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/filter.h>
60 #include <linux/rculist_nulls.h>
61 #include <linux/poll.h>
62 #include <linux/sockptr.h>
63 #include <linux/indirect_call_wrapper.h>
64 #include <linux/atomic.h>
65 #include <linux/refcount.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 /* Define this to get the SOCK_DBG debugging facility. */
80 #define SOCK_DEBUGGING
82 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
83 printk(KERN_DEBUG msg); } while (0)
85 /* Validate arguments and do nothing */
86 static inline __printf(2, 3)
87 void SOCK_DEBUG(const struct sock
*sk
, const char *msg
, ...)
92 /* This is the per-socket lock. The spinlock provides a synchronization
93 * between user contexts and software interrupt processing, whereas the
94 * mini-semaphore synchronizes multiple users amongst themselves.
101 * We express the mutex-alike socket_lock semantics
102 * to the lock validator by explicitly managing
103 * the slock as a lock variant (in addition to
106 #ifdef CONFIG_DEBUG_LOCK_ALLOC
107 struct lockdep_map dep_map
;
115 typedef __u32 __bitwise __portpair
;
116 typedef __u64 __bitwise __addrpair
;
119 * struct sock_common - minimal network layer representation of sockets
120 * @skc_daddr: Foreign IPv4 addr
121 * @skc_rcv_saddr: Bound local IPv4 addr
122 * @skc_addrpair: 8-byte-aligned __u64 union of @skc_daddr & @skc_rcv_saddr
123 * @skc_hash: hash value used with various protocol lookup tables
124 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
125 * @skc_dport: placeholder for inet_dport/tw_dport
126 * @skc_num: placeholder for inet_num/tw_num
127 * @skc_portpair: __u32 union of @skc_dport & @skc_num
128 * @skc_family: network address family
129 * @skc_state: Connection state
130 * @skc_reuse: %SO_REUSEADDR setting
131 * @skc_reuseport: %SO_REUSEPORT setting
132 * @skc_ipv6only: socket is IPV6 only
133 * @skc_net_refcnt: socket is using net ref counting
134 * @skc_bound_dev_if: bound device index if != 0
135 * @skc_bind_node: bind hash linkage for various protocol lookup tables
136 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
137 * @skc_prot: protocol handlers inside a network family
138 * @skc_net: reference to the network namespace of this socket
139 * @skc_v6_daddr: IPV6 destination address
140 * @skc_v6_rcv_saddr: IPV6 source address
141 * @skc_cookie: socket's cookie value
142 * @skc_node: main hash linkage for various protocol lookup tables
143 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
144 * @skc_tx_queue_mapping: tx queue number for this connection
145 * @skc_rx_queue_mapping: rx queue number for this connection
146 * @skc_flags: place holder for sk_flags
147 * %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
148 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
149 * @skc_listener: connection request listener socket (aka rsk_listener)
150 * [union with @skc_flags]
151 * @skc_tw_dr: (aka tw_dr) ptr to &struct inet_timewait_death_row
152 * [union with @skc_flags]
153 * @skc_incoming_cpu: record/match cpu processing incoming packets
154 * @skc_rcv_wnd: (aka rsk_rcv_wnd) TCP receive window size (possibly scaled)
155 * [union with @skc_incoming_cpu]
156 * @skc_tw_rcv_nxt: (aka tw_rcv_nxt) TCP window next expected seq number
157 * [union with @skc_incoming_cpu]
158 * @skc_refcnt: reference count
160 * This is the minimal network layer representation of sockets, the header
161 * for struct sock and struct inet_timewait_sock.
164 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
165 * address on 64bit arches : cf INET_MATCH()
168 __addrpair skc_addrpair
;
171 __be32 skc_rcv_saddr
;
175 unsigned int skc_hash
;
176 __u16 skc_u16hashes
[2];
178 /* skc_dport && skc_num must be grouped as well */
180 __portpair skc_portpair
;
187 unsigned short skc_family
;
188 volatile unsigned char skc_state
;
189 unsigned char skc_reuse
:4;
190 unsigned char skc_reuseport
:1;
191 unsigned char skc_ipv6only
:1;
192 unsigned char skc_net_refcnt
:1;
193 int skc_bound_dev_if
;
195 struct hlist_node skc_bind_node
;
196 struct hlist_node skc_portaddr_node
;
198 struct proto
*skc_prot
;
199 possible_net_t skc_net
;
201 #if IS_ENABLED(CONFIG_IPV6)
202 struct in6_addr skc_v6_daddr
;
203 struct in6_addr skc_v6_rcv_saddr
;
206 atomic64_t skc_cookie
;
208 /* following fields are padding to force
209 * offset(struct sock, sk_refcnt) == 128 on 64bit arches
210 * assuming IPV6 is enabled. We use this padding differently
211 * for different kind of 'sockets'
214 unsigned long skc_flags
;
215 struct sock
*skc_listener
; /* request_sock */
216 struct inet_timewait_death_row
*skc_tw_dr
; /* inet_timewait_sock */
219 * fields between dontcopy_begin/dontcopy_end
220 * are not copied in sock_copy()
223 int skc_dontcopy_begin
[0];
226 struct hlist_node skc_node
;
227 struct hlist_nulls_node skc_nulls_node
;
229 unsigned short skc_tx_queue_mapping
;
230 #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
231 unsigned short skc_rx_queue_mapping
;
234 int skc_incoming_cpu
;
236 u32 skc_tw_rcv_nxt
; /* struct tcp_timewait_sock */
239 refcount_t skc_refcnt
;
241 int skc_dontcopy_end
[0];
244 u32 skc_window_clamp
;
245 u32 skc_tw_snd_nxt
; /* struct tcp_timewait_sock */
250 struct bpf_local_storage
;
253 * struct sock - network layer representation of sockets
254 * @__sk_common: shared layout with inet_timewait_sock
255 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
256 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
257 * @sk_lock: synchronizer
258 * @sk_kern_sock: True if sock is using kernel lock classes
259 * @sk_rcvbuf: size of receive buffer in bytes
260 * @sk_wq: sock wait queue and async head
261 * @sk_rx_dst: receive input route used by early demux
262 * @sk_rx_dst_ifindex: ifindex for @sk_rx_dst
263 * @sk_rx_dst_cookie: cookie for @sk_rx_dst
264 * @sk_dst_cache: destination cache
265 * @sk_dst_pending_confirm: need to confirm neighbour
266 * @sk_policy: flow policy
267 * @sk_rx_skb_cache: cache copy of recently accessed RX skb
268 * @sk_receive_queue: incoming packets
269 * @sk_wmem_alloc: transmit queue bytes committed
270 * @sk_tsq_flags: TCP Small Queues flags
271 * @sk_write_queue: Packet sending queue
272 * @sk_omem_alloc: "o" is "option" or "other"
273 * @sk_wmem_queued: persistent queue size
274 * @sk_forward_alloc: space allocated forward
275 * @sk_napi_id: id of the last napi context to receive data for sk
276 * @sk_ll_usec: usecs to busypoll when there is no data
277 * @sk_allocation: allocation mode
278 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
279 * @sk_pacing_status: Pacing status (requested, handled by sch_fq)
280 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
281 * @sk_sndbuf: size of send buffer in bytes
282 * @__sk_flags_offset: empty field used to determine location of bitfield
283 * @sk_padding: unused element for alignment
284 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
285 * @sk_no_check_rx: allow zero checksum in RX packets
286 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
287 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
288 * @sk_route_forced_caps: static, forced route capabilities
289 * (set in tcp_init_sock())
290 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
291 * @sk_gso_max_size: Maximum GSO segment size to build
292 * @sk_gso_max_segs: Maximum number of GSO segments
293 * @sk_pacing_shift: scaling factor for TCP Small Queues
294 * @sk_lingertime: %SO_LINGER l_linger setting
295 * @sk_backlog: always used with the per-socket spinlock held
296 * @sk_callback_lock: used with the callbacks in the end of this struct
297 * @sk_error_queue: rarely used
298 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
299 * IPV6_ADDRFORM for instance)
300 * @sk_err: last error
301 * @sk_err_soft: errors that don't cause failure but are the cause of a
302 * persistent failure not just 'timed out'
303 * @sk_drops: raw/udp drops counter
304 * @sk_ack_backlog: current listen backlog
305 * @sk_max_ack_backlog: listen backlog set in listen()
306 * @sk_uid: user id of owner
307 * @sk_prefer_busy_poll: prefer busypolling over softirq processing
308 * @sk_busy_poll_budget: napi processing budget when busypolling
309 * @sk_priority: %SO_PRIORITY setting
310 * @sk_type: socket type (%SOCK_STREAM, etc)
311 * @sk_protocol: which protocol this socket belongs in this network family
312 * @sk_peer_lock: lock protecting @sk_peer_pid and @sk_peer_cred
313 * @sk_peer_pid: &struct pid for this socket's peer
314 * @sk_peer_cred: %SO_PEERCRED setting
315 * @sk_rcvlowat: %SO_RCVLOWAT setting
316 * @sk_rcvtimeo: %SO_RCVTIMEO setting
317 * @sk_sndtimeo: %SO_SNDTIMEO setting
318 * @sk_txhash: computed flow hash for use on transmit
319 * @sk_filter: socket filtering instructions
320 * @sk_timer: sock cleanup timer
321 * @sk_stamp: time stamp of last packet received
322 * @sk_stamp_seq: lock for accessing sk_stamp on 32 bit architectures only
323 * @sk_tsflags: SO_TIMESTAMPING flags
324 * @sk_bind_phc: SO_TIMESTAMPING bind PHC index of PTP virtual clock
326 * @sk_tskey: counter to disambiguate concurrent tstamp requests
327 * @sk_zckey: counter to order MSG_ZEROCOPY notifications
328 * @sk_socket: Identd and reporting IO signals
329 * @sk_user_data: RPC layer private data
330 * @sk_frag: cached page frag
331 * @sk_peek_off: current peek_offset value
332 * @sk_send_head: front of stuff to transmit
333 * @tcp_rtx_queue: TCP re-transmit queue [union with @sk_send_head]
334 * @sk_tx_skb_cache: cache copy of recently accessed TX skb
335 * @sk_security: used by security modules
336 * @sk_mark: generic packet mark
337 * @sk_cgrp_data: cgroup data for this cgroup
338 * @sk_memcg: this socket's memory cgroup association
339 * @sk_write_pending: a write to stream socket waits to start
340 * @sk_state_change: callback to indicate change in the state of the sock
341 * @sk_data_ready: callback to indicate there is data to be processed
342 * @sk_write_space: callback to indicate there is bf sending space available
343 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
344 * @sk_backlog_rcv: callback to process the backlog
345 * @sk_validate_xmit_skb: ptr to an optional validate function
346 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
347 * @sk_reuseport_cb: reuseport group container
348 * @sk_bpf_storage: ptr to cache and control for bpf_sk_storage
349 * @sk_rcu: used during RCU grace period
350 * @sk_clockid: clockid used by time-based scheduling (SO_TXTIME)
351 * @sk_txtime_deadline_mode: set deadline mode for SO_TXTIME
352 * @sk_txtime_report_errors: set report errors mode for SO_TXTIME
353 * @sk_txtime_unused: unused txtime flags
357 * Now struct inet_timewait_sock also uses sock_common, so please just
358 * don't add nothing before this first member (__sk_common) --acme
360 struct sock_common __sk_common
;
361 #define sk_node __sk_common.skc_node
362 #define sk_nulls_node __sk_common.skc_nulls_node
363 #define sk_refcnt __sk_common.skc_refcnt
364 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
365 #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
366 #define sk_rx_queue_mapping __sk_common.skc_rx_queue_mapping
369 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
370 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
371 #define sk_hash __sk_common.skc_hash
372 #define sk_portpair __sk_common.skc_portpair
373 #define sk_num __sk_common.skc_num
374 #define sk_dport __sk_common.skc_dport
375 #define sk_addrpair __sk_common.skc_addrpair
376 #define sk_daddr __sk_common.skc_daddr
377 #define sk_rcv_saddr __sk_common.skc_rcv_saddr
378 #define sk_family __sk_common.skc_family
379 #define sk_state __sk_common.skc_state
380 #define sk_reuse __sk_common.skc_reuse
381 #define sk_reuseport __sk_common.skc_reuseport
382 #define sk_ipv6only __sk_common.skc_ipv6only
383 #define sk_net_refcnt __sk_common.skc_net_refcnt
384 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
385 #define sk_bind_node __sk_common.skc_bind_node
386 #define sk_prot __sk_common.skc_prot
387 #define sk_net __sk_common.skc_net
388 #define sk_v6_daddr __sk_common.skc_v6_daddr
389 #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
390 #define sk_cookie __sk_common.skc_cookie
391 #define sk_incoming_cpu __sk_common.skc_incoming_cpu
392 #define sk_flags __sk_common.skc_flags
393 #define sk_rxhash __sk_common.skc_rxhash
395 socket_lock_t sk_lock
;
398 struct sk_buff_head sk_error_queue
;
399 struct sk_buff
*sk_rx_skb_cache
;
400 struct sk_buff_head sk_receive_queue
;
402 * The backlog queue is special, it is always used with
403 * the per-socket spinlock held and requires low latency
404 * access. Therefore we special case it's implementation.
405 * Note : rmem_alloc is in this structure to fill a hole
406 * on 64bit arches, not because its logically part of
412 struct sk_buff
*head
;
413 struct sk_buff
*tail
;
415 #define sk_rmem_alloc sk_backlog.rmem_alloc
417 int sk_forward_alloc
;
418 #ifdef CONFIG_NET_RX_BUSY_POLL
419 unsigned int sk_ll_usec
;
420 /* ===== mostly read cache line ===== */
421 unsigned int sk_napi_id
;
425 struct sk_filter __rcu
*sk_filter
;
427 struct socket_wq __rcu
*sk_wq
;
429 struct socket_wq
*sk_wq_raw
;
433 struct xfrm_policy __rcu
*sk_policy
[2];
435 struct dst_entry __rcu
*sk_rx_dst
;
436 int sk_rx_dst_ifindex
;
437 u32 sk_rx_dst_cookie
;
439 struct dst_entry __rcu
*sk_dst_cache
;
440 atomic_t sk_omem_alloc
;
443 /* ===== cache line for TX ===== */
445 refcount_t sk_wmem_alloc
;
446 unsigned long sk_tsq_flags
;
448 struct sk_buff
*sk_send_head
;
449 struct rb_root tcp_rtx_queue
;
451 struct sk_buff
*sk_tx_skb_cache
;
452 struct sk_buff_head sk_write_queue
;
454 int sk_write_pending
;
455 __u32 sk_dst_pending_confirm
;
456 u32 sk_pacing_status
; /* see enum sk_pacing */
458 struct timer_list sk_timer
;
461 unsigned long sk_pacing_rate
; /* bytes per second */
462 unsigned long sk_max_pacing_rate
;
463 struct page_frag sk_frag
;
464 netdev_features_t sk_route_caps
;
465 netdev_features_t sk_route_nocaps
;
466 netdev_features_t sk_route_forced_caps
;
468 unsigned int sk_gso_max_size
;
473 * Because of non atomicity rules, all
474 * changes are protected by socket lock.
485 unsigned long sk_lingertime
;
486 struct proto
*sk_prot_creator
;
487 rwlock_t sk_callback_lock
;
491 u32 sk_max_ack_backlog
;
493 #ifdef CONFIG_NET_RX_BUSY_POLL
494 u8 sk_prefer_busy_poll
;
495 u16 sk_busy_poll_budget
;
497 spinlock_t sk_peer_lock
;
498 struct pid
*sk_peer_pid
;
499 const struct cred
*sk_peer_cred
;
503 #if BITS_PER_LONG==32
504 seqlock_t sk_stamp_seq
;
513 u8 sk_txtime_deadline_mode
: 1,
514 sk_txtime_report_errors
: 1,
515 sk_txtime_unused
: 6;
517 struct socket
*sk_socket
;
519 #ifdef CONFIG_SECURITY
522 struct sock_cgroup_data sk_cgrp_data
;
523 struct mem_cgroup
*sk_memcg
;
524 void (*sk_state_change
)(struct sock
*sk
);
525 void (*sk_data_ready
)(struct sock
*sk
);
526 void (*sk_write_space
)(struct sock
*sk
);
527 void (*sk_error_report
)(struct sock
*sk
);
528 int (*sk_backlog_rcv
)(struct sock
*sk
,
529 struct sk_buff
*skb
);
530 #ifdef CONFIG_SOCK_VALIDATE_XMIT
531 struct sk_buff
* (*sk_validate_xmit_skb
)(struct sock
*sk
,
532 struct net_device
*dev
,
533 struct sk_buff
*skb
);
535 void (*sk_destruct
)(struct sock
*sk
);
536 struct sock_reuseport __rcu
*sk_reuseport_cb
;
537 #ifdef CONFIG_BPF_SYSCALL
538 struct bpf_local_storage __rcu
*sk_bpf_storage
;
540 struct rcu_head sk_rcu
;
545 SK_PACING_NEEDED
= 1,
549 /* Pointer stored in sk_user_data might not be suitable for copying
550 * when cloning the socket. For instance, it can point to a reference
551 * counted object. sk_user_data bottom bit is set if pointer must not
554 #define SK_USER_DATA_NOCOPY 1UL
555 #define SK_USER_DATA_BPF 2UL /* Managed by BPF */
556 #define SK_USER_DATA_PTRMASK ~(SK_USER_DATA_NOCOPY | SK_USER_DATA_BPF)
559 * sk_user_data_is_nocopy - Test if sk_user_data pointer must not be copied
562 static inline bool sk_user_data_is_nocopy(const struct sock
*sk
)
564 return ((uintptr_t)sk
->sk_user_data
& SK_USER_DATA_NOCOPY
);
567 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
569 #define rcu_dereference_sk_user_data(sk) \
571 void *__tmp = rcu_dereference(__sk_user_data((sk))); \
572 (void *)((uintptr_t)__tmp & SK_USER_DATA_PTRMASK); \
574 #define rcu_assign_sk_user_data(sk, ptr) \
576 uintptr_t __tmp = (uintptr_t)(ptr); \
577 WARN_ON_ONCE(__tmp & ~SK_USER_DATA_PTRMASK); \
578 rcu_assign_pointer(__sk_user_data((sk)), __tmp); \
580 #define rcu_assign_sk_user_data_nocopy(sk, ptr) \
582 uintptr_t __tmp = (uintptr_t)(ptr); \
583 WARN_ON_ONCE(__tmp & ~SK_USER_DATA_PTRMASK); \
584 rcu_assign_pointer(__sk_user_data((sk)), \
585 __tmp | SK_USER_DATA_NOCOPY); \
589 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
590 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
591 * on a socket means that the socket will reuse everybody else's port
592 * without looking at the other's sk_reuse value.
595 #define SK_NO_REUSE 0
596 #define SK_CAN_REUSE 1
597 #define SK_FORCE_REUSE 2
599 int sk_set_peek_off(struct sock
*sk
, int val
);
601 static inline int sk_peek_offset(struct sock
*sk
, int flags
)
603 if (unlikely(flags
& MSG_PEEK
)) {
604 return READ_ONCE(sk
->sk_peek_off
);
610 static inline void sk_peek_offset_bwd(struct sock
*sk
, int val
)
612 s32 off
= READ_ONCE(sk
->sk_peek_off
);
614 if (unlikely(off
>= 0)) {
615 off
= max_t(s32
, off
- val
, 0);
616 WRITE_ONCE(sk
->sk_peek_off
, off
);
620 static inline void sk_peek_offset_fwd(struct sock
*sk
, int val
)
622 sk_peek_offset_bwd(sk
, -val
);
626 * Hashed lists helper routines
628 static inline struct sock
*sk_entry(const struct hlist_node
*node
)
630 return hlist_entry(node
, struct sock
, sk_node
);
633 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
635 return hlist_entry(head
->first
, struct sock
, sk_node
);
638 static inline struct sock
*sk_head(const struct hlist_head
*head
)
640 return hlist_empty(head
) ? NULL
: __sk_head(head
);
643 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
645 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
648 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
650 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
653 static inline struct sock
*sk_next(const struct sock
*sk
)
655 return hlist_entry_safe(sk
->sk_node
.next
, struct sock
, sk_node
);
658 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
660 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
661 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
662 struct sock
, sk_nulls_node
) :
666 static inline bool sk_unhashed(const struct sock
*sk
)
668 return hlist_unhashed(&sk
->sk_node
);
671 static inline bool sk_hashed(const struct sock
*sk
)
673 return !sk_unhashed(sk
);
676 static inline void sk_node_init(struct hlist_node
*node
)
681 static inline void sk_nulls_node_init(struct hlist_nulls_node
*node
)
686 static inline void __sk_del_node(struct sock
*sk
)
688 __hlist_del(&sk
->sk_node
);
691 /* NB: equivalent to hlist_del_init_rcu */
692 static inline bool __sk_del_node_init(struct sock
*sk
)
696 sk_node_init(&sk
->sk_node
);
702 /* Grab socket reference count. This operation is valid only
703 when sk is ALREADY grabbed f.e. it is found in hash table
704 or a list and the lookup is made under lock preventing hash table
708 static __always_inline
void sock_hold(struct sock
*sk
)
710 refcount_inc(&sk
->sk_refcnt
);
713 /* Ungrab socket in the context, which assumes that socket refcnt
714 cannot hit zero, f.e. it is true in context of any socketcall.
716 static __always_inline
void __sock_put(struct sock
*sk
)
718 refcount_dec(&sk
->sk_refcnt
);
721 static inline bool sk_del_node_init(struct sock
*sk
)
723 bool rc
= __sk_del_node_init(sk
);
726 /* paranoid for a while -acme */
727 WARN_ON(refcount_read(&sk
->sk_refcnt
) == 1);
732 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
734 static inline bool __sk_nulls_del_node_init_rcu(struct sock
*sk
)
737 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
743 static inline bool sk_nulls_del_node_init_rcu(struct sock
*sk
)
745 bool rc
= __sk_nulls_del_node_init_rcu(sk
);
748 /* paranoid for a while -acme */
749 WARN_ON(refcount_read(&sk
->sk_refcnt
) == 1);
755 static inline void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
757 hlist_add_head(&sk
->sk_node
, list
);
760 static inline void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
763 __sk_add_node(sk
, list
);
766 static inline void sk_add_node_rcu(struct sock
*sk
, struct hlist_head
*list
)
769 if (IS_ENABLED(CONFIG_IPV6
) && sk
->sk_reuseport
&&
770 sk
->sk_family
== AF_INET6
)
771 hlist_add_tail_rcu(&sk
->sk_node
, list
);
773 hlist_add_head_rcu(&sk
->sk_node
, list
);
776 static inline void sk_add_node_tail_rcu(struct sock
*sk
, struct hlist_head
*list
)
779 hlist_add_tail_rcu(&sk
->sk_node
, list
);
782 static inline void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
784 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
787 static inline void __sk_nulls_add_node_tail_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
789 hlist_nulls_add_tail_rcu(&sk
->sk_nulls_node
, list
);
792 static inline void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
795 __sk_nulls_add_node_rcu(sk
, list
);
798 static inline void __sk_del_bind_node(struct sock
*sk
)
800 __hlist_del(&sk
->sk_bind_node
);
803 static inline void sk_add_bind_node(struct sock
*sk
,
804 struct hlist_head
*list
)
806 hlist_add_head(&sk
->sk_bind_node
, list
);
809 #define sk_for_each(__sk, list) \
810 hlist_for_each_entry(__sk, list, sk_node)
811 #define sk_for_each_rcu(__sk, list) \
812 hlist_for_each_entry_rcu(__sk, list, sk_node)
813 #define sk_nulls_for_each(__sk, node, list) \
814 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
815 #define sk_nulls_for_each_rcu(__sk, node, list) \
816 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
817 #define sk_for_each_from(__sk) \
818 hlist_for_each_entry_from(__sk, sk_node)
819 #define sk_nulls_for_each_from(__sk, node) \
820 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
821 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
822 #define sk_for_each_safe(__sk, tmp, list) \
823 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
824 #define sk_for_each_bound(__sk, list) \
825 hlist_for_each_entry(__sk, list, sk_bind_node)
828 * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset
829 * @tpos: the type * to use as a loop cursor.
830 * @pos: the &struct hlist_node to use as a loop cursor.
831 * @head: the head for your list.
832 * @offset: offset of hlist_node within the struct.
835 #define sk_for_each_entry_offset_rcu(tpos, pos, head, offset) \
836 for (pos = rcu_dereference(hlist_first_rcu(head)); \
838 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
839 pos = rcu_dereference(hlist_next_rcu(pos)))
841 static inline struct user_namespace
*sk_user_ns(struct sock
*sk
)
843 /* Careful only use this in a context where these parameters
844 * can not change and must all be valid, such as recvmsg from
847 return sk
->sk_socket
->file
->f_cred
->user_ns
;
861 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
862 SOCK_DBG
, /* %SO_DEBUG setting */
863 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
864 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
865 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
866 SOCK_MEMALLOC
, /* VM depends on this socket for swapping */
867 SOCK_TIMESTAMPING_RX_SOFTWARE
, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
868 SOCK_FASYNC
, /* fasync() active */
870 SOCK_ZEROCOPY
, /* buffers from userspace */
871 SOCK_WIFI_STATUS
, /* push wifi status to userspace */
872 SOCK_NOFCS
, /* Tell NIC not to do the Ethernet FCS.
873 * Will use last 4 bytes of packet sent from
874 * user-space instead.
876 SOCK_FILTER_LOCKED
, /* Filter cannot be changed anymore */
877 SOCK_SELECT_ERR_QUEUE
, /* Wake select on error queue */
878 SOCK_RCU_FREE
, /* wait rcu grace period in sk_destruct() */
880 SOCK_XDP
, /* XDP is attached */
881 SOCK_TSTAMP_NEW
, /* Indicates 64 bit timestamps always */
884 #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
886 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
888 nsk
->sk_flags
= osk
->sk_flags
;
891 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
893 __set_bit(flag
, &sk
->sk_flags
);
896 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
898 __clear_bit(flag
, &sk
->sk_flags
);
901 static inline void sock_valbool_flag(struct sock
*sk
, enum sock_flags bit
,
905 sock_set_flag(sk
, bit
);
907 sock_reset_flag(sk
, bit
);
910 static inline bool sock_flag(const struct sock
*sk
, enum sock_flags flag
)
912 return test_bit(flag
, &sk
->sk_flags
);
916 DECLARE_STATIC_KEY_FALSE(memalloc_socks_key
);
917 static inline int sk_memalloc_socks(void)
919 return static_branch_unlikely(&memalloc_socks_key
);
922 void __receive_sock(struct file
*file
);
925 static inline int sk_memalloc_socks(void)
930 static inline void __receive_sock(struct file
*file
)
934 static inline gfp_t
sk_gfp_mask(const struct sock
*sk
, gfp_t gfp_mask
)
936 return gfp_mask
| (sk
->sk_allocation
& __GFP_MEMALLOC
);
939 static inline void sk_acceptq_removed(struct sock
*sk
)
941 WRITE_ONCE(sk
->sk_ack_backlog
, sk
->sk_ack_backlog
- 1);
944 static inline void sk_acceptq_added(struct sock
*sk
)
946 WRITE_ONCE(sk
->sk_ack_backlog
, sk
->sk_ack_backlog
+ 1);
949 /* Note: If you think the test should be:
950 * return READ_ONCE(sk->sk_ack_backlog) >= READ_ONCE(sk->sk_max_ack_backlog);
951 * Then please take a look at commit 64a146513f8f ("[NET]: Revert incorrect accept queue backlog changes.")
953 static inline bool sk_acceptq_is_full(const struct sock
*sk
)
955 return READ_ONCE(sk
->sk_ack_backlog
) > READ_ONCE(sk
->sk_max_ack_backlog
);
959 * Compute minimal free write space needed to queue new packets.
961 static inline int sk_stream_min_wspace(const struct sock
*sk
)
963 return READ_ONCE(sk
->sk_wmem_queued
) >> 1;
966 static inline int sk_stream_wspace(const struct sock
*sk
)
968 return READ_ONCE(sk
->sk_sndbuf
) - READ_ONCE(sk
->sk_wmem_queued
);
971 static inline void sk_wmem_queued_add(struct sock
*sk
, int val
)
973 WRITE_ONCE(sk
->sk_wmem_queued
, sk
->sk_wmem_queued
+ val
);
976 void sk_stream_write_space(struct sock
*sk
);
978 /* OOB backlog add */
979 static inline void __sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
981 /* dont let skb dst not refcounted, we are going to leave rcu lock */
984 if (!sk
->sk_backlog
.tail
)
985 WRITE_ONCE(sk
->sk_backlog
.head
, skb
);
987 sk
->sk_backlog
.tail
->next
= skb
;
989 WRITE_ONCE(sk
->sk_backlog
.tail
, skb
);
994 * Take into account size of receive queue and backlog queue
995 * Do not take into account this skb truesize,
996 * to allow even a single big packet to come.
998 static inline bool sk_rcvqueues_full(const struct sock
*sk
, unsigned int limit
)
1000 unsigned int qsize
= sk
->sk_backlog
.len
+ atomic_read(&sk
->sk_rmem_alloc
);
1002 return qsize
> limit
;
1005 /* The per-socket spinlock must be held here. */
1006 static inline __must_check
int sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
,
1009 if (sk_rcvqueues_full(sk
, limit
))
1013 * If the skb was allocated from pfmemalloc reserves, only
1014 * allow SOCK_MEMALLOC sockets to use it as this socket is
1015 * helping free memory
1017 if (skb_pfmemalloc(skb
) && !sock_flag(sk
, SOCK_MEMALLOC
))
1020 __sk_add_backlog(sk
, skb
);
1021 sk
->sk_backlog
.len
+= skb
->truesize
;
1025 int __sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
1027 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1029 if (sk_memalloc_socks() && skb_pfmemalloc(skb
))
1030 return __sk_backlog_rcv(sk
, skb
);
1032 return sk
->sk_backlog_rcv(sk
, skb
);
1035 static inline void sk_incoming_cpu_update(struct sock
*sk
)
1037 int cpu
= raw_smp_processor_id();
1039 if (unlikely(READ_ONCE(sk
->sk_incoming_cpu
) != cpu
))
1040 WRITE_ONCE(sk
->sk_incoming_cpu
, cpu
);
1043 static inline void sock_rps_record_flow_hash(__u32 hash
)
1046 struct rps_sock_flow_table
*sock_flow_table
;
1049 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
1050 rps_record_sock_flow(sock_flow_table
, hash
);
1055 static inline void sock_rps_record_flow(const struct sock
*sk
)
1058 if (static_branch_unlikely(&rfs_needed
)) {
1059 /* Reading sk->sk_rxhash might incur an expensive cache line
1062 * TCP_ESTABLISHED does cover almost all states where RFS
1063 * might be useful, and is cheaper [1] than testing :
1064 * IPv4: inet_sk(sk)->inet_daddr
1065 * IPv6: ipv6_addr_any(&sk->sk_v6_daddr)
1066 * OR an additional socket flag
1067 * [1] : sk_state and sk_prot are in the same cache line.
1069 if (sk
->sk_state
== TCP_ESTABLISHED
)
1070 sock_rps_record_flow_hash(sk
->sk_rxhash
);
1075 static inline void sock_rps_save_rxhash(struct sock
*sk
,
1076 const struct sk_buff
*skb
)
1079 if (unlikely(sk
->sk_rxhash
!= skb
->hash
))
1080 sk
->sk_rxhash
= skb
->hash
;
1084 static inline void sock_rps_reset_rxhash(struct sock
*sk
)
1091 #define sk_wait_event(__sk, __timeo, __condition, __wait) \
1093 release_sock(__sk); \
1094 __rc = __condition; \
1096 *(__timeo) = wait_woken(__wait, \
1097 TASK_INTERRUPTIBLE, \
1100 sched_annotate_sleep(); \
1102 __rc = __condition; \
1106 int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
1107 int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
1108 void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
1109 int sk_stream_error(struct sock
*sk
, int flags
, int err
);
1110 void sk_stream_kill_queues(struct sock
*sk
);
1111 void sk_set_memalloc(struct sock
*sk
);
1112 void sk_clear_memalloc(struct sock
*sk
);
1114 void __sk_flush_backlog(struct sock
*sk
);
1116 static inline bool sk_flush_backlog(struct sock
*sk
)
1118 if (unlikely(READ_ONCE(sk
->sk_backlog
.tail
))) {
1119 __sk_flush_backlog(sk
);
1125 int sk_wait_data(struct sock
*sk
, long *timeo
, const struct sk_buff
*skb
);
1127 struct request_sock_ops
;
1128 struct timewait_sock_ops
;
1129 struct inet_hashinfo
;
1130 struct raw_hashinfo
;
1131 struct smc_hashinfo
;
1136 * caches using SLAB_TYPESAFE_BY_RCU should let .next pointer from nulls nodes
1137 * un-modified. Special care is taken when initializing object to zero.
1139 static inline void sk_prot_clear_nulls(struct sock
*sk
, int size
)
1141 if (offsetof(struct sock
, sk_node
.next
) != 0)
1142 memset(sk
, 0, offsetof(struct sock
, sk_node
.next
));
1143 memset(&sk
->sk_node
.pprev
, 0,
1144 size
- offsetof(struct sock
, sk_node
.pprev
));
1147 /* Networking protocol blocks we attach to sockets.
1148 * socket layer -> transport layer interface
1151 void (*close
)(struct sock
*sk
,
1153 int (*pre_connect
)(struct sock
*sk
,
1154 struct sockaddr
*uaddr
,
1156 int (*connect
)(struct sock
*sk
,
1157 struct sockaddr
*uaddr
,
1159 int (*disconnect
)(struct sock
*sk
, int flags
);
1161 struct sock
* (*accept
)(struct sock
*sk
, int flags
, int *err
,
1164 int (*ioctl
)(struct sock
*sk
, int cmd
,
1166 int (*init
)(struct sock
*sk
);
1167 void (*destroy
)(struct sock
*sk
);
1168 void (*shutdown
)(struct sock
*sk
, int how
);
1169 int (*setsockopt
)(struct sock
*sk
, int level
,
1170 int optname
, sockptr_t optval
,
1171 unsigned int optlen
);
1172 int (*getsockopt
)(struct sock
*sk
, int level
,
1173 int optname
, char __user
*optval
,
1174 int __user
*option
);
1175 void (*keepalive
)(struct sock
*sk
, int valbool
);
1176 #ifdef CONFIG_COMPAT
1177 int (*compat_ioctl
)(struct sock
*sk
,
1178 unsigned int cmd
, unsigned long arg
);
1180 int (*sendmsg
)(struct sock
*sk
, struct msghdr
*msg
,
1182 int (*recvmsg
)(struct sock
*sk
, struct msghdr
*msg
,
1183 size_t len
, int noblock
, int flags
,
1185 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
1186 int offset
, size_t size
, int flags
);
1187 int (*bind
)(struct sock
*sk
,
1188 struct sockaddr
*addr
, int addr_len
);
1189 int (*bind_add
)(struct sock
*sk
,
1190 struct sockaddr
*addr
, int addr_len
);
1192 int (*backlog_rcv
) (struct sock
*sk
,
1193 struct sk_buff
*skb
);
1194 bool (*bpf_bypass_getsockopt
)(int level
,
1197 void (*release_cb
)(struct sock
*sk
);
1199 /* Keeping track of sk's, looking them up, and port selection methods. */
1200 int (*hash
)(struct sock
*sk
);
1201 void (*unhash
)(struct sock
*sk
);
1202 void (*rehash
)(struct sock
*sk
);
1203 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
1204 #ifdef CONFIG_BPF_SYSCALL
1205 int (*psock_update_sk_prot
)(struct sock
*sk
,
1206 struct sk_psock
*psock
,
1210 /* Keeping track of sockets in use */
1211 #ifdef CONFIG_PROC_FS
1212 unsigned int inuse_idx
;
1215 bool (*stream_memory_free
)(const struct sock
*sk
, int wake
);
1216 bool (*sock_is_readable
)(struct sock
*sk
);
1217 /* Memory pressure */
1218 void (*enter_memory_pressure
)(struct sock
*sk
);
1219 void (*leave_memory_pressure
)(struct sock
*sk
);
1220 atomic_long_t
*memory_allocated
; /* Current allocated memory. */
1221 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
1223 * Pressure flag: try to collapse.
1224 * Technical note: it is used by multiple contexts non atomically.
1225 * All the __sk_mem_schedule() is of this nature: accounting
1226 * is strict, actions are advisory and have some latency.
1228 unsigned long *memory_pressure
;
1233 u32 sysctl_wmem_offset
;
1234 u32 sysctl_rmem_offset
;
1239 struct kmem_cache
*slab
;
1240 unsigned int obj_size
;
1241 slab_flags_t slab_flags
;
1242 unsigned int useroffset
; /* Usercopy region offset */
1243 unsigned int usersize
; /* Usercopy region size */
1245 unsigned int __percpu
*orphan_count
;
1247 struct request_sock_ops
*rsk_prot
;
1248 struct timewait_sock_ops
*twsk_prot
;
1251 struct inet_hashinfo
*hashinfo
;
1252 struct udp_table
*udp_table
;
1253 struct raw_hashinfo
*raw_hash
;
1254 struct smc_hashinfo
*smc_hash
;
1257 struct module
*owner
;
1261 struct list_head node
;
1262 #ifdef SOCK_REFCNT_DEBUG
1265 int (*diag_destroy
)(struct sock
*sk
, int err
);
1266 } __randomize_layout
;
1268 int proto_register(struct proto
*prot
, int alloc_slab
);
1269 void proto_unregister(struct proto
*prot
);
1270 int sock_load_diag_module(int family
, int protocol
);
1272 #ifdef SOCK_REFCNT_DEBUG
1273 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
1275 atomic_inc(&sk
->sk_prot
->socks
);
1278 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
1280 atomic_dec(&sk
->sk_prot
->socks
);
1281 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
1282 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
1285 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
1287 if (refcount_read(&sk
->sk_refcnt
) != 1)
1288 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
1289 sk
->sk_prot
->name
, sk
, refcount_read(&sk
->sk_refcnt
));
1291 #else /* SOCK_REFCNT_DEBUG */
1292 #define sk_refcnt_debug_inc(sk) do { } while (0)
1293 #define sk_refcnt_debug_dec(sk) do { } while (0)
1294 #define sk_refcnt_debug_release(sk) do { } while (0)
1295 #endif /* SOCK_REFCNT_DEBUG */
1297 INDIRECT_CALLABLE_DECLARE(bool tcp_stream_memory_free(const struct sock
*sk
, int wake
));
1299 static inline bool __sk_stream_memory_free(const struct sock
*sk
, int wake
)
1301 if (READ_ONCE(sk
->sk_wmem_queued
) >= READ_ONCE(sk
->sk_sndbuf
))
1305 return sk
->sk_prot
->stream_memory_free
?
1306 INDIRECT_CALL_1(sk
->sk_prot
->stream_memory_free
,
1307 tcp_stream_memory_free
,
1310 return sk
->sk_prot
->stream_memory_free
?
1311 sk
->sk_prot
->stream_memory_free(sk
, wake
) : true;
1315 static inline bool sk_stream_memory_free(const struct sock
*sk
)
1317 return __sk_stream_memory_free(sk
, 0);
1320 static inline bool __sk_stream_is_writeable(const struct sock
*sk
, int wake
)
1322 return sk_stream_wspace(sk
) >= sk_stream_min_wspace(sk
) &&
1323 __sk_stream_memory_free(sk
, wake
);
1326 static inline bool sk_stream_is_writeable(const struct sock
*sk
)
1328 return __sk_stream_is_writeable(sk
, 0);
1331 static inline int sk_under_cgroup_hierarchy(struct sock
*sk
,
1332 struct cgroup
*ancestor
)
1334 #ifdef CONFIG_SOCK_CGROUP_DATA
1335 return cgroup_is_descendant(sock_cgroup_ptr(&sk
->sk_cgrp_data
),
1342 static inline bool sk_has_memory_pressure(const struct sock
*sk
)
1344 return sk
->sk_prot
->memory_pressure
!= NULL
;
1347 static inline bool sk_under_memory_pressure(const struct sock
*sk
)
1349 if (!sk
->sk_prot
->memory_pressure
)
1352 if (mem_cgroup_sockets_enabled
&& sk
->sk_memcg
&&
1353 mem_cgroup_under_socket_pressure(sk
->sk_memcg
))
1356 return !!*sk
->sk_prot
->memory_pressure
;
1360 sk_memory_allocated(const struct sock
*sk
)
1362 return atomic_long_read(sk
->sk_prot
->memory_allocated
);
1366 sk_memory_allocated_add(struct sock
*sk
, int amt
)
1368 return atomic_long_add_return(amt
, sk
->sk_prot
->memory_allocated
);
1372 sk_memory_allocated_sub(struct sock
*sk
, int amt
)
1374 atomic_long_sub(amt
, sk
->sk_prot
->memory_allocated
);
1377 #define SK_ALLOC_PERCPU_COUNTER_BATCH 16
1379 static inline void sk_sockets_allocated_dec(struct sock
*sk
)
1381 percpu_counter_add_batch(sk
->sk_prot
->sockets_allocated
, -1,
1382 SK_ALLOC_PERCPU_COUNTER_BATCH
);
1385 static inline void sk_sockets_allocated_inc(struct sock
*sk
)
1387 percpu_counter_add_batch(sk
->sk_prot
->sockets_allocated
, 1,
1388 SK_ALLOC_PERCPU_COUNTER_BATCH
);
1392 sk_sockets_allocated_read_positive(struct sock
*sk
)
1394 return percpu_counter_read_positive(sk
->sk_prot
->sockets_allocated
);
1398 proto_sockets_allocated_sum_positive(struct proto
*prot
)
1400 return percpu_counter_sum_positive(prot
->sockets_allocated
);
1404 proto_memory_allocated(struct proto
*prot
)
1406 return atomic_long_read(prot
->memory_allocated
);
1410 proto_memory_pressure(struct proto
*prot
)
1412 if (!prot
->memory_pressure
)
1414 return !!*prot
->memory_pressure
;
1418 #ifdef CONFIG_PROC_FS
1419 /* Called with local bh disabled */
1420 void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int inc
);
1421 int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
1422 int sock_inuse_get(struct net
*net
);
1424 static inline void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
1431 /* With per-bucket locks this operation is not-atomic, so that
1432 * this version is not worse.
1434 static inline int __sk_prot_rehash(struct sock
*sk
)
1436 sk
->sk_prot
->unhash(sk
);
1437 return sk
->sk_prot
->hash(sk
);
1440 /* About 10 seconds */
1441 #define SOCK_DESTROY_TIME (10*HZ)
1443 /* Sockets 0-1023 can't be bound to unless you are superuser */
1444 #define PROT_SOCK 1024
1446 #define SHUTDOWN_MASK 3
1447 #define RCV_SHUTDOWN 1
1448 #define SEND_SHUTDOWN 2
1450 #define SOCK_BINDADDR_LOCK 4
1451 #define SOCK_BINDPORT_LOCK 8
1453 struct socket_alloc
{
1454 struct socket socket
;
1455 struct inode vfs_inode
;
1458 static inline struct socket
*SOCKET_I(struct inode
*inode
)
1460 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
1463 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
1465 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
1469 * Functions for memory accounting
1471 int __sk_mem_raise_allocated(struct sock
*sk
, int size
, int amt
, int kind
);
1472 int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
1473 void __sk_mem_reduce_allocated(struct sock
*sk
, int amount
);
1474 void __sk_mem_reclaim(struct sock
*sk
, int amount
);
1476 /* We used to have PAGE_SIZE here, but systems with 64KB pages
1477 * do not necessarily have 16x time more memory than 4KB ones.
1479 #define SK_MEM_QUANTUM 4096
1480 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1481 #define SK_MEM_SEND 0
1482 #define SK_MEM_RECV 1
1484 /* sysctl_mem values are in pages, we convert them in SK_MEM_QUANTUM units */
1485 static inline long sk_prot_mem_limits(const struct sock
*sk
, int index
)
1487 long val
= READ_ONCE(sk
->sk_prot
->sysctl_mem
[index
]);
1489 #if PAGE_SIZE > SK_MEM_QUANTUM
1490 val
<<= PAGE_SHIFT
- SK_MEM_QUANTUM_SHIFT
;
1491 #elif PAGE_SIZE < SK_MEM_QUANTUM
1492 val
>>= SK_MEM_QUANTUM_SHIFT
- PAGE_SHIFT
;
1497 static inline int sk_mem_pages(int amt
)
1499 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
1502 static inline bool sk_has_account(struct sock
*sk
)
1504 /* return true if protocol supports memory accounting */
1505 return !!sk
->sk_prot
->memory_allocated
;
1508 static inline bool sk_wmem_schedule(struct sock
*sk
, int size
)
1510 if (!sk_has_account(sk
))
1512 return size
<= sk
->sk_forward_alloc
||
1513 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
1517 sk_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
, int size
)
1519 if (!sk_has_account(sk
))
1521 return size
<= sk
->sk_forward_alloc
||
1522 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
) ||
1523 skb_pfmemalloc(skb
);
1526 static inline void sk_mem_reclaim(struct sock
*sk
)
1528 if (!sk_has_account(sk
))
1530 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
1531 __sk_mem_reclaim(sk
, sk
->sk_forward_alloc
);
1534 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
1536 if (!sk_has_account(sk
))
1538 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
1539 __sk_mem_reclaim(sk
, sk
->sk_forward_alloc
- 1);
1542 static inline void sk_mem_charge(struct sock
*sk
, int size
)
1544 if (!sk_has_account(sk
))
1546 sk
->sk_forward_alloc
-= size
;
1549 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
1551 if (!sk_has_account(sk
))
1553 sk
->sk_forward_alloc
+= size
;
1555 /* Avoid a possible overflow.
1556 * TCP send queues can make this happen, if sk_mem_reclaim()
1557 * is not called and more than 2 GBytes are released at once.
1559 * If we reach 2 MBytes, reclaim 1 MBytes right now, there is
1560 * no need to hold that much forward allocation anyway.
1562 if (unlikely(sk
->sk_forward_alloc
>= 1 << 21))
1563 __sk_mem_reclaim(sk
, 1 << 20);
1566 DECLARE_STATIC_KEY_FALSE(tcp_tx_skb_cache_key
);
1567 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
1569 sk_wmem_queued_add(sk
, -skb
->truesize
);
1570 sk_mem_uncharge(sk
, skb
->truesize
);
1571 if (static_branch_unlikely(&tcp_tx_skb_cache_key
) &&
1572 !sk
->sk_tx_skb_cache
&& !skb_cloned(skb
)) {
1574 skb_zcopy_clear(skb
, true);
1575 sk
->sk_tx_skb_cache
= skb
;
1581 static inline void sock_release_ownership(struct sock
*sk
)
1583 if (sk
->sk_lock
.owned
) {
1584 sk
->sk_lock
.owned
= 0;
1586 /* The sk_lock has mutex_unlock() semantics: */
1587 mutex_release(&sk
->sk_lock
.dep_map
, _RET_IP_
);
1592 * Macro so as to not evaluate some arguments when
1593 * lockdep is not enabled.
1595 * Mark both the sk_lock and the sk_lock.slock as a
1596 * per-address-family lock class.
1598 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1600 sk->sk_lock.owned = 0; \
1601 init_waitqueue_head(&sk->sk_lock.wq); \
1602 spin_lock_init(&(sk)->sk_lock.slock); \
1603 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1604 sizeof((sk)->sk_lock)); \
1605 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1607 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1610 static inline bool lockdep_sock_is_held(const struct sock
*sk
)
1612 return lockdep_is_held(&sk
->sk_lock
) ||
1613 lockdep_is_held(&sk
->sk_lock
.slock
);
1616 void lock_sock_nested(struct sock
*sk
, int subclass
);
1618 static inline void lock_sock(struct sock
*sk
)
1620 lock_sock_nested(sk
, 0);
1623 void __lock_sock(struct sock
*sk
);
1624 void __release_sock(struct sock
*sk
);
1625 void release_sock(struct sock
*sk
);
1627 /* BH context may only use the following locking interface. */
1628 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1629 #define bh_lock_sock_nested(__sk) \
1630 spin_lock_nested(&((__sk)->sk_lock.slock), \
1631 SINGLE_DEPTH_NESTING)
1632 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1634 bool __lock_sock_fast(struct sock
*sk
) __acquires(&sk
->sk_lock
.slock
);
1637 * lock_sock_fast - fast version of lock_sock
1640 * This version should be used for very small section, where process wont block
1641 * return false if fast path is taken:
1643 * sk_lock.slock locked, owned = 0, BH disabled
1645 * return true if slow path is taken:
1647 * sk_lock.slock unlocked, owned = 1, BH enabled
1649 static inline bool lock_sock_fast(struct sock
*sk
)
1651 /* The sk_lock has mutex_lock() semantics here. */
1652 mutex_acquire(&sk
->sk_lock
.dep_map
, 0, 0, _RET_IP_
);
1654 return __lock_sock_fast(sk
);
1657 /* fast socket lock variant for caller already holding a [different] socket lock */
1658 static inline bool lock_sock_fast_nested(struct sock
*sk
)
1660 mutex_acquire(&sk
->sk_lock
.dep_map
, SINGLE_DEPTH_NESTING
, 0, _RET_IP_
);
1662 return __lock_sock_fast(sk
);
1666 * unlock_sock_fast - complement of lock_sock_fast
1670 * fast unlock socket for user context.
1671 * If slow mode is on, we call regular release_sock()
1673 static inline void unlock_sock_fast(struct sock
*sk
, bool slow
)
1674 __releases(&sk
->sk_lock
.slock
)
1678 __release(&sk
->sk_lock
.slock
);
1680 mutex_release(&sk
->sk_lock
.dep_map
, _RET_IP_
);
1681 spin_unlock_bh(&sk
->sk_lock
.slock
);
1685 /* Used by processes to "lock" a socket state, so that
1686 * interrupts and bottom half handlers won't change it
1687 * from under us. It essentially blocks any incoming
1688 * packets, so that we won't get any new data or any
1689 * packets that change the state of the socket.
1691 * While locked, BH processing will add new packets to
1692 * the backlog queue. This queue is processed by the
1693 * owner of the socket lock right before it is released.
1695 * Since ~2.3.5 it is also exclusive sleep lock serializing
1696 * accesses from user process context.
1699 static inline void sock_owned_by_me(const struct sock
*sk
)
1701 #ifdef CONFIG_LOCKDEP
1702 WARN_ON_ONCE(!lockdep_sock_is_held(sk
) && debug_locks
);
1706 static inline bool sock_owned_by_user(const struct sock
*sk
)
1708 sock_owned_by_me(sk
);
1709 return sk
->sk_lock
.owned
;
1712 static inline bool sock_owned_by_user_nocheck(const struct sock
*sk
)
1714 return sk
->sk_lock
.owned
;
1717 /* no reclassification while locks are held */
1718 static inline bool sock_allow_reclassification(const struct sock
*csk
)
1720 struct sock
*sk
= (struct sock
*)csk
;
1722 return !sk
->sk_lock
.owned
&& !spin_is_locked(&sk
->sk_lock
.slock
);
1725 struct sock
*sk_alloc(struct net
*net
, int family
, gfp_t priority
,
1726 struct proto
*prot
, int kern
);
1727 void sk_free(struct sock
*sk
);
1728 void sk_destruct(struct sock
*sk
);
1729 struct sock
*sk_clone_lock(const struct sock
*sk
, const gfp_t priority
);
1730 void sk_free_unlock_clone(struct sock
*sk
);
1732 struct sk_buff
*sock_wmalloc(struct sock
*sk
, unsigned long size
, int force
,
1734 void __sock_wfree(struct sk_buff
*skb
);
1735 void sock_wfree(struct sk_buff
*skb
);
1736 struct sk_buff
*sock_omalloc(struct sock
*sk
, unsigned long size
,
1738 void skb_orphan_partial(struct sk_buff
*skb
);
1739 void sock_rfree(struct sk_buff
*skb
);
1740 void sock_efree(struct sk_buff
*skb
);
1742 void sock_edemux(struct sk_buff
*skb
);
1743 void sock_pfree(struct sk_buff
*skb
);
1745 #define sock_edemux sock_efree
1748 int sock_setsockopt(struct socket
*sock
, int level
, int op
,
1749 sockptr_t optval
, unsigned int optlen
);
1751 int sock_getsockopt(struct socket
*sock
, int level
, int op
,
1752 char __user
*optval
, int __user
*optlen
);
1753 int sock_gettstamp(struct socket
*sock
, void __user
*userstamp
,
1754 bool timeval
, bool time32
);
1755 struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
, unsigned long size
,
1756 int noblock
, int *errcode
);
1757 struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
, unsigned long header_len
,
1758 unsigned long data_len
, int noblock
,
1759 int *errcode
, int max_page_order
);
1760 void *sock_kmalloc(struct sock
*sk
, int size
, gfp_t priority
);
1761 void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
1762 void sock_kzfree_s(struct sock
*sk
, void *mem
, int size
);
1763 void sk_send_sigurg(struct sock
*sk
);
1765 struct sockcm_cookie
{
1771 static inline void sockcm_init(struct sockcm_cookie
*sockc
,
1772 const struct sock
*sk
)
1774 *sockc
= (struct sockcm_cookie
) { .tsflags
= sk
->sk_tsflags
};
1777 int __sock_cmsg_send(struct sock
*sk
, struct msghdr
*msg
, struct cmsghdr
*cmsg
,
1778 struct sockcm_cookie
*sockc
);
1779 int sock_cmsg_send(struct sock
*sk
, struct msghdr
*msg
,
1780 struct sockcm_cookie
*sockc
);
1783 * Functions to fill in entries in struct proto_ops when a protocol
1784 * does not implement a particular function.
1786 int sock_no_bind(struct socket
*, struct sockaddr
*, int);
1787 int sock_no_connect(struct socket
*, struct sockaddr
*, int, int);
1788 int sock_no_socketpair(struct socket
*, struct socket
*);
1789 int sock_no_accept(struct socket
*, struct socket
*, int, bool);
1790 int sock_no_getname(struct socket
*, struct sockaddr
*, int);
1791 int sock_no_ioctl(struct socket
*, unsigned int, unsigned long);
1792 int sock_no_listen(struct socket
*, int);
1793 int sock_no_shutdown(struct socket
*, int);
1794 int sock_no_sendmsg(struct socket
*, struct msghdr
*, size_t);
1795 int sock_no_sendmsg_locked(struct sock
*sk
, struct msghdr
*msg
, size_t len
);
1796 int sock_no_recvmsg(struct socket
*, struct msghdr
*, size_t, int);
1797 int sock_no_mmap(struct file
*file
, struct socket
*sock
,
1798 struct vm_area_struct
*vma
);
1799 ssize_t
sock_no_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
1800 size_t size
, int flags
);
1801 ssize_t
sock_no_sendpage_locked(struct sock
*sk
, struct page
*page
,
1802 int offset
, size_t size
, int flags
);
1805 * Functions to fill in entries in struct proto_ops when a protocol
1806 * uses the inet style.
1808 int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1809 char __user
*optval
, int __user
*optlen
);
1810 int sock_common_recvmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
,
1812 int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1813 sockptr_t optval
, unsigned int optlen
);
1815 void sk_common_release(struct sock
*sk
);
1818 * Default socket callbacks and setup code
1821 /* Initialise core socket variables */
1822 void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1825 * Socket reference counting postulates.
1827 * * Each user of socket SHOULD hold a reference count.
1828 * * Each access point to socket (an hash table bucket, reference from a list,
1829 * running timer, skb in flight MUST hold a reference count.
1830 * * When reference count hits 0, it means it will never increase back.
1831 * * When reference count hits 0, it means that no references from
1832 * outside exist to this socket and current process on current CPU
1833 * is last user and may/should destroy this socket.
1834 * * sk_free is called from any context: process, BH, IRQ. When
1835 * it is called, socket has no references from outside -> sk_free
1836 * may release descendant resources allocated by the socket, but
1837 * to the time when it is called, socket is NOT referenced by any
1838 * hash tables, lists etc.
1839 * * Packets, delivered from outside (from network or from another process)
1840 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1841 * when they sit in queue. Otherwise, packets will leak to hole, when
1842 * socket is looked up by one cpu and unhasing is made by another CPU.
1843 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1844 * (leak to backlog). Packet socket does all the processing inside
1845 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1846 * use separate SMP lock, so that they are prone too.
1849 /* Ungrab socket and destroy it, if it was the last reference. */
1850 static inline void sock_put(struct sock
*sk
)
1852 if (refcount_dec_and_test(&sk
->sk_refcnt
))
1855 /* Generic version of sock_put(), dealing with all sockets
1856 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
1858 void sock_gen_put(struct sock
*sk
);
1860 int __sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
, const int nested
,
1861 unsigned int trim_cap
, bool refcounted
);
1862 static inline int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
1865 return __sk_receive_skb(sk
, skb
, nested
, 1, true);
1868 static inline void sk_tx_queue_set(struct sock
*sk
, int tx_queue
)
1870 /* sk_tx_queue_mapping accept only upto a 16-bit value */
1871 if (WARN_ON_ONCE((unsigned short)tx_queue
>= USHRT_MAX
))
1873 sk
->sk_tx_queue_mapping
= tx_queue
;
1876 #define NO_QUEUE_MAPPING USHRT_MAX
1878 static inline void sk_tx_queue_clear(struct sock
*sk
)
1880 sk
->sk_tx_queue_mapping
= NO_QUEUE_MAPPING
;
1883 static inline int sk_tx_queue_get(const struct sock
*sk
)
1885 if (sk
&& sk
->sk_tx_queue_mapping
!= NO_QUEUE_MAPPING
)
1886 return sk
->sk_tx_queue_mapping
;
1891 static inline void sk_rx_queue_set(struct sock
*sk
, const struct sk_buff
*skb
)
1893 #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
1894 if (skb_rx_queue_recorded(skb
)) {
1895 u16 rx_queue
= skb_get_rx_queue(skb
);
1897 if (WARN_ON_ONCE(rx_queue
== NO_QUEUE_MAPPING
))
1900 sk
->sk_rx_queue_mapping
= rx_queue
;
1905 static inline void sk_rx_queue_clear(struct sock
*sk
)
1907 #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
1908 sk
->sk_rx_queue_mapping
= NO_QUEUE_MAPPING
;
1912 static inline int sk_rx_queue_get(const struct sock
*sk
)
1914 #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
1915 if (sk
&& sk
->sk_rx_queue_mapping
!= NO_QUEUE_MAPPING
)
1916 return sk
->sk_rx_queue_mapping
;
1922 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1924 sk
->sk_socket
= sock
;
1927 static inline wait_queue_head_t
*sk_sleep(struct sock
*sk
)
1929 BUILD_BUG_ON(offsetof(struct socket_wq
, wait
) != 0);
1930 return &rcu_dereference_raw(sk
->sk_wq
)->wait
;
1932 /* Detach socket from process context.
1933 * Announce socket dead, detach it from wait queue and inode.
1934 * Note that parent inode held reference count on this struct sock,
1935 * we do not release it in this function, because protocol
1936 * probably wants some additional cleanups or even continuing
1937 * to work with this socket (TCP).
1939 static inline void sock_orphan(struct sock
*sk
)
1941 write_lock_bh(&sk
->sk_callback_lock
);
1942 sock_set_flag(sk
, SOCK_DEAD
);
1943 sk_set_socket(sk
, NULL
);
1945 write_unlock_bh(&sk
->sk_callback_lock
);
1948 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1950 WARN_ON(parent
->sk
);
1951 write_lock_bh(&sk
->sk_callback_lock
);
1952 rcu_assign_pointer(sk
->sk_wq
, &parent
->wq
);
1954 sk_set_socket(sk
, parent
);
1955 sk
->sk_uid
= SOCK_INODE(parent
)->i_uid
;
1956 security_sock_graft(sk
, parent
);
1957 write_unlock_bh(&sk
->sk_callback_lock
);
1960 kuid_t
sock_i_uid(struct sock
*sk
);
1961 unsigned long sock_i_ino(struct sock
*sk
);
1963 static inline kuid_t
sock_net_uid(const struct net
*net
, const struct sock
*sk
)
1965 return sk
? sk
->sk_uid
: make_kuid(net
->user_ns
, 0);
1968 static inline u32
net_tx_rndhash(void)
1970 u32 v
= prandom_u32();
1975 static inline void sk_set_txhash(struct sock
*sk
)
1977 /* This pairs with READ_ONCE() in skb_set_hash_from_sk() */
1978 WRITE_ONCE(sk
->sk_txhash
, net_tx_rndhash());
1981 static inline bool sk_rethink_txhash(struct sock
*sk
)
1983 if (sk
->sk_txhash
) {
1990 static inline struct dst_entry
*
1991 __sk_dst_get(struct sock
*sk
)
1993 return rcu_dereference_check(sk
->sk_dst_cache
,
1994 lockdep_sock_is_held(sk
));
1997 static inline struct dst_entry
*
1998 sk_dst_get(struct sock
*sk
)
2000 struct dst_entry
*dst
;
2003 dst
= rcu_dereference(sk
->sk_dst_cache
);
2004 if (dst
&& !atomic_inc_not_zero(&dst
->__refcnt
))
2010 static inline void __dst_negative_advice(struct sock
*sk
)
2012 struct dst_entry
*ndst
, *dst
= __sk_dst_get(sk
);
2014 if (dst
&& dst
->ops
->negative_advice
) {
2015 ndst
= dst
->ops
->negative_advice(dst
);
2018 rcu_assign_pointer(sk
->sk_dst_cache
, ndst
);
2019 sk_tx_queue_clear(sk
);
2020 sk
->sk_dst_pending_confirm
= 0;
2025 static inline void dst_negative_advice(struct sock
*sk
)
2027 sk_rethink_txhash(sk
);
2028 __dst_negative_advice(sk
);
2032 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
2034 struct dst_entry
*old_dst
;
2036 sk_tx_queue_clear(sk
);
2037 sk
->sk_dst_pending_confirm
= 0;
2038 old_dst
= rcu_dereference_protected(sk
->sk_dst_cache
,
2039 lockdep_sock_is_held(sk
));
2040 rcu_assign_pointer(sk
->sk_dst_cache
, dst
);
2041 dst_release(old_dst
);
2045 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
2047 struct dst_entry
*old_dst
;
2049 sk_tx_queue_clear(sk
);
2050 sk
->sk_dst_pending_confirm
= 0;
2051 old_dst
= xchg((__force
struct dst_entry
**)&sk
->sk_dst_cache
, dst
);
2052 dst_release(old_dst
);
2056 __sk_dst_reset(struct sock
*sk
)
2058 __sk_dst_set(sk
, NULL
);
2062 sk_dst_reset(struct sock
*sk
)
2064 sk_dst_set(sk
, NULL
);
2067 struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
2069 struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
2071 static inline void sk_dst_confirm(struct sock
*sk
)
2073 if (!READ_ONCE(sk
->sk_dst_pending_confirm
))
2074 WRITE_ONCE(sk
->sk_dst_pending_confirm
, 1);
2077 static inline void sock_confirm_neigh(struct sk_buff
*skb
, struct neighbour
*n
)
2079 if (skb_get_dst_pending_confirm(skb
)) {
2080 struct sock
*sk
= skb
->sk
;
2081 unsigned long now
= jiffies
;
2083 /* avoid dirtying neighbour */
2084 if (READ_ONCE(n
->confirmed
) != now
)
2085 WRITE_ONCE(n
->confirmed
, now
);
2086 if (sk
&& READ_ONCE(sk
->sk_dst_pending_confirm
))
2087 WRITE_ONCE(sk
->sk_dst_pending_confirm
, 0);
2091 bool sk_mc_loop(struct sock
*sk
);
2093 static inline bool sk_can_gso(const struct sock
*sk
)
2095 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
2098 void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
2100 static inline void sk_nocaps_add(struct sock
*sk
, netdev_features_t flags
)
2102 sk
->sk_route_nocaps
|= flags
;
2103 sk
->sk_route_caps
&= ~flags
;
2106 static inline int skb_do_copy_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
2107 struct iov_iter
*from
, char *to
,
2108 int copy
, int offset
)
2110 if (skb
->ip_summed
== CHECKSUM_NONE
) {
2112 if (!csum_and_copy_from_iter_full(to
, copy
, &csum
, from
))
2114 skb
->csum
= csum_block_add(skb
->csum
, csum
, offset
);
2115 } else if (sk
->sk_route_caps
& NETIF_F_NOCACHE_COPY
) {
2116 if (!copy_from_iter_full_nocache(to
, copy
, from
))
2118 } else if (!copy_from_iter_full(to
, copy
, from
))
2124 static inline int skb_add_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
2125 struct iov_iter
*from
, int copy
)
2127 int err
, offset
= skb
->len
;
2129 err
= skb_do_copy_data_nocache(sk
, skb
, from
, skb_put(skb
, copy
),
2132 __skb_trim(skb
, offset
);
2137 static inline int skb_copy_to_page_nocache(struct sock
*sk
, struct iov_iter
*from
,
2138 struct sk_buff
*skb
,
2144 err
= skb_do_copy_data_nocache(sk
, skb
, from
, page_address(page
) + off
,
2150 skb
->data_len
+= copy
;
2151 skb
->truesize
+= copy
;
2152 sk_wmem_queued_add(sk
, copy
);
2153 sk_mem_charge(sk
, copy
);
2158 * sk_wmem_alloc_get - returns write allocations
2161 * Return: sk_wmem_alloc minus initial offset of one
2163 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
2165 return refcount_read(&sk
->sk_wmem_alloc
) - 1;
2169 * sk_rmem_alloc_get - returns read allocations
2172 * Return: sk_rmem_alloc
2174 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
2176 return atomic_read(&sk
->sk_rmem_alloc
);
2180 * sk_has_allocations - check if allocations are outstanding
2183 * Return: true if socket has write or read allocations
2185 static inline bool sk_has_allocations(const struct sock
*sk
)
2187 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
2191 * skwq_has_sleeper - check if there are any waiting processes
2192 * @wq: struct socket_wq
2194 * Return: true if socket_wq has waiting processes
2196 * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
2197 * barrier call. They were added due to the race found within the tcp code.
2199 * Consider following tcp code paths::
2202 * sys_select receive packet
2204 * __add_wait_queue update tp->rcv_nxt
2206 * tp->rcv_nxt check sock_def_readable
2208 * schedule rcu_read_lock();
2209 * wq = rcu_dereference(sk->sk_wq);
2210 * if (wq && waitqueue_active(&wq->wait))
2211 * wake_up_interruptible(&wq->wait)
2215 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
2216 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
2217 * could then endup calling schedule and sleep forever if there are no more
2218 * data on the socket.
2221 static inline bool skwq_has_sleeper(struct socket_wq
*wq
)
2223 return wq
&& wq_has_sleeper(&wq
->wait
);
2227 * sock_poll_wait - place memory barrier behind the poll_wait call.
2229 * @sock: socket to wait on
2232 * See the comments in the wq_has_sleeper function.
2234 static inline void sock_poll_wait(struct file
*filp
, struct socket
*sock
,
2237 if (!poll_does_not_wait(p
)) {
2238 poll_wait(filp
, &sock
->wq
.wait
, p
);
2239 /* We need to be sure we are in sync with the
2240 * socket flags modification.
2242 * This memory barrier is paired in the wq_has_sleeper.
2248 static inline void skb_set_hash_from_sk(struct sk_buff
*skb
, struct sock
*sk
)
2250 /* This pairs with WRITE_ONCE() in sk_set_txhash() */
2251 u32 txhash
= READ_ONCE(sk
->sk_txhash
);
2259 void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
);
2262 * Queue a received datagram if it will fit. Stream and sequenced
2263 * protocols can't normally use this as they need to fit buffers in
2264 * and play with them.
2266 * Inlined as it's very short and called for pretty much every
2267 * packet ever received.
2269 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
2273 skb
->destructor
= sock_rfree
;
2274 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
2275 sk_mem_charge(sk
, skb
->truesize
);
2278 static inline __must_check
bool skb_set_owner_sk_safe(struct sk_buff
*skb
, struct sock
*sk
)
2280 if (sk
&& refcount_inc_not_zero(&sk
->sk_refcnt
)) {
2282 skb
->destructor
= sock_efree
;
2289 static inline void skb_prepare_for_gro(struct sk_buff
*skb
)
2291 if (skb
->destructor
!= sock_wfree
) {
2298 void sk_reset_timer(struct sock
*sk
, struct timer_list
*timer
,
2299 unsigned long expires
);
2301 void sk_stop_timer(struct sock
*sk
, struct timer_list
*timer
);
2303 void sk_stop_timer_sync(struct sock
*sk
, struct timer_list
*timer
);
2305 int __sk_queue_drop_skb(struct sock
*sk
, struct sk_buff_head
*sk_queue
,
2306 struct sk_buff
*skb
, unsigned int flags
,
2307 void (*destructor
)(struct sock
*sk
,
2308 struct sk_buff
*skb
));
2309 int __sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
2310 int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
2312 int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
);
2313 struct sk_buff
*sock_dequeue_err_skb(struct sock
*sk
);
2316 * Recover an error report and clear atomically
2319 static inline int sock_error(struct sock
*sk
)
2323 /* Avoid an atomic operation for the common case.
2324 * This is racy since another cpu/thread can change sk_err under us.
2326 if (likely(data_race(!sk
->sk_err
)))
2329 err
= xchg(&sk
->sk_err
, 0);
2333 void sk_error_report(struct sock
*sk
);
2335 static inline unsigned long sock_wspace(struct sock
*sk
)
2339 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
2340 amt
= sk
->sk_sndbuf
- refcount_read(&sk
->sk_wmem_alloc
);
2348 * We use sk->sk_wq_raw, from contexts knowing this
2349 * pointer is not NULL and cannot disappear/change.
2351 static inline void sk_set_bit(int nr
, struct sock
*sk
)
2353 if ((nr
== SOCKWQ_ASYNC_NOSPACE
|| nr
== SOCKWQ_ASYNC_WAITDATA
) &&
2354 !sock_flag(sk
, SOCK_FASYNC
))
2357 set_bit(nr
, &sk
->sk_wq_raw
->flags
);
2360 static inline void sk_clear_bit(int nr
, struct sock
*sk
)
2362 if ((nr
== SOCKWQ_ASYNC_NOSPACE
|| nr
== SOCKWQ_ASYNC_WAITDATA
) &&
2363 !sock_flag(sk
, SOCK_FASYNC
))
2366 clear_bit(nr
, &sk
->sk_wq_raw
->flags
);
2369 static inline void sk_wake_async(const struct sock
*sk
, int how
, int band
)
2371 if (sock_flag(sk
, SOCK_FASYNC
)) {
2373 sock_wake_async(rcu_dereference(sk
->sk_wq
), how
, band
);
2378 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2379 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2380 * Note: for send buffers, TCP works better if we can build two skbs at
2383 #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2385 #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2386 #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
2388 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
2392 if (sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)
2395 val
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
2397 WRITE_ONCE(sk
->sk_sndbuf
, max_t(u32
, val
, SOCK_MIN_SNDBUF
));
2400 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
,
2401 bool force_schedule
);
2404 * sk_page_frag - return an appropriate page_frag
2407 * Use the per task page_frag instead of the per socket one for
2408 * optimization when we know that we're in process context and own
2409 * everything that's associated with %current.
2411 * Both direct reclaim and page faults can nest inside other
2412 * socket operations and end up recursing into sk_page_frag()
2413 * while it's already in use: explicitly avoid task page_frag
2414 * usage if the caller is potentially doing any of them.
2415 * This assumes that page fault handlers use the GFP_NOFS flags.
2417 * Return: a per task page_frag if context allows that,
2418 * otherwise a per socket one.
2420 static inline struct page_frag
*sk_page_frag(struct sock
*sk
)
2422 if ((sk
->sk_allocation
& (__GFP_DIRECT_RECLAIM
| __GFP_MEMALLOC
| __GFP_FS
)) ==
2423 (__GFP_DIRECT_RECLAIM
| __GFP_FS
))
2424 return ¤t
->task_frag
;
2426 return &sk
->sk_frag
;
2429 bool sk_page_frag_refill(struct sock
*sk
, struct page_frag
*pfrag
);
2432 * Default write policy as shown to user space via poll/select/SIGIO
2434 static inline bool sock_writeable(const struct sock
*sk
)
2436 return refcount_read(&sk
->sk_wmem_alloc
) < (READ_ONCE(sk
->sk_sndbuf
) >> 1);
2439 static inline gfp_t
gfp_any(void)
2441 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
2444 static inline gfp_t
gfp_memcg_charge(void)
2446 return in_softirq() ? GFP_NOWAIT
: GFP_KERNEL
;
2449 static inline long sock_rcvtimeo(const struct sock
*sk
, bool noblock
)
2451 return noblock
? 0 : sk
->sk_rcvtimeo
;
2454 static inline long sock_sndtimeo(const struct sock
*sk
, bool noblock
)
2456 return noblock
? 0 : sk
->sk_sndtimeo
;
2459 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
2461 int v
= waitall
? len
: min_t(int, READ_ONCE(sk
->sk_rcvlowat
), len
);
2466 /* Alas, with timeout socket operations are not restartable.
2467 * Compare this to poll().
2469 static inline int sock_intr_errno(long timeo
)
2471 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
2474 struct sock_skb_cb
{
2478 /* Store sock_skb_cb at the end of skb->cb[] so protocol families
2479 * using skb->cb[] would keep using it directly and utilize its
2480 * alignement guarantee.
2482 #define SOCK_SKB_CB_OFFSET ((sizeof_field(struct sk_buff, cb) - \
2483 sizeof(struct sock_skb_cb)))
2485 #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2486 SOCK_SKB_CB_OFFSET))
2488 #define sock_skb_cb_check_size(size) \
2489 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
2492 sock_skb_set_dropcount(const struct sock
*sk
, struct sk_buff
*skb
)
2494 SOCK_SKB_CB(skb
)->dropcount
= sock_flag(sk
, SOCK_RXQ_OVFL
) ?
2495 atomic_read(&sk
->sk_drops
) : 0;
2498 static inline void sk_drops_add(struct sock
*sk
, const struct sk_buff
*skb
)
2500 int segs
= max_t(u16
, 1, skb_shinfo(skb
)->gso_segs
);
2502 atomic_add(segs
, &sk
->sk_drops
);
2505 static inline ktime_t
sock_read_timestamp(struct sock
*sk
)
2507 #if BITS_PER_LONG==32
2512 seq
= read_seqbegin(&sk
->sk_stamp_seq
);
2514 } while (read_seqretry(&sk
->sk_stamp_seq
, seq
));
2518 return READ_ONCE(sk
->sk_stamp
);
2522 static inline void sock_write_timestamp(struct sock
*sk
, ktime_t kt
)
2524 #if BITS_PER_LONG==32
2525 write_seqlock(&sk
->sk_stamp_seq
);
2527 write_sequnlock(&sk
->sk_stamp_seq
);
2529 WRITE_ONCE(sk
->sk_stamp
, kt
);
2533 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
2534 struct sk_buff
*skb
);
2535 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
2536 struct sk_buff
*skb
);
2539 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
2541 ktime_t kt
= skb
->tstamp
;
2542 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
2545 * generate control messages if
2546 * - receive time stamping in software requested
2547 * - software time stamp available and wanted
2548 * - hardware time stamps available and wanted
2550 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
2551 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RX_SOFTWARE
) ||
2552 (kt
&& sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
) ||
2553 (hwtstamps
->hwtstamp
&&
2554 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
)))
2555 __sock_recv_timestamp(msg
, sk
, skb
);
2557 sock_write_timestamp(sk
, kt
);
2559 if (sock_flag(sk
, SOCK_WIFI_STATUS
) && skb
->wifi_acked_valid
)
2560 __sock_recv_wifi_status(msg
, sk
, skb
);
2563 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2564 struct sk_buff
*skb
);
2566 #define SK_DEFAULT_STAMP (-1L * NSEC_PER_SEC)
2567 static inline void sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2568 struct sk_buff
*skb
)
2570 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2571 (1UL << SOCK_RCVTSTAMP))
2572 #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2573 SOF_TIMESTAMPING_RAW_HARDWARE)
2575 if (sk
->sk_flags
& FLAGS_TS_OR_DROPS
|| sk
->sk_tsflags
& TSFLAGS_ANY
)
2576 __sock_recv_ts_and_drops(msg
, sk
, skb
);
2577 else if (unlikely(sock_flag(sk
, SOCK_TIMESTAMP
)))
2578 sock_write_timestamp(sk
, skb
->tstamp
);
2579 else if (unlikely(sk
->sk_stamp
== SK_DEFAULT_STAMP
))
2580 sock_write_timestamp(sk
, 0);
2583 void __sock_tx_timestamp(__u16 tsflags
, __u8
*tx_flags
);
2586 * _sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2587 * @sk: socket sending this packet
2588 * @tsflags: timestamping flags to use
2589 * @tx_flags: completed with instructions for time stamping
2590 * @tskey: filled in with next sk_tskey (not for TCP, which uses seqno)
2592 * Note: callers should take care of initial ``*tx_flags`` value (usually 0)
2594 static inline void _sock_tx_timestamp(struct sock
*sk
, __u16 tsflags
,
2595 __u8
*tx_flags
, __u32
*tskey
)
2597 if (unlikely(tsflags
)) {
2598 __sock_tx_timestamp(tsflags
, tx_flags
);
2599 if (tsflags
& SOF_TIMESTAMPING_OPT_ID
&& tskey
&&
2600 tsflags
& SOF_TIMESTAMPING_TX_RECORD_MASK
)
2601 *tskey
= atomic_inc_return(&sk
->sk_tskey
) - 1;
2603 if (unlikely(sock_flag(sk
, SOCK_WIFI_STATUS
)))
2604 *tx_flags
|= SKBTX_WIFI_STATUS
;
2607 static inline void sock_tx_timestamp(struct sock
*sk
, __u16 tsflags
,
2610 _sock_tx_timestamp(sk
, tsflags
, tx_flags
, NULL
);
2613 static inline void skb_setup_tx_timestamp(struct sk_buff
*skb
, __u16 tsflags
)
2615 _sock_tx_timestamp(skb
->sk
, tsflags
, &skb_shinfo(skb
)->tx_flags
,
2616 &skb_shinfo(skb
)->tskey
);
2619 DECLARE_STATIC_KEY_FALSE(tcp_rx_skb_cache_key
);
2621 * sk_eat_skb - Release a skb if it is no longer needed
2622 * @sk: socket to eat this skb from
2623 * @skb: socket buffer to eat
2625 * This routine must be called with interrupts disabled or with the socket
2626 * locked so that the sk_buff queue operation is ok.
2628 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
)
2630 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2631 if (static_branch_unlikely(&tcp_rx_skb_cache_key
) &&
2632 !sk
->sk_rx_skb_cache
) {
2633 sk
->sk_rx_skb_cache
= skb
;
2641 struct net
*sock_net(const struct sock
*sk
)
2643 return read_pnet(&sk
->sk_net
);
2647 void sock_net_set(struct sock
*sk
, struct net
*net
)
2649 write_pnet(&sk
->sk_net
, net
);
2653 skb_sk_is_prefetched(struct sk_buff
*skb
)
2656 return skb
->destructor
== sock_pfree
;
2659 #endif /* CONFIG_INET */
2662 /* This helper checks if a socket is a full socket,
2663 * ie _not_ a timewait or request socket.
2665 static inline bool sk_fullsock(const struct sock
*sk
)
2667 return (1 << sk
->sk_state
) & ~(TCPF_TIME_WAIT
| TCPF_NEW_SYN_RECV
);
2671 sk_is_refcounted(struct sock
*sk
)
2673 /* Only full sockets have sk->sk_flags. */
2674 return !sk_fullsock(sk
) || !sock_flag(sk
, SOCK_RCU_FREE
);
2678 * skb_steal_sock - steal a socket from an sk_buff
2679 * @skb: sk_buff to steal the socket from
2680 * @refcounted: is set to true if the socket is reference-counted
2682 static inline struct sock
*
2683 skb_steal_sock(struct sk_buff
*skb
, bool *refcounted
)
2686 struct sock
*sk
= skb
->sk
;
2689 if (skb_sk_is_prefetched(skb
))
2690 *refcounted
= sk_is_refcounted(sk
);
2691 skb
->destructor
= NULL
;
2695 *refcounted
= false;
2699 /* Checks if this SKB belongs to an HW offloaded socket
2700 * and whether any SW fallbacks are required based on dev.
2701 * Check decrypted mark in case skb_orphan() cleared socket.
2703 static inline struct sk_buff
*sk_validate_xmit_skb(struct sk_buff
*skb
,
2704 struct net_device
*dev
)
2706 #ifdef CONFIG_SOCK_VALIDATE_XMIT
2707 struct sock
*sk
= skb
->sk
;
2709 if (sk
&& sk_fullsock(sk
) && sk
->sk_validate_xmit_skb
) {
2710 skb
= sk
->sk_validate_xmit_skb(sk
, dev
, skb
);
2711 #ifdef CONFIG_TLS_DEVICE
2712 } else if (unlikely(skb
->decrypted
)) {
2713 pr_warn_ratelimited("unencrypted skb with no associated socket - dropping\n");
2723 /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2724 * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2726 static inline bool sk_listener(const struct sock
*sk
)
2728 return (1 << sk
->sk_state
) & (TCPF_LISTEN
| TCPF_NEW_SYN_RECV
);
2731 void sock_enable_timestamp(struct sock
*sk
, enum sock_flags flag
);
2732 int sock_recv_errqueue(struct sock
*sk
, struct msghdr
*msg
, int len
, int level
,
2735 bool sk_ns_capable(const struct sock
*sk
,
2736 struct user_namespace
*user_ns
, int cap
);
2737 bool sk_capable(const struct sock
*sk
, int cap
);
2738 bool sk_net_capable(const struct sock
*sk
, int cap
);
2740 void sk_get_meminfo(const struct sock
*sk
, u32
*meminfo
);
2742 /* Take into consideration the size of the struct sk_buff overhead in the
2743 * determination of these values, since that is non-constant across
2744 * platforms. This makes socket queueing behavior and performance
2745 * not depend upon such differences.
2747 #define _SK_MEM_PACKETS 256
2748 #define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
2749 #define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2750 #define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2752 extern __u32 sysctl_wmem_max
;
2753 extern __u32 sysctl_rmem_max
;
2755 extern int sysctl_tstamp_allow_data
;
2756 extern int sysctl_optmem_max
;
2758 extern __u32 sysctl_wmem_default
;
2759 extern __u32 sysctl_rmem_default
;
2761 #define SKB_FRAG_PAGE_ORDER get_order(32768)
2762 DECLARE_STATIC_KEY_FALSE(net_high_order_alloc_disable_key
);
2764 static inline int sk_get_wmem0(const struct sock
*sk
, const struct proto
*proto
)
2766 /* Does this proto have per netns sysctl_wmem ? */
2767 if (proto
->sysctl_wmem_offset
)
2768 return READ_ONCE(*(int *)((void *)sock_net(sk
) + proto
->sysctl_wmem_offset
));
2770 return READ_ONCE(*proto
->sysctl_wmem
);
2773 static inline int sk_get_rmem0(const struct sock
*sk
, const struct proto
*proto
)
2775 /* Does this proto have per netns sysctl_rmem ? */
2776 if (proto
->sysctl_rmem_offset
)
2777 return READ_ONCE(*(int *)((void *)sock_net(sk
) + proto
->sysctl_rmem_offset
));
2779 return READ_ONCE(*proto
->sysctl_rmem
);
2782 /* Default TCP Small queue budget is ~1 ms of data (1sec >> 10)
2783 * Some wifi drivers need to tweak it to get more chunks.
2784 * They can use this helper from their ndo_start_xmit()
2786 static inline void sk_pacing_shift_update(struct sock
*sk
, int val
)
2788 if (!sk
|| !sk_fullsock(sk
) || READ_ONCE(sk
->sk_pacing_shift
) == val
)
2790 WRITE_ONCE(sk
->sk_pacing_shift
, val
);
2793 /* if a socket is bound to a device, check that the given device
2794 * index is either the same or that the socket is bound to an L3
2795 * master device and the given device index is also enslaved to
2798 static inline bool sk_dev_equal_l3scope(struct sock
*sk
, int dif
)
2802 if (!sk
->sk_bound_dev_if
|| sk
->sk_bound_dev_if
== dif
)
2805 mdif
= l3mdev_master_ifindex_by_index(sock_net(sk
), dif
);
2806 if (mdif
&& mdif
== sk
->sk_bound_dev_if
)
2812 void sock_def_readable(struct sock
*sk
);
2814 int sock_bindtoindex(struct sock
*sk
, int ifindex
, bool lock_sk
);
2815 void sock_set_timestamp(struct sock
*sk
, int optname
, bool valbool
);
2816 int sock_set_timestamping(struct sock
*sk
, int optname
,
2817 struct so_timestamping timestamping
);
2819 void sock_enable_timestamps(struct sock
*sk
);
2820 void sock_no_linger(struct sock
*sk
);
2821 void sock_set_keepalive(struct sock
*sk
);
2822 void sock_set_priority(struct sock
*sk
, u32 priority
);
2823 void sock_set_rcvbuf(struct sock
*sk
, int val
);
2824 void sock_set_mark(struct sock
*sk
, u32 val
);
2825 void sock_set_reuseaddr(struct sock
*sk
);
2826 void sock_set_reuseport(struct sock
*sk
);
2827 void sock_set_sndtimeo(struct sock
*sk
, s64 secs
);
2829 int sock_bind_add(struct sock
*sk
, struct sockaddr
*addr
, int addr_len
);
2831 static inline bool sk_is_readable(struct sock
*sk
)
2833 if (sk
->sk_prot
->sock_is_readable
)
2834 return sk
->sk_prot
->sock_is_readable(sk
);
2837 #endif /* _SOCK_H */