http.c revision 254650 
1/*-
2 * Copyright (c) 2000-2013 Dag-Erling Sm��rgrav
3 * All rights reserved.
4 *
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.
16 *
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.
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD: head/lib/libfetch/http.c 254650 2013-08-22 07:43:36Z des $");
31
32/*
33 * The following copyright applies to the base64 code:
34 *
35 *-
36 * Copyright 1997 Massachusetts Institute of Technology
37 *
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
48 * warranty.
49 *
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
61 * SUCH DAMAGE.
62 */
63
64#include <sys/param.h>
65#include <sys/socket.h>
66#include <sys/time.h>
67
68#include <ctype.h>
69#include <err.h>
70#include <errno.h>
71#include <locale.h>
72#include <netdb.h>
73#include <stdarg.h>
74#include <stdio.h>
75#include <stdlib.h>
76#include <string.h>
77#include <time.h>
78#include <unistd.h>
79
80#ifdef WITH_SSL
81#include <openssl/md5.h>
82#define MD5Init(c) MD5_Init(c)
83#define MD5Update(c, data, len) MD5_Update(c, data, len)
84#define MD5Final(md, c) MD5_Final(md, c)
85#else
86#include <md5.h>
87#endif
88
89#include <netinet/in.h>
90#include <netinet/tcp.h>
91
92#include "fetch.h"
93#include "common.h"
94#include "httperr.h"
95
96/* Maximum number of redirects to follow */
97#define MAX_REDIRECT 20
98
99/* Symbolic names for reply codes we care about */
100#define HTTP_OK			200
101#define HTTP_PARTIAL		206
102#define HTTP_MOVED_PERM		301
103#define HTTP_MOVED_TEMP		302
104#define HTTP_SEE_OTHER		303
105#define HTTP_NOT_MODIFIED	304
106#define HTTP_USE_PROXY		305
107#define HTTP_TEMP_REDIRECT	307
108#define HTTP_PERM_REDIRECT	308
109#define HTTP_NEED_AUTH		401
110#define HTTP_NEED_PROXY_AUTH	407
111#define HTTP_BAD_RANGE		416
112#define HTTP_PROTOCOL_ERROR	999
113
114#define HTTP_REDIRECT(xyz) ((xyz) == HTTP_MOVED_PERM \
115			    || (xyz) == HTTP_MOVED_TEMP \
116			    || (xyz) == HTTP_TEMP_REDIRECT \
117			    || (xyz) == HTTP_USE_PROXY \
118			    || (xyz) == HTTP_SEE_OTHER)
119
120#define HTTP_ERROR(xyz) ((xyz) > 400 && (xyz) < 599)
121
122
123/*****************************************************************************
124 * I/O functions for decoding chunked streams
125 */
126
127struct httpio
128{
129	conn_t		*conn;		/* connection */
130	int		 chunked;	/* chunked mode */
131	char		*buf;		/* chunk buffer */
132	size_t		 bufsize;	/* size of chunk buffer */
133	ssize_t		 buflen;	/* amount of data currently in buffer */
134	int		 bufpos;	/* current read offset in buffer */
135	int		 eof;		/* end-of-file flag */
136	int		 error;		/* error flag */
137	size_t		 chunksize;	/* remaining size of current chunk */
138#ifndef NDEBUG
139	size_t		 total;
140#endif
141};
142
143/*
144 * Get next chunk header
145 */
146static int
147http_new_chunk(struct httpio *io)
148{
149	char *p;
150
151	if (fetch_getln(io->conn) == -1)
152		return (-1);
153
154	if (io->conn->buflen < 2 || !isxdigit((unsigned char)*io->conn->buf))
155		return (-1);
156
157	for (p = io->conn->buf; *p && !isspace((unsigned char)*p); ++p) {
158		if (*p == ';')
159			break;
160		if (!isxdigit((unsigned char)*p))
161			return (-1);
162		if (isdigit((unsigned char)*p)) {
163			io->chunksize = io->chunksize * 16 +
164			    *p - '0';
165		} else {
166			io->chunksize = io->chunksize * 16 +
167			    10 + tolower((unsigned char)*p) - 'a';
168		}
169	}
170
171#ifndef NDEBUG
172	if (fetchDebug) {
173		io->total += io->chunksize;
174		if (io->chunksize == 0)
175			fprintf(stderr, "%s(): end of last chunk\n", __func__);
176		else
177			fprintf(stderr, "%s(): new chunk: %lu (%lu)\n",
178			    __func__, (unsigned long)io->chunksize,
179			    (unsigned long)io->total);
180	}
181#endif
182
183	return (io->chunksize);
184}
185
186/*
187 * Grow the input buffer to at least len bytes
188 */
189static inline int
190http_growbuf(struct httpio *io, size_t len)
191{
192	char *tmp;
193
194	if (io->bufsize >= len)
195		return (0);
196
197	if ((tmp = realloc(io->buf, len)) == NULL)
198		return (-1);
199	io->buf = tmp;
200	io->bufsize = len;
201	return (0);
202}
203
204/*
205 * Fill the input buffer, do chunk decoding on the fly
206 */
207static int
208http_fillbuf(struct httpio *io, size_t len)
209{
210	ssize_t nbytes;
211
212	if (io->error)
213		return (-1);
214	if (io->eof)
215		return (0);
216
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);
223		}
224		io->buflen = nbytes;
225		io->bufpos = 0;
226		return (io->buflen);
227	}
228
229	if (io->chunksize == 0) {
230		switch (http_new_chunk(io)) {
231		case -1:
232			io->error = 1;
233			return (-1);
234		case 0:
235			io->eof = 1;
236			return (0);
237		}
238	}
239
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);
247	}
248	io->buflen = nbytes;
249	io->chunksize -= io->buflen;
250
251	if (io->chunksize == 0) {
252		char endl[2];
253
254		if (fetch_read(io->conn, endl, 2) != 2 ||
255		    endl[0] != '\r' || endl[1] != '\n')
256			return (-1);
257	}
258
259	io->bufpos = 0;
260
261	return (io->buflen);
262}
263
264/*
265 * Read function
266 */
267static int
268http_readfn(void *v, char *buf, int len)
269{
270	struct httpio *io = (struct httpio *)v;
271	int l, pos;
272
273	if (io->error)
274		return (-1);
275	if (io->eof)
276		return (0);
277
278	for (pos = 0; len > 0; pos += l, len -= l) {
279		/* empty buffer */
280		if (!io->buf || io->bufpos == io->buflen)
281			if (http_fillbuf(io, len) < 1)
282				break;
283		l = io->buflen - io->bufpos;
284		if (len < l)
285			l = len;
286		memcpy(buf + pos, io->buf + io->bufpos, l);
287		io->bufpos += l;
288	}
289
290	if (!pos && io->error) {
291		if (io->error == EINTR)
292			io->error = 0;
293		return (-1);
294	}
295	return (pos);
296}
297
298/*
299 * Write function
300 */
301static int
302http_writefn(void *v, const char *buf, int len)
303{
304	struct httpio *io = (struct httpio *)v;
305
306	return (fetch_write(io->conn, buf, len));
307}
308
309/*
310 * Close function
311 */
312static int
313http_closefn(void *v)
314{
315	struct httpio *io = (struct httpio *)v;
316	int r;
317
318	r = fetch_close(io->conn);
319	if (io->buf)
320		free(io->buf);
321	free(io);
322	return (r);
323}
324
325/*
326 * Wrap a file descriptor up
327 */
328static FILE *
329http_funopen(conn_t *conn, int chunked)
330{
331	struct httpio *io;
332	FILE *f;
333
334	if ((io = calloc(1, sizeof(*io))) == NULL) {
335		fetch_syserr();
336		return (NULL);
337	}
338	io->conn = conn;
339	io->chunked = chunked;
340	f = funopen(io, http_readfn, http_writefn, NULL, http_closefn);
341	if (f == NULL) {
342		fetch_syserr();
343		free(io);
344		return (NULL);
345	}
346	return (f);
347}
348
349
350/*****************************************************************************
351 * Helper functions for talking to the server and parsing its replies
352 */
353
354/* Header types */
355typedef enum {
356	hdr_syserror = -2,
357	hdr_error = -1,
358	hdr_end = 0,
359	hdr_unknown = 1,
360	hdr_content_length,
361	hdr_content_range,
362	hdr_last_modified,
363	hdr_location,
364	hdr_transfer_encoding,
365	hdr_www_authenticate,
366	hdr_proxy_authenticate,
367} hdr_t;
368
369/* Names of interesting headers */
370static struct {
371	hdr_t		 num;
372	const char	*name;
373} hdr_names[] = {
374	{ hdr_content_length,		"Content-Length" },
375	{ hdr_content_range,		"Content-Range" },
376	{ hdr_last_modified,		"Last-Modified" },
377	{ hdr_location,			"Location" },
378	{ hdr_transfer_encoding,	"Transfer-Encoding" },
379	{ hdr_www_authenticate,		"WWW-Authenticate" },
380	{ hdr_proxy_authenticate,	"Proxy-Authenticate" },
381	{ hdr_unknown,			NULL },
382};
383
384/*
385 * Send a formatted line; optionally echo to terminal
386 */
387static int
388http_cmd(conn_t *conn, const char *fmt, ...)
389{
390	va_list ap;
391	size_t len;
392	char *msg;
393	int r;
394
395	va_start(ap, fmt);
396	len = vasprintf(&msg, fmt, ap);
397	va_end(ap);
398
399	if (msg == NULL) {
400		errno = ENOMEM;
401		fetch_syserr();
402		return (-1);
403	}
404
405	r = fetch_putln(conn, msg, len);
406	free(msg);
407
408	if (r == -1) {
409		fetch_syserr();
410		return (-1);
411	}
412
413	return (0);
414}
415
416/*
417 * Get and parse status line
418 */
419static int
420http_get_reply(conn_t *conn)
421{
422	char *p;
423
424	if (fetch_getln(conn) == -1)
425		return (-1);
426	/*
427	 * A valid status line looks like "HTTP/m.n xyz reason" where m
428	 * and n are the major and minor protocol version numbers and xyz
429	 * is the reply code.
430	 * Unfortunately, there are servers out there (NCSA 1.5.1, to name
431	 * just one) that do not send a version number, so we can't rely
432	 * on finding one, but if we do, insist on it being 1.0 or 1.1.
433	 * We don't care about the reason phrase.
434	 */
435	if (strncmp(conn->buf, "HTTP", 4) != 0)
436		return (HTTP_PROTOCOL_ERROR);
437	p = conn->buf + 4;
438	if (*p == '/') {
439		if (p[1] != '1' || p[2] != '.' || (p[3] != '0' && p[3] != '1'))
440			return (HTTP_PROTOCOL_ERROR);
441		p += 4;
442	}
443	if (*p != ' ' ||
444	    !isdigit((unsigned char)p[1]) ||
445	    !isdigit((unsigned char)p[2]) ||
446	    !isdigit((unsigned char)p[3]))
447		return (HTTP_PROTOCOL_ERROR);
448
449	conn->err = (p[1] - '0') * 100 + (p[2] - '0') * 10 + (p[3] - '0');
450	return (conn->err);
451}
452
453/*
454 * Check a header; if the type matches the given string, return a pointer
455 * to the beginning of the value.
456 */
457static const char *
458http_match(const char *str, const char *hdr)
459{
460	while (*str && *hdr &&
461	    tolower((unsigned char)*str++) == tolower((unsigned char)*hdr++))
462		/* nothing */;
463	if (*str || *hdr != ':')
464		return (NULL);
465	while (*hdr && isspace((unsigned char)*++hdr))
466		/* nothing */;
467	return (hdr);
468}
469
470
471/*
472 * Get the next header and return the appropriate symbolic code.  We
473 * need to read one line ahead for checking for a continuation line
474 * belonging to the current header (continuation lines start with
475 * white space).
476 *
477 * We get called with a fresh line already in the conn buffer, either
478 * from the previous http_next_header() invocation, or, the first
479 * time, from a fetch_getln() performed by our caller.
480 *
481 * This stops when we encounter an empty line (we dont read beyond the header
482 * area).
483 *
484 * Note that the "headerbuf" is just a place to return the result. Its
485 * contents are not used for the next call. This means that no cleanup
486 * is needed when ie doing another connection, just call the cleanup when
487 * fully done to deallocate memory.
488 */
489
490/* Limit the max number of continuation lines to some reasonable value */
491#define HTTP_MAX_CONT_LINES 10
492
493/* Place into which to build a header from one or several lines */
494typedef struct {
495	char	*buf;		/* buffer */
496	size_t	 bufsize;	/* buffer size */
497	size_t	 buflen;	/* length of buffer contents */
498} http_headerbuf_t;
499
500static void
501init_http_headerbuf(http_headerbuf_t *buf)
502{
503	buf->buf = NULL;
504	buf->bufsize = 0;
505	buf->buflen = 0;
506}
507
508static void
509clean_http_headerbuf(http_headerbuf_t *buf)
510{
511	if (buf->buf)
512		free(buf->buf);
513	init_http_headerbuf(buf);
514}
515
516/* Remove whitespace at the end of the buffer */
517static void
518http_conn_trimright(conn_t *conn)
519{
520	while (conn->buflen &&
521	       isspace((unsigned char)conn->buf[conn->buflen - 1]))
522		conn->buflen--;
523	conn->buf[conn->buflen] = '\0';
524}
525
526static hdr_t
527http_next_header(conn_t *conn, http_headerbuf_t *hbuf, const char **p)
528{
529	unsigned int i, len;
530
531	/*
532	 * Have to do the stripping here because of the first line. So
533	 * it's done twice for the subsequent lines. No big deal
534	 */
535	http_conn_trimright(conn);
536	if (conn->buflen == 0)
537		return (hdr_end);
538
539	/* Copy the line to the headerbuf */
540	if (hbuf->bufsize < conn->buflen + 1) {
541		if ((hbuf->buf = realloc(hbuf->buf, conn->buflen + 1)) == NULL)
542			return (hdr_syserror);
543		hbuf->bufsize = conn->buflen + 1;
544	}
545	strcpy(hbuf->buf, conn->buf);
546	hbuf->buflen = conn->buflen;
547
548	/*
549	 * Fetch possible continuation lines. Stop at 1st non-continuation
550	 * and leave it in the conn buffer
551	 */
552	for (i = 0; i < HTTP_MAX_CONT_LINES; i++) {
553		if (fetch_getln(conn) == -1)
554			return (hdr_syserror);
555
556		/*
557		 * Note: we carry on the idea from the previous version
558		 * that a pure whitespace line is equivalent to an empty
559		 * one (so it's not continuation and will be handled when
560		 * we are called next)
561		 */
562		http_conn_trimright(conn);
563		if (conn->buf[0] != ' ' && conn->buf[0] != "\t"[0])
564			break;
565
566		/* Got a continuation line. Concatenate to previous */
567		len = hbuf->buflen + conn->buflen;
568		if (hbuf->bufsize < len + 1) {
569			len *= 2;
570			if ((hbuf->buf = realloc(hbuf->buf, len + 1)) == NULL)
571				return (hdr_syserror);
572			hbuf->bufsize = len + 1;
573		}
574		strcpy(hbuf->buf + hbuf->buflen, conn->buf);
575		hbuf->buflen += conn->buflen;
576	}
577
578	/*
579	 * We could check for malformed headers but we don't really care.
580	 * A valid header starts with a token immediately followed by a
581	 * colon; a token is any sequence of non-control, non-whitespace
582	 * characters except "()<>@,;:\\\"{}".
583	 */
584	for (i = 0; hdr_names[i].num != hdr_unknown; i++)
585		if ((*p = http_match(hdr_names[i].name, hbuf->buf)) != NULL)
586			return (hdr_names[i].num);
587
588	return (hdr_unknown);
589}
590
591/**************************
592 * [Proxy-]Authenticate header parsing
593 */
594
595/*
596 * Read doublequote-delimited string into output buffer obuf (allocated
597 * by caller, whose responsibility it is to ensure that it's big enough)
598 * cp points to the first char after the initial '"'
599 * Handles \ quoting
600 * Returns pointer to the first char after the terminating double quote, or
601 * NULL for error.
602 */
603static const char *
604http_parse_headerstring(const char *cp, char *obuf)
605{
606	for (;;) {
607		switch (*cp) {
608		case 0: /* Unterminated string */
609			*obuf = 0;
610			return (NULL);
611		case '"': /* Ending quote */
612			*obuf = 0;
613			return (++cp);
614		case '\\':
615			if (*++cp == 0) {
616				*obuf = 0;
617				return (NULL);
618			}
619			/* FALLTHROUGH */
620		default:
621			*obuf++ = *cp++;
622		}
623	}
624}
625
626/* Http auth challenge schemes */
627typedef enum {HTTPAS_UNKNOWN, HTTPAS_BASIC,HTTPAS_DIGEST} http_auth_schemes_t;
628
629/* Data holder for a Basic or Digest challenge. */
630typedef struct {
631	http_auth_schemes_t scheme;
632	char	*realm;
633	char	*qop;
634	char	*nonce;
635	char	*opaque;
636	char	*algo;
637	int	 stale;
638	int	 nc; /* Nonce count */
639} http_auth_challenge_t;
640
641static void
642init_http_auth_challenge(http_auth_challenge_t *b)
643{
644	b->scheme = HTTPAS_UNKNOWN;
645	b->realm = b->qop = b->nonce = b->opaque = b->algo = NULL;
646	b->stale = b->nc = 0;
647}
648
649static void
650clean_http_auth_challenge(http_auth_challenge_t *b)
651{
652	if (b->realm)
653		free(b->realm);
654	if (b->qop)
655		free(b->qop);
656	if (b->nonce)
657		free(b->nonce);
658	if (b->opaque)
659		free(b->opaque);
660	if (b->algo)
661		free(b->algo);
662	init_http_auth_challenge(b);
663}
664
665/* Data holder for an array of challenges offered in an http response. */
666#define MAX_CHALLENGES 10
667typedef struct {
668	http_auth_challenge_t *challenges[MAX_CHALLENGES];
669	int	count; /* Number of parsed challenges in the array */
670	int	valid; /* We did parse an authenticate header */
671} http_auth_challenges_t;
672
673static void
674init_http_auth_challenges(http_auth_challenges_t *cs)
675{
676	int i;
677	for (i = 0; i < MAX_CHALLENGES; i++)
678		cs->challenges[i] = NULL;
679	cs->count = cs->valid = 0;
680}
681
682static void
683clean_http_auth_challenges(http_auth_challenges_t *cs)
684{
685	int i;
686	/* We rely on non-zero pointers being allocated, not on the count */
687	for (i = 0; i < MAX_CHALLENGES; i++) {
688		if (cs->challenges[i] != NULL) {
689			clean_http_auth_challenge(cs->challenges[i]);
690			free(cs->challenges[i]);
691		}
692	}
693	init_http_auth_challenges(cs);
694}
695
696/*
697 * Enumeration for lexical elements. Separators will be returned as their own
698 * ascii value
699 */
700typedef enum {HTTPHL_WORD=256, HTTPHL_STRING=257, HTTPHL_END=258,
701	      HTTPHL_ERROR = 259} http_header_lex_t;
702
703/*
704 * Determine what kind of token comes next and return possible value
705 * in buf, which is supposed to have been allocated big enough by
706 * caller. Advance input pointer and return element type.
707 */
708static int
709http_header_lex(const char **cpp, char *buf)
710{
711	size_t l;
712	/* Eat initial whitespace */
713	*cpp += strspn(*cpp, " \t");
714	if (**cpp == 0)
715		return (HTTPHL_END);
716
717	/* Separator ? */
718	if (**cpp == ',' || **cpp == '=')
719		return (*((*cpp)++));
720
721	/* String ? */
722	if (**cpp == '"') {
723		*cpp = http_parse_headerstring(++*cpp, buf);
724		if (*cpp == NULL)
725			return (HTTPHL_ERROR);
726		return (HTTPHL_STRING);
727	}
728
729	/* Read other token, until separator or whitespace */
730	l = strcspn(*cpp, " \t,=");
731	memcpy(buf, *cpp, l);
732	buf[l] = 0;
733	*cpp += l;
734	return (HTTPHL_WORD);
735}
736
737/*
738 * Read challenges from http xxx-authenticate header and accumulate them
739 * in the challenges list structure.
740 *
741 * Headers with multiple challenges are specified by rfc2617, but
742 * servers (ie: squid) often send them in separate headers instead,
743 * which in turn is forbidden by the http spec (multiple headers with
744 * the same name are only allowed for pure comma-separated lists, see
745 * rfc2616 sec 4.2).
746 *
747 * We support both approaches anyway
748 */
749static int
750http_parse_authenticate(const char *cp, http_auth_challenges_t *cs)
751{
752	int ret = -1;
753	http_header_lex_t lex;
754	char *key = malloc(strlen(cp) + 1);
755	char *value = malloc(strlen(cp) + 1);
756	char *buf = malloc(strlen(cp) + 1);
757
758	if (key == NULL || value == NULL || buf == NULL) {
759		fetch_syserr();
760		goto out;
761	}
762
763	/* In any case we've seen the header and we set the valid bit */
764	cs->valid = 1;
765
766	/* Need word first */
767	lex = http_header_lex(&cp, key);
768	if (lex != HTTPHL_WORD)
769		goto out;
770
771	/* Loop on challenges */
772	for (; cs->count < MAX_CHALLENGES; cs->count++) {
773		cs->challenges[cs->count] =
774			malloc(sizeof(http_auth_challenge_t));
775		if (cs->challenges[cs->count] == NULL) {
776			fetch_syserr();
777			goto out;
778		}
779		init_http_auth_challenge(cs->challenges[cs->count]);
780		if (!strcasecmp(key, "basic")) {
781			cs->challenges[cs->count]->scheme = HTTPAS_BASIC;
782		} else if (!strcasecmp(key, "digest")) {
783			cs->challenges[cs->count]->scheme = HTTPAS_DIGEST;
784		} else {
785			cs->challenges[cs->count]->scheme = HTTPAS_UNKNOWN;
786			/*
787			 * Continue parsing as basic or digest may
788			 * follow, and the syntax is the same for
789			 * all. We'll just ignore this one when
790			 * looking at the list
791			 */
792		}
793
794		/* Loop on attributes */
795		for (;;) {
796			/* Key */
797			lex = http_header_lex(&cp, key);
798			if (lex != HTTPHL_WORD)
799				goto out;
800
801			/* Equal sign */
802			lex = http_header_lex(&cp, buf);
803			if (lex != '=')
804				goto out;
805
806			/* Value */
807			lex = http_header_lex(&cp, value);
808			if (lex != HTTPHL_WORD && lex != HTTPHL_STRING)
809				goto out;
810
811			if (!strcasecmp(key, "realm"))
812				cs->challenges[cs->count]->realm =
813					strdup(value);
814			else if (!strcasecmp(key, "qop"))
815				cs->challenges[cs->count]->qop =
816					strdup(value);
817			else if (!strcasecmp(key, "nonce"))
818				cs->challenges[cs->count]->nonce =
819					strdup(value);
820			else if (!strcasecmp(key, "opaque"))
821				cs->challenges[cs->count]->opaque =
822					strdup(value);
823			else if (!strcasecmp(key, "algorithm"))
824				cs->challenges[cs->count]->algo =
825					strdup(value);
826			else if (!strcasecmp(key, "stale"))
827				cs->challenges[cs->count]->stale =
828					strcasecmp(value, "no");
829			/* Else ignore unknown attributes */
830
831			/* Comma or Next challenge or End */
832			lex = http_header_lex(&cp, key);
833			/*
834			 * If we get a word here, this is the beginning of the
835			 * next challenge. Break the attributes loop
836			 */
837			if (lex == HTTPHL_WORD)
838				break;
839
840			if (lex == HTTPHL_END) {
841				/* End while looking for ',' is normal exit */
842				cs->count++;
843				ret = 0;
844				goto out;
845			}
846			/* Anything else is an error */
847			if (lex != ',')
848				goto out;
849
850		} /* End attributes loop */
851	} /* End challenge loop */
852
853	/*
854	 * Challenges max count exceeded. This really can't happen
855	 * with normal data, something's fishy -> error
856	 */
857
858out:
859	if (key)
860		free(key);
861	if (value)
862		free(value);
863	if (buf)
864		free(buf);
865	return (ret);
866}
867
868
869/*
870 * Parse a last-modified header
871 */
872static int
873http_parse_mtime(const char *p, time_t *mtime)
874{
875	char locale[64], *r;
876	struct tm tm;
877
878	strncpy(locale, setlocale(LC_TIME, NULL), sizeof(locale));
879	setlocale(LC_TIME, "C");
880	r = strptime(p, "%a, %d %b %Y %H:%M:%S GMT", &tm);
881	/* XXX should add support for date-2 and date-3 */
882	setlocale(LC_TIME, locale);
883	if (r == NULL)
884		return (-1);
885	DEBUG(fprintf(stderr, "last modified: [%04d-%02d-%02d "
886		  "%02d:%02d:%02d]\n",
887		  tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday,
888		  tm.tm_hour, tm.tm_min, tm.tm_sec));
889	*mtime = timegm(&tm);
890	return (0);
891}
892
893/*
894 * Parse a content-length header
895 */
896static int
897http_parse_length(const char *p, off_t *length)
898{
899	off_t len;
900
901	for (len = 0; *p && isdigit((unsigned char)*p); ++p)
902		len = len * 10 + (*p - '0');
903	if (*p)
904		return (-1);
905	DEBUG(fprintf(stderr, "content length: [%lld]\n",
906	    (long long)len));
907	*length = len;
908	return (0);
909}
910
911/*
912 * Parse a content-range header
913 */
914static int
915http_parse_range(const char *p, off_t *offset, off_t *length, off_t *size)
916{
917	off_t first, last, len;
918
919	if (strncasecmp(p, "bytes ", 6) != 0)
920		return (-1);
921	p += 6;
922	if (*p == '*') {
923		first = last = -1;
924		++p;
925	} else {
926		for (first = 0; *p && isdigit((unsigned char)*p); ++p)
927			first = first * 10 + *p - '0';
928		if (*p != '-')
929			return (-1);
930		for (last = 0, ++p; *p && isdigit((unsigned char)*p); ++p)
931			last = last * 10 + *p - '0';
932	}
933	if (first > last || *p != '/')
934		return (-1);
935	for (len = 0, ++p; *p && isdigit((unsigned char)*p); ++p)
936		len = len * 10 + *p - '0';
937	if (*p || len < last - first + 1)
938		return (-1);
939	if (first == -1) {
940		DEBUG(fprintf(stderr, "content range: [*/%lld]\n",
941		    (long long)len));
942		*length = 0;
943	} else {
944		DEBUG(fprintf(stderr, "content range: [%lld-%lld/%lld]\n",
945		    (long long)first, (long long)last, (long long)len));
946		*length = last - first + 1;
947	}
948	*offset = first;
949	*size = len;
950	return (0);
951}
952
953
954/*****************************************************************************
955 * Helper functions for authorization
956 */
957
958/*
959 * Base64 encoding
960 */
961static char *
962http_base64(const char *src)
963{
964	static const char base64[] =
965	    "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
966	    "abcdefghijklmnopqrstuvwxyz"
967	    "0123456789+/";
968	char *str, *dst;
969	size_t l;
970	int t, r;
971
972	l = strlen(src);
973	if ((str = malloc(((l + 2) / 3) * 4 + 1)) == NULL)
974		return (NULL);
975	dst = str;
976	r = 0;
977
978	while (l >= 3) {
979		t = (src[0] << 16) | (src[1] << 8) | src[2];
980		dst[0] = base64[(t >> 18) & 0x3f];
981		dst[1] = base64[(t >> 12) & 0x3f];
982		dst[2] = base64[(t >> 6) & 0x3f];
983		dst[3] = base64[(t >> 0) & 0x3f];
984		src += 3; l -= 3;
985		dst += 4; r += 4;
986	}
987
988	switch (l) {
989	case 2:
990		t = (src[0] << 16) | (src[1] << 8);
991		dst[0] = base64[(t >> 18) & 0x3f];
992		dst[1] = base64[(t >> 12) & 0x3f];
993		dst[2] = base64[(t >> 6) & 0x3f];
994		dst[3] = '=';
995		dst += 4;
996		r += 4;
997		break;
998	case 1:
999		t = src[0] << 16;
1000		dst[0] = base64[(t >> 18) & 0x3f];
1001		dst[1] = base64[(t >> 12) & 0x3f];
1002		dst[2] = dst[3] = '=';
1003		dst += 4;
1004		r += 4;
1005		break;
1006	case 0:
1007		break;
1008	}
1009
1010	*dst = 0;
1011	return (str);
1012}
1013
1014
1015/*
1016 * Extract authorization parameters from environment value.
1017 * The value is like scheme:realm:user:pass
1018 */
1019typedef struct {
1020	char	*scheme;
1021	char	*realm;
1022	char	*user;
1023	char	*password;
1024} http_auth_params_t;
1025
1026static void
1027init_http_auth_params(http_auth_params_t *s)
1028{
1029	s->scheme = s->realm = s->user = s->password = 0;
1030}
1031
1032static void
1033clean_http_auth_params(http_auth_params_t *s)
1034{
1035	if (s->scheme)
1036		free(s->scheme);
1037	if (s->realm)
1038		free(s->realm);
1039	if (s->user)
1040		free(s->user);
1041	if (s->password)
1042		free(s->password);
1043	init_http_auth_params(s);
1044}
1045
1046static int
1047http_authfromenv(const char *p, http_auth_params_t *parms)
1048{
1049	int ret = -1;
1050	char *v, *ve;
1051	char *str = strdup(p);
1052
1053	if (str == NULL) {
1054		fetch_syserr();
1055		return (-1);
1056	}
1057	v = str;
1058
1059	if ((ve = strchr(v, ':')) == NULL)
1060		goto out;
1061
1062	*ve = 0;
1063	if ((parms->scheme = strdup(v)) == NULL) {
1064		fetch_syserr();
1065		goto out;
1066	}
1067	v = ve + 1;
1068
1069	if ((ve = strchr(v, ':')) == NULL)
1070		goto out;
1071
1072	*ve = 0;
1073	if ((parms->realm = strdup(v)) == NULL) {
1074		fetch_syserr();
1075		goto out;
1076	}
1077	v = ve + 1;
1078
1079	if ((ve = strchr(v, ':')) == NULL)
1080		goto out;
1081
1082	*ve = 0;
1083	if ((parms->user = strdup(v)) == NULL) {
1084		fetch_syserr();
1085		goto out;
1086	}
1087	v = ve + 1;
1088
1089
1090	if ((parms->password = strdup(v)) == NULL) {
1091		fetch_syserr();
1092		goto out;
1093	}
1094	ret = 0;
1095out:
1096	if (ret == -1)
1097		clean_http_auth_params(parms);
1098	if (str)
1099		free(str);
1100	return (ret);
1101}
1102
1103
1104/*
1105 * Digest response: the code to compute the digest is taken from the
1106 * sample implementation in RFC2616
1107 */
1108#define IN const
1109#define OUT
1110
1111#define HASHLEN 16
1112typedef char HASH[HASHLEN];
1113#define HASHHEXLEN 32
1114typedef char HASHHEX[HASHHEXLEN+1];
1115
1116static const char *hexchars = "0123456789abcdef";
1117static void
1118CvtHex(IN HASH Bin, OUT HASHHEX Hex)
1119{
1120	unsigned short i;
1121	unsigned char j;
1122
1123	for (i = 0; i < HASHLEN; i++) {
1124		j = (Bin[i] >> 4) & 0xf;
1125		Hex[i*2] = hexchars[j];
1126		j = Bin[i] & 0xf;
1127		Hex[i*2+1] = hexchars[j];
1128	};
1129	Hex[HASHHEXLEN] = '\0';
1130};
1131
1132/* calculate H(A1) as per spec */
1133static void
1134DigestCalcHA1(
1135	IN char * pszAlg,
1136	IN char * pszUserName,
1137	IN char * pszRealm,
1138	IN char * pszPassword,
1139	IN char * pszNonce,
1140	IN char * pszCNonce,
1141	OUT HASHHEX SessionKey
1142	)
1143{
1144	MD5_CTX Md5Ctx;
1145	HASH HA1;
1146
1147	MD5Init(&Md5Ctx);
1148	MD5Update(&Md5Ctx, pszUserName, strlen(pszUserName));
1149	MD5Update(&Md5Ctx, ":", 1);
1150	MD5Update(&Md5Ctx, pszRealm, strlen(pszRealm));
1151	MD5Update(&Md5Ctx, ":", 1);
1152	MD5Update(&Md5Ctx, pszPassword, strlen(pszPassword));
1153	MD5Final(HA1, &Md5Ctx);
1154	if (strcasecmp(pszAlg, "md5-sess") == 0) {
1155
1156		MD5Init(&Md5Ctx);
1157		MD5Update(&Md5Ctx, HA1, HASHLEN);
1158		MD5Update(&Md5Ctx, ":", 1);
1159		MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce));
1160		MD5Update(&Md5Ctx, ":", 1);
1161		MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce));
1162		MD5Final(HA1, &Md5Ctx);
1163	};
1164	CvtHex(HA1, SessionKey);
1165}
1166
1167/* calculate request-digest/response-digest as per HTTP Digest spec */
1168static void
1169DigestCalcResponse(
1170	IN HASHHEX HA1,           /* H(A1) */
1171	IN char * pszNonce,       /* nonce from server */
1172	IN char * pszNonceCount,  /* 8 hex digits */
1173	IN char * pszCNonce,      /* client nonce */
1174	IN char * pszQop,         /* qop-value: "", "auth", "auth-int" */
1175	IN char * pszMethod,      /* method from the request */
1176	IN char * pszDigestUri,   /* requested URL */
1177	IN HASHHEX HEntity,       /* H(entity body) if qop="auth-int" */
1178	OUT HASHHEX Response      /* request-digest or response-digest */
1179	)
1180{
1181/*	DEBUG(fprintf(stderr,
1182		      "Calc: HA1[%s] Nonce[%s] qop[%s] method[%s] URI[%s]\n",
1183		      HA1, pszNonce, pszQop, pszMethod, pszDigestUri));*/
1184	MD5_CTX Md5Ctx;
1185	HASH HA2;
1186	HASH RespHash;
1187	HASHHEX HA2Hex;
1188
1189	// calculate H(A2)
1190	MD5Init(&Md5Ctx);
1191	MD5Update(&Md5Ctx, pszMethod, strlen(pszMethod));
1192	MD5Update(&Md5Ctx, ":", 1);
1193	MD5Update(&Md5Ctx, pszDigestUri, strlen(pszDigestUri));
1194	if (strcasecmp(pszQop, "auth-int") == 0) {
1195		MD5Update(&Md5Ctx, ":", 1);
1196		MD5Update(&Md5Ctx, HEntity, HASHHEXLEN);
1197	};
1198	MD5Final(HA2, &Md5Ctx);
1199	CvtHex(HA2, HA2Hex);
1200
1201	// calculate response
1202	MD5Init(&Md5Ctx);
1203	MD5Update(&Md5Ctx, HA1, HASHHEXLEN);
1204	MD5Update(&Md5Ctx, ":", 1);
1205	MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce));
1206	MD5Update(&Md5Ctx, ":", 1);
1207	if (*pszQop) {
1208		MD5Update(&Md5Ctx, pszNonceCount, strlen(pszNonceCount));
1209		MD5Update(&Md5Ctx, ":", 1);
1210		MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce));
1211		MD5Update(&Md5Ctx, ":", 1);
1212		MD5Update(&Md5Ctx, pszQop, strlen(pszQop));
1213		MD5Update(&Md5Ctx, ":", 1);
1214	};
1215	MD5Update(&Md5Ctx, HA2Hex, HASHHEXLEN);
1216	MD5Final(RespHash, &Md5Ctx);
1217	CvtHex(RespHash, Response);
1218}
1219
1220/*
1221 * Generate/Send a Digest authorization header
1222 * This looks like: [Proxy-]Authorization: credentials
1223 *
1224 *  credentials      = "Digest" digest-response
1225 *  digest-response  = 1#( username | realm | nonce | digest-uri
1226 *                      | response | [ algorithm ] | [cnonce] |
1227 *                      [opaque] | [message-qop] |
1228 *                          [nonce-count]  | [auth-param] )
1229 *  username         = "username" "=" username-value
1230 *  username-value   = quoted-string
1231 *  digest-uri       = "uri" "=" digest-uri-value
1232 *  digest-uri-value = request-uri   ; As specified by HTTP/1.1
1233 *  message-qop      = "qop" "=" qop-value
1234 *  cnonce           = "cnonce" "=" cnonce-value
1235 *  cnonce-value     = nonce-value
1236 *  nonce-count      = "nc" "=" nc-value
1237 *  nc-value         = 8LHEX
1238 *  response         = "response" "=" request-digest
1239 *  request-digest = <"> 32LHEX <">
1240 */
1241static int
1242http_digest_auth(conn_t *conn, const char *hdr, http_auth_challenge_t *c,
1243		 http_auth_params_t *parms, struct url *url)
1244{
1245	int r;
1246	char noncecount[10];
1247	char cnonce[40];
1248	char *options = 0;
1249
1250	if (!c->realm || !c->nonce) {
1251		DEBUG(fprintf(stderr, "realm/nonce not set in challenge\n"));
1252		return(-1);
1253	}
1254	if (!c->algo)
1255		c->algo = strdup("");
1256
1257	if (asprintf(&options, "%s%s%s%s",
1258		     *c->algo? ",algorithm=" : "", c->algo,
1259		     c->opaque? ",opaque=" : "", c->opaque?c->opaque:"")== -1)
1260		return (-1);
1261
1262	if (!c->qop) {
1263		c->qop = strdup("");
1264		*noncecount = 0;
1265		*cnonce = 0;
1266	} else {
1267		c->nc++;
1268		sprintf(noncecount, "%08x", c->nc);
1269		/* We don't try very hard with the cnonce ... */
1270		sprintf(cnonce, "%x%lx", getpid(), (unsigned long)time(0));
1271	}
1272
1273	HASHHEX HA1;
1274	DigestCalcHA1(c->algo, parms->user, c->realm,
1275		      parms->password, c->nonce, cnonce, HA1);
1276	DEBUG(fprintf(stderr, "HA1: [%s]\n", HA1));
1277	HASHHEX digest;
1278	DigestCalcResponse(HA1, c->nonce, noncecount, cnonce, c->qop,
1279			   "GET", url->doc, "", digest);
1280
1281	if (c->qop[0]) {
1282		r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\","
1283			     "nonce=\"%s\",uri=\"%s\",response=\"%s\","
1284			     "qop=\"auth\", cnonce=\"%s\", nc=%s%s",
1285			     hdr, parms->user, c->realm,
1286			     c->nonce, url->doc, digest,
1287			     cnonce, noncecount, options);
1288	} else {
1289		r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\","
1290			     "nonce=\"%s\",uri=\"%s\",response=\"%s\"%s",
1291			     hdr, parms->user, c->realm,
1292			     c->nonce, url->doc, digest, options);
1293	}
1294	if (options)
1295		free(options);
1296	return (r);
1297}
1298
1299/*
1300 * Encode username and password
1301 */
1302static int
1303http_basic_auth(conn_t *conn, const char *hdr, const char *usr, const char *pwd)
1304{
1305	char *upw, *auth;
1306	int r;
1307
1308	DEBUG(fprintf(stderr, "basic: usr: [%s]\n", usr));
1309	DEBUG(fprintf(stderr, "basic: pwd: [%s]\n", pwd));
1310	if (asprintf(&upw, "%s:%s", usr, pwd) == -1)
1311		return (-1);
1312	auth = http_base64(upw);
1313	free(upw);
1314	if (auth == NULL)
1315		return (-1);
1316	r = http_cmd(conn, "%s: Basic %s", hdr, auth);
1317	free(auth);
1318	return (r);
1319}
1320
1321/*
1322 * Chose the challenge to answer and call the appropriate routine to
1323 * produce the header.
1324 */
1325static int
1326http_authorize(conn_t *conn, const char *hdr, http_auth_challenges_t *cs,
1327	       http_auth_params_t *parms, struct url *url)
1328{
1329	http_auth_challenge_t *basic = NULL;
1330	http_auth_challenge_t *digest = NULL;
1331	int i;
1332
1333	/* If user or pass are null we're not happy */
1334	if (!parms->user || !parms->password) {
1335		DEBUG(fprintf(stderr, "NULL usr or pass\n"));
1336		return (-1);
1337	}
1338
1339	/* Look for a Digest and a Basic challenge */
1340	for (i = 0; i < cs->count; i++) {
1341		if (cs->challenges[i]->scheme == HTTPAS_BASIC)
1342			basic = cs->challenges[i];
1343		if (cs->challenges[i]->scheme == HTTPAS_DIGEST)
1344			digest = cs->challenges[i];
1345	}
1346
1347	/* Error if "Digest" was specified and there is no Digest challenge */
1348	if (!digest && (parms->scheme &&
1349			!strcasecmp(parms->scheme, "digest"))) {
1350		DEBUG(fprintf(stderr,
1351			      "Digest auth in env, not supported by peer\n"));
1352		return (-1);
1353	}
1354	/*
1355	 * If "basic" was specified in the environment, or there is no Digest
1356	 * challenge, do the basic thing. Don't need a challenge for this,
1357	 * so no need to check basic!=NULL
1358	 */
1359	if (!digest || (parms->scheme && !strcasecmp(parms->scheme,"basic")))
1360		return (http_basic_auth(conn,hdr,parms->user,parms->password));
1361
1362	/* Else, prefer digest. We just checked that it's not NULL */
1363	return (http_digest_auth(conn, hdr, digest, parms, url));
1364}
1365
1366/*****************************************************************************
1367 * Helper functions for connecting to a server or proxy
1368 */
1369
1370/*
1371 * Connect to the correct HTTP server or proxy.
1372 */
1373static conn_t *
1374http_connect(struct url *URL, struct url *purl, const char *flags)
1375{
1376	struct url *curl;
1377	conn_t *conn;
1378	int verbose;
1379	int af, val;
1380
1381#ifdef INET6
1382	af = AF_UNSPEC;
1383#else
1384	af = AF_INET;
1385#endif
1386
1387	verbose = CHECK_FLAG('v');
1388	if (CHECK_FLAG('4'))
1389		af = AF_INET;
1390#ifdef INET6
1391	else if (CHECK_FLAG('6'))
1392		af = AF_INET6;
1393#endif
1394
1395	curl = (purl != NULL) ? purl : URL;
1396
1397	if ((conn = fetch_connect(curl->host, curl->port, af, verbose)) == NULL)
1398		/* fetch_connect() has already set an error code */
1399		return (NULL);
1400	if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 && purl) {
1401		http_cmd(conn, "CONNECT %s:%d HTTP/1.1",
1402		    URL->host, URL->port);
1403		http_cmd(conn, "Host: %s:%d",
1404		    URL->host, URL->port);
1405		http_cmd(conn, "");
1406		if (http_get_reply(conn) != HTTP_OK) {
1407			fetch_close(conn);
1408			return (NULL);
1409		}
1410		http_get_reply(conn);
1411	}
1412	if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 &&
1413	    fetch_ssl(conn, URL, verbose) == -1) {
1414		fetch_close(conn);
1415		/* grrr */
1416		errno = EAUTH;
1417		fetch_syserr();
1418		return (NULL);
1419	}
1420
1421	val = 1;
1422	setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, sizeof(val));
1423
1424	return (conn);
1425}
1426
1427static struct url *
1428http_get_proxy(struct url * url, const char *flags)
1429{
1430	struct url *purl;
1431	char *p;
1432
1433	if (flags != NULL && strchr(flags, 'd') != NULL)
1434		return (NULL);
1435	if (fetch_no_proxy_match(url->host))
1436		return (NULL);
1437	if (((p = getenv("HTTP_PROXY")) || (p = getenv("http_proxy"))) &&
1438	    *p && (purl = fetchParseURL(p))) {
1439		if (!*purl->scheme)
1440			strcpy(purl->scheme, SCHEME_HTTP);
1441		if (!purl->port)
1442			purl->port = fetch_default_proxy_port(purl->scheme);
1443		if (strcasecmp(purl->scheme, SCHEME_HTTP) == 0)
1444			return (purl);
1445		fetchFreeURL(purl);
1446	}
1447	return (NULL);
1448}
1449
1450static void
1451http_print_html(FILE *out, FILE *in)
1452{
1453	size_t len;
1454	char *line, *p, *q;
1455	int comment, tag;
1456
1457	comment = tag = 0;
1458	while ((line = fgetln(in, &len)) != NULL) {
1459		while (len && isspace((unsigned char)line[len - 1]))
1460			--len;
1461		for (p = q = line; q < line + len; ++q) {
1462			if (comment && *q == '-') {
1463				if (q + 2 < line + len &&
1464				    strcmp(q, "-->") == 0) {
1465					tag = comment = 0;
1466					q += 2;
1467				}
1468			} else if (tag && !comment && *q == '>') {
1469				p = q + 1;
1470				tag = 0;
1471			} else if (!tag && *q == '<') {
1472				if (q > p)
1473					fwrite(p, q - p, 1, out);
1474				tag = 1;
1475				if (q + 3 < line + len &&
1476				    strcmp(q, "<!--") == 0) {
1477					comment = 1;
1478					q += 3;
1479				}
1480			}
1481		}
1482		if (!tag && q > p)
1483			fwrite(p, q - p, 1, out);
1484		fputc('\n', out);
1485	}
1486}
1487
1488
1489/*****************************************************************************
1490 * Core
1491 */
1492
1493/*
1494 * Send a request and process the reply
1495 *
1496 * XXX This function is way too long, the do..while loop should be split
1497 * XXX off into a separate function.
1498 */
1499FILE *
1500http_request(struct url *URL, const char *op, struct url_stat *us,
1501	struct url *purl, const char *flags)
1502{
1503	char timebuf[80];
1504	char hbuf[MAXHOSTNAMELEN + 7], *host;
1505	conn_t *conn;
1506	struct url *url, *new;
1507	int chunked, direct, ims, noredirect, verbose;
1508	int e, i, n, val;
1509	off_t offset, clength, length, size;
1510	time_t mtime;
1511	const char *p;
1512	FILE *f;
1513	hdr_t h;
1514	struct tm *timestruct;
1515	http_headerbuf_t headerbuf;
1516	http_auth_challenges_t server_challenges;
1517	http_auth_challenges_t proxy_challenges;
1518
1519	/* The following calls don't allocate anything */
1520	init_http_headerbuf(&headerbuf);
1521	init_http_auth_challenges(&server_challenges);
1522	init_http_auth_challenges(&proxy_challenges);
1523
1524	direct = CHECK_FLAG('d');
1525	noredirect = CHECK_FLAG('A');
1526	verbose = CHECK_FLAG('v');
1527	ims = CHECK_FLAG('i');
1528
1529	if (direct && purl) {
1530		fetchFreeURL(purl);
1531		purl = NULL;
1532	}
1533
1534	/* try the provided URL first */
1535	url = URL;
1536
1537	n = MAX_REDIRECT;
1538	i = 0;
1539
1540	e = HTTP_PROTOCOL_ERROR;
1541	do {
1542		new = NULL;
1543		chunked = 0;
1544		offset = 0;
1545		clength = -1;
1546		length = -1;
1547		size = -1;
1548		mtime = 0;
1549
1550		/* check port */
1551		if (!url->port)
1552			url->port = fetch_default_port(url->scheme);
1553
1554		/* were we redirected to an FTP URL? */
1555		if (purl == NULL && strcmp(url->scheme, SCHEME_FTP) == 0) {
1556			if (strcmp(op, "GET") == 0)
1557				return (ftp_request(url, "RETR", us, purl, flags));
1558			else if (strcmp(op, "HEAD") == 0)
1559				return (ftp_request(url, "STAT", us, purl, flags));
1560		}
1561
1562		/* connect to server or proxy */
1563		if ((conn = http_connect(url, purl, flags)) == NULL)
1564			goto ouch;
1565
1566		host = url->host;
1567#ifdef INET6
1568		if (strchr(url->host, ':')) {
1569			snprintf(hbuf, sizeof(hbuf), "[%s]", url->host);
1570			host = hbuf;
1571		}
1572#endif
1573		if (url->port != fetch_default_port(url->scheme)) {
1574			if (host != hbuf) {
1575				strcpy(hbuf, host);
1576				host = hbuf;
1577			}
1578			snprintf(hbuf + strlen(hbuf),
1579			    sizeof(hbuf) - strlen(hbuf), ":%d", url->port);
1580		}
1581
1582		/* send request */
1583		if (verbose)
1584			fetch_info("requesting %s://%s%s",
1585			    url->scheme, host, url->doc);
1586		if (purl && strcasecmp(URL->scheme, SCHEME_HTTPS) != 0) {
1587			http_cmd(conn, "%s %s://%s%s HTTP/1.1",
1588			    op, url->scheme, host, url->doc);
1589		} else {
1590			http_cmd(conn, "%s %s HTTP/1.1",
1591			    op, url->doc);
1592		}
1593
1594		if (ims && url->ims_time) {
1595			timestruct = gmtime((time_t *)&url->ims_time);
1596			(void)strftime(timebuf, 80, "%a, %d %b %Y %T GMT",
1597			    timestruct);
1598			if (verbose)
1599				fetch_info("If-Modified-Since: %s", timebuf);
1600			http_cmd(conn, "If-Modified-Since: %s", timebuf);
1601		}
1602		/* virtual host */
1603		http_cmd(conn, "Host: %s", host);
1604
1605		/*
1606		 * Proxy authorization: we only send auth after we received
1607		 * a 407 error. We do not first try basic anyway (changed
1608		 * when support was added for digest-auth)
1609		 */
1610		if (purl && proxy_challenges.valid) {
1611			http_auth_params_t aparams;
1612			init_http_auth_params(&aparams);
1613			if (*purl->user || *purl->pwd) {
1614				aparams.user = purl->user ?
1615					strdup(purl->user) : strdup("");
1616				aparams.password = purl->pwd?
1617					strdup(purl->pwd) : strdup("");
1618			} else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL &&
1619				   *p != '\0') {
1620				if (http_authfromenv(p, &aparams) < 0) {
1621					http_seterr(HTTP_NEED_PROXY_AUTH);
1622					goto ouch;
1623				}
1624			}
1625			http_authorize(conn, "Proxy-Authorization",
1626				       &proxy_challenges, &aparams, url);
1627			clean_http_auth_params(&aparams);
1628		}
1629
1630		/*
1631		 * Server authorization: we never send "a priori"
1632		 * Basic auth, which used to be done if user/pass were
1633		 * set in the url. This would be weird because we'd send the
1634		 * password in the clear even if Digest is finally to be
1635		 * used (it would have made more sense for the
1636		 * pre-digest version to do this when Basic was specified
1637		 * in the environment)
1638		 */
1639		if (server_challenges.valid) {
1640			http_auth_params_t aparams;
1641			init_http_auth_params(&aparams);
1642			if (*url->user || *url->pwd) {
1643				aparams.user = url->user ?
1644					strdup(url->user) : strdup("");
1645				aparams.password = url->pwd ?
1646					strdup(url->pwd) : strdup("");
1647			} else if ((p = getenv("HTTP_AUTH")) != NULL &&
1648				   *p != '\0') {
1649				if (http_authfromenv(p, &aparams) < 0) {
1650					http_seterr(HTTP_NEED_AUTH);
1651					goto ouch;
1652				}
1653			} else if (fetchAuthMethod &&
1654				   fetchAuthMethod(url) == 0) {
1655				aparams.user = url->user ?
1656					strdup(url->user) : strdup("");
1657				aparams.password = url->pwd ?
1658					strdup(url->pwd) : strdup("");
1659			} else {
1660				http_seterr(HTTP_NEED_AUTH);
1661				goto ouch;
1662			}
1663			http_authorize(conn, "Authorization",
1664				       &server_challenges, &aparams, url);
1665			clean_http_auth_params(&aparams);
1666		}
1667
1668		/* other headers */
1669		if ((p = getenv("HTTP_ACCEPT")) != NULL) {
1670			if (*p != '\0')
1671				http_cmd(conn, "Accept: %s", p);
1672		} else {
1673			http_cmd(conn, "Accept: */*");
1674		}
1675		if ((p = getenv("HTTP_REFERER")) != NULL && *p != '\0') {
1676			if (strcasecmp(p, "auto") == 0)
1677				http_cmd(conn, "Referer: %s://%s%s",
1678				    url->scheme, host, url->doc);
1679			else
1680				http_cmd(conn, "Referer: %s", p);
1681		}
1682		if ((p = getenv("HTTP_USER_AGENT")) != NULL && *p != '\0')
1683			http_cmd(conn, "User-Agent: %s", p);
1684		else
1685			http_cmd(conn, "User-Agent: %s " _LIBFETCH_VER, getprogname());
1686		if (url->offset > 0)
1687			http_cmd(conn, "Range: bytes=%lld-", (long long)url->offset);
1688		http_cmd(conn, "Connection: close");
1689		http_cmd(conn, "");
1690
1691		/*
1692		 * Force the queued request to be dispatched.  Normally, one
1693		 * would do this with shutdown(2) but squid proxies can be
1694		 * configured to disallow such half-closed connections.  To
1695		 * be compatible with such configurations, fiddle with socket
1696		 * options to force the pending data to be written.
1697		 */
1698		val = 0;
1699		setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val,
1700			   sizeof(val));
1701		val = 1;
1702		setsockopt(conn->sd, IPPROTO_TCP, TCP_NODELAY, &val,
1703			   sizeof(val));
1704
1705		/* get reply */
1706		switch (http_get_reply(conn)) {
1707		case HTTP_OK:
1708		case HTTP_PARTIAL:
1709		case HTTP_NOT_MODIFIED:
1710			/* fine */
1711			break;
1712		case HTTP_MOVED_PERM:
1713		case HTTP_MOVED_TEMP:
1714		case HTTP_SEE_OTHER:
1715		case HTTP_USE_PROXY:
1716			/*
1717			 * Not so fine, but we still have to read the
1718			 * headers to get the new location.
1719			 */
1720			break;
1721		case HTTP_NEED_AUTH:
1722			if (server_challenges.valid) {
1723				/*
1724				 * We already sent out authorization code,
1725				 * so there's nothing more we can do.
1726				 */
1727				http_seterr(conn->err);
1728				goto ouch;
1729			}
1730			/* try again, but send the password this time */
1731			if (verbose)
1732				fetch_info("server requires authorization");
1733			break;
1734		case HTTP_NEED_PROXY_AUTH:
1735			if (proxy_challenges.valid) {
1736				/*
1737				 * We already sent our proxy
1738				 * authorization code, so there's
1739				 * nothing more we can do. */
1740				http_seterr(conn->err);
1741				goto ouch;
1742			}
1743			/* try again, but send the password this time */
1744			if (verbose)
1745				fetch_info("proxy requires authorization");
1746			break;
1747		case HTTP_BAD_RANGE:
1748			/*
1749			 * This can happen if we ask for 0 bytes because
1750			 * we already have the whole file.  Consider this
1751			 * a success for now, and check sizes later.
1752			 */
1753			break;
1754		case HTTP_PROTOCOL_ERROR:
1755			/* fall through */
1756		case -1:
1757			fetch_syserr();
1758			goto ouch;
1759		default:
1760			http_seterr(conn->err);
1761			if (!verbose)
1762				goto ouch;
1763			/* fall through so we can get the full error message */
1764		}
1765
1766		/* get headers. http_next_header expects one line readahead */
1767		if (fetch_getln(conn) == -1) {
1768			fetch_syserr();
1769			goto ouch;
1770		}
1771		do {
1772			switch ((h = http_next_header(conn, &headerbuf, &p))) {
1773			case hdr_syserror:
1774				fetch_syserr();
1775				goto ouch;
1776			case hdr_error:
1777				http_seterr(HTTP_PROTOCOL_ERROR);
1778				goto ouch;
1779			case hdr_content_length:
1780				http_parse_length(p, &clength);
1781				break;
1782			case hdr_content_range:
1783				http_parse_range(p, &offset, &length, &size);
1784				break;
1785			case hdr_last_modified:
1786				http_parse_mtime(p, &mtime);
1787				break;
1788			case hdr_location:
1789				if (!HTTP_REDIRECT(conn->err))
1790					break;
1791				/*
1792				 * if the A flag is set, we don't follow
1793				 * temporary redirects.
1794				 */
1795				if (noredirect &&
1796				    conn->err != HTTP_MOVED_PERM &&
1797				    conn->err != HTTP_PERM_REDIRECT &&
1798				    conn->err != HTTP_USE_PROXY) {
1799					n = 1;
1800					break;
1801				}
1802				if (new)
1803					free(new);
1804				if (verbose)
1805					fetch_info("%d redirect to %s", conn->err, p);
1806				if (*p == '/')
1807					/* absolute path */
1808					new = fetchMakeURL(url->scheme, url->host, url->port, p,
1809					    url->user, url->pwd);
1810				else
1811					new = fetchParseURL(p);
1812				if (new == NULL) {
1813					/* XXX should set an error code */
1814					DEBUG(fprintf(stderr, "failed to parse new URL\n"));
1815					goto ouch;
1816				}
1817
1818				/* Only copy credentials if the host matches */
1819				if (!strcmp(new->host, url->host) && !*new->user && !*new->pwd) {
1820					strcpy(new->user, url->user);
1821					strcpy(new->pwd, url->pwd);
1822				}
1823				new->offset = url->offset;
1824				new->length = url->length;
1825				break;
1826			case hdr_transfer_encoding:
1827				/* XXX weak test*/
1828				chunked = (strcasecmp(p, "chunked") == 0);
1829				break;
1830			case hdr_www_authenticate:
1831				if (conn->err != HTTP_NEED_AUTH)
1832					break;
1833				if (http_parse_authenticate(p, &server_challenges) == 0)
1834					++n;
1835				break;
1836			case hdr_proxy_authenticate:
1837				if (conn->err != HTTP_NEED_PROXY_AUTH)
1838					break;
1839				if (http_parse_authenticate(p, &proxy_challenges) == 0)
1840					++n;
1841				break;
1842			case hdr_end:
1843				/* fall through */
1844			case hdr_unknown:
1845				/* ignore */
1846				break;
1847			}
1848		} while (h > hdr_end);
1849
1850		/* we need to provide authentication */
1851		if (conn->err == HTTP_NEED_AUTH ||
1852		    conn->err == HTTP_NEED_PROXY_AUTH) {
1853			e = conn->err;
1854			if ((conn->err == HTTP_NEED_AUTH &&
1855			     !server_challenges.valid) ||
1856			    (conn->err == HTTP_NEED_PROXY_AUTH &&
1857			     !proxy_challenges.valid)) {
1858				/* 401/7 but no www/proxy-authenticate ?? */
1859				DEBUG(fprintf(stderr, "401/7 and no auth header\n"));
1860				goto ouch;
1861			}
1862			fetch_close(conn);
1863			conn = NULL;
1864			continue;
1865		}
1866
1867		/* requested range not satisfiable */
1868		if (conn->err == HTTP_BAD_RANGE) {
1869			if (url->offset == size && url->length == 0) {
1870				/* asked for 0 bytes; fake it */
1871				offset = url->offset;
1872				clength = -1;
1873				conn->err = HTTP_OK;
1874				break;
1875			} else {
1876				http_seterr(conn->err);
1877				goto ouch;
1878			}
1879		}
1880
1881		/* we have a hit or an error */
1882		if (conn->err == HTTP_OK
1883		    || conn->err == HTTP_NOT_MODIFIED
1884		    || conn->err == HTTP_PARTIAL
1885		    || HTTP_ERROR(conn->err))
1886			break;
1887
1888		/* all other cases: we got a redirect */
1889		e = conn->err;
1890		clean_http_auth_challenges(&server_challenges);
1891		fetch_close(conn);
1892		conn = NULL;
1893		if (!new) {
1894			DEBUG(fprintf(stderr, "redirect with no new location\n"));
1895			break;
1896		}
1897		if (url != URL)
1898			fetchFreeURL(url);
1899		url = new;
1900	} while (++i < n);
1901
1902	/* we failed, or ran out of retries */
1903	if (conn == NULL) {
1904		http_seterr(e);
1905		goto ouch;
1906	}
1907
1908	DEBUG(fprintf(stderr, "offset %lld, length %lld,"
1909		  " size %lld, clength %lld\n",
1910		  (long long)offset, (long long)length,
1911		  (long long)size, (long long)clength));
1912
1913	if (conn->err == HTTP_NOT_MODIFIED) {
1914		http_seterr(HTTP_NOT_MODIFIED);
1915		return (NULL);
1916	}
1917
1918	/* check for inconsistencies */
1919	if (clength != -1 && length != -1 && clength != length) {
1920		http_seterr(HTTP_PROTOCOL_ERROR);
1921		goto ouch;
1922	}
1923	if (clength == -1)
1924		clength = length;
1925	if (clength != -1)
1926		length = offset + clength;
1927	if (length != -1 && size != -1 && length != size) {
1928		http_seterr(HTTP_PROTOCOL_ERROR);
1929		goto ouch;
1930	}
1931	if (size == -1)
1932		size = length;
1933
1934	/* fill in stats */
1935	if (us) {
1936		us->size = size;
1937		us->atime = us->mtime = mtime;
1938	}
1939
1940	/* too far? */
1941	if (URL->offset > 0 && offset > URL->offset) {
1942		http_seterr(HTTP_PROTOCOL_ERROR);
1943		goto ouch;
1944	}
1945
1946	/* report back real offset and size */
1947	URL->offset = offset;
1948	URL->length = clength;
1949
1950	/* wrap it up in a FILE */
1951	if ((f = http_funopen(conn, chunked)) == NULL) {
1952		fetch_syserr();
1953		goto ouch;
1954	}
1955
1956	if (url != URL)
1957		fetchFreeURL(url);
1958	if (purl)
1959		fetchFreeURL(purl);
1960
1961	if (HTTP_ERROR(conn->err)) {
1962		http_print_html(stderr, f);
1963		fclose(f);
1964		f = NULL;
1965	}
1966	clean_http_headerbuf(&headerbuf);
1967	clean_http_auth_challenges(&server_challenges);
1968	clean_http_auth_challenges(&proxy_challenges);
1969	return (f);
1970
1971ouch:
1972	if (url != URL)
1973		fetchFreeURL(url);
1974	if (purl)
1975		fetchFreeURL(purl);
1976	if (conn != NULL)
1977		fetch_close(conn);
1978	clean_http_headerbuf(&headerbuf);
1979	clean_http_auth_challenges(&server_challenges);
1980	clean_http_auth_challenges(&proxy_challenges);
1981	return (NULL);
1982}
1983
1984
1985/*****************************************************************************
1986 * Entry points
1987 */
1988
1989/*
1990 * Retrieve and stat a file by HTTP
1991 */
1992FILE *
1993fetchXGetHTTP(struct url *URL, struct url_stat *us, const char *flags)
1994{
1995	return (http_request(URL, "GET", us, http_get_proxy(URL, flags), flags));
1996}
1997
1998/*
1999 * Retrieve a file by HTTP
2000 */
2001FILE *
2002fetchGetHTTP(struct url *URL, const char *flags)
2003{
2004	return (fetchXGetHTTP(URL, NULL, flags));
2005}
2006
2007/*
2008 * Store a file by HTTP
2009 */
2010FILE *
2011fetchPutHTTP(struct url *URL __unused, const char *flags __unused)
2012{
2013	warnx("fetchPutHTTP(): not implemented");
2014	return (NULL);
2015}
2016
2017/*
2018 * Get an HTTP document's metadata
2019 */
2020int
2021fetchStatHTTP(struct url *URL, struct url_stat *us, const char *flags)
2022{
2023	FILE *f;
2024
2025	f = http_request(URL, "HEAD", us, http_get_proxy(URL, flags), flags);
2026	if (f == NULL)
2027		return (-1);
2028	fclose(f);
2029	return (0);
2030}
2031
2032/*
2033 * List a directory
2034 */
2035struct url_ent *
2036fetchListHTTP(struct url *url __unused, const char *flags __unused)
2037{
2038	warnx("fetchListHTTP(): not implemented");
2039	return (NULL);
2040}
2041