2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the AF_INET socket handler.
8 * Version: @(#)sock.h 1.0.4 05/13/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Florian La Roche <flla@stud.uni-sb.de>
16 * Alan Cox : Volatiles in skbuff pointers. See
17 * skbuff comments. May be overdone,
18 * better to prove they can be removed
20 * Alan Cox : Added a zapped field for tcp to note
21 * a socket is reset and must stay shut up
22 * Alan Cox : New fields for options
23 * Pauline Middelink : identd support
24 * Alan Cox : Eliminate low level recv/recvfrom
25 * David S. Miller : New socket lookup architecture.
26 * Steve Whitehouse: Default routines for sock_ops
27 * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made
28 * protinfo be just a void pointer, as the
29 * protocol specific parts were moved to
30 * respective headers and ipv4/v6, etc now
31 * use private slabcaches for its socks
32 * Pedro Hortas : New flags field for socket options
35 * This program is free software; you can redistribute it and/or
36 * modify it under the terms of the GNU General Public License
37 * as published by the Free Software Foundation; either version
38 * 2 of the License, or (at your option) any later version.
43 #include <linux/hardirq.h>
44 #include <linux/kernel.h>
45 #include <linux/list.h>
46 #include <linux/list_nulls.h>
47 #include <linux/timer.h>
48 #include <linux/cache.h>
49 #include <linux/bitops.h>
50 #include <linux/lockdep.h>
51 #include <linux/netdevice.h>
52 #include <linux/skbuff.h> /* struct sk_buff */
54 #include <linux/security.h>
55 #include <linux/slab.h>
56 #include <linux/uaccess.h>
57 #include <linux/memcontrol.h>
58 #include <linux/res_counter.h>
59 #include <linux/static_key.h>
60 #include <linux/aio.h>
61 #include <linux/sched.h>
63 #include <linux/filter.h>
64 #include <linux/rculist_nulls.h>
65 #include <linux/poll.h>
67 #include <linux/atomic.h>
69 #include <net/checksum.h>
74 int mem_cgroup_sockets_init(struct mem_cgroup
*memcg
, struct cgroup_subsys
*ss
);
75 void mem_cgroup_sockets_destroy(struct mem_cgroup
*memcg
);
78 int mem_cgroup_sockets_init(struct mem_cgroup
*memcg
, struct cgroup_subsys
*ss
)
83 void mem_cgroup_sockets_destroy(struct mem_cgroup
*memcg
)
88 * This structure really needs to be cleaned up.
89 * Most of it is for TCP, and not used by any of
90 * the other protocols.
93 /* Define this to get the SOCK_DBG debugging facility. */
94 #define SOCK_DEBUGGING
96 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
97 printk(KERN_DEBUG msg); } while (0)
99 /* Validate arguments and do nothing */
100 static inline __printf(2, 3)
101 void SOCK_DEBUG(const struct sock
*sk
, const char *msg
, ...)
106 /* This is the per-socket lock. The spinlock provides a synchronization
107 * between user contexts and software interrupt processing, whereas the
108 * mini-semaphore synchronizes multiple users amongst themselves.
113 wait_queue_head_t wq
;
115 * We express the mutex-alike socket_lock semantics
116 * to the lock validator by explicitly managing
117 * the slock as a lock variant (in addition to
120 #ifdef CONFIG_DEBUG_LOCK_ALLOC
121 struct lockdep_map dep_map
;
129 typedef __u32 __bitwise __portpair
;
130 typedef __u64 __bitwise __addrpair
;
133 * struct sock_common - minimal network layer representation of sockets
134 * @skc_daddr: Foreign IPv4 addr
135 * @skc_rcv_saddr: Bound local IPv4 addr
136 * @skc_hash: hash value used with various protocol lookup tables
137 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
138 * @skc_dport: placeholder for inet_dport/tw_dport
139 * @skc_num: placeholder for inet_num/tw_num
140 * @skc_family: network address family
141 * @skc_state: Connection state
142 * @skc_reuse: %SO_REUSEADDR setting
143 * @skc_bound_dev_if: bound device index if != 0
144 * @skc_bind_node: bind hash linkage for various protocol lookup tables
145 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
146 * @skc_prot: protocol handlers inside a network family
147 * @skc_net: reference to the network namespace of this socket
148 * @skc_node: main hash linkage for various protocol lookup tables
149 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
150 * @skc_tx_queue_mapping: tx queue number for this connection
151 * @skc_refcnt: reference count
153 * This is the minimal network layer representation of sockets, the header
154 * for struct sock and struct inet_timewait_sock.
157 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
158 * address on 64bit arches : cf INET_MATCH() and INET_TW_MATCH()
161 __addrpair skc_addrpair
;
164 __be32 skc_rcv_saddr
;
168 unsigned int skc_hash
;
169 __u16 skc_u16hashes
[2];
171 /* skc_dport && skc_num must be grouped as well */
173 __portpair skc_portpair
;
180 unsigned short skc_family
;
181 volatile unsigned char skc_state
;
182 unsigned char skc_reuse
;
183 int skc_bound_dev_if
;
185 struct hlist_node skc_bind_node
;
186 struct hlist_nulls_node skc_portaddr_node
;
188 struct proto
*skc_prot
;
193 * fields between dontcopy_begin/dontcopy_end
194 * are not copied in sock_copy()
197 int skc_dontcopy_begin
[0];
200 struct hlist_node skc_node
;
201 struct hlist_nulls_node skc_nulls_node
;
203 int skc_tx_queue_mapping
;
206 int skc_dontcopy_end
[0];
212 * struct sock - network layer representation of sockets
213 * @__sk_common: shared layout with inet_timewait_sock
214 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
215 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
216 * @sk_lock: synchronizer
217 * @sk_rcvbuf: size of receive buffer in bytes
218 * @sk_wq: sock wait queue and async head
219 * @sk_rx_dst: receive input route used by early tcp demux
220 * @sk_dst_cache: destination cache
221 * @sk_dst_lock: destination cache lock
222 * @sk_policy: flow policy
223 * @sk_receive_queue: incoming packets
224 * @sk_wmem_alloc: transmit queue bytes committed
225 * @sk_write_queue: Packet sending queue
226 * @sk_async_wait_queue: DMA copied packets
227 * @sk_omem_alloc: "o" is "option" or "other"
228 * @sk_wmem_queued: persistent queue size
229 * @sk_forward_alloc: space allocated forward
230 * @sk_allocation: allocation mode
231 * @sk_sndbuf: size of send buffer in bytes
232 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
233 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
234 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
235 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
236 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
237 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
238 * @sk_gso_max_size: Maximum GSO segment size to build
239 * @sk_gso_max_segs: Maximum number of GSO segments
240 * @sk_lingertime: %SO_LINGER l_linger setting
241 * @sk_backlog: always used with the per-socket spinlock held
242 * @sk_callback_lock: used with the callbacks in the end of this struct
243 * @sk_error_queue: rarely used
244 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
245 * IPV6_ADDRFORM for instance)
246 * @sk_err: last error
247 * @sk_err_soft: errors that don't cause failure but are the cause of a
248 * persistent failure not just 'timed out'
249 * @sk_drops: raw/udp drops counter
250 * @sk_ack_backlog: current listen backlog
251 * @sk_max_ack_backlog: listen backlog set in listen()
252 * @sk_priority: %SO_PRIORITY setting
253 * @sk_cgrp_prioidx: socket group's priority map index
254 * @sk_type: socket type (%SOCK_STREAM, etc)
255 * @sk_protocol: which protocol this socket belongs in this network family
256 * @sk_peer_pid: &struct pid for this socket's peer
257 * @sk_peer_cred: %SO_PEERCRED setting
258 * @sk_rcvlowat: %SO_RCVLOWAT setting
259 * @sk_rcvtimeo: %SO_RCVTIMEO setting
260 * @sk_sndtimeo: %SO_SNDTIMEO setting
261 * @sk_rxhash: flow hash received from netif layer
262 * @sk_filter: socket filtering instructions
263 * @sk_protinfo: private area, net family specific, when not using slab
264 * @sk_timer: sock cleanup timer
265 * @sk_stamp: time stamp of last packet received
266 * @sk_socket: Identd and reporting IO signals
267 * @sk_user_data: RPC layer private data
268 * @sk_frag: cached page frag
269 * @sk_peek_off: current peek_offset value
270 * @sk_send_head: front of stuff to transmit
271 * @sk_security: used by security modules
272 * @sk_mark: generic packet mark
273 * @sk_classid: this socket's cgroup classid
274 * @sk_cgrp: this socket's cgroup-specific proto data
275 * @sk_write_pending: a write to stream socket waits to start
276 * @sk_state_change: callback to indicate change in the state of the sock
277 * @sk_data_ready: callback to indicate there is data to be processed
278 * @sk_write_space: callback to indicate there is bf sending space available
279 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
280 * @sk_backlog_rcv: callback to process the backlog
281 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
285 * Now struct inet_timewait_sock also uses sock_common, so please just
286 * don't add nothing before this first member (__sk_common) --acme
288 struct sock_common __sk_common
;
289 #define sk_node __sk_common.skc_node
290 #define sk_nulls_node __sk_common.skc_nulls_node
291 #define sk_refcnt __sk_common.skc_refcnt
292 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
294 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
295 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
296 #define sk_hash __sk_common.skc_hash
297 #define sk_family __sk_common.skc_family
298 #define sk_state __sk_common.skc_state
299 #define sk_reuse __sk_common.skc_reuse
300 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
301 #define sk_bind_node __sk_common.skc_bind_node
302 #define sk_prot __sk_common.skc_prot
303 #define sk_net __sk_common.skc_net
304 socket_lock_t sk_lock
;
305 struct sk_buff_head sk_receive_queue
;
307 * The backlog queue is special, it is always used with
308 * the per-socket spinlock held and requires low latency
309 * access. Therefore we special case it's implementation.
310 * Note : rmem_alloc is in this structure to fill a hole
311 * on 64bit arches, not because its logically part of
317 struct sk_buff
*head
;
318 struct sk_buff
*tail
;
320 #define sk_rmem_alloc sk_backlog.rmem_alloc
321 int sk_forward_alloc
;
328 struct sk_filter __rcu
*sk_filter
;
329 struct socket_wq __rcu
*sk_wq
;
331 #ifdef CONFIG_NET_DMA
332 struct sk_buff_head sk_async_wait_queue
;
336 struct xfrm_policy
*sk_policy
[2];
338 unsigned long sk_flags
;
339 struct dst_entry
*sk_rx_dst
;
340 struct dst_entry
*sk_dst_cache
;
341 spinlock_t sk_dst_lock
;
342 atomic_t sk_wmem_alloc
;
343 atomic_t sk_omem_alloc
;
345 struct sk_buff_head sk_write_queue
;
346 kmemcheck_bitfield_begin(flags
);
347 unsigned int sk_shutdown
: 2,
352 kmemcheck_bitfield_end(flags
);
355 netdev_features_t sk_route_caps
;
356 netdev_features_t sk_route_nocaps
;
358 unsigned int sk_gso_max_size
;
361 unsigned long sk_lingertime
;
362 struct sk_buff_head sk_error_queue
;
363 struct proto
*sk_prot_creator
;
364 rwlock_t sk_callback_lock
;
367 unsigned short sk_ack_backlog
;
368 unsigned short sk_max_ack_backlog
;
370 #ifdef CONFIG_CGROUPS
371 __u32 sk_cgrp_prioidx
;
373 struct pid
*sk_peer_pid
;
374 const struct cred
*sk_peer_cred
;
378 struct timer_list sk_timer
;
380 struct socket
*sk_socket
;
382 struct page_frag sk_frag
;
383 struct sk_buff
*sk_send_head
;
385 int sk_write_pending
;
386 #ifdef CONFIG_SECURITY
391 struct cg_proto
*sk_cgrp
;
392 void (*sk_state_change
)(struct sock
*sk
);
393 void (*sk_data_ready
)(struct sock
*sk
, int bytes
);
394 void (*sk_write_space
)(struct sock
*sk
);
395 void (*sk_error_report
)(struct sock
*sk
);
396 int (*sk_backlog_rcv
)(struct sock
*sk
,
397 struct sk_buff
*skb
);
398 void (*sk_destruct
)(struct sock
*sk
);
402 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
403 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
404 * on a socket means that the socket will reuse everybody else's port
405 * without looking at the other's sk_reuse value.
408 #define SK_NO_REUSE 0
409 #define SK_CAN_REUSE 1
410 #define SK_FORCE_REUSE 2
412 static inline int sk_peek_offset(struct sock
*sk
, int flags
)
414 if ((flags
& MSG_PEEK
) && (sk
->sk_peek_off
>= 0))
415 return sk
->sk_peek_off
;
420 static inline void sk_peek_offset_bwd(struct sock
*sk
, int val
)
422 if (sk
->sk_peek_off
>= 0) {
423 if (sk
->sk_peek_off
>= val
)
424 sk
->sk_peek_off
-= val
;
430 static inline void sk_peek_offset_fwd(struct sock
*sk
, int val
)
432 if (sk
->sk_peek_off
>= 0)
433 sk
->sk_peek_off
+= val
;
437 * Hashed lists helper routines
439 static inline struct sock
*sk_entry(const struct hlist_node
*node
)
441 return hlist_entry(node
, struct sock
, sk_node
);
444 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
446 return hlist_entry(head
->first
, struct sock
, sk_node
);
449 static inline struct sock
*sk_head(const struct hlist_head
*head
)
451 return hlist_empty(head
) ? NULL
: __sk_head(head
);
454 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
456 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
459 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
461 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
464 static inline struct sock
*sk_next(const struct sock
*sk
)
466 return sk
->sk_node
.next
?
467 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
470 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
472 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
473 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
474 struct sock
, sk_nulls_node
) :
478 static inline bool sk_unhashed(const struct sock
*sk
)
480 return hlist_unhashed(&sk
->sk_node
);
483 static inline bool sk_hashed(const struct sock
*sk
)
485 return !sk_unhashed(sk
);
488 static inline void sk_node_init(struct hlist_node
*node
)
493 static inline void sk_nulls_node_init(struct hlist_nulls_node
*node
)
498 static inline void __sk_del_node(struct sock
*sk
)
500 __hlist_del(&sk
->sk_node
);
503 /* NB: equivalent to hlist_del_init_rcu */
504 static inline bool __sk_del_node_init(struct sock
*sk
)
508 sk_node_init(&sk
->sk_node
);
514 /* Grab socket reference count. This operation is valid only
515 when sk is ALREADY grabbed f.e. it is found in hash table
516 or a list and the lookup is made under lock preventing hash table
520 static inline void sock_hold(struct sock
*sk
)
522 atomic_inc(&sk
->sk_refcnt
);
525 /* Ungrab socket in the context, which assumes that socket refcnt
526 cannot hit zero, f.e. it is true in context of any socketcall.
528 static inline void __sock_put(struct sock
*sk
)
530 atomic_dec(&sk
->sk_refcnt
);
533 static inline bool sk_del_node_init(struct sock
*sk
)
535 bool rc
= __sk_del_node_init(sk
);
538 /* paranoid for a while -acme */
539 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
544 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
546 static inline bool __sk_nulls_del_node_init_rcu(struct sock
*sk
)
549 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
555 static inline bool sk_nulls_del_node_init_rcu(struct sock
*sk
)
557 bool rc
= __sk_nulls_del_node_init_rcu(sk
);
560 /* paranoid for a while -acme */
561 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
567 static inline void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
569 hlist_add_head(&sk
->sk_node
, list
);
572 static inline void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
575 __sk_add_node(sk
, list
);
578 static inline void sk_add_node_rcu(struct sock
*sk
, struct hlist_head
*list
)
581 hlist_add_head_rcu(&sk
->sk_node
, list
);
584 static inline void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
586 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
589 static inline void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
592 __sk_nulls_add_node_rcu(sk
, list
);
595 static inline void __sk_del_bind_node(struct sock
*sk
)
597 __hlist_del(&sk
->sk_bind_node
);
600 static inline void sk_add_bind_node(struct sock
*sk
,
601 struct hlist_head
*list
)
603 hlist_add_head(&sk
->sk_bind_node
, list
);
606 #define sk_for_each(__sk, node, list) \
607 hlist_for_each_entry(__sk, node, list, sk_node)
608 #define sk_for_each_rcu(__sk, node, list) \
609 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
610 #define sk_nulls_for_each(__sk, node, list) \
611 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
612 #define sk_nulls_for_each_rcu(__sk, node, list) \
613 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
614 #define sk_for_each_from(__sk, node) \
615 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
616 hlist_for_each_entry_from(__sk, node, sk_node)
617 #define sk_nulls_for_each_from(__sk, node) \
618 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
619 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
620 #define sk_for_each_safe(__sk, node, tmp, list) \
621 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
622 #define sk_for_each_bound(__sk, node, list) \
623 hlist_for_each_entry(__sk, node, list, sk_bind_node)
625 static inline struct user_namespace
*sk_user_ns(struct sock
*sk
)
627 /* Careful only use this in a context where these parameters
628 * can not change and must all be valid, such as recvmsg from
631 return sk
->sk_socket
->file
->f_cred
->user_ns
;
645 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
646 SOCK_DBG
, /* %SO_DEBUG setting */
647 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
648 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
649 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
650 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
651 SOCK_MEMALLOC
, /* VM depends on this socket for swapping */
652 SOCK_TIMESTAMPING_TX_HARDWARE
, /* %SOF_TIMESTAMPING_TX_HARDWARE */
653 SOCK_TIMESTAMPING_TX_SOFTWARE
, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
654 SOCK_TIMESTAMPING_RX_HARDWARE
, /* %SOF_TIMESTAMPING_RX_HARDWARE */
655 SOCK_TIMESTAMPING_RX_SOFTWARE
, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
656 SOCK_TIMESTAMPING_SOFTWARE
, /* %SOF_TIMESTAMPING_SOFTWARE */
657 SOCK_TIMESTAMPING_RAW_HARDWARE
, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
658 SOCK_TIMESTAMPING_SYS_HARDWARE
, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
659 SOCK_FASYNC
, /* fasync() active */
661 SOCK_ZEROCOPY
, /* buffers from userspace */
662 SOCK_WIFI_STATUS
, /* push wifi status to userspace */
663 SOCK_NOFCS
, /* Tell NIC not to do the Ethernet FCS.
664 * Will use last 4 bytes of packet sent from
665 * user-space instead.
669 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
671 nsk
->sk_flags
= osk
->sk_flags
;
674 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
676 __set_bit(flag
, &sk
->sk_flags
);
679 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
681 __clear_bit(flag
, &sk
->sk_flags
);
684 static inline bool sock_flag(const struct sock
*sk
, enum sock_flags flag
)
686 return test_bit(flag
, &sk
->sk_flags
);
690 extern struct static_key memalloc_socks
;
691 static inline int sk_memalloc_socks(void)
693 return static_key_false(&memalloc_socks
);
697 static inline int sk_memalloc_socks(void)
704 static inline gfp_t
sk_gfp_atomic(struct sock
*sk
, gfp_t gfp_mask
)
706 return GFP_ATOMIC
| (sk
->sk_allocation
& __GFP_MEMALLOC
);
709 static inline void sk_acceptq_removed(struct sock
*sk
)
711 sk
->sk_ack_backlog
--;
714 static inline void sk_acceptq_added(struct sock
*sk
)
716 sk
->sk_ack_backlog
++;
719 static inline bool sk_acceptq_is_full(const struct sock
*sk
)
721 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
725 * Compute minimal free write space needed to queue new packets.
727 static inline int sk_stream_min_wspace(const struct sock
*sk
)
729 return sk
->sk_wmem_queued
>> 1;
732 static inline int sk_stream_wspace(const struct sock
*sk
)
734 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
737 extern void sk_stream_write_space(struct sock
*sk
);
739 static inline bool sk_stream_memory_free(const struct sock
*sk
)
741 return sk
->sk_wmem_queued
< sk
->sk_sndbuf
;
744 /* OOB backlog add */
745 static inline void __sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
747 /* dont let skb dst not refcounted, we are going to leave rcu lock */
750 if (!sk
->sk_backlog
.tail
)
751 sk
->sk_backlog
.head
= skb
;
753 sk
->sk_backlog
.tail
->next
= skb
;
755 sk
->sk_backlog
.tail
= skb
;
760 * Take into account size of receive queue and backlog queue
761 * Do not take into account this skb truesize,
762 * to allow even a single big packet to come.
764 static inline bool sk_rcvqueues_full(const struct sock
*sk
, const struct sk_buff
*skb
,
767 unsigned int qsize
= sk
->sk_backlog
.len
+ atomic_read(&sk
->sk_rmem_alloc
);
769 return qsize
> limit
;
772 /* The per-socket spinlock must be held here. */
773 static inline __must_check
int sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
,
776 if (sk_rcvqueues_full(sk
, skb
, limit
))
779 __sk_add_backlog(sk
, skb
);
780 sk
->sk_backlog
.len
+= skb
->truesize
;
784 extern int __sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
786 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
788 if (sk_memalloc_socks() && skb_pfmemalloc(skb
))
789 return __sk_backlog_rcv(sk
, skb
);
791 return sk
->sk_backlog_rcv(sk
, skb
);
794 static inline void sock_rps_record_flow(const struct sock
*sk
)
797 struct rps_sock_flow_table
*sock_flow_table
;
800 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
801 rps_record_sock_flow(sock_flow_table
, sk
->sk_rxhash
);
806 static inline void sock_rps_reset_flow(const struct sock
*sk
)
809 struct rps_sock_flow_table
*sock_flow_table
;
812 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
813 rps_reset_sock_flow(sock_flow_table
, sk
->sk_rxhash
);
818 static inline void sock_rps_save_rxhash(struct sock
*sk
,
819 const struct sk_buff
*skb
)
822 if (unlikely(sk
->sk_rxhash
!= skb
->rxhash
)) {
823 sock_rps_reset_flow(sk
);
824 sk
->sk_rxhash
= skb
->rxhash
;
829 static inline void sock_rps_reset_rxhash(struct sock
*sk
)
832 sock_rps_reset_flow(sk
);
837 #define sk_wait_event(__sk, __timeo, __condition) \
839 release_sock(__sk); \
840 __rc = __condition; \
842 *(__timeo) = schedule_timeout(*(__timeo)); \
845 __rc = __condition; \
849 extern int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
850 extern int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
851 extern void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
852 extern int sk_stream_error(struct sock
*sk
, int flags
, int err
);
853 extern void sk_stream_kill_queues(struct sock
*sk
);
854 extern void sk_set_memalloc(struct sock
*sk
);
855 extern void sk_clear_memalloc(struct sock
*sk
);
857 extern int sk_wait_data(struct sock
*sk
, long *timeo
);
859 struct request_sock_ops
;
860 struct timewait_sock_ops
;
861 struct inet_hashinfo
;
865 /* Networking protocol blocks we attach to sockets.
866 * socket layer -> transport layer interface
867 * transport -> network interface is defined by struct inet_proto
870 void (*close
)(struct sock
*sk
,
872 int (*connect
)(struct sock
*sk
,
873 struct sockaddr
*uaddr
,
875 int (*disconnect
)(struct sock
*sk
, int flags
);
877 struct sock
* (*accept
)(struct sock
*sk
, int flags
, int *err
);
879 int (*ioctl
)(struct sock
*sk
, int cmd
,
881 int (*init
)(struct sock
*sk
);
882 void (*destroy
)(struct sock
*sk
);
883 void (*shutdown
)(struct sock
*sk
, int how
);
884 int (*setsockopt
)(struct sock
*sk
, int level
,
885 int optname
, char __user
*optval
,
886 unsigned int optlen
);
887 int (*getsockopt
)(struct sock
*sk
, int level
,
888 int optname
, char __user
*optval
,
891 int (*compat_setsockopt
)(struct sock
*sk
,
893 int optname
, char __user
*optval
,
894 unsigned int optlen
);
895 int (*compat_getsockopt
)(struct sock
*sk
,
897 int optname
, char __user
*optval
,
899 int (*compat_ioctl
)(struct sock
*sk
,
900 unsigned int cmd
, unsigned long arg
);
902 int (*sendmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
903 struct msghdr
*msg
, size_t len
);
904 int (*recvmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
906 size_t len
, int noblock
, int flags
,
908 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
909 int offset
, size_t size
, int flags
);
910 int (*bind
)(struct sock
*sk
,
911 struct sockaddr
*uaddr
, int addr_len
);
913 int (*backlog_rcv
) (struct sock
*sk
,
914 struct sk_buff
*skb
);
916 void (*release_cb
)(struct sock
*sk
);
917 void (*mtu_reduced
)(struct sock
*sk
);
919 /* Keeping track of sk's, looking them up, and port selection methods. */
920 void (*hash
)(struct sock
*sk
);
921 void (*unhash
)(struct sock
*sk
);
922 void (*rehash
)(struct sock
*sk
);
923 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
924 void (*clear_sk
)(struct sock
*sk
, int size
);
926 /* Keeping track of sockets in use */
927 #ifdef CONFIG_PROC_FS
928 unsigned int inuse_idx
;
931 /* Memory pressure */
932 void (*enter_memory_pressure
)(struct sock
*sk
);
933 atomic_long_t
*memory_allocated
; /* Current allocated memory. */
934 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
936 * Pressure flag: try to collapse.
937 * Technical note: it is used by multiple contexts non atomically.
938 * All the __sk_mem_schedule() is of this nature: accounting
939 * is strict, actions are advisory and have some latency.
941 int *memory_pressure
;
948 struct kmem_cache
*slab
;
949 unsigned int obj_size
;
952 struct percpu_counter
*orphan_count
;
954 struct request_sock_ops
*rsk_prot
;
955 struct timewait_sock_ops
*twsk_prot
;
958 struct inet_hashinfo
*hashinfo
;
959 struct udp_table
*udp_table
;
960 struct raw_hashinfo
*raw_hash
;
963 struct module
*owner
;
967 struct list_head node
;
968 #ifdef SOCK_REFCNT_DEBUG
971 #ifdef CONFIG_MEMCG_KMEM
973 * cgroup specific init/deinit functions. Called once for all
974 * protocols that implement it, from cgroups populate function.
975 * This function has to setup any files the protocol want to
976 * appear in the kmem cgroup filesystem.
978 int (*init_cgroup
)(struct mem_cgroup
*memcg
,
979 struct cgroup_subsys
*ss
);
980 void (*destroy_cgroup
)(struct mem_cgroup
*memcg
);
981 struct cg_proto
*(*proto_cgroup
)(struct mem_cgroup
*memcg
);
986 * Bits in struct cg_proto.flags
988 enum cg_proto_flags
{
989 /* Currently active and new sockets should be assigned to cgroups */
991 /* It was ever activated; we must disarm static keys on destruction */
992 MEMCG_SOCK_ACTIVATED
,
996 void (*enter_memory_pressure
)(struct sock
*sk
);
997 struct res_counter
*memory_allocated
; /* Current allocated memory. */
998 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
999 int *memory_pressure
;
1001 unsigned long flags
;
1003 * memcg field is used to find which memcg we belong directly
1004 * Each memcg struct can hold more than one cg_proto, so container_of
1007 * The elegant solution would be having an inverse function to
1008 * proto_cgroup in struct proto, but that means polluting the structure
1009 * for everybody, instead of just for memcg users.
1011 struct mem_cgroup
*memcg
;
1014 extern int proto_register(struct proto
*prot
, int alloc_slab
);
1015 extern void proto_unregister(struct proto
*prot
);
1017 static inline bool memcg_proto_active(struct cg_proto
*cg_proto
)
1019 return test_bit(MEMCG_SOCK_ACTIVE
, &cg_proto
->flags
);
1022 static inline bool memcg_proto_activated(struct cg_proto
*cg_proto
)
1024 return test_bit(MEMCG_SOCK_ACTIVATED
, &cg_proto
->flags
);
1027 #ifdef SOCK_REFCNT_DEBUG
1028 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
1030 atomic_inc(&sk
->sk_prot
->socks
);
1033 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
1035 atomic_dec(&sk
->sk_prot
->socks
);
1036 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
1037 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
1040 inline void sk_refcnt_debug_release(const struct sock
*sk
)
1042 if (atomic_read(&sk
->sk_refcnt
) != 1)
1043 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
1044 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
1046 #else /* SOCK_REFCNT_DEBUG */
1047 #define sk_refcnt_debug_inc(sk) do { } while (0)
1048 #define sk_refcnt_debug_dec(sk) do { } while (0)
1049 #define sk_refcnt_debug_release(sk) do { } while (0)
1050 #endif /* SOCK_REFCNT_DEBUG */
1052 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
1053 extern struct static_key memcg_socket_limit_enabled
;
1054 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
1055 struct cg_proto
*cg_proto
)
1057 return proto
->proto_cgroup(parent_mem_cgroup(cg_proto
->memcg
));
1059 #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
1061 #define mem_cgroup_sockets_enabled 0
1062 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
1063 struct cg_proto
*cg_proto
)
1070 static inline bool sk_has_memory_pressure(const struct sock
*sk
)
1072 return sk
->sk_prot
->memory_pressure
!= NULL
;
1075 static inline bool sk_under_memory_pressure(const struct sock
*sk
)
1077 if (!sk
->sk_prot
->memory_pressure
)
1080 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1081 return !!*sk
->sk_cgrp
->memory_pressure
;
1083 return !!*sk
->sk_prot
->memory_pressure
;
1086 static inline void sk_leave_memory_pressure(struct sock
*sk
)
1088 int *memory_pressure
= sk
->sk_prot
->memory_pressure
;
1090 if (!memory_pressure
)
1093 if (*memory_pressure
)
1094 *memory_pressure
= 0;
1096 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1097 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1098 struct proto
*prot
= sk
->sk_prot
;
1100 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1101 if (*cg_proto
->memory_pressure
)
1102 *cg_proto
->memory_pressure
= 0;
1107 static inline void sk_enter_memory_pressure(struct sock
*sk
)
1109 if (!sk
->sk_prot
->enter_memory_pressure
)
1112 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1113 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1114 struct proto
*prot
= sk
->sk_prot
;
1116 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1117 cg_proto
->enter_memory_pressure(sk
);
1120 sk
->sk_prot
->enter_memory_pressure(sk
);
1123 static inline long sk_prot_mem_limits(const struct sock
*sk
, int index
)
1125 long *prot
= sk
->sk_prot
->sysctl_mem
;
1126 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1127 prot
= sk
->sk_cgrp
->sysctl_mem
;
1131 static inline void memcg_memory_allocated_add(struct cg_proto
*prot
,
1135 struct res_counter
*fail
;
1138 ret
= res_counter_charge_nofail(prot
->memory_allocated
,
1139 amt
<< PAGE_SHIFT
, &fail
);
1141 *parent_status
= OVER_LIMIT
;
1144 static inline void memcg_memory_allocated_sub(struct cg_proto
*prot
,
1147 res_counter_uncharge(prot
->memory_allocated
, amt
<< PAGE_SHIFT
);
1150 static inline u64
memcg_memory_allocated_read(struct cg_proto
*prot
)
1153 ret
= res_counter_read_u64(prot
->memory_allocated
, RES_USAGE
);
1154 return ret
>> PAGE_SHIFT
;
1158 sk_memory_allocated(const struct sock
*sk
)
1160 struct proto
*prot
= sk
->sk_prot
;
1161 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1162 return memcg_memory_allocated_read(sk
->sk_cgrp
);
1164 return atomic_long_read(prot
->memory_allocated
);
1168 sk_memory_allocated_add(struct sock
*sk
, int amt
, int *parent_status
)
1170 struct proto
*prot
= sk
->sk_prot
;
1172 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1173 memcg_memory_allocated_add(sk
->sk_cgrp
, amt
, parent_status
);
1174 /* update the root cgroup regardless */
1175 atomic_long_add_return(amt
, prot
->memory_allocated
);
1176 return memcg_memory_allocated_read(sk
->sk_cgrp
);
1179 return atomic_long_add_return(amt
, prot
->memory_allocated
);
1183 sk_memory_allocated_sub(struct sock
*sk
, int amt
)
1185 struct proto
*prot
= sk
->sk_prot
;
1187 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1188 memcg_memory_allocated_sub(sk
->sk_cgrp
, amt
);
1190 atomic_long_sub(amt
, prot
->memory_allocated
);
1193 static inline void sk_sockets_allocated_dec(struct sock
*sk
)
1195 struct proto
*prot
= sk
->sk_prot
;
1197 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1198 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1200 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1201 percpu_counter_dec(cg_proto
->sockets_allocated
);
1204 percpu_counter_dec(prot
->sockets_allocated
);
1207 static inline void sk_sockets_allocated_inc(struct sock
*sk
)
1209 struct proto
*prot
= sk
->sk_prot
;
1211 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1212 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1214 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1215 percpu_counter_inc(cg_proto
->sockets_allocated
);
1218 percpu_counter_inc(prot
->sockets_allocated
);
1222 sk_sockets_allocated_read_positive(struct sock
*sk
)
1224 struct proto
*prot
= sk
->sk_prot
;
1226 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1227 return percpu_counter_read_positive(sk
->sk_cgrp
->sockets_allocated
);
1229 return percpu_counter_read_positive(prot
->sockets_allocated
);
1233 proto_sockets_allocated_sum_positive(struct proto
*prot
)
1235 return percpu_counter_sum_positive(prot
->sockets_allocated
);
1239 proto_memory_allocated(struct proto
*prot
)
1241 return atomic_long_read(prot
->memory_allocated
);
1245 proto_memory_pressure(struct proto
*prot
)
1247 if (!prot
->memory_pressure
)
1249 return !!*prot
->memory_pressure
;
1253 #ifdef CONFIG_PROC_FS
1254 /* Called with local bh disabled */
1255 extern void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int inc
);
1256 extern int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
1258 static inline void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
1265 /* With per-bucket locks this operation is not-atomic, so that
1266 * this version is not worse.
1268 static inline void __sk_prot_rehash(struct sock
*sk
)
1270 sk
->sk_prot
->unhash(sk
);
1271 sk
->sk_prot
->hash(sk
);
1274 void sk_prot_clear_portaddr_nulls(struct sock
*sk
, int size
);
1276 /* About 10 seconds */
1277 #define SOCK_DESTROY_TIME (10*HZ)
1279 /* Sockets 0-1023 can't be bound to unless you are superuser */
1280 #define PROT_SOCK 1024
1282 #define SHUTDOWN_MASK 3
1283 #define RCV_SHUTDOWN 1
1284 #define SEND_SHUTDOWN 2
1286 #define SOCK_SNDBUF_LOCK 1
1287 #define SOCK_RCVBUF_LOCK 2
1288 #define SOCK_BINDADDR_LOCK 4
1289 #define SOCK_BINDPORT_LOCK 8
1291 /* sock_iocb: used to kick off async processing of socket ios */
1293 struct list_head list
;
1297 struct socket
*sock
;
1299 struct scm_cookie
*scm
;
1300 struct msghdr
*msg
, async_msg
;
1301 struct kiocb
*kiocb
;
1304 static inline struct sock_iocb
*kiocb_to_siocb(struct kiocb
*iocb
)
1306 return (struct sock_iocb
*)iocb
->private;
1309 static inline struct kiocb
*siocb_to_kiocb(struct sock_iocb
*si
)
1314 struct socket_alloc
{
1315 struct socket socket
;
1316 struct inode vfs_inode
;
1319 static inline struct socket
*SOCKET_I(struct inode
*inode
)
1321 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
1324 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
1326 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
1330 * Functions for memory accounting
1332 extern int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
1333 extern void __sk_mem_reclaim(struct sock
*sk
);
1335 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1336 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1337 #define SK_MEM_SEND 0
1338 #define SK_MEM_RECV 1
1340 static inline int sk_mem_pages(int amt
)
1342 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
1345 static inline bool sk_has_account(struct sock
*sk
)
1347 /* return true if protocol supports memory accounting */
1348 return !!sk
->sk_prot
->memory_allocated
;
1351 static inline bool sk_wmem_schedule(struct sock
*sk
, int size
)
1353 if (!sk_has_account(sk
))
1355 return size
<= sk
->sk_forward_alloc
||
1356 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
1360 sk_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
, int size
)
1362 if (!sk_has_account(sk
))
1364 return size
<= sk
->sk_forward_alloc
||
1365 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
) ||
1366 skb_pfmemalloc(skb
);
1369 static inline void sk_mem_reclaim(struct sock
*sk
)
1371 if (!sk_has_account(sk
))
1373 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
1374 __sk_mem_reclaim(sk
);
1377 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
1379 if (!sk_has_account(sk
))
1381 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
1382 __sk_mem_reclaim(sk
);
1385 static inline void sk_mem_charge(struct sock
*sk
, int size
)
1387 if (!sk_has_account(sk
))
1389 sk
->sk_forward_alloc
-= size
;
1392 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
1394 if (!sk_has_account(sk
))
1396 sk
->sk_forward_alloc
+= size
;
1399 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
1401 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
1402 sk
->sk_wmem_queued
-= skb
->truesize
;
1403 sk_mem_uncharge(sk
, skb
->truesize
);
1407 /* Used by processes to "lock" a socket state, so that
1408 * interrupts and bottom half handlers won't change it
1409 * from under us. It essentially blocks any incoming
1410 * packets, so that we won't get any new data or any
1411 * packets that change the state of the socket.
1413 * While locked, BH processing will add new packets to
1414 * the backlog queue. This queue is processed by the
1415 * owner of the socket lock right before it is released.
1417 * Since ~2.3.5 it is also exclusive sleep lock serializing
1418 * accesses from user process context.
1420 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1423 * Macro so as to not evaluate some arguments when
1424 * lockdep is not enabled.
1426 * Mark both the sk_lock and the sk_lock.slock as a
1427 * per-address-family lock class.
1429 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1431 sk->sk_lock.owned = 0; \
1432 init_waitqueue_head(&sk->sk_lock.wq); \
1433 spin_lock_init(&(sk)->sk_lock.slock); \
1434 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1435 sizeof((sk)->sk_lock)); \
1436 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1438 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1441 extern void lock_sock_nested(struct sock
*sk
, int subclass
);
1443 static inline void lock_sock(struct sock
*sk
)
1445 lock_sock_nested(sk
, 0);
1448 extern void release_sock(struct sock
*sk
);
1450 /* BH context may only use the following locking interface. */
1451 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1452 #define bh_lock_sock_nested(__sk) \
1453 spin_lock_nested(&((__sk)->sk_lock.slock), \
1454 SINGLE_DEPTH_NESTING)
1455 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1457 extern bool lock_sock_fast(struct sock
*sk
);
1459 * unlock_sock_fast - complement of lock_sock_fast
1463 * fast unlock socket for user context.
1464 * If slow mode is on, we call regular release_sock()
1466 static inline void unlock_sock_fast(struct sock
*sk
, bool slow
)
1471 spin_unlock_bh(&sk
->sk_lock
.slock
);
1475 extern struct sock
*sk_alloc(struct net
*net
, int family
,
1477 struct proto
*prot
);
1478 extern void sk_free(struct sock
*sk
);
1479 extern void sk_release_kernel(struct sock
*sk
);
1480 extern struct sock
*sk_clone_lock(const struct sock
*sk
,
1481 const gfp_t priority
);
1483 extern struct sk_buff
*sock_wmalloc(struct sock
*sk
,
1484 unsigned long size
, int force
,
1486 extern struct sk_buff
*sock_rmalloc(struct sock
*sk
,
1487 unsigned long size
, int force
,
1489 extern void sock_wfree(struct sk_buff
*skb
);
1490 extern void sock_rfree(struct sk_buff
*skb
);
1491 extern void sock_edemux(struct sk_buff
*skb
);
1493 extern int sock_setsockopt(struct socket
*sock
, int level
,
1494 int op
, char __user
*optval
,
1495 unsigned int optlen
);
1497 extern int sock_getsockopt(struct socket
*sock
, int level
,
1498 int op
, char __user
*optval
,
1499 int __user
*optlen
);
1500 extern struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
,
1504 extern struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
,
1505 unsigned long header_len
,
1506 unsigned long data_len
,
1509 extern void *sock_kmalloc(struct sock
*sk
, int size
,
1511 extern void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
1512 extern void sk_send_sigurg(struct sock
*sk
);
1515 * Functions to fill in entries in struct proto_ops when a protocol
1516 * does not implement a particular function.
1518 extern int sock_no_bind(struct socket
*,
1519 struct sockaddr
*, int);
1520 extern int sock_no_connect(struct socket
*,
1521 struct sockaddr
*, int, int);
1522 extern int sock_no_socketpair(struct socket
*,
1524 extern int sock_no_accept(struct socket
*,
1525 struct socket
*, int);
1526 extern int sock_no_getname(struct socket
*,
1527 struct sockaddr
*, int *, int);
1528 extern unsigned int sock_no_poll(struct file
*, struct socket
*,
1529 struct poll_table_struct
*);
1530 extern int sock_no_ioctl(struct socket
*, unsigned int,
1532 extern int sock_no_listen(struct socket
*, int);
1533 extern int sock_no_shutdown(struct socket
*, int);
1534 extern int sock_no_getsockopt(struct socket
*, int , int,
1535 char __user
*, int __user
*);
1536 extern int sock_no_setsockopt(struct socket
*, int, int,
1537 char __user
*, unsigned int);
1538 extern int sock_no_sendmsg(struct kiocb
*, struct socket
*,
1539 struct msghdr
*, size_t);
1540 extern int sock_no_recvmsg(struct kiocb
*, struct socket
*,
1541 struct msghdr
*, size_t, int);
1542 extern int sock_no_mmap(struct file
*file
,
1543 struct socket
*sock
,
1544 struct vm_area_struct
*vma
);
1545 extern ssize_t
sock_no_sendpage(struct socket
*sock
,
1547 int offset
, size_t size
,
1551 * Functions to fill in entries in struct proto_ops when a protocol
1552 * uses the inet style.
1554 extern int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1555 char __user
*optval
, int __user
*optlen
);
1556 extern int sock_common_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
1557 struct msghdr
*msg
, size_t size
, int flags
);
1558 extern int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1559 char __user
*optval
, unsigned int optlen
);
1560 extern int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
1561 int optname
, char __user
*optval
, int __user
*optlen
);
1562 extern int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
1563 int optname
, char __user
*optval
, unsigned int optlen
);
1565 extern void sk_common_release(struct sock
*sk
);
1568 * Default socket callbacks and setup code
1571 /* Initialise core socket variables */
1572 extern void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1574 extern void sk_filter_release_rcu(struct rcu_head
*rcu
);
1577 * sk_filter_release - release a socket filter
1578 * @fp: filter to remove
1580 * Remove a filter from a socket and release its resources.
1583 static inline void sk_filter_release(struct sk_filter
*fp
)
1585 if (atomic_dec_and_test(&fp
->refcnt
))
1586 call_rcu(&fp
->rcu
, sk_filter_release_rcu
);
1589 static inline void sk_filter_uncharge(struct sock
*sk
, struct sk_filter
*fp
)
1591 unsigned int size
= sk_filter_len(fp
);
1593 atomic_sub(size
, &sk
->sk_omem_alloc
);
1594 sk_filter_release(fp
);
1597 static inline void sk_filter_charge(struct sock
*sk
, struct sk_filter
*fp
)
1599 atomic_inc(&fp
->refcnt
);
1600 atomic_add(sk_filter_len(fp
), &sk
->sk_omem_alloc
);
1604 * Socket reference counting postulates.
1606 * * Each user of socket SHOULD hold a reference count.
1607 * * Each access point to socket (an hash table bucket, reference from a list,
1608 * running timer, skb in flight MUST hold a reference count.
1609 * * When reference count hits 0, it means it will never increase back.
1610 * * When reference count hits 0, it means that no references from
1611 * outside exist to this socket and current process on current CPU
1612 * is last user and may/should destroy this socket.
1613 * * sk_free is called from any context: process, BH, IRQ. When
1614 * it is called, socket has no references from outside -> sk_free
1615 * may release descendant resources allocated by the socket, but
1616 * to the time when it is called, socket is NOT referenced by any
1617 * hash tables, lists etc.
1618 * * Packets, delivered from outside (from network or from another process)
1619 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1620 * when they sit in queue. Otherwise, packets will leak to hole, when
1621 * socket is looked up by one cpu and unhasing is made by another CPU.
1622 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1623 * (leak to backlog). Packet socket does all the processing inside
1624 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1625 * use separate SMP lock, so that they are prone too.
1628 /* Ungrab socket and destroy it, if it was the last reference. */
1629 static inline void sock_put(struct sock
*sk
)
1631 if (atomic_dec_and_test(&sk
->sk_refcnt
))
1635 extern int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
1638 static inline void sk_tx_queue_set(struct sock
*sk
, int tx_queue
)
1640 sk
->sk_tx_queue_mapping
= tx_queue
;
1643 static inline void sk_tx_queue_clear(struct sock
*sk
)
1645 sk
->sk_tx_queue_mapping
= -1;
1648 static inline int sk_tx_queue_get(const struct sock
*sk
)
1650 return sk
? sk
->sk_tx_queue_mapping
: -1;
1653 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1655 sk_tx_queue_clear(sk
);
1656 sk
->sk_socket
= sock
;
1659 static inline wait_queue_head_t
*sk_sleep(struct sock
*sk
)
1661 BUILD_BUG_ON(offsetof(struct socket_wq
, wait
) != 0);
1662 return &rcu_dereference_raw(sk
->sk_wq
)->wait
;
1664 /* Detach socket from process context.
1665 * Announce socket dead, detach it from wait queue and inode.
1666 * Note that parent inode held reference count on this struct sock,
1667 * we do not release it in this function, because protocol
1668 * probably wants some additional cleanups or even continuing
1669 * to work with this socket (TCP).
1671 static inline void sock_orphan(struct sock
*sk
)
1673 write_lock_bh(&sk
->sk_callback_lock
);
1674 sock_set_flag(sk
, SOCK_DEAD
);
1675 sk_set_socket(sk
, NULL
);
1677 write_unlock_bh(&sk
->sk_callback_lock
);
1680 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1682 write_lock_bh(&sk
->sk_callback_lock
);
1683 sk
->sk_wq
= parent
->wq
;
1685 sk_set_socket(sk
, parent
);
1686 security_sock_graft(sk
, parent
);
1687 write_unlock_bh(&sk
->sk_callback_lock
);
1690 extern kuid_t
sock_i_uid(struct sock
*sk
);
1691 extern unsigned long sock_i_ino(struct sock
*sk
);
1693 static inline struct dst_entry
*
1694 __sk_dst_get(struct sock
*sk
)
1696 return rcu_dereference_check(sk
->sk_dst_cache
, sock_owned_by_user(sk
) ||
1697 lockdep_is_held(&sk
->sk_lock
.slock
));
1700 static inline struct dst_entry
*
1701 sk_dst_get(struct sock
*sk
)
1703 struct dst_entry
*dst
;
1706 dst
= rcu_dereference(sk
->sk_dst_cache
);
1713 extern void sk_reset_txq(struct sock
*sk
);
1715 static inline void dst_negative_advice(struct sock
*sk
)
1717 struct dst_entry
*ndst
, *dst
= __sk_dst_get(sk
);
1719 if (dst
&& dst
->ops
->negative_advice
) {
1720 ndst
= dst
->ops
->negative_advice(dst
);
1723 rcu_assign_pointer(sk
->sk_dst_cache
, ndst
);
1730 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1732 struct dst_entry
*old_dst
;
1734 sk_tx_queue_clear(sk
);
1736 * This can be called while sk is owned by the caller only,
1737 * with no state that can be checked in a rcu_dereference_check() cond
1739 old_dst
= rcu_dereference_raw(sk
->sk_dst_cache
);
1740 rcu_assign_pointer(sk
->sk_dst_cache
, dst
);
1741 dst_release(old_dst
);
1745 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1747 spin_lock(&sk
->sk_dst_lock
);
1748 __sk_dst_set(sk
, dst
);
1749 spin_unlock(&sk
->sk_dst_lock
);
1753 __sk_dst_reset(struct sock
*sk
)
1755 __sk_dst_set(sk
, NULL
);
1759 sk_dst_reset(struct sock
*sk
)
1761 spin_lock(&sk
->sk_dst_lock
);
1763 spin_unlock(&sk
->sk_dst_lock
);
1766 extern struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1768 extern struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1770 static inline bool sk_can_gso(const struct sock
*sk
)
1772 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1775 extern void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1777 static inline void sk_nocaps_add(struct sock
*sk
, netdev_features_t flags
)
1779 sk
->sk_route_nocaps
|= flags
;
1780 sk
->sk_route_caps
&= ~flags
;
1783 static inline int skb_do_copy_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1784 char __user
*from
, char *to
,
1785 int copy
, int offset
)
1787 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1789 __wsum csum
= csum_and_copy_from_user(from
, to
, copy
, 0, &err
);
1792 skb
->csum
= csum_block_add(skb
->csum
, csum
, offset
);
1793 } else if (sk
->sk_route_caps
& NETIF_F_NOCACHE_COPY
) {
1794 if (!access_ok(VERIFY_READ
, from
, copy
) ||
1795 __copy_from_user_nocache(to
, from
, copy
))
1797 } else if (copy_from_user(to
, from
, copy
))
1803 static inline int skb_add_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1804 char __user
*from
, int copy
)
1806 int err
, offset
= skb
->len
;
1808 err
= skb_do_copy_data_nocache(sk
, skb
, from
, skb_put(skb
, copy
),
1811 __skb_trim(skb
, offset
);
1816 static inline int skb_copy_to_page_nocache(struct sock
*sk
, char __user
*from
,
1817 struct sk_buff
*skb
,
1823 err
= skb_do_copy_data_nocache(sk
, skb
, from
, page_address(page
) + off
,
1829 skb
->data_len
+= copy
;
1830 skb
->truesize
+= copy
;
1831 sk
->sk_wmem_queued
+= copy
;
1832 sk_mem_charge(sk
, copy
);
1836 static inline int skb_copy_to_page(struct sock
*sk
, char __user
*from
,
1837 struct sk_buff
*skb
, struct page
*page
,
1840 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1842 __wsum csum
= csum_and_copy_from_user(from
,
1843 page_address(page
) + off
,
1847 skb
->csum
= csum_block_add(skb
->csum
, csum
, skb
->len
);
1848 } else if (copy_from_user(page_address(page
) + off
, from
, copy
))
1852 skb
->data_len
+= copy
;
1853 skb
->truesize
+= copy
;
1854 sk
->sk_wmem_queued
+= copy
;
1855 sk_mem_charge(sk
, copy
);
1860 * sk_wmem_alloc_get - returns write allocations
1863 * Returns sk_wmem_alloc minus initial offset of one
1865 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
1867 return atomic_read(&sk
->sk_wmem_alloc
) - 1;
1871 * sk_rmem_alloc_get - returns read allocations
1874 * Returns sk_rmem_alloc
1876 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
1878 return atomic_read(&sk
->sk_rmem_alloc
);
1882 * sk_has_allocations - check if allocations are outstanding
1885 * Returns true if socket has write or read allocations
1887 static inline bool sk_has_allocations(const struct sock
*sk
)
1889 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
1893 * wq_has_sleeper - check if there are any waiting processes
1894 * @wq: struct socket_wq
1896 * Returns true if socket_wq has waiting processes
1898 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1899 * barrier call. They were added due to the race found within the tcp code.
1901 * Consider following tcp code paths:
1905 * sys_select receive packet
1907 * __add_wait_queue update tp->rcv_nxt
1909 * tp->rcv_nxt check sock_def_readable
1911 * schedule rcu_read_lock();
1912 * wq = rcu_dereference(sk->sk_wq);
1913 * if (wq && waitqueue_active(&wq->wait))
1914 * wake_up_interruptible(&wq->wait)
1918 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1919 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1920 * could then endup calling schedule and sleep forever if there are no more
1921 * data on the socket.
1924 static inline bool wq_has_sleeper(struct socket_wq
*wq
)
1926 /* We need to be sure we are in sync with the
1927 * add_wait_queue modifications to the wait queue.
1929 * This memory barrier is paired in the sock_poll_wait.
1932 return wq
&& waitqueue_active(&wq
->wait
);
1936 * sock_poll_wait - place memory barrier behind the poll_wait call.
1938 * @wait_address: socket wait queue
1941 * See the comments in the wq_has_sleeper function.
1943 static inline void sock_poll_wait(struct file
*filp
,
1944 wait_queue_head_t
*wait_address
, poll_table
*p
)
1946 if (!poll_does_not_wait(p
) && wait_address
) {
1947 poll_wait(filp
, wait_address
, p
);
1948 /* We need to be sure we are in sync with the
1949 * socket flags modification.
1951 * This memory barrier is paired in the wq_has_sleeper.
1958 * Queue a received datagram if it will fit. Stream and sequenced
1959 * protocols can't normally use this as they need to fit buffers in
1960 * and play with them.
1962 * Inlined as it's very short and called for pretty much every
1963 * packet ever received.
1966 static inline void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
1970 skb
->destructor
= sock_wfree
;
1972 * We used to take a refcount on sk, but following operation
1973 * is enough to guarantee sk_free() wont free this sock until
1974 * all in-flight packets are completed
1976 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
1979 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
1983 skb
->destructor
= sock_rfree
;
1984 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
1985 sk_mem_charge(sk
, skb
->truesize
);
1988 extern void sk_reset_timer(struct sock
*sk
, struct timer_list
*timer
,
1989 unsigned long expires
);
1991 extern void sk_stop_timer(struct sock
*sk
, struct timer_list
*timer
);
1993 extern int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
1995 extern int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
);
1998 * Recover an error report and clear atomically
2001 static inline int sock_error(struct sock
*sk
)
2004 if (likely(!sk
->sk_err
))
2006 err
= xchg(&sk
->sk_err
, 0);
2010 static inline unsigned long sock_wspace(struct sock
*sk
)
2014 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
2015 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
2022 static inline void sk_wake_async(struct sock
*sk
, int how
, int band
)
2024 if (sock_flag(sk
, SOCK_FASYNC
))
2025 sock_wake_async(sk
->sk_socket
, how
, band
);
2028 #define SOCK_MIN_SNDBUF 2048
2030 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
2031 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
2033 #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
2035 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
2037 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
2038 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
2039 sk
->sk_sndbuf
= max(sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
2043 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
);
2046 * sk_page_frag - return an appropriate page_frag
2049 * If socket allocation mode allows current thread to sleep, it means its
2050 * safe to use the per task page_frag instead of the per socket one.
2052 static inline struct page_frag
*sk_page_frag(struct sock
*sk
)
2054 if (sk
->sk_allocation
& __GFP_WAIT
)
2055 return ¤t
->task_frag
;
2057 return &sk
->sk_frag
;
2060 extern bool sk_page_frag_refill(struct sock
*sk
, struct page_frag
*pfrag
);
2063 * Default write policy as shown to user space via poll/select/SIGIO
2065 static inline bool sock_writeable(const struct sock
*sk
)
2067 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
2070 static inline gfp_t
gfp_any(void)
2072 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
2075 static inline long sock_rcvtimeo(const struct sock
*sk
, bool noblock
)
2077 return noblock
? 0 : sk
->sk_rcvtimeo
;
2080 static inline long sock_sndtimeo(const struct sock
*sk
, bool noblock
)
2082 return noblock
? 0 : sk
->sk_sndtimeo
;
2085 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
2087 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
2090 /* Alas, with timeout socket operations are not restartable.
2091 * Compare this to poll().
2093 static inline int sock_intr_errno(long timeo
)
2095 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
2098 extern void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
2099 struct sk_buff
*skb
);
2100 extern void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
2101 struct sk_buff
*skb
);
2104 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
2106 ktime_t kt
= skb
->tstamp
;
2107 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
2110 * generate control messages if
2111 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2112 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2113 * - software time stamp available and wanted
2114 * (SOCK_TIMESTAMPING_SOFTWARE)
2115 * - hardware time stamps available and wanted
2116 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2117 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2119 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
2120 sock_flag(sk
, SOCK_TIMESTAMPING_RX_SOFTWARE
) ||
2121 (kt
.tv64
&& sock_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
)) ||
2122 (hwtstamps
->hwtstamp
.tv64
&&
2123 sock_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
)) ||
2124 (hwtstamps
->syststamp
.tv64
&&
2125 sock_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
)))
2126 __sock_recv_timestamp(msg
, sk
, skb
);
2130 if (sock_flag(sk
, SOCK_WIFI_STATUS
) && skb
->wifi_acked_valid
)
2131 __sock_recv_wifi_status(msg
, sk
, skb
);
2134 extern void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2135 struct sk_buff
*skb
);
2137 static inline void sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2138 struct sk_buff
*skb
)
2140 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2141 (1UL << SOCK_RCVTSTAMP) | \
2142 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2143 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
2144 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
2145 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2147 if (sk
->sk_flags
& FLAGS_TS_OR_DROPS
)
2148 __sock_recv_ts_and_drops(msg
, sk
, skb
);
2150 sk
->sk_stamp
= skb
->tstamp
;
2154 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2155 * @sk: socket sending this packet
2156 * @tx_flags: filled with instructions for time stamping
2158 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
2159 * parameters are invalid.
2161 extern int sock_tx_timestamp(struct sock
*sk
, __u8
*tx_flags
);
2164 * sk_eat_skb - Release a skb if it is no longer needed
2165 * @sk: socket to eat this skb from
2166 * @skb: socket buffer to eat
2167 * @copied_early: flag indicating whether DMA operations copied this data early
2169 * This routine must be called with interrupts disabled or with the socket
2170 * locked so that the sk_buff queue operation is ok.
2172 #ifdef CONFIG_NET_DMA
2173 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, bool copied_early
)
2175 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2179 __skb_queue_tail(&sk
->sk_async_wait_queue
, skb
);
2182 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, bool copied_early
)
2184 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2190 struct net
*sock_net(const struct sock
*sk
)
2192 return read_pnet(&sk
->sk_net
);
2196 void sock_net_set(struct sock
*sk
, struct net
*net
)
2198 write_pnet(&sk
->sk_net
, net
);
2202 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
2203 * They should not hold a reference to a namespace in order to allow
2205 * Sockets after sk_change_net should be released using sk_release_kernel
2207 static inline void sk_change_net(struct sock
*sk
, struct net
*net
)
2209 put_net(sock_net(sk
));
2210 sock_net_set(sk
, hold_net(net
));
2213 static inline struct sock
*skb_steal_sock(struct sk_buff
*skb
)
2216 struct sock
*sk
= skb
->sk
;
2218 skb
->destructor
= NULL
;
2225 extern void sock_enable_timestamp(struct sock
*sk
, int flag
);
2226 extern int sock_get_timestamp(struct sock
*, struct timeval __user
*);
2227 extern int sock_get_timestampns(struct sock
*, struct timespec __user
*);
2230 * Enable debug/info messages
2232 extern int net_msg_warn
;
2233 #define NETDEBUG(fmt, args...) \
2234 do { if (net_msg_warn) printk(fmt,##args); } while (0)
2236 #define LIMIT_NETDEBUG(fmt, args...) \
2237 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
2239 extern __u32 sysctl_wmem_max
;
2240 extern __u32 sysctl_rmem_max
;
2242 extern int sysctl_optmem_max
;
2244 extern __u32 sysctl_wmem_default
;
2245 extern __u32 sysctl_rmem_default
;
2247 #endif /* _SOCK_H */