source: mainline/uspace/srv/vfs/vfs_ops.c@ a63ff7d

lfn serial ticket/834-toolchain-update topic/msim-upgrade topic/simplify-dev-export
Last change on this file since a63ff7d was 49d871ea, checked in by Jakub Jermar <jakub@…>, 15 years ago

Merge from lp:~jakub/helenos/fs.
  • Property mode set to 100644
File size: 34.5 KB
Line 
1/*
2 * Copyright (c) 2008 Jakub Jermar
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 *
9 * - Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * - Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * - The name of the author may not be used to endorse or promote products
15 * derived from this software without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 */
28
29/** @addtogroup fs
30 * @{
31 */
32
33/**
34 * @file vfs_ops.c
35 * @brief Operations that VFS offers to its clients.
36 */
37
38#include "vfs.h"
39#include <ipc/ipc.h>
40#include <macros.h>
41#include <stdint.h>
42#include <async.h>
43#include <errno.h>
44#include <stdio.h>
45#include <stdlib.h>
46#include <str.h>
47#include <bool.h>
48#include <fibril_synch.h>
49#include <adt/list.h>
50#include <unistd.h>
51#include <ctype.h>
52#include <fcntl.h>
53#include <assert.h>
54#include <vfs/canonify.h>
55
56/* Forward declarations of static functions. */
57static int vfs_truncate_internal(fs_handle_t, dev_handle_t, fs_index_t, aoff64_t);
58
59/**
60 * This rwlock prevents the race between a triplet-to-VFS-node resolution and a
61 * concurrent VFS operation which modifies the file system namespace.
62 */
63FIBRIL_RWLOCK_INITIALIZE(namespace_rwlock);
64
65vfs_pair_t rootfs = {
66 .fs_handle = 0,
67 .dev_handle = 0
68};
69
70static void vfs_mount_internal(ipc_callid_t rid, dev_handle_t dev_handle,
71 fs_handle_t fs_handle, char *mp, char *opts)
72{
73 vfs_lookup_res_t mp_res;
74 vfs_lookup_res_t mr_res;
75 vfs_node_t *mp_node = NULL;
76 vfs_node_t *mr_node;
77 fs_index_t rindex;
78 size_t rsize;
79 unsigned rlnkcnt;
80 ipcarg_t rc;
81 int phone;
82 aid_t msg;
83 ipc_call_t answer;
84
85 /* Resolve the path to the mountpoint. */
86 fibril_rwlock_write_lock(&namespace_rwlock);
87 if (rootfs.fs_handle) {
88 /* We already have the root FS. */
89 if (str_cmp(mp, "/") == 0) {
90 /* Trying to mount root FS over root FS */
91 fibril_rwlock_write_unlock(&namespace_rwlock);
92 ipc_answer_0(rid, EBUSY);
93 return;
94 }
95
96 rc = vfs_lookup_internal(mp, L_MP, &mp_res, NULL);
97 if (rc != EOK) {
98 /* The lookup failed for some reason. */
99 fibril_rwlock_write_unlock(&namespace_rwlock);
100 ipc_answer_0(rid, rc);
101 return;
102 }
103
104 mp_node = vfs_node_get(&mp_res);
105 if (!mp_node) {
106 fibril_rwlock_write_unlock(&namespace_rwlock);
107 ipc_answer_0(rid, ENOMEM);
108 return;
109 }
110
111 /*
112 * Now we hold a reference to mp_node.
113 * It will be dropped upon the corresponding VFS_IN_UNMOUNT.
114 * This prevents the mount point from being deleted.
115 */
116 } else {
117 /* We still don't have the root file system mounted. */
118 if (str_cmp(mp, "/") == 0) {
119 /*
120 * For this simple, but important case,
121 * we are almost done.
122 */
123
124 /* Tell the mountee that it is being mounted. */
125 phone = vfs_grab_phone(fs_handle);
126 msg = async_send_1(phone, VFS_OUT_MOUNTED,
127 (ipcarg_t) dev_handle, &answer);
128 /* send the mount options */
129 rc = async_data_write_start(phone, (void *)opts,
130 str_size(opts));
131 if (rc != EOK) {
132 async_wait_for(msg, NULL);
133 vfs_release_phone(fs_handle, phone);
134 fibril_rwlock_write_unlock(&namespace_rwlock);
135 ipc_answer_0(rid, rc);
136 return;
137 }
138 async_wait_for(msg, &rc);
139 vfs_release_phone(fs_handle, phone);
140
141 if (rc != EOK) {
142 fibril_rwlock_write_unlock(&namespace_rwlock);
143 ipc_answer_0(rid, rc);
144 return;
145 }
146
147 rindex = (fs_index_t) IPC_GET_ARG1(answer);
148 rsize = (size_t) IPC_GET_ARG2(answer);
149 rlnkcnt = (unsigned) IPC_GET_ARG3(answer);
150
151 mr_res.triplet.fs_handle = fs_handle;
152 mr_res.triplet.dev_handle = dev_handle;
153 mr_res.triplet.index = rindex;
154 mr_res.size = rsize;
155 mr_res.lnkcnt = rlnkcnt;
156 mr_res.type = VFS_NODE_DIRECTORY;
157
158 rootfs.fs_handle = fs_handle;
159 rootfs.dev_handle = dev_handle;
160
161 /* Add reference to the mounted root. */
162 mr_node = vfs_node_get(&mr_res);
163 assert(mr_node);
164
165 fibril_rwlock_write_unlock(&namespace_rwlock);
166 ipc_answer_0(rid, rc);
167 return;
168 } else {
169 /*
170 * We can't resolve this without the root filesystem
171 * being mounted first.
172 */
173 fibril_rwlock_write_unlock(&namespace_rwlock);
174 ipc_answer_0(rid, ENOENT);
175 return;
176 }
177 }
178
179 /*
180 * At this point, we have all necessary pieces: file system and device
181 * handles, and we know the mount point VFS node.
182 */
183
184 int mountee_phone = vfs_grab_phone(fs_handle);
185 assert(mountee_phone >= 0);
186
187 phone = vfs_grab_phone(mp_res.triplet.fs_handle);
188 msg = async_send_4(phone, VFS_OUT_MOUNT,
189 (ipcarg_t) mp_res.triplet.dev_handle,
190 (ipcarg_t) mp_res.triplet.index,
191 (ipcarg_t) fs_handle,
192 (ipcarg_t) dev_handle, &answer);
193
194 /* send connection */
195 rc = async_req_1_0(phone, IPC_M_CONNECTION_CLONE, mountee_phone);
196 if (rc != EOK) {
197 async_wait_for(msg, NULL);
198 vfs_release_phone(fs_handle, mountee_phone);
199 vfs_release_phone(mp_res.triplet.fs_handle, phone);
200 /* Mount failed, drop reference to mp_node. */
201 if (mp_node)
202 vfs_node_put(mp_node);
203 ipc_answer_0(rid, rc);
204 fibril_rwlock_write_unlock(&namespace_rwlock);
205 return;
206 }
207
208 vfs_release_phone(fs_handle, mountee_phone);
209
210 /* send the mount options */
211 rc = async_data_write_start(phone, (void *)opts, str_size(opts));
212 if (rc != EOK) {
213 async_wait_for(msg, NULL);
214 vfs_release_phone(mp_res.triplet.fs_handle, phone);
215 /* Mount failed, drop reference to mp_node. */
216 if (mp_node)
217 vfs_node_put(mp_node);
218 fibril_rwlock_write_unlock(&namespace_rwlock);
219 ipc_answer_0(rid, rc);
220 return;
221 }
222 async_wait_for(msg, &rc);
223 vfs_release_phone(mp_res.triplet.fs_handle, phone);
224
225 if (rc == EOK) {
226 rindex = (fs_index_t) IPC_GET_ARG1(answer);
227 rsize = (size_t) IPC_GET_ARG2(answer);
228 rlnkcnt = (unsigned) IPC_GET_ARG3(answer);
229
230 mr_res.triplet.fs_handle = fs_handle;
231 mr_res.triplet.dev_handle = dev_handle;
232 mr_res.triplet.index = rindex;
233 mr_res.size = rsize;
234 mr_res.lnkcnt = rlnkcnt;
235 mr_res.type = VFS_NODE_DIRECTORY;
236
237 /* Add reference to the mounted root. */
238 mr_node = vfs_node_get(&mr_res);
239 assert(mr_node);
240 } else {
241 /* Mount failed, drop reference to mp_node. */
242 if (mp_node)
243 vfs_node_put(mp_node);
244 }
245
246 ipc_answer_0(rid, rc);
247 fibril_rwlock_write_unlock(&namespace_rwlock);
248}
249
250void vfs_mount(ipc_callid_t rid, ipc_call_t *request)
251{
252 /*
253 * We expect the library to do the device-name to device-handle
254 * translation for us, thus the device handle will arrive as ARG1
255 * in the request.
256 */
257 dev_handle_t dev_handle = (dev_handle_t) IPC_GET_ARG1(*request);
258
259 /*
260 * Mount flags are passed as ARG2.
261 */
262 unsigned int flags = (unsigned int) IPC_GET_ARG2(*request);
263
264 /*
265 * For now, don't make use of ARG3, but it can be used to
266 * carry mount options in the future.
267 */
268
269 /* We want the client to send us the mount point. */
270 char *mp;
271 int rc = async_data_write_accept((void **) &mp, true, 0, MAX_PATH_LEN,
272 0, NULL);
273 if (rc != EOK) {
274 ipc_answer_0(rid, rc);
275 return;
276 }
277
278 /* Now we expect to receive the mount options. */
279 char *opts;
280 rc = async_data_write_accept((void **) &opts, true, 0, MAX_MNTOPTS_LEN,
281 0, NULL);
282 if (rc != EOK) {
283 free(mp);
284 ipc_answer_0(rid, rc);
285 return;
286 }
287
288 /*
289 * Now, we expect the client to send us data with the name of the file
290 * system.
291 */
292 char *fs_name;
293 rc = async_data_write_accept((void **) &fs_name, true, 0, FS_NAME_MAXLEN,
294 0, NULL);
295 if (rc != EOK) {
296 free(mp);
297 free(opts);
298 ipc_answer_0(rid, rc);
299 return;
300 }
301
302 /*
303 * Wait for IPC_M_PING so that we can return an error if we don't know
304 * fs_name.
305 */
306 ipc_call_t data;
307 ipc_callid_t callid = async_get_call(&data);
308 if (IPC_GET_METHOD(data) != IPC_M_PING) {
309 ipc_answer_0(callid, ENOTSUP);
310 ipc_answer_0(rid, ENOTSUP);
311 free(mp);
312 free(opts);
313 free(fs_name);
314 return;
315 }
316
317 /*
318 * Check if we know a file system with the same name as is in fs_name.
319 * This will also give us its file system handle.
320 */
321 fibril_mutex_lock(&fs_head_lock);
322 fs_handle_t fs_handle;
323recheck:
324 fs_handle = fs_name_to_handle(fs_name, false);
325 if (!fs_handle) {
326 if (flags & IPC_FLAG_BLOCKING) {
327 fibril_condvar_wait(&fs_head_cv, &fs_head_lock);
328 goto recheck;
329 }
330
331 fibril_mutex_unlock(&fs_head_lock);
332 ipc_answer_0(callid, ENOENT);
333 ipc_answer_0(rid, ENOENT);
334 free(mp);
335 free(fs_name);
336 free(opts);
337 return;
338 }
339 fibril_mutex_unlock(&fs_head_lock);
340
341 /* Acknowledge that we know fs_name. */
342 ipc_answer_0(callid, EOK);
343
344 /* Do the mount */
345 vfs_mount_internal(rid, dev_handle, fs_handle, mp, opts);
346 free(mp);
347 free(fs_name);
348 free(opts);
349}
350
351void vfs_unmount(ipc_callid_t rid, ipc_call_t *request)
352{
353 int rc;
354 char *mp;
355 vfs_lookup_res_t mp_res;
356 vfs_lookup_res_t mr_res;
357 vfs_node_t *mr_node;
358 int phone;
359
360 /*
361 * Receive the mount point path.
362 */
363 rc = async_data_write_accept((void **) &mp, true, 0, MAX_PATH_LEN,
364 0, NULL);
365 if (rc != EOK)
366 ipc_answer_0(rid, rc);
367
368 /*
369 * Taking the namespace lock will do two things for us. First, it will
370 * prevent races with other lookup operations. Second, it will stop new
371 * references to already existing VFS nodes and creation of new VFS
372 * nodes. This is because new references are added as a result of some
373 * lookup operation or at least of some operation which is protected by
374 * the namespace lock.
375 */
376 fibril_rwlock_write_lock(&namespace_rwlock);
377
378 /*
379 * Lookup the mounted root and instantiate it.
380 */
381 rc = vfs_lookup_internal(mp, L_ROOT, &mr_res, NULL);
382 if (rc != EOK) {
383 fibril_rwlock_write_unlock(&namespace_rwlock);
384 free(mp);
385 ipc_answer_0(rid, rc);
386 return;
387 }
388 mr_node = vfs_node_get(&mr_res);
389 if (!mr_node) {
390 fibril_rwlock_write_unlock(&namespace_rwlock);
391 free(mp);
392 ipc_answer_0(rid, ENOMEM);
393 return;
394 }
395
396 /*
397 * Count the total number of references for the mounted file system. We
398 * are expecting at least two. One which we got above and one which we
399 * got when the file system was mounted. If we find more, it means that
400 * the file system cannot be gracefully unmounted at the moment because
401 * someone is working with it.
402 */
403 if (vfs_nodes_refcount_sum_get(mr_node->fs_handle,
404 mr_node->dev_handle) != 2) {
405 fibril_rwlock_write_unlock(&namespace_rwlock);
406 vfs_node_put(mr_node);
407 free(mp);
408 ipc_answer_0(rid, EBUSY);
409 return;
410 }
411
412 if (str_cmp(mp, "/") == 0) {
413
414 /*
415 * Unmounting the root file system.
416 *
417 * In this case, there is no mount point node and we send
418 * VFS_OUT_UNMOUNTED directly to the mounted file system.
419 */
420
421 free(mp);
422 phone = vfs_grab_phone(mr_node->fs_handle);
423 rc = async_req_1_0(phone, VFS_OUT_UNMOUNTED,
424 mr_node->dev_handle);
425 vfs_release_phone(mr_node->fs_handle, phone);
426 if (rc != EOK) {
427 fibril_rwlock_write_unlock(&namespace_rwlock);
428 vfs_node_put(mr_node);
429 ipc_answer_0(rid, rc);
430 return;
431 }
432 rootfs.fs_handle = 0;
433 rootfs.dev_handle = 0;
434 } else {
435
436 /*
437 * Unmounting a non-root file system.
438 *
439 * We have a regular mount point node representing the parent
440 * file system, so we delegate the operation to it.
441 */
442
443 rc = vfs_lookup_internal(mp, L_MP, &mp_res, NULL);
444 free(mp);
445 if (rc != EOK) {
446 fibril_rwlock_write_unlock(&namespace_rwlock);
447 vfs_node_put(mr_node);
448 ipc_answer_0(rid, rc);
449 return;
450 }
451 vfs_node_t *mp_node = vfs_node_get(&mp_res);
452 if (!mp_node) {
453 fibril_rwlock_write_unlock(&namespace_rwlock);
454 vfs_node_put(mr_node);
455 ipc_answer_0(rid, ENOMEM);
456 return;
457 }
458
459 phone = vfs_grab_phone(mp_node->fs_handle);
460 rc = async_req_2_0(phone, VFS_OUT_UNMOUNT, mp_node->dev_handle,
461 mp_node->index);
462 vfs_release_phone(mp_node->fs_handle, phone);
463 if (rc != EOK) {
464 fibril_rwlock_write_unlock(&namespace_rwlock);
465 vfs_node_put(mp_node);
466 vfs_node_put(mr_node);
467 ipc_answer_0(rid, rc);
468 return;
469 }
470
471 /* Drop the reference we got above. */
472 vfs_node_put(mp_node);
473 /* Drop the reference from when the file system was mounted. */
474 vfs_node_put(mp_node);
475 }
476
477
478 /*
479 * All went well, the mounted file system was successfully unmounted.
480 * The only thing left is to forget the unmounted root VFS node.
481 */
482 vfs_node_forget(mr_node);
483
484 fibril_rwlock_write_unlock(&namespace_rwlock);
485 ipc_answer_0(rid, EOK);
486}
487
488void vfs_open(ipc_callid_t rid, ipc_call_t *request)
489{
490 if (!vfs_files_init()) {
491 ipc_answer_0(rid, ENOMEM);
492 return;
493 }
494
495 /*
496 * The POSIX interface is open(path, oflag, mode).
497 * We can receive oflags and mode along with the VFS_IN_OPEN call;
498 * the path will need to arrive in another call.
499 *
500 * We also receive one private, non-POSIX set of flags called lflag
501 * used to pass information to vfs_lookup_internal().
502 */
503 int lflag = IPC_GET_ARG1(*request);
504 int oflag = IPC_GET_ARG2(*request);
505 int mode = IPC_GET_ARG3(*request);
506
507 /* Ignore mode for now. */
508 (void) mode;
509
510 /*
511 * Make sure that we are called with exactly one of L_FILE and
512 * L_DIRECTORY. Make sure that the user does not pass L_OPEN,
513 * L_ROOT or L_MP.
514 */
515 if (((lflag & (L_FILE | L_DIRECTORY)) == 0) ||
516 ((lflag & (L_FILE | L_DIRECTORY)) == (L_FILE | L_DIRECTORY)) ||
517 (lflag & (L_OPEN | L_ROOT | L_MP))) {
518 ipc_answer_0(rid, EINVAL);
519 return;
520 }
521
522 if (oflag & O_CREAT)
523 lflag |= L_CREATE;
524 if (oflag & O_EXCL)
525 lflag |= L_EXCLUSIVE;
526
527 char *path;
528 int rc = async_data_write_accept((void **) &path, true, 0, 0, 0, NULL);
529 if (rc != EOK) {
530 ipc_answer_0(rid, rc);
531 return;
532 }
533
534 /*
535 * Avoid the race condition in which the file can be deleted before we
536 * find/create-and-lock the VFS node corresponding to the looked-up
537 * triplet.
538 */
539 if (lflag & L_CREATE)
540 fibril_rwlock_write_lock(&namespace_rwlock);
541 else
542 fibril_rwlock_read_lock(&namespace_rwlock);
543
544 /* The path is now populated and we can call vfs_lookup_internal(). */
545 vfs_lookup_res_t lr;
546 rc = vfs_lookup_internal(path, lflag | L_OPEN, &lr, NULL);
547 if (rc != EOK) {
548 if (lflag & L_CREATE)
549 fibril_rwlock_write_unlock(&namespace_rwlock);
550 else
551 fibril_rwlock_read_unlock(&namespace_rwlock);
552 ipc_answer_0(rid, rc);
553 free(path);
554 return;
555 }
556
557 /* Path is no longer needed. */
558 free(path);
559
560 vfs_node_t *node = vfs_node_get(&lr);
561 if (lflag & L_CREATE)
562 fibril_rwlock_write_unlock(&namespace_rwlock);
563 else
564 fibril_rwlock_read_unlock(&namespace_rwlock);
565
566 /* Truncate the file if requested and if necessary. */
567 if (oflag & O_TRUNC) {
568 fibril_rwlock_write_lock(&node->contents_rwlock);
569 if (node->size) {
570 rc = vfs_truncate_internal(node->fs_handle,
571 node->dev_handle, node->index, 0);
572 if (rc) {
573 fibril_rwlock_write_unlock(&node->contents_rwlock);
574 vfs_node_put(node);
575 ipc_answer_0(rid, rc);
576 return;
577 }
578 node->size = 0;
579 }
580 fibril_rwlock_write_unlock(&node->contents_rwlock);
581 }
582
583 /*
584 * Get ourselves a file descriptor and the corresponding vfs_file_t
585 * structure.
586 */
587 int fd = vfs_fd_alloc((oflag & O_DESC) != 0);
588 if (fd < 0) {
589 vfs_node_put(node);
590 ipc_answer_0(rid, fd);
591 return;
592 }
593 vfs_file_t *file = vfs_file_get(fd);
594 assert(file);
595 file->node = node;
596 if (oflag & O_APPEND)
597 file->append = true;
598
599 /*
600 * The following increase in reference count is for the fact that the
601 * file is being opened and that a file structure is pointing to it.
602 * It is necessary so that the file will not disappear when
603 * vfs_node_put() is called. The reference will be dropped by the
604 * respective VFS_IN_CLOSE.
605 */
606 vfs_node_addref(node);
607 vfs_node_put(node);
608
609 /* Success! Return the new file descriptor to the client. */
610 ipc_answer_1(rid, EOK, fd);
611}
612
613void vfs_open_node(ipc_callid_t rid, ipc_call_t *request)
614{
615 // FIXME: check for sanity of the supplied fs, dev and index
616
617 if (!vfs_files_init()) {
618 ipc_answer_0(rid, ENOMEM);
619 return;
620 }
621
622 /*
623 * The interface is open_node(fs, dev, index, oflag).
624 */
625 vfs_lookup_res_t lr;
626
627 lr.triplet.fs_handle = IPC_GET_ARG1(*request);
628 lr.triplet.dev_handle = IPC_GET_ARG2(*request);
629 lr.triplet.index = IPC_GET_ARG3(*request);
630 int oflag = IPC_GET_ARG4(*request);
631
632 fibril_rwlock_read_lock(&namespace_rwlock);
633
634 int rc = vfs_open_node_internal(&lr);
635 if (rc != EOK) {
636 fibril_rwlock_read_unlock(&namespace_rwlock);
637 ipc_answer_0(rid, rc);
638 return;
639 }
640
641 vfs_node_t *node = vfs_node_get(&lr);
642 fibril_rwlock_read_unlock(&namespace_rwlock);
643
644 /* Truncate the file if requested and if necessary. */
645 if (oflag & O_TRUNC) {
646 fibril_rwlock_write_lock(&node->contents_rwlock);
647 if (node->size) {
648 rc = vfs_truncate_internal(node->fs_handle,
649 node->dev_handle, node->index, 0);
650 if (rc) {
651 fibril_rwlock_write_unlock(&node->contents_rwlock);
652 vfs_node_put(node);
653 ipc_answer_0(rid, rc);
654 return;
655 }
656 node->size = 0;
657 }
658 fibril_rwlock_write_unlock(&node->contents_rwlock);
659 }
660
661 /*
662 * Get ourselves a file descriptor and the corresponding vfs_file_t
663 * structure.
664 */
665 int fd = vfs_fd_alloc((oflag & O_DESC) != 0);
666 if (fd < 0) {
667 vfs_node_put(node);
668 ipc_answer_0(rid, fd);
669 return;
670 }
671 vfs_file_t *file = vfs_file_get(fd);
672 file->node = node;
673 if (oflag & O_APPEND)
674 file->append = true;
675
676 /*
677 * The following increase in reference count is for the fact that the
678 * file is being opened and that a file structure is pointing to it.
679 * It is necessary so that the file will not disappear when
680 * vfs_node_put() is called. The reference will be dropped by the
681 * respective VFS_IN_CLOSE.
682 */
683 vfs_node_addref(node);
684 vfs_node_put(node);
685
686 /* Success! Return the new file descriptor to the client. */
687 ipc_answer_1(rid, EOK, fd);
688}
689
690void vfs_sync(ipc_callid_t rid, ipc_call_t *request)
691{
692 int fd = IPC_GET_ARG1(*request);
693
694 /* Lookup the file structure corresponding to the file descriptor. */
695 vfs_file_t *file = vfs_file_get(fd);
696 if (!file) {
697 ipc_answer_0(rid, ENOENT);
698 return;
699 }
700
701 /*
702 * Lock the open file structure so that no other thread can manipulate
703 * the same open file at a time.
704 */
705 fibril_mutex_lock(&file->lock);
706 int fs_phone = vfs_grab_phone(file->node->fs_handle);
707
708 /* Make a VFS_OUT_SYMC request at the destination FS server. */
709 aid_t msg;
710 ipc_call_t answer;
711 msg = async_send_2(fs_phone, VFS_OUT_SYNC, file->node->dev_handle,
712 file->node->index, &answer);
713
714 /* Wait for reply from the FS server. */
715 ipcarg_t rc;
716 async_wait_for(msg, &rc);
717
718 vfs_release_phone(file->node->fs_handle, fs_phone);
719 fibril_mutex_unlock(&file->lock);
720
721 ipc_answer_0(rid, rc);
722}
723
724int vfs_close_internal(vfs_file_t *file)
725{
726 /*
727 * Lock the open file structure so that no other thread can manipulate
728 * the same open file at a time.
729 */
730 fibril_mutex_lock(&file->lock);
731
732 if (file->refcnt <= 1) {
733 /* Only close the file on the destination FS server
734 if there are no more file descriptors (except the
735 present one) pointing to this file. */
736
737 int fs_phone = vfs_grab_phone(file->node->fs_handle);
738
739 /* Make a VFS_OUT_CLOSE request at the destination FS server. */
740 aid_t msg;
741 ipc_call_t answer;
742 msg = async_send_2(fs_phone, VFS_OUT_CLOSE, file->node->dev_handle,
743 file->node->index, &answer);
744
745 /* Wait for reply from the FS server. */
746 ipcarg_t rc;
747 async_wait_for(msg, &rc);
748
749 vfs_release_phone(file->node->fs_handle, fs_phone);
750 fibril_mutex_unlock(&file->lock);
751
752 return IPC_GET_ARG1(answer);
753 }
754
755 fibril_mutex_unlock(&file->lock);
756 return EOK;
757}
758
759void vfs_close(ipc_callid_t rid, ipc_call_t *request)
760{
761 int fd = IPC_GET_ARG1(*request);
762
763 /* Lookup the file structure corresponding to the file descriptor. */
764 vfs_file_t *file = vfs_file_get(fd);
765 if (!file) {
766 ipc_answer_0(rid, ENOENT);
767 return;
768 }
769
770 int ret = vfs_close_internal(file);
771 if (ret != EOK)
772 ipc_answer_0(rid, ret);
773
774 ret = vfs_fd_free(fd);
775 ipc_answer_0(rid, ret);
776}
777
778static void vfs_rdwr(ipc_callid_t rid, ipc_call_t *request, bool read)
779{
780
781 /*
782 * The following code strongly depends on the fact that the files data
783 * structure can be only accessed by a single fibril and all file
784 * operations are serialized (i.e. the reads and writes cannot
785 * interleave and a file cannot be closed while it is being read).
786 *
787 * Additional synchronization needs to be added once the table of
788 * open files supports parallel access!
789 */
790
791 int fd = IPC_GET_ARG1(*request);
792
793 /* Lookup the file structure corresponding to the file descriptor. */
794 vfs_file_t *file = vfs_file_get(fd);
795 if (!file) {
796 ipc_answer_0(rid, ENOENT);
797 return;
798 }
799
800 /*
801 * Lock the open file structure so that no other thread can manipulate
802 * the same open file at a time.
803 */
804 fibril_mutex_lock(&file->lock);
805
806 /*
807 * Lock the file's node so that no other client can read/write to it at
808 * the same time.
809 */
810 if (read)
811 fibril_rwlock_read_lock(&file->node->contents_rwlock);
812 else
813 fibril_rwlock_write_lock(&file->node->contents_rwlock);
814
815 if (file->node->type == VFS_NODE_DIRECTORY) {
816 /*
817 * Make sure that no one is modifying the namespace
818 * while we are in readdir().
819 */
820 assert(read);
821 fibril_rwlock_read_lock(&namespace_rwlock);
822 }
823
824 int fs_phone = vfs_grab_phone(file->node->fs_handle);
825
826 /*
827 * Make a VFS_READ/VFS_WRITE request at the destination FS server
828 * and forward the IPC_M_DATA_READ/IPC_M_DATA_WRITE request to the
829 * destination FS server. The call will be routed as if sent by
830 * ourselves. Note that call arguments are immutable in this case so we
831 * don't have to bother.
832 */
833 ipcarg_t rc;
834 ipc_call_t answer;
835 if (read) {
836 if (file->append)
837 file->pos = file->node->size;
838
839 rc = async_data_read_forward_3_1(fs_phone, VFS_OUT_READ,
840 file->node->dev_handle, file->node->index, file->pos,
841 &answer);
842 } else {
843 rc = async_data_write_forward_3_1(fs_phone, VFS_OUT_WRITE,
844 file->node->dev_handle, file->node->index, file->pos,
845 &answer);
846 }
847
848 vfs_release_phone(file->node->fs_handle, fs_phone);
849
850 size_t bytes = IPC_GET_ARG1(answer);
851
852 if (file->node->type == VFS_NODE_DIRECTORY)
853 fibril_rwlock_read_unlock(&namespace_rwlock);
854
855 /* Unlock the VFS node. */
856 if (read)
857 fibril_rwlock_read_unlock(&file->node->contents_rwlock);
858 else {
859 /* Update the cached version of node's size. */
860 if (rc == EOK)
861 file->node->size = IPC_GET_ARG2(answer);
862 fibril_rwlock_write_unlock(&file->node->contents_rwlock);
863 }
864
865 /* Update the position pointer and unlock the open file. */
866 if (rc == EOK)
867 file->pos += bytes;
868 fibril_mutex_unlock(&file->lock);
869
870 /*
871 * FS server's reply is the final result of the whole operation we
872 * return to the client.
873 */
874 ipc_answer_1(rid, rc, bytes);
875}
876
877void vfs_read(ipc_callid_t rid, ipc_call_t *request)
878{
879 vfs_rdwr(rid, request, true);
880}
881
882void vfs_write(ipc_callid_t rid, ipc_call_t *request)
883{
884 vfs_rdwr(rid, request, false);
885}
886
887void vfs_seek(ipc_callid_t rid, ipc_call_t *request)
888{
889 int fd = (int) IPC_GET_ARG1(*request);
890 off64_t off =
891 (off64_t) MERGE_LOUP32(IPC_GET_ARG2(*request), IPC_GET_ARG3(*request));
892 int whence = (int) IPC_GET_ARG4(*request);
893
894 /* Lookup the file structure corresponding to the file descriptor. */
895 vfs_file_t *file = vfs_file_get(fd);
896 if (!file) {
897 ipc_answer_0(rid, ENOENT);
898 return;
899 }
900
901 fibril_mutex_lock(&file->lock);
902
903 off64_t newoff;
904 switch (whence) {
905 case SEEK_SET:
906 if (off >= 0) {
907 file->pos = (aoff64_t) off;
908 fibril_mutex_unlock(&file->lock);
909 ipc_answer_1(rid, EOK, off);
910 return;
911 }
912 break;
913 case SEEK_CUR:
914 if ((off >= 0) && (file->pos + off < file->pos)) {
915 fibril_mutex_unlock(&file->lock);
916 ipc_answer_0(rid, EOVERFLOW);
917 return;
918 }
919
920 if ((off < 0) && (file->pos < (aoff64_t) -off)) {
921 fibril_mutex_unlock(&file->lock);
922 ipc_answer_0(rid, EOVERFLOW);
923 return;
924 }
925
926 file->pos += off;
927 newoff = (file->pos > OFF64_MAX) ? OFF64_MAX : file->pos;
928
929 fibril_mutex_unlock(&file->lock);
930 ipc_answer_2(rid, EOK, LOWER32(newoff), UPPER32(newoff));
931 return;
932 case SEEK_END:
933 fibril_rwlock_read_lock(&file->node->contents_rwlock);
934 aoff64_t size = file->node->size;
935
936 if ((off >= 0) && (size + off < size)) {
937 fibril_rwlock_read_unlock(&file->node->contents_rwlock);
938 fibril_mutex_unlock(&file->lock);
939 ipc_answer_0(rid, EOVERFLOW);
940 return;
941 }
942
943 if ((off < 0) && (size < (aoff64_t) -off)) {
944 fibril_rwlock_read_unlock(&file->node->contents_rwlock);
945 fibril_mutex_unlock(&file->lock);
946 ipc_answer_0(rid, EOVERFLOW);
947 return;
948 }
949
950 file->pos = size + off;
951 newoff = (file->pos > OFF64_MAX) ? OFF64_MAX : file->pos;
952
953 fibril_rwlock_read_unlock(&file->node->contents_rwlock);
954 fibril_mutex_unlock(&file->lock);
955 ipc_answer_2(rid, EOK, LOWER32(newoff), UPPER32(newoff));
956 return;
957 }
958
959 fibril_mutex_unlock(&file->lock);
960 ipc_answer_0(rid, EINVAL);
961}
962
963int vfs_truncate_internal(fs_handle_t fs_handle, dev_handle_t dev_handle,
964 fs_index_t index, aoff64_t size)
965{
966 ipcarg_t rc;
967 int fs_phone;
968
969 fs_phone = vfs_grab_phone(fs_handle);
970 rc = async_req_4_0(fs_phone, VFS_OUT_TRUNCATE, (ipcarg_t) dev_handle,
971 (ipcarg_t) index, LOWER32(size), UPPER32(size));
972 vfs_release_phone(fs_handle, fs_phone);
973 return (int)rc;
974}
975
976void vfs_truncate(ipc_callid_t rid, ipc_call_t *request)
977{
978 int fd = IPC_GET_ARG1(*request);
979 aoff64_t size =
980 (aoff64_t) MERGE_LOUP32(IPC_GET_ARG2(*request), IPC_GET_ARG3(*request));
981 int rc;
982
983 vfs_file_t *file = vfs_file_get(fd);
984 if (!file) {
985 ipc_answer_0(rid, ENOENT);
986 return;
987 }
988 fibril_mutex_lock(&file->lock);
989
990 fibril_rwlock_write_lock(&file->node->contents_rwlock);
991 rc = vfs_truncate_internal(file->node->fs_handle,
992 file->node->dev_handle, file->node->index, size);
993 if (rc == EOK)
994 file->node->size = size;
995 fibril_rwlock_write_unlock(&file->node->contents_rwlock);
996
997 fibril_mutex_unlock(&file->lock);
998 ipc_answer_0(rid, (ipcarg_t)rc);
999}
1000
1001void vfs_fstat(ipc_callid_t rid, ipc_call_t *request)
1002{
1003 int fd = IPC_GET_ARG1(*request);
1004 ipcarg_t rc;
1005
1006 vfs_file_t *file = vfs_file_get(fd);
1007 if (!file) {
1008 ipc_answer_0(rid, ENOENT);
1009 return;
1010 }
1011
1012 ipc_callid_t callid;
1013 if (!async_data_read_receive(&callid, NULL)) {
1014 ipc_answer_0(callid, EINVAL);
1015 ipc_answer_0(rid, EINVAL);
1016 return;
1017 }
1018
1019 fibril_mutex_lock(&file->lock);
1020
1021 int fs_phone = vfs_grab_phone(file->node->fs_handle);
1022
1023 aid_t msg;
1024 msg = async_send_3(fs_phone, VFS_OUT_STAT, file->node->dev_handle,
1025 file->node->index, true, NULL);
1026 ipc_forward_fast(callid, fs_phone, 0, 0, 0, IPC_FF_ROUTE_FROM_ME);
1027 async_wait_for(msg, &rc);
1028 vfs_release_phone(file->node->fs_handle, fs_phone);
1029
1030 fibril_mutex_unlock(&file->lock);
1031 ipc_answer_0(rid, rc);
1032}
1033
1034void vfs_stat(ipc_callid_t rid, ipc_call_t *request)
1035{
1036 char *path;
1037 int rc = async_data_write_accept((void **) &path, true, 0, 0, 0, NULL);
1038 if (rc != EOK) {
1039 ipc_answer_0(rid, rc);
1040 return;
1041 }
1042
1043 ipc_callid_t callid;
1044 if (!async_data_read_receive(&callid, NULL)) {
1045 free(path);
1046 ipc_answer_0(callid, EINVAL);
1047 ipc_answer_0(rid, EINVAL);
1048 return;
1049 }
1050
1051 vfs_lookup_res_t lr;
1052 fibril_rwlock_read_lock(&namespace_rwlock);
1053 rc = vfs_lookup_internal(path, L_NONE, &lr, NULL);
1054 free(path);
1055 if (rc != EOK) {
1056 fibril_rwlock_read_unlock(&namespace_rwlock);
1057 ipc_answer_0(callid, rc);
1058 ipc_answer_0(rid, rc);
1059 return;
1060 }
1061 vfs_node_t *node = vfs_node_get(&lr);
1062 if (!node) {
1063 fibril_rwlock_read_unlock(&namespace_rwlock);
1064 ipc_answer_0(callid, ENOMEM);
1065 ipc_answer_0(rid, ENOMEM);
1066 return;
1067 }
1068
1069 fibril_rwlock_read_unlock(&namespace_rwlock);
1070
1071 int fs_phone = vfs_grab_phone(node->fs_handle);
1072 aid_t msg;
1073 msg = async_send_3(fs_phone, VFS_OUT_STAT, node->dev_handle,
1074 node->index, false, NULL);
1075 ipc_forward_fast(callid, fs_phone, 0, 0, 0, IPC_FF_ROUTE_FROM_ME);
1076
1077 ipcarg_t rv;
1078 async_wait_for(msg, &rv);
1079 vfs_release_phone(node->fs_handle, fs_phone);
1080
1081 ipc_answer_0(rid, rv);
1082
1083 vfs_node_put(node);
1084}
1085
1086void vfs_mkdir(ipc_callid_t rid, ipc_call_t *request)
1087{
1088 int mode = IPC_GET_ARG1(*request);
1089
1090 char *path;
1091 int rc = async_data_write_accept((void **) &path, true, 0, 0, 0, NULL);
1092 if (rc != EOK) {
1093 ipc_answer_0(rid, rc);
1094 return;
1095 }
1096
1097 /* Ignore mode for now. */
1098 (void) mode;
1099
1100 fibril_rwlock_write_lock(&namespace_rwlock);
1101 int lflag = L_DIRECTORY | L_CREATE | L_EXCLUSIVE;
1102 rc = vfs_lookup_internal(path, lflag, NULL, NULL);
1103 fibril_rwlock_write_unlock(&namespace_rwlock);
1104 free(path);
1105 ipc_answer_0(rid, rc);
1106}
1107
1108void vfs_unlink(ipc_callid_t rid, ipc_call_t *request)
1109{
1110 int lflag = IPC_GET_ARG1(*request);
1111
1112 char *path;
1113 int rc = async_data_write_accept((void **) &path, true, 0, 0, 0, NULL);
1114 if (rc != EOK) {
1115 ipc_answer_0(rid, rc);
1116 return;
1117 }
1118
1119 fibril_rwlock_write_lock(&namespace_rwlock);
1120 lflag &= L_DIRECTORY; /* sanitize lflag */
1121 vfs_lookup_res_t lr;
1122 rc = vfs_lookup_internal(path, lflag | L_UNLINK, &lr, NULL);
1123 free(path);
1124 if (rc != EOK) {
1125 fibril_rwlock_write_unlock(&namespace_rwlock);
1126 ipc_answer_0(rid, rc);
1127 return;
1128 }
1129
1130 /*
1131 * The name has already been unlinked by vfs_lookup_internal().
1132 * We have to get and put the VFS node to ensure that it is
1133 * VFS_OUT_DESTROY'ed after the last reference to it is dropped.
1134 */
1135 vfs_node_t *node = vfs_node_get(&lr);
1136 fibril_mutex_lock(&nodes_mutex);
1137 node->lnkcnt--;
1138 fibril_mutex_unlock(&nodes_mutex);
1139 fibril_rwlock_write_unlock(&namespace_rwlock);
1140 vfs_node_put(node);
1141 ipc_answer_0(rid, EOK);
1142}
1143
1144void vfs_rename(ipc_callid_t rid, ipc_call_t *request)
1145{
1146 /* Retrieve the old path. */
1147 char *old;
1148 int rc = async_data_write_accept((void **) &old, true, 0, 0, 0, NULL);
1149 if (rc != EOK) {
1150 ipc_answer_0(rid, rc);
1151 return;
1152 }
1153
1154 /* Retrieve the new path. */
1155 char *new;
1156 rc = async_data_write_accept((void **) &new, true, 0, 0, 0, NULL);
1157 if (rc != EOK) {
1158 free(old);
1159 ipc_answer_0(rid, rc);
1160 return;
1161 }
1162
1163 size_t olen;
1164 size_t nlen;
1165 char *oldc = canonify(old, &olen);
1166 char *newc = canonify(new, &nlen);
1167
1168 if ((!oldc) || (!newc)) {
1169 ipc_answer_0(rid, EINVAL);
1170 free(old);
1171 free(new);
1172 return;
1173 }
1174
1175 oldc[olen] = '\0';
1176 newc[nlen] = '\0';
1177
1178 if ((!str_lcmp(newc, oldc, str_length(oldc))) &&
1179 ((newc[str_length(oldc)] == '/') ||
1180 (str_length(oldc) == 1) ||
1181 (str_length(oldc) == str_length(newc)))) {
1182 /*
1183 * oldc is a prefix of newc and either
1184 * - newc continues with a / where oldc ends, or
1185 * - oldc was / itself, or
1186 * - oldc and newc are equal.
1187 */
1188 ipc_answer_0(rid, EINVAL);
1189 free(old);
1190 free(new);
1191 return;
1192 }
1193
1194 vfs_lookup_res_t old_lr;
1195 vfs_lookup_res_t new_lr;
1196 vfs_lookup_res_t new_par_lr;
1197 fibril_rwlock_write_lock(&namespace_rwlock);
1198
1199 /* Lookup the node belonging to the old file name. */
1200 rc = vfs_lookup_internal(oldc, L_NONE, &old_lr, NULL);
1201 if (rc != EOK) {
1202 fibril_rwlock_write_unlock(&namespace_rwlock);
1203 ipc_answer_0(rid, rc);
1204 free(old);
1205 free(new);
1206 return;
1207 }
1208
1209 vfs_node_t *old_node = vfs_node_get(&old_lr);
1210 if (!old_node) {
1211 fibril_rwlock_write_unlock(&namespace_rwlock);
1212 ipc_answer_0(rid, ENOMEM);
1213 free(old);
1214 free(new);
1215 return;
1216 }
1217
1218 /* Determine the path to the parent of the node with the new name. */
1219 char *parentc = str_dup(newc);
1220 if (!parentc) {
1221 fibril_rwlock_write_unlock(&namespace_rwlock);
1222 ipc_answer_0(rid, rc);
1223 free(old);
1224 free(new);
1225 return;
1226 }
1227
1228 char *lastsl = str_rchr(parentc + 1, '/');
1229 if (lastsl)
1230 *lastsl = '\0';
1231 else
1232 parentc[1] = '\0';
1233
1234 /* Lookup parent of the new file name. */
1235 rc = vfs_lookup_internal(parentc, L_NONE, &new_par_lr, NULL);
1236 free(parentc); /* not needed anymore */
1237 if (rc != EOK) {
1238 fibril_rwlock_write_unlock(&namespace_rwlock);
1239 ipc_answer_0(rid, rc);
1240 free(old);
1241 free(new);
1242 return;
1243 }
1244
1245 /* Check whether linking to the same file system instance. */
1246 if ((old_node->fs_handle != new_par_lr.triplet.fs_handle) ||
1247 (old_node->dev_handle != new_par_lr.triplet.dev_handle)) {
1248 fibril_rwlock_write_unlock(&namespace_rwlock);
1249 ipc_answer_0(rid, EXDEV); /* different file systems */
1250 free(old);
1251 free(new);
1252 return;
1253 }
1254
1255 /* Destroy the old link for the new name. */
1256 vfs_node_t *new_node = NULL;
1257 rc = vfs_lookup_internal(newc, L_UNLINK, &new_lr, NULL);
1258
1259 switch (rc) {
1260 case ENOENT:
1261 /* simply not in our way */
1262 break;
1263 case EOK:
1264 new_node = vfs_node_get(&new_lr);
1265 if (!new_node) {
1266 fibril_rwlock_write_unlock(&namespace_rwlock);
1267 ipc_answer_0(rid, ENOMEM);
1268 free(old);
1269 free(new);
1270 return;
1271 }
1272 fibril_mutex_lock(&nodes_mutex);
1273 new_node->lnkcnt--;
1274 fibril_mutex_unlock(&nodes_mutex);
1275 break;
1276 default:
1277 fibril_rwlock_write_unlock(&namespace_rwlock);
1278 ipc_answer_0(rid, ENOTEMPTY);
1279 free(old);
1280 free(new);
1281 return;
1282 }
1283
1284 /* Create the new link for the new name. */
1285 rc = vfs_lookup_internal(newc, L_LINK, NULL, NULL, old_node->index);
1286 if (rc != EOK) {
1287 fibril_rwlock_write_unlock(&namespace_rwlock);
1288 if (new_node)
1289 vfs_node_put(new_node);
1290 ipc_answer_0(rid, rc);
1291 free(old);
1292 free(new);
1293 return;
1294 }
1295
1296 fibril_mutex_lock(&nodes_mutex);
1297 old_node->lnkcnt++;
1298 fibril_mutex_unlock(&nodes_mutex);
1299
1300 /* Destroy the link for the old name. */
1301 rc = vfs_lookup_internal(oldc, L_UNLINK, NULL, NULL);
1302 if (rc != EOK) {
1303 fibril_rwlock_write_unlock(&namespace_rwlock);
1304 vfs_node_put(old_node);
1305 if (new_node)
1306 vfs_node_put(new_node);
1307 ipc_answer_0(rid, rc);
1308 free(old);
1309 free(new);
1310 return;
1311 }
1312
1313 fibril_mutex_lock(&nodes_mutex);
1314 old_node->lnkcnt--;
1315 fibril_mutex_unlock(&nodes_mutex);
1316 fibril_rwlock_write_unlock(&namespace_rwlock);
1317 vfs_node_put(old_node);
1318
1319 if (new_node)
1320 vfs_node_put(new_node);
1321
1322 free(old);
1323 free(new);
1324 ipc_answer_0(rid, EOK);
1325}
1326
1327void vfs_dup(ipc_callid_t rid, ipc_call_t *request)
1328{
1329 int oldfd = IPC_GET_ARG1(*request);
1330 int newfd = IPC_GET_ARG2(*request);
1331
1332 /* Lookup the file structure corresponding to oldfd. */
1333 vfs_file_t *oldfile = vfs_file_get(oldfd);
1334 if (!oldfile) {
1335 ipc_answer_0(rid, EBADF);
1336 return;
1337 }
1338
1339 /* If the file descriptors are the same, do nothing. */
1340 if (oldfd == newfd) {
1341 ipc_answer_1(rid, EOK, newfd);
1342 return;
1343 }
1344
1345 /*
1346 * Lock the open file structure so that no other thread can manipulate
1347 * the same open file at a time.
1348 */
1349 fibril_mutex_lock(&oldfile->lock);
1350
1351 /* Lookup an open file structure possibly corresponding to newfd. */
1352 vfs_file_t *newfile = vfs_file_get(newfd);
1353 if (newfile) {
1354 /* Close the originally opened file. */
1355 int ret = vfs_close_internal(newfile);
1356 if (ret != EOK) {
1357 fibril_mutex_unlock(&oldfile->lock);
1358 ipc_answer_0(rid, ret);
1359 return;
1360 }
1361
1362 ret = vfs_fd_free(newfd);
1363 if (ret != EOK) {
1364 fibril_mutex_unlock(&oldfile->lock);
1365 ipc_answer_0(rid, ret);
1366 return;
1367 }
1368 }
1369
1370 /* Assign the old file to newfd. */
1371 int ret = vfs_fd_assign(oldfile, newfd);
1372 fibril_mutex_unlock(&oldfile->lock);
1373
1374 if (ret != EOK)
1375 ipc_answer_0(rid, ret);
1376 else
1377 ipc_answer_1(rid, EOK, newfd);
1378}
1379
1380/**
1381 * @}
1382 */
Note: See TracBrowser for help on using the repository browser.