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_cgrp_prioidx: socket group's priority map index
230 * @sk_type: socket type (%SOCK_STREAM, etc)
231 * @sk_protocol: which protocol this socket belongs in this network family
232 * @sk_peer_pid: &struct pid for this socket's peer
233 * @sk_peer_cred: %SO_PEERCRED setting
234 * @sk_rcvlowat: %SO_RCVLOWAT setting
235 * @sk_rcvtimeo: %SO_RCVTIMEO setting
236 * @sk_sndtimeo: %SO_SNDTIMEO setting
237 * @sk_rxhash: flow hash received from netif layer
238 * @sk_filter: socket filtering instructions
239 * @sk_protinfo: private area, net family specific, when not using slab
240 * @sk_timer: sock cleanup timer
241 * @sk_stamp: time stamp of last packet received
242 * @sk_socket: Identd and reporting IO signals
243 * @sk_user_data: RPC layer private data
244 * @sk_sndmsg_page: cached page for sendmsg
245 * @sk_sndmsg_off: cached offset for sendmsg
246 * @sk_send_head: front of stuff to transmit
247 * @sk_security: used by security modules
248 * @sk_mark: generic packet mark
249 * @sk_classid: this socket's cgroup classid
250 * @sk_cgrp: this socket's cgroup-specific proto data
251 * @sk_write_pending: a write to stream socket waits to start
252 * @sk_state_change: callback to indicate change in the state of the sock
253 * @sk_data_ready: callback to indicate there is data to be processed
254 * @sk_write_space: callback to indicate there is bf sending space available
255 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
256 * @sk_backlog_rcv: callback to process the backlog
257 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
261 * Now struct inet_timewait_sock also uses sock_common, so please just
262 * don't add nothing before this first member (__sk_common) --acme
264 struct sock_common __sk_common
;
265 #define sk_node __sk_common.skc_node
266 #define sk_nulls_node __sk_common.skc_nulls_node
267 #define sk_refcnt __sk_common.skc_refcnt
268 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
270 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
271 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
272 #define sk_hash __sk_common.skc_hash
273 #define sk_family __sk_common.skc_family
274 #define sk_state __sk_common.skc_state
275 #define sk_reuse __sk_common.skc_reuse
276 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
277 #define sk_bind_node __sk_common.skc_bind_node
278 #define sk_prot __sk_common.skc_prot
279 #define sk_net __sk_common.skc_net
280 socket_lock_t sk_lock
;
281 struct sk_buff_head sk_receive_queue
;
283 * The backlog queue is special, it is always used with
284 * the per-socket spinlock held and requires low latency
285 * access. Therefore we special case it's implementation.
286 * Note : rmem_alloc is in this structure to fill a hole
287 * on 64bit arches, not because its logically part of
293 struct sk_buff
*head
;
294 struct sk_buff
*tail
;
296 #define sk_rmem_alloc sk_backlog.rmem_alloc
297 int sk_forward_alloc
;
304 struct sk_filter __rcu
*sk_filter
;
305 struct socket_wq __rcu
*sk_wq
;
307 #ifdef CONFIG_NET_DMA
308 struct sk_buff_head sk_async_wait_queue
;
312 struct xfrm_policy
*sk_policy
[2];
314 unsigned long sk_flags
;
315 struct dst_entry
*sk_dst_cache
;
316 spinlock_t sk_dst_lock
;
317 atomic_t sk_wmem_alloc
;
318 atomic_t sk_omem_alloc
;
320 struct sk_buff_head sk_write_queue
;
321 kmemcheck_bitfield_begin(flags
);
322 unsigned int sk_shutdown
: 2,
327 kmemcheck_bitfield_end(flags
);
330 netdev_features_t sk_route_caps
;
331 netdev_features_t sk_route_nocaps
;
333 unsigned int sk_gso_max_size
;
335 unsigned long sk_lingertime
;
336 struct sk_buff_head sk_error_queue
;
337 struct proto
*sk_prot_creator
;
338 rwlock_t sk_callback_lock
;
341 unsigned short sk_ack_backlog
;
342 unsigned short sk_max_ack_backlog
;
344 #ifdef CONFIG_CGROUPS
345 __u32 sk_cgrp_prioidx
;
347 struct pid
*sk_peer_pid
;
348 const struct cred
*sk_peer_cred
;
352 struct timer_list sk_timer
;
354 struct socket
*sk_socket
;
356 struct page
*sk_sndmsg_page
;
357 struct sk_buff
*sk_send_head
;
359 int sk_write_pending
;
360 #ifdef CONFIG_SECURITY
365 struct cg_proto
*sk_cgrp
;
366 void (*sk_state_change
)(struct sock
*sk
);
367 void (*sk_data_ready
)(struct sock
*sk
, int bytes
);
368 void (*sk_write_space
)(struct sock
*sk
);
369 void (*sk_error_report
)(struct sock
*sk
);
370 int (*sk_backlog_rcv
)(struct sock
*sk
,
371 struct sk_buff
*skb
);
372 void (*sk_destruct
)(struct sock
*sk
);
376 * Hashed lists helper routines
378 static inline struct sock
*sk_entry(const struct hlist_node
*node
)
380 return hlist_entry(node
, struct sock
, sk_node
);
383 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
385 return hlist_entry(head
->first
, struct sock
, sk_node
);
388 static inline struct sock
*sk_head(const struct hlist_head
*head
)
390 return hlist_empty(head
) ? NULL
: __sk_head(head
);
393 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
395 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
398 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
400 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
403 static inline struct sock
*sk_next(const struct sock
*sk
)
405 return sk
->sk_node
.next
?
406 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
409 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
411 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
412 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
413 struct sock
, sk_nulls_node
) :
417 static inline int sk_unhashed(const struct sock
*sk
)
419 return hlist_unhashed(&sk
->sk_node
);
422 static inline int sk_hashed(const struct sock
*sk
)
424 return !sk_unhashed(sk
);
427 static __inline__
void sk_node_init(struct hlist_node
*node
)
432 static __inline__
void sk_nulls_node_init(struct hlist_nulls_node
*node
)
437 static __inline__
void __sk_del_node(struct sock
*sk
)
439 __hlist_del(&sk
->sk_node
);
442 /* NB: equivalent to hlist_del_init_rcu */
443 static __inline__
int __sk_del_node_init(struct sock
*sk
)
447 sk_node_init(&sk
->sk_node
);
453 /* Grab socket reference count. This operation is valid only
454 when sk is ALREADY grabbed f.e. it is found in hash table
455 or a list and the lookup is made under lock preventing hash table
459 static inline void sock_hold(struct sock
*sk
)
461 atomic_inc(&sk
->sk_refcnt
);
464 /* Ungrab socket in the context, which assumes that socket refcnt
465 cannot hit zero, f.e. it is true in context of any socketcall.
467 static inline void __sock_put(struct sock
*sk
)
469 atomic_dec(&sk
->sk_refcnt
);
472 static __inline__
int sk_del_node_init(struct sock
*sk
)
474 int rc
= __sk_del_node_init(sk
);
477 /* paranoid for a while -acme */
478 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
483 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
485 static __inline__
int __sk_nulls_del_node_init_rcu(struct sock
*sk
)
488 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
494 static __inline__
int sk_nulls_del_node_init_rcu(struct sock
*sk
)
496 int rc
= __sk_nulls_del_node_init_rcu(sk
);
499 /* paranoid for a while -acme */
500 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
506 static __inline__
void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
508 hlist_add_head(&sk
->sk_node
, list
);
511 static __inline__
void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
514 __sk_add_node(sk
, list
);
517 static __inline__
void sk_add_node_rcu(struct sock
*sk
, struct hlist_head
*list
)
520 hlist_add_head_rcu(&sk
->sk_node
, list
);
523 static __inline__
void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
525 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
528 static __inline__
void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
531 __sk_nulls_add_node_rcu(sk
, list
);
534 static __inline__
void __sk_del_bind_node(struct sock
*sk
)
536 __hlist_del(&sk
->sk_bind_node
);
539 static __inline__
void sk_add_bind_node(struct sock
*sk
,
540 struct hlist_head
*list
)
542 hlist_add_head(&sk
->sk_bind_node
, list
);
545 #define sk_for_each(__sk, node, list) \
546 hlist_for_each_entry(__sk, node, list, sk_node)
547 #define sk_for_each_rcu(__sk, node, list) \
548 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
549 #define sk_nulls_for_each(__sk, node, list) \
550 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
551 #define sk_nulls_for_each_rcu(__sk, node, list) \
552 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
553 #define sk_for_each_from(__sk, node) \
554 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
555 hlist_for_each_entry_from(__sk, node, sk_node)
556 #define sk_nulls_for_each_from(__sk, node) \
557 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
558 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
559 #define sk_for_each_safe(__sk, node, tmp, list) \
560 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
561 #define sk_for_each_bound(__sk, node, list) \
562 hlist_for_each_entry(__sk, node, list, sk_bind_node)
575 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
576 SOCK_DBG
, /* %SO_DEBUG setting */
577 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
578 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
579 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
580 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
581 SOCK_TIMESTAMPING_TX_HARDWARE
, /* %SOF_TIMESTAMPING_TX_HARDWARE */
582 SOCK_TIMESTAMPING_TX_SOFTWARE
, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
583 SOCK_TIMESTAMPING_RX_HARDWARE
, /* %SOF_TIMESTAMPING_RX_HARDWARE */
584 SOCK_TIMESTAMPING_RX_SOFTWARE
, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
585 SOCK_TIMESTAMPING_SOFTWARE
, /* %SOF_TIMESTAMPING_SOFTWARE */
586 SOCK_TIMESTAMPING_RAW_HARDWARE
, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
587 SOCK_TIMESTAMPING_SYS_HARDWARE
, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
588 SOCK_FASYNC
, /* fasync() active */
590 SOCK_ZEROCOPY
, /* buffers from userspace */
591 SOCK_WIFI_STATUS
, /* push wifi status to userspace */
594 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
596 nsk
->sk_flags
= osk
->sk_flags
;
599 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
601 __set_bit(flag
, &sk
->sk_flags
);
604 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
606 __clear_bit(flag
, &sk
->sk_flags
);
609 static inline int sock_flag(struct sock
*sk
, enum sock_flags flag
)
611 return test_bit(flag
, &sk
->sk_flags
);
614 static inline void sk_acceptq_removed(struct sock
*sk
)
616 sk
->sk_ack_backlog
--;
619 static inline void sk_acceptq_added(struct sock
*sk
)
621 sk
->sk_ack_backlog
++;
624 static inline int sk_acceptq_is_full(struct sock
*sk
)
626 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
630 * Compute minimal free write space needed to queue new packets.
632 static inline int sk_stream_min_wspace(struct sock
*sk
)
634 return sk
->sk_wmem_queued
>> 1;
637 static inline int sk_stream_wspace(struct sock
*sk
)
639 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
642 extern void sk_stream_write_space(struct sock
*sk
);
644 static inline int sk_stream_memory_free(struct sock
*sk
)
646 return sk
->sk_wmem_queued
< sk
->sk_sndbuf
;
649 /* OOB backlog add */
650 static inline void __sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
652 /* dont let skb dst not refcounted, we are going to leave rcu lock */
655 if (!sk
->sk_backlog
.tail
)
656 sk
->sk_backlog
.head
= skb
;
658 sk
->sk_backlog
.tail
->next
= skb
;
660 sk
->sk_backlog
.tail
= skb
;
665 * Take into account size of receive queue and backlog queue
666 * Do not take into account this skb truesize,
667 * to allow even a single big packet to come.
669 static inline bool sk_rcvqueues_full(const struct sock
*sk
, const struct sk_buff
*skb
)
671 unsigned int qsize
= sk
->sk_backlog
.len
+ atomic_read(&sk
->sk_rmem_alloc
);
673 return qsize
> sk
->sk_rcvbuf
;
676 /* The per-socket spinlock must be held here. */
677 static inline __must_check
int sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
679 if (sk_rcvqueues_full(sk
, skb
))
682 __sk_add_backlog(sk
, skb
);
683 sk
->sk_backlog
.len
+= skb
->truesize
;
687 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
689 return sk
->sk_backlog_rcv(sk
, skb
);
692 static inline void sock_rps_record_flow(const struct sock
*sk
)
695 struct rps_sock_flow_table
*sock_flow_table
;
698 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
699 rps_record_sock_flow(sock_flow_table
, sk
->sk_rxhash
);
704 static inline void sock_rps_reset_flow(const struct sock
*sk
)
707 struct rps_sock_flow_table
*sock_flow_table
;
710 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
711 rps_reset_sock_flow(sock_flow_table
, sk
->sk_rxhash
);
716 static inline void sock_rps_save_rxhash(struct sock
*sk
,
717 const struct sk_buff
*skb
)
720 if (unlikely(sk
->sk_rxhash
!= skb
->rxhash
)) {
721 sock_rps_reset_flow(sk
);
722 sk
->sk_rxhash
= skb
->rxhash
;
727 static inline void sock_rps_reset_rxhash(struct sock
*sk
)
730 sock_rps_reset_flow(sk
);
735 #define sk_wait_event(__sk, __timeo, __condition) \
737 release_sock(__sk); \
738 __rc = __condition; \
740 *(__timeo) = schedule_timeout(*(__timeo)); \
743 __rc = __condition; \
747 extern int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
748 extern int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
749 extern void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
750 extern int sk_stream_error(struct sock
*sk
, int flags
, int err
);
751 extern void sk_stream_kill_queues(struct sock
*sk
);
753 extern int sk_wait_data(struct sock
*sk
, long *timeo
);
755 struct request_sock_ops
;
756 struct timewait_sock_ops
;
757 struct inet_hashinfo
;
761 /* Networking protocol blocks we attach to sockets.
762 * socket layer -> transport layer interface
763 * transport -> network interface is defined by struct inet_proto
766 void (*close
)(struct sock
*sk
,
768 int (*connect
)(struct sock
*sk
,
769 struct sockaddr
*uaddr
,
771 int (*disconnect
)(struct sock
*sk
, int flags
);
773 struct sock
* (*accept
) (struct sock
*sk
, int flags
, int *err
);
775 int (*ioctl
)(struct sock
*sk
, int cmd
,
777 int (*init
)(struct sock
*sk
);
778 void (*destroy
)(struct sock
*sk
);
779 void (*shutdown
)(struct sock
*sk
, int how
);
780 int (*setsockopt
)(struct sock
*sk
, int level
,
781 int optname
, char __user
*optval
,
782 unsigned int optlen
);
783 int (*getsockopt
)(struct sock
*sk
, int level
,
784 int optname
, char __user
*optval
,
787 int (*compat_setsockopt
)(struct sock
*sk
,
789 int optname
, char __user
*optval
,
790 unsigned int optlen
);
791 int (*compat_getsockopt
)(struct sock
*sk
,
793 int optname
, char __user
*optval
,
795 int (*compat_ioctl
)(struct sock
*sk
,
796 unsigned int cmd
, unsigned long arg
);
798 int (*sendmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
799 struct msghdr
*msg
, size_t len
);
800 int (*recvmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
802 size_t len
, int noblock
, int flags
,
804 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
805 int offset
, size_t size
, int flags
);
806 int (*bind
)(struct sock
*sk
,
807 struct sockaddr
*uaddr
, int addr_len
);
809 int (*backlog_rcv
) (struct sock
*sk
,
810 struct sk_buff
*skb
);
812 /* Keeping track of sk's, looking them up, and port selection methods. */
813 void (*hash
)(struct sock
*sk
);
814 void (*unhash
)(struct sock
*sk
);
815 void (*rehash
)(struct sock
*sk
);
816 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
817 void (*clear_sk
)(struct sock
*sk
, int size
);
819 /* Keeping track of sockets in use */
820 #ifdef CONFIG_PROC_FS
821 unsigned int inuse_idx
;
824 /* Memory pressure */
825 void (*enter_memory_pressure
)(struct sock
*sk
);
826 atomic_long_t
*memory_allocated
; /* Current allocated memory. */
827 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
829 * Pressure flag: try to collapse.
830 * Technical note: it is used by multiple contexts non atomically.
831 * All the __sk_mem_schedule() is of this nature: accounting
832 * is strict, actions are advisory and have some latency.
834 int *memory_pressure
;
841 struct kmem_cache
*slab
;
842 unsigned int obj_size
;
845 struct percpu_counter
*orphan_count
;
847 struct request_sock_ops
*rsk_prot
;
848 struct timewait_sock_ops
*twsk_prot
;
851 struct inet_hashinfo
*hashinfo
;
852 struct udp_table
*udp_table
;
853 struct raw_hashinfo
*raw_hash
;
856 struct module
*owner
;
860 struct list_head node
;
861 #ifdef SOCK_REFCNT_DEBUG
864 #ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
866 * cgroup specific init/deinit functions. Called once for all
867 * protocols that implement it, from cgroups populate function.
868 * This function has to setup any files the protocol want to
869 * appear in the kmem cgroup filesystem.
871 int (*init_cgroup
)(struct cgroup
*cgrp
,
872 struct cgroup_subsys
*ss
);
873 void (*destroy_cgroup
)(struct cgroup
*cgrp
,
874 struct cgroup_subsys
*ss
);
875 struct cg_proto
*(*proto_cgroup
)(struct mem_cgroup
*memcg
);
880 void (*enter_memory_pressure
)(struct sock
*sk
);
881 struct res_counter
*memory_allocated
; /* Current allocated memory. */
882 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
883 int *memory_pressure
;
886 * memcg field is used to find which memcg we belong directly
887 * Each memcg struct can hold more than one cg_proto, so container_of
890 * The elegant solution would be having an inverse function to
891 * proto_cgroup in struct proto, but that means polluting the structure
892 * for everybody, instead of just for memcg users.
894 struct mem_cgroup
*memcg
;
897 extern int proto_register(struct proto
*prot
, int alloc_slab
);
898 extern void proto_unregister(struct proto
*prot
);
900 #ifdef SOCK_REFCNT_DEBUG
901 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
903 atomic_inc(&sk
->sk_prot
->socks
);
906 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
908 atomic_dec(&sk
->sk_prot
->socks
);
909 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
910 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
913 inline void sk_refcnt_debug_release(const struct sock
*sk
)
915 if (atomic_read(&sk
->sk_refcnt
) != 1)
916 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
917 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
919 #else /* SOCK_REFCNT_DEBUG */
920 #define sk_refcnt_debug_inc(sk) do { } while (0)
921 #define sk_refcnt_debug_dec(sk) do { } while (0)
922 #define sk_refcnt_debug_release(sk) do { } while (0)
923 #endif /* SOCK_REFCNT_DEBUG */
925 #ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
926 extern struct jump_label_key memcg_socket_limit_enabled
;
927 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
928 struct cg_proto
*cg_proto
)
930 return proto
->proto_cgroup(parent_mem_cgroup(cg_proto
->memcg
));
932 #define mem_cgroup_sockets_enabled static_branch(&memcg_socket_limit_enabled)
934 #define mem_cgroup_sockets_enabled 0
935 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
936 struct cg_proto
*cg_proto
)
943 static inline bool sk_has_memory_pressure(const struct sock
*sk
)
945 return sk
->sk_prot
->memory_pressure
!= NULL
;
948 static inline bool sk_under_memory_pressure(const struct sock
*sk
)
950 if (!sk
->sk_prot
->memory_pressure
)
953 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
954 return !!*sk
->sk_cgrp
->memory_pressure
;
956 return !!*sk
->sk_prot
->memory_pressure
;
959 static inline void sk_leave_memory_pressure(struct sock
*sk
)
961 int *memory_pressure
= sk
->sk_prot
->memory_pressure
;
963 if (!memory_pressure
)
966 if (*memory_pressure
)
967 *memory_pressure
= 0;
969 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
970 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
971 struct proto
*prot
= sk
->sk_prot
;
973 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
974 if (*cg_proto
->memory_pressure
)
975 *cg_proto
->memory_pressure
= 0;
980 static inline void sk_enter_memory_pressure(struct sock
*sk
)
982 if (!sk
->sk_prot
->enter_memory_pressure
)
985 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
986 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
987 struct proto
*prot
= sk
->sk_prot
;
989 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
990 cg_proto
->enter_memory_pressure(sk
);
993 sk
->sk_prot
->enter_memory_pressure(sk
);
996 static inline long sk_prot_mem_limits(const struct sock
*sk
, int index
)
998 long *prot
= sk
->sk_prot
->sysctl_mem
;
999 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1000 prot
= sk
->sk_cgrp
->sysctl_mem
;
1004 static inline void memcg_memory_allocated_add(struct cg_proto
*prot
,
1008 struct res_counter
*fail
;
1011 ret
= res_counter_charge(prot
->memory_allocated
,
1012 amt
<< PAGE_SHIFT
, &fail
);
1015 *parent_status
= OVER_LIMIT
;
1018 static inline void memcg_memory_allocated_sub(struct cg_proto
*prot
,
1021 res_counter_uncharge(prot
->memory_allocated
, amt
<< PAGE_SHIFT
);
1024 static inline u64
memcg_memory_allocated_read(struct cg_proto
*prot
)
1027 ret
= res_counter_read_u64(prot
->memory_allocated
, RES_USAGE
);
1028 return ret
>> PAGE_SHIFT
;
1032 sk_memory_allocated(const struct sock
*sk
)
1034 struct proto
*prot
= sk
->sk_prot
;
1035 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1036 return memcg_memory_allocated_read(sk
->sk_cgrp
);
1038 return atomic_long_read(prot
->memory_allocated
);
1042 sk_memory_allocated_add(struct sock
*sk
, int amt
, int *parent_status
)
1044 struct proto
*prot
= sk
->sk_prot
;
1046 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1047 memcg_memory_allocated_add(sk
->sk_cgrp
, amt
, parent_status
);
1048 /* update the root cgroup regardless */
1049 atomic_long_add_return(amt
, prot
->memory_allocated
);
1050 return memcg_memory_allocated_read(sk
->sk_cgrp
);
1053 return atomic_long_add_return(amt
, prot
->memory_allocated
);
1057 sk_memory_allocated_sub(struct sock
*sk
, int amt
, int parent_status
)
1059 struct proto
*prot
= sk
->sk_prot
;
1061 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
&&
1062 parent_status
!= OVER_LIMIT
) /* Otherwise was uncharged already */
1063 memcg_memory_allocated_sub(sk
->sk_cgrp
, amt
);
1065 atomic_long_sub(amt
, prot
->memory_allocated
);
1068 static inline void sk_sockets_allocated_dec(struct sock
*sk
)
1070 struct proto
*prot
= sk
->sk_prot
;
1072 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1073 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1075 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1076 percpu_counter_dec(cg_proto
->sockets_allocated
);
1079 percpu_counter_dec(prot
->sockets_allocated
);
1082 static inline void sk_sockets_allocated_inc(struct sock
*sk
)
1084 struct proto
*prot
= sk
->sk_prot
;
1086 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1087 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1089 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1090 percpu_counter_inc(cg_proto
->sockets_allocated
);
1093 percpu_counter_inc(prot
->sockets_allocated
);
1097 sk_sockets_allocated_read_positive(struct sock
*sk
)
1099 struct proto
*prot
= sk
->sk_prot
;
1101 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1102 return percpu_counter_sum_positive(sk
->sk_cgrp
->sockets_allocated
);
1104 return percpu_counter_sum_positive(prot
->sockets_allocated
);
1108 proto_sockets_allocated_sum_positive(struct proto
*prot
)
1110 return percpu_counter_sum_positive(prot
->sockets_allocated
);
1114 proto_memory_allocated(struct proto
*prot
)
1116 return atomic_long_read(prot
->memory_allocated
);
1120 proto_memory_pressure(struct proto
*prot
)
1122 if (!prot
->memory_pressure
)
1124 return !!*prot
->memory_pressure
;
1128 #ifdef CONFIG_PROC_FS
1129 /* Called with local bh disabled */
1130 extern void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int inc
);
1131 extern int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
1133 static void inline sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
1140 /* With per-bucket locks this operation is not-atomic, so that
1141 * this version is not worse.
1143 static inline void __sk_prot_rehash(struct sock
*sk
)
1145 sk
->sk_prot
->unhash(sk
);
1146 sk
->sk_prot
->hash(sk
);
1149 void sk_prot_clear_portaddr_nulls(struct sock
*sk
, int size
);
1151 /* About 10 seconds */
1152 #define SOCK_DESTROY_TIME (10*HZ)
1154 /* Sockets 0-1023 can't be bound to unless you are superuser */
1155 #define PROT_SOCK 1024
1157 #define SHUTDOWN_MASK 3
1158 #define RCV_SHUTDOWN 1
1159 #define SEND_SHUTDOWN 2
1161 #define SOCK_SNDBUF_LOCK 1
1162 #define SOCK_RCVBUF_LOCK 2
1163 #define SOCK_BINDADDR_LOCK 4
1164 #define SOCK_BINDPORT_LOCK 8
1166 /* sock_iocb: used to kick off async processing of socket ios */
1168 struct list_head list
;
1172 struct socket
*sock
;
1174 struct scm_cookie
*scm
;
1175 struct msghdr
*msg
, async_msg
;
1176 struct kiocb
*kiocb
;
1179 static inline struct sock_iocb
*kiocb_to_siocb(struct kiocb
*iocb
)
1181 return (struct sock_iocb
*)iocb
->private;
1184 static inline struct kiocb
*siocb_to_kiocb(struct sock_iocb
*si
)
1189 struct socket_alloc
{
1190 struct socket socket
;
1191 struct inode vfs_inode
;
1194 static inline struct socket
*SOCKET_I(struct inode
*inode
)
1196 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
1199 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
1201 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
1205 * Functions for memory accounting
1207 extern int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
1208 extern void __sk_mem_reclaim(struct sock
*sk
);
1210 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1211 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1212 #define SK_MEM_SEND 0
1213 #define SK_MEM_RECV 1
1215 static inline int sk_mem_pages(int amt
)
1217 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
1220 static inline int sk_has_account(struct sock
*sk
)
1222 /* return true if protocol supports memory accounting */
1223 return !!sk
->sk_prot
->memory_allocated
;
1226 static inline int sk_wmem_schedule(struct sock
*sk
, int size
)
1228 if (!sk_has_account(sk
))
1230 return size
<= sk
->sk_forward_alloc
||
1231 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
1234 static inline int sk_rmem_schedule(struct sock
*sk
, int size
)
1236 if (!sk_has_account(sk
))
1238 return size
<= sk
->sk_forward_alloc
||
1239 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
);
1242 static inline void sk_mem_reclaim(struct sock
*sk
)
1244 if (!sk_has_account(sk
))
1246 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
1247 __sk_mem_reclaim(sk
);
1250 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
1252 if (!sk_has_account(sk
))
1254 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
1255 __sk_mem_reclaim(sk
);
1258 static inline void sk_mem_charge(struct sock
*sk
, int size
)
1260 if (!sk_has_account(sk
))
1262 sk
->sk_forward_alloc
-= size
;
1265 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
1267 if (!sk_has_account(sk
))
1269 sk
->sk_forward_alloc
+= size
;
1272 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
1274 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
1275 sk
->sk_wmem_queued
-= skb
->truesize
;
1276 sk_mem_uncharge(sk
, skb
->truesize
);
1280 /* Used by processes to "lock" a socket state, so that
1281 * interrupts and bottom half handlers won't change it
1282 * from under us. It essentially blocks any incoming
1283 * packets, so that we won't get any new data or any
1284 * packets that change the state of the socket.
1286 * While locked, BH processing will add new packets to
1287 * the backlog queue. This queue is processed by the
1288 * owner of the socket lock right before it is released.
1290 * Since ~2.3.5 it is also exclusive sleep lock serializing
1291 * accesses from user process context.
1293 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1296 * Macro so as to not evaluate some arguments when
1297 * lockdep is not enabled.
1299 * Mark both the sk_lock and the sk_lock.slock as a
1300 * per-address-family lock class.
1302 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1304 sk->sk_lock.owned = 0; \
1305 init_waitqueue_head(&sk->sk_lock.wq); \
1306 spin_lock_init(&(sk)->sk_lock.slock); \
1307 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1308 sizeof((sk)->sk_lock)); \
1309 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1311 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1314 extern void lock_sock_nested(struct sock
*sk
, int subclass
);
1316 static inline void lock_sock(struct sock
*sk
)
1318 lock_sock_nested(sk
, 0);
1321 extern void release_sock(struct sock
*sk
);
1323 /* BH context may only use the following locking interface. */
1324 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1325 #define bh_lock_sock_nested(__sk) \
1326 spin_lock_nested(&((__sk)->sk_lock.slock), \
1327 SINGLE_DEPTH_NESTING)
1328 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1330 extern bool lock_sock_fast(struct sock
*sk
);
1332 * unlock_sock_fast - complement of lock_sock_fast
1336 * fast unlock socket for user context.
1337 * If slow mode is on, we call regular release_sock()
1339 static inline void unlock_sock_fast(struct sock
*sk
, bool slow
)
1344 spin_unlock_bh(&sk
->sk_lock
.slock
);
1348 extern struct sock
*sk_alloc(struct net
*net
, int family
,
1350 struct proto
*prot
);
1351 extern void sk_free(struct sock
*sk
);
1352 extern void sk_release_kernel(struct sock
*sk
);
1353 extern struct sock
*sk_clone_lock(const struct sock
*sk
,
1354 const gfp_t priority
);
1356 extern struct sk_buff
*sock_wmalloc(struct sock
*sk
,
1357 unsigned long size
, int force
,
1359 extern struct sk_buff
*sock_rmalloc(struct sock
*sk
,
1360 unsigned long size
, int force
,
1362 extern void sock_wfree(struct sk_buff
*skb
);
1363 extern void sock_rfree(struct sk_buff
*skb
);
1365 extern int sock_setsockopt(struct socket
*sock
, int level
,
1366 int op
, char __user
*optval
,
1367 unsigned int optlen
);
1369 extern int sock_getsockopt(struct socket
*sock
, int level
,
1370 int op
, char __user
*optval
,
1371 int __user
*optlen
);
1372 extern struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
,
1376 extern struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
,
1377 unsigned long header_len
,
1378 unsigned long data_len
,
1381 extern void *sock_kmalloc(struct sock
*sk
, int size
,
1383 extern void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
1384 extern void sk_send_sigurg(struct sock
*sk
);
1386 #ifdef CONFIG_CGROUPS
1387 extern void sock_update_classid(struct sock
*sk
);
1389 static inline void sock_update_classid(struct sock
*sk
)
1395 * Functions to fill in entries in struct proto_ops when a protocol
1396 * does not implement a particular function.
1398 extern int sock_no_bind(struct socket
*,
1399 struct sockaddr
*, int);
1400 extern int sock_no_connect(struct socket
*,
1401 struct sockaddr
*, int, int);
1402 extern int sock_no_socketpair(struct socket
*,
1404 extern int sock_no_accept(struct socket
*,
1405 struct socket
*, int);
1406 extern int sock_no_getname(struct socket
*,
1407 struct sockaddr
*, int *, int);
1408 extern unsigned int sock_no_poll(struct file
*, struct socket
*,
1409 struct poll_table_struct
*);
1410 extern int sock_no_ioctl(struct socket
*, unsigned int,
1412 extern int sock_no_listen(struct socket
*, int);
1413 extern int sock_no_shutdown(struct socket
*, int);
1414 extern int sock_no_getsockopt(struct socket
*, int , int,
1415 char __user
*, int __user
*);
1416 extern int sock_no_setsockopt(struct socket
*, int, int,
1417 char __user
*, unsigned int);
1418 extern int sock_no_sendmsg(struct kiocb
*, struct socket
*,
1419 struct msghdr
*, size_t);
1420 extern int sock_no_recvmsg(struct kiocb
*, struct socket
*,
1421 struct msghdr
*, size_t, int);
1422 extern int sock_no_mmap(struct file
*file
,
1423 struct socket
*sock
,
1424 struct vm_area_struct
*vma
);
1425 extern ssize_t
sock_no_sendpage(struct socket
*sock
,
1427 int offset
, size_t size
,
1431 * Functions to fill in entries in struct proto_ops when a protocol
1432 * uses the inet style.
1434 extern int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1435 char __user
*optval
, int __user
*optlen
);
1436 extern int sock_common_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
1437 struct msghdr
*msg
, size_t size
, int flags
);
1438 extern int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1439 char __user
*optval
, unsigned int optlen
);
1440 extern int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
1441 int optname
, char __user
*optval
, int __user
*optlen
);
1442 extern int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
1443 int optname
, char __user
*optval
, unsigned int optlen
);
1445 extern void sk_common_release(struct sock
*sk
);
1448 * Default socket callbacks and setup code
1451 /* Initialise core socket variables */
1452 extern void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1454 extern void sk_filter_release_rcu(struct rcu_head
*rcu
);
1457 * sk_filter_release - release a socket filter
1458 * @fp: filter to remove
1460 * Remove a filter from a socket and release its resources.
1463 static inline void sk_filter_release(struct sk_filter
*fp
)
1465 if (atomic_dec_and_test(&fp
->refcnt
))
1466 call_rcu(&fp
->rcu
, sk_filter_release_rcu
);
1469 static inline void sk_filter_uncharge(struct sock
*sk
, struct sk_filter
*fp
)
1471 unsigned int size
= sk_filter_len(fp
);
1473 atomic_sub(size
, &sk
->sk_omem_alloc
);
1474 sk_filter_release(fp
);
1477 static inline void sk_filter_charge(struct sock
*sk
, struct sk_filter
*fp
)
1479 atomic_inc(&fp
->refcnt
);
1480 atomic_add(sk_filter_len(fp
), &sk
->sk_omem_alloc
);
1484 * Socket reference counting postulates.
1486 * * Each user of socket SHOULD hold a reference count.
1487 * * Each access point to socket (an hash table bucket, reference from a list,
1488 * running timer, skb in flight MUST hold a reference count.
1489 * * When reference count hits 0, it means it will never increase back.
1490 * * When reference count hits 0, it means that no references from
1491 * outside exist to this socket and current process on current CPU
1492 * is last user and may/should destroy this socket.
1493 * * sk_free is called from any context: process, BH, IRQ. When
1494 * it is called, socket has no references from outside -> sk_free
1495 * may release descendant resources allocated by the socket, but
1496 * to the time when it is called, socket is NOT referenced by any
1497 * hash tables, lists etc.
1498 * * Packets, delivered from outside (from network or from another process)
1499 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1500 * when they sit in queue. Otherwise, packets will leak to hole, when
1501 * socket is looked up by one cpu and unhasing is made by another CPU.
1502 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1503 * (leak to backlog). Packet socket does all the processing inside
1504 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1505 * use separate SMP lock, so that they are prone too.
1508 /* Ungrab socket and destroy it, if it was the last reference. */
1509 static inline void sock_put(struct sock
*sk
)
1511 if (atomic_dec_and_test(&sk
->sk_refcnt
))
1515 extern int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
1518 static inline void sk_tx_queue_set(struct sock
*sk
, int tx_queue
)
1520 sk
->sk_tx_queue_mapping
= tx_queue
;
1523 static inline void sk_tx_queue_clear(struct sock
*sk
)
1525 sk
->sk_tx_queue_mapping
= -1;
1528 static inline int sk_tx_queue_get(const struct sock
*sk
)
1530 return sk
? sk
->sk_tx_queue_mapping
: -1;
1533 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1535 sk_tx_queue_clear(sk
);
1536 sk
->sk_socket
= sock
;
1539 static inline wait_queue_head_t
*sk_sleep(struct sock
*sk
)
1541 BUILD_BUG_ON(offsetof(struct socket_wq
, wait
) != 0);
1542 return &rcu_dereference_raw(sk
->sk_wq
)->wait
;
1544 /* Detach socket from process context.
1545 * Announce socket dead, detach it from wait queue and inode.
1546 * Note that parent inode held reference count on this struct sock,
1547 * we do not release it in this function, because protocol
1548 * probably wants some additional cleanups or even continuing
1549 * to work with this socket (TCP).
1551 static inline void sock_orphan(struct sock
*sk
)
1553 write_lock_bh(&sk
->sk_callback_lock
);
1554 sock_set_flag(sk
, SOCK_DEAD
);
1555 sk_set_socket(sk
, NULL
);
1557 write_unlock_bh(&sk
->sk_callback_lock
);
1560 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1562 write_lock_bh(&sk
->sk_callback_lock
);
1563 sk
->sk_wq
= parent
->wq
;
1565 sk_set_socket(sk
, parent
);
1566 security_sock_graft(sk
, parent
);
1567 write_unlock_bh(&sk
->sk_callback_lock
);
1570 extern int sock_i_uid(struct sock
*sk
);
1571 extern unsigned long sock_i_ino(struct sock
*sk
);
1573 static inline struct dst_entry
*
1574 __sk_dst_get(struct sock
*sk
)
1576 return rcu_dereference_check(sk
->sk_dst_cache
, sock_owned_by_user(sk
) ||
1577 lockdep_is_held(&sk
->sk_lock
.slock
));
1580 static inline struct dst_entry
*
1581 sk_dst_get(struct sock
*sk
)
1583 struct dst_entry
*dst
;
1586 dst
= rcu_dereference(sk
->sk_dst_cache
);
1593 extern void sk_reset_txq(struct sock
*sk
);
1595 static inline void dst_negative_advice(struct sock
*sk
)
1597 struct dst_entry
*ndst
, *dst
= __sk_dst_get(sk
);
1599 if (dst
&& dst
->ops
->negative_advice
) {
1600 ndst
= dst
->ops
->negative_advice(dst
);
1603 rcu_assign_pointer(sk
->sk_dst_cache
, ndst
);
1610 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1612 struct dst_entry
*old_dst
;
1614 sk_tx_queue_clear(sk
);
1616 * This can be called while sk is owned by the caller only,
1617 * with no state that can be checked in a rcu_dereference_check() cond
1619 old_dst
= rcu_dereference_raw(sk
->sk_dst_cache
);
1620 rcu_assign_pointer(sk
->sk_dst_cache
, dst
);
1621 dst_release(old_dst
);
1625 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1627 spin_lock(&sk
->sk_dst_lock
);
1628 __sk_dst_set(sk
, dst
);
1629 spin_unlock(&sk
->sk_dst_lock
);
1633 __sk_dst_reset(struct sock
*sk
)
1635 __sk_dst_set(sk
, NULL
);
1639 sk_dst_reset(struct sock
*sk
)
1641 spin_lock(&sk
->sk_dst_lock
);
1643 spin_unlock(&sk
->sk_dst_lock
);
1646 extern struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1648 extern struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1650 static inline int sk_can_gso(const struct sock
*sk
)
1652 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1655 extern void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1657 static inline void sk_nocaps_add(struct sock
*sk
, netdev_features_t flags
)
1659 sk
->sk_route_nocaps
|= flags
;
1660 sk
->sk_route_caps
&= ~flags
;
1663 static inline int skb_do_copy_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1664 char __user
*from
, char *to
,
1665 int copy
, int offset
)
1667 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1669 __wsum csum
= csum_and_copy_from_user(from
, to
, copy
, 0, &err
);
1672 skb
->csum
= csum_block_add(skb
->csum
, csum
, offset
);
1673 } else if (sk
->sk_route_caps
& NETIF_F_NOCACHE_COPY
) {
1674 if (!access_ok(VERIFY_READ
, from
, copy
) ||
1675 __copy_from_user_nocache(to
, from
, copy
))
1677 } else if (copy_from_user(to
, from
, copy
))
1683 static inline int skb_add_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1684 char __user
*from
, int copy
)
1686 int err
, offset
= skb
->len
;
1688 err
= skb_do_copy_data_nocache(sk
, skb
, from
, skb_put(skb
, copy
),
1691 __skb_trim(skb
, offset
);
1696 static inline int skb_copy_to_page_nocache(struct sock
*sk
, char __user
*from
,
1697 struct sk_buff
*skb
,
1703 err
= skb_do_copy_data_nocache(sk
, skb
, from
, page_address(page
) + off
,
1709 skb
->data_len
+= copy
;
1710 skb
->truesize
+= copy
;
1711 sk
->sk_wmem_queued
+= copy
;
1712 sk_mem_charge(sk
, copy
);
1716 static inline int skb_copy_to_page(struct sock
*sk
, char __user
*from
,
1717 struct sk_buff
*skb
, struct page
*page
,
1720 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1722 __wsum csum
= csum_and_copy_from_user(from
,
1723 page_address(page
) + off
,
1727 skb
->csum
= csum_block_add(skb
->csum
, csum
, skb
->len
);
1728 } else if (copy_from_user(page_address(page
) + off
, from
, copy
))
1732 skb
->data_len
+= copy
;
1733 skb
->truesize
+= copy
;
1734 sk
->sk_wmem_queued
+= copy
;
1735 sk_mem_charge(sk
, copy
);
1740 * sk_wmem_alloc_get - returns write allocations
1743 * Returns sk_wmem_alloc minus initial offset of one
1745 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
1747 return atomic_read(&sk
->sk_wmem_alloc
) - 1;
1751 * sk_rmem_alloc_get - returns read allocations
1754 * Returns sk_rmem_alloc
1756 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
1758 return atomic_read(&sk
->sk_rmem_alloc
);
1762 * sk_has_allocations - check if allocations are outstanding
1765 * Returns true if socket has write or read allocations
1767 static inline int sk_has_allocations(const struct sock
*sk
)
1769 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
1773 * wq_has_sleeper - check if there are any waiting processes
1774 * @wq: struct socket_wq
1776 * Returns true if socket_wq has waiting processes
1778 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1779 * barrier call. They were added due to the race found within the tcp code.
1781 * Consider following tcp code paths:
1785 * sys_select receive packet
1787 * __add_wait_queue update tp->rcv_nxt
1789 * tp->rcv_nxt check sock_def_readable
1791 * schedule rcu_read_lock();
1792 * wq = rcu_dereference(sk->sk_wq);
1793 * if (wq && waitqueue_active(&wq->wait))
1794 * wake_up_interruptible(&wq->wait)
1798 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1799 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1800 * could then endup calling schedule and sleep forever if there are no more
1801 * data on the socket.
1804 static inline bool wq_has_sleeper(struct socket_wq
*wq
)
1808 * We need to be sure we are in sync with the
1809 * add_wait_queue modifications to the wait queue.
1811 * This memory barrier is paired in the sock_poll_wait.
1814 return wq
&& waitqueue_active(&wq
->wait
);
1818 * sock_poll_wait - place memory barrier behind the poll_wait call.
1820 * @wait_address: socket wait queue
1823 * See the comments in the wq_has_sleeper function.
1825 static inline void sock_poll_wait(struct file
*filp
,
1826 wait_queue_head_t
*wait_address
, poll_table
*p
)
1828 if (p
&& wait_address
) {
1829 poll_wait(filp
, wait_address
, p
);
1831 * We need to be sure we are in sync with the
1832 * socket flags modification.
1834 * This memory barrier is paired in the wq_has_sleeper.
1841 * Queue a received datagram if it will fit. Stream and sequenced
1842 * protocols can't normally use this as they need to fit buffers in
1843 * and play with them.
1845 * Inlined as it's very short and called for pretty much every
1846 * packet ever received.
1849 static inline void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
1853 skb
->destructor
= sock_wfree
;
1855 * We used to take a refcount on sk, but following operation
1856 * is enough to guarantee sk_free() wont free this sock until
1857 * all in-flight packets are completed
1859 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
1862 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
1866 skb
->destructor
= sock_rfree
;
1867 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
1868 sk_mem_charge(sk
, skb
->truesize
);
1871 extern void sk_reset_timer(struct sock
*sk
, struct timer_list
* timer
,
1872 unsigned long expires
);
1874 extern void sk_stop_timer(struct sock
*sk
, struct timer_list
* timer
);
1876 extern int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
1878 extern int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
);
1881 * Recover an error report and clear atomically
1884 static inline int sock_error(struct sock
*sk
)
1887 if (likely(!sk
->sk_err
))
1889 err
= xchg(&sk
->sk_err
, 0);
1893 static inline unsigned long sock_wspace(struct sock
*sk
)
1897 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
1898 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
1905 static inline void sk_wake_async(struct sock
*sk
, int how
, int band
)
1907 if (sock_flag(sk
, SOCK_FASYNC
))
1908 sock_wake_async(sk
->sk_socket
, how
, band
);
1911 #define SOCK_MIN_SNDBUF 2048
1913 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1914 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1916 #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
1918 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
1920 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
1921 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
1922 sk
->sk_sndbuf
= max(sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
1926 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
);
1928 static inline struct page
*sk_stream_alloc_page(struct sock
*sk
)
1930 struct page
*page
= NULL
;
1932 page
= alloc_pages(sk
->sk_allocation
, 0);
1934 sk_enter_memory_pressure(sk
);
1935 sk_stream_moderate_sndbuf(sk
);
1941 * Default write policy as shown to user space via poll/select/SIGIO
1943 static inline int sock_writeable(const struct sock
*sk
)
1945 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
1948 static inline gfp_t
gfp_any(void)
1950 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
1953 static inline long sock_rcvtimeo(const struct sock
*sk
, int noblock
)
1955 return noblock
? 0 : sk
->sk_rcvtimeo
;
1958 static inline long sock_sndtimeo(const struct sock
*sk
, int noblock
)
1960 return noblock
? 0 : sk
->sk_sndtimeo
;
1963 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
1965 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
1968 /* Alas, with timeout socket operations are not restartable.
1969 * Compare this to poll().
1971 static inline int sock_intr_errno(long timeo
)
1973 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
1976 extern void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
1977 struct sk_buff
*skb
);
1978 extern void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
1979 struct sk_buff
*skb
);
1981 static __inline__
void
1982 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
1984 ktime_t kt
= skb
->tstamp
;
1985 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
1988 * generate control messages if
1989 * - receive time stamping in software requested (SOCK_RCVTSTAMP
1990 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
1991 * - software time stamp available and wanted
1992 * (SOCK_TIMESTAMPING_SOFTWARE)
1993 * - hardware time stamps available and wanted
1994 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
1995 * SOCK_TIMESTAMPING_RAW_HARDWARE)
1997 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
1998 sock_flag(sk
, SOCK_TIMESTAMPING_RX_SOFTWARE
) ||
1999 (kt
.tv64
&& sock_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
)) ||
2000 (hwtstamps
->hwtstamp
.tv64
&&
2001 sock_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
)) ||
2002 (hwtstamps
->syststamp
.tv64
&&
2003 sock_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
)))
2004 __sock_recv_timestamp(msg
, sk
, skb
);
2008 if (sock_flag(sk
, SOCK_WIFI_STATUS
) && skb
->wifi_acked_valid
)
2009 __sock_recv_wifi_status(msg
, sk
, skb
);
2012 extern void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2013 struct sk_buff
*skb
);
2015 static inline void sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2016 struct sk_buff
*skb
)
2018 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2019 (1UL << SOCK_RCVTSTAMP) | \
2020 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2021 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
2022 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
2023 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2025 if (sk
->sk_flags
& FLAGS_TS_OR_DROPS
)
2026 __sock_recv_ts_and_drops(msg
, sk
, skb
);
2028 sk
->sk_stamp
= skb
->tstamp
;
2032 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2033 * @sk: socket sending this packet
2034 * @tx_flags: filled with instructions for time stamping
2036 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
2037 * parameters are invalid.
2039 extern int sock_tx_timestamp(struct sock
*sk
, __u8
*tx_flags
);
2042 * sk_eat_skb - Release a skb if it is no longer needed
2043 * @sk: socket to eat this skb from
2044 * @skb: socket buffer to eat
2045 * @copied_early: flag indicating whether DMA operations copied this data early
2047 * This routine must be called with interrupts disabled or with the socket
2048 * locked so that the sk_buff queue operation is ok.
2050 #ifdef CONFIG_NET_DMA
2051 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, int copied_early
)
2053 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2057 __skb_queue_tail(&sk
->sk_async_wait_queue
, skb
);
2060 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, int copied_early
)
2062 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2068 struct net
*sock_net(const struct sock
*sk
)
2070 return read_pnet(&sk
->sk_net
);
2074 void sock_net_set(struct sock
*sk
, struct net
*net
)
2076 write_pnet(&sk
->sk_net
, net
);
2080 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
2081 * They should not hold a reference to a namespace in order to allow
2083 * Sockets after sk_change_net should be released using sk_release_kernel
2085 static inline void sk_change_net(struct sock
*sk
, struct net
*net
)
2087 put_net(sock_net(sk
));
2088 sock_net_set(sk
, hold_net(net
));
2091 static inline struct sock
*skb_steal_sock(struct sk_buff
*skb
)
2093 if (unlikely(skb
->sk
)) {
2094 struct sock
*sk
= skb
->sk
;
2096 skb
->destructor
= NULL
;
2103 extern void sock_enable_timestamp(struct sock
*sk
, int flag
);
2104 extern int sock_get_timestamp(struct sock
*, struct timeval __user
*);
2105 extern int sock_get_timestampns(struct sock
*, struct timespec __user
*);
2108 * Enable debug/info messages
2110 extern int net_msg_warn
;
2111 #define NETDEBUG(fmt, args...) \
2112 do { if (net_msg_warn) printk(fmt,##args); } while (0)
2114 #define LIMIT_NETDEBUG(fmt, args...) \
2115 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
2117 extern __u32 sysctl_wmem_max
;
2118 extern __u32 sysctl_rmem_max
;
2120 extern void sk_init(void);
2122 extern int sysctl_optmem_max
;
2124 extern __u32 sysctl_wmem_default
;
2125 extern __u32 sysctl_rmem_default
;
2127 #endif /* _SOCK_H */