/src/gdal/curl/lib/urlapi.c
Line | Count | Source (jump to first uncovered line) |
1 | | /*************************************************************************** |
2 | | * _ _ ____ _ |
3 | | * Project ___| | | | _ \| | |
4 | | * / __| | | | |_) | | |
5 | | * | (__| |_| | _ <| |___ |
6 | | * \___|\___/|_| \_\_____| |
7 | | * |
8 | | * Copyright (C) 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 | | * SPDX-License-Identifier: curl |
22 | | * |
23 | | ***************************************************************************/ |
24 | | |
25 | | #include "curl_setup.h" |
26 | | |
27 | | #include "urldata.h" |
28 | | #include "urlapi-int.h" |
29 | | #include "strcase.h" |
30 | | #include "url.h" |
31 | | #include "escape.h" |
32 | | #include "curl_ctype.h" |
33 | | #include "curlx/inet_pton.h" |
34 | | #include "inet_ntop.h" |
35 | | #include "strdup.h" |
36 | | #include "idn.h" |
37 | | #include "curlx/strparse.h" |
38 | | #include "curl_memrchr.h" |
39 | | |
40 | | /* The last 3 #include files should be in this order */ |
41 | | #include "curl_printf.h" |
42 | | #include "curl_memory.h" |
43 | | #include "memdebug.h" |
44 | | |
45 | | /* MS-DOS/Windows style drive prefix, eg c: in c:foo */ |
46 | | #define STARTS_WITH_DRIVE_PREFIX(str) \ |
47 | | ((('a' <= str[0] && str[0] <= 'z') || \ |
48 | | ('A' <= str[0] && str[0] <= 'Z')) && \ |
49 | | (str[1] == ':')) |
50 | | |
51 | | /* MS-DOS/Windows style drive prefix, optionally with |
52 | | * a '|' instead of ':', followed by a slash or NUL */ |
53 | | #define STARTS_WITH_URL_DRIVE_PREFIX(str) \ |
54 | 0 | ((('a' <= (str)[0] && (str)[0] <= 'z') || \ |
55 | 0 | ('A' <= (str)[0] && (str)[0] <= 'Z')) && \ |
56 | 0 | ((str)[1] == ':' || (str)[1] == '|') && \ |
57 | 0 | ((str)[2] == '/' || (str)[2] == '\\' || (str)[2] == 0)) |
58 | | |
59 | | /* scheme is not URL encoded, the longest libcurl supported ones are... */ |
60 | 17.9k | #define MAX_SCHEME_LEN 40 |
61 | | |
62 | | /* |
63 | | * If USE_IPV6 is disabled, we still want to parse IPv6 addresses, so make |
64 | | * sure we have _some_ value for AF_INET6 without polluting our fake value |
65 | | * everywhere. |
66 | | */ |
67 | | #if !defined(USE_IPV6) && !defined(AF_INET6) |
68 | | #define AF_INET6 (AF_INET + 1) |
69 | | #endif |
70 | | |
71 | | /* Internal representation of CURLU. Point to URL-encoded strings. */ |
72 | | struct Curl_URL { |
73 | | char *scheme; |
74 | | char *user; |
75 | | char *password; |
76 | | char *options; /* IMAP only? */ |
77 | | char *host; |
78 | | char *zoneid; /* for numerical IPv6 addresses */ |
79 | | char *port; |
80 | | char *path; |
81 | | char *query; |
82 | | char *fragment; |
83 | | unsigned short portnum; /* the numerical version (if 'port' is set) */ |
84 | | BIT(query_present); /* to support blank */ |
85 | | BIT(fragment_present); /* to support blank */ |
86 | | BIT(guessed_scheme); /* when a URL without scheme is parsed */ |
87 | | }; |
88 | | |
89 | 0 | #define DEFAULT_SCHEME "https" |
90 | | |
91 | | static CURLUcode parseurl_and_replace(const char *url, CURLU *u, |
92 | | unsigned int flags); |
93 | | |
94 | | static void free_urlhandle(struct Curl_URL *u) |
95 | 3.69k | { |
96 | 3.69k | free(u->scheme); |
97 | 3.69k | free(u->user); |
98 | 3.69k | free(u->password); |
99 | 3.69k | free(u->options); |
100 | 3.69k | free(u->host); |
101 | 3.69k | free(u->zoneid); |
102 | 3.69k | free(u->port); |
103 | 3.69k | free(u->path); |
104 | 3.69k | free(u->query); |
105 | 3.69k | free(u->fragment); |
106 | 3.69k | } |
107 | | |
108 | | /* |
109 | | * Find the separator at the end of the hostname, or the '?' in cases like |
110 | | * http://www.example.com?id=2380 |
111 | | */ |
112 | | static const char *find_host_sep(const char *url) |
113 | 0 | { |
114 | | /* Find the start of the hostname */ |
115 | 0 | const char *sep = strstr(url, "//"); |
116 | 0 | if(!sep) |
117 | 0 | sep = url; |
118 | 0 | else |
119 | 0 | sep += 2; |
120 | | |
121 | | /* Find first / or ? */ |
122 | 0 | while(*sep && *sep != '/' && *sep != '?') |
123 | 0 | sep++; |
124 | |
|
125 | 0 | return sep; |
126 | 0 | } |
127 | | |
128 | | /* convert CURLcode to CURLUcode */ |
129 | 0 | #define cc2cu(x) ((x) == CURLE_TOO_LARGE ? CURLUE_TOO_LARGE : \ |
130 | 0 | CURLUE_OUT_OF_MEMORY) |
131 | | |
132 | | /* urlencode_str() writes data into an output dynbuf and URL-encodes the |
133 | | * spaces in the source URL accordingly. |
134 | | * |
135 | | * URL encoding should be skipped for hostnames, otherwise IDN resolution |
136 | | * will fail. |
137 | | */ |
138 | | static CURLUcode urlencode_str(struct dynbuf *o, const char *url, |
139 | | size_t len, bool relative, |
140 | | bool query) |
141 | 1.44k | { |
142 | | /* we must add this with whitespace-replacing */ |
143 | 1.44k | bool left = !query; |
144 | 1.44k | const unsigned char *iptr; |
145 | 1.44k | const unsigned char *host_sep = (const unsigned char *) url; |
146 | 1.44k | CURLcode result = CURLE_OK; |
147 | | |
148 | 1.44k | if(!relative) { |
149 | 0 | size_t n; |
150 | 0 | host_sep = (const unsigned char *) find_host_sep(url); |
151 | | |
152 | | /* output the first piece as-is */ |
153 | 0 | n = (const char *)host_sep - url; |
154 | 0 | result = curlx_dyn_addn(o, url, n); |
155 | 0 | len -= n; |
156 | 0 | } |
157 | | |
158 | 39.8k | for(iptr = host_sep; len && !result; iptr++, len--) { |
159 | 38.3k | if(*iptr == ' ') { |
160 | 0 | if(left) |
161 | 0 | result = curlx_dyn_addn(o, "%20", 3); |
162 | 0 | else |
163 | 0 | result = curlx_dyn_addn(o, "+", 1); |
164 | 0 | } |
165 | 38.3k | else if((*iptr < ' ') || (*iptr >= 0x7f)) { |
166 | 5.66k | unsigned char out[3]={'%'}; |
167 | 5.66k | Curl_hexbyte(&out[1], *iptr, TRUE); |
168 | 5.66k | result = curlx_dyn_addn(o, out, 3); |
169 | 5.66k | } |
170 | 32.7k | else { |
171 | 32.7k | result = curlx_dyn_addn(o, iptr, 1); |
172 | 32.7k | if(*iptr == '?') |
173 | 0 | left = FALSE; |
174 | 32.7k | } |
175 | 38.3k | } |
176 | | |
177 | 1.44k | if(result) |
178 | 0 | return cc2cu(result); |
179 | 1.44k | return CURLUE_OK; |
180 | 1.44k | } |
181 | | |
182 | | /* |
183 | | * Returns the length of the scheme if the given URL is absolute (as opposed |
184 | | * to relative). Stores the scheme in the buffer if TRUE and 'buf' is |
185 | | * non-NULL. The buflen must be larger than MAX_SCHEME_LEN if buf is set. |
186 | | * |
187 | | * If 'guess_scheme' is TRUE, it means the URL might be provided without |
188 | | * scheme. |
189 | | */ |
190 | | size_t Curl_is_absolute_url(const char *url, char *buf, size_t buflen, |
191 | | bool guess_scheme) |
192 | 3.62k | { |
193 | 3.62k | size_t i = 0; |
194 | 3.62k | DEBUGASSERT(!buf || (buflen > MAX_SCHEME_LEN)); |
195 | 3.62k | (void)buflen; /* only used in debug-builds */ |
196 | 3.62k | if(buf) |
197 | 1.77k | buf[0] = 0; /* always leave a defined value in buf */ |
198 | | #ifdef _WIN32 |
199 | | if(guess_scheme && STARTS_WITH_DRIVE_PREFIX(url)) |
200 | | return 0; |
201 | | #endif |
202 | 3.62k | if(ISALPHA(url[0])) |
203 | 17.9k | for(i = 1; i < MAX_SCHEME_LEN; ++i) { |
204 | 17.9k | char s = url[i]; |
205 | 17.9k | if(s && (ISALNUM(s) || (s == '+') || (s == '-') || (s == '.') )) { |
206 | | /* RFC 3986 3.1 explains: |
207 | | scheme = ALPHA *( ALPHA / DIGIT / "+" / "-" / "." ) |
208 | | */ |
209 | 14.4k | } |
210 | 3.53k | else { |
211 | 3.53k | break; |
212 | 3.53k | } |
213 | 17.9k | } |
214 | 3.62k | if(i && (url[i] == ':') && ((url[i + 1] == '/') || !guess_scheme)) { |
215 | | /* If this does not guess scheme, the scheme always ends with the colon so |
216 | | that this also detects data: URLs etc. In guessing mode, data: could |
217 | | be the hostname "data" with a specified port number. */ |
218 | | |
219 | | /* the length of the scheme is the name part only */ |
220 | 3.48k | size_t len = i; |
221 | 3.48k | if(buf) { |
222 | 1.74k | Curl_strntolower(buf, url, i); |
223 | 1.74k | buf[i] = 0; |
224 | 1.74k | } |
225 | 3.48k | return len; |
226 | 3.48k | } |
227 | 140 | return 0; |
228 | 3.62k | } |
229 | | |
230 | | /* |
231 | | * Concatenate a relative URL onto a base URL making it absolute. |
232 | | */ |
233 | | static CURLUcode redirect_url(const char *base, const char *relurl, |
234 | | CURLU *u, unsigned int flags) |
235 | 0 | { |
236 | 0 | struct dynbuf urlbuf; |
237 | 0 | bool host_changed = FALSE; |
238 | 0 | const char *useurl = relurl; |
239 | 0 | const char *cutoff = NULL; |
240 | 0 | size_t prelen; |
241 | 0 | CURLUcode uc; |
242 | | |
243 | | /* protsep points to the start of the hostname, after [scheme]:// */ |
244 | 0 | const char *protsep = base + strlen(u->scheme) + 3; |
245 | 0 | DEBUGASSERT(base && relurl && u); /* all set here */ |
246 | 0 | if(!base) |
247 | 0 | return CURLUE_MALFORMED_INPUT; /* should never happen */ |
248 | | |
249 | | /* handle different relative URL types */ |
250 | 0 | switch(relurl[0]) { |
251 | 0 | case '/': |
252 | 0 | if(relurl[1] == '/') { |
253 | | /* protocol-relative URL: //example.com/path */ |
254 | 0 | cutoff = protsep; |
255 | 0 | useurl = &relurl[2]; |
256 | 0 | host_changed = TRUE; |
257 | 0 | } |
258 | 0 | else |
259 | | /* absolute /path */ |
260 | 0 | cutoff = strchr(protsep, '/'); |
261 | 0 | break; |
262 | | |
263 | 0 | case '#': |
264 | | /* fragment-only change */ |
265 | 0 | if(u->fragment) |
266 | 0 | cutoff = strchr(protsep, '#'); |
267 | 0 | break; |
268 | | |
269 | 0 | default: |
270 | | /* path or query-only change */ |
271 | 0 | if(u->query && u->query[0]) |
272 | | /* remove existing query */ |
273 | 0 | cutoff = strchr(protsep, '?'); |
274 | 0 | else if(u->fragment && u->fragment[0]) |
275 | | /* Remove existing fragment */ |
276 | 0 | cutoff = strchr(protsep, '#'); |
277 | |
|
278 | 0 | if(relurl[0] != '?') { |
279 | | /* append a relative path after the last slash */ |
280 | 0 | cutoff = memrchr(protsep, '/', |
281 | 0 | cutoff ? (size_t)(cutoff - protsep) : strlen(protsep)); |
282 | 0 | if(cutoff) |
283 | 0 | cutoff++; /* truncate after last slash */ |
284 | 0 | } |
285 | 0 | break; |
286 | 0 | } |
287 | | |
288 | 0 | prelen = cutoff ? (size_t)(cutoff - base) : strlen(base); |
289 | | |
290 | | /* build new URL */ |
291 | 0 | curlx_dyn_init(&urlbuf, CURL_MAX_INPUT_LENGTH); |
292 | |
|
293 | 0 | if(!curlx_dyn_addn(&urlbuf, base, prelen) && |
294 | 0 | !urlencode_str(&urlbuf, useurl, strlen(useurl), !host_changed, FALSE)) { |
295 | 0 | uc = parseurl_and_replace(curlx_dyn_ptr(&urlbuf), u, |
296 | 0 | flags & ~CURLU_PATH_AS_IS); |
297 | 0 | } |
298 | 0 | else |
299 | 0 | uc = CURLUE_OUT_OF_MEMORY; |
300 | |
|
301 | 0 | curlx_dyn_free(&urlbuf); |
302 | 0 | return uc; |
303 | 0 | } |
304 | | |
305 | | /* scan for byte values <= 31, 127 and sometimes space */ |
306 | | CURLUcode Curl_junkscan(const char *url, size_t *urllen, bool allowspace) |
307 | 1.84k | { |
308 | 1.84k | size_t n = strlen(url); |
309 | 1.84k | size_t i; |
310 | 1.84k | unsigned char control; |
311 | 1.84k | const unsigned char *p = (const unsigned char *)url; |
312 | 1.84k | if(n > CURL_MAX_INPUT_LENGTH) |
313 | 0 | return CURLUE_MALFORMED_INPUT; |
314 | | |
315 | 1.84k | control = allowspace ? 0x1f : 0x20; |
316 | 254k | for(i = 0; i < n; i++) { |
317 | 252k | if(p[i] <= control || p[i] == 127) |
318 | 75 | return CURLUE_MALFORMED_INPUT; |
319 | 252k | } |
320 | 1.77k | *urllen = n; |
321 | 1.77k | return CURLUE_OK; |
322 | 1.84k | } |
323 | | |
324 | | /* |
325 | | * parse_hostname_login() |
326 | | * |
327 | | * Parse the login details (username, password and options) from the URL and |
328 | | * strip them out of the hostname |
329 | | * |
330 | | */ |
331 | | static CURLUcode parse_hostname_login(struct Curl_URL *u, |
332 | | const char *login, |
333 | | size_t len, |
334 | | unsigned int flags, |
335 | | size_t *offset) /* to the hostname */ |
336 | 1.77k | { |
337 | 1.77k | CURLUcode result = CURLUE_OK; |
338 | 1.77k | CURLcode ccode; |
339 | 1.77k | char *userp = NULL; |
340 | 1.77k | char *passwdp = NULL; |
341 | 1.77k | char *optionsp = NULL; |
342 | 1.77k | const struct Curl_handler *h = NULL; |
343 | | |
344 | | /* At this point, we assume all the other special cases have been taken |
345 | | * care of, so the host is at most |
346 | | * |
347 | | * [user[:password][;options]]@]hostname |
348 | | * |
349 | | * We need somewhere to put the embedded details, so do that first. |
350 | | */ |
351 | 1.77k | char *ptr; |
352 | | |
353 | 1.77k | DEBUGASSERT(login); |
354 | | |
355 | 1.77k | *offset = 0; |
356 | 1.77k | ptr = memchr(login, '@', len); |
357 | 1.77k | if(!ptr) |
358 | 293 | goto out; |
359 | | |
360 | | /* We will now try to extract the |
361 | | * possible login information in a string like: |
362 | | * ftp://user:password@ftp.site.example:8021/README */ |
363 | 1.48k | ptr++; |
364 | | |
365 | | /* if this is a known scheme, get some details */ |
366 | 1.48k | if(u->scheme) |
367 | 1.47k | h = Curl_get_scheme_handler(u->scheme); |
368 | | |
369 | | /* We could use the login information in the URL so extract it. Only parse |
370 | | options if the handler says we should. Note that 'h' might be NULL! */ |
371 | 1.48k | ccode = Curl_parse_login_details(login, ptr - login - 1, |
372 | 1.48k | &userp, &passwdp, |
373 | 1.48k | (h && (h->flags & PROTOPT_URLOPTIONS)) ? |
374 | 0 | &optionsp : NULL); |
375 | 1.48k | if(ccode) { |
376 | 0 | result = CURLUE_BAD_LOGIN; |
377 | 0 | goto out; |
378 | 0 | } |
379 | | |
380 | 1.48k | if(userp) { |
381 | 1.48k | if(flags & CURLU_DISALLOW_USER) { |
382 | | /* Option DISALLOW_USER is set and URL contains username. */ |
383 | 0 | result = CURLUE_USER_NOT_ALLOWED; |
384 | 0 | goto out; |
385 | 0 | } |
386 | 1.48k | free(u->user); |
387 | 1.48k | u->user = userp; |
388 | 1.48k | } |
389 | | |
390 | 1.48k | if(passwdp) { |
391 | 1.34k | free(u->password); |
392 | 1.34k | u->password = passwdp; |
393 | 1.34k | } |
394 | | |
395 | 1.48k | if(optionsp) { |
396 | 0 | free(u->options); |
397 | 0 | u->options = optionsp; |
398 | 0 | } |
399 | | |
400 | | /* the hostname starts at this offset */ |
401 | 1.48k | *offset = ptr - login; |
402 | 1.48k | return CURLUE_OK; |
403 | | |
404 | 293 | out: |
405 | | |
406 | 293 | free(userp); |
407 | 293 | free(passwdp); |
408 | 293 | free(optionsp); |
409 | 293 | u->user = NULL; |
410 | 293 | u->password = NULL; |
411 | 293 | u->options = NULL; |
412 | | |
413 | 293 | return result; |
414 | 1.48k | } |
415 | | |
416 | | UNITTEST CURLUcode Curl_parse_port(struct Curl_URL *u, struct dynbuf *host, |
417 | | bool has_scheme) |
418 | 1.77k | { |
419 | 1.77k | const char *portptr; |
420 | 1.77k | char *hostname = curlx_dyn_ptr(host); |
421 | | /* |
422 | | * Find the end of an IPv6 address on the ']' ending bracket. |
423 | | */ |
424 | 1.77k | if(hostname[0] == '[') { |
425 | 0 | portptr = strchr(hostname, ']'); |
426 | 0 | if(!portptr) |
427 | 0 | return CURLUE_BAD_IPV6; |
428 | 0 | portptr++; |
429 | | /* this is a RFC2732-style specified IP-address */ |
430 | 0 | if(*portptr) { |
431 | 0 | if(*portptr != ':') |
432 | 0 | return CURLUE_BAD_PORT_NUMBER; |
433 | 0 | } |
434 | 0 | else |
435 | 0 | portptr = NULL; |
436 | 0 | } |
437 | 1.77k | else |
438 | 1.77k | portptr = strchr(hostname, ':'); |
439 | | |
440 | 1.77k | if(portptr) { |
441 | 22 | curl_off_t port; |
442 | 22 | size_t keep = portptr - hostname; |
443 | | |
444 | | /* Browser behavior adaptation. If there is a colon with no digits after, |
445 | | just cut off the name there which makes us ignore the colon and just |
446 | | use the default port. Firefox, Chrome and Safari all do that. |
447 | | |
448 | | Do not do it if the URL has no scheme, to make something that looks like |
449 | | a scheme not work! |
450 | | */ |
451 | 22 | curlx_dyn_setlen(host, keep); |
452 | 22 | portptr++; |
453 | 22 | if(!*portptr) |
454 | 6 | return has_scheme ? CURLUE_OK : CURLUE_BAD_PORT_NUMBER; |
455 | | |
456 | 16 | if(curlx_str_number(&portptr, &port, 0xffff) || *portptr) |
457 | 15 | return CURLUE_BAD_PORT_NUMBER; |
458 | | |
459 | 1 | u->portnum = (unsigned short) port; |
460 | | /* generate a new port number string to get rid of leading zeroes etc */ |
461 | 1 | free(u->port); |
462 | 1 | u->port = aprintf("%" CURL_FORMAT_CURL_OFF_T, port); |
463 | 1 | if(!u->port) |
464 | 0 | return CURLUE_OUT_OF_MEMORY; |
465 | 1 | } |
466 | | |
467 | 1.75k | return CURLUE_OK; |
468 | 1.77k | } |
469 | | |
470 | | /* this assumes 'hostname' now starts with [ */ |
471 | | static CURLUcode ipv6_parse(struct Curl_URL *u, char *hostname, |
472 | | size_t hlen) /* length of hostname */ |
473 | 0 | { |
474 | 0 | size_t len; |
475 | 0 | DEBUGASSERT(*hostname == '['); |
476 | 0 | if(hlen < 4) /* '[::]' is the shortest possible valid string */ |
477 | 0 | return CURLUE_BAD_IPV6; |
478 | 0 | hostname++; |
479 | 0 | hlen -= 2; |
480 | | |
481 | | /* only valid IPv6 letters are ok */ |
482 | 0 | len = strspn(hostname, "0123456789abcdefABCDEF:."); |
483 | |
|
484 | 0 | if(hlen != len) { |
485 | 0 | hlen = len; |
486 | 0 | if(hostname[len] == '%') { |
487 | | /* this could now be '%[zone id]' */ |
488 | 0 | char zoneid[16]; |
489 | 0 | int i = 0; |
490 | 0 | char *h = &hostname[len + 1]; |
491 | | /* pass '25' if present and is a URL encoded percent sign */ |
492 | 0 | if(!strncmp(h, "25", 2) && h[2] && (h[2] != ']')) |
493 | 0 | h += 2; |
494 | 0 | while(*h && (*h != ']') && (i < 15)) |
495 | 0 | zoneid[i++] = *h++; |
496 | 0 | if(!i || (']' != *h)) |
497 | 0 | return CURLUE_BAD_IPV6; |
498 | 0 | zoneid[i] = 0; |
499 | 0 | u->zoneid = strdup(zoneid); |
500 | 0 | if(!u->zoneid) |
501 | 0 | return CURLUE_OUT_OF_MEMORY; |
502 | 0 | hostname[len] = ']'; /* insert end bracket */ |
503 | 0 | hostname[len + 1] = 0; /* terminate the hostname */ |
504 | 0 | } |
505 | 0 | else |
506 | 0 | return CURLUE_BAD_IPV6; |
507 | | /* hostname is fine */ |
508 | 0 | } |
509 | | |
510 | | /* Normalize the IPv6 address */ |
511 | 0 | { |
512 | 0 | char dest[16]; /* fits a binary IPv6 address */ |
513 | 0 | hostname[hlen] = 0; /* end the address there */ |
514 | 0 | if(1 != curlx_inet_pton(AF_INET6, hostname, dest)) |
515 | 0 | return CURLUE_BAD_IPV6; |
516 | 0 | if(Curl_inet_ntop(AF_INET6, dest, hostname, hlen)) { |
517 | 0 | hlen = strlen(hostname); /* might be shorter now */ |
518 | 0 | hostname[hlen + 1] = 0; |
519 | 0 | } |
520 | 0 | hostname[hlen] = ']'; /* restore ending bracket */ |
521 | 0 | } |
522 | 0 | return CURLUE_OK; |
523 | 0 | } |
524 | | |
525 | | static CURLUcode hostname_check(struct Curl_URL *u, char *hostname, |
526 | | size_t hlen) /* length of hostname */ |
527 | 1.74k | { |
528 | 1.74k | size_t len; |
529 | 1.74k | DEBUGASSERT(hostname); |
530 | | |
531 | 1.74k | if(!hlen) |
532 | 0 | return CURLUE_NO_HOST; |
533 | 1.74k | else if(hostname[0] == '[') |
534 | 0 | return ipv6_parse(u, hostname, hlen); |
535 | 1.74k | else { |
536 | | /* letters from the second string are not ok */ |
537 | 1.74k | len = strcspn(hostname, " \r\n\t/:#?!@{}[]\\$\'\"^`*<>=;,+&()%"); |
538 | 1.74k | if(hlen != len) |
539 | | /* hostname with bad content */ |
540 | 10 | return CURLUE_BAD_HOSTNAME; |
541 | 1.74k | } |
542 | 1.73k | return CURLUE_OK; |
543 | 1.74k | } |
544 | | |
545 | | /* |
546 | | * Handle partial IPv4 numerical addresses and different bases, like |
547 | | * '16843009', '0x7f', '0x7f.1' '0177.1.1.1' etc. |
548 | | * |
549 | | * If the given input string is syntactically wrong IPv4 or any part for |
550 | | * example is too big, this function returns HOST_NAME. |
551 | | * |
552 | | * Output the "normalized" version of that input string in plain quad decimal |
553 | | * integers. |
554 | | * |
555 | | * Returns the host type. |
556 | | */ |
557 | | |
558 | 0 | #define HOST_ERROR -1 /* out of memory */ |
559 | | |
560 | 3.49k | #define HOST_NAME 1 |
561 | 10 | #define HOST_IPV4 2 |
562 | 0 | #define HOST_IPV6 3 |
563 | | |
564 | | static int ipv4_normalize(struct dynbuf *host) |
565 | 1.75k | { |
566 | 1.75k | bool done = FALSE; |
567 | 1.75k | int n = 0; |
568 | 1.75k | const char *c = curlx_dyn_ptr(host); |
569 | 1.75k | unsigned int parts[4] = {0, 0, 0, 0}; |
570 | 1.75k | CURLcode result = CURLE_OK; |
571 | | |
572 | 1.75k | if(*c == '[') |
573 | 0 | return HOST_IPV6; |
574 | | |
575 | 1.75k | while(!done) { |
576 | 1.75k | int rc; |
577 | 1.75k | curl_off_t l; |
578 | 1.75k | if(*c == '0') { |
579 | 3 | if(c[1] == 'x') { |
580 | 0 | c += 2; /* skip the prefix */ |
581 | 0 | rc = curlx_str_hex(&c, &l, UINT_MAX); |
582 | 0 | } |
583 | 3 | else |
584 | 3 | rc = curlx_str_octal(&c, &l, UINT_MAX); |
585 | 3 | } |
586 | 1.74k | else |
587 | 1.74k | rc = curlx_str_number(&c, &l, UINT_MAX); |
588 | | |
589 | 1.75k | if(rc) |
590 | 1.68k | return HOST_NAME; |
591 | | |
592 | 68 | parts[n] = (unsigned int)l; |
593 | | |
594 | 68 | switch(*c) { |
595 | 0 | case '.': |
596 | 0 | if(n == 3) |
597 | 0 | return HOST_NAME; |
598 | 0 | n++; |
599 | 0 | c++; |
600 | 0 | break; |
601 | | |
602 | 5 | case '\0': |
603 | 5 | done = TRUE; |
604 | 5 | break; |
605 | | |
606 | 63 | default: |
607 | 63 | return HOST_NAME; |
608 | 68 | } |
609 | 68 | } |
610 | | |
611 | 5 | switch(n) { |
612 | 5 | case 0: /* a -- 32 bits */ |
613 | 5 | curlx_dyn_reset(host); |
614 | | |
615 | 5 | result = curlx_dyn_addf(host, "%u.%u.%u.%u", |
616 | 5 | (parts[0] >> 24), |
617 | 5 | ((parts[0] >> 16) & 0xff), |
618 | 5 | ((parts[0] >> 8) & 0xff), |
619 | 5 | (parts[0] & 0xff)); |
620 | 5 | break; |
621 | 0 | case 1: /* a.b -- 8.24 bits */ |
622 | 0 | if((parts[0] > 0xff) || (parts[1] > 0xffffff)) |
623 | 0 | return HOST_NAME; |
624 | 0 | curlx_dyn_reset(host); |
625 | 0 | result = curlx_dyn_addf(host, "%u.%u.%u.%u", |
626 | 0 | (parts[0]), |
627 | 0 | ((parts[1] >> 16) & 0xff), |
628 | 0 | ((parts[1] >> 8) & 0xff), |
629 | 0 | (parts[1] & 0xff)); |
630 | 0 | break; |
631 | 0 | case 2: /* a.b.c -- 8.8.16 bits */ |
632 | 0 | if((parts[0] > 0xff) || (parts[1] > 0xff) || (parts[2] > 0xffff)) |
633 | 0 | return HOST_NAME; |
634 | 0 | curlx_dyn_reset(host); |
635 | 0 | result = curlx_dyn_addf(host, "%u.%u.%u.%u", |
636 | 0 | (parts[0]), |
637 | 0 | (parts[1]), |
638 | 0 | ((parts[2] >> 8) & 0xff), |
639 | 0 | (parts[2] & 0xff)); |
640 | 0 | break; |
641 | 0 | case 3: /* a.b.c.d -- 8.8.8.8 bits */ |
642 | 0 | if((parts[0] > 0xff) || (parts[1] > 0xff) || (parts[2] > 0xff) || |
643 | 0 | (parts[3] > 0xff)) |
644 | 0 | return HOST_NAME; |
645 | 0 | curlx_dyn_reset(host); |
646 | 0 | result = curlx_dyn_addf(host, "%u.%u.%u.%u", |
647 | 0 | (parts[0]), |
648 | 0 | (parts[1]), |
649 | 0 | (parts[2]), |
650 | 0 | (parts[3])); |
651 | 0 | break; |
652 | 5 | } |
653 | 5 | if(result) |
654 | 0 | return HOST_ERROR; |
655 | 5 | return HOST_IPV4; |
656 | 5 | } |
657 | | |
658 | | /* if necessary, replace the host content with a URL decoded version */ |
659 | | static CURLUcode urldecode_host(struct dynbuf *host) |
660 | 1.74k | { |
661 | 1.74k | char *per = NULL; |
662 | 1.74k | const char *hostname = curlx_dyn_ptr(host); |
663 | 1.74k | per = strchr(hostname, '%'); |
664 | 1.74k | if(!per) |
665 | | /* nothing to decode */ |
666 | 1.74k | return CURLUE_OK; |
667 | 4 | else { |
668 | | /* encoded */ |
669 | 4 | size_t dlen; |
670 | 4 | char *decoded; |
671 | 4 | CURLcode result = Curl_urldecode(hostname, 0, &decoded, &dlen, |
672 | 4 | REJECT_CTRL); |
673 | 4 | if(result) |
674 | 0 | return CURLUE_BAD_HOSTNAME; |
675 | 4 | curlx_dyn_reset(host); |
676 | 4 | result = curlx_dyn_addn(host, decoded, dlen); |
677 | 4 | free(decoded); |
678 | 4 | if(result) |
679 | 0 | return cc2cu(result); |
680 | 4 | } |
681 | | |
682 | 4 | return CURLUE_OK; |
683 | 1.74k | } |
684 | | |
685 | | static CURLUcode parse_authority(struct Curl_URL *u, |
686 | | const char *auth, size_t authlen, |
687 | | unsigned int flags, |
688 | | struct dynbuf *host, |
689 | | bool has_scheme) |
690 | 1.77k | { |
691 | 1.77k | size_t offset; |
692 | 1.77k | CURLUcode uc; |
693 | 1.77k | CURLcode result; |
694 | | |
695 | | /* |
696 | | * Parse the login details and strip them out of the hostname. |
697 | | */ |
698 | 1.77k | uc = parse_hostname_login(u, auth, authlen, flags, &offset); |
699 | 1.77k | if(uc) |
700 | 0 | goto out; |
701 | | |
702 | 1.77k | result = curlx_dyn_addn(host, auth + offset, authlen - offset); |
703 | 1.77k | if(result) { |
704 | 0 | uc = cc2cu(result); |
705 | 0 | goto out; |
706 | 0 | } |
707 | | |
708 | 1.77k | uc = Curl_parse_port(u, host, has_scheme); |
709 | 1.77k | if(uc) |
710 | 21 | goto out; |
711 | | |
712 | 1.75k | if(!curlx_dyn_len(host)) |
713 | 2 | return CURLUE_NO_HOST; |
714 | | |
715 | 1.75k | switch(ipv4_normalize(host)) { |
716 | 5 | case HOST_IPV4: |
717 | 5 | break; |
718 | 0 | case HOST_IPV6: |
719 | 0 | uc = ipv6_parse(u, curlx_dyn_ptr(host), curlx_dyn_len(host)); |
720 | 0 | break; |
721 | 1.74k | case HOST_NAME: |
722 | 1.74k | uc = urldecode_host(host); |
723 | 1.74k | if(!uc) |
724 | 1.74k | uc = hostname_check(u, curlx_dyn_ptr(host), curlx_dyn_len(host)); |
725 | 1.74k | break; |
726 | 0 | case HOST_ERROR: |
727 | 0 | uc = CURLUE_OUT_OF_MEMORY; |
728 | 0 | break; |
729 | 0 | default: |
730 | 0 | uc = CURLUE_BAD_HOSTNAME; /* Bad IPv4 address even */ |
731 | 0 | break; |
732 | 1.75k | } |
733 | | |
734 | 1.77k | out: |
735 | 1.77k | return uc; |
736 | 1.75k | } |
737 | | |
738 | | /* used for HTTP/2 server push */ |
739 | | CURLUcode Curl_url_set_authority(CURLU *u, const char *authority) |
740 | 0 | { |
741 | 0 | CURLUcode result; |
742 | 0 | struct dynbuf host; |
743 | |
|
744 | 0 | DEBUGASSERT(authority); |
745 | 0 | curlx_dyn_init(&host, CURL_MAX_INPUT_LENGTH); |
746 | |
|
747 | 0 | result = parse_authority(u, authority, strlen(authority), |
748 | 0 | CURLU_DISALLOW_USER, &host, !!u->scheme); |
749 | 0 | if(result) |
750 | 0 | curlx_dyn_free(&host); |
751 | 0 | else { |
752 | 0 | free(u->host); |
753 | 0 | u->host = curlx_dyn_ptr(&host); |
754 | 0 | } |
755 | 0 | return result; |
756 | 0 | } |
757 | | |
758 | | /* |
759 | | * "Remove Dot Segments" |
760 | | * https://datatracker.ietf.org/doc/html/rfc3986#section-5.2.4 |
761 | | */ |
762 | | |
763 | | static bool is_dot(const char **str, size_t *clen) |
764 | 3.52k | { |
765 | 3.52k | const char *p = *str; |
766 | 3.52k | if(*p == '.') { |
767 | 155 | (*str)++; |
768 | 155 | (*clen)--; |
769 | 155 | return TRUE; |
770 | 155 | } |
771 | 3.36k | else if((*clen >= 3) && |
772 | 3.36k | (p[0] == '%') && (p[1] == '2') && ((p[2] | 0x20) == 'e')) { |
773 | 0 | *str += 3; |
774 | 0 | *clen -= 3; |
775 | 0 | return TRUE; |
776 | 0 | } |
777 | 3.36k | return FALSE; |
778 | 3.52k | } |
779 | | |
780 | 45.0k | #define ISSLASH(x) ((x) == '/') |
781 | | |
782 | | /* |
783 | | * dedotdotify() |
784 | | * @unittest: 1395 |
785 | | * |
786 | | * This function gets a null-terminated path with dot and dotdot sequences |
787 | | * passed in and strips them off according to the rules in RFC 3986 section |
788 | | * 5.2.4. |
789 | | * |
790 | | * The function handles a path. It should not contain the query nor fragment. |
791 | | * |
792 | | * RETURNS |
793 | | * |
794 | | * Zero for success and 'out' set to an allocated dedotdotified string. |
795 | | */ |
796 | | UNITTEST int dedotdotify(const char *input, size_t clen, char **outp); |
797 | | UNITTEST int dedotdotify(const char *input, size_t clen, char **outp) |
798 | 1.57k | { |
799 | 1.57k | struct dynbuf out; |
800 | 1.57k | CURLcode result = CURLE_OK; |
801 | | |
802 | 1.57k | *outp = NULL; |
803 | | /* the path always starts with a slash, and a slash has not dot */ |
804 | 1.57k | if(clen < 2) |
805 | 0 | return 0; |
806 | | |
807 | 1.57k | curlx_dyn_init(&out, clen + 1); |
808 | | |
809 | | /* A. If the input buffer begins with a prefix of "../" or "./", then |
810 | | remove that prefix from the input buffer; otherwise, */ |
811 | 1.57k | if(is_dot(&input, &clen)) { |
812 | 0 | const char *p = input; |
813 | 0 | size_t blen = clen; |
814 | |
|
815 | 0 | if(!clen) |
816 | | /* . [end] */ |
817 | 0 | goto end; |
818 | 0 | else if(ISSLASH(*p)) { |
819 | | /* one dot followed by a slash */ |
820 | 0 | input = p + 1; |
821 | 0 | clen--; |
822 | 0 | } |
823 | | |
824 | | /* D. if the input buffer consists only of "." or "..", then remove |
825 | | that from the input buffer; otherwise, */ |
826 | 0 | else if(is_dot(&p, &blen)) { |
827 | 0 | if(!blen) |
828 | | /* .. [end] */ |
829 | 0 | goto end; |
830 | 0 | else if(ISSLASH(*p)) { |
831 | | /* ../ */ |
832 | 0 | input = p + 1; |
833 | 0 | clen = blen - 1; |
834 | 0 | } |
835 | 0 | } |
836 | 0 | } |
837 | | |
838 | 46.4k | while(clen && !result) { /* until end of path content */ |
839 | 44.8k | if(ISSLASH(*input)) { |
840 | 1.79k | const char *p = &input[1]; |
841 | 1.79k | size_t blen = clen - 1; |
842 | | /* B. if the input buffer begins with a prefix of "/./" or "/.", where |
843 | | "." is a complete path segment, then replace that prefix with "/" in |
844 | | the input buffer; otherwise, */ |
845 | 1.79k | if(is_dot(&p, &blen)) { |
846 | 155 | if(!blen) { /* /. */ |
847 | 1 | result = curlx_dyn_addn(&out, "/", 1); |
848 | 1 | break; |
849 | 1 | } |
850 | 154 | else if(ISSLASH(*p)) { /* /./ */ |
851 | 0 | input = p; |
852 | 0 | clen = blen; |
853 | 0 | continue; |
854 | 0 | } |
855 | | |
856 | | /* C. if the input buffer begins with a prefix of "/../" or "/..", |
857 | | where ".." is a complete path segment, then replace that prefix |
858 | | with "/" in the input buffer and remove the last segment and its |
859 | | preceding "/" (if any) from the output buffer; otherwise, */ |
860 | 154 | else if(is_dot(&p, &blen) && (ISSLASH(*p) || !blen)) { |
861 | | /* remove the last segment from the output buffer */ |
862 | 0 | size_t len = curlx_dyn_len(&out); |
863 | 0 | if(len) { |
864 | 0 | char *ptr = curlx_dyn_ptr(&out); |
865 | 0 | char *last = memrchr(ptr, '/', len); |
866 | 0 | if(last) |
867 | | /* trim the output at the slash */ |
868 | 0 | curlx_dyn_setlen(&out, last - ptr); |
869 | 0 | } |
870 | |
|
871 | 0 | if(blen) { /* /../ */ |
872 | 0 | input = p; |
873 | 0 | clen = blen; |
874 | 0 | continue; |
875 | 0 | } |
876 | 0 | result = curlx_dyn_addn(&out, "/", 1); |
877 | 0 | break; |
878 | 0 | } |
879 | 155 | } |
880 | 1.79k | } |
881 | | |
882 | | /* E. move the first path segment in the input buffer to the end of |
883 | | the output buffer, including the initial "/" character (if any) and |
884 | | any subsequent characters up to, but not including, the next "/" |
885 | | character or the end of the input buffer. */ |
886 | | |
887 | 44.8k | result = curlx_dyn_addn(&out, input, 1); |
888 | 44.8k | input++; |
889 | 44.8k | clen--; |
890 | 44.8k | } |
891 | 1.57k | end: |
892 | 1.57k | if(!result) { |
893 | 1.57k | if(curlx_dyn_len(&out)) |
894 | 1.57k | *outp = curlx_dyn_ptr(&out); |
895 | 0 | else { |
896 | 0 | *outp = strdup(""); |
897 | 0 | if(!*outp) |
898 | 0 | return 1; |
899 | 0 | } |
900 | 1.57k | } |
901 | 1.57k | return result ? 1 : 0; /* success */ |
902 | 1.57k | } |
903 | | |
904 | | static CURLUcode parseurl(const char *url, CURLU *u, unsigned int flags) |
905 | 1.84k | { |
906 | 1.84k | const char *path; |
907 | 1.84k | size_t pathlen; |
908 | 1.84k | char *query = NULL; |
909 | 1.84k | char *fragment = NULL; |
910 | 1.84k | char schemebuf[MAX_SCHEME_LEN + 1]; |
911 | 1.84k | size_t schemelen = 0; |
912 | 1.84k | size_t urllen; |
913 | 1.84k | CURLUcode result = CURLUE_OK; |
914 | 1.84k | size_t fraglen = 0; |
915 | 1.84k | struct dynbuf host; |
916 | | |
917 | 1.84k | DEBUGASSERT(url); |
918 | | |
919 | 1.84k | curlx_dyn_init(&host, CURL_MAX_INPUT_LENGTH); |
920 | | |
921 | 1.84k | result = Curl_junkscan(url, &urllen, !!(flags & CURLU_ALLOW_SPACE)); |
922 | 1.84k | if(result) |
923 | 75 | goto fail; |
924 | | |
925 | 1.77k | schemelen = Curl_is_absolute_url(url, schemebuf, sizeof(schemebuf), |
926 | 1.77k | flags & (CURLU_GUESS_SCHEME| |
927 | 1.77k | CURLU_DEFAULT_SCHEME)); |
928 | | |
929 | | /* handle the file: scheme */ |
930 | 1.77k | if(schemelen && !strcmp(schemebuf, "file")) { |
931 | 0 | bool uncpath = FALSE; |
932 | 0 | if(urllen <= 6) { |
933 | | /* file:/ is not enough to actually be a complete file: URL */ |
934 | 0 | result = CURLUE_BAD_FILE_URL; |
935 | 0 | goto fail; |
936 | 0 | } |
937 | | |
938 | | /* path has been allocated large enough to hold this */ |
939 | 0 | path = &url[5]; |
940 | 0 | pathlen = urllen - 5; |
941 | |
|
942 | 0 | u->scheme = strdup("file"); |
943 | 0 | if(!u->scheme) { |
944 | 0 | result = CURLUE_OUT_OF_MEMORY; |
945 | 0 | goto fail; |
946 | 0 | } |
947 | | |
948 | | /* Extra handling URLs with an authority component (i.e. that start with |
949 | | * "file://") |
950 | | * |
951 | | * We allow omitted hostname (e.g. file:/<path>) -- valid according to |
952 | | * RFC 8089, but not the (current) WHAT-WG URL spec. |
953 | | */ |
954 | 0 | if(path[0] == '/' && path[1] == '/') { |
955 | | /* swallow the two slashes */ |
956 | 0 | const char *ptr = &path[2]; |
957 | | |
958 | | /* |
959 | | * According to RFC 8089, a file: URL can be reliably dereferenced if: |
960 | | * |
961 | | * o it has no/blank hostname, or |
962 | | * |
963 | | * o the hostname matches "localhost" (case-insensitively), or |
964 | | * |
965 | | * o the hostname is a FQDN that resolves to this machine, or |
966 | | * |
967 | | * o it is an UNC String transformed to an URI (Windows only, RFC 8089 |
968 | | * Appendix E.3). |
969 | | * |
970 | | * For brevity, we only consider URLs with empty, "localhost", or |
971 | | * "127.0.0.1" hostnames as local, otherwise as an UNC String. |
972 | | * |
973 | | * Additionally, there is an exception for URLs with a Windows drive |
974 | | * letter in the authority (which was accidentally omitted from RFC 8089 |
975 | | * Appendix E, but believe me, it was meant to be there. --MK) |
976 | | */ |
977 | 0 | if(ptr[0] != '/' && !STARTS_WITH_URL_DRIVE_PREFIX(ptr)) { |
978 | | /* the URL includes a hostname, it must match "localhost" or |
979 | | "127.0.0.1" to be valid */ |
980 | 0 | if(checkprefix("localhost/", ptr) || |
981 | 0 | checkprefix("127.0.0.1/", ptr)) { |
982 | 0 | ptr += 9; /* now points to the slash after the host */ |
983 | 0 | } |
984 | 0 | else { |
985 | | #ifdef _WIN32 |
986 | | size_t len; |
987 | | |
988 | | /* the hostname, NetBIOS computer name, can not contain disallowed |
989 | | chars, and the delimiting slash character must be appended to the |
990 | | hostname */ |
991 | | path = strpbrk(ptr, "/\\:*?\"<>|"); |
992 | | if(!path || *path != '/') { |
993 | | result = CURLUE_BAD_FILE_URL; |
994 | | goto fail; |
995 | | } |
996 | | |
997 | | len = path - ptr; |
998 | | if(len) { |
999 | | CURLcode code = curlx_dyn_addn(&host, ptr, len); |
1000 | | if(code) { |
1001 | | result = cc2cu(code); |
1002 | | goto fail; |
1003 | | } |
1004 | | uncpath = TRUE; |
1005 | | } |
1006 | | |
1007 | | ptr -= 2; /* now points to the // before the host in UNC */ |
1008 | | #else |
1009 | | /* Invalid file://hostname/, expected localhost or 127.0.0.1 or |
1010 | | none */ |
1011 | 0 | result = CURLUE_BAD_FILE_URL; |
1012 | 0 | goto fail; |
1013 | 0 | #endif |
1014 | 0 | } |
1015 | 0 | } |
1016 | | |
1017 | 0 | path = ptr; |
1018 | 0 | pathlen = urllen - (ptr - url); |
1019 | 0 | } |
1020 | | |
1021 | 0 | if(!uncpath) |
1022 | | /* no host for file: URLs by default */ |
1023 | 0 | curlx_dyn_reset(&host); |
1024 | |
|
1025 | 0 | #if !defined(_WIN32) && !defined(MSDOS) && !defined(__CYGWIN__) |
1026 | | /* Do not allow Windows drive letters when not in Windows. |
1027 | | * This catches both "file:/c:" and "file:c:" */ |
1028 | 0 | if(('/' == path[0] && STARTS_WITH_URL_DRIVE_PREFIX(&path[1])) || |
1029 | 0 | STARTS_WITH_URL_DRIVE_PREFIX(path)) { |
1030 | | /* File drive letters are only accepted in MS-DOS/Windows */ |
1031 | 0 | result = CURLUE_BAD_FILE_URL; |
1032 | 0 | goto fail; |
1033 | 0 | } |
1034 | | #else |
1035 | | /* If the path starts with a slash and a drive letter, ditch the slash */ |
1036 | | if('/' == path[0] && STARTS_WITH_URL_DRIVE_PREFIX(&path[1])) { |
1037 | | /* This cannot be done with strcpy, as the memory chunks overlap! */ |
1038 | | path++; |
1039 | | pathlen--; |
1040 | | } |
1041 | | #endif |
1042 | |
|
1043 | 0 | } |
1044 | 1.77k | else { |
1045 | | /* clear path */ |
1046 | 1.77k | const char *schemep = NULL; |
1047 | 1.77k | const char *hostp; |
1048 | 1.77k | size_t hostlen; |
1049 | | |
1050 | 1.77k | if(schemelen) { |
1051 | 1.74k | int i = 0; |
1052 | 1.74k | const char *p = &url[schemelen + 1]; |
1053 | 5.21k | while((*p == '/') && (i < 4)) { |
1054 | 3.47k | p++; |
1055 | 3.47k | i++; |
1056 | 3.47k | } |
1057 | | |
1058 | 1.74k | schemep = schemebuf; |
1059 | 1.74k | if(!Curl_get_scheme_handler(schemep) && |
1060 | 1.74k | !(flags & CURLU_NON_SUPPORT_SCHEME)) { |
1061 | 0 | result = CURLUE_UNSUPPORTED_SCHEME; |
1062 | 0 | goto fail; |
1063 | 0 | } |
1064 | | |
1065 | 1.74k | if((i < 1) || (i > 3)) { |
1066 | | /* less than one or more than three slashes */ |
1067 | 0 | result = CURLUE_BAD_SLASHES; |
1068 | 0 | goto fail; |
1069 | 0 | } |
1070 | 1.74k | hostp = p; /* hostname starts here */ |
1071 | 1.74k | } |
1072 | 34 | else { |
1073 | | /* no scheme! */ |
1074 | | |
1075 | 34 | if(!(flags & (CURLU_DEFAULT_SCHEME|CURLU_GUESS_SCHEME))) { |
1076 | 0 | result = CURLUE_BAD_SCHEME; |
1077 | 0 | goto fail; |
1078 | 0 | } |
1079 | 34 | if(flags & CURLU_DEFAULT_SCHEME) |
1080 | 0 | schemep = DEFAULT_SCHEME; |
1081 | | |
1082 | | /* |
1083 | | * The URL was badly formatted, let's try without scheme specified. |
1084 | | */ |
1085 | 34 | hostp = url; |
1086 | 34 | } |
1087 | | |
1088 | 1.77k | if(schemep) { |
1089 | 1.74k | u->scheme = strdup(schemep); |
1090 | 1.74k | if(!u->scheme) { |
1091 | 0 | result = CURLUE_OUT_OF_MEMORY; |
1092 | 0 | goto fail; |
1093 | 0 | } |
1094 | 1.74k | } |
1095 | | |
1096 | | /* find the end of the hostname + port number */ |
1097 | 1.77k | hostlen = strcspn(hostp, "/?#"); |
1098 | 1.77k | path = &hostp[hostlen]; |
1099 | | |
1100 | | /* this pathlen also contains the query and the fragment */ |
1101 | 1.77k | pathlen = urllen - (path - url); |
1102 | 1.77k | if(hostlen) { |
1103 | | |
1104 | 1.77k | result = parse_authority(u, hostp, hostlen, flags, &host, schemelen); |
1105 | 1.77k | if(result) |
1106 | 33 | goto fail; |
1107 | | |
1108 | 1.74k | if((flags & CURLU_GUESS_SCHEME) && !schemep) { |
1109 | 23 | const char *hostname = curlx_dyn_ptr(&host); |
1110 | | /* legacy curl-style guess based on hostname */ |
1111 | 23 | if(checkprefix("ftp.", hostname)) |
1112 | 0 | schemep = "ftp"; |
1113 | 23 | else if(checkprefix("dict.", hostname)) |
1114 | 0 | schemep = "dict"; |
1115 | 23 | else if(checkprefix("ldap.", hostname)) |
1116 | 0 | schemep = "ldap"; |
1117 | 23 | else if(checkprefix("imap.", hostname)) |
1118 | 0 | schemep = "imap"; |
1119 | 23 | else if(checkprefix("smtp.", hostname)) |
1120 | 0 | schemep = "smtp"; |
1121 | 23 | else if(checkprefix("pop3.", hostname)) |
1122 | 0 | schemep = "pop3"; |
1123 | 23 | else |
1124 | 23 | schemep = "http"; |
1125 | | |
1126 | 23 | u->scheme = strdup(schemep); |
1127 | 23 | if(!u->scheme) { |
1128 | 0 | result = CURLUE_OUT_OF_MEMORY; |
1129 | 0 | goto fail; |
1130 | 0 | } |
1131 | 23 | u->guessed_scheme = TRUE; |
1132 | 23 | } |
1133 | 1.74k | } |
1134 | 0 | else if(flags & CURLU_NO_AUTHORITY) { |
1135 | | /* allowed to be empty. */ |
1136 | 0 | if(curlx_dyn_add(&host, "")) { |
1137 | 0 | result = CURLUE_OUT_OF_MEMORY; |
1138 | 0 | goto fail; |
1139 | 0 | } |
1140 | 0 | } |
1141 | 0 | else { |
1142 | 0 | result = CURLUE_NO_HOST; |
1143 | 0 | goto fail; |
1144 | 0 | } |
1145 | 1.77k | } |
1146 | | |
1147 | 1.74k | fragment = strchr(path, '#'); |
1148 | 1.74k | if(fragment) { |
1149 | 5 | fraglen = pathlen - (fragment - path); |
1150 | 5 | u->fragment_present = TRUE; |
1151 | 5 | if(fraglen > 1) { |
1152 | | /* skip the leading '#' in the copy but include the terminating null */ |
1153 | 4 | if(flags & CURLU_URLENCODE) { |
1154 | 0 | struct dynbuf enc; |
1155 | 0 | curlx_dyn_init(&enc, CURL_MAX_INPUT_LENGTH); |
1156 | 0 | result = urlencode_str(&enc, fragment + 1, fraglen - 1, TRUE, FALSE); |
1157 | 0 | if(result) |
1158 | 0 | goto fail; |
1159 | 0 | u->fragment = curlx_dyn_ptr(&enc); |
1160 | 0 | } |
1161 | 4 | else { |
1162 | 4 | u->fragment = Curl_memdup0(fragment + 1, fraglen - 1); |
1163 | 4 | if(!u->fragment) { |
1164 | 0 | result = CURLUE_OUT_OF_MEMORY; |
1165 | 0 | goto fail; |
1166 | 0 | } |
1167 | 4 | } |
1168 | 4 | } |
1169 | | /* after this, pathlen still contains the query */ |
1170 | 5 | pathlen -= fraglen; |
1171 | 5 | } |
1172 | | |
1173 | 1.74k | query = memchr(path, '?', pathlen); |
1174 | 1.74k | if(query) { |
1175 | 25 | size_t qlen = fragment ? (size_t)(fragment - query) : |
1176 | 25 | pathlen - (query - path); |
1177 | 25 | pathlen -= qlen; |
1178 | 25 | u->query_present = TRUE; |
1179 | 25 | if(qlen > 1) { |
1180 | 25 | if(flags & CURLU_URLENCODE) { |
1181 | 0 | struct dynbuf enc; |
1182 | 0 | curlx_dyn_init(&enc, CURL_MAX_INPUT_LENGTH); |
1183 | | /* skip the leading question mark */ |
1184 | 0 | result = urlencode_str(&enc, query + 1, qlen - 1, TRUE, TRUE); |
1185 | 0 | if(result) |
1186 | 0 | goto fail; |
1187 | 0 | u->query = curlx_dyn_ptr(&enc); |
1188 | 0 | } |
1189 | 25 | else { |
1190 | 25 | u->query = Curl_memdup0(query + 1, qlen - 1); |
1191 | 25 | if(!u->query) { |
1192 | 0 | result = CURLUE_OUT_OF_MEMORY; |
1193 | 0 | goto fail; |
1194 | 0 | } |
1195 | 25 | } |
1196 | 25 | } |
1197 | 0 | else { |
1198 | | /* single byte query */ |
1199 | 0 | u->query = strdup(""); |
1200 | 0 | if(!u->query) { |
1201 | 0 | result = CURLUE_OUT_OF_MEMORY; |
1202 | 0 | goto fail; |
1203 | 0 | } |
1204 | 0 | } |
1205 | 25 | } |
1206 | | |
1207 | 1.74k | if(pathlen && (flags & CURLU_URLENCODE)) { |
1208 | 0 | struct dynbuf enc; |
1209 | 0 | curlx_dyn_init(&enc, CURL_MAX_INPUT_LENGTH); |
1210 | 0 | result = urlencode_str(&enc, path, pathlen, TRUE, FALSE); |
1211 | 0 | if(result) |
1212 | 0 | goto fail; |
1213 | 0 | pathlen = curlx_dyn_len(&enc); |
1214 | 0 | path = u->path = curlx_dyn_ptr(&enc); |
1215 | 0 | } |
1216 | | |
1217 | 1.74k | if(pathlen <= 1) { |
1218 | | /* there is no path left or just the slash, unset */ |
1219 | 166 | path = NULL; |
1220 | 166 | } |
1221 | 1.57k | else { |
1222 | 1.57k | if(!u->path) { |
1223 | 1.57k | u->path = Curl_memdup0(path, pathlen); |
1224 | 1.57k | if(!u->path) { |
1225 | 0 | result = CURLUE_OUT_OF_MEMORY; |
1226 | 0 | goto fail; |
1227 | 0 | } |
1228 | 1.57k | path = u->path; |
1229 | 1.57k | } |
1230 | 0 | else if(flags & CURLU_URLENCODE) |
1231 | | /* it might have encoded more than just the path so cut it */ |
1232 | 0 | u->path[pathlen] = 0; |
1233 | | |
1234 | 1.57k | if(!(flags & CURLU_PATH_AS_IS)) { |
1235 | | /* remove ../ and ./ sequences according to RFC3986 */ |
1236 | 1.57k | char *dedot; |
1237 | 1.57k | int err = dedotdotify(path, pathlen, &dedot); |
1238 | 1.57k | if(err) { |
1239 | 0 | result = CURLUE_OUT_OF_MEMORY; |
1240 | 0 | goto fail; |
1241 | 0 | } |
1242 | 1.57k | if(dedot) { |
1243 | 1.57k | free(u->path); |
1244 | 1.57k | u->path = dedot; |
1245 | 1.57k | } |
1246 | 1.57k | } |
1247 | 1.57k | } |
1248 | | |
1249 | 1.74k | u->host = curlx_dyn_ptr(&host); |
1250 | | |
1251 | 1.74k | return result; |
1252 | 108 | fail: |
1253 | 108 | curlx_dyn_free(&host); |
1254 | 108 | free_urlhandle(u); |
1255 | 108 | return result; |
1256 | 1.74k | } |
1257 | | |
1258 | | /* |
1259 | | * Parse the URL and, if successful, replace everything in the Curl_URL struct. |
1260 | | */ |
1261 | | static CURLUcode parseurl_and_replace(const char *url, CURLU *u, |
1262 | | unsigned int flags) |
1263 | 1.84k | { |
1264 | 1.84k | CURLUcode result; |
1265 | 1.84k | CURLU tmpurl; |
1266 | 1.84k | memset(&tmpurl, 0, sizeof(tmpurl)); |
1267 | 1.84k | result = parseurl(url, &tmpurl, flags); |
1268 | 1.84k | if(!result) { |
1269 | 1.74k | free_urlhandle(u); |
1270 | 1.74k | *u = tmpurl; |
1271 | 1.74k | } |
1272 | 1.84k | return result; |
1273 | 1.84k | } |
1274 | | |
1275 | | /* |
1276 | | */ |
1277 | | CURLU *curl_url(void) |
1278 | 1.85k | { |
1279 | 1.85k | return calloc(1, sizeof(struct Curl_URL)); |
1280 | 1.85k | } |
1281 | | |
1282 | | void curl_url_cleanup(CURLU *u) |
1283 | 3.76k | { |
1284 | 3.76k | if(u) { |
1285 | 1.85k | free_urlhandle(u); |
1286 | 1.85k | free(u); |
1287 | 1.85k | } |
1288 | 3.76k | } |
1289 | | |
1290 | | #define DUP(dest, src, name) \ |
1291 | 0 | do { \ |
1292 | 0 | if(src->name) { \ |
1293 | 0 | dest->name = strdup(src->name); \ |
1294 | 0 | if(!dest->name) \ |
1295 | 0 | goto fail; \ |
1296 | 0 | } \ |
1297 | 0 | } while(0) |
1298 | | |
1299 | | CURLU *curl_url_dup(const CURLU *in) |
1300 | 0 | { |
1301 | 0 | struct Curl_URL *u = calloc(1, sizeof(struct Curl_URL)); |
1302 | 0 | if(u) { |
1303 | 0 | DUP(u, in, scheme); |
1304 | 0 | DUP(u, in, user); |
1305 | 0 | DUP(u, in, password); |
1306 | 0 | DUP(u, in, options); |
1307 | 0 | DUP(u, in, host); |
1308 | 0 | DUP(u, in, port); |
1309 | 0 | DUP(u, in, path); |
1310 | 0 | DUP(u, in, query); |
1311 | 0 | DUP(u, in, fragment); |
1312 | 0 | DUP(u, in, zoneid); |
1313 | 0 | u->portnum = in->portnum; |
1314 | 0 | u->fragment_present = in->fragment_present; |
1315 | 0 | u->query_present = in->query_present; |
1316 | 0 | } |
1317 | 0 | return u; |
1318 | 0 | fail: |
1319 | 0 | curl_url_cleanup(u); |
1320 | 0 | return NULL; |
1321 | 0 | } |
1322 | | |
1323 | | CURLUcode curl_url_get(const CURLU *u, CURLUPart what, |
1324 | | char **part, unsigned int flags) |
1325 | 14.0k | { |
1326 | 14.0k | const char *ptr; |
1327 | 14.0k | CURLUcode ifmissing = CURLUE_UNKNOWN_PART; |
1328 | 14.0k | char portbuf[7]; |
1329 | 14.0k | bool urldecode = (flags & CURLU_URLDECODE) ? 1 : 0; |
1330 | 14.0k | bool urlencode = (flags & CURLU_URLENCODE) ? 1 : 0; |
1331 | 14.0k | bool punycode = FALSE; |
1332 | 14.0k | bool depunyfy = FALSE; |
1333 | 14.0k | bool plusdecode = FALSE; |
1334 | 14.0k | (void)flags; |
1335 | 14.0k | if(!u) |
1336 | 0 | return CURLUE_BAD_HANDLE; |
1337 | 14.0k | if(!part) |
1338 | 0 | return CURLUE_BAD_PARTPOINTER; |
1339 | 14.0k | *part = NULL; |
1340 | | |
1341 | 14.0k | switch(what) { |
1342 | 1.74k | case CURLUPART_SCHEME: |
1343 | 1.74k | ptr = u->scheme; |
1344 | 1.74k | ifmissing = CURLUE_NO_SCHEME; |
1345 | 1.74k | urldecode = FALSE; /* never for schemes */ |
1346 | 1.74k | if((flags & CURLU_NO_GUESS_SCHEME) && u->guessed_scheme) |
1347 | 0 | return CURLUE_NO_SCHEME; |
1348 | 1.74k | break; |
1349 | 1.74k | case CURLUPART_USER: |
1350 | 1.44k | ptr = u->user; |
1351 | 1.44k | ifmissing = CURLUE_NO_USER; |
1352 | 1.44k | break; |
1353 | 1.44k | case CURLUPART_PASSWORD: |
1354 | 1.44k | ptr = u->password; |
1355 | 1.44k | ifmissing = CURLUE_NO_PASSWORD; |
1356 | 1.44k | break; |
1357 | 1.44k | case CURLUPART_OPTIONS: |
1358 | 1.44k | ptr = u->options; |
1359 | 1.44k | ifmissing = CURLUE_NO_OPTIONS; |
1360 | 1.44k | break; |
1361 | 1.74k | case CURLUPART_HOST: |
1362 | 1.74k | ptr = u->host; |
1363 | 1.74k | ifmissing = CURLUE_NO_HOST; |
1364 | 1.74k | punycode = (flags & CURLU_PUNYCODE) ? 1 : 0; |
1365 | 1.74k | depunyfy = (flags & CURLU_PUNY2IDN) ? 1 : 0; |
1366 | 1.74k | break; |
1367 | 0 | case CURLUPART_ZONEID: |
1368 | 0 | ptr = u->zoneid; |
1369 | 0 | ifmissing = CURLUE_NO_ZONEID; |
1370 | 0 | break; |
1371 | 1.44k | case CURLUPART_PORT: |
1372 | 1.44k | ptr = u->port; |
1373 | 1.44k | ifmissing = CURLUE_NO_PORT; |
1374 | 1.44k | urldecode = FALSE; /* never for port */ |
1375 | 1.44k | if(!ptr && (flags & CURLU_DEFAULT_PORT) && u->scheme) { |
1376 | | /* there is no stored port number, but asked to deliver |
1377 | | a default one for the scheme */ |
1378 | 1.44k | const struct Curl_handler *h = Curl_get_scheme_handler(u->scheme); |
1379 | 1.44k | if(h) { |
1380 | 1.44k | msnprintf(portbuf, sizeof(portbuf), "%u", h->defport); |
1381 | 1.44k | ptr = portbuf; |
1382 | 1.44k | } |
1383 | 1.44k | } |
1384 | 1 | else if(ptr && u->scheme) { |
1385 | | /* there is a stored port number, but ask to inhibit if |
1386 | | it matches the default one for the scheme */ |
1387 | 1 | const struct Curl_handler *h = Curl_get_scheme_handler(u->scheme); |
1388 | 1 | if(h && (h->defport == u->portnum) && |
1389 | 1 | (flags & CURLU_NO_DEFAULT_PORT)) |
1390 | 0 | ptr = NULL; |
1391 | 1 | } |
1392 | 1.44k | break; |
1393 | 1.44k | case CURLUPART_PATH: |
1394 | 1.44k | ptr = u->path; |
1395 | 1.44k | if(!ptr) |
1396 | 149 | ptr = "/"; |
1397 | 1.44k | break; |
1398 | 1.44k | case CURLUPART_QUERY: |
1399 | 1.44k | ptr = u->query; |
1400 | 1.44k | ifmissing = CURLUE_NO_QUERY; |
1401 | 1.44k | plusdecode = urldecode; |
1402 | 1.44k | if(ptr && !ptr[0] && !(flags & CURLU_GET_EMPTY)) |
1403 | | /* there was a blank query and the user do not ask for it */ |
1404 | 0 | ptr = NULL; |
1405 | 1.44k | break; |
1406 | 0 | case CURLUPART_FRAGMENT: |
1407 | 0 | ptr = u->fragment; |
1408 | 0 | ifmissing = CURLUE_NO_FRAGMENT; |
1409 | 0 | if(!ptr && u->fragment_present && flags & CURLU_GET_EMPTY) |
1410 | | /* there was a blank fragment and the user asks for it */ |
1411 | 0 | ptr = ""; |
1412 | 0 | break; |
1413 | 1.84k | case CURLUPART_URL: { |
1414 | 1.84k | char *url; |
1415 | 1.84k | const char *scheme; |
1416 | 1.84k | char *options = u->options; |
1417 | 1.84k | char *port = u->port; |
1418 | 1.84k | char *allochost = NULL; |
1419 | 1.84k | bool show_fragment = |
1420 | 1.84k | u->fragment || (u->fragment_present && flags & CURLU_GET_EMPTY); |
1421 | 1.84k | bool show_query = |
1422 | 1.84k | (u->query && u->query[0]) || |
1423 | 1.84k | (u->query_present && flags & CURLU_GET_EMPTY); |
1424 | 1.84k | punycode = (flags & CURLU_PUNYCODE) ? 1 : 0; |
1425 | 1.84k | depunyfy = (flags & CURLU_PUNY2IDN) ? 1 : 0; |
1426 | 1.84k | if(u->scheme && strcasecompare("file", u->scheme)) { |
1427 | 0 | url = aprintf("file://%s%s%s%s%s", |
1428 | 0 | u->path, |
1429 | 0 | show_query ? "?": "", |
1430 | 0 | u->query ? u->query : "", |
1431 | 0 | show_fragment ? "#": "", |
1432 | 0 | u->fragment ? u->fragment : ""); |
1433 | 0 | } |
1434 | 1.84k | else if(!u->host) |
1435 | 107 | return CURLUE_NO_HOST; |
1436 | 1.74k | else { |
1437 | 1.74k | const struct Curl_handler *h = NULL; |
1438 | 1.74k | char schemebuf[MAX_SCHEME_LEN + 5]; |
1439 | 1.74k | if(u->scheme) |
1440 | 1.74k | scheme = u->scheme; |
1441 | 0 | else if(flags & CURLU_DEFAULT_SCHEME) |
1442 | 0 | scheme = DEFAULT_SCHEME; |
1443 | 0 | else |
1444 | 0 | return CURLUE_NO_SCHEME; |
1445 | | |
1446 | 1.74k | h = Curl_get_scheme_handler(scheme); |
1447 | 1.74k | if(!port && (flags & CURLU_DEFAULT_PORT)) { |
1448 | | /* there is no stored port number, but asked to deliver |
1449 | | a default one for the scheme */ |
1450 | 0 | if(h) { |
1451 | 0 | msnprintf(portbuf, sizeof(portbuf), "%u", h->defport); |
1452 | 0 | port = portbuf; |
1453 | 0 | } |
1454 | 0 | } |
1455 | 1.74k | else if(port) { |
1456 | | /* there is a stored port number, but asked to inhibit if it matches |
1457 | | the default one for the scheme */ |
1458 | 1 | if(h && (h->defport == u->portnum) && |
1459 | 1 | (flags & CURLU_NO_DEFAULT_PORT)) |
1460 | 0 | port = NULL; |
1461 | 1 | } |
1462 | | |
1463 | 1.74k | if(h && !(h->flags & PROTOPT_URLOPTIONS)) |
1464 | 1.44k | options = NULL; |
1465 | | |
1466 | 1.74k | if(u->host[0] == '[') { |
1467 | 0 | if(u->zoneid) { |
1468 | | /* make it '[ host %25 zoneid ]' */ |
1469 | 0 | struct dynbuf enc; |
1470 | 0 | size_t hostlen = strlen(u->host); |
1471 | 0 | curlx_dyn_init(&enc, CURL_MAX_INPUT_LENGTH); |
1472 | 0 | if(curlx_dyn_addf(&enc, "%.*s%%25%s]", (int)hostlen - 1, u->host, |
1473 | 0 | u->zoneid)) |
1474 | 0 | return CURLUE_OUT_OF_MEMORY; |
1475 | 0 | allochost = curlx_dyn_ptr(&enc); |
1476 | 0 | } |
1477 | 0 | } |
1478 | 1.74k | else if(urlencode) { |
1479 | 0 | allochost = curl_easy_escape(NULL, u->host, 0); |
1480 | 0 | if(!allochost) |
1481 | 0 | return CURLUE_OUT_OF_MEMORY; |
1482 | 0 | } |
1483 | 1.74k | else if(punycode) { |
1484 | 0 | if(!Curl_is_ASCII_name(u->host)) { |
1485 | 0 | #ifndef USE_IDN |
1486 | 0 | return CURLUE_LACKS_IDN; |
1487 | | #else |
1488 | | CURLcode result = Curl_idn_decode(u->host, &allochost); |
1489 | | if(result) |
1490 | | return (result == CURLE_OUT_OF_MEMORY) ? |
1491 | | CURLUE_OUT_OF_MEMORY : CURLUE_BAD_HOSTNAME; |
1492 | | #endif |
1493 | 0 | } |
1494 | 0 | } |
1495 | 1.74k | else if(depunyfy) { |
1496 | 0 | if(Curl_is_ASCII_name(u->host)) { |
1497 | 0 | #ifndef USE_IDN |
1498 | 0 | return CURLUE_LACKS_IDN; |
1499 | | #else |
1500 | | CURLcode result = Curl_idn_encode(u->host, &allochost); |
1501 | | if(result) |
1502 | | /* this is the most likely error */ |
1503 | | return (result == CURLE_OUT_OF_MEMORY) ? |
1504 | | CURLUE_OUT_OF_MEMORY : CURLUE_BAD_HOSTNAME; |
1505 | | #endif |
1506 | 0 | } |
1507 | 0 | } |
1508 | | |
1509 | 1.74k | if(!(flags & CURLU_NO_GUESS_SCHEME) || !u->guessed_scheme) |
1510 | 1.74k | msnprintf(schemebuf, sizeof(schemebuf), "%s://", scheme); |
1511 | 0 | else |
1512 | 0 | schemebuf[0] = 0; |
1513 | | |
1514 | 1.74k | url = aprintf("%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s", |
1515 | 1.74k | schemebuf, |
1516 | 1.74k | u->user ? u->user : "", |
1517 | 1.74k | u->password ? ":": "", |
1518 | 1.74k | u->password ? u->password : "", |
1519 | 1.74k | options ? ";" : "", |
1520 | 1.74k | options ? options : "", |
1521 | 1.74k | (u->user || u->password || options) ? "@": "", |
1522 | 1.74k | allochost ? allochost : u->host, |
1523 | 1.74k | port ? ":": "", |
1524 | 1.74k | port ? port : "", |
1525 | 1.74k | u->path ? u->path : "/", |
1526 | 1.74k | show_query ? "?": "", |
1527 | 1.74k | u->query ? u->query : "", |
1528 | 1.74k | show_fragment ? "#": "", |
1529 | 1.74k | u->fragment ? u->fragment : ""); |
1530 | 1.74k | free(allochost); |
1531 | 1.74k | } |
1532 | 1.74k | if(!url) |
1533 | 0 | return CURLUE_OUT_OF_MEMORY; |
1534 | 1.74k | *part = url; |
1535 | 1.74k | return CURLUE_OK; |
1536 | 1.74k | } |
1537 | 0 | default: |
1538 | 0 | ptr = NULL; |
1539 | 0 | break; |
1540 | 14.0k | } |
1541 | 12.1k | if(ptr) { |
1542 | 8.81k | size_t partlen = strlen(ptr); |
1543 | 8.81k | size_t i = 0; |
1544 | 8.81k | *part = Curl_memdup0(ptr, partlen); |
1545 | 8.81k | if(!*part) |
1546 | 0 | return CURLUE_OUT_OF_MEMORY; |
1547 | 8.81k | if(plusdecode) { |
1548 | | /* convert + to space */ |
1549 | 0 | char *plus = *part; |
1550 | 0 | for(i = 0; i < partlen; ++plus, i++) { |
1551 | 0 | if(*plus == '+') |
1552 | 0 | *plus = ' '; |
1553 | 0 | } |
1554 | 0 | } |
1555 | 8.81k | if(urldecode) { |
1556 | 0 | char *decoded; |
1557 | 0 | size_t dlen; |
1558 | | /* this unconditional rejection of control bytes is documented |
1559 | | API behavior */ |
1560 | 0 | CURLcode res = Curl_urldecode(*part, 0, &decoded, &dlen, REJECT_CTRL); |
1561 | 0 | free(*part); |
1562 | 0 | if(res) { |
1563 | 0 | *part = NULL; |
1564 | 0 | return CURLUE_URLDECODE; |
1565 | 0 | } |
1566 | 0 | *part = decoded; |
1567 | 0 | partlen = dlen; |
1568 | 0 | } |
1569 | 8.81k | if(urlencode) { |
1570 | 1.44k | struct dynbuf enc; |
1571 | 1.44k | CURLUcode uc; |
1572 | 1.44k | curlx_dyn_init(&enc, CURL_MAX_INPUT_LENGTH); |
1573 | 1.44k | uc = urlencode_str(&enc, *part, partlen, TRUE, what == CURLUPART_QUERY); |
1574 | 1.44k | if(uc) |
1575 | 0 | return uc; |
1576 | 1.44k | free(*part); |
1577 | 1.44k | *part = curlx_dyn_ptr(&enc); |
1578 | 1.44k | } |
1579 | 7.37k | else if(punycode) { |
1580 | 0 | if(!Curl_is_ASCII_name(u->host)) { |
1581 | 0 | #ifndef USE_IDN |
1582 | 0 | return CURLUE_LACKS_IDN; |
1583 | | #else |
1584 | | char *allochost; |
1585 | | CURLcode result = Curl_idn_decode(*part, &allochost); |
1586 | | if(result) |
1587 | | return (result == CURLE_OUT_OF_MEMORY) ? |
1588 | | CURLUE_OUT_OF_MEMORY : CURLUE_BAD_HOSTNAME; |
1589 | | free(*part); |
1590 | | *part = allochost; |
1591 | | #endif |
1592 | 0 | } |
1593 | 0 | } |
1594 | 7.37k | else if(depunyfy) { |
1595 | 0 | if(Curl_is_ASCII_name(u->host)) { |
1596 | 0 | #ifndef USE_IDN |
1597 | 0 | return CURLUE_LACKS_IDN; |
1598 | | #else |
1599 | | char *allochost; |
1600 | | CURLcode result = Curl_idn_encode(*part, &allochost); |
1601 | | if(result) |
1602 | | return (result == CURLE_OUT_OF_MEMORY) ? |
1603 | | CURLUE_OUT_OF_MEMORY : CURLUE_BAD_HOSTNAME; |
1604 | | free(*part); |
1605 | | *part = allochost; |
1606 | | #endif |
1607 | 0 | } |
1608 | 0 | } |
1609 | | |
1610 | 8.81k | return CURLUE_OK; |
1611 | 8.81k | } |
1612 | 3.33k | else |
1613 | 3.33k | return ifmissing; |
1614 | 12.1k | } |
1615 | | |
1616 | | static CURLUcode set_url_scheme(CURLU *u, const char *scheme, |
1617 | | unsigned int flags) |
1618 | 0 | { |
1619 | 0 | size_t plen = strlen(scheme); |
1620 | 0 | const char *s = scheme; |
1621 | 0 | if((plen > MAX_SCHEME_LEN) || (plen < 1)) |
1622 | | /* too long or too short */ |
1623 | 0 | return CURLUE_BAD_SCHEME; |
1624 | | /* verify that it is a fine scheme */ |
1625 | 0 | if(!(flags & CURLU_NON_SUPPORT_SCHEME) && !Curl_get_scheme_handler(scheme)) |
1626 | 0 | return CURLUE_UNSUPPORTED_SCHEME; |
1627 | 0 | if(ISALPHA(*s)) { |
1628 | | /* ALPHA *( ALPHA / DIGIT / "+" / "-" / "." ) */ |
1629 | 0 | while(--plen) { |
1630 | 0 | if(ISALNUM(*s) || (*s == '+') || (*s == '-') || (*s == '.')) |
1631 | 0 | s++; /* fine */ |
1632 | 0 | else |
1633 | 0 | return CURLUE_BAD_SCHEME; |
1634 | 0 | } |
1635 | 0 | } |
1636 | 0 | else |
1637 | 0 | return CURLUE_BAD_SCHEME; |
1638 | 0 | u->guessed_scheme = FALSE; |
1639 | 0 | return CURLUE_OK; |
1640 | 0 | } |
1641 | | |
1642 | | static CURLUcode set_url_port(CURLU *u, const char *provided_port) |
1643 | 0 | { |
1644 | 0 | char *tmp; |
1645 | 0 | curl_off_t port; |
1646 | 0 | if(!ISDIGIT(provided_port[0])) |
1647 | | /* not a number */ |
1648 | 0 | return CURLUE_BAD_PORT_NUMBER; |
1649 | 0 | if(curlx_str_number(&provided_port, &port, 0xffff) || *provided_port) |
1650 | | /* weirdly provided number, not good! */ |
1651 | 0 | return CURLUE_BAD_PORT_NUMBER; |
1652 | 0 | tmp = aprintf("%" CURL_FORMAT_CURL_OFF_T, port); |
1653 | 0 | if(!tmp) |
1654 | 0 | return CURLUE_OUT_OF_MEMORY; |
1655 | 0 | free(u->port); |
1656 | 0 | u->port = tmp; |
1657 | 0 | u->portnum = (unsigned short)port; |
1658 | 0 | return CURLUE_OK; |
1659 | 0 | } |
1660 | | |
1661 | | static CURLUcode set_url(CURLU *u, const char *url, size_t part_size, |
1662 | | unsigned int flags) |
1663 | 1.85k | { |
1664 | | /* |
1665 | | * Allow a new URL to replace the existing (if any) contents. |
1666 | | * |
1667 | | * If the existing contents is enough for a URL, allow a relative URL to |
1668 | | * replace it. |
1669 | | */ |
1670 | 1.85k | CURLUcode uc; |
1671 | 1.85k | char *oldurl = NULL; |
1672 | | |
1673 | 1.85k | if(!part_size) { |
1674 | | /* a blank URL is not a valid URL unless we already have a complete one |
1675 | | and this is a redirect */ |
1676 | 1 | if(!curl_url_get(u, CURLUPART_URL, &oldurl, flags)) { |
1677 | | /* success, meaning the "" is a fine relative URL, but nothing |
1678 | | changes */ |
1679 | 0 | free(oldurl); |
1680 | 0 | return CURLUE_OK; |
1681 | 0 | } |
1682 | 1 | return CURLUE_MALFORMED_INPUT; |
1683 | 1 | } |
1684 | | |
1685 | | /* if the new thing is absolute or the old one is not (we could not get an |
1686 | | * absolute URL in 'oldurl'), then replace the existing with the new. */ |
1687 | 1.84k | if(Curl_is_absolute_url(url, NULL, 0, |
1688 | 1.84k | flags & (CURLU_GUESS_SCHEME|CURLU_DEFAULT_SCHEME)) |
1689 | 1.84k | || curl_url_get(u, CURLUPART_URL, &oldurl, flags)) { |
1690 | 1.84k | return parseurl_and_replace(url, u, flags); |
1691 | 1.84k | } |
1692 | 0 | DEBUGASSERT(oldurl); /* it is set here */ |
1693 | | /* apply the relative part to create a new URL */ |
1694 | 0 | uc = redirect_url(oldurl, url, u, flags); |
1695 | 0 | free(oldurl); |
1696 | 0 | return uc; |
1697 | 1.84k | } |
1698 | | |
1699 | | CURLUcode curl_url_set(CURLU *u, CURLUPart what, |
1700 | | const char *part, unsigned int flags) |
1701 | 4.27k | { |
1702 | 4.27k | char **storep = NULL; |
1703 | 4.27k | bool urlencode = (flags & CURLU_URLENCODE) ? 1 : 0; |
1704 | 4.27k | bool plusencode = FALSE; |
1705 | 4.27k | bool urlskipslash = FALSE; |
1706 | 4.27k | bool leadingslash = FALSE; |
1707 | 4.27k | bool appendquery = FALSE; |
1708 | 4.27k | bool equalsencode = FALSE; |
1709 | 4.27k | size_t nalloc; |
1710 | | |
1711 | 4.27k | if(!u) |
1712 | 0 | return CURLUE_BAD_HANDLE; |
1713 | 4.27k | if(!part) { |
1714 | | /* setting a part to NULL clears it */ |
1715 | 0 | switch(what) { |
1716 | 0 | case CURLUPART_URL: |
1717 | 0 | break; |
1718 | 0 | case CURLUPART_SCHEME: |
1719 | 0 | storep = &u->scheme; |
1720 | 0 | u->guessed_scheme = FALSE; |
1721 | 0 | break; |
1722 | 0 | case CURLUPART_USER: |
1723 | 0 | storep = &u->user; |
1724 | 0 | break; |
1725 | 0 | case CURLUPART_PASSWORD: |
1726 | 0 | storep = &u->password; |
1727 | 0 | break; |
1728 | 0 | case CURLUPART_OPTIONS: |
1729 | 0 | storep = &u->options; |
1730 | 0 | break; |
1731 | 0 | case CURLUPART_HOST: |
1732 | 0 | storep = &u->host; |
1733 | 0 | break; |
1734 | 0 | case CURLUPART_ZONEID: |
1735 | 0 | storep = &u->zoneid; |
1736 | 0 | break; |
1737 | 0 | case CURLUPART_PORT: |
1738 | 0 | u->portnum = 0; |
1739 | 0 | storep = &u->port; |
1740 | 0 | break; |
1741 | 0 | case CURLUPART_PATH: |
1742 | 0 | storep = &u->path; |
1743 | 0 | break; |
1744 | 0 | case CURLUPART_QUERY: |
1745 | 0 | storep = &u->query; |
1746 | 0 | u->query_present = FALSE; |
1747 | 0 | break; |
1748 | 0 | case CURLUPART_FRAGMENT: |
1749 | 0 | storep = &u->fragment; |
1750 | 0 | u->fragment_present = FALSE; |
1751 | 0 | break; |
1752 | 0 | default: |
1753 | 0 | return CURLUE_UNKNOWN_PART; |
1754 | 0 | } |
1755 | 0 | if(storep && *storep) { |
1756 | 0 | Curl_safefree(*storep); |
1757 | 0 | } |
1758 | 0 | else if(!storep) { |
1759 | 0 | free_urlhandle(u); |
1760 | 0 | memset(u, 0, sizeof(struct Curl_URL)); |
1761 | 0 | } |
1762 | 0 | return CURLUE_OK; |
1763 | 0 | } |
1764 | | |
1765 | 4.27k | nalloc = strlen(part); |
1766 | 4.27k | if(nalloc > CURL_MAX_INPUT_LENGTH) |
1767 | | /* excessive input length */ |
1768 | 0 | return CURLUE_MALFORMED_INPUT; |
1769 | | |
1770 | 4.27k | switch(what) { |
1771 | 0 | case CURLUPART_SCHEME: { |
1772 | 0 | CURLUcode status = set_url_scheme(u, part, flags); |
1773 | 0 | if(status) |
1774 | 0 | return status; |
1775 | 0 | storep = &u->scheme; |
1776 | 0 | urlencode = FALSE; /* never */ |
1777 | 0 | break; |
1778 | 0 | } |
1779 | 1.24k | case CURLUPART_USER: |
1780 | 1.24k | storep = &u->user; |
1781 | 1.24k | break; |
1782 | 1.17k | case CURLUPART_PASSWORD: |
1783 | 1.17k | storep = &u->password; |
1784 | 1.17k | break; |
1785 | 0 | case CURLUPART_OPTIONS: |
1786 | 0 | storep = &u->options; |
1787 | 0 | break; |
1788 | 0 | case CURLUPART_HOST: |
1789 | 0 | storep = &u->host; |
1790 | 0 | Curl_safefree(u->zoneid); |
1791 | 0 | break; |
1792 | 0 | case CURLUPART_ZONEID: |
1793 | 0 | storep = &u->zoneid; |
1794 | 0 | break; |
1795 | 0 | case CURLUPART_PORT: |
1796 | 0 | return set_url_port(u, part); |
1797 | 0 | case CURLUPART_PATH: |
1798 | 0 | urlskipslash = TRUE; |
1799 | 0 | leadingslash = TRUE; /* enforce */ |
1800 | 0 | storep = &u->path; |
1801 | 0 | break; |
1802 | 0 | case CURLUPART_QUERY: |
1803 | 0 | plusencode = urlencode; |
1804 | 0 | appendquery = (flags & CURLU_APPENDQUERY) ? 1 : 0; |
1805 | 0 | equalsencode = appendquery; |
1806 | 0 | storep = &u->query; |
1807 | 0 | u->query_present = TRUE; |
1808 | 0 | break; |
1809 | 0 | case CURLUPART_FRAGMENT: |
1810 | 0 | storep = &u->fragment; |
1811 | 0 | u->fragment_present = TRUE; |
1812 | 0 | break; |
1813 | 1.85k | case CURLUPART_URL: { |
1814 | 1.85k | return set_url(u, part, nalloc, flags); |
1815 | 0 | } |
1816 | 0 | default: |
1817 | 0 | return CURLUE_UNKNOWN_PART; |
1818 | 4.27k | } |
1819 | 2.42k | DEBUGASSERT(storep); |
1820 | 2.42k | { |
1821 | 2.42k | const char *newp; |
1822 | 2.42k | struct dynbuf enc; |
1823 | 2.42k | curlx_dyn_init(&enc, nalloc * 3 + 1 + leadingslash); |
1824 | | |
1825 | 2.42k | if(leadingslash && (part[0] != '/')) { |
1826 | 0 | CURLcode result = curlx_dyn_addn(&enc, "/", 1); |
1827 | 0 | if(result) |
1828 | 0 | return cc2cu(result); |
1829 | 0 | } |
1830 | 2.42k | if(urlencode) { |
1831 | 2.42k | const unsigned char *i; |
1832 | | |
1833 | 150k | for(i = (const unsigned char *)part; *i; i++) { |
1834 | 148k | CURLcode result; |
1835 | 148k | if((*i == ' ') && plusencode) { |
1836 | 0 | result = curlx_dyn_addn(&enc, "+", 1); |
1837 | 0 | if(result) |
1838 | 0 | return CURLUE_OUT_OF_MEMORY; |
1839 | 0 | } |
1840 | 148k | else if(ISUNRESERVED(*i) || |
1841 | 148k | ((*i == '/') && urlskipslash) || |
1842 | 148k | ((*i == '=') && equalsencode)) { |
1843 | 82.4k | if((*i == '=') && equalsencode) |
1844 | | /* only skip the first equals sign */ |
1845 | 0 | equalsencode = FALSE; |
1846 | 82.4k | result = curlx_dyn_addn(&enc, i, 1); |
1847 | 82.4k | if(result) |
1848 | 0 | return cc2cu(result); |
1849 | 82.4k | } |
1850 | 65.8k | else { |
1851 | 65.8k | unsigned char out[3]={'%'}; |
1852 | 65.8k | Curl_hexbyte(&out[1], *i, TRUE); |
1853 | 65.8k | result = curlx_dyn_addn(&enc, out, 3); |
1854 | 65.8k | if(result) |
1855 | 0 | return cc2cu(result); |
1856 | 65.8k | } |
1857 | 148k | } |
1858 | 2.42k | } |
1859 | 0 | else { |
1860 | 0 | char *p; |
1861 | 0 | CURLcode result = curlx_dyn_add(&enc, part); |
1862 | 0 | if(result) |
1863 | 0 | return cc2cu(result); |
1864 | 0 | p = curlx_dyn_ptr(&enc); |
1865 | 0 | while(*p) { |
1866 | | /* make sure percent encoded are lower case */ |
1867 | 0 | if((*p == '%') && ISXDIGIT(p[1]) && ISXDIGIT(p[2]) && |
1868 | 0 | (ISUPPER(p[1]) || ISUPPER(p[2]))) { |
1869 | 0 | p[1] = Curl_raw_tolower(p[1]); |
1870 | 0 | p[2] = Curl_raw_tolower(p[2]); |
1871 | 0 | p += 3; |
1872 | 0 | } |
1873 | 0 | else |
1874 | 0 | p++; |
1875 | 0 | } |
1876 | 0 | } |
1877 | 2.42k | newp = curlx_dyn_ptr(&enc); |
1878 | | |
1879 | 2.42k | if(appendquery && newp) { |
1880 | | /* Append the 'newp' string onto the old query. Add a '&' separator if |
1881 | | none is present at the end of the existing query already */ |
1882 | |
|
1883 | 0 | size_t querylen = u->query ? strlen(u->query) : 0; |
1884 | 0 | bool addamperand = querylen && (u->query[querylen -1] != '&'); |
1885 | 0 | if(querylen) { |
1886 | 0 | struct dynbuf qbuf; |
1887 | 0 | curlx_dyn_init(&qbuf, CURL_MAX_INPUT_LENGTH); |
1888 | |
|
1889 | 0 | if(curlx_dyn_addn(&qbuf, u->query, querylen)) /* add original query */ |
1890 | 0 | goto nomem; |
1891 | | |
1892 | 0 | if(addamperand) { |
1893 | 0 | if(curlx_dyn_addn(&qbuf, "&", 1)) |
1894 | 0 | goto nomem; |
1895 | 0 | } |
1896 | 0 | if(curlx_dyn_add(&qbuf, newp)) |
1897 | 0 | goto nomem; |
1898 | 0 | curlx_dyn_free(&enc); |
1899 | 0 | free(*storep); |
1900 | 0 | *storep = curlx_dyn_ptr(&qbuf); |
1901 | 0 | return CURLUE_OK; |
1902 | 0 | nomem: |
1903 | 0 | curlx_dyn_free(&enc); |
1904 | 0 | return CURLUE_OUT_OF_MEMORY; |
1905 | 0 | } |
1906 | 0 | } |
1907 | | |
1908 | 2.42k | else if(what == CURLUPART_HOST) { |
1909 | 0 | size_t n = curlx_dyn_len(&enc); |
1910 | 0 | if(!n && (flags & CURLU_NO_AUTHORITY)) { |
1911 | | /* Skip hostname check, it is allowed to be empty. */ |
1912 | 0 | } |
1913 | 0 | else { |
1914 | 0 | bool bad = FALSE; |
1915 | 0 | if(!n) |
1916 | 0 | bad = TRUE; /* empty hostname is not okay */ |
1917 | 0 | else if(!urlencode) { |
1918 | | /* if the host name part was not URL encoded here, it was set ready |
1919 | | URL encoded so we need to decode it to check */ |
1920 | 0 | size_t dlen; |
1921 | 0 | char *decoded = NULL; |
1922 | 0 | CURLcode result = |
1923 | 0 | Curl_urldecode(newp, n, &decoded, &dlen, REJECT_CTRL); |
1924 | 0 | if(result || hostname_check(u, decoded, dlen)) |
1925 | 0 | bad = TRUE; |
1926 | 0 | free(decoded); |
1927 | 0 | } |
1928 | 0 | else if(hostname_check(u, (char *)CURL_UNCONST(newp), n)) |
1929 | 0 | bad = TRUE; |
1930 | 0 | if(bad) { |
1931 | 0 | curlx_dyn_free(&enc); |
1932 | 0 | return CURLUE_BAD_HOSTNAME; |
1933 | 0 | } |
1934 | 0 | } |
1935 | 0 | } |
1936 | | |
1937 | 2.42k | free(*storep); |
1938 | 2.42k | *storep = (char *)CURL_UNCONST(newp); |
1939 | 2.42k | } |
1940 | 0 | return CURLUE_OK; |
1941 | 2.42k | } |