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_reuseport: %SO_REUSEPORT setting
144 * @skc_bound_dev_if: bound device index if != 0
145 * @skc_bind_node: bind hash linkage for various protocol lookup tables
146 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
147 * @skc_prot: protocol handlers inside a network family
148 * @skc_net: reference to the network namespace of this socket
149 * @skc_node: main hash linkage for various protocol lookup tables
150 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
151 * @skc_tx_queue_mapping: tx queue number for this connection
152 * @skc_refcnt: reference count
154 * This is the minimal network layer representation of sockets, the header
155 * for struct sock and struct inet_timewait_sock.
158 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
159 * address on 64bit arches : cf INET_MATCH() and INET_TW_MATCH()
162 __addrpair skc_addrpair
;
165 __be32 skc_rcv_saddr
;
169 unsigned int skc_hash
;
170 __u16 skc_u16hashes
[2];
172 /* skc_dport && skc_num must be grouped as well */
174 __portpair skc_portpair
;
181 unsigned short skc_family
;
182 volatile unsigned char skc_state
;
183 unsigned char skc_reuse
:4;
184 unsigned char skc_reuseport
:4;
185 int skc_bound_dev_if
;
187 struct hlist_node skc_bind_node
;
188 struct hlist_nulls_node skc_portaddr_node
;
190 struct proto
*skc_prot
;
195 * fields between dontcopy_begin/dontcopy_end
196 * are not copied in sock_copy()
199 int skc_dontcopy_begin
[0];
202 struct hlist_node skc_node
;
203 struct hlist_nulls_node skc_nulls_node
;
205 int skc_tx_queue_mapping
;
208 int skc_dontcopy_end
[0];
214 * struct sock - network layer representation of sockets
215 * @__sk_common: shared layout with inet_timewait_sock
216 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
217 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
218 * @sk_lock: synchronizer
219 * @sk_rcvbuf: size of receive buffer in bytes
220 * @sk_wq: sock wait queue and async head
221 * @sk_rx_dst: receive input route used by early tcp demux
222 * @sk_dst_cache: destination cache
223 * @sk_dst_lock: destination cache lock
224 * @sk_policy: flow policy
225 * @sk_receive_queue: incoming packets
226 * @sk_wmem_alloc: transmit queue bytes committed
227 * @sk_write_queue: Packet sending queue
228 * @sk_async_wait_queue: DMA copied packets
229 * @sk_omem_alloc: "o" is "option" or "other"
230 * @sk_wmem_queued: persistent queue size
231 * @sk_forward_alloc: space allocated forward
232 * @sk_napi_id: id of the last napi context to receive data for sk
233 * @sk_ll_usec: usecs to busypoll when there is no data
234 * @sk_allocation: allocation mode
235 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
236 * @sk_sndbuf: size of send buffer in bytes
237 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
238 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
239 * @sk_no_check: %SO_NO_CHECK setting, whether or not checkup packets
240 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
241 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
242 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
243 * @sk_gso_max_size: Maximum GSO segment size to build
244 * @sk_gso_max_segs: Maximum number of GSO segments
245 * @sk_lingertime: %SO_LINGER l_linger setting
246 * @sk_backlog: always used with the per-socket spinlock held
247 * @sk_callback_lock: used with the callbacks in the end of this struct
248 * @sk_error_queue: rarely used
249 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
250 * IPV6_ADDRFORM for instance)
251 * @sk_err: last error
252 * @sk_err_soft: errors that don't cause failure but are the cause of a
253 * persistent failure not just 'timed out'
254 * @sk_drops: raw/udp drops counter
255 * @sk_ack_backlog: current listen backlog
256 * @sk_max_ack_backlog: listen backlog set in listen()
257 * @sk_priority: %SO_PRIORITY setting
258 * @sk_cgrp_prioidx: socket group's priority map index
259 * @sk_type: socket type (%SOCK_STREAM, etc)
260 * @sk_protocol: which protocol this socket belongs in this network family
261 * @sk_peer_pid: &struct pid for this socket's peer
262 * @sk_peer_cred: %SO_PEERCRED setting
263 * @sk_rcvlowat: %SO_RCVLOWAT setting
264 * @sk_rcvtimeo: %SO_RCVTIMEO setting
265 * @sk_sndtimeo: %SO_SNDTIMEO setting
266 * @sk_rxhash: flow hash received from netif layer
267 * @sk_filter: socket filtering instructions
268 * @sk_protinfo: private area, net family specific, when not using slab
269 * @sk_timer: sock cleanup timer
270 * @sk_stamp: time stamp of last packet received
271 * @sk_socket: Identd and reporting IO signals
272 * @sk_user_data: RPC layer private data
273 * @sk_frag: cached page frag
274 * @sk_peek_off: current peek_offset value
275 * @sk_send_head: front of stuff to transmit
276 * @sk_security: used by security modules
277 * @sk_mark: generic packet mark
278 * @sk_classid: this socket's cgroup classid
279 * @sk_cgrp: this socket's cgroup-specific proto data
280 * @sk_write_pending: a write to stream socket waits to start
281 * @sk_state_change: callback to indicate change in the state of the sock
282 * @sk_data_ready: callback to indicate there is data to be processed
283 * @sk_write_space: callback to indicate there is bf sending space available
284 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
285 * @sk_backlog_rcv: callback to process the backlog
286 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
290 * Now struct inet_timewait_sock also uses sock_common, so please just
291 * don't add nothing before this first member (__sk_common) --acme
293 struct sock_common __sk_common
;
294 #define sk_node __sk_common.skc_node
295 #define sk_nulls_node __sk_common.skc_nulls_node
296 #define sk_refcnt __sk_common.skc_refcnt
297 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
299 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
300 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
301 #define sk_hash __sk_common.skc_hash
302 #define sk_family __sk_common.skc_family
303 #define sk_state __sk_common.skc_state
304 #define sk_reuse __sk_common.skc_reuse
305 #define sk_reuseport __sk_common.skc_reuseport
306 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
307 #define sk_bind_node __sk_common.skc_bind_node
308 #define sk_prot __sk_common.skc_prot
309 #define sk_net __sk_common.skc_net
310 socket_lock_t sk_lock
;
311 struct sk_buff_head sk_receive_queue
;
313 * The backlog queue is special, it is always used with
314 * the per-socket spinlock held and requires low latency
315 * access. Therefore we special case it's implementation.
316 * Note : rmem_alloc is in this structure to fill a hole
317 * on 64bit arches, not because its logically part of
323 struct sk_buff
*head
;
324 struct sk_buff
*tail
;
326 #define sk_rmem_alloc sk_backlog.rmem_alloc
327 int sk_forward_alloc
;
331 #ifdef CONFIG_NET_RX_BUSY_POLL
332 unsigned int sk_napi_id
;
333 unsigned int sk_ll_usec
;
338 struct sk_filter __rcu
*sk_filter
;
339 struct socket_wq __rcu
*sk_wq
;
341 #ifdef CONFIG_NET_DMA
342 struct sk_buff_head sk_async_wait_queue
;
346 struct xfrm_policy
*sk_policy
[2];
348 unsigned long sk_flags
;
349 struct dst_entry
*sk_rx_dst
;
350 struct dst_entry __rcu
*sk_dst_cache
;
351 spinlock_t sk_dst_lock
;
352 atomic_t sk_wmem_alloc
;
353 atomic_t sk_omem_alloc
;
355 struct sk_buff_head sk_write_queue
;
356 kmemcheck_bitfield_begin(flags
);
357 unsigned int sk_shutdown
: 2,
362 kmemcheck_bitfield_end(flags
);
365 u32 sk_pacing_rate
; /* bytes per second */
366 netdev_features_t sk_route_caps
;
367 netdev_features_t sk_route_nocaps
;
369 unsigned int sk_gso_max_size
;
372 unsigned long sk_lingertime
;
373 struct sk_buff_head sk_error_queue
;
374 struct proto
*sk_prot_creator
;
375 rwlock_t sk_callback_lock
;
378 unsigned short sk_ack_backlog
;
379 unsigned short sk_max_ack_backlog
;
381 #if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
382 __u32 sk_cgrp_prioidx
;
384 struct pid
*sk_peer_pid
;
385 const struct cred
*sk_peer_cred
;
389 struct timer_list sk_timer
;
391 struct socket
*sk_socket
;
393 struct page_frag sk_frag
;
394 struct sk_buff
*sk_send_head
;
396 int sk_write_pending
;
397 #ifdef CONFIG_SECURITY
402 struct cg_proto
*sk_cgrp
;
403 void (*sk_state_change
)(struct sock
*sk
);
404 void (*sk_data_ready
)(struct sock
*sk
, int bytes
);
405 void (*sk_write_space
)(struct sock
*sk
);
406 void (*sk_error_report
)(struct sock
*sk
);
407 int (*sk_backlog_rcv
)(struct sock
*sk
,
408 struct sk_buff
*skb
);
409 void (*sk_destruct
)(struct sock
*sk
);
412 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
414 #define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
415 #define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
418 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
419 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
420 * on a socket means that the socket will reuse everybody else's port
421 * without looking at the other's sk_reuse value.
424 #define SK_NO_REUSE 0
425 #define SK_CAN_REUSE 1
426 #define SK_FORCE_REUSE 2
428 static inline int sk_peek_offset(struct sock
*sk
, int flags
)
430 if ((flags
& MSG_PEEK
) && (sk
->sk_peek_off
>= 0))
431 return sk
->sk_peek_off
;
436 static inline void sk_peek_offset_bwd(struct sock
*sk
, int val
)
438 if (sk
->sk_peek_off
>= 0) {
439 if (sk
->sk_peek_off
>= val
)
440 sk
->sk_peek_off
-= val
;
446 static inline void sk_peek_offset_fwd(struct sock
*sk
, int val
)
448 if (sk
->sk_peek_off
>= 0)
449 sk
->sk_peek_off
+= val
;
453 * Hashed lists helper routines
455 static inline struct sock
*sk_entry(const struct hlist_node
*node
)
457 return hlist_entry(node
, struct sock
, sk_node
);
460 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
462 return hlist_entry(head
->first
, struct sock
, sk_node
);
465 static inline struct sock
*sk_head(const struct hlist_head
*head
)
467 return hlist_empty(head
) ? NULL
: __sk_head(head
);
470 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
472 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
475 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
477 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
480 static inline struct sock
*sk_next(const struct sock
*sk
)
482 return sk
->sk_node
.next
?
483 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
486 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
488 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
489 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
490 struct sock
, sk_nulls_node
) :
494 static inline bool sk_unhashed(const struct sock
*sk
)
496 return hlist_unhashed(&sk
->sk_node
);
499 static inline bool sk_hashed(const struct sock
*sk
)
501 return !sk_unhashed(sk
);
504 static inline void sk_node_init(struct hlist_node
*node
)
509 static inline void sk_nulls_node_init(struct hlist_nulls_node
*node
)
514 static inline void __sk_del_node(struct sock
*sk
)
516 __hlist_del(&sk
->sk_node
);
519 /* NB: equivalent to hlist_del_init_rcu */
520 static inline bool __sk_del_node_init(struct sock
*sk
)
524 sk_node_init(&sk
->sk_node
);
530 /* Grab socket reference count. This operation is valid only
531 when sk is ALREADY grabbed f.e. it is found in hash table
532 or a list and the lookup is made under lock preventing hash table
536 static inline void sock_hold(struct sock
*sk
)
538 atomic_inc(&sk
->sk_refcnt
);
541 /* Ungrab socket in the context, which assumes that socket refcnt
542 cannot hit zero, f.e. it is true in context of any socketcall.
544 static inline void __sock_put(struct sock
*sk
)
546 atomic_dec(&sk
->sk_refcnt
);
549 static inline bool sk_del_node_init(struct sock
*sk
)
551 bool rc
= __sk_del_node_init(sk
);
554 /* paranoid for a while -acme */
555 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
560 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
562 static inline bool __sk_nulls_del_node_init_rcu(struct sock
*sk
)
565 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
571 static inline bool sk_nulls_del_node_init_rcu(struct sock
*sk
)
573 bool rc
= __sk_nulls_del_node_init_rcu(sk
);
576 /* paranoid for a while -acme */
577 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
583 static inline void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
585 hlist_add_head(&sk
->sk_node
, list
);
588 static inline void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
591 __sk_add_node(sk
, list
);
594 static inline void sk_add_node_rcu(struct sock
*sk
, struct hlist_head
*list
)
597 hlist_add_head_rcu(&sk
->sk_node
, list
);
600 static inline void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
602 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
605 static inline void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
608 __sk_nulls_add_node_rcu(sk
, list
);
611 static inline void __sk_del_bind_node(struct sock
*sk
)
613 __hlist_del(&sk
->sk_bind_node
);
616 static inline void sk_add_bind_node(struct sock
*sk
,
617 struct hlist_head
*list
)
619 hlist_add_head(&sk
->sk_bind_node
, list
);
622 #define sk_for_each(__sk, list) \
623 hlist_for_each_entry(__sk, list, sk_node)
624 #define sk_for_each_rcu(__sk, list) \
625 hlist_for_each_entry_rcu(__sk, list, sk_node)
626 #define sk_nulls_for_each(__sk, node, list) \
627 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
628 #define sk_nulls_for_each_rcu(__sk, node, list) \
629 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
630 #define sk_for_each_from(__sk) \
631 hlist_for_each_entry_from(__sk, sk_node)
632 #define sk_nulls_for_each_from(__sk, node) \
633 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
634 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
635 #define sk_for_each_safe(__sk, tmp, list) \
636 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
637 #define sk_for_each_bound(__sk, list) \
638 hlist_for_each_entry(__sk, list, sk_bind_node)
640 static inline struct user_namespace
*sk_user_ns(struct sock
*sk
)
642 /* Careful only use this in a context where these parameters
643 * can not change and must all be valid, such as recvmsg from
646 return sk
->sk_socket
->file
->f_cred
->user_ns
;
660 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
661 SOCK_DBG
, /* %SO_DEBUG setting */
662 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
663 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
664 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
665 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
666 SOCK_MEMALLOC
, /* VM depends on this socket for swapping */
667 SOCK_TIMESTAMPING_TX_HARDWARE
, /* %SOF_TIMESTAMPING_TX_HARDWARE */
668 SOCK_TIMESTAMPING_TX_SOFTWARE
, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
669 SOCK_TIMESTAMPING_RX_HARDWARE
, /* %SOF_TIMESTAMPING_RX_HARDWARE */
670 SOCK_TIMESTAMPING_RX_SOFTWARE
, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
671 SOCK_TIMESTAMPING_SOFTWARE
, /* %SOF_TIMESTAMPING_SOFTWARE */
672 SOCK_TIMESTAMPING_RAW_HARDWARE
, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
673 SOCK_TIMESTAMPING_SYS_HARDWARE
, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
674 SOCK_FASYNC
, /* fasync() active */
676 SOCK_ZEROCOPY
, /* buffers from userspace */
677 SOCK_WIFI_STATUS
, /* push wifi status to userspace */
678 SOCK_NOFCS
, /* Tell NIC not to do the Ethernet FCS.
679 * Will use last 4 bytes of packet sent from
680 * user-space instead.
682 SOCK_FILTER_LOCKED
, /* Filter cannot be changed anymore */
683 SOCK_SELECT_ERR_QUEUE
, /* Wake select on error queue */
686 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
688 nsk
->sk_flags
= osk
->sk_flags
;
691 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
693 __set_bit(flag
, &sk
->sk_flags
);
696 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
698 __clear_bit(flag
, &sk
->sk_flags
);
701 static inline bool sock_flag(const struct sock
*sk
, enum sock_flags flag
)
703 return test_bit(flag
, &sk
->sk_flags
);
707 extern struct static_key memalloc_socks
;
708 static inline int sk_memalloc_socks(void)
710 return static_key_false(&memalloc_socks
);
714 static inline int sk_memalloc_socks(void)
721 static inline gfp_t
sk_gfp_atomic(struct sock
*sk
, gfp_t gfp_mask
)
723 return GFP_ATOMIC
| (sk
->sk_allocation
& __GFP_MEMALLOC
);
726 static inline void sk_acceptq_removed(struct sock
*sk
)
728 sk
->sk_ack_backlog
--;
731 static inline void sk_acceptq_added(struct sock
*sk
)
733 sk
->sk_ack_backlog
++;
736 static inline bool sk_acceptq_is_full(const struct sock
*sk
)
738 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
742 * Compute minimal free write space needed to queue new packets.
744 static inline int sk_stream_min_wspace(const struct sock
*sk
)
746 return sk
->sk_wmem_queued
>> 1;
749 static inline int sk_stream_wspace(const struct sock
*sk
)
751 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
754 extern void sk_stream_write_space(struct sock
*sk
);
756 /* OOB backlog add */
757 static inline void __sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
759 /* dont let skb dst not refcounted, we are going to leave rcu lock */
762 if (!sk
->sk_backlog
.tail
)
763 sk
->sk_backlog
.head
= skb
;
765 sk
->sk_backlog
.tail
->next
= skb
;
767 sk
->sk_backlog
.tail
= skb
;
772 * Take into account size of receive queue and backlog queue
773 * Do not take into account this skb truesize,
774 * to allow even a single big packet to come.
776 static inline bool sk_rcvqueues_full(const struct sock
*sk
, const struct sk_buff
*skb
,
779 unsigned int qsize
= sk
->sk_backlog
.len
+ atomic_read(&sk
->sk_rmem_alloc
);
781 return qsize
> limit
;
784 /* The per-socket spinlock must be held here. */
785 static inline __must_check
int sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
,
788 if (sk_rcvqueues_full(sk
, skb
, limit
))
791 __sk_add_backlog(sk
, skb
);
792 sk
->sk_backlog
.len
+= skb
->truesize
;
796 extern int __sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
798 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
800 if (sk_memalloc_socks() && skb_pfmemalloc(skb
))
801 return __sk_backlog_rcv(sk
, skb
);
803 return sk
->sk_backlog_rcv(sk
, skb
);
806 static inline void sock_rps_record_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_record_sock_flow(sock_flow_table
, sk
->sk_rxhash
);
818 static inline void sock_rps_reset_flow(const struct sock
*sk
)
821 struct rps_sock_flow_table
*sock_flow_table
;
824 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
825 rps_reset_sock_flow(sock_flow_table
, sk
->sk_rxhash
);
830 static inline void sock_rps_save_rxhash(struct sock
*sk
,
831 const struct sk_buff
*skb
)
834 if (unlikely(sk
->sk_rxhash
!= skb
->rxhash
)) {
835 sock_rps_reset_flow(sk
);
836 sk
->sk_rxhash
= skb
->rxhash
;
841 static inline void sock_rps_reset_rxhash(struct sock
*sk
)
844 sock_rps_reset_flow(sk
);
849 #define sk_wait_event(__sk, __timeo, __condition) \
851 release_sock(__sk); \
852 __rc = __condition; \
854 *(__timeo) = schedule_timeout(*(__timeo)); \
857 __rc = __condition; \
861 extern int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
862 extern int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
863 extern void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
864 extern int sk_stream_error(struct sock
*sk
, int flags
, int err
);
865 extern void sk_stream_kill_queues(struct sock
*sk
);
866 extern void sk_set_memalloc(struct sock
*sk
);
867 extern void sk_clear_memalloc(struct sock
*sk
);
869 extern int sk_wait_data(struct sock
*sk
, long *timeo
);
871 struct request_sock_ops
;
872 struct timewait_sock_ops
;
873 struct inet_hashinfo
;
878 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
879 * un-modified. Special care is taken when initializing object to zero.
881 static inline void sk_prot_clear_nulls(struct sock
*sk
, int size
)
883 if (offsetof(struct sock
, sk_node
.next
) != 0)
884 memset(sk
, 0, offsetof(struct sock
, sk_node
.next
));
885 memset(&sk
->sk_node
.pprev
, 0,
886 size
- offsetof(struct sock
, sk_node
.pprev
));
889 /* Networking protocol blocks we attach to sockets.
890 * socket layer -> transport layer interface
891 * transport -> network interface is defined by struct inet_proto
894 void (*close
)(struct sock
*sk
,
896 int (*connect
)(struct sock
*sk
,
897 struct sockaddr
*uaddr
,
899 int (*disconnect
)(struct sock
*sk
, int flags
);
901 struct sock
* (*accept
)(struct sock
*sk
, int flags
, int *err
);
903 int (*ioctl
)(struct sock
*sk
, int cmd
,
905 int (*init
)(struct sock
*sk
);
906 void (*destroy
)(struct sock
*sk
);
907 void (*shutdown
)(struct sock
*sk
, int how
);
908 int (*setsockopt
)(struct sock
*sk
, int level
,
909 int optname
, char __user
*optval
,
910 unsigned int optlen
);
911 int (*getsockopt
)(struct sock
*sk
, int level
,
912 int optname
, char __user
*optval
,
915 int (*compat_setsockopt
)(struct sock
*sk
,
917 int optname
, char __user
*optval
,
918 unsigned int optlen
);
919 int (*compat_getsockopt
)(struct sock
*sk
,
921 int optname
, char __user
*optval
,
923 int (*compat_ioctl
)(struct sock
*sk
,
924 unsigned int cmd
, unsigned long arg
);
926 int (*sendmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
927 struct msghdr
*msg
, size_t len
);
928 int (*recvmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
930 size_t len
, int noblock
, int flags
,
932 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
933 int offset
, size_t size
, int flags
);
934 int (*bind
)(struct sock
*sk
,
935 struct sockaddr
*uaddr
, int addr_len
);
937 int (*backlog_rcv
) (struct sock
*sk
,
938 struct sk_buff
*skb
);
940 void (*release_cb
)(struct sock
*sk
);
942 /* Keeping track of sk's, looking them up, and port selection methods. */
943 void (*hash
)(struct sock
*sk
);
944 void (*unhash
)(struct sock
*sk
);
945 void (*rehash
)(struct sock
*sk
);
946 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
947 void (*clear_sk
)(struct sock
*sk
, int size
);
949 /* Keeping track of sockets in use */
950 #ifdef CONFIG_PROC_FS
951 unsigned int inuse_idx
;
954 bool (*stream_memory_free
)(const struct sock
*sk
);
955 /* Memory pressure */
956 void (*enter_memory_pressure
)(struct sock
*sk
);
957 atomic_long_t
*memory_allocated
; /* Current allocated memory. */
958 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
960 * Pressure flag: try to collapse.
961 * Technical note: it is used by multiple contexts non atomically.
962 * All the __sk_mem_schedule() is of this nature: accounting
963 * is strict, actions are advisory and have some latency.
965 int *memory_pressure
;
972 struct kmem_cache
*slab
;
973 unsigned int obj_size
;
976 struct percpu_counter
*orphan_count
;
978 struct request_sock_ops
*rsk_prot
;
979 struct timewait_sock_ops
*twsk_prot
;
982 struct inet_hashinfo
*hashinfo
;
983 struct udp_table
*udp_table
;
984 struct raw_hashinfo
*raw_hash
;
987 struct module
*owner
;
991 struct list_head node
;
992 #ifdef SOCK_REFCNT_DEBUG
995 #ifdef CONFIG_MEMCG_KMEM
997 * cgroup specific init/deinit functions. Called once for all
998 * protocols that implement it, from cgroups populate function.
999 * This function has to setup any files the protocol want to
1000 * appear in the kmem cgroup filesystem.
1002 int (*init_cgroup
)(struct mem_cgroup
*memcg
,
1003 struct cgroup_subsys
*ss
);
1004 void (*destroy_cgroup
)(struct mem_cgroup
*memcg
);
1005 struct cg_proto
*(*proto_cgroup
)(struct mem_cgroup
*memcg
);
1010 * Bits in struct cg_proto.flags
1012 enum cg_proto_flags
{
1013 /* Currently active and new sockets should be assigned to cgroups */
1015 /* It was ever activated; we must disarm static keys on destruction */
1016 MEMCG_SOCK_ACTIVATED
,
1020 void (*enter_memory_pressure
)(struct sock
*sk
);
1021 struct res_counter
*memory_allocated
; /* Current allocated memory. */
1022 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
1023 int *memory_pressure
;
1025 unsigned long flags
;
1027 * memcg field is used to find which memcg we belong directly
1028 * Each memcg struct can hold more than one cg_proto, so container_of
1031 * The elegant solution would be having an inverse function to
1032 * proto_cgroup in struct proto, but that means polluting the structure
1033 * for everybody, instead of just for memcg users.
1035 struct mem_cgroup
*memcg
;
1038 extern int proto_register(struct proto
*prot
, int alloc_slab
);
1039 extern void proto_unregister(struct proto
*prot
);
1041 static inline bool memcg_proto_active(struct cg_proto
*cg_proto
)
1043 return test_bit(MEMCG_SOCK_ACTIVE
, &cg_proto
->flags
);
1046 static inline bool memcg_proto_activated(struct cg_proto
*cg_proto
)
1048 return test_bit(MEMCG_SOCK_ACTIVATED
, &cg_proto
->flags
);
1051 #ifdef SOCK_REFCNT_DEBUG
1052 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
1054 atomic_inc(&sk
->sk_prot
->socks
);
1057 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
1059 atomic_dec(&sk
->sk_prot
->socks
);
1060 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
1061 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
1064 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
1066 if (atomic_read(&sk
->sk_refcnt
) != 1)
1067 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
1068 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
1070 #else /* SOCK_REFCNT_DEBUG */
1071 #define sk_refcnt_debug_inc(sk) do { } while (0)
1072 #define sk_refcnt_debug_dec(sk) do { } while (0)
1073 #define sk_refcnt_debug_release(sk) do { } while (0)
1074 #endif /* SOCK_REFCNT_DEBUG */
1076 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
1077 extern struct static_key memcg_socket_limit_enabled
;
1078 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
1079 struct cg_proto
*cg_proto
)
1081 return proto
->proto_cgroup(parent_mem_cgroup(cg_proto
->memcg
));
1083 #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
1085 #define mem_cgroup_sockets_enabled 0
1086 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
1087 struct cg_proto
*cg_proto
)
1093 static inline bool sk_stream_memory_free(const struct sock
*sk
)
1095 if (sk
->sk_wmem_queued
>= sk
->sk_sndbuf
)
1098 return sk
->sk_prot
->stream_memory_free
?
1099 sk
->sk_prot
->stream_memory_free(sk
) : true;
1102 static inline bool sk_stream_is_writeable(const struct sock
*sk
)
1104 return sk_stream_wspace(sk
) >= sk_stream_min_wspace(sk
) &&
1105 sk_stream_memory_free(sk
);
1109 static inline bool sk_has_memory_pressure(const struct sock
*sk
)
1111 return sk
->sk_prot
->memory_pressure
!= NULL
;
1114 static inline bool sk_under_memory_pressure(const struct sock
*sk
)
1116 if (!sk
->sk_prot
->memory_pressure
)
1119 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1120 return !!*sk
->sk_cgrp
->memory_pressure
;
1122 return !!*sk
->sk_prot
->memory_pressure
;
1125 static inline void sk_leave_memory_pressure(struct sock
*sk
)
1127 int *memory_pressure
= sk
->sk_prot
->memory_pressure
;
1129 if (!memory_pressure
)
1132 if (*memory_pressure
)
1133 *memory_pressure
= 0;
1135 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1136 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1137 struct proto
*prot
= sk
->sk_prot
;
1139 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1140 if (*cg_proto
->memory_pressure
)
1141 *cg_proto
->memory_pressure
= 0;
1146 static inline void sk_enter_memory_pressure(struct sock
*sk
)
1148 if (!sk
->sk_prot
->enter_memory_pressure
)
1151 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1152 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1153 struct proto
*prot
= sk
->sk_prot
;
1155 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1156 cg_proto
->enter_memory_pressure(sk
);
1159 sk
->sk_prot
->enter_memory_pressure(sk
);
1162 static inline long sk_prot_mem_limits(const struct sock
*sk
, int index
)
1164 long *prot
= sk
->sk_prot
->sysctl_mem
;
1165 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1166 prot
= sk
->sk_cgrp
->sysctl_mem
;
1170 static inline void memcg_memory_allocated_add(struct cg_proto
*prot
,
1174 struct res_counter
*fail
;
1177 ret
= res_counter_charge_nofail(prot
->memory_allocated
,
1178 amt
<< PAGE_SHIFT
, &fail
);
1180 *parent_status
= OVER_LIMIT
;
1183 static inline void memcg_memory_allocated_sub(struct cg_proto
*prot
,
1186 res_counter_uncharge(prot
->memory_allocated
, amt
<< PAGE_SHIFT
);
1189 static inline u64
memcg_memory_allocated_read(struct cg_proto
*prot
)
1192 ret
= res_counter_read_u64(prot
->memory_allocated
, RES_USAGE
);
1193 return ret
>> PAGE_SHIFT
;
1197 sk_memory_allocated(const struct sock
*sk
)
1199 struct proto
*prot
= sk
->sk_prot
;
1200 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1201 return memcg_memory_allocated_read(sk
->sk_cgrp
);
1203 return atomic_long_read(prot
->memory_allocated
);
1207 sk_memory_allocated_add(struct sock
*sk
, int amt
, int *parent_status
)
1209 struct proto
*prot
= sk
->sk_prot
;
1211 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1212 memcg_memory_allocated_add(sk
->sk_cgrp
, amt
, parent_status
);
1213 /* update the root cgroup regardless */
1214 atomic_long_add_return(amt
, prot
->memory_allocated
);
1215 return memcg_memory_allocated_read(sk
->sk_cgrp
);
1218 return atomic_long_add_return(amt
, prot
->memory_allocated
);
1222 sk_memory_allocated_sub(struct sock
*sk
, int amt
)
1224 struct proto
*prot
= sk
->sk_prot
;
1226 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1227 memcg_memory_allocated_sub(sk
->sk_cgrp
, amt
);
1229 atomic_long_sub(amt
, prot
->memory_allocated
);
1232 static inline void sk_sockets_allocated_dec(struct sock
*sk
)
1234 struct proto
*prot
= sk
->sk_prot
;
1236 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1237 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1239 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1240 percpu_counter_dec(cg_proto
->sockets_allocated
);
1243 percpu_counter_dec(prot
->sockets_allocated
);
1246 static inline void sk_sockets_allocated_inc(struct sock
*sk
)
1248 struct proto
*prot
= sk
->sk_prot
;
1250 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1251 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1253 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1254 percpu_counter_inc(cg_proto
->sockets_allocated
);
1257 percpu_counter_inc(prot
->sockets_allocated
);
1261 sk_sockets_allocated_read_positive(struct sock
*sk
)
1263 struct proto
*prot
= sk
->sk_prot
;
1265 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1266 return percpu_counter_read_positive(sk
->sk_cgrp
->sockets_allocated
);
1268 return percpu_counter_read_positive(prot
->sockets_allocated
);
1272 proto_sockets_allocated_sum_positive(struct proto
*prot
)
1274 return percpu_counter_sum_positive(prot
->sockets_allocated
);
1278 proto_memory_allocated(struct proto
*prot
)
1280 return atomic_long_read(prot
->memory_allocated
);
1284 proto_memory_pressure(struct proto
*prot
)
1286 if (!prot
->memory_pressure
)
1288 return !!*prot
->memory_pressure
;
1292 #ifdef CONFIG_PROC_FS
1293 /* Called with local bh disabled */
1294 extern void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int inc
);
1295 extern int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
1297 static inline void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
1304 /* With per-bucket locks this operation is not-atomic, so that
1305 * this version is not worse.
1307 static inline void __sk_prot_rehash(struct sock
*sk
)
1309 sk
->sk_prot
->unhash(sk
);
1310 sk
->sk_prot
->hash(sk
);
1313 void sk_prot_clear_portaddr_nulls(struct sock
*sk
, int size
);
1315 /* About 10 seconds */
1316 #define SOCK_DESTROY_TIME (10*HZ)
1318 /* Sockets 0-1023 can't be bound to unless you are superuser */
1319 #define PROT_SOCK 1024
1321 #define SHUTDOWN_MASK 3
1322 #define RCV_SHUTDOWN 1
1323 #define SEND_SHUTDOWN 2
1325 #define SOCK_SNDBUF_LOCK 1
1326 #define SOCK_RCVBUF_LOCK 2
1327 #define SOCK_BINDADDR_LOCK 4
1328 #define SOCK_BINDPORT_LOCK 8
1330 /* sock_iocb: used to kick off async processing of socket ios */
1332 struct list_head list
;
1336 struct socket
*sock
;
1338 struct scm_cookie
*scm
;
1339 struct msghdr
*msg
, async_msg
;
1340 struct kiocb
*kiocb
;
1343 static inline struct sock_iocb
*kiocb_to_siocb(struct kiocb
*iocb
)
1345 return (struct sock_iocb
*)iocb
->private;
1348 static inline struct kiocb
*siocb_to_kiocb(struct sock_iocb
*si
)
1353 struct socket_alloc
{
1354 struct socket socket
;
1355 struct inode vfs_inode
;
1358 static inline struct socket
*SOCKET_I(struct inode
*inode
)
1360 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
1363 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
1365 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
1369 * Functions for memory accounting
1371 extern int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
1372 extern void __sk_mem_reclaim(struct sock
*sk
);
1374 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1375 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1376 #define SK_MEM_SEND 0
1377 #define SK_MEM_RECV 1
1379 static inline int sk_mem_pages(int amt
)
1381 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
1384 static inline bool sk_has_account(struct sock
*sk
)
1386 /* return true if protocol supports memory accounting */
1387 return !!sk
->sk_prot
->memory_allocated
;
1390 static inline bool sk_wmem_schedule(struct sock
*sk
, int size
)
1392 if (!sk_has_account(sk
))
1394 return size
<= sk
->sk_forward_alloc
||
1395 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
1399 sk_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
, int size
)
1401 if (!sk_has_account(sk
))
1403 return size
<= sk
->sk_forward_alloc
||
1404 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
) ||
1405 skb_pfmemalloc(skb
);
1408 static inline void sk_mem_reclaim(struct sock
*sk
)
1410 if (!sk_has_account(sk
))
1412 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
1413 __sk_mem_reclaim(sk
);
1416 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
1418 if (!sk_has_account(sk
))
1420 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
1421 __sk_mem_reclaim(sk
);
1424 static inline void sk_mem_charge(struct sock
*sk
, int size
)
1426 if (!sk_has_account(sk
))
1428 sk
->sk_forward_alloc
-= size
;
1431 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
1433 if (!sk_has_account(sk
))
1435 sk
->sk_forward_alloc
+= size
;
1438 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
1440 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
1441 sk
->sk_wmem_queued
-= skb
->truesize
;
1442 sk_mem_uncharge(sk
, skb
->truesize
);
1446 /* Used by processes to "lock" a socket state, so that
1447 * interrupts and bottom half handlers won't change it
1448 * from under us. It essentially blocks any incoming
1449 * packets, so that we won't get any new data or any
1450 * packets that change the state of the socket.
1452 * While locked, BH processing will add new packets to
1453 * the backlog queue. This queue is processed by the
1454 * owner of the socket lock right before it is released.
1456 * Since ~2.3.5 it is also exclusive sleep lock serializing
1457 * accesses from user process context.
1459 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1461 static inline void sock_release_ownership(struct sock
*sk
)
1463 sk
->sk_lock
.owned
= 0;
1467 * Macro so as to not evaluate some arguments when
1468 * lockdep is not enabled.
1470 * Mark both the sk_lock and the sk_lock.slock as a
1471 * per-address-family lock class.
1473 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1475 sk->sk_lock.owned = 0; \
1476 init_waitqueue_head(&sk->sk_lock.wq); \
1477 spin_lock_init(&(sk)->sk_lock.slock); \
1478 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1479 sizeof((sk)->sk_lock)); \
1480 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1482 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1485 extern void lock_sock_nested(struct sock
*sk
, int subclass
);
1487 static inline void lock_sock(struct sock
*sk
)
1489 lock_sock_nested(sk
, 0);
1492 extern void release_sock(struct sock
*sk
);
1494 /* BH context may only use the following locking interface. */
1495 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1496 #define bh_lock_sock_nested(__sk) \
1497 spin_lock_nested(&((__sk)->sk_lock.slock), \
1498 SINGLE_DEPTH_NESTING)
1499 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1501 extern bool lock_sock_fast(struct sock
*sk
);
1503 * unlock_sock_fast - complement of lock_sock_fast
1507 * fast unlock socket for user context.
1508 * If slow mode is on, we call regular release_sock()
1510 static inline void unlock_sock_fast(struct sock
*sk
, bool slow
)
1515 spin_unlock_bh(&sk
->sk_lock
.slock
);
1519 extern struct sock
*sk_alloc(struct net
*net
, int family
,
1521 struct proto
*prot
);
1522 extern void sk_free(struct sock
*sk
);
1523 extern void sk_release_kernel(struct sock
*sk
);
1524 extern struct sock
*sk_clone_lock(const struct sock
*sk
,
1525 const gfp_t priority
);
1527 extern struct sk_buff
*sock_wmalloc(struct sock
*sk
,
1528 unsigned long size
, int force
,
1530 extern struct sk_buff
*sock_rmalloc(struct sock
*sk
,
1531 unsigned long size
, int force
,
1533 extern void sock_wfree(struct sk_buff
*skb
);
1534 extern void skb_orphan_partial(struct sk_buff
*skb
);
1535 extern void sock_rfree(struct sk_buff
*skb
);
1536 extern void sock_edemux(struct sk_buff
*skb
);
1538 extern int sock_setsockopt(struct socket
*sock
, int level
,
1539 int op
, char __user
*optval
,
1540 unsigned int optlen
);
1542 extern int sock_getsockopt(struct socket
*sock
, int level
,
1543 int op
, char __user
*optval
,
1544 int __user
*optlen
);
1545 extern struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
,
1549 extern struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
,
1550 unsigned long header_len
,
1551 unsigned long data_len
,
1554 int max_page_order
);
1555 extern void *sock_kmalloc(struct sock
*sk
, int size
,
1557 extern void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
1558 extern void sk_send_sigurg(struct sock
*sk
);
1561 * Functions to fill in entries in struct proto_ops when a protocol
1562 * does not implement a particular function.
1564 extern int sock_no_bind(struct socket
*,
1565 struct sockaddr
*, int);
1566 extern int sock_no_connect(struct socket
*,
1567 struct sockaddr
*, int, int);
1568 extern int sock_no_socketpair(struct socket
*,
1570 extern int sock_no_accept(struct socket
*,
1571 struct socket
*, int);
1572 extern int sock_no_getname(struct socket
*,
1573 struct sockaddr
*, int *, int);
1574 extern unsigned int sock_no_poll(struct file
*, struct socket
*,
1575 struct poll_table_struct
*);
1576 extern int sock_no_ioctl(struct socket
*, unsigned int,
1578 extern int sock_no_listen(struct socket
*, int);
1579 extern int sock_no_shutdown(struct socket
*, int);
1580 extern int sock_no_getsockopt(struct socket
*, int , int,
1581 char __user
*, int __user
*);
1582 extern int sock_no_setsockopt(struct socket
*, int, int,
1583 char __user
*, unsigned int);
1584 extern int sock_no_sendmsg(struct kiocb
*, struct socket
*,
1585 struct msghdr
*, size_t);
1586 extern int sock_no_recvmsg(struct kiocb
*, struct socket
*,
1587 struct msghdr
*, size_t, int);
1588 extern int sock_no_mmap(struct file
*file
,
1589 struct socket
*sock
,
1590 struct vm_area_struct
*vma
);
1591 extern ssize_t
sock_no_sendpage(struct socket
*sock
,
1593 int offset
, size_t size
,
1597 * Functions to fill in entries in struct proto_ops when a protocol
1598 * uses the inet style.
1600 extern int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1601 char __user
*optval
, int __user
*optlen
);
1602 extern int sock_common_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
1603 struct msghdr
*msg
, size_t size
, int flags
);
1604 extern int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1605 char __user
*optval
, unsigned int optlen
);
1606 extern int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
1607 int optname
, char __user
*optval
, int __user
*optlen
);
1608 extern int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
1609 int optname
, char __user
*optval
, unsigned int optlen
);
1611 extern void sk_common_release(struct sock
*sk
);
1614 * Default socket callbacks and setup code
1617 /* Initialise core socket variables */
1618 extern void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1620 extern void sk_filter_release_rcu(struct rcu_head
*rcu
);
1623 * sk_filter_release - release a socket filter
1624 * @fp: filter to remove
1626 * Remove a filter from a socket and release its resources.
1629 static inline void sk_filter_release(struct sk_filter
*fp
)
1631 if (atomic_dec_and_test(&fp
->refcnt
))
1632 call_rcu(&fp
->rcu
, sk_filter_release_rcu
);
1635 static inline void sk_filter_uncharge(struct sock
*sk
, struct sk_filter
*fp
)
1637 atomic_sub(sk_filter_size(fp
->len
), &sk
->sk_omem_alloc
);
1638 sk_filter_release(fp
);
1641 static inline void sk_filter_charge(struct sock
*sk
, struct sk_filter
*fp
)
1643 atomic_inc(&fp
->refcnt
);
1644 atomic_add(sk_filter_size(fp
->len
), &sk
->sk_omem_alloc
);
1648 * Socket reference counting postulates.
1650 * * Each user of socket SHOULD hold a reference count.
1651 * * Each access point to socket (an hash table bucket, reference from a list,
1652 * running timer, skb in flight MUST hold a reference count.
1653 * * When reference count hits 0, it means it will never increase back.
1654 * * When reference count hits 0, it means that no references from
1655 * outside exist to this socket and current process on current CPU
1656 * is last user and may/should destroy this socket.
1657 * * sk_free is called from any context: process, BH, IRQ. When
1658 * it is called, socket has no references from outside -> sk_free
1659 * may release descendant resources allocated by the socket, but
1660 * to the time when it is called, socket is NOT referenced by any
1661 * hash tables, lists etc.
1662 * * Packets, delivered from outside (from network or from another process)
1663 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1664 * when they sit in queue. Otherwise, packets will leak to hole, when
1665 * socket is looked up by one cpu and unhasing is made by another CPU.
1666 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1667 * (leak to backlog). Packet socket does all the processing inside
1668 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1669 * use separate SMP lock, so that they are prone too.
1672 /* Ungrab socket and destroy it, if it was the last reference. */
1673 static inline void sock_put(struct sock
*sk
)
1675 if (atomic_dec_and_test(&sk
->sk_refcnt
))
1679 extern int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
1682 static inline void sk_tx_queue_set(struct sock
*sk
, int tx_queue
)
1684 sk
->sk_tx_queue_mapping
= tx_queue
;
1687 static inline void sk_tx_queue_clear(struct sock
*sk
)
1689 sk
->sk_tx_queue_mapping
= -1;
1692 static inline int sk_tx_queue_get(const struct sock
*sk
)
1694 return sk
? sk
->sk_tx_queue_mapping
: -1;
1697 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1699 sk_tx_queue_clear(sk
);
1700 sk
->sk_socket
= sock
;
1703 static inline wait_queue_head_t
*sk_sleep(struct sock
*sk
)
1705 BUILD_BUG_ON(offsetof(struct socket_wq
, wait
) != 0);
1706 return &rcu_dereference_raw(sk
->sk_wq
)->wait
;
1708 /* Detach socket from process context.
1709 * Announce socket dead, detach it from wait queue and inode.
1710 * Note that parent inode held reference count on this struct sock,
1711 * we do not release it in this function, because protocol
1712 * probably wants some additional cleanups or even continuing
1713 * to work with this socket (TCP).
1715 static inline void sock_orphan(struct sock
*sk
)
1717 write_lock_bh(&sk
->sk_callback_lock
);
1718 sock_set_flag(sk
, SOCK_DEAD
);
1719 sk_set_socket(sk
, NULL
);
1721 write_unlock_bh(&sk
->sk_callback_lock
);
1724 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1726 write_lock_bh(&sk
->sk_callback_lock
);
1727 sk
->sk_wq
= parent
->wq
;
1729 sk_set_socket(sk
, parent
);
1730 security_sock_graft(sk
, parent
);
1731 write_unlock_bh(&sk
->sk_callback_lock
);
1734 extern kuid_t
sock_i_uid(struct sock
*sk
);
1735 extern unsigned long sock_i_ino(struct sock
*sk
);
1737 static inline struct dst_entry
*
1738 __sk_dst_get(struct sock
*sk
)
1740 return rcu_dereference_check(sk
->sk_dst_cache
, sock_owned_by_user(sk
) ||
1741 lockdep_is_held(&sk
->sk_lock
.slock
));
1744 static inline struct dst_entry
*
1745 sk_dst_get(struct sock
*sk
)
1747 struct dst_entry
*dst
;
1750 dst
= rcu_dereference(sk
->sk_dst_cache
);
1751 if (dst
&& !atomic_inc_not_zero(&dst
->__refcnt
))
1757 extern void sk_reset_txq(struct sock
*sk
);
1759 static inline void dst_negative_advice(struct sock
*sk
)
1761 struct dst_entry
*ndst
, *dst
= __sk_dst_get(sk
);
1763 if (dst
&& dst
->ops
->negative_advice
) {
1764 ndst
= dst
->ops
->negative_advice(dst
);
1767 rcu_assign_pointer(sk
->sk_dst_cache
, ndst
);
1774 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1776 struct dst_entry
*old_dst
;
1778 sk_tx_queue_clear(sk
);
1780 * This can be called while sk is owned by the caller only,
1781 * with no state that can be checked in a rcu_dereference_check() cond
1783 old_dst
= rcu_dereference_raw(sk
->sk_dst_cache
);
1784 rcu_assign_pointer(sk
->sk_dst_cache
, dst
);
1785 dst_release(old_dst
);
1789 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1791 struct dst_entry
*old_dst
;
1793 sk_tx_queue_clear(sk
);
1794 old_dst
= xchg(&sk
->sk_dst_cache
, dst
);
1795 dst_release(old_dst
);
1799 __sk_dst_reset(struct sock
*sk
)
1801 __sk_dst_set(sk
, NULL
);
1805 sk_dst_reset(struct sock
*sk
)
1807 sk_dst_set(sk
, NULL
);
1810 extern struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1812 extern struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1814 static inline bool sk_can_gso(const struct sock
*sk
)
1816 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1819 extern void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1821 static inline void sk_nocaps_add(struct sock
*sk
, netdev_features_t flags
)
1823 sk
->sk_route_nocaps
|= flags
;
1824 sk
->sk_route_caps
&= ~flags
;
1827 static inline int skb_do_copy_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1828 char __user
*from
, char *to
,
1829 int copy
, int offset
)
1831 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1833 __wsum csum
= csum_and_copy_from_user(from
, to
, copy
, 0, &err
);
1836 skb
->csum
= csum_block_add(skb
->csum
, csum
, offset
);
1837 } else if (sk
->sk_route_caps
& NETIF_F_NOCACHE_COPY
) {
1838 if (!access_ok(VERIFY_READ
, from
, copy
) ||
1839 __copy_from_user_nocache(to
, from
, copy
))
1841 } else if (copy_from_user(to
, from
, copy
))
1847 static inline int skb_add_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1848 char __user
*from
, int copy
)
1850 int err
, offset
= skb
->len
;
1852 err
= skb_do_copy_data_nocache(sk
, skb
, from
, skb_put(skb
, copy
),
1855 __skb_trim(skb
, offset
);
1860 static inline int skb_copy_to_page_nocache(struct sock
*sk
, char __user
*from
,
1861 struct sk_buff
*skb
,
1867 err
= skb_do_copy_data_nocache(sk
, skb
, from
, page_address(page
) + off
,
1873 skb
->data_len
+= copy
;
1874 skb
->truesize
+= copy
;
1875 sk
->sk_wmem_queued
+= copy
;
1876 sk_mem_charge(sk
, copy
);
1880 static inline int skb_copy_to_page(struct sock
*sk
, char __user
*from
,
1881 struct sk_buff
*skb
, struct page
*page
,
1884 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1886 __wsum csum
= csum_and_copy_from_user(from
,
1887 page_address(page
) + off
,
1891 skb
->csum
= csum_block_add(skb
->csum
, csum
, skb
->len
);
1892 } else if (copy_from_user(page_address(page
) + off
, from
, copy
))
1896 skb
->data_len
+= copy
;
1897 skb
->truesize
+= copy
;
1898 sk
->sk_wmem_queued
+= copy
;
1899 sk_mem_charge(sk
, copy
);
1904 * sk_wmem_alloc_get - returns write allocations
1907 * Returns sk_wmem_alloc minus initial offset of one
1909 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
1911 return atomic_read(&sk
->sk_wmem_alloc
) - 1;
1915 * sk_rmem_alloc_get - returns read allocations
1918 * Returns sk_rmem_alloc
1920 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
1922 return atomic_read(&sk
->sk_rmem_alloc
);
1926 * sk_has_allocations - check if allocations are outstanding
1929 * Returns true if socket has write or read allocations
1931 static inline bool sk_has_allocations(const struct sock
*sk
)
1933 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
1937 * wq_has_sleeper - check if there are any waiting processes
1938 * @wq: struct socket_wq
1940 * Returns true if socket_wq has waiting processes
1942 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1943 * barrier call. They were added due to the race found within the tcp code.
1945 * Consider following tcp code paths:
1949 * sys_select receive packet
1951 * __add_wait_queue update tp->rcv_nxt
1953 * tp->rcv_nxt check sock_def_readable
1955 * schedule rcu_read_lock();
1956 * wq = rcu_dereference(sk->sk_wq);
1957 * if (wq && waitqueue_active(&wq->wait))
1958 * wake_up_interruptible(&wq->wait)
1962 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1963 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1964 * could then endup calling schedule and sleep forever if there are no more
1965 * data on the socket.
1968 static inline bool wq_has_sleeper(struct socket_wq
*wq
)
1970 /* We need to be sure we are in sync with the
1971 * add_wait_queue modifications to the wait queue.
1973 * This memory barrier is paired in the sock_poll_wait.
1976 return wq
&& waitqueue_active(&wq
->wait
);
1980 * sock_poll_wait - place memory barrier behind the poll_wait call.
1982 * @wait_address: socket wait queue
1985 * See the comments in the wq_has_sleeper function.
1987 static inline void sock_poll_wait(struct file
*filp
,
1988 wait_queue_head_t
*wait_address
, poll_table
*p
)
1990 if (!poll_does_not_wait(p
) && wait_address
) {
1991 poll_wait(filp
, wait_address
, p
);
1992 /* We need to be sure we are in sync with the
1993 * socket flags modification.
1995 * This memory barrier is paired in the wq_has_sleeper.
2002 * Queue a received datagram if it will fit. Stream and sequenced
2003 * protocols can't normally use this as they need to fit buffers in
2004 * and play with them.
2006 * Inlined as it's very short and called for pretty much every
2007 * packet ever received.
2010 static inline void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
2014 skb
->destructor
= sock_wfree
;
2016 * We used to take a refcount on sk, but following operation
2017 * is enough to guarantee sk_free() wont free this sock until
2018 * all in-flight packets are completed
2020 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
2023 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
2027 skb
->destructor
= sock_rfree
;
2028 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
2029 sk_mem_charge(sk
, skb
->truesize
);
2032 extern void sk_reset_timer(struct sock
*sk
, struct timer_list
*timer
,
2033 unsigned long expires
);
2035 extern void sk_stop_timer(struct sock
*sk
, struct timer_list
*timer
);
2037 extern int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
2039 extern int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
);
2042 * Recover an error report and clear atomically
2045 static inline int sock_error(struct sock
*sk
)
2048 if (likely(!sk
->sk_err
))
2050 err
= xchg(&sk
->sk_err
, 0);
2054 static inline unsigned long sock_wspace(struct sock
*sk
)
2058 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
2059 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
2066 static inline void sk_wake_async(struct sock
*sk
, int how
, int band
)
2068 if (sock_flag(sk
, SOCK_FASYNC
))
2069 sock_wake_async(sk
->sk_socket
, how
, band
);
2072 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2073 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2074 * Note: for send buffers, TCP works better if we can build two skbs at
2077 #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2079 #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2080 #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
2082 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
2084 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
2085 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
2086 sk
->sk_sndbuf
= max_t(u32
, sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
2090 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
);
2093 * sk_page_frag - return an appropriate page_frag
2096 * If socket allocation mode allows current thread to sleep, it means its
2097 * safe to use the per task page_frag instead of the per socket one.
2099 static inline struct page_frag
*sk_page_frag(struct sock
*sk
)
2101 if (sk
->sk_allocation
& __GFP_WAIT
)
2102 return ¤t
->task_frag
;
2104 return &sk
->sk_frag
;
2107 extern bool sk_page_frag_refill(struct sock
*sk
, struct page_frag
*pfrag
);
2110 * Default write policy as shown to user space via poll/select/SIGIO
2112 static inline bool sock_writeable(const struct sock
*sk
)
2114 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
2117 static inline gfp_t
gfp_any(void)
2119 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
2122 static inline long sock_rcvtimeo(const struct sock
*sk
, bool noblock
)
2124 return noblock
? 0 : sk
->sk_rcvtimeo
;
2127 static inline long sock_sndtimeo(const struct sock
*sk
, bool noblock
)
2129 return noblock
? 0 : sk
->sk_sndtimeo
;
2132 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
2134 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
2137 /* Alas, with timeout socket operations are not restartable.
2138 * Compare this to poll().
2140 static inline int sock_intr_errno(long timeo
)
2142 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
2145 extern void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
2146 struct sk_buff
*skb
);
2147 extern void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
2148 struct sk_buff
*skb
);
2151 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
2153 ktime_t kt
= skb
->tstamp
;
2154 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
2157 * generate control messages if
2158 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2159 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2160 * - software time stamp available and wanted
2161 * (SOCK_TIMESTAMPING_SOFTWARE)
2162 * - hardware time stamps available and wanted
2163 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2164 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2166 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
2167 sock_flag(sk
, SOCK_TIMESTAMPING_RX_SOFTWARE
) ||
2168 (kt
.tv64
&& sock_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
)) ||
2169 (hwtstamps
->hwtstamp
.tv64
&&
2170 sock_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
)) ||
2171 (hwtstamps
->syststamp
.tv64
&&
2172 sock_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
)))
2173 __sock_recv_timestamp(msg
, sk
, skb
);
2177 if (sock_flag(sk
, SOCK_WIFI_STATUS
) && skb
->wifi_acked_valid
)
2178 __sock_recv_wifi_status(msg
, sk
, skb
);
2181 extern void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2182 struct sk_buff
*skb
);
2184 static inline void sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2185 struct sk_buff
*skb
)
2187 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2188 (1UL << SOCK_RCVTSTAMP) | \
2189 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2190 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
2191 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
2192 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2194 if (sk
->sk_flags
& FLAGS_TS_OR_DROPS
)
2195 __sock_recv_ts_and_drops(msg
, sk
, skb
);
2197 sk
->sk_stamp
= skb
->tstamp
;
2201 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2202 * @sk: socket sending this packet
2203 * @tx_flags: filled with instructions for time stamping
2205 * Currently only depends on SOCK_TIMESTAMPING* flags.
2207 extern void sock_tx_timestamp(struct sock
*sk
, __u8
*tx_flags
);
2210 * sk_eat_skb - Release a skb if it is no longer needed
2211 * @sk: socket to eat this skb from
2212 * @skb: socket buffer to eat
2213 * @copied_early: flag indicating whether DMA operations copied this data early
2215 * This routine must be called with interrupts disabled or with the socket
2216 * locked so that the sk_buff queue operation is ok.
2218 #ifdef CONFIG_NET_DMA
2219 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, bool copied_early
)
2221 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2225 __skb_queue_tail(&sk
->sk_async_wait_queue
, skb
);
2228 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, bool copied_early
)
2230 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2236 struct net
*sock_net(const struct sock
*sk
)
2238 return read_pnet(&sk
->sk_net
);
2242 void sock_net_set(struct sock
*sk
, struct net
*net
)
2244 write_pnet(&sk
->sk_net
, net
);
2248 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
2249 * They should not hold a reference to a namespace in order to allow
2251 * Sockets after sk_change_net should be released using sk_release_kernel
2253 static inline void sk_change_net(struct sock
*sk
, struct net
*net
)
2255 put_net(sock_net(sk
));
2256 sock_net_set(sk
, hold_net(net
));
2259 static inline struct sock
*skb_steal_sock(struct sk_buff
*skb
)
2262 struct sock
*sk
= skb
->sk
;
2264 skb
->destructor
= NULL
;
2271 extern void sock_enable_timestamp(struct sock
*sk
, int flag
);
2272 extern int sock_get_timestamp(struct sock
*, struct timeval __user
*);
2273 extern int sock_get_timestampns(struct sock
*, struct timespec __user
*);
2274 extern int sock_recv_errqueue(struct sock
*sk
, struct msghdr
*msg
, int len
,
2275 int level
, int type
);
2277 bool sk_ns_capable(const struct sock
*sk
,
2278 struct user_namespace
*user_ns
, int cap
);
2279 bool sk_capable(const struct sock
*sk
, int cap
);
2280 bool sk_net_capable(const struct sock
*sk
, int cap
);
2283 * Enable debug/info messages
2285 extern int net_msg_warn
;
2286 #define NETDEBUG(fmt, args...) \
2287 do { if (net_msg_warn) printk(fmt,##args); } while (0)
2289 #define LIMIT_NETDEBUG(fmt, args...) \
2290 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
2292 extern __u32 sysctl_wmem_max
;
2293 extern __u32 sysctl_rmem_max
;
2295 extern int sysctl_optmem_max
;
2297 extern __u32 sysctl_wmem_default
;
2298 extern __u32 sysctl_rmem_default
;
2300 #endif /* _SOCK_H */