1: /***************************************************************************
2: * _ _ ____ _
3: * Project ___| | | | _ \| |
4: * / __| | | | |_) | |
5: * | (__| |_| | _ <| |___
6: * \___|\___/|_| \_\_____|
7: *
8: * Copyright (C) 1998 - 2020, Daniel Stenberg, <daniel@haxx.se>, et al.
9: *
10: * This software is licensed as described in the file COPYING, which
11: * you should have received as part of this distribution. The terms
12: * are also available at https://curl.haxx.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; /* zero 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: {
502: char *portptr = NULL;
503: char endbracket;
504: int len;
505:
506: /*
507: * Find the end of an IPv6 address, either on the ']' ending bracket or
508: * a percent-encoded zone index.
509: */
510: if(1 == sscanf(hostname, "[%*45[0123456789abcdefABCDEF:.]%c%n",
511: &endbracket, &len)) {
512: if(']' == endbracket)
513: portptr = &hostname[len];
514: else if('%' == endbracket) {
515: int zonelen = len;
516: if(1 == sscanf(hostname + zonelen, "%*[^]]%c%n", &endbracket, &len)) {
517: if(']' != endbracket)
518: return CURLUE_MALFORMED_INPUT;
519: portptr = &hostname[--zonelen + len + 1];
520: }
521: else
522: return CURLUE_MALFORMED_INPUT;
523: }
524: else
525: return CURLUE_MALFORMED_INPUT;
526:
527: /* this is a RFC2732-style specified IP-address */
528: if(portptr && *portptr) {
529: if(*portptr != ':')
530: return CURLUE_MALFORMED_INPUT;
531: }
532: else
533: portptr = NULL;
534: }
535: else
536: portptr = strchr(hostname, ':');
537:
538: if(portptr) {
539: char *rest;
540: long port;
541: char portbuf[7];
542:
543: /* Browser behavior adaptation. If there's a colon with no digits after,
544: just cut off the name there which makes us ignore the colon and just
545: use the default port. Firefox, Chrome and Safari all do that. */
546: if(!portptr[1]) {
547: *portptr = '\0';
548: return CURLUE_OK;
549: }
550:
551: if(!ISDIGIT(portptr[1]))
552: return CURLUE_BAD_PORT_NUMBER;
553:
554: port = strtol(portptr + 1, &rest, 10); /* Port number must be decimal */
555:
556: if((port <= 0) || (port > 0xffff))
557: /* Single unix standard says port numbers are 16 bits long, but we don't
558: treat port zero as OK. */
559: return CURLUE_BAD_PORT_NUMBER;
560:
561: if(rest[0])
562: return CURLUE_BAD_PORT_NUMBER;
563:
564: *portptr++ = '\0'; /* cut off the name there */
565: *rest = 0;
566: /* generate a new port number string to get rid of leading zeroes etc */
567: msnprintf(portbuf, sizeof(portbuf), "%ld", port);
568: u->portnum = port;
569: u->port = strdup(portbuf);
570: if(!u->port)
571: return CURLUE_OUT_OF_MEMORY;
572: }
573:
574: return CURLUE_OK;
575: }
576:
577: /* scan for byte values < 31 or 127 */
578: static CURLUcode junkscan(const char *part)
579: {
580: if(part) {
581: static const char badbytes[]={
582: /* */ 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
583: 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
584: 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
585: 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f,
586: 0x7f,
587: 0x00 /* zero terminate */
588: };
589: size_t n = strlen(part);
590: size_t nfine = strcspn(part, badbytes);
591: if(nfine != n)
592: /* since we don't know which part is scanned, return a generic error
593: code */
594: return CURLUE_MALFORMED_INPUT;
595: }
596: return CURLUE_OK;
597: }
598:
599: static CURLUcode hostname_check(struct Curl_URL *u, char *hostname)
600: {
601: size_t len;
602: size_t hlen = strlen(hostname);
603:
604: if(hostname[0] == '[') {
605: #ifdef ENABLE_IPV6
606: char dest[16]; /* fits a binary IPv6 address */
607: #endif
608: const char *l = "0123456789abcdefABCDEF:.";
609: if(hlen < 5) /* '[::1]' is the shortest possible valid string */
610: return CURLUE_MALFORMED_INPUT;
611: hostname++;
612: hlen -= 2;
613:
614: if(hostname[hlen] != ']')
615: return CURLUE_MALFORMED_INPUT;
616:
617: /* only valid letters are ok */
618: len = strspn(hostname, l);
619: if(hlen != len) {
620: hlen = len;
621: if(hostname[len] == '%') {
622: /* this could now be '%[zone id]' */
623: char zoneid[16];
624: int i = 0;
625: char *h = &hostname[len + 1];
626: /* pass '25' if present and is a url encoded percent sign */
627: if(!strncmp(h, "25", 2) && h[2] && (h[2] != ']'))
628: h += 2;
629: while(*h && (*h != ']') && (i < 15))
630: zoneid[i++] = *h++;
631: if(!i || (']' != *h))
632: return CURLUE_MALFORMED_INPUT;
633: zoneid[i] = 0;
634: u->zoneid = strdup(zoneid);
635: if(!u->zoneid)
636: return CURLUE_OUT_OF_MEMORY;
637: hostname[len] = ']'; /* insert end bracket */
638: hostname[len + 1] = 0; /* terminate the hostname */
639: }
640: else
641: return CURLUE_MALFORMED_INPUT;
642: /* hostname is fine */
643: }
644: #ifdef ENABLE_IPV6
645: hostname[hlen] = 0; /* end the address there */
646: if(1 != Curl_inet_pton(AF_INET6, hostname, dest))
647: return CURLUE_MALFORMED_INPUT;
648: hostname[hlen] = ']'; /* restore ending bracket */
649: #endif
650: }
651: else {
652: /* letters from the second string is not ok */
653: len = strcspn(hostname, " ");
654: if(hlen != len)
655: /* hostname with bad content */
656: return CURLUE_MALFORMED_INPUT;
657: }
658: if(!hostname[0])
659: return CURLUE_NO_HOST;
660: return CURLUE_OK;
661: }
662:
663: #define HOSTNAME_END(x) (((x) == '/') || ((x) == '?') || ((x) == '#'))
664:
665: static CURLUcode seturl(const char *url, CURLU *u, unsigned int flags)
666: {
667: char *path;
668: bool path_alloced = FALSE;
669: char *hostname;
670: char *query = NULL;
671: char *fragment = NULL;
672: CURLUcode result;
673: bool url_has_scheme = FALSE;
674: char schemebuf[MAX_SCHEME_LEN + 1];
675: const char *schemep = NULL;
676: size_t schemelen = 0;
677: size_t urllen;
678:
679: if(!url)
680: return CURLUE_MALFORMED_INPUT;
681:
682: /*************************************************************
683: * Parse the URL.
684: ************************************************************/
685: /* allocate scratch area */
686: urllen = strlen(url);
687: if(urllen > CURL_MAX_INPUT_LENGTH)
688: /* excessive input length */
689: return CURLUE_MALFORMED_INPUT;
690:
691: path = u->scratch = malloc(urllen * 2 + 2);
692: if(!path)
693: return CURLUE_OUT_OF_MEMORY;
694:
695: hostname = &path[urllen + 1];
696: hostname[0] = 0;
697:
698: if(Curl_is_absolute_url(url, schemebuf, sizeof(schemebuf))) {
699: url_has_scheme = TRUE;
700: schemelen = strlen(schemebuf);
701: }
702:
703: /* handle the file: scheme */
704: if(url_has_scheme && strcasecompare(schemebuf, "file")) {
705: /* path has been allocated large enough to hold this */
706: strcpy(path, &url[5]);
707:
708: hostname = NULL; /* no host for file: URLs */
709: u->scheme = strdup("file");
710: if(!u->scheme)
711: return CURLUE_OUT_OF_MEMORY;
712:
713: /* Extra handling URLs with an authority component (i.e. that start with
714: * "file://")
715: *
716: * We allow omitted hostname (e.g. file:/<path>) -- valid according to
717: * RFC 8089, but not the (current) WHAT-WG URL spec.
718: */
719: if(path[0] == '/' && path[1] == '/') {
720: /* swallow the two slashes */
721: char *ptr = &path[2];
722:
723: /*
724: * According to RFC 8089, a file: URL can be reliably dereferenced if:
725: *
726: * o it has no/blank hostname, or
727: *
728: * o the hostname matches "localhost" (case-insensitively), or
729: *
730: * o the hostname is a FQDN that resolves to this machine.
731: *
732: * For brevity, we only consider URLs with empty, "localhost", or
733: * "127.0.0.1" hostnames as local.
734: *
735: * Additionally, there is an exception for URLs with a Windows drive
736: * letter in the authority (which was accidentally omitted from RFC 8089
737: * Appendix E, but believe me, it was meant to be there. --MK)
738: */
739: if(ptr[0] != '/' && !STARTS_WITH_URL_DRIVE_PREFIX(ptr)) {
740: /* the URL includes a host name, it must match "localhost" or
741: "127.0.0.1" to be valid */
742: if(!checkprefix("localhost/", ptr) &&
743: !checkprefix("127.0.0.1/", ptr)) {
744: /* Invalid file://hostname/, expected localhost or 127.0.0.1 or
745: none */
746: return CURLUE_MALFORMED_INPUT;
747: }
748: ptr += 9; /* now points to the slash after the host */
749: }
750:
751: path = ptr;
752: }
753:
754: #if !defined(MSDOS) && !defined(WIN32) && !defined(__CYGWIN__)
755: /* Don't allow Windows drive letters when not in Windows.
756: * This catches both "file:/c:" and "file:c:" */
757: if(('/' == path[0] && STARTS_WITH_URL_DRIVE_PREFIX(&path[1])) ||
758: STARTS_WITH_URL_DRIVE_PREFIX(path)) {
759: /* File drive letters are only accepted in MSDOS/Windows */
760: return CURLUE_MALFORMED_INPUT;
761: }
762: #else
763: /* If the path starts with a slash and a drive letter, ditch the slash */
764: if('/' == path[0] && STARTS_WITH_URL_DRIVE_PREFIX(&path[1])) {
765: /* This cannot be done with strcpy, as the memory chunks overlap! */
766: memmove(path, &path[1], strlen(&path[1]) + 1);
767: }
768: #endif
769:
770: }
771: else {
772: /* clear path */
773: const char *p;
774: const char *hostp;
775: size_t len;
776: path[0] = 0;
777:
778: if(url_has_scheme) {
779: int i = 0;
780: p = &url[schemelen + 1];
781: while(p && (*p == '/') && (i < 4)) {
782: p++;
783: i++;
784: }
785: if((i < 1) || (i>3))
786: /* less than one or more than three slashes */
787: return CURLUE_MALFORMED_INPUT;
788:
789: schemep = schemebuf;
790: if(!Curl_builtin_scheme(schemep) &&
791: !(flags & CURLU_NON_SUPPORT_SCHEME))
792: return CURLUE_UNSUPPORTED_SCHEME;
793:
794: if(junkscan(schemep))
795: return CURLUE_MALFORMED_INPUT;
796:
797: }
798: else {
799: /* no scheme! */
800:
801: if(!(flags & (CURLU_DEFAULT_SCHEME|CURLU_GUESS_SCHEME)))
802: return CURLUE_MALFORMED_INPUT;
803: if(flags & CURLU_DEFAULT_SCHEME)
804: schemep = DEFAULT_SCHEME;
805:
806: /*
807: * The URL was badly formatted, let's try without scheme specified.
808: */
809: p = url;
810: }
811: hostp = p; /* host name starts here */
812:
813: while(*p && !HOSTNAME_END(*p)) /* find end of host name */
814: p++;
815:
816: len = p - hostp;
817: if(len) {
818: memcpy(hostname, hostp, len);
819: hostname[len] = 0;
820: }
821: else {
822: if(!(flags & CURLU_NO_AUTHORITY))
823: return CURLUE_MALFORMED_INPUT;
824: }
825:
826: len = strlen(p);
827: memcpy(path, p, len);
828: path[len] = 0;
829:
830: if(schemep) {
831: u->scheme = strdup(schemep);
832: if(!u->scheme)
833: return CURLUE_OUT_OF_MEMORY;
834: }
835: }
836:
837: if(junkscan(path))
838: return CURLUE_MALFORMED_INPUT;
839:
840: if((flags & CURLU_URLENCODE) && path[0]) {
841: /* worst case output length is 3x the original! */
842: char *newp = malloc(strlen(path) * 3);
843: if(!newp)
844: return CURLUE_OUT_OF_MEMORY;
845: path_alloced = TRUE;
846: strcpy_url(newp, path, TRUE); /* consider it relative */
847: u->temppath = path = newp;
848: }
849:
850: fragment = strchr(path, '#');
851: if(fragment) {
852: *fragment++ = 0;
853: if(fragment[0]) {
854: u->fragment = strdup(fragment);
855: if(!u->fragment)
856: return CURLUE_OUT_OF_MEMORY;
857: }
858: }
859:
860: query = strchr(path, '?');
861: if(query) {
862: *query++ = 0;
863: /* done even if the query part is a blank string */
864: u->query = strdup(query);
865: if(!u->query)
866: return CURLUE_OUT_OF_MEMORY;
867: }
868:
869: if(!path[0])
870: /* if there's no path left set, unset */
871: path = NULL;
872: else {
873: if(!(flags & CURLU_PATH_AS_IS)) {
874: /* remove ../ and ./ sequences according to RFC3986 */
875: char *newp = Curl_dedotdotify(path);
876: if(!newp)
877: return CURLUE_OUT_OF_MEMORY;
878:
879: if(strcmp(newp, path)) {
880: /* if we got a new version */
881: if(path_alloced)
882: Curl_safefree(u->temppath);
883: u->temppath = path = newp;
884: path_alloced = TRUE;
885: }
886: else
887: free(newp);
888: }
889:
890: u->path = path_alloced?path:strdup(path);
891: if(!u->path)
892: return CURLUE_OUT_OF_MEMORY;
893: u->temppath = NULL; /* used now */
894: }
895:
896: if(hostname) {
897: /*
898: * Parse the login details and strip them out of the host name.
899: */
900: if(junkscan(hostname))
901: return CURLUE_MALFORMED_INPUT;
902:
903: result = parse_hostname_login(u, &hostname, flags);
904: if(result)
905: return result;
906:
907: result = Curl_parse_port(u, hostname);
908: if(result)
909: return result;
910:
911: if(0 == strlen(hostname) && (flags & CURLU_NO_AUTHORITY)) {
912: /* Skip hostname check, it's allowed to be empty. */
913: }
914: else {
915: result = hostname_check(u, hostname);
916: if(result)
917: return result;
918: }
919:
920: u->host = strdup(hostname);
921: if(!u->host)
922: return CURLUE_OUT_OF_MEMORY;
923:
924: if((flags & CURLU_GUESS_SCHEME) && !schemep) {
925: /* legacy curl-style guess based on host name */
926: if(checkprefix("ftp.", hostname))
927: schemep = "ftp";
928: else if(checkprefix("dict.", hostname))
929: schemep = "dict";
930: else if(checkprefix("ldap.", hostname))
931: schemep = "ldap";
932: else if(checkprefix("imap.", hostname))
933: schemep = "imap";
934: else if(checkprefix("smtp.", hostname))
935: schemep = "smtp";
936: else if(checkprefix("pop3.", hostname))
937: schemep = "pop3";
938: else
939: schemep = "http";
940:
941: u->scheme = strdup(schemep);
942: if(!u->scheme)
943: return CURLUE_OUT_OF_MEMORY;
944: }
945: }
946:
947: Curl_safefree(u->scratch);
948: Curl_safefree(u->temppath);
949:
950: return CURLUE_OK;
951: }
952:
953: /*
954: * Parse the URL and set the relevant members of the Curl_URL struct.
955: */
956: static CURLUcode parseurl(const char *url, CURLU *u, unsigned int flags)
957: {
958: CURLUcode result = seturl(url, u, flags);
959: if(result) {
960: free_urlhandle(u);
961: memset(u, 0, sizeof(struct Curl_URL));
962: }
963: return result;
964: }
965:
966: /*
967: */
968: CURLU *curl_url(void)
969: {
970: return calloc(sizeof(struct Curl_URL), 1);
971: }
972:
973: void curl_url_cleanup(CURLU *u)
974: {
975: if(u) {
976: free_urlhandle(u);
977: free(u);
978: }
979: }
980:
981: #define DUP(dest, src, name) \
982: if(src->name) { \
983: dest->name = strdup(src->name); \
984: if(!dest->name) \
985: goto fail; \
986: }
987:
988: CURLU *curl_url_dup(CURLU *in)
989: {
990: struct Curl_URL *u = calloc(sizeof(struct Curl_URL), 1);
991: if(u) {
992: DUP(u, in, scheme);
993: DUP(u, in, user);
994: DUP(u, in, password);
995: DUP(u, in, options);
996: DUP(u, in, host);
997: DUP(u, in, port);
998: DUP(u, in, path);
999: DUP(u, in, query);
1000: DUP(u, in, fragment);
1001: u->portnum = in->portnum;
1002: }
1003: return u;
1004: fail:
1005: curl_url_cleanup(u);
1006: return NULL;
1007: }
1008:
1009: CURLUcode curl_url_get(CURLU *u, CURLUPart what,
1010: char **part, unsigned int flags)
1011: {
1012: char *ptr;
1013: CURLUcode ifmissing = CURLUE_UNKNOWN_PART;
1014: char portbuf[7];
1015: bool urldecode = (flags & CURLU_URLDECODE)?1:0;
1016: bool plusdecode = FALSE;
1017: (void)flags;
1018: if(!u)
1019: return CURLUE_BAD_HANDLE;
1020: if(!part)
1021: return CURLUE_BAD_PARTPOINTER;
1022: *part = NULL;
1023:
1024: switch(what) {
1025: case CURLUPART_SCHEME:
1026: ptr = u->scheme;
1027: ifmissing = CURLUE_NO_SCHEME;
1028: urldecode = FALSE; /* never for schemes */
1029: break;
1030: case CURLUPART_USER:
1031: ptr = u->user;
1032: ifmissing = CURLUE_NO_USER;
1033: break;
1034: case CURLUPART_PASSWORD:
1035: ptr = u->password;
1036: ifmissing = CURLUE_NO_PASSWORD;
1037: break;
1038: case CURLUPART_OPTIONS:
1039: ptr = u->options;
1040: ifmissing = CURLUE_NO_OPTIONS;
1041: break;
1042: case CURLUPART_HOST:
1043: ptr = u->host;
1044: ifmissing = CURLUE_NO_HOST;
1045: break;
1046: case CURLUPART_ZONEID:
1047: ptr = u->zoneid;
1048: break;
1049: case CURLUPART_PORT:
1050: ptr = u->port;
1051: ifmissing = CURLUE_NO_PORT;
1052: urldecode = FALSE; /* never for port */
1053: if(!ptr && (flags & CURLU_DEFAULT_PORT) && u->scheme) {
1054: /* there's no stored port number, but asked to deliver
1055: a default one for the scheme */
1056: const struct Curl_handler *h =
1057: Curl_builtin_scheme(u->scheme);
1058: if(h) {
1059: msnprintf(portbuf, sizeof(portbuf), "%ld", h->defport);
1060: ptr = portbuf;
1061: }
1062: }
1063: else if(ptr && u->scheme) {
1064: /* there is a stored port number, but ask to inhibit if
1065: it matches the default one for the scheme */
1066: const struct Curl_handler *h =
1067: Curl_builtin_scheme(u->scheme);
1068: if(h && (h->defport == u->portnum) &&
1069: (flags & CURLU_NO_DEFAULT_PORT))
1070: ptr = NULL;
1071: }
1072: break;
1073: case CURLUPART_PATH:
1074: ptr = u->path;
1075: if(!ptr) {
1076: ptr = u->path = strdup("/");
1077: if(!u->path)
1078: return CURLUE_OUT_OF_MEMORY;
1079: }
1080: break;
1081: case CURLUPART_QUERY:
1082: ptr = u->query;
1083: ifmissing = CURLUE_NO_QUERY;
1084: plusdecode = urldecode;
1085: break;
1086: case CURLUPART_FRAGMENT:
1087: ptr = u->fragment;
1088: ifmissing = CURLUE_NO_FRAGMENT;
1089: break;
1090: case CURLUPART_URL: {
1091: char *url;
1092: char *scheme;
1093: char *options = u->options;
1094: char *port = u->port;
1095: char *allochost = NULL;
1096: if(u->scheme && strcasecompare("file", u->scheme)) {
1097: url = aprintf("file://%s%s%s",
1098: u->path,
1099: u->fragment? "#": "",
1100: u->fragment? u->fragment : "");
1101: }
1102: else if(!u->host)
1103: return CURLUE_NO_HOST;
1104: else {
1105: const struct Curl_handler *h = NULL;
1106: if(u->scheme)
1107: scheme = u->scheme;
1108: else if(flags & CURLU_DEFAULT_SCHEME)
1109: scheme = (char *) DEFAULT_SCHEME;
1110: else
1111: return CURLUE_NO_SCHEME;
1112:
1113: h = Curl_builtin_scheme(scheme);
1114: if(!port && (flags & CURLU_DEFAULT_PORT)) {
1115: /* there's no stored port number, but asked to deliver
1116: a default one for the scheme */
1117: if(h) {
1118: msnprintf(portbuf, sizeof(portbuf), "%ld", h->defport);
1119: port = portbuf;
1120: }
1121: }
1122: else if(port) {
1123: /* there is a stored port number, but asked to inhibit if it matches
1124: the default one for the scheme */
1125: if(h && (h->defport == u->portnum) &&
1126: (flags & CURLU_NO_DEFAULT_PORT))
1127: port = NULL;
1128: }
1129:
1130: if(h && !(h->flags & PROTOPT_URLOPTIONS))
1131: options = NULL;
1132:
1133: if((u->host[0] == '[') && u->zoneid) {
1134: /* make it '[ host %25 zoneid ]' */
1135: size_t hostlen = strlen(u->host);
1136: size_t alen = hostlen + 3 + strlen(u->zoneid) + 1;
1137: allochost = malloc(alen);
1138: if(!allochost)
1139: return CURLUE_OUT_OF_MEMORY;
1140: memcpy(allochost, u->host, hostlen - 1);
1141: msnprintf(&allochost[hostlen - 1], alen - hostlen + 1,
1142: "%%25%s]", u->zoneid);
1143: }
1144:
1145: url = aprintf("%s://%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s",
1146: scheme,
1147: u->user ? u->user : "",
1148: u->password ? ":": "",
1149: u->password ? u->password : "",
1150: options ? ";" : "",
1151: options ? options : "",
1152: (u->user || u->password || options) ? "@": "",
1153: allochost ? allochost : u->host,
1154: port ? ":": "",
1155: port ? port : "",
1156: (u->path && (u->path[0] != '/')) ? "/": "",
1157: u->path ? u->path : "/",
1158: (u->query && u->query[0]) ? "?": "",
1159: (u->query && u->query[0]) ? u->query : "",
1160: u->fragment? "#": "",
1161: u->fragment? u->fragment : "");
1162: free(allochost);
1163: }
1164: if(!url)
1165: return CURLUE_OUT_OF_MEMORY;
1166: *part = url;
1167: return CURLUE_OK;
1168: }
1169: default:
1170: ptr = NULL;
1171: break;
1172: }
1173: if(ptr) {
1174: *part = strdup(ptr);
1175: if(!*part)
1176: return CURLUE_OUT_OF_MEMORY;
1177: if(plusdecode) {
1178: /* convert + to space */
1179: char *plus;
1180: for(plus = *part; *plus; ++plus) {
1181: if(*plus == '+')
1182: *plus = ' ';
1183: }
1184: }
1185: if(urldecode) {
1186: char *decoded;
1187: size_t dlen;
1188: CURLcode res = Curl_urldecode(NULL, *part, 0, &decoded, &dlen, TRUE);
1189: free(*part);
1190: if(res) {
1191: *part = NULL;
1192: return CURLUE_URLDECODE;
1193: }
1194: *part = decoded;
1195: }
1196: return CURLUE_OK;
1197: }
1198: else
1199: return ifmissing;
1200: }
1201:
1202: CURLUcode curl_url_set(CURLU *u, CURLUPart what,
1203: const char *part, unsigned int flags)
1204: {
1205: char **storep = NULL;
1206: long port = 0;
1207: bool urlencode = (flags & CURLU_URLENCODE)? 1 : 0;
1208: bool plusencode = FALSE;
1209: bool urlskipslash = FALSE;
1210: bool appendquery = FALSE;
1211: bool equalsencode = FALSE;
1212:
1213: if(!u)
1214: return CURLUE_BAD_HANDLE;
1215: if(!part) {
1216: /* setting a part to NULL clears it */
1217: switch(what) {
1218: case CURLUPART_URL:
1219: break;
1220: case CURLUPART_SCHEME:
1221: storep = &u->scheme;
1222: break;
1223: case CURLUPART_USER:
1224: storep = &u->user;
1225: break;
1226: case CURLUPART_PASSWORD:
1227: storep = &u->password;
1228: break;
1229: case CURLUPART_OPTIONS:
1230: storep = &u->options;
1231: break;
1232: case CURLUPART_HOST:
1233: storep = &u->host;
1234: break;
1235: case CURLUPART_ZONEID:
1236: storep = &u->zoneid;
1237: break;
1238: case CURLUPART_PORT:
1239: u->portnum = 0;
1240: storep = &u->port;
1241: break;
1242: case CURLUPART_PATH:
1243: storep = &u->path;
1244: break;
1245: case CURLUPART_QUERY:
1246: storep = &u->query;
1247: break;
1248: case CURLUPART_FRAGMENT:
1249: storep = &u->fragment;
1250: break;
1251: default:
1252: return CURLUE_UNKNOWN_PART;
1253: }
1254: if(storep && *storep) {
1255: free(*storep);
1256: *storep = NULL;
1257: }
1258: return CURLUE_OK;
1259: }
1260:
1261: switch(what) {
1262: case CURLUPART_SCHEME:
1263: if(strlen(part) > MAX_SCHEME_LEN)
1264: /* too long */
1265: return CURLUE_MALFORMED_INPUT;
1266: if(!(flags & CURLU_NON_SUPPORT_SCHEME) &&
1267: /* verify that it is a fine scheme */
1268: !Curl_builtin_scheme(part))
1269: return CURLUE_UNSUPPORTED_SCHEME;
1270: storep = &u->scheme;
1271: urlencode = FALSE; /* never */
1272: break;
1273: case CURLUPART_USER:
1274: storep = &u->user;
1275: break;
1276: case CURLUPART_PASSWORD:
1277: storep = &u->password;
1278: break;
1279: case CURLUPART_OPTIONS:
1280: storep = &u->options;
1281: break;
1282: case CURLUPART_HOST:
1283: storep = &u->host;
1284: free(u->zoneid);
1285: u->zoneid = NULL;
1286: break;
1287: case CURLUPART_ZONEID:
1288: storep = &u->zoneid;
1289: break;
1290: case CURLUPART_PORT:
1291: {
1292: char *endp;
1293: urlencode = FALSE; /* never */
1294: port = strtol(part, &endp, 10); /* Port number must be decimal */
1295: if((port <= 0) || (port > 0xffff))
1296: return CURLUE_BAD_PORT_NUMBER;
1297: if(*endp)
1298: /* weirdly provided number, not good! */
1299: return CURLUE_MALFORMED_INPUT;
1300: storep = &u->port;
1301: }
1302: break;
1303: case CURLUPART_PATH:
1304: urlskipslash = TRUE;
1305: storep = &u->path;
1306: break;
1307: case CURLUPART_QUERY:
1308: plusencode = urlencode;
1309: appendquery = (flags & CURLU_APPENDQUERY)?1:0;
1310: equalsencode = appendquery;
1311: storep = &u->query;
1312: break;
1313: case CURLUPART_FRAGMENT:
1314: storep = &u->fragment;
1315: break;
1316: case CURLUPART_URL: {
1317: /*
1318: * Allow a new URL to replace the existing (if any) contents.
1319: *
1320: * If the existing contents is enough for a URL, allow a relative URL to
1321: * replace it.
1322: */
1323: CURLUcode result;
1324: char *oldurl;
1325: char *redired_url;
1326: CURLU *handle2;
1327:
1328: if(Curl_is_absolute_url(part, NULL, MAX_SCHEME_LEN + 1)) {
1329: handle2 = curl_url();
1330: if(!handle2)
1331: return CURLUE_OUT_OF_MEMORY;
1332: result = parseurl(part, handle2, flags);
1333: if(!result)
1334: mv_urlhandle(handle2, u);
1335: else
1336: curl_url_cleanup(handle2);
1337: return result;
1338: }
1339: /* extract the full "old" URL to do the redirect on */
1340: result = curl_url_get(u, CURLUPART_URL, &oldurl, flags);
1341: if(result) {
1342: /* couldn't get the old URL, just use the new! */
1343: handle2 = curl_url();
1344: if(!handle2)
1345: return CURLUE_OUT_OF_MEMORY;
1346: result = parseurl(part, handle2, flags);
1347: if(!result)
1348: mv_urlhandle(handle2, u);
1349: else
1350: curl_url_cleanup(handle2);
1351: return result;
1352: }
1353:
1354: /* apply the relative part to create a new URL */
1355: redired_url = concat_url(oldurl, part);
1356: free(oldurl);
1357: if(!redired_url)
1358: return CURLUE_OUT_OF_MEMORY;
1359:
1360: /* now parse the new URL */
1361: handle2 = curl_url();
1362: if(!handle2) {
1363: free(redired_url);
1364: return CURLUE_OUT_OF_MEMORY;
1365: }
1366: result = parseurl(redired_url, handle2, flags);
1367: free(redired_url);
1368: if(!result)
1369: mv_urlhandle(handle2, u);
1370: else
1371: curl_url_cleanup(handle2);
1372: return result;
1373: }
1374: default:
1375: return CURLUE_UNKNOWN_PART;
1376: }
1377: DEBUGASSERT(storep);
1378: {
1379: const char *newp = part;
1380: size_t nalloc = strlen(part);
1381:
1382: if(nalloc > CURL_MAX_INPUT_LENGTH)
1383: /* excessive input length */
1384: return CURLUE_MALFORMED_INPUT;
1385:
1386: if(urlencode) {
1387: const unsigned char *i;
1388: char *o;
1389: bool free_part = FALSE;
1390: char *enc = malloc(nalloc * 3 + 1); /* for worst case! */
1391: if(!enc)
1392: return CURLUE_OUT_OF_MEMORY;
1393: if(plusencode) {
1394: /* space to plus */
1395: i = (const unsigned char *)part;
1396: for(o = enc; *i; ++o, ++i)
1397: *o = (*i == ' ') ? '+' : *i;
1398: *o = 0; /* zero terminate */
1399: part = strdup(enc);
1400: if(!part) {
1401: free(enc);
1402: return CURLUE_OUT_OF_MEMORY;
1403: }
1404: free_part = TRUE;
1405: }
1406: for(i = (const unsigned char *)part, o = enc; *i; i++) {
1407: if(Curl_isunreserved(*i) ||
1408: ((*i == '/') && urlskipslash) ||
1409: ((*i == '=') && equalsencode) ||
1410: ((*i == '+') && plusencode)) {
1411: if((*i == '=') && equalsencode)
1412: /* only skip the first equals sign */
1413: equalsencode = FALSE;
1414: *o = *i;
1415: o++;
1416: }
1417: else {
1418: msnprintf(o, 4, "%%%02x", *i);
1419: o += 3;
1420: }
1421: }
1422: *o = 0; /* zero terminate */
1423: newp = enc;
1424: if(free_part)
1425: free((char *)part);
1426: }
1427: else {
1428: char *p;
1429: newp = strdup(part);
1430: if(!newp)
1431: return CURLUE_OUT_OF_MEMORY;
1432: p = (char *)newp;
1433: while(*p) {
1434: /* make sure percent encoded are lower case */
1435: if((*p == '%') && ISXDIGIT(p[1]) && ISXDIGIT(p[2]) &&
1436: (ISUPPER(p[1]) || ISUPPER(p[2]))) {
1437: p[1] = (char)TOLOWER(p[1]);
1438: p[2] = (char)TOLOWER(p[2]);
1439: p += 3;
1440: }
1441: else
1442: p++;
1443: }
1444: }
1445:
1446: if(appendquery) {
1447: /* Append the string onto the old query. Add a '&' separator if none is
1448: present at the end of the exsting query already */
1449: size_t querylen = u->query ? strlen(u->query) : 0;
1450: bool addamperand = querylen && (u->query[querylen -1] != '&');
1451: if(querylen) {
1452: size_t newplen = strlen(newp);
1453: char *p = malloc(querylen + addamperand + newplen + 1);
1454: if(!p) {
1455: free((char *)newp);
1456: return CURLUE_OUT_OF_MEMORY;
1457: }
1458: strcpy(p, u->query); /* original query */
1459: if(addamperand)
1460: p[querylen] = '&'; /* ampersand */
1461: strcpy(&p[querylen + addamperand], newp); /* new suffix */
1462: free((char *)newp);
1463: free(*storep);
1464: *storep = p;
1465: return CURLUE_OK;
1466: }
1467: }
1468:
1469: if(what == CURLUPART_HOST) {
1470: if(0 == strlen(newp) && (flags & CURLU_NO_AUTHORITY)) {
1471: /* Skip hostname check, it's allowed to be empty. */
1472: }
1473: else {
1474: if(hostname_check(u, (char *)newp)) {
1475: free((char *)newp);
1476: return CURLUE_MALFORMED_INPUT;
1477: }
1478: }
1479: }
1480:
1481: free(*storep);
1482: *storep = (char *)newp;
1483: }
1484: /* set after the string, to make it not assigned if the allocation above
1485: fails */
1486: if(port)
1487: u->portnum = port;
1488: return CURLUE_OK;
1489: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>
![[BACK]](/icons/cvsweb/back.gif){kind=icon}
Return to urlapi.c CVS log ![[TXT]](/icons/cvsweb/text.gif){kind=icon}
![[DIR]](/icons/cvsweb/dir.gif){kind=icon}
Up to [ELWIX - Embedded LightWeight unIX -] / embedaddon / curl / lib