if (sk
->sk_wmem_queued
>= sk
->sk_sndbuf
)
1145 return sk
->sk_prot
->stream_memory_free
?
1146 sk
->sk_prot
->stream_memory_free(sk
) : true;
1149 static inline bool sk_stream_is_writeable(const struct sock
*sk
)
1151 return sk_stream_wspace(sk
) >= sk_stream_min_wspace(sk
) &&
1152 sk_stream_memory_free(sk
);
1155 static inline int sk_under_cgroup_hierarchy(struct sock
*sk
,
1156 struct cgroup
*ancestor
)
1158 #ifdef CONFIG_SOCK_CGROUP_DATA
1159 return cgroup_is_descendant(sock_cgroup_ptr(&sk
->sk_cgrp_data
),
1166 static inline bool sk_has_memory_pressure(const struct sock
*sk
)
1168 return sk
->sk_prot
->memory_pressure
!= NULL
;
1171 static inline bool sk_under_memory_pressure(const struct sock
*sk
)
1173 if (!sk
->sk_prot
->memory_pressure
)
1176 if (mem_cgroup_sockets_enabled
&& sk
->sk_memcg
&&
1177 mem_cgroup_under_socket_pressure(sk
->sk_memcg
))
1180 return !!*sk
->sk_prot
->memory_pressure
;
1183 static inline void sk_leave_memory_pressure(struct sock
*sk
)
1185 int *memory_pressure
= sk
->sk_prot
->memory_pressure
;
1187 if (!memory_pressure
)
1190 if (*memory_pressure
)
1191 *memory_pressure
= 0;
1194 static inline void sk_enter_memory_pressure(struct sock
*sk
)
1196 if (!sk
->sk_prot
->enter_memory_pressure
)
1199 sk
->sk_prot
->enter_memory_pressure(sk
);
1203 sk_memory_allocated(const struct sock
*sk
)
1205 return atomic_long_read(sk
->sk_prot
->memory_allocated
);
1209 sk_memory_allocated_add(struct sock
*sk
, int amt
)
1211 return atomic_long_add_return(amt
, sk
->sk_prot
->memory_allocated
);
1215 sk_memory_allocated_sub(struct sock
*sk
, int amt
)
1217 atomic_long_sub(amt
, sk
->sk_prot
->memory_allocated
);
1220 static inline void sk_sockets_allocated_dec(struct sock
*sk
)
1222 percpu_counter_dec(sk
->sk_prot
->sockets_allocated
);
1225 static inline void sk_sockets_allocated_inc(struct sock
*sk
)
1227 percpu_counter_inc(sk
->sk_prot
->sockets_allocated
);
1231 sk_sockets_allocated_read_positive(struct sock
*sk
)
1233 return percpu_counter_read_positive(sk
->sk_prot
->sockets_allocated
);
1237 proto_sockets_allocated_sum_positive(struct proto
*prot
)
1239 return percpu_counter_sum_positive(prot
->sockets_allocated
);
1243 proto_memory_allocated(struct proto
*prot
)
1245 return atomic_long_read(prot
->memory_allocated
);
1249 proto_memory_pressure(struct proto
*prot
)
1251 if (!prot
->memory_pressure
)
1253 return !!*prot
->memory_pressure
;
1257 #ifdef CONFIG_PROC_FS
1258 /* Called with local bh disabled */
1259 void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int inc
);
1260 int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
1262 static inline void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
1269 /* With per-bucket locks this operation is not-atomic, so that
1270 * this version is not worse.
1272 static inline int __sk_prot_rehash(struct sock
*sk
)
1274 sk
->sk_prot
->unhash(sk
);
1275 return sk
->sk_prot
->hash(sk
);
1278 /* About 10 seconds */
1279 #define SOCK_DESTROY_TIME (10*HZ)
1281 /* Sockets 0-1023 can't be bound to unless you are superuser */
1282 #define PROT_SOCK 1024
1284 #define SHUTDOWN_MASK 3
1285 #define RCV_SHUTDOWN 1
1286 #define SEND_SHUTDOWN 2
1288 #define SOCK_SNDBUF_LOCK 1
1289 #define SOCK_RCVBUF_LOCK 2
1290 #define SOCK_BINDADDR_LOCK 4
1291 #define SOCK_BINDPORT_LOCK 8
1293 struct socket_alloc
{
1294 struct socket socket
;
1295 struct inode vfs_inode
;
1298 static inline struct socket
*SOCKET_I(struct inode
*inode
)
1300 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
1303 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
1305 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
1309 * Functions for memory accounting
1311 int __sk_mem_raise_allocated(struct sock
*sk
, int size
, int amt
, int kind
);
1312 int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
1313 void __sk_mem_reduce_allocated(struct sock
*sk
, int amount
);
1314 void __sk_mem_reclaim(struct sock
*sk
, int amount
);
1316 /* We used to have PAGE_SIZE here, but systems with 64KB pages
1317 * do not necessarily have 16x time more memory than 4KB ones.
1319 #define SK_MEM_QUANTUM 4096
1320 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1321 #define SK_MEM_SEND 0
1322 #define SK_MEM_RECV 1
1324 /* sysctl_mem values are in pages, we convert them in SK_MEM_QUANTUM units */
1325 static inline long sk_prot_mem_limits(const struct sock
*sk
, int index
)
1327 long val
= sk
->sk_prot
->sysctl_mem
[index
];
1329 #if PAGE_SIZE > SK_MEM_QUANTUM
1330 val
<<= PAGE_SHIFT
- SK_MEM_QUANTUM_SHIFT
;
1331 #elif PAGE_SIZE < SK_MEM_QUANTUM
1332 val
>>= SK_MEM_QUANTUM_SHIFT
- PAGE_SHIFT
;
1337 static inline int sk_mem_pages(int amt
)
1339 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
1342 static inline bool sk_has_account(struct sock
*sk
)
1344 /* return true if protocol supports memory accounting */
1345 return !!sk
->sk_prot
->memory_allocated
;
1348 static inline bool sk_wmem_schedule(struct sock
*sk
, int size
)
1350 if (!sk_has_account(sk
))
1352 return size
<= sk
->sk_forward_alloc
||
1353 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
1357 sk_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
, int size
)
1359 if (!sk_has_account(sk
))
1361 return size
<= sk
->sk_forward_alloc
||
1362 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
) ||
1363 skb_pfmemalloc(skb
);
1366 static inline void sk_mem_reclaim(struct sock
*sk
)
1368 if (!sk_has_account(sk
))
1370 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
1371 __sk_mem_reclaim(sk
, sk
->sk_forward_alloc
);
1374 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
1376 if (!sk_has_account(sk
))
1378 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
1379 __sk_mem_reclaim(sk
, sk
->sk_forward_alloc
- 1);
1382 static inline void sk_mem_charge(struct sock
*sk
, int size
)
1384 if (!sk_has_account(sk
))
1386 sk
->sk_forward_alloc
-= size
;
1389 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
1391 if (!sk_has_account(sk
))
1393 sk
->sk_forward_alloc
+= size
;
1395 /* Avoid a possible overflow.
1396 * TCP send queues can make this happen, if sk_mem_reclaim()
1397 * is not called and more than 2 GBytes are released at once.
1399 * If we reach 2 MBytes, reclaim 1 MBytes right now, there is
1400 * no need to hold that much forward allocation anyway.
1402 if (unlikely(sk
->sk_forward_alloc
>= 1 << 21))
1403 __sk_mem_reclaim(sk
, 1 << 20);
1406 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
1408 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
1409 sk
->sk_wmem_queued
-= skb
->truesize
;
1410 sk_mem_uncharge(sk
, skb
->truesize
);
1414 static inline void sock_release_ownership(struct sock
*sk
)
1416 if (sk
->sk_lock
.owned
) {
1417 sk
->sk_lock
.owned
= 0;
1419 /* The sk_lock has mutex_unlock() semantics: */
1420 mutex_release(&sk
->sk_lock
.dep_map
, 1, _RET_IP_
);
1425 * Macro so as to not evaluate some arguments when
1426 * lockdep is not enabled.
1428 * Mark both the sk_lock and the sk_lock.slock as a
1429 * per-address-family lock class.
1431 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1433 sk->sk_lock.owned = 0; \
1434 init_waitqueue_head(&sk->sk_lock.wq); \
1435 spin_lock_init(&(sk)->sk_lock.slock); \
1436 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1437 sizeof((sk)->sk_lock)); \
1438 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1440 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1443 #ifdef CONFIG_LOCKDEP
1444 static inline bool lockdep_sock_is_held(const struct sock
*csk
)
1446 struct sock
*sk
= (struct sock
*)csk
;
1448 return lockdep_is_held(&sk
->sk_lock
) ||
1449 lockdep_is_held(&sk
->sk_lock
.slock
);
1453 void lock_sock_nested(struct sock
*sk
, int subclass
);
1455 static inline void lock_sock(struct sock
*sk
)
1457 lock_sock_nested(sk
, 0);
1460 void release_sock(struct sock
*sk
);
1462 /* BH context may only use the following locking interface. */
1463 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1464 #define bh_lock_sock_nested(__sk) \
1465 spin_lock_nested(&((__sk)->sk_lock.slock), \
1466 SINGLE_DEPTH_NESTING)
1467 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1469 bool lock_sock_fast(struct sock
*sk
);
1471 * unlock_sock_fast - complement of lock_sock_fast
1475 * fast unlock socket for user context.
1476 * If slow mode is on, we call regular release_sock()
1478 static inline void unlock_sock_fast(struct sock
*sk
, bool slow
)
1483 spin_unlock_bh(&sk
->sk_lock
.slock
);
1486 /* Used by processes to "lock" a socket state, so that
1487 * interrupts and bottom half handlers won't change it
1488 * from under us. It essentially blocks any incoming
1489 * packets, so that we won't get any new data or any
1490 * packets that change the state of the socket.
1492 * While locked, BH processing will add new packets to
1493 * the backlog queue. This queue is processed by the
1494 * owner of the socket lock right before it is released.
1496 * Since ~2.3.5 it is also exclusive sleep lock serializing
1497 * accesses from user process context.
1500 static inline void sock_owned_by_me(const struct sock
*sk
)
1502 #ifdef CONFIG_LOCKDEP
1503 WARN_ON_ONCE(!lockdep_sock_is_held(sk
) && debug_locks
);
1507 static inline bool sock_owned_by_user(const struct sock
*sk
)
1509 sock_owned_by_me(sk
);
1510 return sk
->sk_lock
.owned
;
1513 /* no reclassification while locks are held */
1514 static inline bool sock_allow_reclassification(const struct sock
*csk
)
1516 struct sock
*sk
= (struct sock
*)csk
;
1518 return !sk
->sk_lock
.owned
&& !spin_is_locked(&sk
->sk_lock
.slock
);
1521 struct sock
*sk_alloc(struct net
*net
, int family
, gfp_t priority
,
1522 struct proto
*prot
, int kern
);
1523 void sk_free(struct sock
*sk
);
1524 void sk_destruct(struct sock
*sk
);
1525 struct sock
*sk_clone_lock(const struct sock
*sk
, const gfp_t priority
);
1527 struct sk_buff
*sock_wmalloc(struct sock
*sk
, unsigned long size
, int force
,
1529 void __sock_wfree(struct sk_buff
*skb
);
1530 void sock_wfree(struct sk_buff
*skb
);
1531 void skb_orphan_partial(struct sk_buff
*skb
);
1532 void sock_rfree(struct sk_buff
*skb
);
1533 void sock_efree(struct sk_buff
*skb
);
1535 void sock_edemux(struct sk_buff
*skb
);
1537 #define sock_edemux(skb) sock_efree(skb)
1540 int sock_setsockopt(struct socket
*sock
, int level
, int op
,
1541 char __user
*optval
, unsigned int optlen
);
1543 int sock_getsockopt(struct socket
*sock
, int level
, int op
,
1544 char __user
*optval
, int __user
*optlen
);
1545 struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
, unsigned long size
,
1546 int noblock
, int *errcode
);
1547 struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
, unsigned long header_len
,
1548 unsigned long data_len
, int noblock
,
1549 int *errcode
, int max_page_order
);
1550 void *sock_kmalloc(struct sock
*sk
, int size
, gfp_t priority
);
1551 void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
1552 void sock_kzfree_s(struct sock
*sk
, void *mem
, int size
);
1553 void sk_send_sigurg(struct sock
*sk
);
1555 struct sockcm_cookie
{
1560 int __sock_cmsg_send(struct sock
*sk
, struct msghdr
*msg
, struct cmsghdr
*cmsg
,
1561 struct sockcm_cookie
*sockc
);
1562 int sock_cmsg_send(struct sock
*sk
, struct msghdr
*msg
,
1563 struct sockcm_cookie
*sockc
);
1566 * Functions to fill in entries in struct proto_ops when a protocol
1567 * does not implement a particular function.
1569 int sock_no_bind(struct socket
*, struct sockaddr
*, int);
1570 int sock_no_connect(struct socket
*, struct sockaddr
*, int, int);
1571 int sock_no_socketpair(struct socket
*, struct socket
*);
1572 int sock_no_accept(struct socket
*, struct socket
*, int);
1573 int sock_no_getname(struct socket
*, struct sockaddr
*, int *, int);
1574 unsigned int sock_no_poll(struct file
*, struct socket
*,
1575 struct poll_table_struct
*);
1576 int sock_no_ioctl(struct socket
*, unsigned int, unsigned long);
1577 int sock_no_listen(struct socket
*, int);
1578 int sock_no_shutdown(struct socket
*, int);
1579 int sock_no_getsockopt(struct socket
*, int , int, char __user
*, int __user
*);
1580 int sock_no_setsockopt(struct socket
*, int, int, char __user
*, unsigned int);
1581 int sock_no_sendmsg(struct socket
*, struct msghdr
*, size_t);
1582 int sock_no_recvmsg(struct socket
*, struct msghdr
*, size_t, int);
1583 int sock_no_mmap(struct file
*file
, struct socket
*sock
,
1584 struct vm_area_struct
*vma
);
1585 ssize_t
sock_no_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
1586 size_t size
, int flags
);
1589 * Functions to fill in entries in struct proto_ops when a protocol
1590 * uses the inet style.
1592 int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1593 char __user
*optval
, int __user
*optlen
);
1594 int sock_common_recvmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
,
1596 int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1597 char __user
*optval
, unsigned int optlen
);
1598 int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
1599 int optname
, char __user
*optval
, int __user
*optlen
);
1600 int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
1601 int optname
, char __user
*optval
, unsigned int optlen
);
1603 void sk_common_release(struct sock
*sk
);
1606 * Default socket callbacks and setup code
1609 /* Initialise core socket variables */
1610 void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1613 * Socket reference counting postulates.
1615 * * Each user of socket SHOULD hold a reference count.
1616 * * Each access point to socket (an hash table bucket, reference from a list,
1617 * running timer, skb in flight MUST hold a reference count.
1618 * * When reference count hits 0, it means it will never increase back.
1619 * * When reference count hits 0, it means that no references from
1620 * outside exist to this socket and current process on current CPU
1621 * is last user and may/should destroy this socket.
1622 * * sk_free is called from any context: process, BH, IRQ. When
1623 * it is called, socket has no references from outside -> sk_free
1624 * may release descendant resources allocated by the socket, but
1625 * to the time when it is called, socket is NOT referenced by any
1626 * hash tables, lists etc.
1627 * * Packets, delivered from outside (from network or from another process)
1628 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1629 * when they sit in queue. Otherwise, packets will leak to hole, when
1630 * socket is looked up by one cpu and unhasing is made by another CPU.
1631 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1632 * (leak to backlog). Packet socket does all the processing inside
1633 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1634 * use separate SMP lock, so that they are prone too.
1637 /* Ungrab socket and destroy it, if it was the last reference. */
1638 static inline void sock_put(struct sock
*sk
)
1640 if (atomic_dec_and_test(&sk
->sk_refcnt
))
1643 /* Generic version of sock_put(), dealing with all sockets
1644 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
1646 void sock_gen_put(struct sock
*sk
);
1648 int __sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
, const int nested
,
1649 unsigned int trim_cap
, bool refcounted
);
1650 static inline int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
1653 return __sk_receive_skb(sk
, skb
, nested
, 1, true);
1656 static inline void sk_tx_queue_set(struct sock
*sk
, int tx_queue
)
1658 sk
->sk_tx_queue_mapping
= tx_queue
;
1661 static inline void sk_tx_queue_clear(struct sock
*sk
)
1663 sk
->sk_tx_queue_mapping
= -1;
1666 static inline int sk_tx_queue_get(const struct sock
*sk
)
1668 return sk
? sk
->sk_tx_queue_mapping
: -1;
1671 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1673 sk_tx_queue_clear(sk
);
1674 sk
->sk_socket
= sock
;
1677 static inline wait_queue_head_t
*sk_sleep(struct sock
*sk
)
1679 BUILD_BUG_ON(offsetof(struct socket_wq
, wait
) != 0);
1680 return &rcu_dereference_raw(sk
->sk_wq
)->wait
;
1682 /* Detach socket from process context.
1683 * Announce socket dead, detach it from wait queue and inode.
1684 * Note that parent inode held reference count on this struct sock,
1685 * we do not release it in this function, because protocol
1686 * probably wants some additional cleanups or even continuing
1687 * to work with this socket (TCP).
1689 static inline void sock_orphan(struct sock
*sk
)
1691 write_lock_bh(&sk
->sk_callback_lock
);
1692 sock_set_flag(sk
, SOCK_DEAD
);
1693 sk_set_socket(sk
, NULL
);
1695 write_unlock_bh(&sk
->sk_callback_lock
);
1698 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1700 write_lock_bh(&sk
->sk_callback_lock
);
1701 sk
->sk_wq
= parent
->wq
;
1703 sk_set_socket(sk
, parent
);
1704 sk
->sk_uid
= SOCK_INODE(parent
)->i_uid
;
1705 security_sock_graft(sk
, parent
);
1706 write_unlock_bh(&sk
->sk_callback_lock
);
1709 kuid_t
sock_i_uid(struct sock
*sk
);
1710 unsigned long sock_i_ino(struct sock
*sk
);
1712 static inline kuid_t
sock_net_uid(const struct net
*net
, const struct sock
*sk
)
1714 return sk
? sk
->sk_uid
: make_kuid(net
->user_ns
, 0);
1717 static inline u32
net_tx_rndhash(void)
1719 u32 v
= prandom_u32();
1724 static inline void sk_set_txhash(struct sock
*sk
)
1726 sk
->sk_txhash
= net_tx_rndhash();
1729 static inline void sk_rethink_txhash(struct sock
*sk
)
1735 static inline struct dst_entry
*
1736 __sk_dst_get(struct sock
*sk
)
1738 return rcu_dereference_check(sk
->sk_dst_cache
,
1739 lockdep_sock_is_held(sk
));
1742 static inline struct dst_entry
*
1743 sk_dst_get(struct sock
*sk
)
1745 struct dst_entry
*dst
;
1748 dst
= rcu_dereference(sk
->sk_dst_cache
);
1749 if (dst
&& !atomic_inc_not_zero(&dst
->__refcnt
))
1755 static inline void dst_negative_advice(struct sock
*sk
)
1757 struct dst_entry
*ndst
, *dst
= __sk_dst_get(sk
);
1759 sk_rethink_txhash(sk
);
1761 if (dst
&& dst
->ops
->negative_advice
) {
1762 ndst
= dst
->ops
->negative_advice(dst
);
1765 rcu_assign_pointer(sk
->sk_dst_cache
, ndst
);
1766 sk_tx_queue_clear(sk
);
1767 sk
->sk_dst_pending_confirm
= 0;
1773 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1775 struct dst_entry
*old_dst
;
1777 sk_tx_queue_clear(sk
);
1778 sk
->sk_dst_pending_confirm
= 0;
1780 * This can be called while sk is owned by the caller only,
1781 * with no state that can be checked in a rcu_dereference_check() cond
1783 old_dst
= rcu_dereference_raw(sk
->sk_dst_cache
);
1784 rcu_assign_pointer(sk
->sk_dst_cache
, dst
);
1785 dst_release(old_dst
);
1789 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1791 struct dst_entry
*old_dst
;
1793 sk_tx_queue_clear(sk
);
1794 sk
->sk_dst_pending_confirm
= 0;
1795 old_dst
= xchg((__force
struct dst_entry
**)&sk
->sk_dst_cache
, dst
);
1796 dst_release(old_dst
);
1800 __sk_dst_reset(struct sock
*sk
)
1802 __sk_dst_set(sk
, NULL
);
1806 sk_dst_reset(struct sock
*sk
)
1808 sk_dst_set(sk
, NULL
);
1811 struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1813 struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1815 static inline void sk_dst_confirm(struct sock
*sk
)
1817 if (!sk
->sk_dst_pending_confirm
)
1818 sk
->sk_dst_pending_confirm
= 1;
1821 bool sk_mc_loop(struct sock
*sk
);
1823 static inline bool sk_can_gso(const struct sock
*sk
)
1825 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1828 void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1830 static inline void sk_nocaps_add(struct sock
*sk
, netdev_features_t flags
)
1832 sk
->sk_route_nocaps
|= flags
;
1833 sk
->sk_route_caps
&= ~flags
;
1836 static inline bool sk_check_csum_caps(struct sock
*sk
)
1838 return (sk
->sk_route_caps
& NETIF_F_HW_CSUM
) ||
1839 (sk
->sk_family
== PF_INET
&&
1840 (sk
->sk_route_caps
& NETIF_F_IP_CSUM
)) ||
1841 (sk
->sk_family
== PF_INET6
&&
1842 (sk
->sk_route_caps
& NETIF_F_IPV6_CSUM
));
1845 static inline int skb_do_copy_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1846 struct iov_iter
*from
, char *to
,
1847 int copy
, int offset
)
1849 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1851 if (!csum_and_copy_from_iter_full(to
, copy
, &csum
, from
))
1853 skb
->csum
= csum_block_add(skb
->csum
, csum
, offset
);
1854 } else if (sk
->sk_route_caps
& NETIF_F_NOCACHE_COPY
) {
1855 if (!copy_from_iter_full_nocache(to
, copy
, from
))
1857 } else if (!copy_from_iter_full(to
, copy
, from
))
1863 static inline int skb_add_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1864 struct iov_iter
*from
, int copy
)
1866 int err
, offset
= skb
->len
;
1868 err
= skb_do_copy_data_nocache(sk
, skb
, from
, skb_put(skb
, copy
),
1871 __skb_trim(skb
, offset
);
1876 static inline int skb_copy_to_page_nocache(struct sock
*sk
, struct iov_iter
*from
,
1877 struct sk_buff
*skb
,
1883 err
= skb_do_copy_data_nocache(sk
, skb
, from
, page_address(page
) + off
,
1889 skb
->data_len
+= copy
;
1890 skb
->truesize
+= copy
;
1891 sk
->sk_wmem_queued
+= copy
;
1892 sk_mem_charge(sk
, copy
);
1897 * sk_wmem_alloc_get - returns write allocations
1900 * Returns sk_wmem_alloc minus initial offset of one
1902 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
1904 return atomic_read(&sk
->sk_wmem_alloc
) - 1;
1908 * sk_rmem_alloc_get - returns read allocations
1911 * Returns sk_rmem_alloc
1913 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
1915 return atomic_read(&sk
->sk_rmem_alloc
);
1919 * sk_has_allocations - check if allocations are outstanding
1922 * Returns true if socket has write or read allocations
1924 static inline bool sk_has_allocations(const struct sock
*sk
)
1926 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
1930 * skwq_has_sleeper - check if there are any waiting processes
1931 * @wq: struct socket_wq
1933 * Returns true if socket_wq has waiting processes
1935 * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
1936 * barrier call. They were added due to the race found within the tcp code.
1938 * Consider following tcp code paths:
1942 * sys_select receive packet
1944 * __add_wait_queue update tp->rcv_nxt
1946 * tp->rcv_nxt check sock_def_readable
1948 * schedule rcu_read_lock();
1949 * wq = rcu_dereference(sk->sk_wq);
1950 * if (wq && waitqueue_active(&wq->wait))
1951 * wake_up_interruptible(&wq->wait)
1955 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1956 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1957 * could then endup calling schedule and sleep forever if there are no more
1958 * data on the socket.
1961 static inline bool skwq_has_sleeper(struct socket_wq
*wq
)
1963 return wq
&& wq_has_sleeper(&wq
->wait
);
1967 * sock_poll_wait - place memory barrier behind the poll_wait call.
1969 * @wait_address: socket wait queue
1972 * See the comments in the wq_has_sleeper function.
1974 static inline void sock_poll_wait(struct file
*filp
,
1975 wait_queue_head_t
*wait_address
, poll_table
*p
)
1977 if (!poll_does_not_wait(p
) && wait_address
) {
1978 poll_wait(filp
, wait_address
, p
);
1979 /* We need to be sure we are in sync with the
1980 * socket flags modification.
1982 * This memory barrier is paired in the wq_has_sleeper.
1988 static inline void skb_set_hash_from_sk(struct sk_buff
*skb
, struct sock
*sk
)
1990 if (sk
->sk_txhash
) {
1992 skb
->hash
= sk
->sk_txhash
;
1996 void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
);
1999 * Queue a received datagram if it will fit. Stream and sequenced
2000 * protocols can't normally use this as they need to fit buffers in
2001 * and play with them.
2003 * Inlined as it's very short and called for pretty much every
2004 * packet ever received.
2006 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
2010 skb
->destructor
= sock_rfree
;
2011 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
2012 sk_mem_charge(sk
, skb
->truesize
);
2015 void sk_reset_timer(struct sock
*sk
, struct timer_list
*timer
,
2016 unsigned long expires
);
2018 void sk_stop_timer(struct sock
*sk
, struct timer_list
*timer
);
2020 int __sk_queue_drop_skb(struct sock
*sk
, struct sk_buff
*skb
,
2022 void (*destructor
)(struct sock
*sk
,
2023 struct sk_buff
*skb
));
2024 int __sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
2025 int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
2027 int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
);
2028 struct sk_buff
*sock_dequeue_err_skb(struct sock
*sk
);
2031 * Recover an error report and clear atomically
2034 static inline int sock_error(struct sock
*sk
)
2037 if (likely(!sk
->sk_err
))
2039 err
= xchg(&sk
->sk_err
, 0);
2043 static inline unsigned long sock_wspace(struct sock
*sk
)
2047 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
2048 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
2056 * We use sk->sk_wq_raw, from contexts knowing this
2057 * pointer is not NULL and cannot disappear/change.
2059 static inline void sk_set_bit(int nr
, struct sock
*sk
)
2061 if ((nr
== SOCKWQ_ASYNC_NOSPACE
|| nr
== SOCKWQ_ASYNC_WAITDATA
) &&
2062 !sock_flag(sk
, SOCK_FASYNC
))
2065 set_bit(nr
, &sk
->sk_wq_raw
->flags
);
2068 static inline void sk_clear_bit(int nr
, struct sock
*sk
)
2070 if ((nr
== SOCKWQ_ASYNC_NOSPACE
|| nr
== SOCKWQ_ASYNC_WAITDATA
) &&
2071 !sock_flag(sk
, SOCK_FASYNC
))
2074 clear_bit(nr
, &sk
->sk_wq_raw
->flags
);
2077 static inline void sk_wake_async(const struct sock
*sk
, int how
, int band
)
2079 if (sock_flag(sk
, SOCK_FASYNC
)) {
2081 sock_wake_async(rcu_dereference(sk
->sk_wq
), how
, band
);
2086 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2087 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2088 * Note: for send buffers, TCP works better if we can build two skbs at
2091 #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2093 #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2094 #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
2096 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
2098 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
2099 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
2100 sk
->sk_sndbuf
= max_t(u32
, sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
2104 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
,
2105 bool force_schedule
);
2108 * sk_page_frag - return an appropriate page_frag
2111 * If socket allocation mode allows current thread to sleep, it means its
2112 * safe to use the per task page_frag instead of the per socket one.
2114 static inline struct page_frag
*sk_page_frag(struct sock
*sk
)
2116 if (gfpflags_allow_blocking(sk
->sk_allocation
))
2117 return ¤t
->task_frag
;
2119 return &sk
->sk_frag
;
2122 bool sk_page_frag_refill(struct sock
*sk
, struct page_frag
*pfrag
);
2125 * Default write policy as shown to user space via poll/select/SIGIO
2127 static inline bool sock_writeable(const struct sock
*sk
)
2129 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
2132 static inline gfp_t
gfp_any(void)
2134 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
2137 static inline long sock_rcvtimeo(const struct sock
*sk
, bool noblock
)
2139 return noblock
? 0 : sk
->sk_rcvtimeo
;
2142 static inline long sock_sndtimeo(const struct sock
*sk
, bool noblock
)
2144 return noblock
? 0 : sk
->sk_sndtimeo
;
2147 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
2149 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
2152 /* Alas, with timeout socket operations are not restartable.
2153 * Compare this to poll().
2155 static inline int sock_intr_errno(long timeo
)
2157 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
2160 struct sock_skb_cb
{
2164 /* Store sock_skb_cb at the end of skb->cb[] so protocol families
2165 * using skb->cb[] would keep using it directly and utilize its
2166 * alignement guarantee.
2168 #define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \
2169 sizeof(struct sock_skb_cb)))
2171 #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2172 SOCK_SKB_CB_OFFSET))
2174 #define sock_skb_cb_check_size(size) \
2175 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
2178 sock_skb_set_dropcount(const struct sock
*sk
, struct sk_buff
*skb
)
2180 SOCK_SKB_CB(skb
)->dropcount
= sock_flag(sk
, SOCK_RXQ_OVFL
) ?
2181 atomic_read(&sk
->sk_drops
) : 0;
2184 static inline void sk_drops_add(struct sock
*sk
, const struct sk_buff
*skb
)
2186 int segs
= max_t(u16
, 1, skb_shinfo(skb
)->gso_segs
);
2188 atomic_add(segs
, &sk
->sk_drops
);
2191 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
2192 struct sk_buff
*skb
);
2193 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
2194 struct sk_buff
*skb
);
2197 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
2199 ktime_t kt
= skb
->tstamp
;
2200 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
2203 * generate control messages if
2204 * - receive time stamping in software requested
2205 * - software time stamp available and wanted
2206 * - hardware time stamps available and wanted
2208 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
2209 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RX_SOFTWARE
) ||
2210 (kt
&& sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
) ||
2211 (hwtstamps
->hwtstamp
&&
2212 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
)))
2213 __sock_recv_timestamp(msg
, sk
, skb
);
2217 if (sock_flag(sk
, SOCK_WIFI_STATUS
) && skb
->wifi_acked_valid
)
2218 __sock_recv_wifi_status(msg
, sk
, skb
);
2221 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2222 struct sk_buff
*skb
);
2224 static inline void sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2225 struct sk_buff
*skb
)
2227 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2228 (1UL << SOCK_RCVTSTAMP))
2229 #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2230 SOF_TIMESTAMPING_RAW_HARDWARE)
2232 if (sk
->sk_flags
& FLAGS_TS_OR_DROPS
|| sk
->sk_tsflags
& TSFLAGS_ANY
)
2233 __sock_recv_ts_and_drops(msg
, sk
, skb
);
2235 sk
->sk_stamp
= skb
->tstamp
;
2238 void __sock_tx_timestamp(__u16 tsflags
, __u8
*tx_flags
);
2241 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2242 * @sk: socket sending this packet
2243 * @tsflags: timestamping flags to use
2244 * @tx_flags: completed with instructions for time stamping
2246 * Note : callers should take care of initial *tx_flags value (usually 0)
2248 static inline void sock_tx_timestamp(const struct sock
*sk
, __u16 tsflags
,
2251 if (unlikely(tsflags
))
2252 __sock_tx_timestamp(tsflags
, tx_flags
);
2253 if (unlikely(sock_flag(sk
, SOCK_WIFI_STATUS
)))
2254 *tx_flags
|= SKBTX_WIFI_STATUS
;
2258 * sk_eat_skb - Release a skb if it is no longer needed
2259 * @sk: socket to eat this skb from
2260 * @skb: socket buffer to eat
2262 * This routine must be called with interrupts disabled or with the socket
2263 * locked so that the sk_buff queue operation is ok.
2265 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
)
2267 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2272 struct net
*sock_net(const struct sock
*sk
)
2274 return read_pnet(&sk
->sk_net
);
2278 void sock_net_set(struct sock
*sk
, struct net
*net
)
2280 write_pnet(&sk
->sk_net
, net
);
2283 static inline struct sock
*skb_steal_sock(struct sk_buff
*skb
)
2286 struct sock
*sk
= skb
->sk
;
2288 skb
->destructor
= NULL
;
2295 /* This helper checks if a socket is a full socket,
2296 * ie _not_ a timewait or request socket.
2298 static inline bool sk_fullsock(const struct sock
*sk
)
2300 return (1 << sk
->sk_state
) & ~(TCPF_TIME_WAIT
| TCPF_NEW_SYN_RECV
);
2303 /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2304 * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2306 static inline bool sk_listener(const struct sock
*sk
)
2308 return (1 << sk
->sk_state
) & (TCPF_LISTEN
| TCPF_NEW_SYN_RECV
);
2312 * sk_state_load - read sk->sk_state for lockless contexts
2313 * @sk: socket pointer
2315 * Paired with sk_state_store(). Used in places we do not hold socket lock :
2316 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
2318 static inline int sk_state_load(const struct sock
*sk
)
2320 return smp_load_acquire(&sk
->sk_state
);
2324 * sk_state_store - update sk->sk_state
2325 * @sk: socket pointer
2326 * @newstate: new state
2328 * Paired with sk_state_load(). Should be used in contexts where
2329 * state change might impact lockless readers.
2331 static inline void sk_state_store(struct sock
*sk
, int newstate
)
2333 smp_store_release(&sk
->sk_state
, newstate
);
2336 void sock_enable_timestamp(struct sock
*sk
, int flag
);
2337 int sock_get_timestamp(struct sock
*, struct timeval __user
*);
2338 int sock_get_timestampns(struct sock
*, struct timespec __user
*);
2339 int sock_recv_errqueue(struct sock
*sk
, struct msghdr
*msg
, int len
, int level
,
2342 bool sk_ns_capable(const struct sock
*sk
,
2343 struct user_namespace
*user_ns
, int cap
);
2344 bool sk_capable(const struct sock
*sk
, int cap
);
2345 bool sk_net_capable(const struct sock
*sk
, int cap
);
2347 extern __u32 sysctl_wmem_max
;
2348 extern __u32 sysctl_rmem_max
;
2350 extern int sysctl_tstamp_allow_data
;
2351 extern int sysctl_optmem_max
;
2353 extern __u32 sysctl_wmem_default
;
2354 extern __u32 sysctl_rmem_default
;
2356 #endif /* _SOCK_H */