1 /* Licensed to the Apache Software Foundation (ASF) under one or more
2 * contributor license agreements. See the NOTICE file distributed with
3 * this work for additional information regarding copyright ownership.
4 * The ASF licenses this file to You under the Apache License, Version 2.0
5 * (the "License"); you may not use this file except in compliance with
6 * the License. You may obtain a copy of the License at
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
18 * util.c: string utility things
21 * 1995-96 Many changes by the Apache Software Foundation
26 * #define DEBUG to trace all cfg_open*()/cfg_closefile() calls
27 * #define DEBUG_CFG_LINES to trace every line read from the config files
31 #include "apr_strings.h"
34 #define APR_WANT_STDIO
35 #define APR_WANT_STRFUNC
42 #include <netdb.h> /* for gethostbyname() */
47 #include "ap_config.h"
48 #include "apr_base64.h"
50 #include "http_main.h"
52 #include "http_protocol.h"
53 #include "http_config.h"
54 #include "util_ebcdic.h"
63 /* A bunch of functions in util.c scan strings looking for certain characters.
64 * To make that more efficient we encode a lookup table. The test_char_table
65 * is generated automatically by gen_test_char.c.
67 #include "test_char.h"
69 /* we assume the folks using this ensure 0 <= c < 256... which means
70 * you need a cast to (unsigned char) first, you can't just plug a
71 * char in here and get it to work, because if char is signed then it
72 * will first be sign extended.
74 #define TEST_CHAR(c, f) (test_char_table[(unsigned)(c)] & (f))
76 /* Win32/NetWare/OS2 need to check for both forward and back slashes
77 * in ap_getparents() and ap_escape_url.
79 #ifdef CASE_BLIND_FILESYSTEM
80 #define IS_SLASH(s) ((s == '/') || (s == '\\'))
82 #define IS_SLASH(s) (s == '/')
87 * Examine a field value (such as a media-/content-type) string and return
88 * it sans any parameters; e.g., strip off any ';charset=foo' and the like.
90 AP_DECLARE(char *) ap_field_noparam(apr_pool_t *p, const char *intype)
94 if (intype == NULL) return NULL;
96 semi = ap_strchr_c(intype, ';');
98 return apr_pstrdup(p, intype);
101 while ((semi > intype) && apr_isspace(semi[-1])) {
104 return apr_pstrndup(p, intype, semi - intype);
108 AP_DECLARE(char *) ap_ht_time(apr_pool_t *p, apr_time_t t, const char *fmt,
112 char ts[MAX_STRING_LEN];
113 char tf[MAX_STRING_LEN];
120 apr_time_exp_gmt(&xt, t);
121 /* Convert %Z to "GMT" and %z to "+0000";
122 * on hosts that do not have a time zone string in struct tm,
123 * strftime must assume its argument is local time.
125 for(strp = tf, f = fmt; strp < tf + sizeof(tf) - 6 && (*strp = *f)
127 if (*f != '%') continue;
138 case 'z': /* common extension */
152 apr_time_exp_lt(&xt, t);
155 /* check return code? */
156 apr_strftime(ts, &retcode, MAX_STRING_LEN, fmt, &xt);
157 ts[MAX_STRING_LEN - 1] = '\0';
158 return apr_pstrdup(p, ts);
161 /* Roy owes Rob beer. */
162 /* Rob owes Roy dinner. */
164 /* These legacy comments would make a lot more sense if Roy hadn't
165 * replaced the old later_than() routine with util_date.c.
167 * Well, okay, they still wouldn't make any sense.
170 /* Match = 0, NoMatch = 1, Abort = -1
171 * Based loosely on sections of wildmat.c by Rich Salz
172 * Hmmm... shouldn't this really go component by component?
174 AP_DECLARE(int) ap_strcmp_match(const char *str, const char *expected)
178 for (x = 0, y = 0; expected[y]; ++y, ++x) {
179 if ((!str[x]) && (expected[y] != '*'))
181 if (expected[y] == '*') {
182 while (expected[++y] == '*');
187 if ((ret = ap_strcmp_match(&str[x++], &expected[y])) != 1)
192 else if ((expected[y] != '?') && (str[x] != expected[y]))
195 return (str[x] != '\0');
198 AP_DECLARE(int) ap_strcasecmp_match(const char *str, const char *expected)
202 for (x = 0, y = 0; expected[y]; ++y, ++x) {
203 if (!str[x] && expected[y] != '*')
205 if (expected[y] == '*') {
206 while (expected[++y] == '*');
211 if ((ret = ap_strcasecmp_match(&str[x++], &expected[y])) != 1)
216 else if (expected[y] != '?'
217 && apr_tolower(str[x]) != apr_tolower(expected[y]))
220 return (str[x] != '\0');
223 /* We actually compare the canonical root to this root, (but we don't
224 * waste time checking the case), since every use of this function in
225 * httpd-2.1 tests if the path is 'proper', meaning we've already passed
226 * it through apr_filepath_merge, or we haven't.
228 AP_DECLARE(int) ap_os_is_path_absolute(apr_pool_t *p, const char *dir)
231 const char *ourdir = dir;
232 if (apr_filepath_root(&newpath, &dir, 0, p) != APR_SUCCESS
233 || strncmp(newpath, ourdir, strlen(newpath)) != 0) {
239 AP_DECLARE(int) ap_is_matchexp(const char *str)
243 for (x = 0; str[x]; x++)
244 if ((str[x] == '*') || (str[x] == '?'))
250 * Here's a pool-based interface to the POSIX-esque ap_regcomp().
251 * Note that we return ap_regex_t instead of being passed one.
252 * The reason is that if you use an already-used ap_regex_t structure,
253 * the memory that you've already allocated gets forgotten, and
254 * regfree() doesn't clear it. So we don't allow it.
257 static apr_status_t regex_cleanup(void *preg)
259 ap_regfree((ap_regex_t *) preg);
263 AP_DECLARE(ap_regex_t *) ap_pregcomp(apr_pool_t *p, const char *pattern,
266 ap_regex_t *preg = apr_palloc(p, sizeof *preg);
268 if (ap_regcomp(preg, pattern, cflags)) {
272 apr_pool_cleanup_register(p, (void *) preg, regex_cleanup,
273 apr_pool_cleanup_null);
278 AP_DECLARE(void) ap_pregfree(apr_pool_t *p, ap_regex_t *reg)
281 apr_pool_cleanup_kill(p, (void *) reg, regex_cleanup);
285 * Similar to standard strstr() but we ignore case in this version.
286 * Based on the strstr() implementation further below.
288 AP_DECLARE(char *) ap_strcasestr(const char *s1, const char *s2)
296 for ( ; (*s1 != '\0') && (apr_tolower(*s1) != apr_tolower(*s2)); s1++);
300 /* found first character of s2, see if the rest matches */
303 for (++p1, ++p2; apr_tolower(*p1) == apr_tolower(*p2); ++p1, ++p2) {
305 /* both strings ended together */
310 /* second string ended, a match */
313 /* didn't find a match here, try starting at next character in s1 */
320 * Returns an offsetted pointer in bigstring immediately after
321 * prefix. Returns bigstring if bigstring doesn't start with
322 * prefix or if prefix is longer than bigstring while still matching.
323 * NOTE: pointer returned is relative to bigstring, so we
324 * can use standard pointer comparisons in the calling function
325 * (eg: test if ap_stripprefix(a,b) == a)
327 AP_DECLARE(const char *) ap_stripprefix(const char *bigstring,
336 while (*p1 && *prefix) {
337 if (*p1++ != *prefix++)
343 /* hit the end of bigstring! */
347 /* This function substitutes for $0-$9, filling in regular expression
348 * submatches. Pass it the same nmatch and pmatch arguments that you
349 * passed ap_regexec(). pmatch should not be greater than the maximum number
350 * of subexpressions - i.e. one more than the re_nsub member of ap_regex_t.
352 * input should be the string with the $-expressions, source should be the
353 * string that was matched against.
355 * It returns the substituted string, or NULL on error.
357 * Parts of this code are based on Henry Spencer's regsub(), from his
358 * AT&T V8 regexp package.
361 AP_DECLARE(char *) ap_pregsub(apr_pool_t *p, const char *input,
362 const char *source, size_t nmatch,
363 ap_regmatch_t pmatch[])
365 const char *src = input;
374 return apr_pstrdup(p, src);
376 /* First pass, find the size */
380 while ((c = *src++) != '\0') {
383 else if (c == '$' && apr_isdigit(*src))
388 if (no > 9) { /* Ordinary character. */
389 if (c == '\\' && (*src == '$' || *src == '&'))
393 else if (no < nmatch && pmatch[no].rm_so < pmatch[no].rm_eo) {
394 len += pmatch[no].rm_eo - pmatch[no].rm_so;
399 dest = dst = apr_pcalloc(p, len + 1);
401 /* Now actually fill in the string */
405 while ((c = *src++) != '\0') {
408 else if (c == '$' && apr_isdigit(*src))
413 if (no > 9) { /* Ordinary character. */
414 if (c == '\\' && (*src == '$' || *src == '&'))
418 else if (no < nmatch && pmatch[no].rm_so < pmatch[no].rm_eo) {
419 len = pmatch[no].rm_eo - pmatch[no].rm_so;
420 memcpy(dst, source + pmatch[no].rm_so, len);
431 * Parse .. so we don't compromise security
433 AP_DECLARE(void) ap_getparents(char *name)
438 /* Four paseses, as per RFC 1808 */
439 /* a) remove ./ path segments */
440 for (next = name; *next && (*next != '.'); next++) {
443 l = w = first_dot = next - name;
444 while (name[l] != '\0') {
445 if (name[l] == '.' && IS_SLASH(name[l + 1])
446 && (l == 0 || IS_SLASH(name[l - 1])))
449 name[w++] = name[l++];
452 /* b) remove trailing . path, segment */
453 if (w == 1 && name[0] == '.')
455 else if (w > 1 && name[w - 1] == '.' && IS_SLASH(name[w - 2]))
459 /* c) remove all xx/../ segments. (including leading ../ and /../) */
462 while (name[l] != '\0') {
463 if (name[l] == '.' && name[l + 1] == '.' && IS_SLASH(name[l + 2])
464 && (l == 0 || IS_SLASH(name[l - 1]))) {
465 register int m = l + 3, n;
469 while (l >= 0 && !IS_SLASH(name[l]))
476 while ((name[n] = name[m]))
483 /* d) remove trailing xx/.. segment. */
484 if (l == 2 && name[0] == '.' && name[1] == '.')
486 else if (l > 2 && name[l - 1] == '.' && name[l - 2] == '.'
487 && IS_SLASH(name[l - 3])) {
490 while (l >= 0 && !IS_SLASH(name[l]))
500 AP_DECLARE(void) ap_no2slash(char *name)
506 #ifdef HAVE_UNC_PATHS
507 /* Check for UNC names. Leave leading two slashes. */
508 if (s[0] == '/' && s[1] == '/')
513 if ((*d++ = *s) == '/') {
527 * copy at most n leading directories of s into d
528 * d should be at least as large as s plus 1 extra byte
530 * the return value is the ever useful pointer to the trailing \0 of d
532 * MODIFIED FOR HAVE_DRIVE_LETTERS and NETWARE environments,
533 * so that if n == 0, "/" is returned in d with n == 1
534 * and s == "e:/test.html", "e:/" is returned in d
535 * *** See also directory_walk in modules/http/http_request.c
538 * /a/b, 0 ==> / (true for all platforms)
547 * c:/a/b 3 ==> c:/a/b
548 * c:/a/b 4 ==> c:/a/b
550 AP_DECLARE(char *) ap_make_dirstr_prefix(char *d, const char *s, int n)
559 if (*s == '\0' || (*s == '/' && (--n) == 0)) {
571 * return the parent directory name including trailing / of the file s
573 AP_DECLARE(char *) ap_make_dirstr_parent(apr_pool_t *p, const char *s)
575 const char *last_slash = ap_strrchr_c(s, '/');
579 if (last_slash == NULL) {
580 return apr_pstrdup(p, "");
582 l = (last_slash - s) + 1;
583 d = apr_palloc(p, l + 1);
590 AP_DECLARE(int) ap_count_dirs(const char *path)
594 for (x = 0, n = 0; path[x]; x++)
600 AP_DECLARE(char *) ap_getword_nc(apr_pool_t *atrans, char **line, char stop)
602 return ap_getword(atrans, (const char **) line, stop);
605 AP_DECLARE(char *) ap_getword(apr_pool_t *atrans, const char **line, char stop)
607 const char *pos = *line;
611 while ((*pos != stop) && *pos) {
616 res = (char *)apr_palloc(atrans, len + 1);
617 memcpy(res, *line, len);
621 while (*pos == stop) {
630 AP_DECLARE(char *) ap_getword_white_nc(apr_pool_t *atrans, char **line)
632 return ap_getword_white(atrans, (const char **) line);
635 AP_DECLARE(char *) ap_getword_white(apr_pool_t *atrans, const char **line)
637 const char *pos = *line;
641 while (!apr_isspace(*pos) && *pos) {
646 res = (char *)apr_palloc(atrans, len + 1);
647 memcpy(res, *line, len);
650 while (apr_isspace(*pos)) {
659 AP_DECLARE(char *) ap_getword_nulls_nc(apr_pool_t *atrans, char **line,
662 return ap_getword_nulls(atrans, (const char **) line, stop);
665 AP_DECLARE(char *) ap_getword_nulls(apr_pool_t *atrans, const char **line,
668 const char *pos = ap_strchr_c(*line, stop);
672 res = apr_pstrdup(atrans, *line);
673 *line += strlen(*line);
677 res = apr_pstrndup(atrans, *line, pos - *line);
686 /* Get a word, (new) config-file style --- quoted strings and backslashes
690 static char *substring_conf(apr_pool_t *p, const char *start, int len,
693 char *result = apr_palloc(p, len + 2);
697 for (i = 0; i < len; ++i) {
698 if (start[i] == '\\' && (start[i + 1] == '\\'
699 || (quote && start[i + 1] == quote)))
700 *resp++ = start[++i];
706 #if RESOLVE_ENV_PER_TOKEN
707 return (char *)ap_resolve_env(p,result);
713 AP_DECLARE(char *) ap_getword_conf_nc(apr_pool_t *p, char **line)
715 return ap_getword_conf(p, (const char **) line);
718 AP_DECLARE(char *) ap_getword_conf(apr_pool_t *p, const char **line)
720 const char *str = *line, *strend;
724 while (*str && apr_isspace(*str))
732 if ((quote = *str) == '"' || quote == '\'') {
734 while (*strend && *strend != quote) {
735 if (*strend == '\\' && strend[1] &&
736 (strend[1] == quote || strend[1] == '\\')) {
743 res = substring_conf(p, str + 1, strend - str - 1, quote);
745 if (*strend == quote)
750 while (*strend && !apr_isspace(*strend))
753 res = substring_conf(p, str, strend - str, 0);
756 while (*strend && apr_isspace(*strend))
762 /* Check a string for any ${ENV} environment variable
763 * construct and replace each them by the value of
764 * that environment variable, if it exists. If the
765 * environment value does not exist, leave the ${ENV}
766 * construct alone; it means something else.
768 AP_DECLARE(const char *) ap_resolve_env(apr_pool_t *p, const char * word)
770 # define SMALL_EXPANSION 5
775 } *result, *current, sresult[SMALL_EXPANSION];
777 const char *s, *e, *ep;
781 s = ap_strchr_c(word, '$');
786 /* well, actually something to do */
787 ep = word + strlen(word);
789 result = current = &(sresult[spc++]);
790 current->next = NULL;
791 current->string = word;
792 current->len = s - word;
793 outlen = current->len;
796 /* prepare next entry */
798 current->next = (spc < SMALL_EXPANSION)
800 : (struct sll *)apr_palloc(p,
801 sizeof(*current->next));
802 current = current->next;
803 current->next = NULL;
808 if (s[1] == '{' && (e = ap_strchr_c(s, '}'))) {
809 word = getenv(apr_pstrndup(p, s+2, e-s-2));
811 current->string = word;
812 current->len = strlen(word);
813 outlen += current->len;
817 current->len = e - s + 1;
818 outlen += current->len;
823 current->string = s++;
830 s = ap_strchr_c(s, '$');
831 current->string = word;
832 current->len = s ? s - word : ep - word;
833 outlen += current->len;
837 /* assemble result */
838 res_buf = cp = apr_palloc(p, outlen + 1);
841 memcpy(cp, result->string, result->len);
844 result = result->next;
846 res_buf[outlen] = '\0';
851 AP_DECLARE(int) ap_cfg_closefile(ap_configfile_t *cfp)
854 ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, NULL,
855 "Done with config file %s", cfp->name);
857 return (cfp->close == NULL) ? 0 : cfp->close(cfp->param);
860 static apr_status_t cfg_close(void *param)
862 apr_file_t *cfp = (apr_file_t *) param;
863 return (apr_file_close(cfp));
866 static int cfg_getch(void *param)
869 apr_file_t *cfp = (apr_file_t *) param;
870 if (apr_file_getc(&ch, cfp) == APR_SUCCESS)
875 static void *cfg_getstr(void *buf, size_t bufsiz, void *param)
877 apr_file_t *cfp = (apr_file_t *) param;
879 rv = apr_file_gets(buf, bufsiz, cfp);
880 if (rv == APR_SUCCESS) {
886 /* Open a ap_configfile_t as FILE, return open ap_configfile_t struct pointer */
887 AP_DECLARE(apr_status_t) ap_pcfg_openfile(ap_configfile_t **ret_cfg,
888 apr_pool_t *p, const char *name)
890 ap_configfile_t *new_cfg;
891 apr_file_t *file = NULL;
899 ap_log_error(APLOG_MARK, APLOG_ERR, 0, NULL,
900 "Internal error: pcfg_openfile() called with NULL filename");
904 status = apr_file_open(&file, name, APR_READ | APR_BUFFERED,
907 ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, NULL,
908 "Opening config file %s (%s)",
909 name, (status != APR_SUCCESS) ?
910 apr_strerror(status, buf, sizeof(buf)) : "successful");
912 if (status != APR_SUCCESS)
915 status = apr_file_info_get(&finfo, APR_FINFO_TYPE, file);
916 if (status != APR_SUCCESS)
919 if (finfo.filetype != APR_REG &&
920 #if defined(WIN32) || defined(OS2) || defined(NETWARE)
921 strcasecmp(apr_filepath_name_get(name), "nul") != 0) {
923 strcmp(name, "/dev/null") != 0) {
924 #endif /* WIN32 || OS2 */
925 ap_log_error(APLOG_MARK, APLOG_ERR, 0, NULL,
926 "Access to file %s denied by server: not a regular file",
928 apr_file_close(file);
933 /* Some twisted character [no pun intended] at MS decided that a
934 * zero width joiner as the lead wide character would be ideal for
935 * describing Unicode text files. This was further convoluted to
936 * another MSism that the same character mapped into utf-8, EF BB BF
937 * would signify utf-8 text files.
939 * Since MS configuration files are all protecting utf-8 encoded
940 * Unicode path, file and resource names, we already have the correct
941 * WinNT encoding. But at least eat the stupid three bytes up front.
944 unsigned char buf[4];
946 status = apr_file_read(file, buf, &len);
947 if ((status != APR_SUCCESS) || (len < 3)
948 || memcmp(buf, "\xEF\xBB\xBF", 3) != 0) {
950 apr_file_seek(file, APR_SET, &zero);
955 new_cfg = apr_palloc(p, sizeof(*new_cfg));
956 new_cfg->param = file;
957 new_cfg->name = apr_pstrdup(p, name);
958 new_cfg->getch = (int (*)(void *)) cfg_getch;
959 new_cfg->getstr = (void *(*)(void *, size_t, void *)) cfg_getstr;
960 new_cfg->close = (int (*)(void *)) cfg_close;
961 new_cfg->line_number = 0;
967 /* Allocate a ap_configfile_t handle with user defined functions and params */
968 AP_DECLARE(ap_configfile_t *) ap_pcfg_open_custom(apr_pool_t *p,
971 int(*getch)(void *param),
972 void *(*getstr) (void *buf, size_t bufsiz, void *param),
973 int(*close_func)(void *param))
975 ap_configfile_t *new_cfg = apr_palloc(p, sizeof(*new_cfg));
977 ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, NULL,
978 "Opening config handler %s", descr);
980 new_cfg->param = param;
981 new_cfg->name = descr;
982 new_cfg->getch = getch;
983 new_cfg->getstr = getstr;
984 new_cfg->close = close_func;
985 new_cfg->line_number = 0;
989 /* Read one character from a configfile_t */
990 AP_DECLARE(int) ap_cfg_getc(ap_configfile_t *cfp)
992 register int ch = cfp->getch(cfp->param);
998 /* Read one line from open ap_configfile_t, strip LF, increase line number */
999 /* If custom handler does not define a getstr() function, read char by char */
1000 AP_DECLARE(int) ap_cfg_getline(char *buf, size_t bufsize, ap_configfile_t *cfp)
1002 /* If a "get string" function is defined, use it */
1003 if (cfp->getstr != NULL) {
1007 size_t cbufsize = bufsize;
1011 if (cfp->getstr(cbuf, cbufsize, cfp->param) == NULL)
1015 * check for line continuation,
1016 * i.e. match [^\\]\\[\r]\n only
1019 while (cp < cbuf+cbufsize && *cp != '\0')
1021 if (cp > cbuf && cp[-1] == LF) {
1023 if (cp > cbuf && cp[-1] == CR)
1025 if (cp > cbuf && cp[-1] == '\\') {
1027 if (!(cp > cbuf && cp[-1] == '\\')) {
1029 * line continuation requested -
1030 * then remove backslash and continue
1032 cbufsize -= (cp-cbuf);
1038 * no real continuation because escaped -
1039 * then just remove escape character
1041 for ( ; cp < cbuf+cbufsize && *cp != '\0'; cp++)
1050 * Leading and trailing white space is eliminated completely
1053 while (apr_isspace(*src))
1055 /* blast trailing whitespace */
1056 dst = &src[strlen(src)];
1057 while (--dst >= src && apr_isspace(*dst))
1059 /* Zap leading whitespace by shifting */
1061 for (dst = buf; (*dst++ = *src++) != '\0'; )
1064 #ifdef DEBUG_CFG_LINES
1065 ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, NULL, "Read config: %s", buf);
1069 /* No "get string" function defined; read character by character */
1071 register size_t i = 0;
1074 /* skip leading whitespace */
1076 c = cfp->getch(cfp->param);
1077 } while (c == '\t' || c == ' ');
1083 /* too small, assume caller is crazy */
1088 if ((c == '\t') || (c == ' ')) {
1090 while ((c == '\t') || (c == ' '))
1091 c = cfp->getch(cfp->param);
1094 /* silently ignore CR (_assume_ that a LF follows) */
1095 c = cfp->getch(cfp->param);
1098 /* increase line number and return on LF */
1101 if (c == EOF || c == 0x4 || c == LF || i >= (bufsize - 2)) {
1103 * check for line continuation
1105 if (i > 0 && buf[i-1] == '\\') {
1107 if (!(i > 0 && buf[i-1] == '\\')) {
1108 /* line is continued */
1109 c = cfp->getch(cfp->param);
1112 /* else nothing needs be done because
1113 * then the backslash is escaped and
1114 * we just strip to a single one
1117 /* blast trailing whitespace */
1118 while (i > 0 && apr_isspace(buf[i - 1]))
1121 #ifdef DEBUG_CFG_LINES
1122 ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, NULL,
1123 "Read config: %s", buf);
1129 c = cfp->getch(cfp->param);
1134 /* Size an HTTP header field list item, as separated by a comma.
1135 * The return value is a pointer to the beginning of the non-empty list item
1136 * within the original string (or NULL if there is none) and the address
1137 * of field is shifted to the next non-comma, non-whitespace character.
1138 * len is the length of the item excluding any beginning whitespace.
1140 AP_DECLARE(const char *) ap_size_list_item(const char **field, int *len)
1142 const unsigned char *ptr = (const unsigned char *)*field;
1143 const unsigned char *token;
1144 int in_qpair, in_qstr, in_com;
1146 /* Find first non-comma, non-whitespace byte */
1148 while (*ptr == ',' || apr_isspace(*ptr))
1153 /* Find the end of this item, skipping over dead bits */
1155 for (in_qpair = in_qstr = in_com = 0;
1156 *ptr && (in_qpair || in_qstr || in_com || *ptr != ',');
1164 case '\\': in_qpair = 1; /* quoted-pair */
1166 case '"' : if (!in_com) /* quoted string delim */
1169 case '(' : if (!in_qstr) /* comment (may nest) */
1172 case ')' : if (in_com) /* end comment */
1180 if ((*len = (ptr - token)) == 0) {
1181 *field = (const char *)ptr;
1185 /* Advance field pointer to the next non-comma, non-white byte */
1187 while (*ptr == ',' || apr_isspace(*ptr))
1190 *field = (const char *)ptr;
1191 return (const char *)token;
1194 /* Retrieve an HTTP header field list item, as separated by a comma,
1195 * while stripping insignificant whitespace and lowercasing anything not in
1196 * a quoted string or comment. The return value is a new string containing
1197 * the converted list item (or NULL if none) and the address pointed to by
1198 * field is shifted to the next non-comma, non-whitespace.
1200 AP_DECLARE(char *) ap_get_list_item(apr_pool_t *p, const char **field)
1202 const char *tok_start;
1203 const unsigned char *ptr;
1206 int addspace = 0, in_qpair = 0, in_qstr = 0, in_com = 0, tok_len = 0;
1208 /* Find the beginning and maximum length of the list item so that
1209 * we can allocate a buffer for the new string and reset the field.
1211 if ((tok_start = ap_size_list_item(field, &tok_len)) == NULL) {
1214 token = apr_palloc(p, tok_len + 1);
1216 /* Scan the token again, but this time copy only the good bytes.
1217 * We skip extra whitespace and any whitespace around a '=', '/',
1218 * or ';' and lowercase normal characters not within a comment,
1219 * quoted-string or quoted-pair.
1221 for (ptr = (const unsigned char *)tok_start, pos = (unsigned char *)token;
1222 *ptr && (in_qpair || in_qstr || in_com || *ptr != ',');
1231 case '\\': in_qpair = 1;
1237 case '"' : if (!in_com)
1244 case '(' : if (!in_qstr)
1251 case ')' : if (in_com)
1257 case '\t': if (addspace)
1259 if (in_com || in_qstr)
1266 case ';' : if (!(in_com || in_qstr))
1270 default : if (addspace == 1)
1272 *pos++ = (in_com || in_qstr) ? *ptr
1273 : apr_tolower(*ptr);
1284 /* Find an item in canonical form (lowercase, no extra spaces) within
1285 * an HTTP field value list. Returns 1 if found, 0 if not found.
1286 * This would be much more efficient if we stored header fields as
1287 * an array of list items as they are received instead of a plain string.
1289 AP_DECLARE(int) ap_find_list_item(apr_pool_t *p, const char *line,
1292 const unsigned char *pos;
1293 const unsigned char *ptr = (const unsigned char *)line;
1294 int good = 0, addspace = 0, in_qpair = 0, in_qstr = 0, in_com = 0;
1299 do { /* loop for each item in line's list */
1301 /* Find first non-comma, non-whitespace byte */
1303 while (*ptr == ',' || apr_isspace(*ptr))
1307 good = 1; /* until proven otherwise for this item */
1309 break; /* no items left and nothing good found */
1311 /* We skip extra whitespace and any whitespace around a '=', '/',
1312 * or ';' and lowercase normal characters not within a comment,
1313 * quoted-string or quoted-pair.
1315 for (pos = (const unsigned char *)tok;
1316 *ptr && (in_qpair || in_qstr || in_com || *ptr != ',');
1322 good = (*pos++ == *ptr);
1326 case '\\': in_qpair = 1;
1328 good = good && (*pos++ == ' ');
1329 good = good && (*pos++ == *ptr);
1332 case '"' : if (!in_com)
1335 good = good && (*pos++ == ' ');
1336 good = good && (*pos++ == *ptr);
1339 case '(' : if (!in_qstr)
1342 good = good && (*pos++ == ' ');
1343 good = good && (*pos++ == *ptr);
1346 case ')' : if (in_com)
1348 good = good && (*pos++ == *ptr);
1352 case '\t': if (addspace || !good)
1354 if (in_com || in_qstr)
1355 good = (*pos++ == *ptr);
1361 case ';' : if (!(in_com || in_qstr))
1363 good = good && (*pos++ == *ptr);
1365 default : if (!good)
1368 good = (*pos++ == ' ');
1369 if (in_com || in_qstr)
1370 good = good && (*pos++ == *ptr);
1372 good = good && (*pos++ == apr_tolower(*ptr));
1379 good = 0; /* not good if only a prefix was matched */
1381 } while (*ptr && !good);
1387 /* Retrieve a token, spacing over it and returning a pointer to
1388 * the first non-white byte afterwards. Note that these tokens
1389 * are delimited by semis and commas; and can also be delimited
1390 * by whitespace at the caller's option.
1393 AP_DECLARE(char *) ap_get_token(apr_pool_t *p, const char **accept_line,
1396 const char *ptr = *accept_line;
1397 const char *tok_start;
1401 /* Find first non-white byte */
1403 while (*ptr && apr_isspace(*ptr))
1408 /* find token end, skipping over quoted strings.
1409 * (comments are already gone).
1412 while (*ptr && (accept_white || !apr_isspace(*ptr))
1413 && *ptr != ';' && *ptr != ',') {
1420 tok_len = ptr - tok_start;
1421 token = apr_pstrndup(p, tok_start, tok_len);
1423 /* Advance accept_line pointer to the next non-white byte */
1425 while (*ptr && apr_isspace(*ptr))
1433 /* find http tokens, see the definition of token from RFC2068 */
1434 AP_DECLARE(int) ap_find_token(apr_pool_t *p, const char *line, const char *tok)
1436 const unsigned char *start_token;
1437 const unsigned char *s;
1442 s = (const unsigned char *)line;
1444 /* find start of token, skip all stop characters, note NUL
1445 * isn't a token stop, so we don't need to test for it
1447 while (TEST_CHAR(*s, T_HTTP_TOKEN_STOP)) {
1454 /* find end of the token */
1455 while (*s && !TEST_CHAR(*s, T_HTTP_TOKEN_STOP)) {
1458 if (!strncasecmp((const char *)start_token, (const char *)tok,
1469 AP_DECLARE(int) ap_find_last_token(apr_pool_t *p, const char *line,
1472 int llen, tlen, lidx;
1477 llen = strlen(line);
1482 (lidx > 0 && !(apr_isspace(line[lidx - 1]) || line[lidx - 1] == ',')))
1485 return (strncasecmp(&line[lidx], tok, tlen) == 0);
1488 AP_DECLARE(char *) ap_escape_shell_cmd(apr_pool_t *p, const char *str)
1492 const unsigned char *s;
1494 cmd = apr_palloc(p, 2 * strlen(str) + 1); /* Be safe */
1495 d = (unsigned char *)cmd;
1496 s = (const unsigned char *)str;
1499 #if defined(OS2) || defined(WIN32)
1501 * Newlines to Win32/OS2 CreateProcess() are ill advised.
1502 * Convert them to spaces since they are effectively white
1503 * space to most applications
1505 if (*s == '\r' || *s == '\n') {
1511 if (TEST_CHAR(*s, T_ESCAPE_SHELL_CMD)) {
1521 static char x2c(const char *what)
1523 register char digit;
1525 #if !APR_CHARSET_EBCDIC
1526 digit = ((what[0] >= 'A') ? ((what[0] & 0xdf) - 'A') + 10
1529 digit += (what[1] >= 'A' ? ((what[1] & 0xdf) - 'A') + 10
1531 #else /*APR_CHARSET_EBCDIC*/
1538 digit = apr_xlate_conv_byte(ap_hdrs_from_ascii,
1539 0xFF & strtol(xstr, NULL, 16));
1540 #endif /*APR_CHARSET_EBCDIC*/
1545 * Unescapes a URL, leaving reserved characters intact.
1546 * Returns 0 on success, non-zero on error
1548 * bad % escape returns HTTP_BAD_REQUEST
1550 * decoding %00 or a forbidden character returns HTTP_NOT_FOUND
1553 static int unescape_url(char *url, const char *forbid, const char *reserved)
1555 register int badesc, badpath;
1560 /* Initial scan for first '%'. Don't bother writing values before
1562 y = strchr(url, '%');
1566 for (x = y; *y; ++x, ++y) {
1571 if (!apr_isxdigit(*(y + 1)) || !apr_isxdigit(*(y + 2))) {
1577 decoded = x2c(y + 1);
1578 if ((decoded == '\0')
1579 || (forbid && ap_strchr_c(forbid, decoded))) {
1584 else if (reserved && ap_strchr_c(reserved, decoded)) {
1598 return HTTP_BAD_REQUEST;
1601 return HTTP_NOT_FOUND;
1607 AP_DECLARE(int) ap_unescape_url(char *url)
1610 #ifdef CASE_BLIND_FILESYSTEM
1611 return unescape_url(url, "/\\", NULL);
1613 return unescape_url(url, "/", NULL);
1616 AP_DECLARE(int) ap_unescape_url_keep2f(char *url)
1618 /* AllowEncodedSlashes (corrected) */
1619 return unescape_url(url, NULL, "/");
1622 /* IFDEF these out until they've been thought through.
1623 * Just a germ of an API extension for now
1625 AP_DECLARE(int) ap_unescape_url_proxy(char *url)
1627 /* leave RFC1738 reserved characters intact, * so proxied URLs
1628 * don't get mangled. Where does that leave encoded '&' ?
1630 return unescape_url(url, NULL, "/;?");
1632 AP_DECLARE(int) ap_unescape_url_reserved(char *url, const char *reserved)
1634 return unescape_url(url, NULL, reserved);
1638 AP_DECLARE(char *) ap_construct_server(apr_pool_t *p, const char *hostname,
1639 apr_port_t port, const request_rec *r)
1641 if (ap_is_default_port(port, r)) {
1642 return apr_pstrdup(p, hostname);
1645 return apr_psprintf(p, "%s:%u", hostname, port);
1649 AP_DECLARE(int) ap_unescape_all(char *url)
1651 return unescape_url(url, NULL, NULL);
1654 /* c2x takes an unsigned, and expects the caller has guaranteed that
1655 * 0 <= what < 256... which usually means that you have to cast to
1656 * unsigned char first, because (unsigned)(char)(x) first goes through
1657 * signed extension to an int before the unsigned cast.
1659 * The reason for this assumption is to assist gcc code generation --
1660 * the unsigned char -> unsigned extension is already done earlier in
1661 * both uses of this code, so there's no need to waste time doing it
1664 static const char c2x_table[] = "0123456789abcdef";
1666 static APR_INLINE unsigned char *c2x(unsigned what, unsigned char prefix,
1667 unsigned char *where)
1669 #if APR_CHARSET_EBCDIC
1670 what = apr_xlate_conv_byte(ap_hdrs_to_ascii, (unsigned char)what);
1671 #endif /*APR_CHARSET_EBCDIC*/
1673 *where++ = c2x_table[what >> 4];
1674 *where++ = c2x_table[what & 0xf];
1679 * escape_path_segment() escapes a path segment, as defined in RFC 1808. This
1680 * routine is (should be) OS independent.
1682 * os_escape_path() converts an OS path to a URL, in an OS dependent way. In all
1683 * cases if a ':' occurs before the first '/' in the URL, the URL should be
1684 * prefixed with "./" (or the ':' escaped). In the case of Unix, this means
1685 * leaving '/' alone, but otherwise doing what escape_path_segment() does. For
1686 * efficiency reasons, we don't use escape_path_segment(), which is provided for
1687 * reference. Again, RFC 1808 is where this stuff is defined.
1689 * If partial is set, os_escape_path() assumes that the path will be appended to
1690 * something with a '/' in it (and thus does not prefix "./").
1693 AP_DECLARE(char *) ap_escape_path_segment_b(char *copy, const char *segment)
1695 const unsigned char *s = (const unsigned char *)segment;
1696 unsigned char *d = (unsigned char *)copy;
1700 if (TEST_CHAR(c, T_ESCAPE_PATH_SEGMENT)) {
1712 AP_DECLARE(char *) ap_escape_path_segment(apr_pool_t *p, const char *segment)
1714 return ap_escape_path_segment_b(apr_palloc(p, 3 * strlen(segment) + 1), segment);
1717 AP_DECLARE(char *) ap_os_escape_path(apr_pool_t *p, const char *path, int partial)
1719 char *copy = apr_palloc(p, 3 * strlen(path) + 3);
1720 const unsigned char *s = (const unsigned char *)path;
1721 unsigned char *d = (unsigned char *)copy;
1725 const char *colon = ap_strchr_c(path, ':');
1726 const char *slash = ap_strchr_c(path, '/');
1728 if (colon && (!slash || colon < slash)) {
1734 if (TEST_CHAR(c, T_OS_ESCAPE_PATH)) {
1746 /* ap_escape_uri is now a macro for os_escape_path */
1748 AP_DECLARE(char *) ap_escape_html(apr_pool_t *p, const char *s)
1753 /* first, count the number of extra characters */
1754 for (i = 0, j = 0; s[i] != '\0'; i++)
1755 if (s[i] == '<' || s[i] == '>')
1757 else if (s[i] == '&')
1759 else if (s[i] == '"')
1763 return apr_pstrmemdup(p, s, i);
1765 x = apr_palloc(p, i + j + 1);
1766 for (i = 0, j = 0; s[i] != '\0'; i++, j++)
1768 memcpy(&x[j], "<", 4);
1771 else if (s[i] == '>') {
1772 memcpy(&x[j], ">", 4);
1775 else if (s[i] == '&') {
1776 memcpy(&x[j], "&", 5);
1779 else if (s[i] == '"') {
1780 memcpy(&x[j], """, 6);
1790 AP_DECLARE(char *) ap_escape_logitem(apr_pool_t *p, const char *str)
1794 const unsigned char *s;
1800 ret = apr_palloc(p, 4 * strlen(str) + 1); /* Be safe */
1801 d = (unsigned char *)ret;
1802 s = (const unsigned char *)str;
1805 if (TEST_CHAR(*s, T_ESCAPE_LOGITEM)) {
1841 AP_DECLARE(apr_size_t) ap_escape_errorlog_item(char *dest, const char *source,
1844 unsigned char *d, *ep;
1845 const unsigned char *s;
1847 if (!source || !buflen) { /* be safe */
1851 d = (unsigned char *)dest;
1852 s = (const unsigned char *)source;
1853 ep = d + buflen - 1;
1855 for (; d < ep && *s; ++s) {
1857 if (TEST_CHAR(*s, T_ESCAPE_LOGITEM)) {
1883 case '"': /* no need for this in error log */
1888 ep = --d; /* break the for loop as well */
1901 return (d - (unsigned char *)dest);
1904 AP_DECLARE(int) ap_is_directory(apr_pool_t *p, const char *path)
1908 if (apr_stat(&finfo, path, APR_FINFO_TYPE, p) != APR_SUCCESS)
1909 return 0; /* in error condition, just return no */
1911 return (finfo.filetype == APR_DIR);
1914 AP_DECLARE(int) ap_is_rdirectory(apr_pool_t *p, const char *path)
1918 if (apr_stat(&finfo, path, APR_FINFO_LINK | APR_FINFO_TYPE, p) != APR_SUCCESS)
1919 return 0; /* in error condition, just return no */
1921 return (finfo.filetype == APR_DIR);
1924 AP_DECLARE(char *) ap_make_full_path(apr_pool_t *a, const char *src1,
1927 apr_size_t len1, len2;
1930 len1 = strlen(src1);
1931 len2 = strlen(src2);
1932 /* allocate +3 for '/' delimiter, trailing NULL and overallocate
1933 * one extra byte to allow the caller to add a trailing '/'
1935 path = (char *)apr_palloc(a, len1 + len2 + 3);
1938 memcpy(path + 1, src2, len2 + 1);
1942 memcpy(path, src1, len1);
1944 if (next[-1] != '/') {
1947 memcpy(next, src2, len2 + 1);
1953 * Check for an absoluteURI syntax (see section 3.2 in RFC2068).
1955 AP_DECLARE(int) ap_is_url(const char *u)
1959 for (x = 0; u[x] != ':'; x++) {
1961 ((!apr_isalpha(u[x])) && (!apr_isdigit(u[x])) &&
1962 (u[x] != '+') && (u[x] != '-') && (u[x] != '.'))) {
1967 return (x ? 1 : 0); /* If the first character is ':', it's broken, too */
1970 AP_DECLARE(int) ap_ind(const char *s, char c)
1972 const char *p = ap_strchr_c(s, c);
1979 AP_DECLARE(int) ap_rind(const char *s, char c)
1981 const char *p = ap_strrchr_c(s, c);
1988 AP_DECLARE(void) ap_str_tolower(char *str)
1991 *str = apr_tolower(*str);
1997 * We must return a FQDN
1999 char *ap_get_local_host(apr_pool_t *a)
2001 #ifndef MAXHOSTNAMELEN
2002 #define MAXHOSTNAMELEN 256
2004 char str[MAXHOSTNAMELEN + 1];
2005 char *server_hostname = NULL;
2006 apr_sockaddr_t *sockaddr;
2009 if (apr_gethostname(str, sizeof(str) - 1, a) != APR_SUCCESS) {
2010 ap_log_perror(APLOG_MARK, APLOG_STARTUP | APLOG_WARNING, 0, a,
2011 "%s: apr_gethostname() failed to determine ServerName",
2014 str[sizeof(str) - 1] = '\0';
2015 if (apr_sockaddr_info_get(&sockaddr, str, APR_UNSPEC, 0, 0, a) == APR_SUCCESS) {
2016 if ( (apr_getnameinfo(&hostname, sockaddr, 0) == APR_SUCCESS) &&
2017 (ap_strchr_c(hostname, '.')) ) {
2018 server_hostname = apr_pstrdup(a, hostname);
2019 return server_hostname;
2020 } else if (ap_strchr_c(str, '.')) {
2021 server_hostname = apr_pstrdup(a, str);
2023 apr_sockaddr_ip_get(&hostname, sockaddr);
2024 server_hostname = apr_pstrdup(a, hostname);
2027 ap_log_perror(APLOG_MARK, APLOG_STARTUP | APLOG_WARNING, 0, a,
2028 "%s: apr_sockaddr_info_get() failed for %s",
2029 ap_server_argv0, str);
2033 if (!server_hostname)
2034 server_hostname = apr_pstrdup(a, "127.0.0.1");
2036 ap_log_perror(APLOG_MARK, APLOG_ALERT|APLOG_STARTUP, 0, a,
2037 "%s: Could not reliably determine the server's fully qualified "
2038 "domain name, using %s for ServerName",
2039 ap_server_argv0, server_hostname);
2041 return server_hostname;
2044 /* simple 'pool' alloc()ing glue to apr_base64.c
2046 AP_DECLARE(char *) ap_pbase64decode(apr_pool_t *p, const char *bufcoded)
2051 decoded = (char *) apr_palloc(p, 1 + apr_base64_decode_len(bufcoded));
2052 l = apr_base64_decode(decoded, bufcoded);
2053 decoded[l] = '\0'; /* make binary sequence into string */
2058 AP_DECLARE(char *) ap_pbase64encode(apr_pool_t *p, char *string)
2061 int l = strlen(string);
2063 encoded = (char *) apr_palloc(p, 1 + apr_base64_encode_len(l));
2064 l = apr_base64_encode(encoded, string, l);
2065 encoded[l] = '\0'; /* make binary sequence into string */
2070 /* we want to downcase the type/subtype for comparison purposes
2071 * but nothing else because ;parameter=foo values are case sensitive.
2072 * XXX: in truth we want to downcase parameter names... but really,
2073 * apache has never handled parameters and such correctly. You
2074 * also need to compress spaces and such to be able to compare
2077 AP_DECLARE(void) ap_content_type_tolower(char *str)
2081 semi = strchr(str, ';');
2086 ap_str_tolower(str);
2094 * Given a string, replace any bare " with \" .
2096 AP_DECLARE(char *) ap_escape_quotes(apr_pool_t *p, const char *instring)
2099 const char *inchr = instring;
2100 char *outchr, *outstring;
2103 * Look through the input string, jogging the length of the output
2104 * string up by an extra byte each time we find an unescaped ".
2106 while (*inchr != '\0') {
2108 if (*inchr == '"') {
2112 * If we find a slosh, and it's not the last byte in the string,
2113 * it's escaping something - advance past both bytes.
2115 if ((*inchr == '\\') && (inchr[1] != '\0')) {
2121 outstring = apr_palloc(p, newlen + 1);
2125 * Now copy the input string to the output string, inserting a slosh
2126 * in front of every " that doesn't already have one.
2128 while (*inchr != '\0') {
2129 if ((*inchr == '\\') && (inchr[1] != '\0')) {
2130 *outchr++ = *inchr++;
2131 *outchr++ = *inchr++;
2133 if (*inchr == '"') {
2136 if (*inchr != '\0') {
2137 *outchr++ = *inchr++;
2145 * Given a string, append the PID deliminated by delim.
2146 * Usually used to create a pid-appended filepath name
2147 * (eg: /a/b/foo -> /a/b/foo.6726). A function, and not
2148 * a macro, to avoid unistd.h dependency
2150 AP_DECLARE(char *) ap_append_pid(apr_pool_t *p, const char *string,
2153 return apr_psprintf(p, "%s%s%" APR_PID_T_FMT, string,