1 |
/* |
2 |
* Boa, an http server |
3 |
* Copyright (C) 1995 Paul Phillips <paulp@go2net.com> |
4 |
* Some changes Copyright (C) 1996,99 Larry Doolittle <ldoolitt@boa.org> |
5 |
* Some changes Copyright (C) 1996-2002 Jon Nelson <jnelson@boa.org> |
6 |
* Portions Copyright (C) 2002 Nikos Mavroyanopoulos <nmav@gnutls.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: get.c,v 1.7 2002/09/23 19:28:41 nmav Exp $*/ |
25 |
|
26 |
#include "boa.h" |
27 |
#include "socket.h" |
28 |
|
29 |
/* local prototypes */ |
30 |
int get_cachedir_file(request * req, struct stat *statbuf); |
31 |
int index_directory(request * req, char *dest_filename); |
32 |
static int gzip_open(request * req); |
33 |
|
34 |
/* |
35 |
* Name: init_get |
36 |
* Description: Initializes a non-script GET or HEAD request. |
37 |
* |
38 |
* Return values: |
39 |
* 0: finished or error, request will be freed |
40 |
* 1: successfully initialized, added to ready queue |
41 |
*/ |
42 |
|
43 |
int init_get(server_params * params, request * req) |
44 |
{ |
45 |
int data_fd, saved_errno; |
46 |
struct stat statbuf; |
47 |
volatile int bytes; |
48 |
|
49 |
data_fd = open(req->pathname, O_RDONLY); |
50 |
saved_errno = errno; /* might not get used */ |
51 |
|
52 |
#ifdef GUNZIP |
53 |
if (data_fd == -1 && errno == ENOENT) { |
54 |
/* cannot open */ |
55 |
/* it's either a gunzipped file or a directory */ |
56 |
data_fd = gzip_open(req); |
57 |
|
58 |
/* since the gzip file exists then spawn the gunzip. |
59 |
*/ |
60 |
|
61 |
req->is_cgi = GUNZIP_CGI; |
62 |
if (data_fd==1) return init_cgi(req); |
63 |
|
64 |
/* If the return value was not -1 then return it. |
65 |
*/ |
66 |
if (data_fd!=-1) return data_fd; |
67 |
} |
68 |
#endif |
69 |
|
70 |
if (data_fd == -1) { |
71 |
log_error_doc(req); |
72 |
errno = saved_errno; |
73 |
perror("document open"); |
74 |
|
75 |
if (saved_errno == ENOENT) |
76 |
send_r_not_found(req); |
77 |
else if (saved_errno == EACCES) |
78 |
send_r_forbidden(req); |
79 |
else |
80 |
send_r_bad_request(req); |
81 |
return 0; |
82 |
} |
83 |
|
84 |
if (fstat( data_fd, &statbuf) == -1) { |
85 |
/* this is quite impossible, since the file |
86 |
* was opened before. |
87 |
*/ |
88 |
close(data_fd); |
89 |
send_r_not_found(req); |
90 |
return 0; |
91 |
} |
92 |
|
93 |
if (S_ISDIR(statbuf.st_mode)) { /* directory */ |
94 |
close(data_fd); /* close dir */ |
95 |
|
96 |
if (req->pathname[strlen(req->pathname) - 1] != '/') { |
97 |
char buffer[3 * MAX_PATH_LENGTH + 128]; |
98 |
char * proto; |
99 |
int port; |
100 |
|
101 |
if (req->secure) proto = "https"; |
102 |
else proto = "http"; |
103 |
|
104 |
port = params->server_s[req->secure].port; |
105 |
|
106 |
sprintf(buffer, "%s://%s:%d%s/", proto, req->hostname, |
107 |
port, req->request_uri); |
108 |
|
109 |
send_r_moved_perm(req, buffer); |
110 |
return 0; |
111 |
} |
112 |
data_fd = get_dir(req, &statbuf); /* updates statbuf */ |
113 |
|
114 |
if (data_fd == -1) /* couldn't do it */ |
115 |
return 0; /* errors reported by get_dir */ |
116 |
else if (data_fd <= 1) |
117 |
/* data_fd == 0 -> close it down, 1 -> continue */ |
118 |
return data_fd; |
119 |
/* else, data_fd contains the fd of the file... */ |
120 |
} |
121 |
|
122 |
if (req->if_modified_since && |
123 |
!modified_since(&(statbuf.st_mtime), req->if_modified_since)) { |
124 |
send_r_not_modified(req); |
125 |
close(data_fd); |
126 |
return 0; |
127 |
} |
128 |
req->filesize = statbuf.st_size; |
129 |
req->last_modified = statbuf.st_mtime; |
130 |
|
131 |
if (req->range_stop == 0) |
132 |
req->range_stop = statbuf.st_size; |
133 |
|
134 |
/* out of range! */ |
135 |
if (req->range_start > statbuf.st_size || |
136 |
req->range_stop > statbuf.st_size || |
137 |
req->range_stop < req->range_start) { |
138 |
send_r_range_unsatisfiable(req); |
139 |
close(data_fd); |
140 |
return 0; |
141 |
} |
142 |
|
143 |
if (req->method == M_HEAD || req->filesize == 0) { |
144 |
send_r_request_ok(req); |
145 |
close(data_fd); |
146 |
return 0; |
147 |
} |
148 |
|
149 |
if (req->range_stop > MAX_FILE_MMAP) { |
150 |
|
151 |
if ( req->range_start == 0 && req->range_stop == statbuf.st_size) |
152 |
send_r_request_ok(req); /* All's well */ |
153 |
else { |
154 |
/* if ranges were used, then lseek to the start given |
155 |
*/ |
156 |
if ( lseek( data_fd, req->range_start, SEEK_SET) == (off_t) -1) { |
157 |
close(data_fd); |
158 |
send_r_not_found(req); |
159 |
return 0; |
160 |
} |
161 |
send_r_request_partial(req);/* All's well */ |
162 |
} |
163 |
|
164 |
req->status = PIPE_READ; |
165 |
req->cgi_status = CGI_BUFFER; |
166 |
req->data_fd = data_fd; |
167 |
req_flush(req); /* this should *always* complete due to |
168 |
the size of the I/O buffers */ |
169 |
req->header_line = req->header_end = req->buffer; |
170 |
req->pipe_range_stop = req->range_stop; |
171 |
return 1; |
172 |
} |
173 |
|
174 |
if (req->range_stop == 0) { /* done */ |
175 |
send_r_request_ok(req); /* All's well *so far* */ |
176 |
close(data_fd); |
177 |
return 1; |
178 |
} |
179 |
|
180 |
/* NOTE: I (Jon Nelson) tried performing a read(2) |
181 |
* into the output buffer provided the file data would |
182 |
* fit, before mmapping, and if successful, writing that |
183 |
* and stopping there -- all to avoid the cost |
184 |
* of a mmap. Oddly, it was *slower* in benchmarks. |
185 |
*/ |
186 |
#ifdef USE_MMAP_LIST |
187 |
req->mmap_entry_var = find_mmap(data_fd, &statbuf); |
188 |
if (req->mmap_entry_var == NULL) { |
189 |
req->buffer_end = 0; |
190 |
if (errno == ENOENT) |
191 |
send_r_not_found(req); |
192 |
else if (errno == EACCES) |
193 |
send_r_forbidden(req); |
194 |
else |
195 |
send_r_bad_request(req); |
196 |
close(data_fd); |
197 |
return 0; |
198 |
} |
199 |
req->data_mem = req->mmap_entry_var->mmap; |
200 |
#else |
201 |
req->data_mem = mmap(0, req->range_stop, PROT_READ, MAP_OPTIONS, data_fd, 0); |
202 |
#endif |
203 |
close(data_fd); /* close data file */ |
204 |
|
205 |
if ( req->data_mem == MAP_FAILED) { |
206 |
boa_perror(req, "mmap"); |
207 |
return 0; |
208 |
} |
209 |
|
210 |
if ( req->range_start == 0 && req->range_stop == statbuf.st_size) |
211 |
send_r_request_ok(req); /* All's well */ |
212 |
else |
213 |
send_r_request_partial(req);/* All's well */ |
214 |
|
215 |
bytes = BUFFER_SIZE - req->buffer_end; |
216 |
|
217 |
/* bytes is now how much the buffer can hold |
218 |
* after the headers |
219 |
*/ |
220 |
req->filepos = req->range_start; |
221 |
|
222 |
if (bytes > 0) { |
223 |
if (bytes > req->range_stop - req->range_start) |
224 |
bytes = req->range_stop - req->range_start; |
225 |
|
226 |
if (setjmp(params->env) == 0) { |
227 |
params->handle_sigbus = 1; |
228 |
memcpy(req->buffer + req->buffer_end, &req->data_mem[req->filepos], |
229 |
bytes); |
230 |
params->handle_sigbus = 0; |
231 |
/* OK, SIGBUS **after** this point is very bad! */ |
232 |
} else { |
233 |
char buf[30]; |
234 |
/* sigbus! */ |
235 |
log_error_doc(req); |
236 |
reset_output_buffer(req); |
237 |
send_r_error(req); |
238 |
get_commonlog_time( buf); |
239 |
fprintf(stderr, "%sGot SIGBUS in memcpy!\n", |
240 |
buf); |
241 |
return 0; |
242 |
} |
243 |
req->buffer_end += bytes; |
244 |
req->filepos += bytes; |
245 |
if (req->range_stop == req->filepos) { |
246 |
req_flush(req); |
247 |
req->status = DONE; |
248 |
} |
249 |
} |
250 |
|
251 |
/* We lose statbuf here, so make sure response has been sent */ |
252 |
return 1; |
253 |
} |
254 |
|
255 |
#ifdef GUNZIP |
256 |
|
257 |
/* If a file could not be opened, then this function |
258 |
* is called, which tries to open the gziped file. |
259 |
* Return values: |
260 |
* 0: finished or error, request will be freed |
261 |
* 1: success |
262 |
* -1: failed |
263 |
*/ |
264 |
static int gzip_open(request * req) |
265 |
{ |
266 |
char gzip_pathname[MAX_PATH_LENGTH]; |
267 |
int len, data_fd; |
268 |
|
269 |
len = strlen(req->pathname); |
270 |
|
271 |
memcpy(gzip_pathname, req->pathname, len); |
272 |
memcpy(gzip_pathname + len, ".gz", 3); |
273 |
gzip_pathname[len + 3] = '\0'; |
274 |
data_fd = open(gzip_pathname, O_RDONLY); |
275 |
if (data_fd != -1) { |
276 |
close(data_fd); |
277 |
|
278 |
req->response_status = R_REQUEST_OK; |
279 |
if (req->pathname) |
280 |
free(req->pathname); |
281 |
req->pathname = strdup(gzip_pathname); |
282 |
if (!req->pathname) { |
283 |
log_error_time(); |
284 |
perror("strdup"); |
285 |
send_r_error(req); |
286 |
return 0; |
287 |
} |
288 |
if (!req->simple) { |
289 |
req_write(req, "HTTP/1.0 200 OK-GUNZIP\r\n"); |
290 |
print_http_headers(req); |
291 |
print_content_type(req); |
292 |
print_last_modified(req); |
293 |
req_write(req, "\r\n"); |
294 |
req_flush(req); |
295 |
} |
296 |
if (req->method == M_HEAD) |
297 |
return 0; |
298 |
|
299 |
return 1; |
300 |
|
301 |
} |
302 |
|
303 |
return -1; |
304 |
} |
305 |
#endif |
306 |
|
307 |
/* |
308 |
* Name: process_get |
309 |
* Description: Writes a chunk of data to the socket. |
310 |
* |
311 |
* Return values: |
312 |
* -1: request blocked, move to blocked queue |
313 |
* 0: EOF or error, close it down |
314 |
* 1: successful write, recycle in ready queue |
315 |
*/ |
316 |
|
317 |
int process_get(server_params * params, request * req) |
318 |
{ |
319 |
int bytes_written; |
320 |
volatile int bytes_to_write; |
321 |
|
322 |
bytes_to_write = req->range_stop - req->filepos; |
323 |
if (bytes_to_write > SOCKETBUF_SIZE) |
324 |
bytes_to_write = SOCKETBUF_SIZE; |
325 |
|
326 |
|
327 |
if (setjmp(params->env) == 0) { |
328 |
params->handle_sigbus = 1; |
329 |
|
330 |
bytes_written = |
331 |
socket_send(req, req->data_mem + req->filepos, bytes_to_write); |
332 |
|
333 |
params->handle_sigbus = 0; |
334 |
/* OK, SIGBUS **after** this point is very bad! */ |
335 |
} else { |
336 |
char buf[30]; |
337 |
/* sigbus! */ |
338 |
log_error_doc(req); |
339 |
/* sending an error here is inappropriate |
340 |
* if we are here, the file is mmapped, and thus, |
341 |
* a content-length has been sent. If we send fewer bytes |
342 |
* the client knows there has been a problem. |
343 |
* We run the risk of accidentally sending the right number |
344 |
* of bytes (or a few too many) and the client |
345 |
* won't be the wiser. |
346 |
*/ |
347 |
req->status = DEAD; |
348 |
get_commonlog_time( buf); |
349 |
fprintf(stderr, "%sGot SIGBUS in write(2)!\n", buf); |
350 |
return 0; |
351 |
} |
352 |
|
353 |
if (bytes_written < 0) { |
354 |
if (errno == EWOULDBLOCK || errno == EAGAIN) |
355 |
return -1; |
356 |
/* request blocked at the pipe level, but keep going */ |
357 |
else { |
358 |
if (errno != EPIPE) { |
359 |
log_error_doc(req); |
360 |
/* Can generate lots of log entries, */ |
361 |
perror("write"); |
362 |
/* OK to disable if your logs get too big */ |
363 |
} |
364 |
req->status = DEAD; |
365 |
return 0; |
366 |
} |
367 |
} |
368 |
req->filepos += bytes_written; |
369 |
|
370 |
if (req->filepos == req->range_stop) { /* EOF */ |
371 |
return 0; |
372 |
} else |
373 |
return 1; /* more to do */ |
374 |
} |
375 |
|
376 |
/* |
377 |
* Name: get_dir |
378 |
* Description: Called from process_get if the request is a directory. |
379 |
* statbuf must describe directory on input, since we may need its |
380 |
* device, inode, and mtime. |
381 |
* statbuf is updated, since we may need to check mtimes of a cache. |
382 |
* returns: |
383 |
* -1 error |
384 |
* 0 cgi (either gunzip or auto-generated) |
385 |
* >0 file descriptor of file |
386 |
*/ |
387 |
|
388 |
int get_dir(request * req, struct stat *statbuf) |
389 |
{ |
390 |
|
391 |
char pathname_with_index[MAX_PATH_LENGTH]; |
392 |
int data_fd; |
393 |
|
394 |
if (directory_index) { /* look for index.html first?? */ |
395 |
strcpy(pathname_with_index, req->pathname); |
396 |
strcat(pathname_with_index, directory_index); |
397 |
/* |
398 |
sprintf(pathname_with_index, "%s%s", req->pathname, directory_index); |
399 |
*/ |
400 |
|
401 |
data_fd = open(pathname_with_index, O_RDONLY); |
402 |
|
403 |
if (data_fd != -1) { /* user's index file */ |
404 |
strcpy(req->request_uri, directory_index); /* for mimetype */ |
405 |
fstat(data_fd, statbuf); |
406 |
return data_fd; |
407 |
} |
408 |
if (errno == EACCES) { |
409 |
send_r_forbidden(req); |
410 |
return -1; |
411 |
} else if (errno != ENOENT) { |
412 |
/* if there is an error *other* than EACCES or ENOENT */ |
413 |
send_r_not_found(req); |
414 |
return -1; |
415 |
} |
416 |
#ifdef GUNZIP |
417 |
/* if we are here, trying index.html didn't work |
418 |
* try index.html.gz |
419 |
*/ |
420 |
strcat(pathname_with_index, ".gz"); |
421 |
data_fd = open(pathname_with_index, O_RDONLY); |
422 |
if (data_fd != -1) { /* user's index file */ |
423 |
close(data_fd); |
424 |
|
425 |
req->response_status = R_REQUEST_OK; |
426 |
SQUASH_KA(req); |
427 |
if (req->pathname) |
428 |
free(req->pathname); |
429 |
req->pathname = strdup(pathname_with_index); |
430 |
if (!req->pathname) { |
431 |
log_error_time(); |
432 |
perror("strdup"); |
433 |
send_r_error(req); |
434 |
return 0; |
435 |
} |
436 |
if (!req->simple) { |
437 |
req_write(req, "HTTP/1.0 200 OK-GUNZIP\r\n"); |
438 |
print_http_headers(req); |
439 |
print_last_modified(req); |
440 |
req_write(req, "Content-Type: "); |
441 |
req_write(req, get_mime_type(directory_index)); |
442 |
req_write(req, "\r\n\r\n"); |
443 |
req_flush(req); |
444 |
} |
445 |
if (req->method == M_HEAD) |
446 |
return 0; |
447 |
req->is_cgi = GUNZIP_CGI; |
448 |
return init_cgi(req); |
449 |
} |
450 |
#endif |
451 |
} |
452 |
|
453 |
/* only here if index.html, index.html.gz don't exist */ |
454 |
if (dirmaker != NULL) { /* don't look for index.html... maybe automake? */ |
455 |
req->response_status = R_REQUEST_OK; |
456 |
SQUASH_KA(req); |
457 |
|
458 |
/* the indexer should take care of all headers */ |
459 |
if (!req->simple) { |
460 |
req_write(req, "HTTP/1.0 200 OK\r\n"); |
461 |
print_http_headers(req); |
462 |
print_last_modified(req); |
463 |
req_write(req, "Content-Type: text/html\r\n\r\n"); |
464 |
req_flush(req); |
465 |
} |
466 |
if (req->method == M_HEAD) |
467 |
return 0; |
468 |
|
469 |
req->is_cgi = INDEXER_CGI; |
470 |
return init_cgi(req); |
471 |
/* in this case, 0 means success */ |
472 |
} else if (cachedir) { |
473 |
return get_cachedir_file(req, statbuf); |
474 |
} else { /* neither index.html nor autogenerate are allowed */ |
475 |
send_r_forbidden(req); |
476 |
return -1; /* nothing worked */ |
477 |
} |
478 |
} |
479 |
|
480 |
int get_cachedir_file(request * req, struct stat *statbuf) |
481 |
{ |
482 |
|
483 |
char pathname_with_index[MAX_PATH_LENGTH]; |
484 |
int data_fd; |
485 |
time_t real_dir_mtime; |
486 |
|
487 |
real_dir_mtime = statbuf->st_mtime; |
488 |
sprintf(pathname_with_index, "%s/dir.%d.%ld", |
489 |
cachedir, (int) statbuf->st_dev, statbuf->st_ino); |
490 |
data_fd = open(pathname_with_index, O_RDONLY); |
491 |
|
492 |
if (data_fd != -1) { /* index cache */ |
493 |
|
494 |
fstat(data_fd, statbuf); |
495 |
if (statbuf->st_mtime > real_dir_mtime) { |
496 |
statbuf->st_mtime = real_dir_mtime; /* lie */ |
497 |
strcpy(req->request_uri, directory_index); /* for mimetype */ |
498 |
return data_fd; |
499 |
} |
500 |
close(data_fd); |
501 |
unlink(pathname_with_index); /* cache is stale, delete it */ |
502 |
} |
503 |
if (index_directory(req, pathname_with_index) == -1) |
504 |
return -1; |
505 |
|
506 |
data_fd = open(pathname_with_index, O_RDONLY); /* Last chance */ |
507 |
if (data_fd != -1) { |
508 |
strcpy(req->request_uri, directory_index); /* for mimetype */ |
509 |
fstat(data_fd, statbuf); |
510 |
statbuf->st_mtime = real_dir_mtime; /* lie */ |
511 |
return data_fd; |
512 |
} |
513 |
|
514 |
boa_perror(req, "re-opening dircache"); |
515 |
return -1; /* Nothing worked. */ |
516 |
|
517 |
} |
518 |
|
519 |
/* |
520 |
* Name: index_directory |
521 |
* Description: Called from get_cachedir_file if a directory html |
522 |
* has to be generated on the fly |
523 |
* returns -1 for problem, else 0 |
524 |
* This version is the fastest, ugliest, and most accurate yet. |
525 |
* It solves the "stale size or type" problem by not ever giving |
526 |
* the size or type. This also speeds it up since no per-file |
527 |
* stat() is required. |
528 |
*/ |
529 |
|
530 |
int index_directory(request * req, char *dest_filename) |
531 |
{ |
532 |
DIR *request_dir; |
533 |
FILE *fdstream; |
534 |
struct dirent *dirbuf; |
535 |
int bytes = 0; |
536 |
char *escname = NULL; |
537 |
|
538 |
if (chdir(req->pathname) == -1) { |
539 |
if (errno == EACCES || errno == EPERM) { |
540 |
send_r_forbidden(req); |
541 |
} else { |
542 |
log_error_doc(req); |
543 |
perror("chdir"); |
544 |
send_r_bad_request(req); |
545 |
} |
546 |
return -1; |
547 |
} |
548 |
|
549 |
request_dir = opendir("."); |
550 |
if (request_dir == NULL) { |
551 |
int errno_save = errno; |
552 |
send_r_error(req); |
553 |
log_error_time(); |
554 |
fprintf(stderr, "directory \"%s\": ", req->pathname); |
555 |
errno = errno_save; |
556 |
perror("opendir"); |
557 |
return -1; |
558 |
} |
559 |
|
560 |
fdstream = fopen(dest_filename, "w"); |
561 |
if (fdstream == NULL) { |
562 |
boa_perror(req, "dircache fopen"); |
563 |
closedir(request_dir); |
564 |
return -1; |
565 |
} |
566 |
|
567 |
bytes += fprintf(fdstream, |
568 |
"<HTML><HEAD>\n<TITLE>Index of %s</TITLE>\n</HEAD>\n\n", |
569 |
req->request_uri); |
570 |
bytes += |
571 |
fprintf(fdstream, "<BODY>\n\n<H2>Index of %s</H2>\n\n<PRE>\n", |
572 |
req->request_uri); |
573 |
|
574 |
while ((dirbuf = readdir(request_dir))) { |
575 |
if (!strcmp(dirbuf->d_name, ".")) |
576 |
continue; |
577 |
|
578 |
if (!strcmp(dirbuf->d_name, "..")) { |
579 |
bytes += fprintf(fdstream, |
580 |
" [DIR] <A HREF=\"../\">Parent Directory</A>\n"); |
581 |
continue; |
582 |
} |
583 |
|
584 |
if ((escname = escape_string(dirbuf->d_name, NULL)) != NULL) { |
585 |
bytes += fprintf(fdstream, " <A HREF=\"%s\">%s</A>\n", |
586 |
escname, dirbuf->d_name); |
587 |
free(escname); |
588 |
escname = NULL; |
589 |
} |
590 |
} |
591 |
closedir(request_dir); |
592 |
bytes += fprintf(fdstream, "</PRE>\n\n</BODY>\n</HTML>\n"); |
593 |
|
594 |
fclose(fdstream); |
595 |
|
596 |
chdir(server_root); |
597 |
|
598 |
req->filesize = bytes; /* for logging transfer size */ |
599 |
return 0; /* success */ |
600 |
} |