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 /* 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.
165 __addrpair skc_addrpair
;
168 __be32 skc_rcv_saddr
;
172 unsigned int skc_hash
;
173 __u16 skc_u16hashes
[2];
175 /* skc_dport && skc_num must be grouped as well */
177 __portpair skc_portpair
;
184 unsigned short skc_family
;
185 volatile unsigned char skc_state
;
186 unsigned char skc_reuse
:4;
187 unsigned char skc_reuseport
:1;
188 unsigned char skc_ipv6only
:1;
189 unsigned char skc_net_refcnt
:1;
190 int skc_bound_dev_if
;
192 struct hlist_node skc_bind_node
;
193 struct hlist_node skc_portaddr_node
;
195 struct proto
*skc_prot
;
196 possible_net_t skc_net
;
198 #if IS_ENABLED(CONFIG_IPV6)
199 struct in6_addr skc_v6_daddr
;
200 struct in6_addr skc_v6_rcv_saddr
;
203 atomic64_t skc_cookie
;
205 /* following fields are padding to force
206 * offset(struct sock, sk_refcnt) == 128 on 64bit arches
207 * assuming IPV6 is enabled. We use this padding differently
208 * for different kind of 'sockets'
211 unsigned long skc_flags
;
212 struct sock
*skc_listener
; /* request_sock */
213 struct inet_timewait_death_row
*skc_tw_dr
; /* inet_timewait_sock */
216 * fields between dontcopy_begin/dontcopy_end
217 * are not copied in sock_copy()
220 int skc_dontcopy_begin
[0];
223 struct hlist_node skc_node
;
224 struct hlist_nulls_node skc_nulls_node
;
226 unsigned short skc_tx_queue_mapping
;
227 #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
228 unsigned short skc_rx_queue_mapping
;
231 int skc_incoming_cpu
;
233 u32 skc_tw_rcv_nxt
; /* struct tcp_timewait_sock */
236 refcount_t skc_refcnt
;
238 int skc_dontcopy_end
[0];
241 u32 skc_window_clamp
;
242 u32 skc_tw_snd_nxt
; /* struct tcp_timewait_sock */
247 struct bpf_local_storage
;
251 * struct sock - network layer representation of sockets
252 * @__sk_common: shared layout with inet_timewait_sock
253 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
254 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
255 * @sk_lock: synchronizer
256 * @sk_kern_sock: True if sock is using kernel lock classes
257 * @sk_rcvbuf: size of receive buffer in bytes
258 * @sk_wq: sock wait queue and async head
259 * @sk_rx_dst: receive input route used by early demux
260 * @sk_rx_dst_ifindex: ifindex for @sk_rx_dst
261 * @sk_rx_dst_cookie: cookie for @sk_rx_dst
262 * @sk_dst_cache: destination cache
263 * @sk_dst_pending_confirm: need to confirm neighbour
264 * @sk_policy: flow policy
265 * @sk_receive_queue: incoming packets
266 * @sk_wmem_alloc: transmit queue bytes committed
267 * @sk_tsq_flags: TCP Small Queues flags
268 * @sk_write_queue: Packet sending queue
269 * @sk_omem_alloc: "o" is "option" or "other"
270 * @sk_wmem_queued: persistent queue size
271 * @sk_forward_alloc: space allocated forward
272 * @sk_reserved_mem: space reserved and non-reclaimable for the socket
273 * @sk_napi_id: id of the last napi context to receive data for sk
274 * @sk_ll_usec: usecs to busypoll when there is no data
275 * @sk_allocation: allocation mode
276 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
277 * @sk_pacing_status: Pacing status (requested, handled by sch_fq)
278 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
279 * @sk_sndbuf: size of send buffer in bytes
280 * @__sk_flags_offset: empty field used to determine location of bitfield
281 * @sk_padding: unused element for alignment
282 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
283 * @sk_no_check_rx: allow zero checksum in RX packets
284 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
285 * @sk_gso_disabled: if set, NETIF_F_GSO_MASK is forbidden.
286 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
287 * @sk_gso_max_size: Maximum GSO segment size to build
288 * @sk_gso_max_segs: Maximum number of GSO segments
289 * @sk_pacing_shift: scaling factor for TCP Small Queues
290 * @sk_lingertime: %SO_LINGER l_linger setting
291 * @sk_backlog: always used with the per-socket spinlock held
292 * @sk_callback_lock: used with the callbacks in the end of this struct
293 * @sk_error_queue: rarely used
294 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
295 * IPV6_ADDRFORM for instance)
296 * @sk_err: last error
297 * @sk_err_soft: errors that don't cause failure but are the cause of a
298 * persistent failure not just 'timed out'
299 * @sk_drops: raw/udp drops counter
300 * @sk_ack_backlog: current listen backlog
301 * @sk_max_ack_backlog: listen backlog set in listen()
302 * @sk_uid: user id of owner
303 * @sk_prefer_busy_poll: prefer busypolling over softirq processing
304 * @sk_busy_poll_budget: napi processing budget when busypolling
305 * @sk_priority: %SO_PRIORITY setting
306 * @sk_type: socket type (%SOCK_STREAM, etc)
307 * @sk_protocol: which protocol this socket belongs in this network family
308 * @sk_peer_lock: lock protecting @sk_peer_pid and @sk_peer_cred
309 * @sk_peer_pid: &struct pid for this socket's peer
310 * @sk_peer_cred: %SO_PEERCRED setting
311 * @sk_rcvlowat: %SO_RCVLOWAT setting
312 * @sk_rcvtimeo: %SO_RCVTIMEO setting
313 * @sk_sndtimeo: %SO_SNDTIMEO setting
314 * @sk_txhash: computed flow hash for use on transmit
315 * @sk_txrehash: enable TX hash rethink
316 * @sk_filter: socket filtering instructions
317 * @sk_timer: sock cleanup timer
318 * @sk_stamp: time stamp of last packet received
319 * @sk_stamp_seq: lock for accessing sk_stamp on 32 bit architectures only
320 * @sk_tsflags: SO_TIMESTAMPING flags
321 * @sk_use_task_frag: allow sk_page_frag() to use current->task_frag.
322 * Sockets that can be used under memory reclaim should
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. Write-protected by @sk_callback_lock.
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_security: used by security modules
335 * @sk_mark: generic packet mark
336 * @sk_cgrp_data: cgroup data for this cgroup
337 * @sk_memcg: this socket's memory cgroup association
338 * @sk_write_pending: a write to stream socket waits to start
339 * @sk_wait_pending: number of threads blocked on this socket
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
354 * @ns_tracker: tracker for netns reference
355 * @sk_bind2_node: bind node in the bhash2 table
359 * Now struct inet_timewait_sock also uses sock_common, so please just
360 * don't add nothing before this first member (__sk_common) --acme
362 struct sock_common __sk_common
;
363 #define sk_node __sk_common.skc_node
364 #define sk_nulls_node __sk_common.skc_nulls_node
365 #define sk_refcnt __sk_common.skc_refcnt
366 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
367 #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
368 #define sk_rx_queue_mapping __sk_common.skc_rx_queue_mapping
371 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
372 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
373 #define sk_hash __sk_common.skc_hash
374 #define sk_portpair __sk_common.skc_portpair
375 #define sk_num __sk_common.skc_num
376 #define sk_dport __sk_common.skc_dport
377 #define sk_addrpair __sk_common.skc_addrpair
378 #define sk_daddr __sk_common.skc_daddr
379 #define sk_rcv_saddr __sk_common.skc_rcv_saddr
380 #define sk_family __sk_common.skc_family
381 #define sk_state __sk_common.skc_state
382 #define sk_reuse __sk_common.skc_reuse
383 #define sk_reuseport __sk_common.skc_reuseport
384 #define sk_ipv6only __sk_common.skc_ipv6only
385 #define sk_net_refcnt __sk_common.skc_net_refcnt
386 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
387 #define sk_bind_node __sk_common.skc_bind_node
388 #define sk_prot __sk_common.skc_prot
389 #define sk_net __sk_common.skc_net
390 #define sk_v6_daddr __sk_common.skc_v6_daddr
391 #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
392 #define sk_cookie __sk_common.skc_cookie
393 #define sk_incoming_cpu __sk_common.skc_incoming_cpu
394 #define sk_flags __sk_common.skc_flags
395 #define sk_rxhash __sk_common.skc_rxhash
397 /* early demux fields */
398 struct dst_entry __rcu
*sk_rx_dst
;
399 int sk_rx_dst_ifindex
;
400 u32 sk_rx_dst_cookie
;
402 socket_lock_t sk_lock
;
405 struct sk_buff_head sk_error_queue
;
406 struct sk_buff_head sk_receive_queue
;
408 * The backlog queue is special, it is always used with
409 * the per-socket spinlock held and requires low latency
410 * access. Therefore we special case it's implementation.
411 * Note : rmem_alloc is in this structure to fill a hole
412 * on 64bit arches, not because its logically part of
418 struct sk_buff
*head
;
419 struct sk_buff
*tail
;
422 #define sk_rmem_alloc sk_backlog.rmem_alloc
424 int sk_forward_alloc
;
426 #ifdef CONFIG_NET_RX_BUSY_POLL
427 unsigned int sk_ll_usec
;
428 /* ===== mostly read cache line ===== */
429 unsigned int sk_napi_id
;
434 struct sk_filter __rcu
*sk_filter
;
436 struct socket_wq __rcu
*sk_wq
;
438 struct socket_wq
*sk_wq_raw
;
442 struct xfrm_policy __rcu
*sk_policy
[2];
445 struct dst_entry __rcu
*sk_dst_cache
;
446 atomic_t sk_omem_alloc
;
449 /* ===== cache line for TX ===== */
451 refcount_t sk_wmem_alloc
;
452 unsigned long sk_tsq_flags
;
454 struct sk_buff
*sk_send_head
;
455 struct rb_root tcp_rtx_queue
;
457 struct sk_buff_head sk_write_queue
;
459 int sk_write_pending
;
460 __u32 sk_dst_pending_confirm
;
461 u32 sk_pacing_status
; /* see enum sk_pacing */
463 struct timer_list sk_timer
;
466 unsigned long sk_pacing_rate
; /* bytes per second */
467 unsigned long sk_max_pacing_rate
;
468 struct page_frag sk_frag
;
469 netdev_features_t sk_route_caps
;
471 unsigned int sk_gso_max_size
;
476 * Because of non atomicity rules, all
477 * changes are protected by socket lock.
479 u8 sk_gso_disabled
: 1,
488 unsigned long sk_lingertime
;
489 struct proto
*sk_prot_creator
;
490 rwlock_t sk_callback_lock
;
494 u32 sk_max_ack_backlog
;
497 #ifdef CONFIG_NET_RX_BUSY_POLL
498 u8 sk_prefer_busy_poll
;
499 u16 sk_busy_poll_budget
;
501 spinlock_t sk_peer_lock
;
503 struct pid
*sk_peer_pid
;
504 const struct cred
*sk_peer_cred
;
508 #if BITS_PER_LONG==32
509 seqlock_t sk_stamp_seq
;
517 u8 sk_txtime_deadline_mode
: 1,
518 sk_txtime_report_errors
: 1,
519 sk_txtime_unused
: 6;
520 bool sk_use_task_frag
;
522 struct socket
*sk_socket
;
524 #ifdef CONFIG_SECURITY
527 struct sock_cgroup_data sk_cgrp_data
;
528 struct mem_cgroup
*sk_memcg
;
529 void (*sk_state_change
)(struct sock
*sk
);
530 void (*sk_data_ready
)(struct sock
*sk
);
531 void (*sk_write_space
)(struct sock
*sk
);
532 void (*sk_error_report
)(struct sock
*sk
);
533 int (*sk_backlog_rcv
)(struct sock
*sk
,
534 struct sk_buff
*skb
);
535 #ifdef CONFIG_SOCK_VALIDATE_XMIT
536 struct sk_buff
* (*sk_validate_xmit_skb
)(struct sock
*sk
,
537 struct net_device
*dev
,
538 struct sk_buff
*skb
);
540 void (*sk_destruct
)(struct sock
*sk
);
541 struct sock_reuseport __rcu
*sk_reuseport_cb
;
542 #ifdef CONFIG_BPF_SYSCALL
543 struct bpf_local_storage __rcu
*sk_bpf_storage
;
545 struct rcu_head sk_rcu
;
546 netns_tracker ns_tracker
;
547 struct hlist_node sk_bind2_node
;
552 SK_PACING_NEEDED
= 1,
556 /* flag bits in sk_user_data
558 * - SK_USER_DATA_NOCOPY: Pointer stored in sk_user_data might
559 * not be suitable for copying when cloning the socket. For instance,
560 * it can point to a reference counted object. sk_user_data bottom
561 * bit is set if pointer must not be copied.
563 * - SK_USER_DATA_BPF: Mark whether sk_user_data field is
564 * managed/owned by a BPF reuseport array. This bit should be set
565 * when sk_user_data's sk is added to the bpf's reuseport_array.
567 * - SK_USER_DATA_PSOCK: Mark whether pointer stored in
568 * sk_user_data points to psock type. This bit should be set
569 * when sk_user_data is assigned to a psock object.
571 #define SK_USER_DATA_NOCOPY 1UL
572 #define SK_USER_DATA_BPF 2UL
573 #define SK_USER_DATA_PSOCK 4UL
574 #define SK_USER_DATA_PTRMASK ~(SK_USER_DATA_NOCOPY | SK_USER_DATA_BPF |\
578 * sk_user_data_is_nocopy - Test if sk_user_data pointer must not be copied
581 static inline bool sk_user_data_is_nocopy(const struct sock
*sk
)
583 return ((uintptr_t)sk
->sk_user_data
& SK_USER_DATA_NOCOPY
);
586 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
589 * __locked_read_sk_user_data_with_flags - return the pointer
590 * only if argument flags all has been set in sk_user_data. Otherwise
596 * The caller must be holding sk->sk_callback_lock.
599 __locked_read_sk_user_data_with_flags(const struct sock
*sk
,
602 uintptr_t sk_user_data
=
603 (uintptr_t)rcu_dereference_check(__sk_user_data(sk
),
604 lockdep_is_held(&sk
->sk_callback_lock
));
606 WARN_ON_ONCE(flags
& SK_USER_DATA_PTRMASK
);
608 if ((sk_user_data
& flags
) == flags
)
609 return (void *)(sk_user_data
& SK_USER_DATA_PTRMASK
);
614 * __rcu_dereference_sk_user_data_with_flags - return the pointer
615 * only if argument flags all has been set in sk_user_data. Otherwise
622 __rcu_dereference_sk_user_data_with_flags(const struct sock
*sk
,
625 uintptr_t sk_user_data
= (uintptr_t)rcu_dereference(__sk_user_data(sk
));
627 WARN_ON_ONCE(flags
& SK_USER_DATA_PTRMASK
);
629 if ((sk_user_data
& flags
) == flags
)
630 return (void *)(sk_user_data
& SK_USER_DATA_PTRMASK
);
634 #define rcu_dereference_sk_user_data(sk) \
635 __rcu_dereference_sk_user_data_with_flags(sk, 0)
636 #define __rcu_assign_sk_user_data_with_flags(sk, ptr, flags) \
638 uintptr_t __tmp1 = (uintptr_t)(ptr), \
639 __tmp2 = (uintptr_t)(flags); \
640 WARN_ON_ONCE(__tmp1 & ~SK_USER_DATA_PTRMASK); \
641 WARN_ON_ONCE(__tmp2 & SK_USER_DATA_PTRMASK); \
642 rcu_assign_pointer(__sk_user_data((sk)), \
645 #define rcu_assign_sk_user_data(sk, ptr) \
646 __rcu_assign_sk_user_data_with_flags(sk, ptr, 0)
649 struct net
*sock_net(const struct sock
*sk
)
651 return read_pnet(&sk
->sk_net
);
655 void sock_net_set(struct sock
*sk
, struct net
*net
)
657 write_pnet(&sk
->sk_net
, net
);
661 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
662 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
663 * on a socket means that the socket will reuse everybody else's port
664 * without looking at the other's sk_reuse value.
667 #define SK_NO_REUSE 0
668 #define SK_CAN_REUSE 1
669 #define SK_FORCE_REUSE 2
671 int sk_set_peek_off(struct sock
*sk
, int val
);
673 static inline int sk_peek_offset(const struct sock
*sk
, int flags
)
675 if (unlikely(flags
& MSG_PEEK
)) {
676 return READ_ONCE(sk
->sk_peek_off
);
682 static inline void sk_peek_offset_bwd(struct sock
*sk
, int val
)
684 s32 off
= READ_ONCE(sk
->sk_peek_off
);
686 if (unlikely(off
>= 0)) {
687 off
= max_t(s32
, off
- val
, 0);
688 WRITE_ONCE(sk
->sk_peek_off
, off
);
692 static inline void sk_peek_offset_fwd(struct sock
*sk
, int val
)
694 sk_peek_offset_bwd(sk
, -val
);
698 * Hashed lists helper routines
700 static inline struct sock
*sk_entry(const struct hlist_node
*node
)
702 return hlist_entry(node
, struct sock
, sk_node
);
705 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
707 return hlist_entry(head
->first
, struct sock
, sk_node
);
710 static inline struct sock
*sk_head(const struct hlist_head
*head
)
712 return hlist_empty(head
) ? NULL
: __sk_head(head
);
715 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
717 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
720 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
722 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
725 static inline struct sock
*sk_next(const struct sock
*sk
)
727 return hlist_entry_safe(sk
->sk_node
.next
, struct sock
, sk_node
);
730 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
732 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
733 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
734 struct sock
, sk_nulls_node
) :
738 static inline bool sk_unhashed(const struct sock
*sk
)
740 return hlist_unhashed(&sk
->sk_node
);
743 static inline bool sk_hashed(const struct sock
*sk
)
745 return !sk_unhashed(sk
);
748 static inline void sk_node_init(struct hlist_node
*node
)
753 static inline void __sk_del_node(struct sock
*sk
)
755 __hlist_del(&sk
->sk_node
);
758 /* NB: equivalent to hlist_del_init_rcu */
759 static inline bool __sk_del_node_init(struct sock
*sk
)
763 sk_node_init(&sk
->sk_node
);
769 /* Grab socket reference count. This operation is valid only
770 when sk is ALREADY grabbed f.e. it is found in hash table
771 or a list and the lookup is made under lock preventing hash table
775 static __always_inline
void sock_hold(struct sock
*sk
)
777 refcount_inc(&sk
->sk_refcnt
);
780 /* Ungrab socket in the context, which assumes that socket refcnt
781 cannot hit zero, f.e. it is true in context of any socketcall.
783 static __always_inline
void __sock_put(struct sock
*sk
)
785 refcount_dec(&sk
->sk_refcnt
);
788 static inline bool sk_del_node_init(struct sock
*sk
)
790 bool rc
= __sk_del_node_init(sk
);
793 /* paranoid for a while -acme */
794 WARN_ON(refcount_read(&sk
->sk_refcnt
) == 1);
799 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
801 static inline bool __sk_nulls_del_node_init_rcu(struct sock
*sk
)
804 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
810 static inline bool sk_nulls_del_node_init_rcu(struct sock
*sk
)
812 bool rc
= __sk_nulls_del_node_init_rcu(sk
);
815 /* paranoid for a while -acme */
816 WARN_ON(refcount_read(&sk
->sk_refcnt
) == 1);
822 static inline void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
824 hlist_add_head(&sk
->sk_node
, list
);
827 static inline void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
830 __sk_add_node(sk
, list
);
833 static inline void sk_add_node_rcu(struct sock
*sk
, struct hlist_head
*list
)
836 if (IS_ENABLED(CONFIG_IPV6
) && sk
->sk_reuseport
&&
837 sk
->sk_family
== AF_INET6
)
838 hlist_add_tail_rcu(&sk
->sk_node
, list
);
840 hlist_add_head_rcu(&sk
->sk_node
, list
);
843 static inline void sk_add_node_tail_rcu(struct sock
*sk
, struct hlist_head
*list
)
846 hlist_add_tail_rcu(&sk
->sk_node
, list
);
849 static inline void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
851 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
854 static inline void __sk_nulls_add_node_tail_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
856 hlist_nulls_add_tail_rcu(&sk
->sk_nulls_node
, list
);
859 static inline void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
862 __sk_nulls_add_node_rcu(sk
, list
);
865 static inline void __sk_del_bind_node(struct sock
*sk
)
867 __hlist_del(&sk
->sk_bind_node
);
870 static inline void sk_add_bind_node(struct sock
*sk
,
871 struct hlist_head
*list
)
873 hlist_add_head(&sk
->sk_bind_node
, list
);
876 static inline void __sk_del_bind2_node(struct sock
*sk
)
878 __hlist_del(&sk
->sk_bind2_node
);
881 static inline void sk_add_bind2_node(struct sock
*sk
, struct hlist_head
*list
)
883 hlist_add_head(&sk
->sk_bind2_node
, list
);
886 #define sk_for_each(__sk, list) \
887 hlist_for_each_entry(__sk, list, sk_node)
888 #define sk_for_each_rcu(__sk, list) \
889 hlist_for_each_entry_rcu(__sk, list, sk_node)
890 #define sk_nulls_for_each(__sk, node, list) \
891 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
892 #define sk_nulls_for_each_rcu(__sk, node, list) \
893 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
894 #define sk_for_each_from(__sk) \
895 hlist_for_each_entry_from(__sk, sk_node)
896 #define sk_nulls_for_each_from(__sk, node) \
897 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
898 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
899 #define sk_for_each_safe(__sk, tmp, list) \
900 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
901 #define sk_for_each_bound(__sk, list) \
902 hlist_for_each_entry(__sk, list, sk_bind_node)
903 #define sk_for_each_bound_bhash2(__sk, list) \
904 hlist_for_each_entry(__sk, list, sk_bind2_node)
907 * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset
908 * @tpos: the type * to use as a loop cursor.
909 * @pos: the &struct hlist_node to use as a loop cursor.
910 * @head: the head for your list.
911 * @offset: offset of hlist_node within the struct.
914 #define sk_for_each_entry_offset_rcu(tpos, pos, head, offset) \
915 for (pos = rcu_dereference(hlist_first_rcu(head)); \
917 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
918 pos = rcu_dereference(hlist_next_rcu(pos)))
920 static inline struct user_namespace
*sk_user_ns(const struct sock
*sk
)
922 /* Careful only use this in a context where these parameters
923 * can not change and must all be valid, such as recvmsg from
926 return sk
->sk_socket
->file
->f_cred
->user_ns
;
940 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
941 SOCK_DBG
, /* %SO_DEBUG setting */
942 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
943 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
944 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
945 SOCK_MEMALLOC
, /* VM depends on this socket for swapping */
946 SOCK_TIMESTAMPING_RX_SOFTWARE
, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
947 SOCK_FASYNC
, /* fasync() active */
949 SOCK_ZEROCOPY
, /* buffers from userspace */
950 SOCK_WIFI_STATUS
, /* push wifi status to userspace */
951 SOCK_NOFCS
, /* Tell NIC not to do the Ethernet FCS.
952 * Will use last 4 bytes of packet sent from
953 * user-space instead.
955 SOCK_FILTER_LOCKED
, /* Filter cannot be changed anymore */
956 SOCK_SELECT_ERR_QUEUE
, /* Wake select on error queue */
957 SOCK_RCU_FREE
, /* wait rcu grace period in sk_destruct() */
959 SOCK_XDP
, /* XDP is attached */
960 SOCK_TSTAMP_NEW
, /* Indicates 64 bit timestamps always */
961 SOCK_RCVMARK
, /* Receive SO_MARK ancillary data with packet */
964 #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
966 static inline void sock_copy_flags(struct sock
*nsk
, const struct sock
*osk
)
968 nsk
->sk_flags
= osk
->sk_flags
;
971 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
973 __set_bit(flag
, &sk
->sk_flags
);
976 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
978 __clear_bit(flag
, &sk
->sk_flags
);
981 static inline void sock_valbool_flag(struct sock
*sk
, enum sock_flags bit
,
985 sock_set_flag(sk
, bit
);
987 sock_reset_flag(sk
, bit
);
990 static inline bool sock_flag(const struct sock
*sk
, enum sock_flags flag
)
992 return test_bit(flag
, &sk
->sk_flags
);
996 DECLARE_STATIC_KEY_FALSE(memalloc_socks_key
);
997 static inline int sk_memalloc_socks(void)
999 return static_branch_unlikely(&memalloc_socks_key
);
1002 void __receive_sock(struct file
*file
);
1005 static inline int sk_memalloc_socks(void)
1010 static inline void __receive_sock(struct file
*file
)
1014 static inline gfp_t
sk_gfp_mask(const struct sock
*sk
, gfp_t gfp_mask
)
1016 return gfp_mask
| (sk
->sk_allocation
& __GFP_MEMALLOC
);
1019 static inline void sk_acceptq_removed(struct sock
*sk
)
1021 WRITE_ONCE(sk
->sk_ack_backlog
, sk
->sk_ack_backlog
- 1);
1024 static inline void sk_acceptq_added(struct sock
*sk
)
1026 WRITE_ONCE(sk
->sk_ack_backlog
, sk
->sk_ack_backlog
+ 1);
1029 /* Note: If you think the test should be:
1030 * return READ_ONCE(sk->sk_ack_backlog) >= READ_ONCE(sk->sk_max_ack_backlog);
1031 * Then please take a look at commit 64a146513f8f ("[NET]: Revert incorrect accept queue backlog changes.")
1033 static inline bool sk_acceptq_is_full(const struct sock
*sk
)
1035 return READ_ONCE(sk
->sk_ack_backlog
) > READ_ONCE(sk
->sk_max_ack_backlog
);
1039 * Compute minimal free write space needed to queue new packets.
1041 static inline int sk_stream_min_wspace(const struct sock
*sk
)
1043 return READ_ONCE(sk
->sk_wmem_queued
) >> 1;
1046 static inline int sk_stream_wspace(const struct sock
*sk
)
1048 return READ_ONCE(sk
->sk_sndbuf
) - READ_ONCE(sk
->sk_wmem_queued
);
1051 static inline void sk_wmem_queued_add(struct sock
*sk
, int val
)
1053 WRITE_ONCE(sk
->sk_wmem_queued
, sk
->sk_wmem_queued
+ val
);
1056 static inline void sk_forward_alloc_add(struct sock
*sk
, int val
)
1058 /* Paired with lockless reads of sk->sk_forward_alloc */
1059 WRITE_ONCE(sk
->sk_forward_alloc
, sk
->sk_forward_alloc
+ val
);
1062 void sk_stream_write_space(struct sock
*sk
);
1064 /* OOB backlog add */
1065 static inline void __sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
1067 /* dont let skb dst not refcounted, we are going to leave rcu lock */
1070 if (!sk
->sk_backlog
.tail
)
1071 WRITE_ONCE(sk
->sk_backlog
.head
, skb
);
1073 sk
->sk_backlog
.tail
->next
= skb
;
1075 WRITE_ONCE(sk
->sk_backlog
.tail
, skb
);
1080 * Take into account size of receive queue and backlog queue
1081 * Do not take into account this skb truesize,
1082 * to allow even a single big packet to come.
1084 static inline bool sk_rcvqueues_full(const struct sock
*sk
, unsigned int limit
)
1086 unsigned int qsize
= sk
->sk_backlog
.len
+ atomic_read(&sk
->sk_rmem_alloc
);
1088 return qsize
> limit
;
1091 /* The per-socket spinlock must be held here. */
1092 static inline __must_check
int sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
,
1095 if (sk_rcvqueues_full(sk
, limit
))
1099 * If the skb was allocated from pfmemalloc reserves, only
1100 * allow SOCK_MEMALLOC sockets to use it as this socket is
1101 * helping free memory
1103 if (skb_pfmemalloc(skb
) && !sock_flag(sk
, SOCK_MEMALLOC
))
1106 __sk_add_backlog(sk
, skb
);
1107 sk
->sk_backlog
.len
+= skb
->truesize
;
1111 int __sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
1113 INDIRECT_CALLABLE_DECLARE(int tcp_v4_do_rcv(struct sock
*sk
, struct sk_buff
*skb
));
1114 INDIRECT_CALLABLE_DECLARE(int tcp_v6_do_rcv(struct sock
*sk
, struct sk_buff
*skb
));
1116 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1118 if (sk_memalloc_socks() && skb_pfmemalloc(skb
))
1119 return __sk_backlog_rcv(sk
, skb
);
1121 return INDIRECT_CALL_INET(sk
->sk_backlog_rcv
,
1127 static inline void sk_incoming_cpu_update(struct sock
*sk
)
1129 int cpu
= raw_smp_processor_id();
1131 if (unlikely(READ_ONCE(sk
->sk_incoming_cpu
) != cpu
))
1132 WRITE_ONCE(sk
->sk_incoming_cpu
, cpu
);
1135 static inline void sock_rps_record_flow_hash(__u32 hash
)
1138 struct rps_sock_flow_table
*sock_flow_table
;
1141 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
1142 rps_record_sock_flow(sock_flow_table
, hash
);
1147 static inline void sock_rps_record_flow(const struct sock
*sk
)
1150 if (static_branch_unlikely(&rfs_needed
)) {
1151 /* Reading sk->sk_rxhash might incur an expensive cache line
1154 * TCP_ESTABLISHED does cover almost all states where RFS
1155 * might be useful, and is cheaper [1] than testing :
1156 * IPv4: inet_sk(sk)->inet_daddr
1157 * IPv6: ipv6_addr_any(&sk->sk_v6_daddr)
1158 * OR an additional socket flag
1159 * [1] : sk_state and sk_prot are in the same cache line.
1161 if (sk
->sk_state
== TCP_ESTABLISHED
) {
1162 /* This READ_ONCE() is paired with the WRITE_ONCE()
1163 * from sock_rps_save_rxhash() and sock_rps_reset_rxhash().
1165 sock_rps_record_flow_hash(READ_ONCE(sk
->sk_rxhash
));
1171 static inline void sock_rps_save_rxhash(struct sock
*sk
,
1172 const struct sk_buff
*skb
)
1175 /* The following WRITE_ONCE() is paired with the READ_ONCE()
1176 * here, and another one in sock_rps_record_flow().
1178 if (unlikely(READ_ONCE(sk
->sk_rxhash
) != skb
->hash
))
1179 WRITE_ONCE(sk
->sk_rxhash
, skb
->hash
);
1183 static inline void sock_rps_reset_rxhash(struct sock
*sk
)
1186 /* Paired with READ_ONCE() in sock_rps_record_flow() */
1187 WRITE_ONCE(sk
->sk_rxhash
, 0);
1191 #define sk_wait_event(__sk, __timeo, __condition, __wait) \
1193 __sk->sk_wait_pending++; \
1194 release_sock(__sk); \
1195 __rc = __condition; \
1197 *(__timeo) = wait_woken(__wait, \
1198 TASK_INTERRUPTIBLE, \
1201 sched_annotate_sleep(); \
1203 __sk->sk_wait_pending--; \
1204 __rc = __condition; \
1208 int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
1209 int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
1210 void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
1211 int sk_stream_error(struct sock
*sk
, int flags
, int err
);
1212 void sk_stream_kill_queues(struct sock
*sk
);
1213 void sk_set_memalloc(struct sock
*sk
);
1214 void sk_clear_memalloc(struct sock
*sk
);
1216 void __sk_flush_backlog(struct sock
*sk
);
1218 static inline bool sk_flush_backlog(struct sock
*sk
)
1220 if (unlikely(READ_ONCE(sk
->sk_backlog
.tail
))) {
1221 __sk_flush_backlog(sk
);
1227 int sk_wait_data(struct sock
*sk
, long *timeo
, const struct sk_buff
*skb
);
1229 struct request_sock_ops
;
1230 struct timewait_sock_ops
;
1231 struct inet_hashinfo
;
1232 struct raw_hashinfo
;
1233 struct smc_hashinfo
;
1238 * caches using SLAB_TYPESAFE_BY_RCU should let .next pointer from nulls nodes
1239 * un-modified. Special care is taken when initializing object to zero.
1241 static inline void sk_prot_clear_nulls(struct sock
*sk
, int size
)
1243 if (offsetof(struct sock
, sk_node
.next
) != 0)
1244 memset(sk
, 0, offsetof(struct sock
, sk_node
.next
));
1245 memset(&sk
->sk_node
.pprev
, 0,
1246 size
- offsetof(struct sock
, sk_node
.pprev
));
1249 /* Networking protocol blocks we attach to sockets.
1250 * socket layer -> transport layer interface
1253 void (*close
)(struct sock
*sk
,
1255 int (*pre_connect
)(struct sock
*sk
,
1256 struct sockaddr
*uaddr
,
1258 int (*connect
)(struct sock
*sk
,
1259 struct sockaddr
*uaddr
,
1261 int (*disconnect
)(struct sock
*sk
, int flags
);
1263 struct sock
* (*accept
)(struct sock
*sk
, int flags
, int *err
,
1266 int (*ioctl
)(struct sock
*sk
, int cmd
,
1268 int (*init
)(struct sock
*sk
);
1269 void (*destroy
)(struct sock
*sk
);
1270 void (*shutdown
)(struct sock
*sk
, int how
);
1271 int (*setsockopt
)(struct sock
*sk
, int level
,
1272 int optname
, sockptr_t optval
,
1273 unsigned int optlen
);
1274 int (*getsockopt
)(struct sock
*sk
, int level
,
1275 int optname
, char __user
*optval
,
1276 int __user
*option
);
1277 void (*keepalive
)(struct sock
*sk
, int valbool
);
1278 #ifdef CONFIG_COMPAT
1279 int (*compat_ioctl
)(struct sock
*sk
,
1280 unsigned int cmd
, unsigned long arg
);
1282 int (*sendmsg
)(struct sock
*sk
, struct msghdr
*msg
,
1284 int (*recvmsg
)(struct sock
*sk
, struct msghdr
*msg
,
1285 size_t len
, int flags
, int *addr_len
);
1286 void (*splice_eof
)(struct socket
*sock
);
1287 int (*bind
)(struct sock
*sk
,
1288 struct sockaddr
*addr
, int addr_len
);
1289 int (*bind_add
)(struct sock
*sk
,
1290 struct sockaddr
*addr
, int addr_len
);
1292 int (*backlog_rcv
) (struct sock
*sk
,
1293 struct sk_buff
*skb
);
1294 bool (*bpf_bypass_getsockopt
)(int level
,
1297 void (*release_cb
)(struct sock
*sk
);
1299 /* Keeping track of sk's, looking them up, and port selection methods. */
1300 int (*hash
)(struct sock
*sk
);
1301 void (*unhash
)(struct sock
*sk
);
1302 void (*rehash
)(struct sock
*sk
);
1303 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
1304 void (*put_port
)(struct sock
*sk
);
1305 #ifdef CONFIG_BPF_SYSCALL
1306 int (*psock_update_sk_prot
)(struct sock
*sk
,
1307 struct sk_psock
*psock
,
1311 /* Keeping track of sockets in use */
1312 #ifdef CONFIG_PROC_FS
1313 unsigned int inuse_idx
;
1316 #if IS_ENABLED(CONFIG_MPTCP)
1317 int (*forward_alloc_get
)(const struct sock
*sk
);
1320 bool (*stream_memory_free
)(const struct sock
*sk
, int wake
);
1321 bool (*sock_is_readable
)(struct sock
*sk
);
1322 /* Memory pressure */
1323 void (*enter_memory_pressure
)(struct sock
*sk
);
1324 void (*leave_memory_pressure
)(struct sock
*sk
);
1325 atomic_long_t
*memory_allocated
; /* Current allocated memory. */
1326 int __percpu
*per_cpu_fw_alloc
;
1327 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
1330 * Pressure flag: try to collapse.
1331 * Technical note: it is used by multiple contexts non atomically.
1332 * Make sure to use READ_ONCE()/WRITE_ONCE() for all reads/writes.
1333 * All the __sk_mem_schedule() is of this nature: accounting
1334 * is strict, actions are advisory and have some latency.
1336 unsigned long *memory_pressure
;
1341 u32 sysctl_wmem_offset
;
1342 u32 sysctl_rmem_offset
;
1347 struct kmem_cache
*slab
;
1348 unsigned int obj_size
;
1349 unsigned int ipv6_pinfo_offset
;
1350 slab_flags_t slab_flags
;
1351 unsigned int useroffset
; /* Usercopy region offset */
1352 unsigned int usersize
; /* Usercopy region size */
1354 unsigned int __percpu
*orphan_count
;
1356 struct request_sock_ops
*rsk_prot
;
1357 struct timewait_sock_ops
*twsk_prot
;
1360 struct inet_hashinfo
*hashinfo
;
1361 struct udp_table
*udp_table
;
1362 struct raw_hashinfo
*raw_hash
;
1363 struct smc_hashinfo
*smc_hash
;
1366 struct module
*owner
;
1370 struct list_head node
;
1371 int (*diag_destroy
)(struct sock
*sk
, int err
);
1372 } __randomize_layout
;
1374 int proto_register(struct proto
*prot
, int alloc_slab
);
1375 void proto_unregister(struct proto
*prot
);
1376 int sock_load_diag_module(int family
, int protocol
);
1378 INDIRECT_CALLABLE_DECLARE(bool tcp_stream_memory_free(const struct sock
*sk
, int wake
));
1380 static inline int sk_forward_alloc_get(const struct sock
*sk
)
1382 #if IS_ENABLED(CONFIG_MPTCP)
1383 if (sk
->sk_prot
->forward_alloc_get
)
1384 return sk
->sk_prot
->forward_alloc_get(sk
);
1386 return READ_ONCE(sk
->sk_forward_alloc
);
1389 static inline bool __sk_stream_memory_free(const struct sock
*sk
, int wake
)
1391 if (READ_ONCE(sk
->sk_wmem_queued
) >= READ_ONCE(sk
->sk_sndbuf
))
1394 return sk
->sk_prot
->stream_memory_free
?
1395 INDIRECT_CALL_INET_1(sk
->sk_prot
->stream_memory_free
,
1396 tcp_stream_memory_free
, sk
, wake
) : true;
1399 static inline bool sk_stream_memory_free(const struct sock
*sk
)
1401 return __sk_stream_memory_free(sk
, 0);
1404 static inline bool __sk_stream_is_writeable(const struct sock
*sk
, int wake
)
1406 return sk_stream_wspace(sk
) >= sk_stream_min_wspace(sk
) &&
1407 __sk_stream_memory_free(sk
, wake
);
1410 static inline bool sk_stream_is_writeable(const struct sock
*sk
)
1412 return __sk_stream_is_writeable(sk
, 0);
1415 static inline int sk_under_cgroup_hierarchy(struct sock
*sk
,
1416 struct cgroup
*ancestor
)
1418 #ifdef CONFIG_SOCK_CGROUP_DATA
1419 return cgroup_is_descendant(sock_cgroup_ptr(&sk
->sk_cgrp_data
),
1426 static inline bool sk_has_memory_pressure(const struct sock
*sk
)
1428 return sk
->sk_prot
->memory_pressure
!= NULL
;
1431 static inline bool sk_under_global_memory_pressure(const struct sock
*sk
)
1433 return sk
->sk_prot
->memory_pressure
&&
1434 !!READ_ONCE(*sk
->sk_prot
->memory_pressure
);
1437 static inline bool sk_under_memory_pressure(const struct sock
*sk
)
1439 if (!sk
->sk_prot
->memory_pressure
)
1442 if (mem_cgroup_sockets_enabled
&& sk
->sk_memcg
&&
1443 mem_cgroup_under_socket_pressure(sk
->sk_memcg
))
1446 return !!READ_ONCE(*sk
->sk_prot
->memory_pressure
);
1450 proto_memory_allocated(const struct proto
*prot
)
1452 return max(0L, atomic_long_read(prot
->memory_allocated
));
1456 sk_memory_allocated(const struct sock
*sk
)
1458 return proto_memory_allocated(sk
->sk_prot
);
1461 /* 1 MB per cpu, in page units */
1462 #define SK_MEMORY_PCPU_RESERVE (1 << (20 - PAGE_SHIFT))
1465 sk_memory_allocated_add(struct sock
*sk
, int amt
)
1470 local_reserve
= __this_cpu_add_return(*sk
->sk_prot
->per_cpu_fw_alloc
, amt
);
1471 if (local_reserve
>= SK_MEMORY_PCPU_RESERVE
) {
1472 __this_cpu_sub(*sk
->sk_prot
->per_cpu_fw_alloc
, local_reserve
);
1473 atomic_long_add(local_reserve
, sk
->sk_prot
->memory_allocated
);
1479 sk_memory_allocated_sub(struct sock
*sk
, int amt
)
1484 local_reserve
= __this_cpu_sub_return(*sk
->sk_prot
->per_cpu_fw_alloc
, amt
);
1485 if (local_reserve
<= -SK_MEMORY_PCPU_RESERVE
) {
1486 __this_cpu_sub(*sk
->sk_prot
->per_cpu_fw_alloc
, local_reserve
);
1487 atomic_long_add(local_reserve
, sk
->sk_prot
->memory_allocated
);
1492 #define SK_ALLOC_PERCPU_COUNTER_BATCH 16
1494 static inline void sk_sockets_allocated_dec(struct sock
*sk
)
1496 percpu_counter_add_batch(sk
->sk_prot
->sockets_allocated
, -1,
1497 SK_ALLOC_PERCPU_COUNTER_BATCH
);
1500 static inline void sk_sockets_allocated_inc(struct sock
*sk
)
1502 percpu_counter_add_batch(sk
->sk_prot
->sockets_allocated
, 1,
1503 SK_ALLOC_PERCPU_COUNTER_BATCH
);
1507 sk_sockets_allocated_read_positive(struct sock
*sk
)
1509 return percpu_counter_read_positive(sk
->sk_prot
->sockets_allocated
);
1513 proto_sockets_allocated_sum_positive(struct proto
*prot
)
1515 return percpu_counter_sum_positive(prot
->sockets_allocated
);
1519 proto_memory_pressure(struct proto
*prot
)
1521 if (!prot
->memory_pressure
)
1523 return !!READ_ONCE(*prot
->memory_pressure
);
1527 #ifdef CONFIG_PROC_FS
1528 #define PROTO_INUSE_NR 64 /* should be enough for the first time */
1531 int val
[PROTO_INUSE_NR
];
1534 static inline void sock_prot_inuse_add(const struct net
*net
,
1535 const struct proto
*prot
, int val
)
1537 this_cpu_add(net
->core
.prot_inuse
->val
[prot
->inuse_idx
], val
);
1540 static inline void sock_inuse_add(const struct net
*net
, int val
)
1542 this_cpu_add(net
->core
.prot_inuse
->all
, val
);
1545 int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
1546 int sock_inuse_get(struct net
*net
);
1548 static inline void sock_prot_inuse_add(const struct net
*net
,
1549 const struct proto
*prot
, int val
)
1553 static inline void sock_inuse_add(const struct net
*net
, int val
)
1559 /* With per-bucket locks this operation is not-atomic, so that
1560 * this version is not worse.
1562 static inline int __sk_prot_rehash(struct sock
*sk
)
1564 sk
->sk_prot
->unhash(sk
);
1565 return sk
->sk_prot
->hash(sk
);
1568 /* About 10 seconds */
1569 #define SOCK_DESTROY_TIME (10*HZ)
1571 /* Sockets 0-1023 can't be bound to unless you are superuser */
1572 #define PROT_SOCK 1024
1574 #define SHUTDOWN_MASK 3
1575 #define RCV_SHUTDOWN 1
1576 #define SEND_SHUTDOWN 2
1578 #define SOCK_BINDADDR_LOCK 4
1579 #define SOCK_BINDPORT_LOCK 8
1581 struct socket_alloc
{
1582 struct socket socket
;
1583 struct inode vfs_inode
;
1586 static inline struct socket
*SOCKET_I(struct inode
*inode
)
1588 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
1591 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
1593 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
1597 * Functions for memory accounting
1599 int __sk_mem_raise_allocated(struct sock
*sk
, int size
, int amt
, int kind
);
1600 int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
1601 void __sk_mem_reduce_allocated(struct sock
*sk
, int amount
);
1602 void __sk_mem_reclaim(struct sock
*sk
, int amount
);
1604 #define SK_MEM_SEND 0
1605 #define SK_MEM_RECV 1
1607 /* sysctl_mem values are in pages */
1608 static inline long sk_prot_mem_limits(const struct sock
*sk
, int index
)
1610 return READ_ONCE(sk
->sk_prot
->sysctl_mem
[index
]);
1613 static inline int sk_mem_pages(int amt
)
1615 return (amt
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
1618 static inline bool sk_has_account(struct sock
*sk
)
1620 /* return true if protocol supports memory accounting */
1621 return !!sk
->sk_prot
->memory_allocated
;
1624 static inline bool sk_wmem_schedule(struct sock
*sk
, int size
)
1628 if (!sk_has_account(sk
))
1630 delta
= size
- sk
->sk_forward_alloc
;
1631 return delta
<= 0 || __sk_mem_schedule(sk
, delta
, SK_MEM_SEND
);
1635 sk_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
, int size
)
1639 if (!sk_has_account(sk
))
1641 delta
= size
- sk
->sk_forward_alloc
;
1642 return delta
<= 0 || __sk_mem_schedule(sk
, delta
, SK_MEM_RECV
) ||
1643 skb_pfmemalloc(skb
);
1646 static inline int sk_unused_reserved_mem(const struct sock
*sk
)
1650 if (likely(!sk
->sk_reserved_mem
))
1653 unused_mem
= sk
->sk_reserved_mem
- sk
->sk_wmem_queued
-
1654 atomic_read(&sk
->sk_rmem_alloc
);
1656 return unused_mem
> 0 ? unused_mem
: 0;
1659 static inline void sk_mem_reclaim(struct sock
*sk
)
1663 if (!sk_has_account(sk
))
1666 reclaimable
= sk
->sk_forward_alloc
- sk_unused_reserved_mem(sk
);
1668 if (reclaimable
>= (int)PAGE_SIZE
)
1669 __sk_mem_reclaim(sk
, reclaimable
);
1672 static inline void sk_mem_reclaim_final(struct sock
*sk
)
1674 sk
->sk_reserved_mem
= 0;
1678 static inline void sk_mem_charge(struct sock
*sk
, int size
)
1680 if (!sk_has_account(sk
))
1682 sk_forward_alloc_add(sk
, -size
);
1685 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
1687 if (!sk_has_account(sk
))
1689 sk_forward_alloc_add(sk
, size
);
1694 * Macro so as to not evaluate some arguments when
1695 * lockdep is not enabled.
1697 * Mark both the sk_lock and the sk_lock.slock as a
1698 * per-address-family lock class.
1700 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1702 sk->sk_lock.owned = 0; \
1703 init_waitqueue_head(&sk->sk_lock.wq); \
1704 spin_lock_init(&(sk)->sk_lock.slock); \
1705 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1706 sizeof((sk)->sk_lock)); \
1707 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1709 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1712 static inline bool lockdep_sock_is_held(const struct sock
*sk
)
1714 return lockdep_is_held(&sk
->sk_lock
) ||
1715 lockdep_is_held(&sk
->sk_lock
.slock
);
1718 void lock_sock_nested(struct sock
*sk
, int subclass
);
1720 static inline void lock_sock(struct sock
*sk
)
1722 lock_sock_nested(sk
, 0);
1725 void __lock_sock(struct sock
*sk
);
1726 void __release_sock(struct sock
*sk
);
1727 void release_sock(struct sock
*sk
);
1729 /* BH context may only use the following locking interface. */
1730 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1731 #define bh_lock_sock_nested(__sk) \
1732 spin_lock_nested(&((__sk)->sk_lock.slock), \
1733 SINGLE_DEPTH_NESTING)
1734 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1736 bool __lock_sock_fast(struct sock
*sk
) __acquires(&sk
->sk_lock
.slock
);
1739 * lock_sock_fast - fast version of lock_sock
1742 * This version should be used for very small section, where process wont block
1743 * return false if fast path is taken:
1745 * sk_lock.slock locked, owned = 0, BH disabled
1747 * return true if slow path is taken:
1749 * sk_lock.slock unlocked, owned = 1, BH enabled
1751 static inline bool lock_sock_fast(struct sock
*sk
)
1753 /* The sk_lock has mutex_lock() semantics here. */
1754 mutex_acquire(&sk
->sk_lock
.dep_map
, 0, 0, _RET_IP_
);
1756 return __lock_sock_fast(sk
);
1759 /* fast socket lock variant for caller already holding a [different] socket lock */
1760 static inline bool lock_sock_fast_nested(struct sock
*sk
)
1762 mutex_acquire(&sk
->sk_lock
.dep_map
, SINGLE_DEPTH_NESTING
, 0, _RET_IP_
);
1764 return __lock_sock_fast(sk
);
1768 * unlock_sock_fast - complement of lock_sock_fast
1772 * fast unlock socket for user context.
1773 * If slow mode is on, we call regular release_sock()
1775 static inline void unlock_sock_fast(struct sock
*sk
, bool slow
)
1776 __releases(&sk
->sk_lock
.slock
)
1780 __release(&sk
->sk_lock
.slock
);
1782 mutex_release(&sk
->sk_lock
.dep_map
, _RET_IP_
);
1783 spin_unlock_bh(&sk
->sk_lock
.slock
);
1787 void sockopt_lock_sock(struct sock
*sk
);
1788 void sockopt_release_sock(struct sock
*sk
);
1789 bool sockopt_ns_capable(struct user_namespace
*ns
, int cap
);
1790 bool sockopt_capable(int cap
);
1792 /* Used by processes to "lock" a socket state, so that
1793 * interrupts and bottom half handlers won't change it
1794 * from under us. It essentially blocks any incoming
1795 * packets, so that we won't get any new data or any
1796 * packets that change the state of the socket.
1798 * While locked, BH processing will add new packets to
1799 * the backlog queue. This queue is processed by the
1800 * owner of the socket lock right before it is released.
1802 * Since ~2.3.5 it is also exclusive sleep lock serializing
1803 * accesses from user process context.
1806 static inline void sock_owned_by_me(const struct sock
*sk
)
1808 #ifdef CONFIG_LOCKDEP
1809 WARN_ON_ONCE(!lockdep_sock_is_held(sk
) && debug_locks
);
1813 static inline bool sock_owned_by_user(const struct sock
*sk
)
1815 sock_owned_by_me(sk
);
1816 return sk
->sk_lock
.owned
;
1819 static inline bool sock_owned_by_user_nocheck(const struct sock
*sk
)
1821 return sk
->sk_lock
.owned
;
1824 static inline void sock_release_ownership(struct sock
*sk
)
1826 if (sock_owned_by_user_nocheck(sk
)) {
1827 sk
->sk_lock
.owned
= 0;
1829 /* The sk_lock has mutex_unlock() semantics: */
1830 mutex_release(&sk
->sk_lock
.dep_map
, _RET_IP_
);
1834 /* no reclassification while locks are held */
1835 static inline bool sock_allow_reclassification(const struct sock
*csk
)
1837 struct sock
*sk
= (struct sock
*)csk
;
1839 return !sock_owned_by_user_nocheck(sk
) &&
1840 !spin_is_locked(&sk
->sk_lock
.slock
);
1843 struct sock
*sk_alloc(struct net
*net
, int family
, gfp_t priority
,
1844 struct proto
*prot
, int kern
);
1845 void sk_free(struct sock
*sk
);
1846 void sk_destruct(struct sock
*sk
);
1847 struct sock
*sk_clone_lock(const struct sock
*sk
, const gfp_t priority
);
1848 void sk_free_unlock_clone(struct sock
*sk
);
1850 struct sk_buff
*sock_wmalloc(struct sock
*sk
, unsigned long size
, int force
,
1852 void __sock_wfree(struct sk_buff
*skb
);
1853 void sock_wfree(struct sk_buff
*skb
);
1854 struct sk_buff
*sock_omalloc(struct sock
*sk
, unsigned long size
,
1856 void skb_orphan_partial(struct sk_buff
*skb
);
1857 void sock_rfree(struct sk_buff
*skb
);
1858 void sock_efree(struct sk_buff
*skb
);
1860 void sock_edemux(struct sk_buff
*skb
);
1861 void sock_pfree(struct sk_buff
*skb
);
1863 #define sock_edemux sock_efree
1866 int sk_setsockopt(struct sock
*sk
, int level
, int optname
,
1867 sockptr_t optval
, unsigned int optlen
);
1868 int sock_setsockopt(struct socket
*sock
, int level
, int op
,
1869 sockptr_t optval
, unsigned int optlen
);
1871 int sk_getsockopt(struct sock
*sk
, int level
, int optname
,
1872 sockptr_t optval
, sockptr_t optlen
);
1873 int sock_getsockopt(struct socket
*sock
, int level
, int op
,
1874 char __user
*optval
, int __user
*optlen
);
1875 int sock_gettstamp(struct socket
*sock
, void __user
*userstamp
,
1876 bool timeval
, bool time32
);
1877 struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
, unsigned long header_len
,
1878 unsigned long data_len
, int noblock
,
1879 int *errcode
, int max_page_order
);
1881 static inline struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
,
1883 int noblock
, int *errcode
)
1885 return sock_alloc_send_pskb(sk
, size
, 0, noblock
, errcode
, 0);
1888 void *sock_kmalloc(struct sock
*sk
, int size
, gfp_t priority
);
1889 void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
1890 void sock_kzfree_s(struct sock
*sk
, void *mem
, int size
);
1891 void sk_send_sigurg(struct sock
*sk
);
1893 static inline void sock_replace_proto(struct sock
*sk
, struct proto
*proto
)
1896 clear_bit(SOCK_SUPPORT_ZC
, &sk
->sk_socket
->flags
);
1897 WRITE_ONCE(sk
->sk_prot
, proto
);
1900 struct sockcm_cookie
{
1906 static inline void sockcm_init(struct sockcm_cookie
*sockc
,
1907 const struct sock
*sk
)
1909 *sockc
= (struct sockcm_cookie
) {
1910 .tsflags
= READ_ONCE(sk
->sk_tsflags
)
1914 int __sock_cmsg_send(struct sock
*sk
, struct cmsghdr
*cmsg
,
1915 struct sockcm_cookie
*sockc
);
1916 int sock_cmsg_send(struct sock
*sk
, struct msghdr
*msg
,
1917 struct sockcm_cookie
*sockc
);
1920 * Functions to fill in entries in struct proto_ops when a protocol
1921 * does not implement a particular function.
1923 int sock_no_bind(struct socket
*, struct sockaddr
*, int);
1924 int sock_no_connect(struct socket
*, struct sockaddr
*, int, int);
1925 int sock_no_socketpair(struct socket
*, struct socket
*);
1926 int sock_no_accept(struct socket
*, struct socket
*, int, bool);
1927 int sock_no_getname(struct socket
*, struct sockaddr
*, int);
1928 int sock_no_ioctl(struct socket
*, unsigned int, unsigned long);
1929 int sock_no_listen(struct socket
*, int);
1930 int sock_no_shutdown(struct socket
*, int);
1931 int sock_no_sendmsg(struct socket
*, struct msghdr
*, size_t);
1932 int sock_no_sendmsg_locked(struct sock
*sk
, struct msghdr
*msg
, size_t len
);
1933 int sock_no_recvmsg(struct socket
*, struct msghdr
*, size_t, int);
1934 int sock_no_mmap(struct file
*file
, struct socket
*sock
,
1935 struct vm_area_struct
*vma
);
1938 * Functions to fill in entries in struct proto_ops when a protocol
1939 * uses the inet style.
1941 int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1942 char __user
*optval
, int __user
*optlen
);
1943 int sock_common_recvmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
,
1945 int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1946 sockptr_t optval
, unsigned int optlen
);
1948 void sk_common_release(struct sock
*sk
);
1951 * Default socket callbacks and setup code
1954 /* Initialise core socket variables using an explicit uid. */
1955 void sock_init_data_uid(struct socket
*sock
, struct sock
*sk
, kuid_t uid
);
1957 /* Initialise core socket variables.
1958 * Assumes struct socket *sock is embedded in a struct socket_alloc.
1960 void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1963 * Socket reference counting postulates.
1965 * * Each user of socket SHOULD hold a reference count.
1966 * * Each access point to socket (an hash table bucket, reference from a list,
1967 * running timer, skb in flight MUST hold a reference count.
1968 * * When reference count hits 0, it means it will never increase back.
1969 * * When reference count hits 0, it means that no references from
1970 * outside exist to this socket and current process on current CPU
1971 * is last user and may/should destroy this socket.
1972 * * sk_free is called from any context: process, BH, IRQ. When
1973 * it is called, socket has no references from outside -> sk_free
1974 * may release descendant resources allocated by the socket, but
1975 * to the time when it is called, socket is NOT referenced by any
1976 * hash tables, lists etc.
1977 * * Packets, delivered from outside (from network or from another process)
1978 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1979 * when they sit in queue. Otherwise, packets will leak to hole, when
1980 * socket is looked up by one cpu and unhasing is made by another CPU.
1981 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1982 * (leak to backlog). Packet socket does all the processing inside
1983 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1984 * use separate SMP lock, so that they are prone too.
1987 /* Ungrab socket and destroy it, if it was the last reference. */
1988 static inline void sock_put(struct sock
*sk
)
1990 if (refcount_dec_and_test(&sk
->sk_refcnt
))
1993 /* Generic version of sock_put(), dealing with all sockets
1994 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
1996 void sock_gen_put(struct sock
*sk
);
1998 int __sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
, const int nested
,
1999 unsigned int trim_cap
, bool refcounted
);
2000 static inline int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
2003 return __sk_receive_skb(sk
, skb
, nested
, 1, true);
2006 static inline void sk_tx_queue_set(struct sock
*sk
, int tx_queue
)
2008 /* sk_tx_queue_mapping accept only upto a 16-bit value */
2009 if (WARN_ON_ONCE((unsigned short)tx_queue
>= USHRT_MAX
))
2011 sk
->sk_tx_queue_mapping
= tx_queue
;
2014 #define NO_QUEUE_MAPPING USHRT_MAX
2016 static inline void sk_tx_queue_clear(struct sock
*sk
)
2018 sk
->sk_tx_queue_mapping
= NO_QUEUE_MAPPING
;
2021 static inline int sk_tx_queue_get(const struct sock
*sk
)
2023 if (sk
&& sk
->sk_tx_queue_mapping
!= NO_QUEUE_MAPPING
)
2024 return sk
->sk_tx_queue_mapping
;
2029 static inline void __sk_rx_queue_set(struct sock
*sk
,
2030 const struct sk_buff
*skb
,
2033 #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
2034 if (skb_rx_queue_recorded(skb
)) {
2035 u16 rx_queue
= skb_get_rx_queue(skb
);
2038 unlikely(READ_ONCE(sk
->sk_rx_queue_mapping
) != rx_queue
))
2039 WRITE_ONCE(sk
->sk_rx_queue_mapping
, rx_queue
);
2044 static inline void sk_rx_queue_set(struct sock
*sk
, const struct sk_buff
*skb
)
2046 __sk_rx_queue_set(sk
, skb
, true);
2049 static inline void sk_rx_queue_update(struct sock
*sk
, const struct sk_buff
*skb
)
2051 __sk_rx_queue_set(sk
, skb
, false);
2054 static inline void sk_rx_queue_clear(struct sock
*sk
)
2056 #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
2057 WRITE_ONCE(sk
->sk_rx_queue_mapping
, NO_QUEUE_MAPPING
);
2061 static inline int sk_rx_queue_get(const struct sock
*sk
)
2063 #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
2065 int res
= READ_ONCE(sk
->sk_rx_queue_mapping
);
2067 if (res
!= NO_QUEUE_MAPPING
)
2075 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
2077 sk
->sk_socket
= sock
;
2080 static inline wait_queue_head_t
*sk_sleep(struct sock
*sk
)
2082 BUILD_BUG_ON(offsetof(struct socket_wq
, wait
) != 0);
2083 return &rcu_dereference_raw(sk
->sk_wq
)->wait
;
2085 /* Detach socket from process context.
2086 * Announce socket dead, detach it from wait queue and inode.
2087 * Note that parent inode held reference count on this struct sock,
2088 * we do not release it in this function, because protocol
2089 * probably wants some additional cleanups or even continuing
2090 * to work with this socket (TCP).
2092 static inline void sock_orphan(struct sock
*sk
)
2094 write_lock_bh(&sk
->sk_callback_lock
);
2095 sock_set_flag(sk
, SOCK_DEAD
);
2096 sk_set_socket(sk
, NULL
);
2098 write_unlock_bh(&sk
->sk_callback_lock
);
2101 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
2103 WARN_ON(parent
->sk
);
2104 write_lock_bh(&sk
->sk_callback_lock
);
2105 rcu_assign_pointer(sk
->sk_wq
, &parent
->wq
);
2107 sk_set_socket(sk
, parent
);
2108 sk
->sk_uid
= SOCK_INODE(parent
)->i_uid
;
2109 security_sock_graft(sk
, parent
);
2110 write_unlock_bh(&sk
->sk_callback_lock
);
2113 kuid_t
sock_i_uid(struct sock
*sk
);
2114 unsigned long __sock_i_ino(struct sock
*sk
);
2115 unsigned long sock_i_ino(struct sock
*sk
);
2117 static inline kuid_t
sock_net_uid(const struct net
*net
, const struct sock
*sk
)
2119 return sk
? sk
->sk_uid
: make_kuid(net
->user_ns
, 0);
2122 static inline u32
net_tx_rndhash(void)
2124 u32 v
= get_random_u32();
2129 static inline void sk_set_txhash(struct sock
*sk
)
2131 /* This pairs with READ_ONCE() in skb_set_hash_from_sk() */
2132 WRITE_ONCE(sk
->sk_txhash
, net_tx_rndhash());
2135 static inline bool sk_rethink_txhash(struct sock
*sk
)
2137 if (sk
->sk_txhash
&& sk
->sk_txrehash
== SOCK_TXREHASH_ENABLED
) {
2144 static inline struct dst_entry
*
2145 __sk_dst_get(struct sock
*sk
)
2147 return rcu_dereference_check(sk
->sk_dst_cache
,
2148 lockdep_sock_is_held(sk
));
2151 static inline struct dst_entry
*
2152 sk_dst_get(struct sock
*sk
)
2154 struct dst_entry
*dst
;
2157 dst
= rcu_dereference(sk
->sk_dst_cache
);
2158 if (dst
&& !rcuref_get(&dst
->__rcuref
))
2164 static inline void __dst_negative_advice(struct sock
*sk
)
2166 struct dst_entry
*ndst
, *dst
= __sk_dst_get(sk
);
2168 if (dst
&& dst
->ops
->negative_advice
) {
2169 ndst
= dst
->ops
->negative_advice(dst
);
2172 rcu_assign_pointer(sk
->sk_dst_cache
, ndst
);
2173 sk_tx_queue_clear(sk
);
2174 sk
->sk_dst_pending_confirm
= 0;
2179 static inline void dst_negative_advice(struct sock
*sk
)
2181 sk_rethink_txhash(sk
);
2182 __dst_negative_advice(sk
);
2186 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
2188 struct dst_entry
*old_dst
;
2190 sk_tx_queue_clear(sk
);
2191 sk
->sk_dst_pending_confirm
= 0;
2192 old_dst
= rcu_dereference_protected(sk
->sk_dst_cache
,
2193 lockdep_sock_is_held(sk
));
2194 rcu_assign_pointer(sk
->sk_dst_cache
, dst
);
2195 dst_release(old_dst
);
2199 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
2201 struct dst_entry
*old_dst
;
2203 sk_tx_queue_clear(sk
);
2204 sk
->sk_dst_pending_confirm
= 0;
2205 old_dst
= xchg((__force
struct dst_entry
**)&sk
->sk_dst_cache
, dst
);
2206 dst_release(old_dst
);
2210 __sk_dst_reset(struct sock
*sk
)
2212 __sk_dst_set(sk
, NULL
);
2216 sk_dst_reset(struct sock
*sk
)
2218 sk_dst_set(sk
, NULL
);
2221 struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
2223 struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
2225 static inline void sk_dst_confirm(struct sock
*sk
)
2227 if (!READ_ONCE(sk
->sk_dst_pending_confirm
))
2228 WRITE_ONCE(sk
->sk_dst_pending_confirm
, 1);
2231 static inline void sock_confirm_neigh(struct sk_buff
*skb
, struct neighbour
*n
)
2233 if (skb_get_dst_pending_confirm(skb
)) {
2234 struct sock
*sk
= skb
->sk
;
2236 if (sk
&& READ_ONCE(sk
->sk_dst_pending_confirm
))
2237 WRITE_ONCE(sk
->sk_dst_pending_confirm
, 0);
2242 bool sk_mc_loop(struct sock
*sk
);
2244 static inline bool sk_can_gso(const struct sock
*sk
)
2246 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
2249 void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
2251 static inline void sk_gso_disable(struct sock
*sk
)
2253 sk
->sk_gso_disabled
= 1;
2254 sk
->sk_route_caps
&= ~NETIF_F_GSO_MASK
;
2257 static inline int skb_do_copy_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
2258 struct iov_iter
*from
, char *to
,
2259 int copy
, int offset
)
2261 if (skb
->ip_summed
== CHECKSUM_NONE
) {
2263 if (!csum_and_copy_from_iter_full(to
, copy
, &csum
, from
))
2265 skb
->csum
= csum_block_add(skb
->csum
, csum
, offset
);
2266 } else if (sk
->sk_route_caps
& NETIF_F_NOCACHE_COPY
) {
2267 if (!copy_from_iter_full_nocache(to
, copy
, from
))
2269 } else if (!copy_from_iter_full(to
, copy
, from
))
2275 static inline int skb_add_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
2276 struct iov_iter
*from
, int copy
)
2278 int err
, offset
= skb
->len
;
2280 err
= skb_do_copy_data_nocache(sk
, skb
, from
, skb_put(skb
, copy
),
2283 __skb_trim(skb
, offset
);
2288 static inline int skb_copy_to_page_nocache(struct sock
*sk
, struct iov_iter
*from
,
2289 struct sk_buff
*skb
,
2295 err
= skb_do_copy_data_nocache(sk
, skb
, from
, page_address(page
) + off
,
2300 skb_len_add(skb
, copy
);
2301 sk_wmem_queued_add(sk
, copy
);
2302 sk_mem_charge(sk
, copy
);
2307 * sk_wmem_alloc_get - returns write allocations
2310 * Return: sk_wmem_alloc minus initial offset of one
2312 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
2314 return refcount_read(&sk
->sk_wmem_alloc
) - 1;
2318 * sk_rmem_alloc_get - returns read allocations
2321 * Return: sk_rmem_alloc
2323 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
2325 return atomic_read(&sk
->sk_rmem_alloc
);
2329 * sk_has_allocations - check if allocations are outstanding
2332 * Return: true if socket has write or read allocations
2334 static inline bool sk_has_allocations(const struct sock
*sk
)
2336 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
2340 * skwq_has_sleeper - check if there are any waiting processes
2341 * @wq: struct socket_wq
2343 * Return: true if socket_wq has waiting processes
2345 * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
2346 * barrier call. They were added due to the race found within the tcp code.
2348 * Consider following tcp code paths::
2351 * sys_select receive packet
2353 * __add_wait_queue update tp->rcv_nxt
2355 * tp->rcv_nxt check sock_def_readable
2357 * schedule rcu_read_lock();
2358 * wq = rcu_dereference(sk->sk_wq);
2359 * if (wq && waitqueue_active(&wq->wait))
2360 * wake_up_interruptible(&wq->wait)
2364 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
2365 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
2366 * could then endup calling schedule and sleep forever if there are no more
2367 * data on the socket.
2370 static inline bool skwq_has_sleeper(struct socket_wq
*wq
)
2372 return wq
&& wq_has_sleeper(&wq
->wait
);
2376 * sock_poll_wait - place memory barrier behind the poll_wait call.
2378 * @sock: socket to wait on
2381 * See the comments in the wq_has_sleeper function.
2383 static inline void sock_poll_wait(struct file
*filp
, struct socket
*sock
,
2386 if (!poll_does_not_wait(p
)) {
2387 poll_wait(filp
, &sock
->wq
.wait
, p
);
2388 /* We need to be sure we are in sync with the
2389 * socket flags modification.
2391 * This memory barrier is paired in the wq_has_sleeper.
2397 static inline void skb_set_hash_from_sk(struct sk_buff
*skb
, struct sock
*sk
)
2399 /* This pairs with WRITE_ONCE() in sk_set_txhash() */
2400 u32 txhash
= READ_ONCE(sk
->sk_txhash
);
2408 void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
);
2411 * Queue a received datagram if it will fit. Stream and sequenced
2412 * protocols can't normally use this as they need to fit buffers in
2413 * and play with them.
2415 * Inlined as it's very short and called for pretty much every
2416 * packet ever received.
2418 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
2422 skb
->destructor
= sock_rfree
;
2423 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
2424 sk_mem_charge(sk
, skb
->truesize
);
2427 static inline __must_check
bool skb_set_owner_sk_safe(struct sk_buff
*skb
, struct sock
*sk
)
2429 if (sk
&& refcount_inc_not_zero(&sk
->sk_refcnt
)) {
2431 skb
->destructor
= sock_efree
;
2438 static inline struct sk_buff
*skb_clone_and_charge_r(struct sk_buff
*skb
, struct sock
*sk
)
2440 skb
= skb_clone(skb
, sk_gfp_mask(sk
, GFP_ATOMIC
));
2442 if (sk_rmem_schedule(sk
, skb
, skb
->truesize
)) {
2443 skb_set_owner_r(skb
, sk
);
2451 static inline void skb_prepare_for_gro(struct sk_buff
*skb
)
2453 if (skb
->destructor
!= sock_wfree
) {
2460 void sk_reset_timer(struct sock
*sk
, struct timer_list
*timer
,
2461 unsigned long expires
);
2463 void sk_stop_timer(struct sock
*sk
, struct timer_list
*timer
);
2465 void sk_stop_timer_sync(struct sock
*sk
, struct timer_list
*timer
);
2467 int __sk_queue_drop_skb(struct sock
*sk
, struct sk_buff_head
*sk_queue
,
2468 struct sk_buff
*skb
, unsigned int flags
,
2469 void (*destructor
)(struct sock
*sk
,
2470 struct sk_buff
*skb
));
2471 int __sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
2473 int sock_queue_rcv_skb_reason(struct sock
*sk
, struct sk_buff
*skb
,
2474 enum skb_drop_reason
*reason
);
2476 static inline int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
)
2478 return sock_queue_rcv_skb_reason(sk
, skb
, NULL
);
2481 int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
);
2482 struct sk_buff
*sock_dequeue_err_skb(struct sock
*sk
);
2485 * Recover an error report and clear atomically
2488 static inline int sock_error(struct sock
*sk
)
2492 /* Avoid an atomic operation for the common case.
2493 * This is racy since another cpu/thread can change sk_err under us.
2495 if (likely(data_race(!sk
->sk_err
)))
2498 err
= xchg(&sk
->sk_err
, 0);
2502 void sk_error_report(struct sock
*sk
);
2504 static inline unsigned long sock_wspace(struct sock
*sk
)
2508 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
2509 amt
= sk
->sk_sndbuf
- refcount_read(&sk
->sk_wmem_alloc
);
2517 * We use sk->sk_wq_raw, from contexts knowing this
2518 * pointer is not NULL and cannot disappear/change.
2520 static inline void sk_set_bit(int nr
, struct sock
*sk
)
2522 if ((nr
== SOCKWQ_ASYNC_NOSPACE
|| nr
== SOCKWQ_ASYNC_WAITDATA
) &&
2523 !sock_flag(sk
, SOCK_FASYNC
))
2526 set_bit(nr
, &sk
->sk_wq_raw
->flags
);
2529 static inline void sk_clear_bit(int nr
, struct sock
*sk
)
2531 if ((nr
== SOCKWQ_ASYNC_NOSPACE
|| nr
== SOCKWQ_ASYNC_WAITDATA
) &&
2532 !sock_flag(sk
, SOCK_FASYNC
))
2535 clear_bit(nr
, &sk
->sk_wq_raw
->flags
);
2538 static inline void sk_wake_async(const struct sock
*sk
, int how
, int band
)
2540 if (sock_flag(sk
, SOCK_FASYNC
)) {
2542 sock_wake_async(rcu_dereference(sk
->sk_wq
), how
, band
);
2547 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2548 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2549 * Note: for send buffers, TCP works better if we can build two skbs at
2552 #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2554 #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2555 #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
2557 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
2561 if (sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)
2564 val
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
2565 val
= max_t(u32
, val
, sk_unused_reserved_mem(sk
));
2567 WRITE_ONCE(sk
->sk_sndbuf
, max_t(u32
, val
, SOCK_MIN_SNDBUF
));
2571 * sk_page_frag - return an appropriate page_frag
2574 * Use the per task page_frag instead of the per socket one for
2575 * optimization when we know that we're in process context and own
2576 * everything that's associated with %current.
2578 * Both direct reclaim and page faults can nest inside other
2579 * socket operations and end up recursing into sk_page_frag()
2580 * while it's already in use: explicitly avoid task page_frag
2581 * when users disable sk_use_task_frag.
2583 * Return: a per task page_frag if context allows that,
2584 * otherwise a per socket one.
2586 static inline struct page_frag
*sk_page_frag(struct sock
*sk
)
2588 if (sk
->sk_use_task_frag
)
2589 return ¤t
->task_frag
;
2591 return &sk
->sk_frag
;
2594 bool sk_page_frag_refill(struct sock
*sk
, struct page_frag
*pfrag
);
2597 * Default write policy as shown to user space via poll/select/SIGIO
2599 static inline bool sock_writeable(const struct sock
*sk
)
2601 return refcount_read(&sk
->sk_wmem_alloc
) < (READ_ONCE(sk
->sk_sndbuf
) >> 1);
2604 static inline gfp_t
gfp_any(void)
2606 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
2609 static inline gfp_t
gfp_memcg_charge(void)
2611 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
2614 static inline long sock_rcvtimeo(const struct sock
*sk
, bool noblock
)
2616 return noblock
? 0 : sk
->sk_rcvtimeo
;
2619 static inline long sock_sndtimeo(const struct sock
*sk
, bool noblock
)
2621 return noblock
? 0 : sk
->sk_sndtimeo
;
2624 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
2626 int v
= waitall
? len
: min_t(int, READ_ONCE(sk
->sk_rcvlowat
), len
);
2631 /* Alas, with timeout socket operations are not restartable.
2632 * Compare this to poll().
2634 static inline int sock_intr_errno(long timeo
)
2636 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
2639 struct sock_skb_cb
{
2643 /* Store sock_skb_cb at the end of skb->cb[] so protocol families
2644 * using skb->cb[] would keep using it directly and utilize its
2645 * alignement guarantee.
2647 #define SOCK_SKB_CB_OFFSET ((sizeof_field(struct sk_buff, cb) - \
2648 sizeof(struct sock_skb_cb)))
2650 #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2651 SOCK_SKB_CB_OFFSET))
2653 #define sock_skb_cb_check_size(size) \
2654 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
2657 sock_skb_set_dropcount(const struct sock
*sk
, struct sk_buff
*skb
)
2659 SOCK_SKB_CB(skb
)->dropcount
= sock_flag(sk
, SOCK_RXQ_OVFL
) ?
2660 atomic_read(&sk
->sk_drops
) : 0;
2663 static inline void sk_drops_add(struct sock
*sk
, const struct sk_buff
*skb
)
2665 int segs
= max_t(u16
, 1, skb_shinfo(skb
)->gso_segs
);
2667 atomic_add(segs
, &sk
->sk_drops
);
2670 static inline ktime_t
sock_read_timestamp(struct sock
*sk
)
2672 #if BITS_PER_LONG==32
2677 seq
= read_seqbegin(&sk
->sk_stamp_seq
);
2679 } while (read_seqretry(&sk
->sk_stamp_seq
, seq
));
2683 return READ_ONCE(sk
->sk_stamp
);
2687 static inline void sock_write_timestamp(struct sock
*sk
, ktime_t kt
)
2689 #if BITS_PER_LONG==32
2690 write_seqlock(&sk
->sk_stamp_seq
);
2692 write_sequnlock(&sk
->sk_stamp_seq
);
2694 WRITE_ONCE(sk
->sk_stamp
, kt
);
2698 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
2699 struct sk_buff
*skb
);
2700 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
2701 struct sk_buff
*skb
);
2704 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
2706 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
2707 u32 tsflags
= READ_ONCE(sk
->sk_tsflags
);
2708 ktime_t kt
= skb
->tstamp
;
2710 * generate control messages if
2711 * - receive time stamping in software requested
2712 * - software time stamp available and wanted
2713 * - hardware time stamps available and wanted
2715 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
2716 (tsflags
& SOF_TIMESTAMPING_RX_SOFTWARE
) ||
2717 (kt
&& tsflags
& SOF_TIMESTAMPING_SOFTWARE
) ||
2718 (hwtstamps
->hwtstamp
&&
2719 (tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
)))
2720 __sock_recv_timestamp(msg
, sk
, skb
);
2722 sock_write_timestamp(sk
, kt
);
2724 if (sock_flag(sk
, SOCK_WIFI_STATUS
) && skb_wifi_acked_valid(skb
))
2725 __sock_recv_wifi_status(msg
, sk
, skb
);
2728 void __sock_recv_cmsgs(struct msghdr
*msg
, struct sock
*sk
,
2729 struct sk_buff
*skb
);
2731 #define SK_DEFAULT_STAMP (-1L * NSEC_PER_SEC)
2732 static inline void sock_recv_cmsgs(struct msghdr
*msg
, struct sock
*sk
,
2733 struct sk_buff
*skb
)
2735 #define FLAGS_RECV_CMSGS ((1UL << SOCK_RXQ_OVFL) | \
2736 (1UL << SOCK_RCVTSTAMP) | \
2737 (1UL << SOCK_RCVMARK))
2738 #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2739 SOF_TIMESTAMPING_RAW_HARDWARE)
2741 if (sk
->sk_flags
& FLAGS_RECV_CMSGS
||
2742 READ_ONCE(sk
->sk_tsflags
) & TSFLAGS_ANY
)
2743 __sock_recv_cmsgs(msg
, sk
, skb
);
2744 else if (unlikely(sock_flag(sk
, SOCK_TIMESTAMP
)))
2745 sock_write_timestamp(sk
, skb
->tstamp
);
2746 else if (unlikely(sock_read_timestamp(sk
) == SK_DEFAULT_STAMP
))
2747 sock_write_timestamp(sk
, 0);
2750 void __sock_tx_timestamp(__u16 tsflags
, __u8
*tx_flags
);
2753 * _sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2754 * @sk: socket sending this packet
2755 * @tsflags: timestamping flags to use
2756 * @tx_flags: completed with instructions for time stamping
2757 * @tskey: filled in with next sk_tskey (not for TCP, which uses seqno)
2759 * Note: callers should take care of initial ``*tx_flags`` value (usually 0)
2761 static inline void _sock_tx_timestamp(struct sock
*sk
, __u16 tsflags
,
2762 __u8
*tx_flags
, __u32
*tskey
)
2764 if (unlikely(tsflags
)) {
2765 __sock_tx_timestamp(tsflags
, tx_flags
);
2766 if (tsflags
& SOF_TIMESTAMPING_OPT_ID
&& tskey
&&
2767 tsflags
& SOF_TIMESTAMPING_TX_RECORD_MASK
)
2768 *tskey
= atomic_inc_return(&sk
->sk_tskey
) - 1;
2770 if (unlikely(sock_flag(sk
, SOCK_WIFI_STATUS
)))
2771 *tx_flags
|= SKBTX_WIFI_STATUS
;
2774 static inline void sock_tx_timestamp(struct sock
*sk
, __u16 tsflags
,
2777 _sock_tx_timestamp(sk
, tsflags
, tx_flags
, NULL
);
2780 static inline void skb_setup_tx_timestamp(struct sk_buff
*skb
, __u16 tsflags
)
2782 _sock_tx_timestamp(skb
->sk
, tsflags
, &skb_shinfo(skb
)->tx_flags
,
2783 &skb_shinfo(skb
)->tskey
);
2786 static inline bool sk_is_tcp(const struct sock
*sk
)
2788 return sk
->sk_type
== SOCK_STREAM
&& sk
->sk_protocol
== IPPROTO_TCP
;
2792 * sk_eat_skb - Release a skb if it is no longer needed
2793 * @sk: socket to eat this skb from
2794 * @skb: socket buffer to eat
2796 * This routine must be called with interrupts disabled or with the socket
2797 * locked so that the sk_buff queue operation is ok.
2799 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
)
2801 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2806 skb_sk_is_prefetched(struct sk_buff
*skb
)
2809 return skb
->destructor
== sock_pfree
;
2812 #endif /* CONFIG_INET */
2815 /* This helper checks if a socket is a full socket,
2816 * ie _not_ a timewait or request socket.
2818 static inline bool sk_fullsock(const struct sock
*sk
)
2820 return (1 << sk
->sk_state
) & ~(TCPF_TIME_WAIT
| TCPF_NEW_SYN_RECV
);
2824 sk_is_refcounted(struct sock
*sk
)
2826 /* Only full sockets have sk->sk_flags. */
2827 return !sk_fullsock(sk
) || !sock_flag(sk
, SOCK_RCU_FREE
);
2831 * skb_steal_sock - steal a socket from an sk_buff
2832 * @skb: sk_buff to steal the socket from
2833 * @refcounted: is set to true if the socket is reference-counted
2834 * @prefetched: is set to true if the socket was assigned from bpf
2836 static inline struct sock
*
2837 skb_steal_sock(struct sk_buff
*skb
, bool *refcounted
, bool *prefetched
)
2840 struct sock
*sk
= skb
->sk
;
2843 *prefetched
= skb_sk_is_prefetched(skb
);
2845 *refcounted
= sk_is_refcounted(sk
);
2846 skb
->destructor
= NULL
;
2850 *prefetched
= false;
2851 *refcounted
= false;
2855 /* Checks if this SKB belongs to an HW offloaded socket
2856 * and whether any SW fallbacks are required based on dev.
2857 * Check decrypted mark in case skb_orphan() cleared socket.
2859 static inline struct sk_buff
*sk_validate_xmit_skb(struct sk_buff
*skb
,
2860 struct net_device
*dev
)
2862 #ifdef CONFIG_SOCK_VALIDATE_XMIT
2863 struct sock
*sk
= skb
->sk
;
2865 if (sk
&& sk_fullsock(sk
) && sk
->sk_validate_xmit_skb
) {
2866 skb
= sk
->sk_validate_xmit_skb(sk
, dev
, skb
);
2867 #ifdef CONFIG_TLS_DEVICE
2868 } else if (unlikely(skb
->decrypted
)) {
2869 pr_warn_ratelimited("unencrypted skb with no associated socket - dropping\n");
2879 /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2880 * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2882 static inline bool sk_listener(const struct sock
*sk
)
2884 return (1 << sk
->sk_state
) & (TCPF_LISTEN
| TCPF_NEW_SYN_RECV
);
2887 void sock_enable_timestamp(struct sock
*sk
, enum sock_flags flag
);
2888 int sock_recv_errqueue(struct sock
*sk
, struct msghdr
*msg
, int len
, int level
,
2891 bool sk_ns_capable(const struct sock
*sk
,
2892 struct user_namespace
*user_ns
, int cap
);
2893 bool sk_capable(const struct sock
*sk
, int cap
);
2894 bool sk_net_capable(const struct sock
*sk
, int cap
);
2896 void sk_get_meminfo(const struct sock
*sk
, u32
*meminfo
);
2898 /* Take into consideration the size of the struct sk_buff overhead in the
2899 * determination of these values, since that is non-constant across
2900 * platforms. This makes socket queueing behavior and performance
2901 * not depend upon such differences.
2903 #define _SK_MEM_PACKETS 256
2904 #define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
2905 #define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2906 #define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2908 extern __u32 sysctl_wmem_max
;
2909 extern __u32 sysctl_rmem_max
;
2911 extern int sysctl_tstamp_allow_data
;
2912 extern int sysctl_optmem_max
;
2914 extern __u32 sysctl_wmem_default
;
2915 extern __u32 sysctl_rmem_default
;
2917 #define SKB_FRAG_PAGE_ORDER get_order(32768)
2918 DECLARE_STATIC_KEY_FALSE(net_high_order_alloc_disable_key
);
2920 static inline int sk_get_wmem0(const struct sock
*sk
, const struct proto
*proto
)
2922 /* Does this proto have per netns sysctl_wmem ? */
2923 if (proto
->sysctl_wmem_offset
)
2924 return READ_ONCE(*(int *)((void *)sock_net(sk
) + proto
->sysctl_wmem_offset
));
2926 return READ_ONCE(*proto
->sysctl_wmem
);
2929 static inline int sk_get_rmem0(const struct sock
*sk
, const struct proto
*proto
)
2931 /* Does this proto have per netns sysctl_rmem ? */
2932 if (proto
->sysctl_rmem_offset
)
2933 return READ_ONCE(*(int *)((void *)sock_net(sk
) + proto
->sysctl_rmem_offset
));
2935 return READ_ONCE(*proto
->sysctl_rmem
);
2938 /* Default TCP Small queue budget is ~1 ms of data (1sec >> 10)
2939 * Some wifi drivers need to tweak it to get more chunks.
2940 * They can use this helper from their ndo_start_xmit()
2942 static inline void sk_pacing_shift_update(struct sock
*sk
, int val
)
2944 if (!sk
|| !sk_fullsock(sk
) || READ_ONCE(sk
->sk_pacing_shift
) == val
)
2946 WRITE_ONCE(sk
->sk_pacing_shift
, val
);
2949 /* if a socket is bound to a device, check that the given device
2950 * index is either the same or that the socket is bound to an L3
2951 * master device and the given device index is also enslaved to
2954 static inline bool sk_dev_equal_l3scope(struct sock
*sk
, int dif
)
2956 int bound_dev_if
= READ_ONCE(sk
->sk_bound_dev_if
);
2959 if (!bound_dev_if
|| bound_dev_if
== dif
)
2962 mdif
= l3mdev_master_ifindex_by_index(sock_net(sk
), dif
);
2963 if (mdif
&& mdif
== bound_dev_if
)
2969 void sock_def_readable(struct sock
*sk
);
2971 int sock_bindtoindex(struct sock
*sk
, int ifindex
, bool lock_sk
);
2972 void sock_set_timestamp(struct sock
*sk
, int optname
, bool valbool
);
2973 int sock_set_timestamping(struct sock
*sk
, int optname
,
2974 struct so_timestamping timestamping
);
2976 void sock_enable_timestamps(struct sock
*sk
);
2977 void sock_no_linger(struct sock
*sk
);
2978 void sock_set_keepalive(struct sock
*sk
);
2979 void sock_set_priority(struct sock
*sk
, u32 priority
);
2980 void sock_set_rcvbuf(struct sock
*sk
, int val
);
2981 void sock_set_mark(struct sock
*sk
, u32 val
);
2982 void sock_set_reuseaddr(struct sock
*sk
);
2983 void sock_set_reuseport(struct sock
*sk
);
2984 void sock_set_sndtimeo(struct sock
*sk
, s64 secs
);
2986 int sock_bind_add(struct sock
*sk
, struct sockaddr
*addr
, int addr_len
);
2988 int sock_get_timeout(long timeo
, void *optval
, bool old_timeval
);
2989 int sock_copy_user_timeval(struct __kernel_sock_timeval
*tv
,
2990 sockptr_t optval
, int optlen
, bool old_timeval
);
2992 int sock_ioctl_inout(struct sock
*sk
, unsigned int cmd
,
2993 void __user
*arg
, void *karg
, size_t size
);
2994 int sk_ioctl(struct sock
*sk
, unsigned int cmd
, void __user
*arg
);
2995 static inline bool sk_is_readable(struct sock
*sk
)
2997 if (sk
->sk_prot
->sock_is_readable
)
2998 return sk
->sk_prot
->sock_is_readable(sk
);
3001 #endif /* _SOCK_H */