blob: 7f2098b41b464dcb730650f772b052f4df49b5ef [file] [log] [blame]
lh9ed821d2023-04-07 01:36:19 -07001/*
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/lockdep.h>
50#include <linux/netdevice.h>
51#include <linux/skbuff.h> /* struct sk_buff */
52#include <linux/mm.h>
53#include <linux/security.h>
54#include <linux/slab.h>
55#include <linux/uaccess.h>
56#include <linux/memcontrol.h>
57#include <linux/res_counter.h>
58#include <linux/static_key.h>
59#include <linux/aio.h>
60#include <linux/sched.h>
61
62#include <linux/filter.h>
63#include <linux/rculist_nulls.h>
64#include <linux/poll.h>
65
66#include <linux/atomic.h>
67#include <net/dst.h>
68#include <net/checksum.h>
69
70#include <net/SI/sock_track.h>
71
72struct cgroup;
73struct cgroup_subsys;
74#ifdef CONFIG_NET
75int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss);
76void mem_cgroup_sockets_destroy(struct cgroup *cgrp);
77#else
78static inline
79int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss)
80{
81 return 0;
82}
83static inline
84void mem_cgroup_sockets_destroy(struct cgroup *cgrp)
85{
86}
87#endif
xf.li6c8fc1e2023-08-12 00:11:09 -070088
89#ifdef CONFIG_IPC_SOCKET
90extern int socket_rpmsg_debug;
91
92#define sk_soc_dbg(format, arg...) if(socket_rpmsg_debug == 1) \
93 printk(KERN_DEBUG "[socket_rpmsg]<%s>: " format "\n" ,__func__ , ## arg)
94#define sk_soc_info(format, arg...) if(socket_rpmsg_debug == 1) \
95 printk(KERN_INFO "[socket_rpmsg]<%s>: " format "\n" ,__func__ , ## arg)
96
97#endif
lh9ed821d2023-04-07 01:36:19 -070098/*
99 * This structure really needs to be cleaned up.
100 * Most of it is for TCP, and not used by any of
101 * the other protocols.
102 */
103
104/* Define this to get the SOCK_DBG debugging facility. */
105#define SOCK_DEBUGGING
106#ifdef SOCK_DEBUGGING
107#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
108 printk(KERN_DEBUG msg); } while (0)
109#else
110/* Validate arguments and do nothing */
111static inline __printf(2, 3)
112void SOCK_DEBUG(struct sock *sk, const char *msg, ...)
113{
114}
115#endif
116
117/* This is the per-socket lock. The spinlock provides a synchronization
118 * between user contexts and software interrupt processing, whereas the
119 * mini-semaphore synchronizes multiple users amongst themselves.
120 */
121typedef struct {
122 spinlock_t slock;
123 int owned;
124 wait_queue_head_t wq;
125 /*
126 * We express the mutex-alike socket_lock semantics
127 * to the lock validator by explicitly managing
128 * the slock as a lock variant (in addition to
129 * the slock itself):
130 */
131#ifdef CONFIG_DEBUG_LOCK_ALLOC
132 struct lockdep_map dep_map;
133#endif
134} socket_lock_t;
135
136struct sock;
137struct proto;
138struct net;
139
xf.li6c8fc1e2023-08-12 00:11:09 -0700140#ifdef CONFIG_IPC_SOCKET
141
142enum sock_flags;
143extern int unix_is_ipc_socket(struct sock *sock);
144extern int sock_soc_test_flags(struct sock *other, enum sock_flags flag);
145extern int sock_soc_sock_put(struct sock *other);
146extern void sock_soc_sock_hold(struct sock *other);
147
148#endif
lh9ed821d2023-04-07 01:36:19 -0700149/**
150 * struct sock_common - minimal network layer representation of sockets
151 * @skc_daddr: Foreign IPv4 addr
152 * @skc_rcv_saddr: Bound local IPv4 addr
153 * @skc_hash: hash value used with various protocol lookup tables
154 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
155 * @skc_family: network address family
156 * @skc_state: Connection state
157 * @skc_reuse: %SO_REUSEADDR setting
158 * @skc_bound_dev_if: bound device index if != 0
159 * @skc_bind_node: bind hash linkage for various protocol lookup tables
160 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
161 * @skc_prot: protocol handlers inside a network family
162 * @skc_net: reference to the network namespace of this socket
163 * @skc_node: main hash linkage for various protocol lookup tables
164 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
165 * @skc_tx_queue_mapping: tx queue number for this connection
166 * @skc_refcnt: reference count
167 *
168 * This is the minimal network layer representation of sockets, the header
169 * for struct sock and struct inet_timewait_sock.
170 */
171struct sock_common {
172 /* skc_daddr and skc_rcv_saddr must be grouped :
173 * cf INET_MATCH() and INET_TW_MATCH()
174 */
175 __be32 skc_daddr;
176 __be32 skc_rcv_saddr;
177
178 union {
179 unsigned int skc_hash;
180 __u16 skc_u16hashes[2];
181 };
182 unsigned short skc_family;
183 volatile unsigned char skc_state;
184 unsigned char skc_reuse;
185 int skc_bound_dev_if;
186 union {
187 struct hlist_node skc_bind_node;
188 struct hlist_nulls_node skc_portaddr_node;
189 };
190 struct proto *skc_prot;
191#ifdef CONFIG_NET_NS
192 struct net *skc_net;
193#endif
194 /*
195 * fields between dontcopy_begin/dontcopy_end
196 * are not copied in sock_copy()
197 */
198 /* private: */
199 int skc_dontcopy_begin[0];
200 /* public: */
201 union {
202 struct hlist_node skc_node;
203 struct hlist_nulls_node skc_nulls_node;
204 };
205 int skc_tx_queue_mapping;
206 atomic_t skc_refcnt;
207 /* private: */
208 int skc_dontcopy_end[0];
209 /* public: */
210};
211
212struct cg_proto;
213/**
214 * struct sock - network layer representation of sockets
215 * @__sk_common: shared layout with inet_timewait_sock
216 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
217 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
218 * @sk_lock: synchronizer
219 * @sk_rcvbuf: size of receive buffer in bytes
220 * @sk_wq: sock wait queue and async head
221 * @sk_dst_cache: destination cache
222 * @sk_dst_lock: destination cache lock
223 * @sk_policy: flow policy
224 * @sk_receive_queue: incoming packets
225 * @sk_wmem_alloc: transmit queue bytes committed
226 * @sk_write_queue: Packet sending queue
227 * @sk_async_wait_queue: DMA copied packets
228 * @sk_omem_alloc: "o" is "option" or "other"
229 * @sk_wmem_queued: persistent queue size
230 * @sk_forward_alloc: space allocated forward
231 * @sk_allocation: allocation mode
232 * @sk_sndbuf: size of send buffer in bytes
233 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
234 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
235 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
236 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
237 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
238 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
239 * @sk_gso_max_size: Maximum GSO segment size to build
240 * @sk_gso_max_segs: Maximum number of GSO segments
241 * @sk_lingertime: %SO_LINGER l_linger setting
242 * @sk_backlog: always used with the per-socket spinlock held
243 * @sk_callback_lock: used with the callbacks in the end of this struct
244 * @sk_error_queue: rarely used
245 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
246 * IPV6_ADDRFORM for instance)
247 * @sk_err: last error
248 * @sk_err_soft: errors that don't cause failure but are the cause of a
249 * persistent failure not just 'timed out'
250 * @sk_drops: raw/udp drops counter
251 * @sk_ack_backlog: current listen backlog
252 * @sk_max_ack_backlog: listen backlog set in listen()
253 * @sk_priority: %SO_PRIORITY setting
254 * @sk_cgrp_prioidx: socket group's priority map index
255 * @sk_type: socket type (%SOCK_STREAM, etc)
256 * @sk_protocol: which protocol this socket belongs in this network family
257 * @sk_peer_pid: &struct pid for this socket's peer
258 * @sk_peer_cred: %SO_PEERCRED setting
259 * @sk_rcvlowat: %SO_RCVLOWAT setting
260 * @sk_rcvtimeo: %SO_RCVTIMEO setting
261 * @sk_sndtimeo: %SO_SNDTIMEO setting
262 * @sk_rxhash: flow hash received from netif layer
263 * @sk_filter: socket filtering instructions
264 * @sk_protinfo: private area, net family specific, when not using slab
265 * @sk_timer: sock cleanup timer
266 * @sk_stamp: time stamp of last packet received
267 * @sk_socket: Identd and reporting IO signals
268 * @sk_user_data: RPC layer private data
269 * @sk_sndmsg_page: cached page for sendmsg
270 * @sk_sndmsg_off: cached offset for sendmsg
271 * @sk_peek_off: current peek_offset value
272 * @sk_send_head: front of stuff to transmit
273 * @sk_security: used by security modules
274 * @sk_mark: generic packet mark
275 * @sk_classid: this socket's cgroup classid
276 * @sk_cgrp: this socket's cgroup-specific proto data
277 * @sk_write_pending: a write to stream socket waits to start
278 * @sk_state_change: callback to indicate change in the state of the sock
279 * @sk_data_ready: callback to indicate there is data to be processed
280 * @sk_write_space: callback to indicate there is bf sending space available
281 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
282 * @sk_backlog_rcv: callback to process the backlog
283 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
284 */
285struct sock {
286 /*
287 * Now struct inet_timewait_sock also uses sock_common, so please just
288 * don't add nothing before this first member (__sk_common) --acme
289 */
290 struct sock_common __sk_common;
291#define sk_node __sk_common.skc_node
292#define sk_nulls_node __sk_common.skc_nulls_node
293#define sk_refcnt __sk_common.skc_refcnt
294#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
295
296#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
297#define sk_dontcopy_end __sk_common.skc_dontcopy_end
298#define sk_hash __sk_common.skc_hash
299#define sk_family __sk_common.skc_family
300#define sk_state __sk_common.skc_state
301#define sk_reuse __sk_common.skc_reuse
302#define sk_bound_dev_if __sk_common.skc_bound_dev_if
303#define sk_bind_node __sk_common.skc_bind_node
304#define sk_prot __sk_common.skc_prot
305#define sk_net __sk_common.skc_net
306 socket_lock_t sk_lock;
307 struct sk_buff_head sk_receive_queue;
308 /*
309 * The backlog queue is special, it is always used with
310 * the per-socket spinlock held and requires low latency
311 * access. Therefore we special case it's implementation.
312 * Note : rmem_alloc is in this structure to fill a hole
313 * on 64bit arches, not because its logically part of
314 * backlog.
315 */
316 struct {
317 atomic_t rmem_alloc;
318 int len;
319 struct sk_buff *head;
320 struct sk_buff *tail;
321 } sk_backlog;
322#define sk_rmem_alloc sk_backlog.rmem_alloc
323 int sk_forward_alloc;
324#ifdef CONFIG_RPS
325 __u32 sk_rxhash;
326#endif
327 atomic_t sk_drops;
328 int sk_rcvbuf;
329 struct sk_filter __rcu *sk_filter;
330 struct socket_wq __rcu *sk_wq;
331
332#ifdef CONFIG_NET_DMA
333 struct sk_buff_head sk_async_wait_queue;
334#endif
335
336#ifdef CONFIG_XFRM
337 struct xfrm_policy *sk_policy[2];
338#endif
339 unsigned long sk_flags;
340 struct dst_entry *sk_dst_cache;
341 spinlock_t sk_dst_lock;
342 atomic_t sk_wmem_alloc;
343 atomic_t sk_omem_alloc;
344 int sk_sndbuf;
345 struct sk_buff_head sk_write_queue;
346 kmemcheck_bitfield_begin(flags);
347 unsigned int sk_shutdown : 2,
348 sk_no_check : 2,
349 sk_userlocks : 4,
350 sk_protocol : 8,
351#define SK_PROTOCOL_MAX U8_MAX //CVE-2015-8543
352 sk_type : 16;
353 kmemcheck_bitfield_end(flags);
354 int sk_wmem_queued;
355 gfp_t sk_allocation;
356 netdev_features_t sk_route_caps;
357 netdev_features_t sk_route_nocaps;
358 int sk_gso_type;
359 unsigned int sk_gso_max_size;
360 u16 sk_gso_max_segs;
361 int sk_rcvlowat;
362 unsigned long sk_lingertime;
363 struct sk_buff_head sk_error_queue;
364 struct proto *sk_prot_creator;
365 rwlock_t sk_callback_lock;
366 int sk_err,
367 sk_err_soft;
368 unsigned short sk_ack_backlog;
369 unsigned short sk_max_ack_backlog;
370 __u32 sk_priority;
371#ifdef CONFIG_CGROUPS
372 __u32 sk_cgrp_prioidx;
373#endif
374 struct pid *sk_peer_pid;
375 const struct cred *sk_peer_cred;
376 long sk_rcvtimeo;
377 long sk_sndtimeo;
378 void *sk_protinfo;
379 struct timer_list sk_timer;
380 ktime_t sk_stamp;
381 struct socket *sk_socket;
382 void *sk_user_data;
383 struct page *sk_sndmsg_page;
384 struct sk_buff *sk_send_head;
385 __u32 sk_sndmsg_off;
386 __s32 sk_peek_off;
387 int sk_write_pending;
388#ifdef CONFIG_SECURITY
389 void *sk_security;
390#endif
391 __u32 sk_mark;
392 u32 sk_classid;
393 struct cg_proto *sk_cgrp;
394 unsigned long sk_send_sum; //µ½Ä¿Ç°ÎªÖ¹·¢ËͳöÈ¥µÄÊý¾Ý°ü×ܸöÊý
395 int conn_list_init; //¼Ç¼µÄconnÁ´±íÊÇ·ñ³õʼ»¯
396 struct list_head conn_head; //connÁ´±í
397 void (*sk_state_change)(struct sock *sk);
398 void (*sk_data_ready)(struct sock *sk, int bytes);
399 void (*sk_write_space)(struct sock *sk);
400 void (*sk_error_report)(struct sock *sk);
401 int (*sk_backlog_rcv)(struct sock *sk,
402 struct sk_buff *skb);
403 void (*sk_destruct)(struct sock *sk);
xf.li6c8fc1e2023-08-12 00:11:09 -0700404#ifdef CONFIG_IPC_SOCKET
405 int ipc_flag;
406 int sk_fd;
407 struct sock* proxy_sock;
408 u8 closed;
409#endif
lh9ed821d2023-04-07 01:36:19 -0700410};
411
412static inline int sk_peek_offset(struct sock *sk, int flags)
413{
414 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
415 return sk->sk_peek_off;
416 else
417 return 0;
418}
419
420static inline void sk_peek_offset_bwd(struct sock *sk, int val)
421{
422 if (sk->sk_peek_off >= 0) {
423 if (sk->sk_peek_off >= val)
424 sk->sk_peek_off -= val;
425 else
426 sk->sk_peek_off = 0;
427 }
428}
429
430static inline void sk_peek_offset_fwd(struct sock *sk, int val)
431{
432 if (sk->sk_peek_off >= 0)
433 sk->sk_peek_off += val;
434}
435
436/*
437 * Hashed lists helper routines
438 */
439static inline struct sock *sk_entry(const struct hlist_node *node)
440{
441 return hlist_entry(node, struct sock, sk_node);
442}
443
444static inline struct sock *__sk_head(const struct hlist_head *head)
445{
446 return hlist_entry(head->first, struct sock, sk_node);
447}
448
449static inline struct sock *sk_head(const struct hlist_head *head)
450{
451 return hlist_empty(head) ? NULL : __sk_head(head);
452}
453
454static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
455{
456 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
457}
458
459static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
460{
461 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
462}
463
464static inline struct sock *sk_next(const struct sock *sk)
465{
466 return sk->sk_node.next ?
467 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
468}
469
470static inline struct sock *sk_nulls_next(const struct sock *sk)
471{
472 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
473 hlist_nulls_entry(sk->sk_nulls_node.next,
474 struct sock, sk_nulls_node) :
475 NULL;
476}
477
478static inline int sk_unhashed(const struct sock *sk)
479{
480 return hlist_unhashed(&sk->sk_node);
481}
482
483static inline int sk_hashed(const struct sock *sk)
484{
485 return !sk_unhashed(sk);
486}
487
488static __inline__ void sk_node_init(struct hlist_node *node)
489{
490 node->pprev = NULL;
491}
492
493static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
494{
495 node->pprev = NULL;
496}
497
498static __inline__ void __sk_del_node(struct sock *sk)
499{
500 __hlist_del(&sk->sk_node);
501}
502
503/* NB: equivalent to hlist_del_init_rcu */
504static __inline__ int __sk_del_node_init(struct sock *sk)
505{
506 if (sk_hashed(sk)) {
507 __sk_del_node(sk);
508 sk_node_init(&sk->sk_node);
509 return 1;
510 }
511 return 0;
512}
513
514/* Grab socket reference count. This operation is valid only
515 when sk is ALREADY grabbed f.e. it is found in hash table
516 or a list and the lookup is made under lock preventing hash table
517 modifications.
518 */
519
520static inline void sock_hold(struct sock *sk)
521{
522 atomic_inc(&sk->sk_refcnt);
xf.li6c8fc1e2023-08-12 00:11:09 -0700523#ifdef CONFIG_IPC_SOCKET
524 sk_soc_dbg("sock_hold sk=%x, sk->sk_refcnt=%d\n", sk, sk->sk_refcnt);
525#endif
lh9ed821d2023-04-07 01:36:19 -0700526}
527
528/* Ungrab socket in the context, which assumes that socket refcnt
529 cannot hit zero, f.e. it is true in context of any socketcall.
530 */
531static inline void __sock_put(struct sock *sk)
532{
xf.li6c8fc1e2023-08-12 00:11:09 -0700533#ifdef CONFIG_IPC_SOCKET
534 sk_soc_dbg("__sock_put sk=%x, sk->sk_refcnt=%d\n",sk, sk->sk_refcnt);
535#endif
lh9ed821d2023-04-07 01:36:19 -0700536 atomic_dec(&sk->sk_refcnt);
537}
538
539static __inline__ int sk_del_node_init(struct sock *sk)
540{
541 int rc = __sk_del_node_init(sk);
542
543 if (rc) {
544 /* paranoid for a while -acme */
545 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
546 __sock_put(sk);
547 }
548 return rc;
549}
550#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
551
552static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
553{
554 if (sk_hashed(sk)) {
555 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
556 return 1;
557 }
558 return 0;
559}
560
561static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
562{
563 int rc = __sk_nulls_del_node_init_rcu(sk);
564
565 if (rc) {
566 /* paranoid for a while -acme */
567 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
568 __sock_put(sk);
569 }
570 return rc;
571}
572
573static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
574{
575 hlist_add_head(&sk->sk_node, list);
576}
577
578static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
579{
580 sock_hold(sk);
581 __sk_add_node(sk, list);
582}
583
584static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
585{
586 sock_hold(sk);
587 hlist_add_head_rcu(&sk->sk_node, list);
588}
589
590static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
591{
592 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
593}
594
595static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
596{
597 sock_hold(sk);
598 __sk_nulls_add_node_rcu(sk, list);
599}
600
601static __inline__ void __sk_del_bind_node(struct sock *sk)
602{
603 __hlist_del(&sk->sk_bind_node);
604}
605
606static __inline__ void sk_add_bind_node(struct sock *sk,
607 struct hlist_head *list)
608{
609 hlist_add_head(&sk->sk_bind_node, list);
610}
611
612#define sk_for_each(__sk, node, list) \
613 hlist_for_each_entry(__sk, node, list, sk_node)
614#define sk_for_each_rcu(__sk, node, list) \
615 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
616#define sk_nulls_for_each(__sk, node, list) \
617 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
618#define sk_nulls_for_each_rcu(__sk, node, list) \
619 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
620#define sk_for_each_from(__sk, node) \
621 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
622 hlist_for_each_entry_from(__sk, node, sk_node)
623#define sk_nulls_for_each_from(__sk, node) \
624 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
625 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
626#define sk_for_each_safe(__sk, node, tmp, list) \
627 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
628#define sk_for_each_bound(__sk, node, list) \
629 hlist_for_each_entry(__sk, node, list, sk_bind_node)
630
631/* Sock flags */
632enum sock_flags {
633 SOCK_DEAD,
634 SOCK_DONE,
635 SOCK_URGINLINE,
636 SOCK_KEEPOPEN,
637 SOCK_LINGER,
638 SOCK_DESTROY,
639 SOCK_BROADCAST,
640 SOCK_TIMESTAMP,
641 SOCK_ZAPPED,
642 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
643 SOCK_DBG, /* %SO_DEBUG setting */
644 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
645 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
646 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
647 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
648 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
649 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
650 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
651 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
652 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
653 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
654 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
655 SOCK_FASYNC, /* fasync() active */
656 SOCK_RXQ_OVFL,
657 SOCK_ZEROCOPY, /* buffers from userspace */
658 SOCK_WIFI_STATUS, /* push wifi status to userspace */
659 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
660 * Will use last 4 bytes of packet sent from
661 * user-space instead.
662 */
xf.li6c8fc1e2023-08-12 00:11:09 -0700663#ifdef CONFIG_IPC_SOCKET
664 SOCK_IPCSOCK, /*Indicates whether it is a cross core socket */
665 SOCK_IPC_LOCAL,
666#endif
lh9ed821d2023-04-07 01:36:19 -0700667};
668
669static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
670{
671 nsk->sk_flags = osk->sk_flags;
672}
673
674static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
675{
676 __set_bit(flag, &sk->sk_flags);
677}
678
679static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
680{
681 __clear_bit(flag, &sk->sk_flags);
682}
683
684static inline int sock_flag(struct sock *sk, enum sock_flags flag)
685{
xf.li6c8fc1e2023-08-12 00:11:09 -0700686#ifdef CONFIG_IPC_SOCKET
687 if(unix_is_ipc_socket(sk)){
688 return sock_soc_test_flags(sk, flag);
689 }
690 else{
691#endif
lh9ed821d2023-04-07 01:36:19 -0700692 return test_bit(flag, &sk->sk_flags);
xf.li6c8fc1e2023-08-12 00:11:09 -0700693#ifdef CONFIG_IPC_SOCKET
694 }
695#endif
lh9ed821d2023-04-07 01:36:19 -0700696}
697
698static inline void sk_acceptq_removed(struct sock *sk)
699{
700 sk->sk_ack_backlog--;
701}
702
703static inline void sk_acceptq_added(struct sock *sk)
704{
705 sk->sk_ack_backlog++;
706}
707
708static inline int sk_acceptq_is_full(struct sock *sk)
709{
710 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
711}
712
713/*
714 * Compute minimal free write space needed to queue new packets.
715 */
716static inline int sk_stream_min_wspace(struct sock *sk)
717{
718 return sk->sk_wmem_queued >> 1;
719}
720
721static inline int sk_stream_wspace(struct sock *sk)
722{
723 return sk->sk_sndbuf - sk->sk_wmem_queued;
724}
725
726extern void sk_stream_write_space(struct sock *sk);
727
728static inline int sk_stream_memory_free(struct sock *sk)
729{
730 return sk->sk_wmem_queued < sk->sk_sndbuf;
731}
732
733/* OOB backlog add */
734static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
735{
736 /* dont let skb dst not refcounted, we are going to leave rcu lock */
737 skb_dst_force(skb);
738
739 if (!sk->sk_backlog.tail)
740 sk->sk_backlog.head = skb;
741 else
742 sk->sk_backlog.tail->next = skb;
743
744 sk->sk_backlog.tail = skb;
745 skb->next = NULL;
746}
747
748/*
749 * Take into account size of receive queue and backlog queue
750 * Do not take into account this skb truesize,
751 * to allow even a single big packet to come.
752 */
753static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb)
754{
755 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
756
757 return qsize > sk->sk_rcvbuf;
758}
759
760/* The per-socket spinlock must be held here. */
761static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb)
762{
763 if (sk_rcvqueues_full(sk, skb))
764 {
765 TCP_SOCK_TRACK(sk, TCP_RECV_BUFF_FULL);
766 return -ENOBUFS;
767 }
768
769 __sk_add_backlog(sk, skb);
770 sk->sk_backlog.len += skb->truesize;
771 return 0;
772}
773
774static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
775{
776 return sk->sk_backlog_rcv(sk, skb);
777}
778
779static inline void sock_rps_record_flow(const struct sock *sk)
780{
781#ifdef CONFIG_RPS
782 struct rps_sock_flow_table *sock_flow_table;
783
784 rcu_read_lock();
785 sock_flow_table = rcu_dereference(rps_sock_flow_table);
786 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
787 rcu_read_unlock();
788#endif
789}
790
791static inline void sock_rps_reset_flow(const struct sock *sk)
792{
793#ifdef CONFIG_RPS
794 struct rps_sock_flow_table *sock_flow_table;
795
796 rcu_read_lock();
797 sock_flow_table = rcu_dereference(rps_sock_flow_table);
798 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
799 rcu_read_unlock();
800#endif
801}
802
803static inline void sock_rps_save_rxhash(struct sock *sk,
804 const struct sk_buff *skb)
805{
806#ifdef CONFIG_RPS
807 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
808 sock_rps_reset_flow(sk);
809 sk->sk_rxhash = skb->rxhash;
810 }
811#endif
812}
813
814static inline void sock_rps_reset_rxhash(struct sock *sk)
815{
816#ifdef CONFIG_RPS
817 sock_rps_reset_flow(sk);
818 sk->sk_rxhash = 0;
819#endif
820}
821
822#define sk_wait_event(__sk, __timeo, __condition) \
823 ({ int __rc; \
824 release_sock(__sk); \
825 __rc = __condition; \
826 if (!__rc) { \
827 *(__timeo) = schedule_timeout(*(__timeo)); \
828 } \
829 lock_sock(__sk); \
830 __rc = __condition; \
831 __rc; \
832 })
833
834extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
835extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
836extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
837extern int sk_stream_error(struct sock *sk, int flags, int err);
838extern void sk_stream_kill_queues(struct sock *sk);
839
840extern int sk_wait_data(struct sock *sk, long *timeo);
841
842struct request_sock_ops;
843struct timewait_sock_ops;
844struct inet_hashinfo;
845struct raw_hashinfo;
846struct module;
847
848/*
849 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
850 * un-modified. Special care is taken when initializing object to zero.
851 */
852static inline void sk_prot_clear_nulls(struct sock *sk, int size)
853{
854 if (offsetof(struct sock, sk_node.next) != 0)
855 memset(sk, 0, offsetof(struct sock, sk_node.next));
856 memset(&sk->sk_node.pprev, 0,
857 size - offsetof(struct sock, sk_node.pprev));
858}
859
860/* Networking protocol blocks we attach to sockets.
861 * socket layer -> transport layer interface
862 * transport -> network interface is defined by struct inet_proto
863 */
864struct proto {
865 void (*close)(struct sock *sk,
866 long timeout);
867 int (*connect)(struct sock *sk,
868 struct sockaddr *uaddr,
869 int addr_len);
870 int (*disconnect)(struct sock *sk, int flags);
871
872 struct sock * (*accept) (struct sock *sk, int flags, int *err);
873
874 int (*ioctl)(struct sock *sk, int cmd,
875 unsigned long arg);
876 int (*init)(struct sock *sk);
877 void (*destroy)(struct sock *sk);
878 void (*shutdown)(struct sock *sk, int how);
879 int (*setsockopt)(struct sock *sk, int level,
880 int optname, char __user *optval,
881 unsigned int optlen);
882 int (*getsockopt)(struct sock *sk, int level,
883 int optname, char __user *optval,
884 int __user *option);
885#ifdef CONFIG_COMPAT
886 int (*compat_setsockopt)(struct sock *sk,
887 int level,
888 int optname, char __user *optval,
889 unsigned int optlen);
890 int (*compat_getsockopt)(struct sock *sk,
891 int level,
892 int optname, char __user *optval,
893 int __user *option);
894 int (*compat_ioctl)(struct sock *sk,
895 unsigned int cmd, unsigned long arg);
896#endif
897 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
898 struct msghdr *msg, size_t len);
899 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
900 struct msghdr *msg,
901 size_t len, int noblock, int flags,
902 int *addr_len);
903 int (*sendpage)(struct sock *sk, struct page *page,
904 int offset, size_t size, int flags);
905 int (*bind)(struct sock *sk,
906 struct sockaddr *uaddr, int addr_len);
907
908 int (*backlog_rcv) (struct sock *sk,
909 struct sk_buff *skb);
910
911 /* Keeping track of sk's, looking them up, and port selection methods. */
912 void (*hash)(struct sock *sk);
913 void (*unhash)(struct sock *sk);
914 void (*rehash)(struct sock *sk);
915 int (*get_port)(struct sock *sk, unsigned short snum);
916 void (*clear_sk)(struct sock *sk, int size);
917
918 /* Keeping track of sockets in use */
919#ifdef CONFIG_PROC_FS
920 unsigned int inuse_idx;
921#endif
922
923 /* Memory pressure */
924 void (*enter_memory_pressure)(struct sock *sk);
925 atomic_long_t *memory_allocated; /* Current allocated memory. */
926 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
927 /*
928 * Pressure flag: try to collapse.
929 * Technical note: it is used by multiple contexts non atomically.
930 * All the __sk_mem_schedule() is of this nature: accounting
931 * is strict, actions are advisory and have some latency.
932 */
933 int *memory_pressure;
934 long *sysctl_mem;
935 int *sysctl_wmem;
936 int *sysctl_rmem;
937 int max_header;
938 bool no_autobind;
939
940 struct kmem_cache *slab;
941 unsigned int obj_size;
942 int slab_flags;
943
944 struct percpu_counter *orphan_count;
945
946 struct request_sock_ops *rsk_prot;
947 struct timewait_sock_ops *twsk_prot;
948
949 union {
950 struct inet_hashinfo *hashinfo;
951 struct udp_table *udp_table;
952 struct raw_hashinfo *raw_hash;
953 } h;
954
955 struct module *owner;
956
957 char name[32];
958
959 struct list_head node;
960#ifdef SOCK_REFCNT_DEBUG
961 atomic_t socks;
962#endif
963#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
964 /*
965 * cgroup specific init/deinit functions. Called once for all
966 * protocols that implement it, from cgroups populate function.
967 * This function has to setup any files the protocol want to
968 * appear in the kmem cgroup filesystem.
969 */
970 int (*init_cgroup)(struct cgroup *cgrp,
971 struct cgroup_subsys *ss);
972 void (*destroy_cgroup)(struct cgroup *cgrp);
973 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
974#endif
975};
976
977struct cg_proto {
978 void (*enter_memory_pressure)(struct sock *sk);
979 struct res_counter *memory_allocated; /* Current allocated memory. */
980 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
981 int *memory_pressure;
982 long *sysctl_mem;
983 /*
984 * memcg field is used to find which memcg we belong directly
985 * Each memcg struct can hold more than one cg_proto, so container_of
986 * won't really cut.
987 *
988 * The elegant solution would be having an inverse function to
989 * proto_cgroup in struct proto, but that means polluting the structure
990 * for everybody, instead of just for memcg users.
991 */
992 struct mem_cgroup *memcg;
993};
994
995extern int proto_register(struct proto *prot, int alloc_slab);
996extern void proto_unregister(struct proto *prot);
997
998#ifdef SOCK_REFCNT_DEBUG
999static inline void sk_refcnt_debug_inc(struct sock *sk)
1000{
1001 atomic_inc(&sk->sk_prot->socks);
1002}
1003
1004static inline void sk_refcnt_debug_dec(struct sock *sk)
1005{
1006 atomic_dec(&sk->sk_prot->socks);
1007 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1008 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1009}
1010
1011static inline void sk_refcnt_debug_release(const struct sock *sk)
1012{
1013 if (atomic_read(&sk->sk_refcnt) != 1)
1014 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1015 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1016}
1017#else /* SOCK_REFCNT_DEBUG */
1018#define sk_refcnt_debug_inc(sk) do { } while (0)
1019#define sk_refcnt_debug_dec(sk) do { } while (0)
1020#define sk_refcnt_debug_release(sk) do { } while (0)
1021#endif /* SOCK_REFCNT_DEBUG */
1022
1023#if defined(CONFIG_CGROUP_MEM_RES_CTLR_KMEM) && defined(CONFIG_NET)
1024extern struct static_key memcg_socket_limit_enabled;
1025static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1026 struct cg_proto *cg_proto)
1027{
1028 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1029}
1030#define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
1031#else
1032#define mem_cgroup_sockets_enabled 0
1033static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1034 struct cg_proto *cg_proto)
1035{
1036 return NULL;
1037}
1038#endif
1039
1040
1041static inline bool sk_has_memory_pressure(const struct sock *sk)
1042{
1043 return sk->sk_prot->memory_pressure != NULL;
1044}
1045
1046static inline bool sk_under_memory_pressure(const struct sock *sk)
1047{
1048 if (!sk->sk_prot->memory_pressure)
1049 return false;
1050
1051 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1052 return !!*sk->sk_cgrp->memory_pressure;
1053
1054 return !!*sk->sk_prot->memory_pressure;
1055}
1056
1057static inline void sk_leave_memory_pressure(struct sock *sk)
1058{
1059 int *memory_pressure = sk->sk_prot->memory_pressure;
1060
1061 if (!memory_pressure)
1062 return;
1063
1064 if (*memory_pressure)
1065 *memory_pressure = 0;
1066
1067 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1068 struct cg_proto *cg_proto = sk->sk_cgrp;
1069 struct proto *prot = sk->sk_prot;
1070
1071 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1072 if (*cg_proto->memory_pressure)
1073 *cg_proto->memory_pressure = 0;
1074 }
1075
1076}
1077
1078static inline void sk_enter_memory_pressure(struct sock *sk)
1079{
1080 if (!sk->sk_prot->enter_memory_pressure)
1081 return;
1082
1083 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1084 struct cg_proto *cg_proto = sk->sk_cgrp;
1085 struct proto *prot = sk->sk_prot;
1086
1087 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1088 cg_proto->enter_memory_pressure(sk);
1089 }
1090
1091 sk->sk_prot->enter_memory_pressure(sk);
1092}
1093
1094static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1095{
1096 long *prot = sk->sk_prot->sysctl_mem;
1097 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1098 prot = sk->sk_cgrp->sysctl_mem;
1099 return prot[index];
1100}
1101
1102static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1103 unsigned long amt,
1104 int *parent_status)
1105{
1106 struct res_counter *fail;
1107 int ret;
1108
1109 ret = res_counter_charge_nofail(prot->memory_allocated,
1110 amt << PAGE_SHIFT, &fail);
1111 if (ret < 0)
1112 *parent_status = OVER_LIMIT;
1113}
1114
1115static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1116 unsigned long amt)
1117{
1118 res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
1119}
1120
1121static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1122{
1123 u64 ret;
1124 ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
1125 return ret >> PAGE_SHIFT;
1126}
1127
1128static inline long
1129sk_memory_allocated(const struct sock *sk)
1130{
1131 struct proto *prot = sk->sk_prot;
1132 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1133 return memcg_memory_allocated_read(sk->sk_cgrp);
1134
1135 return atomic_long_read(prot->memory_allocated);
1136}
1137
1138static inline long
1139sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
1140{
1141 struct proto *prot = sk->sk_prot;
1142
1143 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1144 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1145 /* update the root cgroup regardless */
1146 atomic_long_add_return(amt, prot->memory_allocated);
1147 return memcg_memory_allocated_read(sk->sk_cgrp);
1148 }
1149
1150 return atomic_long_add_return(amt, prot->memory_allocated);
1151}
1152
1153static inline void
1154sk_memory_allocated_sub(struct sock *sk, int amt)
1155{
1156 struct proto *prot = sk->sk_prot;
1157
1158 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1159 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1160
1161 atomic_long_sub(amt, prot->memory_allocated);
1162}
1163
1164static inline void sk_sockets_allocated_dec(struct sock *sk)
1165{
1166 struct proto *prot = sk->sk_prot;
1167
1168 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1169 struct cg_proto *cg_proto = sk->sk_cgrp;
1170
1171 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1172 percpu_counter_dec(cg_proto->sockets_allocated);
1173 }
1174
1175 percpu_counter_dec(prot->sockets_allocated);
1176}
1177
1178static inline void sk_sockets_allocated_inc(struct sock *sk)
1179{
1180 struct proto *prot = sk->sk_prot;
1181
1182 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1183 struct cg_proto *cg_proto = sk->sk_cgrp;
1184
1185 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1186 percpu_counter_inc(cg_proto->sockets_allocated);
1187 }
1188
1189 percpu_counter_inc(prot->sockets_allocated);
1190}
1191
1192static inline int
1193sk_sockets_allocated_read_positive(struct sock *sk)
1194{
1195 struct proto *prot = sk->sk_prot;
1196
1197 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1198 return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
1199
1200 return percpu_counter_read_positive(prot->sockets_allocated);
1201}
1202
1203static inline int
1204proto_sockets_allocated_sum_positive(struct proto *prot)
1205{
1206 return percpu_counter_sum_positive(prot->sockets_allocated);
1207}
1208
1209static inline long
1210proto_memory_allocated(struct proto *prot)
1211{
1212 return atomic_long_read(prot->memory_allocated);
1213}
1214
1215static inline bool
1216proto_memory_pressure(struct proto *prot)
1217{
1218 if (!prot->memory_pressure)
1219 return false;
1220 return !!*prot->memory_pressure;
1221}
1222
1223
1224#ifdef CONFIG_PROC_FS
1225/* Called with local bh disabled */
1226extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1227extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
1228#else
1229static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
1230 int inc)
1231{
1232}
1233#endif
1234
1235
1236/* With per-bucket locks this operation is not-atomic, so that
1237 * this version is not worse.
1238 */
1239static inline void __sk_prot_rehash(struct sock *sk)
1240{
1241 sk->sk_prot->unhash(sk);
1242 sk->sk_prot->hash(sk);
1243}
1244
1245void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1246
1247/* About 10 seconds */
1248#define SOCK_DESTROY_TIME (10*HZ)
1249
1250/* Sockets 0-1023 can't be bound to unless you are superuser */
1251#define PROT_SOCK 1024
1252
1253#define SHUTDOWN_MASK 3
1254#define RCV_SHUTDOWN 1
1255#define SEND_SHUTDOWN 2
1256
1257#define SOCK_SNDBUF_LOCK 1
1258#define SOCK_RCVBUF_LOCK 2
1259#define SOCK_BINDADDR_LOCK 4
1260#define SOCK_BINDPORT_LOCK 8
1261
1262/* sock_iocb: used to kick off async processing of socket ios */
1263struct sock_iocb {
1264 struct list_head list;
1265
1266 int flags;
1267 int size;
1268 struct socket *sock;
1269 struct sock *sk;
1270 struct scm_cookie *scm;
1271 struct msghdr *msg, async_msg;
1272 struct kiocb *kiocb;
1273};
1274
1275static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1276{
1277 return (struct sock_iocb *)iocb->private;
1278}
1279
1280static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1281{
1282 return si->kiocb;
1283}
1284
1285struct socket_alloc {
1286 struct socket socket;
1287 struct inode vfs_inode;
1288};
1289
1290static inline struct socket *SOCKET_I(struct inode *inode)
1291{
1292 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1293}
1294
1295static inline struct inode *SOCK_INODE(struct socket *socket)
1296{
1297 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1298}
1299
1300/*
1301 * Functions for memory accounting
1302 */
1303extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
1304extern void __sk_mem_reclaim(struct sock *sk);
1305
1306#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1307#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1308#define SK_MEM_SEND 0
1309#define SK_MEM_RECV 1
1310
1311static inline int sk_mem_pages(int amt)
1312{
1313 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1314}
1315
1316static inline int sk_has_account(struct sock *sk)
1317{
1318 /* return true if protocol supports memory accounting */
1319 return !!sk->sk_prot->memory_allocated;
1320}
1321
1322static inline int sk_wmem_schedule(struct sock *sk, int size)
1323{
1324 if (!sk_has_account(sk))
1325 return 1;
1326 return size <= sk->sk_forward_alloc ||
1327 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1328}
1329
1330static inline int sk_rmem_schedule(struct sock *sk, int size)
1331{
1332 if (!sk_has_account(sk))
1333 return 1;
1334 return size <= sk->sk_forward_alloc ||
1335 __sk_mem_schedule(sk, size, SK_MEM_RECV);
1336}
1337
1338static inline void sk_mem_reclaim(struct sock *sk)
1339{
1340 if (!sk_has_account(sk))
1341 return;
1342 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1343 __sk_mem_reclaim(sk);
1344}
1345
1346static inline void sk_mem_reclaim_partial(struct sock *sk)
1347{
1348 if (!sk_has_account(sk))
1349 return;
1350 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1351 __sk_mem_reclaim(sk);
1352}
1353
1354static inline void sk_mem_charge(struct sock *sk, int size)
1355{
1356 if (!sk_has_account(sk))
1357 return;
1358 sk->sk_forward_alloc -= size;
1359}
1360
1361static inline void sk_mem_uncharge(struct sock *sk, int size)
1362{
1363 if (!sk_has_account(sk))
1364 return;
1365 sk->sk_forward_alloc += size;
1366}
1367
1368static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1369{
1370 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1371 sk->sk_wmem_queued -= skb->truesize;
1372 sk_mem_uncharge(sk, skb->truesize);
1373 __kfree_skb(skb);
1374}
1375
1376/* Used by processes to "lock" a socket state, so that
1377 * interrupts and bottom half handlers won't change it
1378 * from under us. It essentially blocks any incoming
1379 * packets, so that we won't get any new data or any
1380 * packets that change the state of the socket.
1381 *
1382 * While locked, BH processing will add new packets to
1383 * the backlog queue. This queue is processed by the
1384 * owner of the socket lock right before it is released.
1385 *
1386 * Since ~2.3.5 it is also exclusive sleep lock serializing
1387 * accesses from user process context.
1388 */
1389#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1390
1391/*
1392 * Macro so as to not evaluate some arguments when
1393 * lockdep is not enabled.
1394 *
1395 * Mark both the sk_lock and the sk_lock.slock as a
1396 * per-address-family lock class.
1397 */
1398#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1399do { \
1400 sk->sk_lock.owned = 0; \
1401 init_waitqueue_head(&sk->sk_lock.wq); \
1402 spin_lock_init(&(sk)->sk_lock.slock); \
1403 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1404 sizeof((sk)->sk_lock)); \
1405 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1406 (skey), (sname)); \
1407 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1408} while (0)
1409
1410extern void lock_sock_nested(struct sock *sk, int subclass);
1411
1412static inline void lock_sock(struct sock *sk)
1413{
1414 lock_sock_nested(sk, 0);
1415}
1416
1417extern void release_sock(struct sock *sk);
1418
1419/* BH context may only use the following locking interface. */
1420#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1421#define bh_lock_sock_nested(__sk) \
1422 spin_lock_nested(&((__sk)->sk_lock.slock), \
1423 SINGLE_DEPTH_NESTING)
1424#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1425
1426extern bool lock_sock_fast(struct sock *sk);
1427/**
1428 * unlock_sock_fast - complement of lock_sock_fast
1429 * @sk: socket
1430 * @slow: slow mode
1431 *
1432 * fast unlock socket for user context.
1433 * If slow mode is on, we call regular release_sock()
1434 */
1435static inline void unlock_sock_fast(struct sock *sk, bool slow)
1436{
1437 if (slow)
1438 release_sock(sk);
1439 else
1440 spin_unlock_bh(&sk->sk_lock.slock);
1441}
1442
1443
1444extern struct sock *sk_alloc(struct net *net, int family,
1445 gfp_t priority,
1446 struct proto *prot);
1447extern void sk_free(struct sock *sk);
1448extern void sk_release_kernel(struct sock *sk);
1449extern struct sock *sk_clone_lock(const struct sock *sk,
1450 const gfp_t priority);
1451
1452extern struct sk_buff *sock_wmalloc(struct sock *sk,
1453 unsigned long size, int force,
1454 gfp_t priority);
1455extern struct sk_buff *sock_rmalloc(struct sock *sk,
1456 unsigned long size, int force,
1457 gfp_t priority);
1458extern void sock_wfree(struct sk_buff *skb);
1459extern void sock_rfree(struct sk_buff *skb);
1460
1461extern int sock_setsockopt(struct socket *sock, int level,
1462 int op, char __user *optval,
1463 unsigned int optlen);
1464
1465extern int sock_getsockopt(struct socket *sock, int level,
1466 int op, char __user *optval,
1467 int __user *optlen);
1468extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1469 unsigned long size,
1470 int noblock,
1471 int *errcode);
1472extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1473 unsigned long header_len,
1474 unsigned long data_len,
1475 int noblock,
1476 int *errcode);
1477extern void *sock_kmalloc(struct sock *sk, int size,
1478 gfp_t priority);
1479extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1480extern void sk_send_sigurg(struct sock *sk);
1481
1482#ifdef CONFIG_CGROUPS
1483extern void sock_update_classid(struct sock *sk);
1484#else
1485static inline void sock_update_classid(struct sock *sk)
1486{
1487}
1488#endif
1489
1490/*
1491 * Functions to fill in entries in struct proto_ops when a protocol
1492 * does not implement a particular function.
1493 */
1494extern int sock_no_bind(struct socket *,
1495 struct sockaddr *, int);
1496extern int sock_no_connect(struct socket *,
1497 struct sockaddr *, int, int);
1498extern int sock_no_socketpair(struct socket *,
1499 struct socket *);
1500extern int sock_no_accept(struct socket *,
1501 struct socket *, int);
1502extern int sock_no_getname(struct socket *,
1503 struct sockaddr *, int *, int);
1504extern unsigned int sock_no_poll(struct file *, struct socket *,
1505 struct poll_table_struct *);
1506extern int sock_no_ioctl(struct socket *, unsigned int,
1507 unsigned long);
1508extern int sock_no_listen(struct socket *, int);
1509extern int sock_no_shutdown(struct socket *, int);
1510extern int sock_no_getsockopt(struct socket *, int , int,
1511 char __user *, int __user *);
1512extern int sock_no_setsockopt(struct socket *, int, int,
1513 char __user *, unsigned int);
1514extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1515 struct msghdr *, size_t);
1516extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1517 struct msghdr *, size_t, int);
1518extern int sock_no_mmap(struct file *file,
1519 struct socket *sock,
1520 struct vm_area_struct *vma);
1521extern ssize_t sock_no_sendpage(struct socket *sock,
1522 struct page *page,
1523 int offset, size_t size,
1524 int flags);
1525
1526/*
1527 * Functions to fill in entries in struct proto_ops when a protocol
1528 * uses the inet style.
1529 */
1530extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1531 char __user *optval, int __user *optlen);
1532extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1533 struct msghdr *msg, size_t size, int flags);
1534extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
1535 char __user *optval, unsigned int optlen);
1536extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1537 int optname, char __user *optval, int __user *optlen);
1538extern int compat_sock_common_setsockopt(struct socket *sock, int level,
1539 int optname, char __user *optval, unsigned int optlen);
1540
1541extern void sk_common_release(struct sock *sk);
1542
1543/*
1544 * Default socket callbacks and setup code
1545 */
1546
1547/* Initialise core socket variables */
1548extern void sock_init_data(struct socket *sock, struct sock *sk);
1549
1550extern void sk_filter_release_rcu(struct rcu_head *rcu);
1551
1552/**
1553 * sk_filter_release - release a socket filter
1554 * @fp: filter to remove
1555 *
1556 * Remove a filter from a socket and release its resources.
1557 */
1558
1559static inline void sk_filter_release(struct sk_filter *fp)
1560{
1561 if (atomic_dec_and_test(&fp->refcnt))
1562 call_rcu(&fp->rcu, sk_filter_release_rcu);
1563}
1564
1565static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1566{
1567 unsigned int size = sk_filter_len(fp);
1568
1569 atomic_sub(size, &sk->sk_omem_alloc);
1570 sk_filter_release(fp);
1571}
1572
1573static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1574{
1575 atomic_inc(&fp->refcnt);
1576 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1577}
1578
1579/*
1580 * Socket reference counting postulates.
1581 *
1582 * * Each user of socket SHOULD hold a reference count.
1583 * * Each access point to socket (an hash table bucket, reference from a list,
1584 * running timer, skb in flight MUST hold a reference count.
1585 * * When reference count hits 0, it means it will never increase back.
1586 * * When reference count hits 0, it means that no references from
1587 * outside exist to this socket and current process on current CPU
1588 * is last user and may/should destroy this socket.
1589 * * sk_free is called from any context: process, BH, IRQ. When
1590 * it is called, socket has no references from outside -> sk_free
1591 * may release descendant resources allocated by the socket, but
1592 * to the time when it is called, socket is NOT referenced by any
1593 * hash tables, lists etc.
1594 * * Packets, delivered from outside (from network or from another process)
1595 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1596 * when they sit in queue. Otherwise, packets will leak to hole, when
1597 * socket is looked up by one cpu and unhasing is made by another CPU.
1598 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1599 * (leak to backlog). Packet socket does all the processing inside
1600 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1601 * use separate SMP lock, so that they are prone too.
1602 */
1603
1604/* Ungrab socket and destroy it, if it was the last reference. */
1605//ʵÏַŵ½sock.cÖУ¬ÒÔ±ãÌí¼Ó×ÔÑдúÂë
1606void sock_put(struct sock *sk);
1607#if 0
1608static inline void sock_put(struct sock *sk)
1609{
1610 if (atomic_dec_and_test(&sk->sk_refcnt))
1611 sk_free(sk);
1612}
1613#endif
1614extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1615 const int nested);
1616
1617static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1618{
1619 sk->sk_tx_queue_mapping = tx_queue;
1620}
1621
1622static inline void sk_tx_queue_clear(struct sock *sk)
1623{
1624 sk->sk_tx_queue_mapping = -1;
1625}
1626
1627static inline int sk_tx_queue_get(const struct sock *sk)
1628{
1629 return sk ? sk->sk_tx_queue_mapping : -1;
1630}
1631
1632static inline void sk_set_socket(struct sock *sk, struct socket *sock)
xf.li6c8fc1e2023-08-12 00:11:09 -07001633{
1634#ifdef CONFIG_IPC_SOCKET
1635 sk_soc_dbg("sk_set_socket sk=%x, sock=%x \n", sk, sock);
1636#endif
lh9ed821d2023-04-07 01:36:19 -07001637 sk_tx_queue_clear(sk);
1638 sk->sk_socket = sock;
1639}
1640
1641static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1642{
1643 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1644 return &rcu_dereference_raw(sk->sk_wq)->wait;
1645}
1646/* Detach socket from process context.
1647 * Announce socket dead, detach it from wait queue and inode.
1648 * Note that parent inode held reference count on this struct sock,
1649 * we do not release it in this function, because protocol
1650 * probably wants some additional cleanups or even continuing
1651 * to work with this socket (TCP).
1652 */
1653static inline void sock_orphan(struct sock *sk)
1654{
1655 write_lock_bh(&sk->sk_callback_lock);
1656 sock_set_flag(sk, SOCK_DEAD);
1657 sk_set_socket(sk, NULL);
1658 sk->sk_wq = NULL;
1659 write_unlock_bh(&sk->sk_callback_lock);
1660}
1661
1662static inline void sock_graft(struct sock *sk, struct socket *parent)
1663{
1664 write_lock_bh(&sk->sk_callback_lock);
1665 sk->sk_wq = parent->wq;
1666 parent->sk = sk;
1667 sk_set_socket(sk, parent);
1668 security_sock_graft(sk, parent);
1669 write_unlock_bh(&sk->sk_callback_lock);
1670}
1671
1672extern int sock_i_uid(struct sock *sk);
1673extern unsigned long sock_i_ino(struct sock *sk);
1674
1675static inline struct dst_entry *
1676__sk_dst_get(struct sock *sk)
1677{
1678 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
1679 lockdep_is_held(&sk->sk_lock.slock));
1680}
1681
1682static inline struct dst_entry *
1683sk_dst_get(struct sock *sk)
1684{
1685 struct dst_entry *dst;
1686
1687 rcu_read_lock();
1688 dst = rcu_dereference(sk->sk_dst_cache);
1689 if (dst)
1690 dst_hold(dst);
1691 rcu_read_unlock();
1692 return dst;
1693}
1694
1695extern void sk_reset_txq(struct sock *sk);
1696
1697static inline void dst_negative_advice(struct sock *sk)
1698{
1699 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1700
1701 if (dst && dst->ops->negative_advice) {
1702 ndst = dst->ops->negative_advice(dst);
1703
1704 if (ndst != dst) {
1705 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1706 sk_reset_txq(sk);
1707 }
1708 }
1709}
1710
1711static inline void
1712__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1713{
1714 struct dst_entry *old_dst;
1715
1716 sk_tx_queue_clear(sk);
1717 /*
1718 * This can be called while sk is owned by the caller only,
1719 * with no state that can be checked in a rcu_dereference_check() cond
1720 */
1721 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1722 rcu_assign_pointer(sk->sk_dst_cache, dst);
1723 dst_release(old_dst);
1724}
1725
1726static inline void
1727sk_dst_set(struct sock *sk, struct dst_entry *dst)
1728{
1729 spin_lock(&sk->sk_dst_lock);
1730 __sk_dst_set(sk, dst);
1731 spin_unlock(&sk->sk_dst_lock);
1732}
1733
1734static inline void
1735__sk_dst_reset(struct sock *sk)
1736{
1737 __sk_dst_set(sk, NULL);
1738}
1739
1740static inline void
1741sk_dst_reset(struct sock *sk)
1742{
1743 spin_lock(&sk->sk_dst_lock);
1744 __sk_dst_reset(sk);
1745 spin_unlock(&sk->sk_dst_lock);
1746}
1747
1748extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1749
1750extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1751
1752static inline int sk_can_gso(const struct sock *sk)
1753{
1754 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1755}
1756
1757extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1758
1759static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
1760{
1761 sk->sk_route_nocaps |= flags;
1762 sk->sk_route_caps &= ~flags;
1763}
1764
1765static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1766 char __user *from, char *to,
1767 int copy, int offset)
1768{
1769 if (skb->ip_summed == CHECKSUM_NONE) {
1770 int err = 0;
1771 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1772 if (err)
1773 return err;
1774 skb->csum = csum_block_add(skb->csum, csum, offset);
1775 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1776 if (!access_ok(VERIFY_READ, from, copy) ||
1777 __copy_from_user_nocache(to, from, copy))
1778 return -EFAULT;
1779 } else if (copy_from_user(to, from, copy))
1780 return -EFAULT;
1781
1782 return 0;
1783}
1784
1785static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1786 char __user *from, int copy)
1787{
1788 int err, offset = skb->len;
1789
1790 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1791 copy, offset);
1792 if (err)
1793 __skb_trim(skb, offset);
1794
1795 return err;
1796}
1797
1798static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1799 struct sk_buff *skb,
1800 struct page *page,
1801 int off, int copy)
1802{
1803 int err;
1804
1805 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1806 copy, skb->len);
1807 if (err)
1808 return err;
1809
1810 skb->len += copy;
1811 skb->data_len += copy;
1812 skb->truesize += copy;
1813 sk->sk_wmem_queued += copy;
1814 sk_mem_charge(sk, copy);
1815 return 0;
1816}
1817
1818static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1819 struct sk_buff *skb, struct page *page,
1820 int off, int copy)
1821{
1822 if (skb->ip_summed == CHECKSUM_NONE) {
1823 int err = 0;
1824 __wsum csum = csum_and_copy_from_user(from,
1825 page_address(page) + off,
1826 copy, 0, &err);
1827 if (err)
1828 return err;
1829 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1830 } else if (copy_from_user(page_address(page) + off, from, copy))
1831 return -EFAULT;
1832
1833 skb->len += copy;
1834 skb->data_len += copy;
1835 skb->truesize += copy;
1836 sk->sk_wmem_queued += copy;
1837 sk_mem_charge(sk, copy);
1838 return 0;
1839}
1840
1841/**
1842 * sk_wmem_alloc_get - returns write allocations
1843 * @sk: socket
1844 *
1845 * Returns sk_wmem_alloc minus initial offset of one
1846 */
1847static inline int sk_wmem_alloc_get(const struct sock *sk)
1848{
1849 return atomic_read(&sk->sk_wmem_alloc) - 1;
1850}
1851
1852/**
1853 * sk_rmem_alloc_get - returns read allocations
1854 * @sk: socket
1855 *
1856 * Returns sk_rmem_alloc
1857 */
1858static inline int sk_rmem_alloc_get(const struct sock *sk)
1859{
1860 return atomic_read(&sk->sk_rmem_alloc);
1861}
1862
1863/**
1864 * sk_has_allocations - check if allocations are outstanding
1865 * @sk: socket
1866 *
1867 * Returns true if socket has write or read allocations
1868 */
1869static inline int sk_has_allocations(const struct sock *sk)
1870{
1871 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1872}
1873
1874/**
1875 * wq_has_sleeper - check if there are any waiting processes
1876 * @wq: struct socket_wq
1877 *
1878 * Returns true if socket_wq has waiting processes
1879 *
1880 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1881 * barrier call. They were added due to the race found within the tcp code.
1882 *
1883 * Consider following tcp code paths:
1884 *
1885 * CPU1 CPU2
1886 *
1887 * sys_select receive packet
1888 * ... ...
1889 * __add_wait_queue update tp->rcv_nxt
1890 * ... ...
1891 * tp->rcv_nxt check sock_def_readable
1892 * ... {
1893 * schedule rcu_read_lock();
1894 * wq = rcu_dereference(sk->sk_wq);
1895 * if (wq && waitqueue_active(&wq->wait))
1896 * wake_up_interruptible(&wq->wait)
1897 * ...
1898 * }
1899 *
1900 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1901 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1902 * could then endup calling schedule and sleep forever if there are no more
1903 * data on the socket.
1904 *
1905 */
1906static inline bool wq_has_sleeper(struct socket_wq *wq)
1907{
1908
1909 /*
1910 * We need to be sure we are in sync with the
1911 * add_wait_queue modifications to the wait queue.
1912 *
1913 * This memory barrier is paired in the sock_poll_wait.
1914 */
1915 smp_mb();
1916 return wq && waitqueue_active(&wq->wait);
1917}
1918
1919/**
1920 * sock_poll_wait - place memory barrier behind the poll_wait call.
1921 * @filp: file
1922 * @wait_address: socket wait queue
1923 * @p: poll_table
1924 *
1925 * See the comments in the wq_has_sleeper function.
1926 */
1927static inline void sock_poll_wait(struct file *filp,
1928 wait_queue_head_t *wait_address, poll_table *p)
1929{
1930 if (!poll_does_not_wait(p) && wait_address) {
1931 poll_wait(filp, wait_address, p);
1932 /*
1933 * We need to be sure we are in sync with the
1934 * socket flags modification.
1935 *
1936 * This memory barrier is paired in the wq_has_sleeper.
1937 */
1938 smp_mb();
1939 }
1940}
1941
1942/*
1943 * Queue a received datagram if it will fit. Stream and sequenced
1944 * protocols can't normally use this as they need to fit buffers in
1945 * and play with them.
1946 *
1947 * Inlined as it's very short and called for pretty much every
1948 * packet ever received.
1949 */
1950
1951static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1952{
1953 skb_orphan(skb);
1954 skb->sk = sk;
1955 skb->destructor = sock_wfree;
1956 /*
1957 * We used to take a refcount on sk, but following operation
1958 * is enough to guarantee sk_free() wont free this sock until
1959 * all in-flight packets are completed
1960 */
1961 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1962}
1963
1964static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1965{
1966 skb_orphan(skb);
1967 skb->sk = sk;
1968 skb->destructor = sock_rfree;
1969 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1970 sk_mem_charge(sk, skb->truesize);
1971}
1972
1973extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1974 unsigned long expires);
1975
1976extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1977
1978extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1979
1980extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1981
1982/*
1983 * Recover an error report and clear atomically
1984 */
1985
1986static inline int sock_error(struct sock *sk)
1987{
1988 int err;
1989 if (likely(!sk->sk_err))
1990 return 0;
1991 err = xchg(&sk->sk_err, 0);
1992 return -err;
1993}
1994
1995static inline unsigned long sock_wspace(struct sock *sk)
1996{
1997 int amt = 0;
1998
1999 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2000 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
2001 if (amt < 0)
2002 amt = 0;
2003 }
2004 return amt;
2005}
2006
2007static inline void sk_wake_async(struct sock *sk, int how, int band)
2008{
2009 if (sock_flag(sk, SOCK_FASYNC))
2010 sock_wake_async(sk->sk_socket, how, band);
2011}
2012
2013#define SOCK_MIN_SNDBUF 2048
2014/*
2015 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
2016 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
2017 */
2018#define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
2019
2020static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2021{
2022 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
2023 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
2024 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
2025 }
2026}
2027
2028struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
2029
2030static inline struct page *sk_stream_alloc_page(struct sock *sk)
2031{
2032 struct page *page = NULL;
2033
2034 page = alloc_pages(sk->sk_allocation, 0);
2035 if (!page) {
2036 sk_enter_memory_pressure(sk);
2037 sk_stream_moderate_sndbuf(sk);
2038 }
2039 netslab_inc(SOCK_ALLOC_PAGES);
2040 return page;
2041}
2042
2043/*
2044 * Default write policy as shown to user space via poll/select/SIGIO
2045 */
2046static inline int sock_writeable(const struct sock *sk)
2047{
2048 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
2049}
2050
2051static inline gfp_t gfp_any(void)
2052{
2053 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
2054}
2055
2056static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
2057{
2058 return noblock ? 0 : sk->sk_rcvtimeo;
2059}
2060
2061static inline long sock_sndtimeo(const struct sock *sk, int noblock)
2062{
2063 return noblock ? 0 : sk->sk_sndtimeo;
2064}
2065
2066static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2067{
2068 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2069}
2070
2071/* Alas, with timeout socket operations are not restartable.
2072 * Compare this to poll().
2073 */
2074static inline int sock_intr_errno(long timeo)
2075{
2076 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2077}
2078
2079extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2080 struct sk_buff *skb);
2081extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2082 struct sk_buff *skb);
2083
2084static __inline__ void
2085sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2086{
2087 ktime_t kt = skb->tstamp;
2088 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
2089
2090 /*
2091 * generate control messages if
2092 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2093 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2094 * - software time stamp available and wanted
2095 * (SOCK_TIMESTAMPING_SOFTWARE)
2096 * - hardware time stamps available and wanted
2097 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2098 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2099 */
2100 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2101 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2102 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2103 (hwtstamps->hwtstamp.tv64 &&
2104 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2105 (hwtstamps->syststamp.tv64 &&
2106 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
2107 __sock_recv_timestamp(msg, sk, skb);
2108 else
2109 sk->sk_stamp = kt;
2110
2111 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2112 __sock_recv_wifi_status(msg, sk, skb);
2113}
2114
2115extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2116 struct sk_buff *skb);
2117
2118static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2119 struct sk_buff *skb)
2120{
2121#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2122 (1UL << SOCK_RCVTSTAMP) | \
2123 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2124 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
2125 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
2126 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2127
2128 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2129 __sock_recv_ts_and_drops(msg, sk, skb);
2130 else
2131 sk->sk_stamp = skb->tstamp;
2132}
2133
2134/**
2135 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2136 * @sk: socket sending this packet
2137 * @tx_flags: filled with instructions for time stamping
2138 *
2139 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
2140 * parameters are invalid.
2141 */
2142extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
2143
2144/**
2145 * sk_eat_skb - Release a skb if it is no longer needed
2146 * @sk: socket to eat this skb from
2147 * @skb: socket buffer to eat
2148 * @copied_early: flag indicating whether DMA operations copied this data early
2149 *
2150 * This routine must be called with interrupts disabled or with the socket
2151 * locked so that the sk_buff queue operation is ok.
2152*/
2153#ifdef CONFIG_NET_DMA
2154static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
2155{
2156 __skb_unlink(skb, &sk->sk_receive_queue);
2157 if (!copied_early)
2158 __kfree_skb(skb);
2159 else
2160 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2161}
2162#else
2163static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
2164{
2165 __skb_unlink(skb, &sk->sk_receive_queue);
2166 __kfree_skb(skb);
2167}
2168#endif
2169
2170static inline
2171struct net *sock_net(const struct sock *sk)
2172{
2173 return read_pnet(&sk->sk_net);
2174}
2175
2176static inline
2177void sock_net_set(struct sock *sk, struct net *net)
2178{
2179 write_pnet(&sk->sk_net, net);
2180}
2181
2182/*
2183 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
2184 * They should not hold a reference to a namespace in order to allow
2185 * to stop it.
2186 * Sockets after sk_change_net should be released using sk_release_kernel
2187 */
2188static inline void sk_change_net(struct sock *sk, struct net *net)
2189{
2190 put_net(sock_net(sk));
2191 sock_net_set(sk, hold_net(net));
2192}
2193
2194static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2195{
2196 if (unlikely(skb->sk)) {
2197 struct sock *sk = skb->sk;
2198
2199 skb->destructor = NULL;
2200 skb->sk = NULL;
2201 return sk;
2202 }
2203 return NULL;
2204}
2205
2206extern void sock_enable_timestamp(struct sock *sk, int flag);
2207extern int sock_get_timestamp(struct sock *, struct timeval __user *);
2208extern int sock_get_timestampns(struct sock *, struct timespec __user *);
2209
2210/*
2211 * Enable debug/info messages
2212 */
2213extern int net_msg_warn;
2214#define NETDEBUG(fmt, args...) \
2215 do { if (net_msg_warn) printk(fmt,##args); } while (0)
2216
2217#define LIMIT_NETDEBUG(fmt, args...) \
2218 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
2219
2220extern __u32 sysctl_wmem_max;
2221extern __u32 sysctl_rmem_max;
2222
2223extern void sk_init(void);
2224
2225extern int sysctl_optmem_max;
2226
2227extern __u32 sysctl_wmem_default;
2228extern __u32 sysctl_rmem_default;
2229
2230#endif /* _SOCK_H */