bozohttpd.c revision 1.125 1 /* $NetBSD: bozohttpd.c,v 1.125 2021/02/11 09:23:55 mrg Exp $ */
2
3 /* $eterna: bozohttpd.c,v 1.178 2011/11/18 09:21:15 mrg Exp $ */
4
5 /*
6 * Copyright (c) 1997-2020 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 * - access lists via libwrap via inetd/tcpd
44 * - virtual hosting
45 * - not that we do not even pretend to understand MIME, but
46 * rely only on the HTTP specification
47 * - ipv6 support
48 * - automatic `index.html' generation
49 * - configurable server name
50 * - directory index generation
51 * - daemon mode (lacks libwrap support)
52 * - .htpasswd support
53 */
54
55 /*
56 * requirements for minimal http/1.1 (at least, as documented in
57 * RFC 2616 (HTTP/1.1):
58 *
59 * - 14.11: content-encoding handling. [1]
60 *
61 * - 14.13: content-length handling. this is only a SHOULD header
62 * thus we could just not send it ever. [1]
63 *
64 * - 14.17: content-type handling. [1]
65 *
66 * - 14.28: if-unmodified-since handling. if-modified-since is
67 * done since, shouldn't be too hard for this one.
68 *
69 * [1] need to revisit to ensure proper behaviour
70 *
71 * and the following is a list of features that we do not need
72 * to have due to other limits, or are too lazy. there are more
73 * of these than are listed, but these are of particular note,
74 * and could perhaps be implemented.
75 *
76 * - 3.5/3.6: content/transfer codings. probably can ignore
77 * this? we "SHOULD"n't. but 4.4 says we should ignore a
78 * `content-length' header upon reciept of a `transfer-encoding'
79 * header.
80 *
81 * - 5.1.1: request methods. only MUST support GET and HEAD,
82 * but there are new ones besides POST that are currently
83 * supported: OPTIONS PUT DELETE TRACE and CONNECT, plus
84 * extensions not yet known?
85 *
86 * - 10.1: we can ignore informational status codes
87 *
88 * - 10.3.3/10.3.4/10.3.8: just use '302' codes always.
89 *
90 * - 14.1/14.2/14.3/14.27: we do not support Accept: headers.
91 * just ignore them and send the request anyway. they are
92 * only SHOULD.
93 *
94 * - 14.5/14.16/14.35: only support simple ranges: %d- and %d-%d
95 * would be nice to support more.
96 *
97 * - 14.9: we aren't a cache.
98 *
99 * - 14.15: content-md5 would be nice.
100 *
101 * - 14.24/14.26/14.27: if-match, if-none-match, if-range. be
102 * nice to support this.
103 *
104 * - 14.44: Vary: seems unneeded. ignore it for now.
105 */
106
107 #ifndef INDEX_HTML
108 #define INDEX_HTML "index.html"
109 #endif
110 #ifndef SERVER_SOFTWARE
111 #define SERVER_SOFTWARE "bozohttpd/20210210"
112 #endif
113 #ifndef PUBLIC_HTML
114 #define PUBLIC_HTML "public_html"
115 #endif
116
117 #ifndef USE_ARG
118 #define USE_ARG(x) /*LINTED*/(void)&(x)
119 #endif
120
121 /*
122 * And so it begins ..
123 */
124
125 #include <sys/param.h>
126 #include <sys/socket.h>
127 #include <sys/time.h>
128 #include <sys/mman.h>
129
130 #include <arpa/inet.h>
131
132 #include <ctype.h>
133 #include <dirent.h>
134 #include <errno.h>
135 #include <fcntl.h>
136 #include <netdb.h>
137 #include <pwd.h>
138 #include <grp.h>
139 #include <stdarg.h>
140 #include <stdlib.h>
141 #include <strings.h>
142 #include <string.h>
143 #include <syslog.h>
144 #include <time.h>
145 #include <unistd.h>
146
147 #include "bozohttpd.h"
148
149 #ifndef SSL_TIMEOUT
150 #define SSL_TIMEOUT "30" /* wait for 30 seconds for ssl handshake */
151 #endif
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 /*
173 * List of special file that we should never serve.
174 */
175 struct {
176 const char *file;
177 const char *name;
178 } specials[] = {
179 { DIRECT_ACCESS_FILE, "rejected direct access request" },
180 { REDIRECT_FILE, "rejected redirect request" },
181 { ABSREDIRECT_FILE, "rejected absredirect request" },
182 { REMAP_FILE, "rejected remap request" },
183 { AUTH_FILE, "rejected authfile request" },
184 { NULL, NULL },
185 };
186
187 volatile sig_atomic_t bozo_timeout_hit;
188
189 /*
190 * check there's enough space in the prefs and names arrays.
191 */
192 static int
193 size_arrays(bozohttpd_t *httpd, bozoprefs_t *bozoprefs, size_t needed)
194 {
195 size_t len = sizeof(char *) * needed;
196
197 if (bozoprefs->size == 0) {
198 /* only get here first time around */
199 bozoprefs->name = bozomalloc(httpd, len);
200 bozoprefs->value = bozomalloc(httpd, len);
201 } else if (bozoprefs->count == bozoprefs->size) {
202 /* only uses 'needed' when filled array */
203 bozoprefs->name = bozorealloc(httpd, bozoprefs->name, len);
204 bozoprefs->value = bozorealloc(httpd, bozoprefs->value, len);
205 }
206
207 bozoprefs->size = needed;
208 return 1;
209 }
210
211 static ssize_t
212 findvar(bozoprefs_t *bozoprefs, const char *name)
213 {
214 size_t i;
215
216 for (i = 0; i < bozoprefs->count; i++)
217 if (strcmp(bozoprefs->name[i], name) == 0)
218 return (ssize_t)i;
219 return -1;
220 }
221
222 int
223 bozo_set_pref(bozohttpd_t *httpd, bozoprefs_t *bozoprefs,
224 const char *name, const char *value)
225 {
226 ssize_t i;
227
228 if ((i = findvar(bozoprefs, name)) < 0) {
229 /* add the element to the array */
230 if (!size_arrays(httpd, bozoprefs, bozoprefs->size + 15))
231 return 0;
232 i = bozoprefs->count++;
233 bozoprefs->name[i] = bozostrdup(httpd, NULL, name);
234 } else {
235 /* replace the element in the array */
236 free(bozoprefs->value[i]);
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 || in == 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 ? *query : "", *proto ? *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_free);
342 if (request->hr_file_free != request->hr_oldfile)
343 free(request->hr_oldfile);
344 free(request->hr_query);
345 free(request->hr_host);
346 bozo_user_free(request->hr_user);
347 bozo_auth_cleanup(request);
348 for (hdr = SIMPLEQ_FIRST(&request->hr_headers); hdr;
349 hdr = SIMPLEQ_NEXT(hdr, h_next)) {
350 free(hdr->h_value);
351 free(hdr->h_header);
352 free(ohdr);
353 ohdr = hdr;
354 }
355 free(ohdr);
356 ohdr = NULL;
357 for (hdr = SIMPLEQ_FIRST(&request->hr_replheaders); hdr;
358 hdr = SIMPLEQ_NEXT(hdr, h_next)) {
359 free(hdr->h_value);
360 free(hdr->h_header);
361 free(ohdr);
362 ohdr = hdr;
363 }
364 free(ohdr);
365
366 free(request);
367 }
368
369 /*
370 * send a HTTP/1.1 408 response if we timeout.
371 */
372 /* ARGSUSED */
373 static void
374 alarmer(int sig)
375 {
376 USE_ARG(sig);
377 bozo_timeout_hit = 1;
378 }
379
380
381 /*
382 * set a timeout for "ssl", "initial", "header", or "request".
383 */
384 int
385 bozo_set_timeout(bozohttpd_t *httpd, bozoprefs_t *prefs,
386 const char *target, const char *val)
387 {
388 const char **cur, *timeouts[] = {
389 "ssl timeout",
390 "initial timeout",
391 "header timeout",
392 "request timeout",
393 NULL,
394 };
395 /* adjust minlen if more timeouts appear with conflicting names */
396 const size_t minlen = 1;
397 size_t len = strlen(target);
398
399 for (cur = timeouts; len >= minlen && *cur; cur++) {
400 if (strncmp(target, *cur, len) == 0) {
401 bozo_set_pref(httpd, prefs, *cur, val);
402 return 0;
403 }
404 }
405 return 1;
406 }
407
408 /*
409 * a list of header quirks: currently, a list of headers that
410 * can't be folded into a single line.
411 */
412 const char *header_quirks[] = { "WWW-Authenticate", NULL };
413
414 /*
415 * add or merge this header (val: str) into the requests list
416 */
417 static bozoheaders_t *
418 addmerge_header(bozo_httpreq_t *request, struct qheaders *headers,
419 const char *val, const char *str, ssize_t len)
420 {
421 struct bozohttpd_t *httpd = request->hr_httpd;
422 struct bozoheaders *hdr = NULL;
423 const char **quirk;
424
425 USE_ARG(len);
426 for (quirk = header_quirks; *quirk; quirk++)
427 if (strcasecmp(*quirk, val) == 0)
428 break;
429
430 if (*quirk == NULL) {
431 /* do we exist already? */
432 SIMPLEQ_FOREACH(hdr, headers, h_next) {
433 if (strcasecmp(val, hdr->h_header) == 0)
434 break;
435 }
436 }
437
438 if (hdr) {
439 /* yup, merge it in */
440 char *nval;
441
442 bozoasprintf(httpd, &nval, "%s, %s", hdr->h_value, str);
443 free(hdr->h_value);
444 hdr->h_value = nval;
445 } else {
446 /* nope, create a new one */
447
448 hdr = bozomalloc(httpd, sizeof *hdr);
449 hdr->h_header = bozostrdup(httpd, request, val);
450 if (str && *str)
451 hdr->h_value = bozostrdup(httpd, request, str);
452 else
453 hdr->h_value = bozostrdup(httpd, request, " ");
454
455 SIMPLEQ_INSERT_TAIL(headers, hdr, h_next);
456 request->hr_nheaders++;
457 }
458
459 return hdr;
460 }
461
462 bozoheaders_t *
463 addmerge_reqheader(bozo_httpreq_t *request, const char *val, const char *str,
464 ssize_t len)
465 {
466
467 return addmerge_header(request, &request->hr_headers, val, str, len);
468 }
469
470 bozoheaders_t *
471 addmerge_replheader(bozo_httpreq_t *request, const char *val, const char *str,
472 ssize_t len)
473 {
474
475 return addmerge_header(request, &request->hr_replheaders,
476 val, str, len);
477 }
478
479 /*
480 * as the prototype string is not constant (eg, "HTTP/1.1" is equivalent
481 * to "HTTP/001.01"), we MUST parse this.
482 */
483 static int
484 process_proto(bozo_httpreq_t *request, const char *proto)
485 {
486 struct bozohttpd_t *httpd = request->hr_httpd;
487 char majorstr[16], *minorstr;
488 int majorint, minorint;
489
490 if (proto == NULL) {
491 got_proto_09:
492 request->hr_proto = httpd->consts.http_09;
493 debug((httpd, DEBUG_FAT, "request %s is http/0.9",
494 request->hr_file));
495 return 0;
496 }
497
498 if (strncasecmp(proto, "HTTP/", 5) != 0)
499 goto bad;
500 strncpy(majorstr, proto + 5, sizeof(majorstr)-1);
501 majorstr[sizeof(majorstr)-1] = 0;
502 minorstr = strchr(majorstr, '.');
503 if (minorstr == NULL)
504 goto bad;
505 *minorstr++ = 0;
506
507 majorint = atoi(majorstr);
508 minorint = atoi(minorstr);
509
510 switch (majorint) {
511 case 0:
512 if (minorint != 9)
513 break;
514 goto got_proto_09;
515 case 1:
516 if (minorint == 0)
517 request->hr_proto = httpd->consts.http_10;
518 else if (minorint == 1)
519 request->hr_proto = httpd->consts.http_11;
520 else
521 break;
522
523 debug((httpd, DEBUG_FAT, "request %s is %s",
524 request->hr_file, request->hr_proto));
525 SIMPLEQ_INIT(&request->hr_headers);
526 request->hr_nheaders = 0;
527 return 0;
528 }
529 bad:
530 return bozo_http_error(httpd, 404, NULL, "unknown prototype");
531 }
532
533 /*
534 * process each type of HTTP method, setting this HTTP requests
535 * method type.
536 */
537 static struct method_map {
538 const char *name;
539 int type;
540 } method_map[] = {
541 { "GET", HTTP_GET, },
542 { "POST", HTTP_POST, },
543 { "HEAD", HTTP_HEAD, },
544 #if 0 /* other non-required http/1.1 methods */
545 { "OPTIONS", HTTP_OPTIONS, },
546 { "PUT", HTTP_PUT, },
547 { "DELETE", HTTP_DELETE, },
548 { "TRACE", HTTP_TRACE, },
549 { "CONNECT", HTTP_CONNECT, },
550 #endif
551 { NULL, 0, },
552 };
553
554 static int
555 process_method(bozo_httpreq_t *request, const char *method)
556 {
557 struct bozohttpd_t *httpd = request->hr_httpd;
558 struct method_map *mmp;
559
560 if (request->hr_proto == httpd->consts.http_11)
561 request->hr_allow = "GET, HEAD, POST";
562
563 for (mmp = method_map; mmp->name; mmp++)
564 if (strcasecmp(method, mmp->name) == 0) {
565 request->hr_method = mmp->type;
566 request->hr_methodstr = mmp->name;
567 return 0;
568 }
569
570 return bozo_http_error(httpd, 404, request, "unknown method");
571 }
572
573 /* check header byte count */
574 static int
575 bozo_got_header_length(bozo_httpreq_t *request, size_t len)
576 {
577
578 if (len > BOZO_HEADERS_MAX_SIZE - request->hr_header_bytes)
579 return bozo_http_error(request->hr_httpd, 413, request,
580 "too many headers");
581
582 request->hr_header_bytes += len;
583
584 return 0;
585 }
586
587 /*
588 * This function reads a http request from stdin, returning a pointer to a
589 * bozo_httpreq_t structure, describing the request.
590 */
591 bozo_httpreq_t *
592 bozo_read_request(bozohttpd_t *httpd)
593 {
594 struct sigaction sa;
595 char *str, *val, *method, *file, *proto, *query;
596 char *host, *addr, *port;
597 char bufport[10];
598 char hbuf[NI_MAXHOST], abuf[NI_MAXHOST];
599 struct sockaddr_storage ss;
600 ssize_t len;
601 int line = 0;
602 socklen_t slen;
603 bozo_httpreq_t *request;
604 struct timespec ots, ts;
605
606 /*
607 * if we're in daemon mode, bozo_daemon_fork() will return here twice
608 * for each call. once in the child, returning 0, and once in the
609 * parent, returning 1 for each child.
610 */
611 if (bozo_daemon_fork(httpd))
612 return NULL;
613
614 request = bozomalloc(httpd, sizeof(*request));
615 memset(request, 0, sizeof(*request));
616 request->hr_httpd = httpd;
617 request->hr_allow = request->hr_host = NULL;
618 request->hr_content_type = request->hr_content_length = NULL;
619 request->hr_range = NULL;
620 request->hr_last_byte_pos = -1;
621 request->hr_if_modified_since = NULL;
622 request->hr_virthostname = NULL;
623 request->hr_file_free = NULL;
624 request->hr_file = NULL;
625 request->hr_oldfile = NULL;
626 SIMPLEQ_INIT(&request->hr_replheaders);
627 bozo_auth_init(request);
628
629 slen = sizeof(ss);
630 if (getpeername(0, (struct sockaddr *)(void *)&ss, &slen) < 0)
631 host = addr = NULL;
632 else {
633 if (getnameinfo((struct sockaddr *)(void *)&ss, slen,
634 abuf, sizeof abuf, NULL, 0, NI_NUMERICHOST) == 0)
635 addr = abuf;
636 else
637 addr = NULL;
638 if (httpd->numeric == 0 &&
639 getnameinfo((struct sockaddr *)(void *)&ss, slen,
640 hbuf, sizeof hbuf, NULL, 0, 0) == 0)
641 host = hbuf;
642 else
643 host = NULL;
644 }
645 if (host != NULL)
646 request->hr_remotehost = bozostrdup(httpd, request, host);
647 if (addr != NULL)
648 request->hr_remoteaddr = bozostrdup(httpd, request, addr);
649 slen = sizeof(ss);
650
651 /*
652 * Override the bound port from the request value, so it works even
653 * if passed through a proxy that doesn't rewrite the port.
654 */
655 if (httpd->bindport) {
656 if (strcmp(httpd->bindport, BOZO_HTTP_PORT) != 0)
657 port = httpd->bindport;
658 else
659 port = NULL;
660 } else {
661 if (getsockname(0, (struct sockaddr *)(void *)&ss, &slen) < 0)
662 port = NULL;
663 else {
664 if (getnameinfo((struct sockaddr *)(void *)&ss, slen,
665 NULL, 0, bufport, sizeof bufport,
666 NI_NUMERICSERV) == 0)
667 port = bufport;
668 else
669 port = NULL;
670 }
671 }
672 if (port != NULL)
673 request->hr_serverport = bozostrdup(httpd, request, port);
674
675 /*
676 * setup a timer to make sure the request is not hung
677 */
678 sa.sa_handler = alarmer;
679 sigemptyset(&sa.sa_mask);
680 sigaddset(&sa.sa_mask, SIGALRM);
681 sa.sa_flags = 0;
682 sigaction(SIGALRM, &sa, NULL);
683
684 if (clock_gettime(CLOCK_MONOTONIC, &ots) != 0) {
685 bozo_http_error(httpd, 500, NULL, "clock_gettime failed");
686 goto cleanup;
687 }
688
689 /*
690 * now to try to setup SSL, and upon failure parent can signal the
691 * caller there was no request to process and it will wait for
692 * another.
693 */
694 if (bozo_ssl_accept(httpd))
695 return NULL;
696
697 alarm(httpd->initial_timeout);
698 while ((str = bozodgetln(httpd, STDIN_FILENO, &len, bozo_read)) != NULL) {
699 alarm(0);
700
701 if (clock_gettime(CLOCK_MONOTONIC, &ts) != 0) {
702 bozo_http_error(httpd, 500, NULL, "clock_gettime failed");
703 goto cleanup;
704 }
705 /*
706 * don't timeout if old tv_sec is not more than current
707 * tv_sec, or if current tv_sec is less than the request
708 * timeout (these shouldn't happen, but the first could
709 * if monotonic time is not available.)
710 *
711 * the other timeout and header size checks should ensure
712 * that even if time it set backwards or forwards a very
713 * long way, timeout will eventually happen, even if this
714 * one fails.
715 */
716 if (ts.tv_sec > ots.tv_sec &&
717 ts.tv_sec > httpd->request_timeout &&
718 ts.tv_sec - httpd->request_timeout > ots.tv_sec)
719 bozo_timeout_hit = 1;
720
721 if (bozo_timeout_hit) {
722 bozo_http_error(httpd, 408, NULL, "request timed out");
723 goto cleanup;
724 }
725 line++;
726
727 if (line == 1) {
728 if (len < 1) {
729 bozo_http_error(httpd, 404, NULL, "null method");
730 goto cleanup;
731 }
732 bozowarn(httpd,
733 "got request ``%s'' from host %s to port %s",
734 str,
735 host ? host : addr ? addr : "<local>",
736 port ? port : "<stdin>");
737
738 /* we allocate return space in file and query only */
739 parse_request(httpd, str, &method, &file, &query, &proto);
740 request->hr_file_free = request->hr_file = file;
741 request->hr_query = query;
742 if (method == NULL) {
743 bozo_http_error(httpd, 404, NULL, "null method");
744 goto cleanup;
745 }
746 if (file == NULL) {
747 bozo_http_error(httpd, 404, NULL, "null file");
748 goto cleanup;
749 }
750
751 /*
752 * note that we parse the proto first, so that we
753 * can more properly parse the method and the url.
754 */
755
756 if (process_proto(request, proto) ||
757 process_method(request, method)) {
758 goto cleanup;
759 }
760
761 debug((httpd, DEBUG_FAT, "got file \"%s\" query \"%s\"",
762 request->hr_file,
763 request->hr_query ? request->hr_query : "<none>"));
764
765 /* http/0.9 has no header processing */
766 if (request->hr_proto == httpd->consts.http_09)
767 break;
768 } else { /* incoming headers */
769 bozoheaders_t *hdr;
770
771 if (*str == '\0')
772 break;
773
774 val = bozostrnsep(&str, ":", &len);
775 debug((httpd, DEBUG_EXPLODING, "read_req2: after "
776 "bozostrnsep: str `%s' val `%s'",
777 str ? str : "<null>", val ? val : "<null>"));
778 if (val == NULL || len == -1) {
779 bozo_http_error(httpd, 404, request, "no header");
780 goto cleanup;
781 }
782 if (str == NULL) {
783 bozo_http_error(httpd, 404, request,
784 "malformed header");
785 goto cleanup;
786 }
787 while (*str == ' ' || *str == '\t')
788 len--, str++;
789 while (*val == ' ' || *val == '\t')
790 val++;
791
792 if (bozo_got_header_length(request, len))
793 goto cleanup;
794
795 if (bozo_auth_check_headers(request, val, str, len))
796 goto next_header;
797
798 hdr = addmerge_reqheader(request, val, str, len);
799
800 if (strcasecmp(hdr->h_header, "content-type") == 0)
801 request->hr_content_type = hdr->h_value;
802 else if (strcasecmp(hdr->h_header, "content-length") == 0)
803 request->hr_content_length = hdr->h_value;
804 else if (strcasecmp(hdr->h_header, "host") == 0) {
805 if (request->hr_host) {
806 /* RFC 7230 (HTTP/1.1): 5.4 */
807 bozo_http_error(httpd, 400, request,
808 "Only allow one Host: header");
809 goto cleanup;
810 }
811 request->hr_host = bozostrdup(httpd, request,
812 hdr->h_value);
813 }
814 /* RFC 2616 (HTTP/1.1): 14.20 */
815 else if (strcasecmp(hdr->h_header, "expect") == 0) {
816 bozo_http_error(httpd, 417, request,
817 "we don't support Expect:");
818 goto cleanup;
819 }
820 else if (strcasecmp(hdr->h_header, "referrer") == 0 ||
821 strcasecmp(hdr->h_header, "referer") == 0)
822 request->hr_referrer = hdr->h_value;
823 else if (strcasecmp(hdr->h_header, "range") == 0)
824 request->hr_range = hdr->h_value;
825 else if (strcasecmp(hdr->h_header,
826 "if-modified-since") == 0)
827 request->hr_if_modified_since = hdr->h_value;
828 else if (strcasecmp(hdr->h_header,
829 "accept-encoding") == 0)
830 request->hr_accept_encoding = hdr->h_value;
831
832 debug((httpd, DEBUG_FAT, "adding header %s: %s",
833 hdr->h_header, hdr->h_value));
834 }
835 next_header:
836 alarm(httpd->header_timeout);
837 }
838
839 /* now, clear it all out */
840 alarm(0);
841 signal(SIGALRM, SIG_DFL);
842
843 /* RFC1945, 8.3 */
844 if (request->hr_method == HTTP_POST &&
845 request->hr_content_length == NULL) {
846 bozo_http_error(httpd, 400, request, "missing content length");
847 goto cleanup;
848 }
849
850 /* RFC 2616 (HTTP/1.1), 14.23 & 19.6.1.1 */
851 if (request->hr_proto == httpd->consts.http_11 &&
852 /*(strncasecmp(request->hr_file, "http://", 7) != 0) &&*/
853 request->hr_host == NULL) {
854 bozo_http_error(httpd, 400, request, "missing Host header");
855 goto cleanup;
856 }
857
858 if (request->hr_range != NULL) {
859 debug((httpd, DEBUG_FAT, "hr_range: %s", request->hr_range));
860 /* support only simple ranges %d- and %d-%d */
861 if (strchr(request->hr_range, ',') == NULL) {
862 const char *rstart, *dash;
863
864 rstart = strchr(request->hr_range, '=');
865 if (rstart != NULL) {
866 rstart++;
867 dash = strchr(rstart, '-');
868 if (dash != NULL && dash != rstart) {
869 dash++;
870 request->hr_have_range = 1;
871 request->hr_first_byte_pos =
872 strtoll(rstart, NULL, 10);
873 if (request->hr_first_byte_pos < 0)
874 request->hr_first_byte_pos = 0;
875 if (*dash != '\0') {
876 request->hr_last_byte_pos =
877 strtoll(dash, NULL, 10);
878 if (request->hr_last_byte_pos < 0)
879 request->hr_last_byte_pos = -1;
880 }
881 }
882 }
883 }
884 }
885
886 debug((httpd, DEBUG_FAT, "bozo_read_request returns url %s in request",
887 request->hr_file));
888 return request;
889
890 cleanup:
891 bozo_clean_request(request);
892
893 return NULL;
894 }
895
896 static int
897 mmap_and_write_part(bozohttpd_t *httpd, int fd, off_t first_byte_pos, size_t sz)
898 {
899 size_t mappedsz, wroffset;
900 off_t mappedoffset;
901 char *addr;
902 void *mappedaddr;
903
904 /*
905 * we need to ensure that both the size *and* offset arguments to
906 * mmap() are page-aligned. our formala for this is:
907 *
908 * input offset: first_byte_pos
909 * input size: sz
910 *
911 * mapped offset = page align truncate (input offset)
912 * mapped size =
913 * page align extend (input offset - mapped offset + input size)
914 * write offset = input offset - mapped offset
915 *
916 * we use the write offset in all writes
917 */
918 mappedoffset = first_byte_pos & ~((off_t)httpd->page_size - 1);
919 mappedsz = (size_t)
920 (first_byte_pos - mappedoffset + sz + httpd->page_size - 1) &
921 ~(httpd->page_size - 1);
922 wroffset = (size_t)(first_byte_pos - mappedoffset);
923
924 addr = mmap(0, mappedsz, PROT_READ, MAP_SHARED, fd, mappedoffset);
925 if (addr == MAP_FAILED) {
926 bozowarn(httpd, "mmap failed: %s", strerror(errno));
927 return -1;
928 }
929 mappedaddr = addr;
930
931 #ifdef MADV_SEQUENTIAL
932 (void)madvise(addr, sz, MADV_SEQUENTIAL);
933 #endif
934 while (sz > BOZO_WRSZ) {
935 if (bozo_write(httpd, STDOUT_FILENO, addr + wroffset,
936 BOZO_WRSZ) != BOZO_WRSZ) {
937 bozowarn(httpd, "write failed: %s", strerror(errno));
938 goto out;
939 }
940 debug((httpd, DEBUG_OBESE, "wrote %d bytes", BOZO_WRSZ));
941 sz -= BOZO_WRSZ;
942 addr += BOZO_WRSZ;
943 }
944 if (sz && (size_t)bozo_write(httpd, STDOUT_FILENO, addr + wroffset,
945 sz) != sz) {
946 bozowarn(httpd, "final write failed: %s", strerror(errno));
947 goto out;
948 }
949 debug((httpd, DEBUG_OBESE, "wrote %d bytes", (int)sz));
950 out:
951 if (munmap(mappedaddr, mappedsz) < 0) {
952 bozowarn(httpd, "munmap failed");
953 return -1;
954 }
955
956 return 0;
957 }
958
959 static int
960 parse_http_date(const char *val, time_t *timestamp)
961 {
962 char *remainder;
963 struct tm tm;
964
965 if ((remainder = strptime(val, "%a, %d %b %Y %T GMT", &tm)) == NULL &&
966 (remainder = strptime(val, "%a, %d-%b-%y %T GMT", &tm)) == NULL &&
967 (remainder = strptime(val, "%a %b %d %T %Y", &tm)) == NULL)
968 return 0; /* Invalid HTTP date format */
969
970 if (*remainder)
971 return 0; /* No trailing garbage */
972
973 *timestamp = timegm(&tm);
974 return 1;
975 }
976
977 /*
978 * given an url, encode it ala rfc 3986. ie, escape ? and friends.
979 * note that this function returns a static buffer, and thus needs
980 * to be updated for any sort of parallel processing. escape only
981 * chosen characters for absolute redirects
982 */
983 char *
984 bozo_escape_rfc3986(bozohttpd_t *httpd, const char *url, int absolute)
985 {
986 static char *buf;
987 static size_t buflen = 0;
988 size_t len;
989 const char *s;
990 char *d;
991
992 len = strlen(url);
993 if (buflen < len * 3 + 1) {
994 buflen = len * 3 + 1;
995 buf = bozorealloc(httpd, buf, buflen);
996 }
997
998 for (s = url, d = buf; *s;) {
999 if (*s & 0x80)
1000 goto encode_it;
1001 switch (*s) {
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 case '+':
1016 case ',':
1017 case ';':
1018 case '=':
1019 case '%':
1020 case '"':
1021 if (absolute)
1022 goto leave_it;
1023 /*FALLTHROUGH*/
1024 case '\n':
1025 case '\r':
1026 case ' ':
1027 encode_it:
1028 snprintf(d, 4, "%%%02X", (unsigned char)*s++);
1029 d += 3;
1030 break;
1031 default:
1032 leave_it:
1033 *d++ = *s++;
1034 break;
1035 }
1036 }
1037 *d = 0;
1038
1039 return buf;
1040 }
1041
1042 /*
1043 * do automatic redirection -- if there are query parameters or userdir for
1044 * the URL we will tack these on to the new (redirected) URL.
1045 */
1046 static void
1047 handle_redirect(bozo_httpreq_t *request, const char *url, int absolute)
1048 {
1049 bozohttpd_t *httpd = request->hr_httpd;
1050 char *finalurl, *urlbuf;
1051 #ifndef NO_USER_SUPPORT
1052 char *userbuf;
1053 #endif /* !NO_USER_SUPPORT */
1054 char portbuf[20];
1055 const char *scheme, *query, *quest;
1056 const char *hostname = BOZOHOST(httpd, request);
1057 int absproto = 0; /* absolute redirect provides own schema */
1058
1059 if (url == NULL) {
1060 bozoasprintf(httpd, &urlbuf, "/%s/", request->hr_file);
1061 url = urlbuf;
1062 } else
1063 urlbuf = NULL;
1064
1065 #ifndef NO_USER_SUPPORT
1066 if (request->hr_user && !absolute) {
1067 bozoasprintf(httpd, &userbuf, "/~%s%s", request->hr_user, url);
1068 url = userbuf;
1069 } else
1070 userbuf = NULL;
1071 #endif /* !NO_USER_SUPPORT */
1072
1073 if (absolute) {
1074 char *sep = NULL;
1075 const char *s;
1076
1077 /*
1078 * absolute redirect may specify own protocol i.e. to redirect
1079 * to another schema like https:// or ftp://.
1080 * Details: RFC 3986, section 3.
1081 */
1082
1083 /* 1. check if url contains :// */
1084 sep = strstr(url, "://");
1085
1086 /*
1087 * RFC 3986, section 3.1:
1088 * scheme = ALPHA *( ALPHA / DIGIT / "+" / "-" / "." )
1089 */
1090 if (sep) {
1091 for (s = url; s != sep;) {
1092 if (!isalnum((int)*s) &&
1093 *s != '+' && *s != '-' && *s != '.')
1094 break;
1095 if (++s == sep) {
1096 absproto = 1;
1097 }
1098 }
1099 }
1100 }
1101
1102 /* construct final redirection url */
1103
1104 scheme = absproto ? "" : httpd->sslinfo ? "https://" : "http://";
1105
1106 if (absolute) {
1107 hostname = "";
1108 portbuf[0] = '\0';
1109 } else {
1110 const char *defport = httpd->sslinfo ? BOZO_HTTPS_PORT : BOZO_HTTP_PORT;
1111
1112 if (request->hr_serverport &&
1113 strcmp(request->hr_serverport, defport) != 0)
1114 snprintf(portbuf, sizeof(portbuf), ":%s",
1115 request->hr_serverport);
1116 else
1117 portbuf[0] = '\0';
1118 }
1119
1120 url = bozo_escape_rfc3986(httpd, url, absolute);
1121
1122 if (request->hr_query && strlen(request->hr_query)) {
1123 query = request->hr_query;
1124 quest = "?";
1125 } else {
1126 query = quest = "";
1127 }
1128
1129 bozoasprintf(httpd, &finalurl, "%s%s%s%s%s%s",
1130 scheme, hostname, portbuf, url, quest, query);
1131
1132 bozowarn(httpd, "redirecting %s", finalurl);
1133 debug((httpd, DEBUG_FAT, "redirecting %s", finalurl));
1134
1135 bozo_printf(httpd, "%s 301 Document Moved\r\n", request->hr_proto);
1136 if (request->hr_proto != httpd->consts.http_09)
1137 bozo_print_header(request, NULL, "text/html", NULL);
1138 if (request->hr_proto != httpd->consts.http_09)
1139 bozo_printf(httpd, "Location: %s\r\n", finalurl);
1140 bozo_printf(httpd, "\r\n");
1141 if (request->hr_method == HTTP_HEAD)
1142 goto head;
1143 bozo_printf(httpd, "<html><head><title>Document Moved</title></head>\n");
1144 bozo_printf(httpd, "<body><h1>Document Moved</h1>\n");
1145 bozo_printf(httpd, "This document had moved <a href=\"%s\">here</a>\n",
1146 finalurl);
1147 bozo_printf(httpd, "</body></html>\n");
1148 head:
1149 bozo_flush(httpd, stdout);
1150 free(urlbuf);
1151 free(finalurl);
1152 #ifndef NO_USER_SUPPORT
1153 free(userbuf);
1154 #endif /* !NO_USER_SUPPORT */
1155 }
1156
1157 /*
1158 * Like strncmp(), but s_esc may contain characters escaped by \.
1159 * The len argument does not include the backslashes used for escaping,
1160 * that is: it gives the raw len, after unescaping the string.
1161 */
1162 static int
1163 esccmp(const char *s_plain, const char *s_esc, size_t len)
1164 {
1165 bool esc = false;
1166
1167 while (len) {
1168 if (!esc && *s_esc == '\\') {
1169 esc = true;
1170 s_esc++;
1171 continue;
1172 }
1173 esc = false;
1174 if (*s_plain == 0 || *s_esc == 0 || *s_plain != *s_esc)
1175 return *s_esc - *s_plain;
1176 s_esc++;
1177 s_plain++;
1178 len--;
1179 }
1180 return 0;
1181 }
1182
1183 /*
1184 * Check if the request refers to a uri that is mapped via a .bzremap.
1185 * We have /requested/path:/re/mapped/to/this.html lines in there,
1186 * and the : separator may be use in the left hand side escaped with
1187 * \ to encode a path containig a : character.
1188 */
1189 static void
1190 check_remap(bozo_httpreq_t *request)
1191 {
1192 bozohttpd_t *httpd = request->hr_httpd;
1193 char *file = request->hr_file, *newfile;
1194 void *fmap;
1195 const char *replace = NULL, *map_to = NULL, *p;
1196 struct stat st;
1197 int mapfile;
1198 size_t avail, len, rlen, reqlen, num_esc = 0;
1199 bool escaped = false;
1200
1201 mapfile = open(REMAP_FILE, O_RDONLY, 0);
1202 if (mapfile == -1)
1203 return;
1204 debug((httpd, DEBUG_FAT, "remap file found"));
1205 if (fstat(mapfile, &st) == -1) {
1206 bozowarn(httpd, "could not stat " REMAP_FILE ", errno: %d",
1207 errno);
1208 goto out;
1209 }
1210
1211 fmap = mmap(NULL, st.st_size, PROT_READ, MAP_SHARED, mapfile, 0);
1212 if (fmap == MAP_FAILED) {
1213 bozowarn(httpd, "could not mmap " REMAP_FILE ", error %d",
1214 errno);
1215 goto out;
1216 }
1217 reqlen = strlen(file);
1218 for (p = fmap, avail = st.st_size; avail; ) {
1219 /*
1220 * We have lines like:
1221 * /this/url:/replacement/that/url
1222 * If we find a matching left hand side, replace will point
1223 * to it and len will be its length. map_to will point to
1224 * the right hand side and rlen wil be its length.
1225 * If we have no match, both pointers will be NULL.
1226 */
1227
1228 /* skip empty lines */
1229 while ((*p == '\r' || *p == '\n') && avail) {
1230 p++;
1231 avail--;
1232 }
1233 replace = p;
1234 escaped = false;
1235 while (avail) {
1236 if (*p == '\r' || *p == '\n')
1237 break;
1238 if (!escaped && *p == ':')
1239 break;
1240 if (escaped) {
1241 escaped = false;
1242 num_esc++;
1243 } else if (*p == '\\') {
1244 escaped = true;
1245 }
1246 p++;
1247 avail--;
1248 }
1249 if (!avail || *p != ':') {
1250 replace = NULL;
1251 map_to = NULL;
1252 break;
1253 }
1254 len = p - replace - num_esc;
1255 /*
1256 * reqlen < len: the left hand side is too long, can't be a
1257 * match
1258 * reqlen == len: full string has to match
1259 * reqlen > len: make sure there is a path separator at 'len'
1260 * avail < 2: we are at eof, missing right hand side
1261 */
1262 if (avail < 2 || reqlen < len ||
1263 (reqlen == len && esccmp(file, replace, len) != 0) ||
1264 (reqlen > len && (file[len] != '/' ||
1265 esccmp(file, replace, len) != 0))) {
1266
1267 /* non-match, skip to end of line and continue */
1268 while (*p != '\r' && *p != '\n' && avail) {
1269 p++;
1270 avail--;
1271 }
1272 replace = NULL;
1273 map_to = NULL;
1274 continue;
1275 }
1276 p++;
1277 avail--;
1278
1279 /* found a match, parse the target */
1280 map_to = p;
1281 while (*p != '\r' && *p != '\n' && avail) {
1282 p++;
1283 avail--;
1284 }
1285 rlen = p - map_to;
1286 break;
1287 }
1288
1289 if (replace && map_to) {
1290 newfile = bozomalloc(httpd, strlen(file) + rlen - len + 1);
1291 memcpy(newfile, map_to, rlen);
1292 strcpy(newfile+rlen, file + len);
1293 debug((httpd, DEBUG_NORMAL, "remapping found '%s'",
1294 newfile));
1295 free(request->hr_file_free);
1296 request->hr_file_free = request->hr_file = newfile;
1297 }
1298
1299 munmap(fmap, st.st_size);
1300 out:
1301 close(mapfile);
1302 }
1303
1304 /*
1305 * deal with virtual host names; we do this:
1306 * if we have a virtual path root (httpd->virtbase), and we are given a
1307 * virtual host spec (Host: ho.st or http://ho.st/), see if this
1308 * directory exists under httpd->virtbase. if it does, use this as the
1309 # new slashdir.
1310 */
1311 static int
1312 check_virtual(bozo_httpreq_t *request)
1313 {
1314 bozohttpd_t *httpd = request->hr_httpd;
1315 char *file = request->hr_file, *s;
1316 size_t len;
1317
1318 /*
1319 * convert http://virtual.host/ to request->hr_host
1320 */
1321 debug((httpd, DEBUG_OBESE,
1322 "checking for http:// virtual host in '%s'", file));
1323 if (strncasecmp(file, "http://", 7) == 0) {
1324 /* we would do virtual hosting here? */
1325 file += 7;
1326 /* RFC 2616 (HTTP/1.1), 5.2: URI takes precedence over Host: */
1327 free(request->hr_host);
1328 request->hr_host = bozostrdup(httpd, request, file);
1329 if ((s = strchr(request->hr_host, '/')) != NULL)
1330 *s = '\0';
1331 s = strchr(file, '/');
1332 free(request->hr_file_free);
1333 request->hr_file_free = request->hr_file =
1334 bozostrdup(httpd, request, s ? s : "/");
1335 debug((httpd, DEBUG_OBESE, "got host '%s' file is now '%s'",
1336 request->hr_host, request->hr_file));
1337 } else if (!request->hr_host)
1338 goto use_slashdir;
1339
1340 /*
1341 * canonicalise hr_host - that is, remove any :80.
1342 */
1343 len = strlen(request->hr_host);
1344 if (len > 3 &&
1345 strcmp(request->hr_host + len - 3, ":" BOZO_HTTP_PORT) == 0) {
1346 request->hr_host[len - 3] = '\0';
1347 len = strlen(request->hr_host);
1348 }
1349
1350 if (!httpd->virtbase) {
1351 /*
1352 * if we don't use vhost support, then set virthostname if
1353 * user supplied Host header. It will be used for possible
1354 * redirections
1355 */
1356 if (request->hr_host) {
1357 s = strrchr(request->hr_host, ':');
1358 if (s != NULL)
1359 /*
1360 * truncate Host: as we want to copy it
1361 * without port part
1362 */
1363 *s = '\0';
1364 request->hr_virthostname = bozostrdup(httpd, request,
1365 request->hr_host);
1366 if (s != NULL)
1367 /* fix Host: again, if we truncated it */
1368 *s = ':';
1369 }
1370 goto use_slashdir;
1371 }
1372
1373 /*
1374 * ok, we have a virtual host, use opendir(3) to find a case
1375 * insensitive match for the virtual host we are asked for.
1376 * note that if the virtual host is the same as the master,
1377 * we don't need to do anything special.
1378 */
1379 debug((httpd, DEBUG_OBESE,
1380 "check_virtual: checking host `%s' under httpd->virtbase `%s' "
1381 "for file `%s'",
1382 request->hr_host, httpd->virtbase, request->hr_file));
1383 if (strncasecmp(httpd->virthostname, request->hr_host, len) != 0) {
1384 s = NULL;
1385 DIR *dirp;
1386 struct dirent *d;
1387
1388 if ((dirp = opendir(httpd->virtbase)) != NULL) {
1389 while ((d = readdir(dirp)) != NULL) {
1390 if (strcmp(d->d_name, ".") == 0 ||
1391 strcmp(d->d_name, "..") == 0) {
1392 continue;
1393 }
1394 debug((httpd, DEBUG_OBESE, "looking at dir '%s'",
1395 d->d_name));
1396 if (strcmp(d->d_name, request->hr_host) == 0) {
1397 /* found it, punch it */
1398 debug((httpd, DEBUG_OBESE, "found it punch it"));
1399 request->hr_virthostname =
1400 bozostrdup(httpd, request, d->d_name);
1401 bozoasprintf(httpd, &s, "%s/%s",
1402 httpd->virtbase,
1403 request->hr_virthostname);
1404 break;
1405 }
1406 }
1407 closedir(dirp);
1408 }
1409 else {
1410 debug((httpd, DEBUG_FAT, "opendir %s failed: %s",
1411 httpd->virtbase, strerror(errno)));
1412 }
1413 if (s == 0) {
1414 if (httpd->unknown_slash)
1415 goto use_slashdir;
1416 return bozo_http_error(httpd, 404, request,
1417 "unknown URL");
1418 }
1419 } else
1420 use_slashdir:
1421 s = httpd->slashdir;
1422
1423 /*
1424 * ok, nailed the correct slashdir, chdir to it
1425 */
1426 if (chdir(s) < 0)
1427 return bozo_http_error(httpd, 404, request,
1428 "can't chdir to slashdir");
1429
1430 /*
1431 * is there a mapping for this request?
1432 */
1433 check_remap(request);
1434
1435 return 0;
1436 }
1437
1438 /*
1439 * checks to see if this request has a valid .bzredirect file. returns
1440 * 0 when no redirection happend, or 1 when handle_redirect() has been
1441 * called, -1 on error.
1442 */
1443 static int
1444 check_bzredirect(bozo_httpreq_t *request)
1445 {
1446 bozohttpd_t *httpd = request->hr_httpd;
1447 struct stat sb;
1448 char dir[MAXPATHLEN], redir[MAXPATHLEN], redirpath[MAXPATHLEN + 1],
1449 path[MAXPATHLEN + 1];
1450 char *basename, *finalredir;
1451 int rv, absolute;
1452
1453 /*
1454 * if this pathname is really a directory, but doesn't end in /,
1455 * use it as the directory to look for the redir file.
1456 */
1457 if ((size_t)snprintf(dir, sizeof(dir), "%s", request->hr_file + 1) >=
1458 sizeof(dir)) {
1459 bozo_http_error(httpd, 404, request, "file path too long");
1460 return -1;
1461 }
1462 debug((httpd, DEBUG_FAT, "check_bzredirect: dir %s", dir));
1463 basename = strrchr(dir, '/');
1464
1465 if ((!basename || basename[1] != '\0') &&
1466 lstat(dir, &sb) == 0 && S_ISDIR(sb.st_mode)) {
1467 strcpy(path, dir);
1468 basename = dir;
1469 } else if (basename == NULL) {
1470 strcpy(path, ".");
1471 strcpy(dir, "");
1472 basename = request->hr_file + 1;
1473 } else {
1474 *basename++ = '\0';
1475 strcpy(path, dir);
1476 }
1477 if (bozo_check_special_files(request, basename, true))
1478 return -1;
1479
1480 debug((httpd, DEBUG_FAT, "check_bzredirect: path %s", path));
1481
1482 if ((size_t)snprintf(redir, sizeof(redir), "%s/%s", path,
1483 REDIRECT_FILE) >= sizeof(redir)) {
1484 return bozo_http_error(httpd, 404, request,
1485 "redirectfile path too long");
1486 }
1487 if (lstat(redir, &sb) == 0) {
1488 if (!S_ISLNK(sb.st_mode))
1489 return 0;
1490 absolute = 0;
1491 } else {
1492 if ((size_t)snprintf(redir, sizeof(redir), "%s/%s", path,
1493 ABSREDIRECT_FILE) >= sizeof(redir)) {
1494 bozo_http_error(httpd, 404, request,
1495 "redirectfile path too long");
1496 return -1;
1497 }
1498 if (lstat(redir, &sb) < 0 || !S_ISLNK(sb.st_mode))
1499 return 0;
1500 absolute = 1;
1501 }
1502 debug((httpd, DEBUG_FAT, "check_bzredirect: calling readlink"));
1503 rv = readlink(redir, redirpath, sizeof redirpath - 1);
1504 if (rv == -1 || rv == 0) {
1505 debug((httpd, DEBUG_FAT, "readlink failed"));
1506 return 0;
1507 }
1508 redirpath[rv] = '\0';
1509 debug((httpd, DEBUG_FAT, "readlink returned \"%s\"", redirpath));
1510
1511 /* check if we need authentication */
1512 snprintf(path, sizeof(path), "%s/", dir);
1513 if (bozo_auth_check(request, path))
1514 return 1;
1515
1516 /* now we have the link pointer, redirect to the real place */
1517 if (!absolute && redirpath[0] != '/') {
1518 if ((size_t)snprintf(finalredir = redir, sizeof(redir), "%s%s/%s",
1519 (strlen(dir) > 0 ? "/" : ""), dir, redirpath) >= sizeof(redir)) {
1520 bozo_http_error(httpd, 404, request,
1521 "redirect path too long");
1522 return -1;
1523 }
1524 } else
1525 finalredir = redirpath;
1526
1527 debug((httpd, DEBUG_FAT, "check_bzredirect: new redir %s", finalredir));
1528 handle_redirect(request, finalredir, absolute);
1529 return 1;
1530 }
1531
1532 /* this fixes the %HH hack that RFC2396 requires. */
1533 int
1534 bozo_decode_url_percent(bozo_httpreq_t *request, char *str)
1535 {
1536 bozohttpd_t *httpd = request->hr_httpd;
1537 char *s, *t, buf[3];
1538 char *end; /* if end is not-zero, we don't translate beyond that */
1539
1540 end = str + strlen(str);
1541
1542 /* fast forward to the first % */
1543 if ((s = strchr(str, '%')) == NULL)
1544 return 0;
1545
1546 t = s;
1547 do {
1548 if (end && s >= end) {
1549 debug((httpd, DEBUG_EXPLODING,
1550 "fu_%%: past end, filling out.."));
1551 while (*s)
1552 *t++ = *s++;
1553 break;
1554 }
1555 debug((httpd, DEBUG_EXPLODING,
1556 "fu_%%: got s == %%, s[1]s[2] == %c%c",
1557 s[1], s[2]));
1558 if (s[1] == '\0' || s[2] == '\0')
1559 return bozo_http_error(httpd, 400, request,
1560 "percent hack missing two chars afterwards");
1561 if (s[1] == '0' && s[2] == '0')
1562 return bozo_http_error(httpd, 404, request,
1563 "percent hack was %00");
1564 if (s[1] == '2' && (s[2] == 'f' || s[2] == 'F'))
1565 return bozo_http_error(httpd, 404, request,
1566 "percent hack was %2f (/)");
1567
1568 buf[0] = *++s;
1569 buf[1] = *++s;
1570 buf[2] = '\0';
1571 s++;
1572 *t = (char)strtol(buf, NULL, 16);
1573 debug((httpd, DEBUG_EXPLODING,
1574 "fu_%%: strtol put '%02x' into *t", *t));
1575 if (*t++ == '\0')
1576 return bozo_http_error(httpd, 400, request,
1577 "percent hack got a 0 back");
1578
1579 while (*s && *s != '%') {
1580 if (end && s >= end)
1581 break;
1582 *t++ = *s++;
1583 }
1584 } while (*s);
1585 *t = '\0';
1586
1587 debug((httpd, DEBUG_FAT, "bozo_decode_url_percent returns `%s'",
1588 request->hr_file));
1589
1590 return 0;
1591 }
1592
1593 /*
1594 * transform_request does this:
1595 * - ``expand'' %20 crapola
1596 * - punt if it doesn't start with /
1597 * - look for "http://myname/" and deal with it.
1598 * - maybe call bozo_process_cgi()
1599 * - check for ~user and call bozo_user_transform() if so
1600 * - if the length > 1, check for trailing slash. if so,
1601 * add the index.html file
1602 * - if the length is 1, return the index.html file
1603 * - disallow anything ending up with a file starting
1604 * at "/" or having ".." in it.
1605 * - anything else is a really weird internal error
1606 * - returns malloced file to serve, if unhandled
1607 */
1608 static int
1609 transform_request(bozo_httpreq_t *request, int *isindex)
1610 {
1611 bozohttpd_t *httpd = request->hr_httpd;
1612 char *file, *newfile = NULL;
1613 size_t len;
1614
1615 file = NULL;
1616 *isindex = 0;
1617 debug((httpd, DEBUG_FAT, "tf_req: file %s", request->hr_file));
1618
1619 if (bozo_decode_url_percent(request, request->hr_file) ||
1620 check_virtual(request))
1621 goto bad_done;
1622
1623 file = request->hr_file;
1624
1625 if (file[0] != '/') {
1626 bozo_http_error(httpd, 404, request, "unknown URL");
1627 goto bad_done;
1628 }
1629
1630 /* omit additional slashes at the beginning */
1631 while (file[1] == '/')
1632 file++;
1633
1634 /* fix file provided by user as it's used in other handlers */
1635 request->hr_file = file;
1636
1637 len = strlen(file);
1638
1639 #ifndef NO_USER_SUPPORT
1640 /* first of all expand user path */
1641 if (len > 1 && httpd->enable_users && file[1] == '~') {
1642 if (file[2] == '\0') {
1643 bozo_http_error(httpd, 404, request,
1644 "missing username");
1645 goto bad_done;
1646 }
1647 if (strchr(file + 2, '/') == NULL) {
1648 char *userredirecturl;
1649
1650 bozoasprintf(httpd, &userredirecturl, "%s/", file);
1651 handle_redirect(request, userredirecturl, 0);
1652 free(userredirecturl);
1653 return 0;
1654 }
1655 debug((httpd, DEBUG_FAT, "calling bozo_user_transform"));
1656
1657 if (!bozo_user_transform(request))
1658 return 0;
1659
1660 file = request->hr_file;
1661 len = strlen(file);
1662 }
1663 #endif /* NO_USER_SUPPORT */
1664
1665
1666 switch (check_bzredirect(request)) {
1667 case -1:
1668 goto bad_done;
1669 case 0:
1670 break;
1671 default:
1672 return 0;
1673 }
1674
1675 if (len > 1) {
1676 debug((httpd, DEBUG_FAT, "file[len-1] == %c", file[len-1]));
1677 if (file[len-1] == '/') { /* append index.html */
1678 *isindex = 1;
1679 debug((httpd, DEBUG_FAT, "appending index.html"));
1680 newfile = bozomalloc(httpd,
1681 len + strlen(httpd->index_html) + 1);
1682 strcpy(newfile, file + 1);
1683 strcat(newfile, httpd->index_html);
1684 } else
1685 newfile = bozostrdup(httpd, request, file + 1);
1686 } else if (len == 1) {
1687 debug((httpd, DEBUG_EXPLODING, "tf_req: len == 1"));
1688 newfile = bozostrdup(httpd, request, httpd->index_html);
1689 *isindex = 1;
1690 } else { /* len == 0 ? */
1691 bozo_http_error(httpd, 500, request, "request->hr_file is nul");
1692 goto bad_done;
1693 }
1694
1695 if (newfile == NULL) {
1696 bozo_http_error(httpd, 500, request, "internal failure");
1697 goto bad_done;
1698 }
1699
1700 /*
1701 * stop traversing outside our domain
1702 *
1703 * XXX true security only comes from our parent using chroot(2)
1704 * before execve(2)'ing us. or our own built in chroot(2) support.
1705 */
1706
1707 debug((httpd, DEBUG_FAT, "newfile: %s", newfile));
1708
1709 if (*newfile == '/' || strcmp(newfile, "..") == 0 ||
1710 strstr(newfile, "/..") || strstr(newfile, "../")) {
1711 bozo_http_error(httpd, 403, request, "illegal request");
1712 goto bad_done;
1713 }
1714
1715 if (bozo_auth_check(request, newfile))
1716 goto bad_done;
1717
1718 if (strlen(newfile)) {
1719 request->hr_oldfile = request->hr_file_free;
1720 request->hr_file = newfile;
1721 }
1722
1723 if (bozo_process_cgi(request) ||
1724 bozo_process_lua(request))
1725 return 0;
1726
1727 debug((httpd, DEBUG_FAT, "transform_request set: %s", newfile));
1728 return 1;
1729
1730 bad_done:
1731 debug((httpd, DEBUG_FAT, "transform_request returning: 0"));
1732 free(newfile);
1733 return 0;
1734 }
1735
1736 /*
1737 * can_gzip checks if the request supports and prefers gzip encoding.
1738 *
1739 * XXX: we do not consider the associated q with gzip in making our
1740 * decision which is broken.
1741 */
1742
1743 static int
1744 can_gzip(bozo_httpreq_t *request)
1745 {
1746 const char *pos;
1747 const char *tmp;
1748 size_t len;
1749
1750 /* First we decide if the request can be gzipped at all. */
1751
1752 /* not if we already are encoded... */
1753 tmp = bozo_content_encoding(request, request->hr_file);
1754 if (tmp && *tmp)
1755 return 0;
1756
1757 /* not if we are not asking for the whole file... */
1758 if (request->hr_last_byte_pos != -1 || request->hr_have_range)
1759 return 0;
1760
1761 /* Then we determine if gzip is on the cards. */
1762
1763 for (pos = request->hr_accept_encoding; pos && *pos; pos += len) {
1764 while (*pos == ' ')
1765 pos++;
1766
1767 len = strcspn(pos, ";,");
1768
1769 if ((len == 4 && strncasecmp("gzip", pos, 4) == 0) ||
1770 (len == 6 && strncasecmp("x-gzip", pos, 6) == 0))
1771 return 1;
1772
1773 if (pos[len] == ';')
1774 len += strcspn(&pos[len], ",");
1775
1776 if (pos[len])
1777 len++;
1778 }
1779
1780 return 0;
1781 }
1782
1783 /*
1784 * bozo_process_request does the following:
1785 * - check the request is valid
1786 * - process cgi-bin if necessary
1787 * - transform a filename if necesarry
1788 * - return the HTTP request
1789 */
1790 void
1791 bozo_process_request(bozo_httpreq_t *request)
1792 {
1793 bozohttpd_t *httpd = request->hr_httpd;
1794 struct stat sb;
1795 time_t timestamp;
1796 char *file;
1797 const char *type, *encoding;
1798 int fd, isindex;
1799
1800 /*
1801 * note that transform_request chdir()'s if required. also note
1802 * that cgi is handed here. if transform_request() returns 0
1803 * then the request has been handled already.
1804 */
1805 if (transform_request(request, &isindex) == 0)
1806 return;
1807
1808 fd = -1;
1809 encoding = NULL;
1810 if (can_gzip(request)) {
1811 bozoasprintf(httpd, &file, "%s.gz", request->hr_file);
1812 fd = open(file, O_RDONLY);
1813 if (fd >= 0)
1814 encoding = "gzip";
1815 free(file);
1816 }
1817
1818 file = request->hr_file;
1819
1820 if (fd < 0)
1821 fd = open(file, O_RDONLY);
1822
1823 if (fd < 0) {
1824 debug((httpd, DEBUG_FAT, "open failed: %s", strerror(errno)));
1825 switch (errno) {
1826 case EPERM:
1827 case EACCES:
1828 bozo_http_error(httpd, 403, request,
1829 "no permission to open file");
1830 break;
1831 case ENAMETOOLONG:
1832 /*FALLTHROUGH*/
1833 case ENOENT:
1834 if (!bozo_dir_index(request, file, isindex))
1835 bozo_http_error(httpd, 404, request, "no file");
1836 break;
1837 default:
1838 bozo_http_error(httpd, 500, request, "open file");
1839 }
1840 goto cleanup_nofd;
1841 }
1842 if (fstat(fd, &sb) < 0) {
1843 bozo_http_error(httpd, 500, request, "can't fstat");
1844 goto cleanup;
1845 }
1846 if (S_ISDIR(sb.st_mode)) {
1847 handle_redirect(request, NULL, 0);
1848 goto cleanup;
1849 }
1850
1851 if (request->hr_if_modified_since &&
1852 parse_http_date(request->hr_if_modified_since, ×tamp) &&
1853 timestamp >= sb.st_mtime) {
1854 /* XXX ignore subsecond of timestamp */
1855 bozo_printf(httpd, "%s 304 Not Modified\r\n",
1856 request->hr_proto);
1857 bozo_printf(httpd, "\r\n");
1858 bozo_flush(httpd, stdout);
1859 goto cleanup;
1860 }
1861
1862 /* validate requested range */
1863 if (request->hr_last_byte_pos == -1 ||
1864 request->hr_last_byte_pos >= sb.st_size)
1865 request->hr_last_byte_pos = sb.st_size - 1;
1866 if (request->hr_have_range &&
1867 request->hr_first_byte_pos > request->hr_last_byte_pos) {
1868 request->hr_have_range = 0; /* punt */
1869 request->hr_first_byte_pos = 0;
1870 request->hr_last_byte_pos = sb.st_size - 1;
1871 }
1872 debug((httpd, DEBUG_FAT, "have_range %d first_pos %lld last_pos %lld",
1873 request->hr_have_range,
1874 (long long)request->hr_first_byte_pos,
1875 (long long)request->hr_last_byte_pos));
1876 if (request->hr_have_range)
1877 bozo_printf(httpd, "%s 206 Partial Content\r\n",
1878 request->hr_proto);
1879 else
1880 bozo_printf(httpd, "%s 200 OK\r\n", request->hr_proto);
1881
1882 if (request->hr_proto != httpd->consts.http_09) {
1883 type = bozo_content_type(request, file);
1884 if (!encoding)
1885 encoding = bozo_content_encoding(request, file);
1886
1887 bozo_print_header(request, &sb, type, encoding);
1888 bozo_printf(httpd, "\r\n");
1889 }
1890 bozo_flush(httpd, stdout);
1891
1892 if (request->hr_method != HTTP_HEAD) {
1893 off_t szleft, cur_byte_pos;
1894
1895 szleft =
1896 request->hr_last_byte_pos - request->hr_first_byte_pos + 1;
1897 cur_byte_pos = request->hr_first_byte_pos;
1898
1899 retry:
1900 while (szleft) {
1901 size_t sz;
1902
1903 if ((off_t)httpd->mmapsz < szleft)
1904 sz = httpd->mmapsz;
1905 else
1906 sz = (size_t)szleft;
1907 if (mmap_and_write_part(httpd, fd, cur_byte_pos, sz)) {
1908 if (errno == ENOMEM) {
1909 httpd->mmapsz /= 2;
1910 if (httpd->mmapsz >= httpd->page_size)
1911 goto retry;
1912 }
1913 goto cleanup;
1914 }
1915 cur_byte_pos += sz;
1916 szleft -= sz;
1917 }
1918 }
1919 cleanup:
1920 close(fd);
1921 cleanup_nofd:
1922 /* If SSL enabled send close_notify. */
1923 bozo_ssl_shutdown(request->hr_httpd);
1924 close(STDIN_FILENO);
1925 close(STDOUT_FILENO);
1926 /*close(STDERR_FILENO);*/
1927 }
1928
1929 /* make sure we're not trying to access special files */
1930 int
1931 bozo_check_special_files(bozo_httpreq_t *request, const char *name, bool doerror)
1932 {
1933 bozohttpd_t *httpd = request->hr_httpd;
1934 size_t i;
1935 int error = 0;
1936
1937 for (i = 0; specials[i].file; i++) {
1938 if (strcmp(name, specials[i].file) == 0) {
1939 if (doerror) {
1940 error = bozo_http_error(httpd, 403, request,
1941 specials[i].name);
1942 } else {
1943 error = -1;
1944 }
1945 }
1946 }
1947
1948 return error;
1949 }
1950
1951 /* generic header printing routine */
1952 void
1953 bozo_print_header(bozo_httpreq_t *request,
1954 struct stat *sbp, const char *type, const char *encoding)
1955 {
1956 bozohttpd_t *httpd = request->hr_httpd;
1957 off_t len;
1958 char date[40];
1959 bozoheaders_t *hdr;
1960
1961 SIMPLEQ_FOREACH(hdr, &request->hr_replheaders, h_next) {
1962 bozo_printf(httpd, "%s: %s\r\n", hdr->h_header,
1963 hdr->h_value);
1964 }
1965
1966 bozo_printf(httpd, "Date: %s\r\n", bozo_http_date(date, sizeof(date)));
1967 bozo_printf(httpd, "Server: %s\r\n", httpd->server_software);
1968 bozo_printf(httpd, "Accept-Ranges: bytes\r\n");
1969 if (sbp) {
1970 char filedate[40];
1971 struct tm *tm;
1972
1973 tm = gmtime(&sbp->st_mtime);
1974 strftime(filedate, sizeof filedate,
1975 "%a, %d %b %Y %H:%M:%S GMT", tm);
1976 bozo_printf(httpd, "Last-Modified: %s\r\n", filedate);
1977 }
1978 if (type && *type)
1979 bozo_printf(httpd, "Content-Type: %s\r\n", type);
1980 if (encoding && *encoding)
1981 bozo_printf(httpd, "Content-Encoding: %s\r\n", encoding);
1982 if (sbp) {
1983 if (request->hr_have_range) {
1984 len = request->hr_last_byte_pos -
1985 request->hr_first_byte_pos +1;
1986 bozo_printf(httpd,
1987 "Content-Range: bytes %qd-%qd/%qd\r\n",
1988 (long long) request->hr_first_byte_pos,
1989 (long long) request->hr_last_byte_pos,
1990 (long long) sbp->st_size);
1991 } else
1992 len = sbp->st_size;
1993 bozo_printf(httpd, "Content-Length: %qd\r\n", (long long)len);
1994 }
1995 if (request->hr_proto == httpd->consts.http_11)
1996 bozo_printf(httpd, "Connection: close\r\n");
1997 bozo_flush(httpd, stdout);
1998 }
1999
2000 #ifndef NO_DEBUG
2001 void
2002 debug__(bozohttpd_t *httpd, int level, const char *fmt, ...)
2003 {
2004 va_list ap;
2005 int savederrno;
2006
2007 /* only log if the level is low enough */
2008 if (httpd->debug < level)
2009 return;
2010
2011 savederrno = errno;
2012 va_start(ap, fmt);
2013 if (httpd->logstderr) {
2014 vfprintf(stderr, fmt, ap);
2015 fputs("\n", stderr);
2016 } else
2017 vsyslog(LOG_DEBUG, fmt, ap);
2018 va_end(ap);
2019 errno = savederrno;
2020 }
2021 #endif /* NO_DEBUG */
2022
2023 /* these are like warn() and err(), except for syslog not stderr */
2024 void
2025 bozowarn(bozohttpd_t *httpd, const char *fmt, ...)
2026 {
2027 va_list ap;
2028
2029 va_start(ap, fmt);
2030 if (httpd->logstderr || isatty(STDERR_FILENO)) {
2031 //fputs("warning: ", stderr);
2032 vfprintf(stderr, fmt, ap);
2033 fputs("\n", stderr);
2034 } else
2035 vsyslog(LOG_INFO, fmt, ap);
2036 va_end(ap);
2037 }
2038
2039 void
2040 bozoerr(bozohttpd_t *httpd, int code, const char *fmt, ...)
2041 {
2042 va_list ap;
2043
2044 va_start(ap, fmt);
2045 if (httpd->logstderr || isatty(STDERR_FILENO)) {
2046 //fputs("error: ", stderr);
2047 vfprintf(stderr, fmt, ap);
2048 fputs("\n", stderr);
2049 } else
2050 vsyslog(LOG_ERR, fmt, ap);
2051 va_end(ap);
2052 exit(code);
2053 }
2054
2055 void
2056 bozoasprintf(bozohttpd_t *httpd, char **str, const char *fmt, ...)
2057 {
2058 va_list ap;
2059 int e;
2060
2061 va_start(ap, fmt);
2062 e = vasprintf(str, fmt, ap);
2063 va_end(ap);
2064
2065 if (e < 0)
2066 bozoerr(httpd, EXIT_FAILURE, "asprintf");
2067 }
2068
2069 /*
2070 * this escapes HTML tags. returns allocated escaped
2071 * string if needed, or NULL on allocation failure or
2072 * lack of escape need.
2073 * call with NULL httpd in error paths, to avoid recursive
2074 * malloc failure. call with valid httpd in normal paths
2075 * to get automatic allocation failure handling.
2076 */
2077 char *
2078 bozo_escape_html(bozohttpd_t *httpd, const char *url)
2079 {
2080 int i, j;
2081 char *tmp;
2082 size_t len;
2083
2084 for (i = 0, j = 0; url[i]; i++) {
2085 switch (url[i]) {
2086 case '<':
2087 case '>':
2088 j += 4;
2089 break;
2090 case '&':
2091 j += 5;
2092 break;
2093 case '"':
2094 j += 6;
2095 break;
2096 }
2097 }
2098
2099 if (j == 0)
2100 return NULL;
2101
2102 /*
2103 * we need to handle being called from different
2104 * pathnames.
2105 */
2106 len = strlen(url) + j;
2107 if (httpd)
2108 tmp = bozomalloc(httpd, len);
2109 else if ((tmp = malloc(len)) == 0)
2110 return NULL;
2111
2112 for (i = 0, j = 0; url[i]; i++) {
2113 switch (url[i]) {
2114 case '<':
2115 memcpy(tmp + j, "<", 4);
2116 j += 4;
2117 break;
2118 case '>':
2119 memcpy(tmp + j, ">", 4);
2120 j += 4;
2121 break;
2122 case '&':
2123 memcpy(tmp + j, "&", 5);
2124 j += 5;
2125 break;
2126 case '"':
2127 memcpy(tmp + j, """, 6);
2128 j += 6;
2129 break;
2130 default:
2131 tmp[j++] = url[i];
2132 }
2133 }
2134 tmp[j] = 0;
2135
2136 return tmp;
2137 }
2138
2139 /* short map between error code, and short/long messages */
2140 static struct errors_map {
2141 int code; /* HTTP return code */
2142 const char *shortmsg; /* short version of message */
2143 const char *longmsg; /* long version of message */
2144 } errors_map[] = {
2145 { 200, "200 OK", "The request was valid", },
2146 { 400, "400 Bad Request", "The request was not valid", },
2147 { 401, "401 Unauthorized", "No authorization", },
2148 { 403, "403 Forbidden", "Access to this item has been denied",},
2149 { 404, "404 Not Found", "This item has not been found", },
2150 { 408, "408 Request Timeout", "This request took too long", },
2151 { 413, "413 Payload Too Large", "Use smaller requests", },
2152 { 417, "417 Expectation Failed","Expectations not available", },
2153 { 420, "420 Enhance Your Calm","Chill, Winston", },
2154 { 500, "500 Internal Error", "An error occured on the server", },
2155 { 501, "501 Not Implemented", "This request is not available", },
2156 { 0, NULL, NULL, },
2157 };
2158
2159 static const char *help = "DANGER! WILL ROBINSON! DANGER!";
2160
2161 static const char *
2162 http_errors_short(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->shortmsg);
2169 return (help);
2170 }
2171
2172 static const char *
2173 http_errors_long(int code)
2174 {
2175 struct errors_map *ep;
2176
2177 for (ep = errors_map; ep->code; ep++)
2178 if (ep->code == code)
2179 return (ep->longmsg);
2180 return (help);
2181 }
2182
2183 #ifndef NO_BLOCKLIST_SUPPORT
2184 static struct blocklist *blstate;
2185
2186 void
2187 pfilter_notify(const int what, const int code)
2188 {
2189
2190 if (blstate == NULL)
2191 blstate = blocklist_open();
2192
2193 if (blstate == NULL)
2194 return;
2195
2196 (void)blocklist_r(blstate, what, 0, http_errors_short(code));
2197 }
2198 #endif /* !NO_BLOCKLIST_SUPPORT */
2199
2200 /* the follow functions and variables are used in handling HTTP errors */
2201 int
2202 bozo_http_error(bozohttpd_t *httpd, int code, bozo_httpreq_t *request,
2203 const char *msg)
2204 {
2205 char portbuf[20];
2206 const char *header = http_errors_short(code);
2207 const char *reason = http_errors_long(code);
2208 const char *proto = (request && request->hr_proto) ?
2209 request->hr_proto : httpd->consts.http_11;
2210 int size;
2211 bozoheaders_t *hdr;
2212
2213 USE_ARG(msg);
2214
2215 debug((httpd, DEBUG_FAT, "bozo_http_error %d: %s", code, msg));
2216 if (header == NULL || reason == NULL) {
2217 bozoerr(httpd, 1,
2218 "bozo_http_error() failed (short = %p, long = %p)",
2219 header, reason);
2220 return code;
2221 }
2222
2223 if (request && request->hr_serverport &&
2224 strcmp(request->hr_serverport, BOZO_HTTP_PORT) != 0)
2225 snprintf(portbuf, sizeof(portbuf), ":%s",
2226 request->hr_serverport);
2227 else
2228 portbuf[0] = '\0';
2229
2230 if (request && request->hr_file) {
2231 char *file = NULL, *user = NULL;
2232 int file_alloc = 0;
2233 const char *hostname = BOZOHOST(httpd, request);
2234
2235 /* bozo_escape_html() failure here is just too bad. */
2236 file = bozo_escape_html(NULL, request->hr_file);
2237 if (file == NULL)
2238 file = request->hr_file;
2239 else
2240 file_alloc = 1;
2241
2242 #ifndef NO_USER_SUPPORT
2243 if (request->hr_user != NULL) {
2244 char *user_escaped;
2245
2246 user_escaped = bozo_escape_html(NULL, request->hr_user);
2247 if (user_escaped == NULL)
2248 user_escaped = request->hr_user;
2249 /* expand username to ~user/ */
2250 bozoasprintf(httpd, &user, "~%s/", user_escaped);
2251 if (user_escaped != request->hr_user)
2252 free(user_escaped);
2253 }
2254 #endif /* !NO_USER_SUPPORT */
2255
2256 size = snprintf(httpd->errorbuf, BUFSIZ,
2257 "<html><head><title>%s</title></head>\n"
2258 "<body><h1>%s</h1>\n"
2259 "%s%s: <pre>%s</pre>\n"
2260 "<hr><address><a href=\"//%s%s/\">%s%s</a></address>\n"
2261 "</body></html>\n",
2262 header, header,
2263 user ? user : "", file,
2264 reason, hostname, portbuf, hostname, portbuf);
2265 free(user);
2266 if (size >= (int)BUFSIZ) {
2267 bozowarn(httpd,
2268 "bozo_http_error buffer too small, truncated");
2269 size = (int)BUFSIZ;
2270 }
2271
2272 if (file_alloc)
2273 free(file);
2274 } else
2275 size = 0;
2276
2277 bozo_printf(httpd, "%s %s\r\n", proto, header);
2278
2279 if (request) {
2280 bozo_auth_check_401(request, code);
2281 SIMPLEQ_FOREACH(hdr, &request->hr_replheaders, h_next) {
2282 bozo_printf(httpd, "%s: %s\r\n", hdr->h_header,
2283 hdr->h_value);
2284 }
2285 }
2286
2287 bozo_printf(httpd, "Content-Type: text/html\r\n");
2288 bozo_printf(httpd, "Content-Length: %d\r\n", size);
2289 bozo_printf(httpd, "Server: %s\r\n", httpd->server_software);
2290 if (request && request->hr_allow)
2291 bozo_printf(httpd, "Allow: %s\r\n", request->hr_allow);
2292 /* RFC 7231 (HTTP/1.1) 6.5.7 */
2293 if (code == 408 && request &&
2294 request->hr_proto == httpd->consts.http_11)
2295 bozo_printf(httpd, "Connection: close\r\n");
2296 bozo_printf(httpd, "\r\n");
2297 /* According to the RFC 2616 sec. 9.4 HEAD method MUST NOT return a
2298 * message-body in the response */
2299 if (size && request && request->hr_method != HTTP_HEAD)
2300 bozo_printf(httpd, "%s", httpd->errorbuf);
2301 bozo_flush(httpd, stdout);
2302
2303 #ifndef NO_BLOCKLIST_SUPPORT
2304 switch(code) {
2305
2306 case 401:
2307 pfilter_notify(BLOCKLIST_AUTH_FAIL, code);
2308 break;
2309
2310 case 403:
2311 pfilter_notify(BLOCKLIST_ABUSIVE_BEHAVIOR, code);
2312 break;
2313 }
2314 #endif /* !NO_BLOCKLIST_SUPPORT */
2315
2316 return code;
2317 }
2318
2319 /* Below are various modified libc functions */
2320
2321 /*
2322 * returns -1 in lenp if the string ran out before finding a delimiter,
2323 * but is otherwise the same as strsep. Note that the length must be
2324 * correctly passed in.
2325 */
2326 char *
2327 bozostrnsep(char **strp, const char *delim, ssize_t *lenp)
2328 {
2329 char *s;
2330 const char *spanp;
2331 int c, sc;
2332 char *tok;
2333
2334 if ((s = *strp) == NULL)
2335 return (NULL);
2336 for (tok = s;;) {
2337 if (lenp && --(*lenp) == -1)
2338 return (NULL);
2339 c = *s++;
2340 spanp = delim;
2341 do {
2342 if ((sc = *spanp++) == c) {
2343 if (c == 0)
2344 s = NULL;
2345 else
2346 s[-1] = '\0';
2347 *strp = s;
2348 return (tok);
2349 }
2350 } while (sc != 0);
2351 }
2352 /* NOTREACHED */
2353 }
2354
2355 /*
2356 * inspired by fgetln(3), but works for fd's. should work identically
2357 * except it, however, does *not* return the newline, and it does nul
2358 * terminate the string.
2359 */
2360 char *
2361 bozodgetln(bozohttpd_t *httpd, int fd, ssize_t *lenp,
2362 ssize_t (*readfn)(bozohttpd_t *, int, void *, size_t))
2363 {
2364 ssize_t len;
2365 int got_cr = 0;
2366 char c, *nbuffer;
2367
2368 /* initialise */
2369 if (httpd->getln_buflen == 0) {
2370 /* should be plenty for most requests */
2371 httpd->getln_buflen = 128;
2372 httpd->getln_buffer = malloc((size_t)httpd->getln_buflen);
2373 if (httpd->getln_buffer == NULL) {
2374 httpd->getln_buflen = 0;
2375 return NULL;
2376 }
2377 }
2378 len = 0;
2379
2380 /*
2381 * we *have* to read one byte at a time, to not break cgi
2382 * programs (for we pass stdin off to them). could fix this
2383 * by becoming a fd-passing program instead of just exec'ing
2384 * the program
2385 *
2386 * the above is no longer true, we are the fd-passing
2387 * program already.
2388 */
2389 for (; readfn(httpd, fd, &c, 1) == 1; ) {
2390 debug((httpd, DEBUG_EXPLODING, "bozodgetln read %c", c));
2391
2392 if (len >= httpd->getln_buflen - 1) {
2393 httpd->getln_buflen *= 2;
2394 debug((httpd, DEBUG_EXPLODING, "bozodgetln: "
2395 "reallocating buffer to buflen %zu",
2396 httpd->getln_buflen));
2397 nbuffer = bozorealloc(httpd, httpd->getln_buffer,
2398 (size_t)httpd->getln_buflen);
2399 httpd->getln_buffer = nbuffer;
2400 }
2401
2402 httpd->getln_buffer[len++] = c;
2403 if (c == '\r') {
2404 got_cr = 1;
2405 continue;
2406 } else if (c == '\n') {
2407 /*
2408 * HTTP/1.1 spec says to ignore CR and treat
2409 * LF as the real line terminator. even though
2410 * the same spec defines CRLF as the line
2411 * terminator, it is recommended in section 19.3
2412 * to do the LF trick for tolerance.
2413 */
2414 if (got_cr)
2415 len -= 2;
2416 else
2417 len -= 1;
2418 break;
2419 }
2420
2421 }
2422 httpd->getln_buffer[len] = '\0';
2423 debug((httpd, DEBUG_OBESE, "bozodgetln returns: '%s' with len %zd",
2424 httpd->getln_buffer, len));
2425 *lenp = len;
2426 return httpd->getln_buffer;
2427 }
2428
2429 /*
2430 * allocation frontends with error handling.
2431 *
2432 * note that these may access members of the httpd and/or request.
2433 */
2434 void *
2435 bozorealloc(bozohttpd_t *httpd, void *ptr, size_t size)
2436 {
2437 void *p;
2438
2439 p = realloc(ptr, size);
2440 if (p)
2441 return p;
2442
2443 bozo_http_error(httpd, 500, NULL, "memory allocation failure");
2444 exit(EXIT_FAILURE);
2445 }
2446
2447 void *
2448 bozomalloc(bozohttpd_t *httpd, size_t size)
2449 {
2450 void *p;
2451
2452 p = malloc(size);
2453 if (p)
2454 return p;
2455
2456 bozo_http_error(httpd, 500, NULL, "memory allocation failure");
2457 exit(EXIT_FAILURE);
2458 }
2459
2460 char *
2461 bozostrdup(bozohttpd_t *httpd, bozo_httpreq_t *request, const char *str)
2462 {
2463 char *p;
2464
2465 p = strdup(str);
2466 if (p)
2467 return p;
2468
2469 if (!request)
2470 bozoerr(httpd, EXIT_FAILURE, "strdup");
2471
2472 bozo_http_error(httpd, 500, request, "memory allocation failure");
2473 exit(EXIT_FAILURE);
2474 }
2475
2476 /* set default values in bozohttpd_t struct */
2477 int
2478 bozo_init_httpd(bozohttpd_t *httpd)
2479 {
2480 /* make sure everything is clean */
2481 (void) memset(httpd, 0x0, sizeof(*httpd));
2482
2483 /* constants */
2484 httpd->consts.http_09 = "HTTP/0.9";
2485 httpd->consts.http_10 = "HTTP/1.0";
2486 httpd->consts.http_11 = "HTTP/1.1";
2487 httpd->consts.text_plain = "text/plain";
2488
2489 /* mmap region size */
2490 httpd->mmapsz = BOZO_MMAPSZ;
2491
2492 /* error buffer for bozo_http_error() */
2493 if ((httpd->errorbuf = malloc(BUFSIZ)) == NULL) {
2494 fprintf(stderr,
2495 "bozohttpd: memory_allocation failure\n");
2496 return 0;
2497 }
2498 #ifndef NO_LUA_SUPPORT
2499 SIMPLEQ_INIT(&httpd->lua_states);
2500 #endif
2501 return 1;
2502 }
2503
2504 /* set default values in bozoprefs_t struct */
2505 int
2506 bozo_init_prefs(bozohttpd_t *httpd, bozoprefs_t *prefs)
2507 {
2508 int rv = 0;
2509
2510 /* make sure everything is clean */
2511 (void) memset(prefs, 0x0, sizeof(*prefs));
2512
2513 /* set up default values */
2514 if (!bozo_set_pref(httpd, prefs, "server software", SERVER_SOFTWARE))
2515 rv = 1;
2516 if (!bozo_set_pref(httpd, prefs, "index.html", INDEX_HTML))
2517 rv = 1;
2518 if (!bozo_set_pref(httpd, prefs, "public_html", PUBLIC_HTML))
2519 rv = 1;
2520 if (!bozo_set_pref(httpd, prefs, "ssl timeout", SSL_TIMEOUT))
2521 rv = 1;
2522 if (!bozo_set_pref(httpd, prefs, "initial timeout", INITIAL_TIMEOUT))
2523 rv = 1;
2524 if (!bozo_set_pref(httpd, prefs, "header timeout", HEADER_WAIT_TIME))
2525 rv = 1;
2526 if (!bozo_set_pref(httpd, prefs, "request timeout", TOTAL_MAX_REQ_TIME))
2527 rv = 1;
2528
2529 return rv;
2530 }
2531
2532 /* set default values */
2533 int
2534 bozo_set_defaults(bozohttpd_t *httpd, bozoprefs_t *prefs)
2535 {
2536 return bozo_init_httpd(httpd) && bozo_init_prefs(httpd, prefs);
2537 }
2538
2539 /* set the virtual host name, port and root */
2540 int
2541 bozo_setup(bozohttpd_t *httpd, bozoprefs_t *prefs, const char *vhost,
2542 const char *root)
2543 {
2544 struct passwd *pw;
2545 extern char **environ;
2546 static char *cleanenv[1] = { NULL };
2547 uid_t uid;
2548 int uidset = 0;
2549 char *chrootdir;
2550 char *username;
2551 char *portnum;
2552 char *cp;
2553 int dirtyenv;
2554
2555 dirtyenv = 0;
2556
2557 if (vhost == NULL) {
2558 httpd->virthostname = bozomalloc(httpd, MAXHOSTNAMELEN+1);
2559 if (gethostname(httpd->virthostname, MAXHOSTNAMELEN+1) < 0)
2560 bozoerr(httpd, 1, "gethostname");
2561 httpd->virthostname[MAXHOSTNAMELEN] = '\0';
2562 } else {
2563 httpd->virthostname = bozostrdup(httpd, NULL, vhost);
2564 }
2565 httpd->slashdir = bozostrdup(httpd, NULL, root);
2566 if ((portnum = bozo_get_pref(prefs, "port number")) != NULL) {
2567 httpd->bindport = bozostrdup(httpd, NULL, portnum);
2568 }
2569
2570 /* go over preferences now */
2571 if ((cp = bozo_get_pref(prefs, "numeric")) != NULL &&
2572 strcmp(cp, "true") == 0) {
2573 httpd->numeric = 1;
2574 }
2575 if ((cp = bozo_get_pref(prefs, "log to stderr")) != NULL &&
2576 strcmp(cp, "true") == 0) {
2577 httpd->logstderr = 1;
2578 }
2579 if ((cp = bozo_get_pref(prefs, "bind address")) != NULL) {
2580 httpd->bindaddress = bozostrdup(httpd, NULL, cp);
2581 }
2582 if ((cp = bozo_get_pref(prefs, "background")) != NULL) {
2583 httpd->background = atoi(cp);
2584 }
2585 if ((cp = bozo_get_pref(prefs, "foreground")) != NULL &&
2586 strcmp(cp, "true") == 0) {
2587 httpd->foreground = 1;
2588 }
2589 if ((cp = bozo_get_pref(prefs, "pid file")) != NULL) {
2590 httpd->pidfile = bozostrdup(httpd, NULL, cp);
2591 }
2592 if ((cp = bozo_get_pref(prefs, "unknown slash")) != NULL &&
2593 strcmp(cp, "true") == 0) {
2594 httpd->unknown_slash = 1;
2595 }
2596 if ((cp = bozo_get_pref(prefs, "virtual base")) != NULL) {
2597 httpd->virtbase = bozostrdup(httpd, NULL, cp);
2598 }
2599 if ((cp = bozo_get_pref(prefs, "enable users")) != NULL &&
2600 strcmp(cp, "true") == 0) {
2601 httpd->enable_users = 1;
2602 }
2603 if ((cp = bozo_get_pref(prefs, "enable user cgibin")) != NULL &&
2604 strcmp(cp, "true") == 0) {
2605 httpd->enable_cgi_users = 1;
2606 }
2607 if ((cp = bozo_get_pref(prefs, "dirty environment")) != NULL &&
2608 strcmp(cp, "true") == 0) {
2609 dirtyenv = 1;
2610 }
2611 if ((cp = bozo_get_pref(prefs, "hide dots")) != NULL &&
2612 strcmp(cp, "true") == 0) {
2613 httpd->hide_dots = 1;
2614 }
2615 if ((cp = bozo_get_pref(prefs, "directory indexing")) != NULL &&
2616 strcmp(cp, "true") == 0) {
2617 httpd->dir_indexing = 1;
2618 }
2619 if ((cp = bozo_get_pref(prefs, "directory index readme")) != NULL) {
2620 httpd->dir_readme = bozostrdup(httpd, NULL, cp);
2621 }
2622 if ((cp = bozo_get_pref(prefs, "public_html")) != NULL) {
2623 httpd->public_html = bozostrdup(httpd, NULL, cp);
2624 }
2625 if ((cp = bozo_get_pref(prefs, "ssl timeout")) != NULL) {
2626 httpd->ssl_timeout = atoi(cp);
2627 }
2628 if ((cp = bozo_get_pref(prefs, "initial timeout")) != NULL) {
2629 httpd->initial_timeout = atoi(cp);
2630 }
2631 if ((cp = bozo_get_pref(prefs, "header timeout")) != NULL) {
2632 httpd->header_timeout = atoi(cp);
2633 }
2634 if ((cp = bozo_get_pref(prefs, "request timeout")) != NULL) {
2635 httpd->request_timeout = atoi(cp);
2636 }
2637 httpd->server_software =
2638 bozostrdup(httpd, NULL, bozo_get_pref(prefs, "server software"));
2639 httpd->index_html =
2640 bozostrdup(httpd, NULL, bozo_get_pref(prefs, "index.html"));
2641
2642 /*
2643 * initialise ssl and daemon mode if necessary.
2644 */
2645 bozo_ssl_init(httpd);
2646 bozo_daemon_init(httpd);
2647
2648 username = bozo_get_pref(prefs, "username");
2649 if (username != NULL) {
2650 if ((pw = getpwnam(username)) == NULL)
2651 bozoerr(httpd, 1, "getpwnam(%s): %s", username,
2652 strerror(errno));
2653 if (initgroups(pw->pw_name, pw->pw_gid) == -1)
2654 bozoerr(httpd, 1, "initgroups: %s", strerror(errno));
2655 if (setgid(pw->pw_gid) == -1)
2656 bozoerr(httpd, 1, "setgid(%u): %s", pw->pw_gid,
2657 strerror(errno));
2658 uid = pw->pw_uid;
2659 uidset = 1;
2660 }
2661 /*
2662 * handle chroot.
2663 */
2664 if ((chrootdir = bozo_get_pref(prefs, "chroot dir")) != NULL) {
2665 httpd->rootdir = bozostrdup(httpd, NULL, chrootdir);
2666 if (chdir(httpd->rootdir) == -1)
2667 bozoerr(httpd, 1, "chdir(%s): %s", httpd->rootdir,
2668 strerror(errno));
2669 if (chroot(httpd->rootdir) == -1)
2670 bozoerr(httpd, 1, "chroot(%s): %s", httpd->rootdir,
2671 strerror(errno));
2672 }
2673
2674 if (uidset && setuid(uid) == -1)
2675 bozoerr(httpd, 1, "setuid(%d): %s", uid, strerror(errno));
2676
2677 /*
2678 * prevent info leakage between different compartments.
2679 * some PATH values in the environment would be invalided
2680 * by chroot. cross-user settings might result in undesirable
2681 * effects.
2682 */
2683 if ((chrootdir != NULL || username != NULL) && !dirtyenv)
2684 environ = cleanenv;
2685
2686 #ifdef _SC_PAGESIZE
2687 httpd->page_size = (long)sysconf(_SC_PAGESIZE);
2688 #else
2689 httpd->page_size = 4096;
2690 #endif
2691 debug((httpd, DEBUG_OBESE, "myname is %s, slashdir is %s",
2692 httpd->virthostname, httpd->slashdir));
2693
2694 return 1;
2695 }
2696
2697 int
2698 bozo_get_version(char *buf, size_t size)
2699 {
2700 return snprintf(buf, size, "%s", SERVER_SOFTWARE);
2701 }
2702