1 /*
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.
5 *
6 * Definitions for the AF_INET socket handler.
7 *
8 * Version: @(#)sock.h 1.0.4 05/13/93
9 *
10 * Authors: Ross Biro
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>
14 *
15 * Fixes:
16 * Alan Cox : Volatiles in skbuff pointers. See
17 * skbuff comments. May be overdone,
18 * better to prove they can be removed
19 * than the reverse.
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
33 *
34 *
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.
39 */
40 #ifndef _SOCK_H
41 #define _SOCK_H
42
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 */
53 #include <linux/mm.h>
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>
62
63 #include <linux/filter.h>
64 #include <linux/rculist_nulls.h>
65 #include <linux/poll.h>
66
67 #include <linux/atomic.h>
68 #include <net/dst.h>
69 #include <net/checksum.h>
70 #include <net/tcp_states.h>
71 #include <linux/net_tstamp.h>
72
73 struct cgroup;
74 struct cgroup_subsys;
75 #ifdef CONFIG_NET
76 int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss);
77 void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg);
78 #else
79 static inline
mem_cgroup_sockets_init(struct mem_cgroup * memcg,struct cgroup_subsys * ss)80 int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
81 {
82 return 0;
83 }
84 static inline
mem_cgroup_sockets_destroy(struct mem_cgroup * memcg)85 void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
86 {
87 }
88 #endif
89 /*
90 * This structure really needs to be cleaned up.
91 * Most of it is for TCP, and not used by any of
92 * the other protocols.
93 */
94
95 /* Define this to get the SOCK_DBG debugging facility. */
96 #define SOCK_DEBUGGING
97 #ifdef SOCK_DEBUGGING
98 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
99 printk(KERN_DEBUG msg); } while (0)
100 #else
101 /* Validate arguments and do nothing */
102 static inline __printf(2, 3)
SOCK_DEBUG(const struct sock * sk,const char * msg,...)103 void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
104 {
105 }
106 #endif
107
108 /* This is the per-socket lock. The spinlock provides a synchronization
109 * between user contexts and software interrupt processing, whereas the
110 * mini-semaphore synchronizes multiple users amongst themselves.
111 */
112 typedef struct {
113 spinlock_t slock;
114 int owned;
115 wait_queue_head_t wq;
116 /*
117 * We express the mutex-alike socket_lock semantics
118 * to the lock validator by explicitly managing
119 * the slock as a lock variant (in addition to
120 * the slock itself):
121 */
122 #ifdef CONFIG_DEBUG_LOCK_ALLOC
123 struct lockdep_map dep_map;
124 #endif
125 } socket_lock_t;
126
127 struct sock;
128 struct proto;
129 struct net;
130
131 typedef __u32 __bitwise __portpair;
132 typedef __u64 __bitwise __addrpair;
133
134 /**
135 * struct sock_common - minimal network layer representation of sockets
136 * @skc_daddr: Foreign IPv4 addr
137 * @skc_rcv_saddr: Bound local IPv4 addr
138 * @skc_hash: hash value used with various protocol lookup tables
139 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
140 * @skc_dport: placeholder for inet_dport/tw_dport
141 * @skc_num: placeholder for inet_num/tw_num
142 * @skc_family: network address family
143 * @skc_state: Connection state
144 * @skc_reuse: %SO_REUSEADDR setting
145 * @skc_reuseport: %SO_REUSEPORT setting
146 * @skc_bound_dev_if: bound device index if != 0
147 * @skc_bind_node: bind hash linkage for various protocol lookup tables
148 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
149 * @skc_prot: protocol handlers inside a network family
150 * @skc_net: reference to the network namespace of this socket
151 * @skc_node: main hash linkage for various protocol lookup tables
152 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
153 * @skc_tx_queue_mapping: tx queue number for this connection
154 * @skc_refcnt: reference count
155 *
156 * This is the minimal network layer representation of sockets, the header
157 * for struct sock and struct inet_timewait_sock.
158 */
159 struct sock_common {
160 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
161 * address on 64bit arches : cf INET_MATCH()
162 */
163 union {
164 __addrpair skc_addrpair;
165 struct {
166 __be32 skc_daddr;
167 __be32 skc_rcv_saddr;
168 };
169 };
170 union {
171 unsigned int skc_hash;
172 __u16 skc_u16hashes[2];
173 };
174 /* skc_dport && skc_num must be grouped as well */
175 union {
176 __portpair skc_portpair;
177 struct {
178 __be16 skc_dport;
179 __u16 skc_num;
180 };
181 };
182
183 unsigned short skc_family;
184 volatile unsigned char skc_state;
185 unsigned char skc_reuse:4;
186 unsigned char skc_reuseport:1;
187 unsigned char skc_ipv6only:1;
188 int skc_bound_dev_if;
189 union {
190 struct hlist_node skc_bind_node;
191 struct hlist_nulls_node skc_portaddr_node;
192 };
193 struct proto *skc_prot;
194 #ifdef CONFIG_NET_NS
195 struct net *skc_net;
196 #endif
197
198 #if IS_ENABLED(CONFIG_IPV6)
199 struct in6_addr skc_v6_daddr;
200 struct in6_addr skc_v6_rcv_saddr;
201 #endif
202
203 /*
204 * fields between dontcopy_begin/dontcopy_end
205 * are not copied in sock_copy()
206 */
207 /* private: */
208 int skc_dontcopy_begin[0];
209 /* public: */
210 union {
211 struct hlist_node skc_node;
212 struct hlist_nulls_node skc_nulls_node;
213 };
214 int skc_tx_queue_mapping;
215 atomic_t skc_refcnt;
216 /* private: */
217 int skc_dontcopy_end[0];
218 /* public: */
219 };
220
221 struct cg_proto;
222 /**
223 * struct sock - network layer representation of sockets
224 * @__sk_common: shared layout with inet_timewait_sock
225 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
226 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
227 * @sk_lock: synchronizer
228 * @sk_rcvbuf: size of receive buffer in bytes
229 * @sk_wq: sock wait queue and async head
230 * @sk_rx_dst: receive input route used by early demux
231 * @sk_dst_cache: destination cache
232 * @sk_dst_lock: destination cache lock
233 * @sk_policy: flow policy
234 * @sk_receive_queue: incoming packets
235 * @sk_wmem_alloc: transmit queue bytes committed
236 * @sk_write_queue: Packet sending queue
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
300 */
301 struct sock {
302 /*
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
305 */
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
311
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
332
333 socket_lock_t sk_lock;
334 struct sk_buff_head sk_receive_queue;
335 /*
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
341 * backlog.
342 */
343 struct {
344 atomic_t rmem_alloc;
345 int len;
346 struct sk_buff *head;
347 struct sk_buff *tail;
348 } sk_backlog;
349 #define sk_rmem_alloc sk_backlog.rmem_alloc
350 int sk_forward_alloc;
351 #ifdef CONFIG_RPS
352 __u32 sk_rxhash;
353 #endif
354 __u32 sk_txhash;
355 #ifdef CONFIG_NET_RX_BUSY_POLL
356 unsigned int sk_napi_id;
357 unsigned int sk_ll_usec;
358 #endif
359 atomic_t sk_drops;
360 int sk_rcvbuf;
361
362 struct sk_filter __rcu *sk_filter;
363 struct socket_wq __rcu *sk_wq;
364
365 #ifdef CONFIG_XFRM
366 struct xfrm_policy *sk_policy[2];
367 #endif
368 unsigned long sk_flags;
369 struct dst_entry *sk_rx_dst;
370 struct dst_entry __rcu *sk_dst_cache;
371 spinlock_t sk_dst_lock;
372 atomic_t sk_wmem_alloc;
373 atomic_t sk_omem_alloc;
374 int sk_sndbuf;
375 struct sk_buff_head sk_write_queue;
376 kmemcheck_bitfield_begin(flags);
377 unsigned int sk_shutdown : 2,
378 sk_no_check_tx : 1,
379 sk_no_check_rx : 1,
380 sk_userlocks : 4,
381 sk_protocol : 8,
382 #define SK_PROTOCOL_MAX U8_MAX
383 sk_type : 16;
384 kmemcheck_bitfield_end(flags);
385 int sk_wmem_queued;
386 gfp_t sk_allocation;
387 u32 sk_pacing_rate; /* bytes per second */
388 u32 sk_max_pacing_rate;
389 netdev_features_t sk_route_caps;
390 netdev_features_t sk_route_nocaps;
391 int sk_gso_type;
392 unsigned int sk_gso_max_size;
393 u16 sk_gso_max_segs;
394 int sk_rcvlowat;
395 unsigned long sk_lingertime;
396 struct sk_buff_head sk_error_queue;
397 struct proto *sk_prot_creator;
398 rwlock_t sk_callback_lock;
399 int sk_err,
400 sk_err_soft;
401 unsigned short sk_ack_backlog;
402 unsigned short sk_max_ack_backlog;
403 __u32 sk_priority;
404 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
405 __u32 sk_cgrp_prioidx;
406 #endif
407 struct pid *sk_peer_pid;
408 const struct cred *sk_peer_cred;
409 long sk_rcvtimeo;
410 long sk_sndtimeo;
411 void *sk_protinfo;
412 struct timer_list sk_timer;
413 ktime_t sk_stamp;
414 u16 sk_tsflags;
415 u32 sk_tskey;
416 struct socket *sk_socket;
417 void *sk_user_data;
418 struct page_frag sk_frag;
419 struct sk_buff *sk_send_head;
420 __s32 sk_peek_off;
421 int sk_write_pending;
422 #ifdef CONFIG_SECURITY
423 void *sk_security;
424 #endif
425 __u32 sk_mark;
426 kuid_t sk_uid;
427 u32 sk_classid;
428 struct cg_proto *sk_cgrp;
429 void (*sk_state_change)(struct sock *sk);
430 void (*sk_data_ready)(struct sock *sk);
431 void (*sk_write_space)(struct sock *sk);
432 void (*sk_error_report)(struct sock *sk);
433 int (*sk_backlog_rcv)(struct sock *sk,
434 struct sk_buff *skb);
435 void (*sk_destruct)(struct sock *sk);
436 };
437
438 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
439
440 #define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
441 #define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
442
443 /*
444 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
445 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
446 * on a socket means that the socket will reuse everybody else's port
447 * without looking at the other's sk_reuse value.
448 */
449
450 #define SK_NO_REUSE 0
451 #define SK_CAN_REUSE 1
452 #define SK_FORCE_REUSE 2
453
sk_peek_offset(struct sock * sk,int flags)454 static inline int sk_peek_offset(struct sock *sk, int flags)
455 {
456 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
457 return sk->sk_peek_off;
458 else
459 return 0;
460 }
461
sk_peek_offset_bwd(struct sock * sk,int val)462 static inline void sk_peek_offset_bwd(struct sock *sk, int val)
463 {
464 if (sk->sk_peek_off >= 0) {
465 if (sk->sk_peek_off >= val)
466 sk->sk_peek_off -= val;
467 else
468 sk->sk_peek_off = 0;
469 }
470 }
471
sk_peek_offset_fwd(struct sock * sk,int val)472 static inline void sk_peek_offset_fwd(struct sock *sk, int val)
473 {
474 if (sk->sk_peek_off >= 0)
475 sk->sk_peek_off += val;
476 }
477
478 /*
479 * Hashed lists helper routines
480 */
sk_entry(const struct hlist_node * node)481 static inline struct sock *sk_entry(const struct hlist_node *node)
482 {
483 return hlist_entry(node, struct sock, sk_node);
484 }
485
__sk_head(const struct hlist_head * head)486 static inline struct sock *__sk_head(const struct hlist_head *head)
487 {
488 return hlist_entry(head->first, struct sock, sk_node);
489 }
490
sk_head(const struct hlist_head * head)491 static inline struct sock *sk_head(const struct hlist_head *head)
492 {
493 return hlist_empty(head) ? NULL : __sk_head(head);
494 }
495
__sk_nulls_head(const struct hlist_nulls_head * head)496 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
497 {
498 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
499 }
500
sk_nulls_head(const struct hlist_nulls_head * head)501 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
502 {
503 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
504 }
505
sk_next(const struct sock * sk)506 static inline struct sock *sk_next(const struct sock *sk)
507 {
508 return sk->sk_node.next ?
509 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
510 }
511
sk_nulls_next(const struct sock * sk)512 static inline struct sock *sk_nulls_next(const struct sock *sk)
513 {
514 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
515 hlist_nulls_entry(sk->sk_nulls_node.next,
516 struct sock, sk_nulls_node) :
517 NULL;
518 }
519
sk_unhashed(const struct sock * sk)520 static inline bool sk_unhashed(const struct sock *sk)
521 {
522 return hlist_unhashed(&sk->sk_node);
523 }
524
sk_hashed(const struct sock * sk)525 static inline bool sk_hashed(const struct sock *sk)
526 {
527 return !sk_unhashed(sk);
528 }
529
sk_node_init(struct hlist_node * node)530 static inline void sk_node_init(struct hlist_node *node)
531 {
532 node->pprev = NULL;
533 }
534
sk_nulls_node_init(struct hlist_nulls_node * node)535 static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
536 {
537 node->pprev = NULL;
538 }
539
__sk_del_node(struct sock * sk)540 static inline void __sk_del_node(struct sock *sk)
541 {
542 __hlist_del(&sk->sk_node);
543 }
544
545 /* NB: equivalent to hlist_del_init_rcu */
__sk_del_node_init(struct sock * sk)546 static inline bool __sk_del_node_init(struct sock *sk)
547 {
548 if (sk_hashed(sk)) {
549 __sk_del_node(sk);
550 sk_node_init(&sk->sk_node);
551 return true;
552 }
553 return false;
554 }
555
556 /* Grab socket reference count. This operation is valid only
557 when sk is ALREADY grabbed f.e. it is found in hash table
558 or a list and the lookup is made under lock preventing hash table
559 modifications.
560 */
561
sock_hold(struct sock * sk)562 static inline void sock_hold(struct sock *sk)
563 {
564 atomic_inc(&sk->sk_refcnt);
565 }
566
567 /* Ungrab socket in the context, which assumes that socket refcnt
568 cannot hit zero, f.e. it is true in context of any socketcall.
569 */
__sock_put(struct sock * sk)570 static inline void __sock_put(struct sock *sk)
571 {
572 atomic_dec(&sk->sk_refcnt);
573 }
574
sk_del_node_init(struct sock * sk)575 static inline bool sk_del_node_init(struct sock *sk)
576 {
577 bool rc = __sk_del_node_init(sk);
578
579 if (rc) {
580 /* paranoid for a while -acme */
581 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
582 __sock_put(sk);
583 }
584 return rc;
585 }
586 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
587
__sk_nulls_del_node_init_rcu(struct sock * sk)588 static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
589 {
590 if (sk_hashed(sk)) {
591 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
592 return true;
593 }
594 return false;
595 }
596
sk_nulls_del_node_init_rcu(struct sock * sk)597 static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
598 {
599 bool rc = __sk_nulls_del_node_init_rcu(sk);
600
601 if (rc) {
602 /* paranoid for a while -acme */
603 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
604 __sock_put(sk);
605 }
606 return rc;
607 }
608
__sk_add_node(struct sock * sk,struct hlist_head * list)609 static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
610 {
611 hlist_add_head(&sk->sk_node, list);
612 }
613
sk_add_node(struct sock * sk,struct hlist_head * list)614 static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
615 {
616 sock_hold(sk);
617 __sk_add_node(sk, list);
618 }
619
sk_add_node_rcu(struct sock * sk,struct hlist_head * list)620 static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
621 {
622 sock_hold(sk);
623 hlist_add_head_rcu(&sk->sk_node, list);
624 }
625
__sk_nulls_add_node_rcu(struct sock * sk,struct hlist_nulls_head * list)626 static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
627 {
628 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
629 }
630
sk_nulls_add_node_rcu(struct sock * sk,struct hlist_nulls_head * list)631 static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
632 {
633 sock_hold(sk);
634 __sk_nulls_add_node_rcu(sk, list);
635 }
636
__sk_del_bind_node(struct sock * sk)637 static inline void __sk_del_bind_node(struct sock *sk)
638 {
639 __hlist_del(&sk->sk_bind_node);
640 }
641
sk_add_bind_node(struct sock * sk,struct hlist_head * list)642 static inline void sk_add_bind_node(struct sock *sk,
643 struct hlist_head *list)
644 {
645 hlist_add_head(&sk->sk_bind_node, list);
646 }
647
648 #define sk_for_each(__sk, list) \
649 hlist_for_each_entry(__sk, list, sk_node)
650 #define sk_for_each_rcu(__sk, list) \
651 hlist_for_each_entry_rcu(__sk, list, sk_node)
652 #define sk_nulls_for_each(__sk, node, list) \
653 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
654 #define sk_nulls_for_each_rcu(__sk, node, list) \
655 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
656 #define sk_for_each_from(__sk) \
657 hlist_for_each_entry_from(__sk, sk_node)
658 #define sk_nulls_for_each_from(__sk, node) \
659 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
660 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
661 #define sk_for_each_safe(__sk, tmp, list) \
662 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
663 #define sk_for_each_bound(__sk, list) \
664 hlist_for_each_entry(__sk, list, sk_bind_node)
665
666 /**
667 * sk_nulls_for_each_entry_offset - iterate over a list at a given struct offset
668 * @tpos: the type * to use as a loop cursor.
669 * @pos: the &struct hlist_node to use as a loop cursor.
670 * @head: the head for your list.
671 * @offset: offset of hlist_node within the struct.
672 *
673 */
674 #define sk_nulls_for_each_entry_offset(tpos, pos, head, offset) \
675 for (pos = (head)->first; \
676 (!is_a_nulls(pos)) && \
677 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
678 pos = pos->next)
679
sk_user_ns(struct sock * sk)680 static inline struct user_namespace *sk_user_ns(struct sock *sk)
681 {
682 /* Careful only use this in a context where these parameters
683 * can not change and must all be valid, such as recvmsg from
684 * userspace.
685 */
686 return sk->sk_socket->file->f_cred->user_ns;
687 }
688
689 /* Sock flags */
690 enum sock_flags {
691 SOCK_DEAD,
692 SOCK_DONE,
693 SOCK_URGINLINE,
694 SOCK_KEEPOPEN,
695 SOCK_LINGER,
696 SOCK_DESTROY,
697 SOCK_BROADCAST,
698 SOCK_TIMESTAMP,
699 SOCK_ZAPPED,
700 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
701 SOCK_DBG, /* %SO_DEBUG setting */
702 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
703 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
704 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
705 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
706 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
707 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
708 SOCK_FASYNC, /* fasync() active */
709 SOCK_RXQ_OVFL,
710 SOCK_ZEROCOPY, /* buffers from userspace */
711 SOCK_WIFI_STATUS, /* push wifi status to userspace */
712 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
713 * Will use last 4 bytes of packet sent from
714 * user-space instead.
715 */
716 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
717 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
718 };
719
720 #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
721
sock_copy_flags(struct sock * nsk,struct sock * osk)722 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
723 {
724 nsk->sk_flags = osk->sk_flags;
725 }
726
sock_set_flag(struct sock * sk,enum sock_flags flag)727 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
728 {
729 __set_bit(flag, &sk->sk_flags);
730 }
731
sock_reset_flag(struct sock * sk,enum sock_flags flag)732 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
733 {
734 __clear_bit(flag, &sk->sk_flags);
735 }
736
sock_flag(const struct sock * sk,enum sock_flags flag)737 static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
738 {
739 return test_bit(flag, &sk->sk_flags);
740 }
741
742 #ifdef CONFIG_NET
743 extern struct static_key memalloc_socks;
sk_memalloc_socks(void)744 static inline int sk_memalloc_socks(void)
745 {
746 return static_key_false(&memalloc_socks);
747 }
748 #else
749
sk_memalloc_socks(void)750 static inline int sk_memalloc_socks(void)
751 {
752 return 0;
753 }
754
755 #endif
756
sk_gfp_atomic(struct sock * sk,gfp_t gfp_mask)757 static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
758 {
759 return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
760 }
761
sk_acceptq_removed(struct sock * sk)762 static inline void sk_acceptq_removed(struct sock *sk)
763 {
764 sk->sk_ack_backlog--;
765 }
766
sk_acceptq_added(struct sock * sk)767 static inline void sk_acceptq_added(struct sock *sk)
768 {
769 sk->sk_ack_backlog++;
770 }
771
sk_acceptq_is_full(const struct sock * sk)772 static inline bool sk_acceptq_is_full(const struct sock *sk)
773 {
774 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
775 }
776
777 /*
778 * Compute minimal free write space needed to queue new packets.
779 */
sk_stream_min_wspace(const struct sock * sk)780 static inline int sk_stream_min_wspace(const struct sock *sk)
781 {
782 return sk->sk_wmem_queued >> 1;
783 }
784
sk_stream_wspace(const struct sock * sk)785 static inline int sk_stream_wspace(const struct sock *sk)
786 {
787 return sk->sk_sndbuf - sk->sk_wmem_queued;
788 }
789
790 void sk_stream_write_space(struct sock *sk);
791
792 /* OOB backlog add */
__sk_add_backlog(struct sock * sk,struct sk_buff * skb)793 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
794 {
795 /* dont let skb dst not refcounted, we are going to leave rcu lock */
796 skb_dst_force_safe(skb);
797
798 if (!sk->sk_backlog.tail)
799 sk->sk_backlog.head = skb;
800 else
801 sk->sk_backlog.tail->next = skb;
802
803 sk->sk_backlog.tail = skb;
804 skb->next = NULL;
805 }
806
807 /*
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.
811 */
sk_rcvqueues_full(const struct sock * sk,unsigned int limit)812 static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
813 {
814 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
815
816 return qsize > limit;
817 }
818
819 /* The per-socket spinlock must be held here. */
sk_add_backlog(struct sock * sk,struct sk_buff * skb,unsigned int limit)820 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
821 unsigned int limit)
822 {
823 if (sk_rcvqueues_full(sk, limit))
824 return -ENOBUFS;
825
826 /*
827 * If the skb was allocated from pfmemalloc reserves, only
828 * allow SOCK_MEMALLOC sockets to use it as this socket is
829 * helping free memory
830 */
831 if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC))
832 return -ENOMEM;
833
834 __sk_add_backlog(sk, skb);
835 sk->sk_backlog.len += skb->truesize;
836 return 0;
837 }
838
839 int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
840
sk_backlog_rcv(struct sock * sk,struct sk_buff * skb)841 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
842 {
843 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
844 return __sk_backlog_rcv(sk, skb);
845
846 return sk->sk_backlog_rcv(sk, skb);
847 }
848
sock_rps_record_flow_hash(__u32 hash)849 static inline void sock_rps_record_flow_hash(__u32 hash)
850 {
851 #ifdef CONFIG_RPS
852 struct rps_sock_flow_table *sock_flow_table;
853
854 rcu_read_lock();
855 sock_flow_table = rcu_dereference(rps_sock_flow_table);
856 rps_record_sock_flow(sock_flow_table, hash);
857 rcu_read_unlock();
858 #endif
859 }
860
sock_rps_reset_flow_hash(__u32 hash)861 static inline void sock_rps_reset_flow_hash(__u32 hash)
862 {
863 #ifdef CONFIG_RPS
864 struct rps_sock_flow_table *sock_flow_table;
865
866 rcu_read_lock();
867 sock_flow_table = rcu_dereference(rps_sock_flow_table);
868 rps_reset_sock_flow(sock_flow_table, hash);
869 rcu_read_unlock();
870 #endif
871 }
872
sock_rps_record_flow(const struct sock * sk)873 static inline void sock_rps_record_flow(const struct sock *sk)
874 {
875 #ifdef CONFIG_RPS
876 sock_rps_record_flow_hash(sk->sk_rxhash);
877 #endif
878 }
879
sock_rps_reset_flow(const struct sock * sk)880 static inline void sock_rps_reset_flow(const struct sock *sk)
881 {
882 #ifdef CONFIG_RPS
883 sock_rps_reset_flow_hash(sk->sk_rxhash);
884 #endif
885 }
886
sock_rps_save_rxhash(struct sock * sk,const struct sk_buff * skb)887 static inline void sock_rps_save_rxhash(struct sock *sk,
888 const struct sk_buff *skb)
889 {
890 #ifdef CONFIG_RPS
891 if (unlikely(sk->sk_rxhash != skb->hash)) {
892 sock_rps_reset_flow(sk);
893 sk->sk_rxhash = skb->hash;
894 }
895 #endif
896 }
897
sock_rps_reset_rxhash(struct sock * sk)898 static inline void sock_rps_reset_rxhash(struct sock *sk)
899 {
900 #ifdef CONFIG_RPS
901 sock_rps_reset_flow(sk);
902 sk->sk_rxhash = 0;
903 #endif
904 }
905
906 #define sk_wait_event(__sk, __timeo, __condition) \
907 ({ int __rc; \
908 release_sock(__sk); \
909 __rc = __condition; \
910 if (!__rc) { \
911 *(__timeo) = schedule_timeout(*(__timeo)); \
912 } \
913 lock_sock(__sk); \
914 __rc = __condition; \
915 __rc; \
916 })
917
918 int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
919 int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
920 void sk_stream_wait_close(struct sock *sk, long timeo_p);
921 int sk_stream_error(struct sock *sk, int flags, int err);
922 void sk_stream_kill_queues(struct sock *sk);
923 void sk_set_memalloc(struct sock *sk);
924 void sk_clear_memalloc(struct sock *sk);
925
926 int sk_wait_data(struct sock *sk, long *timeo);
927
928 struct request_sock_ops;
929 struct timewait_sock_ops;
930 struct inet_hashinfo;
931 struct raw_hashinfo;
932 struct module;
933
934 /*
935 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
936 * un-modified. Special care is taken when initializing object to zero.
937 */
sk_prot_clear_nulls(struct sock * sk,int size)938 static inline void sk_prot_clear_nulls(struct sock *sk, int size)
939 {
940 if (offsetof(struct sock, sk_node.next) != 0)
941 memset(sk, 0, offsetof(struct sock, sk_node.next));
942 memset(&sk->sk_node.pprev, 0,
943 size - offsetof(struct sock, sk_node.pprev));
944 }
945
946 /* Networking protocol blocks we attach to sockets.
947 * socket layer -> transport layer interface
948 * transport -> network interface is defined by struct inet_proto
949 */
950 struct proto {
951 void (*close)(struct sock *sk,
952 long timeout);
953 int (*connect)(struct sock *sk,
954 struct sockaddr *uaddr,
955 int addr_len);
956 int (*disconnect)(struct sock *sk, int flags);
957
958 struct sock * (*accept)(struct sock *sk, int flags, int *err);
959
960 int (*ioctl)(struct sock *sk, int cmd,
961 unsigned long arg);
962 int (*init)(struct sock *sk);
963 void (*destroy)(struct sock *sk);
964 void (*shutdown)(struct sock *sk, int how);
965 int (*setsockopt)(struct sock *sk, int level,
966 int optname, char __user *optval,
967 unsigned int optlen);
968 int (*getsockopt)(struct sock *sk, int level,
969 int optname, char __user *optval,
970 int __user *option);
971 #ifdef CONFIG_COMPAT
972 int (*compat_setsockopt)(struct sock *sk,
973 int level,
974 int optname, char __user *optval,
975 unsigned int optlen);
976 int (*compat_getsockopt)(struct sock *sk,
977 int level,
978 int optname, char __user *optval,
979 int __user *option);
980 int (*compat_ioctl)(struct sock *sk,
981 unsigned int cmd, unsigned long arg);
982 #endif
983 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
984 struct msghdr *msg, size_t len);
985 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
986 struct msghdr *msg,
987 size_t len, int noblock, int flags,
988 int *addr_len);
989 int (*sendpage)(struct sock *sk, struct page *page,
990 int offset, size_t size, int flags);
991 int (*bind)(struct sock *sk,
992 struct sockaddr *uaddr, int addr_len);
993
994 int (*backlog_rcv) (struct sock *sk,
995 struct sk_buff *skb);
996
997 void (*release_cb)(struct sock *sk);
998
999 /* Keeping track of sk's, looking them up, and port selection methods. */
1000 void (*hash)(struct sock *sk);
1001 void (*unhash)(struct sock *sk);
1002 void (*rehash)(struct sock *sk);
1003 int (*get_port)(struct sock *sk, unsigned short snum);
1004 void (*clear_sk)(struct sock *sk, int size);
1005
1006 /* Keeping track of sockets in use */
1007 #ifdef CONFIG_PROC_FS
1008 unsigned int inuse_idx;
1009 #endif
1010
1011 bool (*stream_memory_free)(const struct sock *sk);
1012 /* Memory pressure */
1013 void (*enter_memory_pressure)(struct sock *sk);
1014 atomic_long_t *memory_allocated; /* Current allocated memory. */
1015 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
1016 /*
1017 * Pressure flag: try to collapse.
1018 * Technical note: it is used by multiple contexts non atomically.
1019 * All the __sk_mem_schedule() is of this nature: accounting
1020 * is strict, actions are advisory and have some latency.
1021 */
1022 int *memory_pressure;
1023 long *sysctl_mem;
1024 int *sysctl_wmem;
1025 int *sysctl_rmem;
1026 int max_header;
1027 bool no_autobind;
1028
1029 struct kmem_cache *slab;
1030 unsigned int obj_size;
1031 int slab_flags;
1032
1033 struct percpu_counter *orphan_count;
1034
1035 struct request_sock_ops *rsk_prot;
1036 struct timewait_sock_ops *twsk_prot;
1037
1038 union {
1039 struct inet_hashinfo *hashinfo;
1040 struct udp_table *udp_table;
1041 struct raw_hashinfo *raw_hash;
1042 } h;
1043
1044 struct module *owner;
1045
1046 char name[32];
1047
1048 struct list_head node;
1049 #ifdef SOCK_REFCNT_DEBUG
1050 atomic_t socks;
1051 #endif
1052 #ifdef CONFIG_MEMCG_KMEM
1053 /*
1054 * cgroup specific init/deinit functions. Called once for all
1055 * protocols that implement it, from cgroups populate function.
1056 * This function has to setup any files the protocol want to
1057 * appear in the kmem cgroup filesystem.
1058 */
1059 int (*init_cgroup)(struct mem_cgroup *memcg,
1060 struct cgroup_subsys *ss);
1061 void (*destroy_cgroup)(struct mem_cgroup *memcg);
1062 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
1063 #endif
1064 int (*diag_destroy)(struct sock *sk, int err);
1065 };
1066
1067 /*
1068 * Bits in struct cg_proto.flags
1069 */
1070 enum cg_proto_flags {
1071 /* Currently active and new sockets should be assigned to cgroups */
1072 MEMCG_SOCK_ACTIVE,
1073 /* It was ever activated; we must disarm static keys on destruction */
1074 MEMCG_SOCK_ACTIVATED,
1075 };
1076
1077 struct cg_proto {
1078 struct res_counter memory_allocated; /* Current allocated memory. */
1079 struct percpu_counter sockets_allocated; /* Current number of sockets. */
1080 int memory_pressure;
1081 long sysctl_mem[3];
1082 unsigned long flags;
1083 /*
1084 * memcg field is used to find which memcg we belong directly
1085 * Each memcg struct can hold more than one cg_proto, so container_of
1086 * won't really cut.
1087 *
1088 * The elegant solution would be having an inverse function to
1089 * proto_cgroup in struct proto, but that means polluting the structure
1090 * for everybody, instead of just for memcg users.
1091 */
1092 struct mem_cgroup *memcg;
1093 };
1094
1095 int proto_register(struct proto *prot, int alloc_slab);
1096 void proto_unregister(struct proto *prot);
1097
memcg_proto_active(struct cg_proto * cg_proto)1098 static inline bool memcg_proto_active(struct cg_proto *cg_proto)
1099 {
1100 return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
1101 }
1102
memcg_proto_activated(struct cg_proto * cg_proto)1103 static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
1104 {
1105 return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
1106 }
1107
1108 #ifdef SOCK_REFCNT_DEBUG
sk_refcnt_debug_inc(struct sock * sk)1109 static inline void sk_refcnt_debug_inc(struct sock *sk)
1110 {
1111 atomic_inc(&sk->sk_prot->socks);
1112 }
1113
sk_refcnt_debug_dec(struct sock * sk)1114 static inline void sk_refcnt_debug_dec(struct sock *sk)
1115 {
1116 atomic_dec(&sk->sk_prot->socks);
1117 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1118 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1119 }
1120
sk_refcnt_debug_release(const struct sock * sk)1121 static inline void sk_refcnt_debug_release(const struct sock *sk)
1122 {
1123 if (atomic_read(&sk->sk_refcnt) != 1)
1124 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1125 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1126 }
1127 #else /* SOCK_REFCNT_DEBUG */
1128 #define sk_refcnt_debug_inc(sk) do { } while (0)
1129 #define sk_refcnt_debug_dec(sk) do { } while (0)
1130 #define sk_refcnt_debug_release(sk) do { } while (0)
1131 #endif /* SOCK_REFCNT_DEBUG */
1132
1133 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
1134 extern struct static_key memcg_socket_limit_enabled;
parent_cg_proto(struct proto * proto,struct cg_proto * cg_proto)1135 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1136 struct cg_proto *cg_proto)
1137 {
1138 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1139 }
1140 #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
1141 #else
1142 #define mem_cgroup_sockets_enabled 0
parent_cg_proto(struct proto * proto,struct cg_proto * cg_proto)1143 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1144 struct cg_proto *cg_proto)
1145 {
1146 return NULL;
1147 }
1148 #endif
1149
sk_stream_memory_free(const struct sock * sk)1150 static inline bool sk_stream_memory_free(const struct sock *sk)
1151 {
1152 if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1153 return false;
1154
1155 return sk->sk_prot->stream_memory_free ?
1156 sk->sk_prot->stream_memory_free(sk) : true;
1157 }
1158
sk_stream_is_writeable(const struct sock * sk)1159 static inline bool sk_stream_is_writeable(const struct sock *sk)
1160 {
1161 return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1162 sk_stream_memory_free(sk);
1163 }
1164
1165
sk_has_memory_pressure(const struct sock * sk)1166 static inline bool sk_has_memory_pressure(const struct sock *sk)
1167 {
1168 return sk->sk_prot->memory_pressure != NULL;
1169 }
1170
sk_under_memory_pressure(const struct sock * sk)1171 static inline bool sk_under_memory_pressure(const struct sock *sk)
1172 {
1173 if (!sk->sk_prot->memory_pressure)
1174 return false;
1175
1176 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1177 return !!sk->sk_cgrp->memory_pressure;
1178
1179 return !!*sk->sk_prot->memory_pressure;
1180 }
1181
sk_leave_memory_pressure(struct sock * sk)1182 static inline void sk_leave_memory_pressure(struct sock *sk)
1183 {
1184 int *memory_pressure = sk->sk_prot->memory_pressure;
1185
1186 if (!memory_pressure)
1187 return;
1188
1189 if (*memory_pressure)
1190 *memory_pressure = 0;
1191
1192 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1193 struct cg_proto *cg_proto = sk->sk_cgrp;
1194 struct proto *prot = sk->sk_prot;
1195
1196 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1197 cg_proto->memory_pressure = 0;
1198 }
1199
1200 }
1201
sk_enter_memory_pressure(struct sock * sk)1202 static inline void sk_enter_memory_pressure(struct sock *sk)
1203 {
1204 if (!sk->sk_prot->enter_memory_pressure)
1205 return;
1206
1207 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1208 struct cg_proto *cg_proto = sk->sk_cgrp;
1209 struct proto *prot = sk->sk_prot;
1210
1211 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1212 cg_proto->memory_pressure = 1;
1213 }
1214
1215 sk->sk_prot->enter_memory_pressure(sk);
1216 }
1217
sk_prot_mem_limits(const struct sock * sk,int index)1218 static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1219 {
1220 long *prot = sk->sk_prot->sysctl_mem;
1221 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1222 prot = sk->sk_cgrp->sysctl_mem;
1223 return prot[index];
1224 }
1225
memcg_memory_allocated_add(struct cg_proto * prot,unsigned long amt,int * parent_status)1226 static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1227 unsigned long amt,
1228 int *parent_status)
1229 {
1230 struct res_counter *fail;
1231 int ret;
1232
1233 ret = res_counter_charge_nofail(&prot->memory_allocated,
1234 amt << PAGE_SHIFT, &fail);
1235 if (ret < 0)
1236 *parent_status = OVER_LIMIT;
1237 }
1238
memcg_memory_allocated_sub(struct cg_proto * prot,unsigned long amt)1239 static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1240 unsigned long amt)
1241 {
1242 res_counter_uncharge(&prot->memory_allocated, amt << PAGE_SHIFT);
1243 }
1244
memcg_memory_allocated_read(struct cg_proto * prot)1245 static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1246 {
1247 u64 ret;
1248 ret = res_counter_read_u64(&prot->memory_allocated, RES_USAGE);
1249 return ret >> PAGE_SHIFT;
1250 }
1251
1252 static inline long
sk_memory_allocated(const struct sock * sk)1253 sk_memory_allocated(const struct sock *sk)
1254 {
1255 struct proto *prot = sk->sk_prot;
1256 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1257 return memcg_memory_allocated_read(sk->sk_cgrp);
1258
1259 return atomic_long_read(prot->memory_allocated);
1260 }
1261
1262 static inline long
sk_memory_allocated_add(struct sock * sk,int amt,int * parent_status)1263 sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
1264 {
1265 struct proto *prot = sk->sk_prot;
1266
1267 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1268 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1269 /* update the root cgroup regardless */
1270 atomic_long_add_return(amt, prot->memory_allocated);
1271 return memcg_memory_allocated_read(sk->sk_cgrp);
1272 }
1273
1274 return atomic_long_add_return(amt, prot->memory_allocated);
1275 }
1276
1277 static inline void
sk_memory_allocated_sub(struct sock * sk,int amt)1278 sk_memory_allocated_sub(struct sock *sk, int amt)
1279 {
1280 struct proto *prot = sk->sk_prot;
1281
1282 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1283 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1284
1285 atomic_long_sub(amt, prot->memory_allocated);
1286 }
1287
sk_sockets_allocated_dec(struct sock * sk)1288 static inline void sk_sockets_allocated_dec(struct sock *sk)
1289 {
1290 struct proto *prot = sk->sk_prot;
1291
1292 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1293 struct cg_proto *cg_proto = sk->sk_cgrp;
1294
1295 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1296 percpu_counter_dec(&cg_proto->sockets_allocated);
1297 }
1298
1299 percpu_counter_dec(prot->sockets_allocated);
1300 }
1301
sk_sockets_allocated_inc(struct sock * sk)1302 static inline void sk_sockets_allocated_inc(struct sock *sk)
1303 {
1304 struct proto *prot = sk->sk_prot;
1305
1306 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1307 struct cg_proto *cg_proto = sk->sk_cgrp;
1308
1309 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1310 percpu_counter_inc(&cg_proto->sockets_allocated);
1311 }
1312
1313 percpu_counter_inc(prot->sockets_allocated);
1314 }
1315
1316 static inline int
sk_sockets_allocated_read_positive(struct sock * sk)1317 sk_sockets_allocated_read_positive(struct sock *sk)
1318 {
1319 struct proto *prot = sk->sk_prot;
1320
1321 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1322 return percpu_counter_read_positive(&sk->sk_cgrp->sockets_allocated);
1323
1324 return percpu_counter_read_positive(prot->sockets_allocated);
1325 }
1326
1327 static inline int
proto_sockets_allocated_sum_positive(struct proto * prot)1328 proto_sockets_allocated_sum_positive(struct proto *prot)
1329 {
1330 return percpu_counter_sum_positive(prot->sockets_allocated);
1331 }
1332
1333 static inline long
proto_memory_allocated(struct proto * prot)1334 proto_memory_allocated(struct proto *prot)
1335 {
1336 return atomic_long_read(prot->memory_allocated);
1337 }
1338
1339 static inline bool
proto_memory_pressure(struct proto * prot)1340 proto_memory_pressure(struct proto *prot)
1341 {
1342 if (!prot->memory_pressure)
1343 return false;
1344 return !!*prot->memory_pressure;
1345 }
1346
1347
1348 #ifdef CONFIG_PROC_FS
1349 /* Called with local bh disabled */
1350 void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1351 int sock_prot_inuse_get(struct net *net, struct proto *proto);
1352 #else
sock_prot_inuse_add(struct net * net,struct proto * prot,int inc)1353 static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
1354 int inc)
1355 {
1356 }
1357 #endif
1358
1359
1360 /* With per-bucket locks this operation is not-atomic, so that
1361 * this version is not worse.
1362 */
__sk_prot_rehash(struct sock * sk)1363 static inline void __sk_prot_rehash(struct sock *sk)
1364 {
1365 sk->sk_prot->unhash(sk);
1366 sk->sk_prot->hash(sk);
1367 }
1368
1369 void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1370
1371 /* About 10 seconds */
1372 #define SOCK_DESTROY_TIME (10*HZ)
1373
1374 /* Sockets 0-1023 can't be bound to unless you are superuser */
1375 #define PROT_SOCK 1024
1376
1377 #define SHUTDOWN_MASK 3
1378 #define RCV_SHUTDOWN 1
1379 #define SEND_SHUTDOWN 2
1380
1381 #define SOCK_SNDBUF_LOCK 1
1382 #define SOCK_RCVBUF_LOCK 2
1383 #define SOCK_BINDADDR_LOCK 4
1384 #define SOCK_BINDPORT_LOCK 8
1385
1386 /* sock_iocb: used to kick off async processing of socket ios */
1387 struct sock_iocb {
1388 struct list_head list;
1389
1390 int flags;
1391 int size;
1392 struct socket *sock;
1393 struct sock *sk;
1394 struct scm_cookie *scm;
1395 struct msghdr *msg, async_msg;
1396 struct kiocb *kiocb;
1397 };
1398
kiocb_to_siocb(struct kiocb * iocb)1399 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1400 {
1401 return (struct sock_iocb *)iocb->private;
1402 }
1403
siocb_to_kiocb(struct sock_iocb * si)1404 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1405 {
1406 return si->kiocb;
1407 }
1408
1409 struct socket_alloc {
1410 struct socket socket;
1411 struct inode vfs_inode;
1412 };
1413
SOCKET_I(struct inode * inode)1414 static inline struct socket *SOCKET_I(struct inode *inode)
1415 {
1416 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1417 }
1418
SOCK_INODE(struct socket * socket)1419 static inline struct inode *SOCK_INODE(struct socket *socket)
1420 {
1421 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1422 }
1423
1424 /*
1425 * Functions for memory accounting
1426 */
1427 int __sk_mem_schedule(struct sock *sk, int size, int kind);
1428 void __sk_mem_reclaim(struct sock *sk);
1429
1430 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1431 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1432 #define SK_MEM_SEND 0
1433 #define SK_MEM_RECV 1
1434
sk_mem_pages(int amt)1435 static inline int sk_mem_pages(int amt)
1436 {
1437 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1438 }
1439
sk_has_account(struct sock * sk)1440 static inline bool sk_has_account(struct sock *sk)
1441 {
1442 /* return true if protocol supports memory accounting */
1443 return !!sk->sk_prot->memory_allocated;
1444 }
1445
sk_wmem_schedule(struct sock * sk,int size)1446 static inline bool sk_wmem_schedule(struct sock *sk, int size)
1447 {
1448 if (!sk_has_account(sk))
1449 return true;
1450 return size <= sk->sk_forward_alloc ||
1451 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1452 }
1453
1454 static inline bool
sk_rmem_schedule(struct sock * sk,struct sk_buff * skb,int size)1455 sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
1456 {
1457 if (!sk_has_account(sk))
1458 return true;
1459 return size<= sk->sk_forward_alloc ||
1460 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1461 skb_pfmemalloc(skb);
1462 }
1463
sk_mem_reclaim(struct sock * sk)1464 static inline void sk_mem_reclaim(struct sock *sk)
1465 {
1466 if (!sk_has_account(sk))
1467 return;
1468 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1469 __sk_mem_reclaim(sk);
1470 }
1471
sk_mem_reclaim_partial(struct sock * sk)1472 static inline void sk_mem_reclaim_partial(struct sock *sk)
1473 {
1474 if (!sk_has_account(sk))
1475 return;
1476 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1477 __sk_mem_reclaim(sk);
1478 }
1479
sk_mem_charge(struct sock * sk,int size)1480 static inline void sk_mem_charge(struct sock *sk, int size)
1481 {
1482 if (!sk_has_account(sk))
1483 return;
1484 sk->sk_forward_alloc -= size;
1485 }
1486
sk_mem_uncharge(struct sock * sk,int size)1487 static inline void sk_mem_uncharge(struct sock *sk, int size)
1488 {
1489 if (!sk_has_account(sk))
1490 return;
1491 sk->sk_forward_alloc += size;
1492 }
1493
sk_wmem_free_skb(struct sock * sk,struct sk_buff * skb)1494 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1495 {
1496 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1497 sk->sk_wmem_queued -= skb->truesize;
1498 sk_mem_uncharge(sk, skb->truesize);
1499 __kfree_skb(skb);
1500 }
1501
1502 /* Used by processes to "lock" a socket state, so that
1503 * interrupts and bottom half handlers won't change it
1504 * from under us. It essentially blocks any incoming
1505 * packets, so that we won't get any new data or any
1506 * packets that change the state of the socket.
1507 *
1508 * While locked, BH processing will add new packets to
1509 * the backlog queue. This queue is processed by the
1510 * owner of the socket lock right before it is released.
1511 *
1512 * Since ~2.3.5 it is also exclusive sleep lock serializing
1513 * accesses from user process context.
1514 */
1515 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1516
sock_release_ownership(struct sock * sk)1517 static inline void sock_release_ownership(struct sock *sk)
1518 {
1519 sk->sk_lock.owned = 0;
1520 }
1521
1522 /*
1523 * Macro so as to not evaluate some arguments when
1524 * lockdep is not enabled.
1525 *
1526 * Mark both the sk_lock and the sk_lock.slock as a
1527 * per-address-family lock class.
1528 */
1529 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1530 do { \
1531 sk->sk_lock.owned = 0; \
1532 init_waitqueue_head(&sk->sk_lock.wq); \
1533 spin_lock_init(&(sk)->sk_lock.slock); \
1534 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1535 sizeof((sk)->sk_lock)); \
1536 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1537 (skey), (sname)); \
1538 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1539 } while (0)
1540
1541 void lock_sock_nested(struct sock *sk, int subclass);
1542
lock_sock(struct sock * sk)1543 static inline void lock_sock(struct sock *sk)
1544 {
1545 lock_sock_nested(sk, 0);
1546 }
1547
1548 void release_sock(struct sock *sk);
1549
1550 /* BH context may only use the following locking interface. */
1551 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1552 #define bh_lock_sock_nested(__sk) \
1553 spin_lock_nested(&((__sk)->sk_lock.slock), \
1554 SINGLE_DEPTH_NESTING)
1555 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1556
1557 bool lock_sock_fast(struct sock *sk);
1558 /**
1559 * unlock_sock_fast - complement of lock_sock_fast
1560 * @sk: socket
1561 * @slow: slow mode
1562 *
1563 * fast unlock socket for user context.
1564 * If slow mode is on, we call regular release_sock()
1565 */
unlock_sock_fast(struct sock * sk,bool slow)1566 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1567 {
1568 if (slow)
1569 release_sock(sk);
1570 else
1571 spin_unlock_bh(&sk->sk_lock.slock);
1572 }
1573
1574
1575 struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1576 struct proto *prot);
1577 void sk_free(struct sock *sk);
1578 void sk_release_kernel(struct sock *sk);
1579 struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
1580
1581 struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1582 gfp_t priority);
1583 void sock_wfree(struct sk_buff *skb);
1584 void skb_orphan_partial(struct sk_buff *skb);
1585 void sock_rfree(struct sk_buff *skb);
1586 void sock_efree(struct sk_buff *skb);
1587 #ifdef CONFIG_INET
1588 void sock_edemux(struct sk_buff *skb);
1589 #else
1590 #define sock_edemux(skb) sock_efree(skb)
1591 #endif
1592
1593 int sock_setsockopt(struct socket *sock, int level, int op,
1594 char __user *optval, unsigned int optlen);
1595
1596 int sock_getsockopt(struct socket *sock, int level, int op,
1597 char __user *optval, int __user *optlen);
1598 struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1599 int noblock, int *errcode);
1600 struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1601 unsigned long data_len, int noblock,
1602 int *errcode, int max_page_order);
1603 void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1604 void sock_kfree_s(struct sock *sk, void *mem, int size);
1605 void sk_send_sigurg(struct sock *sk);
1606
1607 /*
1608 * Functions to fill in entries in struct proto_ops when a protocol
1609 * does not implement a particular function.
1610 */
1611 int sock_no_bind(struct socket *, struct sockaddr *, int);
1612 int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1613 int sock_no_socketpair(struct socket *, struct socket *);
1614 int sock_no_accept(struct socket *, struct socket *, int);
1615 int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1616 unsigned int sock_no_poll(struct file *, struct socket *,
1617 struct poll_table_struct *);
1618 int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1619 int sock_no_listen(struct socket *, int);
1620 int sock_no_shutdown(struct socket *, int);
1621 int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1622 int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
1623 int sock_no_sendmsg(struct kiocb *, struct socket *, struct msghdr *, size_t);
1624 int sock_no_recvmsg(struct kiocb *, struct socket *, struct msghdr *, size_t,
1625 int);
1626 int sock_no_mmap(struct file *file, struct socket *sock,
1627 struct vm_area_struct *vma);
1628 ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1629 size_t size, int flags);
1630
1631 /*
1632 * Functions to fill in entries in struct proto_ops when a protocol
1633 * uses the inet style.
1634 */
1635 int sock_common_getsockopt(struct socket *sock, int level, int optname,
1636 char __user *optval, int __user *optlen);
1637 int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1638 struct msghdr *msg, size_t size, int flags);
1639 int sock_common_setsockopt(struct socket *sock, int level, int optname,
1640 char __user *optval, unsigned int optlen);
1641 int compat_sock_common_getsockopt(struct socket *sock, int level,
1642 int optname, char __user *optval, int __user *optlen);
1643 int compat_sock_common_setsockopt(struct socket *sock, int level,
1644 int optname, char __user *optval, unsigned int optlen);
1645
1646 void sk_common_release(struct sock *sk);
1647
1648 /*
1649 * Default socket callbacks and setup code
1650 */
1651
1652 /* Initialise core socket variables */
1653 void sock_init_data(struct socket *sock, struct sock *sk);
1654
1655 /*
1656 * Socket reference counting postulates.
1657 *
1658 * * Each user of socket SHOULD hold a reference count.
1659 * * Each access point to socket (an hash table bucket, reference from a list,
1660 * running timer, skb in flight MUST hold a reference count.
1661 * * When reference count hits 0, it means it will never increase back.
1662 * * When reference count hits 0, it means that no references from
1663 * outside exist to this socket and current process on current CPU
1664 * is last user and may/should destroy this socket.
1665 * * sk_free is called from any context: process, BH, IRQ. When
1666 * it is called, socket has no references from outside -> sk_free
1667 * may release descendant resources allocated by the socket, but
1668 * to the time when it is called, socket is NOT referenced by any
1669 * hash tables, lists etc.
1670 * * Packets, delivered from outside (from network or from another process)
1671 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1672 * when they sit in queue. Otherwise, packets will leak to hole, when
1673 * socket is looked up by one cpu and unhasing is made by another CPU.
1674 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1675 * (leak to backlog). Packet socket does all the processing inside
1676 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1677 * use separate SMP lock, so that they are prone too.
1678 */
1679
1680 /* Ungrab socket and destroy it, if it was the last reference. */
sock_put(struct sock * sk)1681 static inline void sock_put(struct sock *sk)
1682 {
1683 if (atomic_dec_and_test(&sk->sk_refcnt))
1684 sk_free(sk);
1685 }
1686 /* Generic version of sock_put(), dealing with all sockets
1687 * (TCP_TIMEWAIT, ESTABLISHED...)
1688 */
1689 void sock_gen_put(struct sock *sk);
1690
1691 int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested);
1692
sk_tx_queue_set(struct sock * sk,int tx_queue)1693 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1694 {
1695 sk->sk_tx_queue_mapping = tx_queue;
1696 }
1697
sk_tx_queue_clear(struct sock * sk)1698 static inline void sk_tx_queue_clear(struct sock *sk)
1699 {
1700 sk->sk_tx_queue_mapping = -1;
1701 }
1702
sk_tx_queue_get(const struct sock * sk)1703 static inline int sk_tx_queue_get(const struct sock *sk)
1704 {
1705 return sk ? sk->sk_tx_queue_mapping : -1;
1706 }
1707
sk_set_socket(struct sock * sk,struct socket * sock)1708 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1709 {
1710 sk_tx_queue_clear(sk);
1711 sk->sk_socket = sock;
1712 }
1713
sk_sleep(struct sock * sk)1714 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1715 {
1716 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1717 return &rcu_dereference_raw(sk->sk_wq)->wait;
1718 }
1719 /* Detach socket from process context.
1720 * Announce socket dead, detach it from wait queue and inode.
1721 * Note that parent inode held reference count on this struct sock,
1722 * we do not release it in this function, because protocol
1723 * probably wants some additional cleanups or even continuing
1724 * to work with this socket (TCP).
1725 */
sock_orphan(struct sock * sk)1726 static inline void sock_orphan(struct sock *sk)
1727 {
1728 write_lock_bh(&sk->sk_callback_lock);
1729 sock_set_flag(sk, SOCK_DEAD);
1730 sk_set_socket(sk, NULL);
1731 sk->sk_wq = NULL;
1732 write_unlock_bh(&sk->sk_callback_lock);
1733 }
1734
sock_graft(struct sock * sk,struct socket * parent)1735 static inline void sock_graft(struct sock *sk, struct socket *parent)
1736 {
1737 write_lock_bh(&sk->sk_callback_lock);
1738 sk->sk_wq = parent->wq;
1739 parent->sk = sk;
1740 sk_set_socket(sk, parent);
1741 sk->sk_uid = SOCK_INODE(parent)->i_uid;
1742 security_sock_graft(sk, parent);
1743 write_unlock_bh(&sk->sk_callback_lock);
1744 }
1745
1746 kuid_t sock_i_uid(struct sock *sk);
1747 unsigned long sock_i_ino(struct sock *sk);
1748
sock_net_uid(const struct net * net,const struct sock * sk)1749 static inline kuid_t sock_net_uid(const struct net *net, const struct sock *sk)
1750 {
1751 return sk ? sk->sk_uid : make_kuid(net->user_ns, 0);
1752 }
1753
1754 static inline struct dst_entry *
__sk_dst_get(struct sock * sk)1755 __sk_dst_get(struct sock *sk)
1756 {
1757 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
1758 lockdep_is_held(&sk->sk_lock.slock));
1759 }
1760
1761 static inline struct dst_entry *
sk_dst_get(struct sock * sk)1762 sk_dst_get(struct sock *sk)
1763 {
1764 struct dst_entry *dst;
1765
1766 rcu_read_lock();
1767 dst = rcu_dereference(sk->sk_dst_cache);
1768 if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1769 dst = NULL;
1770 rcu_read_unlock();
1771 return dst;
1772 }
1773
dst_negative_advice(struct sock * sk)1774 static inline void dst_negative_advice(struct sock *sk)
1775 {
1776 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1777
1778 if (dst && dst->ops->negative_advice) {
1779 ndst = dst->ops->negative_advice(dst);
1780
1781 if (ndst != dst) {
1782 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1783 sk_tx_queue_clear(sk);
1784 }
1785 }
1786 }
1787
1788 static inline void
__sk_dst_set(struct sock * sk,struct dst_entry * dst)1789 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1790 {
1791 struct dst_entry *old_dst;
1792
1793 sk_tx_queue_clear(sk);
1794 /*
1795 * This can be called while sk is owned by the caller only,
1796 * with no state that can be checked in a rcu_dereference_check() cond
1797 */
1798 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1799 rcu_assign_pointer(sk->sk_dst_cache, dst);
1800 dst_release(old_dst);
1801 }
1802
1803 static inline void
sk_dst_set(struct sock * sk,struct dst_entry * dst)1804 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1805 {
1806 struct dst_entry *old_dst;
1807
1808 sk_tx_queue_clear(sk);
1809 old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
1810 dst_release(old_dst);
1811 }
1812
1813 static inline void
__sk_dst_reset(struct sock * sk)1814 __sk_dst_reset(struct sock *sk)
1815 {
1816 __sk_dst_set(sk, NULL);
1817 }
1818
1819 static inline void
sk_dst_reset(struct sock * sk)1820 sk_dst_reset(struct sock *sk)
1821 {
1822 sk_dst_set(sk, NULL);
1823 }
1824
1825 struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1826
1827 struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1828
1829 bool sk_mc_loop(struct sock *sk);
1830
sk_can_gso(const struct sock * sk)1831 static inline bool sk_can_gso(const struct sock *sk)
1832 {
1833 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1834 }
1835
1836 void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1837
sk_nocaps_add(struct sock * sk,netdev_features_t flags)1838 static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
1839 {
1840 sk->sk_route_nocaps |= flags;
1841 sk->sk_route_caps &= ~flags;
1842 }
1843
skb_do_copy_data_nocache(struct sock * sk,struct sk_buff * skb,char __user * from,char * to,int copy,int offset)1844 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1845 char __user *from, char *to,
1846 int copy, int offset)
1847 {
1848 if (skb->ip_summed == CHECKSUM_NONE) {
1849 int err = 0;
1850 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1851 if (err)
1852 return err;
1853 skb->csum = csum_block_add(skb->csum, csum, offset);
1854 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1855 if (!access_ok(VERIFY_READ, from, copy) ||
1856 __copy_from_user_nocache(to, from, copy))
1857 return -EFAULT;
1858 } else if (copy_from_user(to, from, copy))
1859 return -EFAULT;
1860
1861 return 0;
1862 }
1863
skb_add_data_nocache(struct sock * sk,struct sk_buff * skb,char __user * from,int copy)1864 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1865 char __user *from, int copy)
1866 {
1867 int err, offset = skb->len;
1868
1869 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1870 copy, offset);
1871 if (err)
1872 __skb_trim(skb, offset);
1873
1874 return err;
1875 }
1876
skb_copy_to_page_nocache(struct sock * sk,char __user * from,struct sk_buff * skb,struct page * page,int off,int copy)1877 static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1878 struct sk_buff *skb,
1879 struct page *page,
1880 int off, int copy)
1881 {
1882 int err;
1883
1884 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1885 copy, skb->len);
1886 if (err)
1887 return err;
1888
1889 skb->len += copy;
1890 skb->data_len += copy;
1891 skb->truesize += copy;
1892 sk->sk_wmem_queued += copy;
1893 sk_mem_charge(sk, copy);
1894 return 0;
1895 }
1896
skb_copy_to_page(struct sock * sk,char __user * from,struct sk_buff * skb,struct page * page,int off,int copy)1897 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1898 struct sk_buff *skb, struct page *page,
1899 int off, int copy)
1900 {
1901 if (skb->ip_summed == CHECKSUM_NONE) {
1902 int err = 0;
1903 __wsum csum = csum_and_copy_from_user(from,
1904 page_address(page) + off,
1905 copy, 0, &err);
1906 if (err)
1907 return err;
1908 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1909 } else if (copy_from_user(page_address(page) + off, from, copy))
1910 return -EFAULT;
1911
1912 skb->len += copy;
1913 skb->data_len += copy;
1914 skb->truesize += copy;
1915 sk->sk_wmem_queued += copy;
1916 sk_mem_charge(sk, copy);
1917 return 0;
1918 }
1919
1920 /**
1921 * sk_wmem_alloc_get - returns write allocations
1922 * @sk: socket
1923 *
1924 * Returns sk_wmem_alloc minus initial offset of one
1925 */
sk_wmem_alloc_get(const struct sock * sk)1926 static inline int sk_wmem_alloc_get(const struct sock *sk)
1927 {
1928 return atomic_read(&sk->sk_wmem_alloc) - 1;
1929 }
1930
1931 /**
1932 * sk_rmem_alloc_get - returns read allocations
1933 * @sk: socket
1934 *
1935 * Returns sk_rmem_alloc
1936 */
sk_rmem_alloc_get(const struct sock * sk)1937 static inline int sk_rmem_alloc_get(const struct sock *sk)
1938 {
1939 return atomic_read(&sk->sk_rmem_alloc);
1940 }
1941
1942 /**
1943 * sk_has_allocations - check if allocations are outstanding
1944 * @sk: socket
1945 *
1946 * Returns true if socket has write or read allocations
1947 */
sk_has_allocations(const struct sock * sk)1948 static inline bool sk_has_allocations(const struct sock *sk)
1949 {
1950 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1951 }
1952
1953 /**
1954 * wq_has_sleeper - check if there are any waiting processes
1955 * @wq: struct socket_wq
1956 *
1957 * Returns true if socket_wq has waiting processes
1958 *
1959 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1960 * barrier call. They were added due to the race found within the tcp code.
1961 *
1962 * Consider following tcp code paths:
1963 *
1964 * CPU1 CPU2
1965 *
1966 * sys_select receive packet
1967 * ... ...
1968 * __add_wait_queue update tp->rcv_nxt
1969 * ... ...
1970 * tp->rcv_nxt check sock_def_readable
1971 * ... {
1972 * schedule rcu_read_lock();
1973 * wq = rcu_dereference(sk->sk_wq);
1974 * if (wq && waitqueue_active(&wq->wait))
1975 * wake_up_interruptible(&wq->wait)
1976 * ...
1977 * }
1978 *
1979 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1980 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1981 * could then endup calling schedule and sleep forever if there are no more
1982 * data on the socket.
1983 *
1984 */
wq_has_sleeper(struct socket_wq * wq)1985 static inline bool wq_has_sleeper(struct socket_wq *wq)
1986 {
1987 /* We need to be sure we are in sync with the
1988 * add_wait_queue modifications to the wait queue.
1989 *
1990 * This memory barrier is paired in the sock_poll_wait.
1991 */
1992 smp_mb();
1993 return wq && waitqueue_active(&wq->wait);
1994 }
1995
1996 /**
1997 * sock_poll_wait - place memory barrier behind the poll_wait call.
1998 * @filp: file
1999 * @wait_address: socket wait queue
2000 * @p: poll_table
2001 *
2002 * See the comments in the wq_has_sleeper function.
2003 */
sock_poll_wait(struct file * filp,wait_queue_head_t * wait_address,poll_table * p)2004 static inline void sock_poll_wait(struct file *filp,
2005 wait_queue_head_t *wait_address, poll_table *p)
2006 {
2007 if (!poll_does_not_wait(p) && wait_address) {
2008 poll_wait(filp, wait_address, p);
2009 /* We need to be sure we are in sync with the
2010 * socket flags modification.
2011 *
2012 * This memory barrier is paired in the wq_has_sleeper.
2013 */
2014 smp_mb();
2015 }
2016 }
2017
skb_set_hash_from_sk(struct sk_buff * skb,struct sock * sk)2018 static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
2019 {
2020 if (sk->sk_txhash) {
2021 skb->l4_hash = 1;
2022 skb->hash = sk->sk_txhash;
2023 }
2024 }
2025
2026 /*
2027 * Queue a received datagram if it will fit. Stream and sequenced
2028 * protocols can't normally use this as they need to fit buffers in
2029 * and play with them.
2030 *
2031 * Inlined as it's very short and called for pretty much every
2032 * packet ever received.
2033 */
2034
skb_set_owner_w(struct sk_buff * skb,struct sock * sk)2035 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
2036 {
2037 skb_orphan(skb);
2038 skb->sk = sk;
2039 skb->destructor = sock_wfree;
2040 skb_set_hash_from_sk(skb, sk);
2041 /*
2042 * We used to take a refcount on sk, but following operation
2043 * is enough to guarantee sk_free() wont free this sock until
2044 * all in-flight packets are completed
2045 */
2046 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
2047 }
2048
skb_set_owner_r(struct sk_buff * skb,struct sock * sk)2049 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2050 {
2051 skb_orphan(skb);
2052 skb->sk = sk;
2053 skb->destructor = sock_rfree;
2054 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
2055 sk_mem_charge(sk, skb->truesize);
2056 }
2057
2058 void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2059 unsigned long expires);
2060
2061 void sk_stop_timer(struct sock *sk, struct timer_list *timer);
2062
2063 int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
2064
2065 int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
2066 struct sk_buff *sock_dequeue_err_skb(struct sock *sk);
2067
2068 /*
2069 * Recover an error report and clear atomically
2070 */
2071
sock_error(struct sock * sk)2072 static inline int sock_error(struct sock *sk)
2073 {
2074 int err;
2075 if (likely(!sk->sk_err))
2076 return 0;
2077 err = xchg(&sk->sk_err, 0);
2078 return -err;
2079 }
2080
sock_wspace(struct sock * sk)2081 static inline unsigned long sock_wspace(struct sock *sk)
2082 {
2083 int amt = 0;
2084
2085 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2086 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
2087 if (amt < 0)
2088 amt = 0;
2089 }
2090 return amt;
2091 }
2092
sk_wake_async(struct sock * sk,int how,int band)2093 static inline void sk_wake_async(struct sock *sk, int how, int band)
2094 {
2095 if (sock_flag(sk, SOCK_FASYNC))
2096 sock_wake_async(sk->sk_socket, how, band);
2097 }
2098
2099 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2100 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2101 * Note: for send buffers, TCP works better if we can build two skbs at
2102 * minimum.
2103 */
2104 #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2105
2106 #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2107 #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
2108
sk_stream_moderate_sndbuf(struct sock * sk)2109 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2110 {
2111 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
2112 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
2113 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
2114 }
2115 }
2116
2117 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
2118
2119 /**
2120 * sk_page_frag - return an appropriate page_frag
2121 * @sk: socket
2122 *
2123 * If socket allocation mode allows current thread to sleep, it means its
2124 * safe to use the per task page_frag instead of the per socket one.
2125 */
sk_page_frag(struct sock * sk)2126 static inline struct page_frag *sk_page_frag(struct sock *sk)
2127 {
2128 if (sk->sk_allocation & __GFP_WAIT)
2129 return ¤t->task_frag;
2130
2131 return &sk->sk_frag;
2132 }
2133
2134 bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
2135
2136 /*
2137 * Default write policy as shown to user space via poll/select/SIGIO
2138 */
sock_writeable(const struct sock * sk)2139 static inline bool sock_writeable(const struct sock *sk)
2140 {
2141 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
2142 }
2143
gfp_any(void)2144 static inline gfp_t gfp_any(void)
2145 {
2146 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
2147 }
2148
sock_rcvtimeo(const struct sock * sk,bool noblock)2149 static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
2150 {
2151 return noblock ? 0 : sk->sk_rcvtimeo;
2152 }
2153
sock_sndtimeo(const struct sock * sk,bool noblock)2154 static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
2155 {
2156 return noblock ? 0 : sk->sk_sndtimeo;
2157 }
2158
sock_rcvlowat(const struct sock * sk,int waitall,int len)2159 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2160 {
2161 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2162 }
2163
2164 /* Alas, with timeout socket operations are not restartable.
2165 * Compare this to poll().
2166 */
sock_intr_errno(long timeo)2167 static inline int sock_intr_errno(long timeo)
2168 {
2169 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2170 }
2171
2172 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2173 struct sk_buff *skb);
2174 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2175 struct sk_buff *skb);
2176
2177 static inline void
sock_recv_timestamp(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)2178 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2179 {
2180 ktime_t kt = skb->tstamp;
2181 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
2182
2183 /*
2184 * generate control messages if
2185 * - receive time stamping in software requested
2186 * - software time stamp available and wanted
2187 * - hardware time stamps available and wanted
2188 */
2189 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2190 (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
2191 (kt.tv64 && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
2192 (hwtstamps->hwtstamp.tv64 &&
2193 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
2194 __sock_recv_timestamp(msg, sk, skb);
2195 else
2196 sk->sk_stamp = kt;
2197
2198 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2199 __sock_recv_wifi_status(msg, sk, skb);
2200 }
2201
2202 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2203 struct sk_buff *skb);
2204
sock_recv_ts_and_drops(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)2205 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2206 struct sk_buff *skb)
2207 {
2208 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2209 (1UL << SOCK_RCVTSTAMP))
2210 #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2211 SOF_TIMESTAMPING_RAW_HARDWARE)
2212
2213 if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY)
2214 __sock_recv_ts_and_drops(msg, sk, skb);
2215 else
2216 sk->sk_stamp = skb->tstamp;
2217 }
2218
2219 void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags);
2220
2221 /**
2222 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2223 * @sk: socket sending this packet
2224 * @tx_flags: completed with instructions for time stamping
2225 *
2226 * Note : callers should take care of initial *tx_flags value (usually 0)
2227 */
sock_tx_timestamp(const struct sock * sk,__u8 * tx_flags)2228 static inline void sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
2229 {
2230 if (unlikely(sk->sk_tsflags))
2231 __sock_tx_timestamp(sk, tx_flags);
2232 if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS)))
2233 *tx_flags |= SKBTX_WIFI_STATUS;
2234 }
2235
2236 /**
2237 * sk_eat_skb - Release a skb if it is no longer needed
2238 * @sk: socket to eat this skb from
2239 * @skb: socket buffer to eat
2240 *
2241 * This routine must be called with interrupts disabled or with the socket
2242 * locked so that the sk_buff queue operation is ok.
2243 */
sk_eat_skb(struct sock * sk,struct sk_buff * skb)2244 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
2245 {
2246 __skb_unlink(skb, &sk->sk_receive_queue);
2247 __kfree_skb(skb);
2248 }
2249
2250 static inline
sock_net(const struct sock * sk)2251 struct net *sock_net(const struct sock *sk)
2252 {
2253 return read_pnet(&sk->sk_net);
2254 }
2255
2256 static inline
sock_net_set(struct sock * sk,struct net * net)2257 void sock_net_set(struct sock *sk, struct net *net)
2258 {
2259 write_pnet(&sk->sk_net, net);
2260 }
2261
2262 /*
2263 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
2264 * They should not hold a reference to a namespace in order to allow
2265 * to stop it.
2266 * Sockets after sk_change_net should be released using sk_release_kernel
2267 */
sk_change_net(struct sock * sk,struct net * net)2268 static inline void sk_change_net(struct sock *sk, struct net *net)
2269 {
2270 struct net *current_net = sock_net(sk);
2271
2272 if (!net_eq(current_net, net)) {
2273 put_net(current_net);
2274 sock_net_set(sk, hold_net(net));
2275 }
2276 }
2277
skb_steal_sock(struct sk_buff * skb)2278 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2279 {
2280 if (skb->sk) {
2281 struct sock *sk = skb->sk;
2282
2283 skb->destructor = NULL;
2284 skb->sk = NULL;
2285 return sk;
2286 }
2287 return NULL;
2288 }
2289
2290 /* This helper checks if a socket is a full socket,
2291 * ie _not_ a timewait or request socket.
2292 * TODO: Check for TCPF_NEW_SYN_RECV when that starts to exist.
2293 */
sk_fullsock(const struct sock * sk)2294 static inline bool sk_fullsock(const struct sock *sk)
2295 {
2296 return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT);
2297 }
2298
2299 void sock_enable_timestamp(struct sock *sk, int flag);
2300 int sock_get_timestamp(struct sock *, struct timeval __user *);
2301 int sock_get_timestampns(struct sock *, struct timespec __user *);
2302 int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2303 int type);
2304
2305 bool sk_ns_capable(const struct sock *sk,
2306 struct user_namespace *user_ns, int cap);
2307 bool sk_capable(const struct sock *sk, int cap);
2308 bool sk_net_capable(const struct sock *sk, int cap);
2309
2310 /*
2311 * Enable debug/info messages
2312 */
2313 extern int net_msg_warn;
2314 #define NETDEBUG(fmt, args...) \
2315 do { if (net_msg_warn) printk(fmt,##args); } while (0)
2316
2317 #define LIMIT_NETDEBUG(fmt, args...) \
2318 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
2319
2320 extern __u32 sysctl_wmem_max;
2321 extern __u32 sysctl_rmem_max;
2322
2323 extern int sysctl_optmem_max;
2324
2325 extern __u32 sysctl_wmem_default;
2326 extern __u32 sysctl_rmem_default;
2327
2328 #endif /* _SOCK_H */
2329