Sync with maint
[git.git] / urlmatch.c
blob3d4c54b5cd5d6f6cb4e0ea97550799eb9d46f033
1 #include "cache.h"
2 #include "urlmatch.h"
4 #define URL_ALPHA "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz"
5 #define URL_DIGIT "0123456789"
6 #define URL_ALPHADIGIT URL_ALPHA URL_DIGIT
7 #define URL_SCHEME_CHARS URL_ALPHADIGIT "+.-"
8 #define URL_HOST_CHARS URL_ALPHADIGIT ".-[:]" /* IPv6 literals need [:] */
9 #define URL_UNSAFE_CHARS " <>\"%{}|\\^`" /* plus 0x00-0x1F,0x7F-0xFF */
10 #define URL_GEN_RESERVED ":/?#[]@"
11 #define URL_SUB_RESERVED "!$&'()*+,;="
12 #define URL_RESERVED URL_GEN_RESERVED URL_SUB_RESERVED /* only allowed delims */
14 static int append_normalized_escapes(struct strbuf *buf,
15 const char *from,
16 size_t from_len,
17 const char *esc_extra,
18 const char *esc_ok)
21 * Append to strbuf 'buf' characters from string 'from' with length
22 * 'from_len' while unescaping characters that do not need to be escaped
23 * and escaping characters that do. The set of characters to escape
24 * (the complement of which is unescaped) starts out as the RFC 3986
25 * unsafe characters (0x00-0x1F,0x7F-0xFF," <>\"#%{}|\\^`"). If
26 * 'esc_extra' is not NULL, those additional characters will also always
27 * be escaped. If 'esc_ok' is not NULL, those characters will be left
28 * escaped if found that way, but will not be unescaped otherwise (used
29 * for delimiters). If a %-escape sequence is encountered that is not
30 * followed by 2 hexadecimal digits, the sequence is invalid and
31 * false (0) will be returned. Otherwise true (1) will be returned for
32 * success.
34 * Note that all %-escape sequences will be normalized to UPPERCASE
35 * as indicated in RFC 3986. Unless included in esc_extra or esc_ok
36 * alphanumerics and "-._~" will always be unescaped as per RFC 3986.
39 while (from_len) {
40 int ch = *from++;
41 int was_esc = 0;
43 from_len--;
44 if (ch == '%') {
45 if (from_len < 2 ||
46 !isxdigit((unsigned char)from[0]) ||
47 !isxdigit((unsigned char)from[1]))
48 return 0;
49 ch = hexval_table[(unsigned char)*from++] << 4;
50 ch |= hexval_table[(unsigned char)*from++];
51 from_len -= 2;
52 was_esc = 1;
54 if ((unsigned char)ch <= 0x1F || (unsigned char)ch >= 0x7F ||
55 strchr(URL_UNSAFE_CHARS, ch) ||
56 (esc_extra && strchr(esc_extra, ch)) ||
57 (was_esc && strchr(esc_ok, ch)))
58 strbuf_addf(buf, "%%%02X", (unsigned char)ch);
59 else
60 strbuf_addch(buf, ch);
63 return 1;
66 char *url_normalize(const char *url, struct url_info *out_info)
69 * Normalize NUL-terminated url using the following rules:
71 * 1. Case-insensitive parts of url will be converted to lower case
72 * 2. %-encoded characters that do not need to be will be unencoded
73 * 3. Characters that are not %-encoded and must be will be encoded
74 * 4. All %-encodings will be converted to upper case hexadecimal
75 * 5. Leading 0s are removed from port numbers
76 * 6. If the default port for the scheme is given it will be removed
77 * 7. A path part (including empty) not starting with '/' has one added
78 * 8. Any dot segments (. or ..) in the path are resolved and removed
79 * 9. IPv6 host literals are allowed (but not normalized or validated)
81 * The rules are based on information in RFC 3986.
83 * Please note this function requires a full URL including a scheme
84 * and host part (except for file: URLs which may have an empty host).
86 * The return value is a newly allocated string that must be freed
87 * or NULL if the url is not valid.
89 * If out_info is non-NULL, the url and err fields therein will always
90 * be set. If a non-NULL value is returned, it will be stored in
91 * out_info->url as well, out_info->err will be set to NULL and the
92 * other fields of *out_info will also be filled in. If a NULL value
93 * is returned, NULL will be stored in out_info->url and out_info->err
94 * will be set to a brief, translated, error message, but no other
95 * fields will be filled in.
97 * This is NOT a URL validation function. Full URL validation is NOT
98 * performed. Some invalid host names are passed through this function
99 * undetected. However, most all other problems that make a URL invalid
100 * will be detected (including a missing host for non file: URLs).
103 size_t url_len = strlen(url);
104 struct strbuf norm;
105 size_t spanned;
106 size_t scheme_len, user_off=0, user_len=0, passwd_off=0, passwd_len=0;
107 size_t host_off=0, host_len=0, port_len=0, path_off, path_len, result_len;
108 const char *slash_ptr, *at_ptr, *colon_ptr, *path_start;
109 char *result;
112 * Copy lowercased scheme and :// suffix, %-escapes are not allowed
113 * First character of scheme must be URL_ALPHA
115 spanned = strspn(url, URL_SCHEME_CHARS);
116 if (!spanned || !isalpha(url[0]) || spanned + 3 > url_len ||
117 url[spanned] != ':' || url[spanned+1] != '/' || url[spanned+2] != '/') {
118 if (out_info) {
119 out_info->url = NULL;
120 out_info->err = _("invalid URL scheme name or missing '://' suffix");
122 return NULL; /* Bad scheme and/or missing "://" part */
124 strbuf_init(&norm, url_len);
125 scheme_len = spanned;
126 spanned += 3;
127 url_len -= spanned;
128 while (spanned--)
129 strbuf_addch(&norm, tolower(*url++));
133 * Copy any username:password if present normalizing %-escapes
135 at_ptr = strchr(url, '@');
136 slash_ptr = url + strcspn(url, "/?#");
137 if (at_ptr && at_ptr < slash_ptr) {
138 user_off = norm.len;
139 if (at_ptr > url) {
140 if (!append_normalized_escapes(&norm, url, at_ptr - url,
141 "", URL_RESERVED)) {
142 if (out_info) {
143 out_info->url = NULL;
144 out_info->err = _("invalid %XX escape sequence");
146 strbuf_release(&norm);
147 return NULL;
149 colon_ptr = strchr(norm.buf + scheme_len + 3, ':');
150 if (colon_ptr) {
151 passwd_off = (colon_ptr + 1) - norm.buf;
152 passwd_len = norm.len - passwd_off;
153 user_len = (passwd_off - 1) - (scheme_len + 3);
154 } else {
155 user_len = norm.len - (scheme_len + 3);
158 strbuf_addch(&norm, '@');
159 url_len -= (++at_ptr - url);
160 url = at_ptr;
165 * Copy the host part excluding any port part, no %-escapes allowed
167 if (!url_len || strchr(":/?#", *url)) {
168 /* Missing host invalid for all URL schemes except file */
169 if (strncmp(norm.buf, "file:", 5)) {
170 if (out_info) {
171 out_info->url = NULL;
172 out_info->err = _("missing host and scheme is not 'file:'");
174 strbuf_release(&norm);
175 return NULL;
177 } else {
178 host_off = norm.len;
180 colon_ptr = slash_ptr - 1;
181 while (colon_ptr > url && *colon_ptr != ':' && *colon_ptr != ']')
182 colon_ptr--;
183 if (*colon_ptr != ':') {
184 colon_ptr = slash_ptr;
185 } else if (!host_off && colon_ptr < slash_ptr && colon_ptr + 1 != slash_ptr) {
186 /* file: URLs may not have a port number */
187 if (out_info) {
188 out_info->url = NULL;
189 out_info->err = _("a 'file:' URL may not have a port number");
191 strbuf_release(&norm);
192 return NULL;
194 spanned = strspn(url, URL_HOST_CHARS);
195 if (spanned < colon_ptr - url) {
196 /* Host name has invalid characters */
197 if (out_info) {
198 out_info->url = NULL;
199 out_info->err = _("invalid characters in host name");
201 strbuf_release(&norm);
202 return NULL;
204 while (url < colon_ptr) {
205 strbuf_addch(&norm, tolower(*url++));
206 url_len--;
211 * Check the port part and copy if not the default (after removing any
212 * leading 0s); no %-escapes allowed
214 if (colon_ptr < slash_ptr) {
215 /* skip the ':' and leading 0s but not the last one if all 0s */
216 url++;
217 url += strspn(url, "0");
218 if (url == slash_ptr && url[-1] == '0')
219 url--;
220 if (url == slash_ptr) {
221 /* Skip ":" port with no number, it's same as default */
222 } else if (slash_ptr - url == 2 &&
223 !strncmp(norm.buf, "http:", 5) &&
224 !strncmp(url, "80", 2)) {
225 /* Skip http :80 as it's the default */
226 } else if (slash_ptr - url == 3 &&
227 !strncmp(norm.buf, "https:", 6) &&
228 !strncmp(url, "443", 3)) {
229 /* Skip https :443 as it's the default */
230 } else {
232 * Port number must be all digits with leading 0s removed
233 * and since all the protocols we deal with have a 16-bit
234 * port number it must also be in the range 1..65535
235 * 0 is not allowed because that means "next available"
236 * on just about every system and therefore cannot be used
238 unsigned long pnum = 0;
239 spanned = strspn(url, URL_DIGIT);
240 if (spanned < slash_ptr - url) {
241 /* port number has invalid characters */
242 if (out_info) {
243 out_info->url = NULL;
244 out_info->err = _("invalid port number");
246 strbuf_release(&norm);
247 return NULL;
249 if (slash_ptr - url <= 5)
250 pnum = strtoul(url, NULL, 10);
251 if (pnum == 0 || pnum > 65535) {
252 /* port number not in range 1..65535 */
253 if (out_info) {
254 out_info->url = NULL;
255 out_info->err = _("invalid port number");
257 strbuf_release(&norm);
258 return NULL;
260 strbuf_addch(&norm, ':');
261 strbuf_add(&norm, url, slash_ptr - url);
262 port_len = slash_ptr - url;
264 url_len -= slash_ptr - colon_ptr;
265 url = slash_ptr;
267 if (host_off)
268 host_len = norm.len - host_off;
272 * Now copy the path resolving any . and .. segments being careful not
273 * to corrupt the URL by unescaping any delimiters, but do add an
274 * initial '/' if it's missing and do normalize any %-escape sequences.
276 path_off = norm.len;
277 path_start = norm.buf + path_off;
278 strbuf_addch(&norm, '/');
279 if (*url == '/') {
280 url++;
281 url_len--;
283 for (;;) {
284 const char *seg_start;
285 size_t seg_start_off = norm.len;
286 const char *next_slash = url + strcspn(url, "/?#");
287 int skip_add_slash = 0;
290 * RFC 3689 indicates that any . or .. segments should be
291 * unescaped before being checked for.
293 if (!append_normalized_escapes(&norm, url, next_slash - url, "",
294 URL_RESERVED)) {
295 if (out_info) {
296 out_info->url = NULL;
297 out_info->err = _("invalid %XX escape sequence");
299 strbuf_release(&norm);
300 return NULL;
303 seg_start = norm.buf + seg_start_off;
304 if (!strcmp(seg_start, ".")) {
305 /* ignore a . segment; be careful not to remove initial '/' */
306 if (seg_start == path_start + 1) {
307 strbuf_setlen(&norm, norm.len - 1);
308 skip_add_slash = 1;
309 } else {
310 strbuf_setlen(&norm, norm.len - 2);
312 } else if (!strcmp(seg_start, "..")) {
314 * ignore a .. segment and remove the previous segment;
315 * be careful not to remove initial '/' from path
317 const char *prev_slash = norm.buf + norm.len - 3;
318 if (prev_slash == path_start) {
319 /* invalid .. because no previous segment to remove */
320 if (out_info) {
321 out_info->url = NULL;
322 out_info->err = _("invalid '..' path segment");
324 strbuf_release(&norm);
325 return NULL;
327 while (*--prev_slash != '/') {}
328 if (prev_slash == path_start) {
329 strbuf_setlen(&norm, prev_slash - norm.buf + 1);
330 skip_add_slash = 1;
331 } else {
332 strbuf_setlen(&norm, prev_slash - norm.buf);
335 url_len -= next_slash - url;
336 url = next_slash;
337 /* if the next char is not '/' done with the path */
338 if (*url != '/')
339 break;
340 url++;
341 url_len--;
342 if (!skip_add_slash)
343 strbuf_addch(&norm, '/');
345 path_len = norm.len - path_off;
349 * Now simply copy the rest, if any, only normalizing %-escapes and
350 * being careful not to corrupt the URL by unescaping any delimiters.
352 if (*url) {
353 if (!append_normalized_escapes(&norm, url, url_len, "", URL_RESERVED)) {
354 if (out_info) {
355 out_info->url = NULL;
356 out_info->err = _("invalid %XX escape sequence");
358 strbuf_release(&norm);
359 return NULL;
364 result = strbuf_detach(&norm, &result_len);
365 if (out_info) {
366 out_info->url = result;
367 out_info->err = NULL;
368 out_info->url_len = result_len;
369 out_info->scheme_len = scheme_len;
370 out_info->user_off = user_off;
371 out_info->user_len = user_len;
372 out_info->passwd_off = passwd_off;
373 out_info->passwd_len = passwd_len;
374 out_info->host_off = host_off;
375 out_info->host_len = host_len;
376 out_info->port_len = port_len;
377 out_info->path_off = path_off;
378 out_info->path_len = path_len;
380 return result;
383 static size_t url_match_prefix(const char *url,
384 const char *url_prefix,
385 size_t url_prefix_len)
388 * url_prefix matches url if url_prefix is an exact match for url or it
389 * is a prefix of url and the match ends on a path component boundary.
390 * Both url and url_prefix are considered to have an implicit '/' on the
391 * end for matching purposes if they do not already.
393 * url must be NUL terminated. url_prefix_len is the length of
394 * url_prefix which need not be NUL terminated.
396 * The return value is the length of the match in characters (including
397 * the final '/' even if it's implicit) or 0 for no match.
399 * Passing NULL as url and/or url_prefix will always cause 0 to be
400 * returned without causing any faults.
402 if (!url || !url_prefix)
403 return 0;
404 if (!url_prefix_len || (url_prefix_len == 1 && *url_prefix == '/'))
405 return (!*url || *url == '/') ? 1 : 0;
406 if (url_prefix[url_prefix_len - 1] == '/')
407 url_prefix_len--;
408 if (strncmp(url, url_prefix, url_prefix_len))
409 return 0;
410 if ((strlen(url) == url_prefix_len) || (url[url_prefix_len] == '/'))
411 return url_prefix_len + 1;
412 return 0;
415 int match_urls(const struct url_info *url,
416 const struct url_info *url_prefix,
417 int *exactusermatch)
420 * url_prefix matches url if the scheme, host and port of url_prefix
421 * are the same as those of url and the path portion of url_prefix
422 * is the same as the path portion of url or it is a prefix that
423 * matches at a '/' boundary. If url_prefix contains a user name,
424 * that must also exactly match the user name in url.
426 * If the user, host, port and path match in this fashion, the returned
427 * value is the length of the path match including any implicit
428 * final '/'. For example, "http://me@example.com/path" is matched by
429 * "http://example.com" with a path length of 1.
431 * If there is a match and exactusermatch is not NULL, then
432 * *exactusermatch will be set to true if both url and url_prefix
433 * contained a user name or false if url_prefix did not have a
434 * user name. If there is no match *exactusermatch is left untouched.
436 int usermatched = 0;
437 int pathmatchlen;
439 if (!url || !url_prefix || !url->url || !url_prefix->url)
440 return 0;
442 /* check the scheme */
443 if (url_prefix->scheme_len != url->scheme_len ||
444 strncmp(url->url, url_prefix->url, url->scheme_len))
445 return 0; /* schemes do not match */
447 /* check the user name if url_prefix has one */
448 if (url_prefix->user_off) {
449 if (!url->user_off || url->user_len != url_prefix->user_len ||
450 strncmp(url->url + url->user_off,
451 url_prefix->url + url_prefix->user_off,
452 url->user_len))
453 return 0; /* url_prefix has a user but it's not a match */
454 usermatched = 1;
457 /* check the host and port */
458 if (url_prefix->host_len != url->host_len ||
459 strncmp(url->url + url->host_off,
460 url_prefix->url + url_prefix->host_off, url->host_len))
461 return 0; /* host names and/or ports do not match */
463 /* check the path */
464 pathmatchlen = url_match_prefix(
465 url->url + url->path_off,
466 url_prefix->url + url_prefix->path_off,
467 url_prefix->url_len - url_prefix->path_off);
469 if (pathmatchlen && exactusermatch)
470 *exactusermatch = usermatched;
471 return pathmatchlen;
474 int urlmatch_config_entry(const char *var, const char *value, void *cb)
476 struct string_list_item *item;
477 struct urlmatch_config *collect = cb;
478 struct urlmatch_item *matched;
479 struct url_info *url = &collect->url;
480 const char *key, *dot;
481 struct strbuf synthkey = STRBUF_INIT;
482 size_t matched_len = 0;
483 int user_matched = 0;
484 int retval;
486 if (!skip_prefix(var, collect->section, &key) || *(key++) != '.') {
487 if (collect->cascade_fn)
488 return collect->cascade_fn(var, value, cb);
489 return 0; /* not interested */
491 dot = strrchr(key, '.');
492 if (dot) {
493 char *config_url, *norm_url;
494 struct url_info norm_info;
496 config_url = xmemdupz(key, dot - key);
497 norm_url = url_normalize(config_url, &norm_info);
498 free(config_url);
499 if (!norm_url)
500 return 0;
501 matched_len = match_urls(url, &norm_info, &user_matched);
502 free(norm_url);
503 if (!matched_len)
504 return 0;
505 key = dot + 1;
508 if (collect->key && strcmp(key, collect->key))
509 return 0;
511 item = string_list_insert(&collect->vars, key);
512 if (!item->util) {
513 matched = xcalloc(1, sizeof(*matched));
514 item->util = matched;
515 } else {
516 matched = item->util;
518 * Is our match shorter? Is our match the same
519 * length, and without user while the current
520 * candidate is with user? Then we cannot use it.
522 if (matched_len < matched->matched_len ||
523 ((matched_len == matched->matched_len) &&
524 (!user_matched && matched->user_matched)))
525 return 0;
526 /* Otherwise, replace it with this one. */
529 matched->matched_len = matched_len;
530 matched->user_matched = user_matched;
531 strbuf_addstr(&synthkey, collect->section);
532 strbuf_addch(&synthkey, '.');
533 strbuf_addstr(&synthkey, key);
534 retval = collect->collect_fn(synthkey.buf, value, collect->cb);
536 strbuf_release(&synthkey);
537 return retval;