1/*	$NetBSD: socket.h,v 1.3.4.1 2012/06/05 21:15:29 bouyer Exp $	*/
2
3/*
4 * Copyright (C) 2004-2009, 2011, 2012  Internet Systems Consortium, Inc. ("ISC")
5 * Copyright (C) 1998-2002  Internet Software Consortium.
6 *
7 * Permission to use, copy, modify, and/or distribute this software for any
8 * purpose with or without fee is hereby granted, provided that the above
9 * copyright notice and this permission notice appear in all copies.
10 *
11 * THE SOFTWARE IS PROVIDED "AS IS" AND ISC DISCLAIMS ALL WARRANTIES WITH
12 * REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY
13 * AND FITNESS.  IN NO EVENT SHALL ISC BE LIABLE FOR ANY SPECIAL, DIRECT,
14 * INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM
15 * LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
16 * OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
17 * PERFORMANCE OF THIS SOFTWARE.
18 */
19
20/* Id */
21
22#ifndef ISC_SOCKET_H
23#define ISC_SOCKET_H 1
24
25/*****
26 ***** Module Info
27 *****/
28
29/*! \file isc/socket.h
30 * \brief Provides TCP and UDP sockets for network I/O.  The sockets are event
31 * sources in the task system.
32 *
33 * When I/O completes, a completion event for the socket is posted to the
34 * event queue of the task which requested the I/O.
35 *
36 * \li MP:
37 *	The module ensures appropriate synchronization of data structures it
38 *	creates and manipulates.
39 *	Clients of this module must not be holding a socket's task's lock when
40 *	making a call that affects that socket.  Failure to follow this rule
41 *	can result in deadlock.
42 *	The caller must ensure that isc_socketmgr_destroy() is called only
43 *	once for a given manager.
44 *
45 * \li Reliability:
46 *	No anticipated impact.
47 *
48 * \li Resources:
49 *	TBS
50 *
51 * \li Security:
52 *	No anticipated impact.
53 *
54 * \li Standards:
55 *	None.
56 */
57
58/***
59 *** Imports
60 ***/
61
62#include <isc/lang.h>
63#include <isc/types.h>
64#include <isc/event.h>
65#include <isc/eventclass.h>
66#include <isc/time.h>
67#include <isc/region.h>
68#include <isc/sockaddr.h>
69#include <isc/xml.h>
70
71ISC_LANG_BEGINDECLS
72
73/***
74 *** Constants
75 ***/
76
77/*%
78 * Maximum number of buffers in a scatter/gather read/write.  The operating
79 * system in use must support at least this number (plus one on some.)
80 */
81#define ISC_SOCKET_MAXSCATTERGATHER	8
82
83/*%
84 * In isc_socket_bind() set socket option SO_REUSEADDR prior to calling
85 * bind() if a non zero port is specified (AF_INET and AF_INET6).
86 */
87#define ISC_SOCKET_REUSEADDRESS		0x01U
88
89/*%
90 * Statistics counters.  Used as isc_statscounter_t values.
91 */
92enum {
93	isc_sockstatscounter_udp4open = 0,
94	isc_sockstatscounter_udp6open = 1,
95	isc_sockstatscounter_tcp4open = 2,
96	isc_sockstatscounter_tcp6open = 3,
97	isc_sockstatscounter_unixopen = 4,
98
99	isc_sockstatscounter_udp4openfail = 5,
100	isc_sockstatscounter_udp6openfail = 6,
101	isc_sockstatscounter_tcp4openfail = 7,
102	isc_sockstatscounter_tcp6openfail = 8,
103	isc_sockstatscounter_unixopenfail = 9,
104
105	isc_sockstatscounter_udp4close = 10,
106	isc_sockstatscounter_udp6close = 11,
107	isc_sockstatscounter_tcp4close = 12,
108	isc_sockstatscounter_tcp6close = 13,
109	isc_sockstatscounter_unixclose = 14,
110	isc_sockstatscounter_fdwatchclose = 15,
111
112	isc_sockstatscounter_udp4bindfail = 16,
113	isc_sockstatscounter_udp6bindfail = 17,
114	isc_sockstatscounter_tcp4bindfail = 18,
115	isc_sockstatscounter_tcp6bindfail = 19,
116	isc_sockstatscounter_unixbindfail = 20,
117	isc_sockstatscounter_fdwatchbindfail = 21,
118
119	isc_sockstatscounter_udp4connect = 22,
120	isc_sockstatscounter_udp6connect = 23,
121	isc_sockstatscounter_tcp4connect = 24,
122	isc_sockstatscounter_tcp6connect = 25,
123	isc_sockstatscounter_unixconnect = 26,
124	isc_sockstatscounter_fdwatchconnect = 27,
125
126	isc_sockstatscounter_udp4connectfail = 28,
127	isc_sockstatscounter_udp6connectfail = 29,
128	isc_sockstatscounter_tcp4connectfail = 30,
129	isc_sockstatscounter_tcp6connectfail = 31,
130	isc_sockstatscounter_unixconnectfail = 32,
131	isc_sockstatscounter_fdwatchconnectfail = 33,
132
133	isc_sockstatscounter_tcp4accept = 34,
134	isc_sockstatscounter_tcp6accept = 35,
135	isc_sockstatscounter_unixaccept = 36,
136
137	isc_sockstatscounter_tcp4acceptfail = 37,
138	isc_sockstatscounter_tcp6acceptfail = 38,
139	isc_sockstatscounter_unixacceptfail = 39,
140
141	isc_sockstatscounter_udp4sendfail = 40,
142	isc_sockstatscounter_udp6sendfail = 41,
143	isc_sockstatscounter_tcp4sendfail = 42,
144	isc_sockstatscounter_tcp6sendfail = 43,
145	isc_sockstatscounter_unixsendfail = 44,
146	isc_sockstatscounter_fdwatchsendfail = 45,
147
148	isc_sockstatscounter_udp4recvfail = 46,
149	isc_sockstatscounter_udp6recvfail = 47,
150	isc_sockstatscounter_tcp4recvfail = 48,
151	isc_sockstatscounter_tcp6recvfail = 49,
152	isc_sockstatscounter_unixrecvfail = 50,
153	isc_sockstatscounter_fdwatchrecvfail = 51,
154
155	isc_sockstatscounter_max = 52
156};
157
158/***
159 *** Types
160 ***/
161
162struct isc_socketevent {
163	ISC_EVENT_COMMON(isc_socketevent_t);
164	isc_result_t		result;		/*%< OK, EOF, whatever else */
165	unsigned int		minimum;	/*%< minimum i/o for event */
166	unsigned int		n;		/*%< bytes read or written */
167	unsigned int		offset;		/*%< offset into buffer list */
168	isc_region_t		region;		/*%< for single-buffer i/o */
169	isc_bufferlist_t	bufferlist;	/*%< list of buffers */
170	isc_sockaddr_t		address;	/*%< source address */
171	isc_time_t		timestamp;	/*%< timestamp of packet recv */
172	struct in6_pktinfo	pktinfo;	/*%< ipv6 pktinfo */
173	isc_uint32_t		attributes;	/*%< see below */
174	isc_eventdestructor_t   destroy;	/*%< original destructor */
175};
176
177typedef struct isc_socket_newconnev isc_socket_newconnev_t;
178struct isc_socket_newconnev {
179	ISC_EVENT_COMMON(isc_socket_newconnev_t);
180	isc_socket_t *		newsocket;
181	isc_result_t		result;		/*%< OK, EOF, whatever else */
182	isc_sockaddr_t		address;	/*%< source address */
183};
184
185typedef struct isc_socket_connev isc_socket_connev_t;
186struct isc_socket_connev {
187	ISC_EVENT_COMMON(isc_socket_connev_t);
188	isc_result_t		result;		/*%< OK, EOF, whatever else */
189};
190
191/*@{*/
192/*!
193 * _ATTACHED:	Internal use only.
194 * _TRUNC:	Packet was truncated on receive.
195 * _CTRUNC:	Packet control information was truncated.  This can
196 *		indicate that the packet is not complete, even though
197 *		all the data is valid.
198 * _TIMESTAMP:	The timestamp member is valid.
199 * _PKTINFO:	The pktinfo member is valid.
200 * _MULTICAST:	The UDP packet was received via a multicast transmission.
201 */
202#define ISC_SOCKEVENTATTR_ATTACHED		0x80000000U /* internal */
203#define ISC_SOCKEVENTATTR_TRUNC			0x00800000U /* public */
204#define ISC_SOCKEVENTATTR_CTRUNC		0x00400000U /* public */
205#define ISC_SOCKEVENTATTR_TIMESTAMP		0x00200000U /* public */
206#define ISC_SOCKEVENTATTR_PKTINFO		0x00100000U /* public */
207#define ISC_SOCKEVENTATTR_MULTICAST		0x00080000U /* public */
208/*@}*/
209
210#define ISC_SOCKEVENT_ANYEVENT  (0)
211#define ISC_SOCKEVENT_RECVDONE	(ISC_EVENTCLASS_SOCKET + 1)
212#define ISC_SOCKEVENT_SENDDONE	(ISC_EVENTCLASS_SOCKET + 2)
213#define ISC_SOCKEVENT_NEWCONN	(ISC_EVENTCLASS_SOCKET + 3)
214#define ISC_SOCKEVENT_CONNECT	(ISC_EVENTCLASS_SOCKET + 4)
215
216/*
217 * Internal events.
218 */
219#define ISC_SOCKEVENT_INTR	(ISC_EVENTCLASS_SOCKET + 256)
220#define ISC_SOCKEVENT_INTW	(ISC_EVENTCLASS_SOCKET + 257)
221
222typedef enum {
223	isc_sockettype_udp = 1,
224	isc_sockettype_tcp = 2,
225	isc_sockettype_unix = 3,
226	isc_sockettype_fdwatch = 4
227} isc_sockettype_t;
228
229/*@{*/
230/*!
231 * How a socket should be shutdown in isc_socket_shutdown() calls.
232 */
233#define ISC_SOCKSHUT_RECV	0x00000001	/*%< close read side */
234#define ISC_SOCKSHUT_SEND	0x00000002	/*%< close write side */
235#define ISC_SOCKSHUT_ALL	0x00000003	/*%< close them all */
236/*@}*/
237
238/*@{*/
239/*!
240 * What I/O events to cancel in isc_socket_cancel() calls.
241 */
242#define ISC_SOCKCANCEL_RECV	0x00000001	/*%< cancel recv */
243#define ISC_SOCKCANCEL_SEND	0x00000002	/*%< cancel send */
244#define ISC_SOCKCANCEL_ACCEPT	0x00000004	/*%< cancel accept */
245#define ISC_SOCKCANCEL_CONNECT	0x00000008	/*%< cancel connect */
246#define ISC_SOCKCANCEL_ALL	0x0000000f	/*%< cancel everything */
247/*@}*/
248
249/*@{*/
250/*!
251 * Flags for isc_socket_send() and isc_socket_recv() calls.
252 */
253#define ISC_SOCKFLAG_IMMEDIATE	0x00000001	/*%< send event only if needed */
254#define ISC_SOCKFLAG_NORETRY	0x00000002	/*%< drop failed UDP sends */
255/*@}*/
256
257/*@{*/
258/*!
259 * Flags for fdwatchcreate.
260 */
261#define ISC_SOCKFDWATCH_READ	0x00000001	/*%< watch for readable */
262#define ISC_SOCKFDWATCH_WRITE	0x00000002	/*%< watch for writable */
263/*@}*/
264
265/*% Socket and socket manager methods */
266typedef struct isc_socketmgrmethods {
267	void		(*destroy)(isc_socketmgr_t **managerp);
268	isc_result_t	(*socketcreate)(isc_socketmgr_t *manager, int pf,
269					isc_sockettype_t type,
270					isc_socket_t **socketp);
271	isc_result_t    (*fdwatchcreate)(isc_socketmgr_t *manager, int fd,
272					 int flags,
273					 isc_sockfdwatch_t callback,
274					 void *cbarg, isc_task_t *task,
275					 isc_socket_t **socketp);
276} isc_socketmgrmethods_t;
277
278typedef struct isc_socketmethods {
279	void		(*attach)(isc_socket_t *socket,
280				  isc_socket_t **socketp);
281	void		(*detach)(isc_socket_t **socketp);
282	isc_result_t	(*bind)(isc_socket_t *sock, isc_sockaddr_t *sockaddr,
283				unsigned int options);
284	isc_result_t	(*sendto)(isc_socket_t *sock, isc_region_t *region,
285				  isc_task_t *task, isc_taskaction_t action,
286				  const void *arg, isc_sockaddr_t *address,
287				  struct in6_pktinfo *pktinfo);
288	isc_result_t	(*connect)(isc_socket_t *sock, isc_sockaddr_t *addr,
289				   isc_task_t *task, isc_taskaction_t action,
290				   const void *arg);
291	isc_result_t	(*recv)(isc_socket_t *sock, isc_region_t *region,
292				unsigned int minimum, isc_task_t *task,
293				isc_taskaction_t action, const void *arg);
294	void		(*cancel)(isc_socket_t *sock, isc_task_t *task,
295				  unsigned int how);
296	isc_result_t	(*getsockname)(isc_socket_t *sock,
297				       isc_sockaddr_t *addressp);
298	isc_sockettype_t (*gettype)(isc_socket_t *sock);
299	void		(*ipv6only)(isc_socket_t *sock, isc_boolean_t yes);
300	isc_result_t    (*fdwatchpoke)(isc_socket_t *sock, int flags);
301	isc_result_t		(*dup)(isc_socket_t *socket,
302				  isc_socket_t **socketp);
303	int 		(*getfd)(isc_socket_t *socket);
304} isc_socketmethods_t;
305
306/*%
307 * This structure is actually just the common prefix of a socket manager
308 * object implementation's version of an isc_socketmgr_t.
309 * \brief
310 * Direct use of this structure by clients is forbidden.  socket implementations
311 * may change the structure.  'magic' must be ISCAPI_SOCKETMGR_MAGIC for any
312 * of the isc_socket_ routines to work.  socket implementations must maintain
313 * all socket invariants.
314 * In effect, this definition is used only for non-BIND9 version ("export")
315 * of the library, and the export version does not work for win32.  So, to avoid
316 * the definition conflict with win32/socket.c, we enable this definition only
317 * for non-Win32 (i.e. Unix) platforms.
318 */
319#ifndef WIN32
320struct isc_socketmgr {
321	unsigned int		impmagic;
322	unsigned int		magic;
323	isc_socketmgrmethods_t	*methods;
324};
325#endif
326
327#define ISCAPI_SOCKETMGR_MAGIC		ISC_MAGIC('A','s','m','g')
328#define ISCAPI_SOCKETMGR_VALID(m)	((m) != NULL && \
329					 (m)->magic == ISCAPI_SOCKETMGR_MAGIC)
330
331/*%
332 * This is the common prefix of a socket object.  The same note as
333 * that for the socketmgr structure applies.
334 */
335#ifndef WIN32
336struct isc_socket {
337	unsigned int		impmagic;
338	unsigned int		magic;
339	isc_socketmethods_t	*methods;
340};
341#endif
342
343#define ISCAPI_SOCKET_MAGIC	ISC_MAGIC('A','s','c','t')
344#define ISCAPI_SOCKET_VALID(s)	((s) != NULL && \
345				 (s)->magic == ISCAPI_SOCKET_MAGIC)
346
347/***
348 *** Socket and Socket Manager Functions
349 ***
350 *** Note: all Ensures conditions apply only if the result is success for
351 *** those functions which return an isc_result.
352 ***/
353
354isc_result_t
355isc_socket_fdwatchcreate(isc_socketmgr_t *manager,
356			 int fd,
357			 int flags,
358			 isc_sockfdwatch_t callback,
359			 void *cbarg,
360			 isc_task_t *task,
361			 isc_socket_t **socketp);
362/*%<
363 * Create a new file descriptor watch socket managed by 'manager'.
364 *
365 * Note:
366 *
367 *\li   'fd' is the already-opened file descriptor.
368 *\li	This function is not available on Windows.
369 *\li	The callback function is called "in-line" - this means the function
370 *	needs to return as fast as possible, as all other I/O will be suspended
371 *	until the callback completes.
372 *
373 * Requires:
374 *
375 *\li	'manager' is a valid manager
376 *
377 *\li	'socketp' is a valid pointer, and *socketp == NULL
378 *
379 *\li	'fd' be opened.
380 *
381 * Ensures:
382 *
383 *	'*socketp' is attached to the newly created fdwatch socket
384 *
385 * Returns:
386 *
387 *\li	#ISC_R_SUCCESS
388 *\li	#ISC_R_NOMEMORY
389 *\li	#ISC_R_NORESOURCES
390 *\li	#ISC_R_UNEXPECTED
391 */
392
393isc_result_t
394isc_socket_fdwatchpoke(isc_socket_t *sock,
395		       int flags);
396/*%<
397 * Poke a file descriptor watch socket informing the manager that it
398 * should restart watching the socket
399 *
400 * Note:
401 *
402 *\li   'sock' is the socket returned by isc_socket_fdwatchcreate
403 *
404 *\li   'flags' indicates what the manager should watch for on the socket
405 *      in addition to what it may already be watching.  It can be one or
406 *      both of ISC_SOCKFDWATCH_READ and ISC_SOCKFDWATCH_WRITE.  To
407 *      temporarily disable watching on a socket the value indicating
408 *      no more data should be returned from the call back routine.
409 *
410 *\li	This function is not available on Windows.
411 *
412 * Requires:
413 *
414 *\li	'sock' is a valid isc socket
415 *
416 *
417 * Returns:
418 *
419 *\li	#ISC_R_SUCCESS
420 */
421
422isc_result_t
423isc_socket_create(isc_socketmgr_t *manager,
424		  int pf,
425		  isc_sockettype_t type,
426		  isc_socket_t **socketp);
427/*%<
428 * Create a new 'type' socket managed by 'manager'.
429 *
430 * For isc_sockettype_fdwatch sockets you should use isc_socket_fdwatchcreate()
431 * rather than isc_socket_create().
432 *
433 * Note:
434 *
435 *\li	'pf' is the desired protocol family, e.g. PF_INET or PF_INET6.
436 *
437 * Requires:
438 *
439 *\li	'manager' is a valid manager
440 *
441 *\li	'socketp' is a valid pointer, and *socketp == NULL
442 *
443 *\li	'type' is not isc_sockettype_fdwatch
444 *
445 * Ensures:
446 *
447 *	'*socketp' is attached to the newly created socket
448 *
449 * Returns:
450 *
451 *\li	#ISC_R_SUCCESS
452 *\li	#ISC_R_NOMEMORY
453 *\li	#ISC_R_NORESOURCES
454 *\li	#ISC_R_UNEXPECTED
455 */
456
457isc_result_t
458isc_socket_dup(isc_socket_t *sock0, isc_socket_t **socketp);
459/*%<
460 * Duplicate an existing socket, reusing its file descriptor.
461 */
462
463void
464isc_socket_cancel(isc_socket_t *sock, isc_task_t *task,
465		  unsigned int how);
466/*%<
467 * Cancel pending I/O of the type specified by "how".
468 *
469 * Note: if "task" is NULL, then the cancel applies to all tasks using the
470 * socket.
471 *
472 * Requires:
473 *
474 * \li	"socket" is a valid socket
475 *
476 * \li	"task" is NULL or a valid task
477 *
478 * "how" is a bitmask describing the type of cancelation to perform.
479 * The type ISC_SOCKCANCEL_ALL will cancel all pending I/O on this
480 * socket.
481 *
482 * \li ISC_SOCKCANCEL_RECV:
483 *	Cancel pending isc_socket_recv() calls.
484 *
485 * \li ISC_SOCKCANCEL_SEND:
486 *	Cancel pending isc_socket_send() and isc_socket_sendto() calls.
487 *
488 * \li ISC_SOCKCANCEL_ACCEPT:
489 *	Cancel pending isc_socket_accept() calls.
490 *
491 * \li ISC_SOCKCANCEL_CONNECT:
492 *	Cancel pending isc_socket_connect() call.
493 */
494
495void
496isc_socket_shutdown(isc_socket_t *sock, unsigned int how);
497/*%<
498 * Shutdown 'socket' according to 'how'.
499 *
500 * Requires:
501 *
502 * \li	'socket' is a valid socket.
503 *
504 * \li	'task' is NULL or is a valid task.
505 *
506 * \li	If 'how' is 'ISC_SOCKSHUT_RECV' or 'ISC_SOCKSHUT_ALL' then
507 *
508 *		The read queue must be empty.
509 *
510 *		No further read requests may be made.
511 *
512 * \li	If 'how' is 'ISC_SOCKSHUT_SEND' or 'ISC_SOCKSHUT_ALL' then
513 *
514 *		The write queue must be empty.
515 *
516 *		No further write requests may be made.
517 */
518
519void
520isc_socket_attach(isc_socket_t *sock, isc_socket_t **socketp);
521/*%<
522 * Attach *socketp to socket.
523 *
524 * Requires:
525 *
526 * \li	'socket' is a valid socket.
527 *
528 * \li	'socketp' points to a NULL socket.
529 *
530 * Ensures:
531 *
532 * \li	*socketp is attached to socket.
533 */
534
535void
536isc_socket_detach(isc_socket_t **socketp);
537/*%<
538 * Detach *socketp from its socket.
539 *
540 * Requires:
541 *
542 * \li	'socketp' points to a valid socket.
543 *
544 * \li	If '*socketp' is the last reference to the socket,
545 *	then:
546 *
547 *		There must be no pending I/O requests.
548 *
549 * Ensures:
550 *
551 * \li	*socketp is NULL.
552 *
553 * \li	If '*socketp' is the last reference to the socket,
554 *	then:
555 *
556 *		The socket will be shutdown (both reading and writing)
557 *		for all tasks.
558 *
559 *		All resources used by the socket have been freed
560 */
561
562isc_result_t
563isc_socket_open(isc_socket_t *sock);
564/*%<
565 * Open a new socket file descriptor of the given socket structure.  It simply
566 * opens a new descriptor; all of the other parameters including the socket
567 * type are inherited from the existing socket.  This function is provided to
568 * avoid overhead of destroying and creating sockets when many short-lived
569 * sockets are frequently opened and closed.  When the efficiency is not an
570 * issue, it should be safer to detach the unused socket and re-create a new
571 * one.  This optimization may not be available for some systems, in which
572 * case this function will return ISC_R_NOTIMPLEMENTED and must not be used.
573 *
574 * isc_socket_open() should not be called on sockets created by
575 * isc_socket_fdwatchcreate().
576 *
577 * Requires:
578 *
579 * \li	there must be no other reference to this socket.
580 *
581 * \li	'socket' is a valid and previously closed by isc_socket_close()
582 *
583 * \li  'sock->type' is not isc_sockettype_fdwatch
584 *
585 * Returns:
586 *	Same as isc_socket_create().
587 * \li	ISC_R_NOTIMPLEMENTED
588 */
589
590isc_result_t
591isc_socket_close(isc_socket_t *sock);
592/*%<
593 * Close a socket file descriptor of the given socket structure.  This function
594 * is provided as an alternative to destroying an unused socket when overhead
595 * destroying/re-creating sockets can be significant, and is expected to be
596 * used with isc_socket_open().  This optimization may not be available for some
597 * systems, in which case this function will return ISC_R_NOTIMPLEMENTED and
598 * must not be used.
599 *
600 * isc_socket_close() should not be called on sockets created by
601 * isc_socket_fdwatchcreate().
602 *
603 * Requires:
604 *
605 * \li	The socket must have a valid descriptor.
606 *
607 * \li	There must be no other reference to this socket.
608 *
609 * \li	There must be no pending I/O requests.
610 *
611 * \li  'sock->type' is not isc_sockettype_fdwatch
612 *
613 * Returns:
614 * \li	#ISC_R_NOTIMPLEMENTED
615 */
616
617isc_result_t
618isc_socket_bind(isc_socket_t *sock, isc_sockaddr_t *addressp,
619		unsigned int options);
620/*%<
621 * Bind 'socket' to '*addressp'.
622 *
623 * Requires:
624 *
625 * \li	'socket' is a valid socket
626 *
627 * \li	'addressp' points to a valid isc_sockaddr.
628 *
629 * Returns:
630 *
631 * \li	ISC_R_SUCCESS
632 * \li	ISC_R_NOPERM
633 * \li	ISC_R_ADDRNOTAVAIL
634 * \li	ISC_R_ADDRINUSE
635 * \li	ISC_R_BOUND
636 * \li	ISC_R_UNEXPECTED
637 */
638
639isc_result_t
640isc_socket_filter(isc_socket_t *sock, const char *filter);
641/*%<
642 * Inform the kernel that it should perform accept filtering.
643 * If filter is NULL the current filter will be removed.:w
644 */
645
646isc_result_t
647isc_socket_listen(isc_socket_t *sock, unsigned int backlog);
648/*%<
649 * Set listen mode on the socket.  After this call, the only function that
650 * can be used (other than attach and detach) is isc_socket_accept().
651 *
652 * Notes:
653 *
654 * \li	'backlog' is as in the UNIX system call listen() and may be
655 *	ignored by non-UNIX implementations.
656 *
657 * \li	If 'backlog' is zero, a reasonable system default is used, usually
658 *	SOMAXCONN.
659 *
660 * Requires:
661 *
662 * \li	'socket' is a valid, bound TCP socket or a valid, bound UNIX socket.
663 *
664 * Returns:
665 *
666 * \li	ISC_R_SUCCESS
667 * \li	ISC_R_UNEXPECTED
668 */
669
670isc_result_t
671isc_socket_accept(isc_socket_t *sock,
672		  isc_task_t *task, isc_taskaction_t action, const void *arg);
673/*%<
674 * Queue accept event.  When a new connection is received, the task will
675 * get an ISC_SOCKEVENT_NEWCONN event with the sender set to the listen
676 * socket.  The new socket structure is sent inside the isc_socket_newconnev_t
677 * event type, and is attached to the task 'task'.
678 *
679 * REQUIRES:
680 * \li	'socket' is a valid TCP socket that isc_socket_listen() was called
681 *	on.
682 *
683 * \li	'task' is a valid task
684 *
685 * \li	'action' is a valid action
686 *
687 * RETURNS:
688 * \li	ISC_R_SUCCESS
689 * \li	ISC_R_NOMEMORY
690 * \li	ISC_R_UNEXPECTED
691 */
692
693isc_result_t
694isc_socket_connect(isc_socket_t *sock, isc_sockaddr_t *addressp,
695		   isc_task_t *task, isc_taskaction_t action,
696		   const void *arg);
697/*%<
698 * Connect 'socket' to peer with address *saddr.  When the connection
699 * succeeds, or when an error occurs, a CONNECT event with action 'action'
700 * and arg 'arg' will be posted to the event queue for 'task'.
701 *
702 * Requires:
703 *
704 * \li	'socket' is a valid TCP socket
705 *
706 * \li	'addressp' points to a valid isc_sockaddr
707 *
708 * \li	'task' is a valid task
709 *
710 * \li	'action' is a valid action
711 *
712 * Returns:
713 *
714 * \li	ISC_R_SUCCESS
715 * \li	ISC_R_NOMEMORY
716 * \li	ISC_R_UNEXPECTED
717 *
718 * Posted event's result code:
719 *
720 * \li	ISC_R_SUCCESS
721 * \li	ISC_R_TIMEDOUT
722 * \li	ISC_R_CONNREFUSED
723 * \li	ISC_R_NETUNREACH
724 * \li	ISC_R_UNEXPECTED
725 */
726
727isc_result_t
728isc_socket_getpeername(isc_socket_t *sock, isc_sockaddr_t *addressp);
729/*%<
730 * Get the name of the peer connected to 'socket'.
731 *
732 * Requires:
733 *
734 * \li	'socket' is a valid TCP socket.
735 *
736 * Returns:
737 *
738 * \li	ISC_R_SUCCESS
739 * \li	ISC_R_TOOSMALL
740 * \li	ISC_R_UNEXPECTED
741 */
742
743isc_result_t
744isc_socket_getsockname(isc_socket_t *sock, isc_sockaddr_t *addressp);
745/*%<
746 * Get the name of 'socket'.
747 *
748 * Requires:
749 *
750 * \li	'socket' is a valid socket.
751 *
752 * Returns:
753 *
754 * \li	ISC_R_SUCCESS
755 * \li	ISC_R_TOOSMALL
756 * \li	ISC_R_UNEXPECTED
757 */
758
759/*@{*/
760isc_result_t
761isc_socket_recv(isc_socket_t *sock, isc_region_t *region,
762		unsigned int minimum,
763		isc_task_t *task, isc_taskaction_t action, const void *arg);
764isc_result_t
765isc_socket_recvv(isc_socket_t *sock, isc_bufferlist_t *buflist,
766		 unsigned int minimum,
767		 isc_task_t *task, isc_taskaction_t action, const void *arg);
768
769isc_result_t
770isc_socket_recv2(isc_socket_t *sock, isc_region_t *region,
771		 unsigned int minimum, isc_task_t *task,
772		 isc_socketevent_t *event, unsigned int flags);
773
774/*!
775 * Receive from 'socket', storing the results in region.
776 *
777 * Notes:
778 *
779 *\li	Let 'length' refer to the length of 'region' or to the sum of all
780 *	available regions in the list of buffers '*buflist'.
781 *
782 *\li	If 'minimum' is non-zero and at least that many bytes are read,
783 *	the completion event will be posted to the task 'task.'  If minimum
784 *	is zero, the exact number of bytes requested in the region must
785 * 	be read for an event to be posted.  This only makes sense for TCP
786 *	connections, and is always set to 1 byte for UDP.
787 *
788 *\li	The read will complete when the desired number of bytes have been
789 *	read, if end-of-input occurs, or if an error occurs.  A read done
790 *	event with the given 'action' and 'arg' will be posted to the
791 *	event queue of 'task'.
792 *
793 *\li	The caller may not modify 'region', the buffers which are passed
794 *	into this function, or any data they refer to until the completion
795 *	event is received.
796 *
797 *\li	For isc_socket_recvv():
798 *	On successful completion, '*buflist' will be empty, and the list of
799 *	all buffers will be returned in the done event's 'bufferlist'
800 *	member.  On error return, '*buflist' will be unchanged.
801 *
802 *\li	For isc_socket_recv2():
803 *	'event' is not NULL, and the non-socket specific fields are
804 *	expected to be initialized.
805 *
806 *\li	For isc_socket_recv2():
807 *	The only defined value for 'flags' is ISC_SOCKFLAG_IMMEDIATE.  If
808 *	set and the operation completes, the return value will be
809 *	ISC_R_SUCCESS and the event will be filled in and not sent.  If the
810 *	operation does not complete, the return value will be
811 *	ISC_R_INPROGRESS and the event will be sent when the operation
812 *	completes.
813 *
814 * Requires:
815 *
816 *\li	'socket' is a valid, bound socket.
817 *
818 *\li	For isc_socket_recv():
819 *	'region' is a valid region
820 *
821 *\li	For isc_socket_recvv():
822 *	'buflist' is non-NULL, and '*buflist' contain at least one buffer.
823 *
824 *\li	'task' is a valid task
825 *
826 *\li	For isc_socket_recv() and isc_socket_recvv():
827 *	action != NULL and is a valid action
828 *
829 *\li	For isc_socket_recv2():
830 *	event != NULL
831 *
832 * Returns:
833 *
834 *\li	#ISC_R_SUCCESS
835 *\li	#ISC_R_INPROGRESS
836 *\li	#ISC_R_NOMEMORY
837 *\li	#ISC_R_UNEXPECTED
838 *
839 * Event results:
840 *
841 *\li	#ISC_R_SUCCESS
842 *\li	#ISC_R_UNEXPECTED
843 *\li	XXX needs other net-type errors
844 */
845/*@}*/
846
847/*@{*/
848isc_result_t
849isc_socket_send(isc_socket_t *sock, isc_region_t *region,
850		isc_task_t *task, isc_taskaction_t action, const void *arg);
851isc_result_t
852isc_socket_sendto(isc_socket_t *sock, isc_region_t *region,
853		  isc_task_t *task, isc_taskaction_t action, const void *arg,
854		  isc_sockaddr_t *address, struct in6_pktinfo *pktinfo);
855isc_result_t
856isc_socket_sendv(isc_socket_t *sock, isc_bufferlist_t *buflist,
857		 isc_task_t *task, isc_taskaction_t action, const void *arg);
858isc_result_t
859isc_socket_sendtov(isc_socket_t *sock, isc_bufferlist_t *buflist,
860		   isc_task_t *task, isc_taskaction_t action, const void *arg,
861		   isc_sockaddr_t *address, struct in6_pktinfo *pktinfo);
862isc_result_t
863isc_socket_sendto2(isc_socket_t *sock, isc_region_t *region,
864		   isc_task_t *task,
865		   isc_sockaddr_t *address, struct in6_pktinfo *pktinfo,
866		   isc_socketevent_t *event, unsigned int flags);
867
868/*!
869 * Send the contents of 'region' to the socket's peer.
870 *
871 * Notes:
872 *
873 *\li	Shutting down the requestor's task *may* result in any
874 *	still pending writes being dropped or completed, depending on the
875 *	underlying OS implementation.
876 *
877 *\li	If 'action' is NULL, then no completion event will be posted.
878 *
879 *\li	The caller may not modify 'region', the buffers which are passed
880 *	into this function, or any data they refer to until the completion
881 *	event is received.
882 *
883 *\li	For isc_socket_sendv() and isc_socket_sendtov():
884 *	On successful completion, '*buflist' will be empty, and the list of
885 *	all buffers will be returned in the done event's 'bufferlist'
886 *	member.  On error return, '*buflist' will be unchanged.
887 *
888 *\li	For isc_socket_sendto2():
889 *	'event' is not NULL, and the non-socket specific fields are
890 *	expected to be initialized.
891 *
892 *\li	For isc_socket_sendto2():
893 *	The only defined values for 'flags' are ISC_SOCKFLAG_IMMEDIATE
894 *	and ISC_SOCKFLAG_NORETRY.
895 *
896 *\li	If ISC_SOCKFLAG_IMMEDIATE is set and the operation completes, the
897 *	return value will be ISC_R_SUCCESS and the event will be filled
898 *	in and not sent.  If the operation does not complete, the return
899 *	value will be ISC_R_INPROGRESS and the event will be sent when
900 *	the operation completes.
901 *
902 *\li	ISC_SOCKFLAG_NORETRY can only be set for UDP sockets.  If set
903 *	and the send operation fails due to a transient error, the send
904 *	will not be retried and the error will be indicated in the event.
905 *	Using this option along with ISC_SOCKFLAG_IMMEDIATE allows the caller
906 *	to specify a region that is allocated on the stack.
907 *
908 * Requires:
909 *
910 *\li	'socket' is a valid, bound socket.
911 *
912 *\li	For isc_socket_send():
913 *	'region' is a valid region
914 *
915 *\li	For isc_socket_sendv() and isc_socket_sendtov():
916 *	'buflist' is non-NULL, and '*buflist' contain at least one buffer.
917 *
918 *\li	'task' is a valid task
919 *
920 *\li	For isc_socket_sendv(), isc_socket_sendtov(), isc_socket_send(), and
921 *	isc_socket_sendto():
922 *	action == NULL or is a valid action
923 *
924 *\li	For isc_socket_sendto2():
925 *	event != NULL
926 *
927 * Returns:
928 *
929 *\li	#ISC_R_SUCCESS
930 *\li	#ISC_R_INPROGRESS
931 *\li	#ISC_R_NOMEMORY
932 *\li	#ISC_R_UNEXPECTED
933 *
934 * Event results:
935 *
936 *\li	#ISC_R_SUCCESS
937 *\li	#ISC_R_UNEXPECTED
938 *\li	XXX needs other net-type errors
939 */
940/*@}*/
941
942isc_result_t
943isc_socketmgr_createinctx(isc_mem_t *mctx, isc_appctx_t *actx,
944			  isc_socketmgr_t **managerp);
945
946isc_result_t
947isc_socketmgr_create(isc_mem_t *mctx, isc_socketmgr_t **managerp);
948
949isc_result_t
950isc_socketmgr_create2(isc_mem_t *mctx, isc_socketmgr_t **managerp,
951		      unsigned int maxsocks);
952/*%<
953 * Create a socket manager.  If "maxsocks" is non-zero, it specifies the
954 * maximum number of sockets that the created manager should handle.
955 * isc_socketmgr_create() is equivalent of isc_socketmgr_create2() with
956 * "maxsocks" being zero.
957 * isc_socketmgr_createinctx() also associates the new manager with the
958 * specified application context.
959 *
960 * Notes:
961 *
962 *\li	All memory will be allocated in memory context 'mctx'.
963 *
964 * Requires:
965 *
966 *\li	'mctx' is a valid memory context.
967 *
968 *\li	'managerp' points to a NULL isc_socketmgr_t.
969 *
970 *\li	'actx' is a valid application context (for createinctx()).
971 *
972 * Ensures:
973 *
974 *\li	'*managerp' is a valid isc_socketmgr_t.
975 *
976 * Returns:
977 *
978 *\li	#ISC_R_SUCCESS
979 *\li	#ISC_R_NOMEMORY
980 *\li	#ISC_R_UNEXPECTED
981 *\li	#ISC_R_NOTIMPLEMENTED
982 */
983
984isc_result_t
985isc_socketmgr_getmaxsockets(isc_socketmgr_t *manager, unsigned int *nsockp);
986/*%<
987 * Returns in "*nsockp" the maximum number of sockets this manager may open.
988 *
989 * Requires:
990 *
991 *\li	'*manager' is a valid isc_socketmgr_t.
992 *\li	'nsockp' is not NULL.
993 *
994 * Returns:
995 *
996 *\li	#ISC_R_SUCCESS
997 *\li	#ISC_R_NOTIMPLEMENTED
998 */
999
1000void
1001isc_socketmgr_setstats(isc_socketmgr_t *manager, isc_stats_t *stats);
1002/*%<
1003 * Set a general socket statistics counter set 'stats' for 'manager'.
1004 *
1005 * Requires:
1006 * \li	'manager' is valid, hasn't opened any socket, and doesn't have
1007 *	stats already set.
1008 *
1009 *\li	stats is a valid statistics supporting socket statistics counters
1010 *	(see above).
1011 */
1012
1013void
1014isc_socketmgr_destroy(isc_socketmgr_t **managerp);
1015/*%<
1016 * Destroy a socket manager.
1017 *
1018 * Notes:
1019 *
1020 *\li	This routine blocks until there are no sockets left in the manager,
1021 *	so if the caller holds any socket references using the manager, it
1022 *	must detach them before calling isc_socketmgr_destroy() or it will
1023 *	block forever.
1024 *
1025 * Requires:
1026 *
1027 *\li	'*managerp' is a valid isc_socketmgr_t.
1028 *
1029 *\li	All sockets managed by this manager are fully detached.
1030 *
1031 * Ensures:
1032 *
1033 *\li	*managerp == NULL
1034 *
1035 *\li	All resources used by the manager have been freed.
1036 */
1037
1038isc_sockettype_t
1039isc_socket_gettype(isc_socket_t *sock);
1040/*%<
1041 * Returns the socket type for "sock."
1042 *
1043 * Requires:
1044 *
1045 *\li	"sock" is a valid socket.
1046 */
1047
1048/*@{*/
1049isc_boolean_t
1050isc_socket_isbound(isc_socket_t *sock);
1051
1052void
1053isc_socket_ipv6only(isc_socket_t *sock, isc_boolean_t yes);
1054/*%<
1055 * If the socket is an IPv6 socket set/clear the IPV6_IPV6ONLY socket
1056 * option if the host OS supports this option.
1057 *
1058 * Requires:
1059 *\li	'sock' is a valid socket.
1060 */
1061/*@}*/
1062
1063void
1064isc_socket_cleanunix(isc_sockaddr_t *addr, isc_boolean_t active);
1065
1066/*%<
1067 * Cleanup UNIX domain sockets in the file-system.  If 'active' is true
1068 * then just unlink the socket.  If 'active' is false try to determine
1069 * if there is a listener of the socket or not.  If no listener is found
1070 * then unlink socket.
1071 *
1072 * Prior to unlinking the path is tested to see if it a socket.
1073 *
1074 * Note: there are a number of race conditions which cannot be avoided
1075 *       both in the filesystem and any application using UNIX domain
1076 *	 sockets (e.g. socket is tested between bind() and listen(),
1077 *	 the socket is deleted and replaced in the file-system between
1078 *	 stat() and unlink()).
1079 */
1080
1081isc_result_t
1082isc_socket_permunix(isc_sockaddr_t *sockaddr, isc_uint32_t perm,
1083		    isc_uint32_t owner, isc_uint32_t group);
1084/*%<
1085 * Set ownership and file permissions on the UNIX domain socket.
1086 *
1087 * Note: On Solaris and SunOS this secures the directory containing
1088 *       the socket as Solaris and SunOS do not honour the filesystem
1089 *	 permissions on the socket.
1090 *
1091 * Requires:
1092 * \li	'sockaddr' to be a valid UNIX domain sockaddr.
1093 *
1094 * Returns:
1095 * \li	#ISC_R_SUCCESS
1096 * \li	#ISC_R_FAILURE
1097 */
1098
1099void isc_socket_setname(isc_socket_t *socket, const char *name, void *tag);
1100/*%<
1101 * Set the name and optional tag for a socket.  This allows tracking of the
1102 * owner or purpose for this socket, and is useful for tracing and statistics
1103 * reporting.
1104 */
1105
1106const char *isc_socket_getname(isc_socket_t *socket);
1107/*%<
1108 * Get the name associated with a socket, if any.
1109 */
1110
1111void *isc_socket_gettag(isc_socket_t *socket);
1112/*%<
1113 * Get the tag associated with a socket, if any.
1114 */
1115
1116int isc_socket_getfd(isc_socket_t *socket);
1117/*%<
1118 * Get the file descriptor associated with a socket
1119 */
1120
1121void
1122isc__socketmgr_setreserved(isc_socketmgr_t *mgr, isc_uint32_t);
1123/*%<
1124 * Temporary.  For use by named only.
1125 */
1126
1127void
1128isc__socketmgr_maxudp(isc_socketmgr_t *mgr, int maxudp);
1129/*%<
1130 * Test interface. Drop UDP packet > 'maxudp'.
1131 */
1132
1133#ifdef HAVE_LIBXML2
1134
1135void
1136isc_socketmgr_renderxml(isc_socketmgr_t *mgr, xmlTextWriterPtr writer);
1137/*%<
1138 * Render internal statistics and other state into the XML document.
1139 */
1140
1141#endif /* HAVE_LIBXML2 */
1142
1143#ifdef USE_SOCKETIMPREGISTER
1144/*%<
1145 * See isc_socketmgr_create() above.
1146 */
1147typedef isc_result_t
1148(*isc_socketmgrcreatefunc_t)(isc_mem_t *mctx, isc_socketmgr_t **managerp);
1149
1150isc_result_t
1151isc_socket_register(isc_socketmgrcreatefunc_t createfunc);
1152/*%<
1153 * Register a new socket I/O implementation and add it to the list of
1154 * supported implementations.  This function must be called when a different
1155 * event library is used than the one contained in the ISC library.
1156 */
1157
1158isc_result_t
1159isc__socket_register(void);
1160/*%<
1161 * A short cut function that specifies the socket I/O module in the ISC
1162 * library for isc_socket_register().  An application that uses the ISC library
1163 * usually do not have to care about this function: it would call
1164 * isc_lib_register(), which internally calls this function.
1165 */
1166#endif /* USE_SOCKETIMPREGISTER */
1167
1168ISC_LANG_ENDDECLS
1169
1170#endif /* ISC_SOCKET_H */
1171