xref: /haiku/src/add-ons/kernel/file_systems/bfs/kernel_interface.cpp (revision b289aaf66bbf6e173aa90fa194fc256965f1b34d)
1 /*
2  * Copyright 2001-2010, Axel Dörfler, axeld@pinc-software.de.
3  * This file may be used under the terms of the MIT License.
4  */
5 
6 
7 //!	file system interface to Haiku's vnode layer
8 
9 
10 #include "Debug.h"
11 #include "Volume.h"
12 #include "Inode.h"
13 #include "Index.h"
14 #include "BPlusTree.h"
15 #include "Query.h"
16 #include "Attribute.h"
17 #include "bfs_control.h"
18 #include "bfs_disk_system.h"
19 
20 // TODO: temporary solution as long as there is no public I/O requests API
21 #ifndef BFS_SHELL
22 #	include <io_requests.h>
23 #endif
24 
25 #define BFS_IO_SIZE	65536
26 
27 
28 struct identify_cookie {
29 	disk_super_block super_block;
30 };
31 
32 extern void fill_stat_buffer(Inode* inode, struct stat& stat);
33 
34 
35 static void
36 fill_stat_time(const bfs_inode& node, struct stat& stat)
37 {
38 	stat.st_atim.tv_sec = real_time_clock();
39 	stat.st_atim.tv_nsec = real_time_clock_usecs() % 1000000;
40 
41 	stat.st_mtim.tv_sec = bfs_inode::ToSecs(node.LastModifiedTime());
42 	stat.st_mtim.tv_nsec = bfs_inode::ToUsecs(node.LastModifiedTime());
43 	stat.st_crtim.tv_sec = bfs_inode::ToSecs(node.CreateTime());
44 	stat.st_crtim.tv_nsec = bfs_inode::ToUsecs(node.CreateTime());
45 
46 	// For BeOS compatibility, if on-disk ctime is invalid, fall back to mtime:
47 	bigtime_t changeTime = node.StatusChangeTime();
48 	if (changeTime < node.LastModifiedTime())
49 		stat.st_ctim = stat.st_mtim;
50 	else {
51 		stat.st_ctim.tv_sec = bfs_inode::ToSecs(changeTime);
52 		stat.st_ctim.tv_nsec = bfs_inode::ToUsecs(changeTime);
53 	}
54 }
55 
56 
57 void
58 fill_stat_buffer(Inode* inode, struct stat& stat)
59 {
60 	const bfs_inode& node = inode->Node();
61 
62 	stat.st_dev = inode->GetVolume()->ID();
63 	stat.st_ino = inode->ID();
64 	stat.st_nlink = 1;
65 	stat.st_blksize = BFS_IO_SIZE;
66 
67 	stat.st_uid = node.UserID();
68 	stat.st_gid = node.GroupID();
69 	stat.st_mode = node.Mode();
70 	stat.st_type = node.Type();
71 
72 	fill_stat_time(node, stat);
73 
74 	if (inode->IsSymLink() && (inode->Flags() & INODE_LONG_SYMLINK) == 0) {
75 		// symlinks report the size of the link here
76 		stat.st_size = strlen(node.short_symlink);
77 	} else
78 		stat.st_size = inode->Size();
79 
80 	stat.st_blocks = inode->AllocatedSize() / 512;
81 }
82 
83 
84 //!	bfs_io() callback hook
85 static status_t
86 iterative_io_get_vecs_hook(void* cookie, io_request* request, off_t offset,
87 	size_t size, struct file_io_vec* vecs, size_t* _count)
88 {
89 	Inode* inode = (Inode*)cookie;
90 
91 	return file_map_translate(inode->Map(), offset, size, vecs, _count,
92 		inode->GetVolume()->BlockSize());
93 }
94 
95 
96 //!	bfs_io() callback hook
97 static status_t
98 iterative_io_finished_hook(void* cookie, io_request* request, status_t status,
99 	bool partialTransfer, size_t bytesTransferred)
100 {
101 	Inode* inode = (Inode*)cookie;
102 	rw_lock_read_unlock(&inode->Lock());
103 	return B_OK;
104 }
105 
106 
107 //	#pragma mark - Scanning
108 
109 
110 static float
111 bfs_identify_partition(int fd, partition_data* partition, void** _cookie)
112 {
113 	disk_super_block superBlock;
114 	status_t status = Volume::Identify(fd, &superBlock);
115 	if (status != B_OK)
116 		return -1;
117 
118 	identify_cookie* cookie = new(std::nothrow) identify_cookie;
119 	if (cookie == NULL)
120 		return -1;
121 
122 	memcpy(&cookie->super_block, &superBlock, sizeof(disk_super_block));
123 
124 	*_cookie = cookie;
125 	return 0.8f;
126 }
127 
128 
129 static status_t
130 bfs_scan_partition(int fd, partition_data* partition, void* _cookie)
131 {
132 	identify_cookie* cookie = (identify_cookie*)_cookie;
133 
134 	partition->status = B_PARTITION_VALID;
135 	partition->flags |= B_PARTITION_FILE_SYSTEM;
136 	partition->content_size = cookie->super_block.NumBlocks()
137 		* cookie->super_block.BlockSize();
138 	partition->block_size = cookie->super_block.BlockSize();
139 	partition->content_name = strdup(cookie->super_block.name);
140 	if (partition->content_name == NULL)
141 		return B_NO_MEMORY;
142 
143 	return B_OK;
144 }
145 
146 
147 static void
148 bfs_free_identify_partition_cookie(partition_data* partition, void* _cookie)
149 {
150 	identify_cookie* cookie = (identify_cookie*)_cookie;
151 	delete cookie;
152 }
153 
154 
155 //	#pragma mark -
156 
157 
158 static status_t
159 bfs_mount(fs_volume* _volume, const char* device, uint32 flags,
160 	const char* args, ino_t* _rootID)
161 {
162 	FUNCTION();
163 
164 	Volume* volume = new(std::nothrow) Volume(_volume);
165 	if (volume == NULL)
166 		return B_NO_MEMORY;
167 
168 	status_t status = volume->Mount(device, flags);
169 	if (status != B_OK) {
170 		delete volume;
171 		RETURN_ERROR(status);
172 	}
173 
174 	_volume->private_volume = volume;
175 	_volume->ops = &gBFSVolumeOps;
176 	*_rootID = volume->ToVnode(volume->Root());
177 
178 	INFORM(("mounted \"%s\" (root node at %" B_PRIdINO ", device = %s)\n",
179 		volume->Name(), *_rootID, device));
180 	return B_OK;
181 }
182 
183 
184 static status_t
185 bfs_unmount(fs_volume* _volume)
186 {
187 	FUNCTION();
188 	Volume* volume = (Volume*)_volume->private_volume;
189 
190 	status_t status = volume->Unmount();
191 	delete volume;
192 
193 	RETURN_ERROR(status);
194 }
195 
196 
197 static status_t
198 bfs_read_fs_stat(fs_volume* _volume, struct fs_info* info)
199 {
200 	FUNCTION();
201 
202 	Volume* volume = (Volume*)_volume->private_volume;
203 	MutexLocker locker(volume->Lock());
204 
205 	// File system flags.
206 	info->flags = B_FS_IS_PERSISTENT | B_FS_HAS_ATTR | B_FS_HAS_MIME
207 		| (volume->IndicesNode() != NULL ? B_FS_HAS_QUERY : 0)
208 		| (volume->IsReadOnly() ? B_FS_IS_READONLY : 0);
209 
210 	info->io_size = BFS_IO_SIZE;
211 		// whatever is appropriate here?
212 
213 	info->block_size = volume->BlockSize();
214 	info->total_blocks = volume->NumBlocks();
215 	info->free_blocks = volume->FreeBlocks();
216 
217 	// Volume name
218 	strlcpy(info->volume_name, volume->Name(), sizeof(info->volume_name));
219 
220 	// File system name
221 	strlcpy(info->fsh_name, "bfs", sizeof(info->fsh_name));
222 
223 	return B_OK;
224 }
225 
226 
227 static status_t
228 bfs_write_fs_stat(fs_volume* _volume, const struct fs_info* info, uint32 mask)
229 {
230 	FUNCTION_START(("mask = %ld\n", mask));
231 
232 	Volume* volume = (Volume*)_volume->private_volume;
233 	if (volume->IsReadOnly())
234 		return B_READ_ONLY_DEVICE;
235 
236 	MutexLocker locker(volume->Lock());
237 
238 	status_t status = B_BAD_VALUE;
239 
240 	if (mask & FS_WRITE_FSINFO_NAME) {
241 		disk_super_block& superBlock = volume->SuperBlock();
242 
243 		strncpy(superBlock.name, info->volume_name,
244 			sizeof(superBlock.name) - 1);
245 		superBlock.name[sizeof(superBlock.name) - 1] = '\0';
246 
247 		status = volume->WriteSuperBlock();
248 	}
249 	return status;
250 }
251 
252 
253 static status_t
254 bfs_sync(fs_volume* _volume)
255 {
256 	FUNCTION();
257 
258 	Volume* volume = (Volume*)_volume->private_volume;
259 	return volume->Sync();
260 }
261 
262 
263 //	#pragma mark -
264 
265 
266 /*!	Reads in the node from disk and creates an inode object from it.
267 */
268 static status_t
269 bfs_get_vnode(fs_volume* _volume, ino_t id, fs_vnode* _node, int* _type,
270 	uint32* _flags, bool reenter)
271 {
272 	//FUNCTION_START(("ino_t = %Ld\n", id));
273 	Volume* volume = (Volume*)_volume->private_volume;
274 
275 	// first inode may be after the log area, we don't go through
276 	// the hassle and try to load an earlier block from disk
277 	if (id < volume->ToBlock(volume->Log()) + volume->Log().Length()
278 		|| id > volume->NumBlocks()) {
279 		INFORM(("inode at %" B_PRIdINO " requested!\n", id));
280 		return B_ERROR;
281 	}
282 
283 	CachedBlock cached(volume, id);
284 	bfs_inode* node = (bfs_inode*)cached.Block();
285 	if (node == NULL) {
286 		FATAL(("could not read inode: %" B_PRIdINO "\n", id));
287 		return B_IO_ERROR;
288 	}
289 
290 	status_t status = node->InitCheck(volume);
291 	if (status != B_OK) {
292 		if ((node->Flags() & INODE_DELETED) != 0) {
293 			INFORM(("inode at %" B_PRIdINO " is already deleted!\n", id));
294 		} else {
295 			FATAL(("inode at %" B_PRIdINO " could not be read: %s!\n", id,
296 				strerror(status)));
297 		}
298 		return status;
299 	}
300 
301 	Inode* inode = new(std::nothrow) Inode(volume, id);
302 	if (inode == NULL)
303 		return B_NO_MEMORY;
304 
305 	status = inode->InitCheck(false);
306 	if (status != B_OK)
307 		delete inode;
308 
309 	if (status == B_OK) {
310 		_node->private_node = inode;
311 		_node->ops = &gBFSVnodeOps;
312 		*_type = inode->Mode();
313 		*_flags = 0;
314 	}
315 
316 	return status;
317 }
318 
319 
320 static status_t
321 bfs_put_vnode(fs_volume* _volume, fs_vnode* _node, bool reenter)
322 {
323 	Volume* volume = (Volume*)_volume->private_volume;
324 	Inode* inode = (Inode*)_node->private_node;
325 
326 	// since a directory's size can be changed without having it opened,
327 	// we need to take care about their preallocated blocks here
328 	if (!volume->IsReadOnly() && !volume->IsCheckingThread()
329 		&& inode->NeedsTrimming()) {
330 		Transaction transaction(volume, inode->BlockNumber());
331 
332 		if (inode->TrimPreallocation(transaction) == B_OK)
333 			transaction.Done();
334 		else if (transaction.HasParent()) {
335 			// TODO: for now, we don't let sub-transactions fail
336 			transaction.Done();
337 		}
338 	}
339 
340 	delete inode;
341 	return B_OK;
342 }
343 
344 
345 static status_t
346 bfs_remove_vnode(fs_volume* _volume, fs_vnode* _node, bool reenter)
347 {
348 	FUNCTION();
349 
350 	Volume* volume = (Volume*)_volume->private_volume;
351 	Inode* inode = (Inode*)_node->private_node;
352 
353 	// If the inode isn't in use anymore, we were called before
354 	// bfs_unlink() returns - in this case, we can just use the
355 	// transaction which has already deleted the inode.
356 	Transaction transaction(volume, volume->ToBlock(inode->Parent()));
357 
358 	// The file system check functionality uses this flag to prevent the space
359 	// used up by the inode from being freed - this flag is set only in
360 	// situations where this does not cause any harm as the block bitmap will
361 	// get fixed anyway in this case).
362 	if ((inode->Flags() & INODE_DONT_FREE_SPACE) != 0) {
363 		delete inode;
364 		return B_OK;
365 	}
366 
367 	ASSERT((inode->Flags() & INODE_DELETED) != 0);
368 
369 	status_t status = inode->Free(transaction);
370 	if (status == B_OK) {
371 		status = transaction.Done();
372 	} else if (transaction.HasParent()) {
373 		// TODO: for now, we don't let sub-transactions fail
374 		status = transaction.Done();
375 	}
376 
377 	volume->RemovedInodes().Remove(inode);
378 
379 	// TODO: the VFS currently does not allow this to fail
380 	delete inode;
381 
382 	return status;
383 }
384 
385 
386 static bool
387 bfs_can_page(fs_volume* _volume, fs_vnode* _v, void* _cookie)
388 {
389 	// TODO: we're obviously not even asked...
390 	return false;
391 }
392 
393 
394 static status_t
395 bfs_read_pages(fs_volume* _volume, fs_vnode* _node, void* _cookie,
396 	off_t pos, const iovec* vecs, size_t count, size_t* _numBytes)
397 {
398 	Volume* volume = (Volume*)_volume->private_volume;
399 	Inode* inode = (Inode*)_node->private_node;
400 
401 	if (inode->FileCache() == NULL)
402 		RETURN_ERROR(B_BAD_VALUE);
403 
404 	InodeReadLocker _(inode);
405 
406 	uint32 vecIndex = 0;
407 	size_t vecOffset = 0;
408 	size_t bytesLeft = *_numBytes;
409 	status_t status;
410 
411 	while (true) {
412 		file_io_vec fileVecs[8];
413 		uint32 fileVecCount = 8;
414 
415 		status = file_map_translate(inode->Map(), pos, bytesLeft, fileVecs,
416 			&fileVecCount, 0);
417 		if (status != B_OK && status != B_BUFFER_OVERFLOW)
418 			break;
419 
420 		bool bufferOverflow = status == B_BUFFER_OVERFLOW;
421 
422 		size_t bytes = bytesLeft;
423 		status = read_file_io_vec_pages(volume->Device(), fileVecs,
424 			fileVecCount, vecs, count, &vecIndex, &vecOffset, &bytes);
425 		if (status != B_OK || !bufferOverflow)
426 			break;
427 
428 		pos += bytes;
429 		bytesLeft -= bytes;
430 	}
431 
432 	return status;
433 }
434 
435 
436 static status_t
437 bfs_write_pages(fs_volume* _volume, fs_vnode* _node, void* _cookie,
438 	off_t pos, const iovec* vecs, size_t count, size_t* _numBytes)
439 {
440 	Volume* volume = (Volume*)_volume->private_volume;
441 	Inode* inode = (Inode*)_node->private_node;
442 
443 	if (volume->IsReadOnly())
444 		return B_READ_ONLY_DEVICE;
445 
446 	if (inode->FileCache() == NULL)
447 		RETURN_ERROR(B_BAD_VALUE);
448 
449 	InodeReadLocker _(inode);
450 
451 	uint32 vecIndex = 0;
452 	size_t vecOffset = 0;
453 	size_t bytesLeft = *_numBytes;
454 	status_t status;
455 
456 	while (true) {
457 		file_io_vec fileVecs[8];
458 		uint32 fileVecCount = 8;
459 
460 		status = file_map_translate(inode->Map(), pos, bytesLeft, fileVecs,
461 			&fileVecCount, 0);
462 		if (status != B_OK && status != B_BUFFER_OVERFLOW)
463 			break;
464 
465 		bool bufferOverflow = status == B_BUFFER_OVERFLOW;
466 
467 		size_t bytes = bytesLeft;
468 		status = write_file_io_vec_pages(volume->Device(), fileVecs,
469 			fileVecCount, vecs, count, &vecIndex, &vecOffset, &bytes);
470 		if (status != B_OK || !bufferOverflow)
471 			break;
472 
473 		pos += bytes;
474 		bytesLeft -= bytes;
475 	}
476 
477 	return status;
478 }
479 
480 
481 static status_t
482 bfs_io(fs_volume* _volume, fs_vnode* _node, void* _cookie, io_request* request)
483 {
484 	Volume* volume = (Volume*)_volume->private_volume;
485 	Inode* inode = (Inode*)_node->private_node;
486 
487 #ifndef BFS_SHELL
488 	if (io_request_is_write(request) && volume->IsReadOnly()) {
489 		notify_io_request(request, B_READ_ONLY_DEVICE);
490 		return B_READ_ONLY_DEVICE;
491 	}
492 #endif
493 
494 	if (inode->FileCache() == NULL) {
495 #ifndef BFS_SHELL
496 		notify_io_request(request, B_BAD_VALUE);
497 #endif
498 		RETURN_ERROR(B_BAD_VALUE);
499 	}
500 
501 	// We lock the node here and will unlock it in the "finished" hook.
502 	rw_lock_read_lock(&inode->Lock());
503 
504 	return do_iterative_fd_io(volume->Device(), request,
505 		iterative_io_get_vecs_hook, iterative_io_finished_hook, inode);
506 }
507 
508 
509 static status_t
510 bfs_get_file_map(fs_volume* _volume, fs_vnode* _node, off_t offset, size_t size,
511 	struct file_io_vec* vecs, size_t* _count)
512 {
513 	Volume* volume = (Volume*)_volume->private_volume;
514 	Inode* inode = (Inode*)_node->private_node;
515 
516 	int32 blockShift = volume->BlockShift();
517 	uint32 index = 0, max = *_count;
518 	block_run run;
519 	off_t fileOffset;
520 
521 	//FUNCTION_START(("offset = %Ld, size = %lu\n", offset, size));
522 
523 	while (true) {
524 		status_t status = inode->FindBlockRun(offset, run, fileOffset);
525 		if (status != B_OK)
526 			return status;
527 
528 		vecs[index].offset = volume->ToOffset(run) + offset - fileOffset;
529 		vecs[index].length = (run.Length() << blockShift) - offset + fileOffset;
530 
531 		// are we already done?
532 		if (size <= vecs[index].length
533 			|| offset + vecs[index].length >= inode->Size()) {
534 			if (offset + vecs[index].length > inode->Size()) {
535 				// make sure the extent ends with the last official file
536 				// block (without taking any preallocations into account)
537 				vecs[index].length = round_up(inode->Size() - offset,
538 					volume->BlockSize());
539 			}
540 			*_count = index + 1;
541 			return B_OK;
542 		}
543 
544 		offset += vecs[index].length;
545 		size -= vecs[index].length;
546 		index++;
547 
548 		if (index >= max) {
549 			// we're out of file_io_vecs; let's bail out
550 			*_count = index;
551 			return B_BUFFER_OVERFLOW;
552 		}
553 	}
554 
555 	// can never get here
556 	return B_ERROR;
557 }
558 
559 
560 //	#pragma mark -
561 
562 
563 static status_t
564 bfs_lookup(fs_volume* _volume, fs_vnode* _directory, const char* file,
565 	ino_t* _vnodeID)
566 {
567 	Volume* volume = (Volume*)_volume->private_volume;
568 	Inode* directory = (Inode*)_directory->private_node;
569 
570 	InodeReadLocker locker(directory);
571 
572 	// check access permissions
573 	status_t status = directory->CheckPermissions(X_OK);
574 	if (status != B_OK)
575 		RETURN_ERROR(status);
576 
577 	BPlusTree* tree = directory->Tree();
578 	if (tree == NULL)
579 		RETURN_ERROR(B_BAD_VALUE);
580 
581 	status = tree->Find((uint8*)file, (uint16)strlen(file), _vnodeID);
582 	if (status != B_OK) {
583 		//PRINT(("bfs_walk() could not find %Ld:\"%s\": %s\n", directory->BlockNumber(), file, strerror(status)));
584 		return status;
585 	}
586 
587 	entry_cache_add(volume->ID(), directory->ID(), file, *_vnodeID);
588 
589 	locker.Unlock();
590 
591 	Inode* inode;
592 	status = get_vnode(volume->FSVolume(), *_vnodeID, (void**)&inode);
593 	if (status != B_OK) {
594 		REPORT_ERROR(status);
595 		return B_ENTRY_NOT_FOUND;
596 	}
597 
598 	return B_OK;
599 }
600 
601 
602 static status_t
603 bfs_get_vnode_name(fs_volume* _volume, fs_vnode* _node, char* buffer,
604 	size_t bufferSize)
605 {
606 	Inode* inode = (Inode*)_node->private_node;
607 
608 	return inode->GetName(buffer, bufferSize);
609 }
610 
611 
612 static status_t
613 bfs_ioctl(fs_volume* _volume, fs_vnode* _node, void* _cookie, ulong cmd,
614 	void* buffer, size_t bufferLength)
615 {
616 	FUNCTION_START(("node = %p, cmd = %lu, buf = %p, len = %ld\n", _node, cmd,
617 		buffer, bufferLength));
618 
619 	Volume* volume = (Volume*)_volume->private_volume;
620 
621 	switch (cmd) {
622 		case BFS_IOCTL_VERSION:
623 		{
624 			uint32 version = 0x10000;
625 			return user_memcpy(buffer, &version, sizeof(uint32));
626 		}
627 		case BFS_IOCTL_START_CHECKING:
628 		{
629 			// start checking
630 			BlockAllocator& allocator = volume->Allocator();
631 			check_control control;
632 			if (user_memcpy(&control, buffer, sizeof(check_control)) != B_OK)
633 				return B_BAD_ADDRESS;
634 
635 			status_t status = allocator.StartChecking(&control);
636 			if (status == B_OK) {
637 				file_cookie* cookie = (file_cookie*)_cookie;
638 				cookie->open_mode |= BFS_OPEN_MODE_CHECKING;
639 			}
640 
641 			return status;
642 		}
643 		case BFS_IOCTL_STOP_CHECKING:
644 		{
645 			// stop checking
646 			BlockAllocator& allocator = volume->Allocator();
647 			check_control control;
648 
649 			status_t status = allocator.StopChecking(&control);
650 			if (status == B_OK) {
651 				file_cookie* cookie = (file_cookie*)_cookie;
652 				cookie->open_mode &= ~BFS_OPEN_MODE_CHECKING;
653 			}
654 			if (status == B_OK)
655 				status = user_memcpy(buffer, &control, sizeof(check_control));
656 
657 			return status;
658 		}
659 		case BFS_IOCTL_CHECK_NEXT_NODE:
660 		{
661 			// check next
662 			BlockAllocator& allocator = volume->Allocator();
663 			check_control control;
664 
665 			status_t status = allocator.CheckNextNode(&control);
666 			if (status == B_OK)
667 				status = user_memcpy(buffer, &control, sizeof(check_control));
668 
669 			return status;
670 		}
671 		case BFS_IOCTL_UPDATE_BOOT_BLOCK:
672 		{
673 			// let's makebootable (or anyone else) update the boot block
674 			// while BFS is mounted
675 			update_boot_block update;
676 			if (bufferLength != sizeof(update_boot_block))
677 				return B_BAD_VALUE;
678 			if (user_memcpy(&update, buffer, sizeof(update_boot_block)) != B_OK)
679 				return B_BAD_ADDRESS;
680 			if (update.offset < offsetof(disk_super_block, pad_to_block)
681 				|| update.length + update.offset > 512)
682 				return B_BAD_VALUE;
683 			if (user_memcpy((uint8*)&volume->SuperBlock() + update.offset,
684 					update.data, update.length) != B_OK)
685 				return B_BAD_ADDRESS;
686 
687 			return volume->WriteSuperBlock();
688 		}
689 
690 #ifdef DEBUG_FRAGMENTER
691 		case 56741:
692 		{
693 			BlockAllocator& allocator = volume->Allocator();
694 			allocator.Fragment();
695 			return B_OK;
696 		}
697 #endif
698 
699 #ifdef DEBUG
700 		case 56742:
701 		{
702 			// allocate all free blocks and zero them out
703 			// (a test for the BlockAllocator)!
704 			BlockAllocator& allocator = volume->Allocator();
705 			Transaction transaction(volume, 0);
706 			CachedBlock cached(volume);
707 			block_run run;
708 			while (allocator.AllocateBlocks(transaction, 8, 0, 64, 1, run)
709 					== B_OK) {
710 				PRINT(("write block_run(%ld, %d, %d)\n", run.allocation_group,
711 					run.start, run.length));
712 				for (int32 i = 0;i < run.length;i++) {
713 					uint8* block = cached.SetToWritable(transaction, run);
714 					if (block != NULL)
715 						memset(block, 0, volume->BlockSize());
716 				}
717 			}
718 			return B_OK;
719 		}
720 #endif
721 	}
722 	return B_BAD_VALUE;
723 }
724 
725 
726 /*!	Sets the open-mode flags for the open file cookie - only
727 	supports O_APPEND currently, but that should be sufficient
728 	for a file system.
729 */
730 static status_t
731 bfs_set_flags(fs_volume* _volume, fs_vnode* _node, void* _cookie, int flags)
732 {
733 	FUNCTION_START(("node = %p, flags = %d", _node, flags));
734 
735 	file_cookie* cookie = (file_cookie*)_cookie;
736 	cookie->open_mode = (cookie->open_mode & ~O_APPEND) | (flags & O_APPEND);
737 
738 	return B_OK;
739 }
740 
741 
742 static status_t
743 bfs_fsync(fs_volume* _volume, fs_vnode* _node)
744 {
745 	FUNCTION();
746 
747 	Inode* inode = (Inode*)_node->private_node;
748 	return inode->Sync();
749 }
750 
751 
752 static status_t
753 bfs_read_stat(fs_volume* _volume, fs_vnode* _node, struct stat* stat)
754 {
755 	FUNCTION();
756 
757 	Inode* inode = (Inode*)_node->private_node;
758 	fill_stat_buffer(inode, *stat);
759 	return B_OK;
760 }
761 
762 
763 static status_t
764 bfs_write_stat(fs_volume* _volume, fs_vnode* _node, const struct stat* stat,
765 	uint32 mask)
766 {
767 	FUNCTION();
768 
769 	Volume* volume = (Volume*)_volume->private_volume;
770 	Inode* inode = (Inode*)_node->private_node;
771 
772 	if (volume->IsReadOnly())
773 		return B_READ_ONLY_DEVICE;
774 
775 	// TODO: we should definitely check a bit more if the new stats are
776 	//	valid - or even better, the VFS should check this before calling us
777 
778 	status_t status = inode->CheckPermissions(W_OK);
779 	if (status < B_OK)
780 		RETURN_ERROR(status);
781 
782 	Transaction transaction(volume, inode->BlockNumber());
783 	inode->WriteLockInTransaction(transaction);
784 
785 	bfs_inode& node = inode->Node();
786 	bool updateTime = false;
787 
788 	if ((mask & B_STAT_SIZE) != 0) {
789 		// Since WSTAT_SIZE is the only thing that can fail directly, we
790 		// do it first, so that the inode state will still be consistent
791 		// with the on-disk version
792 		if (inode->IsDirectory())
793 			return B_IS_A_DIRECTORY;
794 		if (!inode->IsFile())
795 			return B_BAD_VALUE;
796 
797 		if (inode->Size() != stat->st_size) {
798 			off_t oldSize = inode->Size();
799 
800 			status = inode->SetFileSize(transaction, stat->st_size);
801 			if (status != B_OK)
802 				return status;
803 
804 			// fill the new blocks (if any) with zeros
805 			if ((mask & B_STAT_SIZE_INSECURE) == 0) {
806 				// We must not keep the inode locked during a write operation,
807 				// or else we might deadlock.
808 				rw_lock_write_unlock(&inode->Lock());
809 				inode->FillGapWithZeros(oldSize, inode->Size());
810 				rw_lock_write_lock(&inode->Lock());
811 			}
812 
813 			if (!inode->IsDeleted()) {
814 				Index index(volume);
815 				index.UpdateSize(transaction, inode);
816 
817 				updateTime = true;
818 			}
819 		}
820 	}
821 
822 	if ((mask & B_STAT_MODE) != 0) {
823 		PRINT(("original mode = %ld, stat->st_mode = %d\n", node.Mode(),
824 			stat->st_mode));
825 		node.mode = HOST_ENDIAN_TO_BFS_INT32((node.Mode() & ~S_IUMSK)
826 			| (stat->st_mode & S_IUMSK));
827 		updateTime = true;
828 	}
829 
830 	if ((mask & B_STAT_UID) != 0) {
831 		node.uid = HOST_ENDIAN_TO_BFS_INT32(stat->st_uid);
832 		updateTime = true;
833 	}
834 	if ((mask & B_STAT_GID) != 0) {
835 		node.gid = HOST_ENDIAN_TO_BFS_INT32(stat->st_gid);
836 		updateTime = true;
837 	}
838 
839 	if ((mask & B_STAT_MODIFICATION_TIME) != 0) {
840 		if (!inode->InLastModifiedIndex()) {
841 			// directory modification times are not part of the index
842 			node.last_modified_time
843 				= HOST_ENDIAN_TO_BFS_INT64(bfs_inode::ToInode(stat->st_mtim));
844 		} else if (!inode->IsDeleted()) {
845 			// Index::UpdateLastModified() will set the new time in the inode
846 			Index index(volume);
847 			index.UpdateLastModified(transaction, inode,
848 				bfs_inode::ToInode(stat->st_mtim));
849 		}
850 	}
851 	if ((mask & B_STAT_CREATION_TIME) != 0) {
852 		node.create_time
853 			= HOST_ENDIAN_TO_BFS_INT64(bfs_inode::ToInode(stat->st_crtim));
854 	}
855 	if ((mask & B_STAT_CHANGE_TIME) != 0 || updateTime) {
856 		bigtime_t newTime;
857 		if ((mask & B_STAT_CHANGE_TIME) == 0)
858 			newTime = bfs_inode::ToInode(real_time_clock_usecs());
859 		else
860 			newTime = bfs_inode::ToInode(stat->st_ctim);
861 
862 		node.status_change_time = HOST_ENDIAN_TO_BFS_INT64(newTime);
863 	}
864 
865 	status = inode->WriteBack(transaction);
866 	if (status == B_OK)
867 		status = transaction.Done();
868 	if (status == B_OK)
869 		notify_stat_changed(volume->ID(), inode->ID(), mask);
870 
871 	return status;
872 }
873 
874 
875 status_t
876 bfs_create(fs_volume* _volume, fs_vnode* _directory, const char* name,
877 	int openMode, int mode, void** _cookie, ino_t* _vnodeID)
878 {
879 	FUNCTION_START(("name = \"%s\", perms = %d, openMode = %d\n", name, mode,
880 		openMode));
881 
882 	Volume* volume = (Volume*)_volume->private_volume;
883 	Inode* directory = (Inode*)_directory->private_node;
884 
885 	if (volume->IsReadOnly())
886 		return B_READ_ONLY_DEVICE;
887 
888 	if (!directory->IsDirectory())
889 		RETURN_ERROR(B_BAD_TYPE);
890 
891 	// We are creating the cookie at this point, so that we don't have
892 	// to remove the inode if we don't have enough free memory later...
893 	file_cookie* cookie = new(std::nothrow) file_cookie;
894 	if (cookie == NULL)
895 		RETURN_ERROR(B_NO_MEMORY);
896 
897 	// initialize the cookie
898 	cookie->open_mode = openMode;
899 	cookie->last_size = 0;
900 	cookie->last_notification = system_time();
901 
902 	Transaction transaction(volume, directory->BlockNumber());
903 
904 	Inode* inode;
905 	bool created;
906 	status_t status = Inode::Create(transaction, directory, name,
907 		S_FILE | (mode & S_IUMSK), openMode, 0, &created, _vnodeID, &inode);
908 
909 	// Disable the file cache, if requested?
910 	if (status == B_OK && (openMode & O_NOCACHE) != 0
911 		&& inode->FileCache() != NULL) {
912 		status = file_cache_disable(inode->FileCache());
913 	}
914 
915 	entry_cache_add(volume->ID(), directory->ID(), name, *_vnodeID);
916 
917 	if (status == B_OK)
918 		status = transaction.Done();
919 
920 	if (status == B_OK) {
921 		// register the cookie
922 		*_cookie = cookie;
923 
924 		if (created) {
925 			notify_entry_created(volume->ID(), directory->ID(), name,
926 				*_vnodeID);
927 		}
928 	} else {
929 		entry_cache_remove(volume->ID(), directory->ID(), name);
930 		delete cookie;
931 	}
932 
933 	return status;
934 }
935 
936 
937 static status_t
938 bfs_create_symlink(fs_volume* _volume, fs_vnode* _directory, const char* name,
939 	const char* path, int mode)
940 {
941 	FUNCTION_START(("name = \"%s\", path = \"%s\"\n", name, path));
942 
943 	Volume* volume = (Volume*)_volume->private_volume;
944 	Inode* directory = (Inode*)_directory->private_node;
945 
946 	if (volume->IsReadOnly())
947 		return B_READ_ONLY_DEVICE;
948 
949 	if (!directory->IsDirectory())
950 		RETURN_ERROR(B_BAD_TYPE);
951 
952 	status_t status = directory->CheckPermissions(W_OK);
953 	if (status < B_OK)
954 		RETURN_ERROR(status);
955 
956 	Transaction transaction(volume, directory->BlockNumber());
957 
958 	Inode* link;
959 	off_t id;
960 	status = Inode::Create(transaction, directory, name, S_SYMLINK | 0777,
961 		0, 0, NULL, &id, &link);
962 	if (status < B_OK)
963 		RETURN_ERROR(status);
964 
965 	size_t length = strlen(path);
966 	if (length < SHORT_SYMLINK_NAME_LENGTH) {
967 		strcpy(link->Node().short_symlink, path);
968 	} else {
969 		link->Node().flags |= HOST_ENDIAN_TO_BFS_INT32(INODE_LONG_SYMLINK
970 			| INODE_LOGGED);
971 
972 		// links usually don't have a file cache attached - but we now need one
973 		link->SetFileCache(file_cache_create(volume->ID(), link->ID(), 0));
974 		link->SetMap(file_map_create(volume->ID(), link->ID(), 0));
975 
976 		// The following call will have to write the inode back, so
977 		// we don't have to do that here...
978 		status = link->WriteAt(transaction, 0, (const uint8*)path, &length);
979 	}
980 
981 	if (status == B_OK)
982 		status = link->WriteBack(transaction);
983 
984 	// Inode::Create() left the inode locked in memory, and also doesn't
985 	// publish links
986 	publish_vnode(volume->FSVolume(), id, link, &gBFSVnodeOps, link->Mode(), 0);
987 	put_vnode(volume->FSVolume(), id);
988 
989 	if (status == B_OK) {
990 		entry_cache_add(volume->ID(), directory->ID(), name, id);
991 
992 		status = transaction.Done();
993 		if (status == B_OK)
994 			notify_entry_created(volume->ID(), directory->ID(), name, id);
995 		else
996 			entry_cache_remove(volume->ID(), directory->ID(), name);
997 	}
998 
999 	return status;
1000 }
1001 
1002 
1003 status_t
1004 bfs_link(fs_volume* _volume, fs_vnode* dir, const char* name, fs_vnode* node)
1005 {
1006 	FUNCTION_START(("name = \"%s\"\n", name));
1007 
1008 	// This one won't be implemented in a binary compatible BFS
1009 	return B_NOT_SUPPORTED;
1010 }
1011 
1012 
1013 status_t
1014 bfs_unlink(fs_volume* _volume, fs_vnode* _directory, const char* name)
1015 {
1016 	FUNCTION_START(("name = \"%s\"\n", name));
1017 
1018 	if (!strcmp(name, "..") || !strcmp(name, "."))
1019 		return B_NOT_ALLOWED;
1020 
1021 	Volume* volume = (Volume*)_volume->private_volume;
1022 	Inode* directory = (Inode*)_directory->private_node;
1023 
1024 	status_t status = directory->CheckPermissions(W_OK);
1025 	if (status < B_OK)
1026 		return status;
1027 
1028 	Transaction transaction(volume, directory->BlockNumber());
1029 
1030 	off_t id;
1031 	status = directory->Remove(transaction, name, &id);
1032 	if (status == B_OK) {
1033 		entry_cache_remove(volume->ID(), directory->ID(), name);
1034 
1035 		status = transaction.Done();
1036 		if (status == B_OK)
1037 			notify_entry_removed(volume->ID(), directory->ID(), name, id);
1038 		else
1039 			entry_cache_add(volume->ID(), directory->ID(), name, id);
1040 	}
1041 	return status;
1042 }
1043 
1044 
1045 status_t
1046 bfs_rename(fs_volume* _volume, fs_vnode* _oldDir, const char* oldName,
1047 	fs_vnode* _newDir, const char* newName)
1048 {
1049 	FUNCTION_START(("oldDir = %p, oldName = \"%s\", newDir = %p, newName = "
1050 		"\"%s\"\n", _oldDir, oldName, _newDir, newName));
1051 
1052 	Volume* volume = (Volume*)_volume->private_volume;
1053 	Inode* oldDirectory = (Inode*)_oldDir->private_node;
1054 	Inode* newDirectory = (Inode*)_newDir->private_node;
1055 
1056 	// are we already done?
1057 	if (oldDirectory == newDirectory && !strcmp(oldName, newName))
1058 		return B_OK;
1059 
1060 	Transaction transaction(volume, oldDirectory->BlockNumber());
1061 
1062 	oldDirectory->WriteLockInTransaction(transaction);
1063 	if (oldDirectory != newDirectory)
1064 		newDirectory->WriteLockInTransaction(transaction);
1065 
1066 	// are we allowed to do what we've been told?
1067 	status_t status = oldDirectory->CheckPermissions(W_OK);
1068 	if (status == B_OK)
1069 		status = newDirectory->CheckPermissions(W_OK);
1070 	if (status != B_OK)
1071 		return status;
1072 
1073 	// Get the directory's tree, and a pointer to the inode which should be
1074 	// changed
1075 	BPlusTree* tree = oldDirectory->Tree();
1076 	if (tree == NULL)
1077 		RETURN_ERROR(B_BAD_VALUE);
1078 
1079 	off_t id;
1080 	status = tree->Find((const uint8*)oldName, strlen(oldName), &id);
1081 	if (status != B_OK)
1082 		RETURN_ERROR(status);
1083 
1084 	Vnode vnode(volume, id);
1085 	Inode* inode;
1086 	if (vnode.Get(&inode) != B_OK)
1087 		return B_IO_ERROR;
1088 
1089 	// Don't move a directory into one of its children - we soar up
1090 	// from the newDirectory to either the root node or the old
1091 	// directory, whichever comes first.
1092 	// If we meet our inode on that way, we have to bail out.
1093 
1094 	if (oldDirectory != newDirectory) {
1095 		ino_t parent = newDirectory->ID();
1096 		ino_t root = volume->RootNode()->ID();
1097 
1098 		while (true) {
1099 			if (parent == id)
1100 				return B_BAD_VALUE;
1101 			else if (parent == root || parent == oldDirectory->ID())
1102 				break;
1103 
1104 			Vnode vnode(volume, parent);
1105 			Inode* parentNode;
1106 			if (vnode.Get(&parentNode) != B_OK)
1107 				return B_ERROR;
1108 
1109 			parent = volume->ToVnode(parentNode->Parent());
1110 		}
1111 	}
1112 
1113 	// Everything okay? Then lets get to work...
1114 
1115 	// First, try to make sure there is nothing that will stop us in
1116 	// the target directory - since this is the only non-critical
1117 	// failure, we will test this case first
1118 	BPlusTree* newTree = tree;
1119 	if (newDirectory != oldDirectory) {
1120 		newTree = newDirectory->Tree();
1121 		if (newTree == NULL)
1122 			RETURN_ERROR(B_BAD_VALUE);
1123 	}
1124 
1125 	status = newTree->Insert(transaction, (const uint8*)newName,
1126 		strlen(newName), id);
1127 	if (status == B_NAME_IN_USE) {
1128 		// If there is already a file with that name, we have to remove
1129 		// it, as long it's not a directory with files in it
1130 		off_t clobber;
1131 		if (newTree->Find((const uint8*)newName, strlen(newName), &clobber)
1132 				< B_OK)
1133 			return B_NAME_IN_USE;
1134 		if (clobber == id)
1135 			return B_BAD_VALUE;
1136 
1137 		Vnode vnode(volume, clobber);
1138 		Inode* other;
1139 		if (vnode.Get(&other) < B_OK)
1140 			return B_NAME_IN_USE;
1141 
1142 		// only allowed, if either both nodes are directories or neither is
1143 		if (inode->IsDirectory() != other->IsDirectory())
1144 			return other->IsDirectory() ? B_IS_A_DIRECTORY : B_NOT_A_DIRECTORY;
1145 
1146 		status = newDirectory->Remove(transaction, newName, NULL,
1147 			other->IsDirectory());
1148 		if (status < B_OK)
1149 			return status;
1150 
1151 		entry_cache_remove(volume->ID(), newDirectory->ID(), newName);
1152 
1153 		notify_entry_removed(volume->ID(), newDirectory->ID(), newName,
1154 			clobber);
1155 
1156 		status = newTree->Insert(transaction, (const uint8*)newName,
1157 			strlen(newName), id);
1158 	}
1159 	if (status != B_OK)
1160 		return status;
1161 
1162 	inode->WriteLockInTransaction(transaction);
1163 
1164 	volume->UpdateLiveQueriesRenameMove(inode, oldDirectory->ID(), oldName,
1165 		newDirectory->ID(), newName);
1166 
1167 	// update the name only when they differ
1168 	bool nameUpdated = false;
1169 	if (strcmp(oldName, newName)) {
1170 		status = inode->SetName(transaction, newName);
1171 		if (status == B_OK) {
1172 			Index index(volume);
1173 			index.UpdateName(transaction, oldName, newName, inode);
1174 			nameUpdated = true;
1175 		}
1176 	}
1177 
1178 	if (status == B_OK) {
1179 		status = tree->Remove(transaction, (const uint8*)oldName,
1180 			strlen(oldName), id);
1181 		if (status == B_OK) {
1182 			inode->Parent() = newDirectory->BlockRun();
1183 
1184 			// if it's a directory, update the parent directory pointer
1185 			// in its tree if necessary
1186 			BPlusTree* movedTree = inode->Tree();
1187 			if (oldDirectory != newDirectory
1188 				&& inode->IsDirectory()
1189 				&& movedTree != NULL) {
1190 				status = movedTree->Replace(transaction, (const uint8*)"..",
1191 					2, newDirectory->ID());
1192 
1193 				if (status == B_OK) {
1194 					// update/add the cache entry for the parent
1195 					entry_cache_add(volume->ID(), id, "..", newDirectory->ID());
1196 				}
1197 			}
1198 
1199 			if (status == B_OK && newDirectory != oldDirectory)
1200 				status = oldDirectory->ContainerContentsChanged(transaction);
1201 			if (status == B_OK)
1202 				status = newDirectory->ContainerContentsChanged(transaction);
1203 
1204 			if (status == B_OK)
1205 				status = inode->WriteBack(transaction);
1206 
1207 			if (status == B_OK) {
1208 				entry_cache_remove(volume->ID(), oldDirectory->ID(), oldName);
1209 				entry_cache_add(volume->ID(), newDirectory->ID(), newName, id);
1210 
1211 				status = transaction.Done();
1212 				if (status == B_OK) {
1213 					notify_entry_moved(volume->ID(), oldDirectory->ID(),
1214 						oldName, newDirectory->ID(), newName, id);
1215 					return B_OK;
1216 				}
1217 
1218 				entry_cache_remove(volume->ID(), oldDirectory->ID(), newName);
1219 				entry_cache_add(volume->ID(), newDirectory->ID(), oldName, id);
1220 			}
1221 		}
1222 	}
1223 
1224 	return status;
1225 }
1226 
1227 
1228 static status_t
1229 bfs_open(fs_volume* _volume, fs_vnode* _node, int openMode, void** _cookie)
1230 {
1231 	FUNCTION();
1232 
1233 	Volume* volume = (Volume*)_volume->private_volume;
1234 	Inode* inode = (Inode*)_node->private_node;
1235 
1236 	// Opening a directory read-only is allowed, although you can't read
1237 	// any data from it.
1238 	if (inode->IsDirectory() && (openMode & O_RWMASK) != O_RDONLY)
1239 		return B_IS_A_DIRECTORY;
1240 	if ((openMode & O_DIRECTORY) != 0 && !inode->IsDirectory())
1241 		return B_NOT_A_DIRECTORY;
1242 
1243 	status_t status = inode->CheckPermissions(open_mode_to_access(openMode)
1244 		| ((openMode & O_TRUNC) != 0 ? W_OK : 0));
1245 	if (status != B_OK)
1246 		RETURN_ERROR(status);
1247 
1248 	file_cookie* cookie = new(std::nothrow) file_cookie;
1249 	if (cookie == NULL)
1250 		RETURN_ERROR(B_NO_MEMORY);
1251 	ObjectDeleter<file_cookie> cookieDeleter(cookie);
1252 
1253 	// initialize the cookie
1254 	cookie->open_mode = openMode & BFS_OPEN_MODE_USER_MASK;
1255 	cookie->last_size = inode->Size();
1256 	cookie->last_notification = system_time();
1257 
1258 	// Disable the file cache, if requested?
1259 	CObjectDeleter<void> fileCacheEnabler(file_cache_enable);
1260 	if ((openMode & O_NOCACHE) != 0 && inode->FileCache() != NULL) {
1261 		status = file_cache_disable(inode->FileCache());
1262 		if (status != B_OK)
1263 			return status;
1264 		fileCacheEnabler.SetTo(inode->FileCache());
1265 	}
1266 
1267 	// Should we truncate the file?
1268 	if ((openMode & O_TRUNC) != 0) {
1269 		if ((openMode & O_RWMASK) == O_RDONLY)
1270 			return B_NOT_ALLOWED;
1271 
1272 		Transaction transaction(volume, inode->BlockNumber());
1273 		inode->WriteLockInTransaction(transaction);
1274 
1275 		status_t status = inode->SetFileSize(transaction, 0);
1276 		if (status == B_OK)
1277 			status = inode->WriteBack(transaction);
1278 		if (status == B_OK)
1279 			status = transaction.Done();
1280 		if (status != B_OK)
1281 			return status;
1282 	}
1283 
1284 	fileCacheEnabler.Detach();
1285 	cookieDeleter.Detach();
1286 	*_cookie = cookie;
1287 	return B_OK;
1288 }
1289 
1290 
1291 static status_t
1292 bfs_read(fs_volume* _volume, fs_vnode* _node, void* _cookie, off_t pos,
1293 	void* buffer, size_t* _length)
1294 {
1295 	//FUNCTION();
1296 	Inode* inode = (Inode*)_node->private_node;
1297 
1298 	if (!inode->HasUserAccessableStream()) {
1299 		*_length = 0;
1300 		return inode->IsDirectory() ? B_IS_A_DIRECTORY : B_BAD_VALUE;
1301 	}
1302 
1303 	return inode->ReadAt(pos, (uint8*)buffer, _length);
1304 }
1305 
1306 
1307 static status_t
1308 bfs_write(fs_volume* _volume, fs_vnode* _node, void* _cookie, off_t pos,
1309 	const void* buffer, size_t* _length)
1310 {
1311 	//FUNCTION();
1312 	Volume* volume = (Volume*)_volume->private_volume;
1313 	Inode* inode = (Inode*)_node->private_node;
1314 
1315 	if (volume->IsReadOnly())
1316 		return B_READ_ONLY_DEVICE;
1317 
1318 	if (!inode->HasUserAccessableStream()) {
1319 		*_length = 0;
1320 		return inode->IsDirectory() ? B_IS_A_DIRECTORY : B_BAD_VALUE;
1321 	}
1322 
1323 	file_cookie* cookie = (file_cookie*)_cookie;
1324 
1325 	if (cookie->open_mode & O_APPEND)
1326 		pos = inode->Size();
1327 
1328 	Transaction transaction;
1329 		// We are not starting the transaction here, since
1330 		// it might not be needed at all (the contents of
1331 		// regular files aren't logged)
1332 
1333 	status_t status = inode->WriteAt(transaction, pos, (const uint8*)buffer,
1334 		_length);
1335 	if (status == B_OK)
1336 		status = transaction.Done();
1337 	if (status == B_OK) {
1338 		InodeReadLocker locker(inode);
1339 
1340 		// periodically notify if the file size has changed
1341 		// TODO: should we better test for a change in the last_modified time only?
1342 		if (!inode->IsDeleted() && cookie->last_size != inode->Size()
1343 			&& system_time() > cookie->last_notification
1344 					+ INODE_NOTIFICATION_INTERVAL) {
1345 			notify_stat_changed(volume->ID(), inode->ID(),
1346 				B_STAT_MODIFICATION_TIME | B_STAT_SIZE | B_STAT_INTERIM_UPDATE);
1347 			cookie->last_size = inode->Size();
1348 			cookie->last_notification = system_time();
1349 		}
1350 	}
1351 
1352 	return status;
1353 }
1354 
1355 
1356 static status_t
1357 bfs_close(fs_volume* _volume, fs_vnode* _node, void* _cookie)
1358 {
1359 	FUNCTION();
1360 	return B_OK;
1361 }
1362 
1363 
1364 static status_t
1365 bfs_free_cookie(fs_volume* _volume, fs_vnode* _node, void* _cookie)
1366 {
1367 	FUNCTION();
1368 
1369 	file_cookie* cookie = (file_cookie*)_cookie;
1370 	Volume* volume = (Volume*)_volume->private_volume;
1371 	Inode* inode = (Inode*)_node->private_node;
1372 
1373 	Transaction transaction;
1374 	bool needsTrimming = false;
1375 
1376 	if (!volume->IsReadOnly() && !volume->IsCheckingThread()) {
1377 		InodeReadLocker locker(inode);
1378 		needsTrimming = inode->NeedsTrimming();
1379 
1380 		if ((cookie->open_mode & O_RWMASK) != 0
1381 			&& !inode->IsDeleted()
1382 			&& (needsTrimming
1383 				|| inode->OldLastModified() != inode->LastModified()
1384 				|| (inode->InSizeIndex()
1385 					// TODO: this can prevent the size update notification
1386 					// for nodes not in the index!
1387 					&& inode->OldSize() != inode->Size()))) {
1388 			locker.Unlock();
1389 			transaction.Start(volume, inode->BlockNumber());
1390 		}
1391 	}
1392 
1393 	status_t status = transaction.IsStarted() ? B_OK : B_ERROR;
1394 
1395 	if (status == B_OK) {
1396 		inode->WriteLockInTransaction(transaction);
1397 
1398 		// trim the preallocated blocks and update the size,
1399 		// and last_modified indices if needed
1400 		bool changedSize = false, changedTime = false;
1401 		Index index(volume);
1402 
1403 		if (needsTrimming) {
1404 			status = inode->TrimPreallocation(transaction);
1405 			if (status < B_OK) {
1406 				FATAL(("Could not trim preallocated blocks: inode %" B_PRIdINO
1407 					", transaction %d: %s!\n", inode->ID(),
1408 					(int)transaction.ID(), strerror(status)));
1409 
1410 				// we still want this transaction to succeed
1411 				status = B_OK;
1412 			}
1413 		}
1414 		if (inode->OldSize() != inode->Size()) {
1415 			if (inode->InSizeIndex())
1416 				index.UpdateSize(transaction, inode);
1417 			changedSize = true;
1418 		}
1419 		if (inode->OldLastModified() != inode->LastModified()) {
1420 			if (inode->InLastModifiedIndex()) {
1421 				index.UpdateLastModified(transaction, inode,
1422 					inode->LastModified());
1423 			}
1424 			changedTime = true;
1425 
1426 			// updating the index doesn't write back the inode
1427 			inode->WriteBack(transaction);
1428 		}
1429 
1430 		if (changedSize || changedTime) {
1431 			notify_stat_changed(volume->ID(), inode->ID(),
1432 				(changedTime ? B_STAT_MODIFICATION_TIME : 0)
1433 				| (changedSize ? B_STAT_SIZE : 0));
1434 		}
1435 	}
1436 	if (status == B_OK)
1437 		transaction.Done();
1438 
1439 	if ((cookie->open_mode & BFS_OPEN_MODE_CHECKING) != 0) {
1440 		// "chkbfs" exited abnormally, so we have to stop it here...
1441 		FATAL(("check process was aborted!\n"));
1442 		volume->Allocator().StopChecking(NULL);
1443 	}
1444 
1445 	if ((cookie->open_mode & O_NOCACHE) != 0 && inode->FileCache() != NULL)
1446 		file_cache_enable(inode->FileCache());
1447 
1448 	delete cookie;
1449 	return B_OK;
1450 }
1451 
1452 
1453 /*!	Checks access permissions, return B_NOT_ALLOWED if the action
1454 	is not allowed.
1455 */
1456 static status_t
1457 bfs_access(fs_volume* _volume, fs_vnode* _node, int accessMode)
1458 {
1459 	//FUNCTION();
1460 
1461 	Inode* inode = (Inode*)_node->private_node;
1462 	status_t status = inode->CheckPermissions(accessMode);
1463 	if (status < B_OK)
1464 		RETURN_ERROR(status);
1465 
1466 	return B_OK;
1467 }
1468 
1469 
1470 static status_t
1471 bfs_read_link(fs_volume* _volume, fs_vnode* _node, char* buffer,
1472 	size_t* _bufferSize)
1473 {
1474 	FUNCTION();
1475 
1476 	Inode* inode = (Inode*)_node->private_node;
1477 
1478 	if (!inode->IsSymLink())
1479 		RETURN_ERROR(B_BAD_VALUE);
1480 
1481 	if ((inode->Flags() & INODE_LONG_SYMLINK) != 0) {
1482 		if (inode->Size() < *_bufferSize)
1483 			*_bufferSize = inode->Size();
1484 
1485 		status_t status = inode->ReadAt(0, (uint8*)buffer, _bufferSize);
1486 		if (status < B_OK)
1487 			RETURN_ERROR(status);
1488 
1489 		return B_OK;
1490 	}
1491 
1492 	size_t linkLen = strlen(inode->Node().short_symlink);
1493 	if (linkLen < *_bufferSize)
1494 		*_bufferSize = linkLen;
1495 
1496 	return user_memcpy(buffer, inode->Node().short_symlink, *_bufferSize);
1497 }
1498 
1499 
1500 //	#pragma mark - Directory functions
1501 
1502 
1503 static status_t
1504 bfs_create_dir(fs_volume* _volume, fs_vnode* _directory, const char* name,
1505 	int mode)
1506 {
1507 	FUNCTION_START(("name = \"%s\", perms = %d\n", name, mode));
1508 
1509 	Volume* volume = (Volume*)_volume->private_volume;
1510 	Inode* directory = (Inode*)_directory->private_node;
1511 
1512 	if (volume->IsReadOnly())
1513 		return B_READ_ONLY_DEVICE;
1514 
1515 	if (!directory->IsDirectory())
1516 		RETURN_ERROR(B_BAD_TYPE);
1517 
1518 	status_t status = directory->CheckPermissions(W_OK);
1519 	if (status < B_OK)
1520 		RETURN_ERROR(status);
1521 
1522 	Transaction transaction(volume, directory->BlockNumber());
1523 
1524 	// Inode::Create() locks the inode if we pass the "id" parameter, but we
1525 	// need it anyway
1526 	off_t id;
1527 	status = Inode::Create(transaction, directory, name,
1528 		S_DIRECTORY | (mode & S_IUMSK), 0, 0, NULL, &id);
1529 	if (status == B_OK) {
1530 		put_vnode(volume->FSVolume(), id);
1531 
1532 		entry_cache_add(volume->ID(), directory->ID(), name, id);
1533 
1534 		status = transaction.Done();
1535 		if (status == B_OK)
1536 			notify_entry_created(volume->ID(), directory->ID(), name, id);
1537 		else
1538 			entry_cache_remove(volume->ID(), directory->ID(), name);
1539 	}
1540 
1541 	return status;
1542 }
1543 
1544 
1545 static status_t
1546 bfs_remove_dir(fs_volume* _volume, fs_vnode* _directory, const char* name)
1547 {
1548 	FUNCTION_START(("name = \"%s\"\n", name));
1549 
1550 	Volume* volume = (Volume*)_volume->private_volume;
1551 	Inode* directory = (Inode*)_directory->private_node;
1552 
1553 	Transaction transaction(volume, directory->BlockNumber());
1554 
1555 	off_t id;
1556 	status_t status = directory->Remove(transaction, name, &id, true);
1557 	if (status == B_OK) {
1558 		// Remove the cache entry for the directory and potentially also
1559 		// the parent entry still belonging to the directory
1560 		entry_cache_remove(volume->ID(), directory->ID(), name);
1561 		entry_cache_remove(volume->ID(), id, "..");
1562 
1563 		status = transaction.Done();
1564 		if (status == B_OK)
1565 			notify_entry_removed(volume->ID(), directory->ID(), name, id);
1566 		else {
1567 			entry_cache_add(volume->ID(), directory->ID(), name, id);
1568 			entry_cache_add(volume->ID(), id, "..", id);
1569 		}
1570 	}
1571 
1572 	return status;
1573 }
1574 
1575 
1576 /*!	Opens a directory ready to be traversed.
1577 	bfs_open_dir() is also used by bfs_open_index_dir().
1578 */
1579 static status_t
1580 bfs_open_dir(fs_volume* _volume, fs_vnode* _node, void** _cookie)
1581 {
1582 	FUNCTION();
1583 
1584 	Inode* inode = (Inode*)_node->private_node;
1585 	status_t status = inode->CheckPermissions(R_OK);
1586 	if (status < B_OK)
1587 		RETURN_ERROR(status);
1588 
1589 	// we don't ask here for directories only, because the bfs_open_index_dir()
1590 	// function utilizes us (so we must be able to open indices as well)
1591 	if (!inode->IsContainer())
1592 		RETURN_ERROR(B_NOT_A_DIRECTORY);
1593 
1594 	BPlusTree* tree = inode->Tree();
1595 	if (tree == NULL)
1596 		RETURN_ERROR(B_BAD_VALUE);
1597 
1598 	TreeIterator* iterator = new(std::nothrow) TreeIterator(tree);
1599 	if (iterator == NULL)
1600 		RETURN_ERROR(B_NO_MEMORY);
1601 
1602 	*_cookie = iterator;
1603 	return B_OK;
1604 }
1605 
1606 
1607 static status_t
1608 bfs_read_dir(fs_volume* _volume, fs_vnode* _node, void* _cookie,
1609 	struct dirent* dirent, size_t bufferSize, uint32* _num)
1610 {
1611 	FUNCTION();
1612 
1613 	TreeIterator* iterator = (TreeIterator*)_cookie;
1614 
1615 	uint16 length;
1616 	ino_t id;
1617 	status_t status = iterator->GetNextEntry(dirent->d_name, &length,
1618 		bufferSize, &id);
1619 	if (status == B_ENTRY_NOT_FOUND) {
1620 		*_num = 0;
1621 		return B_OK;
1622 	} else if (status != B_OK)
1623 		RETURN_ERROR(status);
1624 
1625 	Volume* volume = (Volume*)_volume->private_volume;
1626 
1627 	dirent->d_dev = volume->ID();
1628 	dirent->d_ino = id;
1629 
1630 	dirent->d_reclen = sizeof(struct dirent) + length;
1631 
1632 	*_num = 1;
1633 	return B_OK;
1634 }
1635 
1636 
1637 /*!	Sets the TreeIterator back to the beginning of the directory. */
1638 static status_t
1639 bfs_rewind_dir(fs_volume* /*_volume*/, fs_vnode* /*node*/, void* _cookie)
1640 {
1641 	FUNCTION();
1642 	TreeIterator* iterator = (TreeIterator*)_cookie;
1643 
1644 	return iterator->Rewind();
1645 }
1646 
1647 
1648 static status_t
1649 bfs_close_dir(fs_volume* /*_volume*/, fs_vnode* /*node*/, void* /*_cookie*/)
1650 {
1651 	FUNCTION();
1652 	return B_OK;
1653 }
1654 
1655 
1656 static status_t
1657 bfs_free_dir_cookie(fs_volume* _volume, fs_vnode* node, void* _cookie)
1658 {
1659 	delete (TreeIterator*)_cookie;
1660 	return B_OK;
1661 }
1662 
1663 
1664 //	#pragma mark - Attribute functions
1665 
1666 
1667 static status_t
1668 bfs_open_attr_dir(fs_volume* _volume, fs_vnode* _node, void** _cookie)
1669 {
1670 	Inode* inode = (Inode*)_node->private_node;
1671 
1672 	FUNCTION();
1673 
1674 	AttributeIterator* iterator = new(std::nothrow) AttributeIterator(inode);
1675 	if (iterator == NULL)
1676 		RETURN_ERROR(B_NO_MEMORY);
1677 
1678 	*_cookie = iterator;
1679 	return B_OK;
1680 }
1681 
1682 
1683 static status_t
1684 bfs_close_attr_dir(fs_volume* _volume, fs_vnode* node, void* cookie)
1685 {
1686 	FUNCTION();
1687 	return B_OK;
1688 }
1689 
1690 
1691 static status_t
1692 bfs_free_attr_dir_cookie(fs_volume* _volume, fs_vnode* node, void* _cookie)
1693 {
1694 	FUNCTION();
1695 	AttributeIterator* iterator = (AttributeIterator*)_cookie;
1696 
1697 	delete iterator;
1698 	return B_OK;
1699 }
1700 
1701 
1702 static status_t
1703 bfs_rewind_attr_dir(fs_volume* _volume, fs_vnode* _node, void* _cookie)
1704 {
1705 	FUNCTION();
1706 
1707 	AttributeIterator* iterator = (AttributeIterator*)_cookie;
1708 	RETURN_ERROR(iterator->Rewind());
1709 }
1710 
1711 
1712 static status_t
1713 bfs_read_attr_dir(fs_volume* _volume, fs_vnode* node, void* _cookie,
1714 	struct dirent* dirent, size_t bufferSize, uint32* _num)
1715 {
1716 	FUNCTION();
1717 	AttributeIterator* iterator = (AttributeIterator*)_cookie;
1718 
1719 	uint32 type;
1720 	size_t length;
1721 	status_t status = iterator->GetNext(dirent->d_name, &length, &type,
1722 		&dirent->d_ino);
1723 	if (status == B_ENTRY_NOT_FOUND) {
1724 		*_num = 0;
1725 		return B_OK;
1726 	} else if (status != B_OK) {
1727 		RETURN_ERROR(status);
1728 	}
1729 
1730 	Volume* volume = (Volume*)_volume->private_volume;
1731 
1732 	dirent->d_dev = volume->ID();
1733 	dirent->d_reclen = sizeof(struct dirent) + length;
1734 
1735 	*_num = 1;
1736 	return B_OK;
1737 }
1738 
1739 
1740 static status_t
1741 bfs_create_attr(fs_volume* _volume, fs_vnode* _node, const char* name,
1742 	uint32 type, int openMode, void** _cookie)
1743 {
1744 	FUNCTION();
1745 
1746 	Volume* volume = (Volume*)_volume->private_volume;
1747 	if (volume->IsReadOnly())
1748 		return B_READ_ONLY_DEVICE;
1749 
1750 	Inode* inode = (Inode*)_node->private_node;
1751 	Attribute attribute(inode);
1752 
1753 	return attribute.Create(name, type, openMode, (attr_cookie**)_cookie);
1754 }
1755 
1756 
1757 static status_t
1758 bfs_open_attr(fs_volume* _volume, fs_vnode* _node, const char* name,
1759 	int openMode, void** _cookie)
1760 {
1761 	FUNCTION();
1762 
1763 	Inode* inode = (Inode*)_node->private_node;
1764 	Attribute attribute(inode);
1765 
1766 	return attribute.Open(name, openMode, (attr_cookie**)_cookie);
1767 }
1768 
1769 
1770 static status_t
1771 bfs_close_attr(fs_volume* _volume, fs_vnode* _file, void* cookie)
1772 {
1773 	return B_OK;
1774 }
1775 
1776 
1777 static status_t
1778 bfs_free_attr_cookie(fs_volume* _volume, fs_vnode* _file, void* cookie)
1779 {
1780 	delete (attr_cookie*)cookie;
1781 	return B_OK;
1782 }
1783 
1784 
1785 static status_t
1786 bfs_read_attr(fs_volume* _volume, fs_vnode* _file, void* _cookie, off_t pos,
1787 	void* buffer, size_t* _length)
1788 {
1789 	FUNCTION();
1790 
1791 	attr_cookie* cookie = (attr_cookie*)_cookie;
1792 	Inode* inode = (Inode*)_file->private_node;
1793 
1794 	Attribute attribute(inode, cookie);
1795 
1796 	return attribute.Read(cookie, pos, (uint8*)buffer, _length);
1797 }
1798 
1799 
1800 static status_t
1801 bfs_write_attr(fs_volume* _volume, fs_vnode* _file, void* _cookie,
1802 	off_t pos, const void* buffer, size_t* _length)
1803 {
1804 	FUNCTION();
1805 
1806 	attr_cookie* cookie = (attr_cookie*)_cookie;
1807 	Volume* volume = (Volume*)_volume->private_volume;
1808 	Inode* inode = (Inode*)_file->private_node;
1809 
1810 	Transaction transaction(volume, inode->BlockNumber());
1811 	Attribute attribute(inode, cookie);
1812 
1813 	bool created;
1814 	status_t status = attribute.Write(transaction, cookie, pos,
1815 		(const uint8*)buffer, _length, &created);
1816 	if (status == B_OK) {
1817 		status = transaction.Done();
1818 		if (status == B_OK) {
1819 			notify_attribute_changed(volume->ID(), inode->ID(), cookie->name,
1820 				created ? B_ATTR_CREATED : B_ATTR_CHANGED);
1821 			notify_stat_changed(volume->ID(), inode->ID(), B_STAT_CHANGE_TIME);
1822 		}
1823 	}
1824 
1825 	return status;
1826 }
1827 
1828 
1829 static status_t
1830 bfs_read_attr_stat(fs_volume* _volume, fs_vnode* _file, void* _cookie,
1831 	struct stat* stat)
1832 {
1833 	FUNCTION();
1834 
1835 	attr_cookie* cookie = (attr_cookie*)_cookie;
1836 	Inode* inode = (Inode*)_file->private_node;
1837 
1838 	Attribute attribute(inode, cookie);
1839 
1840 	return attribute.Stat(*stat);
1841 }
1842 
1843 
1844 static status_t
1845 bfs_write_attr_stat(fs_volume* _volume, fs_vnode* file, void* cookie,
1846 	const struct stat* stat, int statMask)
1847 {
1848 	// TODO: Implement (at least setting the size)!
1849 	return EOPNOTSUPP;
1850 }
1851 
1852 
1853 static status_t
1854 bfs_rename_attr(fs_volume* _volume, fs_vnode* fromFile, const char* fromName,
1855 	fs_vnode* toFile, const char* toName)
1856 {
1857 	FUNCTION_START(("name = \"%s\", to = \"%s\"\n", fromName, toName));
1858 
1859 	// TODO: implement bfs_rename_attr()!
1860 	// There will probably be an API to move one attribute to another file,
1861 	// making that function much more complicated - oh joy ;-)
1862 
1863 	return EOPNOTSUPP;
1864 }
1865 
1866 
1867 static status_t
1868 bfs_remove_attr(fs_volume* _volume, fs_vnode* _node, const char* name)
1869 {
1870 	FUNCTION_START(("name = \"%s\"\n", name));
1871 
1872 	Volume* volume = (Volume*)_volume->private_volume;
1873 	Inode* inode = (Inode*)_node->private_node;
1874 
1875 	status_t status = inode->CheckPermissions(W_OK);
1876 	if (status != B_OK)
1877 		return status;
1878 
1879 	Transaction transaction(volume, inode->BlockNumber());
1880 
1881 	status = inode->RemoveAttribute(transaction, name);
1882 	if (status == B_OK)
1883 		status = transaction.Done();
1884 	if (status == B_OK) {
1885 		notify_attribute_changed(volume->ID(), inode->ID(), name,
1886 			B_ATTR_REMOVED);
1887 	}
1888 
1889 	return status;
1890 }
1891 
1892 
1893 //	#pragma mark - Special Nodes
1894 
1895 
1896 status_t
1897 bfs_create_special_node(fs_volume* _volume, fs_vnode* _directory,
1898 	const char* name, fs_vnode* subVnode, mode_t mode, uint32 flags,
1899 	fs_vnode* _superVnode, ino_t* _nodeID)
1900 {
1901 	// no need to support entry-less nodes
1902 	if (name == NULL)
1903 		return B_UNSUPPORTED;
1904 
1905 	FUNCTION_START(("name = \"%s\", mode = %d, flags = 0x%lx, subVnode: %p\n",
1906 		name, mode, flags, subVnode));
1907 
1908 	Volume* volume = (Volume*)_volume->private_volume;
1909 	Inode* directory = (Inode*)_directory->private_node;
1910 
1911 	if (volume->IsReadOnly())
1912 		return B_READ_ONLY_DEVICE;
1913 
1914 	if (!directory->IsDirectory())
1915 		RETURN_ERROR(B_BAD_TYPE);
1916 
1917 	status_t status = directory->CheckPermissions(W_OK);
1918 	if (status < B_OK)
1919 		RETURN_ERROR(status);
1920 
1921 	Transaction transaction(volume, directory->BlockNumber());
1922 
1923 	off_t id;
1924 	Inode* inode;
1925 	status = Inode::Create(transaction, directory, name, mode, O_EXCL, 0, NULL,
1926 		&id, &inode, subVnode ? subVnode->ops : NULL, flags);
1927 	if (status == B_OK) {
1928 		_superVnode->private_node = inode;
1929 		_superVnode->ops = &gBFSVnodeOps;
1930 		*_nodeID = id;
1931 
1932 		entry_cache_add(volume->ID(), directory->ID(), name, id);
1933 
1934 		status = transaction.Done();
1935 		if (status == B_OK)
1936 			notify_entry_created(volume->ID(), directory->ID(), name, id);
1937 		else
1938 			entry_cache_remove(volume->ID(), directory->ID(), name);
1939 	}
1940 
1941 	return status;
1942 }
1943 
1944 
1945 //	#pragma mark - Index functions
1946 
1947 
1948 static status_t
1949 bfs_open_index_dir(fs_volume* _volume, void** _cookie)
1950 {
1951 	FUNCTION();
1952 
1953 	Volume* volume = (Volume*)_volume->private_volume;
1954 
1955 	if (volume->IndicesNode() == NULL) {
1956 		// This volume does not have any indices
1957 		RETURN_ERROR(B_ENTRY_NOT_FOUND);
1958 	}
1959 
1960 	// Since the indices root node is just a directory, and we are storing
1961 	// a pointer to it in our Volume object, we can just use the directory
1962 	// traversal functions.
1963 	// In fact we're storing it in the Volume object for that reason.
1964 
1965 	fs_vnode indicesNode;
1966 	indicesNode.private_node = volume->IndicesNode();
1967 
1968 	RETURN_ERROR(bfs_open_dir(_volume, &indicesNode, _cookie));
1969 }
1970 
1971 
1972 static status_t
1973 bfs_close_index_dir(fs_volume* _volume, void* _cookie)
1974 {
1975 	FUNCTION();
1976 
1977 	Volume* volume = (Volume*)_volume->private_volume;
1978 
1979 	fs_vnode indicesNode;
1980 	indicesNode.private_node = volume->IndicesNode();
1981 
1982 	RETURN_ERROR(bfs_close_dir(_volume, &indicesNode, _cookie));
1983 }
1984 
1985 
1986 static status_t
1987 bfs_free_index_dir_cookie(fs_volume* _volume, void* _cookie)
1988 {
1989 	FUNCTION();
1990 
1991 	Volume* volume = (Volume*)_volume->private_volume;
1992 
1993 	fs_vnode indicesNode;
1994 	indicesNode.private_node = volume->IndicesNode();
1995 
1996 	RETURN_ERROR(bfs_free_dir_cookie(_volume, &indicesNode, _cookie));
1997 }
1998 
1999 
2000 static status_t
2001 bfs_rewind_index_dir(fs_volume* _volume, void* _cookie)
2002 {
2003 	FUNCTION();
2004 
2005 	Volume* volume = (Volume*)_volume->private_volume;
2006 
2007 	fs_vnode indicesNode;
2008 	indicesNode.private_node = volume->IndicesNode();
2009 
2010 	RETURN_ERROR(bfs_rewind_dir(_volume, &indicesNode, _cookie));
2011 }
2012 
2013 
2014 static status_t
2015 bfs_read_index_dir(fs_volume* _volume, void* _cookie, struct dirent* dirent,
2016 	size_t bufferSize, uint32* _num)
2017 {
2018 	FUNCTION();
2019 
2020 	Volume* volume = (Volume*)_volume->private_volume;
2021 
2022 	fs_vnode indicesNode;
2023 	indicesNode.private_node = volume->IndicesNode();
2024 
2025 	RETURN_ERROR(bfs_read_dir(_volume, &indicesNode, _cookie, dirent,
2026 		bufferSize, _num));
2027 }
2028 
2029 
2030 static status_t
2031 bfs_create_index(fs_volume* _volume, const char* name, uint32 type,
2032 	uint32 flags)
2033 {
2034 	FUNCTION_START(("name = \"%s\", type = %ld, flags = %ld\n", name, type, flags));
2035 
2036 	Volume* volume = (Volume*)_volume->private_volume;
2037 
2038 	if (volume->IsReadOnly())
2039 		return B_READ_ONLY_DEVICE;
2040 
2041 	// only root users are allowed to create indices
2042 	if (geteuid() != 0)
2043 		return B_NOT_ALLOWED;
2044 
2045 	Transaction transaction(volume, volume->Indices());
2046 
2047 	Index index(volume);
2048 	status_t status = index.Create(transaction, name, type);
2049 
2050 	if (status == B_OK)
2051 		status = transaction.Done();
2052 
2053 	RETURN_ERROR(status);
2054 }
2055 
2056 
2057 static status_t
2058 bfs_remove_index(fs_volume* _volume, const char* name)
2059 {
2060 	FUNCTION();
2061 
2062 	Volume* volume = (Volume*)_volume->private_volume;
2063 
2064 	if (volume->IsReadOnly())
2065 		return B_READ_ONLY_DEVICE;
2066 
2067 	// only root users are allowed to remove indices
2068 	if (geteuid() != 0)
2069 		return B_NOT_ALLOWED;
2070 
2071 	Inode* indices = volume->IndicesNode();
2072 	if (indices == NULL)
2073 		return B_ENTRY_NOT_FOUND;
2074 
2075 	Transaction transaction(volume, volume->Indices());
2076 
2077 	status_t status = indices->Remove(transaction, name);
2078 	if (status == B_OK)
2079 		status = transaction.Done();
2080 
2081 	RETURN_ERROR(status);
2082 }
2083 
2084 
2085 static status_t
2086 bfs_stat_index(fs_volume* _volume, const char* name, struct stat* stat)
2087 {
2088 	FUNCTION_START(("name = %s\n", name));
2089 
2090 	Volume* volume = (Volume*)_volume->private_volume;
2091 
2092 	Index index(volume);
2093 	status_t status = index.SetTo(name);
2094 	if (status < B_OK)
2095 		RETURN_ERROR(status);
2096 
2097 	bfs_inode& node = index.Node()->Node();
2098 
2099 	stat->st_type = index.Type();
2100 	stat->st_mode = node.Mode();
2101 
2102 	stat->st_size = node.data.Size();
2103 	stat->st_blocks = index.Node()->AllocatedSize() / 512;
2104 
2105 	stat->st_nlink = 1;
2106 	stat->st_blksize = 65536;
2107 
2108 	stat->st_uid = node.UserID();
2109 	stat->st_gid = node.GroupID();
2110 
2111 	fill_stat_time(node, *stat);
2112 
2113 	return B_OK;
2114 }
2115 
2116 
2117 //	#pragma mark - Query functions
2118 
2119 
2120 static status_t
2121 bfs_open_query(fs_volume* _volume, const char* queryString, uint32 flags,
2122 	port_id port, uint32 token, void** _cookie)
2123 {
2124 	FUNCTION_START(("bfs_open_query(\"%s\", flags = %lu, port_id = %ld, token = %ld)\n",
2125 		queryString, flags, port, token));
2126 
2127 	Volume* volume = (Volume*)_volume->private_volume;
2128 
2129 	Expression* expression = new(std::nothrow) Expression((char*)queryString);
2130 	if (expression == NULL)
2131 		RETURN_ERROR(B_NO_MEMORY);
2132 
2133 	if (expression->InitCheck() < B_OK) {
2134 		INFORM(("Could not parse query \"%s\", stopped at: \"%s\"\n",
2135 			queryString, expression->Position()));
2136 
2137 		delete expression;
2138 		RETURN_ERROR(B_BAD_VALUE);
2139 	}
2140 
2141 	Query* query = new(std::nothrow) Query(volume, expression, flags);
2142 	if (query == NULL) {
2143 		delete expression;
2144 		RETURN_ERROR(B_NO_MEMORY);
2145 	}
2146 
2147 	if (flags & B_LIVE_QUERY)
2148 		query->SetLiveMode(port, token);
2149 
2150 	*_cookie = (void*)query;
2151 
2152 	return B_OK;
2153 }
2154 
2155 
2156 static status_t
2157 bfs_close_query(fs_volume* _volume, void* cookie)
2158 {
2159 	FUNCTION();
2160 	return B_OK;
2161 }
2162 
2163 
2164 static status_t
2165 bfs_free_query_cookie(fs_volume* _volume, void* cookie)
2166 {
2167 	FUNCTION();
2168 
2169 	Query* query = (Query*)cookie;
2170 	Expression* expression = query->GetExpression();
2171 	delete query;
2172 	delete expression;
2173 
2174 	return B_OK;
2175 }
2176 
2177 
2178 static status_t
2179 bfs_read_query(fs_volume* /*_volume*/, void* cookie, struct dirent* dirent,
2180 	size_t bufferSize, uint32* _num)
2181 {
2182 	FUNCTION();
2183 	Query* query = (Query*)cookie;
2184 	status_t status = query->GetNextEntry(dirent, bufferSize);
2185 	if (status == B_OK)
2186 		*_num = 1;
2187 	else if (status == B_ENTRY_NOT_FOUND)
2188 		*_num = 0;
2189 	else
2190 		return status;
2191 
2192 	return B_OK;
2193 }
2194 
2195 
2196 static status_t
2197 bfs_rewind_query(fs_volume* /*_volume*/, void* cookie)
2198 {
2199 	FUNCTION();
2200 
2201 	Query* query = (Query*)cookie;
2202 	return query->Rewind();
2203 }
2204 
2205 
2206 //	#pragma mark -
2207 
2208 
2209 static uint32
2210 bfs_get_supported_operations(partition_data* partition, uint32 mask)
2211 {
2212 	// TODO: We should at least check the partition size.
2213 	return B_DISK_SYSTEM_SUPPORTS_INITIALIZING
2214 		| B_DISK_SYSTEM_SUPPORTS_CONTENT_NAME
2215 		| B_DISK_SYSTEM_SUPPORTS_WRITING;
2216 }
2217 
2218 
2219 static status_t
2220 bfs_initialize(int fd, partition_id partitionID, const char* name,
2221 	const char* parameterString, off_t /*partitionSize*/, disk_job_id job)
2222 {
2223 	// check name
2224 	status_t status = check_volume_name(name);
2225 	if (status != B_OK)
2226 		return status;
2227 
2228 	// parse parameters
2229 	initialize_parameters parameters;
2230 	status = parse_initialize_parameters(parameterString, parameters);
2231 	if (status != B_OK)
2232 		return status;
2233 
2234 	update_disk_device_job_progress(job, 0);
2235 
2236 	// initialize the volume
2237 	Volume volume(NULL);
2238 	status = volume.Initialize(fd, name, parameters.blockSize,
2239 		parameters.flags);
2240 	if (status < B_OK) {
2241 		INFORM(("Initializing volume failed: %s\n", strerror(status)));
2242 		return status;
2243 	}
2244 
2245 	// rescan partition
2246 	status = scan_partition(partitionID);
2247 	if (status != B_OK)
2248 		return status;
2249 
2250 	update_disk_device_job_progress(job, 1);
2251 
2252 	// print some info, if desired
2253 	if (parameters.verbose) {
2254 		disk_super_block super = volume.SuperBlock();
2255 
2256 		INFORM(("Disk was initialized successfully.\n"));
2257 		INFORM(("\tname: \"%s\"\n", super.name));
2258 		INFORM(("\tnum blocks: %" B_PRIdOFF "\n", super.NumBlocks()));
2259 		INFORM(("\tused blocks: %" B_PRIdOFF "\n", super.UsedBlocks()));
2260 		INFORM(("\tblock size: %u bytes\n", (unsigned)super.BlockSize()));
2261 		INFORM(("\tnum allocation groups: %d\n",
2262 			(int)super.AllocationGroups()));
2263 		INFORM(("\tallocation group size: %ld blocks\n",
2264 			1L << super.AllocationGroupShift()));
2265 		INFORM(("\tlog size: %u blocks\n", super.log_blocks.Length()));
2266 	}
2267 
2268 	return B_OK;
2269 }
2270 
2271 
2272 //	#pragma mark -
2273 
2274 
2275 static status_t
2276 bfs_std_ops(int32 op, ...)
2277 {
2278 	switch (op) {
2279 		case B_MODULE_INIT:
2280 #ifdef BFS_DEBUGGER_COMMANDS
2281 			add_debugger_commands();
2282 #endif
2283 			return B_OK;
2284 		case B_MODULE_UNINIT:
2285 #ifdef BFS_DEBUGGER_COMMANDS
2286 			remove_debugger_commands();
2287 #endif
2288 			return B_OK;
2289 
2290 		default:
2291 			return B_ERROR;
2292 	}
2293 }
2294 
2295 fs_volume_ops gBFSVolumeOps = {
2296 	&bfs_unmount,
2297 	&bfs_read_fs_stat,
2298 	&bfs_write_fs_stat,
2299 	&bfs_sync,
2300 	&bfs_get_vnode,
2301 
2302 	/* index directory & index operations */
2303 	&bfs_open_index_dir,
2304 	&bfs_close_index_dir,
2305 	&bfs_free_index_dir_cookie,
2306 	&bfs_read_index_dir,
2307 	&bfs_rewind_index_dir,
2308 
2309 	&bfs_create_index,
2310 	&bfs_remove_index,
2311 	&bfs_stat_index,
2312 
2313 	/* query operations */
2314 	&bfs_open_query,
2315 	&bfs_close_query,
2316 	&bfs_free_query_cookie,
2317 	&bfs_read_query,
2318 	&bfs_rewind_query,
2319 };
2320 
2321 fs_vnode_ops gBFSVnodeOps = {
2322 	/* vnode operations */
2323 	&bfs_lookup,
2324 	&bfs_get_vnode_name,
2325 	&bfs_put_vnode,
2326 	&bfs_remove_vnode,
2327 
2328 	/* VM file access */
2329 	&bfs_can_page,
2330 	&bfs_read_pages,
2331 	&bfs_write_pages,
2332 
2333 	&bfs_io,
2334 	NULL,	// cancel_io()
2335 
2336 	&bfs_get_file_map,
2337 
2338 	&bfs_ioctl,
2339 	&bfs_set_flags,
2340 	NULL,	// fs_select
2341 	NULL,	// fs_deselect
2342 	&bfs_fsync,
2343 
2344 	&bfs_read_link,
2345 	&bfs_create_symlink,
2346 
2347 	&bfs_link,
2348 	&bfs_unlink,
2349 	&bfs_rename,
2350 
2351 	&bfs_access,
2352 	&bfs_read_stat,
2353 	&bfs_write_stat,
2354 
2355 	/* file operations */
2356 	&bfs_create,
2357 	&bfs_open,
2358 	&bfs_close,
2359 	&bfs_free_cookie,
2360 	&bfs_read,
2361 	&bfs_write,
2362 
2363 	/* directory operations */
2364 	&bfs_create_dir,
2365 	&bfs_remove_dir,
2366 	&bfs_open_dir,
2367 	&bfs_close_dir,
2368 	&bfs_free_dir_cookie,
2369 	&bfs_read_dir,
2370 	&bfs_rewind_dir,
2371 
2372 	/* attribute directory operations */
2373 	&bfs_open_attr_dir,
2374 	&bfs_close_attr_dir,
2375 	&bfs_free_attr_dir_cookie,
2376 	&bfs_read_attr_dir,
2377 	&bfs_rewind_attr_dir,
2378 
2379 	/* attribute operations */
2380 	&bfs_create_attr,
2381 	&bfs_open_attr,
2382 	&bfs_close_attr,
2383 	&bfs_free_attr_cookie,
2384 	&bfs_read_attr,
2385 	&bfs_write_attr,
2386 
2387 	&bfs_read_attr_stat,
2388 	&bfs_write_attr_stat,
2389 	&bfs_rename_attr,
2390 	&bfs_remove_attr,
2391 
2392 	/* special nodes */
2393 	&bfs_create_special_node
2394 };
2395 
2396 static file_system_module_info sBeFileSystem = {
2397 	{
2398 		"file_systems/bfs" B_CURRENT_FS_API_VERSION,
2399 		0,
2400 		bfs_std_ops,
2401 	},
2402 
2403 	"bfs",						// short_name
2404 	"Be File System",			// pretty_name
2405 
2406 	// DDM flags
2407 	0
2408 //	| B_DISK_SYSTEM_SUPPORTS_CHECKING
2409 //	| B_DISK_SYSTEM_SUPPORTS_REPAIRING
2410 //	| B_DISK_SYSTEM_SUPPORTS_RESIZING
2411 //	| B_DISK_SYSTEM_SUPPORTS_MOVING
2412 //	| B_DISK_SYSTEM_SUPPORTS_SETTING_CONTENT_NAME
2413 //	| B_DISK_SYSTEM_SUPPORTS_SETTING_CONTENT_PARAMETERS
2414 	| B_DISK_SYSTEM_SUPPORTS_INITIALIZING
2415 	| B_DISK_SYSTEM_SUPPORTS_CONTENT_NAME
2416 //	| B_DISK_SYSTEM_SUPPORTS_DEFRAGMENTING
2417 //	| B_DISK_SYSTEM_SUPPORTS_DEFRAGMENTING_WHILE_MOUNTED
2418 //	| B_DISK_SYSTEM_SUPPORTS_CHECKING_WHILE_MOUNTED
2419 //	| B_DISK_SYSTEM_SUPPORTS_REPAIRING_WHILE_MOUNTED
2420 //	| B_DISK_SYSTEM_SUPPORTS_RESIZING_WHILE_MOUNTED
2421 //	| B_DISK_SYSTEM_SUPPORTS_MOVING_WHILE_MOUNTED
2422 //	| B_DISK_SYSTEM_SUPPORTS_SETTING_CONTENT_NAME_WHILE_MOUNTED
2423 //	| B_DISK_SYSTEM_SUPPORTS_SETTING_CONTENT_PARAMETERS_WHILE_MOUNTED
2424 	| B_DISK_SYSTEM_SUPPORTS_WRITING
2425 	,
2426 
2427 	// scanning
2428 	bfs_identify_partition,
2429 	bfs_scan_partition,
2430 	bfs_free_identify_partition_cookie,
2431 	NULL,	// free_partition_content_cookie()
2432 
2433 	&bfs_mount,
2434 
2435 	/* capability querying operations */
2436 	&bfs_get_supported_operations,
2437 
2438 	NULL,	// validate_resize
2439 	NULL,	// validate_move
2440 	NULL,	// validate_set_content_name
2441 	NULL,	// validate_set_content_parameters
2442 	NULL,	// validate_initialize,
2443 
2444 	/* shadow partition modification */
2445 	NULL,	// shadow_changed
2446 
2447 	/* writing */
2448 	NULL,	// defragment
2449 	NULL,	// repair
2450 	NULL,	// resize
2451 	NULL,	// move
2452 	NULL,	// set_content_name
2453 	NULL,	// set_content_parameters
2454 	bfs_initialize,
2455 };
2456 
2457 module_info* modules[] = {
2458 	(module_info*)&sBeFileSystem,
2459 	NULL,
2460 };
2461