1 /*
2  * Copyright (c) 2020 Omar Polo <op@omarpolo.com>
3  *
4  * Permission to use, copy, modify, and distribute this software for any
5  * purpose with or without fee is hereby granted, provided that the above
6  * copyright notice and this permission notice appear in all copies.
7  *
8  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
9  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
10  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
11  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
12  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
13  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
14  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
15  */
16 
17 #include <ctype.h>
18 #include <string.h>
19 
20 #include "gmid.h"
21 
22 /*
23  * Notes from RFC3986
24  *
25  * => gemini://tanso.net/rfc/rfc3986.txt
26  *
27  *
28  * ABNF
29  * ====
30  *
31  * pct-encoded	"%" HEXDIG HEXDIG
32  *
33  * reserved	= gen-delims / sub-delimis
34  * gen-delims	= ":" / "/" / "?" / "#" / "[" / "]" / "@"
35  * sub-delims	= "!" / "$" / "&" / "'" / "(" / ")"
36  * 		/ "*" / "+" / "," / ";" / "="
37  *
38  * unreserved	= ALPHA / DIGIT / "-" / "." / "_" / "~"
39  *
40  * URI		= scheme ":" hier-part [ "?" query ] [ "#" fragment ]
41  *
42  * hier-part	= "//" authority path-abempty
43  * 		/ path-absolute
44  * 		/ path-rootless
45  * 		/ path-empty
46  *
47  * scheme	= ALPHA *( ALPHA / DIGIT / "+" / "-" / "." )
48  *
49  * authority	= [ userinfo "@" ] host [ ":" port ]
50  *
51  * (note that userinfo isn't used for Gemini URL)
52  *
53  * host		= IP-literal / IPv4address / reg-name
54  * reg-name	= *( unreserved / pct-encoded / sub-delims )
55  *
56  * port		= *DIGIT
57  *
58  * path		= path-abemty	; begins with "/" or is empty
59  * 		/ path-absolute	; begins with "/" but not "//"
60  * 		/ path-noscheme	; begins with a non-colon segment
61  * 		/ path-rootless ; begins with a segment
62  * 		/ path-empty	; zero characters
63  *
64  * path-abemty		= *( "/" segment )
65  * path-absolute	= "/" [ segment-nz *( "/" segment ) ]
66  * path-noscheme	= ; not used
67  * path-rootless	= ; not used
68  * path-empty		= ; not used
69  *
70  * segment		= *pchar
71  * segment-nz	= 1*pchar
72  * segment-nz-nc	= 1*( unreserved / pct-encoded / sub-delims / "@" )
73  * pchar		= unreserved / pct-encoded / sub-delims / ":" / "@"
74  *
75  * query		= *( pchar / "/" / "?" )
76  *
77  * fragment		= *( pchar / "/" / "?" )
78  *
79  *
80  * EXAMPLE
81  * =======
82  *
83  *    foo://example.com:8042/over/there?name=ferret#nose
84  *    \_/   \______________/\_________/ \_________/ \__/
85  *     |           |            |            |        |
86  *  scheme     authority       path        query   fragment
87  *
88  */
89 
90 struct parser {
91 	char		*uri;
92 	struct uri	*parsed;
93 	const char	*err;
94 };
95 
96 #define CONT_BYTE(b) ((b & 0xC0) == 0x80)
97 
98 /* XXX: these macros will expand multiple times their argument */
99 
100 #define UNRESERVED(p)				\
101 	(isalnum(p)				\
102 	    || p == '-'				\
103 	    || p == '.'				\
104 	    || p == '_'				\
105 	    || p == '~')
106 
107 #define SUB_DELIMITERS(p)			\
108 	(p == '!'				\
109 	    || p == '$'				\
110 	    || p == '&'				\
111 	    || p == '\''			\
112 	    || p == '('				\
113 	    || p == ')'				\
114 	    || p == '*'				\
115 	    || p == '+'				\
116 	    || p == ','				\
117 	    || p == ';'				\
118 	    || p == '=')
119 
120 /* NOTE: the increments are one less what they should be, because the
121  * caller will add one byte after we return. */
122 static int
123 valid_multibyte_utf8(struct parser *p)
124 {
125 	uint32_t c;
126 	uint8_t s;
127 
128 	c = 0;
129 	s = *p->uri;
130 
131 	if ((s & 0xE0) == 0xC0) {
132 		if (!CONT_BYTE(*(p->uri+1)))
133 			return 0;
134 		c = ((s & 0x1F) << 6) | (*(p->uri+1) & 0x3F);
135 		p->uri += 1;
136 	} else if ((s & 0xF0) == 0xE0) {
137 		if (!CONT_BYTE(*(p->uri+1)) ||
138 		    !CONT_BYTE(*(p->uri+2)))
139 			return 0;
140 		c = (s & 0x0F) << 12
141 			| ((*(p->uri+1) & 0x3F) << 6)
142 			| ((*(p->uri+2) & 0x3F));
143 		p->uri += 2;
144 	} else if ((s & 0xF8) == 0xF0) {
145 		if (!CONT_BYTE(*(p->uri+1)) ||
146 		    !CONT_BYTE(*(p->uri+2)) ||
147 		    !CONT_BYTE(*(p->uri+3)))
148 			return 0;
149 		c = (s & 0x07) << 18
150 			| ((*(p->uri+1) & 0x3F) << 12)
151 			| ((*(p->uri+2) & 0x3F) << 6)
152 			| ((*(p->uri+3) & 0x3F));
153 		p->uri += 3;
154 	} else
155 		return 0;
156 
157 	return (((0x080 <= c) && (c <= 0x7FF))
158 	    || (((0x800 <= c) && (c <= 0xFFFF)))
159 	    || (((0x10000 <= c) && (c <= 0x10FFFF))));
160 }
161 
162 static int
163 parse_pct_encoded(struct parser *p)
164 {
165 	if (*p->uri != '%')
166 		return 0;
167 
168 	if (!isxdigit(*(p->uri+1)) || !isxdigit(*(p->uri+2))) {
169 		p->err = "illegal percent-encoding";
170 		return 0;
171 	}
172 
173 	sscanf(p->uri+1, "%2hhx", p->uri);
174 	memmove(p->uri+1, p->uri+3, strlen(p->uri+3)+1);
175 	if (*p->uri == '\0') {
176 		p->err = "illegal percent-encoding";
177 		return 0;
178 	}
179 
180 	return 1;
181 }
182 
183 /* ALPHA *( ALPHA / DIGIT / "+" / "-" / "." ) "://" */
184 static int
185 parse_scheme(struct parser *p)
186 {
187 	p->parsed->schema = p->uri;
188 
189 	if (!isalpha(*p->uri)) {
190 		p->err = "illegal character in scheme";
191 		return 0;
192 	}
193 
194 	p->uri++;
195 	while (isalnum(*p->uri)
196 	    || *p->uri == '+'
197 	    || *p->uri == '-'
198 	    || *p->uri == '.')
199 		p->uri++;
200 
201 	if (*p->uri != ':') {
202 		p->err = "illegal character in scheme";
203 		return 0;
204 	}
205 
206 	*p->uri = '\0';
207 	if (*(++p->uri) != '/' || *(++p->uri) != '/') {
208 		p->err = "invalid marker after scheme";
209 		return 0;
210 	}
211 
212 	p->uri++;
213 	return 1;
214 }
215 
216 /* *DIGIT */
217 static int
218 parse_port(struct parser *p)
219 {
220 	uint32_t i = 0;
221 
222 	p->parsed->port = p->uri;
223 
224 	for (; isdigit(*p->uri); p->uri++) {
225 		i = i * 10 + *p->uri - '0';
226 		if (i > UINT16_MAX) {
227 			p->err = "port number too large";
228 			return 0;
229 		}
230 	}
231 
232 	if (*p->uri != '/' && *p->uri != '\0') {
233 		p->err = "illegal character in port number";
234 		return 0;
235 	}
236 
237 	p->parsed->port_no = i;
238 
239 	if (*p->uri != '\0') {
240 		*p->uri = '\0';
241 		p->uri++;
242 	}
243 
244 	return 1;
245 }
246 
247 /* TODO: add support for ip-literal and ipv4addr ? */
248 /* *( unreserved / sub-delims / pct-encoded ) */
249 static int
250 parse_authority(struct parser *p)
251 {
252 	p->parsed->host = p->uri;
253 
254 	while (UNRESERVED(*p->uri)
255 	    || SUB_DELIMITERS(*p->uri)
256 	    || parse_pct_encoded(p))
257 		p->uri++;
258 
259 	if (p->err != NULL)
260 		return 0;
261 
262 	if (*p->uri == ':') {
263 		*p->uri = '\0';
264 		p->uri++;
265 		return parse_port(p);
266 	}
267 
268 	if (*p->uri == '/') {
269 		*p->uri = '\0';
270 		p->uri++;
271 		return 1;
272 	}
273 
274 	if (*p->uri == '\0')
275 		return 1;
276 
277 	p->err = "illegal character in authority section";
278 	return 0;
279 }
280 
281 /* Routine for path_clean.  Elide the pointed .. with the preceding
282  * element.  Return 0 if it's not possible.  incr is the length of
283  * the increment, 3 for ../ and 2 for .. */
284 static int
285 path_elide_dotdot(char *path, char *i, int incr)
286 {
287 	char *j;
288 
289 	if (i == path)
290 		return 0;
291 	for (j = i-2; j != path && *j != '/'; j--)
292                 /* noop */ ;
293 	if (*j == '/')
294 		j++;
295 	i += incr;
296 	memmove(j, i, strlen(i)+1);
297 	return 1;
298 }
299 
300 /*
301  * Use an algorithm similar to the one implemented in go' path.Clean:
302  *
303  * 1. Replace multiple slashes with a single slash
304  * 2. Eliminate each . path name element
305  * 3. Eliminate each inner .. along with the non-.. element that precedes it
306  * 4. Eliminate trailing .. if possible or error (go would only discard)
307  *
308  * Unlike path.Clean, this function return the empty string if the
309  * original path is equivalent to "/".
310  */
311 static int
312 path_clean(char *path)
313 {
314 	char *i;
315 
316 	/* 1. replace multiple slashes with a single one */
317 	for (i = path; *i; ++i) {
318 		if (*i == '/' && *(i+1) == '/') {
319 			memmove(i, i+1, strlen(i)); /* move also the \0 */
320 			i--;
321 		}
322 	}
323 
324 	/* 2. eliminate each . path name element */
325 	for (i = path; *i; ++i) {
326 		if ((i == path || *i == '/') && *(i+1) == '.' &&
327 		    *(i+2) == '/') {
328 			/* move also the \0 */
329 			memmove(i, i+2, strlen(i)-1);
330 			i--;
331 		}
332 	}
333 	if (!strcmp(path, ".") || !strcmp(path, "/.")) {
334 		*path = '\0';
335 		return 1;
336 	}
337 
338 	/* 3. eliminate each inner .. along with the preceding non-.. */
339 	for (i = strstr(path, "../"); i != NULL; i = strstr(path, ".."))
340 		if (!path_elide_dotdot(path, i, 3))
341 			return 0;
342 
343 	/* 4. eliminate trailing ..*/
344 	if ((i = strstr(path, "..")) != NULL)
345 		if (!path_elide_dotdot(path, i, 2))
346 			return 0;
347 
348 	return 1;
349 }
350 
351 static int
352 parse_query(struct parser *p)
353 {
354 	p->parsed->query = p->uri;
355 	if (*p->uri == '\0')
356 		return 1;
357 
358 	while (UNRESERVED(*p->uri)
359 	    || SUB_DELIMITERS(*p->uri)
360 	    || *p->uri == '/'
361 	    || *p->uri == '?'
362 	    || parse_pct_encoded(p)
363 	    || valid_multibyte_utf8(p))
364 		p->uri++;
365 
366 	if (p->err != NULL)
367 		return 0;
368 
369 	if (*p->uri != '\0' && *p->uri != '#') {
370 		p->err = "illegal character in query";
371 		return 0;
372 	}
373 
374 	if (*p->uri != '\0') {
375 		*p->uri = '\0';
376 		p->uri++;
377 	}
378 
379 	return 1;
380 }
381 
382 /* don't even bother */
383 static int
384 parse_fragment(struct parser *p)
385 {
386 	p->parsed->fragment = p->uri;
387 	return 1;
388 }
389 
390 /* XXX: is it too broad? */
391 /* *(pchar / "/") */
392 static int
393 parse_path(struct parser *p)
394 {
395 	char c;
396 
397 	p->parsed->path = p->uri;
398 	if (*p->uri == '\0') {
399 		p->parsed->query = p->parsed->fragment = p->uri;
400 		return 1;
401 	}
402 
403 	while (UNRESERVED(*p->uri)
404 	    || SUB_DELIMITERS(*p->uri)
405 	    || *p->uri == '/'
406 	    || parse_pct_encoded(p)
407 	    || valid_multibyte_utf8(p))
408 		p->uri++;
409 
410 	if (p->err != NULL)
411 		return 0;
412 
413 	if (*p->uri != '\0' && *p->uri != '?' && *p->uri != '#') {
414 		p->err = "illegal character in path";
415 		return 0;
416 	}
417 
418 	if (*p->uri != '\0') {
419 		c = *p->uri;
420 		*p->uri = '\0';
421 		p->uri++;
422 
423 		if (c == '#') {
424 			if (!parse_fragment(p))
425 				return 0;
426 		} else
427 			if (!parse_query(p) || !parse_fragment(p))
428 				return 0;
429 	}
430 
431 	if (!path_clean(p->parsed->path)) {
432 		p->err = "illegal path";
433 		return 0;
434 	}
435 
436 	return 1;
437 }
438 
439 int
440 parse_uri(char *uri, struct uri *ret, const char **err_ret)
441 {
442 	char *end;
443 	struct parser p = {uri, ret, NULL};
444 
445 	bzero(ret, sizeof(*ret));
446 
447 	/* initialize optional stuff to the empty string */
448 	end = uri + strlen(uri);
449 	p.parsed->port = end;
450 	p.parsed->path = end;
451 	p.parsed->query = end;
452 	p.parsed->fragment = end;
453 
454 	if (!parse_scheme(&p) || !parse_authority(&p) || !parse_path(&p)) {
455 		*err_ret = p.err;
456 		return 0;
457 	}
458 
459 	*err_ret = NULL;
460 	return 1;
461 }
462 
463 int
464 trim_req_uri(char *uri)
465 {
466 	char *i;
467 
468 	if ((i = strstr(uri, "\r\n")) == NULL)
469 		return 0;
470 	*i = '\0';
471 	return 1;
472 }