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>
63 #include <linux/rbtree.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_pacing_shift: scaling factor for TCP Small Queues
271 * @sk_lingertime: %SO_LINGER l_linger setting
272 * @sk_backlog: always used with the per-socket spinlock held
273 * @sk_callback_lock: used with the callbacks in the end of this struct
274 * @sk_error_queue: rarely used
275 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
276 * IPV6_ADDRFORM for instance)
277 * @sk_err: last error
278 * @sk_err_soft: errors that don't cause failure but are the cause of a
279 * persistent failure not just 'timed out'
280 * @sk_drops: raw/udp drops counter
281 * @sk_ack_backlog: current listen backlog
282 * @sk_max_ack_backlog: listen backlog set in listen()
283 * @sk_uid: user id of owner
284 * @sk_priority: %SO_PRIORITY setting
285 * @sk_type: socket type (%SOCK_STREAM, etc)
286 * @sk_protocol: which protocol this socket belongs in this network family
287 * @sk_peer_pid: &struct pid for this socket's peer
288 * @sk_peer_cred: %SO_PEERCRED setting
289 * @sk_rcvlowat: %SO_RCVLOWAT setting
290 * @sk_rcvtimeo: %SO_RCVTIMEO setting
291 * @sk_sndtimeo: %SO_SNDTIMEO setting
292 * @sk_txhash: computed flow hash for use on transmit
293 * @sk_filter: socket filtering instructions
294 * @sk_timer: sock cleanup timer
295 * @sk_stamp: time stamp of last packet received
296 * @sk_stamp_seq: lock for accessing sk_stamp on 32 bit architectures only
297 * @sk_tsflags: SO_TIMESTAMPING socket options
298 * @sk_tskey: counter to disambiguate concurrent tstamp requests
299 * @sk_zckey: counter to order MSG_ZEROCOPY notifications
300 * @sk_socket: Identd and reporting IO signals
301 * @sk_user_data: RPC layer private data
302 * @sk_frag: cached page frag
303 * @sk_peek_off: current peek_offset value
304 * @sk_send_head: front of stuff to transmit
305 * @sk_security: used by security modules
306 * @sk_mark: generic packet mark
307 * @sk_cgrp_data: cgroup data for this cgroup
308 * @sk_memcg: this socket's memory cgroup association
309 * @sk_write_pending: a write to stream socket waits to start
310 * @sk_state_change: callback to indicate change in the state of the sock
311 * @sk_data_ready: callback to indicate there is data to be processed
312 * @sk_write_space: callback to indicate there is bf sending space available
313 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
314 * @sk_backlog_rcv: callback to process the backlog
315 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
316 * @sk_reuseport_cb: reuseport group container
317 * @sk_rcu: used during RCU grace period
321 * Now struct inet_timewait_sock also uses sock_common, so please just
322 * don't add nothing before this first member (__sk_common) --acme
324 struct sock_common __sk_common
;
325 #define sk_node __sk_common.skc_node
326 #define sk_nulls_node __sk_common.skc_nulls_node
327 #define sk_refcnt __sk_common.skc_refcnt
328 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
330 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
331 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
332 #define sk_hash __sk_common.skc_hash
333 #define sk_portpair __sk_common.skc_portpair
334 #define sk_num __sk_common.skc_num
335 #define sk_dport __sk_common.skc_dport
336 #define sk_addrpair __sk_common.skc_addrpair
337 #define sk_daddr __sk_common.skc_daddr
338 #define sk_rcv_saddr __sk_common.skc_rcv_saddr
339 #define sk_family __sk_common.skc_family
340 #define sk_state __sk_common.skc_state
341 #define sk_reuse __sk_common.skc_reuse
342 #define sk_reuseport __sk_common.skc_reuseport
343 #define sk_ipv6only __sk_common.skc_ipv6only
344 #define sk_net_refcnt __sk_common.skc_net_refcnt
345 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
346 #define sk_bind_node __sk_common.skc_bind_node
347 #define sk_prot __sk_common.skc_prot
348 #define sk_net __sk_common.skc_net
349 #define sk_v6_daddr __sk_common.skc_v6_daddr
350 #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
351 #define sk_cookie __sk_common.skc_cookie
352 #define sk_incoming_cpu __sk_common.skc_incoming_cpu
353 #define sk_flags __sk_common.skc_flags
354 #define sk_rxhash __sk_common.skc_rxhash
356 socket_lock_t sk_lock
;
359 struct sk_buff_head sk_error_queue
;
360 struct sk_buff_head sk_receive_queue
;
362 * The backlog queue is special, it is always used with
363 * the per-socket spinlock held and requires low latency
364 * access. Therefore we special case it's implementation.
365 * Note : rmem_alloc is in this structure to fill a hole
366 * on 64bit arches, not because its logically part of
372 struct sk_buff
*head
;
373 struct sk_buff
*tail
;
375 #define sk_rmem_alloc sk_backlog.rmem_alloc
377 int sk_forward_alloc
;
378 #ifdef CONFIG_NET_RX_BUSY_POLL
379 unsigned int sk_ll_usec
;
380 /* ===== mostly read cache line ===== */
381 unsigned int sk_napi_id
;
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 atomic_t sk_omem_alloc
;
398 /* ===== cache line for TX ===== */
400 refcount_t sk_wmem_alloc
;
401 unsigned long sk_tsq_flags
;
403 struct sk_buff
*sk_send_head
;
404 struct rb_root tcp_rtx_queue
;
406 struct sk_buff_head sk_write_queue
;
408 int sk_write_pending
;
409 __u32 sk_dst_pending_confirm
;
410 u32 sk_pacing_status
; /* see enum sk_pacing */
412 struct timer_list sk_timer
;
415 u32 sk_pacing_rate
; /* bytes per second */
416 u32 sk_max_pacing_rate
;
417 struct page_frag sk_frag
;
418 netdev_features_t sk_route_caps
;
419 netdev_features_t sk_route_nocaps
;
421 unsigned int sk_gso_max_size
;
426 * Because of non atomicity rules, all
427 * changes are protected by socket lock.
429 unsigned int __sk_flags_offset
[0];
430 #ifdef __BIG_ENDIAN_BITFIELD
431 #define SK_FL_PROTO_SHIFT 16
432 #define SK_FL_PROTO_MASK 0x00ff0000
434 #define SK_FL_TYPE_SHIFT 0
435 #define SK_FL_TYPE_MASK 0x0000ffff
437 #define SK_FL_PROTO_SHIFT 8
438 #define SK_FL_PROTO_MASK 0x0000ff00
440 #define SK_FL_TYPE_SHIFT 16
441 #define SK_FL_TYPE_MASK 0xffff0000
444 unsigned int sk_padding
: 1,
451 #define SK_PROTOCOL_MAX U8_MAX
454 unsigned long sk_lingertime
;
455 struct proto
*sk_prot_creator
;
456 rwlock_t sk_callback_lock
;
460 u32 sk_max_ack_backlog
;
462 struct pid
*sk_peer_pid
;
463 const struct cred
*sk_peer_cred
;
466 #if BITS_PER_LONG==32
467 seqlock_t sk_stamp_seq
;
473 struct socket
*sk_socket
;
475 #ifdef CONFIG_SECURITY
478 struct sock_cgroup_data sk_cgrp_data
;
479 struct mem_cgroup
*sk_memcg
;
480 void (*sk_state_change
)(struct sock
*sk
);
481 void (*sk_data_ready
)(struct sock
*sk
);
482 void (*sk_write_space
)(struct sock
*sk
);
483 void (*sk_error_report
)(struct sock
*sk
);
484 int (*sk_backlog_rcv
)(struct sock
*sk
,
485 struct sk_buff
*skb
);
486 void (*sk_destruct
)(struct sock
*sk
);
487 struct sock_reuseport __rcu
*sk_reuseport_cb
;
488 struct rcu_head sk_rcu
;
493 SK_PACING_NEEDED
= 1,
497 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
499 #define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
500 #define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
503 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
504 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
505 * on a socket means that the socket will reuse everybody else's port
506 * without looking at the other's sk_reuse value.
509 #define SK_NO_REUSE 0
510 #define SK_CAN_REUSE 1
511 #define SK_FORCE_REUSE 2
513 int sk_set_peek_off(struct sock
*sk
, int val
);
515 static inline int sk_peek_offset(struct sock
*sk
, int flags
)
517 if (unlikely(flags
& MSG_PEEK
)) {
518 return READ_ONCE(sk
->sk_peek_off
);
524 static inline void sk_peek_offset_bwd(struct sock
*sk
, int val
)
526 s32 off
= READ_ONCE(sk
->sk_peek_off
);
528 if (unlikely(off
>= 0)) {
529 off
= max_t(s32
, off
- val
, 0);
530 WRITE_ONCE(sk
->sk_peek_off
, off
);
534 static inline void sk_peek_offset_fwd(struct sock
*sk
, int val
)
536 sk_peek_offset_bwd(sk
, -val
);
540 * Hashed lists helper routines
542 static inline struct sock
*sk_entry(const struct hlist_node
*node
)
544 return hlist_entry(node
, struct sock
, sk_node
);
547 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
549 return hlist_entry(head
->first
, struct sock
, sk_node
);
552 static inline struct sock
*sk_head(const struct hlist_head
*head
)
554 return hlist_empty(head
) ? NULL
: __sk_head(head
);
557 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
559 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
562 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
564 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
567 static inline struct sock
*sk_next(const struct sock
*sk
)
569 return hlist_entry_safe(sk
->sk_node
.next
, struct sock
, sk_node
);
572 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
574 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
575 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
576 struct sock
, sk_nulls_node
) :
580 static inline bool sk_unhashed(const struct sock
*sk
)
582 return hlist_unhashed(&sk
->sk_node
);
585 static inline bool sk_hashed(const struct sock
*sk
)
587 return !sk_unhashed(sk
);
590 static inline void sk_node_init(struct hlist_node
*node
)
595 static inline void sk_nulls_node_init(struct hlist_nulls_node
*node
)
600 static inline void __sk_del_node(struct sock
*sk
)
602 __hlist_del(&sk
->sk_node
);
605 /* NB: equivalent to hlist_del_init_rcu */
606 static inline bool __sk_del_node_init(struct sock
*sk
)
610 sk_node_init(&sk
->sk_node
);
616 /* Grab socket reference count. This operation is valid only
617 when sk is ALREADY grabbed f.e. it is found in hash table
618 or a list and the lookup is made under lock preventing hash table
622 static __always_inline
void sock_hold(struct sock
*sk
)
624 refcount_inc(&sk
->sk_refcnt
);
627 /* Ungrab socket in the context, which assumes that socket refcnt
628 cannot hit zero, f.e. it is true in context of any socketcall.
630 static __always_inline
void __sock_put(struct sock
*sk
)
632 refcount_dec(&sk
->sk_refcnt
);
635 static inline bool sk_del_node_init(struct sock
*sk
)
637 bool rc
= __sk_del_node_init(sk
);
640 /* paranoid for a while -acme */
641 WARN_ON(refcount_read(&sk
->sk_refcnt
) == 1);
646 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
648 static inline bool __sk_nulls_del_node_init_rcu(struct sock
*sk
)
651 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
657 static inline bool sk_nulls_del_node_init_rcu(struct sock
*sk
)
659 bool rc
= __sk_nulls_del_node_init_rcu(sk
);
662 /* paranoid for a while -acme */
663 WARN_ON(refcount_read(&sk
->sk_refcnt
) == 1);
669 static inline void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
671 hlist_add_head(&sk
->sk_node
, list
);
674 static inline void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
677 __sk_add_node(sk
, list
);
680 static inline void sk_add_node_rcu(struct sock
*sk
, struct hlist_head
*list
)
683 if (IS_ENABLED(CONFIG_IPV6
) && sk
->sk_reuseport
&&
684 sk
->sk_family
== AF_INET6
)
685 hlist_add_tail_rcu(&sk
->sk_node
, list
);
687 hlist_add_head_rcu(&sk
->sk_node
, list
);
690 static inline void sk_add_node_tail_rcu(struct sock
*sk
, struct hlist_head
*list
)
693 hlist_add_tail_rcu(&sk
->sk_node
, list
);
696 static inline void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
698 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
701 static inline void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
704 __sk_nulls_add_node_rcu(sk
, list
);
707 static inline void __sk_del_bind_node(struct sock
*sk
)
709 __hlist_del(&sk
->sk_bind_node
);
712 static inline void sk_add_bind_node(struct sock
*sk
,
713 struct hlist_head
*list
)
715 hlist_add_head(&sk
->sk_bind_node
, list
);
718 #define sk_for_each(__sk, list) \
719 hlist_for_each_entry(__sk, list, sk_node)
720 #define sk_for_each_rcu(__sk, list) \
721 hlist_for_each_entry_rcu(__sk, list, sk_node)
722 #define sk_nulls_for_each(__sk, node, list) \
723 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
724 #define sk_nulls_for_each_rcu(__sk, node, list) \
725 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
726 #define sk_for_each_from(__sk) \
727 hlist_for_each_entry_from(__sk, sk_node)
728 #define sk_nulls_for_each_from(__sk, node) \
729 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
730 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
731 #define sk_for_each_safe(__sk, tmp, list) \
732 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
733 #define sk_for_each_bound(__sk, list) \
734 hlist_for_each_entry(__sk, list, sk_bind_node)
737 * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset
738 * @tpos: the type * to use as a loop cursor.
739 * @pos: the &struct hlist_node to use as a loop cursor.
740 * @head: the head for your list.
741 * @offset: offset of hlist_node within the struct.
744 #define sk_for_each_entry_offset_rcu(tpos, pos, head, offset) \
745 for (pos = rcu_dereference(hlist_first_rcu(head)); \
747 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
748 pos = rcu_dereference(hlist_next_rcu(pos)))
750 static inline struct user_namespace
*sk_user_ns(struct sock
*sk
)
752 /* Careful only use this in a context where these parameters
753 * can not change and must all be valid, such as recvmsg from
756 return sk
->sk_socket
->file
->f_cred
->user_ns
;
770 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
771 SOCK_DBG
, /* %SO_DEBUG setting */
772 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
773 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
774 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
775 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
776 SOCK_MEMALLOC
, /* VM depends on this socket for swapping */
777 SOCK_TIMESTAMPING_RX_SOFTWARE
, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
778 SOCK_FASYNC
, /* fasync() active */
780 SOCK_ZEROCOPY
, /* buffers from userspace */
781 SOCK_WIFI_STATUS
, /* push wifi status to userspace */
782 SOCK_NOFCS
, /* Tell NIC not to do the Ethernet FCS.
783 * Will use last 4 bytes of packet sent from
784 * user-space instead.
786 SOCK_FILTER_LOCKED
, /* Filter cannot be changed anymore */
787 SOCK_SELECT_ERR_QUEUE
, /* Wake select on error queue */
788 SOCK_RCU_FREE
, /* wait rcu grace period in sk_destruct() */
791 #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
793 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
795 nsk
->sk_flags
= osk
->sk_flags
;
798 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
800 __set_bit(flag
, &sk
->sk_flags
);
803 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
805 __clear_bit(flag
, &sk
->sk_flags
);
808 static inline bool sock_flag(const struct sock
*sk
, enum sock_flags flag
)
810 return test_bit(flag
, &sk
->sk_flags
);
814 extern struct static_key memalloc_socks
;
815 static inline int sk_memalloc_socks(void)
817 return static_key_false(&memalloc_socks
);
821 static inline int sk_memalloc_socks(void)
828 static inline gfp_t
sk_gfp_mask(const struct sock
*sk
, gfp_t gfp_mask
)
830 return gfp_mask
| (sk
->sk_allocation
& __GFP_MEMALLOC
);
833 static inline void sk_acceptq_removed(struct sock
*sk
)
835 sk
->sk_ack_backlog
--;
838 static inline void sk_acceptq_added(struct sock
*sk
)
840 sk
->sk_ack_backlog
++;
843 static inline bool sk_acceptq_is_full(const struct sock
*sk
)
845 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
849 * Compute minimal free write space needed to queue new packets.
851 static inline int sk_stream_min_wspace(const struct sock
*sk
)
853 return sk
->sk_wmem_queued
>> 1;
856 static inline int sk_stream_wspace(const struct sock
*sk
)
858 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
861 void sk_stream_write_space(struct sock
*sk
);
863 /* OOB backlog add */
864 static inline void __sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
866 /* dont let skb dst not refcounted, we are going to leave rcu lock */
869 if (!sk
->sk_backlog
.tail
)
870 sk
->sk_backlog
.head
= skb
;
872 sk
->sk_backlog
.tail
->next
= skb
;
874 sk
->sk_backlog
.tail
= skb
;
879 * Take into account size of receive queue and backlog queue
880 * Do not take into account this skb truesize,
881 * to allow even a single big packet to come.
883 static inline bool sk_rcvqueues_full(const struct sock
*sk
, unsigned int limit
)
885 unsigned int qsize
= sk
->sk_backlog
.len
+ atomic_read(&sk
->sk_rmem_alloc
);
887 return qsize
> limit
;
890 /* The per-socket spinlock must be held here. */
891 static inline __must_check
int sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
,
894 if (sk_rcvqueues_full(sk
, limit
))
898 * If the skb was allocated from pfmemalloc reserves, only
899 * allow SOCK_MEMALLOC sockets to use it as this socket is
900 * helping free memory
902 if (skb_pfmemalloc(skb
) && !sock_flag(sk
, SOCK_MEMALLOC
))
905 __sk_add_backlog(sk
, skb
);
906 sk
->sk_backlog
.len
+= skb
->truesize
;
910 int __sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
912 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
914 if (sk_memalloc_socks() && skb_pfmemalloc(skb
))
915 return __sk_backlog_rcv(sk
, skb
);
917 return sk
->sk_backlog_rcv(sk
, skb
);
920 static inline void sk_incoming_cpu_update(struct sock
*sk
)
922 int cpu
= raw_smp_processor_id();
924 if (unlikely(READ_ONCE(sk
->sk_incoming_cpu
) != cpu
))
925 WRITE_ONCE(sk
->sk_incoming_cpu
, cpu
);
928 static inline void sock_rps_record_flow_hash(__u32 hash
)
931 struct rps_sock_flow_table
*sock_flow_table
;
934 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
935 rps_record_sock_flow(sock_flow_table
, hash
);
940 static inline void sock_rps_record_flow(const struct sock
*sk
)
943 if (static_key_false(&rfs_needed
)) {
944 /* Reading sk->sk_rxhash might incur an expensive cache line
947 * TCP_ESTABLISHED does cover almost all states where RFS
948 * might be useful, and is cheaper [1] than testing :
949 * IPv4: inet_sk(sk)->inet_daddr
950 * IPv6: ipv6_addr_any(&sk->sk_v6_daddr)
951 * OR an additional socket flag
952 * [1] : sk_state and sk_prot are in the same cache line.
954 if (sk
->sk_state
== TCP_ESTABLISHED
)
955 sock_rps_record_flow_hash(sk
->sk_rxhash
);
960 static inline void sock_rps_save_rxhash(struct sock
*sk
,
961 const struct sk_buff
*skb
)
964 if (unlikely(sk
->sk_rxhash
!= skb
->hash
))
965 sk
->sk_rxhash
= skb
->hash
;
969 static inline void sock_rps_reset_rxhash(struct sock
*sk
)
976 #define sk_wait_event(__sk, __timeo, __condition, __wait) \
978 release_sock(__sk); \
979 __rc = __condition; \
981 *(__timeo) = wait_woken(__wait, \
982 TASK_INTERRUPTIBLE, \
985 sched_annotate_sleep(); \
987 __rc = __condition; \
991 int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
992 int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
993 void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
994 int sk_stream_error(struct sock
*sk
, int flags
, int err
);
995 void sk_stream_kill_queues(struct sock
*sk
);
996 void sk_set_memalloc(struct sock
*sk
);
997 void sk_clear_memalloc(struct sock
*sk
);
999 void __sk_flush_backlog(struct sock
*sk
);
1001 static inline bool sk_flush_backlog(struct sock
*sk
)
1003 if (unlikely(READ_ONCE(sk
->sk_backlog
.tail
))) {
1004 __sk_flush_backlog(sk
);
1010 int sk_wait_data(struct sock
*sk
, long *timeo
, const struct sk_buff
*skb
);
1012 struct request_sock_ops
;
1013 struct timewait_sock_ops
;
1014 struct inet_hashinfo
;
1015 struct raw_hashinfo
;
1016 struct smc_hashinfo
;
1020 * caches using SLAB_TYPESAFE_BY_RCU should let .next pointer from nulls nodes
1021 * un-modified. Special care is taken when initializing object to zero.
1023 static inline void sk_prot_clear_nulls(struct sock
*sk
, int size
)
1025 if (offsetof(struct sock
, sk_node
.next
) != 0)
1026 memset(sk
, 0, offsetof(struct sock
, sk_node
.next
));
1027 memset(&sk
->sk_node
.pprev
, 0,
1028 size
- offsetof(struct sock
, sk_node
.pprev
));
1031 /* Networking protocol blocks we attach to sockets.
1032 * socket layer -> transport layer interface
1035 void (*close
)(struct sock
*sk
,
1037 int (*connect
)(struct sock
*sk
,
1038 struct sockaddr
*uaddr
,
1040 int (*disconnect
)(struct sock
*sk
, int flags
);
1042 struct sock
* (*accept
)(struct sock
*sk
, int flags
, int *err
,
1045 int (*ioctl
)(struct sock
*sk
, int cmd
,
1047 int (*init
)(struct sock
*sk
);
1048 void (*destroy
)(struct sock
*sk
);
1049 void (*shutdown
)(struct sock
*sk
, int how
);
1050 int (*setsockopt
)(struct sock
*sk
, int level
,
1051 int optname
, char __user
*optval
,
1052 unsigned int optlen
);
1053 int (*getsockopt
)(struct sock
*sk
, int level
,
1054 int optname
, char __user
*optval
,
1055 int __user
*option
);
1056 void (*keepalive
)(struct sock
*sk
, int valbool
);
1057 #ifdef CONFIG_COMPAT
1058 int (*compat_setsockopt
)(struct sock
*sk
,
1060 int optname
, char __user
*optval
,
1061 unsigned int optlen
);
1062 int (*compat_getsockopt
)(struct sock
*sk
,
1064 int optname
, char __user
*optval
,
1065 int __user
*option
);
1066 int (*compat_ioctl
)(struct sock
*sk
,
1067 unsigned int cmd
, unsigned long arg
);
1069 int (*sendmsg
)(struct sock
*sk
, struct msghdr
*msg
,
1071 int (*recvmsg
)(struct sock
*sk
, struct msghdr
*msg
,
1072 size_t len
, int noblock
, int flags
,
1074 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
1075 int offset
, size_t size
, int flags
);
1076 int (*bind
)(struct sock
*sk
,
1077 struct sockaddr
*uaddr
, int addr_len
);
1079 int (*backlog_rcv
) (struct sock
*sk
,
1080 struct sk_buff
*skb
);
1082 void (*release_cb
)(struct sock
*sk
);
1084 /* Keeping track of sk's, looking them up, and port selection methods. */
1085 int (*hash
)(struct sock
*sk
);
1086 void (*unhash
)(struct sock
*sk
);
1087 void (*rehash
)(struct sock
*sk
);
1088 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
1090 /* Keeping track of sockets in use */
1091 #ifdef CONFIG_PROC_FS
1092 unsigned int inuse_idx
;
1095 bool (*stream_memory_free
)(const struct sock
*sk
);
1096 /* Memory pressure */
1097 void (*enter_memory_pressure
)(struct sock
*sk
);
1098 void (*leave_memory_pressure
)(struct sock
*sk
);
1099 atomic_long_t
*memory_allocated
; /* Current allocated memory. */
1100 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
1102 * Pressure flag: try to collapse.
1103 * Technical note: it is used by multiple contexts non atomically.
1104 * All the __sk_mem_schedule() is of this nature: accounting
1105 * is strict, actions are advisory and have some latency.
1107 unsigned long *memory_pressure
;
1112 u32 sysctl_wmem_offset
;
1113 u32 sysctl_rmem_offset
;
1118 struct kmem_cache
*slab
;
1119 unsigned int obj_size
;
1120 slab_flags_t slab_flags
;
1122 struct percpu_counter
*orphan_count
;
1124 struct request_sock_ops
*rsk_prot
;
1125 struct timewait_sock_ops
*twsk_prot
;
1128 struct inet_hashinfo
*hashinfo
;
1129 struct udp_table
*udp_table
;
1130 struct raw_hashinfo
*raw_hash
;
1131 struct smc_hashinfo
*smc_hash
;
1134 struct module
*owner
;
1138 struct list_head node
;
1139 #ifdef SOCK_REFCNT_DEBUG
1142 int (*diag_destroy
)(struct sock
*sk
, int err
);
1143 } __randomize_layout
;
1145 int proto_register(struct proto
*prot
, int alloc_slab
);
1146 void proto_unregister(struct proto
*prot
);
1148 #ifdef SOCK_REFCNT_DEBUG
1149 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
1151 atomic_inc(&sk
->sk_prot
->socks
);
1154 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
1156 atomic_dec(&sk
->sk_prot
->socks
);
1157 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
1158 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
1161 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
1163 if (refcount_read(&sk
->sk_refcnt
) != 1)
1164 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
1165 sk
->sk_prot
->name
, sk
, refcount_read(&sk
->sk_refcnt
));
1167 #else /* SOCK_REFCNT_DEBUG */
1168 #define sk_refcnt_debug_inc(sk) do { } while (0)
1169 #define sk_refcnt_debug_dec(sk) do { } while (0)
1170 #define sk_refcnt_debug_release(sk) do { } while (0)
1171 #endif /* SOCK_REFCNT_DEBUG */
1173 static inline bool sk_stream_memory_free(const struct sock
*sk
)
1175 if (sk
->sk_wmem_queued
>= sk
->sk_sndbuf
)
1178 return sk
->sk_prot
->stream_memory_free
?
1179 sk
->sk_prot
->stream_memory_free(sk
) : true;
1182 static inline bool sk_stream_is_writeable(const struct sock
*sk
)
1184 return sk_stream_wspace(sk
) >= sk_stream_min_wspace(sk
) &&
1185 sk_stream_memory_free(sk
);
1188 static inline int sk_under_cgroup_hierarchy(struct sock
*sk
,
1189 struct cgroup
*ancestor
)
1191 #ifdef CONFIG_SOCK_CGROUP_DATA
1192 return cgroup_is_descendant(sock_cgroup_ptr(&sk
->sk_cgrp_data
),
1199 static inline bool sk_has_memory_pressure(const struct sock
*sk
)
1201 return sk
->sk_prot
->memory_pressure
!= NULL
;
1204 static inline bool sk_under_memory_pressure(const struct sock
*sk
)
1206 if (!sk
->sk_prot
->memory_pressure
)
1209 if (mem_cgroup_sockets_enabled
&& sk
->sk_memcg
&&
1210 mem_cgroup_under_socket_pressure(sk
->sk_memcg
))
1213 return !!*sk
->sk_prot
->memory_pressure
;
1217 sk_memory_allocated(const struct sock
*sk
)
1219 return atomic_long_read(sk
->sk_prot
->memory_allocated
);
1223 sk_memory_allocated_add(struct sock
*sk
, int amt
)
1225 return atomic_long_add_return(amt
, sk
->sk_prot
->memory_allocated
);
1229 sk_memory_allocated_sub(struct sock
*sk
, int amt
)
1231 atomic_long_sub(amt
, sk
->sk_prot
->memory_allocated
);
1234 static inline void sk_sockets_allocated_dec(struct sock
*sk
)
1236 percpu_counter_dec(sk
->sk_prot
->sockets_allocated
);
1239 static inline void sk_sockets_allocated_inc(struct sock
*sk
)
1241 percpu_counter_inc(sk
->sk_prot
->sockets_allocated
);
1245 sk_sockets_allocated_read_positive(struct sock
*sk
)
1247 return percpu_counter_read_positive(sk
->sk_prot
->sockets_allocated
);
1251 proto_sockets_allocated_sum_positive(struct proto
*prot
)
1253 return percpu_counter_sum_positive(prot
->sockets_allocated
);
1257 proto_memory_allocated(struct proto
*prot
)
1259 return atomic_long_read(prot
->memory_allocated
);
1263 proto_memory_pressure(struct proto
*prot
)
1265 if (!prot
->memory_pressure
)
1267 return !!*prot
->memory_pressure
;
1271 #ifdef CONFIG_PROC_FS
1272 /* Called with local bh disabled */
1273 void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int inc
);
1274 int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
1276 static inline void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
1283 /* With per-bucket locks this operation is not-atomic, so that
1284 * this version is not worse.
1286 static inline int __sk_prot_rehash(struct sock
*sk
)
1288 sk
->sk_prot
->unhash(sk
);
1289 return sk
->sk_prot
->hash(sk
);
1292 /* About 10 seconds */
1293 #define SOCK_DESTROY_TIME (10*HZ)
1295 /* Sockets 0-1023 can't be bound to unless you are superuser */
1296 #define PROT_SOCK 1024
1298 #define SHUTDOWN_MASK 3
1299 #define RCV_SHUTDOWN 1
1300 #define SEND_SHUTDOWN 2
1302 #define SOCK_SNDBUF_LOCK 1
1303 #define SOCK_RCVBUF_LOCK 2
1304 #define SOCK_BINDADDR_LOCK 4
1305 #define SOCK_BINDPORT_LOCK 8
1307 struct socket_alloc
{
1308 struct socket socket
;
1309 struct inode vfs_inode
;
1312 static inline struct socket
*SOCKET_I(struct inode
*inode
)
1314 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
1317 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
1319 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
1323 * Functions for memory accounting
1325 int __sk_mem_raise_allocated(struct sock
*sk
, int size
, int amt
, int kind
);
1326 int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
1327 void __sk_mem_reduce_allocated(struct sock
*sk
, int amount
);
1328 void __sk_mem_reclaim(struct sock
*sk
, int amount
);
1330 /* We used to have PAGE_SIZE here, but systems with 64KB pages
1331 * do not necessarily have 16x time more memory than 4KB ones.
1333 #define SK_MEM_QUANTUM 4096
1334 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1335 #define SK_MEM_SEND 0
1336 #define SK_MEM_RECV 1
1338 /* sysctl_mem values are in pages, we convert them in SK_MEM_QUANTUM units */
1339 static inline long sk_prot_mem_limits(const struct sock
*sk
, int index
)
1341 long val
= sk
->sk_prot
->sysctl_mem
[index
];
1343 #if PAGE_SIZE > SK_MEM_QUANTUM
1344 val
<<= PAGE_SHIFT
- SK_MEM_QUANTUM_SHIFT
;
1345 #elif PAGE_SIZE < SK_MEM_QUANTUM
1346 val
>>= SK_MEM_QUANTUM_SHIFT
- PAGE_SHIFT
;
1351 static inline int sk_mem_pages(int amt
)
1353 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
1356 static inline bool sk_has_account(struct sock
*sk
)
1358 /* return true if protocol supports memory accounting */
1359 return !!sk
->sk_prot
->memory_allocated
;
1362 static inline bool sk_wmem_schedule(struct sock
*sk
, int size
)
1364 if (!sk_has_account(sk
))
1366 return size
<= sk
->sk_forward_alloc
||
1367 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
1371 sk_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
, int size
)
1373 if (!sk_has_account(sk
))
1375 return size
<= sk
->sk_forward_alloc
||
1376 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
) ||
1377 skb_pfmemalloc(skb
);
1380 static inline void sk_mem_reclaim(struct sock
*sk
)
1382 if (!sk_has_account(sk
))
1384 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
1385 __sk_mem_reclaim(sk
, sk
->sk_forward_alloc
);
1388 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
1390 if (!sk_has_account(sk
))
1392 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
1393 __sk_mem_reclaim(sk
, sk
->sk_forward_alloc
- 1);
1396 static inline void sk_mem_charge(struct sock
*sk
, int size
)
1398 if (!sk_has_account(sk
))
1400 sk
->sk_forward_alloc
-= size
;
1403 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
1405 if (!sk_has_account(sk
))
1407 sk
->sk_forward_alloc
+= size
;
1409 /* Avoid a possible overflow.
1410 * TCP send queues can make this happen, if sk_mem_reclaim()
1411 * is not called and more than 2 GBytes are released at once.
1413 * If we reach 2 MBytes, reclaim 1 MBytes right now, there is
1414 * no need to hold that much forward allocation anyway.
1416 if (unlikely(sk
->sk_forward_alloc
>= 1 << 21))
1417 __sk_mem_reclaim(sk
, 1 << 20);
1420 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
1422 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
1423 sk
->sk_wmem_queued
-= skb
->truesize
;
1424 sk_mem_uncharge(sk
, skb
->truesize
);
1428 static inline void sock_release_ownership(struct sock
*sk
)
1430 if (sk
->sk_lock
.owned
) {
1431 sk
->sk_lock
.owned
= 0;
1433 /* The sk_lock has mutex_unlock() semantics: */
1434 mutex_release(&sk
->sk_lock
.dep_map
, 1, _RET_IP_
);
1439 * Macro so as to not evaluate some arguments when
1440 * lockdep is not enabled.
1442 * Mark both the sk_lock and the sk_lock.slock as a
1443 * per-address-family lock class.
1445 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1447 sk->sk_lock.owned = 0; \
1448 init_waitqueue_head(&sk->sk_lock.wq); \
1449 spin_lock_init(&(sk)->sk_lock.slock); \
1450 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1451 sizeof((sk)->sk_lock)); \
1452 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1454 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1457 #ifdef CONFIG_LOCKDEP
1458 static inline bool lockdep_sock_is_held(const struct sock
*csk
)
1460 struct sock
*sk
= (struct sock
*)csk
;
1462 return lockdep_is_held(&sk
->sk_lock
) ||
1463 lockdep_is_held(&sk
->sk_lock
.slock
);
1467 void lock_sock_nested(struct sock
*sk
, int subclass
);
1469 static inline void lock_sock(struct sock
*sk
)
1471 lock_sock_nested(sk
, 0);
1474 void __release_sock(struct sock
*sk
);
1475 void release_sock(struct sock
*sk
);
1477 /* BH context may only use the following locking interface. */
1478 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1479 #define bh_lock_sock_nested(__sk) \
1480 spin_lock_nested(&((__sk)->sk_lock.slock), \
1481 SINGLE_DEPTH_NESTING)
1482 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1484 bool lock_sock_fast(struct sock
*sk
);
1486 * unlock_sock_fast - complement of lock_sock_fast
1490 * fast unlock socket for user context.
1491 * If slow mode is on, we call regular release_sock()
1493 static inline void unlock_sock_fast(struct sock
*sk
, bool slow
)
1498 spin_unlock_bh(&sk
->sk_lock
.slock
);
1501 /* Used by processes to "lock" a socket state, so that
1502 * interrupts and bottom half handlers won't change it
1503 * from under us. It essentially blocks any incoming
1504 * packets, so that we won't get any new data or any
1505 * packets that change the state of the socket.
1507 * While locked, BH processing will add new packets to
1508 * the backlog queue. This queue is processed by the
1509 * owner of the socket lock right before it is released.
1511 * Since ~2.3.5 it is also exclusive sleep lock serializing
1512 * accesses from user process context.
1515 static inline void sock_owned_by_me(const struct sock
*sk
)
1517 #ifdef CONFIG_LOCKDEP
1518 WARN_ON_ONCE(!lockdep_sock_is_held(sk
) && debug_locks
);
1522 static inline bool sock_owned_by_user(const struct sock
*sk
)
1524 sock_owned_by_me(sk
);
1525 return sk
->sk_lock
.owned
;
1528 static inline bool sock_owned_by_user_nocheck(const struct sock
*sk
)
1530 return sk
->sk_lock
.owned
;
1533 /* no reclassification while locks are held */
1534 static inline bool sock_allow_reclassification(const struct sock
*csk
)
1536 struct sock
*sk
= (struct sock
*)csk
;
1538 return !sk
->sk_lock
.owned
&& !spin_is_locked(&sk
->sk_lock
.slock
);
1541 struct sock
*sk_alloc(struct net
*net
, int family
, gfp_t priority
,
1542 struct proto
*prot
, int kern
);
1543 void sk_free(struct sock
*sk
);
1544 void sk_destruct(struct sock
*sk
);
1545 struct sock
*sk_clone_lock(const struct sock
*sk
, const gfp_t priority
);
1546 void sk_free_unlock_clone(struct sock
*sk
);
1548 struct sk_buff
*sock_wmalloc(struct sock
*sk
, unsigned long size
, int force
,
1550 void __sock_wfree(struct sk_buff
*skb
);
1551 void sock_wfree(struct sk_buff
*skb
);
1552 struct sk_buff
*sock_omalloc(struct sock
*sk
, unsigned long size
,
1554 void skb_orphan_partial(struct sk_buff
*skb
);
1555 void sock_rfree(struct sk_buff
*skb
);
1556 void sock_efree(struct sk_buff
*skb
);
1558 void sock_edemux(struct sk_buff
*skb
);
1560 #define sock_edemux sock_efree
1563 int sock_setsockopt(struct socket
*sock
, int level
, int op
,
1564 char __user
*optval
, unsigned int optlen
);
1566 int sock_getsockopt(struct socket
*sock
, int level
, int op
,
1567 char __user
*optval
, int __user
*optlen
);
1568 struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
, unsigned long size
,
1569 int noblock
, int *errcode
);
1570 struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
, unsigned long header_len
,
1571 unsigned long data_len
, int noblock
,
1572 int *errcode
, int max_page_order
);
1573 void *sock_kmalloc(struct sock
*sk
, int size
, gfp_t priority
);
1574 void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
1575 void sock_kzfree_s(struct sock
*sk
, void *mem
, int size
);
1576 void sk_send_sigurg(struct sock
*sk
);
1578 struct sockcm_cookie
{
1583 int __sock_cmsg_send(struct sock
*sk
, struct msghdr
*msg
, struct cmsghdr
*cmsg
,
1584 struct sockcm_cookie
*sockc
);
1585 int sock_cmsg_send(struct sock
*sk
, struct msghdr
*msg
,
1586 struct sockcm_cookie
*sockc
);
1589 * Functions to fill in entries in struct proto_ops when a protocol
1590 * does not implement a particular function.
1592 int sock_no_bind(struct socket
*, struct sockaddr
*, int);
1593 int sock_no_connect(struct socket
*, struct sockaddr
*, int, int);
1594 int sock_no_socketpair(struct socket
*, struct socket
*);
1595 int sock_no_accept(struct socket
*, struct socket
*, int, bool);
1596 int sock_no_getname(struct socket
*, struct sockaddr
*, int *, int);
1597 unsigned int sock_no_poll(struct file
*, struct socket
*,
1598 struct poll_table_struct
*);
1599 int sock_no_ioctl(struct socket
*, unsigned int, unsigned long);
1600 int sock_no_listen(struct socket
*, int);
1601 int sock_no_shutdown(struct socket
*, int);
1602 int sock_no_getsockopt(struct socket
*, int , int, char __user
*, int __user
*);
1603 int sock_no_setsockopt(struct socket
*, int, int, char __user
*, unsigned int);
1604 int sock_no_sendmsg(struct socket
*, struct msghdr
*, size_t);
1605 int sock_no_sendmsg_locked(struct sock
*sk
, struct msghdr
*msg
, size_t len
);
1606 int sock_no_recvmsg(struct socket
*, struct msghdr
*, size_t, int);
1607 int sock_no_mmap(struct file
*file
, struct socket
*sock
,
1608 struct vm_area_struct
*vma
);
1609 ssize_t
sock_no_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
1610 size_t size
, int flags
);
1611 ssize_t
sock_no_sendpage_locked(struct sock
*sk
, struct page
*page
,
1612 int offset
, size_t size
, int flags
);
1615 * Functions to fill in entries in struct proto_ops when a protocol
1616 * uses the inet style.
1618 int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1619 char __user
*optval
, int __user
*optlen
);
1620 int sock_common_recvmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
,
1622 int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1623 char __user
*optval
, unsigned int optlen
);
1624 int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
1625 int optname
, char __user
*optval
, int __user
*optlen
);
1626 int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
1627 int optname
, char __user
*optval
, unsigned int optlen
);
1629 void sk_common_release(struct sock
*sk
);
1632 * Default socket callbacks and setup code
1635 /* Initialise core socket variables */
1636 void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1639 * Socket reference counting postulates.
1641 * * Each user of socket SHOULD hold a reference count.
1642 * * Each access point to socket (an hash table bucket, reference from a list,
1643 * running timer, skb in flight MUST hold a reference count.
1644 * * When reference count hits 0, it means it will never increase back.
1645 * * When reference count hits 0, it means that no references from
1646 * outside exist to this socket and current process on current CPU
1647 * is last user and may/should destroy this socket.
1648 * * sk_free is called from any context: process, BH, IRQ. When
1649 * it is called, socket has no references from outside -> sk_free
1650 * may release descendant resources allocated by the socket, but
1651 * to the time when it is called, socket is NOT referenced by any
1652 * hash tables, lists etc.
1653 * * Packets, delivered from outside (from network or from another process)
1654 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1655 * when they sit in queue. Otherwise, packets will leak to hole, when
1656 * socket is looked up by one cpu and unhasing is made by another CPU.
1657 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1658 * (leak to backlog). Packet socket does all the processing inside
1659 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1660 * use separate SMP lock, so that they are prone too.
1663 /* Ungrab socket and destroy it, if it was the last reference. */
1664 static inline void sock_put(struct sock
*sk
)
1666 if (refcount_dec_and_test(&sk
->sk_refcnt
))
1669 /* Generic version of sock_put(), dealing with all sockets
1670 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
1672 void sock_gen_put(struct sock
*sk
);
1674 int __sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
, const int nested
,
1675 unsigned int trim_cap
, bool refcounted
);
1676 static inline int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
1679 return __sk_receive_skb(sk
, skb
, nested
, 1, true);
1682 static inline void sk_tx_queue_set(struct sock
*sk
, int tx_queue
)
1684 sk
->sk_tx_queue_mapping
= tx_queue
;
1687 static inline void sk_tx_queue_clear(struct sock
*sk
)
1689 sk
->sk_tx_queue_mapping
= -1;
1692 static inline int sk_tx_queue_get(const struct sock
*sk
)
1694 return sk
? sk
->sk_tx_queue_mapping
: -1;
1697 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1699 sk_tx_queue_clear(sk
);
1700 sk
->sk_socket
= sock
;
1703 static inline wait_queue_head_t
*sk_sleep(struct sock
*sk
)
1705 BUILD_BUG_ON(offsetof(struct socket_wq
, wait
) != 0);
1706 return &rcu_dereference_raw(sk
->sk_wq
)->wait
;
1708 /* Detach socket from process context.
1709 * Announce socket dead, detach it from wait queue and inode.
1710 * Note that parent inode held reference count on this struct sock,
1711 * we do not release it in this function, because protocol
1712 * probably wants some additional cleanups or even continuing
1713 * to work with this socket (TCP).
1715 static inline void sock_orphan(struct sock
*sk
)
1717 write_lock_bh(&sk
->sk_callback_lock
);
1718 sock_set_flag(sk
, SOCK_DEAD
);
1719 sk_set_socket(sk
, NULL
);
1721 write_unlock_bh(&sk
->sk_callback_lock
);
1724 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1726 WARN_ON(parent
->sk
);
1727 write_lock_bh(&sk
->sk_callback_lock
);
1728 sk
->sk_wq
= parent
->wq
;
1730 sk_set_socket(sk
, parent
);
1731 sk
->sk_uid
= SOCK_INODE(parent
)->i_uid
;
1732 security_sock_graft(sk
, parent
);
1733 write_unlock_bh(&sk
->sk_callback_lock
);
1736 kuid_t
sock_i_uid(struct sock
*sk
);
1737 unsigned long sock_i_ino(struct sock
*sk
);
1739 static inline kuid_t
sock_net_uid(const struct net
*net
, const struct sock
*sk
)
1741 return sk
? sk
->sk_uid
: make_kuid(net
->user_ns
, 0);
1744 static inline u32
net_tx_rndhash(void)
1746 u32 v
= prandom_u32();
1751 static inline void sk_set_txhash(struct sock
*sk
)
1753 sk
->sk_txhash
= net_tx_rndhash();
1756 static inline void sk_rethink_txhash(struct sock
*sk
)
1762 static inline struct dst_entry
*
1763 __sk_dst_get(struct sock
*sk
)
1765 return rcu_dereference_check(sk
->sk_dst_cache
,
1766 lockdep_sock_is_held(sk
));
1769 static inline struct dst_entry
*
1770 sk_dst_get(struct sock
*sk
)
1772 struct dst_entry
*dst
;
1775 dst
= rcu_dereference(sk
->sk_dst_cache
);
1776 if (dst
&& !atomic_inc_not_zero(&dst
->__refcnt
))
1782 static inline void dst_negative_advice(struct sock
*sk
)
1784 struct dst_entry
*ndst
, *dst
= __sk_dst_get(sk
);
1786 sk_rethink_txhash(sk
);
1788 if (dst
&& dst
->ops
->negative_advice
) {
1789 ndst
= dst
->ops
->negative_advice(dst
);
1792 rcu_assign_pointer(sk
->sk_dst_cache
, ndst
);
1793 sk_tx_queue_clear(sk
);
1794 sk
->sk_dst_pending_confirm
= 0;
1800 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1802 struct dst_entry
*old_dst
;
1804 sk_tx_queue_clear(sk
);
1805 sk
->sk_dst_pending_confirm
= 0;
1806 old_dst
= rcu_dereference_protected(sk
->sk_dst_cache
,
1807 lockdep_sock_is_held(sk
));
1808 rcu_assign_pointer(sk
->sk_dst_cache
, dst
);
1809 dst_release(old_dst
);
1813 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1815 struct dst_entry
*old_dst
;
1817 sk_tx_queue_clear(sk
);
1818 sk
->sk_dst_pending_confirm
= 0;
1819 old_dst
= xchg((__force
struct dst_entry
**)&sk
->sk_dst_cache
, dst
);
1820 dst_release(old_dst
);
1824 __sk_dst_reset(struct sock
*sk
)
1826 __sk_dst_set(sk
, NULL
);
1830 sk_dst_reset(struct sock
*sk
)
1832 sk_dst_set(sk
, NULL
);
1835 struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1837 struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1839 static inline void sk_dst_confirm(struct sock
*sk
)
1841 if (!sk
->sk_dst_pending_confirm
)
1842 sk
->sk_dst_pending_confirm
= 1;
1845 static inline void sock_confirm_neigh(struct sk_buff
*skb
, struct neighbour
*n
)
1847 if (skb_get_dst_pending_confirm(skb
)) {
1848 struct sock
*sk
= skb
->sk
;
1849 unsigned long now
= jiffies
;
1851 /* avoid dirtying neighbour */
1852 if (n
->confirmed
!= now
)
1854 if (sk
&& sk
->sk_dst_pending_confirm
)
1855 sk
->sk_dst_pending_confirm
= 0;
1859 bool sk_mc_loop(struct sock
*sk
);
1861 static inline bool sk_can_gso(const struct sock
*sk
)
1863 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1866 void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1868 static inline void sk_nocaps_add(struct sock
*sk
, netdev_features_t flags
)
1870 sk
->sk_route_nocaps
|= flags
;
1871 sk
->sk_route_caps
&= ~flags
;
1874 static inline bool sk_check_csum_caps(struct sock
*sk
)
1876 return (sk
->sk_route_caps
& NETIF_F_HW_CSUM
) ||
1877 (sk
->sk_family
== PF_INET
&&
1878 (sk
->sk_route_caps
& NETIF_F_IP_CSUM
)) ||
1879 (sk
->sk_family
== PF_INET6
&&
1880 (sk
->sk_route_caps
& NETIF_F_IPV6_CSUM
));
1883 static inline int skb_do_copy_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1884 struct iov_iter
*from
, char *to
,
1885 int copy
, int offset
)
1887 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1889 if (!csum_and_copy_from_iter_full(to
, copy
, &csum
, from
))
1891 skb
->csum
= csum_block_add(skb
->csum
, csum
, offset
);
1892 } else if (sk
->sk_route_caps
& NETIF_F_NOCACHE_COPY
) {
1893 if (!copy_from_iter_full_nocache(to
, copy
, from
))
1895 } else if (!copy_from_iter_full(to
, copy
, from
))
1901 static inline int skb_add_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1902 struct iov_iter
*from
, int copy
)
1904 int err
, offset
= skb
->len
;
1906 err
= skb_do_copy_data_nocache(sk
, skb
, from
, skb_put(skb
, copy
),
1909 __skb_trim(skb
, offset
);
1914 static inline int skb_copy_to_page_nocache(struct sock
*sk
, struct iov_iter
*from
,
1915 struct sk_buff
*skb
,
1921 err
= skb_do_copy_data_nocache(sk
, skb
, from
, page_address(page
) + off
,
1927 skb
->data_len
+= copy
;
1928 skb
->truesize
+= copy
;
1929 sk
->sk_wmem_queued
+= copy
;
1930 sk_mem_charge(sk
, copy
);
1935 * sk_wmem_alloc_get - returns write allocations
1938 * Returns sk_wmem_alloc minus initial offset of one
1940 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
1942 return refcount_read(&sk
->sk_wmem_alloc
) - 1;
1946 * sk_rmem_alloc_get - returns read allocations
1949 * Returns sk_rmem_alloc
1951 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
1953 return atomic_read(&sk
->sk_rmem_alloc
);
1957 * sk_has_allocations - check if allocations are outstanding
1960 * Returns true if socket has write or read allocations
1962 static inline bool sk_has_allocations(const struct sock
*sk
)
1964 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
1968 * skwq_has_sleeper - check if there are any waiting processes
1969 * @wq: struct socket_wq
1971 * Returns true if socket_wq has waiting processes
1973 * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
1974 * barrier call. They were added due to the race found within the tcp code.
1976 * Consider following tcp code paths::
1979 * sys_select receive packet
1981 * __add_wait_queue update tp->rcv_nxt
1983 * tp->rcv_nxt check sock_def_readable
1985 * schedule rcu_read_lock();
1986 * wq = rcu_dereference(sk->sk_wq);
1987 * if (wq && waitqueue_active(&wq->wait))
1988 * wake_up_interruptible(&wq->wait)
1992 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1993 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1994 * could then endup calling schedule and sleep forever if there are no more
1995 * data on the socket.
1998 static inline bool skwq_has_sleeper(struct socket_wq
*wq
)
2000 return wq
&& wq_has_sleeper(&wq
->wait
);
2004 * sock_poll_wait - place memory barrier behind the poll_wait call.
2006 * @wait_address: socket wait queue
2009 * See the comments in the wq_has_sleeper function.
2011 static inline void sock_poll_wait(struct file
*filp
,
2012 wait_queue_head_t
*wait_address
, poll_table
*p
)
2014 if (!poll_does_not_wait(p
) && wait_address
) {
2015 poll_wait(filp
, wait_address
, p
);
2016 /* We need to be sure we are in sync with the
2017 * socket flags modification.
2019 * This memory barrier is paired in the wq_has_sleeper.
2025 static inline void skb_set_hash_from_sk(struct sk_buff
*skb
, struct sock
*sk
)
2027 if (sk
->sk_txhash
) {
2029 skb
->hash
= sk
->sk_txhash
;
2033 void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
);
2036 * Queue a received datagram if it will fit. Stream and sequenced
2037 * protocols can't normally use this as they need to fit buffers in
2038 * and play with them.
2040 * Inlined as it's very short and called for pretty much every
2041 * packet ever received.
2043 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
2047 skb
->destructor
= sock_rfree
;
2048 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
2049 sk_mem_charge(sk
, skb
->truesize
);
2052 void sk_reset_timer(struct sock
*sk
, struct timer_list
*timer
,
2053 unsigned long expires
);
2055 void sk_stop_timer(struct sock
*sk
, struct timer_list
*timer
);
2057 int __sk_queue_drop_skb(struct sock
*sk
, struct sk_buff_head
*sk_queue
,
2058 struct sk_buff
*skb
, unsigned int flags
,
2059 void (*destructor
)(struct sock
*sk
,
2060 struct sk_buff
*skb
));
2061 int __sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
2062 int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
2064 int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
);
2065 struct sk_buff
*sock_dequeue_err_skb(struct sock
*sk
);
2068 * Recover an error report and clear atomically
2071 static inline int sock_error(struct sock
*sk
)
2074 if (likely(!sk
->sk_err
))
2076 err
= xchg(&sk
->sk_err
, 0);
2080 static inline unsigned long sock_wspace(struct sock
*sk
)
2084 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
2085 amt
= sk
->sk_sndbuf
- refcount_read(&sk
->sk_wmem_alloc
);
2093 * We use sk->sk_wq_raw, from contexts knowing this
2094 * pointer is not NULL and cannot disappear/change.
2096 static inline void sk_set_bit(int nr
, struct sock
*sk
)
2098 if ((nr
== SOCKWQ_ASYNC_NOSPACE
|| nr
== SOCKWQ_ASYNC_WAITDATA
) &&
2099 !sock_flag(sk
, SOCK_FASYNC
))
2102 set_bit(nr
, &sk
->sk_wq_raw
->flags
);
2105 static inline void sk_clear_bit(int nr
, struct sock
*sk
)
2107 if ((nr
== SOCKWQ_ASYNC_NOSPACE
|| nr
== SOCKWQ_ASYNC_WAITDATA
) &&
2108 !sock_flag(sk
, SOCK_FASYNC
))
2111 clear_bit(nr
, &sk
->sk_wq_raw
->flags
);
2114 static inline void sk_wake_async(const struct sock
*sk
, int how
, int band
)
2116 if (sock_flag(sk
, SOCK_FASYNC
)) {
2118 sock_wake_async(rcu_dereference(sk
->sk_wq
), how
, band
);
2123 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2124 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2125 * Note: for send buffers, TCP works better if we can build two skbs at
2128 #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2130 #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2131 #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
2133 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
2135 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
2136 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
2137 sk
->sk_sndbuf
= max_t(u32
, sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
2141 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
,
2142 bool force_schedule
);
2145 * sk_page_frag - return an appropriate page_frag
2148 * Use the per task page_frag instead of the per socket one for
2149 * optimization when we know that we're in the normal context and owns
2150 * everything that's associated with %current.
2152 * gfpflags_allow_blocking() isn't enough here as direct reclaim may nest
2153 * inside other socket operations and end up recursing into sk_page_frag()
2154 * while it's already in use.
2156 static inline struct page_frag
*sk_page_frag(struct sock
*sk
)
2158 if (gfpflags_normal_context(sk
->sk_allocation
))
2159 return ¤t
->task_frag
;
2161 return &sk
->sk_frag
;
2164 bool sk_page_frag_refill(struct sock
*sk
, struct page_frag
*pfrag
);
2167 * Default write policy as shown to user space via poll/select/SIGIO
2169 static inline bool sock_writeable(const struct sock
*sk
)
2171 return refcount_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
2174 static inline gfp_t
gfp_any(void)
2176 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
2179 static inline long sock_rcvtimeo(const struct sock
*sk
, bool noblock
)
2181 return noblock
? 0 : sk
->sk_rcvtimeo
;
2184 static inline long sock_sndtimeo(const struct sock
*sk
, bool noblock
)
2186 return noblock
? 0 : sk
->sk_sndtimeo
;
2189 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
2191 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
2194 /* Alas, with timeout socket operations are not restartable.
2195 * Compare this to poll().
2197 static inline int sock_intr_errno(long timeo
)
2199 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
2202 struct sock_skb_cb
{
2206 /* Store sock_skb_cb at the end of skb->cb[] so protocol families
2207 * using skb->cb[] would keep using it directly and utilize its
2208 * alignement guarantee.
2210 #define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \
2211 sizeof(struct sock_skb_cb)))
2213 #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2214 SOCK_SKB_CB_OFFSET))
2216 #define sock_skb_cb_check_size(size) \
2217 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
2220 sock_skb_set_dropcount(const struct sock
*sk
, struct sk_buff
*skb
)
2222 SOCK_SKB_CB(skb
)->dropcount
= sock_flag(sk
, SOCK_RXQ_OVFL
) ?
2223 atomic_read(&sk
->sk_drops
) : 0;
2226 static inline void sk_drops_add(struct sock
*sk
, const struct sk_buff
*skb
)
2228 int segs
= max_t(u16
, 1, skb_shinfo(skb
)->gso_segs
);
2230 atomic_add(segs
, &sk
->sk_drops
);
2233 static inline ktime_t
sock_read_timestamp(struct sock
*sk
)
2235 #if BITS_PER_LONG==32
2240 seq
= read_seqbegin(&sk
->sk_stamp_seq
);
2242 } while (read_seqretry(&sk
->sk_stamp_seq
, seq
));
2246 return sk
->sk_stamp
;
2250 static inline void sock_write_timestamp(struct sock
*sk
, ktime_t kt
)
2252 #if BITS_PER_LONG==32
2253 write_seqlock(&sk
->sk_stamp_seq
);
2255 write_sequnlock(&sk
->sk_stamp_seq
);
2261 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
2262 struct sk_buff
*skb
);
2263 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
2264 struct sk_buff
*skb
);
2267 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
2269 ktime_t kt
= skb
->tstamp
;
2270 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
2273 * generate control messages if
2274 * - receive time stamping in software requested
2275 * - software time stamp available and wanted
2276 * - hardware time stamps available and wanted
2278 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
2279 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RX_SOFTWARE
) ||
2280 (kt
&& sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
) ||
2281 (hwtstamps
->hwtstamp
&&
2282 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
)))
2283 __sock_recv_timestamp(msg
, sk
, skb
);
2285 sock_write_timestamp(sk
, kt
);
2287 if (sock_flag(sk
, SOCK_WIFI_STATUS
) && skb
->wifi_acked_valid
)
2288 __sock_recv_wifi_status(msg
, sk
, skb
);
2291 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2292 struct sk_buff
*skb
);
2294 #define SK_DEFAULT_STAMP (-1L * NSEC_PER_SEC)
2295 static inline void sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2296 struct sk_buff
*skb
)
2298 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2299 (1UL << SOCK_RCVTSTAMP))
2300 #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2301 SOF_TIMESTAMPING_RAW_HARDWARE)
2303 if (sk
->sk_flags
& FLAGS_TS_OR_DROPS
|| sk
->sk_tsflags
& TSFLAGS_ANY
)
2304 __sock_recv_ts_and_drops(msg
, sk
, skb
);
2305 else if (unlikely(sock_flag(sk
, SOCK_TIMESTAMP
)))
2306 sock_write_timestamp(sk
, skb
->tstamp
);
2307 else if (unlikely(sk
->sk_stamp
== SK_DEFAULT_STAMP
))
2308 sock_write_timestamp(sk
, 0);
2311 void __sock_tx_timestamp(__u16 tsflags
, __u8
*tx_flags
);
2314 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2315 * @sk: socket sending this packet
2316 * @tsflags: timestamping flags to use
2317 * @tx_flags: completed with instructions for time stamping
2319 * Note: callers should take care of initial ``*tx_flags`` value (usually 0)
2321 static inline void sock_tx_timestamp(const struct sock
*sk
, __u16 tsflags
,
2324 if (unlikely(tsflags
))
2325 __sock_tx_timestamp(tsflags
, tx_flags
);
2326 if (unlikely(sock_flag(sk
, SOCK_WIFI_STATUS
)))
2327 *tx_flags
|= SKBTX_WIFI_STATUS
;
2331 * sk_eat_skb - Release a skb if it is no longer needed
2332 * @sk: socket to eat this skb from
2333 * @skb: socket buffer to eat
2335 * This routine must be called with interrupts disabled or with the socket
2336 * locked so that the sk_buff queue operation is ok.
2338 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
)
2340 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2345 struct net
*sock_net(const struct sock
*sk
)
2347 return read_pnet(&sk
->sk_net
);
2351 void sock_net_set(struct sock
*sk
, struct net
*net
)
2353 write_pnet(&sk
->sk_net
, net
);
2356 static inline struct sock
*skb_steal_sock(struct sk_buff
*skb
)
2359 struct sock
*sk
= skb
->sk
;
2361 skb
->destructor
= NULL
;
2368 /* This helper checks if a socket is a full socket,
2369 * ie _not_ a timewait or request socket.
2371 static inline bool sk_fullsock(const struct sock
*sk
)
2373 return (1 << sk
->sk_state
) & ~(TCPF_TIME_WAIT
| TCPF_NEW_SYN_RECV
);
2376 /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2377 * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2379 static inline bool sk_listener(const struct sock
*sk
)
2381 return (1 << sk
->sk_state
) & (TCPF_LISTEN
| TCPF_NEW_SYN_RECV
);
2385 * sk_state_load - read sk->sk_state for lockless contexts
2386 * @sk: socket pointer
2388 * Paired with sk_state_store(). Used in places we do not hold socket lock :
2389 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
2391 static inline int sk_state_load(const struct sock
*sk
)
2393 return smp_load_acquire(&sk
->sk_state
);
2397 * sk_state_store - update sk->sk_state
2398 * @sk: socket pointer
2399 * @newstate: new state
2401 * Paired with sk_state_load(). Should be used in contexts where
2402 * state change might impact lockless readers.
2404 static inline void sk_state_store(struct sock
*sk
, int newstate
)
2406 smp_store_release(&sk
->sk_state
, newstate
);
2409 void sock_enable_timestamp(struct sock
*sk
, int flag
);
2410 int sock_get_timestamp(struct sock
*, struct timeval __user
*);
2411 int sock_get_timestampns(struct sock
*, struct timespec __user
*);
2412 int sock_recv_errqueue(struct sock
*sk
, struct msghdr
*msg
, int len
, int level
,
2415 bool sk_ns_capable(const struct sock
*sk
,
2416 struct user_namespace
*user_ns
, int cap
);
2417 bool sk_capable(const struct sock
*sk
, int cap
);
2418 bool sk_net_capable(const struct sock
*sk
, int cap
);
2420 void sk_get_meminfo(const struct sock
*sk
, u32
*meminfo
);
2422 /* Take into consideration the size of the struct sk_buff overhead in the
2423 * determination of these values, since that is non-constant across
2424 * platforms. This makes socket queueing behavior and performance
2425 * not depend upon such differences.
2427 #define _SK_MEM_PACKETS 256
2428 #define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
2429 #define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2430 #define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2432 extern __u32 sysctl_wmem_max
;
2433 extern __u32 sysctl_rmem_max
;
2435 extern int sysctl_tstamp_allow_data
;
2436 extern int sysctl_optmem_max
;
2438 extern __u32 sysctl_wmem_default
;
2439 extern __u32 sysctl_rmem_default
;
2441 static inline int sk_get_wmem0(const struct sock
*sk
, const struct proto
*proto
)
2443 /* Does this proto have per netns sysctl_wmem ? */
2444 if (proto
->sysctl_wmem_offset
)
2445 return *(int *)((void *)sock_net(sk
) + proto
->sysctl_wmem_offset
);
2447 return *proto
->sysctl_wmem
;
2450 static inline int sk_get_rmem0(const struct sock
*sk
, const struct proto
*proto
)
2452 /* Does this proto have per netns sysctl_rmem ? */
2453 if (proto
->sysctl_rmem_offset
)
2454 return *(int *)((void *)sock_net(sk
) + proto
->sysctl_rmem_offset
);
2456 return *proto
->sysctl_rmem
;
2459 #endif /* _SOCK_H */