bozohttpd.c revision 1.91 1 /* $NetBSD: bozohttpd.c,v 1.91 2018/11/21 09:37:02 mrg Exp $ */
2
3 /* $eterna: bozohttpd.c,v 1.178 2011/11/18 09:21:15 mrg Exp $ */
4
5 /*
6 * Copyright (c) 1997-2018 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/20181121"
113 #endif
114
115 #ifndef PUBLIC_HTML
116 #define PUBLIC_HTML "public_html"
117 #endif
118
119 #ifndef USE_ARG
120 #define USE_ARG(x) /*LINTED*/(void)&(x)
121 #endif
122
123 /*
124 * And so it begins ..
125 */
126
127 #include <sys/param.h>
128 #include <sys/socket.h>
129 #include <sys/time.h>
130 #include <sys/mman.h>
131
132 #include <arpa/inet.h>
133
134 #include <ctype.h>
135 #include <dirent.h>
136 #include <errno.h>
137 #include <fcntl.h>
138 #include <netdb.h>
139 #include <pwd.h>
140 #include <grp.h>
141 #include <signal.h>
142 #include <stdarg.h>
143 #include <stdlib.h>
144 #include <stdbool.h>
145 #include <string.h>
146 #include <syslog.h>
147 #include <time.h>
148 #include <unistd.h>
149
150 #include "bozohttpd.h"
151
152 #ifndef INITIAL_TIMEOUT
153 #define INITIAL_TIMEOUT "30" /* wait for 30 seconds initially */
154 #endif
155 #ifndef HEADER_WAIT_TIME
156 #define HEADER_WAIT_TIME "10" /* need more headers every 10 seconds */
157 #endif
158 #ifndef TOTAL_MAX_REQ_TIME
159 #define TOTAL_MAX_REQ_TIME "600" /* must have total request in 600 */
160 #endif /* seconds */
161
162 /* if monotonic time is not available try real time. */
163 #ifndef CLOCK_MONOTONIC
164 #define CLOCK_MONOTONIC CLOCK_REALTIME
165 #endif
166
167 /* variables and functions */
168 #ifndef LOG_FTP
169 #define LOG_FTP LOG_DAEMON
170 #endif
171
172 volatile sig_atomic_t timeout_hit;
173
174 /*
175 * check there's enough space in the prefs and names arrays.
176 */
177 static int
178 size_arrays(bozoprefs_t *bozoprefs, size_t needed)
179 {
180 char **temp;
181
182 if (bozoprefs->size == 0) {
183 /* only get here first time around */
184 bozoprefs->name = calloc(sizeof(char *), needed);
185 if (bozoprefs->name == NULL)
186 return 0;
187 bozoprefs->value = calloc(sizeof(char *), needed);
188 if (bozoprefs->value == NULL) {
189 free(bozoprefs->name);
190 return 0;
191 }
192 bozoprefs->size = needed;
193 } else if (bozoprefs->count == bozoprefs->size) {
194 /* only uses 'needed' when filled array */
195 temp = realloc(bozoprefs->name, sizeof(char *) * needed);
196 if (temp == NULL)
197 return 0;
198 bozoprefs->name = temp;
199 temp = realloc(bozoprefs->value, sizeof(char *) * needed);
200 if (temp == NULL)
201 return 0;
202 bozoprefs->value = temp;
203 bozoprefs->size += needed;
204 }
205 return 1;
206 }
207
208 static ssize_t
209 findvar(bozoprefs_t *bozoprefs, const char *name)
210 {
211 size_t i;
212
213 for (i = 0; i < bozoprefs->count; i++)
214 if (strcmp(bozoprefs->name[i], name) == 0)
215 return (ssize_t)i;
216 return -1;
217 }
218
219 int
220 bozo_set_pref(bozohttpd_t *httpd, bozoprefs_t *bozoprefs,
221 const char *name, const char *value)
222 {
223 ssize_t i;
224
225 if ((i = findvar(bozoprefs, name)) < 0) {
226 /* add the element to the array */
227 if (!size_arrays(bozoprefs, bozoprefs->size + 15))
228 return 0;
229 i = bozoprefs->count++;
230 bozoprefs->name[i] = bozostrdup(httpd, NULL, name);
231 } else {
232 /* replace the element in the array */
233 if (bozoprefs->value[i]) {
234 free(bozoprefs->value[i]);
235 bozoprefs->value[i] = NULL;
236 }
237 }
238 bozoprefs->value[i] = bozostrdup(httpd, NULL, value);
239 return 1;
240 }
241
242 /*
243 * get a variable's value, or NULL
244 */
245 char *
246 bozo_get_pref(bozoprefs_t *bozoprefs, const char *name)
247 {
248 ssize_t i;
249
250 i = findvar(bozoprefs, name);
251 return i < 0 ? NULL : bozoprefs->value[i];
252 }
253
254 char *
255 bozo_http_date(char *date, size_t datelen)
256 {
257 struct tm *tm;
258 time_t now;
259
260 /* Sun, 06 Nov 1994 08:49:37 GMT */
261 now = time(NULL);
262 tm = gmtime(&now); /* HTTP/1.1 spec rev 06 sez GMT only */
263 strftime(date, datelen, "%a, %d %b %Y %H:%M:%S GMT", tm);
264 return date;
265 }
266
267 /*
268 * convert "in" into the three parts of a request (first line).
269 * we allocate into file and query, but return pointers into
270 * "in" for proto and method.
271 */
272 static void
273 parse_request(bozohttpd_t *httpd, char *in, char **method, char **file,
274 char **query, char **proto)
275 {
276 ssize_t len;
277 char *val;
278
279 USE_ARG(httpd);
280 debug((httpd, DEBUG_EXPLODING, "parse in: %s", in));
281 *method = *file = *query = *proto = NULL;
282
283 len = (ssize_t)strlen(in);
284 val = bozostrnsep(&in, " \t\n\r", &len);
285 if (len < 1 || val == NULL)
286 return;
287 *method = val;
288
289 while (*in == ' ' || *in == '\t')
290 in++;
291 val = bozostrnsep(&in, " \t\n\r", &len);
292 if (len < 1) {
293 if (len == 0)
294 *file = val;
295 else
296 *file = in;
297 } else {
298 *file = val;
299
300 *query = strchr(*file, '?');
301 if (*query)
302 *(*query)++ = '\0';
303
304 if (in) {
305 while (*in && (*in == ' ' || *in == '\t'))
306 in++;
307 if (*in)
308 *proto = in;
309 }
310 }
311
312 /* allocate private copies */
313 *file = bozostrdup(httpd, NULL, *file);
314 if (*query)
315 *query = bozostrdup(httpd, NULL, *query);
316
317 debug((httpd, DEBUG_FAT,
318 "url: method: \"%s\" file: \"%s\" query: \"%s\" proto: \"%s\"",
319 *method, *file, *query, *proto));
320 }
321
322 /*
323 * cleanup a bozo_httpreq_t after use
324 */
325 void
326 bozo_clean_request(bozo_httpreq_t *request)
327 {
328 struct bozoheaders *hdr, *ohdr = NULL;
329
330 if (request == NULL)
331 return;
332
333 /* If SSL enabled cleanup SSL structure. */
334 bozo_ssl_destroy(request->hr_httpd);
335
336 /* clean up request */
337 free(request->hr_remotehost);
338 free(request->hr_remoteaddr);
339 free(request->hr_serverport);
340 free(request->hr_virthostname);
341 free(request->hr_file);
342 free(request->hr_oldfile);
343 free(request->hr_query);
344 free(request->hr_host);
345 bozo_user_free(request->hr_user);
346 bozo_auth_cleanup(request);
347 for (hdr = SIMPLEQ_FIRST(&request->hr_headers); hdr;
348 hdr = SIMPLEQ_NEXT(hdr, h_next)) {
349 free(hdr->h_value);
350 free(hdr->h_header);
351 free(ohdr);
352 ohdr = hdr;
353 }
354 free(ohdr);
355 ohdr = NULL;
356 for (hdr = SIMPLEQ_FIRST(&request->hr_replheaders); hdr;
357 hdr = SIMPLEQ_NEXT(hdr, h_next)) {
358 free(hdr->h_value);
359 free(hdr->h_header);
360 free(ohdr);
361 ohdr = hdr;
362 }
363 free(ohdr);
364
365 free(request);
366 }
367
368 /*
369 * send a HTTP/1.1 408 response if we timeout.
370 */
371 /* ARGSUSED */
372 static void
373 alarmer(int sig)
374 {
375 timeout_hit = 1;
376 }
377
378
379 /*
380 * set a timeout for "initial", "header", or "request".
381 */
382 int
383 bozo_set_timeout(bozohttpd_t *httpd, bozoprefs_t *prefs,
384 const char *target, const char *time)
385 {
386 const char *cur, *timeouts[] = {
387 "initial timeout",
388 "header timeout",
389 "request timeout",
390 NULL,
391 };
392 /* adjust minlen if more timeouts appear with conflicting names */
393 const size_t minlen = 1;
394 size_t len = strlen(target);
395
396 for (cur = timeouts[0]; len >= minlen && *cur; cur++) {
397 if (strncmp(target, cur, len) == 0) {
398 bozo_set_pref(httpd, prefs, cur, time);
399 return 0;
400 }
401 }
402 return 1;
403 }
404
405 /*
406 * a list of header quirks: currently, a list of headers that
407 * can't be folded into a single line.
408 */
409 const char *header_quirks[] = { "WWW-Authenticate", NULL };
410
411 /*
412 * add or merge this header (val: str) into the requests list
413 */
414 static bozoheaders_t *
415 addmerge_header(bozo_httpreq_t *request, struct qheaders *headers,
416 const char *val, const char *str, ssize_t len)
417 {
418 struct bozohttpd_t *httpd = request->hr_httpd;
419 struct bozoheaders *hdr = NULL;
420 const char **quirk;
421
422 USE_ARG(len);
423 for (quirk = header_quirks; *quirk; quirk++)
424 if (strcasecmp(*quirk, val) == 0)
425 break;
426
427 if (*quirk == NULL) {
428 /* do we exist already? */
429 SIMPLEQ_FOREACH(hdr, headers, h_next) {
430 if (strcasecmp(val, hdr->h_header) == 0)
431 break;
432 }
433 }
434
435 if (hdr) {
436 /* yup, merge it in */
437 char *nval;
438
439 bozoasprintf(httpd, &nval, "%s, %s", hdr->h_value, str);
440 free(hdr->h_value);
441 hdr->h_value = nval;
442 } else {
443 /* nope, create a new one */
444
445 hdr = bozomalloc(httpd, sizeof *hdr);
446 hdr->h_header = bozostrdup(httpd, request, val);
447 if (str && *str)
448 hdr->h_value = bozostrdup(httpd, request, str);
449 else
450 hdr->h_value = bozostrdup(httpd, request, " ");
451
452 SIMPLEQ_INSERT_TAIL(headers, hdr, h_next);
453 request->hr_nheaders++;
454 }
455
456 return hdr;
457 }
458
459 bozoheaders_t *
460 addmerge_reqheader(bozo_httpreq_t *request, const char *val, const char *str,
461 ssize_t len)
462 {
463
464 return addmerge_header(request, &request->hr_headers, val, str, len);
465 }
466
467 bozoheaders_t *
468 addmerge_replheader(bozo_httpreq_t *request, const char *val, const char *str,
469 ssize_t len)
470 {
471
472 return addmerge_header(request, &request->hr_replheaders,
473 val, str, len);
474 }
475
476 /*
477 * as the prototype string is not constant (eg, "HTTP/1.1" is equivalent
478 * to "HTTP/001.01"), we MUST parse this.
479 */
480 static int
481 process_proto(bozo_httpreq_t *request, const char *proto)
482 {
483 struct bozohttpd_t *httpd = request->hr_httpd;
484 char majorstr[16], *minorstr;
485 int majorint, minorint;
486
487 if (proto == NULL) {
488 got_proto_09:
489 request->hr_proto = httpd->consts.http_09;
490 debug((httpd, DEBUG_FAT, "request %s is http/0.9",
491 request->hr_file));
492 return 0;
493 }
494
495 if (strncasecmp(proto, "HTTP/", 5) != 0)
496 goto bad;
497 strncpy(majorstr, proto + 5, sizeof majorstr);
498 majorstr[sizeof(majorstr)-1] = 0;
499 minorstr = strchr(majorstr, '.');
500 if (minorstr == NULL)
501 goto bad;
502 *minorstr++ = 0;
503
504 majorint = atoi(majorstr);
505 minorint = atoi(minorstr);
506
507 switch (majorint) {
508 case 0:
509 if (minorint != 9)
510 break;
511 goto got_proto_09;
512 case 1:
513 if (minorint == 0)
514 request->hr_proto = httpd->consts.http_10;
515 else if (minorint == 1)
516 request->hr_proto = httpd->consts.http_11;
517 else
518 break;
519
520 debug((httpd, DEBUG_FAT, "request %s is %s",
521 request->hr_file, request->hr_proto));
522 SIMPLEQ_INIT(&request->hr_headers);
523 request->hr_nheaders = 0;
524 return 0;
525 }
526 bad:
527 return bozo_http_error(httpd, 404, NULL, "unknown prototype");
528 }
529
530 /*
531 * process each type of HTTP method, setting this HTTP requests
532 * method type.
533 */
534 static struct method_map {
535 const char *name;
536 int type;
537 } method_map[] = {
538 { "GET", HTTP_GET, },
539 { "POST", HTTP_POST, },
540 { "HEAD", HTTP_HEAD, },
541 #if 0 /* other non-required http/1.1 methods */
542 { "OPTIONS", HTTP_OPTIONS, },
543 { "PUT", HTTP_PUT, },
544 { "DELETE", HTTP_DELETE, },
545 { "TRACE", HTTP_TRACE, },
546 { "CONNECT", HTTP_CONNECT, },
547 #endif
548 { NULL, 0, },
549 };
550
551 static int
552 process_method(bozo_httpreq_t *request, const char *method)
553 {
554 struct bozohttpd_t *httpd = request->hr_httpd;
555 struct method_map *mmp;
556
557 if (request->hr_proto == httpd->consts.http_11)
558 request->hr_allow = "GET, HEAD, POST";
559
560 for (mmp = method_map; mmp->name; mmp++)
561 if (strcasecmp(method, mmp->name) == 0) {
562 request->hr_method = mmp->type;
563 request->hr_methodstr = mmp->name;
564 return 0;
565 }
566
567 return bozo_http_error(httpd, 404, request, "unknown method");
568 }
569
570 /* check header byte count */
571 static int
572 bozo_got_header_length(bozo_httpreq_t *request, size_t len)
573 {
574 request->hr_header_bytes += len;
575 if (request->hr_header_bytes < BOZO_HEADERS_MAX_SIZE)
576 return 0;
577
578 return bozo_http_error(request->hr_httpd, 413, request,
579 "too many headers");
580 }
581
582 /*
583 * This function reads a http request from stdin, returning a pointer to a
584 * bozo_httpreq_t structure, describing the request.
585 */
586 bozo_httpreq_t *
587 bozo_read_request(bozohttpd_t *httpd)
588 {
589 struct sigaction sa;
590 char *str, *val, *method, *file, *proto, *query;
591 char *host, *addr, *port;
592 char bufport[10];
593 char hbuf[NI_MAXHOST], abuf[NI_MAXHOST];
594 struct sockaddr_storage ss;
595 ssize_t len;
596 int line = 0;
597 socklen_t slen;
598 bozo_httpreq_t *request;
599 struct timespec ots, ts;
600
601 /*
602 * if we're in daemon mode, bozo_daemon_fork() will return here twice
603 * for each call. once in the child, returning 0, and once in the
604 * parent, returning 1. for each child, then we can setup SSL, and
605 * the parent can signal the caller there was no request to process
606 * and it will wait for another.
607 */
608 if (bozo_daemon_fork(httpd))
609 return NULL;
610 if (bozo_ssl_accept(httpd))
611 return NULL;
612
613 request = bozomalloc(httpd, sizeof(*request));
614 memset(request, 0, sizeof(*request));
615 request->hr_httpd = httpd;
616 request->hr_allow = request->hr_host = NULL;
617 request->hr_content_type = request->hr_content_length = NULL;
618 request->hr_range = NULL;
619 request->hr_last_byte_pos = -1;
620 request->hr_if_modified_since = NULL;
621 request->hr_virthostname = NULL;
622 request->hr_file = NULL;
623 request->hr_oldfile = NULL;
624 SIMPLEQ_INIT(&request->hr_replheaders);
625 bozo_auth_init(request);
626
627 slen = sizeof(ss);
628 if (getpeername(0, (struct sockaddr *)(void *)&ss, &slen) < 0)
629 host = addr = NULL;
630 else {
631 if (getnameinfo((struct sockaddr *)(void *)&ss, slen,
632 abuf, sizeof abuf, NULL, 0, NI_NUMERICHOST) == 0)
633 addr = abuf;
634 else
635 addr = NULL;
636 if (httpd->numeric == 0 &&
637 getnameinfo((struct sockaddr *)(void *)&ss, slen,
638 hbuf, sizeof hbuf, NULL, 0, 0) == 0)
639 host = hbuf;
640 else
641 host = NULL;
642 }
643 if (host != NULL)
644 request->hr_remotehost = bozostrdup(httpd, request, host);
645 if (addr != NULL)
646 request->hr_remoteaddr = bozostrdup(httpd, request, addr);
647 slen = sizeof(ss);
648
649 /*
650 * Override the bound port from the request value, so it works even
651 * if passed through a proxy that doesn't rewrite the port.
652 */
653 if (httpd->bindport) {
654 if (strcmp(httpd->bindport, "80") != 0)
655 port = httpd->bindport;
656 else
657 port = NULL;
658 } else {
659 if (getsockname(0, (struct sockaddr *)(void *)&ss, &slen) < 0)
660 port = NULL;
661 else {
662 if (getnameinfo((struct sockaddr *)(void *)&ss, slen,
663 NULL, 0, bufport, sizeof bufport,
664 NI_NUMERICSERV) == 0)
665 port = bufport;
666 else
667 port = NULL;
668 }
669 }
670 if (port != NULL)
671 request->hr_serverport = bozostrdup(httpd, request, port);
672
673 /*
674 * setup a timer to make sure the request is not hung
675 */
676 sa.sa_handler = alarmer;
677 sigemptyset(&sa.sa_mask);
678 sigaddset(&sa.sa_mask, SIGALRM);
679 sa.sa_flags = 0;
680 sigaction(SIGALRM, &sa, NULL);
681
682 if (clock_gettime(CLOCK_MONOTONIC, &ots) != 0) {
683 (void)bozo_http_error(httpd, 500, NULL,
684 "clock_gettime failed");
685 goto cleanup;
686 }
687
688 alarm(httpd->initial_timeout);
689 while ((str = bozodgetln(httpd, STDIN_FILENO, &len, bozo_read)) != NULL) {
690 alarm(0);
691
692 if (clock_gettime(CLOCK_MONOTONIC, &ts) != 0) {
693 (void)bozo_http_error(httpd, 500, NULL,
694 "clock_gettime failed");
695 goto cleanup;
696 }
697 /*
698 * don't timeout if old tv_sec is not more than current
699 * tv_sec, or if current tv_sec is less than the request
700 * timeout (these shouldn't happen, but the first could
701 * if monotonic time is not available.)
702 *
703 * the other timeout and header size checks should ensure
704 * that even if time it set backwards or forwards a very
705 * long way, timeout will eventually happen, even if this
706 * one fails.
707 */
708 if (ts.tv_sec > ots.tv_sec &&
709 ts.tv_sec > httpd->request_timeout &&
710 ts.tv_sec - httpd->request_timeout > ots.tv_sec)
711 timeout_hit = 1;
712
713 if (timeout_hit) {
714 (void)bozo_http_error(httpd, 408, NULL,
715 "request timed out");
716 goto cleanup;
717 }
718 line++;
719
720 if (line == 1) {
721 if (len < 1) {
722 (void)bozo_http_error(httpd, 404, NULL,
723 "null method");
724 goto cleanup;
725 }
726 bozowarn(httpd,
727 "got request ``%s'' from host %s to port %s",
728 str,
729 host ? host : addr ? addr : "<local>",
730 port ? port : "<stdin>");
731
732 /* we allocate return space in file and query only */
733 parse_request(httpd, str, &method, &file, &query, &proto);
734 request->hr_file = file;
735 request->hr_query = query;
736 if (method == NULL) {
737 (void)bozo_http_error(httpd, 404, NULL,
738 "null method");
739 goto cleanup;
740 }
741 if (file == NULL) {
742 (void)bozo_http_error(httpd, 404, NULL,
743 "null file");
744 goto cleanup;
745 }
746
747 /*
748 * note that we parse the proto first, so that we
749 * can more properly parse the method and the url.
750 */
751
752 if (process_proto(request, proto) ||
753 process_method(request, method)) {
754 goto cleanup;
755 }
756
757 debug((httpd, DEBUG_FAT, "got file \"%s\" query \"%s\"",
758 request->hr_file,
759 request->hr_query ? request->hr_query : "<none>"));
760
761 /* http/0.9 has no header processing */
762 if (request->hr_proto == httpd->consts.http_09)
763 break;
764 } else { /* incoming headers */
765 bozoheaders_t *hdr;
766
767 if (*str == '\0')
768 break;
769
770 val = bozostrnsep(&str, ":", &len);
771 debug((httpd, DEBUG_EXPLODING,
772 "read_req2: after bozostrnsep: str ``%s'' val ``%s''",
773 str, val));
774 if (val == NULL || len == -1) {
775 (void)bozo_http_error(httpd, 404, request,
776 "no header");
777 goto cleanup;
778 }
779 while (*str == ' ' || *str == '\t')
780 len--, str++;
781 while (*val == ' ' || *val == '\t')
782 val++;
783
784 if (bozo_got_header_length(request, len))
785 goto cleanup;
786
787 if (bozo_auth_check_headers(request, val, str, len))
788 goto next_header;
789
790 hdr = addmerge_reqheader(request, val, str, len);
791
792 if (strcasecmp(hdr->h_header, "content-type") == 0)
793 request->hr_content_type = hdr->h_value;
794 else if (strcasecmp(hdr->h_header, "content-length") == 0)
795 request->hr_content_length = hdr->h_value;
796 else if (strcasecmp(hdr->h_header, "host") == 0) {
797 if (request->hr_host) {
798 /* RFC 7230 (HTTP/1.1): 5.4 */
799 (void)bozo_http_error(httpd, 400, request,
800 "Only allow one Host: header");
801 goto cleanup;
802 }
803 request->hr_host = bozostrdup(httpd, request,
804 hdr->h_value);
805 }
806 /* RFC 2616 (HTTP/1.1): 14.20 */
807 else if (strcasecmp(hdr->h_header, "expect") == 0) {
808 (void)bozo_http_error(httpd, 417, request,
809 "we don't support Expect:");
810 goto cleanup;
811 }
812 else if (strcasecmp(hdr->h_header, "referrer") == 0 ||
813 strcasecmp(hdr->h_header, "referer") == 0)
814 request->hr_referrer = hdr->h_value;
815 else if (strcasecmp(hdr->h_header, "range") == 0)
816 request->hr_range = hdr->h_value;
817 else if (strcasecmp(hdr->h_header,
818 "if-modified-since") == 0)
819 request->hr_if_modified_since = hdr->h_value;
820 else if (strcasecmp(hdr->h_header,
821 "accept-encoding") == 0)
822 request->hr_accept_encoding = hdr->h_value;
823
824 debug((httpd, DEBUG_FAT, "adding header %s: %s",
825 hdr->h_header, hdr->h_value));
826 }
827 next_header:
828 alarm(httpd->header_timeout);
829 }
830
831 /* now, clear it all out */
832 alarm(0);
833 signal(SIGALRM, SIG_DFL);
834
835 /* RFC1945, 8.3 */
836 if (request->hr_method == HTTP_POST &&
837 request->hr_content_length == NULL) {
838 (void)bozo_http_error(httpd, 400, request,
839 "missing content length");
840 goto cleanup;
841 }
842
843 /* RFC 2616 (HTTP/1.1), 14.23 & 19.6.1.1 */
844 if (request->hr_proto == httpd->consts.http_11 &&
845 /*(strncasecmp(request->hr_file, "http://", 7) != 0) &&*/
846 request->hr_host == NULL) {
847 (void)bozo_http_error(httpd, 400, request,
848 "missing Host header");
849 goto cleanup;
850 }
851
852 if (request->hr_range != NULL) {
853 debug((httpd, DEBUG_FAT, "hr_range: %s", request->hr_range));
854 /* support only simple ranges %d- and %d-%d */
855 if (strchr(request->hr_range, ',') == NULL) {
856 const char *rstart, *dash;
857
858 rstart = strchr(request->hr_range, '=');
859 if (rstart != NULL) {
860 rstart++;
861 dash = strchr(rstart, '-');
862 if (dash != NULL && dash != rstart) {
863 dash++;
864 request->hr_have_range = 1;
865 request->hr_first_byte_pos =
866 strtoll(rstart, NULL, 10);
867 if (request->hr_first_byte_pos < 0)
868 request->hr_first_byte_pos = 0;
869 if (*dash != '\0') {
870 request->hr_last_byte_pos =
871 strtoll(dash, NULL, 10);
872 if (request->hr_last_byte_pos < 0)
873 request->hr_last_byte_pos = -1;
874 }
875 }
876 }
877 }
878 }
879
880 debug((httpd, DEBUG_FAT, "bozo_read_request returns url %s in request",
881 request->hr_file));
882 return request;
883
884 cleanup:
885 bozo_clean_request(request);
886
887 return NULL;
888 }
889
890 static int
891 mmap_and_write_part(bozohttpd_t *httpd, int fd, off_t first_byte_pos, size_t sz)
892 {
893 size_t mappedsz, wroffset;
894 off_t mappedoffset;
895 char *addr;
896 void *mappedaddr;
897
898 /*
899 * we need to ensure that both the size *and* offset arguments to
900 * mmap() are page-aligned. our formala for this is:
901 *
902 * input offset: first_byte_pos
903 * input size: sz
904 *
905 * mapped offset = page align truncate (input offset)
906 * mapped size =
907 * page align extend (input offset - mapped offset + input size)
908 * write offset = input offset - mapped offset
909 *
910 * we use the write offset in all writes
911 */
912 mappedoffset = first_byte_pos & ~(httpd->page_size - 1);
913 mappedsz = (size_t)
914 (first_byte_pos - mappedoffset + sz + httpd->page_size - 1) &
915 ~(httpd->page_size - 1);
916 wroffset = (size_t)(first_byte_pos - mappedoffset);
917
918 addr = mmap(0, mappedsz, PROT_READ, MAP_SHARED, fd, mappedoffset);
919 if (addr == (char *)-1) {
920 bozowarn(httpd, "mmap failed: %s", strerror(errno));
921 return -1;
922 }
923 mappedaddr = addr;
924
925 #ifdef MADV_SEQUENTIAL
926 (void)madvise(addr, sz, MADV_SEQUENTIAL);
927 #endif
928 while (sz > BOZO_WRSZ) {
929 if (bozo_write(httpd, STDOUT_FILENO, addr + wroffset,
930 BOZO_WRSZ) != BOZO_WRSZ) {
931 bozowarn(httpd, "write failed: %s", strerror(errno));
932 goto out;
933 }
934 debug((httpd, DEBUG_OBESE, "wrote %d bytes", BOZO_WRSZ));
935 sz -= BOZO_WRSZ;
936 addr += BOZO_WRSZ;
937 }
938 if (sz && (size_t)bozo_write(httpd, STDOUT_FILENO, addr + wroffset,
939 sz) != sz) {
940 bozowarn(httpd, "final write failed: %s", strerror(errno));
941 goto out;
942 }
943 debug((httpd, DEBUG_OBESE, "wrote %d bytes", (int)sz));
944 out:
945 if (munmap(mappedaddr, mappedsz) < 0) {
946 bozowarn(httpd, "munmap failed");
947 return -1;
948 }
949
950 return 0;
951 }
952
953 static int
954 parse_http_date(const char *val, time_t *timestamp)
955 {
956 char *remainder;
957 struct tm tm;
958
959 if ((remainder = strptime(val, "%a, %d %b %Y %T GMT", &tm)) == NULL &&
960 (remainder = strptime(val, "%a, %d-%b-%y %T GMT", &tm)) == NULL &&
961 (remainder = strptime(val, "%a %b %d %T %Y", &tm)) == NULL)
962 return 0; /* Invalid HTTP date format */
963
964 if (*remainder)
965 return 0; /* No trailing garbage */
966
967 *timestamp = timegm(&tm);
968 return 1;
969 }
970
971 /*
972 * given an url, encode it ala rfc 3986. ie, escape ? and friends.
973 * note that this function returns a static buffer, and thus needs
974 * to be updated for any sort of parallel processing. escape only
975 * chosen characters for absolute redirects
976 */
977 char *
978 bozo_escape_rfc3986(bozohttpd_t *httpd, const char *url, int absolute)
979 {
980 static char *buf;
981 static size_t buflen = 0;
982 size_t len;
983 const char *s;
984 char *d;
985
986 len = strlen(url);
987 if (buflen < len * 3 + 1) {
988 buflen = len * 3 + 1;
989 buf = bozorealloc(httpd, buf, buflen);
990 }
991
992 for (len = 0, s = url, d = buf; *s;) {
993 if (*s & 0x80)
994 goto encode_it;
995 switch (*s) {
996 case ':':
997 case '?':
998 case '#':
999 case '[':
1000 case ']':
1001 case '@':
1002 case '!':
1003 case '$':
1004 case '&':
1005 case '\'':
1006 case '(':
1007 case ')':
1008 case '*':
1009 case '+':
1010 case ',':
1011 case ';':
1012 case '=':
1013 case '%':
1014 case '"':
1015 if (absolute)
1016 goto leave_it;
1017 case '\n':
1018 case '\r':
1019 case ' ':
1020 encode_it:
1021 snprintf(d, 4, "%%%02X", *s++);
1022 d += 3;
1023 len += 3;
1024 break;
1025 leave_it:
1026 default:
1027 *d++ = *s++;
1028 len++;
1029 break;
1030 }
1031 }
1032 buf[len] = 0;
1033
1034 return buf;
1035 }
1036
1037 /*
1038 * do automatic redirection -- if there are query parameters or userdir for
1039 * the URL we will tack these on to the new (redirected) URL.
1040 */
1041 static void
1042 handle_redirect(bozo_httpreq_t *request, const char *url, int absolute)
1043 {
1044 bozohttpd_t *httpd = request->hr_httpd;
1045 char *finalurl, *urlbuf;
1046 #ifndef NO_USER_SUPPORT
1047 char *userbuf;
1048 #endif /* !NO_USER_SUPPORT */
1049 char portbuf[20];
1050 const char *scheme, *query, *quest;
1051 const char *hostname = BOZOHOST(httpd, request);
1052 int absproto = 0; /* absolute redirect provides own schema */
1053
1054 if (url == NULL) {
1055 bozoasprintf(httpd, &urlbuf, "/%s/", request->hr_file);
1056 url = urlbuf;
1057 } else
1058 urlbuf = NULL;
1059
1060 #ifndef NO_USER_SUPPORT
1061 if (request->hr_user && !absolute) {
1062 bozoasprintf(httpd, &userbuf, "/~%s%s", request->hr_user, url);
1063 url = userbuf;
1064 } else
1065 userbuf = NULL;
1066 #endif /* !NO_USER_SUPPORT */
1067
1068 if (absolute) {
1069 char *sep = NULL;
1070 const char *s;
1071
1072 /*
1073 * absolute redirect may specify own protocol i.e. to redirect
1074 * to another schema like https:// or ftp://.
1075 * Details: RFC 3986, section 3.
1076 */
1077
1078 /* 1. check if url contains :// */
1079 sep = strstr(url, "://");
1080
1081 /*
1082 * RFC 3986, section 3.1:
1083 * scheme = ALPHA *( ALPHA / DIGIT / "+" / "-" / "." )
1084 */
1085 if (sep) {
1086 for (s = url; s != sep;) {
1087 if (!isalnum((int)*s) &&
1088 *s != '+' && *s != '-' && *s != '.')
1089 break;
1090 if (++s == sep) {
1091 absproto = 1;
1092 }
1093 }
1094 }
1095 }
1096
1097 /* construct final redirection url */
1098
1099 scheme = absproto ? "" : httpd->sslinfo ? "https://" : "http://";
1100
1101 if (absolute) {
1102 hostname = "";
1103 portbuf[0] = '\0';
1104 } else {
1105 const char *defport = httpd->sslinfo ? "443" : "80";
1106
1107 if (request->hr_serverport &&
1108 strcmp(request->hr_serverport, defport) != 0)
1109 snprintf(portbuf, sizeof(portbuf), ":%s",
1110 request->hr_serverport);
1111 else
1112 portbuf[0] = '\0';
1113 }
1114
1115 url = bozo_escape_rfc3986(httpd, url, absolute);
1116
1117 if (request->hr_query && strlen(request->hr_query)) {
1118 query = request->hr_query;
1119 quest = "?";
1120 } else {
1121 query = quest = "";
1122 }
1123
1124 bozoasprintf(httpd, &finalurl, "%s%s%s%s%s%s",
1125 scheme, hostname, portbuf, url, quest, query);
1126
1127 bozowarn(httpd, "redirecting %s", finalurl);
1128 debug((httpd, DEBUG_FAT, "redirecting %s", finalurl));
1129
1130 bozo_printf(httpd, "%s 301 Document Moved\r\n", request->hr_proto);
1131 if (request->hr_proto != httpd->consts.http_09)
1132 bozo_print_header(request, NULL, "text/html", NULL);
1133 if (request->hr_proto != httpd->consts.http_09)
1134 bozo_printf(httpd, "Location: %s\r\n", finalurl);
1135 bozo_printf(httpd, "\r\n");
1136 if (request->hr_method == HTTP_HEAD)
1137 goto head;
1138 bozo_printf(httpd, "<html><head><title>Document Moved</title></head>\n");
1139 bozo_printf(httpd, "<body><h1>Document Moved</h1>\n");
1140 bozo_printf(httpd, "This document had moved <a href=\"%s\">here</a>\n",
1141 finalurl);
1142 bozo_printf(httpd, "</body></html>\n");
1143 head:
1144 bozo_flush(httpd, stdout);
1145 free(urlbuf);
1146 free(finalurl);
1147 #ifndef NO_USER_SUPPORT
1148 free(userbuf);
1149 #endif /* !NO_USER_SUPPORT */
1150 }
1151
1152 /*
1153 * Like strncmp(), but s_esc may contain characters escaped by \.
1154 * The len argument does not include the backslashes used for escaping,
1155 * that is: it gives the raw len, after unescaping the string.
1156 */
1157 static int
1158 esccmp(const char *s_plain, const char *s_esc, size_t len)
1159 {
1160 bool esc = false;
1161
1162 while (len) {
1163 if (!esc && *s_esc == '\\') {
1164 esc = true;
1165 s_esc++;
1166 continue;
1167 }
1168 esc = false;
1169 if (*s_plain == 0 || *s_esc == 0 || *s_plain != *s_esc)
1170 return *s_esc - *s_plain;
1171 s_esc++;
1172 s_plain++;
1173 len--;
1174 }
1175 return 0;
1176 }
1177
1178 /*
1179 * Check if the request refers to a uri that is mapped via a .bzremap.
1180 * We have /requested/path:/re/mapped/to/this.html lines in there,
1181 * and the : separator may be use in the left hand side escaped with
1182 * \ to encode a path containig a : character.
1183 */
1184 static void
1185 check_mapping(bozo_httpreq_t *request)
1186 {
1187 bozohttpd_t *httpd = request->hr_httpd;
1188 char *file = request->hr_file, *newfile;
1189 void *fmap;
1190 const char *replace, *map_to, *p;
1191 struct stat st;
1192 int mapfile;
1193 size_t avail, len, rlen, reqlen, num_esc = 0;
1194 bool escaped = false;
1195
1196 mapfile = open(REMAP_FILE, O_RDONLY, 0);
1197 if (mapfile == -1)
1198 return;
1199 debug((httpd, DEBUG_FAT, "remap file found"));
1200 if (fstat(mapfile, &st) == -1) {
1201 bozowarn(httpd, "could not stat " REMAP_FILE ", errno: %d",
1202 errno);
1203 close(mapfile);
1204 return;
1205 }
1206
1207 fmap = mmap(NULL, st.st_size, PROT_READ, 0, mapfile, 0);
1208 if (fmap == NULL) {
1209 bozowarn(httpd, "could not mmap " REMAP_FILE ", error %d",
1210 errno);
1211 close(mapfile);
1212 return;
1213 }
1214 reqlen = strlen(file);
1215 for (p = fmap, avail = st.st_size; avail; ) {
1216 /*
1217 * We have lines like:
1218 * /this/url:/replacement/that/url
1219 * If we find a matching left hand side, replace will point
1220 * to it and len will be its length. map_to will point to
1221 * the right hand side and rlen wil be its length.
1222 * If we have no match, both pointers will be NULL.
1223 */
1224
1225 /* skip empty lines */
1226 while ((*p == '\r' || *p == '\n') && avail) {
1227 p++;
1228 avail--;
1229 }
1230 replace = p;
1231 escaped = false;
1232 while (avail) {
1233 if (*p == '\r' || *p == '\n')
1234 break;
1235 if (!escaped && *p == ':')
1236 break;
1237 if (escaped) {
1238 escaped = false;
1239 num_esc++;
1240 } else if (*p == '\\') {
1241 escaped = true;
1242 }
1243 p++;
1244 avail--;
1245 }
1246 if (!avail || *p != ':') {
1247 replace = NULL;
1248 map_to = NULL;
1249 break;
1250 }
1251 len = p - replace - num_esc;
1252 /*
1253 * reqlen < len: the left hand side is too long, can't be a
1254 * match
1255 * reqlen == len: full string has to match
1256 * reqlen > len: make sure there is a path separator at 'len'
1257 * avail < 2: we are at eof, missing right hand side
1258 */
1259 if (avail < 2 || reqlen < len ||
1260 (reqlen == len && esccmp(file, replace, len) != 0) ||
1261 (reqlen > len && (file[len] != '/' ||
1262 esccmp(file, replace, len) != 0))) {
1263
1264 /* non-match, skip to end of line and continue */
1265 while (*p != '\r' && *p != '\n' && avail) {
1266 p++;
1267 avail--;
1268 }
1269 replace = NULL;
1270 map_to = NULL;
1271 continue;
1272 }
1273 p++;
1274 avail--;
1275
1276 /* found a match, parse the target */
1277 map_to = p;
1278 while (*p != '\r' && *p != '\n' && avail) {
1279 p++;
1280 avail--;
1281 }
1282 rlen = p - map_to;
1283 break;
1284 }
1285
1286 if (replace && map_to) {
1287 newfile = bozomalloc(httpd, strlen(file) + rlen - len + 1);
1288 memcpy(newfile, map_to, rlen);
1289 strcpy(newfile+rlen, file + len);
1290 debug((httpd, DEBUG_NORMAL, "remapping found ``%s'' ",
1291 newfile));
1292 free(request->hr_file);
1293 request->hr_file = newfile;
1294 }
1295
1296 munmap(fmap, st.st_size);
1297 close(mapfile);
1298 }
1299
1300 /*
1301 * deal with virtual host names; we do this:
1302 * if we have a virtual path root (httpd->virtbase), and we are given a
1303 * virtual host spec (Host: ho.st or http://ho.st/), see if this
1304 * directory exists under httpd->virtbase. if it does, use this as the
1305 # new slashdir.
1306 */
1307 static int
1308 check_virtual(bozo_httpreq_t *request)
1309 {
1310 bozohttpd_t *httpd = request->hr_httpd;
1311 char *file = request->hr_file, *s;
1312 size_t len;
1313
1314 /*
1315 * convert http://virtual.host/ to request->hr_host
1316 */
1317 debug((httpd, DEBUG_OBESE, "checking for http:// virtual host in ``%s''",
1318 file));
1319 if (strncasecmp(file, "http://", 7) == 0) {
1320 /* we would do virtual hosting here? */
1321 file += 7;
1322 /* RFC 2616 (HTTP/1.1), 5.2: URI takes precedence over Host: */
1323 free(request->hr_host);
1324 request->hr_host = bozostrdup(httpd, request, file);
1325 if ((s = strchr(request->hr_host, '/')) != NULL)
1326 *s = '\0';
1327 s = strchr(file, '/');
1328 free(request->hr_file);
1329 request->hr_file = bozostrdup(httpd, request, s ? s : "/");
1330 debug((httpd, DEBUG_OBESE, "got host ``%s'' file is now ``%s''",
1331 request->hr_host, request->hr_file));
1332 } else if (!request->hr_host)
1333 goto use_slashdir;
1334
1335 /*
1336 * canonicalise hr_host - that is, remove any :80.
1337 */
1338 len = strlen(request->hr_host);
1339 if (len > 3 && strcmp(request->hr_host + len - 3, ":80") == 0) {
1340 request->hr_host[len - 3] = '\0';
1341 len = strlen(request->hr_host);
1342 }
1343
1344 if (!httpd->virtbase) {
1345
1346 /*
1347 * if we don't use vhost support, then set virthostname if
1348 * user supplied Host header. It will be used for possible
1349 * redirections
1350 */
1351
1352 if (request->hr_host) {
1353 s = strrchr(request->hr_host, ':');
1354 if (s != NULL)
1355 /* truncate Host: as we want to copy it without port part */
1356 *s = '\0';
1357 request->hr_virthostname = bozostrdup(httpd, request,
1358 request->hr_host);
1359 if (s != NULL)
1360 /* fix Host: again, if we truncated it */
1361 *s = ':';
1362 }
1363
1364 goto use_slashdir;
1365 }
1366
1367 /*
1368 * ok, we have a virtual host, use opendir(3) to find a case
1369 * insensitive match for the virtual host we are asked for.
1370 * note that if the virtual host is the same as the master,
1371 * we don't need to do anything special.
1372 */
1373 debug((httpd, DEBUG_OBESE,
1374 "check_virtual: checking host `%s' under httpd->virtbase `%s' "
1375 "for file `%s'",
1376 request->hr_host, httpd->virtbase, request->hr_file));
1377 if (strncasecmp(httpd->virthostname, request->hr_host, len) != 0) {
1378 s = 0;
1379 DIR *dirp;
1380 struct dirent *d;
1381
1382 if ((dirp = opendir(httpd->virtbase)) != NULL) {
1383 while ((d = readdir(dirp)) != NULL) {
1384 if (strcmp(d->d_name, ".") == 0 ||
1385 strcmp(d->d_name, "..") == 0) {
1386 continue;
1387 }
1388 debug((httpd, DEBUG_OBESE, "looking at dir``%s''",
1389 d->d_name));
1390 if (strcmp(d->d_name, request->hr_host) == 0) {
1391 /* found it, punch it */
1392 debug((httpd, DEBUG_OBESE, "found it punch it"));
1393 request->hr_virthostname =
1394 bozostrdup(httpd, request, d->d_name);
1395 bozoasprintf(httpd, &s, "%s/%s",
1396 httpd->virtbase,
1397 request->hr_virthostname);
1398 break;
1399 }
1400 }
1401 closedir(dirp);
1402 }
1403 else {
1404 debug((httpd, DEBUG_FAT, "opendir %s failed: %s",
1405 httpd->virtbase, strerror(errno)));
1406 }
1407 if (s == 0) {
1408 if (httpd->unknown_slash)
1409 goto use_slashdir;
1410 return bozo_http_error(httpd, 404, request,
1411 "unknown URL");
1412 }
1413 } else
1414 use_slashdir:
1415 s = httpd->slashdir;
1416
1417 /*
1418 * ok, nailed the correct slashdir, chdir to it
1419 */
1420 if (chdir(s) < 0)
1421 return bozo_http_error(httpd, 404, request,
1422 "can't chdir to slashdir");
1423
1424 /*
1425 * is there a mapping for this request?
1426 */
1427 check_mapping(request);
1428
1429 return 0;
1430 }
1431
1432 /*
1433 * checks to see if this request has a valid .bzredirect file. returns
1434 * 0 when no redirection happend, or 1 when handle_redirect() has been
1435 * called, -1 on error.
1436 */
1437 static int
1438 check_bzredirect(bozo_httpreq_t *request)
1439 {
1440 bozohttpd_t *httpd = request->hr_httpd;
1441 struct stat sb;
1442 char dir[MAXPATHLEN], redir[MAXPATHLEN], redirpath[MAXPATHLEN + 1],
1443 path[MAXPATHLEN];
1444 char *basename, *finalredir;
1445 int rv, absolute;
1446
1447 /*
1448 * if this pathname is really a directory, but doesn't end in /,
1449 * use it as the directory to look for the redir file.
1450 */
1451 if ((size_t)snprintf(dir, sizeof(dir), "%s", request->hr_file + 1) >=
1452 sizeof(dir))
1453 return bozo_http_error(httpd, 404, request,
1454 "file path too long");
1455 debug((httpd, DEBUG_FAT, "check_bzredirect: dir %s", dir));
1456 basename = strrchr(dir, '/');
1457
1458 if ((!basename || basename[1] != '\0') &&
1459 lstat(dir, &sb) == 0 && S_ISDIR(sb.st_mode)) {
1460 strcpy(path, dir);
1461 basename = dir;
1462 } else if (basename == NULL) {
1463 strcpy(path, ".");
1464 strcpy(dir, "");
1465 basename = request->hr_file + 1;
1466 } else {
1467 *basename++ = '\0';
1468 strcpy(path, dir);
1469 }
1470 if (bozo_check_special_files(request, basename))
1471 return -1;
1472
1473 debug((httpd, DEBUG_FAT, "check_bzredirect: path %s", path));
1474
1475 if ((size_t)snprintf(redir, sizeof(redir), "%s/%s", path,
1476 REDIRECT_FILE) >= sizeof(redir)) {
1477 return bozo_http_error(httpd, 404, request,
1478 "redirectfile path too long");
1479 return -1;
1480 }
1481 if (lstat(redir, &sb) == 0) {
1482 if (!S_ISLNK(sb.st_mode))
1483 return 0;
1484 absolute = 0;
1485 } else {
1486 if((size_t)snprintf(redir, sizeof(redir), "%s/%s", path,
1487 ABSREDIRECT_FILE) >= sizeof(redir)) {
1488 return bozo_http_error(httpd, 404, request,
1489 "redirectfile path too long");
1490 }
1491 if (lstat(redir, &sb) < 0 || !S_ISLNK(sb.st_mode))
1492 return 0;
1493 absolute = 1;
1494 }
1495 debug((httpd, DEBUG_FAT, "check_bzredirect: calling readlink"));
1496 rv = readlink(redir, redirpath, sizeof redirpath - 1);
1497 if (rv == -1 || rv == 0) {
1498 debug((httpd, DEBUG_FAT, "readlink failed"));
1499 return 0;
1500 }
1501 redirpath[rv] = '\0';
1502 debug((httpd, DEBUG_FAT, "readlink returned \"%s\"", redirpath));
1503
1504 /* check if we need authentication */
1505 snprintf(path, sizeof(path), "%s/", dir);
1506 if (bozo_auth_check(request, path))
1507 return 1;
1508
1509 /* now we have the link pointer, redirect to the real place */
1510 if (!absolute && redirpath[0] != '/') {
1511 if ((size_t)snprintf(finalredir = redir, sizeof(redir), "%s%s/%s",
1512 (strlen(dir) > 0 ? "/" : ""), dir, redirpath) >= sizeof(redir)) {
1513 return bozo_http_error(httpd, 404, request,
1514 "redirect path too long");
1515 }
1516 } else
1517 finalredir = redirpath;
1518
1519 debug((httpd, DEBUG_FAT, "check_bzredirect: new redir %s", finalredir));
1520 handle_redirect(request, finalredir, absolute);
1521 return 1;
1522 }
1523
1524 /* this fixes the %HH hack that RFC2396 requires. */
1525 int
1526 bozo_decode_url_percent(bozo_httpreq_t *request, char *str)
1527 {
1528 bozohttpd_t *httpd = request->hr_httpd;
1529 char *s, *t, buf[3];
1530 char *end; /* if end is not-zero, we don't translate beyond that */
1531
1532 end = str + strlen(str);
1533
1534 /* fast forward to the first % */
1535 if ((s = strchr(str, '%')) == NULL)
1536 return 0;
1537
1538 t = s;
1539 do {
1540 if (end && s >= end) {
1541 debug((httpd, DEBUG_EXPLODING,
1542 "fu_%%: past end, filling out.."));
1543 while (*s)
1544 *t++ = *s++;
1545 break;
1546 }
1547 debug((httpd, DEBUG_EXPLODING,
1548 "fu_%%: got s == %%, s[1]s[2] == %c%c",
1549 s[1], s[2]));
1550 if (s[1] == '\0' || s[2] == '\0')
1551 return bozo_http_error(httpd, 400, request,
1552 "percent hack missing two chars afterwards");
1553 if (s[1] == '0' && s[2] == '0')
1554 return bozo_http_error(httpd, 404, request,
1555 "percent hack was %00");
1556 if (s[1] == '2' && s[2] == 'f')
1557 return bozo_http_error(httpd, 404, request,
1558 "percent hack was %2f (/)");
1559
1560 buf[0] = *++s;
1561 buf[1] = *++s;
1562 buf[2] = '\0';
1563 s++;
1564 *t = (char)strtol(buf, NULL, 16);
1565 debug((httpd, DEBUG_EXPLODING,
1566 "fu_%%: strtol put '%02x' into *t", *t));
1567 if (*t++ == '\0')
1568 return bozo_http_error(httpd, 400, request,
1569 "percent hack got a 0 back");
1570
1571 while (*s && *s != '%') {
1572 if (end && s >= end)
1573 break;
1574 *t++ = *s++;
1575 }
1576 } while (*s);
1577 *t = '\0';
1578
1579 debug((httpd, DEBUG_FAT, "bozo_decode_url_percent returns `%s'",
1580 request->hr_file));
1581
1582 return 0;
1583 }
1584
1585 /*
1586 * transform_request does this:
1587 * - ``expand'' %20 crapola
1588 * - punt if it doesn't start with /
1589 * - look for "http://myname/" and deal with it.
1590 * - maybe call bozo_process_cgi()
1591 * - check for ~user and call bozo_user_transform() if so
1592 * - if the length > 1, check for trailing slash. if so,
1593 * add the index.html file
1594 * - if the length is 1, return the index.html file
1595 * - disallow anything ending up with a file starting
1596 * at "/" or having ".." in it.
1597 * - anything else is a really weird internal error
1598 * - returns malloced file to serve, if unhandled
1599 */
1600 static int
1601 transform_request(bozo_httpreq_t *request, int *isindex)
1602 {
1603 bozohttpd_t *httpd = request->hr_httpd;
1604 char *file, *newfile = NULL;
1605 size_t len;
1606
1607 file = NULL;
1608 *isindex = 0;
1609 debug((httpd, DEBUG_FAT, "tf_req: file %s", request->hr_file));
1610 if (bozo_decode_url_percent(request, request->hr_file)) {
1611 goto bad_done;
1612 }
1613 if (check_virtual(request)) {
1614 goto bad_done;
1615 }
1616 file = request->hr_file;
1617
1618 if (file[0] != '/') {
1619 (void)bozo_http_error(httpd, 404, request, "unknown URL");
1620 goto bad_done;
1621 }
1622
1623 /* omit additional slashes at the beginning */
1624 while (file[1] == '/')
1625 file++;
1626
1627 /* fix file provided by user as it's used in other handlers */
1628 request->hr_file = file;
1629
1630 len = strlen(file);
1631
1632 #ifndef NO_USER_SUPPORT
1633 /* first of all expand user path */
1634 if (len > 1 && httpd->enable_users && file[1] == '~') {
1635 if (file[2] == '\0') {
1636 (void)bozo_http_error(httpd, 404, request,
1637 "missing username");
1638 goto bad_done;
1639 }
1640 if (strchr(file + 2, '/') == NULL) {
1641 char *userredirecturl;
1642 bozoasprintf(httpd, &userredirecturl, "%s/", file);
1643 handle_redirect(request, userredirecturl, 0);
1644 free(userredirecturl);
1645 return 0;
1646 }
1647 debug((httpd, DEBUG_FAT, "calling bozo_user_transform"));
1648
1649 if (!bozo_user_transform(request))
1650 return 0;
1651
1652 file = request->hr_file;
1653 len = strlen(file);
1654 }
1655 #endif /* NO_USER_SUPPORT */
1656
1657
1658 switch (check_bzredirect(request)) {
1659 case -1:
1660 goto bad_done;
1661 case 0:
1662 break;
1663 default:
1664 return 0;
1665 }
1666
1667 if (len > 1) {
1668 debug((httpd, DEBUG_FAT, "file[len-1] == %c", file[len-1]));
1669 if (file[len-1] == '/') { /* append index.html */
1670 *isindex = 1;
1671 debug((httpd, DEBUG_FAT, "appending index.html"));
1672 newfile = bozomalloc(httpd,
1673 len + strlen(httpd->index_html) + 1);
1674 strcpy(newfile, file + 1);
1675 strcat(newfile, httpd->index_html);
1676 } else
1677 newfile = bozostrdup(httpd, request, file + 1);
1678 } else if (len == 1) {
1679 debug((httpd, DEBUG_EXPLODING, "tf_req: len == 1"));
1680 newfile = bozostrdup(httpd, request, httpd->index_html);
1681 *isindex = 1;
1682 } else { /* len == 0 ? */
1683 (void)bozo_http_error(httpd, 500, request,
1684 "request->hr_file is nul?");
1685 goto bad_done;
1686 }
1687
1688 if (newfile == NULL) {
1689 (void)bozo_http_error(httpd, 500, request, "internal failure");
1690 goto bad_done;
1691 }
1692
1693 /*
1694 * stop traversing outside our domain
1695 *
1696 * XXX true security only comes from our parent using chroot(2)
1697 * before execve(2)'ing us. or our own built in chroot(2) support.
1698 */
1699
1700 debug((httpd, DEBUG_FAT, "newfile: %s", newfile));
1701
1702 if (*newfile == '/' || strcmp(newfile, "..") == 0 ||
1703 strstr(newfile, "/..") || strstr(newfile, "../")) {
1704 (void)bozo_http_error(httpd, 403, request, "illegal request");
1705 goto bad_done;
1706 }
1707
1708 if (bozo_auth_check(request, newfile))
1709 goto bad_done;
1710
1711 if (strlen(newfile)) {
1712 request->hr_oldfile = request->hr_file;
1713 request->hr_file = newfile;
1714 }
1715
1716 if (bozo_process_cgi(request))
1717 return 0;
1718
1719 if (bozo_process_lua(request))
1720 return 0;
1721
1722 debug((httpd, DEBUG_FAT, "transform_request set: %s", newfile));
1723 return 1;
1724 bad_done:
1725 debug((httpd, DEBUG_FAT, "transform_request returning: 0"));
1726 free(newfile);
1727 return 0;
1728 }
1729
1730 /*
1731 * can_gzip checks if the request supports and prefers gzip encoding.
1732 *
1733 * XXX: we do not consider the associated q with gzip in making our
1734 * decision which is broken.
1735 */
1736
1737 static int
1738 can_gzip(bozo_httpreq_t *request)
1739 {
1740 const char *pos;
1741 const char *tmp;
1742 size_t len;
1743
1744 /* First we decide if the request can be gzipped at all. */
1745
1746 /* not if we already are encoded... */
1747 tmp = bozo_content_encoding(request, request->hr_file);
1748 if (tmp && *tmp)
1749 return 0;
1750
1751 /* not if we are not asking for the whole file... */
1752 if (request->hr_last_byte_pos != -1 || request->hr_have_range)
1753 return 0;
1754
1755 /* Then we determine if gzip is on the cards. */
1756
1757 for (pos = request->hr_accept_encoding; pos && *pos; pos += len) {
1758 while (*pos == ' ')
1759 pos++;
1760
1761 len = strcspn(pos, ";,");
1762
1763 if ((len == 4 && strncasecmp("gzip", pos, 4) == 0) ||
1764 (len == 6 && strncasecmp("x-gzip", pos, 6) == 0))
1765 return 1;
1766
1767 if (pos[len] == ';')
1768 len += strcspn(&pos[len], ",");
1769
1770 if (pos[len])
1771 len++;
1772 }
1773
1774 return 0;
1775 }
1776
1777 /*
1778 * bozo_process_request does the following:
1779 * - check the request is valid
1780 * - process cgi-bin if necessary
1781 * - transform a filename if necesarry
1782 * - return the HTTP request
1783 */
1784 void
1785 bozo_process_request(bozo_httpreq_t *request)
1786 {
1787 bozohttpd_t *httpd = request->hr_httpd;
1788 struct stat sb;
1789 time_t timestamp;
1790 char *file;
1791 const char *type, *encoding;
1792 int fd, isindex;
1793
1794 /*
1795 * note that transform_request chdir()'s if required. also note
1796 * that cgi is handed here. if transform_request() returns 0
1797 * then the request has been handled already.
1798 */
1799 if (transform_request(request, &isindex) == 0)
1800 return;
1801
1802 fd = -1;
1803 encoding = NULL;
1804 if (can_gzip(request)) {
1805 bozoasprintf(httpd, &file, "%s.gz", request->hr_file);
1806 fd = open(file, O_RDONLY);
1807 if (fd >= 0)
1808 encoding = "gzip";
1809 free(file);
1810 }
1811
1812 file = request->hr_file;
1813
1814 if (fd < 0)
1815 fd = open(file, O_RDONLY);
1816
1817 if (fd < 0) {
1818 debug((httpd, DEBUG_FAT, "open failed: %s", strerror(errno)));
1819 switch (errno) {
1820 case EPERM:
1821 case EACCES:
1822 (void)bozo_http_error(httpd, 403, request,
1823 "no permission to open file");
1824 break;
1825 case ENAMETOOLONG:
1826 /*FALLTHROUGH*/
1827 case ENOENT:
1828 if (!bozo_dir_index(request, file, isindex))
1829 (void)bozo_http_error(httpd, 404, request,
1830 "no file");
1831 break;
1832 default:
1833 (void)bozo_http_error(httpd, 500, request, "open file");
1834 }
1835 goto cleanup_nofd;
1836 }
1837 if (fstat(fd, &sb) < 0) {
1838 (void)bozo_http_error(httpd, 500, request, "can't fstat");
1839 goto cleanup;
1840 }
1841 if (S_ISDIR(sb.st_mode)) {
1842 handle_redirect(request, NULL, 0);
1843 goto cleanup;
1844 }
1845
1846 if (request->hr_if_modified_since &&
1847 parse_http_date(request->hr_if_modified_since, ×tamp) &&
1848 timestamp >= sb.st_mtime) {
1849 /* XXX ignore subsecond of timestamp */
1850 bozo_printf(httpd, "%s 304 Not Modified\r\n",
1851 request->hr_proto);
1852 bozo_printf(httpd, "\r\n");
1853 bozo_flush(httpd, stdout);
1854 goto cleanup;
1855 }
1856
1857 /* validate requested range */
1858 if (request->hr_last_byte_pos == -1 ||
1859 request->hr_last_byte_pos >= sb.st_size)
1860 request->hr_last_byte_pos = sb.st_size - 1;
1861 if (request->hr_have_range &&
1862 request->hr_first_byte_pos > request->hr_last_byte_pos) {
1863 request->hr_have_range = 0; /* punt */
1864 request->hr_first_byte_pos = 0;
1865 request->hr_last_byte_pos = sb.st_size - 1;
1866 }
1867 debug((httpd, DEBUG_FAT, "have_range %d first_pos %lld last_pos %lld",
1868 request->hr_have_range,
1869 (long long)request->hr_first_byte_pos,
1870 (long long)request->hr_last_byte_pos));
1871 if (request->hr_have_range)
1872 bozo_printf(httpd, "%s 206 Partial Content\r\n",
1873 request->hr_proto);
1874 else
1875 bozo_printf(httpd, "%s 200 OK\r\n", request->hr_proto);
1876
1877 if (request->hr_proto != httpd->consts.http_09) {
1878 type = bozo_content_type(request, file);
1879 if (!encoding)
1880 encoding = bozo_content_encoding(request, file);
1881
1882 bozo_print_header(request, &sb, type, encoding);
1883 bozo_printf(httpd, "\r\n");
1884 }
1885 bozo_flush(httpd, stdout);
1886
1887 if (request->hr_method != HTTP_HEAD) {
1888 off_t szleft, cur_byte_pos;
1889
1890 szleft =
1891 request->hr_last_byte_pos - request->hr_first_byte_pos + 1;
1892 cur_byte_pos = request->hr_first_byte_pos;
1893
1894 retry:
1895 while (szleft) {
1896 size_t sz;
1897
1898 if ((off_t)httpd->mmapsz < szleft)
1899 sz = httpd->mmapsz;
1900 else
1901 sz = (size_t)szleft;
1902 if (mmap_and_write_part(httpd, fd, cur_byte_pos, sz)) {
1903 if (errno == ENOMEM) {
1904 httpd->mmapsz /= 2;
1905 if (httpd->mmapsz >= httpd->page_size)
1906 goto retry;
1907 }
1908 goto cleanup;
1909 }
1910 cur_byte_pos += sz;
1911 szleft -= sz;
1912 }
1913 }
1914 cleanup:
1915 close(fd);
1916 cleanup_nofd:
1917 close(STDIN_FILENO);
1918 close(STDOUT_FILENO);
1919 /*close(STDERR_FILENO);*/
1920 }
1921
1922 /* make sure we're not trying to access special files */
1923 int
1924 bozo_check_special_files(bozo_httpreq_t *request, const char *name)
1925 {
1926 bozohttpd_t *httpd = request->hr_httpd;
1927
1928 /* ensure basename(name) != special files */
1929 if (strcmp(name, DIRECT_ACCESS_FILE) == 0)
1930 return bozo_http_error(httpd, 403, request,
1931 "no permission to open direct access file");
1932 if (strcmp(name, REDIRECT_FILE) == 0)
1933 return bozo_http_error(httpd, 403, request,
1934 "no permission to open redirect file");
1935 if (strcmp(name, ABSREDIRECT_FILE) == 0)
1936 return bozo_http_error(httpd, 403, request,
1937 "no permission to open redirect file");
1938 if (strcmp(name, REMAP_FILE) == 0)
1939 return bozo_http_error(httpd, 403, request,
1940 "no permission to open redirect file");
1941 if (strcmp(name, AUTH_FILE) == 0)
1942 return bozo_http_error(httpd, 403, request,
1943 "no permission to open authfile");
1944
1945 return 0;
1946 }
1947
1948 /* generic header printing routine */
1949 void
1950 bozo_print_header(bozo_httpreq_t *request,
1951 struct stat *sbp, const char *type, const char *encoding)
1952 {
1953 bozohttpd_t *httpd = request->hr_httpd;
1954 off_t len;
1955 char date[40];
1956 bozoheaders_t *hdr;
1957
1958 SIMPLEQ_FOREACH(hdr, &request->hr_replheaders, h_next) {
1959 bozo_printf(httpd, "%s: %s\r\n", hdr->h_header,
1960 hdr->h_value);
1961 }
1962
1963 bozo_printf(httpd, "Date: %s\r\n", bozo_http_date(date, sizeof(date)));
1964 bozo_printf(httpd, "Server: %s\r\n", httpd->server_software);
1965 bozo_printf(httpd, "Accept-Ranges: bytes\r\n");
1966 if (sbp) {
1967 char filedate[40];
1968 struct tm *tm;
1969
1970 tm = gmtime(&sbp->st_mtime);
1971 strftime(filedate, sizeof filedate,
1972 "%a, %d %b %Y %H:%M:%S GMT", tm);
1973 bozo_printf(httpd, "Last-Modified: %s\r\n", filedate);
1974 }
1975 if (type && *type)
1976 bozo_printf(httpd, "Content-Type: %s\r\n", type);
1977 if (encoding && *encoding)
1978 bozo_printf(httpd, "Content-Encoding: %s\r\n", encoding);
1979 if (sbp) {
1980 if (request->hr_have_range) {
1981 len = request->hr_last_byte_pos -
1982 request->hr_first_byte_pos +1;
1983 bozo_printf(httpd,
1984 "Content-Range: bytes %qd-%qd/%qd\r\n",
1985 (long long) request->hr_first_byte_pos,
1986 (long long) request->hr_last_byte_pos,
1987 (long long) sbp->st_size);
1988 } else
1989 len = sbp->st_size;
1990 bozo_printf(httpd, "Content-Length: %qd\r\n", (long long)len);
1991 }
1992 if (request->hr_proto == httpd->consts.http_11)
1993 bozo_printf(httpd, "Connection: close\r\n");
1994 bozo_flush(httpd, stdout);
1995 }
1996
1997 #ifndef NO_DEBUG
1998 void
1999 debug__(bozohttpd_t *httpd, int level, const char *fmt, ...)
2000 {
2001 va_list ap;
2002 int savederrno;
2003
2004 /* only log if the level is low enough */
2005 if (httpd->debug < level)
2006 return;
2007
2008 savederrno = errno;
2009 va_start(ap, fmt);
2010 if (httpd->logstderr) {
2011 vfprintf(stderr, fmt, ap);
2012 fputs("\n", stderr);
2013 } else
2014 vsyslog(LOG_DEBUG, fmt, ap);
2015 va_end(ap);
2016 errno = savederrno;
2017 }
2018 #endif /* NO_DEBUG */
2019
2020 /* these are like warn() and err(), except for syslog not stderr */
2021 void
2022 bozowarn(bozohttpd_t *httpd, const char *fmt, ...)
2023 {
2024 va_list ap;
2025
2026 va_start(ap, fmt);
2027 if (httpd->logstderr || isatty(STDERR_FILENO)) {
2028 //fputs("warning: ", stderr);
2029 vfprintf(stderr, fmt, ap);
2030 fputs("\n", stderr);
2031 } else
2032 vsyslog(LOG_INFO, fmt, ap);
2033 va_end(ap);
2034 }
2035
2036 void
2037 bozoerr(bozohttpd_t *httpd, int code, const char *fmt, ...)
2038 {
2039 va_list ap;
2040
2041 va_start(ap, fmt);
2042 if (httpd->logstderr || isatty(STDERR_FILENO)) {
2043 //fputs("error: ", stderr);
2044 vfprintf(stderr, fmt, ap);
2045 fputs("\n", stderr);
2046 } else
2047 vsyslog(LOG_ERR, fmt, ap);
2048 va_end(ap);
2049 exit(code);
2050 }
2051
2052 void
2053 bozoasprintf(bozohttpd_t *httpd, char **str, const char *fmt, ...)
2054 {
2055 va_list ap;
2056 int e;
2057
2058 va_start(ap, fmt);
2059 e = vasprintf(str, fmt, ap);
2060 va_end(ap);
2061
2062 if (e < 0)
2063 bozoerr(httpd, EXIT_FAILURE, "asprintf");
2064 }
2065
2066 /*
2067 * this escapes HTML tags. returns allocated escaped
2068 * string if needed, or NULL on allocation failure or
2069 * lack of escape need.
2070 * call with NULL httpd in error paths, to avoid recursive
2071 * malloc failure. call with valid httpd in normal paths
2072 * to get automatic allocation failure handling.
2073 */
2074 char *
2075 bozo_escape_html(bozohttpd_t *httpd, const char *url)
2076 {
2077 int i, j;
2078 char *tmp;
2079 size_t len;
2080
2081 for (i = 0, j = 0; url[i]; i++) {
2082 switch (url[i]) {
2083 case '<':
2084 case '>':
2085 j += 4;
2086 break;
2087 case '&':
2088 j += 5;
2089 break;
2090 }
2091 }
2092
2093 if (j == 0)
2094 return NULL;
2095
2096 /*
2097 * we need to handle being called from different
2098 * pathnames.
2099 */
2100 len = strlen(url) + j;
2101 if (httpd)
2102 tmp = bozomalloc(httpd, len);
2103 else if ((tmp = malloc(len)) == 0)
2104 return NULL;
2105
2106 for (i = 0, j = 0; url[i]; i++) {
2107 switch (url[i]) {
2108 case '<':
2109 memcpy(tmp + j, "<", 4);
2110 j += 4;
2111 break;
2112 case '>':
2113 memcpy(tmp + j, ">", 4);
2114 j += 4;
2115 break;
2116 case '&':
2117 memcpy(tmp + j, "&", 5);
2118 j += 5;
2119 break;
2120 default:
2121 tmp[j++] = url[i];
2122 }
2123 }
2124 tmp[j] = 0;
2125
2126 return tmp;
2127 }
2128
2129 /* short map between error code, and short/long messages */
2130 static struct errors_map {
2131 int code; /* HTTP return code */
2132 const char *shortmsg; /* short version of message */
2133 const char *longmsg; /* long version of message */
2134 } errors_map[] = {
2135 { 400, "400 Bad Request", "The request was not valid", },
2136 { 401, "401 Unauthorized", "No authorization", },
2137 { 403, "403 Forbidden", "Access to this item has been denied",},
2138 { 404, "404 Not Found", "This item has not been found", },
2139 { 408, "408 Request Timeout", "This request took too long", },
2140 { 413, "413 Payload Too Large", "Use smaller requests", },
2141 { 417, "417 Expectation Failed","Expectations not available", },
2142 { 420, "420 Enhance Your Calm","Chill, Winston", },
2143 { 500, "500 Internal Error", "An error occured on the server", },
2144 { 501, "501 Not Implemented", "This request is not available", },
2145 { 0, NULL, NULL, },
2146 };
2147
2148 static const char *help = "DANGER! WILL ROBINSON! DANGER!";
2149
2150 static const char *
2151 http_errors_short(int code)
2152 {
2153 struct errors_map *ep;
2154
2155 for (ep = errors_map; ep->code; ep++)
2156 if (ep->code == code)
2157 return (ep->shortmsg);
2158 return (help);
2159 }
2160
2161 static const char *
2162 http_errors_long(int code)
2163 {
2164 struct errors_map *ep;
2165
2166 for (ep = errors_map; ep->code; ep++)
2167 if (ep->code == code)
2168 return (ep->longmsg);
2169 return (help);
2170 }
2171
2172 /* the follow functions and variables are used in handling HTTP errors */
2173 /* ARGSUSED */
2174 int
2175 bozo_http_error(bozohttpd_t *httpd, int code, bozo_httpreq_t *request,
2176 const char *msg)
2177 {
2178 char portbuf[20];
2179 const char *header = http_errors_short(code);
2180 const char *reason = http_errors_long(code);
2181 const char *proto = (request && request->hr_proto) ?
2182 request->hr_proto : httpd->consts.http_11;
2183 int size;
2184 bozoheaders_t *hdr;
2185
2186 debug((httpd, DEBUG_FAT, "bozo_http_error %d: %s", code, msg));
2187 if (header == NULL || reason == NULL) {
2188 bozoerr(httpd, 1,
2189 "bozo_http_error() failed (short = %p, long = %p)",
2190 header, reason);
2191 return code;
2192 }
2193
2194 if (request && request->hr_serverport &&
2195 strcmp(request->hr_serverport, "80") != 0)
2196 snprintf(portbuf, sizeof(portbuf), ":%s",
2197 request->hr_serverport);
2198 else
2199 portbuf[0] = '\0';
2200
2201 if (request && request->hr_file) {
2202 char *file = NULL, *user = NULL;
2203 int file_alloc = 0;
2204 const char *hostname = BOZOHOST(httpd, request);
2205
2206 /* bozo_escape_html() failure here is just too bad. */
2207 file = bozo_escape_html(NULL, request->hr_file);
2208 if (file == NULL)
2209 file = request->hr_file;
2210 else
2211 file_alloc = 1;
2212
2213 #ifndef NO_USER_SUPPORT
2214 if (request->hr_user != NULL) {
2215 char *user_escaped;
2216
2217 user_escaped = bozo_escape_html(NULL, request->hr_user);
2218 if (user_escaped == NULL)
2219 user_escaped = request->hr_user;
2220 /* expand username to ~user/ */
2221 bozoasprintf(httpd, &user, "~%s/", user_escaped);
2222 if (user_escaped != request->hr_user)
2223 free(user_escaped);
2224 }
2225 #endif /* !NO_USER_SUPPORT */
2226
2227 size = snprintf(httpd->errorbuf, BUFSIZ,
2228 "<html><head><title>%s</title></head>\n"
2229 "<body><h1>%s</h1>\n"
2230 "%s%s: <pre>%s</pre>\n"
2231 "<hr><address><a href=\"//%s%s/\">%s%s</a></address>\n"
2232 "</body></html>\n",
2233 header, header,
2234 user ? user : "", file,
2235 reason, hostname, portbuf, hostname, portbuf);
2236 free(user);
2237 if (size >= (int)BUFSIZ) {
2238 bozowarn(httpd,
2239 "bozo_http_error buffer too small, truncated");
2240 size = (int)BUFSIZ;
2241 }
2242
2243 if (file_alloc)
2244 free(file);
2245 } else
2246 size = 0;
2247
2248 bozo_printf(httpd, "%s %s\r\n", proto, header);
2249
2250 if (request) {
2251 bozo_auth_check_401(request, code);
2252 SIMPLEQ_FOREACH(hdr, &request->hr_replheaders, h_next) {
2253 bozo_printf(httpd, "%s: %s\r\n", hdr->h_header,
2254 hdr->h_value);
2255 }
2256 }
2257
2258 bozo_printf(httpd, "Content-Type: text/html\r\n");
2259 bozo_printf(httpd, "Content-Length: %d\r\n", size);
2260 bozo_printf(httpd, "Server: %s\r\n", httpd->server_software);
2261 if (request && request->hr_allow)
2262 bozo_printf(httpd, "Allow: %s\r\n", request->hr_allow);
2263 /* RFC 7231 (HTTP/1.1) 6.5.7 */
2264 if (code == 408 && request->hr_proto == httpd->consts.http_11)
2265 bozo_printf(httpd, "Connection: close\r\n");
2266 bozo_printf(httpd, "\r\n");
2267 /* According to the RFC 2616 sec. 9.4 HEAD method MUST NOT return a
2268 * message-body in the response */
2269 if (size && request && request->hr_method != HTTP_HEAD)
2270 bozo_printf(httpd, "%s", httpd->errorbuf);
2271 bozo_flush(httpd, stdout);
2272
2273 return code;
2274 }
2275
2276 /* Below are various modified libc functions */
2277
2278 /*
2279 * returns -1 in lenp if the string ran out before finding a delimiter,
2280 * but is otherwise the same as strsep. Note that the length must be
2281 * correctly passed in.
2282 */
2283 char *
2284 bozostrnsep(char **strp, const char *delim, ssize_t *lenp)
2285 {
2286 char *s;
2287 const char *spanp;
2288 int c, sc;
2289 char *tok;
2290
2291 if ((s = *strp) == NULL)
2292 return (NULL);
2293 for (tok = s;;) {
2294 if (lenp && --(*lenp) == -1)
2295 return (NULL);
2296 c = *s++;
2297 spanp = delim;
2298 do {
2299 if ((sc = *spanp++) == c) {
2300 if (c == 0)
2301 s = NULL;
2302 else
2303 s[-1] = '\0';
2304 *strp = s;
2305 return (tok);
2306 }
2307 } while (sc != 0);
2308 }
2309 /* NOTREACHED */
2310 }
2311
2312 /*
2313 * inspired by fgetln(3), but works for fd's. should work identically
2314 * except it, however, does *not* return the newline, and it does nul
2315 * terminate the string.
2316 */
2317 char *
2318 bozodgetln(bozohttpd_t *httpd, int fd, ssize_t *lenp,
2319 ssize_t (*readfn)(bozohttpd_t *, int, void *, size_t))
2320 {
2321 ssize_t len;
2322 int got_cr = 0;
2323 char c, *nbuffer;
2324
2325 /* initialise */
2326 if (httpd->getln_buflen == 0) {
2327 /* should be plenty for most requests */
2328 httpd->getln_buflen = 128;
2329 httpd->getln_buffer = malloc((size_t)httpd->getln_buflen);
2330 if (httpd->getln_buffer == NULL) {
2331 httpd->getln_buflen = 0;
2332 return NULL;
2333 }
2334 }
2335 len = 0;
2336
2337 /*
2338 * we *have* to read one byte at a time, to not break cgi
2339 * programs (for we pass stdin off to them). could fix this
2340 * by becoming a fd-passing program instead of just exec'ing
2341 * the program
2342 *
2343 * the above is no longer true, we are the fd-passing
2344 * program already.
2345 */
2346 for (; readfn(httpd, fd, &c, 1) == 1; ) {
2347 debug((httpd, DEBUG_EXPLODING, "bozodgetln read %c", c));
2348
2349 if (len >= httpd->getln_buflen - 1) {
2350 httpd->getln_buflen *= 2;
2351 debug((httpd, DEBUG_EXPLODING, "bozodgetln: "
2352 "reallocating buffer to buflen %zu",
2353 httpd->getln_buflen));
2354 nbuffer = bozorealloc(httpd, httpd->getln_buffer,
2355 (size_t)httpd->getln_buflen);
2356 httpd->getln_buffer = nbuffer;
2357 }
2358
2359 httpd->getln_buffer[len++] = c;
2360 if (c == '\r') {
2361 got_cr = 1;
2362 continue;
2363 } else if (c == '\n') {
2364 /*
2365 * HTTP/1.1 spec says to ignore CR and treat
2366 * LF as the real line terminator. even though
2367 * the same spec defines CRLF as the line
2368 * terminator, it is recommended in section 19.3
2369 * to do the LF trick for tolerance.
2370 */
2371 if (got_cr)
2372 len -= 2;
2373 else
2374 len -= 1;
2375 break;
2376 }
2377
2378 }
2379 httpd->getln_buffer[len] = '\0';
2380 debug((httpd, DEBUG_OBESE, "bozodgetln returns: ``%s'' with len %zd",
2381 httpd->getln_buffer, len));
2382 *lenp = len;
2383 return httpd->getln_buffer;
2384 }
2385
2386 void *
2387 bozorealloc(bozohttpd_t *httpd, void *ptr, size_t size)
2388 {
2389 void *p;
2390
2391 p = realloc(ptr, size);
2392 if (p)
2393 return p;
2394
2395 (void)bozo_http_error(httpd, 500, NULL, "memory allocation failure");
2396 exit(EXIT_FAILURE);
2397 }
2398
2399 void *
2400 bozomalloc(bozohttpd_t *httpd, size_t size)
2401 {
2402 void *p;
2403
2404 p = malloc(size);
2405 if (p)
2406 return p;
2407
2408 (void)bozo_http_error(httpd, 500, NULL, "memory allocation failure");
2409 exit(EXIT_FAILURE);
2410 }
2411
2412 char *
2413 bozostrdup(bozohttpd_t *httpd, bozo_httpreq_t *request, const char *str)
2414 {
2415 char *p;
2416
2417 p = strdup(str);
2418 if (p)
2419 return p;
2420
2421 if (!request)
2422 bozoerr(httpd, EXIT_FAILURE, "strdup");
2423
2424 (void)bozo_http_error(httpd, 500, request, "memory allocation failure");
2425 exit(EXIT_FAILURE);
2426 }
2427
2428 /* set default values in bozohttpd_t struct */
2429 int
2430 bozo_init_httpd(bozohttpd_t *httpd)
2431 {
2432 /* make sure everything is clean */
2433 (void) memset(httpd, 0x0, sizeof(*httpd));
2434
2435 /* constants */
2436 httpd->consts.http_09 = "HTTP/0.9";
2437 httpd->consts.http_10 = "HTTP/1.0";
2438 httpd->consts.http_11 = "HTTP/1.1";
2439 httpd->consts.text_plain = "text/plain";
2440
2441 /* mmap region size */
2442 httpd->mmapsz = BOZO_MMAPSZ;
2443
2444 /* error buffer for bozo_http_error() */
2445 if ((httpd->errorbuf = malloc(BUFSIZ)) == NULL) {
2446 (void) fprintf(stderr,
2447 "bozohttpd: memory_allocation failure\n");
2448 return 0;
2449 }
2450 #ifndef NO_LUA_SUPPORT
2451 SIMPLEQ_INIT(&httpd->lua_states);
2452 #endif
2453 return 1;
2454 }
2455
2456 /* set default values in bozoprefs_t struct */
2457 int
2458 bozo_init_prefs(bozohttpd_t *httpd, bozoprefs_t *prefs)
2459 {
2460 int rv = 0;
2461
2462 /* make sure everything is clean */
2463 (void) memset(prefs, 0x0, sizeof(*prefs));
2464
2465 /* set up default values */
2466 if (!bozo_set_pref(httpd, prefs, "server software", SERVER_SOFTWARE))
2467 rv = 1;
2468 if (!bozo_set_pref(httpd, prefs, "index.html", INDEX_HTML))
2469 rv = 1;
2470 if (!bozo_set_pref(httpd, prefs, "public_html", PUBLIC_HTML))
2471 rv = 1;
2472 if (!bozo_set_pref(httpd, prefs, "initial timeout", INITIAL_TIMEOUT))
2473 rv = 1;
2474 if (!bozo_set_pref(httpd, prefs, "header timeout", HEADER_WAIT_TIME))
2475 rv = 1;
2476 if (!bozo_set_pref(httpd, prefs, "request timeout", TOTAL_MAX_REQ_TIME))
2477 rv = 1;
2478
2479 return rv;
2480 }
2481
2482 /* set default values */
2483 int
2484 bozo_set_defaults(bozohttpd_t *httpd, bozoprefs_t *prefs)
2485 {
2486 return bozo_init_httpd(httpd) && bozo_init_prefs(httpd, prefs);
2487 }
2488
2489 /* set the virtual host name, port and root */
2490 int
2491 bozo_setup(bozohttpd_t *httpd, bozoprefs_t *prefs, const char *vhost,
2492 const char *root)
2493 {
2494 struct passwd *pw;
2495 extern char **environ;
2496 static char *cleanenv[1] = { NULL };
2497 uid_t uid;
2498 int uidset = 0;
2499 char *chrootdir;
2500 char *username;
2501 char *portnum;
2502 char *cp;
2503 int dirtyenv;
2504
2505 dirtyenv = 0;
2506
2507 if (vhost == NULL) {
2508 httpd->virthostname = bozomalloc(httpd, MAXHOSTNAMELEN+1);
2509 if (gethostname(httpd->virthostname, MAXHOSTNAMELEN+1) < 0)
2510 bozoerr(httpd, 1, "gethostname");
2511 httpd->virthostname[MAXHOSTNAMELEN] = '\0';
2512 } else {
2513 httpd->virthostname = bozostrdup(httpd, NULL, vhost);
2514 }
2515 httpd->slashdir = bozostrdup(httpd, NULL, root);
2516 if ((portnum = bozo_get_pref(prefs, "port number")) != NULL) {
2517 httpd->bindport = bozostrdup(httpd, NULL, portnum);
2518 }
2519
2520 /* go over preferences now */
2521 if ((cp = bozo_get_pref(prefs, "numeric")) != NULL &&
2522 strcmp(cp, "true") == 0) {
2523 httpd->numeric = 1;
2524 }
2525 if ((cp = bozo_get_pref(prefs, "log to stderr")) != NULL &&
2526 strcmp(cp, "true") == 0) {
2527 httpd->logstderr = 1;
2528 }
2529 if ((cp = bozo_get_pref(prefs, "bind address")) != NULL) {
2530 httpd->bindaddress = bozostrdup(httpd, NULL, cp);
2531 }
2532 if ((cp = bozo_get_pref(prefs, "background")) != NULL) {
2533 httpd->background = atoi(cp);
2534 }
2535 if ((cp = bozo_get_pref(prefs, "foreground")) != NULL &&
2536 strcmp(cp, "true") == 0) {
2537 httpd->foreground = 1;
2538 }
2539 if ((cp = bozo_get_pref(prefs, "pid file")) != NULL) {
2540 httpd->pidfile = bozostrdup(httpd, NULL, cp);
2541 }
2542 if ((cp = bozo_get_pref(prefs, "unknown slash")) != NULL &&
2543 strcmp(cp, "true") == 0) {
2544 httpd->unknown_slash = 1;
2545 }
2546 if ((cp = bozo_get_pref(prefs, "virtual base")) != NULL) {
2547 httpd->virtbase = bozostrdup(httpd, NULL, cp);
2548 }
2549 if ((cp = bozo_get_pref(prefs, "enable users")) != NULL &&
2550 strcmp(cp, "true") == 0) {
2551 httpd->enable_users = 1;
2552 }
2553 if ((cp = bozo_get_pref(prefs, "enable user cgibin")) != NULL &&
2554 strcmp(cp, "true") == 0) {
2555 httpd->enable_cgi_users = 1;
2556 }
2557 if ((cp = bozo_get_pref(prefs, "dirty environment")) != NULL &&
2558 strcmp(cp, "true") == 0) {
2559 dirtyenv = 1;
2560 }
2561 if ((cp = bozo_get_pref(prefs, "hide dots")) != NULL &&
2562 strcmp(cp, "true") == 0) {
2563 httpd->hide_dots = 1;
2564 }
2565 if ((cp = bozo_get_pref(prefs, "directory indexing")) != NULL &&
2566 strcmp(cp, "true") == 0) {
2567 httpd->dir_indexing = 1;
2568 }
2569 if ((cp = bozo_get_pref(prefs, "public_html")) != NULL) {
2570 httpd->public_html = bozostrdup(httpd, NULL, cp);
2571 }
2572 if ((cp = bozo_get_pref(prefs, "initial timeout")) != NULL) {
2573 httpd->initial_timeout = atoi(cp);
2574 }
2575 if ((cp = bozo_get_pref(prefs, "header timeout")) != NULL) {
2576 httpd->header_timeout = atoi(cp);
2577 }
2578 if ((cp = bozo_get_pref(prefs, "request timeout")) != NULL) {
2579 httpd->request_timeout = atoi(cp);
2580 }
2581 httpd->server_software =
2582 bozostrdup(httpd, NULL, bozo_get_pref(prefs, "server software"));
2583 httpd->index_html =
2584 bozostrdup(httpd, NULL, bozo_get_pref(prefs, "index.html"));
2585
2586 /*
2587 * initialise ssl and daemon mode if necessary.
2588 */
2589 bozo_ssl_init(httpd);
2590 bozo_daemon_init(httpd);
2591
2592 username = bozo_get_pref(prefs, "username");
2593 if (username != NULL) {
2594 if ((pw = getpwnam(username)) == NULL)
2595 bozoerr(httpd, 1, "getpwnam(%s): %s", username,
2596 strerror(errno));
2597 if (initgroups(pw->pw_name, pw->pw_gid) == -1)
2598 bozoerr(httpd, 1, "initgroups: %s", strerror(errno));
2599 if (setgid(pw->pw_gid) == -1)
2600 bozoerr(httpd, 1, "setgid(%u): %s", pw->pw_gid,
2601 strerror(errno));
2602 uid = pw->pw_uid;
2603 uidset = 1;
2604 }
2605 /*
2606 * handle chroot.
2607 */
2608 if ((chrootdir = bozo_get_pref(prefs, "chroot dir")) != NULL) {
2609 httpd->rootdir = bozostrdup(httpd, NULL, chrootdir);
2610 if (chdir(httpd->rootdir) == -1)
2611 bozoerr(httpd, 1, "chdir(%s): %s", httpd->rootdir,
2612 strerror(errno));
2613 if (chroot(httpd->rootdir) == -1)
2614 bozoerr(httpd, 1, "chroot(%s): %s", httpd->rootdir,
2615 strerror(errno));
2616 }
2617
2618 if (uidset && setuid(uid) == -1)
2619 bozoerr(httpd, 1, "setuid(%d): %s", uid, strerror(errno));
2620
2621 /*
2622 * prevent info leakage between different compartments.
2623 * some PATH values in the environment would be invalided
2624 * by chroot. cross-user settings might result in undesirable
2625 * effects.
2626 */
2627 if ((chrootdir != NULL || username != NULL) && !dirtyenv)
2628 environ = cleanenv;
2629
2630 #ifdef _SC_PAGESIZE
2631 httpd->page_size = (long)sysconf(_SC_PAGESIZE);
2632 #else
2633 httpd->page_size = 4096;
2634 #endif
2635 debug((httpd, DEBUG_OBESE, "myname is %s, slashdir is %s",
2636 httpd->virthostname, httpd->slashdir));
2637
2638 return 1;
2639 }
2640
2641 int
2642 bozo_get_version(char *buf, size_t size)
2643 {
2644 return snprintf(buf, size, "%s", SERVER_SOFTWARE);
2645 }
2646