2 * Copyright (c) 2000-2004 Dag-Erling Coïdan Smørgrav
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer
10 * in this position and unchanged.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * 3. The name of the author may not be used to endorse or promote products
15 * derived from this software without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
33 * The following copyright applies to the base64 code:
36 * Copyright 1997 Massachusetts Institute of Technology
38 * Permission to use, copy, modify, and distribute this software and
39 * its documentation for any purpose and without fee is hereby
40 * granted, provided that both the above copyright notice and this
41 * permission notice appear in all copies, that both the above
42 * copyright notice and this permission notice appear in all
43 * supporting documentation, and that the name of M.I.T. not be used
44 * in advertising or publicity pertaining to distribution of the
45 * software without specific, written prior permission. M.I.T. makes
46 * no representations about the suitability of this software for any
47 * purpose. It is provided "as is" without express or implied
50 * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''. M.I.T. DISCLAIMS
51 * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE,
52 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
53 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT
54 * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
55 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
56 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
57 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
58 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
59 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
60 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64 #include <sys/param.h>
65 #include <sys/socket.h>
81 #include <netinet/in.h>
82 #include <netinet/tcp.h>
88 /* Maximum number of redirects to follow */
89 #define MAX_REDIRECT 5
91 /* Symbolic names for reply codes we care about */
93 #define HTTP_PARTIAL 206
94 #define HTTP_MOVED_PERM 301
95 #define HTTP_MOVED_TEMP 302
96 #define HTTP_SEE_OTHER 303
97 #define HTTP_NOT_MODIFIED 304
98 #define HTTP_TEMP_REDIRECT 307
99 #define HTTP_NEED_AUTH 401
100 #define HTTP_NEED_PROXY_AUTH 407
101 #define HTTP_BAD_RANGE 416
102 #define HTTP_PROTOCOL_ERROR 999
104 #define HTTP_REDIRECT(xyz) ((xyz) == HTTP_MOVED_PERM \
105 || (xyz) == HTTP_MOVED_TEMP \
106 || (xyz) == HTTP_TEMP_REDIRECT \
107 || (xyz) == HTTP_SEE_OTHER)
109 #define HTTP_ERROR(xyz) ((xyz) > 400 && (xyz) < 599)
112 /*****************************************************************************
113 * I/O functions for decoding chunked streams
118 conn_t *conn; /* connection */
119 int chunked; /* chunked mode */
120 char *buf; /* chunk buffer */
121 size_t bufsize; /* size of chunk buffer */
122 ssize_t buflen; /* amount of data currently in buffer */
123 int bufpos; /* current read offset in buffer */
124 int eof; /* end-of-file flag */
125 int error; /* error flag */
126 size_t chunksize; /* remaining size of current chunk */
133 * Get next chunk header
136 http_new_chunk(struct httpio *io)
140 if (fetch_getln(io->conn) == -1)
143 if (io->conn->buflen < 2 || !isxdigit((unsigned char)*io->conn->buf))
146 for (p = io->conn->buf; *p && !isspace((unsigned char)*p); ++p) {
149 if (!isxdigit((unsigned char)*p))
151 if (isdigit((unsigned char)*p)) {
152 io->chunksize = io->chunksize * 16 +
155 io->chunksize = io->chunksize * 16 +
156 10 + tolower((unsigned char)*p) - 'a';
162 io->total += io->chunksize;
163 if (io->chunksize == 0)
164 fprintf(stderr, "%s(): end of last chunk\n", __func__);
166 fprintf(stderr, "%s(): new chunk: %lu (%lu)\n",
167 __func__, (unsigned long)io->chunksize,
168 (unsigned long)io->total);
172 return (io->chunksize);
176 * Grow the input buffer to at least len bytes
179 http_growbuf(struct httpio *io, size_t len)
183 if (io->bufsize >= len)
186 if ((tmp = realloc(io->buf, len)) == NULL)
194 * Fill the input buffer, do chunk decoding on the fly
197 http_fillbuf(struct httpio *io, size_t len)
204 if (io->chunked == 0) {
205 if (http_growbuf(io, len) == -1)
207 if ((io->buflen = fetch_read(io->conn, io->buf, len)) == -1) {
215 if (io->chunksize == 0) {
216 switch (http_new_chunk(io)) {
226 if (len > io->chunksize)
228 if (http_growbuf(io, len) == -1)
230 if ((io->buflen = fetch_read(io->conn, io->buf, len)) == -1) {
234 io->chunksize -= io->buflen;
236 if (io->chunksize == 0) {
239 if (fetch_read(io->conn, endl, 2) != 2 ||
240 endl[0] != '\r' || endl[1] != '\n')
253 http_readfn(void *v, char *buf, int len)
255 struct httpio *io = (struct httpio *)v;
263 for (pos = 0; len > 0; pos += l, len -= l) {
265 if (!io->buf || io->bufpos == io->buflen)
266 if (http_fillbuf(io, len) < 1)
268 l = io->buflen - io->bufpos;
271 memcpy(buf + pos, io->buf + io->bufpos, l);
275 if (!pos && io->error)
284 http_writefn(void *v, const char *buf, int len)
286 struct httpio *io = (struct httpio *)v;
288 return (fetch_write(io->conn, buf, len));
295 http_closefn(void *v)
297 struct httpio *io = (struct httpio *)v;
300 r = fetch_close(io->conn);
308 * Wrap a file descriptor up
311 http_funopen(conn_t *conn, int chunked)
316 if ((io = calloc(1, sizeof(*io))) == NULL) {
321 io->chunked = chunked;
322 f = funopen(io, http_readfn, http_writefn, NULL, http_closefn);
332 /*****************************************************************************
333 * Helper functions for talking to the server and parsing its replies
346 hdr_transfer_encoding,
347 hdr_www_authenticate,
348 hdr_proxy_authenticate,
351 /* Names of interesting headers */
356 { hdr_content_length, "Content-Length" },
357 { hdr_content_range, "Content-Range" },
358 { hdr_last_modified, "Last-Modified" },
359 { hdr_location, "Location" },
360 { hdr_transfer_encoding, "Transfer-Encoding" },
361 { hdr_www_authenticate, "WWW-Authenticate" },
362 { hdr_proxy_authenticate, "Proxy-Authenticate" },
363 { hdr_unknown, NULL },
367 * Send a formatted line; optionally echo to terminal
370 http_cmd(conn_t *conn, const char *fmt, ...)
378 len = vasprintf(&msg, fmt, ap);
387 r = fetch_putln(conn, msg, len);
399 * Get and parse status line
402 http_get_reply(conn_t *conn)
406 if (fetch_getln(conn) == -1)
409 * A valid status line looks like "HTTP/m.n xyz reason" where m
410 * and n are the major and minor protocol version numbers and xyz
412 * Unfortunately, there are servers out there (NCSA 1.5.1, to name
413 * just one) that do not send a version number, so we can't rely
414 * on finding one, but if we do, insist on it being 1.0 or 1.1.
415 * We don't care about the reason phrase.
417 if (strncmp(conn->buf, "HTTP", 4) != 0)
418 return (HTTP_PROTOCOL_ERROR);
421 if (p[1] != '1' || p[2] != '.' || (p[3] != '0' && p[3] != '1'))
422 return (HTTP_PROTOCOL_ERROR);
426 !isdigit((unsigned char)p[1]) ||
427 !isdigit((unsigned char)p[2]) ||
428 !isdigit((unsigned char)p[3]))
429 return (HTTP_PROTOCOL_ERROR);
431 conn->err = (p[1] - '0') * 100 + (p[2] - '0') * 10 + (p[3] - '0');
436 * Check a header; if the type matches the given string, return a pointer
437 * to the beginning of the value.
440 http_match(const char *str, const char *hdr)
442 while (*str && *hdr &&
443 tolower((unsigned char)*str++) == tolower((unsigned char)*hdr++))
445 if (*str || *hdr != ':')
447 while (*hdr && isspace((unsigned char)*++hdr))
454 * Get the next header and return the appropriate symbolic code. We
455 * need to read one line ahead for checking for a continuation line
456 * belonging to the current header (continuation lines start with
459 * We get called with a fresh line already in the conn buffer, either
460 * from the previous http_next_header() invocation, or, the first
461 * time, from a fetch_getln() performed by our caller.
463 * This stops when we encounter an empty line (we dont read beyond the header
466 * Note that the "headerbuf" is just a place to return the result. Its
467 * contents are not used for the next call. This means that no cleanup
468 * is needed when ie doing another connection, just call the cleanup when
469 * fully done to deallocate memory.
472 /* Limit the max number of continuation lines to some reasonable value */
473 #define HTTP_MAX_CONT_LINES 10
475 /* Place into which to build a header from one or several lines */
477 char *buf; /* buffer */
478 size_t bufsize; /* buffer size */
479 size_t buflen; /* length of buffer contents */
483 init_http_headerbuf(http_headerbuf_t *buf)
491 clean_http_headerbuf(http_headerbuf_t *buf)
495 init_http_headerbuf(buf);
498 /* Remove whitespace at the end of the buffer */
500 http_conn_trimright(conn_t *conn)
502 while (conn->buflen &&
503 isspace((unsigned char)conn->buf[conn->buflen - 1]))
505 conn->buf[conn->buflen] = '\0';
509 http_next_header(conn_t *conn, http_headerbuf_t *hbuf, const char **p)
514 * Have to do the stripping here because of the first line. So
515 * it's done twice for the subsequent lines. No big deal
517 http_conn_trimright(conn);
518 if (conn->buflen == 0)
521 /* Copy the line to the headerbuf */
522 if (hbuf->bufsize < conn->buflen + 1) {
523 if ((hbuf->buf = realloc(hbuf->buf, conn->buflen + 1)) == NULL)
524 return (hdr_syserror);
525 hbuf->bufsize = conn->buflen + 1;
527 strcpy(hbuf->buf, conn->buf);
528 hbuf->buflen = conn->buflen;
531 * Fetch possible continuation lines. Stop at 1st non-continuation
532 * and leave it in the conn buffer
534 for (i = 0; i < HTTP_MAX_CONT_LINES; i++) {
535 if (fetch_getln(conn) == -1)
536 return (hdr_syserror);
539 * Note: we carry on the idea from the previous version
540 * that a pure whitespace line is equivalent to an empty
541 * one (so it's not continuation and will be handled when
542 * we are called next)
544 http_conn_trimright(conn);
545 if (conn->buf[0] != ' ' && conn->buf[0] != "\t"[0])
548 /* Got a continuation line. Concatenate to previous */
549 len = hbuf->buflen + conn->buflen;
550 if (hbuf->bufsize < len + 1) {
552 if ((hbuf->buf = realloc(hbuf->buf, len + 1)) == NULL)
553 return (hdr_syserror);
554 hbuf->bufsize = len + 1;
556 strcpy(hbuf->buf + hbuf->buflen, conn->buf);
557 hbuf->buflen += conn->buflen;
561 * We could check for malformed headers but we don't really care.
562 * A valid header starts with a token immediately followed by a
563 * colon; a token is any sequence of non-control, non-whitespace
564 * characters except "()<>@,;:\\\"{}".
566 for (i = 0; hdr_names[i].num != hdr_unknown; i++)
567 if ((*p = http_match(hdr_names[i].name, hbuf->buf)) != NULL)
568 return (hdr_names[i].num);
570 return (hdr_unknown);
573 /**************************
574 * [Proxy-]Authenticate header parsing
578 * Read doublequote-delimited string into output buffer obuf (allocated
579 * by caller, whose responsibility it is to ensure that it's big enough)
580 * cp points to the first char after the initial '"'
582 * Returns pointer to the first char after the terminating double quote, or
586 http_parse_headerstring(const char *cp, char *obuf)
590 case 0: /* Unterminated string */
593 case '"': /* Ending quote */
608 /* Http auth challenge schemes */
609 typedef enum {HTTPAS_UNKNOWN, HTTPAS_BASIC,HTTPAS_DIGEST} http_auth_schemes_t;
611 /* Data holder for a Basic or Digest challenge. */
613 http_auth_schemes_t scheme;
620 int nc; /* Nonce count */
621 } http_auth_challenge_t;
624 init_http_auth_challenge(http_auth_challenge_t *b)
626 b->scheme = HTTPAS_UNKNOWN;
627 b->realm = b->qop = b->nonce = b->opaque = b->algo = NULL;
628 b->stale = b->nc = 0;
632 clean_http_auth_challenge(http_auth_challenge_t *b)
644 init_http_auth_challenge(b);
647 /* Data holder for an array of challenges offered in an http response. */
648 #define MAX_CHALLENGES 10
650 http_auth_challenge_t *challenges[MAX_CHALLENGES];
651 int count; /* Number of parsed challenges in the array */
652 int valid; /* We did parse an authenticate header */
653 } http_auth_challenges_t;
656 init_http_auth_challenges(http_auth_challenges_t *cs)
659 for (i = 0; i < MAX_CHALLENGES; i++)
660 cs->challenges[i] = NULL;
661 cs->count = cs->valid = 0;
665 clean_http_auth_challenges(http_auth_challenges_t *cs)
668 /* We rely on non-zero pointers being allocated, not on the count */
669 for (i = 0; i < MAX_CHALLENGES; i++) {
670 if (cs->challenges[i] != NULL) {
671 clean_http_auth_challenge(cs->challenges[i]);
672 free(cs->challenges[i]);
675 init_http_auth_challenges(cs);
679 * Enumeration for lexical elements. Separators will be returned as their own
682 typedef enum {HTTPHL_WORD=256, HTTPHL_STRING=257, HTTPHL_END=258,
683 HTTPHL_ERROR = 259} http_header_lex_t;
686 * Determine what kind of token comes next and return possible value
687 * in buf, which is supposed to have been allocated big enough by
688 * caller. Advance input pointer and return element type.
691 http_header_lex(const char **cpp, char *buf)
694 /* Eat initial whitespace */
695 *cpp += strspn(*cpp, " \t");
700 if (**cpp == ',' || **cpp == '=')
701 return (*((*cpp)++));
705 *cpp = http_parse_headerstring(++*cpp, buf);
707 return (HTTPHL_ERROR);
708 return (HTTPHL_STRING);
711 /* Read other token, until separator or whitespace */
712 l = strcspn(*cpp, " \t,=");
713 memcpy(buf, *cpp, l);
716 return (HTTPHL_WORD);
720 * Read challenges from http xxx-authenticate header and accumulate them
721 * in the challenges list structure.
723 * Headers with multiple challenges are specified by rfc2617, but
724 * servers (ie: squid) often send them in separate headers instead,
725 * which in turn is forbidden by the http spec (multiple headers with
726 * the same name are only allowed for pure comma-separated lists, see
729 * We support both approaches anyway
732 http_parse_authenticate(const char *cp, http_auth_challenges_t *cs)
735 http_header_lex_t lex;
736 char *key = malloc(strlen(cp) + 1);
737 char *value = malloc(strlen(cp) + 1);
738 char *buf = malloc(strlen(cp) + 1);
740 if (key == NULL || value == NULL || buf == NULL) {
745 /* In any case we've seen the header and we set the valid bit */
748 /* Need word first */
749 lex = http_header_lex(&cp, key);
750 if (lex != HTTPHL_WORD)
753 /* Loop on challenges */
754 for (; cs->count < MAX_CHALLENGES; cs->count++) {
755 cs->challenges[cs->count] =
756 malloc(sizeof(http_auth_challenge_t));
757 if (cs->challenges[cs->count] == NULL) {
761 init_http_auth_challenge(cs->challenges[cs->count]);
762 if (!strcasecmp(key, "basic")) {
763 cs->challenges[cs->count]->scheme = HTTPAS_BASIC;
764 } else if (!strcasecmp(key, "digest")) {
765 cs->challenges[cs->count]->scheme = HTTPAS_DIGEST;
767 cs->challenges[cs->count]->scheme = HTTPAS_UNKNOWN;
769 * Continue parsing as basic or digest may
770 * follow, and the syntax is the same for
771 * all. We'll just ignore this one when
772 * looking at the list
776 /* Loop on attributes */
779 lex = http_header_lex(&cp, key);
780 if (lex != HTTPHL_WORD)
784 lex = http_header_lex(&cp, buf);
789 lex = http_header_lex(&cp, value);
790 if (lex != HTTPHL_WORD && lex != HTTPHL_STRING)
793 if (!strcasecmp(key, "realm"))
794 cs->challenges[cs->count]->realm =
796 else if (!strcasecmp(key, "qop"))
797 cs->challenges[cs->count]->qop =
799 else if (!strcasecmp(key, "nonce"))
800 cs->challenges[cs->count]->nonce =
802 else if (!strcasecmp(key, "opaque"))
803 cs->challenges[cs->count]->opaque =
805 else if (!strcasecmp(key, "algorithm"))
806 cs->challenges[cs->count]->algo =
808 else if (!strcasecmp(key, "stale"))
809 cs->challenges[cs->count]->stale =
810 strcasecmp(value, "no");
811 /* Else ignore unknown attributes */
813 /* Comma or Next challenge or End */
814 lex = http_header_lex(&cp, key);
816 * If we get a word here, this is the beginning of the
817 * next challenge. Break the attributes loop
819 if (lex == HTTPHL_WORD)
822 if (lex == HTTPHL_END) {
823 /* End while looking for ',' is normal exit */
828 /* Anything else is an error */
832 } /* End attributes loop */
833 } /* End challenge loop */
836 * Challenges max count exceeded. This really can't happen
837 * with normal data, something's fishy -> error
852 * Parse a last-modified header
855 http_parse_mtime(const char *p, time_t *mtime)
860 strncpy(locale, setlocale(LC_TIME, NULL), sizeof(locale));
861 setlocale(LC_TIME, "C");
862 r = strptime(p, "%a, %d %b %Y %H:%M:%S GMT", &tm);
863 /* XXX should add support for date-2 and date-3 */
864 setlocale(LC_TIME, locale);
867 DEBUG(fprintf(stderr, "last modified: [%04d-%02d-%02d "
869 tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday,
870 tm.tm_hour, tm.tm_min, tm.tm_sec));
871 *mtime = timegm(&tm);
876 * Parse a content-length header
879 http_parse_length(const char *p, off_t *length)
883 for (len = 0; *p && isdigit((unsigned char)*p); ++p)
884 len = len * 10 + (*p - '0');
887 DEBUG(fprintf(stderr, "content length: [%lld]\n",
894 * Parse a content-range header
897 http_parse_range(const char *p, off_t *offset, off_t *length, off_t *size)
899 off_t first, last, len;
901 if (strncasecmp(p, "bytes ", 6) != 0)
908 for (first = 0; *p && isdigit((unsigned char)*p); ++p)
909 first = first * 10 + *p - '0';
912 for (last = 0, ++p; *p && isdigit((unsigned char)*p); ++p)
913 last = last * 10 + *p - '0';
915 if (first > last || *p != '/')
917 for (len = 0, ++p; *p && isdigit((unsigned char)*p); ++p)
918 len = len * 10 + *p - '0';
919 if (*p || len < last - first + 1)
922 DEBUG(fprintf(stderr, "content range: [*/%lld]\n",
926 DEBUG(fprintf(stderr, "content range: [%lld-%lld/%lld]\n",
927 (long long)first, (long long)last, (long long)len));
928 *length = last - first + 1;
936 /*****************************************************************************
937 * Helper functions for authorization
944 http_base64(const char *src)
946 static const char base64[] =
947 "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
948 "abcdefghijklmnopqrstuvwxyz"
955 if ((str = malloc(((l + 2) / 3) * 4 + 1)) == NULL)
961 t = (src[0] << 16) | (src[1] << 8) | src[2];
962 dst[0] = base64[(t >> 18) & 0x3f];
963 dst[1] = base64[(t >> 12) & 0x3f];
964 dst[2] = base64[(t >> 6) & 0x3f];
965 dst[3] = base64[(t >> 0) & 0x3f];
972 t = (src[0] << 16) | (src[1] << 8);
973 dst[0] = base64[(t >> 18) & 0x3f];
974 dst[1] = base64[(t >> 12) & 0x3f];
975 dst[2] = base64[(t >> 6) & 0x3f];
982 dst[0] = base64[(t >> 18) & 0x3f];
983 dst[1] = base64[(t >> 12) & 0x3f];
984 dst[2] = dst[3] = '=';
998 * Extract authorization parameters from environment value.
999 * The value is like scheme:realm:user:pass
1006 } http_auth_params_t;
1009 init_http_auth_params(http_auth_params_t *s)
1011 s->scheme = s->realm = s->user = s->password = 0;
1015 clean_http_auth_params(http_auth_params_t *s)
1025 init_http_auth_params(s);
1029 http_authfromenv(const char *p, http_auth_params_t *parms)
1033 char *str = strdup(p);
1041 if ((ve = strchr(v, ':')) == NULL)
1045 if ((parms->scheme = strdup(v)) == NULL) {
1051 if ((ve = strchr(v, ':')) == NULL)
1055 if ((parms->realm = strdup(v)) == NULL) {
1061 if ((ve = strchr(v, ':')) == NULL)
1065 if ((parms->user = strdup(v)) == NULL) {
1072 if ((parms->password = strdup(v)) == NULL) {
1079 clean_http_auth_params(parms);
1087 * Digest response: the code to compute the digest is taken from the
1088 * sample implementation in RFC2616
1094 typedef char HASH[HASHLEN];
1095 #define HASHHEXLEN 32
1096 typedef char HASHHEX[HASHHEXLEN+1];
1098 static const char *hexchars = "0123456789abcdef";
1100 CvtHex(IN HASH Bin, OUT HASHHEX Hex)
1105 for (i = 0; i < HASHLEN; i++) {
1106 j = (Bin[i] >> 4) & 0xf;
1107 Hex[i*2] = hexchars[j];
1109 Hex[i*2+1] = hexchars[j];
1111 Hex[HASHHEXLEN] = '\0';
1114 /* calculate H(A1) as per spec */
1118 IN char * pszUserName,
1120 IN char * pszPassword,
1122 IN char * pszCNonce,
1123 OUT HASHHEX SessionKey
1130 MD5Update(&Md5Ctx, pszUserName, strlen(pszUserName));
1131 MD5Update(&Md5Ctx, ":", 1);
1132 MD5Update(&Md5Ctx, pszRealm, strlen(pszRealm));
1133 MD5Update(&Md5Ctx, ":", 1);
1134 MD5Update(&Md5Ctx, pszPassword, strlen(pszPassword));
1135 MD5Final(HA1, &Md5Ctx);
1136 if (strcasecmp(pszAlg, "md5-sess") == 0) {
1139 MD5Update(&Md5Ctx, HA1, HASHLEN);
1140 MD5Update(&Md5Ctx, ":", 1);
1141 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce));
1142 MD5Update(&Md5Ctx, ":", 1);
1143 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce));
1144 MD5Final(HA1, &Md5Ctx);
1146 CvtHex(HA1, SessionKey);
1149 /* calculate request-digest/response-digest as per HTTP Digest spec */
1152 IN HASHHEX HA1, /* H(A1) */
1153 IN char * pszNonce, /* nonce from server */
1154 IN char * pszNonceCount, /* 8 hex digits */
1155 IN char * pszCNonce, /* client nonce */
1156 IN char * pszQop, /* qop-value: "", "auth", "auth-int" */
1157 IN char * pszMethod, /* method from the request */
1158 IN char * pszDigestUri, /* requested URL */
1159 IN HASHHEX HEntity, /* H(entity body) if qop="auth-int" */
1160 OUT HASHHEX Response /* request-digest or response-digest */
1163 /* DEBUG(fprintf(stderr,
1164 "Calc: HA1[%s] Nonce[%s] qop[%s] method[%s] URI[%s]\n",
1165 HA1, pszNonce, pszQop, pszMethod, pszDigestUri));*/
1173 MD5Update(&Md5Ctx, pszMethod, strlen(pszMethod));
1174 MD5Update(&Md5Ctx, ":", 1);
1175 MD5Update(&Md5Ctx, pszDigestUri, strlen(pszDigestUri));
1176 if (strcasecmp(pszQop, "auth-int") == 0) {
1177 MD5Update(&Md5Ctx, ":", 1);
1178 MD5Update(&Md5Ctx, HEntity, HASHHEXLEN);
1180 MD5Final(HA2, &Md5Ctx);
1181 CvtHex(HA2, HA2Hex);
1183 // calculate response
1185 MD5Update(&Md5Ctx, HA1, HASHHEXLEN);
1186 MD5Update(&Md5Ctx, ":", 1);
1187 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce));
1188 MD5Update(&Md5Ctx, ":", 1);
1190 MD5Update(&Md5Ctx, pszNonceCount, strlen(pszNonceCount));
1191 MD5Update(&Md5Ctx, ":", 1);
1192 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce));
1193 MD5Update(&Md5Ctx, ":", 1);
1194 MD5Update(&Md5Ctx, pszQop, strlen(pszQop));
1195 MD5Update(&Md5Ctx, ":", 1);
1197 MD5Update(&Md5Ctx, HA2Hex, HASHHEXLEN);
1198 MD5Final(RespHash, &Md5Ctx);
1199 CvtHex(RespHash, Response);
1203 * Generate/Send a Digest authorization header
1204 * This looks like: [Proxy-]Authorization: credentials
1206 * credentials = "Digest" digest-response
1207 * digest-response = 1#( username | realm | nonce | digest-uri
1208 * | response | [ algorithm ] | [cnonce] |
1209 * [opaque] | [message-qop] |
1210 * [nonce-count] | [auth-param] )
1211 * username = "username" "=" username-value
1212 * username-value = quoted-string
1213 * digest-uri = "uri" "=" digest-uri-value
1214 * digest-uri-value = request-uri ; As specified by HTTP/1.1
1215 * message-qop = "qop" "=" qop-value
1216 * cnonce = "cnonce" "=" cnonce-value
1217 * cnonce-value = nonce-value
1218 * nonce-count = "nc" "=" nc-value
1220 * response = "response" "=" request-digest
1221 * request-digest = <"> 32LHEX <">
1224 http_digest_auth(conn_t *conn, const char *hdr, http_auth_challenge_t *c,
1225 http_auth_params_t *parms, struct url *url)
1228 char noncecount[10];
1232 if (!c->realm || !c->nonce) {
1233 DEBUG(fprintf(stderr, "realm/nonce not set in challenge\n"));
1237 c->algo = strdup("");
1239 if (asprintf(&options, "%s%s%s%s",
1240 *c->algo? ",algorithm=" : "", c->algo,
1241 c->opaque? ",opaque=" : "", c->opaque?c->opaque:"")== -1)
1245 c->qop = strdup("");
1250 sprintf(noncecount, "%08x", c->nc);
1251 /* We don't try very hard with the cnonce ... */
1252 sprintf(cnonce, "%x%lx", getpid(), (unsigned long)time(0));
1256 DigestCalcHA1(c->algo, parms->user, c->realm,
1257 parms->password, c->nonce, cnonce, HA1);
1258 DEBUG(fprintf(stderr, "HA1: [%s]\n", HA1));
1260 DigestCalcResponse(HA1, c->nonce, noncecount, cnonce, c->qop,
1261 "GET", url->doc, "", digest);
1264 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\","
1265 "nonce=\"%s\",uri=\"%s\",response=\"%s\","
1266 "qop=\"auth\", cnonce=\"%s\", nc=%s%s",
1267 hdr, parms->user, c->realm,
1268 c->nonce, url->doc, digest,
1269 cnonce, noncecount, options);
1271 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\","
1272 "nonce=\"%s\",uri=\"%s\",response=\"%s\"%s",
1273 hdr, parms->user, c->realm,
1274 c->nonce, url->doc, digest, options);
1282 * Encode username and password
1285 http_basic_auth(conn_t *conn, const char *hdr, const char *usr, const char *pwd)
1290 DEBUG(fprintf(stderr, "basic: usr: [%s]\n", usr));
1291 DEBUG(fprintf(stderr, "basic: pwd: [%s]\n", pwd));
1292 if (asprintf(&upw, "%s:%s", usr, pwd) == -1)
1294 auth = http_base64(upw);
1298 r = http_cmd(conn, "%s: Basic %s", hdr, auth);
1304 * Chose the challenge to answer and call the appropriate routine to
1305 * produce the header.
1308 http_authorize(conn_t *conn, const char *hdr, http_auth_challenges_t *cs,
1309 http_auth_params_t *parms, struct url *url)
1311 http_auth_challenge_t *basic = NULL;
1312 http_auth_challenge_t *digest = NULL;
1315 /* If user or pass are null we're not happy */
1316 if (!parms->user || !parms->password) {
1317 DEBUG(fprintf(stderr, "NULL usr or pass\n"));
1321 /* Look for a Digest and a Basic challenge */
1322 for (i = 0; i < cs->count; i++) {
1323 if (cs->challenges[i]->scheme == HTTPAS_BASIC)
1324 basic = cs->challenges[i];
1325 if (cs->challenges[i]->scheme == HTTPAS_DIGEST)
1326 digest = cs->challenges[i];
1329 /* Error if "Digest" was specified and there is no Digest challenge */
1330 if (!digest && (parms->scheme &&
1331 !strcasecmp(parms->scheme, "digest"))) {
1332 DEBUG(fprintf(stderr,
1333 "Digest auth in env, not supported by peer\n"));
1337 * If "basic" was specified in the environment, or there is no Digest
1338 * challenge, do the basic thing. Don't need a challenge for this,
1339 * so no need to check basic!=NULL
1341 if (!digest || (parms->scheme && !strcasecmp(parms->scheme,"basic")))
1342 return (http_basic_auth(conn,hdr,parms->user,parms->password));
1344 /* Else, prefer digest. We just checked that it's not NULL */
1345 return (http_digest_auth(conn, hdr, digest, parms, url));
1348 /*****************************************************************************
1349 * Helper functions for connecting to a server or proxy
1353 * Connect to the correct HTTP server or proxy.
1356 http_connect(struct url *URL, struct url *purl, const char *flags)
1368 verbose = CHECK_FLAG('v');
1369 if (CHECK_FLAG('4'))
1372 else if (CHECK_FLAG('6'))
1376 if (purl && strcasecmp(URL->scheme, SCHEME_HTTPS) != 0) {
1378 } else if (strcasecmp(URL->scheme, SCHEME_FTP) == 0) {
1379 /* can't talk http to an ftp server */
1380 /* XXX should set an error code */
1384 if ((conn = fetch_connect(URL->host, URL->port, af, verbose)) == NULL)
1385 /* fetch_connect() has already set an error code */
1387 if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 &&
1388 fetch_ssl(conn, verbose) == -1) {
1397 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, sizeof(val));
1403 http_get_proxy(struct url * url, const char *flags)
1408 if (flags != NULL && strchr(flags, 'd') != NULL)
1410 if (fetch_no_proxy_match(url->host))
1412 if (((p = getenv("HTTP_PROXY")) || (p = getenv("http_proxy"))) &&
1413 *p && (purl = fetchParseURL(p))) {
1415 strcpy(purl->scheme, SCHEME_HTTP);
1417 purl->port = fetch_default_proxy_port(purl->scheme);
1418 if (strcasecmp(purl->scheme, SCHEME_HTTP) == 0)
1426 http_print_html(FILE *out, FILE *in)
1433 while ((line = fgetln(in, &len)) != NULL) {
1434 while (len && isspace((unsigned char)line[len - 1]))
1436 for (p = q = line; q < line + len; ++q) {
1437 if (comment && *q == '-') {
1438 if (q + 2 < line + len &&
1439 strcmp(q, "-->") == 0) {
1443 } else if (tag && !comment && *q == '>') {
1446 } else if (!tag && *q == '<') {
1448 fwrite(p, q - p, 1, out);
1450 if (q + 3 < line + len &&
1451 strcmp(q, "<!--") == 0) {
1458 fwrite(p, q - p, 1, out);
1464 /*****************************************************************************
1469 * Send a request and process the reply
1471 * XXX This function is way too long, the do..while loop should be split
1472 * XXX off into a separate function.
1475 http_request(struct url *URL, const char *op, struct url_stat *us,
1476 struct url *purl, const char *flags)
1479 char hbuf[MAXHOSTNAMELEN + 7], *host;
1481 struct url *url, *new;
1482 int chunked, direct, ims, noredirect, verbose;
1484 off_t offset, clength, length, size;
1489 struct tm *timestruct;
1490 http_headerbuf_t headerbuf;
1491 http_auth_challenges_t server_challenges;
1492 http_auth_challenges_t proxy_challenges;
1494 /* The following calls don't allocate anything */
1495 init_http_headerbuf(&headerbuf);
1496 init_http_auth_challenges(&server_challenges);
1497 init_http_auth_challenges(&proxy_challenges);
1499 direct = CHECK_FLAG('d');
1500 noredirect = CHECK_FLAG('A');
1501 verbose = CHECK_FLAG('v');
1502 ims = CHECK_FLAG('i');
1504 if (direct && purl) {
1509 /* try the provided URL first */
1512 /* if the A flag is set, we only get one try */
1513 n = noredirect ? 1 : MAX_REDIRECT;
1516 e = HTTP_PROTOCOL_ERROR;
1528 url->port = fetch_default_port(url->scheme);
1530 /* were we redirected to an FTP URL? */
1531 if (purl == NULL && strcmp(url->scheme, SCHEME_FTP) == 0) {
1532 if (strcmp(op, "GET") == 0)
1533 return (ftp_request(url, "RETR", us, purl, flags));
1534 else if (strcmp(op, "HEAD") == 0)
1535 return (ftp_request(url, "STAT", us, purl, flags));
1538 /* connect to server or proxy */
1539 if ((conn = http_connect(url, purl, flags)) == NULL)
1544 if (strchr(url->host, ':')) {
1545 snprintf(hbuf, sizeof(hbuf), "[%s]", url->host);
1549 if (url->port != fetch_default_port(url->scheme)) {
1554 snprintf(hbuf + strlen(hbuf),
1555 sizeof(hbuf) - strlen(hbuf), ":%d", url->port);
1560 fetch_info("requesting %s://%s%s",
1561 url->scheme, host, url->doc);
1563 http_cmd(conn, "%s %s://%s%s HTTP/1.1",
1564 op, url->scheme, host, url->doc);
1566 http_cmd(conn, "%s %s HTTP/1.1",
1570 if (ims && url->ims_time) {
1571 timestruct = gmtime((time_t *)&url->ims_time);
1572 (void)strftime(timebuf, 80, "%a, %d %b %Y %T GMT",
1575 fetch_info("If-Modified-Since: %s", timebuf);
1576 http_cmd(conn, "If-Modified-Since: %s", timebuf);
1579 http_cmd(conn, "Host: %s", host);
1582 * Proxy authorization: we only send auth after we received
1583 * a 407 error. We do not first try basic anyway (changed
1584 * when support was added for digest-auth)
1586 if (purl && proxy_challenges.valid) {
1587 http_auth_params_t aparams;
1588 init_http_auth_params(&aparams);
1589 if (*purl->user || *purl->pwd) {
1590 aparams.user = purl->user ?
1591 strdup(purl->user) : strdup("");
1592 aparams.password = purl->pwd?
1593 strdup(purl->pwd) : strdup("");
1594 } else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL &&
1596 if (http_authfromenv(p, &aparams) < 0) {
1597 http_seterr(HTTP_NEED_PROXY_AUTH);
1601 http_authorize(conn, "Proxy-Authorization",
1602 &proxy_challenges, &aparams, url);
1603 clean_http_auth_params(&aparams);
1607 * Server authorization: we never send "a priori"
1608 * Basic auth, which used to be done if user/pass were
1609 * set in the url. This would be weird because we'd send the
1610 * password in the clear even if Digest is finally to be
1611 * used (it would have made more sense for the
1612 * pre-digest version to do this when Basic was specified
1613 * in the environment)
1615 if (server_challenges.valid) {
1616 http_auth_params_t aparams;
1617 init_http_auth_params(&aparams);
1618 if (*url->user || *url->pwd) {
1619 aparams.user = url->user ?
1620 strdup(url->user) : strdup("");
1621 aparams.password = url->pwd ?
1622 strdup(url->pwd) : strdup("");
1623 } else if ((p = getenv("HTTP_AUTH")) != NULL &&
1625 if (http_authfromenv(p, &aparams) < 0) {
1626 http_seterr(HTTP_NEED_AUTH);
1629 } else if (fetchAuthMethod &&
1630 fetchAuthMethod(url) == 0) {
1631 aparams.user = url->user ?
1632 strdup(url->user) : strdup("");
1633 aparams.password = url->pwd ?
1634 strdup(url->pwd) : strdup("");
1636 http_seterr(HTTP_NEED_AUTH);
1639 http_authorize(conn, "Authorization",
1640 &server_challenges, &aparams, url);
1641 clean_http_auth_params(&aparams);
1645 if ((p = getenv("HTTP_REFERER")) != NULL && *p != '\0') {
1646 if (strcasecmp(p, "auto") == 0)
1647 http_cmd(conn, "Referer: %s://%s%s",
1648 url->scheme, host, url->doc);
1650 http_cmd(conn, "Referer: %s", p);
1652 if ((p = getenv("HTTP_USER_AGENT")) != NULL && *p != '\0')
1653 http_cmd(conn, "User-Agent: %s", p);
1655 http_cmd(conn, "User-Agent: %s " _LIBFETCH_VER, getprogname());
1656 if (url->offset > 0)
1657 http_cmd(conn, "Range: bytes=%lld-", (long long)url->offset);
1658 http_cmd(conn, "Connection: close");
1662 * Force the queued request to be dispatched. Normally, one
1663 * would do this with shutdown(2) but squid proxies can be
1664 * configured to disallow such half-closed connections. To
1665 * be compatible with such configurations, fiddle with socket
1666 * options to force the pending data to be written.
1669 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val,
1672 setsockopt(conn->sd, IPPROTO_TCP, TCP_NODELAY, &val,
1676 switch (http_get_reply(conn)) {
1679 case HTTP_NOT_MODIFIED:
1682 case HTTP_MOVED_PERM:
1683 case HTTP_MOVED_TEMP:
1684 case HTTP_SEE_OTHER:
1686 * Not so fine, but we still have to read the
1687 * headers to get the new location.
1690 case HTTP_NEED_AUTH:
1691 if (server_challenges.valid) {
1693 * We already sent out authorization code,
1694 * so there's nothing more we can do.
1696 http_seterr(conn->err);
1699 /* try again, but send the password this time */
1701 fetch_info("server requires authorization");
1703 case HTTP_NEED_PROXY_AUTH:
1704 if (proxy_challenges.valid) {
1706 * We already sent our proxy
1707 * authorization code, so there's
1708 * nothing more we can do. */
1709 http_seterr(conn->err);
1712 /* try again, but send the password this time */
1714 fetch_info("proxy requires authorization");
1716 case HTTP_BAD_RANGE:
1718 * This can happen if we ask for 0 bytes because
1719 * we already have the whole file. Consider this
1720 * a success for now, and check sizes later.
1723 case HTTP_PROTOCOL_ERROR:
1729 http_seterr(conn->err);
1732 /* fall through so we can get the full error message */
1735 /* get headers. http_next_header expects one line readahead */
1736 if (fetch_getln(conn) == -1) {
1741 switch ((h = http_next_header(conn, &headerbuf, &p))) {
1746 http_seterr(HTTP_PROTOCOL_ERROR);
1748 case hdr_content_length:
1749 http_parse_length(p, &clength);
1751 case hdr_content_range:
1752 http_parse_range(p, &offset, &length, &size);
1754 case hdr_last_modified:
1755 http_parse_mtime(p, &mtime);
1758 if (!HTTP_REDIRECT(conn->err))
1763 fetch_info("%d redirect to %s", conn->err, p);
1766 new = fetchMakeURL(url->scheme, url->host, url->port, p,
1767 url->user, url->pwd);
1769 new = fetchParseURL(p);
1771 /* XXX should set an error code */
1772 DEBUG(fprintf(stderr, "failed to parse new URL\n"));
1775 if (!*new->user && !*new->pwd) {
1776 strcpy(new->user, url->user);
1777 strcpy(new->pwd, url->pwd);
1779 new->offset = url->offset;
1780 new->length = url->length;
1782 case hdr_transfer_encoding:
1784 chunked = (strcasecmp(p, "chunked") == 0);
1786 case hdr_www_authenticate:
1787 if (conn->err != HTTP_NEED_AUTH)
1789 if (http_parse_authenticate(p, &server_challenges) == 0)
1792 case hdr_proxy_authenticate:
1793 if (conn->err != HTTP_NEED_PROXY_AUTH)
1795 if (http_parse_authenticate(p, &proxy_challenges) == 0)
1804 } while (h > hdr_end);
1806 /* we need to provide authentication */
1807 if (conn->err == HTTP_NEED_AUTH ||
1808 conn->err == HTTP_NEED_PROXY_AUTH) {
1810 if ((conn->err == HTTP_NEED_AUTH &&
1811 !server_challenges.valid) ||
1812 (conn->err == HTTP_NEED_PROXY_AUTH &&
1813 !proxy_challenges.valid)) {
1814 /* 401/7 but no www/proxy-authenticate ?? */
1815 DEBUG(fprintf(stderr, "401/7 and no auth header\n"));
1823 /* requested range not satisfiable */
1824 if (conn->err == HTTP_BAD_RANGE) {
1825 if (url->offset == size && url->length == 0) {
1826 /* asked for 0 bytes; fake it */
1827 offset = url->offset;
1829 conn->err = HTTP_OK;
1832 http_seterr(conn->err);
1837 /* we have a hit or an error */
1838 if (conn->err == HTTP_OK
1839 || conn->err == HTTP_NOT_MODIFIED
1840 || conn->err == HTTP_PARTIAL
1841 || HTTP_ERROR(conn->err))
1844 /* all other cases: we got a redirect */
1846 clean_http_auth_challenges(&server_challenges);
1850 DEBUG(fprintf(stderr, "redirect with no new location\n"));
1858 /* we failed, or ran out of retries */
1864 DEBUG(fprintf(stderr, "offset %lld, length %lld,"
1865 " size %lld, clength %lld\n",
1866 (long long)offset, (long long)length,
1867 (long long)size, (long long)clength));
1869 if (conn->err == HTTP_NOT_MODIFIED) {
1870 http_seterr(HTTP_NOT_MODIFIED);
1874 /* check for inconsistencies */
1875 if (clength != -1 && length != -1 && clength != length) {
1876 http_seterr(HTTP_PROTOCOL_ERROR);
1882 length = offset + clength;
1883 if (length != -1 && size != -1 && length != size) {
1884 http_seterr(HTTP_PROTOCOL_ERROR);
1893 us->atime = us->mtime = mtime;
1897 if (URL->offset > 0 && offset > URL->offset) {
1898 http_seterr(HTTP_PROTOCOL_ERROR);
1902 /* report back real offset and size */
1903 URL->offset = offset;
1904 URL->length = clength;
1906 /* wrap it up in a FILE */
1907 if ((f = http_funopen(conn, chunked)) == NULL) {
1917 if (HTTP_ERROR(conn->err)) {
1918 http_print_html(stderr, f);
1922 clean_http_headerbuf(&headerbuf);
1923 clean_http_auth_challenges(&server_challenges);
1924 clean_http_auth_challenges(&proxy_challenges);
1934 clean_http_headerbuf(&headerbuf);
1935 clean_http_auth_challenges(&server_challenges);
1936 clean_http_auth_challenges(&proxy_challenges);
1941 /*****************************************************************************
1946 * Retrieve and stat a file by HTTP
1949 fetchXGetHTTP(struct url *URL, struct url_stat *us, const char *flags)
1951 return (http_request(URL, "GET", us, http_get_proxy(URL, flags), flags));
1955 * Retrieve a file by HTTP
1958 fetchGetHTTP(struct url *URL, const char *flags)
1960 return (fetchXGetHTTP(URL, NULL, flags));
1964 * Store a file by HTTP
1967 fetchPutHTTP(struct url *URL __unused, const char *flags __unused)
1969 warnx("fetchPutHTTP(): not implemented");
1974 * Get an HTTP document's metadata
1977 fetchStatHTTP(struct url *URL, struct url_stat *us, const char *flags)
1981 f = http_request(URL, "HEAD", us, http_get_proxy(URL, flags), flags);
1992 fetchListHTTP(struct url *url __unused, const char *flags __unused)
1994 warnx("fetchListHTTP(): not implemented");