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_flags: place holder for sk_flags
154 * %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
155 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
156 * @skc_incoming_cpu: record/match cpu processing incoming packets
157 * @skc_refcnt: reference count
159 * This is the minimal network layer representation of sockets, the header
160 * for struct sock and struct inet_timewait_sock.
163 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
164 * address on 64bit arches : cf INET_MATCH()
167 __addrpair skc_addrpair
;
170 __be32 skc_rcv_saddr
;
174 unsigned int skc_hash
;
175 __u16 skc_u16hashes
[2];
177 /* skc_dport && skc_num must be grouped as well */
179 __portpair skc_portpair
;
186 unsigned short skc_family
;
187 volatile unsigned char skc_state
;
188 unsigned char skc_reuse
:4;
189 unsigned char skc_reuseport
:1;
190 unsigned char skc_ipv6only
:1;
191 unsigned char skc_net_refcnt
:1;
192 int skc_bound_dev_if
;
194 struct hlist_node skc_bind_node
;
195 struct hlist_nulls_node skc_portaddr_node
;
197 struct proto
*skc_prot
;
198 possible_net_t skc_net
;
200 #if IS_ENABLED(CONFIG_IPV6)
201 struct in6_addr skc_v6_daddr
;
202 struct in6_addr skc_v6_rcv_saddr
;
205 atomic64_t skc_cookie
;
207 /* following fields are padding to force
208 * offset(struct sock, sk_refcnt) == 128 on 64bit arches
209 * assuming IPV6 is enabled. We use this padding differently
210 * for different kind of 'sockets'
213 unsigned long skc_flags
;
214 struct sock
*skc_listener
; /* request_sock */
215 struct inet_timewait_death_row
*skc_tw_dr
; /* inet_timewait_sock */
218 * fields between dontcopy_begin/dontcopy_end
219 * are not copied in sock_copy()
222 int skc_dontcopy_begin
[0];
225 struct hlist_node skc_node
;
226 struct hlist_nulls_node skc_nulls_node
;
228 int skc_tx_queue_mapping
;
230 int skc_incoming_cpu
;
232 u32 skc_tw_rcv_nxt
; /* struct tcp_timewait_sock */
237 int skc_dontcopy_end
[0];
240 u32 skc_window_clamp
;
241 u32 skc_tw_snd_nxt
; /* struct tcp_timewait_sock */
248 * struct sock - network layer representation of sockets
249 * @__sk_common: shared layout with inet_timewait_sock
250 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
251 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
252 * @sk_lock: synchronizer
253 * @sk_rcvbuf: size of receive buffer in bytes
254 * @sk_wq: sock wait queue and async head
255 * @sk_rx_dst: receive input route used by early demux
256 * @sk_dst_cache: destination cache
257 * @sk_policy: flow policy
258 * @sk_receive_queue: incoming packets
259 * @sk_wmem_alloc: transmit queue bytes committed
260 * @sk_write_queue: Packet sending queue
261 * @sk_omem_alloc: "o" is "option" or "other"
262 * @sk_wmem_queued: persistent queue size
263 * @sk_forward_alloc: space allocated forward
264 * @sk_napi_id: id of the last napi context to receive data for sk
265 * @sk_ll_usec: usecs to busypoll when there is no data
266 * @sk_allocation: allocation mode
267 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
268 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
269 * @sk_sndbuf: size of send buffer in bytes
270 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
271 * @sk_no_check_rx: allow zero checksum in RX packets
272 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
273 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
274 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
275 * @sk_gso_max_size: Maximum GSO segment size to build
276 * @sk_gso_max_segs: Maximum number of GSO segments
277 * @sk_lingertime: %SO_LINGER l_linger setting
278 * @sk_backlog: always used with the per-socket spinlock held
279 * @sk_callback_lock: used with the callbacks in the end of this struct
280 * @sk_error_queue: rarely used
281 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
282 * IPV6_ADDRFORM for instance)
283 * @sk_err: last error
284 * @sk_err_soft: errors that don't cause failure but are the cause of a
285 * persistent failure not just 'timed out'
286 * @sk_drops: raw/udp drops counter
287 * @sk_ack_backlog: current listen backlog
288 * @sk_max_ack_backlog: listen backlog set in listen()
289 * @sk_priority: %SO_PRIORITY setting
290 * @sk_cgrp_prioidx: socket group's priority map index
291 * @sk_type: socket type (%SOCK_STREAM, etc)
292 * @sk_protocol: which protocol this socket belongs in this network family
293 * @sk_peer_pid: &struct pid for this socket's peer
294 * @sk_peer_cred: %SO_PEERCRED setting
295 * @sk_rcvlowat: %SO_RCVLOWAT setting
296 * @sk_rcvtimeo: %SO_RCVTIMEO setting
297 * @sk_sndtimeo: %SO_SNDTIMEO setting
298 * @sk_txhash: computed flow hash for use on transmit
299 * @sk_filter: socket filtering instructions
300 * @sk_timer: sock cleanup timer
301 * @sk_stamp: time stamp of last packet received
302 * @sk_tsflags: SO_TIMESTAMPING socket options
303 * @sk_tskey: counter to disambiguate concurrent tstamp requests
304 * @sk_socket: Identd and reporting IO signals
305 * @sk_user_data: RPC layer private data
306 * @sk_frag: cached page frag
307 * @sk_peek_off: current peek_offset value
308 * @sk_send_head: front of stuff to transmit
309 * @sk_security: used by security modules
310 * @sk_mark: generic packet mark
311 * @sk_classid: this socket's cgroup classid
312 * @sk_cgrp: this socket's cgroup-specific proto data
313 * @sk_write_pending: a write to stream socket waits to start
314 * @sk_state_change: callback to indicate change in the state of the sock
315 * @sk_data_ready: callback to indicate there is data to be processed
316 * @sk_write_space: callback to indicate there is bf sending space available
317 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
318 * @sk_backlog_rcv: callback to process the backlog
319 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
323 * Now struct inet_timewait_sock also uses sock_common, so please just
324 * don't add nothing before this first member (__sk_common) --acme
326 struct sock_common __sk_common
;
327 #define sk_node __sk_common.skc_node
328 #define sk_nulls_node __sk_common.skc_nulls_node
329 #define sk_refcnt __sk_common.skc_refcnt
330 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
332 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
333 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
334 #define sk_hash __sk_common.skc_hash
335 #define sk_portpair __sk_common.skc_portpair
336 #define sk_num __sk_common.skc_num
337 #define sk_dport __sk_common.skc_dport
338 #define sk_addrpair __sk_common.skc_addrpair
339 #define sk_daddr __sk_common.skc_daddr
340 #define sk_rcv_saddr __sk_common.skc_rcv_saddr
341 #define sk_family __sk_common.skc_family
342 #define sk_state __sk_common.skc_state
343 #define sk_reuse __sk_common.skc_reuse
344 #define sk_reuseport __sk_common.skc_reuseport
345 #define sk_ipv6only __sk_common.skc_ipv6only
346 #define sk_net_refcnt __sk_common.skc_net_refcnt
347 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
348 #define sk_bind_node __sk_common.skc_bind_node
349 #define sk_prot __sk_common.skc_prot
350 #define sk_net __sk_common.skc_net
351 #define sk_v6_daddr __sk_common.skc_v6_daddr
352 #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
353 #define sk_cookie __sk_common.skc_cookie
354 #define sk_incoming_cpu __sk_common.skc_incoming_cpu
355 #define sk_flags __sk_common.skc_flags
356 #define sk_rxhash __sk_common.skc_rxhash
358 socket_lock_t sk_lock
;
359 struct sk_buff_head sk_receive_queue
;
361 * The backlog queue is special, it is always used with
362 * the per-socket spinlock held and requires low latency
363 * access. Therefore we special case it's implementation.
364 * Note : rmem_alloc is in this structure to fill a hole
365 * on 64bit arches, not because its logically part of
371 struct sk_buff
*head
;
372 struct sk_buff
*tail
;
374 #define sk_rmem_alloc sk_backlog.rmem_alloc
375 int sk_forward_alloc
;
378 #ifdef CONFIG_NET_RX_BUSY_POLL
379 unsigned int sk_napi_id
;
380 unsigned int sk_ll_usec
;
385 struct sk_filter __rcu
*sk_filter
;
387 struct socket_wq __rcu
*sk_wq
;
388 struct socket_wq
*sk_wq_raw
;
391 struct xfrm_policy __rcu
*sk_policy
[2];
393 struct dst_entry
*sk_rx_dst
;
394 struct dst_entry __rcu
*sk_dst_cache
;
395 /* Note: 32bit hole on 64bit arches */
396 atomic_t sk_wmem_alloc
;
397 atomic_t sk_omem_alloc
;
399 struct sk_buff_head sk_write_queue
;
400 kmemcheck_bitfield_begin(flags
);
401 unsigned int sk_shutdown
: 2,
407 #define SK_PROTOCOL_MAX U8_MAX
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
;
428 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
429 __u32 sk_cgrp_prioidx
;
431 struct pid
*sk_peer_pid
;
432 const struct cred
*sk_peer_cred
;
435 struct timer_list sk_timer
;
439 struct socket
*sk_socket
;
441 struct page_frag sk_frag
;
442 struct sk_buff
*sk_send_head
;
444 int sk_write_pending
;
445 #ifdef CONFIG_SECURITY
450 #ifdef CONFIG_CGROUP_NET_CLASSID
453 struct cg_proto
*sk_cgrp
;
454 /* START_OF_KNOX_NPA */
458 __be32 sk_udp_daddr_v6
[4];
459 __be32 sk_udp_saddr_v6
[4];
462 char domain_name
[255];
463 /* END_OF_KNOX_NPA */
464 void (*sk_state_change
)(struct sock
*sk
);
465 void (*sk_data_ready
)(struct sock
*sk
);
466 void (*sk_write_space
)(struct sock
*sk
);
467 void (*sk_error_report
)(struct sock
*sk
);
468 int (*sk_backlog_rcv
)(struct sock
*sk
,
469 struct sk_buff
*skb
);
470 void (*sk_destruct
)(struct sock
*sk
);
473 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
475 #define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
476 #define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
479 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
480 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
481 * on a socket means that the socket will reuse everybody else's port
482 * without looking at the other's sk_reuse value.
485 #define SK_NO_REUSE 0
486 #define SK_CAN_REUSE 1
487 #define SK_FORCE_REUSE 2
489 static inline int sk_peek_offset(struct sock
*sk
, int flags
)
491 if ((flags
& MSG_PEEK
) && (sk
->sk_peek_off
>= 0))
492 return sk
->sk_peek_off
;
497 static inline void sk_peek_offset_bwd(struct sock
*sk
, int val
)
499 if (sk
->sk_peek_off
>= 0) {
500 if (sk
->sk_peek_off
>= val
)
501 sk
->sk_peek_off
-= val
;
507 static inline void sk_peek_offset_fwd(struct sock
*sk
, int val
)
509 if (sk
->sk_peek_off
>= 0)
510 sk
->sk_peek_off
+= val
;
514 * Hashed lists helper routines
516 static inline struct sock
*sk_entry(const struct hlist_node
*node
)
518 return hlist_entry(node
, struct sock
, sk_node
);
521 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
523 return hlist_entry(head
->first
, struct sock
, sk_node
);
526 static inline struct sock
*sk_head(const struct hlist_head
*head
)
528 return hlist_empty(head
) ? NULL
: __sk_head(head
);
531 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
533 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
536 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
538 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
541 static inline struct sock
*sk_next(const struct sock
*sk
)
543 return sk
->sk_node
.next
?
544 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
547 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
549 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
550 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
551 struct sock
, sk_nulls_node
) :
555 static inline bool sk_unhashed(const struct sock
*sk
)
557 return hlist_unhashed(&sk
->sk_node
);
560 static inline bool sk_hashed(const struct sock
*sk
)
562 return !sk_unhashed(sk
);
565 static inline void sk_node_init(struct hlist_node
*node
)
570 static inline void sk_nulls_node_init(struct hlist_nulls_node
*node
)
575 static inline void __sk_del_node(struct sock
*sk
)
577 __hlist_del(&sk
->sk_node
);
580 /* NB: equivalent to hlist_del_init_rcu */
581 static inline bool __sk_del_node_init(struct sock
*sk
)
585 sk_node_init(&sk
->sk_node
);
591 /* Grab socket reference count. This operation is valid only
592 when sk is ALREADY grabbed f.e. it is found in hash table
593 or a list and the lookup is made under lock preventing hash table
597 static inline void sock_hold(struct sock
*sk
)
599 atomic_inc(&sk
->sk_refcnt
);
602 /* Ungrab socket in the context, which assumes that socket refcnt
603 cannot hit zero, f.e. it is true in context of any socketcall.
605 static inline void __sock_put(struct sock
*sk
)
607 atomic_dec(&sk
->sk_refcnt
);
610 static inline bool sk_del_node_init(struct sock
*sk
)
612 bool rc
= __sk_del_node_init(sk
);
615 /* paranoid for a while -acme */
616 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
621 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
623 static inline bool __sk_nulls_del_node_init_rcu(struct sock
*sk
)
626 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
632 static inline bool sk_nulls_del_node_init_rcu(struct sock
*sk
)
634 bool rc
= __sk_nulls_del_node_init_rcu(sk
);
637 /* paranoid for a while -acme */
638 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
644 static inline void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
646 hlist_add_head(&sk
->sk_node
, list
);
649 static inline void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
652 __sk_add_node(sk
, list
);
655 static inline void sk_add_node_rcu(struct sock
*sk
, struct hlist_head
*list
)
658 hlist_add_head_rcu(&sk
->sk_node
, list
);
661 static inline void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
663 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
666 static inline void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
669 __sk_nulls_add_node_rcu(sk
, list
);
672 static inline void __sk_del_bind_node(struct sock
*sk
)
674 __hlist_del(&sk
->sk_bind_node
);
677 static inline void sk_add_bind_node(struct sock
*sk
,
678 struct hlist_head
*list
)
680 hlist_add_head(&sk
->sk_bind_node
, list
);
683 #define sk_for_each(__sk, list) \
684 hlist_for_each_entry(__sk, list, sk_node)
685 #define sk_for_each_rcu(__sk, list) \
686 hlist_for_each_entry_rcu(__sk, list, sk_node)
687 #define sk_nulls_for_each(__sk, node, list) \
688 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
689 #define sk_nulls_for_each_rcu(__sk, node, list) \
690 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
691 #define sk_for_each_from(__sk) \
692 hlist_for_each_entry_from(__sk, sk_node)
693 #define sk_nulls_for_each_from(__sk, node) \
694 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
695 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
696 #define sk_for_each_safe(__sk, tmp, list) \
697 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
698 #define sk_for_each_bound(__sk, list) \
699 hlist_for_each_entry(__sk, list, sk_bind_node)
702 * sk_nulls_for_each_entry_offset - iterate over a list at a given struct offset
703 * @tpos: the type * to use as a loop cursor.
704 * @pos: the &struct hlist_node to use as a loop cursor.
705 * @head: the head for your list.
706 * @offset: offset of hlist_node within the struct.
709 #define sk_nulls_for_each_entry_offset(tpos, pos, head, offset) \
710 for (pos = (head)->first; \
711 (!is_a_nulls(pos)) && \
712 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
715 static inline struct user_namespace
*sk_user_ns(struct sock
*sk
)
717 /* Careful only use this in a context where these parameters
718 * can not change and must all be valid, such as recvmsg from
721 return sk
->sk_socket
->file
->f_cred
->user_ns
;
735 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
736 SOCK_DBG
, /* %SO_DEBUG setting */
737 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
738 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
739 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
740 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
741 SOCK_MEMALLOC
, /* VM depends on this socket for swapping */
742 SOCK_TIMESTAMPING_RX_SOFTWARE
, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
743 SOCK_FASYNC
, /* fasync() active */
745 SOCK_ZEROCOPY
, /* buffers from userspace */
746 SOCK_WIFI_STATUS
, /* push wifi status to userspace */
747 SOCK_NOFCS
, /* Tell NIC not to do the Ethernet FCS.
748 * Will use last 4 bytes of packet sent from
749 * user-space instead.
751 SOCK_FILTER_LOCKED
, /* Filter cannot be changed anymore */
752 SOCK_SELECT_ERR_QUEUE
, /* Wake select on error queue */
754 SOCK_MPTCP
, /* MPTCP set on this socket */
758 #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
760 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
762 nsk
->sk_flags
= osk
->sk_flags
;
765 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
767 __set_bit(flag
, &sk
->sk_flags
);
770 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
772 __clear_bit(flag
, &sk
->sk_flags
);
775 static inline bool sock_flag(const struct sock
*sk
, enum sock_flags flag
)
777 return test_bit(flag
, &sk
->sk_flags
);
781 extern struct static_key memalloc_socks
;
782 static inline int sk_memalloc_socks(void)
784 return static_key_false(&memalloc_socks
);
788 static inline int sk_memalloc_socks(void)
795 static inline gfp_t
sk_gfp_atomic(const struct sock
*sk
, gfp_t gfp_mask
)
797 return GFP_ATOMIC
| (sk
->sk_allocation
& __GFP_MEMALLOC
);
800 static inline void sk_acceptq_removed(struct sock
*sk
)
802 sk
->sk_ack_backlog
--;
805 static inline void sk_acceptq_added(struct sock
*sk
)
807 sk
->sk_ack_backlog
++;
810 static inline bool sk_acceptq_is_full(const struct sock
*sk
)
812 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
816 * Compute minimal free write space needed to queue new packets.
818 static inline int sk_stream_min_wspace(const struct sock
*sk
)
820 return sk
->sk_wmem_queued
>> 1;
823 static inline int sk_stream_wspace(const struct sock
*sk
)
825 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
828 void sk_stream_write_space(struct sock
*sk
);
830 /* OOB backlog add */
831 static inline void __sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
833 /* dont let skb dst not refcounted, we are going to leave rcu lock */
834 skb_dst_force_safe(skb
);
836 if (!sk
->sk_backlog
.tail
)
837 sk
->sk_backlog
.head
= skb
;
839 sk
->sk_backlog
.tail
->next
= skb
;
841 sk
->sk_backlog
.tail
= skb
;
846 * Take into account size of receive queue and backlog queue
847 * Do not take into account this skb truesize,
848 * to allow even a single big packet to come.
850 static inline bool sk_rcvqueues_full(const struct sock
*sk
, unsigned int limit
)
852 unsigned int qsize
= sk
->sk_backlog
.len
+ atomic_read(&sk
->sk_rmem_alloc
);
854 return qsize
> limit
;
857 /* The per-socket spinlock must be held here. */
858 static inline __must_check
int sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
,
861 if (sk_rcvqueues_full(sk
, limit
))
865 * If the skb was allocated from pfmemalloc reserves, only
866 * allow SOCK_MEMALLOC sockets to use it as this socket is
867 * helping free memory
869 if (skb_pfmemalloc(skb
) && !sock_flag(sk
, SOCK_MEMALLOC
))
872 __sk_add_backlog(sk
, skb
);
873 sk
->sk_backlog
.len
+= skb
->truesize
;
877 int __sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
879 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
881 if (sk_memalloc_socks() && skb_pfmemalloc(skb
))
882 return __sk_backlog_rcv(sk
, skb
);
884 return sk
->sk_backlog_rcv(sk
, skb
);
887 static inline void sk_incoming_cpu_update(struct sock
*sk
)
889 sk
->sk_incoming_cpu
= raw_smp_processor_id();
892 static inline void sock_rps_record_flow_hash(__u32 hash
)
895 struct rps_sock_flow_table
*sock_flow_table
;
898 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
899 rps_record_sock_flow(sock_flow_table
, hash
);
904 static inline void sock_rps_record_flow(const struct sock
*sk
)
907 sock_rps_record_flow_hash(sk
->sk_rxhash
);
911 static inline void sock_rps_save_rxhash(struct sock
*sk
,
912 const struct sk_buff
*skb
)
915 if (unlikely(sk
->sk_rxhash
!= skb
->hash
))
916 sk
->sk_rxhash
= skb
->hash
;
920 static inline void sock_rps_reset_rxhash(struct sock
*sk
)
927 #define sk_wait_event(__sk, __timeo, __condition) \
929 release_sock(__sk); \
930 __rc = __condition; \
932 *(__timeo) = schedule_timeout(*(__timeo)); \
934 sched_annotate_sleep(); \
936 __rc = __condition; \
940 int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
941 int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
942 void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
943 int sk_stream_error(struct sock
*sk
, int flags
, int err
);
944 void sk_stream_kill_queues(struct sock
*sk
);
945 void sk_set_memalloc(struct sock
*sk
);
946 void sk_clear_memalloc(struct sock
*sk
);
948 int sk_wait_data(struct sock
*sk
, long *timeo
, const struct sk_buff
*skb
);
950 /* START - needed for MPTCP */
951 struct sock
*sk_prot_alloc(struct proto
*prot
, gfp_t priority
, int family
);
952 void sock_lock_init(struct sock
*sk
);
954 extern struct lock_class_key af_callback_keys
[AF_MAX
];
955 extern char *const af_family_clock_key_strings
[AF_MAX
+1];
957 #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
958 /* END - needed for MPTCP */
961 struct request_sock_ops
;
962 struct timewait_sock_ops
;
963 struct inet_hashinfo
;
968 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
969 * un-modified. Special care is taken when initializing object to zero.
971 static inline void sk_prot_clear_nulls(struct sock
*sk
, int size
)
973 if (offsetof(struct sock
, sk_node
.next
) != 0)
974 memset(sk
, 0, offsetof(struct sock
, sk_node
.next
));
975 memset(&sk
->sk_node
.pprev
, 0,
976 size
- offsetof(struct sock
, sk_node
.pprev
));
979 /* Networking protocol blocks we attach to sockets.
980 * socket layer -> transport layer interface
983 void (*close
)(struct sock
*sk
,
985 int (*connect
)(struct sock
*sk
,
986 struct sockaddr
*uaddr
,
988 int (*disconnect
)(struct sock
*sk
, int flags
);
990 struct sock
* (*accept
)(struct sock
*sk
, int flags
, int *err
);
992 int (*ioctl
)(struct sock
*sk
, int cmd
,
994 int (*init
)(struct sock
*sk
);
995 void (*destroy
)(struct sock
*sk
);
996 void (*shutdown
)(struct sock
*sk
, int how
);
997 int (*setsockopt
)(struct sock
*sk
, int level
,
998 int optname
, char __user
*optval
,
999 unsigned int optlen
);
1000 int (*getsockopt
)(struct sock
*sk
, int level
,
1001 int optname
, char __user
*optval
,
1002 int __user
*option
);
1003 #ifdef CONFIG_COMPAT
1004 int (*compat_setsockopt
)(struct sock
*sk
,
1006 int optname
, char __user
*optval
,
1007 unsigned int optlen
);
1008 int (*compat_getsockopt
)(struct sock
*sk
,
1010 int optname
, char __user
*optval
,
1011 int __user
*option
);
1012 int (*compat_ioctl
)(struct sock
*sk
,
1013 unsigned int cmd
, unsigned long arg
);
1015 int (*sendmsg
)(struct sock
*sk
, struct msghdr
*msg
,
1017 int (*recvmsg
)(struct sock
*sk
, struct msghdr
*msg
,
1018 size_t len
, int noblock
, int flags
,
1020 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
1021 int offset
, size_t size
, int flags
);
1022 int (*bind
)(struct sock
*sk
,
1023 struct sockaddr
*uaddr
, int addr_len
);
1025 int (*backlog_rcv
) (struct sock
*sk
,
1026 struct sk_buff
*skb
);
1028 void (*release_cb
)(struct sock
*sk
);
1030 /* Keeping track of sk's, looking them up, and port selection methods. */
1031 void (*hash
)(struct sock
*sk
);
1032 void (*unhash
)(struct sock
*sk
);
1033 void (*rehash
)(struct sock
*sk
);
1034 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
1035 void (*clear_sk
)(struct sock
*sk
, int size
);
1037 void (*copy_sk
)(struct sock
*nsk
, const struct sock
*osk
);
1040 /* Keeping track of sockets in use */
1041 #ifdef CONFIG_PROC_FS
1042 unsigned int inuse_idx
;
1045 bool (*stream_memory_free
)(const struct sock
*sk
);
1046 /* Memory pressure */
1047 void (*enter_memory_pressure
)(struct sock
*sk
);
1048 atomic_long_t
*memory_allocated
; /* Current allocated memory. */
1049 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
1051 * Pressure flag: try to collapse.
1052 * Technical note: it is used by multiple contexts non atomically.
1053 * All the __sk_mem_schedule() is of this nature: accounting
1054 * is strict, actions are advisory and have some latency.
1056 int *memory_pressure
;
1063 struct kmem_cache
*slab
;
1064 unsigned int obj_size
;
1067 struct percpu_counter
*orphan_count
;
1069 struct request_sock_ops
*rsk_prot
;
1070 struct timewait_sock_ops
*twsk_prot
;
1073 struct inet_hashinfo
*hashinfo
;
1074 struct udp_table
*udp_table
;
1075 struct raw_hashinfo
*raw_hash
;
1078 struct module
*owner
;
1082 struct list_head node
;
1083 #ifdef SOCK_REFCNT_DEBUG
1086 #ifdef CONFIG_MEMCG_KMEM
1088 * cgroup specific init/deinit functions. Called once for all
1089 * protocols that implement it, from cgroups populate function.
1090 * This function has to setup any files the protocol want to
1091 * appear in the kmem cgroup filesystem.
1093 int (*init_cgroup
)(struct mem_cgroup
*memcg
,
1094 struct cgroup_subsys
*ss
);
1095 void (*destroy_cgroup
)(struct mem_cgroup
*memcg
);
1096 struct cg_proto
*(*proto_cgroup
)(struct mem_cgroup
*memcg
);
1098 int (*diag_destroy
)(struct sock
*sk
, int err
);
1101 int proto_register(struct proto
*prot
, int alloc_slab
);
1102 void proto_unregister(struct proto
*prot
);
1104 #ifdef SOCK_REFCNT_DEBUG
1105 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
1107 atomic_inc(&sk
->sk_prot
->socks
);
1110 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
1112 atomic_dec(&sk
->sk_prot
->socks
);
1113 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
1114 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
1117 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
1119 if (atomic_read(&sk
->sk_refcnt
) != 1)
1120 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
1121 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
1123 #else /* SOCK_REFCNT_DEBUG */
1124 #define sk_refcnt_debug_inc(sk) do { } while (0)
1125 #define sk_refcnt_debug_dec(sk) do { } while (0)
1126 #define sk_refcnt_debug_release(sk) do { } while (0)
1127 #endif /* SOCK_REFCNT_DEBUG */
1129 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
1130 extern struct static_key memcg_socket_limit_enabled
;
1131 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
1132 struct cg_proto
*cg_proto
)
1134 return proto
->proto_cgroup(parent_mem_cgroup(cg_proto
->memcg
));
1136 #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
1138 #define mem_cgroup_sockets_enabled 0
1139 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
1140 struct cg_proto
*cg_proto
)
1146 static inline bool sk_stream_memory_free(const struct sock
*sk
)
1148 if (sk
->sk_wmem_queued
>= sk
->sk_sndbuf
)
1151 return sk
->sk_prot
->stream_memory_free
?
1152 sk
->sk_prot
->stream_memory_free(sk
) : true;
1155 static inline bool sk_stream_is_writeable(const struct sock
*sk
)
1157 return sk_stream_wspace(sk
) >= sk_stream_min_wspace(sk
) &&
1158 sk_stream_memory_free(sk
);
1162 static inline bool sk_has_memory_pressure(const struct sock
*sk
)
1164 return sk
->sk_prot
->memory_pressure
!= NULL
;
1167 static inline bool sk_under_memory_pressure(const struct sock
*sk
)
1169 if (!sk
->sk_prot
->memory_pressure
)
1172 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1173 return !!sk
->sk_cgrp
->memory_pressure
;
1175 return !!*sk
->sk_prot
->memory_pressure
;
1178 static inline void sk_leave_memory_pressure(struct sock
*sk
)
1180 int *memory_pressure
= sk
->sk_prot
->memory_pressure
;
1182 if (!memory_pressure
)
1185 if (*memory_pressure
)
1186 *memory_pressure
= 0;
1188 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1189 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1190 struct proto
*prot
= sk
->sk_prot
;
1192 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1193 cg_proto
->memory_pressure
= 0;
1198 static inline void sk_enter_memory_pressure(struct sock
*sk
)
1200 if (!sk
->sk_prot
->enter_memory_pressure
)
1203 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1204 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1205 struct proto
*prot
= sk
->sk_prot
;
1207 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1208 cg_proto
->memory_pressure
= 1;
1211 sk
->sk_prot
->enter_memory_pressure(sk
);
1214 static inline long sk_prot_mem_limits(const struct sock
*sk
, int index
)
1216 long *prot
= sk
->sk_prot
->sysctl_mem
;
1217 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1218 prot
= sk
->sk_cgrp
->sysctl_mem
;
1222 static inline void memcg_memory_allocated_add(struct cg_proto
*prot
,
1226 page_counter_charge(&prot
->memory_allocated
, amt
);
1228 if (page_counter_read(&prot
->memory_allocated
) >
1229 prot
->memory_allocated
.limit
)
1230 *parent_status
= OVER_LIMIT
;
1233 static inline void memcg_memory_allocated_sub(struct cg_proto
*prot
,
1236 page_counter_uncharge(&prot
->memory_allocated
, amt
);
1240 sk_memory_allocated(const struct sock
*sk
)
1242 struct proto
*prot
= sk
->sk_prot
;
1244 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1245 return page_counter_read(&sk
->sk_cgrp
->memory_allocated
);
1247 return atomic_long_read(prot
->memory_allocated
);
1251 sk_memory_allocated_add(struct sock
*sk
, int amt
, int *parent_status
)
1253 struct proto
*prot
= sk
->sk_prot
;
1255 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1256 memcg_memory_allocated_add(sk
->sk_cgrp
, amt
, parent_status
);
1257 /* update the root cgroup regardless */
1258 atomic_long_add_return(amt
, prot
->memory_allocated
);
1259 return page_counter_read(&sk
->sk_cgrp
->memory_allocated
);
1262 return atomic_long_add_return(amt
, prot
->memory_allocated
);
1266 sk_memory_allocated_sub(struct sock
*sk
, int amt
)
1268 struct proto
*prot
= sk
->sk_prot
;
1270 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1271 memcg_memory_allocated_sub(sk
->sk_cgrp
, amt
);
1273 atomic_long_sub(amt
, prot
->memory_allocated
);
1276 static inline void sk_sockets_allocated_dec(struct sock
*sk
)
1278 struct proto
*prot
= sk
->sk_prot
;
1280 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1281 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1283 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1284 percpu_counter_dec(&cg_proto
->sockets_allocated
);
1287 percpu_counter_dec(prot
->sockets_allocated
);
1290 static inline void sk_sockets_allocated_inc(struct sock
*sk
)
1292 struct proto
*prot
= sk
->sk_prot
;
1294 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1295 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1297 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1298 percpu_counter_inc(&cg_proto
->sockets_allocated
);
1301 percpu_counter_inc(prot
->sockets_allocated
);
1305 sk_sockets_allocated_read_positive(struct sock
*sk
)
1307 struct proto
*prot
= sk
->sk_prot
;
1309 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1310 return percpu_counter_read_positive(&sk
->sk_cgrp
->sockets_allocated
);
1312 return percpu_counter_read_positive(prot
->sockets_allocated
);
1316 proto_sockets_allocated_sum_positive(struct proto
*prot
)
1318 return percpu_counter_sum_positive(prot
->sockets_allocated
);
1322 proto_memory_allocated(struct proto
*prot
)
1324 return atomic_long_read(prot
->memory_allocated
);
1328 proto_memory_pressure(struct proto
*prot
)
1330 if (!prot
->memory_pressure
)
1332 return !!*prot
->memory_pressure
;
1336 #ifdef CONFIG_PROC_FS
1337 /* Called with local bh disabled */
1338 void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int inc
);
1339 int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
1341 static inline void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
1348 /* With per-bucket locks this operation is not-atomic, so that
1349 * this version is not worse.
1351 static inline void __sk_prot_rehash(struct sock
*sk
)
1353 sk
->sk_prot
->unhash(sk
);
1354 sk
->sk_prot
->hash(sk
);
1357 void sk_prot_clear_portaddr_nulls(struct sock
*sk
, int size
);
1359 /* About 10 seconds */
1360 #define SOCK_DESTROY_TIME (10*HZ)
1362 /* Sockets 0-1023 can't be bound to unless you are superuser */
1363 #define PROT_SOCK 1024
1365 #define SHUTDOWN_MASK 3
1366 #define RCV_SHUTDOWN 1
1367 #define SEND_SHUTDOWN 2
1369 #define SOCK_SNDBUF_LOCK 1
1370 #define SOCK_RCVBUF_LOCK 2
1371 #define SOCK_BINDADDR_LOCK 4
1372 #define SOCK_BINDPORT_LOCK 8
1374 struct socket_alloc
{
1375 struct socket socket
;
1376 struct inode vfs_inode
;
1379 static inline struct socket
*SOCKET_I(struct inode
*inode
)
1381 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
1384 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
1386 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
1390 * Functions for memory accounting
1392 int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
1393 void __sk_mem_reclaim(struct sock
*sk
, int amount
);
1395 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1396 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1397 #define SK_MEM_SEND 0
1398 #define SK_MEM_RECV 1
1400 static inline int sk_mem_pages(int amt
)
1402 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
1405 static inline bool sk_has_account(struct sock
*sk
)
1407 /* return true if protocol supports memory accounting */
1408 return !!sk
->sk_prot
->memory_allocated
;
1411 static inline bool sk_wmem_schedule(struct sock
*sk
, int size
)
1413 if (!sk_has_account(sk
))
1415 return size
<= sk
->sk_forward_alloc
||
1416 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
1420 sk_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
, int size
)
1422 if (!sk_has_account(sk
))
1424 return size
<= sk
->sk_forward_alloc
||
1425 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
) ||
1426 skb_pfmemalloc(skb
);
1429 static inline void sk_mem_reclaim(struct sock
*sk
)
1431 if (!sk_has_account(sk
))
1433 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
1434 __sk_mem_reclaim(sk
, sk
->sk_forward_alloc
);
1437 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
1439 if (!sk_has_account(sk
))
1441 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
1442 __sk_mem_reclaim(sk
, sk
->sk_forward_alloc
- 1);
1445 static inline void sk_mem_charge(struct sock
*sk
, int size
)
1447 if (!sk_has_account(sk
))
1449 sk
->sk_forward_alloc
-= size
;
1452 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
1454 if (!sk_has_account(sk
))
1456 sk
->sk_forward_alloc
+= size
;
1458 /* Avoid a possible overflow.
1459 * TCP send queues can make this happen, if sk_mem_reclaim()
1460 * is not called and more than 2 GBytes are released at once.
1462 * If we reach 2 MBytes, reclaim 1 MBytes right now, there is
1463 * no need to hold that much forward allocation anyway.
1465 if (unlikely(sk
->sk_forward_alloc
>= 1 << 21))
1466 __sk_mem_reclaim(sk
, 1 << 20);
1469 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
1471 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
1472 sk
->sk_wmem_queued
-= skb
->truesize
;
1473 sk_mem_uncharge(sk
, skb
->truesize
);
1477 /* Used by processes to "lock" a socket state, so that
1478 * interrupts and bottom half handlers won't change it
1479 * from under us. It essentially blocks any incoming
1480 * packets, so that we won't get any new data or any
1481 * packets that change the state of the socket.
1483 * While locked, BH processing will add new packets to
1484 * the backlog queue. This queue is processed by the
1485 * owner of the socket lock right before it is released.
1487 * Since ~2.3.5 it is also exclusive sleep lock serializing
1488 * accesses from user process context.
1490 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1492 static inline void sock_release_ownership(struct sock
*sk
)
1494 sk
->sk_lock
.owned
= 0;
1498 * Macro so as to not evaluate some arguments when
1499 * lockdep is not enabled.
1501 * Mark both the sk_lock and the sk_lock.slock as a
1502 * per-address-family lock class.
1504 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1506 sk->sk_lock.owned = 0; \
1507 init_waitqueue_head(&sk->sk_lock.wq); \
1508 spin_lock_init(&(sk)->sk_lock.slock); \
1509 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1510 sizeof((sk)->sk_lock)); \
1511 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1513 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1516 void lock_sock_nested(struct sock
*sk
, int subclass
);
1518 static inline void lock_sock(struct sock
*sk
)
1520 lock_sock_nested(sk
, 0);
1523 void release_sock(struct sock
*sk
);
1525 /* BH context may only use the following locking interface. */
1526 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1527 #define bh_lock_sock_nested(__sk) \
1528 spin_lock_nested(&((__sk)->sk_lock.slock), \
1529 SINGLE_DEPTH_NESTING)
1530 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1532 bool lock_sock_fast(struct sock
*sk
);
1534 * unlock_sock_fast - complement of lock_sock_fast
1538 * fast unlock socket for user context.
1539 * If slow mode is on, we call regular release_sock()
1541 static inline void unlock_sock_fast(struct sock
*sk
, bool slow
)
1546 spin_unlock_bh(&sk
->sk_lock
.slock
);
1550 struct sock
*sk_alloc(struct net
*net
, int family
, gfp_t priority
,
1551 struct proto
*prot
, int kern
);
1552 void sk_free(struct sock
*sk
);
1553 void sk_destruct(struct sock
*sk
);
1554 struct sock
*sk_clone_lock(const struct sock
*sk
, const gfp_t priority
);
1556 struct sk_buff
*sock_wmalloc(struct sock
*sk
, unsigned long size
, int force
,
1558 void sock_wfree(struct sk_buff
*skb
);
1559 void skb_orphan_partial(struct sk_buff
*skb
);
1560 void sock_rfree(struct sk_buff
*skb
);
1561 void sock_efree(struct sk_buff
*skb
);
1563 void sock_edemux(struct sk_buff
*skb
);
1565 #define sock_edemux(skb) sock_efree(skb)
1568 int sock_setsockopt(struct socket
*sock
, int level
, int op
,
1569 char __user
*optval
, unsigned int optlen
);
1571 int sock_getsockopt(struct socket
*sock
, int level
, int op
,
1572 char __user
*optval
, int __user
*optlen
);
1573 struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
, unsigned long size
,
1574 int noblock
, int *errcode
);
1575 struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
, unsigned long header_len
,
1576 unsigned long data_len
, int noblock
,
1577 int *errcode
, int max_page_order
);
1578 void *sock_kmalloc(struct sock
*sk
, int size
, gfp_t priority
);
1579 void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
1580 void sock_kzfree_s(struct sock
*sk
, void *mem
, int size
);
1581 void sk_send_sigurg(struct sock
*sk
);
1583 struct sockcm_cookie
{
1587 int sock_cmsg_send(struct sock
*sk
, struct msghdr
*msg
,
1588 struct sockcm_cookie
*sockc
);
1591 * Functions to fill in entries in struct proto_ops when a protocol
1592 * does not implement a particular function.
1594 int sock_no_bind(struct socket
*, struct sockaddr
*, int);
1595 int sock_no_connect(struct socket
*, struct sockaddr
*, int, int);
1596 int sock_no_socketpair(struct socket
*, struct socket
*);
1597 int sock_no_accept(struct socket
*, struct socket
*, int);
1598 int sock_no_getname(struct socket
*, struct sockaddr
*, int *, int);
1599 unsigned int sock_no_poll(struct file
*, struct socket
*,
1600 struct poll_table_struct
*);
1601 int sock_no_ioctl(struct socket
*, unsigned int, unsigned long);
1602 int sock_no_listen(struct socket
*, int);
1603 int sock_no_shutdown(struct socket
*, int);
1604 int sock_no_getsockopt(struct socket
*, int , int, char __user
*, int __user
*);
1605 int sock_no_setsockopt(struct socket
*, int, int, char __user
*, unsigned int);
1606 int sock_no_sendmsg(struct socket
*, struct msghdr
*, size_t);
1607 int sock_no_recvmsg(struct socket
*, struct msghdr
*, size_t, int);
1608 int sock_no_mmap(struct file
*file
, struct socket
*sock
,
1609 struct vm_area_struct
*vma
);
1610 ssize_t
sock_no_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
1611 size_t size
, int flags
);
1614 * Functions to fill in entries in struct proto_ops when a protocol
1615 * uses the inet style.
1617 int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1618 char __user
*optval
, int __user
*optlen
);
1619 int sock_common_recvmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
,
1621 int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1622 char __user
*optval
, unsigned int optlen
);
1623 int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
1624 int optname
, char __user
*optval
, int __user
*optlen
);
1625 int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
1626 int optname
, char __user
*optval
, unsigned int optlen
);
1628 void sk_common_release(struct sock
*sk
);
1631 * Default socket callbacks and setup code
1634 /* Initialise core socket variables */
1635 void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1638 * Socket reference counting postulates.
1640 * * Each user of socket SHOULD hold a reference count.
1641 * * Each access point to socket (an hash table bucket, reference from a list,
1642 * running timer, skb in flight MUST hold a reference count.
1643 * * When reference count hits 0, it means it will never increase back.
1644 * * When reference count hits 0, it means that no references from
1645 * outside exist to this socket and current process on current CPU
1646 * is last user and may/should destroy this socket.
1647 * * sk_free is called from any context: process, BH, IRQ. When
1648 * it is called, socket has no references from outside -> sk_free
1649 * may release descendant resources allocated by the socket, but
1650 * to the time when it is called, socket is NOT referenced by any
1651 * hash tables, lists etc.
1652 * * Packets, delivered from outside (from network or from another process)
1653 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1654 * when they sit in queue. Otherwise, packets will leak to hole, when
1655 * socket is looked up by one cpu and unhasing is made by another CPU.
1656 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1657 * (leak to backlog). Packet socket does all the processing inside
1658 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1659 * use separate SMP lock, so that they are prone too.
1662 /* Ungrab socket and destroy it, if it was the last reference. */
1663 static inline void sock_put(struct sock
*sk
)
1665 if (atomic_dec_and_test(&sk
->sk_refcnt
))
1668 /* Generic version of sock_put(), dealing with all sockets
1669 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
1671 void sock_gen_put(struct sock
*sk
);
1673 int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
, const int nested
);
1675 static inline void sk_tx_queue_set(struct sock
*sk
, int tx_queue
)
1677 sk
->sk_tx_queue_mapping
= tx_queue
;
1680 static inline void sk_tx_queue_clear(struct sock
*sk
)
1682 sk
->sk_tx_queue_mapping
= -1;
1685 static inline int sk_tx_queue_get(const struct sock
*sk
)
1687 return sk
? sk
->sk_tx_queue_mapping
: -1;
1690 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1692 sk_tx_queue_clear(sk
);
1693 sk
->sk_socket
= sock
;
1696 static inline wait_queue_head_t
*sk_sleep(struct sock
*sk
)
1698 BUILD_BUG_ON(offsetof(struct socket_wq
, wait
) != 0);
1699 return &rcu_dereference_raw(sk
->sk_wq
)->wait
;
1701 /* Detach socket from process context.
1702 * Announce socket dead, detach it from wait queue and inode.
1703 * Note that parent inode held reference count on this struct sock,
1704 * we do not release it in this function, because protocol
1705 * probably wants some additional cleanups or even continuing
1706 * to work with this socket (TCP).
1708 static inline void sock_orphan(struct sock
*sk
)
1710 write_lock_bh(&sk
->sk_callback_lock
);
1711 sock_set_flag(sk
, SOCK_DEAD
);
1712 sk_set_socket(sk
, NULL
);
1714 write_unlock_bh(&sk
->sk_callback_lock
);
1717 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1719 write_lock_bh(&sk
->sk_callback_lock
);
1720 sk
->sk_wq
= parent
->wq
;
1722 sk_set_socket(sk
, parent
);
1723 sk
->sk_uid
= SOCK_INODE(parent
)->i_uid
;
1724 security_sock_graft(sk
, parent
);
1725 write_unlock_bh(&sk
->sk_callback_lock
);
1728 kuid_t
sock_i_uid(struct sock
*sk
);
1729 unsigned long sock_i_ino(struct sock
*sk
);
1731 static inline kuid_t
sock_net_uid(const struct net
*net
, const struct sock
*sk
)
1733 return sk
? sk
->sk_uid
: make_kuid(net
->user_ns
, 0);
1737 static inline u32
net_tx_rndhash(void)
1739 u32 v
= prandom_u32();
1744 static inline void sk_set_txhash(struct sock
*sk
)
1746 sk
->sk_txhash
= net_tx_rndhash();
1749 static inline void sk_rethink_txhash(struct sock
*sk
)
1755 static inline struct dst_entry
*
1756 __sk_dst_get(struct sock
*sk
)
1758 return rcu_dereference_check(sk
->sk_dst_cache
, sock_owned_by_user(sk
) ||
1759 lockdep_is_held(&sk
->sk_lock
.slock
));
1762 static inline struct dst_entry
*
1763 sk_dst_get(struct sock
*sk
)
1765 struct dst_entry
*dst
;
1768 dst
= rcu_dereference(sk
->sk_dst_cache
);
1769 if (dst
&& !atomic_inc_not_zero(&dst
->__refcnt
))
1775 static inline void dst_negative_advice(struct sock
*sk
)
1777 struct dst_entry
*ndst
, *dst
= __sk_dst_get(sk
);
1779 sk_rethink_txhash(sk
);
1781 if (dst
&& dst
->ops
->negative_advice
) {
1782 ndst
= dst
->ops
->negative_advice(dst
);
1785 rcu_assign_pointer(sk
->sk_dst_cache
, ndst
);
1786 sk_tx_queue_clear(sk
);
1792 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1794 struct dst_entry
*old_dst
;
1796 sk_tx_queue_clear(sk
);
1798 * This can be called while sk is owned by the caller only,
1799 * with no state that can be checked in a rcu_dereference_check() cond
1801 old_dst
= rcu_dereference_raw(sk
->sk_dst_cache
);
1802 rcu_assign_pointer(sk
->sk_dst_cache
, dst
);
1803 dst_release(old_dst
);
1807 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1809 struct dst_entry
*old_dst
;
1811 sk_tx_queue_clear(sk
);
1812 old_dst
= xchg((__force
struct dst_entry
**)&sk
->sk_dst_cache
, dst
);
1813 dst_release(old_dst
);
1817 __sk_dst_reset(struct sock
*sk
)
1819 __sk_dst_set(sk
, NULL
);
1823 sk_dst_reset(struct sock
*sk
)
1825 sk_dst_set(sk
, NULL
);
1828 struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1830 struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1832 bool sk_mc_loop(struct sock
*sk
);
1834 static inline bool sk_can_gso(const struct sock
*sk
)
1836 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1839 void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1841 static inline void sk_nocaps_add(struct sock
*sk
, netdev_features_t flags
)
1843 sk
->sk_route_nocaps
|= flags
;
1844 sk
->sk_route_caps
&= ~flags
;
1847 static inline int skb_do_copy_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1848 struct iov_iter
*from
, char *to
,
1849 int copy
, int offset
)
1851 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1853 if (csum_and_copy_from_iter(to
, copy
, &csum
, from
) != copy
)
1855 skb
->csum
= csum_block_add(skb
->csum
, csum
, offset
);
1856 } else if (sk
->sk_route_caps
& NETIF_F_NOCACHE_COPY
) {
1857 if (copy_from_iter_nocache(to
, copy
, from
) != copy
)
1859 } else if (copy_from_iter(to
, copy
, from
) != copy
)
1865 static inline int skb_add_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1866 struct iov_iter
*from
, int copy
)
1868 int err
, offset
= skb
->len
;
1870 err
= skb_do_copy_data_nocache(sk
, skb
, from
, skb_put(skb
, copy
),
1873 __skb_trim(skb
, offset
);
1878 static inline int skb_copy_to_page_nocache(struct sock
*sk
, struct iov_iter
*from
,
1879 struct sk_buff
*skb
,
1885 err
= skb_do_copy_data_nocache(sk
, skb
, from
, page_address(page
) + off
,
1891 skb
->data_len
+= copy
;
1892 skb
->truesize
+= copy
;
1893 sk
->sk_wmem_queued
+= copy
;
1894 sk_mem_charge(sk
, copy
);
1899 * sk_wmem_alloc_get - returns write allocations
1902 * Returns sk_wmem_alloc minus initial offset of one
1904 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
1906 return atomic_read(&sk
->sk_wmem_alloc
) - 1;
1910 * sk_rmem_alloc_get - returns read allocations
1913 * Returns sk_rmem_alloc
1915 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
1917 return atomic_read(&sk
->sk_rmem_alloc
);
1921 * sk_has_allocations - check if allocations are outstanding
1924 * Returns true if socket has write or read allocations
1926 static inline bool sk_has_allocations(const struct sock
*sk
)
1928 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
1932 * wq_has_sleeper - check if there are any waiting processes
1933 * @wq: struct socket_wq
1935 * Returns true if socket_wq has waiting processes
1937 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1938 * barrier call. They were added due to the race found within the tcp code.
1940 * Consider following tcp code paths:
1944 * sys_select receive packet
1946 * __add_wait_queue update tp->rcv_nxt
1948 * tp->rcv_nxt check sock_def_readable
1950 * schedule rcu_read_lock();
1951 * wq = rcu_dereference(sk->sk_wq);
1952 * if (wq && waitqueue_active(&wq->wait))
1953 * wake_up_interruptible(&wq->wait)
1957 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1958 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1959 * could then endup calling schedule and sleep forever if there are no more
1960 * data on the socket.
1963 static inline bool wq_has_sleeper(struct socket_wq
*wq
)
1965 /* We need to be sure we are in sync with the
1966 * add_wait_queue modifications to the wait queue.
1968 * This memory barrier is paired in the sock_poll_wait.
1971 return wq
&& waitqueue_active(&wq
->wait
);
1975 * sock_poll_wait - place memory barrier behind the poll_wait call.
1977 * @wait_address: socket wait queue
1980 * See the comments in the wq_has_sleeper function.
1982 static inline void sock_poll_wait(struct file
*filp
,
1983 wait_queue_head_t
*wait_address
, poll_table
*p
)
1985 if (!poll_does_not_wait(p
) && wait_address
) {
1986 poll_wait(filp
, wait_address
, p
);
1987 /* We need to be sure we are in sync with the
1988 * socket flags modification.
1990 * This memory barrier is paired in the wq_has_sleeper.
1996 static inline void skb_set_hash_from_sk(struct sk_buff
*skb
, struct sock
*sk
)
1998 if (sk
->sk_txhash
) {
2000 skb
->hash
= sk
->sk_txhash
;
2004 void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
);
2007 * Queue a received datagram if it will fit. Stream and sequenced
2008 * protocols can't normally use this as they need to fit buffers in
2009 * and play with them.
2011 * Inlined as it's very short and called for pretty much every
2012 * packet ever received.
2014 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
2018 skb
->destructor
= sock_rfree
;
2019 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
2020 sk_mem_charge(sk
, skb
->truesize
);
2023 void sk_reset_timer(struct sock
*sk
, struct timer_list
*timer
,
2024 unsigned long expires
);
2026 void sk_stop_timer(struct sock
*sk
, struct timer_list
*timer
);
2028 int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
2030 int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
);
2031 struct sk_buff
*sock_dequeue_err_skb(struct sock
*sk
);
2034 * Recover an error report and clear atomically
2037 static inline int sock_error(struct sock
*sk
)
2040 if (likely(!sk
->sk_err
))
2042 err
= xchg(&sk
->sk_err
, 0);
2046 static inline unsigned long sock_wspace(struct sock
*sk
)
2050 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
2051 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
2059 * We use sk->sk_wq_raw, from contexts knowing this
2060 * pointer is not NULL and cannot disappear/change.
2062 static inline void sk_set_bit(int nr
, struct sock
*sk
)
2064 set_bit(nr
, &sk
->sk_wq_raw
->flags
);
2067 static inline void sk_clear_bit(int nr
, struct sock
*sk
)
2069 clear_bit(nr
, &sk
->sk_wq_raw
->flags
);
2072 static inline void sk_wake_async(const struct sock
*sk
, int how
, int band
)
2074 if (sock_flag(sk
, SOCK_FASYNC
)) {
2076 sock_wake_async(rcu_dereference(sk
->sk_wq
), how
, band
);
2081 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2082 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2083 * Note: for send buffers, TCP works better if we can build two skbs at
2086 #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2088 #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2089 #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
2091 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
2093 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
2094 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
2095 sk
->sk_sndbuf
= max_t(u32
, sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
2099 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
,
2100 bool force_schedule
);
2103 * sk_page_frag - return an appropriate page_frag
2106 * If socket allocation mode allows current thread to sleep, it means its
2107 * safe to use the per task page_frag instead of the per socket one.
2109 static inline struct page_frag
*sk_page_frag(struct sock
*sk
)
2111 if (gfpflags_allow_blocking(sk
->sk_allocation
))
2112 return ¤t
->task_frag
;
2114 return &sk
->sk_frag
;
2117 bool sk_page_frag_refill(struct sock
*sk
, struct page_frag
*pfrag
);
2120 * Default write policy as shown to user space via poll/select/SIGIO
2122 static inline bool sock_writeable(const struct sock
*sk
)
2124 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
2127 static inline gfp_t
gfp_any(void)
2129 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
2132 static inline long sock_rcvtimeo(const struct sock
*sk
, bool noblock
)
2134 return noblock
? 0 : sk
->sk_rcvtimeo
;
2137 static inline long sock_sndtimeo(const struct sock
*sk
, bool noblock
)
2139 return noblock
? 0 : sk
->sk_sndtimeo
;
2142 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
2144 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
2147 /* Alas, with timeout socket operations are not restartable.
2148 * Compare this to poll().
2150 static inline int sock_intr_errno(long timeo
)
2152 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
2155 struct sock_skb_cb
{
2159 /* Store sock_skb_cb at the end of skb->cb[] so protocol families
2160 * using skb->cb[] would keep using it directly and utilize its
2161 * alignement guarantee.
2163 #define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \
2164 sizeof(struct sock_skb_cb)))
2166 #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2167 SOCK_SKB_CB_OFFSET))
2169 #define sock_skb_cb_check_size(size) \
2170 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
2173 sock_skb_set_dropcount(const struct sock
*sk
, struct sk_buff
*skb
)
2175 SOCK_SKB_CB(skb
)->dropcount
= atomic_read(&sk
->sk_drops
);
2178 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
2179 struct sk_buff
*skb
);
2180 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
2181 struct sk_buff
*skb
);
2184 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
2186 ktime_t kt
= skb
->tstamp
;
2187 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
2190 * generate control messages if
2191 * - receive time stamping in software requested
2192 * - software time stamp available and wanted
2193 * - hardware time stamps available and wanted
2195 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
2196 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RX_SOFTWARE
) ||
2197 (kt
.tv64
&& sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
) ||
2198 (hwtstamps
->hwtstamp
.tv64
&&
2199 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
)))
2200 __sock_recv_timestamp(msg
, sk
, skb
);
2204 if (sock_flag(sk
, SOCK_WIFI_STATUS
) && skb
->wifi_acked_valid
)
2205 __sock_recv_wifi_status(msg
, sk
, skb
);
2208 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2209 struct sk_buff
*skb
);
2211 static inline void sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2212 struct sk_buff
*skb
)
2214 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2215 (1UL << SOCK_RCVTSTAMP))
2216 #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2217 SOF_TIMESTAMPING_RAW_HARDWARE)
2219 if (sk
->sk_flags
& FLAGS_TS_OR_DROPS
|| sk
->sk_tsflags
& TSFLAGS_ANY
)
2220 __sock_recv_ts_and_drops(msg
, sk
, skb
);
2222 sk
->sk_stamp
= skb
->tstamp
;
2225 void __sock_tx_timestamp(const struct sock
*sk
, __u8
*tx_flags
);
2228 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2229 * @sk: socket sending this packet
2230 * @tx_flags: completed with instructions for time stamping
2232 * Note : callers should take care of initial *tx_flags value (usually 0)
2234 static inline void sock_tx_timestamp(const struct sock
*sk
, __u8
*tx_flags
)
2236 if (unlikely(sk
->sk_tsflags
))
2237 __sock_tx_timestamp(sk
, tx_flags
);
2238 if (unlikely(sock_flag(sk
, SOCK_WIFI_STATUS
)))
2239 *tx_flags
|= SKBTX_WIFI_STATUS
;
2243 * sk_eat_skb - Release a skb if it is no longer needed
2244 * @sk: socket to eat this skb from
2245 * @skb: socket buffer to eat
2247 * This routine must be called with interrupts disabled or with the socket
2248 * locked so that the sk_buff queue operation is ok.
2250 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
)
2252 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2257 struct net
*sock_net(const struct sock
*sk
)
2259 return read_pnet(&sk
->sk_net
);
2263 void sock_net_set(struct sock
*sk
, struct net
*net
)
2265 write_pnet(&sk
->sk_net
, net
);
2268 static inline struct sock
*skb_steal_sock(struct sk_buff
*skb
)
2271 struct sock
*sk
= skb
->sk
;
2273 skb
->destructor
= NULL
;
2280 /* This helper checks if a socket is a full socket,
2281 * ie _not_ a timewait or request socket.
2283 static inline bool sk_fullsock(const struct sock
*sk
)
2285 return (1 << sk
->sk_state
) & ~(TCPF_TIME_WAIT
| TCPF_NEW_SYN_RECV
);
2288 /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2289 * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2291 static inline bool sk_listener(const struct sock
*sk
)
2293 return (1 << sk
->sk_state
) & (TCPF_LISTEN
| TCPF_NEW_SYN_RECV
);
2297 * sk_state_load - read sk->sk_state for lockless contexts
2298 * @sk: socket pointer
2300 * Paired with sk_state_store(). Used in places we do not hold socket lock :
2301 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
2303 static inline int sk_state_load(const struct sock
*sk
)
2305 return smp_load_acquire(&sk
->sk_state
);
2309 * sk_state_store - update sk->sk_state
2310 * @sk: socket pointer
2311 * @newstate: new state
2313 * Paired with sk_state_load(). Should be used in contexts where
2314 * state change might impact lockless readers.
2316 static inline void sk_state_store(struct sock
*sk
, int newstate
)
2318 smp_store_release(&sk
->sk_state
, newstate
);
2321 void sock_enable_timestamp(struct sock
*sk
, int flag
);
2322 int sock_get_timestamp(struct sock
*, struct timeval __user
*);
2323 int sock_get_timestampns(struct sock
*, struct timespec __user
*);
2324 int sock_recv_errqueue(struct sock
*sk
, struct msghdr
*msg
, int len
, int level
,
2327 bool sk_ns_capable(const struct sock
*sk
,
2328 struct user_namespace
*user_ns
, int cap
);
2329 bool sk_capable(const struct sock
*sk
, int cap
);
2330 bool sk_net_capable(const struct sock
*sk
, int cap
);
2332 extern __u32 sysctl_wmem_max
;
2333 extern __u32 sysctl_rmem_max
;
2335 extern int sysctl_tstamp_allow_data
;
2336 extern int sysctl_optmem_max
;
2338 extern __u32 sysctl_wmem_default
;
2339 extern __u32 sysctl_rmem_default
;
2341 #endif /* _SOCK_H */