aboutsummaryrefslogtreecommitdiffstats
path: root/include/net/sock.h
diff options
context:
space:
mode:
Diffstat (limited to 'include/net/sock.h')
-rw-r--r--include/net/sock.h1297
1 files changed, 1297 insertions, 0 deletions
diff --git a/include/net/sock.h b/include/net/sock.h
new file mode 100644
index 000000000000..be81cabd0da3
--- /dev/null
+++ b/include/net/sock.h
@@ -0,0 +1,1297 @@
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, <bir7@leland.Stanford.Edu>
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/config.h>
44#include <linux/list.h>
45#include <linux/timer.h>
46#include <linux/cache.h>
47#include <linux/module.h>
48#include <linux/netdevice.h>
49#include <linux/skbuff.h> /* struct sk_buff */
50#include <linux/security.h>
51
52#include <linux/filter.h>
53
54#include <asm/atomic.h>
55#include <net/dst.h>
56#include <net/checksum.h>
57
58/*
59 * This structure really needs to be cleaned up.
60 * Most of it is for TCP, and not used by any of
61 * the other protocols.
62 */
63
64/* Define this to get the SOCK_DBG debugging facility. */
65#define SOCK_DEBUGGING
66#ifdef SOCK_DEBUGGING
67#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
68 printk(KERN_DEBUG msg); } while (0)
69#else
70#define SOCK_DEBUG(sk, msg...) do { } while (0)
71#endif
72
73/* This is the per-socket lock. The spinlock provides a synchronization
74 * between user contexts and software interrupt processing, whereas the
75 * mini-semaphore synchronizes multiple users amongst themselves.
76 */
77struct sock_iocb;
78typedef struct {
79 spinlock_t slock;
80 struct sock_iocb *owner;
81 wait_queue_head_t wq;
82} socket_lock_t;
83
84#define sock_lock_init(__sk) \
85do { spin_lock_init(&((__sk)->sk_lock.slock)); \
86 (__sk)->sk_lock.owner = NULL; \
87 init_waitqueue_head(&((__sk)->sk_lock.wq)); \
88} while(0)
89
90struct sock;
91
92/**
93 * struct sock_common - minimal network layer representation of sockets
94 * @skc_family - network address family
95 * @skc_state - Connection state
96 * @skc_reuse - %SO_REUSEADDR setting
97 * @skc_bound_dev_if - bound device index if != 0
98 * @skc_node - main hash linkage for various protocol lookup tables
99 * @skc_bind_node - bind hash linkage for various protocol lookup tables
100 * @skc_refcnt - reference count
101 *
102 * This is the minimal network layer representation of sockets, the header
103 * for struct sock and struct tcp_tw_bucket.
104 */
105struct sock_common {
106 unsigned short skc_family;
107 volatile unsigned char skc_state;
108 unsigned char skc_reuse;
109 int skc_bound_dev_if;
110 struct hlist_node skc_node;
111 struct hlist_node skc_bind_node;
112 atomic_t skc_refcnt;
113};
114
115/**
116 * struct sock - network layer representation of sockets
117 * @__sk_common - shared layout with tcp_tw_bucket
118 * @sk_shutdown - mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
119 * @sk_userlocks - %SO_SNDBUF and %SO_RCVBUF settings
120 * @sk_lock - synchronizer
121 * @sk_rcvbuf - size of receive buffer in bytes
122 * @sk_sleep - sock wait queue
123 * @sk_dst_cache - destination cache
124 * @sk_dst_lock - destination cache lock
125 * @sk_policy - flow policy
126 * @sk_rmem_alloc - receive queue bytes committed
127 * @sk_receive_queue - incoming packets
128 * @sk_wmem_alloc - transmit queue bytes committed
129 * @sk_write_queue - Packet sending queue
130 * @sk_omem_alloc - "o" is "option" or "other"
131 * @sk_wmem_queued - persistent queue size
132 * @sk_forward_alloc - space allocated forward
133 * @sk_allocation - allocation mode
134 * @sk_sndbuf - size of send buffer in bytes
135 * @sk_flags - %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
136 * @sk_no_check - %SO_NO_CHECK setting, wether or not checkup packets
137 * @sk_route_caps - route capabilities (e.g. %NETIF_F_TSO)
138 * @sk_lingertime - %SO_LINGER l_linger setting
139 * @sk_hashent - hash entry in several tables (e.g. tcp_ehash)
140 * @sk_backlog - always used with the per-socket spinlock held
141 * @sk_callback_lock - used with the callbacks in the end of this struct
142 * @sk_error_queue - rarely used
143 * @sk_prot - protocol handlers inside a network family
144 * @sk_err - last error
145 * @sk_err_soft - errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
146 * @sk_ack_backlog - current listen backlog
147 * @sk_max_ack_backlog - listen backlog set in listen()
148 * @sk_priority - %SO_PRIORITY setting
149 * @sk_type - socket type (%SOCK_STREAM, etc)
150 * @sk_protocol - which protocol this socket belongs in this network family
151 * @sk_peercred - %SO_PEERCRED setting
152 * @sk_rcvlowat - %SO_RCVLOWAT setting
153 * @sk_rcvtimeo - %SO_RCVTIMEO setting
154 * @sk_sndtimeo - %SO_SNDTIMEO setting
155 * @sk_filter - socket filtering instructions
156 * @sk_protinfo - private area, net family specific, when not using slab
157 * @sk_timer - sock cleanup timer
158 * @sk_stamp - time stamp of last packet received
159 * @sk_socket - Identd and reporting IO signals
160 * @sk_user_data - RPC layer private data
161 * @sk_sndmsg_page - cached page for sendmsg
162 * @sk_sndmsg_off - cached offset for sendmsg
163 * @sk_send_head - front of stuff to transmit
164 * @sk_write_pending - a write to stream socket waits to start
165 * @sk_state_change - callback to indicate change in the state of the sock
166 * @sk_data_ready - callback to indicate there is data to be processed
167 * @sk_write_space - callback to indicate there is bf sending space available
168 * @sk_error_report - callback to indicate errors (e.g. %MSG_ERRQUEUE)
169 * @sk_backlog_rcv - callback to process the backlog
170 * @sk_destruct - called at sock freeing time, i.e. when all refcnt == 0
171 */
172struct sock {
173 /*
174 * Now struct tcp_tw_bucket also uses sock_common, so please just
175 * don't add nothing before this first member (__sk_common) --acme
176 */
177 struct sock_common __sk_common;
178#define sk_family __sk_common.skc_family
179#define sk_state __sk_common.skc_state
180#define sk_reuse __sk_common.skc_reuse
181#define sk_bound_dev_if __sk_common.skc_bound_dev_if
182#define sk_node __sk_common.skc_node
183#define sk_bind_node __sk_common.skc_bind_node
184#define sk_refcnt __sk_common.skc_refcnt
185 unsigned char sk_shutdown : 2,
186 sk_no_check : 2,
187 sk_userlocks : 4;
188 unsigned char sk_protocol;
189 unsigned short sk_type;
190 int sk_rcvbuf;
191 socket_lock_t sk_lock;
192 wait_queue_head_t *sk_sleep;
193 struct dst_entry *sk_dst_cache;
194 struct xfrm_policy *sk_policy[2];
195 rwlock_t sk_dst_lock;
196 atomic_t sk_rmem_alloc;
197 atomic_t sk_wmem_alloc;
198 atomic_t sk_omem_alloc;
199 struct sk_buff_head sk_receive_queue;
200 struct sk_buff_head sk_write_queue;
201 int sk_wmem_queued;
202 int sk_forward_alloc;
203 unsigned int sk_allocation;
204 int sk_sndbuf;
205 int sk_route_caps;
206 int sk_hashent;
207 unsigned long sk_flags;
208 unsigned long sk_lingertime;
209 /*
210 * The backlog queue is special, it is always used with
211 * the per-socket spinlock held and requires low latency
212 * access. Therefore we special case it's implementation.
213 */
214 struct {
215 struct sk_buff *head;
216 struct sk_buff *tail;
217 } sk_backlog;
218 struct sk_buff_head sk_error_queue;
219 struct proto *sk_prot;
220 rwlock_t sk_callback_lock;
221 int sk_err,
222 sk_err_soft;
223 unsigned short sk_ack_backlog;
224 unsigned short sk_max_ack_backlog;
225 __u32 sk_priority;
226 struct ucred sk_peercred;
227 int sk_rcvlowat;
228 long sk_rcvtimeo;
229 long sk_sndtimeo;
230 struct sk_filter *sk_filter;
231 void *sk_protinfo;
232 struct timer_list sk_timer;
233 struct timeval sk_stamp;
234 struct socket *sk_socket;
235 void *sk_user_data;
236 struct page *sk_sndmsg_page;
237 struct sk_buff *sk_send_head;
238 __u32 sk_sndmsg_off;
239 int sk_write_pending;
240 void *sk_security;
241 void (*sk_state_change)(struct sock *sk);
242 void (*sk_data_ready)(struct sock *sk, int bytes);
243 void (*sk_write_space)(struct sock *sk);
244 void (*sk_error_report)(struct sock *sk);
245 int (*sk_backlog_rcv)(struct sock *sk,
246 struct sk_buff *skb);
247 void (*sk_destruct)(struct sock *sk);
248};
249
250/*
251 * Hashed lists helper routines
252 */
253static inline struct sock *__sk_head(struct hlist_head *head)
254{
255 return hlist_entry(head->first, struct sock, sk_node);
256}
257
258static inline struct sock *sk_head(struct hlist_head *head)
259{
260 return hlist_empty(head) ? NULL : __sk_head(head);
261}
262
263static inline struct sock *sk_next(struct sock *sk)
264{
265 return sk->sk_node.next ?
266 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
267}
268
269static inline int sk_unhashed(struct sock *sk)
270{
271 return hlist_unhashed(&sk->sk_node);
272}
273
274static inline int sk_hashed(struct sock *sk)
275{
276 return sk->sk_node.pprev != NULL;
277}
278
279static __inline__ void sk_node_init(struct hlist_node *node)
280{
281 node->pprev = NULL;
282}
283
284static __inline__ void __sk_del_node(struct sock *sk)
285{
286 __hlist_del(&sk->sk_node);
287}
288
289static __inline__ int __sk_del_node_init(struct sock *sk)
290{
291 if (sk_hashed(sk)) {
292 __sk_del_node(sk);
293 sk_node_init(&sk->sk_node);
294 return 1;
295 }
296 return 0;
297}
298
299/* Grab socket reference count. This operation is valid only
300 when sk is ALREADY grabbed f.e. it is found in hash table
301 or a list and the lookup is made under lock preventing hash table
302 modifications.
303 */
304
305static inline void sock_hold(struct sock *sk)
306{
307 atomic_inc(&sk->sk_refcnt);
308}
309
310/* Ungrab socket in the context, which assumes that socket refcnt
311 cannot hit zero, f.e. it is true in context of any socketcall.
312 */
313static inline void __sock_put(struct sock *sk)
314{
315 atomic_dec(&sk->sk_refcnt);
316}
317
318static __inline__ int sk_del_node_init(struct sock *sk)
319{
320 int rc = __sk_del_node_init(sk);
321
322 if (rc) {
323 /* paranoid for a while -acme */
324 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
325 __sock_put(sk);
326 }
327 return rc;
328}
329
330static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
331{
332 hlist_add_head(&sk->sk_node, list);
333}
334
335static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
336{
337 sock_hold(sk);
338 __sk_add_node(sk, list);
339}
340
341static __inline__ void __sk_del_bind_node(struct sock *sk)
342{
343 __hlist_del(&sk->sk_bind_node);
344}
345
346static __inline__ void sk_add_bind_node(struct sock *sk,
347 struct hlist_head *list)
348{
349 hlist_add_head(&sk->sk_bind_node, list);
350}
351
352#define sk_for_each(__sk, node, list) \
353 hlist_for_each_entry(__sk, node, list, sk_node)
354#define sk_for_each_from(__sk, node) \
355 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
356 hlist_for_each_entry_from(__sk, node, sk_node)
357#define sk_for_each_continue(__sk, node) \
358 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
359 hlist_for_each_entry_continue(__sk, node, sk_node)
360#define sk_for_each_safe(__sk, node, tmp, list) \
361 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
362#define sk_for_each_bound(__sk, node, list) \
363 hlist_for_each_entry(__sk, node, list, sk_bind_node)
364
365/* Sock flags */
366enum sock_flags {
367 SOCK_DEAD,
368 SOCK_DONE,
369 SOCK_URGINLINE,
370 SOCK_KEEPOPEN,
371 SOCK_LINGER,
372 SOCK_DESTROY,
373 SOCK_BROADCAST,
374 SOCK_TIMESTAMP,
375 SOCK_ZAPPED,
376 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
377 SOCK_DBG, /* %SO_DEBUG setting */
378 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
379 SOCK_NO_LARGESEND, /* whether to sent large segments or not */
380 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
381 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
382};
383
384static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
385{
386 __set_bit(flag, &sk->sk_flags);
387}
388
389static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
390{
391 __clear_bit(flag, &sk->sk_flags);
392}
393
394static inline int sock_flag(struct sock *sk, enum sock_flags flag)
395{
396 return test_bit(flag, &sk->sk_flags);
397}
398
399static inline void sk_acceptq_removed(struct sock *sk)
400{
401 sk->sk_ack_backlog--;
402}
403
404static inline void sk_acceptq_added(struct sock *sk)
405{
406 sk->sk_ack_backlog++;
407}
408
409static inline int sk_acceptq_is_full(struct sock *sk)
410{
411 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
412}
413
414/*
415 * Compute minimal free write space needed to queue new packets.
416 */
417static inline int sk_stream_min_wspace(struct sock *sk)
418{
419 return sk->sk_wmem_queued / 2;
420}
421
422static inline int sk_stream_wspace(struct sock *sk)
423{
424 return sk->sk_sndbuf - sk->sk_wmem_queued;
425}
426
427extern void sk_stream_write_space(struct sock *sk);
428
429static inline int sk_stream_memory_free(struct sock *sk)
430{
431 return sk->sk_wmem_queued < sk->sk_sndbuf;
432}
433
434extern void sk_stream_rfree(struct sk_buff *skb);
435
436static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk)
437{
438 skb->sk = sk;
439 skb->destructor = sk_stream_rfree;
440 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
441 sk->sk_forward_alloc -= skb->truesize;
442}
443
444static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb)
445{
446 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
447 sk->sk_wmem_queued -= skb->truesize;
448 sk->sk_forward_alloc += skb->truesize;
449 __kfree_skb(skb);
450}
451
452/* The per-socket spinlock must be held here. */
453#define sk_add_backlog(__sk, __skb) \
454do { if (!(__sk)->sk_backlog.tail) { \
455 (__sk)->sk_backlog.head = \
456 (__sk)->sk_backlog.tail = (__skb); \
457 } else { \
458 ((__sk)->sk_backlog.tail)->next = (__skb); \
459 (__sk)->sk_backlog.tail = (__skb); \
460 } \
461 (__skb)->next = NULL; \
462} while(0)
463
464#define sk_wait_event(__sk, __timeo, __condition) \
465({ int rc; \
466 release_sock(__sk); \
467 rc = __condition; \
468 if (!rc) { \
469 *(__timeo) = schedule_timeout(*(__timeo)); \
470 rc = __condition; \
471 } \
472 lock_sock(__sk); \
473 rc; \
474})
475
476extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
477extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
478extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
479extern int sk_stream_error(struct sock *sk, int flags, int err);
480extern void sk_stream_kill_queues(struct sock *sk);
481
482extern int sk_wait_data(struct sock *sk, long *timeo);
483
484/* Networking protocol blocks we attach to sockets.
485 * socket layer -> transport layer interface
486 * transport -> network interface is defined by struct inet_proto
487 */
488struct proto {
489 void (*close)(struct sock *sk,
490 long timeout);
491 int (*connect)(struct sock *sk,
492 struct sockaddr *uaddr,
493 int addr_len);
494 int (*disconnect)(struct sock *sk, int flags);
495
496 struct sock * (*accept) (struct sock *sk, int flags, int *err);
497
498 int (*ioctl)(struct sock *sk, int cmd,
499 unsigned long arg);
500 int (*init)(struct sock *sk);
501 int (*destroy)(struct sock *sk);
502 void (*shutdown)(struct sock *sk, int how);
503 int (*setsockopt)(struct sock *sk, int level,
504 int optname, char __user *optval,
505 int optlen);
506 int (*getsockopt)(struct sock *sk, int level,
507 int optname, char __user *optval,
508 int __user *option);
509 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
510 struct msghdr *msg, size_t len);
511 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
512 struct msghdr *msg,
513 size_t len, int noblock, int flags,
514 int *addr_len);
515 int (*sendpage)(struct sock *sk, struct page *page,
516 int offset, size_t size, int flags);
517 int (*bind)(struct sock *sk,
518 struct sockaddr *uaddr, int addr_len);
519
520 int (*backlog_rcv) (struct sock *sk,
521 struct sk_buff *skb);
522
523 /* Keeping track of sk's, looking them up, and port selection methods. */
524 void (*hash)(struct sock *sk);
525 void (*unhash)(struct sock *sk);
526 int (*get_port)(struct sock *sk, unsigned short snum);
527
528 /* Memory pressure */
529 void (*enter_memory_pressure)(void);
530 atomic_t *memory_allocated; /* Current allocated memory. */
531 atomic_t *sockets_allocated; /* Current number of sockets. */
532 /*
533 * Pressure flag: try to collapse.
534 * Technical note: it is used by multiple contexts non atomically.
535 * All the sk_stream_mem_schedule() is of this nature: accounting
536 * is strict, actions are advisory and have some latency.
537 */
538 int *memory_pressure;
539 int *sysctl_mem;
540 int *sysctl_wmem;
541 int *sysctl_rmem;
542 int max_header;
543
544 kmem_cache_t *slab;
545 unsigned int obj_size;
546
547 struct module *owner;
548
549 char name[32];
550
551 struct list_head node;
552
553 struct {
554 int inuse;
555 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
556 } stats[NR_CPUS];
557};
558
559extern int proto_register(struct proto *prot, int alloc_slab);
560extern void proto_unregister(struct proto *prot);
561
562/* Called with local bh disabled */
563static __inline__ void sock_prot_inc_use(struct proto *prot)
564{
565 prot->stats[smp_processor_id()].inuse++;
566}
567
568static __inline__ void sock_prot_dec_use(struct proto *prot)
569{
570 prot->stats[smp_processor_id()].inuse--;
571}
572
573/* About 10 seconds */
574#define SOCK_DESTROY_TIME (10*HZ)
575
576/* Sockets 0-1023 can't be bound to unless you are superuser */
577#define PROT_SOCK 1024
578
579#define SHUTDOWN_MASK 3
580#define RCV_SHUTDOWN 1
581#define SEND_SHUTDOWN 2
582
583#define SOCK_SNDBUF_LOCK 1
584#define SOCK_RCVBUF_LOCK 2
585#define SOCK_BINDADDR_LOCK 4
586#define SOCK_BINDPORT_LOCK 8
587
588/* sock_iocb: used to kick off async processing of socket ios */
589struct sock_iocb {
590 struct list_head list;
591
592 int flags;
593 int size;
594 struct socket *sock;
595 struct sock *sk;
596 struct scm_cookie *scm;
597 struct msghdr *msg, async_msg;
598 struct iovec async_iov;
599 struct kiocb *kiocb;
600};
601
602static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
603{
604 return (struct sock_iocb *)iocb->private;
605}
606
607static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
608{
609 return si->kiocb;
610}
611
612struct socket_alloc {
613 struct socket socket;
614 struct inode vfs_inode;
615};
616
617static inline struct socket *SOCKET_I(struct inode *inode)
618{
619 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
620}
621
622static inline struct inode *SOCK_INODE(struct socket *socket)
623{
624 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
625}
626
627extern void __sk_stream_mem_reclaim(struct sock *sk);
628extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
629
630#define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
631
632static inline int sk_stream_pages(int amt)
633{
634 return (amt + SK_STREAM_MEM_QUANTUM - 1) / SK_STREAM_MEM_QUANTUM;
635}
636
637static inline void sk_stream_mem_reclaim(struct sock *sk)
638{
639 if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM)
640 __sk_stream_mem_reclaim(sk);
641}
642
643static inline void sk_stream_writequeue_purge(struct sock *sk)
644{
645 struct sk_buff *skb;
646
647 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
648 sk_stream_free_skb(sk, skb);
649 sk_stream_mem_reclaim(sk);
650}
651
652static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb)
653{
654 return (int)skb->truesize <= sk->sk_forward_alloc ||
655 sk_stream_mem_schedule(sk, skb->truesize, 1);
656}
657
658/* Used by processes to "lock" a socket state, so that
659 * interrupts and bottom half handlers won't change it
660 * from under us. It essentially blocks any incoming
661 * packets, so that we won't get any new data or any
662 * packets that change the state of the socket.
663 *
664 * While locked, BH processing will add new packets to
665 * the backlog queue. This queue is processed by the
666 * owner of the socket lock right before it is released.
667 *
668 * Since ~2.3.5 it is also exclusive sleep lock serializing
669 * accesses from user process context.
670 */
671#define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
672
673extern void FASTCALL(lock_sock(struct sock *sk));
674extern void FASTCALL(release_sock(struct sock *sk));
675
676/* BH context may only use the following locking interface. */
677#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
678#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
679
680extern struct sock *sk_alloc(int family, int priority,
681 struct proto *prot, int zero_it);
682extern void sk_free(struct sock *sk);
683
684extern struct sk_buff *sock_wmalloc(struct sock *sk,
685 unsigned long size, int force,
686 int priority);
687extern struct sk_buff *sock_rmalloc(struct sock *sk,
688 unsigned long size, int force,
689 int priority);
690extern void sock_wfree(struct sk_buff *skb);
691extern void sock_rfree(struct sk_buff *skb);
692
693extern int sock_setsockopt(struct socket *sock, int level,
694 int op, char __user *optval,
695 int optlen);
696
697extern int sock_getsockopt(struct socket *sock, int level,
698 int op, char __user *optval,
699 int __user *optlen);
700extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
701 unsigned long size,
702 int noblock,
703 int *errcode);
704extern void *sock_kmalloc(struct sock *sk, int size, int priority);
705extern void sock_kfree_s(struct sock *sk, void *mem, int size);
706extern void sk_send_sigurg(struct sock *sk);
707
708/*
709 * Functions to fill in entries in struct proto_ops when a protocol
710 * does not implement a particular function.
711 */
712extern int sock_no_bind(struct socket *,
713 struct sockaddr *, int);
714extern int sock_no_connect(struct socket *,
715 struct sockaddr *, int, int);
716extern int sock_no_socketpair(struct socket *,
717 struct socket *);
718extern int sock_no_accept(struct socket *,
719 struct socket *, int);
720extern int sock_no_getname(struct socket *,
721 struct sockaddr *, int *, int);
722extern unsigned int sock_no_poll(struct file *, struct socket *,
723 struct poll_table_struct *);
724extern int sock_no_ioctl(struct socket *, unsigned int,
725 unsigned long);
726extern int sock_no_listen(struct socket *, int);
727extern int sock_no_shutdown(struct socket *, int);
728extern int sock_no_getsockopt(struct socket *, int , int,
729 char __user *, int __user *);
730extern int sock_no_setsockopt(struct socket *, int, int,
731 char __user *, int);
732extern int sock_no_sendmsg(struct kiocb *, struct socket *,
733 struct msghdr *, size_t);
734extern int sock_no_recvmsg(struct kiocb *, struct socket *,
735 struct msghdr *, size_t, int);
736extern int sock_no_mmap(struct file *file,
737 struct socket *sock,
738 struct vm_area_struct *vma);
739extern ssize_t sock_no_sendpage(struct socket *sock,
740 struct page *page,
741 int offset, size_t size,
742 int flags);
743
744/*
745 * Functions to fill in entries in struct proto_ops when a protocol
746 * uses the inet style.
747 */
748extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
749 char __user *optval, int __user *optlen);
750extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
751 struct msghdr *msg, size_t size, int flags);
752extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
753 char __user *optval, int optlen);
754
755extern void sk_common_release(struct sock *sk);
756
757/*
758 * Default socket callbacks and setup code
759 */
760
761/* Initialise core socket variables */
762extern void sock_init_data(struct socket *sock, struct sock *sk);
763
764/**
765 * sk_filter - run a packet through a socket filter
766 * @sk: sock associated with &sk_buff
767 * @skb: buffer to filter
768 * @needlock: set to 1 if the sock is not locked by caller.
769 *
770 * Run the filter code and then cut skb->data to correct size returned by
771 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
772 * than pkt_len we keep whole skb->data. This is the socket level
773 * wrapper to sk_run_filter. It returns 0 if the packet should
774 * be accepted or -EPERM if the packet should be tossed.
775 *
776 */
777
778static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock)
779{
780 int err;
781
782 err = security_sock_rcv_skb(sk, skb);
783 if (err)
784 return err;
785
786 if (sk->sk_filter) {
787 struct sk_filter *filter;
788
789 if (needlock)
790 bh_lock_sock(sk);
791
792 filter = sk->sk_filter;
793 if (filter) {
794 int pkt_len = sk_run_filter(skb, filter->insns,
795 filter->len);
796 if (!pkt_len)
797 err = -EPERM;
798 else
799 skb_trim(skb, pkt_len);
800 }
801
802 if (needlock)
803 bh_unlock_sock(sk);
804 }
805 return err;
806}
807
808/**
809 * sk_filter_release: Release a socket filter
810 * @sk: socket
811 * @fp: filter to remove
812 *
813 * Remove a filter from a socket and release its resources.
814 */
815
816static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
817{
818 unsigned int size = sk_filter_len(fp);
819
820 atomic_sub(size, &sk->sk_omem_alloc);
821
822 if (atomic_dec_and_test(&fp->refcnt))
823 kfree(fp);
824}
825
826static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
827{
828 atomic_inc(&fp->refcnt);
829 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
830}
831
832/*
833 * Socket reference counting postulates.
834 *
835 * * Each user of socket SHOULD hold a reference count.
836 * * Each access point to socket (an hash table bucket, reference from a list,
837 * running timer, skb in flight MUST hold a reference count.
838 * * When reference count hits 0, it means it will never increase back.
839 * * When reference count hits 0, it means that no references from
840 * outside exist to this socket and current process on current CPU
841 * is last user and may/should destroy this socket.
842 * * sk_free is called from any context: process, BH, IRQ. When
843 * it is called, socket has no references from outside -> sk_free
844 * may release descendant resources allocated by the socket, but
845 * to the time when it is called, socket is NOT referenced by any
846 * hash tables, lists etc.
847 * * Packets, delivered from outside (from network or from another process)
848 * and enqueued on receive/error queues SHOULD NOT grab reference count,
849 * when they sit in queue. Otherwise, packets will leak to hole, when
850 * socket is looked up by one cpu and unhasing is made by another CPU.
851 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
852 * (leak to backlog). Packet socket does all the processing inside
853 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
854 * use separate SMP lock, so that they are prone too.
855 */
856
857/* Ungrab socket and destroy it, if it was the last reference. */
858static inline void sock_put(struct sock *sk)
859{
860 if (atomic_dec_and_test(&sk->sk_refcnt))
861 sk_free(sk);
862}
863
864/* Detach socket from process context.
865 * Announce socket dead, detach it from wait queue and inode.
866 * Note that parent inode held reference count on this struct sock,
867 * we do not release it in this function, because protocol
868 * probably wants some additional cleanups or even continuing
869 * to work with this socket (TCP).
870 */
871static inline void sock_orphan(struct sock *sk)
872{
873 write_lock_bh(&sk->sk_callback_lock);
874 sock_set_flag(sk, SOCK_DEAD);
875 sk->sk_socket = NULL;
876 sk->sk_sleep = NULL;
877 write_unlock_bh(&sk->sk_callback_lock);
878}
879
880static inline void sock_graft(struct sock *sk, struct socket *parent)
881{
882 write_lock_bh(&sk->sk_callback_lock);
883 sk->sk_sleep = &parent->wait;
884 parent->sk = sk;
885 sk->sk_socket = parent;
886 write_unlock_bh(&sk->sk_callback_lock);
887}
888
889extern int sock_i_uid(struct sock *sk);
890extern unsigned long sock_i_ino(struct sock *sk);
891
892static inline struct dst_entry *
893__sk_dst_get(struct sock *sk)
894{
895 return sk->sk_dst_cache;
896}
897
898static inline struct dst_entry *
899sk_dst_get(struct sock *sk)
900{
901 struct dst_entry *dst;
902
903 read_lock(&sk->sk_dst_lock);
904 dst = sk->sk_dst_cache;
905 if (dst)
906 dst_hold(dst);
907 read_unlock(&sk->sk_dst_lock);
908 return dst;
909}
910
911static inline void
912__sk_dst_set(struct sock *sk, struct dst_entry *dst)
913{
914 struct dst_entry *old_dst;
915
916 old_dst = sk->sk_dst_cache;
917 sk->sk_dst_cache = dst;
918 dst_release(old_dst);
919}
920
921static inline void
922sk_dst_set(struct sock *sk, struct dst_entry *dst)
923{
924 write_lock(&sk->sk_dst_lock);
925 __sk_dst_set(sk, dst);
926 write_unlock(&sk->sk_dst_lock);
927}
928
929static inline void
930__sk_dst_reset(struct sock *sk)
931{
932 struct dst_entry *old_dst;
933
934 old_dst = sk->sk_dst_cache;
935 sk->sk_dst_cache = NULL;
936 dst_release(old_dst);
937}
938
939static inline void
940sk_dst_reset(struct sock *sk)
941{
942 write_lock(&sk->sk_dst_lock);
943 __sk_dst_reset(sk);
944 write_unlock(&sk->sk_dst_lock);
945}
946
947static inline struct dst_entry *
948__sk_dst_check(struct sock *sk, u32 cookie)
949{
950 struct dst_entry *dst = sk->sk_dst_cache;
951
952 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
953 sk->sk_dst_cache = NULL;
954 dst_release(dst);
955 return NULL;
956 }
957
958 return dst;
959}
960
961static inline struct dst_entry *
962sk_dst_check(struct sock *sk, u32 cookie)
963{
964 struct dst_entry *dst = sk_dst_get(sk);
965
966 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
967 sk_dst_reset(sk);
968 dst_release(dst);
969 return NULL;
970 }
971
972 return dst;
973}
974
975static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
976{
977 sk->sk_wmem_queued += skb->truesize;
978 sk->sk_forward_alloc -= skb->truesize;
979}
980
981static inline int skb_copy_to_page(struct sock *sk, char __user *from,
982 struct sk_buff *skb, struct page *page,
983 int off, int copy)
984{
985 if (skb->ip_summed == CHECKSUM_NONE) {
986 int err = 0;
987 unsigned int csum = csum_and_copy_from_user(from,
988 page_address(page) + off,
989 copy, 0, &err);
990 if (err)
991 return err;
992 skb->csum = csum_block_add(skb->csum, csum, skb->len);
993 } else if (copy_from_user(page_address(page) + off, from, copy))
994 return -EFAULT;
995
996 skb->len += copy;
997 skb->data_len += copy;
998 skb->truesize += copy;
999 sk->sk_wmem_queued += copy;
1000 sk->sk_forward_alloc -= copy;
1001 return 0;
1002}
1003
1004/*
1005 * Queue a received datagram if it will fit. Stream and sequenced
1006 * protocols can't normally use this as they need to fit buffers in
1007 * and play with them.
1008 *
1009 * Inlined as it's very short and called for pretty much every
1010 * packet ever received.
1011 */
1012
1013static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1014{
1015 sock_hold(sk);
1016 skb->sk = sk;
1017 skb->destructor = sock_wfree;
1018 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1019}
1020
1021static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1022{
1023 skb->sk = sk;
1024 skb->destructor = sock_rfree;
1025 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1026}
1027
1028extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1029 unsigned long expires);
1030
1031extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1032
1033static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1034{
1035 int err = 0;
1036 int skb_len;
1037
1038 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1039 number of warnings when compiling with -W --ANK
1040 */
1041 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1042 (unsigned)sk->sk_rcvbuf) {
1043 err = -ENOMEM;
1044 goto out;
1045 }
1046
1047 /* It would be deadlock, if sock_queue_rcv_skb is used
1048 with socket lock! We assume that users of this
1049 function are lock free.
1050 */
1051 err = sk_filter(sk, skb, 1);
1052 if (err)
1053 goto out;
1054
1055 skb->dev = NULL;
1056 skb_set_owner_r(skb, sk);
1057
1058 /* Cache the SKB length before we tack it onto the receive
1059 * queue. Once it is added it no longer belongs to us and
1060 * may be freed by other threads of control pulling packets
1061 * from the queue.
1062 */
1063 skb_len = skb->len;
1064
1065 skb_queue_tail(&sk->sk_receive_queue, skb);
1066
1067 if (!sock_flag(sk, SOCK_DEAD))
1068 sk->sk_data_ready(sk, skb_len);
1069out:
1070 return err;
1071}
1072
1073static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1074{
1075 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1076 number of warnings when compiling with -W --ANK
1077 */
1078 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1079 (unsigned)sk->sk_rcvbuf)
1080 return -ENOMEM;
1081 skb_set_owner_r(skb, sk);
1082 skb_queue_tail(&sk->sk_error_queue, skb);
1083 if (!sock_flag(sk, SOCK_DEAD))
1084 sk->sk_data_ready(sk, skb->len);
1085 return 0;
1086}
1087
1088/*
1089 * Recover an error report and clear atomically
1090 */
1091
1092static inline int sock_error(struct sock *sk)
1093{
1094 int err = xchg(&sk->sk_err, 0);
1095 return -err;
1096}
1097
1098static inline unsigned long sock_wspace(struct sock *sk)
1099{
1100 int amt = 0;
1101
1102 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1103 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1104 if (amt < 0)
1105 amt = 0;
1106 }
1107 return amt;
1108}
1109
1110static inline void sk_wake_async(struct sock *sk, int how, int band)
1111{
1112 if (sk->sk_socket && sk->sk_socket->fasync_list)
1113 sock_wake_async(sk->sk_socket, how, band);
1114}
1115
1116#define SOCK_MIN_SNDBUF 2048
1117#define SOCK_MIN_RCVBUF 256
1118
1119static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1120{
1121 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1122 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2);
1123 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1124 }
1125}
1126
1127static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk,
1128 int size, int mem, int gfp)
1129{
1130 struct sk_buff *skb = alloc_skb(size + sk->sk_prot->max_header, gfp);
1131
1132 if (skb) {
1133 skb->truesize += mem;
1134 if (sk->sk_forward_alloc >= (int)skb->truesize ||
1135 sk_stream_mem_schedule(sk, skb->truesize, 0)) {
1136 skb_reserve(skb, sk->sk_prot->max_header);
1137 return skb;
1138 }
1139 __kfree_skb(skb);
1140 } else {
1141 sk->sk_prot->enter_memory_pressure();
1142 sk_stream_moderate_sndbuf(sk);
1143 }
1144 return NULL;
1145}
1146
1147static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk,
1148 int size, int gfp)
1149{
1150 return sk_stream_alloc_pskb(sk, size, 0, gfp);
1151}
1152
1153static inline struct page *sk_stream_alloc_page(struct sock *sk)
1154{
1155 struct page *page = NULL;
1156
1157 if (sk->sk_forward_alloc >= (int)PAGE_SIZE ||
1158 sk_stream_mem_schedule(sk, PAGE_SIZE, 0))
1159 page = alloc_pages(sk->sk_allocation, 0);
1160 else {
1161 sk->sk_prot->enter_memory_pressure();
1162 sk_stream_moderate_sndbuf(sk);
1163 }
1164 return page;
1165}
1166
1167#define sk_stream_for_retrans_queue(skb, sk) \
1168 for (skb = (sk)->sk_write_queue.next; \
1169 (skb != (sk)->sk_send_head) && \
1170 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1171 skb = skb->next)
1172
1173/*
1174 * Default write policy as shown to user space via poll/select/SIGIO
1175 */
1176static inline int sock_writeable(const struct sock *sk)
1177{
1178 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1179}
1180
1181static inline int gfp_any(void)
1182{
1183 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1184}
1185
1186static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1187{
1188 return noblock ? 0 : sk->sk_rcvtimeo;
1189}
1190
1191static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1192{
1193 return noblock ? 0 : sk->sk_sndtimeo;
1194}
1195
1196static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1197{
1198 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1199}
1200
1201/* Alas, with timeout socket operations are not restartable.
1202 * Compare this to poll().
1203 */
1204static inline int sock_intr_errno(long timeo)
1205{
1206 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1207}
1208
1209static __inline__ void
1210sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1211{
1212 struct timeval *stamp = &skb->stamp;
1213 if (sock_flag(sk, SOCK_RCVTSTAMP)) {
1214 /* Race occurred between timestamp enabling and packet
1215 receiving. Fill in the current time for now. */
1216 if (stamp->tv_sec == 0)
1217 do_gettimeofday(stamp);
1218 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval),
1219 stamp);
1220 } else
1221 sk->sk_stamp = *stamp;
1222}
1223
1224/**
1225 * sk_eat_skb - Release a skb if it is no longer needed
1226 * @sk - socket to eat this skb from
1227 * @skb - socket buffer to eat
1228 *
1229 * This routine must be called with interrupts disabled or with the socket
1230 * locked so that the sk_buff queue operation is ok.
1231*/
1232static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1233{
1234 __skb_unlink(skb, &sk->sk_receive_queue);
1235 __kfree_skb(skb);
1236}
1237
1238extern void sock_enable_timestamp(struct sock *sk);
1239extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1240
1241/*
1242 * Enable debug/info messages
1243 */
1244
1245#if 0
1246#define NETDEBUG(x) do { } while (0)
1247#define LIMIT_NETDEBUG(x) do {} while(0)
1248#else
1249#define NETDEBUG(x) do { x; } while (0)
1250#define LIMIT_NETDEBUG(x) do { if (net_ratelimit()) { x; } } while(0)
1251#endif
1252
1253/*
1254 * Macros for sleeping on a socket. Use them like this:
1255 *
1256 * SOCK_SLEEP_PRE(sk)
1257 * if (condition)
1258 * schedule();
1259 * SOCK_SLEEP_POST(sk)
1260 *
1261 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1262 * and when the last use of them in DECnet has gone, I'm intending to
1263 * remove them.
1264 */
1265
1266#define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1267 DECLARE_WAITQUEUE(wait, tsk); \
1268 tsk->state = TASK_INTERRUPTIBLE; \
1269 add_wait_queue((sk)->sk_sleep, &wait); \
1270 release_sock(sk);
1271
1272#define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1273 remove_wait_queue((sk)->sk_sleep, &wait); \
1274 lock_sock(sk); \
1275 }
1276
1277static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1278{
1279 if (valbool)
1280 sock_set_flag(sk, bit);
1281 else
1282 sock_reset_flag(sk, bit);
1283}
1284
1285extern __u32 sysctl_wmem_max;
1286extern __u32 sysctl_rmem_max;
1287
1288#ifdef CONFIG_NET
1289int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
1290#else
1291static inline int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg)
1292{
1293 return -ENODEV;
1294}
1295#endif
1296
1297#endif /* _SOCK_H */