1 /* Copyright (c) 2015 Samsung Electronics Co., Ltd. */
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * Definitions for the AF_INET socket handler.
9 * Version: @(#)sock.h 1.0.4 05/13/93
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Florian La Roche <flla@stud.uni-sb.de>
17 * Alan Cox : Volatiles in skbuff pointers. See
18 * skbuff comments. May be overdone,
19 * better to prove they can be removed
21 * Alan Cox : Added a zapped field for tcp to note
22 * a socket is reset and must stay shut up
23 * Alan Cox : New fields for options
24 * Pauline Middelink : identd support
25 * Alan Cox : Eliminate low level recv/recvfrom
26 * David S. Miller : New socket lookup architecture.
27 * Steve Whitehouse: Default routines for sock_ops
28 * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made
29 * protinfo be just a void pointer, as the
30 * protocol specific parts were moved to
31 * respective headers and ipv4/v6, etc now
32 * use private slabcaches for its socks
33 * Pedro Hortas : New flags field for socket options
36 * This program is free software; you can redistribute it and/or
37 * modify it under the terms of the GNU General Public License
38 * as published by the Free Software Foundation; either version
39 * 2 of the License, or (at your option) any later version.
43 * KwnagHyun Kim <kh0304.kim@samsung.com> 2015/07/08
44 * Baesung Park <baesung.park@samsung.com> 2015/07/08
45 * Vignesh Saravanaperumal <vignesh1.s@samsung.com> 2015/07/08
46 * Add codes to share UID/PID information
53 #include <linux/hardirq.h>
54 #include <linux/kernel.h>
55 #include <linux/list.h>
56 #include <linux/list_nulls.h>
57 #include <linux/timer.h>
58 #include <linux/cache.h>
59 #include <linux/bitops.h>
60 #include <linux/lockdep.h>
61 #include <linux/netdevice.h>
62 #include <linux/skbuff.h> /* struct sk_buff */
64 #include <linux/security.h>
65 #include <linux/slab.h>
66 #include <linux/uaccess.h>
67 #include <linux/memcontrol.h>
68 #include <linux/res_counter.h>
69 #include <linux/static_key.h>
70 #include <linux/aio.h>
71 #include <linux/sched.h>
73 #include <linux/filter.h>
74 #include <linux/rculist_nulls.h>
75 #include <linux/poll.h>
77 #include <linux/atomic.h>
79 #include <net/checksum.h>
81 #define TCP_BACKLOG_SCALE 4
86 int mem_cgroup_sockets_init(struct mem_cgroup
*memcg
, struct cgroup_subsys
*ss
);
87 void mem_cgroup_sockets_destroy(struct mem_cgroup
*memcg
);
90 int mem_cgroup_sockets_init(struct mem_cgroup
*memcg
, struct cgroup_subsys
*ss
)
95 void mem_cgroup_sockets_destroy(struct mem_cgroup
*memcg
)
100 * This structure really needs to be cleaned up.
101 * Most of it is for TCP, and not used by any of
102 * the other protocols.
105 /* Define this to get the SOCK_DBG debugging facility. */
106 #define SOCK_DEBUGGING
107 #ifdef SOCK_DEBUGGING
108 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
109 printk(KERN_DEBUG msg); } while (0)
111 /* Validate arguments and do nothing */
112 static inline __printf(2, 3)
113 void SOCK_DEBUG(const struct sock
*sk
, const char *msg
, ...)
118 /* This is the per-socket lock. The spinlock provides a synchronization
119 * between user contexts and software interrupt processing, whereas the
120 * mini-semaphore synchronizes multiple users amongst themselves.
125 wait_queue_head_t wq
;
127 * We express the mutex-alike socket_lock semantics
128 * to the lock validator by explicitly managing
129 * the slock as a lock variant (in addition to
132 #ifdef CONFIG_DEBUG_LOCK_ALLOC
133 struct lockdep_map dep_map
;
141 typedef __u32 __bitwise __portpair
;
142 typedef __u64 __bitwise __addrpair
;
145 * struct sock_common - minimal network layer representation of sockets
146 * @skc_daddr: Foreign IPv4 addr
147 * @skc_rcv_saddr: Bound local IPv4 addr
148 * @skc_hash: hash value used with various protocol lookup tables
149 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
150 * @skc_dport: placeholder for inet_dport/tw_dport
151 * @skc_num: placeholder for inet_num/tw_num
152 * @skc_family: network address family
153 * @skc_state: Connection state
154 * @skc_reuse: %SO_REUSEADDR setting
155 * @skc_reuseport: %SO_REUSEPORT setting
156 * @skc_bound_dev_if: bound device index if != 0
157 * @skc_bind_node: bind hash linkage for various protocol lookup tables
158 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
159 * @skc_prot: protocol handlers inside a network family
160 * @skc_net: reference to the network namespace of this socket
161 * @skc_node: main hash linkage for various protocol lookup tables
162 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
163 * @skc_tx_queue_mapping: tx queue number for this connection
164 * @skc_refcnt: reference count
166 * This is the minimal network layer representation of sockets, the header
167 * for struct sock and struct inet_timewait_sock.
170 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
171 * address on 64bit arches : cf INET_MATCH() and INET_TW_MATCH()
174 __addrpair skc_addrpair
;
177 __be32 skc_rcv_saddr
;
181 unsigned int skc_hash
;
182 __u16 skc_u16hashes
[2];
184 /* skc_dport && skc_num must be grouped as well */
186 __portpair skc_portpair
;
193 unsigned short skc_family
;
194 volatile unsigned char skc_state
;
195 unsigned char skc_reuse
:4;
196 unsigned char skc_reuseport
:4;
197 int skc_bound_dev_if
;
199 struct hlist_node skc_bind_node
;
200 struct hlist_nulls_node skc_portaddr_node
;
202 struct proto
*skc_prot
;
207 * fields between dontcopy_begin/dontcopy_end
208 * are not copied in sock_copy()
211 int skc_dontcopy_begin
[0];
214 struct hlist_node skc_node
;
215 struct hlist_nulls_node skc_nulls_node
;
217 int skc_tx_queue_mapping
;
220 int skc_dontcopy_end
[0];
226 * struct sock - network layer representation of sockets
227 * @__sk_common: shared layout with inet_timewait_sock
228 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
229 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
230 * @sk_lock: synchronizer
231 * @sk_rcvbuf: size of receive buffer in bytes
232 * @sk_wq: sock wait queue and async head
233 * @sk_rx_dst: receive input route used by early tcp demux
234 * @sk_dst_cache: destination cache
235 * @sk_dst_lock: destination cache lock
236 * @sk_policy: flow policy
237 * @sk_receive_queue: incoming packets
238 * @sk_wmem_alloc: transmit queue bytes committed
239 * @sk_write_queue: Packet sending queue
240 * @sk_async_wait_queue: DMA copied packets
241 * @sk_omem_alloc: "o" is "option" or "other"
242 * @sk_wmem_queued: persistent queue size
243 * @sk_forward_alloc: space allocated forward
244 * @sk_allocation: allocation mode
245 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
246 * @sk_sndbuf: size of send buffer in bytes
247 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
248 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
249 * @sk_no_check: %SO_NO_CHECK setting, whether or not checkup 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_filter: socket filtering instructions
278 * @sk_protinfo: private area, net family specific, when not using slab
279 * @sk_timer: sock cleanup timer
280 * @sk_stamp: time stamp of last packet received
281 * @sk_socket: Identd and reporting IO signals
282 * @sk_user_data: RPC layer private data
283 * @sk_frag: cached page frag
284 * @sk_peek_off: current peek_offset value
285 * @sk_send_head: front of stuff to transmit
286 * @sk_security: used by security modules
287 * @sk_mark: generic packet mark
288 * @sk_classid: this socket's cgroup classid
289 * @sk_cgrp: this socket's cgroup-specific proto data
290 * @sk_write_pending: a write to stream socket waits to start
291 * @sk_state_change: callback to indicate change in the state of the sock
292 * @sk_data_ready: callback to indicate there is data to be processed
293 * @sk_write_space: callback to indicate there is bf sending space available
294 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
295 * @sk_backlog_rcv: callback to process the backlog
296 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
300 * Now struct inet_timewait_sock also uses sock_common, so please just
301 * don't add nothing before this first member (__sk_common) --acme
303 struct sock_common __sk_common
;
304 #define sk_node __sk_common.skc_node
305 #define sk_nulls_node __sk_common.skc_nulls_node
306 #define sk_refcnt __sk_common.skc_refcnt
307 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
309 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
310 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
311 #define sk_hash __sk_common.skc_hash
312 #define sk_family __sk_common.skc_family
313 #define sk_state __sk_common.skc_state
314 #define sk_reuse __sk_common.skc_reuse
315 #define sk_reuseport __sk_common.skc_reuseport
316 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
317 #define sk_bind_node __sk_common.skc_bind_node
318 #define sk_prot __sk_common.skc_prot
319 #define sk_net __sk_common.skc_net
320 socket_lock_t sk_lock
;
321 struct sk_buff_head sk_receive_queue
;
323 * The backlog queue is special, it is always used with
324 * the per-socket spinlock held and requires low latency
325 * access. Therefore we special case it's implementation.
326 * Note : rmem_alloc is in this structure to fill a hole
327 * on 64bit arches, not because its logically part of
333 struct sk_buff
*head
;
334 struct sk_buff
*tail
;
336 #define sk_rmem_alloc sk_backlog.rmem_alloc
337 int sk_forward_alloc
;
344 struct sk_filter __rcu
*sk_filter
;
345 struct socket_wq __rcu
*sk_wq
;
347 #ifdef CONFIG_NET_DMA
348 struct sk_buff_head sk_async_wait_queue
;
352 struct xfrm_policy
*sk_policy
[2];
354 unsigned long sk_flags
;
355 struct dst_entry
*sk_rx_dst
;
356 struct dst_entry __rcu
*sk_dst_cache
;
357 spinlock_t sk_dst_lock
;
358 atomic_t sk_wmem_alloc
;
359 atomic_t sk_omem_alloc
;
361 struct sk_buff_head sk_write_queue
;
362 kmemcheck_bitfield_begin(flags
);
363 unsigned int sk_shutdown
: 2,
367 #define SK_PROTOCOL_MAX U8_MAX
369 kmemcheck_bitfield_end(flags
);
372 u32 sk_pacing_rate
; /* bytes per second */
373 netdev_features_t sk_route_caps
;
374 netdev_features_t sk_route_nocaps
;
376 unsigned int sk_gso_max_size
;
379 unsigned long sk_lingertime
;
380 struct sk_buff_head sk_error_queue
;
381 struct proto
*sk_prot_creator
;
382 rwlock_t sk_callback_lock
;
385 unsigned short sk_ack_backlog
;
386 unsigned short sk_max_ack_backlog
;
388 #if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
389 __u32 sk_cgrp_prioidx
;
391 struct pid
*sk_peer_pid
;
392 const struct cred
*sk_peer_cred
;
396 struct timer_list sk_timer
;
398 struct socket
*sk_socket
;
400 struct page_frag sk_frag
;
401 struct sk_buff
*sk_send_head
;
403 int sk_write_pending
;
404 #ifdef CONFIG_SECURITY
410 struct cg_proto
*sk_cgrp
;
413 void (*sk_state_change
)(struct sock
*sk
);
414 void (*sk_data_ready
)(struct sock
*sk
, int bytes
);
415 void (*sk_write_space
)(struct sock
*sk
);
416 void (*sk_error_report
)(struct sock
*sk
);
417 int (*sk_backlog_rcv
)(struct sock
*sk
,
418 struct sk_buff
*skb
);
419 void (*sk_destruct
)(struct sock
*sk
);
423 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
424 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
425 * on a socket means that the socket will reuse everybody else's port
426 * without looking at the other's sk_reuse value.
429 #define SK_NO_REUSE 0
430 #define SK_CAN_REUSE 1
431 #define SK_FORCE_REUSE 2
433 static inline int sk_peek_offset(struct sock
*sk
, int flags
)
435 if ((flags
& MSG_PEEK
) && (sk
->sk_peek_off
>= 0))
436 return sk
->sk_peek_off
;
441 static inline void sk_peek_offset_bwd(struct sock
*sk
, int val
)
443 if (sk
->sk_peek_off
>= 0) {
444 if (sk
->sk_peek_off
>= val
)
445 sk
->sk_peek_off
-= val
;
451 static inline void sk_peek_offset_fwd(struct sock
*sk
, int val
)
453 if (sk
->sk_peek_off
>= 0)
454 sk
->sk_peek_off
+= val
;
458 * Hashed lists helper routines
460 static inline struct sock
*sk_entry(const struct hlist_node
*node
)
462 return hlist_entry(node
, struct sock
, sk_node
);
465 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
467 return hlist_entry(head
->first
, struct sock
, sk_node
);
470 static inline struct sock
*sk_head(const struct hlist_head
*head
)
472 return hlist_empty(head
) ? NULL
: __sk_head(head
);
475 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
477 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
480 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
482 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
485 static inline struct sock
*sk_next(const struct sock
*sk
)
487 return sk
->sk_node
.next
?
488 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
491 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
493 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
494 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
495 struct sock
, sk_nulls_node
) :
499 static inline bool sk_unhashed(const struct sock
*sk
)
501 return hlist_unhashed(&sk
->sk_node
);
504 static inline bool sk_hashed(const struct sock
*sk
)
506 return !sk_unhashed(sk
);
509 static inline void sk_node_init(struct hlist_node
*node
)
514 static inline void sk_nulls_node_init(struct hlist_nulls_node
*node
)
519 static inline void __sk_del_node(struct sock
*sk
)
521 __hlist_del(&sk
->sk_node
);
524 /* NB: equivalent to hlist_del_init_rcu */
525 static inline bool __sk_del_node_init(struct sock
*sk
)
529 sk_node_init(&sk
->sk_node
);
535 /* Grab socket reference count. This operation is valid only
536 when sk is ALREADY grabbed f.e. it is found in hash table
537 or a list and the lookup is made under lock preventing hash table
541 static inline void sock_hold(struct sock
*sk
)
543 atomic_inc(&sk
->sk_refcnt
);
546 /* Ungrab socket in the context, which assumes that socket refcnt
547 cannot hit zero, f.e. it is true in context of any socketcall.
549 static inline void __sock_put(struct sock
*sk
)
551 atomic_dec(&sk
->sk_refcnt
);
554 static inline bool sk_del_node_init(struct sock
*sk
)
556 bool rc
= __sk_del_node_init(sk
);
559 /* paranoid for a while -acme */
560 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
565 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
567 static inline bool __sk_nulls_del_node_init_rcu(struct sock
*sk
)
570 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
576 static inline bool sk_nulls_del_node_init_rcu(struct sock
*sk
)
578 bool rc
= __sk_nulls_del_node_init_rcu(sk
);
581 /* paranoid for a while -acme */
582 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
588 static inline void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
590 hlist_add_head(&sk
->sk_node
, list
);
593 static inline void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
596 __sk_add_node(sk
, list
);
599 static inline void sk_add_node_rcu(struct sock
*sk
, struct hlist_head
*list
)
602 hlist_add_head_rcu(&sk
->sk_node
, list
);
605 static inline void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
607 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
610 static inline void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
613 __sk_nulls_add_node_rcu(sk
, list
);
616 static inline void __sk_del_bind_node(struct sock
*sk
)
618 __hlist_del(&sk
->sk_bind_node
);
621 static inline void sk_add_bind_node(struct sock
*sk
,
622 struct hlist_head
*list
)
624 hlist_add_head(&sk
->sk_bind_node
, list
);
627 #define sk_for_each(__sk, list) \
628 hlist_for_each_entry(__sk, list, sk_node)
629 #define sk_for_each_rcu(__sk, list) \
630 hlist_for_each_entry_rcu(__sk, list, sk_node)
631 #define sk_nulls_for_each(__sk, node, list) \
632 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
633 #define sk_nulls_for_each_rcu(__sk, node, list) \
634 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
635 #define sk_for_each_from(__sk) \
636 hlist_for_each_entry_from(__sk, sk_node)
637 #define sk_nulls_for_each_from(__sk, node) \
638 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
639 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
640 #define sk_for_each_safe(__sk, tmp, list) \
641 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
642 #define sk_for_each_bound(__sk, list) \
643 hlist_for_each_entry(__sk, list, sk_bind_node)
645 static inline struct user_namespace
*sk_user_ns(struct sock
*sk
)
647 /* Careful only use this in a context where these parameters
648 * can not change and must all be valid, such as recvmsg from
651 return sk
->sk_socket
->file
->f_cred
->user_ns
;
665 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
666 SOCK_DBG
, /* %SO_DEBUG setting */
667 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
668 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
669 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
670 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
671 SOCK_MEMALLOC
, /* VM depends on this socket for swapping */
672 SOCK_TIMESTAMPING_TX_HARDWARE
, /* %SOF_TIMESTAMPING_TX_HARDWARE */
673 SOCK_TIMESTAMPING_TX_SOFTWARE
, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
674 SOCK_TIMESTAMPING_RX_HARDWARE
, /* %SOF_TIMESTAMPING_RX_HARDWARE */
675 SOCK_TIMESTAMPING_RX_SOFTWARE
, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
676 SOCK_TIMESTAMPING_SOFTWARE
, /* %SOF_TIMESTAMPING_SOFTWARE */
677 SOCK_TIMESTAMPING_RAW_HARDWARE
, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
678 SOCK_TIMESTAMPING_SYS_HARDWARE
, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
679 SOCK_FASYNC
, /* fasync() active */
681 SOCK_ZEROCOPY
, /* buffers from userspace */
682 SOCK_WIFI_STATUS
, /* push wifi status to userspace */
683 SOCK_NOFCS
, /* Tell NIC not to do the Ethernet FCS.
684 * Will use last 4 bytes of packet sent from
685 * user-space instead.
687 SOCK_FILTER_LOCKED
, /* Filter cannot be changed anymore */
688 SOCK_SELECT_ERR_QUEUE
, /* Wake select on error queue */
691 #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
693 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
695 nsk
->sk_flags
= osk
->sk_flags
;
698 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
700 __set_bit(flag
, &sk
->sk_flags
);
703 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
705 __clear_bit(flag
, &sk
->sk_flags
);
708 static inline bool sock_flag(const struct sock
*sk
, enum sock_flags flag
)
710 return test_bit(flag
, &sk
->sk_flags
);
714 extern struct static_key memalloc_socks
;
715 static inline int sk_memalloc_socks(void)
717 return static_key_false(&memalloc_socks
);
721 static inline int sk_memalloc_socks(void)
728 static inline gfp_t
sk_gfp_atomic(struct sock
*sk
, gfp_t gfp_mask
)
730 return GFP_ATOMIC
| (sk
->sk_allocation
& __GFP_MEMALLOC
);
733 static inline void sk_acceptq_removed(struct sock
*sk
)
735 sk
->sk_ack_backlog
--;
738 static inline void sk_acceptq_added(struct sock
*sk
)
740 sk
->sk_ack_backlog
++;
743 static inline bool sk_acceptq_is_full(const struct sock
*sk
)
745 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
749 * Compute minimal free write space needed to queue new packets.
751 static inline int sk_stream_min_wspace(const struct sock
*sk
)
753 return sk
->sk_wmem_queued
>> 1;
756 static inline int sk_stream_wspace(const struct sock
*sk
)
758 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
761 extern void sk_stream_write_space(struct sock
*sk
);
763 static inline bool sk_stream_memory_free(const struct sock
*sk
)
765 return sk
->sk_wmem_queued
< sk
->sk_sndbuf
;
768 /* OOB backlog add */
769 static inline void __sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
771 /* dont let skb dst not refcounted, we are going to leave rcu lock */
774 if (!sk
->sk_backlog
.tail
)
775 sk
->sk_backlog
.head
= skb
;
777 sk
->sk_backlog
.tail
->next
= skb
;
779 sk
->sk_backlog
.tail
= skb
;
784 * Take into account size of receive queue and backlog queue
785 * Do not take into account this skb truesize,
786 * to allow even a single big packet to come.
788 static inline bool sk_rcvqueues_full(const struct sock
*sk
, const struct sk_buff
*skb
,
791 unsigned int qsize
= sk
->sk_backlog
.len
+ atomic_read(&sk
->sk_rmem_alloc
);
793 return qsize
> limit
;
796 /* The per-socket spinlock must be held here. */
797 static inline __must_check
int sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
,
800 if (sk_rcvqueues_full(sk
, skb
, limit
* TCP_BACKLOG_SCALE
))
804 * If the skb was allocated from pfmemalloc reserves, only
805 * allow SOCK_MEMALLOC sockets to use it as this socket is
806 * helping free memory
808 if (skb_pfmemalloc(skb
) && !sock_flag(sk
, SOCK_MEMALLOC
))
811 __sk_add_backlog(sk
, skb
);
812 sk
->sk_backlog
.len
+= skb
->truesize
;
816 extern int __sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
818 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
820 if (sk_memalloc_socks() && skb_pfmemalloc(skb
))
821 return __sk_backlog_rcv(sk
, skb
);
823 return sk
->sk_backlog_rcv(sk
, skb
);
826 static inline void sock_rps_record_flow(const struct sock
*sk
)
829 struct rps_sock_flow_table
*sock_flow_table
;
832 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
833 rps_record_sock_flow(sock_flow_table
, sk
->sk_rxhash
);
838 static inline void sock_rps_reset_flow(const struct sock
*sk
)
841 struct rps_sock_flow_table
*sock_flow_table
;
844 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
845 rps_reset_sock_flow(sock_flow_table
, sk
->sk_rxhash
);
850 static inline void sock_rps_save_rxhash(struct sock
*sk
,
851 const struct sk_buff
*skb
)
854 if (unlikely(sk
->sk_rxhash
!= skb
->rxhash
)) {
855 sock_rps_reset_flow(sk
);
856 sk
->sk_rxhash
= skb
->rxhash
;
861 static inline void sock_rps_reset_rxhash(struct sock
*sk
)
864 sock_rps_reset_flow(sk
);
869 #define sk_wait_event(__sk, __timeo, __condition) \
871 release_sock(__sk); \
872 __rc = __condition; \
874 *(__timeo) = schedule_timeout(*(__timeo)); \
877 __rc = __condition; \
881 extern int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
882 extern int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
883 extern void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
884 extern int sk_stream_error(struct sock
*sk
, int flags
, int err
);
885 extern void sk_stream_kill_queues(struct sock
*sk
);
886 extern void sk_set_memalloc(struct sock
*sk
);
887 extern void sk_clear_memalloc(struct sock
*sk
);
889 extern int sk_wait_data(struct sock
*sk
, long *timeo
);
891 struct request_sock_ops
;
892 struct timewait_sock_ops
;
893 struct inet_hashinfo
;
898 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
899 * un-modified. Special care is taken when initializing object to zero.
901 static inline void sk_prot_clear_nulls(struct sock
*sk
, int size
)
903 if (offsetof(struct sock
, sk_node
.next
) != 0)
904 memset(sk
, 0, offsetof(struct sock
, sk_node
.next
));
905 memset(&sk
->sk_node
.pprev
, 0,
906 size
- offsetof(struct sock
, sk_node
.pprev
));
909 /* Networking protocol blocks we attach to sockets.
910 * socket layer -> transport layer interface
911 * transport -> network interface is defined by struct inet_proto
914 void (*close
)(struct sock
*sk
,
916 int (*connect
)(struct sock
*sk
,
917 struct sockaddr
*uaddr
,
919 int (*disconnect
)(struct sock
*sk
, int flags
);
921 struct sock
* (*accept
)(struct sock
*sk
, int flags
, int *err
);
923 int (*ioctl
)(struct sock
*sk
, int cmd
,
925 int (*init
)(struct sock
*sk
);
926 void (*destroy
)(struct sock
*sk
);
927 void (*shutdown
)(struct sock
*sk
, int how
);
928 int (*setsockopt
)(struct sock
*sk
, int level
,
929 int optname
, char __user
*optval
,
930 unsigned int optlen
);
931 int (*getsockopt
)(struct sock
*sk
, int level
,
932 int optname
, char __user
*optval
,
935 int (*compat_setsockopt
)(struct sock
*sk
,
937 int optname
, char __user
*optval
,
938 unsigned int optlen
);
939 int (*compat_getsockopt
)(struct sock
*sk
,
941 int optname
, char __user
*optval
,
943 int (*compat_ioctl
)(struct sock
*sk
,
944 unsigned int cmd
, unsigned long arg
);
946 int (*sendmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
947 struct msghdr
*msg
, size_t len
);
948 int (*recvmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
950 size_t len
, int noblock
, int flags
,
952 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
953 int offset
, size_t size
, int flags
);
954 int (*bind
)(struct sock
*sk
,
955 struct sockaddr
*uaddr
, int addr_len
);
957 int (*backlog_rcv
) (struct sock
*sk
,
958 struct sk_buff
*skb
);
960 void (*release_cb
)(struct sock
*sk
);
962 /* Keeping track of sk's, looking them up, and port selection methods. */
963 void (*hash
)(struct sock
*sk
);
964 void (*unhash
)(struct sock
*sk
);
965 void (*rehash
)(struct sock
*sk
);
966 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
967 void (*clear_sk
)(struct sock
*sk
, int size
);
969 /* Keeping track of sockets in use */
970 #ifdef CONFIG_PROC_FS
971 unsigned int inuse_idx
;
974 /* Memory pressure */
975 void (*enter_memory_pressure
)(struct sock
*sk
);
976 atomic_long_t
*memory_allocated
; /* Current allocated memory. */
977 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
979 * Pressure flag: try to collapse.
980 * Technical note: it is used by multiple contexts non atomically.
981 * All the __sk_mem_schedule() is of this nature: accounting
982 * is strict, actions are advisory and have some latency.
984 int *memory_pressure
;
991 struct kmem_cache
*slab
;
992 unsigned int obj_size
;
995 struct percpu_counter
*orphan_count
;
997 struct request_sock_ops
*rsk_prot
;
998 struct timewait_sock_ops
*twsk_prot
;
1001 struct inet_hashinfo
*hashinfo
;
1002 struct udp_table
*udp_table
;
1003 struct raw_hashinfo
*raw_hash
;
1006 struct module
*owner
;
1010 struct list_head node
;
1011 #ifdef SOCK_REFCNT_DEBUG
1014 #ifdef CONFIG_MEMCG_KMEM
1016 * cgroup specific init/deinit functions. Called once for all
1017 * protocols that implement it, from cgroups populate function.
1018 * This function has to setup any files the protocol want to
1019 * appear in the kmem cgroup filesystem.
1021 int (*init_cgroup
)(struct mem_cgroup
*memcg
,
1022 struct cgroup_subsys
*ss
);
1023 void (*destroy_cgroup
)(struct mem_cgroup
*memcg
);
1024 struct cg_proto
*(*proto_cgroup
)(struct mem_cgroup
*memcg
);
1026 int (*diag_destroy
)(struct sock
*sk
, int err
);
1030 * Bits in struct cg_proto.flags
1032 enum cg_proto_flags
{
1033 /* Currently active and new sockets should be assigned to cgroups */
1035 /* It was ever activated; we must disarm static keys on destruction */
1036 MEMCG_SOCK_ACTIVATED
,
1040 void (*enter_memory_pressure
)(struct sock
*sk
);
1041 struct res_counter
*memory_allocated
; /* Current allocated memory. */
1042 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
1043 int *memory_pressure
;
1045 unsigned long flags
;
1047 * memcg field is used to find which memcg we belong directly
1048 * Each memcg struct can hold more than one cg_proto, so container_of
1051 * The elegant solution would be having an inverse function to
1052 * proto_cgroup in struct proto, but that means polluting the structure
1053 * for everybody, instead of just for memcg users.
1055 struct mem_cgroup
*memcg
;
1058 extern int proto_register(struct proto
*prot
, int alloc_slab
);
1059 extern void proto_unregister(struct proto
*prot
);
1061 static inline bool memcg_proto_active(struct cg_proto
*cg_proto
)
1063 return test_bit(MEMCG_SOCK_ACTIVE
, &cg_proto
->flags
);
1066 static inline bool memcg_proto_activated(struct cg_proto
*cg_proto
)
1068 return test_bit(MEMCG_SOCK_ACTIVATED
, &cg_proto
->flags
);
1071 #ifdef SOCK_REFCNT_DEBUG
1072 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
1074 atomic_inc(&sk
->sk_prot
->socks
);
1077 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
1079 atomic_dec(&sk
->sk_prot
->socks
);
1080 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
1081 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
1084 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
1086 if (atomic_read(&sk
->sk_refcnt
) != 1)
1087 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
1088 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
1090 #else /* SOCK_REFCNT_DEBUG */
1091 #define sk_refcnt_debug_inc(sk) do { } while (0)
1092 #define sk_refcnt_debug_dec(sk) do { } while (0)
1093 #define sk_refcnt_debug_release(sk) do { } while (0)
1094 #endif /* SOCK_REFCNT_DEBUG */
1096 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
1097 extern struct static_key memcg_socket_limit_enabled
;
1098 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
1099 struct cg_proto
*cg_proto
)
1101 return proto
->proto_cgroup(parent_mem_cgroup(cg_proto
->memcg
));
1103 #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
1105 #define mem_cgroup_sockets_enabled 0
1106 static inline struct cg_proto
*parent_cg_proto(struct proto
*proto
,
1107 struct cg_proto
*cg_proto
)
1114 static inline bool sk_has_memory_pressure(const struct sock
*sk
)
1116 return sk
->sk_prot
->memory_pressure
!= NULL
;
1119 static inline bool sk_under_memory_pressure(const struct sock
*sk
)
1121 if (!sk
->sk_prot
->memory_pressure
)
1124 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1125 return !!*sk
->sk_cgrp
->memory_pressure
;
1127 return !!*sk
->sk_prot
->memory_pressure
;
1130 static inline void sk_leave_memory_pressure(struct sock
*sk
)
1132 int *memory_pressure
= sk
->sk_prot
->memory_pressure
;
1134 if (!memory_pressure
)
1137 if (*memory_pressure
)
1138 *memory_pressure
= 0;
1140 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1141 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1142 struct proto
*prot
= sk
->sk_prot
;
1144 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1145 if (*cg_proto
->memory_pressure
)
1146 *cg_proto
->memory_pressure
= 0;
1151 static inline void sk_enter_memory_pressure(struct sock
*sk
)
1153 if (!sk
->sk_prot
->enter_memory_pressure
)
1156 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1157 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1158 struct proto
*prot
= sk
->sk_prot
;
1160 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1161 cg_proto
->enter_memory_pressure(sk
);
1164 sk
->sk_prot
->enter_memory_pressure(sk
);
1167 static inline long sk_prot_mem_limits(const struct sock
*sk
, int index
)
1169 long *prot
= sk
->sk_prot
->sysctl_mem
;
1170 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1171 prot
= sk
->sk_cgrp
->sysctl_mem
;
1175 static inline void memcg_memory_allocated_add(struct cg_proto
*prot
,
1179 struct res_counter
*fail
;
1182 ret
= res_counter_charge_nofail(prot
->memory_allocated
,
1183 amt
<< PAGE_SHIFT
, &fail
);
1185 *parent_status
= OVER_LIMIT
;
1188 static inline void memcg_memory_allocated_sub(struct cg_proto
*prot
,
1191 res_counter_uncharge(prot
->memory_allocated
, amt
<< PAGE_SHIFT
);
1194 static inline u64
memcg_memory_allocated_read(struct cg_proto
*prot
)
1197 ret
= res_counter_read_u64(prot
->memory_allocated
, RES_USAGE
);
1198 return ret
>> PAGE_SHIFT
;
1202 sk_memory_allocated(const struct sock
*sk
)
1204 struct proto
*prot
= sk
->sk_prot
;
1205 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1206 return memcg_memory_allocated_read(sk
->sk_cgrp
);
1208 return atomic_long_read(prot
->memory_allocated
);
1212 sk_memory_allocated_add(struct sock
*sk
, int amt
, int *parent_status
)
1214 struct proto
*prot
= sk
->sk_prot
;
1216 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1217 memcg_memory_allocated_add(sk
->sk_cgrp
, amt
, parent_status
);
1218 /* update the root cgroup regardless */
1219 atomic_long_add_return(amt
, prot
->memory_allocated
);
1220 return memcg_memory_allocated_read(sk
->sk_cgrp
);
1223 return atomic_long_add_return(amt
, prot
->memory_allocated
);
1227 sk_memory_allocated_sub(struct sock
*sk
, int amt
)
1229 struct proto
*prot
= sk
->sk_prot
;
1231 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1232 memcg_memory_allocated_sub(sk
->sk_cgrp
, amt
);
1234 atomic_long_sub(amt
, prot
->memory_allocated
);
1237 static inline void sk_sockets_allocated_dec(struct sock
*sk
)
1239 struct proto
*prot
= sk
->sk_prot
;
1241 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1242 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1244 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1245 percpu_counter_dec(cg_proto
->sockets_allocated
);
1248 percpu_counter_dec(prot
->sockets_allocated
);
1251 static inline void sk_sockets_allocated_inc(struct sock
*sk
)
1253 struct proto
*prot
= sk
->sk_prot
;
1255 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
) {
1256 struct cg_proto
*cg_proto
= sk
->sk_cgrp
;
1258 for (; cg_proto
; cg_proto
= parent_cg_proto(prot
, cg_proto
))
1259 percpu_counter_inc(cg_proto
->sockets_allocated
);
1262 percpu_counter_inc(prot
->sockets_allocated
);
1266 sk_sockets_allocated_read_positive(struct sock
*sk
)
1268 struct proto
*prot
= sk
->sk_prot
;
1270 if (mem_cgroup_sockets_enabled
&& sk
->sk_cgrp
)
1271 return percpu_counter_read_positive(sk
->sk_cgrp
->sockets_allocated
);
1273 return percpu_counter_read_positive(prot
->sockets_allocated
);
1277 proto_sockets_allocated_sum_positive(struct proto
*prot
)
1279 return percpu_counter_sum_positive(prot
->sockets_allocated
);
1283 proto_memory_allocated(struct proto
*prot
)
1285 return atomic_long_read(prot
->memory_allocated
);
1289 proto_memory_pressure(struct proto
*prot
)
1291 if (!prot
->memory_pressure
)
1293 return !!*prot
->memory_pressure
;
1297 #ifdef CONFIG_PROC_FS
1298 /* Called with local bh disabled */
1299 extern void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int inc
);
1300 extern int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
1302 static inline void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
1309 /* With per-bucket locks this operation is not-atomic, so that
1310 * this version is not worse.
1312 static inline void __sk_prot_rehash(struct sock
*sk
)
1314 sk
->sk_prot
->unhash(sk
);
1315 sk
->sk_prot
->hash(sk
);
1318 void sk_prot_clear_portaddr_nulls(struct sock
*sk
, int size
);
1320 /* About 10 seconds */
1321 #define SOCK_DESTROY_TIME (10*HZ)
1323 /* Sockets 0-1023 can't be bound to unless you are superuser */
1324 #define PROT_SOCK 1024
1326 #define SHUTDOWN_MASK 3
1327 #define RCV_SHUTDOWN 1
1328 #define SEND_SHUTDOWN 2
1330 #define SOCK_SNDBUF_LOCK 1
1331 #define SOCK_RCVBUF_LOCK 2
1332 #define SOCK_BINDADDR_LOCK 4
1333 #define SOCK_BINDPORT_LOCK 8
1335 /* sock_iocb: used to kick off async processing of socket ios */
1337 struct list_head list
;
1341 struct socket
*sock
;
1343 struct scm_cookie
*scm
;
1344 struct msghdr
*msg
, async_msg
;
1345 struct kiocb
*kiocb
;
1348 static inline struct sock_iocb
*kiocb_to_siocb(struct kiocb
*iocb
)
1350 return (struct sock_iocb
*)iocb
->private;
1353 static inline struct kiocb
*siocb_to_kiocb(struct sock_iocb
*si
)
1358 struct socket_alloc
{
1359 struct socket socket
;
1360 struct inode vfs_inode
;
1363 static inline struct socket
*SOCKET_I(struct inode
*inode
)
1365 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
1368 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
1370 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
1374 * Functions for memory accounting
1376 extern int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
1377 void __sk_mem_reclaim(struct sock
*sk
, int amount
);
1379 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1380 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1381 #define SK_MEM_SEND 0
1382 #define SK_MEM_RECV 1
1384 static inline int sk_mem_pages(int amt
)
1386 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
1389 static inline bool sk_has_account(struct sock
*sk
)
1391 /* return true if protocol supports memory accounting */
1392 return !!sk
->sk_prot
->memory_allocated
;
1395 static inline bool sk_wmem_schedule(struct sock
*sk
, int size
)
1397 if (!sk_has_account(sk
))
1399 return size
<= sk
->sk_forward_alloc
||
1400 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
1404 sk_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
, int size
)
1406 if (!sk_has_account(sk
))
1408 return size
<= sk
->sk_forward_alloc
||
1409 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
) ||
1410 skb_pfmemalloc(skb
);
1413 static inline void sk_mem_reclaim(struct sock
*sk
)
1415 if (!sk_has_account(sk
))
1417 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
1418 __sk_mem_reclaim(sk
, sk
->sk_forward_alloc
);
1421 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
1423 if (!sk_has_account(sk
))
1425 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
1426 __sk_mem_reclaim(sk
, sk
->sk_forward_alloc
- 1);
1429 static inline void sk_mem_charge(struct sock
*sk
, int size
)
1431 if (!sk_has_account(sk
))
1433 sk
->sk_forward_alloc
-= size
;
1436 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
1438 if (!sk_has_account(sk
))
1440 sk
->sk_forward_alloc
+= size
;
1442 /* Avoid a possible overflow.
1443 * TCP send queues can make this happen, if sk_mem_reclaim()
1444 * is not called and more than 2 GBytes are released at once.
1446 * If we reach 2 MBytes, reclaim 1 MBytes right now, there is
1447 * no need to hold that much forward allocation anyway.
1449 if (unlikely(sk
->sk_forward_alloc
>= 1 << 21))
1450 __sk_mem_reclaim(sk
, 1 << 20);
1453 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
1455 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
1456 sk
->sk_wmem_queued
-= skb
->truesize
;
1457 sk_mem_uncharge(sk
, skb
->truesize
);
1461 /* Used by processes to "lock" a socket state, so that
1462 * interrupts and bottom half handlers won't change it
1463 * from under us. It essentially blocks any incoming
1464 * packets, so that we won't get any new data or any
1465 * packets that change the state of the socket.
1467 * While locked, BH processing will add new packets to
1468 * the backlog queue. This queue is processed by the
1469 * owner of the socket lock right before it is released.
1471 * Since ~2.3.5 it is also exclusive sleep lock serializing
1472 * accesses from user process context.
1474 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1476 static inline void sock_release_ownership(struct sock
*sk
)
1478 sk
->sk_lock
.owned
= 0;
1482 * Macro so as to not evaluate some arguments when
1483 * lockdep is not enabled.
1485 * Mark both the sk_lock and the sk_lock.slock as a
1486 * per-address-family lock class.
1488 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1490 sk->sk_lock.owned = 0; \
1491 init_waitqueue_head(&sk->sk_lock.wq); \
1492 spin_lock_init(&(sk)->sk_lock.slock); \
1493 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1494 sizeof((sk)->sk_lock)); \
1495 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1497 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1500 extern void lock_sock_nested(struct sock
*sk
, int subclass
);
1502 static inline void lock_sock(struct sock
*sk
)
1504 lock_sock_nested(sk
, 0);
1507 extern void release_sock(struct sock
*sk
);
1509 /* BH context may only use the following locking interface. */
1510 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1511 #define bh_lock_sock_nested(__sk) \
1512 spin_lock_nested(&((__sk)->sk_lock.slock), \
1513 SINGLE_DEPTH_NESTING)
1514 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1516 extern bool lock_sock_fast(struct sock
*sk
);
1518 * unlock_sock_fast - complement of lock_sock_fast
1522 * fast unlock socket for user context.
1523 * If slow mode is on, we call regular release_sock()
1525 static inline void unlock_sock_fast(struct sock
*sk
, bool slow
)
1530 spin_unlock_bh(&sk
->sk_lock
.slock
);
1534 extern struct sock
*sk_alloc(struct net
*net
, int family
,
1536 struct proto
*prot
);
1537 extern void sk_free(struct sock
*sk
);
1538 extern void sk_release_kernel(struct sock
*sk
);
1539 extern struct sock
*sk_clone_lock(const struct sock
*sk
,
1540 const gfp_t priority
);
1542 extern struct sk_buff
*sock_wmalloc(struct sock
*sk
,
1543 unsigned long size
, int force
,
1545 extern struct sk_buff
*sock_rmalloc(struct sock
*sk
,
1546 unsigned long size
, int force
,
1548 extern void sock_wfree(struct sk_buff
*skb
);
1549 extern void sock_rfree(struct sk_buff
*skb
);
1550 extern void sock_edemux(struct sk_buff
*skb
);
1552 extern int sock_setsockopt(struct socket
*sock
, int level
,
1553 int op
, char __user
*optval
,
1554 unsigned int optlen
);
1556 extern int sock_getsockopt(struct socket
*sock
, int level
,
1557 int op
, char __user
*optval
,
1558 int __user
*optlen
);
1559 extern struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
,
1563 extern struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
,
1564 unsigned long header_len
,
1565 unsigned long data_len
,
1568 extern void *sock_kmalloc(struct sock
*sk
, int size
,
1570 extern void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
1571 extern void sk_send_sigurg(struct sock
*sk
);
1574 * Functions to fill in entries in struct proto_ops when a protocol
1575 * does not implement a particular function.
1577 extern int sock_no_bind(struct socket
*,
1578 struct sockaddr
*, int);
1579 extern int sock_no_connect(struct socket
*,
1580 struct sockaddr
*, int, int);
1581 extern int sock_no_socketpair(struct socket
*,
1583 extern int sock_no_accept(struct socket
*,
1584 struct socket
*, int);
1585 extern int sock_no_getname(struct socket
*,
1586 struct sockaddr
*, int *, int);
1587 extern unsigned int sock_no_poll(struct file
*, struct socket
*,
1588 struct poll_table_struct
*);
1589 extern int sock_no_ioctl(struct socket
*, unsigned int,
1591 extern int sock_no_listen(struct socket
*, int);
1592 extern int sock_no_shutdown(struct socket
*, int);
1593 extern int sock_no_getsockopt(struct socket
*, int , int,
1594 char __user
*, int __user
*);
1595 extern int sock_no_setsockopt(struct socket
*, int, int,
1596 char __user
*, unsigned int);
1597 extern int sock_no_sendmsg(struct kiocb
*, struct socket
*,
1598 struct msghdr
*, size_t);
1599 extern int sock_no_recvmsg(struct kiocb
*, struct socket
*,
1600 struct msghdr
*, size_t, int);
1601 extern int sock_no_mmap(struct file
*file
,
1602 struct socket
*sock
,
1603 struct vm_area_struct
*vma
);
1604 extern ssize_t
sock_no_sendpage(struct socket
*sock
,
1606 int offset
, size_t size
,
1610 * Functions to fill in entries in struct proto_ops when a protocol
1611 * uses the inet style.
1613 extern int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1614 char __user
*optval
, int __user
*optlen
);
1615 extern int sock_common_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
1616 struct msghdr
*msg
, size_t size
, int flags
);
1617 extern int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1618 char __user
*optval
, unsigned int optlen
);
1619 extern int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
1620 int optname
, char __user
*optval
, int __user
*optlen
);
1621 extern int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
1622 int optname
, char __user
*optval
, unsigned int optlen
);
1624 extern void sk_common_release(struct sock
*sk
);
1627 * Default socket callbacks and setup code
1630 /* Initialise core socket variables */
1631 extern void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1633 extern void sk_filter_release_rcu(struct rcu_head
*rcu
);
1636 * sk_filter_release - release a socket filter
1637 * @fp: filter to remove
1639 * Remove a filter from a socket and release its resources.
1642 static inline void sk_filter_release(struct sk_filter
*fp
)
1644 if (atomic_dec_and_test(&fp
->refcnt
))
1645 call_rcu(&fp
->rcu
, sk_filter_release_rcu
);
1648 static inline void sk_filter_uncharge(struct sock
*sk
, struct sk_filter
*fp
)
1650 unsigned int size
= sk_filter_len(fp
);
1652 atomic_sub(size
, &sk
->sk_omem_alloc
);
1653 sk_filter_release(fp
);
1656 static inline void sk_filter_charge(struct sock
*sk
, struct sk_filter
*fp
)
1658 atomic_inc(&fp
->refcnt
);
1659 atomic_add(sk_filter_len(fp
), &sk
->sk_omem_alloc
);
1663 * Socket reference counting postulates.
1665 * * Each user of socket SHOULD hold a reference count.
1666 * * Each access point to socket (an hash table bucket, reference from a list,
1667 * running timer, skb in flight MUST hold a reference count.
1668 * * When reference count hits 0, it means it will never increase back.
1669 * * When reference count hits 0, it means that no references from
1670 * outside exist to this socket and current process on current CPU
1671 * is last user and may/should destroy this socket.
1672 * * sk_free is called from any context: process, BH, IRQ. When
1673 * it is called, socket has no references from outside -> sk_free
1674 * may release descendant resources allocated by the socket, but
1675 * to the time when it is called, socket is NOT referenced by any
1676 * hash tables, lists etc.
1677 * * Packets, delivered from outside (from network or from another process)
1678 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1679 * when they sit in queue. Otherwise, packets will leak to hole, when
1680 * socket is looked up by one cpu and unhasing is made by another CPU.
1681 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1682 * (leak to backlog). Packet socket does all the processing inside
1683 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1684 * use separate SMP lock, so that they are prone too.
1687 /* Ungrab socket and destroy it, if it was the last reference. */
1688 static inline void sock_put(struct sock
*sk
)
1690 if (atomic_dec_and_test(&sk
->sk_refcnt
))
1694 extern int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
1697 static inline void sk_tx_queue_set(struct sock
*sk
, int tx_queue
)
1699 sk
->sk_tx_queue_mapping
= tx_queue
;
1702 static inline void sk_tx_queue_clear(struct sock
*sk
)
1704 sk
->sk_tx_queue_mapping
= -1;
1707 static inline int sk_tx_queue_get(const struct sock
*sk
)
1709 return sk
? sk
->sk_tx_queue_mapping
: -1;
1712 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1714 sk_tx_queue_clear(sk
);
1715 sk
->sk_socket
= sock
;
1718 static inline wait_queue_head_t
*sk_sleep(struct sock
*sk
)
1720 BUILD_BUG_ON(offsetof(struct socket_wq
, wait
) != 0);
1721 return &rcu_dereference_raw(sk
->sk_wq
)->wait
;
1723 /* Detach socket from process context.
1724 * Announce socket dead, detach it from wait queue and inode.
1725 * Note that parent inode held reference count on this struct sock,
1726 * we do not release it in this function, because protocol
1727 * probably wants some additional cleanups or even continuing
1728 * to work with this socket (TCP).
1730 static inline void sock_orphan(struct sock
*sk
)
1732 write_lock_bh(&sk
->sk_callback_lock
);
1733 sock_set_flag(sk
, SOCK_DEAD
);
1734 sk_set_socket(sk
, NULL
);
1736 write_unlock_bh(&sk
->sk_callback_lock
);
1739 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1741 write_lock_bh(&sk
->sk_callback_lock
);
1742 sk
->sk_wq
= parent
->wq
;
1744 sk_set_socket(sk
, parent
);
1745 sk
->sk_uid
= SOCK_INODE(parent
)->i_uid
;
1746 security_sock_graft(sk
, parent
);
1747 write_unlock_bh(&sk
->sk_callback_lock
);
1750 extern kuid_t
sock_i_uid(struct sock
*sk
);
1751 extern unsigned long sock_i_ino(struct sock
*sk
);
1753 static inline kuid_t
sock_net_uid(const struct net
*net
, const struct sock
*sk
)
1755 return sk
? sk
->sk_uid
: make_kuid(net
->user_ns
, 0);
1758 static inline struct dst_entry
*
1759 __sk_dst_get(struct sock
*sk
)
1761 return rcu_dereference_check(sk
->sk_dst_cache
, sock_owned_by_user(sk
) ||
1762 lockdep_is_held(&sk
->sk_lock
.slock
));
1765 static inline struct dst_entry
*
1766 sk_dst_get(struct sock
*sk
)
1768 struct dst_entry
*dst
;
1771 dst
= rcu_dereference(sk
->sk_dst_cache
);
1772 if (dst
&& !atomic_inc_not_zero(&dst
->__refcnt
))
1778 extern void sk_reset_txq(struct sock
*sk
);
1780 static inline void dst_negative_advice(struct sock
*sk
)
1782 struct dst_entry
*ndst
, *dst
= __sk_dst_get(sk
);
1784 if (dst
&& dst
->ops
->negative_advice
) {
1785 ndst
= dst
->ops
->negative_advice(dst
);
1788 rcu_assign_pointer(sk
->sk_dst_cache
, ndst
);
1795 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1797 struct dst_entry
*old_dst
;
1799 sk_tx_queue_clear(sk
);
1801 * This can be called while sk is owned by the caller only,
1802 * with no state that can be checked in a rcu_dereference_check() cond
1804 old_dst
= rcu_dereference_raw(sk
->sk_dst_cache
);
1805 rcu_assign_pointer(sk
->sk_dst_cache
, dst
);
1806 dst_release(old_dst
);
1810 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1812 struct dst_entry
*old_dst
;
1814 sk_tx_queue_clear(sk
);
1815 old_dst
= xchg((__force
struct dst_entry
**)&sk
->sk_dst_cache
, dst
);
1816 dst_release(old_dst
);
1820 __sk_dst_reset(struct sock
*sk
)
1822 __sk_dst_set(sk
, NULL
);
1826 sk_dst_reset(struct sock
*sk
)
1828 sk_dst_set(sk
, NULL
);
1831 extern struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1833 extern struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1835 static inline bool sk_can_gso(const struct sock
*sk
)
1837 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1840 extern void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1842 static inline void sk_nocaps_add(struct sock
*sk
, netdev_features_t flags
)
1844 sk
->sk_route_nocaps
|= flags
;
1845 sk
->sk_route_caps
&= ~flags
;
1848 static inline int skb_do_copy_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1849 char __user
*from
, char *to
,
1850 int copy
, int offset
)
1852 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1854 __wsum csum
= csum_and_copy_from_user(from
, to
, copy
, 0, &err
);
1857 skb
->csum
= csum_block_add(skb
->csum
, csum
, offset
);
1858 } else if (sk
->sk_route_caps
& NETIF_F_NOCACHE_COPY
) {
1859 if (!access_ok(VERIFY_READ
, from
, copy
) ||
1860 __copy_from_user_nocache(to
, from
, copy
))
1862 } else if (copy_from_user(to
, from
, copy
))
1868 static inline int skb_add_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1869 char __user
*from
, int copy
)
1871 int err
, offset
= skb
->len
;
1873 err
= skb_do_copy_data_nocache(sk
, skb
, from
, skb_put(skb
, copy
),
1876 __skb_trim(skb
, offset
);
1881 static inline int skb_copy_to_page_nocache(struct sock
*sk
, char __user
*from
,
1882 struct sk_buff
*skb
,
1888 err
= skb_do_copy_data_nocache(sk
, skb
, from
, page_address(page
) + off
,
1894 skb
->data_len
+= copy
;
1895 skb
->truesize
+= copy
;
1896 sk
->sk_wmem_queued
+= copy
;
1897 sk_mem_charge(sk
, copy
);
1901 static inline int skb_copy_to_page(struct sock
*sk
, char __user
*from
,
1902 struct sk_buff
*skb
, struct page
*page
,
1905 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1907 __wsum csum
= csum_and_copy_from_user(from
,
1908 page_address(page
) + off
,
1912 skb
->csum
= csum_block_add(skb
->csum
, csum
, skb
->len
);
1913 } else if (copy_from_user(page_address(page
) + off
, from
, copy
))
1917 skb
->data_len
+= copy
;
1918 skb
->truesize
+= copy
;
1919 sk
->sk_wmem_queued
+= copy
;
1920 sk_mem_charge(sk
, copy
);
1925 * sk_wmem_alloc_get - returns write allocations
1928 * Returns sk_wmem_alloc minus initial offset of one
1930 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
1932 return atomic_read(&sk
->sk_wmem_alloc
) - 1;
1936 * sk_rmem_alloc_get - returns read allocations
1939 * Returns sk_rmem_alloc
1941 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
1943 return atomic_read(&sk
->sk_rmem_alloc
);
1947 * sk_has_allocations - check if allocations are outstanding
1950 * Returns true if socket has write or read allocations
1952 static inline bool sk_has_allocations(const struct sock
*sk
)
1954 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
1958 * wq_has_sleeper - check if there are any waiting processes
1959 * @wq: struct socket_wq
1961 * Returns true if socket_wq has waiting processes
1963 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1964 * barrier call. They were added due to the race found within the tcp code.
1966 * Consider following tcp code paths:
1970 * sys_select receive packet
1972 * __add_wait_queue update tp->rcv_nxt
1974 * tp->rcv_nxt check sock_def_readable
1976 * schedule rcu_read_lock();
1977 * wq = rcu_dereference(sk->sk_wq);
1978 * if (wq && waitqueue_active(&wq->wait))
1979 * wake_up_interruptible(&wq->wait)
1983 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1984 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1985 * could then endup calling schedule and sleep forever if there are no more
1986 * data on the socket.
1989 static inline bool wq_has_sleeper(struct socket_wq
*wq
)
1991 /* We need to be sure we are in sync with the
1992 * add_wait_queue modifications to the wait queue.
1994 * This memory barrier is paired in the sock_poll_wait.
1997 return wq
&& waitqueue_active(&wq
->wait
);
2001 * sock_poll_wait - place memory barrier behind the poll_wait call.
2003 * @wait_address: socket wait queue
2006 * See the comments in the wq_has_sleeper function.
2008 static inline void sock_poll_wait(struct file
*filp
,
2009 wait_queue_head_t
*wait_address
, poll_table
*p
)
2011 if (!poll_does_not_wait(p
) && wait_address
) {
2012 poll_wait(filp
, wait_address
, p
);
2013 /* We need to be sure we are in sync with the
2014 * socket flags modification.
2016 * This memory barrier is paired in the wq_has_sleeper.
2023 * Queue a received datagram if it will fit. Stream and sequenced
2024 * protocols can't normally use this as they need to fit buffers in
2025 * and play with them.
2027 * Inlined as it's very short and called for pretty much every
2028 * packet ever received.
2031 static inline void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
2035 skb
->destructor
= sock_wfree
;
2037 * We used to take a refcount on sk, but following operation
2038 * is enough to guarantee sk_free() wont free this sock until
2039 * all in-flight packets are completed
2041 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
2044 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
2048 skb
->destructor
= sock_rfree
;
2049 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
2050 sk_mem_charge(sk
, skb
->truesize
);
2053 extern void sk_reset_timer(struct sock
*sk
, struct timer_list
*timer
,
2054 unsigned long expires
);
2056 extern void sk_stop_timer(struct sock
*sk
, struct timer_list
*timer
);
2058 extern int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
2060 extern int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
);
2063 * Recover an error report and clear atomically
2066 static inline int sock_error(struct sock
*sk
)
2069 if (likely(!sk
->sk_err
))
2071 err
= xchg(&sk
->sk_err
, 0);
2075 static inline unsigned long sock_wspace(struct sock
*sk
)
2079 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
2080 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
2087 static inline void sk_wake_async(struct sock
*sk
, int how
, int band
)
2089 if (sock_flag(sk
, SOCK_FASYNC
))
2090 sock_wake_async(sk
->sk_socket
, how
, band
);
2093 #define SOCK_MIN_SNDBUF 2048
2095 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
2096 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
2098 #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
2100 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
2102 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
2103 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
2104 sk
->sk_sndbuf
= max(sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
2108 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
);
2111 * sk_page_frag - return an appropriate page_frag
2114 * If socket allocation mode allows current thread to sleep, it means its
2115 * safe to use the per task page_frag instead of the per socket one.
2117 static inline struct page_frag
*sk_page_frag(struct sock
*sk
)
2119 if (sk
->sk_allocation
& __GFP_WAIT
)
2120 return ¤t
->task_frag
;
2122 return &sk
->sk_frag
;
2125 extern bool sk_page_frag_refill(struct sock
*sk
, struct page_frag
*pfrag
);
2128 * Default write policy as shown to user space via poll/select/SIGIO
2130 static inline bool sock_writeable(const struct sock
*sk
)
2132 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
2135 static inline gfp_t
gfp_any(void)
2137 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
2140 static inline long sock_rcvtimeo(const struct sock
*sk
, bool noblock
)
2142 return noblock
? 0 : sk
->sk_rcvtimeo
;
2145 static inline long sock_sndtimeo(const struct sock
*sk
, bool noblock
)
2147 return noblock
? 0 : sk
->sk_sndtimeo
;
2150 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
2152 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
2155 /* Alas, with timeout socket operations are not restartable.
2156 * Compare this to poll().
2158 static inline int sock_intr_errno(long timeo
)
2160 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
2163 extern void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
2164 struct sk_buff
*skb
);
2165 extern void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
2166 struct sk_buff
*skb
);
2169 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
2171 ktime_t kt
= skb
->tstamp
;
2172 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
2175 * generate control messages if
2176 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2177 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2178 * - software time stamp available and wanted
2179 * (SOCK_TIMESTAMPING_SOFTWARE)
2180 * - hardware time stamps available and wanted
2181 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2182 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2184 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
2185 sock_flag(sk
, SOCK_TIMESTAMPING_RX_SOFTWARE
) ||
2186 (kt
.tv64
&& sock_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
)) ||
2187 (hwtstamps
->hwtstamp
.tv64
&&
2188 sock_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
)) ||
2189 (hwtstamps
->syststamp
.tv64
&&
2190 sock_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
)))
2191 __sock_recv_timestamp(msg
, sk
, skb
);
2195 if (sock_flag(sk
, SOCK_WIFI_STATUS
) && skb
->wifi_acked_valid
)
2196 __sock_recv_wifi_status(msg
, sk
, skb
);
2199 extern void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2200 struct sk_buff
*skb
);
2202 static inline void sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
2203 struct sk_buff
*skb
)
2205 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2206 (1UL << SOCK_RCVTSTAMP) | \
2207 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2208 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
2209 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
2210 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2212 if (sk
->sk_flags
& FLAGS_TS_OR_DROPS
)
2213 __sock_recv_ts_and_drops(msg
, sk
, skb
);
2215 sk
->sk_stamp
= skb
->tstamp
;
2219 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2220 * @sk: socket sending this packet
2221 * @tx_flags: filled with instructions for time stamping
2223 * Currently only depends on SOCK_TIMESTAMPING* flags.
2225 extern void sock_tx_timestamp(struct sock
*sk
, __u8
*tx_flags
);
2228 * sk_eat_skb - Release a skb if it is no longer needed
2229 * @sk: socket to eat this skb from
2230 * @skb: socket buffer to eat
2231 * @copied_early: flag indicating whether DMA operations copied this data early
2233 * This routine must be called with interrupts disabled or with the socket
2234 * locked so that the sk_buff queue operation is ok.
2236 #ifdef CONFIG_NET_DMA
2237 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, bool copied_early
)
2239 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2243 __skb_queue_tail(&sk
->sk_async_wait_queue
, skb
);
2246 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, bool copied_early
)
2248 __skb_unlink(skb
, &sk
->sk_receive_queue
);
2254 struct net
*sock_net(const struct sock
*sk
)
2256 return read_pnet(&sk
->sk_net
);
2260 void sock_net_set(struct sock
*sk
, struct net
*net
)
2262 write_pnet(&sk
->sk_net
, net
);
2266 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
2267 * They should not hold a reference to a namespace in order to allow
2269 * Sockets after sk_change_net should be released using sk_release_kernel
2271 static inline void sk_change_net(struct sock
*sk
, struct net
*net
)
2273 put_net(sock_net(sk
));
2274 sock_net_set(sk
, hold_net(net
));
2277 static inline struct sock
*skb_steal_sock(struct sk_buff
*skb
)
2280 struct sock
*sk
= skb
->sk
;
2282 skb
->destructor
= NULL
;
2289 extern void sock_enable_timestamp(struct sock
*sk
, int flag
);
2290 extern int sock_get_timestamp(struct sock
*, struct timeval __user
*);
2291 extern int sock_get_timestampns(struct sock
*, struct timespec __user
*);
2293 bool sk_ns_capable(const struct sock
*sk
,
2294 struct user_namespace
*user_ns
, int cap
);
2295 bool sk_capable(const struct sock
*sk
, int cap
);
2296 bool sk_net_capable(const struct sock
*sk
, int cap
);
2299 * Enable debug/info messages
2301 extern int net_msg_warn
;
2302 #define NETDEBUG(fmt, args...) \
2303 do { if (net_msg_warn) printk(fmt,##args); } while (0)
2305 #define LIMIT_NETDEBUG(fmt, args...) \
2306 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
2308 extern __u32 sysctl_wmem_max
;
2309 extern __u32 sysctl_rmem_max
;
2311 extern int sysctl_optmem_max
;
2313 extern __u32 sysctl_wmem_default
;
2314 extern __u32 sysctl_rmem_default
;
2316 #endif /* _SOCK_H */