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>
69 #include <linux/refcount.h>
71 #include <net/checksum.h>
72 #include <net/tcp_states.h>
73 #include <linux/net_tstamp.h>
77 * This structure really needs to be cleaned up.
78 * Most of it is for TCP, and not used by any of
79 * the other protocols.
82 /* Define this to get the SOCK_DBG debugging facility. */
83 #define SOCK_DEBUGGING
85 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
86 printk(KERN_DEBUG msg); } while (0)
88 /* Validate arguments and do nothing */
89 static inline __printf(2, 3)
90 void SOCK_DEBUG(const struct sock
*sk
, const char *msg
, ...)
95 /* This is the per-socket lock. The spinlock provides a synchronization
96 * between user contexts and software interrupt processing, whereas the
97 * mini-semaphore synchronizes multiple users amongst themselves.
102 wait_queue_head_t wq
;
104 * We express the mutex-alike socket_lock semantics
105 * to the lock validator by explicitly managing
106 * the slock as a lock variant (in addition to
109 #ifdef CONFIG_DEBUG_LOCK_ALLOC
110 struct lockdep_map dep_map
;
118 typedef __u32 __bitwise __portpair
;
119 typedef __u64 __bitwise __addrpair
;
122 * struct sock_common - minimal network layer representation of sockets
123 * @skc_daddr: Foreign IPv4 addr
124 * @skc_rcv_saddr: Bound local IPv4 addr
125 * @skc_hash: hash value used with various protocol lookup tables
126 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
127 * @skc_dport: placeholder for inet_dport/tw_dport
128 * @skc_num: placeholder for inet_num/tw_num
129 * @skc_family: network address family
130 * @skc_state: Connection state
131 * @skc_reuse: %SO_REUSEADDR setting
132 * @skc_reuseport: %SO_REUSEPORT setting
133 * @skc_bound_dev_if: bound device index if != 0
134 * @skc_bind_node: bind hash linkage for various protocol lookup tables
135 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
136 * @skc_prot: protocol handlers inside a network family
137 * @skc_net: reference to the network namespace of this socket
138 * @skc_node: main hash linkage for various protocol lookup tables
139 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
140 * @skc_tx_queue_mapping: tx queue number for this connection
141 * @skc_flags: place holder for sk_flags
142 * %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
143 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
144 * @skc_incoming_cpu: record/match cpu processing incoming packets
145 * @skc_refcnt: reference count
147 * This is the minimal network layer representation of sockets, the header
148 * for struct sock and struct inet_timewait_sock.
151 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
152 * address on 64bit arches : cf INET_MATCH()
155 __addrpair skc_addrpair
;
158 __be32 skc_rcv_saddr
;
162 unsigned int skc_hash
;
163 __u16 skc_u16hashes
[2];
165 /* skc_dport && skc_num must be grouped as well */
167 __portpair skc_portpair
;
174 unsigned short skc_family
;
175 volatile unsigned char skc_state
;
176 unsigned char skc_reuse
:4;
177 unsigned char skc_reuseport
:1;
178 unsigned char skc_ipv6only
:1;
179 unsigned char skc_net_refcnt
:1;
180 int skc_bound_dev_if
;
182 struct hlist_node skc_bind_node
;
183 struct hlist_node skc_portaddr_node
;
185 struct proto
*skc_prot
;
186 possible_net_t skc_net
;
188 #if IS_ENABLED(CONFIG_IPV6)
189 struct in6_addr skc_v6_daddr
;
190 struct in6_addr skc_v6_rcv_saddr
;
193 atomic64_t skc_cookie
;
195 /* following fields are padding to force
196 * offset(struct sock, sk_refcnt) == 128 on 64bit arches
197 * assuming IPV6 is enabled. We use this padding differently
198 * for different kind of 'sockets'
201 unsigned long skc_flags
;
202 struct sock
*skc_listener
; /* request_sock */
203 struct inet_timewait_death_row
*skc_tw_dr
; /* inet_timewait_sock */
206 * fields between dontcopy_begin/dontcopy_end
207 * are not copied in sock_copy()
210 int skc_dontcopy_begin
[0];
213 struct hlist_node skc_node
;
214 struct hlist_nulls_node skc_nulls_node
;
216 int skc_tx_queue_mapping
;
218 int skc_incoming_cpu
;
220 u32 skc_tw_rcv_nxt
; /* struct tcp_timewait_sock */
223 refcount_t skc_refcnt
;
225 int skc_dontcopy_end
[0];
228 u32 skc_window_clamp
;
229 u32 skc_tw_snd_nxt
; /* struct tcp_timewait_sock */
235 * struct sock - network layer representation of sockets
236 * @__sk_common: shared layout with inet_timewait_sock
237 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
238 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
239 * @sk_lock: synchronizer
240 * @sk_kern_sock: True if sock is using kernel lock classes
241 * @sk_rcvbuf: size of receive buffer in bytes
242 * @sk_wq: sock wait queue and async head
243 * @sk_rx_dst: receive input route used by early demux
244 * @sk_dst_cache: destination cache
245 * @sk_dst_pending_confirm: need to confirm neighbour
246 * @sk_policy: flow policy
247 * @sk_receive_queue: incoming packets
248 * @sk_wmem_alloc: transmit queue bytes committed
249 * @sk_tsq_flags: TCP Small Queues flags
250 * @sk_write_queue: Packet sending queue
251 * @sk_omem_alloc: "o" is "option" or "other"
252 * @sk_wmem_queued: persistent queue size
253 * @sk_forward_alloc: space allocated forward
254 * @sk_napi_id: id of the last napi context to receive data for sk
255 * @sk_ll_usec: usecs to busypoll when there is no data
256 * @sk_allocation: allocation mode
257 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
258 * @sk_pacing_status: Pacing status (requested, handled by sch_fq)
259 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
260 * @sk_sndbuf: size of send buffer in bytes
261 * @__sk_flags_offset: empty field used to determine location of bitfield
262 * @sk_padding: unused element for alignment
263 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
264 * @sk_no_check_rx: allow zero checksum in RX packets
265 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
266 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
267 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
268 * @sk_gso_max_size: Maximum GSO segment size to build
269 * @sk_gso_max_segs: Maximum number of GSO segments
270 * @sk_lingertime: %SO_LINGER l_linger setting
271 * @sk_backlog: always used with the per-socket spinlock held
272 * @sk_callback_lock: used with the callbacks in the end of this struct
273 * @sk_error_queue: rarely used
274 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
275 * IPV6_ADDRFORM for instance)
276 * @sk_err: last error
277 * @sk_err_soft: errors that don't cause failure but are the cause of a
278 * persistent failure not just 'timed out'
279 * @sk_drops: raw/udp drops counter
280 * @sk_ack_backlog: current listen backlog
281 * @sk_max_ack_backlog: listen backlog set in listen()
282 * @sk_uid: user id of owner
283 * @sk_priority: %SO_PRIORITY setting
284 * @sk_type: socket type (%SOCK_STREAM, etc)
285 * @sk_protocol: which protocol this socket belongs in this network family
286 * @sk_peer_pid: &struct pid for this socket's peer
287 * @sk_peer_cred: %SO_PEERCRED setting
288 * @sk_rcvlowat: %SO_RCVLOWAT setting
289 * @sk_rcvtimeo: %SO_RCVTIMEO setting
290 * @sk_sndtimeo: %SO_SNDTIMEO setting
291 * @sk_txhash: computed flow hash for use on transmit
292 * @sk_filter: socket filtering instructions
293 * @sk_timer: sock cleanup timer
294 * @sk_stamp: time stamp of last packet received
295 * @sk_tsflags: SO_TIMESTAMPING socket options
296 * @sk_tskey: counter to disambiguate concurrent tstamp requests
297 * @sk_socket: Identd and reporting IO signals
298 * @sk_user_data: RPC layer private data
299 * @sk_frag: cached page frag
300 * @sk_peek_off: current peek_offset value
301 * @sk_send_head: front of stuff to transmit
302 * @sk_security: used by security modules
303 * @sk_mark: generic packet mark
304 * @sk_cgrp_data: cgroup data for this cgroup
305 * @sk_memcg: this socket's memory cgroup association
306 * @sk_write_pending: a write to stream socket waits to start
307 * @sk_state_change: callback to indicate change in the state of the sock
308 * @sk_data_ready: callback to indicate there is data to be processed
309 * @sk_write_space: callback to indicate there is bf sending space available
310 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
311 * @sk_backlog_rcv: callback to process the backlog
312 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
313 * @sk_reuseport_cb: reuseport group container
314 * @sk_rcu: used during RCU grace period
318 * Now struct inet_timewait_sock also uses sock_common, so please just
319 * don't add nothing before this first member (__sk_common) --acme
321 struct sock_common __sk_common
;
322 #define sk_node __sk_common.skc_node
323 #define sk_nulls_node __sk_common.skc_nulls_node
324 #define sk_refcnt __sk_common.skc_refcnt
325 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
327 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
328 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
329 #define sk_hash __sk_common.skc_hash
330 #define sk_portpair __sk_common.skc_portpair
331 #define sk_num __sk_common.skc_num
332 #define sk_dport __sk_common.skc_dport
333 #define sk_addrpair __sk_common.skc_addrpair
334 #define sk_daddr __sk_common.skc_daddr
335 #define sk_rcv_saddr __sk_common.skc_rcv_saddr
336 #define sk_family __sk_common.skc_family
337 #define sk_state __sk_common.skc_state
338 #define sk_reuse __sk_common.skc_reuse
339 #define sk_reuseport __sk_common.skc_reuseport
340 #define sk_ipv6only __sk_common.skc_ipv6only
341 #define sk_net_refcnt __sk_common.skc_net_refcnt
342 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
343 #define sk_bind_node __sk_common.skc_bind_node
344 #define sk_prot __sk_common.skc_prot
345 #define sk_net __sk_common.skc_net
346 #define sk_v6_daddr __sk_common.skc_v6_daddr
347 #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
348 #define sk_cookie __sk_common.skc_cookie
349 #define sk_incoming_cpu __sk_common.skc_incoming_cpu
350 #define sk_flags __sk_common.skc_flags
351 #define sk_rxhash __sk_common.skc_rxhash
353 socket_lock_t sk_lock
;
356 struct sk_buff_head sk_error_queue
;
357 struct sk_buff_head sk_receive_queue
;
359 * The backlog queue is special, it is always used with
360 * the per-socket spinlock held and requires low latency
361 * access. Therefore we special case it's implementation.
362 * Note : rmem_alloc is in this structure to fill a hole
363 * on 64bit arches, not because its logically part of
369 struct sk_buff
*head
;
370 struct sk_buff
*tail
;
372 #define sk_rmem_alloc sk_backlog.rmem_alloc
374 int sk_forward_alloc
;
375 #ifdef CONFIG_NET_RX_BUSY_POLL
376 unsigned int sk_ll_usec
;
377 /* ===== mostly read cache line ===== */
378 unsigned int sk_napi_id
;
382 struct sk_filter __rcu
*sk_filter
;
384 struct socket_wq __rcu
*sk_wq
;
385 struct socket_wq
*sk_wq_raw
;
388 struct xfrm_policy __rcu
*sk_policy
[2];
390 struct dst_entry
*sk_rx_dst
;
391 struct dst_entry __rcu
*sk_dst_cache
;
392 atomic_t sk_omem_alloc
;
395 /* ===== cache line for TX ===== */
397 refcount_t sk_wmem_alloc
;
398 unsigned long sk_tsq_flags
;
399 struct sk_buff
*sk_send_head
;
400 struct sk_buff_head sk_write_queue
;
402 int sk_write_pending
;
403 __u32 sk_dst_pending_confirm
;
404 u32 sk_pacing_status
; /* see enum sk_pacing */
406 struct timer_list sk_timer
;
409 u32 sk_pacing_rate
; /* bytes per second */
410 u32 sk_max_pacing_rate
;
411 struct page_frag sk_frag
;
412 netdev_features_t sk_route_caps
;
413 netdev_features_t sk_route_nocaps
;
415 unsigned int sk_gso_max_size
;
420 * Because of non atomicity rules, all
421 * changes are protected by socket lock.
423 unsigned int __sk_flags_offset
[0];
424 #ifdef __BIG_ENDIAN_BITFIELD
425 #define SK_FL_PROTO_SHIFT 16
426 #define SK_FL_PROTO_MASK 0x00ff0000
428 #define SK_FL_TYPE_SHIFT 0
429 #define SK_FL_TYPE_MASK 0x0000ffff
431 #define SK_FL_PROTO_SHIFT 8
432 #define SK_FL_PROTO_MASK 0x0000ff00
434 #define SK_FL_TYPE_SHIFT 16
435 #define SK_FL_TYPE_MASK 0xffff0000
438 kmemcheck_bitfield_begin(flags
);
439 unsigned int sk_padding
: 1,
446 #define SK_PROTOCOL_MAX U8_MAX
447 kmemcheck_bitfield_end(flags
);
450 unsigned long sk_lingertime
;
451 struct proto
*sk_prot_creator
;
452 rwlock_t sk_callback_lock
;
456 u32 sk_max_ack_backlog
;
458 struct pid
*sk_peer_pid
;
459 const struct cred
*sk_peer_cred
;
465 struct socket
*sk_socket
;
467 #ifdef CONFIG_SECURITY
470 struct sock_cgroup_data sk_cgrp_data
;
471 struct mem_cgroup
*sk_memcg
;
472 void (*sk_state_change
)(struct sock
*sk
);
473 void (*sk_data_ready
)(struct sock
*sk
);
474 void (*sk_write_space
)(struct sock
*sk
);
475 void (*sk_error_report
)(struct sock
*sk
);
476 int (*sk_backlog_rcv
)(struct sock
*sk
,
477 struct sk_buff
*skb
);
478 void (*sk_destruct
)(struct sock
*sk
);
479 struct sock_reuseport __rcu
*sk_reuseport_cb
;
480 struct rcu_head sk_rcu
;
485 SK_PACING_NEEDED
= 1,
489 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
491 #define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
492 #define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
495 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
496 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
497 * on a socket means that the socket will reuse everybody else's port
498 * without looking at the other's sk_reuse value.
501 #define SK_NO_REUSE 0
502 #define SK_CAN_REUSE 1
503 #define SK_FORCE_REUSE 2
505 int sk_set_peek_off(struct sock
*sk
, int val
);
507 static inline int sk_peek_offset(struct sock
*sk
, int flags
)
509 if (unlikely(flags
& MSG_PEEK
)) {
510 s32 off
= READ_ONCE(sk
->sk_peek_off
);
518 static inline void sk_peek_offset_bwd(struct sock
*sk
, int val
)
520 s32 off
= READ_ONCE(sk
->sk_peek_off
);
522 if (unlikely(off
>= 0)) {
523 off
= max_t(s32
, off
- val
, 0);
524 WRITE_ONCE(sk
->sk_peek_off
, off
);
528 static inline void sk_peek_offset_fwd(struct sock
*sk
, int val
)
530 sk_peek_offset_bwd(sk
, -val
);
534 * Hashed lists helper routines
536 static inline struct sock
*sk_entry(const struct hlist_node
*node
)
538 return hlist_entry(node
, struct sock
, sk_node
);
541 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
543 return hlist_entry(head
->first
, struct sock
, sk_node
);
546 static inline struct sock
*sk_head(const struct hlist_head
*head
)
548 return hlist_empty(head
) ? NULL
: __sk_head(head
);
551 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
553 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
556 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
558 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
561 static inline struct sock
*sk_next(const struct sock
*sk
)
563 return hlist_entry_safe(sk
->sk_node
.next
, struct sock
, sk_node
);
566 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
568 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
569 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
570 struct sock
, sk_nulls_node
) :
574 static inline bool sk_unhashed(const struct sock
*sk
)
576 return hlist_unhashed(&sk
->sk_node
);
579 static inline bool sk_hashed(const struct sock
*sk
)
581 return !sk_unhashed(sk
);
584 static inline void sk_node_init(struct hlist_node
*node
)
589 static inline void sk_nulls_node_init(struct hlist_nulls_node
*node
)
594 static inline void __sk_del_node(struct sock
*sk
)
596 __hlist_del(&sk
->sk_node
);
599 /* NB: equivalent to hlist_del_init_rcu */
600 static inline bool __sk_del_node_init(struct sock
*sk
)
604 sk_node_init(&sk
->sk_node
);
610 /* Grab socket reference count. This operation is valid only
611 when sk is ALREADY grabbed f.e. it is found in hash table
612 or a list and the lookup is made under lock preventing hash table
616 static __always_inline
void sock_hold(struct sock
*sk
)
618 refcount_inc(&sk
->sk_refcnt
);
621 /* Ungrab socket in the context, which assumes that socket refcnt
622 cannot hit zero, f.e. it is true in context of any socketcall.
624 static __always_inline
void __sock_put(struct sock
*sk
)
626 refcount_dec(&sk
->sk_refcnt
);
629 static inline bool sk_del_node_init(struct sock
*sk
)
631 bool rc
= __sk_del_node_init(sk
);
634 /* paranoid for a while -acme */
635 WARN_ON(refcount_read(&sk
->sk_refcnt
) == 1);
640 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
642 static inline bool __sk_nulls_del_node_init_rcu(struct sock
*sk
)
645 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
651 static inline bool sk_nulls_del_node_init_rcu(struct sock
*sk
)
653 bool rc
= __sk_nulls_del_node_init_rcu(sk
);
656 /* paranoid for a while -acme */
657 WARN_ON(refcount_read(&sk
->sk_refcnt
) == 1);
663 static inline void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
665 hlist_add_head(&sk
->sk_node
, list
);
668 static inline void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
671 __sk_add_node(sk
, list
);
674 static inline void sk_add_node_rcu(struct sock
*sk
, struct hlist_head
*list
)
677 if (IS_ENABLED(CONFIG_IPV6
) && sk
->sk_reuseport
&&
678 sk
->sk_family
== AF_INET6
)
679 hlist_add_tail_rcu(&sk
->sk_node
, list
);
681 hlist_add_head_rcu(&sk
->sk_node
, list
);
684 static inline void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
686 if (IS_ENABLED(CONFIG_IPV6
) && sk
->sk_reuseport
&&
687 sk
->sk_family
== AF_INET6
)
688 hlist_nulls_add_tail_rcu(&sk
->sk_nulls_node
, list
);
690 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
693 static inline void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
696 __sk_nulls_add_node_rcu(sk
, list
);
699 static inline void __sk_del_bind_node(struct sock
*sk
)
701 __hlist_del(&sk
->sk_bind_node
);
704 static inline void sk_add_bind_node(struct sock
*sk
,
705 struct hlist_head
*list
)
707 hlist_add_head(&sk
->sk_bind_node
, list
);
710 #define sk_for_each(__sk, list) \
711 hlist_for_each_entry(__sk, list, sk_node)
712 #define sk_for_each_rcu(__sk, list) \
713 hlist_for_each_entry_rcu(__sk, list, sk_node)
714 #define sk_nulls_for_each(__sk, node, list) \
715 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
716 #define sk_nulls_for_each_rcu(__sk, node, list) \
717 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
718 #define sk_for_each_from(__sk) \
719 hlist_for_each_entry_from(__sk, sk_node)
720 #define sk_nulls_for_each_from(__sk, node) \
721 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
722 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
723 #define sk_for_each_safe(__sk, tmp, list) \
724 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
725 #define sk_for_each_bound(__sk, list) \
726 hlist_for_each_entry(__sk, list, sk_bind_node)
729 * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset
730 * @tpos: the type * to use as a loop cursor.
731 * @pos: the &struct hlist_node to use as a loop cursor.
732 * @head: the head for your list.
733 * @offset: offset of hlist_node within the struct.
736 #define sk_for_each_entry_offset_rcu(tpos, pos, head, offset) \
737 for (pos = rcu_dereference((head)->first); \
739 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
740 pos = rcu_dereference(pos->next))
742 static inline struct user_namespace
*sk_user_ns(struct sock
*sk
)
744 /* Careful only use this in a context where these parameters
745 * can not change and must all be valid, such as recvmsg from
748 return sk
->sk_socket
->file
->f_cred
->user_ns
;
762 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
763 SOCK_DBG
, /* %SO_DEBUG setting */
764 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
765 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
766 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
767 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
768 SOCK_MEMALLOC
, /* VM depends on this socket for swapping */
769 SOCK_TIMESTAMPING_RX_SOFTWARE
, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
770 SOCK_FASYNC
, /* fasync() active */
772 SOCK_ZEROCOPY
, /* buffers from userspace */
773 SOCK_WIFI_STATUS
, /* push wifi status to userspace */
774 SOCK_NOFCS
, /* Tell NIC not to do the Ethernet FCS.
775 * Will use last 4 bytes of packet sent from
776 * user-space instead.
778 SOCK_FILTER_LOCKED
, /* Filter cannot be changed anymore */
779 SOCK_SELECT_ERR_QUEUE
, /* Wake select on error queue */
780 SOCK_RCU_FREE
, /* wait rcu grace period in sk_destruct() */
783 #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
785 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
787 nsk
->sk_flags
= osk
->sk_flags
;
790 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
792 __set_bit(flag
, &sk
->sk_flags
);
795 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
797 __clear_bit(flag
, &sk
->sk_flags
);
800 static inline bool sock_flag(const struct sock
*sk
, enum sock_flags flag
)
802 return test_bit(flag
, &sk
->sk_flags
);
806 extern struct static_key memalloc_socks
;
807 static inline int sk_memalloc_socks(void)
809 return static_key_false(&memalloc_socks
);
813 static inline int sk_memalloc_socks(void)
820 static inline gfp_t
sk_gfp_mask(const struct sock
*sk
, gfp_t gfp_mask
)
822 return gfp_mask
| (sk
->sk_allocation
& __GFP_MEMALLOC
);
825 static inline void sk_acceptq_removed(struct sock
*sk
)
827 sk
->sk_ack_backlog
--;
830 static inline void sk_acceptq_added(struct sock
*sk
)
832 sk
->sk_ack_backlog
++;
835 static inline bool sk_acceptq_is_full(const struct sock
*sk
)
837 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
841 * Compute minimal free write space needed to queue new packets.
843 static inline int sk_stream_min_wspace(const struct sock
*sk
)
845 return sk
->sk_wmem_queued
>> 1;
848 static inline int sk_stream_wspace(const struct sock
*sk
)
850 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
853 void sk_stream_write_space(struct sock
*sk
);
855 /* OOB backlog add */
856 static inline void __sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
858 /* dont let skb dst not refcounted, we are going to leave rcu lock */
859 skb_dst_force_safe(skb
);
861 if (!sk
->sk_backlog
.tail
)
862 sk
->sk_backlog
.head
= skb
;
864 sk
->sk_backlog
.tail
->next
= skb
;
866 sk
->sk_backlog
.tail
= skb
;
871 * Take into account size of receive queue and backlog queue
872 * Do not take into account this skb truesize,
873 * to allow even a single big packet to come.
875 static inline bool sk_rcvqueues_full(const struct sock
*sk
, unsigned int limit
)
877 unsigned int qsize
= sk
->sk_backlog
.len
+ atomic_read(&sk
->sk_rmem_alloc
);
879 return qsize
> limit
;
882 /* The per-socket spinlock must be held here. */
883 static inline __must_check
int sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
,
886 if (sk_rcvqueues_full(sk
, limit
))
890 * If the skb was allocated from pfmemalloc reserves, only
891 * allow SOCK_MEMALLOC sockets to use it as this socket is
892 * helping free memory
894 if (skb_pfmemalloc(skb
) && !sock_flag(sk
, SOCK_MEMALLOC
))
897 __sk_add_backlog(sk
, skb
);
898 sk
->sk_backlog
.len
+= skb
->truesize
;
902 int __sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
904 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
906 if (sk_memalloc_socks() && skb_pfmemalloc(skb
))
907 return __sk_backlog_rcv(sk
, skb
);
909 return sk
->sk_backlog_rcv(sk
, skb
);
912 static inline void sk_incoming_cpu_update(struct sock
*sk
)
914 int cpu
= raw_smp_processor_id();
916 if (unlikely(sk
->sk_incoming_cpu
!= cpu
))
917 sk
->sk_incoming_cpu
= cpu
;
920 static inline void sock_rps_record_flow_hash(__u32 hash
)
923 struct rps_sock_flow_table
*sock_flow_table
;
926 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
927 rps_record_sock_flow(sock_flow_table
, hash
);
932 static inline void sock_rps_record_flow(const struct sock
*sk
)
935 if (static_key_false(&rfs_needed
)) {
936 /* Reading sk->sk_rxhash might incur an expensive cache line
939 * TCP_ESTABLISHED does cover almost all states where RFS
940 * might be useful, and is cheaper [1] than testing :
941 * IPv4: inet_sk(sk)->inet_daddr
942 * IPv6: ipv6_addr_any(&sk->sk_v6_daddr)
943 * OR an additional socket flag
944 * [1] : sk_state and sk_prot are in the same cache line.
946 if (sk
->sk_state
== TCP_ESTABLISHED
)
947 sock_rps_record_flow_hash(sk
->sk_rxhash
);
952 static inline void sock_rps_save_rxhash(struct sock
*sk
,
953 const struct sk_buff
*skb
)
956 if (unlikely(sk
->sk_rxhash
!= skb
->hash
))
957 sk
->sk_rxhash
= skb
->hash
;
961 static inline void sock_rps_reset_rxhash(struct sock
*sk
)
968 #define sk_wait_event(__sk, __timeo, __condition, __wait) \
970 release_sock(__sk); \
971 __rc = __condition; \
973 *(__timeo) = wait_woken(__wait, \
974 TASK_INTERRUPTIBLE, \
977 sched_annotate_sleep(); \
979 __rc = __condition; \
983 int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
984 int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
985 void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
986 int sk_stream_error(struct sock
*sk
, int flags
, int err
);
987 void sk_stream_kill_queues(struct sock
*sk
);
988 void sk_set_memalloc(struct sock
*sk
);
989 void sk_clear_memalloc(struct sock
*sk
);
991 void __sk_flush_backlog(struct sock
*sk
);
993 static inline bool sk_flush_backlog(struct sock
*sk
)
995 if (unlikely(READ_ONCE(sk
->sk_backlog
.tail
))) {
996 __sk_flush_backlog(sk
);
1002 int sk_wait_data(struct sock
*sk
, long *timeo
, const struct sk_buff
*skb
);
1004 struct request_sock_ops
;
1005 struct timewait_sock_ops
;
1006 struct inet_hashinfo
;
1007 struct raw_hashinfo
;
1008 struct smc_hashinfo
;
1012 * caches using SLAB_TYPESAFE_BY_RCU should let .next pointer from nulls nodes
1013 * un-modified. Special care is taken when initializing object to zero.
1015 static inline void sk_prot_clear_nulls(struct sock
*sk
, int size
)
1017 if (offsetof(struct sock
, sk_node
.next
) != 0)
1018 memset(sk
, 0, offsetof(struct sock
, sk_node
.next
));
1019 memset(&sk
->sk_node
.pprev
, 0,
1020 size
- offsetof(struct sock
, sk_node
.pprev
));
1023 /* Networking protocol blocks we attach to sockets.
1024 * socket layer -> transport layer interface
1027 void (*close
)(struct sock
*sk
,
1029 int (*connect
)(struct sock
*sk
,
1030 struct sockaddr
*uaddr
,
1032 int (*disconnect
)(struct sock
*sk
, int flags
);
1034 struct sock
* (*accept
)(struct sock
*sk
, int flags
, int *err
,
1037 int (*ioctl
)(struct sock
*sk
, int cmd
,
1039 int (*init
)(struct sock
*sk
);
1040 void (*destroy
)(struct sock
*sk
);
1041 void (*shutdown
)(struct sock
*sk
, int how
);
1042 int (*setsockopt
)(struct sock
*sk
, int level
,
1043 int optname
, char __user
*optval
,
1044 unsigned int optlen
);
1045 int (*getsockopt
)(struct sock
*sk
, int level
,
1046 int optname
, char __user
*optval
,
1047 int __user
*option
);
1048 void (*keepalive
)(struct sock
*sk
, int valbool
);
1049 #ifdef CONFIG_COMPAT
1050 int (*compat_setsockopt
)(struct sock
*sk
,
1052 int optname
, char __user
*optval
,
1053 unsigned int optlen
);
1054 int (*compat_getsockopt
)(struct sock
*sk
,
1056 int optname
, char __user
*optval
,
1057 int __user
*option
);
1058 int (*compat_ioctl
)(struct sock
*sk
,
1059 unsigned int cmd
, unsigned long arg
);
1061 int (*sendmsg
)(struct sock
*sk
, struct msghdr
*msg
,
1063 int (*recvmsg
)(struct sock
*sk
, struct msghdr
*msg
,
1064 size_t len
, int noblock
, int flags
,
1066 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
1067 int offset
, size_t size
, int flags
);
1068 int (*bind
)(struct sock
*sk
,
1069 struct sockaddr
*uaddr
, int addr_len
);
1071 int (*backlog_rcv
) (struct sock
*sk
,
1072 struct sk_buff
*skb
);
1074 void (*release_cb
)(struct sock
*sk
);
1076 /* Keeping track of sk's, looking them up, and port selection methods. */
1077 int (*hash
)(struct sock
*sk
);
1078 void (*unhash
)(struct sock
*sk
);
1079 void (*rehash
)(struct sock
*sk
);
1080 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
1082 /* Keeping track of sockets in use */
1083 #ifdef CONFIG_PROC_FS
1084 unsigned int inuse_idx
;
1087 bool (*stream_memory_free
)(const struct sock
*sk
);
1088 /* Memory pressure */
1089 void (*enter_memory_pressure
)(struct sock
*sk
);
1090 void (*leave_memory_pressure
)(struct sock
*sk
);
1091 atomic_long_t
*memory_allocated
; /* Current allocated memory. */
1092 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
1094 * Pressure flag: try to collapse.
1095 * Technical note: it is used by multiple contexts non atomically.
1096 * All the __sk_mem_schedule() is of this nature: accounting
1097 * is strict, actions are advisory and have some latency.
1099 unsigned long *memory_pressure
;
1106 struct kmem_cache
*slab
;
1107 unsigned int obj_size
;
1110 struct percpu_counter
*orphan_count
;
1112 struct request_sock_ops
*rsk_prot
;
1113 struct timewait_sock_ops
*twsk_prot
;
1116 struct inet_hashinfo
*hashinfo
;
1117 struct udp_table
*udp_table
;
1118 struct raw_hashinfo
*raw_hash
;
1119 struct smc_hashinfo
*smc_hash
;
1122 struct module
*owner
;
1126 struct list_head node
;
1127 #ifdef SOCK_REFCNT_DEBUG
1130 int (*diag_destroy
)(struct sock
*sk
, int err
);
1133 int proto_register(struct proto
*prot
, int alloc_slab
);
1134 void proto_unregister(struct proto
*prot
);
1136 #ifdef SOCK_REFCNT_DEBUG
1137 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
1139 atomic_inc(&sk
->sk_prot
->socks
);
1142 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
1144 atomic_dec(&sk
->sk_prot
->socks
);
1145 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
1146 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
1149 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
1151 if (refcount_read(&sk
->sk_refcnt
) != 1)
1152 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
1153 sk
->sk_prot
->name
, sk
, refcount_read(&sk
->sk_refcnt
));
1155 #else /* SOCK_REFCNT_DEBUG */
1156 #define sk_refcnt_debug_inc(sk) do { } while (0)
1157 #define sk_refcnt_debug_dec(sk) do { } while (0)
1158 #define sk_refcnt_debug_release(sk) do { } while (0)
1159 #endif /* SOCK_REFCNT_DEBUG */
1161 static inline bool sk_stream_memory_free(const struct sock
*sk
)
1163 if (sk
->sk_wmem_queued
>= sk
->sk_sndbuf
)
1166 return sk
->sk_prot
->stream_memory_free
?
1167 sk
->sk_prot
->stream_memory_free(sk
) : true;
1170 static inline bool sk_stream_is_writeable(const struct sock
*sk
)
1172 return sk_stream_wspace(sk
) >= sk_stream_min_wspace(sk
) &&
1173 sk_stream_memory_free(sk
);
1176 static inline int sk_under_cgroup_hierarchy(struct sock
*sk
,
1177 struct cgroup
*ancestor
)
1179 #ifdef CONFIG_SOCK_CGROUP_DATA
1180 return cgroup_is_descendant(sock_cgroup_ptr(&sk
->sk_cgrp_data
),
1187 static inline bool sk_has_memory_pressure(const struct sock
*sk
)
1189 return sk
->sk_prot
->memory_pressure
!= NULL
;
1192 static inline bool sk_under_memory_pressure(const struct sock
*sk
)
1194 if (!sk
->sk_prot
->memory_pressure
)
1197 if (mem_cgroup_sockets_enabled
&& sk
->sk_memcg
&&
1198 mem_cgroup_under_socket_pressure(sk
->sk_memcg
))
1201 return !!*sk
->sk_prot
->memory_pressure
;
1205 sk_memory_allocated(const struct sock
*sk
)
1207 return atomic_long_read(sk
->sk_prot
->memory_allocated
);
1211 sk_memory_allocated_add(struct sock
*sk
, int amt
)
1213 return atomic_long_add_return(amt
, sk
->sk_prot
->memory_allocated
);
1217 sk_memory_allocated_sub(struct sock
*sk
, int amt
)
1219 atomic_long_sub(amt
, sk
->sk_prot
->memory_allocated
);
1222 static inline void sk_sockets_allocated_dec(struct sock
*sk
)
1224 percpu_counter_dec(sk
->sk_prot
->sockets_allocated
);
1227 static inline void sk_sockets_allocated_inc(struct sock
*sk
)
1229 percpu_counter_inc(sk
->sk_prot
->sockets_allocated
);
1233 sk_sockets_allocated_read_positive(struct sock
*sk
)
1235 return percpu_counter_read_positive(sk
->sk_prot
->sockets_allocated
);
1239 proto_sockets_allocated_sum_positive(struct proto
*prot
)
1241 return percpu_counter_sum_positive(prot
->sockets_allocated
);
1245 proto_memory_allocated(struct proto
*prot
)
1247 return atomic_long_read(prot
->memory_allocated
);
1251 proto_memory_pressure(struct proto
*prot
)
1253 if (!prot
->memory_pressure
)
1255 return !!*prot
->memory_pressure
;
1259 #ifdef CONFIG_PROC_FS
1260 /* Called with local bh disabled */
1261 void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int inc
);
1262 int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
1264 static inline void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
1271 /* With per-bucket locks this operation is not-atomic, so that
1272 * this version is not worse.
1274 static inline int __sk_prot_rehash(struct sock
*sk
)
1276 sk
->sk_prot
->unhash(sk
);
1277 return sk
->sk_prot
->hash(sk
);
1280 /* About 10 seconds */
1281 #define SOCK_DESTROY_TIME (10*HZ)
1283 /* Sockets 0-1023 can't be bound to unless you are superuser */
1284 #define PROT_SOCK 1024
1286 #define SHUTDOWN_MASK 3
1287 #define RCV_SHUTDOWN 1
1288 #define SEND_SHUTDOWN 2
1290 #define SOCK_SNDBUF_LOCK 1
1291 #define SOCK_RCVBUF_LOCK 2
1292 #define SOCK_BINDADDR_LOCK 4
1293 #define SOCK_BINDPORT_LOCK 8
1295 struct socket_alloc
{
1296 struct socket socket
;
1297 struct inode vfs_inode
;
1300 static inline struct socket
*SOCKET_I(struct inode
*inode
)
1302 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
1305 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
1307 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
1311 * Functions for memory accounting
1313 int __sk_mem_raise_allocated(struct sock
*sk
, int size
, int amt
, int kind
);
1314 int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
1315 void __sk_mem_reduce_allocated(struct sock
*sk
, int amount
);
1316 void __sk_mem_reclaim(struct sock
*sk
, int amount
);
1318 /* We used to have PAGE_SIZE here, but systems with 64KB pages
1319 * do not necessarily have 16x time more memory than 4KB ones.
1321 #define SK_MEM_QUANTUM 4096
1322 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1323 #define SK_MEM_SEND 0
1324 #define SK_MEM_RECV 1
1326 /* sysctl_mem values are in pages, we convert them in SK_MEM_QUANTUM units */
1327 static inline long sk_prot_mem_limits(const struct sock
*sk
, int index
)
1329 long val
= sk
->sk_prot
->sysctl_mem
[index
];
1331 #if PAGE_SIZE > SK_MEM_QUANTUM
1332 val
<<= PAGE_SHIFT
- SK_MEM_QUANTUM_SHIFT
;
1333 #elif PAGE_SIZE < SK_MEM_QUANTUM
1334 val
>>= SK_MEM_QUANTUM_SHIFT
- PAGE_SHIFT
;
1339 static inline int sk_mem_pages(int amt
)
1341 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
1344 static inline bool sk_has_account(struct sock
*sk
)
1346 /* return true if protocol supports memory accounting */
1347 return !!sk
->sk_prot
->memory_allocated
;
1350 static inline bool sk_wmem_schedule(struct sock
*sk
, int size
)
1352 if (!sk_has_account(sk
))
1354 return size
<= sk
->sk_forward_alloc
||
1355 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
1359 sk_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
, int size
)
1361 if (!sk_has_account(sk
))
1363 return size
<= sk
->sk_forward_alloc
||
1364 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
) ||
1365 skb_pfmemalloc(skb
);
1368 static inline void sk_mem_reclaim(struct sock
*sk
)
1370 if (!sk_has_account(sk
))
1372 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
1373 __sk_mem_reclaim(sk
, sk
->sk_forward_alloc
);
1376 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
1378 if (!sk_has_account(sk
))
1380 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
1381 __sk_mem_reclaim(sk
, sk
->sk_forward_alloc
- 1);
1384 static inline void sk_mem_charge(struct sock
*sk
, int size
)
1386 if (!sk_has_account(sk
))
1388 sk
->sk_forward_alloc
-= size
;
1391 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
1393 if (!sk_has_account(sk
))
1395 sk
->sk_forward_alloc
+= size
;
1397 /* Avoid a possible overflow.
1398 * TCP send queues can make this happen, if sk_mem_reclaim()
1399 * is not called and more than 2 GBytes are released at once.
1401 * If we reach 2 MBytes, reclaim 1 MBytes right now, there is
1402 * no need to hold that much forward allocation anyway.
1404 if (unlikely(sk
->sk_forward_alloc
>= 1 << 21))
1405 __sk_mem_reclaim(sk
, 1 << 20);
1408 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
1410 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
1411 sk
->sk_wmem_queued
-= skb
->truesize
;
1412 sk_mem_uncharge(sk
, skb
->truesize
);
1416 static inline void sock_release_ownership(struct sock
*sk
)
1418 if (sk
->sk_lock
.owned
) {
1419 sk
->sk_lock
.owned
= 0;
1421 /* The sk_lock has mutex_unlock() semantics: */
1422 mutex_release(&sk
->sk_lock
.dep_map
, 1, _RET_IP_
);
1427 * Macro so as to not evaluate some arguments when
1428 * lockdep is not enabled.
1430 * Mark both the sk_lock and the sk_lock.slock as a
1431 * per-address-family lock class.
1433 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1435 sk->sk_lock.owned = 0; \
1436 init_waitqueue_head(&sk->sk_lock.wq); \
1437 spin_lock_init(&(sk)->sk_lock.slock); \
1438 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1439 sizeof((sk)->sk_lock)); \
1440 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1442 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1445 #ifdef CONFIG_LOCKDEP
1446 static inline bool lockdep_sock_is_held(const struct sock
*csk
)
1448 struct sock
*sk
= (struct sock
*)csk
;
1450 return lockdep_is_held(&sk
->sk_lock
) ||
1451 lockdep_is_held(&sk
->sk_lock
.slock
);
1455 void lock_sock_nested(struct sock
*sk
, int subclass
);
1457 static inline void lock_sock(struct sock
*sk
)
1459 lock_sock_nested(sk
, 0);
1462 void release_sock(struct sock
*sk
);
1464 /* BH context may only use the following locking interface. */
1465 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1466 #define bh_lock_sock_nested(__sk) \
1467 spin_lock_nested(&((__sk)->sk_lock.slock), \
1468 SINGLE_DEPTH_NESTING)
1469 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1471 bool lock_sock_fast(struct sock
*sk
);
1473 * unlock_sock_fast - complement of lock_sock_fast
1477 * fast unlock socket for user context.
1478 * If slow mode is on, we call regular release_sock()
1480 static inline void unlock_sock_fast(struct sock
*sk
, bool slow
)
1485 spin_unlock_bh(&sk
->sk_lock
.slock
);
1488 /* Used by processes to "lock" a socket state, so that
1489 * interrupts and bottom half handlers won't change it
1490 * from under us. It essentially blocks any incoming
1491 * packets, so that we won't get any new data or any
1492 * packets that change the state of the socket.
1494 * While locked, BH processing will add new packets to
1495 * the backlog queue. This queue is processed by the
1496 * owner of the socket lock right before it is released.
1498 * Since ~2.3.5 it is also exclusive sleep lock serializing
1499 * accesses from user process context.
1502 static inline void sock_owned_by_me(const struct sock
*sk
)
1504 #ifdef CONFIG_LOCKDEP
1505 WARN_ON_ONCE(!lockdep_sock_is_held(sk
) && debug_locks
);
1509 static inline bool sock_owned_by_user(const struct sock
*sk
)
1511 sock_owned_by_me(sk
);
1512 return sk
->sk_lock
.owned
;
1515 /* no reclassification while locks are held */
1516 static inline bool sock_allow_reclassification(const struct sock
*csk
)
1518 struct sock
*sk
= (struct sock
*)csk
;
1520 return !sk
->sk_lock
.owned
&& !spin_is_locked(&sk
->sk_lock
.slock
);
1523 struct sock
*sk_alloc(struct net
*net
, int family
, gfp_t priority
,
1524 struct proto
*prot
, int kern
);
1525 void sk_free(struct sock
*sk
);
1526 void sk_destruct(struct sock
*sk
);
1527 struct sock
*sk_clone_lock(const struct sock
*sk
, const gfp_t priority
);
1528 void sk_free_unlock_clone(struct sock
*sk
);
1530 struct sk_buff
*sock_wmalloc(struct sock
*sk
, unsigned long size
, int force
,
1532 void __sock_wfree(struct sk_buff
*skb
);
1533 void sock_wfree(struct sk_buff
*skb
);
1534 void skb_orphan_partial(struct sk_buff
*skb
);
1535 void sock_rfree(struct sk_buff
*skb
);
1536 void sock_efree(struct sk_buff
*skb
);
1538 void sock_edemux(struct sk_buff
*skb
);
1540 #define sock_edemux sock_efree
1543 int sock_setsockopt(struct socket
*sock
, int level
, int op
,
1544 char __user
*optval
, unsigned int optlen
);
1546 int sock_getsockopt(struct socket
*sock
, int level
, int op
,
1547 char __user
*optval
, int __user
*optlen
);
1548 struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
, unsigned long size
,
1549 int noblock
, int *errcode
);
1550 struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
, unsigned long header_len
,
1551 unsigned long data_len
, int noblock
,
1552 int *errcode
, int max_page_order
);
1553 void *sock_kmalloc(struct sock
*sk
, int size
, gfp_t priority
);
1554 void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
1555 void sock_kzfree_s(struct sock
*sk
, void *mem
, int size
);
1556 void sk_send_sigurg(struct sock
*sk
);
1558 struct sockcm_cookie
{
1563 int __sock_cmsg_send(struct sock
*sk
, struct msghdr
*msg
, struct cmsghdr
*cmsg
,
1564 struct sockcm_cookie
*sockc
);
1565 int sock_cmsg_send(struct sock
*sk
, struct msghdr
*msg
,
1566 struct sockcm_cookie
*sockc
);
1569 * Functions to fill in entries in struct proto_ops when a protocol
1570 * does not implement a particular function.
1572 int sock_no_bind(struct socket
*, struct sockaddr
*, int);
1573 int sock_no_connect(struct socket
*, struct sockaddr
*, int, int);
1574 int sock_no_socketpair(struct socket
*, struct socket
*);
1575 int sock_no_accept(struct socket
*, struct socket
*, int, bool);
1576 int sock_no_getname(struct socket
*, struct sockaddr
*, int *, int);
1577 unsigned int sock_no_poll(struct file
*, struct socket
*,
1578 struct poll_table_struct
*);
1579 int sock_no_ioctl(struct socket
*, unsigned int, unsigned long);
1580 int sock_no_listen(struct socket
*, int);
1581 int sock_no_shutdown(struct socket
*, int);
1582 int sock_no_getsockopt(struct socket
*, int , int, char __user
*, int __user
*);
1583 int sock_no_setsockopt(struct socket
*, int, int, char __user
*, unsigned int);
1584 int sock_no_sendmsg(struct socket
*, struct msghdr
*, size_t);
1585 int sock_no_recvmsg(struct socket
*, struct msghdr
*, size_t, int);
1586 int sock_no_mmap(struct file
*file
, struct socket
*sock
,
1587 struct vm_area_struct
*vma
);
1588 ssize_t
sock_no_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
1589 size_t size
, int flags
);
1592 * Functions to fill in entries in struct proto_ops when a protocol
1593 * uses the inet style.
1595 int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1596 char __user
*optval
, int __user
*optlen
);
1597 int sock_common_recvmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
,
1599 int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1600 char __user
*optval
, unsigned int optlen
);
1601 int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
1602 int optname
, char __user
*optval
, int __user
*optlen
);
1603 int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
1604 int optname
, char __user
*optval
, unsigned int optlen
);
1606 void sk_common_release(struct sock
*sk
);
1609 * Default socket callbacks and setup code
1612 /* Initialise core socket variables */
1613 void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1616 * Socket reference counting postulates.
1618 * * Each user of socket SHOULD hold a reference count.
1619 * * Each access point to socket (an hash table bucket, reference from a list,
1620 * running timer, skb in flight MUST hold a reference count.
1621 * * When reference count hits 0, it means it will never increase back.
1622 * * When reference count hits 0, it means that no references from
1623 * outside exist to this socket and current process on current CPU
1624 * is last user and may/should destroy this socket.
1625 * * sk_free is called from any context: process, BH, IRQ. When
1626 * it is called, socket has no references from outside -> sk_free
1627 * may release descendant resources allocated by the socket, but
1628 * to the time when it is called, socket is NOT referenced by any
1629 * hash tables, lists etc.
1630 * * Packets, delivered from outside (from network or from another process)
1631 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1632 * when they sit in queue. Otherwise, packets will leak to hole, when
1633 * socket is looked up by one cpu and unhasing is made by another CPU.
1634 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1635 * (leak to backlog). Packet socket does all the processing inside
1636 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1637 * use separate SMP lock, so that they are prone too.
1640 /* Ungrab socket and destroy it, if it was the last reference. */
1641 static inline void sock_put(struct sock
*sk
)
1643 if (refcount_dec_and_test(&sk
->sk_refcnt
))
1646 /* Generic version of sock_put(), dealing with all sockets
1647 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
1649 void sock_gen_put(struct sock
*sk
);
1651 int __sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
, const int nested
,
1652 unsigned int trim_cap
, bool refcounted
);
1653 static inline int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
1656 return __sk_receive_skb(sk
, skb
, nested
, 1, true);
1659 static inline void sk_tx_queue_set(struct sock
*sk
, int tx_queue
)
1661 sk
->sk_tx_queue_mapping
= tx_queue
;
1664 static inline void sk_tx_queue_clear(struct sock
*sk
)
1666 sk
->sk_tx_queue_mapping
= -1;
1669 static inline int sk_tx_queue_get(const struct sock
*sk
)
1671 return sk
? sk
->sk_tx_queue_mapping
: -1;
1674 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1676 sk_tx_queue_clear(sk
);
1677 sk
->sk_socket
= sock
;
1680 static inline wait_queue_head_t
*sk_sleep(struct sock
*sk
)
1682 BUILD_BUG_ON(offsetof(struct socket_wq
, wait
) != 0);
1683 return &rcu_dereference_raw(sk
->sk_wq
)->wait
;
1685 /* Detach socket from process context.
1686 * Announce socket dead, detach it from wait queue and inode.
1687 * Note that parent inode held reference count on this struct sock,
1688 * we do not release it in this function, because protocol
1689 * probably wants some additional cleanups or even continuing
1690 * to work with this socket (TCP).
1692 static inline void sock_orphan(struct sock
*sk
)
1694 write_lock_bh(&sk
->sk_callback_lock
);
1695 sock_set_flag(sk
, SOCK_DEAD
);
1696 sk_set_socket(sk
, NULL
);
1698 write_unlock_bh(&sk
->sk_callback_lock
);
1701 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1703 WARN_ON(parent
->sk
);
1704 write_lock_bh(&sk
->sk_callback_lock
);
1705 sk
->sk_wq
= parent
->wq
;
1707 sk_set_socket(sk
, parent
);
1708 sk
->sk_uid
= SOCK_INODE(parent
)->i_uid
;
1709 security_sock_graft(sk
, parent
);
1710 write_unlock_bh(&sk
->sk_callback_lock
);
1713 kuid_t
sock_i_uid(struct sock
*sk
);
1714 unsigned long sock_i_ino(struct sock
*sk
);
1716 static inline kuid_t
sock_net_uid(const struct net
*net
, const struct sock
*sk
)
1718 return sk
? sk
->sk_uid
: make_kuid(net
->user_ns
, 0);
1721 static inline u32
net_tx_rndhash(void)
1723 u32 v
= prandom_u32();
1728 static inline void sk_set_txhash(struct sock
*sk
)
1730 sk
->sk_txhash
= net_tx_rndhash();
1733 static inline void sk_rethink_txhash(struct sock
*sk
)
1739 static inline struct dst_entry
*
1740 __sk_dst_get(struct sock
*sk
)
1742 return rcu_dereference_check(sk
->sk_dst_cache
,
1743 lockdep_sock_is_held(sk
));
1746 static inline struct dst_entry
*
1747 sk_dst_get(struct sock
*sk
)
1749 struct dst_entry
*dst
;
1752 dst
= rcu_dereference(sk
->sk_dst_cache
);
1753 if (dst
&& !atomic_inc_not_zero(&dst
->__refcnt
))
1759 static inline void dst_negative_advice(struct sock
*sk
)
1761 struct dst_entry
*ndst
, *dst
= __sk_dst_get(sk
);
1763 sk_rethink_txhash(sk
);
1765 if (dst
&& dst
->ops
->negative_advice
) {
1766 ndst
= dst
->ops
->negative_advice(dst
);
1769 rcu_assign_pointer(sk
->sk_dst_cache
, ndst
);
1770 sk_tx_queue_clear(sk
);
1771 sk
->sk_dst_pending_confirm
= 0;
1777 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1779 struct dst_entry
*old_dst
;
1781 sk_tx_queue_clear(sk
);
1782 sk
->sk_dst_pending_confirm
= 0;
1783 old_dst
= rcu_dereference_protected(sk
->sk_dst_cache
,
1784 lockdep_sock_is_held(sk
));
1785 rcu_assign_pointer(sk
->sk_dst_cache
, dst
);
1786 dst_release(old_dst
);
1790 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1792 struct dst_entry
*old_dst
;
1794 sk_tx_queue_clear(sk
);
1795 sk
->sk_dst_pending_confirm
= 0;
1796 old_dst
= xchg((__force
struct dst_entry
**)&sk
->sk_dst_cache
, dst
);
1797 dst_release(old_dst
);
1801 __sk_dst_reset(struct sock
*sk
)
1803 __sk_dst_set(sk
, NULL
);
1807 sk_dst_reset(struct sock
*sk
)
1809 sk_dst_set(sk
, NULL
);
1812 struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1814 struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1816 static inline void sk_dst_confirm(struct sock
*sk
)
1818 if (!sk
->sk_dst_pending_confirm
)
1819 sk
->sk_dst_pending_confirm
= 1;
1822 static inline void sock_confirm_neigh(struct sk_buff
*skb
, struct neighbour
*n
)
1824 if (skb_get_dst_pending_confirm(skb
)) {
1825 struct sock
*sk
= skb
->sk
;
1826 unsigned long now
= jiffies
;
1828 /* avoid dirtying neighbour */
1829 if (n
->confirmed
!= now
)
1831 if (sk
&& sk
->sk_dst_pending_confirm
)
1832 sk
->sk_dst_pending_confirm
= 0;
1836 bool sk_mc_loop(struct sock
*sk
);
1838 static inline bool sk_can_gso(const struct sock
*sk
)
1840 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1843 void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1845 static inline void sk_nocaps_add(struct sock
*sk
, netdev_features_t flags
)
1847 sk
->sk_route_nocaps
|= flags
;
1848 sk
->sk_route_caps
&= ~flags
;
1851 static inline bool sk_check_csum_caps(struct sock
*sk
)
1853 return (sk
->sk_route_caps
& NETIF_F_HW_CSUM
) ||
1854 (sk
->sk_family
== PF_INET
&&
1855 (sk
->sk_route_caps
& NETIF_F_IP_CSUM
)) ||
1856 (sk
->sk_family
== PF_INET6
&&
1857 (sk
->sk_route_caps
& NETIF_F_IPV6_CSUM
));
1860 static inline int skb_do_copy_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1861 struct iov_iter
*from
, char *to
,
1862 int copy
, int offset
)
1864 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1866 if (!csum_and_copy_from_iter_full(to
, copy
, &csum
, from
))
1868 skb
->csum
= csum_block_add(skb
->csum
, csum
, offset
);
1869 } else if (sk
->sk_route_caps
& NETIF_F_NOCACHE_COPY
) {
1870 if (!copy_from_iter_full_nocache(to
, copy
, from
))
1872 } else if (!copy_from_iter_full(to
, copy
, from
))
1878 static inline int skb_add_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1879 struct iov_iter
*from
, int copy
)
1881 int err
, offset
= skb
->len
;
1883 err
= skb_do_copy_data_nocache(sk
, skb
, from
, skb_put(skb
, copy
),
1886 __skb_trim(skb
, offset
);
1891 static inline int skb_copy_to_page_nocache(struct sock
*sk
, struct iov_iter
*from
,
1892 struct sk_buff
*skb
,
1898 err
= skb_do_copy_data_nocache(sk
, skb
, from
, page_address(page
) + off
,
1904 skb
->data_len
+= copy
;
1905 skb
->truesize
+= copy
;
1906 sk
->sk_wmem_queued
+= copy
;
1907 sk_mem_charge(sk
, copy
);
1912 * sk_wmem_alloc_get - returns write allocations
1915 * Returns sk_wmem_alloc minus initial offset of one
1917 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
1919 return refcount_read(&sk
->sk_wmem_alloc
) - 1;
1923 * sk_rmem_alloc_get - returns read allocations
1926 * Returns sk_rmem_alloc
1928 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
1930 return atomic_read(&sk
->sk_rmem_alloc
);
1934 * sk_has_allocations - check if allocations are outstanding
1937 * Returns true if socket has write or read allocations
1939 static inline bool sk_has_allocations(const struct sock
*sk
)
1941 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
1945 * skwq_has_sleeper - check if there are any waiting processes
1946 * @wq: struct socket_wq
1948 * Returns true if socket_wq has waiting processes
1950 * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
1951 * barrier call. They were added due to the race found within the tcp code.
1953 * Consider following tcp code paths::
1956 * sys_select receive packet
1958 * __add_wait_queue update tp->rcv_nxt
1960 * tp->rcv_nxt check sock_def_readable
1962 * schedule rcu_read_lock();
1963 * wq = rcu_dereference(sk->sk_wq);
1964 * if (wq && waitqueue_active(&wq->wait))
1965 * wake_up_interruptible(&wq->wait)
1969 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1970 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1971 * could then endup calling schedule and sleep forever if there are no more
1972 * data on the socket.
1975 static inline bool skwq_has_sleeper(struct socket_wq
*wq
)
1977 return wq
&& wq_has_sleeper(&wq
->wait
);
1981 * sock_poll_wait - place memory barrier behind the poll_wait call.
1983 * @wait_address: socket wait queue
1986 * See the comments in the wq_has_sleeper function.
1988 static inline void sock_poll_wait(struct file
*filp
,
1989 wait_queue_head_t
*wait_address
, poll_table
*p
)
1991 if (!poll_does_not_wait(p
) && wait_address
) {
1992 poll_wait(filp
, wait_address
, p
);
1993 /* We need to be sure we are in sync with the
1994 * socket flags modification.
1996 * This memory barrier is paired in the wq_has_sleeper.
2002 static inline void skb_set_hash_from_sk(struct sk_buff
*skb
, struct sock
*sk
)
2004 if (sk
->sk_txhash
) {
2006 skb
->hash
= sk
->sk_txhash
;
2010 void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
);
2013 * Queue a received datagram if it will fit. Stream and sequenced
2014 * protocols can't normally use this as they need to fit buffers in
2015 * and play with them.
2017 * Inlined as it's very short and called for pretty much every
2018 * packet ever received.
2020 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
2024 skb
->destructor
= sock_rfree
;
2025 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
2026 sk_mem_charge(sk
, skb
->truesize
);
2029 void sk_reset_timer(struct sock
*sk
, struct timer_list
*timer
,
2030 unsigned long expires
);
2032 void sk_stop_timer(struct sock
*sk
, struct timer_list
*timer
);
2034 int __sk_queue_drop_skb(struct sock
*sk
, struct sk_buff_head
*sk_queue
,
2035 struct sk_buff
*skb
, unsigned int flags
,
2036 void (*destructor
)(struct sock
*sk
,
2037 struct sk_buff
*skb
));
2038 int __sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
2039 int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
2041 int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
);
2042 struct sk_buff
*sock_dequeue_err_skb(struct sock
*sk
);
2045 * Recover an error report and clear atomically
2048 static inline int sock_error(struct sock
*sk
)
2051 if (likely(!sk
->sk_err
))
2053 err
= xchg(&sk
->sk_err
, 0);
2057 static inline unsigned long sock_wspace(struct sock
*sk
)
2061 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
2062 amt
= sk
->sk_sndbuf
- refcount_read(&sk
->sk_wmem_alloc
);
2070 * We use sk->sk_wq_raw, from contexts knowing this
2071 * pointer is not NULL and cannot disappear/change.
2073 static inline void sk_set_bit(int nr
, struct sock
*sk
)
2075 if ((nr
== SOCKWQ_ASYNC_NOSPACE
|| nr
== SOCKWQ_ASYNC_WAITDATA
) &&
2076 !sock_flag(sk
, SOCK_FASYNC
))
2079 set_bit(nr
, &sk
->sk_wq_raw
->flags
);
2082 static inline void sk_clear_bit(int nr
, struct sock
*sk
)
2084 if ((nr
== SOCKWQ_ASYNC_NOSPACE
|| nr
== SOCKWQ_ASYNC_WAITDATA
) &&
2085 !sock_flag(sk
, SOCK_FASYNC
))
2088 clear_bit(nr
, &sk
->sk_wq_raw
->flags
);
2091 static inline void sk_wake_async(const struct sock
*sk
, int how
, int band
)
2093 if (sock_flag(sk
, SOCK_FASYNC
)) {
2095 sock_wake_async(rcu_dereference(sk
->sk_wq
), how
, band
);
2100 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2101 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2102 * Note: for send buffers, TCP works better if we can build two skbs at
2105 #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2107 #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2108 #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
2110 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
2112 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
2113 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
2114 sk
->sk_sndbuf
= max_t(u32
, sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
2118 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
,
2119 bool force_schedule
);
2122 * sk_page_frag - return an appropriate page_frag
2125 * If socket allocation mode allows current thread to sleep, it means its
2126 * safe to use the per task page_frag instead of the per socket one.
2128 static inline struct page_frag
*sk_page_frag(struct sock
*sk
)
2130 if (gfpflags_allow_blocking(sk
->sk_allocation
))
2131 return ¤t
->task_frag
;
2133 return &sk
->sk_frag
;
2136 bool sk_page_frag_refill(struct sock
*sk
, struct page_frag
*pfrag
);
2139 * Default write policy as shown to user space via poll/select/SIGIO
2141 static inline bool sock_writeable(const struct sock
*sk
)
2143 return refcount_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
2146 static inline gfp_t
gfp_any(void)
2148 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
2151 static inline long sock_rcvtimeo(const struct sock
*sk
, bool noblock
)
2153 return noblock
? 0 : sk
->sk_rcvtimeo
;
2156 static inline long sock_sndtimeo(const struct sock
*sk
, bool noblock
)
2158 return noblock
? 0 : sk
->sk_sndtimeo
;
2161 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
2163 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
2166 /* Alas, with timeout socket operations are not restartable.
2167 * Compare this to poll().
2169 static inline int sock_intr_errno(long timeo
)
2171 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
2174 struct sock_skb_cb
{
2178 /* Store sock_skb_cb at the end of skb->cb[] so protocol families
2179 * using skb->cb[] would keep using it directly and utilize its
2180 * alignement guarantee.
2182 #define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \
2183 sizeof(struct sock_skb_cb)))
2185 #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2186 SOCK_SKB_CB_OFFSET))
2188 #define sock_skb_cb_check_size(size) \
2189 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
2192 sock_skb_set_dropcount(const struct sock
*sk
, struct sk_buff
*skb
)
2194 SOCK_SKB_CB(skb
)->dropcount
= sock_flag(sk
, SOCK_RXQ_OVFL
) ?
2195 atomic_read(&sk
->sk_drops
) : 0;
2198 static inline void sk_drops_add(struct sock
*sk
, const struct sk_buff
*skb
)
2200 int segs
= max_t(u16
, 1, skb_shinfo(skb
)->gso_segs
);
2202 atomic_add(segs
, &sk
->sk_drops
);
2205 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
2206 struct sk_buff
*skb
);
2207 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
2208 struct sk_buff
*skb
);
2211 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
2213 ktime_t kt
= skb
->tstamp
;
2214 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
2217 * generate control messages if
2218 * - receive time stamping in software requested
2219 * - software time stamp available and wanted
2220 * - hardware time stamps available and wanted
2222 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
2223 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RX_SOFTWARE
) ||
2224 (kt
&& sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
) ||
2225 (hwtstamps
->hwtstamp
&&
2226 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
)))
2227 __sock_recv_timestamp(msg
, sk
, skb
);
2231 if (sock_flag(sk
, SOCK_WIFI_STATUS
) && skb
->wifi_acked_valid
)
2232 __sock_recv_wifi_status(msg
, sk
, skb
);
2235 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2236 struct sk_buff
*skb
);
2238 #define SK_DEFAULT_STAMP (-1L * NSEC_PER_SEC)
2239 static inline void sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2240 struct sk_buff
*skb
)
2242 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2243 (1UL << SOCK_RCVTSTAMP))
2244 #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2245 SOF_TIMESTAMPING_RAW_HARDWARE)
2247 if (sk
->sk_flags
& FLAGS_TS_OR_DROPS
|| sk
->sk_tsflags
& TSFLAGS_ANY
)
2248 __sock_recv_ts_and_drops(msg
, sk
, skb
);
2249 else if (unlikely(sock_flag(sk
, SOCK_TIMESTAMP
)))
2250 sk
->sk_stamp
= skb
->tstamp
;
2251 else if (unlikely(sk
->sk_stamp
== SK_DEFAULT_STAMP
))
2255 void __sock_tx_timestamp(__u16 tsflags
, __u8
*tx_flags
);
2258 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2259 * @sk: socket sending this packet
2260 * @tsflags: timestamping flags to use
2261 * @tx_flags: completed with instructions for time stamping
2263 * Note: callers should take care of initial ``*tx_flags`` value (usually 0)
2265 static inline void sock_tx_timestamp(const struct sock
*sk
, __u16 tsflags
,
2268 if (unlikely(tsflags
))
2269 __sock_tx_timestamp(tsflags
, tx_flags
);
2270 if (unlikely(sock_flag(sk
, SOCK_WIFI_STATUS
)))
2271 *tx_flags
|= SKBTX_WIFI_STATUS
;
2275 * sk_eat_skb - Release a skb if it is no longer needed
2276 * @sk: socket to eat this skb from
2277 * @skb: socket buffer to eat
2279 * This routine must be called with interrupts disabled or with the socket
2280 * locked so that the sk_buff queue operation is ok.
2282 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
)
2284 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2289 struct net
*sock_net(const struct sock
*sk
)
2291 return read_pnet(&sk
->sk_net
);
2295 void sock_net_set(struct sock
*sk
, struct net
*net
)
2297 write_pnet(&sk
->sk_net
, net
);
2300 static inline struct sock
*skb_steal_sock(struct sk_buff
*skb
)
2303 struct sock
*sk
= skb
->sk
;
2305 skb
->destructor
= NULL
;
2312 /* This helper checks if a socket is a full socket,
2313 * ie _not_ a timewait or request socket.
2315 static inline bool sk_fullsock(const struct sock
*sk
)
2317 return (1 << sk
->sk_state
) & ~(TCPF_TIME_WAIT
| TCPF_NEW_SYN_RECV
);
2320 /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2321 * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2323 static inline bool sk_listener(const struct sock
*sk
)
2325 return (1 << sk
->sk_state
) & (TCPF_LISTEN
| TCPF_NEW_SYN_RECV
);
2329 * sk_state_load - read sk->sk_state for lockless contexts
2330 * @sk: socket pointer
2332 * Paired with sk_state_store(). Used in places we do not hold socket lock :
2333 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
2335 static inline int sk_state_load(const struct sock
*sk
)
2337 return smp_load_acquire(&sk
->sk_state
);
2341 * sk_state_store - update sk->sk_state
2342 * @sk: socket pointer
2343 * @newstate: new state
2345 * Paired with sk_state_load(). Should be used in contexts where
2346 * state change might impact lockless readers.
2348 static inline void sk_state_store(struct sock
*sk
, int newstate
)
2350 smp_store_release(&sk
->sk_state
, newstate
);
2353 void sock_enable_timestamp(struct sock
*sk
, int flag
);
2354 int sock_get_timestamp(struct sock
*, struct timeval __user
*);
2355 int sock_get_timestampns(struct sock
*, struct timespec __user
*);
2356 int sock_recv_errqueue(struct sock
*sk
, struct msghdr
*msg
, int len
, int level
,
2359 bool sk_ns_capable(const struct sock
*sk
,
2360 struct user_namespace
*user_ns
, int cap
);
2361 bool sk_capable(const struct sock
*sk
, int cap
);
2362 bool sk_net_capable(const struct sock
*sk
, int cap
);
2364 void sk_get_meminfo(const struct sock
*sk
, u32
*meminfo
);
2366 extern __u32 sysctl_wmem_max
;
2367 extern __u32 sysctl_rmem_max
;
2369 extern int sysctl_tstamp_allow_data
;
2370 extern int sysctl_optmem_max
;
2372 extern __u32 sysctl_wmem_default
;
2373 extern __u32 sysctl_rmem_default
;
2375 #endif /* _SOCK_H */