blob: f39ec1efc7277569d78e4e1c9387e80e3fec22e0 [file] [log] [blame]
Willy Tarreaubaaee002006-06-26 02:48:02 +02001/*
2 * General purpose functions.
3 *
Willy Tarreau348238b2010-01-18 15:05:57 +01004 * Copyright 2000-2010 Willy Tarreau <w@1wt.eu>
Willy Tarreaubaaee002006-06-26 02:48:02 +02005 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 *
11 */
12
Baruch Siache1651b22020-07-24 07:52:20 +030013#if (defined(__ELF__) && !defined(__linux__)) || defined(USE_DL)
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +010014#define _GNU_SOURCE
15#include <dlfcn.h>
16#include <link.h>
17#endif
18
Willy Tarreau2e74c3f2007-12-02 18:45:09 +010019#include <ctype.h>
Willy Tarreau16e01562016-08-09 16:46:18 +020020#include <errno.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020021#include <netdb.h>
Willy Tarreau9a7bea52012-04-27 11:16:50 +020022#include <stdarg.h>
Willy Tarreaudd2f85e2012-09-02 22:34:23 +020023#include <stdio.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020024#include <stdlib.h>
25#include <string.h>
Thierry Fournier93127942016-01-20 18:49:45 +010026#include <time.h>
Willy Tarreau16e01562016-08-09 16:46:18 +020027#include <unistd.h>
Willy Tarreau127f9662007-12-06 00:53:51 +010028#include <sys/socket.h>
Willy Tarreau37101052019-05-20 16:48:20 +020029#include <sys/stat.h>
30#include <sys/types.h>
Willy Tarreau127f9662007-12-06 00:53:51 +010031#include <sys/un.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020032#include <netinet/in.h>
33#include <arpa/inet.h>
34
Willy Tarreau30053062020-08-20 16:39:14 +020035#if (__GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 16))
36#include <sys/auxv.h>
37#endif
38
Willy Tarreau48fbcae2020-06-03 18:09:46 +020039#include <import/eb32sctree.h>
Willy Tarreaub2551052020-06-09 09:07:15 +020040#include <import/eb32tree.h>
Willy Tarreau48fbcae2020-06-03 18:09:46 +020041
Willy Tarreau4c7e4b72020-05-27 12:58:42 +020042#include <haproxy/api.h>
Willy Tarreauc13ed532020-06-02 10:22:45 +020043#include <haproxy/chunk.h>
Willy Tarreau7c18b542020-06-11 09:23:02 +020044#include <haproxy/dgram.h>
Willy Tarreauf268ee82020-06-04 17:05:57 +020045#include <haproxy/global.h>
Willy Tarreau86416052020-06-04 09:20:54 +020046#include <haproxy/hlua.h>
Willy Tarreau213e9902020-06-04 14:58:24 +020047#include <haproxy/listener.h>
Willy Tarreau7a00efb2020-06-02 17:02:59 +020048#include <haproxy/namespace.h>
Christopher Faulet9553de72021-02-26 09:12:50 +010049#include <haproxy/net_helper.h>
Willy Tarreau5fc93282020-09-16 18:25:03 +020050#include <haproxy/protocol.h>
Emeric Brunc9437992021-02-12 19:42:55 +010051#include <haproxy/resolvers.h>
Willy Tarreau586f71b2020-12-11 15:54:36 +010052#include <haproxy/sock.h>
Willy Tarreau209108d2020-06-04 20:30:20 +020053#include <haproxy/ssl_sock.h>
Willy Tarreau5e539c92020-06-04 20:45:39 +020054#include <haproxy/stream_interface.h>
Willy Tarreaucea0e1b2020-06-04 17:25:40 +020055#include <haproxy/task.h>
Willy Tarreau48fbcae2020-06-03 18:09:46 +020056#include <haproxy/tools.h>
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +010057
Thierry Fournier93127942016-01-20 18:49:45 +010058/* This macro returns false if the test __x is false. Many
59 * of the following parsing function must be abort the processing
60 * if it returns 0, so this macro is useful for writing light code.
61 */
62#define RET0_UNLESS(__x) do { if (!(__x)) return 0; } while (0)
63
Willy Tarreau56adcf22012-12-23 18:00:29 +010064/* enough to store NB_ITOA_STR integers of :
Willy Tarreau72d759c2007-10-25 12:14:10 +020065 * 2^64-1 = 18446744073709551615 or
66 * -2^63 = -9223372036854775808
Willy Tarreaue7239b52009-03-29 13:41:58 +020067 *
68 * The HTML version needs room for adding the 25 characters
69 * '<span class="rls"></span>' around digits at positions 3N+1 in order
70 * to add spacing at up to 6 positions : 18 446 744 073 709 551 615
Willy Tarreau72d759c2007-10-25 12:14:10 +020071 */
Christopher Faulet99bca652017-11-14 16:47:26 +010072THREAD_LOCAL char itoa_str[NB_ITOA_STR][171];
73THREAD_LOCAL int itoa_idx = 0; /* index of next itoa_str to use */
Willy Tarreaubaaee002006-06-26 02:48:02 +020074
Willy Tarreau588297f2014-06-16 15:16:40 +020075/* sometimes we'll need to quote strings (eg: in stats), and we don't expect
76 * to quote strings larger than a max configuration line.
77 */
Christopher Faulet99bca652017-11-14 16:47:26 +010078THREAD_LOCAL char quoted_str[NB_QSTR][QSTR_SIZE + 1];
79THREAD_LOCAL int quoted_idx = 0;
Willy Tarreau588297f2014-06-16 15:16:40 +020080
Willy Tarreau06e69b52021-03-02 14:01:35 +010081/* thread-local PRNG state. It's modified to start from a different sequence
82 * on all threads upon startup. It must not be used or anything beyond getting
83 * statistical values as it's 100% predictable.
84 */
85THREAD_LOCAL unsigned int statistical_prng_state = 2463534242U;
86
Willy Tarreaubaaee002006-06-26 02:48:02 +020087/*
William Lallemande7340ec2012-01-24 11:15:39 +010088 * unsigned long long ASCII representation
89 *
90 * return the last char '\0' or NULL if no enough
91 * space in dst
92 */
93char *ulltoa(unsigned long long n, char *dst, size_t size)
94{
95 int i = 0;
96 char *res;
97
98 switch(n) {
99 case 1ULL ... 9ULL:
100 i = 0;
101 break;
102
103 case 10ULL ... 99ULL:
104 i = 1;
105 break;
106
107 case 100ULL ... 999ULL:
108 i = 2;
109 break;
110
111 case 1000ULL ... 9999ULL:
112 i = 3;
113 break;
114
115 case 10000ULL ... 99999ULL:
116 i = 4;
117 break;
118
119 case 100000ULL ... 999999ULL:
120 i = 5;
121 break;
122
123 case 1000000ULL ... 9999999ULL:
124 i = 6;
125 break;
126
127 case 10000000ULL ... 99999999ULL:
128 i = 7;
129 break;
130
131 case 100000000ULL ... 999999999ULL:
132 i = 8;
133 break;
134
135 case 1000000000ULL ... 9999999999ULL:
136 i = 9;
137 break;
138
139 case 10000000000ULL ... 99999999999ULL:
140 i = 10;
141 break;
142
143 case 100000000000ULL ... 999999999999ULL:
144 i = 11;
145 break;
146
147 case 1000000000000ULL ... 9999999999999ULL:
148 i = 12;
149 break;
150
151 case 10000000000000ULL ... 99999999999999ULL:
152 i = 13;
153 break;
154
155 case 100000000000000ULL ... 999999999999999ULL:
156 i = 14;
157 break;
158
159 case 1000000000000000ULL ... 9999999999999999ULL:
160 i = 15;
161 break;
162
163 case 10000000000000000ULL ... 99999999999999999ULL:
164 i = 16;
165 break;
166
167 case 100000000000000000ULL ... 999999999999999999ULL:
168 i = 17;
169 break;
170
171 case 1000000000000000000ULL ... 9999999999999999999ULL:
172 i = 18;
173 break;
174
175 case 10000000000000000000ULL ... ULLONG_MAX:
176 i = 19;
177 break;
178 }
179 if (i + 2 > size) // (i + 1) + '\0'
180 return NULL; // too long
181 res = dst + i + 1;
182 *res = '\0';
183 for (; i >= 0; i--) {
184 dst[i] = n % 10ULL + '0';
185 n /= 10ULL;
186 }
187 return res;
188}
189
190/*
191 * unsigned long ASCII representation
192 *
193 * return the last char '\0' or NULL if no enough
194 * space in dst
195 */
196char *ultoa_o(unsigned long n, char *dst, size_t size)
197{
198 int i = 0;
199 char *res;
200
201 switch (n) {
202 case 0U ... 9UL:
203 i = 0;
204 break;
205
206 case 10U ... 99UL:
207 i = 1;
208 break;
209
210 case 100U ... 999UL:
211 i = 2;
212 break;
213
214 case 1000U ... 9999UL:
215 i = 3;
216 break;
217
218 case 10000U ... 99999UL:
219 i = 4;
220 break;
221
222 case 100000U ... 999999UL:
223 i = 5;
224 break;
225
226 case 1000000U ... 9999999UL:
227 i = 6;
228 break;
229
230 case 10000000U ... 99999999UL:
231 i = 7;
232 break;
233
234 case 100000000U ... 999999999UL:
235 i = 8;
236 break;
237#if __WORDSIZE == 32
238
239 case 1000000000ULL ... ULONG_MAX:
240 i = 9;
241 break;
242
243#elif __WORDSIZE == 64
244
245 case 1000000000ULL ... 9999999999UL:
246 i = 9;
247 break;
248
249 case 10000000000ULL ... 99999999999UL:
250 i = 10;
251 break;
252
253 case 100000000000ULL ... 999999999999UL:
254 i = 11;
255 break;
256
257 case 1000000000000ULL ... 9999999999999UL:
258 i = 12;
259 break;
260
261 case 10000000000000ULL ... 99999999999999UL:
262 i = 13;
263 break;
264
265 case 100000000000000ULL ... 999999999999999UL:
266 i = 14;
267 break;
268
269 case 1000000000000000ULL ... 9999999999999999UL:
270 i = 15;
271 break;
272
273 case 10000000000000000ULL ... 99999999999999999UL:
274 i = 16;
275 break;
276
277 case 100000000000000000ULL ... 999999999999999999UL:
278 i = 17;
279 break;
280
281 case 1000000000000000000ULL ... 9999999999999999999UL:
282 i = 18;
283 break;
284
285 case 10000000000000000000ULL ... ULONG_MAX:
286 i = 19;
287 break;
288
289#endif
290 }
291 if (i + 2 > size) // (i + 1) + '\0'
292 return NULL; // too long
293 res = dst + i + 1;
294 *res = '\0';
295 for (; i >= 0; i--) {
296 dst[i] = n % 10U + '0';
297 n /= 10U;
298 }
299 return res;
300}
301
302/*
303 * signed long ASCII representation
304 *
305 * return the last char '\0' or NULL if no enough
306 * space in dst
307 */
308char *ltoa_o(long int n, char *dst, size_t size)
309{
310 char *pos = dst;
311
312 if (n < 0) {
313 if (size < 3)
314 return NULL; // min size is '-' + digit + '\0' but another test in ultoa
315 *pos = '-';
316 pos++;
317 dst = ultoa_o(-n, pos, size - 1);
318 } else {
319 dst = ultoa_o(n, dst, size);
320 }
321 return dst;
322}
323
324/*
325 * signed long long ASCII representation
326 *
327 * return the last char '\0' or NULL if no enough
328 * space in dst
329 */
330char *lltoa(long long n, char *dst, size_t size)
331{
332 char *pos = dst;
333
334 if (n < 0) {
335 if (size < 3)
336 return NULL; // min size is '-' + digit + '\0' but another test in ulltoa
337 *pos = '-';
338 pos++;
339 dst = ulltoa(-n, pos, size - 1);
340 } else {
341 dst = ulltoa(n, dst, size);
342 }
343 return dst;
344}
345
346/*
347 * write a ascii representation of a unsigned into dst,
348 * return a pointer to the last character
349 * Pad the ascii representation with '0', using size.
350 */
351char *utoa_pad(unsigned int n, char *dst, size_t size)
352{
353 int i = 0;
354 char *ret;
355
356 switch(n) {
357 case 0U ... 9U:
358 i = 0;
359 break;
360
361 case 10U ... 99U:
362 i = 1;
363 break;
364
365 case 100U ... 999U:
366 i = 2;
367 break;
368
369 case 1000U ... 9999U:
370 i = 3;
371 break;
372
373 case 10000U ... 99999U:
374 i = 4;
375 break;
376
377 case 100000U ... 999999U:
378 i = 5;
379 break;
380
381 case 1000000U ... 9999999U:
382 i = 6;
383 break;
384
385 case 10000000U ... 99999999U:
386 i = 7;
387 break;
388
389 case 100000000U ... 999999999U:
390 i = 8;
391 break;
392
393 case 1000000000U ... 4294967295U:
394 i = 9;
395 break;
396 }
397 if (i + 2 > size) // (i + 1) + '\0'
398 return NULL; // too long
399 if (i < size)
400 i = size - 2; // padding - '\0'
401
402 ret = dst + i + 1;
403 *ret = '\0';
404 for (; i >= 0; i--) {
405 dst[i] = n % 10U + '0';
406 n /= 10U;
407 }
408 return ret;
409}
410
411/*
Willy Tarreaubaaee002006-06-26 02:48:02 +0200412 * copies at most <size-1> chars from <src> to <dst>. Last char is always
413 * set to 0, unless <size> is 0. The number of chars copied is returned
414 * (excluding the terminating zero).
415 * This code has been optimized for size and speed : on x86, it's 45 bytes
416 * long, uses only registers, and consumes only 4 cycles per char.
417 */
418int strlcpy2(char *dst, const char *src, int size)
419{
420 char *orig = dst;
421 if (size) {
422 while (--size && (*dst = *src)) {
423 src++; dst++;
424 }
425 *dst = 0;
426 }
427 return dst - orig;
428}
429
430/*
Willy Tarreau72d759c2007-10-25 12:14:10 +0200431 * This function simply returns a locally allocated string containing
Willy Tarreaubaaee002006-06-26 02:48:02 +0200432 * the ascii representation for number 'n' in decimal.
433 */
Emeric Brun3a7fce52010-01-04 14:54:38 +0100434char *ultoa_r(unsigned long n, char *buffer, int size)
Willy Tarreaubaaee002006-06-26 02:48:02 +0200435{
436 char *pos;
437
Willy Tarreau72d759c2007-10-25 12:14:10 +0200438 pos = buffer + size - 1;
Willy Tarreaubaaee002006-06-26 02:48:02 +0200439 *pos-- = '\0';
440
441 do {
442 *pos-- = '0' + n % 10;
443 n /= 10;
Willy Tarreau72d759c2007-10-25 12:14:10 +0200444 } while (n && pos >= buffer);
Willy Tarreaubaaee002006-06-26 02:48:02 +0200445 return pos + 1;
446}
447
Willy Tarreau91092e52007-10-25 16:58:42 +0200448/*
Willy Tarreaue7239b52009-03-29 13:41:58 +0200449 * This function simply returns a locally allocated string containing
Thierry FOURNIER763a5d82015-07-06 23:09:52 +0200450 * the ascii representation for number 'n' in decimal.
451 */
452char *lltoa_r(long long int in, char *buffer, int size)
453{
454 char *pos;
455 int neg = 0;
456 unsigned long long int n;
457
458 pos = buffer + size - 1;
459 *pos-- = '\0';
460
461 if (in < 0) {
462 neg = 1;
463 n = -in;
464 }
465 else
466 n = in;
467
468 do {
469 *pos-- = '0' + n % 10;
470 n /= 10;
471 } while (n && pos >= buffer);
472 if (neg && pos > buffer)
473 *pos-- = '-';
474 return pos + 1;
475}
476
477/*
478 * This function simply returns a locally allocated string containing
Thierry FOURNIER1480bd82015-06-06 19:14:59 +0200479 * the ascii representation for signed number 'n' in decimal.
480 */
481char *sltoa_r(long n, char *buffer, int size)
482{
483 char *pos;
484
485 if (n >= 0)
486 return ultoa_r(n, buffer, size);
487
488 pos = ultoa_r(-n, buffer + 1, size - 1) - 1;
489 *pos = '-';
490 return pos;
491}
492
493/*
494 * This function simply returns a locally allocated string containing
Willy Tarreaue7239b52009-03-29 13:41:58 +0200495 * the ascii representation for number 'n' in decimal, formatted for
496 * HTML output with tags to create visual grouping by 3 digits. The
497 * output needs to support at least 171 characters.
498 */
499const char *ulltoh_r(unsigned long long n, char *buffer, int size)
500{
501 char *start;
502 int digit = 0;
503
504 start = buffer + size;
505 *--start = '\0';
506
507 do {
508 if (digit == 3 && start >= buffer + 7)
509 memcpy(start -= 7, "</span>", 7);
510
511 if (start >= buffer + 1) {
512 *--start = '0' + n % 10;
513 n /= 10;
514 }
515
516 if (digit == 3 && start >= buffer + 18)
517 memcpy(start -= 18, "<span class=\"rls\">", 18);
518
519 if (digit++ == 3)
520 digit = 1;
521 } while (n && start > buffer);
522 return start;
523}
524
525/*
Willy Tarreau91092e52007-10-25 16:58:42 +0200526 * This function simply returns a locally allocated string containing the ascii
527 * representation for number 'n' in decimal, unless n is 0 in which case it
528 * returns the alternate string (or an empty string if the alternate string is
529 * NULL). It use is intended for limits reported in reports, where it's
530 * desirable not to display anything if there is no limit. Warning! it shares
531 * the same vector as ultoa_r().
532 */
533const char *limit_r(unsigned long n, char *buffer, int size, const char *alt)
534{
535 return (n) ? ultoa_r(n, buffer, size) : (alt ? alt : "");
536}
537
Willy Tarreau588297f2014-06-16 15:16:40 +0200538/* returns a locally allocated string containing the quoted encoding of the
539 * input string. The output may be truncated to QSTR_SIZE chars, but it is
540 * guaranteed that the string will always be properly terminated. Quotes are
541 * encoded by doubling them as is commonly done in CSV files. QSTR_SIZE must
542 * always be at least 4 chars.
543 */
544const char *qstr(const char *str)
545{
546 char *ret = quoted_str[quoted_idx];
547 char *p, *end;
548
549 if (++quoted_idx >= NB_QSTR)
550 quoted_idx = 0;
551
552 p = ret;
553 end = ret + QSTR_SIZE;
554
555 *p++ = '"';
556
557 /* always keep 3 chars to support passing "" and the ending " */
558 while (*str && p < end - 3) {
559 if (*str == '"') {
560 *p++ = '"';
561 *p++ = '"';
562 }
563 else
564 *p++ = *str;
565 str++;
566 }
567 *p++ = '"';
568 return ret;
569}
570
Robert Tsai81ae1952007-12-05 10:47:29 +0100571/*
Willy Tarreaubaaee002006-06-26 02:48:02 +0200572 * Returns non-zero if character <s> is a hex digit (0-9, a-f, A-F), else zero.
573 *
574 * It looks like this one would be a good candidate for inlining, but this is
575 * not interesting because it around 35 bytes long and often called multiple
576 * times within the same function.
577 */
578int ishex(char s)
579{
580 s -= '0';
581 if ((unsigned char)s <= 9)
582 return 1;
583 s -= 'A' - '0';
584 if ((unsigned char)s <= 5)
585 return 1;
586 s -= 'a' - 'A';
587 if ((unsigned char)s <= 5)
588 return 1;
589 return 0;
590}
591
Willy Tarreau3ca1a882015-01-15 18:43:49 +0100592/* rounds <i> down to the closest value having max 2 digits */
593unsigned int round_2dig(unsigned int i)
594{
595 unsigned int mul = 1;
596
597 while (i >= 100) {
598 i /= 10;
599 mul *= 10;
600 }
601 return i * mul;
602}
603
Willy Tarreau2e74c3f2007-12-02 18:45:09 +0100604/*
605 * Checks <name> for invalid characters. Valid chars are [A-Za-z0-9_:.-]. If an
606 * invalid character is found, a pointer to it is returned. If everything is
607 * fine, NULL is returned.
608 */
609const char *invalid_char(const char *name)
610{
611 if (!*name)
612 return name;
613
614 while (*name) {
Willy Tarreau90807112020-02-25 08:16:33 +0100615 if (!isalnum((unsigned char)*name) && *name != '.' && *name != ':' &&
Willy Tarreau2e74c3f2007-12-02 18:45:09 +0100616 *name != '_' && *name != '-')
617 return name;
618 name++;
619 }
620 return NULL;
621}
Willy Tarreaubaaee002006-06-26 02:48:02 +0200622
623/*
Frédéric Lécailleb82f7422017-04-13 18:24:23 +0200624 * Checks <name> for invalid characters. Valid chars are [_.-] and those
625 * accepted by <f> function.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +0200626 * If an invalid character is found, a pointer to it is returned.
627 * If everything is fine, NULL is returned.
628 */
Frédéric Lécailleb82f7422017-04-13 18:24:23 +0200629static inline const char *__invalid_char(const char *name, int (*f)(int)) {
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +0200630
631 if (!*name)
632 return name;
633
634 while (*name) {
Willy Tarreau90807112020-02-25 08:16:33 +0100635 if (!f((unsigned char)*name) && *name != '.' &&
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +0200636 *name != '_' && *name != '-')
637 return name;
638
639 name++;
640 }
641
642 return NULL;
643}
644
645/*
Frédéric Lécailleb82f7422017-04-13 18:24:23 +0200646 * Checks <name> for invalid characters. Valid chars are [A-Za-z0-9_.-].
647 * If an invalid character is found, a pointer to it is returned.
648 * If everything is fine, NULL is returned.
649 */
650const char *invalid_domainchar(const char *name) {
651 return __invalid_char(name, isalnum);
652}
653
654/*
655 * Checks <name> for invalid characters. Valid chars are [A-Za-z_.-].
656 * If an invalid character is found, a pointer to it is returned.
657 * If everything is fine, NULL is returned.
658 */
659const char *invalid_prefix_char(const char *name) {
Thierry Fournierf7b7c3e2018-03-26 11:54:39 +0200660 return __invalid_char(name, isalnum);
Frédéric Lécailleb82f7422017-04-13 18:24:23 +0200661}
662
663/*
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100664 * converts <str> to a struct sockaddr_storage* provided by the caller. The
Willy Tarreau24709282013-03-10 21:32:12 +0100665 * caller must have zeroed <sa> first, and may have set sa->ss_family to force
666 * parse a specific address format. If the ss_family is 0 or AF_UNSPEC, then
667 * the function tries to guess the address family from the syntax. If the
668 * family is forced and the format doesn't match, an error is returned. The
Willy Tarreaufab5a432011-03-04 15:31:53 +0100669 * string is assumed to contain only an address, no port. The address can be a
670 * dotted IPv4 address, an IPv6 address, a host name, or empty or "*" to
671 * indicate INADDR_ANY. NULL is returned if the host part cannot be resolved.
672 * The return address will only have the address family and the address set,
673 * all other fields remain zero. The string is not supposed to be modified.
Thierry FOURNIER58639a02014-11-25 12:02:25 +0100674 * The IPv6 '::' address is IN6ADDR_ANY. If <resolve> is non-zero, the hostname
675 * is resolved, otherwise only IP addresses are resolved, and anything else
Willy Tarreauecde7df2016-11-02 22:37:03 +0100676 * returns NULL. If the address contains a port, this one is preserved.
Willy Tarreaubaaee002006-06-26 02:48:02 +0200677 */
Thierry FOURNIER58639a02014-11-25 12:02:25 +0100678struct sockaddr_storage *str2ip2(const char *str, struct sockaddr_storage *sa, int resolve)
Willy Tarreaubaaee002006-06-26 02:48:02 +0200679{
Willy Tarreaufab5a432011-03-04 15:31:53 +0100680 struct hostent *he;
mildisff5d5102015-10-26 18:50:08 +0100681 /* max IPv6 length, including brackets and terminating NULL */
682 char tmpip[48];
Willy Tarreauecde7df2016-11-02 22:37:03 +0100683 int port = get_host_port(sa);
mildisff5d5102015-10-26 18:50:08 +0100684
685 /* check IPv6 with square brackets */
686 if (str[0] == '[') {
687 size_t iplength = strlen(str);
688
689 if (iplength < 4) {
690 /* minimal size is 4 when using brackets "[::]" */
691 goto fail;
692 }
693 else if (iplength >= sizeof(tmpip)) {
694 /* IPv6 literal can not be larger than tmpip */
695 goto fail;
696 }
697 else {
698 if (str[iplength - 1] != ']') {
699 /* if address started with bracket, it should end with bracket */
700 goto fail;
701 }
702 else {
703 memcpy(tmpip, str + 1, iplength - 2);
704 tmpip[iplength - 2] = '\0';
705 str = tmpip;
706 }
707 }
708 }
Willy Tarreaufab5a432011-03-04 15:31:53 +0100709
Willy Tarreaufab5a432011-03-04 15:31:53 +0100710 /* Any IPv6 address */
711 if (str[0] == ':' && str[1] == ':' && !str[2]) {
Willy Tarreau24709282013-03-10 21:32:12 +0100712 if (!sa->ss_family || sa->ss_family == AF_UNSPEC)
713 sa->ss_family = AF_INET6;
714 else if (sa->ss_family != AF_INET6)
715 goto fail;
Willy Tarreauecde7df2016-11-02 22:37:03 +0100716 set_host_port(sa, port);
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100717 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100718 }
719
Willy Tarreau24709282013-03-10 21:32:12 +0100720 /* Any address for the family, defaults to IPv4 */
Willy Tarreaufab5a432011-03-04 15:31:53 +0100721 if (!str[0] || (str[0] == '*' && !str[1])) {
Willy Tarreau24709282013-03-10 21:32:12 +0100722 if (!sa->ss_family || sa->ss_family == AF_UNSPEC)
723 sa->ss_family = AF_INET;
Willy Tarreauecde7df2016-11-02 22:37:03 +0100724 set_host_port(sa, port);
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100725 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100726 }
727
728 /* check for IPv6 first */
Willy Tarreau24709282013-03-10 21:32:12 +0100729 if ((!sa->ss_family || sa->ss_family == AF_UNSPEC || sa->ss_family == AF_INET6) &&
730 inet_pton(AF_INET6, str, &((struct sockaddr_in6 *)sa)->sin6_addr)) {
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100731 sa->ss_family = AF_INET6;
Willy Tarreauecde7df2016-11-02 22:37:03 +0100732 set_host_port(sa, port);
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100733 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100734 }
735
736 /* then check for IPv4 */
Willy Tarreau24709282013-03-10 21:32:12 +0100737 if ((!sa->ss_family || sa->ss_family == AF_UNSPEC || sa->ss_family == AF_INET) &&
738 inet_pton(AF_INET, str, &((struct sockaddr_in *)sa)->sin_addr)) {
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100739 sa->ss_family = AF_INET;
Willy Tarreauecde7df2016-11-02 22:37:03 +0100740 set_host_port(sa, port);
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100741 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100742 }
743
Thierry FOURNIER58639a02014-11-25 12:02:25 +0100744 if (!resolve)
745 return NULL;
746
Emeric Brund30e9a12020-12-23 18:49:16 +0100747 if (!resolv_hostname_validation(str, NULL))
Baptiste Assmanna68ca962015-04-14 01:15:08 +0200748 return NULL;
749
David du Colombierd5f43282011-03-17 10:40:16 +0100750#ifdef USE_GETADDRINFO
Nenad Merdanovic88afe032014-04-14 15:56:58 +0200751 if (global.tune.options & GTUNE_USE_GAI) {
David du Colombierd5f43282011-03-17 10:40:16 +0100752 struct addrinfo hints, *result;
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100753 int success = 0;
David du Colombierd5f43282011-03-17 10:40:16 +0100754
755 memset(&result, 0, sizeof(result));
756 memset(&hints, 0, sizeof(hints));
Willy Tarreau24709282013-03-10 21:32:12 +0100757 hints.ai_family = sa->ss_family ? sa->ss_family : AF_UNSPEC;
David du Colombierd5f43282011-03-17 10:40:16 +0100758 hints.ai_socktype = SOCK_DGRAM;
Dmitry Sivachenkoeab7f392015-10-02 01:01:58 +0200759 hints.ai_flags = 0;
David du Colombierd5f43282011-03-17 10:40:16 +0100760 hints.ai_protocol = 0;
761
762 if (getaddrinfo(str, NULL, &hints, &result) == 0) {
Willy Tarreau24709282013-03-10 21:32:12 +0100763 if (!sa->ss_family || sa->ss_family == AF_UNSPEC)
764 sa->ss_family = result->ai_family;
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100765 else if (sa->ss_family != result->ai_family) {
766 freeaddrinfo(result);
Willy Tarreau24709282013-03-10 21:32:12 +0100767 goto fail;
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100768 }
Willy Tarreau24709282013-03-10 21:32:12 +0100769
David du Colombierd5f43282011-03-17 10:40:16 +0100770 switch (result->ai_family) {
771 case AF_INET:
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100772 memcpy((struct sockaddr_in *)sa, result->ai_addr, result->ai_addrlen);
Willy Tarreauecde7df2016-11-02 22:37:03 +0100773 set_host_port(sa, port);
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100774 success = 1;
775 break;
David du Colombierd5f43282011-03-17 10:40:16 +0100776 case AF_INET6:
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100777 memcpy((struct sockaddr_in6 *)sa, result->ai_addr, result->ai_addrlen);
Willy Tarreauecde7df2016-11-02 22:37:03 +0100778 set_host_port(sa, port);
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100779 success = 1;
780 break;
David du Colombierd5f43282011-03-17 10:40:16 +0100781 }
782 }
783
Sean Carey58ea0392013-02-15 23:39:18 +0100784 if (result)
785 freeaddrinfo(result);
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100786
787 if (success)
788 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100789 }
David du Colombierd5f43282011-03-17 10:40:16 +0100790#endif
Nenad Merdanovic88afe032014-04-14 15:56:58 +0200791 /* try to resolve an IPv4/IPv6 hostname */
792 he = gethostbyname(str);
793 if (he) {
794 if (!sa->ss_family || sa->ss_family == AF_UNSPEC)
795 sa->ss_family = he->h_addrtype;
796 else if (sa->ss_family != he->h_addrtype)
797 goto fail;
798
799 switch (sa->ss_family) {
800 case AF_INET:
801 ((struct sockaddr_in *)sa)->sin_addr = *(struct in_addr *) *(he->h_addr_list);
Willy Tarreauecde7df2016-11-02 22:37:03 +0100802 set_host_port(sa, port);
Nenad Merdanovic88afe032014-04-14 15:56:58 +0200803 return sa;
804 case AF_INET6:
805 ((struct sockaddr_in6 *)sa)->sin6_addr = *(struct in6_addr *) *(he->h_addr_list);
Willy Tarreauecde7df2016-11-02 22:37:03 +0100806 set_host_port(sa, port);
Nenad Merdanovic88afe032014-04-14 15:56:58 +0200807 return sa;
808 }
809 }
810
David du Colombierd5f43282011-03-17 10:40:16 +0100811 /* unsupported address family */
Willy Tarreau24709282013-03-10 21:32:12 +0100812 fail:
Willy Tarreaufab5a432011-03-04 15:31:53 +0100813 return NULL;
814}
815
816/*
Willy Tarreaud4448bc2013-02-20 15:55:15 +0100817 * Converts <str> to a locally allocated struct sockaddr_storage *, and a port
818 * range or offset consisting in two integers that the caller will have to
819 * check to find the relevant input format. The following format are supported :
820 *
821 * String format | address | port | low | high
822 * addr | <addr> | 0 | 0 | 0
823 * addr: | <addr> | 0 | 0 | 0
824 * addr:port | <addr> | <port> | <port> | <port>
825 * addr:pl-ph | <addr> | <pl> | <pl> | <ph>
826 * addr:+port | <addr> | <port> | 0 | <port>
827 * addr:-port | <addr> |-<port> | <port> | 0
828 *
829 * The detection of a port range or increment by the caller is made by
830 * comparing <low> and <high>. If both are equal, then port 0 means no port
831 * was specified. The caller may pass NULL for <low> and <high> if it is not
832 * interested in retrieving port ranges.
833 *
834 * Note that <addr> above may also be :
835 * - empty ("") => family will be AF_INET and address will be INADDR_ANY
836 * - "*" => family will be AF_INET and address will be INADDR_ANY
837 * - "::" => family will be AF_INET6 and address will be IN6ADDR_ANY
838 * - a host name => family and address will depend on host name resolving.
839 *
Willy Tarreau24709282013-03-10 21:32:12 +0100840 * A prefix may be passed in before the address above to force the family :
841 * - "ipv4@" => force address to resolve as IPv4 and fail if not possible.
842 * - "ipv6@" => force address to resolve as IPv6 and fail if not possible.
843 * - "unix@" => force address to be a path to a UNIX socket even if the
844 * path does not start with a '/'
Willy Tarreauccfccef2014-05-10 01:49:15 +0200845 * - 'abns@' -> force address to belong to the abstract namespace (Linux
846 * only). These sockets are just like Unix sockets but without
847 * the need for an underlying file system. The address is a
848 * string. Technically it's like a Unix socket with a zero in
849 * the first byte of the address.
Willy Tarreau40aa0702013-03-10 23:51:38 +0100850 * - "fd@" => an integer must follow, and is a file descriptor number.
Willy Tarreau24709282013-03-10 21:32:12 +0100851 *
mildisff5d5102015-10-26 18:50:08 +0100852 * IPv6 addresses can be declared with or without square brackets. When using
853 * square brackets for IPv6 addresses, the port separator (colon) is optional.
854 * If not using square brackets, and in order to avoid any ambiguity with
855 * IPv6 addresses, the last colon ':' is mandatory even when no port is specified.
856 * NULL is returned if the address cannot be parsed. The <low> and <high> ports
857 * are always initialized if non-null, even for non-IP families.
Willy Tarreaud393a622013-03-04 18:22:00 +0100858 *
859 * If <pfx> is non-null, it is used as a string prefix before any path-based
860 * address (typically the path to a unix socket).
Willy Tarreau40aa0702013-03-10 23:51:38 +0100861 *
Willy Tarreau72b8c1f2015-09-08 15:50:19 +0200862 * if <fqdn> is non-null, it will be filled with :
863 * - a pointer to the FQDN of the server name to resolve if there's one, and
864 * that the caller will have to free(),
865 * - NULL if there was an explicit address that doesn't require resolution.
866 *
Willy Tarreaucd3a55912020-09-04 15:30:46 +0200867 * Hostnames are only resolved if <opts> has PA_O_RESOLVE. Otherwise <fqdn> is
868 * still honored so it is possible for the caller to know whether a resolution
869 * failed by clearing this flag and checking if <fqdn> was filled, indicating
870 * the need for a resolution.
Thierry FOURNIER7fe3be72015-09-26 20:03:36 +0200871 *
Willy Tarreau40aa0702013-03-10 23:51:38 +0100872 * When a file descriptor is passed, its value is put into the s_addr part of
Willy Tarreaua5b325f2020-09-04 16:44:20 +0200873 * the address when cast to sockaddr_in and the address family is
874 * AF_CUST_EXISTING_FD.
Willy Tarreaua93e5c72020-09-15 14:01:16 +0200875 *
Willy Tarreau5fc93282020-09-16 18:25:03 +0200876 * The matching protocol will be set into <proto> if non-null.
877 *
Willy Tarreaua93e5c72020-09-15 14:01:16 +0200878 * Any known file descriptor is also assigned to <fd> if non-null, otherwise it
879 * is forced to -1.
Willy Tarreaufab5a432011-03-04 15:31:53 +0100880 */
Willy Tarreau5fc93282020-09-16 18:25:03 +0200881struct sockaddr_storage *str2sa_range(const char *str, int *port, int *low, int *high, int *fd,
882 struct protocol **proto, char **err,
883 const char *pfx, char **fqdn, unsigned int opts)
Willy Tarreaufab5a432011-03-04 15:31:53 +0100884{
Christopher Faulet1bc04c72017-10-29 20:14:08 +0100885 static THREAD_LOCAL struct sockaddr_storage ss;
David du Colombier6f5ccb12011-03-10 22:26:24 +0100886 struct sockaddr_storage *ret = NULL;
Willy Tarreau5fc93282020-09-16 18:25:03 +0200887 struct protocol *new_proto = NULL;
Willy Tarreau24709282013-03-10 21:32:12 +0100888 char *back, *str2;
Willy Tarreaud4448bc2013-02-20 15:55:15 +0100889 char *port1, *port2;
890 int portl, porth, porta;
Willy Tarreauccfccef2014-05-10 01:49:15 +0200891 int abstract = 0;
Willy Tarreaua93e5c72020-09-15 14:01:16 +0200892 int new_fd = -1;
Willy Tarreaue835bd82020-09-16 11:35:47 +0200893 int sock_type, ctrl_type;
Willy Tarreaud4448bc2013-02-20 15:55:15 +0100894
895 portl = porth = porta = 0;
Willy Tarreau72b8c1f2015-09-08 15:50:19 +0200896 if (fqdn)
897 *fqdn = NULL;
Willy Tarreaubaaee002006-06-26 02:48:02 +0200898
Willy Tarreaudad36a32013-03-11 01:20:04 +0100899 str2 = back = env_expand(strdup(str));
Willy Tarreaudf350f12013-03-01 20:22:54 +0100900 if (str2 == NULL) {
901 memprintf(err, "out of memory in '%s'\n", __FUNCTION__);
Willy Tarreaud5191e72010-02-09 20:50:45 +0100902 goto out;
Willy Tarreaudf350f12013-03-01 20:22:54 +0100903 }
Willy Tarreaubaaee002006-06-26 02:48:02 +0200904
Willy Tarreau9f69f462015-09-08 16:01:25 +0200905 if (!*str2) {
906 memprintf(err, "'%s' resolves to an empty address (environment variable missing?)\n", str);
907 goto out;
908 }
909
Willy Tarreau24709282013-03-10 21:32:12 +0100910 memset(&ss, 0, sizeof(ss));
911
Willy Tarreaue835bd82020-09-16 11:35:47 +0200912 /* prepare the default socket types */
913 if ((opts & (PA_O_STREAM|PA_O_DGRAM)) == PA_O_DGRAM)
914 sock_type = ctrl_type = SOCK_DGRAM;
915 else
916 sock_type = ctrl_type = SOCK_STREAM;
917
918 if (strncmp(str2, "stream+", 7) == 0) {
919 str2 += 7;
920 sock_type = ctrl_type = SOCK_STREAM;
921 }
922 else if (strncmp(str2, "dgram+", 6) == 0) {
923 str2 += 6;
924 sock_type = ctrl_type = SOCK_DGRAM;
925 }
926
Willy Tarreau24709282013-03-10 21:32:12 +0100927 if (strncmp(str2, "unix@", 5) == 0) {
928 str2 += 5;
Willy Tarreauccfccef2014-05-10 01:49:15 +0200929 abstract = 0;
Willy Tarreau24709282013-03-10 21:32:12 +0100930 ss.ss_family = AF_UNIX;
931 }
Willy Tarreauccfccef2014-05-10 01:49:15 +0200932 else if (strncmp(str2, "abns@", 5) == 0) {
933 str2 += 5;
934 abstract = 1;
935 ss.ss_family = AF_UNIX;
936 }
Willy Tarreau24709282013-03-10 21:32:12 +0100937 else if (strncmp(str2, "ipv4@", 5) == 0) {
938 str2 += 5;
939 ss.ss_family = AF_INET;
940 }
941 else if (strncmp(str2, "ipv6@", 5) == 0) {
942 str2 += 5;
943 ss.ss_family = AF_INET6;
944 }
Emeric Brun3835c0d2020-07-07 09:46:09 +0200945 else if (strncmp(str2, "udp4@", 5) == 0) {
946 str2 += 5;
947 ss.ss_family = AF_INET;
Willy Tarreaue835bd82020-09-16 11:35:47 +0200948 sock_type = ctrl_type = SOCK_DGRAM;
Emeric Brun3835c0d2020-07-07 09:46:09 +0200949 }
950 else if (strncmp(str2, "udp6@", 5) == 0) {
951 str2 += 5;
952 ss.ss_family = AF_INET6;
Willy Tarreaue835bd82020-09-16 11:35:47 +0200953 sock_type = ctrl_type = SOCK_DGRAM;
Emeric Brun3835c0d2020-07-07 09:46:09 +0200954 }
955 else if (strncmp(str2, "udp@", 4) == 0) {
956 str2 += 4;
957 ss.ss_family = AF_UNSPEC;
Willy Tarreaue835bd82020-09-16 11:35:47 +0200958 sock_type = ctrl_type = SOCK_DGRAM;
Emeric Brun3835c0d2020-07-07 09:46:09 +0200959 }
Frédéric Lécaille10caf652020-11-23 11:36:57 +0100960 else if (strncmp(str2, "quic4@", 6) == 0) {
961 str2 += 6;
962 ss.ss_family = AF_INET;
963 sock_type = SOCK_DGRAM;
964 ctrl_type = SOCK_STREAM;
965 }
966 else if (strncmp(str2, "quic6@", 6) == 0) {
967 str2 += 6;
968 ss.ss_family = AF_INET6;
969 sock_type = SOCK_DGRAM;
970 ctrl_type = SOCK_STREAM;
971 }
Willy Tarreau5a7beed2020-09-04 16:54:05 +0200972 else if (strncmp(str2, "fd@", 3) == 0) {
973 str2 += 3;
974 ss.ss_family = AF_CUST_EXISTING_FD;
975 }
976 else if (strncmp(str2, "sockpair@", 9) == 0) {
977 str2 += 9;
978 ss.ss_family = AF_CUST_SOCKPAIR;
979 }
Willy Tarreau24709282013-03-10 21:32:12 +0100980 else if (*str2 == '/') {
981 ss.ss_family = AF_UNIX;
982 }
983 else
984 ss.ss_family = AF_UNSPEC;
985
Willy Tarreau5a7beed2020-09-04 16:54:05 +0200986 if (ss.ss_family == AF_CUST_SOCKPAIR) {
Willy Tarreaua215be22020-09-16 10:14:16 +0200987 struct sockaddr_storage ss2;
988 socklen_t addr_len;
William Lallemand2fe7dd02018-09-11 16:51:29 +0200989 char *endptr;
990
Willy Tarreaua93e5c72020-09-15 14:01:16 +0200991 new_fd = strtol(str2, &endptr, 10);
992 if (!*str2 || new_fd < 0 || *endptr) {
William Lallemand2fe7dd02018-09-11 16:51:29 +0200993 memprintf(err, "file descriptor '%s' is not a valid integer in '%s'\n", str2, str);
994 goto out;
995 }
Willy Tarreaua93e5c72020-09-15 14:01:16 +0200996
Willy Tarreaua215be22020-09-16 10:14:16 +0200997 /* just verify that it's a socket */
998 addr_len = sizeof(ss2);
999 if (getsockname(new_fd, (struct sockaddr *)&ss2, &addr_len) == -1) {
1000 memprintf(err, "cannot use file descriptor '%d' : %s.\n", new_fd, strerror(errno));
1001 goto out;
1002 }
1003
Willy Tarreaua93e5c72020-09-15 14:01:16 +02001004 ((struct sockaddr_in *)&ss)->sin_addr.s_addr = new_fd;
1005 ((struct sockaddr_in *)&ss)->sin_port = 0;
William Lallemand2fe7dd02018-09-11 16:51:29 +02001006 }
Willy Tarreau5a7beed2020-09-04 16:54:05 +02001007 else if (ss.ss_family == AF_CUST_EXISTING_FD) {
Willy Tarreau40aa0702013-03-10 23:51:38 +01001008 char *endptr;
1009
Willy Tarreaua93e5c72020-09-15 14:01:16 +02001010 new_fd = strtol(str2, &endptr, 10);
1011 if (!*str2 || new_fd < 0 || *endptr) {
Willy Tarreaudad36a32013-03-11 01:20:04 +01001012 memprintf(err, "file descriptor '%s' is not a valid integer in '%s'\n", str2, str);
Willy Tarreau40aa0702013-03-10 23:51:38 +01001013 goto out;
1014 }
Willy Tarreaua93e5c72020-09-15 14:01:16 +02001015
Willy Tarreau6edc7222020-09-15 17:41:56 +02001016 if (opts & PA_O_SOCKET_FD) {
1017 socklen_t addr_len;
1018 int type;
1019
1020 addr_len = sizeof(ss);
1021 if (getsockname(new_fd, (struct sockaddr *)&ss, &addr_len) == -1) {
1022 memprintf(err, "cannot use file descriptor '%d' : %s.\n", new_fd, strerror(errno));
1023 goto out;
1024 }
1025
1026 addr_len = sizeof(type);
1027 if (getsockopt(new_fd, SOL_SOCKET, SO_TYPE, &type, &addr_len) != 0 ||
Willy Tarreaue835bd82020-09-16 11:35:47 +02001028 (type == SOCK_STREAM) != (sock_type == SOCK_STREAM)) {
Willy Tarreau6edc7222020-09-15 17:41:56 +02001029 memprintf(err, "socket on file descriptor '%d' is of the wrong type.\n", new_fd);
1030 goto out;
1031 }
1032
1033 porta = portl = porth = get_host_port(&ss);
1034 } else if (opts & PA_O_RAW_FD) {
1035 ((struct sockaddr_in *)&ss)->sin_addr.s_addr = new_fd;
1036 ((struct sockaddr_in *)&ss)->sin_port = 0;
1037 } else {
1038 memprintf(err, "a file descriptor is not acceptable here in '%s'\n", str);
1039 goto out;
1040 }
Willy Tarreau40aa0702013-03-10 23:51:38 +01001041 }
1042 else if (ss.ss_family == AF_UNIX) {
Willy Tarreau8daa9202019-06-16 18:16:33 +02001043 struct sockaddr_un *un = (struct sockaddr_un *)&ss;
Willy Tarreau15586382013-03-04 19:48:14 +01001044 int prefix_path_len;
1045 int max_path_len;
Willy Tarreau94ef3f32014-04-14 14:49:00 +02001046 int adr_len;
Willy Tarreau15586382013-03-04 19:48:14 +01001047
1048 /* complete unix socket path name during startup or soft-restart is
1049 * <unix_bind_prefix><path>.<pid>.<bak|tmp>
1050 */
Willy Tarreauccfccef2014-05-10 01:49:15 +02001051 prefix_path_len = (pfx && !abstract) ? strlen(pfx) : 0;
Willy Tarreau8daa9202019-06-16 18:16:33 +02001052 max_path_len = (sizeof(un->sun_path) - 1) -
Willy Tarreau327ea5a2020-02-11 06:43:37 +01001053 (abstract ? 0 : prefix_path_len + 1 + 5 + 1 + 3);
Willy Tarreau15586382013-03-04 19:48:14 +01001054
Willy Tarreau94ef3f32014-04-14 14:49:00 +02001055 adr_len = strlen(str2);
1056 if (adr_len > max_path_len) {
Willy Tarreau15586382013-03-04 19:48:14 +01001057 memprintf(err, "socket path '%s' too long (max %d)\n", str, max_path_len);
1058 goto out;
1059 }
1060
Willy Tarreauccfccef2014-05-10 01:49:15 +02001061 /* when abstract==1, we skip the first zero and copy all bytes except the trailing zero */
Willy Tarreau8daa9202019-06-16 18:16:33 +02001062 memset(un->sun_path, 0, sizeof(un->sun_path));
Willy Tarreau94ef3f32014-04-14 14:49:00 +02001063 if (prefix_path_len)
Willy Tarreau8daa9202019-06-16 18:16:33 +02001064 memcpy(un->sun_path, pfx, prefix_path_len);
1065 memcpy(un->sun_path + prefix_path_len + abstract, str2, adr_len + 1 - abstract);
Willy Tarreau15586382013-03-04 19:48:14 +01001066 }
Willy Tarreau24709282013-03-10 21:32:12 +01001067 else { /* IPv4 and IPv6 */
mildisff5d5102015-10-26 18:50:08 +01001068 char *end = str2 + strlen(str2);
1069 char *chr;
Willy Tarreau72b8c1f2015-09-08 15:50:19 +02001070
mildisff5d5102015-10-26 18:50:08 +01001071 /* search for : or ] whatever comes first */
1072 for (chr = end-1; chr > str2; chr--) {
1073 if (*chr == ']' || *chr == ':')
1074 break;
1075 }
1076
1077 if (*chr == ':') {
1078 /* Found a colon before a closing-bracket, must be a port separator.
1079 * This guarantee backward compatibility.
1080 */
Willy Tarreau7f96a842020-09-15 11:12:44 +02001081 if (!(opts & PA_O_PORT_OK)) {
1082 memprintf(err, "port specification not permitted here in '%s'", str);
1083 goto out;
1084 }
mildisff5d5102015-10-26 18:50:08 +01001085 *chr++ = '\0';
1086 port1 = chr;
1087 }
1088 else {
1089 /* Either no colon and no closing-bracket
1090 * or directly ending with a closing-bracket.
1091 * However, no port.
1092 */
Willy Tarreau7f96a842020-09-15 11:12:44 +02001093 if (opts & PA_O_PORT_MAND) {
1094 memprintf(err, "missing port specification in '%s'", str);
1095 goto out;
1096 }
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001097 port1 = "";
mildisff5d5102015-10-26 18:50:08 +01001098 }
Willy Tarreaubaaee002006-06-26 02:48:02 +02001099
Willy Tarreau90807112020-02-25 08:16:33 +01001100 if (isdigit((unsigned char)*port1)) { /* single port or range */
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001101 port2 = strchr(port1, '-');
Willy Tarreau7f96a842020-09-15 11:12:44 +02001102 if (port2) {
1103 if (!(opts & PA_O_PORT_RANGE)) {
1104 memprintf(err, "port range not permitted here in '%s'", str);
1105 goto out;
1106 }
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001107 *port2++ = '\0';
Willy Tarreau7f96a842020-09-15 11:12:44 +02001108 }
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001109 else
1110 port2 = port1;
1111 portl = atoi(port1);
1112 porth = atoi(port2);
Willy Tarreau7f96a842020-09-15 11:12:44 +02001113
1114 if (portl < !!(opts & PA_O_PORT_MAND) || portl > 65535) {
1115 memprintf(err, "invalid port '%s'", port1);
1116 goto out;
1117 }
1118
1119 if (porth < !!(opts & PA_O_PORT_MAND) || porth > 65535) {
1120 memprintf(err, "invalid port '%s'", port2);
1121 goto out;
1122 }
1123
1124 if (portl > porth) {
1125 memprintf(err, "invalid port range '%d-%d'", portl, porth);
1126 goto out;
1127 }
1128
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001129 porta = portl;
1130 }
1131 else if (*port1 == '-') { /* negative offset */
Willy Tarreau7f96a842020-09-15 11:12:44 +02001132 if (!(opts & PA_O_PORT_OFS)) {
1133 memprintf(err, "port offset not permitted here in '%s'", str);
1134 goto out;
1135 }
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001136 portl = atoi(port1 + 1);
1137 porta = -portl;
1138 }
1139 else if (*port1 == '+') { /* positive offset */
Willy Tarreau7f96a842020-09-15 11:12:44 +02001140 if (!(opts & PA_O_PORT_OFS)) {
1141 memprintf(err, "port offset not permitted here in '%s'", str);
1142 goto out;
1143 }
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001144 porth = atoi(port1 + 1);
1145 porta = porth;
1146 }
1147 else if (*port1) { /* other any unexpected char */
Willy Tarreaudad36a32013-03-11 01:20:04 +01001148 memprintf(err, "invalid character '%c' in port number '%s' in '%s'\n", *port1, port1, str);
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001149 goto out;
1150 }
Willy Tarreau7f96a842020-09-15 11:12:44 +02001151 else if (opts & PA_O_PORT_MAND) {
1152 memprintf(err, "missing port specification in '%s'", str);
1153 goto out;
1154 }
Willy Tarreauceccdd72016-11-02 22:27:10 +01001155
1156 /* first try to parse the IP without resolving. If it fails, it
1157 * tells us we need to keep a copy of the FQDN to resolve later
1158 * and to enable DNS. In this case we can proceed if <fqdn> is
Willy Tarreaucd3a55912020-09-04 15:30:46 +02001159 * set or if PA_O_RESOLVE is set, otherwise it's an error.
Willy Tarreauceccdd72016-11-02 22:27:10 +01001160 */
1161 if (str2ip2(str2, &ss, 0) == NULL) {
Willy Tarreaucd3a55912020-09-04 15:30:46 +02001162 if ((!(opts & PA_O_RESOLVE) && !fqdn) ||
1163 ((opts & PA_O_RESOLVE) && str2ip2(str2, &ss, 1) == NULL)) {
Willy Tarreauceccdd72016-11-02 22:27:10 +01001164 memprintf(err, "invalid address: '%s' in '%s'\n", str2, str);
1165 goto out;
1166 }
Willy Tarreau72b8c1f2015-09-08 15:50:19 +02001167
Willy Tarreauceccdd72016-11-02 22:27:10 +01001168 if (fqdn) {
1169 if (str2 != back)
1170 memmove(back, str2, strlen(str2) + 1);
1171 *fqdn = back;
1172 back = NULL;
1173 }
Willy Tarreau72b8c1f2015-09-08 15:50:19 +02001174 }
Willy Tarreauceccdd72016-11-02 22:27:10 +01001175 set_host_port(&ss, porta);
Willy Tarreaue4c58c82013-03-06 15:28:17 +01001176 }
Willy Tarreaufab5a432011-03-04 15:31:53 +01001177
Willy Tarreaue835bd82020-09-16 11:35:47 +02001178 if (ctrl_type == SOCK_STREAM && !(opts & PA_O_STREAM)) {
1179 memprintf(err, "stream-type socket not acceptable in '%s'\n", str);
1180 goto out;
1181 }
1182 else if (ctrl_type == SOCK_DGRAM && !(opts & PA_O_DGRAM)) {
1183 memprintf(err, "dgram-type socket not acceptable in '%s'\n", str);
1184 goto out;
1185 }
1186
Willy Tarreau65ec4e32020-09-16 19:17:08 +02001187 if (proto || (opts & PA_O_CONNECT)) {
Willy Tarreau5fc93282020-09-16 18:25:03 +02001188 /* Note: if the caller asks for a proto, we must find one,
1189 * except if we return with an fqdn that will resolve later,
1190 * in which case the address is not known yet (this is only
1191 * for servers actually).
1192 */
Willy Tarreaub2ffc992020-09-16 21:37:31 +02001193 new_proto = protocol_lookup(ss.ss_family,
Willy Tarreauaf9609b2020-09-16 22:04:46 +02001194 sock_type == SOCK_DGRAM,
1195 ctrl_type == SOCK_DGRAM);
Willy Tarreaub2ffc992020-09-16 21:37:31 +02001196
Willy Tarreau5fc93282020-09-16 18:25:03 +02001197 if (!new_proto && (!fqdn || !*fqdn)) {
1198 memprintf(err, "unsupported protocol family %d for address '%s'", ss.ss_family, str);
1199 goto out;
1200 }
Willy Tarreau65ec4e32020-09-16 19:17:08 +02001201
1202 if ((opts & PA_O_CONNECT) && new_proto && !new_proto->connect) {
1203 memprintf(err, "connect() not supported for this protocol family %d used by address '%s'", ss.ss_family, str);
1204 goto out;
1205 }
Willy Tarreau5fc93282020-09-16 18:25:03 +02001206 }
1207
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001208 ret = &ss;
Willy Tarreaud5191e72010-02-09 20:50:45 +01001209 out:
Willy Tarreau48ef4c92017-01-06 18:32:38 +01001210 if (port)
1211 *port = porta;
Willy Tarreaud4448bc2013-02-20 15:55:15 +01001212 if (low)
1213 *low = portl;
1214 if (high)
1215 *high = porth;
Willy Tarreaua93e5c72020-09-15 14:01:16 +02001216 if (fd)
1217 *fd = new_fd;
Willy Tarreau5fc93282020-09-16 18:25:03 +02001218 if (proto)
1219 *proto = new_proto;
Willy Tarreau24709282013-03-10 21:32:12 +01001220 free(back);
Willy Tarreaud5191e72010-02-09 20:50:45 +01001221 return ret;
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02001222}
1223
Thayne McCombs92149f92020-11-20 01:28:26 -07001224/* converts <addr> and <port> into a string representation of the address and port. This is sort
1225 * of an inverse of str2sa_range, with some restrictions. The supported families are AF_INET,
1226 * AF_INET6, AF_UNIX, and AF_CUST_SOCKPAIR. If the family is unsopported NULL is returned.
1227 * If map_ports is true, then the sign of the port is included in the output, to indicate it is
1228 * relative to the incoming port. AF_INET and AF_INET6 will be in the form "<addr>:<port>".
1229 * AF_UNIX will either be just the path (if using a pathname) or "abns@<path>" if it is abstract.
1230 * AF_CUST_SOCKPAIR will be of the form "sockpair@<fd>".
1231 *
1232 * The returned char* is allocated, and it is the responsibility of the caller to free it.
1233 */
1234char * sa2str(const struct sockaddr_storage *addr, int port, int map_ports)
1235{
1236 char buffer[INET6_ADDRSTRLEN];
1237 char *out = NULL;
1238 const void *ptr;
1239 const char *path;
1240
1241 switch (addr->ss_family) {
1242 case AF_INET:
1243 ptr = &((struct sockaddr_in *)addr)->sin_addr;
1244 break;
1245 case AF_INET6:
1246 ptr = &((struct sockaddr_in6 *)addr)->sin6_addr;
1247 break;
1248 case AF_UNIX:
1249 path = ((struct sockaddr_un *)addr)->sun_path;
1250 if (path[0] == '\0') {
1251 const int max_length = sizeof(struct sockaddr_un) - offsetof(struct sockaddr_un, sun_path) - 1;
1252 return memprintf(&out, "abns@%.*s", max_length, path+1);
1253 } else {
1254 return strdup(path);
1255 }
1256 case AF_CUST_SOCKPAIR:
1257 return memprintf(&out, "sockpair@%d", ((struct sockaddr_in *)addr)->sin_addr.s_addr);
1258 default:
1259 return NULL;
1260 }
1261 inet_ntop(addr->ss_family, ptr, buffer, get_addr_len(addr));
1262 if (map_ports)
1263 return memprintf(&out, "%s:%+d", buffer, port);
1264 else
1265 return memprintf(&out, "%s:%d", buffer, port);
1266}
1267
1268
Willy Tarreau2937c0d2010-01-26 17:36:17 +01001269/* converts <str> to a struct in_addr containing a network mask. It can be
1270 * passed in dotted form (255.255.255.0) or in CIDR form (24). It returns 1
Jarno Huuskonen577d5ac2017-05-21 17:32:21 +03001271 * if the conversion succeeds otherwise zero.
Willy Tarreau2937c0d2010-01-26 17:36:17 +01001272 */
1273int str2mask(const char *str, struct in_addr *mask)
1274{
1275 if (strchr(str, '.') != NULL) { /* dotted notation */
1276 if (!inet_pton(AF_INET, str, mask))
1277 return 0;
1278 }
1279 else { /* mask length */
1280 char *err;
1281 unsigned long len = strtol(str, &err, 10);
1282
1283 if (!*str || (err && *err) || (unsigned)len > 32)
1284 return 0;
Tim Duesterhus8575f722018-01-25 16:24:48 +01001285
1286 len2mask4(len, mask);
Willy Tarreau2937c0d2010-01-26 17:36:17 +01001287 }
1288 return 1;
1289}
1290
Tim Duesterhus47185172018-01-25 16:24:49 +01001291/* converts <str> to a struct in6_addr containing a network mask. It can be
Tim Duesterhus5e642862018-02-20 17:02:18 +01001292 * passed in quadruplet form (ffff:ffff::) or in CIDR form (64). It returns 1
Tim Duesterhus47185172018-01-25 16:24:49 +01001293 * if the conversion succeeds otherwise zero.
1294 */
1295int str2mask6(const char *str, struct in6_addr *mask)
1296{
1297 if (strchr(str, ':') != NULL) { /* quadruplet notation */
1298 if (!inet_pton(AF_INET6, str, mask))
1299 return 0;
1300 }
1301 else { /* mask length */
1302 char *err;
1303 unsigned long len = strtol(str, &err, 10);
1304
1305 if (!*str || (err && *err) || (unsigned)len > 128)
1306 return 0;
1307
1308 len2mask6(len, mask);
1309 }
1310 return 1;
1311}
1312
Thierry FOURNIERb0504632013-12-14 15:39:02 +01001313/* convert <cidr> to struct in_addr <mask>. It returns 1 if the conversion
1314 * succeeds otherwise zero.
1315 */
1316int cidr2dotted(int cidr, struct in_addr *mask) {
1317
1318 if (cidr < 0 || cidr > 32)
1319 return 0;
1320
1321 mask->s_addr = cidr ? htonl(~0UL << (32 - cidr)) : 0;
1322 return 1;
1323}
1324
Thierry Fournier70473a52016-02-17 17:12:14 +01001325/* Convert mask from bit length form to in_addr form.
1326 * This function never fails.
1327 */
1328void len2mask4(int len, struct in_addr *addr)
1329{
1330 if (len >= 32) {
1331 addr->s_addr = 0xffffffff;
1332 return;
1333 }
1334 if (len <= 0) {
1335 addr->s_addr = 0x00000000;
1336 return;
1337 }
1338 addr->s_addr = 0xffffffff << (32 - len);
1339 addr->s_addr = htonl(addr->s_addr);
1340}
1341
1342/* Convert mask from bit length form to in6_addr form.
1343 * This function never fails.
1344 */
1345void len2mask6(int len, struct in6_addr *addr)
1346{
1347 len2mask4(len, (struct in_addr *)&addr->s6_addr[0]); /* msb */
1348 len -= 32;
1349 len2mask4(len, (struct in_addr *)&addr->s6_addr[4]);
1350 len -= 32;
1351 len2mask4(len, (struct in_addr *)&addr->s6_addr[8]);
1352 len -= 32;
1353 len2mask4(len, (struct in_addr *)&addr->s6_addr[12]); /* lsb */
1354}
1355
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02001356/*
Willy Tarreaud077a8e2007-05-08 18:28:09 +02001357 * converts <str> to two struct in_addr* which must be pre-allocated.
Willy Tarreaubaaee002006-06-26 02:48:02 +02001358 * The format is "addr[/mask]", where "addr" cannot be empty, and mask
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05001359 * is optional and either in the dotted or CIDR notation.
Willy Tarreaubaaee002006-06-26 02:48:02 +02001360 * Note: "addr" can also be a hostname. Returns 1 if OK, 0 if error.
1361 */
Thierry FOURNIERfc7ac7b2014-02-11 15:23:04 +01001362int str2net(const char *str, int resolve, struct in_addr *addr, struct in_addr *mask)
Willy Tarreaubaaee002006-06-26 02:48:02 +02001363{
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001364 __label__ out_free, out_err;
1365 char *c, *s;
1366 int ret_val;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001367
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001368 s = strdup(str);
1369 if (!s)
1370 return 0;
1371
Willy Tarreaubaaee002006-06-26 02:48:02 +02001372 memset(mask, 0, sizeof(*mask));
1373 memset(addr, 0, sizeof(*addr));
Willy Tarreaubaaee002006-06-26 02:48:02 +02001374
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001375 if ((c = strrchr(s, '/')) != NULL) {
Willy Tarreaubaaee002006-06-26 02:48:02 +02001376 *c++ = '\0';
1377 /* c points to the mask */
Willy Tarreau2937c0d2010-01-26 17:36:17 +01001378 if (!str2mask(c, mask))
1379 goto out_err;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001380 }
1381 else {
Willy Tarreauebd61602006-12-30 11:54:15 +01001382 mask->s_addr = ~0U;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001383 }
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001384 if (!inet_pton(AF_INET, s, addr)) {
Willy Tarreaubaaee002006-06-26 02:48:02 +02001385 struct hostent *he;
1386
Thierry FOURNIERfc7ac7b2014-02-11 15:23:04 +01001387 if (!resolve)
1388 goto out_err;
1389
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001390 if ((he = gethostbyname(s)) == NULL) {
1391 goto out_err;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001392 }
1393 else
1394 *addr = *(struct in_addr *) *(he->h_addr_list);
1395 }
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001396
1397 ret_val = 1;
1398 out_free:
1399 free(s);
1400 return ret_val;
1401 out_err:
1402 ret_val = 0;
1403 goto out_free;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001404}
1405
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001406
1407/*
Willy Tarreau6d20e282012-04-27 22:49:47 +02001408 * converts <str> to two struct in6_addr* which must be pre-allocated.
1409 * The format is "addr[/mask]", where "addr" cannot be empty, and mask
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05001410 * is an optional number of bits (128 being the default).
Willy Tarreau6d20e282012-04-27 22:49:47 +02001411 * Returns 1 if OK, 0 if error.
1412 */
1413int str62net(const char *str, struct in6_addr *addr, unsigned char *mask)
1414{
1415 char *c, *s;
1416 int ret_val = 0;
1417 char *err;
1418 unsigned long len = 128;
1419
1420 s = strdup(str);
1421 if (!s)
1422 return 0;
1423
1424 memset(mask, 0, sizeof(*mask));
1425 memset(addr, 0, sizeof(*addr));
1426
1427 if ((c = strrchr(s, '/')) != NULL) {
1428 *c++ = '\0'; /* c points to the mask */
1429 if (!*c)
1430 goto out_free;
1431
1432 len = strtoul(c, &err, 10);
1433 if ((err && *err) || (unsigned)len > 128)
1434 goto out_free;
1435 }
1436 *mask = len; /* OK we have a valid mask in <len> */
1437
1438 if (!inet_pton(AF_INET6, s, addr))
1439 goto out_free;
1440
1441 ret_val = 1;
1442 out_free:
1443 free(s);
1444 return ret_val;
1445}
1446
1447
1448/*
David du Colombier6f5ccb12011-03-10 22:26:24 +01001449 * Parse IPv4 address found in url.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001450 */
David du Colombier6f5ccb12011-03-10 22:26:24 +01001451int url2ipv4(const char *addr, struct in_addr *dst)
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001452{
1453 int saw_digit, octets, ch;
1454 u_char tmp[4], *tp;
1455 const char *cp = addr;
1456
1457 saw_digit = 0;
1458 octets = 0;
1459 *(tp = tmp) = 0;
1460
1461 while (*addr) {
1462 unsigned char digit = (ch = *addr++) - '0';
1463 if (digit > 9 && ch != '.')
1464 break;
1465 if (digit <= 9) {
1466 u_int new = *tp * 10 + digit;
1467 if (new > 255)
1468 return 0;
1469 *tp = new;
1470 if (!saw_digit) {
1471 if (++octets > 4)
1472 return 0;
1473 saw_digit = 1;
1474 }
1475 } else if (ch == '.' && saw_digit) {
1476 if (octets == 4)
1477 return 0;
1478 *++tp = 0;
1479 saw_digit = 0;
1480 } else
1481 return 0;
1482 }
1483
1484 if (octets < 4)
1485 return 0;
1486
1487 memcpy(&dst->s_addr, tmp, 4);
1488 return addr-cp-1;
1489}
1490
1491/*
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001492 * Resolve destination server from URL. Convert <str> to a sockaddr_storage.
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05001493 * <out> contain the code of the detected scheme, the start and length of
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001494 * the hostname. Actually only http and https are supported. <out> can be NULL.
1495 * This function returns the consumed length. It is useful if you parse complete
1496 * url like http://host:port/path, because the consumed length corresponds to
1497 * the first character of the path. If the conversion fails, it returns -1.
1498 *
1499 * This function tries to resolve the DNS name if haproxy is in starting mode.
1500 * So, this function may be used during the configuration parsing.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001501 */
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001502int url2sa(const char *url, int ulen, struct sockaddr_storage *addr, struct split_url *out)
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001503{
1504 const char *curr = url, *cp = url;
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001505 const char *end;
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001506 int ret, url_code = 0;
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001507 unsigned long long int http_code = 0;
1508 int default_port;
1509 struct hostent *he;
1510 char *p;
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001511
1512 /* Firstly, try to find :// pattern */
1513 while (curr < url+ulen && url_code != 0x3a2f2f) {
1514 url_code = ((url_code & 0xffff) << 8);
1515 url_code += (unsigned char)*curr++;
1516 }
1517
1518 /* Secondly, if :// pattern is found, verify parsed stuff
1519 * before pattern is matching our http pattern.
1520 * If so parse ip address and port in uri.
1521 *
1522 * WARNING: Current code doesn't support dynamic async dns resolver.
1523 */
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001524 if (url_code != 0x3a2f2f)
1525 return -1;
1526
1527 /* Copy scheme, and utrn to lower case. */
1528 while (cp < curr - 3)
1529 http_code = (http_code << 8) + *cp++;
1530 http_code |= 0x2020202020202020ULL; /* Turn everything to lower case */
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001531
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001532 /* HTTP or HTTPS url matching */
1533 if (http_code == 0x2020202068747470ULL) {
1534 default_port = 80;
1535 if (out)
1536 out->scheme = SCH_HTTP;
1537 }
1538 else if (http_code == 0x2020206874747073ULL) {
1539 default_port = 443;
1540 if (out)
1541 out->scheme = SCH_HTTPS;
1542 }
1543 else
1544 return -1;
1545
1546 /* If the next char is '[', the host address is IPv6. */
1547 if (*curr == '[') {
1548 curr++;
1549
1550 /* Check trash size */
1551 if (trash.size < ulen)
1552 return -1;
1553
1554 /* Look for ']' and copy the address in a trash buffer. */
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001555 p = trash.area;
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001556 for (end = curr;
1557 end < url + ulen && *end != ']';
1558 end++, p++)
1559 *p = *end;
1560 if (*end != ']')
1561 return -1;
1562 *p = '\0';
1563
1564 /* Update out. */
1565 if (out) {
1566 out->host = curr;
1567 out->host_len = end - curr;
1568 }
1569
1570 /* Try IPv6 decoding. */
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001571 if (!inet_pton(AF_INET6, trash.area, &((struct sockaddr_in6 *)addr)->sin6_addr))
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001572 return -1;
1573 end++;
1574
1575 /* Decode port. */
1576 if (*end == ':') {
1577 end++;
1578 default_port = read_uint(&end, url + ulen);
1579 }
1580 ((struct sockaddr_in6 *)addr)->sin6_port = htons(default_port);
1581 ((struct sockaddr_in6 *)addr)->sin6_family = AF_INET6;
1582 return end - url;
1583 }
1584 else {
1585 /* We are looking for IP address. If you want to parse and
1586 * resolve hostname found in url, you can use str2sa_range(), but
1587 * be warned this can slow down global daemon performances
1588 * while handling lagging dns responses.
1589 */
1590 ret = url2ipv4(curr, &((struct sockaddr_in *)addr)->sin_addr);
1591 if (ret) {
1592 /* Update out. */
1593 if (out) {
1594 out->host = curr;
1595 out->host_len = ret;
1596 }
1597
1598 curr += ret;
1599
1600 /* Decode port. */
1601 if (*curr == ':') {
1602 curr++;
1603 default_port = read_uint(&curr, url + ulen);
1604 }
1605 ((struct sockaddr_in *)addr)->sin_port = htons(default_port);
1606
1607 /* Set family. */
1608 ((struct sockaddr_in *)addr)->sin_family = AF_INET;
1609 return curr - url;
1610 }
1611 else if (global.mode & MODE_STARTING) {
1612 /* The IPv4 and IPv6 decoding fails, maybe the url contain name. Try to execute
1613 * synchronous DNS request only if HAProxy is in the start state.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001614 */
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001615
1616 /* look for : or / or end */
1617 for (end = curr;
1618 end < url + ulen && *end != '/' && *end != ':';
1619 end++);
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001620 memcpy(trash.area, curr, end - curr);
1621 trash.area[end - curr] = '\0';
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001622
1623 /* try to resolve an IPv4/IPv6 hostname */
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001624 he = gethostbyname(trash.area);
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001625 if (!he)
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001626 return -1;
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001627
1628 /* Update out. */
1629 if (out) {
1630 out->host = curr;
1631 out->host_len = end - curr;
1632 }
1633
1634 /* Decode port. */
1635 if (*end == ':') {
1636 end++;
1637 default_port = read_uint(&end, url + ulen);
1638 }
1639
1640 /* Copy IP address, set port and family. */
1641 switch (he->h_addrtype) {
1642 case AF_INET:
1643 ((struct sockaddr_in *)addr)->sin_addr = *(struct in_addr *) *(he->h_addr_list);
1644 ((struct sockaddr_in *)addr)->sin_port = htons(default_port);
1645 ((struct sockaddr_in *)addr)->sin_family = AF_INET;
1646 return end - url;
1647
1648 case AF_INET6:
1649 ((struct sockaddr_in6 *)addr)->sin6_addr = *(struct in6_addr *) *(he->h_addr_list);
1650 ((struct sockaddr_in6 *)addr)->sin6_port = htons(default_port);
1651 ((struct sockaddr_in6 *)addr)->sin6_family = AF_INET6;
1652 return end - url;
1653 }
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001654 }
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001655 }
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001656 return -1;
1657}
1658
Willy Tarreau631f01c2011-09-05 00:36:48 +02001659/* Tries to convert a sockaddr_storage address to text form. Upon success, the
1660 * address family is returned so that it's easy for the caller to adapt to the
1661 * output format. Zero is returned if the address family is not supported. -1
1662 * is returned upon error, with errno set. AF_INET, AF_INET6 and AF_UNIX are
1663 * supported.
1664 */
Willy Tarreaud5ec4bf2019-04-25 17:48:16 +02001665int addr_to_str(const struct sockaddr_storage *addr, char *str, int size)
Willy Tarreau631f01c2011-09-05 00:36:48 +02001666{
1667
Willy Tarreaud5ec4bf2019-04-25 17:48:16 +02001668 const void *ptr;
Willy Tarreau631f01c2011-09-05 00:36:48 +02001669
1670 if (size < 5)
1671 return 0;
1672 *str = '\0';
1673
1674 switch (addr->ss_family) {
1675 case AF_INET:
1676 ptr = &((struct sockaddr_in *)addr)->sin_addr;
1677 break;
1678 case AF_INET6:
1679 ptr = &((struct sockaddr_in6 *)addr)->sin6_addr;
1680 break;
1681 case AF_UNIX:
1682 memcpy(str, "unix", 5);
1683 return addr->ss_family;
1684 default:
1685 return 0;
1686 }
1687
1688 if (inet_ntop(addr->ss_family, ptr, str, size))
1689 return addr->ss_family;
1690
1691 /* failed */
1692 return -1;
1693}
1694
Simon Horman75ab8bd2014-06-16 09:39:41 +09001695/* Tries to convert a sockaddr_storage port to text form. Upon success, the
1696 * address family is returned so that it's easy for the caller to adapt to the
1697 * output format. Zero is returned if the address family is not supported. -1
1698 * is returned upon error, with errno set. AF_INET, AF_INET6 and AF_UNIX are
1699 * supported.
1700 */
Willy Tarreaud5ec4bf2019-04-25 17:48:16 +02001701int port_to_str(const struct sockaddr_storage *addr, char *str, int size)
Simon Horman75ab8bd2014-06-16 09:39:41 +09001702{
1703
1704 uint16_t port;
1705
1706
Willy Tarreaud7dad1b2017-01-06 16:46:22 +01001707 if (size < 6)
Simon Horman75ab8bd2014-06-16 09:39:41 +09001708 return 0;
1709 *str = '\0';
1710
1711 switch (addr->ss_family) {
1712 case AF_INET:
1713 port = ((struct sockaddr_in *)addr)->sin_port;
1714 break;
1715 case AF_INET6:
1716 port = ((struct sockaddr_in6 *)addr)->sin6_port;
1717 break;
1718 case AF_UNIX:
1719 memcpy(str, "unix", 5);
1720 return addr->ss_family;
1721 default:
1722 return 0;
1723 }
1724
1725 snprintf(str, size, "%u", ntohs(port));
1726 return addr->ss_family;
1727}
1728
Willy Tarreau16e01562016-08-09 16:46:18 +02001729/* check if the given address is local to the system or not. It will return
1730 * -1 when it's not possible to know, 0 when the address is not local, 1 when
1731 * it is. We don't want to iterate over all interfaces for this (and it is not
1732 * portable). So instead we try to bind in UDP to this address on a free non
1733 * privileged port and to connect to the same address, port 0 (connect doesn't
1734 * care). If it succeeds, we own the address. Note that non-inet addresses are
1735 * considered local since they're most likely AF_UNIX.
1736 */
1737int addr_is_local(const struct netns_entry *ns,
1738 const struct sockaddr_storage *orig)
1739{
1740 struct sockaddr_storage addr;
1741 int result;
1742 int fd;
1743
1744 if (!is_inet_addr(orig))
1745 return 1;
1746
1747 memcpy(&addr, orig, sizeof(addr));
1748 set_host_port(&addr, 0);
1749
1750 fd = my_socketat(ns, addr.ss_family, SOCK_DGRAM, IPPROTO_UDP);
1751 if (fd < 0)
1752 return -1;
1753
1754 result = -1;
1755 if (bind(fd, (struct sockaddr *)&addr, get_addr_len(&addr)) == 0) {
1756 if (connect(fd, (struct sockaddr *)&addr, get_addr_len(&addr)) == -1)
1757 result = 0; // fail, non-local address
1758 else
1759 result = 1; // success, local address
1760 }
1761 else {
1762 if (errno == EADDRNOTAVAIL)
1763 result = 0; // definitely not local :-)
1764 }
1765 close(fd);
1766
1767 return result;
1768}
1769
Willy Tarreaubaaee002006-06-26 02:48:02 +02001770/* will try to encode the string <string> replacing all characters tagged in
1771 * <map> with the hexadecimal representation of their ASCII-code (2 digits)
1772 * prefixed by <escape>, and will store the result between <start> (included)
1773 * and <stop> (excluded), and will always terminate the string with a '\0'
1774 * before <stop>. The position of the '\0' is returned if the conversion
1775 * completes. If bytes are missing between <start> and <stop>, then the
1776 * conversion will be incomplete and truncated. If <stop> <= <start>, the '\0'
1777 * cannot even be stored so we return <start> without writing the 0.
1778 * The input string must also be zero-terminated.
1779 */
1780const char hextab[16] = "0123456789ABCDEF";
1781char *encode_string(char *start, char *stop,
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001782 const char escape, const long *map,
Willy Tarreaubaaee002006-06-26 02:48:02 +02001783 const char *string)
1784{
1785 if (start < stop) {
1786 stop--; /* reserve one byte for the final '\0' */
1787 while (start < stop && *string != '\0') {
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001788 if (!ha_bit_test((unsigned char)(*string), map))
Willy Tarreaubaaee002006-06-26 02:48:02 +02001789 *start++ = *string;
1790 else {
1791 if (start + 3 >= stop)
1792 break;
1793 *start++ = escape;
1794 *start++ = hextab[(*string >> 4) & 15];
1795 *start++ = hextab[*string & 15];
1796 }
1797 string++;
1798 }
1799 *start = '\0';
1800 }
1801 return start;
1802}
1803
Thierry FOURNIERe059ec92014-03-17 12:01:13 +01001804/*
1805 * Same behavior as encode_string() above, except that it encodes chunk
1806 * <chunk> instead of a string.
1807 */
1808char *encode_chunk(char *start, char *stop,
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001809 const char escape, const long *map,
Willy Tarreau83061a82018-07-13 11:56:34 +02001810 const struct buffer *chunk)
Thierry FOURNIERe059ec92014-03-17 12:01:13 +01001811{
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001812 char *str = chunk->area;
1813 char *end = chunk->area + chunk->data;
Thierry FOURNIERe059ec92014-03-17 12:01:13 +01001814
1815 if (start < stop) {
1816 stop--; /* reserve one byte for the final '\0' */
1817 while (start < stop && str < end) {
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001818 if (!ha_bit_test((unsigned char)(*str), map))
Thierry FOURNIERe059ec92014-03-17 12:01:13 +01001819 *start++ = *str;
1820 else {
1821 if (start + 3 >= stop)
1822 break;
1823 *start++ = escape;
1824 *start++ = hextab[(*str >> 4) & 15];
1825 *start++ = hextab[*str & 15];
1826 }
1827 str++;
1828 }
1829 *start = '\0';
1830 }
1831 return start;
1832}
1833
Dragan Dosen0edd1092016-02-12 13:23:02 +01001834/*
1835 * Tries to prefix characters tagged in the <map> with the <escape>
Dragan Dosen1a5d0602016-07-22 16:00:31 +02001836 * character. The input <string> must be zero-terminated. The result will
1837 * be stored between <start> (included) and <stop> (excluded). This
1838 * function will always try to terminate the resulting string with a '\0'
1839 * before <stop>, and will return its position if the conversion
1840 * completes.
1841 */
1842char *escape_string(char *start, char *stop,
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001843 const char escape, const long *map,
Dragan Dosen1a5d0602016-07-22 16:00:31 +02001844 const char *string)
1845{
1846 if (start < stop) {
1847 stop--; /* reserve one byte for the final '\0' */
1848 while (start < stop && *string != '\0') {
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001849 if (!ha_bit_test((unsigned char)(*string), map))
Dragan Dosen1a5d0602016-07-22 16:00:31 +02001850 *start++ = *string;
1851 else {
1852 if (start + 2 >= stop)
1853 break;
1854 *start++ = escape;
1855 *start++ = *string;
1856 }
1857 string++;
1858 }
1859 *start = '\0';
1860 }
1861 return start;
1862}
1863
1864/*
1865 * Tries to prefix characters tagged in the <map> with the <escape>
Dragan Dosen0edd1092016-02-12 13:23:02 +01001866 * character. <chunk> contains the input to be escaped. The result will be
1867 * stored between <start> (included) and <stop> (excluded). The function
1868 * will always try to terminate the resulting string with a '\0' before
1869 * <stop>, and will return its position if the conversion completes.
1870 */
1871char *escape_chunk(char *start, char *stop,
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001872 const char escape, const long *map,
Willy Tarreau83061a82018-07-13 11:56:34 +02001873 const struct buffer *chunk)
Dragan Dosen0edd1092016-02-12 13:23:02 +01001874{
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001875 char *str = chunk->area;
1876 char *end = chunk->area + chunk->data;
Dragan Dosen0edd1092016-02-12 13:23:02 +01001877
1878 if (start < stop) {
1879 stop--; /* reserve one byte for the final '\0' */
1880 while (start < stop && str < end) {
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001881 if (!ha_bit_test((unsigned char)(*str), map))
Dragan Dosen0edd1092016-02-12 13:23:02 +01001882 *start++ = *str;
1883 else {
1884 if (start + 2 >= stop)
1885 break;
1886 *start++ = escape;
1887 *start++ = *str;
1888 }
1889 str++;
1890 }
1891 *start = '\0';
1892 }
1893 return start;
1894}
1895
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001896/* Check a string for using it in a CSV output format. If the string contains
1897 * one of the following four char <">, <,>, CR or LF, the string is
1898 * encapsulated between <"> and the <"> are escaped by a <""> sequence.
1899 * <str> is the input string to be escaped. The function assumes that
1900 * the input string is null-terminated.
1901 *
1902 * If <quote> is 0, the result is returned escaped but without double quote.
Willy Tarreau898529b2016-01-06 18:07:04 +01001903 * It is useful if the escaped string is used between double quotes in the
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001904 * format.
1905 *
Willy Tarreau898529b2016-01-06 18:07:04 +01001906 * printf("..., \"%s\", ...\r\n", csv_enc(str, 0, &trash));
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001907 *
Willy Tarreaub631c292016-01-08 10:04:08 +01001908 * If <quote> is 1, the converter puts the quotes only if any reserved character
1909 * is present. If <quote> is 2, the converter always puts the quotes.
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001910 *
Willy Tarreau83061a82018-07-13 11:56:34 +02001911 * <output> is a struct buffer used for storing the output string.
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001912 *
Willy Tarreau898529b2016-01-06 18:07:04 +01001913 * The function returns the converted string on its output. If an error
1914 * occurs, the function returns an empty string. This type of output is useful
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001915 * for using the function directly as printf() argument.
1916 *
1917 * If the output buffer is too short to contain the input string, the result
1918 * is truncated.
Willy Tarreau898529b2016-01-06 18:07:04 +01001919 *
Willy Tarreaub631c292016-01-08 10:04:08 +01001920 * This function appends the encoding to the existing output chunk, and it
1921 * guarantees that it starts immediately at the first available character of
1922 * the chunk. Please use csv_enc() instead if you want to replace the output
1923 * chunk.
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001924 */
Willy Tarreau83061a82018-07-13 11:56:34 +02001925const char *csv_enc_append(const char *str, int quote, struct buffer *output)
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001926{
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001927 char *end = output->area + output->size;
1928 char *out = output->area + output->data;
Willy Tarreau898529b2016-01-06 18:07:04 +01001929 char *ptr = out;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001930
Willy Tarreaub631c292016-01-08 10:04:08 +01001931 if (quote == 1) {
1932 /* automatic quoting: first verify if we'll have to quote the string */
1933 if (!strpbrk(str, "\n\r,\""))
1934 quote = 0;
1935 }
1936
1937 if (quote)
1938 *ptr++ = '"';
1939
Willy Tarreau898529b2016-01-06 18:07:04 +01001940 while (*str && ptr < end - 2) { /* -2 for reserving space for <"> and \0. */
1941 *ptr = *str;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001942 if (*str == '"') {
Willy Tarreau898529b2016-01-06 18:07:04 +01001943 ptr++;
1944 if (ptr >= end - 2) {
1945 ptr--;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001946 break;
1947 }
Willy Tarreau898529b2016-01-06 18:07:04 +01001948 *ptr = '"';
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001949 }
Willy Tarreau898529b2016-01-06 18:07:04 +01001950 ptr++;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001951 str++;
1952 }
1953
Willy Tarreaub631c292016-01-08 10:04:08 +01001954 if (quote)
1955 *ptr++ = '"';
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001956
Willy Tarreau898529b2016-01-06 18:07:04 +01001957 *ptr = '\0';
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001958 output->data = ptr - output->area;
Willy Tarreau898529b2016-01-06 18:07:04 +01001959 return out;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001960}
1961
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001962/* Decode an URL-encoded string in-place. The resulting string might
1963 * be shorter. If some forbidden characters are found, the conversion is
Thierry FOURNIER5068d962013-10-04 16:27:27 +02001964 * aborted, the string is truncated before the issue and a negative value is
1965 * returned, otherwise the operation returns the length of the decoded string.
Willy Tarreau62ba9ba2020-04-23 17:54:47 +02001966 * If the 'in_form' argument is non-nul the string is assumed to be part of
1967 * an "application/x-www-form-urlencoded" encoded string, and the '+' will be
1968 * turned to a space. If it's zero, this will only be done after a question
1969 * mark ('?').
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001970 */
Willy Tarreau62ba9ba2020-04-23 17:54:47 +02001971int url_decode(char *string, int in_form)
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001972{
1973 char *in, *out;
Thierry FOURNIER5068d962013-10-04 16:27:27 +02001974 int ret = -1;
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001975
1976 in = string;
1977 out = string;
1978 while (*in) {
1979 switch (*in) {
1980 case '+' :
Willy Tarreau62ba9ba2020-04-23 17:54:47 +02001981 *out++ = in_form ? ' ' : *in;
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001982 break;
1983 case '%' :
1984 if (!ishex(in[1]) || !ishex(in[2]))
1985 goto end;
1986 *out++ = (hex2i(in[1]) << 4) + hex2i(in[2]);
1987 in += 2;
1988 break;
Willy Tarreau62ba9ba2020-04-23 17:54:47 +02001989 case '?':
1990 in_form = 1;
1991 /* fall through */
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001992 default:
1993 *out++ = *in;
1994 break;
1995 }
1996 in++;
1997 }
Thierry FOURNIER5068d962013-10-04 16:27:27 +02001998 ret = out - string; /* success */
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001999 end:
2000 *out = 0;
2001 return ret;
2002}
Willy Tarreaubaaee002006-06-26 02:48:02 +02002003
Willy Tarreau6911fa42007-03-04 18:06:08 +01002004unsigned int str2ui(const char *s)
2005{
2006 return __str2ui(s);
2007}
2008
2009unsigned int str2uic(const char *s)
2010{
2011 return __str2uic(s);
2012}
2013
2014unsigned int strl2ui(const char *s, int len)
2015{
2016 return __strl2ui(s, len);
2017}
2018
2019unsigned int strl2uic(const char *s, int len)
2020{
2021 return __strl2uic(s, len);
2022}
2023
Willy Tarreau4ec83cd2010-10-15 23:19:55 +02002024unsigned int read_uint(const char **s, const char *end)
2025{
2026 return __read_uint(s, end);
2027}
2028
Thierry FOURNIER763a5d82015-07-06 23:09:52 +02002029/* This function reads an unsigned integer from the string pointed to by <s> and
2030 * returns it. The <s> pointer is adjusted to point to the first unread char. The
2031 * function automatically stops at <end>. If the number overflows, the 2^64-1
2032 * value is returned.
2033 */
2034unsigned long long int read_uint64(const char **s, const char *end)
2035{
2036 const char *ptr = *s;
2037 unsigned long long int i = 0, tmp;
2038 unsigned int j;
2039
2040 while (ptr < end) {
2041
2042 /* read next char */
2043 j = *ptr - '0';
2044 if (j > 9)
2045 goto read_uint64_end;
2046
2047 /* add char to the number and check overflow. */
2048 tmp = i * 10;
2049 if (tmp / 10 != i) {
2050 i = ULLONG_MAX;
2051 goto read_uint64_eat;
2052 }
2053 if (ULLONG_MAX - tmp < j) {
2054 i = ULLONG_MAX;
2055 goto read_uint64_eat;
2056 }
2057 i = tmp + j;
2058 ptr++;
2059 }
2060read_uint64_eat:
2061 /* eat each numeric char */
2062 while (ptr < end) {
2063 if ((unsigned int)(*ptr - '0') > 9)
2064 break;
2065 ptr++;
2066 }
2067read_uint64_end:
2068 *s = ptr;
2069 return i;
2070}
2071
2072/* This function reads an integer from the string pointed to by <s> and returns
2073 * it. The <s> pointer is adjusted to point to the first unread char. The function
2074 * automatically stops at <end>. Il the number is bigger than 2^63-2, the 2^63-1
2075 * value is returned. If the number is lowest than -2^63-1, the -2^63 value is
2076 * returned.
2077 */
2078long long int read_int64(const char **s, const char *end)
2079{
2080 unsigned long long int i = 0;
2081 int neg = 0;
2082
2083 /* Look for minus char. */
2084 if (**s == '-') {
2085 neg = 1;
2086 (*s)++;
2087 }
2088 else if (**s == '+')
2089 (*s)++;
2090
2091 /* convert as positive number. */
2092 i = read_uint64(s, end);
2093
2094 if (neg) {
2095 if (i > 0x8000000000000000ULL)
2096 return LLONG_MIN;
2097 return -i;
2098 }
2099 if (i > 0x7fffffffffffffffULL)
2100 return LLONG_MAX;
2101 return i;
2102}
2103
Willy Tarreau6911fa42007-03-04 18:06:08 +01002104/* This one is 7 times faster than strtol() on athlon with checks.
2105 * It returns the value of the number composed of all valid digits read,
2106 * and can process negative numbers too.
2107 */
2108int strl2ic(const char *s, int len)
2109{
2110 int i = 0;
Willy Tarreau3f0c9762007-10-25 09:42:24 +02002111 int j, k;
Willy Tarreau6911fa42007-03-04 18:06:08 +01002112
2113 if (len > 0) {
2114 if (*s != '-') {
2115 /* positive number */
2116 while (len-- > 0) {
2117 j = (*s++) - '0';
Willy Tarreau3f0c9762007-10-25 09:42:24 +02002118 k = i * 10;
Willy Tarreau6911fa42007-03-04 18:06:08 +01002119 if (j > 9)
2120 break;
Willy Tarreau3f0c9762007-10-25 09:42:24 +02002121 i = k + j;
Willy Tarreau6911fa42007-03-04 18:06:08 +01002122 }
2123 } else {
2124 /* negative number */
2125 s++;
2126 while (--len > 0) {
2127 j = (*s++) - '0';
Willy Tarreau3f0c9762007-10-25 09:42:24 +02002128 k = i * 10;
Willy Tarreau6911fa42007-03-04 18:06:08 +01002129 if (j > 9)
2130 break;
Willy Tarreau3f0c9762007-10-25 09:42:24 +02002131 i = k - j;
Willy Tarreau6911fa42007-03-04 18:06:08 +01002132 }
2133 }
2134 }
2135 return i;
2136}
2137
2138
2139/* This function reads exactly <len> chars from <s> and converts them to a
2140 * signed integer which it stores into <ret>. It accurately detects any error
2141 * (truncated string, invalid chars, overflows). It is meant to be used in
2142 * applications designed for hostile environments. It returns zero when the
2143 * number has successfully been converted, non-zero otherwise. When an error
2144 * is returned, the <ret> value is left untouched. It is yet 5 to 40 times
2145 * faster than strtol().
2146 */
2147int strl2irc(const char *s, int len, int *ret)
2148{
2149 int i = 0;
2150 int j;
2151
2152 if (!len)
2153 return 1;
2154
2155 if (*s != '-') {
2156 /* positive number */
2157 while (len-- > 0) {
2158 j = (*s++) - '0';
2159 if (j > 9) return 1; /* invalid char */
2160 if (i > INT_MAX / 10) return 1; /* check for multiply overflow */
2161 i = i * 10;
2162 if (i + j < i) return 1; /* check for addition overflow */
2163 i = i + j;
2164 }
2165 } else {
2166 /* negative number */
2167 s++;
2168 while (--len > 0) {
2169 j = (*s++) - '0';
2170 if (j > 9) return 1; /* invalid char */
2171 if (i < INT_MIN / 10) return 1; /* check for multiply overflow */
2172 i = i * 10;
2173 if (i - j > i) return 1; /* check for subtract overflow */
2174 i = i - j;
2175 }
2176 }
2177 *ret = i;
2178 return 0;
2179}
2180
2181
2182/* This function reads exactly <len> chars from <s> and converts them to a
2183 * signed integer which it stores into <ret>. It accurately detects any error
2184 * (truncated string, invalid chars, overflows). It is meant to be used in
2185 * applications designed for hostile environments. It returns zero when the
2186 * number has successfully been converted, non-zero otherwise. When an error
2187 * is returned, the <ret> value is left untouched. It is about 3 times slower
William Dauchy060ffc82021-02-06 20:47:51 +01002188 * than strl2irc().
Willy Tarreau6911fa42007-03-04 18:06:08 +01002189 */
Willy Tarreau6911fa42007-03-04 18:06:08 +01002190
2191int strl2llrc(const char *s, int len, long long *ret)
2192{
2193 long long i = 0;
2194 int j;
2195
2196 if (!len)
2197 return 1;
2198
2199 if (*s != '-') {
2200 /* positive number */
2201 while (len-- > 0) {
2202 j = (*s++) - '0';
2203 if (j > 9) return 1; /* invalid char */
2204 if (i > LLONG_MAX / 10LL) return 1; /* check for multiply overflow */
2205 i = i * 10LL;
2206 if (i + j < i) return 1; /* check for addition overflow */
2207 i = i + j;
2208 }
2209 } else {
2210 /* negative number */
2211 s++;
2212 while (--len > 0) {
2213 j = (*s++) - '0';
2214 if (j > 9) return 1; /* invalid char */
2215 if (i < LLONG_MIN / 10LL) return 1; /* check for multiply overflow */
2216 i = i * 10LL;
2217 if (i - j > i) return 1; /* check for subtract overflow */
2218 i = i - j;
2219 }
2220 }
2221 *ret = i;
2222 return 0;
2223}
2224
Thierry FOURNIER511e9472014-01-23 17:40:34 +01002225/* This function is used with pat_parse_dotted_ver(). It converts a string
2226 * composed by two number separated by a dot. Each part must contain in 16 bits
2227 * because internally they will be represented as a 32-bit quantity stored in
2228 * a 64-bit integer. It returns zero when the number has successfully been
2229 * converted, non-zero otherwise. When an error is returned, the <ret> value
2230 * is left untouched.
2231 *
2232 * "1.3" -> 0x0000000000010003
2233 * "65535.65535" -> 0x00000000ffffffff
2234 */
2235int strl2llrc_dotted(const char *text, int len, long long *ret)
2236{
2237 const char *end = &text[len];
2238 const char *p;
2239 long long major, minor;
2240
2241 /* Look for dot. */
2242 for (p = text; p < end; p++)
2243 if (*p == '.')
2244 break;
2245
2246 /* Convert major. */
2247 if (strl2llrc(text, p - text, &major) != 0)
2248 return 1;
2249
2250 /* Check major. */
2251 if (major >= 65536)
2252 return 1;
2253
2254 /* Convert minor. */
2255 minor = 0;
2256 if (p < end)
2257 if (strl2llrc(p + 1, end - (p + 1), &minor) != 0)
2258 return 1;
2259
2260 /* Check minor. */
2261 if (minor >= 65536)
2262 return 1;
2263
2264 /* Compose value. */
2265 *ret = (major << 16) | (minor & 0xffff);
2266 return 0;
2267}
2268
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002269/* This function parses a time value optionally followed by a unit suffix among
2270 * "d", "h", "m", "s", "ms" or "us". It converts the value into the unit
2271 * expected by the caller. The computation does its best to avoid overflows.
2272 * The value is returned in <ret> if everything is fine, and a NULL is returned
2273 * by the function. In case of error, a pointer to the error is returned and
2274 * <ret> is left untouched. Values are automatically rounded up when needed.
Willy Tarreau9faebe32019-06-07 19:00:37 +02002275 * Values resulting in values larger than or equal to 2^31 after conversion are
2276 * reported as an overflow as value PARSE_TIME_OVER. Non-null values resulting
2277 * in an underflow are reported as an underflow as value PARSE_TIME_UNDER.
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002278 */
2279const char *parse_time_err(const char *text, unsigned *ret, unsigned unit_flags)
2280{
Willy Tarreau9faebe32019-06-07 19:00:37 +02002281 unsigned long long imult, idiv;
2282 unsigned long long omult, odiv;
2283 unsigned long long value, result;
Christopher Fauletc20ad0d2020-12-11 09:23:07 +01002284 const char *str = text;
2285
2286 if (!isdigit((unsigned char)*text))
2287 return text;
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002288
2289 omult = odiv = 1;
2290
2291 switch (unit_flags & TIME_UNIT_MASK) {
2292 case TIME_UNIT_US: omult = 1000000; break;
2293 case TIME_UNIT_MS: omult = 1000; break;
2294 case TIME_UNIT_S: break;
2295 case TIME_UNIT_MIN: odiv = 60; break;
2296 case TIME_UNIT_HOUR: odiv = 3600; break;
2297 case TIME_UNIT_DAY: odiv = 86400; break;
2298 default: break;
2299 }
2300
2301 value = 0;
2302
2303 while (1) {
2304 unsigned int j;
2305
2306 j = *text - '0';
2307 if (j > 9)
2308 break;
2309 text++;
2310 value *= 10;
2311 value += j;
2312 }
2313
2314 imult = idiv = 1;
2315 switch (*text) {
2316 case '\0': /* no unit = default unit */
2317 imult = omult = idiv = odiv = 1;
Christopher Fauletc20ad0d2020-12-11 09:23:07 +01002318 goto end;
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002319 case 's': /* second = unscaled unit */
2320 break;
2321 case 'u': /* microsecond : "us" */
2322 if (text[1] == 's') {
2323 idiv = 1000000;
2324 text++;
2325 }
Christopher Fauletc20ad0d2020-12-11 09:23:07 +01002326 return text;
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002327 case 'm': /* millisecond : "ms" or minute: "m" */
2328 if (text[1] == 's') {
2329 idiv = 1000;
2330 text++;
2331 } else
2332 imult = 60;
2333 break;
2334 case 'h': /* hour : "h" */
2335 imult = 3600;
2336 break;
2337 case 'd': /* day : "d" */
2338 imult = 86400;
2339 break;
2340 default:
2341 return text;
2342 break;
2343 }
Christopher Fauletc20ad0d2020-12-11 09:23:07 +01002344 if (*(++text) != '\0') {
2345 ha_warning("unexpected character '%c' after the timer value '%s', only "
2346 "(us=microseconds,ms=milliseconds,s=seconds,m=minutes,h=hours,d=days) are supported."
2347 " This will be reported as an error in next versions.\n", *text, str);
2348 }
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002349
Christopher Fauletc20ad0d2020-12-11 09:23:07 +01002350 end:
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002351 if (omult % idiv == 0) { omult /= idiv; idiv = 1; }
2352 if (idiv % omult == 0) { idiv /= omult; omult = 1; }
2353 if (imult % odiv == 0) { imult /= odiv; odiv = 1; }
2354 if (odiv % imult == 0) { odiv /= imult; imult = 1; }
2355
Willy Tarreau9faebe32019-06-07 19:00:37 +02002356 result = (value * (imult * omult) + (idiv * odiv - 1)) / (idiv * odiv);
2357 if (result >= 0x80000000)
2358 return PARSE_TIME_OVER;
2359 if (!result && value)
2360 return PARSE_TIME_UNDER;
2361 *ret = result;
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002362 return NULL;
2363}
Willy Tarreau6911fa42007-03-04 18:06:08 +01002364
Emeric Brun39132b22010-01-04 14:57:24 +01002365/* this function converts the string starting at <text> to an unsigned int
2366 * stored in <ret>. If an error is detected, the pointer to the unexpected
Joseph Herlant32b83272018-11-15 11:58:28 -08002367 * character is returned. If the conversion is successful, NULL is returned.
Emeric Brun39132b22010-01-04 14:57:24 +01002368 */
2369const char *parse_size_err(const char *text, unsigned *ret) {
2370 unsigned value = 0;
2371
Christopher Faulet82635a02020-12-11 09:30:45 +01002372 if (!isdigit((unsigned char)*text))
2373 return text;
2374
Emeric Brun39132b22010-01-04 14:57:24 +01002375 while (1) {
2376 unsigned int j;
2377
2378 j = *text - '0';
2379 if (j > 9)
2380 break;
2381 if (value > ~0U / 10)
2382 return text;
2383 value *= 10;
2384 if (value > (value + j))
2385 return text;
2386 value += j;
2387 text++;
2388 }
2389
2390 switch (*text) {
2391 case '\0':
2392 break;
2393 case 'K':
2394 case 'k':
2395 if (value > ~0U >> 10)
2396 return text;
2397 value = value << 10;
2398 break;
2399 case 'M':
2400 case 'm':
2401 if (value > ~0U >> 20)
2402 return text;
2403 value = value << 20;
2404 break;
2405 case 'G':
2406 case 'g':
2407 if (value > ~0U >> 30)
2408 return text;
2409 value = value << 30;
2410 break;
2411 default:
2412 return text;
2413 }
2414
Godbach58048a22015-01-28 17:36:16 +08002415 if (*text != '\0' && *++text != '\0')
2416 return text;
2417
Emeric Brun39132b22010-01-04 14:57:24 +01002418 *ret = value;
2419 return NULL;
2420}
2421
Willy Tarreau126d4062013-12-03 17:50:47 +01002422/*
2423 * Parse binary string written in hexadecimal (source) and store the decoded
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05002424 * result into binstr and set binstrlen to the length of binstr. Memory for
Willy Tarreau126d4062013-12-03 17:50:47 +01002425 * binstr is allocated by the function. In case of error, returns 0 with an
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05002426 * error message in err. In success case, it returns the consumed length.
Willy Tarreau126d4062013-12-03 17:50:47 +01002427 */
2428int parse_binary(const char *source, char **binstr, int *binstrlen, char **err)
2429{
2430 int len;
2431 const char *p = source;
2432 int i,j;
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002433 int alloc;
Willy Tarreau126d4062013-12-03 17:50:47 +01002434
2435 len = strlen(source);
2436 if (len % 2) {
2437 memprintf(err, "an even number of hex digit is expected");
2438 return 0;
2439 }
2440
2441 len = len >> 1;
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002442
Willy Tarreau126d4062013-12-03 17:50:47 +01002443 if (!*binstr) {
Tim Duesterhuse52b6e52020-09-12 20:26:43 +02002444 *binstr = calloc(len, sizeof(**binstr));
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002445 if (!*binstr) {
2446 memprintf(err, "out of memory while loading string pattern");
2447 return 0;
2448 }
2449 alloc = 1;
Willy Tarreau126d4062013-12-03 17:50:47 +01002450 }
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002451 else {
2452 if (*binstrlen < len) {
Joseph Herlant76dbe782018-11-15 12:01:22 -08002453 memprintf(err, "no space available in the buffer. expect %d, provides %d",
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002454 len, *binstrlen);
2455 return 0;
2456 }
2457 alloc = 0;
2458 }
2459 *binstrlen = len;
Willy Tarreau126d4062013-12-03 17:50:47 +01002460
2461 i = j = 0;
2462 while (j < len) {
2463 if (!ishex(p[i++]))
2464 goto bad_input;
2465 if (!ishex(p[i++]))
2466 goto bad_input;
2467 (*binstr)[j++] = (hex2i(p[i-2]) << 4) + hex2i(p[i-1]);
2468 }
Thierry FOURNIERee330af2014-01-21 11:36:14 +01002469 return len << 1;
Willy Tarreau126d4062013-12-03 17:50:47 +01002470
2471bad_input:
2472 memprintf(err, "an hex digit is expected (found '%c')", p[i-1]);
Willy Tarreau61cfdf42021-02-20 10:46:51 +01002473 if (alloc)
2474 ha_free(binstr);
Willy Tarreau126d4062013-12-03 17:50:47 +01002475 return 0;
2476}
2477
Willy Tarreau946ba592009-05-10 15:41:18 +02002478/* copies at most <n> characters from <src> and always terminates with '\0' */
2479char *my_strndup(const char *src, int n)
2480{
2481 int len = 0;
2482 char *ret;
2483
2484 while (len < n && src[len])
2485 len++;
2486
Vincent Bernat3c2f2f22016-04-03 13:48:42 +02002487 ret = malloc(len + 1);
Willy Tarreau946ba592009-05-10 15:41:18 +02002488 if (!ret)
2489 return ret;
2490 memcpy(ret, src, len);
2491 ret[len] = '\0';
2492 return ret;
2493}
2494
Baptiste Assmannbb77c8e2013-10-06 23:24:13 +02002495/*
2496 * search needle in haystack
2497 * returns the pointer if found, returns NULL otherwise
2498 */
2499const void *my_memmem(const void *haystack, size_t haystacklen, const void *needle, size_t needlelen)
2500{
2501 const void *c = NULL;
2502 unsigned char f;
2503
2504 if ((haystack == NULL) || (needle == NULL) || (haystacklen < needlelen))
2505 return NULL;
2506
2507 f = *(char *)needle;
2508 c = haystack;
2509 while ((c = memchr(c, f, haystacklen - (c - haystack))) != NULL) {
2510 if ((haystacklen - (c - haystack)) < needlelen)
2511 return NULL;
2512
2513 if (memcmp(c, needle, needlelen) == 0)
2514 return c;
2515 ++c;
2516 }
2517 return NULL;
2518}
2519
Ilya Shipitsinc02a23f2020-05-06 00:53:22 +05002520/* get length of the initial segment consisting entirely of bytes in <accept> */
Christopher Faulet5eb96cb2020-04-15 10:23:01 +02002521size_t my_memspn(const void *str, size_t len, const void *accept, size_t acceptlen)
2522{
2523 size_t ret = 0;
2524
2525 while (ret < len && memchr(accept, *((int *)str), acceptlen)) {
2526 str++;
2527 ret++;
2528 }
2529 return ret;
2530}
2531
Ilya Shipitsinc02a23f2020-05-06 00:53:22 +05002532/* get length of the initial segment consisting entirely of bytes not in <rejcet> */
Christopher Faulet5eb96cb2020-04-15 10:23:01 +02002533size_t my_memcspn(const void *str, size_t len, const void *reject, size_t rejectlen)
2534{
2535 size_t ret = 0;
2536
2537 while (ret < len) {
2538 if(memchr(reject, *((int *)str), rejectlen))
2539 return ret;
2540 str++;
2541 ret++;
2542 }
2543 return ret;
2544}
2545
Willy Tarreau482b00d2009-10-04 22:48:42 +02002546/* This function returns the first unused key greater than or equal to <key> in
2547 * ID tree <root>. Zero is returned if no place is found.
2548 */
2549unsigned int get_next_id(struct eb_root *root, unsigned int key)
2550{
2551 struct eb32_node *used;
2552
2553 do {
2554 used = eb32_lookup_ge(root, key);
2555 if (!used || used->key > key)
2556 return key; /* key is available */
2557 key++;
2558 } while (key);
2559 return key;
2560}
2561
Willy Tarreau9c1e15d2017-11-15 18:51:29 +01002562/* dump the full tree to <file> in DOT format for debugging purposes. Will
2563 * optionally highlight node <subj> if found, depending on operation <op> :
2564 * 0 : nothing
2565 * >0 : insertion, node/leaf are surrounded in red
2566 * <0 : removal, node/leaf are dashed with no background
2567 * Will optionally add "desc" as a label on the graph if set and non-null.
2568 */
2569void eb32sc_to_file(FILE *file, struct eb_root *root, const struct eb32sc_node *subj, int op, const char *desc)
Willy Tarreaued3cda02017-11-15 15:04:05 +01002570{
2571 struct eb32sc_node *node;
2572 unsigned long scope = -1;
2573
Willy Tarreau9c1e15d2017-11-15 18:51:29 +01002574 fprintf(file, "digraph ebtree {\n");
2575
2576 if (desc && *desc) {
2577 fprintf(file,
2578 " fontname=\"fixed\";\n"
2579 " fontsize=8;\n"
2580 " label=\"%s\";\n", desc);
2581 }
2582
Willy Tarreaued3cda02017-11-15 15:04:05 +01002583 fprintf(file,
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002584 " node [fontname=\"fixed\" fontsize=8 shape=\"box\" style=\"filled\" color=\"black\" fillcolor=\"white\"];\n"
2585 " edge [fontname=\"fixed\" fontsize=8 style=\"solid\" color=\"magenta\" dir=\"forward\"];\n"
Willy Tarreaued3cda02017-11-15 15:04:05 +01002586 " \"%lx_n\" [label=\"root\\n%lx\"]\n", (long)eb_root_to_node(root), (long)root
2587 );
2588
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002589 fprintf(file, " \"%lx_n\" -> \"%lx_%c\" [taillabel=\"L\"];\n",
Willy Tarreaued3cda02017-11-15 15:04:05 +01002590 (long)eb_root_to_node(root),
2591 (long)eb_root_to_node(eb_clrtag(root->b[0])),
Willy Tarreaued3cda02017-11-15 15:04:05 +01002592 eb_gettag(root->b[0]) == EB_LEAF ? 'l' : 'n');
2593
2594 node = eb32sc_first(root, scope);
2595 while (node) {
2596 if (node->node.node_p) {
2597 /* node part is used */
Willy Tarreau9c1e15d2017-11-15 18:51:29 +01002598 fprintf(file, " \"%lx_n\" [label=\"%lx\\nkey=%u\\nscope=%lx\\nbit=%d\" fillcolor=\"lightskyblue1\" %s];\n",
2599 (long)node, (long)node, node->key, node->node_s, node->node.bit,
2600 (node == subj) ? (op < 0 ? "color=\"red\" style=\"dashed\"" : op > 0 ? "color=\"red\"" : "") : "");
Willy Tarreaued3cda02017-11-15 15:04:05 +01002601
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002602 fprintf(file, " \"%lx_n\" -> \"%lx_n\" [taillabel=\"%c\"];\n",
Willy Tarreaued3cda02017-11-15 15:04:05 +01002603 (long)node,
2604 (long)eb_root_to_node(eb_clrtag(node->node.node_p)),
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002605 eb_gettag(node->node.node_p) ? 'R' : 'L');
Willy Tarreaued3cda02017-11-15 15:04:05 +01002606
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002607 fprintf(file, " \"%lx_n\" -> \"%lx_%c\" [taillabel=\"L\"];\n",
Willy Tarreaued3cda02017-11-15 15:04:05 +01002608 (long)node,
2609 (long)eb_root_to_node(eb_clrtag(node->node.branches.b[0])),
Willy Tarreaued3cda02017-11-15 15:04:05 +01002610 eb_gettag(node->node.branches.b[0]) == EB_LEAF ? 'l' : 'n');
2611
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002612 fprintf(file, " \"%lx_n\" -> \"%lx_%c\" [taillabel=\"R\"];\n",
Willy Tarreaued3cda02017-11-15 15:04:05 +01002613 (long)node,
2614 (long)eb_root_to_node(eb_clrtag(node->node.branches.b[1])),
Willy Tarreaued3cda02017-11-15 15:04:05 +01002615 eb_gettag(node->node.branches.b[1]) == EB_LEAF ? 'l' : 'n');
2616 }
2617
Willy Tarreau9c1e15d2017-11-15 18:51:29 +01002618 fprintf(file, " \"%lx_l\" [label=\"%lx\\nkey=%u\\nscope=%lx\\npfx=%u\" fillcolor=\"yellow\" %s];\n",
2619 (long)node, (long)node, node->key, node->leaf_s, node->node.pfx,
2620 (node == subj) ? (op < 0 ? "color=\"red\" style=\"dashed\"" : op > 0 ? "color=\"red\"" : "") : "");
Willy Tarreaued3cda02017-11-15 15:04:05 +01002621
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002622 fprintf(file, " \"%lx_l\" -> \"%lx_n\" [taillabel=\"%c\"];\n",
Willy Tarreaued3cda02017-11-15 15:04:05 +01002623 (long)node,
2624 (long)eb_root_to_node(eb_clrtag(node->node.leaf_p)),
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002625 eb_gettag(node->node.leaf_p) ? 'R' : 'L');
Willy Tarreaued3cda02017-11-15 15:04:05 +01002626 node = eb32sc_next(node, scope);
2627 }
2628 fprintf(file, "}\n");
2629}
2630
Willy Tarreau348238b2010-01-18 15:05:57 +01002631/* This function compares a sample word possibly followed by blanks to another
2632 * clean word. The compare is case-insensitive. 1 is returned if both are equal,
2633 * otherwise zero. This intends to be used when checking HTTP headers for some
2634 * values. Note that it validates a word followed only by blanks but does not
2635 * validate a word followed by blanks then other chars.
2636 */
2637int word_match(const char *sample, int slen, const char *word, int wlen)
2638{
2639 if (slen < wlen)
2640 return 0;
2641
2642 while (wlen) {
2643 char c = *sample ^ *word;
2644 if (c && c != ('A' ^ 'a'))
2645 return 0;
2646 sample++;
2647 word++;
2648 slen--;
2649 wlen--;
2650 }
2651
2652 while (slen) {
2653 if (*sample != ' ' && *sample != '\t')
2654 return 0;
2655 sample++;
2656 slen--;
2657 }
2658 return 1;
2659}
Willy Tarreau482b00d2009-10-04 22:48:42 +02002660
Willy Tarreaud54bbdc2009-09-07 11:00:31 +02002661/* Converts any text-formatted IPv4 address to a host-order IPv4 address. It
2662 * is particularly fast because it avoids expensive operations such as
2663 * multiplies, which are optimized away at the end. It requires a properly
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05002664 * formatted address though (3 points).
Willy Tarreaud54bbdc2009-09-07 11:00:31 +02002665 */
2666unsigned int inetaddr_host(const char *text)
2667{
2668 const unsigned int ascii_zero = ('0' << 24) | ('0' << 16) | ('0' << 8) | '0';
2669 register unsigned int dig100, dig10, dig1;
2670 int s;
2671 const char *p, *d;
2672
2673 dig1 = dig10 = dig100 = ascii_zero;
2674 s = 24;
2675
2676 p = text;
2677 while (1) {
2678 if (((unsigned)(*p - '0')) <= 9) {
2679 p++;
2680 continue;
2681 }
2682
2683 /* here, we have a complete byte between <text> and <p> (exclusive) */
2684 if (p == text)
2685 goto end;
2686
2687 d = p - 1;
2688 dig1 |= (unsigned int)(*d << s);
2689 if (d == text)
2690 goto end;
2691
2692 d--;
2693 dig10 |= (unsigned int)(*d << s);
2694 if (d == text)
2695 goto end;
2696
2697 d--;
2698 dig100 |= (unsigned int)(*d << s);
2699 end:
2700 if (!s || *p != '.')
2701 break;
2702
2703 s -= 8;
2704 text = ++p;
2705 }
2706
2707 dig100 -= ascii_zero;
2708 dig10 -= ascii_zero;
2709 dig1 -= ascii_zero;
2710 return ((dig100 * 10) + dig10) * 10 + dig1;
2711}
2712
2713/*
2714 * Idem except the first unparsed character has to be passed in <stop>.
2715 */
2716unsigned int inetaddr_host_lim(const char *text, const char *stop)
2717{
2718 const unsigned int ascii_zero = ('0' << 24) | ('0' << 16) | ('0' << 8) | '0';
2719 register unsigned int dig100, dig10, dig1;
2720 int s;
2721 const char *p, *d;
2722
2723 dig1 = dig10 = dig100 = ascii_zero;
2724 s = 24;
2725
2726 p = text;
2727 while (1) {
2728 if (((unsigned)(*p - '0')) <= 9 && p < stop) {
2729 p++;
2730 continue;
2731 }
2732
2733 /* here, we have a complete byte between <text> and <p> (exclusive) */
2734 if (p == text)
2735 goto end;
2736
2737 d = p - 1;
2738 dig1 |= (unsigned int)(*d << s);
2739 if (d == text)
2740 goto end;
2741
2742 d--;
2743 dig10 |= (unsigned int)(*d << s);
2744 if (d == text)
2745 goto end;
2746
2747 d--;
2748 dig100 |= (unsigned int)(*d << s);
2749 end:
2750 if (!s || p == stop || *p != '.')
2751 break;
2752
2753 s -= 8;
2754 text = ++p;
2755 }
2756
2757 dig100 -= ascii_zero;
2758 dig10 -= ascii_zero;
2759 dig1 -= ascii_zero;
2760 return ((dig100 * 10) + dig10) * 10 + dig1;
2761}
2762
2763/*
2764 * Idem except the pointer to first unparsed byte is returned into <ret> which
2765 * must not be NULL.
2766 */
Willy Tarreau74172752010-10-15 23:21:42 +02002767unsigned int inetaddr_host_lim_ret(char *text, char *stop, char **ret)
Willy Tarreaud54bbdc2009-09-07 11:00:31 +02002768{
2769 const unsigned int ascii_zero = ('0' << 24) | ('0' << 16) | ('0' << 8) | '0';
2770 register unsigned int dig100, dig10, dig1;
2771 int s;
Willy Tarreau74172752010-10-15 23:21:42 +02002772 char *p, *d;
Willy Tarreaud54bbdc2009-09-07 11:00:31 +02002773
2774 dig1 = dig10 = dig100 = ascii_zero;
2775 s = 24;
2776
2777 p = text;
2778 while (1) {
2779 if (((unsigned)(*p - '0')) <= 9 && p < stop) {
2780 p++;
2781 continue;
2782 }
2783
2784 /* here, we have a complete byte between <text> and <p> (exclusive) */
2785 if (p == text)
2786 goto end;
2787
2788 d = p - 1;
2789 dig1 |= (unsigned int)(*d << s);
2790 if (d == text)
2791 goto end;
2792
2793 d--;
2794 dig10 |= (unsigned int)(*d << s);
2795 if (d == text)
2796 goto end;
2797
2798 d--;
2799 dig100 |= (unsigned int)(*d << s);
2800 end:
2801 if (!s || p == stop || *p != '.')
2802 break;
2803
2804 s -= 8;
2805 text = ++p;
2806 }
2807
2808 *ret = p;
2809 dig100 -= ascii_zero;
2810 dig10 -= ascii_zero;
2811 dig1 -= ascii_zero;
2812 return ((dig100 * 10) + dig10) * 10 + dig1;
2813}
2814
Willy Tarreauf0b38bf2010-06-06 13:22:23 +02002815/* Convert a fixed-length string to an IP address. Returns 0 in case of error,
2816 * or the number of chars read in case of success. Maybe this could be replaced
2817 * by one of the functions above. Also, apparently this function does not support
2818 * hosts above 255 and requires exactly 4 octets.
Willy Tarreau075415a2013-12-12 11:29:39 +01002819 * The destination is only modified on success.
Willy Tarreauf0b38bf2010-06-06 13:22:23 +02002820 */
2821int buf2ip(const char *buf, size_t len, struct in_addr *dst)
2822{
2823 const char *addr;
2824 int saw_digit, octets, ch;
2825 u_char tmp[4], *tp;
2826 const char *cp = buf;
2827
2828 saw_digit = 0;
2829 octets = 0;
2830 *(tp = tmp) = 0;
2831
2832 for (addr = buf; addr - buf < len; addr++) {
2833 unsigned char digit = (ch = *addr) - '0';
2834
2835 if (digit > 9 && ch != '.')
2836 break;
2837
2838 if (digit <= 9) {
2839 u_int new = *tp * 10 + digit;
2840
2841 if (new > 255)
2842 return 0;
2843
2844 *tp = new;
2845
2846 if (!saw_digit) {
2847 if (++octets > 4)
2848 return 0;
2849 saw_digit = 1;
2850 }
2851 } else if (ch == '.' && saw_digit) {
2852 if (octets == 4)
2853 return 0;
2854
2855 *++tp = 0;
2856 saw_digit = 0;
2857 } else
2858 return 0;
2859 }
2860
2861 if (octets < 4)
2862 return 0;
2863
2864 memcpy(&dst->s_addr, tmp, 4);
2865 return addr - cp;
2866}
2867
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002868/* This function converts the string in <buf> of the len <len> to
2869 * struct in6_addr <dst> which must be allocated by the caller.
2870 * This function returns 1 in success case, otherwise zero.
Willy Tarreau075415a2013-12-12 11:29:39 +01002871 * The destination is only modified on success.
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002872 */
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002873int buf2ip6(const char *buf, size_t len, struct in6_addr *dst)
2874{
Thierry FOURNIERcd659912013-12-11 12:33:54 +01002875 char null_term_ip6[INET6_ADDRSTRLEN + 1];
Willy Tarreau075415a2013-12-12 11:29:39 +01002876 struct in6_addr out;
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002877
Thierry FOURNIERcd659912013-12-11 12:33:54 +01002878 if (len > INET6_ADDRSTRLEN)
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002879 return 0;
2880
2881 memcpy(null_term_ip6, buf, len);
2882 null_term_ip6[len] = '\0';
2883
Willy Tarreau075415a2013-12-12 11:29:39 +01002884 if (!inet_pton(AF_INET6, null_term_ip6, &out))
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002885 return 0;
2886
Willy Tarreau075415a2013-12-12 11:29:39 +01002887 *dst = out;
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002888 return 1;
2889}
2890
Willy Tarreauacf95772010-06-14 19:09:21 +02002891/* To be used to quote config arg positions. Returns the short string at <ptr>
2892 * surrounded by simple quotes if <ptr> is valid and non-empty, or "end of line"
2893 * if ptr is NULL or empty. The string is locally allocated.
2894 */
2895const char *quote_arg(const char *ptr)
2896{
Christopher Faulet1bc04c72017-10-29 20:14:08 +01002897 static THREAD_LOCAL char val[32];
Willy Tarreauacf95772010-06-14 19:09:21 +02002898 int i;
2899
2900 if (!ptr || !*ptr)
2901 return "end of line";
2902 val[0] = '\'';
Willy Tarreaude2dd6b2013-01-24 02:14:42 +01002903 for (i = 1; i < sizeof(val) - 2 && *ptr; i++)
Willy Tarreauacf95772010-06-14 19:09:21 +02002904 val[i] = *ptr++;
2905 val[i++] = '\'';
2906 val[i] = '\0';
2907 return val;
2908}
2909
Willy Tarreau5b180202010-07-18 10:40:48 +02002910/* returns an operator among STD_OP_* for string <str> or < 0 if unknown */
2911int get_std_op(const char *str)
2912{
2913 int ret = -1;
2914
2915 if (*str == 'e' && str[1] == 'q')
2916 ret = STD_OP_EQ;
2917 else if (*str == 'n' && str[1] == 'e')
2918 ret = STD_OP_NE;
2919 else if (*str == 'l') {
2920 if (str[1] == 'e') ret = STD_OP_LE;
2921 else if (str[1] == 't') ret = STD_OP_LT;
2922 }
2923 else if (*str == 'g') {
2924 if (str[1] == 'e') ret = STD_OP_GE;
2925 else if (str[1] == 't') ret = STD_OP_GT;
2926 }
2927
2928 if (ret == -1 || str[2] != '\0')
2929 return -1;
2930 return ret;
2931}
2932
Willy Tarreau4c14eaa2010-11-24 14:01:45 +01002933/* hash a 32-bit integer to another 32-bit integer */
2934unsigned int full_hash(unsigned int a)
2935{
2936 return __full_hash(a);
2937}
2938
Willy Tarreauf3241112019-02-26 09:56:22 +01002939/* Return the bit position in mask <m> of the nth bit set of rank <r>, between
2940 * 0 and LONGBITS-1 included, starting from the left. For example ranks 0,1,2,3
2941 * for mask 0x55 will be 6, 4, 2 and 0 respectively. This algorithm is based on
2942 * a popcount variant and is described here :
2943 * https://graphics.stanford.edu/~seander/bithacks.html
2944 */
2945unsigned int mask_find_rank_bit(unsigned int r, unsigned long m)
2946{
2947 unsigned long a, b, c, d;
2948 unsigned int s;
2949 unsigned int t;
2950
2951 a = m - ((m >> 1) & ~0UL/3);
2952 b = (a & ~0UL/5) + ((a >> 2) & ~0UL/5);
2953 c = (b + (b >> 4)) & ~0UL/0x11;
2954 d = (c + (c >> 8)) & ~0UL/0x101;
2955
2956 r++; // make r be 1..64
2957
2958 t = 0;
2959 s = LONGBITS;
2960 if (s > 32) {
Willy Tarreau9b6be3b2019-03-18 16:31:18 +01002961 unsigned long d2 = (d >> 16) >> 16;
2962 t = d2 + (d2 >> 16);
Willy Tarreauf3241112019-02-26 09:56:22 +01002963 s -= ((t - r) & 256) >> 3; r -= (t & ((t - r) >> 8));
2964 }
2965
2966 t = (d >> (s - 16)) & 0xff;
2967 s -= ((t - r) & 256) >> 4; r -= (t & ((t - r) >> 8));
2968 t = (c >> (s - 8)) & 0xf;
2969 s -= ((t - r) & 256) >> 5; r -= (t & ((t - r) >> 8));
2970 t = (b >> (s - 4)) & 0x7;
2971 s -= ((t - r) & 256) >> 6; r -= (t & ((t - r) >> 8));
2972 t = (a >> (s - 2)) & 0x3;
2973 s -= ((t - r) & 256) >> 7; r -= (t & ((t - r) >> 8));
2974 t = (m >> (s - 1)) & 0x1;
2975 s -= ((t - r) & 256) >> 8;
2976
2977 return s - 1;
2978}
2979
2980/* Same as mask_find_rank_bit() above but makes use of pre-computed bitmaps
2981 * based on <m>, in <a..d>. These ones must be updated whenever <m> changes
2982 * using mask_prep_rank_map() below.
2983 */
2984unsigned int mask_find_rank_bit_fast(unsigned int r, unsigned long m,
2985 unsigned long a, unsigned long b,
2986 unsigned long c, unsigned long d)
2987{
2988 unsigned int s;
2989 unsigned int t;
2990
2991 r++; // make r be 1..64
2992
2993 t = 0;
2994 s = LONGBITS;
2995 if (s > 32) {
Willy Tarreau9b6be3b2019-03-18 16:31:18 +01002996 unsigned long d2 = (d >> 16) >> 16;
2997 t = d2 + (d2 >> 16);
Willy Tarreauf3241112019-02-26 09:56:22 +01002998 s -= ((t - r) & 256) >> 3; r -= (t & ((t - r) >> 8));
2999 }
3000
3001 t = (d >> (s - 16)) & 0xff;
3002 s -= ((t - r) & 256) >> 4; r -= (t & ((t - r) >> 8));
3003 t = (c >> (s - 8)) & 0xf;
3004 s -= ((t - r) & 256) >> 5; r -= (t & ((t - r) >> 8));
3005 t = (b >> (s - 4)) & 0x7;
3006 s -= ((t - r) & 256) >> 6; r -= (t & ((t - r) >> 8));
3007 t = (a >> (s - 2)) & 0x3;
3008 s -= ((t - r) & 256) >> 7; r -= (t & ((t - r) >> 8));
3009 t = (m >> (s - 1)) & 0x1;
3010 s -= ((t - r) & 256) >> 8;
3011
3012 return s - 1;
3013}
3014
3015/* Prepare the bitmaps used by the fast implementation of the find_rank_bit()
3016 * above.
3017 */
3018void mask_prep_rank_map(unsigned long m,
3019 unsigned long *a, unsigned long *b,
3020 unsigned long *c, unsigned long *d)
3021{
3022 *a = m - ((m >> 1) & ~0UL/3);
3023 *b = (*a & ~0UL/5) + ((*a >> 2) & ~0UL/5);
3024 *c = (*b + (*b >> 4)) & ~0UL/0x11;
3025 *d = (*c + (*c >> 8)) & ~0UL/0x101;
3026}
3027
David du Colombier4f92d322011-03-24 11:09:31 +01003028/* Return non-zero if IPv4 address is part of the network,
Willy Tarreaueec1d382016-07-13 11:59:39 +02003029 * otherwise zero. Note that <addr> may not necessarily be aligned
3030 * while the two other ones must.
David du Colombier4f92d322011-03-24 11:09:31 +01003031 */
Willy Tarreaueec1d382016-07-13 11:59:39 +02003032int in_net_ipv4(const void *addr, const struct in_addr *mask, const struct in_addr *net)
David du Colombier4f92d322011-03-24 11:09:31 +01003033{
Willy Tarreaueec1d382016-07-13 11:59:39 +02003034 struct in_addr addr_copy;
3035
3036 memcpy(&addr_copy, addr, sizeof(addr_copy));
3037 return((addr_copy.s_addr & mask->s_addr) == (net->s_addr & mask->s_addr));
David du Colombier4f92d322011-03-24 11:09:31 +01003038}
3039
3040/* Return non-zero if IPv6 address is part of the network,
Willy Tarreaueec1d382016-07-13 11:59:39 +02003041 * otherwise zero. Note that <addr> may not necessarily be aligned
3042 * while the two other ones must.
David du Colombier4f92d322011-03-24 11:09:31 +01003043 */
Willy Tarreaueec1d382016-07-13 11:59:39 +02003044int in_net_ipv6(const void *addr, const struct in6_addr *mask, const struct in6_addr *net)
David du Colombier4f92d322011-03-24 11:09:31 +01003045{
3046 int i;
Willy Tarreaueec1d382016-07-13 11:59:39 +02003047 struct in6_addr addr_copy;
David du Colombier4f92d322011-03-24 11:09:31 +01003048
Willy Tarreaueec1d382016-07-13 11:59:39 +02003049 memcpy(&addr_copy, addr, sizeof(addr_copy));
David du Colombier4f92d322011-03-24 11:09:31 +01003050 for (i = 0; i < sizeof(struct in6_addr) / sizeof(int); i++)
Willy Tarreaueec1d382016-07-13 11:59:39 +02003051 if (((((int *)&addr_copy)[i] & ((int *)mask)[i])) !=
David du Colombier4f92d322011-03-24 11:09:31 +01003052 (((int *)net)[i] & ((int *)mask)[i]))
3053 return 0;
3054 return 1;
3055}
3056
3057/* RFC 4291 prefix */
3058const char rfc4291_pfx[] = { 0x00, 0x00, 0x00, 0x00,
3059 0x00, 0x00, 0x00, 0x00,
3060 0x00, 0x00, 0xFF, 0xFF };
3061
Joseph Herlant32b83272018-11-15 11:58:28 -08003062/* Map IPv4 address on IPv6 address, as specified in RFC 3513.
Thierry FOURNIER4a04dc32013-11-28 16:33:15 +01003063 * Input and output may overlap.
3064 */
David du Colombier4f92d322011-03-24 11:09:31 +01003065void v4tov6(struct in6_addr *sin6_addr, struct in_addr *sin_addr)
3066{
Thierry FOURNIER4a04dc32013-11-28 16:33:15 +01003067 struct in_addr tmp_addr;
3068
3069 tmp_addr.s_addr = sin_addr->s_addr;
David du Colombier4f92d322011-03-24 11:09:31 +01003070 memcpy(sin6_addr->s6_addr, rfc4291_pfx, sizeof(rfc4291_pfx));
Thierry FOURNIER4a04dc32013-11-28 16:33:15 +01003071 memcpy(sin6_addr->s6_addr+12, &tmp_addr.s_addr, 4);
David du Colombier4f92d322011-03-24 11:09:31 +01003072}
3073
Joseph Herlant32b83272018-11-15 11:58:28 -08003074/* Map IPv6 address on IPv4 address, as specified in RFC 3513.
David du Colombier4f92d322011-03-24 11:09:31 +01003075 * Return true if conversion is possible and false otherwise.
3076 */
3077int v6tov4(struct in_addr *sin_addr, struct in6_addr *sin6_addr)
3078{
3079 if (memcmp(sin6_addr->s6_addr, rfc4291_pfx, sizeof(rfc4291_pfx)) == 0) {
3080 memcpy(&(sin_addr->s_addr), &(sin6_addr->s6_addr[12]),
3081 sizeof(struct in_addr));
3082 return 1;
3083 }
3084
3085 return 0;
3086}
3087
Baptiste Assmann08b24cf2016-01-23 23:39:12 +01003088/* compare two struct sockaddr_storage and return:
3089 * 0 (true) if the addr is the same in both
3090 * 1 (false) if the addr is not the same in both
3091 * -1 (unable) if one of the addr is not AF_INET*
3092 */
3093int ipcmp(struct sockaddr_storage *ss1, struct sockaddr_storage *ss2)
3094{
3095 if ((ss1->ss_family != AF_INET) && (ss1->ss_family != AF_INET6))
3096 return -1;
3097
3098 if ((ss2->ss_family != AF_INET) && (ss2->ss_family != AF_INET6))
3099 return -1;
3100
3101 if (ss1->ss_family != ss2->ss_family)
3102 return 1;
3103
3104 switch (ss1->ss_family) {
3105 case AF_INET:
3106 return memcmp(&((struct sockaddr_in *)ss1)->sin_addr,
3107 &((struct sockaddr_in *)ss2)->sin_addr,
3108 sizeof(struct in_addr)) != 0;
3109 case AF_INET6:
3110 return memcmp(&((struct sockaddr_in6 *)ss1)->sin6_addr,
3111 &((struct sockaddr_in6 *)ss2)->sin6_addr,
3112 sizeof(struct in6_addr)) != 0;
3113 }
3114
3115 return 1;
3116}
3117
Christopher Faulet9553de72021-02-26 09:12:50 +01003118/* compare a struct sockaddr_storage to a struct net_addr and return :
3119 * 0 (true) if <addr> is matching <net>
3120 * 1 (false) if <addr> is not matching <net>
3121 * -1 (unable) if <addr> or <net> is not AF_INET*
3122 */
3123int ipcmp2net(const struct sockaddr_storage *addr, const struct net_addr *net)
3124{
3125 if ((addr->ss_family != AF_INET) && (addr->ss_family != AF_INET6))
3126 return -1;
3127
3128 if ((net->family != AF_INET) && (net->family != AF_INET6))
3129 return -1;
3130
3131 if (addr->ss_family != net->family)
3132 return 1;
3133
3134 if (addr->ss_family == AF_INET &&
3135 (((struct sockaddr_in *)addr)->sin_addr.s_addr & net->addr.v4.mask.s_addr) == net->addr.v4.ip.s_addr)
3136 return 0;
3137 else {
3138 const struct in6_addr *addr6 = &(((const struct sockaddr_in6*)addr)->sin6_addr);
3139 const struct in6_addr *nip6 = &net->addr.v6.ip;
3140 const struct in6_addr *nmask6 = &net->addr.v6.mask;
3141
3142 if ((read_u32(&addr6->s6_addr[0]) & read_u32(&nmask6->s6_addr[0])) == read_u32(&nip6->s6_addr[0]) &&
3143 (read_u32(&addr6->s6_addr[4]) & read_u32(&nmask6->s6_addr[4])) == read_u32(&nip6->s6_addr[4]) &&
3144 (read_u32(&addr6->s6_addr[8]) & read_u32(&nmask6->s6_addr[8])) == read_u32(&nip6->s6_addr[8]) &&
3145 (read_u32(&addr6->s6_addr[12]) & read_u32(&nmask6->s6_addr[12])) == read_u32(&nip6->s6_addr[12]))
3146 return 0;
3147 }
3148
3149 return 1;
3150}
3151
Baptiste Assmann08396c82016-01-31 00:27:17 +01003152/* copy IP address from <source> into <dest>
Willy Tarreaudc3a9e82016-11-04 18:47:01 +01003153 * The caller must allocate and clear <dest> before calling.
3154 * The source must be in either AF_INET or AF_INET6 family, or the destination
3155 * address will be undefined. If the destination address used to hold a port,
3156 * it is preserved, so that this function can be used to switch to another
3157 * address family with no risk. Returns a pointer to the destination.
Baptiste Assmann08396c82016-01-31 00:27:17 +01003158 */
3159struct sockaddr_storage *ipcpy(struct sockaddr_storage *source, struct sockaddr_storage *dest)
3160{
Willy Tarreaudc3a9e82016-11-04 18:47:01 +01003161 int prev_port;
3162
3163 prev_port = get_net_port(dest);
3164 memset(dest, 0, sizeof(*dest));
Baptiste Assmann08396c82016-01-31 00:27:17 +01003165 dest->ss_family = source->ss_family;
3166
3167 /* copy new addr and apply it */
3168 switch (source->ss_family) {
3169 case AF_INET:
3170 ((struct sockaddr_in *)dest)->sin_addr.s_addr = ((struct sockaddr_in *)source)->sin_addr.s_addr;
Willy Tarreaudc3a9e82016-11-04 18:47:01 +01003171 ((struct sockaddr_in *)dest)->sin_port = prev_port;
Baptiste Assmann08396c82016-01-31 00:27:17 +01003172 break;
3173 case AF_INET6:
3174 memcpy(((struct sockaddr_in6 *)dest)->sin6_addr.s6_addr, ((struct sockaddr_in6 *)source)->sin6_addr.s6_addr, sizeof(struct in6_addr));
Willy Tarreaudc3a9e82016-11-04 18:47:01 +01003175 ((struct sockaddr_in6 *)dest)->sin6_port = prev_port;
Baptiste Assmann08396c82016-01-31 00:27:17 +01003176 break;
3177 }
3178
3179 return dest;
3180}
3181
William Lallemand421f5b52012-02-06 18:15:57 +01003182char *human_time(int t, short hz_div) {
3183 static char rv[sizeof("24855d23h")+1]; // longest of "23h59m" and "59m59s"
3184 char *p = rv;
Willy Tarreau761b3d52014-04-14 14:53:06 +02003185 char *end = rv + sizeof(rv);
William Lallemand421f5b52012-02-06 18:15:57 +01003186 int cnt=2; // print two numbers
3187
3188 if (unlikely(t < 0 || hz_div <= 0)) {
Willy Tarreau761b3d52014-04-14 14:53:06 +02003189 snprintf(p, end - p, "?");
William Lallemand421f5b52012-02-06 18:15:57 +01003190 return rv;
3191 }
3192
3193 if (unlikely(hz_div > 1))
3194 t /= hz_div;
3195
3196 if (t >= DAY) {
Willy Tarreau761b3d52014-04-14 14:53:06 +02003197 p += snprintf(p, end - p, "%dd", t / DAY);
William Lallemand421f5b52012-02-06 18:15:57 +01003198 cnt--;
3199 }
3200
3201 if (cnt && t % DAY / HOUR) {
Willy Tarreau761b3d52014-04-14 14:53:06 +02003202 p += snprintf(p, end - p, "%dh", t % DAY / HOUR);
William Lallemand421f5b52012-02-06 18:15:57 +01003203 cnt--;
3204 }
3205
3206 if (cnt && t % HOUR / MINUTE) {
Willy Tarreau761b3d52014-04-14 14:53:06 +02003207 p += snprintf(p, end - p, "%dm", t % HOUR / MINUTE);
William Lallemand421f5b52012-02-06 18:15:57 +01003208 cnt--;
3209 }
3210
3211 if ((cnt && t % MINUTE) || !t) // also display '0s'
Willy Tarreau761b3d52014-04-14 14:53:06 +02003212 p += snprintf(p, end - p, "%ds", t % MINUTE / SEC);
William Lallemand421f5b52012-02-06 18:15:57 +01003213
3214 return rv;
3215}
3216
3217const char *monthname[12] = {
3218 "Jan", "Feb", "Mar", "Apr", "May", "Jun",
3219 "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
3220};
3221
3222/* date2str_log: write a date in the format :
3223 * sprintf(str, "%02d/%s/%04d:%02d:%02d:%02d.%03d",
3224 * tm.tm_mday, monthname[tm.tm_mon], tm.tm_year+1900,
3225 * tm.tm_hour, tm.tm_min, tm.tm_sec, (int)date.tv_usec/1000);
3226 *
3227 * without using sprintf. return a pointer to the last char written (\0) or
3228 * NULL if there isn't enough space.
3229 */
Willy Tarreauf16cb412018-09-04 19:08:48 +02003230char *date2str_log(char *dst, const struct tm *tm, const struct timeval *date, size_t size)
William Lallemand421f5b52012-02-06 18:15:57 +01003231{
3232
3233 if (size < 25) /* the size is fixed: 24 chars + \0 */
3234 return NULL;
3235
3236 dst = utoa_pad((unsigned int)tm->tm_mday, dst, 3); // day
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003237 if (!dst)
3238 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003239 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003240
William Lallemand421f5b52012-02-06 18:15:57 +01003241 memcpy(dst, monthname[tm->tm_mon], 3); // month
3242 dst += 3;
3243 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003244
William Lallemand421f5b52012-02-06 18:15:57 +01003245 dst = utoa_pad((unsigned int)tm->tm_year+1900, dst, 5); // year
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003246 if (!dst)
3247 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003248 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003249
William Lallemand421f5b52012-02-06 18:15:57 +01003250 dst = utoa_pad((unsigned int)tm->tm_hour, dst, 3); // hour
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003251 if (!dst)
3252 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003253 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003254
William Lallemand421f5b52012-02-06 18:15:57 +01003255 dst = utoa_pad((unsigned int)tm->tm_min, dst, 3); // minutes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003256 if (!dst)
3257 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003258 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003259
William Lallemand421f5b52012-02-06 18:15:57 +01003260 dst = utoa_pad((unsigned int)tm->tm_sec, dst, 3); // secondes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003261 if (!dst)
3262 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003263 *dst++ = '.';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003264
Willy Tarreau7d9421d2020-02-29 09:08:02 +01003265 dst = utoa_pad((unsigned int)(date->tv_usec/1000)%1000, dst, 4); // milliseconds
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003266 if (!dst)
3267 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003268 *dst = '\0';
3269
3270 return dst;
3271}
3272
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003273/* Base year used to compute leap years */
3274#define TM_YEAR_BASE 1900
3275
3276/* Return the difference in seconds between two times (leap seconds are ignored).
3277 * Retrieved from glibc 2.18 source code.
3278 */
3279static int my_tm_diff(const struct tm *a, const struct tm *b)
3280{
3281 /* Compute intervening leap days correctly even if year is negative.
3282 * Take care to avoid int overflow in leap day calculations,
3283 * but it's OK to assume that A and B are close to each other.
3284 */
3285 int a4 = (a->tm_year >> 2) + (TM_YEAR_BASE >> 2) - ! (a->tm_year & 3);
3286 int b4 = (b->tm_year >> 2) + (TM_YEAR_BASE >> 2) - ! (b->tm_year & 3);
3287 int a100 = a4 / 25 - (a4 % 25 < 0);
3288 int b100 = b4 / 25 - (b4 % 25 < 0);
3289 int a400 = a100 >> 2;
3290 int b400 = b100 >> 2;
3291 int intervening_leap_days = (a4 - b4) - (a100 - b100) + (a400 - b400);
3292 int years = a->tm_year - b->tm_year;
3293 int days = (365 * years + intervening_leap_days
3294 + (a->tm_yday - b->tm_yday));
3295 return (60 * (60 * (24 * days + (a->tm_hour - b->tm_hour))
3296 + (a->tm_min - b->tm_min))
3297 + (a->tm_sec - b->tm_sec));
3298}
3299
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003300/* Return the GMT offset for a specific local time.
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003301 * Both t and tm must represent the same time.
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003302 * The string returned has the same format as returned by strftime(... "%z", tm).
3303 * Offsets are kept in an internal cache for better performances.
3304 */
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003305const char *get_gmt_offset(time_t t, struct tm *tm)
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003306{
3307 /* Cache offsets from GMT (depending on whether DST is active or not) */
Christopher Faulet1bc04c72017-10-29 20:14:08 +01003308 static THREAD_LOCAL char gmt_offsets[2][5+1] = { "", "" };
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003309
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003310 char *gmt_offset;
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003311 struct tm tm_gmt;
3312 int diff;
3313 int isdst = tm->tm_isdst;
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003314
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003315 /* Pretend DST not active if its status is unknown */
3316 if (isdst < 0)
3317 isdst = 0;
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003318
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003319 /* Fetch the offset and initialize it if needed */
3320 gmt_offset = gmt_offsets[isdst & 0x01];
3321 if (unlikely(!*gmt_offset)) {
3322 get_gmtime(t, &tm_gmt);
3323 diff = my_tm_diff(tm, &tm_gmt);
3324 if (diff < 0) {
3325 diff = -diff;
3326 *gmt_offset = '-';
3327 } else {
3328 *gmt_offset = '+';
3329 }
Willy Tarreaue112c8a2019-10-29 10:16:11 +01003330 diff %= 86400U;
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003331 diff /= 60; /* Convert to minutes */
3332 snprintf(gmt_offset+1, 4+1, "%02d%02d", diff/60, diff%60);
3333 }
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003334
Willy Tarreaue112c8a2019-10-29 10:16:11 +01003335 return gmt_offset;
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003336}
3337
William Lallemand421f5b52012-02-06 18:15:57 +01003338/* gmt2str_log: write a date in the format :
3339 * "%02d/%s/%04d:%02d:%02d:%02d +0000" without using snprintf
3340 * return a pointer to the last char written (\0) or
3341 * NULL if there isn't enough space.
3342 */
3343char *gmt2str_log(char *dst, struct tm *tm, size_t size)
3344{
Yuxans Yao4e25b012012-10-19 10:36:09 +08003345 if (size < 27) /* the size is fixed: 26 chars + \0 */
William Lallemand421f5b52012-02-06 18:15:57 +01003346 return NULL;
3347
3348 dst = utoa_pad((unsigned int)tm->tm_mday, dst, 3); // day
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003349 if (!dst)
3350 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003351 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003352
William Lallemand421f5b52012-02-06 18:15:57 +01003353 memcpy(dst, monthname[tm->tm_mon], 3); // month
3354 dst += 3;
3355 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003356
William Lallemand421f5b52012-02-06 18:15:57 +01003357 dst = utoa_pad((unsigned int)tm->tm_year+1900, dst, 5); // year
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003358 if (!dst)
3359 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003360 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003361
William Lallemand421f5b52012-02-06 18:15:57 +01003362 dst = utoa_pad((unsigned int)tm->tm_hour, dst, 3); // hour
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003363 if (!dst)
3364 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003365 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003366
William Lallemand421f5b52012-02-06 18:15:57 +01003367 dst = utoa_pad((unsigned int)tm->tm_min, dst, 3); // minutes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003368 if (!dst)
3369 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003370 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003371
William Lallemand421f5b52012-02-06 18:15:57 +01003372 dst = utoa_pad((unsigned int)tm->tm_sec, dst, 3); // secondes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003373 if (!dst)
3374 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003375 *dst++ = ' ';
3376 *dst++ = '+';
3377 *dst++ = '0';
3378 *dst++ = '0';
3379 *dst++ = '0';
3380 *dst++ = '0';
3381 *dst = '\0';
3382
3383 return dst;
3384}
3385
Yuxans Yao4e25b012012-10-19 10:36:09 +08003386/* localdate2str_log: write a date in the format :
3387 * "%02d/%s/%04d:%02d:%02d:%02d +0000(local timezone)" without using snprintf
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003388 * Both t and tm must represent the same time.
3389 * return a pointer to the last char written (\0) or
3390 * NULL if there isn't enough space.
Yuxans Yao4e25b012012-10-19 10:36:09 +08003391 */
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003392char *localdate2str_log(char *dst, time_t t, struct tm *tm, size_t size)
Yuxans Yao4e25b012012-10-19 10:36:09 +08003393{
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003394 const char *gmt_offset;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003395 if (size < 27) /* the size is fixed: 26 chars + \0 */
3396 return NULL;
3397
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003398 gmt_offset = get_gmt_offset(t, tm);
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003399
Yuxans Yao4e25b012012-10-19 10:36:09 +08003400 dst = utoa_pad((unsigned int)tm->tm_mday, dst, 3); // day
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003401 if (!dst)
3402 return NULL;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003403 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003404
Yuxans Yao4e25b012012-10-19 10:36:09 +08003405 memcpy(dst, monthname[tm->tm_mon], 3); // month
3406 dst += 3;
3407 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003408
Yuxans Yao4e25b012012-10-19 10:36:09 +08003409 dst = utoa_pad((unsigned int)tm->tm_year+1900, dst, 5); // year
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003410 if (!dst)
3411 return NULL;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003412 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003413
Yuxans Yao4e25b012012-10-19 10:36:09 +08003414 dst = utoa_pad((unsigned int)tm->tm_hour, dst, 3); // hour
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003415 if (!dst)
3416 return NULL;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003417 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003418
Yuxans Yao4e25b012012-10-19 10:36:09 +08003419 dst = utoa_pad((unsigned int)tm->tm_min, dst, 3); // minutes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003420 if (!dst)
3421 return NULL;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003422 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003423
Yuxans Yao4e25b012012-10-19 10:36:09 +08003424 dst = utoa_pad((unsigned int)tm->tm_sec, dst, 3); // secondes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003425 if (!dst)
3426 return NULL;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003427 *dst++ = ' ';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003428
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003429 memcpy(dst, gmt_offset, 5); // Offset from local time to GMT
Yuxans Yao4e25b012012-10-19 10:36:09 +08003430 dst += 5;
3431 *dst = '\0';
3432
3433 return dst;
3434}
3435
Willy Tarreaucb1949b2017-07-19 19:05:29 +02003436/* Returns the number of seconds since 01/01/1970 0:0:0 GMT for GMT date <tm>.
3437 * It is meant as a portable replacement for timegm() for use with valid inputs.
3438 * Returns undefined results for invalid dates (eg: months out of range 0..11).
3439 */
3440time_t my_timegm(const struct tm *tm)
3441{
3442 /* Each month has 28, 29, 30 or 31 days, or 28+N. The date in the year
3443 * is thus (current month - 1)*28 + cumulated_N[month] to count the
3444 * sum of the extra N days for elapsed months. The sum of all these N
3445 * days doesn't exceed 30 for a complete year (366-12*28) so it fits
3446 * in a 5-bit word. This means that with 60 bits we can represent a
3447 * matrix of all these values at once, which is fast and efficient to
3448 * access. The extra February day for leap years is not counted here.
3449 *
3450 * Jan : none = 0 (0)
3451 * Feb : Jan = 3 (3)
3452 * Mar : Jan..Feb = 3 (3 + 0)
3453 * Apr : Jan..Mar = 6 (3 + 0 + 3)
3454 * May : Jan..Apr = 8 (3 + 0 + 3 + 2)
3455 * Jun : Jan..May = 11 (3 + 0 + 3 + 2 + 3)
3456 * Jul : Jan..Jun = 13 (3 + 0 + 3 + 2 + 3 + 2)
3457 * Aug : Jan..Jul = 16 (3 + 0 + 3 + 2 + 3 + 2 + 3)
3458 * Sep : Jan..Aug = 19 (3 + 0 + 3 + 2 + 3 + 2 + 3 + 3)
3459 * Oct : Jan..Sep = 21 (3 + 0 + 3 + 2 + 3 + 2 + 3 + 3 + 2)
3460 * Nov : Jan..Oct = 24 (3 + 0 + 3 + 2 + 3 + 2 + 3 + 3 + 2 + 3)
3461 * Dec : Jan..Nov = 26 (3 + 0 + 3 + 2 + 3 + 2 + 3 + 3 + 2 + 3 + 2)
3462 */
3463 uint64_t extra =
3464 ( 0ULL << 0*5) + ( 3ULL << 1*5) + ( 3ULL << 2*5) + /* Jan, Feb, Mar, */
3465 ( 6ULL << 3*5) + ( 8ULL << 4*5) + (11ULL << 5*5) + /* Apr, May, Jun, */
3466 (13ULL << 6*5) + (16ULL << 7*5) + (19ULL << 8*5) + /* Jul, Aug, Sep, */
3467 (21ULL << 9*5) + (24ULL << 10*5) + (26ULL << 11*5); /* Oct, Nov, Dec, */
3468
3469 unsigned int y = tm->tm_year + 1900;
3470 unsigned int m = tm->tm_mon;
3471 unsigned long days = 0;
3472
3473 /* days since 1/1/1970 for full years */
3474 days += days_since_zero(y) - days_since_zero(1970);
3475
3476 /* days for full months in the current year */
3477 days += 28 * m + ((extra >> (m * 5)) & 0x1f);
3478
3479 /* count + 1 after March for leap years. A leap year is a year multiple
3480 * of 4, unless it's multiple of 100 without being multiple of 400. 2000
3481 * is leap, 1900 isn't, 1904 is.
3482 */
3483 if ((m > 1) && !(y & 3) && ((y % 100) || !(y % 400)))
3484 days++;
3485
3486 days += tm->tm_mday - 1;
3487 return days * 86400ULL + tm->tm_hour * 3600 + tm->tm_min * 60 + tm->tm_sec;
3488}
3489
Thierry Fournier93127942016-01-20 18:49:45 +01003490/* This function check a char. It returns true and updates
3491 * <date> and <len> pointer to the new position if the
3492 * character is found.
3493 */
3494static inline int parse_expect_char(const char **date, int *len, char c)
3495{
3496 if (*len < 1 || **date != c)
3497 return 0;
3498 (*len)--;
3499 (*date)++;
3500 return 1;
3501}
3502
3503/* This function expects a string <str> of len <l>. It return true and updates.
3504 * <date> and <len> if the string matches, otherwise, it returns false.
3505 */
3506static inline int parse_strcmp(const char **date, int *len, char *str, int l)
3507{
3508 if (*len < l || strncmp(*date, str, l) != 0)
3509 return 0;
3510 (*len) -= l;
3511 (*date) += l;
3512 return 1;
3513}
3514
3515/* This macro converts 3 chars name in integer. */
3516#define STR2I3(__a, __b, __c) ((__a) * 65536 + (__b) * 256 + (__c))
3517
3518/* day-name = %x4D.6F.6E ; "Mon", case-sensitive
3519 * / %x54.75.65 ; "Tue", case-sensitive
3520 * / %x57.65.64 ; "Wed", case-sensitive
3521 * / %x54.68.75 ; "Thu", case-sensitive
3522 * / %x46.72.69 ; "Fri", case-sensitive
3523 * / %x53.61.74 ; "Sat", case-sensitive
3524 * / %x53.75.6E ; "Sun", case-sensitive
3525 *
3526 * This array must be alphabetically sorted
3527 */
3528static inline int parse_http_dayname(const char **date, int *len, struct tm *tm)
3529{
3530 if (*len < 3)
3531 return 0;
3532 switch (STR2I3((*date)[0], (*date)[1], (*date)[2])) {
3533 case STR2I3('M','o','n'): tm->tm_wday = 1; break;
3534 case STR2I3('T','u','e'): tm->tm_wday = 2; break;
3535 case STR2I3('W','e','d'): tm->tm_wday = 3; break;
3536 case STR2I3('T','h','u'): tm->tm_wday = 4; break;
3537 case STR2I3('F','r','i'): tm->tm_wday = 5; break;
3538 case STR2I3('S','a','t'): tm->tm_wday = 6; break;
3539 case STR2I3('S','u','n'): tm->tm_wday = 7; break;
3540 default: return 0;
3541 }
3542 *len -= 3;
3543 *date += 3;
3544 return 1;
3545}
3546
3547/* month = %x4A.61.6E ; "Jan", case-sensitive
3548 * / %x46.65.62 ; "Feb", case-sensitive
3549 * / %x4D.61.72 ; "Mar", case-sensitive
3550 * / %x41.70.72 ; "Apr", case-sensitive
3551 * / %x4D.61.79 ; "May", case-sensitive
3552 * / %x4A.75.6E ; "Jun", case-sensitive
3553 * / %x4A.75.6C ; "Jul", case-sensitive
3554 * / %x41.75.67 ; "Aug", case-sensitive
3555 * / %x53.65.70 ; "Sep", case-sensitive
3556 * / %x4F.63.74 ; "Oct", case-sensitive
3557 * / %x4E.6F.76 ; "Nov", case-sensitive
3558 * / %x44.65.63 ; "Dec", case-sensitive
3559 *
3560 * This array must be alphabetically sorted
3561 */
3562static inline int parse_http_monthname(const char **date, int *len, struct tm *tm)
3563{
3564 if (*len < 3)
3565 return 0;
3566 switch (STR2I3((*date)[0], (*date)[1], (*date)[2])) {
3567 case STR2I3('J','a','n'): tm->tm_mon = 0; break;
3568 case STR2I3('F','e','b'): tm->tm_mon = 1; break;
3569 case STR2I3('M','a','r'): tm->tm_mon = 2; break;
3570 case STR2I3('A','p','r'): tm->tm_mon = 3; break;
3571 case STR2I3('M','a','y'): tm->tm_mon = 4; break;
3572 case STR2I3('J','u','n'): tm->tm_mon = 5; break;
3573 case STR2I3('J','u','l'): tm->tm_mon = 6; break;
3574 case STR2I3('A','u','g'): tm->tm_mon = 7; break;
3575 case STR2I3('S','e','p'): tm->tm_mon = 8; break;
3576 case STR2I3('O','c','t'): tm->tm_mon = 9; break;
3577 case STR2I3('N','o','v'): tm->tm_mon = 10; break;
3578 case STR2I3('D','e','c'): tm->tm_mon = 11; break;
3579 default: return 0;
3580 }
3581 *len -= 3;
3582 *date += 3;
3583 return 1;
3584}
3585
3586/* day-name-l = %x4D.6F.6E.64.61.79 ; "Monday", case-sensitive
3587 * / %x54.75.65.73.64.61.79 ; "Tuesday", case-sensitive
3588 * / %x57.65.64.6E.65.73.64.61.79 ; "Wednesday", case-sensitive
3589 * / %x54.68.75.72.73.64.61.79 ; "Thursday", case-sensitive
3590 * / %x46.72.69.64.61.79 ; "Friday", case-sensitive
3591 * / %x53.61.74.75.72.64.61.79 ; "Saturday", case-sensitive
3592 * / %x53.75.6E.64.61.79 ; "Sunday", case-sensitive
3593 *
3594 * This array must be alphabetically sorted
3595 */
3596static inline int parse_http_ldayname(const char **date, int *len, struct tm *tm)
3597{
3598 if (*len < 6) /* Minimum length. */
3599 return 0;
3600 switch (STR2I3((*date)[0], (*date)[1], (*date)[2])) {
3601 case STR2I3('M','o','n'):
3602 RET0_UNLESS(parse_strcmp(date, len, "Monday", 6));
3603 tm->tm_wday = 1;
3604 return 1;
3605 case STR2I3('T','u','e'):
3606 RET0_UNLESS(parse_strcmp(date, len, "Tuesday", 7));
3607 tm->tm_wday = 2;
3608 return 1;
3609 case STR2I3('W','e','d'):
3610 RET0_UNLESS(parse_strcmp(date, len, "Wednesday", 9));
3611 tm->tm_wday = 3;
3612 return 1;
3613 case STR2I3('T','h','u'):
3614 RET0_UNLESS(parse_strcmp(date, len, "Thursday", 8));
3615 tm->tm_wday = 4;
3616 return 1;
3617 case STR2I3('F','r','i'):
3618 RET0_UNLESS(parse_strcmp(date, len, "Friday", 6));
3619 tm->tm_wday = 5;
3620 return 1;
3621 case STR2I3('S','a','t'):
3622 RET0_UNLESS(parse_strcmp(date, len, "Saturday", 8));
3623 tm->tm_wday = 6;
3624 return 1;
3625 case STR2I3('S','u','n'):
3626 RET0_UNLESS(parse_strcmp(date, len, "Sunday", 6));
3627 tm->tm_wday = 7;
3628 return 1;
3629 }
3630 return 0;
3631}
3632
3633/* This function parses exactly 1 digit and returns the numeric value in "digit". */
3634static inline int parse_digit(const char **date, int *len, int *digit)
3635{
3636 if (*len < 1 || **date < '0' || **date > '9')
3637 return 0;
3638 *digit = (**date - '0');
3639 (*date)++;
3640 (*len)--;
3641 return 1;
3642}
3643
3644/* This function parses exactly 2 digits and returns the numeric value in "digit". */
3645static inline int parse_2digit(const char **date, int *len, int *digit)
3646{
3647 int value;
3648
3649 RET0_UNLESS(parse_digit(date, len, &value));
3650 (*digit) = value * 10;
3651 RET0_UNLESS(parse_digit(date, len, &value));
3652 (*digit) += value;
3653
3654 return 1;
3655}
3656
3657/* This function parses exactly 4 digits and returns the numeric value in "digit". */
3658static inline int parse_4digit(const char **date, int *len, int *digit)
3659{
3660 int value;
3661
3662 RET0_UNLESS(parse_digit(date, len, &value));
3663 (*digit) = value * 1000;
3664
3665 RET0_UNLESS(parse_digit(date, len, &value));
3666 (*digit) += value * 100;
3667
3668 RET0_UNLESS(parse_digit(date, len, &value));
3669 (*digit) += value * 10;
3670
3671 RET0_UNLESS(parse_digit(date, len, &value));
3672 (*digit) += value;
3673
3674 return 1;
3675}
3676
3677/* time-of-day = hour ":" minute ":" second
3678 * ; 00:00:00 - 23:59:60 (leap second)
3679 *
3680 * hour = 2DIGIT
3681 * minute = 2DIGIT
3682 * second = 2DIGIT
3683 */
3684static inline int parse_http_time(const char **date, int *len, struct tm *tm)
3685{
3686 RET0_UNLESS(parse_2digit(date, len, &tm->tm_hour)); /* hour 2DIGIT */
3687 RET0_UNLESS(parse_expect_char(date, len, ':')); /* expect ":" */
3688 RET0_UNLESS(parse_2digit(date, len, &tm->tm_min)); /* min 2DIGIT */
3689 RET0_UNLESS(parse_expect_char(date, len, ':')); /* expect ":" */
3690 RET0_UNLESS(parse_2digit(date, len, &tm->tm_sec)); /* sec 2DIGIT */
3691 return 1;
3692}
3693
3694/* From RFC7231
3695 * https://tools.ietf.org/html/rfc7231#section-7.1.1.1
3696 *
3697 * IMF-fixdate = day-name "," SP date1 SP time-of-day SP GMT
3698 * ; fixed length/zone/capitalization subset of the format
3699 * ; see Section 3.3 of [RFC5322]
3700 *
3701 *
3702 * date1 = day SP month SP year
3703 * ; e.g., 02 Jun 1982
3704 *
3705 * day = 2DIGIT
3706 * year = 4DIGIT
3707 *
3708 * GMT = %x47.4D.54 ; "GMT", case-sensitive
3709 *
3710 * time-of-day = hour ":" minute ":" second
3711 * ; 00:00:00 - 23:59:60 (leap second)
3712 *
3713 * hour = 2DIGIT
3714 * minute = 2DIGIT
3715 * second = 2DIGIT
3716 *
3717 * DIGIT = decimal 0-9
3718 */
3719int parse_imf_date(const char *date, int len, struct tm *tm)
3720{
David Carlier327298c2016-11-20 10:42:38 +00003721 /* tm_gmtoff, if present, ought to be zero'ed */
3722 memset(tm, 0, sizeof(*tm));
3723
Thierry Fournier93127942016-01-20 18:49:45 +01003724 RET0_UNLESS(parse_http_dayname(&date, &len, tm)); /* day-name */
3725 RET0_UNLESS(parse_expect_char(&date, &len, ',')); /* expect "," */
3726 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3727 RET0_UNLESS(parse_2digit(&date, &len, &tm->tm_mday)); /* day 2DIGIT */
3728 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3729 RET0_UNLESS(parse_http_monthname(&date, &len, tm)); /* Month */
3730 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3731 RET0_UNLESS(parse_4digit(&date, &len, &tm->tm_year)); /* year = 4DIGIT */
3732 tm->tm_year -= 1900;
3733 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3734 RET0_UNLESS(parse_http_time(&date, &len, tm)); /* Parse time. */
3735 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3736 RET0_UNLESS(parse_strcmp(&date, &len, "GMT", 3)); /* GMT = %x47.4D.54 ; "GMT", case-sensitive */
3737 tm->tm_isdst = -1;
Thierry Fournier93127942016-01-20 18:49:45 +01003738 return 1;
3739}
3740
3741/* From RFC7231
3742 * https://tools.ietf.org/html/rfc7231#section-7.1.1.1
3743 *
3744 * rfc850-date = day-name-l "," SP date2 SP time-of-day SP GMT
3745 * date2 = day "-" month "-" 2DIGIT
3746 * ; e.g., 02-Jun-82
3747 *
3748 * day = 2DIGIT
3749 */
3750int parse_rfc850_date(const char *date, int len, struct tm *tm)
3751{
3752 int year;
3753
David Carlier327298c2016-11-20 10:42:38 +00003754 /* tm_gmtoff, if present, ought to be zero'ed */
3755 memset(tm, 0, sizeof(*tm));
3756
Thierry Fournier93127942016-01-20 18:49:45 +01003757 RET0_UNLESS(parse_http_ldayname(&date, &len, tm)); /* Read the day name */
3758 RET0_UNLESS(parse_expect_char(&date, &len, ',')); /* expect "," */
3759 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3760 RET0_UNLESS(parse_2digit(&date, &len, &tm->tm_mday)); /* day 2DIGIT */
3761 RET0_UNLESS(parse_expect_char(&date, &len, '-')); /* expect "-" */
3762 RET0_UNLESS(parse_http_monthname(&date, &len, tm)); /* Month */
3763 RET0_UNLESS(parse_expect_char(&date, &len, '-')); /* expect "-" */
3764
3765 /* year = 2DIGIT
3766 *
3767 * Recipients of a timestamp value in rfc850-(*date) format, which uses a
3768 * two-digit year, MUST interpret a timestamp that appears to be more
3769 * than 50 years in the future as representing the most recent year in
3770 * the past that had the same last two digits.
3771 */
3772 RET0_UNLESS(parse_2digit(&date, &len, &tm->tm_year));
3773
3774 /* expect SP */
3775 if (!parse_expect_char(&date, &len, ' ')) {
3776 /* Maybe we have the date with 4 digits. */
3777 RET0_UNLESS(parse_2digit(&date, &len, &year));
3778 tm->tm_year = (tm->tm_year * 100 + year) - 1900;
3779 /* expect SP */
3780 RET0_UNLESS(parse_expect_char(&date, &len, ' '));
3781 } else {
3782 /* I fix 60 as pivot: >60: +1900, <60: +2000. Note that the
3783 * tm_year is the number of year since 1900, so for +1900, we
3784 * do nothing, and for +2000, we add 100.
3785 */
3786 if (tm->tm_year <= 60)
3787 tm->tm_year += 100;
3788 }
3789
3790 RET0_UNLESS(parse_http_time(&date, &len, tm)); /* Parse time. */
3791 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3792 RET0_UNLESS(parse_strcmp(&date, &len, "GMT", 3)); /* GMT = %x47.4D.54 ; "GMT", case-sensitive */
3793 tm->tm_isdst = -1;
Thierry Fournier93127942016-01-20 18:49:45 +01003794
3795 return 1;
3796}
3797
3798/* From RFC7231
3799 * https://tools.ietf.org/html/rfc7231#section-7.1.1.1
3800 *
3801 * asctime-date = day-name SP date3 SP time-of-day SP year
3802 * date3 = month SP ( 2DIGIT / ( SP 1DIGIT ))
3803 * ; e.g., Jun 2
3804 *
3805 * HTTP-date is case sensitive. A sender MUST NOT generate additional
3806 * whitespace in an HTTP-date beyond that specifically included as SP in
3807 * the grammar.
3808 */
3809int parse_asctime_date(const char *date, int len, struct tm *tm)
3810{
David Carlier327298c2016-11-20 10:42:38 +00003811 /* tm_gmtoff, if present, ought to be zero'ed */
3812 memset(tm, 0, sizeof(*tm));
3813
Thierry Fournier93127942016-01-20 18:49:45 +01003814 RET0_UNLESS(parse_http_dayname(&date, &len, tm)); /* day-name */
3815 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3816 RET0_UNLESS(parse_http_monthname(&date, &len, tm)); /* expect month */
3817 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3818
3819 /* expect SP and 1DIGIT or 2DIGIT */
3820 if (parse_expect_char(&date, &len, ' '))
3821 RET0_UNLESS(parse_digit(&date, &len, &tm->tm_mday));
3822 else
3823 RET0_UNLESS(parse_2digit(&date, &len, &tm->tm_mday));
3824
3825 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3826 RET0_UNLESS(parse_http_time(&date, &len, tm)); /* Parse time. */
3827 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3828 RET0_UNLESS(parse_4digit(&date, &len, &tm->tm_year)); /* year = 4DIGIT */
3829 tm->tm_year -= 1900;
3830 tm->tm_isdst = -1;
Thierry Fournier93127942016-01-20 18:49:45 +01003831 return 1;
3832}
3833
3834/* From RFC7231
3835 * https://tools.ietf.org/html/rfc7231#section-7.1.1.1
3836 *
3837 * HTTP-date = IMF-fixdate / obs-date
3838 * obs-date = rfc850-date / asctime-date
3839 *
3840 * parses an HTTP date in the RFC format and is accepted
3841 * alternatives. <date> is the strinf containing the date,
3842 * len is the len of the string. <tm> is filled with the
3843 * parsed time. We must considers this time as GMT.
3844 */
3845int parse_http_date(const char *date, int len, struct tm *tm)
3846{
3847 if (parse_imf_date(date, len, tm))
3848 return 1;
3849
3850 if (parse_rfc850_date(date, len, tm))
3851 return 1;
3852
3853 if (parse_asctime_date(date, len, tm))
3854 return 1;
3855
3856 return 0;
3857}
3858
Willy Tarreau4deeb102021-01-29 10:47:52 +01003859/* print the time <ns> in a short form (exactly 7 chars) at the end of buffer
3860 * <out>. "-" is printed if the value is zero, "inf" if larger than 1000 years.
3861 * It returns the new buffer length, or 0 if it doesn't fit. The value will be
3862 * surrounded by <pfx> and <sfx> respectively if not NULL.
3863 */
3864int print_time_short(struct buffer *out, const char *pfx, uint64_t ns, const char *sfx)
3865{
3866 double val = ns; // 52 bits of mantissa keep ns accuracy over 52 days
3867 const char *unit;
3868
3869 if (!pfx)
3870 pfx = "";
3871 if (!sfx)
3872 sfx = "";
3873
3874 do {
3875 unit = " - "; if (val <= 0.0) break;
3876 unit = "ns"; if (val < 1000.0) break;
3877 unit = "us"; val /= 1000.0; if (val < 1000.0) break;
3878 unit = "ms"; val /= 1000.0; if (val < 1000.0) break;
3879 unit = "s "; val /= 1000.0; if (val < 60.0) break;
3880 unit = "m "; val /= 60.0; if (val < 60.0) break;
3881 unit = "h "; val /= 60.0; if (val < 24.0) break;
3882 unit = "d "; val /= 24.0; if (val < 365.0) break;
3883 unit = "yr"; val /= 365.0; if (val < 1000.0) break;
3884 unit = " inf "; val = 0.0; break;
3885 } while (0);
3886
3887 if (val <= 0.0)
3888 return chunk_appendf(out, "%s%7s%s", pfx, unit, sfx);
3889 else if (val < 10.0)
3890 return chunk_appendf(out, "%s%1.3f%s%s", pfx, val, unit, sfx);
3891 else if (val < 100.0)
3892 return chunk_appendf(out, "%s%2.2f%s%s", pfx, val, unit, sfx);
3893 else
3894 return chunk_appendf(out, "%s%3.1f%s%s", pfx, val, unit, sfx);
3895}
3896
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003897/* Dynamically allocates a string of the proper length to hold the formatted
3898 * output. NULL is returned on error. The caller is responsible for freeing the
3899 * memory area using free(). The resulting string is returned in <out> if the
3900 * pointer is not NULL. A previous version of <out> might be used to build the
3901 * new string, and it will be freed before returning if it is not NULL, which
3902 * makes it possible to build complex strings from iterative calls without
3903 * having to care about freeing intermediate values, as in the example below :
3904 *
3905 * memprintf(&err, "invalid argument: '%s'", arg);
3906 * ...
3907 * memprintf(&err, "parser said : <%s>\n", *err);
3908 * ...
3909 * free(*err);
3910 *
3911 * This means that <err> must be initialized to NULL before first invocation.
3912 * The return value also holds the allocated string, which eases error checking
3913 * and immediate consumption. If the output pointer is not used, NULL must be
Willy Tarreaueb6cead2012-09-20 19:43:14 +02003914 * passed instead and it will be ignored. The returned message will then also
3915 * be NULL so that the caller does not have to bother with freeing anything.
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003916 *
3917 * It is also convenient to use it without any free except the last one :
3918 * err = NULL;
3919 * if (!fct1(err)) report(*err);
3920 * if (!fct2(err)) report(*err);
3921 * if (!fct3(err)) report(*err);
3922 * free(*err);
Christopher Faulet93a518f2017-10-24 11:25:33 +02003923 *
3924 * memprintf relies on memvprintf. This last version can be called from any
3925 * function with variadic arguments.
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003926 */
Christopher Faulet93a518f2017-10-24 11:25:33 +02003927char *memvprintf(char **out, const char *format, va_list orig_args)
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003928{
3929 va_list args;
3930 char *ret = NULL;
3931 int allocated = 0;
3932 int needed = 0;
3933
Willy Tarreaueb6cead2012-09-20 19:43:14 +02003934 if (!out)
3935 return NULL;
3936
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003937 do {
Willy Tarreaue0609f52019-03-29 19:13:23 +01003938 char buf1;
3939
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003940 /* vsnprintf() will return the required length even when the
3941 * target buffer is NULL. We do this in a loop just in case
3942 * intermediate evaluations get wrong.
3943 */
Christopher Faulet93a518f2017-10-24 11:25:33 +02003944 va_copy(args, orig_args);
Willy Tarreaue0609f52019-03-29 19:13:23 +01003945 needed = vsnprintf(ret ? ret : &buf1, allocated, format, args);
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003946 va_end(args);
Willy Tarreau1b2fed62013-04-01 22:48:54 +02003947 if (needed < allocated) {
3948 /* Note: on Solaris 8, the first iteration always
3949 * returns -1 if allocated is zero, so we force a
3950 * retry.
3951 */
3952 if (!allocated)
3953 needed = 0;
3954 else
3955 break;
3956 }
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003957
Willy Tarreau1b2fed62013-04-01 22:48:54 +02003958 allocated = needed + 1;
Hubert Verstraete831962e2016-06-28 22:44:26 +02003959 ret = my_realloc2(ret, allocated);
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003960 } while (ret);
3961
3962 if (needed < 0) {
3963 /* an error was encountered */
Willy Tarreau61cfdf42021-02-20 10:46:51 +01003964 ha_free(&ret);
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003965 }
3966
3967 if (out) {
3968 free(*out);
3969 *out = ret;
3970 }
3971
3972 return ret;
3973}
William Lallemand421f5b52012-02-06 18:15:57 +01003974
Christopher Faulet93a518f2017-10-24 11:25:33 +02003975char *memprintf(char **out, const char *format, ...)
3976{
3977 va_list args;
3978 char *ret = NULL;
3979
3980 va_start(args, format);
3981 ret = memvprintf(out, format, args);
3982 va_end(args);
3983
3984 return ret;
3985}
3986
Willy Tarreau21c705b2012-09-14 11:40:36 +02003987/* Used to add <level> spaces before each line of <out>, unless there is only one line.
3988 * The input argument is automatically freed and reassigned. The result will have to be
Willy Tarreau70eec382012-10-10 08:56:47 +02003989 * freed by the caller. It also supports being passed a NULL which results in the same
3990 * output.
Willy Tarreau21c705b2012-09-14 11:40:36 +02003991 * Example of use :
3992 * parse(cmd, &err); (callee: memprintf(&err, ...))
3993 * fprintf(stderr, "Parser said: %s\n", indent_error(&err));
3994 * free(err);
3995 */
3996char *indent_msg(char **out, int level)
3997{
3998 char *ret, *in, *p;
3999 int needed = 0;
4000 int lf = 0;
4001 int lastlf = 0;
4002 int len;
4003
Willy Tarreau70eec382012-10-10 08:56:47 +02004004 if (!out || !*out)
4005 return NULL;
4006
Willy Tarreau21c705b2012-09-14 11:40:36 +02004007 in = *out - 1;
4008 while ((in = strchr(in + 1, '\n')) != NULL) {
4009 lastlf = in - *out;
4010 lf++;
4011 }
4012
4013 if (!lf) /* single line, no LF, return it as-is */
4014 return *out;
4015
4016 len = strlen(*out);
4017
4018 if (lf == 1 && lastlf == len - 1) {
4019 /* single line, LF at end, strip it and return as-is */
4020 (*out)[lastlf] = 0;
4021 return *out;
4022 }
4023
4024 /* OK now we have at least one LF, we need to process the whole string
4025 * as a multi-line string. What we'll do :
4026 * - prefix with an LF if there is none
4027 * - add <level> spaces before each line
4028 * This means at most ( 1 + level + (len-lf) + lf*<1+level) ) =
4029 * 1 + level + len + lf * level = 1 + level * (lf + 1) + len.
4030 */
4031
4032 needed = 1 + level * (lf + 1) + len + 1;
4033 p = ret = malloc(needed);
4034 in = *out;
4035
4036 /* skip initial LFs */
4037 while (*in == '\n')
4038 in++;
4039
4040 /* copy each line, prefixed with LF and <level> spaces, and without the trailing LF */
4041 while (*in) {
4042 *p++ = '\n';
4043 memset(p, ' ', level);
4044 p += level;
4045 do {
4046 *p++ = *in++;
4047 } while (*in && *in != '\n');
4048 if (*in)
4049 in++;
4050 }
4051 *p = 0;
4052
4053 free(*out);
4054 *out = ret;
4055
4056 return ret;
4057}
4058
Willy Tarreaua2c99112019-08-21 13:17:37 +02004059/* makes a copy of message <in> into <out>, with each line prefixed with <pfx>
4060 * and end of lines replaced with <eol> if not 0. The first line to indent has
4061 * to be indicated in <first> (starts at zero), so that it is possible to skip
4062 * indenting the first line if it has to be appended after an existing message.
4063 * Empty strings are never indented, and NULL strings are considered empty both
4064 * for <in> and <pfx>. It returns non-zero if an EOL was appended as the last
4065 * character, non-zero otherwise.
4066 */
4067int append_prefixed_str(struct buffer *out, const char *in, const char *pfx, char eol, int first)
4068{
4069 int bol, lf;
4070 int pfxlen = pfx ? strlen(pfx) : 0;
4071
4072 if (!in)
4073 return 0;
4074
4075 bol = 1;
4076 lf = 0;
4077 while (*in) {
4078 if (bol && pfxlen) {
4079 if (first > 0)
4080 first--;
4081 else
4082 b_putblk(out, pfx, pfxlen);
4083 bol = 0;
4084 }
4085
4086 lf = (*in == '\n');
4087 bol |= lf;
4088 b_putchr(out, (lf && eol) ? eol : *in);
4089 in++;
4090 }
4091 return lf;
4092}
4093
Willy Tarreau9d22e562019-03-29 18:49:09 +01004094/* removes environment variable <name> from the environment as found in
4095 * environ. This is only provided as an alternative for systems without
4096 * unsetenv() (old Solaris and AIX versions). THIS IS NOT THREAD SAFE.
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05004097 * The principle is to scan environ for each occurrence of variable name
Willy Tarreau9d22e562019-03-29 18:49:09 +01004098 * <name> and to replace the matching pointers with the last pointer of
4099 * the array (since variables are not ordered).
4100 * It always returns 0 (success).
4101 */
4102int my_unsetenv(const char *name)
4103{
4104 extern char **environ;
4105 char **p = environ;
4106 int vars;
4107 int next;
4108 int len;
4109
4110 len = strlen(name);
4111 for (vars = 0; p[vars]; vars++)
4112 ;
4113 next = 0;
4114 while (next < vars) {
4115 if (strncmp(p[next], name, len) != 0 || p[next][len] != '=') {
4116 next++;
4117 continue;
4118 }
4119 if (next < vars - 1)
4120 p[next] = p[vars - 1];
4121 p[--vars] = NULL;
4122 }
4123 return 0;
4124}
4125
Willy Tarreaudad36a32013-03-11 01:20:04 +01004126/* Convert occurrences of environment variables in the input string to their
4127 * corresponding value. A variable is identified as a series of alphanumeric
4128 * characters or underscores following a '$' sign. The <in> string must be
4129 * free()able. NULL returns NULL. The resulting string might be reallocated if
4130 * some expansion is made. Variable names may also be enclosed into braces if
4131 * needed (eg: to concatenate alphanum characters).
4132 */
4133char *env_expand(char *in)
4134{
4135 char *txt_beg;
4136 char *out;
4137 char *txt_end;
4138 char *var_beg;
4139 char *var_end;
4140 char *value;
4141 char *next;
4142 int out_len;
4143 int val_len;
4144
4145 if (!in)
4146 return in;
4147
4148 value = out = NULL;
4149 out_len = 0;
4150
4151 txt_beg = in;
4152 do {
4153 /* look for next '$' sign in <in> */
4154 for (txt_end = txt_beg; *txt_end && *txt_end != '$'; txt_end++);
4155
4156 if (!*txt_end && !out) /* end and no expansion performed */
4157 return in;
4158
4159 val_len = 0;
4160 next = txt_end;
4161 if (*txt_end == '$') {
4162 char save;
4163
4164 var_beg = txt_end + 1;
4165 if (*var_beg == '{')
4166 var_beg++;
4167
4168 var_end = var_beg;
Willy Tarreau90807112020-02-25 08:16:33 +01004169 while (isalnum((unsigned char)*var_end) || *var_end == '_') {
Willy Tarreaudad36a32013-03-11 01:20:04 +01004170 var_end++;
4171 }
4172
4173 next = var_end;
4174 if (*var_end == '}' && (var_beg > txt_end + 1))
4175 next++;
4176
4177 /* get value of the variable name at this location */
4178 save = *var_end;
4179 *var_end = '\0';
4180 value = getenv(var_beg);
4181 *var_end = save;
4182 val_len = value ? strlen(value) : 0;
4183 }
4184
Hubert Verstraete831962e2016-06-28 22:44:26 +02004185 out = my_realloc2(out, out_len + (txt_end - txt_beg) + val_len + 1);
Willy Tarreaudad36a32013-03-11 01:20:04 +01004186 if (txt_end > txt_beg) {
4187 memcpy(out + out_len, txt_beg, txt_end - txt_beg);
4188 out_len += txt_end - txt_beg;
4189 }
4190 if (val_len) {
4191 memcpy(out + out_len, value, val_len);
4192 out_len += val_len;
4193 }
4194 out[out_len] = 0;
4195 txt_beg = next;
4196 } while (*txt_beg);
4197
4198 /* here we know that <out> was allocated and that we don't need <in> anymore */
4199 free(in);
4200 return out;
4201}
4202
de Lafond Guillaume88c278f2013-04-15 19:27:10 +02004203
4204/* same as strstr() but case-insensitive and with limit length */
4205const char *strnistr(const char *str1, int len_str1, const char *str2, int len_str2)
4206{
4207 char *pptr, *sptr, *start;
Willy Tarreauc8746532014-05-28 23:05:07 +02004208 unsigned int slen, plen;
4209 unsigned int tmp1, tmp2;
de Lafond Guillaume88c278f2013-04-15 19:27:10 +02004210
4211 if (str1 == NULL || len_str1 == 0) // search pattern into an empty string => search is not found
4212 return NULL;
4213
4214 if (str2 == NULL || len_str2 == 0) // pattern is empty => every str1 match
4215 return str1;
4216
4217 if (len_str1 < len_str2) // pattern is longer than string => search is not found
4218 return NULL;
4219
4220 for (tmp1 = 0, start = (char *)str1, pptr = (char *)str2, slen = len_str1, plen = len_str2; slen >= plen; start++, slen--) {
Willy Tarreauf278eec2020-07-05 21:46:32 +02004221 while (toupper((unsigned char)*start) != toupper((unsigned char)*str2)) {
de Lafond Guillaume88c278f2013-04-15 19:27:10 +02004222 start++;
4223 slen--;
4224 tmp1++;
4225
4226 if (tmp1 >= len_str1)
4227 return NULL;
4228
4229 /* if pattern longer than string */
4230 if (slen < plen)
4231 return NULL;
4232 }
4233
4234 sptr = start;
4235 pptr = (char *)str2;
4236
4237 tmp2 = 0;
Willy Tarreauf278eec2020-07-05 21:46:32 +02004238 while (toupper((unsigned char)*sptr) == toupper((unsigned char)*pptr)) {
de Lafond Guillaume88c278f2013-04-15 19:27:10 +02004239 sptr++;
4240 pptr++;
4241 tmp2++;
4242
4243 if (*pptr == '\0' || tmp2 == len_str2) /* end of pattern found */
4244 return start;
4245 if (*sptr == '\0' || tmp2 == len_str1) /* end of string found and the pattern is not fully found */
4246 return NULL;
4247 }
4248 }
4249 return NULL;
4250}
4251
Thierry FOURNIER317e1c42014-08-12 10:20:47 +02004252/* This function read the next valid utf8 char.
4253 * <s> is the byte srray to be decode, <len> is its length.
4254 * The function returns decoded char encoded like this:
4255 * The 4 msb are the return code (UTF8_CODE_*), the 4 lsb
4256 * are the length read. The decoded character is stored in <c>.
4257 */
4258unsigned char utf8_next(const char *s, int len, unsigned int *c)
4259{
4260 const unsigned char *p = (unsigned char *)s;
4261 int dec;
4262 unsigned char code = UTF8_CODE_OK;
4263
4264 if (len < 1)
4265 return UTF8_CODE_OK;
4266
4267 /* Check the type of UTF8 sequence
4268 *
4269 * 0... .... 0x00 <= x <= 0x7f : 1 byte: ascii char
4270 * 10.. .... 0x80 <= x <= 0xbf : invalid sequence
4271 * 110. .... 0xc0 <= x <= 0xdf : 2 bytes
4272 * 1110 .... 0xe0 <= x <= 0xef : 3 bytes
4273 * 1111 0... 0xf0 <= x <= 0xf7 : 4 bytes
4274 * 1111 10.. 0xf8 <= x <= 0xfb : 5 bytes
4275 * 1111 110. 0xfc <= x <= 0xfd : 6 bytes
4276 * 1111 111. 0xfe <= x <= 0xff : invalid sequence
4277 */
4278 switch (*p) {
4279 case 0x00 ... 0x7f:
4280 *c = *p;
4281 return UTF8_CODE_OK | 1;
4282
4283 case 0x80 ... 0xbf:
4284 *c = *p;
4285 return UTF8_CODE_BADSEQ | 1;
4286
4287 case 0xc0 ... 0xdf:
4288 if (len < 2) {
4289 *c = *p;
4290 return UTF8_CODE_BADSEQ | 1;
4291 }
4292 *c = *p & 0x1f;
4293 dec = 1;
4294 break;
4295
4296 case 0xe0 ... 0xef:
4297 if (len < 3) {
4298 *c = *p;
4299 return UTF8_CODE_BADSEQ | 1;
4300 }
4301 *c = *p & 0x0f;
4302 dec = 2;
4303 break;
4304
4305 case 0xf0 ... 0xf7:
4306 if (len < 4) {
4307 *c = *p;
4308 return UTF8_CODE_BADSEQ | 1;
4309 }
4310 *c = *p & 0x07;
4311 dec = 3;
4312 break;
4313
4314 case 0xf8 ... 0xfb:
4315 if (len < 5) {
4316 *c = *p;
4317 return UTF8_CODE_BADSEQ | 1;
4318 }
4319 *c = *p & 0x03;
4320 dec = 4;
4321 break;
4322
4323 case 0xfc ... 0xfd:
4324 if (len < 6) {
4325 *c = *p;
4326 return UTF8_CODE_BADSEQ | 1;
4327 }
4328 *c = *p & 0x01;
4329 dec = 5;
4330 break;
4331
4332 case 0xfe ... 0xff:
4333 default:
4334 *c = *p;
4335 return UTF8_CODE_BADSEQ | 1;
4336 }
4337
4338 p++;
4339
4340 while (dec > 0) {
4341
4342 /* need 0x10 for the 2 first bits */
4343 if ( ( *p & 0xc0 ) != 0x80 )
4344 return UTF8_CODE_BADSEQ | ((p-(unsigned char *)s)&0xffff);
4345
4346 /* add data at char */
4347 *c = ( *c << 6 ) | ( *p & 0x3f );
4348
4349 dec--;
4350 p++;
4351 }
4352
4353 /* Check ovelong encoding.
4354 * 1 byte : 5 + 6 : 11 : 0x80 ... 0x7ff
4355 * 2 bytes : 4 + 6 + 6 : 16 : 0x800 ... 0xffff
4356 * 3 bytes : 3 + 6 + 6 + 6 : 21 : 0x10000 ... 0x1fffff
4357 */
Thierry FOURNIER9e7ec082015-03-12 19:32:38 +01004358 if (( *c <= 0x7f && (p-(unsigned char *)s) > 1) ||
Thierry FOURNIER317e1c42014-08-12 10:20:47 +02004359 (*c >= 0x80 && *c <= 0x7ff && (p-(unsigned char *)s) > 2) ||
4360 (*c >= 0x800 && *c <= 0xffff && (p-(unsigned char *)s) > 3) ||
4361 (*c >= 0x10000 && *c <= 0x1fffff && (p-(unsigned char *)s) > 4))
4362 code |= UTF8_CODE_OVERLONG;
4363
4364 /* Check invalid UTF8 range. */
4365 if ((*c >= 0xd800 && *c <= 0xdfff) ||
4366 (*c >= 0xfffe && *c <= 0xffff))
4367 code |= UTF8_CODE_INVRANGE;
4368
4369 return code | ((p-(unsigned char *)s)&0x0f);
4370}
4371
Maxime de Roucydc887852016-05-13 23:52:54 +02004372/* append a copy of string <str> (in a wordlist) at the end of the list <li>
4373 * On failure : return 0 and <err> filled with an error message.
4374 * The caller is responsible for freeing the <err> and <str> copy
4375 * memory area using free()
4376 */
4377int list_append_word(struct list *li, const char *str, char **err)
4378{
4379 struct wordlist *wl;
4380
4381 wl = calloc(1, sizeof(*wl));
4382 if (!wl) {
4383 memprintf(err, "out of memory");
4384 goto fail_wl;
4385 }
4386
4387 wl->s = strdup(str);
4388 if (!wl->s) {
4389 memprintf(err, "out of memory");
4390 goto fail_wl_s;
4391 }
4392
4393 LIST_ADDQ(li, &wl->list);
4394
4395 return 1;
4396
4397fail_wl_s:
4398 free(wl->s);
4399fail_wl:
4400 free(wl);
4401 return 0;
4402}
4403
Willy Tarreau37101052019-05-20 16:48:20 +02004404/* indicates if a memory location may safely be read or not. The trick consists
4405 * in performing a harmless syscall using this location as an input and letting
4406 * the operating system report whether it's OK or not. For this we have the
4407 * stat() syscall, which will return EFAULT when the memory location supposed
4408 * to contain the file name is not readable. If it is readable it will then
4409 * either return 0 if the area contains an existing file name, or -1 with
4410 * another code. This must not be abused, and some audit systems might detect
4411 * this as abnormal activity. It's used only for unsafe dumps.
4412 */
4413int may_access(const void *ptr)
4414{
4415 struct stat buf;
4416
4417 if (stat(ptr, &buf) == 0)
4418 return 1;
4419 if (errno == EFAULT)
4420 return 0;
4421 return 1;
4422}
4423
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004424/* print a string of text buffer to <out>. The format is :
4425 * Non-printable chars \t, \n, \r and \e are * encoded in C format.
4426 * Other non-printable chars are encoded "\xHH". Space, '\', and '=' are also escaped.
4427 * Print stopped if null char or <bsize> is reached, or if no more place in the chunk.
4428 */
Willy Tarreau83061a82018-07-13 11:56:34 +02004429int dump_text(struct buffer *out, const char *buf, int bsize)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004430{
4431 unsigned char c;
4432 int ptr = 0;
4433
4434 while (buf[ptr] && ptr < bsize) {
4435 c = buf[ptr];
Willy Tarreau90807112020-02-25 08:16:33 +01004436 if (isprint((unsigned char)c) && isascii((unsigned char)c) && c != '\\' && c != ' ' && c != '=') {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004437 if (out->data > out->size - 1)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004438 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004439 out->area[out->data++] = c;
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004440 }
4441 else if (c == '\t' || c == '\n' || c == '\r' || c == '\e' || c == '\\' || c == ' ' || c == '=') {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004442 if (out->data > out->size - 2)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004443 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004444 out->area[out->data++] = '\\';
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004445 switch (c) {
4446 case ' ': c = ' '; break;
4447 case '\t': c = 't'; break;
4448 case '\n': c = 'n'; break;
4449 case '\r': c = 'r'; break;
4450 case '\e': c = 'e'; break;
4451 case '\\': c = '\\'; break;
4452 case '=': c = '='; break;
4453 }
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004454 out->area[out->data++] = c;
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004455 }
4456 else {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004457 if (out->data > out->size - 4)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004458 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004459 out->area[out->data++] = '\\';
4460 out->area[out->data++] = 'x';
4461 out->area[out->data++] = hextab[(c >> 4) & 0xF];
4462 out->area[out->data++] = hextab[c & 0xF];
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004463 }
4464 ptr++;
4465 }
4466
4467 return ptr;
4468}
4469
4470/* print a buffer in hexa.
4471 * Print stopped if <bsize> is reached, or if no more place in the chunk.
4472 */
Willy Tarreau83061a82018-07-13 11:56:34 +02004473int dump_binary(struct buffer *out, const char *buf, int bsize)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004474{
4475 unsigned char c;
4476 int ptr = 0;
4477
4478 while (ptr < bsize) {
4479 c = buf[ptr];
4480
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004481 if (out->data > out->size - 2)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004482 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004483 out->area[out->data++] = hextab[(c >> 4) & 0xF];
4484 out->area[out->data++] = hextab[c & 0xF];
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004485
4486 ptr++;
4487 }
4488 return ptr;
4489}
4490
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004491/* Appends into buffer <out> a hex dump of memory area <buf> for <len> bytes,
4492 * prepending each line with prefix <pfx>. The output is *not* initialized.
4493 * The output will not wrap pas the buffer's end so it is more optimal if the
4494 * caller makes sure the buffer is aligned first. A trailing zero will always
4495 * be appended (and not counted) if there is room for it. The caller must make
Willy Tarreau37101052019-05-20 16:48:20 +02004496 * sure that the area is dumpable first. If <unsafe> is non-null, the memory
4497 * locations are checked first for being readable.
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004498 */
Willy Tarreau37101052019-05-20 16:48:20 +02004499void dump_hex(struct buffer *out, const char *pfx, const void *buf, int len, int unsafe)
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004500{
4501 const unsigned char *d = buf;
4502 int i, j, start;
4503
4504 d = (const unsigned char *)(((unsigned long)buf) & -16);
4505 start = ((unsigned long)buf) & 15;
4506
4507 for (i = 0; i < start + len; i += 16) {
4508 chunk_appendf(out, (sizeof(void *) == 4) ? "%s%8p: " : "%s%16p: ", pfx, d + i);
4509
Willy Tarreau37101052019-05-20 16:48:20 +02004510 // 0: unchecked, 1: checked safe, 2: danger
4511 unsafe = !!unsafe;
4512 if (unsafe && !may_access(d + i))
4513 unsafe = 2;
4514
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004515 for (j = 0; j < 16; j++) {
Willy Tarreau37101052019-05-20 16:48:20 +02004516 if ((i + j < start) || (i + j >= start + len))
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004517 chunk_strcat(out, "'' ");
Willy Tarreau37101052019-05-20 16:48:20 +02004518 else if (unsafe > 1)
4519 chunk_strcat(out, "** ");
4520 else
4521 chunk_appendf(out, "%02x ", d[i + j]);
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004522
4523 if (j == 7)
4524 chunk_strcat(out, "- ");
4525 }
4526 chunk_strcat(out, " ");
4527 for (j = 0; j < 16; j++) {
Willy Tarreau37101052019-05-20 16:48:20 +02004528 if ((i + j < start) || (i + j >= start + len))
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004529 chunk_strcat(out, "'");
Willy Tarreau37101052019-05-20 16:48:20 +02004530 else if (unsafe > 1)
4531 chunk_strcat(out, "*");
Willy Tarreau90807112020-02-25 08:16:33 +01004532 else if (isprint((unsigned char)d[i + j]))
Willy Tarreau37101052019-05-20 16:48:20 +02004533 chunk_appendf(out, "%c", d[i + j]);
4534 else
4535 chunk_strcat(out, ".");
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004536 }
4537 chunk_strcat(out, "\n");
4538 }
4539}
4540
Willy Tarreau762fb3e2020-03-03 15:57:10 +01004541/* dumps <pfx> followed by <n> bytes from <addr> in hex form into buffer <buf>
4542 * enclosed in brackets after the address itself, formatted on 14 chars
4543 * including the "0x" prefix. This is meant to be used as a prefix for code
4544 * areas. For example:
4545 * "0x7f10b6557690 [48 c7 c0 0f 00 00 00 0f]"
4546 * It relies on may_access() to know if the bytes are dumpable, otherwise "--"
4547 * is emitted. A NULL <pfx> will be considered empty.
4548 */
4549void dump_addr_and_bytes(struct buffer *buf, const char *pfx, const void *addr, int n)
4550{
4551 int ok = 0;
4552 int i;
4553
4554 chunk_appendf(buf, "%s%#14lx [", pfx ? pfx : "", (long)addr);
4555
4556 for (i = 0; i < n; i++) {
4557 if (i == 0 || (((long)(addr + i) ^ (long)(addr)) & 4096))
4558 ok = may_access(addr + i);
4559 if (ok)
4560 chunk_appendf(buf, "%02x%s", ((uint8_t*)addr)[i], (i<n-1) ? " " : "]");
4561 else
4562 chunk_appendf(buf, "--%s", (i<n-1) ? " " : "]");
4563 }
4564}
4565
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004566/* print a line of text buffer (limited to 70 bytes) to <out>. The format is :
4567 * <2 spaces> <offset=5 digits> <space or plus> <space> <70 chars max> <\n>
4568 * which is 60 chars per line. Non-printable chars \t, \n, \r and \e are
4569 * encoded in C format. Other non-printable chars are encoded "\xHH". Original
4570 * lines are respected within the limit of 70 output chars. Lines that are
4571 * continuation of a previous truncated line begin with "+" instead of " "
4572 * after the offset. The new pointer is returned.
4573 */
Willy Tarreau83061a82018-07-13 11:56:34 +02004574int dump_text_line(struct buffer *out, const char *buf, int bsize, int len,
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004575 int *line, int ptr)
4576{
4577 int end;
4578 unsigned char c;
4579
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004580 end = out->data + 80;
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004581 if (end > out->size)
4582 return ptr;
4583
4584 chunk_appendf(out, " %05d%c ", ptr, (ptr == *line) ? ' ' : '+');
4585
4586 while (ptr < len && ptr < bsize) {
4587 c = buf[ptr];
Willy Tarreau90807112020-02-25 08:16:33 +01004588 if (isprint((unsigned char)c) && isascii((unsigned char)c) && c != '\\') {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004589 if (out->data > end - 2)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004590 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004591 out->area[out->data++] = c;
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004592 } else if (c == '\t' || c == '\n' || c == '\r' || c == '\e' || c == '\\') {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004593 if (out->data > end - 3)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004594 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004595 out->area[out->data++] = '\\';
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004596 switch (c) {
4597 case '\t': c = 't'; break;
4598 case '\n': c = 'n'; break;
4599 case '\r': c = 'r'; break;
4600 case '\e': c = 'e'; break;
4601 case '\\': c = '\\'; break;
4602 }
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004603 out->area[out->data++] = c;
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004604 } else {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004605 if (out->data > end - 5)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004606 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004607 out->area[out->data++] = '\\';
4608 out->area[out->data++] = 'x';
4609 out->area[out->data++] = hextab[(c >> 4) & 0xF];
4610 out->area[out->data++] = hextab[c & 0xF];
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004611 }
4612 if (buf[ptr++] == '\n') {
4613 /* we had a line break, let's return now */
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004614 out->area[out->data++] = '\n';
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004615 *line = ptr;
4616 return ptr;
4617 }
4618 }
4619 /* we have an incomplete line, we return it as-is */
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004620 out->area[out->data++] = '\n';
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004621 return ptr;
4622}
4623
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004624/* displays a <len> long memory block at <buf>, assuming first byte of <buf>
Willy Tarreaued936c52017-04-27 18:03:20 +02004625 * has address <baseaddr>. String <pfx> may be placed as a prefix in front of
4626 * each line. It may be NULL if unused. The output is emitted to file <out>.
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004627 */
Willy Tarreaued936c52017-04-27 18:03:20 +02004628void debug_hexdump(FILE *out, const char *pfx, const char *buf,
4629 unsigned int baseaddr, int len)
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004630{
Willy Tarreau73459792017-04-11 07:58:08 +02004631 unsigned int i;
4632 int b, j;
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004633
4634 for (i = 0; i < (len + (baseaddr & 15)); i += 16) {
4635 b = i - (baseaddr & 15);
Willy Tarreaued936c52017-04-27 18:03:20 +02004636 fprintf(out, "%s%08x: ", pfx ? pfx : "", i + (baseaddr & ~15));
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004637 for (j = 0; j < 8; j++) {
4638 if (b + j >= 0 && b + j < len)
4639 fprintf(out, "%02x ", (unsigned char)buf[b + j]);
4640 else
4641 fprintf(out, " ");
4642 }
4643
4644 if (b + j >= 0 && b + j < len)
4645 fputc('-', out);
4646 else
4647 fputc(' ', out);
4648
4649 for (j = 8; j < 16; j++) {
4650 if (b + j >= 0 && b + j < len)
4651 fprintf(out, " %02x", (unsigned char)buf[b + j]);
4652 else
4653 fprintf(out, " ");
4654 }
4655
4656 fprintf(out, " ");
4657 for (j = 0; j < 16; j++) {
4658 if (b + j >= 0 && b + j < len) {
4659 if (isprint((unsigned char)buf[b + j]))
4660 fputc((unsigned char)buf[b + j], out);
4661 else
4662 fputc('.', out);
4663 }
4664 else
4665 fputc(' ', out);
4666 }
4667 fputc('\n', out);
4668 }
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004669}
4670
Willy Tarreaubb869862020-04-16 10:52:41 +02004671/* Tries to report the executable path name on platforms supporting this. If
4672 * not found or not possible, returns NULL.
4673 */
4674const char *get_exec_path()
4675{
4676 const char *ret = NULL;
4677
4678#if (__GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 16))
4679 long execfn = getauxval(AT_EXECFN);
4680
4681 if (execfn && execfn != ENOENT)
4682 ret = (const char *)execfn;
4683#endif
4684 return ret;
4685}
4686
Baruch Siache1651b22020-07-24 07:52:20 +03004687#if (defined(__ELF__) && !defined(__linux__)) || defined(USE_DL)
Willy Tarreau9133e482020-03-04 10:19:36 +01004688/* calls dladdr() or dladdr1() on <addr> and <dli>. If dladdr1 is available,
4689 * also returns the symbol size in <size>, otherwise returns 0 there.
4690 */
4691static int dladdr_and_size(const void *addr, Dl_info *dli, size_t *size)
4692{
4693 int ret;
Willy Tarreau62af9c82020-03-10 07:51:48 +01004694#if (__GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 3)) // most detailed one
Willy Tarreau9133e482020-03-04 10:19:36 +01004695 const ElfW(Sym) *sym;
4696
4697 ret = dladdr1(addr, dli, (void **)&sym, RTLD_DL_SYMENT);
4698 if (ret)
4699 *size = sym ? sym->st_size : 0;
4700#else
4701 ret = dladdr(addr, dli);
4702 *size = 0;
4703#endif
4704 return ret;
4705}
4706#endif
4707
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004708/* Tries to append to buffer <buf> some indications about the symbol at address
4709 * <addr> using the following form:
4710 * lib:+0xoffset (unresolvable address from lib's base)
4711 * main+0xoffset (unresolvable address from main (+/-))
4712 * lib:main+0xoffset (unresolvable lib address from main (+/-))
4713 * name (resolved exact exec address)
4714 * lib:name (resolved exact lib address)
4715 * name+0xoffset/0xsize (resolved address within exec symbol)
4716 * lib:name+0xoffset/0xsize (resolved address within lib symbol)
4717 *
4718 * The file name (lib or executable) is limited to what lies between the last
4719 * '/' and the first following '.'. An optional prefix <pfx> is prepended before
4720 * the output if not null. The file is not dumped when it's the same as the one
Baruch Siache1651b22020-07-24 07:52:20 +03004721 * that contains the "main" symbol, or when __ELF__ && USE_DL are not set.
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004722 *
4723 * The symbol's base address is returned, or NULL when unresolved, in order to
4724 * allow the caller to match it against known ones.
4725 */
Willy Tarreau45fd1032021-01-20 14:37:59 +01004726const void *resolve_sym_name(struct buffer *buf, const char *pfx, const void *addr)
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004727{
4728 const struct {
4729 const void *func;
4730 const char *name;
4731 } fcts[] = {
4732 { .func = process_stream, .name = "process_stream" },
4733 { .func = task_run_applet, .name = "task_run_applet" },
4734 { .func = si_cs_io_cb, .name = "si_cs_io_cb" },
Willy Tarreau586f71b2020-12-11 15:54:36 +01004735 { .func = sock_conn_iocb, .name = "sock_conn_iocb" },
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004736 { .func = dgram_fd_handler, .name = "dgram_fd_handler" },
4737 { .func = listener_accept, .name = "listener_accept" },
Willy Tarreaud597ec22021-01-29 14:29:06 +01004738 { .func = manage_global_listener_queue, .name = "manage_global_listener_queue" },
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004739 { .func = poller_pipe_io_handler, .name = "poller_pipe_io_handler" },
4740 { .func = mworker_accept_wrapper, .name = "mworker_accept_wrapper" },
Willy Tarreau02922e12021-01-29 12:27:57 +01004741 { .func = session_expire_embryonic, .name = "session_expire_embryonic" },
Willy Tarreaufb5401f2021-01-29 12:25:23 +01004742#ifdef USE_THREAD
4743 { .func = accept_queue_process, .name = "accept_queue_process" },
4744#endif
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004745#ifdef USE_LUA
4746 { .func = hlua_process_task, .name = "hlua_process_task" },
4747#endif
Ilya Shipitsinbdec3ba2020-11-14 01:56:34 +05004748#ifdef SSL_MODE_ASYNC
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004749 { .func = ssl_async_fd_free, .name = "ssl_async_fd_free" },
4750 { .func = ssl_async_fd_handler, .name = "ssl_async_fd_handler" },
4751#endif
4752 };
4753
Baruch Siache1651b22020-07-24 07:52:20 +03004754#if (defined(__ELF__) && !defined(__linux__)) || defined(USE_DL)
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004755 Dl_info dli, dli_main;
Willy Tarreau9133e482020-03-04 10:19:36 +01004756 size_t size;
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004757 const char *fname, *p;
4758#endif
4759 int i;
4760
4761 if (pfx)
4762 chunk_appendf(buf, "%s", pfx);
4763
4764 for (i = 0; i < sizeof(fcts) / sizeof(fcts[0]); i++) {
4765 if (addr == fcts[i].func) {
4766 chunk_appendf(buf, "%s", fcts[i].name);
4767 return addr;
4768 }
4769 }
4770
Baruch Siache1651b22020-07-24 07:52:20 +03004771#if (defined(__ELF__) && !defined(__linux__)) || defined(USE_DL)
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004772 /* Now let's try to be smarter */
Willy Tarreau9133e482020-03-04 10:19:36 +01004773 if (!dladdr_and_size(addr, &dli, &size))
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004774 goto unknown;
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004775
4776 /* 1. prefix the library name if it's not the same object as the one
4777 * that contains the main function. The name is picked between last '/'
4778 * and first following '.'.
4779 */
4780 if (!dladdr(main, &dli_main))
4781 dli_main.dli_fbase = NULL;
4782
4783 if (dli_main.dli_fbase != dli.dli_fbase) {
4784 fname = dli.dli_fname;
4785 p = strrchr(fname, '/');
4786 if (p++)
4787 fname = p;
4788 p = strchr(fname, '.');
4789 if (!p)
4790 p = fname + strlen(fname);
4791
4792 chunk_appendf(buf, "%.*s:", (int)(long)(p - fname), fname);
4793 }
4794
4795 /* 2. symbol name */
4796 if (dli.dli_sname) {
4797 /* known, dump it and return symbol's address (exact or relative) */
4798 chunk_appendf(buf, "%s", dli.dli_sname);
4799 if (addr != dli.dli_saddr) {
4800 chunk_appendf(buf, "+%#lx", (long)(addr - dli.dli_saddr));
Willy Tarreau9133e482020-03-04 10:19:36 +01004801 if (size)
4802 chunk_appendf(buf, "/%#lx", (long)size);
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004803 }
4804 return dli.dli_saddr;
4805 }
4806 else if (dli_main.dli_fbase != dli.dli_fbase) {
4807 /* unresolved symbol from a known library, report relative offset */
4808 chunk_appendf(buf, "+%#lx", (long)(addr - dli.dli_fbase));
4809 return NULL;
4810 }
Baruch Siache1651b22020-07-24 07:52:20 +03004811#endif /* __ELF__ && !__linux__ || USE_DL */
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004812 unknown:
4813 /* unresolved symbol from the main file, report relative offset to main */
4814 if ((void*)addr < (void*)main)
4815 chunk_appendf(buf, "main-%#lx", (long)((void*)main - addr));
4816 else
4817 chunk_appendf(buf, "main+%#lx", (long)(addr - (void*)main));
4818 return NULL;
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004819}
4820
Frédéric Lécaille3b717162019-02-25 15:04:22 +01004821/*
4822 * Allocate an array of unsigned int with <nums> as address from <str> string
Ilya Shipitsin46a030c2020-07-05 16:36:08 +05004823 * made of integer separated by dot characters.
Frédéric Lécaille3b717162019-02-25 15:04:22 +01004824 *
4825 * First, initializes the value with <sz> as address to 0 and initializes the
4826 * array with <nums> as address to NULL. Then allocates the array with <nums> as
4827 * address updating <sz> pointed value to the size of this array.
4828 *
4829 * Returns 1 if succeeded, 0 if not.
4830 */
4831int parse_dotted_uints(const char *str, unsigned int **nums, size_t *sz)
4832{
4833 unsigned int *n;
4834 const char *s, *end;
4835
4836 s = str;
4837 *sz = 0;
4838 end = str + strlen(str);
4839 *nums = n = NULL;
4840
4841 while (1) {
4842 unsigned int r;
4843
4844 if (s >= end)
4845 break;
4846
4847 r = read_uint(&s, end);
4848 /* Expected characters after having read an uint: '\0' or '.',
4849 * if '.', must not be terminal.
4850 */
Christopher Faulet4b524122021-02-11 10:42:41 +01004851 if (*s != '\0'&& (*s++ != '.' || s == end)) {
4852 free(n);
Frédéric Lécaille3b717162019-02-25 15:04:22 +01004853 return 0;
Christopher Faulet4b524122021-02-11 10:42:41 +01004854 }
Frédéric Lécaille3b717162019-02-25 15:04:22 +01004855
Frédéric Lécaille12a71842019-02-26 18:19:48 +01004856 n = my_realloc2(n, (*sz + 1) * sizeof *n);
Frédéric Lécaille3b717162019-02-25 15:04:22 +01004857 if (!n)
4858 return 0;
4859
4860 n[(*sz)++] = r;
4861 }
4862 *nums = n;
4863
4864 return 1;
4865}
4866
Willy Tarreau4d589e72019-08-23 19:02:26 +02004867
4868/* returns the number of bytes needed to encode <v> as a varint. An inline
4869 * version exists for use with constants (__varint_bytes()).
4870 */
4871int varint_bytes(uint64_t v)
4872{
4873 int len = 1;
4874
4875 if (v >= 240) {
4876 v = (v - 240) >> 4;
4877 while (1) {
4878 len++;
4879 if (v < 128)
4880 break;
4881 v = (v - 128) >> 7;
4882 }
4883 }
4884 return len;
4885}
4886
Willy Tarreau52bf8392020-03-08 00:42:37 +01004887
4888/* Random number generator state, see below */
Willy Tarreau1544c142020-03-12 00:31:18 +01004889static uint64_t ha_random_state[2] ALIGNED(2*sizeof(uint64_t));
Willy Tarreau52bf8392020-03-08 00:42:37 +01004890
4891/* This is a thread-safe implementation of xoroshiro128** described below:
4892 * http://prng.di.unimi.it/
4893 * It features a 2^128 long sequence, returns 64 high-quality bits on each call,
4894 * supports fast jumps and passes all common quality tests. It is thread-safe,
4895 * uses a double-cas on 64-bit architectures supporting it, and falls back to a
4896 * local lock on other ones.
4897 */
4898uint64_t ha_random64()
4899{
4900 uint64_t result;
Willy Tarreau1544c142020-03-12 00:31:18 +01004901 uint64_t old[2] ALIGNED(2*sizeof(uint64_t));
4902 uint64_t new[2] ALIGNED(2*sizeof(uint64_t));
Willy Tarreau52bf8392020-03-08 00:42:37 +01004903
4904#if defined(USE_THREAD) && (!defined(HA_CAS_IS_8B) || !defined(HA_HAVE_CAS_DW))
4905 static HA_SPINLOCK_T rand_lock;
4906
4907 HA_SPIN_LOCK(OTHER_LOCK, &rand_lock);
4908#endif
4909
4910 old[0] = ha_random_state[0];
4911 old[1] = ha_random_state[1];
4912
4913#if defined(USE_THREAD) && defined(HA_CAS_IS_8B) && defined(HA_HAVE_CAS_DW)
4914 do {
4915#endif
4916 result = rotl64(old[0] * 5, 7) * 9;
4917 new[1] = old[0] ^ old[1];
4918 new[0] = rotl64(old[0], 24) ^ new[1] ^ (new[1] << 16); // a, b
4919 new[1] = rotl64(new[1], 37); // c
4920
4921#if defined(USE_THREAD) && defined(HA_CAS_IS_8B) && defined(HA_HAVE_CAS_DW)
4922 } while (unlikely(!_HA_ATOMIC_DWCAS(ha_random_state, old, new)));
4923#else
4924 ha_random_state[0] = new[0];
4925 ha_random_state[1] = new[1];
4926#if defined(USE_THREAD)
4927 HA_SPIN_UNLOCK(OTHER_LOCK, &rand_lock);
4928#endif
4929#endif
4930 return result;
4931}
4932
4933/* seeds the random state using up to <len> bytes from <seed>, starting with
4934 * the first non-zero byte.
4935 */
4936void ha_random_seed(const unsigned char *seed, size_t len)
4937{
4938 size_t pos;
4939
4940 /* the seed must not be all zeroes, so we pre-fill it with alternating
4941 * bits and overwrite part of them with the block starting at the first
4942 * non-zero byte from the seed.
4943 */
4944 memset(ha_random_state, 0x55, sizeof(ha_random_state));
4945
4946 for (pos = 0; pos < len; pos++)
4947 if (seed[pos] != 0)
4948 break;
4949
4950 if (pos == len)
4951 return;
4952
4953 seed += pos;
4954 len -= pos;
4955
4956 if (len > sizeof(ha_random_state))
4957 len = sizeof(ha_random_state);
4958
4959 memcpy(ha_random_state, seed, len);
4960}
4961
4962/* This causes a jump to (dist * 2^96) places in the pseudo-random sequence,
4963 * and is equivalent to calling ha_random64() as many times. It is used to
4964 * provide non-overlapping sequences of 2^96 numbers (~7*10^28) to up to 2^32
4965 * different generators (i.e. different processes after a fork). The <dist>
4966 * argument is the distance to jump to and is used in a loop so it rather not
4967 * be too large if the processing time is a concern.
4968 *
4969 * BEWARE: this function is NOT thread-safe and must not be called during
4970 * concurrent accesses to ha_random64().
4971 */
4972void ha_random_jump96(uint32_t dist)
4973{
4974 while (dist--) {
4975 uint64_t s0 = 0;
4976 uint64_t s1 = 0;
4977 int b;
4978
4979 for (b = 0; b < 64; b++) {
4980 if ((0xd2a98b26625eee7bULL >> b) & 1) {
4981 s0 ^= ha_random_state[0];
4982 s1 ^= ha_random_state[1];
4983 }
4984 ha_random64();
4985 }
4986
4987 for (b = 0; b < 64; b++) {
4988 if ((0xdddf9b1090aa7ac1ULL >> b) & 1) {
4989 s0 ^= ha_random_state[0];
4990 s1 ^= ha_random_state[1];
4991 }
4992 ha_random64();
4993 }
4994 ha_random_state[0] = s0;
4995 ha_random_state[1] = s1;
4996 }
4997}
4998
Willy Tarreauee3bcdd2020-03-08 17:48:17 +01004999/* Generates an RFC4122 UUID into chunk <output> which must be at least 37
5000 * bytes large.
5001 */
5002void ha_generate_uuid(struct buffer *output)
5003{
5004 uint32_t rnd[4];
5005 uint64_t last;
5006
5007 last = ha_random64();
5008 rnd[0] = last;
5009 rnd[1] = last >> 32;
5010
5011 last = ha_random64();
5012 rnd[2] = last;
5013 rnd[3] = last >> 32;
5014
5015 chunk_printf(output, "%8.8x-%4.4x-%4.4x-%4.4x-%12.12llx",
5016 rnd[0],
5017 rnd[1] & 0xFFFF,
5018 ((rnd[1] >> 16u) & 0xFFF) | 0x4000, // highest 4 bits indicate the uuid version
5019 (rnd[2] & 0x3FFF) | 0x8000, // the highest 2 bits indicate the UUID variant (10),
5020 (long long)((rnd[2] >> 14u) | ((uint64_t) rnd[3] << 18u)) & 0xFFFFFFFFFFFFull);
5021}
5022
5023
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005024/* only used by parse_line() below. It supports writing in place provided that
5025 * <in> is updated to the next location before calling it. In that case, the
5026 * char at <in> may be overwritten.
5027 */
5028#define EMIT_CHAR(x) \
5029 do { \
5030 char __c = (char)(x); \
5031 if ((opts & PARSE_OPT_INPLACE) && out+outpos > in) \
5032 err |= PARSE_ERR_OVERLAP; \
5033 if (outpos >= outmax) \
5034 err |= PARSE_ERR_TOOLARGE; \
5035 if (!err) \
5036 out[outpos] = __c; \
5037 outpos++; \
5038 } while (0)
5039
Ilya Shipitsin46a030c2020-07-05 16:36:08 +05005040/* Parse <in>, copy it into <out> split into isolated words whose pointers
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005041 * are put in <args>. If more than <outlen> bytes have to be emitted, the
5042 * extraneous ones are not emitted but <outlen> is updated so that the caller
5043 * knows how much to realloc. Similarly, <args> are not updated beyond <nbargs>
5044 * but the returned <nbargs> indicates how many were found. All trailing args
Willy Tarreau61dd44b2020-06-25 07:35:42 +02005045 * up to <nbargs> point to the trailing zero, and as long as <nbargs> is > 0,
5046 * it is guaranteed that at least one arg will point to the zero. It is safe
5047 * to call it with a NULL <args> if <nbargs> is 0.
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005048 *
5049 * <out> may overlap with <in> provided that it never goes further, in which
5050 * case the parser will accept to perform in-place parsing and unquoting/
5051 * unescaping but only if environment variables do not lead to expansion that
5052 * causes overlapping, otherwise the input string being destroyed, the error
5053 * will not be recoverable. Note that even during out-of-place <in> will
5054 * experience temporary modifications in-place for variable resolution and must
5055 * be writable, and will also receive zeroes to delimit words when using
5056 * in-place copy. Parsing options <opts> taken from PARSE_OPT_*. Return value
5057 * is zero on success otherwise a bitwise-or of PARSE_ERR_*. Upon error, the
5058 * starting point of the first invalid character sequence or unmatched
5059 * quote/brace is reported in <errptr> if not NULL. When using in-place parsing
5060 * error reporting might be difficult since zeroes will have been inserted into
5061 * the string. One solution for the caller may consist in replacing all args
5062 * delimiters with spaces in this case.
5063 */
5064uint32_t parse_line(char *in, char *out, size_t *outlen, char **args, int *nbargs, uint32_t opts, char **errptr)
5065{
5066 char *quote = NULL;
5067 char *brace = NULL;
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005068 char *word_expand = NULL;
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005069 unsigned char hex1, hex2;
5070 size_t outmax = *outlen;
Willy Tarreau61dd44b2020-06-25 07:35:42 +02005071 int argsmax = *nbargs - 1;
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005072 size_t outpos = 0;
5073 int squote = 0;
5074 int dquote = 0;
5075 int arg = 0;
5076 uint32_t err = 0;
5077
5078 *nbargs = 0;
5079 *outlen = 0;
5080
Willy Tarreau61dd44b2020-06-25 07:35:42 +02005081 /* argsmax may be -1 here, protecting args[] from any write */
5082 if (arg < argsmax)
5083 args[arg] = out;
5084
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005085 while (1) {
5086 if (*in >= '-' && *in != '\\') {
5087 /* speedup: directly send all regular chars starting
5088 * with '-', '.', '/', alnum etc...
5089 */
5090 EMIT_CHAR(*in++);
5091 continue;
5092 }
5093 else if (*in == '\0' || *in == '\n' || *in == '\r') {
5094 /* end of line */
5095 break;
5096 }
5097 else if (*in == '#' && (opts & PARSE_OPT_SHARP) && !squote && !dquote) {
5098 /* comment */
5099 break;
5100 }
5101 else if (*in == '"' && !squote && (opts & PARSE_OPT_DQUOTE)) { /* double quote outside single quotes */
5102 if (dquote) {
5103 dquote = 0;
5104 quote = NULL;
5105 }
5106 else {
5107 dquote = 1;
5108 quote = in;
5109 }
5110 in++;
5111 continue;
5112 }
5113 else if (*in == '\'' && !dquote && (opts & PARSE_OPT_SQUOTE)) { /* single quote outside double quotes */
5114 if (squote) {
5115 squote = 0;
5116 quote = NULL;
5117 }
5118 else {
5119 squote = 1;
5120 quote = in;
5121 }
5122 in++;
5123 continue;
5124 }
5125 else if (*in == '\\' && !squote && (opts & PARSE_OPT_BKSLASH)) {
5126 /* first, we'll replace \\, \<space>, \#, \r, \n, \t, \xXX with their
5127 * C equivalent value but only when they have a special meaning and within
5128 * double quotes for some of them. Other combinations left unchanged (eg: \1).
5129 */
5130 char tosend = *in;
5131
5132 switch (in[1]) {
5133 case ' ':
5134 case '\\':
5135 tosend = in[1];
5136 in++;
5137 break;
5138
5139 case 't':
5140 tosend = '\t';
5141 in++;
5142 break;
5143
5144 case 'n':
5145 tosend = '\n';
5146 in++;
5147 break;
5148
5149 case 'r':
5150 tosend = '\r';
5151 in++;
5152 break;
5153
5154 case '#':
5155 /* escaping of "#" only if comments are supported */
5156 if (opts & PARSE_OPT_SHARP)
5157 in++;
5158 tosend = *in;
5159 break;
5160
5161 case '\'':
5162 /* escaping of "'" only outside single quotes and only if single quotes are supported */
5163 if (opts & PARSE_OPT_SQUOTE && !squote)
5164 in++;
5165 tosend = *in;
5166 break;
5167
5168 case '"':
5169 /* escaping of '"' only outside single quotes and only if double quotes are supported */
5170 if (opts & PARSE_OPT_DQUOTE && !squote)
5171 in++;
5172 tosend = *in;
5173 break;
5174
5175 case '$':
5176 /* escaping of '$' only inside double quotes and only if env supported */
5177 if (opts & PARSE_OPT_ENV && dquote)
5178 in++;
5179 tosend = *in;
5180 break;
5181
5182 case 'x':
5183 if (!ishex(in[2]) || !ishex(in[3])) {
5184 /* invalid or incomplete hex sequence */
5185 err |= PARSE_ERR_HEX;
5186 if (errptr)
5187 *errptr = in;
5188 goto leave;
5189 }
Willy Tarreauf278eec2020-07-05 21:46:32 +02005190 hex1 = toupper((unsigned char)in[2]) - '0';
5191 hex2 = toupper((unsigned char)in[3]) - '0';
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005192 if (hex1 > 9) hex1 -= 'A' - '9' - 1;
5193 if (hex2 > 9) hex2 -= 'A' - '9' - 1;
5194 tosend = (hex1 << 4) + hex2;
5195 in += 3;
5196 break;
5197
5198 default:
5199 /* other combinations are not escape sequences */
5200 break;
5201 }
5202
5203 in++;
5204 EMIT_CHAR(tosend);
5205 }
5206 else if (isspace((unsigned char)*in) && !squote && !dquote) {
5207 /* a non-escaped space is an argument separator */
5208 while (isspace((unsigned char)*in))
5209 in++;
5210 EMIT_CHAR(0);
5211 arg++;
5212 if (arg < argsmax)
5213 args[arg] = out + outpos;
5214 else
5215 err |= PARSE_ERR_TOOMANY;
5216 }
5217 else if (*in == '$' && (opts & PARSE_OPT_ENV) && (dquote || !(opts & PARSE_OPT_DQUOTE))) {
5218 /* environment variables are evaluated anywhere, or only
5219 * inside double quotes if they are supported.
5220 */
5221 char *var_name;
5222 char save_char;
5223 char *value;
5224
5225 in++;
5226
5227 if (*in == '{')
5228 brace = in++;
5229
5230 if (!isalpha((unsigned char)*in) && *in != '_') {
5231 /* unacceptable character in variable name */
5232 err |= PARSE_ERR_VARNAME;
5233 if (errptr)
5234 *errptr = in;
5235 goto leave;
5236 }
5237
5238 var_name = in;
5239 while (isalnum((unsigned char)*in) || *in == '_')
5240 in++;
5241
5242 save_char = *in;
5243 *in = '\0';
5244 value = getenv(var_name);
5245 *in = save_char;
5246
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005247 /* support for '[*]' sequence to force word expansion,
5248 * only available inside braces */
5249 if (*in == '[' && brace && (opts & PARSE_OPT_WORD_EXPAND)) {
5250 word_expand = in++;
5251
5252 if (*in++ != '*' || *in++ != ']') {
5253 err |= PARSE_ERR_WRONG_EXPAND;
5254 if (errptr)
5255 *errptr = word_expand;
5256 goto leave;
5257 }
5258 }
5259
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005260 if (brace) {
5261 if (*in != '}') {
5262 /* unmatched brace */
5263 err |= PARSE_ERR_BRACE;
5264 if (errptr)
5265 *errptr = brace;
5266 goto leave;
5267 }
5268 in++;
5269 brace = NULL;
5270 }
5271
5272 if (value) {
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005273 while (*value) {
5274 /* expand as individual parameters on a space character */
Willy Tarreaufe2cc412020-10-01 18:04:40 +02005275 if (word_expand && isspace((unsigned char)*value)) {
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005276 EMIT_CHAR(0);
5277 ++arg;
5278 if (arg < argsmax)
5279 args[arg] = out + outpos;
5280 else
5281 err |= PARSE_ERR_TOOMANY;
5282
5283 /* skip consecutive spaces */
Willy Tarreaufe2cc412020-10-01 18:04:40 +02005284 while (isspace((unsigned char)*++value))
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005285 ;
5286 } else {
5287 EMIT_CHAR(*value++);
5288 }
5289 }
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005290 }
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005291 word_expand = NULL;
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005292 }
5293 else {
5294 /* any other regular char */
5295 EMIT_CHAR(*in++);
5296 }
5297 }
5298
5299 /* end of output string */
5300 EMIT_CHAR(0);
5301 arg++;
5302
5303 if (quote) {
5304 /* unmatched quote */
5305 err |= PARSE_ERR_QUOTE;
5306 if (errptr)
5307 *errptr = quote;
5308 goto leave;
5309 }
5310 leave:
5311 *nbargs = arg;
5312 *outlen = outpos;
5313
Willy Tarreau61dd44b2020-06-25 07:35:42 +02005314 /* empty all trailing args by making them point to the trailing zero,
5315 * at least the last one in any case.
5316 */
5317 if (arg > argsmax)
5318 arg = argsmax;
5319
5320 while (arg >= 0 && arg <= argsmax)
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005321 args[arg++] = out + outpos - 1;
5322
5323 return err;
5324}
5325#undef EMIT_CHAR
5326
Willy Tarreauc54e5ad2020-06-25 09:15:40 +02005327/* This is used to sanitize an input line that's about to be used for error reporting.
5328 * It will adjust <line> to print approximately <width> chars around <pos>, trying to
5329 * preserve the beginning, with leading or trailing "..." when the line is truncated.
5330 * If non-printable chars are present in the output. It returns the new offset <pos>
5331 * in the modified line. Non-printable characters are replaced with '?'. <width> must
5332 * be at least 6 to support two "..." otherwise the result is undefined. The line
5333 * itself must have at least 7 chars allocated for the same reason.
5334 */
5335size_t sanitize_for_printing(char *line, size_t pos, size_t width)
5336{
5337 size_t shift = 0;
5338 char *out = line;
5339 char *in = line;
5340 char *end = line + width;
5341
5342 if (pos >= width) {
5343 /* if we have to shift, we'll be out of context, so let's
5344 * try to put <pos> at the center of width.
5345 */
5346 shift = pos - width / 2;
5347 in += shift + 3;
5348 end = out + width - 3;
5349 out[0] = out[1] = out[2] = '.';
5350 out += 3;
5351 }
5352
5353 while (out < end && *in) {
5354 if (isspace((unsigned char)*in))
5355 *out++ = ' ';
5356 else if (isprint((unsigned char)*in))
5357 *out++ = *in;
5358 else
5359 *out++ = '?';
5360 in++;
5361 }
5362
5363 if (end < line + width) {
5364 out[0] = out[1] = out[2] = '.';
5365 out += 3;
5366 }
5367
5368 *out++ = 0;
5369 return pos - shift;
5370}
Willy Tarreau06e69b52021-03-02 14:01:35 +01005371
Willy Tarreaue33c4b32021-03-12 18:59:31 +01005372/* Update array <fp> with the fingerprint of word <word> by counting the
Willy Tarreauba2c4452021-03-12 09:01:52 +01005373 * transitions between characters. <fp> is a 1024-entries array indexed as
5374 * 32*from+to. Positions for 'from' and 'to' are:
Willy Tarreau9294e882021-03-15 09:34:27 +01005375 * 1..26=letter, 27=digit, 28=other/begin/end.
5376 * Row "from=0" is used to mark the character's presence. Others unused.
Willy Tarreauba2c4452021-03-12 09:01:52 +01005377 */
Willy Tarreaue33c4b32021-03-12 18:59:31 +01005378void update_word_fingerprint(uint8_t *fp, const char *word)
Willy Tarreauba2c4452021-03-12 09:01:52 +01005379{
5380 const char *p;
5381 int from, to;
5382 int c;
5383
Willy Tarreauba2c4452021-03-12 09:01:52 +01005384 from = 28; // begin
5385 for (p = word; *p; p++) {
5386 c = tolower(*p);
5387 switch(c) {
Willy Tarreau9294e882021-03-15 09:34:27 +01005388 case 'a'...'z': to = c - 'a' + 1; break;
5389 case 'A'...'Z': to = tolower(c) - 'a' + 1; break;
5390 case '0'...'9': to = 27; break;
5391 default: to = 28; break;
Willy Tarreauba2c4452021-03-12 09:01:52 +01005392 }
Willy Tarreau9294e882021-03-15 09:34:27 +01005393 fp[to] = 1;
Willy Tarreauba2c4452021-03-12 09:01:52 +01005394 fp[32 * from + to]++;
5395 from = to;
5396 }
5397 to = 28; // end
5398 fp[32 * from + to]++;
5399}
5400
Willy Tarreaue33c4b32021-03-12 18:59:31 +01005401/* Initialize array <fp> with the fingerprint of word <word> by counting the
5402 * transitions between characters. <fp> is a 1024-entries array indexed as
5403 * 32*from+to. Positions for 'from' and 'to' are:
5404 * 0..25=letter, 26=digit, 27=other, 28=begin, 29=end, others unused.
5405 */
5406void make_word_fingerprint(uint8_t *fp, const char *word)
5407{
5408 memset(fp, 0, 1024);
5409 update_word_fingerprint(fp, word);
5410}
5411
Willy Tarreauba2c4452021-03-12 09:01:52 +01005412/* Return the distance between two word fingerprints created by function
5413 * make_word_fingerprint(). It's a positive integer calculated as the sum of
Willy Tarreau714c4c12021-03-15 09:44:53 +01005414 * the differences between each location.
Willy Tarreauba2c4452021-03-12 09:01:52 +01005415 */
5416int word_fingerprint_distance(const uint8_t *fp1, const uint8_t *fp2)
5417{
5418 int i, k, dist = 0;
5419
5420 for (i = 0; i < 1024; i++) {
5421 k = (int)fp1[i] - (int)fp2[i];
Willy Tarreau714c4c12021-03-15 09:44:53 +01005422 dist += abs(k);
Willy Tarreauba2c4452021-03-12 09:01:52 +01005423 }
5424 return dist;
5425}
5426
Willy Tarreau06e69b52021-03-02 14:01:35 +01005427static int init_tools_per_thread()
5428{
5429 /* Let's make each thread start from a different position */
5430 statistical_prng_state += tid * MAX_THREADS;
5431 if (!statistical_prng_state)
5432 statistical_prng_state++;
5433 return 1;
5434}
5435REGISTER_PER_THREAD_INIT(init_tools_per_thread);
Willy Tarreauc54e5ad2020-06-25 09:15:40 +02005436
Willy Tarreaubaaee002006-06-26 02:48:02 +02005437/*
5438 * Local variables:
5439 * c-indent-level: 8
5440 * c-basic-offset: 8
5441 * End:
5442 */