1 | /*************************************************************************** |
2 | * _ _ ____ _ |
3 | * Project ___| | | | _ \| | |
4 | * / __| | | | |_) | | |
5 | * | (__| |_| | _ <| |___ |
6 | * \___|\___/|_| \_\_____| |
7 | * |
8 | * Copyright (C) 1998 - 2021, 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 | #include "curl_setup.h" |
24 | |
25 | #include "urldata.h" |
26 | #include "urlapi-int.h" |
27 | #include "strcase.h" |
28 | #include "dotdot.h" |
29 | #include "url.h" |
30 | #include "escape.h" |
31 | #include "curl_ctype.h" |
32 | #include "inet_pton.h" |
33 | |
34 | /* The last 3 #include files should be in this order */ |
35 | #include "curl_printf.h" |
36 | #include "curl_memory.h" |
37 | #include "memdebug.h" |
38 | |
39 | /* MSDOS/Windows style drive prefix, eg c: in c:foo */ |
40 | #define STARTS_WITH_DRIVE_PREFIX(str) \ |
41 | ((('a' <= str[0] && str[0] <= 'z') || \ |
42 | ('A' <= str[0] && str[0] <= 'Z')) && \ |
43 | (str[1] == ':')) |
44 | |
45 | /* MSDOS/Windows style drive prefix, optionally with |
46 | * a '|' instead of ':', followed by a slash or NUL */ |
47 | #define STARTS_WITH_URL_DRIVE_PREFIX(str) \ |
48 | ((('a' <= (str)[0] && (str)[0] <= 'z') || \ |
49 | ('A' <= (str)[0] && (str)[0] <= 'Z')) && \ |
50 | ((str)[1] == ':' || (str)[1] == '|') && \ |
51 | ((str)[2] == '/' || (str)[2] == '\\' || (str)[2] == 0)) |
52 | |
53 | /* Internal representation of CURLU. Point to URL-encoded strings. */ |
54 | struct Curl_URL { |
55 | char *scheme; |
56 | char *user; |
57 | char *password; |
58 | char *options; /* IMAP only? */ |
59 | char *host; |
60 | char *zoneid; /* for numerical IPv6 addresses */ |
61 | char *port; |
62 | char *path; |
63 | char *query; |
64 | char *fragment; |
65 | |
66 | char *scratch; /* temporary scratch area */ |
67 | char *temppath; /* temporary path pointer */ |
68 | long portnum; /* the numerical version */ |
69 | }; |
70 | |
71 | #define DEFAULT_SCHEME "https" |
72 | |
73 | static void free_urlhandle(struct Curl_URL *u) |
74 | { |
75 | free(u->scheme); |
76 | free(u->user); |
77 | free(u->password); |
78 | free(u->options); |
79 | free(u->host); |
80 | free(u->zoneid); |
81 | free(u->port); |
82 | free(u->path); |
83 | free(u->query); |
84 | free(u->fragment); |
85 | free(u->scratch); |
86 | free(u->temppath); |
87 | } |
88 | |
89 | /* move the full contents of one handle onto another and |
90 | free the original */ |
91 | static void mv_urlhandle(struct Curl_URL *from, |
92 | struct Curl_URL *to) |
93 | { |
94 | free_urlhandle(to); |
95 | *to = *from; |
96 | free(from); |
97 | } |
98 | |
99 | /* |
100 | * Find the separator at the end of the host name, or the '?' in cases like |
101 | * http://www.url.com?id=2380 |
102 | */ |
103 | static const char *find_host_sep(const char *url) |
104 | { |
105 | const char *sep; |
106 | const char *query; |
107 | |
108 | /* Find the start of the hostname */ |
109 | sep = strstr(url, "//" ); |
110 | if(!sep) |
111 | sep = url; |
112 | else |
113 | sep += 2; |
114 | |
115 | query = strchr(sep, '?'); |
116 | sep = strchr(sep, '/'); |
117 | |
118 | if(!sep) |
119 | sep = url + strlen(url); |
120 | |
121 | if(!query) |
122 | query = url + strlen(url); |
123 | |
124 | return sep < query ? sep : query; |
125 | } |
126 | |
127 | /* |
128 | * Decide in an encoding-independent manner whether a character in an |
129 | * URL must be escaped. The same criterion must be used in strlen_url() |
130 | * and strcpy_url(). |
131 | */ |
132 | static bool urlchar_needs_escaping(int c) |
133 | { |
134 | return !(ISCNTRL(c) || ISSPACE(c) || ISGRAPH(c)); |
135 | } |
136 | |
137 | /* |
138 | * strlen_url() returns the length of the given URL if the spaces within the |
139 | * URL were properly URL encoded. |
140 | * URL encoding should be skipped for host names, otherwise IDN resolution |
141 | * will fail. |
142 | */ |
143 | static size_t strlen_url(const char *url, bool relative) |
144 | { |
145 | const unsigned char *ptr; |
146 | size_t newlen = 0; |
147 | bool left = TRUE; /* left side of the ? */ |
148 | const unsigned char *host_sep = (const unsigned char *) url; |
149 | |
150 | if(!relative) |
151 | host_sep = (const unsigned char *) find_host_sep(url); |
152 | |
153 | for(ptr = (unsigned char *)url; *ptr; ptr++) { |
154 | |
155 | if(ptr < host_sep) { |
156 | ++newlen; |
157 | continue; |
158 | } |
159 | |
160 | switch(*ptr) { |
161 | case '?': |
162 | left = FALSE; |
163 | /* FALLTHROUGH */ |
164 | default: |
165 | if(urlchar_needs_escaping(*ptr)) |
166 | newlen += 2; |
167 | newlen++; |
168 | break; |
169 | case ' ': |
170 | if(left) |
171 | newlen += 3; |
172 | else |
173 | newlen++; |
174 | break; |
175 | } |
176 | } |
177 | return newlen; |
178 | } |
179 | |
180 | /* strcpy_url() copies a url to a output buffer and URL-encodes the spaces in |
181 | * the source URL accordingly. |
182 | * URL encoding should be skipped for host names, otherwise IDN resolution |
183 | * will fail. |
184 | */ |
185 | static void strcpy_url(char *output, const char *url, bool relative) |
186 | { |
187 | /* we must add this with whitespace-replacing */ |
188 | bool left = TRUE; |
189 | const unsigned char *iptr; |
190 | char *optr = output; |
191 | const unsigned char *host_sep = (const unsigned char *) url; |
192 | |
193 | if(!relative) |
194 | host_sep = (const unsigned char *) find_host_sep(url); |
195 | |
196 | for(iptr = (unsigned char *)url; /* read from here */ |
197 | *iptr; /* until zero byte */ |
198 | iptr++) { |
199 | |
200 | if(iptr < host_sep) { |
201 | *optr++ = *iptr; |
202 | continue; |
203 | } |
204 | |
205 | switch(*iptr) { |
206 | case '?': |
207 | left = FALSE; |
208 | /* FALLTHROUGH */ |
209 | default: |
210 | if(urlchar_needs_escaping(*iptr)) { |
211 | msnprintf(optr, 4, "%%%02x" , *iptr); |
212 | optr += 3; |
213 | } |
214 | else |
215 | *optr++=*iptr; |
216 | break; |
217 | case ' ': |
218 | if(left) { |
219 | *optr++='%'; /* add a '%' */ |
220 | *optr++='2'; /* add a '2' */ |
221 | *optr++='0'; /* add a '0' */ |
222 | } |
223 | else |
224 | *optr++='+'; /* add a '+' here */ |
225 | break; |
226 | } |
227 | } |
228 | *optr = 0; /* null-terminate output buffer */ |
229 | |
230 | } |
231 | |
232 | /* |
233 | * Returns true if the given URL is absolute (as opposed to relative) within |
234 | * the buffer size. Returns the scheme in the buffer if TRUE and 'buf' is |
235 | * non-NULL. |
236 | */ |
237 | bool Curl_is_absolute_url(const char *url, char *buf, size_t buflen) |
238 | { |
239 | size_t i; |
240 | #ifdef WIN32 |
241 | if(STARTS_WITH_DRIVE_PREFIX(url)) |
242 | return FALSE; |
243 | #endif |
244 | for(i = 0; i < buflen && url[i]; ++i) { |
245 | char s = url[i]; |
246 | if((s == ':') && (url[i + 1] == '/')) { |
247 | if(buf) |
248 | buf[i] = 0; |
249 | return TRUE; |
250 | } |
251 | /* RFC 3986 3.1 explains: |
252 | scheme = ALPHA *( ALPHA / DIGIT / "+" / "-" / "." ) |
253 | */ |
254 | else if(ISALNUM(s) || (s == '+') || (s == '-') || (s == '.') ) { |
255 | if(buf) |
256 | buf[i] = (char)TOLOWER(s); |
257 | } |
258 | else |
259 | break; |
260 | } |
261 | return FALSE; |
262 | } |
263 | |
264 | /* |
265 | * Concatenate a relative URL to a base URL making it absolute. |
266 | * URL-encodes any spaces. |
267 | * The returned pointer must be freed by the caller unless NULL |
268 | * (returns NULL on out of memory). |
269 | */ |
270 | static char *concat_url(const char *base, const char *relurl) |
271 | { |
272 | /*** |
273 | TRY to append this new path to the old URL |
274 | to the right of the host part. Oh crap, this is doomed to cause |
275 | problems in the future... |
276 | */ |
277 | char *newest; |
278 | char *protsep; |
279 | char *pathsep; |
280 | size_t newlen; |
281 | bool host_changed = FALSE; |
282 | |
283 | const char *useurl = relurl; |
284 | size_t urllen; |
285 | |
286 | /* we must make our own copy of the URL to play with, as it may |
287 | point to read-only data */ |
288 | char *url_clone = strdup(base); |
289 | |
290 | if(!url_clone) |
291 | return NULL; /* skip out of this NOW */ |
292 | |
293 | /* protsep points to the start of the host name */ |
294 | protsep = strstr(url_clone, "//" ); |
295 | if(!protsep) |
296 | protsep = url_clone; |
297 | else |
298 | protsep += 2; /* pass the slashes */ |
299 | |
300 | if('/' != relurl[0]) { |
301 | int level = 0; |
302 | |
303 | /* First we need to find out if there's a ?-letter in the URL, |
304 | and cut it and the right-side of that off */ |
305 | pathsep = strchr(protsep, '?'); |
306 | if(pathsep) |
307 | *pathsep = 0; |
308 | |
309 | /* we have a relative path to append to the last slash if there's one |
310 | available, or if the new URL is just a query string (starts with a |
311 | '?') we append the new one at the end of the entire currently worked |
312 | out URL */ |
313 | if(useurl[0] != '?') { |
314 | pathsep = strrchr(protsep, '/'); |
315 | if(pathsep) |
316 | *pathsep = 0; |
317 | } |
318 | |
319 | /* Check if there's any slash after the host name, and if so, remember |
320 | that position instead */ |
321 | pathsep = strchr(protsep, '/'); |
322 | if(pathsep) |
323 | protsep = pathsep + 1; |
324 | else |
325 | protsep = NULL; |
326 | |
327 | /* now deal with one "./" or any amount of "../" in the newurl |
328 | and act accordingly */ |
329 | |
330 | if((useurl[0] == '.') && (useurl[1] == '/')) |
331 | useurl += 2; /* just skip the "./" */ |
332 | |
333 | while((useurl[0] == '.') && |
334 | (useurl[1] == '.') && |
335 | (useurl[2] == '/')) { |
336 | level++; |
337 | useurl += 3; /* pass the "../" */ |
338 | } |
339 | |
340 | if(protsep) { |
341 | while(level--) { |
342 | /* cut off one more level from the right of the original URL */ |
343 | pathsep = strrchr(protsep, '/'); |
344 | if(pathsep) |
345 | *pathsep = 0; |
346 | else { |
347 | *protsep = 0; |
348 | break; |
349 | } |
350 | } |
351 | } |
352 | } |
353 | else { |
354 | /* We got a new absolute path for this server */ |
355 | |
356 | if(relurl[1] == '/') { |
357 | /* the new URL starts with //, just keep the protocol part from the |
358 | original one */ |
359 | *protsep = 0; |
360 | useurl = &relurl[2]; /* we keep the slashes from the original, so we |
361 | skip the new ones */ |
362 | host_changed = TRUE; |
363 | } |
364 | else { |
365 | /* cut off the original URL from the first slash, or deal with URLs |
366 | without slash */ |
367 | pathsep = strchr(protsep, '/'); |
368 | if(pathsep) { |
369 | /* When people use badly formatted URLs, such as |
370 | "http://www.url.com?dir=/home/daniel" we must not use the first |
371 | slash, if there's a ?-letter before it! */ |
372 | char *sep = strchr(protsep, '?'); |
373 | if(sep && (sep < pathsep)) |
374 | pathsep = sep; |
375 | *pathsep = 0; |
376 | } |
377 | else { |
378 | /* There was no slash. Now, since we might be operating on a badly |
379 | formatted URL, such as "http://www.url.com?id=2380" which doesn't |
380 | use a slash separator as it is supposed to, we need to check for a |
381 | ?-letter as well! */ |
382 | pathsep = strchr(protsep, '?'); |
383 | if(pathsep) |
384 | *pathsep = 0; |
385 | } |
386 | } |
387 | } |
388 | |
389 | /* If the new part contains a space, this is a mighty stupid redirect |
390 | but we still make an effort to do "right". To the left of a '?' |
391 | letter we replace each space with %20 while it is replaced with '+' |
392 | on the right side of the '?' letter. |
393 | */ |
394 | newlen = strlen_url(useurl, !host_changed); |
395 | |
396 | urllen = strlen(url_clone); |
397 | |
398 | newest = malloc(urllen + 1 + /* possible slash */ |
399 | newlen + 1 /* zero byte */); |
400 | |
401 | if(!newest) { |
402 | free(url_clone); /* don't leak this */ |
403 | return NULL; |
404 | } |
405 | |
406 | /* copy over the root url part */ |
407 | memcpy(newest, url_clone, urllen); |
408 | |
409 | /* check if we need to append a slash */ |
410 | if(('/' == useurl[0]) || (protsep && !*protsep) || ('?' == useurl[0])) |
411 | ; |
412 | else |
413 | newest[urllen++]='/'; |
414 | |
415 | /* then append the new piece on the right side */ |
416 | strcpy_url(&newest[urllen], useurl, !host_changed); |
417 | |
418 | free(url_clone); |
419 | |
420 | return newest; |
421 | } |
422 | |
423 | /* |
424 | * parse_hostname_login() |
425 | * |
426 | * Parse the login details (user name, password and options) from the URL and |
427 | * strip them out of the host name |
428 | * |
429 | */ |
430 | static CURLUcode parse_hostname_login(struct Curl_URL *u, |
431 | char **hostname, |
432 | unsigned int flags) |
433 | { |
434 | CURLUcode result = CURLUE_OK; |
435 | CURLcode ccode; |
436 | char *userp = NULL; |
437 | char *passwdp = NULL; |
438 | char *optionsp = NULL; |
439 | const struct Curl_handler *h = NULL; |
440 | |
441 | /* At this point, we're hoping all the other special cases have |
442 | * been taken care of, so conn->host.name is at most |
443 | * [user[:password][;options]]@]hostname |
444 | * |
445 | * We need somewhere to put the embedded details, so do that first. |
446 | */ |
447 | |
448 | char *ptr = strchr(*hostname, '@'); |
449 | char *login = *hostname; |
450 | |
451 | if(!ptr) |
452 | goto out; |
453 | |
454 | /* We will now try to extract the |
455 | * possible login information in a string like: |
456 | * ftp://user:password@ftp.my.site:8021/README */ |
457 | *hostname = ++ptr; |
458 | |
459 | /* if this is a known scheme, get some details */ |
460 | if(u->scheme) |
461 | h = Curl_builtin_scheme(u->scheme); |
462 | |
463 | /* We could use the login information in the URL so extract it. Only parse |
464 | options if the handler says we should. Note that 'h' might be NULL! */ |
465 | ccode = Curl_parse_login_details(login, ptr - login - 1, |
466 | &userp, &passwdp, |
467 | (h && (h->flags & PROTOPT_URLOPTIONS)) ? |
468 | &optionsp:NULL); |
469 | if(ccode) { |
470 | result = CURLUE_MALFORMED_INPUT; |
471 | goto out; |
472 | } |
473 | |
474 | if(userp) { |
475 | if(flags & CURLU_DISALLOW_USER) { |
476 | /* Option DISALLOW_USER is set and url contains username. */ |
477 | result = CURLUE_USER_NOT_ALLOWED; |
478 | goto out; |
479 | } |
480 | |
481 | u->user = userp; |
482 | } |
483 | |
484 | if(passwdp) |
485 | u->password = passwdp; |
486 | |
487 | if(optionsp) |
488 | u->options = optionsp; |
489 | |
490 | return CURLUE_OK; |
491 | out: |
492 | |
493 | free(userp); |
494 | free(passwdp); |
495 | free(optionsp); |
496 | |
497 | return result; |
498 | } |
499 | |
500 | UNITTEST CURLUcode Curl_parse_port(struct Curl_URL *u, char *hostname, |
501 | bool has_scheme) |
502 | { |
503 | char *portptr = NULL; |
504 | char endbracket; |
505 | int len; |
506 | |
507 | /* |
508 | * Find the end of an IPv6 address, either on the ']' ending bracket or |
509 | * a percent-encoded zone index. |
510 | */ |
511 | if(1 == sscanf(hostname, "[%*45[0123456789abcdefABCDEF:.]%c%n" , |
512 | &endbracket, &len)) { |
513 | if(']' == endbracket) |
514 | portptr = &hostname[len]; |
515 | else if('%' == endbracket) { |
516 | int zonelen = len; |
517 | if(1 == sscanf(hostname + zonelen, "%*[^]]%c%n" , &endbracket, &len)) { |
518 | if(']' != endbracket) |
519 | return CURLUE_MALFORMED_INPUT; |
520 | portptr = &hostname[--zonelen + len + 1]; |
521 | } |
522 | else |
523 | return CURLUE_MALFORMED_INPUT; |
524 | } |
525 | else |
526 | return CURLUE_MALFORMED_INPUT; |
527 | |
528 | /* this is a RFC2732-style specified IP-address */ |
529 | if(portptr && *portptr) { |
530 | if(*portptr != ':') |
531 | return CURLUE_MALFORMED_INPUT; |
532 | } |
533 | else |
534 | portptr = NULL; |
535 | } |
536 | else |
537 | portptr = strchr(hostname, ':'); |
538 | |
539 | if(portptr) { |
540 | char *rest; |
541 | long port; |
542 | char portbuf[7]; |
543 | |
544 | /* Browser behavior adaptation. If there's a colon with no digits after, |
545 | just cut off the name there which makes us ignore the colon and just |
546 | use the default port. Firefox, Chrome and Safari all do that. |
547 | |
548 | Don't do it if the URL has no scheme, to make something that looks like |
549 | a scheme not work! |
550 | */ |
551 | if(!portptr[1]) { |
552 | *portptr = '\0'; |
553 | return has_scheme ? CURLUE_OK : CURLUE_BAD_PORT_NUMBER; |
554 | } |
555 | |
556 | if(!ISDIGIT(portptr[1])) |
557 | return CURLUE_BAD_PORT_NUMBER; |
558 | |
559 | port = strtol(portptr + 1, &rest, 10); /* Port number must be decimal */ |
560 | |
561 | if((port <= 0) || (port > 0xffff)) |
562 | /* Single unix standard says port numbers are 16 bits long, but we don't |
563 | treat port zero as OK. */ |
564 | return CURLUE_BAD_PORT_NUMBER; |
565 | |
566 | if(rest[0]) |
567 | return CURLUE_BAD_PORT_NUMBER; |
568 | |
569 | *portptr++ = '\0'; /* cut off the name there */ |
570 | *rest = 0; |
571 | /* generate a new port number string to get rid of leading zeroes etc */ |
572 | msnprintf(portbuf, sizeof(portbuf), "%ld" , port); |
573 | u->portnum = port; |
574 | u->port = strdup(portbuf); |
575 | if(!u->port) |
576 | return CURLUE_OUT_OF_MEMORY; |
577 | } |
578 | |
579 | return CURLUE_OK; |
580 | } |
581 | |
582 | /* scan for byte values < 31 or 127 */ |
583 | static bool junkscan(const char *part, unsigned int flags) |
584 | { |
585 | if(part) { |
586 | static const char badbytes[]={ |
587 | /* */ 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, |
588 | 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, |
589 | 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, |
590 | 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, |
591 | 0x7f, 0x00 /* null-terminate */ |
592 | }; |
593 | size_t n = strlen(part); |
594 | size_t nfine = strcspn(part, badbytes); |
595 | if(nfine != n) |
596 | /* since we don't know which part is scanned, return a generic error |
597 | code */ |
598 | return TRUE; |
599 | if(!(flags & CURLU_ALLOW_SPACE) && strchr(part, ' ')) |
600 | return TRUE; |
601 | } |
602 | return FALSE; |
603 | } |
604 | |
605 | static CURLUcode hostname_check(struct Curl_URL *u, char *hostname) |
606 | { |
607 | size_t len; |
608 | size_t hlen = strlen(hostname); |
609 | |
610 | if(hostname[0] == '[') { |
611 | #ifdef ENABLE_IPV6 |
612 | char dest[16]; /* fits a binary IPv6 address */ |
613 | #endif |
614 | const char *l = "0123456789abcdefABCDEF:." ; |
615 | if(hlen < 4) /* '[::]' is the shortest possible valid string */ |
616 | return CURLUE_MALFORMED_INPUT; |
617 | hostname++; |
618 | hlen -= 2; |
619 | |
620 | if(hostname[hlen] != ']') |
621 | return CURLUE_MALFORMED_INPUT; |
622 | |
623 | /* only valid letters are ok */ |
624 | len = strspn(hostname, l); |
625 | if(hlen != len) { |
626 | hlen = len; |
627 | if(hostname[len] == '%') { |
628 | /* this could now be '%[zone id]' */ |
629 | char zoneid[16]; |
630 | int i = 0; |
631 | char *h = &hostname[len + 1]; |
632 | /* pass '25' if present and is a url encoded percent sign */ |
633 | if(!strncmp(h, "25" , 2) && h[2] && (h[2] != ']')) |
634 | h += 2; |
635 | while(*h && (*h != ']') && (i < 15)) |
636 | zoneid[i++] = *h++; |
637 | if(!i || (']' != *h)) |
638 | return CURLUE_MALFORMED_INPUT; |
639 | zoneid[i] = 0; |
640 | u->zoneid = strdup(zoneid); |
641 | if(!u->zoneid) |
642 | return CURLUE_OUT_OF_MEMORY; |
643 | hostname[len] = ']'; /* insert end bracket */ |
644 | hostname[len + 1] = 0; /* terminate the hostname */ |
645 | } |
646 | else |
647 | return CURLUE_MALFORMED_INPUT; |
648 | /* hostname is fine */ |
649 | } |
650 | #ifdef ENABLE_IPV6 |
651 | hostname[hlen] = 0; /* end the address there */ |
652 | if(1 != Curl_inet_pton(AF_INET6, hostname, dest)) |
653 | return CURLUE_MALFORMED_INPUT; |
654 | hostname[hlen] = ']'; /* restore ending bracket */ |
655 | #endif |
656 | } |
657 | else { |
658 | /* letters from the second string is not ok */ |
659 | len = strcspn(hostname, " " ); |
660 | if(hlen != len) |
661 | /* hostname with bad content */ |
662 | return CURLUE_MALFORMED_INPUT; |
663 | } |
664 | if(!hostname[0]) |
665 | return CURLUE_NO_HOST; |
666 | return CURLUE_OK; |
667 | } |
668 | |
669 | #define HOSTNAME_END(x) (((x) == '/') || ((x) == '?') || ((x) == '#')) |
670 | |
671 | /* |
672 | * Handle partial IPv4 numerical addresses and different bases, like |
673 | * '16843009', '0x7f', '0x7f.1' '0177.1.1.1' etc. |
674 | * |
675 | * If the given input string is syntactically wrong or any part for example is |
676 | * too big, this function returns FALSE and doesn't create any output. |
677 | * |
678 | * Output the "normalized" version of that input string in plain quad decimal |
679 | * integers and return TRUE. |
680 | */ |
681 | static bool ipv4_normalize(const char *hostname, char *outp, size_t olen) |
682 | { |
683 | bool done = FALSE; |
684 | int n = 0; |
685 | const char *c = hostname; |
686 | unsigned long parts[4] = {0, 0, 0, 0}; |
687 | |
688 | while(!done) { |
689 | char *endp; |
690 | unsigned long l; |
691 | if((*c < '0') || (*c > '9')) |
692 | /* most importantly this doesn't allow a leading plus or minus */ |
693 | return FALSE; |
694 | l = strtoul(c, &endp, 0); |
695 | |
696 | /* overflow or nothing parsed at all */ |
697 | if(((l == ULONG_MAX) && (errno == ERANGE)) || (endp == c)) |
698 | return FALSE; |
699 | |
700 | #if SIZEOF_LONG > 4 |
701 | /* a value larger than 32 bits */ |
702 | if(l > UINT_MAX) |
703 | return FALSE; |
704 | #endif |
705 | |
706 | parts[n] = l; |
707 | c = endp; |
708 | |
709 | switch (*c) { |
710 | case '.' : |
711 | if(n == 3) |
712 | return FALSE; |
713 | n++; |
714 | c++; |
715 | break; |
716 | |
717 | case '\0': |
718 | done = TRUE; |
719 | break; |
720 | |
721 | default: |
722 | return FALSE; |
723 | } |
724 | } |
725 | |
726 | /* this is deemed a valid IPv4 numerical address */ |
727 | |
728 | switch(n) { |
729 | case 0: /* a -- 32 bits */ |
730 | msnprintf(outp, olen, "%u.%u.%u.%u" , |
731 | parts[0] >> 24, (parts[0] >> 16) & 0xff, |
732 | (parts[0] >> 8) & 0xff, parts[0] & 0xff); |
733 | break; |
734 | case 1: /* a.b -- 8.24 bits */ |
735 | if((parts[0] > 0xff) || (parts[1] > 0xffffff)) |
736 | return FALSE; |
737 | msnprintf(outp, olen, "%u.%u.%u.%u" , |
738 | parts[0], (parts[1] >> 16) & 0xff, |
739 | (parts[1] >> 8) & 0xff, parts[1] & 0xff); |
740 | break; |
741 | case 2: /* a.b.c -- 8.8.16 bits */ |
742 | if((parts[0] > 0xff) || (parts[1] > 0xff) || (parts[2] > 0xffff)) |
743 | return FALSE; |
744 | msnprintf(outp, olen, "%u.%u.%u.%u" , |
745 | parts[0], parts[1], (parts[2] >> 8) & 0xff, |
746 | parts[2] & 0xff); |
747 | break; |
748 | case 3: /* a.b.c.d -- 8.8.8.8 bits */ |
749 | if((parts[0] > 0xff) || (parts[1] > 0xff) || (parts[2] > 0xff) || |
750 | (parts[3] > 0xff)) |
751 | return FALSE; |
752 | msnprintf(outp, olen, "%u.%u.%u.%u" , |
753 | parts[0], parts[1], parts[2], parts[3]); |
754 | break; |
755 | } |
756 | return TRUE; |
757 | } |
758 | |
759 | static CURLUcode seturl(const char *url, CURLU *u, unsigned int flags) |
760 | { |
761 | char *path; |
762 | bool path_alloced = FALSE; |
763 | char *hostname; |
764 | char *query = NULL; |
765 | char *fragment = NULL; |
766 | CURLUcode result; |
767 | bool url_has_scheme = FALSE; |
768 | char schemebuf[MAX_SCHEME_LEN + 1]; |
769 | const char *schemep = NULL; |
770 | size_t schemelen = 0; |
771 | size_t urllen; |
772 | |
773 | DEBUGASSERT(url); |
774 | |
775 | /************************************************************* |
776 | * Parse the URL. |
777 | ************************************************************/ |
778 | /* allocate scratch area */ |
779 | urllen = strlen(url); |
780 | if(urllen > CURL_MAX_INPUT_LENGTH) |
781 | /* excessive input length */ |
782 | return CURLUE_MALFORMED_INPUT; |
783 | |
784 | path = u->scratch = malloc(urllen * 2 + 2); |
785 | if(!path) |
786 | return CURLUE_OUT_OF_MEMORY; |
787 | |
788 | hostname = &path[urllen + 1]; |
789 | hostname[0] = 0; |
790 | |
791 | if(Curl_is_absolute_url(url, schemebuf, sizeof(schemebuf))) { |
792 | url_has_scheme = TRUE; |
793 | schemelen = strlen(schemebuf); |
794 | } |
795 | |
796 | /* handle the file: scheme */ |
797 | if(url_has_scheme && strcasecompare(schemebuf, "file" )) { |
798 | /* path has been allocated large enough to hold this */ |
799 | strcpy(path, &url[5]); |
800 | |
801 | hostname = NULL; /* no host for file: URLs */ |
802 | u->scheme = strdup("file" ); |
803 | if(!u->scheme) |
804 | return CURLUE_OUT_OF_MEMORY; |
805 | |
806 | /* Extra handling URLs with an authority component (i.e. that start with |
807 | * "file://") |
808 | * |
809 | * We allow omitted hostname (e.g. file:/<path>) -- valid according to |
810 | * RFC 8089, but not the (current) WHAT-WG URL spec. |
811 | */ |
812 | if(path[0] == '/' && path[1] == '/') { |
813 | /* swallow the two slashes */ |
814 | char *ptr = &path[2]; |
815 | |
816 | /* |
817 | * According to RFC 8089, a file: URL can be reliably dereferenced if: |
818 | * |
819 | * o it has no/blank hostname, or |
820 | * |
821 | * o the hostname matches "localhost" (case-insensitively), or |
822 | * |
823 | * o the hostname is a FQDN that resolves to this machine. |
824 | * |
825 | * For brevity, we only consider URLs with empty, "localhost", or |
826 | * "127.0.0.1" hostnames as local. |
827 | * |
828 | * Additionally, there is an exception for URLs with a Windows drive |
829 | * letter in the authority (which was accidentally omitted from RFC 8089 |
830 | * Appendix E, but believe me, it was meant to be there. --MK) |
831 | */ |
832 | if(ptr[0] != '/' && !STARTS_WITH_URL_DRIVE_PREFIX(ptr)) { |
833 | /* the URL includes a host name, it must match "localhost" or |
834 | "127.0.0.1" to be valid */ |
835 | if(!checkprefix("localhost/" , ptr) && |
836 | !checkprefix("127.0.0.1/" , ptr)) { |
837 | /* Invalid file://hostname/, expected localhost or 127.0.0.1 or |
838 | none */ |
839 | return CURLUE_MALFORMED_INPUT; |
840 | } |
841 | ptr += 9; /* now points to the slash after the host */ |
842 | } |
843 | |
844 | path = ptr; |
845 | } |
846 | |
847 | #if !defined(MSDOS) && !defined(WIN32) && !defined(__CYGWIN__) |
848 | /* Don't allow Windows drive letters when not in Windows. |
849 | * This catches both "file:/c:" and "file:c:" */ |
850 | if(('/' == path[0] && STARTS_WITH_URL_DRIVE_PREFIX(&path[1])) || |
851 | STARTS_WITH_URL_DRIVE_PREFIX(path)) { |
852 | /* File drive letters are only accepted in MSDOS/Windows */ |
853 | return CURLUE_MALFORMED_INPUT; |
854 | } |
855 | #else |
856 | /* If the path starts with a slash and a drive letter, ditch the slash */ |
857 | if('/' == path[0] && STARTS_WITH_URL_DRIVE_PREFIX(&path[1])) { |
858 | /* This cannot be done with strcpy, as the memory chunks overlap! */ |
859 | memmove(path, &path[1], strlen(&path[1]) + 1); |
860 | } |
861 | #endif |
862 | |
863 | } |
864 | else { |
865 | /* clear path */ |
866 | const char *p; |
867 | const char *hostp; |
868 | size_t len; |
869 | path[0] = 0; |
870 | |
871 | if(url_has_scheme) { |
872 | int i = 0; |
873 | p = &url[schemelen + 1]; |
874 | while(p && (*p == '/') && (i < 4)) { |
875 | p++; |
876 | i++; |
877 | } |
878 | if((i < 1) || (i>3)) |
879 | /* less than one or more than three slashes */ |
880 | return CURLUE_MALFORMED_INPUT; |
881 | |
882 | schemep = schemebuf; |
883 | if(!Curl_builtin_scheme(schemep) && |
884 | !(flags & CURLU_NON_SUPPORT_SCHEME)) |
885 | return CURLUE_UNSUPPORTED_SCHEME; |
886 | |
887 | if(junkscan(schemep, flags)) |
888 | return CURLUE_MALFORMED_INPUT; |
889 | } |
890 | else { |
891 | /* no scheme! */ |
892 | |
893 | if(!(flags & (CURLU_DEFAULT_SCHEME|CURLU_GUESS_SCHEME))) |
894 | return CURLUE_MALFORMED_INPUT; |
895 | if(flags & CURLU_DEFAULT_SCHEME) |
896 | schemep = DEFAULT_SCHEME; |
897 | |
898 | /* |
899 | * The URL was badly formatted, let's try without scheme specified. |
900 | */ |
901 | p = url; |
902 | } |
903 | hostp = p; /* host name starts here */ |
904 | |
905 | while(*p && !HOSTNAME_END(*p)) /* find end of host name */ |
906 | p++; |
907 | |
908 | len = p - hostp; |
909 | if(len) { |
910 | memcpy(hostname, hostp, len); |
911 | hostname[len] = 0; |
912 | } |
913 | else { |
914 | if(!(flags & CURLU_NO_AUTHORITY)) |
915 | return CURLUE_MALFORMED_INPUT; |
916 | } |
917 | |
918 | len = strlen(p); |
919 | memcpy(path, p, len); |
920 | path[len] = 0; |
921 | |
922 | if(schemep) { |
923 | u->scheme = strdup(schemep); |
924 | if(!u->scheme) |
925 | return CURLUE_OUT_OF_MEMORY; |
926 | } |
927 | } |
928 | |
929 | if(junkscan(path, flags)) |
930 | return CURLUE_MALFORMED_INPUT; |
931 | |
932 | if((flags & CURLU_URLENCODE) && path[0]) { |
933 | /* worst case output length is 3x the original! */ |
934 | char *newp = malloc(strlen(path) * 3); |
935 | if(!newp) |
936 | return CURLUE_OUT_OF_MEMORY; |
937 | path_alloced = TRUE; |
938 | strcpy_url(newp, path, TRUE); /* consider it relative */ |
939 | u->temppath = path = newp; |
940 | } |
941 | |
942 | fragment = strchr(path, '#'); |
943 | if(fragment) { |
944 | *fragment++ = 0; |
945 | if(fragment[0]) { |
946 | u->fragment = strdup(fragment); |
947 | if(!u->fragment) |
948 | return CURLUE_OUT_OF_MEMORY; |
949 | } |
950 | } |
951 | |
952 | query = strchr(path, '?'); |
953 | if(query) { |
954 | *query++ = 0; |
955 | /* done even if the query part is a blank string */ |
956 | u->query = strdup(query); |
957 | if(!u->query) |
958 | return CURLUE_OUT_OF_MEMORY; |
959 | } |
960 | |
961 | if(!path[0]) |
962 | /* if there's no path left set, unset */ |
963 | path = NULL; |
964 | else { |
965 | if(!(flags & CURLU_PATH_AS_IS)) { |
966 | /* remove ../ and ./ sequences according to RFC3986 */ |
967 | char *newp = Curl_dedotdotify(path); |
968 | if(!newp) |
969 | return CURLUE_OUT_OF_MEMORY; |
970 | |
971 | if(strcmp(newp, path)) { |
972 | /* if we got a new version */ |
973 | if(path_alloced) |
974 | Curl_safefree(u->temppath); |
975 | u->temppath = path = newp; |
976 | path_alloced = TRUE; |
977 | } |
978 | else |
979 | free(newp); |
980 | } |
981 | |
982 | u->path = path_alloced?path:strdup(path); |
983 | if(!u->path) |
984 | return CURLUE_OUT_OF_MEMORY; |
985 | u->temppath = NULL; /* used now */ |
986 | } |
987 | |
988 | if(hostname) { |
989 | char normalized_ipv4[sizeof("255.255.255.255" ) + 1]; |
990 | /* |
991 | * Parse the login details and strip them out of the host name. |
992 | */ |
993 | if(junkscan(hostname, flags)) |
994 | return CURLUE_MALFORMED_INPUT; |
995 | |
996 | result = parse_hostname_login(u, &hostname, flags); |
997 | if(result) |
998 | return result; |
999 | |
1000 | result = Curl_parse_port(u, hostname, url_has_scheme); |
1001 | if(result) |
1002 | return result; |
1003 | |
1004 | if(0 == strlen(hostname) && (flags & CURLU_NO_AUTHORITY)) { |
1005 | /* Skip hostname check, it's allowed to be empty. */ |
1006 | } |
1007 | else { |
1008 | result = hostname_check(u, hostname); |
1009 | if(result) |
1010 | return result; |
1011 | } |
1012 | |
1013 | if(ipv4_normalize(hostname, normalized_ipv4, sizeof(normalized_ipv4))) |
1014 | u->host = strdup(normalized_ipv4); |
1015 | else |
1016 | u->host = strdup(hostname); |
1017 | if(!u->host) |
1018 | return CURLUE_OUT_OF_MEMORY; |
1019 | |
1020 | if((flags & CURLU_GUESS_SCHEME) && !schemep) { |
1021 | /* legacy curl-style guess based on host name */ |
1022 | if(checkprefix("ftp." , hostname)) |
1023 | schemep = "ftp" ; |
1024 | else if(checkprefix("dict." , hostname)) |
1025 | schemep = "dict" ; |
1026 | else if(checkprefix("ldap." , hostname)) |
1027 | schemep = "ldap" ; |
1028 | else if(checkprefix("imap." , hostname)) |
1029 | schemep = "imap" ; |
1030 | else if(checkprefix("smtp." , hostname)) |
1031 | schemep = "smtp" ; |
1032 | else if(checkprefix("pop3." , hostname)) |
1033 | schemep = "pop3" ; |
1034 | else |
1035 | schemep = "http" ; |
1036 | |
1037 | u->scheme = strdup(schemep); |
1038 | if(!u->scheme) |
1039 | return CURLUE_OUT_OF_MEMORY; |
1040 | } |
1041 | } |
1042 | |
1043 | Curl_safefree(u->scratch); |
1044 | Curl_safefree(u->temppath); |
1045 | |
1046 | return CURLUE_OK; |
1047 | } |
1048 | |
1049 | /* |
1050 | * Parse the URL and set the relevant members of the Curl_URL struct. |
1051 | */ |
1052 | static CURLUcode parseurl(const char *url, CURLU *u, unsigned int flags) |
1053 | { |
1054 | CURLUcode result = seturl(url, u, flags); |
1055 | if(result) { |
1056 | free_urlhandle(u); |
1057 | memset(u, 0, sizeof(struct Curl_URL)); |
1058 | } |
1059 | return result; |
1060 | } |
1061 | |
1062 | /* |
1063 | */ |
1064 | CURLU *curl_url(void) |
1065 | { |
1066 | return calloc(sizeof(struct Curl_URL), 1); |
1067 | } |
1068 | |
1069 | void curl_url_cleanup(CURLU *u) |
1070 | { |
1071 | if(u) { |
1072 | free_urlhandle(u); |
1073 | free(u); |
1074 | } |
1075 | } |
1076 | |
1077 | #define DUP(dest, src, name) \ |
1078 | do { \ |
1079 | if(src->name) { \ |
1080 | dest->name = strdup(src->name); \ |
1081 | if(!dest->name) \ |
1082 | goto fail; \ |
1083 | } \ |
1084 | } while(0) |
1085 | |
1086 | CURLU *curl_url_dup(CURLU *in) |
1087 | { |
1088 | struct Curl_URL *u = calloc(sizeof(struct Curl_URL), 1); |
1089 | if(u) { |
1090 | DUP(u, in, scheme); |
1091 | DUP(u, in, user); |
1092 | DUP(u, in, password); |
1093 | DUP(u, in, options); |
1094 | DUP(u, in, host); |
1095 | DUP(u, in, port); |
1096 | DUP(u, in, path); |
1097 | DUP(u, in, query); |
1098 | DUP(u, in, fragment); |
1099 | u->portnum = in->portnum; |
1100 | } |
1101 | return u; |
1102 | fail: |
1103 | curl_url_cleanup(u); |
1104 | return NULL; |
1105 | } |
1106 | |
1107 | CURLUcode curl_url_get(CURLU *u, CURLUPart what, |
1108 | char **part, unsigned int flags) |
1109 | { |
1110 | char *ptr; |
1111 | CURLUcode ifmissing = CURLUE_UNKNOWN_PART; |
1112 | char portbuf[7]; |
1113 | bool urldecode = (flags & CURLU_URLDECODE)?1:0; |
1114 | bool plusdecode = FALSE; |
1115 | (void)flags; |
1116 | if(!u) |
1117 | return CURLUE_BAD_HANDLE; |
1118 | if(!part) |
1119 | return CURLUE_BAD_PARTPOINTER; |
1120 | *part = NULL; |
1121 | |
1122 | switch(what) { |
1123 | case CURLUPART_SCHEME: |
1124 | ptr = u->scheme; |
1125 | ifmissing = CURLUE_NO_SCHEME; |
1126 | urldecode = FALSE; /* never for schemes */ |
1127 | break; |
1128 | case CURLUPART_USER: |
1129 | ptr = u->user; |
1130 | ifmissing = CURLUE_NO_USER; |
1131 | break; |
1132 | case CURLUPART_PASSWORD: |
1133 | ptr = u->password; |
1134 | ifmissing = CURLUE_NO_PASSWORD; |
1135 | break; |
1136 | case CURLUPART_OPTIONS: |
1137 | ptr = u->options; |
1138 | ifmissing = CURLUE_NO_OPTIONS; |
1139 | break; |
1140 | case CURLUPART_HOST: |
1141 | ptr = u->host; |
1142 | ifmissing = CURLUE_NO_HOST; |
1143 | break; |
1144 | case CURLUPART_ZONEID: |
1145 | ptr = u->zoneid; |
1146 | break; |
1147 | case CURLUPART_PORT: |
1148 | ptr = u->port; |
1149 | ifmissing = CURLUE_NO_PORT; |
1150 | urldecode = FALSE; /* never for port */ |
1151 | if(!ptr && (flags & CURLU_DEFAULT_PORT) && u->scheme) { |
1152 | /* there's no stored port number, but asked to deliver |
1153 | a default one for the scheme */ |
1154 | const struct Curl_handler *h = |
1155 | Curl_builtin_scheme(u->scheme); |
1156 | if(h) { |
1157 | msnprintf(portbuf, sizeof(portbuf), "%u" , h->defport); |
1158 | ptr = portbuf; |
1159 | } |
1160 | } |
1161 | else if(ptr && u->scheme) { |
1162 | /* there is a stored port number, but ask to inhibit if |
1163 | it matches the default one for the scheme */ |
1164 | const struct Curl_handler *h = |
1165 | Curl_builtin_scheme(u->scheme); |
1166 | if(h && (h->defport == u->portnum) && |
1167 | (flags & CURLU_NO_DEFAULT_PORT)) |
1168 | ptr = NULL; |
1169 | } |
1170 | break; |
1171 | case CURLUPART_PATH: |
1172 | ptr = u->path; |
1173 | if(!ptr) { |
1174 | ptr = u->path = strdup("/" ); |
1175 | if(!u->path) |
1176 | return CURLUE_OUT_OF_MEMORY; |
1177 | } |
1178 | break; |
1179 | case CURLUPART_QUERY: |
1180 | ptr = u->query; |
1181 | ifmissing = CURLUE_NO_QUERY; |
1182 | plusdecode = urldecode; |
1183 | break; |
1184 | case CURLUPART_FRAGMENT: |
1185 | ptr = u->fragment; |
1186 | ifmissing = CURLUE_NO_FRAGMENT; |
1187 | break; |
1188 | case CURLUPART_URL: { |
1189 | char *url; |
1190 | char *scheme; |
1191 | char *options = u->options; |
1192 | char *port = u->port; |
1193 | char *allochost = NULL; |
1194 | if(u->scheme && strcasecompare("file" , u->scheme)) { |
1195 | url = aprintf("file://%s%s%s" , |
1196 | u->path, |
1197 | u->fragment? "#" : "" , |
1198 | u->fragment? u->fragment : "" ); |
1199 | } |
1200 | else if(!u->host) |
1201 | return CURLUE_NO_HOST; |
1202 | else { |
1203 | const struct Curl_handler *h = NULL; |
1204 | if(u->scheme) |
1205 | scheme = u->scheme; |
1206 | else if(flags & CURLU_DEFAULT_SCHEME) |
1207 | scheme = (char *) DEFAULT_SCHEME; |
1208 | else |
1209 | return CURLUE_NO_SCHEME; |
1210 | |
1211 | h = Curl_builtin_scheme(scheme); |
1212 | if(!port && (flags & CURLU_DEFAULT_PORT)) { |
1213 | /* there's no stored port number, but asked to deliver |
1214 | a default one for the scheme */ |
1215 | if(h) { |
1216 | msnprintf(portbuf, sizeof(portbuf), "%u" , h->defport); |
1217 | port = portbuf; |
1218 | } |
1219 | } |
1220 | else if(port) { |
1221 | /* there is a stored port number, but asked to inhibit if it matches |
1222 | the default one for the scheme */ |
1223 | if(h && (h->defport == u->portnum) && |
1224 | (flags & CURLU_NO_DEFAULT_PORT)) |
1225 | port = NULL; |
1226 | } |
1227 | |
1228 | if(h && !(h->flags & PROTOPT_URLOPTIONS)) |
1229 | options = NULL; |
1230 | |
1231 | if((u->host[0] == '[') && u->zoneid) { |
1232 | /* make it '[ host %25 zoneid ]' */ |
1233 | size_t hostlen = strlen(u->host); |
1234 | size_t alen = hostlen + 3 + strlen(u->zoneid) + 1; |
1235 | allochost = malloc(alen); |
1236 | if(!allochost) |
1237 | return CURLUE_OUT_OF_MEMORY; |
1238 | memcpy(allochost, u->host, hostlen - 1); |
1239 | msnprintf(&allochost[hostlen - 1], alen - hostlen + 1, |
1240 | "%%25%s]" , u->zoneid); |
1241 | } |
1242 | |
1243 | url = aprintf("%s://%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s" , |
1244 | scheme, |
1245 | u->user ? u->user : "" , |
1246 | u->password ? ":" : "" , |
1247 | u->password ? u->password : "" , |
1248 | options ? ";" : "" , |
1249 | options ? options : "" , |
1250 | (u->user || u->password || options) ? "@" : "" , |
1251 | allochost ? allochost : u->host, |
1252 | port ? ":" : "" , |
1253 | port ? port : "" , |
1254 | (u->path && (u->path[0] != '/')) ? "/" : "" , |
1255 | u->path ? u->path : "/" , |
1256 | (u->query && u->query[0]) ? "?" : "" , |
1257 | (u->query && u->query[0]) ? u->query : "" , |
1258 | u->fragment? "#" : "" , |
1259 | u->fragment? u->fragment : "" ); |
1260 | free(allochost); |
1261 | } |
1262 | if(!url) |
1263 | return CURLUE_OUT_OF_MEMORY; |
1264 | *part = url; |
1265 | return CURLUE_OK; |
1266 | } |
1267 | default: |
1268 | ptr = NULL; |
1269 | break; |
1270 | } |
1271 | if(ptr) { |
1272 | *part = strdup(ptr); |
1273 | if(!*part) |
1274 | return CURLUE_OUT_OF_MEMORY; |
1275 | if(plusdecode) { |
1276 | /* convert + to space */ |
1277 | char *plus; |
1278 | for(plus = *part; *plus; ++plus) { |
1279 | if(*plus == '+') |
1280 | *plus = ' '; |
1281 | } |
1282 | } |
1283 | if(urldecode) { |
1284 | char *decoded; |
1285 | size_t dlen; |
1286 | /* this unconditional rejection of control bytes is documented |
1287 | API behavior */ |
1288 | CURLcode res = Curl_urldecode(NULL, *part, 0, &decoded, &dlen, |
1289 | REJECT_CTRL); |
1290 | free(*part); |
1291 | if(res) { |
1292 | *part = NULL; |
1293 | return CURLUE_URLDECODE; |
1294 | } |
1295 | *part = decoded; |
1296 | } |
1297 | return CURLUE_OK; |
1298 | } |
1299 | else |
1300 | return ifmissing; |
1301 | } |
1302 | |
1303 | CURLUcode curl_url_set(CURLU *u, CURLUPart what, |
1304 | const char *part, unsigned int flags) |
1305 | { |
1306 | char **storep = NULL; |
1307 | long port = 0; |
1308 | bool urlencode = (flags & CURLU_URLENCODE)? 1 : 0; |
1309 | bool plusencode = FALSE; |
1310 | bool urlskipslash = FALSE; |
1311 | bool appendquery = FALSE; |
1312 | bool equalsencode = FALSE; |
1313 | |
1314 | if(!u) |
1315 | return CURLUE_BAD_HANDLE; |
1316 | if(!part) { |
1317 | /* setting a part to NULL clears it */ |
1318 | switch(what) { |
1319 | case CURLUPART_URL: |
1320 | break; |
1321 | case CURLUPART_SCHEME: |
1322 | storep = &u->scheme; |
1323 | break; |
1324 | case CURLUPART_USER: |
1325 | storep = &u->user; |
1326 | break; |
1327 | case CURLUPART_PASSWORD: |
1328 | storep = &u->password; |
1329 | break; |
1330 | case CURLUPART_OPTIONS: |
1331 | storep = &u->options; |
1332 | break; |
1333 | case CURLUPART_HOST: |
1334 | storep = &u->host; |
1335 | break; |
1336 | case CURLUPART_ZONEID: |
1337 | storep = &u->zoneid; |
1338 | break; |
1339 | case CURLUPART_PORT: |
1340 | u->portnum = 0; |
1341 | storep = &u->port; |
1342 | break; |
1343 | case CURLUPART_PATH: |
1344 | storep = &u->path; |
1345 | break; |
1346 | case CURLUPART_QUERY: |
1347 | storep = &u->query; |
1348 | break; |
1349 | case CURLUPART_FRAGMENT: |
1350 | storep = &u->fragment; |
1351 | break; |
1352 | default: |
1353 | return CURLUE_UNKNOWN_PART; |
1354 | } |
1355 | if(storep && *storep) { |
1356 | Curl_safefree(*storep); |
1357 | } |
1358 | return CURLUE_OK; |
1359 | } |
1360 | |
1361 | switch(what) { |
1362 | case CURLUPART_SCHEME: |
1363 | if(strlen(part) > MAX_SCHEME_LEN) |
1364 | /* too long */ |
1365 | return CURLUE_MALFORMED_INPUT; |
1366 | if(!(flags & CURLU_NON_SUPPORT_SCHEME) && |
1367 | /* verify that it is a fine scheme */ |
1368 | !Curl_builtin_scheme(part)) |
1369 | return CURLUE_UNSUPPORTED_SCHEME; |
1370 | storep = &u->scheme; |
1371 | urlencode = FALSE; /* never */ |
1372 | break; |
1373 | case CURLUPART_USER: |
1374 | storep = &u->user; |
1375 | break; |
1376 | case CURLUPART_PASSWORD: |
1377 | storep = &u->password; |
1378 | break; |
1379 | case CURLUPART_OPTIONS: |
1380 | storep = &u->options; |
1381 | break; |
1382 | case CURLUPART_HOST: |
1383 | storep = &u->host; |
1384 | Curl_safefree(u->zoneid); |
1385 | break; |
1386 | case CURLUPART_ZONEID: |
1387 | storep = &u->zoneid; |
1388 | break; |
1389 | case CURLUPART_PORT: |
1390 | { |
1391 | char *endp; |
1392 | urlencode = FALSE; /* never */ |
1393 | port = strtol(part, &endp, 10); /* Port number must be decimal */ |
1394 | if((port <= 0) || (port > 0xffff)) |
1395 | return CURLUE_BAD_PORT_NUMBER; |
1396 | if(*endp) |
1397 | /* weirdly provided number, not good! */ |
1398 | return CURLUE_MALFORMED_INPUT; |
1399 | storep = &u->port; |
1400 | } |
1401 | break; |
1402 | case CURLUPART_PATH: |
1403 | urlskipslash = TRUE; |
1404 | storep = &u->path; |
1405 | break; |
1406 | case CURLUPART_QUERY: |
1407 | plusencode = urlencode; |
1408 | appendquery = (flags & CURLU_APPENDQUERY)?1:0; |
1409 | equalsencode = appendquery; |
1410 | storep = &u->query; |
1411 | break; |
1412 | case CURLUPART_FRAGMENT: |
1413 | storep = &u->fragment; |
1414 | break; |
1415 | case CURLUPART_URL: { |
1416 | /* |
1417 | * Allow a new URL to replace the existing (if any) contents. |
1418 | * |
1419 | * If the existing contents is enough for a URL, allow a relative URL to |
1420 | * replace it. |
1421 | */ |
1422 | CURLUcode result; |
1423 | char *oldurl; |
1424 | char *redired_url; |
1425 | CURLU *handle2; |
1426 | |
1427 | if(Curl_is_absolute_url(part, NULL, MAX_SCHEME_LEN + 1)) { |
1428 | handle2 = curl_url(); |
1429 | if(!handle2) |
1430 | return CURLUE_OUT_OF_MEMORY; |
1431 | result = parseurl(part, handle2, flags); |
1432 | if(!result) |
1433 | mv_urlhandle(handle2, u); |
1434 | else |
1435 | curl_url_cleanup(handle2); |
1436 | return result; |
1437 | } |
1438 | /* extract the full "old" URL to do the redirect on */ |
1439 | result = curl_url_get(u, CURLUPART_URL, &oldurl, flags); |
1440 | if(result) { |
1441 | /* couldn't get the old URL, just use the new! */ |
1442 | handle2 = curl_url(); |
1443 | if(!handle2) |
1444 | return CURLUE_OUT_OF_MEMORY; |
1445 | result = parseurl(part, handle2, flags); |
1446 | if(!result) |
1447 | mv_urlhandle(handle2, u); |
1448 | else |
1449 | curl_url_cleanup(handle2); |
1450 | return result; |
1451 | } |
1452 | |
1453 | /* apply the relative part to create a new URL */ |
1454 | redired_url = concat_url(oldurl, part); |
1455 | free(oldurl); |
1456 | if(!redired_url) |
1457 | return CURLUE_OUT_OF_MEMORY; |
1458 | |
1459 | /* now parse the new URL */ |
1460 | handle2 = curl_url(); |
1461 | if(!handle2) { |
1462 | free(redired_url); |
1463 | return CURLUE_OUT_OF_MEMORY; |
1464 | } |
1465 | result = parseurl(redired_url, handle2, flags); |
1466 | free(redired_url); |
1467 | if(!result) |
1468 | mv_urlhandle(handle2, u); |
1469 | else |
1470 | curl_url_cleanup(handle2); |
1471 | return result; |
1472 | } |
1473 | default: |
1474 | return CURLUE_UNKNOWN_PART; |
1475 | } |
1476 | DEBUGASSERT(storep); |
1477 | { |
1478 | const char *newp = part; |
1479 | size_t nalloc = strlen(part); |
1480 | |
1481 | if(nalloc > CURL_MAX_INPUT_LENGTH) |
1482 | /* excessive input length */ |
1483 | return CURLUE_MALFORMED_INPUT; |
1484 | |
1485 | if(urlencode) { |
1486 | const unsigned char *i; |
1487 | char *o; |
1488 | char *enc = malloc(nalloc * 3 + 1); /* for worst case! */ |
1489 | if(!enc) |
1490 | return CURLUE_OUT_OF_MEMORY; |
1491 | for(i = (const unsigned char *)part, o = enc; *i; i++) { |
1492 | if((*i == ' ') && plusencode) { |
1493 | *o = '+'; |
1494 | o++; |
1495 | } |
1496 | else if(Curl_isunreserved(*i) || |
1497 | ((*i == '/') && urlskipslash) || |
1498 | ((*i == '=') && equalsencode)) { |
1499 | if((*i == '=') && equalsencode) |
1500 | /* only skip the first equals sign */ |
1501 | equalsencode = FALSE; |
1502 | *o = *i; |
1503 | o++; |
1504 | } |
1505 | else { |
1506 | msnprintf(o, 4, "%%%02x" , *i); |
1507 | o += 3; |
1508 | } |
1509 | } |
1510 | *o = 0; /* null-terminate */ |
1511 | newp = enc; |
1512 | } |
1513 | else { |
1514 | char *p; |
1515 | newp = strdup(part); |
1516 | if(!newp) |
1517 | return CURLUE_OUT_OF_MEMORY; |
1518 | p = (char *)newp; |
1519 | while(*p) { |
1520 | /* make sure percent encoded are lower case */ |
1521 | if((*p == '%') && ISXDIGIT(p[1]) && ISXDIGIT(p[2]) && |
1522 | (ISUPPER(p[1]) || ISUPPER(p[2]))) { |
1523 | p[1] = (char)TOLOWER(p[1]); |
1524 | p[2] = (char)TOLOWER(p[2]); |
1525 | p += 3; |
1526 | } |
1527 | else |
1528 | p++; |
1529 | } |
1530 | } |
1531 | |
1532 | if(appendquery) { |
1533 | /* Append the string onto the old query. Add a '&' separator if none is |
1534 | present at the end of the exsting query already */ |
1535 | size_t querylen = u->query ? strlen(u->query) : 0; |
1536 | bool addamperand = querylen && (u->query[querylen -1] != '&'); |
1537 | if(querylen) { |
1538 | size_t newplen = strlen(newp); |
1539 | char *p = malloc(querylen + addamperand + newplen + 1); |
1540 | if(!p) { |
1541 | free((char *)newp); |
1542 | return CURLUE_OUT_OF_MEMORY; |
1543 | } |
1544 | strcpy(p, u->query); /* original query */ |
1545 | if(addamperand) |
1546 | p[querylen] = '&'; /* ampersand */ |
1547 | strcpy(&p[querylen + addamperand], newp); /* new suffix */ |
1548 | free((char *)newp); |
1549 | free(*storep); |
1550 | *storep = p; |
1551 | return CURLUE_OK; |
1552 | } |
1553 | } |
1554 | |
1555 | if(what == CURLUPART_HOST) { |
1556 | if(0 == strlen(newp) && (flags & CURLU_NO_AUTHORITY)) { |
1557 | /* Skip hostname check, it's allowed to be empty. */ |
1558 | } |
1559 | else { |
1560 | if(hostname_check(u, (char *)newp)) { |
1561 | free((char *)newp); |
1562 | return CURLUE_MALFORMED_INPUT; |
1563 | } |
1564 | } |
1565 | } |
1566 | |
1567 | free(*storep); |
1568 | *storep = (char *)newp; |
1569 | } |
1570 | /* set after the string, to make it not assigned if the allocation above |
1571 | fails */ |
1572 | if(port) |
1573 | u->portnum = port; |
1574 | return CURLUE_OK; |
1575 | } |
1576 | |