2 * Copyright (c) 2000-2011 Dag-Erling 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.
30 * The following copyright applies to the base64 code:
33 * Copyright 1997 Massachusetts Institute of Technology
35 * Permission to use, copy, modify, and distribute this software and
36 * its documentation for any purpose and without fee is hereby
37 * granted, provided that both the above copyright notice and this
38 * permission notice appear in all copies, that both the above
39 * copyright notice and this permission notice appear in all
40 * supporting documentation, and that the name of M.I.T. not be used
41 * in advertising or publicity pertaining to distribution of the
42 * software without specific, written prior permission. M.I.T. makes
43 * no representations about the suitability of this software for any
44 * purpose. It is provided "as is" without express or implied
47 * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''. M.I.T. DISCLAIMS
48 * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE,
49 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
50 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT
51 * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
52 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
53 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
54 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
55 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
56 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
57 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
61 #include <sys/param.h>
62 #include <sys/socket.h>
78 #include <openssl/md5.h>
79 #define MD5Init(c) MD5_Init(c)
80 #define MD5Update(c, data, len) MD5_Update(c, data, len)
81 #define MD5Final(md, c) MD5_Final(md, c)
86 #include <netinet/in.h>
87 #include <netinet/tcp.h>
93 /* Maximum number of redirects to follow */
94 #define MAX_REDIRECT 20
96 /* Symbolic names for reply codes we care about */
98 #define HTTP_PARTIAL 206
99 #define HTTP_MOVED_PERM 301
100 #define HTTP_MOVED_TEMP 302
101 #define HTTP_SEE_OTHER 303
102 #define HTTP_NOT_MODIFIED 304
103 #define HTTP_USE_PROXY 305
104 #define HTTP_TEMP_REDIRECT 307
105 #define HTTP_PERM_REDIRECT 308
106 #define HTTP_NEED_AUTH 401
107 #define HTTP_NEED_PROXY_AUTH 407
108 #define HTTP_BAD_RANGE 416
109 #define HTTP_PROTOCOL_ERROR 999
111 #define HTTP_REDIRECT(xyz) ((xyz) == HTTP_MOVED_PERM \
112 || (xyz) == HTTP_MOVED_TEMP \
113 || (xyz) == HTTP_TEMP_REDIRECT \
114 || (xyz) == HTTP_USE_PROXY \
115 || (xyz) == HTTP_SEE_OTHER)
117 #define HTTP_ERROR(xyz) ((xyz) > 400 && (xyz) < 599)
120 /*****************************************************************************
121 * I/O functions for decoding chunked streams
126 conn_t *conn; /* connection */
127 int chunked; /* chunked mode */
128 char *buf; /* chunk buffer */
129 size_t bufsize; /* size of chunk buffer */
130 ssize_t buflen; /* amount of data currently in buffer */
131 int bufpos; /* current read offset in buffer */
132 int eof; /* end-of-file flag */
133 int error; /* error flag */
134 size_t chunksize; /* remaining size of current chunk */
140 static int http_cmd(conn_t *, const char *, ...) __printflike(2, 3);
143 * Get next chunk header
146 http_new_chunk(struct httpio *io)
150 if (fetch_getln(io->conn) == -1)
153 if (io->conn->buflen < 2 || !isxdigit((unsigned char)*io->conn->buf))
156 for (p = io->conn->buf; *p && !isspace((unsigned char)*p); ++p) {
159 if (!isxdigit((unsigned char)*p))
161 if (isdigit((unsigned char)*p)) {
162 io->chunksize = io->chunksize * 16 +
165 io->chunksize = io->chunksize * 16 +
166 10 + tolower((unsigned char)*p) - 'a';
172 io->total += io->chunksize;
173 if (io->chunksize == 0)
174 fprintf(stderr, "%s(): end of last chunk\n", __func__);
176 fprintf(stderr, "%s(): new chunk: %lu (%lu)\n",
177 __func__, (unsigned long)io->chunksize,
178 (unsigned long)io->total);
182 return (io->chunksize);
186 * Grow the input buffer to at least len bytes
189 http_growbuf(struct httpio *io, size_t len)
193 if (io->bufsize >= len)
196 if ((tmp = realloc(io->buf, len)) == NULL)
204 * Fill the input buffer, do chunk decoding on the fly
207 http_fillbuf(struct httpio *io, size_t len)
216 if (io->chunked == 0) {
217 if (http_growbuf(io, len) == -1)
219 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) {
228 if (io->chunksize == 0) {
229 switch (http_new_chunk(io)) {
239 if (len > io->chunksize)
241 if (http_growbuf(io, len) == -1)
243 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) {
248 io->chunksize -= io->buflen;
250 if (io->chunksize == 0) {
253 if (fetch_read(io->conn, endl, 2) != 2 ||
254 endl[0] != '\r' || endl[1] != '\n')
267 http_readfn(void *v, char *buf, int len)
269 struct httpio *io = (struct httpio *)v;
277 for (pos = 0; len > 0; pos += l, len -= l) {
279 if (!io->buf || io->bufpos == io->buflen)
280 if (http_fillbuf(io, len) < 1)
282 l = io->buflen - io->bufpos;
285 memcpy(buf + pos, io->buf + io->bufpos, l);
289 if (!pos && io->error) {
290 if (io->error == EINTR)
301 http_writefn(void *v, const char *buf, int len)
303 struct httpio *io = (struct httpio *)v;
305 return (fetch_write(io->conn, buf, len));
312 http_closefn(void *v)
314 struct httpio *io = (struct httpio *)v;
317 r = fetch_close(io->conn);
325 * Wrap a file descriptor up
328 http_funopen(conn_t *conn, int chunked)
333 if ((io = calloc(1, sizeof(*io))) == NULL) {
338 io->chunked = chunked;
339 f = funopen(io, http_readfn, http_writefn, NULL, http_closefn);
349 /*****************************************************************************
350 * Helper functions for talking to the server and parsing its replies
363 hdr_transfer_encoding,
364 hdr_www_authenticate,
365 hdr_proxy_authenticate,
368 /* Names of interesting headers */
373 { hdr_content_length, "Content-Length" },
374 { hdr_content_range, "Content-Range" },
375 { hdr_last_modified, "Last-Modified" },
376 { hdr_location, "Location" },
377 { hdr_transfer_encoding, "Transfer-Encoding" },
378 { hdr_www_authenticate, "WWW-Authenticate" },
379 { hdr_proxy_authenticate, "Proxy-Authenticate" },
380 { hdr_unknown, NULL },
384 * Send a formatted line; optionally echo to terminal
387 http_cmd(conn_t *conn, const char *fmt, ...)
395 len = vasprintf(&msg, fmt, ap);
404 r = fetch_putln(conn, msg, len);
416 * Get and parse status line
419 http_get_reply(conn_t *conn)
423 if (fetch_getln(conn) == -1)
426 * A valid status line looks like "HTTP/m.n xyz reason" where m
427 * and n are the major and minor protocol version numbers and xyz
429 * Unfortunately, there are servers out there (NCSA 1.5.1, to name
430 * just one) that do not send a version number, so we can't rely
431 * on finding one, but if we do, insist on it being 1.0 or 1.1.
432 * We don't care about the reason phrase.
434 if (strncmp(conn->buf, "HTTP", 4) != 0)
435 return (HTTP_PROTOCOL_ERROR);
438 if (p[1] != '1' || p[2] != '.' || (p[3] != '0' && p[3] != '1'))
439 return (HTTP_PROTOCOL_ERROR);
443 !isdigit((unsigned char)p[1]) ||
444 !isdigit((unsigned char)p[2]) ||
445 !isdigit((unsigned char)p[3]))
446 return (HTTP_PROTOCOL_ERROR);
448 conn->err = (p[1] - '0') * 100 + (p[2] - '0') * 10 + (p[3] - '0');
453 * Check a header; if the type matches the given string, return a pointer
454 * to the beginning of the value.
457 http_match(const char *str, const char *hdr)
459 while (*str && *hdr &&
460 tolower((unsigned char)*str++) == tolower((unsigned char)*hdr++))
462 if (*str || *hdr != ':')
464 while (*hdr && isspace((unsigned char)*++hdr))
471 * Get the next header and return the appropriate symbolic code. We
472 * need to read one line ahead for checking for a continuation line
473 * belonging to the current header (continuation lines start with
476 * We get called with a fresh line already in the conn buffer, either
477 * from the previous http_next_header() invocation, or, the first
478 * time, from a fetch_getln() performed by our caller.
480 * This stops when we encounter an empty line (we dont read beyond the header
483 * Note that the "headerbuf" is just a place to return the result. Its
484 * contents are not used for the next call. This means that no cleanup
485 * is needed when ie doing another connection, just call the cleanup when
486 * fully done to deallocate memory.
489 /* Limit the max number of continuation lines to some reasonable value */
490 #define HTTP_MAX_CONT_LINES 10
492 /* Place into which to build a header from one or several lines */
494 char *buf; /* buffer */
495 size_t bufsize; /* buffer size */
496 size_t buflen; /* length of buffer contents */
500 init_http_headerbuf(http_headerbuf_t *buf)
508 clean_http_headerbuf(http_headerbuf_t *buf)
512 init_http_headerbuf(buf);
515 /* Remove whitespace at the end of the buffer */
517 http_conn_trimright(conn_t *conn)
519 while (conn->buflen &&
520 isspace((unsigned char)conn->buf[conn->buflen - 1]))
522 conn->buf[conn->buflen] = '\0';
526 http_next_header(conn_t *conn, http_headerbuf_t *hbuf, const char **p)
531 * Have to do the stripping here because of the first line. So
532 * it's done twice for the subsequent lines. No big deal
534 http_conn_trimright(conn);
535 if (conn->buflen == 0)
538 /* Copy the line to the headerbuf */
539 if (hbuf->bufsize < conn->buflen + 1) {
540 if ((hbuf->buf = realloc(hbuf->buf, conn->buflen + 1)) == NULL)
541 return (hdr_syserror);
542 hbuf->bufsize = conn->buflen + 1;
544 strcpy(hbuf->buf, conn->buf);
545 hbuf->buflen = conn->buflen;
548 * Fetch possible continuation lines. Stop at 1st non-continuation
549 * and leave it in the conn buffer
551 for (i = 0; i < HTTP_MAX_CONT_LINES; i++) {
552 if (fetch_getln(conn) == -1)
553 return (hdr_syserror);
556 * Note: we carry on the idea from the previous version
557 * that a pure whitespace line is equivalent to an empty
558 * one (so it's not continuation and will be handled when
559 * we are called next)
561 http_conn_trimright(conn);
562 if (conn->buf[0] != ' ' && conn->buf[0] != "\t"[0])
565 /* Got a continuation line. Concatenate to previous */
566 len = hbuf->buflen + conn->buflen;
567 if (hbuf->bufsize < len + 1) {
569 if ((hbuf->buf = realloc(hbuf->buf, len + 1)) == NULL)
570 return (hdr_syserror);
571 hbuf->bufsize = len + 1;
573 strcpy(hbuf->buf + hbuf->buflen, conn->buf);
574 hbuf->buflen += conn->buflen;
578 * We could check for malformed headers but we don't really care.
579 * A valid header starts with a token immediately followed by a
580 * colon; a token is any sequence of non-control, non-whitespace
581 * characters except "()<>@,;:\\\"{}".
583 for (i = 0; hdr_names[i].num != hdr_unknown; i++)
584 if ((*p = http_match(hdr_names[i].name, hbuf->buf)) != NULL)
585 return (hdr_names[i].num);
587 return (hdr_unknown);
590 /**************************
591 * [Proxy-]Authenticate header parsing
595 * Read doublequote-delimited string into output buffer obuf (allocated
596 * by caller, whose responsibility it is to ensure that it's big enough)
597 * cp points to the first char after the initial '"'
599 * Returns pointer to the first char after the terminating double quote, or
603 http_parse_headerstring(const char *cp, char *obuf)
607 case 0: /* Unterminated string */
610 case '"': /* Ending quote */
625 /* Http auth challenge schemes */
626 typedef enum {HTTPAS_UNKNOWN, HTTPAS_BASIC,HTTPAS_DIGEST} http_auth_schemes_t;
628 /* Data holder for a Basic or Digest challenge. */
630 http_auth_schemes_t scheme;
637 int nc; /* Nonce count */
638 } http_auth_challenge_t;
641 init_http_auth_challenge(http_auth_challenge_t *b)
643 b->scheme = HTTPAS_UNKNOWN;
644 b->realm = b->qop = b->nonce = b->opaque = b->algo = NULL;
645 b->stale = b->nc = 0;
649 clean_http_auth_challenge(http_auth_challenge_t *b)
661 init_http_auth_challenge(b);
664 /* Data holder for an array of challenges offered in an http response. */
665 #define MAX_CHALLENGES 10
667 http_auth_challenge_t *challenges[MAX_CHALLENGES];
668 int count; /* Number of parsed challenges in the array */
669 int valid; /* We did parse an authenticate header */
670 } http_auth_challenges_t;
673 init_http_auth_challenges(http_auth_challenges_t *cs)
676 for (i = 0; i < MAX_CHALLENGES; i++)
677 cs->challenges[i] = NULL;
678 cs->count = cs->valid = 0;
682 clean_http_auth_challenges(http_auth_challenges_t *cs)
685 /* We rely on non-zero pointers being allocated, not on the count */
686 for (i = 0; i < MAX_CHALLENGES; i++) {
687 if (cs->challenges[i] != NULL) {
688 clean_http_auth_challenge(cs->challenges[i]);
689 free(cs->challenges[i]);
692 init_http_auth_challenges(cs);
696 * Enumeration for lexical elements. Separators will be returned as their own
699 typedef enum {HTTPHL_WORD=256, HTTPHL_STRING=257, HTTPHL_END=258,
700 HTTPHL_ERROR = 259} http_header_lex_t;
703 * Determine what kind of token comes next and return possible value
704 * in buf, which is supposed to have been allocated big enough by
705 * caller. Advance input pointer and return element type.
708 http_header_lex(const char **cpp, char *buf)
711 /* Eat initial whitespace */
712 *cpp += strspn(*cpp, " \t");
717 if (**cpp == ',' || **cpp == '=')
718 return (*((*cpp)++));
722 *cpp = http_parse_headerstring(++*cpp, buf);
724 return (HTTPHL_ERROR);
725 return (HTTPHL_STRING);
728 /* Read other token, until separator or whitespace */
729 l = strcspn(*cpp, " \t,=");
730 memcpy(buf, *cpp, l);
733 return (HTTPHL_WORD);
737 * Read challenges from http xxx-authenticate header and accumulate them
738 * in the challenges list structure.
740 * Headers with multiple challenges are specified by rfc2617, but
741 * servers (ie: squid) often send them in separate headers instead,
742 * which in turn is forbidden by the http spec (multiple headers with
743 * the same name are only allowed for pure comma-separated lists, see
746 * We support both approaches anyway
749 http_parse_authenticate(const char *cp, http_auth_challenges_t *cs)
752 http_header_lex_t lex;
753 char *key = malloc(strlen(cp) + 1);
754 char *value = malloc(strlen(cp) + 1);
755 char *buf = malloc(strlen(cp) + 1);
757 if (key == NULL || value == NULL || buf == NULL) {
762 /* In any case we've seen the header and we set the valid bit */
765 /* Need word first */
766 lex = http_header_lex(&cp, key);
767 if (lex != HTTPHL_WORD)
770 /* Loop on challenges */
771 for (; cs->count < MAX_CHALLENGES; cs->count++) {
772 cs->challenges[cs->count] =
773 malloc(sizeof(http_auth_challenge_t));
774 if (cs->challenges[cs->count] == NULL) {
778 init_http_auth_challenge(cs->challenges[cs->count]);
779 if (!strcasecmp(key, "basic")) {
780 cs->challenges[cs->count]->scheme = HTTPAS_BASIC;
781 } else if (!strcasecmp(key, "digest")) {
782 cs->challenges[cs->count]->scheme = HTTPAS_DIGEST;
784 cs->challenges[cs->count]->scheme = HTTPAS_UNKNOWN;
786 * Continue parsing as basic or digest may
787 * follow, and the syntax is the same for
788 * all. We'll just ignore this one when
789 * looking at the list
793 /* Loop on attributes */
796 lex = http_header_lex(&cp, key);
797 if (lex != HTTPHL_WORD)
801 lex = http_header_lex(&cp, buf);
806 lex = http_header_lex(&cp, value);
807 if (lex != HTTPHL_WORD && lex != HTTPHL_STRING)
810 if (!strcasecmp(key, "realm"))
811 cs->challenges[cs->count]->realm =
813 else if (!strcasecmp(key, "qop"))
814 cs->challenges[cs->count]->qop =
816 else if (!strcasecmp(key, "nonce"))
817 cs->challenges[cs->count]->nonce =
819 else if (!strcasecmp(key, "opaque"))
820 cs->challenges[cs->count]->opaque =
822 else if (!strcasecmp(key, "algorithm"))
823 cs->challenges[cs->count]->algo =
825 else if (!strcasecmp(key, "stale"))
826 cs->challenges[cs->count]->stale =
827 strcasecmp(value, "no");
828 /* Else ignore unknown attributes */
830 /* Comma or Next challenge or End */
831 lex = http_header_lex(&cp, key);
833 * If we get a word here, this is the beginning of the
834 * next challenge. Break the attributes loop
836 if (lex == HTTPHL_WORD)
839 if (lex == HTTPHL_END) {
840 /* End while looking for ',' is normal exit */
845 /* Anything else is an error */
849 } /* End attributes loop */
850 } /* End challenge loop */
853 * Challenges max count exceeded. This really can't happen
854 * with normal data, something's fishy -> error
869 * Parse a last-modified header
872 http_parse_mtime(const char *p, time_t *mtime)
877 strncpy(locale, setlocale(LC_TIME, NULL), sizeof(locale));
878 setlocale(LC_TIME, "C");
879 r = strptime(p, "%a, %d %b %Y %H:%M:%S GMT", &tm);
880 /* XXX should add support for date-2 and date-3 */
881 setlocale(LC_TIME, locale);
884 DEBUG(fprintf(stderr, "last modified: [%04d-%02d-%02d "
886 tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday,
887 tm.tm_hour, tm.tm_min, tm.tm_sec));
888 *mtime = timegm(&tm);
893 * Parse a content-length header
896 http_parse_length(const char *p, off_t *length)
900 for (len = 0; *p && isdigit((unsigned char)*p); ++p)
901 len = len * 10 + (*p - '0');
904 DEBUG(fprintf(stderr, "content length: [%lld]\n",
911 * Parse a content-range header
914 http_parse_range(const char *p, off_t *offset, off_t *length, off_t *size)
916 off_t first, last, len;
918 if (strncasecmp(p, "bytes ", 6) != 0)
925 for (first = 0; *p && isdigit((unsigned char)*p); ++p)
926 first = first * 10 + *p - '0';
929 for (last = 0, ++p; *p && isdigit((unsigned char)*p); ++p)
930 last = last * 10 + *p - '0';
932 if (first > last || *p != '/')
934 for (len = 0, ++p; *p && isdigit((unsigned char)*p); ++p)
935 len = len * 10 + *p - '0';
936 if (*p || len < last - first + 1)
939 DEBUG(fprintf(stderr, "content range: [*/%lld]\n",
943 DEBUG(fprintf(stderr, "content range: [%lld-%lld/%lld]\n",
944 (long long)first, (long long)last, (long long)len));
945 *length = last - first + 1;
953 /*****************************************************************************
954 * Helper functions for authorization
961 http_base64(const char *src)
963 static const char base64[] =
964 "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
965 "abcdefghijklmnopqrstuvwxyz"
972 if ((str = malloc(((l + 2) / 3) * 4 + 1)) == NULL)
978 t = (src[0] << 16) | (src[1] << 8) | src[2];
979 dst[0] = base64[(t >> 18) & 0x3f];
980 dst[1] = base64[(t >> 12) & 0x3f];
981 dst[2] = base64[(t >> 6) & 0x3f];
982 dst[3] = base64[(t >> 0) & 0x3f];
989 t = (src[0] << 16) | (src[1] << 8);
990 dst[0] = base64[(t >> 18) & 0x3f];
991 dst[1] = base64[(t >> 12) & 0x3f];
992 dst[2] = base64[(t >> 6) & 0x3f];
999 dst[0] = base64[(t >> 18) & 0x3f];
1000 dst[1] = base64[(t >> 12) & 0x3f];
1001 dst[2] = dst[3] = '=';
1015 * Extract authorization parameters from environment value.
1016 * The value is like scheme:realm:user:pass
1023 } http_auth_params_t;
1026 init_http_auth_params(http_auth_params_t *s)
1028 s->scheme = s->realm = s->user = s->password = NULL;
1032 clean_http_auth_params(http_auth_params_t *s)
1042 init_http_auth_params(s);
1046 http_authfromenv(const char *p, http_auth_params_t *parms)
1050 char *str = strdup(p);
1058 if ((ve = strchr(v, ':')) == NULL)
1062 if ((parms->scheme = strdup(v)) == NULL) {
1068 if ((ve = strchr(v, ':')) == NULL)
1072 if ((parms->realm = strdup(v)) == NULL) {
1078 if ((ve = strchr(v, ':')) == NULL)
1082 if ((parms->user = strdup(v)) == NULL) {
1089 if ((parms->password = strdup(v)) == NULL) {
1096 clean_http_auth_params(parms);
1104 * Digest response: the code to compute the digest is taken from the
1105 * sample implementation in RFC2616
1111 typedef char HASH[HASHLEN];
1112 #define HASHHEXLEN 32
1113 typedef char HASHHEX[HASHHEXLEN+1];
1115 static const char *hexchars = "0123456789abcdef";
1117 CvtHex(IN HASH Bin, OUT HASHHEX Hex)
1122 for (i = 0; i < HASHLEN; i++) {
1123 j = (Bin[i] >> 4) & 0xf;
1124 Hex[i*2] = hexchars[j];
1126 Hex[i*2+1] = hexchars[j];
1128 Hex[HASHHEXLEN] = '\0';
1131 /* calculate H(A1) as per spec */
1135 IN char * pszUserName,
1137 IN char * pszPassword,
1139 IN char * pszCNonce,
1140 OUT HASHHEX SessionKey
1147 MD5Update(&Md5Ctx, pszUserName, strlen(pszUserName));
1148 MD5Update(&Md5Ctx, ":", 1);
1149 MD5Update(&Md5Ctx, pszRealm, strlen(pszRealm));
1150 MD5Update(&Md5Ctx, ":", 1);
1151 MD5Update(&Md5Ctx, pszPassword, strlen(pszPassword));
1152 MD5Final(HA1, &Md5Ctx);
1153 if (strcasecmp(pszAlg, "md5-sess") == 0) {
1156 MD5Update(&Md5Ctx, HA1, HASHLEN);
1157 MD5Update(&Md5Ctx, ":", 1);
1158 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce));
1159 MD5Update(&Md5Ctx, ":", 1);
1160 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce));
1161 MD5Final(HA1, &Md5Ctx);
1163 CvtHex(HA1, SessionKey);
1166 /* calculate request-digest/response-digest as per HTTP Digest spec */
1169 IN HASHHEX HA1, /* H(A1) */
1170 IN char * pszNonce, /* nonce from server */
1171 IN char * pszNonceCount, /* 8 hex digits */
1172 IN char * pszCNonce, /* client nonce */
1173 IN char * pszQop, /* qop-value: "", "auth", "auth-int" */
1174 IN char * pszMethod, /* method from the request */
1175 IN char * pszDigestUri, /* requested URL */
1176 IN HASHHEX HEntity, /* H(entity body) if qop="auth-int" */
1177 OUT HASHHEX Response /* request-digest or response-digest */
1180 /* DEBUG(fprintf(stderr,
1181 "Calc: HA1[%s] Nonce[%s] qop[%s] method[%s] URI[%s]\n",
1182 HA1, pszNonce, pszQop, pszMethod, pszDigestUri));*/
1190 MD5Update(&Md5Ctx, pszMethod, strlen(pszMethod));
1191 MD5Update(&Md5Ctx, ":", 1);
1192 MD5Update(&Md5Ctx, pszDigestUri, strlen(pszDigestUri));
1193 if (strcasecmp(pszQop, "auth-int") == 0) {
1194 MD5Update(&Md5Ctx, ":", 1);
1195 MD5Update(&Md5Ctx, HEntity, HASHHEXLEN);
1197 MD5Final(HA2, &Md5Ctx);
1198 CvtHex(HA2, HA2Hex);
1200 // calculate response
1202 MD5Update(&Md5Ctx, HA1, HASHHEXLEN);
1203 MD5Update(&Md5Ctx, ":", 1);
1204 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce));
1205 MD5Update(&Md5Ctx, ":", 1);
1207 MD5Update(&Md5Ctx, pszNonceCount, strlen(pszNonceCount));
1208 MD5Update(&Md5Ctx, ":", 1);
1209 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce));
1210 MD5Update(&Md5Ctx, ":", 1);
1211 MD5Update(&Md5Ctx, pszQop, strlen(pszQop));
1212 MD5Update(&Md5Ctx, ":", 1);
1214 MD5Update(&Md5Ctx, HA2Hex, HASHHEXLEN);
1215 MD5Final(RespHash, &Md5Ctx);
1216 CvtHex(RespHash, Response);
1220 * Generate/Send a Digest authorization header
1221 * This looks like: [Proxy-]Authorization: credentials
1223 * credentials = "Digest" digest-response
1224 * digest-response = 1#( username | realm | nonce | digest-uri
1225 * | response | [ algorithm ] | [cnonce] |
1226 * [opaque] | [message-qop] |
1227 * [nonce-count] | [auth-param] )
1228 * username = "username" "=" username-value
1229 * username-value = quoted-string
1230 * digest-uri = "uri" "=" digest-uri-value
1231 * digest-uri-value = request-uri ; As specified by HTTP/1.1
1232 * message-qop = "qop" "=" qop-value
1233 * cnonce = "cnonce" "=" cnonce-value
1234 * cnonce-value = nonce-value
1235 * nonce-count = "nc" "=" nc-value
1237 * response = "response" "=" request-digest
1238 * request-digest = <"> 32LHEX <">
1241 http_digest_auth(conn_t *conn, const char *hdr, http_auth_challenge_t *c,
1242 http_auth_params_t *parms, struct url *url)
1245 char noncecount[10];
1247 char *options = NULL;
1249 if (!c->realm || !c->nonce) {
1250 DEBUG(fprintf(stderr, "realm/nonce not set in challenge\n"));
1254 c->algo = strdup("");
1256 if (asprintf(&options, "%s%s%s%s",
1257 *c->algo? ",algorithm=" : "", c->algo,
1258 c->opaque? ",opaque=" : "", c->opaque?c->opaque:"")== -1)
1262 c->qop = strdup("");
1267 sprintf(noncecount, "%08x", c->nc);
1268 /* We don't try very hard with the cnonce ... */
1269 sprintf(cnonce, "%x%lx", getpid(), (unsigned long)time(0));
1273 DigestCalcHA1(c->algo, parms->user, c->realm,
1274 parms->password, c->nonce, cnonce, HA1);
1275 DEBUG(fprintf(stderr, "HA1: [%s]\n", HA1));
1277 DigestCalcResponse(HA1, c->nonce, noncecount, cnonce, c->qop,
1278 "GET", url->doc, "", digest);
1281 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\","
1282 "nonce=\"%s\",uri=\"%s\",response=\"%s\","
1283 "qop=\"auth\", cnonce=\"%s\", nc=%s%s",
1284 hdr, parms->user, c->realm,
1285 c->nonce, url->doc, digest,
1286 cnonce, noncecount, options);
1288 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\","
1289 "nonce=\"%s\",uri=\"%s\",response=\"%s\"%s",
1290 hdr, parms->user, c->realm,
1291 c->nonce, url->doc, digest, options);
1299 * Encode username and password
1302 http_basic_auth(conn_t *conn, const char *hdr, const char *usr, const char *pwd)
1307 DEBUG(fprintf(stderr, "basic: usr: [%s]\n", usr));
1308 DEBUG(fprintf(stderr, "basic: pwd: [%s]\n", pwd));
1309 if (asprintf(&upw, "%s:%s", usr, pwd) == -1)
1311 auth = http_base64(upw);
1315 r = http_cmd(conn, "%s: Basic %s", hdr, auth);
1321 * Chose the challenge to answer and call the appropriate routine to
1322 * produce the header.
1325 http_authorize(conn_t *conn, const char *hdr, http_auth_challenges_t *cs,
1326 http_auth_params_t *parms, struct url *url)
1328 http_auth_challenge_t *digest = NULL;
1331 /* If user or pass are null we're not happy */
1332 if (!parms->user || !parms->password) {
1333 DEBUG(fprintf(stderr, "NULL usr or pass\n"));
1337 /* Look for a Digest */
1338 for (i = 0; i < cs->count; i++) {
1339 if (cs->challenges[i]->scheme == HTTPAS_DIGEST)
1340 digest = cs->challenges[i];
1343 /* Error if "Digest" was specified and there is no Digest challenge */
1344 if (!digest && (parms->scheme &&
1345 !strcasecmp(parms->scheme, "digest"))) {
1346 DEBUG(fprintf(stderr,
1347 "Digest auth in env, not supported by peer\n"));
1351 * If "basic" was specified in the environment, or there is no Digest
1352 * challenge, do the basic thing. Don't need a challenge for this,
1353 * so no need to check basic!=NULL
1355 if (!digest || (parms->scheme && !strcasecmp(parms->scheme,"basic")))
1356 return (http_basic_auth(conn,hdr,parms->user,parms->password));
1358 /* Else, prefer digest. We just checked that it's not NULL */
1359 return (http_digest_auth(conn, hdr, digest, parms, url));
1362 /*****************************************************************************
1363 * Helper functions for connecting to a server or proxy
1367 * Connect to the correct HTTP server or proxy.
1370 http_connect(struct url *URL, struct url *purl, const char *flags)
1382 verbose = CHECK_FLAG('v');
1383 if (CHECK_FLAG('4'))
1386 else if (CHECK_FLAG('6'))
1390 if (purl && strcasecmp(URL->scheme, SCHEME_HTTPS) != 0) {
1392 } else if (strcasecmp(URL->scheme, SCHEME_FTP) == 0) {
1393 /* can't talk http to an ftp server */
1394 /* XXX should set an error code */
1398 if ((conn = fetch_connect(URL->host, URL->port, af, verbose)) == NULL)
1399 /* fetch_connect() has already set an error code */
1401 if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 &&
1402 fetch_ssl(conn, verbose) == -1) {
1411 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, sizeof(val));
1417 http_get_proxy(struct url * url, const char *flags)
1422 if (flags != NULL && strchr(flags, 'd') != NULL)
1424 if (fetch_no_proxy_match(url->host))
1426 if (((p = getenv("HTTP_PROXY")) || (p = getenv("http_proxy"))) &&
1427 *p && (purl = fetchParseURL(p))) {
1429 strcpy(purl->scheme, SCHEME_HTTP);
1431 purl->port = fetch_default_proxy_port(purl->scheme);
1432 if (strcasecmp(purl->scheme, SCHEME_HTTP) == 0)
1440 http_print_html(FILE *out, FILE *in)
1447 while ((line = fgetln(in, &len)) != NULL) {
1448 while (len && isspace((unsigned char)line[len - 1]))
1450 for (p = q = line; q < line + len; ++q) {
1451 if (comment && *q == '-') {
1452 if (q + 2 < line + len &&
1453 strcmp(q, "-->") == 0) {
1457 } else if (tag && !comment && *q == '>') {
1460 } else if (!tag && *q == '<') {
1462 fwrite(p, q - p, 1, out);
1464 if (q + 3 < line + len &&
1465 strcmp(q, "<!--") == 0) {
1472 fwrite(p, q - p, 1, out);
1478 /*****************************************************************************
1483 * Send a request and process the reply
1485 * XXX This function is way too long, the do..while loop should be split
1486 * XXX off into a separate function.
1489 http_request(struct url *URL, const char *op, struct url_stat *us,
1490 struct url *purl, const char *flags)
1493 char hbuf[MAXHOSTNAMELEN + 7], *host;
1495 struct url *url, *new;
1496 int chunked, direct, ims, noredirect, verbose;
1498 off_t offset, clength, length, size;
1503 struct tm *timestruct;
1504 http_headerbuf_t headerbuf;
1505 http_auth_challenges_t server_challenges;
1506 http_auth_challenges_t proxy_challenges;
1508 /* The following calls don't allocate anything */
1509 init_http_headerbuf(&headerbuf);
1510 init_http_auth_challenges(&server_challenges);
1511 init_http_auth_challenges(&proxy_challenges);
1513 direct = CHECK_FLAG('d');
1514 noredirect = CHECK_FLAG('A');
1515 verbose = CHECK_FLAG('v');
1516 ims = CHECK_FLAG('i');
1518 if (direct && purl) {
1523 /* try the provided URL first */
1529 e = HTTP_PROTOCOL_ERROR;
1541 url->port = fetch_default_port(url->scheme);
1543 /* were we redirected to an FTP URL? */
1544 if (purl == NULL && strcmp(url->scheme, SCHEME_FTP) == 0) {
1545 if (strcmp(op, "GET") == 0)
1546 return (ftp_request(url, "RETR", us, purl, flags));
1547 else if (strcmp(op, "HEAD") == 0)
1548 return (ftp_request(url, "STAT", us, purl, flags));
1551 /* connect to server or proxy */
1552 if ((conn = http_connect(url, purl, flags)) == NULL)
1557 if (strchr(url->host, ':')) {
1558 snprintf(hbuf, sizeof(hbuf), "[%s]", url->host);
1562 if (url->port != fetch_default_port(url->scheme)) {
1567 snprintf(hbuf + strlen(hbuf),
1568 sizeof(hbuf) - strlen(hbuf), ":%d", url->port);
1573 fetch_info("requesting %s://%s%s",
1574 url->scheme, host, url->doc);
1576 http_cmd(conn, "%s %s://%s%s HTTP/1.1",
1577 op, url->scheme, host, url->doc);
1579 http_cmd(conn, "%s %s HTTP/1.1",
1583 if (ims && url->ims_time) {
1584 timestruct = gmtime((time_t *)&url->ims_time);
1585 (void)strftime(timebuf, 80, "%a, %d %b %Y %T GMT",
1588 fetch_info("If-Modified-Since: %s", timebuf);
1589 http_cmd(conn, "If-Modified-Since: %s", timebuf);
1592 http_cmd(conn, "Host: %s", host);
1595 * Proxy authorization: we only send auth after we received
1596 * a 407 error. We do not first try basic anyway (changed
1597 * when support was added for digest-auth)
1599 if (purl && proxy_challenges.valid) {
1600 http_auth_params_t aparams;
1601 init_http_auth_params(&aparams);
1602 if (*purl->user || *purl->pwd) {
1603 aparams.user = purl->user ?
1604 strdup(purl->user) : strdup("");
1605 aparams.password = purl->pwd?
1606 strdup(purl->pwd) : strdup("");
1607 } else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL &&
1609 if (http_authfromenv(p, &aparams) < 0) {
1610 http_seterr(HTTP_NEED_PROXY_AUTH);
1614 http_authorize(conn, "Proxy-Authorization",
1615 &proxy_challenges, &aparams, url);
1616 clean_http_auth_params(&aparams);
1620 * Server authorization: we never send "a priori"
1621 * Basic auth, which used to be done if user/pass were
1622 * set in the url. This would be weird because we'd send the
1623 * password in the clear even if Digest is finally to be
1624 * used (it would have made more sense for the
1625 * pre-digest version to do this when Basic was specified
1626 * in the environment)
1628 if (server_challenges.valid) {
1629 http_auth_params_t aparams;
1630 init_http_auth_params(&aparams);
1631 if (*url->user || *url->pwd) {
1632 aparams.user = url->user ?
1633 strdup(url->user) : strdup("");
1634 aparams.password = url->pwd ?
1635 strdup(url->pwd) : strdup("");
1636 } else if ((p = getenv("HTTP_AUTH")) != NULL &&
1638 if (http_authfromenv(p, &aparams) < 0) {
1639 http_seterr(HTTP_NEED_AUTH);
1642 } else if (fetchAuthMethod &&
1643 fetchAuthMethod(url) == 0) {
1644 aparams.user = url->user ?
1645 strdup(url->user) : strdup("");
1646 aparams.password = url->pwd ?
1647 strdup(url->pwd) : strdup("");
1649 http_seterr(HTTP_NEED_AUTH);
1652 http_authorize(conn, "Authorization",
1653 &server_challenges, &aparams, url);
1654 clean_http_auth_params(&aparams);
1658 if ((p = getenv("HTTP_REFERER")) != NULL && *p != '\0') {
1659 if (strcasecmp(p, "auto") == 0)
1660 http_cmd(conn, "Referer: %s://%s%s",
1661 url->scheme, host, url->doc);
1663 http_cmd(conn, "Referer: %s", p);
1665 if ((p = getenv("HTTP_USER_AGENT")) != NULL && *p != '\0')
1666 http_cmd(conn, "User-Agent: %s", p);
1668 http_cmd(conn, "User-Agent: %s " _LIBFETCH_VER, getprogname());
1669 if (url->offset > 0)
1670 http_cmd(conn, "Range: bytes=%lld-", (long long)url->offset);
1671 http_cmd(conn, "Connection: close");
1672 http_cmd(conn, "%s", "");
1675 * Force the queued request to be dispatched. Normally, one
1676 * would do this with shutdown(2) but squid proxies can be
1677 * configured to disallow such half-closed connections. To
1678 * be compatible with such configurations, fiddle with socket
1679 * options to force the pending data to be written.
1682 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val,
1685 setsockopt(conn->sd, IPPROTO_TCP, TCP_NODELAY, &val,
1689 switch (http_get_reply(conn)) {
1692 case HTTP_NOT_MODIFIED:
1695 case HTTP_MOVED_PERM:
1696 case HTTP_MOVED_TEMP:
1697 case HTTP_SEE_OTHER:
1698 case HTTP_USE_PROXY:
1700 * Not so fine, but we still have to read the
1701 * headers to get the new location.
1704 case HTTP_NEED_AUTH:
1705 if (server_challenges.valid) {
1707 * We already sent out authorization code,
1708 * so there's nothing more we can do.
1710 http_seterr(conn->err);
1713 /* try again, but send the password this time */
1715 fetch_info("server requires authorization");
1717 case HTTP_NEED_PROXY_AUTH:
1718 if (proxy_challenges.valid) {
1720 * We already sent our proxy
1721 * authorization code, so there's
1722 * nothing more we can do. */
1723 http_seterr(conn->err);
1726 /* try again, but send the password this time */
1728 fetch_info("proxy requires authorization");
1730 case HTTP_BAD_RANGE:
1732 * This can happen if we ask for 0 bytes because
1733 * we already have the whole file. Consider this
1734 * a success for now, and check sizes later.
1737 case HTTP_PROTOCOL_ERROR:
1743 http_seterr(conn->err);
1746 /* fall through so we can get the full error message */
1749 /* get headers. http_next_header expects one line readahead */
1750 if (fetch_getln(conn) == -1) {
1755 switch ((h = http_next_header(conn, &headerbuf, &p))) {
1760 http_seterr(HTTP_PROTOCOL_ERROR);
1762 case hdr_content_length:
1763 http_parse_length(p, &clength);
1765 case hdr_content_range:
1766 http_parse_range(p, &offset, &length, &size);
1768 case hdr_last_modified:
1769 http_parse_mtime(p, &mtime);
1772 if (!HTTP_REDIRECT(conn->err))
1775 * if the A flag is set, we don't follow
1776 * temporary redirects.
1779 conn->err != HTTP_MOVED_PERM &&
1780 conn->err != HTTP_PERM_REDIRECT &&
1781 conn->err != HTTP_USE_PROXY) {
1788 fetch_info("%d redirect to %s", conn->err, p);
1791 new = fetchMakeURL(url->scheme, url->host, url->port, p,
1792 url->user, url->pwd);
1794 new = fetchParseURL(p);
1796 /* XXX should set an error code */
1797 DEBUG(fprintf(stderr, "failed to parse new URL\n"));
1801 /* Only copy credentials if the host matches */
1802 if (!strcmp(new->host, url->host) && !*new->user && !*new->pwd) {
1803 strcpy(new->user, url->user);
1804 strcpy(new->pwd, url->pwd);
1806 new->offset = url->offset;
1807 new->length = url->length;
1809 case hdr_transfer_encoding:
1811 chunked = (strcasecmp(p, "chunked") == 0);
1813 case hdr_www_authenticate:
1814 if (conn->err != HTTP_NEED_AUTH)
1816 if (http_parse_authenticate(p, &server_challenges) == 0)
1819 case hdr_proxy_authenticate:
1820 if (conn->err != HTTP_NEED_PROXY_AUTH)
1822 if (http_parse_authenticate(p, &proxy_challenges) == 0)
1831 } while (h > hdr_end);
1833 /* we need to provide authentication */
1834 if (conn->err == HTTP_NEED_AUTH ||
1835 conn->err == HTTP_NEED_PROXY_AUTH) {
1837 if ((conn->err == HTTP_NEED_AUTH &&
1838 !server_challenges.valid) ||
1839 (conn->err == HTTP_NEED_PROXY_AUTH &&
1840 !proxy_challenges.valid)) {
1841 /* 401/7 but no www/proxy-authenticate ?? */
1842 DEBUG(fprintf(stderr, "401/7 and no auth header\n"));
1850 /* requested range not satisfiable */
1851 if (conn->err == HTTP_BAD_RANGE) {
1852 if (url->offset == size && url->length == 0) {
1853 /* asked for 0 bytes; fake it */
1854 offset = url->offset;
1856 conn->err = HTTP_OK;
1859 http_seterr(conn->err);
1864 /* we have a hit or an error */
1865 if (conn->err == HTTP_OK
1866 || conn->err == HTTP_NOT_MODIFIED
1867 || conn->err == HTTP_PARTIAL
1868 || HTTP_ERROR(conn->err))
1871 /* all other cases: we got a redirect */
1873 clean_http_auth_challenges(&server_challenges);
1877 DEBUG(fprintf(stderr, "redirect with no new location\n"));
1885 /* we failed, or ran out of retries */
1891 DEBUG(fprintf(stderr, "offset %lld, length %lld,"
1892 " size %lld, clength %lld\n",
1893 (long long)offset, (long long)length,
1894 (long long)size, (long long)clength));
1896 if (conn->err == HTTP_NOT_MODIFIED) {
1897 http_seterr(HTTP_NOT_MODIFIED);
1901 /* check for inconsistencies */
1902 if (clength != -1 && length != -1 && clength != length) {
1903 http_seterr(HTTP_PROTOCOL_ERROR);
1909 length = offset + clength;
1910 if (length != -1 && size != -1 && length != size) {
1911 http_seterr(HTTP_PROTOCOL_ERROR);
1920 us->atime = us->mtime = mtime;
1924 if (URL->offset > 0 && offset > URL->offset) {
1925 http_seterr(HTTP_PROTOCOL_ERROR);
1929 /* report back real offset and size */
1930 URL->offset = offset;
1931 URL->length = clength;
1933 /* wrap it up in a FILE */
1934 if ((f = http_funopen(conn, chunked)) == NULL) {
1944 if (HTTP_ERROR(conn->err)) {
1945 http_print_html(stderr, f);
1949 clean_http_headerbuf(&headerbuf);
1950 clean_http_auth_challenges(&server_challenges);
1951 clean_http_auth_challenges(&proxy_challenges);
1961 clean_http_headerbuf(&headerbuf);
1962 clean_http_auth_challenges(&server_challenges);
1963 clean_http_auth_challenges(&proxy_challenges);
1968 /*****************************************************************************
1973 * Retrieve and stat a file by HTTP
1976 fetchXGetHTTP(struct url *URL, struct url_stat *us, const char *flags)
1978 return (http_request(URL, "GET", us, http_get_proxy(URL, flags), flags));
1982 * Retrieve a file by HTTP
1985 fetchGetHTTP(struct url *URL, const char *flags)
1987 return (fetchXGetHTTP(URL, NULL, flags));
1991 * Store a file by HTTP
1994 fetchPutHTTP(struct url *URL __unused, const char *flags __unused)
1996 warnx("fetchPutHTTP(): not implemented");
2001 * Get an HTTP document's metadata
2004 fetchStatHTTP(struct url *URL, struct url_stat *us, const char *flags)
2008 f = http_request(URL, "HEAD", us, http_get_proxy(URL, flags), flags);
2019 fetchListHTTP(struct url *url __unused, const char *flags __unused)
2021 warnx("fetchListHTTP(): not implemented");