1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * Definitions for the AF_INET socket handler.
9 * Version: @(#)sock.h 1.0.4 05/13/93
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Florian La Roche <flla@stud.uni-sb.de>
17 * Alan Cox : Volatiles in skbuff pointers. See
18 * skbuff comments. May be overdone,
19 * better to prove they can be removed
21 * Alan Cox : Added a zapped field for tcp to note
22 * a socket is reset and must stay shut up
23 * Alan Cox : New fields for options
24 * Pauline Middelink : identd support
25 * Alan Cox : Eliminate low level recv/recvfrom
26 * David S. Miller : New socket lookup architecture.
27 * Steve Whitehouse: Default routines for sock_ops
28 * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made
29 * protinfo be just a void pointer, as the
30 * protocol specific parts were moved to
31 * respective headers and ipv4/v6, etc now
32 * use private slabcaches for its socks
33 * Pedro Hortas : New flags field for socket options
38 #include <linux/hardirq.h>
39 #include <linux/kernel.h>
40 #include <linux/list.h>
41 #include <linux/list_nulls.h>
42 #include <linux/timer.h>
43 #include <linux/cache.h>
44 #include <linux/bitops.h>
45 #include <linux/lockdep.h>
46 #include <linux/netdevice.h>
47 #include <linux/skbuff.h> /* struct sk_buff */
49 #include <linux/security.h>
50 #include <linux/slab.h>
51 #include <linux/uaccess.h>
52 #include <linux/page_counter.h>
53 #include <linux/memcontrol.h>
54 #include <linux/static_key.h>
55 #include <linux/sched.h>
56 #include <linux/wait.h>
57 #include <linux/cgroup-defs.h>
58 #include <linux/rbtree.h>
59 #include <linux/rculist_nulls.h>
60 #include <linux/poll.h>
61 #include <linux/sockptr.h>
62 #include <linux/indirect_call_wrapper.h>
63 #include <linux/atomic.h>
64 #include <linux/refcount.h>
65 #include <linux/llist.h>
67 #include <net/checksum.h>
68 #include <net/tcp_states.h>
69 #include <linux/net_tstamp.h>
70 #include <net/l3mdev.h>
71 #include <uapi/linux/socket.h>
74 * This structure really needs to be cleaned up.
75 * Most of it is for TCP, and not used by any of
76 * the other protocols.
79 /* This is the per-socket lock. The spinlock provides a synchronization
80 * between user contexts and software interrupt processing, whereas the
81 * mini-semaphore synchronizes multiple users amongst themselves.
88 * We express the mutex-alike socket_lock semantics
89 * to the lock validator by explicitly managing
90 * the slock as a lock variant (in addition to
93 #ifdef CONFIG_DEBUG_LOCK_ALLOC
94 struct lockdep_map dep_map;
102 typedef __u32 __bitwise __portpair;
103 typedef __u64 __bitwise __addrpair;
106 * struct sock_common - minimal network layer representation of sockets
107 * @skc_daddr: Foreign IPv4 addr
108 * @skc_rcv_saddr: Bound local IPv4 addr
109 * @skc_addrpair: 8-byte-aligned __u64 union of @skc_daddr & @skc_rcv_saddr
110 * @skc_hash: hash value used with various protocol lookup tables
111 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
112 * @skc_dport: placeholder for inet_dport/tw_dport
113 * @skc_num: placeholder for inet_num/tw_num
114 * @skc_portpair: __u32 union of @skc_dport & @skc_num
115 * @skc_family: network address family
116 * @skc_state: Connection state
117 * @skc_reuse: %SO_REUSEADDR setting
118 * @skc_reuseport: %SO_REUSEPORT setting
119 * @skc_ipv6only: socket is IPV6 only
120 * @skc_net_refcnt: socket is using net ref counting
121 * @skc_bound_dev_if: bound device index if != 0
122 * @skc_bind_node: bind hash linkage for various protocol lookup tables
123 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
124 * @skc_prot: protocol handlers inside a network family
125 * @skc_net: reference to the network namespace of this socket
126 * @skc_v6_daddr: IPV6 destination address
127 * @skc_v6_rcv_saddr: IPV6 source address
128 * @skc_cookie: socket's cookie value
129 * @skc_node: main hash linkage for various protocol lookup tables
130 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
131 * @skc_tx_queue_mapping: tx queue number for this connection
132 * @skc_rx_queue_mapping: rx queue number for this connection
133 * @skc_flags: place holder for sk_flags
134 * %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
135 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
136 * @skc_listener: connection request listener socket (aka rsk_listener)
137 * [union with @skc_flags]
138 * @skc_tw_dr: (aka tw_dr) ptr to &struct inet_timewait_death_row
139 * [union with @skc_flags]
140 * @skc_incoming_cpu: record/match cpu processing incoming packets
141 * @skc_rcv_wnd: (aka rsk_rcv_wnd) TCP receive window size (possibly scaled)
142 * [union with @skc_incoming_cpu]
143 * @skc_tw_rcv_nxt: (aka tw_rcv_nxt) TCP window next expected seq number
144 * [union with @skc_incoming_cpu]
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.
152 __addrpair skc_addrpair;
155 __be32 skc_rcv_saddr;
159 unsigned int skc_hash;
160 __u16 skc_u16hashes[2];
162 /* skc_dport && skc_num must be grouped as well */
164 __portpair skc_portpair;
171 unsigned short skc_family;
172 volatile unsigned char skc_state;
173 unsigned char skc_reuse:4;
174 unsigned char skc_reuseport:1;
175 unsigned char skc_ipv6only:1;
176 unsigned char skc_net_refcnt:1;
177 int skc_bound_dev_if;
179 struct hlist_node skc_bind_node;
180 struct hlist_node skc_portaddr_node;
182 struct proto *skc_prot;
183 possible_net_t skc_net;
185 #if IS_ENABLED(CONFIG_IPV6)
186 struct in6_addr skc_v6_daddr;
187 struct in6_addr skc_v6_rcv_saddr;
190 atomic64_t skc_cookie;
192 /* following fields are padding to force
193 * offset(struct sock, sk_refcnt) == 128 on 64bit arches
194 * assuming IPV6 is enabled. We use this padding differently
195 * for different kind of 'sockets'
198 unsigned long skc_flags;
199 struct sock *skc_listener; /* request_sock */
200 struct inet_timewait_death_row *skc_tw_dr; /* inet_timewait_sock */
203 * fields between dontcopy_begin/dontcopy_end
204 * are not copied in sock_copy()
207 int skc_dontcopy_begin[0];
210 struct hlist_node skc_node;
211 struct hlist_nulls_node skc_nulls_node;
213 unsigned short skc_tx_queue_mapping;
214 #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
215 unsigned short skc_rx_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 */
234 struct bpf_local_storage;
238 * struct sock - network layer representation of sockets
239 * @__sk_common: shared layout with inet_timewait_sock
240 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
241 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
242 * @sk_lock: synchronizer
243 * @sk_kern_sock: True if sock is using kernel lock classes
244 * @sk_rcvbuf: size of receive buffer in bytes
245 * @sk_wq: sock wait queue and async head
246 * @sk_rx_dst: receive input route used by early demux
247 * @sk_rx_dst_ifindex: ifindex for @sk_rx_dst
248 * @sk_rx_dst_cookie: cookie for @sk_rx_dst
249 * @sk_dst_cache: destination cache
250 * @sk_dst_pending_confirm: need to confirm neighbour
251 * @sk_policy: flow policy
252 * @sk_receive_queue: incoming packets
253 * @sk_wmem_alloc: transmit queue bytes committed
254 * @sk_tsq_flags: TCP Small Queues flags
255 * @sk_write_queue: Packet sending queue
256 * @sk_omem_alloc: "o" is "option" or "other"
257 * @sk_wmem_queued: persistent queue size
258 * @sk_forward_alloc: space allocated forward
259 * @sk_reserved_mem: space reserved and non-reclaimable for the socket
260 * @sk_napi_id: id of the last napi context to receive data for sk
261 * @sk_ll_usec: usecs to busypoll when there is no data
262 * @sk_allocation: allocation mode
263 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
264 * @sk_pacing_status: Pacing status (requested, handled by sch_fq)
265 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
266 * @sk_sndbuf: size of send buffer in bytes
267 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
268 * @sk_no_check_rx: allow zero checksum in RX packets
269 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
270 * @sk_gso_disabled: if set, NETIF_F_GSO_MASK is forbidden.
271 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
272 * @sk_gso_max_size: Maximum GSO segment size to build
273 * @sk_gso_max_segs: Maximum number of GSO segments
274 * @sk_pacing_shift: scaling factor for TCP Small Queues
275 * @sk_lingertime: %SO_LINGER l_linger setting
276 * @sk_backlog: always used with the per-socket spinlock held
277 * @sk_callback_lock: used with the callbacks in the end of this struct
278 * @sk_error_queue: rarely used
279 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
280 * IPV6_ADDRFORM for instance)
281 * @sk_err: last error
282 * @sk_err_soft: errors that don't cause failure but are the cause of a
283 * persistent failure not just 'timed out'
284 * @sk_drops: raw/udp drops counter
285 * @sk_ack_backlog: current listen backlog
286 * @sk_max_ack_backlog: listen backlog set in listen()
287 * @sk_uid: user id of owner
288 * @sk_prefer_busy_poll: prefer busypolling over softirq processing
289 * @sk_busy_poll_budget: napi processing budget when busypolling
290 * @sk_priority: %SO_PRIORITY setting
291 * @sk_type: socket type (%SOCK_STREAM, etc)
292 * @sk_protocol: which protocol this socket belongs in this network family
293 * @sk_peer_lock: lock protecting @sk_peer_pid and @sk_peer_cred
294 * @sk_peer_pid: &struct pid for this socket's peer
295 * @sk_peer_cred: %SO_PEERCRED setting
296 * @sk_rcvlowat: %SO_RCVLOWAT setting
297 * @sk_rcvtimeo: %SO_RCVTIMEO setting
298 * @sk_sndtimeo: %SO_SNDTIMEO setting
299 * @sk_txhash: computed flow hash for use on transmit
300 * @sk_txrehash: enable TX hash rethink
301 * @sk_filter: socket filtering instructions
302 * @sk_timer: sock cleanup timer
303 * @sk_stamp: time stamp of last packet received
304 * @sk_stamp_seq: lock for accessing sk_stamp on 32 bit architectures only
305 * @sk_tsflags: SO_TIMESTAMPING flags
306 * @sk_use_task_frag: allow sk_page_frag() to use current->task_frag.
307 * Sockets that can be used under memory reclaim should
309 * @sk_bind_phc: SO_TIMESTAMPING bind PHC index of PTP virtual clock
311 * @sk_tskey: counter to disambiguate concurrent tstamp requests
312 * @sk_zckey: counter to order MSG_ZEROCOPY notifications
313 * @sk_socket: Identd and reporting IO signals
314 * @sk_user_data: RPC layer private data. Write-protected by @sk_callback_lock.
315 * @sk_frag: cached page frag
316 * @sk_peek_off: current peek_offset value
317 * @sk_send_head: front of stuff to transmit
318 * @tcp_rtx_queue: TCP re-transmit queue [union with @sk_send_head]
319 * @sk_security: used by security modules
320 * @sk_mark: generic packet mark
321 * @sk_cgrp_data: cgroup data for this cgroup
322 * @sk_memcg: this socket's memory cgroup association
323 * @sk_write_pending: a write to stream socket waits to start
324 * @sk_disconnects: number of disconnect operations performed on this sock
325 * @sk_state_change: callback to indicate change in the state of the sock
326 * @sk_data_ready: callback to indicate there is data to be processed
327 * @sk_write_space: callback to indicate there is bf sending space available
328 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
329 * @sk_backlog_rcv: callback to process the backlog
330 * @sk_validate_xmit_skb: ptr to an optional validate function
331 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
332 * @sk_reuseport_cb: reuseport group container
333 * @sk_bpf_storage: ptr to cache and control for bpf_sk_storage
334 * @sk_rcu: used during RCU grace period
335 * @sk_clockid: clockid used by time-based scheduling (SO_TXTIME)
336 * @sk_txtime_deadline_mode: set deadline mode for SO_TXTIME
337 * @sk_txtime_report_errors: set report errors mode for SO_TXTIME
338 * @sk_txtime_unused: unused txtime flags
339 * @ns_tracker: tracker for netns reference
343 * Now struct inet_timewait_sock also uses sock_common, so please just
344 * don't add nothing before this first member (__sk_common) --acme
346 struct sock_common __sk_common;
347 #define sk_node __sk_common.skc_node
348 #define sk_nulls_node __sk_common.skc_nulls_node
349 #define sk_refcnt __sk_common.skc_refcnt
350 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
351 #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
352 #define sk_rx_queue_mapping __sk_common.skc_rx_queue_mapping
355 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
356 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
357 #define sk_hash __sk_common.skc_hash
358 #define sk_portpair __sk_common.skc_portpair
359 #define sk_num __sk_common.skc_num
360 #define sk_dport __sk_common.skc_dport
361 #define sk_addrpair __sk_common.skc_addrpair
362 #define sk_daddr __sk_common.skc_daddr
363 #define sk_rcv_saddr __sk_common.skc_rcv_saddr
364 #define sk_family __sk_common.skc_family
365 #define sk_state __sk_common.skc_state
366 #define sk_reuse __sk_common.skc_reuse
367 #define sk_reuseport __sk_common.skc_reuseport
368 #define sk_ipv6only __sk_common.skc_ipv6only
369 #define sk_net_refcnt __sk_common.skc_net_refcnt
370 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
371 #define sk_bind_node __sk_common.skc_bind_node
372 #define sk_prot __sk_common.skc_prot
373 #define sk_net __sk_common.skc_net
374 #define sk_v6_daddr __sk_common.skc_v6_daddr
375 #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
376 #define sk_cookie __sk_common.skc_cookie
377 #define sk_incoming_cpu __sk_common.skc_incoming_cpu
378 #define sk_flags __sk_common.skc_flags
379 #define sk_rxhash __sk_common.skc_rxhash
381 __cacheline_group_begin(sock_write_rx);
385 struct sk_buff_head sk_error_queue;
386 struct sk_buff_head sk_receive_queue;
388 * The backlog queue is special, it is always used with
389 * the per-socket spinlock held and requires low latency
390 * access. Therefore we special case it's implementation.
391 * Note : rmem_alloc is in this structure to fill a hole
392 * on 64bit arches, not because its logically part of
398 struct sk_buff *head;
399 struct sk_buff *tail;
401 #define sk_rmem_alloc sk_backlog.rmem_alloc
403 __cacheline_group_end(sock_write_rx);
405 __cacheline_group_begin(sock_read_rx);
406 /* early demux fields */
407 struct dst_entry __rcu *sk_rx_dst;
408 int sk_rx_dst_ifindex;
409 u32 sk_rx_dst_cookie;
411 #ifdef CONFIG_NET_RX_BUSY_POLL
412 unsigned int sk_ll_usec;
413 unsigned int sk_napi_id;
414 u16 sk_busy_poll_budget;
415 u8 sk_prefer_busy_poll;
420 struct sk_filter __rcu *sk_filter;
422 struct socket_wq __rcu *sk_wq;
424 struct socket_wq *sk_wq_raw;
428 void (*sk_data_ready)(struct sock *sk);
431 __cacheline_group_end(sock_read_rx);
433 __cacheline_group_begin(sock_read_rxtx);
435 struct socket *sk_socket;
436 struct mem_cgroup *sk_memcg;
438 struct xfrm_policy __rcu *sk_policy[2];
440 __cacheline_group_end(sock_read_rxtx);
442 __cacheline_group_begin(sock_write_rxtx);
443 socket_lock_t sk_lock;
445 int sk_forward_alloc;
447 __cacheline_group_end(sock_write_rxtx);
449 __cacheline_group_begin(sock_write_tx);
450 int sk_write_pending;
451 atomic_t sk_omem_alloc;
455 refcount_t sk_wmem_alloc;
456 unsigned long sk_tsq_flags;
458 struct sk_buff *sk_send_head;
459 struct rb_root tcp_rtx_queue;
461 struct sk_buff_head sk_write_queue;
462 u32 sk_dst_pending_confirm;
463 u32 sk_pacing_status; /* see enum sk_pacing */
464 struct page_frag sk_frag;
465 struct timer_list sk_timer;
467 unsigned long sk_pacing_rate; /* bytes per second */
470 __cacheline_group_end(sock_write_tx);
472 __cacheline_group_begin(sock_read_tx);
473 unsigned long sk_max_pacing_rate;
477 struct dst_entry __rcu *sk_dst_cache;
478 netdev_features_t sk_route_caps;
479 #ifdef CONFIG_SOCK_VALIDATE_XMIT
480 struct sk_buff* (*sk_validate_xmit_skb)(struct sock *sk,
481 struct net_device *dev,
482 struct sk_buff *skb);
486 unsigned int sk_gso_max_size;
490 bool sk_use_task_frag;
491 __cacheline_group_end(sock_read_tx);
494 * Because of non atomicity rules, all
495 * changes are protected by socket lock.
497 u8 sk_gso_disabled : 1,
504 unsigned long sk_lingertime;
505 struct proto *sk_prot_creator;
506 rwlock_t sk_callback_lock;
509 u32 sk_max_ack_backlog;
511 spinlock_t sk_peer_lock;
513 struct pid *sk_peer_pid;
514 const struct cred *sk_peer_cred;
517 #if BITS_PER_LONG==32
518 seqlock_t sk_stamp_seq;
524 u8 sk_txtime_deadline_mode : 1,
525 sk_txtime_report_errors : 1,
526 sk_txtime_unused : 6;
529 #ifdef CONFIG_SECURITY
532 struct sock_cgroup_data sk_cgrp_data;
533 void (*sk_state_change)(struct sock *sk);
534 void (*sk_write_space)(struct sock *sk);
535 void (*sk_error_report)(struct sock *sk);
536 int (*sk_backlog_rcv)(struct sock *sk,
537 struct sk_buff *skb);
538 void (*sk_destruct)(struct sock *sk);
539 struct sock_reuseport __rcu *sk_reuseport_cb;
540 #ifdef CONFIG_BPF_SYSCALL
541 struct bpf_local_storage __rcu *sk_bpf_storage;
543 struct rcu_head sk_rcu;
544 netns_tracker ns_tracker;
549 SK_PACING_NEEDED = 1,
553 /* flag bits in sk_user_data
555 * - SK_USER_DATA_NOCOPY: Pointer stored in sk_user_data might
556 * not be suitable for copying when cloning the socket. For instance,
557 * it can point to a reference counted object. sk_user_data bottom
558 * bit is set if pointer must not be copied.
560 * - SK_USER_DATA_BPF: Mark whether sk_user_data field is
561 * managed/owned by a BPF reuseport array. This bit should be set
562 * when sk_user_data's sk is added to the bpf's reuseport_array.
564 * - SK_USER_DATA_PSOCK: Mark whether pointer stored in
565 * sk_user_data points to psock type. This bit should be set
566 * when sk_user_data is assigned to a psock object.
568 #define SK_USER_DATA_NOCOPY 1UL
569 #define SK_USER_DATA_BPF 2UL
570 #define SK_USER_DATA_PSOCK 4UL
571 #define SK_USER_DATA_PTRMASK ~(SK_USER_DATA_NOCOPY | SK_USER_DATA_BPF |\
575 * sk_user_data_is_nocopy - Test if sk_user_data pointer must not be copied
578 static inline bool sk_user_data_is_nocopy(const struct sock *sk)
580 return ((uintptr_t)sk->sk_user_data & SK_USER_DATA_NOCOPY);
583 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
586 * __locked_read_sk_user_data_with_flags - return the pointer
587 * only if argument flags all has been set in sk_user_data. Otherwise
593 * The caller must be holding sk->sk_callback_lock.
596 __locked_read_sk_user_data_with_flags(const struct sock *sk,
599 uintptr_t sk_user_data =
600 (uintptr_t)rcu_dereference_check(__sk_user_data(sk),
601 lockdep_is_held(&sk->sk_callback_lock));
603 WARN_ON_ONCE(flags & SK_USER_DATA_PTRMASK);
605 if ((sk_user_data & flags) == flags)
606 return (void *)(sk_user_data & SK_USER_DATA_PTRMASK);
611 * __rcu_dereference_sk_user_data_with_flags - return the pointer
612 * only if argument flags all has been set in sk_user_data. Otherwise
619 __rcu_dereference_sk_user_data_with_flags(const struct sock *sk,
622 uintptr_t sk_user_data = (uintptr_t)rcu_dereference(__sk_user_data(sk));
624 WARN_ON_ONCE(flags & SK_USER_DATA_PTRMASK);
626 if ((sk_user_data & flags) == flags)
627 return (void *)(sk_user_data & SK_USER_DATA_PTRMASK);
631 #define rcu_dereference_sk_user_data(sk) \
632 __rcu_dereference_sk_user_data_with_flags(sk, 0)
633 #define __rcu_assign_sk_user_data_with_flags(sk, ptr, flags) \
635 uintptr_t __tmp1 = (uintptr_t)(ptr), \
636 __tmp2 = (uintptr_t)(flags); \
637 WARN_ON_ONCE(__tmp1 & ~SK_USER_DATA_PTRMASK); \
638 WARN_ON_ONCE(__tmp2 & SK_USER_DATA_PTRMASK); \
639 rcu_assign_pointer(__sk_user_data((sk)), \
642 #define rcu_assign_sk_user_data(sk, ptr) \
643 __rcu_assign_sk_user_data_with_flags(sk, ptr, 0)
646 struct net *sock_net(const struct sock *sk)
648 return read_pnet(&sk->sk_net);
652 void sock_net_set(struct sock *sk, struct net *net)
654 write_pnet(&sk->sk_net, net);
658 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
659 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
660 * on a socket means that the socket will reuse everybody else's port
661 * without looking at the other's sk_reuse value.
664 #define SK_NO_REUSE 0
665 #define SK_CAN_REUSE 1
666 #define SK_FORCE_REUSE 2
668 int sk_set_peek_off(struct sock *sk, int val);
670 static inline int sk_peek_offset(const struct sock *sk, int flags)
672 if (unlikely(flags & MSG_PEEK)) {
673 return READ_ONCE(sk->sk_peek_off);
679 static inline void sk_peek_offset_bwd(struct sock *sk, int val)
681 s32 off = READ_ONCE(sk->sk_peek_off);
683 if (unlikely(off >= 0)) {
684 off = max_t(s32, off - val, 0);
685 WRITE_ONCE(sk->sk_peek_off, off);
689 static inline void sk_peek_offset_fwd(struct sock *sk, int val)
691 sk_peek_offset_bwd(sk, -val);
695 * Hashed lists helper routines
697 static inline struct sock *sk_entry(const struct hlist_node *node)
699 return hlist_entry(node, struct sock, sk_node);
702 static inline struct sock *__sk_head(const struct hlist_head *head)
704 return hlist_entry(head->first, struct sock, sk_node);
707 static inline struct sock *sk_head(const struct hlist_head *head)
709 return hlist_empty(head) ? NULL : __sk_head(head);
712 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
714 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
717 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
719 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
722 static inline struct sock *sk_next(const struct sock *sk)
724 return hlist_entry_safe(sk->sk_node.next, struct sock, sk_node);
727 static inline struct sock *sk_nulls_next(const struct sock *sk)
729 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
730 hlist_nulls_entry(sk->sk_nulls_node.next,
731 struct sock, sk_nulls_node) :
735 static inline bool sk_unhashed(const struct sock *sk)
737 return hlist_unhashed(&sk->sk_node);
740 static inline bool sk_hashed(const struct sock *sk)
742 return !sk_unhashed(sk);
745 static inline void sk_node_init(struct hlist_node *node)
750 static inline void __sk_del_node(struct sock *sk)
752 __hlist_del(&sk->sk_node);
755 /* NB: equivalent to hlist_del_init_rcu */
756 static inline bool __sk_del_node_init(struct sock *sk)
760 sk_node_init(&sk->sk_node);
766 /* Grab socket reference count. This operation is valid only
767 when sk is ALREADY grabbed f.e. it is found in hash table
768 or a list and the lookup is made under lock preventing hash table
772 static __always_inline void sock_hold(struct sock *sk)
774 refcount_inc(&sk->sk_refcnt);
777 /* Ungrab socket in the context, which assumes that socket refcnt
778 cannot hit zero, f.e. it is true in context of any socketcall.
780 static __always_inline void __sock_put(struct sock *sk)
782 refcount_dec(&sk->sk_refcnt);
785 static inline bool sk_del_node_init(struct sock *sk)
787 bool rc = __sk_del_node_init(sk);
790 /* paranoid for a while -acme */
791 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
796 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
798 static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
801 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
807 static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
809 bool rc = __sk_nulls_del_node_init_rcu(sk);
812 /* paranoid for a while -acme */
813 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
819 static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
821 hlist_add_head(&sk->sk_node, list);
824 static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
827 __sk_add_node(sk, list);
830 static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
833 if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport &&
834 sk->sk_family == AF_INET6)
835 hlist_add_tail_rcu(&sk->sk_node, list);
837 hlist_add_head_rcu(&sk->sk_node, list);
840 static inline void sk_add_node_tail_rcu(struct sock *sk, struct hlist_head *list)
843 hlist_add_tail_rcu(&sk->sk_node, list);
846 static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
848 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
851 static inline void __sk_nulls_add_node_tail_rcu(struct sock *sk, struct hlist_nulls_head *list)
853 hlist_nulls_add_tail_rcu(&sk->sk_nulls_node, list);
856 static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
859 __sk_nulls_add_node_rcu(sk, list);
862 static inline void __sk_del_bind_node(struct sock *sk)
864 __hlist_del(&sk->sk_bind_node);
867 static inline void sk_add_bind_node(struct sock *sk,
868 struct hlist_head *list)
870 hlist_add_head(&sk->sk_bind_node, list);
873 #define sk_for_each(__sk, list) \
874 hlist_for_each_entry(__sk, list, sk_node)
875 #define sk_for_each_rcu(__sk, list) \
876 hlist_for_each_entry_rcu(__sk, list, sk_node)
877 #define sk_nulls_for_each(__sk, node, list) \
878 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
879 #define sk_nulls_for_each_rcu(__sk, node, list) \
880 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
881 #define sk_for_each_from(__sk) \
882 hlist_for_each_entry_from(__sk, sk_node)
883 #define sk_nulls_for_each_from(__sk, node) \
884 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
885 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
886 #define sk_for_each_safe(__sk, tmp, list) \
887 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
888 #define sk_for_each_bound(__sk, list) \
889 hlist_for_each_entry(__sk, list, sk_bind_node)
892 * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset
893 * @tpos: the type * to use as a loop cursor.
894 * @pos: the &struct hlist_node to use as a loop cursor.
895 * @head: the head for your list.
896 * @offset: offset of hlist_node within the struct.
899 #define sk_for_each_entry_offset_rcu(tpos, pos, head, offset) \
900 for (pos = rcu_dereference(hlist_first_rcu(head)); \
902 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
903 pos = rcu_dereference(hlist_next_rcu(pos)))
905 static inline struct user_namespace *sk_user_ns(const struct sock *sk)
907 /* Careful only use this in a context where these parameters
908 * can not change and must all be valid, such as recvmsg from
911 return sk->sk_socket->file->f_cred->user_ns;
925 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
926 SOCK_DBG, /* %SO_DEBUG setting */
927 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
928 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
929 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
930 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
931 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
932 SOCK_FASYNC, /* fasync() active */
934 SOCK_ZEROCOPY, /* buffers from userspace */
935 SOCK_WIFI_STATUS, /* push wifi status to userspace */
936 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
937 * Will use last 4 bytes of packet sent from
938 * user-space instead.
940 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
941 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
942 SOCK_RCU_FREE, /* wait rcu grace period in sk_destruct() */
944 SOCK_XDP, /* XDP is attached */
945 SOCK_TSTAMP_NEW, /* Indicates 64 bit timestamps always */
946 SOCK_RCVMARK, /* Receive SO_MARK ancillary data with packet */
949 #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
951 static inline void sock_copy_flags(struct sock *nsk, const struct sock *osk)
953 nsk->sk_flags = osk->sk_flags;
956 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
958 __set_bit(flag, &sk->sk_flags);
961 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
963 __clear_bit(flag, &sk->sk_flags);
966 static inline void sock_valbool_flag(struct sock *sk, enum sock_flags bit,
970 sock_set_flag(sk, bit);
972 sock_reset_flag(sk, bit);
975 static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
977 return test_bit(flag, &sk->sk_flags);
981 DECLARE_STATIC_KEY_FALSE(memalloc_socks_key);
982 static inline int sk_memalloc_socks(void)
984 return static_branch_unlikely(&memalloc_socks_key);
987 void __receive_sock(struct file *file);
990 static inline int sk_memalloc_socks(void)
995 static inline void __receive_sock(struct file *file)
999 static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask)
1001 return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC);
1004 static inline void sk_acceptq_removed(struct sock *sk)
1006 WRITE_ONCE(sk->sk_ack_backlog, sk->sk_ack_backlog - 1);
1009 static inline void sk_acceptq_added(struct sock *sk)
1011 WRITE_ONCE(sk->sk_ack_backlog, sk->sk_ack_backlog + 1);
1014 /* Note: If you think the test should be:
1015 * return READ_ONCE(sk->sk_ack_backlog) >= READ_ONCE(sk->sk_max_ack_backlog);
1016 * Then please take a look at commit 64a146513f8f ("[NET]: Revert incorrect accept queue backlog changes.")
1018 static inline bool sk_acceptq_is_full(const struct sock *sk)
1020 return READ_ONCE(sk->sk_ack_backlog) > READ_ONCE(sk->sk_max_ack_backlog);
1024 * Compute minimal free write space needed to queue new packets.
1026 static inline int sk_stream_min_wspace(const struct sock *sk)
1028 return READ_ONCE(sk->sk_wmem_queued) >> 1;
1031 static inline int sk_stream_wspace(const struct sock *sk)
1033 return READ_ONCE(sk->sk_sndbuf) - READ_ONCE(sk->sk_wmem_queued);
1036 static inline void sk_wmem_queued_add(struct sock *sk, int val)
1038 WRITE_ONCE(sk->sk_wmem_queued, sk->sk_wmem_queued + val);
1041 static inline void sk_forward_alloc_add(struct sock *sk, int val)
1043 /* Paired with lockless reads of sk->sk_forward_alloc */
1044 WRITE_ONCE(sk->sk_forward_alloc, sk->sk_forward_alloc + val);
1047 void sk_stream_write_space(struct sock *sk);
1049 /* OOB backlog add */
1050 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
1052 /* dont let skb dst not refcounted, we are going to leave rcu lock */
1055 if (!sk->sk_backlog.tail)
1056 WRITE_ONCE(sk->sk_backlog.head, skb);
1058 sk->sk_backlog.tail->next = skb;
1060 WRITE_ONCE(sk->sk_backlog.tail, skb);
1065 * Take into account size of receive queue and backlog queue
1066 * Do not take into account this skb truesize,
1067 * to allow even a single big packet to come.
1069 static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
1071 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
1073 return qsize > limit;
1076 /* The per-socket spinlock must be held here. */
1077 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
1080 if (sk_rcvqueues_full(sk, limit))
1084 * If the skb was allocated from pfmemalloc reserves, only
1085 * allow SOCK_MEMALLOC sockets to use it as this socket is
1086 * helping free memory
1088 if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC))
1091 __sk_add_backlog(sk, skb);
1092 sk->sk_backlog.len += skb->truesize;
1096 int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
1098 INDIRECT_CALLABLE_DECLARE(int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb));
1099 INDIRECT_CALLABLE_DECLARE(int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb));
1101 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1103 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
1104 return __sk_backlog_rcv(sk, skb);
1106 return INDIRECT_CALL_INET(sk->sk_backlog_rcv,
1112 static inline void sk_incoming_cpu_update(struct sock *sk)
1114 int cpu = raw_smp_processor_id();
1116 if (unlikely(READ_ONCE(sk->sk_incoming_cpu) != cpu))
1117 WRITE_ONCE(sk->sk_incoming_cpu, cpu);
1121 static inline void sock_rps_save_rxhash(struct sock *sk,
1122 const struct sk_buff *skb)
1125 /* The following WRITE_ONCE() is paired with the READ_ONCE()
1126 * here, and another one in sock_rps_record_flow().
1128 if (unlikely(READ_ONCE(sk->sk_rxhash) != skb->hash))
1129 WRITE_ONCE(sk->sk_rxhash, skb->hash);
1133 static inline void sock_rps_reset_rxhash(struct sock *sk)
1136 /* Paired with READ_ONCE() in sock_rps_record_flow() */
1137 WRITE_ONCE(sk->sk_rxhash, 0);
1141 #define sk_wait_event(__sk, __timeo, __condition, __wait) \
1142 ({ int __rc, __dis = __sk->sk_disconnects; \
1143 release_sock(__sk); \
1144 __rc = __condition; \
1146 *(__timeo) = wait_woken(__wait, \
1147 TASK_INTERRUPTIBLE, \
1150 sched_annotate_sleep(); \
1152 __rc = __dis == __sk->sk_disconnects ? __condition : -EPIPE; \
1156 int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
1157 int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
1158 void sk_stream_wait_close(struct sock *sk, long timeo_p);
1159 int sk_stream_error(struct sock *sk, int flags, int err);
1160 void sk_stream_kill_queues(struct sock *sk);
1161 void sk_set_memalloc(struct sock *sk);
1162 void sk_clear_memalloc(struct sock *sk);
1164 void __sk_flush_backlog(struct sock *sk);
1166 static inline bool sk_flush_backlog(struct sock *sk)
1168 if (unlikely(READ_ONCE(sk->sk_backlog.tail))) {
1169 __sk_flush_backlog(sk);
1175 int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb);
1177 struct request_sock_ops;
1178 struct timewait_sock_ops;
1179 struct inet_hashinfo;
1180 struct raw_hashinfo;
1181 struct smc_hashinfo;
1186 * caches using SLAB_TYPESAFE_BY_RCU should let .next pointer from nulls nodes
1187 * un-modified. Special care is taken when initializing object to zero.
1189 static inline void sk_prot_clear_nulls(struct sock *sk, int size)
1191 if (offsetof(struct sock, sk_node.next) != 0)
1192 memset(sk, 0, offsetof(struct sock, sk_node.next));
1193 memset(&sk->sk_node.pprev, 0,
1194 size - offsetof(struct sock, sk_node.pprev));
1197 /* Networking protocol blocks we attach to sockets.
1198 * socket layer -> transport layer interface
1201 void (*close)(struct sock *sk,
1203 int (*pre_connect)(struct sock *sk,
1204 struct sockaddr *uaddr,
1206 int (*connect)(struct sock *sk,
1207 struct sockaddr *uaddr,
1209 int (*disconnect)(struct sock *sk, int flags);
1211 struct sock * (*accept)(struct sock *sk, int flags, int *err,
1214 int (*ioctl)(struct sock *sk, int cmd,
1216 int (*init)(struct sock *sk);
1217 void (*destroy)(struct sock *sk);
1218 void (*shutdown)(struct sock *sk, int how);
1219 int (*setsockopt)(struct sock *sk, int level,
1220 int optname, sockptr_t optval,
1221 unsigned int optlen);
1222 int (*getsockopt)(struct sock *sk, int level,
1223 int optname, char __user *optval,
1224 int __user *option);
1225 void (*keepalive)(struct sock *sk, int valbool);
1226 #ifdef CONFIG_COMPAT
1227 int (*compat_ioctl)(struct sock *sk,
1228 unsigned int cmd, unsigned long arg);
1230 int (*sendmsg)(struct sock *sk, struct msghdr *msg,
1232 int (*recvmsg)(struct sock *sk, struct msghdr *msg,
1233 size_t len, int flags, int *addr_len);
1234 void (*splice_eof)(struct socket *sock);
1235 int (*bind)(struct sock *sk,
1236 struct sockaddr *addr, int addr_len);
1237 int (*bind_add)(struct sock *sk,
1238 struct sockaddr *addr, int addr_len);
1240 int (*backlog_rcv) (struct sock *sk,
1241 struct sk_buff *skb);
1242 bool (*bpf_bypass_getsockopt)(int level,
1245 void (*release_cb)(struct sock *sk);
1247 /* Keeping track of sk's, looking them up, and port selection methods. */
1248 int (*hash)(struct sock *sk);
1249 void (*unhash)(struct sock *sk);
1250 void (*rehash)(struct sock *sk);
1251 int (*get_port)(struct sock *sk, unsigned short snum);
1252 void (*put_port)(struct sock *sk);
1253 #ifdef CONFIG_BPF_SYSCALL
1254 int (*psock_update_sk_prot)(struct sock *sk,
1255 struct sk_psock *psock,
1259 /* Keeping track of sockets in use */
1260 #ifdef CONFIG_PROC_FS
1261 unsigned int inuse_idx;
1264 #if IS_ENABLED(CONFIG_MPTCP)
1265 int (*forward_alloc_get)(const struct sock *sk);
1268 bool (*stream_memory_free)(const struct sock *sk, int wake);
1269 bool (*sock_is_readable)(struct sock *sk);
1270 /* Memory pressure */
1271 void (*enter_memory_pressure)(struct sock *sk);
1272 void (*leave_memory_pressure)(struct sock *sk);
1273 atomic_long_t *memory_allocated; /* Current allocated memory. */
1274 int __percpu *per_cpu_fw_alloc;
1275 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
1278 * Pressure flag: try to collapse.
1279 * Technical note: it is used by multiple contexts non atomically.
1280 * Make sure to use READ_ONCE()/WRITE_ONCE() for all reads/writes.
1281 * All the __sk_mem_schedule() is of this nature: accounting
1282 * is strict, actions are advisory and have some latency.
1284 unsigned long *memory_pressure;
1289 u32 sysctl_wmem_offset;
1290 u32 sysctl_rmem_offset;
1295 struct kmem_cache *slab;
1296 unsigned int obj_size;
1297 unsigned int ipv6_pinfo_offset;
1298 slab_flags_t slab_flags;
1299 unsigned int useroffset; /* Usercopy region offset */
1300 unsigned int usersize; /* Usercopy region size */
1302 unsigned int __percpu *orphan_count;
1304 struct request_sock_ops *rsk_prot;
1305 struct timewait_sock_ops *twsk_prot;
1308 struct inet_hashinfo *hashinfo;
1309 struct udp_table *udp_table;
1310 struct raw_hashinfo *raw_hash;
1311 struct smc_hashinfo *smc_hash;
1314 struct module *owner;
1318 struct list_head node;
1319 int (*diag_destroy)(struct sock *sk, int err);
1320 } __randomize_layout;
1322 int proto_register(struct proto *prot, int alloc_slab);
1323 void proto_unregister(struct proto *prot);
1324 int sock_load_diag_module(int family, int protocol);
1326 INDIRECT_CALLABLE_DECLARE(bool tcp_stream_memory_free(const struct sock *sk, int wake));
1328 static inline int sk_forward_alloc_get(const struct sock *sk)
1330 #if IS_ENABLED(CONFIG_MPTCP)
1331 if (sk->sk_prot->forward_alloc_get)
1332 return sk->sk_prot->forward_alloc_get(sk);
1334 return READ_ONCE(sk->sk_forward_alloc);
1337 static inline bool __sk_stream_memory_free(const struct sock *sk, int wake)
1339 if (READ_ONCE(sk->sk_wmem_queued) >= READ_ONCE(sk->sk_sndbuf))
1342 return sk->sk_prot->stream_memory_free ?
1343 INDIRECT_CALL_INET_1(sk->sk_prot->stream_memory_free,
1344 tcp_stream_memory_free, sk, wake) : true;
1347 static inline bool sk_stream_memory_free(const struct sock *sk)
1349 return __sk_stream_memory_free(sk, 0);
1352 static inline bool __sk_stream_is_writeable(const struct sock *sk, int wake)
1354 return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1355 __sk_stream_memory_free(sk, wake);
1358 static inline bool sk_stream_is_writeable(const struct sock *sk)
1360 return __sk_stream_is_writeable(sk, 0);
1363 static inline int sk_under_cgroup_hierarchy(struct sock *sk,
1364 struct cgroup *ancestor)
1366 #ifdef CONFIG_SOCK_CGROUP_DATA
1367 return cgroup_is_descendant(sock_cgroup_ptr(&sk->sk_cgrp_data),
1374 static inline bool sk_has_memory_pressure(const struct sock *sk)
1376 return sk->sk_prot->memory_pressure != NULL;
1379 static inline bool sk_under_global_memory_pressure(const struct sock *sk)
1381 return sk->sk_prot->memory_pressure &&
1382 !!READ_ONCE(*sk->sk_prot->memory_pressure);
1385 static inline bool sk_under_memory_pressure(const struct sock *sk)
1387 if (!sk->sk_prot->memory_pressure)
1390 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
1391 mem_cgroup_under_socket_pressure(sk->sk_memcg))
1394 return !!READ_ONCE(*sk->sk_prot->memory_pressure);
1398 proto_memory_allocated(const struct proto *prot)
1400 return max(0L, atomic_long_read(prot->memory_allocated));
1404 sk_memory_allocated(const struct sock *sk)
1406 return proto_memory_allocated(sk->sk_prot);
1409 /* 1 MB per cpu, in page units */
1410 #define SK_MEMORY_PCPU_RESERVE (1 << (20 - PAGE_SHIFT))
1411 extern int sysctl_mem_pcpu_rsv;
1414 sk_memory_allocated_add(struct sock *sk, int amt)
1419 local_reserve = __this_cpu_add_return(*sk->sk_prot->per_cpu_fw_alloc, amt);
1420 if (local_reserve >= READ_ONCE(sysctl_mem_pcpu_rsv)) {
1421 __this_cpu_sub(*sk->sk_prot->per_cpu_fw_alloc, local_reserve);
1422 atomic_long_add(local_reserve, sk->sk_prot->memory_allocated);
1428 sk_memory_allocated_sub(struct sock *sk, int amt)
1433 local_reserve = __this_cpu_sub_return(*sk->sk_prot->per_cpu_fw_alloc, amt);
1434 if (local_reserve <= -READ_ONCE(sysctl_mem_pcpu_rsv)) {
1435 __this_cpu_sub(*sk->sk_prot->per_cpu_fw_alloc, local_reserve);
1436 atomic_long_add(local_reserve, sk->sk_prot->memory_allocated);
1441 #define SK_ALLOC_PERCPU_COUNTER_BATCH 16
1443 static inline void sk_sockets_allocated_dec(struct sock *sk)
1445 percpu_counter_add_batch(sk->sk_prot->sockets_allocated, -1,
1446 SK_ALLOC_PERCPU_COUNTER_BATCH);
1449 static inline void sk_sockets_allocated_inc(struct sock *sk)
1451 percpu_counter_add_batch(sk->sk_prot->sockets_allocated, 1,
1452 SK_ALLOC_PERCPU_COUNTER_BATCH);
1456 sk_sockets_allocated_read_positive(struct sock *sk)
1458 return percpu_counter_read_positive(sk->sk_prot->sockets_allocated);
1462 proto_sockets_allocated_sum_positive(struct proto *prot)
1464 return percpu_counter_sum_positive(prot->sockets_allocated);
1468 proto_memory_pressure(struct proto *prot)
1470 if (!prot->memory_pressure)
1472 return !!READ_ONCE(*prot->memory_pressure);
1476 #ifdef CONFIG_PROC_FS
1477 #define PROTO_INUSE_NR 64 /* should be enough for the first time */
1480 int val[PROTO_INUSE_NR];
1483 static inline void sock_prot_inuse_add(const struct net *net,
1484 const struct proto *prot, int val)
1486 this_cpu_add(net->core.prot_inuse->val[prot->inuse_idx], val);
1489 static inline void sock_inuse_add(const struct net *net, int val)
1491 this_cpu_add(net->core.prot_inuse->all, val);
1494 int sock_prot_inuse_get(struct net *net, struct proto *proto);
1495 int sock_inuse_get(struct net *net);
1497 static inline void sock_prot_inuse_add(const struct net *net,
1498 const struct proto *prot, int val)
1502 static inline void sock_inuse_add(const struct net *net, int val)
1508 /* With per-bucket locks this operation is not-atomic, so that
1509 * this version is not worse.
1511 static inline int __sk_prot_rehash(struct sock *sk)
1513 sk->sk_prot->unhash(sk);
1514 return sk->sk_prot->hash(sk);
1517 /* About 10 seconds */
1518 #define SOCK_DESTROY_TIME (10*HZ)
1520 /* Sockets 0-1023 can't be bound to unless you are superuser */
1521 #define PROT_SOCK 1024
1523 #define SHUTDOWN_MASK 3
1524 #define RCV_SHUTDOWN 1
1525 #define SEND_SHUTDOWN 2
1527 #define SOCK_BINDADDR_LOCK 4
1528 #define SOCK_BINDPORT_LOCK 8
1530 struct socket_alloc {
1531 struct socket socket;
1532 struct inode vfs_inode;
1535 static inline struct socket *SOCKET_I(struct inode *inode)
1537 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1540 static inline struct inode *SOCK_INODE(struct socket *socket)
1542 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1546 * Functions for memory accounting
1548 int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind);
1549 int __sk_mem_schedule(struct sock *sk, int size, int kind);
1550 void __sk_mem_reduce_allocated(struct sock *sk, int amount);
1551 void __sk_mem_reclaim(struct sock *sk, int amount);
1553 #define SK_MEM_SEND 0
1554 #define SK_MEM_RECV 1
1556 /* sysctl_mem values are in pages */
1557 static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1559 return READ_ONCE(sk->sk_prot->sysctl_mem[index]);
1562 static inline int sk_mem_pages(int amt)
1564 return (amt + PAGE_SIZE - 1) >> PAGE_SHIFT;
1567 static inline bool sk_has_account(struct sock *sk)
1569 /* return true if protocol supports memory accounting */
1570 return !!sk->sk_prot->memory_allocated;
1573 static inline bool sk_wmem_schedule(struct sock *sk, int size)
1577 if (!sk_has_account(sk))
1579 delta = size - sk->sk_forward_alloc;
1580 return delta <= 0 || __sk_mem_schedule(sk, delta, SK_MEM_SEND);
1584 sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
1588 if (!sk_has_account(sk))
1590 delta = size - sk->sk_forward_alloc;
1591 return delta <= 0 || __sk_mem_schedule(sk, delta, SK_MEM_RECV) ||
1592 skb_pfmemalloc(skb);
1595 static inline int sk_unused_reserved_mem(const struct sock *sk)
1599 if (likely(!sk->sk_reserved_mem))
1602 unused_mem = sk->sk_reserved_mem - sk->sk_wmem_queued -
1603 atomic_read(&sk->sk_rmem_alloc);
1605 return unused_mem > 0 ? unused_mem : 0;
1608 static inline void sk_mem_reclaim(struct sock *sk)
1612 if (!sk_has_account(sk))
1615 reclaimable = sk->sk_forward_alloc - sk_unused_reserved_mem(sk);
1617 if (reclaimable >= (int)PAGE_SIZE)
1618 __sk_mem_reclaim(sk, reclaimable);
1621 static inline void sk_mem_reclaim_final(struct sock *sk)
1623 sk->sk_reserved_mem = 0;
1627 static inline void sk_mem_charge(struct sock *sk, int size)
1629 if (!sk_has_account(sk))
1631 sk_forward_alloc_add(sk, -size);
1634 static inline void sk_mem_uncharge(struct sock *sk, int size)
1636 if (!sk_has_account(sk))
1638 sk_forward_alloc_add(sk, size);
1643 * Macro so as to not evaluate some arguments when
1644 * lockdep is not enabled.
1646 * Mark both the sk_lock and the sk_lock.slock as a
1647 * per-address-family lock class.
1649 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1651 sk->sk_lock.owned = 0; \
1652 init_waitqueue_head(&sk->sk_lock.wq); \
1653 spin_lock_init(&(sk)->sk_lock.slock); \
1654 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1655 sizeof((sk)->sk_lock)); \
1656 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1658 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1661 static inline bool lockdep_sock_is_held(const struct sock *sk)
1663 return lockdep_is_held(&sk->sk_lock) ||
1664 lockdep_is_held(&sk->sk_lock.slock);
1667 void lock_sock_nested(struct sock *sk, int subclass);
1669 static inline void lock_sock(struct sock *sk)
1671 lock_sock_nested(sk, 0);
1674 void __lock_sock(struct sock *sk);
1675 void __release_sock(struct sock *sk);
1676 void release_sock(struct sock *sk);
1678 /* BH context may only use the following locking interface. */
1679 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1680 #define bh_lock_sock_nested(__sk) \
1681 spin_lock_nested(&((__sk)->sk_lock.slock), \
1682 SINGLE_DEPTH_NESTING)
1683 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1685 bool __lock_sock_fast(struct sock *sk) __acquires(&sk->sk_lock.slock);
1688 * lock_sock_fast - fast version of lock_sock
1691 * This version should be used for very small section, where process wont block
1692 * return false if fast path is taken:
1694 * sk_lock.slock locked, owned = 0, BH disabled
1696 * return true if slow path is taken:
1698 * sk_lock.slock unlocked, owned = 1, BH enabled
1700 static inline bool lock_sock_fast(struct sock *sk)
1702 /* The sk_lock has mutex_lock() semantics here. */
1703 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
1705 return __lock_sock_fast(sk);
1708 /* fast socket lock variant for caller already holding a [different] socket lock */
1709 static inline bool lock_sock_fast_nested(struct sock *sk)
1711 mutex_acquire(&sk->sk_lock.dep_map, SINGLE_DEPTH_NESTING, 0, _RET_IP_);
1713 return __lock_sock_fast(sk);
1717 * unlock_sock_fast - complement of lock_sock_fast
1721 * fast unlock socket for user context.
1722 * If slow mode is on, we call regular release_sock()
1724 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1725 __releases(&sk->sk_lock.slock)
1729 __release(&sk->sk_lock.slock);
1731 mutex_release(&sk->sk_lock.dep_map, _RET_IP_);
1732 spin_unlock_bh(&sk->sk_lock.slock);
1736 void sockopt_lock_sock(struct sock *sk);
1737 void sockopt_release_sock(struct sock *sk);
1738 bool sockopt_ns_capable(struct user_namespace *ns, int cap);
1739 bool sockopt_capable(int cap);
1741 /* Used by processes to "lock" a socket state, so that
1742 * interrupts and bottom half handlers won't change it
1743 * from under us. It essentially blocks any incoming
1744 * packets, so that we won't get any new data or any
1745 * packets that change the state of the socket.
1747 * While locked, BH processing will add new packets to
1748 * the backlog queue. This queue is processed by the
1749 * owner of the socket lock right before it is released.
1751 * Since ~2.3.5 it is also exclusive sleep lock serializing
1752 * accesses from user process context.
1755 static inline void sock_owned_by_me(const struct sock *sk)
1757 #ifdef CONFIG_LOCKDEP
1758 WARN_ON_ONCE(!lockdep_sock_is_held(sk) && debug_locks);
1762 static inline bool sock_owned_by_user(const struct sock *sk)
1764 sock_owned_by_me(sk);
1765 return sk->sk_lock.owned;
1768 static inline bool sock_owned_by_user_nocheck(const struct sock *sk)
1770 return sk->sk_lock.owned;
1773 static inline void sock_release_ownership(struct sock *sk)
1775 DEBUG_NET_WARN_ON_ONCE(!sock_owned_by_user_nocheck(sk));
1776 sk->sk_lock.owned = 0;
1778 /* The sk_lock has mutex_unlock() semantics: */
1779 mutex_release(&sk->sk_lock.dep_map, _RET_IP_);
1782 /* no reclassification while locks are held */
1783 static inline bool sock_allow_reclassification(const struct sock *csk)
1785 struct sock *sk = (struct sock *)csk;
1787 return !sock_owned_by_user_nocheck(sk) &&
1788 !spin_is_locked(&sk->sk_lock.slock);
1791 struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1792 struct proto *prot, int kern);
1793 void sk_free(struct sock *sk);
1794 void sk_destruct(struct sock *sk);
1795 struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
1796 void sk_free_unlock_clone(struct sock *sk);
1798 struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1800 void __sock_wfree(struct sk_buff *skb);
1801 void sock_wfree(struct sk_buff *skb);
1802 struct sk_buff *sock_omalloc(struct sock *sk, unsigned long size,
1804 void skb_orphan_partial(struct sk_buff *skb);
1805 void sock_rfree(struct sk_buff *skb);
1806 void sock_efree(struct sk_buff *skb);
1808 void sock_edemux(struct sk_buff *skb);
1809 void sock_pfree(struct sk_buff *skb);
1811 #define sock_edemux sock_efree
1814 int sk_setsockopt(struct sock *sk, int level, int optname,
1815 sockptr_t optval, unsigned int optlen);
1816 int sock_setsockopt(struct socket *sock, int level, int op,
1817 sockptr_t optval, unsigned int optlen);
1818 int do_sock_setsockopt(struct socket *sock, bool compat, int level,
1819 int optname, sockptr_t optval, int optlen);
1820 int do_sock_getsockopt(struct socket *sock, bool compat, int level,
1821 int optname, sockptr_t optval, sockptr_t optlen);
1823 int sk_getsockopt(struct sock *sk, int level, int optname,
1824 sockptr_t optval, sockptr_t optlen);
1825 int sock_gettstamp(struct socket *sock, void __user *userstamp,
1826 bool timeval, bool time32);
1827 struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1828 unsigned long data_len, int noblock,
1829 int *errcode, int max_page_order);
1831 static inline struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1833 int noblock, int *errcode)
1835 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
1838 void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1839 void sock_kfree_s(struct sock *sk, void *mem, int size);
1840 void sock_kzfree_s(struct sock *sk, void *mem, int size);
1841 void sk_send_sigurg(struct sock *sk);
1843 static inline void sock_replace_proto(struct sock *sk, struct proto *proto)
1846 clear_bit(SOCK_SUPPORT_ZC, &sk->sk_socket->flags);
1847 WRITE_ONCE(sk->sk_prot, proto);
1850 struct sockcm_cookie {
1856 static inline void sockcm_init(struct sockcm_cookie *sockc,
1857 const struct sock *sk)
1859 *sockc = (struct sockcm_cookie) {
1860 .tsflags = READ_ONCE(sk->sk_tsflags)
1864 int __sock_cmsg_send(struct sock *sk, struct cmsghdr *cmsg,
1865 struct sockcm_cookie *sockc);
1866 int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1867 struct sockcm_cookie *sockc);
1870 * Functions to fill in entries in struct proto_ops when a protocol
1871 * does not implement a particular function.
1873 int sock_no_bind(struct socket *, struct sockaddr *, int);
1874 int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1875 int sock_no_socketpair(struct socket *, struct socket *);
1876 int sock_no_accept(struct socket *, struct socket *, int, bool);
1877 int sock_no_getname(struct socket *, struct sockaddr *, int);
1878 int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1879 int sock_no_listen(struct socket *, int);
1880 int sock_no_shutdown(struct socket *, int);
1881 int sock_no_sendmsg(struct socket *, struct msghdr *, size_t);
1882 int sock_no_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t len);
1883 int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int);
1884 int sock_no_mmap(struct file *file, struct socket *sock,
1885 struct vm_area_struct *vma);
1888 * Functions to fill in entries in struct proto_ops when a protocol
1889 * uses the inet style.
1891 int sock_common_getsockopt(struct socket *sock, int level, int optname,
1892 char __user *optval, int __user *optlen);
1893 int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1895 int sock_common_setsockopt(struct socket *sock, int level, int optname,
1896 sockptr_t optval, unsigned int optlen);
1898 void sk_common_release(struct sock *sk);
1901 * Default socket callbacks and setup code
1904 /* Initialise core socket variables using an explicit uid. */
1905 void sock_init_data_uid(struct socket *sock, struct sock *sk, kuid_t uid);
1907 /* Initialise core socket variables.
1908 * Assumes struct socket *sock is embedded in a struct socket_alloc.
1910 void sock_init_data(struct socket *sock, struct sock *sk);
1913 * Socket reference counting postulates.
1915 * * Each user of socket SHOULD hold a reference count.
1916 * * Each access point to socket (an hash table bucket, reference from a list,
1917 * running timer, skb in flight MUST hold a reference count.
1918 * * When reference count hits 0, it means it will never increase back.
1919 * * When reference count hits 0, it means that no references from
1920 * outside exist to this socket and current process on current CPU
1921 * is last user and may/should destroy this socket.
1922 * * sk_free is called from any context: process, BH, IRQ. When
1923 * it is called, socket has no references from outside -> sk_free
1924 * may release descendant resources allocated by the socket, but
1925 * to the time when it is called, socket is NOT referenced by any
1926 * hash tables, lists etc.
1927 * * Packets, delivered from outside (from network or from another process)
1928 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1929 * when they sit in queue. Otherwise, packets will leak to hole, when
1930 * socket is looked up by one cpu and unhasing is made by another CPU.
1931 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1932 * (leak to backlog). Packet socket does all the processing inside
1933 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1934 * use separate SMP lock, so that they are prone too.
1937 /* Ungrab socket and destroy it, if it was the last reference. */
1938 static inline void sock_put(struct sock *sk)
1940 if (refcount_dec_and_test(&sk->sk_refcnt))
1943 /* Generic version of sock_put(), dealing with all sockets
1944 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
1946 void sock_gen_put(struct sock *sk);
1948 int __sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested,
1949 unsigned int trim_cap, bool refcounted);
1950 static inline int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1953 return __sk_receive_skb(sk, skb, nested, 1, true);
1956 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1958 /* sk_tx_queue_mapping accept only upto a 16-bit value */
1959 if (WARN_ON_ONCE((unsigned short)tx_queue >= USHRT_MAX))
1961 /* Paired with READ_ONCE() in sk_tx_queue_get() and
1962 * other WRITE_ONCE() because socket lock might be not held.
1964 WRITE_ONCE(sk->sk_tx_queue_mapping, tx_queue);
1967 #define NO_QUEUE_MAPPING USHRT_MAX
1969 static inline void sk_tx_queue_clear(struct sock *sk)
1971 /* Paired with READ_ONCE() in sk_tx_queue_get() and
1972 * other WRITE_ONCE() because socket lock might be not held.
1974 WRITE_ONCE(sk->sk_tx_queue_mapping, NO_QUEUE_MAPPING);
1977 static inline int sk_tx_queue_get(const struct sock *sk)
1980 /* Paired with WRITE_ONCE() in sk_tx_queue_clear()
1981 * and sk_tx_queue_set().
1983 int val = READ_ONCE(sk->sk_tx_queue_mapping);
1985 if (val != NO_QUEUE_MAPPING)
1991 static inline void __sk_rx_queue_set(struct sock *sk,
1992 const struct sk_buff *skb,
1995 #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
1996 if (skb_rx_queue_recorded(skb)) {
1997 u16 rx_queue = skb_get_rx_queue(skb);
2000 unlikely(READ_ONCE(sk->sk_rx_queue_mapping) != rx_queue))
2001 WRITE_ONCE(sk->sk_rx_queue_mapping, rx_queue);
2006 static inline void sk_rx_queue_set(struct sock *sk, const struct sk_buff *skb)
2008 __sk_rx_queue_set(sk, skb, true);
2011 static inline void sk_rx_queue_update(struct sock *sk, const struct sk_buff *skb)
2013 __sk_rx_queue_set(sk, skb, false);
2016 static inline void sk_rx_queue_clear(struct sock *sk)
2018 #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
2019 WRITE_ONCE(sk->sk_rx_queue_mapping, NO_QUEUE_MAPPING);
2023 static inline int sk_rx_queue_get(const struct sock *sk)
2025 #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
2027 int res = READ_ONCE(sk->sk_rx_queue_mapping);
2029 if (res != NO_QUEUE_MAPPING)
2037 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
2039 sk->sk_socket = sock;
2042 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
2044 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
2045 return &rcu_dereference_raw(sk->sk_wq)->wait;
2047 /* Detach socket from process context.
2048 * Announce socket dead, detach it from wait queue and inode.
2049 * Note that parent inode held reference count on this struct sock,
2050 * we do not release it in this function, because protocol
2051 * probably wants some additional cleanups or even continuing
2052 * to work with this socket (TCP).
2054 static inline void sock_orphan(struct sock *sk)
2056 write_lock_bh(&sk->sk_callback_lock);
2057 sock_set_flag(sk, SOCK_DEAD);
2058 sk_set_socket(sk, NULL);
2060 write_unlock_bh(&sk->sk_callback_lock);
2063 static inline void sock_graft(struct sock *sk, struct socket *parent)
2065 WARN_ON(parent->sk);
2066 write_lock_bh(&sk->sk_callback_lock);
2067 rcu_assign_pointer(sk->sk_wq, &parent->wq);
2069 sk_set_socket(sk, parent);
2070 sk->sk_uid = SOCK_INODE(parent)->i_uid;
2071 security_sock_graft(sk, parent);
2072 write_unlock_bh(&sk->sk_callback_lock);
2075 kuid_t sock_i_uid(struct sock *sk);
2076 unsigned long __sock_i_ino(struct sock *sk);
2077 unsigned long sock_i_ino(struct sock *sk);
2079 static inline kuid_t sock_net_uid(const struct net *net, const struct sock *sk)
2081 return sk ? sk->sk_uid : make_kuid(net->user_ns, 0);
2084 static inline u32 net_tx_rndhash(void)
2086 u32 v = get_random_u32();
2091 static inline void sk_set_txhash(struct sock *sk)
2093 /* This pairs with READ_ONCE() in skb_set_hash_from_sk() */
2094 WRITE_ONCE(sk->sk_txhash, net_tx_rndhash());
2097 static inline bool sk_rethink_txhash(struct sock *sk)
2099 if (sk->sk_txhash && sk->sk_txrehash == SOCK_TXREHASH_ENABLED) {
2106 static inline struct dst_entry *
2107 __sk_dst_get(const struct sock *sk)
2109 return rcu_dereference_check(sk->sk_dst_cache,
2110 lockdep_sock_is_held(sk));
2113 static inline struct dst_entry *
2114 sk_dst_get(const struct sock *sk)
2116 struct dst_entry *dst;
2119 dst = rcu_dereference(sk->sk_dst_cache);
2120 if (dst && !rcuref_get(&dst->__rcuref))
2126 static inline void __dst_negative_advice(struct sock *sk)
2128 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
2130 if (dst && dst->ops->negative_advice) {
2131 ndst = dst->ops->negative_advice(dst);
2134 rcu_assign_pointer(sk->sk_dst_cache, ndst);
2135 sk_tx_queue_clear(sk);
2136 WRITE_ONCE(sk->sk_dst_pending_confirm, 0);
2141 static inline void dst_negative_advice(struct sock *sk)
2143 sk_rethink_txhash(sk);
2144 __dst_negative_advice(sk);
2148 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
2150 struct dst_entry *old_dst;
2152 sk_tx_queue_clear(sk);
2153 WRITE_ONCE(sk->sk_dst_pending_confirm, 0);
2154 old_dst = rcu_dereference_protected(sk->sk_dst_cache,
2155 lockdep_sock_is_held(sk));
2156 rcu_assign_pointer(sk->sk_dst_cache, dst);
2157 dst_release(old_dst);
2161 sk_dst_set(struct sock *sk, struct dst_entry *dst)
2163 struct dst_entry *old_dst;
2165 sk_tx_queue_clear(sk);
2166 WRITE_ONCE(sk->sk_dst_pending_confirm, 0);
2167 old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
2168 dst_release(old_dst);
2172 __sk_dst_reset(struct sock *sk)
2174 __sk_dst_set(sk, NULL);
2178 sk_dst_reset(struct sock *sk)
2180 sk_dst_set(sk, NULL);
2183 struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
2185 struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
2187 static inline void sk_dst_confirm(struct sock *sk)
2189 if (!READ_ONCE(sk->sk_dst_pending_confirm))
2190 WRITE_ONCE(sk->sk_dst_pending_confirm, 1);
2193 static inline void sock_confirm_neigh(struct sk_buff *skb, struct neighbour *n)
2195 if (skb_get_dst_pending_confirm(skb)) {
2196 struct sock *sk = skb->sk;
2198 if (sk && READ_ONCE(sk->sk_dst_pending_confirm))
2199 WRITE_ONCE(sk->sk_dst_pending_confirm, 0);
2204 bool sk_mc_loop(const struct sock *sk);
2206 static inline bool sk_can_gso(const struct sock *sk)
2208 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
2211 void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
2213 static inline void sk_gso_disable(struct sock *sk)
2215 sk->sk_gso_disabled = 1;
2216 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
2219 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
2220 struct iov_iter *from, char *to,
2221 int copy, int offset)
2223 if (skb->ip_summed == CHECKSUM_NONE) {
2225 if (!csum_and_copy_from_iter_full(to, copy, &csum, from))
2227 skb->csum = csum_block_add(skb->csum, csum, offset);
2228 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
2229 if (!copy_from_iter_full_nocache(to, copy, from))
2231 } else if (!copy_from_iter_full(to, copy, from))
2237 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
2238 struct iov_iter *from, int copy)
2240 int err, offset = skb->len;
2242 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
2245 __skb_trim(skb, offset);
2250 static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from,
2251 struct sk_buff *skb,
2257 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
2262 skb_len_add(skb, copy);
2263 sk_wmem_queued_add(sk, copy);
2264 sk_mem_charge(sk, copy);
2269 * sk_wmem_alloc_get - returns write allocations
2272 * Return: sk_wmem_alloc minus initial offset of one
2274 static inline int sk_wmem_alloc_get(const struct sock *sk)
2276 return refcount_read(&sk->sk_wmem_alloc) - 1;
2280 * sk_rmem_alloc_get - returns read allocations
2283 * Return: sk_rmem_alloc
2285 static inline int sk_rmem_alloc_get(const struct sock *sk)
2287 return atomic_read(&sk->sk_rmem_alloc);
2291 * sk_has_allocations - check if allocations are outstanding
2294 * Return: true if socket has write or read allocations
2296 static inline bool sk_has_allocations(const struct sock *sk)
2298 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
2302 * skwq_has_sleeper - check if there are any waiting processes
2303 * @wq: struct socket_wq
2305 * Return: true if socket_wq has waiting processes
2307 * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
2308 * barrier call. They were added due to the race found within the tcp code.
2310 * Consider following tcp code paths::
2313 * sys_select receive packet
2315 * __add_wait_queue update tp->rcv_nxt
2317 * tp->rcv_nxt check sock_def_readable
2319 * schedule rcu_read_lock();
2320 * wq = rcu_dereference(sk->sk_wq);
2321 * if (wq && waitqueue_active(&wq->wait))
2322 * wake_up_interruptible(&wq->wait)
2326 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
2327 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
2328 * could then endup calling schedule and sleep forever if there are no more
2329 * data on the socket.
2332 static inline bool skwq_has_sleeper(struct socket_wq *wq)
2334 return wq && wq_has_sleeper(&wq->wait);
2338 * sock_poll_wait - place memory barrier behind the poll_wait call.
2340 * @sock: socket to wait on
2343 * See the comments in the wq_has_sleeper function.
2345 static inline void sock_poll_wait(struct file *filp, struct socket *sock,
2348 if (!poll_does_not_wait(p)) {
2349 poll_wait(filp, &sock->wq.wait, p);
2350 /* We need to be sure we are in sync with the
2351 * socket flags modification.
2353 * This memory barrier is paired in the wq_has_sleeper.
2359 static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
2361 /* This pairs with WRITE_ONCE() in sk_set_txhash() */
2362 u32 txhash = READ_ONCE(sk->sk_txhash);
2370 void skb_set_owner_w(struct sk_buff *skb, struct sock *sk);
2373 * Queue a received datagram if it will fit. Stream and sequenced
2374 * protocols can't normally use this as they need to fit buffers in
2375 * and play with them.
2377 * Inlined as it's very short and called for pretty much every
2378 * packet ever received.
2380 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2384 skb->destructor = sock_rfree;
2385 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
2386 sk_mem_charge(sk, skb->truesize);
2389 static inline __must_check bool skb_set_owner_sk_safe(struct sk_buff *skb, struct sock *sk)
2391 if (sk && refcount_inc_not_zero(&sk->sk_refcnt)) {
2393 skb->destructor = sock_efree;
2400 static inline struct sk_buff *skb_clone_and_charge_r(struct sk_buff *skb, struct sock *sk)
2402 skb = skb_clone(skb, sk_gfp_mask(sk, GFP_ATOMIC));
2404 if (sk_rmem_schedule(sk, skb, skb->truesize)) {
2405 skb_set_owner_r(skb, sk);
2413 static inline void skb_prepare_for_gro(struct sk_buff *skb)
2415 if (skb->destructor != sock_wfree) {
2422 void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2423 unsigned long expires);
2425 void sk_stop_timer(struct sock *sk, struct timer_list *timer);
2427 void sk_stop_timer_sync(struct sock *sk, struct timer_list *timer);
2429 int __sk_queue_drop_skb(struct sock *sk, struct sk_buff_head *sk_queue,
2430 struct sk_buff *skb, unsigned int flags,
2431 void (*destructor)(struct sock *sk,
2432 struct sk_buff *skb));
2433 int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
2435 int sock_queue_rcv_skb_reason(struct sock *sk, struct sk_buff *skb,
2436 enum skb_drop_reason *reason);
2438 static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
2440 return sock_queue_rcv_skb_reason(sk, skb, NULL);
2443 int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
2444 struct sk_buff *sock_dequeue_err_skb(struct sock *sk);
2447 * Recover an error report and clear atomically
2450 static inline int sock_error(struct sock *sk)
2454 /* Avoid an atomic operation for the common case.
2455 * This is racy since another cpu/thread can change sk_err under us.
2457 if (likely(data_race(!sk->sk_err)))
2460 err = xchg(&sk->sk_err, 0);
2464 void sk_error_report(struct sock *sk);
2466 static inline unsigned long sock_wspace(struct sock *sk)
2470 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2471 amt = sk->sk_sndbuf - refcount_read(&sk->sk_wmem_alloc);
2479 * We use sk->sk_wq_raw, from contexts knowing this
2480 * pointer is not NULL and cannot disappear/change.
2482 static inline void sk_set_bit(int nr, struct sock *sk)
2484 if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
2485 !sock_flag(sk, SOCK_FASYNC))
2488 set_bit(nr, &sk->sk_wq_raw->flags);
2491 static inline void sk_clear_bit(int nr, struct sock *sk)
2493 if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
2494 !sock_flag(sk, SOCK_FASYNC))
2497 clear_bit(nr, &sk->sk_wq_raw->flags);
2500 static inline void sk_wake_async(const struct sock *sk, int how, int band)
2502 if (sock_flag(sk, SOCK_FASYNC)) {
2504 sock_wake_async(rcu_dereference(sk->sk_wq), how, band);
2509 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2510 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2511 * Note: for send buffers, TCP works better if we can build two skbs at
2514 #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2516 #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2517 #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
2519 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2523 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
2526 val = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
2527 val = max_t(u32, val, sk_unused_reserved_mem(sk));
2529 WRITE_ONCE(sk->sk_sndbuf, max_t(u32, val, SOCK_MIN_SNDBUF));
2533 * sk_page_frag - return an appropriate page_frag
2536 * Use the per task page_frag instead of the per socket one for
2537 * optimization when we know that we're in process context and own
2538 * everything that's associated with %current.
2540 * Both direct reclaim and page faults can nest inside other
2541 * socket operations and end up recursing into sk_page_frag()
2542 * while it's already in use: explicitly avoid task page_frag
2543 * when users disable sk_use_task_frag.
2545 * Return: a per task page_frag if context allows that,
2546 * otherwise a per socket one.
2548 static inline struct page_frag *sk_page_frag(struct sock *sk)
2550 if (sk->sk_use_task_frag)
2551 return ¤t->task_frag;
2553 return &sk->sk_frag;
2556 bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
2559 * Default write policy as shown to user space via poll/select/SIGIO
2561 static inline bool sock_writeable(const struct sock *sk)
2563 return refcount_read(&sk->sk_wmem_alloc) < (READ_ONCE(sk->sk_sndbuf) >> 1);
2566 static inline gfp_t gfp_any(void)
2568 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
2571 static inline gfp_t gfp_memcg_charge(void)
2573 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
2576 static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
2578 return noblock ? 0 : sk->sk_rcvtimeo;
2581 static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
2583 return noblock ? 0 : sk->sk_sndtimeo;
2586 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2588 int v = waitall ? len : min_t(int, READ_ONCE(sk->sk_rcvlowat), len);
2593 /* Alas, with timeout socket operations are not restartable.
2594 * Compare this to poll().
2596 static inline int sock_intr_errno(long timeo)
2598 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2601 struct sock_skb_cb {
2605 /* Store sock_skb_cb at the end of skb->cb[] so protocol families
2606 * using skb->cb[] would keep using it directly and utilize its
2607 * alignement guarantee.
2609 #define SOCK_SKB_CB_OFFSET ((sizeof_field(struct sk_buff, cb) - \
2610 sizeof(struct sock_skb_cb)))
2612 #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2613 SOCK_SKB_CB_OFFSET))
2615 #define sock_skb_cb_check_size(size) \
2616 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
2619 sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb)
2621 SOCK_SKB_CB(skb)->dropcount = sock_flag(sk, SOCK_RXQ_OVFL) ?
2622 atomic_read(&sk->sk_drops) : 0;
2625 static inline void sk_drops_add(struct sock *sk, const struct sk_buff *skb)
2627 int segs = max_t(u16, 1, skb_shinfo(skb)->gso_segs);
2629 atomic_add(segs, &sk->sk_drops);
2632 static inline ktime_t sock_read_timestamp(struct sock *sk)
2634 #if BITS_PER_LONG==32
2639 seq = read_seqbegin(&sk->sk_stamp_seq);
2641 } while (read_seqretry(&sk->sk_stamp_seq, seq));
2645 return READ_ONCE(sk->sk_stamp);
2649 static inline void sock_write_timestamp(struct sock *sk, ktime_t kt)
2651 #if BITS_PER_LONG==32
2652 write_seqlock(&sk->sk_stamp_seq);
2654 write_sequnlock(&sk->sk_stamp_seq);
2656 WRITE_ONCE(sk->sk_stamp, kt);
2660 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2661 struct sk_buff *skb);
2662 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2663 struct sk_buff *skb);
2666 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2668 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
2669 u32 tsflags = READ_ONCE(sk->sk_tsflags);
2670 ktime_t kt = skb->tstamp;
2672 * generate control messages if
2673 * - receive time stamping in software requested
2674 * - software time stamp available and wanted
2675 * - hardware time stamps available and wanted
2677 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2678 (tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
2679 (kt && tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
2680 (hwtstamps->hwtstamp &&
2681 (tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
2682 __sock_recv_timestamp(msg, sk, skb);
2684 sock_write_timestamp(sk, kt);
2686 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb_wifi_acked_valid(skb))
2687 __sock_recv_wifi_status(msg, sk, skb);
2690 void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
2691 struct sk_buff *skb);
2693 #define SK_DEFAULT_STAMP (-1L * NSEC_PER_SEC)
2694 static inline void sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
2695 struct sk_buff *skb)
2697 #define FLAGS_RECV_CMSGS ((1UL << SOCK_RXQ_OVFL) | \
2698 (1UL << SOCK_RCVTSTAMP) | \
2699 (1UL << SOCK_RCVMARK))
2700 #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2701 SOF_TIMESTAMPING_RAW_HARDWARE)
2703 if (sk->sk_flags & FLAGS_RECV_CMSGS ||
2704 READ_ONCE(sk->sk_tsflags) & TSFLAGS_ANY)
2705 __sock_recv_cmsgs(msg, sk, skb);
2706 else if (unlikely(sock_flag(sk, SOCK_TIMESTAMP)))
2707 sock_write_timestamp(sk, skb->tstamp);
2708 else if (unlikely(sock_read_timestamp(sk) == SK_DEFAULT_STAMP))
2709 sock_write_timestamp(sk, 0);
2712 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags);
2715 * _sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2716 * @sk: socket sending this packet
2717 * @tsflags: timestamping flags to use
2718 * @tx_flags: completed with instructions for time stamping
2719 * @tskey: filled in with next sk_tskey (not for TCP, which uses seqno)
2721 * Note: callers should take care of initial ``*tx_flags`` value (usually 0)
2723 static inline void _sock_tx_timestamp(struct sock *sk, __u16 tsflags,
2724 __u8 *tx_flags, __u32 *tskey)
2726 if (unlikely(tsflags)) {
2727 __sock_tx_timestamp(tsflags, tx_flags);
2728 if (tsflags & SOF_TIMESTAMPING_OPT_ID && tskey &&
2729 tsflags & SOF_TIMESTAMPING_TX_RECORD_MASK)
2730 *tskey = atomic_inc_return(&sk->sk_tskey) - 1;
2732 if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS)))
2733 *tx_flags |= SKBTX_WIFI_STATUS;
2736 static inline void sock_tx_timestamp(struct sock *sk, __u16 tsflags,
2739 _sock_tx_timestamp(sk, tsflags, tx_flags, NULL);
2742 static inline void skb_setup_tx_timestamp(struct sk_buff *skb, __u16 tsflags)
2744 _sock_tx_timestamp(skb->sk, tsflags, &skb_shinfo(skb)->tx_flags,
2745 &skb_shinfo(skb)->tskey);
2748 static inline bool sk_is_inet(const struct sock *sk)
2750 int family = READ_ONCE(sk->sk_family);
2752 return family == AF_INET || family == AF_INET6;
2755 static inline bool sk_is_tcp(const struct sock *sk)
2757 return sk_is_inet(sk) &&
2758 sk->sk_type == SOCK_STREAM &&
2759 sk->sk_protocol == IPPROTO_TCP;
2762 static inline bool sk_is_udp(const struct sock *sk)
2764 return sk_is_inet(sk) &&
2765 sk->sk_type == SOCK_DGRAM &&
2766 sk->sk_protocol == IPPROTO_UDP;
2769 static inline bool sk_is_stream_unix(const struct sock *sk)
2771 return sk->sk_family == AF_UNIX && sk->sk_type == SOCK_STREAM;
2775 * sk_eat_skb - Release a skb if it is no longer needed
2776 * @sk: socket to eat this skb from
2777 * @skb: socket buffer to eat
2779 * This routine must be called with interrupts disabled or with the socket
2780 * locked so that the sk_buff queue operation is ok.
2782 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
2784 __skb_unlink(skb, &sk->sk_receive_queue);
2789 skb_sk_is_prefetched(struct sk_buff *skb)
2792 return skb->destructor == sock_pfree;
2795 #endif /* CONFIG_INET */
2798 /* This helper checks if a socket is a full socket,
2799 * ie _not_ a timewait or request socket.
2801 static inline bool sk_fullsock(const struct sock *sk)
2803 return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV);
2807 sk_is_refcounted(struct sock *sk)
2809 /* Only full sockets have sk->sk_flags. */
2810 return !sk_fullsock(sk) || !sock_flag(sk, SOCK_RCU_FREE);
2813 /* Checks if this SKB belongs to an HW offloaded socket
2814 * and whether any SW fallbacks are required based on dev.
2815 * Check decrypted mark in case skb_orphan() cleared socket.
2817 static inline struct sk_buff *sk_validate_xmit_skb(struct sk_buff *skb,
2818 struct net_device *dev)
2820 #ifdef CONFIG_SOCK_VALIDATE_XMIT
2821 struct sock *sk = skb->sk;
2823 if (sk && sk_fullsock(sk) && sk->sk_validate_xmit_skb) {
2824 skb = sk->sk_validate_xmit_skb(sk, dev, skb);
2825 #ifdef CONFIG_TLS_DEVICE
2826 } else if (unlikely(skb->decrypted)) {
2827 pr_warn_ratelimited("unencrypted skb with no associated socket - dropping\n");
2837 /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2838 * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2840 static inline bool sk_listener(const struct sock *sk)
2842 return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV);
2845 void sock_enable_timestamp(struct sock *sk, enum sock_flags flag);
2846 int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2849 bool sk_ns_capable(const struct sock *sk,
2850 struct user_namespace *user_ns, int cap);
2851 bool sk_capable(const struct sock *sk, int cap);
2852 bool sk_net_capable(const struct sock *sk, int cap);
2854 void sk_get_meminfo(const struct sock *sk, u32 *meminfo);
2856 /* Take into consideration the size of the struct sk_buff overhead in the
2857 * determination of these values, since that is non-constant across
2858 * platforms. This makes socket queueing behavior and performance
2859 * not depend upon such differences.
2861 #define _SK_MEM_PACKETS 256
2862 #define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
2863 #define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2864 #define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2866 extern __u32 sysctl_wmem_max;
2867 extern __u32 sysctl_rmem_max;
2869 extern int sysctl_tstamp_allow_data;
2871 extern __u32 sysctl_wmem_default;
2872 extern __u32 sysctl_rmem_default;
2874 #define SKB_FRAG_PAGE_ORDER get_order(32768)
2875 DECLARE_STATIC_KEY_FALSE(net_high_order_alloc_disable_key);
2877 static inline int sk_get_wmem0(const struct sock *sk, const struct proto *proto)
2879 /* Does this proto have per netns sysctl_wmem ? */
2880 if (proto->sysctl_wmem_offset)
2881 return READ_ONCE(*(int *)((void *)sock_net(sk) + proto->sysctl_wmem_offset));
2883 return READ_ONCE(*proto->sysctl_wmem);
2886 static inline int sk_get_rmem0(const struct sock *sk, const struct proto *proto)
2888 /* Does this proto have per netns sysctl_rmem ? */
2889 if (proto->sysctl_rmem_offset)
2890 return READ_ONCE(*(int *)((void *)sock_net(sk) + proto->sysctl_rmem_offset));
2892 return READ_ONCE(*proto->sysctl_rmem);
2895 /* Default TCP Small queue budget is ~1 ms of data (1sec >> 10)
2896 * Some wifi drivers need to tweak it to get more chunks.
2897 * They can use this helper from their ndo_start_xmit()
2899 static inline void sk_pacing_shift_update(struct sock *sk, int val)
2901 if (!sk || !sk_fullsock(sk) || READ_ONCE(sk->sk_pacing_shift) == val)
2903 WRITE_ONCE(sk->sk_pacing_shift, val);
2906 /* if a socket is bound to a device, check that the given device
2907 * index is either the same or that the socket is bound to an L3
2908 * master device and the given device index is also enslaved to
2911 static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
2913 int bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
2916 if (!bound_dev_if || bound_dev_if == dif)
2919 mdif = l3mdev_master_ifindex_by_index(sock_net(sk), dif);
2920 if (mdif && mdif == bound_dev_if)
2926 void sock_def_readable(struct sock *sk);
2928 int sock_bindtoindex(struct sock *sk, int ifindex, bool lock_sk);
2929 void sock_set_timestamp(struct sock *sk, int optname, bool valbool);
2930 int sock_set_timestamping(struct sock *sk, int optname,
2931 struct so_timestamping timestamping);
2933 void sock_enable_timestamps(struct sock *sk);
2934 void sock_no_linger(struct sock *sk);
2935 void sock_set_keepalive(struct sock *sk);
2936 void sock_set_priority(struct sock *sk, u32 priority);
2937 void sock_set_rcvbuf(struct sock *sk, int val);
2938 void sock_set_mark(struct sock *sk, u32 val);
2939 void sock_set_reuseaddr(struct sock *sk);
2940 void sock_set_reuseport(struct sock *sk);
2941 void sock_set_sndtimeo(struct sock *sk, s64 secs);
2943 int sock_bind_add(struct sock *sk, struct sockaddr *addr, int addr_len);
2945 int sock_get_timeout(long timeo, void *optval, bool old_timeval);
2946 int sock_copy_user_timeval(struct __kernel_sock_timeval *tv,
2947 sockptr_t optval, int optlen, bool old_timeval);
2949 int sock_ioctl_inout(struct sock *sk, unsigned int cmd,
2950 void __user *arg, void *karg, size_t size);
2951 int sk_ioctl(struct sock *sk, unsigned int cmd, void __user *arg);
2952 static inline bool sk_is_readable(struct sock *sk)
2954 if (sk->sk_prot->sock_is_readable)
2955 return sk->sk_prot->sock_is_readable(sk);
2958 #endif /* _SOCK_H */