1 /*
2 * http_date - HTTP date manipulation
3 *
4 * Copyright(c) 2015 Glenn Strauss gstrauss()gluelogic.com All rights reserved
5 * License: BSD 3-clause (same as lighttpd)
6 */
7 #include "http_date.h"
8
9 #include "sys-time.h"
10 #include <string.h> /* strlen() */
11
12 #include "buffer.h" /* light_isdigit() */
13 #include "log.h" /* log_epoch_secs */
14
15 /**
16 * https://tools.ietf.org/html/rfc7231
17 * [RFC7231] 7.1.1.1 Date/Time Formats
18 * Prior to 1995, there were three different formats commonly used by
19 * servers to communicate timestamps. For compatibility with old
20 * implementations, all three are defined here. The preferred format is
21 * a fixed-length and single-zone subset of the date and time
22 * specification used by the Internet Message Format [RFC5322].
23 * HTTP-date = IMF-fixdate / obs-date
24 * An example of the preferred format is
25 * Sun, 06 Nov 1994 08:49:37 GMT ; IMF-fixdate
26 *
27 *
28 * (intended for use with strftime() and strptime())
29 * "%a, %d %b %Y %T GMT"
30 */
31
32
33 static const char datestrs[] =
34 /*0 10 20 30 40 50 60 70 80 90*/
35 "\0\x0A\x14\x1E\x28\x32\x3c\x46\x50\x5A"
36 "SunMonTueWedThuFriSat"
37 "JanFebMarAprMayJunJulAugSepOctNovDec";
38
39
40 __attribute_cold__
41 static const char *
http_date_parse_RFC_850(const char * s,struct tm * const tm)42 http_date_parse_RFC_850 (const char *s, struct tm * const tm)
43 {
44 /* RFC 7231 7.1.1.1.
45 * Recipients of a timestamp value in rfc850-date format, which uses a
46 * two-digit year, MUST interpret a timestamp that appears to be more
47 * than 50 years in the future as representing the most recent year in
48 * the past that had the same last two digits.
49 */
50 static unix_time64_t tm_year_last_check;
51 static int tm_year_cur;
52 static int tm_year_base;
53 /* (log_epoch_secs is a global variable, maintained elsewhere) */
54 /* (optimization: check for year change no more than once per min) */
55 if (log_epoch_secs >= tm_year_last_check + 60) {
56 struct tm tm_cur;
57 if (NULL != gmtime64_r(&log_epoch_secs, &tm_cur)) {
58 tm_year_last_check = log_epoch_secs;
59 if (tm_cur.tm_year != tm_year_cur) {
60 tm_year_cur = tm_cur.tm_year;
61 tm_year_base = tm_year_cur - (tm_year_cur % 100);
62 }
63 }
64 }
65
66 /* Note: does not validate numerical ranges of
67 * tm_mday, tm_hour, tm_min, tm_sec */
68 /* Note: does not validate tm_wday beyond first three chars */
69
70 tm->tm_isdst = 0;
71 tm->tm_yday = 0;
72 tm->tm_wday = 0;
73 tm->tm_mon = 0;
74
75 const char * const tens = datestrs;
76
77 const char *p = tens + 10;
78 do {
79 if (s[0] == p[0] && s[1] == p[1] && s[2] == p[2]) break;
80 p += 3;
81 } while (++tm->tm_wday < 7);
82 if (7 == tm->tm_wday) return NULL;
83
84 s += 3;
85 while (*s != ',' && *s != '\0') ++s;
86
87 if (s[0] != ',' || s[1] != ' '
88 || !light_isdigit(s[2]) || !light_isdigit(s[3]))
89 return NULL;
90 tm->tm_mday = tens[(s[2]-'0')] + (s[3]-'0');
91
92 if ( s[4] != '-') return NULL;
93 p = tens + 10 + sizeof("SunMonTueWedThuFriSat")-1;
94 do {
95 if (s[5] == p[0] && s[6] == p[1] && s[7] == p[2]) break;
96 p += 3;
97 } while (++tm->tm_mon < 12);
98 if (12 == tm->tm_mon) return NULL;
99
100 if (s[8] != '-' || !light_isdigit(s[9]) || !light_isdigit(s[10]))
101 return NULL;
102 tm->tm_year = tens[(s[9]-'0')] + (s[10]-'0') + tm_year_base;
103 if (tm->tm_year > tm_year_cur + 50) tm->tm_year -= 100;
104
105 if (s[11] != ' ' || !light_isdigit(s[12]) || !light_isdigit(s[13]))
106 return NULL;
107 tm->tm_hour = tens[(s[12]-'0')] + (s[13]-'0');
108
109 if (s[14] != ':' || !light_isdigit(s[15]) || !light_isdigit(s[16]))
110 return NULL;
111 tm->tm_min = tens[(s[15]-'0')] + (s[16]-'0');
112
113 if (s[17] != ':' || !light_isdigit(s[18]) || !light_isdigit(s[19]))
114 return NULL;
115 tm->tm_sec = tens[(s[18]-'0')] + (s[19]-'0');
116
117 if (s[20] != ' ' || s[21] != 'G' || s[22] != 'M' || s[23] != 'T')
118 return NULL;
119
120 return s+24; /*(24 chars from ',' following the variable len wday)*/
121 }
122
123
124 __attribute_cold__
125 static const char *
http_date_parse_asctime(const char * const s,struct tm * const tm)126 http_date_parse_asctime (const char * const s, struct tm * const tm)
127 {
128 /* Note: does not validate numerical ranges of
129 * tm_mday, tm_hour, tm_min, tm_sec */
130
131 tm->tm_isdst = 0;
132 tm->tm_yday = 0;
133 tm->tm_wday = 0;
134 tm->tm_mon = 0;
135
136 const char * const tens = datestrs;
137
138 const char *p = tens + 10;
139 do {
140 if (s[0] == p[0] && s[1] == p[1] && s[2] == p[2]) break;
141 p += 3;
142 } while (++tm->tm_wday < 7);
143 if (7 == tm->tm_wday) return NULL;
144
145 if (s[3] != ' ') return NULL;
146 p = tens + 10 + sizeof("SunMonTueWedThuFriSat")-1;
147 do {
148 if (s[4] == p[0] && s[5] == p[1] && s[6] == p[2]) break;
149 p += 3;
150 } while (++tm->tm_mon < 12);
151 if (12 == tm->tm_mon) return NULL;
152
153 if (s[7] != ' ' || (s[8] != ' ' && !light_isdigit(s[8]))
154 || !light_isdigit(s[9]))
155 return NULL;
156 tm->tm_mday = (s[8] == ' ' ? 0 : tens[(s[8]-'0')]) + (s[9]-'0');
157
158 if (s[10] != ' ' || !light_isdigit(s[11]) || !light_isdigit(s[12]))
159 return NULL;
160 tm->tm_hour = tens[(s[11]-'0')] + (s[12]-'0');
161
162 if (s[13] != ':' || !light_isdigit(s[14]) || !light_isdigit(s[15]))
163 return NULL;
164 tm->tm_min = tens[(s[14]-'0')] + (s[15]-'0');
165
166 if (s[16] != ':' || !light_isdigit(s[17]) || !light_isdigit(s[18]))
167 return NULL;
168 tm->tm_sec = tens[(s[17]-'0')] + (s[18]-'0');
169
170 if (s[19] != ' ' || !light_isdigit(s[20]) || !light_isdigit(s[21])
171 || !light_isdigit(s[22]) || !light_isdigit(s[23])) return NULL;
172 tm->tm_year =(tens[(s[20]-'0')] + (s[21]-'0'))*100
173 + tens[(s[22]-'0')] + (s[23]-'0') - 1900;
174
175 return s+24;
176 }
177
178
179 static const char *
http_date_parse_IMF_fixdate(const char * const s,struct tm * const tm)180 http_date_parse_IMF_fixdate (const char * const s, struct tm * const tm)
181 {
182 /* Note: does not validate numerical ranges of
183 * tm_mday, tm_hour, tm_min, tm_sec */
184
185 tm->tm_isdst = 0;
186 tm->tm_yday = 0;
187 tm->tm_wday = 0;
188 tm->tm_mon = 0;
189
190 const char * const tens = datestrs;
191
192 const char *p = tens + 10;
193 do {
194 if (s[0] == p[0] && s[1] == p[1] && s[2] == p[2]) break;
195 p += 3;
196 } while (++tm->tm_wday < 7);
197 if (7 == tm->tm_wday) return NULL;
198
199 if (s[3] != ',' || s[4] != ' '
200 || !light_isdigit(s[5]) || !light_isdigit(s[6]))
201 return NULL;
202 tm->tm_mday = tens[(s[5]-'0')] + (s[6]-'0');
203
204 if ( s[7] != ' ') return NULL;
205 p = tens + 10 + sizeof("SunMonTueWedThuFriSat")-1;
206 do {
207 if (s[8] == p[0] && s[9] == p[1] && s[10] == p[2]) break;
208 p += 3;
209 } while (++tm->tm_mon < 12);
210 if (12 == tm->tm_mon) return NULL;
211
212 if (s[11] != ' ' || !light_isdigit(s[12]) || !light_isdigit(s[13])
213 || !light_isdigit(s[14]) || !light_isdigit(s[15])) return NULL;
214 tm->tm_year =(tens[(s[12]-'0')] + (s[13]-'0'))*100
215 + tens[(s[14]-'0')] + (s[15]-'0') - 1900;
216
217 if (s[16] != ' ' || !light_isdigit(s[17]) || !light_isdigit(s[18]))
218 return NULL;
219 tm->tm_hour = tens[(s[17]-'0')] + (s[18]-'0');
220
221 if (s[19] != ':' || !light_isdigit(s[20]) || !light_isdigit(s[21]))
222 return NULL;
223 tm->tm_min = tens[(s[20]-'0')] + (s[21]-'0');
224
225 if (s[22] != ':' || !light_isdigit(s[23]) || !light_isdigit(s[24]))
226 return NULL;
227 tm->tm_sec = tens[(s[23]-'0')] + (s[24]-'0');
228
229 if (s[25] != ' ' || s[26] != 'G' || s[27] != 'M' || s[28] != 'T')
230 return NULL;
231
232 return s+29;
233 }
234
235
236 static const char *
http_date_str_to_tm(const char * const s,const uint32_t len,struct tm * const tm)237 http_date_str_to_tm (const char * const s, const uint32_t len,
238 struct tm * const tm)
239 {
240
241 /* attempt strptime() using multiple date formats
242 * support RFC 822,1123,7231; RFC 850; and ANSI C asctime() date strings,
243 * as required by [RFC7231] https://tools.ietf.org/html/rfc7231#section-7.1
244 * [RFC7231] 7.1.1.1 Date/Time Formats
245 * HTTP-date = IMF-fixdate / obs-date
246 * [...]
247 * A recipient that parses a timestamp value in an HTTP header field
248 * MUST accept all three HTTP-date formats.
249 */
250
251 /* employ specialized strptime()
252 * - HTTP expected date formats are known, so not needed as input param
253 * - HTTP expected date string content is in C locale and is case-sensitive
254 * - returns (const char *) instead of strptime() (char *) return type
255 * - returns NULL if error (if date string could not be parsed)
256 * Note: internal implementation requires '\0'-terminated string, or at
257 * least one valid char after partial match of RFC 850 or asctime formats */
258 if (len == 29)
259 return http_date_parse_IMF_fixdate(s, tm);
260 else if (len > 29)
261 return http_date_parse_RFC_850(s, tm);
262 else /* len < 29 */
263 return http_date_parse_asctime(s, tm);
264 }
265
266
267 uint32_t
http_date_time_to_str(char * const s,const size_t sz,const unix_time64_t t)268 http_date_time_to_str (char * const s, const size_t sz, const unix_time64_t t)
269 {
270 /*('max' is expected to be >= 30 (IMF-fixdate is 29 chars + '\0'))*/
271 struct tm tm;
272 const char fmt[] = "%a, %d %b %Y %T GMT"; /*IMF-fixdate fmt*/
273 return (__builtin_expect( (NULL != gmtime64_r(&t, &tm)), 1))
274 ? (uint32_t)strftime(s, sz, fmt, &tm)
275 : 0;
276 }
277
278
279 int
http_date_if_modified_since(const char * const ifmod,const uint32_t ifmodlen,const unix_time64_t lmtime)280 http_date_if_modified_since (const char * const ifmod, const uint32_t ifmodlen,
281 const unix_time64_t lmtime)
282 {
283 struct tm ifmodtm;
284 if (NULL == http_date_str_to_tm(ifmod, ifmodlen, &ifmodtm))
285 return 1; /* date parse error */
286 const time_t ifmtime = timegm(&ifmodtm);
287 #if HAS_TIME_BITS64
288 return (lmtime > ifmtime);
289 #else
290 return (TIME64_CAST(lmtime) > TIME64_CAST(ifmtime) || ifmtime==(time_t)-1);
291 #endif
292 /* returns 0 if not modified since,
293 * returns 1 if modified since or date parse error */
294 }
295