1 /***************************************************************************
2 * _ _ ____ _
3 * Project ___| | | | _ \| |
4 * / __| | | | |_) | |
5 * | (__| |_| | _ <| |___
6 * \___|\___/|_| \_\_____|
7 *
8 * Copyright (C) Daniel Stenberg, <daniel@haxx.se>, 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
file_setup_connection(struct Curl_easy * data,struct connectdata * conn)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 */
file_connect(struct Curl_easy * data,bool * done)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 CURLcode result;
154
155 if(file->path) {
156 /* already connected.
157 * the handler->connect_it() is normally only called once, but
158 * FILE does a special check on setting up the connection which
159 * calls this explicitly. */
160 *done = TRUE;
161 return CURLE_OK;
162 }
163
164 result = Curl_urldecode(data->state.up.path, 0, &real_path,
165 &real_path_len, REJECT_ZERO);
166 if(result)
167 return result;
168
169 #ifdef DOS_FILESYSTEM
170 /* If the first character is a slash, and there's
171 something that looks like a drive at the beginning of
172 the path, skip the slash. If we remove the initial
173 slash in all cases, paths without drive letters end up
174 relative to the current directory which isn't how
175 browsers work.
176
177 Some browsers accept | instead of : as the drive letter
178 separator, so we do too.
179
180 On other platforms, we need the slash to indicate an
181 absolute pathname. On Windows, absolute paths start
182 with a drive letter.
183 */
184 actual_path = real_path;
185 if((actual_path[0] == '/') &&
186 actual_path[1] &&
187 (actual_path[2] == ':' || actual_path[2] == '|')) {
188 actual_path[2] = ':';
189 actual_path++;
190 real_path_len--;
191 }
192
193 /* change path separators from '/' to '\\' for DOS, Windows and OS/2 */
194 for(i = 0; i < real_path_len; ++i)
195 if(actual_path[i] == '/')
196 actual_path[i] = '\\';
197 else if(!actual_path[i]) { /* binary zero */
198 Curl_safefree(real_path);
199 return CURLE_URL_MALFORMAT;
200 }
201
202 fd = open_readonly(actual_path, O_RDONLY|O_BINARY);
203 file->path = actual_path;
204 #else
205 if(memchr(real_path, 0, real_path_len)) {
206 /* binary zeroes indicate foul play */
207 Curl_safefree(real_path);
208 return CURLE_URL_MALFORMAT;
209 }
210
211 #ifdef AMIGA_FILESYSTEM
212 /*
213 * A leading slash in an AmigaDOS path denotes the parent
214 * directory, and hence we block this as it is relative.
215 * Absolute paths start with 'volumename:', so we check for
216 * this first. Failing that, we treat the path as a real unix
217 * path, but only if the application was compiled with -lunix.
218 */
219 fd = -1;
220 file->path = real_path;
221
222 if(real_path[0] == '/') {
223 extern int __unix_path_semantics;
224 if(strchr(real_path + 1, ':')) {
225 /* Amiga absolute path */
226 fd = open_readonly(real_path + 1, O_RDONLY);
227 file->path++;
228 }
229 else if(__unix_path_semantics) {
230 /* -lunix fallback */
231 fd = open_readonly(real_path, O_RDONLY);
232 }
233 }
234 #else
235 fd = open_readonly(real_path, O_RDONLY);
236 file->path = real_path;
237 #endif
238 #endif
239 Curl_safefree(file->freepath);
240 file->freepath = real_path; /* free this when done */
241
242 file->fd = fd;
243 if(!data->state.upload && (fd == -1)) {
244 failf(data, "Couldn't open file %s", data->state.up.path);
245 file_done(data, CURLE_FILE_COULDNT_READ_FILE, FALSE);
246 return CURLE_FILE_COULDNT_READ_FILE;
247 }
248 *done = TRUE;
249
250 return CURLE_OK;
251 }
252
file_done(struct Curl_easy * data,CURLcode status,bool premature)253 static CURLcode file_done(struct Curl_easy *data,
254 CURLcode status, bool premature)
255 {
256 struct FILEPROTO *file = data->req.p.file;
257 (void)status; /* not used */
258 (void)premature; /* not used */
259
260 if(file) {
261 Curl_safefree(file->freepath);
262 file->path = NULL;
263 if(file->fd != -1)
264 close(file->fd);
265 file->fd = -1;
266 }
267
268 return CURLE_OK;
269 }
270
file_disconnect(struct Curl_easy * data,struct connectdata * conn,bool dead_connection)271 static CURLcode file_disconnect(struct Curl_easy *data,
272 struct connectdata *conn,
273 bool dead_connection)
274 {
275 (void)dead_connection; /* not used */
276 (void)conn;
277 return file_done(data, CURLE_OK, FALSE);
278 }
279
280 #ifdef DOS_FILESYSTEM
281 #define DIRSEP '\\'
282 #else
283 #define DIRSEP '/'
284 #endif
285
file_upload(struct Curl_easy * data)286 static CURLcode file_upload(struct Curl_easy *data)
287 {
288 struct FILEPROTO *file = data->req.p.file;
289 const char *dir = strchr(file->path, DIRSEP);
290 int fd;
291 int mode;
292 CURLcode result = CURLE_OK;
293 char *buf = data->state.buffer;
294 curl_off_t bytecount = 0;
295 struct_stat file_stat;
296 const char *buf2;
297
298 /*
299 * Since FILE: doesn't do the full init, we need to provide some extra
300 * assignments here.
301 */
302 data->req.upload_fromhere = buf;
303
304 if(!dir)
305 return CURLE_FILE_COULDNT_READ_FILE; /* fix: better error code */
306
307 if(!dir[1])
308 return CURLE_FILE_COULDNT_READ_FILE; /* fix: better error code */
309
310 #ifdef O_BINARY
311 #define MODE_DEFAULT O_WRONLY|O_CREAT|O_BINARY
312 #else
313 #define MODE_DEFAULT O_WRONLY|O_CREAT
314 #endif
315
316 if(data->state.resume_from)
317 mode = MODE_DEFAULT|O_APPEND;
318 else
319 mode = MODE_DEFAULT|O_TRUNC;
320
321 fd = open(file->path, mode, data->set.new_file_perms);
322 if(fd < 0) {
323 failf(data, "Can't open %s for writing", file->path);
324 return CURLE_WRITE_ERROR;
325 }
326
327 if(-1 != data->state.infilesize)
328 /* known size of data to "upload" */
329 Curl_pgrsSetUploadSize(data, data->state.infilesize);
330
331 /* treat the negative resume offset value as the case of "-" */
332 if(data->state.resume_from < 0) {
333 if(fstat(fd, &file_stat)) {
334 close(fd);
335 failf(data, "Can't get the size of %s", file->path);
336 return CURLE_WRITE_ERROR;
337 }
338 data->state.resume_from = (curl_off_t)file_stat.st_size;
339 }
340
341 while(!result) {
342 size_t nread;
343 ssize_t nwrite;
344 size_t readcount;
345 result = Curl_fillreadbuffer(data, data->set.buffer_size, &readcount);
346 if(result)
347 break;
348
349 if(!readcount)
350 break;
351
352 nread = readcount;
353
354 /* skip bytes before resume point */
355 if(data->state.resume_from) {
356 if((curl_off_t)nread <= data->state.resume_from) {
357 data->state.resume_from -= nread;
358 nread = 0;
359 buf2 = buf;
360 }
361 else {
362 buf2 = buf + data->state.resume_from;
363 nread -= (size_t)data->state.resume_from;
364 data->state.resume_from = 0;
365 }
366 }
367 else
368 buf2 = buf;
369
370 /* write the data to the target */
371 nwrite = write(fd, buf2, nread);
372 if((size_t)nwrite != nread) {
373 result = CURLE_SEND_ERROR;
374 break;
375 }
376
377 bytecount += nread;
378
379 Curl_pgrsSetUploadCounter(data, bytecount);
380
381 if(Curl_pgrsUpdate(data))
382 result = CURLE_ABORTED_BY_CALLBACK;
383 else
384 result = Curl_speedcheck(data, Curl_now());
385 }
386 if(!result && Curl_pgrsUpdate(data))
387 result = CURLE_ABORTED_BY_CALLBACK;
388
389 close(fd);
390
391 return result;
392 }
393
394 /*
395 * file_do() is the protocol-specific function for the do-phase, separated
396 * from the connect-phase above. Other protocols merely setup the transfer in
397 * the do-phase, to have it done in the main transfer loop but since some
398 * platforms we support don't allow select()ing etc on file handles (as
399 * opposed to sockets) we instead perform the whole do-operation in this
400 * function.
401 */
file_do(struct Curl_easy * data,bool * done)402 static CURLcode file_do(struct Curl_easy *data, bool *done)
403 {
404 /* This implementation ignores the host name in conformance with
405 RFC 1738. Only local files (reachable via the standard file system)
406 are supported. This means that files on remotely mounted directories
407 (via NFS, Samba, NT sharing) can be accessed through a file:// URL
408 */
409 CURLcode result = CURLE_OK;
410 struct_stat statbuf; /* struct_stat instead of struct stat just to allow the
411 Windows version to have a different struct without
412 having to redefine the simple word 'stat' */
413 curl_off_t expected_size = -1;
414 bool size_known;
415 bool fstated = FALSE;
416 char *buf = data->state.buffer;
417 curl_off_t bytecount = 0;
418 int fd;
419 struct FILEPROTO *file;
420
421 *done = TRUE; /* unconditionally */
422
423 Curl_pgrsStartNow(data);
424
425 if(data->state.upload)
426 return file_upload(data);
427
428 file = data->req.p.file;
429
430 /* get the fd from the connection phase */
431 fd = file->fd;
432
433 /* VMS: This only works reliable for STREAMLF files */
434 if(-1 != fstat(fd, &statbuf)) {
435 if(!S_ISDIR(statbuf.st_mode))
436 expected_size = statbuf.st_size;
437 /* and store the modification time */
438 data->info.filetime = statbuf.st_mtime;
439 fstated = TRUE;
440 }
441
442 if(fstated && !data->state.range && data->set.timecondition) {
443 if(!Curl_meets_timecondition(data, data->info.filetime)) {
444 *done = TRUE;
445 return CURLE_OK;
446 }
447 }
448
449 if(fstated) {
450 time_t filetime;
451 struct tm buffer;
452 const struct tm *tm = &buffer;
453 char header[80];
454 int headerlen;
455 char accept_ranges[24]= { "Accept-ranges: bytes\r\n" };
456 if(expected_size >= 0) {
457 headerlen = msnprintf(header, sizeof(header),
458 "Content-Length: %" CURL_FORMAT_CURL_OFF_T "\r\n",
459 expected_size);
460 result = Curl_client_write(data, CLIENTWRITE_HEADER, header, headerlen);
461 if(result)
462 return result;
463
464 result = Curl_client_write(data, CLIENTWRITE_HEADER,
465 accept_ranges, strlen(accept_ranges));
466 if(result != CURLE_OK)
467 return result;
468 }
469
470 filetime = (time_t)statbuf.st_mtime;
471 result = Curl_gmtime(filetime, &buffer);
472 if(result)
473 return result;
474
475 /* format: "Tue, 15 Nov 1994 12:45:26 GMT" */
476 headerlen = msnprintf(header, sizeof(header),
477 "Last-Modified: %s, %02d %s %4d %02d:%02d:%02d GMT\r\n%s",
478 Curl_wkday[tm->tm_wday?tm->tm_wday-1:6],
479 tm->tm_mday,
480 Curl_month[tm->tm_mon],
481 tm->tm_year + 1900,
482 tm->tm_hour,
483 tm->tm_min,
484 tm->tm_sec,
485 data->req.no_body ? "": "\r\n");
486 result = Curl_client_write(data, CLIENTWRITE_HEADER, header, headerlen);
487 if(result)
488 return result;
489 /* set the file size to make it available post transfer */
490 Curl_pgrsSetDownloadSize(data, expected_size);
491 if(data->req.no_body)
492 return result;
493 }
494
495 /* Check whether file range has been specified */
496 result = Curl_range(data);
497 if(result)
498 return result;
499
500 /* Adjust the start offset in case we want to get the N last bytes
501 * of the stream if the filesize could be determined */
502 if(data->state.resume_from < 0) {
503 if(!fstated) {
504 failf(data, "Can't get the size of file.");
505 return CURLE_READ_ERROR;
506 }
507 data->state.resume_from += (curl_off_t)statbuf.st_size;
508 }
509
510 if(data->state.resume_from > 0) {
511 /* We check explicitly if we have a start offset, because
512 * expected_size may be -1 if we don't know how large the file is,
513 * in which case we should not adjust it. */
514 if(data->state.resume_from <= expected_size)
515 expected_size -= data->state.resume_from;
516 else {
517 failf(data, "failed to resume file:// transfer");
518 return CURLE_BAD_DOWNLOAD_RESUME;
519 }
520 }
521
522 /* A high water mark has been specified so we obey... */
523 if(data->req.maxdownload > 0)
524 expected_size = data->req.maxdownload;
525
526 if(!fstated || (expected_size <= 0))
527 size_known = FALSE;
528 else
529 size_known = TRUE;
530
531 /* The following is a shortcut implementation of file reading
532 this is both more efficient than the former call to download() and
533 it avoids problems with select() and recv() on file descriptors
534 in Winsock */
535 if(size_known)
536 Curl_pgrsSetDownloadSize(data, expected_size);
537
538 if(data->state.resume_from) {
539 if(data->state.resume_from !=
540 lseek(fd, data->state.resume_from, SEEK_SET))
541 return CURLE_BAD_DOWNLOAD_RESUME;
542 }
543
544 Curl_pgrsTime(data, TIMER_STARTTRANSFER);
545
546 while(!result) {
547 ssize_t nread;
548 /* Don't fill a whole buffer if we want less than all data */
549 size_t bytestoread;
550
551 if(size_known) {
552 bytestoread = (expected_size < data->set.buffer_size) ?
553 curlx_sotouz(expected_size) : (size_t)data->set.buffer_size;
554 }
555 else
556 bytestoread = data->set.buffer_size-1;
557
558 nread = read(fd, buf, bytestoread);
559
560 if(nread > 0)
561 buf[nread] = 0;
562
563 if(nread <= 0 || (size_known && (expected_size == 0)))
564 break;
565
566 bytecount += nread;
567 if(size_known)
568 expected_size -= nread;
569
570 result = Curl_client_write(data, CLIENTWRITE_BODY, buf, nread);
571 if(result)
572 return result;
573
574 result = Curl_pgrsSetDownloadCounter(data, bytecount);
575 if(result)
576 return result;
577
578 if(Curl_pgrsUpdate(data))
579 result = CURLE_ABORTED_BY_CALLBACK;
580 else
581 result = Curl_speedcheck(data, Curl_now());
582 }
583 if(Curl_pgrsUpdate(data))
584 result = CURLE_ABORTED_BY_CALLBACK;
585
586 return result;
587 }
588
589 #endif
590