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>
58 #include <linux/static_key.h>
59 #include <linux/aio.h>
60 #include <linux/sched.h>
62 #include <linux/filter.h>
63 #include <linux/rculist_nulls.h>
64 #include <linux/poll.h>
66 #include <linux/atomic.h>
68 #include <net/checksum.h>
73 int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss);
74 void mem_cgroup_sockets_destroy(struct cgroup *cgrp);
77 int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss)
82 void mem_cgroup_sockets_destroy(struct cgroup *cgrp)
87 * This structure really needs to be cleaned up.
88 * Most of it is for TCP, and not used by any of
89 * the other protocols.
92 /* Define this to get the SOCK_DBG debugging facility. */
93 #define SOCK_DEBUGGING
95 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
96 printk(KERN_DEBUG msg); } while (0)
98 /* Validate arguments and do nothing */
99 static inline __printf(2, 3)
100 void SOCK_DEBUG(struct sock *sk, const char *msg, ...)
105 /* This is the per-socket lock. The spinlock provides a synchronization
106 * between user contexts and software interrupt processing, whereas the
107 * mini-semaphore synchronizes multiple users amongst themselves.
112 wait_queue_head_t wq;
114 * We express the mutex-alike socket_lock semantics
115 * to the lock validator by explicitly managing
116 * the slock as a lock variant (in addition to
119 #ifdef CONFIG_DEBUG_LOCK_ALLOC
120 struct lockdep_map dep_map;
129 * struct sock_common - minimal network layer representation of sockets
130 * @skc_daddr: Foreign IPv4 addr
131 * @skc_rcv_saddr: Bound local IPv4 addr
132 * @skc_hash: hash value used with various protocol lookup tables
133 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
134 * @skc_family: network address family
135 * @skc_state: Connection state
136 * @skc_reuse: %SO_REUSEADDR setting
137 * @skc_bound_dev_if: bound device index if != 0
138 * @skc_bind_node: bind hash linkage for various protocol lookup tables
139 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
140 * @skc_prot: protocol handlers inside a network family
141 * @skc_net: reference to the network namespace of this socket
142 * @skc_node: main hash linkage for various protocol lookup tables
143 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
144 * @skc_tx_queue_mapping: tx queue number for this connection
145 * @skc_refcnt: reference count
147 * This is the minimal network layer representation of sockets, the header
148 * for struct sock and struct inet_timewait_sock.
151 /* skc_daddr and skc_rcv_saddr must be grouped :
152 * cf INET_MATCH() and INET_TW_MATCH()
155 __be32 skc_rcv_saddr;
158 unsigned int skc_hash;
159 __u16 skc_u16hashes[2];
161 unsigned short skc_family;
162 volatile unsigned char skc_state;
163 unsigned char skc_reuse;
164 int skc_bound_dev_if;
166 struct hlist_node skc_bind_node;
167 struct hlist_nulls_node skc_portaddr_node;
169 struct proto *skc_prot;
174 * fields between dontcopy_begin/dontcopy_end
175 * are not copied in sock_copy()
178 int skc_dontcopy_begin[0];
181 struct hlist_node skc_node;
182 struct hlist_nulls_node skc_nulls_node;
184 int skc_tx_queue_mapping;
187 int skc_dontcopy_end[0];
193 * struct sock - network layer representation of sockets
194 * @__sk_common: shared layout with inet_timewait_sock
195 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
196 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
197 * @sk_lock: synchronizer
198 * @sk_rcvbuf: size of receive buffer in bytes
199 * @sk_wq: sock wait queue and async head
200 * @sk_dst_cache: destination cache
201 * @sk_dst_lock: destination cache lock
202 * @sk_policy: flow policy
203 * @sk_receive_queue: incoming packets
204 * @sk_wmem_alloc: transmit queue bytes committed
205 * @sk_write_queue: Packet sending queue
206 * @sk_async_wait_queue: DMA copied packets
207 * @sk_omem_alloc: "o" is "option" or "other"
208 * @sk_wmem_queued: persistent queue size
209 * @sk_forward_alloc: space allocated forward
210 * @sk_allocation: allocation mode
211 * @sk_sndbuf: size of send buffer in bytes
212 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
213 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
214 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
215 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
216 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
217 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
218 * @sk_gso_max_size: Maximum GSO segment size to build
219 * @sk_lingertime: %SO_LINGER l_linger setting
220 * @sk_backlog: always used with the per-socket spinlock held
221 * @sk_callback_lock: used with the callbacks in the end of this struct
222 * @sk_error_queue: rarely used
223 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
224 * IPV6_ADDRFORM for instance)
225 * @sk_err: last error
226 * @sk_err_soft: errors that don't cause failure but are the cause of a
227 * persistent failure not just 'timed out'
228 * @sk_drops: raw/udp drops counter
229 * @sk_ack_backlog: current listen backlog
230 * @sk_max_ack_backlog: listen backlog set in listen()
231 * @sk_priority: %SO_PRIORITY setting
232 * @sk_cgrp_prioidx: socket group's priority map index
233 * @sk_type: socket type (%SOCK_STREAM, etc)
234 * @sk_protocol: which protocol this socket belongs in this network family
235 * @sk_peer_pid: &struct pid for this socket's peer
236 * @sk_peer_cred: %SO_PEERCRED setting
237 * @sk_rcvlowat: %SO_RCVLOWAT setting
238 * @sk_rcvtimeo: %SO_RCVTIMEO setting
239 * @sk_sndtimeo: %SO_SNDTIMEO setting
240 * @sk_rxhash: flow hash received from netif layer
241 * @sk_filter: socket filtering instructions
242 * @sk_protinfo: private area, net family specific, when not using slab
243 * @sk_timer: sock cleanup timer
244 * @sk_stamp: time stamp of last packet received
245 * @sk_socket: Identd and reporting IO signals
246 * @sk_user_data: RPC layer private data
247 * @sk_sndmsg_page: cached page for sendmsg
248 * @sk_sndmsg_off: cached offset for sendmsg
249 * @sk_peek_off: current peek_offset value
250 * @sk_send_head: front of stuff to transmit
251 * @sk_security: used by security modules
252 * @sk_mark: generic packet mark
253 * @sk_classid: this socket's cgroup classid
254 * @sk_cgrp: this socket's cgroup-specific proto data
255 * @sk_write_pending: a write to stream socket waits to start
256 * @sk_state_change: callback to indicate change in the state of the sock
257 * @sk_data_ready: callback to indicate there is data to be processed
258 * @sk_write_space: callback to indicate there is bf sending space available
259 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
260 * @sk_backlog_rcv: callback to process the backlog
261 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
265 * Now struct inet_timewait_sock also uses sock_common, so please just
266 * don't add nothing before this first member (__sk_common) --acme
268 struct sock_common __sk_common;
269 #define sk_node __sk_common.skc_node
270 #define sk_nulls_node __sk_common.skc_nulls_node
271 #define sk_refcnt __sk_common.skc_refcnt
272 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
274 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
275 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
276 #define sk_hash __sk_common.skc_hash
277 #define sk_family __sk_common.skc_family
278 #define sk_state __sk_common.skc_state
279 #define sk_reuse __sk_common.skc_reuse
280 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
281 #define sk_bind_node __sk_common.skc_bind_node
282 #define sk_prot __sk_common.skc_prot
283 #define sk_net __sk_common.skc_net
284 socket_lock_t sk_lock;
285 struct sk_buff_head sk_receive_queue;
287 * The backlog queue is special, it is always used with
288 * the per-socket spinlock held and requires low latency
289 * access. Therefore we special case it's implementation.
290 * Note : rmem_alloc is in this structure to fill a hole
291 * on 64bit arches, not because its logically part of
297 struct sk_buff *head;
298 struct sk_buff *tail;
300 #define sk_rmem_alloc sk_backlog.rmem_alloc
301 int sk_forward_alloc;
308 struct sk_filter __rcu *sk_filter;
309 struct socket_wq __rcu *sk_wq;
311 #ifdef CONFIG_NET_DMA
312 struct sk_buff_head sk_async_wait_queue;
316 struct xfrm_policy *sk_policy[2];
318 unsigned long sk_flags;
319 struct dst_entry *sk_dst_cache;
320 spinlock_t sk_dst_lock;
321 atomic_t sk_wmem_alloc;
322 atomic_t sk_omem_alloc;
324 struct sk_buff_head sk_write_queue;
325 kmemcheck_bitfield_begin(flags);
326 unsigned int sk_shutdown : 2,
331 kmemcheck_bitfield_end(flags);
334 netdev_features_t sk_route_caps;
335 netdev_features_t sk_route_nocaps;
337 unsigned int sk_gso_max_size;
339 unsigned long sk_lingertime;
340 struct sk_buff_head sk_error_queue;
341 struct proto *sk_prot_creator;
342 rwlock_t sk_callback_lock;
345 unsigned short sk_ack_backlog;
346 unsigned short sk_max_ack_backlog;
348 #ifdef CONFIG_CGROUPS
349 __u32 sk_cgrp_prioidx;
351 struct pid *sk_peer_pid;
352 const struct cred *sk_peer_cred;
356 struct timer_list sk_timer;
358 struct socket *sk_socket;
360 struct page *sk_sndmsg_page;
361 struct sk_buff *sk_send_head;
364 int sk_write_pending;
365 #ifdef CONFIG_SECURITY
370 struct cg_proto *sk_cgrp;
371 void (*sk_state_change)(struct sock *sk);
372 void (*sk_data_ready)(struct sock *sk, int bytes);
373 void (*sk_write_space)(struct sock *sk);
374 void (*sk_error_report)(struct sock *sk);
375 int (*sk_backlog_rcv)(struct sock *sk,
376 struct sk_buff *skb);
377 void (*sk_destruct)(struct sock *sk);
380 static inline int sk_peek_offset(struct sock *sk, int flags)
382 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
383 return sk->sk_peek_off;
388 static inline void sk_peek_offset_bwd(struct sock *sk, int val)
390 if (sk->sk_peek_off >= 0) {
391 if (sk->sk_peek_off >= val)
392 sk->sk_peek_off -= val;
398 static inline void sk_peek_offset_fwd(struct sock *sk, int val)
400 if (sk->sk_peek_off >= 0)
401 sk->sk_peek_off += val;
405 * Hashed lists helper routines
407 static inline struct sock *sk_entry(const struct hlist_node *node)
409 return hlist_entry(node, struct sock, sk_node);
412 static inline struct sock *__sk_head(const struct hlist_head *head)
414 return hlist_entry(head->first, struct sock, sk_node);
417 static inline struct sock *sk_head(const struct hlist_head *head)
419 return hlist_empty(head) ? NULL : __sk_head(head);
422 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
424 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
427 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
429 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
432 static inline struct sock *sk_next(const struct sock *sk)
434 return sk->sk_node.next ?
435 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
438 static inline struct sock *sk_nulls_next(const struct sock *sk)
440 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
441 hlist_nulls_entry(sk->sk_nulls_node.next,
442 struct sock, sk_nulls_node) :
446 static inline int sk_unhashed(const struct sock *sk)
448 return hlist_unhashed(&sk->sk_node);
451 static inline int sk_hashed(const struct sock *sk)
453 return !sk_unhashed(sk);
456 static __inline__ void sk_node_init(struct hlist_node *node)
461 static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
466 static __inline__ void __sk_del_node(struct sock *sk)
468 __hlist_del(&sk->sk_node);
471 /* NB: equivalent to hlist_del_init_rcu */
472 static __inline__ int __sk_del_node_init(struct sock *sk)
476 sk_node_init(&sk->sk_node);
482 /* Grab socket reference count. This operation is valid only
483 when sk is ALREADY grabbed f.e. it is found in hash table
484 or a list and the lookup is made under lock preventing hash table
488 static inline void sock_hold(struct sock *sk)
490 atomic_inc(&sk->sk_refcnt);
493 /* Ungrab socket in the context, which assumes that socket refcnt
494 cannot hit zero, f.e. it is true in context of any socketcall.
496 static inline void __sock_put(struct sock *sk)
498 atomic_dec(&sk->sk_refcnt);
501 static __inline__ int sk_del_node_init(struct sock *sk)
503 int rc = __sk_del_node_init(sk);
506 /* paranoid for a while -acme */
507 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
512 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
514 static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
517 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
523 static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
525 int rc = __sk_nulls_del_node_init_rcu(sk);
528 /* paranoid for a while -acme */
529 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
535 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
537 hlist_add_head(&sk->sk_node, list);
540 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
543 __sk_add_node(sk, list);
546 static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
549 hlist_add_head_rcu(&sk->sk_node, list);
552 static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
554 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
557 static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
560 __sk_nulls_add_node_rcu(sk, list);
563 static __inline__ void __sk_del_bind_node(struct sock *sk)
565 __hlist_del(&sk->sk_bind_node);
568 static __inline__ void sk_add_bind_node(struct sock *sk,
569 struct hlist_head *list)
571 hlist_add_head(&sk->sk_bind_node, list);
574 #define sk_for_each(__sk, node, list) \
575 hlist_for_each_entry(__sk, node, list, sk_node)
576 #define sk_for_each_rcu(__sk, node, list) \
577 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
578 #define sk_nulls_for_each(__sk, node, list) \
579 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
580 #define sk_nulls_for_each_rcu(__sk, node, list) \
581 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
582 #define sk_for_each_from(__sk, node) \
583 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
584 hlist_for_each_entry_from(__sk, node, sk_node)
585 #define sk_nulls_for_each_from(__sk, node) \
586 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
587 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
588 #define sk_for_each_safe(__sk, node, tmp, list) \
589 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
590 #define sk_for_each_bound(__sk, node, list) \
591 hlist_for_each_entry(__sk, node, list, sk_bind_node)
604 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
605 SOCK_DBG, /* %SO_DEBUG setting */
606 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
607 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
608 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
609 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
610 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
611 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
612 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
613 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
614 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
615 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
616 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
617 SOCK_FASYNC, /* fasync() active */
619 SOCK_ZEROCOPY, /* buffers from userspace */
620 SOCK_WIFI_STATUS, /* push wifi status to userspace */
621 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
622 * Will use last 4 bytes of packet sent from
623 * user-space instead.
627 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
629 nsk->sk_flags = osk->sk_flags;
632 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
634 __set_bit(flag, &sk->sk_flags);
637 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
639 __clear_bit(flag, &sk->sk_flags);
642 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
644 return test_bit(flag, &sk->sk_flags);
647 static inline void sk_acceptq_removed(struct sock *sk)
649 sk->sk_ack_backlog--;
652 static inline void sk_acceptq_added(struct sock *sk)
654 sk->sk_ack_backlog++;
657 static inline int sk_acceptq_is_full(struct sock *sk)
659 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
663 * Compute minimal free write space needed to queue new packets.
665 static inline int sk_stream_min_wspace(struct sock *sk)
667 return sk->sk_wmem_queued >> 1;
670 static inline int sk_stream_wspace(struct sock *sk)
672 return sk->sk_sndbuf - sk->sk_wmem_queued;
675 extern void sk_stream_write_space(struct sock *sk);
677 static inline int sk_stream_memory_free(struct sock *sk)
679 return sk->sk_wmem_queued < sk->sk_sndbuf;
682 /* OOB backlog add */
683 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
685 /* dont let skb dst not refcounted, we are going to leave rcu lock */
688 if (!sk->sk_backlog.tail)
689 sk->sk_backlog.head = skb;
691 sk->sk_backlog.tail->next = skb;
693 sk->sk_backlog.tail = skb;
698 * Take into account size of receive queue and backlog queue
699 * Do not take into account this skb truesize,
700 * to allow even a single big packet to come.
702 static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb)
704 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
706 return qsize > sk->sk_rcvbuf;
709 /* The per-socket spinlock must be held here. */
710 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb)
712 if (sk_rcvqueues_full(sk, skb))
715 __sk_add_backlog(sk, skb);
716 sk->sk_backlog.len += skb->truesize;
720 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
722 return sk->sk_backlog_rcv(sk, skb);
725 static inline void sock_rps_record_flow(const struct sock *sk)
728 struct rps_sock_flow_table *sock_flow_table;
731 sock_flow_table = rcu_dereference(rps_sock_flow_table);
732 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
737 static inline void sock_rps_reset_flow(const struct sock *sk)
740 struct rps_sock_flow_table *sock_flow_table;
743 sock_flow_table = rcu_dereference(rps_sock_flow_table);
744 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
749 static inline void sock_rps_save_rxhash(struct sock *sk,
750 const struct sk_buff *skb)
753 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
754 sock_rps_reset_flow(sk);
755 sk->sk_rxhash = skb->rxhash;
760 static inline void sock_rps_reset_rxhash(struct sock *sk)
763 sock_rps_reset_flow(sk);
768 #define sk_wait_event(__sk, __timeo, __condition) \
770 release_sock(__sk); \
771 __rc = __condition; \
773 *(__timeo) = schedule_timeout(*(__timeo)); \
776 __rc = __condition; \
780 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
781 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
782 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
783 extern int sk_stream_error(struct sock *sk, int flags, int err);
784 extern void sk_stream_kill_queues(struct sock *sk);
786 extern int sk_wait_data(struct sock *sk, long *timeo);
788 struct request_sock_ops;
789 struct timewait_sock_ops;
790 struct inet_hashinfo;
794 /* Networking protocol blocks we attach to sockets.
795 * socket layer -> transport layer interface
796 * transport -> network interface is defined by struct inet_proto
799 void (*close)(struct sock *sk,
801 int (*connect)(struct sock *sk,
802 struct sockaddr *uaddr,
804 int (*disconnect)(struct sock *sk, int flags);
806 struct sock * (*accept) (struct sock *sk, int flags, int *err);
808 int (*ioctl)(struct sock *sk, int cmd,
810 int (*init)(struct sock *sk);
811 void (*destroy)(struct sock *sk);
812 void (*shutdown)(struct sock *sk, int how);
813 int (*setsockopt)(struct sock *sk, int level,
814 int optname, char __user *optval,
815 unsigned int optlen);
816 int (*getsockopt)(struct sock *sk, int level,
817 int optname, char __user *optval,
820 int (*compat_setsockopt)(struct sock *sk,
822 int optname, char __user *optval,
823 unsigned int optlen);
824 int (*compat_getsockopt)(struct sock *sk,
826 int optname, char __user *optval,
828 int (*compat_ioctl)(struct sock *sk,
829 unsigned int cmd, unsigned long arg);
831 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
832 struct msghdr *msg, size_t len);
833 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
835 size_t len, int noblock, int flags,
837 int (*sendpage)(struct sock *sk, struct page *page,
838 int offset, size_t size, int flags);
839 int (*bind)(struct sock *sk,
840 struct sockaddr *uaddr, int addr_len);
842 int (*backlog_rcv) (struct sock *sk,
843 struct sk_buff *skb);
845 /* Keeping track of sk's, looking them up, and port selection methods. */
846 void (*hash)(struct sock *sk);
847 void (*unhash)(struct sock *sk);
848 void (*rehash)(struct sock *sk);
849 int (*get_port)(struct sock *sk, unsigned short snum);
850 void (*clear_sk)(struct sock *sk, int size);
852 /* Keeping track of sockets in use */
853 #ifdef CONFIG_PROC_FS
854 unsigned int inuse_idx;
857 /* Memory pressure */
858 void (*enter_memory_pressure)(struct sock *sk);
859 atomic_long_t *memory_allocated; /* Current allocated memory. */
860 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
862 * Pressure flag: try to collapse.
863 * Technical note: it is used by multiple contexts non atomically.
864 * All the __sk_mem_schedule() is of this nature: accounting
865 * is strict, actions are advisory and have some latency.
867 int *memory_pressure;
874 struct kmem_cache *slab;
875 unsigned int obj_size;
878 struct percpu_counter *orphan_count;
880 struct request_sock_ops *rsk_prot;
881 struct timewait_sock_ops *twsk_prot;
884 struct inet_hashinfo *hashinfo;
885 struct udp_table *udp_table;
886 struct raw_hashinfo *raw_hash;
889 struct module *owner;
893 struct list_head node;
894 #ifdef SOCK_REFCNT_DEBUG
897 #ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
899 * cgroup specific init/deinit functions. Called once for all
900 * protocols that implement it, from cgroups populate function.
901 * This function has to setup any files the protocol want to
902 * appear in the kmem cgroup filesystem.
904 int (*init_cgroup)(struct cgroup *cgrp,
905 struct cgroup_subsys *ss);
906 void (*destroy_cgroup)(struct cgroup *cgrp);
907 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
912 void (*enter_memory_pressure)(struct sock *sk);
913 struct res_counter *memory_allocated; /* Current allocated memory. */
914 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
915 int *memory_pressure;
918 * memcg field is used to find which memcg we belong directly
919 * Each memcg struct can hold more than one cg_proto, so container_of
922 * The elegant solution would be having an inverse function to
923 * proto_cgroup in struct proto, but that means polluting the structure
924 * for everybody, instead of just for memcg users.
926 struct mem_cgroup *memcg;
929 extern int proto_register(struct proto *prot, int alloc_slab);
930 extern void proto_unregister(struct proto *prot);
932 #ifdef SOCK_REFCNT_DEBUG
933 static inline void sk_refcnt_debug_inc(struct sock *sk)
935 atomic_inc(&sk->sk_prot->socks);
938 static inline void sk_refcnt_debug_dec(struct sock *sk)
940 atomic_dec(&sk->sk_prot->socks);
941 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
942 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
945 inline void sk_refcnt_debug_release(const struct sock *sk)
947 if (atomic_read(&sk->sk_refcnt) != 1)
948 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
949 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
951 #else /* SOCK_REFCNT_DEBUG */
952 #define sk_refcnt_debug_inc(sk) do { } while (0)
953 #define sk_refcnt_debug_dec(sk) do { } while (0)
954 #define sk_refcnt_debug_release(sk) do { } while (0)
955 #endif /* SOCK_REFCNT_DEBUG */
957 #if defined(CONFIG_CGROUP_MEM_RES_CTLR_KMEM) && defined(CONFIG_NET)
958 extern struct static_key memcg_socket_limit_enabled;
959 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
960 struct cg_proto *cg_proto)
962 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
964 #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
966 #define mem_cgroup_sockets_enabled 0
967 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
968 struct cg_proto *cg_proto)
975 static inline bool sk_has_memory_pressure(const struct sock *sk)
977 return sk->sk_prot->memory_pressure != NULL;
980 static inline bool sk_under_memory_pressure(const struct sock *sk)
982 if (!sk->sk_prot->memory_pressure)
985 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
986 return !!*sk->sk_cgrp->memory_pressure;
988 return !!*sk->sk_prot->memory_pressure;
991 static inline void sk_leave_memory_pressure(struct sock *sk)
993 int *memory_pressure = sk->sk_prot->memory_pressure;
995 if (!memory_pressure)
998 if (*memory_pressure)
999 *memory_pressure = 0;
1001 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1002 struct cg_proto *cg_proto = sk->sk_cgrp;
1003 struct proto *prot = sk->sk_prot;
1005 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1006 if (*cg_proto->memory_pressure)
1007 *cg_proto->memory_pressure = 0;
1012 static inline void sk_enter_memory_pressure(struct sock *sk)
1014 if (!sk->sk_prot->enter_memory_pressure)
1017 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1018 struct cg_proto *cg_proto = sk->sk_cgrp;
1019 struct proto *prot = sk->sk_prot;
1021 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1022 cg_proto->enter_memory_pressure(sk);
1025 sk->sk_prot->enter_memory_pressure(sk);
1028 static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1030 long *prot = sk->sk_prot->sysctl_mem;
1031 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1032 prot = sk->sk_cgrp->sysctl_mem;
1036 static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1040 struct res_counter *fail;
1043 ret = res_counter_charge_nofail(prot->memory_allocated,
1044 amt << PAGE_SHIFT, &fail);
1046 *parent_status = OVER_LIMIT;
1049 static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1052 res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
1055 static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1058 ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
1059 return ret >> PAGE_SHIFT;
1063 sk_memory_allocated(const struct sock *sk)
1065 struct proto *prot = sk->sk_prot;
1066 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1067 return memcg_memory_allocated_read(sk->sk_cgrp);
1069 return atomic_long_read(prot->memory_allocated);
1073 sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
1075 struct proto *prot = sk->sk_prot;
1077 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1078 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1079 /* update the root cgroup regardless */
1080 atomic_long_add_return(amt, prot->memory_allocated);
1081 return memcg_memory_allocated_read(sk->sk_cgrp);
1084 return atomic_long_add_return(amt, prot->memory_allocated);
1088 sk_memory_allocated_sub(struct sock *sk, int amt)
1090 struct proto *prot = sk->sk_prot;
1092 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1093 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1095 atomic_long_sub(amt, prot->memory_allocated);
1098 static inline void sk_sockets_allocated_dec(struct sock *sk)
1100 struct proto *prot = sk->sk_prot;
1102 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1103 struct cg_proto *cg_proto = sk->sk_cgrp;
1105 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1106 percpu_counter_dec(cg_proto->sockets_allocated);
1109 percpu_counter_dec(prot->sockets_allocated);
1112 static inline void sk_sockets_allocated_inc(struct sock *sk)
1114 struct proto *prot = sk->sk_prot;
1116 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1117 struct cg_proto *cg_proto = sk->sk_cgrp;
1119 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1120 percpu_counter_inc(cg_proto->sockets_allocated);
1123 percpu_counter_inc(prot->sockets_allocated);
1127 sk_sockets_allocated_read_positive(struct sock *sk)
1129 struct proto *prot = sk->sk_prot;
1131 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1132 return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
1134 return percpu_counter_read_positive(prot->sockets_allocated);
1138 proto_sockets_allocated_sum_positive(struct proto *prot)
1140 return percpu_counter_sum_positive(prot->sockets_allocated);
1144 proto_memory_allocated(struct proto *prot)
1146 return atomic_long_read(prot->memory_allocated);
1150 proto_memory_pressure(struct proto *prot)
1152 if (!prot->memory_pressure)
1154 return !!*prot->memory_pressure;
1158 #ifdef CONFIG_PROC_FS
1159 /* Called with local bh disabled */
1160 extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1161 extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
1163 static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
1170 /* With per-bucket locks this operation is not-atomic, so that
1171 * this version is not worse.
1173 static inline void __sk_prot_rehash(struct sock *sk)
1175 sk->sk_prot->unhash(sk);
1176 sk->sk_prot->hash(sk);
1179 void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1181 /* About 10 seconds */
1182 #define SOCK_DESTROY_TIME (10*HZ)
1184 /* Sockets 0-1023 can't be bound to unless you are superuser */
1185 #define PROT_SOCK 1024
1187 #define SHUTDOWN_MASK 3
1188 #define RCV_SHUTDOWN 1
1189 #define SEND_SHUTDOWN 2
1191 #define SOCK_SNDBUF_LOCK 1
1192 #define SOCK_RCVBUF_LOCK 2
1193 #define SOCK_BINDADDR_LOCK 4
1194 #define SOCK_BINDPORT_LOCK 8
1196 /* sock_iocb: used to kick off async processing of socket ios */
1198 struct list_head list;
1202 struct socket *sock;
1204 struct scm_cookie *scm;
1205 struct msghdr *msg, async_msg;
1206 struct kiocb *kiocb;
1209 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1211 return (struct sock_iocb *)iocb->private;
1214 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1219 struct socket_alloc {
1220 struct socket socket;
1221 struct inode vfs_inode;
1224 static inline struct socket *SOCKET_I(struct inode *inode)
1226 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1229 static inline struct inode *SOCK_INODE(struct socket *socket)
1231 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1235 * Functions for memory accounting
1237 extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
1238 extern void __sk_mem_reclaim(struct sock *sk);
1240 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1241 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1242 #define SK_MEM_SEND 0
1243 #define SK_MEM_RECV 1
1245 static inline int sk_mem_pages(int amt)
1247 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1250 static inline int sk_has_account(struct sock *sk)
1252 /* return true if protocol supports memory accounting */
1253 return !!sk->sk_prot->memory_allocated;
1256 static inline int sk_wmem_schedule(struct sock *sk, int size)
1258 if (!sk_has_account(sk))
1260 return size <= sk->sk_forward_alloc ||
1261 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1264 static inline int sk_rmem_schedule(struct sock *sk, int size)
1266 if (!sk_has_account(sk))
1268 return size <= sk->sk_forward_alloc ||
1269 __sk_mem_schedule(sk, size, SK_MEM_RECV);
1272 static inline void sk_mem_reclaim(struct sock *sk)
1274 if (!sk_has_account(sk))
1276 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1277 __sk_mem_reclaim(sk);
1280 static inline void sk_mem_reclaim_partial(struct sock *sk)
1282 if (!sk_has_account(sk))
1284 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1285 __sk_mem_reclaim(sk);
1288 static inline void sk_mem_charge(struct sock *sk, int size)
1290 if (!sk_has_account(sk))
1292 sk->sk_forward_alloc -= size;
1295 static inline void sk_mem_uncharge(struct sock *sk, int size)
1297 if (!sk_has_account(sk))
1299 sk->sk_forward_alloc += size;
1302 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1304 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1305 sk->sk_wmem_queued -= skb->truesize;
1306 sk_mem_uncharge(sk, skb->truesize);
1310 /* Used by processes to "lock" a socket state, so that
1311 * interrupts and bottom half handlers won't change it
1312 * from under us. It essentially blocks any incoming
1313 * packets, so that we won't get any new data or any
1314 * packets that change the state of the socket.
1316 * While locked, BH processing will add new packets to
1317 * the backlog queue. This queue is processed by the
1318 * owner of the socket lock right before it is released.
1320 * Since ~2.3.5 it is also exclusive sleep lock serializing
1321 * accesses from user process context.
1323 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1326 * Macro so as to not evaluate some arguments when
1327 * lockdep is not enabled.
1329 * Mark both the sk_lock and the sk_lock.slock as a
1330 * per-address-family lock class.
1332 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1334 sk->sk_lock.owned = 0; \
1335 init_waitqueue_head(&sk->sk_lock.wq); \
1336 spin_lock_init(&(sk)->sk_lock.slock); \
1337 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1338 sizeof((sk)->sk_lock)); \
1339 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1341 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1344 extern void lock_sock_nested(struct sock *sk, int subclass);
1346 static inline void lock_sock(struct sock *sk)
1348 lock_sock_nested(sk, 0);
1351 extern void release_sock(struct sock *sk);
1353 /* BH context may only use the following locking interface. */
1354 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1355 #define bh_lock_sock_nested(__sk) \
1356 spin_lock_nested(&((__sk)->sk_lock.slock), \
1357 SINGLE_DEPTH_NESTING)
1358 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1360 extern bool lock_sock_fast(struct sock *sk);
1362 * unlock_sock_fast - complement of lock_sock_fast
1366 * fast unlock socket for user context.
1367 * If slow mode is on, we call regular release_sock()
1369 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1374 spin_unlock_bh(&sk->sk_lock.slock);
1378 extern struct sock *sk_alloc(struct net *net, int family,
1380 struct proto *prot);
1381 extern void sk_free(struct sock *sk);
1382 extern void sk_release_kernel(struct sock *sk);
1383 extern struct sock *sk_clone_lock(const struct sock *sk,
1384 const gfp_t priority);
1386 extern struct sk_buff *sock_wmalloc(struct sock *sk,
1387 unsigned long size, int force,
1389 extern struct sk_buff *sock_rmalloc(struct sock *sk,
1390 unsigned long size, int force,
1392 extern void sock_wfree(struct sk_buff *skb);
1393 extern void sock_rfree(struct sk_buff *skb);
1395 extern int sock_setsockopt(struct socket *sock, int level,
1396 int op, char __user *optval,
1397 unsigned int optlen);
1399 extern int sock_getsockopt(struct socket *sock, int level,
1400 int op, char __user *optval,
1401 int __user *optlen);
1402 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1406 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1407 unsigned long header_len,
1408 unsigned long data_len,
1411 extern void *sock_kmalloc(struct sock *sk, int size,
1413 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1414 extern void sk_send_sigurg(struct sock *sk);
1416 #ifdef CONFIG_CGROUPS
1417 extern void sock_update_classid(struct sock *sk);
1419 static inline void sock_update_classid(struct sock *sk)
1425 * Functions to fill in entries in struct proto_ops when a protocol
1426 * does not implement a particular function.
1428 extern int sock_no_bind(struct socket *,
1429 struct sockaddr *, int);
1430 extern int sock_no_connect(struct socket *,
1431 struct sockaddr *, int, int);
1432 extern int sock_no_socketpair(struct socket *,
1434 extern int sock_no_accept(struct socket *,
1435 struct socket *, int);
1436 extern int sock_no_getname(struct socket *,
1437 struct sockaddr *, int *, int);
1438 extern unsigned int sock_no_poll(struct file *, struct socket *,
1439 struct poll_table_struct *);
1440 extern int sock_no_ioctl(struct socket *, unsigned int,
1442 extern int sock_no_listen(struct socket *, int);
1443 extern int sock_no_shutdown(struct socket *, int);
1444 extern int sock_no_getsockopt(struct socket *, int , int,
1445 char __user *, int __user *);
1446 extern int sock_no_setsockopt(struct socket *, int, int,
1447 char __user *, unsigned int);
1448 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1449 struct msghdr *, size_t);
1450 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1451 struct msghdr *, size_t, int);
1452 extern int sock_no_mmap(struct file *file,
1453 struct socket *sock,
1454 struct vm_area_struct *vma);
1455 extern ssize_t sock_no_sendpage(struct socket *sock,
1457 int offset, size_t size,
1461 * Functions to fill in entries in struct proto_ops when a protocol
1462 * uses the inet style.
1464 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1465 char __user *optval, int __user *optlen);
1466 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1467 struct msghdr *msg, size_t size, int flags);
1468 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
1469 char __user *optval, unsigned int optlen);
1470 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1471 int optname, char __user *optval, int __user *optlen);
1472 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
1473 int optname, char __user *optval, unsigned int optlen);
1475 extern void sk_common_release(struct sock *sk);
1478 * Default socket callbacks and setup code
1481 /* Initialise core socket variables */
1482 extern void sock_init_data(struct socket *sock, struct sock *sk);
1484 extern void sk_filter_release_rcu(struct rcu_head *rcu);
1487 * sk_filter_release - release a socket filter
1488 * @fp: filter to remove
1490 * Remove a filter from a socket and release its resources.
1493 static inline void sk_filter_release(struct sk_filter *fp)
1495 if (atomic_dec_and_test(&fp->refcnt))
1496 call_rcu(&fp->rcu, sk_filter_release_rcu);
1499 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1501 unsigned int size = sk_filter_len(fp);
1503 atomic_sub(size, &sk->sk_omem_alloc);
1504 sk_filter_release(fp);
1507 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1509 atomic_inc(&fp->refcnt);
1510 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1514 * Socket reference counting postulates.
1516 * * Each user of socket SHOULD hold a reference count.
1517 * * Each access point to socket (an hash table bucket, reference from a list,
1518 * running timer, skb in flight MUST hold a reference count.
1519 * * When reference count hits 0, it means it will never increase back.
1520 * * When reference count hits 0, it means that no references from
1521 * outside exist to this socket and current process on current CPU
1522 * is last user and may/should destroy this socket.
1523 * * sk_free is called from any context: process, BH, IRQ. When
1524 * it is called, socket has no references from outside -> sk_free
1525 * may release descendant resources allocated by the socket, but
1526 * to the time when it is called, socket is NOT referenced by any
1527 * hash tables, lists etc.
1528 * * Packets, delivered from outside (from network or from another process)
1529 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1530 * when they sit in queue. Otherwise, packets will leak to hole, when
1531 * socket is looked up by one cpu and unhasing is made by another CPU.
1532 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1533 * (leak to backlog). Packet socket does all the processing inside
1534 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1535 * use separate SMP lock, so that they are prone too.
1538 /* Ungrab socket and destroy it, if it was the last reference. */
1539 static inline void sock_put(struct sock *sk)
1541 if (atomic_dec_and_test(&sk->sk_refcnt))
1545 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1548 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1550 sk->sk_tx_queue_mapping = tx_queue;
1553 static inline void sk_tx_queue_clear(struct sock *sk)
1555 sk->sk_tx_queue_mapping = -1;
1558 static inline int sk_tx_queue_get(const struct sock *sk)
1560 return sk ? sk->sk_tx_queue_mapping : -1;
1563 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1565 sk_tx_queue_clear(sk);
1566 sk->sk_socket = sock;
1569 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1571 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1572 return &rcu_dereference_raw(sk->sk_wq)->wait;
1574 /* Detach socket from process context.
1575 * Announce socket dead, detach it from wait queue and inode.
1576 * Note that parent inode held reference count on this struct sock,
1577 * we do not release it in this function, because protocol
1578 * probably wants some additional cleanups or even continuing
1579 * to work with this socket (TCP).
1581 static inline void sock_orphan(struct sock *sk)
1583 write_lock_bh(&sk->sk_callback_lock);
1584 sock_set_flag(sk, SOCK_DEAD);
1585 sk_set_socket(sk, NULL);
1587 write_unlock_bh(&sk->sk_callback_lock);
1590 static inline void sock_graft(struct sock *sk, struct socket *parent)
1592 write_lock_bh(&sk->sk_callback_lock);
1593 sk->sk_wq = parent->wq;
1595 sk_set_socket(sk, parent);
1596 security_sock_graft(sk, parent);
1597 write_unlock_bh(&sk->sk_callback_lock);
1600 extern int sock_i_uid(struct sock *sk);
1601 extern unsigned long sock_i_ino(struct sock *sk);
1603 static inline struct dst_entry *
1604 __sk_dst_get(struct sock *sk)
1606 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
1607 lockdep_is_held(&sk->sk_lock.slock));
1610 static inline struct dst_entry *
1611 sk_dst_get(struct sock *sk)
1613 struct dst_entry *dst;
1616 dst = rcu_dereference(sk->sk_dst_cache);
1623 extern void sk_reset_txq(struct sock *sk);
1625 static inline void dst_negative_advice(struct sock *sk)
1627 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1629 if (dst && dst->ops->negative_advice) {
1630 ndst = dst->ops->negative_advice(dst);
1633 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1640 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1642 struct dst_entry *old_dst;
1644 sk_tx_queue_clear(sk);
1646 * This can be called while sk is owned by the caller only,
1647 * with no state that can be checked in a rcu_dereference_check() cond
1649 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1650 rcu_assign_pointer(sk->sk_dst_cache, dst);
1651 dst_release(old_dst);
1655 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1657 spin_lock(&sk->sk_dst_lock);
1658 __sk_dst_set(sk, dst);
1659 spin_unlock(&sk->sk_dst_lock);
1663 __sk_dst_reset(struct sock *sk)
1665 __sk_dst_set(sk, NULL);
1669 sk_dst_reset(struct sock *sk)
1671 spin_lock(&sk->sk_dst_lock);
1673 spin_unlock(&sk->sk_dst_lock);
1676 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1678 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1680 static inline int sk_can_gso(const struct sock *sk)
1682 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1685 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1687 static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
1689 sk->sk_route_nocaps |= flags;
1690 sk->sk_route_caps &= ~flags;
1693 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1694 char __user *from, char *to,
1695 int copy, int offset)
1697 if (skb->ip_summed == CHECKSUM_NONE) {
1699 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1702 skb->csum = csum_block_add(skb->csum, csum, offset);
1703 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1704 if (!access_ok(VERIFY_READ, from, copy) ||
1705 __copy_from_user_nocache(to, from, copy))
1707 } else if (copy_from_user(to, from, copy))
1713 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1714 char __user *from, int copy)
1716 int err, offset = skb->len;
1718 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1721 __skb_trim(skb, offset);
1726 static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1727 struct sk_buff *skb,
1733 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1739 skb->data_len += copy;
1740 skb->truesize += copy;
1741 sk->sk_wmem_queued += copy;
1742 sk_mem_charge(sk, copy);
1746 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1747 struct sk_buff *skb, struct page *page,
1750 if (skb->ip_summed == CHECKSUM_NONE) {
1752 __wsum csum = csum_and_copy_from_user(from,
1753 page_address(page) + off,
1757 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1758 } else if (copy_from_user(page_address(page) + off, from, copy))
1762 skb->data_len += copy;
1763 skb->truesize += copy;
1764 sk->sk_wmem_queued += copy;
1765 sk_mem_charge(sk, copy);
1770 * sk_wmem_alloc_get - returns write allocations
1773 * Returns sk_wmem_alloc minus initial offset of one
1775 static inline int sk_wmem_alloc_get(const struct sock *sk)
1777 return atomic_read(&sk->sk_wmem_alloc) - 1;
1781 * sk_rmem_alloc_get - returns read allocations
1784 * Returns sk_rmem_alloc
1786 static inline int sk_rmem_alloc_get(const struct sock *sk)
1788 return atomic_read(&sk->sk_rmem_alloc);
1792 * sk_has_allocations - check if allocations are outstanding
1795 * Returns true if socket has write or read allocations
1797 static inline int sk_has_allocations(const struct sock *sk)
1799 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1803 * wq_has_sleeper - check if there are any waiting processes
1804 * @wq: struct socket_wq
1806 * Returns true if socket_wq has waiting processes
1808 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1809 * barrier call. They were added due to the race found within the tcp code.
1811 * Consider following tcp code paths:
1815 * sys_select receive packet
1817 * __add_wait_queue update tp->rcv_nxt
1819 * tp->rcv_nxt check sock_def_readable
1821 * schedule rcu_read_lock();
1822 * wq = rcu_dereference(sk->sk_wq);
1823 * if (wq && waitqueue_active(&wq->wait))
1824 * wake_up_interruptible(&wq->wait)
1828 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1829 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1830 * could then endup calling schedule and sleep forever if there are no more
1831 * data on the socket.
1834 static inline bool wq_has_sleeper(struct socket_wq *wq)
1838 * We need to be sure we are in sync with the
1839 * add_wait_queue modifications to the wait queue.
1841 * This memory barrier is paired in the sock_poll_wait.
1844 return wq && waitqueue_active(&wq->wait);
1848 * sock_poll_wait - place memory barrier behind the poll_wait call.
1850 * @wait_address: socket wait queue
1853 * See the comments in the wq_has_sleeper function.
1855 static inline void sock_poll_wait(struct file *filp,
1856 wait_queue_head_t *wait_address, poll_table *p)
1858 if (!poll_does_not_wait(p) && wait_address) {
1859 poll_wait(filp, wait_address, p);
1861 * We need to be sure we are in sync with the
1862 * socket flags modification.
1864 * This memory barrier is paired in the wq_has_sleeper.
1871 * Queue a received datagram if it will fit. Stream and sequenced
1872 * protocols can't normally use this as they need to fit buffers in
1873 * and play with them.
1875 * Inlined as it's very short and called for pretty much every
1876 * packet ever received.
1879 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1883 skb->destructor = sock_wfree;
1885 * We used to take a refcount on sk, but following operation
1886 * is enough to guarantee sk_free() wont free this sock until
1887 * all in-flight packets are completed
1889 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1892 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1896 skb->destructor = sock_rfree;
1897 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1898 sk_mem_charge(sk, skb->truesize);
1901 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1902 unsigned long expires);
1904 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1906 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1908 extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1911 * Recover an error report and clear atomically
1914 static inline int sock_error(struct sock *sk)
1917 if (likely(!sk->sk_err))
1919 err = xchg(&sk->sk_err, 0);
1923 static inline unsigned long sock_wspace(struct sock *sk)
1927 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1928 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1935 static inline void sk_wake_async(struct sock *sk, int how, int band)
1937 if (sock_flag(sk, SOCK_FASYNC))
1938 sock_wake_async(sk->sk_socket, how, band);
1941 #define SOCK_MIN_SNDBUF 2048
1943 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1944 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1946 #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
1948 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1950 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1951 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1952 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1956 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1958 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1960 struct page *page = NULL;
1962 page = alloc_pages(sk->sk_allocation, 0);
1964 sk_enter_memory_pressure(sk);
1965 sk_stream_moderate_sndbuf(sk);
1971 * Default write policy as shown to user space via poll/select/SIGIO
1973 static inline int sock_writeable(const struct sock *sk)
1975 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1978 static inline gfp_t gfp_any(void)
1980 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1983 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1985 return noblock ? 0 : sk->sk_rcvtimeo;
1988 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1990 return noblock ? 0 : sk->sk_sndtimeo;
1993 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1995 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1998 /* Alas, with timeout socket operations are not restartable.
1999 * Compare this to poll().
2001 static inline int sock_intr_errno(long timeo)
2003 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2006 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2007 struct sk_buff *skb);
2008 extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2009 struct sk_buff *skb);
2011 static __inline__ void
2012 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2014 ktime_t kt = skb->tstamp;
2015 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
2018 * generate control messages if
2019 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2020 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2021 * - software time stamp available and wanted
2022 * (SOCK_TIMESTAMPING_SOFTWARE)
2023 * - hardware time stamps available and wanted
2024 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2025 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2027 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2028 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2029 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2030 (hwtstamps->hwtstamp.tv64 &&
2031 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2032 (hwtstamps->syststamp.tv64 &&
2033 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
2034 __sock_recv_timestamp(msg, sk, skb);
2038 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2039 __sock_recv_wifi_status(msg, sk, skb);
2042 extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2043 struct sk_buff *skb);
2045 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2046 struct sk_buff *skb)
2048 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2049 (1UL << SOCK_RCVTSTAMP) | \
2050 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2051 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
2052 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
2053 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2055 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2056 __sock_recv_ts_and_drops(msg, sk, skb);
2058 sk->sk_stamp = skb->tstamp;
2062 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2063 * @sk: socket sending this packet
2064 * @tx_flags: filled with instructions for time stamping
2066 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
2067 * parameters are invalid.
2069 extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
2072 * sk_eat_skb - Release a skb if it is no longer needed
2073 * @sk: socket to eat this skb from
2074 * @skb: socket buffer to eat
2075 * @copied_early: flag indicating whether DMA operations copied this data early
2077 * This routine must be called with interrupts disabled or with the socket
2078 * locked so that the sk_buff queue operation is ok.
2080 #ifdef CONFIG_NET_DMA
2081 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
2083 __skb_unlink(skb, &sk->sk_receive_queue);
2087 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2090 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
2092 __skb_unlink(skb, &sk->sk_receive_queue);
2098 struct net *sock_net(const struct sock *sk)
2100 return read_pnet(&sk->sk_net);
2104 void sock_net_set(struct sock *sk, struct net *net)
2106 write_pnet(&sk->sk_net, net);
2110 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
2111 * They should not hold a reference to a namespace in order to allow
2113 * Sockets after sk_change_net should be released using sk_release_kernel
2115 static inline void sk_change_net(struct sock *sk, struct net *net)
2117 put_net(sock_net(sk));
2118 sock_net_set(sk, hold_net(net));
2121 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2123 if (unlikely(skb->sk)) {
2124 struct sock *sk = skb->sk;
2126 skb->destructor = NULL;
2133 extern void sock_enable_timestamp(struct sock *sk, int flag);
2134 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
2135 extern int sock_get_timestampns(struct sock *, struct timespec __user *);
2138 * Enable debug/info messages
2140 extern int net_msg_warn;
2141 #define NETDEBUG(fmt, args...) \
2142 do { if (net_msg_warn) printk(fmt,##args); } while (0)
2144 #define LIMIT_NETDEBUG(fmt, args...) \
2145 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
2147 extern __u32 sysctl_wmem_max;
2148 extern __u32 sysctl_rmem_max;
2150 extern void sk_init(void);
2152 extern int sysctl_optmem_max;
2154 extern __u32 sysctl_wmem_default;
2155 extern __u32 sysctl_rmem_default;
2157 #endif /* _SOCK_H */