1 |
/* |
2 |
* Hydra, 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 1.13 2002/09/28 16:32:37 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 |
|
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 |
if (data_fd == -1) { |
52 |
log_error_doc(req); |
53 |
errno = saved_errno; |
54 |
perror("document open"); |
55 |
|
56 |
if (saved_errno == ENOENT) |
57 |
send_r_not_found(req); |
58 |
else if (saved_errno == EACCES) |
59 |
send_r_forbidden(req); |
60 |
else |
61 |
send_r_bad_request(req); |
62 |
return 0; |
63 |
} |
64 |
|
65 |
if (fstat( data_fd, &statbuf) == -1) { |
66 |
/* this is quite impossible, since the file |
67 |
* was opened before. |
68 |
*/ |
69 |
close(data_fd); |
70 |
send_r_not_found(req); |
71 |
return 0; |
72 |
} |
73 |
|
74 |
if (S_ISDIR(statbuf.st_mode)) { /* directory */ |
75 |
close(data_fd); /* close dir */ |
76 |
|
77 |
if (req->pathname[strlen(req->pathname) - 1] != '/') { |
78 |
char buffer[3 * MAX_PATH_LENGTH + 128]; |
79 |
|
80 |
create_url( buffer, sizeof(buffer), req->secure, req->hostname, |
81 |
params->server_s[req->secure].port, req->request_uri); |
82 |
|
83 |
send_r_moved_perm(req, buffer); |
84 |
return 0; |
85 |
} |
86 |
data_fd = get_dir(req, &statbuf); /* updates statbuf */ |
87 |
|
88 |
if (data_fd == -1) /* couldn't do it */ |
89 |
return 0; /* errors reported by get_dir */ |
90 |
else if (data_fd <= 1) |
91 |
/* data_fd == 0 -> close it down, 1 -> continue */ |
92 |
return data_fd; |
93 |
/* else, data_fd contains the fd of the file... */ |
94 |
} |
95 |
|
96 |
if (req->if_modified_since && |
97 |
!modified_since(&(statbuf.st_mtime), req->if_modified_since)) { |
98 |
send_r_not_modified(req); |
99 |
close(data_fd); |
100 |
return 0; |
101 |
} |
102 |
req->filesize = statbuf.st_size; |
103 |
req->last_modified = statbuf.st_mtime; |
104 |
|
105 |
if (req->range_stop == 0) |
106 |
req->range_stop = statbuf.st_size; |
107 |
|
108 |
/* out of range! */ |
109 |
if (req->range_start > statbuf.st_size || |
110 |
req->range_stop > statbuf.st_size || |
111 |
req->range_stop < req->range_start) { |
112 |
send_r_range_unsatisfiable(req); |
113 |
close(data_fd); |
114 |
return 0; |
115 |
} |
116 |
|
117 |
if (req->method == M_HEAD || req->filesize == 0) { |
118 |
send_r_request_ok(req); |
119 |
close(data_fd); |
120 |
return 0; |
121 |
} |
122 |
|
123 |
if (req->range_stop > max_file_size_cache) { |
124 |
|
125 |
if ( req->range_start == 0 && req->range_stop == statbuf.st_size) |
126 |
send_r_request_ok(req); /* All's well */ |
127 |
else { |
128 |
/* if ranges were used, then lseek to the start given |
129 |
*/ |
130 |
if ( lseek( data_fd, req->range_start, SEEK_SET) == (off_t) -1) { |
131 |
close(data_fd); |
132 |
send_r_not_found(req); |
133 |
return 0; |
134 |
} |
135 |
send_r_request_partial(req);/* All's well */ |
136 |
} |
137 |
|
138 |
req->status = PIPE_READ; |
139 |
req->cgi_status = CGI_BUFFER; |
140 |
req->data_fd = data_fd; |
141 |
req_flush(req); /* this should *always* complete due to |
142 |
the size of the I/O buffers */ |
143 |
req->header_line = req->header_end = req->buffer; |
144 |
req->pipe_range_stop = req->range_stop; |
145 |
return 1; |
146 |
} |
147 |
|
148 |
if (req->range_stop == 0) { /* done */ |
149 |
send_r_request_ok(req); /* All's well *so far* */ |
150 |
close(data_fd); |
151 |
return 1; |
152 |
} |
153 |
|
154 |
/* NOTE: I (Jon Nelson) tried performing a read(2) |
155 |
* into the output buffer provided the file data would |
156 |
* fit, before mmapping, and if successful, writing that |
157 |
* and stopping there -- all to avoid the cost |
158 |
* of a mmap. Oddly, it was *slower* in benchmarks. |
159 |
*/ |
160 |
if (max_files_cache > 0) { |
161 |
req->mmap_entry_var = find_mmap(data_fd, &statbuf); |
162 |
if (req->mmap_entry_var == NULL) { |
163 |
req->buffer_end = 0; |
164 |
if (errno == ENOENT) |
165 |
send_r_not_found(req); |
166 |
else if (errno == EACCES) |
167 |
send_r_forbidden(req); |
168 |
else |
169 |
send_r_bad_request(req); |
170 |
close(data_fd); |
171 |
return 0; |
172 |
} |
173 |
req->data_mem = req->mmap_entry_var->mmap; |
174 |
} else { /* File caching is disabled. |
175 |
*/ |
176 |
req->data_mem = mmap(0, req->range_stop, PROT_READ, MAP_OPTIONS, data_fd, 0); |
177 |
} |
178 |
|
179 |
close(data_fd); /* close data file */ |
180 |
|
181 |
if ( req->data_mem == MAP_FAILED) { |
182 |
boa_perror(req, "mmap"); |
183 |
return 0; |
184 |
} |
185 |
|
186 |
if ( req->range_start == 0 && req->range_stop == statbuf.st_size) |
187 |
send_r_request_ok(req); /* All's well */ |
188 |
else |
189 |
send_r_request_partial(req);/* All's well */ |
190 |
|
191 |
bytes = BUFFER_SIZE - req->buffer_end; |
192 |
|
193 |
/* bytes is now how much the buffer can hold |
194 |
* after the headers |
195 |
*/ |
196 |
req->filepos = req->range_start; |
197 |
|
198 |
if (bytes > 0) { |
199 |
if (bytes > req->range_stop - req->range_start) |
200 |
bytes = req->range_stop - req->range_start; |
201 |
|
202 |
if (setjmp(params->env) == 0) { |
203 |
params->handle_sigbus = 1; |
204 |
memcpy(req->buffer + req->buffer_end, &req->data_mem[req->filepos], |
205 |
bytes); |
206 |
params->handle_sigbus = 0; |
207 |
/* OK, SIGBUS **after** this point is very bad! */ |
208 |
} else { |
209 |
char buf[30]; |
210 |
/* sigbus! */ |
211 |
log_error_doc(req); |
212 |
reset_output_buffer(req); |
213 |
send_r_error(req); |
214 |
get_commonlog_time( buf); |
215 |
fprintf(stderr, "%sGot SIGBUS in memcpy!\n", |
216 |
buf); |
217 |
return 0; |
218 |
} |
219 |
req->buffer_end += bytes; |
220 |
req->filepos += bytes; |
221 |
if (req->range_stop == req->filepos) { |
222 |
req_flush(req); |
223 |
req->status = DONE; |
224 |
} |
225 |
} |
226 |
|
227 |
/* We lose statbuf here, so make sure response has been sent */ |
228 |
return 1; |
229 |
} |
230 |
|
231 |
|
232 |
/* |
233 |
* Name: process_get |
234 |
* Description: Writes a chunk of data to the socket. |
235 |
* |
236 |
* Return values: |
237 |
* -1: request blocked, move to blocked queue |
238 |
* 0: EOF or error, close it down |
239 |
* 1: successful write, recycle in ready queue |
240 |
*/ |
241 |
|
242 |
int process_get(server_params * params, request * req) |
243 |
{ |
244 |
int bytes_written; |
245 |
volatile int bytes_to_write; |
246 |
|
247 |
bytes_to_write = req->range_stop - req->filepos; |
248 |
if (bytes_to_write > SOCKETBUF_SIZE) |
249 |
bytes_to_write = SOCKETBUF_SIZE; |
250 |
|
251 |
|
252 |
if (setjmp(params->env) == 0) { |
253 |
params->handle_sigbus = 1; |
254 |
|
255 |
bytes_written = |
256 |
socket_send(req, req->data_mem + req->filepos, bytes_to_write); |
257 |
|
258 |
params->handle_sigbus = 0; |
259 |
/* OK, SIGBUS **after** this point is very bad! */ |
260 |
} else { |
261 |
char buf[30]; |
262 |
/* sigbus! */ |
263 |
log_error_doc(req); |
264 |
/* sending an error here is inappropriate |
265 |
* if we are here, the file is mmapped, and thus, |
266 |
* a content-length has been sent. If we send fewer bytes |
267 |
* the client knows there has been a problem. |
268 |
* We run the risk of accidentally sending the right number |
269 |
* of bytes (or a few too many) and the client |
270 |
* won't be the wiser. |
271 |
*/ |
272 |
req->status = DEAD; |
273 |
get_commonlog_time( buf); |
274 |
fprintf(stderr, "%sGot SIGBUS in write(2)!\n", buf); |
275 |
return 0; |
276 |
} |
277 |
|
278 |
if (bytes_written < 0) { |
279 |
if (errno == EWOULDBLOCK || errno == EAGAIN) |
280 |
return -1; |
281 |
/* request blocked at the pipe level, but keep going */ |
282 |
else { |
283 |
if (errno != EPIPE) { |
284 |
log_error_doc(req); |
285 |
/* Can generate lots of log entries, */ |
286 |
perror("write"); |
287 |
/* OK to disable if your logs get too big */ |
288 |
} |
289 |
req->status = DEAD; |
290 |
return 0; |
291 |
} |
292 |
} |
293 |
req->filepos += bytes_written; |
294 |
|
295 |
if (req->filepos == req->range_stop) { /* EOF */ |
296 |
return 0; |
297 |
} else |
298 |
return 1; /* more to do */ |
299 |
} |
300 |
|
301 |
/* |
302 |
* Name: get_dir |
303 |
* Description: Called from process_get if the request is a directory. |
304 |
* statbuf must describe directory on input, since we may need its |
305 |
* device, inode, and mtime. |
306 |
* statbuf is updated, since we may need to check mtimes of a cache. |
307 |
* returns: |
308 |
* -1 error |
309 |
* 0 cgi (either gunzip or auto-generated) |
310 |
* >0 file descriptor of file |
311 |
*/ |
312 |
|
313 |
int get_dir(request * req, struct stat *statbuf) |
314 |
{ |
315 |
|
316 |
char* directory_index; |
317 |
int data_fd; |
318 |
|
319 |
directory_index = find_and_open_directory_index( req->pathname, 0, &data_fd); |
320 |
|
321 |
if (directory_index) { /* look for index.html first?? */ |
322 |
if (data_fd != -1) { /* user's index file */ |
323 |
int ret; |
324 |
|
325 |
/* Check if we can execute the file |
326 |
*/ |
327 |
|
328 |
strcat(req->request_uri, directory_index); |
329 |
|
330 |
ret = is_executable_cgi( req, directory_index); |
331 |
if ( ret != 0) { /* it is a CGI */ |
332 |
close( data_fd); /* we don't need it */ |
333 |
if (ret == -1) { |
334 |
send_r_error(req); |
335 |
return -1; |
336 |
} |
337 |
return init_cgi(req); |
338 |
} |
339 |
|
340 |
/* Not a cgi */ |
341 |
|
342 |
fstat(data_fd, statbuf); |
343 |
return data_fd; |
344 |
} |
345 |
if (errno == EACCES) { |
346 |
send_r_forbidden(req); |
347 |
return -1; |
348 |
} else if (errno != ENOENT) { |
349 |
/* if there is an error *other* than EACCES or ENOENT */ |
350 |
send_r_not_found(req); |
351 |
return -1; |
352 |
} |
353 |
} |
354 |
|
355 |
/* only here if index.html, index.html.gz don't exist */ |
356 |
if (dirmaker != NULL) { /* don't look for index.html... maybe automake? */ |
357 |
req->response_status = R_REQUEST_OK; |
358 |
SQUASH_KA(req); |
359 |
|
360 |
/* the indexer should take care of all headers */ |
361 |
if (!req->simple) { |
362 |
req_write(req, HTTP_VERSION" 200 OK\r\n"); |
363 |
print_http_headers(req); |
364 |
print_last_modified(req); |
365 |
req_write(req, "Content-Type: " TEXT_HTML CRLF CRLF); |
366 |
req_flush(req); |
367 |
} |
368 |
if (req->method == M_HEAD) |
369 |
return 0; |
370 |
|
371 |
req->is_cgi = INDEXER_CGI; |
372 |
return init_cgi(req); |
373 |
/* in this case, 0 means success */ |
374 |
} else if (cachedir) { |
375 |
return get_cachedir_file(req, statbuf); |
376 |
} else { /* neither index.html nor autogenerate are allowed */ |
377 |
send_r_forbidden(req); |
378 |
return -1; /* nothing worked */ |
379 |
} |
380 |
} |
381 |
|
382 |
int get_cachedir_file(request * req, struct stat *statbuf) |
383 |
{ |
384 |
|
385 |
char pathname_with_index[MAX_PATH_LENGTH]; |
386 |
int data_fd; |
387 |
time_t real_dir_mtime; |
388 |
|
389 |
real_dir_mtime = statbuf->st_mtime; |
390 |
sprintf(pathname_with_index, "%s/dir.%d.%ld", |
391 |
cachedir, (int) statbuf->st_dev, statbuf->st_ino); |
392 |
data_fd = open(pathname_with_index, O_RDONLY); |
393 |
|
394 |
if (data_fd != -1) { /* index cache */ |
395 |
|
396 |
fstat(data_fd, statbuf); |
397 |
if (statbuf->st_mtime > real_dir_mtime) { |
398 |
statbuf->st_mtime = real_dir_mtime; /* lie */ |
399 |
strcpy(req->request_uri, find_default_directory_index()); /* for mimetype */ |
400 |
return data_fd; |
401 |
} |
402 |
close(data_fd); |
403 |
unlink(pathname_with_index); /* cache is stale, delete it */ |
404 |
} |
405 |
if (index_directory(req, pathname_with_index) == -1) |
406 |
return -1; |
407 |
|
408 |
data_fd = open(pathname_with_index, O_RDONLY); /* Last chance */ |
409 |
if (data_fd != -1) { |
410 |
strcpy(req->request_uri, find_default_directory_index()); /* for mimetype */ |
411 |
fstat(data_fd, statbuf); |
412 |
statbuf->st_mtime = real_dir_mtime; /* lie */ |
413 |
return data_fd; |
414 |
} |
415 |
|
416 |
boa_perror(req, "re-opening dircache"); |
417 |
return -1; /* Nothing worked. */ |
418 |
|
419 |
} |
420 |
|
421 |
/* |
422 |
* Name: index_directory |
423 |
* Description: Called from get_cachedir_file if a directory html |
424 |
* has to be generated on the fly |
425 |
* returns -1 for problem, else 0 |
426 |
* This version is the fastest, ugliest, and most accurate yet. |
427 |
* It solves the "stale size or type" problem by not ever giving |
428 |
* the size or type. This also speeds it up since no per-file |
429 |
* stat() is required. |
430 |
*/ |
431 |
|
432 |
int index_directory(request * req, char *dest_filename) |
433 |
{ |
434 |
DIR *request_dir; |
435 |
FILE *fdstream; |
436 |
struct dirent *dirbuf; |
437 |
int bytes = 0; |
438 |
char *escname = NULL; |
439 |
|
440 |
if (chdir(req->pathname) == -1) { |
441 |
if (errno == EACCES || errno == EPERM) { |
442 |
send_r_forbidden(req); |
443 |
} else { |
444 |
log_error_doc(req); |
445 |
perror("chdir"); |
446 |
send_r_bad_request(req); |
447 |
} |
448 |
return -1; |
449 |
} |
450 |
|
451 |
request_dir = opendir("."); |
452 |
if (request_dir == NULL) { |
453 |
int errno_save = errno; |
454 |
send_r_error(req); |
455 |
log_error_time(); |
456 |
fprintf(stderr, "directory \"%s\": ", req->pathname); |
457 |
errno = errno_save; |
458 |
perror("opendir"); |
459 |
return -1; |
460 |
} |
461 |
|
462 |
fdstream = fopen(dest_filename, "w"); |
463 |
if (fdstream == NULL) { |
464 |
boa_perror(req, "dircache fopen"); |
465 |
closedir(request_dir); |
466 |
return -1; |
467 |
} |
468 |
|
469 |
bytes += fprintf(fdstream, |
470 |
"<HTML><HEAD>\n<TITLE>Index of %s</TITLE>\n</HEAD>\n\n", |
471 |
req->request_uri); |
472 |
bytes += |
473 |
fprintf(fdstream, "<BODY>\n\n<H2>Index of %s</H2>\n\n<PRE>\n", |
474 |
req->request_uri); |
475 |
|
476 |
while ((dirbuf = readdir(request_dir))) { |
477 |
if (!strcmp(dirbuf->d_name, ".")) |
478 |
continue; |
479 |
|
480 |
if (!strcmp(dirbuf->d_name, "..")) { |
481 |
bytes += fprintf(fdstream, |
482 |
" [DIR] <A HREF=\"../\">Parent Directory</A>\n"); |
483 |
continue; |
484 |
} |
485 |
|
486 |
if ((escname = escape_string(dirbuf->d_name, NULL)) != NULL) { |
487 |
bytes += fprintf(fdstream, " <A HREF=\"%s\">%s</A>\n", |
488 |
escname, dirbuf->d_name); |
489 |
free(escname); |
490 |
escname = NULL; |
491 |
} |
492 |
} |
493 |
closedir(request_dir); |
494 |
bytes += fprintf(fdstream, "</PRE>\n\n</BODY>\n</HTML>\n"); |
495 |
|
496 |
fclose(fdstream); |
497 |
|
498 |
chdir(server_root); |
499 |
|
500 |
req->filesize = bytes; /* for logging transfer size */ |
501 |
return 0; /* success */ |
502 |
} |