1 /*
2  * services/listen_dnsport.h - listen on port 53 for incoming DNS queries.
3  *
4  * Copyright (c) 2007, NLnet Labs. All rights reserved.
5  *
6  * This software is open source.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  *
12  * Redistributions of source code must retain the above copyright notice,
13  * this list of conditions and the following disclaimer.
14  *
15  * Redistributions in binary form must reproduce the above copyright notice,
16  * this list of conditions and the following disclaimer in the documentation
17  * and/or other materials provided with the distribution.
18  *
19  * Neither the name of the NLNET LABS nor the names of its contributors may
20  * be used to endorse or promote products derived from this software without
21  * specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
24  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
25  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
26  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
27  * HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
28  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
29  * TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
30  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34  */
35 
36 /**
37  * \file
38  *
39  * This file has functions to get queries from clients.
40  */
41 
42 #ifndef LISTEN_DNSPORT_H
43 #define LISTEN_DNSPORT_H
44 
45 #include "util/netevent.h"
46 #include "daemon/acl_list.h"
47 #ifdef HAVE_NGHTTP2_NGHTTP2_H
48 #include <nghttp2/nghttp2.h>
49 #endif
50 struct listen_list;
51 struct config_file;
52 struct addrinfo;
53 struct sldns_buffer;
54 struct tcl_list;
55 
56 /**
57  * Listening for queries structure.
58  * Contains list of query-listen sockets.
59  */
60 struct listen_dnsport {
61 	/** Base for select calls */
62 	struct comm_base* base;
63 
64 	/** buffer shared by UDP connections, since there is only one
65 	    datagram at any time. */
66 	struct sldns_buffer* udp_buff;
67 #ifdef USE_DNSCRYPT
68 	struct sldns_buffer* dnscrypt_udp_buff;
69 #endif
70 	/** list of comm points used to get incoming events */
71 	struct listen_list* cps;
72 };
73 
74 /**
75  * Single linked list to store event points.
76  */
77 struct listen_list {
78 	/** next in list */
79 	struct listen_list* next;
80 	/** event info */
81 	struct comm_point* com;
82 };
83 
84 /**
85  * type of ports
86  */
87 enum listen_type {
88 	/** udp type */
89 	listen_type_udp,
90 	/** tcp type */
91 	listen_type_tcp,
92 	/** udp ipv6 (v4mapped) for use with ancillary data */
93 	listen_type_udpancil,
94 	/** ssl over tcp type */
95 	listen_type_ssl,
96 	/** udp type  + dnscrypt*/
97 	listen_type_udp_dnscrypt,
98 	/** tcp type + dnscrypt */
99 	listen_type_tcp_dnscrypt,
100 	/** udp ipv6 (v4mapped) for use with ancillary data + dnscrypt*/
101 	listen_type_udpancil_dnscrypt,
102 	/** HTTP(2) over TLS over TCP */
103 	listen_type_http
104 };
105 
106 /*
107  * socket properties (just like NSD nsd_socket structure definition)
108  */
109 struct unbound_socket {
110 	/** socket-address structure */
111 	struct addrinfo* addr;
112 	/** socket descriptor returned by socket() syscall */
113 	int s;
114 	/** address family (AF_INET/IF_INET6) */
115 	int fam;
116 	/** ACL on the socket (listening interface) */
117 	struct acl_addr* acl;
118 };
119 
120 /**
121  * Single linked list to store shared ports that have been
122  * opened for use by all threads.
123  */
124 struct listen_port {
125 	/** next in list */
126 	struct listen_port* next;
127 	/** file descriptor, open and ready for use */
128 	int fd;
129 	/** type of file descriptor, udp or tcp */
130 	enum listen_type ftype;
131 	/** if the port should support PROXYv2 */
132 	int pp2_enabled;
133 	/** fill in unbound_socket structure for every opened socket at
134 	 * Unbound startup */
135 	struct unbound_socket* socket;
136 };
137 
138 /**
139  * Create shared listening ports
140  * Getaddrinfo, create socket, bind and listen to zero or more
141  * interfaces for IP4 and/or IP6, for UDP and/or TCP.
142  * On the given port number. It creates the sockets.
143  * @param cfg: settings on what ports to open.
144  * @param ifs: interfaces to open, array of IP addresses, "ip[@port]".
145  * @param num_ifs: length of ifs.
146  * @param reuseport: set to true if you want reuseport, or NULL to not have it,
147  *   set to false on exit if reuseport failed to apply (because of no
148  *   kernel support).
149  * @return: linked list of ports or NULL on error.
150  */
151 struct listen_port* listening_ports_open(struct config_file* cfg,
152 	char** ifs, int num_ifs, int* reuseport);
153 
154 /**
155  * Close and delete the (list of) listening ports.
156  */
157 void listening_ports_free(struct listen_port* list);
158 
159 struct config_strlist;
160 /**
161  * Resolve interface names in config and store result IP addresses
162  * @param ifs: array of interfaces.  The list of interface names, if not NULL.
163  * @param num_ifs: length of ifs array.
164  * @param list: if not NULL, this is used as the list of interface names.
165  * @param resif: string array (malloced array of malloced strings) with
166  * 	result.  NULL if cfg has none.
167  * @param num_resif: length of resif.  Zero if cfg has zero num_ifs.
168  * @return 0 on failure.
169  */
170 int resolve_interface_names(char** ifs, int num_ifs,
171 	struct config_strlist* list, char*** resif, int* num_resif);
172 
173 /**
174  * Create commpoints with for this thread for the shared ports.
175  * @param base: the comm_base that provides event functionality.
176  *	for default all ifs.
177  * @param ports: the list of shared ports.
178  * @param bufsize: size of datagram buffer.
179  * @param tcp_accept_count: max number of simultaneous TCP connections
180  * 	from clients.
181  * @param tcp_idle_timeout: idle timeout for TCP connections in msec.
182  * @param harden_large_queries: whether query size should be limited.
183  * @param http_max_streams: maximum number of HTTP/2 streams per connection.
184  * @param http_endpoint: HTTP endpoint to service queries on
185  * @param http_notls: no TLS for http downstream
186  * @param tcp_conn_limit: TCP connection limit info.
187  * @param sslctx: nonNULL if ssl context.
188  * @param dtenv: nonNULL if dnstap enabled.
189  * @param cb: callback function when a request arrives. It is passed
190  *	  the packet and user argument. Return true to send a reply.
191  * @param cb_arg: user data argument for callback function.
192  * @return: the malloced listening structure, ready for use. NULL on error.
193  */
194 struct listen_dnsport*
195 listen_create(struct comm_base* base, struct listen_port* ports,
196 	size_t bufsize, int tcp_accept_count, int tcp_idle_timeout,
197 	int harden_large_queries, uint32_t http_max_streams,
198 	char* http_endpoint, int http_notls, struct tcl_list* tcp_conn_limit,
199 	void* sslctx, struct dt_env* dtenv, comm_point_callback_type* cb,
200 	void *cb_arg);
201 
202 /**
203  * delete the listening structure
204  * @param listen: listening structure.
205  */
206 void listen_delete(struct listen_dnsport* listen);
207 
208 /** setup the locks for the listen ports */
209 void listen_setup_locks(void);
210 /** desetup the locks for the listen ports */
211 void listen_desetup_locks(void);
212 
213 /**
214  * delete listen_list of commpoints. Calls commpointdelete() on items.
215  * This may close the fds or not depending on flags.
216  * @param list: to delete.
217  */
218 void listen_list_delete(struct listen_list* list);
219 
220 /**
221  * get memory size used by the listening structs
222  * @param listen: listening structure.
223  * @return: size in bytes.
224  */
225 size_t listen_get_mem(struct listen_dnsport* listen);
226 
227 /**
228  * stop accept handlers for TCP (until enabled again)
229  * @param listen: listening structure.
230  */
231 void listen_stop_accept(struct listen_dnsport* listen);
232 
233 /**
234  * start accept handlers for TCP (was stopped before)
235  * @param listen: listening structure.
236  */
237 void listen_start_accept(struct listen_dnsport* listen);
238 
239 /**
240  * Create and bind nonblocking UDP socket
241  * @param family: for socket call.
242  * @param socktype: for socket call.
243  * @param addr: for bind call.
244  * @param addrlen: for bind call.
245  * @param v6only: if enabled, IP6 sockets get IP6ONLY option set.
246  * 	if enabled with value 2 IP6ONLY option is disabled.
247  * @param inuse: on error, this is set true if the port was in use.
248  * @param noproto: on error, this is set true if cause is that the
249 	IPv6 proto (family) is not available.
250  * @param rcv: set size on rcvbuf with socket option, if 0 it is not set.
251  * @param snd: set size on sndbuf with socket option, if 0 it is not set.
252  * @param listen: if true, this is a listening UDP port, eg port 53, and
253  * 	set SO_REUSEADDR on it.
254  * @param reuseport: if nonNULL and true, try to set SO_REUSEPORT on
255  * 	listening UDP port.  Set to false on return if it failed to do so.
256  * @param transparent: set IP_TRANSPARENT socket option.
257  * @param freebind: set IP_FREEBIND socket option.
258  * @param use_systemd: if true, fetch sockets from systemd.
259  * @param dscp: DSCP to use.
260  * @return: the socket. -1 on error.
261  */
262 int create_udp_sock(int family, int socktype, struct sockaddr* addr,
263 	socklen_t addrlen, int v6only, int* inuse, int* noproto, int rcv,
264 	int snd, int listen, int* reuseport, int transparent, int freebind, int use_systemd, int dscp);
265 
266 /**
267  * Create and bind TCP listening socket
268  * @param addr: address info ready to make socket.
269  * @param v6only: enable ip6 only flag on ip6 sockets.
270  * @param noproto: if error caused by lack of protocol support.
271  * @param reuseport: if nonNULL and true, try to set SO_REUSEPORT on
272  * 	listening UDP port.  Set to false on return if it failed to do so.
273  * @param transparent: set IP_TRANSPARENT socket option.
274  * @param mss: maximum segment size of the socket. if zero, leaves the default.
275  * @param nodelay: if true set TCP_NODELAY and TCP_QUICKACK socket options.
276  * @param freebind: set IP_FREEBIND socket option.
277  * @param use_systemd: if true, fetch sockets from systemd.
278  * @param dscp: DSCP to use.
279  * @return: the socket. -1 on error.
280  */
281 int create_tcp_accept_sock(struct addrinfo *addr, int v6only, int* noproto,
282 	int* reuseport, int transparent, int mss, int nodelay, int freebind,
283 	int use_systemd, int dscp);
284 
285 /**
286  * Create and bind local listening socket
287  * @param path: path to the socket.
288  * @param noproto: on error, this is set true if cause is that local sockets
289  *	are not supported.
290  * @param use_systemd: if true, fetch sockets from systemd.
291  * @return: the socket. -1 on error.
292  */
293 int create_local_accept_sock(const char* path, int* noproto, int use_systemd);
294 
295 /**
296  * TCP request info.  List of requests outstanding on the channel, that
297  * are asked for but not yet answered back.
298  */
299 struct tcp_req_info {
300 	/** the TCP comm point for this.  Its buffer is used for read/write */
301 	struct comm_point* cp;
302 	/** the buffer to use to spool reply from mesh into,
303 	 * it can then be copied to the result list and written.
304 	 * it is a pointer to the shared udp buffer. */
305 	struct sldns_buffer* spool_buffer;
306 	/** are we in worker_handle function call (for recursion callback)*/
307 	int in_worker_handle;
308 	/** is the comm point dropped (by worker handle).
309 	 * That means we have to disconnect the channel. */
310 	int is_drop;
311 	/** is the comm point set to send_reply (by mesh new client in worker
312 	 * handle), if so answer is available in c.buffer */
313 	int is_reply;
314 	/** read channel has closed, just write pending results */
315 	int read_is_closed;
316 	/** read again */
317 	int read_again;
318 	/** number of outstanding requests */
319 	int num_open_req;
320 	/** list of outstanding requests */
321 	struct tcp_req_open_item* open_req_list;
322 	/** number of pending writeable results */
323 	int num_done_req;
324 	/** list of pending writable result packets, malloced one at a time */
325 	struct tcp_req_done_item* done_req_list;
326 };
327 
328 /**
329  * List of open items in TCP channel
330  */
331 struct tcp_req_open_item {
332 	/** next in list */
333 	struct tcp_req_open_item* next;
334 	/** the mesh area of the mesh_state */
335 	struct mesh_area* mesh;
336 	/** the mesh state */
337 	struct mesh_state* mesh_state;
338 };
339 
340 /**
341  * List of done items in TCP channel
342  */
343 struct tcp_req_done_item {
344 	/** next in list */
345 	struct tcp_req_done_item* next;
346 	/** the buffer with packet contents */
347 	uint8_t* buf;
348 	/** length of the buffer */
349 	size_t len;
350 };
351 
352 /**
353  * Create tcp request info structure that keeps track of open
354  * requests on the TCP channel that are resolved at the same time,
355  * and the pending results that have to get written back to that client.
356  * @param spoolbuf: shared buffer
357  * @return new structure or NULL on alloc failure.
358  */
359 struct tcp_req_info* tcp_req_info_create(struct sldns_buffer* spoolbuf);
360 
361 /**
362  * Delete tcp request structure.  Called by owning commpoint.
363  * Removes mesh entry references and stored results from the lists.
364  * @param req: the tcp request info
365  */
366 void tcp_req_info_delete(struct tcp_req_info* req);
367 
368 /**
369  * Clear tcp request structure.  Removes list entries, sets it up ready
370  * for the next connection.
371  * @param req: tcp request info structure.
372  */
373 void tcp_req_info_clear(struct tcp_req_info* req);
374 
375 /**
376  * Remove mesh state entry from list in tcp_req_info.
377  * caller has to manage the mesh state reply entry in the mesh state.
378  * @param req: the tcp req info that has the entry removed from the list.
379  * @param m: the state removed from the list.
380  */
381 void tcp_req_info_remove_mesh_state(struct tcp_req_info* req,
382 	struct mesh_state* m);
383 
384 /**
385  * Handle write done of the last result packet
386  * @param req: the tcp req info.
387  */
388 void tcp_req_info_handle_writedone(struct tcp_req_info* req);
389 
390 /**
391  * Handle read done of a new request from the client
392  * @param req: the tcp req info.
393  */
394 void tcp_req_info_handle_readdone(struct tcp_req_info* req);
395 
396 /**
397  * Add mesh state to the tcp req list of open requests.
398  * So the comm_reply can be removed off the mesh reply list when
399  * the tcp channel has to be closed (for other reasons then that that
400  * request was done, eg. channel closed by client or some format error).
401  * @param req: tcp req info structure.  It keeps track of the simultaneous
402  * 	requests and results on a tcp (or TLS) channel.
403  * @param mesh: mesh area for the state.
404  * @param m: mesh state to add.
405  * @return 0 on failure (malloc failure).
406  */
407 int tcp_req_info_add_meshstate(struct tcp_req_info* req,
408 	struct mesh_area* mesh, struct mesh_state* m);
409 
410 /**
411  * Send reply on tcp simultaneous answer channel.  May queue it up.
412  * @param req: request info structure.
413  */
414 void tcp_req_info_send_reply(struct tcp_req_info* req);
415 
416 /** the read channel has closed
417  * @param req: request. remaining queries are looked up and answered.
418  * @return zero if nothing to do, just close the tcp.
419  */
420 int tcp_req_info_handle_read_close(struct tcp_req_info* req);
421 
422 /** get the size of currently used tcp stream wait buffers (in bytes) */
423 size_t tcp_req_info_get_stream_buffer_size(void);
424 
425 /** get the size of currently used HTTP2 query buffers (in bytes) */
426 size_t http2_get_query_buffer_size(void);
427 /** get the size of currently used HTTP2 response buffers (in bytes) */
428 size_t http2_get_response_buffer_size(void);
429 
430 #ifdef HAVE_NGHTTP2
431 /**
432  * Create nghttp2 callbacks to handle HTTP2 requests.
433  * @return malloc'ed struct, NULL on failure
434  */
435 nghttp2_session_callbacks* http2_req_callbacks_create(void);
436 
437 /** Free http2 stream buffers and decrease buffer counters */
438 void http2_req_stream_clear(struct http2_stream* h2_stream);
439 
440 /**
441  * DNS response ready to be submitted to nghttp2, to be prepared for sending
442  * out. Response is stored in c->buffer. Copy to rbuffer because the c->buffer
443  * might be used before this will be send out.
444  * @param h2_session: http2 session, containing c->buffer which contains answer
445  * @param h2_stream: http2 stream, containing buffer to store answer in
446  * @return 0 on error, 1 otherwise
447  */
448 int http2_submit_dns_response(struct http2_session* h2_session);
449 #else
450 int http2_submit_dns_response(void* v);
451 #endif /* HAVE_NGHTTP2 */
452 
453 char* set_ip_dscp(int socket, int addrfamily, int ds);
454 
455 /** for debug and profiling purposes only
456  * @param ub_sock: the structure containing created socket info we want to print or log for
457  */
458 void verbose_print_unbound_socket(struct unbound_socket* ub_sock);
459 
460 #endif /* LISTEN_DNSPORT_H */
461