]>
Commit | Line | Data |
---|---|---|
b5d9ee5c MM |
1 | /* |
2 | * BIRD Internet Routing Daemon -- Unix I/O | |
3 | * | |
38a608c5 | 4 | * (c) 1998--2004 Martin Mares <mj@ucw.cz> |
b1a1faba | 5 | * (c) 2004 Ondrej Filip <feela@network.cz> |
b5d9ee5c MM |
6 | * |
7 | * Can be freely distributed and used under the terms of the GNU GPL. | |
8 | */ | |
9 | ||
607d9914 OZ |
10 | /* Unfortunately, some glibc versions hide parts of RFC 3542 API |
11 | if _GNU_SOURCE is not defined. */ | |
12 | #define _GNU_SOURCE 1 | |
13 | ||
b5d9ee5c MM |
14 | #include <stdio.h> |
15 | #include <stdlib.h> | |
01b776e1 | 16 | #include <time.h> |
b5d9ee5c MM |
17 | #include <sys/time.h> |
18 | #include <sys/types.h> | |
19 | #include <sys/socket.h> | |
46a82e9c | 20 | #include <sys/uio.h> |
b93abffa | 21 | #include <sys/un.h> |
b5d9ee5c | 22 | #include <unistd.h> |
a0b176e3 | 23 | #include <fcntl.h> |
b5d9ee5c | 24 | #include <errno.h> |
05476c4d | 25 | #include <net/if.h> |
d0e9b36d | 26 | #include <netinet/in.h> |
48e5f32d OZ |
27 | #include <netinet/tcp.h> |
28 | #include <netinet/udp.h> | |
93e868c7 | 29 | #include <netinet/icmp6.h> |
b5d9ee5c MM |
30 | |
31 | #include "nest/bird.h" | |
32 | #include "lib/lists.h" | |
33 | #include "lib/resource.h" | |
34 | #include "lib/timer.h" | |
35 | #include "lib/socket.h" | |
e8f73195 | 36 | #include "lib/event.h" |
afa8937a | 37 | #include "lib/string.h" |
b5d9ee5c MM |
38 | #include "nest/iface.h" |
39 | ||
40 | #include "lib/unix.h" | |
a2867cd9 | 41 | #include "lib/sysio.h" |
b5d9ee5c | 42 | |
ea89da38 | 43 | /* Maximum number of calls of tx handler for one socket in one |
4323099d OZ |
44 | * select iteration. Should be small enough to not monopolize CPU by |
45 | * one protocol instance. | |
46 | */ | |
47 | #define MAX_STEPS 4 | |
48 | ||
ea89da38 OZ |
49 | /* Maximum number of calls of rx handler for all sockets in one select |
50 | iteration. RX callbacks are often much more costly so we limit | |
51 | this to gen small latencies */ | |
52 | #define MAX_RX_STEPS 4 | |
53 | ||
a9c986f9 MM |
54 | /* |
55 | * Tracked Files | |
56 | */ | |
57 | ||
58 | struct rfile { | |
59 | resource r; | |
60 | FILE *f; | |
61 | }; | |
62 | ||
63 | static void | |
64 | rf_free(resource *r) | |
65 | { | |
66 | struct rfile *a = (struct rfile *) r; | |
67 | ||
68 | fclose(a->f); | |
69 | } | |
70 | ||
71 | static void | |
72 | rf_dump(resource *r) | |
73 | { | |
74 | struct rfile *a = (struct rfile *) r; | |
75 | ||
76 | debug("(FILE *%p)\n", a->f); | |
77 | } | |
78 | ||
79 | static struct resclass rf_class = { | |
80 | "FILE", | |
81 | sizeof(struct rfile), | |
82 | rf_free, | |
e81b440f | 83 | rf_dump, |
acb60628 | 84 | NULL, |
e81b440f | 85 | NULL |
a9c986f9 MM |
86 | }; |
87 | ||
88 | void * | |
f78056fb | 89 | tracked_fopen(pool *p, char *name, char *mode) |
a9c986f9 MM |
90 | { |
91 | FILE *f = fopen(name, mode); | |
92 | ||
93 | if (f) | |
94 | { | |
95 | struct rfile *r = ralloc(p, &rf_class); | |
96 | r->f = f; | |
97 | } | |
98 | return f; | |
99 | } | |
100 | ||
525fa2c1 MM |
101 | /** |
102 | * DOC: Timers | |
103 | * | |
104 | * Timers are resources which represent a wish of a module to call | |
105 | * a function at the specified time. The platform dependent code | |
58f7d004 | 106 | * doesn't guarantee exact timing, only that a timer function |
525fa2c1 MM |
107 | * won't be called before the requested time. |
108 | * | |
fd91ae33 OZ |
109 | * In BIRD, time is represented by values of the &bird_clock_t type |
110 | * which are integral numbers interpreted as a relative number of seconds since | |
111 | * some fixed time point in past. The current time can be read | |
112 | * from variable @now with reasonable accuracy and is monotonic. There is also | |
113 | * a current 'absolute' time in variable @now_real reported by OS. | |
525fa2c1 MM |
114 | * |
115 | * Each timer is described by a &timer structure containing a pointer | |
116 | * to the handler function (@hook), data private to this function (@data), | |
117 | * time the function should be called at (@expires, 0 for inactive timers), | |
118 | * for the other fields see |timer.h|. | |
b5d9ee5c MM |
119 | */ |
120 | ||
121 | #define NEAR_TIMER_LIMIT 4 | |
122 | ||
b5d9ee5c MM |
123 | static list near_timers, far_timers; |
124 | static bird_clock_t first_far_timer = TIME_INFINITY; | |
125 | ||
002b6423 | 126 | /* now must be different from 0, because 0 is a special value in timer->expires */ |
a92cf57d | 127 | bird_clock_t now = 1, now_real, boot_time; |
fd91ae33 OZ |
128 | |
129 | static void | |
130 | update_times_plain(void) | |
131 | { | |
132 | bird_clock_t new_time = time(NULL); | |
133 | int delta = new_time - now_real; | |
134 | ||
135 | if ((delta >= 0) && (delta < 60)) | |
136 | now += delta; | |
137 | else if (now_real != 0) | |
138 | log(L_WARN "Time jump, delta %d s", delta); | |
139 | ||
140 | now_real = new_time; | |
141 | } | |
142 | ||
143 | static void | |
144 | update_times_gettime(void) | |
145 | { | |
146 | struct timespec ts; | |
147 | int rv; | |
148 | ||
149 | rv = clock_gettime(CLOCK_MONOTONIC, &ts); | |
150 | if (rv != 0) | |
151 | die("clock_gettime: %m"); | |
152 | ||
153 | if (ts.tv_sec != now) { | |
154 | if (ts.tv_sec < now) | |
155 | log(L_ERR "Monotonic timer is broken"); | |
156 | ||
157 | now = ts.tv_sec; | |
158 | now_real = time(NULL); | |
159 | } | |
160 | } | |
161 | ||
162 | static int clock_monotonic_available; | |
163 | ||
164 | static inline void | |
165 | update_times(void) | |
166 | { | |
167 | if (clock_monotonic_available) | |
168 | update_times_gettime(); | |
169 | else | |
170 | update_times_plain(); | |
171 | } | |
172 | ||
173 | static inline void | |
174 | init_times(void) | |
175 | { | |
176 | struct timespec ts; | |
177 | clock_monotonic_available = (clock_gettime(CLOCK_MONOTONIC, &ts) == 0); | |
178 | if (!clock_monotonic_available) | |
179 | log(L_WARN "Monotonic timer is missing"); | |
180 | } | |
181 | ||
b5d9ee5c MM |
182 | |
183 | static void | |
184 | tm_free(resource *r) | |
185 | { | |
186 | timer *t = (timer *) r; | |
187 | ||
188 | tm_stop(t); | |
189 | } | |
190 | ||
191 | static void | |
192 | tm_dump(resource *r) | |
193 | { | |
194 | timer *t = (timer *) r; | |
195 | ||
e8f73195 | 196 | debug("(code %p, data %p, ", t->hook, t->data); |
af847acc MM |
197 | if (t->randomize) |
198 | debug("rand %d, ", t->randomize); | |
199 | if (t->recurrent) | |
200 | debug("recur %d, ", t->recurrent); | |
b5d9ee5c MM |
201 | if (t->expires) |
202 | debug("expires in %d sec)\n", t->expires - now); | |
203 | else | |
204 | debug("inactive)\n"); | |
205 | } | |
206 | ||
207 | static struct resclass tm_class = { | |
208 | "Timer", | |
209 | sizeof(timer), | |
210 | tm_free, | |
e81b440f | 211 | tm_dump, |
acb60628 | 212 | NULL, |
e81b440f | 213 | NULL |
b5d9ee5c MM |
214 | }; |
215 | ||
525fa2c1 MM |
216 | /** |
217 | * tm_new - create a timer | |
218 | * @p: pool | |
219 | * | |
220 | * This function creates a new timer resource and returns | |
221 | * a pointer to it. To use the timer, you need to fill in | |
222 | * the structure fields and call tm_start() to start timing. | |
223 | */ | |
b5d9ee5c MM |
224 | timer * |
225 | tm_new(pool *p) | |
226 | { | |
227 | timer *t = ralloc(p, &tm_class); | |
b5d9ee5c MM |
228 | return t; |
229 | } | |
230 | ||
231 | static inline void | |
232 | tm_insert_near(timer *t) | |
233 | { | |
234 | node *n = HEAD(near_timers); | |
235 | ||
236 | while (n->next && (SKIP_BACK(timer, n, n)->expires < t->expires)) | |
237 | n = n->next; | |
238 | insert_node(&t->n, n->prev); | |
239 | } | |
240 | ||
525fa2c1 MM |
241 | /** |
242 | * tm_start - start a timer | |
243 | * @t: timer | |
244 | * @after: number of seconds the timer should be run after | |
245 | * | |
246 | * This function schedules the hook function of the timer to | |
247 | * be called after @after seconds. If the timer has been already | |
248 | * started, it's @expire time is replaced by the new value. | |
249 | * | |
250 | * You can have set the @randomize field of @t, the timeout | |
251 | * will be increased by a random number of seconds chosen | |
252 | * uniformly from range 0 .. @randomize. | |
253 | * | |
254 | * You can call tm_start() from the handler function of the timer | |
255 | * to request another run of the timer. Also, you can set the @recurrent | |
256 | * field to have the timer re-added automatically with the same timeout. | |
257 | */ | |
b5d9ee5c MM |
258 | void |
259 | tm_start(timer *t, unsigned after) | |
260 | { | |
261 | bird_clock_t when; | |
262 | ||
263 | if (t->randomize) | |
af847acc | 264 | after += random() % (t->randomize + 1); |
b5d9ee5c MM |
265 | when = now + after; |
266 | if (t->expires == when) | |
267 | return; | |
268 | if (t->expires) | |
269 | rem_node(&t->n); | |
270 | t->expires = when; | |
271 | if (after <= NEAR_TIMER_LIMIT) | |
272 | tm_insert_near(t); | |
273 | else | |
274 | { | |
275 | if (!first_far_timer || first_far_timer > when) | |
276 | first_far_timer = when; | |
277 | add_tail(&far_timers, &t->n); | |
278 | } | |
279 | } | |
280 | ||
525fa2c1 MM |
281 | /** |
282 | * tm_stop - stop a timer | |
283 | * @t: timer | |
284 | * | |
285 | * This function stops a timer. If the timer is already stopped, | |
286 | * nothing happens. | |
287 | */ | |
b5d9ee5c MM |
288 | void |
289 | tm_stop(timer *t) | |
290 | { | |
291 | if (t->expires) | |
292 | { | |
293 | rem_node(&t->n); | |
294 | t->expires = 0; | |
295 | } | |
296 | } | |
297 | ||
298 | static void | |
299 | tm_dump_them(char *name, list *l) | |
300 | { | |
301 | node *n; | |
302 | timer *t; | |
303 | ||
304 | debug("%s timers:\n", name); | |
305 | WALK_LIST(n, *l) | |
306 | { | |
307 | t = SKIP_BACK(timer, n, n); | |
308 | debug("%p ", t); | |
309 | tm_dump(&t->r); | |
310 | } | |
311 | debug("\n"); | |
312 | } | |
313 | ||
314 | void | |
315 | tm_dump_all(void) | |
316 | { | |
317 | tm_dump_them("Near", &near_timers); | |
318 | tm_dump_them("Far", &far_timers); | |
319 | } | |
320 | ||
321 | static inline time_t | |
322 | tm_first_shot(void) | |
323 | { | |
324 | time_t x = first_far_timer; | |
325 | ||
326 | if (!EMPTY_LIST(near_timers)) | |
327 | { | |
328 | timer *t = SKIP_BACK(timer, n, HEAD(near_timers)); | |
329 | if (t->expires < x) | |
330 | x = t->expires; | |
331 | } | |
332 | return x; | |
333 | } | |
334 | ||
8bcb5fb1 OZ |
335 | void io_log_event(void *hook, void *data); |
336 | ||
b5d9ee5c MM |
337 | static void |
338 | tm_shot(void) | |
339 | { | |
340 | timer *t; | |
341 | node *n, *m; | |
342 | ||
343 | if (first_far_timer <= now) | |
344 | { | |
28a9a189 | 345 | bird_clock_t limit = now + NEAR_TIMER_LIMIT; |
b5d9ee5c MM |
346 | first_far_timer = TIME_INFINITY; |
347 | n = HEAD(far_timers); | |
348 | while (m = n->next) | |
349 | { | |
350 | t = SKIP_BACK(timer, n, n); | |
351 | if (t->expires <= limit) | |
352 | { | |
353 | rem_node(n); | |
354 | tm_insert_near(t); | |
355 | } | |
356 | else if (t->expires < first_far_timer) | |
357 | first_far_timer = t->expires; | |
358 | n = m; | |
359 | } | |
360 | } | |
361 | while ((n = HEAD(near_timers)) -> next) | |
362 | { | |
af847acc | 363 | int delay; |
b5d9ee5c MM |
364 | t = SKIP_BACK(timer, n, n); |
365 | if (t->expires > now) | |
366 | break; | |
367 | rem_node(n); | |
af847acc | 368 | delay = t->expires - now; |
b5d9ee5c | 369 | t->expires = 0; |
af847acc MM |
370 | if (t->recurrent) |
371 | { | |
372 | int i = t->recurrent - delay; | |
373 | if (i < 0) | |
374 | i = 0; | |
375 | tm_start(t, i); | |
376 | } | |
8bcb5fb1 | 377 | io_log_event(t->hook, t->data); |
b5d9ee5c MM |
378 | t->hook(t); |
379 | } | |
380 | } | |
381 | ||
0d3effcf OF |
382 | /** |
383 | * tm_parse_datetime - parse a date and time | |
384 | * @x: datetime string | |
385 | * | |
386 | * tm_parse_datetime() takes a textual representation of | |
387 | * a date and time (dd-mm-yyyy hh:mm:ss) | |
388 | * and converts it to the corresponding value of type &bird_clock_t. | |
389 | */ | |
390 | bird_clock_t | |
391 | tm_parse_datetime(char *x) | |
392 | { | |
393 | struct tm tm; | |
394 | int n; | |
395 | time_t t; | |
396 | ||
397 | if (sscanf(x, "%d-%d-%d %d:%d:%d%n", &tm.tm_mday, &tm.tm_mon, &tm.tm_year, &tm.tm_hour, &tm.tm_min, &tm.tm_sec, &n) != 6 || x[n]) | |
398 | return tm_parse_date(x); | |
399 | tm.tm_mon--; | |
400 | tm.tm_year -= 1900; | |
401 | t = mktime(&tm); | |
402 | if (t == (time_t) -1) | |
403 | return 0; | |
404 | return t; | |
405 | } | |
525fa2c1 MM |
406 | /** |
407 | * tm_parse_date - parse a date | |
408 | * @x: date string | |
409 | * | |
410 | * tm_parse_date() takes a textual representation of a date (dd-mm-yyyy) | |
411 | * and converts it to the corresponding value of type &bird_clock_t. | |
412 | */ | |
913f7dc9 MM |
413 | bird_clock_t |
414 | tm_parse_date(char *x) | |
415 | { | |
416 | struct tm tm; | |
417 | int n; | |
418 | time_t t; | |
419 | ||
420 | if (sscanf(x, "%d-%d-%d%n", &tm.tm_mday, &tm.tm_mon, &tm.tm_year, &n) != 3 || x[n]) | |
421 | return 0; | |
422 | tm.tm_mon--; | |
423 | tm.tm_year -= 1900; | |
424 | tm.tm_hour = tm.tm_min = tm.tm_sec = 0; | |
425 | t = mktime(&tm); | |
426 | if (t == (time_t) -1) | |
427 | return 0; | |
428 | return t; | |
429 | } | |
430 | ||
c37e7851 OZ |
431 | static void |
432 | tm_format_reltime(char *x, struct tm *tm, bird_clock_t delta) | |
913f7dc9 | 433 | { |
c37e7851 OZ |
434 | static char *month_names[12] = { "Jan", "Feb", "Mar", "Apr", "May", "Jun", |
435 | "Jul", "Aug", "Sep", "Oct", "Nov", "Dec" }; | |
913f7dc9 | 436 | |
c37e7851 OZ |
437 | if (delta < 20*3600) |
438 | bsprintf(x, "%02d:%02d", tm->tm_hour, tm->tm_min); | |
439 | else if (delta < 360*86400) | |
440 | bsprintf(x, "%s%02d", month_names[tm->tm_mon], tm->tm_mday); | |
441 | else | |
442 | bsprintf(x, "%d", tm->tm_year+1900); | |
913f7dc9 MM |
443 | } |
444 | ||
c37e7851 OZ |
445 | #include "conf/conf.h" |
446 | ||
525fa2c1 MM |
447 | /** |
448 | * tm_format_datetime - convert date and time to textual representation | |
449 | * @x: destination buffer of size %TM_DATETIME_BUFFER_SIZE | |
450 | * @t: time | |
451 | * | |
fd91ae33 OZ |
452 | * This function formats the given relative time value @t to a textual |
453 | * date/time representation (dd-mm-yyyy hh:mm:ss) in real time. | |
525fa2c1 | 454 | */ |
7a88832e | 455 | void |
c37e7851 | 456 | tm_format_datetime(char *x, struct timeformat *fmt_spec, bird_clock_t t) |
7a88832e | 457 | { |
c37e7851 | 458 | const char *fmt_used; |
7a88832e | 459 | struct tm *tm; |
fd91ae33 OZ |
460 | bird_clock_t delta = now - t; |
461 | t = now_real - delta; | |
7a88832e | 462 | tm = localtime(&t); |
7a88832e | 463 | |
c37e7851 OZ |
464 | if (fmt_spec->fmt1 == NULL) |
465 | return tm_format_reltime(x, tm, delta); | |
afa8937a | 466 | |
c37e7851 OZ |
467 | if ((fmt_spec->limit == 0) || (delta < fmt_spec->limit)) |
468 | fmt_used = fmt_spec->fmt1; | |
afa8937a | 469 | else |
c37e7851 OZ |
470 | fmt_used = fmt_spec->fmt2; |
471 | ||
472 | int rv = strftime(x, TM_DATETIME_BUFFER_SIZE, fmt_used, tm); | |
473 | if (((rv == 0) && fmt_used[0]) || (rv == TM_DATETIME_BUFFER_SIZE)) | |
474 | strcpy(x, "<too-long>"); | |
afa8937a MM |
475 | } |
476 | ||
05476c4d | 477 | |
525fa2c1 MM |
478 | /** |
479 | * DOC: Sockets | |
480 | * | |
481 | * Socket resources represent network connections. Their data structure (&socket) | |
482 | * contains a lot of fields defining the exact type of the socket, the local and | |
483 | * remote addresses and ports, pointers to socket buffers and finally pointers to | |
484 | * hook functions to be called when new data have arrived to the receive buffer | |
485 | * (@rx_hook), when the contents of the transmit buffer have been transmitted | |
486 | * (@tx_hook) and when an error or connection close occurs (@err_hook). | |
487 | * | |
38a608c5 | 488 | * Freeing of sockets from inside socket hooks is perfectly safe. |
b5d9ee5c MM |
489 | */ |
490 | ||
abae6e9c MM |
491 | #ifndef SOL_IP |
492 | #define SOL_IP IPPROTO_IP | |
493 | #endif | |
494 | ||
b1a1faba OF |
495 | #ifndef SOL_IPV6 |
496 | #define SOL_IPV6 IPPROTO_IPV6 | |
497 | #endif | |
498 | ||
48e5f32d OZ |
499 | #ifndef SOL_ICMPV6 |
500 | #define SOL_ICMPV6 IPPROTO_ICMPV6 | |
501 | #endif | |
502 | ||
503 | ||
05476c4d OZ |
504 | /* |
505 | * Sockaddr helper functions | |
506 | */ | |
38a608c5 | 507 | |
05476c4d OZ |
508 | static inline int sockaddr_length(int af) |
509 | { return (af == AF_INET) ? sizeof(struct sockaddr_in) : sizeof(struct sockaddr_in6); } | |
510 | ||
511 | static inline void | |
512 | sockaddr_fill4(struct sockaddr_in *sa, ip_addr a, struct iface *ifa, uint port) | |
38a608c5 | 513 | { |
05476c4d OZ |
514 | memset(sa, 0, sizeof(struct sockaddr_in)); |
515 | #ifdef HAVE_SIN_LEN | |
516 | sa->sin_len = sizeof(struct sockaddr_in); | |
517 | #endif | |
518 | sa->sin_family = AF_INET; | |
519 | sa->sin_port = htons(port); | |
520 | sa->sin_addr = ipa_to_in4(a); | |
38a608c5 | 521 | } |
b5d9ee5c | 522 | |
05476c4d OZ |
523 | static inline void |
524 | sockaddr_fill6(struct sockaddr_in6 *sa, ip_addr a, struct iface *ifa, uint port) | |
b5d9ee5c | 525 | { |
05476c4d OZ |
526 | memset(sa, 0, sizeof(struct sockaddr_in6)); |
527 | #ifdef SIN6_LEN | |
528 | sa->sin6_len = sizeof(struct sockaddr_in6); | |
529 | #endif | |
530 | sa->sin6_family = AF_INET6; | |
531 | sa->sin6_port = htons(port); | |
532 | sa->sin6_flowinfo = 0; | |
533 | sa->sin6_addr = ipa_to_in6(a); | |
534 | ||
535 | if (ifa && ipa_is_link_local(a)) | |
536 | sa->sin6_scope_id = ifa->index; | |
4da25acb | 537 | } |
b5d9ee5c | 538 | |
05476c4d OZ |
539 | void |
540 | sockaddr_fill(sockaddr *sa, int af, ip_addr a, struct iface *ifa, uint port) | |
4da25acb | 541 | { |
05476c4d OZ |
542 | if (af == AF_INET) |
543 | sockaddr_fill4((struct sockaddr_in *) sa, a, ifa, port); | |
544 | else if (af == AF_INET6) | |
545 | sockaddr_fill6((struct sockaddr_in6 *) sa, a, ifa, port); | |
546 | else | |
547 | bug("Unknown AF"); | |
4da25acb MM |
548 | } |
549 | ||
05476c4d OZ |
550 | static inline void |
551 | sockaddr_read4(struct sockaddr_in *sa, ip_addr *a, struct iface **ifa, uint *port) | |
4da25acb | 552 | { |
05476c4d OZ |
553 | *port = ntohs(sa->sin_port); |
554 | *a = ipa_from_in4(sa->sin_addr); | |
b5d9ee5c MM |
555 | } |
556 | ||
05476c4d OZ |
557 | static inline void |
558 | sockaddr_read6(struct sockaddr_in6 *sa, ip_addr *a, struct iface **ifa, uint *port) | |
48e5f32d | 559 | { |
05476c4d OZ |
560 | *port = ntohs(sa->sin6_port); |
561 | *a = ipa_from_in6(sa->sin6_addr); | |
48e5f32d | 562 | |
05476c4d OZ |
563 | if (ifa && ipa_is_link_local(*a)) |
564 | *ifa = if_find_by_index(sa->sin6_scope_id); | |
48e5f32d OZ |
565 | } |
566 | ||
05476c4d OZ |
567 | int |
568 | sockaddr_read(sockaddr *sa, int af, ip_addr *a, struct iface **ifa, uint *port) | |
48e5f32d | 569 | { |
05476c4d OZ |
570 | if (sa->sa.sa_family != af) |
571 | goto fail; | |
48e5f32d | 572 | |
05476c4d OZ |
573 | if (af == AF_INET) |
574 | sockaddr_read4((struct sockaddr_in *) sa, a, ifa, port); | |
575 | else if (af == AF_INET6) | |
576 | sockaddr_read6((struct sockaddr_in6 *) sa, a, ifa, port); | |
577 | else | |
578 | goto fail; | |
48e5f32d | 579 | |
05476c4d | 580 | return 0; |
48e5f32d | 581 | |
05476c4d OZ |
582 | fail: |
583 | *a = IPA_NONE; | |
584 | *port = 0; | |
585 | return -1; | |
48e5f32d OZ |
586 | } |
587 | ||
48e5f32d | 588 | |
05476c4d OZ |
589 | /* |
590 | * IPv6 multicast syscalls | |
591 | */ | |
4da25acb | 592 | |
05476c4d | 593 | /* Fortunately standardized in RFC 3493 */ |
b5d9ee5c | 594 | |
05476c4d OZ |
595 | #define INIT_MREQ6(maddr,ifa) \ |
596 | { .ipv6mr_multiaddr = ipa_to_in6(maddr), .ipv6mr_interface = ifa->index } | |
b5d9ee5c | 597 | |
05476c4d OZ |
598 | static inline int |
599 | sk_setup_multicast6(sock *s) | |
b5d9ee5c | 600 | { |
05476c4d OZ |
601 | int index = s->iface->index; |
602 | int ttl = s->ttl; | |
603 | int n = 0; | |
b5d9ee5c | 604 | |
05476c4d OZ |
605 | if (setsockopt(s->fd, SOL_IPV6, IPV6_MULTICAST_IF, &index, sizeof(index)) < 0) |
606 | ERR("IPV6_MULTICAST_IF"); | |
b5d9ee5c | 607 | |
05476c4d OZ |
608 | if (setsockopt(s->fd, SOL_IPV6, IPV6_MULTICAST_HOPS, &ttl, sizeof(ttl)) < 0) |
609 | ERR("IPV6_MULTICAST_HOPS"); | |
4f22c981 | 610 | |
05476c4d OZ |
611 | if (setsockopt(s->fd, SOL_IPV6, IPV6_MULTICAST_LOOP, &n, sizeof(n)) < 0) |
612 | ERR("IPV6_MULTICAST_LOOP"); | |
4f22c981 | 613 | |
05476c4d | 614 | return 0; |
061ab802 OZ |
615 | } |
616 | ||
05476c4d OZ |
617 | static inline int |
618 | sk_join_group6(sock *s, ip_addr maddr) | |
4f22c981 | 619 | { |
05476c4d | 620 | struct ipv6_mreq mr = INIT_MREQ6(maddr, s->iface); |
eb1451a3 | 621 | |
05476c4d OZ |
622 | if (setsockopt(s->fd, SOL_IPV6, IPV6_JOIN_GROUP, &mr, sizeof(mr)) < 0) |
623 | ERR("IPV6_JOIN_GROUP"); | |
4f22c981 | 624 | |
05476c4d | 625 | return 0; |
b5d9ee5c MM |
626 | } |
627 | ||
05476c4d OZ |
628 | static inline int |
629 | sk_leave_group6(sock *s, ip_addr maddr) | |
b5d9ee5c | 630 | { |
05476c4d | 631 | struct ipv6_mreq mr = INIT_MREQ6(maddr, s->iface); |
b5d9ee5c | 632 | |
05476c4d OZ |
633 | if (setsockopt(s->fd, SOL_IPV6, IPV6_LEAVE_GROUP, &mr, sizeof(mr)) < 0) |
634 | ERR("IPV6_LEAVE_GROUP"); | |
635 | ||
636 | return 0; | |
637 | } | |
4f22c981 | 638 | |
bed41728 | 639 | |
05476c4d OZ |
640 | /* |
641 | * IPv6 packet control messages | |
642 | */ | |
bed41728 | 643 | |
05476c4d | 644 | /* Also standardized, in RFC 3542 */ |
bed41728 | 645 | |
dcc60494 OZ |
646 | /* |
647 | * RFC 2292 uses IPV6_PKTINFO for both the socket option and the cmsg | |
648 | * type, RFC 3542 changed the socket option to IPV6_RECVPKTINFO. If we | |
649 | * don't have IPV6_RECVPKTINFO we suppose the OS implements the older | |
650 | * RFC and we use IPV6_PKTINFO. | |
651 | */ | |
652 | #ifndef IPV6_RECVPKTINFO | |
653 | #define IPV6_RECVPKTINFO IPV6_PKTINFO | |
654 | #endif | |
70e212f9 OZ |
655 | /* |
656 | * Same goes for IPV6_HOPLIMIT -> IPV6_RECVHOPLIMIT. | |
657 | */ | |
658 | #ifndef IPV6_RECVHOPLIMIT | |
659 | #define IPV6_RECVHOPLIMIT IPV6_HOPLIMIT | |
660 | #endif | |
dcc60494 | 661 | |
70e212f9 | 662 | |
05476c4d OZ |
663 | #define CMSG6_SPACE_PKTINFO CMSG_SPACE(sizeof(struct in6_pktinfo)) |
664 | #define CMSG6_SPACE_TTL CMSG_SPACE(sizeof(int)) | |
bed41728 | 665 | |
05476c4d OZ |
666 | static inline int |
667 | sk_request_cmsg6_pktinfo(sock *s) | |
668 | { | |
669 | int y = 1; | |
70e212f9 | 670 | |
05476c4d OZ |
671 | if (setsockopt(s->fd, SOL_IPV6, IPV6_RECVPKTINFO, &y, sizeof(y)) < 0) |
672 | ERR("IPV6_RECVPKTINFO"); | |
673 | ||
674 | return 0; | |
bed41728 OZ |
675 | } |
676 | ||
05476c4d OZ |
677 | static inline int |
678 | sk_request_cmsg6_ttl(sock *s) | |
bed41728 | 679 | { |
05476c4d | 680 | int y = 1; |
bed41728 | 681 | |
05476c4d OZ |
682 | if (setsockopt(s->fd, SOL_IPV6, IPV6_RECVHOPLIMIT, &y, sizeof(y)) < 0) |
683 | ERR("IPV6_RECVHOPLIMIT"); | |
70e212f9 | 684 | |
05476c4d OZ |
685 | return 0; |
686 | } | |
70e212f9 | 687 | |
05476c4d OZ |
688 | static inline void |
689 | sk_process_cmsg6_pktinfo(sock *s, struct cmsghdr *cm) | |
690 | { | |
691 | if (cm->cmsg_type == IPV6_PKTINFO) | |
70e212f9 | 692 | { |
05476c4d OZ |
693 | struct in6_pktinfo *pi = (struct in6_pktinfo *) CMSG_DATA(cm); |
694 | s->laddr = ipa_from_in6(pi->ipi6_addr); | |
695 | s->lifindex = pi->ipi6_ifindex; | |
70e212f9 | 696 | } |
05476c4d | 697 | } |
70e212f9 | 698 | |
05476c4d OZ |
699 | static inline void |
700 | sk_process_cmsg6_ttl(sock *s, struct cmsghdr *cm) | |
701 | { | |
702 | if (cm->cmsg_type == IPV6_HOPLIMIT) | |
703 | s->rcv_ttl = * (int *) CMSG_DATA(cm); | |
bed41728 OZ |
704 | } |
705 | ||
05476c4d OZ |
706 | static inline void |
707 | sk_prepare_cmsgs6(sock *s, struct msghdr *msg, void *cbuf, size_t cbuflen) | |
bed41728 OZ |
708 | { |
709 | struct cmsghdr *cm; | |
710 | struct in6_pktinfo *pi; | |
8945f73d | 711 | int controllen = 0; |
bed41728 | 712 | |
bed41728 OZ |
713 | msg->msg_control = cbuf; |
714 | msg->msg_controllen = cbuflen; | |
715 | ||
716 | cm = CMSG_FIRSTHDR(msg); | |
48e5f32d | 717 | cm->cmsg_level = SOL_IPV6; |
bed41728 OZ |
718 | cm->cmsg_type = IPV6_PKTINFO; |
719 | cm->cmsg_len = CMSG_LEN(sizeof(*pi)); | |
8945f73d | 720 | controllen += CMSG_SPACE(sizeof(*pi)); |
bed41728 OZ |
721 | |
722 | pi = (struct in6_pktinfo *) CMSG_DATA(cm); | |
bed41728 | 723 | pi->ipi6_ifindex = s->iface ? s->iface->index : 0; |
05476c4d | 724 | pi->ipi6_addr = ipa_to_in6(s->saddr); |
bed41728 | 725 | |
8945f73d | 726 | msg->msg_controllen = controllen; |
bed41728 | 727 | } |
48e5f32d | 728 | |
bed41728 | 729 | |
05476c4d OZ |
730 | /* |
731 | * Miscellaneous socket syscalls | |
732 | */ | |
733 | ||
734 | static inline int | |
735 | sk_set_ttl4(sock *s, int ttl) | |
a39b165e | 736 | { |
05476c4d OZ |
737 | if (setsockopt(s->fd, SOL_IP, IP_TTL, &ttl, sizeof(ttl)) < 0) |
738 | ERR("IP_TTL"); | |
739 | ||
740 | return 0; | |
a39b165e OZ |
741 | } |
742 | ||
05476c4d OZ |
743 | static inline int |
744 | sk_set_ttl6(sock *s, int ttl) | |
745 | { | |
746 | if (setsockopt(s->fd, SOL_IPV6, IPV6_UNICAST_HOPS, &ttl, sizeof(ttl)) < 0) | |
747 | ERR("IPV6_UNICAST_HOPS"); | |
38a608c5 | 748 | |
05476c4d OZ |
749 | return 0; |
750 | } | |
751 | ||
752 | static inline int | |
753 | sk_set_tos4(sock *s, int tos) | |
b5d9ee5c | 754 | { |
05476c4d OZ |
755 | if (setsockopt(s->fd, SOL_IP, IP_TOS, &tos, sizeof(tos)) < 0) |
756 | ERR("IP_TOS"); | |
b5d9ee5c | 757 | |
05476c4d OZ |
758 | return 0; |
759 | } | |
ef4a50be | 760 | |
05476c4d OZ |
761 | static inline int |
762 | sk_set_tos6(sock *s, int tos) | |
763 | { | |
764 | if (setsockopt(s->fd, SOL_IPV6, IPV6_TCLASS, &tos, sizeof(tos)) < 0) | |
765 | ERR("IPV6_TCLASS"); | |
48e5f32d | 766 | |
05476c4d OZ |
767 | return 0; |
768 | } | |
48e5f32d | 769 | |
88a183c6 OZ |
770 | static inline byte * |
771 | sk_skip_ip_header(byte *pkt, int *len) | |
772 | { | |
773 | if ((*len < 20) || ((*pkt & 0xf0) != 0x40)) | |
774 | return NULL; | |
775 | ||
776 | int hlen = (*pkt & 0x0f) * 4; | |
777 | if ((hlen < 20) || (hlen > *len)) | |
778 | return NULL; | |
779 | ||
780 | *len -= hlen; | |
781 | return pkt + hlen; | |
782 | } | |
783 | ||
784 | byte * | |
785 | sk_rx_buffer(sock *s, int *len) | |
786 | { | |
787 | if (sk_is_ipv4(s) && (s->type == SK_IP)) | |
788 | return sk_skip_ip_header(s->rbuf, len); | |
789 | else | |
790 | return s->rbuf; | |
791 | } | |
792 | ||
48e5f32d | 793 | |
05476c4d OZ |
794 | /* |
795 | * Public socket functions | |
796 | */ | |
48e5f32d | 797 | |
05476c4d OZ |
798 | /** |
799 | * sk_setup_multicast - enable multicast for given socket | |
800 | * @s: socket | |
801 | * | |
802 | * Prepare transmission of multicast packets for given datagram socket. | |
803 | * The socket must have defined @iface. | |
804 | * | |
805 | * Result: 0 for success, -1 for an error. | |
806 | */ | |
48e5f32d | 807 | |
05476c4d OZ |
808 | int |
809 | sk_setup_multicast(sock *s) | |
810 | { | |
811 | ASSERT(s->iface); | |
48e5f32d | 812 | |
05476c4d OZ |
813 | if (sk_is_ipv4(s)) |
814 | return sk_setup_multicast4(s); | |
815 | else | |
816 | return sk_setup_multicast6(s); | |
817 | } | |
48e5f32d | 818 | |
05476c4d OZ |
819 | /** |
820 | * sk_join_group - join multicast group for given socket | |
821 | * @s: socket | |
822 | * @maddr: multicast address | |
823 | * | |
824 | * Join multicast group for given datagram socket and associated interface. | |
825 | * The socket must have defined @iface. | |
826 | * | |
827 | * Result: 0 for success, -1 for an error. | |
828 | */ | |
789772ed | 829 | |
05476c4d OZ |
830 | int |
831 | sk_join_group(sock *s, ip_addr maddr) | |
832 | { | |
833 | if (sk_is_ipv4(s)) | |
834 | return sk_join_group4(s, maddr); | |
835 | else | |
836 | return sk_join_group6(s, maddr); | |
837 | } | |
ef4a50be | 838 | |
05476c4d OZ |
839 | /** |
840 | * sk_leave_group - leave multicast group for given socket | |
841 | * @s: socket | |
842 | * @maddr: multicast address | |
843 | * | |
844 | * Leave multicast group for given datagram socket and associated interface. | |
845 | * The socket must have defined @iface. | |
846 | * | |
847 | * Result: 0 for success, -1 for an error. | |
848 | */ | |
789772ed | 849 | |
05476c4d OZ |
850 | int |
851 | sk_leave_group(sock *s, ip_addr maddr) | |
852 | { | |
853 | if (sk_is_ipv4(s)) | |
854 | return sk_leave_group4(s, maddr); | |
855 | else | |
856 | return sk_leave_group6(s, maddr); | |
b5d9ee5c MM |
857 | } |
858 | ||
a39b165e | 859 | /** |
05476c4d OZ |
860 | * sk_setup_broadcast - enable broadcast for given socket |
861 | * @s: socket | |
862 | * | |
863 | * Allow reception and transmission of broadcast packets for given datagram | |
864 | * socket. The socket must have defined @iface. For transmission, packets should | |
865 | * be send to @brd address of @iface. | |
866 | * | |
867 | * Result: 0 for success, -1 for an error. | |
868 | */ | |
869 | ||
870 | int | |
871 | sk_setup_broadcast(sock *s) | |
872 | { | |
873 | int y = 1; | |
874 | ||
875 | if (setsockopt(s->fd, SOL_SOCKET, SO_BROADCAST, &y, sizeof(y)) < 0) | |
876 | ERR("SO_BROADCAST"); | |
877 | ||
878 | return 0; | |
879 | } | |
880 | ||
881 | /** | |
882 | * sk_set_ttl - set transmit TTL for given socket | |
a39b165e OZ |
883 | * @s: socket |
884 | * @ttl: TTL value | |
885 | * | |
05476c4d OZ |
886 | * Set TTL for already opened connections when TTL was not set before. Useful |
887 | * for accepted connections when different ones should have different TTL. | |
a39b165e OZ |
888 | * |
889 | * Result: 0 for success, -1 for an error. | |
890 | */ | |
891 | ||
892 | int | |
893 | sk_set_ttl(sock *s, int ttl) | |
894 | { | |
a39b165e | 895 | s->ttl = ttl; |
a39b165e | 896 | |
05476c4d OZ |
897 | if (sk_is_ipv4(s)) |
898 | return sk_set_ttl4(s, ttl); | |
899 | else | |
900 | return sk_set_ttl6(s, ttl); | |
a39b165e OZ |
901 | } |
902 | ||
b1b19433 | 903 | /** |
05476c4d | 904 | * sk_set_min_ttl - set minimal accepted TTL for given socket |
b1b19433 OZ |
905 | * @s: socket |
906 | * @ttl: TTL value | |
907 | * | |
05476c4d OZ |
908 | * Set minimal accepted TTL for given socket. Can be used for TTL security. |
909 | * implementations. | |
b1b19433 OZ |
910 | * |
911 | * Result: 0 for success, -1 for an error. | |
912 | */ | |
913 | ||
914 | int | |
915 | sk_set_min_ttl(sock *s, int ttl) | |
916 | { | |
05476c4d OZ |
917 | if (sk_is_ipv4(s)) |
918 | return sk_set_min_ttl4(s, ttl); | |
919 | else | |
920 | return sk_set_min_ttl6(s, ttl); | |
b1b19433 | 921 | } |
d51aa281 | 922 | |
05476c4d | 923 | #if 0 |
d51aa281 | 924 | /** |
05476c4d | 925 | * sk_set_md5_auth - add / remove MD5 security association for given socket |
d51aa281 OZ |
926 | * @s: socket |
927 | * @a: IP address of the other side | |
eb1451a3 | 928 | * @ifa: Interface for link-local IP address |
d51aa281 OZ |
929 | * @passwd: password used for MD5 authentication |
930 | * | |
05476c4d OZ |
931 | * In TCP MD5 handling code in kernel, there is a set of pairs (address, |
932 | * password) used to choose password according to address of the other side. | |
933 | * This function is useful for listening socket, for active sockets it is enough | |
934 | * to set s->password field. | |
d51aa281 OZ |
935 | * |
936 | * When called with passwd != NULL, the new pair is added, | |
937 | * When called with passwd == NULL, the existing pair is removed. | |
938 | * | |
939 | * Result: 0 for success, -1 for an error. | |
940 | */ | |
941 | ||
942 | int | |
eb1451a3 | 943 | sk_set_md5_auth(sock *s, ip_addr a, struct iface *ifa, char *passwd) |
05476c4d OZ |
944 | { DUMMY; } |
945 | #endif | |
f9c799a0 | 946 | |
05476c4d OZ |
947 | /** |
948 | * sk_set_ipv6_checksum - specify IPv6 checksum offset for given socket | |
949 | * @s: socket | |
950 | * @offset: offset | |
951 | * | |
952 | * Specify IPv6 checksum field offset for given raw IPv6 socket. After that, the | |
953 | * kernel will automatically fill it for outgoing packets and check it for | |
954 | * incoming packets. Should not be used on ICMPv6 sockets, where the position is | |
955 | * known to the kernel. | |
956 | * | |
957 | * Result: 0 for success, -1 for an error. | |
958 | */ | |
f9c799a0 | 959 | |
4ac7c834 OZ |
960 | int |
961 | sk_set_ipv6_checksum(sock *s, int offset) | |
962 | { | |
48e5f32d | 963 | if (setsockopt(s->fd, SOL_IPV6, IPV6_CHECKSUM, &offset, sizeof(offset)) < 0) |
05476c4d | 964 | ERR("IPV6_CHECKSUM"); |
4ac7c834 OZ |
965 | |
966 | return 0; | |
967 | } | |
968 | ||
93e868c7 | 969 | int |
05476c4d | 970 | sk_set_icmp6_filter(sock *s, int p1, int p2) |
93e868c7 OZ |
971 | { |
972 | /* a bit of lame interface, but it is here only for Radv */ | |
973 | struct icmp6_filter f; | |
974 | ||
975 | ICMP6_FILTER_SETBLOCKALL(&f); | |
976 | ICMP6_FILTER_SETPASS(p1, &f); | |
977 | ICMP6_FILTER_SETPASS(p2, &f); | |
978 | ||
48e5f32d | 979 | if (setsockopt(s->fd, SOL_ICMPV6, ICMP6_FILTER, &f, sizeof(f)) < 0) |
05476c4d | 980 | ERR("ICMP6_FILTER"); |
93e868c7 OZ |
981 | |
982 | return 0; | |
983 | } | |
984 | ||
05476c4d OZ |
985 | void |
986 | sk_log_error(sock *s, const char *p) | |
987 | { | |
988 | log(L_ERR "%s: Socket error: %s%#m", p, s->err); | |
989 | } | |
990 | ||
991 | ||
992 | /* | |
993 | * Actual struct birdsock code | |
994 | */ | |
995 | ||
996 | static list sock_list; | |
997 | static struct birdsock *current_sock; | |
998 | static struct birdsock *stored_sock; | |
999 | static int sock_recalc_fdsets_p; | |
1000 | ||
1001 | static inline sock * | |
1002 | sk_next(sock *s) | |
1003 | { | |
1004 | if (!s->n.next->next) | |
1005 | return NULL; | |
1006 | else | |
1007 | return SKIP_BACK(sock, n, s->n.next); | |
1008 | } | |
1009 | ||
1010 | static void | |
1011 | sk_alloc_bufs(sock *s) | |
1012 | { | |
1013 | if (!s->rbuf && s->rbsize) | |
1014 | s->rbuf = s->rbuf_alloc = xmalloc(s->rbsize); | |
1015 | s->rpos = s->rbuf; | |
1016 | if (!s->tbuf && s->tbsize) | |
1017 | s->tbuf = s->tbuf_alloc = xmalloc(s->tbsize); | |
1018 | s->tpos = s->ttx = s->tbuf; | |
1019 | } | |
1020 | ||
1021 | static void | |
1022 | sk_free_bufs(sock *s) | |
1023 | { | |
1024 | if (s->rbuf_alloc) | |
1025 | { | |
1026 | xfree(s->rbuf_alloc); | |
1027 | s->rbuf = s->rbuf_alloc = NULL; | |
1028 | } | |
1029 | if (s->tbuf_alloc) | |
1030 | { | |
1031 | xfree(s->tbuf_alloc); | |
1032 | s->tbuf = s->tbuf_alloc = NULL; | |
1033 | } | |
1034 | } | |
1035 | ||
1036 | static void | |
1037 | sk_free(resource *r) | |
1038 | { | |
1039 | sock *s = (sock *) r; | |
1040 | ||
1041 | sk_free_bufs(s); | |
1042 | if (s->fd >= 0) | |
1043 | { | |
1044 | close(s->fd); | |
1045 | ||
1046 | /* FIXME: we should call sk_stop() for SKF_THREAD sockets */ | |
1047 | if (s->flags & SKF_THREAD) | |
1048 | return; | |
1049 | ||
1050 | if (s == current_sock) | |
1051 | current_sock = sk_next(s); | |
1052 | if (s == stored_sock) | |
1053 | stored_sock = sk_next(s); | |
1054 | rem_node(&s->n); | |
1055 | sock_recalc_fdsets_p = 1; | |
1056 | } | |
1057 | } | |
1058 | ||
1059 | void | |
1060 | sk_set_rbsize(sock *s, uint val) | |
1061 | { | |
1062 | ASSERT(s->rbuf_alloc == s->rbuf); | |
1063 | ||
1064 | if (s->rbsize == val) | |
1065 | return; | |
1066 | ||
1067 | s->rbsize = val; | |
1068 | xfree(s->rbuf_alloc); | |
1069 | s->rbuf_alloc = xmalloc(val); | |
1070 | s->rpos = s->rbuf = s->rbuf_alloc; | |
1071 | } | |
1072 | ||
1073 | void | |
1074 | sk_set_tbsize(sock *s, uint val) | |
1075 | { | |
1076 | ASSERT(s->tbuf_alloc == s->tbuf); | |
1077 | ||
1078 | if (s->tbsize == val) | |
1079 | return; | |
1080 | ||
1081 | byte *old_tbuf = s->tbuf; | |
1082 | ||
1083 | s->tbsize = val; | |
1084 | s->tbuf = s->tbuf_alloc = xrealloc(s->tbuf_alloc, val); | |
1085 | s->tpos = s->tbuf + (s->tpos - old_tbuf); | |
1086 | s->ttx = s->tbuf + (s->ttx - old_tbuf); | |
1087 | } | |
1088 | ||
1089 | void | |
1090 | sk_set_tbuf(sock *s, void *tbuf) | |
1091 | { | |
1092 | s->tbuf = tbuf ?: s->tbuf_alloc; | |
1093 | s->ttx = s->tpos = s->tbuf; | |
1094 | } | |
1095 | ||
1096 | void | |
1097 | sk_reallocate(sock *s) | |
1098 | { | |
1099 | sk_free_bufs(s); | |
1100 | sk_alloc_bufs(s); | |
1101 | } | |
1102 | ||
1103 | static void | |
1104 | sk_dump(resource *r) | |
1105 | { | |
1106 | sock *s = (sock *) r; | |
1107 | static char *sk_type_names[] = { "TCP<", "TCP>", "TCP", "UDP", NULL, "IP", NULL, "MAGIC", "UNIX<", "UNIX", "DEL!" }; | |
1108 | ||
1109 | debug("(%s, ud=%p, sa=%08x, sp=%d, da=%08x, dp=%d, tos=%d, ttl=%d, if=%s)\n", | |
1110 | sk_type_names[s->type], | |
1111 | s->data, | |
1112 | s->saddr, | |
1113 | s->sport, | |
1114 | s->daddr, | |
1115 | s->dport, | |
1116 | s->tos, | |
1117 | s->ttl, | |
1118 | s->iface ? s->iface->name : "none"); | |
1119 | } | |
1120 | ||
1121 | static struct resclass sk_class = { | |
1122 | "Socket", | |
1123 | sizeof(sock), | |
1124 | sk_free, | |
1125 | sk_dump, | |
1126 | NULL, | |
1127 | NULL | |
1128 | }; | |
1129 | ||
1130 | /** | |
1131 | * sk_new - create a socket | |
1132 | * @p: pool | |
1133 | * | |
1134 | * This function creates a new socket resource. If you want to use it, | |
1135 | * you need to fill in all the required fields of the structure and | |
1136 | * call sk_open() to do the actual opening of the socket. | |
1137 | * | |
1138 | * The real function name is sock_new(), sk_new() is a macro wrapper | |
1139 | * to avoid collision with OpenSSL. | |
1140 | */ | |
1141 | sock * | |
1142 | sock_new(pool *p) | |
1143 | { | |
1144 | sock *s = ralloc(p, &sk_class); | |
1145 | s->pool = p; | |
1146 | // s->saddr = s->daddr = IPA_NONE; | |
1147 | s->tos = s->priority = s->ttl = -1; | |
1148 | s->fd = -1; | |
1149 | return s; | |
1150 | } | |
1151 | ||
1152 | static int | |
1153 | sk_setup(sock *s) | |
f9c799a0 | 1154 | { |
05476c4d OZ |
1155 | int y = 1; |
1156 | int fd = s->fd; | |
f9c799a0 | 1157 | |
05476c4d OZ |
1158 | if (fcntl(fd, F_SETFL, O_NONBLOCK) < 0) |
1159 | ERR("O_NONBLOCK"); | |
f9c799a0 | 1160 | |
05476c4d OZ |
1161 | if (!s->af) |
1162 | return 0; | |
f9c799a0 | 1163 | |
05476c4d OZ |
1164 | if (ipa_nonzero(s->saddr) && !(s->flags & SKF_BIND)) |
1165 | s->flags |= SKF_PKTINFO; | |
f9c799a0 | 1166 | |
05476c4d OZ |
1167 | #ifdef CONFIG_USE_HDRINCL |
1168 | if (sk_is_ipv4(s) && (s->type == SK_IP) && (s->flags & SKF_PKTINFO)) | |
1169 | { | |
1170 | s->flags &= ~SKF_PKTINFO; | |
1171 | s->flags |= SKF_HDRINCL; | |
1172 | if (setsockopt(fd, SOL_IP, IP_HDRINCL, &y, sizeof(y)) < 0) | |
1173 | ERR("IP_HDRINCL"); | |
1174 | } | |
48e5f32d OZ |
1175 | #endif |
1176 | ||
05476c4d OZ |
1177 | if (s->iface) |
1178 | { | |
1179 | #ifdef SO_BINDTODEVICE | |
1180 | struct ifreq ifr; | |
1181 | strcpy(ifr.ifr_name, s->iface->name); | |
1182 | if (setsockopt(s->fd, SOL_SOCKET, SO_BINDTODEVICE, &ifr, sizeof(ifr)) < 0) | |
1183 | ERR("SO_BINDTODEVICE"); | |
1184 | #endif | |
f1aceff5 | 1185 | |
05476c4d OZ |
1186 | #ifdef CONFIG_UNIX_DONTROUTE |
1187 | if (setsockopt(s->fd, SOL_SOCKET, SO_DONTROUTE, &y, sizeof(y)) < 0) | |
1188 | ERR("SO_DONTROUTE"); | |
1189 | #endif | |
1190 | } | |
f9c799a0 | 1191 | |
05476c4d OZ |
1192 | if (s->priority >= 0) |
1193 | if (sk_set_priority(s, s->priority) < 0) | |
f9c799a0 | 1194 | return -1; |
f9c799a0 | 1195 | |
05476c4d OZ |
1196 | if (sk_is_ipv4(s)) |
1197 | { | |
1198 | if (s->flags & SKF_LADDR_RX) | |
1199 | if (sk_request_cmsg4_pktinfo(s) < 0) | |
1200 | return -1; | |
f9c799a0 | 1201 | |
05476c4d OZ |
1202 | if (s->flags & SKF_TTL_RX) |
1203 | if (sk_request_cmsg4_ttl(s) < 0) | |
1204 | return -1; | |
f9c799a0 | 1205 | |
05476c4d OZ |
1206 | if ((s->type == SK_UDP) || (s->type == SK_IP)) |
1207 | if (sk_disable_mtu_disc4(s) < 0) | |
1208 | return -1; | |
f9c799a0 | 1209 | |
05476c4d OZ |
1210 | if (s->ttl >= 0) |
1211 | if (sk_set_ttl4(s, s->ttl) < 0) | |
1212 | return -1; | |
f9c799a0 | 1213 | |
05476c4d OZ |
1214 | if (s->tos >= 0) |
1215 | if (sk_set_tos4(s, s->tos) < 0) | |
1216 | return -1; | |
1217 | } | |
f9c799a0 | 1218 | |
05476c4d OZ |
1219 | if (sk_is_ipv6(s)) |
1220 | { | |
1221 | if (s->flags & SKF_V6ONLY) | |
1222 | if (setsockopt(fd, SOL_IPV6, IPV6_V6ONLY, &y, sizeof(y)) < 0) | |
1223 | ERR("IPV6_V6ONLY"); | |
f9c799a0 | 1224 | |
05476c4d OZ |
1225 | if (s->flags & SKF_LADDR_RX) |
1226 | if (sk_request_cmsg6_pktinfo(s) < 0) | |
1227 | return -1; | |
f9c799a0 | 1228 | |
05476c4d OZ |
1229 | if (s->flags & SKF_TTL_RX) |
1230 | if (sk_request_cmsg6_ttl(s) < 0) | |
1231 | return -1; | |
f9c799a0 | 1232 | |
05476c4d OZ |
1233 | if ((s->type == SK_UDP) || (s->type == SK_IP)) |
1234 | if (sk_disable_mtu_disc6(s) < 0) | |
1235 | return -1; | |
f9c799a0 | 1236 | |
05476c4d OZ |
1237 | if (s->ttl >= 0) |
1238 | if (sk_set_ttl6(s, s->ttl) < 0) | |
1239 | return -1; | |
f9c799a0 | 1240 | |
05476c4d OZ |
1241 | if (s->tos >= 0) |
1242 | if (sk_set_tos6(s, s->tos) < 0) | |
1243 | return -1; | |
1244 | } | |
f9c799a0 OZ |
1245 | |
1246 | return 0; | |
1247 | } | |
1248 | ||
05476c4d OZ |
1249 | static void |
1250 | sk_insert(sock *s) | |
f9c799a0 | 1251 | { |
05476c4d OZ |
1252 | add_tail(&sock_list, &s->n); |
1253 | sock_recalc_fdsets_p = 1; | |
f9c799a0 OZ |
1254 | } |
1255 | ||
b93abffa | 1256 | static void |
b5d9ee5c MM |
1257 | sk_tcp_connected(sock *s) |
1258 | { | |
05476c4d OZ |
1259 | sockaddr sa; |
1260 | int sa_len = sizeof(sa); | |
1261 | ||
1262 | if ((getsockname(s->fd, &sa.sa, &sa_len) < 0) || | |
1263 | (sockaddr_read(&sa, s->af, &s->saddr, &s->iface, &s->sport) < 0)) | |
1264 | log(L_WARN "SOCK: Cannot get local IP address for TCP>"); | |
9be9a264 | 1265 | |
b5d9ee5c MM |
1266 | s->type = SK_TCP; |
1267 | sk_alloc_bufs(s); | |
320f4173 | 1268 | s->tx_hook(s); |
b5d9ee5c MM |
1269 | } |
1270 | ||
b93abffa | 1271 | static int |
05476c4d | 1272 | sk_passive_connected(sock *s, int type) |
b93abffa | 1273 | { |
05476c4d OZ |
1274 | sockaddr loc_sa, rem_sa; |
1275 | int loc_sa_len = sizeof(loc_sa); | |
1276 | int rem_sa_len = sizeof(rem_sa); | |
cf31112f | 1277 | |
05476c4d OZ |
1278 | int fd = accept(s->fd, ((type == SK_TCP) ? &rem_sa.sa : NULL), &rem_sa_len); |
1279 | if (fd < 0) | |
1280 | { | |
1281 | if ((errno != EINTR) && (errno != EAGAIN)) | |
c025b852 | 1282 | s->err_hook(s, errno); |
05476c4d OZ |
1283 | return 0; |
1284 | } | |
1285 | ||
1286 | sock *t = sk_new(s->pool); | |
1287 | t->type = type; | |
1288 | t->fd = fd; | |
1289 | t->af = s->af; | |
1290 | t->ttl = s->ttl; | |
1291 | t->tos = s->tos; | |
1292 | t->rbsize = s->rbsize; | |
1293 | t->tbsize = s->tbsize; | |
1294 | ||
1295 | if (type == SK_TCP) | |
1296 | { | |
1297 | if ((getsockname(fd, &loc_sa.sa, &loc_sa_len) < 0) || | |
1298 | (sockaddr_read(&loc_sa, s->af, &t->saddr, &t->iface, &t->sport) < 0)) | |
1299 | log(L_WARN "SOCK: Cannot get local IP address for TCP<"); | |
1300 | ||
1301 | if (sockaddr_read(&rem_sa, s->af, &t->daddr, &t->iface, &t->dport) < 0) | |
1302 | log(L_WARN "SOCK: Cannot get remote IP address for TCP<"); | |
1303 | } | |
1304 | ||
1305 | if (sk_setup(t) < 0) | |
1306 | { | |
1307 | /* FIXME: Call err_hook instead ? */ | |
1308 | log(L_ERR "SOCK: Incoming connection: %s%#m", t->err); | |
1309 | ||
1310 | /* FIXME: handle it better in rfree() */ | |
9c89560e | 1311 | close(t->fd); |
05476c4d OZ |
1312 | t->fd = -1; |
1313 | rfree(t); | |
1314 | return 1; | |
1315 | } | |
1316 | ||
1317 | sk_insert(t); | |
1318 | sk_alloc_bufs(t); | |
1319 | s->rx_hook(t, 0); | |
1320 | return 1; | |
b93abffa MM |
1321 | } |
1322 | ||
525fa2c1 MM |
1323 | /** |
1324 | * sk_open - open a socket | |
1325 | * @s: socket | |
1326 | * | |
1327 | * This function takes a socket resource created by sk_new() and | |
1328 | * initialized by the user and binds a corresponding network connection | |
1329 | * to it. | |
1330 | * | |
1331 | * Result: 0 for success, -1 for an error. | |
1332 | */ | |
b5d9ee5c MM |
1333 | int |
1334 | sk_open(sock *s) | |
1335 | { | |
05476c4d OZ |
1336 | int af = BIRD_AF; |
1337 | int fd = -1; | |
48e5f32d OZ |
1338 | int do_bind = 0; |
1339 | int bind_port = 0; | |
1340 | ip_addr bind_addr = IPA_NONE; | |
1341 | sockaddr sa; | |
b5d9ee5c | 1342 | |
48e5f32d | 1343 | switch (s->type) |
05476c4d OZ |
1344 | { |
1345 | case SK_TCP_ACTIVE: | |
1346 | s->ttx = ""; /* Force s->ttx != s->tpos */ | |
1347 | /* Fall thru */ | |
1348 | case SK_TCP_PASSIVE: | |
1349 | fd = socket(af, SOCK_STREAM, IPPROTO_TCP); | |
1350 | bind_port = s->sport; | |
1351 | bind_addr = s->saddr; | |
1352 | do_bind = bind_port || ipa_nonzero(bind_addr); | |
1353 | break; | |
9c89560e | 1354 | |
05476c4d OZ |
1355 | case SK_UDP: |
1356 | fd = socket(af, SOCK_DGRAM, IPPROTO_UDP); | |
1357 | bind_port = s->sport; | |
1358 | bind_addr = (s->flags & SKF_BIND) ? s->saddr : IPA_NONE; | |
1359 | do_bind = 1; | |
1360 | break; | |
1361 | ||
1362 | case SK_IP: | |
1363 | fd = socket(af, SOCK_RAW, s->dport); | |
1364 | bind_port = 0; | |
1365 | bind_addr = (s->flags & SKF_BIND) ? s->saddr : IPA_NONE; | |
1366 | do_bind = ipa_nonzero(bind_addr); | |
1367 | break; | |
1368 | ||
1369 | case SK_MAGIC: | |
1370 | af = 0; | |
1371 | fd = s->fd; | |
1372 | break; | |
1373 | ||
1374 | default: | |
1375 | bug("sk_open() called for invalid sock type %d", s->type); | |
1376 | } | |
1377 | ||
b5d9ee5c | 1378 | if (fd < 0) |
05476c4d OZ |
1379 | ERR("socket"); |
1380 | ||
1381 | s->af = af; | |
b5d9ee5c MM |
1382 | s->fd = fd; |
1383 | ||
05476c4d OZ |
1384 | if (sk_setup(s) < 0) |
1385 | goto err; | |
38a608c5 | 1386 | |
48e5f32d | 1387 | if (do_bind) |
05476c4d OZ |
1388 | { |
1389 | if (bind_port) | |
b5d9ee5c | 1390 | { |
05476c4d OZ |
1391 | int y = 1; |
1392 | ||
1393 | if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &y, sizeof(y)) < 0) | |
1394 | ERR2("SO_REUSEADDR"); | |
48e5f32d | 1395 | |
8931425d | 1396 | #ifdef CONFIG_NO_IFACE_BIND |
05476c4d OZ |
1397 | /* Workaround missing ability to bind to an iface */ |
1398 | if ((s->type == SK_UDP) && s->iface && ipa_zero(bind_addr)) | |
1399 | { | |
1400 | if (setsockopt(fd, SOL_SOCKET, SO_REUSEPORT, &y, sizeof(y)) < 0) | |
1401 | ERR2("SO_REUSEPORT"); | |
1402 | } | |
8931425d | 1403 | #endif |
b5d9ee5c | 1404 | } |
9c89560e OZ |
1405 | #ifdef IP_PORTRANGE |
1406 | else if (s->flags & SKF_HIGH_PORT) | |
1407 | { | |
1408 | int range = IP_PORTRANGE_HIGH; | |
1409 | if (setsockopt(fd, IPPROTO_IP, IP_PORTRANGE, &range, sizeof(range)) < 0) | |
1410 | log(L_WARN "Socket error: %s%#m", "IP_PORTRANGE"); | |
1411 | } | |
1412 | #endif | |
48e5f32d | 1413 | |
05476c4d OZ |
1414 | sockaddr_fill(&sa, af, bind_addr, s->iface, bind_port); |
1415 | if (bind(fd, &sa.sa, SA_LEN(sa)) < 0) | |
1416 | ERR2("bind"); | |
1417 | } | |
d51aa281 OZ |
1418 | |
1419 | if (s->password) | |
05476c4d OZ |
1420 | if (sk_set_md5_auth(s, s->daddr, s->iface, s->password) < 0) |
1421 | goto err; | |
d51aa281 | 1422 | |
48e5f32d | 1423 | switch (s->type) |
05476c4d OZ |
1424 | { |
1425 | case SK_TCP_ACTIVE: | |
1426 | sockaddr_fill(&sa, af, s->daddr, s->iface, s->dport); | |
1427 | if (connect(fd, &sa.sa, SA_LEN(sa)) >= 0) | |
1428 | sk_tcp_connected(s); | |
1429 | else if (errno != EINTR && errno != EAGAIN && errno != EINPROGRESS && | |
1430 | errno != ECONNREFUSED && errno != EHOSTUNREACH && errno != ENETUNREACH) | |
1431 | ERR2("connect"); | |
1432 | break; | |
1433 | ||
1434 | case SK_TCP_PASSIVE: | |
1435 | if (listen(fd, 8) < 0) | |
1436 | ERR2("listen"); | |
1437 | break; | |
1438 | ||
1439 | case SK_MAGIC: | |
1440 | break; | |
1441 | ||
1442 | default: | |
1443 | sk_alloc_bufs(s); | |
1444 | } | |
b5d9ee5c | 1445 | |
bf139664 OZ |
1446 | if (!(s->flags & SKF_THREAD)) |
1447 | sk_insert(s); | |
b5d9ee5c MM |
1448 | return 0; |
1449 | ||
05476c4d | 1450 | err: |
b5d9ee5c MM |
1451 | close(fd); |
1452 | s->fd = -1; | |
1453 | return -1; | |
1454 | } | |
1455 | ||
05476c4d | 1456 | int |
b93abffa MM |
1457 | sk_open_unix(sock *s, char *name) |
1458 | { | |
b93abffa | 1459 | struct sockaddr_un sa; |
05476c4d OZ |
1460 | int fd; |
1461 | ||
1462 | /* We are sloppy during error (leak fd and not set s->err), but we die anyway */ | |
b93abffa MM |
1463 | |
1464 | fd = socket(AF_UNIX, SOCK_STREAM, 0); | |
1465 | if (fd < 0) | |
05476c4d OZ |
1466 | return -1; |
1467 | ||
1468 | if (fcntl(fd, F_SETFL, O_NONBLOCK) < 0) | |
1469 | return -1; | |
68fa95cf | 1470 | |
97e46d28 | 1471 | /* Path length checked in test_old_bird() */ |
b93abffa | 1472 | sa.sun_family = AF_UNIX; |
97c6fa02 | 1473 | strcpy(sa.sun_path, name); |
05476c4d | 1474 | |
0b3bf4b1 | 1475 | if (bind(fd, (struct sockaddr *) &sa, SUN_LEN(&sa)) < 0) |
05476c4d OZ |
1476 | return -1; |
1477 | ||
1478 | if (listen(fd, 8) < 0) | |
1479 | return -1; | |
1480 | ||
1481 | s->fd = fd; | |
38a608c5 | 1482 | sk_insert(s); |
05476c4d OZ |
1483 | return 0; |
1484 | } | |
1485 | ||
1486 | ||
1487 | #define CMSG_RX_SPACE MAX(CMSG4_SPACE_PKTINFO+CMSG4_SPACE_TTL, \ | |
1488 | CMSG6_SPACE_PKTINFO+CMSG6_SPACE_TTL) | |
1489 | #define CMSG_TX_SPACE MAX(CMSG4_SPACE_PKTINFO,CMSG6_SPACE_PKTINFO) | |
1490 | ||
1491 | static void | |
1492 | sk_prepare_cmsgs(sock *s, struct msghdr *msg, void *cbuf, size_t cbuflen) | |
1493 | { | |
1494 | if (sk_is_ipv4(s)) | |
1495 | sk_prepare_cmsgs4(s, msg, cbuf, cbuflen); | |
1496 | else | |
1497 | sk_prepare_cmsgs6(s, msg, cbuf, cbuflen); | |
1498 | } | |
1499 | ||
1500 | static void | |
1501 | sk_process_cmsgs(sock *s, struct msghdr *msg) | |
1502 | { | |
1503 | struct cmsghdr *cm; | |
1504 | ||
1505 | s->laddr = IPA_NONE; | |
1506 | s->lifindex = 0; | |
1507 | s->rcv_ttl = -1; | |
1508 | ||
1509 | for (cm = CMSG_FIRSTHDR(msg); cm != NULL; cm = CMSG_NXTHDR(msg, cm)) | |
1510 | { | |
1511 | if ((cm->cmsg_level == SOL_IP) && sk_is_ipv4(s)) | |
1512 | { | |
1513 | sk_process_cmsg4_pktinfo(s, cm); | |
1514 | sk_process_cmsg4_ttl(s, cm); | |
1515 | } | |
b93abffa | 1516 | |
05476c4d OZ |
1517 | if ((cm->cmsg_level == SOL_IPV6) && sk_is_ipv6(s)) |
1518 | { | |
1519 | sk_process_cmsg6_pktinfo(s, cm); | |
1520 | sk_process_cmsg6_ttl(s, cm); | |
1521 | } | |
1522 | } | |
b93abffa MM |
1523 | } |
1524 | ||
48e5f32d OZ |
1525 | |
1526 | static inline int | |
1527 | sk_sendmsg(sock *s) | |
1528 | { | |
1529 | struct iovec iov = {s->tbuf, s->tpos - s->tbuf}; | |
1530 | byte cmsg_buf[CMSG_TX_SPACE]; | |
1531 | sockaddr dst; | |
1532 | ||
05476c4d | 1533 | sockaddr_fill(&dst, s->af, s->daddr, s->iface, s->dport); |
48e5f32d OZ |
1534 | |
1535 | struct msghdr msg = { | |
05476c4d OZ |
1536 | .msg_name = &dst.sa, |
1537 | .msg_namelen = SA_LEN(dst), | |
48e5f32d OZ |
1538 | .msg_iov = &iov, |
1539 | .msg_iovlen = 1 | |
1540 | }; | |
1541 | ||
1542 | #ifdef CONFIG_USE_HDRINCL | |
1543 | byte hdr[20]; | |
1544 | struct iovec iov2[2] = { {hdr, 20}, iov }; | |
1545 | ||
1546 | if (s->flags & SKF_HDRINCL) | |
1547 | { | |
05476c4d | 1548 | sk_prepare_ip_header(s, hdr, iov.iov_len); |
48e5f32d OZ |
1549 | msg.msg_iov = iov2; |
1550 | msg.msg_iovlen = 2; | |
1551 | } | |
1552 | #endif | |
1553 | ||
1554 | if (s->flags & SKF_PKTINFO) | |
05476c4d | 1555 | sk_prepare_cmsgs(s, &msg, cmsg_buf, sizeof(cmsg_buf)); |
48e5f32d OZ |
1556 | |
1557 | return sendmsg(s->fd, &msg, 0); | |
1558 | } | |
1559 | ||
1560 | static inline int | |
1561 | sk_recvmsg(sock *s) | |
1562 | { | |
1563 | struct iovec iov = {s->rbuf, s->rbsize}; | |
1564 | byte cmsg_buf[CMSG_RX_SPACE]; | |
1565 | sockaddr src; | |
1566 | ||
1567 | struct msghdr msg = { | |
05476c4d OZ |
1568 | .msg_name = &src.sa, |
1569 | .msg_namelen = sizeof(src), // XXXX ?? | |
48e5f32d OZ |
1570 | .msg_iov = &iov, |
1571 | .msg_iovlen = 1, | |
1572 | .msg_control = cmsg_buf, | |
1573 | .msg_controllen = sizeof(cmsg_buf), | |
1574 | .msg_flags = 0 | |
1575 | }; | |
1576 | ||
1577 | int rv = recvmsg(s->fd, &msg, 0); | |
1578 | if (rv < 0) | |
1579 | return rv; | |
1580 | ||
1581 | //ifdef IPV4 | |
1582 | // if (cf_type == SK_IP) | |
1583 | // rv = ipv4_skip_header(pbuf, rv); | |
1584 | //endif | |
1585 | ||
05476c4d OZ |
1586 | sockaddr_read(&src, s->af, &s->faddr, NULL, &s->fport); |
1587 | sk_process_cmsgs(s, &msg); | |
48e5f32d OZ |
1588 | |
1589 | if (msg.msg_flags & MSG_TRUNC) | |
1590 | s->flags |= SKF_TRUNCATED; | |
1591 | else | |
1592 | s->flags &= ~SKF_TRUNCATED; | |
1593 | ||
1594 | return rv; | |
1595 | } | |
1596 | ||
1597 | ||
353729f5 OZ |
1598 | static inline void reset_tx_buffer(sock *s) { s->ttx = s->tpos = s->tbuf; } |
1599 | ||
b5d9ee5c MM |
1600 | static int |
1601 | sk_maybe_write(sock *s) | |
1602 | { | |
1603 | int e; | |
1604 | ||
1605 | switch (s->type) | |
05476c4d OZ |
1606 | { |
1607 | case SK_TCP: | |
1608 | case SK_MAGIC: | |
1609 | case SK_UNIX: | |
1610 | while (s->ttx != s->tpos) | |
b5d9ee5c | 1611 | { |
05476c4d OZ |
1612 | e = write(s->fd, s->ttx, s->tpos - s->ttx); |
1613 | ||
1614 | if (e < 0) | |
1615 | { | |
1616 | if (errno != EINTR && errno != EAGAIN) | |
b5d9ee5c | 1617 | { |
05476c4d OZ |
1618 | reset_tx_buffer(s); |
1619 | /* EPIPE is just a connection close notification during TX */ | |
1620 | s->err_hook(s, (errno != EPIPE) ? errno : 0); | |
1621 | return -1; | |
b5d9ee5c | 1622 | } |
05476c4d OZ |
1623 | return 0; |
1624 | } | |
1625 | s->ttx += e; | |
1626 | } | |
1627 | reset_tx_buffer(s); | |
1628 | return 1; | |
1629 | ||
1630 | case SK_UDP: | |
1631 | case SK_IP: | |
1632 | { | |
1633 | if (s->tbuf == s->tpos) | |
b5d9ee5c | 1634 | return 1; |
05476c4d OZ |
1635 | |
1636 | e = sk_sendmsg(s); | |
1637 | ||
1638 | if (e < 0) | |
1639 | { | |
1640 | if (errno != EINTR && errno != EAGAIN) | |
1641 | { | |
1642 | reset_tx_buffer(s); | |
1643 | s->err_hook(s, errno); | |
1644 | return -1; | |
1645 | } | |
1646 | ||
1647 | if (!s->tx_hook) | |
1648 | reset_tx_buffer(s); | |
1649 | return 0; | |
b5d9ee5c | 1650 | } |
05476c4d OZ |
1651 | reset_tx_buffer(s); |
1652 | return 1; | |
b5d9ee5c | 1653 | } |
05476c4d OZ |
1654 | default: |
1655 | bug("sk_maybe_write: unknown socket type %d", s->type); | |
1656 | } | |
b5d9ee5c MM |
1657 | } |
1658 | ||
ea89da38 OZ |
1659 | int |
1660 | sk_rx_ready(sock *s) | |
1661 | { | |
1662 | fd_set rd, wr; | |
1663 | struct timeval timo; | |
1664 | int rv; | |
1665 | ||
1666 | FD_ZERO(&rd); | |
1667 | FD_ZERO(&wr); | |
1668 | FD_SET(s->fd, &rd); | |
1669 | ||
1670 | timo.tv_sec = 0; | |
1671 | timo.tv_usec = 0; | |
1672 | ||
1673 | redo: | |
1674 | rv = select(s->fd+1, &rd, &wr, NULL, &timo); | |
9c89560e | 1675 | |
ea89da38 OZ |
1676 | if ((rv < 0) && (errno == EINTR || errno == EAGAIN)) |
1677 | goto redo; | |
1678 | ||
1679 | return rv; | |
1680 | } | |
1681 | ||
525fa2c1 MM |
1682 | /** |
1683 | * sk_send - send data to a socket | |
1684 | * @s: socket | |
1685 | * @len: number of bytes to send | |
1686 | * | |
1687 | * This function sends @len bytes of data prepared in the | |
1688 | * transmit buffer of the socket @s to the network connection. | |
1689 | * If the packet can be sent immediately, it does so and returns | |
1690 | * 1, else it queues the packet for later processing, returns 0 | |
1691 | * and calls the @tx_hook of the socket when the tranmission | |
1692 | * takes place. | |
1693 | */ | |
b5d9ee5c MM |
1694 | int |
1695 | sk_send(sock *s, unsigned len) | |
1696 | { | |
b5d9ee5c MM |
1697 | s->ttx = s->tbuf; |
1698 | s->tpos = s->tbuf + len; | |
1699 | return sk_maybe_write(s); | |
1700 | } | |
1701 | ||
525fa2c1 MM |
1702 | /** |
1703 | * sk_send_to - send data to a specific destination | |
1704 | * @s: socket | |
1705 | * @len: number of bytes to send | |
1706 | * @addr: IP address to send the packet to | |
1707 | * @port: port to send the packet to | |
1708 | * | |
2e9b2421 | 1709 | * This is a sk_send() replacement for connection-less packet sockets |
525fa2c1 | 1710 | * which allows destination of the packet to be chosen dynamically. |
48e5f32d | 1711 | * Raw IP sockets should use 0 for @port. |
525fa2c1 | 1712 | */ |
b5d9ee5c MM |
1713 | int |
1714 | sk_send_to(sock *s, unsigned len, ip_addr addr, unsigned port) | |
1715 | { | |
353729f5 | 1716 | s->daddr = addr; |
48e5f32d OZ |
1717 | if (port) |
1718 | s->dport = port; | |
1719 | ||
b5d9ee5c MM |
1720 | s->ttx = s->tbuf; |
1721 | s->tpos = s->tbuf + len; | |
1722 | return sk_maybe_write(s); | |
1723 | } | |
1724 | ||
353729f5 OZ |
1725 | /* |
1726 | int | |
1727 | sk_send_full(sock *s, unsigned len, struct iface *ifa, | |
1728 | ip_addr saddr, ip_addr daddr, unsigned dport) | |
1729 | { | |
1730 | s->iface = ifa; | |
1731 | s->saddr = saddr; | |
1732 | s->daddr = daddr; | |
1733 | s->dport = dport; | |
1734 | s->ttx = s->tbuf; | |
1735 | s->tpos = s->tbuf + len; | |
1736 | return sk_maybe_write(s); | |
1737 | } | |
1738 | */ | |
1739 | ||
6a8d3f1c OZ |
1740 | /* sk_read() and sk_write() are called from BFD's event loop */ |
1741 | ||
1742 | int | |
b5d9ee5c MM |
1743 | sk_read(sock *s) |
1744 | { | |
1745 | switch (s->type) | |
05476c4d OZ |
1746 | { |
1747 | case SK_TCP_PASSIVE: | |
1748 | return sk_passive_connected(s, SK_TCP); | |
1749 | ||
1750 | case SK_UNIX_PASSIVE: | |
1751 | return sk_passive_connected(s, SK_UNIX); | |
1752 | ||
1753 | case SK_TCP: | |
1754 | case SK_UNIX: | |
b5d9ee5c | 1755 | { |
05476c4d OZ |
1756 | int c = read(s->fd, s->rpos, s->rbuf + s->rbsize - s->rpos); |
1757 | ||
1758 | if (c < 0) | |
b93abffa | 1759 | { |
05476c4d OZ |
1760 | if (errno != EINTR && errno != EAGAIN) |
1761 | s->err_hook(s, errno); | |
b5d9ee5c | 1762 | } |
05476c4d OZ |
1763 | else if (!c) |
1764 | s->err_hook(s, 0); | |
1765 | else | |
b5d9ee5c | 1766 | { |
05476c4d OZ |
1767 | s->rpos += c; |
1768 | if (s->rx_hook(s, s->rpos - s->rbuf)) | |
1769 | { | |
1770 | /* We need to be careful since the socket could have been deleted by the hook */ | |
1771 | if (current_sock == s) | |
1772 | s->rpos = s->rbuf; | |
1773 | } | |
1774 | return 1; | |
b5d9ee5c | 1775 | } |
05476c4d OZ |
1776 | return 0; |
1777 | } | |
353729f5 | 1778 | |
05476c4d OZ |
1779 | case SK_MAGIC: |
1780 | return s->rx_hook(s, 0); | |
b5d9ee5c | 1781 | |
05476c4d OZ |
1782 | default: |
1783 | { | |
1784 | int e = sk_recvmsg(s); | |
353729f5 | 1785 | |
05476c4d OZ |
1786 | if (e < 0) |
1787 | { | |
1788 | if (errno != EINTR && errno != EAGAIN) | |
1789 | s->err_hook(s, errno); | |
1790 | return 0; | |
b5d9ee5c | 1791 | } |
05476c4d OZ |
1792 | |
1793 | s->rpos = s->rbuf + e; | |
1794 | s->rx_hook(s, e); | |
1795 | return 1; | |
b5d9ee5c | 1796 | } |
05476c4d | 1797 | } |
b5d9ee5c MM |
1798 | } |
1799 | ||
6a8d3f1c | 1800 | int |
b5d9ee5c MM |
1801 | sk_write(sock *s) |
1802 | { | |
320f4173 | 1803 | switch (s->type) |
05476c4d OZ |
1804 | { |
1805 | case SK_TCP_ACTIVE: | |
320f4173 | 1806 | { |
05476c4d OZ |
1807 | sockaddr sa; |
1808 | sockaddr_fill(&sa, s->af, s->daddr, s->iface, s->dport); | |
1809 | ||
1810 | if (connect(s->fd, &sa.sa, SA_LEN(sa)) >= 0 || errno == EISCONN) | |
1811 | sk_tcp_connected(s); | |
1812 | else if (errno != EINTR && errno != EAGAIN && errno != EINPROGRESS) | |
1813 | s->err_hook(s, errno); | |
38a608c5 | 1814 | return 0; |
320f4173 | 1815 | } |
05476c4d OZ |
1816 | |
1817 | default: | |
1818 | if (s->ttx != s->tpos && sk_maybe_write(s) > 0) | |
1819 | { | |
1820 | if (s->tx_hook) | |
1821 | s->tx_hook(s); | |
1822 | return 1; | |
1823 | } | |
1824 | return 0; | |
1825 | } | |
b5d9ee5c MM |
1826 | } |
1827 | ||
1828 | void | |
1829 | sk_dump_all(void) | |
1830 | { | |
1831 | node *n; | |
1832 | sock *s; | |
1833 | ||
1834 | debug("Open sockets:\n"); | |
1835 | WALK_LIST(n, sock_list) | |
05476c4d OZ |
1836 | { |
1837 | s = SKIP_BACK(sock, n, n); | |
1838 | debug("%p ", s); | |
1839 | sk_dump(&s->r); | |
1840 | } | |
b5d9ee5c MM |
1841 | debug("\n"); |
1842 | } | |
1843 | ||
b5d9ee5c | 1844 | |
8bcb5fb1 OZ |
1845 | /* |
1846 | * Internal event log and watchdog | |
1847 | */ | |
1848 | ||
1849 | #define EVENT_LOG_LENGTH 32 | |
1850 | ||
1851 | struct event_log_entry | |
1852 | { | |
1853 | void *hook; | |
1854 | void *data; | |
1855 | btime timestamp; | |
1856 | btime duration; | |
1857 | }; | |
1858 | ||
1859 | static struct event_log_entry event_log[EVENT_LOG_LENGTH]; | |
1860 | static struct event_log_entry *event_open; | |
1861 | static int event_log_pos, event_log_num, watchdog_active; | |
1862 | static btime last_time; | |
1863 | static btime loop_time; | |
1864 | ||
1865 | static void | |
1866 | io_update_time(void) | |
1867 | { | |
1868 | struct timespec ts; | |
1869 | int rv; | |
1870 | ||
1871 | if (!clock_monotonic_available) | |
1872 | return; | |
1873 | ||
1874 | /* | |
1875 | * This is third time-tracking procedure (after update_times() above and | |
1876 | * times_update() in BFD), dedicated to internal event log and latency | |
1877 | * tracking. Hopefully, we consolidate these sometimes. | |
1878 | */ | |
1879 | ||
1880 | rv = clock_gettime(CLOCK_MONOTONIC, &ts); | |
1881 | if (rv < 0) | |
1882 | die("clock_gettime: %m"); | |
1883 | ||
1884 | last_time = ((s64) ts.tv_sec S) + (ts.tv_nsec / 1000); | |
1885 | ||
1886 | if (event_open) | |
1887 | { | |
1888 | event_open->duration = last_time - event_open->timestamp; | |
1889 | ||
1890 | if (event_open->duration > config->latency_limit) | |
1891 | log(L_WARN "Event 0x%p 0x%p took %d ms", | |
1892 | event_open->hook, event_open->data, (int) (event_open->duration TO_MS)); | |
1893 | ||
1894 | event_open = NULL; | |
1895 | } | |
1896 | } | |
1897 | ||
1898 | /** | |
1899 | * io_log_event - mark approaching event into event log | |
1900 | * @hook: event hook address | |
1901 | * @data: event data address | |
1902 | * | |
1903 | * Store info (hook, data, timestamp) about the following internal event into | |
1904 | * a circular event log (@event_log). When latency tracking is enabled, the log | |
1905 | * entry is kept open (in @event_open) so the duration can be filled later. | |
1906 | */ | |
1907 | void | |
1908 | io_log_event(void *hook, void *data) | |
1909 | { | |
1910 | if (config->latency_debug) | |
1911 | io_update_time(); | |
1912 | ||
1913 | struct event_log_entry *en = event_log + event_log_pos; | |
1914 | ||
1915 | en->hook = hook; | |
1916 | en->data = data; | |
1917 | en->timestamp = last_time; | |
1918 | en->duration = 0; | |
1919 | ||
1920 | event_log_num++; | |
1921 | event_log_pos++; | |
1922 | event_log_pos %= EVENT_LOG_LENGTH; | |
1923 | ||
1924 | event_open = config->latency_debug ? en : NULL; | |
1925 | } | |
1926 | ||
1927 | static inline void | |
1928 | io_close_event(void) | |
1929 | { | |
1930 | if (event_open) | |
1931 | io_update_time(); | |
1932 | } | |
1933 | ||
1934 | void | |
1935 | io_log_dump(void) | |
1936 | { | |
1937 | int i; | |
1938 | ||
1939 | log(L_DEBUG "Event log:"); | |
1940 | for (i = 0; i < EVENT_LOG_LENGTH; i++) | |
1941 | { | |
1942 | struct event_log_entry *en = event_log + (event_log_pos + i) % EVENT_LOG_LENGTH; | |
1943 | if (en->hook) | |
1944 | log(L_DEBUG " Event 0x%p 0x%p at %8d for %d ms", en->hook, en->data, | |
1945 | (int) ((last_time - en->timestamp) TO_MS), (int) (en->duration TO_MS)); | |
1946 | } | |
1947 | } | |
1948 | ||
1949 | void | |
1950 | watchdog_sigalrm(int sig UNUSED) | |
1951 | { | |
1952 | /* Update last_time and duration, but skip latency check */ | |
1953 | config->latency_limit = 0xffffffff; | |
1954 | io_update_time(); | |
1955 | ||
1956 | /* We want core dump */ | |
1957 | abort(); | |
1958 | } | |
1959 | ||
1960 | static inline void | |
1961 | watchdog_start1(void) | |
1962 | { | |
1963 | io_update_time(); | |
1964 | ||
1965 | loop_time = last_time; | |
1966 | } | |
1967 | ||
1968 | static inline void | |
1969 | watchdog_start(void) | |
1970 | { | |
1971 | io_update_time(); | |
1972 | ||
1973 | loop_time = last_time; | |
1974 | event_log_num = 0; | |
1975 | ||
1976 | if (config->watchdog_timeout) | |
1977 | { | |
1978 | alarm(config->watchdog_timeout); | |
1979 | watchdog_active = 1; | |
1980 | } | |
1981 | } | |
1982 | ||
1983 | static inline void | |
1984 | watchdog_stop(void) | |
1985 | { | |
1986 | io_update_time(); | |
1987 | ||
1988 | if (watchdog_active) | |
1989 | { | |
1990 | alarm(0); | |
1991 | watchdog_active = 0; | |
1992 | } | |
1993 | ||
1994 | btime duration = last_time - loop_time; | |
1995 | if (duration > config->watchdog_warning) | |
1996 | log(L_WARN "I/O loop cycle took %d ms for %d events", | |
1997 | (int) (duration TO_MS), event_log_num); | |
1998 | } | |
1999 | ||
2000 | ||
b5d9ee5c MM |
2001 | /* |
2002 | * Main I/O Loop | |
2003 | */ | |
2004 | ||
4c9dd1e4 MM |
2005 | volatile int async_config_flag; /* Asynchronous reconfiguration/dump scheduled */ |
2006 | volatile int async_dump_flag; | |
2007 | ||
b5d9ee5c MM |
2008 | void |
2009 | io_init(void) | |
2010 | { | |
2011 | init_list(&near_timers); | |
2012 | init_list(&far_timers); | |
2013 | init_list(&sock_list); | |
e8f73195 | 2014 | init_list(&global_event_list); |
7e5f5ffd | 2015 | krt_io_init(); |
fd91ae33 OZ |
2016 | init_times(); |
2017 | update_times(); | |
a92cf57d | 2018 | boot_time = now; |
fd91ae33 | 2019 | srandom((int) now_real); |
b5d9ee5c MM |
2020 | } |
2021 | ||
ea89da38 OZ |
2022 | static int short_loops = 0; |
2023 | #define SHORT_LOOP_MAX 10 | |
2024 | ||
b5d9ee5c MM |
2025 | void |
2026 | io_loop(void) | |
2027 | { | |
2028 | fd_set rd, wr; | |
2029 | struct timeval timo; | |
2030 | time_t tout; | |
30770df2 | 2031 | int hi, events; |
b5d9ee5c | 2032 | sock *s; |
38a608c5 | 2033 | node *n; |
b5d9ee5c | 2034 | |
8bcb5fb1 | 2035 | watchdog_start1(); |
38a608c5 | 2036 | sock_recalc_fdsets_p = 1; |
b5d9ee5c MM |
2037 | for(;;) |
2038 | { | |
30770df2 | 2039 | events = ev_run_list(&global_event_list); |
fd91ae33 | 2040 | update_times(); |
b5d9ee5c MM |
2041 | tout = tm_first_shot(); |
2042 | if (tout <= now) | |
2043 | { | |
2044 | tm_shot(); | |
2045 | continue; | |
2046 | } | |
a92cf57d | 2047 | timo.tv_sec = events ? 0 : MIN(tout - now, 3); |
30770df2 | 2048 | timo.tv_usec = 0; |
b5d9ee5c | 2049 | |
8bcb5fb1 OZ |
2050 | io_close_event(); |
2051 | ||
38a608c5 MM |
2052 | if (sock_recalc_fdsets_p) |
2053 | { | |
2054 | sock_recalc_fdsets_p = 0; | |
2055 | FD_ZERO(&rd); | |
2056 | FD_ZERO(&wr); | |
2057 | } | |
2058 | ||
b5d9ee5c MM |
2059 | hi = 0; |
2060 | WALK_LIST(n, sock_list) | |
2061 | { | |
2062 | s = SKIP_BACK(sock, n, n); | |
2063 | if (s->rx_hook) | |
2064 | { | |
2065 | FD_SET(s->fd, &rd); | |
2066 | if (s->fd > hi) | |
2067 | hi = s->fd; | |
2068 | } | |
38a608c5 MM |
2069 | else |
2070 | FD_CLR(s->fd, &rd); | |
b5d9ee5c MM |
2071 | if (s->tx_hook && s->ttx != s->tpos) |
2072 | { | |
2073 | FD_SET(s->fd, &wr); | |
2074 | if (s->fd > hi) | |
2075 | hi = s->fd; | |
2076 | } | |
38a608c5 MM |
2077 | else |
2078 | FD_CLR(s->fd, &wr); | |
b5d9ee5c MM |
2079 | } |
2080 | ||
4c9dd1e4 MM |
2081 | /* |
2082 | * Yes, this is racy. But even if the signal comes before this test | |
2083 | * and entering select(), it gets caught on the next timer tick. | |
2084 | */ | |
2085 | ||
2086 | if (async_config_flag) | |
2087 | { | |
8bcb5fb1 | 2088 | io_log_event(async_config, NULL); |
4c9dd1e4 MM |
2089 | async_config(); |
2090 | async_config_flag = 0; | |
f4aabcee | 2091 | continue; |
4c9dd1e4 MM |
2092 | } |
2093 | if (async_dump_flag) | |
2094 | { | |
8bcb5fb1 | 2095 | io_log_event(async_dump, NULL); |
4c9dd1e4 MM |
2096 | async_dump(); |
2097 | async_dump_flag = 0; | |
f4aabcee MM |
2098 | continue; |
2099 | } | |
2100 | if (async_shutdown_flag) | |
2101 | { | |
8bcb5fb1 | 2102 | io_log_event(async_shutdown, NULL); |
f4aabcee MM |
2103 | async_shutdown(); |
2104 | async_shutdown_flag = 0; | |
2105 | continue; | |
4c9dd1e4 MM |
2106 | } |
2107 | ||
2108 | /* And finally enter select() to find active sockets */ | |
8bcb5fb1 | 2109 | watchdog_stop(); |
b5d9ee5c | 2110 | hi = select(hi+1, &rd, &wr, NULL, &timo); |
8bcb5fb1 | 2111 | watchdog_start(); |
ea89da38 | 2112 | |
b5d9ee5c MM |
2113 | if (hi < 0) |
2114 | { | |
2115 | if (errno == EINTR || errno == EAGAIN) | |
2116 | continue; | |
2117 | die("select: %m"); | |
2118 | } | |
2119 | if (hi) | |
2120 | { | |
ea89da38 OZ |
2121 | /* guaranteed to be non-empty */ |
2122 | current_sock = SKIP_BACK(sock, n, HEAD(sock_list)); | |
2123 | ||
38a608c5 | 2124 | while (current_sock) |
b5d9ee5c | 2125 | { |
38a608c5 MM |
2126 | sock *s = current_sock; |
2127 | int e; | |
ea89da38 OZ |
2128 | int steps; |
2129 | ||
2130 | steps = MAX_STEPS; | |
2131 | if ((s->type >= SK_MAGIC) && FD_ISSET(s->fd, &rd) && s->rx_hook) | |
38a608c5 MM |
2132 | do |
2133 | { | |
4323099d | 2134 | steps--; |
8bcb5fb1 | 2135 | io_log_event(s->rx_hook, s->data); |
38a608c5 MM |
2136 | e = sk_read(s); |
2137 | if (s != current_sock) | |
2138 | goto next; | |
2139 | } | |
4323099d OZ |
2140 | while (e && s->rx_hook && steps); |
2141 | ||
2142 | steps = MAX_STEPS; | |
38a608c5 MM |
2143 | if (FD_ISSET(s->fd, &wr)) |
2144 | do | |
2145 | { | |
4323099d | 2146 | steps--; |
8bcb5fb1 | 2147 | io_log_event(s->tx_hook, s->data); |
38a608c5 MM |
2148 | e = sk_write(s); |
2149 | if (s != current_sock) | |
2150 | goto next; | |
2151 | } | |
4323099d | 2152 | while (e && steps); |
38a608c5 MM |
2153 | current_sock = sk_next(s); |
2154 | next: ; | |
b5d9ee5c | 2155 | } |
ea89da38 OZ |
2156 | |
2157 | short_loops++; | |
2158 | if (events && (short_loops < SHORT_LOOP_MAX)) | |
2159 | continue; | |
2160 | short_loops = 0; | |
2161 | ||
2162 | int count = 0; | |
2163 | current_sock = stored_sock; | |
2164 | if (current_sock == NULL) | |
2165 | current_sock = SKIP_BACK(sock, n, HEAD(sock_list)); | |
2166 | ||
2167 | while (current_sock && count < MAX_RX_STEPS) | |
2168 | { | |
2169 | sock *s = current_sock; | |
0479b443 | 2170 | int e UNUSED; |
ea89da38 OZ |
2171 | |
2172 | if ((s->type < SK_MAGIC) && FD_ISSET(s->fd, &rd) && s->rx_hook) | |
2173 | { | |
2174 | count++; | |
8bcb5fb1 | 2175 | io_log_event(s->rx_hook, s->data); |
ea89da38 OZ |
2176 | e = sk_read(s); |
2177 | if (s != current_sock) | |
2178 | goto next2; | |
2179 | } | |
2180 | current_sock = sk_next(s); | |
2181 | next2: ; | |
2182 | } | |
2183 | ||
2184 | stored_sock = current_sock; | |
b5d9ee5c MM |
2185 | } |
2186 | } | |
2187 | } | |
41c8976e OF |
2188 | |
2189 | void | |
2190 | test_old_bird(char *path) | |
2191 | { | |
2192 | int fd; | |
2193 | struct sockaddr_un sa; | |
2194 | ||
2195 | fd = socket(AF_UNIX, SOCK_STREAM, 0); | |
41c8976e OF |
2196 | if (fd < 0) |
2197 | die("Cannot create socket: %m"); | |
97e46d28 OZ |
2198 | if (strlen(path) >= sizeof(sa.sun_path)) |
2199 | die("Socket path too long"); | |
41c8976e OF |
2200 | bzero(&sa, sizeof(sa)); |
2201 | sa.sun_family = AF_UNIX; | |
2202 | strcpy(sa.sun_path, path); | |
2203 | if (connect(fd, (struct sockaddr *) &sa, SUN_LEN(&sa)) == 0) | |
2204 | die("I found another BIRD running."); | |
2205 | close(fd); | |
2206 | } |