blob: 71337aa6412a889d20c1ce2b1a3ad6fc57332fbc [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
devnexen@gmail.comc4e52322021-08-17 12:55:49 +010019#if defined(__FreeBSD__)
20#include <elf.h>
21#include <dlfcn.h>
22extern void *__elf_aux_vector;
23#endif
24
David Carlierbd2cced2021-08-17 08:44:25 +010025#if defined(__NetBSD__)
26#include <sys/exec_elf.h>
27#include <dlfcn.h>
28#endif
29
Willy Tarreau2e74c3f2007-12-02 18:45:09 +010030#include <ctype.h>
Willy Tarreau16e01562016-08-09 16:46:18 +020031#include <errno.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020032#include <netdb.h>
Willy Tarreau9a7bea52012-04-27 11:16:50 +020033#include <stdarg.h>
Willy Tarreaudd2f85e2012-09-02 22:34:23 +020034#include <stdio.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020035#include <stdlib.h>
36#include <string.h>
Thierry Fournier93127942016-01-20 18:49:45 +010037#include <time.h>
Willy Tarreau16e01562016-08-09 16:46:18 +020038#include <unistd.h>
Willy Tarreau127f9662007-12-06 00:53:51 +010039#include <sys/socket.h>
Willy Tarreau37101052019-05-20 16:48:20 +020040#include <sys/stat.h>
41#include <sys/types.h>
Willy Tarreau127f9662007-12-06 00:53:51 +010042#include <sys/un.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020043#include <netinet/in.h>
44#include <arpa/inet.h>
45
David Carlier43a56852022-03-04 15:50:48 +000046#if defined(__linux__) && defined(__GLIBC__) && (__GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 16))
Willy Tarreau30053062020-08-20 16:39:14 +020047#include <sys/auxv.h>
48#endif
49
Willy Tarreau48fbcae2020-06-03 18:09:46 +020050#include <import/eb32sctree.h>
Willy Tarreaub2551052020-06-09 09:07:15 +020051#include <import/eb32tree.h>
Willy Tarreau48fbcae2020-06-03 18:09:46 +020052
Willy Tarreau4c7e4b72020-05-27 12:58:42 +020053#include <haproxy/api.h>
Christopher Faulet6b0a0fb2022-04-04 11:29:28 +020054#include <haproxy/applet.h>
Willy Tarreauc13ed532020-06-02 10:22:45 +020055#include <haproxy/chunk.h>
Christopher Faulet5e29b762022-04-04 08:58:34 +020056#include <haproxy/conn_stream.h>
57#include <haproxy/cs_utils.h>
Willy Tarreau7c18b542020-06-11 09:23:02 +020058#include <haproxy/dgram.h>
Willy Tarreauf268ee82020-06-04 17:05:57 +020059#include <haproxy/global.h>
Willy Tarreau86416052020-06-04 09:20:54 +020060#include <haproxy/hlua.h>
Willy Tarreau213e9902020-06-04 14:58:24 +020061#include <haproxy/listener.h>
Willy Tarreau7a00efb2020-06-02 17:02:59 +020062#include <haproxy/namespace.h>
Christopher Faulet9553de72021-02-26 09:12:50 +010063#include <haproxy/net_helper.h>
Willy Tarreau5fc93282020-09-16 18:25:03 +020064#include <haproxy/protocol.h>
Emeric Brunc9437992021-02-12 19:42:55 +010065#include <haproxy/resolvers.h>
Willy Tarreau586f71b2020-12-11 15:54:36 +010066#include <haproxy/sock.h>
Willy Tarreau209108d2020-06-04 20:30:20 +020067#include <haproxy/ssl_sock.h>
William Lallemand3aeb3f92021-08-21 23:59:56 +020068#include <haproxy/ssl_utils.h>
Willy Tarreaucea0e1b2020-06-04 17:25:40 +020069#include <haproxy/task.h>
Willy Tarreau48fbcae2020-06-03 18:09:46 +020070#include <haproxy/tools.h>
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +010071
Thierry Fournier93127942016-01-20 18:49:45 +010072/* This macro returns false if the test __x is false. Many
73 * of the following parsing function must be abort the processing
74 * if it returns 0, so this macro is useful for writing light code.
75 */
76#define RET0_UNLESS(__x) do { if (!(__x)) return 0; } while (0)
77
Willy Tarreau56adcf22012-12-23 18:00:29 +010078/* enough to store NB_ITOA_STR integers of :
Willy Tarreau72d759c2007-10-25 12:14:10 +020079 * 2^64-1 = 18446744073709551615 or
80 * -2^63 = -9223372036854775808
Willy Tarreaue7239b52009-03-29 13:41:58 +020081 *
82 * The HTML version needs room for adding the 25 characters
83 * '<span class="rls"></span>' around digits at positions 3N+1 in order
84 * to add spacing at up to 6 positions : 18 446 744 073 709 551 615
Willy Tarreau72d759c2007-10-25 12:14:10 +020085 */
Christopher Faulet99bca652017-11-14 16:47:26 +010086THREAD_LOCAL char itoa_str[NB_ITOA_STR][171];
87THREAD_LOCAL int itoa_idx = 0; /* index of next itoa_str to use */
Willy Tarreaubaaee002006-06-26 02:48:02 +020088
Willy Tarreau588297f2014-06-16 15:16:40 +020089/* sometimes we'll need to quote strings (eg: in stats), and we don't expect
90 * to quote strings larger than a max configuration line.
91 */
Christopher Faulet99bca652017-11-14 16:47:26 +010092THREAD_LOCAL char quoted_str[NB_QSTR][QSTR_SIZE + 1];
93THREAD_LOCAL int quoted_idx = 0;
Willy Tarreau588297f2014-06-16 15:16:40 +020094
Willy Tarreau06e69b52021-03-02 14:01:35 +010095/* thread-local PRNG state. It's modified to start from a different sequence
96 * on all threads upon startup. It must not be used or anything beyond getting
97 * statistical values as it's 100% predictable.
98 */
99THREAD_LOCAL unsigned int statistical_prng_state = 2463534242U;
100
Willy Tarreaubaaee002006-06-26 02:48:02 +0200101/*
William Lallemande7340ec2012-01-24 11:15:39 +0100102 * unsigned long long ASCII representation
103 *
104 * return the last char '\0' or NULL if no enough
105 * space in dst
106 */
107char *ulltoa(unsigned long long n, char *dst, size_t size)
108{
109 int i = 0;
110 char *res;
111
112 switch(n) {
113 case 1ULL ... 9ULL:
114 i = 0;
115 break;
116
117 case 10ULL ... 99ULL:
118 i = 1;
119 break;
120
121 case 100ULL ... 999ULL:
122 i = 2;
123 break;
124
125 case 1000ULL ... 9999ULL:
126 i = 3;
127 break;
128
129 case 10000ULL ... 99999ULL:
130 i = 4;
131 break;
132
133 case 100000ULL ... 999999ULL:
134 i = 5;
135 break;
136
137 case 1000000ULL ... 9999999ULL:
138 i = 6;
139 break;
140
141 case 10000000ULL ... 99999999ULL:
142 i = 7;
143 break;
144
145 case 100000000ULL ... 999999999ULL:
146 i = 8;
147 break;
148
149 case 1000000000ULL ... 9999999999ULL:
150 i = 9;
151 break;
152
153 case 10000000000ULL ... 99999999999ULL:
154 i = 10;
155 break;
156
157 case 100000000000ULL ... 999999999999ULL:
158 i = 11;
159 break;
160
161 case 1000000000000ULL ... 9999999999999ULL:
162 i = 12;
163 break;
164
165 case 10000000000000ULL ... 99999999999999ULL:
166 i = 13;
167 break;
168
169 case 100000000000000ULL ... 999999999999999ULL:
170 i = 14;
171 break;
172
173 case 1000000000000000ULL ... 9999999999999999ULL:
174 i = 15;
175 break;
176
177 case 10000000000000000ULL ... 99999999999999999ULL:
178 i = 16;
179 break;
180
181 case 100000000000000000ULL ... 999999999999999999ULL:
182 i = 17;
183 break;
184
185 case 1000000000000000000ULL ... 9999999999999999999ULL:
186 i = 18;
187 break;
188
189 case 10000000000000000000ULL ... ULLONG_MAX:
190 i = 19;
191 break;
192 }
193 if (i + 2 > size) // (i + 1) + '\0'
194 return NULL; // too long
195 res = dst + i + 1;
196 *res = '\0';
197 for (; i >= 0; i--) {
198 dst[i] = n % 10ULL + '0';
199 n /= 10ULL;
200 }
201 return res;
202}
203
204/*
205 * unsigned long ASCII representation
206 *
207 * return the last char '\0' or NULL if no enough
208 * space in dst
209 */
210char *ultoa_o(unsigned long n, char *dst, size_t size)
211{
212 int i = 0;
213 char *res;
214
215 switch (n) {
216 case 0U ... 9UL:
217 i = 0;
218 break;
219
220 case 10U ... 99UL:
221 i = 1;
222 break;
223
224 case 100U ... 999UL:
225 i = 2;
226 break;
227
228 case 1000U ... 9999UL:
229 i = 3;
230 break;
231
232 case 10000U ... 99999UL:
233 i = 4;
234 break;
235
236 case 100000U ... 999999UL:
237 i = 5;
238 break;
239
240 case 1000000U ... 9999999UL:
241 i = 6;
242 break;
243
244 case 10000000U ... 99999999UL:
245 i = 7;
246 break;
247
248 case 100000000U ... 999999999UL:
249 i = 8;
250 break;
251#if __WORDSIZE == 32
252
253 case 1000000000ULL ... ULONG_MAX:
254 i = 9;
255 break;
256
257#elif __WORDSIZE == 64
258
259 case 1000000000ULL ... 9999999999UL:
260 i = 9;
261 break;
262
263 case 10000000000ULL ... 99999999999UL:
264 i = 10;
265 break;
266
267 case 100000000000ULL ... 999999999999UL:
268 i = 11;
269 break;
270
271 case 1000000000000ULL ... 9999999999999UL:
272 i = 12;
273 break;
274
275 case 10000000000000ULL ... 99999999999999UL:
276 i = 13;
277 break;
278
279 case 100000000000000ULL ... 999999999999999UL:
280 i = 14;
281 break;
282
283 case 1000000000000000ULL ... 9999999999999999UL:
284 i = 15;
285 break;
286
287 case 10000000000000000ULL ... 99999999999999999UL:
288 i = 16;
289 break;
290
291 case 100000000000000000ULL ... 999999999999999999UL:
292 i = 17;
293 break;
294
295 case 1000000000000000000ULL ... 9999999999999999999UL:
296 i = 18;
297 break;
298
299 case 10000000000000000000ULL ... ULONG_MAX:
300 i = 19;
301 break;
302
303#endif
304 }
305 if (i + 2 > size) // (i + 1) + '\0'
306 return NULL; // too long
307 res = dst + i + 1;
308 *res = '\0';
309 for (; i >= 0; i--) {
310 dst[i] = n % 10U + '0';
311 n /= 10U;
312 }
313 return res;
314}
315
316/*
317 * signed long ASCII representation
318 *
319 * return the last char '\0' or NULL if no enough
320 * space in dst
321 */
322char *ltoa_o(long int n, char *dst, size_t size)
323{
324 char *pos = dst;
325
326 if (n < 0) {
327 if (size < 3)
328 return NULL; // min size is '-' + digit + '\0' but another test in ultoa
329 *pos = '-';
330 pos++;
331 dst = ultoa_o(-n, pos, size - 1);
332 } else {
333 dst = ultoa_o(n, dst, size);
334 }
335 return dst;
336}
337
338/*
339 * signed long long ASCII representation
340 *
341 * return the last char '\0' or NULL if no enough
342 * space in dst
343 */
344char *lltoa(long long n, char *dst, size_t size)
345{
346 char *pos = dst;
347
348 if (n < 0) {
349 if (size < 3)
350 return NULL; // min size is '-' + digit + '\0' but another test in ulltoa
351 *pos = '-';
352 pos++;
353 dst = ulltoa(-n, pos, size - 1);
354 } else {
355 dst = ulltoa(n, dst, size);
356 }
357 return dst;
358}
359
360/*
361 * write a ascii representation of a unsigned into dst,
362 * return a pointer to the last character
363 * Pad the ascii representation with '0', using size.
364 */
365char *utoa_pad(unsigned int n, char *dst, size_t size)
366{
367 int i = 0;
368 char *ret;
369
370 switch(n) {
371 case 0U ... 9U:
372 i = 0;
373 break;
374
375 case 10U ... 99U:
376 i = 1;
377 break;
378
379 case 100U ... 999U:
380 i = 2;
381 break;
382
383 case 1000U ... 9999U:
384 i = 3;
385 break;
386
387 case 10000U ... 99999U:
388 i = 4;
389 break;
390
391 case 100000U ... 999999U:
392 i = 5;
393 break;
394
395 case 1000000U ... 9999999U:
396 i = 6;
397 break;
398
399 case 10000000U ... 99999999U:
400 i = 7;
401 break;
402
403 case 100000000U ... 999999999U:
404 i = 8;
405 break;
406
407 case 1000000000U ... 4294967295U:
408 i = 9;
409 break;
410 }
411 if (i + 2 > size) // (i + 1) + '\0'
412 return NULL; // too long
413 if (i < size)
414 i = size - 2; // padding - '\0'
415
416 ret = dst + i + 1;
417 *ret = '\0';
418 for (; i >= 0; i--) {
419 dst[i] = n % 10U + '0';
420 n /= 10U;
421 }
422 return ret;
423}
424
425/*
Willy Tarreaubaaee002006-06-26 02:48:02 +0200426 * copies at most <size-1> chars from <src> to <dst>. Last char is always
427 * set to 0, unless <size> is 0. The number of chars copied is returned
428 * (excluding the terminating zero).
429 * This code has been optimized for size and speed : on x86, it's 45 bytes
430 * long, uses only registers, and consumes only 4 cycles per char.
431 */
432int strlcpy2(char *dst, const char *src, int size)
433{
434 char *orig = dst;
435 if (size) {
436 while (--size && (*dst = *src)) {
437 src++; dst++;
438 }
439 *dst = 0;
440 }
441 return dst - orig;
442}
443
444/*
Willy Tarreau72d759c2007-10-25 12:14:10 +0200445 * This function simply returns a locally allocated string containing
Willy Tarreaubaaee002006-06-26 02:48:02 +0200446 * the ascii representation for number 'n' in decimal.
447 */
Emeric Brun3a7fce52010-01-04 14:54:38 +0100448char *ultoa_r(unsigned long n, char *buffer, int size)
Willy Tarreaubaaee002006-06-26 02:48:02 +0200449{
450 char *pos;
451
Willy Tarreau72d759c2007-10-25 12:14:10 +0200452 pos = buffer + size - 1;
Willy Tarreaubaaee002006-06-26 02:48:02 +0200453 *pos-- = '\0';
454
455 do {
456 *pos-- = '0' + n % 10;
457 n /= 10;
Willy Tarreau72d759c2007-10-25 12:14:10 +0200458 } while (n && pos >= buffer);
Willy Tarreaubaaee002006-06-26 02:48:02 +0200459 return pos + 1;
460}
461
Willy Tarreau91092e52007-10-25 16:58:42 +0200462/*
Willy Tarreaue7239b52009-03-29 13:41:58 +0200463 * This function simply returns a locally allocated string containing
Thierry FOURNIER763a5d82015-07-06 23:09:52 +0200464 * the ascii representation for number 'n' in decimal.
465 */
466char *lltoa_r(long long int in, char *buffer, int size)
467{
468 char *pos;
469 int neg = 0;
470 unsigned long long int n;
471
472 pos = buffer + size - 1;
473 *pos-- = '\0';
474
475 if (in < 0) {
476 neg = 1;
477 n = -in;
478 }
479 else
480 n = in;
481
482 do {
483 *pos-- = '0' + n % 10;
484 n /= 10;
485 } while (n && pos >= buffer);
486 if (neg && pos > buffer)
487 *pos-- = '-';
488 return pos + 1;
489}
490
491/*
492 * This function simply returns a locally allocated string containing
Thierry FOURNIER1480bd82015-06-06 19:14:59 +0200493 * the ascii representation for signed number 'n' in decimal.
494 */
495char *sltoa_r(long n, char *buffer, int size)
496{
497 char *pos;
498
499 if (n >= 0)
500 return ultoa_r(n, buffer, size);
501
502 pos = ultoa_r(-n, buffer + 1, size - 1) - 1;
503 *pos = '-';
504 return pos;
505}
506
507/*
508 * This function simply returns a locally allocated string containing
Willy Tarreaue7239b52009-03-29 13:41:58 +0200509 * the ascii representation for number 'n' in decimal, formatted for
510 * HTML output with tags to create visual grouping by 3 digits. The
511 * output needs to support at least 171 characters.
512 */
513const char *ulltoh_r(unsigned long long n, char *buffer, int size)
514{
515 char *start;
516 int digit = 0;
517
518 start = buffer + size;
519 *--start = '\0';
520
521 do {
522 if (digit == 3 && start >= buffer + 7)
523 memcpy(start -= 7, "</span>", 7);
524
525 if (start >= buffer + 1) {
526 *--start = '0' + n % 10;
527 n /= 10;
528 }
529
530 if (digit == 3 && start >= buffer + 18)
531 memcpy(start -= 18, "<span class=\"rls\">", 18);
532
533 if (digit++ == 3)
534 digit = 1;
535 } while (n && start > buffer);
536 return start;
537}
538
539/*
Willy Tarreau91092e52007-10-25 16:58:42 +0200540 * This function simply returns a locally allocated string containing the ascii
541 * representation for number 'n' in decimal, unless n is 0 in which case it
542 * returns the alternate string (or an empty string if the alternate string is
543 * NULL). It use is intended for limits reported in reports, where it's
544 * desirable not to display anything if there is no limit. Warning! it shares
545 * the same vector as ultoa_r().
546 */
547const char *limit_r(unsigned long n, char *buffer, int size, const char *alt)
548{
549 return (n) ? ultoa_r(n, buffer, size) : (alt ? alt : "");
550}
551
Willy Tarreau56d1d8d2021-05-08 10:28:53 +0200552/* Trims the first "%f" float in a string to its minimum number of digits after
553 * the decimal point by trimming trailing zeroes, even dropping the decimal
554 * point if not needed. The string is in <buffer> of length <len>, and the
555 * number is expected to start at or after position <num_start> (the first
556 * point appearing there is considered). A NUL character is always placed at
557 * the end if some trimming occurs. The new buffer length is returned.
558 */
559size_t flt_trim(char *buffer, size_t num_start, size_t len)
560{
561 char *end = buffer + len;
562 char *p = buffer + num_start;
563 char *trim;
564
565 do {
566 if (p >= end)
567 return len;
568 trim = p++;
569 } while (*trim != '.');
570
571 /* For now <trim> is on the decimal point. Let's look for any other
572 * meaningful digit after it.
573 */
574 while (p < end) {
575 if (*p++ != '0')
576 trim = p;
577 }
578
579 if (trim < end)
580 *trim = 0;
581
582 return trim - buffer;
583}
584
Willy Tarreauae03d262021-05-08 07:35:00 +0200585/*
586 * This function simply returns a locally allocated string containing
587 * the ascii representation for number 'n' in decimal with useless trailing
588 * zeroes trimmed.
589 */
590char *ftoa_r(double n, char *buffer, int size)
591{
592 flt_trim(buffer, 0, snprintf(buffer, size, "%f", n));
593 return buffer;
594}
595
Willy Tarreau588297f2014-06-16 15:16:40 +0200596/* returns a locally allocated string containing the quoted encoding of the
597 * input string. The output may be truncated to QSTR_SIZE chars, but it is
598 * guaranteed that the string will always be properly terminated. Quotes are
599 * encoded by doubling them as is commonly done in CSV files. QSTR_SIZE must
600 * always be at least 4 chars.
601 */
602const char *qstr(const char *str)
603{
604 char *ret = quoted_str[quoted_idx];
605 char *p, *end;
606
607 if (++quoted_idx >= NB_QSTR)
608 quoted_idx = 0;
609
610 p = ret;
611 end = ret + QSTR_SIZE;
612
613 *p++ = '"';
614
615 /* always keep 3 chars to support passing "" and the ending " */
616 while (*str && p < end - 3) {
617 if (*str == '"') {
618 *p++ = '"';
619 *p++ = '"';
620 }
621 else
622 *p++ = *str;
623 str++;
624 }
625 *p++ = '"';
626 return ret;
627}
628
Robert Tsai81ae1952007-12-05 10:47:29 +0100629/*
Willy Tarreaubaaee002006-06-26 02:48:02 +0200630 * Returns non-zero if character <s> is a hex digit (0-9, a-f, A-F), else zero.
631 *
632 * It looks like this one would be a good candidate for inlining, but this is
633 * not interesting because it around 35 bytes long and often called multiple
634 * times within the same function.
635 */
636int ishex(char s)
637{
638 s -= '0';
639 if ((unsigned char)s <= 9)
640 return 1;
641 s -= 'A' - '0';
642 if ((unsigned char)s <= 5)
643 return 1;
644 s -= 'a' - 'A';
645 if ((unsigned char)s <= 5)
646 return 1;
647 return 0;
648}
649
Willy Tarreau3ca1a882015-01-15 18:43:49 +0100650/* rounds <i> down to the closest value having max 2 digits */
651unsigned int round_2dig(unsigned int i)
652{
653 unsigned int mul = 1;
654
655 while (i >= 100) {
656 i /= 10;
657 mul *= 10;
658 }
659 return i * mul;
660}
661
Willy Tarreau2e74c3f2007-12-02 18:45:09 +0100662/*
663 * Checks <name> for invalid characters. Valid chars are [A-Za-z0-9_:.-]. If an
664 * invalid character is found, a pointer to it is returned. If everything is
665 * fine, NULL is returned.
666 */
667const char *invalid_char(const char *name)
668{
669 if (!*name)
670 return name;
671
672 while (*name) {
Willy Tarreau90807112020-02-25 08:16:33 +0100673 if (!isalnum((unsigned char)*name) && *name != '.' && *name != ':' &&
Willy Tarreau2e74c3f2007-12-02 18:45:09 +0100674 *name != '_' && *name != '-')
675 return name;
676 name++;
677 }
678 return NULL;
679}
Willy Tarreaubaaee002006-06-26 02:48:02 +0200680
681/*
Frédéric Lécailleb82f7422017-04-13 18:24:23 +0200682 * Checks <name> for invalid characters. Valid chars are [_.-] and those
683 * accepted by <f> function.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +0200684 * If an invalid character is found, a pointer to it is returned.
685 * If everything is fine, NULL is returned.
686 */
Frédéric Lécailleb82f7422017-04-13 18:24:23 +0200687static inline const char *__invalid_char(const char *name, int (*f)(int)) {
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +0200688
689 if (!*name)
690 return name;
691
692 while (*name) {
Willy Tarreau90807112020-02-25 08:16:33 +0100693 if (!f((unsigned char)*name) && *name != '.' &&
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +0200694 *name != '_' && *name != '-')
695 return name;
696
697 name++;
698 }
699
700 return NULL;
701}
702
703/*
Frédéric Lécailleb82f7422017-04-13 18:24:23 +0200704 * Checks <name> for invalid characters. Valid chars are [A-Za-z0-9_.-].
705 * If an invalid character is found, a pointer to it is returned.
706 * If everything is fine, NULL is returned.
707 */
708const char *invalid_domainchar(const char *name) {
709 return __invalid_char(name, isalnum);
710}
711
712/*
713 * Checks <name> for invalid characters. Valid chars are [A-Za-z_.-].
714 * If an invalid character is found, a pointer to it is returned.
715 * If everything is fine, NULL is returned.
716 */
717const char *invalid_prefix_char(const char *name) {
Thierry Fournierf7b7c3e2018-03-26 11:54:39 +0200718 return __invalid_char(name, isalnum);
Frédéric Lécailleb82f7422017-04-13 18:24:23 +0200719}
720
721/*
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100722 * converts <str> to a struct sockaddr_storage* provided by the caller. The
Willy Tarreau24709282013-03-10 21:32:12 +0100723 * caller must have zeroed <sa> first, and may have set sa->ss_family to force
724 * parse a specific address format. If the ss_family is 0 or AF_UNSPEC, then
725 * the function tries to guess the address family from the syntax. If the
726 * family is forced and the format doesn't match, an error is returned. The
Willy Tarreaufab5a432011-03-04 15:31:53 +0100727 * string is assumed to contain only an address, no port. The address can be a
728 * dotted IPv4 address, an IPv6 address, a host name, or empty or "*" to
729 * indicate INADDR_ANY. NULL is returned if the host part cannot be resolved.
730 * The return address will only have the address family and the address set,
731 * all other fields remain zero. The string is not supposed to be modified.
Thierry FOURNIER58639a02014-11-25 12:02:25 +0100732 * The IPv6 '::' address is IN6ADDR_ANY. If <resolve> is non-zero, the hostname
733 * is resolved, otherwise only IP addresses are resolved, and anything else
Willy Tarreauecde7df2016-11-02 22:37:03 +0100734 * returns NULL. If the address contains a port, this one is preserved.
Willy Tarreaubaaee002006-06-26 02:48:02 +0200735 */
Thierry FOURNIER58639a02014-11-25 12:02:25 +0100736struct sockaddr_storage *str2ip2(const char *str, struct sockaddr_storage *sa, int resolve)
Willy Tarreaubaaee002006-06-26 02:48:02 +0200737{
Willy Tarreaufab5a432011-03-04 15:31:53 +0100738 struct hostent *he;
mildisff5d5102015-10-26 18:50:08 +0100739 /* max IPv6 length, including brackets and terminating NULL */
740 char tmpip[48];
Willy Tarreauecde7df2016-11-02 22:37:03 +0100741 int port = get_host_port(sa);
mildisff5d5102015-10-26 18:50:08 +0100742
743 /* check IPv6 with square brackets */
744 if (str[0] == '[') {
745 size_t iplength = strlen(str);
746
747 if (iplength < 4) {
748 /* minimal size is 4 when using brackets "[::]" */
749 goto fail;
750 }
751 else if (iplength >= sizeof(tmpip)) {
752 /* IPv6 literal can not be larger than tmpip */
753 goto fail;
754 }
755 else {
756 if (str[iplength - 1] != ']') {
757 /* if address started with bracket, it should end with bracket */
758 goto fail;
759 }
760 else {
761 memcpy(tmpip, str + 1, iplength - 2);
762 tmpip[iplength - 2] = '\0';
763 str = tmpip;
764 }
765 }
766 }
Willy Tarreaufab5a432011-03-04 15:31:53 +0100767
Willy Tarreaufab5a432011-03-04 15:31:53 +0100768 /* Any IPv6 address */
769 if (str[0] == ':' && str[1] == ':' && !str[2]) {
Willy Tarreau24709282013-03-10 21:32:12 +0100770 if (!sa->ss_family || sa->ss_family == AF_UNSPEC)
771 sa->ss_family = AF_INET6;
772 else if (sa->ss_family != AF_INET6)
773 goto fail;
Willy Tarreauecde7df2016-11-02 22:37:03 +0100774 set_host_port(sa, port);
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100775 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100776 }
777
Willy Tarreau24709282013-03-10 21:32:12 +0100778 /* Any address for the family, defaults to IPv4 */
Willy Tarreaufab5a432011-03-04 15:31:53 +0100779 if (!str[0] || (str[0] == '*' && !str[1])) {
Willy Tarreau24709282013-03-10 21:32:12 +0100780 if (!sa->ss_family || sa->ss_family == AF_UNSPEC)
781 sa->ss_family = AF_INET;
Willy Tarreauecde7df2016-11-02 22:37:03 +0100782 set_host_port(sa, port);
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100783 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100784 }
785
786 /* check for IPv6 first */
Willy Tarreau24709282013-03-10 21:32:12 +0100787 if ((!sa->ss_family || sa->ss_family == AF_UNSPEC || sa->ss_family == AF_INET6) &&
788 inet_pton(AF_INET6, str, &((struct sockaddr_in6 *)sa)->sin6_addr)) {
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100789 sa->ss_family = AF_INET6;
Willy Tarreauecde7df2016-11-02 22:37:03 +0100790 set_host_port(sa, port);
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100791 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100792 }
793
794 /* then check for IPv4 */
Willy Tarreau24709282013-03-10 21:32:12 +0100795 if ((!sa->ss_family || sa->ss_family == AF_UNSPEC || sa->ss_family == AF_INET) &&
796 inet_pton(AF_INET, str, &((struct sockaddr_in *)sa)->sin_addr)) {
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100797 sa->ss_family = AF_INET;
Willy Tarreauecde7df2016-11-02 22:37:03 +0100798 set_host_port(sa, port);
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100799 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100800 }
801
Thierry FOURNIER58639a02014-11-25 12:02:25 +0100802 if (!resolve)
803 return NULL;
804
Emeric Brund30e9a12020-12-23 18:49:16 +0100805 if (!resolv_hostname_validation(str, NULL))
Baptiste Assmanna68ca962015-04-14 01:15:08 +0200806 return NULL;
807
David du Colombierd5f43282011-03-17 10:40:16 +0100808#ifdef USE_GETADDRINFO
Nenad Merdanovic88afe032014-04-14 15:56:58 +0200809 if (global.tune.options & GTUNE_USE_GAI) {
David du Colombierd5f43282011-03-17 10:40:16 +0100810 struct addrinfo hints, *result;
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100811 int success = 0;
David du Colombierd5f43282011-03-17 10:40:16 +0100812
813 memset(&result, 0, sizeof(result));
814 memset(&hints, 0, sizeof(hints));
Willy Tarreau24709282013-03-10 21:32:12 +0100815 hints.ai_family = sa->ss_family ? sa->ss_family : AF_UNSPEC;
David du Colombierd5f43282011-03-17 10:40:16 +0100816 hints.ai_socktype = SOCK_DGRAM;
Dmitry Sivachenkoeab7f392015-10-02 01:01:58 +0200817 hints.ai_flags = 0;
David du Colombierd5f43282011-03-17 10:40:16 +0100818 hints.ai_protocol = 0;
819
820 if (getaddrinfo(str, NULL, &hints, &result) == 0) {
Willy Tarreau24709282013-03-10 21:32:12 +0100821 if (!sa->ss_family || sa->ss_family == AF_UNSPEC)
822 sa->ss_family = result->ai_family;
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100823 else if (sa->ss_family != result->ai_family) {
824 freeaddrinfo(result);
Willy Tarreau24709282013-03-10 21:32:12 +0100825 goto fail;
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100826 }
Willy Tarreau24709282013-03-10 21:32:12 +0100827
David du Colombierd5f43282011-03-17 10:40:16 +0100828 switch (result->ai_family) {
829 case AF_INET:
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100830 memcpy((struct sockaddr_in *)sa, result->ai_addr, result->ai_addrlen);
Willy Tarreauecde7df2016-11-02 22:37:03 +0100831 set_host_port(sa, port);
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100832 success = 1;
833 break;
David du Colombierd5f43282011-03-17 10:40:16 +0100834 case AF_INET6:
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100835 memcpy((struct sockaddr_in6 *)sa, result->ai_addr, result->ai_addrlen);
Willy Tarreauecde7df2016-11-02 22:37:03 +0100836 set_host_port(sa, port);
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100837 success = 1;
838 break;
David du Colombierd5f43282011-03-17 10:40:16 +0100839 }
840 }
841
Sean Carey58ea0392013-02-15 23:39:18 +0100842 if (result)
843 freeaddrinfo(result);
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100844
845 if (success)
846 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100847 }
David du Colombierd5f43282011-03-17 10:40:16 +0100848#endif
Nenad Merdanovic88afe032014-04-14 15:56:58 +0200849 /* try to resolve an IPv4/IPv6 hostname */
850 he = gethostbyname(str);
851 if (he) {
852 if (!sa->ss_family || sa->ss_family == AF_UNSPEC)
853 sa->ss_family = he->h_addrtype;
854 else if (sa->ss_family != he->h_addrtype)
855 goto fail;
856
857 switch (sa->ss_family) {
858 case AF_INET:
859 ((struct sockaddr_in *)sa)->sin_addr = *(struct in_addr *) *(he->h_addr_list);
Willy Tarreauecde7df2016-11-02 22:37:03 +0100860 set_host_port(sa, port);
Nenad Merdanovic88afe032014-04-14 15:56:58 +0200861 return sa;
862 case AF_INET6:
863 ((struct sockaddr_in6 *)sa)->sin6_addr = *(struct in6_addr *) *(he->h_addr_list);
Willy Tarreauecde7df2016-11-02 22:37:03 +0100864 set_host_port(sa, port);
Nenad Merdanovic88afe032014-04-14 15:56:58 +0200865 return sa;
866 }
867 }
868
David du Colombierd5f43282011-03-17 10:40:16 +0100869 /* unsupported address family */
Willy Tarreau24709282013-03-10 21:32:12 +0100870 fail:
Willy Tarreaufab5a432011-03-04 15:31:53 +0100871 return NULL;
872}
873
874/*
Willy Tarreaud4448bc2013-02-20 15:55:15 +0100875 * Converts <str> to a locally allocated struct sockaddr_storage *, and a port
876 * range or offset consisting in two integers that the caller will have to
877 * check to find the relevant input format. The following format are supported :
878 *
879 * String format | address | port | low | high
880 * addr | <addr> | 0 | 0 | 0
881 * addr: | <addr> | 0 | 0 | 0
882 * addr:port | <addr> | <port> | <port> | <port>
883 * addr:pl-ph | <addr> | <pl> | <pl> | <ph>
884 * addr:+port | <addr> | <port> | 0 | <port>
885 * addr:-port | <addr> |-<port> | <port> | 0
886 *
887 * The detection of a port range or increment by the caller is made by
888 * comparing <low> and <high>. If both are equal, then port 0 means no port
889 * was specified. The caller may pass NULL for <low> and <high> if it is not
890 * interested in retrieving port ranges.
891 *
892 * Note that <addr> above may also be :
893 * - empty ("") => family will be AF_INET and address will be INADDR_ANY
894 * - "*" => family will be AF_INET and address will be INADDR_ANY
895 * - "::" => family will be AF_INET6 and address will be IN6ADDR_ANY
896 * - a host name => family and address will depend on host name resolving.
897 *
Willy Tarreau24709282013-03-10 21:32:12 +0100898 * A prefix may be passed in before the address above to force the family :
899 * - "ipv4@" => force address to resolve as IPv4 and fail if not possible.
900 * - "ipv6@" => force address to resolve as IPv6 and fail if not possible.
901 * - "unix@" => force address to be a path to a UNIX socket even if the
902 * path does not start with a '/'
Willy Tarreauccfccef2014-05-10 01:49:15 +0200903 * - 'abns@' -> force address to belong to the abstract namespace (Linux
904 * only). These sockets are just like Unix sockets but without
905 * the need for an underlying file system. The address is a
906 * string. Technically it's like a Unix socket with a zero in
907 * the first byte of the address.
Willy Tarreau40aa0702013-03-10 23:51:38 +0100908 * - "fd@" => an integer must follow, and is a file descriptor number.
Willy Tarreau24709282013-03-10 21:32:12 +0100909 *
mildisff5d5102015-10-26 18:50:08 +0100910 * IPv6 addresses can be declared with or without square brackets. When using
911 * square brackets for IPv6 addresses, the port separator (colon) is optional.
912 * If not using square brackets, and in order to avoid any ambiguity with
913 * IPv6 addresses, the last colon ':' is mandatory even when no port is specified.
914 * NULL is returned if the address cannot be parsed. The <low> and <high> ports
915 * are always initialized if non-null, even for non-IP families.
Willy Tarreaud393a622013-03-04 18:22:00 +0100916 *
917 * If <pfx> is non-null, it is used as a string prefix before any path-based
918 * address (typically the path to a unix socket).
Willy Tarreau40aa0702013-03-10 23:51:38 +0100919 *
Willy Tarreau72b8c1f2015-09-08 15:50:19 +0200920 * if <fqdn> is non-null, it will be filled with :
921 * - a pointer to the FQDN of the server name to resolve if there's one, and
922 * that the caller will have to free(),
923 * - NULL if there was an explicit address that doesn't require resolution.
924 *
Willy Tarreaucd3a55912020-09-04 15:30:46 +0200925 * Hostnames are only resolved if <opts> has PA_O_RESOLVE. Otherwise <fqdn> is
926 * still honored so it is possible for the caller to know whether a resolution
927 * failed by clearing this flag and checking if <fqdn> was filled, indicating
928 * the need for a resolution.
Thierry FOURNIER7fe3be72015-09-26 20:03:36 +0200929 *
Willy Tarreau40aa0702013-03-10 23:51:38 +0100930 * When a file descriptor is passed, its value is put into the s_addr part of
Willy Tarreaua5b325f2020-09-04 16:44:20 +0200931 * the address when cast to sockaddr_in and the address family is
932 * AF_CUST_EXISTING_FD.
Willy Tarreaua93e5c72020-09-15 14:01:16 +0200933 *
Willy Tarreau5fc93282020-09-16 18:25:03 +0200934 * The matching protocol will be set into <proto> if non-null.
935 *
Willy Tarreaua93e5c72020-09-15 14:01:16 +0200936 * Any known file descriptor is also assigned to <fd> if non-null, otherwise it
937 * is forced to -1.
Willy Tarreaufab5a432011-03-04 15:31:53 +0100938 */
Willy Tarreau5fc93282020-09-16 18:25:03 +0200939struct sockaddr_storage *str2sa_range(const char *str, int *port, int *low, int *high, int *fd,
940 struct protocol **proto, char **err,
941 const char *pfx, char **fqdn, unsigned int opts)
Willy Tarreaufab5a432011-03-04 15:31:53 +0100942{
Christopher Faulet1bc04c72017-10-29 20:14:08 +0100943 static THREAD_LOCAL struct sockaddr_storage ss;
David du Colombier6f5ccb12011-03-10 22:26:24 +0100944 struct sockaddr_storage *ret = NULL;
Willy Tarreau5fc93282020-09-16 18:25:03 +0200945 struct protocol *new_proto = NULL;
Willy Tarreau24709282013-03-10 21:32:12 +0100946 char *back, *str2;
Willy Tarreaud4448bc2013-02-20 15:55:15 +0100947 char *port1, *port2;
948 int portl, porth, porta;
Willy Tarreauccfccef2014-05-10 01:49:15 +0200949 int abstract = 0;
Willy Tarreaua93e5c72020-09-15 14:01:16 +0200950 int new_fd = -1;
Willy Tarreaue3b45182021-10-27 17:28:55 +0200951 enum proto_type proto_type;
952 int ctrl_type;
Willy Tarreaud4448bc2013-02-20 15:55:15 +0100953
954 portl = porth = porta = 0;
Willy Tarreau72b8c1f2015-09-08 15:50:19 +0200955 if (fqdn)
956 *fqdn = NULL;
Willy Tarreaubaaee002006-06-26 02:48:02 +0200957
Willy Tarreaudad36a32013-03-11 01:20:04 +0100958 str2 = back = env_expand(strdup(str));
Willy Tarreaudf350f12013-03-01 20:22:54 +0100959 if (str2 == NULL) {
960 memprintf(err, "out of memory in '%s'\n", __FUNCTION__);
Willy Tarreaud5191e72010-02-09 20:50:45 +0100961 goto out;
Willy Tarreaudf350f12013-03-01 20:22:54 +0100962 }
Willy Tarreaubaaee002006-06-26 02:48:02 +0200963
Willy Tarreau9f69f462015-09-08 16:01:25 +0200964 if (!*str2) {
965 memprintf(err, "'%s' resolves to an empty address (environment variable missing?)\n", str);
966 goto out;
967 }
968
Willy Tarreau24709282013-03-10 21:32:12 +0100969 memset(&ss, 0, sizeof(ss));
970
Willy Tarreaue835bd82020-09-16 11:35:47 +0200971 /* prepare the default socket types */
Willy Tarreauf23b1bc2021-03-23 18:36:37 +0100972 if ((opts & (PA_O_STREAM|PA_O_DGRAM)) == PA_O_DGRAM ||
Willy Tarreaue3b45182021-10-27 17:28:55 +0200973 ((opts & (PA_O_STREAM|PA_O_DGRAM)) == (PA_O_DGRAM|PA_O_STREAM) && (opts & PA_O_DEFAULT_DGRAM))) {
974 proto_type = PROTO_TYPE_DGRAM;
975 ctrl_type = SOCK_DGRAM;
976 } else {
977 proto_type = PROTO_TYPE_STREAM;
978 ctrl_type = SOCK_STREAM;
979 }
Willy Tarreaue835bd82020-09-16 11:35:47 +0200980
981 if (strncmp(str2, "stream+", 7) == 0) {
982 str2 += 7;
Willy Tarreaue3b45182021-10-27 17:28:55 +0200983 proto_type = PROTO_TYPE_STREAM;
984 ctrl_type = SOCK_STREAM;
Willy Tarreaue835bd82020-09-16 11:35:47 +0200985 }
986 else if (strncmp(str2, "dgram+", 6) == 0) {
987 str2 += 6;
Willy Tarreaue3b45182021-10-27 17:28:55 +0200988 proto_type = PROTO_TYPE_DGRAM;
989 ctrl_type = SOCK_DGRAM;
Willy Tarreaue835bd82020-09-16 11:35:47 +0200990 }
991
Willy Tarreau24709282013-03-10 21:32:12 +0100992 if (strncmp(str2, "unix@", 5) == 0) {
993 str2 += 5;
Willy Tarreauccfccef2014-05-10 01:49:15 +0200994 abstract = 0;
Willy Tarreau24709282013-03-10 21:32:12 +0100995 ss.ss_family = AF_UNIX;
996 }
Emeric Brunce325c42021-04-02 17:05:09 +0200997 else if (strncmp(str2, "uxdg@", 5) == 0) {
998 str2 += 5;
999 abstract = 0;
1000 ss.ss_family = AF_UNIX;
Willy Tarreaue3b45182021-10-27 17:28:55 +02001001 proto_type = PROTO_TYPE_DGRAM;
1002 ctrl_type = SOCK_DGRAM;
Emeric Brunce325c42021-04-02 17:05:09 +02001003 }
1004 else if (strncmp(str2, "uxst@", 5) == 0) {
1005 str2 += 5;
1006 abstract = 0;
1007 ss.ss_family = AF_UNIX;
Willy Tarreaue3b45182021-10-27 17:28:55 +02001008 proto_type = PROTO_TYPE_STREAM;
1009 ctrl_type = SOCK_STREAM;
Emeric Brunce325c42021-04-02 17:05:09 +02001010 }
Willy Tarreauccfccef2014-05-10 01:49:15 +02001011 else if (strncmp(str2, "abns@", 5) == 0) {
1012 str2 += 5;
1013 abstract = 1;
1014 ss.ss_family = AF_UNIX;
1015 }
Emeric Brunce325c42021-04-02 17:05:09 +02001016 else if (strncmp(str2, "ip@", 3) == 0) {
1017 str2 += 3;
1018 ss.ss_family = AF_UNSPEC;
1019 }
Willy Tarreau24709282013-03-10 21:32:12 +01001020 else if (strncmp(str2, "ipv4@", 5) == 0) {
1021 str2 += 5;
1022 ss.ss_family = AF_INET;
1023 }
1024 else if (strncmp(str2, "ipv6@", 5) == 0) {
1025 str2 += 5;
1026 ss.ss_family = AF_INET6;
1027 }
Emeric Brunce325c42021-04-02 17:05:09 +02001028 else if (strncmp(str2, "tcp4@", 5) == 0) {
1029 str2 += 5;
1030 ss.ss_family = AF_INET;
Willy Tarreaue3b45182021-10-27 17:28:55 +02001031 proto_type = PROTO_TYPE_STREAM;
1032 ctrl_type = SOCK_STREAM;
Emeric Brunce325c42021-04-02 17:05:09 +02001033 }
Emeric Brun3835c0d2020-07-07 09:46:09 +02001034 else if (strncmp(str2, "udp4@", 5) == 0) {
1035 str2 += 5;
1036 ss.ss_family = AF_INET;
Willy Tarreaue3b45182021-10-27 17:28:55 +02001037 proto_type = PROTO_TYPE_DGRAM;
1038 ctrl_type = SOCK_DGRAM;
Emeric Brun3835c0d2020-07-07 09:46:09 +02001039 }
Emeric Brunce325c42021-04-02 17:05:09 +02001040 else if (strncmp(str2, "tcp6@", 5) == 0) {
1041 str2 += 5;
1042 ss.ss_family = AF_INET6;
Willy Tarreaue3b45182021-10-27 17:28:55 +02001043 proto_type = PROTO_TYPE_STREAM;
1044 ctrl_type = SOCK_STREAM;
Emeric Brunce325c42021-04-02 17:05:09 +02001045 }
Emeric Brun3835c0d2020-07-07 09:46:09 +02001046 else if (strncmp(str2, "udp6@", 5) == 0) {
1047 str2 += 5;
1048 ss.ss_family = AF_INET6;
Willy Tarreaue3b45182021-10-27 17:28:55 +02001049 proto_type = PROTO_TYPE_DGRAM;
1050 ctrl_type = SOCK_DGRAM;
Emeric Brun3835c0d2020-07-07 09:46:09 +02001051 }
Emeric Brunce325c42021-04-02 17:05:09 +02001052 else if (strncmp(str2, "tcp@", 4) == 0) {
1053 str2 += 4;
1054 ss.ss_family = AF_UNSPEC;
Willy Tarreaue3b45182021-10-27 17:28:55 +02001055 proto_type = PROTO_TYPE_STREAM;
1056 ctrl_type = SOCK_STREAM;
Emeric Brunce325c42021-04-02 17:05:09 +02001057 }
Emeric Brun3835c0d2020-07-07 09:46:09 +02001058 else if (strncmp(str2, "udp@", 4) == 0) {
1059 str2 += 4;
1060 ss.ss_family = AF_UNSPEC;
Willy Tarreaue3b45182021-10-27 17:28:55 +02001061 proto_type = PROTO_TYPE_DGRAM;
1062 ctrl_type = SOCK_DGRAM;
Emeric Brun3835c0d2020-07-07 09:46:09 +02001063 }
Frédéric Lécaille10caf652020-11-23 11:36:57 +01001064 else if (strncmp(str2, "quic4@", 6) == 0) {
1065 str2 += 6;
1066 ss.ss_family = AF_INET;
Willy Tarreaue3b45182021-10-27 17:28:55 +02001067 proto_type = PROTO_TYPE_DGRAM;
Frédéric Lécaille10caf652020-11-23 11:36:57 +01001068 ctrl_type = SOCK_STREAM;
1069 }
1070 else if (strncmp(str2, "quic6@", 6) == 0) {
1071 str2 += 6;
1072 ss.ss_family = AF_INET6;
Willy Tarreaue3b45182021-10-27 17:28:55 +02001073 proto_type = PROTO_TYPE_DGRAM;
Frédéric Lécaille10caf652020-11-23 11:36:57 +01001074 ctrl_type = SOCK_STREAM;
1075 }
Willy Tarreau5a7beed2020-09-04 16:54:05 +02001076 else if (strncmp(str2, "fd@", 3) == 0) {
1077 str2 += 3;
1078 ss.ss_family = AF_CUST_EXISTING_FD;
1079 }
1080 else if (strncmp(str2, "sockpair@", 9) == 0) {
1081 str2 += 9;
1082 ss.ss_family = AF_CUST_SOCKPAIR;
1083 }
Willy Tarreau24709282013-03-10 21:32:12 +01001084 else if (*str2 == '/') {
1085 ss.ss_family = AF_UNIX;
1086 }
1087 else
1088 ss.ss_family = AF_UNSPEC;
1089
Willy Tarreau5a7beed2020-09-04 16:54:05 +02001090 if (ss.ss_family == AF_CUST_SOCKPAIR) {
Willy Tarreaua215be22020-09-16 10:14:16 +02001091 struct sockaddr_storage ss2;
1092 socklen_t addr_len;
William Lallemand2fe7dd02018-09-11 16:51:29 +02001093 char *endptr;
1094
Willy Tarreaua93e5c72020-09-15 14:01:16 +02001095 new_fd = strtol(str2, &endptr, 10);
1096 if (!*str2 || new_fd < 0 || *endptr) {
William Lallemand2fe7dd02018-09-11 16:51:29 +02001097 memprintf(err, "file descriptor '%s' is not a valid integer in '%s'\n", str2, str);
1098 goto out;
1099 }
Willy Tarreaua93e5c72020-09-15 14:01:16 +02001100
Willy Tarreaua215be22020-09-16 10:14:16 +02001101 /* just verify that it's a socket */
1102 addr_len = sizeof(ss2);
1103 if (getsockname(new_fd, (struct sockaddr *)&ss2, &addr_len) == -1) {
1104 memprintf(err, "cannot use file descriptor '%d' : %s.\n", new_fd, strerror(errno));
1105 goto out;
1106 }
1107
Willy Tarreaua93e5c72020-09-15 14:01:16 +02001108 ((struct sockaddr_in *)&ss)->sin_addr.s_addr = new_fd;
1109 ((struct sockaddr_in *)&ss)->sin_port = 0;
William Lallemand2fe7dd02018-09-11 16:51:29 +02001110 }
Willy Tarreau5a7beed2020-09-04 16:54:05 +02001111 else if (ss.ss_family == AF_CUST_EXISTING_FD) {
Willy Tarreau40aa0702013-03-10 23:51:38 +01001112 char *endptr;
1113
Willy Tarreaua93e5c72020-09-15 14:01:16 +02001114 new_fd = strtol(str2, &endptr, 10);
1115 if (!*str2 || new_fd < 0 || *endptr) {
Willy Tarreaudad36a32013-03-11 01:20:04 +01001116 memprintf(err, "file descriptor '%s' is not a valid integer in '%s'\n", str2, str);
Willy Tarreau40aa0702013-03-10 23:51:38 +01001117 goto out;
1118 }
Willy Tarreaua93e5c72020-09-15 14:01:16 +02001119
Willy Tarreau6edc7222020-09-15 17:41:56 +02001120 if (opts & PA_O_SOCKET_FD) {
1121 socklen_t addr_len;
1122 int type;
1123
1124 addr_len = sizeof(ss);
1125 if (getsockname(new_fd, (struct sockaddr *)&ss, &addr_len) == -1) {
1126 memprintf(err, "cannot use file descriptor '%d' : %s.\n", new_fd, strerror(errno));
1127 goto out;
1128 }
1129
1130 addr_len = sizeof(type);
1131 if (getsockopt(new_fd, SOL_SOCKET, SO_TYPE, &type, &addr_len) != 0 ||
Willy Tarreaue3b45182021-10-27 17:28:55 +02001132 (type == SOCK_STREAM) != (proto_type == PROTO_TYPE_STREAM)) {
Willy Tarreau6edc7222020-09-15 17:41:56 +02001133 memprintf(err, "socket on file descriptor '%d' is of the wrong type.\n", new_fd);
1134 goto out;
1135 }
1136
1137 porta = portl = porth = get_host_port(&ss);
1138 } else if (opts & PA_O_RAW_FD) {
1139 ((struct sockaddr_in *)&ss)->sin_addr.s_addr = new_fd;
1140 ((struct sockaddr_in *)&ss)->sin_port = 0;
1141 } else {
1142 memprintf(err, "a file descriptor is not acceptable here in '%s'\n", str);
1143 goto out;
1144 }
Willy Tarreau40aa0702013-03-10 23:51:38 +01001145 }
1146 else if (ss.ss_family == AF_UNIX) {
Willy Tarreau8daa9202019-06-16 18:16:33 +02001147 struct sockaddr_un *un = (struct sockaddr_un *)&ss;
Willy Tarreau15586382013-03-04 19:48:14 +01001148 int prefix_path_len;
1149 int max_path_len;
Willy Tarreau94ef3f32014-04-14 14:49:00 +02001150 int adr_len;
Willy Tarreau15586382013-03-04 19:48:14 +01001151
1152 /* complete unix socket path name during startup or soft-restart is
1153 * <unix_bind_prefix><path>.<pid>.<bak|tmp>
1154 */
Willy Tarreauccfccef2014-05-10 01:49:15 +02001155 prefix_path_len = (pfx && !abstract) ? strlen(pfx) : 0;
Willy Tarreau8daa9202019-06-16 18:16:33 +02001156 max_path_len = (sizeof(un->sun_path) - 1) -
Willy Tarreau327ea5a2020-02-11 06:43:37 +01001157 (abstract ? 0 : prefix_path_len + 1 + 5 + 1 + 3);
Willy Tarreau15586382013-03-04 19:48:14 +01001158
Willy Tarreau94ef3f32014-04-14 14:49:00 +02001159 adr_len = strlen(str2);
1160 if (adr_len > max_path_len) {
Willy Tarreau15586382013-03-04 19:48:14 +01001161 memprintf(err, "socket path '%s' too long (max %d)\n", str, max_path_len);
1162 goto out;
1163 }
1164
Willy Tarreauccfccef2014-05-10 01:49:15 +02001165 /* when abstract==1, we skip the first zero and copy all bytes except the trailing zero */
Willy Tarreau8daa9202019-06-16 18:16:33 +02001166 memset(un->sun_path, 0, sizeof(un->sun_path));
Willy Tarreau94ef3f32014-04-14 14:49:00 +02001167 if (prefix_path_len)
Willy Tarreau8daa9202019-06-16 18:16:33 +02001168 memcpy(un->sun_path, pfx, prefix_path_len);
1169 memcpy(un->sun_path + prefix_path_len + abstract, str2, adr_len + 1 - abstract);
Willy Tarreau15586382013-03-04 19:48:14 +01001170 }
Willy Tarreau24709282013-03-10 21:32:12 +01001171 else { /* IPv4 and IPv6 */
mildisff5d5102015-10-26 18:50:08 +01001172 char *end = str2 + strlen(str2);
1173 char *chr;
Willy Tarreau72b8c1f2015-09-08 15:50:19 +02001174
mildisff5d5102015-10-26 18:50:08 +01001175 /* search for : or ] whatever comes first */
1176 for (chr = end-1; chr > str2; chr--) {
1177 if (*chr == ']' || *chr == ':')
1178 break;
1179 }
1180
1181 if (*chr == ':') {
1182 /* Found a colon before a closing-bracket, must be a port separator.
1183 * This guarantee backward compatibility.
1184 */
Willy Tarreau7f96a842020-09-15 11:12:44 +02001185 if (!(opts & PA_O_PORT_OK)) {
1186 memprintf(err, "port specification not permitted here in '%s'", str);
1187 goto out;
1188 }
mildisff5d5102015-10-26 18:50:08 +01001189 *chr++ = '\0';
1190 port1 = chr;
1191 }
1192 else {
1193 /* Either no colon and no closing-bracket
1194 * or directly ending with a closing-bracket.
1195 * However, no port.
1196 */
Willy Tarreau7f96a842020-09-15 11:12:44 +02001197 if (opts & PA_O_PORT_MAND) {
1198 memprintf(err, "missing port specification in '%s'", str);
1199 goto out;
1200 }
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001201 port1 = "";
mildisff5d5102015-10-26 18:50:08 +01001202 }
Willy Tarreaubaaee002006-06-26 02:48:02 +02001203
Willy Tarreau90807112020-02-25 08:16:33 +01001204 if (isdigit((unsigned char)*port1)) { /* single port or range */
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001205 port2 = strchr(port1, '-');
Willy Tarreau7f96a842020-09-15 11:12:44 +02001206 if (port2) {
1207 if (!(opts & PA_O_PORT_RANGE)) {
1208 memprintf(err, "port range not permitted here in '%s'", str);
1209 goto out;
1210 }
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001211 *port2++ = '\0';
Willy Tarreau7f96a842020-09-15 11:12:44 +02001212 }
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001213 else
1214 port2 = port1;
1215 portl = atoi(port1);
1216 porth = atoi(port2);
Willy Tarreau7f96a842020-09-15 11:12:44 +02001217
1218 if (portl < !!(opts & PA_O_PORT_MAND) || portl > 65535) {
1219 memprintf(err, "invalid port '%s'", port1);
1220 goto out;
1221 }
1222
1223 if (porth < !!(opts & PA_O_PORT_MAND) || porth > 65535) {
1224 memprintf(err, "invalid port '%s'", port2);
1225 goto out;
1226 }
1227
1228 if (portl > porth) {
1229 memprintf(err, "invalid port range '%d-%d'", portl, porth);
1230 goto out;
1231 }
1232
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001233 porta = portl;
1234 }
1235 else if (*port1 == '-') { /* negative offset */
Willy Tarreau7f96a842020-09-15 11:12:44 +02001236 if (!(opts & PA_O_PORT_OFS)) {
1237 memprintf(err, "port offset not permitted here in '%s'", str);
1238 goto out;
1239 }
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001240 portl = atoi(port1 + 1);
1241 porta = -portl;
1242 }
1243 else if (*port1 == '+') { /* positive offset */
Willy Tarreau7f96a842020-09-15 11:12:44 +02001244 if (!(opts & PA_O_PORT_OFS)) {
1245 memprintf(err, "port offset not permitted here in '%s'", str);
1246 goto out;
1247 }
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001248 porth = atoi(port1 + 1);
1249 porta = porth;
1250 }
1251 else if (*port1) { /* other any unexpected char */
Willy Tarreaudad36a32013-03-11 01:20:04 +01001252 memprintf(err, "invalid character '%c' in port number '%s' in '%s'\n", *port1, port1, str);
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001253 goto out;
1254 }
Willy Tarreau7f96a842020-09-15 11:12:44 +02001255 else if (opts & PA_O_PORT_MAND) {
1256 memprintf(err, "missing port specification in '%s'", str);
1257 goto out;
1258 }
Willy Tarreauceccdd72016-11-02 22:27:10 +01001259
1260 /* first try to parse the IP without resolving. If it fails, it
1261 * tells us we need to keep a copy of the FQDN to resolve later
1262 * and to enable DNS. In this case we can proceed if <fqdn> is
Willy Tarreaucd3a55912020-09-04 15:30:46 +02001263 * set or if PA_O_RESOLVE is set, otherwise it's an error.
Willy Tarreauceccdd72016-11-02 22:27:10 +01001264 */
1265 if (str2ip2(str2, &ss, 0) == NULL) {
Willy Tarreaucd3a55912020-09-04 15:30:46 +02001266 if ((!(opts & PA_O_RESOLVE) && !fqdn) ||
1267 ((opts & PA_O_RESOLVE) && str2ip2(str2, &ss, 1) == NULL)) {
Willy Tarreauceccdd72016-11-02 22:27:10 +01001268 memprintf(err, "invalid address: '%s' in '%s'\n", str2, str);
1269 goto out;
1270 }
Willy Tarreau72b8c1f2015-09-08 15:50:19 +02001271
Willy Tarreauceccdd72016-11-02 22:27:10 +01001272 if (fqdn) {
1273 if (str2 != back)
1274 memmove(back, str2, strlen(str2) + 1);
1275 *fqdn = back;
1276 back = NULL;
1277 }
Willy Tarreau72b8c1f2015-09-08 15:50:19 +02001278 }
Willy Tarreauceccdd72016-11-02 22:27:10 +01001279 set_host_port(&ss, porta);
Willy Tarreaue4c58c82013-03-06 15:28:17 +01001280 }
Willy Tarreaufab5a432011-03-04 15:31:53 +01001281
Willy Tarreaue835bd82020-09-16 11:35:47 +02001282 if (ctrl_type == SOCK_STREAM && !(opts & PA_O_STREAM)) {
Willy Tarreau8d31ab02022-05-09 16:18:26 +02001283 memprintf(err, "stream-type address not acceptable in '%s'\n", str);
Willy Tarreaue835bd82020-09-16 11:35:47 +02001284 goto out;
1285 }
1286 else if (ctrl_type == SOCK_DGRAM && !(opts & PA_O_DGRAM)) {
Willy Tarreau8d31ab02022-05-09 16:18:26 +02001287 memprintf(err, "dgram-type address not acceptable in '%s'\n", str);
Willy Tarreaue835bd82020-09-16 11:35:47 +02001288 goto out;
1289 }
1290
Willy Tarreau65ec4e32020-09-16 19:17:08 +02001291 if (proto || (opts & PA_O_CONNECT)) {
Willy Tarreau5fc93282020-09-16 18:25:03 +02001292 /* Note: if the caller asks for a proto, we must find one,
Emeric Brun26754902021-04-07 14:26:44 +02001293 * except if we inherit from a raw FD (family == AF_CUST_EXISTING_FD)
1294 * orif we return with an fqdn that will resolve later,
Willy Tarreau5fc93282020-09-16 18:25:03 +02001295 * in which case the address is not known yet (this is only
1296 * for servers actually).
1297 */
Willy Tarreaub2ffc992020-09-16 21:37:31 +02001298 new_proto = protocol_lookup(ss.ss_family,
Willy Tarreaue3b45182021-10-27 17:28:55 +02001299 proto_type,
Willy Tarreauaf9609b2020-09-16 22:04:46 +02001300 ctrl_type == SOCK_DGRAM);
Willy Tarreaub2ffc992020-09-16 21:37:31 +02001301
Emeric Brun26754902021-04-07 14:26:44 +02001302 if (!new_proto && (!fqdn || !*fqdn) && (ss.ss_family != AF_CUST_EXISTING_FD)) {
Willy Tarreau2b049b82022-05-20 17:28:30 +02001303 memprintf(err, "unsupported %s protocol for %s family %d address '%s'",
1304 (ctrl_type == SOCK_DGRAM) ? "datagram" : "stream",
1305 (proto_type == PROTO_TYPE_DGRAM) ? "datagram" : "stream",
1306 ss.ss_family,
1307 str);
Willy Tarreau5fc93282020-09-16 18:25:03 +02001308 goto out;
1309 }
Willy Tarreau65ec4e32020-09-16 19:17:08 +02001310
1311 if ((opts & PA_O_CONNECT) && new_proto && !new_proto->connect) {
1312 memprintf(err, "connect() not supported for this protocol family %d used by address '%s'", ss.ss_family, str);
1313 goto out;
1314 }
Willy Tarreau5fc93282020-09-16 18:25:03 +02001315 }
1316
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001317 ret = &ss;
Willy Tarreaud5191e72010-02-09 20:50:45 +01001318 out:
Willy Tarreau48ef4c92017-01-06 18:32:38 +01001319 if (port)
1320 *port = porta;
Willy Tarreaud4448bc2013-02-20 15:55:15 +01001321 if (low)
1322 *low = portl;
1323 if (high)
1324 *high = porth;
Willy Tarreaua93e5c72020-09-15 14:01:16 +02001325 if (fd)
1326 *fd = new_fd;
Willy Tarreau5fc93282020-09-16 18:25:03 +02001327 if (proto)
1328 *proto = new_proto;
Willy Tarreau24709282013-03-10 21:32:12 +01001329 free(back);
Willy Tarreaud5191e72010-02-09 20:50:45 +01001330 return ret;
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02001331}
1332
Thayne McCombs92149f92020-11-20 01:28:26 -07001333/* converts <addr> and <port> into a string representation of the address and port. This is sort
1334 * of an inverse of str2sa_range, with some restrictions. The supported families are AF_INET,
1335 * AF_INET6, AF_UNIX, and AF_CUST_SOCKPAIR. If the family is unsopported NULL is returned.
1336 * If map_ports is true, then the sign of the port is included in the output, to indicate it is
1337 * relative to the incoming port. AF_INET and AF_INET6 will be in the form "<addr>:<port>".
1338 * AF_UNIX will either be just the path (if using a pathname) or "abns@<path>" if it is abstract.
1339 * AF_CUST_SOCKPAIR will be of the form "sockpair@<fd>".
1340 *
1341 * The returned char* is allocated, and it is the responsibility of the caller to free it.
1342 */
1343char * sa2str(const struct sockaddr_storage *addr, int port, int map_ports)
1344{
1345 char buffer[INET6_ADDRSTRLEN];
1346 char *out = NULL;
1347 const void *ptr;
1348 const char *path;
1349
1350 switch (addr->ss_family) {
1351 case AF_INET:
1352 ptr = &((struct sockaddr_in *)addr)->sin_addr;
1353 break;
1354 case AF_INET6:
1355 ptr = &((struct sockaddr_in6 *)addr)->sin6_addr;
1356 break;
1357 case AF_UNIX:
1358 path = ((struct sockaddr_un *)addr)->sun_path;
1359 if (path[0] == '\0') {
1360 const int max_length = sizeof(struct sockaddr_un) - offsetof(struct sockaddr_un, sun_path) - 1;
1361 return memprintf(&out, "abns@%.*s", max_length, path+1);
1362 } else {
1363 return strdup(path);
1364 }
1365 case AF_CUST_SOCKPAIR:
1366 return memprintf(&out, "sockpair@%d", ((struct sockaddr_in *)addr)->sin_addr.s_addr);
1367 default:
1368 return NULL;
1369 }
1370 inet_ntop(addr->ss_family, ptr, buffer, get_addr_len(addr));
1371 if (map_ports)
1372 return memprintf(&out, "%s:%+d", buffer, port);
1373 else
1374 return memprintf(&out, "%s:%d", buffer, port);
1375}
1376
1377
Willy Tarreau2937c0d2010-01-26 17:36:17 +01001378/* converts <str> to a struct in_addr containing a network mask. It can be
1379 * passed in dotted form (255.255.255.0) or in CIDR form (24). It returns 1
Jarno Huuskonen577d5ac2017-05-21 17:32:21 +03001380 * if the conversion succeeds otherwise zero.
Willy Tarreau2937c0d2010-01-26 17:36:17 +01001381 */
1382int str2mask(const char *str, struct in_addr *mask)
1383{
1384 if (strchr(str, '.') != NULL) { /* dotted notation */
1385 if (!inet_pton(AF_INET, str, mask))
1386 return 0;
1387 }
1388 else { /* mask length */
1389 char *err;
1390 unsigned long len = strtol(str, &err, 10);
1391
1392 if (!*str || (err && *err) || (unsigned)len > 32)
1393 return 0;
Tim Duesterhus8575f722018-01-25 16:24:48 +01001394
1395 len2mask4(len, mask);
Willy Tarreau2937c0d2010-01-26 17:36:17 +01001396 }
1397 return 1;
1398}
1399
Tim Duesterhus47185172018-01-25 16:24:49 +01001400/* converts <str> to a struct in6_addr containing a network mask. It can be
Tim Duesterhus5e642862018-02-20 17:02:18 +01001401 * passed in quadruplet form (ffff:ffff::) or in CIDR form (64). It returns 1
Tim Duesterhus47185172018-01-25 16:24:49 +01001402 * if the conversion succeeds otherwise zero.
1403 */
1404int str2mask6(const char *str, struct in6_addr *mask)
1405{
1406 if (strchr(str, ':') != NULL) { /* quadruplet notation */
1407 if (!inet_pton(AF_INET6, str, mask))
1408 return 0;
1409 }
1410 else { /* mask length */
1411 char *err;
1412 unsigned long len = strtol(str, &err, 10);
1413
1414 if (!*str || (err && *err) || (unsigned)len > 128)
1415 return 0;
1416
1417 len2mask6(len, mask);
1418 }
1419 return 1;
1420}
1421
Thierry FOURNIERb0504632013-12-14 15:39:02 +01001422/* convert <cidr> to struct in_addr <mask>. It returns 1 if the conversion
1423 * succeeds otherwise zero.
1424 */
1425int cidr2dotted(int cidr, struct in_addr *mask) {
1426
1427 if (cidr < 0 || cidr > 32)
1428 return 0;
1429
1430 mask->s_addr = cidr ? htonl(~0UL << (32 - cidr)) : 0;
1431 return 1;
1432}
1433
Thierry Fournier70473a52016-02-17 17:12:14 +01001434/* Convert mask from bit length form to in_addr form.
1435 * This function never fails.
1436 */
1437void len2mask4(int len, struct in_addr *addr)
1438{
1439 if (len >= 32) {
1440 addr->s_addr = 0xffffffff;
1441 return;
1442 }
1443 if (len <= 0) {
1444 addr->s_addr = 0x00000000;
1445 return;
1446 }
1447 addr->s_addr = 0xffffffff << (32 - len);
1448 addr->s_addr = htonl(addr->s_addr);
1449}
1450
1451/* Convert mask from bit length form to in6_addr form.
1452 * This function never fails.
1453 */
1454void len2mask6(int len, struct in6_addr *addr)
1455{
1456 len2mask4(len, (struct in_addr *)&addr->s6_addr[0]); /* msb */
1457 len -= 32;
1458 len2mask4(len, (struct in_addr *)&addr->s6_addr[4]);
1459 len -= 32;
1460 len2mask4(len, (struct in_addr *)&addr->s6_addr[8]);
1461 len -= 32;
1462 len2mask4(len, (struct in_addr *)&addr->s6_addr[12]); /* lsb */
1463}
1464
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02001465/*
Willy Tarreaud077a8e2007-05-08 18:28:09 +02001466 * converts <str> to two struct in_addr* which must be pre-allocated.
Willy Tarreaubaaee002006-06-26 02:48:02 +02001467 * The format is "addr[/mask]", where "addr" cannot be empty, and mask
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05001468 * is optional and either in the dotted or CIDR notation.
Willy Tarreaubaaee002006-06-26 02:48:02 +02001469 * Note: "addr" can also be a hostname. Returns 1 if OK, 0 if error.
1470 */
Thierry FOURNIERfc7ac7b2014-02-11 15:23:04 +01001471int str2net(const char *str, int resolve, struct in_addr *addr, struct in_addr *mask)
Willy Tarreaubaaee002006-06-26 02:48:02 +02001472{
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001473 __label__ out_free, out_err;
1474 char *c, *s;
1475 int ret_val;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001476
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001477 s = strdup(str);
1478 if (!s)
1479 return 0;
1480
Willy Tarreaubaaee002006-06-26 02:48:02 +02001481 memset(mask, 0, sizeof(*mask));
1482 memset(addr, 0, sizeof(*addr));
Willy Tarreaubaaee002006-06-26 02:48:02 +02001483
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001484 if ((c = strrchr(s, '/')) != NULL) {
Willy Tarreaubaaee002006-06-26 02:48:02 +02001485 *c++ = '\0';
1486 /* c points to the mask */
Willy Tarreau2937c0d2010-01-26 17:36:17 +01001487 if (!str2mask(c, mask))
1488 goto out_err;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001489 }
1490 else {
Willy Tarreauebd61602006-12-30 11:54:15 +01001491 mask->s_addr = ~0U;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001492 }
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001493 if (!inet_pton(AF_INET, s, addr)) {
Willy Tarreaubaaee002006-06-26 02:48:02 +02001494 struct hostent *he;
1495
Thierry FOURNIERfc7ac7b2014-02-11 15:23:04 +01001496 if (!resolve)
1497 goto out_err;
1498
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001499 if ((he = gethostbyname(s)) == NULL) {
1500 goto out_err;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001501 }
1502 else
1503 *addr = *(struct in_addr *) *(he->h_addr_list);
1504 }
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001505
1506 ret_val = 1;
1507 out_free:
1508 free(s);
1509 return ret_val;
1510 out_err:
1511 ret_val = 0;
1512 goto out_free;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001513}
1514
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001515
1516/*
Willy Tarreau6d20e282012-04-27 22:49:47 +02001517 * converts <str> to two struct in6_addr* which must be pre-allocated.
1518 * The format is "addr[/mask]", where "addr" cannot be empty, and mask
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05001519 * is an optional number of bits (128 being the default).
Willy Tarreau6d20e282012-04-27 22:49:47 +02001520 * Returns 1 if OK, 0 if error.
1521 */
1522int str62net(const char *str, struct in6_addr *addr, unsigned char *mask)
1523{
1524 char *c, *s;
1525 int ret_val = 0;
1526 char *err;
1527 unsigned long len = 128;
1528
1529 s = strdup(str);
1530 if (!s)
1531 return 0;
1532
1533 memset(mask, 0, sizeof(*mask));
1534 memset(addr, 0, sizeof(*addr));
1535
1536 if ((c = strrchr(s, '/')) != NULL) {
1537 *c++ = '\0'; /* c points to the mask */
1538 if (!*c)
1539 goto out_free;
1540
1541 len = strtoul(c, &err, 10);
1542 if ((err && *err) || (unsigned)len > 128)
1543 goto out_free;
1544 }
1545 *mask = len; /* OK we have a valid mask in <len> */
1546
1547 if (!inet_pton(AF_INET6, s, addr))
1548 goto out_free;
1549
1550 ret_val = 1;
1551 out_free:
1552 free(s);
1553 return ret_val;
1554}
1555
1556
1557/*
Willy Tarreau12e10272021-03-25 11:34:40 +01001558 * Parse IPv4 address found in url. Return the number of bytes parsed. It
1559 * expects exactly 4 numbers between 0 and 255 delimited by dots, and returns
1560 * zero in case of mismatch.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001561 */
David du Colombier6f5ccb12011-03-10 22:26:24 +01001562int url2ipv4(const char *addr, struct in_addr *dst)
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001563{
1564 int saw_digit, octets, ch;
1565 u_char tmp[4], *tp;
1566 const char *cp = addr;
1567
1568 saw_digit = 0;
1569 octets = 0;
1570 *(tp = tmp) = 0;
1571
1572 while (*addr) {
Willy Tarreau12e10272021-03-25 11:34:40 +01001573 unsigned char digit = (ch = *addr) - '0';
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001574 if (digit > 9 && ch != '.')
1575 break;
Willy Tarreau12e10272021-03-25 11:34:40 +01001576 addr++;
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001577 if (digit <= 9) {
1578 u_int new = *tp * 10 + digit;
1579 if (new > 255)
1580 return 0;
1581 *tp = new;
1582 if (!saw_digit) {
1583 if (++octets > 4)
1584 return 0;
1585 saw_digit = 1;
1586 }
1587 } else if (ch == '.' && saw_digit) {
1588 if (octets == 4)
1589 return 0;
1590 *++tp = 0;
1591 saw_digit = 0;
1592 } else
1593 return 0;
1594 }
1595
1596 if (octets < 4)
1597 return 0;
1598
1599 memcpy(&dst->s_addr, tmp, 4);
Willy Tarreau12e10272021-03-25 11:34:40 +01001600 return addr - cp;
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001601}
1602
1603/*
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001604 * Resolve destination server from URL. Convert <str> to a sockaddr_storage.
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05001605 * <out> contain the code of the detected scheme, the start and length of
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001606 * the hostname. Actually only http and https are supported. <out> can be NULL.
1607 * This function returns the consumed length. It is useful if you parse complete
1608 * url like http://host:port/path, because the consumed length corresponds to
1609 * the first character of the path. If the conversion fails, it returns -1.
1610 *
1611 * This function tries to resolve the DNS name if haproxy is in starting mode.
1612 * So, this function may be used during the configuration parsing.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001613 */
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001614int url2sa(const char *url, int ulen, struct sockaddr_storage *addr, struct split_url *out)
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001615{
1616 const char *curr = url, *cp = url;
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001617 const char *end;
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001618 int ret, url_code = 0;
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001619 unsigned long long int http_code = 0;
1620 int default_port;
1621 struct hostent *he;
1622 char *p;
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001623
1624 /* Firstly, try to find :// pattern */
1625 while (curr < url+ulen && url_code != 0x3a2f2f) {
1626 url_code = ((url_code & 0xffff) << 8);
1627 url_code += (unsigned char)*curr++;
1628 }
1629
1630 /* Secondly, if :// pattern is found, verify parsed stuff
1631 * before pattern is matching our http pattern.
1632 * If so parse ip address and port in uri.
1633 *
1634 * WARNING: Current code doesn't support dynamic async dns resolver.
1635 */
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001636 if (url_code != 0x3a2f2f)
1637 return -1;
1638
1639 /* Copy scheme, and utrn to lower case. */
1640 while (cp < curr - 3)
1641 http_code = (http_code << 8) + *cp++;
1642 http_code |= 0x2020202020202020ULL; /* Turn everything to lower case */
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001643
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001644 /* HTTP or HTTPS url matching */
1645 if (http_code == 0x2020202068747470ULL) {
1646 default_port = 80;
1647 if (out)
1648 out->scheme = SCH_HTTP;
1649 }
1650 else if (http_code == 0x2020206874747073ULL) {
1651 default_port = 443;
1652 if (out)
1653 out->scheme = SCH_HTTPS;
1654 }
1655 else
1656 return -1;
1657
1658 /* If the next char is '[', the host address is IPv6. */
1659 if (*curr == '[') {
1660 curr++;
1661
1662 /* Check trash size */
1663 if (trash.size < ulen)
1664 return -1;
1665
1666 /* Look for ']' and copy the address in a trash buffer. */
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001667 p = trash.area;
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001668 for (end = curr;
1669 end < url + ulen && *end != ']';
1670 end++, p++)
1671 *p = *end;
1672 if (*end != ']')
1673 return -1;
1674 *p = '\0';
1675
1676 /* Update out. */
1677 if (out) {
1678 out->host = curr;
1679 out->host_len = end - curr;
1680 }
1681
1682 /* Try IPv6 decoding. */
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001683 if (!inet_pton(AF_INET6, trash.area, &((struct sockaddr_in6 *)addr)->sin6_addr))
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001684 return -1;
1685 end++;
1686
1687 /* Decode port. */
William Lallemand3d7a9182022-03-25 17:37:51 +01001688 if (end < url + ulen && *end == ':') {
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001689 end++;
1690 default_port = read_uint(&end, url + ulen);
1691 }
1692 ((struct sockaddr_in6 *)addr)->sin6_port = htons(default_port);
1693 ((struct sockaddr_in6 *)addr)->sin6_family = AF_INET6;
1694 return end - url;
1695 }
1696 else {
William Lallemand8a913742022-02-18 16:13:12 +01001697 /* we need to copy the string into the trash because url2ipv4
1698 * needs a \0 at the end of the string */
1699 if (trash.size < ulen)
1700 return -1;
1701
1702 memcpy(trash.area, curr, ulen - (curr - url));
1703 trash.area[ulen - (curr - url)] = '\0';
1704
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001705 /* We are looking for IP address. If you want to parse and
1706 * resolve hostname found in url, you can use str2sa_range(), but
1707 * be warned this can slow down global daemon performances
1708 * while handling lagging dns responses.
1709 */
William Lallemand8a913742022-02-18 16:13:12 +01001710 ret = url2ipv4(trash.area, &((struct sockaddr_in *)addr)->sin_addr);
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001711 if (ret) {
1712 /* Update out. */
1713 if (out) {
1714 out->host = curr;
1715 out->host_len = ret;
1716 }
1717
William Lallemandb938b772022-03-24 21:59:03 +01001718 curr += ret;
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001719
1720 /* Decode port. */
William Lallemand3d7a9182022-03-25 17:37:51 +01001721 if (curr < url + ulen && *curr == ':') {
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001722 curr++;
1723 default_port = read_uint(&curr, url + ulen);
1724 }
1725 ((struct sockaddr_in *)addr)->sin_port = htons(default_port);
1726
1727 /* Set family. */
1728 ((struct sockaddr_in *)addr)->sin_family = AF_INET;
1729 return curr - url;
1730 }
1731 else if (global.mode & MODE_STARTING) {
1732 /* The IPv4 and IPv6 decoding fails, maybe the url contain name. Try to execute
1733 * synchronous DNS request only if HAProxy is in the start state.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001734 */
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001735
1736 /* look for : or / or end */
1737 for (end = curr;
1738 end < url + ulen && *end != '/' && *end != ':';
1739 end++);
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001740 memcpy(trash.area, curr, end - curr);
1741 trash.area[end - curr] = '\0';
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001742
1743 /* try to resolve an IPv4/IPv6 hostname */
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001744 he = gethostbyname(trash.area);
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001745 if (!he)
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001746 return -1;
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001747
1748 /* Update out. */
1749 if (out) {
1750 out->host = curr;
1751 out->host_len = end - curr;
1752 }
1753
1754 /* Decode port. */
William Lallemand3d7a9182022-03-25 17:37:51 +01001755 if (end < url + ulen && *end == ':') {
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001756 end++;
1757 default_port = read_uint(&end, url + ulen);
1758 }
1759
1760 /* Copy IP address, set port and family. */
1761 switch (he->h_addrtype) {
1762 case AF_INET:
1763 ((struct sockaddr_in *)addr)->sin_addr = *(struct in_addr *) *(he->h_addr_list);
1764 ((struct sockaddr_in *)addr)->sin_port = htons(default_port);
1765 ((struct sockaddr_in *)addr)->sin_family = AF_INET;
1766 return end - url;
1767
1768 case AF_INET6:
1769 ((struct sockaddr_in6 *)addr)->sin6_addr = *(struct in6_addr *) *(he->h_addr_list);
1770 ((struct sockaddr_in6 *)addr)->sin6_port = htons(default_port);
1771 ((struct sockaddr_in6 *)addr)->sin6_family = AF_INET6;
1772 return end - url;
1773 }
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001774 }
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001775 }
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001776 return -1;
1777}
1778
Willy Tarreau631f01c2011-09-05 00:36:48 +02001779/* Tries to convert a sockaddr_storage address to text form. Upon success, the
1780 * address family is returned so that it's easy for the caller to adapt to the
1781 * output format. Zero is returned if the address family is not supported. -1
1782 * is returned upon error, with errno set. AF_INET, AF_INET6 and AF_UNIX are
1783 * supported.
1784 */
Willy Tarreaud5ec4bf2019-04-25 17:48:16 +02001785int addr_to_str(const struct sockaddr_storage *addr, char *str, int size)
Willy Tarreau631f01c2011-09-05 00:36:48 +02001786{
1787
Willy Tarreaud5ec4bf2019-04-25 17:48:16 +02001788 const void *ptr;
Willy Tarreau631f01c2011-09-05 00:36:48 +02001789
1790 if (size < 5)
1791 return 0;
1792 *str = '\0';
1793
1794 switch (addr->ss_family) {
1795 case AF_INET:
1796 ptr = &((struct sockaddr_in *)addr)->sin_addr;
1797 break;
1798 case AF_INET6:
1799 ptr = &((struct sockaddr_in6 *)addr)->sin6_addr;
1800 break;
1801 case AF_UNIX:
1802 memcpy(str, "unix", 5);
1803 return addr->ss_family;
1804 default:
1805 return 0;
1806 }
1807
1808 if (inet_ntop(addr->ss_family, ptr, str, size))
1809 return addr->ss_family;
1810
1811 /* failed */
1812 return -1;
1813}
1814
Simon Horman75ab8bd2014-06-16 09:39:41 +09001815/* Tries to convert a sockaddr_storage port to text form. Upon success, the
1816 * address family is returned so that it's easy for the caller to adapt to the
1817 * output format. Zero is returned if the address family is not supported. -1
1818 * is returned upon error, with errno set. AF_INET, AF_INET6 and AF_UNIX are
1819 * supported.
1820 */
Willy Tarreaud5ec4bf2019-04-25 17:48:16 +02001821int port_to_str(const struct sockaddr_storage *addr, char *str, int size)
Simon Horman75ab8bd2014-06-16 09:39:41 +09001822{
1823
1824 uint16_t port;
1825
1826
Willy Tarreaud7dad1b2017-01-06 16:46:22 +01001827 if (size < 6)
Simon Horman75ab8bd2014-06-16 09:39:41 +09001828 return 0;
1829 *str = '\0';
1830
1831 switch (addr->ss_family) {
1832 case AF_INET:
1833 port = ((struct sockaddr_in *)addr)->sin_port;
1834 break;
1835 case AF_INET6:
1836 port = ((struct sockaddr_in6 *)addr)->sin6_port;
1837 break;
1838 case AF_UNIX:
1839 memcpy(str, "unix", 5);
1840 return addr->ss_family;
1841 default:
1842 return 0;
1843 }
1844
1845 snprintf(str, size, "%u", ntohs(port));
1846 return addr->ss_family;
1847}
1848
Willy Tarreau16e01562016-08-09 16:46:18 +02001849/* check if the given address is local to the system or not. It will return
1850 * -1 when it's not possible to know, 0 when the address is not local, 1 when
1851 * it is. We don't want to iterate over all interfaces for this (and it is not
1852 * portable). So instead we try to bind in UDP to this address on a free non
1853 * privileged port and to connect to the same address, port 0 (connect doesn't
1854 * care). If it succeeds, we own the address. Note that non-inet addresses are
1855 * considered local since they're most likely AF_UNIX.
1856 */
1857int addr_is_local(const struct netns_entry *ns,
1858 const struct sockaddr_storage *orig)
1859{
1860 struct sockaddr_storage addr;
1861 int result;
1862 int fd;
1863
1864 if (!is_inet_addr(orig))
1865 return 1;
1866
1867 memcpy(&addr, orig, sizeof(addr));
1868 set_host_port(&addr, 0);
1869
1870 fd = my_socketat(ns, addr.ss_family, SOCK_DGRAM, IPPROTO_UDP);
1871 if (fd < 0)
1872 return -1;
1873
1874 result = -1;
1875 if (bind(fd, (struct sockaddr *)&addr, get_addr_len(&addr)) == 0) {
1876 if (connect(fd, (struct sockaddr *)&addr, get_addr_len(&addr)) == -1)
1877 result = 0; // fail, non-local address
1878 else
1879 result = 1; // success, local address
1880 }
1881 else {
1882 if (errno == EADDRNOTAVAIL)
1883 result = 0; // definitely not local :-)
1884 }
1885 close(fd);
1886
1887 return result;
1888}
1889
Willy Tarreaubaaee002006-06-26 02:48:02 +02001890/* will try to encode the string <string> replacing all characters tagged in
1891 * <map> with the hexadecimal representation of their ASCII-code (2 digits)
1892 * prefixed by <escape>, and will store the result between <start> (included)
1893 * and <stop> (excluded), and will always terminate the string with a '\0'
1894 * before <stop>. The position of the '\0' is returned if the conversion
1895 * completes. If bytes are missing between <start> and <stop>, then the
1896 * conversion will be incomplete and truncated. If <stop> <= <start>, the '\0'
1897 * cannot even be stored so we return <start> without writing the 0.
1898 * The input string must also be zero-terminated.
1899 */
1900const char hextab[16] = "0123456789ABCDEF";
1901char *encode_string(char *start, char *stop,
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001902 const char escape, const long *map,
Willy Tarreaubaaee002006-06-26 02:48:02 +02001903 const char *string)
1904{
1905 if (start < stop) {
1906 stop--; /* reserve one byte for the final '\0' */
1907 while (start < stop && *string != '\0') {
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001908 if (!ha_bit_test((unsigned char)(*string), map))
Willy Tarreaubaaee002006-06-26 02:48:02 +02001909 *start++ = *string;
1910 else {
1911 if (start + 3 >= stop)
1912 break;
1913 *start++ = escape;
1914 *start++ = hextab[(*string >> 4) & 15];
1915 *start++ = hextab[*string & 15];
1916 }
1917 string++;
1918 }
1919 *start = '\0';
1920 }
1921 return start;
1922}
1923
Thierry FOURNIERe059ec92014-03-17 12:01:13 +01001924/*
1925 * Same behavior as encode_string() above, except that it encodes chunk
1926 * <chunk> instead of a string.
1927 */
1928char *encode_chunk(char *start, char *stop,
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001929 const char escape, const long *map,
Willy Tarreau83061a82018-07-13 11:56:34 +02001930 const struct buffer *chunk)
Thierry FOURNIERe059ec92014-03-17 12:01:13 +01001931{
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001932 char *str = chunk->area;
1933 char *end = chunk->area + chunk->data;
Thierry FOURNIERe059ec92014-03-17 12:01:13 +01001934
1935 if (start < stop) {
1936 stop--; /* reserve one byte for the final '\0' */
1937 while (start < stop && str < end) {
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001938 if (!ha_bit_test((unsigned char)(*str), map))
Thierry FOURNIERe059ec92014-03-17 12:01:13 +01001939 *start++ = *str;
1940 else {
1941 if (start + 3 >= stop)
1942 break;
1943 *start++ = escape;
1944 *start++ = hextab[(*str >> 4) & 15];
1945 *start++ = hextab[*str & 15];
1946 }
1947 str++;
1948 }
1949 *start = '\0';
1950 }
1951 return start;
1952}
1953
Dragan Dosen0edd1092016-02-12 13:23:02 +01001954/*
1955 * Tries to prefix characters tagged in the <map> with the <escape>
Dragan Dosen1a5d0602016-07-22 16:00:31 +02001956 * character. The input <string> must be zero-terminated. The result will
1957 * be stored between <start> (included) and <stop> (excluded). This
1958 * function will always try to terminate the resulting string with a '\0'
1959 * before <stop>, and will return its position if the conversion
1960 * completes.
1961 */
1962char *escape_string(char *start, char *stop,
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001963 const char escape, const long *map,
Dragan Dosen1a5d0602016-07-22 16:00:31 +02001964 const char *string)
1965{
1966 if (start < stop) {
1967 stop--; /* reserve one byte for the final '\0' */
1968 while (start < stop && *string != '\0') {
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001969 if (!ha_bit_test((unsigned char)(*string), map))
Dragan Dosen1a5d0602016-07-22 16:00:31 +02001970 *start++ = *string;
1971 else {
1972 if (start + 2 >= stop)
1973 break;
1974 *start++ = escape;
1975 *start++ = *string;
1976 }
1977 string++;
1978 }
1979 *start = '\0';
1980 }
1981 return start;
1982}
1983
1984/*
1985 * Tries to prefix characters tagged in the <map> with the <escape>
Dragan Dosen0edd1092016-02-12 13:23:02 +01001986 * character. <chunk> contains the input to be escaped. The result will be
1987 * stored between <start> (included) and <stop> (excluded). The function
1988 * will always try to terminate the resulting string with a '\0' before
1989 * <stop>, and will return its position if the conversion completes.
1990 */
1991char *escape_chunk(char *start, char *stop,
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001992 const char escape, const long *map,
Willy Tarreau83061a82018-07-13 11:56:34 +02001993 const struct buffer *chunk)
Dragan Dosen0edd1092016-02-12 13:23:02 +01001994{
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001995 char *str = chunk->area;
1996 char *end = chunk->area + chunk->data;
Dragan Dosen0edd1092016-02-12 13:23:02 +01001997
1998 if (start < stop) {
1999 stop--; /* reserve one byte for the final '\0' */
2000 while (start < stop && str < end) {
Willy Tarreau1bfd6022019-06-07 11:10:07 +02002001 if (!ha_bit_test((unsigned char)(*str), map))
Dragan Dosen0edd1092016-02-12 13:23:02 +01002002 *start++ = *str;
2003 else {
2004 if (start + 2 >= stop)
2005 break;
2006 *start++ = escape;
2007 *start++ = *str;
2008 }
2009 str++;
2010 }
2011 *start = '\0';
2012 }
2013 return start;
2014}
2015
Thierry FOURNIERddea6262015-05-28 16:00:28 +02002016/* Check a string for using it in a CSV output format. If the string contains
2017 * one of the following four char <">, <,>, CR or LF, the string is
2018 * encapsulated between <"> and the <"> are escaped by a <""> sequence.
2019 * <str> is the input string to be escaped. The function assumes that
2020 * the input string is null-terminated.
2021 *
2022 * If <quote> is 0, the result is returned escaped but without double quote.
Willy Tarreau898529b2016-01-06 18:07:04 +01002023 * It is useful if the escaped string is used between double quotes in the
Thierry FOURNIERddea6262015-05-28 16:00:28 +02002024 * format.
2025 *
Willy Tarreau898529b2016-01-06 18:07:04 +01002026 * printf("..., \"%s\", ...\r\n", csv_enc(str, 0, &trash));
Thierry FOURNIERddea6262015-05-28 16:00:28 +02002027 *
Willy Tarreaub631c292016-01-08 10:04:08 +01002028 * If <quote> is 1, the converter puts the quotes only if any reserved character
2029 * is present. If <quote> is 2, the converter always puts the quotes.
Thierry FOURNIERddea6262015-05-28 16:00:28 +02002030 *
Willy Tarreau83061a82018-07-13 11:56:34 +02002031 * <output> is a struct buffer used for storing the output string.
Thierry FOURNIERddea6262015-05-28 16:00:28 +02002032 *
Willy Tarreau898529b2016-01-06 18:07:04 +01002033 * The function returns the converted string on its output. If an error
2034 * occurs, the function returns an empty string. This type of output is useful
Thierry FOURNIERddea6262015-05-28 16:00:28 +02002035 * for using the function directly as printf() argument.
2036 *
2037 * If the output buffer is too short to contain the input string, the result
2038 * is truncated.
Willy Tarreau898529b2016-01-06 18:07:04 +01002039 *
Willy Tarreaub631c292016-01-08 10:04:08 +01002040 * This function appends the encoding to the existing output chunk, and it
2041 * guarantees that it starts immediately at the first available character of
2042 * the chunk. Please use csv_enc() instead if you want to replace the output
2043 * chunk.
Thierry FOURNIERddea6262015-05-28 16:00:28 +02002044 */
Willy Tarreau83061a82018-07-13 11:56:34 +02002045const char *csv_enc_append(const char *str, int quote, struct buffer *output)
Thierry FOURNIERddea6262015-05-28 16:00:28 +02002046{
Willy Tarreau843b7cb2018-07-13 10:54:26 +02002047 char *end = output->area + output->size;
2048 char *out = output->area + output->data;
Willy Tarreau898529b2016-01-06 18:07:04 +01002049 char *ptr = out;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02002050
Willy Tarreaub631c292016-01-08 10:04:08 +01002051 if (quote == 1) {
2052 /* automatic quoting: first verify if we'll have to quote the string */
2053 if (!strpbrk(str, "\n\r,\""))
2054 quote = 0;
2055 }
2056
2057 if (quote)
2058 *ptr++ = '"';
2059
Willy Tarreau898529b2016-01-06 18:07:04 +01002060 while (*str && ptr < end - 2) { /* -2 for reserving space for <"> and \0. */
2061 *ptr = *str;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02002062 if (*str == '"') {
Willy Tarreau898529b2016-01-06 18:07:04 +01002063 ptr++;
2064 if (ptr >= end - 2) {
2065 ptr--;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02002066 break;
2067 }
Willy Tarreau898529b2016-01-06 18:07:04 +01002068 *ptr = '"';
Thierry FOURNIERddea6262015-05-28 16:00:28 +02002069 }
Willy Tarreau898529b2016-01-06 18:07:04 +01002070 ptr++;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02002071 str++;
2072 }
2073
Willy Tarreaub631c292016-01-08 10:04:08 +01002074 if (quote)
2075 *ptr++ = '"';
Thierry FOURNIERddea6262015-05-28 16:00:28 +02002076
Willy Tarreau898529b2016-01-06 18:07:04 +01002077 *ptr = '\0';
Willy Tarreau843b7cb2018-07-13 10:54:26 +02002078 output->data = ptr - output->area;
Willy Tarreau898529b2016-01-06 18:07:04 +01002079 return out;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02002080}
2081
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02002082/* Decode an URL-encoded string in-place. The resulting string might
2083 * be shorter. If some forbidden characters are found, the conversion is
Thierry FOURNIER5068d962013-10-04 16:27:27 +02002084 * aborted, the string is truncated before the issue and a negative value is
2085 * returned, otherwise the operation returns the length of the decoded string.
Willy Tarreau62ba9ba2020-04-23 17:54:47 +02002086 * If the 'in_form' argument is non-nul the string is assumed to be part of
2087 * an "application/x-www-form-urlencoded" encoded string, and the '+' will be
2088 * turned to a space. If it's zero, this will only be done after a question
2089 * mark ('?').
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02002090 */
Willy Tarreau62ba9ba2020-04-23 17:54:47 +02002091int url_decode(char *string, int in_form)
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02002092{
2093 char *in, *out;
Thierry FOURNIER5068d962013-10-04 16:27:27 +02002094 int ret = -1;
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02002095
2096 in = string;
2097 out = string;
2098 while (*in) {
2099 switch (*in) {
2100 case '+' :
Willy Tarreau62ba9ba2020-04-23 17:54:47 +02002101 *out++ = in_form ? ' ' : *in;
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02002102 break;
2103 case '%' :
2104 if (!ishex(in[1]) || !ishex(in[2]))
2105 goto end;
2106 *out++ = (hex2i(in[1]) << 4) + hex2i(in[2]);
2107 in += 2;
2108 break;
Willy Tarreau62ba9ba2020-04-23 17:54:47 +02002109 case '?':
2110 in_form = 1;
2111 /* fall through */
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02002112 default:
2113 *out++ = *in;
2114 break;
2115 }
2116 in++;
2117 }
Thierry FOURNIER5068d962013-10-04 16:27:27 +02002118 ret = out - string; /* success */
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02002119 end:
2120 *out = 0;
2121 return ret;
2122}
Willy Tarreaubaaee002006-06-26 02:48:02 +02002123
Willy Tarreau6911fa42007-03-04 18:06:08 +01002124unsigned int str2ui(const char *s)
2125{
2126 return __str2ui(s);
2127}
2128
2129unsigned int str2uic(const char *s)
2130{
2131 return __str2uic(s);
2132}
2133
2134unsigned int strl2ui(const char *s, int len)
2135{
2136 return __strl2ui(s, len);
2137}
2138
2139unsigned int strl2uic(const char *s, int len)
2140{
2141 return __strl2uic(s, len);
2142}
2143
Willy Tarreau4ec83cd2010-10-15 23:19:55 +02002144unsigned int read_uint(const char **s, const char *end)
2145{
2146 return __read_uint(s, end);
2147}
2148
Thierry FOURNIER763a5d82015-07-06 23:09:52 +02002149/* This function reads an unsigned integer from the string pointed to by <s> and
2150 * returns it. The <s> pointer is adjusted to point to the first unread char. The
2151 * function automatically stops at <end>. If the number overflows, the 2^64-1
2152 * value is returned.
2153 */
2154unsigned long long int read_uint64(const char **s, const char *end)
2155{
2156 const char *ptr = *s;
2157 unsigned long long int i = 0, tmp;
2158 unsigned int j;
2159
2160 while (ptr < end) {
2161
2162 /* read next char */
2163 j = *ptr - '0';
2164 if (j > 9)
2165 goto read_uint64_end;
2166
2167 /* add char to the number and check overflow. */
2168 tmp = i * 10;
2169 if (tmp / 10 != i) {
2170 i = ULLONG_MAX;
2171 goto read_uint64_eat;
2172 }
2173 if (ULLONG_MAX - tmp < j) {
2174 i = ULLONG_MAX;
2175 goto read_uint64_eat;
2176 }
2177 i = tmp + j;
2178 ptr++;
2179 }
2180read_uint64_eat:
2181 /* eat each numeric char */
2182 while (ptr < end) {
2183 if ((unsigned int)(*ptr - '0') > 9)
2184 break;
2185 ptr++;
2186 }
2187read_uint64_end:
2188 *s = ptr;
2189 return i;
2190}
2191
2192/* This function reads an integer from the string pointed to by <s> and returns
2193 * it. The <s> pointer is adjusted to point to the first unread char. The function
2194 * automatically stops at <end>. Il the number is bigger than 2^63-2, the 2^63-1
2195 * value is returned. If the number is lowest than -2^63-1, the -2^63 value is
2196 * returned.
2197 */
2198long long int read_int64(const char **s, const char *end)
2199{
2200 unsigned long long int i = 0;
2201 int neg = 0;
2202
2203 /* Look for minus char. */
2204 if (**s == '-') {
2205 neg = 1;
2206 (*s)++;
2207 }
2208 else if (**s == '+')
2209 (*s)++;
2210
2211 /* convert as positive number. */
2212 i = read_uint64(s, end);
2213
2214 if (neg) {
2215 if (i > 0x8000000000000000ULL)
2216 return LLONG_MIN;
2217 return -i;
2218 }
2219 if (i > 0x7fffffffffffffffULL)
2220 return LLONG_MAX;
2221 return i;
2222}
2223
Willy Tarreau6911fa42007-03-04 18:06:08 +01002224/* This one is 7 times faster than strtol() on athlon with checks.
2225 * It returns the value of the number composed of all valid digits read,
2226 * and can process negative numbers too.
2227 */
2228int strl2ic(const char *s, int len)
2229{
2230 int i = 0;
Willy Tarreau3f0c9762007-10-25 09:42:24 +02002231 int j, k;
Willy Tarreau6911fa42007-03-04 18:06:08 +01002232
2233 if (len > 0) {
2234 if (*s != '-') {
2235 /* positive number */
2236 while (len-- > 0) {
2237 j = (*s++) - '0';
Willy Tarreau3f0c9762007-10-25 09:42:24 +02002238 k = i * 10;
Willy Tarreau6911fa42007-03-04 18:06:08 +01002239 if (j > 9)
2240 break;
Willy Tarreau3f0c9762007-10-25 09:42:24 +02002241 i = k + j;
Willy Tarreau6911fa42007-03-04 18:06:08 +01002242 }
2243 } else {
2244 /* negative number */
2245 s++;
2246 while (--len > 0) {
2247 j = (*s++) - '0';
Willy Tarreau3f0c9762007-10-25 09:42:24 +02002248 k = i * 10;
Willy Tarreau6911fa42007-03-04 18:06:08 +01002249 if (j > 9)
2250 break;
Willy Tarreau3f0c9762007-10-25 09:42:24 +02002251 i = k - j;
Willy Tarreau6911fa42007-03-04 18:06:08 +01002252 }
2253 }
2254 }
2255 return i;
2256}
2257
2258
2259/* This function reads exactly <len> chars from <s> and converts them to a
2260 * signed integer which it stores into <ret>. It accurately detects any error
2261 * (truncated string, invalid chars, overflows). It is meant to be used in
2262 * applications designed for hostile environments. It returns zero when the
2263 * number has successfully been converted, non-zero otherwise. When an error
2264 * is returned, the <ret> value is left untouched. It is yet 5 to 40 times
2265 * faster than strtol().
2266 */
2267int strl2irc(const char *s, int len, int *ret)
2268{
2269 int i = 0;
2270 int j;
2271
2272 if (!len)
2273 return 1;
2274
2275 if (*s != '-') {
2276 /* positive number */
2277 while (len-- > 0) {
2278 j = (*s++) - '0';
2279 if (j > 9) return 1; /* invalid char */
2280 if (i > INT_MAX / 10) return 1; /* check for multiply overflow */
2281 i = i * 10;
2282 if (i + j < i) return 1; /* check for addition overflow */
2283 i = i + j;
2284 }
2285 } else {
2286 /* negative number */
2287 s++;
2288 while (--len > 0) {
2289 j = (*s++) - '0';
2290 if (j > 9) return 1; /* invalid char */
2291 if (i < INT_MIN / 10) return 1; /* check for multiply overflow */
2292 i = i * 10;
2293 if (i - j > i) return 1; /* check for subtract overflow */
2294 i = i - j;
2295 }
2296 }
2297 *ret = i;
2298 return 0;
2299}
2300
2301
2302/* This function reads exactly <len> chars from <s> and converts them to a
2303 * signed integer which it stores into <ret>. It accurately detects any error
2304 * (truncated string, invalid chars, overflows). It is meant to be used in
2305 * applications designed for hostile environments. It returns zero when the
2306 * number has successfully been converted, non-zero otherwise. When an error
2307 * is returned, the <ret> value is left untouched. It is about 3 times slower
William Dauchy060ffc82021-02-06 20:47:51 +01002308 * than strl2irc().
Willy Tarreau6911fa42007-03-04 18:06:08 +01002309 */
Willy Tarreau6911fa42007-03-04 18:06:08 +01002310
2311int strl2llrc(const char *s, int len, long long *ret)
2312{
2313 long long i = 0;
2314 int j;
2315
2316 if (!len)
2317 return 1;
2318
2319 if (*s != '-') {
2320 /* positive number */
2321 while (len-- > 0) {
2322 j = (*s++) - '0';
2323 if (j > 9) return 1; /* invalid char */
2324 if (i > LLONG_MAX / 10LL) return 1; /* check for multiply overflow */
2325 i = i * 10LL;
2326 if (i + j < i) return 1; /* check for addition overflow */
2327 i = i + j;
2328 }
2329 } else {
2330 /* negative number */
2331 s++;
2332 while (--len > 0) {
2333 j = (*s++) - '0';
2334 if (j > 9) return 1; /* invalid char */
2335 if (i < LLONG_MIN / 10LL) return 1; /* check for multiply overflow */
2336 i = i * 10LL;
2337 if (i - j > i) return 1; /* check for subtract overflow */
2338 i = i - j;
2339 }
2340 }
2341 *ret = i;
2342 return 0;
2343}
2344
Thierry FOURNIER511e9472014-01-23 17:40:34 +01002345/* This function is used with pat_parse_dotted_ver(). It converts a string
2346 * composed by two number separated by a dot. Each part must contain in 16 bits
2347 * because internally they will be represented as a 32-bit quantity stored in
2348 * a 64-bit integer. It returns zero when the number has successfully been
2349 * converted, non-zero otherwise. When an error is returned, the <ret> value
2350 * is left untouched.
2351 *
2352 * "1.3" -> 0x0000000000010003
2353 * "65535.65535" -> 0x00000000ffffffff
2354 */
2355int strl2llrc_dotted(const char *text, int len, long long *ret)
2356{
2357 const char *end = &text[len];
2358 const char *p;
2359 long long major, minor;
2360
2361 /* Look for dot. */
2362 for (p = text; p < end; p++)
2363 if (*p == '.')
2364 break;
2365
2366 /* Convert major. */
2367 if (strl2llrc(text, p - text, &major) != 0)
2368 return 1;
2369
2370 /* Check major. */
2371 if (major >= 65536)
2372 return 1;
2373
2374 /* Convert minor. */
2375 minor = 0;
2376 if (p < end)
2377 if (strl2llrc(p + 1, end - (p + 1), &minor) != 0)
2378 return 1;
2379
2380 /* Check minor. */
2381 if (minor >= 65536)
2382 return 1;
2383
2384 /* Compose value. */
2385 *ret = (major << 16) | (minor & 0xffff);
2386 return 0;
2387}
2388
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002389/* This function parses a time value optionally followed by a unit suffix among
2390 * "d", "h", "m", "s", "ms" or "us". It converts the value into the unit
2391 * expected by the caller. The computation does its best to avoid overflows.
2392 * The value is returned in <ret> if everything is fine, and a NULL is returned
2393 * by the function. In case of error, a pointer to the error is returned and
2394 * <ret> is left untouched. Values are automatically rounded up when needed.
Willy Tarreau9faebe32019-06-07 19:00:37 +02002395 * Values resulting in values larger than or equal to 2^31 after conversion are
2396 * reported as an overflow as value PARSE_TIME_OVER. Non-null values resulting
2397 * in an underflow are reported as an underflow as value PARSE_TIME_UNDER.
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002398 */
2399const char *parse_time_err(const char *text, unsigned *ret, unsigned unit_flags)
2400{
Willy Tarreau9faebe32019-06-07 19:00:37 +02002401 unsigned long long imult, idiv;
2402 unsigned long long omult, odiv;
2403 unsigned long long value, result;
Christopher Fauletc20ad0d2020-12-11 09:23:07 +01002404 const char *str = text;
2405
2406 if (!isdigit((unsigned char)*text))
2407 return text;
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002408
2409 omult = odiv = 1;
2410
2411 switch (unit_flags & TIME_UNIT_MASK) {
2412 case TIME_UNIT_US: omult = 1000000; break;
2413 case TIME_UNIT_MS: omult = 1000; break;
2414 case TIME_UNIT_S: break;
2415 case TIME_UNIT_MIN: odiv = 60; break;
2416 case TIME_UNIT_HOUR: odiv = 3600; break;
2417 case TIME_UNIT_DAY: odiv = 86400; break;
2418 default: break;
2419 }
2420
2421 value = 0;
2422
2423 while (1) {
2424 unsigned int j;
2425
2426 j = *text - '0';
2427 if (j > 9)
2428 break;
2429 text++;
2430 value *= 10;
2431 value += j;
2432 }
2433
2434 imult = idiv = 1;
2435 switch (*text) {
2436 case '\0': /* no unit = default unit */
2437 imult = omult = idiv = odiv = 1;
Christopher Fauletc20ad0d2020-12-11 09:23:07 +01002438 goto end;
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002439 case 's': /* second = unscaled unit */
2440 break;
2441 case 'u': /* microsecond : "us" */
2442 if (text[1] == 's') {
2443 idiv = 1000000;
2444 text++;
Thayne McCombsa6838052021-04-02 14:12:43 -06002445 break;
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002446 }
Christopher Fauletc20ad0d2020-12-11 09:23:07 +01002447 return text;
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002448 case 'm': /* millisecond : "ms" or minute: "m" */
2449 if (text[1] == 's') {
2450 idiv = 1000;
2451 text++;
2452 } else
2453 imult = 60;
2454 break;
2455 case 'h': /* hour : "h" */
2456 imult = 3600;
2457 break;
2458 case 'd': /* day : "d" */
2459 imult = 86400;
2460 break;
2461 default:
2462 return text;
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002463 }
Christopher Fauletc20ad0d2020-12-11 09:23:07 +01002464 if (*(++text) != '\0') {
2465 ha_warning("unexpected character '%c' after the timer value '%s', only "
2466 "(us=microseconds,ms=milliseconds,s=seconds,m=minutes,h=hours,d=days) are supported."
2467 " This will be reported as an error in next versions.\n", *text, str);
2468 }
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002469
Christopher Fauletc20ad0d2020-12-11 09:23:07 +01002470 end:
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002471 if (omult % idiv == 0) { omult /= idiv; idiv = 1; }
2472 if (idiv % omult == 0) { idiv /= omult; omult = 1; }
2473 if (imult % odiv == 0) { imult /= odiv; odiv = 1; }
2474 if (odiv % imult == 0) { odiv /= imult; imult = 1; }
2475
Willy Tarreau9faebe32019-06-07 19:00:37 +02002476 result = (value * (imult * omult) + (idiv * odiv - 1)) / (idiv * odiv);
2477 if (result >= 0x80000000)
2478 return PARSE_TIME_OVER;
2479 if (!result && value)
2480 return PARSE_TIME_UNDER;
2481 *ret = result;
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002482 return NULL;
2483}
Willy Tarreau6911fa42007-03-04 18:06:08 +01002484
Emeric Brun39132b22010-01-04 14:57:24 +01002485/* this function converts the string starting at <text> to an unsigned int
2486 * stored in <ret>. If an error is detected, the pointer to the unexpected
Joseph Herlant32b83272018-11-15 11:58:28 -08002487 * character is returned. If the conversion is successful, NULL is returned.
Emeric Brun39132b22010-01-04 14:57:24 +01002488 */
2489const char *parse_size_err(const char *text, unsigned *ret) {
2490 unsigned value = 0;
2491
Christopher Faulet82635a02020-12-11 09:30:45 +01002492 if (!isdigit((unsigned char)*text))
2493 return text;
2494
Emeric Brun39132b22010-01-04 14:57:24 +01002495 while (1) {
2496 unsigned int j;
2497
2498 j = *text - '0';
2499 if (j > 9)
2500 break;
2501 if (value > ~0U / 10)
2502 return text;
2503 value *= 10;
2504 if (value > (value + j))
2505 return text;
2506 value += j;
2507 text++;
2508 }
2509
2510 switch (*text) {
2511 case '\0':
2512 break;
2513 case 'K':
2514 case 'k':
2515 if (value > ~0U >> 10)
2516 return text;
2517 value = value << 10;
2518 break;
2519 case 'M':
2520 case 'm':
2521 if (value > ~0U >> 20)
2522 return text;
2523 value = value << 20;
2524 break;
2525 case 'G':
2526 case 'g':
2527 if (value > ~0U >> 30)
2528 return text;
2529 value = value << 30;
2530 break;
2531 default:
2532 return text;
2533 }
2534
Godbach58048a22015-01-28 17:36:16 +08002535 if (*text != '\0' && *++text != '\0')
2536 return text;
2537
Emeric Brun39132b22010-01-04 14:57:24 +01002538 *ret = value;
2539 return NULL;
2540}
2541
Willy Tarreau126d4062013-12-03 17:50:47 +01002542/*
2543 * Parse binary string written in hexadecimal (source) and store the decoded
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05002544 * result into binstr and set binstrlen to the length of binstr. Memory for
Willy Tarreau126d4062013-12-03 17:50:47 +01002545 * binstr is allocated by the function. In case of error, returns 0 with an
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05002546 * error message in err. In success case, it returns the consumed length.
Willy Tarreau126d4062013-12-03 17:50:47 +01002547 */
2548int parse_binary(const char *source, char **binstr, int *binstrlen, char **err)
2549{
2550 int len;
2551 const char *p = source;
2552 int i,j;
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002553 int alloc;
Willy Tarreau126d4062013-12-03 17:50:47 +01002554
2555 len = strlen(source);
2556 if (len % 2) {
2557 memprintf(err, "an even number of hex digit is expected");
2558 return 0;
2559 }
2560
2561 len = len >> 1;
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002562
Willy Tarreau126d4062013-12-03 17:50:47 +01002563 if (!*binstr) {
Tim Duesterhuse52b6e52020-09-12 20:26:43 +02002564 *binstr = calloc(len, sizeof(**binstr));
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002565 if (!*binstr) {
2566 memprintf(err, "out of memory while loading string pattern");
2567 return 0;
2568 }
2569 alloc = 1;
Willy Tarreau126d4062013-12-03 17:50:47 +01002570 }
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002571 else {
2572 if (*binstrlen < len) {
Joseph Herlant76dbe782018-11-15 12:01:22 -08002573 memprintf(err, "no space available in the buffer. expect %d, provides %d",
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002574 len, *binstrlen);
2575 return 0;
2576 }
2577 alloc = 0;
2578 }
2579 *binstrlen = len;
Willy Tarreau126d4062013-12-03 17:50:47 +01002580
2581 i = j = 0;
2582 while (j < len) {
2583 if (!ishex(p[i++]))
2584 goto bad_input;
2585 if (!ishex(p[i++]))
2586 goto bad_input;
2587 (*binstr)[j++] = (hex2i(p[i-2]) << 4) + hex2i(p[i-1]);
2588 }
Thierry FOURNIERee330af2014-01-21 11:36:14 +01002589 return len << 1;
Willy Tarreau126d4062013-12-03 17:50:47 +01002590
2591bad_input:
2592 memprintf(err, "an hex digit is expected (found '%c')", p[i-1]);
Willy Tarreau61cfdf42021-02-20 10:46:51 +01002593 if (alloc)
2594 ha_free(binstr);
Willy Tarreau126d4062013-12-03 17:50:47 +01002595 return 0;
2596}
2597
Willy Tarreau946ba592009-05-10 15:41:18 +02002598/* copies at most <n> characters from <src> and always terminates with '\0' */
2599char *my_strndup(const char *src, int n)
2600{
2601 int len = 0;
2602 char *ret;
2603
2604 while (len < n && src[len])
2605 len++;
2606
Vincent Bernat3c2f2f22016-04-03 13:48:42 +02002607 ret = malloc(len + 1);
Willy Tarreau946ba592009-05-10 15:41:18 +02002608 if (!ret)
2609 return ret;
2610 memcpy(ret, src, len);
2611 ret[len] = '\0';
2612 return ret;
2613}
2614
Baptiste Assmannbb77c8e2013-10-06 23:24:13 +02002615/*
2616 * search needle in haystack
2617 * returns the pointer if found, returns NULL otherwise
2618 */
2619const void *my_memmem(const void *haystack, size_t haystacklen, const void *needle, size_t needlelen)
2620{
2621 const void *c = NULL;
2622 unsigned char f;
2623
2624 if ((haystack == NULL) || (needle == NULL) || (haystacklen < needlelen))
2625 return NULL;
2626
2627 f = *(char *)needle;
2628 c = haystack;
2629 while ((c = memchr(c, f, haystacklen - (c - haystack))) != NULL) {
2630 if ((haystacklen - (c - haystack)) < needlelen)
2631 return NULL;
2632
2633 if (memcmp(c, needle, needlelen) == 0)
2634 return c;
2635 ++c;
2636 }
2637 return NULL;
2638}
2639
Ilya Shipitsinc02a23f2020-05-06 00:53:22 +05002640/* get length of the initial segment consisting entirely of bytes in <accept> */
Christopher Faulet5eb96cb2020-04-15 10:23:01 +02002641size_t my_memspn(const void *str, size_t len, const void *accept, size_t acceptlen)
2642{
2643 size_t ret = 0;
2644
2645 while (ret < len && memchr(accept, *((int *)str), acceptlen)) {
2646 str++;
2647 ret++;
2648 }
2649 return ret;
2650}
2651
Ilya Shipitsinc02a23f2020-05-06 00:53:22 +05002652/* get length of the initial segment consisting entirely of bytes not in <rejcet> */
Christopher Faulet5eb96cb2020-04-15 10:23:01 +02002653size_t my_memcspn(const void *str, size_t len, const void *reject, size_t rejectlen)
2654{
2655 size_t ret = 0;
2656
2657 while (ret < len) {
2658 if(memchr(reject, *((int *)str), rejectlen))
2659 return ret;
2660 str++;
2661 ret++;
2662 }
2663 return ret;
2664}
2665
Willy Tarreau482b00d2009-10-04 22:48:42 +02002666/* This function returns the first unused key greater than or equal to <key> in
2667 * ID tree <root>. Zero is returned if no place is found.
2668 */
2669unsigned int get_next_id(struct eb_root *root, unsigned int key)
2670{
2671 struct eb32_node *used;
2672
2673 do {
2674 used = eb32_lookup_ge(root, key);
2675 if (!used || used->key > key)
2676 return key; /* key is available */
2677 key++;
2678 } while (key);
2679 return key;
2680}
2681
Willy Tarreau9c1e15d2017-11-15 18:51:29 +01002682/* dump the full tree to <file> in DOT format for debugging purposes. Will
2683 * optionally highlight node <subj> if found, depending on operation <op> :
2684 * 0 : nothing
2685 * >0 : insertion, node/leaf are surrounded in red
2686 * <0 : removal, node/leaf are dashed with no background
2687 * Will optionally add "desc" as a label on the graph if set and non-null.
2688 */
2689void 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 +01002690{
2691 struct eb32sc_node *node;
2692 unsigned long scope = -1;
2693
Willy Tarreau9c1e15d2017-11-15 18:51:29 +01002694 fprintf(file, "digraph ebtree {\n");
2695
2696 if (desc && *desc) {
2697 fprintf(file,
2698 " fontname=\"fixed\";\n"
2699 " fontsize=8;\n"
2700 " label=\"%s\";\n", desc);
2701 }
2702
Willy Tarreaued3cda02017-11-15 15:04:05 +01002703 fprintf(file,
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002704 " node [fontname=\"fixed\" fontsize=8 shape=\"box\" style=\"filled\" color=\"black\" fillcolor=\"white\"];\n"
2705 " edge [fontname=\"fixed\" fontsize=8 style=\"solid\" color=\"magenta\" dir=\"forward\"];\n"
Willy Tarreaued3cda02017-11-15 15:04:05 +01002706 " \"%lx_n\" [label=\"root\\n%lx\"]\n", (long)eb_root_to_node(root), (long)root
2707 );
2708
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002709 fprintf(file, " \"%lx_n\" -> \"%lx_%c\" [taillabel=\"L\"];\n",
Willy Tarreaued3cda02017-11-15 15:04:05 +01002710 (long)eb_root_to_node(root),
2711 (long)eb_root_to_node(eb_clrtag(root->b[0])),
Willy Tarreaued3cda02017-11-15 15:04:05 +01002712 eb_gettag(root->b[0]) == EB_LEAF ? 'l' : 'n');
2713
2714 node = eb32sc_first(root, scope);
2715 while (node) {
2716 if (node->node.node_p) {
2717 /* node part is used */
Willy Tarreau9c1e15d2017-11-15 18:51:29 +01002718 fprintf(file, " \"%lx_n\" [label=\"%lx\\nkey=%u\\nscope=%lx\\nbit=%d\" fillcolor=\"lightskyblue1\" %s];\n",
2719 (long)node, (long)node, node->key, node->node_s, node->node.bit,
2720 (node == subj) ? (op < 0 ? "color=\"red\" style=\"dashed\"" : op > 0 ? "color=\"red\"" : "") : "");
Willy Tarreaued3cda02017-11-15 15:04:05 +01002721
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002722 fprintf(file, " \"%lx_n\" -> \"%lx_n\" [taillabel=\"%c\"];\n",
Willy Tarreaued3cda02017-11-15 15:04:05 +01002723 (long)node,
2724 (long)eb_root_to_node(eb_clrtag(node->node.node_p)),
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002725 eb_gettag(node->node.node_p) ? 'R' : 'L');
Willy Tarreaued3cda02017-11-15 15:04:05 +01002726
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002727 fprintf(file, " \"%lx_n\" -> \"%lx_%c\" [taillabel=\"L\"];\n",
Willy Tarreaued3cda02017-11-15 15:04:05 +01002728 (long)node,
2729 (long)eb_root_to_node(eb_clrtag(node->node.branches.b[0])),
Willy Tarreaued3cda02017-11-15 15:04:05 +01002730 eb_gettag(node->node.branches.b[0]) == EB_LEAF ? 'l' : 'n');
2731
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002732 fprintf(file, " \"%lx_n\" -> \"%lx_%c\" [taillabel=\"R\"];\n",
Willy Tarreaued3cda02017-11-15 15:04:05 +01002733 (long)node,
2734 (long)eb_root_to_node(eb_clrtag(node->node.branches.b[1])),
Willy Tarreaued3cda02017-11-15 15:04:05 +01002735 eb_gettag(node->node.branches.b[1]) == EB_LEAF ? 'l' : 'n');
2736 }
2737
Willy Tarreau9c1e15d2017-11-15 18:51:29 +01002738 fprintf(file, " \"%lx_l\" [label=\"%lx\\nkey=%u\\nscope=%lx\\npfx=%u\" fillcolor=\"yellow\" %s];\n",
2739 (long)node, (long)node, node->key, node->leaf_s, node->node.pfx,
2740 (node == subj) ? (op < 0 ? "color=\"red\" style=\"dashed\"" : op > 0 ? "color=\"red\"" : "") : "");
Willy Tarreaued3cda02017-11-15 15:04:05 +01002741
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002742 fprintf(file, " \"%lx_l\" -> \"%lx_n\" [taillabel=\"%c\"];\n",
Willy Tarreaued3cda02017-11-15 15:04:05 +01002743 (long)node,
2744 (long)eb_root_to_node(eb_clrtag(node->node.leaf_p)),
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002745 eb_gettag(node->node.leaf_p) ? 'R' : 'L');
Willy Tarreaued3cda02017-11-15 15:04:05 +01002746 node = eb32sc_next(node, scope);
2747 }
2748 fprintf(file, "}\n");
2749}
2750
Willy Tarreau348238b2010-01-18 15:05:57 +01002751/* This function compares a sample word possibly followed by blanks to another
2752 * clean word. The compare is case-insensitive. 1 is returned if both are equal,
2753 * otherwise zero. This intends to be used when checking HTTP headers for some
2754 * values. Note that it validates a word followed only by blanks but does not
2755 * validate a word followed by blanks then other chars.
2756 */
2757int word_match(const char *sample, int slen, const char *word, int wlen)
2758{
2759 if (slen < wlen)
2760 return 0;
2761
2762 while (wlen) {
2763 char c = *sample ^ *word;
2764 if (c && c != ('A' ^ 'a'))
2765 return 0;
2766 sample++;
2767 word++;
2768 slen--;
2769 wlen--;
2770 }
2771
2772 while (slen) {
2773 if (*sample != ' ' && *sample != '\t')
2774 return 0;
2775 sample++;
2776 slen--;
2777 }
2778 return 1;
2779}
Willy Tarreau482b00d2009-10-04 22:48:42 +02002780
Willy Tarreaud54bbdc2009-09-07 11:00:31 +02002781/* Converts any text-formatted IPv4 address to a host-order IPv4 address. It
2782 * is particularly fast because it avoids expensive operations such as
2783 * multiplies, which are optimized away at the end. It requires a properly
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05002784 * formatted address though (3 points).
Willy Tarreaud54bbdc2009-09-07 11:00:31 +02002785 */
2786unsigned int inetaddr_host(const char *text)
2787{
2788 const unsigned int ascii_zero = ('0' << 24) | ('0' << 16) | ('0' << 8) | '0';
2789 register unsigned int dig100, dig10, dig1;
2790 int s;
2791 const char *p, *d;
2792
2793 dig1 = dig10 = dig100 = ascii_zero;
2794 s = 24;
2795
2796 p = text;
2797 while (1) {
2798 if (((unsigned)(*p - '0')) <= 9) {
2799 p++;
2800 continue;
2801 }
2802
2803 /* here, we have a complete byte between <text> and <p> (exclusive) */
2804 if (p == text)
2805 goto end;
2806
2807 d = p - 1;
2808 dig1 |= (unsigned int)(*d << s);
2809 if (d == text)
2810 goto end;
2811
2812 d--;
2813 dig10 |= (unsigned int)(*d << s);
2814 if (d == text)
2815 goto end;
2816
2817 d--;
2818 dig100 |= (unsigned int)(*d << s);
2819 end:
2820 if (!s || *p != '.')
2821 break;
2822
2823 s -= 8;
2824 text = ++p;
2825 }
2826
2827 dig100 -= ascii_zero;
2828 dig10 -= ascii_zero;
2829 dig1 -= ascii_zero;
2830 return ((dig100 * 10) + dig10) * 10 + dig1;
2831}
2832
2833/*
2834 * Idem except the first unparsed character has to be passed in <stop>.
2835 */
2836unsigned int inetaddr_host_lim(const char *text, const char *stop)
2837{
2838 const unsigned int ascii_zero = ('0' << 24) | ('0' << 16) | ('0' << 8) | '0';
2839 register unsigned int dig100, dig10, dig1;
2840 int s;
2841 const char *p, *d;
2842
2843 dig1 = dig10 = dig100 = ascii_zero;
2844 s = 24;
2845
2846 p = text;
2847 while (1) {
2848 if (((unsigned)(*p - '0')) <= 9 && p < stop) {
2849 p++;
2850 continue;
2851 }
2852
2853 /* here, we have a complete byte between <text> and <p> (exclusive) */
2854 if (p == text)
2855 goto end;
2856
2857 d = p - 1;
2858 dig1 |= (unsigned int)(*d << s);
2859 if (d == text)
2860 goto end;
2861
2862 d--;
2863 dig10 |= (unsigned int)(*d << s);
2864 if (d == text)
2865 goto end;
2866
2867 d--;
2868 dig100 |= (unsigned int)(*d << s);
2869 end:
2870 if (!s || p == stop || *p != '.')
2871 break;
2872
2873 s -= 8;
2874 text = ++p;
2875 }
2876
2877 dig100 -= ascii_zero;
2878 dig10 -= ascii_zero;
2879 dig1 -= ascii_zero;
2880 return ((dig100 * 10) + dig10) * 10 + dig1;
2881}
2882
2883/*
2884 * Idem except the pointer to first unparsed byte is returned into <ret> which
2885 * must not be NULL.
2886 */
Willy Tarreau74172752010-10-15 23:21:42 +02002887unsigned int inetaddr_host_lim_ret(char *text, char *stop, char **ret)
Willy Tarreaud54bbdc2009-09-07 11:00:31 +02002888{
2889 const unsigned int ascii_zero = ('0' << 24) | ('0' << 16) | ('0' << 8) | '0';
2890 register unsigned int dig100, dig10, dig1;
2891 int s;
Willy Tarreau74172752010-10-15 23:21:42 +02002892 char *p, *d;
Willy Tarreaud54bbdc2009-09-07 11:00:31 +02002893
2894 dig1 = dig10 = dig100 = ascii_zero;
2895 s = 24;
2896
2897 p = text;
2898 while (1) {
2899 if (((unsigned)(*p - '0')) <= 9 && p < stop) {
2900 p++;
2901 continue;
2902 }
2903
2904 /* here, we have a complete byte between <text> and <p> (exclusive) */
2905 if (p == text)
2906 goto end;
2907
2908 d = p - 1;
2909 dig1 |= (unsigned int)(*d << s);
2910 if (d == text)
2911 goto end;
2912
2913 d--;
2914 dig10 |= (unsigned int)(*d << s);
2915 if (d == text)
2916 goto end;
2917
2918 d--;
2919 dig100 |= (unsigned int)(*d << s);
2920 end:
2921 if (!s || p == stop || *p != '.')
2922 break;
2923
2924 s -= 8;
2925 text = ++p;
2926 }
2927
2928 *ret = p;
2929 dig100 -= ascii_zero;
2930 dig10 -= ascii_zero;
2931 dig1 -= ascii_zero;
2932 return ((dig100 * 10) + dig10) * 10 + dig1;
2933}
2934
Willy Tarreauf0b38bf2010-06-06 13:22:23 +02002935/* Convert a fixed-length string to an IP address. Returns 0 in case of error,
2936 * or the number of chars read in case of success. Maybe this could be replaced
2937 * by one of the functions above. Also, apparently this function does not support
2938 * hosts above 255 and requires exactly 4 octets.
Willy Tarreau075415a2013-12-12 11:29:39 +01002939 * The destination is only modified on success.
Willy Tarreauf0b38bf2010-06-06 13:22:23 +02002940 */
2941int buf2ip(const char *buf, size_t len, struct in_addr *dst)
2942{
2943 const char *addr;
2944 int saw_digit, octets, ch;
2945 u_char tmp[4], *tp;
2946 const char *cp = buf;
2947
2948 saw_digit = 0;
2949 octets = 0;
2950 *(tp = tmp) = 0;
2951
2952 for (addr = buf; addr - buf < len; addr++) {
2953 unsigned char digit = (ch = *addr) - '0';
2954
2955 if (digit > 9 && ch != '.')
2956 break;
2957
2958 if (digit <= 9) {
2959 u_int new = *tp * 10 + digit;
2960
2961 if (new > 255)
2962 return 0;
2963
2964 *tp = new;
2965
2966 if (!saw_digit) {
2967 if (++octets > 4)
2968 return 0;
2969 saw_digit = 1;
2970 }
2971 } else if (ch == '.' && saw_digit) {
2972 if (octets == 4)
2973 return 0;
2974
2975 *++tp = 0;
2976 saw_digit = 0;
2977 } else
2978 return 0;
2979 }
2980
2981 if (octets < 4)
2982 return 0;
2983
2984 memcpy(&dst->s_addr, tmp, 4);
2985 return addr - cp;
2986}
2987
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002988/* This function converts the string in <buf> of the len <len> to
2989 * struct in6_addr <dst> which must be allocated by the caller.
2990 * This function returns 1 in success case, otherwise zero.
Willy Tarreau075415a2013-12-12 11:29:39 +01002991 * The destination is only modified on success.
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002992 */
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002993int buf2ip6(const char *buf, size_t len, struct in6_addr *dst)
2994{
Thierry FOURNIERcd659912013-12-11 12:33:54 +01002995 char null_term_ip6[INET6_ADDRSTRLEN + 1];
Willy Tarreau075415a2013-12-12 11:29:39 +01002996 struct in6_addr out;
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002997
Thierry FOURNIERcd659912013-12-11 12:33:54 +01002998 if (len > INET6_ADDRSTRLEN)
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002999 return 0;
3000
3001 memcpy(null_term_ip6, buf, len);
3002 null_term_ip6[len] = '\0';
3003
Willy Tarreau075415a2013-12-12 11:29:39 +01003004 if (!inet_pton(AF_INET6, null_term_ip6, &out))
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01003005 return 0;
3006
Willy Tarreau075415a2013-12-12 11:29:39 +01003007 *dst = out;
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01003008 return 1;
3009}
3010
Willy Tarreauacf95772010-06-14 19:09:21 +02003011/* To be used to quote config arg positions. Returns the short string at <ptr>
3012 * surrounded by simple quotes if <ptr> is valid and non-empty, or "end of line"
3013 * if ptr is NULL or empty. The string is locally allocated.
3014 */
3015const char *quote_arg(const char *ptr)
3016{
Christopher Faulet1bc04c72017-10-29 20:14:08 +01003017 static THREAD_LOCAL char val[32];
Willy Tarreauacf95772010-06-14 19:09:21 +02003018 int i;
3019
3020 if (!ptr || !*ptr)
3021 return "end of line";
3022 val[0] = '\'';
Willy Tarreaude2dd6b2013-01-24 02:14:42 +01003023 for (i = 1; i < sizeof(val) - 2 && *ptr; i++)
Willy Tarreauacf95772010-06-14 19:09:21 +02003024 val[i] = *ptr++;
3025 val[i++] = '\'';
3026 val[i] = '\0';
3027 return val;
3028}
3029
Willy Tarreau5b180202010-07-18 10:40:48 +02003030/* returns an operator among STD_OP_* for string <str> or < 0 if unknown */
3031int get_std_op(const char *str)
3032{
3033 int ret = -1;
3034
3035 if (*str == 'e' && str[1] == 'q')
3036 ret = STD_OP_EQ;
3037 else if (*str == 'n' && str[1] == 'e')
3038 ret = STD_OP_NE;
3039 else if (*str == 'l') {
3040 if (str[1] == 'e') ret = STD_OP_LE;
3041 else if (str[1] == 't') ret = STD_OP_LT;
3042 }
3043 else if (*str == 'g') {
3044 if (str[1] == 'e') ret = STD_OP_GE;
3045 else if (str[1] == 't') ret = STD_OP_GT;
3046 }
3047
3048 if (ret == -1 || str[2] != '\0')
3049 return -1;
3050 return ret;
3051}
3052
Willy Tarreau4c14eaa2010-11-24 14:01:45 +01003053/* hash a 32-bit integer to another 32-bit integer */
3054unsigned int full_hash(unsigned int a)
3055{
3056 return __full_hash(a);
3057}
3058
Willy Tarreauf3241112019-02-26 09:56:22 +01003059/* Return the bit position in mask <m> of the nth bit set of rank <r>, between
3060 * 0 and LONGBITS-1 included, starting from the left. For example ranks 0,1,2,3
3061 * for mask 0x55 will be 6, 4, 2 and 0 respectively. This algorithm is based on
3062 * a popcount variant and is described here :
3063 * https://graphics.stanford.edu/~seander/bithacks.html
3064 */
3065unsigned int mask_find_rank_bit(unsigned int r, unsigned long m)
3066{
3067 unsigned long a, b, c, d;
3068 unsigned int s;
3069 unsigned int t;
3070
3071 a = m - ((m >> 1) & ~0UL/3);
3072 b = (a & ~0UL/5) + ((a >> 2) & ~0UL/5);
3073 c = (b + (b >> 4)) & ~0UL/0x11;
3074 d = (c + (c >> 8)) & ~0UL/0x101;
3075
3076 r++; // make r be 1..64
3077
3078 t = 0;
3079 s = LONGBITS;
3080 if (s > 32) {
Willy Tarreau9b6be3b2019-03-18 16:31:18 +01003081 unsigned long d2 = (d >> 16) >> 16;
3082 t = d2 + (d2 >> 16);
Willy Tarreauf3241112019-02-26 09:56:22 +01003083 s -= ((t - r) & 256) >> 3; r -= (t & ((t - r) >> 8));
3084 }
3085
3086 t = (d >> (s - 16)) & 0xff;
3087 s -= ((t - r) & 256) >> 4; r -= (t & ((t - r) >> 8));
3088 t = (c >> (s - 8)) & 0xf;
3089 s -= ((t - r) & 256) >> 5; r -= (t & ((t - r) >> 8));
3090 t = (b >> (s - 4)) & 0x7;
3091 s -= ((t - r) & 256) >> 6; r -= (t & ((t - r) >> 8));
3092 t = (a >> (s - 2)) & 0x3;
3093 s -= ((t - r) & 256) >> 7; r -= (t & ((t - r) >> 8));
3094 t = (m >> (s - 1)) & 0x1;
3095 s -= ((t - r) & 256) >> 8;
3096
3097 return s - 1;
3098}
3099
3100/* Same as mask_find_rank_bit() above but makes use of pre-computed bitmaps
3101 * based on <m>, in <a..d>. These ones must be updated whenever <m> changes
3102 * using mask_prep_rank_map() below.
3103 */
3104unsigned int mask_find_rank_bit_fast(unsigned int r, unsigned long m,
3105 unsigned long a, unsigned long b,
3106 unsigned long c, unsigned long d)
3107{
3108 unsigned int s;
3109 unsigned int t;
3110
3111 r++; // make r be 1..64
3112
3113 t = 0;
3114 s = LONGBITS;
3115 if (s > 32) {
Willy Tarreau9b6be3b2019-03-18 16:31:18 +01003116 unsigned long d2 = (d >> 16) >> 16;
3117 t = d2 + (d2 >> 16);
Willy Tarreauf3241112019-02-26 09:56:22 +01003118 s -= ((t - r) & 256) >> 3; r -= (t & ((t - r) >> 8));
3119 }
3120
3121 t = (d >> (s - 16)) & 0xff;
3122 s -= ((t - r) & 256) >> 4; r -= (t & ((t - r) >> 8));
3123 t = (c >> (s - 8)) & 0xf;
3124 s -= ((t - r) & 256) >> 5; r -= (t & ((t - r) >> 8));
3125 t = (b >> (s - 4)) & 0x7;
3126 s -= ((t - r) & 256) >> 6; r -= (t & ((t - r) >> 8));
3127 t = (a >> (s - 2)) & 0x3;
3128 s -= ((t - r) & 256) >> 7; r -= (t & ((t - r) >> 8));
3129 t = (m >> (s - 1)) & 0x1;
3130 s -= ((t - r) & 256) >> 8;
3131
3132 return s - 1;
3133}
3134
3135/* Prepare the bitmaps used by the fast implementation of the find_rank_bit()
3136 * above.
3137 */
3138void mask_prep_rank_map(unsigned long m,
3139 unsigned long *a, unsigned long *b,
3140 unsigned long *c, unsigned long *d)
3141{
3142 *a = m - ((m >> 1) & ~0UL/3);
3143 *b = (*a & ~0UL/5) + ((*a >> 2) & ~0UL/5);
3144 *c = (*b + (*b >> 4)) & ~0UL/0x11;
3145 *d = (*c + (*c >> 8)) & ~0UL/0x101;
3146}
3147
David du Colombier4f92d322011-03-24 11:09:31 +01003148/* Return non-zero if IPv4 address is part of the network,
Willy Tarreaueec1d382016-07-13 11:59:39 +02003149 * otherwise zero. Note that <addr> may not necessarily be aligned
3150 * while the two other ones must.
David du Colombier4f92d322011-03-24 11:09:31 +01003151 */
Willy Tarreaueec1d382016-07-13 11:59:39 +02003152int in_net_ipv4(const void *addr, const struct in_addr *mask, const struct in_addr *net)
David du Colombier4f92d322011-03-24 11:09:31 +01003153{
Willy Tarreaueec1d382016-07-13 11:59:39 +02003154 struct in_addr addr_copy;
3155
3156 memcpy(&addr_copy, addr, sizeof(addr_copy));
3157 return((addr_copy.s_addr & mask->s_addr) == (net->s_addr & mask->s_addr));
David du Colombier4f92d322011-03-24 11:09:31 +01003158}
3159
3160/* Return non-zero if IPv6 address is part of the network,
Willy Tarreaueec1d382016-07-13 11:59:39 +02003161 * otherwise zero. Note that <addr> may not necessarily be aligned
3162 * while the two other ones must.
David du Colombier4f92d322011-03-24 11:09:31 +01003163 */
Willy Tarreaueec1d382016-07-13 11:59:39 +02003164int in_net_ipv6(const void *addr, const struct in6_addr *mask, const struct in6_addr *net)
David du Colombier4f92d322011-03-24 11:09:31 +01003165{
3166 int i;
Willy Tarreaueec1d382016-07-13 11:59:39 +02003167 struct in6_addr addr_copy;
David du Colombier4f92d322011-03-24 11:09:31 +01003168
Willy Tarreaueec1d382016-07-13 11:59:39 +02003169 memcpy(&addr_copy, addr, sizeof(addr_copy));
David du Colombier4f92d322011-03-24 11:09:31 +01003170 for (i = 0; i < sizeof(struct in6_addr) / sizeof(int); i++)
Willy Tarreaueec1d382016-07-13 11:59:39 +02003171 if (((((int *)&addr_copy)[i] & ((int *)mask)[i])) !=
David du Colombier4f92d322011-03-24 11:09:31 +01003172 (((int *)net)[i] & ((int *)mask)[i]))
3173 return 0;
3174 return 1;
3175}
3176
3177/* RFC 4291 prefix */
3178const char rfc4291_pfx[] = { 0x00, 0x00, 0x00, 0x00,
3179 0x00, 0x00, 0x00, 0x00,
3180 0x00, 0x00, 0xFF, 0xFF };
3181
Joseph Herlant32b83272018-11-15 11:58:28 -08003182/* Map IPv4 address on IPv6 address, as specified in RFC 3513.
Thierry FOURNIER4a04dc32013-11-28 16:33:15 +01003183 * Input and output may overlap.
3184 */
David du Colombier4f92d322011-03-24 11:09:31 +01003185void v4tov6(struct in6_addr *sin6_addr, struct in_addr *sin_addr)
3186{
Thierry FOURNIER4a04dc32013-11-28 16:33:15 +01003187 struct in_addr tmp_addr;
3188
3189 tmp_addr.s_addr = sin_addr->s_addr;
David du Colombier4f92d322011-03-24 11:09:31 +01003190 memcpy(sin6_addr->s6_addr, rfc4291_pfx, sizeof(rfc4291_pfx));
Thierry FOURNIER4a04dc32013-11-28 16:33:15 +01003191 memcpy(sin6_addr->s6_addr+12, &tmp_addr.s_addr, 4);
David du Colombier4f92d322011-03-24 11:09:31 +01003192}
3193
Joseph Herlant32b83272018-11-15 11:58:28 -08003194/* Map IPv6 address on IPv4 address, as specified in RFC 3513.
David du Colombier4f92d322011-03-24 11:09:31 +01003195 * Return true if conversion is possible and false otherwise.
3196 */
3197int v6tov4(struct in_addr *sin_addr, struct in6_addr *sin6_addr)
3198{
3199 if (memcmp(sin6_addr->s6_addr, rfc4291_pfx, sizeof(rfc4291_pfx)) == 0) {
3200 memcpy(&(sin_addr->s_addr), &(sin6_addr->s6_addr[12]),
3201 sizeof(struct in_addr));
3202 return 1;
3203 }
3204
3205 return 0;
3206}
3207
Baptiste Assmann08b24cf2016-01-23 23:39:12 +01003208/* compare two struct sockaddr_storage and return:
3209 * 0 (true) if the addr is the same in both
3210 * 1 (false) if the addr is not the same in both
3211 * -1 (unable) if one of the addr is not AF_INET*
3212 */
3213int ipcmp(struct sockaddr_storage *ss1, struct sockaddr_storage *ss2)
3214{
3215 if ((ss1->ss_family != AF_INET) && (ss1->ss_family != AF_INET6))
3216 return -1;
3217
3218 if ((ss2->ss_family != AF_INET) && (ss2->ss_family != AF_INET6))
3219 return -1;
3220
3221 if (ss1->ss_family != ss2->ss_family)
3222 return 1;
3223
3224 switch (ss1->ss_family) {
3225 case AF_INET:
3226 return memcmp(&((struct sockaddr_in *)ss1)->sin_addr,
3227 &((struct sockaddr_in *)ss2)->sin_addr,
3228 sizeof(struct in_addr)) != 0;
3229 case AF_INET6:
3230 return memcmp(&((struct sockaddr_in6 *)ss1)->sin6_addr,
3231 &((struct sockaddr_in6 *)ss2)->sin6_addr,
3232 sizeof(struct in6_addr)) != 0;
3233 }
3234
3235 return 1;
3236}
3237
Christopher Faulet9553de72021-02-26 09:12:50 +01003238/* compare a struct sockaddr_storage to a struct net_addr and return :
3239 * 0 (true) if <addr> is matching <net>
3240 * 1 (false) if <addr> is not matching <net>
3241 * -1 (unable) if <addr> or <net> is not AF_INET*
3242 */
3243int ipcmp2net(const struct sockaddr_storage *addr, const struct net_addr *net)
3244{
3245 if ((addr->ss_family != AF_INET) && (addr->ss_family != AF_INET6))
3246 return -1;
3247
3248 if ((net->family != AF_INET) && (net->family != AF_INET6))
3249 return -1;
3250
3251 if (addr->ss_family != net->family)
3252 return 1;
3253
3254 if (addr->ss_family == AF_INET &&
3255 (((struct sockaddr_in *)addr)->sin_addr.s_addr & net->addr.v4.mask.s_addr) == net->addr.v4.ip.s_addr)
3256 return 0;
3257 else {
3258 const struct in6_addr *addr6 = &(((const struct sockaddr_in6*)addr)->sin6_addr);
3259 const struct in6_addr *nip6 = &net->addr.v6.ip;
3260 const struct in6_addr *nmask6 = &net->addr.v6.mask;
3261
3262 if ((read_u32(&addr6->s6_addr[0]) & read_u32(&nmask6->s6_addr[0])) == read_u32(&nip6->s6_addr[0]) &&
3263 (read_u32(&addr6->s6_addr[4]) & read_u32(&nmask6->s6_addr[4])) == read_u32(&nip6->s6_addr[4]) &&
3264 (read_u32(&addr6->s6_addr[8]) & read_u32(&nmask6->s6_addr[8])) == read_u32(&nip6->s6_addr[8]) &&
3265 (read_u32(&addr6->s6_addr[12]) & read_u32(&nmask6->s6_addr[12])) == read_u32(&nip6->s6_addr[12]))
3266 return 0;
3267 }
3268
3269 return 1;
3270}
3271
Baptiste Assmann08396c82016-01-31 00:27:17 +01003272/* copy IP address from <source> into <dest>
Willy Tarreaudc3a9e82016-11-04 18:47:01 +01003273 * The caller must allocate and clear <dest> before calling.
3274 * The source must be in either AF_INET or AF_INET6 family, or the destination
3275 * address will be undefined. If the destination address used to hold a port,
3276 * it is preserved, so that this function can be used to switch to another
3277 * address family with no risk. Returns a pointer to the destination.
Baptiste Assmann08396c82016-01-31 00:27:17 +01003278 */
3279struct sockaddr_storage *ipcpy(struct sockaddr_storage *source, struct sockaddr_storage *dest)
3280{
Willy Tarreaudc3a9e82016-11-04 18:47:01 +01003281 int prev_port;
3282
3283 prev_port = get_net_port(dest);
3284 memset(dest, 0, sizeof(*dest));
Baptiste Assmann08396c82016-01-31 00:27:17 +01003285 dest->ss_family = source->ss_family;
3286
3287 /* copy new addr and apply it */
3288 switch (source->ss_family) {
3289 case AF_INET:
3290 ((struct sockaddr_in *)dest)->sin_addr.s_addr = ((struct sockaddr_in *)source)->sin_addr.s_addr;
Willy Tarreaudc3a9e82016-11-04 18:47:01 +01003291 ((struct sockaddr_in *)dest)->sin_port = prev_port;
Baptiste Assmann08396c82016-01-31 00:27:17 +01003292 break;
3293 case AF_INET6:
3294 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 +01003295 ((struct sockaddr_in6 *)dest)->sin6_port = prev_port;
Baptiste Assmann08396c82016-01-31 00:27:17 +01003296 break;
3297 }
3298
3299 return dest;
3300}
3301
William Lallemand421f5b52012-02-06 18:15:57 +01003302char *human_time(int t, short hz_div) {
3303 static char rv[sizeof("24855d23h")+1]; // longest of "23h59m" and "59m59s"
3304 char *p = rv;
Willy Tarreau761b3d52014-04-14 14:53:06 +02003305 char *end = rv + sizeof(rv);
William Lallemand421f5b52012-02-06 18:15:57 +01003306 int cnt=2; // print two numbers
3307
3308 if (unlikely(t < 0 || hz_div <= 0)) {
Willy Tarreau761b3d52014-04-14 14:53:06 +02003309 snprintf(p, end - p, "?");
William Lallemand421f5b52012-02-06 18:15:57 +01003310 return rv;
3311 }
3312
3313 if (unlikely(hz_div > 1))
3314 t /= hz_div;
3315
3316 if (t >= DAY) {
Willy Tarreau761b3d52014-04-14 14:53:06 +02003317 p += snprintf(p, end - p, "%dd", t / DAY);
William Lallemand421f5b52012-02-06 18:15:57 +01003318 cnt--;
3319 }
3320
3321 if (cnt && t % DAY / HOUR) {
Willy Tarreau761b3d52014-04-14 14:53:06 +02003322 p += snprintf(p, end - p, "%dh", t % DAY / HOUR);
William Lallemand421f5b52012-02-06 18:15:57 +01003323 cnt--;
3324 }
3325
3326 if (cnt && t % HOUR / MINUTE) {
Willy Tarreau761b3d52014-04-14 14:53:06 +02003327 p += snprintf(p, end - p, "%dm", t % HOUR / MINUTE);
William Lallemand421f5b52012-02-06 18:15:57 +01003328 cnt--;
3329 }
3330
3331 if ((cnt && t % MINUTE) || !t) // also display '0s'
Willy Tarreau761b3d52014-04-14 14:53:06 +02003332 p += snprintf(p, end - p, "%ds", t % MINUTE / SEC);
William Lallemand421f5b52012-02-06 18:15:57 +01003333
3334 return rv;
3335}
3336
3337const char *monthname[12] = {
3338 "Jan", "Feb", "Mar", "Apr", "May", "Jun",
3339 "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
3340};
3341
3342/* date2str_log: write a date in the format :
3343 * sprintf(str, "%02d/%s/%04d:%02d:%02d:%02d.%03d",
3344 * tm.tm_mday, monthname[tm.tm_mon], tm.tm_year+1900,
3345 * tm.tm_hour, tm.tm_min, tm.tm_sec, (int)date.tv_usec/1000);
3346 *
3347 * without using sprintf. return a pointer to the last char written (\0) or
3348 * NULL if there isn't enough space.
3349 */
Willy Tarreauf16cb412018-09-04 19:08:48 +02003350char *date2str_log(char *dst, const struct tm *tm, const struct timeval *date, size_t size)
William Lallemand421f5b52012-02-06 18:15:57 +01003351{
3352
3353 if (size < 25) /* the size is fixed: 24 chars + \0 */
3354 return NULL;
3355
3356 dst = utoa_pad((unsigned int)tm->tm_mday, dst, 3); // day
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003357 if (!dst)
3358 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003359 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003360
William Lallemand421f5b52012-02-06 18:15:57 +01003361 memcpy(dst, monthname[tm->tm_mon], 3); // month
3362 dst += 3;
3363 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003364
William Lallemand421f5b52012-02-06 18:15:57 +01003365 dst = utoa_pad((unsigned int)tm->tm_year+1900, dst, 5); // year
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003366 if (!dst)
3367 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003368 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003369
William Lallemand421f5b52012-02-06 18:15:57 +01003370 dst = utoa_pad((unsigned int)tm->tm_hour, dst, 3); // hour
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003371 if (!dst)
3372 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003373 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003374
William Lallemand421f5b52012-02-06 18:15:57 +01003375 dst = utoa_pad((unsigned int)tm->tm_min, dst, 3); // minutes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003376 if (!dst)
3377 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003378 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003379
William Lallemand421f5b52012-02-06 18:15:57 +01003380 dst = utoa_pad((unsigned int)tm->tm_sec, dst, 3); // secondes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003381 if (!dst)
3382 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003383 *dst++ = '.';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003384
Willy Tarreau7d9421d2020-02-29 09:08:02 +01003385 dst = utoa_pad((unsigned int)(date->tv_usec/1000)%1000, dst, 4); // milliseconds
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003386 if (!dst)
3387 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003388 *dst = '\0';
3389
3390 return dst;
3391}
3392
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003393/* Base year used to compute leap years */
3394#define TM_YEAR_BASE 1900
3395
3396/* Return the difference in seconds between two times (leap seconds are ignored).
3397 * Retrieved from glibc 2.18 source code.
3398 */
3399static int my_tm_diff(const struct tm *a, const struct tm *b)
3400{
3401 /* Compute intervening leap days correctly even if year is negative.
3402 * Take care to avoid int overflow in leap day calculations,
3403 * but it's OK to assume that A and B are close to each other.
3404 */
3405 int a4 = (a->tm_year >> 2) + (TM_YEAR_BASE >> 2) - ! (a->tm_year & 3);
3406 int b4 = (b->tm_year >> 2) + (TM_YEAR_BASE >> 2) - ! (b->tm_year & 3);
3407 int a100 = a4 / 25 - (a4 % 25 < 0);
3408 int b100 = b4 / 25 - (b4 % 25 < 0);
3409 int a400 = a100 >> 2;
3410 int b400 = b100 >> 2;
3411 int intervening_leap_days = (a4 - b4) - (a100 - b100) + (a400 - b400);
3412 int years = a->tm_year - b->tm_year;
3413 int days = (365 * years + intervening_leap_days
3414 + (a->tm_yday - b->tm_yday));
3415 return (60 * (60 * (24 * days + (a->tm_hour - b->tm_hour))
3416 + (a->tm_min - b->tm_min))
3417 + (a->tm_sec - b->tm_sec));
3418}
3419
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003420/* Return the GMT offset for a specific local time.
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003421 * Both t and tm must represent the same time.
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003422 * The string returned has the same format as returned by strftime(... "%z", tm).
3423 * Offsets are kept in an internal cache for better performances.
3424 */
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003425const char *get_gmt_offset(time_t t, struct tm *tm)
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003426{
3427 /* Cache offsets from GMT (depending on whether DST is active or not) */
Christopher Faulet1bc04c72017-10-29 20:14:08 +01003428 static THREAD_LOCAL char gmt_offsets[2][5+1] = { "", "" };
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003429
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003430 char *gmt_offset;
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003431 struct tm tm_gmt;
3432 int diff;
3433 int isdst = tm->tm_isdst;
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003434
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003435 /* Pretend DST not active if its status is unknown */
3436 if (isdst < 0)
3437 isdst = 0;
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003438
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003439 /* Fetch the offset and initialize it if needed */
3440 gmt_offset = gmt_offsets[isdst & 0x01];
3441 if (unlikely(!*gmt_offset)) {
3442 get_gmtime(t, &tm_gmt);
3443 diff = my_tm_diff(tm, &tm_gmt);
3444 if (diff < 0) {
3445 diff = -diff;
3446 *gmt_offset = '-';
3447 } else {
3448 *gmt_offset = '+';
3449 }
Willy Tarreaue112c8a2019-10-29 10:16:11 +01003450 diff %= 86400U;
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003451 diff /= 60; /* Convert to minutes */
3452 snprintf(gmt_offset+1, 4+1, "%02d%02d", diff/60, diff%60);
3453 }
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003454
Willy Tarreaue112c8a2019-10-29 10:16:11 +01003455 return gmt_offset;
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003456}
3457
William Lallemand421f5b52012-02-06 18:15:57 +01003458/* gmt2str_log: write a date in the format :
3459 * "%02d/%s/%04d:%02d:%02d:%02d +0000" without using snprintf
3460 * return a pointer to the last char written (\0) or
3461 * NULL if there isn't enough space.
3462 */
3463char *gmt2str_log(char *dst, struct tm *tm, size_t size)
3464{
Yuxans Yao4e25b012012-10-19 10:36:09 +08003465 if (size < 27) /* the size is fixed: 26 chars + \0 */
William Lallemand421f5b52012-02-06 18:15:57 +01003466 return NULL;
3467
3468 dst = utoa_pad((unsigned int)tm->tm_mday, dst, 3); // day
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003469 if (!dst)
3470 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003471 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003472
William Lallemand421f5b52012-02-06 18:15:57 +01003473 memcpy(dst, monthname[tm->tm_mon], 3); // month
3474 dst += 3;
3475 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003476
William Lallemand421f5b52012-02-06 18:15:57 +01003477 dst = utoa_pad((unsigned int)tm->tm_year+1900, dst, 5); // year
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003478 if (!dst)
3479 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003480 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003481
William Lallemand421f5b52012-02-06 18:15:57 +01003482 dst = utoa_pad((unsigned int)tm->tm_hour, dst, 3); // hour
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003483 if (!dst)
3484 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003485 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003486
William Lallemand421f5b52012-02-06 18:15:57 +01003487 dst = utoa_pad((unsigned int)tm->tm_min, dst, 3); // minutes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003488 if (!dst)
3489 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003490 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003491
William Lallemand421f5b52012-02-06 18:15:57 +01003492 dst = utoa_pad((unsigned int)tm->tm_sec, dst, 3); // secondes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003493 if (!dst)
3494 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003495 *dst++ = ' ';
3496 *dst++ = '+';
3497 *dst++ = '0';
3498 *dst++ = '0';
3499 *dst++ = '0';
3500 *dst++ = '0';
3501 *dst = '\0';
3502
3503 return dst;
3504}
3505
Yuxans Yao4e25b012012-10-19 10:36:09 +08003506/* localdate2str_log: write a date in the format :
3507 * "%02d/%s/%04d:%02d:%02d:%02d +0000(local timezone)" without using snprintf
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003508 * Both t and tm must represent the same time.
3509 * return a pointer to the last char written (\0) or
3510 * NULL if there isn't enough space.
Yuxans Yao4e25b012012-10-19 10:36:09 +08003511 */
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003512char *localdate2str_log(char *dst, time_t t, struct tm *tm, size_t size)
Yuxans Yao4e25b012012-10-19 10:36:09 +08003513{
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003514 const char *gmt_offset;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003515 if (size < 27) /* the size is fixed: 26 chars + \0 */
3516 return NULL;
3517
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003518 gmt_offset = get_gmt_offset(t, tm);
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003519
Yuxans Yao4e25b012012-10-19 10:36:09 +08003520 dst = utoa_pad((unsigned int)tm->tm_mday, dst, 3); // day
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003521 if (!dst)
3522 return NULL;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003523 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003524
Yuxans Yao4e25b012012-10-19 10:36:09 +08003525 memcpy(dst, monthname[tm->tm_mon], 3); // month
3526 dst += 3;
3527 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003528
Yuxans Yao4e25b012012-10-19 10:36:09 +08003529 dst = utoa_pad((unsigned int)tm->tm_year+1900, dst, 5); // year
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003530 if (!dst)
3531 return NULL;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003532 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003533
Yuxans Yao4e25b012012-10-19 10:36:09 +08003534 dst = utoa_pad((unsigned int)tm->tm_hour, dst, 3); // hour
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003535 if (!dst)
3536 return NULL;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003537 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003538
Yuxans Yao4e25b012012-10-19 10:36:09 +08003539 dst = utoa_pad((unsigned int)tm->tm_min, dst, 3); // minutes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003540 if (!dst)
3541 return NULL;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003542 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003543
Yuxans Yao4e25b012012-10-19 10:36:09 +08003544 dst = utoa_pad((unsigned int)tm->tm_sec, dst, 3); // secondes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003545 if (!dst)
3546 return NULL;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003547 *dst++ = ' ';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003548
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003549 memcpy(dst, gmt_offset, 5); // Offset from local time to GMT
Yuxans Yao4e25b012012-10-19 10:36:09 +08003550 dst += 5;
3551 *dst = '\0';
3552
3553 return dst;
3554}
3555
Willy Tarreaucb1949b2017-07-19 19:05:29 +02003556/* Returns the number of seconds since 01/01/1970 0:0:0 GMT for GMT date <tm>.
3557 * It is meant as a portable replacement for timegm() for use with valid inputs.
3558 * Returns undefined results for invalid dates (eg: months out of range 0..11).
3559 */
3560time_t my_timegm(const struct tm *tm)
3561{
3562 /* Each month has 28, 29, 30 or 31 days, or 28+N. The date in the year
3563 * is thus (current month - 1)*28 + cumulated_N[month] to count the
3564 * sum of the extra N days for elapsed months. The sum of all these N
3565 * days doesn't exceed 30 for a complete year (366-12*28) so it fits
3566 * in a 5-bit word. This means that with 60 bits we can represent a
3567 * matrix of all these values at once, which is fast and efficient to
3568 * access. The extra February day for leap years is not counted here.
3569 *
3570 * Jan : none = 0 (0)
3571 * Feb : Jan = 3 (3)
3572 * Mar : Jan..Feb = 3 (3 + 0)
3573 * Apr : Jan..Mar = 6 (3 + 0 + 3)
3574 * May : Jan..Apr = 8 (3 + 0 + 3 + 2)
3575 * Jun : Jan..May = 11 (3 + 0 + 3 + 2 + 3)
3576 * Jul : Jan..Jun = 13 (3 + 0 + 3 + 2 + 3 + 2)
3577 * Aug : Jan..Jul = 16 (3 + 0 + 3 + 2 + 3 + 2 + 3)
3578 * Sep : Jan..Aug = 19 (3 + 0 + 3 + 2 + 3 + 2 + 3 + 3)
3579 * Oct : Jan..Sep = 21 (3 + 0 + 3 + 2 + 3 + 2 + 3 + 3 + 2)
3580 * Nov : Jan..Oct = 24 (3 + 0 + 3 + 2 + 3 + 2 + 3 + 3 + 2 + 3)
3581 * Dec : Jan..Nov = 26 (3 + 0 + 3 + 2 + 3 + 2 + 3 + 3 + 2 + 3 + 2)
3582 */
3583 uint64_t extra =
3584 ( 0ULL << 0*5) + ( 3ULL << 1*5) + ( 3ULL << 2*5) + /* Jan, Feb, Mar, */
3585 ( 6ULL << 3*5) + ( 8ULL << 4*5) + (11ULL << 5*5) + /* Apr, May, Jun, */
3586 (13ULL << 6*5) + (16ULL << 7*5) + (19ULL << 8*5) + /* Jul, Aug, Sep, */
3587 (21ULL << 9*5) + (24ULL << 10*5) + (26ULL << 11*5); /* Oct, Nov, Dec, */
3588
3589 unsigned int y = tm->tm_year + 1900;
3590 unsigned int m = tm->tm_mon;
3591 unsigned long days = 0;
3592
3593 /* days since 1/1/1970 for full years */
3594 days += days_since_zero(y) - days_since_zero(1970);
3595
3596 /* days for full months in the current year */
3597 days += 28 * m + ((extra >> (m * 5)) & 0x1f);
3598
3599 /* count + 1 after March for leap years. A leap year is a year multiple
3600 * of 4, unless it's multiple of 100 without being multiple of 400. 2000
3601 * is leap, 1900 isn't, 1904 is.
3602 */
3603 if ((m > 1) && !(y & 3) && ((y % 100) || !(y % 400)))
3604 days++;
3605
3606 days += tm->tm_mday - 1;
3607 return days * 86400ULL + tm->tm_hour * 3600 + tm->tm_min * 60 + tm->tm_sec;
3608}
3609
Thierry Fournier93127942016-01-20 18:49:45 +01003610/* This function check a char. It returns true and updates
3611 * <date> and <len> pointer to the new position if the
3612 * character is found.
3613 */
3614static inline int parse_expect_char(const char **date, int *len, char c)
3615{
3616 if (*len < 1 || **date != c)
3617 return 0;
3618 (*len)--;
3619 (*date)++;
3620 return 1;
3621}
3622
3623/* This function expects a string <str> of len <l>. It return true and updates.
3624 * <date> and <len> if the string matches, otherwise, it returns false.
3625 */
3626static inline int parse_strcmp(const char **date, int *len, char *str, int l)
3627{
3628 if (*len < l || strncmp(*date, str, l) != 0)
3629 return 0;
3630 (*len) -= l;
3631 (*date) += l;
3632 return 1;
3633}
3634
3635/* This macro converts 3 chars name in integer. */
3636#define STR2I3(__a, __b, __c) ((__a) * 65536 + (__b) * 256 + (__c))
3637
3638/* day-name = %x4D.6F.6E ; "Mon", case-sensitive
3639 * / %x54.75.65 ; "Tue", case-sensitive
3640 * / %x57.65.64 ; "Wed", case-sensitive
3641 * / %x54.68.75 ; "Thu", case-sensitive
3642 * / %x46.72.69 ; "Fri", case-sensitive
3643 * / %x53.61.74 ; "Sat", case-sensitive
3644 * / %x53.75.6E ; "Sun", case-sensitive
3645 *
3646 * This array must be alphabetically sorted
3647 */
3648static inline int parse_http_dayname(const char **date, int *len, struct tm *tm)
3649{
3650 if (*len < 3)
3651 return 0;
3652 switch (STR2I3((*date)[0], (*date)[1], (*date)[2])) {
3653 case STR2I3('M','o','n'): tm->tm_wday = 1; break;
3654 case STR2I3('T','u','e'): tm->tm_wday = 2; break;
3655 case STR2I3('W','e','d'): tm->tm_wday = 3; break;
3656 case STR2I3('T','h','u'): tm->tm_wday = 4; break;
3657 case STR2I3('F','r','i'): tm->tm_wday = 5; break;
3658 case STR2I3('S','a','t'): tm->tm_wday = 6; break;
3659 case STR2I3('S','u','n'): tm->tm_wday = 7; break;
3660 default: return 0;
3661 }
3662 *len -= 3;
3663 *date += 3;
3664 return 1;
3665}
3666
3667/* month = %x4A.61.6E ; "Jan", case-sensitive
3668 * / %x46.65.62 ; "Feb", case-sensitive
3669 * / %x4D.61.72 ; "Mar", case-sensitive
3670 * / %x41.70.72 ; "Apr", case-sensitive
3671 * / %x4D.61.79 ; "May", case-sensitive
3672 * / %x4A.75.6E ; "Jun", case-sensitive
3673 * / %x4A.75.6C ; "Jul", case-sensitive
3674 * / %x41.75.67 ; "Aug", case-sensitive
3675 * / %x53.65.70 ; "Sep", case-sensitive
3676 * / %x4F.63.74 ; "Oct", case-sensitive
3677 * / %x4E.6F.76 ; "Nov", case-sensitive
3678 * / %x44.65.63 ; "Dec", case-sensitive
3679 *
3680 * This array must be alphabetically sorted
3681 */
3682static inline int parse_http_monthname(const char **date, int *len, struct tm *tm)
3683{
3684 if (*len < 3)
3685 return 0;
3686 switch (STR2I3((*date)[0], (*date)[1], (*date)[2])) {
3687 case STR2I3('J','a','n'): tm->tm_mon = 0; break;
3688 case STR2I3('F','e','b'): tm->tm_mon = 1; break;
3689 case STR2I3('M','a','r'): tm->tm_mon = 2; break;
3690 case STR2I3('A','p','r'): tm->tm_mon = 3; break;
3691 case STR2I3('M','a','y'): tm->tm_mon = 4; break;
3692 case STR2I3('J','u','n'): tm->tm_mon = 5; break;
3693 case STR2I3('J','u','l'): tm->tm_mon = 6; break;
3694 case STR2I3('A','u','g'): tm->tm_mon = 7; break;
3695 case STR2I3('S','e','p'): tm->tm_mon = 8; break;
3696 case STR2I3('O','c','t'): tm->tm_mon = 9; break;
3697 case STR2I3('N','o','v'): tm->tm_mon = 10; break;
3698 case STR2I3('D','e','c'): tm->tm_mon = 11; break;
3699 default: return 0;
3700 }
3701 *len -= 3;
3702 *date += 3;
3703 return 1;
3704}
3705
3706/* day-name-l = %x4D.6F.6E.64.61.79 ; "Monday", case-sensitive
3707 * / %x54.75.65.73.64.61.79 ; "Tuesday", case-sensitive
3708 * / %x57.65.64.6E.65.73.64.61.79 ; "Wednesday", case-sensitive
3709 * / %x54.68.75.72.73.64.61.79 ; "Thursday", case-sensitive
3710 * / %x46.72.69.64.61.79 ; "Friday", case-sensitive
3711 * / %x53.61.74.75.72.64.61.79 ; "Saturday", case-sensitive
3712 * / %x53.75.6E.64.61.79 ; "Sunday", case-sensitive
3713 *
3714 * This array must be alphabetically sorted
3715 */
3716static inline int parse_http_ldayname(const char **date, int *len, struct tm *tm)
3717{
3718 if (*len < 6) /* Minimum length. */
3719 return 0;
3720 switch (STR2I3((*date)[0], (*date)[1], (*date)[2])) {
3721 case STR2I3('M','o','n'):
3722 RET0_UNLESS(parse_strcmp(date, len, "Monday", 6));
3723 tm->tm_wday = 1;
3724 return 1;
3725 case STR2I3('T','u','e'):
3726 RET0_UNLESS(parse_strcmp(date, len, "Tuesday", 7));
3727 tm->tm_wday = 2;
3728 return 1;
3729 case STR2I3('W','e','d'):
3730 RET0_UNLESS(parse_strcmp(date, len, "Wednesday", 9));
3731 tm->tm_wday = 3;
3732 return 1;
3733 case STR2I3('T','h','u'):
3734 RET0_UNLESS(parse_strcmp(date, len, "Thursday", 8));
3735 tm->tm_wday = 4;
3736 return 1;
3737 case STR2I3('F','r','i'):
3738 RET0_UNLESS(parse_strcmp(date, len, "Friday", 6));
3739 tm->tm_wday = 5;
3740 return 1;
3741 case STR2I3('S','a','t'):
3742 RET0_UNLESS(parse_strcmp(date, len, "Saturday", 8));
3743 tm->tm_wday = 6;
3744 return 1;
3745 case STR2I3('S','u','n'):
3746 RET0_UNLESS(parse_strcmp(date, len, "Sunday", 6));
3747 tm->tm_wday = 7;
3748 return 1;
3749 }
3750 return 0;
3751}
3752
3753/* This function parses exactly 1 digit and returns the numeric value in "digit". */
3754static inline int parse_digit(const char **date, int *len, int *digit)
3755{
3756 if (*len < 1 || **date < '0' || **date > '9')
3757 return 0;
3758 *digit = (**date - '0');
3759 (*date)++;
3760 (*len)--;
3761 return 1;
3762}
3763
3764/* This function parses exactly 2 digits and returns the numeric value in "digit". */
3765static inline int parse_2digit(const char **date, int *len, int *digit)
3766{
3767 int value;
3768
3769 RET0_UNLESS(parse_digit(date, len, &value));
3770 (*digit) = value * 10;
3771 RET0_UNLESS(parse_digit(date, len, &value));
3772 (*digit) += value;
3773
3774 return 1;
3775}
3776
3777/* This function parses exactly 4 digits and returns the numeric value in "digit". */
3778static inline int parse_4digit(const char **date, int *len, int *digit)
3779{
3780 int value;
3781
3782 RET0_UNLESS(parse_digit(date, len, &value));
3783 (*digit) = value * 1000;
3784
3785 RET0_UNLESS(parse_digit(date, len, &value));
3786 (*digit) += value * 100;
3787
3788 RET0_UNLESS(parse_digit(date, len, &value));
3789 (*digit) += value * 10;
3790
3791 RET0_UNLESS(parse_digit(date, len, &value));
3792 (*digit) += value;
3793
3794 return 1;
3795}
3796
3797/* time-of-day = hour ":" minute ":" second
3798 * ; 00:00:00 - 23:59:60 (leap second)
3799 *
3800 * hour = 2DIGIT
3801 * minute = 2DIGIT
3802 * second = 2DIGIT
3803 */
3804static inline int parse_http_time(const char **date, int *len, struct tm *tm)
3805{
3806 RET0_UNLESS(parse_2digit(date, len, &tm->tm_hour)); /* hour 2DIGIT */
3807 RET0_UNLESS(parse_expect_char(date, len, ':')); /* expect ":" */
3808 RET0_UNLESS(parse_2digit(date, len, &tm->tm_min)); /* min 2DIGIT */
3809 RET0_UNLESS(parse_expect_char(date, len, ':')); /* expect ":" */
3810 RET0_UNLESS(parse_2digit(date, len, &tm->tm_sec)); /* sec 2DIGIT */
3811 return 1;
3812}
3813
3814/* From RFC7231
3815 * https://tools.ietf.org/html/rfc7231#section-7.1.1.1
3816 *
3817 * IMF-fixdate = day-name "," SP date1 SP time-of-day SP GMT
3818 * ; fixed length/zone/capitalization subset of the format
3819 * ; see Section 3.3 of [RFC5322]
3820 *
3821 *
3822 * date1 = day SP month SP year
3823 * ; e.g., 02 Jun 1982
3824 *
3825 * day = 2DIGIT
3826 * year = 4DIGIT
3827 *
3828 * GMT = %x47.4D.54 ; "GMT", case-sensitive
3829 *
3830 * time-of-day = hour ":" minute ":" second
3831 * ; 00:00:00 - 23:59:60 (leap second)
3832 *
3833 * hour = 2DIGIT
3834 * minute = 2DIGIT
3835 * second = 2DIGIT
3836 *
3837 * DIGIT = decimal 0-9
3838 */
3839int parse_imf_date(const char *date, int len, struct tm *tm)
3840{
David Carlier327298c2016-11-20 10:42:38 +00003841 /* tm_gmtoff, if present, ought to be zero'ed */
3842 memset(tm, 0, sizeof(*tm));
3843
Thierry Fournier93127942016-01-20 18:49:45 +01003844 RET0_UNLESS(parse_http_dayname(&date, &len, tm)); /* day-name */
3845 RET0_UNLESS(parse_expect_char(&date, &len, ',')); /* expect "," */
3846 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3847 RET0_UNLESS(parse_2digit(&date, &len, &tm->tm_mday)); /* day 2DIGIT */
3848 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3849 RET0_UNLESS(parse_http_monthname(&date, &len, tm)); /* Month */
3850 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3851 RET0_UNLESS(parse_4digit(&date, &len, &tm->tm_year)); /* year = 4DIGIT */
3852 tm->tm_year -= 1900;
3853 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3854 RET0_UNLESS(parse_http_time(&date, &len, tm)); /* Parse time. */
3855 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3856 RET0_UNLESS(parse_strcmp(&date, &len, "GMT", 3)); /* GMT = %x47.4D.54 ; "GMT", case-sensitive */
3857 tm->tm_isdst = -1;
Thierry Fournier93127942016-01-20 18:49:45 +01003858 return 1;
3859}
3860
3861/* From RFC7231
3862 * https://tools.ietf.org/html/rfc7231#section-7.1.1.1
3863 *
3864 * rfc850-date = day-name-l "," SP date2 SP time-of-day SP GMT
3865 * date2 = day "-" month "-" 2DIGIT
3866 * ; e.g., 02-Jun-82
3867 *
3868 * day = 2DIGIT
3869 */
3870int parse_rfc850_date(const char *date, int len, struct tm *tm)
3871{
3872 int year;
3873
David Carlier327298c2016-11-20 10:42:38 +00003874 /* tm_gmtoff, if present, ought to be zero'ed */
3875 memset(tm, 0, sizeof(*tm));
3876
Thierry Fournier93127942016-01-20 18:49:45 +01003877 RET0_UNLESS(parse_http_ldayname(&date, &len, tm)); /* Read the day name */
3878 RET0_UNLESS(parse_expect_char(&date, &len, ',')); /* expect "," */
3879 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3880 RET0_UNLESS(parse_2digit(&date, &len, &tm->tm_mday)); /* day 2DIGIT */
3881 RET0_UNLESS(parse_expect_char(&date, &len, '-')); /* expect "-" */
3882 RET0_UNLESS(parse_http_monthname(&date, &len, tm)); /* Month */
3883 RET0_UNLESS(parse_expect_char(&date, &len, '-')); /* expect "-" */
3884
3885 /* year = 2DIGIT
3886 *
3887 * Recipients of a timestamp value in rfc850-(*date) format, which uses a
3888 * two-digit year, MUST interpret a timestamp that appears to be more
3889 * than 50 years in the future as representing the most recent year in
3890 * the past that had the same last two digits.
3891 */
3892 RET0_UNLESS(parse_2digit(&date, &len, &tm->tm_year));
3893
3894 /* expect SP */
3895 if (!parse_expect_char(&date, &len, ' ')) {
3896 /* Maybe we have the date with 4 digits. */
3897 RET0_UNLESS(parse_2digit(&date, &len, &year));
3898 tm->tm_year = (tm->tm_year * 100 + year) - 1900;
3899 /* expect SP */
3900 RET0_UNLESS(parse_expect_char(&date, &len, ' '));
3901 } else {
3902 /* I fix 60 as pivot: >60: +1900, <60: +2000. Note that the
3903 * tm_year is the number of year since 1900, so for +1900, we
3904 * do nothing, and for +2000, we add 100.
3905 */
3906 if (tm->tm_year <= 60)
3907 tm->tm_year += 100;
3908 }
3909
3910 RET0_UNLESS(parse_http_time(&date, &len, tm)); /* Parse time. */
3911 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3912 RET0_UNLESS(parse_strcmp(&date, &len, "GMT", 3)); /* GMT = %x47.4D.54 ; "GMT", case-sensitive */
3913 tm->tm_isdst = -1;
Thierry Fournier93127942016-01-20 18:49:45 +01003914
3915 return 1;
3916}
3917
3918/* From RFC7231
3919 * https://tools.ietf.org/html/rfc7231#section-7.1.1.1
3920 *
3921 * asctime-date = day-name SP date3 SP time-of-day SP year
3922 * date3 = month SP ( 2DIGIT / ( SP 1DIGIT ))
3923 * ; e.g., Jun 2
3924 *
3925 * HTTP-date is case sensitive. A sender MUST NOT generate additional
3926 * whitespace in an HTTP-date beyond that specifically included as SP in
3927 * the grammar.
3928 */
3929int parse_asctime_date(const char *date, int len, struct tm *tm)
3930{
David Carlier327298c2016-11-20 10:42:38 +00003931 /* tm_gmtoff, if present, ought to be zero'ed */
3932 memset(tm, 0, sizeof(*tm));
3933
Thierry Fournier93127942016-01-20 18:49:45 +01003934 RET0_UNLESS(parse_http_dayname(&date, &len, tm)); /* day-name */
3935 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3936 RET0_UNLESS(parse_http_monthname(&date, &len, tm)); /* expect month */
3937 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3938
3939 /* expect SP and 1DIGIT or 2DIGIT */
3940 if (parse_expect_char(&date, &len, ' '))
3941 RET0_UNLESS(parse_digit(&date, &len, &tm->tm_mday));
3942 else
3943 RET0_UNLESS(parse_2digit(&date, &len, &tm->tm_mday));
3944
3945 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3946 RET0_UNLESS(parse_http_time(&date, &len, tm)); /* Parse time. */
3947 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3948 RET0_UNLESS(parse_4digit(&date, &len, &tm->tm_year)); /* year = 4DIGIT */
3949 tm->tm_year -= 1900;
3950 tm->tm_isdst = -1;
Thierry Fournier93127942016-01-20 18:49:45 +01003951 return 1;
3952}
3953
3954/* From RFC7231
3955 * https://tools.ietf.org/html/rfc7231#section-7.1.1.1
3956 *
3957 * HTTP-date = IMF-fixdate / obs-date
3958 * obs-date = rfc850-date / asctime-date
3959 *
3960 * parses an HTTP date in the RFC format and is accepted
3961 * alternatives. <date> is the strinf containing the date,
3962 * len is the len of the string. <tm> is filled with the
3963 * parsed time. We must considers this time as GMT.
3964 */
3965int parse_http_date(const char *date, int len, struct tm *tm)
3966{
3967 if (parse_imf_date(date, len, tm))
3968 return 1;
3969
3970 if (parse_rfc850_date(date, len, tm))
3971 return 1;
3972
3973 if (parse_asctime_date(date, len, tm))
3974 return 1;
3975
3976 return 0;
3977}
3978
Willy Tarreau4deeb102021-01-29 10:47:52 +01003979/* print the time <ns> in a short form (exactly 7 chars) at the end of buffer
3980 * <out>. "-" is printed if the value is zero, "inf" if larger than 1000 years.
3981 * It returns the new buffer length, or 0 if it doesn't fit. The value will be
3982 * surrounded by <pfx> and <sfx> respectively if not NULL.
3983 */
3984int print_time_short(struct buffer *out, const char *pfx, uint64_t ns, const char *sfx)
3985{
3986 double val = ns; // 52 bits of mantissa keep ns accuracy over 52 days
3987 const char *unit;
3988
3989 if (!pfx)
3990 pfx = "";
3991 if (!sfx)
3992 sfx = "";
3993
3994 do {
3995 unit = " - "; if (val <= 0.0) break;
3996 unit = "ns"; if (val < 1000.0) break;
3997 unit = "us"; val /= 1000.0; if (val < 1000.0) break;
3998 unit = "ms"; val /= 1000.0; if (val < 1000.0) break;
3999 unit = "s "; val /= 1000.0; if (val < 60.0) break;
4000 unit = "m "; val /= 60.0; if (val < 60.0) break;
4001 unit = "h "; val /= 60.0; if (val < 24.0) break;
4002 unit = "d "; val /= 24.0; if (val < 365.0) break;
4003 unit = "yr"; val /= 365.0; if (val < 1000.0) break;
4004 unit = " inf "; val = 0.0; break;
4005 } while (0);
4006
4007 if (val <= 0.0)
4008 return chunk_appendf(out, "%s%7s%s", pfx, unit, sfx);
4009 else if (val < 10.0)
4010 return chunk_appendf(out, "%s%1.3f%s%s", pfx, val, unit, sfx);
4011 else if (val < 100.0)
4012 return chunk_appendf(out, "%s%2.2f%s%s", pfx, val, unit, sfx);
4013 else
4014 return chunk_appendf(out, "%s%3.1f%s%s", pfx, val, unit, sfx);
4015}
4016
Willy Tarreau9a7bea52012-04-27 11:16:50 +02004017/* Dynamically allocates a string of the proper length to hold the formatted
4018 * output. NULL is returned on error. The caller is responsible for freeing the
4019 * memory area using free(). The resulting string is returned in <out> if the
4020 * pointer is not NULL. A previous version of <out> might be used to build the
4021 * new string, and it will be freed before returning if it is not NULL, which
4022 * makes it possible to build complex strings from iterative calls without
4023 * having to care about freeing intermediate values, as in the example below :
4024 *
4025 * memprintf(&err, "invalid argument: '%s'", arg);
4026 * ...
4027 * memprintf(&err, "parser said : <%s>\n", *err);
4028 * ...
4029 * free(*err);
4030 *
4031 * This means that <err> must be initialized to NULL before first invocation.
4032 * The return value also holds the allocated string, which eases error checking
4033 * and immediate consumption. If the output pointer is not used, NULL must be
Willy Tarreaueb6cead2012-09-20 19:43:14 +02004034 * passed instead and it will be ignored. The returned message will then also
4035 * be NULL so that the caller does not have to bother with freeing anything.
Willy Tarreau9a7bea52012-04-27 11:16:50 +02004036 *
4037 * It is also convenient to use it without any free except the last one :
4038 * err = NULL;
4039 * if (!fct1(err)) report(*err);
4040 * if (!fct2(err)) report(*err);
4041 * if (!fct3(err)) report(*err);
4042 * free(*err);
Christopher Faulet93a518f2017-10-24 11:25:33 +02004043 *
4044 * memprintf relies on memvprintf. This last version can be called from any
4045 * function with variadic arguments.
Willy Tarreau9a7bea52012-04-27 11:16:50 +02004046 */
Christopher Faulet93a518f2017-10-24 11:25:33 +02004047char *memvprintf(char **out, const char *format, va_list orig_args)
Willy Tarreau9a7bea52012-04-27 11:16:50 +02004048{
4049 va_list args;
4050 char *ret = NULL;
4051 int allocated = 0;
4052 int needed = 0;
4053
Willy Tarreaueb6cead2012-09-20 19:43:14 +02004054 if (!out)
4055 return NULL;
4056
Willy Tarreau9a7bea52012-04-27 11:16:50 +02004057 do {
Willy Tarreaue0609f52019-03-29 19:13:23 +01004058 char buf1;
4059
Willy Tarreau9a7bea52012-04-27 11:16:50 +02004060 /* vsnprintf() will return the required length even when the
4061 * target buffer is NULL. We do this in a loop just in case
4062 * intermediate evaluations get wrong.
4063 */
Christopher Faulet93a518f2017-10-24 11:25:33 +02004064 va_copy(args, orig_args);
Willy Tarreaue0609f52019-03-29 19:13:23 +01004065 needed = vsnprintf(ret ? ret : &buf1, allocated, format, args);
Willy Tarreau9a7bea52012-04-27 11:16:50 +02004066 va_end(args);
Willy Tarreau1b2fed62013-04-01 22:48:54 +02004067 if (needed < allocated) {
4068 /* Note: on Solaris 8, the first iteration always
4069 * returns -1 if allocated is zero, so we force a
4070 * retry.
4071 */
4072 if (!allocated)
4073 needed = 0;
4074 else
4075 break;
4076 }
Willy Tarreau9a7bea52012-04-27 11:16:50 +02004077
Willy Tarreau1b2fed62013-04-01 22:48:54 +02004078 allocated = needed + 1;
Hubert Verstraete831962e2016-06-28 22:44:26 +02004079 ret = my_realloc2(ret, allocated);
Willy Tarreau9a7bea52012-04-27 11:16:50 +02004080 } while (ret);
4081
4082 if (needed < 0) {
4083 /* an error was encountered */
Willy Tarreau61cfdf42021-02-20 10:46:51 +01004084 ha_free(&ret);
Willy Tarreau9a7bea52012-04-27 11:16:50 +02004085 }
4086
4087 if (out) {
4088 free(*out);
4089 *out = ret;
4090 }
4091
4092 return ret;
4093}
William Lallemand421f5b52012-02-06 18:15:57 +01004094
Christopher Faulet93a518f2017-10-24 11:25:33 +02004095char *memprintf(char **out, const char *format, ...)
4096{
4097 va_list args;
4098 char *ret = NULL;
4099
4100 va_start(args, format);
4101 ret = memvprintf(out, format, args);
4102 va_end(args);
4103
4104 return ret;
4105}
4106
Willy Tarreau21c705b2012-09-14 11:40:36 +02004107/* Used to add <level> spaces before each line of <out>, unless there is only one line.
4108 * The input argument is automatically freed and reassigned. The result will have to be
Willy Tarreau70eec382012-10-10 08:56:47 +02004109 * freed by the caller. It also supports being passed a NULL which results in the same
4110 * output.
Willy Tarreau21c705b2012-09-14 11:40:36 +02004111 * Example of use :
4112 * parse(cmd, &err); (callee: memprintf(&err, ...))
4113 * fprintf(stderr, "Parser said: %s\n", indent_error(&err));
4114 * free(err);
4115 */
4116char *indent_msg(char **out, int level)
4117{
4118 char *ret, *in, *p;
4119 int needed = 0;
4120 int lf = 0;
4121 int lastlf = 0;
4122 int len;
4123
Willy Tarreau70eec382012-10-10 08:56:47 +02004124 if (!out || !*out)
4125 return NULL;
4126
Willy Tarreau21c705b2012-09-14 11:40:36 +02004127 in = *out - 1;
4128 while ((in = strchr(in + 1, '\n')) != NULL) {
4129 lastlf = in - *out;
4130 lf++;
4131 }
4132
4133 if (!lf) /* single line, no LF, return it as-is */
4134 return *out;
4135
4136 len = strlen(*out);
4137
4138 if (lf == 1 && lastlf == len - 1) {
4139 /* single line, LF at end, strip it and return as-is */
4140 (*out)[lastlf] = 0;
4141 return *out;
4142 }
4143
4144 /* OK now we have at least one LF, we need to process the whole string
4145 * as a multi-line string. What we'll do :
4146 * - prefix with an LF if there is none
4147 * - add <level> spaces before each line
4148 * This means at most ( 1 + level + (len-lf) + lf*<1+level) ) =
4149 * 1 + level + len + lf * level = 1 + level * (lf + 1) + len.
4150 */
4151
4152 needed = 1 + level * (lf + 1) + len + 1;
4153 p = ret = malloc(needed);
4154 in = *out;
4155
4156 /* skip initial LFs */
4157 while (*in == '\n')
4158 in++;
4159
4160 /* copy each line, prefixed with LF and <level> spaces, and without the trailing LF */
4161 while (*in) {
4162 *p++ = '\n';
4163 memset(p, ' ', level);
4164 p += level;
4165 do {
4166 *p++ = *in++;
4167 } while (*in && *in != '\n');
4168 if (*in)
4169 in++;
4170 }
4171 *p = 0;
4172
4173 free(*out);
4174 *out = ret;
4175
4176 return ret;
4177}
4178
Willy Tarreaua2c99112019-08-21 13:17:37 +02004179/* makes a copy of message <in> into <out>, with each line prefixed with <pfx>
4180 * and end of lines replaced with <eol> if not 0. The first line to indent has
4181 * to be indicated in <first> (starts at zero), so that it is possible to skip
4182 * indenting the first line if it has to be appended after an existing message.
4183 * Empty strings are never indented, and NULL strings are considered empty both
4184 * for <in> and <pfx>. It returns non-zero if an EOL was appended as the last
4185 * character, non-zero otherwise.
4186 */
4187int append_prefixed_str(struct buffer *out, const char *in, const char *pfx, char eol, int first)
4188{
4189 int bol, lf;
4190 int pfxlen = pfx ? strlen(pfx) : 0;
4191
4192 if (!in)
4193 return 0;
4194
4195 bol = 1;
4196 lf = 0;
4197 while (*in) {
4198 if (bol && pfxlen) {
4199 if (first > 0)
4200 first--;
4201 else
4202 b_putblk(out, pfx, pfxlen);
4203 bol = 0;
4204 }
4205
4206 lf = (*in == '\n');
4207 bol |= lf;
4208 b_putchr(out, (lf && eol) ? eol : *in);
4209 in++;
4210 }
4211 return lf;
4212}
4213
Willy Tarreau9d22e562019-03-29 18:49:09 +01004214/* removes environment variable <name> from the environment as found in
4215 * environ. This is only provided as an alternative for systems without
4216 * unsetenv() (old Solaris and AIX versions). THIS IS NOT THREAD SAFE.
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05004217 * The principle is to scan environ for each occurrence of variable name
Willy Tarreau9d22e562019-03-29 18:49:09 +01004218 * <name> and to replace the matching pointers with the last pointer of
4219 * the array (since variables are not ordered).
4220 * It always returns 0 (success).
4221 */
4222int my_unsetenv(const char *name)
4223{
4224 extern char **environ;
4225 char **p = environ;
4226 int vars;
4227 int next;
4228 int len;
4229
4230 len = strlen(name);
4231 for (vars = 0; p[vars]; vars++)
4232 ;
4233 next = 0;
4234 while (next < vars) {
4235 if (strncmp(p[next], name, len) != 0 || p[next][len] != '=') {
4236 next++;
4237 continue;
4238 }
4239 if (next < vars - 1)
4240 p[next] = p[vars - 1];
4241 p[--vars] = NULL;
4242 }
4243 return 0;
4244}
4245
Willy Tarreaudad36a32013-03-11 01:20:04 +01004246/* Convert occurrences of environment variables in the input string to their
4247 * corresponding value. A variable is identified as a series of alphanumeric
4248 * characters or underscores following a '$' sign. The <in> string must be
4249 * free()able. NULL returns NULL. The resulting string might be reallocated if
4250 * some expansion is made. Variable names may also be enclosed into braces if
4251 * needed (eg: to concatenate alphanum characters).
4252 */
4253char *env_expand(char *in)
4254{
4255 char *txt_beg;
4256 char *out;
4257 char *txt_end;
4258 char *var_beg;
4259 char *var_end;
4260 char *value;
4261 char *next;
4262 int out_len;
4263 int val_len;
4264
4265 if (!in)
4266 return in;
4267
4268 value = out = NULL;
4269 out_len = 0;
4270
4271 txt_beg = in;
4272 do {
4273 /* look for next '$' sign in <in> */
4274 for (txt_end = txt_beg; *txt_end && *txt_end != '$'; txt_end++);
4275
4276 if (!*txt_end && !out) /* end and no expansion performed */
4277 return in;
4278
4279 val_len = 0;
4280 next = txt_end;
4281 if (*txt_end == '$') {
4282 char save;
4283
4284 var_beg = txt_end + 1;
4285 if (*var_beg == '{')
4286 var_beg++;
4287
4288 var_end = var_beg;
Willy Tarreau90807112020-02-25 08:16:33 +01004289 while (isalnum((unsigned char)*var_end) || *var_end == '_') {
Willy Tarreaudad36a32013-03-11 01:20:04 +01004290 var_end++;
4291 }
4292
4293 next = var_end;
4294 if (*var_end == '}' && (var_beg > txt_end + 1))
4295 next++;
4296
4297 /* get value of the variable name at this location */
4298 save = *var_end;
4299 *var_end = '\0';
4300 value = getenv(var_beg);
4301 *var_end = save;
4302 val_len = value ? strlen(value) : 0;
4303 }
4304
Hubert Verstraete831962e2016-06-28 22:44:26 +02004305 out = my_realloc2(out, out_len + (txt_end - txt_beg) + val_len + 1);
Willy Tarreaudad36a32013-03-11 01:20:04 +01004306 if (txt_end > txt_beg) {
4307 memcpy(out + out_len, txt_beg, txt_end - txt_beg);
4308 out_len += txt_end - txt_beg;
4309 }
4310 if (val_len) {
4311 memcpy(out + out_len, value, val_len);
4312 out_len += val_len;
4313 }
4314 out[out_len] = 0;
4315 txt_beg = next;
4316 } while (*txt_beg);
4317
4318 /* here we know that <out> was allocated and that we don't need <in> anymore */
4319 free(in);
4320 return out;
4321}
4322
de Lafond Guillaume88c278f2013-04-15 19:27:10 +02004323
4324/* same as strstr() but case-insensitive and with limit length */
4325const char *strnistr(const char *str1, int len_str1, const char *str2, int len_str2)
4326{
4327 char *pptr, *sptr, *start;
Willy Tarreauc8746532014-05-28 23:05:07 +02004328 unsigned int slen, plen;
4329 unsigned int tmp1, tmp2;
de Lafond Guillaume88c278f2013-04-15 19:27:10 +02004330
4331 if (str1 == NULL || len_str1 == 0) // search pattern into an empty string => search is not found
4332 return NULL;
4333
4334 if (str2 == NULL || len_str2 == 0) // pattern is empty => every str1 match
4335 return str1;
4336
4337 if (len_str1 < len_str2) // pattern is longer than string => search is not found
4338 return NULL;
4339
4340 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 +02004341 while (toupper((unsigned char)*start) != toupper((unsigned char)*str2)) {
de Lafond Guillaume88c278f2013-04-15 19:27:10 +02004342 start++;
4343 slen--;
4344 tmp1++;
4345
4346 if (tmp1 >= len_str1)
4347 return NULL;
4348
4349 /* if pattern longer than string */
4350 if (slen < plen)
4351 return NULL;
4352 }
4353
4354 sptr = start;
4355 pptr = (char *)str2;
4356
4357 tmp2 = 0;
Willy Tarreauf278eec2020-07-05 21:46:32 +02004358 while (toupper((unsigned char)*sptr) == toupper((unsigned char)*pptr)) {
de Lafond Guillaume88c278f2013-04-15 19:27:10 +02004359 sptr++;
4360 pptr++;
4361 tmp2++;
4362
4363 if (*pptr == '\0' || tmp2 == len_str2) /* end of pattern found */
4364 return start;
4365 if (*sptr == '\0' || tmp2 == len_str1) /* end of string found and the pattern is not fully found */
4366 return NULL;
4367 }
4368 }
4369 return NULL;
4370}
4371
Willy Tarreau3ff476e2022-03-30 10:02:56 +02004372/* Returns true if s1 < s2 < s3 otherwise zero. Both s1 and s3 may be NULL and
4373 * in this case only non-null strings are compared. This allows to pass initial
4374 * values in iterators and in sort functions.
4375 */
4376int strordered(const char *s1, const char *s2, const char *s3)
4377{
4378 return (!s1 || strcmp(s1, s2) < 0) && (!s3 || strcmp(s2, s3) < 0);
4379}
4380
Thierry FOURNIER317e1c42014-08-12 10:20:47 +02004381/* This function read the next valid utf8 char.
4382 * <s> is the byte srray to be decode, <len> is its length.
4383 * The function returns decoded char encoded like this:
4384 * The 4 msb are the return code (UTF8_CODE_*), the 4 lsb
4385 * are the length read. The decoded character is stored in <c>.
4386 */
4387unsigned char utf8_next(const char *s, int len, unsigned int *c)
4388{
4389 const unsigned char *p = (unsigned char *)s;
4390 int dec;
4391 unsigned char code = UTF8_CODE_OK;
4392
4393 if (len < 1)
4394 return UTF8_CODE_OK;
4395
4396 /* Check the type of UTF8 sequence
4397 *
4398 * 0... .... 0x00 <= x <= 0x7f : 1 byte: ascii char
4399 * 10.. .... 0x80 <= x <= 0xbf : invalid sequence
4400 * 110. .... 0xc0 <= x <= 0xdf : 2 bytes
4401 * 1110 .... 0xe0 <= x <= 0xef : 3 bytes
4402 * 1111 0... 0xf0 <= x <= 0xf7 : 4 bytes
4403 * 1111 10.. 0xf8 <= x <= 0xfb : 5 bytes
4404 * 1111 110. 0xfc <= x <= 0xfd : 6 bytes
4405 * 1111 111. 0xfe <= x <= 0xff : invalid sequence
4406 */
4407 switch (*p) {
4408 case 0x00 ... 0x7f:
4409 *c = *p;
4410 return UTF8_CODE_OK | 1;
4411
4412 case 0x80 ... 0xbf:
4413 *c = *p;
4414 return UTF8_CODE_BADSEQ | 1;
4415
4416 case 0xc0 ... 0xdf:
4417 if (len < 2) {
4418 *c = *p;
4419 return UTF8_CODE_BADSEQ | 1;
4420 }
4421 *c = *p & 0x1f;
4422 dec = 1;
4423 break;
4424
4425 case 0xe0 ... 0xef:
4426 if (len < 3) {
4427 *c = *p;
4428 return UTF8_CODE_BADSEQ | 1;
4429 }
4430 *c = *p & 0x0f;
4431 dec = 2;
4432 break;
4433
4434 case 0xf0 ... 0xf7:
4435 if (len < 4) {
4436 *c = *p;
4437 return UTF8_CODE_BADSEQ | 1;
4438 }
4439 *c = *p & 0x07;
4440 dec = 3;
4441 break;
4442
4443 case 0xf8 ... 0xfb:
4444 if (len < 5) {
4445 *c = *p;
4446 return UTF8_CODE_BADSEQ | 1;
4447 }
4448 *c = *p & 0x03;
4449 dec = 4;
4450 break;
4451
4452 case 0xfc ... 0xfd:
4453 if (len < 6) {
4454 *c = *p;
4455 return UTF8_CODE_BADSEQ | 1;
4456 }
4457 *c = *p & 0x01;
4458 dec = 5;
4459 break;
4460
4461 case 0xfe ... 0xff:
4462 default:
4463 *c = *p;
4464 return UTF8_CODE_BADSEQ | 1;
4465 }
4466
4467 p++;
4468
4469 while (dec > 0) {
4470
4471 /* need 0x10 for the 2 first bits */
4472 if ( ( *p & 0xc0 ) != 0x80 )
4473 return UTF8_CODE_BADSEQ | ((p-(unsigned char *)s)&0xffff);
4474
4475 /* add data at char */
4476 *c = ( *c << 6 ) | ( *p & 0x3f );
4477
4478 dec--;
4479 p++;
4480 }
4481
4482 /* Check ovelong encoding.
4483 * 1 byte : 5 + 6 : 11 : 0x80 ... 0x7ff
4484 * 2 bytes : 4 + 6 + 6 : 16 : 0x800 ... 0xffff
4485 * 3 bytes : 3 + 6 + 6 + 6 : 21 : 0x10000 ... 0x1fffff
4486 */
Thierry FOURNIER9e7ec082015-03-12 19:32:38 +01004487 if (( *c <= 0x7f && (p-(unsigned char *)s) > 1) ||
Thierry FOURNIER317e1c42014-08-12 10:20:47 +02004488 (*c >= 0x80 && *c <= 0x7ff && (p-(unsigned char *)s) > 2) ||
4489 (*c >= 0x800 && *c <= 0xffff && (p-(unsigned char *)s) > 3) ||
4490 (*c >= 0x10000 && *c <= 0x1fffff && (p-(unsigned char *)s) > 4))
4491 code |= UTF8_CODE_OVERLONG;
4492
4493 /* Check invalid UTF8 range. */
4494 if ((*c >= 0xd800 && *c <= 0xdfff) ||
4495 (*c >= 0xfffe && *c <= 0xffff))
4496 code |= UTF8_CODE_INVRANGE;
4497
4498 return code | ((p-(unsigned char *)s)&0x0f);
4499}
4500
Maxime de Roucydc887852016-05-13 23:52:54 +02004501/* append a copy of string <str> (in a wordlist) at the end of the list <li>
4502 * On failure : return 0 and <err> filled with an error message.
4503 * The caller is responsible for freeing the <err> and <str> copy
4504 * memory area using free()
4505 */
4506int list_append_word(struct list *li, const char *str, char **err)
4507{
4508 struct wordlist *wl;
4509
4510 wl = calloc(1, sizeof(*wl));
4511 if (!wl) {
4512 memprintf(err, "out of memory");
4513 goto fail_wl;
4514 }
4515
4516 wl->s = strdup(str);
4517 if (!wl->s) {
4518 memprintf(err, "out of memory");
4519 goto fail_wl_s;
4520 }
4521
Willy Tarreau2b718102021-04-21 07:32:39 +02004522 LIST_APPEND(li, &wl->list);
Maxime de Roucydc887852016-05-13 23:52:54 +02004523
4524 return 1;
4525
4526fail_wl_s:
4527 free(wl->s);
4528fail_wl:
4529 free(wl);
4530 return 0;
4531}
4532
Willy Tarreau37101052019-05-20 16:48:20 +02004533/* indicates if a memory location may safely be read or not. The trick consists
4534 * in performing a harmless syscall using this location as an input and letting
4535 * the operating system report whether it's OK or not. For this we have the
4536 * stat() syscall, which will return EFAULT when the memory location supposed
4537 * to contain the file name is not readable. If it is readable it will then
4538 * either return 0 if the area contains an existing file name, or -1 with
4539 * another code. This must not be abused, and some audit systems might detect
4540 * this as abnormal activity. It's used only for unsafe dumps.
4541 */
4542int may_access(const void *ptr)
4543{
4544 struct stat buf;
4545
4546 if (stat(ptr, &buf) == 0)
4547 return 1;
4548 if (errno == EFAULT)
4549 return 0;
4550 return 1;
4551}
4552
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004553/* print a string of text buffer to <out>. The format is :
4554 * Non-printable chars \t, \n, \r and \e are * encoded in C format.
4555 * Other non-printable chars are encoded "\xHH". Space, '\', and '=' are also escaped.
4556 * Print stopped if null char or <bsize> is reached, or if no more place in the chunk.
4557 */
Willy Tarreau83061a82018-07-13 11:56:34 +02004558int dump_text(struct buffer *out, const char *buf, int bsize)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004559{
4560 unsigned char c;
Tim Duesterhus18795d42021-08-29 00:58:22 +02004561 size_t ptr = 0;
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004562
Tim Duesterhus18795d42021-08-29 00:58:22 +02004563 while (ptr < bsize && buf[ptr]) {
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004564 c = buf[ptr];
Willy Tarreau90807112020-02-25 08:16:33 +01004565 if (isprint((unsigned char)c) && isascii((unsigned char)c) && c != '\\' && c != ' ' && c != '=') {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004566 if (out->data > out->size - 1)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004567 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004568 out->area[out->data++] = c;
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004569 }
4570 else if (c == '\t' || c == '\n' || c == '\r' || c == '\e' || c == '\\' || c == ' ' || c == '=') {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004571 if (out->data > out->size - 2)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004572 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004573 out->area[out->data++] = '\\';
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004574 switch (c) {
4575 case ' ': c = ' '; break;
4576 case '\t': c = 't'; break;
4577 case '\n': c = 'n'; break;
4578 case '\r': c = 'r'; break;
4579 case '\e': c = 'e'; break;
4580 case '\\': c = '\\'; break;
4581 case '=': c = '='; break;
4582 }
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004583 out->area[out->data++] = c;
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004584 }
4585 else {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004586 if (out->data > out->size - 4)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004587 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004588 out->area[out->data++] = '\\';
4589 out->area[out->data++] = 'x';
4590 out->area[out->data++] = hextab[(c >> 4) & 0xF];
4591 out->area[out->data++] = hextab[c & 0xF];
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004592 }
4593 ptr++;
4594 }
4595
4596 return ptr;
4597}
4598
4599/* print a buffer in hexa.
4600 * Print stopped if <bsize> is reached, or if no more place in the chunk.
4601 */
Willy Tarreau83061a82018-07-13 11:56:34 +02004602int dump_binary(struct buffer *out, const char *buf, int bsize)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004603{
4604 unsigned char c;
4605 int ptr = 0;
4606
4607 while (ptr < bsize) {
4608 c = buf[ptr];
4609
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004610 if (out->data > out->size - 2)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004611 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004612 out->area[out->data++] = hextab[(c >> 4) & 0xF];
4613 out->area[out->data++] = hextab[c & 0xF];
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004614
4615 ptr++;
4616 }
4617 return ptr;
4618}
4619
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004620/* Appends into buffer <out> a hex dump of memory area <buf> for <len> bytes,
4621 * prepending each line with prefix <pfx>. The output is *not* initialized.
4622 * The output will not wrap pas the buffer's end so it is more optimal if the
4623 * caller makes sure the buffer is aligned first. A trailing zero will always
4624 * be appended (and not counted) if there is room for it. The caller must make
Willy Tarreau37101052019-05-20 16:48:20 +02004625 * sure that the area is dumpable first. If <unsafe> is non-null, the memory
4626 * locations are checked first for being readable.
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004627 */
Willy Tarreau37101052019-05-20 16:48:20 +02004628void dump_hex(struct buffer *out, const char *pfx, const void *buf, int len, int unsafe)
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004629{
4630 const unsigned char *d = buf;
4631 int i, j, start;
4632
4633 d = (const unsigned char *)(((unsigned long)buf) & -16);
4634 start = ((unsigned long)buf) & 15;
4635
4636 for (i = 0; i < start + len; i += 16) {
4637 chunk_appendf(out, (sizeof(void *) == 4) ? "%s%8p: " : "%s%16p: ", pfx, d + i);
4638
Willy Tarreau37101052019-05-20 16:48:20 +02004639 // 0: unchecked, 1: checked safe, 2: danger
4640 unsafe = !!unsafe;
4641 if (unsafe && !may_access(d + i))
4642 unsafe = 2;
4643
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004644 for (j = 0; j < 16; j++) {
Willy Tarreau37101052019-05-20 16:48:20 +02004645 if ((i + j < start) || (i + j >= start + len))
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004646 chunk_strcat(out, "'' ");
Willy Tarreau37101052019-05-20 16:48:20 +02004647 else if (unsafe > 1)
4648 chunk_strcat(out, "** ");
4649 else
4650 chunk_appendf(out, "%02x ", d[i + j]);
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004651
4652 if (j == 7)
4653 chunk_strcat(out, "- ");
4654 }
4655 chunk_strcat(out, " ");
4656 for (j = 0; j < 16; j++) {
Willy Tarreau37101052019-05-20 16:48:20 +02004657 if ((i + j < start) || (i + j >= start + len))
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004658 chunk_strcat(out, "'");
Willy Tarreau37101052019-05-20 16:48:20 +02004659 else if (unsafe > 1)
4660 chunk_strcat(out, "*");
Willy Tarreau90807112020-02-25 08:16:33 +01004661 else if (isprint((unsigned char)d[i + j]))
Willy Tarreau37101052019-05-20 16:48:20 +02004662 chunk_appendf(out, "%c", d[i + j]);
4663 else
4664 chunk_strcat(out, ".");
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004665 }
4666 chunk_strcat(out, "\n");
4667 }
4668}
4669
Willy Tarreau762fb3e2020-03-03 15:57:10 +01004670/* dumps <pfx> followed by <n> bytes from <addr> in hex form into buffer <buf>
4671 * enclosed in brackets after the address itself, formatted on 14 chars
4672 * including the "0x" prefix. This is meant to be used as a prefix for code
4673 * areas. For example:
4674 * "0x7f10b6557690 [48 c7 c0 0f 00 00 00 0f]"
4675 * It relies on may_access() to know if the bytes are dumpable, otherwise "--"
4676 * is emitted. A NULL <pfx> will be considered empty.
4677 */
4678void dump_addr_and_bytes(struct buffer *buf, const char *pfx, const void *addr, int n)
4679{
4680 int ok = 0;
4681 int i;
4682
4683 chunk_appendf(buf, "%s%#14lx [", pfx ? pfx : "", (long)addr);
4684
4685 for (i = 0; i < n; i++) {
4686 if (i == 0 || (((long)(addr + i) ^ (long)(addr)) & 4096))
4687 ok = may_access(addr + i);
4688 if (ok)
4689 chunk_appendf(buf, "%02x%s", ((uint8_t*)addr)[i], (i<n-1) ? " " : "]");
4690 else
4691 chunk_appendf(buf, "--%s", (i<n-1) ? " " : "]");
4692 }
4693}
4694
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004695/* print a line of text buffer (limited to 70 bytes) to <out>. The format is :
4696 * <2 spaces> <offset=5 digits> <space or plus> <space> <70 chars max> <\n>
4697 * which is 60 chars per line. Non-printable chars \t, \n, \r and \e are
4698 * encoded in C format. Other non-printable chars are encoded "\xHH". Original
4699 * lines are respected within the limit of 70 output chars. Lines that are
4700 * continuation of a previous truncated line begin with "+" instead of " "
4701 * after the offset. The new pointer is returned.
4702 */
Willy Tarreau83061a82018-07-13 11:56:34 +02004703int dump_text_line(struct buffer *out, const char *buf, int bsize, int len,
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004704 int *line, int ptr)
4705{
4706 int end;
4707 unsigned char c;
4708
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004709 end = out->data + 80;
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004710 if (end > out->size)
4711 return ptr;
4712
4713 chunk_appendf(out, " %05d%c ", ptr, (ptr == *line) ? ' ' : '+');
4714
4715 while (ptr < len && ptr < bsize) {
4716 c = buf[ptr];
Willy Tarreau90807112020-02-25 08:16:33 +01004717 if (isprint((unsigned char)c) && isascii((unsigned char)c) && c != '\\') {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004718 if (out->data > end - 2)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004719 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004720 out->area[out->data++] = c;
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004721 } else if (c == '\t' || c == '\n' || c == '\r' || c == '\e' || c == '\\') {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004722 if (out->data > end - 3)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004723 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004724 out->area[out->data++] = '\\';
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004725 switch (c) {
4726 case '\t': c = 't'; break;
4727 case '\n': c = 'n'; break;
4728 case '\r': c = 'r'; break;
4729 case '\e': c = 'e'; break;
4730 case '\\': c = '\\'; break;
4731 }
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004732 out->area[out->data++] = c;
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004733 } else {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004734 if (out->data > end - 5)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004735 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004736 out->area[out->data++] = '\\';
4737 out->area[out->data++] = 'x';
4738 out->area[out->data++] = hextab[(c >> 4) & 0xF];
4739 out->area[out->data++] = hextab[c & 0xF];
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004740 }
4741 if (buf[ptr++] == '\n') {
4742 /* we had a line break, let's return now */
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004743 out->area[out->data++] = '\n';
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004744 *line = ptr;
4745 return ptr;
4746 }
4747 }
4748 /* we have an incomplete line, we return it as-is */
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004749 out->area[out->data++] = '\n';
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004750 return ptr;
4751}
4752
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004753/* displays a <len> long memory block at <buf>, assuming first byte of <buf>
Willy Tarreaued936c52017-04-27 18:03:20 +02004754 * has address <baseaddr>. String <pfx> may be placed as a prefix in front of
4755 * each line. It may be NULL if unused. The output is emitted to file <out>.
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004756 */
Willy Tarreaued936c52017-04-27 18:03:20 +02004757void debug_hexdump(FILE *out, const char *pfx, const char *buf,
4758 unsigned int baseaddr, int len)
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004759{
Willy Tarreau73459792017-04-11 07:58:08 +02004760 unsigned int i;
4761 int b, j;
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004762
4763 for (i = 0; i < (len + (baseaddr & 15)); i += 16) {
4764 b = i - (baseaddr & 15);
Willy Tarreaued936c52017-04-27 18:03:20 +02004765 fprintf(out, "%s%08x: ", pfx ? pfx : "", i + (baseaddr & ~15));
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004766 for (j = 0; j < 8; j++) {
4767 if (b + j >= 0 && b + j < len)
4768 fprintf(out, "%02x ", (unsigned char)buf[b + j]);
4769 else
4770 fprintf(out, " ");
4771 }
4772
4773 if (b + j >= 0 && b + j < len)
4774 fputc('-', out);
4775 else
4776 fputc(' ', out);
4777
4778 for (j = 8; j < 16; j++) {
4779 if (b + j >= 0 && b + j < len)
4780 fprintf(out, " %02x", (unsigned char)buf[b + j]);
4781 else
4782 fprintf(out, " ");
4783 }
4784
4785 fprintf(out, " ");
4786 for (j = 0; j < 16; j++) {
4787 if (b + j >= 0 && b + j < len) {
4788 if (isprint((unsigned char)buf[b + j]))
4789 fputc((unsigned char)buf[b + j], out);
4790 else
4791 fputc('.', out);
4792 }
4793 else
4794 fputc(' ', out);
4795 }
4796 fputc('\n', out);
4797 }
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004798}
4799
Willy Tarreaubb869862020-04-16 10:52:41 +02004800/* Tries to report the executable path name on platforms supporting this. If
4801 * not found or not possible, returns NULL.
4802 */
4803const char *get_exec_path()
4804{
4805 const char *ret = NULL;
4806
David Carlier43a56852022-03-04 15:50:48 +00004807#if defined(__linux__) && defined(__GLIBC__) && (__GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 16))
Willy Tarreaubb869862020-04-16 10:52:41 +02004808 long execfn = getauxval(AT_EXECFN);
4809
4810 if (execfn && execfn != ENOENT)
4811 ret = (const char *)execfn;
devnexen@gmail.comc4e52322021-08-17 12:55:49 +01004812#elif defined(__FreeBSD__)
4813 Elf_Auxinfo *auxv;
4814 for (auxv = __elf_aux_vector; auxv->a_type != AT_NULL; ++auxv) {
4815 if (auxv->a_type == AT_EXECPATH) {
4816 ret = (const char *)auxv->a_un.a_ptr;
4817 break;
4818 }
4819 }
David Carlierbd2cced2021-08-17 08:44:25 +01004820#elif defined(__NetBSD__)
4821 AuxInfo *auxv;
4822 for (auxv = _dlauxinfo(); auxv->a_type != AT_NULL; ++auxv) {
4823 if (auxv->a_type == AT_SUN_EXECNAME) {
4824 ret = (const char *)auxv->a_v;
4825 break;
4826 }
4827 }
David Carlier7198c702022-05-14 17:15:49 +01004828#elif defined(__sun)
4829 ret = getexecname();
Willy Tarreaubb869862020-04-16 10:52:41 +02004830#endif
4831 return ret;
4832}
4833
Baruch Siache1651b22020-07-24 07:52:20 +03004834#if (defined(__ELF__) && !defined(__linux__)) || defined(USE_DL)
Willy Tarreau9133e482020-03-04 10:19:36 +01004835/* calls dladdr() or dladdr1() on <addr> and <dli>. If dladdr1 is available,
4836 * also returns the symbol size in <size>, otherwise returns 0 there.
4837 */
4838static int dladdr_and_size(const void *addr, Dl_info *dli, size_t *size)
4839{
4840 int ret;
Willy Tarreau7b2108c2021-08-30 10:15:35 +02004841#if defined(__GLIBC__) && (__GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 3)) // most detailed one
Willy Tarreauf3d5c4b2022-01-28 09:42:29 +01004842 const ElfW(Sym) *sym __attribute__((may_alias));
Willy Tarreau9133e482020-03-04 10:19:36 +01004843
4844 ret = dladdr1(addr, dli, (void **)&sym, RTLD_DL_SYMENT);
4845 if (ret)
4846 *size = sym ? sym->st_size : 0;
4847#else
David Carlierae5c42f2021-12-31 08:15:29 +00004848#if defined(__sun)
4849 ret = dladdr((void *)addr, dli);
4850#else
Willy Tarreau9133e482020-03-04 10:19:36 +01004851 ret = dladdr(addr, dli);
David Carlierae5c42f2021-12-31 08:15:29 +00004852#endif
Willy Tarreau9133e482020-03-04 10:19:36 +01004853 *size = 0;
4854#endif
4855 return ret;
4856}
Willy Tarreau64192392021-05-05 09:06:21 +02004857
4858/* Tries to retrieve the address of the first occurrence symbol <name>.
4859 * Note that NULL in return is not always an error as a symbol may have that
4860 * address in special situations.
4861 */
4862void *get_sym_curr_addr(const char *name)
4863{
4864 void *ptr = NULL;
4865
4866#ifdef RTLD_DEFAULT
4867 ptr = dlsym(RTLD_DEFAULT, name);
4868#endif
4869 return ptr;
4870}
4871
4872
4873/* Tries to retrieve the address of the next occurrence of symbol <name>
4874 * Note that NULL in return is not always an error as a symbol may have that
4875 * address in special situations.
4876 */
4877void *get_sym_next_addr(const char *name)
4878{
4879 void *ptr = NULL;
4880
4881#ifdef RTLD_NEXT
4882 ptr = dlsym(RTLD_NEXT, name);
Willy Tarreau9133e482020-03-04 10:19:36 +01004883#endif
Willy Tarreau64192392021-05-05 09:06:21 +02004884 return ptr;
4885}
4886
4887#else /* elf & linux & dl */
4888
4889/* no possible resolving on other platforms at the moment */
4890void *get_sym_curr_addr(const char *name)
4891{
4892 return NULL;
4893}
4894
4895void *get_sym_next_addr(const char *name)
4896{
4897 return NULL;
4898}
4899
4900#endif /* elf & linux & dl */
Willy Tarreau9133e482020-03-04 10:19:36 +01004901
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004902/* Tries to append to buffer <buf> some indications about the symbol at address
4903 * <addr> using the following form:
4904 * lib:+0xoffset (unresolvable address from lib's base)
4905 * main+0xoffset (unresolvable address from main (+/-))
4906 * lib:main+0xoffset (unresolvable lib address from main (+/-))
4907 * name (resolved exact exec address)
4908 * lib:name (resolved exact lib address)
4909 * name+0xoffset/0xsize (resolved address within exec symbol)
4910 * lib:name+0xoffset/0xsize (resolved address within lib symbol)
4911 *
4912 * The file name (lib or executable) is limited to what lies between the last
4913 * '/' and the first following '.'. An optional prefix <pfx> is prepended before
4914 * 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 +03004915 * that contains the "main" symbol, or when __ELF__ && USE_DL are not set.
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004916 *
4917 * The symbol's base address is returned, or NULL when unresolved, in order to
4918 * allow the caller to match it against known ones.
4919 */
Willy Tarreau45fd1032021-01-20 14:37:59 +01004920const void *resolve_sym_name(struct buffer *buf, const char *pfx, const void *addr)
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004921{
4922 const struct {
4923 const void *func;
4924 const char *name;
4925 } fcts[] = {
4926 { .func = process_stream, .name = "process_stream" },
4927 { .func = task_run_applet, .name = "task_run_applet" },
Christopher Faulet4a7764a2022-04-01 16:58:52 +02004928 { .func = cs_conn_io_cb, .name = "cs_conn_io_cb" },
Willy Tarreau586f71b2020-12-11 15:54:36 +01004929 { .func = sock_conn_iocb, .name = "sock_conn_iocb" },
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004930 { .func = dgram_fd_handler, .name = "dgram_fd_handler" },
4931 { .func = listener_accept, .name = "listener_accept" },
Willy Tarreaud597ec22021-01-29 14:29:06 +01004932 { .func = manage_global_listener_queue, .name = "manage_global_listener_queue" },
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004933 { .func = poller_pipe_io_handler, .name = "poller_pipe_io_handler" },
4934 { .func = mworker_accept_wrapper, .name = "mworker_accept_wrapper" },
Willy Tarreau02922e12021-01-29 12:27:57 +01004935 { .func = session_expire_embryonic, .name = "session_expire_embryonic" },
Willy Tarreaufb5401f2021-01-29 12:25:23 +01004936#ifdef USE_THREAD
4937 { .func = accept_queue_process, .name = "accept_queue_process" },
4938#endif
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004939#ifdef USE_LUA
4940 { .func = hlua_process_task, .name = "hlua_process_task" },
4941#endif
Ilya Shipitsinbdec3ba2020-11-14 01:56:34 +05004942#ifdef SSL_MODE_ASYNC
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004943 { .func = ssl_async_fd_free, .name = "ssl_async_fd_free" },
4944 { .func = ssl_async_fd_handler, .name = "ssl_async_fd_handler" },
4945#endif
4946 };
4947
Baruch Siache1651b22020-07-24 07:52:20 +03004948#if (defined(__ELF__) && !defined(__linux__)) || defined(USE_DL)
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004949 Dl_info dli, dli_main;
Willy Tarreau9133e482020-03-04 10:19:36 +01004950 size_t size;
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004951 const char *fname, *p;
4952#endif
4953 int i;
4954
4955 if (pfx)
4956 chunk_appendf(buf, "%s", pfx);
4957
4958 for (i = 0; i < sizeof(fcts) / sizeof(fcts[0]); i++) {
4959 if (addr == fcts[i].func) {
4960 chunk_appendf(buf, "%s", fcts[i].name);
4961 return addr;
4962 }
4963 }
4964
Baruch Siache1651b22020-07-24 07:52:20 +03004965#if (defined(__ELF__) && !defined(__linux__)) || defined(USE_DL)
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004966 /* Now let's try to be smarter */
Willy Tarreau9133e482020-03-04 10:19:36 +01004967 if (!dladdr_and_size(addr, &dli, &size))
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004968 goto unknown;
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004969
4970 /* 1. prefix the library name if it's not the same object as the one
4971 * that contains the main function. The name is picked between last '/'
4972 * and first following '.'.
4973 */
4974 if (!dladdr(main, &dli_main))
4975 dli_main.dli_fbase = NULL;
4976
4977 if (dli_main.dli_fbase != dli.dli_fbase) {
4978 fname = dli.dli_fname;
4979 p = strrchr(fname, '/');
4980 if (p++)
4981 fname = p;
4982 p = strchr(fname, '.');
4983 if (!p)
4984 p = fname + strlen(fname);
4985
4986 chunk_appendf(buf, "%.*s:", (int)(long)(p - fname), fname);
4987 }
4988
4989 /* 2. symbol name */
4990 if (dli.dli_sname) {
4991 /* known, dump it and return symbol's address (exact or relative) */
4992 chunk_appendf(buf, "%s", dli.dli_sname);
4993 if (addr != dli.dli_saddr) {
4994 chunk_appendf(buf, "+%#lx", (long)(addr - dli.dli_saddr));
Willy Tarreau9133e482020-03-04 10:19:36 +01004995 if (size)
4996 chunk_appendf(buf, "/%#lx", (long)size);
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004997 }
4998 return dli.dli_saddr;
4999 }
5000 else if (dli_main.dli_fbase != dli.dli_fbase) {
5001 /* unresolved symbol from a known library, report relative offset */
5002 chunk_appendf(buf, "+%#lx", (long)(addr - dli.dli_fbase));
5003 return NULL;
5004 }
Baruch Siache1651b22020-07-24 07:52:20 +03005005#endif /* __ELF__ && !__linux__ || USE_DL */
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01005006 unknown:
5007 /* unresolved symbol from the main file, report relative offset to main */
5008 if ((void*)addr < (void*)main)
5009 chunk_appendf(buf, "main-%#lx", (long)((void*)main - addr));
5010 else
5011 chunk_appendf(buf, "main+%#lx", (long)(addr - (void*)main));
5012 return NULL;
Willy Tarreau0ebb5112016-12-05 00:10:57 +01005013}
5014
Willy Tarreau6ab7b212021-12-28 09:57:10 +01005015/* On systems where this is supported, let's provide a possibility to enumerate
5016 * the list of object files. The output is appended to a buffer initialized by
5017 * the caller, with one name per line. A trailing zero is always emitted if data
5018 * are written. Only real objects are dumped (executable and .so libs). The
5019 * function returns non-zero if it dumps anything. These functions do not make
5020 * use of the trash so that it is possible for the caller to call them with the
5021 * trash on input. The output format may be platform-specific but at least one
5022 * version must emit raw object file names when argument is zero.
5023 */
5024#if defined(HA_HAVE_DUMP_LIBS)
5025# if defined(HA_HAVE_DL_ITERATE_PHDR)
5026/* the private <data> we pass below is a dump context initialized like this */
5027struct dl_dump_ctx {
5028 struct buffer *buf;
5029 int with_addr;
5030};
5031
5032static int dl_dump_libs_cb(struct dl_phdr_info *info, size_t size, void *data)
5033{
5034 struct dl_dump_ctx *ctx = data;
5035 const char *fname;
5036 size_t p1, p2, beg, end;
5037 int idx;
5038
5039 if (!info || !info->dlpi_name)
5040 goto leave;
5041
5042 if (!*info->dlpi_name)
5043 fname = get_exec_path();
5044 else if (strchr(info->dlpi_name, '/'))
5045 fname = info->dlpi_name;
5046 else
5047 /* else it's a VDSO or similar and we're not interested */
5048 goto leave;
5049
5050 if (!ctx->with_addr)
5051 goto dump_name;
5052
5053 /* virtual addresses are relative to the load address and are per
5054 * pseudo-header, so we have to scan them all to find the furthest
5055 * one from the beginning. In this case we only dump entries if
5056 * they have at least one section.
5057 */
5058 beg = ~0; end = 0;
5059 for (idx = 0; idx < info->dlpi_phnum; idx++) {
5060 if (!info->dlpi_phdr[idx].p_memsz)
5061 continue;
5062 p1 = info->dlpi_phdr[idx].p_vaddr;
5063 if (p1 < beg)
5064 beg = p1;
5065 p2 = p1 + info->dlpi_phdr[idx].p_memsz - 1;
5066 if (p2 > end)
5067 end = p2;
5068 }
5069
5070 if (!idx)
5071 goto leave;
5072
5073 chunk_appendf(ctx->buf, "0x%012llx-0x%012llx (0x%07llx) ",
5074 (ullong)info->dlpi_addr + beg,
5075 (ullong)info->dlpi_addr + end,
5076 (ullong)(end - beg + 1));
5077 dump_name:
5078 chunk_appendf(ctx->buf, "%s\n", fname);
5079 leave:
5080 return 0;
5081}
5082
5083/* dumps lib names and optionally address ranges */
5084int dump_libs(struct buffer *output, int with_addr)
5085{
5086 struct dl_dump_ctx ctx = { .buf = output, .with_addr = with_addr };
5087 size_t old_data = output->data;
5088
5089 dl_iterate_phdr(dl_dump_libs_cb, &ctx);
5090 return output->data != old_data;
5091}
5092# else // no DL_ITERATE_PHDR
5093# error "No dump_libs() function for this platform"
5094# endif
5095#else // no HA_HAVE_DUMP_LIBS
5096
5097/* unsupported platform: do not dump anything */
5098int dump_libs(struct buffer *output, int with_addr)
5099{
5100 return 0;
5101}
5102
5103#endif // HA_HAVE_DUMP_LIBS
5104
Frédéric Lécaille3b717162019-02-25 15:04:22 +01005105/*
5106 * Allocate an array of unsigned int with <nums> as address from <str> string
Ilya Shipitsin46a030c2020-07-05 16:36:08 +05005107 * made of integer separated by dot characters.
Frédéric Lécaille3b717162019-02-25 15:04:22 +01005108 *
5109 * First, initializes the value with <sz> as address to 0 and initializes the
5110 * array with <nums> as address to NULL. Then allocates the array with <nums> as
5111 * address updating <sz> pointed value to the size of this array.
5112 *
5113 * Returns 1 if succeeded, 0 if not.
5114 */
5115int parse_dotted_uints(const char *str, unsigned int **nums, size_t *sz)
5116{
5117 unsigned int *n;
5118 const char *s, *end;
5119
5120 s = str;
5121 *sz = 0;
5122 end = str + strlen(str);
5123 *nums = n = NULL;
5124
5125 while (1) {
5126 unsigned int r;
5127
5128 if (s >= end)
5129 break;
5130
5131 r = read_uint(&s, end);
5132 /* Expected characters after having read an uint: '\0' or '.',
5133 * if '.', must not be terminal.
5134 */
Christopher Faulet4b524122021-02-11 10:42:41 +01005135 if (*s != '\0'&& (*s++ != '.' || s == end)) {
5136 free(n);
Frédéric Lécaille3b717162019-02-25 15:04:22 +01005137 return 0;
Christopher Faulet4b524122021-02-11 10:42:41 +01005138 }
Frédéric Lécaille3b717162019-02-25 15:04:22 +01005139
Frédéric Lécaille12a71842019-02-26 18:19:48 +01005140 n = my_realloc2(n, (*sz + 1) * sizeof *n);
Frédéric Lécaille3b717162019-02-25 15:04:22 +01005141 if (!n)
5142 return 0;
5143
5144 n[(*sz)++] = r;
5145 }
5146 *nums = n;
5147
5148 return 1;
5149}
5150
Willy Tarreau4d589e72019-08-23 19:02:26 +02005151
5152/* returns the number of bytes needed to encode <v> as a varint. An inline
5153 * version exists for use with constants (__varint_bytes()).
5154 */
5155int varint_bytes(uint64_t v)
5156{
5157 int len = 1;
5158
5159 if (v >= 240) {
5160 v = (v - 240) >> 4;
5161 while (1) {
5162 len++;
5163 if (v < 128)
5164 break;
5165 v = (v - 128) >> 7;
5166 }
5167 }
5168 return len;
5169}
5170
Willy Tarreau52bf8392020-03-08 00:42:37 +01005171
5172/* Random number generator state, see below */
Willy Tarreau1544c142020-03-12 00:31:18 +01005173static uint64_t ha_random_state[2] ALIGNED(2*sizeof(uint64_t));
Willy Tarreau52bf8392020-03-08 00:42:37 +01005174
5175/* This is a thread-safe implementation of xoroshiro128** described below:
5176 * http://prng.di.unimi.it/
5177 * It features a 2^128 long sequence, returns 64 high-quality bits on each call,
5178 * supports fast jumps and passes all common quality tests. It is thread-safe,
5179 * uses a double-cas on 64-bit architectures supporting it, and falls back to a
5180 * local lock on other ones.
5181 */
5182uint64_t ha_random64()
5183{
Willy Tarreau1544c142020-03-12 00:31:18 +01005184 uint64_t old[2] ALIGNED(2*sizeof(uint64_t));
5185 uint64_t new[2] ALIGNED(2*sizeof(uint64_t));
Willy Tarreau52bf8392020-03-08 00:42:37 +01005186
5187#if defined(USE_THREAD) && (!defined(HA_CAS_IS_8B) || !defined(HA_HAVE_CAS_DW))
5188 static HA_SPINLOCK_T rand_lock;
5189
5190 HA_SPIN_LOCK(OTHER_LOCK, &rand_lock);
5191#endif
5192
5193 old[0] = ha_random_state[0];
5194 old[1] = ha_random_state[1];
5195
5196#if defined(USE_THREAD) && defined(HA_CAS_IS_8B) && defined(HA_HAVE_CAS_DW)
5197 do {
5198#endif
Willy Tarreau52bf8392020-03-08 00:42:37 +01005199 new[1] = old[0] ^ old[1];
5200 new[0] = rotl64(old[0], 24) ^ new[1] ^ (new[1] << 16); // a, b
5201 new[1] = rotl64(new[1], 37); // c
5202
5203#if defined(USE_THREAD) && defined(HA_CAS_IS_8B) && defined(HA_HAVE_CAS_DW)
5204 } while (unlikely(!_HA_ATOMIC_DWCAS(ha_random_state, old, new)));
5205#else
5206 ha_random_state[0] = new[0];
5207 ha_random_state[1] = new[1];
5208#if defined(USE_THREAD)
5209 HA_SPIN_UNLOCK(OTHER_LOCK, &rand_lock);
5210#endif
5211#endif
Willy Tarreaub2475a12021-05-09 10:26:14 +02005212 return rotl64(old[0] * 5, 7) * 9;
Willy Tarreau52bf8392020-03-08 00:42:37 +01005213}
5214
5215/* seeds the random state using up to <len> bytes from <seed>, starting with
5216 * the first non-zero byte.
5217 */
5218void ha_random_seed(const unsigned char *seed, size_t len)
5219{
5220 size_t pos;
5221
5222 /* the seed must not be all zeroes, so we pre-fill it with alternating
5223 * bits and overwrite part of them with the block starting at the first
5224 * non-zero byte from the seed.
5225 */
5226 memset(ha_random_state, 0x55, sizeof(ha_random_state));
5227
5228 for (pos = 0; pos < len; pos++)
5229 if (seed[pos] != 0)
5230 break;
5231
5232 if (pos == len)
5233 return;
5234
5235 seed += pos;
5236 len -= pos;
5237
5238 if (len > sizeof(ha_random_state))
5239 len = sizeof(ha_random_state);
5240
5241 memcpy(ha_random_state, seed, len);
5242}
5243
5244/* This causes a jump to (dist * 2^96) places in the pseudo-random sequence,
5245 * and is equivalent to calling ha_random64() as many times. It is used to
5246 * provide non-overlapping sequences of 2^96 numbers (~7*10^28) to up to 2^32
5247 * different generators (i.e. different processes after a fork). The <dist>
5248 * argument is the distance to jump to and is used in a loop so it rather not
5249 * be too large if the processing time is a concern.
5250 *
5251 * BEWARE: this function is NOT thread-safe and must not be called during
5252 * concurrent accesses to ha_random64().
5253 */
5254void ha_random_jump96(uint32_t dist)
5255{
5256 while (dist--) {
5257 uint64_t s0 = 0;
5258 uint64_t s1 = 0;
5259 int b;
5260
5261 for (b = 0; b < 64; b++) {
5262 if ((0xd2a98b26625eee7bULL >> b) & 1) {
5263 s0 ^= ha_random_state[0];
5264 s1 ^= ha_random_state[1];
5265 }
5266 ha_random64();
5267 }
5268
5269 for (b = 0; b < 64; b++) {
5270 if ((0xdddf9b1090aa7ac1ULL >> b) & 1) {
5271 s0 ^= ha_random_state[0];
5272 s1 ^= ha_random_state[1];
5273 }
5274 ha_random64();
5275 }
5276 ha_random_state[0] = s0;
5277 ha_random_state[1] = s1;
5278 }
5279}
5280
Willy Tarreauee3bcdd2020-03-08 17:48:17 +01005281/* Generates an RFC4122 UUID into chunk <output> which must be at least 37
5282 * bytes large.
5283 */
5284void ha_generate_uuid(struct buffer *output)
5285{
5286 uint32_t rnd[4];
5287 uint64_t last;
5288
5289 last = ha_random64();
5290 rnd[0] = last;
5291 rnd[1] = last >> 32;
5292
5293 last = ha_random64();
5294 rnd[2] = last;
5295 rnd[3] = last >> 32;
5296
5297 chunk_printf(output, "%8.8x-%4.4x-%4.4x-%4.4x-%12.12llx",
5298 rnd[0],
5299 rnd[1] & 0xFFFF,
5300 ((rnd[1] >> 16u) & 0xFFF) | 0x4000, // highest 4 bits indicate the uuid version
5301 (rnd[2] & 0x3FFF) | 0x8000, // the highest 2 bits indicate the UUID variant (10),
5302 (long long)((rnd[2] >> 14u) | ((uint64_t) rnd[3] << 18u)) & 0xFFFFFFFFFFFFull);
5303}
5304
5305
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005306/* only used by parse_line() below. It supports writing in place provided that
5307 * <in> is updated to the next location before calling it. In that case, the
5308 * char at <in> may be overwritten.
5309 */
5310#define EMIT_CHAR(x) \
5311 do { \
5312 char __c = (char)(x); \
5313 if ((opts & PARSE_OPT_INPLACE) && out+outpos > in) \
5314 err |= PARSE_ERR_OVERLAP; \
5315 if (outpos >= outmax) \
5316 err |= PARSE_ERR_TOOLARGE; \
5317 if (!err) \
5318 out[outpos] = __c; \
5319 outpos++; \
5320 } while (0)
5321
Ilya Shipitsin46a030c2020-07-05 16:36:08 +05005322/* Parse <in>, copy it into <out> split into isolated words whose pointers
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005323 * are put in <args>. If more than <outlen> bytes have to be emitted, the
5324 * extraneous ones are not emitted but <outlen> is updated so that the caller
5325 * knows how much to realloc. Similarly, <args> are not updated beyond <nbargs>
5326 * but the returned <nbargs> indicates how many were found. All trailing args
Willy Tarreau61dd44b2020-06-25 07:35:42 +02005327 * up to <nbargs> point to the trailing zero, and as long as <nbargs> is > 0,
5328 * it is guaranteed that at least one arg will point to the zero. It is safe
5329 * to call it with a NULL <args> if <nbargs> is 0.
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005330 *
5331 * <out> may overlap with <in> provided that it never goes further, in which
5332 * case the parser will accept to perform in-place parsing and unquoting/
5333 * unescaping but only if environment variables do not lead to expansion that
5334 * causes overlapping, otherwise the input string being destroyed, the error
5335 * will not be recoverable. Note that even during out-of-place <in> will
5336 * experience temporary modifications in-place for variable resolution and must
5337 * be writable, and will also receive zeroes to delimit words when using
5338 * in-place copy. Parsing options <opts> taken from PARSE_OPT_*. Return value
5339 * is zero on success otherwise a bitwise-or of PARSE_ERR_*. Upon error, the
5340 * starting point of the first invalid character sequence or unmatched
5341 * quote/brace is reported in <errptr> if not NULL. When using in-place parsing
5342 * error reporting might be difficult since zeroes will have been inserted into
5343 * the string. One solution for the caller may consist in replacing all args
5344 * delimiters with spaces in this case.
5345 */
Maximilian Mader29c6cd72021-06-06 00:50:21 +02005346uint32_t parse_line(char *in, char *out, size_t *outlen, char **args, int *nbargs, uint32_t opts, const char **errptr)
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005347{
5348 char *quote = NULL;
5349 char *brace = NULL;
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005350 char *word_expand = NULL;
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005351 unsigned char hex1, hex2;
5352 size_t outmax = *outlen;
Willy Tarreau61dd44b2020-06-25 07:35:42 +02005353 int argsmax = *nbargs - 1;
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005354 size_t outpos = 0;
5355 int squote = 0;
5356 int dquote = 0;
5357 int arg = 0;
5358 uint32_t err = 0;
5359
5360 *nbargs = 0;
5361 *outlen = 0;
5362
Willy Tarreau61dd44b2020-06-25 07:35:42 +02005363 /* argsmax may be -1 here, protecting args[] from any write */
5364 if (arg < argsmax)
5365 args[arg] = out;
5366
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005367 while (1) {
5368 if (*in >= '-' && *in != '\\') {
5369 /* speedup: directly send all regular chars starting
5370 * with '-', '.', '/', alnum etc...
5371 */
5372 EMIT_CHAR(*in++);
5373 continue;
5374 }
5375 else if (*in == '\0' || *in == '\n' || *in == '\r') {
5376 /* end of line */
5377 break;
5378 }
5379 else if (*in == '#' && (opts & PARSE_OPT_SHARP) && !squote && !dquote) {
5380 /* comment */
5381 break;
5382 }
5383 else if (*in == '"' && !squote && (opts & PARSE_OPT_DQUOTE)) { /* double quote outside single quotes */
5384 if (dquote) {
5385 dquote = 0;
5386 quote = NULL;
5387 }
5388 else {
5389 dquote = 1;
5390 quote = in;
5391 }
5392 in++;
5393 continue;
5394 }
5395 else if (*in == '\'' && !dquote && (opts & PARSE_OPT_SQUOTE)) { /* single quote outside double quotes */
5396 if (squote) {
5397 squote = 0;
5398 quote = NULL;
5399 }
5400 else {
5401 squote = 1;
5402 quote = in;
5403 }
5404 in++;
5405 continue;
5406 }
5407 else if (*in == '\\' && !squote && (opts & PARSE_OPT_BKSLASH)) {
5408 /* first, we'll replace \\, \<space>, \#, \r, \n, \t, \xXX with their
5409 * C equivalent value but only when they have a special meaning and within
5410 * double quotes for some of them. Other combinations left unchanged (eg: \1).
5411 */
5412 char tosend = *in;
5413
5414 switch (in[1]) {
5415 case ' ':
5416 case '\\':
5417 tosend = in[1];
5418 in++;
5419 break;
5420
5421 case 't':
5422 tosend = '\t';
5423 in++;
5424 break;
5425
5426 case 'n':
5427 tosend = '\n';
5428 in++;
5429 break;
5430
5431 case 'r':
5432 tosend = '\r';
5433 in++;
5434 break;
5435
5436 case '#':
5437 /* escaping of "#" only if comments are supported */
5438 if (opts & PARSE_OPT_SHARP)
5439 in++;
5440 tosend = *in;
5441 break;
5442
5443 case '\'':
5444 /* escaping of "'" only outside single quotes and only if single quotes are supported */
5445 if (opts & PARSE_OPT_SQUOTE && !squote)
5446 in++;
5447 tosend = *in;
5448 break;
5449
5450 case '"':
5451 /* escaping of '"' only outside single quotes and only if double quotes are supported */
5452 if (opts & PARSE_OPT_DQUOTE && !squote)
5453 in++;
5454 tosend = *in;
5455 break;
5456
5457 case '$':
5458 /* escaping of '$' only inside double quotes and only if env supported */
5459 if (opts & PARSE_OPT_ENV && dquote)
5460 in++;
5461 tosend = *in;
5462 break;
5463
5464 case 'x':
5465 if (!ishex(in[2]) || !ishex(in[3])) {
5466 /* invalid or incomplete hex sequence */
5467 err |= PARSE_ERR_HEX;
5468 if (errptr)
5469 *errptr = in;
5470 goto leave;
5471 }
Willy Tarreauf278eec2020-07-05 21:46:32 +02005472 hex1 = toupper((unsigned char)in[2]) - '0';
5473 hex2 = toupper((unsigned char)in[3]) - '0';
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005474 if (hex1 > 9) hex1 -= 'A' - '9' - 1;
5475 if (hex2 > 9) hex2 -= 'A' - '9' - 1;
5476 tosend = (hex1 << 4) + hex2;
5477 in += 3;
5478 break;
5479
5480 default:
5481 /* other combinations are not escape sequences */
5482 break;
5483 }
5484
5485 in++;
5486 EMIT_CHAR(tosend);
5487 }
5488 else if (isspace((unsigned char)*in) && !squote && !dquote) {
5489 /* a non-escaped space is an argument separator */
5490 while (isspace((unsigned char)*in))
5491 in++;
5492 EMIT_CHAR(0);
5493 arg++;
5494 if (arg < argsmax)
5495 args[arg] = out + outpos;
5496 else
5497 err |= PARSE_ERR_TOOMANY;
5498 }
5499 else if (*in == '$' && (opts & PARSE_OPT_ENV) && (dquote || !(opts & PARSE_OPT_DQUOTE))) {
5500 /* environment variables are evaluated anywhere, or only
5501 * inside double quotes if they are supported.
5502 */
5503 char *var_name;
5504 char save_char;
Willy Tarreaua46f1af2021-05-06 10:25:11 +02005505 const char *value;
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005506
5507 in++;
5508
5509 if (*in == '{')
5510 brace = in++;
5511
Willy Tarreaua46f1af2021-05-06 10:25:11 +02005512 if (!isalpha((unsigned char)*in) && *in != '_' && *in != '.') {
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005513 /* unacceptable character in variable name */
5514 err |= PARSE_ERR_VARNAME;
5515 if (errptr)
5516 *errptr = in;
5517 goto leave;
5518 }
5519
5520 var_name = in;
Willy Tarreaua46f1af2021-05-06 10:25:11 +02005521 if (*in == '.')
5522 in++;
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005523 while (isalnum((unsigned char)*in) || *in == '_')
5524 in++;
5525
5526 save_char = *in;
5527 *in = '\0';
Willy Tarreaua46f1af2021-05-06 10:25:11 +02005528 if (unlikely(*var_name == '.')) {
5529 /* internal pseudo-variables */
5530 if (strcmp(var_name, ".LINE") == 0)
5531 value = ultoa(global.cfg_curr_line);
5532 else if (strcmp(var_name, ".FILE") == 0)
5533 value = global.cfg_curr_file;
5534 else if (strcmp(var_name, ".SECTION") == 0)
5535 value = global.cfg_curr_section;
5536 else {
5537 /* unsupported internal variable name */
5538 err |= PARSE_ERR_VARNAME;
5539 if (errptr)
5540 *errptr = var_name;
5541 goto leave;
5542 }
5543 } else {
5544 value = getenv(var_name);
5545 }
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005546 *in = save_char;
5547
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005548 /* support for '[*]' sequence to force word expansion,
5549 * only available inside braces */
5550 if (*in == '[' && brace && (opts & PARSE_OPT_WORD_EXPAND)) {
5551 word_expand = in++;
5552
5553 if (*in++ != '*' || *in++ != ']') {
5554 err |= PARSE_ERR_WRONG_EXPAND;
5555 if (errptr)
5556 *errptr = word_expand;
5557 goto leave;
5558 }
5559 }
5560
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005561 if (brace) {
Willy Tarreauec347b12021-11-18 17:42:50 +01005562 if (*in == '-') {
5563 /* default value starts just after the '-' */
5564 if (!value)
5565 value = in + 1;
5566
5567 while (*in && *in != '}')
5568 in++;
5569 if (!*in)
5570 goto no_brace;
5571 *in = 0; // terminate the default value
5572 }
5573 else if (*in != '}') {
5574 no_brace:
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005575 /* unmatched brace */
5576 err |= PARSE_ERR_BRACE;
5577 if (errptr)
5578 *errptr = brace;
5579 goto leave;
5580 }
Willy Tarreauec347b12021-11-18 17:42:50 +01005581
5582 /* brace found, skip it */
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005583 in++;
5584 brace = NULL;
5585 }
5586
5587 if (value) {
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005588 while (*value) {
5589 /* expand as individual parameters on a space character */
Willy Tarreaufe2cc412020-10-01 18:04:40 +02005590 if (word_expand && isspace((unsigned char)*value)) {
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005591 EMIT_CHAR(0);
5592 ++arg;
5593 if (arg < argsmax)
5594 args[arg] = out + outpos;
5595 else
5596 err |= PARSE_ERR_TOOMANY;
5597
5598 /* skip consecutive spaces */
Willy Tarreaufe2cc412020-10-01 18:04:40 +02005599 while (isspace((unsigned char)*++value))
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005600 ;
5601 } else {
5602 EMIT_CHAR(*value++);
5603 }
5604 }
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005605 }
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005606 word_expand = NULL;
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005607 }
5608 else {
5609 /* any other regular char */
5610 EMIT_CHAR(*in++);
5611 }
5612 }
5613
5614 /* end of output string */
5615 EMIT_CHAR(0);
5616 arg++;
5617
5618 if (quote) {
5619 /* unmatched quote */
5620 err |= PARSE_ERR_QUOTE;
5621 if (errptr)
5622 *errptr = quote;
5623 goto leave;
5624 }
5625 leave:
5626 *nbargs = arg;
5627 *outlen = outpos;
5628
Willy Tarreau61dd44b2020-06-25 07:35:42 +02005629 /* empty all trailing args by making them point to the trailing zero,
5630 * at least the last one in any case.
5631 */
5632 if (arg > argsmax)
5633 arg = argsmax;
5634
5635 while (arg >= 0 && arg <= argsmax)
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005636 args[arg++] = out + outpos - 1;
5637
5638 return err;
5639}
5640#undef EMIT_CHAR
5641
Willy Tarreauc54e5ad2020-06-25 09:15:40 +02005642/* This is used to sanitize an input line that's about to be used for error reporting.
5643 * It will adjust <line> to print approximately <width> chars around <pos>, trying to
5644 * preserve the beginning, with leading or trailing "..." when the line is truncated.
5645 * If non-printable chars are present in the output. It returns the new offset <pos>
5646 * in the modified line. Non-printable characters are replaced with '?'. <width> must
5647 * be at least 6 to support two "..." otherwise the result is undefined. The line
5648 * itself must have at least 7 chars allocated for the same reason.
5649 */
5650size_t sanitize_for_printing(char *line, size_t pos, size_t width)
5651{
5652 size_t shift = 0;
5653 char *out = line;
5654 char *in = line;
5655 char *end = line + width;
5656
5657 if (pos >= width) {
5658 /* if we have to shift, we'll be out of context, so let's
5659 * try to put <pos> at the center of width.
5660 */
5661 shift = pos - width / 2;
5662 in += shift + 3;
5663 end = out + width - 3;
5664 out[0] = out[1] = out[2] = '.';
5665 out += 3;
5666 }
5667
5668 while (out < end && *in) {
5669 if (isspace((unsigned char)*in))
5670 *out++ = ' ';
5671 else if (isprint((unsigned char)*in))
5672 *out++ = *in;
5673 else
5674 *out++ = '?';
5675 in++;
5676 }
5677
5678 if (end < line + width) {
5679 out[0] = out[1] = out[2] = '.';
5680 out += 3;
5681 }
5682
5683 *out++ = 0;
5684 return pos - shift;
5685}
Willy Tarreau06e69b52021-03-02 14:01:35 +01005686
Willy Tarreaue33c4b32021-03-12 18:59:31 +01005687/* Update array <fp> with the fingerprint of word <word> by counting the
Willy Tarreauba2c4452021-03-12 09:01:52 +01005688 * transitions between characters. <fp> is a 1024-entries array indexed as
5689 * 32*from+to. Positions for 'from' and 'to' are:
Willy Tarreau9294e882021-03-15 09:34:27 +01005690 * 1..26=letter, 27=digit, 28=other/begin/end.
5691 * Row "from=0" is used to mark the character's presence. Others unused.
Willy Tarreauba2c4452021-03-12 09:01:52 +01005692 */
Willy Tarreaue33c4b32021-03-12 18:59:31 +01005693void update_word_fingerprint(uint8_t *fp, const char *word)
Willy Tarreauba2c4452021-03-12 09:01:52 +01005694{
5695 const char *p;
5696 int from, to;
5697 int c;
5698
Willy Tarreauba2c4452021-03-12 09:01:52 +01005699 from = 28; // begin
5700 for (p = word; *p; p++) {
5701 c = tolower(*p);
5702 switch(c) {
Willy Tarreau9294e882021-03-15 09:34:27 +01005703 case 'a'...'z': to = c - 'a' + 1; break;
5704 case 'A'...'Z': to = tolower(c) - 'a' + 1; break;
5705 case '0'...'9': to = 27; break;
5706 default: to = 28; break;
Willy Tarreauba2c4452021-03-12 09:01:52 +01005707 }
Willy Tarreau9294e882021-03-15 09:34:27 +01005708 fp[to] = 1;
Willy Tarreauba2c4452021-03-12 09:01:52 +01005709 fp[32 * from + to]++;
5710 from = to;
5711 }
5712 to = 28; // end
5713 fp[32 * from + to]++;
5714}
5715
Willy Tarreaue33c4b32021-03-12 18:59:31 +01005716/* Initialize array <fp> with the fingerprint of word <word> by counting the
5717 * transitions between characters. <fp> is a 1024-entries array indexed as
5718 * 32*from+to. Positions for 'from' and 'to' are:
5719 * 0..25=letter, 26=digit, 27=other, 28=begin, 29=end, others unused.
5720 */
5721void make_word_fingerprint(uint8_t *fp, const char *word)
5722{
5723 memset(fp, 0, 1024);
5724 update_word_fingerprint(fp, word);
5725}
5726
Willy Tarreauba2c4452021-03-12 09:01:52 +01005727/* Return the distance between two word fingerprints created by function
5728 * make_word_fingerprint(). It's a positive integer calculated as the sum of
Willy Tarreau714c4c12021-03-15 09:44:53 +01005729 * the differences between each location.
Willy Tarreauba2c4452021-03-12 09:01:52 +01005730 */
5731int word_fingerprint_distance(const uint8_t *fp1, const uint8_t *fp2)
5732{
5733 int i, k, dist = 0;
5734
5735 for (i = 0; i < 1024; i++) {
5736 k = (int)fp1[i] - (int)fp2[i];
Willy Tarreau714c4c12021-03-15 09:44:53 +01005737 dist += abs(k);
Willy Tarreauba2c4452021-03-12 09:01:52 +01005738 }
5739 return dist;
5740}
5741
William Lallemand3aeb3f92021-08-21 23:59:56 +02005742/*
5743 * This function compares the loaded openssl version with a string <version>
5744 * This function use the same return code as compare_current_version:
5745 *
5746 * -1 : the version in argument is older than the current openssl version
5747 * 0 : the version in argument is the same as the current openssl version
5748 * 1 : the version in argument is newer than the current openssl version
5749 *
5750 * Or some errors:
5751 * -2 : openssl is not available on this process
5752 * -3 : the version in argument is not parsable
5753 */
5754int openssl_compare_current_version(const char *version)
5755{
5756#ifdef USE_OPENSSL
5757 int numversion;
5758
5759 numversion = openssl_version_parser(version);
5760 if (numversion == 0)
5761 return -3;
5762
5763 if (numversion < OPENSSL_VERSION_NUMBER)
5764 return -1;
5765 else if (numversion > OPENSSL_VERSION_NUMBER)
5766 return 1;
5767 else
5768 return 0;
5769#else
5770 return -2;
5771#endif
5772}
5773
Remi Tricot-Le Bretonb01179a2021-10-11 15:34:12 +02005774/*
5775 * This function compares the loaded openssl name with a string <name>
5776 * This function returns 0 if the OpenSSL name starts like the passed parameter,
5777 * 1 otherwise.
5778 */
5779int openssl_compare_current_name(const char *name)
5780{
5781#ifdef USE_OPENSSL
5782 int name_len = 0;
5783 const char *openssl_version = OpenSSL_version(OPENSSL_VERSION);
5784
5785 if (name) {
5786 name_len = strlen(name);
5787 if (strlen(name) <= strlen(openssl_version))
5788 return strncmp(openssl_version, name, name_len);
5789 }
5790#endif
5791 return 1;
5792}
5793
Willy Tarreau06e69b52021-03-02 14:01:35 +01005794static int init_tools_per_thread()
5795{
5796 /* Let's make each thread start from a different position */
5797 statistical_prng_state += tid * MAX_THREADS;
5798 if (!statistical_prng_state)
5799 statistical_prng_state++;
5800 return 1;
5801}
5802REGISTER_PER_THREAD_INIT(init_tools_per_thread);
Willy Tarreauc54e5ad2020-06-25 09:15:40 +02005803
Willy Tarreaubaaee002006-06-26 02:48:02 +02005804/*
5805 * Local variables:
5806 * c-indent-level: 8
5807 * c-basic-offset: 8
5808 * End:
5809 */