kmalloc: Avoid code duplication.
[dragonfly.git] / lib / libfetch / http.c
blobb48d89506029055bd4607facc8762fdf03e57124
1 /*-
2 * Copyright (c) 2000-2014 Dag-Erling Smørgrav
3 * All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
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.
28 * $FreeBSD: head/lib/libfetch/http.c 261284 2014-01-30 08:37:23Z des $
32 * The following copyright applies to the base64 code:
35 * Copyright 1997 Massachusetts Institute of Technology
37 * Permission to use, copy, modify, and distribute this software and
38 * its documentation for any purpose and without fee is hereby
39 * granted, provided that both the above copyright notice and this
40 * permission notice appear in all copies, that both the above
41 * copyright notice and this permission notice appear in all
42 * supporting documentation, and that the name of M.I.T. not be used
43 * in advertising or publicity pertaining to distribution of the
44 * software without specific, written prior permission. M.I.T. makes
45 * no representations about the suitability of this software for any
46 * purpose. It is provided "as is" without express or implied
47 * warranty.
49 * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''. M.I.T. DISCLAIMS
50 * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE,
51 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
52 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT
53 * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
54 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
55 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
56 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
57 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
58 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
59 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60 * SUCH DAMAGE.
63 #include <sys/param.h>
64 #include <sys/socket.h>
65 #include <sys/time.h>
67 #include <ctype.h>
68 #include <err.h>
69 #include <errno.h>
70 #include <locale.h>
71 #include <netdb.h>
72 #include <stdarg.h>
73 #include <stdio.h>
74 #include <stdlib.h>
75 #include <string.h>
76 #include <time.h>
77 #include <unistd.h>
79 #ifdef WITH_SSL
80 #include <openssl/md5.h>
81 #define MD5Init(c) MD5_Init(c)
82 #define MD5Update(c, data, len) MD5_Update(c, data, len)
83 #define MD5Final(md, c) MD5_Final(md, c)
84 #else
85 #include <md5.h>
86 #endif
88 #include <netinet/in.h>
89 #include <netinet/tcp.h>
91 #include "fetch.h"
92 #include "common.h"
93 #include "httperr.h"
95 /* Maximum number of redirects to follow */
96 #define MAX_REDIRECT 20
98 /* Symbolic names for reply codes we care about */
99 #define HTTP_OK 200
100 #define HTTP_PARTIAL 206
101 #define HTTP_MOVED_PERM 301
102 #define HTTP_MOVED_TEMP 302
103 #define HTTP_SEE_OTHER 303
104 #define HTTP_NOT_MODIFIED 304
105 #define HTTP_USE_PROXY 305
106 #define HTTP_TEMP_REDIRECT 307
107 #define HTTP_PERM_REDIRECT 308
108 #define HTTP_NEED_AUTH 401
109 #define HTTP_NEED_PROXY_AUTH 407
110 #define HTTP_BAD_RANGE 416
111 #define HTTP_PROTOCOL_ERROR 999
113 #define HTTP_REDIRECT(xyz) ((xyz) == HTTP_MOVED_PERM \
114 || (xyz) == HTTP_MOVED_TEMP \
115 || (xyz) == HTTP_TEMP_REDIRECT \
116 || (xyz) == HTTP_USE_PROXY \
117 || (xyz) == HTTP_SEE_OTHER)
119 #define HTTP_ERROR(xyz) ((xyz) > 400 && (xyz) < 599)
122 /*****************************************************************************
123 * I/O functions for decoding chunked streams
126 struct httpio
128 conn_t *conn; /* connection */
129 int chunked; /* chunked mode */
130 char *buf; /* chunk buffer */
131 size_t bufsize; /* size of chunk buffer */
132 ssize_t buflen; /* amount of data currently in buffer */
133 int bufpos; /* current read offset in buffer */
134 int eof; /* end-of-file flag */
135 int error; /* error flag */
136 size_t chunksize; /* remaining size of current chunk */
137 #ifndef NDEBUG
138 size_t total;
139 #endif
143 * Get next chunk header
145 static int
146 http_new_chunk(struct httpio *io)
148 char *p;
150 if (fetch_getln(io->conn) == -1)
151 return (-1);
153 if (io->conn->buflen < 2 || !isxdigit((unsigned char)*io->conn->buf))
154 return (-1);
156 for (p = io->conn->buf; *p && !isspace((unsigned char)*p); ++p) {
157 if (*p == ';')
158 break;
159 if (!isxdigit((unsigned char)*p))
160 return (-1);
161 if (isdigit((unsigned char)*p)) {
162 io->chunksize = io->chunksize * 16 +
163 *p - '0';
164 } else {
165 io->chunksize = io->chunksize * 16 +
166 10 + tolower((unsigned char)*p) - 'a';
170 #ifndef NDEBUG
171 if (fetchDebug) {
172 io->total += io->chunksize;
173 if (io->chunksize == 0)
174 fprintf(stderr, "%s(): end of last chunk\n", __func__);
175 else
176 fprintf(stderr, "%s(): new chunk: %lu (%lu)\n",
177 __func__, (unsigned long)io->chunksize,
178 (unsigned long)io->total);
180 #endif
182 return (io->chunksize);
186 * Grow the input buffer to at least len bytes
188 static inline int
189 http_growbuf(struct httpio *io, size_t len)
191 char *tmp;
193 if (io->bufsize >= len)
194 return (0);
196 if ((tmp = realloc(io->buf, len)) == NULL)
197 return (-1);
198 io->buf = tmp;
199 io->bufsize = len;
200 return (0);
204 * Fill the input buffer, do chunk decoding on the fly
206 static ssize_t
207 http_fillbuf(struct httpio *io, size_t len)
209 ssize_t nbytes;
210 char ch;
212 if (io->error)
213 return (-1);
214 if (io->eof)
215 return (0);
217 if (io->chunked == 0) {
218 if (http_growbuf(io, len) == -1)
219 return (-1);
220 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) {
221 io->error = errno;
222 return (-1);
224 io->buflen = nbytes;
225 io->bufpos = 0;
226 return (io->buflen);
229 if (io->chunksize == 0) {
230 switch (http_new_chunk(io)) {
231 case -1:
232 io->error = EPROTO;
233 return (-1);
234 case 0:
235 io->eof = 1;
236 return (0);
240 if (len > io->chunksize)
241 len = io->chunksize;
242 if (http_growbuf(io, len) == -1)
243 return (-1);
244 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) {
245 io->error = errno;
246 return (-1);
248 io->buflen = nbytes;
249 io->chunksize -= io->buflen;
251 if (io->chunksize == 0) {
252 if (fetch_read(io->conn, &ch, 1) != 1 || ch != '\r' ||
253 fetch_read(io->conn, &ch, 1) != 1 || ch != '\n')
254 return (-1);
257 io->bufpos = 0;
259 return (io->buflen);
263 * Read function
265 static int
266 http_readfn(void *v, char *buf, int len)
268 struct httpio *io = (struct httpio *)v;
269 int rlen;
271 if (io->error)
272 return (-1);
273 if (io->eof)
274 return (0);
276 /* empty buffer */
277 if (!io->buf || io->bufpos == io->buflen) {
278 if ((rlen = http_fillbuf(io, len)) < 0) {
279 if ((errno = io->error) == EINTR)
280 io->error = 0;
281 return (-1);
282 } else if (rlen == 0) {
283 return (0);
287 rlen = io->buflen - io->bufpos;
288 if (len < rlen)
289 rlen = len;
290 memcpy(buf, io->buf + io->bufpos, rlen);
291 io->bufpos += rlen;
292 return (rlen);
296 * Write function
298 static int
299 http_writefn(void *v, const char *buf, int len)
301 struct httpio *io = (struct httpio *)v;
303 return (fetch_write(io->conn, buf, len));
307 * Close function
309 static int
310 http_closefn(void *v)
312 struct httpio *io = (struct httpio *)v;
313 int r;
315 r = fetch_close(io->conn);
316 if (io->buf)
317 free(io->buf);
318 free(io);
319 return (r);
323 * Wrap a file descriptor up
325 static FILE *
326 http_funopen(conn_t *conn, int chunked)
328 struct httpio *io;
329 FILE *f;
331 if ((io = calloc(1, sizeof(*io))) == NULL) {
332 fetch_syserr();
333 return (NULL);
335 io->conn = conn;
336 io->chunked = chunked;
337 f = funopen(io, http_readfn, http_writefn, NULL, http_closefn);
338 if (f == NULL) {
339 fetch_syserr();
340 free(io);
341 return (NULL);
343 return (f);
347 /*****************************************************************************
348 * Helper functions for talking to the server and parsing its replies
351 /* Header types */
352 typedef enum {
353 hdr_syserror = -2,
354 hdr_error = -1,
355 hdr_end = 0,
356 hdr_unknown = 1,
357 hdr_content_length,
358 hdr_content_range,
359 hdr_last_modified,
360 hdr_location,
361 hdr_transfer_encoding,
362 hdr_www_authenticate,
363 hdr_proxy_authenticate,
364 } hdr_t;
366 /* Names of interesting headers */
367 static struct {
368 hdr_t num;
369 const char *name;
370 } hdr_names[] = {
371 { hdr_content_length, "Content-Length" },
372 { hdr_content_range, "Content-Range" },
373 { hdr_last_modified, "Last-Modified" },
374 { hdr_location, "Location" },
375 { hdr_transfer_encoding, "Transfer-Encoding" },
376 { hdr_www_authenticate, "WWW-Authenticate" },
377 { hdr_proxy_authenticate, "Proxy-Authenticate" },
378 { hdr_unknown, NULL },
382 * Send a formatted line; optionally echo to terminal
384 static int
385 http_cmd(conn_t *conn, const char *fmt, ...)
387 va_list ap;
388 size_t len;
389 char *msg;
390 int r;
392 va_start(ap, fmt);
393 len = vasprintf(&msg, fmt, ap);
394 va_end(ap);
396 if (msg == NULL) {
397 errno = ENOMEM;
398 fetch_syserr();
399 return (-1);
402 r = fetch_putln(conn, msg, len);
403 free(msg);
405 if (r == -1) {
406 fetch_syserr();
407 return (-1);
410 return (0);
414 * Get and parse status line
416 static int
417 http_get_reply(conn_t *conn)
419 char *p;
421 if (fetch_getln(conn) == -1)
422 return (-1);
424 * A valid status line looks like "HTTP/m.n xyz reason" where m
425 * and n are the major and minor protocol version numbers and xyz
426 * is the reply code.
427 * Unfortunately, there are servers out there (NCSA 1.5.1, to name
428 * just one) that do not send a version number, so we can't rely
429 * on finding one, but if we do, insist on it being 1.0 or 1.1.
430 * We don't care about the reason phrase.
432 if (strncmp(conn->buf, "HTTP", 4) != 0)
433 return (HTTP_PROTOCOL_ERROR);
434 p = conn->buf + 4;
435 if (*p == '/') {
436 if (p[1] != '1' || p[2] != '.' || (p[3] != '0' && p[3] != '1'))
437 return (HTTP_PROTOCOL_ERROR);
438 p += 4;
440 if (*p != ' ' ||
441 !isdigit((unsigned char)p[1]) ||
442 !isdigit((unsigned char)p[2]) ||
443 !isdigit((unsigned char)p[3]))
444 return (HTTP_PROTOCOL_ERROR);
446 conn->err = (p[1] - '0') * 100 + (p[2] - '0') * 10 + (p[3] - '0');
447 return (conn->err);
451 * Check a header; if the type matches the given string, return a pointer
452 * to the beginning of the value.
454 static const char *
455 http_match(const char *str, const char *hdr)
457 while (*str && *hdr &&
458 tolower((unsigned char)*str++) == tolower((unsigned char)*hdr++))
459 /* nothing */;
460 if (*str || *hdr != ':')
461 return (NULL);
462 while (*hdr && isspace((unsigned char)*++hdr))
463 /* nothing */;
464 return (hdr);
469 * Get the next header and return the appropriate symbolic code. We
470 * need to read one line ahead for checking for a continuation line
471 * belonging to the current header (continuation lines start with
472 * white space).
474 * We get called with a fresh line already in the conn buffer, either
475 * from the previous http_next_header() invocation, or, the first
476 * time, from a fetch_getln() performed by our caller.
478 * This stops when we encounter an empty line (we dont read beyond the header
479 * area).
481 * Note that the "headerbuf" is just a place to return the result. Its
482 * contents are not used for the next call. This means that no cleanup
483 * is needed when ie doing another connection, just call the cleanup when
484 * fully done to deallocate memory.
487 /* Limit the max number of continuation lines to some reasonable value */
488 #define HTTP_MAX_CONT_LINES 10
490 /* Place into which to build a header from one or several lines */
491 typedef struct {
492 char *buf; /* buffer */
493 size_t bufsize; /* buffer size */
494 size_t buflen; /* length of buffer contents */
495 } http_headerbuf_t;
497 static void
498 init_http_headerbuf(http_headerbuf_t *buf)
500 buf->buf = NULL;
501 buf->bufsize = 0;
502 buf->buflen = 0;
505 static void
506 clean_http_headerbuf(http_headerbuf_t *buf)
508 if (buf->buf)
509 free(buf->buf);
510 init_http_headerbuf(buf);
513 /* Remove whitespace at the end of the buffer */
514 static void
515 http_conn_trimright(conn_t *conn)
517 while (conn->buflen &&
518 isspace((unsigned char)conn->buf[conn->buflen - 1]))
519 conn->buflen--;
520 conn->buf[conn->buflen] = '\0';
523 static hdr_t
524 http_next_header(conn_t *conn, http_headerbuf_t *hbuf, const char **p)
526 unsigned int i, len;
529 * Have to do the stripping here because of the first line. So
530 * it's done twice for the subsequent lines. No big deal
532 http_conn_trimright(conn);
533 if (conn->buflen == 0)
534 return (hdr_end);
536 /* Copy the line to the headerbuf */
537 if (hbuf->bufsize < conn->buflen + 1) {
538 if ((hbuf->buf = realloc(hbuf->buf, conn->buflen + 1)) == NULL)
539 return (hdr_syserror);
540 hbuf->bufsize = conn->buflen + 1;
542 strcpy(hbuf->buf, conn->buf);
543 hbuf->buflen = conn->buflen;
546 * Fetch possible continuation lines. Stop at 1st non-continuation
547 * and leave it in the conn buffer
549 for (i = 0; i < HTTP_MAX_CONT_LINES; i++) {
550 if (fetch_getln(conn) == -1)
551 return (hdr_syserror);
554 * Note: we carry on the idea from the previous version
555 * that a pure whitespace line is equivalent to an empty
556 * one (so it's not continuation and will be handled when
557 * we are called next)
559 http_conn_trimright(conn);
560 if (conn->buf[0] != ' ' && conn->buf[0] != "\t"[0])
561 break;
563 /* Got a continuation line. Concatenate to previous */
564 len = hbuf->buflen + conn->buflen;
565 if (hbuf->bufsize < len + 1) {
566 len *= 2;
567 if ((hbuf->buf = realloc(hbuf->buf, len + 1)) == NULL)
568 return (hdr_syserror);
569 hbuf->bufsize = len + 1;
571 strcpy(hbuf->buf + hbuf->buflen, conn->buf);
572 hbuf->buflen += conn->buflen;
576 * We could check for malformed headers but we don't really care.
577 * A valid header starts with a token immediately followed by a
578 * colon; a token is any sequence of non-control, non-whitespace
579 * characters except "()<>@,;:\\\"{}".
581 for (i = 0; hdr_names[i].num != hdr_unknown; i++)
582 if ((*p = http_match(hdr_names[i].name, hbuf->buf)) != NULL)
583 return (hdr_names[i].num);
585 return (hdr_unknown);
588 /**************************
589 * [Proxy-]Authenticate header parsing
593 * Read doublequote-delimited string into output buffer obuf (allocated
594 * by caller, whose responsibility it is to ensure that it's big enough)
595 * cp points to the first char after the initial '"'
596 * Handles \ quoting
597 * Returns pointer to the first char after the terminating double quote, or
598 * NULL for error.
600 static const char *
601 http_parse_headerstring(const char *cp, char *obuf)
603 for (;;) {
604 switch (*cp) {
605 case 0: /* Unterminated string */
606 *obuf = 0;
607 return (NULL);
608 case '"': /* Ending quote */
609 *obuf = 0;
610 return (++cp);
611 case '\\':
612 if (*++cp == 0) {
613 *obuf = 0;
614 return (NULL);
616 /* FALLTHROUGH */
617 default:
618 *obuf++ = *cp++;
623 /* Http auth challenge schemes */
624 typedef enum {HTTPAS_UNKNOWN, HTTPAS_BASIC,HTTPAS_DIGEST} http_auth_schemes_t;
626 /* Data holder for a Basic or Digest challenge. */
627 typedef struct {
628 http_auth_schemes_t scheme;
629 char *realm;
630 char *qop;
631 char *nonce;
632 char *opaque;
633 char *algo;
634 int stale;
635 int nc; /* Nonce count */
636 } http_auth_challenge_t;
638 static void
639 init_http_auth_challenge(http_auth_challenge_t *b)
641 b->scheme = HTTPAS_UNKNOWN;
642 b->realm = b->qop = b->nonce = b->opaque = b->algo = NULL;
643 b->stale = b->nc = 0;
646 static void
647 clean_http_auth_challenge(http_auth_challenge_t *b)
649 if (b->realm)
650 free(b->realm);
651 if (b->qop)
652 free(b->qop);
653 if (b->nonce)
654 free(b->nonce);
655 if (b->opaque)
656 free(b->opaque);
657 if (b->algo)
658 free(b->algo);
659 init_http_auth_challenge(b);
662 /* Data holder for an array of challenges offered in an http response. */
663 #define MAX_CHALLENGES 10
664 typedef struct {
665 http_auth_challenge_t *challenges[MAX_CHALLENGES];
666 int count; /* Number of parsed challenges in the array */
667 int valid; /* We did parse an authenticate header */
668 } http_auth_challenges_t;
670 static void
671 init_http_auth_challenges(http_auth_challenges_t *cs)
673 int i;
674 for (i = 0; i < MAX_CHALLENGES; i++)
675 cs->challenges[i] = NULL;
676 cs->count = cs->valid = 0;
679 static void
680 clean_http_auth_challenges(http_auth_challenges_t *cs)
682 int i;
683 /* We rely on non-zero pointers being allocated, not on the count */
684 for (i = 0; i < MAX_CHALLENGES; i++) {
685 if (cs->challenges[i] != NULL) {
686 clean_http_auth_challenge(cs->challenges[i]);
687 free(cs->challenges[i]);
690 init_http_auth_challenges(cs);
694 * Enumeration for lexical elements. Separators will be returned as their own
695 * ascii value
697 typedef enum {HTTPHL_WORD=256, HTTPHL_STRING=257, HTTPHL_END=258,
698 HTTPHL_ERROR = 259} http_header_lex_t;
701 * Determine what kind of token comes next and return possible value
702 * in buf, which is supposed to have been allocated big enough by
703 * caller. Advance input pointer and return element type.
705 static int
706 http_header_lex(const char **cpp, char *buf)
708 size_t l;
709 /* Eat initial whitespace */
710 *cpp += strspn(*cpp, " \t");
711 if (**cpp == 0)
712 return (HTTPHL_END);
714 /* Separator ? */
715 if (**cpp == ',' || **cpp == '=')
716 return (*((*cpp)++));
718 /* String ? */
719 if (**cpp == '"') {
720 *cpp = http_parse_headerstring(++*cpp, buf);
721 if (*cpp == NULL)
722 return (HTTPHL_ERROR);
723 return (HTTPHL_STRING);
726 /* Read other token, until separator or whitespace */
727 l = strcspn(*cpp, " \t,=");
728 memcpy(buf, *cpp, l);
729 buf[l] = 0;
730 *cpp += l;
731 return (HTTPHL_WORD);
735 * Read challenges from http xxx-authenticate header and accumulate them
736 * in the challenges list structure.
738 * Headers with multiple challenges are specified by rfc2617, but
739 * servers (ie: squid) often send them in separate headers instead,
740 * which in turn is forbidden by the http spec (multiple headers with
741 * the same name are only allowed for pure comma-separated lists, see
742 * rfc2616 sec 4.2).
744 * We support both approaches anyway
746 static int
747 http_parse_authenticate(const char *cp, http_auth_challenges_t *cs)
749 int ret = -1;
750 http_header_lex_t lex;
751 char *key = malloc(strlen(cp) + 1);
752 char *value = malloc(strlen(cp) + 1);
753 char *buf = malloc(strlen(cp) + 1);
755 if (key == NULL || value == NULL || buf == NULL) {
756 fetch_syserr();
757 goto out;
760 /* In any case we've seen the header and we set the valid bit */
761 cs->valid = 1;
763 /* Need word first */
764 lex = http_header_lex(&cp, key);
765 if (lex != HTTPHL_WORD)
766 goto out;
768 /* Loop on challenges */
769 for (; cs->count < MAX_CHALLENGES; cs->count++) {
770 cs->challenges[cs->count] =
771 malloc(sizeof(http_auth_challenge_t));
772 if (cs->challenges[cs->count] == NULL) {
773 fetch_syserr();
774 goto out;
776 init_http_auth_challenge(cs->challenges[cs->count]);
777 if (!strcasecmp(key, "basic")) {
778 cs->challenges[cs->count]->scheme = HTTPAS_BASIC;
779 } else if (!strcasecmp(key, "digest")) {
780 cs->challenges[cs->count]->scheme = HTTPAS_DIGEST;
781 } else {
782 cs->challenges[cs->count]->scheme = HTTPAS_UNKNOWN;
784 * Continue parsing as basic or digest may
785 * follow, and the syntax is the same for
786 * all. We'll just ignore this one when
787 * looking at the list
791 /* Loop on attributes */
792 for (;;) {
793 /* Key */
794 lex = http_header_lex(&cp, key);
795 if (lex != HTTPHL_WORD)
796 goto out;
798 /* Equal sign */
799 lex = http_header_lex(&cp, buf);
800 if (lex != '=')
801 goto out;
803 /* Value */
804 lex = http_header_lex(&cp, value);
805 if (lex != HTTPHL_WORD && lex != HTTPHL_STRING)
806 goto out;
808 if (!strcasecmp(key, "realm"))
809 cs->challenges[cs->count]->realm =
810 strdup(value);
811 else if (!strcasecmp(key, "qop"))
812 cs->challenges[cs->count]->qop =
813 strdup(value);
814 else if (!strcasecmp(key, "nonce"))
815 cs->challenges[cs->count]->nonce =
816 strdup(value);
817 else if (!strcasecmp(key, "opaque"))
818 cs->challenges[cs->count]->opaque =
819 strdup(value);
820 else if (!strcasecmp(key, "algorithm"))
821 cs->challenges[cs->count]->algo =
822 strdup(value);
823 else if (!strcasecmp(key, "stale"))
824 cs->challenges[cs->count]->stale =
825 strcasecmp(value, "no");
826 /* Else ignore unknown attributes */
828 /* Comma or Next challenge or End */
829 lex = http_header_lex(&cp, key);
831 * If we get a word here, this is the beginning of the
832 * next challenge. Break the attributes loop
834 if (lex == HTTPHL_WORD)
835 break;
837 if (lex == HTTPHL_END) {
838 /* End while looking for ',' is normal exit */
839 cs->count++;
840 ret = 0;
841 goto out;
843 /* Anything else is an error */
844 if (lex != ',')
845 goto out;
847 } /* End attributes loop */
848 } /* End challenge loop */
851 * Challenges max count exceeded. This really can't happen
852 * with normal data, something's fishy -> error
855 out:
856 if (key)
857 free(key);
858 if (value)
859 free(value);
860 if (buf)
861 free(buf);
862 return (ret);
867 * Parse a last-modified header
869 static int
870 http_parse_mtime(const char *p, time_t *mtime)
872 char locale[64], *r;
873 struct tm tm;
875 strncpy(locale, setlocale(LC_TIME, NULL), sizeof(locale));
876 setlocale(LC_TIME, "C");
877 r = strptime(p, "%a, %d %b %Y %H:%M:%S GMT", &tm);
878 /* XXX should add support for date-2 and date-3 */
879 setlocale(LC_TIME, locale);
880 if (r == NULL)
881 return (-1);
882 DEBUG(fprintf(stderr, "last modified: [%04d-%02d-%02d "
883 "%02d:%02d:%02d]\n",
884 tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday,
885 tm.tm_hour, tm.tm_min, tm.tm_sec));
886 *mtime = timegm(&tm);
887 return (0);
891 * Parse a content-length header
893 static int
894 http_parse_length(const char *p, off_t *length)
896 off_t len;
898 for (len = 0; *p && isdigit((unsigned char)*p); ++p)
899 len = len * 10 + (*p - '0');
900 if (*p)
901 return (-1);
902 DEBUG(fprintf(stderr, "content length: [%lld]\n",
903 (long long)len));
904 *length = len;
905 return (0);
909 * Parse a content-range header
911 static int
912 http_parse_range(const char *p, off_t *offset, off_t *length, off_t *size)
914 off_t first, last, len;
916 if (strncasecmp(p, "bytes ", 6) != 0)
917 return (-1);
918 p += 6;
919 if (*p == '*') {
920 first = last = -1;
921 ++p;
922 } else {
923 for (first = 0; *p && isdigit((unsigned char)*p); ++p)
924 first = first * 10 + *p - '0';
925 if (*p != '-')
926 return (-1);
927 for (last = 0, ++p; *p && isdigit((unsigned char)*p); ++p)
928 last = last * 10 + *p - '0';
930 if (first > last || *p != '/')
931 return (-1);
932 for (len = 0, ++p; *p && isdigit((unsigned char)*p); ++p)
933 len = len * 10 + *p - '0';
934 if (*p || len < last - first + 1)
935 return (-1);
936 if (first == -1) {
937 DEBUG(fprintf(stderr, "content range: [*/%lld]\n",
938 (long long)len));
939 *length = 0;
940 } else {
941 DEBUG(fprintf(stderr, "content range: [%lld-%lld/%lld]\n",
942 (long long)first, (long long)last, (long long)len));
943 *length = last - first + 1;
945 *offset = first;
946 *size = len;
947 return (0);
951 /*****************************************************************************
952 * Helper functions for authorization
956 * Base64 encoding
958 static char *
959 http_base64(const char *src)
961 static const char base64[] =
962 "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
963 "abcdefghijklmnopqrstuvwxyz"
964 "0123456789+/";
965 char *str, *dst;
966 size_t l;
967 int t, r;
969 l = strlen(src);
970 if ((str = malloc(((l + 2) / 3) * 4 + 1)) == NULL)
971 return (NULL);
972 dst = str;
973 r = 0;
975 while (l >= 3) {
976 t = (src[0] << 16) | (src[1] << 8) | src[2];
977 dst[0] = base64[(t >> 18) & 0x3f];
978 dst[1] = base64[(t >> 12) & 0x3f];
979 dst[2] = base64[(t >> 6) & 0x3f];
980 dst[3] = base64[(t >> 0) & 0x3f];
981 src += 3; l -= 3;
982 dst += 4; r += 4;
985 switch (l) {
986 case 2:
987 t = (src[0] << 16) | (src[1] << 8);
988 dst[0] = base64[(t >> 18) & 0x3f];
989 dst[1] = base64[(t >> 12) & 0x3f];
990 dst[2] = base64[(t >> 6) & 0x3f];
991 dst[3] = '=';
992 dst += 4;
993 r += 4;
994 break;
995 case 1:
996 t = src[0] << 16;
997 dst[0] = base64[(t >> 18) & 0x3f];
998 dst[1] = base64[(t >> 12) & 0x3f];
999 dst[2] = dst[3] = '=';
1000 dst += 4;
1001 r += 4;
1002 break;
1003 case 0:
1004 break;
1007 *dst = 0;
1008 return (str);
1013 * Extract authorization parameters from environment value.
1014 * The value is like scheme:realm:user:pass
1016 typedef struct {
1017 char *scheme;
1018 char *realm;
1019 char *user;
1020 char *password;
1021 } http_auth_params_t;
1023 static void
1024 init_http_auth_params(http_auth_params_t *s)
1026 s->scheme = s->realm = s->user = s->password = NULL;
1029 static void
1030 clean_http_auth_params(http_auth_params_t *s)
1032 if (s->scheme)
1033 free(s->scheme);
1034 if (s->realm)
1035 free(s->realm);
1036 if (s->user)
1037 free(s->user);
1038 if (s->password)
1039 free(s->password);
1040 init_http_auth_params(s);
1043 static int
1044 http_authfromenv(const char *p, http_auth_params_t *parms)
1046 int ret = -1;
1047 char *v, *ve;
1048 char *str = strdup(p);
1050 if (str == NULL) {
1051 fetch_syserr();
1052 return (-1);
1054 v = str;
1056 if ((ve = strchr(v, ':')) == NULL)
1057 goto out;
1059 *ve = 0;
1060 if ((parms->scheme = strdup(v)) == NULL) {
1061 fetch_syserr();
1062 goto out;
1064 v = ve + 1;
1066 if ((ve = strchr(v, ':')) == NULL)
1067 goto out;
1069 *ve = 0;
1070 if ((parms->realm = strdup(v)) == NULL) {
1071 fetch_syserr();
1072 goto out;
1074 v = ve + 1;
1076 if ((ve = strchr(v, ':')) == NULL)
1077 goto out;
1079 *ve = 0;
1080 if ((parms->user = strdup(v)) == NULL) {
1081 fetch_syserr();
1082 goto out;
1084 v = ve + 1;
1087 if ((parms->password = strdup(v)) == NULL) {
1088 fetch_syserr();
1089 goto out;
1091 ret = 0;
1092 out:
1093 if (ret == -1)
1094 clean_http_auth_params(parms);
1095 if (str)
1096 free(str);
1097 return (ret);
1102 * Digest response: the code to compute the digest is taken from the
1103 * sample implementation in RFC2616
1105 #define IN const
1106 #define OUT
1108 #define HASHLEN 16
1109 typedef char HASH[HASHLEN];
1110 #define HASHHEXLEN 32
1111 typedef char HASHHEX[HASHHEXLEN+1];
1113 static const char *hexchars = "0123456789abcdef";
1114 static void
1115 CvtHex(IN HASH Bin, OUT HASHHEX Hex)
1117 unsigned short i;
1118 unsigned char j;
1120 for (i = 0; i < HASHLEN; i++) {
1121 j = (Bin[i] >> 4) & 0xf;
1122 Hex[i*2] = hexchars[j];
1123 j = Bin[i] & 0xf;
1124 Hex[i*2+1] = hexchars[j];
1126 Hex[HASHHEXLEN] = '\0';
1129 /* calculate H(A1) as per spec */
1130 static void
1131 DigestCalcHA1(
1132 IN char * pszAlg,
1133 IN char * pszUserName,
1134 IN char * pszRealm,
1135 IN char * pszPassword,
1136 IN char * pszNonce,
1137 IN char * pszCNonce,
1138 OUT HASHHEX SessionKey
1141 MD5_CTX Md5Ctx;
1142 HASH HA1;
1144 MD5Init(&Md5Ctx);
1145 MD5Update(&Md5Ctx, pszUserName, strlen(pszUserName));
1146 MD5Update(&Md5Ctx, ":", 1);
1147 MD5Update(&Md5Ctx, pszRealm, strlen(pszRealm));
1148 MD5Update(&Md5Ctx, ":", 1);
1149 MD5Update(&Md5Ctx, pszPassword, strlen(pszPassword));
1150 MD5Final(HA1, &Md5Ctx);
1151 if (strcasecmp(pszAlg, "md5-sess") == 0) {
1153 MD5Init(&Md5Ctx);
1154 MD5Update(&Md5Ctx, HA1, HASHLEN);
1155 MD5Update(&Md5Ctx, ":", 1);
1156 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce));
1157 MD5Update(&Md5Ctx, ":", 1);
1158 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce));
1159 MD5Final(HA1, &Md5Ctx);
1161 CvtHex(HA1, SessionKey);
1164 /* calculate request-digest/response-digest as per HTTP Digest spec */
1165 static void
1166 DigestCalcResponse(
1167 IN HASHHEX HA1, /* H(A1) */
1168 IN char * pszNonce, /* nonce from server */
1169 IN char * pszNonceCount, /* 8 hex digits */
1170 IN char * pszCNonce, /* client nonce */
1171 IN char * pszQop, /* qop-value: "", "auth", "auth-int" */
1172 IN char * pszMethod, /* method from the request */
1173 IN char * pszDigestUri, /* requested URL */
1174 IN HASHHEX HEntity, /* H(entity body) if qop="auth-int" */
1175 OUT HASHHEX Response /* request-digest or response-digest */
1178 /* DEBUG(fprintf(stderr,
1179 "Calc: HA1[%s] Nonce[%s] qop[%s] method[%s] URI[%s]\n",
1180 HA1, pszNonce, pszQop, pszMethod, pszDigestUri));*/
1181 MD5_CTX Md5Ctx;
1182 HASH HA2;
1183 HASH RespHash;
1184 HASHHEX HA2Hex;
1186 // calculate H(A2)
1187 MD5Init(&Md5Ctx);
1188 MD5Update(&Md5Ctx, pszMethod, strlen(pszMethod));
1189 MD5Update(&Md5Ctx, ":", 1);
1190 MD5Update(&Md5Ctx, pszDigestUri, strlen(pszDigestUri));
1191 if (strcasecmp(pszQop, "auth-int") == 0) {
1192 MD5Update(&Md5Ctx, ":", 1);
1193 MD5Update(&Md5Ctx, HEntity, HASHHEXLEN);
1195 MD5Final(HA2, &Md5Ctx);
1196 CvtHex(HA2, HA2Hex);
1198 // calculate response
1199 MD5Init(&Md5Ctx);
1200 MD5Update(&Md5Ctx, HA1, HASHHEXLEN);
1201 MD5Update(&Md5Ctx, ":", 1);
1202 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce));
1203 MD5Update(&Md5Ctx, ":", 1);
1204 if (*pszQop) {
1205 MD5Update(&Md5Ctx, pszNonceCount, strlen(pszNonceCount));
1206 MD5Update(&Md5Ctx, ":", 1);
1207 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce));
1208 MD5Update(&Md5Ctx, ":", 1);
1209 MD5Update(&Md5Ctx, pszQop, strlen(pszQop));
1210 MD5Update(&Md5Ctx, ":", 1);
1212 MD5Update(&Md5Ctx, HA2Hex, HASHHEXLEN);
1213 MD5Final(RespHash, &Md5Ctx);
1214 CvtHex(RespHash, Response);
1218 * Generate/Send a Digest authorization header
1219 * This looks like: [Proxy-]Authorization: credentials
1221 * credentials = "Digest" digest-response
1222 * digest-response = 1#( username | realm | nonce | digest-uri
1223 * | response | [ algorithm ] | [cnonce] |
1224 * [opaque] | [message-qop] |
1225 * [nonce-count] | [auth-param] )
1226 * username = "username" "=" username-value
1227 * username-value = quoted-string
1228 * digest-uri = "uri" "=" digest-uri-value
1229 * digest-uri-value = request-uri ; As specified by HTTP/1.1
1230 * message-qop = "qop" "=" qop-value
1231 * cnonce = "cnonce" "=" cnonce-value
1232 * cnonce-value = nonce-value
1233 * nonce-count = "nc" "=" nc-value
1234 * nc-value = 8LHEX
1235 * response = "response" "=" request-digest
1236 * request-digest = <"> 32LHEX <">
1238 static int
1239 http_digest_auth(conn_t *conn, const char *hdr, http_auth_challenge_t *c,
1240 http_auth_params_t *parms, struct url *url)
1242 int r;
1243 char noncecount[10];
1244 char cnonce[40];
1245 char *options = NULL;
1247 if (!c->realm || !c->nonce) {
1248 DEBUG(fprintf(stderr, "realm/nonce not set in challenge\n"));
1249 return(-1);
1251 if (!c->algo)
1252 c->algo = strdup("");
1254 if (asprintf(&options, "%s%s%s%s",
1255 *c->algo? ",algorithm=" : "", c->algo,
1256 c->opaque? ",opaque=" : "", c->opaque?c->opaque:"")== -1)
1257 return (-1);
1259 if (!c->qop) {
1260 c->qop = strdup("");
1261 *noncecount = 0;
1262 *cnonce = 0;
1263 } else {
1264 c->nc++;
1265 sprintf(noncecount, "%08x", c->nc);
1266 /* We don't try very hard with the cnonce ... */
1267 sprintf(cnonce, "%x%lx", getpid(), (unsigned long)time(0));
1270 HASHHEX HA1;
1271 DigestCalcHA1(c->algo, parms->user, c->realm,
1272 parms->password, c->nonce, cnonce, HA1);
1273 DEBUG(fprintf(stderr, "HA1: [%s]\n", HA1));
1274 HASHHEX digest;
1275 DigestCalcResponse(HA1, c->nonce, noncecount, cnonce, c->qop,
1276 "GET", url->doc, "", digest);
1278 if (c->qop[0]) {
1279 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\","
1280 "nonce=\"%s\",uri=\"%s\",response=\"%s\","
1281 "qop=\"auth\", cnonce=\"%s\", nc=%s%s",
1282 hdr, parms->user, c->realm,
1283 c->nonce, url->doc, digest,
1284 cnonce, noncecount, options);
1285 } else {
1286 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\","
1287 "nonce=\"%s\",uri=\"%s\",response=\"%s\"%s",
1288 hdr, parms->user, c->realm,
1289 c->nonce, url->doc, digest, options);
1291 if (options)
1292 free(options);
1293 return (r);
1297 * Encode username and password
1299 static int
1300 http_basic_auth(conn_t *conn, const char *hdr, const char *usr, const char *pwd)
1302 char *upw, *auth;
1303 int r;
1305 DEBUG(fprintf(stderr, "basic: usr: [%s]\n", usr));
1306 DEBUG(fprintf(stderr, "basic: pwd: [%s]\n", pwd));
1307 if (asprintf(&upw, "%s:%s", usr, pwd) == -1)
1308 return (-1);
1309 auth = http_base64(upw);
1310 free(upw);
1311 if (auth == NULL)
1312 return (-1);
1313 r = http_cmd(conn, "%s: Basic %s", hdr, auth);
1314 free(auth);
1315 return (r);
1319 * Chose the challenge to answer and call the appropriate routine to
1320 * produce the header.
1322 static int
1323 http_authorize(conn_t *conn, const char *hdr, http_auth_challenges_t *cs,
1324 http_auth_params_t *parms, struct url *url)
1326 http_auth_challenge_t *digest = NULL;
1327 int i;
1329 /* If user or pass are null we're not happy */
1330 if (!parms->user || !parms->password) {
1331 DEBUG(fprintf(stderr, "NULL usr or pass\n"));
1332 return (-1);
1335 /* Look for a Digest */
1336 for (i = 0; i < cs->count; i++) {
1337 if (cs->challenges[i]->scheme == HTTPAS_DIGEST)
1338 digest = cs->challenges[i];
1341 /* Error if "Digest" was specified and there is no Digest challenge */
1342 if (!digest && (parms->scheme &&
1343 !strcasecmp(parms->scheme, "digest"))) {
1344 DEBUG(fprintf(stderr,
1345 "Digest auth in env, not supported by peer\n"));
1346 return (-1);
1349 * If "basic" was specified in the environment, or there is no Digest
1350 * challenge, do the basic thing. Don't need a challenge for this,
1351 * so no need to check basic!=NULL
1353 if (!digest || (parms->scheme && !strcasecmp(parms->scheme,"basic")))
1354 return (http_basic_auth(conn,hdr,parms->user,parms->password));
1356 /* Else, prefer digest. We just checked that it's not NULL */
1357 return (http_digest_auth(conn, hdr, digest, parms, url));
1360 /*****************************************************************************
1361 * Helper functions for connecting to a server or proxy
1365 * Connect to the correct HTTP server or proxy.
1367 static conn_t *
1368 http_connect(struct url *URL, struct url *purl, const char *flags)
1370 struct url *curl;
1371 conn_t *conn;
1372 int verbose;
1373 int af, val;
1375 #ifdef INET6
1376 af = AF_UNSPEC;
1377 #else
1378 af = AF_INET;
1379 #endif
1381 verbose = CHECK_FLAG('v');
1382 if (CHECK_FLAG('4'))
1383 af = AF_INET;
1384 #ifdef INET6
1385 else if (CHECK_FLAG('6'))
1386 af = AF_INET6;
1387 #endif
1389 curl = (purl != NULL) ? purl : URL;
1391 if ((conn = fetch_connect(curl->host, curl->port, af, verbose)) == NULL)
1392 /* fetch_connect() has already set an error code */
1393 return (NULL);
1394 if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 && purl) {
1395 http_cmd(conn, "CONNECT %s:%d HTTP/1.1",
1396 URL->host, URL->port);
1397 http_cmd(conn, "Host: %s:%d",
1398 URL->host, URL->port);
1399 http_cmd(conn, "");
1400 if (http_get_reply(conn) != HTTP_OK) {
1401 fetch_close(conn);
1402 return (NULL);
1404 http_get_reply(conn);
1406 if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 &&
1407 fetch_ssl(conn, URL, verbose) == -1) {
1408 fetch_close(conn);
1409 /* grrr */
1410 errno = EAUTH;
1411 fetch_syserr();
1412 return (NULL);
1415 val = 1;
1416 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, sizeof(val));
1418 return (conn);
1421 static struct url *
1422 http_get_proxy(struct url * url, const char *flags)
1424 struct url *purl;
1425 char *p;
1427 if (flags != NULL && strchr(flags, 'd') != NULL)
1428 return (NULL);
1429 if (fetch_no_proxy_match(url->host))
1430 return (NULL);
1431 if (((p = getenv("HTTP_PROXY")) || (p = getenv("http_proxy"))) &&
1432 *p && (purl = fetchParseURL(p))) {
1433 if (!*purl->scheme)
1434 strcpy(purl->scheme, SCHEME_HTTP);
1435 if (!purl->port)
1436 purl->port = fetch_default_proxy_port(purl->scheme);
1437 if (strcasecmp(purl->scheme, SCHEME_HTTP) == 0)
1438 return (purl);
1439 fetchFreeURL(purl);
1441 return (NULL);
1444 static void
1445 http_print_html(FILE *out, FILE *in)
1447 size_t len;
1448 char *line, *p, *q;
1449 int comment, tag;
1451 comment = tag = 0;
1452 while ((line = fgetln(in, &len)) != NULL) {
1453 while (len && isspace((unsigned char)line[len - 1]))
1454 --len;
1455 for (p = q = line; q < line + len; ++q) {
1456 if (comment && *q == '-') {
1457 if (q + 2 < line + len &&
1458 strcmp(q, "-->") == 0) {
1459 tag = comment = 0;
1460 q += 2;
1462 } else if (tag && !comment && *q == '>') {
1463 p = q + 1;
1464 tag = 0;
1465 } else if (!tag && *q == '<') {
1466 if (q > p)
1467 fwrite(p, q - p, 1, out);
1468 tag = 1;
1469 if (q + 3 < line + len &&
1470 strcmp(q, "<!--") == 0) {
1471 comment = 1;
1472 q += 3;
1476 if (!tag && q > p)
1477 fwrite(p, q - p, 1, out);
1478 fputc('\n', out);
1483 /*****************************************************************************
1484 * Core
1488 * Send a request and process the reply
1490 * XXX This function is way too long, the do..while loop should be split
1491 * XXX off into a separate function.
1493 FILE *
1494 http_request(struct url *URL, const char *op, struct url_stat *us,
1495 struct url *purl, const char *flags)
1497 char timebuf[80];
1498 char hbuf[MAXHOSTNAMELEN + 7], *host;
1499 conn_t *conn;
1500 struct url *url, *new;
1501 int chunked, direct, ims, noredirect, verbose;
1502 int e, i, n, val;
1503 off_t offset, clength, length, size;
1504 time_t mtime;
1505 const char *p;
1506 FILE *f;
1507 hdr_t h;
1508 struct tm *timestruct;
1509 http_headerbuf_t headerbuf;
1510 http_auth_challenges_t server_challenges;
1511 http_auth_challenges_t proxy_challenges;
1513 /* The following calls don't allocate anything */
1514 init_http_headerbuf(&headerbuf);
1515 init_http_auth_challenges(&server_challenges);
1516 init_http_auth_challenges(&proxy_challenges);
1518 direct = CHECK_FLAG('d');
1519 noredirect = CHECK_FLAG('A');
1520 verbose = CHECK_FLAG('v');
1521 ims = CHECK_FLAG('i');
1523 if (direct && purl) {
1524 fetchFreeURL(purl);
1525 purl = NULL;
1528 /* try the provided URL first */
1529 url = URL;
1531 n = MAX_REDIRECT;
1532 i = 0;
1534 e = HTTP_PROTOCOL_ERROR;
1535 do {
1536 new = NULL;
1537 chunked = 0;
1538 offset = 0;
1539 clength = -1;
1540 length = -1;
1541 size = -1;
1542 mtime = 0;
1544 /* check port */
1545 if (!url->port)
1546 url->port = fetch_default_port(url->scheme);
1548 /* were we redirected to an FTP URL? */
1549 if (purl == NULL && strcmp(url->scheme, SCHEME_FTP) == 0) {
1550 if (strcmp(op, "GET") == 0)
1551 return (ftp_request(url, "RETR", us, purl, flags));
1552 else if (strcmp(op, "HEAD") == 0)
1553 return (ftp_request(url, "STAT", us, purl, flags));
1556 /* connect to server or proxy */
1557 if ((conn = http_connect(url, purl, flags)) == NULL)
1558 goto ouch;
1560 host = url->host;
1561 #ifdef INET6
1562 if (strchr(url->host, ':')) {
1563 snprintf(hbuf, sizeof(hbuf), "[%s]", url->host);
1564 host = hbuf;
1566 #endif
1567 if (url->port != fetch_default_port(url->scheme)) {
1568 if (host != hbuf) {
1569 strcpy(hbuf, host);
1570 host = hbuf;
1572 snprintf(hbuf + strlen(hbuf),
1573 sizeof(hbuf) - strlen(hbuf), ":%d", url->port);
1576 /* send request */
1577 if (verbose)
1578 fetch_info("requesting %s://%s%s",
1579 url->scheme, host, url->doc);
1580 if (purl && strcasecmp(URL->scheme, SCHEME_HTTPS) != 0) {
1581 http_cmd(conn, "%s %s://%s%s HTTP/1.1",
1582 op, url->scheme, host, url->doc);
1583 } else {
1584 http_cmd(conn, "%s %s HTTP/1.1",
1585 op, url->doc);
1588 if (ims && url->ims_time) {
1589 timestruct = gmtime((time_t *)&url->ims_time);
1590 (void)strftime(timebuf, 80, "%a, %d %b %Y %T GMT",
1591 timestruct);
1592 if (verbose)
1593 fetch_info("If-Modified-Since: %s", timebuf);
1594 http_cmd(conn, "If-Modified-Since: %s", timebuf);
1596 /* virtual host */
1597 http_cmd(conn, "Host: %s", host);
1600 * Proxy authorization: we only send auth after we received
1601 * a 407 error. We do not first try basic anyway (changed
1602 * when support was added for digest-auth)
1604 if (purl && proxy_challenges.valid) {
1605 http_auth_params_t aparams;
1606 init_http_auth_params(&aparams);
1607 if (*purl->user || *purl->pwd) {
1608 aparams.user = purl->user ?
1609 strdup(purl->user) : strdup("");
1610 aparams.password = purl->pwd?
1611 strdup(purl->pwd) : strdup("");
1612 } else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL &&
1613 *p != '\0') {
1614 if (http_authfromenv(p, &aparams) < 0) {
1615 http_seterr(HTTP_NEED_PROXY_AUTH);
1616 goto ouch;
1619 http_authorize(conn, "Proxy-Authorization",
1620 &proxy_challenges, &aparams, url);
1621 clean_http_auth_params(&aparams);
1625 * Server authorization: we never send "a priori"
1626 * Basic auth, which used to be done if user/pass were
1627 * set in the url. This would be weird because we'd send the
1628 * password in the clear even if Digest is finally to be
1629 * used (it would have made more sense for the
1630 * pre-digest version to do this when Basic was specified
1631 * in the environment)
1633 if (server_challenges.valid) {
1634 http_auth_params_t aparams;
1635 init_http_auth_params(&aparams);
1636 if (*url->user || *url->pwd) {
1637 aparams.user = url->user ?
1638 strdup(url->user) : strdup("");
1639 aparams.password = url->pwd ?
1640 strdup(url->pwd) : strdup("");
1641 } else if ((p = getenv("HTTP_AUTH")) != NULL &&
1642 *p != '\0') {
1643 if (http_authfromenv(p, &aparams) < 0) {
1644 http_seterr(HTTP_NEED_AUTH);
1645 goto ouch;
1647 } else if (fetchAuthMethod &&
1648 fetchAuthMethod(url) == 0) {
1649 aparams.user = url->user ?
1650 strdup(url->user) : strdup("");
1651 aparams.password = url->pwd ?
1652 strdup(url->pwd) : strdup("");
1653 } else {
1654 http_seterr(HTTP_NEED_AUTH);
1655 goto ouch;
1657 http_authorize(conn, "Authorization",
1658 &server_challenges, &aparams, url);
1659 clean_http_auth_params(&aparams);
1662 /* other headers */
1663 if ((p = getenv("HTTP_ACCEPT")) != NULL) {
1664 if (*p != '\0')
1665 http_cmd(conn, "Accept: %s", p);
1666 } else {
1667 http_cmd(conn, "Accept: */*");
1669 if ((p = getenv("HTTP_REFERER")) != NULL && *p != '\0') {
1670 if (strcasecmp(p, "auto") == 0)
1671 http_cmd(conn, "Referer: %s://%s%s",
1672 url->scheme, host, url->doc);
1673 else
1674 http_cmd(conn, "Referer: %s", p);
1676 if ((p = getenv("HTTP_USER_AGENT")) != NULL && *p != '\0')
1677 http_cmd(conn, "User-Agent: %s", p);
1678 else
1679 http_cmd(conn, "User-Agent: %s " _LIBFETCH_VER, getprogname());
1680 if (url->offset > 0)
1681 http_cmd(conn, "Range: bytes=%lld-", (long long)url->offset);
1682 http_cmd(conn, "Connection: close");
1683 http_cmd(conn, "");
1686 * Force the queued request to be dispatched. Normally, one
1687 * would do this with shutdown(2) but squid proxies can be
1688 * configured to disallow such half-closed connections. To
1689 * be compatible with such configurations, fiddle with socket
1690 * options to force the pending data to be written.
1692 val = 0;
1693 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val,
1694 sizeof(val));
1695 val = 1;
1696 setsockopt(conn->sd, IPPROTO_TCP, TCP_NODELAY, &val,
1697 sizeof(val));
1699 /* get reply */
1700 switch (http_get_reply(conn)) {
1701 case HTTP_OK:
1702 case HTTP_PARTIAL:
1703 case HTTP_NOT_MODIFIED:
1704 /* fine */
1705 break;
1706 case HTTP_MOVED_PERM:
1707 case HTTP_MOVED_TEMP:
1708 case HTTP_SEE_OTHER:
1709 case HTTP_USE_PROXY:
1711 * Not so fine, but we still have to read the
1712 * headers to get the new location.
1714 break;
1715 case HTTP_NEED_AUTH:
1716 if (server_challenges.valid) {
1718 * We already sent out authorization code,
1719 * so there's nothing more we can do.
1721 http_seterr(conn->err);
1722 goto ouch;
1724 /* try again, but send the password this time */
1725 if (verbose)
1726 fetch_info("server requires authorization");
1727 break;
1728 case HTTP_NEED_PROXY_AUTH:
1729 if (proxy_challenges.valid) {
1731 * We already sent our proxy
1732 * authorization code, so there's
1733 * nothing more we can do. */
1734 http_seterr(conn->err);
1735 goto ouch;
1737 /* try again, but send the password this time */
1738 if (verbose)
1739 fetch_info("proxy requires authorization");
1740 break;
1741 case HTTP_BAD_RANGE:
1743 * This can happen if we ask for 0 bytes because
1744 * we already have the whole file. Consider this
1745 * a success for now, and check sizes later.
1747 break;
1748 case HTTP_PROTOCOL_ERROR:
1749 /* fall through */
1750 case -1:
1751 fetch_syserr();
1752 goto ouch;
1753 default:
1754 http_seterr(conn->err);
1755 if (!verbose)
1756 goto ouch;
1757 /* fall through so we can get the full error message */
1760 /* get headers. http_next_header expects one line readahead */
1761 if (fetch_getln(conn) == -1) {
1762 fetch_syserr();
1763 goto ouch;
1765 do {
1766 switch ((h = http_next_header(conn, &headerbuf, &p))) {
1767 case hdr_syserror:
1768 fetch_syserr();
1769 goto ouch;
1770 case hdr_error:
1771 http_seterr(HTTP_PROTOCOL_ERROR);
1772 goto ouch;
1773 case hdr_content_length:
1774 http_parse_length(p, &clength);
1775 break;
1776 case hdr_content_range:
1777 http_parse_range(p, &offset, &length, &size);
1778 break;
1779 case hdr_last_modified:
1780 http_parse_mtime(p, &mtime);
1781 break;
1782 case hdr_location:
1783 if (!HTTP_REDIRECT(conn->err))
1784 break;
1786 * if the A flag is set, we don't follow
1787 * temporary redirects.
1789 if (noredirect &&
1790 conn->err != HTTP_MOVED_PERM &&
1791 conn->err != HTTP_PERM_REDIRECT &&
1792 conn->err != HTTP_USE_PROXY) {
1793 n = 1;
1794 break;
1796 if (new)
1797 free(new);
1798 if (verbose)
1799 fetch_info("%d redirect to %s", conn->err, p);
1800 if (*p == '/')
1801 /* absolute path */
1802 new = fetchMakeURL(url->scheme, url->host, url->port, p,
1803 url->user, url->pwd);
1804 else
1805 new = fetchParseURL(p);
1806 if (new == NULL) {
1807 /* XXX should set an error code */
1808 DEBUG(fprintf(stderr, "failed to parse new URL\n"));
1809 goto ouch;
1812 /* Only copy credentials if the host matches */
1813 if (!strcmp(new->host, url->host) && !*new->user && !*new->pwd) {
1814 strcpy(new->user, url->user);
1815 strcpy(new->pwd, url->pwd);
1817 new->offset = url->offset;
1818 new->length = url->length;
1819 break;
1820 case hdr_transfer_encoding:
1821 /* XXX weak test*/
1822 chunked = (strcasecmp(p, "chunked") == 0);
1823 break;
1824 case hdr_www_authenticate:
1825 if (conn->err != HTTP_NEED_AUTH)
1826 break;
1827 if (http_parse_authenticate(p, &server_challenges) == 0)
1828 ++n;
1829 break;
1830 case hdr_proxy_authenticate:
1831 if (conn->err != HTTP_NEED_PROXY_AUTH)
1832 break;
1833 if (http_parse_authenticate(p, &proxy_challenges) == 0)
1834 ++n;
1835 break;
1836 case hdr_end:
1837 /* fall through */
1838 case hdr_unknown:
1839 /* ignore */
1840 break;
1842 } while (h > hdr_end);
1844 /* we need to provide authentication */
1845 if (conn->err == HTTP_NEED_AUTH ||
1846 conn->err == HTTP_NEED_PROXY_AUTH) {
1847 e = conn->err;
1848 if ((conn->err == HTTP_NEED_AUTH &&
1849 !server_challenges.valid) ||
1850 (conn->err == HTTP_NEED_PROXY_AUTH &&
1851 !proxy_challenges.valid)) {
1852 /* 401/7 but no www/proxy-authenticate ?? */
1853 DEBUG(fprintf(stderr, "401/7 and no auth header\n"));
1854 goto ouch;
1856 fetch_close(conn);
1857 conn = NULL;
1858 continue;
1861 /* requested range not satisfiable */
1862 if (conn->err == HTTP_BAD_RANGE) {
1863 if (url->offset == size && url->length == 0) {
1864 /* asked for 0 bytes; fake it */
1865 offset = url->offset;
1866 clength = -1;
1867 conn->err = HTTP_OK;
1868 break;
1869 } else {
1870 http_seterr(conn->err);
1871 goto ouch;
1875 /* we have a hit or an error */
1876 if (conn->err == HTTP_OK
1877 || conn->err == HTTP_NOT_MODIFIED
1878 || conn->err == HTTP_PARTIAL
1879 || HTTP_ERROR(conn->err))
1880 break;
1882 /* all other cases: we got a redirect */
1883 e = conn->err;
1884 clean_http_auth_challenges(&server_challenges);
1885 fetch_close(conn);
1886 conn = NULL;
1887 if (!new) {
1888 DEBUG(fprintf(stderr, "redirect with no new location\n"));
1889 break;
1891 if (url != URL)
1892 fetchFreeURL(url);
1893 url = new;
1894 } while (++i < n);
1896 /* we failed, or ran out of retries */
1897 if (conn == NULL) {
1898 http_seterr(e);
1899 goto ouch;
1902 DEBUG(fprintf(stderr, "offset %lld, length %lld,"
1903 " size %lld, clength %lld\n",
1904 (long long)offset, (long long)length,
1905 (long long)size, (long long)clength));
1907 if (conn->err == HTTP_NOT_MODIFIED) {
1908 http_seterr(HTTP_NOT_MODIFIED);
1909 return (NULL);
1912 /* check for inconsistencies */
1913 if (clength != -1 && length != -1 && clength != length) {
1914 http_seterr(HTTP_PROTOCOL_ERROR);
1915 goto ouch;
1917 if (clength == -1)
1918 clength = length;
1919 if (clength != -1)
1920 length = offset + clength;
1921 if (length != -1 && size != -1 && length != size) {
1922 http_seterr(HTTP_PROTOCOL_ERROR);
1923 goto ouch;
1925 if (size == -1)
1926 size = length;
1928 /* fill in stats */
1929 if (us) {
1930 us->size = size;
1931 us->atime = us->mtime = mtime;
1934 /* too far? */
1935 if (URL->offset > 0 && offset > URL->offset) {
1936 http_seterr(HTTP_PROTOCOL_ERROR);
1937 goto ouch;
1940 /* report back real offset and size */
1941 URL->offset = offset;
1942 URL->length = clength;
1944 /* wrap it up in a FILE */
1945 if ((f = http_funopen(conn, chunked)) == NULL) {
1946 fetch_syserr();
1947 goto ouch;
1950 if (url != URL)
1951 fetchFreeURL(url);
1952 if (purl)
1953 fetchFreeURL(purl);
1955 if (HTTP_ERROR(conn->err)) {
1956 http_print_html(stderr, f);
1957 fclose(f);
1958 f = NULL;
1960 clean_http_headerbuf(&headerbuf);
1961 clean_http_auth_challenges(&server_challenges);
1962 clean_http_auth_challenges(&proxy_challenges);
1963 return (f);
1965 ouch:
1966 if (url != URL)
1967 fetchFreeURL(url);
1968 if (purl)
1969 fetchFreeURL(purl);
1970 if (conn != NULL)
1971 fetch_close(conn);
1972 clean_http_headerbuf(&headerbuf);
1973 clean_http_auth_challenges(&server_challenges);
1974 clean_http_auth_challenges(&proxy_challenges);
1975 return (NULL);
1979 /*****************************************************************************
1980 * Entry points
1984 * Retrieve and stat a file by HTTP
1986 FILE *
1987 fetchXGetHTTP(struct url *URL, struct url_stat *us, const char *flags)
1989 return (http_request(URL, "GET", us, http_get_proxy(URL, flags), flags));
1993 * Retrieve a file by HTTP
1995 FILE *
1996 fetchGetHTTP(struct url *URL, const char *flags)
1998 return (fetchXGetHTTP(URL, NULL, flags));
2002 * Store a file by HTTP
2004 FILE *
2005 fetchPutHTTP(struct url *URL __unused, const char *flags __unused)
2007 warnx("fetchPutHTTP(): not implemented");
2008 return (NULL);
2012 * Get an HTTP document's metadata
2015 fetchStatHTTP(struct url *URL, struct url_stat *us, const char *flags)
2017 FILE *f;
2019 f = http_request(URL, "HEAD", us, http_get_proxy(URL, flags), flags);
2020 if (f == NULL)
2021 return (-1);
2022 fclose(f);
2023 return (0);
2027 * List a directory
2029 struct url_ent *
2030 fetchListHTTP(struct url *url __unused, const char *flags __unused)
2032 warnx("fetchListHTTP(): not implemented");
2033 return (NULL);