xref: /netbsd-src/libexec/httpd/bozohttpd.c (revision b7b7574d3bf8eeb51a1fa3977b59142ec6434a55)
1 /*	$NetBSD: bozohttpd.c,v 1.52 2014/07/02 13:58:09 shm Exp $	*/
2 
3 /*	$eterna: bozohttpd.c,v 1.178 2011/11/18 09:21:15 mrg Exp $	*/
4 
5 /*
6  * Copyright (c) 1997-2014 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/20140516"
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, unsigned needed)
175 {
176 	char	**temp;
177 
178 	if (bozoprefs->size == 0) {
179 		/* only get here first time around */
180 		bozoprefs->size = needed;
181 		if ((bozoprefs->name = calloc(sizeof(char *), needed)) == NULL) {
182 			(void) fprintf(stderr, "size_arrays: bad alloc\n");
183 			return 0;
184 		}
185 		if ((bozoprefs->value = calloc(sizeof(char *), needed)) == NULL) {
186 			free(bozoprefs->name);
187 			(void) fprintf(stderr, "size_arrays: bad alloc\n");
188 			return 0;
189 		}
190 	} else if (bozoprefs->c == bozoprefs->size) {
191 		/* only uses 'needed' when filled array */
192 		bozoprefs->size += needed;
193 		temp = realloc(bozoprefs->name, sizeof(char *) * needed);
194 		if (temp == NULL) {
195 			(void) fprintf(stderr, "size_arrays: bad alloc\n");
196 			return 0;
197 		}
198 		bozoprefs->name = temp;
199 		temp = realloc(bozoprefs->value, sizeof(char *) * needed);
200 		if (temp == NULL) {
201 			(void) fprintf(stderr, "size_arrays: bad alloc\n");
202 			return 0;
203 		}
204 		bozoprefs->value = temp;
205 	}
206 	return 1;
207 }
208 
209 static int
210 findvar(bozoprefs_t *bozoprefs, const char *name)
211 {
212 	unsigned	i;
213 
214 	for (i = 0 ; i < bozoprefs->c && strcmp(bozoprefs->name[i], name) != 0; i++)
215 		;
216 	return (i == bozoprefs->c) ? -1 : (int)i;
217 }
218 
219 int
220 bozo_set_pref(bozoprefs_t *bozoprefs, const char *name, const char *value)
221 {
222 	int	i;
223 
224 	if ((i = findvar(bozoprefs, name)) < 0) {
225 		/* add the element to the array */
226 		if (size_arrays(bozoprefs, bozoprefs->size + 15)) {
227 			bozoprefs->name[i = bozoprefs->c++] = strdup(name);
228 		}
229 	} else {
230 		/* replace the element in the array */
231 		if (bozoprefs->value[i]) {
232 			free(bozoprefs->value[i]);
233 			bozoprefs->value[i] = NULL;
234 		}
235 	}
236 	/* sanity checks for range of values go here */
237 	bozoprefs->value[i] = strdup(value);
238 	return 1;
239 }
240 
241 /*
242  * get a variable's value, or NULL
243  */
244 char *
245 bozo_get_pref(bozoprefs_t *bozoprefs, const char *name)
246 {
247 	int	i;
248 
249 	return ((i = findvar(bozoprefs, name)) < 0) ? NULL :
250 			bozoprefs->value[i];
251 }
252 
253 char *
254 bozo_http_date(char *date, size_t datelen)
255 {
256 	struct	tm *tm;
257 	time_t	now;
258 
259 	/* Sun, 06 Nov 1994 08:49:37 GMT */
260 	now = time(NULL);
261 	tm = gmtime(&now);	/* HTTP/1.1 spec rev 06 sez GMT only */
262 	strftime(date, datelen, "%a, %d %b %Y %H:%M:%S GMT", tm);
263 	return date;
264 }
265 
266 /*
267  * convert "in" into the three parts of a request (first line).
268  * we allocate into file and query, but return pointers into
269  * "in" for proto and method.
270  */
271 static void
272 parse_request(bozohttpd_t *httpd, char *in, char **method, char **file,
273 		char **query, char **proto)
274 {
275 	ssize_t	len;
276 	char	*val;
277 
278 	USE_ARG(httpd);
279 	debug((httpd, DEBUG_EXPLODING, "parse in: %s", in));
280 	*method = *file = *query = *proto = NULL;
281 
282 	len = (ssize_t)strlen(in);
283 	val = bozostrnsep(&in, " \t\n\r", &len);
284 	if (len < 1 || val == NULL)
285 		return;
286 	*method = val;
287 
288 	while (*in == ' ' || *in == '\t')
289 		in++;
290 	val = bozostrnsep(&in, " \t\n\r", &len);
291 	if (len < 1) {
292 		if (len == 0)
293 			*file = val;
294 		else
295 			*file = in;
296 	} else {
297 		*file = val;
298 
299 		*query = strchr(*file, '?');
300 		if (*query)
301 			*(*query)++ = '\0';
302 
303 		if (in) {
304 			while (*in && (*in == ' ' || *in == '\t'))
305 				in++;
306 			if (*in)
307 				*proto = in;
308 		}
309 	}
310 
311 	/* allocate private copies */
312 	*file = bozostrdup(httpd, *file);
313 	if (*query)
314 		*query = bozostrdup(httpd, *query);
315 
316 	debug((httpd, DEBUG_FAT,
317 		"url: method: \"%s\" file: \"%s\" query: \"%s\" proto: \"%s\"",
318 		*method, *file, *query, *proto));
319 }
320 
321 /*
322  * cleanup a bozo_httpreq_t after use
323  */
324 void
325 bozo_clean_request(bozo_httpreq_t *request)
326 {
327 	struct bozoheaders *hdr, *ohdr = NULL;
328 
329 	if (request == NULL)
330 		return;
331 
332 	/* If SSL enabled cleanup SSL structure. */
333 	bozo_ssl_destroy(request->hr_httpd);
334 
335 	/* clean up request */
336 	free(request->hr_remotehost);
337 	free(request->hr_remoteaddr);
338 	free(request->hr_serverport);
339 	free(request->hr_virthostname);
340 	free(request->hr_file);
341 	free(request->hr_oldfile);
342 	free(request->hr_query);
343 	free(request->hr_host);
344 	bozo_auth_cleanup(request);
345 	for (hdr = SIMPLEQ_FIRST(&request->hr_headers); hdr;
346 	    hdr = SIMPLEQ_NEXT(hdr, h_next)) {
347 		free(hdr->h_value);
348 		free(hdr->h_header);
349 		free(ohdr);
350 		ohdr = hdr;
351 	}
352 	free(ohdr);
353 
354 	free(request);
355 }
356 
357 /*
358  * send a HTTP/1.1 408 response if we timeout.
359  */
360 /* ARGSUSED */
361 static void
362 alarmer(int sig)
363 {
364 	alarmhit = 1;
365 }
366 
367 /*
368  * add or merge this header (val: str) into the requests list
369  */
370 static bozoheaders_t *
371 addmerge_header(bozo_httpreq_t *request, char *val,
372 		char *str, ssize_t len)
373 {
374 	struct	bozoheaders *hdr;
375 
376 	USE_ARG(len);
377 	/* do we exist already? */
378 	SIMPLEQ_FOREACH(hdr, &request->hr_headers, h_next) {
379 		if (strcasecmp(val, hdr->h_header) == 0)
380 			break;
381 	}
382 
383 	if (hdr) {
384 		/* yup, merge it in */
385 		char *nval;
386 
387 		if (asprintf(&nval, "%s, %s", hdr->h_value, str) == -1) {
388 			(void)bozo_http_error(request->hr_httpd, 500, NULL,
389 			     "memory allocation failure");
390 			return NULL;
391 		}
392 		free(hdr->h_value);
393 		hdr->h_value = nval;
394 	} else {
395 		/* nope, create a new one */
396 
397 		hdr = bozomalloc(request->hr_httpd, sizeof *hdr);
398 		hdr->h_header = bozostrdup(request->hr_httpd, val);
399 		if (str && *str)
400 			hdr->h_value = bozostrdup(request->hr_httpd, str);
401 		else
402 			hdr->h_value = bozostrdup(request->hr_httpd, " ");
403 
404 		SIMPLEQ_INSERT_TAIL(&request->hr_headers, hdr, h_next);
405 		request->hr_nheaders++;
406 	}
407 
408 	return hdr;
409 }
410 
411 /*
412  * as the prototype string is not constant (eg, "HTTP/1.1" is equivalent
413  * to "HTTP/001.01"), we MUST parse this.
414  */
415 static int
416 process_proto(bozo_httpreq_t *request, const char *proto)
417 {
418 	char	majorstr[16], *minorstr;
419 	int	majorint, minorint;
420 
421 	if (proto == NULL) {
422 got_proto_09:
423 		request->hr_proto = request->hr_httpd->consts.http_09;
424 		debug((request->hr_httpd, DEBUG_FAT, "request %s is http/0.9",
425 			request->hr_file));
426 		return 0;
427 	}
428 
429 	if (strncasecmp(proto, "HTTP/", 5) != 0)
430 		goto bad;
431 	strncpy(majorstr, proto + 5, sizeof majorstr);
432 	majorstr[sizeof(majorstr)-1] = 0;
433 	minorstr = strchr(majorstr, '.');
434 	if (minorstr == NULL)
435 		goto bad;
436 	*minorstr++ = 0;
437 
438 	majorint = atoi(majorstr);
439 	minorint = atoi(minorstr);
440 
441 	switch (majorint) {
442 	case 0:
443 		if (minorint != 9)
444 			break;
445 		goto got_proto_09;
446 	case 1:
447 		if (minorint == 0)
448 			request->hr_proto = request->hr_httpd->consts.http_10;
449 		else if (minorint == 1)
450 			request->hr_proto = request->hr_httpd->consts.http_11;
451 		else
452 			break;
453 
454 		debug((request->hr_httpd, DEBUG_FAT, "request %s is %s",
455 		    request->hr_file, request->hr_proto));
456 		SIMPLEQ_INIT(&request->hr_headers);
457 		request->hr_nheaders = 0;
458 		return 0;
459 	}
460 bad:
461 	return bozo_http_error(request->hr_httpd, 404, NULL, "unknown prototype");
462 }
463 
464 /*
465  * process each type of HTTP method, setting this HTTP requests
466  # method type.
467  */
468 static struct method_map {
469 	const char *name;
470 	int	type;
471 } method_map[] = {
472 	{ "GET", 	HTTP_GET, },
473 	{ "POST",	HTTP_POST, },
474 	{ "HEAD",	HTTP_HEAD, },
475 #if 0	/* other non-required http/1.1 methods */
476 	{ "OPTIONS",	HTTP_OPTIONS, },
477 	{ "PUT",	HTTP_PUT, },
478 	{ "DELETE",	HTTP_DELETE, },
479 	{ "TRACE",	HTTP_TRACE, },
480 	{ "CONNECT",	HTTP_CONNECT, },
481 #endif
482 	{ NULL,		0, },
483 };
484 
485 static int
486 process_method(bozo_httpreq_t *request, const char *method)
487 {
488 	struct	method_map *mmp;
489 
490 	if (request->hr_proto == request->hr_httpd->consts.http_11)
491 		request->hr_allow = "GET, HEAD, POST";
492 
493 	for (mmp = method_map; mmp->name; mmp++)
494 		if (strcasecmp(method, mmp->name) == 0) {
495 			request->hr_method = mmp->type;
496 			request->hr_methodstr = mmp->name;
497 			return 0;
498 		}
499 
500 	return bozo_http_error(request->hr_httpd, 404, request, "unknown method");
501 }
502 
503 /*
504  * This function reads a http request from stdin, returning a pointer to a
505  * bozo_httpreq_t structure, describing the request.
506  */
507 bozo_httpreq_t *
508 bozo_read_request(bozohttpd_t *httpd)
509 {
510 	struct	sigaction	sa;
511 	char	*str, *val, *method, *file, *proto, *query;
512 	char	*host, *addr, *port;
513 	char	bufport[10];
514 	char	hbuf[NI_MAXHOST], abuf[NI_MAXHOST];
515 	struct	sockaddr_storage ss;
516 	ssize_t	len;
517 	int	line = 0;
518 	socklen_t slen;
519 	bozo_httpreq_t *request;
520 
521 	/*
522 	 * if we're in daemon mode, bozo_daemon_fork() will return here twice
523 	 * for each call.  once in the child, returning 0, and once in the
524 	 * parent, returning 1.  for each child, then we can setup SSL, and
525 	 * the parent can signal the caller there was no request to process
526 	 * and it will wait for another.
527 	 */
528 	if (bozo_daemon_fork(httpd))
529 		return NULL;
530 	bozo_ssl_accept(httpd);
531 
532 	request = bozomalloc(httpd, sizeof(*request));
533 	memset(request, 0, sizeof(*request));
534 	request->hr_httpd = httpd;
535 	request->hr_allow = request->hr_host = NULL;
536 	request->hr_content_type = request->hr_content_length = NULL;
537 	request->hr_range = NULL;
538 	request->hr_last_byte_pos = -1;
539 	request->hr_if_modified_since = NULL;
540 	request->hr_virthostname = NULL;
541 	request->hr_file = NULL;
542 	request->hr_oldfile = NULL;
543 
544 	slen = sizeof(ss);
545 	if (getpeername(0, (struct sockaddr *)(void *)&ss, &slen) < 0)
546 		host = addr = NULL;
547 	else {
548 		if (getnameinfo((struct sockaddr *)(void *)&ss, slen,
549 		    abuf, sizeof abuf, NULL, 0, NI_NUMERICHOST) == 0)
550 			addr = abuf;
551 		else
552 			addr = NULL;
553 		if (httpd->numeric == 0 &&
554 		    getnameinfo((struct sockaddr *)(void *)&ss, slen,
555 				hbuf, sizeof hbuf, NULL, 0, 0) == 0)
556 			host = hbuf;
557 		else
558 			host = NULL;
559 	}
560 	if (host != NULL)
561 		request->hr_remotehost = bozostrdup(request->hr_httpd, host);
562 	if (addr != NULL)
563 		request->hr_remoteaddr = bozostrdup(request->hr_httpd, addr);
564 	slen = sizeof(ss);
565 
566 	/*
567 	 * Override the bound port from the request value, so it works even
568 	 * if passed through a proxy that doesn't rewrite the port.
569 	 */
570 	if (httpd->bindport) {
571 		if (strcmp(httpd->bindport, "80") != 0)
572 			port = httpd->bindport;
573 		else
574 			port = NULL;
575 	} else {
576 		if (getsockname(0, (struct sockaddr *)(void *)&ss, &slen) < 0)
577 			port = NULL;
578 		else {
579 			if (getnameinfo((struct sockaddr *)(void *)&ss, slen, NULL, 0,
580 					bufport, sizeof bufport, NI_NUMERICSERV) == 0)
581 				port = bufport;
582 			else
583 				port = NULL;
584 		}
585 	}
586 	if (port != NULL)
587 		request->hr_serverport = bozostrdup(request->hr_httpd, port);
588 
589 	/*
590 	 * setup a timer to make sure the request is not hung
591 	 */
592 	sa.sa_handler = alarmer;
593 	sigemptyset(&sa.sa_mask);
594 	sigaddset(&sa.sa_mask, SIGALRM);
595 	sa.sa_flags = 0;
596 	sigaction(SIGALRM, &sa, NULL);	/* XXX */
597 
598 	alarm(MAX_WAIT_TIME);
599 	while ((str = bozodgetln(httpd, STDIN_FILENO, &len, bozo_read)) != NULL) {
600 		alarm(0);
601 		if (alarmhit) {
602 			(void)bozo_http_error(httpd, 408, NULL,
603 					"request timed out");
604 			goto cleanup;
605 		}
606 		line++;
607 
608 		if (line == 1) {
609 
610 			if (len < 1) {
611 				(void)bozo_http_error(httpd, 404, NULL,
612 						"null method");
613 				goto cleanup;
614 			}
615 
616 			bozo_warn(httpd, "got request ``%s'' from host %s to port %s",
617 				str,
618 				host ? host : addr ? addr : "<local>",
619 				port ? port : "<stdin>");
620 
621 			/* we allocate return space in file and query only */
622 			parse_request(httpd, str, &method, &file, &query, &proto);
623 			request->hr_file = file;
624 			request->hr_query = query;
625 			if (method == NULL) {
626 				(void)bozo_http_error(httpd, 404, NULL,
627 						"null method");
628 				goto cleanup;
629 			}
630 			if (file == NULL) {
631 				(void)bozo_http_error(httpd, 404, NULL,
632 						"null file");
633 				goto cleanup;
634 			}
635 
636 			/*
637 			 * note that we parse the proto first, so that we
638 			 * can more properly parse the method and the url.
639 			 */
640 
641 			if (process_proto(request, proto) ||
642 			    process_method(request, method)) {
643 				goto cleanup;
644 			}
645 
646 			debug((httpd, DEBUG_FAT, "got file \"%s\" query \"%s\"",
647 			    request->hr_file,
648 			    request->hr_query ? request->hr_query : "<none>"));
649 
650 			/* http/0.9 has no header processing */
651 			if (request->hr_proto == httpd->consts.http_09)
652 				break;
653 		} else {		/* incoming headers */
654 			bozoheaders_t *hdr;
655 
656 			if (*str == '\0')
657 				break;
658 
659 			val = bozostrnsep(&str, ":", &len);
660 			debug((httpd, DEBUG_EXPLODING,
661 			    "read_req2: after bozostrnsep: str ``%s'' val ``%s''",
662 			    str, val));
663 			if (val == NULL || len == -1) {
664 				(void)bozo_http_error(httpd, 404, request,
665 						"no header");
666 				goto cleanup;
667 			}
668 			while (*str == ' ' || *str == '\t')
669 				len--, str++;
670 			while (*val == ' ' || *val == '\t')
671 				val++;
672 
673 			if (bozo_auth_check_headers(request, val, str, len))
674 				goto next_header;
675 
676 			hdr = addmerge_header(request, val, str, len);
677 
678 			if (strcasecmp(hdr->h_header, "content-type") == 0)
679 				request->hr_content_type = hdr->h_value;
680 			else if (strcasecmp(hdr->h_header, "content-length") == 0)
681 				request->hr_content_length = hdr->h_value;
682 			else if (strcasecmp(hdr->h_header, "host") == 0)
683 				request->hr_host = bozostrdup(httpd, hdr->h_value);
684 			/* RFC 2616 (HTTP/1.1): 14.20 */
685 			else if (strcasecmp(hdr->h_header, "expect") == 0) {
686 				(void)bozo_http_error(httpd, 417, request,
687 						"we don't support Expect:");
688 				goto cleanup;
689 			}
690 			else if (strcasecmp(hdr->h_header, "referrer") == 0 ||
691 			         strcasecmp(hdr->h_header, "referer") == 0)
692 				request->hr_referrer = hdr->h_value;
693 			else if (strcasecmp(hdr->h_header, "range") == 0)
694 				request->hr_range = hdr->h_value;
695 			else if (strcasecmp(hdr->h_header,
696 					"if-modified-since") == 0)
697 				request->hr_if_modified_since = hdr->h_value;
698 			else if (strcasecmp(hdr->h_header,
699 					"accept-encoding") == 0)
700 				request->hr_accept_encoding = hdr->h_value;
701 
702 			debug((httpd, DEBUG_FAT, "adding header %s: %s",
703 			    hdr->h_header, hdr->h_value));
704 		}
705 next_header:
706 		alarm(MAX_WAIT_TIME);
707 	}
708 
709 	/* now, clear it all out */
710 	alarm(0);
711 	signal(SIGALRM, SIG_DFL);
712 
713 	/* RFC1945, 8.3 */
714 	if (request->hr_method == HTTP_POST &&
715 	    request->hr_content_length == NULL) {
716 		(void)bozo_http_error(httpd, 400, request,
717 				"missing content length");
718 		goto cleanup;
719 	}
720 
721 	/* RFC 2616 (HTTP/1.1), 14.23 & 19.6.1.1 */
722 	if (request->hr_proto == httpd->consts.http_11 &&
723 	    /*(strncasecmp(request->hr_file, "http://", 7) != 0) &&*/
724 	    request->hr_host == NULL) {
725 		(void)bozo_http_error(httpd, 400, request,
726 				"missing Host header");
727 		goto cleanup;
728 	}
729 
730 	if (request->hr_range != NULL) {
731 		debug((httpd, DEBUG_FAT, "hr_range: %s", request->hr_range));
732 		/* support only simple ranges %d- and %d-%d */
733 		if (strchr(request->hr_range, ',') == NULL) {
734 			const char *rstart, *dash;
735 
736 			rstart = strchr(request->hr_range, '=');
737 			if (rstart != NULL) {
738 				rstart++;
739 				dash = strchr(rstart, '-');
740 				if (dash != NULL && dash != rstart) {
741 					dash++;
742 					request->hr_have_range = 1;
743 					request->hr_first_byte_pos =
744 					    strtoll(rstart, NULL, 10);
745 					if (request->hr_first_byte_pos < 0)
746 						request->hr_first_byte_pos = 0;
747 					if (*dash != '\0') {
748 						request->hr_last_byte_pos =
749 						    strtoll(dash, NULL, 10);
750 						if (request->hr_last_byte_pos < 0)
751 							request->hr_last_byte_pos = -1;
752 					}
753 				}
754 			}
755 		}
756 	}
757 
758 	debug((httpd, DEBUG_FAT, "bozo_read_request returns url %s in request",
759 	       request->hr_file));
760 	return request;
761 
762 cleanup:
763 	bozo_clean_request(request);
764 
765 	return NULL;
766 }
767 
768 static int
769 mmap_and_write_part(bozohttpd_t *httpd, int fd, off_t first_byte_pos, size_t sz)
770 {
771 	size_t mappedsz, wroffset;
772 	off_t mappedoffset;
773 	char *addr;
774 	void *mappedaddr;
775 
776 	/*
777 	 * we need to ensure that both the size *and* offset arguments to
778 	 * mmap() are page-aligned.  our formala for this is:
779 	 *
780 	 *    input offset: first_byte_pos
781 	 *    input size: sz
782 	 *
783 	 *    mapped offset = page align truncate (input offset)
784 	 *    mapped size   =
785 	 *        page align extend (input offset - mapped offset + input size)
786 	 *    write offset  = input offset - mapped offset
787 	 *
788 	 * we use the write offset in all writes
789 	 */
790 	mappedoffset = first_byte_pos & ~(httpd->page_size - 1);
791 	mappedsz = (size_t)
792 		(first_byte_pos - mappedoffset + sz + httpd->page_size - 1) &
793 		~(httpd->page_size - 1);
794 	wroffset = (size_t)(first_byte_pos - mappedoffset);
795 
796 	addr = mmap(0, mappedsz, PROT_READ, MAP_SHARED, fd, mappedoffset);
797 	if (addr == (char *)-1) {
798 		bozo_warn(httpd, "mmap failed: %s", strerror(errno));
799 		return -1;
800 	}
801 	mappedaddr = addr;
802 
803 #ifdef MADV_SEQUENTIAL
804 	(void)madvise(addr, sz, MADV_SEQUENTIAL);
805 #endif
806 	while (sz > BOZO_WRSZ) {
807 		if (bozo_write(httpd, STDOUT_FILENO, addr + wroffset,
808 				BOZO_WRSZ) != BOZO_WRSZ) {
809 			bozo_warn(httpd, "write failed: %s", strerror(errno));
810 			goto out;
811 		}
812 		debug((httpd, DEBUG_OBESE, "wrote %d bytes", BOZO_WRSZ));
813 		sz -= BOZO_WRSZ;
814 		addr += BOZO_WRSZ;
815 	}
816 	if (sz && (size_t)bozo_write(httpd, STDOUT_FILENO, addr + wroffset,
817 				sz) != sz) {
818 		bozo_warn(httpd, "final write failed: %s", strerror(errno));
819 		goto out;
820 	}
821 	debug((httpd, DEBUG_OBESE, "wrote %d bytes", (int)sz));
822  out:
823 	if (munmap(mappedaddr, mappedsz) < 0) {
824 		bozo_warn(httpd, "munmap failed");
825 		return -1;
826 	}
827 
828 	return 0;
829 }
830 
831 static int
832 parse_http_date(const char *val, time_t *timestamp)
833 {
834 	char *remainder;
835 	struct tm tm;
836 
837 	if ((remainder = strptime(val, "%a, %d %b %Y %T GMT", &tm)) == NULL &&
838 	    (remainder = strptime(val, "%a, %d-%b-%y %T GMT", &tm)) == NULL &&
839 	    (remainder = strptime(val, "%a %b %d %T %Y", &tm)) == NULL)
840 		return 0; /* Invalid HTTP date format */
841 
842 	if (*remainder)
843 		return 0; /* No trailing garbage */
844 
845 	*timestamp = timegm(&tm);
846 	return 1;
847 }
848 
849 /*
850  * given an url, encode it ala rfc 3986.  ie, escape ? and friends.
851  * note that this function returns a static buffer, and thus needs
852  * to be updated for any sort of parallel processing.
853  */
854 char *
855 bozo_escape_rfc3986(bozohttpd_t *httpd, const char *url)
856 {
857 	static char *buf;
858 	static size_t buflen = 0;
859 	size_t len;
860 	const char *s;
861 	char *d;
862 
863 	len = strlen(url);
864 	if (buflen < len * 3 + 1) {
865 		buflen = len * 3 + 1;
866 		buf = bozorealloc(httpd, buf, buflen);
867 	}
868 
869 	if (url == NULL) {
870 		buf[0] = 0;
871 		return buf;
872 	}
873 
874 	for (len = 0, s = url, d = buf; *s;) {
875 		if (*s & 0x80)
876 			goto encode_it;
877 		switch (*s) {
878 		case ':':
879 		case '/':
880 		case '?':
881 		case '#':
882 		case '[':
883 		case ']':
884 		case '@':
885 		case '!':
886 		case '$':
887 		case '&':
888 		case '\'':
889 		case '(':
890 		case ')':
891 		case '*':
892 		case '+':
893 		case ',':
894 		case ';':
895 		case '=':
896 		case '%':
897 		encode_it:
898 			snprintf(d, 4, "%%%2X", *s++);
899 			d += 3;
900 			len += 3;
901 			break;
902 		default:
903 			*d++ = *s++;
904 			len++;
905 			break;
906 		}
907 	}
908 	buf[len] = 0;
909 
910 	return buf;
911 }
912 
913 /*
914  * checks to see if this request has a valid .bzdirect file.  returns
915  * 0 on failure and 1 on success.
916  */
917 static int
918 check_direct_access(bozo_httpreq_t *request)
919 {
920 	FILE *fp;
921 	struct stat sb;
922 	char dir[MAXPATHLEN], dirfile[MAXPATHLEN], *basename;
923 
924 	snprintf(dir, sizeof(dir), "%s", request->hr_file + 1);
925 	debug((request->hr_httpd, DEBUG_FAT, "check_direct_access: dir %s", dir));
926 	basename = strrchr(dir, '/');
927 
928 	if ((!basename || basename[1] != '\0') &&
929 	    lstat(dir, &sb) == 0 && S_ISDIR(sb.st_mode))
930 		/* nothing */;
931 	else if (basename == NULL)
932 		strcpy(dir, ".");
933 	else {
934 		*basename++ = '\0';
935 		bozo_check_special_files(request, basename);
936 	}
937 
938 	snprintf(dirfile, sizeof(dirfile), "%s/%s", dir, DIRECT_ACCESS_FILE);
939 	if (stat(dirfile, &sb) < 0 ||
940 	    (fp = fopen(dirfile, "r")) == NULL)
941 		return 0;
942 	fclose(fp);
943 	return 1;
944 }
945 
946 /*
947  * do automatic redirection -- if there are query parameters for the URL
948  * we will tack these on to the new (redirected) URL.
949  */
950 static void
951 handle_redirect(bozo_httpreq_t *request,
952 		const char *url, int absolute)
953 {
954 	bozohttpd_t *httpd = request->hr_httpd;
955 	char *urlbuf;
956 	char portbuf[20];
957 	const char *hostname = BOZOHOST(httpd, request);
958 	int query = 0;
959 
960 	if (url == NULL) {
961 		if (asprintf(&urlbuf, "/%s/", request->hr_file) < 0)
962 			bozo_err(httpd, 1, "asprintf");
963 		url = urlbuf;
964 	} else
965 		urlbuf = NULL;
966 	url = bozo_escape_rfc3986(request->hr_httpd, url);
967 
968 	if (request->hr_query && strlen(request->hr_query))
969 		query = 1;
970 
971 	if (request->hr_serverport && strcmp(request->hr_serverport, "80") != 0)
972 		snprintf(portbuf, sizeof(portbuf), ":%s",
973 		    request->hr_serverport);
974 	else
975 		portbuf[0] = '\0';
976 	if (absolute)
977 		bozo_warn(httpd, "redirecting %s", url);
978 	else
979 		bozo_warn(httpd, "redirecting %s%s%s", hostname, portbuf, url);
980 	debug((httpd, DEBUG_FAT, "redirecting %s", url));
981 	bozo_printf(httpd, "%s 301 Document Moved\r\n", request->hr_proto);
982 	if (request->hr_proto != httpd->consts.http_09)
983 		bozo_print_header(request, NULL, "text/html", NULL);
984 	if (request->hr_proto != httpd->consts.http_09) {
985 		bozo_printf(httpd, "Location: http://");
986 		if (absolute == 0)
987 			bozo_printf(httpd, "%s%s", hostname, portbuf);
988 		if (query) {
989 			bozo_printf(httpd, "%s?%s\r\n", url, request->hr_query);
990 		} else {
991 			bozo_printf(httpd, "%s\r\n", url);
992 		}
993 	}
994 	bozo_printf(httpd, "\r\n");
995 	if (request->hr_method == HTTP_HEAD)
996 		goto head;
997 	bozo_printf(httpd, "<html><head><title>Document Moved</title></head>\n");
998 	bozo_printf(httpd, "<body><h1>Document Moved</h1>\n");
999 	bozo_printf(httpd, "This document had moved <a href=\"http://");
1000 	if (query) {
1001 		if (absolute)
1002 			bozo_printf(httpd, "%s?%s", url, request->hr_query);
1003 		else
1004 			bozo_printf(httpd, "%s%s%s?%s", hostname,
1005 				    portbuf, url, request->hr_query);
1006 	} else {
1007 		if (absolute)
1008 			bozo_printf(httpd, "%s", url);
1009 		else
1010 			bozo_printf(httpd, "%s%s%s", hostname,
1011 				    portbuf, url);
1012 	}
1013 	bozo_printf(httpd, "\">here</a>\n");
1014 	bozo_printf(httpd, "</body></html>\n");
1015 head:
1016 	bozo_flush(httpd, stdout);
1017 	free(urlbuf);
1018 }
1019 
1020 /*
1021  * deal with virtual host names; we do this:
1022  *	if we have a virtual path root (httpd->virtbase), and we are given a
1023  *	virtual host spec (Host: ho.st or http://ho.st/), see if this
1024  *	directory exists under httpd->virtbase.  if it does, use this as the
1025  #	new slashdir.
1026  */
1027 static int
1028 check_virtual(bozo_httpreq_t *request)
1029 {
1030 	bozohttpd_t *httpd = request->hr_httpd;
1031 	char *file = request->hr_file, *s;
1032 	size_t len;
1033 
1034 	if (!httpd->virtbase)
1035 		goto use_slashdir;
1036 
1037 	/*
1038 	 * convert http://virtual.host/ to request->hr_host
1039 	 */
1040 	debug((httpd, DEBUG_OBESE, "checking for http:// virtual host in ``%s''",
1041 			file));
1042 	if (strncasecmp(file, "http://", 7) == 0) {
1043 		/* we would do virtual hosting here? */
1044 		file += 7;
1045 		/* RFC 2616 (HTTP/1.1), 5.2: URI takes precedence over Host: */
1046 		free(request->hr_host);
1047 		request->hr_host = bozostrdup(request->hr_httpd, file);
1048 		if ((s = strchr(request->hr_host, '/')) != NULL)
1049 			*s = '\0';
1050 		s = strchr(file, '/');
1051 		free(request->hr_file);
1052 		request->hr_file = bozostrdup(request->hr_httpd, s ? s : "/");
1053 		debug((httpd, DEBUG_OBESE, "got host ``%s'' file is now ``%s''",
1054 		    request->hr_host, request->hr_file));
1055 	} else if (!request->hr_host)
1056 		goto use_slashdir;
1057 
1058 	/*
1059 	 * canonicalise hr_host - that is, remove any :80.
1060 	 */
1061 	len = strlen(request->hr_host);
1062 	if (len > 3 && strcmp(request->hr_host + len - 3, ":80") == 0) {
1063 		request->hr_host[len - 3] = '\0';
1064 		len = strlen(request->hr_host);
1065 	}
1066 
1067 	/*
1068 	 * ok, we have a virtual host, use opendir(3) to find a case
1069 	 * insensitive match for the virtual host we are asked for.
1070 	 * note that if the virtual host is the same as the master,
1071 	 * we don't need to do anything special.
1072 	 */
1073 	debug((httpd, DEBUG_OBESE,
1074 	    "check_virtual: checking host `%s' under httpd->virtbase `%s' "
1075 	    "for file `%s'",
1076 	    request->hr_host, httpd->virtbase, request->hr_file));
1077 	if (strncasecmp(httpd->virthostname, request->hr_host, len) != 0) {
1078 		s = 0;
1079 		DIR *dirp;
1080 		struct dirent *d;
1081 
1082 		if ((dirp = opendir(httpd->virtbase)) != NULL) {
1083 			while ((d = readdir(dirp)) != NULL) {
1084 				if (strcmp(d->d_name, ".") == 0 ||
1085 				    strcmp(d->d_name, "..") == 0) {
1086 					continue;
1087 				}
1088 				debug((httpd, DEBUG_OBESE, "looking at dir``%s''",
1089 			 	   d->d_name));
1090 				if (strncasecmp(d->d_name, request->hr_host,
1091 				    len) == 0) {
1092 					/* found it, punch it */
1093 					debug((httpd, DEBUG_OBESE, "found it punch it"));
1094 					request->hr_virthostname =
1095 					    bozostrdup(httpd, d->d_name);
1096 					if (asprintf(&s, "%s/%s", httpd->virtbase,
1097 					    request->hr_virthostname) < 0)
1098 						bozo_err(httpd, 1, "asprintf");
1099 					break;
1100 				}
1101 			}
1102 			closedir(dirp);
1103 		}
1104 		else {
1105 			debug((httpd, DEBUG_FAT, "opendir %s failed: %s",
1106 			    httpd->virtbase, strerror(errno)));
1107 		}
1108 		if (s == 0) {
1109 			if (httpd->unknown_slash)
1110 				goto use_slashdir;
1111 			return bozo_http_error(httpd, 404, request,
1112 						"unknown URL");
1113 		}
1114 	} else
1115 use_slashdir:
1116 		s = httpd->slashdir;
1117 
1118 	/*
1119 	 * ok, nailed the correct slashdir, chdir to it
1120 	 */
1121 	if (chdir(s) < 0)
1122 		return bozo_http_error(httpd, 404, request,
1123 					"can't chdir to slashdir");
1124 	return 0;
1125 }
1126 
1127 /*
1128  * checks to see if this request has a valid .bzredirect file.  returns
1129  * 0 when no redirection happend, or 1 when handle_redirect() has been
1130  * called.
1131  */
1132 static int
1133 check_bzredirect(bozo_httpreq_t *request)
1134 {
1135 	struct stat sb;
1136 	char dir[MAXPATHLEN], redir[MAXPATHLEN], redirpath[MAXPATHLEN + 1],
1137 	    path[MAXPATHLEN];
1138 	char *basename, *finalredir;
1139 	int rv, absolute;
1140 
1141 	/*
1142 	 * if this pathname is really a directory, but doesn't end in /,
1143 	 * use it as the directory to look for the redir file.
1144 	 */
1145 	snprintf(dir, sizeof(dir), "%s", request->hr_file + 1);
1146 	debug((request->hr_httpd, DEBUG_FAT, "check_bzredirect: dir %s", dir));
1147 	basename = strrchr(dir, '/');
1148 
1149 	if ((!basename || basename[1] != '\0') &&
1150 	    lstat(dir, &sb) == 0 && S_ISDIR(sb.st_mode))
1151 		/* nothing */;
1152 	else if (basename == NULL)
1153 		strcpy(dir, ".");
1154 	else {
1155 		*basename++ = '\0';
1156 		bozo_check_special_files(request, basename);
1157 	}
1158 
1159 	snprintf(redir, sizeof(redir), "%s/%s", dir, REDIRECT_FILE);
1160 	if (lstat(redir, &sb) == 0) {
1161 		if (!S_ISLNK(sb.st_mode))
1162 			return 0;
1163 		absolute = 0;
1164 	} else {
1165 		snprintf(redir, sizeof(redir), "%s/%s", dir, ABSREDIRECT_FILE);
1166 		if (lstat(redir, &sb) < 0 || !S_ISLNK(sb.st_mode))
1167 			return 0;
1168 		absolute = 1;
1169 	}
1170 	debug((request->hr_httpd, DEBUG_FAT,
1171 	       "check_bzredirect: calling readlink"));
1172 	rv = readlink(redir, redirpath, sizeof redirpath - 1);
1173 	if (rv == -1 || rv == 0) {
1174 		debug((request->hr_httpd, DEBUG_FAT, "readlink failed"));
1175 		return 0;
1176 	}
1177 	redirpath[rv] = '\0';
1178 	debug((request->hr_httpd, DEBUG_FAT,
1179 	       "readlink returned \"%s\"", redirpath));
1180 
1181 	/* check if we need authentication */
1182 	snprintf(path, sizeof(path), "%s/", dir);
1183 	if (bozo_auth_check(request, path))
1184 		return 1;
1185 
1186 	/* now we have the link pointer, redirect to the real place */
1187 	if (absolute)
1188 		finalredir = redirpath;
1189 	else
1190 		snprintf(finalredir = redir, sizeof(redir), "/%s/%s", dir,
1191 			 redirpath);
1192 
1193 	debug((request->hr_httpd, DEBUG_FAT,
1194 	       "check_bzredirect: new redir %s", finalredir));
1195 	handle_redirect(request, finalredir, absolute);
1196 	return 1;
1197 }
1198 
1199 /* this fixes the %HH hack that RFC2396 requires.  */
1200 static int
1201 fix_url_percent(bozo_httpreq_t *request)
1202 {
1203 	bozohttpd_t *httpd = request->hr_httpd;
1204 	char	*s, *t, buf[3], *url;
1205 	char	*end;	/* if end is not-zero, we don't translate beyond that */
1206 
1207 	url = request->hr_file;
1208 
1209 	end = url + strlen(url);
1210 
1211 	/* fast forward to the first % */
1212 	if ((s = strchr(url, '%')) == NULL)
1213 		return 0;
1214 
1215 	t = s;
1216 	do {
1217 		if (end && s >= end) {
1218 			debug((httpd, DEBUG_EXPLODING,
1219 				"fu_%%: past end, filling out.."));
1220 			while (*s)
1221 				*t++ = *s++;
1222 			break;
1223 		}
1224 		debug((httpd, DEBUG_EXPLODING,
1225 			"fu_%%: got s == %%, s[1]s[2] == %c%c",
1226 			s[1], s[2]));
1227 		if (s[1] == '\0' || s[2] == '\0') {
1228 			(void)bozo_http_error(httpd, 400, request,
1229 			    "percent hack missing two chars afterwards");
1230 			return 1;
1231 		}
1232 		if (s[1] == '0' && s[2] == '0') {
1233 			(void)bozo_http_error(httpd, 404, request,
1234 					"percent hack was %00");
1235 			return 1;
1236 		}
1237 		if (s[1] == '2' && s[2] == 'f') {
1238 			(void)bozo_http_error(httpd, 404, request,
1239 					"percent hack was %2f (/)");
1240 			return 1;
1241 		}
1242 
1243 		buf[0] = *++s;
1244 		buf[1] = *++s;
1245 		buf[2] = '\0';
1246 		s++;
1247 		*t = (char)strtol(buf, NULL, 16);
1248 		debug((httpd, DEBUG_EXPLODING,
1249 				"fu_%%: strtol put '%02x' into *t", *t));
1250 		if (*t++ == '\0') {
1251 			(void)bozo_http_error(httpd, 400, request,
1252 					"percent hack got a 0 back");
1253 			return 1;
1254 		}
1255 
1256 		while (*s && *s != '%') {
1257 			if (end && s >= end)
1258 				break;
1259 			*t++ = *s++;
1260 		}
1261 	} while (*s);
1262 	*t = '\0';
1263 
1264 	debug((httpd, DEBUG_FAT, "fix_url_percent returns %s in url",
1265 			request->hr_file));
1266 
1267 	return 0;
1268 }
1269 
1270 /*
1271  * transform_request does this:
1272  *	- ``expand'' %20 crapola
1273  *	- punt if it doesn't start with /
1274  *	- check httpd->untrustedref / referrer
1275  *	- look for "http://myname/" and deal with it.
1276  *	- maybe call bozo_process_cgi()
1277  *	- check for ~user and call bozo_user_transform() if so
1278  *	- if the length > 1, check for trailing slash.  if so,
1279  *	  add the index.html file
1280  *	- if the length is 1, return the index.html file
1281  *	- disallow anything ending up with a file starting
1282  *	  at "/" or having ".." in it.
1283  *	- anything else is a really weird internal error
1284  *	- returns malloced file to serve, if unhandled
1285  */
1286 static int
1287 transform_request(bozo_httpreq_t *request, int *isindex)
1288 {
1289 	bozohttpd_t *httpd = request->hr_httpd;
1290 	char	*file, *newfile = NULL;
1291 	size_t	len;
1292 	const char *hostname = BOZOHOST(httpd, request);
1293 
1294 	file = NULL;
1295 	*isindex = 0;
1296 	debug((httpd, DEBUG_FAT, "tf_req: file %s", request->hr_file));
1297 	if (fix_url_percent(request)) {
1298 		goto bad_done;
1299 	}
1300 	if (check_virtual(request)) {
1301 		goto bad_done;
1302 	}
1303 	file = request->hr_file;
1304 
1305 	if (file[0] != '/') {
1306 		(void)bozo_http_error(httpd, 404, request, "unknown URL");
1307 		goto bad_done;
1308 	}
1309 
1310 	if (check_bzredirect(request))
1311 		return 0;
1312 
1313 	if (httpd->untrustedref) {
1314 		int to_indexhtml = 0;
1315 
1316 #define TOP_PAGE(x)	(strcmp((x), "/") == 0 || \
1317 			 strcmp((x) + 1, httpd->index_html) == 0 || \
1318 			 strcmp((x) + 1, "favicon.ico") == 0)
1319 
1320 		debug((httpd, DEBUG_EXPLODING, "checking httpd->untrustedref"));
1321 		/*
1322 		 * first check that this path isn't allowed via .bzdirect file,
1323 		 * and then check referrer; make sure that people come via the
1324 		 * real name... otherwise if we aren't looking at / or
1325 		 * /index.html, redirect...  we also special case favicon.ico.
1326 		 */
1327 		if (check_direct_access(request))
1328 			/* nothing */;
1329 		else if (request->hr_referrer) {
1330 			const char *r = request->hr_referrer;
1331 
1332 			debug((httpd, DEBUG_FAT,
1333 				"checking referrer \"%s\" vs virthostname %s",
1334 				r, hostname));
1335 			if (strncmp(r, "http://", 7) != 0 ||
1336 			    (strncasecmp(r + 7, hostname,
1337 			    		 strlen(hostname)) != 0 &&
1338 			     !TOP_PAGE(file)))
1339 				to_indexhtml = 1;
1340 		} else {
1341 			const char *h = request->hr_host;
1342 
1343 			debug((httpd, DEBUG_FAT, "url has no referrer at all"));
1344 			/* if there's no referrer, let / or /index.html past */
1345 			if (!TOP_PAGE(file) ||
1346 			    (h && strncasecmp(h, hostname,
1347 			    		strlen(hostname)) != 0))
1348 				to_indexhtml = 1;
1349 		}
1350 
1351 		if (to_indexhtml) {
1352 			char *slashindexhtml;
1353 
1354 			if (asprintf(&slashindexhtml, "/%s",
1355 					httpd->index_html) < 0)
1356 				bozo_err(httpd, 1, "asprintf");
1357 			debug((httpd, DEBUG_FAT,
1358 				"httpd->untrustedref: redirecting %s to %s",
1359 				file, slashindexhtml));
1360 			handle_redirect(request, slashindexhtml, 0);
1361 			free(slashindexhtml);
1362 			return 0;
1363 		}
1364 	}
1365 
1366 	len = strlen(file);
1367 	if (/*CONSTCOND*/0) {
1368 #ifndef NO_USER_SUPPORT
1369 	} else if (len > 1 && httpd->enable_users && file[1] == '~') {
1370 		if (file[2] == '\0') {
1371 			(void)bozo_http_error(httpd, 404, request,
1372 						"missing username");
1373 			goto bad_done;
1374 		}
1375 		if (strchr(file + 2, '/') == NULL) {
1376 			handle_redirect(request, NULL, 0);
1377 			return 0;
1378 		}
1379 		debug((httpd, DEBUG_FAT, "calling bozo_user_transform"));
1380 
1381 		return bozo_user_transform(request, isindex);
1382 #endif /* NO_USER_SUPPORT */
1383 	} else if (len > 1) {
1384 		debug((httpd, DEBUG_FAT, "file[len-1] == %c", file[len-1]));
1385 		if (file[len-1] == '/') {	/* append index.html */
1386 			*isindex = 1;
1387 			debug((httpd, DEBUG_FAT, "appending index.html"));
1388 			newfile = bozomalloc(httpd,
1389 					len + strlen(httpd->index_html) + 1);
1390 			strcpy(newfile, file + 1);
1391 			strcat(newfile, httpd->index_html);
1392 		} else
1393 			newfile = bozostrdup(request->hr_httpd, file + 1);
1394 	} else if (len == 1) {
1395 		debug((httpd, DEBUG_EXPLODING, "tf_req: len == 1"));
1396 		newfile = bozostrdup(request->hr_httpd, httpd->index_html);
1397 		*isindex = 1;
1398 	} else {	/* len == 0 ? */
1399 		(void)bozo_http_error(httpd, 500, request,
1400 					"request->hr_file is nul?");
1401 		goto bad_done;
1402 	}
1403 
1404 	if (newfile == NULL) {
1405 		(void)bozo_http_error(httpd, 500, request, "internal failure");
1406 		goto bad_done;
1407 	}
1408 
1409 	/*
1410 	 * look for "http://myname/" and deal with it as necessary.
1411 	 */
1412 
1413 	/*
1414 	 * stop traversing outside our domain
1415 	 *
1416 	 * XXX true security only comes from our parent using chroot(2)
1417 	 * before execve(2)'ing us.  or our own built in chroot(2) support.
1418 	 */
1419 	if (*newfile == '/' || strcmp(newfile, "..") == 0 ||
1420 	    strstr(newfile, "/..") || strstr(newfile, "../")) {
1421 		(void)bozo_http_error(httpd, 403, request, "illegal request");
1422 		goto bad_done;
1423 	}
1424 
1425 	if (bozo_auth_check(request, newfile))
1426 		goto bad_done;
1427 
1428 	if (strlen(newfile)) {
1429 		request->hr_oldfile = request->hr_file;
1430 		request->hr_file = newfile;
1431 	}
1432 
1433 	if (bozo_process_cgi(request))
1434 		return 0;
1435 
1436 	if (bozo_process_lua(request))
1437 		return 0;
1438 
1439 	debug((httpd, DEBUG_FAT, "transform_request set: %s", newfile));
1440 	return 1;
1441 bad_done:
1442 	debug((httpd, DEBUG_FAT, "transform_request returning: 0"));
1443 	free(newfile);
1444 	return 0;
1445 }
1446 
1447 /*
1448  * can_gzip checks if the request supports and prefers gzip encoding.
1449  *
1450  * XXX: we do not consider the associated q with gzip in making our
1451  *      decision which is broken.
1452  */
1453 
1454 static int
1455 can_gzip(bozo_httpreq_t *request)
1456 {
1457 	const char	*pos;
1458 	const char	*tmp;
1459 	size_t		 len;
1460 
1461 	/* First we decide if the request can be gzipped at all. */
1462 
1463 	/* not if we already are encoded... */
1464 	tmp = bozo_content_encoding(request, request->hr_file);
1465 	if (tmp && *tmp)
1466 		return 0;
1467 
1468 	/* not if we are not asking for the whole file... */
1469 	if (request->hr_last_byte_pos != -1 || request->hr_have_range)
1470 		return 0;
1471 
1472 	/* Then we determine if gzip is on the cards. */
1473 
1474 	for (pos = request->hr_accept_encoding; pos && *pos; pos += len) {
1475 		while (*pos == ' ')
1476 			pos++;
1477 
1478 		len = strcspn(pos, ";,");
1479 
1480 		if ((len == 4 && strncasecmp("gzip", pos, 4) == 0) ||
1481 		    (len == 6 && strncasecmp("x-gzip", pos, 6) == 0))
1482 			return 1;
1483 
1484 		if (pos[len] == ';')
1485 			len += strcspn(&pos[len], ",");
1486 
1487 		if (pos[len])
1488 			len++;
1489 	}
1490 
1491 	return 0;
1492 }
1493 
1494 /*
1495  * bozo_process_request does the following:
1496  *	- check the request is valid
1497  *	- process cgi-bin if necessary
1498  *	- transform a filename if necesarry
1499  *	- return the HTTP request
1500  */
1501 void
1502 bozo_process_request(bozo_httpreq_t *request)
1503 {
1504 	bozohttpd_t *httpd = request->hr_httpd;
1505 	struct	stat sb;
1506 	time_t timestamp;
1507 	char	*file;
1508 	const char *type, *encoding;
1509 	int	fd, isindex;
1510 
1511 	/*
1512 	 * note that transform_request chdir()'s if required.  also note
1513 	 * that cgi is handed here.  if transform_request() returns 0
1514 	 * then the request has been handled already.
1515 	 */
1516 	if (transform_request(request, &isindex) == 0)
1517 		return;
1518 
1519 	fd = -1;
1520 	encoding = NULL;
1521 	if (can_gzip(request)) {
1522 		asprintf(&file, "%s.gz", request->hr_file);
1523 		fd = open(file, O_RDONLY);
1524 		if (fd >= 0)
1525 			encoding = "gzip";
1526 		free(file);
1527 	}
1528 
1529 	file = request->hr_file;
1530 
1531 	if (fd < 0)
1532 		fd = open(file, O_RDONLY);
1533 
1534 	if (fd < 0) {
1535 		debug((httpd, DEBUG_FAT, "open failed: %s", strerror(errno)));
1536 		switch(errno) {
1537 		case EPERM:
1538 			(void)bozo_http_error(httpd, 403, request,
1539 						"no permission to open file");
1540 			break;
1541 		case ENAMETOOLONG:
1542 			/*FALLTHROUGH*/
1543 		case ENOENT:
1544 			if (!bozo_dir_index(request, file, isindex))
1545 				(void)bozo_http_error(httpd, 404, request,
1546 							"no file");
1547 			break;
1548 		default:
1549 			(void)bozo_http_error(httpd, 500, request, "open file");
1550 		}
1551 		goto cleanup_nofd;
1552 	}
1553 	if (fstat(fd, &sb) < 0) {
1554 		(void)bozo_http_error(httpd, 500, request, "can't fstat");
1555 		goto cleanup;
1556 	}
1557 	if (S_ISDIR(sb.st_mode)) {
1558 		handle_redirect(request, NULL, 0);
1559 		goto cleanup;
1560 	}
1561 
1562 	if (request->hr_if_modified_since &&
1563 	    parse_http_date(request->hr_if_modified_since, &timestamp) &&
1564 	    timestamp >= sb.st_mtime) {
1565 		/* XXX ignore subsecond of timestamp */
1566 		bozo_printf(httpd, "%s 304 Not Modified\r\n",
1567 				request->hr_proto);
1568 		bozo_printf(httpd, "\r\n");
1569 		bozo_flush(httpd, stdout);
1570 		goto cleanup;
1571 	}
1572 
1573 	/* validate requested range */
1574 	if (request->hr_last_byte_pos == -1 ||
1575 	    request->hr_last_byte_pos >= sb.st_size)
1576 		request->hr_last_byte_pos = sb.st_size - 1;
1577 	if (request->hr_have_range &&
1578 	    request->hr_first_byte_pos > request->hr_last_byte_pos) {
1579 		request->hr_have_range = 0;	/* punt */
1580 		request->hr_first_byte_pos = 0;
1581 		request->hr_last_byte_pos = sb.st_size - 1;
1582 	}
1583 	debug((httpd, DEBUG_FAT, "have_range %d first_pos %lld last_pos %lld",
1584 	    request->hr_have_range,
1585 	    (long long)request->hr_first_byte_pos,
1586 	    (long long)request->hr_last_byte_pos));
1587 	if (request->hr_have_range)
1588 		bozo_printf(httpd, "%s 206 Partial Content\r\n",
1589 				request->hr_proto);
1590 	else
1591 		bozo_printf(httpd, "%s 200 OK\r\n", request->hr_proto);
1592 
1593 	if (request->hr_proto != httpd->consts.http_09) {
1594 		type = bozo_content_type(request, file);
1595 		if (!encoding)
1596 			encoding = bozo_content_encoding(request, file);
1597 
1598 		bozo_print_header(request, &sb, type, encoding);
1599 		bozo_printf(httpd, "\r\n");
1600 	}
1601 	bozo_flush(httpd, stdout);
1602 
1603 	if (request->hr_method != HTTP_HEAD) {
1604 		off_t szleft, cur_byte_pos;
1605 
1606 		szleft =
1607 		     request->hr_last_byte_pos - request->hr_first_byte_pos + 1;
1608 		cur_byte_pos = request->hr_first_byte_pos;
1609 
1610  retry:
1611 		while (szleft) {
1612 			size_t sz;
1613 
1614 			/* This should take care of the first unaligned chunk */
1615 			if ((cur_byte_pos & (httpd->page_size - 1)) != 0)
1616 				sz = (size_t)(cur_byte_pos & ~httpd->page_size);
1617 			if ((off_t)httpd->mmapsz < szleft)
1618 				sz = httpd->mmapsz;
1619 			else
1620 				sz = (size_t)szleft;
1621 			if (mmap_and_write_part(httpd, fd, cur_byte_pos, sz)) {
1622 				if (errno == ENOMEM) {
1623 					httpd->mmapsz /= 2;
1624 					if (httpd->mmapsz >= httpd->page_size)
1625 						goto retry;
1626 				}
1627 				goto cleanup;
1628 			}
1629 			cur_byte_pos += sz;
1630 			szleft -= sz;
1631 		}
1632 	}
1633  cleanup:
1634 	close(fd);
1635  cleanup_nofd:
1636 	close(STDIN_FILENO);
1637 	close(STDOUT_FILENO);
1638 	/*close(STDERR_FILENO);*/
1639 }
1640 
1641 /* make sure we're not trying to access special files */
1642 int
1643 bozo_check_special_files(bozo_httpreq_t *request, const char *name)
1644 {
1645 	bozohttpd_t *httpd = request->hr_httpd;
1646 
1647 	/* ensure basename(name) != special files */
1648 	if (strcmp(name, DIRECT_ACCESS_FILE) == 0)
1649 		return bozo_http_error(httpd, 403, request,
1650 		    "no permission to open direct access file");
1651 	if (strcmp(name, REDIRECT_FILE) == 0)
1652 		return bozo_http_error(httpd, 403, request,
1653 		    "no permission to open redirect file");
1654 	if (strcmp(name, ABSREDIRECT_FILE) == 0)
1655 		return bozo_http_error(httpd, 403, request,
1656 		    "no permission to open redirect file");
1657 	return bozo_auth_check_special_files(request, name);
1658 }
1659 
1660 /* generic header printing routine */
1661 void
1662 bozo_print_header(bozo_httpreq_t *request,
1663 		struct stat *sbp, const char *type, const char *encoding)
1664 {
1665 	bozohttpd_t *httpd = request->hr_httpd;
1666 	off_t len;
1667 	char	date[40];
1668 
1669 	bozo_printf(httpd, "Date: %s\r\n", bozo_http_date(date, sizeof(date)));
1670 	bozo_printf(httpd, "Server: %s\r\n", httpd->server_software);
1671 	bozo_printf(httpd, "Accept-Ranges: bytes\r\n");
1672 	if (sbp) {
1673 		char filedate[40];
1674 		struct	tm *tm;
1675 
1676 		tm = gmtime(&sbp->st_mtime);
1677 		strftime(filedate, sizeof filedate,
1678 		    "%a, %d %b %Y %H:%M:%S GMT", tm);
1679 		bozo_printf(httpd, "Last-Modified: %s\r\n", filedate);
1680 	}
1681 	if (type && *type)
1682 		bozo_printf(httpd, "Content-Type: %s\r\n", type);
1683 	if (encoding && *encoding)
1684 		bozo_printf(httpd, "Content-Encoding: %s\r\n", encoding);
1685 	if (sbp) {
1686 		if (request->hr_have_range) {
1687 			len = request->hr_last_byte_pos -
1688 					request->hr_first_byte_pos +1;
1689 			bozo_printf(httpd,
1690 				"Content-Range: bytes %qd-%qd/%qd\r\n",
1691 				(long long) request->hr_first_byte_pos,
1692 				(long long) request->hr_last_byte_pos,
1693 				(long long) sbp->st_size);
1694 		} else
1695 			len = sbp->st_size;
1696 		bozo_printf(httpd, "Content-Length: %qd\r\n", (long long)len);
1697 	}
1698 	if (request && request->hr_proto == httpd->consts.http_11)
1699 		bozo_printf(httpd, "Connection: close\r\n");
1700 	bozo_flush(httpd, stdout);
1701 }
1702 
1703 #ifndef NO_DEBUG
1704 void
1705 debug__(bozohttpd_t *httpd, int level, const char *fmt, ...)
1706 {
1707 	va_list	ap;
1708 	int savederrno;
1709 
1710 	/* only log if the level is low enough */
1711 	if (httpd->debug < level)
1712 		return;
1713 
1714 	savederrno = errno;
1715 	va_start(ap, fmt);
1716 	if (httpd->logstderr) {
1717 		vfprintf(stderr, fmt, ap);
1718 		fputs("\n", stderr);
1719 	} else
1720 		vsyslog(LOG_DEBUG, fmt, ap);
1721 	va_end(ap);
1722 	errno = savederrno;
1723 }
1724 #endif /* NO_DEBUG */
1725 
1726 /* these are like warn() and err(), except for syslog not stderr */
1727 void
1728 bozo_warn(bozohttpd_t *httpd, const char *fmt, ...)
1729 {
1730 	va_list ap;
1731 
1732 	va_start(ap, fmt);
1733 	if (httpd->logstderr || isatty(STDERR_FILENO)) {
1734 		//fputs("warning: ", stderr);
1735 		vfprintf(stderr, fmt, ap);
1736 		fputs("\n", stderr);
1737 	} else
1738 		vsyslog(LOG_INFO, fmt, ap);
1739 	va_end(ap);
1740 }
1741 
1742 void
1743 bozo_err(bozohttpd_t *httpd, int code, const char *fmt, ...)
1744 {
1745 	va_list ap;
1746 
1747 	va_start(ap, fmt);
1748 	if (httpd->logstderr || isatty(STDERR_FILENO)) {
1749 		//fputs("error: ", stderr);
1750 		vfprintf(stderr, fmt, ap);
1751 		fputs("\n", stderr);
1752 	} else
1753 		vsyslog(LOG_ERR, fmt, ap);
1754 	va_end(ap);
1755 	exit(code);
1756 }
1757 
1758 /*
1759  * this escapes HTML tags.  returns allocated escaped
1760  * string if needed, or NULL on allocation failure or
1761  * lack of escape need.
1762  * call with NULL httpd in error paths, to avoid recursive
1763  * malloc failure.  call with valid httpd in normal paths
1764  * to get automatic allocation failure handling.
1765  */
1766 char *
1767 bozo_escape_html(bozohttpd_t *httpd, const char *url)
1768 {
1769 	int	i, j;
1770 	char	*tmp;
1771 	size_t	len;
1772 
1773 	for (i = 0, j = 0; url[i]; i++) {
1774 		switch (url[i]) {
1775 		case '<':
1776 		case '>':
1777 			j += 4;
1778 			break;
1779 		case '&':
1780 			j += 5;
1781 			break;
1782 		}
1783 	}
1784 
1785 	if (j == 0)
1786 		return NULL;
1787 
1788 	/*
1789 	 * we need to handle being called from different
1790 	 * pathnames.
1791 	 */
1792 	len = strlen(url) + j;
1793 	if (httpd)
1794 		tmp = bozomalloc(httpd, len);
1795 	else if ((tmp = malloc(len)) == 0)
1796 			return NULL;
1797 
1798 	for (i = 0, j = 0; url[i]; i++) {
1799 		switch (url[i]) {
1800 		case '<':
1801 			memcpy(tmp + j, "&lt;", 4);
1802 			j += 4;
1803 			break;
1804 		case '>':
1805 			memcpy(tmp + j, "&gt;", 4);
1806 			j += 4;
1807 			break;
1808 		case '&':
1809 			memcpy(tmp + j, "&amp;", 5);
1810 			j += 5;
1811 			break;
1812 		default:
1813 			tmp[j++] = url[i];
1814 		}
1815 	}
1816 	tmp[j] = 0;
1817 
1818 	return tmp;
1819 }
1820 
1821 /* short map between error code, and short/long messages */
1822 static struct errors_map {
1823 	int	code;			/* HTTP return code */
1824 	const char *shortmsg;		/* short version of message */
1825 	const char *longmsg;		/* long version of message */
1826 } errors_map[] = {
1827 	{ 400,	"400 Bad Request",	"The request was not valid", },
1828 	{ 401,	"401 Unauthorized",	"No authorization", },
1829 	{ 403,	"403 Forbidden",	"Access to this item has been denied",},
1830 	{ 404, 	"404 Not Found",	"This item has not been found", },
1831 	{ 408, 	"408 Request Timeout",	"This request took too long", },
1832 	{ 417,	"417 Expectation Failed","Expectations not available", },
1833 	{ 500,	"500 Internal Error",	"An error occured on the server", },
1834 	{ 501,	"501 Not Implemented",	"This request is not available", },
1835 	{ 0,	NULL,			NULL, },
1836 };
1837 
1838 static const char *help = "DANGER! WILL ROBINSON! DANGER!";
1839 
1840 static const char *
1841 http_errors_short(int code)
1842 {
1843 	struct errors_map *ep;
1844 
1845 	for (ep = errors_map; ep->code; ep++)
1846 		if (ep->code == code)
1847 			return (ep->shortmsg);
1848 	return (help);
1849 }
1850 
1851 static const char *
1852 http_errors_long(int code)
1853 {
1854 	struct errors_map *ep;
1855 
1856 	for (ep = errors_map; ep->code; ep++)
1857 		if (ep->code == code)
1858 			return (ep->longmsg);
1859 	return (help);
1860 }
1861 
1862 /* the follow functions and variables are used in handling HTTP errors */
1863 /* ARGSUSED */
1864 int
1865 bozo_http_error(bozohttpd_t *httpd, int code, bozo_httpreq_t *request,
1866 		const char *msg)
1867 {
1868 	char portbuf[20];
1869 	const char *header = http_errors_short(code);
1870 	const char *reason = http_errors_long(code);
1871 	const char *proto = (request && request->hr_proto) ?
1872 				request->hr_proto : httpd->consts.http_11;
1873 	int	size;
1874 
1875 	debug((httpd, DEBUG_FAT, "bozo_http_error %d: %s", code, msg));
1876 	if (header == NULL || reason == NULL) {
1877 		bozo_err(httpd, 1,
1878 			"bozo_http_error() failed (short = %p, long = %p)",
1879 			header, reason);
1880 		return code;
1881 	}
1882 
1883 	if (request && request->hr_serverport &&
1884 	    strcmp(request->hr_serverport, "80") != 0)
1885 		snprintf(portbuf, sizeof(portbuf), ":%s",
1886 				request->hr_serverport);
1887 	else
1888 		portbuf[0] = '\0';
1889 
1890 	if (request && request->hr_file) {
1891 		char *file = NULL;
1892 		const char *hostname = BOZOHOST(httpd, request);
1893 
1894 		/* bozo_escape_html() failure here is just too bad. */
1895 		file = bozo_escape_html(NULL, request->hr_file);
1896 		if (file == NULL)
1897 			file = request->hr_file;
1898 		size = snprintf(httpd->errorbuf, BUFSIZ,
1899 		    "<html><head><title>%s</title></head>\n"
1900 		    "<body><h1>%s</h1>\n"
1901 		    "%s: <pre>%s</pre>\n"
1902  		    "<hr><address><a href=\"http://%s%s/\">%s%s</a></address>\n"
1903 		    "</body></html>\n",
1904 		    header, header, file, reason,
1905 		    hostname, portbuf, hostname, portbuf);
1906 		if (size >= (int)BUFSIZ) {
1907 			bozo_warn(httpd,
1908 				"bozo_http_error buffer too small, truncated");
1909 			size = (int)BUFSIZ;
1910 		}
1911 	} else
1912 		size = 0;
1913 
1914 	bozo_printf(httpd, "%s %s\r\n", proto, header);
1915 	if (request)
1916 		bozo_auth_check_401(request, code);
1917 
1918 	bozo_printf(httpd, "Content-Type: text/html\r\n");
1919 	bozo_printf(httpd, "Content-Length: %d\r\n", size);
1920 	bozo_printf(httpd, "Server: %s\r\n", httpd->server_software);
1921 	if (request && request->hr_allow)
1922 		bozo_printf(httpd, "Allow: %s\r\n", request->hr_allow);
1923 	bozo_printf(httpd, "\r\n");
1924 	/* According to the RFC 2616 sec. 9.4 HEAD method MUST NOT return a
1925 	 * message-body in the response */
1926 	if (size && request && request->hr_method != HTTP_HEAD)
1927 		bozo_printf(httpd, "%s", httpd->errorbuf);
1928 	bozo_flush(httpd, stdout);
1929 
1930 	return code;
1931 }
1932 
1933 /* Below are various modified libc functions */
1934 
1935 /*
1936  * returns -1 in lenp if the string ran out before finding a delimiter,
1937  * but is otherwise the same as strsep.  Note that the length must be
1938  * correctly passed in.
1939  */
1940 char *
1941 bozostrnsep(char **strp, const char *delim, ssize_t	*lenp)
1942 {
1943 	char	*s;
1944 	const	char *spanp;
1945 	int	c, sc;
1946 	char	*tok;
1947 
1948 	if ((s = *strp) == NULL)
1949 		return (NULL);
1950 	for (tok = s;;) {
1951 		if (lenp && --(*lenp) == -1)
1952 			return (NULL);
1953 		c = *s++;
1954 		spanp = delim;
1955 		do {
1956 			if ((sc = *spanp++) == c) {
1957 				if (c == 0)
1958 					s = NULL;
1959 				else
1960 					s[-1] = '\0';
1961 				*strp = s;
1962 				return (tok);
1963 			}
1964 		} while (sc != 0);
1965 	}
1966 	/* NOTREACHED */
1967 }
1968 
1969 /*
1970  * inspired by fgetln(3), but works for fd's.  should work identically
1971  * except it, however, does *not* return the newline, and it does nul
1972  * terminate the string.
1973  */
1974 char *
1975 bozodgetln(bozohttpd_t *httpd, int fd, ssize_t *lenp,
1976 	ssize_t (*readfn)(bozohttpd_t *, int, void *, size_t))
1977 {
1978 	ssize_t	len;
1979 	int	got_cr = 0;
1980 	char	c, *nbuffer;
1981 
1982 	/* initialise */
1983 	if (httpd->getln_buflen == 0) {
1984 		/* should be plenty for most requests */
1985 		httpd->getln_buflen = 128;
1986 		httpd->getln_buffer = malloc((size_t)httpd->getln_buflen);
1987 		if (httpd->getln_buffer == NULL) {
1988 			httpd->getln_buflen = 0;
1989 			return NULL;
1990 		}
1991 	}
1992 	len = 0;
1993 
1994 	/*
1995 	 * we *have* to read one byte at a time, to not break cgi
1996 	 * programs (for we pass stdin off to them).  could fix this
1997 	 * by becoming a fd-passing program instead of just exec'ing
1998 	 * the program
1999 	 *
2000 	 * the above is no longer true, we are the fd-passing
2001 	 * program already.
2002 	 */
2003 	for (; readfn(httpd, fd, &c, 1) == 1; ) {
2004 		debug((httpd, DEBUG_EXPLODING, "bozodgetln read %c", c));
2005 
2006 		if (len >= httpd->getln_buflen - 1) {
2007 			httpd->getln_buflen *= 2;
2008 			debug((httpd, DEBUG_EXPLODING, "bozodgetln: "
2009 				"reallocating buffer to buflen %zu",
2010 				httpd->getln_buflen));
2011 			nbuffer = bozorealloc(httpd, httpd->getln_buffer,
2012 				(size_t)httpd->getln_buflen);
2013 			httpd->getln_buffer = nbuffer;
2014 		}
2015 
2016 		httpd->getln_buffer[len++] = c;
2017 		if (c == '\r') {
2018 			got_cr = 1;
2019 			continue;
2020 		} else if (c == '\n') {
2021 			/*
2022 			 * HTTP/1.1 spec says to ignore CR and treat
2023 			 * LF as the real line terminator.  even though
2024 			 * the same spec defines CRLF as the line
2025 			 * terminator, it is recommended in section 19.3
2026 			 * to do the LF trick for tolerance.
2027 			 */
2028 			if (got_cr)
2029 				len -= 2;
2030 			else
2031 				len -= 1;
2032 			break;
2033 		}
2034 
2035 	}
2036 	httpd->getln_buffer[len] = '\0';
2037 	debug((httpd, DEBUG_OBESE, "bozodgetln returns: ``%s'' with len %zd",
2038 	       httpd->getln_buffer, len));
2039 	*lenp = len;
2040 	return httpd->getln_buffer;
2041 }
2042 
2043 void *
2044 bozorealloc(bozohttpd_t *httpd, void *ptr, size_t size)
2045 {
2046 	void	*p;
2047 
2048 	p = realloc(ptr, size);
2049 	if (p == NULL) {
2050 		(void)bozo_http_error(httpd, 500, NULL,
2051 				"memory allocation failure");
2052 		exit(1);
2053 	}
2054 	return (p);
2055 }
2056 
2057 void *
2058 bozomalloc(bozohttpd_t *httpd, size_t size)
2059 {
2060 	void	*p;
2061 
2062 	p = malloc(size);
2063 	if (p == NULL) {
2064 		(void)bozo_http_error(httpd, 500, NULL,
2065 				"memory allocation failure");
2066 		exit(1);
2067 	}
2068 	return (p);
2069 }
2070 
2071 char *
2072 bozostrdup(bozohttpd_t *httpd, const char *str)
2073 {
2074 	char	*p;
2075 
2076 	p = strdup(str);
2077 	if (p == NULL) {
2078 		(void)bozo_http_error(httpd, 500, NULL,
2079 					"memory allocation failure");
2080 		exit(1);
2081 	}
2082 	return (p);
2083 }
2084 
2085 /* set default values in bozohttpd_t struct */
2086 int
2087 bozo_init_httpd(bozohttpd_t *httpd)
2088 {
2089 	/* make sure everything is clean */
2090 	(void) memset(httpd, 0x0, sizeof(*httpd));
2091 
2092 	/* constants */
2093 	httpd->consts.http_09 = "HTTP/0.9";
2094 	httpd->consts.http_10 = "HTTP/1.0";
2095 	httpd->consts.http_11 = "HTTP/1.1";
2096 	httpd->consts.text_plain = "text/plain";
2097 
2098 	/* mmap region size */
2099 	httpd->mmapsz = BOZO_MMAPSZ;
2100 
2101 	/* error buffer for bozo_http_error() */
2102 	if ((httpd->errorbuf = malloc(BUFSIZ)) == NULL) {
2103 		(void) fprintf(stderr,
2104 			"bozohttpd: memory_allocation failure\n");
2105 		return 0;
2106 	}
2107 #ifndef NO_LUA_SUPPORT
2108 	SIMPLEQ_INIT(&httpd->lua_states);
2109 #endif
2110 	return 1;
2111 }
2112 
2113 /* set default values in bozoprefs_t struct */
2114 int
2115 bozo_init_prefs(bozoprefs_t *prefs)
2116 {
2117 	/* make sure everything is clean */
2118 	(void) memset(prefs, 0x0, sizeof(*prefs));
2119 
2120 	/* set up default values */
2121 	bozo_set_pref(prefs, "server software", SERVER_SOFTWARE);
2122 	bozo_set_pref(prefs, "index.html", INDEX_HTML);
2123 	bozo_set_pref(prefs, "public_html", PUBLIC_HTML);
2124 
2125 	return 1;
2126 }
2127 
2128 /* set default values */
2129 int
2130 bozo_set_defaults(bozohttpd_t *httpd, bozoprefs_t *prefs)
2131 {
2132 	return bozo_init_httpd(httpd) && bozo_init_prefs(prefs);
2133 }
2134 
2135 /* set the virtual host name, port and root */
2136 int
2137 bozo_setup(bozohttpd_t *httpd, bozoprefs_t *prefs, const char *vhost,
2138 		const char *root)
2139 {
2140 	struct passwd	 *pw;
2141 	extern char	**environ;
2142 	static char	 *cleanenv[1] = { NULL };
2143 	uid_t		  uid;
2144 	char		 *chrootdir;
2145 	char		 *username;
2146 	char		 *portnum;
2147 	char		 *cp;
2148 	int		  dirtyenv;
2149 
2150 	dirtyenv = 0;
2151 
2152 	if (vhost == NULL) {
2153 		httpd->virthostname = bozomalloc(httpd, MAXHOSTNAMELEN+1);
2154 		/* XXX we do not check for FQDN here */
2155 		if (gethostname(httpd->virthostname, MAXHOSTNAMELEN+1) < 0)
2156 			bozo_err(httpd, 1, "gethostname");
2157 		httpd->virthostname[MAXHOSTNAMELEN] = '\0';
2158 	} else {
2159 		httpd->virthostname = strdup(vhost);
2160 	}
2161 	httpd->slashdir = strdup(root);
2162 	if ((portnum = bozo_get_pref(prefs, "port number")) != NULL) {
2163 		httpd->bindport = strdup(portnum);
2164 	}
2165 
2166 	/* go over preferences now */
2167 	if ((cp = bozo_get_pref(prefs, "numeric")) != NULL &&
2168 	    strcmp(cp, "true") == 0) {
2169 		httpd->numeric = 1;
2170 	}
2171 	if ((cp = bozo_get_pref(prefs, "trusted referal")) != NULL &&
2172 	    strcmp(cp, "true") == 0) {
2173 		httpd->untrustedref = 1;
2174 	}
2175 	if ((cp = bozo_get_pref(prefs, "log to stderr")) != NULL &&
2176 	    strcmp(cp, "true") == 0) {
2177 		httpd->logstderr = 1;
2178 	}
2179 	if ((cp = bozo_get_pref(prefs, "bind address")) != NULL) {
2180 		httpd->bindaddress = strdup(cp);
2181 	}
2182 	if ((cp = bozo_get_pref(prefs, "background")) != NULL) {
2183 		httpd->background = atoi(cp);
2184 	}
2185 	if ((cp = bozo_get_pref(prefs, "foreground")) != NULL &&
2186 	    strcmp(cp, "true") == 0) {
2187 		httpd->foreground = 1;
2188 	}
2189 	if ((cp = bozo_get_pref(prefs, "pid file")) != NULL) {
2190 		httpd->pidfile = strdup(cp);
2191 	}
2192 	if ((cp = bozo_get_pref(prefs, "unknown slash")) != NULL &&
2193 	    strcmp(cp, "true") == 0) {
2194 		httpd->unknown_slash = 1;
2195 	}
2196 	if ((cp = bozo_get_pref(prefs, "virtual base")) != NULL) {
2197 		httpd->virtbase = strdup(cp);
2198 	}
2199 	if ((cp = bozo_get_pref(prefs, "enable users")) != NULL &&
2200 	    strcmp(cp, "true") == 0) {
2201 		httpd->enable_users = 1;
2202 	}
2203 	if ((cp = bozo_get_pref(prefs, "dirty environment")) != NULL &&
2204 	    strcmp(cp, "true") == 0) {
2205 		dirtyenv = 1;
2206 	}
2207 	if ((cp = bozo_get_pref(prefs, "hide dots")) != NULL &&
2208 	    strcmp(cp, "true") == 0) {
2209 		httpd->hide_dots = 1;
2210 	}
2211 	if ((cp = bozo_get_pref(prefs, "directory indexing")) != NULL &&
2212 	    strcmp(cp, "true") == 0) {
2213 		httpd->dir_indexing = 1;
2214 	}
2215 	if ((cp = bozo_get_pref(prefs, "public_html")) != NULL) {
2216 		httpd->public_html = strdup(cp);
2217 	}
2218 	httpd->server_software =
2219 			strdup(bozo_get_pref(prefs, "server software"));
2220 	httpd->index_html = strdup(bozo_get_pref(prefs, "index.html"));
2221 
2222 	/*
2223 	 * initialise ssl and daemon mode if necessary.
2224 	 */
2225 	bozo_ssl_init(httpd);
2226 	bozo_daemon_init(httpd);
2227 
2228 	if ((username = bozo_get_pref(prefs, "username")) == NULL) {
2229 		if ((pw = getpwuid(uid = 0)) == NULL)
2230 			bozo_err(httpd, 1, "getpwuid(0): %s", strerror(errno));
2231 		httpd->username = strdup(pw->pw_name);
2232 	} else {
2233 		httpd->username = strdup(username);
2234 		if ((pw = getpwnam(httpd->username)) == NULL)
2235 			bozo_err(httpd, 1, "getpwnam(%s): %s", httpd->username,
2236 					strerror(errno));
2237 		if (initgroups(pw->pw_name, pw->pw_gid) == -1)
2238 			bozo_err(httpd, 1, "initgroups: %s", strerror(errno));
2239 		if (setgid(pw->pw_gid) == -1)
2240 			bozo_err(httpd, 1, "setgid(%u): %s", pw->pw_gid,
2241 					strerror(errno));
2242 		uid = pw->pw_uid;
2243 	}
2244 	/*
2245 	 * handle chroot.
2246 	 */
2247 	if ((chrootdir = bozo_get_pref(prefs, "chroot dir")) != NULL) {
2248 		httpd->rootdir = strdup(chrootdir);
2249 		if (chdir(httpd->rootdir) == -1)
2250 			bozo_err(httpd, 1, "chdir(%s): %s", httpd->rootdir,
2251 				strerror(errno));
2252 		if (chroot(httpd->rootdir) == -1)
2253 			bozo_err(httpd, 1, "chroot(%s): %s", httpd->rootdir,
2254 				strerror(errno));
2255 	}
2256 
2257 	if (username != NULL)
2258 		if (setuid(uid) == -1)
2259 			bozo_err(httpd, 1, "setuid(%d): %s", uid,
2260 					strerror(errno));
2261 
2262 	/*
2263 	 * prevent info leakage between different compartments.
2264 	 * some PATH values in the environment would be invalided
2265 	 * by chroot. cross-user settings might result in undesirable
2266 	 * effects.
2267 	 */
2268 	if ((chrootdir != NULL || username != NULL) && !dirtyenv)
2269 		environ = cleanenv;
2270 
2271 #ifdef _SC_PAGESIZE
2272 	httpd->page_size = (long)sysconf(_SC_PAGESIZE);
2273 #else
2274 	httpd->page_size = 4096;
2275 #endif
2276 	debug((httpd, DEBUG_OBESE, "myname is %s, slashdir is %s",
2277 			httpd->virthostname, httpd->slashdir));
2278 
2279 	return 1;
2280 }
2281