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/lockdep.h>
50 #include <linux/netdevice.h>
51 #include <linux/skbuff.h> /* struct sk_buff */
53 #include <linux/security.h>
54 #include <linux/slab.h>
55 #include <linux/uaccess.h>
56 #include <linux/memcontrol.h>
57 #include <linux/res_counter.h>
59 #include <linux/filter.h>
60 #include <linux/rculist_nulls.h>
61 #include <linux/poll.h>
63 #include <linux/atomic.h>
65 #include <net/checksum.h>
70 int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss);
71 void mem_cgroup_sockets_destroy(struct cgroup *cgrp, struct cgroup_subsys *ss);
74 int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss)
79 void mem_cgroup_sockets_destroy(struct cgroup *cgrp, struct cgroup_subsys *ss)
84 * This structure really needs to be cleaned up.
85 * Most of it is for TCP, and not used by any of
86 * the other protocols.
89 /* Define this to get the SOCK_DBG debugging facility. */
90 #define SOCK_DEBUGGING
92 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
93 printk(KERN_DEBUG msg); } while (0)
95 /* Validate arguments and do nothing */
96 static inline __printf(2, 3)
97 void SOCK_DEBUG(struct sock *sk, const char *msg, ...)
102 /* This is the per-socket lock. The spinlock provides a synchronization
103 * between user contexts and software interrupt processing, whereas the
104 * mini-semaphore synchronizes multiple users amongst themselves.
109 wait_queue_head_t wq;
111 * We express the mutex-alike socket_lock semantics
112 * to the lock validator by explicitly managing
113 * the slock as a lock variant (in addition to
116 #ifdef CONFIG_DEBUG_LOCK_ALLOC
117 struct lockdep_map dep_map;
126 * struct sock_common - minimal network layer representation of sockets
127 * @skc_daddr: Foreign IPv4 addr
128 * @skc_rcv_saddr: Bound local IPv4 addr
129 * @skc_hash: hash value used with various protocol lookup tables
130 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
131 * @skc_family: network address family
132 * @skc_state: Connection state
133 * @skc_reuse: %SO_REUSEADDR setting
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_node: main hash linkage for various protocol lookup tables
140 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
141 * @skc_tx_queue_mapping: tx queue number for this connection
142 * @skc_refcnt: reference count
144 * This is the minimal network layer representation of sockets, the header
145 * for struct sock and struct inet_timewait_sock.
148 /* skc_daddr and skc_rcv_saddr must be grouped :
149 * cf INET_MATCH() and INET_TW_MATCH()
152 __be32 skc_rcv_saddr;
155 unsigned int skc_hash;
156 __u16 skc_u16hashes[2];
158 unsigned short skc_family;
159 volatile unsigned char skc_state;
160 unsigned char skc_reuse;
161 int skc_bound_dev_if;
163 struct hlist_node skc_bind_node;
164 struct hlist_nulls_node skc_portaddr_node;
166 struct proto *skc_prot;
171 * fields between dontcopy_begin/dontcopy_end
172 * are not copied in sock_copy()
175 int skc_dontcopy_begin[0];
178 struct hlist_node skc_node;
179 struct hlist_nulls_node skc_nulls_node;
181 int skc_tx_queue_mapping;
184 int skc_dontcopy_end[0];
190 * struct sock - network layer representation of sockets
191 * @__sk_common: shared layout with inet_timewait_sock
192 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
193 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
194 * @sk_lock: synchronizer
195 * @sk_rcvbuf: size of receive buffer in bytes
196 * @sk_wq: sock wait queue and async head
197 * @sk_dst_cache: destination cache
198 * @sk_dst_lock: destination cache lock
199 * @sk_policy: flow policy
200 * @sk_receive_queue: incoming packets
201 * @sk_wmem_alloc: transmit queue bytes committed
202 * @sk_write_queue: Packet sending queue
203 * @sk_async_wait_queue: DMA copied packets
204 * @sk_omem_alloc: "o" is "option" or "other"
205 * @sk_wmem_queued: persistent queue size
206 * @sk_forward_alloc: space allocated forward
207 * @sk_allocation: allocation mode
208 * @sk_sndbuf: size of send buffer in bytes
209 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
210 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
211 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
212 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
213 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
214 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
215 * @sk_gso_max_size: Maximum GSO segment size to build
216 * @sk_lingertime: %SO_LINGER l_linger setting
217 * @sk_backlog: always used with the per-socket spinlock held
218 * @sk_callback_lock: used with the callbacks in the end of this struct
219 * @sk_error_queue: rarely used
220 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
221 * IPV6_ADDRFORM for instance)
222 * @sk_err: last error
223 * @sk_err_soft: errors that don't cause failure but are the cause of a
224 * persistent failure not just 'timed out'
225 * @sk_drops: raw/udp drops counter
226 * @sk_ack_backlog: current listen backlog
227 * @sk_max_ack_backlog: listen backlog set in listen()
228 * @sk_priority: %SO_PRIORITY setting
229 * @sk_type: socket type (%SOCK_STREAM, etc)
230 * @sk_protocol: which protocol this socket belongs in this network family
231 * @sk_peer_pid: &struct pid for this socket's peer
232 * @sk_peer_cred: %SO_PEERCRED setting
233 * @sk_rcvlowat: %SO_RCVLOWAT setting
234 * @sk_rcvtimeo: %SO_RCVTIMEO setting
235 * @sk_sndtimeo: %SO_SNDTIMEO setting
236 * @sk_rxhash: flow hash received from netif layer
237 * @sk_filter: socket filtering instructions
238 * @sk_protinfo: private area, net family specific, when not using slab
239 * @sk_timer: sock cleanup timer
240 * @sk_stamp: time stamp of last packet received
241 * @sk_socket: Identd and reporting IO signals
242 * @sk_user_data: RPC layer private data
243 * @sk_sndmsg_page: cached page for sendmsg
244 * @sk_sndmsg_off: cached offset for sendmsg
245 * @sk_send_head: front of stuff to transmit
246 * @sk_security: used by security modules
247 * @sk_mark: generic packet mark
248 * @sk_classid: this socket's cgroup classid
249 * @sk_cgrp: this socket's cgroup-specific proto data
250 * @sk_write_pending: a write to stream socket waits to start
251 * @sk_state_change: callback to indicate change in the state of the sock
252 * @sk_data_ready: callback to indicate there is data to be processed
253 * @sk_write_space: callback to indicate there is bf sending space available
254 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
255 * @sk_backlog_rcv: callback to process the backlog
256 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
260 * Now struct inet_timewait_sock also uses sock_common, so please just
261 * don't add nothing before this first member (__sk_common) --acme
263 struct sock_common __sk_common;
264 #define sk_node __sk_common.skc_node
265 #define sk_nulls_node __sk_common.skc_nulls_node
266 #define sk_refcnt __sk_common.skc_refcnt
267 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
269 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
270 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
271 #define sk_hash __sk_common.skc_hash
272 #define sk_family __sk_common.skc_family
273 #define sk_state __sk_common.skc_state
274 #define sk_reuse __sk_common.skc_reuse
275 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
276 #define sk_bind_node __sk_common.skc_bind_node
277 #define sk_prot __sk_common.skc_prot
278 #define sk_net __sk_common.skc_net
279 socket_lock_t sk_lock;
280 struct sk_buff_head sk_receive_queue;
282 * The backlog queue is special, it is always used with
283 * the per-socket spinlock held and requires low latency
284 * access. Therefore we special case it's implementation.
285 * Note : rmem_alloc is in this structure to fill a hole
286 * on 64bit arches, not because its logically part of
292 struct sk_buff *head;
293 struct sk_buff *tail;
295 #define sk_rmem_alloc sk_backlog.rmem_alloc
296 int sk_forward_alloc;
303 struct sk_filter __rcu *sk_filter;
304 struct socket_wq __rcu *sk_wq;
306 #ifdef CONFIG_NET_DMA
307 struct sk_buff_head sk_async_wait_queue;
311 struct xfrm_policy *sk_policy[2];
313 unsigned long sk_flags;
314 struct dst_entry *sk_dst_cache;
315 spinlock_t sk_dst_lock;
316 atomic_t sk_wmem_alloc;
317 atomic_t sk_omem_alloc;
319 struct sk_buff_head sk_write_queue;
320 kmemcheck_bitfield_begin(flags);
321 unsigned int sk_shutdown : 2,
326 kmemcheck_bitfield_end(flags);
329 netdev_features_t sk_route_caps;
330 netdev_features_t sk_route_nocaps;
332 unsigned int sk_gso_max_size;
334 unsigned long sk_lingertime;
335 struct sk_buff_head sk_error_queue;
336 struct proto *sk_prot_creator;
337 rwlock_t sk_callback_lock;
340 unsigned short sk_ack_backlog;
341 unsigned short sk_max_ack_backlog;
343 #ifdef CONFIG_CGROUPS
344 __u32 sk_cgrp_prioidx;
346 struct pid *sk_peer_pid;
347 const struct cred *sk_peer_cred;
351 struct timer_list sk_timer;
353 struct socket *sk_socket;
355 struct page *sk_sndmsg_page;
356 struct sk_buff *sk_send_head;
358 int sk_write_pending;
359 #ifdef CONFIG_SECURITY
364 struct cg_proto *sk_cgrp;
365 void (*sk_state_change)(struct sock *sk);
366 void (*sk_data_ready)(struct sock *sk, int bytes);
367 void (*sk_write_space)(struct sock *sk);
368 void (*sk_error_report)(struct sock *sk);
369 int (*sk_backlog_rcv)(struct sock *sk,
370 struct sk_buff *skb);
371 void (*sk_destruct)(struct sock *sk);
375 * Hashed lists helper routines
377 static inline struct sock *sk_entry(const struct hlist_node *node)
379 return hlist_entry(node, struct sock, sk_node);
382 static inline struct sock *__sk_head(const struct hlist_head *head)
384 return hlist_entry(head->first, struct sock, sk_node);
387 static inline struct sock *sk_head(const struct hlist_head *head)
389 return hlist_empty(head) ? NULL : __sk_head(head);
392 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
394 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
397 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
399 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
402 static inline struct sock *sk_next(const struct sock *sk)
404 return sk->sk_node.next ?
405 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
408 static inline struct sock *sk_nulls_next(const struct sock *sk)
410 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
411 hlist_nulls_entry(sk->sk_nulls_node.next,
412 struct sock, sk_nulls_node) :
416 static inline int sk_unhashed(const struct sock *sk)
418 return hlist_unhashed(&sk->sk_node);
421 static inline int sk_hashed(const struct sock *sk)
423 return !sk_unhashed(sk);
426 static __inline__ void sk_node_init(struct hlist_node *node)
431 static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
436 static __inline__ void __sk_del_node(struct sock *sk)
438 __hlist_del(&sk->sk_node);
441 /* NB: equivalent to hlist_del_init_rcu */
442 static __inline__ int __sk_del_node_init(struct sock *sk)
446 sk_node_init(&sk->sk_node);
452 /* Grab socket reference count. This operation is valid only
453 when sk is ALREADY grabbed f.e. it is found in hash table
454 or a list and the lookup is made under lock preventing hash table
458 static inline void sock_hold(struct sock *sk)
460 atomic_inc(&sk->sk_refcnt);
463 /* Ungrab socket in the context, which assumes that socket refcnt
464 cannot hit zero, f.e. it is true in context of any socketcall.
466 static inline void __sock_put(struct sock *sk)
468 atomic_dec(&sk->sk_refcnt);
471 static __inline__ int sk_del_node_init(struct sock *sk)
473 int rc = __sk_del_node_init(sk);
476 /* paranoid for a while -acme */
477 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
482 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
484 static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
487 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
493 static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
495 int rc = __sk_nulls_del_node_init_rcu(sk);
498 /* paranoid for a while -acme */
499 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
505 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
507 hlist_add_head(&sk->sk_node, list);
510 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
513 __sk_add_node(sk, list);
516 static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
519 hlist_add_head_rcu(&sk->sk_node, list);
522 static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
524 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
527 static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
530 __sk_nulls_add_node_rcu(sk, list);
533 static __inline__ void __sk_del_bind_node(struct sock *sk)
535 __hlist_del(&sk->sk_bind_node);
538 static __inline__ void sk_add_bind_node(struct sock *sk,
539 struct hlist_head *list)
541 hlist_add_head(&sk->sk_bind_node, list);
544 #define sk_for_each(__sk, node, list) \
545 hlist_for_each_entry(__sk, node, list, sk_node)
546 #define sk_for_each_rcu(__sk, node, list) \
547 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
548 #define sk_nulls_for_each(__sk, node, list) \
549 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
550 #define sk_nulls_for_each_rcu(__sk, node, list) \
551 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
552 #define sk_for_each_from(__sk, node) \
553 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
554 hlist_for_each_entry_from(__sk, node, sk_node)
555 #define sk_nulls_for_each_from(__sk, node) \
556 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
557 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
558 #define sk_for_each_safe(__sk, node, tmp, list) \
559 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
560 #define sk_for_each_bound(__sk, node, list) \
561 hlist_for_each_entry(__sk, node, list, sk_bind_node)
574 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
575 SOCK_DBG, /* %SO_DEBUG setting */
576 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
577 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
578 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
579 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
580 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
581 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
582 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
583 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
584 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
585 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
586 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
587 SOCK_FASYNC, /* fasync() active */
589 SOCK_ZEROCOPY, /* buffers from userspace */
590 SOCK_WIFI_STATUS, /* push wifi status to userspace */
593 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
595 nsk->sk_flags = osk->sk_flags;
598 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
600 __set_bit(flag, &sk->sk_flags);
603 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
605 __clear_bit(flag, &sk->sk_flags);
608 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
610 return test_bit(flag, &sk->sk_flags);
613 static inline void sk_acceptq_removed(struct sock *sk)
615 sk->sk_ack_backlog--;
618 static inline void sk_acceptq_added(struct sock *sk)
620 sk->sk_ack_backlog++;
623 static inline int sk_acceptq_is_full(struct sock *sk)
625 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
629 * Compute minimal free write space needed to queue new packets.
631 static inline int sk_stream_min_wspace(struct sock *sk)
633 return sk->sk_wmem_queued >> 1;
636 static inline int sk_stream_wspace(struct sock *sk)
638 return sk->sk_sndbuf - sk->sk_wmem_queued;
641 extern void sk_stream_write_space(struct sock *sk);
643 static inline int sk_stream_memory_free(struct sock *sk)
645 return sk->sk_wmem_queued < sk->sk_sndbuf;
648 /* OOB backlog add */
649 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
651 /* dont let skb dst not refcounted, we are going to leave rcu lock */
654 if (!sk->sk_backlog.tail)
655 sk->sk_backlog.head = skb;
657 sk->sk_backlog.tail->next = skb;
659 sk->sk_backlog.tail = skb;
664 * Take into account size of receive queue and backlog queue
666 static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb)
668 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
670 return qsize + skb->truesize > sk->sk_rcvbuf;
673 /* The per-socket spinlock must be held here. */
674 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb)
676 if (sk_rcvqueues_full(sk, skb))
679 __sk_add_backlog(sk, skb);
680 sk->sk_backlog.len += skb->truesize;
684 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
686 return sk->sk_backlog_rcv(sk, skb);
689 static inline void sock_rps_record_flow(const struct sock *sk)
692 struct rps_sock_flow_table *sock_flow_table;
695 sock_flow_table = rcu_dereference(rps_sock_flow_table);
696 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
701 static inline void sock_rps_reset_flow(const struct sock *sk)
704 struct rps_sock_flow_table *sock_flow_table;
707 sock_flow_table = rcu_dereference(rps_sock_flow_table);
708 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
713 static inline void sock_rps_save_rxhash(struct sock *sk,
714 const struct sk_buff *skb)
717 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
718 sock_rps_reset_flow(sk);
719 sk->sk_rxhash = skb->rxhash;
724 static inline void sock_rps_reset_rxhash(struct sock *sk)
727 sock_rps_reset_flow(sk);
732 #define sk_wait_event(__sk, __timeo, __condition) \
734 release_sock(__sk); \
735 __rc = __condition; \
737 *(__timeo) = schedule_timeout(*(__timeo)); \
740 __rc = __condition; \
744 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
745 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
746 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
747 extern int sk_stream_error(struct sock *sk, int flags, int err);
748 extern void sk_stream_kill_queues(struct sock *sk);
750 extern int sk_wait_data(struct sock *sk, long *timeo);
752 struct request_sock_ops;
753 struct timewait_sock_ops;
754 struct inet_hashinfo;
758 /* Networking protocol blocks we attach to sockets.
759 * socket layer -> transport layer interface
760 * transport -> network interface is defined by struct inet_proto
763 void (*close)(struct sock *sk,
765 int (*connect)(struct sock *sk,
766 struct sockaddr *uaddr,
768 int (*disconnect)(struct sock *sk, int flags);
770 struct sock * (*accept) (struct sock *sk, int flags, int *err);
772 int (*ioctl)(struct sock *sk, int cmd,
774 int (*init)(struct sock *sk);
775 void (*destroy)(struct sock *sk);
776 void (*shutdown)(struct sock *sk, int how);
777 int (*setsockopt)(struct sock *sk, int level,
778 int optname, char __user *optval,
779 unsigned int optlen);
780 int (*getsockopt)(struct sock *sk, int level,
781 int optname, char __user *optval,
784 int (*compat_setsockopt)(struct sock *sk,
786 int optname, char __user *optval,
787 unsigned int optlen);
788 int (*compat_getsockopt)(struct sock *sk,
790 int optname, char __user *optval,
792 int (*compat_ioctl)(struct sock *sk,
793 unsigned int cmd, unsigned long arg);
795 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
796 struct msghdr *msg, size_t len);
797 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
799 size_t len, int noblock, int flags,
801 int (*sendpage)(struct sock *sk, struct page *page,
802 int offset, size_t size, int flags);
803 int (*bind)(struct sock *sk,
804 struct sockaddr *uaddr, int addr_len);
806 int (*backlog_rcv) (struct sock *sk,
807 struct sk_buff *skb);
809 /* Keeping track of sk's, looking them up, and port selection methods. */
810 void (*hash)(struct sock *sk);
811 void (*unhash)(struct sock *sk);
812 void (*rehash)(struct sock *sk);
813 int (*get_port)(struct sock *sk, unsigned short snum);
814 void (*clear_sk)(struct sock *sk, int size);
816 /* Keeping track of sockets in use */
817 #ifdef CONFIG_PROC_FS
818 unsigned int inuse_idx;
821 /* Memory pressure */
822 void (*enter_memory_pressure)(struct sock *sk);
823 atomic_long_t *memory_allocated; /* Current allocated memory. */
824 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
826 * Pressure flag: try to collapse.
827 * Technical note: it is used by multiple contexts non atomically.
828 * All the __sk_mem_schedule() is of this nature: accounting
829 * is strict, actions are advisory and have some latency.
831 int *memory_pressure;
838 struct kmem_cache *slab;
839 unsigned int obj_size;
842 struct percpu_counter *orphan_count;
844 struct request_sock_ops *rsk_prot;
845 struct timewait_sock_ops *twsk_prot;
848 struct inet_hashinfo *hashinfo;
849 struct udp_table *udp_table;
850 struct raw_hashinfo *raw_hash;
853 struct module *owner;
857 struct list_head node;
858 #ifdef SOCK_REFCNT_DEBUG
861 #ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
863 * cgroup specific init/deinit functions. Called once for all
864 * protocols that implement it, from cgroups populate function.
865 * This function has to setup any files the protocol want to
866 * appear in the kmem cgroup filesystem.
868 int (*init_cgroup)(struct cgroup *cgrp,
869 struct cgroup_subsys *ss);
870 void (*destroy_cgroup)(struct cgroup *cgrp,
871 struct cgroup_subsys *ss);
872 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
877 void (*enter_memory_pressure)(struct sock *sk);
878 struct res_counter *memory_allocated; /* Current allocated memory. */
879 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
880 int *memory_pressure;
883 * memcg field is used to find which memcg we belong directly
884 * Each memcg struct can hold more than one cg_proto, so container_of
887 * The elegant solution would be having an inverse function to
888 * proto_cgroup in struct proto, but that means polluting the structure
889 * for everybody, instead of just for memcg users.
891 struct mem_cgroup *memcg;
894 extern int proto_register(struct proto *prot, int alloc_slab);
895 extern void proto_unregister(struct proto *prot);
897 #ifdef SOCK_REFCNT_DEBUG
898 static inline void sk_refcnt_debug_inc(struct sock *sk)
900 atomic_inc(&sk->sk_prot->socks);
903 static inline void sk_refcnt_debug_dec(struct sock *sk)
905 atomic_dec(&sk->sk_prot->socks);
906 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
907 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
910 inline void sk_refcnt_debug_release(const struct sock *sk)
912 if (atomic_read(&sk->sk_refcnt) != 1)
913 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
914 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
916 #else /* SOCK_REFCNT_DEBUG */
917 #define sk_refcnt_debug_inc(sk) do { } while (0)
918 #define sk_refcnt_debug_dec(sk) do { } while (0)
919 #define sk_refcnt_debug_release(sk) do { } while (0)
920 #endif /* SOCK_REFCNT_DEBUG */
922 #ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
923 extern struct jump_label_key memcg_socket_limit_enabled;
924 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
925 struct cg_proto *cg_proto)
927 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
929 #define mem_cgroup_sockets_enabled static_branch(&memcg_socket_limit_enabled)
931 #define mem_cgroup_sockets_enabled 0
932 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
933 struct cg_proto *cg_proto)
940 static inline bool sk_has_memory_pressure(const struct sock *sk)
942 return sk->sk_prot->memory_pressure != NULL;
945 static inline bool sk_under_memory_pressure(const struct sock *sk)
947 if (!sk->sk_prot->memory_pressure)
950 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
951 return !!*sk->sk_cgrp->memory_pressure;
953 return !!*sk->sk_prot->memory_pressure;
956 static inline void sk_leave_memory_pressure(struct sock *sk)
958 int *memory_pressure = sk->sk_prot->memory_pressure;
960 if (!memory_pressure)
963 if (*memory_pressure)
964 *memory_pressure = 0;
966 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
967 struct cg_proto *cg_proto = sk->sk_cgrp;
968 struct proto *prot = sk->sk_prot;
970 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
971 if (*cg_proto->memory_pressure)
972 *cg_proto->memory_pressure = 0;
977 static inline void sk_enter_memory_pressure(struct sock *sk)
979 if (!sk->sk_prot->enter_memory_pressure)
982 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
983 struct cg_proto *cg_proto = sk->sk_cgrp;
984 struct proto *prot = sk->sk_prot;
986 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
987 cg_proto->enter_memory_pressure(sk);
990 sk->sk_prot->enter_memory_pressure(sk);
993 static inline long sk_prot_mem_limits(const struct sock *sk, int index)
995 long *prot = sk->sk_prot->sysctl_mem;
996 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
997 prot = sk->sk_cgrp->sysctl_mem;
1001 static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1005 struct res_counter *fail;
1008 ret = res_counter_charge(prot->memory_allocated,
1009 amt << PAGE_SHIFT, &fail);
1012 *parent_status = OVER_LIMIT;
1015 static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1018 res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
1021 static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1024 ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
1025 return ret >> PAGE_SHIFT;
1029 sk_memory_allocated(const struct sock *sk)
1031 struct proto *prot = sk->sk_prot;
1032 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1033 return memcg_memory_allocated_read(sk->sk_cgrp);
1035 return atomic_long_read(prot->memory_allocated);
1039 sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
1041 struct proto *prot = sk->sk_prot;
1043 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1044 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1045 /* update the root cgroup regardless */
1046 atomic_long_add_return(amt, prot->memory_allocated);
1047 return memcg_memory_allocated_read(sk->sk_cgrp);
1050 return atomic_long_add_return(amt, prot->memory_allocated);
1054 sk_memory_allocated_sub(struct sock *sk, int amt, int parent_status)
1056 struct proto *prot = sk->sk_prot;
1058 if (mem_cgroup_sockets_enabled && sk->sk_cgrp &&
1059 parent_status != OVER_LIMIT) /* Otherwise was uncharged already */
1060 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1062 atomic_long_sub(amt, prot->memory_allocated);
1065 static inline void sk_sockets_allocated_dec(struct sock *sk)
1067 struct proto *prot = sk->sk_prot;
1069 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1070 struct cg_proto *cg_proto = sk->sk_cgrp;
1072 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1073 percpu_counter_dec(cg_proto->sockets_allocated);
1076 percpu_counter_dec(prot->sockets_allocated);
1079 static inline void sk_sockets_allocated_inc(struct sock *sk)
1081 struct proto *prot = sk->sk_prot;
1083 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1084 struct cg_proto *cg_proto = sk->sk_cgrp;
1086 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1087 percpu_counter_inc(cg_proto->sockets_allocated);
1090 percpu_counter_inc(prot->sockets_allocated);
1094 sk_sockets_allocated_read_positive(struct sock *sk)
1096 struct proto *prot = sk->sk_prot;
1098 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1099 return percpu_counter_sum_positive(sk->sk_cgrp->sockets_allocated);
1101 return percpu_counter_sum_positive(prot->sockets_allocated);
1105 proto_sockets_allocated_sum_positive(struct proto *prot)
1107 return percpu_counter_sum_positive(prot->sockets_allocated);
1111 proto_memory_allocated(struct proto *prot)
1113 return atomic_long_read(prot->memory_allocated);
1117 proto_memory_pressure(struct proto *prot)
1119 if (!prot->memory_pressure)
1121 return !!*prot->memory_pressure;
1125 #ifdef CONFIG_PROC_FS
1126 /* Called with local bh disabled */
1127 extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1128 extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
1130 static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
1137 /* With per-bucket locks this operation is not-atomic, so that
1138 * this version is not worse.
1140 static inline void __sk_prot_rehash(struct sock *sk)
1142 sk->sk_prot->unhash(sk);
1143 sk->sk_prot->hash(sk);
1146 void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1148 /* About 10 seconds */
1149 #define SOCK_DESTROY_TIME (10*HZ)
1151 /* Sockets 0-1023 can't be bound to unless you are superuser */
1152 #define PROT_SOCK 1024
1154 #define SHUTDOWN_MASK 3
1155 #define RCV_SHUTDOWN 1
1156 #define SEND_SHUTDOWN 2
1158 #define SOCK_SNDBUF_LOCK 1
1159 #define SOCK_RCVBUF_LOCK 2
1160 #define SOCK_BINDADDR_LOCK 4
1161 #define SOCK_BINDPORT_LOCK 8
1163 /* sock_iocb: used to kick off async processing of socket ios */
1165 struct list_head list;
1169 struct socket *sock;
1171 struct scm_cookie *scm;
1172 struct msghdr *msg, async_msg;
1173 struct kiocb *kiocb;
1176 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1178 return (struct sock_iocb *)iocb->private;
1181 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1186 struct socket_alloc {
1187 struct socket socket;
1188 struct inode vfs_inode;
1191 static inline struct socket *SOCKET_I(struct inode *inode)
1193 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1196 static inline struct inode *SOCK_INODE(struct socket *socket)
1198 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1202 * Functions for memory accounting
1204 extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
1205 extern void __sk_mem_reclaim(struct sock *sk);
1207 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1208 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1209 #define SK_MEM_SEND 0
1210 #define SK_MEM_RECV 1
1212 static inline int sk_mem_pages(int amt)
1214 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1217 static inline int sk_has_account(struct sock *sk)
1219 /* return true if protocol supports memory accounting */
1220 return !!sk->sk_prot->memory_allocated;
1223 static inline int sk_wmem_schedule(struct sock *sk, int size)
1225 if (!sk_has_account(sk))
1227 return size <= sk->sk_forward_alloc ||
1228 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1231 static inline int sk_rmem_schedule(struct sock *sk, int size)
1233 if (!sk_has_account(sk))
1235 return size <= sk->sk_forward_alloc ||
1236 __sk_mem_schedule(sk, size, SK_MEM_RECV);
1239 static inline void sk_mem_reclaim(struct sock *sk)
1241 if (!sk_has_account(sk))
1243 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1244 __sk_mem_reclaim(sk);
1247 static inline void sk_mem_reclaim_partial(struct sock *sk)
1249 if (!sk_has_account(sk))
1251 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1252 __sk_mem_reclaim(sk);
1255 static inline void sk_mem_charge(struct sock *sk, int size)
1257 if (!sk_has_account(sk))
1259 sk->sk_forward_alloc -= size;
1262 static inline void sk_mem_uncharge(struct sock *sk, int size)
1264 if (!sk_has_account(sk))
1266 sk->sk_forward_alloc += size;
1269 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1271 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1272 sk->sk_wmem_queued -= skb->truesize;
1273 sk_mem_uncharge(sk, skb->truesize);
1277 /* Used by processes to "lock" a socket state, so that
1278 * interrupts and bottom half handlers won't change it
1279 * from under us. It essentially blocks any incoming
1280 * packets, so that we won't get any new data or any
1281 * packets that change the state of the socket.
1283 * While locked, BH processing will add new packets to
1284 * the backlog queue. This queue is processed by the
1285 * owner of the socket lock right before it is released.
1287 * Since ~2.3.5 it is also exclusive sleep lock serializing
1288 * accesses from user process context.
1290 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1293 * Macro so as to not evaluate some arguments when
1294 * lockdep is not enabled.
1296 * Mark both the sk_lock and the sk_lock.slock as a
1297 * per-address-family lock class.
1299 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1301 sk->sk_lock.owned = 0; \
1302 init_waitqueue_head(&sk->sk_lock.wq); \
1303 spin_lock_init(&(sk)->sk_lock.slock); \
1304 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1305 sizeof((sk)->sk_lock)); \
1306 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1308 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1311 extern void lock_sock_nested(struct sock *sk, int subclass);
1313 static inline void lock_sock(struct sock *sk)
1315 lock_sock_nested(sk, 0);
1318 extern void release_sock(struct sock *sk);
1320 /* BH context may only use the following locking interface. */
1321 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1322 #define bh_lock_sock_nested(__sk) \
1323 spin_lock_nested(&((__sk)->sk_lock.slock), \
1324 SINGLE_DEPTH_NESTING)
1325 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1327 extern bool lock_sock_fast(struct sock *sk);
1329 * unlock_sock_fast - complement of lock_sock_fast
1333 * fast unlock socket for user context.
1334 * If slow mode is on, we call regular release_sock()
1336 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1341 spin_unlock_bh(&sk->sk_lock.slock);
1345 extern struct sock *sk_alloc(struct net *net, int family,
1347 struct proto *prot);
1348 extern void sk_free(struct sock *sk);
1349 extern void sk_release_kernel(struct sock *sk);
1350 extern struct sock *sk_clone_lock(const struct sock *sk,
1351 const gfp_t priority);
1353 extern struct sk_buff *sock_wmalloc(struct sock *sk,
1354 unsigned long size, int force,
1356 extern struct sk_buff *sock_rmalloc(struct sock *sk,
1357 unsigned long size, int force,
1359 extern void sock_wfree(struct sk_buff *skb);
1360 extern void sock_rfree(struct sk_buff *skb);
1362 extern int sock_setsockopt(struct socket *sock, int level,
1363 int op, char __user *optval,
1364 unsigned int optlen);
1366 extern int sock_getsockopt(struct socket *sock, int level,
1367 int op, char __user *optval,
1368 int __user *optlen);
1369 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1373 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1374 unsigned long header_len,
1375 unsigned long data_len,
1378 extern void *sock_kmalloc(struct sock *sk, int size,
1380 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1381 extern void sk_send_sigurg(struct sock *sk);
1383 #ifdef CONFIG_CGROUPS
1384 extern void sock_update_classid(struct sock *sk);
1386 static inline void sock_update_classid(struct sock *sk)
1392 * Functions to fill in entries in struct proto_ops when a protocol
1393 * does not implement a particular function.
1395 extern int sock_no_bind(struct socket *,
1396 struct sockaddr *, int);
1397 extern int sock_no_connect(struct socket *,
1398 struct sockaddr *, int, int);
1399 extern int sock_no_socketpair(struct socket *,
1401 extern int sock_no_accept(struct socket *,
1402 struct socket *, int);
1403 extern int sock_no_getname(struct socket *,
1404 struct sockaddr *, int *, int);
1405 extern unsigned int sock_no_poll(struct file *, struct socket *,
1406 struct poll_table_struct *);
1407 extern int sock_no_ioctl(struct socket *, unsigned int,
1409 extern int sock_no_listen(struct socket *, int);
1410 extern int sock_no_shutdown(struct socket *, int);
1411 extern int sock_no_getsockopt(struct socket *, int , int,
1412 char __user *, int __user *);
1413 extern int sock_no_setsockopt(struct socket *, int, int,
1414 char __user *, unsigned int);
1415 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1416 struct msghdr *, size_t);
1417 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1418 struct msghdr *, size_t, int);
1419 extern int sock_no_mmap(struct file *file,
1420 struct socket *sock,
1421 struct vm_area_struct *vma);
1422 extern ssize_t sock_no_sendpage(struct socket *sock,
1424 int offset, size_t size,
1428 * Functions to fill in entries in struct proto_ops when a protocol
1429 * uses the inet style.
1431 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1432 char __user *optval, int __user *optlen);
1433 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1434 struct msghdr *msg, size_t size, int flags);
1435 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
1436 char __user *optval, unsigned int optlen);
1437 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1438 int optname, char __user *optval, int __user *optlen);
1439 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
1440 int optname, char __user *optval, unsigned int optlen);
1442 extern void sk_common_release(struct sock *sk);
1445 * Default socket callbacks and setup code
1448 /* Initialise core socket variables */
1449 extern void sock_init_data(struct socket *sock, struct sock *sk);
1451 extern void sk_filter_release_rcu(struct rcu_head *rcu);
1454 * sk_filter_release - release a socket filter
1455 * @fp: filter to remove
1457 * Remove a filter from a socket and release its resources.
1460 static inline void sk_filter_release(struct sk_filter *fp)
1462 if (atomic_dec_and_test(&fp->refcnt))
1463 call_rcu(&fp->rcu, sk_filter_release_rcu);
1466 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1468 unsigned int size = sk_filter_len(fp);
1470 atomic_sub(size, &sk->sk_omem_alloc);
1471 sk_filter_release(fp);
1474 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1476 atomic_inc(&fp->refcnt);
1477 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1481 * Socket reference counting postulates.
1483 * * Each user of socket SHOULD hold a reference count.
1484 * * Each access point to socket (an hash table bucket, reference from a list,
1485 * running timer, skb in flight MUST hold a reference count.
1486 * * When reference count hits 0, it means it will never increase back.
1487 * * When reference count hits 0, it means that no references from
1488 * outside exist to this socket and current process on current CPU
1489 * is last user and may/should destroy this socket.
1490 * * sk_free is called from any context: process, BH, IRQ. When
1491 * it is called, socket has no references from outside -> sk_free
1492 * may release descendant resources allocated by the socket, but
1493 * to the time when it is called, socket is NOT referenced by any
1494 * hash tables, lists etc.
1495 * * Packets, delivered from outside (from network or from another process)
1496 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1497 * when they sit in queue. Otherwise, packets will leak to hole, when
1498 * socket is looked up by one cpu and unhasing is made by another CPU.
1499 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1500 * (leak to backlog). Packet socket does all the processing inside
1501 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1502 * use separate SMP lock, so that they are prone too.
1505 /* Ungrab socket and destroy it, if it was the last reference. */
1506 static inline void sock_put(struct sock *sk)
1508 if (atomic_dec_and_test(&sk->sk_refcnt))
1512 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1515 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1517 sk->sk_tx_queue_mapping = tx_queue;
1520 static inline void sk_tx_queue_clear(struct sock *sk)
1522 sk->sk_tx_queue_mapping = -1;
1525 static inline int sk_tx_queue_get(const struct sock *sk)
1527 return sk ? sk->sk_tx_queue_mapping : -1;
1530 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1532 sk_tx_queue_clear(sk);
1533 sk->sk_socket = sock;
1536 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1538 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1539 return &rcu_dereference_raw(sk->sk_wq)->wait;
1541 /* Detach socket from process context.
1542 * Announce socket dead, detach it from wait queue and inode.
1543 * Note that parent inode held reference count on this struct sock,
1544 * we do not release it in this function, because protocol
1545 * probably wants some additional cleanups or even continuing
1546 * to work with this socket (TCP).
1548 static inline void sock_orphan(struct sock *sk)
1550 write_lock_bh(&sk->sk_callback_lock);
1551 sock_set_flag(sk, SOCK_DEAD);
1552 sk_set_socket(sk, NULL);
1554 write_unlock_bh(&sk->sk_callback_lock);
1557 static inline void sock_graft(struct sock *sk, struct socket *parent)
1559 write_lock_bh(&sk->sk_callback_lock);
1560 sk->sk_wq = parent->wq;
1562 sk_set_socket(sk, parent);
1563 security_sock_graft(sk, parent);
1564 write_unlock_bh(&sk->sk_callback_lock);
1567 extern int sock_i_uid(struct sock *sk);
1568 extern unsigned long sock_i_ino(struct sock *sk);
1570 static inline struct dst_entry *
1571 __sk_dst_get(struct sock *sk)
1573 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
1574 lockdep_is_held(&sk->sk_lock.slock));
1577 static inline struct dst_entry *
1578 sk_dst_get(struct sock *sk)
1580 struct dst_entry *dst;
1583 dst = rcu_dereference(sk->sk_dst_cache);
1590 extern void sk_reset_txq(struct sock *sk);
1592 static inline void dst_negative_advice(struct sock *sk)
1594 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1596 if (dst && dst->ops->negative_advice) {
1597 ndst = dst->ops->negative_advice(dst);
1600 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1607 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1609 struct dst_entry *old_dst;
1611 sk_tx_queue_clear(sk);
1613 * This can be called while sk is owned by the caller only,
1614 * with no state that can be checked in a rcu_dereference_check() cond
1616 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1617 rcu_assign_pointer(sk->sk_dst_cache, dst);
1618 dst_release(old_dst);
1622 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1624 spin_lock(&sk->sk_dst_lock);
1625 __sk_dst_set(sk, dst);
1626 spin_unlock(&sk->sk_dst_lock);
1630 __sk_dst_reset(struct sock *sk)
1632 __sk_dst_set(sk, NULL);
1636 sk_dst_reset(struct sock *sk)
1638 spin_lock(&sk->sk_dst_lock);
1640 spin_unlock(&sk->sk_dst_lock);
1643 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1645 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1647 static inline int sk_can_gso(const struct sock *sk)
1649 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1652 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1654 static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
1656 sk->sk_route_nocaps |= flags;
1657 sk->sk_route_caps &= ~flags;
1660 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1661 char __user *from, char *to,
1662 int copy, int offset)
1664 if (skb->ip_summed == CHECKSUM_NONE) {
1666 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1669 skb->csum = csum_block_add(skb->csum, csum, offset);
1670 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1671 if (!access_ok(VERIFY_READ, from, copy) ||
1672 __copy_from_user_nocache(to, from, copy))
1674 } else if (copy_from_user(to, from, copy))
1680 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1681 char __user *from, int copy)
1683 int err, offset = skb->len;
1685 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1688 __skb_trim(skb, offset);
1693 static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1694 struct sk_buff *skb,
1700 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1706 skb->data_len += copy;
1707 skb->truesize += copy;
1708 sk->sk_wmem_queued += copy;
1709 sk_mem_charge(sk, copy);
1713 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1714 struct sk_buff *skb, struct page *page,
1717 if (skb->ip_summed == CHECKSUM_NONE) {
1719 __wsum csum = csum_and_copy_from_user(from,
1720 page_address(page) + off,
1724 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1725 } else if (copy_from_user(page_address(page) + off, from, copy))
1729 skb->data_len += copy;
1730 skb->truesize += copy;
1731 sk->sk_wmem_queued += copy;
1732 sk_mem_charge(sk, copy);
1737 * sk_wmem_alloc_get - returns write allocations
1740 * Returns sk_wmem_alloc minus initial offset of one
1742 static inline int sk_wmem_alloc_get(const struct sock *sk)
1744 return atomic_read(&sk->sk_wmem_alloc) - 1;
1748 * sk_rmem_alloc_get - returns read allocations
1751 * Returns sk_rmem_alloc
1753 static inline int sk_rmem_alloc_get(const struct sock *sk)
1755 return atomic_read(&sk->sk_rmem_alloc);
1759 * sk_has_allocations - check if allocations are outstanding
1762 * Returns true if socket has write or read allocations
1764 static inline int sk_has_allocations(const struct sock *sk)
1766 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1770 * wq_has_sleeper - check if there are any waiting processes
1771 * @wq: struct socket_wq
1773 * Returns true if socket_wq has waiting processes
1775 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1776 * barrier call. They were added due to the race found within the tcp code.
1778 * Consider following tcp code paths:
1782 * sys_select receive packet
1784 * __add_wait_queue update tp->rcv_nxt
1786 * tp->rcv_nxt check sock_def_readable
1788 * schedule rcu_read_lock();
1789 * wq = rcu_dereference(sk->sk_wq);
1790 * if (wq && waitqueue_active(&wq->wait))
1791 * wake_up_interruptible(&wq->wait)
1795 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1796 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1797 * could then endup calling schedule and sleep forever if there are no more
1798 * data on the socket.
1801 static inline bool wq_has_sleeper(struct socket_wq *wq)
1805 * We need to be sure we are in sync with the
1806 * add_wait_queue modifications to the wait queue.
1808 * This memory barrier is paired in the sock_poll_wait.
1811 return wq && waitqueue_active(&wq->wait);
1815 * sock_poll_wait - place memory barrier behind the poll_wait call.
1817 * @wait_address: socket wait queue
1820 * See the comments in the wq_has_sleeper function.
1822 static inline void sock_poll_wait(struct file *filp,
1823 wait_queue_head_t *wait_address, poll_table *p)
1825 if (p && wait_address) {
1826 poll_wait(filp, wait_address, p);
1828 * We need to be sure we are in sync with the
1829 * socket flags modification.
1831 * This memory barrier is paired in the wq_has_sleeper.
1838 * Queue a received datagram if it will fit. Stream and sequenced
1839 * protocols can't normally use this as they need to fit buffers in
1840 * and play with them.
1842 * Inlined as it's very short and called for pretty much every
1843 * packet ever received.
1846 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1850 skb->destructor = sock_wfree;
1852 * We used to take a refcount on sk, but following operation
1853 * is enough to guarantee sk_free() wont free this sock until
1854 * all in-flight packets are completed
1856 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1859 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1863 skb->destructor = sock_rfree;
1864 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1865 sk_mem_charge(sk, skb->truesize);
1868 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1869 unsigned long expires);
1871 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1873 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1875 extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1878 * Recover an error report and clear atomically
1881 static inline int sock_error(struct sock *sk)
1884 if (likely(!sk->sk_err))
1886 err = xchg(&sk->sk_err, 0);
1890 static inline unsigned long sock_wspace(struct sock *sk)
1894 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1895 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1902 static inline void sk_wake_async(struct sock *sk, int how, int band)
1904 if (sock_flag(sk, SOCK_FASYNC))
1905 sock_wake_async(sk->sk_socket, how, band);
1908 #define SOCK_MIN_SNDBUF 2048
1910 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1911 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1913 #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
1915 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1917 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1918 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1919 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1923 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1925 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1927 struct page *page = NULL;
1929 page = alloc_pages(sk->sk_allocation, 0);
1931 sk_enter_memory_pressure(sk);
1932 sk_stream_moderate_sndbuf(sk);
1938 * Default write policy as shown to user space via poll/select/SIGIO
1940 static inline int sock_writeable(const struct sock *sk)
1942 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1945 static inline gfp_t gfp_any(void)
1947 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1950 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1952 return noblock ? 0 : sk->sk_rcvtimeo;
1955 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1957 return noblock ? 0 : sk->sk_sndtimeo;
1960 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1962 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1965 /* Alas, with timeout socket operations are not restartable.
1966 * Compare this to poll().
1968 static inline int sock_intr_errno(long timeo)
1970 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1973 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1974 struct sk_buff *skb);
1975 extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
1976 struct sk_buff *skb);
1978 static __inline__ void
1979 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1981 ktime_t kt = skb->tstamp;
1982 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
1985 * generate control messages if
1986 * - receive time stamping in software requested (SOCK_RCVTSTAMP
1987 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
1988 * - software time stamp available and wanted
1989 * (SOCK_TIMESTAMPING_SOFTWARE)
1990 * - hardware time stamps available and wanted
1991 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
1992 * SOCK_TIMESTAMPING_RAW_HARDWARE)
1994 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
1995 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
1996 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
1997 (hwtstamps->hwtstamp.tv64 &&
1998 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
1999 (hwtstamps->syststamp.tv64 &&
2000 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
2001 __sock_recv_timestamp(msg, sk, skb);
2005 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2006 __sock_recv_wifi_status(msg, sk, skb);
2009 extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2010 struct sk_buff *skb);
2012 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2013 struct sk_buff *skb)
2015 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2016 (1UL << SOCK_RCVTSTAMP) | \
2017 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2018 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
2019 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
2020 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2022 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2023 __sock_recv_ts_and_drops(msg, sk, skb);
2025 sk->sk_stamp = skb->tstamp;
2029 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2030 * @sk: socket sending this packet
2031 * @tx_flags: filled with instructions for time stamping
2033 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
2034 * parameters are invalid.
2036 extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
2039 * sk_eat_skb - Release a skb if it is no longer needed
2040 * @sk: socket to eat this skb from
2041 * @skb: socket buffer to eat
2042 * @copied_early: flag indicating whether DMA operations copied this data early
2044 * This routine must be called with interrupts disabled or with the socket
2045 * locked so that the sk_buff queue operation is ok.
2047 #ifdef CONFIG_NET_DMA
2048 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
2050 __skb_unlink(skb, &sk->sk_receive_queue);
2054 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2057 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
2059 __skb_unlink(skb, &sk->sk_receive_queue);
2065 struct net *sock_net(const struct sock *sk)
2067 return read_pnet(&sk->sk_net);
2071 void sock_net_set(struct sock *sk, struct net *net)
2073 write_pnet(&sk->sk_net, net);
2077 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
2078 * They should not hold a reference to a namespace in order to allow
2080 * Sockets after sk_change_net should be released using sk_release_kernel
2082 static inline void sk_change_net(struct sock *sk, struct net *net)
2084 put_net(sock_net(sk));
2085 sock_net_set(sk, hold_net(net));
2088 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2090 if (unlikely(skb->sk)) {
2091 struct sock *sk = skb->sk;
2093 skb->destructor = NULL;
2100 extern void sock_enable_timestamp(struct sock *sk, int flag);
2101 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
2102 extern int sock_get_timestampns(struct sock *, struct timespec __user *);
2105 * Enable debug/info messages
2107 extern int net_msg_warn;
2108 #define NETDEBUG(fmt, args...) \
2109 do { if (net_msg_warn) printk(fmt,##args); } while (0)
2111 #define LIMIT_NETDEBUG(fmt, args...) \
2112 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
2114 extern __u32 sysctl_wmem_max;
2115 extern __u32 sysctl_rmem_max;
2117 extern void sk_init(void);
2119 extern int sysctl_optmem_max;
2121 extern __u32 sysctl_wmem_default;
2122 extern __u32 sysctl_rmem_default;
2124 #endif /* _SOCK_H */