2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the AF_INET socket handler.
8 * Version: @(#)sock.h 1.0.4 05/13/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Florian La Roche <flla@stud.uni-sb.de>
16 * Alan Cox : Volatiles in skbuff pointers. See
17 * skbuff comments. May be overdone,
18 * better to prove they can be removed
20 * Alan Cox : Added a zapped field for tcp to note
21 * a socket is reset and must stay shut up
22 * Alan Cox : New fields for options
23 * Pauline Middelink : identd support
24 * Alan Cox : Eliminate low level recv/recvfrom
25 * David S. Miller : New socket lookup architecture.
26 * Steve Whitehouse: Default routines for sock_ops
27 * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made
28 * protinfo be just a void pointer, as the
29 * protocol specific parts were moved to
30 * respective headers and ipv4/v6, etc now
31 * use private slabcaches for its socks
32 * Pedro Hortas : New flags field for socket options
35 * This program is free software; you can redistribute it and/or
36 * modify it under the terms of the GNU General Public License
37 * as published by the Free Software Foundation; either version
38 * 2 of the License, or (at your option) any later version.
43 #include <linux/hardirq.h>
44 #include <linux/kernel.h>
45 #include <linux/list.h>
46 #include <linux/list_nulls.h>
47 #include <linux/timer.h>
48 #include <linux/cache.h>
49 #include <linux/bitops.h>
50 #include <linux/lockdep.h>
51 #include <linux/netdevice.h>
52 #include <linux/skbuff.h> /* struct sk_buff */
54 #include <linux/security.h>
55 #include <linux/slab.h>
56 #include <linux/uaccess.h>
57 #include <linux/memcontrol.h>
58 #include <linux/res_counter.h>
59 #include <linux/static_key.h>
60 #include <linux/aio.h>
61 #include <linux/sched.h>
63 #include <linux/filter.h>
64 #include <linux/rculist_nulls.h>
65 #include <linux/poll.h>
67 #include <linux/atomic.h>
69 #include <net/checksum.h>
74 int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss);
75 void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg);
78 int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
83 void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
88 * This structure really needs to be cleaned up.
89 * Most of it is for TCP, and not used by any of
90 * the other protocols.
93 /* Define this to get the SOCK_DBG debugging facility. */
94 #define SOCK_DEBUGGING
96 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
97 printk(KERN_DEBUG msg); } while (0)
99 /* Validate arguments and do nothing */
100 static inline __printf(2, 3)
101 void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
106 /* This is the per-socket lock. The spinlock provides a synchronization
107 * between user contexts and software interrupt processing, whereas the
108 * mini-semaphore synchronizes multiple users amongst themselves.
113 wait_queue_head_t wq;
115 * We express the mutex-alike socket_lock semantics
116 * to the lock validator by explicitly managing
117 * the slock as a lock variant (in addition to
120 #ifdef CONFIG_DEBUG_LOCK_ALLOC
121 struct lockdep_map dep_map;
130 * struct sock_common - minimal network layer representation of sockets
131 * @skc_daddr: Foreign IPv4 addr
132 * @skc_rcv_saddr: Bound local IPv4 addr
133 * @skc_hash: hash value used with various protocol lookup tables
134 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
135 * @skc_family: network address family
136 * @skc_state: Connection state
137 * @skc_reuse: %SO_REUSEADDR setting
138 * @skc_bound_dev_if: bound device index if != 0
139 * @skc_bind_node: bind hash linkage for various protocol lookup tables
140 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
141 * @skc_prot: protocol handlers inside a network family
142 * @skc_net: reference to the network namespace of this socket
143 * @skc_node: main hash linkage for various protocol lookup tables
144 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
145 * @skc_tx_queue_mapping: tx queue number for this connection
146 * @skc_refcnt: reference count
148 * This is the minimal network layer representation of sockets, the header
149 * for struct sock and struct inet_timewait_sock.
152 /* skc_daddr and skc_rcv_saddr must be grouped :
153 * cf INET_MATCH() and INET_TW_MATCH()
156 __be32 skc_rcv_saddr;
159 unsigned int skc_hash;
160 __u16 skc_u16hashes[2];
162 unsigned short skc_family;
163 volatile unsigned char skc_state;
164 unsigned char skc_reuse;
165 int skc_bound_dev_if;
167 struct hlist_node skc_bind_node;
168 struct hlist_nulls_node skc_portaddr_node;
170 struct proto *skc_prot;
175 * fields between dontcopy_begin/dontcopy_end
176 * are not copied in sock_copy()
179 int skc_dontcopy_begin[0];
182 struct hlist_node skc_node;
183 struct hlist_nulls_node skc_nulls_node;
185 int skc_tx_queue_mapping;
188 int skc_dontcopy_end[0];
194 * struct sock - network layer representation of sockets
195 * @__sk_common: shared layout with inet_timewait_sock
196 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
197 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
198 * @sk_lock: synchronizer
199 * @sk_rcvbuf: size of receive buffer in bytes
200 * @sk_wq: sock wait queue and async head
201 * @sk_rx_dst: receive input route used by early tcp demux
202 * @sk_dst_cache: destination cache
203 * @sk_dst_lock: destination cache lock
204 * @sk_policy: flow policy
205 * @sk_receive_queue: incoming packets
206 * @sk_wmem_alloc: transmit queue bytes committed
207 * @sk_write_queue: Packet sending queue
208 * @sk_async_wait_queue: DMA copied packets
209 * @sk_omem_alloc: "o" is "option" or "other"
210 * @sk_wmem_queued: persistent queue size
211 * @sk_forward_alloc: space allocated forward
212 * @sk_allocation: allocation mode
213 * @sk_sndbuf: size of send buffer in bytes
214 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
215 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
216 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
217 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
218 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
219 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
220 * @sk_gso_max_size: Maximum GSO segment size to build
221 * @sk_gso_max_segs: Maximum number of GSO segments
222 * @sk_lingertime: %SO_LINGER l_linger setting
223 * @sk_backlog: always used with the per-socket spinlock held
224 * @sk_callback_lock: used with the callbacks in the end of this struct
225 * @sk_error_queue: rarely used
226 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
227 * IPV6_ADDRFORM for instance)
228 * @sk_err: last error
229 * @sk_err_soft: errors that don't cause failure but are the cause of a
230 * persistent failure not just 'timed out'
231 * @sk_drops: raw/udp drops counter
232 * @sk_ack_backlog: current listen backlog
233 * @sk_max_ack_backlog: listen backlog set in listen()
234 * @sk_priority: %SO_PRIORITY setting
235 * @sk_cgrp_prioidx: socket group's priority map index
236 * @sk_type: socket type (%SOCK_STREAM, etc)
237 * @sk_protocol: which protocol this socket belongs in this network family
238 * @sk_peer_pid: &struct pid for this socket's peer
239 * @sk_peer_cred: %SO_PEERCRED setting
240 * @sk_rcvlowat: %SO_RCVLOWAT setting
241 * @sk_rcvtimeo: %SO_RCVTIMEO setting
242 * @sk_sndtimeo: %SO_SNDTIMEO setting
243 * @sk_rxhash: flow hash received from netif layer
244 * @sk_filter: socket filtering instructions
245 * @sk_protinfo: private area, net family specific, when not using slab
246 * @sk_timer: sock cleanup timer
247 * @sk_stamp: time stamp of last packet received
248 * @sk_socket: Identd and reporting IO signals
249 * @sk_user_data: RPC layer private data
250 * @sk_sndmsg_page: cached page for sendmsg
251 * @sk_sndmsg_off: cached offset for sendmsg
252 * @sk_peek_off: current peek_offset value
253 * @sk_send_head: front of stuff to transmit
254 * @sk_security: used by security modules
255 * @sk_mark: generic packet mark
256 * @sk_classid: this socket's cgroup classid
257 * @sk_cgrp: this socket's cgroup-specific proto data
258 * @sk_write_pending: a write to stream socket waits to start
259 * @sk_state_change: callback to indicate change in the state of the sock
260 * @sk_data_ready: callback to indicate there is data to be processed
261 * @sk_write_space: callback to indicate there is bf sending space available
262 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
263 * @sk_backlog_rcv: callback to process the backlog
264 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
268 * Now struct inet_timewait_sock also uses sock_common, so please just
269 * don't add nothing before this first member (__sk_common) --acme
271 struct sock_common __sk_common;
272 #define sk_node __sk_common.skc_node
273 #define sk_nulls_node __sk_common.skc_nulls_node
274 #define sk_refcnt __sk_common.skc_refcnt
275 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
277 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
278 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
279 #define sk_hash __sk_common.skc_hash
280 #define sk_family __sk_common.skc_family
281 #define sk_state __sk_common.skc_state
282 #define sk_reuse __sk_common.skc_reuse
283 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
284 #define sk_bind_node __sk_common.skc_bind_node
285 #define sk_prot __sk_common.skc_prot
286 #define sk_net __sk_common.skc_net
287 socket_lock_t sk_lock;
288 struct sk_buff_head sk_receive_queue;
290 * The backlog queue is special, it is always used with
291 * the per-socket spinlock held and requires low latency
292 * access. Therefore we special case it's implementation.
293 * Note : rmem_alloc is in this structure to fill a hole
294 * on 64bit arches, not because its logically part of
300 struct sk_buff *head;
301 struct sk_buff *tail;
303 #define sk_rmem_alloc sk_backlog.rmem_alloc
304 int sk_forward_alloc;
311 struct sk_filter __rcu *sk_filter;
312 struct socket_wq __rcu *sk_wq;
314 #ifdef CONFIG_NET_DMA
315 struct sk_buff_head sk_async_wait_queue;
319 struct xfrm_policy *sk_policy[2];
321 unsigned long sk_flags;
322 struct dst_entry *sk_rx_dst;
323 struct dst_entry *sk_dst_cache;
324 spinlock_t sk_dst_lock;
325 atomic_t sk_wmem_alloc;
326 atomic_t sk_omem_alloc;
328 struct sk_buff_head sk_write_queue;
329 kmemcheck_bitfield_begin(flags);
330 unsigned int sk_shutdown : 2,
335 kmemcheck_bitfield_end(flags);
338 netdev_features_t sk_route_caps;
339 netdev_features_t sk_route_nocaps;
341 unsigned int sk_gso_max_size;
344 unsigned long sk_lingertime;
345 struct sk_buff_head sk_error_queue;
346 struct proto *sk_prot_creator;
347 rwlock_t sk_callback_lock;
350 unsigned short sk_ack_backlog;
351 unsigned short sk_max_ack_backlog;
353 #ifdef CONFIG_CGROUPS
354 __u32 sk_cgrp_prioidx;
356 struct pid *sk_peer_pid;
357 const struct cred *sk_peer_cred;
361 struct timer_list sk_timer;
363 struct socket *sk_socket;
365 struct page *sk_sndmsg_page;
366 struct sk_buff *sk_send_head;
369 int sk_write_pending;
370 #ifdef CONFIG_SECURITY
375 struct cg_proto *sk_cgrp;
376 void (*sk_state_change)(struct sock *sk);
377 void (*sk_data_ready)(struct sock *sk, int bytes);
378 void (*sk_write_space)(struct sock *sk);
379 void (*sk_error_report)(struct sock *sk);
380 int (*sk_backlog_rcv)(struct sock *sk,
381 struct sk_buff *skb);
382 void (*sk_destruct)(struct sock *sk);
386 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
387 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
388 * on a socket means that the socket will reuse everybody else's port
389 * without looking at the other's sk_reuse value.
392 #define SK_NO_REUSE 0
393 #define SK_CAN_REUSE 1
394 #define SK_FORCE_REUSE 2
396 static inline int sk_peek_offset(struct sock *sk, int flags)
398 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
399 return sk->sk_peek_off;
404 static inline void sk_peek_offset_bwd(struct sock *sk, int val)
406 if (sk->sk_peek_off >= 0) {
407 if (sk->sk_peek_off >= val)
408 sk->sk_peek_off -= val;
414 static inline void sk_peek_offset_fwd(struct sock *sk, int val)
416 if (sk->sk_peek_off >= 0)
417 sk->sk_peek_off += val;
421 * Hashed lists helper routines
423 static inline struct sock *sk_entry(const struct hlist_node *node)
425 return hlist_entry(node, struct sock, sk_node);
428 static inline struct sock *__sk_head(const struct hlist_head *head)
430 return hlist_entry(head->first, struct sock, sk_node);
433 static inline struct sock *sk_head(const struct hlist_head *head)
435 return hlist_empty(head) ? NULL : __sk_head(head);
438 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
440 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
443 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
445 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
448 static inline struct sock *sk_next(const struct sock *sk)
450 return sk->sk_node.next ?
451 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
454 static inline struct sock *sk_nulls_next(const struct sock *sk)
456 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
457 hlist_nulls_entry(sk->sk_nulls_node.next,
458 struct sock, sk_nulls_node) :
462 static inline bool sk_unhashed(const struct sock *sk)
464 return hlist_unhashed(&sk->sk_node);
467 static inline bool sk_hashed(const struct sock *sk)
469 return !sk_unhashed(sk);
472 static inline void sk_node_init(struct hlist_node *node)
477 static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
482 static inline void __sk_del_node(struct sock *sk)
484 __hlist_del(&sk->sk_node);
487 /* NB: equivalent to hlist_del_init_rcu */
488 static inline bool __sk_del_node_init(struct sock *sk)
492 sk_node_init(&sk->sk_node);
498 /* Grab socket reference count. This operation is valid only
499 when sk is ALREADY grabbed f.e. it is found in hash table
500 or a list and the lookup is made under lock preventing hash table
504 static inline void sock_hold(struct sock *sk)
506 atomic_inc(&sk->sk_refcnt);
509 /* Ungrab socket in the context, which assumes that socket refcnt
510 cannot hit zero, f.e. it is true in context of any socketcall.
512 static inline void __sock_put(struct sock *sk)
514 atomic_dec(&sk->sk_refcnt);
517 static inline bool sk_del_node_init(struct sock *sk)
519 bool rc = __sk_del_node_init(sk);
522 /* paranoid for a while -acme */
523 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
528 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
530 static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
533 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
539 static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
541 bool rc = __sk_nulls_del_node_init_rcu(sk);
544 /* paranoid for a while -acme */
545 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
551 static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
553 hlist_add_head(&sk->sk_node, list);
556 static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
559 __sk_add_node(sk, list);
562 static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
565 hlist_add_head_rcu(&sk->sk_node, list);
568 static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
570 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
573 static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
576 __sk_nulls_add_node_rcu(sk, list);
579 static inline void __sk_del_bind_node(struct sock *sk)
581 __hlist_del(&sk->sk_bind_node);
584 static inline void sk_add_bind_node(struct sock *sk,
585 struct hlist_head *list)
587 hlist_add_head(&sk->sk_bind_node, list);
590 #define sk_for_each(__sk, node, list) \
591 hlist_for_each_entry(__sk, node, list, sk_node)
592 #define sk_for_each_rcu(__sk, node, list) \
593 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
594 #define sk_nulls_for_each(__sk, node, list) \
595 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
596 #define sk_nulls_for_each_rcu(__sk, node, list) \
597 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
598 #define sk_for_each_from(__sk, node) \
599 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
600 hlist_for_each_entry_from(__sk, node, sk_node)
601 #define sk_nulls_for_each_from(__sk, node) \
602 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
603 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
604 #define sk_for_each_safe(__sk, node, tmp, list) \
605 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
606 #define sk_for_each_bound(__sk, node, list) \
607 hlist_for_each_entry(__sk, node, list, sk_bind_node)
609 static inline struct user_namespace *sk_user_ns(struct sock *sk)
611 /* Careful only use this in a context where these parameters
612 * can not change and must all be valid, such as recvmsg from
615 return sk->sk_socket->file->f_cred->user_ns;
629 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
630 SOCK_DBG, /* %SO_DEBUG setting */
631 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
632 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
633 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
634 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
635 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
636 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
637 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
638 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
639 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
640 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
641 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
642 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
643 SOCK_FASYNC, /* fasync() active */
645 SOCK_ZEROCOPY, /* buffers from userspace */
646 SOCK_WIFI_STATUS, /* push wifi status to userspace */
647 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
648 * Will use last 4 bytes of packet sent from
649 * user-space instead.
653 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
655 nsk->sk_flags = osk->sk_flags;
658 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
660 __set_bit(flag, &sk->sk_flags);
663 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
665 __clear_bit(flag, &sk->sk_flags);
668 static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
670 return test_bit(flag, &sk->sk_flags);
674 extern struct static_key memalloc_socks;
675 static inline int sk_memalloc_socks(void)
677 return static_key_false(&memalloc_socks);
681 static inline int sk_memalloc_socks(void)
688 static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
690 return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
693 static inline void sk_acceptq_removed(struct sock *sk)
695 sk->sk_ack_backlog--;
698 static inline void sk_acceptq_added(struct sock *sk)
700 sk->sk_ack_backlog++;
703 static inline bool sk_acceptq_is_full(const struct sock *sk)
705 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
709 * Compute minimal free write space needed to queue new packets.
711 static inline int sk_stream_min_wspace(const struct sock *sk)
713 return sk->sk_wmem_queued >> 1;
716 static inline int sk_stream_wspace(const struct sock *sk)
718 return sk->sk_sndbuf - sk->sk_wmem_queued;
721 extern void sk_stream_write_space(struct sock *sk);
723 static inline bool sk_stream_memory_free(const struct sock *sk)
725 return sk->sk_wmem_queued < sk->sk_sndbuf;
728 /* OOB backlog add */
729 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
731 /* dont let skb dst not refcounted, we are going to leave rcu lock */
734 if (!sk->sk_backlog.tail)
735 sk->sk_backlog.head = skb;
737 sk->sk_backlog.tail->next = skb;
739 sk->sk_backlog.tail = skb;
744 * Take into account size of receive queue and backlog queue
745 * Do not take into account this skb truesize,
746 * to allow even a single big packet to come.
748 static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb,
751 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
753 return qsize > limit;
756 /* The per-socket spinlock must be held here. */
757 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
760 if (sk_rcvqueues_full(sk, skb, limit))
763 __sk_add_backlog(sk, skb);
764 sk->sk_backlog.len += skb->truesize;
768 extern int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
770 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
772 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
773 return __sk_backlog_rcv(sk, skb);
775 return sk->sk_backlog_rcv(sk, skb);
778 static inline void sock_rps_record_flow(const struct sock *sk)
781 struct rps_sock_flow_table *sock_flow_table;
784 sock_flow_table = rcu_dereference(rps_sock_flow_table);
785 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
790 static inline void sock_rps_reset_flow(const struct sock *sk)
793 struct rps_sock_flow_table *sock_flow_table;
796 sock_flow_table = rcu_dereference(rps_sock_flow_table);
797 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
802 static inline void sock_rps_save_rxhash(struct sock *sk,
803 const struct sk_buff *skb)
806 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
807 sock_rps_reset_flow(sk);
808 sk->sk_rxhash = skb->rxhash;
813 static inline void sock_rps_reset_rxhash(struct sock *sk)
816 sock_rps_reset_flow(sk);
821 #define sk_wait_event(__sk, __timeo, __condition) \
823 release_sock(__sk); \
824 __rc = __condition; \
826 *(__timeo) = schedule_timeout(*(__timeo)); \
829 __rc = __condition; \
833 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
834 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
835 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
836 extern int sk_stream_error(struct sock *sk, int flags, int err);
837 extern void sk_stream_kill_queues(struct sock *sk);
838 extern void sk_set_memalloc(struct sock *sk);
839 extern void sk_clear_memalloc(struct sock *sk);
841 extern int sk_wait_data(struct sock *sk, long *timeo);
843 struct request_sock_ops;
844 struct timewait_sock_ops;
845 struct inet_hashinfo;
849 /* Networking protocol blocks we attach to sockets.
850 * socket layer -> transport layer interface
851 * transport -> network interface is defined by struct inet_proto
854 void (*close)(struct sock *sk,
856 int (*connect)(struct sock *sk,
857 struct sockaddr *uaddr,
859 int (*disconnect)(struct sock *sk, int flags);
861 struct sock * (*accept)(struct sock *sk, int flags, int *err);
863 int (*ioctl)(struct sock *sk, int cmd,
865 int (*init)(struct sock *sk);
866 void (*destroy)(struct sock *sk);
867 void (*shutdown)(struct sock *sk, int how);
868 int (*setsockopt)(struct sock *sk, int level,
869 int optname, char __user *optval,
870 unsigned int optlen);
871 int (*getsockopt)(struct sock *sk, int level,
872 int optname, char __user *optval,
875 int (*compat_setsockopt)(struct sock *sk,
877 int optname, char __user *optval,
878 unsigned int optlen);
879 int (*compat_getsockopt)(struct sock *sk,
881 int optname, char __user *optval,
883 int (*compat_ioctl)(struct sock *sk,
884 unsigned int cmd, unsigned long arg);
886 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
887 struct msghdr *msg, size_t len);
888 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
890 size_t len, int noblock, int flags,
892 int (*sendpage)(struct sock *sk, struct page *page,
893 int offset, size_t size, int flags);
894 int (*bind)(struct sock *sk,
895 struct sockaddr *uaddr, int addr_len);
897 int (*backlog_rcv) (struct sock *sk,
898 struct sk_buff *skb);
900 void (*release_cb)(struct sock *sk);
901 void (*mtu_reduced)(struct sock *sk);
903 /* Keeping track of sk's, looking them up, and port selection methods. */
904 void (*hash)(struct sock *sk);
905 void (*unhash)(struct sock *sk);
906 void (*rehash)(struct sock *sk);
907 int (*get_port)(struct sock *sk, unsigned short snum);
908 void (*clear_sk)(struct sock *sk, int size);
910 /* Keeping track of sockets in use */
911 #ifdef CONFIG_PROC_FS
912 unsigned int inuse_idx;
915 /* Memory pressure */
916 void (*enter_memory_pressure)(struct sock *sk);
917 atomic_long_t *memory_allocated; /* Current allocated memory. */
918 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
920 * Pressure flag: try to collapse.
921 * Technical note: it is used by multiple contexts non atomically.
922 * All the __sk_mem_schedule() is of this nature: accounting
923 * is strict, actions are advisory and have some latency.
925 int *memory_pressure;
932 struct kmem_cache *slab;
933 unsigned int obj_size;
936 struct percpu_counter *orphan_count;
938 struct request_sock_ops *rsk_prot;
939 struct timewait_sock_ops *twsk_prot;
942 struct inet_hashinfo *hashinfo;
943 struct udp_table *udp_table;
944 struct raw_hashinfo *raw_hash;
947 struct module *owner;
951 struct list_head node;
952 #ifdef SOCK_REFCNT_DEBUG
955 #ifdef CONFIG_MEMCG_KMEM
957 * cgroup specific init/deinit functions. Called once for all
958 * protocols that implement it, from cgroups populate function.
959 * This function has to setup any files the protocol want to
960 * appear in the kmem cgroup filesystem.
962 int (*init_cgroup)(struct mem_cgroup *memcg,
963 struct cgroup_subsys *ss);
964 void (*destroy_cgroup)(struct mem_cgroup *memcg);
965 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
970 * Bits in struct cg_proto.flags
972 enum cg_proto_flags {
973 /* Currently active and new sockets should be assigned to cgroups */
975 /* It was ever activated; we must disarm static keys on destruction */
976 MEMCG_SOCK_ACTIVATED,
980 void (*enter_memory_pressure)(struct sock *sk);
981 struct res_counter *memory_allocated; /* Current allocated memory. */
982 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
983 int *memory_pressure;
987 * memcg field is used to find which memcg we belong directly
988 * Each memcg struct can hold more than one cg_proto, so container_of
991 * The elegant solution would be having an inverse function to
992 * proto_cgroup in struct proto, but that means polluting the structure
993 * for everybody, instead of just for memcg users.
995 struct mem_cgroup *memcg;
998 extern int proto_register(struct proto *prot, int alloc_slab);
999 extern void proto_unregister(struct proto *prot);
1001 static inline bool memcg_proto_active(struct cg_proto *cg_proto)
1003 return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
1006 static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
1008 return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
1011 #ifdef SOCK_REFCNT_DEBUG
1012 static inline void sk_refcnt_debug_inc(struct sock *sk)
1014 atomic_inc(&sk->sk_prot->socks);
1017 static inline void sk_refcnt_debug_dec(struct sock *sk)
1019 atomic_dec(&sk->sk_prot->socks);
1020 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1021 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1024 inline void sk_refcnt_debug_release(const struct sock *sk)
1026 if (atomic_read(&sk->sk_refcnt) != 1)
1027 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1028 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1030 #else /* SOCK_REFCNT_DEBUG */
1031 #define sk_refcnt_debug_inc(sk) do { } while (0)
1032 #define sk_refcnt_debug_dec(sk) do { } while (0)
1033 #define sk_refcnt_debug_release(sk) do { } while (0)
1034 #endif /* SOCK_REFCNT_DEBUG */
1036 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
1037 extern struct static_key memcg_socket_limit_enabled;
1038 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1039 struct cg_proto *cg_proto)
1041 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1043 #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
1045 #define mem_cgroup_sockets_enabled 0
1046 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1047 struct cg_proto *cg_proto)
1054 static inline bool sk_has_memory_pressure(const struct sock *sk)
1056 return sk->sk_prot->memory_pressure != NULL;
1059 static inline bool sk_under_memory_pressure(const struct sock *sk)
1061 if (!sk->sk_prot->memory_pressure)
1064 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1065 return !!*sk->sk_cgrp->memory_pressure;
1067 return !!*sk->sk_prot->memory_pressure;
1070 static inline void sk_leave_memory_pressure(struct sock *sk)
1072 int *memory_pressure = sk->sk_prot->memory_pressure;
1074 if (!memory_pressure)
1077 if (*memory_pressure)
1078 *memory_pressure = 0;
1080 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1081 struct cg_proto *cg_proto = sk->sk_cgrp;
1082 struct proto *prot = sk->sk_prot;
1084 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1085 if (*cg_proto->memory_pressure)
1086 *cg_proto->memory_pressure = 0;
1091 static inline void sk_enter_memory_pressure(struct sock *sk)
1093 if (!sk->sk_prot->enter_memory_pressure)
1096 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1097 struct cg_proto *cg_proto = sk->sk_cgrp;
1098 struct proto *prot = sk->sk_prot;
1100 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1101 cg_proto->enter_memory_pressure(sk);
1104 sk->sk_prot->enter_memory_pressure(sk);
1107 static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1109 long *prot = sk->sk_prot->sysctl_mem;
1110 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1111 prot = sk->sk_cgrp->sysctl_mem;
1115 static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1119 struct res_counter *fail;
1122 ret = res_counter_charge_nofail(prot->memory_allocated,
1123 amt << PAGE_SHIFT, &fail);
1125 *parent_status = OVER_LIMIT;
1128 static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1131 res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
1134 static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1137 ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
1138 return ret >> PAGE_SHIFT;
1142 sk_memory_allocated(const struct sock *sk)
1144 struct proto *prot = sk->sk_prot;
1145 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1146 return memcg_memory_allocated_read(sk->sk_cgrp);
1148 return atomic_long_read(prot->memory_allocated);
1152 sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
1154 struct proto *prot = sk->sk_prot;
1156 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1157 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1158 /* update the root cgroup regardless */
1159 atomic_long_add_return(amt, prot->memory_allocated);
1160 return memcg_memory_allocated_read(sk->sk_cgrp);
1163 return atomic_long_add_return(amt, prot->memory_allocated);
1167 sk_memory_allocated_sub(struct sock *sk, int amt)
1169 struct proto *prot = sk->sk_prot;
1171 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1172 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1174 atomic_long_sub(amt, prot->memory_allocated);
1177 static inline void sk_sockets_allocated_dec(struct sock *sk)
1179 struct proto *prot = sk->sk_prot;
1181 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1182 struct cg_proto *cg_proto = sk->sk_cgrp;
1184 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1185 percpu_counter_dec(cg_proto->sockets_allocated);
1188 percpu_counter_dec(prot->sockets_allocated);
1191 static inline void sk_sockets_allocated_inc(struct sock *sk)
1193 struct proto *prot = sk->sk_prot;
1195 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1196 struct cg_proto *cg_proto = sk->sk_cgrp;
1198 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1199 percpu_counter_inc(cg_proto->sockets_allocated);
1202 percpu_counter_inc(prot->sockets_allocated);
1206 sk_sockets_allocated_read_positive(struct sock *sk)
1208 struct proto *prot = sk->sk_prot;
1210 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1211 return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
1213 return percpu_counter_read_positive(prot->sockets_allocated);
1217 proto_sockets_allocated_sum_positive(struct proto *prot)
1219 return percpu_counter_sum_positive(prot->sockets_allocated);
1223 proto_memory_allocated(struct proto *prot)
1225 return atomic_long_read(prot->memory_allocated);
1229 proto_memory_pressure(struct proto *prot)
1231 if (!prot->memory_pressure)
1233 return !!*prot->memory_pressure;
1237 #ifdef CONFIG_PROC_FS
1238 /* Called with local bh disabled */
1239 extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1240 extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
1242 static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
1249 /* With per-bucket locks this operation is not-atomic, so that
1250 * this version is not worse.
1252 static inline void __sk_prot_rehash(struct sock *sk)
1254 sk->sk_prot->unhash(sk);
1255 sk->sk_prot->hash(sk);
1258 void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1260 /* About 10 seconds */
1261 #define SOCK_DESTROY_TIME (10*HZ)
1263 /* Sockets 0-1023 can't be bound to unless you are superuser */
1264 #define PROT_SOCK 1024
1266 #define SHUTDOWN_MASK 3
1267 #define RCV_SHUTDOWN 1
1268 #define SEND_SHUTDOWN 2
1270 #define SOCK_SNDBUF_LOCK 1
1271 #define SOCK_RCVBUF_LOCK 2
1272 #define SOCK_BINDADDR_LOCK 4
1273 #define SOCK_BINDPORT_LOCK 8
1275 /* sock_iocb: used to kick off async processing of socket ios */
1277 struct list_head list;
1281 struct socket *sock;
1283 struct scm_cookie *scm;
1284 struct msghdr *msg, async_msg;
1285 struct kiocb *kiocb;
1288 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1290 return (struct sock_iocb *)iocb->private;
1293 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1298 struct socket_alloc {
1299 struct socket socket;
1300 struct inode vfs_inode;
1303 static inline struct socket *SOCKET_I(struct inode *inode)
1305 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1308 static inline struct inode *SOCK_INODE(struct socket *socket)
1310 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1314 * Functions for memory accounting
1316 extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
1317 extern void __sk_mem_reclaim(struct sock *sk);
1319 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1320 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1321 #define SK_MEM_SEND 0
1322 #define SK_MEM_RECV 1
1324 static inline int sk_mem_pages(int amt)
1326 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1329 static inline bool sk_has_account(struct sock *sk)
1331 /* return true if protocol supports memory accounting */
1332 return !!sk->sk_prot->memory_allocated;
1335 static inline bool sk_wmem_schedule(struct sock *sk, int size)
1337 if (!sk_has_account(sk))
1339 return size <= sk->sk_forward_alloc ||
1340 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1344 sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, unsigned int size)
1346 if (!sk_has_account(sk))
1348 return size<= sk->sk_forward_alloc ||
1349 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1350 skb_pfmemalloc(skb);
1353 static inline void sk_mem_reclaim(struct sock *sk)
1355 if (!sk_has_account(sk))
1357 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1358 __sk_mem_reclaim(sk);
1361 static inline void sk_mem_reclaim_partial(struct sock *sk)
1363 if (!sk_has_account(sk))
1365 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1366 __sk_mem_reclaim(sk);
1369 static inline void sk_mem_charge(struct sock *sk, int size)
1371 if (!sk_has_account(sk))
1373 sk->sk_forward_alloc -= size;
1376 static inline void sk_mem_uncharge(struct sock *sk, int size)
1378 if (!sk_has_account(sk))
1380 sk->sk_forward_alloc += size;
1383 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1385 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1386 sk->sk_wmem_queued -= skb->truesize;
1387 sk_mem_uncharge(sk, skb->truesize);
1391 /* Used by processes to "lock" a socket state, so that
1392 * interrupts and bottom half handlers won't change it
1393 * from under us. It essentially blocks any incoming
1394 * packets, so that we won't get any new data or any
1395 * packets that change the state of the socket.
1397 * While locked, BH processing will add new packets to
1398 * the backlog queue. This queue is processed by the
1399 * owner of the socket lock right before it is released.
1401 * Since ~2.3.5 it is also exclusive sleep lock serializing
1402 * accesses from user process context.
1404 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1407 * Macro so as to not evaluate some arguments when
1408 * lockdep is not enabled.
1410 * Mark both the sk_lock and the sk_lock.slock as a
1411 * per-address-family lock class.
1413 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1415 sk->sk_lock.owned = 0; \
1416 init_waitqueue_head(&sk->sk_lock.wq); \
1417 spin_lock_init(&(sk)->sk_lock.slock); \
1418 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1419 sizeof((sk)->sk_lock)); \
1420 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1422 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1425 extern void lock_sock_nested(struct sock *sk, int subclass);
1427 static inline void lock_sock(struct sock *sk)
1429 lock_sock_nested(sk, 0);
1432 extern void release_sock(struct sock *sk);
1434 /* BH context may only use the following locking interface. */
1435 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1436 #define bh_lock_sock_nested(__sk) \
1437 spin_lock_nested(&((__sk)->sk_lock.slock), \
1438 SINGLE_DEPTH_NESTING)
1439 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1441 extern bool lock_sock_fast(struct sock *sk);
1443 * unlock_sock_fast - complement of lock_sock_fast
1447 * fast unlock socket for user context.
1448 * If slow mode is on, we call regular release_sock()
1450 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1455 spin_unlock_bh(&sk->sk_lock.slock);
1459 extern struct sock *sk_alloc(struct net *net, int family,
1461 struct proto *prot);
1462 extern void sk_free(struct sock *sk);
1463 extern void sk_release_kernel(struct sock *sk);
1464 extern struct sock *sk_clone_lock(const struct sock *sk,
1465 const gfp_t priority);
1467 extern struct sk_buff *sock_wmalloc(struct sock *sk,
1468 unsigned long size, int force,
1470 extern struct sk_buff *sock_rmalloc(struct sock *sk,
1471 unsigned long size, int force,
1473 extern void sock_wfree(struct sk_buff *skb);
1474 extern void sock_rfree(struct sk_buff *skb);
1475 extern void sock_edemux(struct sk_buff *skb);
1477 extern int sock_setsockopt(struct socket *sock, int level,
1478 int op, char __user *optval,
1479 unsigned int optlen);
1481 extern int sock_getsockopt(struct socket *sock, int level,
1482 int op, char __user *optval,
1483 int __user *optlen);
1484 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1488 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1489 unsigned long header_len,
1490 unsigned long data_len,
1493 extern void *sock_kmalloc(struct sock *sk, int size,
1495 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1496 extern void sk_send_sigurg(struct sock *sk);
1498 #ifdef CONFIG_CGROUPS
1499 extern void sock_update_classid(struct sock *sk);
1501 static inline void sock_update_classid(struct sock *sk)
1507 * Functions to fill in entries in struct proto_ops when a protocol
1508 * does not implement a particular function.
1510 extern int sock_no_bind(struct socket *,
1511 struct sockaddr *, int);
1512 extern int sock_no_connect(struct socket *,
1513 struct sockaddr *, int, int);
1514 extern int sock_no_socketpair(struct socket *,
1516 extern int sock_no_accept(struct socket *,
1517 struct socket *, int);
1518 extern int sock_no_getname(struct socket *,
1519 struct sockaddr *, int *, int);
1520 extern unsigned int sock_no_poll(struct file *, struct socket *,
1521 struct poll_table_struct *);
1522 extern int sock_no_ioctl(struct socket *, unsigned int,
1524 extern int sock_no_listen(struct socket *, int);
1525 extern int sock_no_shutdown(struct socket *, int);
1526 extern int sock_no_getsockopt(struct socket *, int , int,
1527 char __user *, int __user *);
1528 extern int sock_no_setsockopt(struct socket *, int, int,
1529 char __user *, unsigned int);
1530 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1531 struct msghdr *, size_t);
1532 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1533 struct msghdr *, size_t, int);
1534 extern int sock_no_mmap(struct file *file,
1535 struct socket *sock,
1536 struct vm_area_struct *vma);
1537 extern ssize_t sock_no_sendpage(struct socket *sock,
1539 int offset, size_t size,
1543 * Functions to fill in entries in struct proto_ops when a protocol
1544 * uses the inet style.
1546 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1547 char __user *optval, int __user *optlen);
1548 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1549 struct msghdr *msg, size_t size, int flags);
1550 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
1551 char __user *optval, unsigned int optlen);
1552 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1553 int optname, char __user *optval, int __user *optlen);
1554 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
1555 int optname, char __user *optval, unsigned int optlen);
1557 extern void sk_common_release(struct sock *sk);
1560 * Default socket callbacks and setup code
1563 /* Initialise core socket variables */
1564 extern void sock_init_data(struct socket *sock, struct sock *sk);
1566 extern void sk_filter_release_rcu(struct rcu_head *rcu);
1569 * sk_filter_release - release a socket filter
1570 * @fp: filter to remove
1572 * Remove a filter from a socket and release its resources.
1575 static inline void sk_filter_release(struct sk_filter *fp)
1577 if (atomic_dec_and_test(&fp->refcnt))
1578 call_rcu(&fp->rcu, sk_filter_release_rcu);
1581 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1583 unsigned int size = sk_filter_len(fp);
1585 atomic_sub(size, &sk->sk_omem_alloc);
1586 sk_filter_release(fp);
1589 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1591 atomic_inc(&fp->refcnt);
1592 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1596 * Socket reference counting postulates.
1598 * * Each user of socket SHOULD hold a reference count.
1599 * * Each access point to socket (an hash table bucket, reference from a list,
1600 * running timer, skb in flight MUST hold a reference count.
1601 * * When reference count hits 0, it means it will never increase back.
1602 * * When reference count hits 0, it means that no references from
1603 * outside exist to this socket and current process on current CPU
1604 * is last user and may/should destroy this socket.
1605 * * sk_free is called from any context: process, BH, IRQ. When
1606 * it is called, socket has no references from outside -> sk_free
1607 * may release descendant resources allocated by the socket, but
1608 * to the time when it is called, socket is NOT referenced by any
1609 * hash tables, lists etc.
1610 * * Packets, delivered from outside (from network or from another process)
1611 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1612 * when they sit in queue. Otherwise, packets will leak to hole, when
1613 * socket is looked up by one cpu and unhasing is made by another CPU.
1614 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1615 * (leak to backlog). Packet socket does all the processing inside
1616 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1617 * use separate SMP lock, so that they are prone too.
1620 /* Ungrab socket and destroy it, if it was the last reference. */
1621 static inline void sock_put(struct sock *sk)
1623 if (atomic_dec_and_test(&sk->sk_refcnt))
1627 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1630 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1632 sk->sk_tx_queue_mapping = tx_queue;
1635 static inline void sk_tx_queue_clear(struct sock *sk)
1637 sk->sk_tx_queue_mapping = -1;
1640 static inline int sk_tx_queue_get(const struct sock *sk)
1642 return sk ? sk->sk_tx_queue_mapping : -1;
1645 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1647 sk_tx_queue_clear(sk);
1648 sk->sk_socket = sock;
1651 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1653 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1654 return &rcu_dereference_raw(sk->sk_wq)->wait;
1656 /* Detach socket from process context.
1657 * Announce socket dead, detach it from wait queue and inode.
1658 * Note that parent inode held reference count on this struct sock,
1659 * we do not release it in this function, because protocol
1660 * probably wants some additional cleanups or even continuing
1661 * to work with this socket (TCP).
1663 static inline void sock_orphan(struct sock *sk)
1665 write_lock_bh(&sk->sk_callback_lock);
1666 sock_set_flag(sk, SOCK_DEAD);
1667 sk_set_socket(sk, NULL);
1669 write_unlock_bh(&sk->sk_callback_lock);
1672 static inline void sock_graft(struct sock *sk, struct socket *parent)
1674 write_lock_bh(&sk->sk_callback_lock);
1675 sk->sk_wq = parent->wq;
1677 sk_set_socket(sk, parent);
1678 security_sock_graft(sk, parent);
1679 write_unlock_bh(&sk->sk_callback_lock);
1682 extern kuid_t sock_i_uid(struct sock *sk);
1683 extern unsigned long sock_i_ino(struct sock *sk);
1685 static inline struct dst_entry *
1686 __sk_dst_get(struct sock *sk)
1688 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
1689 lockdep_is_held(&sk->sk_lock.slock));
1692 static inline struct dst_entry *
1693 sk_dst_get(struct sock *sk)
1695 struct dst_entry *dst;
1698 dst = rcu_dereference(sk->sk_dst_cache);
1705 extern void sk_reset_txq(struct sock *sk);
1707 static inline void dst_negative_advice(struct sock *sk)
1709 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1711 if (dst && dst->ops->negative_advice) {
1712 ndst = dst->ops->negative_advice(dst);
1715 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1722 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1724 struct dst_entry *old_dst;
1726 sk_tx_queue_clear(sk);
1728 * This can be called while sk is owned by the caller only,
1729 * with no state that can be checked in a rcu_dereference_check() cond
1731 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1732 rcu_assign_pointer(sk->sk_dst_cache, dst);
1733 dst_release(old_dst);
1737 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1739 spin_lock(&sk->sk_dst_lock);
1740 __sk_dst_set(sk, dst);
1741 spin_unlock(&sk->sk_dst_lock);
1745 __sk_dst_reset(struct sock *sk)
1747 __sk_dst_set(sk, NULL);
1751 sk_dst_reset(struct sock *sk)
1753 spin_lock(&sk->sk_dst_lock);
1755 spin_unlock(&sk->sk_dst_lock);
1758 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1760 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1762 static inline bool sk_can_gso(const struct sock *sk)
1764 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1767 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1769 static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
1771 sk->sk_route_nocaps |= flags;
1772 sk->sk_route_caps &= ~flags;
1775 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1776 char __user *from, char *to,
1777 int copy, int offset)
1779 if (skb->ip_summed == CHECKSUM_NONE) {
1781 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1784 skb->csum = csum_block_add(skb->csum, csum, offset);
1785 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1786 if (!access_ok(VERIFY_READ, from, copy) ||
1787 __copy_from_user_nocache(to, from, copy))
1789 } else if (copy_from_user(to, from, copy))
1795 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1796 char __user *from, int copy)
1798 int err, offset = skb->len;
1800 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1803 __skb_trim(skb, offset);
1808 static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1809 struct sk_buff *skb,
1815 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1821 skb->data_len += copy;
1822 skb->truesize += copy;
1823 sk->sk_wmem_queued += copy;
1824 sk_mem_charge(sk, copy);
1828 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1829 struct sk_buff *skb, struct page *page,
1832 if (skb->ip_summed == CHECKSUM_NONE) {
1834 __wsum csum = csum_and_copy_from_user(from,
1835 page_address(page) + off,
1839 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1840 } else if (copy_from_user(page_address(page) + off, from, copy))
1844 skb->data_len += copy;
1845 skb->truesize += copy;
1846 sk->sk_wmem_queued += copy;
1847 sk_mem_charge(sk, copy);
1852 * sk_wmem_alloc_get - returns write allocations
1855 * Returns sk_wmem_alloc minus initial offset of one
1857 static inline int sk_wmem_alloc_get(const struct sock *sk)
1859 return atomic_read(&sk->sk_wmem_alloc) - 1;
1863 * sk_rmem_alloc_get - returns read allocations
1866 * Returns sk_rmem_alloc
1868 static inline int sk_rmem_alloc_get(const struct sock *sk)
1870 return atomic_read(&sk->sk_rmem_alloc);
1874 * sk_has_allocations - check if allocations are outstanding
1877 * Returns true if socket has write or read allocations
1879 static inline bool sk_has_allocations(const struct sock *sk)
1881 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1885 * wq_has_sleeper - check if there are any waiting processes
1886 * @wq: struct socket_wq
1888 * Returns true if socket_wq has waiting processes
1890 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1891 * barrier call. They were added due to the race found within the tcp code.
1893 * Consider following tcp code paths:
1897 * sys_select receive packet
1899 * __add_wait_queue update tp->rcv_nxt
1901 * tp->rcv_nxt check sock_def_readable
1903 * schedule rcu_read_lock();
1904 * wq = rcu_dereference(sk->sk_wq);
1905 * if (wq && waitqueue_active(&wq->wait))
1906 * wake_up_interruptible(&wq->wait)
1910 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1911 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1912 * could then endup calling schedule and sleep forever if there are no more
1913 * data on the socket.
1916 static inline bool wq_has_sleeper(struct socket_wq *wq)
1918 /* We need to be sure we are in sync with the
1919 * add_wait_queue modifications to the wait queue.
1921 * This memory barrier is paired in the sock_poll_wait.
1924 return wq && waitqueue_active(&wq->wait);
1928 * sock_poll_wait - place memory barrier behind the poll_wait call.
1930 * @wait_address: socket wait queue
1933 * See the comments in the wq_has_sleeper function.
1935 static inline void sock_poll_wait(struct file *filp,
1936 wait_queue_head_t *wait_address, poll_table *p)
1938 if (!poll_does_not_wait(p) && wait_address) {
1939 poll_wait(filp, wait_address, p);
1940 /* We need to be sure we are in sync with the
1941 * socket flags modification.
1943 * This memory barrier is paired in the wq_has_sleeper.
1950 * Queue a received datagram if it will fit. Stream and sequenced
1951 * protocols can't normally use this as they need to fit buffers in
1952 * and play with them.
1954 * Inlined as it's very short and called for pretty much every
1955 * packet ever received.
1958 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1962 skb->destructor = sock_wfree;
1964 * We used to take a refcount on sk, but following operation
1965 * is enough to guarantee sk_free() wont free this sock until
1966 * all in-flight packets are completed
1968 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1971 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1975 skb->destructor = sock_rfree;
1976 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1977 sk_mem_charge(sk, skb->truesize);
1980 extern void sk_reset_timer(struct sock *sk, struct timer_list *timer,
1981 unsigned long expires);
1983 extern void sk_stop_timer(struct sock *sk, struct timer_list *timer);
1985 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1987 extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1990 * Recover an error report and clear atomically
1993 static inline int sock_error(struct sock *sk)
1996 if (likely(!sk->sk_err))
1998 err = xchg(&sk->sk_err, 0);
2002 static inline unsigned long sock_wspace(struct sock *sk)
2006 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2007 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
2014 static inline void sk_wake_async(struct sock *sk, int how, int band)
2016 if (sock_flag(sk, SOCK_FASYNC))
2017 sock_wake_async(sk->sk_socket, how, band);
2020 #define SOCK_MIN_SNDBUF 2048
2022 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
2023 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
2025 #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
2027 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2029 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
2030 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
2031 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
2035 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
2037 static inline struct page *sk_stream_alloc_page(struct sock *sk)
2039 struct page *page = NULL;
2041 page = alloc_pages(sk->sk_allocation, 0);
2043 sk_enter_memory_pressure(sk);
2044 sk_stream_moderate_sndbuf(sk);
2050 * Default write policy as shown to user space via poll/select/SIGIO
2052 static inline bool sock_writeable(const struct sock *sk)
2054 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
2057 static inline gfp_t gfp_any(void)
2059 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
2062 static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
2064 return noblock ? 0 : sk->sk_rcvtimeo;
2067 static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
2069 return noblock ? 0 : sk->sk_sndtimeo;
2072 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2074 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2077 /* Alas, with timeout socket operations are not restartable.
2078 * Compare this to poll().
2080 static inline int sock_intr_errno(long timeo)
2082 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2085 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2086 struct sk_buff *skb);
2087 extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2088 struct sk_buff *skb);
2091 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2093 ktime_t kt = skb->tstamp;
2094 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
2097 * generate control messages if
2098 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2099 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2100 * - software time stamp available and wanted
2101 * (SOCK_TIMESTAMPING_SOFTWARE)
2102 * - hardware time stamps available and wanted
2103 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2104 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2106 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2107 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2108 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2109 (hwtstamps->hwtstamp.tv64 &&
2110 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2111 (hwtstamps->syststamp.tv64 &&
2112 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
2113 __sock_recv_timestamp(msg, sk, skb);
2117 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2118 __sock_recv_wifi_status(msg, sk, skb);
2121 extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2122 struct sk_buff *skb);
2124 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2125 struct sk_buff *skb)
2127 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2128 (1UL << SOCK_RCVTSTAMP) | \
2129 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2130 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
2131 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
2132 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2134 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2135 __sock_recv_ts_and_drops(msg, sk, skb);
2137 sk->sk_stamp = skb->tstamp;
2141 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2142 * @sk: socket sending this packet
2143 * @tx_flags: filled with instructions for time stamping
2145 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
2146 * parameters are invalid.
2148 extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
2151 * sk_eat_skb - Release a skb if it is no longer needed
2152 * @sk: socket to eat this skb from
2153 * @skb: socket buffer to eat
2154 * @copied_early: flag indicating whether DMA operations copied this data early
2156 * This routine must be called with interrupts disabled or with the socket
2157 * locked so that the sk_buff queue operation is ok.
2159 #ifdef CONFIG_NET_DMA
2160 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
2162 __skb_unlink(skb, &sk->sk_receive_queue);
2166 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2169 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
2171 __skb_unlink(skb, &sk->sk_receive_queue);
2177 struct net *sock_net(const struct sock *sk)
2179 return read_pnet(&sk->sk_net);
2183 void sock_net_set(struct sock *sk, struct net *net)
2185 write_pnet(&sk->sk_net, net);
2189 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
2190 * They should not hold a reference to a namespace in order to allow
2192 * Sockets after sk_change_net should be released using sk_release_kernel
2194 static inline void sk_change_net(struct sock *sk, struct net *net)
2196 put_net(sock_net(sk));
2197 sock_net_set(sk, hold_net(net));
2200 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2203 struct sock *sk = skb->sk;
2205 skb->destructor = NULL;
2212 extern void sock_enable_timestamp(struct sock *sk, int flag);
2213 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
2214 extern int sock_get_timestampns(struct sock *, struct timespec __user *);
2217 * Enable debug/info messages
2219 extern int net_msg_warn;
2220 #define NETDEBUG(fmt, args...) \
2221 do { if (net_msg_warn) printk(fmt,##args); } while (0)
2223 #define LIMIT_NETDEBUG(fmt, args...) \
2224 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
2226 extern __u32 sysctl_wmem_max;
2227 extern __u32 sysctl_rmem_max;
2229 extern void sk_init(void);
2231 extern int sysctl_optmem_max;
2233 extern __u32 sysctl_wmem_default;
2234 extern __u32 sysctl_rmem_default;
2236 #endif /* _SOCK_H */