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>
77 int mem_cgroup_sockets_init(struct mem_cgroup
*memcg
, struct cgroup_subsys
*ss
);
78 void mem_cgroup_sockets_destroy(struct mem_cgroup
*memcg
);
81 int mem_cgroup_sockets_init(struct mem_cgroup
*memcg
, struct cgroup_subsys
*ss
)
86 void mem_cgroup_sockets_destroy(struct mem_cgroup
*memcg
)
91 * This structure really needs to be cleaned up.
92 * Most of it is for TCP, and not used by any of
93 * the other protocols.
96 /* Define this to get the SOCK_DBG debugging facility. */
97 #define SOCK_DEBUGGING
99 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
100 printk(KERN_DEBUG msg); } while (0)
102 /* Validate arguments and do nothing */
103 static inline __printf(2, 3)
104 void SOCK_DEBUG(const struct sock
*sk
, const char *msg
, ...)
109 /* This is the per-socket lock. The spinlock provides a synchronization
110 * between user contexts and software interrupt processing, whereas the
111 * mini-semaphore synchronizes multiple users amongst themselves.
116 wait_queue_head_t wq
;
118 * We express the mutex-alike socket_lock semantics
119 * to the lock validator by explicitly managing
120 * the slock as a lock variant (in addition to
123 #ifdef CONFIG_DEBUG_LOCK_ALLOC
124 struct lockdep_map dep_map
;
132 typedef __u32 __bitwise __portpair
;
133 typedef __u64 __bitwise __addrpair
;
136 * struct sock_common - minimal network layer representation of sockets
137 * @skc_daddr: Foreign IPv4 addr
138 * @skc_rcv_saddr: Bound local IPv4 addr
139 * @skc_hash: hash value used with various protocol lookup tables
140 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
141 * @skc_dport: placeholder for inet_dport/tw_dport
142 * @skc_num: placeholder for inet_num/tw_num
143 * @skc_family: network address family
144 * @skc_state: Connection state
145 * @skc_reuse: %SO_REUSEADDR setting
146 * @skc_reuseport: %SO_REUSEPORT setting
147 * @skc_bound_dev_if: bound device index if != 0
148 * @skc_bind_node: bind hash linkage for various protocol lookup tables
149 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
150 * @skc_prot: protocol handlers inside a network family
151 * @skc_net: reference to the network namespace of this socket
152 * @skc_node: main hash linkage for various protocol lookup tables
153 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
154 * @skc_tx_queue_mapping: tx queue number for this connection
155 * @skc_flags: place holder for sk_flags
156 * %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
157 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
158 * @skc_incoming_cpu: record/match cpu processing incoming packets
159 * @skc_refcnt: reference count
161 * This is the minimal network layer representation of sockets, the header
162 * for struct sock and struct inet_timewait_sock.
165 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
166 * address on 64bit arches : cf INET_MATCH()
169 __addrpair skc_addrpair
;
172 __be32 skc_rcv_saddr
;
176 unsigned int skc_hash
;
177 __u16 skc_u16hashes
[2];
179 /* skc_dport && skc_num must be grouped as well */
181 __portpair skc_portpair
;
188 unsigned short skc_family
;
189 volatile unsigned char skc_state
;
190 unsigned char skc_reuse
:4;
191 unsigned char skc_reuseport
:1;
192 unsigned char skc_ipv6only
:1;
193 unsigned char skc_net_refcnt
:1;
194 int skc_bound_dev_if
;
196 struct hlist_node skc_bind_node
;
197 struct hlist_nulls_node skc_portaddr_node
;
199 struct proto
*skc_prot
;
200 possible_net_t skc_net
;
202 #if IS_ENABLED(CONFIG_IPV6)
203 struct in6_addr skc_v6_daddr
;
204 struct in6_addr skc_v6_rcv_saddr
;
207 atomic64_t skc_cookie
;
209 /* following fields are padding to force
210 * offset(struct sock, sk_refcnt) == 128 on 64bit arches
211 * assuming IPV6 is enabled. We use this padding differently
212 * for different kind of 'sockets'
215 unsigned long skc_flags
;
216 struct sock
*skc_listener
; /* request_sock */
217 struct inet_timewait_death_row
*skc_tw_dr
; /* inet_timewait_sock */
220 * fields between dontcopy_begin/dontcopy_end
221 * are not copied in sock_copy()
224 int skc_dontcopy_begin
[0];
227 struct hlist_node skc_node
;
228 struct hlist_nulls_node skc_nulls_node
;
230 int skc_tx_queue_mapping
;
232 int skc_incoming_cpu
;
234 u32 skc_tw_rcv_nxt
; /* struct tcp_timewait_sock */
239 int skc_dontcopy_end
[0];
242 u32 skc_window_clamp
;
243 u32 skc_tw_snd_nxt
; /* struct tcp_timewait_sock */
250 * struct sock - network layer representation of sockets
251 * @__sk_common: shared layout with inet_timewait_sock
252 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
253 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
254 * @sk_lock: synchronizer
255 * @sk_rcvbuf: size of receive buffer in bytes
256 * @sk_wq: sock wait queue and async head
257 * @sk_rx_dst: receive input route used by early demux
258 * @sk_dst_cache: destination cache
259 * @sk_policy: flow policy
260 * @sk_receive_queue: incoming packets
261 * @sk_wmem_alloc: transmit queue bytes committed
262 * @sk_write_queue: Packet sending queue
263 * @sk_omem_alloc: "o" is "option" or "other"
264 * @sk_wmem_queued: persistent queue size
265 * @sk_forward_alloc: space allocated forward
266 * @sk_napi_id: id of the last napi context to receive data for sk
267 * @sk_ll_usec: usecs to busypoll when there is no data
268 * @sk_allocation: allocation mode
269 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
270 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
271 * @sk_sndbuf: size of send buffer in bytes
272 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
273 * @sk_no_check_rx: allow zero checksum in RX packets
274 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
275 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
276 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
277 * @sk_gso_max_size: Maximum GSO segment size to build
278 * @sk_gso_max_segs: Maximum number of GSO segments
279 * @sk_lingertime: %SO_LINGER l_linger setting
280 * @sk_backlog: always used with the per-socket spinlock held
281 * @sk_callback_lock: used with the callbacks in the end of this struct
282 * @sk_error_queue: rarely used
283 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
284 * IPV6_ADDRFORM for instance)
285 * @sk_err: last error
286 * @sk_err_soft: errors that don't cause failure but are the cause of a
287 * persistent failure not just 'timed out'
288 * @sk_drops: raw/udp drops counter
289 * @sk_ack_backlog: current listen backlog
290 * @sk_max_ack_backlog: listen backlog set in listen()
291 * @sk_priority: %SO_PRIORITY setting
292 * @sk_type: socket type (%SOCK_STREAM, etc)
293 * @sk_protocol: which protocol this socket belongs in this network family
294 * @sk_peer_pid: &struct pid for this socket's peer
295 * @sk_peer_cred: %SO_PEERCRED setting
296 * @sk_rcvlowat: %SO_RCVLOWAT setting
297 * @sk_rcvtimeo: %SO_RCVTIMEO setting
298 * @sk_sndtimeo: %SO_SNDTIMEO setting
299 * @sk_txhash: computed flow hash for use on transmit
300 * @sk_filter: socket filtering instructions
301 * @sk_timer: sock cleanup timer
302 * @sk_stamp: time stamp of last packet received
303 * @sk_tsflags: SO_TIMESTAMPING socket options
304 * @sk_tskey: counter to disambiguate concurrent tstamp requests
305 * @sk_socket: Identd and reporting IO signals
306 * @sk_user_data: RPC layer private data
307 * @sk_frag: cached page frag
308 * @sk_peek_off: current peek_offset value
309 * @sk_send_head: front of stuff to transmit
310 * @sk_security: used by security modules
311 * @sk_mark: generic packet mark
312 * @sk_cgrp_data: cgroup data for this cgroup
313 * @sk_cgrp: this socket's cgroup-specific proto data
314 * @sk_write_pending: a write to stream socket waits to start
315 * @sk_state_change: callback to indicate change in the state of the sock
316 * @sk_data_ready: callback to indicate there is data to be processed
317 * @sk_write_space: callback to indicate there is bf sending space available
318 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
319 * @sk_backlog_rcv: callback to process the backlog
320 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
324 * Now struct inet_timewait_sock also uses sock_common, so please just
325 * don't add nothing before this first member (__sk_common) --acme
327 struct sock_common __sk_common
;
328 #define sk_node __sk_common.skc_node
329 #define sk_nulls_node __sk_common.skc_nulls_node
330 #define sk_refcnt __sk_common.skc_refcnt
331 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
333 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
334 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
335 #define sk_hash __sk_common.skc_hash
336 #define sk_portpair __sk_common.skc_portpair
337 #define sk_num __sk_common.skc_num
338 #define sk_dport __sk_common.skc_dport
339 #define sk_addrpair __sk_common.skc_addrpair
340 #define sk_daddr __sk_common.skc_daddr
341 #define sk_rcv_saddr __sk_common.skc_rcv_saddr
342 #define sk_family __sk_common.skc_family
343 #define sk_state __sk_common.skc_state
344 #define sk_reuse __sk_common.skc_reuse
345 #define sk_reuseport __sk_common.skc_reuseport
346 #define sk_ipv6only __sk_common.skc_ipv6only
347 #define sk_net_refcnt __sk_common.skc_net_refcnt
348 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
349 #define sk_bind_node __sk_common.skc_bind_node
350 #define sk_prot __sk_common.skc_prot
351 #define sk_net __sk_common.skc_net
352 #define sk_v6_daddr __sk_common.skc_v6_daddr
353 #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
354 #define sk_cookie __sk_common.skc_cookie
355 #define sk_incoming_cpu __sk_common.skc_incoming_cpu
356 #define sk_flags __sk_common.skc_flags
357 #define sk_rxhash __sk_common.skc_rxhash
359 socket_lock_t sk_lock
;
360 struct sk_buff_head sk_receive_queue
;
362 * The backlog queue is special, it is always used with
363 * the per-socket spinlock held and requires low latency
364 * access. Therefore we special case it's implementation.
365 * Note : rmem_alloc is in this structure to fill a hole
366 * on 64bit arches, not because its logically part of
372 struct sk_buff
*head
;
373 struct sk_buff
*tail
;
375 #define sk_rmem_alloc sk_backlog.rmem_alloc
376 int sk_forward_alloc
;
379 #ifdef CONFIG_NET_RX_BUSY_POLL
380 unsigned int sk_napi_id
;
381 unsigned int sk_ll_usec
;
386 struct sk_filter __rcu
*sk_filter
;
388 struct socket_wq __rcu
*sk_wq
;
389 struct socket_wq
*sk_wq_raw
;
392 struct xfrm_policy
*sk_policy
[2];
394 struct dst_entry
*sk_rx_dst
;
395 struct dst_entry __rcu
*sk_dst_cache
;
396 /* Note: 32bit hole on 64bit arches */
397 atomic_t sk_wmem_alloc
;
398 atomic_t sk_omem_alloc
;
400 struct sk_buff_head sk_write_queue
;
401 kmemcheck_bitfield_begin(flags
);
402 unsigned int sk_shutdown
: 2,
408 kmemcheck_bitfield_end(flags
);
411 u32 sk_pacing_rate
; /* bytes per second */
412 u32 sk_max_pacing_rate
;
413 netdev_features_t sk_route_caps
;
414 netdev_features_t sk_route_nocaps
;
416 unsigned int sk_gso_max_size
;
419 unsigned long sk_lingertime
;
420 struct sk_buff_head sk_error_queue
;
421 struct proto
*sk_prot_creator
;
422 rwlock_t sk_callback_lock
;
426 u32 sk_max_ack_backlog
;
429 struct pid
*sk_peer_pid
;
430 const struct cred
*sk_peer_cred
;
433 struct timer_list sk_timer
;
437 struct socket
*sk_socket
;
439 struct page_frag sk_frag
;
440 struct sk_buff
*sk_send_head
;
442 int sk_write_pending
;
443 #ifdef CONFIG_SECURITY
446 struct sock_cgroup_data sk_cgrp_data
;
447 struct cg_proto
*sk_cgrp
;
448 void (*sk_state_change
)(struct sock
*sk
);
449 void (*sk_data_ready
)(struct sock
*sk
);
450 void (*sk_write_space
)(struct sock
*sk
);
451 void (*sk_error_report
)(struct sock
*sk
);
452 int (*sk_backlog_rcv
)(struct sock
*sk
,
453 struct sk_buff
*skb
);
454 void (*sk_destruct
)(struct sock
*sk
);
457 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
459 #define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
460 #define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
463 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
464 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
465 * on a socket means that the socket will reuse everybody else's port
466 * without looking at the other's sk_reuse value.
469 #define SK_NO_REUSE 0
470 #define SK_CAN_REUSE 1
471 #define SK_FORCE_REUSE 2
473 static inline int sk_peek_offset(struct sock
*sk
, int flags
)
475 if ((flags
& MSG_PEEK
) && (sk
->sk_peek_off
>= 0))
476 return sk
->sk_peek_off
;
481 static inline void sk_peek_offset_bwd(struct sock
*sk
, int val
)
483 if (sk
->sk_peek_off
>= 0) {
484 if (sk
->sk_peek_off
>= val
)
485 sk
->sk_peek_off
-= val
;
491 static inline void sk_peek_offset_fwd(struct sock
*sk
, int val
)
493 if (sk
->sk_peek_off
>= 0)
494 sk
->sk_peek_off
+= val
;
498 * Hashed lists helper routines
500 static inline struct sock
*sk_entry(const struct hlist_node
*node
)
502 return hlist_entry(node
, struct sock
, sk_node
);
505 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
507 return hlist_entry(head
->first
, struct sock
, sk_node
);
510 static inline struct sock
*sk_head(const struct hlist_head
*head
)
512 return hlist_empty(head
) ? NULL
: __sk_head(head
);
515 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
517 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
520 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
522 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
525 static inline struct sock
*sk_next(const struct sock
*sk
)
527 return sk
->sk_node
.next
?
528 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
531 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
533 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
534 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
535 struct sock
, sk_nulls_node
) :
539 static inline bool sk_unhashed(const struct sock
*sk
)
541 return hlist_unhashed(&sk
->sk_node
);
544 static inline bool sk_hashed(const struct sock
*sk
)
546 return !sk_unhashed(sk
);
549 static inline void sk_node_init(struct hlist_node
*node
)
554 static inline void sk_nulls_node_init(struct hlist_nulls_node
*node
)
559 static inline void __sk_del_node(struct sock
*sk
)
561 __hlist_del(&sk
->sk_node
);
564 /* NB: equivalent to hlist_del_init_rcu */
565 static inline bool __sk_del_node_init(struct sock
*sk
)
569 sk_node_init(&sk
->sk_node
);
575 /* Grab socket reference count. This operation is valid only
576 when sk is ALREADY grabbed f.e. it is found in hash table
577 or a list and the lookup is made under lock preventing hash table
581 static inline void sock_hold(struct sock
*sk
)
583 atomic_inc(&sk
->sk_refcnt
);
586 /* Ungrab socket in the context, which assumes that socket refcnt
587 cannot hit zero, f.e. it is true in context of any socketcall.
589 static inline void __sock_put(struct sock
*sk
)
591 atomic_dec(&sk
->sk_refcnt
);
594 static inline bool sk_del_node_init(struct sock
*sk
)
596 bool rc
= __sk_del_node_init(sk
);
599 /* paranoid for a while -acme */
600 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
605 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
607 static inline bool __sk_nulls_del_node_init_rcu(struct sock
*sk
)
610 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
616 static inline bool sk_nulls_del_node_init_rcu(struct sock
*sk
)
618 bool rc
= __sk_nulls_del_node_init_rcu(sk
);
621 /* paranoid for a while -acme */
622 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
628 static inline void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
630 hlist_add_head(&sk
->sk_node
, list
);
633 static inline void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
636 __sk_add_node(sk
, list
);
639 static inline void sk_add_node_rcu(struct sock
*sk
, struct hlist_head
*list
)
642 hlist_add_head_rcu(&sk
->sk_node
, list
);
645 static inline void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
647 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
650 static inline void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
653 __sk_nulls_add_node_rcu(sk
, list
);
656 static inline void __sk_del_bind_node(struct sock
*sk
)
658 __hlist_del(&sk
->sk_bind_node
);
661 static inline void sk_add_bind_node(struct sock
*sk
,
662 struct hlist_head
*list
)
664 hlist_add_head(&sk
->sk_bind_node
, list
);
667 #define sk_for_each(__sk, list) \
668 hlist_for_each_entry(__sk, list, sk_node)
669 #define sk_for_each_rcu(__sk, list) \
670 hlist_for_each_entry_rcu(__sk, list, sk_node)
671 #define sk_nulls_for_each(__sk, node, list) \
672 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
673 #define sk_nulls_for_each_rcu(__sk, node, list) \
674 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
675 #define sk_for_each_from(__sk) \
676 hlist_for_each_entry_from(__sk, sk_node)
677 #define sk_nulls_for_each_from(__sk, node) \
678 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
679 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
680 #define sk_for_each_safe(__sk, tmp, list) \
681 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
682 #define sk_for_each_bound(__sk, list) \
683 hlist_for_each_entry(__sk, list, sk_bind_node)
686 * sk_nulls_for_each_entry_offset - iterate over a list at a given struct offset
687 * @tpos: the type * to use as a loop cursor.
688 * @pos: the &struct hlist_node to use as a loop cursor.
689 * @head: the head for your list.
690 * @offset: offset of hlist_node within the struct.
693 #define sk_nulls_for_each_entry_offset(tpos, pos, head, offset) \
694 for (pos = (head)->first; \
695 (!is_a_nulls(pos)) && \
696 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
699 static inline struct user_namespace
*sk_user_ns(struct sock
*sk
)
701 /* Careful only use this in a context where these parameters
702 * can not change and must all be valid, such as recvmsg from
705 return sk
->sk_socket
->file
->f_cred
->user_ns
;
719 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
720 SOCK_DBG
, /* %SO_DEBUG setting */
721 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
722 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
723 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
724 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
725 SOCK_MEMALLOC
, /* VM depends on this socket for swapping */
726 SOCK_TIMESTAMPING_RX_SOFTWARE
, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
727 SOCK_FASYNC
, /* fasync() active */
729 SOCK_ZEROCOPY
, /* buffers from userspace */
730 SOCK_WIFI_STATUS
, /* push wifi status to userspace */
731 SOCK_NOFCS
, /* Tell NIC not to do the Ethernet FCS.
732 * Will use last 4 bytes of packet sent from
733 * user-space instead.
735 SOCK_FILTER_LOCKED
, /* Filter cannot be changed anymore */
736 SOCK_SELECT_ERR_QUEUE
, /* Wake select on error queue */
739 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
741 nsk
->sk_flags
= osk
->sk_flags
;
744 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
746 __set_bit(flag
, &sk
->sk_flags
);
749 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
751 __clear_bit(flag
, &sk
->sk_flags
);
754 static inline bool sock_flag(const struct sock
*sk
, enum sock_flags flag
)
756 return test_bit(flag
, &sk
->sk_flags
);
760 extern struct static_key memalloc_socks
;
761 static inline int sk_memalloc_socks(void)
763 return static_key_false(&memalloc_socks
);
767 static inline int sk_memalloc_socks(void)
774 static inline gfp_t
sk_gfp_mask(const struct sock
*sk
, gfp_t gfp_mask
)
776 return gfp_mask
| (sk
->sk_allocation
& __GFP_MEMALLOC
);
779 static inline void sk_acceptq_removed(struct sock
*sk
)
781 sk
->sk_ack_backlog
--;
784 static inline void sk_acceptq_added(struct sock
*sk
)
786 sk
->sk_ack_backlog
++;
789 static inline bool sk_acceptq_is_full(const struct sock
*sk
)
791 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
795 * Compute minimal free write space needed to queue new packets.
797 static inline int sk_stream_min_wspace(const struct sock
*sk
)
799 return sk
->sk_wmem_queued
>> 1;
802 static inline int sk_stream_wspace(const struct sock
*sk
)
804 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
807 void sk_stream_write_space(struct sock
*sk
);
809 /* OOB backlog add */
810 static inline void __sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
812 /* dont let skb dst not refcounted, we are going to leave rcu lock */
815 if (!sk
->sk_backlog
.tail
)
816 sk
->sk_backlog
.head
= skb
;
818 sk
->sk_backlog
.tail
->next
= skb
;
820 sk
->sk_backlog
.tail
= skb
;
825 * Take into account size of receive queue and backlog queue
826 * Do not take into account this skb truesize,
827 * to allow even a single big packet to come.
829 static inline bool sk_rcvqueues_full(const struct sock
*sk
, unsigned int limit
)
831 unsigned int qsize
= sk
->sk_backlog
.len
+ atomic_read(&sk
->sk_rmem_alloc
);
833 return qsize
> limit
;
836 /* The per-socket spinlock must be held here. */
837 static inline __must_check
int sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
,
840 if (sk_rcvqueues_full(sk
, limit
))
844 * If the skb was allocated from pfmemalloc reserves, only
845 * allow SOCK_MEMALLOC sockets to use it as this socket is
846 * helping free memory
848 if (skb_pfmemalloc(skb
) && !sock_flag(sk
, SOCK_MEMALLOC
))
851 __sk_add_backlog(sk
, skb
);
852 sk
->sk_backlog
.len
+= skb
->truesize
;
856 int __sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
858 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
860 if (sk_memalloc_socks() && skb_pfmemalloc(skb
))
861 return __sk_backlog_rcv(sk
, skb
);
863 return sk
->sk_backlog_rcv(sk
, skb
);
866 static inline void sk_incoming_cpu_update(struct sock
*sk
)
868 sk
->sk_incoming_cpu
= raw_smp_processor_id();
871 static inline void sock_rps_record_flow_hash(__u32 hash
)
874 struct rps_sock_flow_table
*sock_flow_table
;
877 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
878 rps_record_sock_flow(sock_flow_table
, hash
);
883 static inline void sock_rps_record_flow(const struct sock
*sk
)
886 sock_rps_record_flow_hash(sk
->sk_rxhash
);
890 static inline void sock_rps_save_rxhash(struct sock
*sk
,
891 const struct sk_buff
*skb
)
894 if (unlikely(sk
->sk_rxhash
!= skb
->hash
))
895 sk
->sk_rxhash
= skb
->hash
;
899 static inline void sock_rps_reset_rxhash(struct sock
*sk
)
906 #define sk_wait_event(__sk, __timeo, __condition) \
908 release_sock(__sk); \
909 __rc = __condition; \
911 *(__timeo) = schedule_timeout(*(__timeo)); \
913 sched_annotate_sleep(); \
915 __rc = __condition; \
919 int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
920 int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
921 void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
922 int sk_stream_error(struct sock
*sk
, int flags
, int err
);
923 void sk_stream_kill_queues(struct sock
*sk
);
924 void sk_set_memalloc(struct sock
*sk
);
925 void sk_clear_memalloc(struct sock
*sk
);
927 int sk_wait_data(struct sock
*sk
, long *timeo
, const struct sk_buff
*skb
);
929 struct request_sock_ops
;
930 struct timewait_sock_ops
;
931 struct inet_hashinfo
;
936 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
937 * un-modified. Special care is taken when initializing object to zero.
939 static inline void sk_prot_clear_nulls(struct sock
*sk
, int size
)
941 if (offsetof(struct sock
, sk_node
.next
) != 0)
942 memset(sk
, 0, offsetof(struct sock
, sk_node
.next
));
943 memset(&sk
->sk_node
.pprev
, 0,
944 size
- offsetof(struct sock
, sk_node
.pprev
));
947 /* Networking protocol blocks we attach to sockets.
948 * socket layer -> transport layer interface
951 void (*close
)(struct sock
*sk
,
953 int (*connect
)(struct sock
*sk
,
954 struct sockaddr
*uaddr
,
956 int (*disconnect
)(struct sock
*sk
, int flags
);
958 struct sock
* (*accept
)(struct sock
*sk
, int flags
, int *err
);
960 int (*ioctl
)(struct sock
*sk
, int cmd
,
962 int (*init
)(struct sock
*sk
);
963 void (*destroy
)(struct sock
*sk
);
964 void (*shutdown
)(struct sock
*sk
, int how
);
965 int (*setsockopt
)(struct sock
*sk
, int level
,
966 int optname
, char __user
*optval
,
967 unsigned int optlen
);
968 int (*getsockopt
)(struct sock
*sk
, int level
,
969 int optname
, char __user
*optval
,
972 int (*compat_setsockopt
)(struct sock
*sk
,
974 int optname
, char __user
*optval
,
975 unsigned int optlen
);
976 int (*compat_getsockopt
)(struct sock
*sk
,
978 int optname
, char __user
*optval
,
980 int (*compat_ioctl
)(struct sock
*sk
,
981 unsigned int cmd
, unsigned long arg
);
983 int (*sendmsg
)(struct sock
*sk
, struct msghdr
*msg
,
985 int (*recvmsg
)(struct sock
*sk
, struct msghdr
*msg
,
986 size_t len
, int noblock
, int flags
,
988 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
989 int offset
, size_t size
, int flags
);
990 int (*bind
)(struct sock
*sk
,
991 struct sockaddr
*uaddr
, int addr_len
);
993 int (*backlog_rcv
) (struct sock
*sk
,
994 struct sk_buff
*skb
);
996 void (*release_cb
)(struct sock
*sk
);
998 /* Keeping track of sk's, looking them up, and port selection methods. */
999 void (*hash
)(struct sock
*sk
);
1000 void (*unhash
)(struct sock
*sk
);
1001 void (*rehash
)(struct sock
*sk
);
1002 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
1003 void (*clear_sk
)(struct sock
*sk
, int size
);
1005 /* Keeping track of sockets in use */
1006 #ifdef CONFIG_PROC_FS
1007 unsigned int inuse_idx
;
1010 bool (*stream_memory_free
)(const struct sock
*sk
);
1011 /* Memory pressure */
1012 void (*enter_memory_pressure
)(struct sock
*sk
);
1013 atomic_long_t
*memory_allocated
; /* Current allocated memory. */
1014 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
1016 * Pressure flag: try to collapse.
1017 * Technical note: it is used by multiple contexts non atomically.
1018 * All the __sk_mem_schedule() is of this nature: accounting
1019 * is strict, actions are advisory and have some latency.
1021 int *memory_pressure
;
1028 struct kmem_cache
*slab
;
1029 unsigned int obj_size
;
1032 struct percpu_counter
*orphan_count
;
1034 struct request_sock_ops
*rsk_prot
;
1035 struct timewait_sock_ops
*twsk_prot
;
1038 struct inet_hashinfo
*hashinfo
;
1039 struct udp_table
*udp_table
;
1040 struct raw_hashinfo
*raw_hash
;
1043 struct module
*owner
;
1047 struct list_head node
;
1048 #ifdef SOCK_REFCNT_DEBUG
1051 #ifdef CONFIG_MEMCG_KMEM
1053 * cgroup specific init/deinit functions. Called once for all
1054 * protocols that implement it, from cgroups populate function.
1055 * This function has to setup any files the protocol want to
1056 * appear in the kmem cgroup filesystem.
1058 int (*init_cgroup
)(struct mem_cgroup
*memcg
,
1059 struct cgroup_subsys
*ss
);
1060 void (*destroy_cgroup
)(struct mem_cgroup
*memcg
);
1061 struct cg_proto
*(*proto_cgroup
)(struct mem_cgroup
*memcg
);
1065 int proto_register(struct proto
*prot
, int alloc_slab
);
1066 void proto_unregister(struct proto
*prot
);
1068 #ifdef SOCK_REFCNT_DEBUG
1069 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
1071 atomic_inc(&sk
->sk_prot
->socks
);
1074 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
1076 atomic_dec(&sk
->sk_prot
->socks
);
1077 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
1078 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
1081 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
1083 if (atomic_read(&sk
->sk_refcnt
) != 1)
1084 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
1085 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
1087 #else /* SOCK_REFCNT_DEBUG */
1088 #define sk_refcnt_debug_inc(sk) do { } while (0)
1089 #define sk_refcnt_debug_dec(sk) do { } while (0)
1090 #define sk_refcnt_debug_release(sk) do { } while (0)
1091 #endif /* SOCK_REFCNT_DEBUG */
1093 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
1094 extern struct static_key memcg_socket_limit_enabled
;
1095 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
1096 struct cg_proto
*cg_proto
)
1098 return proto
->proto_cgroup(parent_mem_cgroup(cg_proto
->memcg
));
1100 #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
1102 #define mem_cgroup_sockets_enabled 0
1103 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
1104 struct cg_proto
*cg_proto
)
1110 static inline bool sk_stream_memory_free(const struct sock
*sk
)
1112 if (sk
->sk_wmem_queued
>= sk
->sk_sndbuf
)
1115 return sk
->sk_prot
->stream_memory_free
?
1116 sk
->sk_prot
->stream_memory_free(sk
) : true;
1119 static inline bool sk_stream_is_writeable(const struct sock
*sk
)
1121 return sk_stream_wspace(sk
) >= sk_stream_min_wspace(sk
) &&
1122 sk_stream_memory_free(sk
);
1126 static inline bool sk_has_memory_pressure(const struct sock
*sk
)
1128 return sk
->sk_prot
->memory_pressure
!= NULL
;
1131 static inline bool sk_under_memory_pressure(const struct sock
*sk
)
1133 if (!sk
->sk_prot
->memory_pressure
)
1136 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1137 return !!sk
->sk_cgrp
->memory_pressure
;
1139 return !!*sk
->sk_prot
->memory_pressure
;
1142 static inline void sk_leave_memory_pressure(struct sock
*sk
)
1144 int *memory_pressure
= sk
->sk_prot
->memory_pressure
;
1146 if (!memory_pressure
)
1149 if (*memory_pressure
)
1150 *memory_pressure
= 0;
1152 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1153 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1154 struct proto
*prot
= sk
->sk_prot
;
1156 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1157 cg_proto
->memory_pressure
= 0;
1162 static inline void sk_enter_memory_pressure(struct sock
*sk
)
1164 if (!sk
->sk_prot
->enter_memory_pressure
)
1167 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1168 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1169 struct proto
*prot
= sk
->sk_prot
;
1171 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1172 cg_proto
->memory_pressure
= 1;
1175 sk
->sk_prot
->enter_memory_pressure(sk
);
1178 static inline long sk_prot_mem_limits(const struct sock
*sk
, int index
)
1180 long *prot
= sk
->sk_prot
->sysctl_mem
;
1181 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1182 prot
= sk
->sk_cgrp
->sysctl_mem
;
1186 static inline void memcg_memory_allocated_add(struct cg_proto
*prot
,
1190 page_counter_charge(&prot
->memory_allocated
, amt
);
1192 if (page_counter_read(&prot
->memory_allocated
) >
1193 prot
->memory_allocated
.limit
)
1194 *parent_status
= OVER_LIMIT
;
1197 static inline void memcg_memory_allocated_sub(struct cg_proto
*prot
,
1200 page_counter_uncharge(&prot
->memory_allocated
, amt
);
1204 sk_memory_allocated(const struct sock
*sk
)
1206 struct proto
*prot
= sk
->sk_prot
;
1208 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1209 return page_counter_read(&sk
->sk_cgrp
->memory_allocated
);
1211 return atomic_long_read(prot
->memory_allocated
);
1215 sk_memory_allocated_add(struct sock
*sk
, int amt
, int *parent_status
)
1217 struct proto
*prot
= sk
->sk_prot
;
1219 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1220 memcg_memory_allocated_add(sk
->sk_cgrp
, amt
, parent_status
);
1221 /* update the root cgroup regardless */
1222 atomic_long_add_return(amt
, prot
->memory_allocated
);
1223 return page_counter_read(&sk
->sk_cgrp
->memory_allocated
);
1226 return atomic_long_add_return(amt
, prot
->memory_allocated
);
1230 sk_memory_allocated_sub(struct sock
*sk
, int amt
)
1232 struct proto
*prot
= sk
->sk_prot
;
1234 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1235 memcg_memory_allocated_sub(sk
->sk_cgrp
, amt
);
1237 atomic_long_sub(amt
, prot
->memory_allocated
);
1240 static inline void sk_sockets_allocated_dec(struct sock
*sk
)
1242 struct proto
*prot
= sk
->sk_prot
;
1244 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1245 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1247 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1248 percpu_counter_dec(&cg_proto
->sockets_allocated
);
1251 percpu_counter_dec(prot
->sockets_allocated
);
1254 static inline void sk_sockets_allocated_inc(struct sock
*sk
)
1256 struct proto
*prot
= sk
->sk_prot
;
1258 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1259 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1261 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1262 percpu_counter_inc(&cg_proto
->sockets_allocated
);
1265 percpu_counter_inc(prot
->sockets_allocated
);
1269 sk_sockets_allocated_read_positive(struct sock
*sk
)
1271 struct proto
*prot
= sk
->sk_prot
;
1273 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1274 return percpu_counter_read_positive(&sk
->sk_cgrp
->sockets_allocated
);
1276 return percpu_counter_read_positive(prot
->sockets_allocated
);
1280 proto_sockets_allocated_sum_positive(struct proto
*prot
)
1282 return percpu_counter_sum_positive(prot
->sockets_allocated
);
1286 proto_memory_allocated(struct proto
*prot
)
1288 return atomic_long_read(prot
->memory_allocated
);
1292 proto_memory_pressure(struct proto
*prot
)
1294 if (!prot
->memory_pressure
)
1296 return !!*prot
->memory_pressure
;
1300 #ifdef CONFIG_PROC_FS
1301 /* Called with local bh disabled */
1302 void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int inc
);
1303 int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
1305 static inline void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
1312 /* With per-bucket locks this operation is not-atomic, so that
1313 * this version is not worse.
1315 static inline void __sk_prot_rehash(struct sock
*sk
)
1317 sk
->sk_prot
->unhash(sk
);
1318 sk
->sk_prot
->hash(sk
);
1321 void sk_prot_clear_portaddr_nulls(struct sock
*sk
, int size
);
1323 /* About 10 seconds */
1324 #define SOCK_DESTROY_TIME (10*HZ)
1326 /* Sockets 0-1023 can't be bound to unless you are superuser */
1327 #define PROT_SOCK 1024
1329 #define SHUTDOWN_MASK 3
1330 #define RCV_SHUTDOWN 1
1331 #define SEND_SHUTDOWN 2
1333 #define SOCK_SNDBUF_LOCK 1
1334 #define SOCK_RCVBUF_LOCK 2
1335 #define SOCK_BINDADDR_LOCK 4
1336 #define SOCK_BINDPORT_LOCK 8
1338 struct socket_alloc
{
1339 struct socket socket
;
1340 struct inode vfs_inode
;
1343 static inline struct socket
*SOCKET_I(struct inode
*inode
)
1345 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
1348 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
1350 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
1354 * Functions for memory accounting
1356 int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
1357 void __sk_mem_reclaim(struct sock
*sk
, int amount
);
1359 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1360 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1361 #define SK_MEM_SEND 0
1362 #define SK_MEM_RECV 1
1364 static inline int sk_mem_pages(int amt
)
1366 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
1369 static inline bool sk_has_account(struct sock
*sk
)
1371 /* return true if protocol supports memory accounting */
1372 return !!sk
->sk_prot
->memory_allocated
;
1375 static inline bool sk_wmem_schedule(struct sock
*sk
, int size
)
1377 if (!sk_has_account(sk
))
1379 return size
<= sk
->sk_forward_alloc
||
1380 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
1384 sk_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
, int size
)
1386 if (!sk_has_account(sk
))
1388 return size
<= sk
->sk_forward_alloc
||
1389 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
) ||
1390 skb_pfmemalloc(skb
);
1393 static inline void sk_mem_reclaim(struct sock
*sk
)
1395 if (!sk_has_account(sk
))
1397 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
1398 __sk_mem_reclaim(sk
, sk
->sk_forward_alloc
);
1401 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
1403 if (!sk_has_account(sk
))
1405 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
1406 __sk_mem_reclaim(sk
, sk
->sk_forward_alloc
- 1);
1409 static inline void sk_mem_charge(struct sock
*sk
, int size
)
1411 if (!sk_has_account(sk
))
1413 sk
->sk_forward_alloc
-= size
;
1416 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
1418 if (!sk_has_account(sk
))
1420 sk
->sk_forward_alloc
+= size
;
1423 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
1425 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
1426 sk
->sk_wmem_queued
-= skb
->truesize
;
1427 sk_mem_uncharge(sk
, skb
->truesize
);
1431 /* Used by processes to "lock" a socket state, so that
1432 * interrupts and bottom half handlers won't change it
1433 * from under us. It essentially blocks any incoming
1434 * packets, so that we won't get any new data or any
1435 * packets that change the state of the socket.
1437 * While locked, BH processing will add new packets to
1438 * the backlog queue. This queue is processed by the
1439 * owner of the socket lock right before it is released.
1441 * Since ~2.3.5 it is also exclusive sleep lock serializing
1442 * accesses from user process context.
1444 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1446 static inline void sock_release_ownership(struct sock
*sk
)
1448 sk
->sk_lock
.owned
= 0;
1452 * Macro so as to not evaluate some arguments when
1453 * lockdep is not enabled.
1455 * Mark both the sk_lock and the sk_lock.slock as a
1456 * per-address-family lock class.
1458 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1460 sk->sk_lock.owned = 0; \
1461 init_waitqueue_head(&sk->sk_lock.wq); \
1462 spin_lock_init(&(sk)->sk_lock.slock); \
1463 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1464 sizeof((sk)->sk_lock)); \
1465 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1467 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1470 void lock_sock_nested(struct sock
*sk
, int subclass
);
1472 static inline void lock_sock(struct sock
*sk
)
1474 lock_sock_nested(sk
, 0);
1477 void release_sock(struct sock
*sk
);
1479 /* BH context may only use the following locking interface. */
1480 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1481 #define bh_lock_sock_nested(__sk) \
1482 spin_lock_nested(&((__sk)->sk_lock.slock), \
1483 SINGLE_DEPTH_NESTING)
1484 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1486 bool lock_sock_fast(struct sock
*sk
);
1488 * unlock_sock_fast - complement of lock_sock_fast
1492 * fast unlock socket for user context.
1493 * If slow mode is on, we call regular release_sock()
1495 static inline void unlock_sock_fast(struct sock
*sk
, bool slow
)
1500 spin_unlock_bh(&sk
->sk_lock
.slock
);
1504 struct sock
*sk_alloc(struct net
*net
, int family
, gfp_t priority
,
1505 struct proto
*prot
, int kern
);
1506 void sk_free(struct sock
*sk
);
1507 void sk_destruct(struct sock
*sk
);
1508 struct sock
*sk_clone_lock(const struct sock
*sk
, const gfp_t priority
);
1510 struct sk_buff
*sock_wmalloc(struct sock
*sk
, unsigned long size
, int force
,
1512 void sock_wfree(struct sk_buff
*skb
);
1513 void skb_orphan_partial(struct sk_buff
*skb
);
1514 void sock_rfree(struct sk_buff
*skb
);
1515 void sock_efree(struct sk_buff
*skb
);
1517 void sock_edemux(struct sk_buff
*skb
);
1519 #define sock_edemux(skb) sock_efree(skb)
1522 int sock_setsockopt(struct socket
*sock
, int level
, int op
,
1523 char __user
*optval
, unsigned int optlen
);
1525 int sock_getsockopt(struct socket
*sock
, int level
, int op
,
1526 char __user
*optval
, int __user
*optlen
);
1527 struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
, unsigned long size
,
1528 int noblock
, int *errcode
);
1529 struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
, unsigned long header_len
,
1530 unsigned long data_len
, int noblock
,
1531 int *errcode
, int max_page_order
);
1532 void *sock_kmalloc(struct sock
*sk
, int size
, gfp_t priority
);
1533 void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
1534 void sock_kzfree_s(struct sock
*sk
, void *mem
, int size
);
1535 void sk_send_sigurg(struct sock
*sk
);
1537 struct sockcm_cookie
{
1541 int sock_cmsg_send(struct sock
*sk
, struct msghdr
*msg
,
1542 struct sockcm_cookie
*sockc
);
1545 * Functions to fill in entries in struct proto_ops when a protocol
1546 * does not implement a particular function.
1548 int sock_no_bind(struct socket
*, struct sockaddr
*, int);
1549 int sock_no_connect(struct socket
*, struct sockaddr
*, int, int);
1550 int sock_no_socketpair(struct socket
*, struct socket
*);
1551 int sock_no_accept(struct socket
*, struct socket
*, int);
1552 int sock_no_getname(struct socket
*, struct sockaddr
*, int *, int);
1553 unsigned int sock_no_poll(struct file
*, struct socket
*,
1554 struct poll_table_struct
*);
1555 int sock_no_ioctl(struct socket
*, unsigned int, unsigned long);
1556 int sock_no_listen(struct socket
*, int);
1557 int sock_no_shutdown(struct socket
*, int);
1558 int sock_no_getsockopt(struct socket
*, int , int, char __user
*, int __user
*);
1559 int sock_no_setsockopt(struct socket
*, int, int, char __user
*, unsigned int);
1560 int sock_no_sendmsg(struct socket
*, struct msghdr
*, size_t);
1561 int sock_no_recvmsg(struct socket
*, struct msghdr
*, size_t, int);
1562 int sock_no_mmap(struct file
*file
, struct socket
*sock
,
1563 struct vm_area_struct
*vma
);
1564 ssize_t
sock_no_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
1565 size_t size
, int flags
);
1568 * Functions to fill in entries in struct proto_ops when a protocol
1569 * uses the inet style.
1571 int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1572 char __user
*optval
, int __user
*optlen
);
1573 int sock_common_recvmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
,
1575 int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1576 char __user
*optval
, unsigned int optlen
);
1577 int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
1578 int optname
, char __user
*optval
, int __user
*optlen
);
1579 int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
1580 int optname
, char __user
*optval
, unsigned int optlen
);
1582 void sk_common_release(struct sock
*sk
);
1585 * Default socket callbacks and setup code
1588 /* Initialise core socket variables */
1589 void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1592 * Socket reference counting postulates.
1594 * * Each user of socket SHOULD hold a reference count.
1595 * * Each access point to socket (an hash table bucket, reference from a list,
1596 * running timer, skb in flight MUST hold a reference count.
1597 * * When reference count hits 0, it means it will never increase back.
1598 * * When reference count hits 0, it means that no references from
1599 * outside exist to this socket and current process on current CPU
1600 * is last user and may/should destroy this socket.
1601 * * sk_free is called from any context: process, BH, IRQ. When
1602 * it is called, socket has no references from outside -> sk_free
1603 * may release descendant resources allocated by the socket, but
1604 * to the time when it is called, socket is NOT referenced by any
1605 * hash tables, lists etc.
1606 * * Packets, delivered from outside (from network or from another process)
1607 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1608 * when they sit in queue. Otherwise, packets will leak to hole, when
1609 * socket is looked up by one cpu and unhasing is made by another CPU.
1610 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1611 * (leak to backlog). Packet socket does all the processing inside
1612 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1613 * use separate SMP lock, so that they are prone too.
1616 /* Ungrab socket and destroy it, if it was the last reference. */
1617 static inline void sock_put(struct sock
*sk
)
1619 if (atomic_dec_and_test(&sk
->sk_refcnt
))
1622 /* Generic version of sock_put(), dealing with all sockets
1623 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
1625 void sock_gen_put(struct sock
*sk
);
1627 int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
, const int nested
);
1629 static inline void sk_tx_queue_set(struct sock
*sk
, int tx_queue
)
1631 sk
->sk_tx_queue_mapping
= tx_queue
;
1634 static inline void sk_tx_queue_clear(struct sock
*sk
)
1636 sk
->sk_tx_queue_mapping
= -1;
1639 static inline int sk_tx_queue_get(const struct sock
*sk
)
1641 return sk
? sk
->sk_tx_queue_mapping
: -1;
1644 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1646 sk_tx_queue_clear(sk
);
1647 sk
->sk_socket
= sock
;
1650 static inline wait_queue_head_t
*sk_sleep(struct sock
*sk
)
1652 BUILD_BUG_ON(offsetof(struct socket_wq
, wait
) != 0);
1653 return &rcu_dereference_raw(sk
->sk_wq
)->wait
;
1655 /* Detach socket from process context.
1656 * Announce socket dead, detach it from wait queue and inode.
1657 * Note that parent inode held reference count on this struct sock,
1658 * we do not release it in this function, because protocol
1659 * probably wants some additional cleanups or even continuing
1660 * to work with this socket (TCP).
1662 static inline void sock_orphan(struct sock
*sk
)
1664 write_lock_bh(&sk
->sk_callback_lock
);
1665 sock_set_flag(sk
, SOCK_DEAD
);
1666 sk_set_socket(sk
, NULL
);
1668 write_unlock_bh(&sk
->sk_callback_lock
);
1671 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1673 write_lock_bh(&sk
->sk_callback_lock
);
1674 sk
->sk_wq
= parent
->wq
;
1676 sk_set_socket(sk
, parent
);
1677 security_sock_graft(sk
, parent
);
1678 write_unlock_bh(&sk
->sk_callback_lock
);
1681 kuid_t
sock_i_uid(struct sock
*sk
);
1682 unsigned long sock_i_ino(struct sock
*sk
);
1684 static inline u32
net_tx_rndhash(void)
1686 u32 v
= prandom_u32();
1691 static inline void sk_set_txhash(struct sock
*sk
)
1693 sk
->sk_txhash
= net_tx_rndhash();
1696 static inline void sk_rethink_txhash(struct sock
*sk
)
1702 static inline struct dst_entry
*
1703 __sk_dst_get(struct sock
*sk
)
1705 return rcu_dereference_check(sk
->sk_dst_cache
, sock_owned_by_user(sk
) ||
1706 lockdep_is_held(&sk
->sk_lock
.slock
));
1709 static inline struct dst_entry
*
1710 sk_dst_get(struct sock
*sk
)
1712 struct dst_entry
*dst
;
1715 dst
= rcu_dereference(sk
->sk_dst_cache
);
1716 if (dst
&& !atomic_inc_not_zero(&dst
->__refcnt
))
1722 static inline void dst_negative_advice(struct sock
*sk
)
1724 struct dst_entry
*ndst
, *dst
= __sk_dst_get(sk
);
1726 sk_rethink_txhash(sk
);
1728 if (dst
&& dst
->ops
->negative_advice
) {
1729 ndst
= dst
->ops
->negative_advice(dst
);
1732 rcu_assign_pointer(sk
->sk_dst_cache
, ndst
);
1733 sk_tx_queue_clear(sk
);
1739 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1741 struct dst_entry
*old_dst
;
1743 sk_tx_queue_clear(sk
);
1745 * This can be called while sk is owned by the caller only,
1746 * with no state that can be checked in a rcu_dereference_check() cond
1748 old_dst
= rcu_dereference_raw(sk
->sk_dst_cache
);
1749 rcu_assign_pointer(sk
->sk_dst_cache
, dst
);
1750 dst_release(old_dst
);
1754 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1756 struct dst_entry
*old_dst
;
1758 sk_tx_queue_clear(sk
);
1759 old_dst
= xchg((__force
struct dst_entry
**)&sk
->sk_dst_cache
, dst
);
1760 dst_release(old_dst
);
1764 __sk_dst_reset(struct sock
*sk
)
1766 __sk_dst_set(sk
, NULL
);
1770 sk_dst_reset(struct sock
*sk
)
1772 sk_dst_set(sk
, NULL
);
1775 struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1777 struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1779 bool sk_mc_loop(struct sock
*sk
);
1781 static inline bool sk_can_gso(const struct sock
*sk
)
1783 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1786 void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1788 static inline void sk_nocaps_add(struct sock
*sk
, netdev_features_t flags
)
1790 sk
->sk_route_nocaps
|= flags
;
1791 sk
->sk_route_caps
&= ~flags
;
1794 static inline int skb_do_copy_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1795 struct iov_iter
*from
, char *to
,
1796 int copy
, int offset
)
1798 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1800 if (csum_and_copy_from_iter(to
, copy
, &csum
, from
) != copy
)
1802 skb
->csum
= csum_block_add(skb
->csum
, csum
, offset
);
1803 } else if (sk
->sk_route_caps
& NETIF_F_NOCACHE_COPY
) {
1804 if (copy_from_iter_nocache(to
, copy
, from
) != copy
)
1806 } else if (copy_from_iter(to
, copy
, from
) != copy
)
1812 static inline int skb_add_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1813 struct iov_iter
*from
, int copy
)
1815 int err
, offset
= skb
->len
;
1817 err
= skb_do_copy_data_nocache(sk
, skb
, from
, skb_put(skb
, copy
),
1820 __skb_trim(skb
, offset
);
1825 static inline int skb_copy_to_page_nocache(struct sock
*sk
, struct iov_iter
*from
,
1826 struct sk_buff
*skb
,
1832 err
= skb_do_copy_data_nocache(sk
, skb
, from
, page_address(page
) + off
,
1838 skb
->data_len
+= copy
;
1839 skb
->truesize
+= copy
;
1840 sk
->sk_wmem_queued
+= copy
;
1841 sk_mem_charge(sk
, copy
);
1846 * sk_wmem_alloc_get - returns write allocations
1849 * Returns sk_wmem_alloc minus initial offset of one
1851 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
1853 return atomic_read(&sk
->sk_wmem_alloc
) - 1;
1857 * sk_rmem_alloc_get - returns read allocations
1860 * Returns sk_rmem_alloc
1862 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
1864 return atomic_read(&sk
->sk_rmem_alloc
);
1868 * sk_has_allocations - check if allocations are outstanding
1871 * Returns true if socket has write or read allocations
1873 static inline bool sk_has_allocations(const struct sock
*sk
)
1875 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
1879 * skwq_has_sleeper - check if there are any waiting processes
1880 * @wq: struct socket_wq
1882 * Returns true if socket_wq has waiting processes
1884 * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
1885 * barrier call. They were added due to the race found within the tcp code.
1887 * Consider following tcp code paths:
1891 * sys_select receive packet
1893 * __add_wait_queue update tp->rcv_nxt
1895 * tp->rcv_nxt check sock_def_readable
1897 * schedule rcu_read_lock();
1898 * wq = rcu_dereference(sk->sk_wq);
1899 * if (wq && waitqueue_active(&wq->wait))
1900 * wake_up_interruptible(&wq->wait)
1904 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1905 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1906 * could then endup calling schedule and sleep forever if there are no more
1907 * data on the socket.
1910 static inline bool skwq_has_sleeper(struct socket_wq
*wq
)
1912 return wq
&& wq_has_sleeper(&wq
->wait
);
1916 * sock_poll_wait - place memory barrier behind the poll_wait call.
1918 * @wait_address: socket wait queue
1921 * See the comments in the wq_has_sleeper function.
1923 static inline void sock_poll_wait(struct file
*filp
,
1924 wait_queue_head_t
*wait_address
, poll_table
*p
)
1926 if (!poll_does_not_wait(p
) && wait_address
) {
1927 poll_wait(filp
, wait_address
, p
);
1928 /* We need to be sure we are in sync with the
1929 * socket flags modification.
1931 * This memory barrier is paired in the wq_has_sleeper.
1937 static inline void skb_set_hash_from_sk(struct sk_buff
*skb
, struct sock
*sk
)
1939 if (sk
->sk_txhash
) {
1941 skb
->hash
= sk
->sk_txhash
;
1945 void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
);
1948 * Queue a received datagram if it will fit. Stream and sequenced
1949 * protocols can't normally use this as they need to fit buffers in
1950 * and play with them.
1952 * Inlined as it's very short and called for pretty much every
1953 * packet ever received.
1955 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
1959 skb
->destructor
= sock_rfree
;
1960 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
1961 sk_mem_charge(sk
, skb
->truesize
);
1964 void sk_reset_timer(struct sock
*sk
, struct timer_list
*timer
,
1965 unsigned long expires
);
1967 void sk_stop_timer(struct sock
*sk
, struct timer_list
*timer
);
1969 int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
1971 int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
);
1972 struct sk_buff
*sock_dequeue_err_skb(struct sock
*sk
);
1975 * Recover an error report and clear atomically
1978 static inline int sock_error(struct sock
*sk
)
1981 if (likely(!sk
->sk_err
))
1983 err
= xchg(&sk
->sk_err
, 0);
1987 static inline unsigned long sock_wspace(struct sock
*sk
)
1991 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
1992 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
2000 * We use sk->sk_wq_raw, from contexts knowing this
2001 * pointer is not NULL and cannot disappear/change.
2003 static inline void sk_set_bit(int nr
, struct sock
*sk
)
2005 set_bit(nr
, &sk
->sk_wq_raw
->flags
);
2008 static inline void sk_clear_bit(int nr
, struct sock
*sk
)
2010 clear_bit(nr
, &sk
->sk_wq_raw
->flags
);
2013 static inline void sk_wake_async(const struct sock
*sk
, int how
, int band
)
2015 if (sock_flag(sk
, SOCK_FASYNC
)) {
2017 sock_wake_async(rcu_dereference(sk
->sk_wq
), how
, band
);
2022 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2023 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2024 * Note: for send buffers, TCP works better if we can build two skbs at
2027 #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2029 #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2030 #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
2032 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
2034 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
2035 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
2036 sk
->sk_sndbuf
= max_t(u32
, sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
2040 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
,
2041 bool force_schedule
);
2044 * sk_page_frag - return an appropriate page_frag
2047 * If socket allocation mode allows current thread to sleep, it means its
2048 * safe to use the per task page_frag instead of the per socket one.
2050 static inline struct page_frag
*sk_page_frag(struct sock
*sk
)
2052 if (gfpflags_allow_blocking(sk
->sk_allocation
))
2053 return ¤t
->task_frag
;
2055 return &sk
->sk_frag
;
2058 bool sk_page_frag_refill(struct sock
*sk
, struct page_frag
*pfrag
);
2061 * Default write policy as shown to user space via poll/select/SIGIO
2063 static inline bool sock_writeable(const struct sock
*sk
)
2065 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
2068 static inline gfp_t
gfp_any(void)
2070 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
2073 static inline long sock_rcvtimeo(const struct sock
*sk
, bool noblock
)
2075 return noblock
? 0 : sk
->sk_rcvtimeo
;
2078 static inline long sock_sndtimeo(const struct sock
*sk
, bool noblock
)
2080 return noblock
? 0 : sk
->sk_sndtimeo
;
2083 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
2085 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
2088 /* Alas, with timeout socket operations are not restartable.
2089 * Compare this to poll().
2091 static inline int sock_intr_errno(long timeo
)
2093 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
2096 struct sock_skb_cb
{
2100 /* Store sock_skb_cb at the end of skb->cb[] so protocol families
2101 * using skb->cb[] would keep using it directly and utilize its
2102 * alignement guarantee.
2104 #define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \
2105 sizeof(struct sock_skb_cb)))
2107 #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2108 SOCK_SKB_CB_OFFSET))
2110 #define sock_skb_cb_check_size(size) \
2111 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
2114 sock_skb_set_dropcount(const struct sock
*sk
, struct sk_buff
*skb
)
2116 SOCK_SKB_CB(skb
)->dropcount
= atomic_read(&sk
->sk_drops
);
2119 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
2120 struct sk_buff
*skb
);
2121 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
2122 struct sk_buff
*skb
);
2125 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
2127 ktime_t kt
= skb
->tstamp
;
2128 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
2131 * generate control messages if
2132 * - receive time stamping in software requested
2133 * - software time stamp available and wanted
2134 * - hardware time stamps available and wanted
2136 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
2137 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RX_SOFTWARE
) ||
2138 (kt
.tv64
&& sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
) ||
2139 (hwtstamps
->hwtstamp
.tv64
&&
2140 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
)))
2141 __sock_recv_timestamp(msg
, sk
, skb
);
2145 if (sock_flag(sk
, SOCK_WIFI_STATUS
) && skb
->wifi_acked_valid
)
2146 __sock_recv_wifi_status(msg
, sk
, skb
);
2149 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2150 struct sk_buff
*skb
);
2152 static inline void sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2153 struct sk_buff
*skb
)
2155 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2156 (1UL << SOCK_RCVTSTAMP))
2157 #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2158 SOF_TIMESTAMPING_RAW_HARDWARE)
2160 if (sk
->sk_flags
& FLAGS_TS_OR_DROPS
|| sk
->sk_tsflags
& TSFLAGS_ANY
)
2161 __sock_recv_ts_and_drops(msg
, sk
, skb
);
2163 sk
->sk_stamp
= skb
->tstamp
;
2166 void __sock_tx_timestamp(const struct sock
*sk
, __u8
*tx_flags
);
2169 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2170 * @sk: socket sending this packet
2171 * @tx_flags: completed with instructions for time stamping
2173 * Note : callers should take care of initial *tx_flags value (usually 0)
2175 static inline void sock_tx_timestamp(const struct sock
*sk
, __u8
*tx_flags
)
2177 if (unlikely(sk
->sk_tsflags
))
2178 __sock_tx_timestamp(sk
, tx_flags
);
2179 if (unlikely(sock_flag(sk
, SOCK_WIFI_STATUS
)))
2180 *tx_flags
|= SKBTX_WIFI_STATUS
;
2184 * sk_eat_skb - Release a skb if it is no longer needed
2185 * @sk: socket to eat this skb from
2186 * @skb: socket buffer to eat
2188 * This routine must be called with interrupts disabled or with the socket
2189 * locked so that the sk_buff queue operation is ok.
2191 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
)
2193 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2198 struct net
*sock_net(const struct sock
*sk
)
2200 return read_pnet(&sk
->sk_net
);
2204 void sock_net_set(struct sock
*sk
, struct net
*net
)
2206 write_pnet(&sk
->sk_net
, net
);
2209 static inline struct sock
*skb_steal_sock(struct sk_buff
*skb
)
2212 struct sock
*sk
= skb
->sk
;
2214 skb
->destructor
= NULL
;
2221 /* This helper checks if a socket is a full socket,
2222 * ie _not_ a timewait or request socket.
2224 static inline bool sk_fullsock(const struct sock
*sk
)
2226 return (1 << sk
->sk_state
) & ~(TCPF_TIME_WAIT
| TCPF_NEW_SYN_RECV
);
2229 /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2230 * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2232 static inline bool sk_listener(const struct sock
*sk
)
2234 return (1 << sk
->sk_state
) & (TCPF_LISTEN
| TCPF_NEW_SYN_RECV
);
2238 * sk_state_load - read sk->sk_state for lockless contexts
2239 * @sk: socket pointer
2241 * Paired with sk_state_store(). Used in places we do not hold socket lock :
2242 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
2244 static inline int sk_state_load(const struct sock
*sk
)
2246 return smp_load_acquire(&sk
->sk_state
);
2250 * sk_state_store - update sk->sk_state
2251 * @sk: socket pointer
2252 * @newstate: new state
2254 * Paired with sk_state_load(). Should be used in contexts where
2255 * state change might impact lockless readers.
2257 static inline void sk_state_store(struct sock
*sk
, int newstate
)
2259 smp_store_release(&sk
->sk_state
, newstate
);
2262 void sock_enable_timestamp(struct sock
*sk
, int flag
);
2263 int sock_get_timestamp(struct sock
*, struct timeval __user
*);
2264 int sock_get_timestampns(struct sock
*, struct timespec __user
*);
2265 int sock_recv_errqueue(struct sock
*sk
, struct msghdr
*msg
, int len
, int level
,
2268 bool sk_ns_capable(const struct sock
*sk
,
2269 struct user_namespace
*user_ns
, int cap
);
2270 bool sk_capable(const struct sock
*sk
, int cap
);
2271 bool sk_net_capable(const struct sock
*sk
, int cap
);
2273 extern __u32 sysctl_wmem_max
;
2274 extern __u32 sysctl_rmem_max
;
2276 extern int sysctl_tstamp_allow_data
;
2277 extern int sysctl_optmem_max
;
2279 extern __u32 sysctl_wmem_default
;
2280 extern __u32 sysctl_rmem_default
;
2282 #endif /* _SOCK_H */