blob: f002573d7589d5752c860d19f559889f90b131eb [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
Willy Tarreau2e74c3f2007-12-02 18:45:09 +010013#include <ctype.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020014#include <netdb.h>
Willy Tarreau9a7bea52012-04-27 11:16:50 +020015#include <stdarg.h>
Willy Tarreaudd2f85e2012-09-02 22:34:23 +020016#include <stdio.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020017#include <stdlib.h>
18#include <string.h>
Thierry Fournier93127942016-01-20 18:49:45 +010019#include <time.h>
Willy Tarreau127f9662007-12-06 00:53:51 +010020#include <sys/socket.h>
21#include <sys/un.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020022#include <netinet/in.h>
23#include <arpa/inet.h>
24
Thierry FOURNIERe059ec92014-03-17 12:01:13 +010025#include <common/chunk.h>
Willy Tarreaue3ba5f02006-06-29 18:54:54 +020026#include <common/config.h>
Willy Tarreau2dd0d472006-06-29 17:53:05 +020027#include <common/standard.h>
Thierry Fournier93127942016-01-20 18:49:45 +010028#include <common/tools.h>
Thierry FOURNIER9f95e402014-03-21 14:51:46 +010029#include <types/global.h>
Baptiste Assmanna68ca962015-04-14 01:15:08 +020030#include <proto/dns.h>
Willy Tarreau45cb4fb2009-10-26 21:10:04 +010031#include <eb32tree.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020032
Thierry Fournier93127942016-01-20 18:49:45 +010033/* This macro returns false if the test __x is false. Many
34 * of the following parsing function must be abort the processing
35 * if it returns 0, so this macro is useful for writing light code.
36 */
37#define RET0_UNLESS(__x) do { if (!(__x)) return 0; } while (0)
38
Willy Tarreau56adcf22012-12-23 18:00:29 +010039/* enough to store NB_ITOA_STR integers of :
Willy Tarreau72d759c2007-10-25 12:14:10 +020040 * 2^64-1 = 18446744073709551615 or
41 * -2^63 = -9223372036854775808
Willy Tarreaue7239b52009-03-29 13:41:58 +020042 *
43 * The HTML version needs room for adding the 25 characters
44 * '<span class="rls"></span>' around digits at positions 3N+1 in order
45 * to add spacing at up to 6 positions : 18 446 744 073 709 551 615
Willy Tarreau72d759c2007-10-25 12:14:10 +020046 */
Willy Tarreau56adcf22012-12-23 18:00:29 +010047char itoa_str[NB_ITOA_STR][171];
48int itoa_idx = 0; /* index of next itoa_str to use */
Willy Tarreaubaaee002006-06-26 02:48:02 +020049
Willy Tarreau588297f2014-06-16 15:16:40 +020050/* sometimes we'll need to quote strings (eg: in stats), and we don't expect
51 * to quote strings larger than a max configuration line.
52 */
53char quoted_str[NB_QSTR][QSTR_SIZE + 1];
54int quoted_idx = 0;
55
Willy Tarreaubaaee002006-06-26 02:48:02 +020056/*
William Lallemande7340ec2012-01-24 11:15:39 +010057 * unsigned long long ASCII representation
58 *
59 * return the last char '\0' or NULL if no enough
60 * space in dst
61 */
62char *ulltoa(unsigned long long n, char *dst, size_t size)
63{
64 int i = 0;
65 char *res;
66
67 switch(n) {
68 case 1ULL ... 9ULL:
69 i = 0;
70 break;
71
72 case 10ULL ... 99ULL:
73 i = 1;
74 break;
75
76 case 100ULL ... 999ULL:
77 i = 2;
78 break;
79
80 case 1000ULL ... 9999ULL:
81 i = 3;
82 break;
83
84 case 10000ULL ... 99999ULL:
85 i = 4;
86 break;
87
88 case 100000ULL ... 999999ULL:
89 i = 5;
90 break;
91
92 case 1000000ULL ... 9999999ULL:
93 i = 6;
94 break;
95
96 case 10000000ULL ... 99999999ULL:
97 i = 7;
98 break;
99
100 case 100000000ULL ... 999999999ULL:
101 i = 8;
102 break;
103
104 case 1000000000ULL ... 9999999999ULL:
105 i = 9;
106 break;
107
108 case 10000000000ULL ... 99999999999ULL:
109 i = 10;
110 break;
111
112 case 100000000000ULL ... 999999999999ULL:
113 i = 11;
114 break;
115
116 case 1000000000000ULL ... 9999999999999ULL:
117 i = 12;
118 break;
119
120 case 10000000000000ULL ... 99999999999999ULL:
121 i = 13;
122 break;
123
124 case 100000000000000ULL ... 999999999999999ULL:
125 i = 14;
126 break;
127
128 case 1000000000000000ULL ... 9999999999999999ULL:
129 i = 15;
130 break;
131
132 case 10000000000000000ULL ... 99999999999999999ULL:
133 i = 16;
134 break;
135
136 case 100000000000000000ULL ... 999999999999999999ULL:
137 i = 17;
138 break;
139
140 case 1000000000000000000ULL ... 9999999999999999999ULL:
141 i = 18;
142 break;
143
144 case 10000000000000000000ULL ... ULLONG_MAX:
145 i = 19;
146 break;
147 }
148 if (i + 2 > size) // (i + 1) + '\0'
149 return NULL; // too long
150 res = dst + i + 1;
151 *res = '\0';
152 for (; i >= 0; i--) {
153 dst[i] = n % 10ULL + '0';
154 n /= 10ULL;
155 }
156 return res;
157}
158
159/*
160 * unsigned long ASCII representation
161 *
162 * return the last char '\0' or NULL if no enough
163 * space in dst
164 */
165char *ultoa_o(unsigned long n, char *dst, size_t size)
166{
167 int i = 0;
168 char *res;
169
170 switch (n) {
171 case 0U ... 9UL:
172 i = 0;
173 break;
174
175 case 10U ... 99UL:
176 i = 1;
177 break;
178
179 case 100U ... 999UL:
180 i = 2;
181 break;
182
183 case 1000U ... 9999UL:
184 i = 3;
185 break;
186
187 case 10000U ... 99999UL:
188 i = 4;
189 break;
190
191 case 100000U ... 999999UL:
192 i = 5;
193 break;
194
195 case 1000000U ... 9999999UL:
196 i = 6;
197 break;
198
199 case 10000000U ... 99999999UL:
200 i = 7;
201 break;
202
203 case 100000000U ... 999999999UL:
204 i = 8;
205 break;
206#if __WORDSIZE == 32
207
208 case 1000000000ULL ... ULONG_MAX:
209 i = 9;
210 break;
211
212#elif __WORDSIZE == 64
213
214 case 1000000000ULL ... 9999999999UL:
215 i = 9;
216 break;
217
218 case 10000000000ULL ... 99999999999UL:
219 i = 10;
220 break;
221
222 case 100000000000ULL ... 999999999999UL:
223 i = 11;
224 break;
225
226 case 1000000000000ULL ... 9999999999999UL:
227 i = 12;
228 break;
229
230 case 10000000000000ULL ... 99999999999999UL:
231 i = 13;
232 break;
233
234 case 100000000000000ULL ... 999999999999999UL:
235 i = 14;
236 break;
237
238 case 1000000000000000ULL ... 9999999999999999UL:
239 i = 15;
240 break;
241
242 case 10000000000000000ULL ... 99999999999999999UL:
243 i = 16;
244 break;
245
246 case 100000000000000000ULL ... 999999999999999999UL:
247 i = 17;
248 break;
249
250 case 1000000000000000000ULL ... 9999999999999999999UL:
251 i = 18;
252 break;
253
254 case 10000000000000000000ULL ... ULONG_MAX:
255 i = 19;
256 break;
257
258#endif
259 }
260 if (i + 2 > size) // (i + 1) + '\0'
261 return NULL; // too long
262 res = dst + i + 1;
263 *res = '\0';
264 for (; i >= 0; i--) {
265 dst[i] = n % 10U + '0';
266 n /= 10U;
267 }
268 return res;
269}
270
271/*
272 * signed long ASCII representation
273 *
274 * return the last char '\0' or NULL if no enough
275 * space in dst
276 */
277char *ltoa_o(long int n, char *dst, size_t size)
278{
279 char *pos = dst;
280
281 if (n < 0) {
282 if (size < 3)
283 return NULL; // min size is '-' + digit + '\0' but another test in ultoa
284 *pos = '-';
285 pos++;
286 dst = ultoa_o(-n, pos, size - 1);
287 } else {
288 dst = ultoa_o(n, dst, size);
289 }
290 return dst;
291}
292
293/*
294 * signed long long ASCII representation
295 *
296 * return the last char '\0' or NULL if no enough
297 * space in dst
298 */
299char *lltoa(long long n, char *dst, size_t size)
300{
301 char *pos = dst;
302
303 if (n < 0) {
304 if (size < 3)
305 return NULL; // min size is '-' + digit + '\0' but another test in ulltoa
306 *pos = '-';
307 pos++;
308 dst = ulltoa(-n, pos, size - 1);
309 } else {
310 dst = ulltoa(n, dst, size);
311 }
312 return dst;
313}
314
315/*
316 * write a ascii representation of a unsigned into dst,
317 * return a pointer to the last character
318 * Pad the ascii representation with '0', using size.
319 */
320char *utoa_pad(unsigned int n, char *dst, size_t size)
321{
322 int i = 0;
323 char *ret;
324
325 switch(n) {
326 case 0U ... 9U:
327 i = 0;
328 break;
329
330 case 10U ... 99U:
331 i = 1;
332 break;
333
334 case 100U ... 999U:
335 i = 2;
336 break;
337
338 case 1000U ... 9999U:
339 i = 3;
340 break;
341
342 case 10000U ... 99999U:
343 i = 4;
344 break;
345
346 case 100000U ... 999999U:
347 i = 5;
348 break;
349
350 case 1000000U ... 9999999U:
351 i = 6;
352 break;
353
354 case 10000000U ... 99999999U:
355 i = 7;
356 break;
357
358 case 100000000U ... 999999999U:
359 i = 8;
360 break;
361
362 case 1000000000U ... 4294967295U:
363 i = 9;
364 break;
365 }
366 if (i + 2 > size) // (i + 1) + '\0'
367 return NULL; // too long
368 if (i < size)
369 i = size - 2; // padding - '\0'
370
371 ret = dst + i + 1;
372 *ret = '\0';
373 for (; i >= 0; i--) {
374 dst[i] = n % 10U + '0';
375 n /= 10U;
376 }
377 return ret;
378}
379
380/*
Willy Tarreaubaaee002006-06-26 02:48:02 +0200381 * copies at most <size-1> chars from <src> to <dst>. Last char is always
382 * set to 0, unless <size> is 0. The number of chars copied is returned
383 * (excluding the terminating zero).
384 * This code has been optimized for size and speed : on x86, it's 45 bytes
385 * long, uses only registers, and consumes only 4 cycles per char.
386 */
387int strlcpy2(char *dst, const char *src, int size)
388{
389 char *orig = dst;
390 if (size) {
391 while (--size && (*dst = *src)) {
392 src++; dst++;
393 }
394 *dst = 0;
395 }
396 return dst - orig;
397}
398
399/*
Willy Tarreau72d759c2007-10-25 12:14:10 +0200400 * This function simply returns a locally allocated string containing
Willy Tarreaubaaee002006-06-26 02:48:02 +0200401 * the ascii representation for number 'n' in decimal.
402 */
Emeric Brun3a7fce52010-01-04 14:54:38 +0100403char *ultoa_r(unsigned long n, char *buffer, int size)
Willy Tarreaubaaee002006-06-26 02:48:02 +0200404{
405 char *pos;
406
Willy Tarreau72d759c2007-10-25 12:14:10 +0200407 pos = buffer + size - 1;
Willy Tarreaubaaee002006-06-26 02:48:02 +0200408 *pos-- = '\0';
409
410 do {
411 *pos-- = '0' + n % 10;
412 n /= 10;
Willy Tarreau72d759c2007-10-25 12:14:10 +0200413 } while (n && pos >= buffer);
Willy Tarreaubaaee002006-06-26 02:48:02 +0200414 return pos + 1;
415}
416
Willy Tarreau91092e52007-10-25 16:58:42 +0200417/*
Willy Tarreaue7239b52009-03-29 13:41:58 +0200418 * This function simply returns a locally allocated string containing
Thierry FOURNIER763a5d82015-07-06 23:09:52 +0200419 * the ascii representation for number 'n' in decimal.
420 */
421char *lltoa_r(long long int in, char *buffer, int size)
422{
423 char *pos;
424 int neg = 0;
425 unsigned long long int n;
426
427 pos = buffer + size - 1;
428 *pos-- = '\0';
429
430 if (in < 0) {
431 neg = 1;
432 n = -in;
433 }
434 else
435 n = in;
436
437 do {
438 *pos-- = '0' + n % 10;
439 n /= 10;
440 } while (n && pos >= buffer);
441 if (neg && pos > buffer)
442 *pos-- = '-';
443 return pos + 1;
444}
445
446/*
447 * This function simply returns a locally allocated string containing
Thierry FOURNIER1480bd82015-06-06 19:14:59 +0200448 * the ascii representation for signed number 'n' in decimal.
449 */
450char *sltoa_r(long n, char *buffer, int size)
451{
452 char *pos;
453
454 if (n >= 0)
455 return ultoa_r(n, buffer, size);
456
457 pos = ultoa_r(-n, buffer + 1, size - 1) - 1;
458 *pos = '-';
459 return pos;
460}
461
462/*
463 * This function simply returns a locally allocated string containing
Willy Tarreaue7239b52009-03-29 13:41:58 +0200464 * the ascii representation for number 'n' in decimal, formatted for
465 * HTML output with tags to create visual grouping by 3 digits. The
466 * output needs to support at least 171 characters.
467 */
468const char *ulltoh_r(unsigned long long n, char *buffer, int size)
469{
470 char *start;
471 int digit = 0;
472
473 start = buffer + size;
474 *--start = '\0';
475
476 do {
477 if (digit == 3 && start >= buffer + 7)
478 memcpy(start -= 7, "</span>", 7);
479
480 if (start >= buffer + 1) {
481 *--start = '0' + n % 10;
482 n /= 10;
483 }
484
485 if (digit == 3 && start >= buffer + 18)
486 memcpy(start -= 18, "<span class=\"rls\">", 18);
487
488 if (digit++ == 3)
489 digit = 1;
490 } while (n && start > buffer);
491 return start;
492}
493
494/*
Willy Tarreau91092e52007-10-25 16:58:42 +0200495 * This function simply returns a locally allocated string containing the ascii
496 * representation for number 'n' in decimal, unless n is 0 in which case it
497 * returns the alternate string (or an empty string if the alternate string is
498 * NULL). It use is intended for limits reported in reports, where it's
499 * desirable not to display anything if there is no limit. Warning! it shares
500 * the same vector as ultoa_r().
501 */
502const char *limit_r(unsigned long n, char *buffer, int size, const char *alt)
503{
504 return (n) ? ultoa_r(n, buffer, size) : (alt ? alt : "");
505}
506
Willy Tarreau588297f2014-06-16 15:16:40 +0200507/* returns a locally allocated string containing the quoted encoding of the
508 * input string. The output may be truncated to QSTR_SIZE chars, but it is
509 * guaranteed that the string will always be properly terminated. Quotes are
510 * encoded by doubling them as is commonly done in CSV files. QSTR_SIZE must
511 * always be at least 4 chars.
512 */
513const char *qstr(const char *str)
514{
515 char *ret = quoted_str[quoted_idx];
516 char *p, *end;
517
518 if (++quoted_idx >= NB_QSTR)
519 quoted_idx = 0;
520
521 p = ret;
522 end = ret + QSTR_SIZE;
523
524 *p++ = '"';
525
526 /* always keep 3 chars to support passing "" and the ending " */
527 while (*str && p < end - 3) {
528 if (*str == '"') {
529 *p++ = '"';
530 *p++ = '"';
531 }
532 else
533 *p++ = *str;
534 str++;
535 }
536 *p++ = '"';
537 return ret;
538}
539
Robert Tsai81ae1952007-12-05 10:47:29 +0100540/*
Willy Tarreaubaaee002006-06-26 02:48:02 +0200541 * Returns non-zero if character <s> is a hex digit (0-9, a-f, A-F), else zero.
542 *
543 * It looks like this one would be a good candidate for inlining, but this is
544 * not interesting because it around 35 bytes long and often called multiple
545 * times within the same function.
546 */
547int ishex(char s)
548{
549 s -= '0';
550 if ((unsigned char)s <= 9)
551 return 1;
552 s -= 'A' - '0';
553 if ((unsigned char)s <= 5)
554 return 1;
555 s -= 'a' - 'A';
556 if ((unsigned char)s <= 5)
557 return 1;
558 return 0;
559}
560
Willy Tarreau3ca1a882015-01-15 18:43:49 +0100561/* rounds <i> down to the closest value having max 2 digits */
562unsigned int round_2dig(unsigned int i)
563{
564 unsigned int mul = 1;
565
566 while (i >= 100) {
567 i /= 10;
568 mul *= 10;
569 }
570 return i * mul;
571}
572
Willy Tarreau2e74c3f2007-12-02 18:45:09 +0100573/*
574 * Checks <name> for invalid characters. Valid chars are [A-Za-z0-9_:.-]. If an
575 * invalid character is found, a pointer to it is returned. If everything is
576 * fine, NULL is returned.
577 */
578const char *invalid_char(const char *name)
579{
580 if (!*name)
581 return name;
582
583 while (*name) {
Willy Tarreau88e05812010-03-03 00:16:00 +0100584 if (!isalnum((int)(unsigned char)*name) && *name != '.' && *name != ':' &&
Willy Tarreau2e74c3f2007-12-02 18:45:09 +0100585 *name != '_' && *name != '-')
586 return name;
587 name++;
588 }
589 return NULL;
590}
Willy Tarreaubaaee002006-06-26 02:48:02 +0200591
592/*
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +0200593 * Checks <domainname> for invalid characters. Valid chars are [A-Za-z0-9_.-].
594 * If an invalid character is found, a pointer to it is returned.
595 * If everything is fine, NULL is returned.
596 */
597const char *invalid_domainchar(const char *name) {
598
599 if (!*name)
600 return name;
601
602 while (*name) {
Willy Tarreau88e05812010-03-03 00:16:00 +0100603 if (!isalnum((int)(unsigned char)*name) && *name != '.' &&
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +0200604 *name != '_' && *name != '-')
605 return name;
606
607 name++;
608 }
609
610 return NULL;
611}
612
613/*
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100614 * converts <str> to a struct sockaddr_storage* provided by the caller. The
Willy Tarreau24709282013-03-10 21:32:12 +0100615 * caller must have zeroed <sa> first, and may have set sa->ss_family to force
616 * parse a specific address format. If the ss_family is 0 or AF_UNSPEC, then
617 * the function tries to guess the address family from the syntax. If the
618 * family is forced and the format doesn't match, an error is returned. The
Willy Tarreaufab5a432011-03-04 15:31:53 +0100619 * string is assumed to contain only an address, no port. The address can be a
620 * dotted IPv4 address, an IPv6 address, a host name, or empty or "*" to
621 * indicate INADDR_ANY. NULL is returned if the host part cannot be resolved.
622 * The return address will only have the address family and the address set,
623 * all other fields remain zero. The string is not supposed to be modified.
Thierry FOURNIER58639a02014-11-25 12:02:25 +0100624 * The IPv6 '::' address is IN6ADDR_ANY. If <resolve> is non-zero, the hostname
625 * is resolved, otherwise only IP addresses are resolved, and anything else
626 * returns NULL.
Willy Tarreaubaaee002006-06-26 02:48:02 +0200627 */
Thierry FOURNIER58639a02014-11-25 12:02:25 +0100628struct sockaddr_storage *str2ip2(const char *str, struct sockaddr_storage *sa, int resolve)
Willy Tarreaubaaee002006-06-26 02:48:02 +0200629{
Willy Tarreaufab5a432011-03-04 15:31:53 +0100630 struct hostent *he;
mildisff5d5102015-10-26 18:50:08 +0100631 /* max IPv6 length, including brackets and terminating NULL */
632 char tmpip[48];
633
634 /* check IPv6 with square brackets */
635 if (str[0] == '[') {
636 size_t iplength = strlen(str);
637
638 if (iplength < 4) {
639 /* minimal size is 4 when using brackets "[::]" */
640 goto fail;
641 }
642 else if (iplength >= sizeof(tmpip)) {
643 /* IPv6 literal can not be larger than tmpip */
644 goto fail;
645 }
646 else {
647 if (str[iplength - 1] != ']') {
648 /* if address started with bracket, it should end with bracket */
649 goto fail;
650 }
651 else {
652 memcpy(tmpip, str + 1, iplength - 2);
653 tmpip[iplength - 2] = '\0';
654 str = tmpip;
655 }
656 }
657 }
Willy Tarreaufab5a432011-03-04 15:31:53 +0100658
Willy Tarreaufab5a432011-03-04 15:31:53 +0100659 /* Any IPv6 address */
660 if (str[0] == ':' && str[1] == ':' && !str[2]) {
Willy Tarreau24709282013-03-10 21:32:12 +0100661 if (!sa->ss_family || sa->ss_family == AF_UNSPEC)
662 sa->ss_family = AF_INET6;
663 else if (sa->ss_family != AF_INET6)
664 goto fail;
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100665 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100666 }
667
Willy Tarreau24709282013-03-10 21:32:12 +0100668 /* Any address for the family, defaults to IPv4 */
Willy Tarreaufab5a432011-03-04 15:31:53 +0100669 if (!str[0] || (str[0] == '*' && !str[1])) {
Willy Tarreau24709282013-03-10 21:32:12 +0100670 if (!sa->ss_family || sa->ss_family == AF_UNSPEC)
671 sa->ss_family = AF_INET;
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100672 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100673 }
674
675 /* check for IPv6 first */
Willy Tarreau24709282013-03-10 21:32:12 +0100676 if ((!sa->ss_family || sa->ss_family == AF_UNSPEC || sa->ss_family == AF_INET6) &&
677 inet_pton(AF_INET6, str, &((struct sockaddr_in6 *)sa)->sin6_addr)) {
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100678 sa->ss_family = AF_INET6;
679 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100680 }
681
682 /* then check for IPv4 */
Willy Tarreau24709282013-03-10 21:32:12 +0100683 if ((!sa->ss_family || sa->ss_family == AF_UNSPEC || sa->ss_family == AF_INET) &&
684 inet_pton(AF_INET, str, &((struct sockaddr_in *)sa)->sin_addr)) {
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100685 sa->ss_family = AF_INET;
686 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100687 }
688
Thierry FOURNIER58639a02014-11-25 12:02:25 +0100689 if (!resolve)
690 return NULL;
691
Baptiste Assmanna68ca962015-04-14 01:15:08 +0200692 if (!dns_hostname_validation(str, NULL))
693 return NULL;
694
David du Colombierd5f43282011-03-17 10:40:16 +0100695#ifdef USE_GETADDRINFO
Nenad Merdanovic88afe032014-04-14 15:56:58 +0200696 if (global.tune.options & GTUNE_USE_GAI) {
David du Colombierd5f43282011-03-17 10:40:16 +0100697 struct addrinfo hints, *result;
698
699 memset(&result, 0, sizeof(result));
700 memset(&hints, 0, sizeof(hints));
Willy Tarreau24709282013-03-10 21:32:12 +0100701 hints.ai_family = sa->ss_family ? sa->ss_family : AF_UNSPEC;
David du Colombierd5f43282011-03-17 10:40:16 +0100702 hints.ai_socktype = SOCK_DGRAM;
Dmitry Sivachenkoeab7f392015-10-02 01:01:58 +0200703 hints.ai_flags = 0;
David du Colombierd5f43282011-03-17 10:40:16 +0100704 hints.ai_protocol = 0;
705
706 if (getaddrinfo(str, NULL, &hints, &result) == 0) {
Willy Tarreau24709282013-03-10 21:32:12 +0100707 if (!sa->ss_family || sa->ss_family == AF_UNSPEC)
708 sa->ss_family = result->ai_family;
709 else if (sa->ss_family != result->ai_family)
710 goto fail;
711
David du Colombierd5f43282011-03-17 10:40:16 +0100712 switch (result->ai_family) {
713 case AF_INET:
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100714 memcpy((struct sockaddr_in *)sa, result->ai_addr, result->ai_addrlen);
715 return sa;
David du Colombierd5f43282011-03-17 10:40:16 +0100716 case AF_INET6:
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100717 memcpy((struct sockaddr_in6 *)sa, result->ai_addr, result->ai_addrlen);
718 return sa;
David du Colombierd5f43282011-03-17 10:40:16 +0100719 }
720 }
721
Sean Carey58ea0392013-02-15 23:39:18 +0100722 if (result)
723 freeaddrinfo(result);
Willy Tarreaufab5a432011-03-04 15:31:53 +0100724 }
David du Colombierd5f43282011-03-17 10:40:16 +0100725#endif
Nenad Merdanovic88afe032014-04-14 15:56:58 +0200726 /* try to resolve an IPv4/IPv6 hostname */
727 he = gethostbyname(str);
728 if (he) {
729 if (!sa->ss_family || sa->ss_family == AF_UNSPEC)
730 sa->ss_family = he->h_addrtype;
731 else if (sa->ss_family != he->h_addrtype)
732 goto fail;
733
734 switch (sa->ss_family) {
735 case AF_INET:
736 ((struct sockaddr_in *)sa)->sin_addr = *(struct in_addr *) *(he->h_addr_list);
737 return sa;
738 case AF_INET6:
739 ((struct sockaddr_in6 *)sa)->sin6_addr = *(struct in6_addr *) *(he->h_addr_list);
740 return sa;
741 }
742 }
743
David du Colombierd5f43282011-03-17 10:40:16 +0100744 /* unsupported address family */
Willy Tarreau24709282013-03-10 21:32:12 +0100745 fail:
Willy Tarreaufab5a432011-03-04 15:31:53 +0100746 return NULL;
747}
748
749/*
Willy Tarreaud4448bc2013-02-20 15:55:15 +0100750 * Converts <str> to a locally allocated struct sockaddr_storage *, and a port
751 * range or offset consisting in two integers that the caller will have to
752 * check to find the relevant input format. The following format are supported :
753 *
754 * String format | address | port | low | high
755 * addr | <addr> | 0 | 0 | 0
756 * addr: | <addr> | 0 | 0 | 0
757 * addr:port | <addr> | <port> | <port> | <port>
758 * addr:pl-ph | <addr> | <pl> | <pl> | <ph>
759 * addr:+port | <addr> | <port> | 0 | <port>
760 * addr:-port | <addr> |-<port> | <port> | 0
761 *
762 * The detection of a port range or increment by the caller is made by
763 * comparing <low> and <high>. If both are equal, then port 0 means no port
764 * was specified. The caller may pass NULL for <low> and <high> if it is not
765 * interested in retrieving port ranges.
766 *
767 * Note that <addr> above may also be :
768 * - empty ("") => family will be AF_INET and address will be INADDR_ANY
769 * - "*" => family will be AF_INET and address will be INADDR_ANY
770 * - "::" => family will be AF_INET6 and address will be IN6ADDR_ANY
771 * - a host name => family and address will depend on host name resolving.
772 *
Willy Tarreau24709282013-03-10 21:32:12 +0100773 * A prefix may be passed in before the address above to force the family :
774 * - "ipv4@" => force address to resolve as IPv4 and fail if not possible.
775 * - "ipv6@" => force address to resolve as IPv6 and fail if not possible.
776 * - "unix@" => force address to be a path to a UNIX socket even if the
777 * path does not start with a '/'
Willy Tarreauccfccef2014-05-10 01:49:15 +0200778 * - 'abns@' -> force address to belong to the abstract namespace (Linux
779 * only). These sockets are just like Unix sockets but without
780 * the need for an underlying file system. The address is a
781 * string. Technically it's like a Unix socket with a zero in
782 * the first byte of the address.
Willy Tarreau40aa0702013-03-10 23:51:38 +0100783 * - "fd@" => an integer must follow, and is a file descriptor number.
Willy Tarreau24709282013-03-10 21:32:12 +0100784 *
mildisff5d5102015-10-26 18:50:08 +0100785 * IPv6 addresses can be declared with or without square brackets. When using
786 * square brackets for IPv6 addresses, the port separator (colon) is optional.
787 * If not using square brackets, and in order to avoid any ambiguity with
788 * IPv6 addresses, the last colon ':' is mandatory even when no port is specified.
789 * NULL is returned if the address cannot be parsed. The <low> and <high> ports
790 * are always initialized if non-null, even for non-IP families.
Willy Tarreaud393a622013-03-04 18:22:00 +0100791 *
792 * If <pfx> is non-null, it is used as a string prefix before any path-based
793 * address (typically the path to a unix socket).
Willy Tarreau40aa0702013-03-10 23:51:38 +0100794 *
Willy Tarreau72b8c1f2015-09-08 15:50:19 +0200795 * if <fqdn> is non-null, it will be filled with :
796 * - a pointer to the FQDN of the server name to resolve if there's one, and
797 * that the caller will have to free(),
798 * - NULL if there was an explicit address that doesn't require resolution.
799 *
Thierry FOURNIER7fe3be72015-09-26 20:03:36 +0200800 * Hostnames are only resolved if <resolve> is non-null.
801 *
Willy Tarreau40aa0702013-03-10 23:51:38 +0100802 * When a file descriptor is passed, its value is put into the s_addr part of
803 * the address when cast to sockaddr_in and the address family is AF_UNSPEC.
Willy Tarreaufab5a432011-03-04 15:31:53 +0100804 */
Thierry FOURNIER7fe3be72015-09-26 20:03:36 +0200805struct sockaddr_storage *str2sa_range(const char *str, int *low, int *high, char **err, const char *pfx, char **fqdn, int resolve)
Willy Tarreaufab5a432011-03-04 15:31:53 +0100806{
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100807 static struct sockaddr_storage ss;
David du Colombier6f5ccb12011-03-10 22:26:24 +0100808 struct sockaddr_storage *ret = NULL;
Willy Tarreau24709282013-03-10 21:32:12 +0100809 char *back, *str2;
Willy Tarreaud4448bc2013-02-20 15:55:15 +0100810 char *port1, *port2;
811 int portl, porth, porta;
Willy Tarreauccfccef2014-05-10 01:49:15 +0200812 int abstract = 0;
Willy Tarreaud4448bc2013-02-20 15:55:15 +0100813
814 portl = porth = porta = 0;
Willy Tarreau72b8c1f2015-09-08 15:50:19 +0200815 if (fqdn)
816 *fqdn = NULL;
Willy Tarreaubaaee002006-06-26 02:48:02 +0200817
Willy Tarreaudad36a32013-03-11 01:20:04 +0100818 str2 = back = env_expand(strdup(str));
Willy Tarreaudf350f12013-03-01 20:22:54 +0100819 if (str2 == NULL) {
820 memprintf(err, "out of memory in '%s'\n", __FUNCTION__);
Willy Tarreaud5191e72010-02-09 20:50:45 +0100821 goto out;
Willy Tarreaudf350f12013-03-01 20:22:54 +0100822 }
Willy Tarreaubaaee002006-06-26 02:48:02 +0200823
Willy Tarreau9f69f462015-09-08 16:01:25 +0200824 if (!*str2) {
825 memprintf(err, "'%s' resolves to an empty address (environment variable missing?)\n", str);
826 goto out;
827 }
828
Willy Tarreau24709282013-03-10 21:32:12 +0100829 memset(&ss, 0, sizeof(ss));
830
831 if (strncmp(str2, "unix@", 5) == 0) {
832 str2 += 5;
Willy Tarreauccfccef2014-05-10 01:49:15 +0200833 abstract = 0;
Willy Tarreau24709282013-03-10 21:32:12 +0100834 ss.ss_family = AF_UNIX;
835 }
Willy Tarreauccfccef2014-05-10 01:49:15 +0200836 else if (strncmp(str2, "abns@", 5) == 0) {
837 str2 += 5;
838 abstract = 1;
839 ss.ss_family = AF_UNIX;
840 }
Willy Tarreau24709282013-03-10 21:32:12 +0100841 else if (strncmp(str2, "ipv4@", 5) == 0) {
842 str2 += 5;
843 ss.ss_family = AF_INET;
844 }
845 else if (strncmp(str2, "ipv6@", 5) == 0) {
846 str2 += 5;
847 ss.ss_family = AF_INET6;
848 }
849 else if (*str2 == '/') {
850 ss.ss_family = AF_UNIX;
851 }
852 else
853 ss.ss_family = AF_UNSPEC;
854
Willy Tarreau40aa0702013-03-10 23:51:38 +0100855 if (ss.ss_family == AF_UNSPEC && strncmp(str2, "fd@", 3) == 0) {
856 char *endptr;
857
858 str2 += 3;
859 ((struct sockaddr_in *)&ss)->sin_addr.s_addr = strtol(str2, &endptr, 10);
860
861 if (!*str2 || *endptr) {
Willy Tarreaudad36a32013-03-11 01:20:04 +0100862 memprintf(err, "file descriptor '%s' is not a valid integer in '%s'\n", str2, str);
Willy Tarreau40aa0702013-03-10 23:51:38 +0100863 goto out;
864 }
865
866 /* we return AF_UNSPEC if we use a file descriptor number */
867 ss.ss_family = AF_UNSPEC;
868 }
869 else if (ss.ss_family == AF_UNIX) {
Willy Tarreau15586382013-03-04 19:48:14 +0100870 int prefix_path_len;
871 int max_path_len;
Willy Tarreau94ef3f32014-04-14 14:49:00 +0200872 int adr_len;
Willy Tarreau15586382013-03-04 19:48:14 +0100873
874 /* complete unix socket path name during startup or soft-restart is
875 * <unix_bind_prefix><path>.<pid>.<bak|tmp>
876 */
Willy Tarreauccfccef2014-05-10 01:49:15 +0200877 prefix_path_len = (pfx && !abstract) ? strlen(pfx) : 0;
Willy Tarreau15586382013-03-04 19:48:14 +0100878 max_path_len = (sizeof(((struct sockaddr_un *)&ss)->sun_path) - 1) -
879 (prefix_path_len ? prefix_path_len + 1 + 5 + 1 + 3 : 0);
880
Willy Tarreau94ef3f32014-04-14 14:49:00 +0200881 adr_len = strlen(str2);
882 if (adr_len > max_path_len) {
Willy Tarreau15586382013-03-04 19:48:14 +0100883 memprintf(err, "socket path '%s' too long (max %d)\n", str, max_path_len);
884 goto out;
885 }
886
Willy Tarreauccfccef2014-05-10 01:49:15 +0200887 /* when abstract==1, we skip the first zero and copy all bytes except the trailing zero */
888 memset(((struct sockaddr_un *)&ss)->sun_path, 0, sizeof(((struct sockaddr_un *)&ss)->sun_path));
Willy Tarreau94ef3f32014-04-14 14:49:00 +0200889 if (prefix_path_len)
Willy Tarreau15586382013-03-04 19:48:14 +0100890 memcpy(((struct sockaddr_un *)&ss)->sun_path, pfx, prefix_path_len);
Willy Tarreauccfccef2014-05-10 01:49:15 +0200891 memcpy(((struct sockaddr_un *)&ss)->sun_path + prefix_path_len + abstract, str2, adr_len + 1 - abstract);
Willy Tarreau15586382013-03-04 19:48:14 +0100892 }
Willy Tarreau24709282013-03-10 21:32:12 +0100893 else { /* IPv4 and IPv6 */
Willy Tarreau72b8c1f2015-09-08 15:50:19 +0200894 int use_fqdn = 0;
mildisff5d5102015-10-26 18:50:08 +0100895 char *end = str2 + strlen(str2);
896 char *chr;
Willy Tarreau72b8c1f2015-09-08 15:50:19 +0200897
mildisff5d5102015-10-26 18:50:08 +0100898 /* search for : or ] whatever comes first */
899 for (chr = end-1; chr > str2; chr--) {
900 if (*chr == ']' || *chr == ':')
901 break;
902 }
903
904 if (*chr == ':') {
905 /* Found a colon before a closing-bracket, must be a port separator.
906 * This guarantee backward compatibility.
907 */
908 *chr++ = '\0';
909 port1 = chr;
910 }
911 else {
912 /* Either no colon and no closing-bracket
913 * or directly ending with a closing-bracket.
914 * However, no port.
915 */
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100916 port1 = "";
mildisff5d5102015-10-26 18:50:08 +0100917 }
Willy Tarreaubaaee002006-06-26 02:48:02 +0200918
Willy Tarreau72b8c1f2015-09-08 15:50:19 +0200919 if (str2ip2(str2, &ss, 0) == NULL) {
920 use_fqdn = 1;
Thierry FOURNIER7fe3be72015-09-26 20:03:36 +0200921 if (!resolve || str2ip2(str2, &ss, 1) == NULL) {
Willy Tarreau72b8c1f2015-09-08 15:50:19 +0200922 memprintf(err, "invalid address: '%s' in '%s'\n", str2, str);
923 goto out;
924 }
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100925 }
Willy Tarreaufab5a432011-03-04 15:31:53 +0100926
Willy Tarreaua39d1992013-04-01 20:37:42 +0200927 if (isdigit((int)(unsigned char)*port1)) { /* single port or range */
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100928 port2 = strchr(port1, '-');
929 if (port2)
930 *port2++ = '\0';
931 else
932 port2 = port1;
933 portl = atoi(port1);
934 porth = atoi(port2);
935 porta = portl;
936 }
937 else if (*port1 == '-') { /* negative offset */
938 portl = atoi(port1 + 1);
939 porta = -portl;
940 }
941 else if (*port1 == '+') { /* positive offset */
942 porth = atoi(port1 + 1);
943 porta = porth;
944 }
945 else if (*port1) { /* other any unexpected char */
Willy Tarreaudad36a32013-03-11 01:20:04 +0100946 memprintf(err, "invalid character '%c' in port number '%s' in '%s'\n", *port1, port1, str);
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100947 goto out;
948 }
949 set_host_port(&ss, porta);
Willy Tarreau72b8c1f2015-09-08 15:50:19 +0200950
951 if (use_fqdn && fqdn) {
952 if (str2 != back)
953 memmove(back, str2, strlen(str2) + 1);
954 *fqdn = back;
955 back = NULL;
956 }
Willy Tarreaue4c58c82013-03-06 15:28:17 +0100957 }
Willy Tarreaufab5a432011-03-04 15:31:53 +0100958
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100959 ret = &ss;
Willy Tarreaud5191e72010-02-09 20:50:45 +0100960 out:
Willy Tarreaud4448bc2013-02-20 15:55:15 +0100961 if (low)
962 *low = portl;
963 if (high)
964 *high = porth;
Willy Tarreau24709282013-03-10 21:32:12 +0100965 free(back);
Willy Tarreaud5191e72010-02-09 20:50:45 +0100966 return ret;
Willy Tarreauc6f4ce82009-06-10 11:09:37 +0200967}
968
Willy Tarreau2937c0d2010-01-26 17:36:17 +0100969/* converts <str> to a struct in_addr containing a network mask. It can be
970 * passed in dotted form (255.255.255.0) or in CIDR form (24). It returns 1
971 * if the conversion succeeds otherwise non-zero.
972 */
973int str2mask(const char *str, struct in_addr *mask)
974{
975 if (strchr(str, '.') != NULL) { /* dotted notation */
976 if (!inet_pton(AF_INET, str, mask))
977 return 0;
978 }
979 else { /* mask length */
980 char *err;
981 unsigned long len = strtol(str, &err, 10);
982
983 if (!*str || (err && *err) || (unsigned)len > 32)
984 return 0;
985 if (len)
986 mask->s_addr = htonl(~0UL << (32 - len));
987 else
988 mask->s_addr = 0;
989 }
990 return 1;
991}
992
Thierry FOURNIERb0504632013-12-14 15:39:02 +0100993/* convert <cidr> to struct in_addr <mask>. It returns 1 if the conversion
994 * succeeds otherwise zero.
995 */
996int cidr2dotted(int cidr, struct in_addr *mask) {
997
998 if (cidr < 0 || cidr > 32)
999 return 0;
1000
1001 mask->s_addr = cidr ? htonl(~0UL << (32 - cidr)) : 0;
1002 return 1;
1003}
1004
Thierry Fournier70473a52016-02-17 17:12:14 +01001005/* Convert mask from bit length form to in_addr form.
1006 * This function never fails.
1007 */
1008void len2mask4(int len, struct in_addr *addr)
1009{
1010 if (len >= 32) {
1011 addr->s_addr = 0xffffffff;
1012 return;
1013 }
1014 if (len <= 0) {
1015 addr->s_addr = 0x00000000;
1016 return;
1017 }
1018 addr->s_addr = 0xffffffff << (32 - len);
1019 addr->s_addr = htonl(addr->s_addr);
1020}
1021
1022/* Convert mask from bit length form to in6_addr form.
1023 * This function never fails.
1024 */
1025void len2mask6(int len, struct in6_addr *addr)
1026{
1027 len2mask4(len, (struct in_addr *)&addr->s6_addr[0]); /* msb */
1028 len -= 32;
1029 len2mask4(len, (struct in_addr *)&addr->s6_addr[4]);
1030 len -= 32;
1031 len2mask4(len, (struct in_addr *)&addr->s6_addr[8]);
1032 len -= 32;
1033 len2mask4(len, (struct in_addr *)&addr->s6_addr[12]); /* lsb */
1034}
1035
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02001036/*
Willy Tarreaud077a8e2007-05-08 18:28:09 +02001037 * converts <str> to two struct in_addr* which must be pre-allocated.
Willy Tarreaubaaee002006-06-26 02:48:02 +02001038 * The format is "addr[/mask]", where "addr" cannot be empty, and mask
1039 * is optionnal and either in the dotted or CIDR notation.
1040 * Note: "addr" can also be a hostname. Returns 1 if OK, 0 if error.
1041 */
Thierry FOURNIERfc7ac7b2014-02-11 15:23:04 +01001042int str2net(const char *str, int resolve, struct in_addr *addr, struct in_addr *mask)
Willy Tarreaubaaee002006-06-26 02:48:02 +02001043{
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001044 __label__ out_free, out_err;
1045 char *c, *s;
1046 int ret_val;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001047
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001048 s = strdup(str);
1049 if (!s)
1050 return 0;
1051
Willy Tarreaubaaee002006-06-26 02:48:02 +02001052 memset(mask, 0, sizeof(*mask));
1053 memset(addr, 0, sizeof(*addr));
Willy Tarreaubaaee002006-06-26 02:48:02 +02001054
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001055 if ((c = strrchr(s, '/')) != NULL) {
Willy Tarreaubaaee002006-06-26 02:48:02 +02001056 *c++ = '\0';
1057 /* c points to the mask */
Willy Tarreau2937c0d2010-01-26 17:36:17 +01001058 if (!str2mask(c, mask))
1059 goto out_err;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001060 }
1061 else {
Willy Tarreauebd61602006-12-30 11:54:15 +01001062 mask->s_addr = ~0U;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001063 }
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001064 if (!inet_pton(AF_INET, s, addr)) {
Willy Tarreaubaaee002006-06-26 02:48:02 +02001065 struct hostent *he;
1066
Thierry FOURNIERfc7ac7b2014-02-11 15:23:04 +01001067 if (!resolve)
1068 goto out_err;
1069
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001070 if ((he = gethostbyname(s)) == NULL) {
1071 goto out_err;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001072 }
1073 else
1074 *addr = *(struct in_addr *) *(he->h_addr_list);
1075 }
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001076
1077 ret_val = 1;
1078 out_free:
1079 free(s);
1080 return ret_val;
1081 out_err:
1082 ret_val = 0;
1083 goto out_free;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001084}
1085
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001086
1087/*
Willy Tarreau6d20e282012-04-27 22:49:47 +02001088 * converts <str> to two struct in6_addr* which must be pre-allocated.
1089 * The format is "addr[/mask]", where "addr" cannot be empty, and mask
1090 * is an optionnal number of bits (128 being the default).
1091 * Returns 1 if OK, 0 if error.
1092 */
1093int str62net(const char *str, struct in6_addr *addr, unsigned char *mask)
1094{
1095 char *c, *s;
1096 int ret_val = 0;
1097 char *err;
1098 unsigned long len = 128;
1099
1100 s = strdup(str);
1101 if (!s)
1102 return 0;
1103
1104 memset(mask, 0, sizeof(*mask));
1105 memset(addr, 0, sizeof(*addr));
1106
1107 if ((c = strrchr(s, '/')) != NULL) {
1108 *c++ = '\0'; /* c points to the mask */
1109 if (!*c)
1110 goto out_free;
1111
1112 len = strtoul(c, &err, 10);
1113 if ((err && *err) || (unsigned)len > 128)
1114 goto out_free;
1115 }
1116 *mask = len; /* OK we have a valid mask in <len> */
1117
1118 if (!inet_pton(AF_INET6, s, addr))
1119 goto out_free;
1120
1121 ret_val = 1;
1122 out_free:
1123 free(s);
1124 return ret_val;
1125}
1126
1127
1128/*
David du Colombier6f5ccb12011-03-10 22:26:24 +01001129 * Parse IPv4 address found in url.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001130 */
David du Colombier6f5ccb12011-03-10 22:26:24 +01001131int url2ipv4(const char *addr, struct in_addr *dst)
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001132{
1133 int saw_digit, octets, ch;
1134 u_char tmp[4], *tp;
1135 const char *cp = addr;
1136
1137 saw_digit = 0;
1138 octets = 0;
1139 *(tp = tmp) = 0;
1140
1141 while (*addr) {
1142 unsigned char digit = (ch = *addr++) - '0';
1143 if (digit > 9 && ch != '.')
1144 break;
1145 if (digit <= 9) {
1146 u_int new = *tp * 10 + digit;
1147 if (new > 255)
1148 return 0;
1149 *tp = new;
1150 if (!saw_digit) {
1151 if (++octets > 4)
1152 return 0;
1153 saw_digit = 1;
1154 }
1155 } else if (ch == '.' && saw_digit) {
1156 if (octets == 4)
1157 return 0;
1158 *++tp = 0;
1159 saw_digit = 0;
1160 } else
1161 return 0;
1162 }
1163
1164 if (octets < 4)
1165 return 0;
1166
1167 memcpy(&dst->s_addr, tmp, 4);
1168 return addr-cp-1;
1169}
1170
1171/*
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001172 * Resolve destination server from URL. Convert <str> to a sockaddr_storage.
1173 * <out> contain the code of the dectected scheme, the start and length of
1174 * the hostname. Actually only http and https are supported. <out> can be NULL.
1175 * This function returns the consumed length. It is useful if you parse complete
1176 * url like http://host:port/path, because the consumed length corresponds to
1177 * the first character of the path. If the conversion fails, it returns -1.
1178 *
1179 * This function tries to resolve the DNS name if haproxy is in starting mode.
1180 * So, this function may be used during the configuration parsing.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001181 */
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001182int url2sa(const char *url, int ulen, struct sockaddr_storage *addr, struct split_url *out)
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001183{
1184 const char *curr = url, *cp = url;
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001185 const char *end;
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001186 int ret, url_code = 0;
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001187 unsigned long long int http_code = 0;
1188 int default_port;
1189 struct hostent *he;
1190 char *p;
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001191
1192 /* Firstly, try to find :// pattern */
1193 while (curr < url+ulen && url_code != 0x3a2f2f) {
1194 url_code = ((url_code & 0xffff) << 8);
1195 url_code += (unsigned char)*curr++;
1196 }
1197
1198 /* Secondly, if :// pattern is found, verify parsed stuff
1199 * before pattern is matching our http pattern.
1200 * If so parse ip address and port in uri.
1201 *
1202 * WARNING: Current code doesn't support dynamic async dns resolver.
1203 */
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001204 if (url_code != 0x3a2f2f)
1205 return -1;
1206
1207 /* Copy scheme, and utrn to lower case. */
1208 while (cp < curr - 3)
1209 http_code = (http_code << 8) + *cp++;
1210 http_code |= 0x2020202020202020ULL; /* Turn everything to lower case */
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001211
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001212 /* HTTP or HTTPS url matching */
1213 if (http_code == 0x2020202068747470ULL) {
1214 default_port = 80;
1215 if (out)
1216 out->scheme = SCH_HTTP;
1217 }
1218 else if (http_code == 0x2020206874747073ULL) {
1219 default_port = 443;
1220 if (out)
1221 out->scheme = SCH_HTTPS;
1222 }
1223 else
1224 return -1;
1225
1226 /* If the next char is '[', the host address is IPv6. */
1227 if (*curr == '[') {
1228 curr++;
1229
1230 /* Check trash size */
1231 if (trash.size < ulen)
1232 return -1;
1233
1234 /* Look for ']' and copy the address in a trash buffer. */
1235 p = trash.str;
1236 for (end = curr;
1237 end < url + ulen && *end != ']';
1238 end++, p++)
1239 *p = *end;
1240 if (*end != ']')
1241 return -1;
1242 *p = '\0';
1243
1244 /* Update out. */
1245 if (out) {
1246 out->host = curr;
1247 out->host_len = end - curr;
1248 }
1249
1250 /* Try IPv6 decoding. */
1251 if (!inet_pton(AF_INET6, trash.str, &((struct sockaddr_in6 *)addr)->sin6_addr))
1252 return -1;
1253 end++;
1254
1255 /* Decode port. */
1256 if (*end == ':') {
1257 end++;
1258 default_port = read_uint(&end, url + ulen);
1259 }
1260 ((struct sockaddr_in6 *)addr)->sin6_port = htons(default_port);
1261 ((struct sockaddr_in6 *)addr)->sin6_family = AF_INET6;
1262 return end - url;
1263 }
1264 else {
1265 /* We are looking for IP address. If you want to parse and
1266 * resolve hostname found in url, you can use str2sa_range(), but
1267 * be warned this can slow down global daemon performances
1268 * while handling lagging dns responses.
1269 */
1270 ret = url2ipv4(curr, &((struct sockaddr_in *)addr)->sin_addr);
1271 if (ret) {
1272 /* Update out. */
1273 if (out) {
1274 out->host = curr;
1275 out->host_len = ret;
1276 }
1277
1278 curr += ret;
1279
1280 /* Decode port. */
1281 if (*curr == ':') {
1282 curr++;
1283 default_port = read_uint(&curr, url + ulen);
1284 }
1285 ((struct sockaddr_in *)addr)->sin_port = htons(default_port);
1286
1287 /* Set family. */
1288 ((struct sockaddr_in *)addr)->sin_family = AF_INET;
1289 return curr - url;
1290 }
1291 else if (global.mode & MODE_STARTING) {
1292 /* The IPv4 and IPv6 decoding fails, maybe the url contain name. Try to execute
1293 * synchronous DNS request only if HAProxy is in the start state.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001294 */
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001295
1296 /* look for : or / or end */
1297 for (end = curr;
1298 end < url + ulen && *end != '/' && *end != ':';
1299 end++);
1300 memcpy(trash.str, curr, end - curr);
1301 trash.str[end - curr] = '\0';
1302
1303 /* try to resolve an IPv4/IPv6 hostname */
1304 he = gethostbyname(trash.str);
1305 if (!he)
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001306 return -1;
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001307
1308 /* Update out. */
1309 if (out) {
1310 out->host = curr;
1311 out->host_len = end - curr;
1312 }
1313
1314 /* Decode port. */
1315 if (*end == ':') {
1316 end++;
1317 default_port = read_uint(&end, url + ulen);
1318 }
1319
1320 /* Copy IP address, set port and family. */
1321 switch (he->h_addrtype) {
1322 case AF_INET:
1323 ((struct sockaddr_in *)addr)->sin_addr = *(struct in_addr *) *(he->h_addr_list);
1324 ((struct sockaddr_in *)addr)->sin_port = htons(default_port);
1325 ((struct sockaddr_in *)addr)->sin_family = AF_INET;
1326 return end - url;
1327
1328 case AF_INET6:
1329 ((struct sockaddr_in6 *)addr)->sin6_addr = *(struct in6_addr *) *(he->h_addr_list);
1330 ((struct sockaddr_in6 *)addr)->sin6_port = htons(default_port);
1331 ((struct sockaddr_in6 *)addr)->sin6_family = AF_INET6;
1332 return end - url;
1333 }
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001334 }
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001335 }
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001336 return -1;
1337}
1338
Willy Tarreau631f01c2011-09-05 00:36:48 +02001339/* Tries to convert a sockaddr_storage address to text form. Upon success, the
1340 * address family is returned so that it's easy for the caller to adapt to the
1341 * output format. Zero is returned if the address family is not supported. -1
1342 * is returned upon error, with errno set. AF_INET, AF_INET6 and AF_UNIX are
1343 * supported.
1344 */
1345int addr_to_str(struct sockaddr_storage *addr, char *str, int size)
1346{
1347
1348 void *ptr;
1349
1350 if (size < 5)
1351 return 0;
1352 *str = '\0';
1353
1354 switch (addr->ss_family) {
1355 case AF_INET:
1356 ptr = &((struct sockaddr_in *)addr)->sin_addr;
1357 break;
1358 case AF_INET6:
1359 ptr = &((struct sockaddr_in6 *)addr)->sin6_addr;
1360 break;
1361 case AF_UNIX:
1362 memcpy(str, "unix", 5);
1363 return addr->ss_family;
1364 default:
1365 return 0;
1366 }
1367
1368 if (inet_ntop(addr->ss_family, ptr, str, size))
1369 return addr->ss_family;
1370
1371 /* failed */
1372 return -1;
1373}
1374
Simon Horman75ab8bd2014-06-16 09:39:41 +09001375/* Tries to convert a sockaddr_storage port to text form. Upon success, the
1376 * address family is returned so that it's easy for the caller to adapt to the
1377 * output format. Zero is returned if the address family is not supported. -1
1378 * is returned upon error, with errno set. AF_INET, AF_INET6 and AF_UNIX are
1379 * supported.
1380 */
1381int port_to_str(struct sockaddr_storage *addr, char *str, int size)
1382{
1383
1384 uint16_t port;
1385
1386
1387 if (size < 5)
1388 return 0;
1389 *str = '\0';
1390
1391 switch (addr->ss_family) {
1392 case AF_INET:
1393 port = ((struct sockaddr_in *)addr)->sin_port;
1394 break;
1395 case AF_INET6:
1396 port = ((struct sockaddr_in6 *)addr)->sin6_port;
1397 break;
1398 case AF_UNIX:
1399 memcpy(str, "unix", 5);
1400 return addr->ss_family;
1401 default:
1402 return 0;
1403 }
1404
1405 snprintf(str, size, "%u", ntohs(port));
1406 return addr->ss_family;
1407}
1408
Willy Tarreaubaaee002006-06-26 02:48:02 +02001409/* will try to encode the string <string> replacing all characters tagged in
1410 * <map> with the hexadecimal representation of their ASCII-code (2 digits)
1411 * prefixed by <escape>, and will store the result between <start> (included)
1412 * and <stop> (excluded), and will always terminate the string with a '\0'
1413 * before <stop>. The position of the '\0' is returned if the conversion
1414 * completes. If bytes are missing between <start> and <stop>, then the
1415 * conversion will be incomplete and truncated. If <stop> <= <start>, the '\0'
1416 * cannot even be stored so we return <start> without writing the 0.
1417 * The input string must also be zero-terminated.
1418 */
1419const char hextab[16] = "0123456789ABCDEF";
1420char *encode_string(char *start, char *stop,
1421 const char escape, const fd_set *map,
1422 const char *string)
1423{
1424 if (start < stop) {
1425 stop--; /* reserve one byte for the final '\0' */
1426 while (start < stop && *string != '\0') {
1427 if (!FD_ISSET((unsigned char)(*string), map))
1428 *start++ = *string;
1429 else {
1430 if (start + 3 >= stop)
1431 break;
1432 *start++ = escape;
1433 *start++ = hextab[(*string >> 4) & 15];
1434 *start++ = hextab[*string & 15];
1435 }
1436 string++;
1437 }
1438 *start = '\0';
1439 }
1440 return start;
1441}
1442
Thierry FOURNIERe059ec92014-03-17 12:01:13 +01001443/*
1444 * Same behavior as encode_string() above, except that it encodes chunk
1445 * <chunk> instead of a string.
1446 */
1447char *encode_chunk(char *start, char *stop,
1448 const char escape, const fd_set *map,
1449 const struct chunk *chunk)
1450{
1451 char *str = chunk->str;
1452 char *end = chunk->str + chunk->len;
1453
1454 if (start < stop) {
1455 stop--; /* reserve one byte for the final '\0' */
1456 while (start < stop && str < end) {
1457 if (!FD_ISSET((unsigned char)(*str), map))
1458 *start++ = *str;
1459 else {
1460 if (start + 3 >= stop)
1461 break;
1462 *start++ = escape;
1463 *start++ = hextab[(*str >> 4) & 15];
1464 *start++ = hextab[*str & 15];
1465 }
1466 str++;
1467 }
1468 *start = '\0';
1469 }
1470 return start;
1471}
1472
Dragan Dosen0edd1092016-02-12 13:23:02 +01001473/*
1474 * Tries to prefix characters tagged in the <map> with the <escape>
Dragan Dosen1a5d0602016-07-22 16:00:31 +02001475 * character. The input <string> must be zero-terminated. The result will
1476 * be stored between <start> (included) and <stop> (excluded). This
1477 * function will always try to terminate the resulting string with a '\0'
1478 * before <stop>, and will return its position if the conversion
1479 * completes.
1480 */
1481char *escape_string(char *start, char *stop,
1482 const char escape, const fd_set *map,
1483 const char *string)
1484{
1485 if (start < stop) {
1486 stop--; /* reserve one byte for the final '\0' */
1487 while (start < stop && *string != '\0') {
1488 if (!FD_ISSET((unsigned char)(*string), map))
1489 *start++ = *string;
1490 else {
1491 if (start + 2 >= stop)
1492 break;
1493 *start++ = escape;
1494 *start++ = *string;
1495 }
1496 string++;
1497 }
1498 *start = '\0';
1499 }
1500 return start;
1501}
1502
1503/*
1504 * Tries to prefix characters tagged in the <map> with the <escape>
Dragan Dosen0edd1092016-02-12 13:23:02 +01001505 * character. <chunk> contains the input to be escaped. The result will be
1506 * stored between <start> (included) and <stop> (excluded). The function
1507 * will always try to terminate the resulting string with a '\0' before
1508 * <stop>, and will return its position if the conversion completes.
1509 */
1510char *escape_chunk(char *start, char *stop,
1511 const char escape, const fd_set *map,
1512 const struct chunk *chunk)
1513{
1514 char *str = chunk->str;
1515 char *end = chunk->str + chunk->len;
1516
1517 if (start < stop) {
1518 stop--; /* reserve one byte for the final '\0' */
1519 while (start < stop && str < end) {
1520 if (!FD_ISSET((unsigned char)(*str), map))
1521 *start++ = *str;
1522 else {
1523 if (start + 2 >= stop)
1524 break;
1525 *start++ = escape;
1526 *start++ = *str;
1527 }
1528 str++;
1529 }
1530 *start = '\0';
1531 }
1532 return start;
1533}
1534
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001535/* Check a string for using it in a CSV output format. If the string contains
1536 * one of the following four char <">, <,>, CR or LF, the string is
1537 * encapsulated between <"> and the <"> are escaped by a <""> sequence.
1538 * <str> is the input string to be escaped. The function assumes that
1539 * the input string is null-terminated.
1540 *
1541 * If <quote> is 0, the result is returned escaped but without double quote.
Willy Tarreau898529b2016-01-06 18:07:04 +01001542 * It is useful if the escaped string is used between double quotes in the
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001543 * format.
1544 *
Willy Tarreau898529b2016-01-06 18:07:04 +01001545 * printf("..., \"%s\", ...\r\n", csv_enc(str, 0, &trash));
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001546 *
Willy Tarreaub631c292016-01-08 10:04:08 +01001547 * If <quote> is 1, the converter puts the quotes only if any reserved character
1548 * is present. If <quote> is 2, the converter always puts the quotes.
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001549 *
Willy Tarreau898529b2016-01-06 18:07:04 +01001550 * <output> is a struct chunk used for storing the output string.
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001551 *
Willy Tarreau898529b2016-01-06 18:07:04 +01001552 * The function returns the converted string on its output. If an error
1553 * occurs, the function returns an empty string. This type of output is useful
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001554 * for using the function directly as printf() argument.
1555 *
1556 * If the output buffer is too short to contain the input string, the result
1557 * is truncated.
Willy Tarreau898529b2016-01-06 18:07:04 +01001558 *
Willy Tarreaub631c292016-01-08 10:04:08 +01001559 * This function appends the encoding to the existing output chunk, and it
1560 * guarantees that it starts immediately at the first available character of
1561 * the chunk. Please use csv_enc() instead if you want to replace the output
1562 * chunk.
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001563 */
Willy Tarreau898529b2016-01-06 18:07:04 +01001564const char *csv_enc_append(const char *str, int quote, struct chunk *output)
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001565{
1566 char *end = output->str + output->size;
Willy Tarreaub631c292016-01-08 10:04:08 +01001567 char *out = output->str + output->len;
Willy Tarreau898529b2016-01-06 18:07:04 +01001568 char *ptr = out;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001569
Willy Tarreaub631c292016-01-08 10:04:08 +01001570 if (quote == 1) {
1571 /* automatic quoting: first verify if we'll have to quote the string */
1572 if (!strpbrk(str, "\n\r,\""))
1573 quote = 0;
1574 }
1575
1576 if (quote)
1577 *ptr++ = '"';
1578
Willy Tarreau898529b2016-01-06 18:07:04 +01001579 while (*str && ptr < end - 2) { /* -2 for reserving space for <"> and \0. */
1580 *ptr = *str;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001581 if (*str == '"') {
Willy Tarreau898529b2016-01-06 18:07:04 +01001582 ptr++;
1583 if (ptr >= end - 2) {
1584 ptr--;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001585 break;
1586 }
Willy Tarreau898529b2016-01-06 18:07:04 +01001587 *ptr = '"';
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001588 }
Willy Tarreau898529b2016-01-06 18:07:04 +01001589 ptr++;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001590 str++;
1591 }
1592
Willy Tarreaub631c292016-01-08 10:04:08 +01001593 if (quote)
1594 *ptr++ = '"';
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001595
Willy Tarreau898529b2016-01-06 18:07:04 +01001596 *ptr = '\0';
1597 output->len = ptr - output->str;
1598 return out;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001599}
1600
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001601/* Decode an URL-encoded string in-place. The resulting string might
1602 * be shorter. If some forbidden characters are found, the conversion is
Thierry FOURNIER5068d962013-10-04 16:27:27 +02001603 * aborted, the string is truncated before the issue and a negative value is
1604 * returned, otherwise the operation returns the length of the decoded string.
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001605 */
1606int url_decode(char *string)
1607{
1608 char *in, *out;
Thierry FOURNIER5068d962013-10-04 16:27:27 +02001609 int ret = -1;
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001610
1611 in = string;
1612 out = string;
1613 while (*in) {
1614 switch (*in) {
1615 case '+' :
1616 *out++ = ' ';
1617 break;
1618 case '%' :
1619 if (!ishex(in[1]) || !ishex(in[2]))
1620 goto end;
1621 *out++ = (hex2i(in[1]) << 4) + hex2i(in[2]);
1622 in += 2;
1623 break;
1624 default:
1625 *out++ = *in;
1626 break;
1627 }
1628 in++;
1629 }
Thierry FOURNIER5068d962013-10-04 16:27:27 +02001630 ret = out - string; /* success */
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001631 end:
1632 *out = 0;
1633 return ret;
1634}
Willy Tarreaubaaee002006-06-26 02:48:02 +02001635
Willy Tarreau6911fa42007-03-04 18:06:08 +01001636unsigned int str2ui(const char *s)
1637{
1638 return __str2ui(s);
1639}
1640
1641unsigned int str2uic(const char *s)
1642{
1643 return __str2uic(s);
1644}
1645
1646unsigned int strl2ui(const char *s, int len)
1647{
1648 return __strl2ui(s, len);
1649}
1650
1651unsigned int strl2uic(const char *s, int len)
1652{
1653 return __strl2uic(s, len);
1654}
1655
Willy Tarreau4ec83cd2010-10-15 23:19:55 +02001656unsigned int read_uint(const char **s, const char *end)
1657{
1658 return __read_uint(s, end);
1659}
1660
Thierry FOURNIER763a5d82015-07-06 23:09:52 +02001661/* This function reads an unsigned integer from the string pointed to by <s> and
1662 * returns it. The <s> pointer is adjusted to point to the first unread char. The
1663 * function automatically stops at <end>. If the number overflows, the 2^64-1
1664 * value is returned.
1665 */
1666unsigned long long int read_uint64(const char **s, const char *end)
1667{
1668 const char *ptr = *s;
1669 unsigned long long int i = 0, tmp;
1670 unsigned int j;
1671
1672 while (ptr < end) {
1673
1674 /* read next char */
1675 j = *ptr - '0';
1676 if (j > 9)
1677 goto read_uint64_end;
1678
1679 /* add char to the number and check overflow. */
1680 tmp = i * 10;
1681 if (tmp / 10 != i) {
1682 i = ULLONG_MAX;
1683 goto read_uint64_eat;
1684 }
1685 if (ULLONG_MAX - tmp < j) {
1686 i = ULLONG_MAX;
1687 goto read_uint64_eat;
1688 }
1689 i = tmp + j;
1690 ptr++;
1691 }
1692read_uint64_eat:
1693 /* eat each numeric char */
1694 while (ptr < end) {
1695 if ((unsigned int)(*ptr - '0') > 9)
1696 break;
1697 ptr++;
1698 }
1699read_uint64_end:
1700 *s = ptr;
1701 return i;
1702}
1703
1704/* This function reads an integer from the string pointed to by <s> and returns
1705 * it. The <s> pointer is adjusted to point to the first unread char. The function
1706 * automatically stops at <end>. Il the number is bigger than 2^63-2, the 2^63-1
1707 * value is returned. If the number is lowest than -2^63-1, the -2^63 value is
1708 * returned.
1709 */
1710long long int read_int64(const char **s, const char *end)
1711{
1712 unsigned long long int i = 0;
1713 int neg = 0;
1714
1715 /* Look for minus char. */
1716 if (**s == '-') {
1717 neg = 1;
1718 (*s)++;
1719 }
1720 else if (**s == '+')
1721 (*s)++;
1722
1723 /* convert as positive number. */
1724 i = read_uint64(s, end);
1725
1726 if (neg) {
1727 if (i > 0x8000000000000000ULL)
1728 return LLONG_MIN;
1729 return -i;
1730 }
1731 if (i > 0x7fffffffffffffffULL)
1732 return LLONG_MAX;
1733 return i;
1734}
1735
Willy Tarreau6911fa42007-03-04 18:06:08 +01001736/* This one is 7 times faster than strtol() on athlon with checks.
1737 * It returns the value of the number composed of all valid digits read,
1738 * and can process negative numbers too.
1739 */
1740int strl2ic(const char *s, int len)
1741{
1742 int i = 0;
Willy Tarreau3f0c9762007-10-25 09:42:24 +02001743 int j, k;
Willy Tarreau6911fa42007-03-04 18:06:08 +01001744
1745 if (len > 0) {
1746 if (*s != '-') {
1747 /* positive number */
1748 while (len-- > 0) {
1749 j = (*s++) - '0';
Willy Tarreau3f0c9762007-10-25 09:42:24 +02001750 k = i * 10;
Willy Tarreau6911fa42007-03-04 18:06:08 +01001751 if (j > 9)
1752 break;
Willy Tarreau3f0c9762007-10-25 09:42:24 +02001753 i = k + j;
Willy Tarreau6911fa42007-03-04 18:06:08 +01001754 }
1755 } else {
1756 /* negative number */
1757 s++;
1758 while (--len > 0) {
1759 j = (*s++) - '0';
Willy Tarreau3f0c9762007-10-25 09:42:24 +02001760 k = i * 10;
Willy Tarreau6911fa42007-03-04 18:06:08 +01001761 if (j > 9)
1762 break;
Willy Tarreau3f0c9762007-10-25 09:42:24 +02001763 i = k - j;
Willy Tarreau6911fa42007-03-04 18:06:08 +01001764 }
1765 }
1766 }
1767 return i;
1768}
1769
1770
1771/* This function reads exactly <len> chars from <s> and converts them to a
1772 * signed integer which it stores into <ret>. It accurately detects any error
1773 * (truncated string, invalid chars, overflows). It is meant to be used in
1774 * applications designed for hostile environments. It returns zero when the
1775 * number has successfully been converted, non-zero otherwise. When an error
1776 * is returned, the <ret> value is left untouched. It is yet 5 to 40 times
1777 * faster than strtol().
1778 */
1779int strl2irc(const char *s, int len, int *ret)
1780{
1781 int i = 0;
1782 int j;
1783
1784 if (!len)
1785 return 1;
1786
1787 if (*s != '-') {
1788 /* positive number */
1789 while (len-- > 0) {
1790 j = (*s++) - '0';
1791 if (j > 9) return 1; /* invalid char */
1792 if (i > INT_MAX / 10) return 1; /* check for multiply overflow */
1793 i = i * 10;
1794 if (i + j < i) return 1; /* check for addition overflow */
1795 i = i + j;
1796 }
1797 } else {
1798 /* negative number */
1799 s++;
1800 while (--len > 0) {
1801 j = (*s++) - '0';
1802 if (j > 9) return 1; /* invalid char */
1803 if (i < INT_MIN / 10) return 1; /* check for multiply overflow */
1804 i = i * 10;
1805 if (i - j > i) return 1; /* check for subtract overflow */
1806 i = i - j;
1807 }
1808 }
1809 *ret = i;
1810 return 0;
1811}
1812
1813
1814/* This function reads exactly <len> chars from <s> and converts them to a
1815 * signed integer which it stores into <ret>. It accurately detects any error
1816 * (truncated string, invalid chars, overflows). It is meant to be used in
1817 * applications designed for hostile environments. It returns zero when the
1818 * number has successfully been converted, non-zero otherwise. When an error
1819 * is returned, the <ret> value is left untouched. It is about 3 times slower
1820 * than str2irc().
1821 */
Willy Tarreau6911fa42007-03-04 18:06:08 +01001822
1823int strl2llrc(const char *s, int len, long long *ret)
1824{
1825 long long i = 0;
1826 int j;
1827
1828 if (!len)
1829 return 1;
1830
1831 if (*s != '-') {
1832 /* positive number */
1833 while (len-- > 0) {
1834 j = (*s++) - '0';
1835 if (j > 9) return 1; /* invalid char */
1836 if (i > LLONG_MAX / 10LL) return 1; /* check for multiply overflow */
1837 i = i * 10LL;
1838 if (i + j < i) return 1; /* check for addition overflow */
1839 i = i + j;
1840 }
1841 } else {
1842 /* negative number */
1843 s++;
1844 while (--len > 0) {
1845 j = (*s++) - '0';
1846 if (j > 9) return 1; /* invalid char */
1847 if (i < LLONG_MIN / 10LL) return 1; /* check for multiply overflow */
1848 i = i * 10LL;
1849 if (i - j > i) return 1; /* check for subtract overflow */
1850 i = i - j;
1851 }
1852 }
1853 *ret = i;
1854 return 0;
1855}
1856
Thierry FOURNIER511e9472014-01-23 17:40:34 +01001857/* This function is used with pat_parse_dotted_ver(). It converts a string
1858 * composed by two number separated by a dot. Each part must contain in 16 bits
1859 * because internally they will be represented as a 32-bit quantity stored in
1860 * a 64-bit integer. It returns zero when the number has successfully been
1861 * converted, non-zero otherwise. When an error is returned, the <ret> value
1862 * is left untouched.
1863 *
1864 * "1.3" -> 0x0000000000010003
1865 * "65535.65535" -> 0x00000000ffffffff
1866 */
1867int strl2llrc_dotted(const char *text, int len, long long *ret)
1868{
1869 const char *end = &text[len];
1870 const char *p;
1871 long long major, minor;
1872
1873 /* Look for dot. */
1874 for (p = text; p < end; p++)
1875 if (*p == '.')
1876 break;
1877
1878 /* Convert major. */
1879 if (strl2llrc(text, p - text, &major) != 0)
1880 return 1;
1881
1882 /* Check major. */
1883 if (major >= 65536)
1884 return 1;
1885
1886 /* Convert minor. */
1887 minor = 0;
1888 if (p < end)
1889 if (strl2llrc(p + 1, end - (p + 1), &minor) != 0)
1890 return 1;
1891
1892 /* Check minor. */
1893 if (minor >= 65536)
1894 return 1;
1895
1896 /* Compose value. */
1897 *ret = (major << 16) | (minor & 0xffff);
1898 return 0;
1899}
1900
Willy Tarreaua0d37b62007-12-02 22:00:35 +01001901/* This function parses a time value optionally followed by a unit suffix among
1902 * "d", "h", "m", "s", "ms" or "us". It converts the value into the unit
1903 * expected by the caller. The computation does its best to avoid overflows.
1904 * The value is returned in <ret> if everything is fine, and a NULL is returned
1905 * by the function. In case of error, a pointer to the error is returned and
1906 * <ret> is left untouched. Values are automatically rounded up when needed.
1907 */
1908const char *parse_time_err(const char *text, unsigned *ret, unsigned unit_flags)
1909{
1910 unsigned imult, idiv;
1911 unsigned omult, odiv;
1912 unsigned value;
1913
1914 omult = odiv = 1;
1915
1916 switch (unit_flags & TIME_UNIT_MASK) {
1917 case TIME_UNIT_US: omult = 1000000; break;
1918 case TIME_UNIT_MS: omult = 1000; break;
1919 case TIME_UNIT_S: break;
1920 case TIME_UNIT_MIN: odiv = 60; break;
1921 case TIME_UNIT_HOUR: odiv = 3600; break;
1922 case TIME_UNIT_DAY: odiv = 86400; break;
1923 default: break;
1924 }
1925
1926 value = 0;
1927
1928 while (1) {
1929 unsigned int j;
1930
1931 j = *text - '0';
1932 if (j > 9)
1933 break;
1934 text++;
1935 value *= 10;
1936 value += j;
1937 }
1938
1939 imult = idiv = 1;
1940 switch (*text) {
1941 case '\0': /* no unit = default unit */
1942 imult = omult = idiv = odiv = 1;
1943 break;
1944 case 's': /* second = unscaled unit */
1945 break;
1946 case 'u': /* microsecond : "us" */
1947 if (text[1] == 's') {
1948 idiv = 1000000;
1949 text++;
1950 }
1951 break;
1952 case 'm': /* millisecond : "ms" or minute: "m" */
1953 if (text[1] == 's') {
1954 idiv = 1000;
1955 text++;
1956 } else
1957 imult = 60;
1958 break;
1959 case 'h': /* hour : "h" */
1960 imult = 3600;
1961 break;
1962 case 'd': /* day : "d" */
1963 imult = 86400;
1964 break;
1965 default:
1966 return text;
1967 break;
1968 }
1969
1970 if (omult % idiv == 0) { omult /= idiv; idiv = 1; }
1971 if (idiv % omult == 0) { idiv /= omult; omult = 1; }
1972 if (imult % odiv == 0) { imult /= odiv; odiv = 1; }
1973 if (odiv % imult == 0) { odiv /= imult; imult = 1; }
1974
1975 value = (value * (imult * omult) + (idiv * odiv - 1)) / (idiv * odiv);
1976 *ret = value;
1977 return NULL;
1978}
Willy Tarreau6911fa42007-03-04 18:06:08 +01001979
Emeric Brun39132b22010-01-04 14:57:24 +01001980/* this function converts the string starting at <text> to an unsigned int
1981 * stored in <ret>. If an error is detected, the pointer to the unexpected
1982 * character is returned. If the conversio is succesful, NULL is returned.
1983 */
1984const char *parse_size_err(const char *text, unsigned *ret) {
1985 unsigned value = 0;
1986
1987 while (1) {
1988 unsigned int j;
1989
1990 j = *text - '0';
1991 if (j > 9)
1992 break;
1993 if (value > ~0U / 10)
1994 return text;
1995 value *= 10;
1996 if (value > (value + j))
1997 return text;
1998 value += j;
1999 text++;
2000 }
2001
2002 switch (*text) {
2003 case '\0':
2004 break;
2005 case 'K':
2006 case 'k':
2007 if (value > ~0U >> 10)
2008 return text;
2009 value = value << 10;
2010 break;
2011 case 'M':
2012 case 'm':
2013 if (value > ~0U >> 20)
2014 return text;
2015 value = value << 20;
2016 break;
2017 case 'G':
2018 case 'g':
2019 if (value > ~0U >> 30)
2020 return text;
2021 value = value << 30;
2022 break;
2023 default:
2024 return text;
2025 }
2026
Godbach58048a22015-01-28 17:36:16 +08002027 if (*text != '\0' && *++text != '\0')
2028 return text;
2029
Emeric Brun39132b22010-01-04 14:57:24 +01002030 *ret = value;
2031 return NULL;
2032}
2033
Willy Tarreau126d4062013-12-03 17:50:47 +01002034/*
2035 * Parse binary string written in hexadecimal (source) and store the decoded
2036 * result into binstr and set binstrlen to the lengh of binstr. Memory for
2037 * binstr is allocated by the function. In case of error, returns 0 with an
Thierry FOURNIERee330af2014-01-21 11:36:14 +01002038 * error message in err. In succes case, it returns the consumed length.
Willy Tarreau126d4062013-12-03 17:50:47 +01002039 */
2040int parse_binary(const char *source, char **binstr, int *binstrlen, char **err)
2041{
2042 int len;
2043 const char *p = source;
2044 int i,j;
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002045 int alloc;
Willy Tarreau126d4062013-12-03 17:50:47 +01002046
2047 len = strlen(source);
2048 if (len % 2) {
2049 memprintf(err, "an even number of hex digit is expected");
2050 return 0;
2051 }
2052
2053 len = len >> 1;
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002054
Willy Tarreau126d4062013-12-03 17:50:47 +01002055 if (!*binstr) {
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002056 *binstr = calloc(len, sizeof(char));
2057 if (!*binstr) {
2058 memprintf(err, "out of memory while loading string pattern");
2059 return 0;
2060 }
2061 alloc = 1;
Willy Tarreau126d4062013-12-03 17:50:47 +01002062 }
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002063 else {
2064 if (*binstrlen < len) {
2065 memprintf(err, "no space avalaible in the buffer. expect %d, provides %d",
2066 len, *binstrlen);
2067 return 0;
2068 }
2069 alloc = 0;
2070 }
2071 *binstrlen = len;
Willy Tarreau126d4062013-12-03 17:50:47 +01002072
2073 i = j = 0;
2074 while (j < len) {
2075 if (!ishex(p[i++]))
2076 goto bad_input;
2077 if (!ishex(p[i++]))
2078 goto bad_input;
2079 (*binstr)[j++] = (hex2i(p[i-2]) << 4) + hex2i(p[i-1]);
2080 }
Thierry FOURNIERee330af2014-01-21 11:36:14 +01002081 return len << 1;
Willy Tarreau126d4062013-12-03 17:50:47 +01002082
2083bad_input:
2084 memprintf(err, "an hex digit is expected (found '%c')", p[i-1]);
Andreas Seltenreich93f91c32016-03-03 20:40:37 +01002085 if (alloc) {
2086 free(*binstr);
2087 *binstr = NULL;
2088 }
Willy Tarreau126d4062013-12-03 17:50:47 +01002089 return 0;
2090}
2091
Willy Tarreau946ba592009-05-10 15:41:18 +02002092/* copies at most <n> characters from <src> and always terminates with '\0' */
2093char *my_strndup(const char *src, int n)
2094{
2095 int len = 0;
2096 char *ret;
2097
2098 while (len < n && src[len])
2099 len++;
2100
Vincent Bernat3c2f2f22016-04-03 13:48:42 +02002101 ret = malloc(len + 1);
Willy Tarreau946ba592009-05-10 15:41:18 +02002102 if (!ret)
2103 return ret;
2104 memcpy(ret, src, len);
2105 ret[len] = '\0';
2106 return ret;
2107}
2108
Baptiste Assmannbb77c8e2013-10-06 23:24:13 +02002109/*
2110 * search needle in haystack
2111 * returns the pointer if found, returns NULL otherwise
2112 */
2113const void *my_memmem(const void *haystack, size_t haystacklen, const void *needle, size_t needlelen)
2114{
2115 const void *c = NULL;
2116 unsigned char f;
2117
2118 if ((haystack == NULL) || (needle == NULL) || (haystacklen < needlelen))
2119 return NULL;
2120
2121 f = *(char *)needle;
2122 c = haystack;
2123 while ((c = memchr(c, f, haystacklen - (c - haystack))) != NULL) {
2124 if ((haystacklen - (c - haystack)) < needlelen)
2125 return NULL;
2126
2127 if (memcmp(c, needle, needlelen) == 0)
2128 return c;
2129 ++c;
2130 }
2131 return NULL;
2132}
2133
Willy Tarreau482b00d2009-10-04 22:48:42 +02002134/* This function returns the first unused key greater than or equal to <key> in
2135 * ID tree <root>. Zero is returned if no place is found.
2136 */
2137unsigned int get_next_id(struct eb_root *root, unsigned int key)
2138{
2139 struct eb32_node *used;
2140
2141 do {
2142 used = eb32_lookup_ge(root, key);
2143 if (!used || used->key > key)
2144 return key; /* key is available */
2145 key++;
2146 } while (key);
2147 return key;
2148}
2149
Willy Tarreau348238b2010-01-18 15:05:57 +01002150/* This function compares a sample word possibly followed by blanks to another
2151 * clean word. The compare is case-insensitive. 1 is returned if both are equal,
2152 * otherwise zero. This intends to be used when checking HTTP headers for some
2153 * values. Note that it validates a word followed only by blanks but does not
2154 * validate a word followed by blanks then other chars.
2155 */
2156int word_match(const char *sample, int slen, const char *word, int wlen)
2157{
2158 if (slen < wlen)
2159 return 0;
2160
2161 while (wlen) {
2162 char c = *sample ^ *word;
2163 if (c && c != ('A' ^ 'a'))
2164 return 0;
2165 sample++;
2166 word++;
2167 slen--;
2168 wlen--;
2169 }
2170
2171 while (slen) {
2172 if (*sample != ' ' && *sample != '\t')
2173 return 0;
2174 sample++;
2175 slen--;
2176 }
2177 return 1;
2178}
Willy Tarreau482b00d2009-10-04 22:48:42 +02002179
Willy Tarreaud54bbdc2009-09-07 11:00:31 +02002180/* Converts any text-formatted IPv4 address to a host-order IPv4 address. It
2181 * is particularly fast because it avoids expensive operations such as
2182 * multiplies, which are optimized away at the end. It requires a properly
2183 * formated address though (3 points).
2184 */
2185unsigned int inetaddr_host(const char *text)
2186{
2187 const unsigned int ascii_zero = ('0' << 24) | ('0' << 16) | ('0' << 8) | '0';
2188 register unsigned int dig100, dig10, dig1;
2189 int s;
2190 const char *p, *d;
2191
2192 dig1 = dig10 = dig100 = ascii_zero;
2193 s = 24;
2194
2195 p = text;
2196 while (1) {
2197 if (((unsigned)(*p - '0')) <= 9) {
2198 p++;
2199 continue;
2200 }
2201
2202 /* here, we have a complete byte between <text> and <p> (exclusive) */
2203 if (p == text)
2204 goto end;
2205
2206 d = p - 1;
2207 dig1 |= (unsigned int)(*d << s);
2208 if (d == text)
2209 goto end;
2210
2211 d--;
2212 dig10 |= (unsigned int)(*d << s);
2213 if (d == text)
2214 goto end;
2215
2216 d--;
2217 dig100 |= (unsigned int)(*d << s);
2218 end:
2219 if (!s || *p != '.')
2220 break;
2221
2222 s -= 8;
2223 text = ++p;
2224 }
2225
2226 dig100 -= ascii_zero;
2227 dig10 -= ascii_zero;
2228 dig1 -= ascii_zero;
2229 return ((dig100 * 10) + dig10) * 10 + dig1;
2230}
2231
2232/*
2233 * Idem except the first unparsed character has to be passed in <stop>.
2234 */
2235unsigned int inetaddr_host_lim(const char *text, const char *stop)
2236{
2237 const unsigned int ascii_zero = ('0' << 24) | ('0' << 16) | ('0' << 8) | '0';
2238 register unsigned int dig100, dig10, dig1;
2239 int s;
2240 const char *p, *d;
2241
2242 dig1 = dig10 = dig100 = ascii_zero;
2243 s = 24;
2244
2245 p = text;
2246 while (1) {
2247 if (((unsigned)(*p - '0')) <= 9 && p < stop) {
2248 p++;
2249 continue;
2250 }
2251
2252 /* here, we have a complete byte between <text> and <p> (exclusive) */
2253 if (p == text)
2254 goto end;
2255
2256 d = p - 1;
2257 dig1 |= (unsigned int)(*d << s);
2258 if (d == text)
2259 goto end;
2260
2261 d--;
2262 dig10 |= (unsigned int)(*d << s);
2263 if (d == text)
2264 goto end;
2265
2266 d--;
2267 dig100 |= (unsigned int)(*d << s);
2268 end:
2269 if (!s || p == stop || *p != '.')
2270 break;
2271
2272 s -= 8;
2273 text = ++p;
2274 }
2275
2276 dig100 -= ascii_zero;
2277 dig10 -= ascii_zero;
2278 dig1 -= ascii_zero;
2279 return ((dig100 * 10) + dig10) * 10 + dig1;
2280}
2281
2282/*
2283 * Idem except the pointer to first unparsed byte is returned into <ret> which
2284 * must not be NULL.
2285 */
Willy Tarreau74172752010-10-15 23:21:42 +02002286unsigned int inetaddr_host_lim_ret(char *text, char *stop, char **ret)
Willy Tarreaud54bbdc2009-09-07 11:00:31 +02002287{
2288 const unsigned int ascii_zero = ('0' << 24) | ('0' << 16) | ('0' << 8) | '0';
2289 register unsigned int dig100, dig10, dig1;
2290 int s;
Willy Tarreau74172752010-10-15 23:21:42 +02002291 char *p, *d;
Willy Tarreaud54bbdc2009-09-07 11:00:31 +02002292
2293 dig1 = dig10 = dig100 = ascii_zero;
2294 s = 24;
2295
2296 p = text;
2297 while (1) {
2298 if (((unsigned)(*p - '0')) <= 9 && p < stop) {
2299 p++;
2300 continue;
2301 }
2302
2303 /* here, we have a complete byte between <text> and <p> (exclusive) */
2304 if (p == text)
2305 goto end;
2306
2307 d = p - 1;
2308 dig1 |= (unsigned int)(*d << s);
2309 if (d == text)
2310 goto end;
2311
2312 d--;
2313 dig10 |= (unsigned int)(*d << s);
2314 if (d == text)
2315 goto end;
2316
2317 d--;
2318 dig100 |= (unsigned int)(*d << s);
2319 end:
2320 if (!s || p == stop || *p != '.')
2321 break;
2322
2323 s -= 8;
2324 text = ++p;
2325 }
2326
2327 *ret = p;
2328 dig100 -= ascii_zero;
2329 dig10 -= ascii_zero;
2330 dig1 -= ascii_zero;
2331 return ((dig100 * 10) + dig10) * 10 + dig1;
2332}
2333
Willy Tarreauf0b38bf2010-06-06 13:22:23 +02002334/* Convert a fixed-length string to an IP address. Returns 0 in case of error,
2335 * or the number of chars read in case of success. Maybe this could be replaced
2336 * by one of the functions above. Also, apparently this function does not support
2337 * hosts above 255 and requires exactly 4 octets.
Willy Tarreau075415a2013-12-12 11:29:39 +01002338 * The destination is only modified on success.
Willy Tarreauf0b38bf2010-06-06 13:22:23 +02002339 */
2340int buf2ip(const char *buf, size_t len, struct in_addr *dst)
2341{
2342 const char *addr;
2343 int saw_digit, octets, ch;
2344 u_char tmp[4], *tp;
2345 const char *cp = buf;
2346
2347 saw_digit = 0;
2348 octets = 0;
2349 *(tp = tmp) = 0;
2350
2351 for (addr = buf; addr - buf < len; addr++) {
2352 unsigned char digit = (ch = *addr) - '0';
2353
2354 if (digit > 9 && ch != '.')
2355 break;
2356
2357 if (digit <= 9) {
2358 u_int new = *tp * 10 + digit;
2359
2360 if (new > 255)
2361 return 0;
2362
2363 *tp = new;
2364
2365 if (!saw_digit) {
2366 if (++octets > 4)
2367 return 0;
2368 saw_digit = 1;
2369 }
2370 } else if (ch == '.' && saw_digit) {
2371 if (octets == 4)
2372 return 0;
2373
2374 *++tp = 0;
2375 saw_digit = 0;
2376 } else
2377 return 0;
2378 }
2379
2380 if (octets < 4)
2381 return 0;
2382
2383 memcpy(&dst->s_addr, tmp, 4);
2384 return addr - cp;
2385}
2386
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002387/* This function converts the string in <buf> of the len <len> to
2388 * struct in6_addr <dst> which must be allocated by the caller.
2389 * This function returns 1 in success case, otherwise zero.
Willy Tarreau075415a2013-12-12 11:29:39 +01002390 * The destination is only modified on success.
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002391 */
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002392int buf2ip6(const char *buf, size_t len, struct in6_addr *dst)
2393{
Thierry FOURNIERcd659912013-12-11 12:33:54 +01002394 char null_term_ip6[INET6_ADDRSTRLEN + 1];
Willy Tarreau075415a2013-12-12 11:29:39 +01002395 struct in6_addr out;
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002396
Thierry FOURNIERcd659912013-12-11 12:33:54 +01002397 if (len > INET6_ADDRSTRLEN)
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002398 return 0;
2399
2400 memcpy(null_term_ip6, buf, len);
2401 null_term_ip6[len] = '\0';
2402
Willy Tarreau075415a2013-12-12 11:29:39 +01002403 if (!inet_pton(AF_INET6, null_term_ip6, &out))
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002404 return 0;
2405
Willy Tarreau075415a2013-12-12 11:29:39 +01002406 *dst = out;
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002407 return 1;
2408}
2409
Willy Tarreauacf95772010-06-14 19:09:21 +02002410/* To be used to quote config arg positions. Returns the short string at <ptr>
2411 * surrounded by simple quotes if <ptr> is valid and non-empty, or "end of line"
2412 * if ptr is NULL or empty. The string is locally allocated.
2413 */
2414const char *quote_arg(const char *ptr)
2415{
2416 static char val[32];
2417 int i;
2418
2419 if (!ptr || !*ptr)
2420 return "end of line";
2421 val[0] = '\'';
Willy Tarreaude2dd6b2013-01-24 02:14:42 +01002422 for (i = 1; i < sizeof(val) - 2 && *ptr; i++)
Willy Tarreauacf95772010-06-14 19:09:21 +02002423 val[i] = *ptr++;
2424 val[i++] = '\'';
2425 val[i] = '\0';
2426 return val;
2427}
2428
Willy Tarreau5b180202010-07-18 10:40:48 +02002429/* returns an operator among STD_OP_* for string <str> or < 0 if unknown */
2430int get_std_op(const char *str)
2431{
2432 int ret = -1;
2433
2434 if (*str == 'e' && str[1] == 'q')
2435 ret = STD_OP_EQ;
2436 else if (*str == 'n' && str[1] == 'e')
2437 ret = STD_OP_NE;
2438 else if (*str == 'l') {
2439 if (str[1] == 'e') ret = STD_OP_LE;
2440 else if (str[1] == 't') ret = STD_OP_LT;
2441 }
2442 else if (*str == 'g') {
2443 if (str[1] == 'e') ret = STD_OP_GE;
2444 else if (str[1] == 't') ret = STD_OP_GT;
2445 }
2446
2447 if (ret == -1 || str[2] != '\0')
2448 return -1;
2449 return ret;
2450}
2451
Willy Tarreau4c14eaa2010-11-24 14:01:45 +01002452/* hash a 32-bit integer to another 32-bit integer */
2453unsigned int full_hash(unsigned int a)
2454{
2455 return __full_hash(a);
2456}
2457
David du Colombier4f92d322011-03-24 11:09:31 +01002458/* Return non-zero if IPv4 address is part of the network,
Willy Tarreaueec1d382016-07-13 11:59:39 +02002459 * otherwise zero. Note that <addr> may not necessarily be aligned
2460 * while the two other ones must.
David du Colombier4f92d322011-03-24 11:09:31 +01002461 */
Willy Tarreaueec1d382016-07-13 11:59:39 +02002462int in_net_ipv4(const void *addr, const struct in_addr *mask, const struct in_addr *net)
David du Colombier4f92d322011-03-24 11:09:31 +01002463{
Willy Tarreaueec1d382016-07-13 11:59:39 +02002464 struct in_addr addr_copy;
2465
2466 memcpy(&addr_copy, addr, sizeof(addr_copy));
2467 return((addr_copy.s_addr & mask->s_addr) == (net->s_addr & mask->s_addr));
David du Colombier4f92d322011-03-24 11:09:31 +01002468}
2469
2470/* Return non-zero if IPv6 address is part of the network,
Willy Tarreaueec1d382016-07-13 11:59:39 +02002471 * otherwise zero. Note that <addr> may not necessarily be aligned
2472 * while the two other ones must.
David du Colombier4f92d322011-03-24 11:09:31 +01002473 */
Willy Tarreaueec1d382016-07-13 11:59:39 +02002474int in_net_ipv6(const void *addr, const struct in6_addr *mask, const struct in6_addr *net)
David du Colombier4f92d322011-03-24 11:09:31 +01002475{
2476 int i;
Willy Tarreaueec1d382016-07-13 11:59:39 +02002477 struct in6_addr addr_copy;
David du Colombier4f92d322011-03-24 11:09:31 +01002478
Willy Tarreaueec1d382016-07-13 11:59:39 +02002479 memcpy(&addr_copy, addr, sizeof(addr_copy));
David du Colombier4f92d322011-03-24 11:09:31 +01002480 for (i = 0; i < sizeof(struct in6_addr) / sizeof(int); i++)
Willy Tarreaueec1d382016-07-13 11:59:39 +02002481 if (((((int *)&addr_copy)[i] & ((int *)mask)[i])) !=
David du Colombier4f92d322011-03-24 11:09:31 +01002482 (((int *)net)[i] & ((int *)mask)[i]))
2483 return 0;
2484 return 1;
2485}
2486
2487/* RFC 4291 prefix */
2488const char rfc4291_pfx[] = { 0x00, 0x00, 0x00, 0x00,
2489 0x00, 0x00, 0x00, 0x00,
2490 0x00, 0x00, 0xFF, 0xFF };
2491
Thierry FOURNIER4a04dc32013-11-28 16:33:15 +01002492/* Map IPv4 adress on IPv6 address, as specified in RFC 3513.
2493 * Input and output may overlap.
2494 */
David du Colombier4f92d322011-03-24 11:09:31 +01002495void v4tov6(struct in6_addr *sin6_addr, struct in_addr *sin_addr)
2496{
Thierry FOURNIER4a04dc32013-11-28 16:33:15 +01002497 struct in_addr tmp_addr;
2498
2499 tmp_addr.s_addr = sin_addr->s_addr;
David du Colombier4f92d322011-03-24 11:09:31 +01002500 memcpy(sin6_addr->s6_addr, rfc4291_pfx, sizeof(rfc4291_pfx));
Thierry FOURNIER4a04dc32013-11-28 16:33:15 +01002501 memcpy(sin6_addr->s6_addr+12, &tmp_addr.s_addr, 4);
David du Colombier4f92d322011-03-24 11:09:31 +01002502}
2503
2504/* Map IPv6 adress on IPv4 address, as specified in RFC 3513.
2505 * Return true if conversion is possible and false otherwise.
2506 */
2507int v6tov4(struct in_addr *sin_addr, struct in6_addr *sin6_addr)
2508{
2509 if (memcmp(sin6_addr->s6_addr, rfc4291_pfx, sizeof(rfc4291_pfx)) == 0) {
2510 memcpy(&(sin_addr->s_addr), &(sin6_addr->s6_addr[12]),
2511 sizeof(struct in_addr));
2512 return 1;
2513 }
2514
2515 return 0;
2516}
2517
William Lallemand421f5b52012-02-06 18:15:57 +01002518char *human_time(int t, short hz_div) {
2519 static char rv[sizeof("24855d23h")+1]; // longest of "23h59m" and "59m59s"
2520 char *p = rv;
Willy Tarreau761b3d52014-04-14 14:53:06 +02002521 char *end = rv + sizeof(rv);
William Lallemand421f5b52012-02-06 18:15:57 +01002522 int cnt=2; // print two numbers
2523
2524 if (unlikely(t < 0 || hz_div <= 0)) {
Willy Tarreau761b3d52014-04-14 14:53:06 +02002525 snprintf(p, end - p, "?");
William Lallemand421f5b52012-02-06 18:15:57 +01002526 return rv;
2527 }
2528
2529 if (unlikely(hz_div > 1))
2530 t /= hz_div;
2531
2532 if (t >= DAY) {
Willy Tarreau761b3d52014-04-14 14:53:06 +02002533 p += snprintf(p, end - p, "%dd", t / DAY);
William Lallemand421f5b52012-02-06 18:15:57 +01002534 cnt--;
2535 }
2536
2537 if (cnt && t % DAY / HOUR) {
Willy Tarreau761b3d52014-04-14 14:53:06 +02002538 p += snprintf(p, end - p, "%dh", t % DAY / HOUR);
William Lallemand421f5b52012-02-06 18:15:57 +01002539 cnt--;
2540 }
2541
2542 if (cnt && t % HOUR / MINUTE) {
Willy Tarreau761b3d52014-04-14 14:53:06 +02002543 p += snprintf(p, end - p, "%dm", t % HOUR / MINUTE);
William Lallemand421f5b52012-02-06 18:15:57 +01002544 cnt--;
2545 }
2546
2547 if ((cnt && t % MINUTE) || !t) // also display '0s'
Willy Tarreau761b3d52014-04-14 14:53:06 +02002548 p += snprintf(p, end - p, "%ds", t % MINUTE / SEC);
William Lallemand421f5b52012-02-06 18:15:57 +01002549
2550 return rv;
2551}
2552
2553const char *monthname[12] = {
2554 "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2555 "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2556};
2557
2558/* date2str_log: write a date in the format :
2559 * sprintf(str, "%02d/%s/%04d:%02d:%02d:%02d.%03d",
2560 * tm.tm_mday, monthname[tm.tm_mon], tm.tm_year+1900,
2561 * tm.tm_hour, tm.tm_min, tm.tm_sec, (int)date.tv_usec/1000);
2562 *
2563 * without using sprintf. return a pointer to the last char written (\0) or
2564 * NULL if there isn't enough space.
2565 */
2566char *date2str_log(char *dst, struct tm *tm, struct timeval *date, size_t size)
2567{
2568
2569 if (size < 25) /* the size is fixed: 24 chars + \0 */
2570 return NULL;
2571
2572 dst = utoa_pad((unsigned int)tm->tm_mday, dst, 3); // day
2573 *dst++ = '/';
2574 memcpy(dst, monthname[tm->tm_mon], 3); // month
2575 dst += 3;
2576 *dst++ = '/';
2577 dst = utoa_pad((unsigned int)tm->tm_year+1900, dst, 5); // year
2578 *dst++ = ':';
2579 dst = utoa_pad((unsigned int)tm->tm_hour, dst, 3); // hour
2580 *dst++ = ':';
2581 dst = utoa_pad((unsigned int)tm->tm_min, dst, 3); // minutes
2582 *dst++ = ':';
2583 dst = utoa_pad((unsigned int)tm->tm_sec, dst, 3); // secondes
2584 *dst++ = '.';
2585 utoa_pad((unsigned int)(date->tv_usec/1000), dst, 4); // millisecondes
2586 dst += 3; // only the 3 first digits
2587 *dst = '\0';
2588
2589 return dst;
2590}
2591
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02002592/* Base year used to compute leap years */
2593#define TM_YEAR_BASE 1900
2594
2595/* Return the difference in seconds between two times (leap seconds are ignored).
2596 * Retrieved from glibc 2.18 source code.
2597 */
2598static int my_tm_diff(const struct tm *a, const struct tm *b)
2599{
2600 /* Compute intervening leap days correctly even if year is negative.
2601 * Take care to avoid int overflow in leap day calculations,
2602 * but it's OK to assume that A and B are close to each other.
2603 */
2604 int a4 = (a->tm_year >> 2) + (TM_YEAR_BASE >> 2) - ! (a->tm_year & 3);
2605 int b4 = (b->tm_year >> 2) + (TM_YEAR_BASE >> 2) - ! (b->tm_year & 3);
2606 int a100 = a4 / 25 - (a4 % 25 < 0);
2607 int b100 = b4 / 25 - (b4 % 25 < 0);
2608 int a400 = a100 >> 2;
2609 int b400 = b100 >> 2;
2610 int intervening_leap_days = (a4 - b4) - (a100 - b100) + (a400 - b400);
2611 int years = a->tm_year - b->tm_year;
2612 int days = (365 * years + intervening_leap_days
2613 + (a->tm_yday - b->tm_yday));
2614 return (60 * (60 * (24 * days + (a->tm_hour - b->tm_hour))
2615 + (a->tm_min - b->tm_min))
2616 + (a->tm_sec - b->tm_sec));
2617}
2618
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02002619/* Return the GMT offset for a specific local time.
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02002620 * Both t and tm must represent the same time.
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02002621 * The string returned has the same format as returned by strftime(... "%z", tm).
2622 * Offsets are kept in an internal cache for better performances.
2623 */
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02002624const char *get_gmt_offset(time_t t, struct tm *tm)
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02002625{
2626 /* Cache offsets from GMT (depending on whether DST is active or not) */
2627 static char gmt_offsets[2][5+1] = { "", "" };
2628
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02002629 char *gmt_offset;
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02002630 struct tm tm_gmt;
2631 int diff;
2632 int isdst = tm->tm_isdst;
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02002633
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02002634 /* Pretend DST not active if its status is unknown */
2635 if (isdst < 0)
2636 isdst = 0;
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02002637
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02002638 /* Fetch the offset and initialize it if needed */
2639 gmt_offset = gmt_offsets[isdst & 0x01];
2640 if (unlikely(!*gmt_offset)) {
2641 get_gmtime(t, &tm_gmt);
2642 diff = my_tm_diff(tm, &tm_gmt);
2643 if (diff < 0) {
2644 diff = -diff;
2645 *gmt_offset = '-';
2646 } else {
2647 *gmt_offset = '+';
2648 }
2649 diff /= 60; /* Convert to minutes */
2650 snprintf(gmt_offset+1, 4+1, "%02d%02d", diff/60, diff%60);
2651 }
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02002652
2653 return gmt_offset;
2654}
2655
William Lallemand421f5b52012-02-06 18:15:57 +01002656/* gmt2str_log: write a date in the format :
2657 * "%02d/%s/%04d:%02d:%02d:%02d +0000" without using snprintf
2658 * return a pointer to the last char written (\0) or
2659 * NULL if there isn't enough space.
2660 */
2661char *gmt2str_log(char *dst, struct tm *tm, size_t size)
2662{
Yuxans Yao4e25b012012-10-19 10:36:09 +08002663 if (size < 27) /* the size is fixed: 26 chars + \0 */
William Lallemand421f5b52012-02-06 18:15:57 +01002664 return NULL;
2665
2666 dst = utoa_pad((unsigned int)tm->tm_mday, dst, 3); // day
2667 *dst++ = '/';
2668 memcpy(dst, monthname[tm->tm_mon], 3); // month
2669 dst += 3;
2670 *dst++ = '/';
2671 dst = utoa_pad((unsigned int)tm->tm_year+1900, dst, 5); // year
2672 *dst++ = ':';
2673 dst = utoa_pad((unsigned int)tm->tm_hour, dst, 3); // hour
2674 *dst++ = ':';
2675 dst = utoa_pad((unsigned int)tm->tm_min, dst, 3); // minutes
2676 *dst++ = ':';
2677 dst = utoa_pad((unsigned int)tm->tm_sec, dst, 3); // secondes
2678 *dst++ = ' ';
2679 *dst++ = '+';
2680 *dst++ = '0';
2681 *dst++ = '0';
2682 *dst++ = '0';
2683 *dst++ = '0';
2684 *dst = '\0';
2685
2686 return dst;
2687}
2688
Yuxans Yao4e25b012012-10-19 10:36:09 +08002689/* localdate2str_log: write a date in the format :
2690 * "%02d/%s/%04d:%02d:%02d:%02d +0000(local timezone)" without using snprintf
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02002691 * Both t and tm must represent the same time.
2692 * return a pointer to the last char written (\0) or
2693 * NULL if there isn't enough space.
Yuxans Yao4e25b012012-10-19 10:36:09 +08002694 */
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02002695char *localdate2str_log(char *dst, time_t t, struct tm *tm, size_t size)
Yuxans Yao4e25b012012-10-19 10:36:09 +08002696{
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02002697 const char *gmt_offset;
Yuxans Yao4e25b012012-10-19 10:36:09 +08002698 if (size < 27) /* the size is fixed: 26 chars + \0 */
2699 return NULL;
2700
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02002701 gmt_offset = get_gmt_offset(t, tm);
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02002702
Yuxans Yao4e25b012012-10-19 10:36:09 +08002703 dst = utoa_pad((unsigned int)tm->tm_mday, dst, 3); // day
2704 *dst++ = '/';
2705 memcpy(dst, monthname[tm->tm_mon], 3); // month
2706 dst += 3;
2707 *dst++ = '/';
2708 dst = utoa_pad((unsigned int)tm->tm_year+1900, dst, 5); // year
2709 *dst++ = ':';
2710 dst = utoa_pad((unsigned int)tm->tm_hour, dst, 3); // hour
2711 *dst++ = ':';
2712 dst = utoa_pad((unsigned int)tm->tm_min, dst, 3); // minutes
2713 *dst++ = ':';
2714 dst = utoa_pad((unsigned int)tm->tm_sec, dst, 3); // secondes
2715 *dst++ = ' ';
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02002716 memcpy(dst, gmt_offset, 5); // Offset from local time to GMT
Yuxans Yao4e25b012012-10-19 10:36:09 +08002717 dst += 5;
2718 *dst = '\0';
2719
2720 return dst;
2721}
2722
Thierry Fournier93127942016-01-20 18:49:45 +01002723/* This function check a char. It returns true and updates
2724 * <date> and <len> pointer to the new position if the
2725 * character is found.
2726 */
2727static inline int parse_expect_char(const char **date, int *len, char c)
2728{
2729 if (*len < 1 || **date != c)
2730 return 0;
2731 (*len)--;
2732 (*date)++;
2733 return 1;
2734}
2735
2736/* This function expects a string <str> of len <l>. It return true and updates.
2737 * <date> and <len> if the string matches, otherwise, it returns false.
2738 */
2739static inline int parse_strcmp(const char **date, int *len, char *str, int l)
2740{
2741 if (*len < l || strncmp(*date, str, l) != 0)
2742 return 0;
2743 (*len) -= l;
2744 (*date) += l;
2745 return 1;
2746}
2747
2748/* This macro converts 3 chars name in integer. */
2749#define STR2I3(__a, __b, __c) ((__a) * 65536 + (__b) * 256 + (__c))
2750
2751/* day-name = %x4D.6F.6E ; "Mon", case-sensitive
2752 * / %x54.75.65 ; "Tue", case-sensitive
2753 * / %x57.65.64 ; "Wed", case-sensitive
2754 * / %x54.68.75 ; "Thu", case-sensitive
2755 * / %x46.72.69 ; "Fri", case-sensitive
2756 * / %x53.61.74 ; "Sat", case-sensitive
2757 * / %x53.75.6E ; "Sun", case-sensitive
2758 *
2759 * This array must be alphabetically sorted
2760 */
2761static inline int parse_http_dayname(const char **date, int *len, struct tm *tm)
2762{
2763 if (*len < 3)
2764 return 0;
2765 switch (STR2I3((*date)[0], (*date)[1], (*date)[2])) {
2766 case STR2I3('M','o','n'): tm->tm_wday = 1; break;
2767 case STR2I3('T','u','e'): tm->tm_wday = 2; break;
2768 case STR2I3('W','e','d'): tm->tm_wday = 3; break;
2769 case STR2I3('T','h','u'): tm->tm_wday = 4; break;
2770 case STR2I3('F','r','i'): tm->tm_wday = 5; break;
2771 case STR2I3('S','a','t'): tm->tm_wday = 6; break;
2772 case STR2I3('S','u','n'): tm->tm_wday = 7; break;
2773 default: return 0;
2774 }
2775 *len -= 3;
2776 *date += 3;
2777 return 1;
2778}
2779
2780/* month = %x4A.61.6E ; "Jan", case-sensitive
2781 * / %x46.65.62 ; "Feb", case-sensitive
2782 * / %x4D.61.72 ; "Mar", case-sensitive
2783 * / %x41.70.72 ; "Apr", case-sensitive
2784 * / %x4D.61.79 ; "May", case-sensitive
2785 * / %x4A.75.6E ; "Jun", case-sensitive
2786 * / %x4A.75.6C ; "Jul", case-sensitive
2787 * / %x41.75.67 ; "Aug", case-sensitive
2788 * / %x53.65.70 ; "Sep", case-sensitive
2789 * / %x4F.63.74 ; "Oct", case-sensitive
2790 * / %x4E.6F.76 ; "Nov", case-sensitive
2791 * / %x44.65.63 ; "Dec", case-sensitive
2792 *
2793 * This array must be alphabetically sorted
2794 */
2795static inline int parse_http_monthname(const char **date, int *len, struct tm *tm)
2796{
2797 if (*len < 3)
2798 return 0;
2799 switch (STR2I3((*date)[0], (*date)[1], (*date)[2])) {
2800 case STR2I3('J','a','n'): tm->tm_mon = 0; break;
2801 case STR2I3('F','e','b'): tm->tm_mon = 1; break;
2802 case STR2I3('M','a','r'): tm->tm_mon = 2; break;
2803 case STR2I3('A','p','r'): tm->tm_mon = 3; break;
2804 case STR2I3('M','a','y'): tm->tm_mon = 4; break;
2805 case STR2I3('J','u','n'): tm->tm_mon = 5; break;
2806 case STR2I3('J','u','l'): tm->tm_mon = 6; break;
2807 case STR2I3('A','u','g'): tm->tm_mon = 7; break;
2808 case STR2I3('S','e','p'): tm->tm_mon = 8; break;
2809 case STR2I3('O','c','t'): tm->tm_mon = 9; break;
2810 case STR2I3('N','o','v'): tm->tm_mon = 10; break;
2811 case STR2I3('D','e','c'): tm->tm_mon = 11; break;
2812 default: return 0;
2813 }
2814 *len -= 3;
2815 *date += 3;
2816 return 1;
2817}
2818
2819/* day-name-l = %x4D.6F.6E.64.61.79 ; "Monday", case-sensitive
2820 * / %x54.75.65.73.64.61.79 ; "Tuesday", case-sensitive
2821 * / %x57.65.64.6E.65.73.64.61.79 ; "Wednesday", case-sensitive
2822 * / %x54.68.75.72.73.64.61.79 ; "Thursday", case-sensitive
2823 * / %x46.72.69.64.61.79 ; "Friday", case-sensitive
2824 * / %x53.61.74.75.72.64.61.79 ; "Saturday", case-sensitive
2825 * / %x53.75.6E.64.61.79 ; "Sunday", case-sensitive
2826 *
2827 * This array must be alphabetically sorted
2828 */
2829static inline int parse_http_ldayname(const char **date, int *len, struct tm *tm)
2830{
2831 if (*len < 6) /* Minimum length. */
2832 return 0;
2833 switch (STR2I3((*date)[0], (*date)[1], (*date)[2])) {
2834 case STR2I3('M','o','n'):
2835 RET0_UNLESS(parse_strcmp(date, len, "Monday", 6));
2836 tm->tm_wday = 1;
2837 return 1;
2838 case STR2I3('T','u','e'):
2839 RET0_UNLESS(parse_strcmp(date, len, "Tuesday", 7));
2840 tm->tm_wday = 2;
2841 return 1;
2842 case STR2I3('W','e','d'):
2843 RET0_UNLESS(parse_strcmp(date, len, "Wednesday", 9));
2844 tm->tm_wday = 3;
2845 return 1;
2846 case STR2I3('T','h','u'):
2847 RET0_UNLESS(parse_strcmp(date, len, "Thursday", 8));
2848 tm->tm_wday = 4;
2849 return 1;
2850 case STR2I3('F','r','i'):
2851 RET0_UNLESS(parse_strcmp(date, len, "Friday", 6));
2852 tm->tm_wday = 5;
2853 return 1;
2854 case STR2I3('S','a','t'):
2855 RET0_UNLESS(parse_strcmp(date, len, "Saturday", 8));
2856 tm->tm_wday = 6;
2857 return 1;
2858 case STR2I3('S','u','n'):
2859 RET0_UNLESS(parse_strcmp(date, len, "Sunday", 6));
2860 tm->tm_wday = 7;
2861 return 1;
2862 }
2863 return 0;
2864}
2865
2866/* This function parses exactly 1 digit and returns the numeric value in "digit". */
2867static inline int parse_digit(const char **date, int *len, int *digit)
2868{
2869 if (*len < 1 || **date < '0' || **date > '9')
2870 return 0;
2871 *digit = (**date - '0');
2872 (*date)++;
2873 (*len)--;
2874 return 1;
2875}
2876
2877/* This function parses exactly 2 digits and returns the numeric value in "digit". */
2878static inline int parse_2digit(const char **date, int *len, int *digit)
2879{
2880 int value;
2881
2882 RET0_UNLESS(parse_digit(date, len, &value));
2883 (*digit) = value * 10;
2884 RET0_UNLESS(parse_digit(date, len, &value));
2885 (*digit) += value;
2886
2887 return 1;
2888}
2889
2890/* This function parses exactly 4 digits and returns the numeric value in "digit". */
2891static inline int parse_4digit(const char **date, int *len, int *digit)
2892{
2893 int value;
2894
2895 RET0_UNLESS(parse_digit(date, len, &value));
2896 (*digit) = value * 1000;
2897
2898 RET0_UNLESS(parse_digit(date, len, &value));
2899 (*digit) += value * 100;
2900
2901 RET0_UNLESS(parse_digit(date, len, &value));
2902 (*digit) += value * 10;
2903
2904 RET0_UNLESS(parse_digit(date, len, &value));
2905 (*digit) += value;
2906
2907 return 1;
2908}
2909
2910/* time-of-day = hour ":" minute ":" second
2911 * ; 00:00:00 - 23:59:60 (leap second)
2912 *
2913 * hour = 2DIGIT
2914 * minute = 2DIGIT
2915 * second = 2DIGIT
2916 */
2917static inline int parse_http_time(const char **date, int *len, struct tm *tm)
2918{
2919 RET0_UNLESS(parse_2digit(date, len, &tm->tm_hour)); /* hour 2DIGIT */
2920 RET0_UNLESS(parse_expect_char(date, len, ':')); /* expect ":" */
2921 RET0_UNLESS(parse_2digit(date, len, &tm->tm_min)); /* min 2DIGIT */
2922 RET0_UNLESS(parse_expect_char(date, len, ':')); /* expect ":" */
2923 RET0_UNLESS(parse_2digit(date, len, &tm->tm_sec)); /* sec 2DIGIT */
2924 return 1;
2925}
2926
2927/* From RFC7231
2928 * https://tools.ietf.org/html/rfc7231#section-7.1.1.1
2929 *
2930 * IMF-fixdate = day-name "," SP date1 SP time-of-day SP GMT
2931 * ; fixed length/zone/capitalization subset of the format
2932 * ; see Section 3.3 of [RFC5322]
2933 *
2934 *
2935 * date1 = day SP month SP year
2936 * ; e.g., 02 Jun 1982
2937 *
2938 * day = 2DIGIT
2939 * year = 4DIGIT
2940 *
2941 * GMT = %x47.4D.54 ; "GMT", case-sensitive
2942 *
2943 * time-of-day = hour ":" minute ":" second
2944 * ; 00:00:00 - 23:59:60 (leap second)
2945 *
2946 * hour = 2DIGIT
2947 * minute = 2DIGIT
2948 * second = 2DIGIT
2949 *
2950 * DIGIT = decimal 0-9
2951 */
2952int parse_imf_date(const char *date, int len, struct tm *tm)
2953{
2954 RET0_UNLESS(parse_http_dayname(&date, &len, tm)); /* day-name */
2955 RET0_UNLESS(parse_expect_char(&date, &len, ',')); /* expect "," */
2956 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
2957 RET0_UNLESS(parse_2digit(&date, &len, &tm->tm_mday)); /* day 2DIGIT */
2958 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
2959 RET0_UNLESS(parse_http_monthname(&date, &len, tm)); /* Month */
2960 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
2961 RET0_UNLESS(parse_4digit(&date, &len, &tm->tm_year)); /* year = 4DIGIT */
2962 tm->tm_year -= 1900;
2963 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
2964 RET0_UNLESS(parse_http_time(&date, &len, tm)); /* Parse time. */
2965 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
2966 RET0_UNLESS(parse_strcmp(&date, &len, "GMT", 3)); /* GMT = %x47.4D.54 ; "GMT", case-sensitive */
2967 tm->tm_isdst = -1;
2968 tm->tm_gmtoff = 0;
2969 return 1;
2970}
2971
2972/* From RFC7231
2973 * https://tools.ietf.org/html/rfc7231#section-7.1.1.1
2974 *
2975 * rfc850-date = day-name-l "," SP date2 SP time-of-day SP GMT
2976 * date2 = day "-" month "-" 2DIGIT
2977 * ; e.g., 02-Jun-82
2978 *
2979 * day = 2DIGIT
2980 */
2981int parse_rfc850_date(const char *date, int len, struct tm *tm)
2982{
2983 int year;
2984
2985 RET0_UNLESS(parse_http_ldayname(&date, &len, tm)); /* Read the day name */
2986 RET0_UNLESS(parse_expect_char(&date, &len, ',')); /* expect "," */
2987 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
2988 RET0_UNLESS(parse_2digit(&date, &len, &tm->tm_mday)); /* day 2DIGIT */
2989 RET0_UNLESS(parse_expect_char(&date, &len, '-')); /* expect "-" */
2990 RET0_UNLESS(parse_http_monthname(&date, &len, tm)); /* Month */
2991 RET0_UNLESS(parse_expect_char(&date, &len, '-')); /* expect "-" */
2992
2993 /* year = 2DIGIT
2994 *
2995 * Recipients of a timestamp value in rfc850-(*date) format, which uses a
2996 * two-digit year, MUST interpret a timestamp that appears to be more
2997 * than 50 years in the future as representing the most recent year in
2998 * the past that had the same last two digits.
2999 */
3000 RET0_UNLESS(parse_2digit(&date, &len, &tm->tm_year));
3001
3002 /* expect SP */
3003 if (!parse_expect_char(&date, &len, ' ')) {
3004 /* Maybe we have the date with 4 digits. */
3005 RET0_UNLESS(parse_2digit(&date, &len, &year));
3006 tm->tm_year = (tm->tm_year * 100 + year) - 1900;
3007 /* expect SP */
3008 RET0_UNLESS(parse_expect_char(&date, &len, ' '));
3009 } else {
3010 /* I fix 60 as pivot: >60: +1900, <60: +2000. Note that the
3011 * tm_year is the number of year since 1900, so for +1900, we
3012 * do nothing, and for +2000, we add 100.
3013 */
3014 if (tm->tm_year <= 60)
3015 tm->tm_year += 100;
3016 }
3017
3018 RET0_UNLESS(parse_http_time(&date, &len, tm)); /* Parse time. */
3019 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3020 RET0_UNLESS(parse_strcmp(&date, &len, "GMT", 3)); /* GMT = %x47.4D.54 ; "GMT", case-sensitive */
3021 tm->tm_isdst = -1;
3022 tm->tm_gmtoff = 0;
3023
3024 return 1;
3025}
3026
3027/* From RFC7231
3028 * https://tools.ietf.org/html/rfc7231#section-7.1.1.1
3029 *
3030 * asctime-date = day-name SP date3 SP time-of-day SP year
3031 * date3 = month SP ( 2DIGIT / ( SP 1DIGIT ))
3032 * ; e.g., Jun 2
3033 *
3034 * HTTP-date is case sensitive. A sender MUST NOT generate additional
3035 * whitespace in an HTTP-date beyond that specifically included as SP in
3036 * the grammar.
3037 */
3038int parse_asctime_date(const char *date, int len, struct tm *tm)
3039{
3040 RET0_UNLESS(parse_http_dayname(&date, &len, tm)); /* day-name */
3041 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3042 RET0_UNLESS(parse_http_monthname(&date, &len, tm)); /* expect month */
3043 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3044
3045 /* expect SP and 1DIGIT or 2DIGIT */
3046 if (parse_expect_char(&date, &len, ' '))
3047 RET0_UNLESS(parse_digit(&date, &len, &tm->tm_mday));
3048 else
3049 RET0_UNLESS(parse_2digit(&date, &len, &tm->tm_mday));
3050
3051 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3052 RET0_UNLESS(parse_http_time(&date, &len, tm)); /* Parse time. */
3053 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3054 RET0_UNLESS(parse_4digit(&date, &len, &tm->tm_year)); /* year = 4DIGIT */
3055 tm->tm_year -= 1900;
3056 tm->tm_isdst = -1;
3057 tm->tm_gmtoff = 0;
3058 return 1;
3059}
3060
3061/* From RFC7231
3062 * https://tools.ietf.org/html/rfc7231#section-7.1.1.1
3063 *
3064 * HTTP-date = IMF-fixdate / obs-date
3065 * obs-date = rfc850-date / asctime-date
3066 *
3067 * parses an HTTP date in the RFC format and is accepted
3068 * alternatives. <date> is the strinf containing the date,
3069 * len is the len of the string. <tm> is filled with the
3070 * parsed time. We must considers this time as GMT.
3071 */
3072int parse_http_date(const char *date, int len, struct tm *tm)
3073{
3074 if (parse_imf_date(date, len, tm))
3075 return 1;
3076
3077 if (parse_rfc850_date(date, len, tm))
3078 return 1;
3079
3080 if (parse_asctime_date(date, len, tm))
3081 return 1;
3082
3083 return 0;
3084}
3085
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003086/* Dynamically allocates a string of the proper length to hold the formatted
3087 * output. NULL is returned on error. The caller is responsible for freeing the
3088 * memory area using free(). The resulting string is returned in <out> if the
3089 * pointer is not NULL. A previous version of <out> might be used to build the
3090 * new string, and it will be freed before returning if it is not NULL, which
3091 * makes it possible to build complex strings from iterative calls without
3092 * having to care about freeing intermediate values, as in the example below :
3093 *
3094 * memprintf(&err, "invalid argument: '%s'", arg);
3095 * ...
3096 * memprintf(&err, "parser said : <%s>\n", *err);
3097 * ...
3098 * free(*err);
3099 *
3100 * This means that <err> must be initialized to NULL before first invocation.
3101 * The return value also holds the allocated string, which eases error checking
3102 * and immediate consumption. If the output pointer is not used, NULL must be
Willy Tarreaueb6cead2012-09-20 19:43:14 +02003103 * passed instead and it will be ignored. The returned message will then also
3104 * be NULL so that the caller does not have to bother with freeing anything.
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003105 *
3106 * It is also convenient to use it without any free except the last one :
3107 * err = NULL;
3108 * if (!fct1(err)) report(*err);
3109 * if (!fct2(err)) report(*err);
3110 * if (!fct3(err)) report(*err);
3111 * free(*err);
3112 */
3113char *memprintf(char **out, const char *format, ...)
3114{
3115 va_list args;
3116 char *ret = NULL;
3117 int allocated = 0;
3118 int needed = 0;
3119
Willy Tarreaueb6cead2012-09-20 19:43:14 +02003120 if (!out)
3121 return NULL;
3122
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003123 do {
3124 /* vsnprintf() will return the required length even when the
3125 * target buffer is NULL. We do this in a loop just in case
3126 * intermediate evaluations get wrong.
3127 */
3128 va_start(args, format);
Willy Tarreau1b2fed62013-04-01 22:48:54 +02003129 needed = vsnprintf(ret, allocated, format, args);
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003130 va_end(args);
3131
Willy Tarreau1b2fed62013-04-01 22:48:54 +02003132 if (needed < allocated) {
3133 /* Note: on Solaris 8, the first iteration always
3134 * returns -1 if allocated is zero, so we force a
3135 * retry.
3136 */
3137 if (!allocated)
3138 needed = 0;
3139 else
3140 break;
3141 }
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003142
Willy Tarreau1b2fed62013-04-01 22:48:54 +02003143 allocated = needed + 1;
Hubert Verstraete831962e2016-06-28 22:44:26 +02003144 ret = my_realloc2(ret, allocated);
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003145 } while (ret);
3146
3147 if (needed < 0) {
3148 /* an error was encountered */
3149 free(ret);
3150 ret = NULL;
3151 }
3152
3153 if (out) {
3154 free(*out);
3155 *out = ret;
3156 }
3157
3158 return ret;
3159}
William Lallemand421f5b52012-02-06 18:15:57 +01003160
Willy Tarreau21c705b2012-09-14 11:40:36 +02003161/* Used to add <level> spaces before each line of <out>, unless there is only one line.
3162 * The input argument is automatically freed and reassigned. The result will have to be
Willy Tarreau70eec382012-10-10 08:56:47 +02003163 * freed by the caller. It also supports being passed a NULL which results in the same
3164 * output.
Willy Tarreau21c705b2012-09-14 11:40:36 +02003165 * Example of use :
3166 * parse(cmd, &err); (callee: memprintf(&err, ...))
3167 * fprintf(stderr, "Parser said: %s\n", indent_error(&err));
3168 * free(err);
3169 */
3170char *indent_msg(char **out, int level)
3171{
3172 char *ret, *in, *p;
3173 int needed = 0;
3174 int lf = 0;
3175 int lastlf = 0;
3176 int len;
3177
Willy Tarreau70eec382012-10-10 08:56:47 +02003178 if (!out || !*out)
3179 return NULL;
3180
Willy Tarreau21c705b2012-09-14 11:40:36 +02003181 in = *out - 1;
3182 while ((in = strchr(in + 1, '\n')) != NULL) {
3183 lastlf = in - *out;
3184 lf++;
3185 }
3186
3187 if (!lf) /* single line, no LF, return it as-is */
3188 return *out;
3189
3190 len = strlen(*out);
3191
3192 if (lf == 1 && lastlf == len - 1) {
3193 /* single line, LF at end, strip it and return as-is */
3194 (*out)[lastlf] = 0;
3195 return *out;
3196 }
3197
3198 /* OK now we have at least one LF, we need to process the whole string
3199 * as a multi-line string. What we'll do :
3200 * - prefix with an LF if there is none
3201 * - add <level> spaces before each line
3202 * This means at most ( 1 + level + (len-lf) + lf*<1+level) ) =
3203 * 1 + level + len + lf * level = 1 + level * (lf + 1) + len.
3204 */
3205
3206 needed = 1 + level * (lf + 1) + len + 1;
3207 p = ret = malloc(needed);
3208 in = *out;
3209
3210 /* skip initial LFs */
3211 while (*in == '\n')
3212 in++;
3213
3214 /* copy each line, prefixed with LF and <level> spaces, and without the trailing LF */
3215 while (*in) {
3216 *p++ = '\n';
3217 memset(p, ' ', level);
3218 p += level;
3219 do {
3220 *p++ = *in++;
3221 } while (*in && *in != '\n');
3222 if (*in)
3223 in++;
3224 }
3225 *p = 0;
3226
3227 free(*out);
3228 *out = ret;
3229
3230 return ret;
3231}
3232
Willy Tarreaudad36a32013-03-11 01:20:04 +01003233/* Convert occurrences of environment variables in the input string to their
3234 * corresponding value. A variable is identified as a series of alphanumeric
3235 * characters or underscores following a '$' sign. The <in> string must be
3236 * free()able. NULL returns NULL. The resulting string might be reallocated if
3237 * some expansion is made. Variable names may also be enclosed into braces if
3238 * needed (eg: to concatenate alphanum characters).
3239 */
3240char *env_expand(char *in)
3241{
3242 char *txt_beg;
3243 char *out;
3244 char *txt_end;
3245 char *var_beg;
3246 char *var_end;
3247 char *value;
3248 char *next;
3249 int out_len;
3250 int val_len;
3251
3252 if (!in)
3253 return in;
3254
3255 value = out = NULL;
3256 out_len = 0;
3257
3258 txt_beg = in;
3259 do {
3260 /* look for next '$' sign in <in> */
3261 for (txt_end = txt_beg; *txt_end && *txt_end != '$'; txt_end++);
3262
3263 if (!*txt_end && !out) /* end and no expansion performed */
3264 return in;
3265
3266 val_len = 0;
3267 next = txt_end;
3268 if (*txt_end == '$') {
3269 char save;
3270
3271 var_beg = txt_end + 1;
3272 if (*var_beg == '{')
3273 var_beg++;
3274
3275 var_end = var_beg;
3276 while (isalnum((int)(unsigned char)*var_end) || *var_end == '_') {
3277 var_end++;
3278 }
3279
3280 next = var_end;
3281 if (*var_end == '}' && (var_beg > txt_end + 1))
3282 next++;
3283
3284 /* get value of the variable name at this location */
3285 save = *var_end;
3286 *var_end = '\0';
3287 value = getenv(var_beg);
3288 *var_end = save;
3289 val_len = value ? strlen(value) : 0;
3290 }
3291
Hubert Verstraete831962e2016-06-28 22:44:26 +02003292 out = my_realloc2(out, out_len + (txt_end - txt_beg) + val_len + 1);
Willy Tarreaudad36a32013-03-11 01:20:04 +01003293 if (txt_end > txt_beg) {
3294 memcpy(out + out_len, txt_beg, txt_end - txt_beg);
3295 out_len += txt_end - txt_beg;
3296 }
3297 if (val_len) {
3298 memcpy(out + out_len, value, val_len);
3299 out_len += val_len;
3300 }
3301 out[out_len] = 0;
3302 txt_beg = next;
3303 } while (*txt_beg);
3304
3305 /* here we know that <out> was allocated and that we don't need <in> anymore */
3306 free(in);
3307 return out;
3308}
3309
de Lafond Guillaume88c278f2013-04-15 19:27:10 +02003310
3311/* same as strstr() but case-insensitive and with limit length */
3312const char *strnistr(const char *str1, int len_str1, const char *str2, int len_str2)
3313{
3314 char *pptr, *sptr, *start;
Willy Tarreauc8746532014-05-28 23:05:07 +02003315 unsigned int slen, plen;
3316 unsigned int tmp1, tmp2;
de Lafond Guillaume88c278f2013-04-15 19:27:10 +02003317
3318 if (str1 == NULL || len_str1 == 0) // search pattern into an empty string => search is not found
3319 return NULL;
3320
3321 if (str2 == NULL || len_str2 == 0) // pattern is empty => every str1 match
3322 return str1;
3323
3324 if (len_str1 < len_str2) // pattern is longer than string => search is not found
3325 return NULL;
3326
3327 for (tmp1 = 0, start = (char *)str1, pptr = (char *)str2, slen = len_str1, plen = len_str2; slen >= plen; start++, slen--) {
3328 while (toupper(*start) != toupper(*str2)) {
3329 start++;
3330 slen--;
3331 tmp1++;
3332
3333 if (tmp1 >= len_str1)
3334 return NULL;
3335
3336 /* if pattern longer than string */
3337 if (slen < plen)
3338 return NULL;
3339 }
3340
3341 sptr = start;
3342 pptr = (char *)str2;
3343
3344 tmp2 = 0;
3345 while (toupper(*sptr) == toupper(*pptr)) {
3346 sptr++;
3347 pptr++;
3348 tmp2++;
3349
3350 if (*pptr == '\0' || tmp2 == len_str2) /* end of pattern found */
3351 return start;
3352 if (*sptr == '\0' || tmp2 == len_str1) /* end of string found and the pattern is not fully found */
3353 return NULL;
3354 }
3355 }
3356 return NULL;
3357}
3358
Thierry FOURNIER317e1c42014-08-12 10:20:47 +02003359/* This function read the next valid utf8 char.
3360 * <s> is the byte srray to be decode, <len> is its length.
3361 * The function returns decoded char encoded like this:
3362 * The 4 msb are the return code (UTF8_CODE_*), the 4 lsb
3363 * are the length read. The decoded character is stored in <c>.
3364 */
3365unsigned char utf8_next(const char *s, int len, unsigned int *c)
3366{
3367 const unsigned char *p = (unsigned char *)s;
3368 int dec;
3369 unsigned char code = UTF8_CODE_OK;
3370
3371 if (len < 1)
3372 return UTF8_CODE_OK;
3373
3374 /* Check the type of UTF8 sequence
3375 *
3376 * 0... .... 0x00 <= x <= 0x7f : 1 byte: ascii char
3377 * 10.. .... 0x80 <= x <= 0xbf : invalid sequence
3378 * 110. .... 0xc0 <= x <= 0xdf : 2 bytes
3379 * 1110 .... 0xe0 <= x <= 0xef : 3 bytes
3380 * 1111 0... 0xf0 <= x <= 0xf7 : 4 bytes
3381 * 1111 10.. 0xf8 <= x <= 0xfb : 5 bytes
3382 * 1111 110. 0xfc <= x <= 0xfd : 6 bytes
3383 * 1111 111. 0xfe <= x <= 0xff : invalid sequence
3384 */
3385 switch (*p) {
3386 case 0x00 ... 0x7f:
3387 *c = *p;
3388 return UTF8_CODE_OK | 1;
3389
3390 case 0x80 ... 0xbf:
3391 *c = *p;
3392 return UTF8_CODE_BADSEQ | 1;
3393
3394 case 0xc0 ... 0xdf:
3395 if (len < 2) {
3396 *c = *p;
3397 return UTF8_CODE_BADSEQ | 1;
3398 }
3399 *c = *p & 0x1f;
3400 dec = 1;
3401 break;
3402
3403 case 0xe0 ... 0xef:
3404 if (len < 3) {
3405 *c = *p;
3406 return UTF8_CODE_BADSEQ | 1;
3407 }
3408 *c = *p & 0x0f;
3409 dec = 2;
3410 break;
3411
3412 case 0xf0 ... 0xf7:
3413 if (len < 4) {
3414 *c = *p;
3415 return UTF8_CODE_BADSEQ | 1;
3416 }
3417 *c = *p & 0x07;
3418 dec = 3;
3419 break;
3420
3421 case 0xf8 ... 0xfb:
3422 if (len < 5) {
3423 *c = *p;
3424 return UTF8_CODE_BADSEQ | 1;
3425 }
3426 *c = *p & 0x03;
3427 dec = 4;
3428 break;
3429
3430 case 0xfc ... 0xfd:
3431 if (len < 6) {
3432 *c = *p;
3433 return UTF8_CODE_BADSEQ | 1;
3434 }
3435 *c = *p & 0x01;
3436 dec = 5;
3437 break;
3438
3439 case 0xfe ... 0xff:
3440 default:
3441 *c = *p;
3442 return UTF8_CODE_BADSEQ | 1;
3443 }
3444
3445 p++;
3446
3447 while (dec > 0) {
3448
3449 /* need 0x10 for the 2 first bits */
3450 if ( ( *p & 0xc0 ) != 0x80 )
3451 return UTF8_CODE_BADSEQ | ((p-(unsigned char *)s)&0xffff);
3452
3453 /* add data at char */
3454 *c = ( *c << 6 ) | ( *p & 0x3f );
3455
3456 dec--;
3457 p++;
3458 }
3459
3460 /* Check ovelong encoding.
3461 * 1 byte : 5 + 6 : 11 : 0x80 ... 0x7ff
3462 * 2 bytes : 4 + 6 + 6 : 16 : 0x800 ... 0xffff
3463 * 3 bytes : 3 + 6 + 6 + 6 : 21 : 0x10000 ... 0x1fffff
3464 */
Thierry FOURNIER9e7ec082015-03-12 19:32:38 +01003465 if (( *c <= 0x7f && (p-(unsigned char *)s) > 1) ||
Thierry FOURNIER317e1c42014-08-12 10:20:47 +02003466 (*c >= 0x80 && *c <= 0x7ff && (p-(unsigned char *)s) > 2) ||
3467 (*c >= 0x800 && *c <= 0xffff && (p-(unsigned char *)s) > 3) ||
3468 (*c >= 0x10000 && *c <= 0x1fffff && (p-(unsigned char *)s) > 4))
3469 code |= UTF8_CODE_OVERLONG;
3470
3471 /* Check invalid UTF8 range. */
3472 if ((*c >= 0xd800 && *c <= 0xdfff) ||
3473 (*c >= 0xfffe && *c <= 0xffff))
3474 code |= UTF8_CODE_INVRANGE;
3475
3476 return code | ((p-(unsigned char *)s)&0x0f);
3477}
3478
Maxime de Roucydc887852016-05-13 23:52:54 +02003479/* append a copy of string <str> (in a wordlist) at the end of the list <li>
3480 * On failure : return 0 and <err> filled with an error message.
3481 * The caller is responsible for freeing the <err> and <str> copy
3482 * memory area using free()
3483 */
3484int list_append_word(struct list *li, const char *str, char **err)
3485{
3486 struct wordlist *wl;
3487
3488 wl = calloc(1, sizeof(*wl));
3489 if (!wl) {
3490 memprintf(err, "out of memory");
3491 goto fail_wl;
3492 }
3493
3494 wl->s = strdup(str);
3495 if (!wl->s) {
3496 memprintf(err, "out of memory");
3497 goto fail_wl_s;
3498 }
3499
3500 LIST_ADDQ(li, &wl->list);
3501
3502 return 1;
3503
3504fail_wl_s:
3505 free(wl->s);
3506fail_wl:
3507 free(wl);
3508 return 0;
3509}
3510
Willy Tarreaubaaee002006-06-26 02:48:02 +02003511/*
3512 * Local variables:
3513 * c-indent-level: 8
3514 * c-basic-offset: 8
3515 * End:
3516 */