1 |
/* |
2 |
* Boa, an http server |
3 |
* Copyright (C) 1995 Paul Phillips <paulp@go2net.com> |
4 |
* Some changes Copyright (C) 1996,97 Larry Doolittle <ldoolitt@boa.org> |
5 |
* Some changes Copyright (C) 1996 Charles F. Randall <crandall@goldsys.com> |
6 |
* Some changes Copyright (C) 1997-2002 Jon Nelson <jnelson@boa.org> |
7 |
* |
8 |
* This program is free software; you can redistribute it and/or modify |
9 |
* it under the terms of the GNU General Public License as published by |
10 |
* the Free Software Foundation; either version 1, or (at your option) |
11 |
* any later version. |
12 |
* |
13 |
* This program is distributed in the hope that it will be useful, |
14 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
15 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
16 |
* GNU General Public License for more details. |
17 |
* |
18 |
* You should have received a copy of the GNU General Public License |
19 |
* along with this program; if not, write to the Free Software |
20 |
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
21 |
* |
22 |
*/ |
23 |
|
24 |
/* $Id: cgi.c,v 1.2 2002/09/22 09:07:57 nmav Exp $ */ |
25 |
|
26 |
#include "boa.h" |
27 |
|
28 |
static char *env_gen_extra(const char *key, const char *value, int extra); |
29 |
|
30 |
int verbose_cgi_logs = 0; |
31 |
/* The +1 is for the the NULL in complete_env */ |
32 |
static char *common_cgi_env[COMMON_CGI_COUNT + 1]; |
33 |
|
34 |
/* |
35 |
* Name: create_common_env |
36 |
* |
37 |
* Description: Set up the environment variables that are common to |
38 |
* all CGI scripts |
39 |
*/ |
40 |
|
41 |
void create_common_env() |
42 |
{ |
43 |
int index = 0, i; |
44 |
|
45 |
|
46 |
/* NOTE NOTE NOTE: |
47 |
If you (the reader) someday modify this chunk of code to |
48 |
handle more "common" CGI environment variables, then bump the |
49 |
value COMMON_CGI_COUNT in defines.h UP |
50 |
|
51 |
Also, in the case of document_root and server_admin, two variables |
52 |
that may or may not be defined depending on the way the server |
53 |
is configured, we check for null values and use an empty |
54 |
string to denote a NULL value to the environment, as per the |
55 |
specification. The quote for which follows: |
56 |
|
57 |
"In all cases, a missing environment variable is |
58 |
equivalent to a zero-length (NULL) value, and vice versa." |
59 |
*/ |
60 |
common_cgi_env[index++] = env_gen_extra("PATH", |
61 |
((cgi_path != NULL) ? cgi_path : DEFAULT_PATH), 0); |
62 |
common_cgi_env[index++] = env_gen_extra("SERVER_SOFTWARE", SERVER_VERSION, 0); |
63 |
common_cgi_env[index++] = env_gen_extra("SERVER_NAME", server_name, 0); |
64 |
common_cgi_env[index++] = env_gen_extra("GATEWAY_INTERFACE", CGI_VERSION, 0); |
65 |
|
66 |
/* removed the SERVER_PORT which may change due to SSL support |
67 |
*/ |
68 |
|
69 |
/* NCSA and APACHE added -- not in CGI spec */ |
70 |
/* common_cgi_env[index++] = env_gen_extra("DOCUMENT_ROOT", document_root); */ |
71 |
|
72 |
/* NCSA added */ |
73 |
/* common_cgi_env[index++] = env_gen_extra("SERVER_ROOT", server_root); */ |
74 |
|
75 |
/* APACHE added */ |
76 |
common_cgi_env[index++] = env_gen_extra("SERVER_ADMIN", server_admin, 0); |
77 |
common_cgi_env[index] = NULL; |
78 |
|
79 |
/* Sanity checking -- make *sure* the memory got allocated */ |
80 |
if (index > COMMON_CGI_COUNT) { |
81 |
log_error_time(); |
82 |
fprintf(stderr, "COMMON_CGI_COUNT not high enough.\n"); |
83 |
exit(1); |
84 |
} |
85 |
|
86 |
for(i = 0;i < index;++i) { |
87 |
if (common_cgi_env[i] == NULL) { |
88 |
log_error_time(); |
89 |
fprintf(stderr, "Unable to allocate a component of common_cgi_env - out of memory.\n"); |
90 |
exit(1); |
91 |
} |
92 |
} |
93 |
} |
94 |
|
95 |
void clear_common_env(void) |
96 |
{ |
97 |
int i; |
98 |
|
99 |
for(i = 0;i <= COMMON_CGI_COUNT;++i) { |
100 |
if (common_cgi_env[i] != NULL) { |
101 |
free(common_cgi_env[i]); |
102 |
common_cgi_env[i] = NULL; |
103 |
} |
104 |
} |
105 |
} |
106 |
|
107 |
/* |
108 |
* Name: env_gen_extra |
109 |
* (and via a not-so-tricky #define, env_gen) |
110 |
* This routine calls malloc: please free the memory when you are done |
111 |
*/ |
112 |
static char *env_gen_extra(const char *key, const char *value, int extra) |
113 |
{ |
114 |
char *result; |
115 |
int key_len, value_len; |
116 |
|
117 |
if (value == NULL) /* ServerAdmin may not be defined, eg */ |
118 |
value = ""; |
119 |
key_len = strlen(key); |
120 |
value_len = strlen(value); |
121 |
/* leave room for '=' sign and null terminator */ |
122 |
result = malloc(extra + key_len + value_len + 2); |
123 |
if (result) { |
124 |
memcpy(result + extra, key, key_len); |
125 |
*(result + extra + key_len) = '='; |
126 |
memcpy(result + extra + key_len + 1, value, value_len); |
127 |
*(result + extra + key_len + value_len + 1) = '\0'; |
128 |
} else { |
129 |
log_error_time(); |
130 |
perror("malloc"); |
131 |
log_error_time(); |
132 |
fprintf(stderr, |
133 |
"tried to allocate (key=value) extra=%d: %s=%s\n", |
134 |
extra, key, value); |
135 |
} |
136 |
return result; |
137 |
} |
138 |
|
139 |
/* |
140 |
* Name: add_cgi_env |
141 |
* |
142 |
* Description: adds a variable to CGI's environment |
143 |
* Used for HTTP_ headers |
144 |
*/ |
145 |
|
146 |
int add_cgi_env(request * req, char *key, char *value, int http_prefix) |
147 |
{ |
148 |
char *p; |
149 |
int prefix_len; |
150 |
|
151 |
if (http_prefix) { |
152 |
prefix_len = 5; |
153 |
} else { |
154 |
prefix_len = 0; |
155 |
} |
156 |
|
157 |
if (req->cgi_env_index < CGI_ENV_MAX) { |
158 |
p = env_gen_extra(key, value, prefix_len); |
159 |
if (!p) { |
160 |
log_error_time(); |
161 |
fprintf(stderr, "Unable to generate additional CGI Environment" |
162 |
"variable -- ran out of memory!\n"); |
163 |
} |
164 |
if (prefix_len) |
165 |
memcpy(p, "HTTP_", 5); |
166 |
req->cgi_env[req->cgi_env_index++] = p; |
167 |
return 1; |
168 |
} else { |
169 |
log_error_time(); |
170 |
fprintf(stderr, "Unable to generate additional CGI Environment" |
171 |
"variable -- not enough space!\n"); |
172 |
} |
173 |
return 0; |
174 |
} |
175 |
|
176 |
#define my_add_cgi_env(req, key, value) { \ |
177 |
int ok = add_cgi_env(req, key, value, 0); \ |
178 |
if (!ok) return 0; \ |
179 |
} |
180 |
|
181 |
/* |
182 |
* Name: complete_env |
183 |
* |
184 |
* Description: adds the known client header env variables |
185 |
* and terminates the environment array |
186 |
*/ |
187 |
|
188 |
int complete_env(request * req) |
189 |
{ |
190 |
int i; |
191 |
char buf[22]; |
192 |
|
193 |
for (i = 0; common_cgi_env[i]; i++) |
194 |
req->cgi_env[i] = common_cgi_env[i]; |
195 |
|
196 |
{ |
197 |
char *w; |
198 |
switch (req->method) { |
199 |
case M_POST: |
200 |
w = "POST"; |
201 |
break; |
202 |
case M_HEAD: |
203 |
w = "HEAD"; |
204 |
break; |
205 |
case M_GET: |
206 |
w = "GET"; |
207 |
break; |
208 |
default: |
209 |
w = "UNKNOWN"; |
210 |
break; |
211 |
} |
212 |
my_add_cgi_env(req, "REQUEST_METHOD", w); |
213 |
} |
214 |
|
215 |
if (req->secure) { |
216 |
simple_itoa(ssl_port, buf); |
217 |
} else { |
218 |
simple_itoa(server_port, buf); |
219 |
} |
220 |
my_add_cgi_env( req, "SERVER_PORT", buf); |
221 |
|
222 |
my_add_cgi_env(req, "SERVER_ADDR", req->local_ip_addr); |
223 |
my_add_cgi_env(req, "SERVER_PROTOCOL", req->http_version); |
224 |
my_add_cgi_env(req, "REQUEST_URI", req->request_uri); |
225 |
|
226 |
if (req->path_info) |
227 |
my_add_cgi_env(req, "PATH_INFO", req->path_info); |
228 |
|
229 |
if (req->path_translated) |
230 |
/* while path_translated depends on path_info, |
231 |
* there are cases when path_translated might |
232 |
* not exist when path_info does |
233 |
*/ |
234 |
my_add_cgi_env(req, "PATH_TRANSLATED", req->path_translated); |
235 |
|
236 |
my_add_cgi_env(req, "SCRIPT_NAME", req->script_name); |
237 |
|
238 |
if (req->query_string) |
239 |
my_add_cgi_env(req, "QUERY_STRING", req->query_string); |
240 |
my_add_cgi_env(req, "REMOTE_ADDR", req->remote_ip_addr); |
241 |
|
242 |
simple_itoa(req->remote_port, buf); |
243 |
my_add_cgi_env(req, "REMOTE_PORT", buf); |
244 |
|
245 |
if (req->method == M_POST) { |
246 |
if (req->content_type) { |
247 |
my_add_cgi_env(req, "CONTENT_TYPE", req->content_type); |
248 |
} else { |
249 |
my_add_cgi_env(req, "CONTENT_TYPE", default_type); |
250 |
} |
251 |
if (req->content_length) { |
252 |
my_add_cgi_env(req, "CONTENT_LENGTH", req->content_length); |
253 |
} |
254 |
} |
255 |
#ifdef ACCEPT_ON |
256 |
if (req->accept[0]) |
257 |
my_add_cgi_env(req, "HTTP_ACCEPT", req->accept); |
258 |
#endif |
259 |
|
260 |
if (req->cgi_env_index < CGI_ENV_MAX + 1) { |
261 |
req->cgi_env[req->cgi_env_index] = NULL; /* terminate */ |
262 |
return 1; |
263 |
} |
264 |
log_error_time(); |
265 |
fprintf(stderr, "Not enough space in CGI environment for remainder"\ |
266 |
" of variables.\n"); |
267 |
return 0; |
268 |
} |
269 |
|
270 |
/* |
271 |
* Name: make_args_cgi |
272 |
* |
273 |
* Build argv list for a CGI script according to spec |
274 |
* |
275 |
*/ |
276 |
|
277 |
void create_argv(request * req, char **aargv) |
278 |
{ |
279 |
char *p, *q, *r; |
280 |
int aargc; |
281 |
|
282 |
q = req->query_string; |
283 |
aargv[0] = req->pathname; |
284 |
|
285 |
/* here, we handle a special "indexed" query string. |
286 |
* Taken from the CGI/1.1 SPEC: |
287 |
* This is identified by a GET or HEAD request with a query string |
288 |
* with no *unencoded* '=' in it. |
289 |
* For such a request, I'm supposed to parse the search string |
290 |
* into words, according to the following rules: |
291 |
|
292 |
search-string = search-word *( "+" search-word ) |
293 |
search-word = 1*schar |
294 |
schar = xunreserved | escaped | xreserved |
295 |
xunreserved = alpha | digit | xsafe | extra |
296 |
xsafe = "$" | "-" | "_" | "." |
297 |
xreserved = ";" | "/" | "?" | ":" | "@" | "&" |
298 |
|
299 |
After parsing, each word is URL-decoded, optionally encoded in a system |
300 |
defined manner, and then the argument list |
301 |
is set to the list of words. |
302 |
|
303 |
|
304 |
Thus, schar is alpha|digit|"$"|"-"|"_"|"."|";"|"/"|"?"|":"|"@"|"&" |
305 |
|
306 |
As of this writing, escape.pl escapes the following chars: |
307 |
|
308 |
"-", "_", ".", "!", "~", "*", "'", "(", ")", |
309 |
"0".."9", "A".."Z", "a".."z", |
310 |
";", "/", "?", ":", "@", "&", "=", "+", "\$", "," |
311 |
|
312 |
Which therefore means |
313 |
"=", "+", "~", "!", "*", "'", "(", ")", "," |
314 |
are *not* escaped and should be? |
315 |
Wait, we don't do any escaping, and nor should we. |
316 |
According to the RFC draft, we unescape and then re-escape |
317 |
in a "system defined manner" (here: none). |
318 |
|
319 |
The CGI/1.1 draft (03, latest is 1999???) is very unclear here. |
320 |
|
321 |
I am using the latest published RFC, 2396, for what does and does |
322 |
not need escaping. |
323 |
|
324 |
Since boa builds the argument list and does not call /bin/sh, |
325 |
(boa uses execve for CGI) |
326 |
*/ |
327 |
|
328 |
if (q && !strchr(q, '=')) { |
329 |
/* we have an 'index' style */ |
330 |
q = strdup(q); |
331 |
if (!q) { |
332 |
WARN("unable to strdup 'q' in create_argv!"); |
333 |
} |
334 |
for (aargc = 1; q && (aargc < CGI_ARGC_MAX);) { |
335 |
r = q; |
336 |
/* for an index-style CGI, + is used to seperate arguments |
337 |
* an escaped '+' is of no concern to us |
338 |
*/ |
339 |
if ((p = strchr(q, '+'))) { |
340 |
*p = '\0'; |
341 |
q = p + 1; |
342 |
} else { |
343 |
q = NULL; |
344 |
} |
345 |
if (unescape_uri(r, NULL)) { |
346 |
/* printf("parameter %d: %s\n",aargc,r); */ |
347 |
aargv[aargc++] = r; |
348 |
} |
349 |
} |
350 |
aargv[aargc] = NULL; |
351 |
} else { |
352 |
aargv[1] = NULL; |
353 |
} |
354 |
} |
355 |
|
356 |
/* |
357 |
* Name: init_cgi |
358 |
* |
359 |
* Description: Called for GET/POST requests that refer to ScriptAlias |
360 |
* directories or application/x-httpd-cgi files. Ties stdout to socket, |
361 |
* stdin to data if POST, and execs CGI. |
362 |
* stderr remains tied to our log file; is this good? |
363 |
* |
364 |
* Returns: |
365 |
* 0 - error or NPH, either way the socket is closed |
366 |
* 1 - success |
367 |
*/ |
368 |
|
369 |
int init_cgi(request * req) |
370 |
{ |
371 |
int child_pid; |
372 |
int pipes[2]; |
373 |
int use_pipes = 0; |
374 |
|
375 |
SQUASH_KA(req); |
376 |
|
377 |
if (req->is_cgi==NPH || req->is_cgi==CGI) { |
378 |
if (complete_env(req) == 0) { |
379 |
return 0; |
380 |
} |
381 |
} |
382 |
#ifdef FASCIST_LOGGING |
383 |
{ |
384 |
int i; |
385 |
for (i = 0; i < req->cgi_env_index; ++i) |
386 |
fprintf(stderr, "%s - environment variable for cgi: \"%s\"\n", |
387 |
__FILE__, req->cgi_env[i]); |
388 |
} |
389 |
#endif |
390 |
|
391 |
if (req->is_cgi) { |
392 |
use_pipes = 1; |
393 |
if (pipe(pipes) == -1) { |
394 |
log_error_time(); |
395 |
perror("pipe"); |
396 |
return 0; |
397 |
} |
398 |
|
399 |
/* set the read end of the socket to non-blocking */ |
400 |
if (set_nonblock_fd(pipes[0]) == -1) { |
401 |
log_error_time(); |
402 |
perror("cgi-fcntl"); |
403 |
close(pipes[0]); |
404 |
close(pipes[1]); |
405 |
return 0; |
406 |
} |
407 |
} else { |
408 |
log_error_time(); |
409 |
fprintf(stderr, "Non CGI in init_cgi()!\n"); |
410 |
return 0; |
411 |
} |
412 |
|
413 |
child_pid = fork(); |
414 |
switch(child_pid) { |
415 |
case -1: |
416 |
/* fork unsuccessful */ |
417 |
log_error_time(); |
418 |
perror("fork"); |
419 |
|
420 |
if (use_pipes) { |
421 |
close(pipes[0]); |
422 |
close(pipes[1]); |
423 |
} |
424 |
send_r_error(req); |
425 |
/* FIXME: There is aproblem here. send_r_error would work |
426 |
for NPH and CGI, but not for GUNZIP. Fix that. */ |
427 |
/* i'd like to send_r_error, but.... */ |
428 |
return 0; |
429 |
break; |
430 |
case 0: |
431 |
/* child */ |
432 |
if (req->is_cgi == CGI || req->is_cgi == NPH) { |
433 |
char *foo = strdup(req->pathname); |
434 |
char *c; |
435 |
|
436 |
if (!foo) { |
437 |
WARN("unable to strdup pathname for req->pathname"); |
438 |
_exit(1); |
439 |
} |
440 |
c = strrchr(foo, '/'); |
441 |
if (c) { |
442 |
++c; |
443 |
*c = '\0'; |
444 |
} else { |
445 |
/* we have a serious problem */ |
446 |
log_error_time(); |
447 |
perror("chdir"); |
448 |
if (use_pipes) |
449 |
close(pipes[1]); |
450 |
_exit(1); |
451 |
} |
452 |
if (chdir(foo) != 0) { |
453 |
log_error_time(); |
454 |
perror("chdir"); |
455 |
if (use_pipes) |
456 |
close(pipes[1]); |
457 |
_exit(1); |
458 |
} |
459 |
} |
460 |
if (use_pipes) { |
461 |
close(pipes[0]); |
462 |
/* tie cgi's STDOUT to it's write end of pipe */ |
463 |
if (dup2(pipes[1], STDOUT_FILENO) == -1) { |
464 |
log_error_time(); |
465 |
perror("dup2 - pipes"); |
466 |
close(pipes[1]); |
467 |
_exit(1); |
468 |
} |
469 |
close(pipes[1]); |
470 |
if (set_block_fd(STDOUT_FILENO) == -1) { |
471 |
log_error_time(); |
472 |
perror("cgi-fcntl"); |
473 |
_exit(1); |
474 |
} |
475 |
} else { |
476 |
/* tie stdout to socket */ |
477 |
if (dup2(req->fd, STDOUT_FILENO) == -1) { |
478 |
log_error_time(); |
479 |
perror("dup2 - fd"); |
480 |
_exit(1); |
481 |
} |
482 |
/* Switch socket flags back to blocking */ |
483 |
if (set_block_fd(req->fd) == -1) { |
484 |
log_error_time(); |
485 |
perror("cgi-fcntl"); |
486 |
_exit(1); |
487 |
} |
488 |
} |
489 |
/* tie post_data_fd to POST stdin */ |
490 |
if (req->method == M_POST) { /* tie stdin to file */ |
491 |
lseek(req->post_data_fd, SEEK_SET, 0); |
492 |
dup2(req->post_data_fd, STDIN_FILENO); |
493 |
close(req->post_data_fd); |
494 |
} |
495 |
/* Close access log, so CGI program can't scribble |
496 |
* where it shouldn't |
497 |
*/ |
498 |
close_access_log(); |
499 |
|
500 |
/* |
501 |
* tie STDERR to cgi_log_fd |
502 |
* cgi_log_fd will automatically close, close-on-exec rocks! |
503 |
* if we don't tied STDERR (current log_error) to cgi_log_fd, |
504 |
* then we ought to close it. |
505 |
*/ |
506 |
if (!cgi_log_fd) |
507 |
dup2(devnullfd, STDERR_FILENO); |
508 |
else |
509 |
dup2(cgi_log_fd, STDERR_FILENO); |
510 |
|
511 |
if (req->is_cgi==NPH || req->is_cgi==CGI) { |
512 |
char *aargv[CGI_ARGC_MAX + 1]; |
513 |
create_argv(req, aargv); |
514 |
execve(req->pathname, aargv, req->cgi_env); |
515 |
} else { |
516 |
if (req->is_cgi==INDEXER_CGI) |
517 |
execl(dirmaker, dirmaker, req->pathname, req->request_uri, |
518 |
NULL); |
519 |
#ifdef GUNZIP |
520 |
else /* GUNZIP_CGI */ |
521 |
execl(GUNZIP, GUNZIP, "--stdout", "--decompress", |
522 |
req->pathname, NULL); |
523 |
#endif |
524 |
} |
525 |
/* execve failed */ |
526 |
WARN(req->pathname); |
527 |
_exit(1); |
528 |
break; |
529 |
|
530 |
default: |
531 |
/* parent */ |
532 |
/* if here, fork was successful */ |
533 |
if (verbose_cgi_logs) { |
534 |
log_error_time(); |
535 |
fprintf(stderr, "Forked child \"%s\" pid %d\n", |
536 |
req->pathname, child_pid); |
537 |
} |
538 |
|
539 |
if (req->method == M_POST) { |
540 |
close(req->post_data_fd); /* child closed it too */ |
541 |
req->post_data_fd = 0; |
542 |
} |
543 |
|
544 |
/* NPH, GUNZIP, etc... all go straight to the fd */ |
545 |
if (!use_pipes) |
546 |
return 0; |
547 |
|
548 |
close(pipes[1]); |
549 |
req->data_fd = pipes[0]; |
550 |
|
551 |
req->status = PIPE_READ; |
552 |
if (req->is_cgi == CGI) { |
553 |
req->cgi_status = CGI_PARSE; /* got to parse cgi header */ |
554 |
/* for cgi_header... I get half the buffer! */ |
555 |
req->header_line = req->header_end = |
556 |
(req->buffer + BUFFER_SIZE / 2); |
557 |
} else { |
558 |
req->cgi_status = CGI_BUFFER; |
559 |
/* I get all the buffer! */ |
560 |
req->header_line = req->header_end = req->buffer; |
561 |
} |
562 |
|
563 |
/* reset req->filepos for logging (it's used in pipe.c) */ |
564 |
/* still don't know why req->filesize might be reset though */ |
565 |
req->filepos = 0; |
566 |
break; |
567 |
} |
568 |
|
569 |
return 1; |
570 |
} |