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_send_head: front of stuff to transmit
250 * @sk_security: used by security modules
251 * @sk_mark: generic packet mark
252 * @sk_classid: this socket's cgroup classid
253 * @sk_cgrp: this socket's cgroup-specific proto data
254 * @sk_write_pending: a write to stream socket waits to start
255 * @sk_state_change: callback to indicate change in the state of the sock
256 * @sk_data_ready: callback to indicate there is data to be processed
257 * @sk_write_space: callback to indicate there is bf sending space available
258 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
259 * @sk_backlog_rcv: callback to process the backlog
260 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
264 * Now struct inet_timewait_sock also uses sock_common, so please just
265 * don't add nothing before this first member (__sk_common) --acme
267 struct sock_common __sk_common
;
268 #define sk_node __sk_common.skc_node
269 #define sk_nulls_node __sk_common.skc_nulls_node
270 #define sk_refcnt __sk_common.skc_refcnt
271 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
273 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
274 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
275 #define sk_hash __sk_common.skc_hash
276 #define sk_family __sk_common.skc_family
277 #define sk_state __sk_common.skc_state
278 #define sk_reuse __sk_common.skc_reuse
279 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
280 #define sk_bind_node __sk_common.skc_bind_node
281 #define sk_prot __sk_common.skc_prot
282 #define sk_net __sk_common.skc_net
283 socket_lock_t sk_lock
;
284 struct sk_buff_head sk_receive_queue
;
286 * The backlog queue is special, it is always used with
287 * the per-socket spinlock held and requires low latency
288 * access. Therefore we special case it's implementation.
289 * Note : rmem_alloc is in this structure to fill a hole
290 * on 64bit arches, not because its logically part of
296 struct sk_buff
*head
;
297 struct sk_buff
*tail
;
299 #define sk_rmem_alloc sk_backlog.rmem_alloc
300 int sk_forward_alloc
;
307 struct sk_filter __rcu
*sk_filter
;
308 struct socket_wq __rcu
*sk_wq
;
310 #ifdef CONFIG_NET_DMA
311 struct sk_buff_head sk_async_wait_queue
;
315 struct xfrm_policy
*sk_policy
[2];
317 unsigned long sk_flags
;
318 struct dst_entry
*sk_dst_cache
;
319 spinlock_t sk_dst_lock
;
320 atomic_t sk_wmem_alloc
;
321 atomic_t sk_omem_alloc
;
323 struct sk_buff_head sk_write_queue
;
324 kmemcheck_bitfield_begin(flags
);
325 unsigned int sk_shutdown
: 2,
330 kmemcheck_bitfield_end(flags
);
333 netdev_features_t sk_route_caps
;
334 netdev_features_t sk_route_nocaps
;
336 unsigned int sk_gso_max_size
;
338 unsigned long sk_lingertime
;
339 struct sk_buff_head sk_error_queue
;
340 struct proto
*sk_prot_creator
;
341 rwlock_t sk_callback_lock
;
344 unsigned short sk_ack_backlog
;
345 unsigned short sk_max_ack_backlog
;
347 #ifdef CONFIG_CGROUPS
348 __u32 sk_cgrp_prioidx
;
350 struct pid
*sk_peer_pid
;
351 const struct cred
*sk_peer_cred
;
355 struct timer_list sk_timer
;
357 struct socket
*sk_socket
;
359 struct page
*sk_sndmsg_page
;
360 struct sk_buff
*sk_send_head
;
363 int sk_write_pending
;
364 #ifdef CONFIG_SECURITY
369 struct cg_proto
*sk_cgrp
;
370 void (*sk_state_change
)(struct sock
*sk
);
371 void (*sk_data_ready
)(struct sock
*sk
, int bytes
);
372 void (*sk_write_space
)(struct sock
*sk
);
373 void (*sk_error_report
)(struct sock
*sk
);
374 int (*sk_backlog_rcv
)(struct sock
*sk
,
375 struct sk_buff
*skb
);
376 void (*sk_destruct
)(struct sock
*sk
);
379 static inline int sk_peek_offset(struct sock
*sk
, int flags
)
381 if ((flags
& MSG_PEEK
) && (sk
->sk_peek_off
>= 0))
382 return sk
->sk_peek_off
;
387 static inline void sk_peek_offset_bwd(struct sock
*sk
, int val
)
389 if (sk
->sk_peek_off
>= 0) {
390 if (sk
->sk_peek_off
>= val
)
391 sk
->sk_peek_off
-= val
;
397 static inline void sk_peek_offset_fwd(struct sock
*sk
, int val
)
399 if (sk
->sk_peek_off
>= 0)
400 sk
->sk_peek_off
+= val
;
404 * Hashed lists helper routines
406 static inline struct sock
*sk_entry(const struct hlist_node
*node
)
408 return hlist_entry(node
, struct sock
, sk_node
);
411 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
413 return hlist_entry(head
->first
, struct sock
, sk_node
);
416 static inline struct sock
*sk_head(const struct hlist_head
*head
)
418 return hlist_empty(head
) ? NULL
: __sk_head(head
);
421 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
423 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
426 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
428 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
431 static inline struct sock
*sk_next(const struct sock
*sk
)
433 return sk
->sk_node
.next
?
434 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
437 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
439 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
440 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
441 struct sock
, sk_nulls_node
) :
445 static inline int sk_unhashed(const struct sock
*sk
)
447 return hlist_unhashed(&sk
->sk_node
);
450 static inline int sk_hashed(const struct sock
*sk
)
452 return !sk_unhashed(sk
);
455 static __inline__
void sk_node_init(struct hlist_node
*node
)
460 static __inline__
void sk_nulls_node_init(struct hlist_nulls_node
*node
)
465 static __inline__
void __sk_del_node(struct sock
*sk
)
467 __hlist_del(&sk
->sk_node
);
470 /* NB: equivalent to hlist_del_init_rcu */
471 static __inline__
int __sk_del_node_init(struct sock
*sk
)
475 sk_node_init(&sk
->sk_node
);
481 /* Grab socket reference count. This operation is valid only
482 when sk is ALREADY grabbed f.e. it is found in hash table
483 or a list and the lookup is made under lock preventing hash table
487 static inline void sock_hold(struct sock
*sk
)
489 atomic_inc(&sk
->sk_refcnt
);
492 /* Ungrab socket in the context, which assumes that socket refcnt
493 cannot hit zero, f.e. it is true in context of any socketcall.
495 static inline void __sock_put(struct sock
*sk
)
497 atomic_dec(&sk
->sk_refcnt
);
500 static __inline__
int sk_del_node_init(struct sock
*sk
)
502 int rc
= __sk_del_node_init(sk
);
505 /* paranoid for a while -acme */
506 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
511 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
513 static __inline__
int __sk_nulls_del_node_init_rcu(struct sock
*sk
)
516 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
522 static __inline__
int sk_nulls_del_node_init_rcu(struct sock
*sk
)
524 int rc
= __sk_nulls_del_node_init_rcu(sk
);
527 /* paranoid for a while -acme */
528 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
534 static __inline__
void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
536 hlist_add_head(&sk
->sk_node
, list
);
539 static __inline__
void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
542 __sk_add_node(sk
, list
);
545 static __inline__
void sk_add_node_rcu(struct sock
*sk
, struct hlist_head
*list
)
548 hlist_add_head_rcu(&sk
->sk_node
, list
);
551 static __inline__
void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
553 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
556 static __inline__
void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
559 __sk_nulls_add_node_rcu(sk
, list
);
562 static __inline__
void __sk_del_bind_node(struct sock
*sk
)
564 __hlist_del(&sk
->sk_bind_node
);
567 static __inline__
void sk_add_bind_node(struct sock
*sk
,
568 struct hlist_head
*list
)
570 hlist_add_head(&sk
->sk_bind_node
, list
);
573 #define sk_for_each(__sk, node, list) \
574 hlist_for_each_entry(__sk, node, list, sk_node)
575 #define sk_for_each_rcu(__sk, node, list) \
576 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
577 #define sk_nulls_for_each(__sk, node, list) \
578 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
579 #define sk_nulls_for_each_rcu(__sk, node, list) \
580 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
581 #define sk_for_each_from(__sk, node) \
582 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
583 hlist_for_each_entry_from(__sk, node, sk_node)
584 #define sk_nulls_for_each_from(__sk, node) \
585 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
586 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
587 #define sk_for_each_safe(__sk, node, tmp, list) \
588 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
589 #define sk_for_each_bound(__sk, node, list) \
590 hlist_for_each_entry(__sk, node, list, sk_bind_node)
603 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
604 SOCK_DBG
, /* %SO_DEBUG setting */
605 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
606 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
607 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
608 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
609 SOCK_TIMESTAMPING_TX_HARDWARE
, /* %SOF_TIMESTAMPING_TX_HARDWARE */
610 SOCK_TIMESTAMPING_TX_SOFTWARE
, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
611 SOCK_TIMESTAMPING_RX_HARDWARE
, /* %SOF_TIMESTAMPING_RX_HARDWARE */
612 SOCK_TIMESTAMPING_RX_SOFTWARE
, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
613 SOCK_TIMESTAMPING_SOFTWARE
, /* %SOF_TIMESTAMPING_SOFTWARE */
614 SOCK_TIMESTAMPING_RAW_HARDWARE
, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
615 SOCK_TIMESTAMPING_SYS_HARDWARE
, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
616 SOCK_FASYNC
, /* fasync() active */
618 SOCK_ZEROCOPY
, /* buffers from userspace */
619 SOCK_WIFI_STATUS
, /* push wifi status to userspace */
620 SOCK_NOFCS
, /* Tell NIC not to do the Ethernet FCS.
621 * Will use last 4 bytes of packet sent from
622 * user-space instead.
626 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
628 nsk
->sk_flags
= osk
->sk_flags
;
631 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
633 __set_bit(flag
, &sk
->sk_flags
);
636 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
638 __clear_bit(flag
, &sk
->sk_flags
);
641 static inline int sock_flag(struct sock
*sk
, enum sock_flags flag
)
643 return test_bit(flag
, &sk
->sk_flags
);
646 static inline void sk_acceptq_removed(struct sock
*sk
)
648 sk
->sk_ack_backlog
--;
651 static inline void sk_acceptq_added(struct sock
*sk
)
653 sk
->sk_ack_backlog
++;
656 static inline int sk_acceptq_is_full(struct sock
*sk
)
658 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
662 * Compute minimal free write space needed to queue new packets.
664 static inline int sk_stream_min_wspace(struct sock
*sk
)
666 return sk
->sk_wmem_queued
>> 1;
669 static inline int sk_stream_wspace(struct sock
*sk
)
671 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
674 extern void sk_stream_write_space(struct sock
*sk
);
676 static inline int sk_stream_memory_free(struct sock
*sk
)
678 return sk
->sk_wmem_queued
< sk
->sk_sndbuf
;
681 /* OOB backlog add */
682 static inline void __sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
684 /* dont let skb dst not refcounted, we are going to leave rcu lock */
687 if (!sk
->sk_backlog
.tail
)
688 sk
->sk_backlog
.head
= skb
;
690 sk
->sk_backlog
.tail
->next
= skb
;
692 sk
->sk_backlog
.tail
= skb
;
697 * Take into account size of receive queue and backlog queue
698 * Do not take into account this skb truesize,
699 * to allow even a single big packet to come.
701 static inline bool sk_rcvqueues_full(const struct sock
*sk
, const struct sk_buff
*skb
)
703 unsigned int qsize
= sk
->sk_backlog
.len
+ atomic_read(&sk
->sk_rmem_alloc
);
705 return qsize
> sk
->sk_rcvbuf
;
708 /* The per-socket spinlock must be held here. */
709 static inline __must_check
int sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
711 if (sk_rcvqueues_full(sk
, skb
))
714 __sk_add_backlog(sk
, skb
);
715 sk
->sk_backlog
.len
+= skb
->truesize
;
719 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
721 return sk
->sk_backlog_rcv(sk
, skb
);
724 static inline void sock_rps_record_flow(const struct sock
*sk
)
727 struct rps_sock_flow_table
*sock_flow_table
;
730 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
731 rps_record_sock_flow(sock_flow_table
, sk
->sk_rxhash
);
736 static inline void sock_rps_reset_flow(const struct sock
*sk
)
739 struct rps_sock_flow_table
*sock_flow_table
;
742 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
743 rps_reset_sock_flow(sock_flow_table
, sk
->sk_rxhash
);
748 static inline void sock_rps_save_rxhash(struct sock
*sk
,
749 const struct sk_buff
*skb
)
752 if (unlikely(sk
->sk_rxhash
!= skb
->rxhash
)) {
753 sock_rps_reset_flow(sk
);
754 sk
->sk_rxhash
= skb
->rxhash
;
759 static inline void sock_rps_reset_rxhash(struct sock
*sk
)
762 sock_rps_reset_flow(sk
);
767 #define sk_wait_event(__sk, __timeo, __condition) \
769 release_sock(__sk); \
770 __rc = __condition; \
772 *(__timeo) = schedule_timeout(*(__timeo)); \
775 __rc = __condition; \
779 extern int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
780 extern int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
781 extern void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
782 extern int sk_stream_error(struct sock
*sk
, int flags
, int err
);
783 extern void sk_stream_kill_queues(struct sock
*sk
);
785 extern int sk_wait_data(struct sock
*sk
, long *timeo
);
787 struct request_sock_ops
;
788 struct timewait_sock_ops
;
789 struct inet_hashinfo
;
793 /* Networking protocol blocks we attach to sockets.
794 * socket layer -> transport layer interface
795 * transport -> network interface is defined by struct inet_proto
798 void (*close
)(struct sock
*sk
,
800 int (*connect
)(struct sock
*sk
,
801 struct sockaddr
*uaddr
,
803 int (*disconnect
)(struct sock
*sk
, int flags
);
805 struct sock
* (*accept
) (struct sock
*sk
, int flags
, int *err
);
807 int (*ioctl
)(struct sock
*sk
, int cmd
,
809 int (*init
)(struct sock
*sk
);
810 void (*destroy
)(struct sock
*sk
);
811 void (*shutdown
)(struct sock
*sk
, int how
);
812 int (*setsockopt
)(struct sock
*sk
, int level
,
813 int optname
, char __user
*optval
,
814 unsigned int optlen
);
815 int (*getsockopt
)(struct sock
*sk
, int level
,
816 int optname
, char __user
*optval
,
819 int (*compat_setsockopt
)(struct sock
*sk
,
821 int optname
, char __user
*optval
,
822 unsigned int optlen
);
823 int (*compat_getsockopt
)(struct sock
*sk
,
825 int optname
, char __user
*optval
,
827 int (*compat_ioctl
)(struct sock
*sk
,
828 unsigned int cmd
, unsigned long arg
);
830 int (*sendmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
831 struct msghdr
*msg
, size_t len
);
832 int (*recvmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
834 size_t len
, int noblock
, int flags
,
836 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
837 int offset
, size_t size
, int flags
);
838 int (*bind
)(struct sock
*sk
,
839 struct sockaddr
*uaddr
, int addr_len
);
841 int (*backlog_rcv
) (struct sock
*sk
,
842 struct sk_buff
*skb
);
844 /* Keeping track of sk's, looking them up, and port selection methods. */
845 void (*hash
)(struct sock
*sk
);
846 void (*unhash
)(struct sock
*sk
);
847 void (*rehash
)(struct sock
*sk
);
848 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
849 void (*clear_sk
)(struct sock
*sk
, int size
);
851 /* Keeping track of sockets in use */
852 #ifdef CONFIG_PROC_FS
853 unsigned int inuse_idx
;
856 /* Memory pressure */
857 void (*enter_memory_pressure
)(struct sock
*sk
);
858 atomic_long_t
*memory_allocated
; /* Current allocated memory. */
859 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
861 * Pressure flag: try to collapse.
862 * Technical note: it is used by multiple contexts non atomically.
863 * All the __sk_mem_schedule() is of this nature: accounting
864 * is strict, actions are advisory and have some latency.
866 int *memory_pressure
;
873 struct kmem_cache
*slab
;
874 unsigned int obj_size
;
877 struct percpu_counter
*orphan_count
;
879 struct request_sock_ops
*rsk_prot
;
880 struct timewait_sock_ops
*twsk_prot
;
883 struct inet_hashinfo
*hashinfo
;
884 struct udp_table
*udp_table
;
885 struct raw_hashinfo
*raw_hash
;
888 struct module
*owner
;
892 struct list_head node
;
893 #ifdef SOCK_REFCNT_DEBUG
896 #ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
898 * cgroup specific init/deinit functions. Called once for all
899 * protocols that implement it, from cgroups populate function.
900 * This function has to setup any files the protocol want to
901 * appear in the kmem cgroup filesystem.
903 int (*init_cgroup
)(struct cgroup
*cgrp
,
904 struct cgroup_subsys
*ss
);
905 void (*destroy_cgroup
)(struct cgroup
*cgrp
);
906 struct cg_proto
*(*proto_cgroup
)(struct mem_cgroup
*memcg
);
911 void (*enter_memory_pressure
)(struct sock
*sk
);
912 struct res_counter
*memory_allocated
; /* Current allocated memory. */
913 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
914 int *memory_pressure
;
917 * memcg field is used to find which memcg we belong directly
918 * Each memcg struct can hold more than one cg_proto, so container_of
921 * The elegant solution would be having an inverse function to
922 * proto_cgroup in struct proto, but that means polluting the structure
923 * for everybody, instead of just for memcg users.
925 struct mem_cgroup
*memcg
;
928 extern int proto_register(struct proto
*prot
, int alloc_slab
);
929 extern void proto_unregister(struct proto
*prot
);
931 #ifdef SOCK_REFCNT_DEBUG
932 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
934 atomic_inc(&sk
->sk_prot
->socks
);
937 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
939 atomic_dec(&sk
->sk_prot
->socks
);
940 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
941 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
944 inline void sk_refcnt_debug_release(const struct sock
*sk
)
946 if (atomic_read(&sk
->sk_refcnt
) != 1)
947 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
948 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
950 #else /* SOCK_REFCNT_DEBUG */
951 #define sk_refcnt_debug_inc(sk) do { } while (0)
952 #define sk_refcnt_debug_dec(sk) do { } while (0)
953 #define sk_refcnt_debug_release(sk) do { } while (0)
954 #endif /* SOCK_REFCNT_DEBUG */
956 #if defined(CONFIG_CGROUP_MEM_RES_CTLR_KMEM) && defined(CONFIG_NET)
957 extern struct static_key memcg_socket_limit_enabled
;
958 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
959 struct cg_proto
*cg_proto
)
961 return proto
->proto_cgroup(parent_mem_cgroup(cg_proto
->memcg
));
963 #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
965 #define mem_cgroup_sockets_enabled 0
966 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
967 struct cg_proto
*cg_proto
)
974 static inline bool sk_has_memory_pressure(const struct sock
*sk
)
976 return sk
->sk_prot
->memory_pressure
!= NULL
;
979 static inline bool sk_under_memory_pressure(const struct sock
*sk
)
981 if (!sk
->sk_prot
->memory_pressure
)
984 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
985 return !!*sk
->sk_cgrp
->memory_pressure
;
987 return !!*sk
->sk_prot
->memory_pressure
;
990 static inline void sk_leave_memory_pressure(struct sock
*sk
)
992 int *memory_pressure
= sk
->sk_prot
->memory_pressure
;
994 if (!memory_pressure
)
997 if (*memory_pressure
)
998 *memory_pressure
= 0;
1000 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1001 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1002 struct proto
*prot
= sk
->sk_prot
;
1004 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1005 if (*cg_proto
->memory_pressure
)
1006 *cg_proto
->memory_pressure
= 0;
1011 static inline void sk_enter_memory_pressure(struct sock
*sk
)
1013 if (!sk
->sk_prot
->enter_memory_pressure
)
1016 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1017 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1018 struct proto
*prot
= sk
->sk_prot
;
1020 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1021 cg_proto
->enter_memory_pressure(sk
);
1024 sk
->sk_prot
->enter_memory_pressure(sk
);
1027 static inline long sk_prot_mem_limits(const struct sock
*sk
, int index
)
1029 long *prot
= sk
->sk_prot
->sysctl_mem
;
1030 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1031 prot
= sk
->sk_cgrp
->sysctl_mem
;
1035 static inline void memcg_memory_allocated_add(struct cg_proto
*prot
,
1039 struct res_counter
*fail
;
1042 ret
= res_counter_charge_nofail(prot
->memory_allocated
,
1043 amt
<< PAGE_SHIFT
, &fail
);
1045 *parent_status
= OVER_LIMIT
;
1048 static inline void memcg_memory_allocated_sub(struct cg_proto
*prot
,
1051 res_counter_uncharge(prot
->memory_allocated
, amt
<< PAGE_SHIFT
);
1054 static inline u64
memcg_memory_allocated_read(struct cg_proto
*prot
)
1057 ret
= res_counter_read_u64(prot
->memory_allocated
, RES_USAGE
);
1058 return ret
>> PAGE_SHIFT
;
1062 sk_memory_allocated(const struct sock
*sk
)
1064 struct proto
*prot
= sk
->sk_prot
;
1065 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1066 return memcg_memory_allocated_read(sk
->sk_cgrp
);
1068 return atomic_long_read(prot
->memory_allocated
);
1072 sk_memory_allocated_add(struct sock
*sk
, int amt
, int *parent_status
)
1074 struct proto
*prot
= sk
->sk_prot
;
1076 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1077 memcg_memory_allocated_add(sk
->sk_cgrp
, amt
, parent_status
);
1078 /* update the root cgroup regardless */
1079 atomic_long_add_return(amt
, prot
->memory_allocated
);
1080 return memcg_memory_allocated_read(sk
->sk_cgrp
);
1083 return atomic_long_add_return(amt
, prot
->memory_allocated
);
1087 sk_memory_allocated_sub(struct sock
*sk
, int amt
)
1089 struct proto
*prot
= sk
->sk_prot
;
1091 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1092 memcg_memory_allocated_sub(sk
->sk_cgrp
, amt
);
1094 atomic_long_sub(amt
, prot
->memory_allocated
);
1097 static inline void sk_sockets_allocated_dec(struct sock
*sk
)
1099 struct proto
*prot
= sk
->sk_prot
;
1101 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1102 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1104 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1105 percpu_counter_dec(cg_proto
->sockets_allocated
);
1108 percpu_counter_dec(prot
->sockets_allocated
);
1111 static inline void sk_sockets_allocated_inc(struct sock
*sk
)
1113 struct proto
*prot
= sk
->sk_prot
;
1115 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1116 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1118 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1119 percpu_counter_inc(cg_proto
->sockets_allocated
);
1122 percpu_counter_inc(prot
->sockets_allocated
);
1126 sk_sockets_allocated_read_positive(struct sock
*sk
)
1128 struct proto
*prot
= sk
->sk_prot
;
1130 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1131 return percpu_counter_sum_positive(sk
->sk_cgrp
->sockets_allocated
);
1133 return percpu_counter_sum_positive(prot
->sockets_allocated
);
1137 proto_sockets_allocated_sum_positive(struct proto
*prot
)
1139 return percpu_counter_sum_positive(prot
->sockets_allocated
);
1143 proto_memory_allocated(struct proto
*prot
)
1145 return atomic_long_read(prot
->memory_allocated
);
1149 proto_memory_pressure(struct proto
*prot
)
1151 if (!prot
->memory_pressure
)
1153 return !!*prot
->memory_pressure
;
1157 #ifdef CONFIG_PROC_FS
1158 /* Called with local bh disabled */
1159 extern void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int inc
);
1160 extern int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
1162 static void inline sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
1169 /* With per-bucket locks this operation is not-atomic, so that
1170 * this version is not worse.
1172 static inline void __sk_prot_rehash(struct sock
*sk
)
1174 sk
->sk_prot
->unhash(sk
);
1175 sk
->sk_prot
->hash(sk
);
1178 void sk_prot_clear_portaddr_nulls(struct sock
*sk
, int size
);
1180 /* About 10 seconds */
1181 #define SOCK_DESTROY_TIME (10*HZ)
1183 /* Sockets 0-1023 can't be bound to unless you are superuser */
1184 #define PROT_SOCK 1024
1186 #define SHUTDOWN_MASK 3
1187 #define RCV_SHUTDOWN 1
1188 #define SEND_SHUTDOWN 2
1190 #define SOCK_SNDBUF_LOCK 1
1191 #define SOCK_RCVBUF_LOCK 2
1192 #define SOCK_BINDADDR_LOCK 4
1193 #define SOCK_BINDPORT_LOCK 8
1195 /* sock_iocb: used to kick off async processing of socket ios */
1197 struct list_head list
;
1201 struct socket
*sock
;
1203 struct scm_cookie
*scm
;
1204 struct msghdr
*msg
, async_msg
;
1205 struct kiocb
*kiocb
;
1208 static inline struct sock_iocb
*kiocb_to_siocb(struct kiocb
*iocb
)
1210 return (struct sock_iocb
*)iocb
->private;
1213 static inline struct kiocb
*siocb_to_kiocb(struct sock_iocb
*si
)
1218 struct socket_alloc
{
1219 struct socket socket
;
1220 struct inode vfs_inode
;
1223 static inline struct socket
*SOCKET_I(struct inode
*inode
)
1225 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
1228 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
1230 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
1234 * Functions for memory accounting
1236 extern int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
1237 extern void __sk_mem_reclaim(struct sock
*sk
);
1239 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1240 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1241 #define SK_MEM_SEND 0
1242 #define SK_MEM_RECV 1
1244 static inline int sk_mem_pages(int amt
)
1246 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
1249 static inline int sk_has_account(struct sock
*sk
)
1251 /* return true if protocol supports memory accounting */
1252 return !!sk
->sk_prot
->memory_allocated
;
1255 static inline int sk_wmem_schedule(struct sock
*sk
, int size
)
1257 if (!sk_has_account(sk
))
1259 return size
<= sk
->sk_forward_alloc
||
1260 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
1263 static inline int sk_rmem_schedule(struct sock
*sk
, int size
)
1265 if (!sk_has_account(sk
))
1267 return size
<= sk
->sk_forward_alloc
||
1268 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
);
1271 static inline void sk_mem_reclaim(struct sock
*sk
)
1273 if (!sk_has_account(sk
))
1275 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
1276 __sk_mem_reclaim(sk
);
1279 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
1281 if (!sk_has_account(sk
))
1283 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
1284 __sk_mem_reclaim(sk
);
1287 static inline void sk_mem_charge(struct sock
*sk
, int size
)
1289 if (!sk_has_account(sk
))
1291 sk
->sk_forward_alloc
-= size
;
1294 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
1296 if (!sk_has_account(sk
))
1298 sk
->sk_forward_alloc
+= size
;
1301 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
1303 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
1304 sk
->sk_wmem_queued
-= skb
->truesize
;
1305 sk_mem_uncharge(sk
, skb
->truesize
);
1309 /* Used by processes to "lock" a socket state, so that
1310 * interrupts and bottom half handlers won't change it
1311 * from under us. It essentially blocks any incoming
1312 * packets, so that we won't get any new data or any
1313 * packets that change the state of the socket.
1315 * While locked, BH processing will add new packets to
1316 * the backlog queue. This queue is processed by the
1317 * owner of the socket lock right before it is released.
1319 * Since ~2.3.5 it is also exclusive sleep lock serializing
1320 * accesses from user process context.
1322 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1325 * Macro so as to not evaluate some arguments when
1326 * lockdep is not enabled.
1328 * Mark both the sk_lock and the sk_lock.slock as a
1329 * per-address-family lock class.
1331 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1333 sk->sk_lock.owned = 0; \
1334 init_waitqueue_head(&sk->sk_lock.wq); \
1335 spin_lock_init(&(sk)->sk_lock.slock); \
1336 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1337 sizeof((sk)->sk_lock)); \
1338 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1340 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1343 extern void lock_sock_nested(struct sock
*sk
, int subclass
);
1345 static inline void lock_sock(struct sock
*sk
)
1347 lock_sock_nested(sk
, 0);
1350 extern void release_sock(struct sock
*sk
);
1352 /* BH context may only use the following locking interface. */
1353 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1354 #define bh_lock_sock_nested(__sk) \
1355 spin_lock_nested(&((__sk)->sk_lock.slock), \
1356 SINGLE_DEPTH_NESTING)
1357 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1359 extern bool lock_sock_fast(struct sock
*sk
);
1361 * unlock_sock_fast - complement of lock_sock_fast
1365 * fast unlock socket for user context.
1366 * If slow mode is on, we call regular release_sock()
1368 static inline void unlock_sock_fast(struct sock
*sk
, bool slow
)
1373 spin_unlock_bh(&sk
->sk_lock
.slock
);
1377 extern struct sock
*sk_alloc(struct net
*net
, int family
,
1379 struct proto
*prot
);
1380 extern void sk_free(struct sock
*sk
);
1381 extern void sk_release_kernel(struct sock
*sk
);
1382 extern struct sock
*sk_clone_lock(const struct sock
*sk
,
1383 const gfp_t priority
);
1385 extern struct sk_buff
*sock_wmalloc(struct sock
*sk
,
1386 unsigned long size
, int force
,
1388 extern struct sk_buff
*sock_rmalloc(struct sock
*sk
,
1389 unsigned long size
, int force
,
1391 extern void sock_wfree(struct sk_buff
*skb
);
1392 extern void sock_rfree(struct sk_buff
*skb
);
1394 extern int sock_setsockopt(struct socket
*sock
, int level
,
1395 int op
, char __user
*optval
,
1396 unsigned int optlen
);
1398 extern int sock_getsockopt(struct socket
*sock
, int level
,
1399 int op
, char __user
*optval
,
1400 int __user
*optlen
);
1401 extern struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
,
1405 extern struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
,
1406 unsigned long header_len
,
1407 unsigned long data_len
,
1410 extern void *sock_kmalloc(struct sock
*sk
, int size
,
1412 extern void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
1413 extern void sk_send_sigurg(struct sock
*sk
);
1415 #ifdef CONFIG_CGROUPS
1416 extern void sock_update_classid(struct sock
*sk
);
1418 static inline void sock_update_classid(struct sock
*sk
)
1424 * Functions to fill in entries in struct proto_ops when a protocol
1425 * does not implement a particular function.
1427 extern int sock_no_bind(struct socket
*,
1428 struct sockaddr
*, int);
1429 extern int sock_no_connect(struct socket
*,
1430 struct sockaddr
*, int, int);
1431 extern int sock_no_socketpair(struct socket
*,
1433 extern int sock_no_accept(struct socket
*,
1434 struct socket
*, int);
1435 extern int sock_no_getname(struct socket
*,
1436 struct sockaddr
*, int *, int);
1437 extern unsigned int sock_no_poll(struct file
*, struct socket
*,
1438 struct poll_table_struct
*);
1439 extern int sock_no_ioctl(struct socket
*, unsigned int,
1441 extern int sock_no_listen(struct socket
*, int);
1442 extern int sock_no_shutdown(struct socket
*, int);
1443 extern int sock_no_getsockopt(struct socket
*, int , int,
1444 char __user
*, int __user
*);
1445 extern int sock_no_setsockopt(struct socket
*, int, int,
1446 char __user
*, unsigned int);
1447 extern int sock_no_sendmsg(struct kiocb
*, struct socket
*,
1448 struct msghdr
*, size_t);
1449 extern int sock_no_recvmsg(struct kiocb
*, struct socket
*,
1450 struct msghdr
*, size_t, int);
1451 extern int sock_no_mmap(struct file
*file
,
1452 struct socket
*sock
,
1453 struct vm_area_struct
*vma
);
1454 extern ssize_t
sock_no_sendpage(struct socket
*sock
,
1456 int offset
, size_t size
,
1460 * Functions to fill in entries in struct proto_ops when a protocol
1461 * uses the inet style.
1463 extern int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1464 char __user
*optval
, int __user
*optlen
);
1465 extern int sock_common_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
1466 struct msghdr
*msg
, size_t size
, int flags
);
1467 extern int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1468 char __user
*optval
, unsigned int optlen
);
1469 extern int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
1470 int optname
, char __user
*optval
, int __user
*optlen
);
1471 extern int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
1472 int optname
, char __user
*optval
, unsigned int optlen
);
1474 extern void sk_common_release(struct sock
*sk
);
1477 * Default socket callbacks and setup code
1480 /* Initialise core socket variables */
1481 extern void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1483 extern void sk_filter_release_rcu(struct rcu_head
*rcu
);
1486 * sk_filter_release - release a socket filter
1487 * @fp: filter to remove
1489 * Remove a filter from a socket and release its resources.
1492 static inline void sk_filter_release(struct sk_filter
*fp
)
1494 if (atomic_dec_and_test(&fp
->refcnt
))
1495 call_rcu(&fp
->rcu
, sk_filter_release_rcu
);
1498 static inline void sk_filter_uncharge(struct sock
*sk
, struct sk_filter
*fp
)
1500 unsigned int size
= sk_filter_len(fp
);
1502 atomic_sub(size
, &sk
->sk_omem_alloc
);
1503 sk_filter_release(fp
);
1506 static inline void sk_filter_charge(struct sock
*sk
, struct sk_filter
*fp
)
1508 atomic_inc(&fp
->refcnt
);
1509 atomic_add(sk_filter_len(fp
), &sk
->sk_omem_alloc
);
1513 * Socket reference counting postulates.
1515 * * Each user of socket SHOULD hold a reference count.
1516 * * Each access point to socket (an hash table bucket, reference from a list,
1517 * running timer, skb in flight MUST hold a reference count.
1518 * * When reference count hits 0, it means it will never increase back.
1519 * * When reference count hits 0, it means that no references from
1520 * outside exist to this socket and current process on current CPU
1521 * is last user and may/should destroy this socket.
1522 * * sk_free is called from any context: process, BH, IRQ. When
1523 * it is called, socket has no references from outside -> sk_free
1524 * may release descendant resources allocated by the socket, but
1525 * to the time when it is called, socket is NOT referenced by any
1526 * hash tables, lists etc.
1527 * * Packets, delivered from outside (from network or from another process)
1528 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1529 * when they sit in queue. Otherwise, packets will leak to hole, when
1530 * socket is looked up by one cpu and unhasing is made by another CPU.
1531 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1532 * (leak to backlog). Packet socket does all the processing inside
1533 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1534 * use separate SMP lock, so that they are prone too.
1537 /* Ungrab socket and destroy it, if it was the last reference. */
1538 static inline void sock_put(struct sock
*sk
)
1540 if (atomic_dec_and_test(&sk
->sk_refcnt
))
1544 extern int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
1547 static inline void sk_tx_queue_set(struct sock
*sk
, int tx_queue
)
1549 sk
->sk_tx_queue_mapping
= tx_queue
;
1552 static inline void sk_tx_queue_clear(struct sock
*sk
)
1554 sk
->sk_tx_queue_mapping
= -1;
1557 static inline int sk_tx_queue_get(const struct sock
*sk
)
1559 return sk
? sk
->sk_tx_queue_mapping
: -1;
1562 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1564 sk_tx_queue_clear(sk
);
1565 sk
->sk_socket
= sock
;
1568 static inline wait_queue_head_t
*sk_sleep(struct sock
*sk
)
1570 BUILD_BUG_ON(offsetof(struct socket_wq
, wait
) != 0);
1571 return &rcu_dereference_raw(sk
->sk_wq
)->wait
;
1573 /* Detach socket from process context.
1574 * Announce socket dead, detach it from wait queue and inode.
1575 * Note that parent inode held reference count on this struct sock,
1576 * we do not release it in this function, because protocol
1577 * probably wants some additional cleanups or even continuing
1578 * to work with this socket (TCP).
1580 static inline void sock_orphan(struct sock
*sk
)
1582 write_lock_bh(&sk
->sk_callback_lock
);
1583 sock_set_flag(sk
, SOCK_DEAD
);
1584 sk_set_socket(sk
, NULL
);
1586 write_unlock_bh(&sk
->sk_callback_lock
);
1589 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1591 write_lock_bh(&sk
->sk_callback_lock
);
1592 sk
->sk_wq
= parent
->wq
;
1594 sk_set_socket(sk
, parent
);
1595 security_sock_graft(sk
, parent
);
1596 write_unlock_bh(&sk
->sk_callback_lock
);
1599 extern int sock_i_uid(struct sock
*sk
);
1600 extern unsigned long sock_i_ino(struct sock
*sk
);
1602 static inline struct dst_entry
*
1603 __sk_dst_get(struct sock
*sk
)
1605 return rcu_dereference_check(sk
->sk_dst_cache
, sock_owned_by_user(sk
) ||
1606 lockdep_is_held(&sk
->sk_lock
.slock
));
1609 static inline struct dst_entry
*
1610 sk_dst_get(struct sock
*sk
)
1612 struct dst_entry
*dst
;
1615 dst
= rcu_dereference(sk
->sk_dst_cache
);
1622 extern void sk_reset_txq(struct sock
*sk
);
1624 static inline void dst_negative_advice(struct sock
*sk
)
1626 struct dst_entry
*ndst
, *dst
= __sk_dst_get(sk
);
1628 if (dst
&& dst
->ops
->negative_advice
) {
1629 ndst
= dst
->ops
->negative_advice(dst
);
1632 rcu_assign_pointer(sk
->sk_dst_cache
, ndst
);
1639 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1641 struct dst_entry
*old_dst
;
1643 sk_tx_queue_clear(sk
);
1645 * This can be called while sk is owned by the caller only,
1646 * with no state that can be checked in a rcu_dereference_check() cond
1648 old_dst
= rcu_dereference_raw(sk
->sk_dst_cache
);
1649 rcu_assign_pointer(sk
->sk_dst_cache
, dst
);
1650 dst_release(old_dst
);
1654 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1656 spin_lock(&sk
->sk_dst_lock
);
1657 __sk_dst_set(sk
, dst
);
1658 spin_unlock(&sk
->sk_dst_lock
);
1662 __sk_dst_reset(struct sock
*sk
)
1664 __sk_dst_set(sk
, NULL
);
1668 sk_dst_reset(struct sock
*sk
)
1670 spin_lock(&sk
->sk_dst_lock
);
1672 spin_unlock(&sk
->sk_dst_lock
);
1675 extern struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1677 extern struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1679 static inline int sk_can_gso(const struct sock
*sk
)
1681 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1684 extern void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1686 static inline void sk_nocaps_add(struct sock
*sk
, netdev_features_t flags
)
1688 sk
->sk_route_nocaps
|= flags
;
1689 sk
->sk_route_caps
&= ~flags
;
1692 static inline int skb_do_copy_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1693 char __user
*from
, char *to
,
1694 int copy
, int offset
)
1696 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1698 __wsum csum
= csum_and_copy_from_user(from
, to
, copy
, 0, &err
);
1701 skb
->csum
= csum_block_add(skb
->csum
, csum
, offset
);
1702 } else if (sk
->sk_route_caps
& NETIF_F_NOCACHE_COPY
) {
1703 if (!access_ok(VERIFY_READ
, from
, copy
) ||
1704 __copy_from_user_nocache(to
, from
, copy
))
1706 } else if (copy_from_user(to
, from
, copy
))
1712 static inline int skb_add_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1713 char __user
*from
, int copy
)
1715 int err
, offset
= skb
->len
;
1717 err
= skb_do_copy_data_nocache(sk
, skb
, from
, skb_put(skb
, copy
),
1720 __skb_trim(skb
, offset
);
1725 static inline int skb_copy_to_page_nocache(struct sock
*sk
, char __user
*from
,
1726 struct sk_buff
*skb
,
1732 err
= skb_do_copy_data_nocache(sk
, skb
, from
, page_address(page
) + off
,
1738 skb
->data_len
+= copy
;
1739 skb
->truesize
+= copy
;
1740 sk
->sk_wmem_queued
+= copy
;
1741 sk_mem_charge(sk
, copy
);
1745 static inline int skb_copy_to_page(struct sock
*sk
, char __user
*from
,
1746 struct sk_buff
*skb
, struct page
*page
,
1749 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1751 __wsum csum
= csum_and_copy_from_user(from
,
1752 page_address(page
) + off
,
1756 skb
->csum
= csum_block_add(skb
->csum
, csum
, skb
->len
);
1757 } else if (copy_from_user(page_address(page
) + off
, from
, copy
))
1761 skb
->data_len
+= copy
;
1762 skb
->truesize
+= copy
;
1763 sk
->sk_wmem_queued
+= copy
;
1764 sk_mem_charge(sk
, copy
);
1769 * sk_wmem_alloc_get - returns write allocations
1772 * Returns sk_wmem_alloc minus initial offset of one
1774 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
1776 return atomic_read(&sk
->sk_wmem_alloc
) - 1;
1780 * sk_rmem_alloc_get - returns read allocations
1783 * Returns sk_rmem_alloc
1785 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
1787 return atomic_read(&sk
->sk_rmem_alloc
);
1791 * sk_has_allocations - check if allocations are outstanding
1794 * Returns true if socket has write or read allocations
1796 static inline int sk_has_allocations(const struct sock
*sk
)
1798 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
1802 * wq_has_sleeper - check if there are any waiting processes
1803 * @wq: struct socket_wq
1805 * Returns true if socket_wq has waiting processes
1807 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1808 * barrier call. They were added due to the race found within the tcp code.
1810 * Consider following tcp code paths:
1814 * sys_select receive packet
1816 * __add_wait_queue update tp->rcv_nxt
1818 * tp->rcv_nxt check sock_def_readable
1820 * schedule rcu_read_lock();
1821 * wq = rcu_dereference(sk->sk_wq);
1822 * if (wq && waitqueue_active(&wq->wait))
1823 * wake_up_interruptible(&wq->wait)
1827 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1828 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1829 * could then endup calling schedule and sleep forever if there are no more
1830 * data on the socket.
1833 static inline bool wq_has_sleeper(struct socket_wq
*wq
)
1837 * We need to be sure we are in sync with the
1838 * add_wait_queue modifications to the wait queue.
1840 * This memory barrier is paired in the sock_poll_wait.
1843 return wq
&& waitqueue_active(&wq
->wait
);
1847 * sock_poll_wait - place memory barrier behind the poll_wait call.
1849 * @wait_address: socket wait queue
1852 * See the comments in the wq_has_sleeper function.
1854 static inline void sock_poll_wait(struct file
*filp
,
1855 wait_queue_head_t
*wait_address
, poll_table
*p
)
1857 if (!poll_does_not_wait(p
) && wait_address
) {
1858 poll_wait(filp
, wait_address
, p
);
1860 * We need to be sure we are in sync with the
1861 * socket flags modification.
1863 * This memory barrier is paired in the wq_has_sleeper.
1870 * Queue a received datagram if it will fit. Stream and sequenced
1871 * protocols can't normally use this as they need to fit buffers in
1872 * and play with them.
1874 * Inlined as it's very short and called for pretty much every
1875 * packet ever received.
1878 static inline void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
1882 skb
->destructor
= sock_wfree
;
1884 * We used to take a refcount on sk, but following operation
1885 * is enough to guarantee sk_free() wont free this sock until
1886 * all in-flight packets are completed
1888 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
1891 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
1895 skb
->destructor
= sock_rfree
;
1896 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
1897 sk_mem_charge(sk
, skb
->truesize
);
1900 extern void sk_reset_timer(struct sock
*sk
, struct timer_list
* timer
,
1901 unsigned long expires
);
1903 extern void sk_stop_timer(struct sock
*sk
, struct timer_list
* timer
);
1905 extern int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
1907 extern int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
);
1910 * Recover an error report and clear atomically
1913 static inline int sock_error(struct sock
*sk
)
1916 if (likely(!sk
->sk_err
))
1918 err
= xchg(&sk
->sk_err
, 0);
1922 static inline unsigned long sock_wspace(struct sock
*sk
)
1926 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
1927 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
1934 static inline void sk_wake_async(struct sock
*sk
, int how
, int band
)
1936 if (sock_flag(sk
, SOCK_FASYNC
))
1937 sock_wake_async(sk
->sk_socket
, how
, band
);
1940 #define SOCK_MIN_SNDBUF 2048
1942 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1943 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1945 #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
1947 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
1949 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
1950 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
1951 sk
->sk_sndbuf
= max(sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
1955 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
);
1957 static inline struct page
*sk_stream_alloc_page(struct sock
*sk
)
1959 struct page
*page
= NULL
;
1961 page
= alloc_pages(sk
->sk_allocation
, 0);
1963 sk_enter_memory_pressure(sk
);
1964 sk_stream_moderate_sndbuf(sk
);
1970 * Default write policy as shown to user space via poll/select/SIGIO
1972 static inline int sock_writeable(const struct sock
*sk
)
1974 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
1977 static inline gfp_t
gfp_any(void)
1979 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
1982 static inline long sock_rcvtimeo(const struct sock
*sk
, int noblock
)
1984 return noblock
? 0 : sk
->sk_rcvtimeo
;
1987 static inline long sock_sndtimeo(const struct sock
*sk
, int noblock
)
1989 return noblock
? 0 : sk
->sk_sndtimeo
;
1992 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
1994 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
1997 /* Alas, with timeout socket operations are not restartable.
1998 * Compare this to poll().
2000 static inline int sock_intr_errno(long timeo
)
2002 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
2005 extern void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
2006 struct sk_buff
*skb
);
2007 extern void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
2008 struct sk_buff
*skb
);
2010 static __inline__
void
2011 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
2013 ktime_t kt
= skb
->tstamp
;
2014 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
2017 * generate control messages if
2018 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2019 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2020 * - software time stamp available and wanted
2021 * (SOCK_TIMESTAMPING_SOFTWARE)
2022 * - hardware time stamps available and wanted
2023 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2024 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2026 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
2027 sock_flag(sk
, SOCK_TIMESTAMPING_RX_SOFTWARE
) ||
2028 (kt
.tv64
&& sock_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
)) ||
2029 (hwtstamps
->hwtstamp
.tv64
&&
2030 sock_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
)) ||
2031 (hwtstamps
->syststamp
.tv64
&&
2032 sock_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
)))
2033 __sock_recv_timestamp(msg
, sk
, skb
);
2037 if (sock_flag(sk
, SOCK_WIFI_STATUS
) && skb
->wifi_acked_valid
)
2038 __sock_recv_wifi_status(msg
, sk
, skb
);
2041 extern void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2042 struct sk_buff
*skb
);
2044 static inline void sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2045 struct sk_buff
*skb
)
2047 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2048 (1UL << SOCK_RCVTSTAMP) | \
2049 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2050 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
2051 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
2052 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2054 if (sk
->sk_flags
& FLAGS_TS_OR_DROPS
)
2055 __sock_recv_ts_and_drops(msg
, sk
, skb
);
2057 sk
->sk_stamp
= skb
->tstamp
;
2061 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2062 * @sk: socket sending this packet
2063 * @tx_flags: filled with instructions for time stamping
2065 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
2066 * parameters are invalid.
2068 extern int sock_tx_timestamp(struct sock
*sk
, __u8
*tx_flags
);
2071 * sk_eat_skb - Release a skb if it is no longer needed
2072 * @sk: socket to eat this skb from
2073 * @skb: socket buffer to eat
2074 * @copied_early: flag indicating whether DMA operations copied this data early
2076 * This routine must be called with interrupts disabled or with the socket
2077 * locked so that the sk_buff queue operation is ok.
2079 #ifdef CONFIG_NET_DMA
2080 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, int copied_early
)
2082 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2086 __skb_queue_tail(&sk
->sk_async_wait_queue
, skb
);
2089 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, int copied_early
)
2091 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2097 struct net
*sock_net(const struct sock
*sk
)
2099 return read_pnet(&sk
->sk_net
);
2103 void sock_net_set(struct sock
*sk
, struct net
*net
)
2105 write_pnet(&sk
->sk_net
, net
);
2109 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
2110 * They should not hold a reference to a namespace in order to allow
2112 * Sockets after sk_change_net should be released using sk_release_kernel
2114 static inline void sk_change_net(struct sock
*sk
, struct net
*net
)
2116 put_net(sock_net(sk
));
2117 sock_net_set(sk
, hold_net(net
));
2120 static inline struct sock
*skb_steal_sock(struct sk_buff
*skb
)
2122 if (unlikely(skb
->sk
)) {
2123 struct sock
*sk
= skb
->sk
;
2125 skb
->destructor
= NULL
;
2132 extern void sock_enable_timestamp(struct sock
*sk
, int flag
);
2133 extern int sock_get_timestamp(struct sock
*, struct timeval __user
*);
2134 extern int sock_get_timestampns(struct sock
*, struct timespec __user
*);
2137 * Enable debug/info messages
2139 extern int net_msg_warn
;
2140 #define NETDEBUG(fmt, args...) \
2141 do { if (net_msg_warn) printk(fmt,##args); } while (0)
2143 #define LIMIT_NETDEBUG(fmt, args...) \
2144 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
2146 extern __u32 sysctl_wmem_max
;
2147 extern __u32 sysctl_rmem_max
;
2149 extern void sk_init(void);
2151 extern int sysctl_optmem_max
;
2153 extern __u32 sysctl_wmem_default
;
2154 extern __u32 sysctl_rmem_default
;
2156 #endif /* _SOCK_H */