1/*
2 * Copyright (C) 2004-2009, 2011-2013  Internet Systems Consortium, Inc. ("ISC")
3 * Copyright (C) 1998-2002  Internet Software Consortium.
4 *
5 * Permission to use, copy, modify, and/or distribute this software for any
6 * purpose with or without fee is hereby granted, provided that the above
7 * copyright notice and this permission notice appear in all copies.
8 *
9 * THE SOFTWARE IS PROVIDED "AS IS" AND ISC DISCLAIMS ALL WARRANTIES WITH
10 * REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY
11 * AND FITNESS.  IN NO EVENT SHALL ISC BE LIABLE FOR ANY SPECIAL, DIRECT,
12 * INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM
13 * LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
14 * OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
15 * PERFORMANCE OF THIS SOFTWARE.
16 */
17
18/* $Id$ */
19
20#ifndef ISC_SOCKET_H
21#define ISC_SOCKET_H 1
22
23/*****
24 ***** Module Info
25 *****/
26
27/*! \file isc/socket.h
28 * \brief Provides TCP and UDP sockets for network I/O.  The sockets are event
29 * sources in the task system.
30 *
31 * When I/O completes, a completion event for the socket is posted to the
32 * event queue of the task which requested the I/O.
33 *
34 * \li MP:
35 *	The module ensures appropriate synchronization of data structures it
36 *	creates and manipulates.
37 *	Clients of this module must not be holding a socket's task's lock when
38 *	making a call that affects that socket.  Failure to follow this rule
39 *	can result in deadlock.
40 *	The caller must ensure that isc_socketmgr_destroy() is called only
41 *	once for a given manager.
42 *
43 * \li Reliability:
44 *	No anticipated impact.
45 *
46 * \li Resources:
47 *	TBS
48 *
49 * \li Security:
50 *	No anticipated impact.
51 *
52 * \li Standards:
53 *	None.
54 */
55
56/***
57 *** Imports
58 ***/
59
60#include <isc/lang.h>
61#include <isc/types.h>
62#include <isc/event.h>
63#include <isc/eventclass.h>
64#include <isc/time.h>
65#include <isc/region.h>
66#include <isc/sockaddr.h>
67#include <isc/xml.h>
68
69ISC_LANG_BEGINDECLS
70
71/***
72 *** Constants
73 ***/
74
75/*%
76 * Maximum number of buffers in a scatter/gather read/write.  The operating
77 * system in use must support at least this number (plus one on some.)
78 */
79#define ISC_SOCKET_MAXSCATTERGATHER	8
80
81/*%
82 * In isc_socket_bind() set socket option SO_REUSEADDR prior to calling
83 * bind() if a non zero port is specified (AF_INET and AF_INET6).
84 */
85#define ISC_SOCKET_REUSEADDRESS		0x01U
86
87/*%
88 * Statistics counters.  Used as isc_statscounter_t values.
89 */
90enum {
91	isc_sockstatscounter_udp4open = 0,
92	isc_sockstatscounter_udp6open = 1,
93	isc_sockstatscounter_tcp4open = 2,
94	isc_sockstatscounter_tcp6open = 3,
95	isc_sockstatscounter_unixopen = 4,
96
97	isc_sockstatscounter_udp4openfail = 5,
98	isc_sockstatscounter_udp6openfail = 6,
99	isc_sockstatscounter_tcp4openfail = 7,
100	isc_sockstatscounter_tcp6openfail = 8,
101	isc_sockstatscounter_unixopenfail = 9,
102
103	isc_sockstatscounter_udp4close = 10,
104	isc_sockstatscounter_udp6close = 11,
105	isc_sockstatscounter_tcp4close = 12,
106	isc_sockstatscounter_tcp6close = 13,
107	isc_sockstatscounter_unixclose = 14,
108	isc_sockstatscounter_fdwatchclose = 15,
109
110	isc_sockstatscounter_udp4bindfail = 16,
111	isc_sockstatscounter_udp6bindfail = 17,
112	isc_sockstatscounter_tcp4bindfail = 18,
113	isc_sockstatscounter_tcp6bindfail = 19,
114	isc_sockstatscounter_unixbindfail = 20,
115	isc_sockstatscounter_fdwatchbindfail = 21,
116
117	isc_sockstatscounter_udp4connect = 22,
118	isc_sockstatscounter_udp6connect = 23,
119	isc_sockstatscounter_tcp4connect = 24,
120	isc_sockstatscounter_tcp6connect = 25,
121	isc_sockstatscounter_unixconnect = 26,
122	isc_sockstatscounter_fdwatchconnect = 27,
123
124	isc_sockstatscounter_udp4connectfail = 28,
125	isc_sockstatscounter_udp6connectfail = 29,
126	isc_sockstatscounter_tcp4connectfail = 30,
127	isc_sockstatscounter_tcp6connectfail = 31,
128	isc_sockstatscounter_unixconnectfail = 32,
129	isc_sockstatscounter_fdwatchconnectfail = 33,
130
131	isc_sockstatscounter_tcp4accept = 34,
132	isc_sockstatscounter_tcp6accept = 35,
133	isc_sockstatscounter_unixaccept = 36,
134
135	isc_sockstatscounter_tcp4acceptfail = 37,
136	isc_sockstatscounter_tcp6acceptfail = 38,
137	isc_sockstatscounter_unixacceptfail = 39,
138
139	isc_sockstatscounter_udp4sendfail = 40,
140	isc_sockstatscounter_udp6sendfail = 41,
141	isc_sockstatscounter_tcp4sendfail = 42,
142	isc_sockstatscounter_tcp6sendfail = 43,
143	isc_sockstatscounter_unixsendfail = 44,
144	isc_sockstatscounter_fdwatchsendfail = 45,
145
146	isc_sockstatscounter_udp4recvfail = 46,
147	isc_sockstatscounter_udp6recvfail = 47,
148	isc_sockstatscounter_tcp4recvfail = 48,
149	isc_sockstatscounter_tcp6recvfail = 49,
150	isc_sockstatscounter_unixrecvfail = 50,
151	isc_sockstatscounter_fdwatchrecvfail = 51,
152
153	isc_sockstatscounter_max = 52
154};
155
156/***
157 *** Types
158 ***/
159
160struct isc_socketevent {
161	ISC_EVENT_COMMON(isc_socketevent_t);
162	isc_result_t		result;		/*%< OK, EOF, whatever else */
163	unsigned int		minimum;	/*%< minimum i/o for event */
164	unsigned int		n;		/*%< bytes read or written */
165	unsigned int		offset;		/*%< offset into buffer list */
166	isc_region_t		region;		/*%< for single-buffer i/o */
167	isc_bufferlist_t	bufferlist;	/*%< list of buffers */
168	isc_sockaddr_t		address;	/*%< source address */
169	isc_time_t		timestamp;	/*%< timestamp of packet recv */
170	struct in6_pktinfo	pktinfo;	/*%< ipv6 pktinfo */
171	isc_uint32_t		attributes;	/*%< see below */
172	isc_eventdestructor_t   destroy;	/*%< original destructor */
173};
174
175typedef struct isc_socket_newconnev isc_socket_newconnev_t;
176struct isc_socket_newconnev {
177	ISC_EVENT_COMMON(isc_socket_newconnev_t);
178	isc_socket_t *		newsocket;
179	isc_result_t		result;		/*%< OK, EOF, whatever else */
180	isc_sockaddr_t		address;	/*%< source address */
181};
182
183typedef struct isc_socket_connev isc_socket_connev_t;
184struct isc_socket_connev {
185	ISC_EVENT_COMMON(isc_socket_connev_t);
186	isc_result_t		result;		/*%< OK, EOF, whatever else */
187};
188
189/*@{*/
190/*!
191 * _ATTACHED:	Internal use only.
192 * _TRUNC:	Packet was truncated on receive.
193 * _CTRUNC:	Packet control information was truncated.  This can
194 *		indicate that the packet is not complete, even though
195 *		all the data is valid.
196 * _TIMESTAMP:	The timestamp member is valid.
197 * _PKTINFO:	The pktinfo member is valid.
198 * _MULTICAST:	The UDP packet was received via a multicast transmission.
199 */
200#define ISC_SOCKEVENTATTR_ATTACHED		0x80000000U /* internal */
201#define ISC_SOCKEVENTATTR_TRUNC			0x00800000U /* public */
202#define ISC_SOCKEVENTATTR_CTRUNC		0x00400000U /* public */
203#define ISC_SOCKEVENTATTR_TIMESTAMP		0x00200000U /* public */
204#define ISC_SOCKEVENTATTR_PKTINFO		0x00100000U /* public */
205#define ISC_SOCKEVENTATTR_MULTICAST		0x00080000U /* public */
206/*@}*/
207
208#define ISC_SOCKEVENT_ANYEVENT  (0)
209#define ISC_SOCKEVENT_RECVDONE	(ISC_EVENTCLASS_SOCKET + 1)
210#define ISC_SOCKEVENT_SENDDONE	(ISC_EVENTCLASS_SOCKET + 2)
211#define ISC_SOCKEVENT_NEWCONN	(ISC_EVENTCLASS_SOCKET + 3)
212#define ISC_SOCKEVENT_CONNECT	(ISC_EVENTCLASS_SOCKET + 4)
213
214/*
215 * Internal events.
216 */
217#define ISC_SOCKEVENT_INTR	(ISC_EVENTCLASS_SOCKET + 256)
218#define ISC_SOCKEVENT_INTW	(ISC_EVENTCLASS_SOCKET + 257)
219
220typedef enum {
221	isc_sockettype_udp = 1,
222	isc_sockettype_tcp = 2,
223	isc_sockettype_unix = 3,
224	isc_sockettype_fdwatch = 4
225} isc_sockettype_t;
226
227/*@{*/
228/*!
229 * How a socket should be shutdown in isc_socket_shutdown() calls.
230 */
231#define ISC_SOCKSHUT_RECV	0x00000001	/*%< close read side */
232#define ISC_SOCKSHUT_SEND	0x00000002	/*%< close write side */
233#define ISC_SOCKSHUT_ALL	0x00000003	/*%< close them all */
234/*@}*/
235
236/*@{*/
237/*!
238 * What I/O events to cancel in isc_socket_cancel() calls.
239 */
240#define ISC_SOCKCANCEL_RECV	0x00000001	/*%< cancel recv */
241#define ISC_SOCKCANCEL_SEND	0x00000002	/*%< cancel send */
242#define ISC_SOCKCANCEL_ACCEPT	0x00000004	/*%< cancel accept */
243#define ISC_SOCKCANCEL_CONNECT	0x00000008	/*%< cancel connect */
244#define ISC_SOCKCANCEL_ALL	0x0000000f	/*%< cancel everything */
245/*@}*/
246
247/*@{*/
248/*!
249 * Flags for isc_socket_send() and isc_socket_recv() calls.
250 */
251#define ISC_SOCKFLAG_IMMEDIATE	0x00000001	/*%< send event only if needed */
252#define ISC_SOCKFLAG_NORETRY	0x00000002	/*%< drop failed UDP sends */
253/*@}*/
254
255/*@{*/
256/*!
257 * Flags for fdwatchcreate.
258 */
259#define ISC_SOCKFDWATCH_READ	0x00000001	/*%< watch for readable */
260#define ISC_SOCKFDWATCH_WRITE	0x00000002	/*%< watch for writable */
261/*@}*/
262
263/*% Socket and socket manager methods */
264typedef struct isc_socketmgrmethods {
265	void		(*destroy)(isc_socketmgr_t **managerp);
266	isc_result_t	(*socketcreate)(isc_socketmgr_t *manager, int pf,
267					isc_sockettype_t type,
268					isc_socket_t **socketp);
269	isc_result_t    (*fdwatchcreate)(isc_socketmgr_t *manager, int fd,
270					 int flags,
271					 isc_sockfdwatch_t callback,
272					 void *cbarg, isc_task_t *task,
273					 isc_socket_t **socketp);
274} isc_socketmgrmethods_t;
275
276typedef struct isc_socketmethods {
277	void		(*attach)(isc_socket_t *socket,
278				  isc_socket_t **socketp);
279	void		(*detach)(isc_socket_t **socketp);
280	isc_result_t	(*bind)(isc_socket_t *sock, isc_sockaddr_t *sockaddr,
281				unsigned int options);
282	isc_result_t	(*sendto)(isc_socket_t *sock, isc_region_t *region,
283				  isc_task_t *task, isc_taskaction_t action,
284				  const void *arg, isc_sockaddr_t *address,
285				  struct in6_pktinfo *pktinfo);
286	isc_result_t	(*sendto2)(isc_socket_t *sock, isc_region_t *region,
287				   isc_task_t *task, isc_sockaddr_t *address,
288				   struct in6_pktinfo *pktinfo,
289				   isc_socketevent_t *event,
290				   unsigned int flags);
291	isc_result_t	(*connect)(isc_socket_t *sock, isc_sockaddr_t *addr,
292				   isc_task_t *task, isc_taskaction_t action,
293				   const void *arg);
294	isc_result_t	(*recv)(isc_socket_t *sock, isc_region_t *region,
295				unsigned int minimum, isc_task_t *task,
296				isc_taskaction_t action, const void *arg);
297	isc_result_t	(*recv2)(isc_socket_t *sock, isc_region_t *region,
298				 unsigned int minimum, isc_task_t *task,
299				 isc_socketevent_t *event, unsigned int flags);
300	void		(*cancel)(isc_socket_t *sock, isc_task_t *task,
301				  unsigned int how);
302	isc_result_t	(*getsockname)(isc_socket_t *sock,
303				       isc_sockaddr_t *addressp);
304	isc_sockettype_t (*gettype)(isc_socket_t *sock);
305	void		(*ipv6only)(isc_socket_t *sock, isc_boolean_t yes);
306	isc_result_t    (*fdwatchpoke)(isc_socket_t *sock, int flags);
307	isc_result_t		(*dup)(isc_socket_t *socket,
308				  isc_socket_t **socketp);
309	int 		(*getfd)(isc_socket_t *socket);
310} isc_socketmethods_t;
311
312/*%
313 * This structure is actually just the common prefix of a socket manager
314 * object implementation's version of an isc_socketmgr_t.
315 * \brief
316 * Direct use of this structure by clients is forbidden.  socket implementations
317 * may change the structure.  'magic' must be ISCAPI_SOCKETMGR_MAGIC for any
318 * of the isc_socket_ routines to work.  socket implementations must maintain
319 * all socket invariants.
320 * In effect, this definition is used only for non-BIND9 version ("export")
321 * of the library, and the export version does not work for win32.  So, to avoid
322 * the definition conflict with win32/socket.c, we enable this definition only
323 * for non-Win32 (i.e. Unix) platforms.
324 */
325#ifndef WIN32
326struct isc_socketmgr {
327	unsigned int		impmagic;
328	unsigned int		magic;
329	isc_socketmgrmethods_t	*methods;
330};
331#endif
332
333#define ISCAPI_SOCKETMGR_MAGIC		ISC_MAGIC('A','s','m','g')
334#define ISCAPI_SOCKETMGR_VALID(m)	((m) != NULL && \
335					 (m)->magic == ISCAPI_SOCKETMGR_MAGIC)
336
337/*%
338 * This is the common prefix of a socket object.  The same note as
339 * that for the socketmgr structure applies.
340 */
341#ifndef WIN32
342struct isc_socket {
343	unsigned int		impmagic;
344	unsigned int		magic;
345	isc_socketmethods_t	*methods;
346};
347#endif
348
349#define ISCAPI_SOCKET_MAGIC	ISC_MAGIC('A','s','c','t')
350#define ISCAPI_SOCKET_VALID(s)	((s) != NULL && \
351				 (s)->magic == ISCAPI_SOCKET_MAGIC)
352
353/***
354 *** Socket and Socket Manager Functions
355 ***
356 *** Note: all Ensures conditions apply only if the result is success for
357 *** those functions which return an isc_result.
358 ***/
359
360isc_result_t
361isc_socket_fdwatchcreate(isc_socketmgr_t *manager,
362			 int fd,
363			 int flags,
364			 isc_sockfdwatch_t callback,
365			 void *cbarg,
366			 isc_task_t *task,
367			 isc_socket_t **socketp);
368/*%<
369 * Create a new file descriptor watch socket managed by 'manager'.
370 *
371 * Note:
372 *
373 *\li   'fd' is the already-opened file descriptor.
374 *\li	This function is not available on Windows.
375 *\li	The callback function is called "in-line" - this means the function
376 *	needs to return as fast as possible, as all other I/O will be suspended
377 *	until the callback completes.
378 *
379 * Requires:
380 *
381 *\li	'manager' is a valid manager
382 *
383 *\li	'socketp' is a valid pointer, and *socketp == NULL
384 *
385 *\li	'fd' be opened.
386 *
387 * Ensures:
388 *
389 *	'*socketp' is attached to the newly created fdwatch socket
390 *
391 * Returns:
392 *
393 *\li	#ISC_R_SUCCESS
394 *\li	#ISC_R_NOMEMORY
395 *\li	#ISC_R_NORESOURCES
396 *\li	#ISC_R_UNEXPECTED
397 */
398
399isc_result_t
400isc_socket_fdwatchpoke(isc_socket_t *sock,
401		       int flags);
402/*%<
403 * Poke a file descriptor watch socket informing the manager that it
404 * should restart watching the socket
405 *
406 * Note:
407 *
408 *\li   'sock' is the socket returned by isc_socket_fdwatchcreate
409 *
410 *\li   'flags' indicates what the manager should watch for on the socket
411 *      in addition to what it may already be watching.  It can be one or
412 *      both of ISC_SOCKFDWATCH_READ and ISC_SOCKFDWATCH_WRITE.  To
413 *      temporarily disable watching on a socket the value indicating
414 *      no more data should be returned from the call back routine.
415 *
416 *\li	This function is not available on Windows.
417 *
418 * Requires:
419 *
420 *\li	'sock' is a valid isc socket
421 *
422 *
423 * Returns:
424 *
425 *\li	#ISC_R_SUCCESS
426 */
427
428isc_result_t
429isc_socket_create(isc_socketmgr_t *manager,
430		  int pf,
431		  isc_sockettype_t type,
432		  isc_socket_t **socketp);
433/*%<
434 * Create a new 'type' socket managed by 'manager'.
435 *
436 * For isc_sockettype_fdwatch sockets you should use isc_socket_fdwatchcreate()
437 * rather than isc_socket_create().
438 *
439 * Note:
440 *
441 *\li	'pf' is the desired protocol family, e.g. PF_INET or PF_INET6.
442 *
443 * Requires:
444 *
445 *\li	'manager' is a valid manager
446 *
447 *\li	'socketp' is a valid pointer, and *socketp == NULL
448 *
449 *\li	'type' is not isc_sockettype_fdwatch
450 *
451 * Ensures:
452 *
453 *	'*socketp' is attached to the newly created socket
454 *
455 * Returns:
456 *
457 *\li	#ISC_R_SUCCESS
458 *\li	#ISC_R_NOMEMORY
459 *\li	#ISC_R_NORESOURCES
460 *\li	#ISC_R_UNEXPECTED
461 */
462
463isc_result_t
464isc_socket_dup(isc_socket_t *sock0, isc_socket_t **socketp);
465/*%<
466 * Duplicate an existing socket, reusing its file descriptor.
467 */
468
469void
470isc_socket_cancel(isc_socket_t *sock, isc_task_t *task,
471		  unsigned int how);
472/*%<
473 * Cancel pending I/O of the type specified by "how".
474 *
475 * Note: if "task" is NULL, then the cancel applies to all tasks using the
476 * socket.
477 *
478 * Requires:
479 *
480 * \li	"socket" is a valid socket
481 *
482 * \li	"task" is NULL or a valid task
483 *
484 * "how" is a bitmask describing the type of cancelation to perform.
485 * The type ISC_SOCKCANCEL_ALL will cancel all pending I/O on this
486 * socket.
487 *
488 * \li ISC_SOCKCANCEL_RECV:
489 *	Cancel pending isc_socket_recv() calls.
490 *
491 * \li ISC_SOCKCANCEL_SEND:
492 *	Cancel pending isc_socket_send() and isc_socket_sendto() calls.
493 *
494 * \li ISC_SOCKCANCEL_ACCEPT:
495 *	Cancel pending isc_socket_accept() calls.
496 *
497 * \li ISC_SOCKCANCEL_CONNECT:
498 *	Cancel pending isc_socket_connect() call.
499 */
500
501void
502isc_socket_shutdown(isc_socket_t *sock, unsigned int how);
503/*%<
504 * Shutdown 'socket' according to 'how'.
505 *
506 * Requires:
507 *
508 * \li	'socket' is a valid socket.
509 *
510 * \li	'task' is NULL or is a valid task.
511 *
512 * \li	If 'how' is 'ISC_SOCKSHUT_RECV' or 'ISC_SOCKSHUT_ALL' then
513 *
514 *		The read queue must be empty.
515 *
516 *		No further read requests may be made.
517 *
518 * \li	If 'how' is 'ISC_SOCKSHUT_SEND' or 'ISC_SOCKSHUT_ALL' then
519 *
520 *		The write queue must be empty.
521 *
522 *		No further write requests may be made.
523 */
524
525void
526isc_socket_attach(isc_socket_t *sock, isc_socket_t **socketp);
527/*%<
528 * Attach *socketp to socket.
529 *
530 * Requires:
531 *
532 * \li	'socket' is a valid socket.
533 *
534 * \li	'socketp' points to a NULL socket.
535 *
536 * Ensures:
537 *
538 * \li	*socketp is attached to socket.
539 */
540
541void
542isc_socket_detach(isc_socket_t **socketp);
543/*%<
544 * Detach *socketp from its socket.
545 *
546 * Requires:
547 *
548 * \li	'socketp' points to a valid socket.
549 *
550 * \li	If '*socketp' is the last reference to the socket,
551 *	then:
552 *
553 *		There must be no pending I/O requests.
554 *
555 * Ensures:
556 *
557 * \li	*socketp is NULL.
558 *
559 * \li	If '*socketp' is the last reference to the socket,
560 *	then:
561 *
562 *		The socket will be shutdown (both reading and writing)
563 *		for all tasks.
564 *
565 *		All resources used by the socket have been freed
566 */
567
568isc_result_t
569isc_socket_open(isc_socket_t *sock);
570/*%<
571 * Open a new socket file descriptor of the given socket structure.  It simply
572 * opens a new descriptor; all of the other parameters including the socket
573 * type are inherited from the existing socket.  This function is provided to
574 * avoid overhead of destroying and creating sockets when many short-lived
575 * sockets are frequently opened and closed.  When the efficiency is not an
576 * issue, it should be safer to detach the unused socket and re-create a new
577 * one.  This optimization may not be available for some systems, in which
578 * case this function will return ISC_R_NOTIMPLEMENTED and must not be used.
579 *
580 * isc_socket_open() should not be called on sockets created by
581 * isc_socket_fdwatchcreate().
582 *
583 * Requires:
584 *
585 * \li	there must be no other reference to this socket.
586 *
587 * \li	'socket' is a valid and previously closed by isc_socket_close()
588 *
589 * \li  'sock->type' is not isc_sockettype_fdwatch
590 *
591 * Returns:
592 *	Same as isc_socket_create().
593 * \li	ISC_R_NOTIMPLEMENTED
594 */
595
596isc_result_t
597isc_socket_close(isc_socket_t *sock);
598/*%<
599 * Close a socket file descriptor of the given socket structure.  This function
600 * is provided as an alternative to destroying an unused socket when overhead
601 * destroying/re-creating sockets can be significant, and is expected to be
602 * used with isc_socket_open().  This optimization may not be available for some
603 * systems, in which case this function will return ISC_R_NOTIMPLEMENTED and
604 * must not be used.
605 *
606 * isc_socket_close() should not be called on sockets created by
607 * isc_socket_fdwatchcreate().
608 *
609 * Requires:
610 *
611 * \li	The socket must have a valid descriptor.
612 *
613 * \li	There must be no other reference to this socket.
614 *
615 * \li	There must be no pending I/O requests.
616 *
617 * \li  'sock->type' is not isc_sockettype_fdwatch
618 *
619 * Returns:
620 * \li	#ISC_R_NOTIMPLEMENTED
621 */
622
623isc_result_t
624isc_socket_bind(isc_socket_t *sock, isc_sockaddr_t *addressp,
625		unsigned int options);
626/*%<
627 * Bind 'socket' to '*addressp'.
628 *
629 * Requires:
630 *
631 * \li	'socket' is a valid socket
632 *
633 * \li	'addressp' points to a valid isc_sockaddr.
634 *
635 * Returns:
636 *
637 * \li	ISC_R_SUCCESS
638 * \li	ISC_R_NOPERM
639 * \li	ISC_R_ADDRNOTAVAIL
640 * \li	ISC_R_ADDRINUSE
641 * \li	ISC_R_BOUND
642 * \li	ISC_R_UNEXPECTED
643 */
644
645isc_result_t
646isc_socket_filter(isc_socket_t *sock, const char *filter);
647/*%<
648 * Inform the kernel that it should perform accept filtering.
649 * If filter is NULL the current filter will be removed.:w
650 */
651
652isc_result_t
653isc_socket_listen(isc_socket_t *sock, unsigned int backlog);
654/*%<
655 * Set listen mode on the socket.  After this call, the only function that
656 * can be used (other than attach and detach) is isc_socket_accept().
657 *
658 * Notes:
659 *
660 * \li	'backlog' is as in the UNIX system call listen() and may be
661 *	ignored by non-UNIX implementations.
662 *
663 * \li	If 'backlog' is zero, a reasonable system default is used, usually
664 *	SOMAXCONN.
665 *
666 * Requires:
667 *
668 * \li	'socket' is a valid, bound TCP socket or a valid, bound UNIX socket.
669 *
670 * Returns:
671 *
672 * \li	ISC_R_SUCCESS
673 * \li	ISC_R_UNEXPECTED
674 */
675
676isc_result_t
677isc_socket_accept(isc_socket_t *sock,
678		  isc_task_t *task, isc_taskaction_t action, const void *arg);
679/*%<
680 * Queue accept event.  When a new connection is received, the task will
681 * get an ISC_SOCKEVENT_NEWCONN event with the sender set to the listen
682 * socket.  The new socket structure is sent inside the isc_socket_newconnev_t
683 * event type, and is attached to the task 'task'.
684 *
685 * REQUIRES:
686 * \li	'socket' is a valid TCP socket that isc_socket_listen() was called
687 *	on.
688 *
689 * \li	'task' is a valid task
690 *
691 * \li	'action' is a valid action
692 *
693 * RETURNS:
694 * \li	ISC_R_SUCCESS
695 * \li	ISC_R_NOMEMORY
696 * \li	ISC_R_UNEXPECTED
697 */
698
699isc_result_t
700isc_socket_connect(isc_socket_t *sock, isc_sockaddr_t *addressp,
701		   isc_task_t *task, isc_taskaction_t action,
702		   const void *arg);
703/*%<
704 * Connect 'socket' to peer with address *saddr.  When the connection
705 * succeeds, or when an error occurs, a CONNECT event with action 'action'
706 * and arg 'arg' will be posted to the event queue for 'task'.
707 *
708 * Requires:
709 *
710 * \li	'socket' is a valid TCP socket
711 *
712 * \li	'addressp' points to a valid isc_sockaddr
713 *
714 * \li	'task' is a valid task
715 *
716 * \li	'action' is a valid action
717 *
718 * Returns:
719 *
720 * \li	ISC_R_SUCCESS
721 * \li	ISC_R_NOMEMORY
722 * \li	ISC_R_UNEXPECTED
723 *
724 * Posted event's result code:
725 *
726 * \li	ISC_R_SUCCESS
727 * \li	ISC_R_TIMEDOUT
728 * \li	ISC_R_CONNREFUSED
729 * \li	ISC_R_NETUNREACH
730 * \li	ISC_R_UNEXPECTED
731 */
732
733isc_result_t
734isc_socket_getpeername(isc_socket_t *sock, isc_sockaddr_t *addressp);
735/*%<
736 * Get the name of the peer connected to 'socket'.
737 *
738 * Requires:
739 *
740 * \li	'socket' is a valid TCP socket.
741 *
742 * Returns:
743 *
744 * \li	ISC_R_SUCCESS
745 * \li	ISC_R_TOOSMALL
746 * \li	ISC_R_UNEXPECTED
747 */
748
749isc_result_t
750isc_socket_getsockname(isc_socket_t *sock, isc_sockaddr_t *addressp);
751/*%<
752 * Get the name of 'socket'.
753 *
754 * Requires:
755 *
756 * \li	'socket' is a valid socket.
757 *
758 * Returns:
759 *
760 * \li	ISC_R_SUCCESS
761 * \li	ISC_R_TOOSMALL
762 * \li	ISC_R_UNEXPECTED
763 */
764
765/*@{*/
766isc_result_t
767isc_socket_recv(isc_socket_t *sock, isc_region_t *region,
768		unsigned int minimum,
769		isc_task_t *task, isc_taskaction_t action, const void *arg);
770isc_result_t
771isc_socket_recvv(isc_socket_t *sock, isc_bufferlist_t *buflist,
772		 unsigned int minimum,
773		 isc_task_t *task, isc_taskaction_t action, const void *arg);
774
775isc_result_t
776isc_socket_recv2(isc_socket_t *sock, isc_region_t *region,
777		 unsigned int minimum, isc_task_t *task,
778		 isc_socketevent_t *event, unsigned int flags);
779
780/*!
781 * Receive from 'socket', storing the results in region.
782 *
783 * Notes:
784 *
785 *\li	Let 'length' refer to the length of 'region' or to the sum of all
786 *	available regions in the list of buffers '*buflist'.
787 *
788 *\li	If 'minimum' is non-zero and at least that many bytes are read,
789 *	the completion event will be posted to the task 'task.'  If minimum
790 *	is zero, the exact number of bytes requested in the region must
791 * 	be read for an event to be posted.  This only makes sense for TCP
792 *	connections, and is always set to 1 byte for UDP.
793 *
794 *\li	The read will complete when the desired number of bytes have been
795 *	read, if end-of-input occurs, or if an error occurs.  A read done
796 *	event with the given 'action' and 'arg' will be posted to the
797 *	event queue of 'task'.
798 *
799 *\li	The caller may not modify 'region', the buffers which are passed
800 *	into this function, or any data they refer to until the completion
801 *	event is received.
802 *
803 *\li	For isc_socket_recvv():
804 *	On successful completion, '*buflist' will be empty, and the list of
805 *	all buffers will be returned in the done event's 'bufferlist'
806 *	member.  On error return, '*buflist' will be unchanged.
807 *
808 *\li	For isc_socket_recv2():
809 *	'event' is not NULL, and the non-socket specific fields are
810 *	expected to be initialized.
811 *
812 *\li	For isc_socket_recv2():
813 *	The only defined value for 'flags' is ISC_SOCKFLAG_IMMEDIATE.  If
814 *	set and the operation completes, the return value will be
815 *	ISC_R_SUCCESS and the event will be filled in and not sent.  If the
816 *	operation does not complete, the return value will be
817 *	ISC_R_INPROGRESS and the event will be sent when the operation
818 *	completes.
819 *
820 * Requires:
821 *
822 *\li	'socket' is a valid, bound socket.
823 *
824 *\li	For isc_socket_recv():
825 *	'region' is a valid region
826 *
827 *\li	For isc_socket_recvv():
828 *	'buflist' is non-NULL, and '*buflist' contain at least one buffer.
829 *
830 *\li	'task' is a valid task
831 *
832 *\li	For isc_socket_recv() and isc_socket_recvv():
833 *	action != NULL and is a valid action
834 *
835 *\li	For isc_socket_recv2():
836 *	event != NULL
837 *
838 * Returns:
839 *
840 *\li	#ISC_R_SUCCESS
841 *\li	#ISC_R_INPROGRESS
842 *\li	#ISC_R_NOMEMORY
843 *\li	#ISC_R_UNEXPECTED
844 *
845 * Event results:
846 *
847 *\li	#ISC_R_SUCCESS
848 *\li	#ISC_R_UNEXPECTED
849 *\li	XXX needs other net-type errors
850 */
851/*@}*/
852
853/*@{*/
854isc_result_t
855isc_socket_send(isc_socket_t *sock, isc_region_t *region,
856		isc_task_t *task, isc_taskaction_t action, const void *arg);
857isc_result_t
858isc_socket_sendto(isc_socket_t *sock, isc_region_t *region,
859		  isc_task_t *task, isc_taskaction_t action, const void *arg,
860		  isc_sockaddr_t *address, struct in6_pktinfo *pktinfo);
861isc_result_t
862isc_socket_sendv(isc_socket_t *sock, isc_bufferlist_t *buflist,
863		 isc_task_t *task, isc_taskaction_t action, const void *arg);
864isc_result_t
865isc_socket_sendtov(isc_socket_t *sock, isc_bufferlist_t *buflist,
866		   isc_task_t *task, isc_taskaction_t action, const void *arg,
867		   isc_sockaddr_t *address, struct in6_pktinfo *pktinfo);
868isc_result_t
869isc_socket_sendtov2(isc_socket_t *sock, isc_bufferlist_t *buflist,
870		    isc_task_t *task, isc_taskaction_t action, const void *arg,
871		    isc_sockaddr_t *address, struct in6_pktinfo *pktinfo,
872		    unsigned int flags);
873isc_result_t
874isc_socket_sendto2(isc_socket_t *sock, isc_region_t *region,
875		   isc_task_t *task,
876		   isc_sockaddr_t *address, struct in6_pktinfo *pktinfo,
877		   isc_socketevent_t *event, unsigned int flags);
878
879/*!
880 * Send the contents of 'region' to the socket's peer.
881 *
882 * Notes:
883 *
884 *\li	Shutting down the requestor's task *may* result in any
885 *	still pending writes being dropped or completed, depending on the
886 *	underlying OS implementation.
887 *
888 *\li	If 'action' is NULL, then no completion event will be posted.
889 *
890 *\li	The caller may not modify 'region', the buffers which are passed
891 *	into this function, or any data they refer to until the completion
892 *	event is received.
893 *
894 *\li	For isc_socket_sendv() and isc_socket_sendtov():
895 *	On successful completion, '*buflist' will be empty, and the list of
896 *	all buffers will be returned in the done event's 'bufferlist'
897 *	member.  On error return, '*buflist' will be unchanged.
898 *
899 *\li	For isc_socket_sendto2():
900 *	'event' is not NULL, and the non-socket specific fields are
901 *	expected to be initialized.
902 *
903 *\li	For isc_socket_sendto2():
904 *	The only defined values for 'flags' are ISC_SOCKFLAG_IMMEDIATE
905 *	and ISC_SOCKFLAG_NORETRY.
906 *
907 *\li	If ISC_SOCKFLAG_IMMEDIATE is set and the operation completes, the
908 *	return value will be ISC_R_SUCCESS and the event will be filled
909 *	in and not sent.  If the operation does not complete, the return
910 *	value will be ISC_R_INPROGRESS and the event will be sent when
911 *	the operation completes.
912 *
913 *\li	ISC_SOCKFLAG_NORETRY can only be set for UDP sockets.  If set
914 *	and the send operation fails due to a transient error, the send
915 *	will not be retried and the error will be indicated in the event.
916 *	Using this option along with ISC_SOCKFLAG_IMMEDIATE allows the caller
917 *	to specify a region that is allocated on the stack.
918 *
919 * Requires:
920 *
921 *\li	'socket' is a valid, bound socket.
922 *
923 *\li	For isc_socket_send():
924 *	'region' is a valid region
925 *
926 *\li	For isc_socket_sendv() and isc_socket_sendtov():
927 *	'buflist' is non-NULL, and '*buflist' contain at least one buffer.
928 *
929 *\li	'task' is a valid task
930 *
931 *\li	For isc_socket_sendv(), isc_socket_sendtov(), isc_socket_send(), and
932 *	isc_socket_sendto():
933 *	action == NULL or is a valid action
934 *
935 *\li	For isc_socket_sendto2():
936 *	event != NULL
937 *
938 * Returns:
939 *
940 *\li	#ISC_R_SUCCESS
941 *\li	#ISC_R_INPROGRESS
942 *\li	#ISC_R_NOMEMORY
943 *\li	#ISC_R_UNEXPECTED
944 *
945 * Event results:
946 *
947 *\li	#ISC_R_SUCCESS
948 *\li	#ISC_R_UNEXPECTED
949 *\li	XXX needs other net-type errors
950 */
951/*@}*/
952
953isc_result_t
954isc_socketmgr_createinctx(isc_mem_t *mctx, isc_appctx_t *actx,
955			  isc_socketmgr_t **managerp);
956
957isc_result_t
958isc_socketmgr_create(isc_mem_t *mctx, isc_socketmgr_t **managerp);
959
960isc_result_t
961isc_socketmgr_create2(isc_mem_t *mctx, isc_socketmgr_t **managerp,
962		      unsigned int maxsocks);
963/*%<
964 * Create a socket manager.  If "maxsocks" is non-zero, it specifies the
965 * maximum number of sockets that the created manager should handle.
966 * isc_socketmgr_create() is equivalent of isc_socketmgr_create2() with
967 * "maxsocks" being zero.
968 * isc_socketmgr_createinctx() also associates the new manager with the
969 * specified application context.
970 *
971 * Notes:
972 *
973 *\li	All memory will be allocated in memory context 'mctx'.
974 *
975 * Requires:
976 *
977 *\li	'mctx' is a valid memory context.
978 *
979 *\li	'managerp' points to a NULL isc_socketmgr_t.
980 *
981 *\li	'actx' is a valid application context (for createinctx()).
982 *
983 * Ensures:
984 *
985 *\li	'*managerp' is a valid isc_socketmgr_t.
986 *
987 * Returns:
988 *
989 *\li	#ISC_R_SUCCESS
990 *\li	#ISC_R_NOMEMORY
991 *\li	#ISC_R_UNEXPECTED
992 *\li	#ISC_R_NOTIMPLEMENTED
993 */
994
995isc_result_t
996isc_socketmgr_getmaxsockets(isc_socketmgr_t *manager, unsigned int *nsockp);
997/*%<
998 * Returns in "*nsockp" the maximum number of sockets this manager may open.
999 *
1000 * Requires:
1001 *
1002 *\li	'*manager' is a valid isc_socketmgr_t.
1003 *\li	'nsockp' is not NULL.
1004 *
1005 * Returns:
1006 *
1007 *\li	#ISC_R_SUCCESS
1008 *\li	#ISC_R_NOTIMPLEMENTED
1009 */
1010
1011void
1012isc_socketmgr_setstats(isc_socketmgr_t *manager, isc_stats_t *stats);
1013/*%<
1014 * Set a general socket statistics counter set 'stats' for 'manager'.
1015 *
1016 * Requires:
1017 * \li	'manager' is valid, hasn't opened any socket, and doesn't have
1018 *	stats already set.
1019 *
1020 *\li	stats is a valid statistics supporting socket statistics counters
1021 *	(see above).
1022 */
1023
1024void
1025isc_socketmgr_destroy(isc_socketmgr_t **managerp);
1026/*%<
1027 * Destroy a socket manager.
1028 *
1029 * Notes:
1030 *
1031 *\li	This routine blocks until there are no sockets left in the manager,
1032 *	so if the caller holds any socket references using the manager, it
1033 *	must detach them before calling isc_socketmgr_destroy() or it will
1034 *	block forever.
1035 *
1036 * Requires:
1037 *
1038 *\li	'*managerp' is a valid isc_socketmgr_t.
1039 *
1040 *\li	All sockets managed by this manager are fully detached.
1041 *
1042 * Ensures:
1043 *
1044 *\li	*managerp == NULL
1045 *
1046 *\li	All resources used by the manager have been freed.
1047 */
1048
1049isc_sockettype_t
1050isc_socket_gettype(isc_socket_t *sock);
1051/*%<
1052 * Returns the socket type for "sock."
1053 *
1054 * Requires:
1055 *
1056 *\li	"sock" is a valid socket.
1057 */
1058
1059/*@{*/
1060isc_boolean_t
1061isc_socket_isbound(isc_socket_t *sock);
1062
1063void
1064isc_socket_ipv6only(isc_socket_t *sock, isc_boolean_t yes);
1065/*%<
1066 * If the socket is an IPv6 socket set/clear the IPV6_IPV6ONLY socket
1067 * option if the host OS supports this option.
1068 *
1069 * Requires:
1070 *\li	'sock' is a valid socket.
1071 */
1072/*@}*/
1073
1074void
1075isc_socket_cleanunix(isc_sockaddr_t *addr, isc_boolean_t active);
1076
1077/*%<
1078 * Cleanup UNIX domain sockets in the file-system.  If 'active' is true
1079 * then just unlink the socket.  If 'active' is false try to determine
1080 * if there is a listener of the socket or not.  If no listener is found
1081 * then unlink socket.
1082 *
1083 * Prior to unlinking the path is tested to see if it a socket.
1084 *
1085 * Note: there are a number of race conditions which cannot be avoided
1086 *       both in the filesystem and any application using UNIX domain
1087 *	 sockets (e.g. socket is tested between bind() and listen(),
1088 *	 the socket is deleted and replaced in the file-system between
1089 *	 stat() and unlink()).
1090 */
1091
1092isc_result_t
1093isc_socket_permunix(isc_sockaddr_t *sockaddr, isc_uint32_t perm,
1094		    isc_uint32_t owner, isc_uint32_t group);
1095/*%<
1096 * Set ownership and file permissions on the UNIX domain socket.
1097 *
1098 * Note: On Solaris and SunOS this secures the directory containing
1099 *       the socket as Solaris and SunOS do not honour the filesystem
1100 *	 permissions on the socket.
1101 *
1102 * Requires:
1103 * \li	'sockaddr' to be a valid UNIX domain sockaddr.
1104 *
1105 * Returns:
1106 * \li	#ISC_R_SUCCESS
1107 * \li	#ISC_R_FAILURE
1108 */
1109
1110void isc_socket_setname(isc_socket_t *socket, const char *name, void *tag);
1111/*%<
1112 * Set the name and optional tag for a socket.  This allows tracking of the
1113 * owner or purpose for this socket, and is useful for tracing and statistics
1114 * reporting.
1115 */
1116
1117const char *isc_socket_getname(isc_socket_t *socket);
1118/*%<
1119 * Get the name associated with a socket, if any.
1120 */
1121
1122void *isc_socket_gettag(isc_socket_t *socket);
1123/*%<
1124 * Get the tag associated with a socket, if any.
1125 */
1126
1127int isc_socket_getfd(isc_socket_t *socket);
1128/*%<
1129 * Get the file descriptor associated with a socket
1130 */
1131
1132void
1133isc__socketmgr_setreserved(isc_socketmgr_t *mgr, isc_uint32_t);
1134/*%<
1135 * Temporary.  For use by named only.
1136 */
1137
1138void
1139isc__socketmgr_maxudp(isc_socketmgr_t *mgr, int maxudp);
1140/*%<
1141 * Test interface. Drop UDP packet > 'maxudp'.
1142 */
1143
1144#ifdef HAVE_LIBXML2
1145
1146int
1147isc_socketmgr_renderxml(isc_socketmgr_t *mgr, xmlTextWriterPtr writer);
1148/*%<
1149 * Render internal statistics and other state into the XML document.
1150 */
1151
1152#endif /* HAVE_LIBXML2 */
1153
1154#ifdef USE_SOCKETIMPREGISTER
1155/*%<
1156 * See isc_socketmgr_create() above.
1157 */
1158typedef isc_result_t
1159(*isc_socketmgrcreatefunc_t)(isc_mem_t *mctx, isc_socketmgr_t **managerp);
1160
1161isc_result_t
1162isc_socket_register(isc_socketmgrcreatefunc_t createfunc);
1163/*%<
1164 * Register a new socket I/O implementation and add it to the list of
1165 * supported implementations.  This function must be called when a different
1166 * event library is used than the one contained in the ISC library.
1167 */
1168
1169isc_result_t
1170isc__socket_register(void);
1171/*%<
1172 * A short cut function that specifies the socket I/O module in the ISC
1173 * library for isc_socket_register().  An application that uses the ISC library
1174 * usually do not have to care about this function: it would call
1175 * isc_lib_register(), which internally calls this function.
1176 */
1177#endif /* USE_SOCKETIMPREGISTER */
1178
1179ISC_LANG_ENDDECLS
1180
1181#endif /* ISC_SOCKET_H */
1182