1 | /*************************************************************************** |
2 | * _ _ ____ _ |
3 | * Project ___| | | | _ \| | |
4 | * / __| | | | |_) | | |
5 | * | (__| |_| | _ <| |___ |
6 | * \___|\___/|_| \_\_____| |
7 | * |
8 | * Copyright (C) 1998 - 2022, Daniel Stenberg, <[email protected]>, et al. |
9 | * |
10 | * This software is licensed as described in the file COPYING, which |
11 | * you should have received as part of this distribution. The terms |
12 | * are also available at https://curl.se/docs/copyright.html. |
13 | * |
14 | * You may opt to use, copy, modify, merge, publish, distribute and/or sell |
15 | * copies of the Software, and permit persons to whom the Software is |
16 | * furnished to do so, under the terms of the COPYING file. |
17 | * |
18 | * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY |
19 | * KIND, either express or implied. |
20 | * |
21 | * SPDX-License-Identifier: curl |
22 | * |
23 | ***************************************************************************/ |
24 | |
25 | #include "curl_setup.h" |
26 | |
27 | #ifndef CURL_DISABLE_FILE |
28 | |
29 | #ifdef HAVE_NETINET_IN_H |
30 | #include <netinet/in.h> |
31 | #endif |
32 | #ifdef HAVE_NETDB_H |
33 | #include <netdb.h> |
34 | #endif |
35 | #ifdef HAVE_ARPA_INET_H |
36 | #include <arpa/inet.h> |
37 | #endif |
38 | #ifdef HAVE_NET_IF_H |
39 | #include <net/if.h> |
40 | #endif |
41 | #ifdef HAVE_SYS_IOCTL_H |
42 | #include <sys/ioctl.h> |
43 | #endif |
44 | |
45 | #ifdef HAVE_SYS_PARAM_H |
46 | #include <sys/param.h> |
47 | #endif |
48 | |
49 | #ifdef HAVE_FCNTL_H |
50 | #include <fcntl.h> |
51 | #endif |
52 | |
53 | #include "strtoofft.h" |
54 | #include "urldata.h" |
55 | #include <curl/curl.h> |
56 | #include "progress.h" |
57 | #include "sendf.h" |
58 | #include "escape.h" |
59 | #include "file.h" |
60 | #include "speedcheck.h" |
61 | #include "getinfo.h" |
62 | #include "transfer.h" |
63 | #include "url.h" |
64 | #include "parsedate.h" /* for the week day and month names */ |
65 | #include "warnless.h" |
66 | #include "curl_range.h" |
67 | /* The last 3 #include files should be in this order */ |
68 | #include "curl_printf.h" |
69 | #include "curl_memory.h" |
70 | #include "memdebug.h" |
71 | |
72 | #if defined(WIN32) || defined(MSDOS) || defined(__EMX__) |
73 | #define DOS_FILESYSTEM 1 |
74 | #elif defined(__amigaos4__) |
75 | #define AMIGA_FILESYSTEM 1 |
76 | #endif |
77 | |
78 | #ifdef OPEN_NEEDS_ARG3 |
79 | # define open_readonly(p,f) open((p),(f),(0)) |
80 | #else |
81 | # define open_readonly(p,f) open((p),(f)) |
82 | #endif |
83 | |
84 | /* |
85 | * Forward declarations. |
86 | */ |
87 | |
88 | static CURLcode file_do(struct Curl_easy *data, bool *done); |
89 | static CURLcode file_done(struct Curl_easy *data, |
90 | CURLcode status, bool premature); |
91 | static CURLcode file_connect(struct Curl_easy *data, bool *done); |
92 | static CURLcode file_disconnect(struct Curl_easy *data, |
93 | struct connectdata *conn, |
94 | bool dead_connection); |
95 | static CURLcode file_setup_connection(struct Curl_easy *data, |
96 | struct connectdata *conn); |
97 | |
98 | /* |
99 | * FILE scheme handler. |
100 | */ |
101 | |
102 | const struct Curl_handler Curl_handler_file = { |
103 | "FILE" , /* scheme */ |
104 | file_setup_connection, /* setup_connection */ |
105 | file_do, /* do_it */ |
106 | file_done, /* done */ |
107 | ZERO_NULL, /* do_more */ |
108 | file_connect, /* connect_it */ |
109 | ZERO_NULL, /* connecting */ |
110 | ZERO_NULL, /* doing */ |
111 | ZERO_NULL, /* proto_getsock */ |
112 | ZERO_NULL, /* doing_getsock */ |
113 | ZERO_NULL, /* domore_getsock */ |
114 | ZERO_NULL, /* perform_getsock */ |
115 | file_disconnect, /* disconnect */ |
116 | ZERO_NULL, /* readwrite */ |
117 | ZERO_NULL, /* connection_check */ |
118 | ZERO_NULL, /* attach connection */ |
119 | 0, /* defport */ |
120 | CURLPROTO_FILE, /* protocol */ |
121 | CURLPROTO_FILE, /* family */ |
122 | PROTOPT_NONETWORK | PROTOPT_NOURLQUERY /* flags */ |
123 | }; |
124 | |
125 | |
126 | static CURLcode file_setup_connection(struct Curl_easy *data, |
127 | struct connectdata *conn) |
128 | { |
129 | (void)conn; |
130 | /* allocate the FILE specific struct */ |
131 | data->req.p.file = calloc(1, sizeof(struct FILEPROTO)); |
132 | if(!data->req.p.file) |
133 | return CURLE_OUT_OF_MEMORY; |
134 | |
135 | return CURLE_OK; |
136 | } |
137 | |
138 | /* |
139 | * file_connect() gets called from Curl_protocol_connect() to allow us to |
140 | * do protocol-specific actions at connect-time. We emulate a |
141 | * connect-then-transfer protocol and "connect" to the file here |
142 | */ |
143 | static CURLcode file_connect(struct Curl_easy *data, bool *done) |
144 | { |
145 | char *real_path; |
146 | struct FILEPROTO *file = data->req.p.file; |
147 | int fd; |
148 | #ifdef DOS_FILESYSTEM |
149 | size_t i; |
150 | char *actual_path; |
151 | #endif |
152 | size_t real_path_len; |
153 | |
154 | CURLcode result = Curl_urldecode(data->state.up.path, 0, &real_path, |
155 | &real_path_len, REJECT_ZERO); |
156 | if(result) |
157 | return result; |
158 | |
159 | #ifdef DOS_FILESYSTEM |
160 | /* If the first character is a slash, and there's |
161 | something that looks like a drive at the beginning of |
162 | the path, skip the slash. If we remove the initial |
163 | slash in all cases, paths without drive letters end up |
164 | relative to the current directory which isn't how |
165 | browsers work. |
166 | |
167 | Some browsers accept | instead of : as the drive letter |
168 | separator, so we do too. |
169 | |
170 | On other platforms, we need the slash to indicate an |
171 | absolute pathname. On Windows, absolute paths start |
172 | with a drive letter. |
173 | */ |
174 | actual_path = real_path; |
175 | if((actual_path[0] == '/') && |
176 | actual_path[1] && |
177 | (actual_path[2] == ':' || actual_path[2] == '|')) { |
178 | actual_path[2] = ':'; |
179 | actual_path++; |
180 | real_path_len--; |
181 | } |
182 | |
183 | /* change path separators from '/' to '\\' for DOS, Windows and OS/2 */ |
184 | for(i = 0; i < real_path_len; ++i) |
185 | if(actual_path[i] == '/') |
186 | actual_path[i] = '\\'; |
187 | else if(!actual_path[i]) { /* binary zero */ |
188 | Curl_safefree(real_path); |
189 | return CURLE_URL_MALFORMAT; |
190 | } |
191 | |
192 | fd = open_readonly(actual_path, O_RDONLY|O_BINARY); |
193 | file->path = actual_path; |
194 | #else |
195 | if(memchr(real_path, 0, real_path_len)) { |
196 | /* binary zeroes indicate foul play */ |
197 | Curl_safefree(real_path); |
198 | return CURLE_URL_MALFORMAT; |
199 | } |
200 | |
201 | #ifdef AMIGA_FILESYSTEM |
202 | /* |
203 | * A leading slash in an AmigaDOS path denotes the parent |
204 | * directory, and hence we block this as it is relative. |
205 | * Absolute paths start with 'volumename:', so we check for |
206 | * this first. Failing that, we treat the path as a real unix |
207 | * path, but only if the application was compiled with -lunix. |
208 | */ |
209 | fd = -1; |
210 | file->path = real_path; |
211 | |
212 | if(real_path[0] == '/') { |
213 | extern int __unix_path_semantics; |
214 | if(strchr(real_path + 1, ':')) { |
215 | /* Amiga absolute path */ |
216 | fd = open_readonly(real_path + 1, O_RDONLY); |
217 | file->path++; |
218 | } |
219 | else if(__unix_path_semantics) { |
220 | /* -lunix fallback */ |
221 | fd = open_readonly(real_path, O_RDONLY); |
222 | } |
223 | } |
224 | #else |
225 | fd = open_readonly(real_path, O_RDONLY); |
226 | file->path = real_path; |
227 | #endif |
228 | #endif |
229 | file->freepath = real_path; /* free this when done */ |
230 | |
231 | file->fd = fd; |
232 | if(!data->set.upload && (fd == -1)) { |
233 | failf(data, "Couldn't open file %s" , data->state.up.path); |
234 | file_done(data, CURLE_FILE_COULDNT_READ_FILE, FALSE); |
235 | return CURLE_FILE_COULDNT_READ_FILE; |
236 | } |
237 | *done = TRUE; |
238 | |
239 | return CURLE_OK; |
240 | } |
241 | |
242 | static CURLcode file_done(struct Curl_easy *data, |
243 | CURLcode status, bool premature) |
244 | { |
245 | struct FILEPROTO *file = data->req.p.file; |
246 | (void)status; /* not used */ |
247 | (void)premature; /* not used */ |
248 | |
249 | if(file) { |
250 | Curl_safefree(file->freepath); |
251 | file->path = NULL; |
252 | if(file->fd != -1) |
253 | close(file->fd); |
254 | file->fd = -1; |
255 | } |
256 | |
257 | return CURLE_OK; |
258 | } |
259 | |
260 | static CURLcode file_disconnect(struct Curl_easy *data, |
261 | struct connectdata *conn, |
262 | bool dead_connection) |
263 | { |
264 | (void)dead_connection; /* not used */ |
265 | (void)conn; |
266 | return file_done(data, CURLE_OK, FALSE); |
267 | } |
268 | |
269 | #ifdef DOS_FILESYSTEM |
270 | #define DIRSEP '\\' |
271 | #else |
272 | #define DIRSEP '/' |
273 | #endif |
274 | |
275 | static CURLcode file_upload(struct Curl_easy *data) |
276 | { |
277 | struct FILEPROTO *file = data->req.p.file; |
278 | const char *dir = strchr(file->path, DIRSEP); |
279 | int fd; |
280 | int mode; |
281 | CURLcode result = CURLE_OK; |
282 | char *buf = data->state.buffer; |
283 | curl_off_t bytecount = 0; |
284 | struct_stat file_stat; |
285 | const char *buf2; |
286 | |
287 | /* |
288 | * Since FILE: doesn't do the full init, we need to provide some extra |
289 | * assignments here. |
290 | */ |
291 | data->req.upload_fromhere = buf; |
292 | |
293 | if(!dir) |
294 | return CURLE_FILE_COULDNT_READ_FILE; /* fix: better error code */ |
295 | |
296 | if(!dir[1]) |
297 | return CURLE_FILE_COULDNT_READ_FILE; /* fix: better error code */ |
298 | |
299 | #ifdef O_BINARY |
300 | #define MODE_DEFAULT O_WRONLY|O_CREAT|O_BINARY |
301 | #else |
302 | #define MODE_DEFAULT O_WRONLY|O_CREAT |
303 | #endif |
304 | |
305 | if(data->state.resume_from) |
306 | mode = MODE_DEFAULT|O_APPEND; |
307 | else |
308 | mode = MODE_DEFAULT|O_TRUNC; |
309 | |
310 | fd = open(file->path, mode, data->set.new_file_perms); |
311 | if(fd < 0) { |
312 | failf(data, "Can't open %s for writing" , file->path); |
313 | return CURLE_WRITE_ERROR; |
314 | } |
315 | |
316 | if(-1 != data->state.infilesize) |
317 | /* known size of data to "upload" */ |
318 | Curl_pgrsSetUploadSize(data, data->state.infilesize); |
319 | |
320 | /* treat the negative resume offset value as the case of "-" */ |
321 | if(data->state.resume_from < 0) { |
322 | if(fstat(fd, &file_stat)) { |
323 | close(fd); |
324 | failf(data, "Can't get the size of %s" , file->path); |
325 | return CURLE_WRITE_ERROR; |
326 | } |
327 | data->state.resume_from = (curl_off_t)file_stat.st_size; |
328 | } |
329 | |
330 | while(!result) { |
331 | size_t nread; |
332 | size_t nwrite; |
333 | size_t readcount; |
334 | result = Curl_fillreadbuffer(data, data->set.buffer_size, &readcount); |
335 | if(result) |
336 | break; |
337 | |
338 | if(!readcount) |
339 | break; |
340 | |
341 | nread = readcount; |
342 | |
343 | /*skip bytes before resume point*/ |
344 | if(data->state.resume_from) { |
345 | if((curl_off_t)nread <= data->state.resume_from) { |
346 | data->state.resume_from -= nread; |
347 | nread = 0; |
348 | buf2 = buf; |
349 | } |
350 | else { |
351 | buf2 = buf + data->state.resume_from; |
352 | nread -= (size_t)data->state.resume_from; |
353 | data->state.resume_from = 0; |
354 | } |
355 | } |
356 | else |
357 | buf2 = buf; |
358 | |
359 | /* write the data to the target */ |
360 | nwrite = write(fd, buf2, nread); |
361 | if(nwrite != nread) { |
362 | result = CURLE_SEND_ERROR; |
363 | break; |
364 | } |
365 | |
366 | bytecount += nread; |
367 | |
368 | Curl_pgrsSetUploadCounter(data, bytecount); |
369 | |
370 | if(Curl_pgrsUpdate(data)) |
371 | result = CURLE_ABORTED_BY_CALLBACK; |
372 | else |
373 | result = Curl_speedcheck(data, Curl_now()); |
374 | } |
375 | if(!result && Curl_pgrsUpdate(data)) |
376 | result = CURLE_ABORTED_BY_CALLBACK; |
377 | |
378 | close(fd); |
379 | |
380 | return result; |
381 | } |
382 | |
383 | /* |
384 | * file_do() is the protocol-specific function for the do-phase, separated |
385 | * from the connect-phase above. Other protocols merely setup the transfer in |
386 | * the do-phase, to have it done in the main transfer loop but since some |
387 | * platforms we support don't allow select()ing etc on file handles (as |
388 | * opposed to sockets) we instead perform the whole do-operation in this |
389 | * function. |
390 | */ |
391 | static CURLcode file_do(struct Curl_easy *data, bool *done) |
392 | { |
393 | /* This implementation ignores the host name in conformance with |
394 | RFC 1738. Only local files (reachable via the standard file system) |
395 | are supported. This means that files on remotely mounted directories |
396 | (via NFS, Samba, NT sharing) can be accessed through a file:// URL |
397 | */ |
398 | CURLcode result = CURLE_OK; |
399 | struct_stat statbuf; /* struct_stat instead of struct stat just to allow the |
400 | Windows version to have a different struct without |
401 | having to redefine the simple word 'stat' */ |
402 | curl_off_t expected_size = -1; |
403 | bool size_known; |
404 | bool fstated = FALSE; |
405 | char *buf = data->state.buffer; |
406 | curl_off_t bytecount = 0; |
407 | int fd; |
408 | struct FILEPROTO *file; |
409 | |
410 | *done = TRUE; /* unconditionally */ |
411 | |
412 | Curl_pgrsStartNow(data); |
413 | |
414 | if(data->set.upload) |
415 | return file_upload(data); |
416 | |
417 | file = data->req.p.file; |
418 | |
419 | /* get the fd from the connection phase */ |
420 | fd = file->fd; |
421 | |
422 | /* VMS: This only works reliable for STREAMLF files */ |
423 | if(-1 != fstat(fd, &statbuf)) { |
424 | if(!S_ISDIR(statbuf.st_mode)) |
425 | expected_size = statbuf.st_size; |
426 | /* and store the modification time */ |
427 | data->info.filetime = statbuf.st_mtime; |
428 | fstated = TRUE; |
429 | } |
430 | |
431 | if(fstated && !data->state.range && data->set.timecondition) { |
432 | if(!Curl_meets_timecondition(data, data->info.filetime)) { |
433 | *done = TRUE; |
434 | return CURLE_OK; |
435 | } |
436 | } |
437 | |
438 | if(fstated) { |
439 | time_t filetime; |
440 | struct tm buffer; |
441 | const struct tm *tm = &buffer; |
442 | char header[80]; |
443 | int headerlen; |
444 | char accept_ranges[24]= { "Accept-ranges: bytes\r\n" }; |
445 | if(expected_size >= 0) { |
446 | headerlen = msnprintf(header, sizeof(header), |
447 | "Content-Length: %" CURL_FORMAT_CURL_OFF_T "\r\n" , |
448 | expected_size); |
449 | result = Curl_client_write(data, CLIENTWRITE_HEADER, header, headerlen); |
450 | if(result) |
451 | return result; |
452 | |
453 | result = Curl_client_write(data, CLIENTWRITE_HEADER, |
454 | accept_ranges, strlen(accept_ranges)); |
455 | if(result != CURLE_OK) |
456 | return result; |
457 | } |
458 | |
459 | filetime = (time_t)statbuf.st_mtime; |
460 | result = Curl_gmtime(filetime, &buffer); |
461 | if(result) |
462 | return result; |
463 | |
464 | /* format: "Tue, 15 Nov 1994 12:45:26 GMT" */ |
465 | headerlen = msnprintf(header, sizeof(header), |
466 | "Last-Modified: %s, %02d %s %4d %02d:%02d:%02d GMT\r\n%s" , |
467 | Curl_wkday[tm->tm_wday?tm->tm_wday-1:6], |
468 | tm->tm_mday, |
469 | Curl_month[tm->tm_mon], |
470 | tm->tm_year + 1900, |
471 | tm->tm_hour, |
472 | tm->tm_min, |
473 | tm->tm_sec, |
474 | data->set.opt_no_body ? "" : "\r\n" ); |
475 | result = Curl_client_write(data, CLIENTWRITE_HEADER, header, headerlen); |
476 | if(result) |
477 | return result; |
478 | /* set the file size to make it available post transfer */ |
479 | Curl_pgrsSetDownloadSize(data, expected_size); |
480 | if(data->set.opt_no_body) |
481 | return result; |
482 | } |
483 | |
484 | /* Check whether file range has been specified */ |
485 | result = Curl_range(data); |
486 | if(result) |
487 | return result; |
488 | |
489 | /* Adjust the start offset in case we want to get the N last bytes |
490 | * of the stream if the filesize could be determined */ |
491 | if(data->state.resume_from < 0) { |
492 | if(!fstated) { |
493 | failf(data, "Can't get the size of file." ); |
494 | return CURLE_READ_ERROR; |
495 | } |
496 | data->state.resume_from += (curl_off_t)statbuf.st_size; |
497 | } |
498 | |
499 | if(data->state.resume_from > 0) { |
500 | /* We check explicitly if we have a start offset, because |
501 | * expected_size may be -1 if we don't know how large the file is, |
502 | * in which case we should not adjust it. */ |
503 | if(data->state.resume_from <= expected_size) |
504 | expected_size -= data->state.resume_from; |
505 | else { |
506 | failf(data, "failed to resume file:// transfer" ); |
507 | return CURLE_BAD_DOWNLOAD_RESUME; |
508 | } |
509 | } |
510 | |
511 | /* A high water mark has been specified so we obey... */ |
512 | if(data->req.maxdownload > 0) |
513 | expected_size = data->req.maxdownload; |
514 | |
515 | if(!fstated || (expected_size <= 0)) |
516 | size_known = FALSE; |
517 | else |
518 | size_known = TRUE; |
519 | |
520 | /* The following is a shortcut implementation of file reading |
521 | this is both more efficient than the former call to download() and |
522 | it avoids problems with select() and recv() on file descriptors |
523 | in Winsock */ |
524 | if(size_known) |
525 | Curl_pgrsSetDownloadSize(data, expected_size); |
526 | |
527 | if(data->state.resume_from) { |
528 | if(data->state.resume_from != |
529 | lseek(fd, data->state.resume_from, SEEK_SET)) |
530 | return CURLE_BAD_DOWNLOAD_RESUME; |
531 | } |
532 | |
533 | Curl_pgrsTime(data, TIMER_STARTTRANSFER); |
534 | |
535 | while(!result) { |
536 | ssize_t nread; |
537 | /* Don't fill a whole buffer if we want less than all data */ |
538 | size_t bytestoread; |
539 | |
540 | if(size_known) { |
541 | bytestoread = (expected_size < data->set.buffer_size) ? |
542 | curlx_sotouz(expected_size) : (size_t)data->set.buffer_size; |
543 | } |
544 | else |
545 | bytestoread = data->set.buffer_size-1; |
546 | |
547 | nread = read(fd, buf, bytestoread); |
548 | |
549 | if(nread > 0) |
550 | buf[nread] = 0; |
551 | |
552 | if(nread <= 0 || (size_known && (expected_size == 0))) |
553 | break; |
554 | |
555 | bytecount += nread; |
556 | if(size_known) |
557 | expected_size -= nread; |
558 | |
559 | result = Curl_client_write(data, CLIENTWRITE_BODY, buf, nread); |
560 | if(result) |
561 | return result; |
562 | |
563 | Curl_pgrsSetDownloadCounter(data, bytecount); |
564 | |
565 | if(Curl_pgrsUpdate(data)) |
566 | result = CURLE_ABORTED_BY_CALLBACK; |
567 | else |
568 | result = Curl_speedcheck(data, Curl_now()); |
569 | } |
570 | if(Curl_pgrsUpdate(data)) |
571 | result = CURLE_ABORTED_BY_CALLBACK; |
572 | |
573 | return result; |
574 | } |
575 | |
576 | #endif |
577 | |