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.
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)
206 if (io->chunked == 0) {
207 if (http_growbuf(io, len) == -1)
209 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) {
218 if (io->chunksize == 0) {
219 switch (http_new_chunk(io)) {
229 if (len > io->chunksize)
231 if (http_growbuf(io, len) == -1)
233 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) {
238 io->chunksize -= io->buflen;
240 if (io->chunksize == 0) {
243 if (fetch_read(io->conn, endl, 2) != 2 ||
244 endl[0] != '\r' || endl[1] != '\n')
257 http_readfn(void *v, char *buf, int len)
259 struct httpio *io = (struct httpio *)v;
267 for (pos = 0; len > 0; pos += l, len -= l) {
269 if (!io->buf || io->bufpos == io->buflen)
270 if (http_fillbuf(io, len) < 1)
272 l = io->buflen - io->bufpos;
275 memcpy(buf + pos, io->buf + io->bufpos, l);
279 if (!pos && io->error) {
280 if (io->error == EINTR)
291 http_writefn(void *v, const char *buf, int len)
293 struct httpio *io = (struct httpio *)v;
295 return (fetch_write(io->conn, buf, len));
302 http_closefn(void *v)
304 struct httpio *io = (struct httpio *)v;
307 r = fetch_close(io->conn);
315 * Wrap a file descriptor up
318 http_funopen(conn_t *conn, int chunked)
323 if ((io = calloc(1, sizeof(*io))) == NULL) {
328 io->chunked = chunked;
329 f = funopen(io, http_readfn, http_writefn, NULL, http_closefn);
339 /*****************************************************************************
340 * Helper functions for talking to the server and parsing its replies
353 hdr_transfer_encoding,
354 hdr_www_authenticate,
355 hdr_proxy_authenticate,
358 /* Names of interesting headers */
363 { hdr_content_length, "Content-Length" },
364 { hdr_content_range, "Content-Range" },
365 { hdr_last_modified, "Last-Modified" },
366 { hdr_location, "Location" },
367 { hdr_transfer_encoding, "Transfer-Encoding" },
368 { hdr_www_authenticate, "WWW-Authenticate" },
369 { hdr_proxy_authenticate, "Proxy-Authenticate" },
370 { hdr_unknown, NULL },
374 * Send a formatted line; optionally echo to terminal
377 http_cmd(conn_t *conn, const char *fmt, ...)
385 len = vasprintf(&msg, fmt, ap);
394 r = fetch_putln(conn, msg, len);
406 * Get and parse status line
409 http_get_reply(conn_t *conn)
413 if (fetch_getln(conn) == -1)
416 * A valid status line looks like "HTTP/m.n xyz reason" where m
417 * and n are the major and minor protocol version numbers and xyz
419 * Unfortunately, there are servers out there (NCSA 1.5.1, to name
420 * just one) that do not send a version number, so we can't rely
421 * on finding one, but if we do, insist on it being 1.0 or 1.1.
422 * We don't care about the reason phrase.
424 if (strncmp(conn->buf, "HTTP", 4) != 0)
425 return (HTTP_PROTOCOL_ERROR);
428 if (p[1] != '1' || p[2] != '.' || (p[3] != '0' && p[3] != '1'))
429 return (HTTP_PROTOCOL_ERROR);
433 !isdigit((unsigned char)p[1]) ||
434 !isdigit((unsigned char)p[2]) ||
435 !isdigit((unsigned char)p[3]))
436 return (HTTP_PROTOCOL_ERROR);
438 conn->err = (p[1] - '0') * 100 + (p[2] - '0') * 10 + (p[3] - '0');
443 * Check a header; if the type matches the given string, return a pointer
444 * to the beginning of the value.
447 http_match(const char *str, const char *hdr)
449 while (*str && *hdr &&
450 tolower((unsigned char)*str++) == tolower((unsigned char)*hdr++))
452 if (*str || *hdr != ':')
454 while (*hdr && isspace((unsigned char)*++hdr))
461 * Get the next header and return the appropriate symbolic code. We
462 * need to read one line ahead for checking for a continuation line
463 * belonging to the current header (continuation lines start with
466 * We get called with a fresh line already in the conn buffer, either
467 * from the previous http_next_header() invocation, or, the first
468 * time, from a fetch_getln() performed by our caller.
470 * This stops when we encounter an empty line (we dont read beyond the header
473 * Note that the "headerbuf" is just a place to return the result. Its
474 * contents are not used for the next call. This means that no cleanup
475 * is needed when ie doing another connection, just call the cleanup when
476 * fully done to deallocate memory.
479 /* Limit the max number of continuation lines to some reasonable value */
480 #define HTTP_MAX_CONT_LINES 10
482 /* Place into which to build a header from one or several lines */
484 char *buf; /* buffer */
485 size_t bufsize; /* buffer size */
486 size_t buflen; /* length of buffer contents */
490 init_http_headerbuf(http_headerbuf_t *buf)
498 clean_http_headerbuf(http_headerbuf_t *buf)
502 init_http_headerbuf(buf);
505 /* Remove whitespace at the end of the buffer */
507 http_conn_trimright(conn_t *conn)
509 while (conn->buflen &&
510 isspace((unsigned char)conn->buf[conn->buflen - 1]))
512 conn->buf[conn->buflen] = '\0';
516 http_next_header(conn_t *conn, http_headerbuf_t *hbuf, const char **p)
521 * Have to do the stripping here because of the first line. So
522 * it's done twice for the subsequent lines. No big deal
524 http_conn_trimright(conn);
525 if (conn->buflen == 0)
528 /* Copy the line to the headerbuf */
529 if (hbuf->bufsize < conn->buflen + 1) {
530 if ((hbuf->buf = realloc(hbuf->buf, conn->buflen + 1)) == NULL)
531 return (hdr_syserror);
532 hbuf->bufsize = conn->buflen + 1;
534 strcpy(hbuf->buf, conn->buf);
535 hbuf->buflen = conn->buflen;
538 * Fetch possible continuation lines. Stop at 1st non-continuation
539 * and leave it in the conn buffer
541 for (i = 0; i < HTTP_MAX_CONT_LINES; i++) {
542 if (fetch_getln(conn) == -1)
543 return (hdr_syserror);
546 * Note: we carry on the idea from the previous version
547 * that a pure whitespace line is equivalent to an empty
548 * one (so it's not continuation and will be handled when
549 * we are called next)
551 http_conn_trimright(conn);
552 if (conn->buf[0] != ' ' && conn->buf[0] != "\t"[0])
555 /* Got a continuation line. Concatenate to previous */
556 len = hbuf->buflen + conn->buflen;
557 if (hbuf->bufsize < len + 1) {
559 if ((hbuf->buf = realloc(hbuf->buf, len + 1)) == NULL)
560 return (hdr_syserror);
561 hbuf->bufsize = len + 1;
563 strcpy(hbuf->buf + hbuf->buflen, conn->buf);
564 hbuf->buflen += conn->buflen;
568 * We could check for malformed headers but we don't really care.
569 * A valid header starts with a token immediately followed by a
570 * colon; a token is any sequence of non-control, non-whitespace
571 * characters except "()<>@,;:\\\"{}".
573 for (i = 0; hdr_names[i].num != hdr_unknown; i++)
574 if ((*p = http_match(hdr_names[i].name, hbuf->buf)) != NULL)
575 return (hdr_names[i].num);
577 return (hdr_unknown);
580 /**************************
581 * [Proxy-]Authenticate header parsing
585 * Read doublequote-delimited string into output buffer obuf (allocated
586 * by caller, whose responsibility it is to ensure that it's big enough)
587 * cp points to the first char after the initial '"'
589 * Returns pointer to the first char after the terminating double quote, or
593 http_parse_headerstring(const char *cp, char *obuf)
597 case 0: /* Unterminated string */
600 case '"': /* Ending quote */
615 /* Http auth challenge schemes */
616 typedef enum {HTTPAS_UNKNOWN, HTTPAS_BASIC,HTTPAS_DIGEST} http_auth_schemes_t;
618 /* Data holder for a Basic or Digest challenge. */
620 http_auth_schemes_t scheme;
627 int nc; /* Nonce count */
628 } http_auth_challenge_t;
631 init_http_auth_challenge(http_auth_challenge_t *b)
633 b->scheme = HTTPAS_UNKNOWN;
634 b->realm = b->qop = b->nonce = b->opaque = b->algo = NULL;
635 b->stale = b->nc = 0;
639 clean_http_auth_challenge(http_auth_challenge_t *b)
651 init_http_auth_challenge(b);
654 /* Data holder for an array of challenges offered in an http response. */
655 #define MAX_CHALLENGES 10
657 http_auth_challenge_t *challenges[MAX_CHALLENGES];
658 int count; /* Number of parsed challenges in the array */
659 int valid; /* We did parse an authenticate header */
660 } http_auth_challenges_t;
663 init_http_auth_challenges(http_auth_challenges_t *cs)
666 for (i = 0; i < MAX_CHALLENGES; i++)
667 cs->challenges[i] = NULL;
668 cs->count = cs->valid = 0;
672 clean_http_auth_challenges(http_auth_challenges_t *cs)
675 /* We rely on non-zero pointers being allocated, not on the count */
676 for (i = 0; i < MAX_CHALLENGES; i++) {
677 if (cs->challenges[i] != NULL) {
678 clean_http_auth_challenge(cs->challenges[i]);
679 free(cs->challenges[i]);
682 init_http_auth_challenges(cs);
686 * Enumeration for lexical elements. Separators will be returned as their own
689 typedef enum {HTTPHL_WORD=256, HTTPHL_STRING=257, HTTPHL_END=258,
690 HTTPHL_ERROR = 259} http_header_lex_t;
693 * Determine what kind of token comes next and return possible value
694 * in buf, which is supposed to have been allocated big enough by
695 * caller. Advance input pointer and return element type.
698 http_header_lex(const char **cpp, char *buf)
701 /* Eat initial whitespace */
702 *cpp += strspn(*cpp, " \t");
707 if (**cpp == ',' || **cpp == '=')
708 return (*((*cpp)++));
712 *cpp = http_parse_headerstring(++*cpp, buf);
714 return (HTTPHL_ERROR);
715 return (HTTPHL_STRING);
718 /* Read other token, until separator or whitespace */
719 l = strcspn(*cpp, " \t,=");
720 memcpy(buf, *cpp, l);
723 return (HTTPHL_WORD);
727 * Read challenges from http xxx-authenticate header and accumulate them
728 * in the challenges list structure.
730 * Headers with multiple challenges are specified by rfc2617, but
731 * servers (ie: squid) often send them in separate headers instead,
732 * which in turn is forbidden by the http spec (multiple headers with
733 * the same name are only allowed for pure comma-separated lists, see
736 * We support both approaches anyway
739 http_parse_authenticate(const char *cp, http_auth_challenges_t *cs)
742 http_header_lex_t lex;
743 char *key = malloc(strlen(cp) + 1);
744 char *value = malloc(strlen(cp) + 1);
745 char *buf = malloc(strlen(cp) + 1);
747 if (key == NULL || value == NULL || buf == NULL) {
752 /* In any case we've seen the header and we set the valid bit */
755 /* Need word first */
756 lex = http_header_lex(&cp, key);
757 if (lex != HTTPHL_WORD)
760 /* Loop on challenges */
761 for (; cs->count < MAX_CHALLENGES; cs->count++) {
762 cs->challenges[cs->count] =
763 malloc(sizeof(http_auth_challenge_t));
764 if (cs->challenges[cs->count] == NULL) {
768 init_http_auth_challenge(cs->challenges[cs->count]);
769 if (!strcasecmp(key, "basic")) {
770 cs->challenges[cs->count]->scheme = HTTPAS_BASIC;
771 } else if (!strcasecmp(key, "digest")) {
772 cs->challenges[cs->count]->scheme = HTTPAS_DIGEST;
774 cs->challenges[cs->count]->scheme = HTTPAS_UNKNOWN;
776 * Continue parsing as basic or digest may
777 * follow, and the syntax is the same for
778 * all. We'll just ignore this one when
779 * looking at the list
783 /* Loop on attributes */
786 lex = http_header_lex(&cp, key);
787 if (lex != HTTPHL_WORD)
791 lex = http_header_lex(&cp, buf);
796 lex = http_header_lex(&cp, value);
797 if (lex != HTTPHL_WORD && lex != HTTPHL_STRING)
800 if (!strcasecmp(key, "realm"))
801 cs->challenges[cs->count]->realm =
803 else if (!strcasecmp(key, "qop"))
804 cs->challenges[cs->count]->qop =
806 else if (!strcasecmp(key, "nonce"))
807 cs->challenges[cs->count]->nonce =
809 else if (!strcasecmp(key, "opaque"))
810 cs->challenges[cs->count]->opaque =
812 else if (!strcasecmp(key, "algorithm"))
813 cs->challenges[cs->count]->algo =
815 else if (!strcasecmp(key, "stale"))
816 cs->challenges[cs->count]->stale =
817 strcasecmp(value, "no");
818 /* Else ignore unknown attributes */
820 /* Comma or Next challenge or End */
821 lex = http_header_lex(&cp, key);
823 * If we get a word here, this is the beginning of the
824 * next challenge. Break the attributes loop
826 if (lex == HTTPHL_WORD)
829 if (lex == HTTPHL_END) {
830 /* End while looking for ',' is normal exit */
835 /* Anything else is an error */
839 } /* End attributes loop */
840 } /* End challenge loop */
843 * Challenges max count exceeded. This really can't happen
844 * with normal data, something's fishy -> error
859 * Parse a last-modified header
862 http_parse_mtime(const char *p, time_t *mtime)
867 strncpy(locale, setlocale(LC_TIME, NULL), sizeof(locale));
868 setlocale(LC_TIME, "C");
869 r = strptime(p, "%a, %d %b %Y %H:%M:%S GMT", &tm);
870 /* XXX should add support for date-2 and date-3 */
871 setlocale(LC_TIME, locale);
874 DEBUG(fprintf(stderr, "last modified: [%04d-%02d-%02d "
876 tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday,
877 tm.tm_hour, tm.tm_min, tm.tm_sec));
878 *mtime = timegm(&tm);
883 * Parse a content-length header
886 http_parse_length(const char *p, off_t *length)
890 for (len = 0; *p && isdigit((unsigned char)*p); ++p)
891 len = len * 10 + (*p - '0');
894 DEBUG(fprintf(stderr, "content length: [%lld]\n",
901 * Parse a content-range header
904 http_parse_range(const char *p, off_t *offset, off_t *length, off_t *size)
906 off_t first, last, len;
908 if (strncasecmp(p, "bytes ", 6) != 0)
915 for (first = 0; *p && isdigit((unsigned char)*p); ++p)
916 first = first * 10 + *p - '0';
919 for (last = 0, ++p; *p && isdigit((unsigned char)*p); ++p)
920 last = last * 10 + *p - '0';
922 if (first > last || *p != '/')
924 for (len = 0, ++p; *p && isdigit((unsigned char)*p); ++p)
925 len = len * 10 + *p - '0';
926 if (*p || len < last - first + 1)
929 DEBUG(fprintf(stderr, "content range: [*/%lld]\n",
933 DEBUG(fprintf(stderr, "content range: [%lld-%lld/%lld]\n",
934 (long long)first, (long long)last, (long long)len));
935 *length = last - first + 1;
943 /*****************************************************************************
944 * Helper functions for authorization
951 http_base64(const char *src)
953 static const char base64[] =
954 "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
955 "abcdefghijklmnopqrstuvwxyz"
962 if ((str = malloc(((l + 2) / 3) * 4 + 1)) == NULL)
968 t = (src[0] << 16) | (src[1] << 8) | src[2];
969 dst[0] = base64[(t >> 18) & 0x3f];
970 dst[1] = base64[(t >> 12) & 0x3f];
971 dst[2] = base64[(t >> 6) & 0x3f];
972 dst[3] = base64[(t >> 0) & 0x3f];
979 t = (src[0] << 16) | (src[1] << 8);
980 dst[0] = base64[(t >> 18) & 0x3f];
981 dst[1] = base64[(t >> 12) & 0x3f];
982 dst[2] = base64[(t >> 6) & 0x3f];
989 dst[0] = base64[(t >> 18) & 0x3f];
990 dst[1] = base64[(t >> 12) & 0x3f];
991 dst[2] = dst[3] = '=';
1005 * Extract authorization parameters from environment value.
1006 * The value is like scheme:realm:user:pass
1013 } http_auth_params_t;
1016 init_http_auth_params(http_auth_params_t *s)
1018 s->scheme = s->realm = s->user = s->password = 0;
1022 clean_http_auth_params(http_auth_params_t *s)
1032 init_http_auth_params(s);
1036 http_authfromenv(const char *p, http_auth_params_t *parms)
1040 char *str = strdup(p);
1048 if ((ve = strchr(v, ':')) == NULL)
1052 if ((parms->scheme = strdup(v)) == NULL) {
1058 if ((ve = strchr(v, ':')) == NULL)
1062 if ((parms->realm = strdup(v)) == NULL) {
1068 if ((ve = strchr(v, ':')) == NULL)
1072 if ((parms->user = strdup(v)) == NULL) {
1079 if ((parms->password = strdup(v)) == NULL) {
1086 clean_http_auth_params(parms);
1094 * Digest response: the code to compute the digest is taken from the
1095 * sample implementation in RFC2616
1101 typedef char HASH[HASHLEN];
1102 #define HASHHEXLEN 32
1103 typedef char HASHHEX[HASHHEXLEN+1];
1105 static const char *hexchars = "0123456789abcdef";
1107 CvtHex(IN HASH Bin, OUT HASHHEX Hex)
1112 for (i = 0; i < HASHLEN; i++) {
1113 j = (Bin[i] >> 4) & 0xf;
1114 Hex[i*2] = hexchars[j];
1116 Hex[i*2+1] = hexchars[j];
1118 Hex[HASHHEXLEN] = '\0';
1121 /* calculate H(A1) as per spec */
1125 IN char * pszUserName,
1127 IN char * pszPassword,
1129 IN char * pszCNonce,
1130 OUT HASHHEX SessionKey
1137 MD5Update(&Md5Ctx, pszUserName, strlen(pszUserName));
1138 MD5Update(&Md5Ctx, ":", 1);
1139 MD5Update(&Md5Ctx, pszRealm, strlen(pszRealm));
1140 MD5Update(&Md5Ctx, ":", 1);
1141 MD5Update(&Md5Ctx, pszPassword, strlen(pszPassword));
1142 MD5Final(HA1, &Md5Ctx);
1143 if (strcasecmp(pszAlg, "md5-sess") == 0) {
1146 MD5Update(&Md5Ctx, HA1, HASHLEN);
1147 MD5Update(&Md5Ctx, ":", 1);
1148 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce));
1149 MD5Update(&Md5Ctx, ":", 1);
1150 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce));
1151 MD5Final(HA1, &Md5Ctx);
1153 CvtHex(HA1, SessionKey);
1156 /* calculate request-digest/response-digest as per HTTP Digest spec */
1159 IN HASHHEX HA1, /* H(A1) */
1160 IN char * pszNonce, /* nonce from server */
1161 IN char * pszNonceCount, /* 8 hex digits */
1162 IN char * pszCNonce, /* client nonce */
1163 IN char * pszQop, /* qop-value: "", "auth", "auth-int" */
1164 IN char * pszMethod, /* method from the request */
1165 IN char * pszDigestUri, /* requested URL */
1166 IN HASHHEX HEntity, /* H(entity body) if qop="auth-int" */
1167 OUT HASHHEX Response /* request-digest or response-digest */
1170 /* DEBUG(fprintf(stderr,
1171 "Calc: HA1[%s] Nonce[%s] qop[%s] method[%s] URI[%s]\n",
1172 HA1, pszNonce, pszQop, pszMethod, pszDigestUri));*/
1180 MD5Update(&Md5Ctx, pszMethod, strlen(pszMethod));
1181 MD5Update(&Md5Ctx, ":", 1);
1182 MD5Update(&Md5Ctx, pszDigestUri, strlen(pszDigestUri));
1183 if (strcasecmp(pszQop, "auth-int") == 0) {
1184 MD5Update(&Md5Ctx, ":", 1);
1185 MD5Update(&Md5Ctx, HEntity, HASHHEXLEN);
1187 MD5Final(HA2, &Md5Ctx);
1188 CvtHex(HA2, HA2Hex);
1190 // calculate response
1192 MD5Update(&Md5Ctx, HA1, HASHHEXLEN);
1193 MD5Update(&Md5Ctx, ":", 1);
1194 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce));
1195 MD5Update(&Md5Ctx, ":", 1);
1197 MD5Update(&Md5Ctx, pszNonceCount, strlen(pszNonceCount));
1198 MD5Update(&Md5Ctx, ":", 1);
1199 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce));
1200 MD5Update(&Md5Ctx, ":", 1);
1201 MD5Update(&Md5Ctx, pszQop, strlen(pszQop));
1202 MD5Update(&Md5Ctx, ":", 1);
1204 MD5Update(&Md5Ctx, HA2Hex, HASHHEXLEN);
1205 MD5Final(RespHash, &Md5Ctx);
1206 CvtHex(RespHash, Response);
1210 * Generate/Send a Digest authorization header
1211 * This looks like: [Proxy-]Authorization: credentials
1213 * credentials = "Digest" digest-response
1214 * digest-response = 1#( username | realm | nonce | digest-uri
1215 * | response | [ algorithm ] | [cnonce] |
1216 * [opaque] | [message-qop] |
1217 * [nonce-count] | [auth-param] )
1218 * username = "username" "=" username-value
1219 * username-value = quoted-string
1220 * digest-uri = "uri" "=" digest-uri-value
1221 * digest-uri-value = request-uri ; As specified by HTTP/1.1
1222 * message-qop = "qop" "=" qop-value
1223 * cnonce = "cnonce" "=" cnonce-value
1224 * cnonce-value = nonce-value
1225 * nonce-count = "nc" "=" nc-value
1227 * response = "response" "=" request-digest
1228 * request-digest = <"> 32LHEX <">
1231 http_digest_auth(conn_t *conn, const char *hdr, http_auth_challenge_t *c,
1232 http_auth_params_t *parms, struct url *url)
1235 char noncecount[10];
1239 if (!c->realm || !c->nonce) {
1240 DEBUG(fprintf(stderr, "realm/nonce not set in challenge\n"));
1244 c->algo = strdup("");
1246 if (asprintf(&options, "%s%s%s%s",
1247 *c->algo? ",algorithm=" : "", c->algo,
1248 c->opaque? ",opaque=" : "", c->opaque?c->opaque:"")== -1)
1252 c->qop = strdup("");
1257 sprintf(noncecount, "%08x", c->nc);
1258 /* We don't try very hard with the cnonce ... */
1259 sprintf(cnonce, "%x%lx", getpid(), (unsigned long)time(0));
1263 DigestCalcHA1(c->algo, parms->user, c->realm,
1264 parms->password, c->nonce, cnonce, HA1);
1265 DEBUG(fprintf(stderr, "HA1: [%s]\n", HA1));
1267 DigestCalcResponse(HA1, c->nonce, noncecount, cnonce, c->qop,
1268 "GET", url->doc, "", digest);
1271 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\","
1272 "nonce=\"%s\",uri=\"%s\",response=\"%s\","
1273 "qop=\"auth\", cnonce=\"%s\", nc=%s%s",
1274 hdr, parms->user, c->realm,
1275 c->nonce, url->doc, digest,
1276 cnonce, noncecount, options);
1278 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\","
1279 "nonce=\"%s\",uri=\"%s\",response=\"%s\"%s",
1280 hdr, parms->user, c->realm,
1281 c->nonce, url->doc, digest, options);
1289 * Encode username and password
1292 http_basic_auth(conn_t *conn, const char *hdr, const char *usr, const char *pwd)
1297 DEBUG(fprintf(stderr, "basic: usr: [%s]\n", usr));
1298 DEBUG(fprintf(stderr, "basic: pwd: [%s]\n", pwd));
1299 if (asprintf(&upw, "%s:%s", usr, pwd) == -1)
1301 auth = http_base64(upw);
1305 r = http_cmd(conn, "%s: Basic %s", hdr, auth);
1311 * Chose the challenge to answer and call the appropriate routine to
1312 * produce the header.
1315 http_authorize(conn_t *conn, const char *hdr, http_auth_challenges_t *cs,
1316 http_auth_params_t *parms, struct url *url)
1318 http_auth_challenge_t *basic = NULL;
1319 http_auth_challenge_t *digest = NULL;
1322 /* If user or pass are null we're not happy */
1323 if (!parms->user || !parms->password) {
1324 DEBUG(fprintf(stderr, "NULL usr or pass\n"));
1328 /* Look for a Digest and a Basic challenge */
1329 for (i = 0; i < cs->count; i++) {
1330 if (cs->challenges[i]->scheme == HTTPAS_BASIC)
1331 basic = cs->challenges[i];
1332 if (cs->challenges[i]->scheme == HTTPAS_DIGEST)
1333 digest = cs->challenges[i];
1336 /* Error if "Digest" was specified and there is no Digest challenge */
1337 if (!digest && (parms->scheme &&
1338 !strcasecmp(parms->scheme, "digest"))) {
1339 DEBUG(fprintf(stderr,
1340 "Digest auth in env, not supported by peer\n"));
1344 * If "basic" was specified in the environment, or there is no Digest
1345 * challenge, do the basic thing. Don't need a challenge for this,
1346 * so no need to check basic!=NULL
1348 if (!digest || (parms->scheme && !strcasecmp(parms->scheme,"basic")))
1349 return (http_basic_auth(conn,hdr,parms->user,parms->password));
1351 /* Else, prefer digest. We just checked that it's not NULL */
1352 return (http_digest_auth(conn, hdr, digest, parms, url));
1355 /*****************************************************************************
1356 * Helper functions for connecting to a server or proxy
1360 * Connect to the correct HTTP server or proxy.
1363 http_connect(struct url *URL, struct url *purl, const char *flags)
1375 verbose = CHECK_FLAG('v');
1376 if (CHECK_FLAG('4'))
1379 else if (CHECK_FLAG('6'))
1383 if (purl && strcasecmp(URL->scheme, SCHEME_HTTPS) != 0) {
1385 } else if (strcasecmp(URL->scheme, SCHEME_FTP) == 0) {
1386 /* can't talk http to an ftp server */
1387 /* XXX should set an error code */
1391 if ((conn = fetch_connect(URL->host, URL->port, af, verbose)) == NULL)
1392 /* fetch_connect() has already set an error code */
1394 if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 &&
1395 fetch_ssl(conn, verbose) == -1) {
1404 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, sizeof(val));
1410 http_get_proxy(struct url * url, const char *flags)
1415 if (flags != NULL && strchr(flags, 'd') != NULL)
1417 if (fetch_no_proxy_match(url->host))
1419 if (((p = getenv("HTTP_PROXY")) || (p = getenv("http_proxy"))) &&
1420 *p && (purl = fetchParseURL(p))) {
1422 strcpy(purl->scheme, SCHEME_HTTP);
1424 purl->port = fetch_default_proxy_port(purl->scheme);
1425 if (strcasecmp(purl->scheme, SCHEME_HTTP) == 0)
1433 http_print_html(FILE *out, FILE *in)
1440 while ((line = fgetln(in, &len)) != NULL) {
1441 while (len && isspace((unsigned char)line[len - 1]))
1443 for (p = q = line; q < line + len; ++q) {
1444 if (comment && *q == '-') {
1445 if (q + 2 < line + len &&
1446 strcmp(q, "-->") == 0) {
1450 } else if (tag && !comment && *q == '>') {
1453 } else if (!tag && *q == '<') {
1455 fwrite(p, q - p, 1, out);
1457 if (q + 3 < line + len &&
1458 strcmp(q, "<!--") == 0) {
1465 fwrite(p, q - p, 1, out);
1471 /*****************************************************************************
1476 * Send a request and process the reply
1478 * XXX This function is way too long, the do..while loop should be split
1479 * XXX off into a separate function.
1482 http_request(struct url *URL, const char *op, struct url_stat *us,
1483 struct url *purl, const char *flags)
1486 char hbuf[MAXHOSTNAMELEN + 7], *host;
1488 struct url *url, *new;
1489 int chunked, direct, ims, noredirect, verbose;
1491 off_t offset, clength, length, size;
1496 struct tm *timestruct;
1497 http_headerbuf_t headerbuf;
1498 http_auth_challenges_t server_challenges;
1499 http_auth_challenges_t proxy_challenges;
1501 /* The following calls don't allocate anything */
1502 init_http_headerbuf(&headerbuf);
1503 init_http_auth_challenges(&server_challenges);
1504 init_http_auth_challenges(&proxy_challenges);
1506 direct = CHECK_FLAG('d');
1507 noredirect = CHECK_FLAG('A');
1508 verbose = CHECK_FLAG('v');
1509 ims = CHECK_FLAG('i');
1511 if (direct && purl) {
1516 /* try the provided URL first */
1519 /* if the A flag is set, we only get one try */
1520 n = noredirect ? 1 : MAX_REDIRECT;
1523 e = HTTP_PROTOCOL_ERROR;
1535 url->port = fetch_default_port(url->scheme);
1537 /* were we redirected to an FTP URL? */
1538 if (purl == NULL && strcmp(url->scheme, SCHEME_FTP) == 0) {
1539 if (strcmp(op, "GET") == 0)
1540 return (ftp_request(url, "RETR", us, purl, flags));
1541 else if (strcmp(op, "HEAD") == 0)
1542 return (ftp_request(url, "STAT", us, purl, flags));
1545 /* connect to server or proxy */
1546 if ((conn = http_connect(url, purl, flags)) == NULL)
1551 if (strchr(url->host, ':')) {
1552 snprintf(hbuf, sizeof(hbuf), "[%s]", url->host);
1556 if (url->port != fetch_default_port(url->scheme)) {
1561 snprintf(hbuf + strlen(hbuf),
1562 sizeof(hbuf) - strlen(hbuf), ":%d", url->port);
1567 fetch_info("requesting %s://%s%s",
1568 url->scheme, host, url->doc);
1570 http_cmd(conn, "%s %s://%s%s HTTP/1.1",
1571 op, url->scheme, host, url->doc);
1573 http_cmd(conn, "%s %s HTTP/1.1",
1577 if (ims && url->ims_time) {
1578 timestruct = gmtime((time_t *)&url->ims_time);
1579 (void)strftime(timebuf, 80, "%a, %d %b %Y %T GMT",
1582 fetch_info("If-Modified-Since: %s", timebuf);
1583 http_cmd(conn, "If-Modified-Since: %s", timebuf);
1586 http_cmd(conn, "Host: %s", host);
1589 * Proxy authorization: we only send auth after we received
1590 * a 407 error. We do not first try basic anyway (changed
1591 * when support was added for digest-auth)
1593 if (purl && proxy_challenges.valid) {
1594 http_auth_params_t aparams;
1595 init_http_auth_params(&aparams);
1596 if (*purl->user || *purl->pwd) {
1597 aparams.user = purl->user ?
1598 strdup(purl->user) : strdup("");
1599 aparams.password = purl->pwd?
1600 strdup(purl->pwd) : strdup("");
1601 } else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL &&
1603 if (http_authfromenv(p, &aparams) < 0) {
1604 http_seterr(HTTP_NEED_PROXY_AUTH);
1608 http_authorize(conn, "Proxy-Authorization",
1609 &proxy_challenges, &aparams, url);
1610 clean_http_auth_params(&aparams);
1614 * Server authorization: we never send "a priori"
1615 * Basic auth, which used to be done if user/pass were
1616 * set in the url. This would be weird because we'd send the
1617 * password in the clear even if Digest is finally to be
1618 * used (it would have made more sense for the
1619 * pre-digest version to do this when Basic was specified
1620 * in the environment)
1622 if (server_challenges.valid) {
1623 http_auth_params_t aparams;
1624 init_http_auth_params(&aparams);
1625 if (*url->user || *url->pwd) {
1626 aparams.user = url->user ?
1627 strdup(url->user) : strdup("");
1628 aparams.password = url->pwd ?
1629 strdup(url->pwd) : strdup("");
1630 } else if ((p = getenv("HTTP_AUTH")) != NULL &&
1632 if (http_authfromenv(p, &aparams) < 0) {
1633 http_seterr(HTTP_NEED_AUTH);
1636 } else if (fetchAuthMethod &&
1637 fetchAuthMethod(url) == 0) {
1638 aparams.user = url->user ?
1639 strdup(url->user) : strdup("");
1640 aparams.password = url->pwd ?
1641 strdup(url->pwd) : strdup("");
1643 http_seterr(HTTP_NEED_AUTH);
1646 http_authorize(conn, "Authorization",
1647 &server_challenges, &aparams, url);
1648 clean_http_auth_params(&aparams);
1652 if ((p = getenv("HTTP_REFERER")) != NULL && *p != '\0') {
1653 if (strcasecmp(p, "auto") == 0)
1654 http_cmd(conn, "Referer: %s://%s%s",
1655 url->scheme, host, url->doc);
1657 http_cmd(conn, "Referer: %s", p);
1659 if ((p = getenv("HTTP_USER_AGENT")) != NULL && *p != '\0')
1660 http_cmd(conn, "User-Agent: %s", p);
1662 http_cmd(conn, "User-Agent: %s " _LIBFETCH_VER, getprogname());
1663 if (url->offset > 0)
1664 http_cmd(conn, "Range: bytes=%lld-", (long long)url->offset);
1665 http_cmd(conn, "Connection: close");
1669 * Force the queued request to be dispatched. Normally, one
1670 * would do this with shutdown(2) but squid proxies can be
1671 * configured to disallow such half-closed connections. To
1672 * be compatible with such configurations, fiddle with socket
1673 * options to force the pending data to be written.
1676 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val,
1679 setsockopt(conn->sd, IPPROTO_TCP, TCP_NODELAY, &val,
1683 switch (http_get_reply(conn)) {
1686 case HTTP_NOT_MODIFIED:
1689 case HTTP_MOVED_PERM:
1690 case HTTP_MOVED_TEMP:
1691 case HTTP_SEE_OTHER:
1693 * Not so fine, but we still have to read the
1694 * headers to get the new location.
1697 case HTTP_NEED_AUTH:
1698 if (server_challenges.valid) {
1700 * We already sent out authorization code,
1701 * so there's nothing more we can do.
1703 http_seterr(conn->err);
1706 /* try again, but send the password this time */
1708 fetch_info("server requires authorization");
1710 case HTTP_NEED_PROXY_AUTH:
1711 if (proxy_challenges.valid) {
1713 * We already sent our proxy
1714 * authorization code, so there's
1715 * nothing more we can do. */
1716 http_seterr(conn->err);
1719 /* try again, but send the password this time */
1721 fetch_info("proxy requires authorization");
1723 case HTTP_BAD_RANGE:
1725 * This can happen if we ask for 0 bytes because
1726 * we already have the whole file. Consider this
1727 * a success for now, and check sizes later.
1730 case HTTP_PROTOCOL_ERROR:
1736 http_seterr(conn->err);
1739 /* fall through so we can get the full error message */
1742 /* get headers. http_next_header expects one line readahead */
1743 if (fetch_getln(conn) == -1) {
1748 switch ((h = http_next_header(conn, &headerbuf, &p))) {
1753 http_seterr(HTTP_PROTOCOL_ERROR);
1755 case hdr_content_length:
1756 http_parse_length(p, &clength);
1758 case hdr_content_range:
1759 http_parse_range(p, &offset, &length, &size);
1761 case hdr_last_modified:
1762 http_parse_mtime(p, &mtime);
1765 if (!HTTP_REDIRECT(conn->err))
1770 fetch_info("%d redirect to %s", conn->err, p);
1773 new = fetchMakeURL(url->scheme, url->host, url->port, p,
1774 url->user, url->pwd);
1776 new = fetchParseURL(p);
1778 /* XXX should set an error code */
1779 DEBUG(fprintf(stderr, "failed to parse new URL\n"));
1783 /* Only copy credentials if the host matches */
1784 if (!strcmp(new->host, url->host) && !*new->user && !*new->pwd) {
1785 strcpy(new->user, url->user);
1786 strcpy(new->pwd, url->pwd);
1788 new->offset = url->offset;
1789 new->length = url->length;
1791 case hdr_transfer_encoding:
1793 chunked = (strcasecmp(p, "chunked") == 0);
1795 case hdr_www_authenticate:
1796 if (conn->err != HTTP_NEED_AUTH)
1798 if (http_parse_authenticate(p, &server_challenges) == 0)
1801 case hdr_proxy_authenticate:
1802 if (conn->err != HTTP_NEED_PROXY_AUTH)
1804 if (http_parse_authenticate(p, &proxy_challenges) == 0)
1813 } while (h > hdr_end);
1815 /* we need to provide authentication */
1816 if (conn->err == HTTP_NEED_AUTH ||
1817 conn->err == HTTP_NEED_PROXY_AUTH) {
1819 if ((conn->err == HTTP_NEED_AUTH &&
1820 !server_challenges.valid) ||
1821 (conn->err == HTTP_NEED_PROXY_AUTH &&
1822 !proxy_challenges.valid)) {
1823 /* 401/7 but no www/proxy-authenticate ?? */
1824 DEBUG(fprintf(stderr, "401/7 and no auth header\n"));
1832 /* requested range not satisfiable */
1833 if (conn->err == HTTP_BAD_RANGE) {
1834 if (url->offset == size && url->length == 0) {
1835 /* asked for 0 bytes; fake it */
1836 offset = url->offset;
1838 conn->err = HTTP_OK;
1841 http_seterr(conn->err);
1846 /* we have a hit or an error */
1847 if (conn->err == HTTP_OK
1848 || conn->err == HTTP_NOT_MODIFIED
1849 || conn->err == HTTP_PARTIAL
1850 || HTTP_ERROR(conn->err))
1853 /* all other cases: we got a redirect */
1855 clean_http_auth_challenges(&server_challenges);
1859 DEBUG(fprintf(stderr, "redirect with no new location\n"));
1867 /* we failed, or ran out of retries */
1873 DEBUG(fprintf(stderr, "offset %lld, length %lld,"
1874 " size %lld, clength %lld\n",
1875 (long long)offset, (long long)length,
1876 (long long)size, (long long)clength));
1878 if (conn->err == HTTP_NOT_MODIFIED) {
1879 http_seterr(HTTP_NOT_MODIFIED);
1883 /* check for inconsistencies */
1884 if (clength != -1 && length != -1 && clength != length) {
1885 http_seterr(HTTP_PROTOCOL_ERROR);
1891 length = offset + clength;
1892 if (length != -1 && size != -1 && length != size) {
1893 http_seterr(HTTP_PROTOCOL_ERROR);
1902 us->atime = us->mtime = mtime;
1906 if (URL->offset > 0 && offset > URL->offset) {
1907 http_seterr(HTTP_PROTOCOL_ERROR);
1911 /* report back real offset and size */
1912 URL->offset = offset;
1913 URL->length = clength;
1915 /* wrap it up in a FILE */
1916 if ((f = http_funopen(conn, chunked)) == NULL) {
1926 if (HTTP_ERROR(conn->err)) {
1927 http_print_html(stderr, f);
1931 clean_http_headerbuf(&headerbuf);
1932 clean_http_auth_challenges(&server_challenges);
1933 clean_http_auth_challenges(&proxy_challenges);
1943 clean_http_headerbuf(&headerbuf);
1944 clean_http_auth_challenges(&server_challenges);
1945 clean_http_auth_challenges(&proxy_challenges);
1950 /*****************************************************************************
1955 * Retrieve and stat a file by HTTP
1958 fetchXGetHTTP(struct url *URL, struct url_stat *us, const char *flags)
1960 return (http_request(URL, "GET", us, http_get_proxy(URL, flags), flags));
1964 * Retrieve a file by HTTP
1967 fetchGetHTTP(struct url *URL, const char *flags)
1969 return (fetchXGetHTTP(URL, NULL, flags));
1973 * Store a file by HTTP
1976 fetchPutHTTP(struct url *URL __unused, const char *flags __unused)
1978 warnx("fetchPutHTTP(): not implemented");
1983 * Get an HTTP document's metadata
1986 fetchStatHTTP(struct url *URL, struct url_stat *us, const char *flags)
1990 f = http_request(URL, "HEAD", us, http_get_proxy(URL, flags), flags);
2001 fetchListHTTP(struct url *url __unused, const char *flags __unused)
2003 warnx("fetchListHTTP(): not implemented");