1/***************************************************************************
2 * _ _ ____ _
3 * Project ___| | | | _ \| |
4 * / __| | | | |_) | |
5 * | (__| |_| | _ <| |___
6 * \___|\___/|_| \_\_____|
7 *
8 * Copyright (C) 1998 - 2020, Daniel Stenberg, <daniel@haxx.se>, et al.
9 *
10 * This software is licensed as described in the file COPYING, which
11 * you should have received as part of this distribution. The terms
12 * are also available at https://curl.se/docs/copyright.html.
13 *
14 * You may opt to use, copy, modify, merge, publish, distribute and/or sell
15 * copies of the Software, and permit persons to whom the Software is
16 * furnished to do so, under the terms of the COPYING file.
17 *
18 * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
19 * KIND, either express or implied.
20 *
21 ***************************************************************************/
22/*
23 A brief summary of the date string formats this parser groks:
24
25 RFC 2616 3.3.1
26
27 Sun, 06 Nov 1994 08:49:37 GMT ; RFC 822, updated by RFC 1123
28 Sunday, 06-Nov-94 08:49:37 GMT ; RFC 850, obsoleted by RFC 1036
29 Sun Nov 6 08:49:37 1994 ; ANSI C's asctime() format
30
31 we support dates without week day name:
32
33 06 Nov 1994 08:49:37 GMT
34 06-Nov-94 08:49:37 GMT
35 Nov 6 08:49:37 1994
36
37 without the time zone:
38
39 06 Nov 1994 08:49:37
40 06-Nov-94 08:49:37
41
42 weird order:
43
44 1994 Nov 6 08:49:37 (GNU date fails)
45 GMT 08:49:37 06-Nov-94 Sunday
46 94 6 Nov 08:49:37 (GNU date fails)
47
48 time left out:
49
50 1994 Nov 6
51 06-Nov-94
52 Sun Nov 6 94
53
54 unusual separators:
55
56 1994.Nov.6
57 Sun/Nov/6/94/GMT
58
59 commonly used time zone names:
60
61 Sun, 06 Nov 1994 08:49:37 CET
62 06 Nov 1994 08:49:37 EST
63
64 time zones specified using RFC822 style:
65
66 Sun, 12 Sep 2004 15:05:58 -0700
67 Sat, 11 Sep 2004 21:32:11 +0200
68
69 compact numerical date strings:
70
71 20040912 15:05:58 -0700
72 20040911 +0200
73
74*/
75
76#include "curl_setup.h"
77
78#include <limits.h>
79
80#include <curl/curl.h>
81#include "strcase.h"
82#include "warnless.h"
83#include "parsedate.h"
84
85/*
86 * parsedate()
87 *
88 * Returns:
89 *
90 * PARSEDATE_OK - a fine conversion
91 * PARSEDATE_FAIL - failed to convert
92 * PARSEDATE_LATER - time overflow at the far end of time_t
93 * PARSEDATE_SOONER - time underflow at the low end of time_t
94 */
95
96static int parsedate(const char *date, time_t *output);
97
98#define PARSEDATE_OK 0
99#define PARSEDATE_FAIL -1
100#define PARSEDATE_LATER 1
101#define PARSEDATE_SOONER 2
102
103#if !defined(CURL_DISABLE_PARSEDATE) || !defined(CURL_DISABLE_FTP) || \
104 !defined(CURL_DISABLE_FILE)
105/* These names are also used by FTP and FILE code */
106const char * const Curl_wkday[] =
107{"Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun"};
108const char * const Curl_month[]=
109{ "Jan", "Feb", "Mar", "Apr", "May", "Jun",
110 "Jul", "Aug", "Sep", "Oct", "Nov", "Dec" };
111#endif
112
113#ifndef CURL_DISABLE_PARSEDATE
114static const char * const weekday[] =
115{ "Monday", "Tuesday", "Wednesday", "Thursday",
116 "Friday", "Saturday", "Sunday" };
117
118struct tzinfo {
119 char name[5];
120 int offset; /* +/- in minutes */
121};
122
123/* Here's a bunch of frequently used time zone names. These were supported
124 by the old getdate parser. */
125#define tDAYZONE -60 /* offset for daylight savings time */
126static const struct tzinfo tz[]= {
127 {"GMT", 0}, /* Greenwich Mean */
128 {"UT", 0}, /* Universal Time */
129 {"UTC", 0}, /* Universal (Coordinated) */
130 {"WET", 0}, /* Western European */
131 {"BST", 0 tDAYZONE}, /* British Summer */
132 {"WAT", 60}, /* West Africa */
133 {"AST", 240}, /* Atlantic Standard */
134 {"ADT", 240 tDAYZONE}, /* Atlantic Daylight */
135 {"EST", 300}, /* Eastern Standard */
136 {"EDT", 300 tDAYZONE}, /* Eastern Daylight */
137 {"CST", 360}, /* Central Standard */
138 {"CDT", 360 tDAYZONE}, /* Central Daylight */
139 {"MST", 420}, /* Mountain Standard */
140 {"MDT", 420 tDAYZONE}, /* Mountain Daylight */
141 {"PST", 480}, /* Pacific Standard */
142 {"PDT", 480 tDAYZONE}, /* Pacific Daylight */
143 {"YST", 540}, /* Yukon Standard */
144 {"YDT", 540 tDAYZONE}, /* Yukon Daylight */
145 {"HST", 600}, /* Hawaii Standard */
146 {"HDT", 600 tDAYZONE}, /* Hawaii Daylight */
147 {"CAT", 600}, /* Central Alaska */
148 {"AHST", 600}, /* Alaska-Hawaii Standard */
149 {"NT", 660}, /* Nome */
150 {"IDLW", 720}, /* International Date Line West */
151 {"CET", -60}, /* Central European */
152 {"MET", -60}, /* Middle European */
153 {"MEWT", -60}, /* Middle European Winter */
154 {"MEST", -60 tDAYZONE}, /* Middle European Summer */
155 {"CEST", -60 tDAYZONE}, /* Central European Summer */
156 {"MESZ", -60 tDAYZONE}, /* Middle European Summer */
157 {"FWT", -60}, /* French Winter */
158 {"FST", -60 tDAYZONE}, /* French Summer */
159 {"EET", -120}, /* Eastern Europe, USSR Zone 1 */
160 {"WAST", -420}, /* West Australian Standard */
161 {"WADT", -420 tDAYZONE}, /* West Australian Daylight */
162 {"CCT", -480}, /* China Coast, USSR Zone 7 */
163 {"JST", -540}, /* Japan Standard, USSR Zone 8 */
164 {"EAST", -600}, /* Eastern Australian Standard */
165 {"EADT", -600 tDAYZONE}, /* Eastern Australian Daylight */
166 {"GST", -600}, /* Guam Standard, USSR Zone 9 */
167 {"NZT", -720}, /* New Zealand */
168 {"NZST", -720}, /* New Zealand Standard */
169 {"NZDT", -720 tDAYZONE}, /* New Zealand Daylight */
170 {"IDLE", -720}, /* International Date Line East */
171 /* Next up: Military timezone names. RFC822 allowed these, but (as noted in
172 RFC 1123) had their signs wrong. Here we use the correct signs to match
173 actual military usage.
174 */
175 {"A", 1 * 60}, /* Alpha */
176 {"B", 2 * 60}, /* Bravo */
177 {"C", 3 * 60}, /* Charlie */
178 {"D", 4 * 60}, /* Delta */
179 {"E", 5 * 60}, /* Echo */
180 {"F", 6 * 60}, /* Foxtrot */
181 {"G", 7 * 60}, /* Golf */
182 {"H", 8 * 60}, /* Hotel */
183 {"I", 9 * 60}, /* India */
184 /* "J", Juliet is not used as a timezone, to indicate the observer's local
185 time */
186 {"K", 10 * 60}, /* Kilo */
187 {"L", 11 * 60}, /* Lima */
188 {"M", 12 * 60}, /* Mike */
189 {"N", -1 * 60}, /* November */
190 {"O", -2 * 60}, /* Oscar */
191 {"P", -3 * 60}, /* Papa */
192 {"Q", -4 * 60}, /* Quebec */
193 {"R", -5 * 60}, /* Romeo */
194 {"S", -6 * 60}, /* Sierra */
195 {"T", -7 * 60}, /* Tango */
196 {"U", -8 * 60}, /* Uniform */
197 {"V", -9 * 60}, /* Victor */
198 {"W", -10 * 60}, /* Whiskey */
199 {"X", -11 * 60}, /* X-ray */
200 {"Y", -12 * 60}, /* Yankee */
201 {"Z", 0}, /* Zulu, zero meridian, a.k.a. UTC */
202};
203
204/* returns:
205 -1 no day
206 0 monday - 6 sunday
207*/
208
209static int checkday(const char *check, size_t len)
210{
211 int i;
212 const char * const *what;
213 bool found = FALSE;
214 if(len > 3)
215 what = &weekday[0];
216 else
217 what = &Curl_wkday[0];
218 for(i = 0; i<7; i++) {
219 if(strcasecompare(check, what[0])) {
220 found = TRUE;
221 break;
222 }
223 what++;
224 }
225 return found?i:-1;
226}
227
228static int checkmonth(const char *check)
229{
230 int i;
231 const char * const *what;
232 bool found = FALSE;
233
234 what = &Curl_month[0];
235 for(i = 0; i<12; i++) {
236 if(strcasecompare(check, what[0])) {
237 found = TRUE;
238 break;
239 }
240 what++;
241 }
242 return found?i:-1; /* return the offset or -1, no real offset is -1 */
243}
244
245/* return the time zone offset between GMT and the input one, in number
246 of seconds or -1 if the timezone wasn't found/legal */
247
248static int checktz(const char *check)
249{
250 unsigned int i;
251 const struct tzinfo *what;
252 bool found = FALSE;
253
254 what = tz;
255 for(i = 0; i< sizeof(tz)/sizeof(tz[0]); i++) {
256 if(strcasecompare(check, what->name)) {
257 found = TRUE;
258 break;
259 }
260 what++;
261 }
262 return found?what->offset*60:-1;
263}
264
265static void skip(const char **date)
266{
267 /* skip everything that aren't letters or digits */
268 while(**date && !ISALNUM(**date))
269 (*date)++;
270}
271
272enum assume {
273 DATE_MDAY,
274 DATE_YEAR,
275 DATE_TIME
276};
277
278/*
279 * time2epoch: time stamp to seconds since epoch in GMT time zone. Similar to
280 * mktime but for GMT only.
281 */
282static time_t time2epoch(int sec, int min, int hour,
283 int mday, int mon, int year)
284{
285 static const int month_days_cumulative [12] =
286 { 0, 31, 59, 90, 120, 151, 181, 212, 243, 273, 304, 334 };
287 int leap_days = year - (mon <= 1);
288 leap_days = ((leap_days / 4) - (leap_days / 100) + (leap_days / 400)
289 - (1969 / 4) + (1969 / 100) - (1969 / 400));
290 return ((((time_t) (year - 1970) * 365
291 + leap_days + month_days_cumulative[mon] + mday - 1) * 24
292 + hour) * 60 + min) * 60 + sec;
293}
294
295/*
296 * parsedate()
297 *
298 * Returns:
299 *
300 * PARSEDATE_OK - a fine conversion
301 * PARSEDATE_FAIL - failed to convert
302 * PARSEDATE_LATER - time overflow at the far end of time_t
303 * PARSEDATE_SOONER - time underflow at the low end of time_t
304 */
305
306static int parsedate(const char *date, time_t *output)
307{
308 time_t t = 0;
309 int wdaynum = -1; /* day of the week number, 0-6 (mon-sun) */
310 int monnum = -1; /* month of the year number, 0-11 */
311 int mdaynum = -1; /* day of month, 1 - 31 */
312 int hournum = -1;
313 int minnum = -1;
314 int secnum = -1;
315 int yearnum = -1;
316 int tzoff = -1;
317 enum assume dignext = DATE_MDAY;
318 const char *indate = date; /* save the original pointer */
319 int part = 0; /* max 6 parts */
320
321 while(*date && (part < 6)) {
322 bool found = FALSE;
323
324 skip(&date);
325
326 if(ISALPHA(*date)) {
327 /* a name coming up */
328 char buf[32]="";
329 size_t len;
330 if(sscanf(date, "%31[ABCDEFGHIJKLMNOPQRSTUVWXYZ"
331 "abcdefghijklmnopqrstuvwxyz]", buf))
332 len = strlen(buf);
333 else
334 len = 0;
335
336 if(wdaynum == -1) {
337 wdaynum = checkday(buf, len);
338 if(wdaynum != -1)
339 found = TRUE;
340 }
341 if(!found && (monnum == -1)) {
342 monnum = checkmonth(buf);
343 if(monnum != -1)
344 found = TRUE;
345 }
346
347 if(!found && (tzoff == -1)) {
348 /* this just must be a time zone string */
349 tzoff = checktz(buf);
350 if(tzoff != -1)
351 found = TRUE;
352 }
353
354 if(!found)
355 return PARSEDATE_FAIL; /* bad string */
356
357 date += len;
358 }
359 else if(ISDIGIT(*date)) {
360 /* a digit */
361 int val;
362 char *end;
363 int len = 0;
364 if((secnum == -1) &&
365 (3 == sscanf(date, "%02d:%02d:%02d%n",
366 &hournum, &minnum, &secnum, &len))) {
367 /* time stamp! */
368 date += len;
369 }
370 else if((secnum == -1) &&
371 (2 == sscanf(date, "%02d:%02d%n", &hournum, &minnum, &len))) {
372 /* time stamp without seconds */
373 date += len;
374 secnum = 0;
375 }
376 else {
377 long lval;
378 int error;
379 int old_errno;
380
381 old_errno = errno;
382 errno = 0;
383 lval = strtol(date, &end, 10);
384 error = errno;
385 if(errno != old_errno)
386 errno = old_errno;
387
388 if(error)
389 return PARSEDATE_FAIL;
390
391#if LONG_MAX != INT_MAX
392 if((lval > (long)INT_MAX) || (lval < (long)INT_MIN))
393 return PARSEDATE_FAIL;
394#endif
395
396 val = curlx_sltosi(lval);
397
398 if((tzoff == -1) &&
399 ((end - date) == 4) &&
400 (val <= 1400) &&
401 (indate< date) &&
402 ((date[-1] == '+' || date[-1] == '-'))) {
403 /* four digits and a value less than or equal to 1400 (to take into
404 account all sorts of funny time zone diffs) and it is preceded
405 with a plus or minus. This is a time zone indication. 1400 is
406 picked since +1300 is frequently used and +1400 is mentioned as
407 an edge number in the document "ISO C 200X Proposal: Timezone
408 Functions" at http://david.tribble.com/text/c0xtimezone.html If
409 anyone has a more authoritative source for the exact maximum time
410 zone offsets, please speak up! */
411 found = TRUE;
412 tzoff = (val/100 * 60 + val%100)*60;
413
414 /* the + and - prefix indicates the local time compared to GMT,
415 this we need their reversed math to get what we want */
416 tzoff = date[-1]=='+'?-tzoff:tzoff;
417 }
418
419 if(((end - date) == 8) &&
420 (yearnum == -1) &&
421 (monnum == -1) &&
422 (mdaynum == -1)) {
423 /* 8 digits, no year, month or day yet. This is YYYYMMDD */
424 found = TRUE;
425 yearnum = val/10000;
426 monnum = (val%10000)/100-1; /* month is 0 - 11 */
427 mdaynum = val%100;
428 }
429
430 if(!found && (dignext == DATE_MDAY) && (mdaynum == -1)) {
431 if((val > 0) && (val<32)) {
432 mdaynum = val;
433 found = TRUE;
434 }
435 dignext = DATE_YEAR;
436 }
437
438 if(!found && (dignext == DATE_YEAR) && (yearnum == -1)) {
439 yearnum = val;
440 found = TRUE;
441 if(yearnum < 100) {
442 if(yearnum > 70)
443 yearnum += 1900;
444 else
445 yearnum += 2000;
446 }
447 if(mdaynum == -1)
448 dignext = DATE_MDAY;
449 }
450
451 if(!found)
452 return PARSEDATE_FAIL;
453
454 date = end;
455 }
456 }
457
458 part++;
459 }
460
461 if(-1 == secnum)
462 secnum = minnum = hournum = 0; /* no time, make it zero */
463
464 if((-1 == mdaynum) ||
465 (-1 == monnum) ||
466 (-1 == yearnum))
467 /* lacks vital info, fail */
468 return PARSEDATE_FAIL;
469
470#ifdef HAVE_TIME_T_UNSIGNED
471 if(yearnum < 1970) {
472 /* only positive numbers cannot return earlier */
473 *output = TIME_T_MIN;
474 return PARSEDATE_SOONER;
475 }
476#endif
477
478#if (SIZEOF_TIME_T < 5)
479
480#ifdef HAVE_TIME_T_UNSIGNED
481 /* an unsigned 32 bit time_t can only hold dates to 2106 */
482 if(yearnum > 2105) {
483 *output = TIME_T_MAX;
484 return PARSEDATE_LATER;
485 }
486#else
487 /* a signed 32 bit time_t can only hold dates to the beginning of 2038 */
488 if(yearnum > 2037) {
489 *output = TIME_T_MAX;
490 return PARSEDATE_LATER;
491 }
492 if(yearnum < 1903) {
493 *output = TIME_T_MIN;
494 return PARSEDATE_SOONER;
495 }
496#endif
497
498#else
499 /* The Gregorian calendar was introduced 1582 */
500 if(yearnum < 1583)
501 return PARSEDATE_FAIL;
502#endif
503
504 if((mdaynum > 31) || (monnum > 11) ||
505 (hournum > 23) || (minnum > 59) || (secnum > 60))
506 return PARSEDATE_FAIL; /* clearly an illegal date */
507
508 /* time2epoch() returns a time_t. time_t is often 32 bits, sometimes even on
509 architectures that feature 64 bit 'long' but ultimately time_t is the
510 correct data type to use.
511 */
512 t = time2epoch(secnum, minnum, hournum, mdaynum, monnum, yearnum);
513
514 /* Add the time zone diff between local time zone and GMT. */
515 if(tzoff == -1)
516 tzoff = 0;
517
518 if((tzoff > 0) && (t > TIME_T_MAX - tzoff)) {
519 *output = TIME_T_MAX;
520 return PARSEDATE_LATER; /* time_t overflow */
521 }
522
523 t += tzoff;
524
525 *output = t;
526
527 return PARSEDATE_OK;
528}
529#else
530/* disabled */
531static int parsedate(const char *date, time_t *output)
532{
533 (void)date;
534 *output = 0;
535 return PARSEDATE_OK; /* a lie */
536}
537#endif
538
539time_t curl_getdate(const char *p, const time_t *now)
540{
541 time_t parsed = -1;
542 int rc = parsedate(p, &parsed);
543 (void)now; /* legacy argument from the past that we ignore */
544
545 if(rc == PARSEDATE_OK) {
546 if(parsed == -1)
547 /* avoid returning -1 for a working scenario */
548 parsed++;
549 return parsed;
550 }
551 /* everything else is fail */
552 return -1;
553}
554
555/* Curl_getdate_capped() differs from curl_getdate() in that this will return
556 TIME_T_MAX in case the parsed time value was too big, instead of an
557 error. */
558
559time_t Curl_getdate_capped(const char *p)
560{
561 time_t parsed = -1;
562 int rc = parsedate(p, &parsed);
563
564 switch(rc) {
565 case PARSEDATE_OK:
566 if(parsed == -1)
567 /* avoid returning -1 for a working scenario */
568 parsed++;
569 return parsed;
570 case PARSEDATE_LATER:
571 /* this returns the maximum time value */
572 return parsed;
573 default:
574 return -1; /* everything else is fail */
575 }
576 /* UNREACHABLE */
577}
578
579/*
580 * Curl_gmtime() is a gmtime() replacement for portability. Do not use the
581 * gmtime_r() or gmtime() functions anywhere else but here.
582 *
583 */
584
585CURLcode Curl_gmtime(time_t intime, struct tm *store)
586{
587 const struct tm *tm;
588#ifdef HAVE_GMTIME_R
589 /* thread-safe version */
590 tm = (struct tm *)gmtime_r(&intime, store);
591#else
592 /* !checksrc! disable BANNEDFUNC 1 */
593 tm = gmtime(&intime);
594 if(tm)
595 *store = *tm; /* copy the pointed struct to the local copy */
596#endif
597
598 if(!tm)
599 return CURLE_BAD_FUNCTION_ARGUMENT;
600 return CURLE_OK;
601}
602