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>
62 #include <linux/filter.h>
63 #include <linux/rculist_nulls.h>
64 #include <linux/poll.h>
66 #include <linux/atomic.h>
68 #include <net/checksum.h>
69 #include <net/tcp_states.h>
70 #include <linux/net_tstamp.h>
75 int mem_cgroup_sockets_init(struct mem_cgroup
*memcg
, struct cgroup_subsys
*ss
);
76 void mem_cgroup_sockets_destroy(struct mem_cgroup
*memcg
);
79 int mem_cgroup_sockets_init(struct mem_cgroup
*memcg
, struct cgroup_subsys
*ss
)
84 void mem_cgroup_sockets_destroy(struct mem_cgroup
*memcg
)
89 * This structure really needs to be cleaned up.
90 * Most of it is for TCP, and not used by any of
91 * the other protocols.
94 /* Define this to get the SOCK_DBG debugging facility. */
95 #define SOCK_DEBUGGING
97 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
98 printk(KERN_DEBUG msg); } while (0)
100 /* Validate arguments and do nothing */
101 static inline __printf(2, 3)
102 void SOCK_DEBUG(const struct sock
*sk
, const char *msg
, ...)
107 /* This is the per-socket lock. The spinlock provides a synchronization
108 * between user contexts and software interrupt processing, whereas the
109 * mini-semaphore synchronizes multiple users amongst themselves.
114 wait_queue_head_t wq
;
116 * We express the mutex-alike socket_lock semantics
117 * to the lock validator by explicitly managing
118 * the slock as a lock variant (in addition to
121 #ifdef CONFIG_DEBUG_LOCK_ALLOC
122 struct lockdep_map dep_map
;
130 typedef __u32 __bitwise __portpair
;
131 typedef __u64 __bitwise __addrpair
;
134 * struct sock_common - minimal network layer representation of sockets
135 * @skc_daddr: Foreign IPv4 addr
136 * @skc_rcv_saddr: Bound local IPv4 addr
137 * @skc_hash: hash value used with various protocol lookup tables
138 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
139 * @skc_dport: placeholder for inet_dport/tw_dport
140 * @skc_num: placeholder for inet_num/tw_num
141 * @skc_family: network address family
142 * @skc_state: Connection state
143 * @skc_reuse: %SO_REUSEADDR setting
144 * @skc_reuseport: %SO_REUSEPORT setting
145 * @skc_bound_dev_if: bound device index if != 0
146 * @skc_bind_node: bind hash linkage for various protocol lookup tables
147 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
148 * @skc_prot: protocol handlers inside a network family
149 * @skc_net: reference to the network namespace of this socket
150 * @skc_node: main hash linkage for various protocol lookup tables
151 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
152 * @skc_tx_queue_mapping: tx queue number for this connection
153 * @skc_refcnt: reference count
155 * This is the minimal network layer representation of sockets, the header
156 * for struct sock and struct inet_timewait_sock.
159 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
160 * address on 64bit arches : cf INET_MATCH()
163 __addrpair skc_addrpair
;
166 __be32 skc_rcv_saddr
;
170 unsigned int skc_hash
;
171 __u16 skc_u16hashes
[2];
173 /* skc_dport && skc_num must be grouped as well */
175 __portpair skc_portpair
;
182 unsigned short skc_family
;
183 volatile unsigned char skc_state
;
184 unsigned char skc_reuse
:4;
185 unsigned char skc_reuseport
:1;
186 unsigned char skc_ipv6only
:1;
187 unsigned char skc_net_refcnt
:1;
188 int skc_bound_dev_if
;
190 struct hlist_node skc_bind_node
;
191 struct hlist_nulls_node skc_portaddr_node
;
193 struct proto
*skc_prot
;
194 possible_net_t skc_net
;
196 #if IS_ENABLED(CONFIG_IPV6)
197 struct in6_addr skc_v6_daddr
;
198 struct in6_addr skc_v6_rcv_saddr
;
201 atomic64_t skc_cookie
;
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
;
217 int skc_dontcopy_end
[0];
223 * struct sock - network layer representation of sockets
224 * @__sk_common: shared layout with inet_timewait_sock
225 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
226 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
227 * @sk_lock: synchronizer
228 * @sk_rcvbuf: size of receive buffer in bytes
229 * @sk_wq: sock wait queue and async head
230 * @sk_rx_dst: receive input route used by early demux
231 * @sk_dst_cache: destination cache
232 * @sk_dst_lock: destination cache lock
233 * @sk_policy: flow policy
234 * @sk_receive_queue: incoming packets
235 * @sk_wmem_alloc: transmit queue bytes committed
236 * @sk_write_queue: Packet sending queue
237 * @sk_omem_alloc: "o" is "option" or "other"
238 * @sk_wmem_queued: persistent queue size
239 * @sk_forward_alloc: space allocated forward
240 * @sk_napi_id: id of the last napi context to receive data for sk
241 * @sk_ll_usec: usecs to busypoll when there is no data
242 * @sk_allocation: allocation mode
243 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
244 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
245 * @sk_sndbuf: size of send buffer in bytes
246 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
247 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
248 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
249 * @sk_no_check_rx: allow zero checksum in RX packets
250 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
251 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
252 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
253 * @sk_gso_max_size: Maximum GSO segment size to build
254 * @sk_gso_max_segs: Maximum number of GSO segments
255 * @sk_lingertime: %SO_LINGER l_linger setting
256 * @sk_backlog: always used with the per-socket spinlock held
257 * @sk_callback_lock: used with the callbacks in the end of this struct
258 * @sk_error_queue: rarely used
259 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
260 * IPV6_ADDRFORM for instance)
261 * @sk_err: last error
262 * @sk_err_soft: errors that don't cause failure but are the cause of a
263 * persistent failure not just 'timed out'
264 * @sk_drops: raw/udp drops counter
265 * @sk_ack_backlog: current listen backlog
266 * @sk_max_ack_backlog: listen backlog set in listen()
267 * @sk_priority: %SO_PRIORITY setting
268 * @sk_cgrp_prioidx: socket group's priority map index
269 * @sk_type: socket type (%SOCK_STREAM, etc)
270 * @sk_protocol: which protocol this socket belongs in this network family
271 * @sk_peer_pid: &struct pid for this socket's peer
272 * @sk_peer_cred: %SO_PEERCRED setting
273 * @sk_rcvlowat: %SO_RCVLOWAT setting
274 * @sk_rcvtimeo: %SO_RCVTIMEO setting
275 * @sk_sndtimeo: %SO_SNDTIMEO setting
276 * @sk_rxhash: flow hash received from netif layer
277 * @sk_incoming_cpu: record cpu processing incoming packets
278 * @sk_txhash: computed flow hash for use on transmit
279 * @sk_filter: socket filtering instructions
280 * @sk_timer: sock cleanup timer
281 * @sk_stamp: time stamp of last packet received
282 * @sk_tsflags: SO_TIMESTAMPING socket options
283 * @sk_tskey: counter to disambiguate concurrent tstamp requests
284 * @sk_socket: Identd and reporting IO signals
285 * @sk_user_data: RPC layer private data
286 * @sk_frag: cached page frag
287 * @sk_peek_off: current peek_offset value
288 * @sk_send_head: front of stuff to transmit
289 * @sk_security: used by security modules
290 * @sk_mark: generic packet mark
291 * @sk_classid: this socket's cgroup classid
292 * @sk_cgrp: this socket's cgroup-specific proto data
293 * @sk_write_pending: a write to stream socket waits to start
294 * @sk_state_change: callback to indicate change in the state of the sock
295 * @sk_data_ready: callback to indicate there is data to be processed
296 * @sk_write_space: callback to indicate there is bf sending space available
297 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
298 * @sk_backlog_rcv: callback to process the backlog
299 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
303 * Now struct inet_timewait_sock also uses sock_common, so please just
304 * don't add nothing before this first member (__sk_common) --acme
306 struct sock_common __sk_common
;
307 #define sk_node __sk_common.skc_node
308 #define sk_nulls_node __sk_common.skc_nulls_node
309 #define sk_refcnt __sk_common.skc_refcnt
310 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
312 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
313 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
314 #define sk_hash __sk_common.skc_hash
315 #define sk_portpair __sk_common.skc_portpair
316 #define sk_num __sk_common.skc_num
317 #define sk_dport __sk_common.skc_dport
318 #define sk_addrpair __sk_common.skc_addrpair
319 #define sk_daddr __sk_common.skc_daddr
320 #define sk_rcv_saddr __sk_common.skc_rcv_saddr
321 #define sk_family __sk_common.skc_family
322 #define sk_state __sk_common.skc_state
323 #define sk_reuse __sk_common.skc_reuse
324 #define sk_reuseport __sk_common.skc_reuseport
325 #define sk_ipv6only __sk_common.skc_ipv6only
326 #define sk_net_refcnt __sk_common.skc_net_refcnt
327 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
328 #define sk_bind_node __sk_common.skc_bind_node
329 #define sk_prot __sk_common.skc_prot
330 #define sk_net __sk_common.skc_net
331 #define sk_v6_daddr __sk_common.skc_v6_daddr
332 #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
333 #define sk_cookie __sk_common.skc_cookie
335 socket_lock_t sk_lock
;
336 struct sk_buff_head sk_receive_queue
;
338 * The backlog queue is special, it is always used with
339 * the per-socket spinlock held and requires low latency
340 * access. Therefore we special case it's implementation.
341 * Note : rmem_alloc is in this structure to fill a hole
342 * on 64bit arches, not because its logically part of
348 struct sk_buff
*head
;
349 struct sk_buff
*tail
;
351 #define sk_rmem_alloc sk_backlog.rmem_alloc
352 int sk_forward_alloc
;
358 * Warned : sk_incoming_cpu can be set from softirq,
359 * Do not use this hole without fully understanding possible issues.
363 #ifdef CONFIG_NET_RX_BUSY_POLL
364 unsigned int sk_napi_id
;
365 unsigned int sk_ll_usec
;
370 struct sk_filter __rcu
*sk_filter
;
371 struct socket_wq __rcu
*sk_wq
;
374 struct xfrm_policy
*sk_policy
[2];
376 unsigned long sk_flags
;
377 struct dst_entry
*sk_rx_dst
;
378 struct dst_entry __rcu
*sk_dst_cache
;
379 spinlock_t sk_dst_lock
;
380 atomic_t sk_wmem_alloc
;
381 atomic_t sk_omem_alloc
;
383 struct sk_buff_head sk_write_queue
;
384 kmemcheck_bitfield_begin(flags
);
385 unsigned int sk_shutdown
: 2,
391 kmemcheck_bitfield_end(flags
);
394 u32 sk_pacing_rate
; /* bytes per second */
395 u32 sk_max_pacing_rate
;
396 netdev_features_t sk_route_caps
;
397 netdev_features_t sk_route_nocaps
;
399 unsigned int sk_gso_max_size
;
402 unsigned long sk_lingertime
;
403 struct sk_buff_head sk_error_queue
;
404 struct proto
*sk_prot_creator
;
405 rwlock_t sk_callback_lock
;
409 u32 sk_max_ack_backlog
;
411 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
412 __u32 sk_cgrp_prioidx
;
414 struct pid
*sk_peer_pid
;
415 const struct cred
*sk_peer_cred
;
418 struct timer_list sk_timer
;
422 struct socket
*sk_socket
;
424 struct page_frag sk_frag
;
425 struct sk_buff
*sk_send_head
;
427 int sk_write_pending
;
428 #ifdef CONFIG_SECURITY
432 #ifdef CONFIG_CGROUP_NET_CLASSID
435 struct cg_proto
*sk_cgrp
;
436 void (*sk_state_change
)(struct sock
*sk
);
437 void (*sk_data_ready
)(struct sock
*sk
);
438 void (*sk_write_space
)(struct sock
*sk
);
439 void (*sk_error_report
)(struct sock
*sk
);
440 int (*sk_backlog_rcv
)(struct sock
*sk
,
441 struct sk_buff
*skb
);
442 void (*sk_destruct
)(struct sock
*sk
);
445 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
447 #define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
448 #define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
451 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
452 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
453 * on a socket means that the socket will reuse everybody else's port
454 * without looking at the other's sk_reuse value.
457 #define SK_NO_REUSE 0
458 #define SK_CAN_REUSE 1
459 #define SK_FORCE_REUSE 2
461 static inline int sk_peek_offset(struct sock
*sk
, int flags
)
463 if ((flags
& MSG_PEEK
) && (sk
->sk_peek_off
>= 0))
464 return sk
->sk_peek_off
;
469 static inline void sk_peek_offset_bwd(struct sock
*sk
, int val
)
471 if (sk
->sk_peek_off
>= 0) {
472 if (sk
->sk_peek_off
>= val
)
473 sk
->sk_peek_off
-= val
;
479 static inline void sk_peek_offset_fwd(struct sock
*sk
, int val
)
481 if (sk
->sk_peek_off
>= 0)
482 sk
->sk_peek_off
+= val
;
486 * Hashed lists helper routines
488 static inline struct sock
*sk_entry(const struct hlist_node
*node
)
490 return hlist_entry(node
, struct sock
, sk_node
);
493 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
495 return hlist_entry(head
->first
, struct sock
, sk_node
);
498 static inline struct sock
*sk_head(const struct hlist_head
*head
)
500 return hlist_empty(head
) ? NULL
: __sk_head(head
);
503 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
505 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
508 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
510 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
513 static inline struct sock
*sk_next(const struct sock
*sk
)
515 return sk
->sk_node
.next
?
516 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
519 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
521 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
522 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
523 struct sock
, sk_nulls_node
) :
527 static inline bool sk_unhashed(const struct sock
*sk
)
529 return hlist_unhashed(&sk
->sk_node
);
532 static inline bool sk_hashed(const struct sock
*sk
)
534 return !sk_unhashed(sk
);
537 static inline void sk_node_init(struct hlist_node
*node
)
542 static inline void sk_nulls_node_init(struct hlist_nulls_node
*node
)
547 static inline void __sk_del_node(struct sock
*sk
)
549 __hlist_del(&sk
->sk_node
);
552 /* NB: equivalent to hlist_del_init_rcu */
553 static inline bool __sk_del_node_init(struct sock
*sk
)
557 sk_node_init(&sk
->sk_node
);
563 /* Grab socket reference count. This operation is valid only
564 when sk is ALREADY grabbed f.e. it is found in hash table
565 or a list and the lookup is made under lock preventing hash table
569 static inline void sock_hold(struct sock
*sk
)
571 atomic_inc(&sk
->sk_refcnt
);
574 /* Ungrab socket in the context, which assumes that socket refcnt
575 cannot hit zero, f.e. it is true in context of any socketcall.
577 static inline void __sock_put(struct sock
*sk
)
579 atomic_dec(&sk
->sk_refcnt
);
582 static inline bool sk_del_node_init(struct sock
*sk
)
584 bool rc
= __sk_del_node_init(sk
);
587 /* paranoid for a while -acme */
588 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
593 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
595 static inline bool __sk_nulls_del_node_init_rcu(struct sock
*sk
)
598 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
604 static inline bool sk_nulls_del_node_init_rcu(struct sock
*sk
)
606 bool rc
= __sk_nulls_del_node_init_rcu(sk
);
609 /* paranoid for a while -acme */
610 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
616 static inline void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
618 hlist_add_head(&sk
->sk_node
, list
);
621 static inline void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
624 __sk_add_node(sk
, list
);
627 static inline void sk_add_node_rcu(struct sock
*sk
, struct hlist_head
*list
)
630 hlist_add_head_rcu(&sk
->sk_node
, list
);
633 static inline void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
635 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
638 static inline void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
641 __sk_nulls_add_node_rcu(sk
, list
);
644 static inline void __sk_del_bind_node(struct sock
*sk
)
646 __hlist_del(&sk
->sk_bind_node
);
649 static inline void sk_add_bind_node(struct sock
*sk
,
650 struct hlist_head
*list
)
652 hlist_add_head(&sk
->sk_bind_node
, list
);
655 #define sk_for_each(__sk, list) \
656 hlist_for_each_entry(__sk, list, sk_node)
657 #define sk_for_each_rcu(__sk, list) \
658 hlist_for_each_entry_rcu(__sk, list, sk_node)
659 #define sk_nulls_for_each(__sk, node, list) \
660 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
661 #define sk_nulls_for_each_rcu(__sk, node, list) \
662 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
663 #define sk_for_each_from(__sk) \
664 hlist_for_each_entry_from(__sk, sk_node)
665 #define sk_nulls_for_each_from(__sk, node) \
666 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
667 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
668 #define sk_for_each_safe(__sk, tmp, list) \
669 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
670 #define sk_for_each_bound(__sk, list) \
671 hlist_for_each_entry(__sk, list, sk_bind_node)
674 * sk_nulls_for_each_entry_offset - iterate over a list at a given struct offset
675 * @tpos: the type * to use as a loop cursor.
676 * @pos: the &struct hlist_node to use as a loop cursor.
677 * @head: the head for your list.
678 * @offset: offset of hlist_node within the struct.
681 #define sk_nulls_for_each_entry_offset(tpos, pos, head, offset) \
682 for (pos = (head)->first; \
683 (!is_a_nulls(pos)) && \
684 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
687 static inline struct user_namespace
*sk_user_ns(struct sock
*sk
)
689 /* Careful only use this in a context where these parameters
690 * can not change and must all be valid, such as recvmsg from
693 return sk
->sk_socket
->file
->f_cred
->user_ns
;
707 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
708 SOCK_DBG
, /* %SO_DEBUG setting */
709 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
710 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
711 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
712 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
713 SOCK_MEMALLOC
, /* VM depends on this socket for swapping */
714 SOCK_TIMESTAMPING_RX_SOFTWARE
, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
715 SOCK_FASYNC
, /* fasync() active */
717 SOCK_ZEROCOPY
, /* buffers from userspace */
718 SOCK_WIFI_STATUS
, /* push wifi status to userspace */
719 SOCK_NOFCS
, /* Tell NIC not to do the Ethernet FCS.
720 * Will use last 4 bytes of packet sent from
721 * user-space instead.
723 SOCK_FILTER_LOCKED
, /* Filter cannot be changed anymore */
724 SOCK_SELECT_ERR_QUEUE
, /* Wake select on error queue */
727 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
729 nsk
->sk_flags
= osk
->sk_flags
;
732 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
734 __set_bit(flag
, &sk
->sk_flags
);
737 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
739 __clear_bit(flag
, &sk
->sk_flags
);
742 static inline bool sock_flag(const struct sock
*sk
, enum sock_flags flag
)
744 return test_bit(flag
, &sk
->sk_flags
);
748 extern struct static_key memalloc_socks
;
749 static inline int sk_memalloc_socks(void)
751 return static_key_false(&memalloc_socks
);
755 static inline int sk_memalloc_socks(void)
762 static inline gfp_t
sk_gfp_atomic(struct sock
*sk
, gfp_t gfp_mask
)
764 return GFP_ATOMIC
| (sk
->sk_allocation
& __GFP_MEMALLOC
);
767 static inline void sk_acceptq_removed(struct sock
*sk
)
769 sk
->sk_ack_backlog
--;
772 static inline void sk_acceptq_added(struct sock
*sk
)
774 sk
->sk_ack_backlog
++;
777 static inline bool sk_acceptq_is_full(const struct sock
*sk
)
779 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
783 * Compute minimal free write space needed to queue new packets.
785 static inline int sk_stream_min_wspace(const struct sock
*sk
)
787 return sk
->sk_wmem_queued
>> 1;
790 static inline int sk_stream_wspace(const struct sock
*sk
)
792 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
795 void sk_stream_write_space(struct sock
*sk
);
797 /* OOB backlog add */
798 static inline void __sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
800 /* dont let skb dst not refcounted, we are going to leave rcu lock */
803 if (!sk
->sk_backlog
.tail
)
804 sk
->sk_backlog
.head
= skb
;
806 sk
->sk_backlog
.tail
->next
= skb
;
808 sk
->sk_backlog
.tail
= skb
;
813 * Take into account size of receive queue and backlog queue
814 * Do not take into account this skb truesize,
815 * to allow even a single big packet to come.
817 static inline bool sk_rcvqueues_full(const struct sock
*sk
, unsigned int limit
)
819 unsigned int qsize
= sk
->sk_backlog
.len
+ atomic_read(&sk
->sk_rmem_alloc
);
821 return qsize
> limit
;
824 /* The per-socket spinlock must be held here. */
825 static inline __must_check
int sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
,
828 if (sk_rcvqueues_full(sk
, limit
))
832 * If the skb was allocated from pfmemalloc reserves, only
833 * allow SOCK_MEMALLOC sockets to use it as this socket is
834 * helping free memory
836 if (skb_pfmemalloc(skb
) && !sock_flag(sk
, SOCK_MEMALLOC
))
839 __sk_add_backlog(sk
, skb
);
840 sk
->sk_backlog
.len
+= skb
->truesize
;
844 int __sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
846 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
848 if (sk_memalloc_socks() && skb_pfmemalloc(skb
))
849 return __sk_backlog_rcv(sk
, skb
);
851 return sk
->sk_backlog_rcv(sk
, skb
);
854 static inline void sk_incoming_cpu_update(struct sock
*sk
)
856 sk
->sk_incoming_cpu
= raw_smp_processor_id();
859 static inline void sock_rps_record_flow_hash(__u32 hash
)
862 struct rps_sock_flow_table
*sock_flow_table
;
865 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
866 rps_record_sock_flow(sock_flow_table
, hash
);
871 static inline void sock_rps_record_flow(const struct sock
*sk
)
874 sock_rps_record_flow_hash(sk
->sk_rxhash
);
878 static inline void sock_rps_save_rxhash(struct sock
*sk
,
879 const struct sk_buff
*skb
)
882 if (unlikely(sk
->sk_rxhash
!= skb
->hash
))
883 sk
->sk_rxhash
= skb
->hash
;
887 static inline void sock_rps_reset_rxhash(struct sock
*sk
)
894 #define sk_wait_event(__sk, __timeo, __condition) \
896 release_sock(__sk); \
897 __rc = __condition; \
899 *(__timeo) = schedule_timeout(*(__timeo)); \
901 sched_annotate_sleep(); \
903 __rc = __condition; \
907 int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
908 int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
909 void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
910 int sk_stream_error(struct sock
*sk
, int flags
, int err
);
911 void sk_stream_kill_queues(struct sock
*sk
);
912 void sk_set_memalloc(struct sock
*sk
);
913 void sk_clear_memalloc(struct sock
*sk
);
915 int sk_wait_data(struct sock
*sk
, long *timeo
, const struct sk_buff
*skb
);
917 struct request_sock_ops
;
918 struct timewait_sock_ops
;
919 struct inet_hashinfo
;
924 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
925 * un-modified. Special care is taken when initializing object to zero.
927 static inline void sk_prot_clear_nulls(struct sock
*sk
, int size
)
929 if (offsetof(struct sock
, sk_node
.next
) != 0)
930 memset(sk
, 0, offsetof(struct sock
, sk_node
.next
));
931 memset(&sk
->sk_node
.pprev
, 0,
932 size
- offsetof(struct sock
, sk_node
.pprev
));
935 /* Networking protocol blocks we attach to sockets.
936 * socket layer -> transport layer interface
939 void (*close
)(struct sock
*sk
,
941 int (*connect
)(struct sock
*sk
,
942 struct sockaddr
*uaddr
,
944 int (*disconnect
)(struct sock
*sk
, int flags
);
946 struct sock
* (*accept
)(struct sock
*sk
, int flags
, int *err
);
948 int (*ioctl
)(struct sock
*sk
, int cmd
,
950 int (*init
)(struct sock
*sk
);
951 void (*destroy
)(struct sock
*sk
);
952 void (*shutdown
)(struct sock
*sk
, int how
);
953 int (*setsockopt
)(struct sock
*sk
, int level
,
954 int optname
, char __user
*optval
,
955 unsigned int optlen
);
956 int (*getsockopt
)(struct sock
*sk
, int level
,
957 int optname
, char __user
*optval
,
960 int (*compat_setsockopt
)(struct sock
*sk
,
962 int optname
, char __user
*optval
,
963 unsigned int optlen
);
964 int (*compat_getsockopt
)(struct sock
*sk
,
966 int optname
, char __user
*optval
,
968 int (*compat_ioctl
)(struct sock
*sk
,
969 unsigned int cmd
, unsigned long arg
);
971 int (*sendmsg
)(struct sock
*sk
, struct msghdr
*msg
,
973 int (*recvmsg
)(struct sock
*sk
, struct msghdr
*msg
,
974 size_t len
, int noblock
, int flags
,
976 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
977 int offset
, size_t size
, int flags
);
978 int (*bind
)(struct sock
*sk
,
979 struct sockaddr
*uaddr
, int addr_len
);
981 int (*backlog_rcv
) (struct sock
*sk
,
982 struct sk_buff
*skb
);
984 void (*release_cb
)(struct sock
*sk
);
986 /* Keeping track of sk's, looking them up, and port selection methods. */
987 void (*hash
)(struct sock
*sk
);
988 void (*unhash
)(struct sock
*sk
);
989 void (*rehash
)(struct sock
*sk
);
990 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
991 void (*clear_sk
)(struct sock
*sk
, int size
);
993 /* Keeping track of sockets in use */
994 #ifdef CONFIG_PROC_FS
995 unsigned int inuse_idx
;
998 bool (*stream_memory_free
)(const struct sock
*sk
);
999 /* Memory pressure */
1000 void (*enter_memory_pressure
)(struct sock
*sk
);
1001 atomic_long_t
*memory_allocated
; /* Current allocated memory. */
1002 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
1004 * Pressure flag: try to collapse.
1005 * Technical note: it is used by multiple contexts non atomically.
1006 * All the __sk_mem_schedule() is of this nature: accounting
1007 * is strict, actions are advisory and have some latency.
1009 int *memory_pressure
;
1016 struct kmem_cache
*slab
;
1017 unsigned int obj_size
;
1020 struct percpu_counter
*orphan_count
;
1022 struct request_sock_ops
*rsk_prot
;
1023 struct timewait_sock_ops
*twsk_prot
;
1026 struct inet_hashinfo
*hashinfo
;
1027 struct udp_table
*udp_table
;
1028 struct raw_hashinfo
*raw_hash
;
1031 struct module
*owner
;
1035 struct list_head node
;
1036 #ifdef SOCK_REFCNT_DEBUG
1039 #ifdef CONFIG_MEMCG_KMEM
1041 * cgroup specific init/deinit functions. Called once for all
1042 * protocols that implement it, from cgroups populate function.
1043 * This function has to setup any files the protocol want to
1044 * appear in the kmem cgroup filesystem.
1046 int (*init_cgroup
)(struct mem_cgroup
*memcg
,
1047 struct cgroup_subsys
*ss
);
1048 void (*destroy_cgroup
)(struct mem_cgroup
*memcg
);
1049 struct cg_proto
*(*proto_cgroup
)(struct mem_cgroup
*memcg
);
1053 int proto_register(struct proto
*prot
, int alloc_slab
);
1054 void proto_unregister(struct proto
*prot
);
1056 #ifdef SOCK_REFCNT_DEBUG
1057 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
1059 atomic_inc(&sk
->sk_prot
->socks
);
1062 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
1064 atomic_dec(&sk
->sk_prot
->socks
);
1065 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
1066 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
1069 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
1071 if (atomic_read(&sk
->sk_refcnt
) != 1)
1072 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
1073 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
1075 #else /* SOCK_REFCNT_DEBUG */
1076 #define sk_refcnt_debug_inc(sk) do { } while (0)
1077 #define sk_refcnt_debug_dec(sk) do { } while (0)
1078 #define sk_refcnt_debug_release(sk) do { } while (0)
1079 #endif /* SOCK_REFCNT_DEBUG */
1081 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
1082 extern struct static_key memcg_socket_limit_enabled
;
1083 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
1084 struct cg_proto
*cg_proto
)
1086 return proto
->proto_cgroup(parent_mem_cgroup(cg_proto
->memcg
));
1088 #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
1090 #define mem_cgroup_sockets_enabled 0
1091 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
1092 struct cg_proto
*cg_proto
)
1098 static inline bool sk_stream_memory_free(const struct sock
*sk
)
1100 if (sk
->sk_wmem_queued
>= sk
->sk_sndbuf
)
1103 return sk
->sk_prot
->stream_memory_free
?
1104 sk
->sk_prot
->stream_memory_free(sk
) : true;
1107 static inline bool sk_stream_is_writeable(const struct sock
*sk
)
1109 return sk_stream_wspace(sk
) >= sk_stream_min_wspace(sk
) &&
1110 sk_stream_memory_free(sk
);
1114 static inline bool sk_has_memory_pressure(const struct sock
*sk
)
1116 return sk
->sk_prot
->memory_pressure
!= NULL
;
1119 static inline bool sk_under_memory_pressure(const struct sock
*sk
)
1121 if (!sk
->sk_prot
->memory_pressure
)
1124 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1125 return !!sk
->sk_cgrp
->memory_pressure
;
1127 return !!*sk
->sk_prot
->memory_pressure
;
1130 static inline void sk_leave_memory_pressure(struct sock
*sk
)
1132 int *memory_pressure
= sk
->sk_prot
->memory_pressure
;
1134 if (!memory_pressure
)
1137 if (*memory_pressure
)
1138 *memory_pressure
= 0;
1140 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1141 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1142 struct proto
*prot
= sk
->sk_prot
;
1144 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1145 cg_proto
->memory_pressure
= 0;
1150 static inline void sk_enter_memory_pressure(struct sock
*sk
)
1152 if (!sk
->sk_prot
->enter_memory_pressure
)
1155 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1156 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1157 struct proto
*prot
= sk
->sk_prot
;
1159 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1160 cg_proto
->memory_pressure
= 1;
1163 sk
->sk_prot
->enter_memory_pressure(sk
);
1166 static inline long sk_prot_mem_limits(const struct sock
*sk
, int index
)
1168 long *prot
= sk
->sk_prot
->sysctl_mem
;
1169 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1170 prot
= sk
->sk_cgrp
->sysctl_mem
;
1174 static inline void memcg_memory_allocated_add(struct cg_proto
*prot
,
1178 page_counter_charge(&prot
->memory_allocated
, amt
);
1180 if (page_counter_read(&prot
->memory_allocated
) >
1181 prot
->memory_allocated
.limit
)
1182 *parent_status
= OVER_LIMIT
;
1185 static inline void memcg_memory_allocated_sub(struct cg_proto
*prot
,
1188 page_counter_uncharge(&prot
->memory_allocated
, amt
);
1192 sk_memory_allocated(const struct sock
*sk
)
1194 struct proto
*prot
= sk
->sk_prot
;
1196 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1197 return page_counter_read(&sk
->sk_cgrp
->memory_allocated
);
1199 return atomic_long_read(prot
->memory_allocated
);
1203 sk_memory_allocated_add(struct sock
*sk
, int amt
, int *parent_status
)
1205 struct proto
*prot
= sk
->sk_prot
;
1207 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1208 memcg_memory_allocated_add(sk
->sk_cgrp
, amt
, parent_status
);
1209 /* update the root cgroup regardless */
1210 atomic_long_add_return(amt
, prot
->memory_allocated
);
1211 return page_counter_read(&sk
->sk_cgrp
->memory_allocated
);
1214 return atomic_long_add_return(amt
, prot
->memory_allocated
);
1218 sk_memory_allocated_sub(struct sock
*sk
, int amt
)
1220 struct proto
*prot
= sk
->sk_prot
;
1222 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1223 memcg_memory_allocated_sub(sk
->sk_cgrp
, amt
);
1225 atomic_long_sub(amt
, prot
->memory_allocated
);
1228 static inline void sk_sockets_allocated_dec(struct sock
*sk
)
1230 struct proto
*prot
= sk
->sk_prot
;
1232 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1233 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1235 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1236 percpu_counter_dec(&cg_proto
->sockets_allocated
);
1239 percpu_counter_dec(prot
->sockets_allocated
);
1242 static inline void sk_sockets_allocated_inc(struct sock
*sk
)
1244 struct proto
*prot
= sk
->sk_prot
;
1246 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1247 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1249 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1250 percpu_counter_inc(&cg_proto
->sockets_allocated
);
1253 percpu_counter_inc(prot
->sockets_allocated
);
1257 sk_sockets_allocated_read_positive(struct sock
*sk
)
1259 struct proto
*prot
= sk
->sk_prot
;
1261 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1262 return percpu_counter_read_positive(&sk
->sk_cgrp
->sockets_allocated
);
1264 return percpu_counter_read_positive(prot
->sockets_allocated
);
1268 proto_sockets_allocated_sum_positive(struct proto
*prot
)
1270 return percpu_counter_sum_positive(prot
->sockets_allocated
);
1274 proto_memory_allocated(struct proto
*prot
)
1276 return atomic_long_read(prot
->memory_allocated
);
1280 proto_memory_pressure(struct proto
*prot
)
1282 if (!prot
->memory_pressure
)
1284 return !!*prot
->memory_pressure
;
1288 #ifdef CONFIG_PROC_FS
1289 /* Called with local bh disabled */
1290 void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int inc
);
1291 int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
1293 static inline void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
1300 /* With per-bucket locks this operation is not-atomic, so that
1301 * this version is not worse.
1303 static inline void __sk_prot_rehash(struct sock
*sk
)
1305 sk
->sk_prot
->unhash(sk
);
1306 sk
->sk_prot
->hash(sk
);
1309 void sk_prot_clear_portaddr_nulls(struct sock
*sk
, int size
);
1311 /* About 10 seconds */
1312 #define SOCK_DESTROY_TIME (10*HZ)
1314 /* Sockets 0-1023 can't be bound to unless you are superuser */
1315 #define PROT_SOCK 1024
1317 #define SHUTDOWN_MASK 3
1318 #define RCV_SHUTDOWN 1
1319 #define SEND_SHUTDOWN 2
1321 #define SOCK_SNDBUF_LOCK 1
1322 #define SOCK_RCVBUF_LOCK 2
1323 #define SOCK_BINDADDR_LOCK 4
1324 #define SOCK_BINDPORT_LOCK 8
1326 struct socket_alloc
{
1327 struct socket socket
;
1328 struct inode vfs_inode
;
1331 static inline struct socket
*SOCKET_I(struct inode
*inode
)
1333 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
1336 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
1338 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
1342 * Functions for memory accounting
1344 int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
1345 void __sk_mem_reclaim(struct sock
*sk
, int amount
);
1347 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1348 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1349 #define SK_MEM_SEND 0
1350 #define SK_MEM_RECV 1
1352 static inline int sk_mem_pages(int amt
)
1354 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
1357 static inline bool sk_has_account(struct sock
*sk
)
1359 /* return true if protocol supports memory accounting */
1360 return !!sk
->sk_prot
->memory_allocated
;
1363 static inline bool sk_wmem_schedule(struct sock
*sk
, int size
)
1365 if (!sk_has_account(sk
))
1367 return size
<= sk
->sk_forward_alloc
||
1368 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
1372 sk_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
, int size
)
1374 if (!sk_has_account(sk
))
1376 return size
<= sk
->sk_forward_alloc
||
1377 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
) ||
1378 skb_pfmemalloc(skb
);
1381 static inline void sk_mem_reclaim(struct sock
*sk
)
1383 if (!sk_has_account(sk
))
1385 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
1386 __sk_mem_reclaim(sk
, sk
->sk_forward_alloc
);
1389 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
1391 if (!sk_has_account(sk
))
1393 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
1394 __sk_mem_reclaim(sk
, sk
->sk_forward_alloc
- 1);
1397 static inline void sk_mem_charge(struct sock
*sk
, int size
)
1399 if (!sk_has_account(sk
))
1401 sk
->sk_forward_alloc
-= size
;
1404 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
1406 if (!sk_has_account(sk
))
1408 sk
->sk_forward_alloc
+= size
;
1411 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
1413 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
1414 sk
->sk_wmem_queued
-= skb
->truesize
;
1415 sk_mem_uncharge(sk
, skb
->truesize
);
1419 /* Used by processes to "lock" a socket state, so that
1420 * interrupts and bottom half handlers won't change it
1421 * from under us. It essentially blocks any incoming
1422 * packets, so that we won't get any new data or any
1423 * packets that change the state of the socket.
1425 * While locked, BH processing will add new packets to
1426 * the backlog queue. This queue is processed by the
1427 * owner of the socket lock right before it is released.
1429 * Since ~2.3.5 it is also exclusive sleep lock serializing
1430 * accesses from user process context.
1432 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1434 static inline void sock_release_ownership(struct sock
*sk
)
1436 sk
->sk_lock
.owned
= 0;
1440 * Macro so as to not evaluate some arguments when
1441 * lockdep is not enabled.
1443 * Mark both the sk_lock and the sk_lock.slock as a
1444 * per-address-family lock class.
1446 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1448 sk->sk_lock.owned = 0; \
1449 init_waitqueue_head(&sk->sk_lock.wq); \
1450 spin_lock_init(&(sk)->sk_lock.slock); \
1451 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1452 sizeof((sk)->sk_lock)); \
1453 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1455 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1458 void lock_sock_nested(struct sock
*sk
, int subclass
);
1460 static inline void lock_sock(struct sock
*sk
)
1462 lock_sock_nested(sk
, 0);
1465 void release_sock(struct sock
*sk
);
1467 /* BH context may only use the following locking interface. */
1468 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1469 #define bh_lock_sock_nested(__sk) \
1470 spin_lock_nested(&((__sk)->sk_lock.slock), \
1471 SINGLE_DEPTH_NESTING)
1472 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1474 bool lock_sock_fast(struct sock
*sk
);
1476 * unlock_sock_fast - complement of lock_sock_fast
1480 * fast unlock socket for user context.
1481 * If slow mode is on, we call regular release_sock()
1483 static inline void unlock_sock_fast(struct sock
*sk
, bool slow
)
1488 spin_unlock_bh(&sk
->sk_lock
.slock
);
1492 struct sock
*sk_alloc(struct net
*net
, int family
, gfp_t priority
,
1493 struct proto
*prot
, int kern
);
1494 void sk_free(struct sock
*sk
);
1495 void sk_destruct(struct sock
*sk
);
1496 struct sock
*sk_clone_lock(const struct sock
*sk
, const gfp_t priority
);
1498 struct sk_buff
*sock_wmalloc(struct sock
*sk
, unsigned long size
, int force
,
1500 void sock_wfree(struct sk_buff
*skb
);
1501 void skb_orphan_partial(struct sk_buff
*skb
);
1502 void sock_rfree(struct sk_buff
*skb
);
1503 void sock_efree(struct sk_buff
*skb
);
1505 void sock_edemux(struct sk_buff
*skb
);
1507 #define sock_edemux(skb) sock_efree(skb)
1510 int sock_setsockopt(struct socket
*sock
, int level
, int op
,
1511 char __user
*optval
, unsigned int optlen
);
1513 int sock_getsockopt(struct socket
*sock
, int level
, int op
,
1514 char __user
*optval
, int __user
*optlen
);
1515 struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
, unsigned long size
,
1516 int noblock
, int *errcode
);
1517 struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
, unsigned long header_len
,
1518 unsigned long data_len
, int noblock
,
1519 int *errcode
, int max_page_order
);
1520 void *sock_kmalloc(struct sock
*sk
, int size
, gfp_t priority
);
1521 void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
1522 void sock_kzfree_s(struct sock
*sk
, void *mem
, int size
);
1523 void sk_send_sigurg(struct sock
*sk
);
1526 * Functions to fill in entries in struct proto_ops when a protocol
1527 * does not implement a particular function.
1529 int sock_no_bind(struct socket
*, struct sockaddr
*, int);
1530 int sock_no_connect(struct socket
*, struct sockaddr
*, int, int);
1531 int sock_no_socketpair(struct socket
*, struct socket
*);
1532 int sock_no_accept(struct socket
*, struct socket
*, int);
1533 int sock_no_getname(struct socket
*, struct sockaddr
*, int *, int);
1534 unsigned int sock_no_poll(struct file
*, struct socket
*,
1535 struct poll_table_struct
*);
1536 int sock_no_ioctl(struct socket
*, unsigned int, unsigned long);
1537 int sock_no_listen(struct socket
*, int);
1538 int sock_no_shutdown(struct socket
*, int);
1539 int sock_no_getsockopt(struct socket
*, int , int, char __user
*, int __user
*);
1540 int sock_no_setsockopt(struct socket
*, int, int, char __user
*, unsigned int);
1541 int sock_no_sendmsg(struct socket
*, struct msghdr
*, size_t);
1542 int sock_no_recvmsg(struct socket
*, struct msghdr
*, size_t, int);
1543 int sock_no_mmap(struct file
*file
, struct socket
*sock
,
1544 struct vm_area_struct
*vma
);
1545 ssize_t
sock_no_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
1546 size_t size
, int flags
);
1549 * Functions to fill in entries in struct proto_ops when a protocol
1550 * uses the inet style.
1552 int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1553 char __user
*optval
, int __user
*optlen
);
1554 int sock_common_recvmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
,
1556 int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1557 char __user
*optval
, unsigned int optlen
);
1558 int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
1559 int optname
, char __user
*optval
, int __user
*optlen
);
1560 int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
1561 int optname
, char __user
*optval
, unsigned int optlen
);
1563 void sk_common_release(struct sock
*sk
);
1566 * Default socket callbacks and setup code
1569 /* Initialise core socket variables */
1570 void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1573 * Socket reference counting postulates.
1575 * * Each user of socket SHOULD hold a reference count.
1576 * * Each access point to socket (an hash table bucket, reference from a list,
1577 * running timer, skb in flight MUST hold a reference count.
1578 * * When reference count hits 0, it means it will never increase back.
1579 * * When reference count hits 0, it means that no references from
1580 * outside exist to this socket and current process on current CPU
1581 * is last user and may/should destroy this socket.
1582 * * sk_free is called from any context: process, BH, IRQ. When
1583 * it is called, socket has no references from outside -> sk_free
1584 * may release descendant resources allocated by the socket, but
1585 * to the time when it is called, socket is NOT referenced by any
1586 * hash tables, lists etc.
1587 * * Packets, delivered from outside (from network or from another process)
1588 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1589 * when they sit in queue. Otherwise, packets will leak to hole, when
1590 * socket is looked up by one cpu and unhasing is made by another CPU.
1591 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1592 * (leak to backlog). Packet socket does all the processing inside
1593 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1594 * use separate SMP lock, so that they are prone too.
1597 /* Ungrab socket and destroy it, if it was the last reference. */
1598 static inline void sock_put(struct sock
*sk
)
1600 if (atomic_dec_and_test(&sk
->sk_refcnt
))
1603 /* Generic version of sock_put(), dealing with all sockets
1604 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
1606 void sock_gen_put(struct sock
*sk
);
1608 int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
, const int nested
);
1610 static inline void sk_tx_queue_set(struct sock
*sk
, int tx_queue
)
1612 sk
->sk_tx_queue_mapping
= tx_queue
;
1615 static inline void sk_tx_queue_clear(struct sock
*sk
)
1617 sk
->sk_tx_queue_mapping
= -1;
1620 static inline int sk_tx_queue_get(const struct sock
*sk
)
1622 return sk
? sk
->sk_tx_queue_mapping
: -1;
1625 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1627 sk_tx_queue_clear(sk
);
1628 sk
->sk_socket
= sock
;
1631 static inline wait_queue_head_t
*sk_sleep(struct sock
*sk
)
1633 BUILD_BUG_ON(offsetof(struct socket_wq
, wait
) != 0);
1634 return &rcu_dereference_raw(sk
->sk_wq
)->wait
;
1636 /* Detach socket from process context.
1637 * Announce socket dead, detach it from wait queue and inode.
1638 * Note that parent inode held reference count on this struct sock,
1639 * we do not release it in this function, because protocol
1640 * probably wants some additional cleanups or even continuing
1641 * to work with this socket (TCP).
1643 static inline void sock_orphan(struct sock
*sk
)
1645 write_lock_bh(&sk
->sk_callback_lock
);
1646 sock_set_flag(sk
, SOCK_DEAD
);
1647 sk_set_socket(sk
, NULL
);
1649 write_unlock_bh(&sk
->sk_callback_lock
);
1652 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1654 write_lock_bh(&sk
->sk_callback_lock
);
1655 sk
->sk_wq
= parent
->wq
;
1657 sk_set_socket(sk
, parent
);
1658 security_sock_graft(sk
, parent
);
1659 write_unlock_bh(&sk
->sk_callback_lock
);
1662 kuid_t
sock_i_uid(struct sock
*sk
);
1663 unsigned long sock_i_ino(struct sock
*sk
);
1665 static inline void sk_set_txhash(struct sock
*sk
)
1667 sk
->sk_txhash
= prandom_u32();
1669 if (unlikely(!sk
->sk_txhash
))
1673 static inline void sk_rethink_txhash(struct sock
*sk
)
1679 static inline struct dst_entry
*
1680 __sk_dst_get(struct sock
*sk
)
1682 return rcu_dereference_check(sk
->sk_dst_cache
, sock_owned_by_user(sk
) ||
1683 lockdep_is_held(&sk
->sk_lock
.slock
));
1686 static inline struct dst_entry
*
1687 sk_dst_get(struct sock
*sk
)
1689 struct dst_entry
*dst
;
1692 dst
= rcu_dereference(sk
->sk_dst_cache
);
1693 if (dst
&& !atomic_inc_not_zero(&dst
->__refcnt
))
1699 static inline void dst_negative_advice(struct sock
*sk
)
1701 struct dst_entry
*ndst
, *dst
= __sk_dst_get(sk
);
1703 sk_rethink_txhash(sk
);
1705 if (dst
&& dst
->ops
->negative_advice
) {
1706 ndst
= dst
->ops
->negative_advice(dst
);
1709 rcu_assign_pointer(sk
->sk_dst_cache
, ndst
);
1710 sk_tx_queue_clear(sk
);
1716 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1718 struct dst_entry
*old_dst
;
1720 sk_tx_queue_clear(sk
);
1722 * This can be called while sk is owned by the caller only,
1723 * with no state that can be checked in a rcu_dereference_check() cond
1725 old_dst
= rcu_dereference_raw(sk
->sk_dst_cache
);
1726 rcu_assign_pointer(sk
->sk_dst_cache
, dst
);
1727 dst_release(old_dst
);
1731 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1733 struct dst_entry
*old_dst
;
1735 sk_tx_queue_clear(sk
);
1736 old_dst
= xchg((__force
struct dst_entry
**)&sk
->sk_dst_cache
, dst
);
1737 dst_release(old_dst
);
1741 __sk_dst_reset(struct sock
*sk
)
1743 __sk_dst_set(sk
, NULL
);
1747 sk_dst_reset(struct sock
*sk
)
1749 sk_dst_set(sk
, NULL
);
1752 struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1754 struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1756 bool sk_mc_loop(struct sock
*sk
);
1758 static inline bool sk_can_gso(const struct sock
*sk
)
1760 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1763 void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1765 static inline void sk_nocaps_add(struct sock
*sk
, netdev_features_t flags
)
1767 sk
->sk_route_nocaps
|= flags
;
1768 sk
->sk_route_caps
&= ~flags
;
1771 static inline int skb_do_copy_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1772 struct iov_iter
*from
, char *to
,
1773 int copy
, int offset
)
1775 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1777 if (csum_and_copy_from_iter(to
, copy
, &csum
, from
) != copy
)
1779 skb
->csum
= csum_block_add(skb
->csum
, csum
, offset
);
1780 } else if (sk
->sk_route_caps
& NETIF_F_NOCACHE_COPY
) {
1781 if (copy_from_iter_nocache(to
, copy
, from
) != copy
)
1783 } else if (copy_from_iter(to
, copy
, from
) != copy
)
1789 static inline int skb_add_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1790 struct iov_iter
*from
, int copy
)
1792 int err
, offset
= skb
->len
;
1794 err
= skb_do_copy_data_nocache(sk
, skb
, from
, skb_put(skb
, copy
),
1797 __skb_trim(skb
, offset
);
1802 static inline int skb_copy_to_page_nocache(struct sock
*sk
, struct iov_iter
*from
,
1803 struct sk_buff
*skb
,
1809 err
= skb_do_copy_data_nocache(sk
, skb
, from
, page_address(page
) + off
,
1815 skb
->data_len
+= copy
;
1816 skb
->truesize
+= copy
;
1817 sk
->sk_wmem_queued
+= copy
;
1818 sk_mem_charge(sk
, copy
);
1823 * sk_wmem_alloc_get - returns write allocations
1826 * Returns sk_wmem_alloc minus initial offset of one
1828 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
1830 return atomic_read(&sk
->sk_wmem_alloc
) - 1;
1834 * sk_rmem_alloc_get - returns read allocations
1837 * Returns sk_rmem_alloc
1839 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
1841 return atomic_read(&sk
->sk_rmem_alloc
);
1845 * sk_has_allocations - check if allocations are outstanding
1848 * Returns true if socket has write or read allocations
1850 static inline bool sk_has_allocations(const struct sock
*sk
)
1852 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
1856 * wq_has_sleeper - check if there are any waiting processes
1857 * @wq: struct socket_wq
1859 * Returns true if socket_wq has waiting processes
1861 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1862 * barrier call. They were added due to the race found within the tcp code.
1864 * Consider following tcp code paths:
1868 * sys_select receive packet
1870 * __add_wait_queue update tp->rcv_nxt
1872 * tp->rcv_nxt check sock_def_readable
1874 * schedule rcu_read_lock();
1875 * wq = rcu_dereference(sk->sk_wq);
1876 * if (wq && waitqueue_active(&wq->wait))
1877 * wake_up_interruptible(&wq->wait)
1881 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1882 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1883 * could then endup calling schedule and sleep forever if there are no more
1884 * data on the socket.
1887 static inline bool wq_has_sleeper(struct socket_wq
*wq
)
1889 /* We need to be sure we are in sync with the
1890 * add_wait_queue modifications to the wait queue.
1892 * This memory barrier is paired in the sock_poll_wait.
1895 return wq
&& waitqueue_active(&wq
->wait
);
1899 * sock_poll_wait - place memory barrier behind the poll_wait call.
1901 * @wait_address: socket wait queue
1904 * See the comments in the wq_has_sleeper function.
1906 static inline void sock_poll_wait(struct file
*filp
,
1907 wait_queue_head_t
*wait_address
, poll_table
*p
)
1909 if (!poll_does_not_wait(p
) && wait_address
) {
1910 poll_wait(filp
, wait_address
, p
);
1911 /* We need to be sure we are in sync with the
1912 * socket flags modification.
1914 * This memory barrier is paired in the wq_has_sleeper.
1920 static inline void skb_set_hash_from_sk(struct sk_buff
*skb
, struct sock
*sk
)
1922 if (sk
->sk_txhash
) {
1924 skb
->hash
= sk
->sk_txhash
;
1929 * Queue a received datagram if it will fit. Stream and sequenced
1930 * protocols can't normally use this as they need to fit buffers in
1931 * and play with them.
1933 * Inlined as it's very short and called for pretty much every
1934 * packet ever received.
1937 static inline void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
1941 skb
->destructor
= sock_wfree
;
1942 skb_set_hash_from_sk(skb
, sk
);
1944 * We used to take a refcount on sk, but following operation
1945 * is enough to guarantee sk_free() wont free this sock until
1946 * all in-flight packets are completed
1948 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
1951 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
1955 skb
->destructor
= sock_rfree
;
1956 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
1957 sk_mem_charge(sk
, skb
->truesize
);
1960 void sk_reset_timer(struct sock
*sk
, struct timer_list
*timer
,
1961 unsigned long expires
);
1963 void sk_stop_timer(struct sock
*sk
, struct timer_list
*timer
);
1965 int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
1967 int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
);
1968 struct sk_buff
*sock_dequeue_err_skb(struct sock
*sk
);
1971 * Recover an error report and clear atomically
1974 static inline int sock_error(struct sock
*sk
)
1977 if (likely(!sk
->sk_err
))
1979 err
= xchg(&sk
->sk_err
, 0);
1983 static inline unsigned long sock_wspace(struct sock
*sk
)
1987 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
1988 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
1995 static inline void sk_wake_async(struct sock
*sk
, int how
, int band
)
1997 if (sock_flag(sk
, SOCK_FASYNC
))
1998 sock_wake_async(sk
->sk_socket
, how
, band
);
2001 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2002 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2003 * Note: for send buffers, TCP works better if we can build two skbs at
2006 #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2008 #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2009 #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
2011 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
2013 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
2014 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
2015 sk
->sk_sndbuf
= max_t(u32
, sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
2019 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
,
2020 bool force_schedule
);
2023 * sk_page_frag - return an appropriate page_frag
2026 * If socket allocation mode allows current thread to sleep, it means its
2027 * safe to use the per task page_frag instead of the per socket one.
2029 static inline struct page_frag
*sk_page_frag(struct sock
*sk
)
2031 if (sk
->sk_allocation
& __GFP_WAIT
)
2032 return ¤t
->task_frag
;
2034 return &sk
->sk_frag
;
2037 bool sk_page_frag_refill(struct sock
*sk
, struct page_frag
*pfrag
);
2040 * Default write policy as shown to user space via poll/select/SIGIO
2042 static inline bool sock_writeable(const struct sock
*sk
)
2044 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
2047 static inline gfp_t
gfp_any(void)
2049 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
2052 static inline long sock_rcvtimeo(const struct sock
*sk
, bool noblock
)
2054 return noblock
? 0 : sk
->sk_rcvtimeo
;
2057 static inline long sock_sndtimeo(const struct sock
*sk
, bool noblock
)
2059 return noblock
? 0 : sk
->sk_sndtimeo
;
2062 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
2064 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
2067 /* Alas, with timeout socket operations are not restartable.
2068 * Compare this to poll().
2070 static inline int sock_intr_errno(long timeo
)
2072 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
2075 struct sock_skb_cb
{
2079 /* Store sock_skb_cb at the end of skb->cb[] so protocol families
2080 * using skb->cb[] would keep using it directly and utilize its
2081 * alignement guarantee.
2083 #define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \
2084 sizeof(struct sock_skb_cb)))
2086 #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2087 SOCK_SKB_CB_OFFSET))
2089 #define sock_skb_cb_check_size(size) \
2090 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
2093 sock_skb_set_dropcount(const struct sock
*sk
, struct sk_buff
*skb
)
2095 SOCK_SKB_CB(skb
)->dropcount
= atomic_read(&sk
->sk_drops
);
2098 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
2099 struct sk_buff
*skb
);
2100 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
2101 struct sk_buff
*skb
);
2104 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
2106 ktime_t kt
= skb
->tstamp
;
2107 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
2110 * generate control messages if
2111 * - receive time stamping in software requested
2112 * - software time stamp available and wanted
2113 * - hardware time stamps available and wanted
2115 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
2116 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RX_SOFTWARE
) ||
2117 (kt
.tv64
&& sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
) ||
2118 (hwtstamps
->hwtstamp
.tv64
&&
2119 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
)))
2120 __sock_recv_timestamp(msg
, sk
, skb
);
2124 if (sock_flag(sk
, SOCK_WIFI_STATUS
) && skb
->wifi_acked_valid
)
2125 __sock_recv_wifi_status(msg
, sk
, skb
);
2128 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2129 struct sk_buff
*skb
);
2131 static inline void sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2132 struct sk_buff
*skb
)
2134 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2135 (1UL << SOCK_RCVTSTAMP))
2136 #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2137 SOF_TIMESTAMPING_RAW_HARDWARE)
2139 if (sk
->sk_flags
& FLAGS_TS_OR_DROPS
|| sk
->sk_tsflags
& TSFLAGS_ANY
)
2140 __sock_recv_ts_and_drops(msg
, sk
, skb
);
2142 sk
->sk_stamp
= skb
->tstamp
;
2145 void __sock_tx_timestamp(const struct sock
*sk
, __u8
*tx_flags
);
2148 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2149 * @sk: socket sending this packet
2150 * @tx_flags: completed with instructions for time stamping
2152 * Note : callers should take care of initial *tx_flags value (usually 0)
2154 static inline void sock_tx_timestamp(const struct sock
*sk
, __u8
*tx_flags
)
2156 if (unlikely(sk
->sk_tsflags
))
2157 __sock_tx_timestamp(sk
, tx_flags
);
2158 if (unlikely(sock_flag(sk
, SOCK_WIFI_STATUS
)))
2159 *tx_flags
|= SKBTX_WIFI_STATUS
;
2163 * sk_eat_skb - Release a skb if it is no longer needed
2164 * @sk: socket to eat this skb from
2165 * @skb: socket buffer to eat
2167 * This routine must be called with interrupts disabled or with the socket
2168 * locked so that the sk_buff queue operation is ok.
2170 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
)
2172 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2177 struct net
*sock_net(const struct sock
*sk
)
2179 return read_pnet(&sk
->sk_net
);
2183 void sock_net_set(struct sock
*sk
, struct net
*net
)
2185 write_pnet(&sk
->sk_net
, net
);
2188 static inline struct sock
*skb_steal_sock(struct sk_buff
*skb
)
2191 struct sock
*sk
= skb
->sk
;
2193 skb
->destructor
= NULL
;
2200 /* This helper checks if a socket is a full socket,
2201 * ie _not_ a timewait or request socket.
2203 static inline bool sk_fullsock(const struct sock
*sk
)
2205 return (1 << sk
->sk_state
) & ~(TCPF_TIME_WAIT
| TCPF_NEW_SYN_RECV
);
2208 void sock_enable_timestamp(struct sock
*sk
, int flag
);
2209 int sock_get_timestamp(struct sock
*, struct timeval __user
*);
2210 int sock_get_timestampns(struct sock
*, struct timespec __user
*);
2211 int sock_recv_errqueue(struct sock
*sk
, struct msghdr
*msg
, int len
, int level
,
2214 bool sk_ns_capable(const struct sock
*sk
,
2215 struct user_namespace
*user_ns
, int cap
);
2216 bool sk_capable(const struct sock
*sk
, int cap
);
2217 bool sk_net_capable(const struct sock
*sk
, int cap
);
2219 extern __u32 sysctl_wmem_max
;
2220 extern __u32 sysctl_rmem_max
;
2222 extern int sysctl_tstamp_allow_data
;
2223 extern int sysctl_optmem_max
;
2225 extern __u32 sysctl_wmem_default
;
2226 extern __u32 sysctl_rmem_default
;
2228 #endif /* _SOCK_H */