xref: /freebsd/lib/libfetch/http.c (revision 61e21613)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 2000-2014 Dag-Erling Smørgrav
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer
12  *    in this position and unchanged.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. The name of the author may not be used to endorse or promote products
17  *    derived from this software without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
20  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
21  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
22  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
23  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
24  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
25  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
26  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
27  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
28  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29  */
30 
31 #include <sys/cdefs.h>
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 <stdbool.h>
75 #include <stdio.h>
76 #include <stdlib.h>
77 #include <string.h>
78 #include <time.h>
79 #include <unistd.h>
80 
81 #ifdef WITH_SSL
82 #include <openssl/md5.h>
83 #define MD5Init(c) MD5_Init(c)
84 #define MD5Update(c, data, len) MD5_Update(c, data, len)
85 #define MD5Final(md, c) MD5_Final(md, c)
86 #else
87 #include <md5.h>
88 #endif
89 
90 #include <netinet/in.h>
91 #include <netinet/tcp.h>
92 
93 #include "fetch.h"
94 #include "common.h"
95 #include "httperr.h"
96 
97 /* Maximum number of redirects to follow */
98 #define MAX_REDIRECT 20
99 
100 /* Symbolic names for reply codes we care about */
101 #define HTTP_OK			200
102 #define HTTP_PARTIAL		206
103 #define HTTP_MOVED_PERM		301
104 #define HTTP_MOVED_TEMP		302
105 #define HTTP_SEE_OTHER		303
106 #define HTTP_NOT_MODIFIED	304
107 #define HTTP_USE_PROXY		305
108 #define HTTP_TEMP_REDIRECT	307
109 #define HTTP_PERM_REDIRECT	308
110 #define HTTP_NEED_AUTH		401
111 #define HTTP_NEED_PROXY_AUTH	407
112 #define HTTP_BAD_RANGE		416
113 #define HTTP_PROTOCOL_ERROR	999
114 
115 #define HTTP_REDIRECT(xyz) ((xyz) == HTTP_MOVED_PERM \
116 			    || (xyz) == HTTP_MOVED_TEMP \
117 			    || (xyz) == HTTP_TEMP_REDIRECT \
118 			    || (xyz) == HTTP_PERM_REDIRECT \
119 			    || (xyz) == HTTP_USE_PROXY \
120 			    || (xyz) == HTTP_SEE_OTHER)
121 
122 #define HTTP_ERROR(xyz) ((xyz) >= 400 && (xyz) <= 599)
123 
124 
125 /*****************************************************************************
126  * I/O functions for decoding chunked streams
127  */
128 
129 struct httpio
130 {
131 	conn_t		*conn;		/* connection */
132 	int		 chunked;	/* chunked mode */
133 	char		*buf;		/* chunk buffer */
134 	size_t		 bufsize;	/* size of chunk buffer */
135 	size_t		 buflen;	/* amount of data currently in buffer */
136 	size_t		 bufpos;	/* current read offset in buffer */
137 	int		 eof;		/* end-of-file flag */
138 	int		 error;		/* error flag */
139 	size_t		 chunksize;	/* remaining size of current chunk */
140 #ifndef NDEBUG
141 	size_t		 total;
142 #endif
143 };
144 
145 /*
146  * Get next chunk header
147  */
148 static int
149 http_new_chunk(struct httpio *io)
150 {
151 	char *p;
152 
153 	if (fetch_getln(io->conn) == -1)
154 		return (-1);
155 
156 	if (io->conn->buflen < 2 || !isxdigit((unsigned char)*io->conn->buf))
157 		return (-1);
158 
159 	for (p = io->conn->buf; *p && !isspace((unsigned char)*p); ++p) {
160 		if (*p == ';')
161 			break;
162 		if (!isxdigit((unsigned char)*p))
163 			return (-1);
164 		if (isdigit((unsigned char)*p)) {
165 			io->chunksize = io->chunksize * 16 +
166 			    *p - '0';
167 		} else {
168 			io->chunksize = io->chunksize * 16 +
169 			    10 + tolower((unsigned char)*p) - 'a';
170 		}
171 	}
172 
173 #ifndef NDEBUG
174 	if (fetchDebug) {
175 		io->total += io->chunksize;
176 		if (io->chunksize == 0)
177 			fprintf(stderr, "%s(): end of last chunk\n", __func__);
178 		else
179 			fprintf(stderr, "%s(): new chunk: %lu (%lu)\n",
180 			    __func__, (unsigned long)io->chunksize,
181 			    (unsigned long)io->total);
182 	}
183 #endif
184 
185 	return (io->chunksize);
186 }
187 
188 /*
189  * Grow the input buffer to at least len bytes
190  */
191 static inline int
192 http_growbuf(struct httpio *io, size_t len)
193 {
194 	char *tmp;
195 
196 	if (io->bufsize >= len)
197 		return (0);
198 
199 	if ((tmp = realloc(io->buf, len)) == NULL)
200 		return (-1);
201 	io->buf = tmp;
202 	io->bufsize = len;
203 	return (0);
204 }
205 
206 /*
207  * Fill the input buffer, do chunk decoding on the fly
208  */
209 static ssize_t
210 http_fillbuf(struct httpio *io, size_t len)
211 {
212 	ssize_t nbytes;
213 	char ch;
214 
215 	if (io->error)
216 		return (-1);
217 	if (io->eof)
218 		return (0);
219 
220 	/* not chunked: just fetch the requested amount */
221 	if (io->chunked == 0) {
222 		if (http_growbuf(io, len) == -1)
223 			return (-1);
224 		if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) {
225 			io->error = errno;
226 			return (-1);
227 		}
228 		io->buflen = nbytes;
229 		io->bufpos = 0;
230 		return (io->buflen);
231 	}
232 
233 	/* chunked, but we ran out: get the next chunk header */
234 	if (io->chunksize == 0) {
235 		switch (http_new_chunk(io)) {
236 		case -1:
237 			io->error = EPROTO;
238 			return (-1);
239 		case 0:
240 			io->eof = 1;
241 			return (0);
242 		}
243 	}
244 
245 	/* fetch the requested amount, but no more than the current chunk */
246 	if (len > io->chunksize)
247 		len = io->chunksize;
248 	if (http_growbuf(io, len) == -1)
249 		return (-1);
250 	if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) {
251 		io->error = errno;
252 		return (-1);
253 	}
254 	io->bufpos = 0;
255 	io->buflen = nbytes;
256 	io->chunksize -= nbytes;
257 
258 	if (io->chunksize == 0) {
259 		if (fetch_read(io->conn, &ch, 1) != 1 || ch != '\r' ||
260 		    fetch_read(io->conn, &ch, 1) != 1 || ch != '\n')
261 			return (-1);
262 	}
263 
264 	return (io->buflen);
265 }
266 
267 /*
268  * Read function
269  */
270 static int
271 http_readfn(void *v, char *buf, int len)
272 {
273 	struct httpio *io = (struct httpio *)v;
274 	int rlen;
275 
276 	if (io->error)
277 		return (-1);
278 	if (io->eof)
279 		return (0);
280 
281 	/* empty buffer */
282 	if (!io->buf || io->bufpos == io->buflen) {
283 		if ((rlen = http_fillbuf(io, len)) < 0) {
284 			if ((errno = io->error) == EINTR)
285 				io->error = 0;
286 			return (-1);
287 		} else if (rlen == 0) {
288 			return (0);
289 		}
290 	}
291 
292 	rlen = io->buflen - io->bufpos;
293 	if (len < rlen)
294 		rlen = len;
295 	memcpy(buf, io->buf + io->bufpos, rlen);
296 	io->bufpos += rlen;
297 	return (rlen);
298 }
299 
300 /*
301  * Write function
302  */
303 static int
304 http_writefn(void *v, const char *buf, int len)
305 {
306 	struct httpio *io = (struct httpio *)v;
307 
308 	return (fetch_write(io->conn, buf, len));
309 }
310 
311 /*
312  * Close function
313  */
314 static int
315 http_closefn(void *v)
316 {
317 	struct httpio *io = (struct httpio *)v;
318 	int r;
319 
320 	r = fetch_close(io->conn);
321 	if (io->buf)
322 		free(io->buf);
323 	free(io);
324 	return (r);
325 }
326 
327 /*
328  * Wrap a file descriptor up
329  */
330 static FILE *
331 http_funopen(conn_t *conn, int chunked)
332 {
333 	struct httpio *io;
334 	FILE *f;
335 
336 	if ((io = calloc(1, sizeof(*io))) == NULL) {
337 		fetch_syserr();
338 		return (NULL);
339 	}
340 	io->conn = conn;
341 	io->chunked = chunked;
342 	f = funopen(io, http_readfn, http_writefn, NULL, http_closefn);
343 	if (f == NULL) {
344 		fetch_syserr();
345 		free(io);
346 		return (NULL);
347 	}
348 	return (f);
349 }
350 
351 
352 /*****************************************************************************
353  * Helper functions for talking to the server and parsing its replies
354  */
355 
356 /* Header types */
357 typedef enum {
358 	hdr_syserror = -2,
359 	hdr_error = -1,
360 	hdr_end = 0,
361 	hdr_unknown = 1,
362 	hdr_content_length,
363 	hdr_content_range,
364 	hdr_last_modified,
365 	hdr_location,
366 	hdr_transfer_encoding,
367 	hdr_www_authenticate,
368 	hdr_proxy_authenticate,
369 } hdr_t;
370 
371 /* Names of interesting headers */
372 static struct {
373 	hdr_t		 num;
374 	const char	*name;
375 } hdr_names[] = {
376 	{ hdr_content_length,		"Content-Length" },
377 	{ hdr_content_range,		"Content-Range" },
378 	{ hdr_last_modified,		"Last-Modified" },
379 	{ hdr_location,			"Location" },
380 	{ hdr_transfer_encoding,	"Transfer-Encoding" },
381 	{ hdr_www_authenticate,		"WWW-Authenticate" },
382 	{ hdr_proxy_authenticate,	"Proxy-Authenticate" },
383 	{ hdr_unknown,			NULL },
384 };
385 
386 /*
387  * Send a formatted line; optionally echo to terminal
388  */
389 static int
390 http_cmd(conn_t *conn, const char *fmt, ...)
391 {
392 	va_list ap;
393 	size_t len;
394 	char *msg;
395 	int r;
396 
397 	va_start(ap, fmt);
398 	len = vasprintf(&msg, fmt, ap);
399 	va_end(ap);
400 
401 	if (msg == NULL) {
402 		errno = ENOMEM;
403 		fetch_syserr();
404 		return (-1);
405 	}
406 
407 	r = fetch_putln(conn, msg, len);
408 	free(msg);
409 
410 	if (r == -1) {
411 		fetch_syserr();
412 		return (-1);
413 	}
414 
415 	return (0);
416 }
417 
418 /*
419  * Get and parse status line
420  */
421 static int
422 http_get_reply(conn_t *conn)
423 {
424 	char *p;
425 
426 	if (fetch_getln(conn) == -1)
427 		return (-1);
428 	/*
429 	 * A valid status line looks like "HTTP/m.n xyz reason" where m
430 	 * and n are the major and minor protocol version numbers and xyz
431 	 * is the reply code.
432 	 * Unfortunately, there are servers out there (NCSA 1.5.1, to name
433 	 * just one) that do not send a version number, so we can't rely
434 	 * on finding one, but if we do, insist on it being 1.0 or 1.1.
435 	 * We don't care about the reason phrase.
436 	 */
437 	if (strncmp(conn->buf, "HTTP", 4) != 0)
438 		return (HTTP_PROTOCOL_ERROR);
439 	p = conn->buf + 4;
440 	if (*p == '/') {
441 		if (p[1] != '1' || p[2] != '.' || (p[3] != '0' && p[3] != '1'))
442 			return (HTTP_PROTOCOL_ERROR);
443 		p += 4;
444 	}
445 	if (*p != ' ' ||
446 	    !isdigit((unsigned char)p[1]) ||
447 	    !isdigit((unsigned char)p[2]) ||
448 	    !isdigit((unsigned char)p[3]))
449 		return (HTTP_PROTOCOL_ERROR);
450 
451 	conn->err = (p[1] - '0') * 100 + (p[2] - '0') * 10 + (p[3] - '0');
452 	return (conn->err);
453 }
454 
455 /*
456  * Check a header; if the type matches the given string, return a pointer
457  * to the beginning of the value.
458  */
459 static const char *
460 http_match(const char *str, const char *hdr)
461 {
462 	while (*str && *hdr &&
463 	    tolower((unsigned char)*str++) == tolower((unsigned char)*hdr++))
464 		/* nothing */;
465 	if (*str || *hdr != ':')
466 		return (NULL);
467 	while (*hdr && isspace((unsigned char)*++hdr))
468 		/* nothing */;
469 	return (hdr);
470 }
471 
472 
473 /*
474  * Get the next header and return the appropriate symbolic code.  We
475  * need to read one line ahead for checking for a continuation line
476  * belonging to the current header (continuation lines start with
477  * white space).
478  *
479  * We get called with a fresh line already in the conn buffer, either
480  * from the previous http_next_header() invocation, or, the first
481  * time, from a fetch_getln() performed by our caller.
482  *
483  * This stops when we encounter an empty line (we dont read beyond the header
484  * area).
485  *
486  * Note that the "headerbuf" is just a place to return the result. Its
487  * contents are not used for the next call. This means that no cleanup
488  * is needed when ie doing another connection, just call the cleanup when
489  * fully done to deallocate memory.
490  */
491 
492 /* Limit the max number of continuation lines to some reasonable value */
493 #define HTTP_MAX_CONT_LINES 10
494 
495 /* Place into which to build a header from one or several lines */
496 typedef struct {
497 	char	*buf;		/* buffer */
498 	size_t	 bufsize;	/* buffer size */
499 	size_t	 buflen;	/* length of buffer contents */
500 } http_headerbuf_t;
501 
502 static void
503 init_http_headerbuf(http_headerbuf_t *buf)
504 {
505 	buf->buf = NULL;
506 	buf->bufsize = 0;
507 	buf->buflen = 0;
508 }
509 
510 static void
511 clean_http_headerbuf(http_headerbuf_t *buf)
512 {
513 	if (buf->buf)
514 		free(buf->buf);
515 	init_http_headerbuf(buf);
516 }
517 
518 /* Remove whitespace at the end of the buffer */
519 static void
520 http_conn_trimright(conn_t *conn)
521 {
522 	while (conn->buflen &&
523 	       isspace((unsigned char)conn->buf[conn->buflen - 1]))
524 		conn->buflen--;
525 	conn->buf[conn->buflen] = '\0';
526 }
527 
528 static hdr_t
529 http_next_header(conn_t *conn, http_headerbuf_t *hbuf, const char **p)
530 {
531 	unsigned int i, len;
532 
533 	/*
534 	 * Have to do the stripping here because of the first line. So
535 	 * it's done twice for the subsequent lines. No big deal
536 	 */
537 	http_conn_trimright(conn);
538 	if (conn->buflen == 0)
539 		return (hdr_end);
540 
541 	/* Copy the line to the headerbuf */
542 	if (hbuf->bufsize < conn->buflen + 1) {
543 		if ((hbuf->buf = realloc(hbuf->buf, conn->buflen + 1)) == NULL)
544 			return (hdr_syserror);
545 		hbuf->bufsize = conn->buflen + 1;
546 	}
547 	strcpy(hbuf->buf, conn->buf);
548 	hbuf->buflen = conn->buflen;
549 
550 	/*
551 	 * Fetch possible continuation lines. Stop at 1st non-continuation
552 	 * and leave it in the conn buffer
553 	 */
554 	for (i = 0; i < HTTP_MAX_CONT_LINES; i++) {
555 		if (fetch_getln(conn) == -1)
556 			return (hdr_syserror);
557 
558 		/*
559 		 * Note: we carry on the idea from the previous version
560 		 * that a pure whitespace line is equivalent to an empty
561 		 * one (so it's not continuation and will be handled when
562 		 * we are called next)
563 		 */
564 		http_conn_trimright(conn);
565 		if (conn->buf[0] != ' ' && conn->buf[0] != "\t"[0])
566 			break;
567 
568 		/* Got a continuation line. Concatenate to previous */
569 		len = hbuf->buflen + conn->buflen;
570 		if (hbuf->bufsize < len + 1) {
571 			len *= 2;
572 			if ((hbuf->buf = realloc(hbuf->buf, len + 1)) == NULL)
573 				return (hdr_syserror);
574 			hbuf->bufsize = len + 1;
575 		}
576 		strcpy(hbuf->buf + hbuf->buflen, conn->buf);
577 		hbuf->buflen += conn->buflen;
578 	}
579 
580 	/*
581 	 * We could check for malformed headers but we don't really care.
582 	 * A valid header starts with a token immediately followed by a
583 	 * colon; a token is any sequence of non-control, non-whitespace
584 	 * characters except "()<>@,;:\\\"{}".
585 	 */
586 	for (i = 0; hdr_names[i].num != hdr_unknown; i++)
587 		if ((*p = http_match(hdr_names[i].name, hbuf->buf)) != NULL)
588 			return (hdr_names[i].num);
589 
590 	return (hdr_unknown);
591 }
592 
593 /**************************
594  * [Proxy-]Authenticate header parsing
595  */
596 
597 /*
598  * Read doublequote-delimited string into output buffer obuf (allocated
599  * by caller, whose responsibility it is to ensure that it's big enough)
600  * cp points to the first char after the initial '"'
601  * Handles \ quoting
602  * Returns pointer to the first char after the terminating double quote, or
603  * NULL for error.
604  */
605 static const char *
606 http_parse_headerstring(const char *cp, char *obuf)
607 {
608 	for (;;) {
609 		switch (*cp) {
610 		case 0: /* Unterminated string */
611 			*obuf = 0;
612 			return (NULL);
613 		case '"': /* Ending quote */
614 			*obuf = 0;
615 			return (++cp);
616 		case '\\':
617 			if (*++cp == 0) {
618 				*obuf = 0;
619 				return (NULL);
620 			}
621 			/* FALLTHROUGH */
622 		default:
623 			*obuf++ = *cp++;
624 		}
625 	}
626 }
627 
628 /* Http auth challenge schemes */
629 typedef enum {HTTPAS_UNKNOWN, HTTPAS_BASIC,HTTPAS_DIGEST} http_auth_schemes_t;
630 
631 /* Data holder for a Basic or Digest challenge. */
632 typedef struct {
633 	http_auth_schemes_t scheme;
634 	char	*realm;
635 	char	*qop;
636 	char	*nonce;
637 	char	*opaque;
638 	char	*algo;
639 	int	 stale;
640 	int	 nc; /* Nonce count */
641 } http_auth_challenge_t;
642 
643 static void
644 init_http_auth_challenge(http_auth_challenge_t *b)
645 {
646 	b->scheme = HTTPAS_UNKNOWN;
647 	b->realm = b->qop = b->nonce = b->opaque = b->algo = NULL;
648 	b->stale = b->nc = 0;
649 }
650 
651 static void
652 clean_http_auth_challenge(http_auth_challenge_t *b)
653 {
654 	if (b->realm)
655 		free(b->realm);
656 	if (b->qop)
657 		free(b->qop);
658 	if (b->nonce)
659 		free(b->nonce);
660 	if (b->opaque)
661 		free(b->opaque);
662 	if (b->algo)
663 		free(b->algo);
664 	init_http_auth_challenge(b);
665 }
666 
667 /* Data holder for an array of challenges offered in an http response. */
668 #define MAX_CHALLENGES 10
669 typedef struct {
670 	http_auth_challenge_t *challenges[MAX_CHALLENGES];
671 	int	count; /* Number of parsed challenges in the array */
672 	int	valid; /* We did parse an authenticate header */
673 } http_auth_challenges_t;
674 
675 static void
676 init_http_auth_challenges(http_auth_challenges_t *cs)
677 {
678 	int i;
679 	for (i = 0; i < MAX_CHALLENGES; i++)
680 		cs->challenges[i] = NULL;
681 	cs->count = cs->valid = 0;
682 }
683 
684 static void
685 clean_http_auth_challenges(http_auth_challenges_t *cs)
686 {
687 	int i;
688 	/* We rely on non-zero pointers being allocated, not on the count */
689 	for (i = 0; i < MAX_CHALLENGES; i++) {
690 		if (cs->challenges[i] != NULL) {
691 			clean_http_auth_challenge(cs->challenges[i]);
692 			free(cs->challenges[i]);
693 		}
694 	}
695 	init_http_auth_challenges(cs);
696 }
697 
698 /*
699  * Enumeration for lexical elements. Separators will be returned as their own
700  * ascii value
701  */
702 typedef enum {HTTPHL_WORD=256, HTTPHL_STRING=257, HTTPHL_END=258,
703 	      HTTPHL_ERROR = 259} http_header_lex_t;
704 
705 /*
706  * Determine what kind of token comes next and return possible value
707  * in buf, which is supposed to have been allocated big enough by
708  * caller. Advance input pointer and return element type.
709  */
710 static int
711 http_header_lex(const char **cpp, char *buf)
712 {
713 	size_t l;
714 	/* Eat initial whitespace */
715 	*cpp += strspn(*cpp, " \t");
716 	if (**cpp == 0)
717 		return (HTTPHL_END);
718 
719 	/* Separator ? */
720 	if (**cpp == ',' || **cpp == '=')
721 		return (*((*cpp)++));
722 
723 	/* String ? */
724 	if (**cpp == '"') {
725 		*cpp = http_parse_headerstring(++*cpp, buf);
726 		if (*cpp == NULL)
727 			return (HTTPHL_ERROR);
728 		return (HTTPHL_STRING);
729 	}
730 
731 	/* Read other token, until separator or whitespace */
732 	l = strcspn(*cpp, " \t,=");
733 	memcpy(buf, *cpp, l);
734 	buf[l] = 0;
735 	*cpp += l;
736 	return (HTTPHL_WORD);
737 }
738 
739 /*
740  * Read challenges from http xxx-authenticate header and accumulate them
741  * in the challenges list structure.
742  *
743  * Headers with multiple challenges are specified by rfc2617, but
744  * servers (ie: squid) often send them in separate headers instead,
745  * which in turn is forbidden by the http spec (multiple headers with
746  * the same name are only allowed for pure comma-separated lists, see
747  * rfc2616 sec 4.2).
748  *
749  * We support both approaches anyway
750  */
751 static int
752 http_parse_authenticate(const char *cp, http_auth_challenges_t *cs)
753 {
754 	int ret = -1;
755 	http_header_lex_t lex;
756 	char *key = malloc(strlen(cp) + 1);
757 	char *value = malloc(strlen(cp) + 1);
758 	char *buf = malloc(strlen(cp) + 1);
759 
760 	if (key == NULL || value == NULL || buf == NULL) {
761 		fetch_syserr();
762 		goto out;
763 	}
764 
765 	/* In any case we've seen the header and we set the valid bit */
766 	cs->valid = 1;
767 
768 	/* Need word first */
769 	lex = http_header_lex(&cp, key);
770 	if (lex != HTTPHL_WORD)
771 		goto out;
772 
773 	/* Loop on challenges */
774 	for (; cs->count < MAX_CHALLENGES; cs->count++) {
775 		cs->challenges[cs->count] =
776 			malloc(sizeof(http_auth_challenge_t));
777 		if (cs->challenges[cs->count] == NULL) {
778 			fetch_syserr();
779 			goto out;
780 		}
781 		init_http_auth_challenge(cs->challenges[cs->count]);
782 		if (strcasecmp(key, "basic") == 0) {
783 			cs->challenges[cs->count]->scheme = HTTPAS_BASIC;
784 		} else if (strcasecmp(key, "digest") == 0) {
785 			cs->challenges[cs->count]->scheme = HTTPAS_DIGEST;
786 		} else {
787 			cs->challenges[cs->count]->scheme = HTTPAS_UNKNOWN;
788 			/*
789 			 * Continue parsing as basic or digest may
790 			 * follow, and the syntax is the same for
791 			 * all. We'll just ignore this one when
792 			 * looking at the list
793 			 */
794 		}
795 
796 		/* Loop on attributes */
797 		for (;;) {
798 			/* Key */
799 			lex = http_header_lex(&cp, key);
800 			if (lex != HTTPHL_WORD)
801 				goto out;
802 
803 			/* Equal sign */
804 			lex = http_header_lex(&cp, buf);
805 			if (lex != '=')
806 				goto out;
807 
808 			/* Value */
809 			lex = http_header_lex(&cp, value);
810 			if (lex != HTTPHL_WORD && lex != HTTPHL_STRING)
811 				goto out;
812 
813 			if (strcasecmp(key, "realm") == 0) {
814 				cs->challenges[cs->count]->realm =
815 				    strdup(value);
816 			} else if (strcasecmp(key, "qop") == 0) {
817 				cs->challenges[cs->count]->qop =
818 				    strdup(value);
819 			} else if (strcasecmp(key, "nonce") == 0) {
820 				cs->challenges[cs->count]->nonce =
821 				    strdup(value);
822 			} else if (strcasecmp(key, "opaque") == 0) {
823 				cs->challenges[cs->count]->opaque =
824 				    strdup(value);
825 			} else if (strcasecmp(key, "algorithm") == 0) {
826 				cs->challenges[cs->count]->algo =
827 				    strdup(value);
828 			} else if (strcasecmp(key, "stale") == 0) {
829 				cs->challenges[cs->count]->stale =
830 				    strcasecmp(value, "no");
831 			} else {
832 				/* ignore unknown attributes */
833 			}
834 
835 			/* Comma or Next challenge or End */
836 			lex = http_header_lex(&cp, key);
837 			/*
838 			 * If we get a word here, this is the beginning of the
839 			 * next challenge. Break the attributes loop
840 			 */
841 			if (lex == HTTPHL_WORD)
842 				break;
843 
844 			if (lex == HTTPHL_END) {
845 				/* End while looking for ',' is normal exit */
846 				cs->count++;
847 				ret = 0;
848 				goto out;
849 			}
850 			/* Anything else is an error */
851 			if (lex != ',')
852 				goto out;
853 
854 		} /* End attributes loop */
855 	} /* End challenge loop */
856 
857 	/*
858 	 * Challenges max count exceeded. This really can't happen
859 	 * with normal data, something's fishy -> error
860 	 */
861 
862 out:
863 	if (key)
864 		free(key);
865 	if (value)
866 		free(value);
867 	if (buf)
868 		free(buf);
869 	return (ret);
870 }
871 
872 
873 /*
874  * Parse a last-modified header
875  */
876 static int
877 http_parse_mtime(const char *p, time_t *mtime)
878 {
879 	char locale[64], *r;
880 	struct tm tm;
881 
882 	strlcpy(locale, setlocale(LC_TIME, NULL), sizeof(locale));
883 	setlocale(LC_TIME, "C");
884 	r = strptime(p, "%a, %d %b %Y %H:%M:%S GMT", &tm);
885 	/*
886 	 * Some proxies use UTC in response, but it should still be
887 	 * parsed. RFC2616 states GMT and UTC are exactly equal for HTTP.
888 	 */
889 	if (r == NULL)
890 		r = strptime(p, "%a, %d %b %Y %H:%M:%S UTC", &tm);
891 	/* XXX should add support for date-2 and date-3 */
892 	setlocale(LC_TIME, locale);
893 	if (r == NULL)
894 		return (-1);
895 	DEBUGF("last modified: [%04d-%02d-%02d %02d:%02d:%02d]\n",
896 	    tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday,
897 	    tm.tm_hour, tm.tm_min, tm.tm_sec);
898 	*mtime = timegm(&tm);
899 	return (0);
900 }
901 
902 /*
903  * Parse a content-length header
904  */
905 static int
906 http_parse_length(const char *p, off_t *length)
907 {
908 	off_t len;
909 
910 	for (len = 0; *p && isdigit((unsigned char)*p); ++p)
911 		len = len * 10 + (*p - '0');
912 	if (*p)
913 		return (-1);
914 	DEBUGF("content length: [%lld]\n", (long long)len);
915 	*length = len;
916 	return (0);
917 }
918 
919 /*
920  * Parse a content-range header
921  */
922 static int
923 http_parse_range(const char *p, off_t *offset, off_t *length, off_t *size)
924 {
925 	off_t first, last, len;
926 
927 	if (strncasecmp(p, "bytes ", 6) != 0)
928 		return (-1);
929 	p += 6;
930 	if (*p == '*') {
931 		first = last = -1;
932 		++p;
933 	} else {
934 		for (first = 0; *p && isdigit((unsigned char)*p); ++p)
935 			first = first * 10 + *p - '0';
936 		if (*p != '-')
937 			return (-1);
938 		for (last = 0, ++p; *p && isdigit((unsigned char)*p); ++p)
939 			last = last * 10 + *p - '0';
940 	}
941 	if (first > last || *p != '/')
942 		return (-1);
943 	for (len = 0, ++p; *p && isdigit((unsigned char)*p); ++p)
944 		len = len * 10 + *p - '0';
945 	if (*p || len < last - first + 1)
946 		return (-1);
947 	if (first == -1) {
948 		DEBUGF("content range: [*/%lld]\n", (long long)len);
949 		*length = 0;
950 	} else {
951 		DEBUGF("content range: [%lld-%lld/%lld]\n",
952 		    (long long)first, (long long)last, (long long)len);
953 		*length = last - first + 1;
954 	}
955 	*offset = first;
956 	*size = len;
957 	return (0);
958 }
959 
960 
961 /*****************************************************************************
962  * Helper functions for authorization
963  */
964 
965 /*
966  * Base64 encoding
967  */
968 static char *
969 http_base64(const char *src)
970 {
971 	static const char base64[] =
972 	    "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
973 	    "abcdefghijklmnopqrstuvwxyz"
974 	    "0123456789+/";
975 	char *str, *dst;
976 	size_t l;
977 	int t;
978 
979 	l = strlen(src);
980 	if ((str = malloc(((l + 2) / 3) * 4 + 1)) == NULL)
981 		return (NULL);
982 	dst = str;
983 
984 	while (l >= 3) {
985 		t = (src[0] << 16) | (src[1] << 8) | src[2];
986 		dst[0] = base64[(t >> 18) & 0x3f];
987 		dst[1] = base64[(t >> 12) & 0x3f];
988 		dst[2] = base64[(t >> 6) & 0x3f];
989 		dst[3] = base64[(t >> 0) & 0x3f];
990 		src += 3; l -= 3;
991 		dst += 4;
992 	}
993 
994 	switch (l) {
995 	case 2:
996 		t = (src[0] << 16) | (src[1] << 8);
997 		dst[0] = base64[(t >> 18) & 0x3f];
998 		dst[1] = base64[(t >> 12) & 0x3f];
999 		dst[2] = base64[(t >> 6) & 0x3f];
1000 		dst[3] = '=';
1001 		dst += 4;
1002 		break;
1003 	case 1:
1004 		t = src[0] << 16;
1005 		dst[0] = base64[(t >> 18) & 0x3f];
1006 		dst[1] = base64[(t >> 12) & 0x3f];
1007 		dst[2] = dst[3] = '=';
1008 		dst += 4;
1009 		break;
1010 	case 0:
1011 		break;
1012 	}
1013 
1014 	*dst = 0;
1015 	return (str);
1016 }
1017 
1018 
1019 /*
1020  * Extract authorization parameters from environment value.
1021  * The value is like scheme:realm:user:pass
1022  */
1023 typedef struct {
1024 	char	*scheme;
1025 	char	*realm;
1026 	char	*user;
1027 	char	*password;
1028 } http_auth_params_t;
1029 
1030 static void
1031 init_http_auth_params(http_auth_params_t *s)
1032 {
1033 	s->scheme = s->realm = s->user = s->password = NULL;
1034 }
1035 
1036 static void
1037 clean_http_auth_params(http_auth_params_t *s)
1038 {
1039 	if (s->scheme)
1040 		free(s->scheme);
1041 	if (s->realm)
1042 		free(s->realm);
1043 	if (s->user)
1044 		free(s->user);
1045 	if (s->password)
1046 		free(s->password);
1047 	init_http_auth_params(s);
1048 }
1049 
1050 static int
1051 http_authfromenv(const char *p, http_auth_params_t *parms)
1052 {
1053 	int ret = -1;
1054 	char *v, *ve;
1055 	char *str = strdup(p);
1056 
1057 	if (str == NULL) {
1058 		fetch_syserr();
1059 		return (-1);
1060 	}
1061 	v = str;
1062 
1063 	if ((ve = strchr(v, ':')) == NULL)
1064 		goto out;
1065 
1066 	*ve = 0;
1067 	if ((parms->scheme = strdup(v)) == NULL) {
1068 		fetch_syserr();
1069 		goto out;
1070 	}
1071 	v = ve + 1;
1072 
1073 	if ((ve = strchr(v, ':')) == NULL)
1074 		goto out;
1075 
1076 	*ve = 0;
1077 	if ((parms->realm = strdup(v)) == NULL) {
1078 		fetch_syserr();
1079 		goto out;
1080 	}
1081 	v = ve + 1;
1082 
1083 	if ((ve = strchr(v, ':')) == NULL)
1084 		goto out;
1085 
1086 	*ve = 0;
1087 	if ((parms->user = strdup(v)) == NULL) {
1088 		fetch_syserr();
1089 		goto out;
1090 	}
1091 	v = ve + 1;
1092 
1093 
1094 	if ((parms->password = strdup(v)) == NULL) {
1095 		fetch_syserr();
1096 		goto out;
1097 	}
1098 	ret = 0;
1099 out:
1100 	if (ret == -1)
1101 		clean_http_auth_params(parms);
1102 	if (str)
1103 		free(str);
1104 	return (ret);
1105 }
1106 
1107 
1108 /*
1109  * Digest response: the code to compute the digest is taken from the
1110  * sample implementation in RFC2616
1111  */
1112 #define IN const
1113 #define OUT
1114 
1115 #define HASHLEN 16
1116 typedef char HASH[HASHLEN];
1117 #define HASHHEXLEN 32
1118 typedef char HASHHEX[HASHHEXLEN+1];
1119 
1120 static const char *hexchars = "0123456789abcdef";
1121 static void
1122 CvtHex(IN HASH Bin, OUT HASHHEX Hex)
1123 {
1124 	unsigned short i;
1125 	unsigned char j;
1126 
1127 	for (i = 0; i < HASHLEN; i++) {
1128 		j = (Bin[i] >> 4) & 0xf;
1129 		Hex[i*2] = hexchars[j];
1130 		j = Bin[i] & 0xf;
1131 		Hex[i*2+1] = hexchars[j];
1132 	}
1133 	Hex[HASHHEXLEN] = '\0';
1134 };
1135 
1136 /* calculate H(A1) as per spec */
1137 static void
1138 DigestCalcHA1(
1139 	IN char * pszAlg,
1140 	IN char * pszUserName,
1141 	IN char * pszRealm,
1142 	IN char * pszPassword,
1143 	IN char * pszNonce,
1144 	IN char * pszCNonce,
1145 	OUT HASHHEX SessionKey
1146 	)
1147 {
1148 	MD5_CTX Md5Ctx;
1149 	HASH HA1;
1150 
1151 	MD5Init(&Md5Ctx);
1152 	MD5Update(&Md5Ctx, pszUserName, strlen(pszUserName));
1153 	MD5Update(&Md5Ctx, ":", 1);
1154 	MD5Update(&Md5Ctx, pszRealm, strlen(pszRealm));
1155 	MD5Update(&Md5Ctx, ":", 1);
1156 	MD5Update(&Md5Ctx, pszPassword, strlen(pszPassword));
1157 	MD5Final(HA1, &Md5Ctx);
1158 	if (strcasecmp(pszAlg, "md5-sess") == 0) {
1159 
1160 		MD5Init(&Md5Ctx);
1161 		MD5Update(&Md5Ctx, HA1, HASHLEN);
1162 		MD5Update(&Md5Ctx, ":", 1);
1163 		MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce));
1164 		MD5Update(&Md5Ctx, ":", 1);
1165 		MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce));
1166 		MD5Final(HA1, &Md5Ctx);
1167 	}
1168 	CvtHex(HA1, SessionKey);
1169 }
1170 
1171 /* calculate request-digest/response-digest as per HTTP Digest spec */
1172 static void
1173 DigestCalcResponse(
1174 	IN HASHHEX HA1,           /* H(A1) */
1175 	IN char * pszNonce,       /* nonce from server */
1176 	IN char * pszNonceCount,  /* 8 hex digits */
1177 	IN char * pszCNonce,      /* client nonce */
1178 	IN char * pszQop,         /* qop-value: "", "auth", "auth-int" */
1179 	IN char * pszMethod,      /* method from the request */
1180 	IN char * pszDigestUri,   /* requested URL */
1181 	IN HASHHEX HEntity,       /* H(entity body) if qop="auth-int" */
1182 	OUT HASHHEX Response      /* request-digest or response-digest */
1183 	)
1184 {
1185 #if 0
1186 	DEBUGF("Calc: HA1[%s] Nonce[%s] qop[%s] method[%s] URI[%s]\n",
1187 	    HA1, pszNonce, pszQop, pszMethod, pszDigestUri);
1188 #endif
1189 	MD5_CTX Md5Ctx;
1190 	HASH HA2;
1191 	HASH RespHash;
1192 	HASHHEX HA2Hex;
1193 
1194 	// calculate H(A2)
1195 	MD5Init(&Md5Ctx);
1196 	MD5Update(&Md5Ctx, pszMethod, strlen(pszMethod));
1197 	MD5Update(&Md5Ctx, ":", 1);
1198 	MD5Update(&Md5Ctx, pszDigestUri, strlen(pszDigestUri));
1199 	if (strcasecmp(pszQop, "auth-int") == 0) {
1200 		MD5Update(&Md5Ctx, ":", 1);
1201 		MD5Update(&Md5Ctx, HEntity, HASHHEXLEN);
1202 	}
1203 	MD5Final(HA2, &Md5Ctx);
1204 	CvtHex(HA2, HA2Hex);
1205 
1206 	// calculate response
1207 	MD5Init(&Md5Ctx);
1208 	MD5Update(&Md5Ctx, HA1, HASHHEXLEN);
1209 	MD5Update(&Md5Ctx, ":", 1);
1210 	MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce));
1211 	MD5Update(&Md5Ctx, ":", 1);
1212 	if (*pszQop) {
1213 		MD5Update(&Md5Ctx, pszNonceCount, strlen(pszNonceCount));
1214 		MD5Update(&Md5Ctx, ":", 1);
1215 		MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce));
1216 		MD5Update(&Md5Ctx, ":", 1);
1217 		MD5Update(&Md5Ctx, pszQop, strlen(pszQop));
1218 		MD5Update(&Md5Ctx, ":", 1);
1219 	}
1220 	MD5Update(&Md5Ctx, HA2Hex, HASHHEXLEN);
1221 	MD5Final(RespHash, &Md5Ctx);
1222 	CvtHex(RespHash, Response);
1223 }
1224 
1225 /*
1226  * Generate/Send a Digest authorization header
1227  * This looks like: [Proxy-]Authorization: credentials
1228  *
1229  *  credentials      = "Digest" digest-response
1230  *  digest-response  = 1#( username | realm | nonce | digest-uri
1231  *                      | response | [ algorithm ] | [cnonce] |
1232  *                      [opaque] | [message-qop] |
1233  *                          [nonce-count]  | [auth-param] )
1234  *  username         = "username" "=" username-value
1235  *  username-value   = quoted-string
1236  *  digest-uri       = "uri" "=" digest-uri-value
1237  *  digest-uri-value = request-uri   ; As specified by HTTP/1.1
1238  *  message-qop      = "qop" "=" qop-value
1239  *  cnonce           = "cnonce" "=" cnonce-value
1240  *  cnonce-value     = nonce-value
1241  *  nonce-count      = "nc" "=" nc-value
1242  *  nc-value         = 8LHEX
1243  *  response         = "response" "=" request-digest
1244  *  request-digest = <"> 32LHEX <">
1245  */
1246 static int
1247 http_digest_auth(conn_t *conn, const char *hdr, http_auth_challenge_t *c,
1248 		 http_auth_params_t *parms, struct url *url)
1249 {
1250 	int r;
1251 	char noncecount[10];
1252 	char cnonce[40];
1253 	char *options = NULL;
1254 
1255 	if (!c->realm || !c->nonce) {
1256 		DEBUGF("realm/nonce not set in challenge\n");
1257 		return(-1);
1258 	}
1259 	if (!c->algo)
1260 		c->algo = strdup("");
1261 
1262 	if (asprintf(&options, "%s%s%s%s",
1263 	    *c->algo? ",algorithm=" : "", c->algo,
1264 	    c->opaque? ",opaque=" : "", c->opaque?c->opaque:"") < 0)
1265 		return (-1);
1266 
1267 	if (!c->qop) {
1268 		c->qop = strdup("");
1269 		*noncecount = 0;
1270 		*cnonce = 0;
1271 	} else {
1272 		c->nc++;
1273 		sprintf(noncecount, "%08x", c->nc);
1274 		/* We don't try very hard with the cnonce ... */
1275 		sprintf(cnonce, "%x%lx", getpid(), (unsigned long)time(0));
1276 	}
1277 
1278 	HASHHEX HA1;
1279 	DigestCalcHA1(c->algo, parms->user, c->realm,
1280 		      parms->password, c->nonce, cnonce, HA1);
1281 	DEBUGF("HA1: [%s]\n", HA1);
1282 	HASHHEX digest, null;
1283 	memset(null, 0, sizeof(null));
1284 	DigestCalcResponse(HA1, c->nonce, noncecount, cnonce, c->qop,
1285 			   "GET", url->doc, null, digest);
1286 
1287 	if (c->qop[0]) {
1288 		r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\","
1289 			     "nonce=\"%s\",uri=\"%s\",response=\"%s\","
1290 			     "qop=\"auth\", cnonce=\"%s\", nc=%s%s",
1291 			     hdr, parms->user, c->realm,
1292 			     c->nonce, url->doc, digest,
1293 			     cnonce, noncecount, options);
1294 	} else {
1295 		r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\","
1296 			     "nonce=\"%s\",uri=\"%s\",response=\"%s\"%s",
1297 			     hdr, parms->user, c->realm,
1298 			     c->nonce, url->doc, digest, options);
1299 	}
1300 	if (options)
1301 		free(options);
1302 	return (r);
1303 }
1304 
1305 /*
1306  * Encode username and password
1307  */
1308 static int
1309 http_basic_auth(conn_t *conn, const char *hdr, const char *usr, const char *pwd)
1310 {
1311 	char *upw, *auth;
1312 	int r;
1313 
1314 	DEBUGF("basic: usr: [%s]\n", usr);
1315 	DEBUGF("basic: pwd: [%s]\n", pwd);
1316 	if (asprintf(&upw, "%s:%s", usr, pwd) == -1)
1317 		return (-1);
1318 	auth = http_base64(upw);
1319 	free(upw);
1320 	if (auth == NULL)
1321 		return (-1);
1322 	r = http_cmd(conn, "%s: Basic %s", hdr, auth);
1323 	free(auth);
1324 	return (r);
1325 }
1326 
1327 /*
1328  * Chose the challenge to answer and call the appropriate routine to
1329  * produce the header.
1330  */
1331 static int
1332 http_authorize(conn_t *conn, const char *hdr, http_auth_challenges_t *cs,
1333 	       http_auth_params_t *parms, struct url *url)
1334 {
1335 	http_auth_challenge_t *digest = NULL;
1336 	int i;
1337 
1338 	/* If user or pass are null we're not happy */
1339 	if (!parms->user || !parms->password) {
1340 		DEBUGF("NULL usr or pass\n");
1341 		return (-1);
1342 	}
1343 
1344 	/* Look for a Digest */
1345 	for (i = 0; i < cs->count; i++) {
1346 		if (cs->challenges[i]->scheme == HTTPAS_DIGEST)
1347 			digest = cs->challenges[i];
1348 	}
1349 
1350 	/* Error if "Digest" was specified and there is no Digest challenge */
1351 	if (!digest &&
1352 	    (parms->scheme && strcasecmp(parms->scheme, "digest") == 0)) {
1353 		DEBUGF("Digest auth in env, not supported by peer\n");
1354 		return (-1);
1355 	}
1356 	/*
1357 	 * If "basic" was specified in the environment, or there is no Digest
1358 	 * challenge, do the basic thing. Don't need a challenge for this,
1359 	 * so no need to check basic!=NULL
1360 	 */
1361 	if (!digest ||
1362 	    (parms->scheme && strcasecmp(parms->scheme, "basic") == 0))
1363 		return (http_basic_auth(conn,hdr,parms->user,parms->password));
1364 
1365 	/* Else, prefer digest. We just checked that it's not NULL */
1366 	return (http_digest_auth(conn, hdr, digest, parms, url));
1367 }
1368 
1369 /*****************************************************************************
1370  * Helper functions for connecting to a server or proxy
1371  */
1372 
1373 /*
1374  * Connect to the correct HTTP server or proxy.
1375  */
1376 static conn_t *
1377 http_connect(struct url *URL, struct url *purl, const char *flags)
1378 {
1379 	struct url *curl;
1380 	conn_t *conn;
1381 	hdr_t h;
1382 	http_headerbuf_t headerbuf;
1383 	const char *p;
1384 	int verbose;
1385 	int af, val;
1386 	int serrno;
1387 	bool isproxyauth = false;
1388 	http_auth_challenges_t proxy_challenges;
1389 
1390 #ifdef INET6
1391 	af = AF_UNSPEC;
1392 #else
1393 	af = AF_INET;
1394 #endif
1395 
1396 	verbose = CHECK_FLAG('v');
1397 	if (CHECK_FLAG('4'))
1398 		af = AF_INET;
1399 #ifdef INET6
1400 	else if (CHECK_FLAG('6'))
1401 		af = AF_INET6;
1402 #endif
1403 
1404 	curl = (purl != NULL) ? purl : URL;
1405 
1406 retry:
1407 	if ((conn = fetch_connect(curl->host, curl->port, af, verbose)) == NULL)
1408 		/* fetch_connect() has already set an error code */
1409 		return (NULL);
1410 	init_http_headerbuf(&headerbuf);
1411 	if (strcmp(URL->scheme, SCHEME_HTTPS) == 0 && purl) {
1412 		init_http_auth_challenges(&proxy_challenges);
1413 		http_cmd(conn, "CONNECT %s:%d HTTP/1.1", URL->host, URL->port);
1414 		http_cmd(conn, "Host: %s:%d", URL->host, URL->port);
1415 		if (isproxyauth) {
1416 			http_auth_params_t aparams;
1417 			init_http_auth_params(&aparams);
1418 			if (*purl->user || *purl->pwd) {
1419 				aparams.user = strdup(purl->user);
1420 				aparams.password = strdup(purl->pwd);
1421 			} else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL &&
1422 				    *p != '\0') {
1423 				if (http_authfromenv(p, &aparams) < 0) {
1424 					http_seterr(HTTP_NEED_PROXY_AUTH);
1425 					fetch_syserr();
1426 					goto ouch;
1427 				}
1428 			} else if (fetch_netrc_auth(purl) == 0) {
1429 				aparams.user = strdup(purl->user);
1430 				aparams.password = strdup(purl->pwd);
1431 			} else {
1432 				/*
1433 				 * No auth information found in system - exiting
1434 				 * with warning.
1435 				 */
1436 				warnx("Missing username and/or password set");
1437 				fetch_syserr();
1438 				goto ouch;
1439 			}
1440 			http_authorize(conn, "Proxy-Authorization",
1441 			    &proxy_challenges, &aparams, purl);
1442 			clean_http_auth_params(&aparams);
1443 		}
1444 		http_cmd(conn, "");
1445 		/* Get reply from CONNECT Tunnel attempt */
1446 		int httpreply = http_get_reply(conn);
1447 		if (httpreply != HTTP_OK) {
1448 			http_seterr(httpreply);
1449 			/* If the error is a 407/HTTP_NEED_PROXY_AUTH */
1450 			if (httpreply == HTTP_NEED_PROXY_AUTH &&
1451 			    ! isproxyauth) {
1452 				/* Try again with authentication. */
1453 				clean_http_headerbuf(&headerbuf);
1454 				fetch_close(conn);
1455 				isproxyauth = true;
1456 				goto retry;
1457 			}
1458 			goto ouch;
1459 		}
1460 		/* Read and discard the rest of the proxy response */
1461 		if (fetch_getln(conn) < 0) {
1462 			fetch_syserr();
1463 			goto ouch;
1464 		}
1465 		do {
1466 			switch ((h = http_next_header(conn, &headerbuf, &p))) {
1467 			case hdr_syserror:
1468 				fetch_syserr();
1469 				goto ouch;
1470 			case hdr_error:
1471 				http_seterr(HTTP_PROTOCOL_ERROR);
1472 				goto ouch;
1473 			default:
1474 				/* ignore */ ;
1475 			}
1476 		} while (h > hdr_end);
1477 	}
1478 	if (strcmp(URL->scheme, SCHEME_HTTPS) == 0 &&
1479 	    fetch_ssl(conn, URL, verbose) == -1) {
1480 		/* grrr */
1481 		errno = EAUTH;
1482 		fetch_syserr();
1483 		goto ouch;
1484 	}
1485 
1486 	val = 1;
1487 	setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, sizeof(val));
1488 
1489 	clean_http_headerbuf(&headerbuf);
1490 	return (conn);
1491 ouch:
1492 	serrno = errno;
1493 	clean_http_headerbuf(&headerbuf);
1494 	fetch_close(conn);
1495 	errno = serrno;
1496 	return (NULL);
1497 }
1498 
1499 static struct url *
1500 http_get_proxy(struct url * url, const char *flags)
1501 {
1502 	struct url *purl;
1503 	char *p;
1504 
1505 	if (flags != NULL && strchr(flags, 'd') != NULL)
1506 		return (NULL);
1507 	if (fetch_no_proxy_match(url->host))
1508 		return (NULL);
1509 	if (((p = getenv("HTTP_PROXY")) || (p = getenv("http_proxy"))) &&
1510 	    *p && (purl = fetchParseURL(p))) {
1511 		if (!*purl->scheme)
1512 			strcpy(purl->scheme, SCHEME_HTTP);
1513 		if (!purl->port)
1514 			purl->port = fetch_default_proxy_port(purl->scheme);
1515 		if (strcmp(purl->scheme, SCHEME_HTTP) == 0)
1516 			return (purl);
1517 		fetchFreeURL(purl);
1518 	}
1519 	return (NULL);
1520 }
1521 
1522 static void
1523 http_print_html(FILE *out, FILE *in)
1524 {
1525 	ssize_t len = 0;
1526 	size_t cap;
1527 	char *line = NULL, *p, *q;
1528 	int comment, tag;
1529 
1530 	comment = tag = 0;
1531 	while ((len = getline(&line, &cap, in)) >= 0) {
1532 		while (len && isspace((unsigned char)line[len - 1]))
1533 			--len;
1534 		for (p = q = line; q < line + len; ++q) {
1535 			if (comment && *q == '-') {
1536 				if (q + 2 < line + len &&
1537 				    strcmp(q, "-->") == 0) {
1538 					tag = comment = 0;
1539 					q += 2;
1540 				}
1541 			} else if (tag && !comment && *q == '>') {
1542 				p = q + 1;
1543 				tag = 0;
1544 			} else if (!tag && *q == '<') {
1545 				if (q > p)
1546 					fwrite(p, q - p, 1, out);
1547 				tag = 1;
1548 				if (q + 3 < line + len &&
1549 				    strcmp(q, "<!--") == 0) {
1550 					comment = 1;
1551 					q += 3;
1552 				}
1553 			}
1554 		}
1555 		if (!tag && q > p)
1556 			fwrite(p, q - p, 1, out);
1557 		fputc('\n', out);
1558 	}
1559 
1560 	free(line);
1561 }
1562 
1563 
1564 /*****************************************************************************
1565  * Core
1566  */
1567 
1568 FILE *
1569 http_request(struct url *URL, const char *op, struct url_stat *us,
1570 	struct url *purl, const char *flags)
1571 {
1572 
1573 	return (http_request_body(URL, op, us, purl, flags, NULL, NULL));
1574 }
1575 
1576 /*
1577  * Send a request and process the reply
1578  *
1579  * XXX This function is way too long, the do..while loop should be split
1580  * XXX off into a separate function.
1581  */
1582 FILE *
1583 http_request_body(struct url *URL, const char *op, struct url_stat *us,
1584 	struct url *purl, const char *flags, const char *content_type,
1585 	const char *body)
1586 {
1587 	char timebuf[80];
1588 	char hbuf[MAXHOSTNAMELEN + 7], *host;
1589 	conn_t *conn;
1590 	struct url *url, *new;
1591 	int chunked, direct, ims, noredirect, verbose;
1592 	int e, i, n, val;
1593 	off_t offset, clength, length, size;
1594 	time_t mtime;
1595 	const char *p;
1596 	FILE *f;
1597 	hdr_t h;
1598 	struct tm *timestruct;
1599 	http_headerbuf_t headerbuf;
1600 	http_auth_challenges_t server_challenges;
1601 	http_auth_challenges_t proxy_challenges;
1602 	size_t body_len;
1603 
1604 	/* The following calls don't allocate anything */
1605 	init_http_headerbuf(&headerbuf);
1606 	init_http_auth_challenges(&server_challenges);
1607 	init_http_auth_challenges(&proxy_challenges);
1608 
1609 	direct = CHECK_FLAG('d');
1610 	noredirect = CHECK_FLAG('A');
1611 	verbose = CHECK_FLAG('v');
1612 	ims = CHECK_FLAG('i');
1613 
1614 	if (direct && purl) {
1615 		fetchFreeURL(purl);
1616 		purl = NULL;
1617 	}
1618 
1619 	/* try the provided URL first */
1620 	url = URL;
1621 
1622 	n = MAX_REDIRECT;
1623 	i = 0;
1624 
1625 	e = HTTP_PROTOCOL_ERROR;
1626 	do {
1627 		new = NULL;
1628 		chunked = 0;
1629 		offset = 0;
1630 		clength = -1;
1631 		length = -1;
1632 		size = -1;
1633 		mtime = 0;
1634 
1635 		/* check port */
1636 		if (!url->port)
1637 			url->port = fetch_default_port(url->scheme);
1638 
1639 		/* were we redirected to an FTP URL? */
1640 		if (purl == NULL && strcmp(url->scheme, SCHEME_FTP) == 0) {
1641 			if (strcmp(op, "GET") == 0)
1642 				return (ftp_request(url, "RETR", us, purl, flags));
1643 			else if (strcmp(op, "HEAD") == 0)
1644 				return (ftp_request(url, "STAT", us, purl, flags));
1645 		}
1646 
1647 		/* connect to server or proxy */
1648 		if ((conn = http_connect(url, purl, flags)) == NULL)
1649 			goto ouch;
1650 
1651 		/* append port number only if necessary */
1652 		host = url->host;
1653 		if (url->port != fetch_default_port(url->scheme)) {
1654 			snprintf(hbuf, sizeof(hbuf), "%s:%d", host, url->port);
1655 			host = hbuf;
1656 		}
1657 
1658 		/* send request */
1659 		if (verbose)
1660 			fetch_info("requesting %s://%s%s",
1661 			    url->scheme, host, url->doc);
1662 		if (purl && strcmp(url->scheme, SCHEME_HTTPS) != 0) {
1663 			http_cmd(conn, "%s %s://%s%s HTTP/1.1",
1664 			    op, url->scheme, host, url->doc);
1665 		} else {
1666 			http_cmd(conn, "%s %s HTTP/1.1",
1667 			    op, url->doc);
1668 		}
1669 
1670 		if (ims && url->ims_time) {
1671 			timestruct = gmtime((time_t *)&url->ims_time);
1672 			(void)strftime(timebuf, 80, "%a, %d %b %Y %T GMT",
1673 			    timestruct);
1674 			if (verbose)
1675 				fetch_info("If-Modified-Since: %s", timebuf);
1676 			http_cmd(conn, "If-Modified-Since: %s", timebuf);
1677 		}
1678 		/* virtual host */
1679 		http_cmd(conn, "Host: %s", host);
1680 
1681 		/*
1682 		 * Proxy authorization: we only send auth after we received
1683 		 * a 407 error. We do not first try basic anyway (changed
1684 		 * when support was added for digest-auth)
1685 		 */
1686 		if (purl && proxy_challenges.valid) {
1687 			http_auth_params_t aparams;
1688 			init_http_auth_params(&aparams);
1689 			if (*purl->user || *purl->pwd) {
1690 				aparams.user = strdup(purl->user);
1691 				aparams.password = strdup(purl->pwd);
1692 			} else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL &&
1693 				   *p != '\0') {
1694 				if (http_authfromenv(p, &aparams) < 0) {
1695 					http_seterr(HTTP_NEED_PROXY_AUTH);
1696 					goto ouch;
1697 				}
1698 			} else if (fetch_netrc_auth(purl) == 0) {
1699 				aparams.user = strdup(purl->user);
1700 				aparams.password = strdup(purl->pwd);
1701 			}
1702 			http_authorize(conn, "Proxy-Authorization",
1703 				       &proxy_challenges, &aparams, url);
1704 			clean_http_auth_params(&aparams);
1705 		}
1706 
1707 		/*
1708 		 * Server authorization: we never send "a priori"
1709 		 * Basic auth, which used to be done if user/pass were
1710 		 * set in the url. This would be weird because we'd send the
1711 		 * password in the clear even if Digest is finally to be
1712 		 * used (it would have made more sense for the
1713 		 * pre-digest version to do this when Basic was specified
1714 		 * in the environment)
1715 		 */
1716 		if (server_challenges.valid) {
1717 			http_auth_params_t aparams;
1718 			init_http_auth_params(&aparams);
1719 			if (*url->user || *url->pwd) {
1720 				aparams.user = strdup(url->user);
1721 				aparams.password = strdup(url->pwd);
1722 			} else if ((p = getenv("HTTP_AUTH")) != NULL &&
1723 				   *p != '\0') {
1724 				if (http_authfromenv(p, &aparams) < 0) {
1725 					http_seterr(HTTP_NEED_AUTH);
1726 					goto ouch;
1727 				}
1728 			} else if (fetch_netrc_auth(url) == 0) {
1729 				aparams.user = strdup(url->user);
1730 				aparams.password = strdup(url->pwd);
1731 			} else if (fetchAuthMethod &&
1732 				   fetchAuthMethod(url) == 0) {
1733 				aparams.user = strdup(url->user);
1734 				aparams.password = strdup(url->pwd);
1735 			} else {
1736 				http_seterr(HTTP_NEED_AUTH);
1737 				goto ouch;
1738 			}
1739 			http_authorize(conn, "Authorization",
1740 				       &server_challenges, &aparams, url);
1741 			clean_http_auth_params(&aparams);
1742 		}
1743 
1744 		/* other headers */
1745 		if ((p = getenv("HTTP_ACCEPT")) != NULL) {
1746 			if (*p != '\0')
1747 				http_cmd(conn, "Accept: %s", p);
1748 		} else {
1749 			http_cmd(conn, "Accept: */*");
1750 		}
1751 		if ((p = getenv("HTTP_REFERER")) != NULL && *p != '\0') {
1752 			if (strcasecmp(p, "auto") == 0)
1753 				http_cmd(conn, "Referer: %s://%s%s",
1754 				    url->scheme, host, url->doc);
1755 			else
1756 				http_cmd(conn, "Referer: %s", p);
1757 		}
1758 		if ((p = getenv("HTTP_USER_AGENT")) != NULL) {
1759 			/* no User-Agent if defined but empty */
1760 			if  (*p != '\0')
1761 				http_cmd(conn, "User-Agent: %s", p);
1762 		} else {
1763 			/* default User-Agent */
1764 			http_cmd(conn, "User-Agent: %s " _LIBFETCH_VER,
1765 			    getprogname());
1766 		}
1767 		if (url->offset > 0)
1768 			http_cmd(conn, "Range: bytes=%lld-", (long long)url->offset);
1769 		http_cmd(conn, "Connection: close");
1770 
1771 		if (body) {
1772 			body_len = strlen(body);
1773 			http_cmd(conn, "Content-Length: %zu", body_len);
1774 			if (content_type != NULL)
1775 				http_cmd(conn, "Content-Type: %s", content_type);
1776 		}
1777 
1778 		http_cmd(conn, "");
1779 
1780 		if (body)
1781 			fetch_write(conn, body, body_len);
1782 
1783 		/*
1784 		 * Force the queued request to be dispatched.  Normally, one
1785 		 * would do this with shutdown(2) but squid proxies can be
1786 		 * configured to disallow such half-closed connections.  To
1787 		 * be compatible with such configurations, fiddle with socket
1788 		 * options to force the pending data to be written.
1789 		 */
1790 		val = 0;
1791 		setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val,
1792 			   sizeof(val));
1793 		val = 1;
1794 		setsockopt(conn->sd, IPPROTO_TCP, TCP_NODELAY, &val,
1795 			   sizeof(val));
1796 
1797 		/* get reply */
1798 		switch (http_get_reply(conn)) {
1799 		case HTTP_OK:
1800 		case HTTP_PARTIAL:
1801 		case HTTP_NOT_MODIFIED:
1802 			/* fine */
1803 			break;
1804 		case HTTP_MOVED_PERM:
1805 		case HTTP_MOVED_TEMP:
1806 		case HTTP_TEMP_REDIRECT:
1807 		case HTTP_PERM_REDIRECT:
1808 		case HTTP_SEE_OTHER:
1809 		case HTTP_USE_PROXY:
1810 			/*
1811 			 * Not so fine, but we still have to read the
1812 			 * headers to get the new location.
1813 			 */
1814 			break;
1815 		case HTTP_NEED_AUTH:
1816 			if (server_challenges.valid) {
1817 				/*
1818 				 * We already sent out authorization code,
1819 				 * so there's nothing more we can do.
1820 				 */
1821 				http_seterr(conn->err);
1822 				goto ouch;
1823 			}
1824 			/* try again, but send the password this time */
1825 			if (verbose)
1826 				fetch_info("server requires authorization");
1827 			break;
1828 		case HTTP_NEED_PROXY_AUTH:
1829 			if (proxy_challenges.valid) {
1830 				/*
1831 				 * We already sent our proxy
1832 				 * authorization code, so there's
1833 				 * nothing more we can do. */
1834 				http_seterr(conn->err);
1835 				goto ouch;
1836 			}
1837 			/* try again, but send the password this time */
1838 			if (verbose)
1839 				fetch_info("proxy requires authorization");
1840 			break;
1841 		case HTTP_BAD_RANGE:
1842 			/*
1843 			 * This can happen if we ask for 0 bytes because
1844 			 * we already have the whole file.  Consider this
1845 			 * a success for now, and check sizes later.
1846 			 */
1847 			break;
1848 		case HTTP_PROTOCOL_ERROR:
1849 			/* fall through */
1850 		case -1:
1851 			fetch_syserr();
1852 			goto ouch;
1853 		default:
1854 			http_seterr(conn->err);
1855 			if (!verbose)
1856 				goto ouch;
1857 			/* fall through so we can get the full error message */
1858 		}
1859 
1860 		/* get headers. http_next_header expects one line readahead */
1861 		if (fetch_getln(conn) == -1) {
1862 			fetch_syserr();
1863 			goto ouch;
1864 		}
1865 		do {
1866 			switch ((h = http_next_header(conn, &headerbuf, &p))) {
1867 			case hdr_syserror:
1868 				fetch_syserr();
1869 				goto ouch;
1870 			case hdr_error:
1871 				http_seterr(HTTP_PROTOCOL_ERROR);
1872 				goto ouch;
1873 			case hdr_content_length:
1874 				http_parse_length(p, &clength);
1875 				break;
1876 			case hdr_content_range:
1877 				http_parse_range(p, &offset, &length, &size);
1878 				break;
1879 			case hdr_last_modified:
1880 				http_parse_mtime(p, &mtime);
1881 				break;
1882 			case hdr_location:
1883 				if (!HTTP_REDIRECT(conn->err))
1884 					break;
1885 				/*
1886 				 * if the A flag is set, we don't follow
1887 				 * temporary redirects.
1888 				 */
1889 				if (noredirect &&
1890 				    conn->err != HTTP_MOVED_PERM &&
1891 				    conn->err != HTTP_PERM_REDIRECT &&
1892 				    conn->err != HTTP_USE_PROXY) {
1893 					n = 1;
1894 					break;
1895 				}
1896 				if (new)
1897 					free(new);
1898 				if (verbose)
1899 					fetch_info("%d redirect to %s",
1900 					    conn->err, p);
1901 				if (*p == '/')
1902 					/* absolute path */
1903 					new = fetchMakeURL(url->scheme, url->host,
1904 					    url->port, p, url->user, url->pwd);
1905 				else
1906 					new = fetchParseURL(p);
1907 				if (new == NULL) {
1908 					/* XXX should set an error code */
1909 					DEBUGF("failed to parse new URL\n");
1910 					goto ouch;
1911 				}
1912 
1913 				/* Only copy credentials if the host matches */
1914 				if (strcmp(new->host, url->host) == 0 &&
1915 				    !*new->user && !*new->pwd) {
1916 					strcpy(new->user, url->user);
1917 					strcpy(new->pwd, url->pwd);
1918 				}
1919 				new->offset = url->offset;
1920 				new->length = url->length;
1921 				new->ims_time = url->ims_time;
1922 				break;
1923 			case hdr_transfer_encoding:
1924 				/* XXX weak test*/
1925 				chunked = (strcasecmp(p, "chunked") == 0);
1926 				break;
1927 			case hdr_www_authenticate:
1928 				if (conn->err != HTTP_NEED_AUTH)
1929 					break;
1930 				if (http_parse_authenticate(p, &server_challenges) == 0)
1931 					++n;
1932 				break;
1933 			case hdr_proxy_authenticate:
1934 				if (conn->err != HTTP_NEED_PROXY_AUTH)
1935 					break;
1936 				if (http_parse_authenticate(p, &proxy_challenges) == 0)
1937 					++n;
1938 				break;
1939 			case hdr_end:
1940 				/* fall through */
1941 			case hdr_unknown:
1942 				/* ignore */
1943 				break;
1944 			}
1945 		} while (h > hdr_end);
1946 
1947 		/* we need to provide authentication */
1948 		if (conn->err == HTTP_NEED_AUTH ||
1949 		    conn->err == HTTP_NEED_PROXY_AUTH) {
1950 			e = conn->err;
1951 			if ((conn->err == HTTP_NEED_AUTH &&
1952 			     !server_challenges.valid) ||
1953 			    (conn->err == HTTP_NEED_PROXY_AUTH &&
1954 			     !proxy_challenges.valid)) {
1955 				/* 401/7 but no www/proxy-authenticate ?? */
1956 				DEBUGF("%03d without auth header\n", conn->err);
1957 				goto ouch;
1958 			}
1959 			fetch_close(conn);
1960 			conn = NULL;
1961 			continue;
1962 		}
1963 
1964 		/* requested range not satisfiable */
1965 		if (conn->err == HTTP_BAD_RANGE) {
1966 			if (url->offset > 0 && url->length == 0) {
1967 				/* asked for 0 bytes; fake it */
1968 				offset = url->offset;
1969 				clength = -1;
1970 				conn->err = HTTP_OK;
1971 				break;
1972 			} else {
1973 				http_seterr(conn->err);
1974 				goto ouch;
1975 			}
1976 		}
1977 
1978 		/* we have a hit or an error */
1979 		if (conn->err == HTTP_OK
1980 		    || conn->err == HTTP_NOT_MODIFIED
1981 		    || conn->err == HTTP_PARTIAL
1982 		    || HTTP_ERROR(conn->err))
1983 			break;
1984 
1985 		/* all other cases: we got a redirect */
1986 		e = conn->err;
1987 		clean_http_auth_challenges(&server_challenges);
1988 		fetch_close(conn);
1989 		conn = NULL;
1990 		if (!new) {
1991 			DEBUGF("redirect with no new location\n");
1992 			break;
1993 		}
1994 		if (url != URL)
1995 			fetchFreeURL(url);
1996 		url = new;
1997 	} while (++i < n);
1998 
1999 	/* we failed, or ran out of retries */
2000 	if (conn == NULL) {
2001 		http_seterr(e);
2002 		goto ouch;
2003 	}
2004 
2005 	DEBUGF("offset %lld, length %lld, size %lld, clength %lld\n",
2006 	    (long long)offset, (long long)length,
2007 	    (long long)size, (long long)clength);
2008 
2009 	if (conn->err == HTTP_NOT_MODIFIED) {
2010 		http_seterr(HTTP_NOT_MODIFIED);
2011 		return (NULL);
2012 	}
2013 
2014 	/* check for inconsistencies */
2015 	if (clength != -1 && length != -1 && clength != length) {
2016 		http_seterr(HTTP_PROTOCOL_ERROR);
2017 		goto ouch;
2018 	}
2019 	if (clength == -1)
2020 		clength = length;
2021 	if (clength != -1)
2022 		length = offset + clength;
2023 	if (length != -1 && size != -1 && length != size) {
2024 		http_seterr(HTTP_PROTOCOL_ERROR);
2025 		goto ouch;
2026 	}
2027 	if (size == -1)
2028 		size = length;
2029 
2030 	/* fill in stats */
2031 	if (us) {
2032 		us->size = size;
2033 		us->atime = us->mtime = mtime;
2034 	}
2035 
2036 	/* too far? */
2037 	if (URL->offset > 0 && offset > URL->offset) {
2038 		http_seterr(HTTP_PROTOCOL_ERROR);
2039 		goto ouch;
2040 	}
2041 
2042 	/* report back real offset and size */
2043 	URL->offset = offset;
2044 	URL->length = clength;
2045 
2046 	/* wrap it up in a FILE */
2047 	if ((f = http_funopen(conn, chunked)) == NULL) {
2048 		fetch_syserr();
2049 		goto ouch;
2050 	}
2051 
2052 	if (url != URL)
2053 		fetchFreeURL(url);
2054 	if (purl)
2055 		fetchFreeURL(purl);
2056 
2057 	if (HTTP_ERROR(conn->err)) {
2058 		http_print_html(stderr, f);
2059 		fclose(f);
2060 		f = NULL;
2061 	}
2062 	clean_http_headerbuf(&headerbuf);
2063 	clean_http_auth_challenges(&server_challenges);
2064 	clean_http_auth_challenges(&proxy_challenges);
2065 	return (f);
2066 
2067 ouch:
2068 	if (url != URL)
2069 		fetchFreeURL(url);
2070 	if (purl)
2071 		fetchFreeURL(purl);
2072 	if (conn != NULL)
2073 		fetch_close(conn);
2074 	clean_http_headerbuf(&headerbuf);
2075 	clean_http_auth_challenges(&server_challenges);
2076 	clean_http_auth_challenges(&proxy_challenges);
2077 	return (NULL);
2078 }
2079 
2080 
2081 /*****************************************************************************
2082  * Entry points
2083  */
2084 
2085 /*
2086  * Retrieve and stat a file by HTTP
2087  */
2088 FILE *
2089 fetchXGetHTTP(struct url *URL, struct url_stat *us, const char *flags)
2090 {
2091 	return (http_request(URL, "GET", us, http_get_proxy(URL, flags), flags));
2092 }
2093 
2094 /*
2095  * Retrieve a file by HTTP
2096  */
2097 FILE *
2098 fetchGetHTTP(struct url *URL, const char *flags)
2099 {
2100 	return (fetchXGetHTTP(URL, NULL, flags));
2101 }
2102 
2103 /*
2104  * Store a file by HTTP
2105  */
2106 FILE *
2107 fetchPutHTTP(struct url *URL __unused, const char *flags __unused)
2108 {
2109 	warnx("fetchPutHTTP(): not implemented");
2110 	return (NULL);
2111 }
2112 
2113 /*
2114  * Get an HTTP document's metadata
2115  */
2116 int
2117 fetchStatHTTP(struct url *URL, struct url_stat *us, const char *flags)
2118 {
2119 	FILE *f;
2120 
2121 	f = http_request(URL, "HEAD", us, http_get_proxy(URL, flags), flags);
2122 	if (f == NULL)
2123 		return (-1);
2124 	fclose(f);
2125 	return (0);
2126 }
2127 
2128 /*
2129  * List a directory
2130  */
2131 struct url_ent *
2132 fetchListHTTP(struct url *url __unused, const char *flags __unused)
2133 {
2134 	warnx("fetchListHTTP(): not implemented");
2135 	return (NULL);
2136 }
2137 
2138 /*
2139  * Arbitrary HTTP verb and content requests
2140  */
2141 FILE *
2142 fetchReqHTTP(struct url *URL, const char *method, const char *flags,
2143 	const char *content_type, const char *body)
2144 {
2145 
2146 	return (http_request_body(URL, method, NULL, http_get_proxy(URL, flags),
2147 	    flags, content_type, body));
2148 }
2149