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 |
* |
7 |
* This program is free software; you can redistribute it and/or modify |
8 |
* it under the terms of the GNU General Public License as published by |
9 |
* the Free Software Foundation; either version 1, or (at your option) |
10 |
* any later version. |
11 |
* |
12 |
* This program is distributed in the hope that it will be useful, |
13 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
14 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
15 |
* GNU General Public License for more details. |
16 |
* |
17 |
* You should have received a copy of the GNU General Public License |
18 |
* along with this program; if not, write to the Free Software |
19 |
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
20 |
* |
21 |
*/ |
22 |
|
23 |
/* $Id: get.c,v 1.76.2.5 2002/07/26 03:05:59 jnelson Exp $*/ |
24 |
|
25 |
#include "boa.h" |
26 |
#include "socket.h" |
27 |
|
28 |
/* local prototypes */ |
29 |
int get_cachedir_file(request * req, struct stat *statbuf); |
30 |
int index_directory(request * req, char *dest_filename); |
31 |
static int gzip_open(request * req); |
32 |
|
33 |
/* |
34 |
* Name: init_get |
35 |
* Description: Initializes a non-script GET or HEAD request. |
36 |
* |
37 |
* Return values: |
38 |
* 0: finished or error, request will be freed |
39 |
* 1: successfully initialized, added to ready queue |
40 |
*/ |
41 |
|
42 |
int init_get(server_params * params, request * req) |
43 |
{ |
44 |
int data_fd, saved_errno; |
45 |
struct stat statbuf; |
46 |
volatile int bytes; |
47 |
|
48 |
data_fd = open(req->pathname, O_RDONLY); |
49 |
saved_errno = errno; /* might not get used */ |
50 |
|
51 |
#ifdef GUNZIP |
52 |
if (data_fd == -1 && errno == ENOENT) { |
53 |
/* cannot open */ |
54 |
/* it's either a gunzipped file or a directory */ |
55 |
data_fd = gzip_open(req); |
56 |
|
57 |
/* since the gzip file exists then spawn the gunzip. |
58 |
*/ |
59 |
if (data_fd==1) return init_cgi(req); |
60 |
|
61 |
/* If the return value was not -1 then return it. |
62 |
*/ |
63 |
if (data_fd!=-1) return data_fd; |
64 |
} |
65 |
#endif |
66 |
|
67 |
if (data_fd == -1) { |
68 |
log_error_doc(req); |
69 |
errno = saved_errno; |
70 |
perror("document open"); |
71 |
|
72 |
if (saved_errno == ENOENT) |
73 |
send_r_not_found(req); |
74 |
else if (saved_errno == EACCES) |
75 |
send_r_forbidden(req); |
76 |
else |
77 |
send_r_bad_request(req); |
78 |
return 0; |
79 |
} |
80 |
|
81 |
if (fstat( data_fd, &statbuf) == -1) { |
82 |
/* this is quite impossible, since the file |
83 |
* was opened before. |
84 |
*/ |
85 |
close(data_fd); |
86 |
send_r_bad_request(req); |
87 |
return 0; |
88 |
} |
89 |
|
90 |
if (S_ISDIR(statbuf.st_mode)) { /* directory */ |
91 |
close(data_fd); /* close dir */ |
92 |
|
93 |
if (req->pathname[strlen(req->pathname) - 1] != '/') { |
94 |
char buffer[3 * MAX_PATH_LENGTH + 128]; |
95 |
|
96 |
if (server_port != 80) |
97 |
sprintf(buffer, "http://%s:%d%s/", server_name, |
98 |
server_port, req->request_uri); |
99 |
else |
100 |
sprintf(buffer, "http://%s%s/", server_name, req->request_uri); |
101 |
send_r_moved_perm(req, buffer); |
102 |
return 0; |
103 |
} |
104 |
data_fd = get_dir(req, &statbuf); /* updates statbuf */ |
105 |
|
106 |
if (data_fd == -1) /* couldn't do it */ |
107 |
return 0; /* errors reported by get_dir */ |
108 |
else if (data_fd <= 1) |
109 |
/* data_fd == 0 -> close it down, 1 -> continue */ |
110 |
return data_fd; |
111 |
/* else, data_fd contains the fd of the file... */ |
112 |
} |
113 |
if (req->if_modified_since && |
114 |
!modified_since(&(statbuf.st_mtime), req->if_modified_since)) { |
115 |
send_r_not_modified(req); |
116 |
close(data_fd); |
117 |
return 0; |
118 |
} |
119 |
req->filesize = statbuf.st_size; |
120 |
req->last_modified = statbuf.st_mtime; |
121 |
|
122 |
if (req->method == M_HEAD || req->filesize == 0) { |
123 |
send_r_request_ok(req); |
124 |
close(data_fd); |
125 |
return 0; |
126 |
} |
127 |
|
128 |
if (req->filesize > MAX_FILE_MMAP) { |
129 |
send_r_request_ok(req); /* All's well */ |
130 |
req->status = PIPE_READ; |
131 |
req->cgi_status = CGI_BUFFER; |
132 |
req->data_fd = data_fd; |
133 |
req_flush(req); /* this should *always* complete due to |
134 |
the size of the I/O buffers */ |
135 |
req->header_line = req->header_end = req->buffer; |
136 |
return 1; |
137 |
} |
138 |
|
139 |
if (req->filesize == 0) { /* done */ |
140 |
send_r_request_ok(req); /* All's well *so far* */ |
141 |
close(data_fd); |
142 |
return 1; |
143 |
} |
144 |
|
145 |
/* NOTE: I (Jon Nelson) tried performing a read(2) |
146 |
* into the output buffer provided the file data would |
147 |
* fit, before mmapping, and if successful, writing that |
148 |
* and stopping there -- all to avoid the cost |
149 |
* of a mmap. Oddly, it was *slower* in benchmarks. |
150 |
*/ |
151 |
req->mmap_entry_var = find_mmap(data_fd, &statbuf); |
152 |
if (req->mmap_entry_var == NULL) { |
153 |
req->buffer_end = 0; |
154 |
if (errno == ENOENT) |
155 |
send_r_not_found(req); |
156 |
else if (errno == EACCES) |
157 |
send_r_forbidden(req); |
158 |
else |
159 |
send_r_bad_request(req); |
160 |
close(data_fd); |
161 |
return 0; |
162 |
} |
163 |
req->data_mem = req->mmap_entry_var->mmap; |
164 |
//req->data_mem = mmap(0, statbuf.st_size, PROT_READ, MAP_OPTIONS, data_fd, 0); |
165 |
close(data_fd); /* close data file */ |
166 |
|
167 |
if ((long) req->data_mem == -1) { |
168 |
boa_perror(req, "mmap"); |
169 |
return 0; |
170 |
} |
171 |
|
172 |
send_r_request_ok(req); /* All's well */ |
173 |
|
174 |
bytes = BUFFER_SIZE - req->buffer_end; |
175 |
|
176 |
/* bytes is now how much the buffer can hold |
177 |
* after the headers |
178 |
*/ |
179 |
|
180 |
if (bytes > 0) { |
181 |
if (bytes > req->filesize) |
182 |
bytes = req->filesize; |
183 |
|
184 |
if (setjmp(params->env) == 0) { |
185 |
params->handle_sigbus = 1; |
186 |
memcpy(req->buffer + req->buffer_end, req->data_mem, bytes); |
187 |
params->handle_sigbus = 0; |
188 |
/* OK, SIGBUS **after** this point is very bad! */ |
189 |
} else { |
190 |
/* sigbus! */ |
191 |
log_error_doc(req); |
192 |
reset_output_buffer(req); |
193 |
send_r_error(req); |
194 |
fprintf(stderr, "%sGot SIGBUS in memcpy!\n", |
195 |
get_commonlog_time()); |
196 |
return 0; |
197 |
} |
198 |
req->buffer_end += bytes; |
199 |
req->filepos += bytes; |
200 |
if (req->filesize == req->filepos) { |
201 |
req_flush(req); |
202 |
req->status = DONE; |
203 |
} |
204 |
} |
205 |
|
206 |
/* We lose statbuf here, so make sure response has been sent */ |
207 |
return 1; |
208 |
} |
209 |
|
210 |
#ifdef GUNZIP |
211 |
|
212 |
/* If a file could not be opened, then this function |
213 |
* is called, which tries to open the gziped file. |
214 |
* Return values: |
215 |
* 0: finished or error, request will be freed |
216 |
* 1: success |
217 |
* -1: failed |
218 |
*/ |
219 |
static int gzip_open(request * req) |
220 |
{ |
221 |
char gzip_pathname[MAX_PATH_LENGTH]; |
222 |
int len, data_fd; |
223 |
|
224 |
len = strlen(req->pathname); |
225 |
|
226 |
memcpy(gzip_pathname, req->pathname, len); |
227 |
memcpy(gzip_pathname + len, ".gz", 3); |
228 |
gzip_pathname[len + 3] = '\0'; |
229 |
data_fd = open(gzip_pathname, O_RDONLY); |
230 |
if (data_fd != -1) { |
231 |
close(data_fd); |
232 |
|
233 |
req->response_status = R_REQUEST_OK; |
234 |
if (req->pathname) |
235 |
free(req->pathname); |
236 |
req->pathname = strdup(gzip_pathname); |
237 |
if (!req->pathname) { |
238 |
log_error_time(); |
239 |
perror("strdup"); |
240 |
send_r_error(req); |
241 |
return 0; |
242 |
} |
243 |
if (!req->simple) { |
244 |
req_write(req, "HTTP/1.0 200 OK-GUNZIP\r\n"); |
245 |
print_http_headers(req); |
246 |
print_content_type(req); |
247 |
print_last_modified(req); |
248 |
req_write(req, "\r\n"); |
249 |
req_flush(req); |
250 |
} |
251 |
if (req->method == M_HEAD) |
252 |
return 0; |
253 |
|
254 |
return 1; |
255 |
|
256 |
} |
257 |
|
258 |
return -1; |
259 |
} |
260 |
#endif |
261 |
|
262 |
/* |
263 |
* Name: process_get |
264 |
* Description: Writes a chunk of data to the socket. |
265 |
* |
266 |
* Return values: |
267 |
* -1: request blocked, move to blocked queue |
268 |
* 0: EOF or error, close it down |
269 |
* 1: successful write, recycle in ready queue |
270 |
*/ |
271 |
|
272 |
int process_get(server_params * params, request * req) |
273 |
{ |
274 |
int bytes_written; |
275 |
volatile int bytes_to_write; |
276 |
|
277 |
bytes_to_write = req->filesize - req->filepos; |
278 |
if (bytes_to_write > SOCKETBUF_SIZE) |
279 |
bytes_to_write = SOCKETBUF_SIZE; |
280 |
|
281 |
|
282 |
if (setjmp(params->env) == 0) { |
283 |
params->handle_sigbus = 1; |
284 |
|
285 |
bytes_written = |
286 |
socket_send(req, req->data_mem + req->filepos, bytes_to_write); |
287 |
|
288 |
params->handle_sigbus = 0; |
289 |
/* OK, SIGBUS **after** this point is very bad! */ |
290 |
} else { |
291 |
/* sigbus! */ |
292 |
log_error_doc(req); |
293 |
/* sending an error here is inappropriate |
294 |
* if we are here, the file is mmapped, and thus, |
295 |
* a content-length has been sent. If we send fewer bytes |
296 |
* the client knows there has been a problem. |
297 |
* We run the risk of accidentally sending the right number |
298 |
* of bytes (or a few too many) and the client |
299 |
* won't be the wiser. |
300 |
*/ |
301 |
req->status = DEAD; |
302 |
fprintf(stderr, "%sGot SIGBUS in write(2)!\n", get_commonlog_time()); |
303 |
return 0; |
304 |
} |
305 |
|
306 |
if (bytes_written < 0) { |
307 |
if (errno == EWOULDBLOCK || errno == EAGAIN) |
308 |
return -1; |
309 |
/* request blocked at the pipe level, but keep going */ |
310 |
else { |
311 |
if (errno != EPIPE) { |
312 |
log_error_doc(req); |
313 |
/* Can generate lots of log entries, */ |
314 |
perror("write"); |
315 |
/* OK to disable if your logs get too big */ |
316 |
} |
317 |
req->status = DEAD; |
318 |
return 0; |
319 |
} |
320 |
} |
321 |
req->filepos += bytes_written; |
322 |
|
323 |
if (req->filepos == req->filesize) { /* EOF */ |
324 |
return 0; |
325 |
} else |
326 |
return 1; /* more to do */ |
327 |
} |
328 |
|
329 |
/* |
330 |
* Name: get_dir |
331 |
* Description: Called from process_get if the request is a directory. |
332 |
* statbuf must describe directory on input, since we may need its |
333 |
* device, inode, and mtime. |
334 |
* statbuf is updated, since we may need to check mtimes of a cache. |
335 |
* returns: |
336 |
* -1 error |
337 |
* 0 cgi (either gunzip or auto-generated) |
338 |
* >0 file descriptor of file |
339 |
*/ |
340 |
|
341 |
int get_dir(request * req, struct stat *statbuf) |
342 |
{ |
343 |
|
344 |
char pathname_with_index[MAX_PATH_LENGTH]; |
345 |
int data_fd; |
346 |
|
347 |
if (directory_index) { /* look for index.html first?? */ |
348 |
strcpy(pathname_with_index, req->pathname); |
349 |
strcat(pathname_with_index, directory_index); |
350 |
/* |
351 |
sprintf(pathname_with_index, "%s%s", req->pathname, directory_index); |
352 |
*/ |
353 |
|
354 |
data_fd = open(pathname_with_index, O_RDONLY); |
355 |
|
356 |
if (data_fd != -1) { /* user's index file */ |
357 |
strcpy(req->request_uri, directory_index); /* for mimetype */ |
358 |
fstat(data_fd, statbuf); |
359 |
return data_fd; |
360 |
} |
361 |
if (errno == EACCES) { |
362 |
send_r_forbidden(req); |
363 |
return -1; |
364 |
} else if (errno != ENOENT) { |
365 |
/* if there is an error *other* than EACCES or ENOENT */ |
366 |
send_r_not_found(req); |
367 |
return -1; |
368 |
} |
369 |
#ifdef GUNZIP |
370 |
/* if we are here, trying index.html didn't work |
371 |
* try index.html.gz |
372 |
*/ |
373 |
strcat(pathname_with_index, ".gz"); |
374 |
data_fd = open(pathname_with_index, O_RDONLY); |
375 |
if (data_fd != -1) { /* user's index file */ |
376 |
close(data_fd); |
377 |
|
378 |
req->response_status = R_REQUEST_OK; |
379 |
SQUASH_KA(req); |
380 |
if (req->pathname) |
381 |
free(req->pathname); |
382 |
req->pathname = strdup(pathname_with_index); |
383 |
if (!req->pathname) { |
384 |
log_error_time(); |
385 |
perror("strdup"); |
386 |
send_r_error(req); |
387 |
return 0; |
388 |
} |
389 |
if (!req->simple) { |
390 |
req_write(req, "HTTP/1.0 200 OK-GUNZIP\r\n"); |
391 |
print_http_headers(req); |
392 |
print_last_modified(req); |
393 |
req_write(req, "Content-Type: "); |
394 |
req_write(req, get_mime_type(directory_index)); |
395 |
req_write(req, "\r\n\r\n"); |
396 |
req_flush(req); |
397 |
} |
398 |
if (req->method == M_HEAD) |
399 |
return 0; |
400 |
return init_cgi(req); |
401 |
} |
402 |
#endif |
403 |
} |
404 |
|
405 |
/* only here if index.html, index.html.gz don't exist */ |
406 |
if (dirmaker != NULL) { /* don't look for index.html... maybe automake? */ |
407 |
req->response_status = R_REQUEST_OK; |
408 |
SQUASH_KA(req); |
409 |
|
410 |
/* the indexer should take care of all headers */ |
411 |
if (!req->simple) { |
412 |
req_write(req, "HTTP/1.0 200 OK\r\n"); |
413 |
print_http_headers(req); |
414 |
print_last_modified(req); |
415 |
req_write(req, "Content-Type: text/html\r\n\r\n"); |
416 |
req_flush(req); |
417 |
} |
418 |
if (req->method == M_HEAD) |
419 |
return 0; |
420 |
|
421 |
return init_cgi(req); |
422 |
/* in this case, 0 means success */ |
423 |
} else if (cachedir) { |
424 |
return get_cachedir_file(req, statbuf); |
425 |
} else { /* neither index.html nor autogenerate are allowed */ |
426 |
send_r_forbidden(req); |
427 |
return -1; /* nothing worked */ |
428 |
} |
429 |
} |
430 |
|
431 |
int get_cachedir_file(request * req, struct stat *statbuf) |
432 |
{ |
433 |
|
434 |
char pathname_with_index[MAX_PATH_LENGTH]; |
435 |
int data_fd; |
436 |
time_t real_dir_mtime; |
437 |
|
438 |
real_dir_mtime = statbuf->st_mtime; |
439 |
sprintf(pathname_with_index, "%s/dir.%d.%ld", |
440 |
cachedir, (int) statbuf->st_dev, statbuf->st_ino); |
441 |
data_fd = open(pathname_with_index, O_RDONLY); |
442 |
|
443 |
if (data_fd != -1) { /* index cache */ |
444 |
|
445 |
fstat(data_fd, statbuf); |
446 |
if (statbuf->st_mtime > real_dir_mtime) { |
447 |
statbuf->st_mtime = real_dir_mtime; /* lie */ |
448 |
strcpy(req->request_uri, directory_index); /* for mimetype */ |
449 |
return data_fd; |
450 |
} |
451 |
close(data_fd); |
452 |
unlink(pathname_with_index); /* cache is stale, delete it */ |
453 |
} |
454 |
if (index_directory(req, pathname_with_index) == -1) |
455 |
return -1; |
456 |
|
457 |
data_fd = open(pathname_with_index, O_RDONLY); /* Last chance */ |
458 |
if (data_fd != -1) { |
459 |
strcpy(req->request_uri, directory_index); /* for mimetype */ |
460 |
fstat(data_fd, statbuf); |
461 |
statbuf->st_mtime = real_dir_mtime; /* lie */ |
462 |
return data_fd; |
463 |
} |
464 |
|
465 |
boa_perror(req, "re-opening dircache"); |
466 |
return -1; /* Nothing worked. */ |
467 |
|
468 |
} |
469 |
|
470 |
/* |
471 |
* Name: index_directory |
472 |
* Description: Called from get_cachedir_file if a directory html |
473 |
* has to be generated on the fly |
474 |
* returns -1 for problem, else 0 |
475 |
* This version is the fastest, ugliest, and most accurate yet. |
476 |
* It solves the "stale size or type" problem by not ever giving |
477 |
* the size or type. This also speeds it up since no per-file |
478 |
* stat() is required. |
479 |
*/ |
480 |
|
481 |
int index_directory(request * req, char *dest_filename) |
482 |
{ |
483 |
DIR *request_dir; |
484 |
FILE *fdstream; |
485 |
struct dirent *dirbuf; |
486 |
int bytes = 0; |
487 |
char *escname = NULL; |
488 |
|
489 |
if (chdir(req->pathname) == -1) { |
490 |
if (errno == EACCES || errno == EPERM) { |
491 |
send_r_forbidden(req); |
492 |
} else { |
493 |
log_error_doc(req); |
494 |
perror("chdir"); |
495 |
send_r_bad_request(req); |
496 |
} |
497 |
return -1; |
498 |
} |
499 |
|
500 |
request_dir = opendir("."); |
501 |
if (request_dir == NULL) { |
502 |
int errno_save = errno; |
503 |
send_r_error(req); |
504 |
log_error_time(); |
505 |
fprintf(stderr, "directory \"%s\": ", req->pathname); |
506 |
errno = errno_save; |
507 |
perror("opendir"); |
508 |
return -1; |
509 |
} |
510 |
|
511 |
fdstream = fopen(dest_filename, "w"); |
512 |
if (fdstream == NULL) { |
513 |
boa_perror(req, "dircache fopen"); |
514 |
closedir(request_dir); |
515 |
return -1; |
516 |
} |
517 |
|
518 |
bytes += fprintf(fdstream, |
519 |
"<HTML><HEAD>\n<TITLE>Index of %s</TITLE>\n</HEAD>\n\n", |
520 |
req->request_uri); |
521 |
bytes += |
522 |
fprintf(fdstream, "<BODY>\n\n<H2>Index of %s</H2>\n\n<PRE>\n", |
523 |
req->request_uri); |
524 |
|
525 |
while ((dirbuf = readdir(request_dir))) { |
526 |
if (!strcmp(dirbuf->d_name, ".")) |
527 |
continue; |
528 |
|
529 |
if (!strcmp(dirbuf->d_name, "..")) { |
530 |
bytes += fprintf(fdstream, |
531 |
" [DIR] <A HREF=\"../\">Parent Directory</A>\n"); |
532 |
continue; |
533 |
} |
534 |
|
535 |
if ((escname = escape_string(dirbuf->d_name, NULL)) != NULL) { |
536 |
bytes += fprintf(fdstream, " <A HREF=\"%s\">%s</A>\n", |
537 |
escname, dirbuf->d_name); |
538 |
free(escname); |
539 |
escname = NULL; |
540 |
} |
541 |
} |
542 |
closedir(request_dir); |
543 |
bytes += fprintf(fdstream, "</PRE>\n\n</BODY>\n</HTML>\n"); |
544 |
|
545 |
fclose(fdstream); |
546 |
|
547 |
chdir(server_root); |
548 |
|
549 |
req->filesize = bytes; /* for logging transfer size */ |
550 |
return 0; /* success */ |
551 |
} |