xref: /netbsd-src/libexec/httpd/bozohttpd.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: bozohttpd.c,v 1.87 2018/01/28 13:37:39 maya Exp $	*/
2 
3 /*	$eterna: bozohttpd.c,v 1.178 2011/11/18 09:21:15 mrg Exp $	*/
4 
5 /*
6  * Copyright (c) 1997-2017 Matthew R. Green
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer and
16  *    dedication in the documentation and/or other materials provided
17  *    with the distribution.
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,
24  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
25  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
26  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
27  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  */
32 
33 /* this program is dedicated to the Great God of Processed Cheese */
34 
35 /*
36  * bozohttpd.c:  minimal httpd; provides only these features:
37  *	- HTTP/0.9 (by virtue of ..)
38  *	- HTTP/1.0
39  *	- HTTP/1.1
40  *	- CGI/1.1 this will only be provided for "system" scripts
41  *	- automatic "missing trailing slash" redirections
42  *	- configurable translation of /~user/ to ~user/public_html,
43  *	  however, this does not include cgi-bin support
44  *	- access lists via libwrap via inetd/tcpd
45  *	- virtual hosting
46  *	- not that we do not even pretend to understand MIME, but
47  *	  rely only on the HTTP specification
48  *	- ipv6 support
49  *	- automatic `index.html' generation
50  *	- configurable server name
51  *	- directory index generation
52  *	- daemon mode (lacks libwrap support)
53  *	- .htpasswd support
54  */
55 
56 /*
57  * requirements for minimal http/1.1 (at least, as documented in
58  * RFC 2616 (HTTP/1.1):
59  *
60  *	- 14.11: content-encoding handling. [1]
61  *
62  *	- 14.13: content-length handling.  this is only a SHOULD header
63  *	  thus we could just not send it ever.  [1]
64  *
65  *	- 14.17: content-type handling. [1]
66  *
67  *	- 14.28: if-unmodified-since handling.  if-modified-since is
68  *	  done since, shouldn't be too hard for this one.
69  *
70  * [1] need to revisit to ensure proper behaviour
71  *
72  * and the following is a list of features that we do not need
73  * to have due to other limits, or are too lazy.  there are more
74  * of these than are listed, but these are of particular note,
75  * and could perhaps be implemented.
76  *
77  *	- 3.5/3.6: content/transfer codings.  probably can ignore
78  *	  this?  we "SHOULD"n't.  but 4.4 says we should ignore a
79  *	  `content-length' header upon reciept of a `transfer-encoding'
80  *	  header.
81  *
82  *	- 5.1.1: request methods.  only MUST support GET and HEAD,
83  *	  but there are new ones besides POST that are currently
84  *	  supported: OPTIONS PUT DELETE TRACE and CONNECT, plus
85  *	  extensions not yet known?
86  *
87  * 	- 10.1: we can ignore informational status codes
88  *
89  *	- 10.3.3/10.3.4/10.3.8:  just use '302' codes always.
90  *
91  *	- 14.1/14.2/14.3/14.27: we do not support Accept: headers.
92  *	  just ignore them and send the request anyway.  they are
93  *	  only SHOULD.
94  *
95  *	- 14.5/14.16/14.35: only support simple ranges: %d- and %d-%d
96  *	  would be nice to support more.
97  *
98  *	- 14.9: we aren't a cache.
99  *
100  *	- 14.15: content-md5 would be nice.
101  *
102  *	- 14.24/14.26/14.27: if-match, if-none-match, if-range.  be
103  *	  nice to support this.
104  *
105  *	- 14.44: Vary: seems unneeded.  ignore it for now.
106  */
107 
108 #ifndef INDEX_HTML
109 #define INDEX_HTML		"index.html"
110 #endif
111 #ifndef SERVER_SOFTWARE
112 #define SERVER_SOFTWARE		"bozohttpd/20170201"
113 #endif
114 #ifndef DIRECT_ACCESS_FILE
115 #define DIRECT_ACCESS_FILE	".bzdirect"
116 #endif
117 #ifndef REDIRECT_FILE
118 #define REDIRECT_FILE		".bzredirect"
119 #endif
120 #ifndef ABSREDIRECT_FILE
121 #define ABSREDIRECT_FILE	".bzabsredirect"
122 #endif
123 #ifndef PUBLIC_HTML
124 #define PUBLIC_HTML		"public_html"
125 #endif
126 
127 #ifndef USE_ARG
128 #define USE_ARG(x)	/*LINTED*/(void)&(x)
129 #endif
130 
131 /*
132  * And so it begins ..
133  */
134 
135 #include <sys/param.h>
136 #include <sys/socket.h>
137 #include <sys/time.h>
138 #include <sys/mman.h>
139 
140 #include <arpa/inet.h>
141 
142 #include <ctype.h>
143 #include <dirent.h>
144 #include <errno.h>
145 #include <fcntl.h>
146 #include <netdb.h>
147 #include <pwd.h>
148 #include <grp.h>
149 #include <signal.h>
150 #include <stdarg.h>
151 #include <stdlib.h>
152 #include <string.h>
153 #include <syslog.h>
154 #include <time.h>
155 #include <unistd.h>
156 
157 #include "bozohttpd.h"
158 
159 #ifndef MAX_WAIT_TIME
160 #define	MAX_WAIT_TIME	60	/* hang around for 60 seconds max */
161 #endif
162 
163 /* variables and functions */
164 #ifndef LOG_FTP
165 #define LOG_FTP LOG_DAEMON
166 #endif
167 
168 volatile sig_atomic_t	alarmhit;
169 
170 /*
171  * check there's enough space in the prefs and names arrays.
172  */
173 static int
174 size_arrays(bozoprefs_t *bozoprefs, size_t needed)
175 {
176 	char	**temp;
177 
178 	if (bozoprefs->size == 0) {
179 		/* only get here first time around */
180 		bozoprefs->name = calloc(sizeof(char *), needed);
181 		if (bozoprefs->name == NULL)
182 			return 0;
183 		bozoprefs->value = calloc(sizeof(char *), needed);
184 		if (bozoprefs->value == NULL) {
185 			free(bozoprefs->name);
186 			return 0;
187 		}
188 		bozoprefs->size = needed;
189 	} else if (bozoprefs->count == bozoprefs->size) {
190 		/* only uses 'needed' when filled array */
191 		temp = realloc(bozoprefs->name, sizeof(char *) * needed);
192 		if (temp == NULL)
193 			return 0;
194 		bozoprefs->name = temp;
195 		temp = realloc(bozoprefs->value, sizeof(char *) * needed);
196 		if (temp == NULL)
197 			return 0;
198 		bozoprefs->value = temp;
199 		bozoprefs->size += needed;
200 	}
201 	return 1;
202 }
203 
204 static ssize_t
205 findvar(bozoprefs_t *bozoprefs, const char *name)
206 {
207 	size_t	i;
208 
209 	for (i = 0; i < bozoprefs->count; i++)
210 		if (strcmp(bozoprefs->name[i], name) == 0)
211 			return (ssize_t)i;
212 	return -1;
213 }
214 
215 int
216 bozo_set_pref(bozohttpd_t *httpd, bozoprefs_t *bozoprefs,
217 	      const char *name, const char *value)
218 {
219 	ssize_t	i;
220 
221 	if ((i = findvar(bozoprefs, name)) < 0) {
222 		/* add the element to the array */
223 		if (!size_arrays(bozoprefs, bozoprefs->size + 15))
224 			return 0;
225 		i = bozoprefs->count++;
226 		bozoprefs->name[i] = bozostrdup(httpd, NULL, name);
227 	} else {
228 		/* replace the element in the array */
229 		if (bozoprefs->value[i]) {
230 			free(bozoprefs->value[i]);
231 			bozoprefs->value[i] = NULL;
232 		}
233 	}
234 	bozoprefs->value[i] = bozostrdup(httpd, NULL, value);
235 	return 1;
236 }
237 
238 /*
239  * get a variable's value, or NULL
240  */
241 char *
242 bozo_get_pref(bozoprefs_t *bozoprefs, const char *name)
243 {
244 	ssize_t	i;
245 
246 	i = findvar(bozoprefs, name);
247 	return i < 0 ? NULL : bozoprefs->value[i];
248 }
249 
250 char *
251 bozo_http_date(char *date, size_t datelen)
252 {
253 	struct	tm *tm;
254 	time_t	now;
255 
256 	/* Sun, 06 Nov 1994 08:49:37 GMT */
257 	now = time(NULL);
258 	tm = gmtime(&now);	/* HTTP/1.1 spec rev 06 sez GMT only */
259 	strftime(date, datelen, "%a, %d %b %Y %H:%M:%S GMT", tm);
260 	return date;
261 }
262 
263 /*
264  * convert "in" into the three parts of a request (first line).
265  * we allocate into file and query, but return pointers into
266  * "in" for proto and method.
267  */
268 static void
269 parse_request(bozohttpd_t *httpd, char *in, char **method, char **file,
270 		char **query, char **proto)
271 {
272 	ssize_t	len;
273 	char	*val;
274 
275 	USE_ARG(httpd);
276 	debug((httpd, DEBUG_EXPLODING, "parse in: %s", in));
277 	*method = *file = *query = *proto = NULL;
278 
279 	len = (ssize_t)strlen(in);
280 	val = bozostrnsep(&in, " \t\n\r", &len);
281 	if (len < 1 || val == NULL)
282 		return;
283 	*method = val;
284 
285 	while (*in == ' ' || *in == '\t')
286 		in++;
287 	val = bozostrnsep(&in, " \t\n\r", &len);
288 	if (len < 1) {
289 		if (len == 0)
290 			*file = val;
291 		else
292 			*file = in;
293 	} else {
294 		*file = val;
295 
296 		*query = strchr(*file, '?');
297 		if (*query)
298 			*(*query)++ = '\0';
299 
300 		if (in) {
301 			while (*in && (*in == ' ' || *in == '\t'))
302 				in++;
303 			if (*in)
304 				*proto = in;
305 		}
306 	}
307 
308 	/* allocate private copies */
309 	*file = bozostrdup(httpd, NULL, *file);
310 	if (*query)
311 		*query = bozostrdup(httpd, NULL, *query);
312 
313 	debug((httpd, DEBUG_FAT,
314 		"url: method: \"%s\" file: \"%s\" query: \"%s\" proto: \"%s\"",
315 		*method, *file, *query, *proto));
316 }
317 
318 /*
319  * cleanup a bozo_httpreq_t after use
320  */
321 void
322 bozo_clean_request(bozo_httpreq_t *request)
323 {
324 	struct bozoheaders *hdr, *ohdr = NULL;
325 
326 	if (request == NULL)
327 		return;
328 
329 	/* If SSL enabled cleanup SSL structure. */
330 	bozo_ssl_destroy(request->hr_httpd);
331 
332 	/* clean up request */
333 	free(request->hr_remotehost);
334 	free(request->hr_remoteaddr);
335 	free(request->hr_serverport);
336 	free(request->hr_virthostname);
337 	free(request->hr_file);
338 	free(request->hr_oldfile);
339 	free(request->hr_query);
340 	free(request->hr_host);
341 	bozo_user_free(request->hr_user);
342 	bozo_auth_cleanup(request);
343 	for (hdr = SIMPLEQ_FIRST(&request->hr_headers); hdr;
344 	    hdr = SIMPLEQ_NEXT(hdr, h_next)) {
345 		free(hdr->h_value);
346 		free(hdr->h_header);
347 		free(ohdr);
348 		ohdr = hdr;
349 	}
350 	free(ohdr);
351 	ohdr = NULL;
352 	for (hdr = SIMPLEQ_FIRST(&request->hr_replheaders); hdr;
353 	    hdr = SIMPLEQ_NEXT(hdr, h_next)) {
354 		free(hdr->h_value);
355 		free(hdr->h_header);
356 		free(ohdr);
357 		ohdr = hdr;
358 	}
359 	free(ohdr);
360 
361 	free(request);
362 }
363 
364 /*
365  * send a HTTP/1.1 408 response if we timeout.
366  */
367 /* ARGSUSED */
368 static void
369 alarmer(int sig)
370 {
371 	alarmhit = 1;
372 }
373 
374 /*
375  * a list of header quirks: currently, a list of headers that
376  * can't be folded into a single line.
377  */
378 const char *header_quirks[] = { "WWW-Authenticate", NULL };
379 
380 /*
381  * add or merge this header (val: str) into the requests list
382  */
383 static bozoheaders_t *
384 addmerge_header(bozo_httpreq_t *request, struct qheaders *headers,
385 		const char *val, const char *str, ssize_t len)
386 {
387 	struct	bozohttpd_t *httpd = request->hr_httpd;
388 	struct bozoheaders	 *hdr = NULL;
389 	const char		**quirk;
390 
391 	USE_ARG(len);
392 	for (quirk = header_quirks; *quirk; quirk++)
393 		if (strcasecmp(*quirk, val) == 0)
394 			break;
395 
396 	if (*quirk == NULL) {
397 		/* do we exist already? */
398 		SIMPLEQ_FOREACH(hdr, headers, h_next) {
399 			if (strcasecmp(val, hdr->h_header) == 0)
400 				break;
401 		}
402 	}
403 
404 	if (hdr) {
405 		/* yup, merge it in */
406 		char *nval;
407 
408 		bozoasprintf(httpd, &nval, "%s, %s", hdr->h_value, str);
409 		free(hdr->h_value);
410 		hdr->h_value = nval;
411 	} else {
412 		/* nope, create a new one */
413 
414 		hdr = bozomalloc(httpd, sizeof *hdr);
415 		hdr->h_header = bozostrdup(httpd, request, val);
416 		if (str && *str)
417 			hdr->h_value = bozostrdup(httpd, request, str);
418 		else
419 			hdr->h_value = bozostrdup(httpd, request, " ");
420 
421 		SIMPLEQ_INSERT_TAIL(headers, hdr, h_next);
422 		request->hr_nheaders++;
423 	}
424 
425 	return hdr;
426 }
427 
428 bozoheaders_t *
429 addmerge_reqheader(bozo_httpreq_t *request, const char *val, const char *str,
430 		   ssize_t len)
431 {
432 
433 	return addmerge_header(request, &request->hr_headers, val, str, len);
434 }
435 
436 bozoheaders_t *
437 addmerge_replheader(bozo_httpreq_t *request, const char *val, const char *str,
438 		    ssize_t len)
439 {
440 
441 	return addmerge_header(request, &request->hr_replheaders,
442 	    val, str, len);
443 }
444 
445 /*
446  * as the prototype string is not constant (eg, "HTTP/1.1" is equivalent
447  * to "HTTP/001.01"), we MUST parse this.
448  */
449 static int
450 process_proto(bozo_httpreq_t *request, const char *proto)
451 {
452 	struct	bozohttpd_t *httpd = request->hr_httpd;
453 	char	majorstr[16], *minorstr;
454 	int	majorint, minorint;
455 
456 	if (proto == NULL) {
457 got_proto_09:
458 		request->hr_proto = httpd->consts.http_09;
459 		debug((httpd, DEBUG_FAT, "request %s is http/0.9",
460 			request->hr_file));
461 		return 0;
462 	}
463 
464 	if (strncasecmp(proto, "HTTP/", 5) != 0)
465 		goto bad;
466 	strncpy(majorstr, proto + 5, sizeof majorstr);
467 	majorstr[sizeof(majorstr)-1] = 0;
468 	minorstr = strchr(majorstr, '.');
469 	if (minorstr == NULL)
470 		goto bad;
471 	*minorstr++ = 0;
472 
473 	majorint = atoi(majorstr);
474 	minorint = atoi(minorstr);
475 
476 	switch (majorint) {
477 	case 0:
478 		if (minorint != 9)
479 			break;
480 		goto got_proto_09;
481 	case 1:
482 		if (minorint == 0)
483 			request->hr_proto = httpd->consts.http_10;
484 		else if (minorint == 1)
485 			request->hr_proto = httpd->consts.http_11;
486 		else
487 			break;
488 
489 		debug((httpd, DEBUG_FAT, "request %s is %s",
490 		    request->hr_file, request->hr_proto));
491 		SIMPLEQ_INIT(&request->hr_headers);
492 		request->hr_nheaders = 0;
493 		return 0;
494 	}
495 bad:
496 	return bozo_http_error(httpd, 404, NULL, "unknown prototype");
497 }
498 
499 /*
500  * process each type of HTTP method, setting this HTTP requests
501  * method type.
502  */
503 static struct method_map {
504 	const char *name;
505 	int	type;
506 } method_map[] = {
507 	{ "GET", 	HTTP_GET, },
508 	{ "POST",	HTTP_POST, },
509 	{ "HEAD",	HTTP_HEAD, },
510 #if 0	/* other non-required http/1.1 methods */
511 	{ "OPTIONS",	HTTP_OPTIONS, },
512 	{ "PUT",	HTTP_PUT, },
513 	{ "DELETE",	HTTP_DELETE, },
514 	{ "TRACE",	HTTP_TRACE, },
515 	{ "CONNECT",	HTTP_CONNECT, },
516 #endif
517 	{ NULL,		0, },
518 };
519 
520 static int
521 process_method(bozo_httpreq_t *request, const char *method)
522 {
523 	struct	bozohttpd_t *httpd = request->hr_httpd;
524 	struct	method_map *mmp;
525 
526 	if (request->hr_proto == httpd->consts.http_11)
527 		request->hr_allow = "GET, HEAD, POST";
528 
529 	for (mmp = method_map; mmp->name; mmp++)
530 		if (strcasecmp(method, mmp->name) == 0) {
531 			request->hr_method = mmp->type;
532 			request->hr_methodstr = mmp->name;
533 			return 0;
534 		}
535 
536 	return bozo_http_error(httpd, 404, request, "unknown method");
537 }
538 
539 /*
540  * This function reads a http request from stdin, returning a pointer to a
541  * bozo_httpreq_t structure, describing the request.
542  */
543 bozo_httpreq_t *
544 bozo_read_request(bozohttpd_t *httpd)
545 {
546 	struct	sigaction	sa;
547 	char	*str, *val, *method, *file, *proto, *query;
548 	char	*host, *addr, *port;
549 	char	bufport[10];
550 	char	hbuf[NI_MAXHOST], abuf[NI_MAXHOST];
551 	struct	sockaddr_storage ss;
552 	ssize_t	len;
553 	int	line = 0;
554 	socklen_t slen;
555 	bozo_httpreq_t *request;
556 
557 	/*
558 	 * if we're in daemon mode, bozo_daemon_fork() will return here twice
559 	 * for each call.  once in the child, returning 0, and once in the
560 	 * parent, returning 1.  for each child, then we can setup SSL, and
561 	 * the parent can signal the caller there was no request to process
562 	 * and it will wait for another.
563 	 */
564 	if (bozo_daemon_fork(httpd))
565 		return NULL;
566 	if (bozo_ssl_accept(httpd))
567 		return NULL;
568 
569 	request = bozomalloc(httpd, sizeof(*request));
570 	memset(request, 0, sizeof(*request));
571 	request->hr_httpd = httpd;
572 	request->hr_allow = request->hr_host = NULL;
573 	request->hr_content_type = request->hr_content_length = NULL;
574 	request->hr_range = NULL;
575 	request->hr_last_byte_pos = -1;
576 	request->hr_if_modified_since = NULL;
577 	request->hr_virthostname = NULL;
578 	request->hr_file = NULL;
579 	request->hr_oldfile = NULL;
580 	SIMPLEQ_INIT(&request->hr_replheaders);
581 	bozo_auth_init(request);
582 
583 	slen = sizeof(ss);
584 	if (getpeername(0, (struct sockaddr *)(void *)&ss, &slen) < 0)
585 		host = addr = NULL;
586 	else {
587 		if (getnameinfo((struct sockaddr *)(void *)&ss, slen,
588 		    abuf, sizeof abuf, NULL, 0, NI_NUMERICHOST) == 0)
589 			addr = abuf;
590 		else
591 			addr = NULL;
592 		if (httpd->numeric == 0 &&
593 		    getnameinfo((struct sockaddr *)(void *)&ss, slen,
594 				hbuf, sizeof hbuf, NULL, 0, 0) == 0)
595 			host = hbuf;
596 		else
597 			host = NULL;
598 	}
599 	if (host != NULL)
600 		request->hr_remotehost = bozostrdup(httpd, request, host);
601 	if (addr != NULL)
602 		request->hr_remoteaddr = bozostrdup(httpd, request, addr);
603 	slen = sizeof(ss);
604 
605 	/*
606 	 * Override the bound port from the request value, so it works even
607 	 * if passed through a proxy that doesn't rewrite the port.
608 	 */
609 	if (httpd->bindport) {
610 		if (strcmp(httpd->bindport, "80") != 0)
611 			port = httpd->bindport;
612 		else
613 			port = NULL;
614 	} else {
615 		if (getsockname(0, (struct sockaddr *)(void *)&ss, &slen) < 0)
616 			port = NULL;
617 		else {
618 			if (getnameinfo((struct sockaddr *)(void *)&ss, slen,
619 					NULL, 0, bufport, sizeof bufport,
620 					NI_NUMERICSERV) == 0)
621 				port = bufport;
622 			else
623 				port = NULL;
624 		}
625 	}
626 	if (port != NULL)
627 		request->hr_serverport = bozostrdup(httpd, request, port);
628 
629 	/*
630 	 * setup a timer to make sure the request is not hung
631 	 */
632 	sa.sa_handler = alarmer;
633 	sigemptyset(&sa.sa_mask);
634 	sigaddset(&sa.sa_mask, SIGALRM);
635 	sa.sa_flags = 0;
636 	sigaction(SIGALRM, &sa, NULL);
637 
638 	alarm(MAX_WAIT_TIME);
639 	while ((str = bozodgetln(httpd, STDIN_FILENO, &len, bozo_read)) != NULL) {
640 		alarm(0);
641 		if (alarmhit) {
642 			(void)bozo_http_error(httpd, 408, NULL,
643 					"request timed out");
644 			goto cleanup;
645 		}
646 		line++;
647 
648 		if (line == 1) {
649 
650 			if (len < 1) {
651 				(void)bozo_http_error(httpd, 404, NULL,
652 						"null method");
653 				goto cleanup;
654 			}
655 			bozowarn(httpd,
656 				  "got request ``%s'' from host %s to port %s",
657 				  str,
658 				  host ? host : addr ? addr : "<local>",
659 				  port ? port : "<stdin>");
660 
661 			/* we allocate return space in file and query only */
662 			parse_request(httpd, str, &method, &file, &query, &proto);
663 			request->hr_file = file;
664 			request->hr_query = query;
665 			if (method == NULL) {
666 				(void)bozo_http_error(httpd, 404, NULL,
667 						"null method");
668 				goto cleanup;
669 			}
670 			if (file == NULL) {
671 				(void)bozo_http_error(httpd, 404, NULL,
672 						"null file");
673 				goto cleanup;
674 			}
675 
676 			/*
677 			 * note that we parse the proto first, so that we
678 			 * can more properly parse the method and the url.
679 			 */
680 
681 			if (process_proto(request, proto) ||
682 			    process_method(request, method)) {
683 				goto cleanup;
684 			}
685 
686 			debug((httpd, DEBUG_FAT, "got file \"%s\" query \"%s\"",
687 			    request->hr_file,
688 			    request->hr_query ? request->hr_query : "<none>"));
689 
690 			/* http/0.9 has no header processing */
691 			if (request->hr_proto == httpd->consts.http_09)
692 				break;
693 		} else {		/* incoming headers */
694 			bozoheaders_t *hdr;
695 
696 			if (*str == '\0')
697 				break;
698 
699 			val = bozostrnsep(&str, ":", &len);
700 			debug((httpd, DEBUG_EXPLODING,
701 			    "read_req2: after bozostrnsep: str ``%s'' val ``%s''",
702 			    str, val));
703 			if (val == NULL || len == -1) {
704 				(void)bozo_http_error(httpd, 404, request,
705 						"no header");
706 				goto cleanup;
707 			}
708 			while (*str == ' ' || *str == '\t')
709 				len--, str++;
710 			while (*val == ' ' || *val == '\t')
711 				val++;
712 
713 			if (bozo_auth_check_headers(request, val, str, len))
714 				goto next_header;
715 
716 			hdr = addmerge_reqheader(request, val, str, len);
717 
718 			if (strcasecmp(hdr->h_header, "content-type") == 0)
719 				request->hr_content_type = hdr->h_value;
720 			else if (strcasecmp(hdr->h_header, "content-length") == 0)
721 				request->hr_content_length = hdr->h_value;
722 			else if (strcasecmp(hdr->h_header, "host") == 0)
723 				request->hr_host = bozostrdup(httpd, request,
724 							      hdr->h_value);
725 			/* RFC 2616 (HTTP/1.1): 14.20 */
726 			else if (strcasecmp(hdr->h_header, "expect") == 0) {
727 				(void)bozo_http_error(httpd, 417, request,
728 						"we don't support Expect:");
729 				goto cleanup;
730 			}
731 			else if (strcasecmp(hdr->h_header, "referrer") == 0 ||
732 			         strcasecmp(hdr->h_header, "referer") == 0)
733 				request->hr_referrer = hdr->h_value;
734 			else if (strcasecmp(hdr->h_header, "range") == 0)
735 				request->hr_range = hdr->h_value;
736 			else if (strcasecmp(hdr->h_header,
737 					"if-modified-since") == 0)
738 				request->hr_if_modified_since = hdr->h_value;
739 			else if (strcasecmp(hdr->h_header,
740 					"accept-encoding") == 0)
741 				request->hr_accept_encoding = hdr->h_value;
742 
743 			debug((httpd, DEBUG_FAT, "adding header %s: %s",
744 			    hdr->h_header, hdr->h_value));
745 		}
746 next_header:
747 		alarm(MAX_WAIT_TIME);
748 	}
749 
750 	/* now, clear it all out */
751 	alarm(0);
752 	signal(SIGALRM, SIG_DFL);
753 
754 	/* RFC1945, 8.3 */
755 	if (request->hr_method == HTTP_POST &&
756 	    request->hr_content_length == NULL) {
757 		(void)bozo_http_error(httpd, 400, request,
758 				"missing content length");
759 		goto cleanup;
760 	}
761 
762 	/* RFC 2616 (HTTP/1.1), 14.23 & 19.6.1.1 */
763 	if (request->hr_proto == httpd->consts.http_11 &&
764 	    /*(strncasecmp(request->hr_file, "http://", 7) != 0) &&*/
765 	    request->hr_host == NULL) {
766 		(void)bozo_http_error(httpd, 400, request,
767 				"missing Host header");
768 		goto cleanup;
769 	}
770 
771 	if (request->hr_range != NULL) {
772 		debug((httpd, DEBUG_FAT, "hr_range: %s", request->hr_range));
773 		/* support only simple ranges %d- and %d-%d */
774 		if (strchr(request->hr_range, ',') == NULL) {
775 			const char *rstart, *dash;
776 
777 			rstart = strchr(request->hr_range, '=');
778 			if (rstart != NULL) {
779 				rstart++;
780 				dash = strchr(rstart, '-');
781 				if (dash != NULL && dash != rstart) {
782 					dash++;
783 					request->hr_have_range = 1;
784 					request->hr_first_byte_pos =
785 					    strtoll(rstart, NULL, 10);
786 					if (request->hr_first_byte_pos < 0)
787 						request->hr_first_byte_pos = 0;
788 					if (*dash != '\0') {
789 						request->hr_last_byte_pos =
790 						    strtoll(dash, NULL, 10);
791 						if (request->hr_last_byte_pos < 0)
792 							request->hr_last_byte_pos = -1;
793 					}
794 				}
795 			}
796 		}
797 	}
798 
799 	debug((httpd, DEBUG_FAT, "bozo_read_request returns url %s in request",
800 	       request->hr_file));
801 	return request;
802 
803 cleanup:
804 	bozo_clean_request(request);
805 
806 	return NULL;
807 }
808 
809 static int
810 mmap_and_write_part(bozohttpd_t *httpd, int fd, off_t first_byte_pos, size_t sz)
811 {
812 	size_t mappedsz, wroffset;
813 	off_t mappedoffset;
814 	char *addr;
815 	void *mappedaddr;
816 
817 	/*
818 	 * we need to ensure that both the size *and* offset arguments to
819 	 * mmap() are page-aligned.  our formala for this is:
820 	 *
821 	 *    input offset: first_byte_pos
822 	 *    input size: sz
823 	 *
824 	 *    mapped offset = page align truncate (input offset)
825 	 *    mapped size   =
826 	 *        page align extend (input offset - mapped offset + input size)
827 	 *    write offset  = input offset - mapped offset
828 	 *
829 	 * we use the write offset in all writes
830 	 */
831 	mappedoffset = first_byte_pos & ~(httpd->page_size - 1);
832 	mappedsz = (size_t)
833 		(first_byte_pos - mappedoffset + sz + httpd->page_size - 1) &
834 		~(httpd->page_size - 1);
835 	wroffset = (size_t)(first_byte_pos - mappedoffset);
836 
837 	addr = mmap(0, mappedsz, PROT_READ, MAP_SHARED, fd, mappedoffset);
838 	if (addr == (char *)-1) {
839 		bozowarn(httpd, "mmap failed: %s", strerror(errno));
840 		return -1;
841 	}
842 	mappedaddr = addr;
843 
844 #ifdef MADV_SEQUENTIAL
845 	(void)madvise(addr, sz, MADV_SEQUENTIAL);
846 #endif
847 	while (sz > BOZO_WRSZ) {
848 		if (bozo_write(httpd, STDOUT_FILENO, addr + wroffset,
849 				BOZO_WRSZ) != BOZO_WRSZ) {
850 			bozowarn(httpd, "write failed: %s", strerror(errno));
851 			goto out;
852 		}
853 		debug((httpd, DEBUG_OBESE, "wrote %d bytes", BOZO_WRSZ));
854 		sz -= BOZO_WRSZ;
855 		addr += BOZO_WRSZ;
856 	}
857 	if (sz && (size_t)bozo_write(httpd, STDOUT_FILENO, addr + wroffset,
858 				sz) != sz) {
859 		bozowarn(httpd, "final write failed: %s", strerror(errno));
860 		goto out;
861 	}
862 	debug((httpd, DEBUG_OBESE, "wrote %d bytes", (int)sz));
863  out:
864 	if (munmap(mappedaddr, mappedsz) < 0) {
865 		bozowarn(httpd, "munmap failed");
866 		return -1;
867 	}
868 
869 	return 0;
870 }
871 
872 static int
873 parse_http_date(const char *val, time_t *timestamp)
874 {
875 	char *remainder;
876 	struct tm tm;
877 
878 	if ((remainder = strptime(val, "%a, %d %b %Y %T GMT", &tm)) == NULL &&
879 	    (remainder = strptime(val, "%a, %d-%b-%y %T GMT", &tm)) == NULL &&
880 	    (remainder = strptime(val, "%a %b %d %T %Y", &tm)) == NULL)
881 		return 0; /* Invalid HTTP date format */
882 
883 	if (*remainder)
884 		return 0; /* No trailing garbage */
885 
886 	*timestamp = timegm(&tm);
887 	return 1;
888 }
889 
890 /*
891  * given an url, encode it ala rfc 3986.  ie, escape ? and friends.
892  * note that this function returns a static buffer, and thus needs
893  * to be updated for any sort of parallel processing. escape only
894  * chosen characters for absolute redirects
895  */
896 char *
897 bozo_escape_rfc3986(bozohttpd_t *httpd, const char *url, int absolute)
898 {
899 	static char *buf;
900 	static size_t buflen = 0;
901 	size_t len;
902 	const char *s;
903 	char *d;
904 
905 	len = strlen(url);
906 	if (buflen < len * 3 + 1) {
907 		buflen = len * 3 + 1;
908 		buf = bozorealloc(httpd, buf, buflen);
909 	}
910 
911 	for (len = 0, s = url, d = buf; *s;) {
912 		if (*s & 0x80)
913 			goto encode_it;
914 		switch (*s) {
915 		case ':':
916 		case '?':
917 		case '#':
918 		case '[':
919 		case ']':
920 		case '@':
921 		case '!':
922 		case '$':
923 		case '&':
924 		case '\'':
925 		case '(':
926 		case ')':
927 		case '*':
928 		case '+':
929 		case ',':
930 		case ';':
931 		case '=':
932 		case '%':
933 		case '"':
934 			if (absolute)
935 				goto leave_it;
936 		case '\n':
937 		case '\r':
938 		case ' ':
939 		encode_it:
940 			snprintf(d, 4, "%%%02X", *s++);
941 			d += 3;
942 			len += 3;
943 			break;
944 		leave_it:
945 		default:
946 			*d++ = *s++;
947 			len++;
948 			break;
949 		}
950 	}
951 	buf[len] = 0;
952 
953 	return buf;
954 }
955 
956 /*
957  * do automatic redirection -- if there are query parameters or userdir for
958  * the URL we will tack these on to the new (redirected) URL.
959  */
960 static void
961 handle_redirect(bozo_httpreq_t *request, const char *url, int absolute)
962 {
963 	bozohttpd_t *httpd = request->hr_httpd;
964 	char *finalurl, *urlbuf;
965 #ifndef NO_USER_SUPPORT
966 	char *userbuf;
967 #endif /* !NO_USER_SUPPORT */
968 	char portbuf[20];
969 	const char *scheme, *query, *quest;
970 	const char *hostname = BOZOHOST(httpd, request);
971 	int absproto = 0; /* absolute redirect provides own schema */
972 
973 	if (url == NULL) {
974 		bozoasprintf(httpd, &urlbuf, "/%s/", request->hr_file);
975 		url = urlbuf;
976 	} else
977 		urlbuf = NULL;
978 
979 #ifndef NO_USER_SUPPORT
980 	if (request->hr_user && !absolute) {
981 		bozoasprintf(httpd, &userbuf, "/~%s%s", request->hr_user, url);
982 		url = userbuf;
983 	} else
984 		userbuf = NULL;
985 #endif /* !NO_USER_SUPPORT */
986 
987 	if (absolute) {
988 		char *sep = NULL;
989 		const char *s;
990 
991 		/*
992 		 * absolute redirect may specify own protocol i.e. to redirect
993 		 * to another schema like https:// or ftp://.
994 		 * Details: RFC 3986, section 3.
995 		 */
996 
997 		/* 1. check if url contains :// */
998 		sep = strstr(url, "://");
999 
1000 		/*
1001 		 * RFC 3986, section 3.1:
1002 		 * scheme      = ALPHA *( ALPHA / DIGIT / "+" / "-" / "." )
1003 		 */
1004 		if (sep) {
1005 			for (s = url; s != sep;) {
1006 				if (!isalnum((int)*s) &&
1007 				    *s != '+' && *s != '-' && *s != '.')
1008 					break;
1009 				if (++s == sep) {
1010 					absproto = 1;
1011 				}
1012 			}
1013 		}
1014 	}
1015 
1016 	/* construct final redirection url */
1017 
1018 	scheme = absproto ? "" : httpd->sslinfo ? "https://" : "http://";
1019 
1020 	if (absolute) {
1021 		hostname = "";
1022 		portbuf[0] = '\0';
1023 	} else {
1024 		const char *defport = httpd->sslinfo ? "443" : "80";
1025 
1026 		if (request->hr_serverport &&
1027 		    strcmp(request->hr_serverport, defport) != 0)
1028 			snprintf(portbuf, sizeof(portbuf), ":%s",
1029 			    request->hr_serverport);
1030 		else
1031 			portbuf[0] = '\0';
1032 	}
1033 
1034 	url = bozo_escape_rfc3986(httpd, url, absolute);
1035 
1036 	if (request->hr_query && strlen(request->hr_query)) {
1037 		query = request->hr_query;
1038 		quest = "?";
1039 	} else {
1040 		query = quest = "";
1041 	}
1042 
1043 	bozoasprintf(httpd, &finalurl, "%s%s%s%s%s%s",
1044 		     scheme, hostname, portbuf, url, quest, query);
1045 
1046 	bozowarn(httpd, "redirecting %s", finalurl);
1047 	debug((httpd, DEBUG_FAT, "redirecting %s", finalurl));
1048 
1049 	bozo_printf(httpd, "%s 301 Document Moved\r\n", request->hr_proto);
1050 	if (request->hr_proto != httpd->consts.http_09)
1051 		bozo_print_header(request, NULL, "text/html", NULL);
1052 	if (request->hr_proto != httpd->consts.http_09)
1053 		bozo_printf(httpd, "Location: %s\r\n", finalurl);
1054 	bozo_printf(httpd, "\r\n");
1055 	if (request->hr_method == HTTP_HEAD)
1056 		goto head;
1057 	bozo_printf(httpd, "<html><head><title>Document Moved</title></head>\n");
1058 	bozo_printf(httpd, "<body><h1>Document Moved</h1>\n");
1059 	bozo_printf(httpd, "This document had moved <a href=\"%s\">here</a>\n",
1060 	  finalurl);
1061 	bozo_printf(httpd, "</body></html>\n");
1062 head:
1063 	bozo_flush(httpd, stdout);
1064 	free(urlbuf);
1065 	free(finalurl);
1066 #ifndef NO_USER_SUPPORT
1067 	free(userbuf);
1068 #endif /* !NO_USER_SUPPORT */
1069 }
1070 
1071 /*
1072  * deal with virtual host names; we do this:
1073  *	if we have a virtual path root (httpd->virtbase), and we are given a
1074  *	virtual host spec (Host: ho.st or http://ho.st/), see if this
1075  *	directory exists under httpd->virtbase.  if it does, use this as the
1076  #	new slashdir.
1077  */
1078 static int
1079 check_virtual(bozo_httpreq_t *request)
1080 {
1081 	bozohttpd_t *httpd = request->hr_httpd;
1082 	char *file = request->hr_file, *s;
1083 	size_t len;
1084 
1085 	/*
1086 	 * convert http://virtual.host/ to request->hr_host
1087 	 */
1088 	debug((httpd, DEBUG_OBESE, "checking for http:// virtual host in ``%s''",
1089 			file));
1090 	if (strncasecmp(file, "http://", 7) == 0) {
1091 		/* we would do virtual hosting here? */
1092 		file += 7;
1093 		/* RFC 2616 (HTTP/1.1), 5.2: URI takes precedence over Host: */
1094 		free(request->hr_host);
1095 		request->hr_host = bozostrdup(httpd, request, file);
1096 		if ((s = strchr(request->hr_host, '/')) != NULL)
1097 			*s = '\0';
1098 		s = strchr(file, '/');
1099 		free(request->hr_file);
1100 		request->hr_file = bozostrdup(httpd, request, s ? s : "/");
1101 		debug((httpd, DEBUG_OBESE, "got host ``%s'' file is now ``%s''",
1102 		    request->hr_host, request->hr_file));
1103 	} else if (!request->hr_host)
1104 		goto use_slashdir;
1105 
1106 	/*
1107 	 * canonicalise hr_host - that is, remove any :80.
1108 	 */
1109 	len = strlen(request->hr_host);
1110 	if (len > 3 && strcmp(request->hr_host + len - 3, ":80") == 0) {
1111 		request->hr_host[len - 3] = '\0';
1112 		len = strlen(request->hr_host);
1113 	}
1114 
1115 	if (!httpd->virtbase) {
1116 
1117 		/*
1118 		 * if we don't use vhost support, then set virthostname if
1119 		 * user supplied Host header. It will be used for possible
1120 		 * redirections
1121 		 */
1122 
1123 		if (request->hr_host) {
1124 			s = strrchr(request->hr_host, ':');
1125 			if (s != NULL)
1126 				/* truncate Host: as we want to copy it without port part */
1127 				*s = '\0';
1128 			request->hr_virthostname = bozostrdup(httpd, request,
1129 			  request->hr_host);
1130 			if (s != NULL)
1131 				/* fix Host: again, if we truncated it */
1132 				*s = ':';
1133 		}
1134 
1135 		goto use_slashdir;
1136 	}
1137 
1138 	/*
1139 	 * ok, we have a virtual host, use opendir(3) to find a case
1140 	 * insensitive match for the virtual host we are asked for.
1141 	 * note that if the virtual host is the same as the master,
1142 	 * we don't need to do anything special.
1143 	 */
1144 	debug((httpd, DEBUG_OBESE,
1145 	    "check_virtual: checking host `%s' under httpd->virtbase `%s' "
1146 	    "for file `%s'",
1147 	    request->hr_host, httpd->virtbase, request->hr_file));
1148 	if (strncasecmp(httpd->virthostname, request->hr_host, len) != 0) {
1149 		s = 0;
1150 		DIR *dirp;
1151 		struct dirent *d;
1152 
1153 		if ((dirp = opendir(httpd->virtbase)) != NULL) {
1154 			while ((d = readdir(dirp)) != NULL) {
1155 				if (strcmp(d->d_name, ".") == 0 ||
1156 				    strcmp(d->d_name, "..") == 0) {
1157 					continue;
1158 				}
1159 				debug((httpd, DEBUG_OBESE, "looking at dir``%s''",
1160 			 	   d->d_name));
1161 				if (strcmp(d->d_name, request->hr_host) == 0) {
1162 					/* found it, punch it */
1163 					debug((httpd, DEBUG_OBESE, "found it punch it"));
1164 					request->hr_virthostname =
1165 					    bozostrdup(httpd, request, d->d_name);
1166 					bozoasprintf(httpd, &s, "%s/%s",
1167 					    httpd->virtbase,
1168 					    request->hr_virthostname);
1169 					break;
1170 				}
1171 			}
1172 			closedir(dirp);
1173 		}
1174 		else {
1175 			debug((httpd, DEBUG_FAT, "opendir %s failed: %s",
1176 			    httpd->virtbase, strerror(errno)));
1177 		}
1178 		if (s == 0) {
1179 			if (httpd->unknown_slash)
1180 				goto use_slashdir;
1181 			return bozo_http_error(httpd, 404, request,
1182 						"unknown URL");
1183 		}
1184 	} else
1185 use_slashdir:
1186 		s = httpd->slashdir;
1187 
1188 	/*
1189 	 * ok, nailed the correct slashdir, chdir to it
1190 	 */
1191 	if (chdir(s) < 0)
1192 		return bozo_http_error(httpd, 404, request,
1193 					"can't chdir to slashdir");
1194 	return 0;
1195 }
1196 
1197 /*
1198  * checks to see if this request has a valid .bzredirect file.  returns
1199  * 0 when no redirection happend, or 1 when handle_redirect() has been
1200  * called, -1 on error.
1201  */
1202 static int
1203 check_bzredirect(bozo_httpreq_t *request)
1204 {
1205 	bozohttpd_t *httpd = request->hr_httpd;
1206 	struct stat sb;
1207 	char dir[MAXPATHLEN], redir[MAXPATHLEN], redirpath[MAXPATHLEN + 1],
1208 	    path[MAXPATHLEN];
1209 	char *basename, *finalredir;
1210 	int rv, absolute;
1211 
1212 	/*
1213 	 * if this pathname is really a directory, but doesn't end in /,
1214 	 * use it as the directory to look for the redir file.
1215 	 */
1216 	if((size_t)snprintf(dir, sizeof(dir), "%s", request->hr_file + 1) >=
1217 	  sizeof(dir)) {
1218 		bozo_http_error(httpd, 404, request,
1219 		  "file path too long");
1220 		return -1;
1221 	}
1222 	debug((httpd, DEBUG_FAT, "check_bzredirect: dir %s", dir));
1223 	basename = strrchr(dir, '/');
1224 
1225 	if ((!basename || basename[1] != '\0') &&
1226 	    lstat(dir, &sb) == 0 && S_ISDIR(sb.st_mode)) {
1227 		strcpy(path, dir);
1228 	} else if (basename == NULL) {
1229 		strcpy(path, ".");
1230 		strcpy(dir, "");
1231 	} else {
1232 		*basename++ = '\0';
1233 		bozo_check_special_files(request, basename);
1234 		strcpy(path, dir);
1235 	}
1236 
1237 	debug((httpd, DEBUG_FAT, "check_bzredirect: path %s", path));
1238 
1239 	if ((size_t)snprintf(redir, sizeof(redir), "%s/%s", path,
1240 	  REDIRECT_FILE) >= sizeof(redir)) {
1241 		bozo_http_error(httpd, 404, request,
1242 		    "redirectfile path too long");
1243 		return -1;
1244 	}
1245 	if (lstat(redir, &sb) == 0) {
1246 		if (!S_ISLNK(sb.st_mode))
1247 			return 0;
1248 		absolute = 0;
1249 	} else {
1250 		if((size_t)snprintf(redir, sizeof(redir), "%s/%s", path,
1251 		  ABSREDIRECT_FILE) >= sizeof(redir)) {
1252 			bozo_http_error(httpd, 404, request,
1253 			  "redirectfile path too long");
1254 			return -1;
1255 		}
1256 		if (lstat(redir, &sb) < 0 || !S_ISLNK(sb.st_mode))
1257 			return 0;
1258 		absolute = 1;
1259 	}
1260 	debug((httpd, DEBUG_FAT, "check_bzredirect: calling readlink"));
1261 	rv = readlink(redir, redirpath, sizeof redirpath - 1);
1262 	if (rv == -1 || rv == 0) {
1263 		debug((httpd, DEBUG_FAT, "readlink failed"));
1264 		return 0;
1265 	}
1266 	redirpath[rv] = '\0';
1267 	debug((httpd, DEBUG_FAT, "readlink returned \"%s\"", redirpath));
1268 
1269 	/* check if we need authentication */
1270 	snprintf(path, sizeof(path), "%s/", dir);
1271 	if (bozo_auth_check(request, path))
1272 		return 1;
1273 
1274 	/* now we have the link pointer, redirect to the real place */
1275 	if (!absolute && redirpath[0] != '/') {
1276 		if ((size_t)snprintf(finalredir = redir, sizeof(redir), "%s%s/%s",
1277 		  (strlen(dir) > 0 ? "/" : ""), dir, redirpath) >= sizeof(redir)) {
1278 			bozo_http_error(httpd, 404, request,
1279 			  "redirect path too long");
1280 			return -1;
1281 		}
1282 	} else
1283 		finalredir = redirpath;
1284 
1285 	debug((httpd, DEBUG_FAT, "check_bzredirect: new redir %s", finalredir));
1286 	handle_redirect(request, finalredir, absolute);
1287 	return 1;
1288 }
1289 
1290 /* this fixes the %HH hack that RFC2396 requires.  */
1291 int
1292 bozo_decode_url_percent(bozo_httpreq_t *request, char *str)
1293 {
1294 	bozohttpd_t *httpd = request->hr_httpd;
1295 	char	*s, *t, buf[3];
1296 	char	*end;	/* if end is not-zero, we don't translate beyond that */
1297 
1298 	end = str + strlen(str);
1299 
1300 	/* fast forward to the first % */
1301 	if ((s = strchr(str, '%')) == NULL)
1302 		return 0;
1303 
1304 	t = s;
1305 	do {
1306 		if (end && s >= end) {
1307 			debug((httpd, DEBUG_EXPLODING,
1308 				"fu_%%: past end, filling out.."));
1309 			while (*s)
1310 				*t++ = *s++;
1311 			break;
1312 		}
1313 		debug((httpd, DEBUG_EXPLODING,
1314 			"fu_%%: got s == %%, s[1]s[2] == %c%c",
1315 			s[1], s[2]));
1316 		if (s[1] == '\0' || s[2] == '\0') {
1317 			(void)bozo_http_error(httpd, 400, request,
1318 			    "percent hack missing two chars afterwards");
1319 			return 1;
1320 		}
1321 		if (s[1] == '0' && s[2] == '0') {
1322 			(void)bozo_http_error(httpd, 404, request,
1323 					"percent hack was %00");
1324 			return 1;
1325 		}
1326 		if (s[1] == '2' && s[2] == 'f') {
1327 			(void)bozo_http_error(httpd, 404, request,
1328 					"percent hack was %2f (/)");
1329 			return 1;
1330 		}
1331 
1332 		buf[0] = *++s;
1333 		buf[1] = *++s;
1334 		buf[2] = '\0';
1335 		s++;
1336 		*t = (char)strtol(buf, NULL, 16);
1337 		debug((httpd, DEBUG_EXPLODING,
1338 				"fu_%%: strtol put '%02x' into *t", *t));
1339 		if (*t++ == '\0') {
1340 			(void)bozo_http_error(httpd, 400, request,
1341 					"percent hack got a 0 back");
1342 			return 1;
1343 		}
1344 
1345 		while (*s && *s != '%') {
1346 			if (end && s >= end)
1347 				break;
1348 			*t++ = *s++;
1349 		}
1350 	} while (*s);
1351 	*t = '\0';
1352 
1353 	debug((httpd, DEBUG_FAT, "bozo_decode_url_percent returns `%s'",
1354 			request->hr_file));
1355 
1356 	return 0;
1357 }
1358 
1359 /*
1360  * transform_request does this:
1361  *	- ``expand'' %20 crapola
1362  *	- punt if it doesn't start with /
1363  *	- look for "http://myname/" and deal with it.
1364  *	- maybe call bozo_process_cgi()
1365  *	- check for ~user and call bozo_user_transform() if so
1366  *	- if the length > 1, check for trailing slash.  if so,
1367  *	  add the index.html file
1368  *	- if the length is 1, return the index.html file
1369  *	- disallow anything ending up with a file starting
1370  *	  at "/" or having ".." in it.
1371  *	- anything else is a really weird internal error
1372  *	- returns malloced file to serve, if unhandled
1373  */
1374 static int
1375 transform_request(bozo_httpreq_t *request, int *isindex)
1376 {
1377 	bozohttpd_t *httpd = request->hr_httpd;
1378 	char	*file, *newfile = NULL;
1379 	size_t	len;
1380 
1381 	file = NULL;
1382 	*isindex = 0;
1383 	debug((httpd, DEBUG_FAT, "tf_req: file %s", request->hr_file));
1384 	if (bozo_decode_url_percent(request, request->hr_file)) {
1385 		goto bad_done;
1386 	}
1387 	if (check_virtual(request)) {
1388 		goto bad_done;
1389 	}
1390 	file = request->hr_file;
1391 
1392 	if (file[0] != '/') {
1393 		(void)bozo_http_error(httpd, 404, request, "unknown URL");
1394 		goto bad_done;
1395 	}
1396 
1397 	/* omit additional slashes at the beginning */
1398 	while (file[1] == '/')
1399 		file++;
1400 
1401 	/* fix file provided by user as it's used in other handlers */
1402 	request->hr_file = file;
1403 
1404 	len = strlen(file);
1405 
1406 #ifndef NO_USER_SUPPORT
1407 	/* first of all expand user path */
1408 	if (len > 1 && httpd->enable_users && file[1] == '~') {
1409 		if (file[2] == '\0') {
1410 			(void)bozo_http_error(httpd, 404, request,
1411 						"missing username");
1412 			goto bad_done;
1413 		}
1414 		if (strchr(file + 2, '/') == NULL) {
1415 			char *userredirecturl;
1416 			bozoasprintf(httpd, &userredirecturl, "%s/", file);
1417 			handle_redirect(request, userredirecturl, 0);
1418 			free(userredirecturl);
1419 			return 0;
1420 		}
1421 		debug((httpd, DEBUG_FAT, "calling bozo_user_transform"));
1422 
1423 		if (!bozo_user_transform(request))
1424 			return 0;
1425 
1426 		file = request->hr_file;
1427 		len = strlen(file);
1428 	}
1429 #endif /* NO_USER_SUPPORT */
1430 
1431 
1432 	switch (check_bzredirect(request)) {
1433 	case -1:
1434 		goto bad_done;
1435 	case 1:
1436 		return 0;
1437 	}
1438 
1439 	if (len > 1) {
1440 		debug((httpd, DEBUG_FAT, "file[len-1] == %c", file[len-1]));
1441 		if (file[len-1] == '/') {	/* append index.html */
1442 			*isindex = 1;
1443 			debug((httpd, DEBUG_FAT, "appending index.html"));
1444 			newfile = bozomalloc(httpd,
1445 					len + strlen(httpd->index_html) + 1);
1446 			strcpy(newfile, file + 1);
1447 			strcat(newfile, httpd->index_html);
1448 		} else
1449 			newfile = bozostrdup(httpd, request, file + 1);
1450 	} else if (len == 1) {
1451 		debug((httpd, DEBUG_EXPLODING, "tf_req: len == 1"));
1452 		newfile = bozostrdup(httpd, request, httpd->index_html);
1453 		*isindex = 1;
1454 	} else {	/* len == 0 ? */
1455 		(void)bozo_http_error(httpd, 500, request,
1456 					"request->hr_file is nul?");
1457 		goto bad_done;
1458 	}
1459 
1460 	if (newfile == NULL) {
1461 		(void)bozo_http_error(httpd, 500, request, "internal failure");
1462 		goto bad_done;
1463 	}
1464 
1465 	/*
1466 	 * stop traversing outside our domain
1467 	 *
1468 	 * XXX true security only comes from our parent using chroot(2)
1469 	 * before execve(2)'ing us.  or our own built in chroot(2) support.
1470 	 */
1471 
1472 	debug((httpd, DEBUG_FAT, "newfile: %s", newfile));
1473 
1474 	if (*newfile == '/' || strcmp(newfile, "..") == 0 ||
1475 	    strstr(newfile, "/..") || strstr(newfile, "../")) {
1476 		(void)bozo_http_error(httpd, 403, request, "illegal request");
1477 		goto bad_done;
1478 	}
1479 
1480 	if (bozo_auth_check(request, newfile))
1481 		goto bad_done;
1482 
1483 	if (strlen(newfile)) {
1484 		request->hr_oldfile = request->hr_file;
1485 		request->hr_file = newfile;
1486 	}
1487 
1488 	if (bozo_process_cgi(request))
1489 		return 0;
1490 
1491 	if (bozo_process_lua(request))
1492 		return 0;
1493 
1494 	debug((httpd, DEBUG_FAT, "transform_request set: %s", newfile));
1495 	return 1;
1496 bad_done:
1497 	debug((httpd, DEBUG_FAT, "transform_request returning: 0"));
1498 	free(newfile);
1499 	return 0;
1500 }
1501 
1502 /*
1503  * can_gzip checks if the request supports and prefers gzip encoding.
1504  *
1505  * XXX: we do not consider the associated q with gzip in making our
1506  *      decision which is broken.
1507  */
1508 
1509 static int
1510 can_gzip(bozo_httpreq_t *request)
1511 {
1512 	const char	*pos;
1513 	const char	*tmp;
1514 	size_t		 len;
1515 
1516 	/* First we decide if the request can be gzipped at all. */
1517 
1518 	/* not if we already are encoded... */
1519 	tmp = bozo_content_encoding(request, request->hr_file);
1520 	if (tmp && *tmp)
1521 		return 0;
1522 
1523 	/* not if we are not asking for the whole file... */
1524 	if (request->hr_last_byte_pos != -1 || request->hr_have_range)
1525 		return 0;
1526 
1527 	/* Then we determine if gzip is on the cards. */
1528 
1529 	for (pos = request->hr_accept_encoding; pos && *pos; pos += len) {
1530 		while (*pos == ' ')
1531 			pos++;
1532 
1533 		len = strcspn(pos, ";,");
1534 
1535 		if ((len == 4 && strncasecmp("gzip", pos, 4) == 0) ||
1536 		    (len == 6 && strncasecmp("x-gzip", pos, 6) == 0))
1537 			return 1;
1538 
1539 		if (pos[len] == ';')
1540 			len += strcspn(&pos[len], ",");
1541 
1542 		if (pos[len])
1543 			len++;
1544 	}
1545 
1546 	return 0;
1547 }
1548 
1549 /*
1550  * bozo_process_request does the following:
1551  *	- check the request is valid
1552  *	- process cgi-bin if necessary
1553  *	- transform a filename if necesarry
1554  *	- return the HTTP request
1555  */
1556 void
1557 bozo_process_request(bozo_httpreq_t *request)
1558 {
1559 	bozohttpd_t *httpd = request->hr_httpd;
1560 	struct	stat sb;
1561 	time_t timestamp;
1562 	char	*file;
1563 	const char *type, *encoding;
1564 	int	fd, isindex;
1565 
1566 	/*
1567 	 * note that transform_request chdir()'s if required.  also note
1568 	 * that cgi is handed here.  if transform_request() returns 0
1569 	 * then the request has been handled already.
1570 	 */
1571 	if (transform_request(request, &isindex) == 0)
1572 		return;
1573 
1574 	fd = -1;
1575 	encoding = NULL;
1576 	if (can_gzip(request)) {
1577 		bozoasprintf(httpd, &file, "%s.gz", request->hr_file);
1578 		fd = open(file, O_RDONLY);
1579 		if (fd >= 0)
1580 			encoding = "gzip";
1581 		free(file);
1582 	}
1583 
1584 	file = request->hr_file;
1585 
1586 	if (fd < 0)
1587 		fd = open(file, O_RDONLY);
1588 
1589 	if (fd < 0) {
1590 		debug((httpd, DEBUG_FAT, "open failed: %s", strerror(errno)));
1591 		switch (errno) {
1592 		case EPERM:
1593 		case EACCES:
1594 			(void)bozo_http_error(httpd, 403, request,
1595 						"no permission to open file");
1596 			break;
1597 		case ENAMETOOLONG:
1598 			/*FALLTHROUGH*/
1599 		case ENOENT:
1600 			if (!bozo_dir_index(request, file, isindex))
1601 				(void)bozo_http_error(httpd, 404, request,
1602 							"no file");
1603 			break;
1604 		default:
1605 			(void)bozo_http_error(httpd, 500, request, "open file");
1606 		}
1607 		goto cleanup_nofd;
1608 	}
1609 	if (fstat(fd, &sb) < 0) {
1610 		(void)bozo_http_error(httpd, 500, request, "can't fstat");
1611 		goto cleanup;
1612 	}
1613 	if (S_ISDIR(sb.st_mode)) {
1614 		handle_redirect(request, NULL, 0);
1615 		goto cleanup;
1616 	}
1617 
1618 	if (request->hr_if_modified_since &&
1619 	    parse_http_date(request->hr_if_modified_since, &timestamp) &&
1620 	    timestamp >= sb.st_mtime) {
1621 		/* XXX ignore subsecond of timestamp */
1622 		bozo_printf(httpd, "%s 304 Not Modified\r\n",
1623 				request->hr_proto);
1624 		bozo_printf(httpd, "\r\n");
1625 		bozo_flush(httpd, stdout);
1626 		goto cleanup;
1627 	}
1628 
1629 	/* validate requested range */
1630 	if (request->hr_last_byte_pos == -1 ||
1631 	    request->hr_last_byte_pos >= sb.st_size)
1632 		request->hr_last_byte_pos = sb.st_size - 1;
1633 	if (request->hr_have_range &&
1634 	    request->hr_first_byte_pos > request->hr_last_byte_pos) {
1635 		request->hr_have_range = 0;	/* punt */
1636 		request->hr_first_byte_pos = 0;
1637 		request->hr_last_byte_pos = sb.st_size - 1;
1638 	}
1639 	debug((httpd, DEBUG_FAT, "have_range %d first_pos %lld last_pos %lld",
1640 	    request->hr_have_range,
1641 	    (long long)request->hr_first_byte_pos,
1642 	    (long long)request->hr_last_byte_pos));
1643 	if (request->hr_have_range)
1644 		bozo_printf(httpd, "%s 206 Partial Content\r\n",
1645 				request->hr_proto);
1646 	else
1647 		bozo_printf(httpd, "%s 200 OK\r\n", request->hr_proto);
1648 
1649 	if (request->hr_proto != httpd->consts.http_09) {
1650 		type = bozo_content_type(request, file);
1651 		if (!encoding)
1652 			encoding = bozo_content_encoding(request, file);
1653 
1654 		bozo_print_header(request, &sb, type, encoding);
1655 		bozo_printf(httpd, "\r\n");
1656 	}
1657 	bozo_flush(httpd, stdout);
1658 
1659 	if (request->hr_method != HTTP_HEAD) {
1660 		off_t szleft, cur_byte_pos;
1661 
1662 		szleft =
1663 		     request->hr_last_byte_pos - request->hr_first_byte_pos + 1;
1664 		cur_byte_pos = request->hr_first_byte_pos;
1665 
1666  retry:
1667 		while (szleft) {
1668 			size_t sz;
1669 
1670 			if ((off_t)httpd->mmapsz < szleft)
1671 				sz = httpd->mmapsz;
1672 			else
1673 				sz = (size_t)szleft;
1674 			if (mmap_and_write_part(httpd, fd, cur_byte_pos, sz)) {
1675 				if (errno == ENOMEM) {
1676 					httpd->mmapsz /= 2;
1677 					if (httpd->mmapsz >= httpd->page_size)
1678 						goto retry;
1679 				}
1680 				goto cleanup;
1681 			}
1682 			cur_byte_pos += sz;
1683 			szleft -= sz;
1684 		}
1685 	}
1686  cleanup:
1687 	close(fd);
1688  cleanup_nofd:
1689 	close(STDIN_FILENO);
1690 	close(STDOUT_FILENO);
1691 	/*close(STDERR_FILENO);*/
1692 }
1693 
1694 /* make sure we're not trying to access special files */
1695 int
1696 bozo_check_special_files(bozo_httpreq_t *request, const char *name)
1697 {
1698 	bozohttpd_t *httpd = request->hr_httpd;
1699 
1700 	/* ensure basename(name) != special files */
1701 	if (strcmp(name, DIRECT_ACCESS_FILE) == 0)
1702 		return bozo_http_error(httpd, 403, request,
1703 		    "no permission to open direct access file");
1704 	if (strcmp(name, REDIRECT_FILE) == 0)
1705 		return bozo_http_error(httpd, 403, request,
1706 		    "no permission to open redirect file");
1707 	if (strcmp(name, ABSREDIRECT_FILE) == 0)
1708 		return bozo_http_error(httpd, 403, request,
1709 		    "no permission to open redirect file");
1710 	return bozo_auth_check_special_files(request, name);
1711 }
1712 
1713 /* generic header printing routine */
1714 void
1715 bozo_print_header(bozo_httpreq_t *request,
1716 		struct stat *sbp, const char *type, const char *encoding)
1717 {
1718 	bozohttpd_t *httpd = request->hr_httpd;
1719 	off_t len;
1720 	char	date[40];
1721 	bozoheaders_t *hdr;
1722 
1723 	SIMPLEQ_FOREACH(hdr, &request->hr_replheaders, h_next) {
1724 		bozo_printf(httpd, "%s: %s\r\n", hdr->h_header,
1725 				hdr->h_value);
1726 	}
1727 
1728 	bozo_printf(httpd, "Date: %s\r\n", bozo_http_date(date, sizeof(date)));
1729 	bozo_printf(httpd, "Server: %s\r\n", httpd->server_software);
1730 	bozo_printf(httpd, "Accept-Ranges: bytes\r\n");
1731 	if (sbp) {
1732 		char filedate[40];
1733 		struct	tm *tm;
1734 
1735 		tm = gmtime(&sbp->st_mtime);
1736 		strftime(filedate, sizeof filedate,
1737 		    "%a, %d %b %Y %H:%M:%S GMT", tm);
1738 		bozo_printf(httpd, "Last-Modified: %s\r\n", filedate);
1739 	}
1740 	if (type && *type)
1741 		bozo_printf(httpd, "Content-Type: %s\r\n", type);
1742 	if (encoding && *encoding)
1743 		bozo_printf(httpd, "Content-Encoding: %s\r\n", encoding);
1744 	if (sbp) {
1745 		if (request->hr_have_range) {
1746 			len = request->hr_last_byte_pos -
1747 					request->hr_first_byte_pos +1;
1748 			bozo_printf(httpd,
1749 				"Content-Range: bytes %qd-%qd/%qd\r\n",
1750 				(long long) request->hr_first_byte_pos,
1751 				(long long) request->hr_last_byte_pos,
1752 				(long long) sbp->st_size);
1753 		} else
1754 			len = sbp->st_size;
1755 		bozo_printf(httpd, "Content-Length: %qd\r\n", (long long)len);
1756 	}
1757 	if (request->hr_proto == httpd->consts.http_11)
1758 		bozo_printf(httpd, "Connection: close\r\n");
1759 	bozo_flush(httpd, stdout);
1760 }
1761 
1762 #ifndef NO_DEBUG
1763 void
1764 debug__(bozohttpd_t *httpd, int level, const char *fmt, ...)
1765 {
1766 	va_list	ap;
1767 	int savederrno;
1768 
1769 	/* only log if the level is low enough */
1770 	if (httpd->debug < level)
1771 		return;
1772 
1773 	savederrno = errno;
1774 	va_start(ap, fmt);
1775 	if (httpd->logstderr) {
1776 		vfprintf(stderr, fmt, ap);
1777 		fputs("\n", stderr);
1778 	} else
1779 		vsyslog(LOG_DEBUG, fmt, ap);
1780 	va_end(ap);
1781 	errno = savederrno;
1782 }
1783 #endif /* NO_DEBUG */
1784 
1785 /* these are like warn() and err(), except for syslog not stderr */
1786 void
1787 bozowarn(bozohttpd_t *httpd, const char *fmt, ...)
1788 {
1789 	va_list ap;
1790 
1791 	va_start(ap, fmt);
1792 	if (httpd->logstderr || isatty(STDERR_FILENO)) {
1793 		//fputs("warning: ", stderr);
1794 		vfprintf(stderr, fmt, ap);
1795 		fputs("\n", stderr);
1796 	} else
1797 		vsyslog(LOG_INFO, fmt, ap);
1798 	va_end(ap);
1799 }
1800 
1801 void
1802 bozoerr(bozohttpd_t *httpd, int code, const char *fmt, ...)
1803 {
1804 	va_list ap;
1805 
1806 	va_start(ap, fmt);
1807 	if (httpd->logstderr || isatty(STDERR_FILENO)) {
1808 		//fputs("error: ", stderr);
1809 		vfprintf(stderr, fmt, ap);
1810 		fputs("\n", stderr);
1811 	} else
1812 		vsyslog(LOG_ERR, fmt, ap);
1813 	va_end(ap);
1814 	exit(code);
1815 }
1816 
1817 void
1818 bozoasprintf(bozohttpd_t *httpd, char **str, const char *fmt, ...)
1819 {
1820 	va_list ap;
1821 	int e;
1822 
1823 	va_start(ap, fmt);
1824 	e = vasprintf(str, fmt, ap);
1825 	va_end(ap);
1826 
1827 	if (e < 0)
1828 		bozoerr(httpd, EXIT_FAILURE, "asprintf");
1829 }
1830 
1831 /*
1832  * this escapes HTML tags.  returns allocated escaped
1833  * string if needed, or NULL on allocation failure or
1834  * lack of escape need.
1835  * call with NULL httpd in error paths, to avoid recursive
1836  * malloc failure.  call with valid httpd in normal paths
1837  * to get automatic allocation failure handling.
1838  */
1839 char *
1840 bozo_escape_html(bozohttpd_t *httpd, const char *url)
1841 {
1842 	int	i, j;
1843 	char	*tmp;
1844 	size_t	len;
1845 
1846 	for (i = 0, j = 0; url[i]; i++) {
1847 		switch (url[i]) {
1848 		case '<':
1849 		case '>':
1850 			j += 4;
1851 			break;
1852 		case '&':
1853 			j += 5;
1854 			break;
1855 		}
1856 	}
1857 
1858 	if (j == 0)
1859 		return NULL;
1860 
1861 	/*
1862 	 * we need to handle being called from different
1863 	 * pathnames.
1864 	 */
1865 	len = strlen(url) + j;
1866 	if (httpd)
1867 		tmp = bozomalloc(httpd, len);
1868 	else if ((tmp = malloc(len)) == 0)
1869 			return NULL;
1870 
1871 	for (i = 0, j = 0; url[i]; i++) {
1872 		switch (url[i]) {
1873 		case '<':
1874 			memcpy(tmp + j, "&lt;", 4);
1875 			j += 4;
1876 			break;
1877 		case '>':
1878 			memcpy(tmp + j, "&gt;", 4);
1879 			j += 4;
1880 			break;
1881 		case '&':
1882 			memcpy(tmp + j, "&amp;", 5);
1883 			j += 5;
1884 			break;
1885 		default:
1886 			tmp[j++] = url[i];
1887 		}
1888 	}
1889 	tmp[j] = 0;
1890 
1891 	return tmp;
1892 }
1893 
1894 /* short map between error code, and short/long messages */
1895 static struct errors_map {
1896 	int	code;			/* HTTP return code */
1897 	const char *shortmsg;		/* short version of message */
1898 	const char *longmsg;		/* long version of message */
1899 } errors_map[] = {
1900 	{ 400,	"400 Bad Request",	"The request was not valid", },
1901 	{ 401,	"401 Unauthorized",	"No authorization", },
1902 	{ 403,	"403 Forbidden",	"Access to this item has been denied",},
1903 	{ 404, 	"404 Not Found",	"This item has not been found", },
1904 	{ 408, 	"408 Request Timeout",	"This request took too long", },
1905 	{ 417,	"417 Expectation Failed","Expectations not available", },
1906 	{ 420,	"420 Enhance Your Calm","Chill, Winston", },
1907 	{ 500,	"500 Internal Error",	"An error occured on the server", },
1908 	{ 501,	"501 Not Implemented",	"This request is not available", },
1909 	{ 0,	NULL,			NULL, },
1910 };
1911 
1912 static const char *help = "DANGER! WILL ROBINSON! DANGER!";
1913 
1914 static const char *
1915 http_errors_short(int code)
1916 {
1917 	struct errors_map *ep;
1918 
1919 	for (ep = errors_map; ep->code; ep++)
1920 		if (ep->code == code)
1921 			return (ep->shortmsg);
1922 	return (help);
1923 }
1924 
1925 static const char *
1926 http_errors_long(int code)
1927 {
1928 	struct errors_map *ep;
1929 
1930 	for (ep = errors_map; ep->code; ep++)
1931 		if (ep->code == code)
1932 			return (ep->longmsg);
1933 	return (help);
1934 }
1935 
1936 /* the follow functions and variables are used in handling HTTP errors */
1937 /* ARGSUSED */
1938 int
1939 bozo_http_error(bozohttpd_t *httpd, int code, bozo_httpreq_t *request,
1940 		const char *msg)
1941 {
1942 	char portbuf[20];
1943 	const char *header = http_errors_short(code);
1944 	const char *reason = http_errors_long(code);
1945 	const char *proto = (request && request->hr_proto) ?
1946 				request->hr_proto : httpd->consts.http_11;
1947 	int	size;
1948 	bozoheaders_t *hdr;
1949 
1950 	debug((httpd, DEBUG_FAT, "bozo_http_error %d: %s", code, msg));
1951 	if (header == NULL || reason == NULL) {
1952 		bozoerr(httpd, 1,
1953 			"bozo_http_error() failed (short = %p, long = %p)",
1954 			header, reason);
1955 		return code;
1956 	}
1957 
1958 	if (request && request->hr_serverport &&
1959 	    strcmp(request->hr_serverport, "80") != 0)
1960 		snprintf(portbuf, sizeof(portbuf), ":%s",
1961 				request->hr_serverport);
1962 	else
1963 		portbuf[0] = '\0';
1964 
1965 	if (request && request->hr_file) {
1966 		char *file = NULL, *user = NULL, *user_escaped = NULL;
1967 		int file_alloc = 0;
1968 		const char *hostname = BOZOHOST(httpd, request);
1969 
1970 		/* bozo_escape_html() failure here is just too bad. */
1971 		file = bozo_escape_html(NULL, request->hr_file);
1972 		if (file == NULL)
1973 			file = request->hr_file;
1974 		else
1975 			file_alloc = 1;
1976 
1977 #ifndef NO_USER_SUPPORT
1978 		if (request->hr_user != NULL) {
1979 			user_escaped = bozo_escape_html(NULL, request->hr_user);
1980 			if (user_escaped == NULL)
1981 				user_escaped = request->hr_user;
1982 			/* expand username to ~user/ */
1983 			bozoasprintf(httpd, &user, "~%s/", user_escaped);
1984 			if (user_escaped != request->hr_user)
1985 				free(user_escaped);
1986 		}
1987 #endif /* !NO_USER_SUPPORT */
1988 
1989 		size = snprintf(httpd->errorbuf, BUFSIZ,
1990 		    "<html><head><title>%s</title></head>\n"
1991 		    "<body><h1>%s</h1>\n"
1992 		    "%s%s: <pre>%s</pre>\n"
1993  		    "<hr><address><a href=\"//%s%s/\">%s%s</a></address>\n"
1994 		    "</body></html>\n",
1995 		    header, header,
1996 		    user ? user : "", file,
1997 		    reason, hostname, portbuf, hostname, portbuf);
1998 		free(user);
1999 		if (size >= (int)BUFSIZ) {
2000 			bozowarn(httpd,
2001 				"bozo_http_error buffer too small, truncated");
2002 			size = (int)BUFSIZ;
2003 		}
2004 
2005 		if (file_alloc)
2006 			free(file);
2007 	} else
2008 		size = 0;
2009 
2010 	bozo_printf(httpd, "%s %s\r\n", proto, header);
2011 
2012 	if (request) {
2013 		bozo_auth_check_401(request, code);
2014 		SIMPLEQ_FOREACH(hdr, &request->hr_replheaders, h_next) {
2015 			bozo_printf(httpd, "%s: %s\r\n", hdr->h_header,
2016 					hdr->h_value);
2017 		}
2018 	}
2019 
2020 	bozo_printf(httpd, "Content-Type: text/html\r\n");
2021 	bozo_printf(httpd, "Content-Length: %d\r\n", size);
2022 	bozo_printf(httpd, "Server: %s\r\n", httpd->server_software);
2023 	if (request && request->hr_allow)
2024 		bozo_printf(httpd, "Allow: %s\r\n", request->hr_allow);
2025 	bozo_printf(httpd, "\r\n");
2026 	/* According to the RFC 2616 sec. 9.4 HEAD method MUST NOT return a
2027 	 * message-body in the response */
2028 	if (size && request && request->hr_method != HTTP_HEAD)
2029 		bozo_printf(httpd, "%s", httpd->errorbuf);
2030 	bozo_flush(httpd, stdout);
2031 
2032 	return code;
2033 }
2034 
2035 /* Below are various modified libc functions */
2036 
2037 /*
2038  * returns -1 in lenp if the string ran out before finding a delimiter,
2039  * but is otherwise the same as strsep.  Note that the length must be
2040  * correctly passed in.
2041  */
2042 char *
2043 bozostrnsep(char **strp, const char *delim, ssize_t	*lenp)
2044 {
2045 	char	*s;
2046 	const	char *spanp;
2047 	int	c, sc;
2048 	char	*tok;
2049 
2050 	if ((s = *strp) == NULL)
2051 		return (NULL);
2052 	for (tok = s;;) {
2053 		if (lenp && --(*lenp) == -1)
2054 			return (NULL);
2055 		c = *s++;
2056 		spanp = delim;
2057 		do {
2058 			if ((sc = *spanp++) == c) {
2059 				if (c == 0)
2060 					s = NULL;
2061 				else
2062 					s[-1] = '\0';
2063 				*strp = s;
2064 				return (tok);
2065 			}
2066 		} while (sc != 0);
2067 	}
2068 	/* NOTREACHED */
2069 }
2070 
2071 /*
2072  * inspired by fgetln(3), but works for fd's.  should work identically
2073  * except it, however, does *not* return the newline, and it does nul
2074  * terminate the string.
2075  */
2076 char *
2077 bozodgetln(bozohttpd_t *httpd, int fd, ssize_t *lenp,
2078 	ssize_t (*readfn)(bozohttpd_t *, int, void *, size_t))
2079 {
2080 	ssize_t	len;
2081 	int	got_cr = 0;
2082 	char	c, *nbuffer;
2083 
2084 	/* initialise */
2085 	if (httpd->getln_buflen == 0) {
2086 		/* should be plenty for most requests */
2087 		httpd->getln_buflen = 128;
2088 		httpd->getln_buffer = malloc((size_t)httpd->getln_buflen);
2089 		if (httpd->getln_buffer == NULL) {
2090 			httpd->getln_buflen = 0;
2091 			return NULL;
2092 		}
2093 	}
2094 	len = 0;
2095 
2096 	/*
2097 	 * we *have* to read one byte at a time, to not break cgi
2098 	 * programs (for we pass stdin off to them).  could fix this
2099 	 * by becoming a fd-passing program instead of just exec'ing
2100 	 * the program
2101 	 *
2102 	 * the above is no longer true, we are the fd-passing
2103 	 * program already.
2104 	 */
2105 	for (; readfn(httpd, fd, &c, 1) == 1; ) {
2106 		debug((httpd, DEBUG_EXPLODING, "bozodgetln read %c", c));
2107 
2108 		if (len >= httpd->getln_buflen - 1) {
2109 			httpd->getln_buflen *= 2;
2110 			debug((httpd, DEBUG_EXPLODING, "bozodgetln: "
2111 				"reallocating buffer to buflen %zu",
2112 				httpd->getln_buflen));
2113 			nbuffer = bozorealloc(httpd, httpd->getln_buffer,
2114 				(size_t)httpd->getln_buflen);
2115 			httpd->getln_buffer = nbuffer;
2116 		}
2117 
2118 		httpd->getln_buffer[len++] = c;
2119 		if (c == '\r') {
2120 			got_cr = 1;
2121 			continue;
2122 		} else if (c == '\n') {
2123 			/*
2124 			 * HTTP/1.1 spec says to ignore CR and treat
2125 			 * LF as the real line terminator.  even though
2126 			 * the same spec defines CRLF as the line
2127 			 * terminator, it is recommended in section 19.3
2128 			 * to do the LF trick for tolerance.
2129 			 */
2130 			if (got_cr)
2131 				len -= 2;
2132 			else
2133 				len -= 1;
2134 			break;
2135 		}
2136 
2137 	}
2138 	httpd->getln_buffer[len] = '\0';
2139 	debug((httpd, DEBUG_OBESE, "bozodgetln returns: ``%s'' with len %zd",
2140 	       httpd->getln_buffer, len));
2141 	*lenp = len;
2142 	return httpd->getln_buffer;
2143 }
2144 
2145 void *
2146 bozorealloc(bozohttpd_t *httpd, void *ptr, size_t size)
2147 {
2148 	void	*p;
2149 
2150 	p = realloc(ptr, size);
2151 	if (p)
2152 		return p;
2153 
2154 	(void)bozo_http_error(httpd, 500, NULL, "memory allocation failure");
2155 	exit(EXIT_FAILURE);
2156 }
2157 
2158 void *
2159 bozomalloc(bozohttpd_t *httpd, size_t size)
2160 {
2161 	void	*p;
2162 
2163 	p = malloc(size);
2164 	if (p)
2165 		return p;
2166 
2167 	(void)bozo_http_error(httpd, 500, NULL, "memory allocation failure");
2168 	exit(EXIT_FAILURE);
2169 }
2170 
2171 char *
2172 bozostrdup(bozohttpd_t *httpd, bozo_httpreq_t *request, const char *str)
2173 {
2174 	char	*p;
2175 
2176 	p = strdup(str);
2177 	if (p)
2178 		return p;
2179 
2180 	if (!request)
2181 		bozoerr(httpd, EXIT_FAILURE, "strdup");
2182 
2183 	(void)bozo_http_error(httpd, 500, request, "memory allocation failure");
2184 	exit(EXIT_FAILURE);
2185 }
2186 
2187 /* set default values in bozohttpd_t struct */
2188 int
2189 bozo_init_httpd(bozohttpd_t *httpd)
2190 {
2191 	/* make sure everything is clean */
2192 	(void) memset(httpd, 0x0, sizeof(*httpd));
2193 
2194 	/* constants */
2195 	httpd->consts.http_09 = "HTTP/0.9";
2196 	httpd->consts.http_10 = "HTTP/1.0";
2197 	httpd->consts.http_11 = "HTTP/1.1";
2198 	httpd->consts.text_plain = "text/plain";
2199 
2200 	/* mmap region size */
2201 	httpd->mmapsz = BOZO_MMAPSZ;
2202 
2203 	/* error buffer for bozo_http_error() */
2204 	if ((httpd->errorbuf = malloc(BUFSIZ)) == NULL) {
2205 		(void) fprintf(stderr,
2206 			"bozohttpd: memory_allocation failure\n");
2207 		return 0;
2208 	}
2209 #ifndef NO_LUA_SUPPORT
2210 	SIMPLEQ_INIT(&httpd->lua_states);
2211 #endif
2212 	return 1;
2213 }
2214 
2215 /* set default values in bozoprefs_t struct */
2216 int
2217 bozo_init_prefs(bozohttpd_t *httpd, bozoprefs_t *prefs)
2218 {
2219 	/* make sure everything is clean */
2220 	(void) memset(prefs, 0x0, sizeof(*prefs));
2221 
2222 	/* set up default values */
2223 	if (!bozo_set_pref(httpd, prefs, "server software", SERVER_SOFTWARE) ||
2224 	    !bozo_set_pref(httpd, prefs, "index.html", INDEX_HTML) ||
2225 	    !bozo_set_pref(httpd, prefs, "public_html", PUBLIC_HTML))
2226 		return 0;
2227 
2228 	return 1;
2229 }
2230 
2231 /* set default values */
2232 int
2233 bozo_set_defaults(bozohttpd_t *httpd, bozoprefs_t *prefs)
2234 {
2235 	return bozo_init_httpd(httpd) && bozo_init_prefs(httpd, prefs);
2236 }
2237 
2238 /* set the virtual host name, port and root */
2239 int
2240 bozo_setup(bozohttpd_t *httpd, bozoprefs_t *prefs, const char *vhost,
2241 		const char *root)
2242 {
2243 	struct passwd	 *pw;
2244 	extern char	**environ;
2245 	static char	 *cleanenv[1] = { NULL };
2246 	uid_t		  uid;
2247 	int		  uidset = 0;
2248 	char		 *chrootdir;
2249 	char		 *username;
2250 	char		 *portnum;
2251 	char		 *cp;
2252 	int		  dirtyenv;
2253 
2254 	dirtyenv = 0;
2255 
2256 	if (vhost == NULL) {
2257 		httpd->virthostname = bozomalloc(httpd, MAXHOSTNAMELEN+1);
2258 		if (gethostname(httpd->virthostname, MAXHOSTNAMELEN+1) < 0)
2259 			bozoerr(httpd, 1, "gethostname");
2260 		httpd->virthostname[MAXHOSTNAMELEN] = '\0';
2261 	} else {
2262 		httpd->virthostname = bozostrdup(httpd, NULL, vhost);
2263 	}
2264 	httpd->slashdir = bozostrdup(httpd, NULL, root);
2265 	if ((portnum = bozo_get_pref(prefs, "port number")) != NULL) {
2266 		httpd->bindport = bozostrdup(httpd, NULL, portnum);
2267 	}
2268 
2269 	/* go over preferences now */
2270 	if ((cp = bozo_get_pref(prefs, "numeric")) != NULL &&
2271 	    strcmp(cp, "true") == 0) {
2272 		httpd->numeric = 1;
2273 	}
2274 	if ((cp = bozo_get_pref(prefs, "log to stderr")) != NULL &&
2275 	    strcmp(cp, "true") == 0) {
2276 		httpd->logstderr = 1;
2277 	}
2278 	if ((cp = bozo_get_pref(prefs, "bind address")) != NULL) {
2279 		httpd->bindaddress = bozostrdup(httpd, NULL, cp);
2280 	}
2281 	if ((cp = bozo_get_pref(prefs, "background")) != NULL) {
2282 		httpd->background = atoi(cp);
2283 	}
2284 	if ((cp = bozo_get_pref(prefs, "foreground")) != NULL &&
2285 	    strcmp(cp, "true") == 0) {
2286 		httpd->foreground = 1;
2287 	}
2288 	if ((cp = bozo_get_pref(prefs, "pid file")) != NULL) {
2289 		httpd->pidfile = bozostrdup(httpd, NULL, cp);
2290 	}
2291 	if ((cp = bozo_get_pref(prefs, "unknown slash")) != NULL &&
2292 	    strcmp(cp, "true") == 0) {
2293 		httpd->unknown_slash = 1;
2294 	}
2295 	if ((cp = bozo_get_pref(prefs, "virtual base")) != NULL) {
2296 		httpd->virtbase = bozostrdup(httpd, NULL, cp);
2297 	}
2298 	if ((cp = bozo_get_pref(prefs, "enable users")) != NULL &&
2299 	    strcmp(cp, "true") == 0) {
2300 		httpd->enable_users = 1;
2301 	}
2302 	if ((cp = bozo_get_pref(prefs, "enable user cgibin")) != NULL &&
2303 	    strcmp(cp, "true") == 0) {
2304 		httpd->enable_cgi_users = 1;
2305 	}
2306 	if ((cp = bozo_get_pref(prefs, "dirty environment")) != NULL &&
2307 	    strcmp(cp, "true") == 0) {
2308 		dirtyenv = 1;
2309 	}
2310 	if ((cp = bozo_get_pref(prefs, "hide dots")) != NULL &&
2311 	    strcmp(cp, "true") == 0) {
2312 		httpd->hide_dots = 1;
2313 	}
2314 	if ((cp = bozo_get_pref(prefs, "directory indexing")) != NULL &&
2315 	    strcmp(cp, "true") == 0) {
2316 		httpd->dir_indexing = 1;
2317 	}
2318 	if ((cp = bozo_get_pref(prefs, "public_html")) != NULL) {
2319 		httpd->public_html = bozostrdup(httpd, NULL, cp);
2320 	}
2321 	httpd->server_software =
2322 	    bozostrdup(httpd, NULL, bozo_get_pref(prefs, "server software"));
2323 	httpd->index_html =
2324 	    bozostrdup(httpd, NULL, bozo_get_pref(prefs, "index.html"));
2325 
2326 	/*
2327 	 * initialise ssl and daemon mode if necessary.
2328 	 */
2329 	bozo_ssl_init(httpd);
2330 	bozo_daemon_init(httpd);
2331 
2332 	username = bozo_get_pref(prefs, "username");
2333 	if (username != NULL) {
2334 		if ((pw = getpwnam(username)) == NULL)
2335 			bozoerr(httpd, 1, "getpwnam(%s): %s", username,
2336 				strerror(errno));
2337 		if (initgroups(pw->pw_name, pw->pw_gid) == -1)
2338 			bozoerr(httpd, 1, "initgroups: %s", strerror(errno));
2339 		if (setgid(pw->pw_gid) == -1)
2340 			bozoerr(httpd, 1, "setgid(%u): %s", pw->pw_gid,
2341 				strerror(errno));
2342 		uid = pw->pw_uid;
2343 		uidset = 1;
2344 	}
2345 	/*
2346 	 * handle chroot.
2347 	 */
2348 	if ((chrootdir = bozo_get_pref(prefs, "chroot dir")) != NULL) {
2349 		httpd->rootdir = bozostrdup(httpd, NULL, chrootdir);
2350 		if (chdir(httpd->rootdir) == -1)
2351 			bozoerr(httpd, 1, "chdir(%s): %s", httpd->rootdir,
2352 				strerror(errno));
2353 		if (chroot(httpd->rootdir) == -1)
2354 			bozoerr(httpd, 1, "chroot(%s): %s", httpd->rootdir,
2355 				strerror(errno));
2356 	}
2357 
2358 	if (uidset && setuid(uid) == -1)
2359 		bozoerr(httpd, 1, "setuid(%d): %s", uid, strerror(errno));
2360 
2361 	/*
2362 	 * prevent info leakage between different compartments.
2363 	 * some PATH values in the environment would be invalided
2364 	 * by chroot. cross-user settings might result in undesirable
2365 	 * effects.
2366 	 */
2367 	if ((chrootdir != NULL || username != NULL) && !dirtyenv)
2368 		environ = cleanenv;
2369 
2370 #ifdef _SC_PAGESIZE
2371 	httpd->page_size = (long)sysconf(_SC_PAGESIZE);
2372 #else
2373 	httpd->page_size = 4096;
2374 #endif
2375 	debug((httpd, DEBUG_OBESE, "myname is %s, slashdir is %s",
2376 			httpd->virthostname, httpd->slashdir));
2377 
2378 	return 1;
2379 }
2380 
2381 int
2382 bozo_get_version(char *buf, size_t size)
2383 {
2384 	return snprintf(buf, size, "%s", SERVER_SOFTWARE);
2385 }
2386