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>
70 #include <linux/net_tstamp.h>
75 int mem_cgroup_sockets_init(struct mem_cgroup
*memcg
, struct cgroup_subsys
*ss
);
76 void mem_cgroup_sockets_destroy(struct mem_cgroup
*memcg
);
79 int mem_cgroup_sockets_init(struct mem_cgroup
*memcg
, struct cgroup_subsys
*ss
)
84 void mem_cgroup_sockets_destroy(struct mem_cgroup
*memcg
)
89 * This structure really needs to be cleaned up.
90 * Most of it is for TCP, and not used by any of
91 * the other protocols.
94 /* Define this to get the SOCK_DBG debugging facility. */
95 #define SOCK_DEBUGGING
97 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
98 printk(KERN_DEBUG msg); } while (0)
100 /* Validate arguments and do nothing */
101 static inline __printf(2, 3)
102 void SOCK_DEBUG(const struct sock
*sk
, const char *msg
, ...)
107 /* This is the per-socket lock. The spinlock provides a synchronization
108 * between user contexts and software interrupt processing, whereas the
109 * mini-semaphore synchronizes multiple users amongst themselves.
114 wait_queue_head_t wq
;
116 * We express the mutex-alike socket_lock semantics
117 * to the lock validator by explicitly managing
118 * the slock as a lock variant (in addition to
121 #ifdef CONFIG_DEBUG_LOCK_ALLOC
122 struct lockdep_map dep_map
;
130 typedef __u32 __bitwise __portpair
;
131 typedef __u64 __bitwise __addrpair
;
134 * struct sock_common - minimal network layer representation of sockets
135 * @skc_daddr: Foreign IPv4 addr
136 * @skc_rcv_saddr: Bound local IPv4 addr
137 * @skc_hash: hash value used with various protocol lookup tables
138 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
139 * @skc_dport: placeholder for inet_dport/tw_dport
140 * @skc_num: placeholder for inet_num/tw_num
141 * @skc_family: network address family
142 * @skc_state: Connection state
143 * @skc_reuse: %SO_REUSEADDR setting
144 * @skc_reuseport: %SO_REUSEPORT setting
145 * @skc_bound_dev_if: bound device index if != 0
146 * @skc_bind_node: bind hash linkage for various protocol lookup tables
147 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
148 * @skc_prot: protocol handlers inside a network family
149 * @skc_net: reference to the network namespace of this socket
150 * @skc_node: main hash linkage for various protocol lookup tables
151 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
152 * @skc_tx_queue_mapping: tx queue number for this connection
153 * @skc_refcnt: reference count
155 * This is the minimal network layer representation of sockets, the header
156 * for struct sock and struct inet_timewait_sock.
159 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
160 * address on 64bit arches : cf INET_MATCH()
163 __addrpair skc_addrpair
;
166 __be32 skc_rcv_saddr
;
170 unsigned int skc_hash
;
171 __u16 skc_u16hashes
[2];
173 /* skc_dport && skc_num must be grouped as well */
175 __portpair skc_portpair
;
182 unsigned short skc_family
;
183 volatile unsigned char skc_state
;
184 unsigned char skc_reuse
:4;
185 unsigned char skc_reuseport
:1;
186 unsigned char skc_ipv6only
:1;
187 int skc_bound_dev_if
;
189 struct hlist_node skc_bind_node
;
190 struct hlist_nulls_node skc_portaddr_node
;
192 struct proto
*skc_prot
;
197 #if IS_ENABLED(CONFIG_IPV6)
198 struct in6_addr skc_v6_daddr
;
199 struct in6_addr skc_v6_rcv_saddr
;
203 * fields between dontcopy_begin/dontcopy_end
204 * are not copied in sock_copy()
207 int skc_dontcopy_begin
[0];
210 struct hlist_node skc_node
;
211 struct hlist_nulls_node skc_nulls_node
;
213 int skc_tx_queue_mapping
;
216 int skc_dontcopy_end
[0];
222 * struct sock - network layer representation of sockets
223 * @__sk_common: shared layout with inet_timewait_sock
224 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
225 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
226 * @sk_lock: synchronizer
227 * @sk_rcvbuf: size of receive buffer in bytes
228 * @sk_wq: sock wait queue and async head
229 * @sk_rx_dst: receive input route used by early demux
230 * @sk_dst_cache: destination cache
231 * @sk_dst_lock: destination cache lock
232 * @sk_policy: flow policy
233 * @sk_receive_queue: incoming packets
234 * @sk_wmem_alloc: transmit queue bytes committed
235 * @sk_write_queue: Packet sending queue
236 * @sk_async_wait_queue: DMA copied packets
237 * @sk_omem_alloc: "o" is "option" or "other"
238 * @sk_wmem_queued: persistent queue size
239 * @sk_forward_alloc: space allocated forward
240 * @sk_napi_id: id of the last napi context to receive data for sk
241 * @sk_ll_usec: usecs to busypoll when there is no data
242 * @sk_allocation: allocation mode
243 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
244 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
245 * @sk_sndbuf: size of send buffer in bytes
246 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
247 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
248 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
249 * @sk_no_check_rx: allow zero checksum in RX packets
250 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
251 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
252 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
253 * @sk_gso_max_size: Maximum GSO segment size to build
254 * @sk_gso_max_segs: Maximum number of GSO segments
255 * @sk_lingertime: %SO_LINGER l_linger setting
256 * @sk_backlog: always used with the per-socket spinlock held
257 * @sk_callback_lock: used with the callbacks in the end of this struct
258 * @sk_error_queue: rarely used
259 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
260 * IPV6_ADDRFORM for instance)
261 * @sk_err: last error
262 * @sk_err_soft: errors that don't cause failure but are the cause of a
263 * persistent failure not just 'timed out'
264 * @sk_drops: raw/udp drops counter
265 * @sk_ack_backlog: current listen backlog
266 * @sk_max_ack_backlog: listen backlog set in listen()
267 * @sk_priority: %SO_PRIORITY setting
268 * @sk_cgrp_prioidx: socket group's priority map index
269 * @sk_type: socket type (%SOCK_STREAM, etc)
270 * @sk_protocol: which protocol this socket belongs in this network family
271 * @sk_peer_pid: &struct pid for this socket's peer
272 * @sk_peer_cred: %SO_PEERCRED setting
273 * @sk_rcvlowat: %SO_RCVLOWAT setting
274 * @sk_rcvtimeo: %SO_RCVTIMEO setting
275 * @sk_sndtimeo: %SO_SNDTIMEO setting
276 * @sk_rxhash: flow hash received from netif layer
277 * @sk_txhash: computed flow hash for use on transmit
278 * @sk_filter: socket filtering instructions
279 * @sk_protinfo: private area, net family specific, when not using slab
280 * @sk_timer: sock cleanup timer
281 * @sk_stamp: time stamp of last packet received
282 * @sk_tsflags: SO_TIMESTAMPING socket options
283 * @sk_tskey: counter to disambiguate concurrent tstamp requests
284 * @sk_socket: Identd and reporting IO signals
285 * @sk_user_data: RPC layer private data
286 * @sk_frag: cached page frag
287 * @sk_peek_off: current peek_offset value
288 * @sk_send_head: front of stuff to transmit
289 * @sk_security: used by security modules
290 * @sk_mark: generic packet mark
291 * @sk_classid: this socket's cgroup classid
292 * @sk_cgrp: this socket's cgroup-specific proto data
293 * @sk_write_pending: a write to stream socket waits to start
294 * @sk_state_change: callback to indicate change in the state of the sock
295 * @sk_data_ready: callback to indicate there is data to be processed
296 * @sk_write_space: callback to indicate there is bf sending space available
297 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
298 * @sk_backlog_rcv: callback to process the backlog
299 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
303 * Now struct inet_timewait_sock also uses sock_common, so please just
304 * don't add nothing before this first member (__sk_common) --acme
306 struct sock_common __sk_common
;
307 #define sk_node __sk_common.skc_node
308 #define sk_nulls_node __sk_common.skc_nulls_node
309 #define sk_refcnt __sk_common.skc_refcnt
310 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
312 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
313 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
314 #define sk_hash __sk_common.skc_hash
315 #define sk_portpair __sk_common.skc_portpair
316 #define sk_num __sk_common.skc_num
317 #define sk_dport __sk_common.skc_dport
318 #define sk_addrpair __sk_common.skc_addrpair
319 #define sk_daddr __sk_common.skc_daddr
320 #define sk_rcv_saddr __sk_common.skc_rcv_saddr
321 #define sk_family __sk_common.skc_family
322 #define sk_state __sk_common.skc_state
323 #define sk_reuse __sk_common.skc_reuse
324 #define sk_reuseport __sk_common.skc_reuseport
325 #define sk_ipv6only __sk_common.skc_ipv6only
326 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
327 #define sk_bind_node __sk_common.skc_bind_node
328 #define sk_prot __sk_common.skc_prot
329 #define sk_net __sk_common.skc_net
330 #define sk_v6_daddr __sk_common.skc_v6_daddr
331 #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
333 socket_lock_t sk_lock
;
334 struct sk_buff_head sk_receive_queue
;
336 * The backlog queue is special, it is always used with
337 * the per-socket spinlock held and requires low latency
338 * access. Therefore we special case it's implementation.
339 * Note : rmem_alloc is in this structure to fill a hole
340 * on 64bit arches, not because its logically part of
346 struct sk_buff
*head
;
347 struct sk_buff
*tail
;
349 #define sk_rmem_alloc sk_backlog.rmem_alloc
350 int sk_forward_alloc
;
355 #ifdef CONFIG_NET_RX_BUSY_POLL
356 unsigned int sk_napi_id
;
357 unsigned int sk_ll_usec
;
362 struct sk_filter __rcu
*sk_filter
;
363 struct socket_wq __rcu
*sk_wq
;
365 #ifdef CONFIG_NET_DMA
366 struct sk_buff_head sk_async_wait_queue
;
370 struct xfrm_policy
*sk_policy
[2];
372 unsigned long sk_flags
;
373 struct dst_entry
*sk_rx_dst
;
374 struct dst_entry __rcu
*sk_dst_cache
;
375 spinlock_t sk_dst_lock
;
376 atomic_t sk_wmem_alloc
;
377 atomic_t sk_omem_alloc
;
379 struct sk_buff_head sk_write_queue
;
380 kmemcheck_bitfield_begin(flags
);
381 unsigned int sk_shutdown
: 2,
387 kmemcheck_bitfield_end(flags
);
390 u32 sk_pacing_rate
; /* bytes per second */
391 u32 sk_max_pacing_rate
;
392 netdev_features_t sk_route_caps
;
393 netdev_features_t sk_route_nocaps
;
395 unsigned int sk_gso_max_size
;
398 unsigned long sk_lingertime
;
399 struct sk_buff_head sk_error_queue
;
400 struct proto
*sk_prot_creator
;
401 rwlock_t sk_callback_lock
;
404 unsigned short sk_ack_backlog
;
405 unsigned short sk_max_ack_backlog
;
407 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
408 __u32 sk_cgrp_prioidx
;
410 struct pid
*sk_peer_pid
;
411 const struct cred
*sk_peer_cred
;
415 struct timer_list sk_timer
;
419 struct socket
*sk_socket
;
421 struct page_frag sk_frag
;
422 struct sk_buff
*sk_send_head
;
424 int sk_write_pending
;
425 #ifdef CONFIG_SECURITY
430 struct cg_proto
*sk_cgrp
;
431 void (*sk_state_change
)(struct sock
*sk
);
432 void (*sk_data_ready
)(struct sock
*sk
);
433 void (*sk_write_space
)(struct sock
*sk
);
434 void (*sk_error_report
)(struct sock
*sk
);
435 int (*sk_backlog_rcv
)(struct sock
*sk
,
436 struct sk_buff
*skb
);
437 void (*sk_destruct
)(struct sock
*sk
);
440 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
442 #define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
443 #define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
446 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
447 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
448 * on a socket means that the socket will reuse everybody else's port
449 * without looking at the other's sk_reuse value.
452 #define SK_NO_REUSE 0
453 #define SK_CAN_REUSE 1
454 #define SK_FORCE_REUSE 2
456 static inline int sk_peek_offset(struct sock
*sk
, int flags
)
458 if ((flags
& MSG_PEEK
) && (sk
->sk_peek_off
>= 0))
459 return sk
->sk_peek_off
;
464 static inline void sk_peek_offset_bwd(struct sock
*sk
, int val
)
466 if (sk
->sk_peek_off
>= 0) {
467 if (sk
->sk_peek_off
>= val
)
468 sk
->sk_peek_off
-= val
;
474 static inline void sk_peek_offset_fwd(struct sock
*sk
, int val
)
476 if (sk
->sk_peek_off
>= 0)
477 sk
->sk_peek_off
+= val
;
481 * Hashed lists helper routines
483 static inline struct sock
*sk_entry(const struct hlist_node
*node
)
485 return hlist_entry(node
, struct sock
, sk_node
);
488 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
490 return hlist_entry(head
->first
, struct sock
, sk_node
);
493 static inline struct sock
*sk_head(const struct hlist_head
*head
)
495 return hlist_empty(head
) ? NULL
: __sk_head(head
);
498 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
500 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
503 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
505 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
508 static inline struct sock
*sk_next(const struct sock
*sk
)
510 return sk
->sk_node
.next
?
511 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
514 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
516 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
517 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
518 struct sock
, sk_nulls_node
) :
522 static inline bool sk_unhashed(const struct sock
*sk
)
524 return hlist_unhashed(&sk
->sk_node
);
527 static inline bool sk_hashed(const struct sock
*sk
)
529 return !sk_unhashed(sk
);
532 static inline void sk_node_init(struct hlist_node
*node
)
537 static inline void sk_nulls_node_init(struct hlist_nulls_node
*node
)
542 static inline void __sk_del_node(struct sock
*sk
)
544 __hlist_del(&sk
->sk_node
);
547 /* NB: equivalent to hlist_del_init_rcu */
548 static inline bool __sk_del_node_init(struct sock
*sk
)
552 sk_node_init(&sk
->sk_node
);
558 /* Grab socket reference count. This operation is valid only
559 when sk is ALREADY grabbed f.e. it is found in hash table
560 or a list and the lookup is made under lock preventing hash table
564 static inline void sock_hold(struct sock
*sk
)
566 atomic_inc(&sk
->sk_refcnt
);
569 /* Ungrab socket in the context, which assumes that socket refcnt
570 cannot hit zero, f.e. it is true in context of any socketcall.
572 static inline void __sock_put(struct sock
*sk
)
574 atomic_dec(&sk
->sk_refcnt
);
577 static inline bool sk_del_node_init(struct sock
*sk
)
579 bool rc
= __sk_del_node_init(sk
);
582 /* paranoid for a while -acme */
583 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
588 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
590 static inline bool __sk_nulls_del_node_init_rcu(struct sock
*sk
)
593 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
599 static inline bool sk_nulls_del_node_init_rcu(struct sock
*sk
)
601 bool rc
= __sk_nulls_del_node_init_rcu(sk
);
604 /* paranoid for a while -acme */
605 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
611 static inline void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
613 hlist_add_head(&sk
->sk_node
, list
);
616 static inline void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
619 __sk_add_node(sk
, list
);
622 static inline void sk_add_node_rcu(struct sock
*sk
, struct hlist_head
*list
)
625 hlist_add_head_rcu(&sk
->sk_node
, list
);
628 static inline void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
630 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
633 static inline void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
636 __sk_nulls_add_node_rcu(sk
, list
);
639 static inline void __sk_del_bind_node(struct sock
*sk
)
641 __hlist_del(&sk
->sk_bind_node
);
644 static inline void sk_add_bind_node(struct sock
*sk
,
645 struct hlist_head
*list
)
647 hlist_add_head(&sk
->sk_bind_node
, list
);
650 #define sk_for_each(__sk, list) \
651 hlist_for_each_entry(__sk, list, sk_node)
652 #define sk_for_each_rcu(__sk, list) \
653 hlist_for_each_entry_rcu(__sk, list, sk_node)
654 #define sk_nulls_for_each(__sk, node, list) \
655 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
656 #define sk_nulls_for_each_rcu(__sk, node, list) \
657 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
658 #define sk_for_each_from(__sk) \
659 hlist_for_each_entry_from(__sk, sk_node)
660 #define sk_nulls_for_each_from(__sk, node) \
661 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
662 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
663 #define sk_for_each_safe(__sk, tmp, list) \
664 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
665 #define sk_for_each_bound(__sk, list) \
666 hlist_for_each_entry(__sk, list, sk_bind_node)
669 * sk_nulls_for_each_entry_offset - iterate over a list at a given struct offset
670 * @tpos: the type * to use as a loop cursor.
671 * @pos: the &struct hlist_node to use as a loop cursor.
672 * @head: the head for your list.
673 * @offset: offset of hlist_node within the struct.
676 #define sk_nulls_for_each_entry_offset(tpos, pos, head, offset) \
677 for (pos = (head)->first; \
678 (!is_a_nulls(pos)) && \
679 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
682 static inline struct user_namespace
*sk_user_ns(struct sock
*sk
)
684 /* Careful only use this in a context where these parameters
685 * can not change and must all be valid, such as recvmsg from
688 return sk
->sk_socket
->file
->f_cred
->user_ns
;
702 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
703 SOCK_DBG
, /* %SO_DEBUG setting */
704 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
705 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
706 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
707 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
708 SOCK_MEMALLOC
, /* VM depends on this socket for swapping */
709 SOCK_TIMESTAMPING_RX_SOFTWARE
, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
710 SOCK_FASYNC
, /* fasync() active */
712 SOCK_ZEROCOPY
, /* buffers from userspace */
713 SOCK_WIFI_STATUS
, /* push wifi status to userspace */
714 SOCK_NOFCS
, /* Tell NIC not to do the Ethernet FCS.
715 * Will use last 4 bytes of packet sent from
716 * user-space instead.
718 SOCK_FILTER_LOCKED
, /* Filter cannot be changed anymore */
719 SOCK_SELECT_ERR_QUEUE
, /* Wake select on error queue */
722 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
724 nsk
->sk_flags
= osk
->sk_flags
;
727 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
729 __set_bit(flag
, &sk
->sk_flags
);
732 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
734 __clear_bit(flag
, &sk
->sk_flags
);
737 static inline bool sock_flag(const struct sock
*sk
, enum sock_flags flag
)
739 return test_bit(flag
, &sk
->sk_flags
);
743 extern struct static_key memalloc_socks
;
744 static inline int sk_memalloc_socks(void)
746 return static_key_false(&memalloc_socks
);
750 static inline int sk_memalloc_socks(void)
757 static inline gfp_t
sk_gfp_atomic(struct sock
*sk
, gfp_t gfp_mask
)
759 return GFP_ATOMIC
| (sk
->sk_allocation
& __GFP_MEMALLOC
);
762 static inline void sk_acceptq_removed(struct sock
*sk
)
764 sk
->sk_ack_backlog
--;
767 static inline void sk_acceptq_added(struct sock
*sk
)
769 sk
->sk_ack_backlog
++;
772 static inline bool sk_acceptq_is_full(const struct sock
*sk
)
774 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
778 * Compute minimal free write space needed to queue new packets.
780 static inline int sk_stream_min_wspace(const struct sock
*sk
)
782 return sk
->sk_wmem_queued
>> 1;
785 static inline int sk_stream_wspace(const struct sock
*sk
)
787 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
790 void sk_stream_write_space(struct sock
*sk
);
792 /* OOB backlog add */
793 static inline void __sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
795 /* dont let skb dst not refcounted, we are going to leave rcu lock */
798 if (!sk
->sk_backlog
.tail
)
799 sk
->sk_backlog
.head
= skb
;
801 sk
->sk_backlog
.tail
->next
= skb
;
803 sk
->sk_backlog
.tail
= skb
;
808 * Take into account size of receive queue and backlog queue
809 * Do not take into account this skb truesize,
810 * to allow even a single big packet to come.
812 static inline bool sk_rcvqueues_full(const struct sock
*sk
, unsigned int limit
)
814 unsigned int qsize
= sk
->sk_backlog
.len
+ atomic_read(&sk
->sk_rmem_alloc
);
816 return qsize
> limit
;
819 /* The per-socket spinlock must be held here. */
820 static inline __must_check
int sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
,
823 if (sk_rcvqueues_full(sk
, limit
))
826 __sk_add_backlog(sk
, skb
);
827 sk
->sk_backlog
.len
+= skb
->truesize
;
831 int __sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
833 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
835 if (sk_memalloc_socks() && skb_pfmemalloc(skb
))
836 return __sk_backlog_rcv(sk
, skb
);
838 return sk
->sk_backlog_rcv(sk
, skb
);
841 static inline void sock_rps_record_flow_hash(__u32 hash
)
844 struct rps_sock_flow_table
*sock_flow_table
;
847 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
848 rps_record_sock_flow(sock_flow_table
, hash
);
853 static inline void sock_rps_reset_flow_hash(__u32 hash
)
856 struct rps_sock_flow_table
*sock_flow_table
;
859 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
860 rps_reset_sock_flow(sock_flow_table
, hash
);
865 static inline void sock_rps_record_flow(const struct sock
*sk
)
868 sock_rps_record_flow_hash(sk
->sk_rxhash
);
872 static inline void sock_rps_reset_flow(const struct sock
*sk
)
875 sock_rps_reset_flow_hash(sk
->sk_rxhash
);
879 static inline void sock_rps_save_rxhash(struct sock
*sk
,
880 const struct sk_buff
*skb
)
883 if (unlikely(sk
->sk_rxhash
!= skb
->hash
)) {
884 sock_rps_reset_flow(sk
);
885 sk
->sk_rxhash
= skb
->hash
;
890 static inline void sock_rps_reset_rxhash(struct sock
*sk
)
893 sock_rps_reset_flow(sk
);
898 #define sk_wait_event(__sk, __timeo, __condition) \
900 release_sock(__sk); \
901 __rc = __condition; \
903 *(__timeo) = schedule_timeout(*(__timeo)); \
906 __rc = __condition; \
910 int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
911 int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
912 void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
913 int sk_stream_error(struct sock
*sk
, int flags
, int err
);
914 void sk_stream_kill_queues(struct sock
*sk
);
915 void sk_set_memalloc(struct sock
*sk
);
916 void sk_clear_memalloc(struct sock
*sk
);
918 int sk_wait_data(struct sock
*sk
, long *timeo
);
920 struct request_sock_ops
;
921 struct timewait_sock_ops
;
922 struct inet_hashinfo
;
927 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
928 * un-modified. Special care is taken when initializing object to zero.
930 static inline void sk_prot_clear_nulls(struct sock
*sk
, int size
)
932 if (offsetof(struct sock
, sk_node
.next
) != 0)
933 memset(sk
, 0, offsetof(struct sock
, sk_node
.next
));
934 memset(&sk
->sk_node
.pprev
, 0,
935 size
- offsetof(struct sock
, sk_node
.pprev
));
938 /* Networking protocol blocks we attach to sockets.
939 * socket layer -> transport layer interface
940 * transport -> network interface is defined by struct inet_proto
943 void (*close
)(struct sock
*sk
,
945 int (*connect
)(struct sock
*sk
,
946 struct sockaddr
*uaddr
,
948 int (*disconnect
)(struct sock
*sk
, int flags
);
950 struct sock
* (*accept
)(struct sock
*sk
, int flags
, int *err
);
952 int (*ioctl
)(struct sock
*sk
, int cmd
,
954 int (*init
)(struct sock
*sk
);
955 void (*destroy
)(struct sock
*sk
);
956 void (*shutdown
)(struct sock
*sk
, int how
);
957 int (*setsockopt
)(struct sock
*sk
, int level
,
958 int optname
, char __user
*optval
,
959 unsigned int optlen
);
960 int (*getsockopt
)(struct sock
*sk
, int level
,
961 int optname
, char __user
*optval
,
964 int (*compat_setsockopt
)(struct sock
*sk
,
966 int optname
, char __user
*optval
,
967 unsigned int optlen
);
968 int (*compat_getsockopt
)(struct sock
*sk
,
970 int optname
, char __user
*optval
,
972 int (*compat_ioctl
)(struct sock
*sk
,
973 unsigned int cmd
, unsigned long arg
);
975 int (*sendmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
976 struct msghdr
*msg
, size_t len
);
977 int (*recvmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
979 size_t len
, int noblock
, int flags
,
981 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
982 int offset
, size_t size
, int flags
);
983 int (*bind
)(struct sock
*sk
,
984 struct sockaddr
*uaddr
, int addr_len
);
986 int (*backlog_rcv
) (struct sock
*sk
,
987 struct sk_buff
*skb
);
989 void (*release_cb
)(struct sock
*sk
);
990 void (*mtu_reduced
)(struct sock
*sk
);
992 /* Keeping track of sk's, looking them up, and port selection methods. */
993 void (*hash
)(struct sock
*sk
);
994 void (*unhash
)(struct sock
*sk
);
995 void (*rehash
)(struct sock
*sk
);
996 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
997 void (*clear_sk
)(struct sock
*sk
, int size
);
999 /* Keeping track of sockets in use */
1000 #ifdef CONFIG_PROC_FS
1001 unsigned int inuse_idx
;
1004 bool (*stream_memory_free
)(const struct sock
*sk
);
1005 /* Memory pressure */
1006 void (*enter_memory_pressure
)(struct sock
*sk
);
1007 atomic_long_t
*memory_allocated
; /* Current allocated memory. */
1008 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
1010 * Pressure flag: try to collapse.
1011 * Technical note: it is used by multiple contexts non atomically.
1012 * All the __sk_mem_schedule() is of this nature: accounting
1013 * is strict, actions are advisory and have some latency.
1015 int *memory_pressure
;
1022 struct kmem_cache
*slab
;
1023 unsigned int obj_size
;
1026 struct percpu_counter
*orphan_count
;
1028 struct request_sock_ops
*rsk_prot
;
1029 struct timewait_sock_ops
*twsk_prot
;
1032 struct inet_hashinfo
*hashinfo
;
1033 struct udp_table
*udp_table
;
1034 struct raw_hashinfo
*raw_hash
;
1037 struct module
*owner
;
1041 struct list_head node
;
1042 #ifdef SOCK_REFCNT_DEBUG
1045 #ifdef CONFIG_MEMCG_KMEM
1047 * cgroup specific init/deinit functions. Called once for all
1048 * protocols that implement it, from cgroups populate function.
1049 * This function has to setup any files the protocol want to
1050 * appear in the kmem cgroup filesystem.
1052 int (*init_cgroup
)(struct mem_cgroup
*memcg
,
1053 struct cgroup_subsys
*ss
);
1054 void (*destroy_cgroup
)(struct mem_cgroup
*memcg
);
1055 struct cg_proto
*(*proto_cgroup
)(struct mem_cgroup
*memcg
);
1060 * Bits in struct cg_proto.flags
1062 enum cg_proto_flags
{
1063 /* Currently active and new sockets should be assigned to cgroups */
1065 /* It was ever activated; we must disarm static keys on destruction */
1066 MEMCG_SOCK_ACTIVATED
,
1070 struct res_counter memory_allocated
; /* Current allocated memory. */
1071 struct percpu_counter sockets_allocated
; /* Current number of sockets. */
1072 int memory_pressure
;
1074 unsigned long flags
;
1076 * memcg field is used to find which memcg we belong directly
1077 * Each memcg struct can hold more than one cg_proto, so container_of
1080 * The elegant solution would be having an inverse function to
1081 * proto_cgroup in struct proto, but that means polluting the structure
1082 * for everybody, instead of just for memcg users.
1084 struct mem_cgroup
*memcg
;
1087 int proto_register(struct proto
*prot
, int alloc_slab
);
1088 void proto_unregister(struct proto
*prot
);
1090 static inline bool memcg_proto_active(struct cg_proto
*cg_proto
)
1092 return test_bit(MEMCG_SOCK_ACTIVE
, &cg_proto
->flags
);
1095 static inline bool memcg_proto_activated(struct cg_proto
*cg_proto
)
1097 return test_bit(MEMCG_SOCK_ACTIVATED
, &cg_proto
->flags
);
1100 #ifdef SOCK_REFCNT_DEBUG
1101 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
1103 atomic_inc(&sk
->sk_prot
->socks
);
1106 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
1108 atomic_dec(&sk
->sk_prot
->socks
);
1109 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
1110 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
1113 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
1115 if (atomic_read(&sk
->sk_refcnt
) != 1)
1116 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
1117 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
1119 #else /* SOCK_REFCNT_DEBUG */
1120 #define sk_refcnt_debug_inc(sk) do { } while (0)
1121 #define sk_refcnt_debug_dec(sk) do { } while (0)
1122 #define sk_refcnt_debug_release(sk) do { } while (0)
1123 #endif /* SOCK_REFCNT_DEBUG */
1125 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
1126 extern struct static_key memcg_socket_limit_enabled
;
1127 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
1128 struct cg_proto
*cg_proto
)
1130 return proto
->proto_cgroup(parent_mem_cgroup(cg_proto
->memcg
));
1132 #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
1134 #define mem_cgroup_sockets_enabled 0
1135 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
1136 struct cg_proto
*cg_proto
)
1142 static inline bool sk_stream_memory_free(const struct sock
*sk
)
1144 if (sk
->sk_wmem_queued
>= sk
->sk_sndbuf
)
1147 return sk
->sk_prot
->stream_memory_free
?
1148 sk
->sk_prot
->stream_memory_free(sk
) : true;
1151 static inline bool sk_stream_is_writeable(const struct sock
*sk
)
1153 return sk_stream_wspace(sk
) >= sk_stream_min_wspace(sk
) &&
1154 sk_stream_memory_free(sk
);
1158 static inline bool sk_has_memory_pressure(const struct sock
*sk
)
1160 return sk
->sk_prot
->memory_pressure
!= NULL
;
1163 static inline bool sk_under_memory_pressure(const struct sock
*sk
)
1165 if (!sk
->sk_prot
->memory_pressure
)
1168 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1169 return !!sk
->sk_cgrp
->memory_pressure
;
1171 return !!*sk
->sk_prot
->memory_pressure
;
1174 static inline void sk_leave_memory_pressure(struct sock
*sk
)
1176 int *memory_pressure
= sk
->sk_prot
->memory_pressure
;
1178 if (!memory_pressure
)
1181 if (*memory_pressure
)
1182 *memory_pressure
= 0;
1184 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1185 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1186 struct proto
*prot
= sk
->sk_prot
;
1188 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1189 cg_proto
->memory_pressure
= 0;
1194 static inline void sk_enter_memory_pressure(struct sock
*sk
)
1196 if (!sk
->sk_prot
->enter_memory_pressure
)
1199 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1200 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1201 struct proto
*prot
= sk
->sk_prot
;
1203 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1204 cg_proto
->memory_pressure
= 1;
1207 sk
->sk_prot
->enter_memory_pressure(sk
);
1210 static inline long sk_prot_mem_limits(const struct sock
*sk
, int index
)
1212 long *prot
= sk
->sk_prot
->sysctl_mem
;
1213 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1214 prot
= sk
->sk_cgrp
->sysctl_mem
;
1218 static inline void memcg_memory_allocated_add(struct cg_proto
*prot
,
1222 struct res_counter
*fail
;
1225 ret
= res_counter_charge_nofail(&prot
->memory_allocated
,
1226 amt
<< PAGE_SHIFT
, &fail
);
1228 *parent_status
= OVER_LIMIT
;
1231 static inline void memcg_memory_allocated_sub(struct cg_proto
*prot
,
1234 res_counter_uncharge(&prot
->memory_allocated
, amt
<< PAGE_SHIFT
);
1237 static inline u64
memcg_memory_allocated_read(struct cg_proto
*prot
)
1240 ret
= res_counter_read_u64(&prot
->memory_allocated
, RES_USAGE
);
1241 return ret
>> PAGE_SHIFT
;
1245 sk_memory_allocated(const struct sock
*sk
)
1247 struct proto
*prot
= sk
->sk_prot
;
1248 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1249 return memcg_memory_allocated_read(sk
->sk_cgrp
);
1251 return atomic_long_read(prot
->memory_allocated
);
1255 sk_memory_allocated_add(struct sock
*sk
, int amt
, int *parent_status
)
1257 struct proto
*prot
= sk
->sk_prot
;
1259 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1260 memcg_memory_allocated_add(sk
->sk_cgrp
, amt
, parent_status
);
1261 /* update the root cgroup regardless */
1262 atomic_long_add_return(amt
, prot
->memory_allocated
);
1263 return memcg_memory_allocated_read(sk
->sk_cgrp
);
1266 return atomic_long_add_return(amt
, prot
->memory_allocated
);
1270 sk_memory_allocated_sub(struct sock
*sk
, int amt
)
1272 struct proto
*prot
= sk
->sk_prot
;
1274 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1275 memcg_memory_allocated_sub(sk
->sk_cgrp
, amt
);
1277 atomic_long_sub(amt
, prot
->memory_allocated
);
1280 static inline void sk_sockets_allocated_dec(struct sock
*sk
)
1282 struct proto
*prot
= sk
->sk_prot
;
1284 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1285 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1287 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1288 percpu_counter_dec(&cg_proto
->sockets_allocated
);
1291 percpu_counter_dec(prot
->sockets_allocated
);
1294 static inline void sk_sockets_allocated_inc(struct sock
*sk
)
1296 struct proto
*prot
= sk
->sk_prot
;
1298 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1299 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1301 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1302 percpu_counter_inc(&cg_proto
->sockets_allocated
);
1305 percpu_counter_inc(prot
->sockets_allocated
);
1309 sk_sockets_allocated_read_positive(struct sock
*sk
)
1311 struct proto
*prot
= sk
->sk_prot
;
1313 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1314 return percpu_counter_read_positive(&sk
->sk_cgrp
->sockets_allocated
);
1316 return percpu_counter_read_positive(prot
->sockets_allocated
);
1320 proto_sockets_allocated_sum_positive(struct proto
*prot
)
1322 return percpu_counter_sum_positive(prot
->sockets_allocated
);
1326 proto_memory_allocated(struct proto
*prot
)
1328 return atomic_long_read(prot
->memory_allocated
);
1332 proto_memory_pressure(struct proto
*prot
)
1334 if (!prot
->memory_pressure
)
1336 return !!*prot
->memory_pressure
;
1340 #ifdef CONFIG_PROC_FS
1341 /* Called with local bh disabled */
1342 void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int inc
);
1343 int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
1345 static inline void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
1352 /* With per-bucket locks this operation is not-atomic, so that
1353 * this version is not worse.
1355 static inline void __sk_prot_rehash(struct sock
*sk
)
1357 sk
->sk_prot
->unhash(sk
);
1358 sk
->sk_prot
->hash(sk
);
1361 void sk_prot_clear_portaddr_nulls(struct sock
*sk
, int size
);
1363 /* About 10 seconds */
1364 #define SOCK_DESTROY_TIME (10*HZ)
1366 /* Sockets 0-1023 can't be bound to unless you are superuser */
1367 #define PROT_SOCK 1024
1369 #define SHUTDOWN_MASK 3
1370 #define RCV_SHUTDOWN 1
1371 #define SEND_SHUTDOWN 2
1373 #define SOCK_SNDBUF_LOCK 1
1374 #define SOCK_RCVBUF_LOCK 2
1375 #define SOCK_BINDADDR_LOCK 4
1376 #define SOCK_BINDPORT_LOCK 8
1378 /* sock_iocb: used to kick off async processing of socket ios */
1380 struct list_head list
;
1384 struct socket
*sock
;
1386 struct scm_cookie
*scm
;
1387 struct msghdr
*msg
, async_msg
;
1388 struct kiocb
*kiocb
;
1391 static inline struct sock_iocb
*kiocb_to_siocb(struct kiocb
*iocb
)
1393 return (struct sock_iocb
*)iocb
->private;
1396 static inline struct kiocb
*siocb_to_kiocb(struct sock_iocb
*si
)
1401 struct socket_alloc
{
1402 struct socket socket
;
1403 struct inode vfs_inode
;
1406 static inline struct socket
*SOCKET_I(struct inode
*inode
)
1408 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
1411 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
1413 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
1417 * Functions for memory accounting
1419 int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
1420 void __sk_mem_reclaim(struct sock
*sk
);
1422 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1423 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1424 #define SK_MEM_SEND 0
1425 #define SK_MEM_RECV 1
1427 static inline int sk_mem_pages(int amt
)
1429 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
1432 static inline bool sk_has_account(struct sock
*sk
)
1434 /* return true if protocol supports memory accounting */
1435 return !!sk
->sk_prot
->memory_allocated
;
1438 static inline bool sk_wmem_schedule(struct sock
*sk
, int size
)
1440 if (!sk_has_account(sk
))
1442 return size
<= sk
->sk_forward_alloc
||
1443 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
1447 sk_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
, int size
)
1449 if (!sk_has_account(sk
))
1451 return size
<= sk
->sk_forward_alloc
||
1452 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
) ||
1453 skb_pfmemalloc(skb
);
1456 static inline void sk_mem_reclaim(struct sock
*sk
)
1458 if (!sk_has_account(sk
))
1460 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
1461 __sk_mem_reclaim(sk
);
1464 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
1466 if (!sk_has_account(sk
))
1468 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
1469 __sk_mem_reclaim(sk
);
1472 static inline void sk_mem_charge(struct sock
*sk
, int size
)
1474 if (!sk_has_account(sk
))
1476 sk
->sk_forward_alloc
-= size
;
1479 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
1481 if (!sk_has_account(sk
))
1483 sk
->sk_forward_alloc
+= size
;
1486 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
1488 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
1489 sk
->sk_wmem_queued
-= skb
->truesize
;
1490 sk_mem_uncharge(sk
, skb
->truesize
);
1494 /* Used by processes to "lock" a socket state, so that
1495 * interrupts and bottom half handlers won't change it
1496 * from under us. It essentially blocks any incoming
1497 * packets, so that we won't get any new data or any
1498 * packets that change the state of the socket.
1500 * While locked, BH processing will add new packets to
1501 * the backlog queue. This queue is processed by the
1502 * owner of the socket lock right before it is released.
1504 * Since ~2.3.5 it is also exclusive sleep lock serializing
1505 * accesses from user process context.
1507 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1509 static inline void sock_release_ownership(struct sock
*sk
)
1511 sk
->sk_lock
.owned
= 0;
1515 * Macro so as to not evaluate some arguments when
1516 * lockdep is not enabled.
1518 * Mark both the sk_lock and the sk_lock.slock as a
1519 * per-address-family lock class.
1521 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1523 sk->sk_lock.owned = 0; \
1524 init_waitqueue_head(&sk->sk_lock.wq); \
1525 spin_lock_init(&(sk)->sk_lock.slock); \
1526 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1527 sizeof((sk)->sk_lock)); \
1528 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1530 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1533 void lock_sock_nested(struct sock
*sk
, int subclass
);
1535 static inline void lock_sock(struct sock
*sk
)
1537 lock_sock_nested(sk
, 0);
1540 void release_sock(struct sock
*sk
);
1542 /* BH context may only use the following locking interface. */
1543 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1544 #define bh_lock_sock_nested(__sk) \
1545 spin_lock_nested(&((__sk)->sk_lock.slock), \
1546 SINGLE_DEPTH_NESTING)
1547 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1549 bool lock_sock_fast(struct sock
*sk
);
1551 * unlock_sock_fast - complement of lock_sock_fast
1555 * fast unlock socket for user context.
1556 * If slow mode is on, we call regular release_sock()
1558 static inline void unlock_sock_fast(struct sock
*sk
, bool slow
)
1563 spin_unlock_bh(&sk
->sk_lock
.slock
);
1567 struct sock
*sk_alloc(struct net
*net
, int family
, gfp_t priority
,
1568 struct proto
*prot
);
1569 void sk_free(struct sock
*sk
);
1570 void sk_release_kernel(struct sock
*sk
);
1571 struct sock
*sk_clone_lock(const struct sock
*sk
, const gfp_t priority
);
1573 struct sk_buff
*sock_wmalloc(struct sock
*sk
, unsigned long size
, int force
,
1575 void sock_wfree(struct sk_buff
*skb
);
1576 void skb_orphan_partial(struct sk_buff
*skb
);
1577 void sock_rfree(struct sk_buff
*skb
);
1578 void sock_edemux(struct sk_buff
*skb
);
1580 int sock_setsockopt(struct socket
*sock
, int level
, int op
,
1581 char __user
*optval
, unsigned int optlen
);
1583 int sock_getsockopt(struct socket
*sock
, int level
, int op
,
1584 char __user
*optval
, int __user
*optlen
);
1585 struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
, unsigned long size
,
1586 int noblock
, int *errcode
);
1587 struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
, unsigned long header_len
,
1588 unsigned long data_len
, int noblock
,
1589 int *errcode
, int max_page_order
);
1590 void *sock_kmalloc(struct sock
*sk
, int size
, gfp_t priority
);
1591 void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
1592 void sk_send_sigurg(struct sock
*sk
);
1595 * Functions to fill in entries in struct proto_ops when a protocol
1596 * does not implement a particular function.
1598 int sock_no_bind(struct socket
*, struct sockaddr
*, int);
1599 int sock_no_connect(struct socket
*, struct sockaddr
*, int, int);
1600 int sock_no_socketpair(struct socket
*, struct socket
*);
1601 int sock_no_accept(struct socket
*, struct socket
*, int);
1602 int sock_no_getname(struct socket
*, struct sockaddr
*, int *, int);
1603 unsigned int sock_no_poll(struct file
*, struct socket
*,
1604 struct poll_table_struct
*);
1605 int sock_no_ioctl(struct socket
*, unsigned int, unsigned long);
1606 int sock_no_listen(struct socket
*, int);
1607 int sock_no_shutdown(struct socket
*, int);
1608 int sock_no_getsockopt(struct socket
*, int , int, char __user
*, int __user
*);
1609 int sock_no_setsockopt(struct socket
*, int, int, char __user
*, unsigned int);
1610 int sock_no_sendmsg(struct kiocb
*, struct socket
*, struct msghdr
*, size_t);
1611 int sock_no_recvmsg(struct kiocb
*, struct socket
*, struct msghdr
*, size_t,
1613 int sock_no_mmap(struct file
*file
, struct socket
*sock
,
1614 struct vm_area_struct
*vma
);
1615 ssize_t
sock_no_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
1616 size_t size
, int flags
);
1619 * Functions to fill in entries in struct proto_ops when a protocol
1620 * uses the inet style.
1622 int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1623 char __user
*optval
, int __user
*optlen
);
1624 int sock_common_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
1625 struct msghdr
*msg
, size_t size
, int flags
);
1626 int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1627 char __user
*optval
, unsigned int optlen
);
1628 int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
1629 int optname
, char __user
*optval
, int __user
*optlen
);
1630 int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
1631 int optname
, char __user
*optval
, unsigned int optlen
);
1633 void sk_common_release(struct sock
*sk
);
1636 * Default socket callbacks and setup code
1639 /* Initialise core socket variables */
1640 void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1643 * Socket reference counting postulates.
1645 * * Each user of socket SHOULD hold a reference count.
1646 * * Each access point to socket (an hash table bucket, reference from a list,
1647 * running timer, skb in flight MUST hold a reference count.
1648 * * When reference count hits 0, it means it will never increase back.
1649 * * When reference count hits 0, it means that no references from
1650 * outside exist to this socket and current process on current CPU
1651 * is last user and may/should destroy this socket.
1652 * * sk_free is called from any context: process, BH, IRQ. When
1653 * it is called, socket has no references from outside -> sk_free
1654 * may release descendant resources allocated by the socket, but
1655 * to the time when it is called, socket is NOT referenced by any
1656 * hash tables, lists etc.
1657 * * Packets, delivered from outside (from network or from another process)
1658 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1659 * when they sit in queue. Otherwise, packets will leak to hole, when
1660 * socket is looked up by one cpu and unhasing is made by another CPU.
1661 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1662 * (leak to backlog). Packet socket does all the processing inside
1663 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1664 * use separate SMP lock, so that they are prone too.
1667 /* Ungrab socket and destroy it, if it was the last reference. */
1668 static inline void sock_put(struct sock
*sk
)
1670 if (atomic_dec_and_test(&sk
->sk_refcnt
))
1673 /* Generic version of sock_put(), dealing with all sockets
1674 * (TCP_TIMEWAIT, ESTABLISHED...)
1676 void sock_gen_put(struct sock
*sk
);
1678 int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
, const int nested
);
1680 static inline void sk_tx_queue_set(struct sock
*sk
, int tx_queue
)
1682 sk
->sk_tx_queue_mapping
= tx_queue
;
1685 static inline void sk_tx_queue_clear(struct sock
*sk
)
1687 sk
->sk_tx_queue_mapping
= -1;
1690 static inline int sk_tx_queue_get(const struct sock
*sk
)
1692 return sk
? sk
->sk_tx_queue_mapping
: -1;
1695 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1697 sk_tx_queue_clear(sk
);
1698 sk
->sk_socket
= sock
;
1701 static inline wait_queue_head_t
*sk_sleep(struct sock
*sk
)
1703 BUILD_BUG_ON(offsetof(struct socket_wq
, wait
) != 0);
1704 return &rcu_dereference_raw(sk
->sk_wq
)->wait
;
1706 /* Detach socket from process context.
1707 * Announce socket dead, detach it from wait queue and inode.
1708 * Note that parent inode held reference count on this struct sock,
1709 * we do not release it in this function, because protocol
1710 * probably wants some additional cleanups or even continuing
1711 * to work with this socket (TCP).
1713 static inline void sock_orphan(struct sock
*sk
)
1715 write_lock_bh(&sk
->sk_callback_lock
);
1716 sock_set_flag(sk
, SOCK_DEAD
);
1717 sk_set_socket(sk
, NULL
);
1719 write_unlock_bh(&sk
->sk_callback_lock
);
1722 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1724 write_lock_bh(&sk
->sk_callback_lock
);
1725 sk
->sk_wq
= parent
->wq
;
1727 sk_set_socket(sk
, parent
);
1728 security_sock_graft(sk
, parent
);
1729 write_unlock_bh(&sk
->sk_callback_lock
);
1732 kuid_t
sock_i_uid(struct sock
*sk
);
1733 unsigned long sock_i_ino(struct sock
*sk
);
1735 static inline struct dst_entry
*
1736 __sk_dst_get(struct sock
*sk
)
1738 return rcu_dereference_check(sk
->sk_dst_cache
, sock_owned_by_user(sk
) ||
1739 lockdep_is_held(&sk
->sk_lock
.slock
));
1742 static inline struct dst_entry
*
1743 sk_dst_get(struct sock
*sk
)
1745 struct dst_entry
*dst
;
1748 dst
= rcu_dereference(sk
->sk_dst_cache
);
1749 if (dst
&& !atomic_inc_not_zero(&dst
->__refcnt
))
1755 static inline void dst_negative_advice(struct sock
*sk
)
1757 struct dst_entry
*ndst
, *dst
= __sk_dst_get(sk
);
1759 if (dst
&& dst
->ops
->negative_advice
) {
1760 ndst
= dst
->ops
->negative_advice(dst
);
1763 rcu_assign_pointer(sk
->sk_dst_cache
, ndst
);
1764 sk_tx_queue_clear(sk
);
1770 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1772 struct dst_entry
*old_dst
;
1774 sk_tx_queue_clear(sk
);
1776 * This can be called while sk is owned by the caller only,
1777 * with no state that can be checked in a rcu_dereference_check() cond
1779 old_dst
= rcu_dereference_raw(sk
->sk_dst_cache
);
1780 rcu_assign_pointer(sk
->sk_dst_cache
, dst
);
1781 dst_release(old_dst
);
1785 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1787 struct dst_entry
*old_dst
;
1789 sk_tx_queue_clear(sk
);
1790 old_dst
= xchg((__force
struct dst_entry
**)&sk
->sk_dst_cache
, dst
);
1791 dst_release(old_dst
);
1795 __sk_dst_reset(struct sock
*sk
)
1797 __sk_dst_set(sk
, NULL
);
1801 sk_dst_reset(struct sock
*sk
)
1803 sk_dst_set(sk
, NULL
);
1806 struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1808 struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1810 static inline bool sk_can_gso(const struct sock
*sk
)
1812 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1815 void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1817 static inline void sk_nocaps_add(struct sock
*sk
, netdev_features_t flags
)
1819 sk
->sk_route_nocaps
|= flags
;
1820 sk
->sk_route_caps
&= ~flags
;
1823 static inline int skb_do_copy_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1824 char __user
*from
, char *to
,
1825 int copy
, int offset
)
1827 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1829 __wsum csum
= csum_and_copy_from_user(from
, to
, copy
, 0, &err
);
1832 skb
->csum
= csum_block_add(skb
->csum
, csum
, offset
);
1833 } else if (sk
->sk_route_caps
& NETIF_F_NOCACHE_COPY
) {
1834 if (!access_ok(VERIFY_READ
, from
, copy
) ||
1835 __copy_from_user_nocache(to
, from
, copy
))
1837 } else if (copy_from_user(to
, from
, copy
))
1843 static inline int skb_add_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1844 char __user
*from
, int copy
)
1846 int err
, offset
= skb
->len
;
1848 err
= skb_do_copy_data_nocache(sk
, skb
, from
, skb_put(skb
, copy
),
1851 __skb_trim(skb
, offset
);
1856 static inline int skb_copy_to_page_nocache(struct sock
*sk
, char __user
*from
,
1857 struct sk_buff
*skb
,
1863 err
= skb_do_copy_data_nocache(sk
, skb
, from
, page_address(page
) + off
,
1869 skb
->data_len
+= copy
;
1870 skb
->truesize
+= copy
;
1871 sk
->sk_wmem_queued
+= copy
;
1872 sk_mem_charge(sk
, copy
);
1876 static inline int skb_copy_to_page(struct sock
*sk
, char __user
*from
,
1877 struct sk_buff
*skb
, struct page
*page
,
1880 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1882 __wsum csum
= csum_and_copy_from_user(from
,
1883 page_address(page
) + off
,
1887 skb
->csum
= csum_block_add(skb
->csum
, csum
, skb
->len
);
1888 } else if (copy_from_user(page_address(page
) + off
, from
, copy
))
1892 skb
->data_len
+= copy
;
1893 skb
->truesize
+= copy
;
1894 sk
->sk_wmem_queued
+= copy
;
1895 sk_mem_charge(sk
, copy
);
1900 * sk_wmem_alloc_get - returns write allocations
1903 * Returns sk_wmem_alloc minus initial offset of one
1905 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
1907 return atomic_read(&sk
->sk_wmem_alloc
) - 1;
1911 * sk_rmem_alloc_get - returns read allocations
1914 * Returns sk_rmem_alloc
1916 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
1918 return atomic_read(&sk
->sk_rmem_alloc
);
1922 * sk_has_allocations - check if allocations are outstanding
1925 * Returns true if socket has write or read allocations
1927 static inline bool sk_has_allocations(const struct sock
*sk
)
1929 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
1933 * wq_has_sleeper - check if there are any waiting processes
1934 * @wq: struct socket_wq
1936 * Returns true if socket_wq has waiting processes
1938 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1939 * barrier call. They were added due to the race found within the tcp code.
1941 * Consider following tcp code paths:
1945 * sys_select receive packet
1947 * __add_wait_queue update tp->rcv_nxt
1949 * tp->rcv_nxt check sock_def_readable
1951 * schedule rcu_read_lock();
1952 * wq = rcu_dereference(sk->sk_wq);
1953 * if (wq && waitqueue_active(&wq->wait))
1954 * wake_up_interruptible(&wq->wait)
1958 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1959 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1960 * could then endup calling schedule and sleep forever if there are no more
1961 * data on the socket.
1964 static inline bool wq_has_sleeper(struct socket_wq
*wq
)
1966 /* We need to be sure we are in sync with the
1967 * add_wait_queue modifications to the wait queue.
1969 * This memory barrier is paired in the sock_poll_wait.
1972 return wq
&& waitqueue_active(&wq
->wait
);
1976 * sock_poll_wait - place memory barrier behind the poll_wait call.
1978 * @wait_address: socket wait queue
1981 * See the comments in the wq_has_sleeper function.
1983 static inline void sock_poll_wait(struct file
*filp
,
1984 wait_queue_head_t
*wait_address
, poll_table
*p
)
1986 if (!poll_does_not_wait(p
) && wait_address
) {
1987 poll_wait(filp
, wait_address
, p
);
1988 /* We need to be sure we are in sync with the
1989 * socket flags modification.
1991 * This memory barrier is paired in the wq_has_sleeper.
1997 static inline void skb_set_hash_from_sk(struct sk_buff
*skb
, struct sock
*sk
)
1999 if (sk
->sk_txhash
) {
2001 skb
->hash
= sk
->sk_txhash
;
2006 * Queue a received datagram if it will fit. Stream and sequenced
2007 * protocols can't normally use this as they need to fit buffers in
2008 * and play with them.
2010 * Inlined as it's very short and called for pretty much every
2011 * packet ever received.
2014 static inline void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
2018 skb
->destructor
= sock_wfree
;
2019 skb_set_hash_from_sk(skb
, sk
);
2021 * We used to take a refcount on sk, but following operation
2022 * is enough to guarantee sk_free() wont free this sock until
2023 * all in-flight packets are completed
2025 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
2028 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
2032 skb
->destructor
= sock_rfree
;
2033 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
2034 sk_mem_charge(sk
, skb
->truesize
);
2037 void sk_reset_timer(struct sock
*sk
, struct timer_list
*timer
,
2038 unsigned long expires
);
2040 void sk_stop_timer(struct sock
*sk
, struct timer_list
*timer
);
2042 int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
2044 int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
);
2047 * Recover an error report and clear atomically
2050 static inline int sock_error(struct sock
*sk
)
2053 if (likely(!sk
->sk_err
))
2055 err
= xchg(&sk
->sk_err
, 0);
2059 static inline unsigned long sock_wspace(struct sock
*sk
)
2063 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
2064 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
2071 static inline void sk_wake_async(struct sock
*sk
, int how
, int band
)
2073 if (sock_flag(sk
, SOCK_FASYNC
))
2074 sock_wake_async(sk
->sk_socket
, how
, band
);
2077 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2078 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2079 * Note: for send buffers, TCP works better if we can build two skbs at
2082 #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2084 #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2085 #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
2087 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
2089 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
2090 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
2091 sk
->sk_sndbuf
= max_t(u32
, sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
2095 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
);
2098 * sk_page_frag - return an appropriate page_frag
2101 * If socket allocation mode allows current thread to sleep, it means its
2102 * safe to use the per task page_frag instead of the per socket one.
2104 static inline struct page_frag
*sk_page_frag(struct sock
*sk
)
2106 if (sk
->sk_allocation
& __GFP_WAIT
)
2107 return ¤t
->task_frag
;
2109 return &sk
->sk_frag
;
2112 bool sk_page_frag_refill(struct sock
*sk
, struct page_frag
*pfrag
);
2115 * Default write policy as shown to user space via poll/select/SIGIO
2117 static inline bool sock_writeable(const struct sock
*sk
)
2119 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
2122 static inline gfp_t
gfp_any(void)
2124 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
2127 static inline long sock_rcvtimeo(const struct sock
*sk
, bool noblock
)
2129 return noblock
? 0 : sk
->sk_rcvtimeo
;
2132 static inline long sock_sndtimeo(const struct sock
*sk
, bool noblock
)
2134 return noblock
? 0 : sk
->sk_sndtimeo
;
2137 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
2139 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
2142 /* Alas, with timeout socket operations are not restartable.
2143 * Compare this to poll().
2145 static inline int sock_intr_errno(long timeo
)
2147 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
2150 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
2151 struct sk_buff
*skb
);
2152 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
2153 struct sk_buff
*skb
);
2156 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
2158 ktime_t kt
= skb
->tstamp
;
2159 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
2162 * generate control messages if
2163 * - receive time stamping in software requested
2164 * - software time stamp available and wanted
2165 * - hardware time stamps available and wanted
2167 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
2168 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RX_SOFTWARE
) ||
2170 (sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
||
2171 skb_shinfo(skb
)->tx_flags
& SKBTX_ANY_SW_TSTAMP
)) ||
2172 (hwtstamps
->hwtstamp
.tv64
&&
2173 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
)))
2174 __sock_recv_timestamp(msg
, sk
, skb
);
2178 if (sock_flag(sk
, SOCK_WIFI_STATUS
) && skb
->wifi_acked_valid
)
2179 __sock_recv_wifi_status(msg
, sk
, skb
);
2182 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2183 struct sk_buff
*skb
);
2185 static inline void sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2186 struct sk_buff
*skb
)
2188 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2189 (1UL << SOCK_RCVTSTAMP))
2190 #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2191 SOF_TIMESTAMPING_RAW_HARDWARE)
2193 if (sk
->sk_flags
& FLAGS_TS_OR_DROPS
|| sk
->sk_tsflags
& TSFLAGS_ANY
)
2194 __sock_recv_ts_and_drops(msg
, sk
, skb
);
2196 sk
->sk_stamp
= skb
->tstamp
;
2200 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2201 * @sk: socket sending this packet
2202 * @tx_flags: filled with instructions for time stamping
2204 void sock_tx_timestamp(struct sock
*sk
, __u8
*tx_flags
);
2207 * sk_eat_skb - Release a skb if it is no longer needed
2208 * @sk: socket to eat this skb from
2209 * @skb: socket buffer to eat
2210 * @copied_early: flag indicating whether DMA operations copied this data early
2212 * This routine must be called with interrupts disabled or with the socket
2213 * locked so that the sk_buff queue operation is ok.
2215 #ifdef CONFIG_NET_DMA
2216 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, bool copied_early
)
2218 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2222 __skb_queue_tail(&sk
->sk_async_wait_queue
, skb
);
2225 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, bool copied_early
)
2227 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2233 struct net
*sock_net(const struct sock
*sk
)
2235 return read_pnet(&sk
->sk_net
);
2239 void sock_net_set(struct sock
*sk
, struct net
*net
)
2241 write_pnet(&sk
->sk_net
, net
);
2245 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
2246 * They should not hold a reference to a namespace in order to allow
2248 * Sockets after sk_change_net should be released using sk_release_kernel
2250 static inline void sk_change_net(struct sock
*sk
, struct net
*net
)
2252 struct net
*current_net
= sock_net(sk
);
2254 if (!net_eq(current_net
, net
)) {
2255 put_net(current_net
);
2256 sock_net_set(sk
, hold_net(net
));
2260 static inline struct sock
*skb_steal_sock(struct sk_buff
*skb
)
2263 struct sock
*sk
= skb
->sk
;
2265 skb
->destructor
= NULL
;
2272 void sock_enable_timestamp(struct sock
*sk
, int flag
);
2273 int sock_get_timestamp(struct sock
*, struct timeval __user
*);
2274 int sock_get_timestampns(struct sock
*, struct timespec __user
*);
2275 int sock_recv_errqueue(struct sock
*sk
, struct msghdr
*msg
, int len
, int level
,
2278 bool sk_ns_capable(const struct sock
*sk
,
2279 struct user_namespace
*user_ns
, int cap
);
2280 bool sk_capable(const struct sock
*sk
, int cap
);
2281 bool sk_net_capable(const struct sock
*sk
, int cap
);
2284 * Enable debug/info messages
2286 extern int net_msg_warn
;
2287 #define NETDEBUG(fmt, args...) \
2288 do { if (net_msg_warn) printk(fmt,##args); } while (0)
2290 #define LIMIT_NETDEBUG(fmt, args...) \
2291 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
2293 extern __u32 sysctl_wmem_max
;
2294 extern __u32 sysctl_rmem_max
;
2296 extern int sysctl_optmem_max
;
2298 extern __u32 sysctl_wmem_default
;
2299 extern __u32 sysctl_rmem_default
;
2301 #endif /* _SOCK_H */