xref: /openbmc/linux/fs/ntfs/inode.c (revision d5cb9783536a41df9f9cba5b0a1d78047ed787f7)
1 /**
2  * inode.c - NTFS kernel inode handling. Part of the Linux-NTFS project.
3  *
4  * Copyright (c) 2001-2005 Anton Altaparmakov
5  *
6  * This program/include file is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU General Public License as published
8  * by the Free Software Foundation; either version 2 of the License, or
9  * (at your option) any later version.
10  *
11  * This program/include file is distributed in the hope that it will be
12  * useful, but WITHOUT ANY WARRANTY; without even the implied warranty
13  * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
14  * GNU General Public License for more details.
15  *
16  * You should have received a copy of the GNU General Public License
17  * along with this program (in the main directory of the Linux-NTFS
18  * distribution in the file COPYING); if not, write to the Free Software
19  * Foundation,Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
20  */
21 
22 #include <linux/pagemap.h>
23 #include <linux/buffer_head.h>
24 #include <linux/smp_lock.h>
25 #include <linux/quotaops.h>
26 #include <linux/mount.h>
27 
28 #include "aops.h"
29 #include "dir.h"
30 #include "debug.h"
31 #include "inode.h"
32 #include "attrib.h"
33 #include "lcnalloc.h"
34 #include "malloc.h"
35 #include "mft.h"
36 #include "time.h"
37 #include "ntfs.h"
38 
39 /**
40  * ntfs_test_inode - compare two (possibly fake) inodes for equality
41  * @vi:		vfs inode which to test
42  * @na:		ntfs attribute which is being tested with
43  *
44  * Compare the ntfs attribute embedded in the ntfs specific part of the vfs
45  * inode @vi for equality with the ntfs attribute @na.
46  *
47  * If searching for the normal file/directory inode, set @na->type to AT_UNUSED.
48  * @na->name and @na->name_len are then ignored.
49  *
50  * Return 1 if the attributes match and 0 if not.
51  *
52  * NOTE: This function runs with the inode_lock spin lock held so it is not
53  * allowed to sleep.
54  */
55 int ntfs_test_inode(struct inode *vi, ntfs_attr *na)
56 {
57 	ntfs_inode *ni;
58 
59 	if (vi->i_ino != na->mft_no)
60 		return 0;
61 	ni = NTFS_I(vi);
62 	/* If !NInoAttr(ni), @vi is a normal file or directory inode. */
63 	if (likely(!NInoAttr(ni))) {
64 		/* If not looking for a normal inode this is a mismatch. */
65 		if (unlikely(na->type != AT_UNUSED))
66 			return 0;
67 	} else {
68 		/* A fake inode describing an attribute. */
69 		if (ni->type != na->type)
70 			return 0;
71 		if (ni->name_len != na->name_len)
72 			return 0;
73 		if (na->name_len && memcmp(ni->name, na->name,
74 				na->name_len * sizeof(ntfschar)))
75 			return 0;
76 	}
77 	/* Match! */
78 	return 1;
79 }
80 
81 /**
82  * ntfs_init_locked_inode - initialize an inode
83  * @vi:		vfs inode to initialize
84  * @na:		ntfs attribute which to initialize @vi to
85  *
86  * Initialize the vfs inode @vi with the values from the ntfs attribute @na in
87  * order to enable ntfs_test_inode() to do its work.
88  *
89  * If initializing the normal file/directory inode, set @na->type to AT_UNUSED.
90  * In that case, @na->name and @na->name_len should be set to NULL and 0,
91  * respectively. Although that is not strictly necessary as
92  * ntfs_read_inode_locked() will fill them in later.
93  *
94  * Return 0 on success and -errno on error.
95  *
96  * NOTE: This function runs with the inode_lock spin lock held so it is not
97  * allowed to sleep. (Hence the GFP_ATOMIC allocation.)
98  */
99 static int ntfs_init_locked_inode(struct inode *vi, ntfs_attr *na)
100 {
101 	ntfs_inode *ni = NTFS_I(vi);
102 
103 	vi->i_ino = na->mft_no;
104 
105 	ni->type = na->type;
106 	if (na->type == AT_INDEX_ALLOCATION)
107 		NInoSetMstProtected(ni);
108 
109 	ni->name = na->name;
110 	ni->name_len = na->name_len;
111 
112 	/* If initializing a normal inode, we are done. */
113 	if (likely(na->type == AT_UNUSED)) {
114 		BUG_ON(na->name);
115 		BUG_ON(na->name_len);
116 		return 0;
117 	}
118 
119 	/* It is a fake inode. */
120 	NInoSetAttr(ni);
121 
122 	/*
123 	 * We have I30 global constant as an optimization as it is the name
124 	 * in >99.9% of named attributes! The other <0.1% incur a GFP_ATOMIC
125 	 * allocation but that is ok. And most attributes are unnamed anyway,
126 	 * thus the fraction of named attributes with name != I30 is actually
127 	 * absolutely tiny.
128 	 */
129 	if (na->name_len && na->name != I30) {
130 		unsigned int i;
131 
132 		BUG_ON(!na->name);
133 		i = na->name_len * sizeof(ntfschar);
134 		ni->name = (ntfschar*)kmalloc(i + sizeof(ntfschar), GFP_ATOMIC);
135 		if (!ni->name)
136 			return -ENOMEM;
137 		memcpy(ni->name, na->name, i);
138 		ni->name[i] = 0;
139 	}
140 	return 0;
141 }
142 
143 typedef int (*set_t)(struct inode *, void *);
144 static int ntfs_read_locked_inode(struct inode *vi);
145 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi);
146 static int ntfs_read_locked_index_inode(struct inode *base_vi,
147 		struct inode *vi);
148 
149 /**
150  * ntfs_iget - obtain a struct inode corresponding to a specific normal inode
151  * @sb:		super block of mounted volume
152  * @mft_no:	mft record number / inode number to obtain
153  *
154  * Obtain the struct inode corresponding to a specific normal inode (i.e. a
155  * file or directory).
156  *
157  * If the inode is in the cache, it is just returned with an increased
158  * reference count. Otherwise, a new struct inode is allocated and initialized,
159  * and finally ntfs_read_locked_inode() is called to read in the inode and
160  * fill in the remainder of the inode structure.
161  *
162  * Return the struct inode on success. Check the return value with IS_ERR() and
163  * if true, the function failed and the error code is obtained from PTR_ERR().
164  */
165 struct inode *ntfs_iget(struct super_block *sb, unsigned long mft_no)
166 {
167 	struct inode *vi;
168 	ntfs_attr na;
169 	int err;
170 
171 	na.mft_no = mft_no;
172 	na.type = AT_UNUSED;
173 	na.name = NULL;
174 	na.name_len = 0;
175 
176 	vi = iget5_locked(sb, mft_no, (test_t)ntfs_test_inode,
177 			(set_t)ntfs_init_locked_inode, &na);
178 	if (unlikely(!vi))
179 		return ERR_PTR(-ENOMEM);
180 
181 	err = 0;
182 
183 	/* If this is a freshly allocated inode, need to read it now. */
184 	if (vi->i_state & I_NEW) {
185 		err = ntfs_read_locked_inode(vi);
186 		unlock_new_inode(vi);
187 	}
188 	/*
189 	 * There is no point in keeping bad inodes around if the failure was
190 	 * due to ENOMEM. We want to be able to retry again later.
191 	 */
192 	if (unlikely(err == -ENOMEM)) {
193 		iput(vi);
194 		vi = ERR_PTR(err);
195 	}
196 	return vi;
197 }
198 
199 /**
200  * ntfs_attr_iget - obtain a struct inode corresponding to an attribute
201  * @base_vi:	vfs base inode containing the attribute
202  * @type:	attribute type
203  * @name:	Unicode name of the attribute (NULL if unnamed)
204  * @name_len:	length of @name in Unicode characters (0 if unnamed)
205  *
206  * Obtain the (fake) struct inode corresponding to the attribute specified by
207  * @type, @name, and @name_len, which is present in the base mft record
208  * specified by the vfs inode @base_vi.
209  *
210  * If the attribute inode is in the cache, it is just returned with an
211  * increased reference count. Otherwise, a new struct inode is allocated and
212  * initialized, and finally ntfs_read_locked_attr_inode() is called to read the
213  * attribute and fill in the inode structure.
214  *
215  * Note, for index allocation attributes, you need to use ntfs_index_iget()
216  * instead of ntfs_attr_iget() as working with indices is a lot more complex.
217  *
218  * Return the struct inode of the attribute inode on success. Check the return
219  * value with IS_ERR() and if true, the function failed and the error code is
220  * obtained from PTR_ERR().
221  */
222 struct inode *ntfs_attr_iget(struct inode *base_vi, ATTR_TYPE type,
223 		ntfschar *name, u32 name_len)
224 {
225 	struct inode *vi;
226 	ntfs_attr na;
227 	int err;
228 
229 	/* Make sure no one calls ntfs_attr_iget() for indices. */
230 	BUG_ON(type == AT_INDEX_ALLOCATION);
231 
232 	na.mft_no = base_vi->i_ino;
233 	na.type = type;
234 	na.name = name;
235 	na.name_len = name_len;
236 
237 	vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
238 			(set_t)ntfs_init_locked_inode, &na);
239 	if (unlikely(!vi))
240 		return ERR_PTR(-ENOMEM);
241 
242 	err = 0;
243 
244 	/* If this is a freshly allocated inode, need to read it now. */
245 	if (vi->i_state & I_NEW) {
246 		err = ntfs_read_locked_attr_inode(base_vi, vi);
247 		unlock_new_inode(vi);
248 	}
249 	/*
250 	 * There is no point in keeping bad attribute inodes around. This also
251 	 * simplifies things in that we never need to check for bad attribute
252 	 * inodes elsewhere.
253 	 */
254 	if (unlikely(err)) {
255 		iput(vi);
256 		vi = ERR_PTR(err);
257 	}
258 	return vi;
259 }
260 
261 /**
262  * ntfs_index_iget - obtain a struct inode corresponding to an index
263  * @base_vi:	vfs base inode containing the index related attributes
264  * @name:	Unicode name of the index
265  * @name_len:	length of @name in Unicode characters
266  *
267  * Obtain the (fake) struct inode corresponding to the index specified by @name
268  * and @name_len, which is present in the base mft record specified by the vfs
269  * inode @base_vi.
270  *
271  * If the index inode is in the cache, it is just returned with an increased
272  * reference count.  Otherwise, a new struct inode is allocated and
273  * initialized, and finally ntfs_read_locked_index_inode() is called to read
274  * the index related attributes and fill in the inode structure.
275  *
276  * Return the struct inode of the index inode on success. Check the return
277  * value with IS_ERR() and if true, the function failed and the error code is
278  * obtained from PTR_ERR().
279  */
280 struct inode *ntfs_index_iget(struct inode *base_vi, ntfschar *name,
281 		u32 name_len)
282 {
283 	struct inode *vi;
284 	ntfs_attr na;
285 	int err;
286 
287 	na.mft_no = base_vi->i_ino;
288 	na.type = AT_INDEX_ALLOCATION;
289 	na.name = name;
290 	na.name_len = name_len;
291 
292 	vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
293 			(set_t)ntfs_init_locked_inode, &na);
294 	if (unlikely(!vi))
295 		return ERR_PTR(-ENOMEM);
296 
297 	err = 0;
298 
299 	/* If this is a freshly allocated inode, need to read it now. */
300 	if (vi->i_state & I_NEW) {
301 		err = ntfs_read_locked_index_inode(base_vi, vi);
302 		unlock_new_inode(vi);
303 	}
304 	/*
305 	 * There is no point in keeping bad index inodes around.  This also
306 	 * simplifies things in that we never need to check for bad index
307 	 * inodes elsewhere.
308 	 */
309 	if (unlikely(err)) {
310 		iput(vi);
311 		vi = ERR_PTR(err);
312 	}
313 	return vi;
314 }
315 
316 struct inode *ntfs_alloc_big_inode(struct super_block *sb)
317 {
318 	ntfs_inode *ni;
319 
320 	ntfs_debug("Entering.");
321 	ni = kmem_cache_alloc(ntfs_big_inode_cache, SLAB_NOFS);
322 	if (likely(ni != NULL)) {
323 		ni->state = 0;
324 		return VFS_I(ni);
325 	}
326 	ntfs_error(sb, "Allocation of NTFS big inode structure failed.");
327 	return NULL;
328 }
329 
330 void ntfs_destroy_big_inode(struct inode *inode)
331 {
332 	ntfs_inode *ni = NTFS_I(inode);
333 
334 	ntfs_debug("Entering.");
335 	BUG_ON(ni->page);
336 	if (!atomic_dec_and_test(&ni->count))
337 		BUG();
338 	kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode));
339 }
340 
341 static inline ntfs_inode *ntfs_alloc_extent_inode(void)
342 {
343 	ntfs_inode *ni;
344 
345 	ntfs_debug("Entering.");
346 	ni = kmem_cache_alloc(ntfs_inode_cache, SLAB_NOFS);
347 	if (likely(ni != NULL)) {
348 		ni->state = 0;
349 		return ni;
350 	}
351 	ntfs_error(NULL, "Allocation of NTFS inode structure failed.");
352 	return NULL;
353 }
354 
355 static void ntfs_destroy_extent_inode(ntfs_inode *ni)
356 {
357 	ntfs_debug("Entering.");
358 	BUG_ON(ni->page);
359 	if (!atomic_dec_and_test(&ni->count))
360 		BUG();
361 	kmem_cache_free(ntfs_inode_cache, ni);
362 }
363 
364 /**
365  * __ntfs_init_inode - initialize ntfs specific part of an inode
366  * @sb:		super block of mounted volume
367  * @ni:		freshly allocated ntfs inode which to initialize
368  *
369  * Initialize an ntfs inode to defaults.
370  *
371  * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left
372  * untouched. Make sure to initialize them elsewhere.
373  *
374  * Return zero on success and -ENOMEM on error.
375  */
376 void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni)
377 {
378 	ntfs_debug("Entering.");
379 	rwlock_init(&ni->size_lock);
380 	ni->initialized_size = ni->allocated_size = 0;
381 	ni->seq_no = 0;
382 	atomic_set(&ni->count, 1);
383 	ni->vol = NTFS_SB(sb);
384 	ntfs_init_runlist(&ni->runlist);
385 	init_MUTEX(&ni->mrec_lock);
386 	ni->page = NULL;
387 	ni->page_ofs = 0;
388 	ni->attr_list_size = 0;
389 	ni->attr_list = NULL;
390 	ntfs_init_runlist(&ni->attr_list_rl);
391 	ni->itype.index.bmp_ino = NULL;
392 	ni->itype.index.block_size = 0;
393 	ni->itype.index.vcn_size = 0;
394 	ni->itype.index.collation_rule = 0;
395 	ni->itype.index.block_size_bits = 0;
396 	ni->itype.index.vcn_size_bits = 0;
397 	init_MUTEX(&ni->extent_lock);
398 	ni->nr_extents = 0;
399 	ni->ext.base_ntfs_ino = NULL;
400 }
401 
402 inline ntfs_inode *ntfs_new_extent_inode(struct super_block *sb,
403 		unsigned long mft_no)
404 {
405 	ntfs_inode *ni = ntfs_alloc_extent_inode();
406 
407 	ntfs_debug("Entering.");
408 	if (likely(ni != NULL)) {
409 		__ntfs_init_inode(sb, ni);
410 		ni->mft_no = mft_no;
411 		ni->type = AT_UNUSED;
412 		ni->name = NULL;
413 		ni->name_len = 0;
414 	}
415 	return ni;
416 }
417 
418 /**
419  * ntfs_is_extended_system_file - check if a file is in the $Extend directory
420  * @ctx:	initialized attribute search context
421  *
422  * Search all file name attributes in the inode described by the attribute
423  * search context @ctx and check if any of the names are in the $Extend system
424  * directory.
425  *
426  * Return values:
427  *	   1: file is in $Extend directory
428  *	   0: file is not in $Extend directory
429  *    -errno: failed to determine if the file is in the $Extend directory
430  */
431 static int ntfs_is_extended_system_file(ntfs_attr_search_ctx *ctx)
432 {
433 	int nr_links, err;
434 
435 	/* Restart search. */
436 	ntfs_attr_reinit_search_ctx(ctx);
437 
438 	/* Get number of hard links. */
439 	nr_links = le16_to_cpu(ctx->mrec->link_count);
440 
441 	/* Loop through all hard links. */
442 	while (!(err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0, NULL, 0,
443 			ctx))) {
444 		FILE_NAME_ATTR *file_name_attr;
445 		ATTR_RECORD *attr = ctx->attr;
446 		u8 *p, *p2;
447 
448 		nr_links--;
449 		/*
450 		 * Maximum sanity checking as we are called on an inode that
451 		 * we suspect might be corrupt.
452 		 */
453 		p = (u8*)attr + le32_to_cpu(attr->length);
454 		if (p < (u8*)ctx->mrec || (u8*)p > (u8*)ctx->mrec +
455 				le32_to_cpu(ctx->mrec->bytes_in_use)) {
456 err_corrupt_attr:
457 			ntfs_error(ctx->ntfs_ino->vol->sb, "Corrupt file name "
458 					"attribute. You should run chkdsk.");
459 			return -EIO;
460 		}
461 		if (attr->non_resident) {
462 			ntfs_error(ctx->ntfs_ino->vol->sb, "Non-resident file "
463 					"name. You should run chkdsk.");
464 			return -EIO;
465 		}
466 		if (attr->flags) {
467 			ntfs_error(ctx->ntfs_ino->vol->sb, "File name with "
468 					"invalid flags. You should run "
469 					"chkdsk.");
470 			return -EIO;
471 		}
472 		if (!(attr->data.resident.flags & RESIDENT_ATTR_IS_INDEXED)) {
473 			ntfs_error(ctx->ntfs_ino->vol->sb, "Unindexed file "
474 					"name. You should run chkdsk.");
475 			return -EIO;
476 		}
477 		file_name_attr = (FILE_NAME_ATTR*)((u8*)attr +
478 				le16_to_cpu(attr->data.resident.value_offset));
479 		p2 = (u8*)attr + le32_to_cpu(attr->data.resident.value_length);
480 		if (p2 < (u8*)attr || p2 > p)
481 			goto err_corrupt_attr;
482 		/* This attribute is ok, but is it in the $Extend directory? */
483 		if (MREF_LE(file_name_attr->parent_directory) == FILE_Extend)
484 			return 1;	/* YES, it's an extended system file. */
485 	}
486 	if (unlikely(err != -ENOENT))
487 		return err;
488 	if (unlikely(nr_links)) {
489 		ntfs_error(ctx->ntfs_ino->vol->sb, "Inode hard link count "
490 				"doesn't match number of name attributes. You "
491 				"should run chkdsk.");
492 		return -EIO;
493 	}
494 	return 0;	/* NO, it is not an extended system file. */
495 }
496 
497 /**
498  * ntfs_read_locked_inode - read an inode from its device
499  * @vi:		inode to read
500  *
501  * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode
502  * described by @vi into memory from the device.
503  *
504  * The only fields in @vi that we need to/can look at when the function is
505  * called are i_sb, pointing to the mounted device's super block, and i_ino,
506  * the number of the inode to load.
507  *
508  * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino
509  * for reading and sets up the necessary @vi fields as well as initializing
510  * the ntfs inode.
511  *
512  * Q: What locks are held when the function is called?
513  * A: i_state has I_LOCK set, hence the inode is locked, also
514  *    i_count is set to 1, so it is not going to go away
515  *    i_flags is set to 0 and we have no business touching it.  Only an ioctl()
516  *    is allowed to write to them. We should of course be honouring them but
517  *    we need to do that using the IS_* macros defined in include/linux/fs.h.
518  *    In any case ntfs_read_locked_inode() has nothing to do with i_flags.
519  *
520  * Return 0 on success and -errno on error.  In the error case, the inode will
521  * have had make_bad_inode() executed on it.
522  */
523 static int ntfs_read_locked_inode(struct inode *vi)
524 {
525 	ntfs_volume *vol = NTFS_SB(vi->i_sb);
526 	ntfs_inode *ni;
527 	MFT_RECORD *m;
528 	ATTR_RECORD *a;
529 	STANDARD_INFORMATION *si;
530 	ntfs_attr_search_ctx *ctx;
531 	int err = 0;
532 
533 	ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
534 
535 	/* Setup the generic vfs inode parts now. */
536 
537 	/* This is the optimal IO size (for stat), not the fs block size. */
538 	vi->i_blksize = PAGE_CACHE_SIZE;
539 	/*
540 	 * This is for checking whether an inode has changed w.r.t. a file so
541 	 * that the file can be updated if necessary (compare with f_version).
542 	 */
543 	vi->i_version = 1;
544 
545 	vi->i_uid = vol->uid;
546 	vi->i_gid = vol->gid;
547 	vi->i_mode = 0;
548 
549 	/*
550 	 * Initialize the ntfs specific part of @vi special casing
551 	 * FILE_MFT which we need to do at mount time.
552 	 */
553 	if (vi->i_ino != FILE_MFT)
554 		ntfs_init_big_inode(vi);
555 	ni = NTFS_I(vi);
556 
557 	m = map_mft_record(ni);
558 	if (IS_ERR(m)) {
559 		err = PTR_ERR(m);
560 		goto err_out;
561 	}
562 	ctx = ntfs_attr_get_search_ctx(ni, m);
563 	if (!ctx) {
564 		err = -ENOMEM;
565 		goto unm_err_out;
566 	}
567 
568 	if (!(m->flags & MFT_RECORD_IN_USE)) {
569 		ntfs_error(vi->i_sb, "Inode is not in use!");
570 		goto unm_err_out;
571 	}
572 	if (m->base_mft_record) {
573 		ntfs_error(vi->i_sb, "Inode is an extent inode!");
574 		goto unm_err_out;
575 	}
576 
577 	/* Transfer information from mft record into vfs and ntfs inodes. */
578 	vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
579 
580 	/*
581 	 * FIXME: Keep in mind that link_count is two for files which have both
582 	 * a long file name and a short file name as separate entries, so if
583 	 * we are hiding short file names this will be too high. Either we need
584 	 * to account for the short file names by subtracting them or we need
585 	 * to make sure we delete files even though i_nlink is not zero which
586 	 * might be tricky due to vfs interactions. Need to think about this
587 	 * some more when implementing the unlink command.
588 	 */
589 	vi->i_nlink = le16_to_cpu(m->link_count);
590 	/*
591 	 * FIXME: Reparse points can have the directory bit set even though
592 	 * they would be S_IFLNK. Need to deal with this further below when we
593 	 * implement reparse points / symbolic links but it will do for now.
594 	 * Also if not a directory, it could be something else, rather than
595 	 * a regular file. But again, will do for now.
596 	 */
597 	/* Everyone gets all permissions. */
598 	vi->i_mode |= S_IRWXUGO;
599 	/* If read-only, noone gets write permissions. */
600 	if (IS_RDONLY(vi))
601 		vi->i_mode &= ~S_IWUGO;
602 	if (m->flags & MFT_RECORD_IS_DIRECTORY) {
603 		vi->i_mode |= S_IFDIR;
604 		/*
605 		 * Apply the directory permissions mask set in the mount
606 		 * options.
607 		 */
608 		vi->i_mode &= ~vol->dmask;
609 		/* Things break without this kludge! */
610 		if (vi->i_nlink > 1)
611 			vi->i_nlink = 1;
612 	} else {
613 		vi->i_mode |= S_IFREG;
614 		/* Apply the file permissions mask set in the mount options. */
615 		vi->i_mode &= ~vol->fmask;
616 	}
617 	/*
618 	 * Find the standard information attribute in the mft record. At this
619 	 * stage we haven't setup the attribute list stuff yet, so this could
620 	 * in fact fail if the standard information is in an extent record, but
621 	 * I don't think this actually ever happens.
622 	 */
623 	err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0,
624 			ctx);
625 	if (unlikely(err)) {
626 		if (err == -ENOENT) {
627 			/*
628 			 * TODO: We should be performing a hot fix here (if the
629 			 * recover mount option is set) by creating a new
630 			 * attribute.
631 			 */
632 			ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute "
633 					"is missing.");
634 		}
635 		goto unm_err_out;
636 	}
637 	a = ctx->attr;
638 	/* Get the standard information attribute value. */
639 	si = (STANDARD_INFORMATION*)((u8*)a +
640 			le16_to_cpu(a->data.resident.value_offset));
641 
642 	/* Transfer information from the standard information into vi. */
643 	/*
644 	 * Note: The i_?times do not quite map perfectly onto the NTFS times,
645 	 * but they are close enough, and in the end it doesn't really matter
646 	 * that much...
647 	 */
648 	/*
649 	 * mtime is the last change of the data within the file. Not changed
650 	 * when only metadata is changed, e.g. a rename doesn't affect mtime.
651 	 */
652 	vi->i_mtime = ntfs2utc(si->last_data_change_time);
653 	/*
654 	 * ctime is the last change of the metadata of the file. This obviously
655 	 * always changes, when mtime is changed. ctime can be changed on its
656 	 * own, mtime is then not changed, e.g. when a file is renamed.
657 	 */
658 	vi->i_ctime = ntfs2utc(si->last_mft_change_time);
659 	/*
660 	 * Last access to the data within the file. Not changed during a rename
661 	 * for example but changed whenever the file is written to.
662 	 */
663 	vi->i_atime = ntfs2utc(si->last_access_time);
664 
665 	/* Find the attribute list attribute if present. */
666 	ntfs_attr_reinit_search_ctx(ctx);
667 	err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
668 	if (err) {
669 		if (unlikely(err != -ENOENT)) {
670 			ntfs_error(vi->i_sb, "Failed to lookup attribute list "
671 					"attribute.");
672 			goto unm_err_out;
673 		}
674 	} else /* if (!err) */ {
675 		if (vi->i_ino == FILE_MFT)
676 			goto skip_attr_list_load;
677 		ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino);
678 		NInoSetAttrList(ni);
679 		a = ctx->attr;
680 		if (a->flags & ATTR_IS_ENCRYPTED ||
681 				a->flags & ATTR_COMPRESSION_MASK ||
682 				a->flags & ATTR_IS_SPARSE) {
683 			ntfs_error(vi->i_sb, "Attribute list attribute is "
684 					"compressed/encrypted/sparse.");
685 			goto unm_err_out;
686 		}
687 		/* Now allocate memory for the attribute list. */
688 		ni->attr_list_size = (u32)ntfs_attr_size(a);
689 		ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
690 		if (!ni->attr_list) {
691 			ntfs_error(vi->i_sb, "Not enough memory to allocate "
692 					"buffer for attribute list.");
693 			err = -ENOMEM;
694 			goto unm_err_out;
695 		}
696 		if (a->non_resident) {
697 			NInoSetAttrListNonResident(ni);
698 			if (a->data.non_resident.lowest_vcn) {
699 				ntfs_error(vi->i_sb, "Attribute list has non "
700 						"zero lowest_vcn.");
701 				goto unm_err_out;
702 			}
703 			/*
704 			 * Setup the runlist. No need for locking as we have
705 			 * exclusive access to the inode at this time.
706 			 */
707 			ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
708 					a, NULL);
709 			if (IS_ERR(ni->attr_list_rl.rl)) {
710 				err = PTR_ERR(ni->attr_list_rl.rl);
711 				ni->attr_list_rl.rl = NULL;
712 				ntfs_error(vi->i_sb, "Mapping pairs "
713 						"decompression failed.");
714 				goto unm_err_out;
715 			}
716 			/* Now load the attribute list. */
717 			if ((err = load_attribute_list(vol, &ni->attr_list_rl,
718 					ni->attr_list, ni->attr_list_size,
719 					sle64_to_cpu(a->data.non_resident.
720 					initialized_size)))) {
721 				ntfs_error(vi->i_sb, "Failed to load "
722 						"attribute list attribute.");
723 				goto unm_err_out;
724 			}
725 		} else /* if (!a->non_resident) */ {
726 			if ((u8*)a + le16_to_cpu(a->data.resident.value_offset)
727 					+ le32_to_cpu(
728 					a->data.resident.value_length) >
729 					(u8*)ctx->mrec + vol->mft_record_size) {
730 				ntfs_error(vi->i_sb, "Corrupt attribute list "
731 						"in inode.");
732 				goto unm_err_out;
733 			}
734 			/* Now copy the attribute list. */
735 			memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
736 					a->data.resident.value_offset),
737 					le32_to_cpu(
738 					a->data.resident.value_length));
739 		}
740 	}
741 skip_attr_list_load:
742 	/*
743 	 * If an attribute list is present we now have the attribute list value
744 	 * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes.
745 	 */
746 	if (S_ISDIR(vi->i_mode)) {
747 		loff_t bvi_size;
748 		struct inode *bvi;
749 		ntfs_inode *bni;
750 		INDEX_ROOT *ir;
751 		u8 *ir_end, *index_end;
752 
753 		/* It is a directory, find index root attribute. */
754 		ntfs_attr_reinit_search_ctx(ctx);
755 		err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE,
756 				0, NULL, 0, ctx);
757 		if (unlikely(err)) {
758 			if (err == -ENOENT) {
759 				// FIXME: File is corrupt! Hot-fix with empty
760 				// index root attribute if recovery option is
761 				// set.
762 				ntfs_error(vi->i_sb, "$INDEX_ROOT attribute "
763 						"is missing.");
764 			}
765 			goto unm_err_out;
766 		}
767 		a = ctx->attr;
768 		/* Set up the state. */
769 		if (unlikely(a->non_resident)) {
770 			ntfs_error(vol->sb, "$INDEX_ROOT attribute is not "
771 					"resident.");
772 			goto unm_err_out;
773 		}
774 		/* Ensure the attribute name is placed before the value. */
775 		if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
776 				le16_to_cpu(a->data.resident.value_offset)))) {
777 			ntfs_error(vol->sb, "$INDEX_ROOT attribute name is "
778 					"placed after the attribute value.");
779 			goto unm_err_out;
780 		}
781 		/*
782 		 * Compressed/encrypted index root just means that the newly
783 		 * created files in that directory should be created compressed/
784 		 * encrypted. However index root cannot be both compressed and
785 		 * encrypted.
786 		 */
787 		if (a->flags & ATTR_COMPRESSION_MASK)
788 			NInoSetCompressed(ni);
789 		if (a->flags & ATTR_IS_ENCRYPTED) {
790 			if (a->flags & ATTR_COMPRESSION_MASK) {
791 				ntfs_error(vi->i_sb, "Found encrypted and "
792 						"compressed attribute.");
793 				goto unm_err_out;
794 			}
795 			NInoSetEncrypted(ni);
796 		}
797 		if (a->flags & ATTR_IS_SPARSE)
798 			NInoSetSparse(ni);
799 		ir = (INDEX_ROOT*)((u8*)a +
800 				le16_to_cpu(a->data.resident.value_offset));
801 		ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
802 		if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
803 			ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
804 					"corrupt.");
805 			goto unm_err_out;
806 		}
807 		index_end = (u8*)&ir->index +
808 				le32_to_cpu(ir->index.index_length);
809 		if (index_end > ir_end) {
810 			ntfs_error(vi->i_sb, "Directory index is corrupt.");
811 			goto unm_err_out;
812 		}
813 		if (ir->type != AT_FILE_NAME) {
814 			ntfs_error(vi->i_sb, "Indexed attribute is not "
815 					"$FILE_NAME.");
816 			goto unm_err_out;
817 		}
818 		if (ir->collation_rule != COLLATION_FILE_NAME) {
819 			ntfs_error(vi->i_sb, "Index collation rule is not "
820 					"COLLATION_FILE_NAME.");
821 			goto unm_err_out;
822 		}
823 		ni->itype.index.collation_rule = ir->collation_rule;
824 		ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
825 		if (ni->itype.index.block_size &
826 				(ni->itype.index.block_size - 1)) {
827 			ntfs_error(vi->i_sb, "Index block size (%u) is not a "
828 					"power of two.",
829 					ni->itype.index.block_size);
830 			goto unm_err_out;
831 		}
832 		if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
833 			ntfs_error(vi->i_sb, "Index block size (%u) > "
834 					"PAGE_CACHE_SIZE (%ld) is not "
835 					"supported.  Sorry.",
836 					ni->itype.index.block_size,
837 					PAGE_CACHE_SIZE);
838 			err = -EOPNOTSUPP;
839 			goto unm_err_out;
840 		}
841 		if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
842 			ntfs_error(vi->i_sb, "Index block size (%u) < "
843 					"NTFS_BLOCK_SIZE (%i) is not "
844 					"supported.  Sorry.",
845 					ni->itype.index.block_size,
846 					NTFS_BLOCK_SIZE);
847 			err = -EOPNOTSUPP;
848 			goto unm_err_out;
849 		}
850 		ni->itype.index.block_size_bits =
851 				ffs(ni->itype.index.block_size) - 1;
852 		/* Determine the size of a vcn in the directory index. */
853 		if (vol->cluster_size <= ni->itype.index.block_size) {
854 			ni->itype.index.vcn_size = vol->cluster_size;
855 			ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
856 		} else {
857 			ni->itype.index.vcn_size = vol->sector_size;
858 			ni->itype.index.vcn_size_bits = vol->sector_size_bits;
859 		}
860 
861 		/* Setup the index allocation attribute, even if not present. */
862 		NInoSetMstProtected(ni);
863 		ni->type = AT_INDEX_ALLOCATION;
864 		ni->name = I30;
865 		ni->name_len = 4;
866 
867 		if (!(ir->index.flags & LARGE_INDEX)) {
868 			/* No index allocation. */
869 			vi->i_size = ni->initialized_size =
870 					ni->allocated_size = 0;
871 			/* We are done with the mft record, so we release it. */
872 			ntfs_attr_put_search_ctx(ctx);
873 			unmap_mft_record(ni);
874 			m = NULL;
875 			ctx = NULL;
876 			goto skip_large_dir_stuff;
877 		} /* LARGE_INDEX: Index allocation present. Setup state. */
878 		NInoSetIndexAllocPresent(ni);
879 		/* Find index allocation attribute. */
880 		ntfs_attr_reinit_search_ctx(ctx);
881 		err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4,
882 				CASE_SENSITIVE, 0, NULL, 0, ctx);
883 		if (unlikely(err)) {
884 			if (err == -ENOENT)
885 				ntfs_error(vi->i_sb, "$INDEX_ALLOCATION "
886 						"attribute is not present but "
887 						"$INDEX_ROOT indicated it is.");
888 			else
889 				ntfs_error(vi->i_sb, "Failed to lookup "
890 						"$INDEX_ALLOCATION "
891 						"attribute.");
892 			goto unm_err_out;
893 		}
894 		a = ctx->attr;
895 		if (!a->non_resident) {
896 			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
897 					"is resident.");
898 			goto unm_err_out;
899 		}
900 		/*
901 		 * Ensure the attribute name is placed before the mapping pairs
902 		 * array.
903 		 */
904 		if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
905 				le16_to_cpu(
906 				a->data.non_resident.mapping_pairs_offset)))) {
907 			ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name "
908 					"is placed after the mapping pairs "
909 					"array.");
910 			goto unm_err_out;
911 		}
912 		if (a->flags & ATTR_IS_ENCRYPTED) {
913 			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
914 					"is encrypted.");
915 			goto unm_err_out;
916 		}
917 		if (a->flags & ATTR_IS_SPARSE) {
918 			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
919 					"is sparse.");
920 			goto unm_err_out;
921 		}
922 		if (a->flags & ATTR_COMPRESSION_MASK) {
923 			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
924 					"is compressed.");
925 			goto unm_err_out;
926 		}
927 		if (a->data.non_resident.lowest_vcn) {
928 			ntfs_error(vi->i_sb, "First extent of "
929 					"$INDEX_ALLOCATION attribute has non "
930 					"zero lowest_vcn.");
931 			goto unm_err_out;
932 		}
933 		vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
934 		ni->initialized_size = sle64_to_cpu(
935 				a->data.non_resident.initialized_size);
936 		ni->allocated_size = sle64_to_cpu(
937 				a->data.non_resident.allocated_size);
938 		/*
939 		 * We are done with the mft record, so we release it. Otherwise
940 		 * we would deadlock in ntfs_attr_iget().
941 		 */
942 		ntfs_attr_put_search_ctx(ctx);
943 		unmap_mft_record(ni);
944 		m = NULL;
945 		ctx = NULL;
946 		/* Get the index bitmap attribute inode. */
947 		bvi = ntfs_attr_iget(vi, AT_BITMAP, I30, 4);
948 		if (IS_ERR(bvi)) {
949 			ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
950 			err = PTR_ERR(bvi);
951 			goto unm_err_out;
952 		}
953 		ni->itype.index.bmp_ino = bvi;
954 		bni = NTFS_I(bvi);
955 		if (NInoCompressed(bni) || NInoEncrypted(bni) ||
956 				NInoSparse(bni)) {
957 			ntfs_error(vi->i_sb, "$BITMAP attribute is compressed "
958 					"and/or encrypted and/or sparse.");
959 			goto unm_err_out;
960 		}
961 		/* Consistency check bitmap size vs. index allocation size. */
962 		bvi_size = i_size_read(bvi);
963 		if ((bvi_size << 3) < (vi->i_size >>
964 				ni->itype.index.block_size_bits)) {
965 			ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) "
966 					"for index allocation (0x%llx).",
967 					bvi_size << 3, vi->i_size);
968 			goto unm_err_out;
969 		}
970 skip_large_dir_stuff:
971 		/* Setup the operations for this inode. */
972 		vi->i_op = &ntfs_dir_inode_ops;
973 		vi->i_fop = &ntfs_dir_ops;
974 	} else {
975 		/* It is a file. */
976 		ntfs_attr_reinit_search_ctx(ctx);
977 
978 		/* Setup the data attribute, even if not present. */
979 		ni->type = AT_DATA;
980 		ni->name = NULL;
981 		ni->name_len = 0;
982 
983 		/* Find first extent of the unnamed data attribute. */
984 		err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx);
985 		if (unlikely(err)) {
986 			vi->i_size = ni->initialized_size =
987 					ni->allocated_size = 0;
988 			if (err != -ENOENT) {
989 				ntfs_error(vi->i_sb, "Failed to lookup $DATA "
990 						"attribute.");
991 				goto unm_err_out;
992 			}
993 			/*
994 			 * FILE_Secure does not have an unnamed $DATA
995 			 * attribute, so we special case it here.
996 			 */
997 			if (vi->i_ino == FILE_Secure)
998 				goto no_data_attr_special_case;
999 			/*
1000 			 * Most if not all the system files in the $Extend
1001 			 * system directory do not have unnamed data
1002 			 * attributes so we need to check if the parent
1003 			 * directory of the file is FILE_Extend and if it is
1004 			 * ignore this error. To do this we need to get the
1005 			 * name of this inode from the mft record as the name
1006 			 * contains the back reference to the parent directory.
1007 			 */
1008 			if (ntfs_is_extended_system_file(ctx) > 0)
1009 				goto no_data_attr_special_case;
1010 			// FIXME: File is corrupt! Hot-fix with empty data
1011 			// attribute if recovery option is set.
1012 			ntfs_error(vi->i_sb, "$DATA attribute is missing.");
1013 			goto unm_err_out;
1014 		}
1015 		a = ctx->attr;
1016 		/* Setup the state. */
1017 		if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1018 			if (a->flags & ATTR_COMPRESSION_MASK) {
1019 				NInoSetCompressed(ni);
1020 				if (vol->cluster_size > 4096) {
1021 					ntfs_error(vi->i_sb, "Found "
1022 							"compressed data but "
1023 							"compression is "
1024 							"disabled due to "
1025 							"cluster size (%i) > "
1026 							"4kiB.",
1027 							vol->cluster_size);
1028 					goto unm_err_out;
1029 				}
1030 				if ((a->flags & ATTR_COMPRESSION_MASK)
1031 						!= ATTR_IS_COMPRESSED) {
1032 					ntfs_error(vi->i_sb, "Found unknown "
1033 							"compression method "
1034 							"or corrupt file.");
1035 					goto unm_err_out;
1036 				}
1037 			}
1038 			if (a->flags & ATTR_IS_SPARSE)
1039 				NInoSetSparse(ni);
1040 		}
1041 		if (a->flags & ATTR_IS_ENCRYPTED) {
1042 			if (NInoCompressed(ni)) {
1043 				ntfs_error(vi->i_sb, "Found encrypted and "
1044 						"compressed data.");
1045 				goto unm_err_out;
1046 			}
1047 			NInoSetEncrypted(ni);
1048 		}
1049 		if (a->non_resident) {
1050 			NInoSetNonResident(ni);
1051 			if (NInoCompressed(ni) || NInoSparse(ni)) {
1052 				if (a->data.non_resident.compression_unit !=
1053 						4) {
1054 					ntfs_error(vi->i_sb, "Found "
1055 							"nonstandard "
1056 							"compression unit (%u "
1057 							"instead of 4).  "
1058 							"Cannot handle this.",
1059 							a->data.non_resident.
1060 							compression_unit);
1061 					err = -EOPNOTSUPP;
1062 					goto unm_err_out;
1063 				}
1064 				ni->itype.compressed.block_clusters = 1U <<
1065 						a->data.non_resident.
1066 						compression_unit;
1067 				ni->itype.compressed.block_size = 1U << (
1068 						a->data.non_resident.
1069 						compression_unit +
1070 						vol->cluster_size_bits);
1071 				ni->itype.compressed.block_size_bits = ffs(
1072 						ni->itype.compressed.
1073 						block_size) - 1;
1074 				ni->itype.compressed.size = sle64_to_cpu(
1075 						a->data.non_resident.
1076 						compressed_size);
1077 			}
1078 			if (a->data.non_resident.lowest_vcn) {
1079 				ntfs_error(vi->i_sb, "First extent of $DATA "
1080 						"attribute has non zero "
1081 						"lowest_vcn.");
1082 				goto unm_err_out;
1083 			}
1084 			vi->i_size = sle64_to_cpu(
1085 					a->data.non_resident.data_size);
1086 			ni->initialized_size = sle64_to_cpu(
1087 					a->data.non_resident.initialized_size);
1088 			ni->allocated_size = sle64_to_cpu(
1089 					a->data.non_resident.allocated_size);
1090 		} else { /* Resident attribute. */
1091 			vi->i_size = ni->initialized_size = le32_to_cpu(
1092 					a->data.resident.value_length);
1093 			ni->allocated_size = le32_to_cpu(a->length) -
1094 					le16_to_cpu(
1095 					a->data.resident.value_offset);
1096 			if (vi->i_size > ni->allocated_size) {
1097 				ntfs_error(vi->i_sb, "Resident data attribute "
1098 						"is corrupt (size exceeds "
1099 						"allocation).");
1100 				goto unm_err_out;
1101 			}
1102 		}
1103 no_data_attr_special_case:
1104 		/* We are done with the mft record, so we release it. */
1105 		ntfs_attr_put_search_ctx(ctx);
1106 		unmap_mft_record(ni);
1107 		m = NULL;
1108 		ctx = NULL;
1109 		/* Setup the operations for this inode. */
1110 		vi->i_op = &ntfs_file_inode_ops;
1111 		vi->i_fop = &ntfs_file_ops;
1112 	}
1113 	if (NInoMstProtected(ni))
1114 		vi->i_mapping->a_ops = &ntfs_mst_aops;
1115 	else
1116 		vi->i_mapping->a_ops = &ntfs_aops;
1117 	/*
1118 	 * The number of 512-byte blocks used on disk (for stat). This is in so
1119 	 * far inaccurate as it doesn't account for any named streams or other
1120 	 * special non-resident attributes, but that is how Windows works, too,
1121 	 * so we are at least consistent with Windows, if not entirely
1122 	 * consistent with the Linux Way. Doing it the Linux Way would cause a
1123 	 * significant slowdown as it would involve iterating over all
1124 	 * attributes in the mft record and adding the allocated/compressed
1125 	 * sizes of all non-resident attributes present to give us the Linux
1126 	 * correct size that should go into i_blocks (after division by 512).
1127 	 */
1128 	if (S_ISREG(vi->i_mode) && (NInoCompressed(ni) || NInoSparse(ni)))
1129 		vi->i_blocks = ni->itype.compressed.size >> 9;
1130 	else
1131 		vi->i_blocks = ni->allocated_size >> 9;
1132 	ntfs_debug("Done.");
1133 	return 0;
1134 
1135 unm_err_out:
1136 	if (!err)
1137 		err = -EIO;
1138 	if (ctx)
1139 		ntfs_attr_put_search_ctx(ctx);
1140 	if (m)
1141 		unmap_mft_record(ni);
1142 err_out:
1143 	ntfs_error(vol->sb, "Failed with error code %i.  Marking corrupt "
1144 			"inode 0x%lx as bad.  Run chkdsk.", err, vi->i_ino);
1145 	make_bad_inode(vi);
1146 	if (err != -EOPNOTSUPP && err != -ENOMEM)
1147 		NVolSetErrors(vol);
1148 	return err;
1149 }
1150 
1151 /**
1152  * ntfs_read_locked_attr_inode - read an attribute inode from its base inode
1153  * @base_vi:	base inode
1154  * @vi:		attribute inode to read
1155  *
1156  * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the
1157  * attribute inode described by @vi into memory from the base mft record
1158  * described by @base_ni.
1159  *
1160  * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for
1161  * reading and looks up the attribute described by @vi before setting up the
1162  * necessary fields in @vi as well as initializing the ntfs inode.
1163  *
1164  * Q: What locks are held when the function is called?
1165  * A: i_state has I_LOCK set, hence the inode is locked, also
1166  *    i_count is set to 1, so it is not going to go away
1167  *
1168  * Return 0 on success and -errno on error.  In the error case, the inode will
1169  * have had make_bad_inode() executed on it.
1170  *
1171  * Note this cannot be called for AT_INDEX_ALLOCATION.
1172  */
1173 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi)
1174 {
1175 	ntfs_volume *vol = NTFS_SB(vi->i_sb);
1176 	ntfs_inode *ni, *base_ni;
1177 	MFT_RECORD *m;
1178 	ATTR_RECORD *a;
1179 	ntfs_attr_search_ctx *ctx;
1180 	int err = 0;
1181 
1182 	ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1183 
1184 	ntfs_init_big_inode(vi);
1185 
1186 	ni	= NTFS_I(vi);
1187 	base_ni = NTFS_I(base_vi);
1188 
1189 	/* Just mirror the values from the base inode. */
1190 	vi->i_blksize	= base_vi->i_blksize;
1191 	vi->i_version	= base_vi->i_version;
1192 	vi->i_uid	= base_vi->i_uid;
1193 	vi->i_gid	= base_vi->i_gid;
1194 	vi->i_nlink	= base_vi->i_nlink;
1195 	vi->i_mtime	= base_vi->i_mtime;
1196 	vi->i_ctime	= base_vi->i_ctime;
1197 	vi->i_atime	= base_vi->i_atime;
1198 	vi->i_generation = ni->seq_no = base_ni->seq_no;
1199 
1200 	/* Set inode type to zero but preserve permissions. */
1201 	vi->i_mode	= base_vi->i_mode & ~S_IFMT;
1202 
1203 	m = map_mft_record(base_ni);
1204 	if (IS_ERR(m)) {
1205 		err = PTR_ERR(m);
1206 		goto err_out;
1207 	}
1208 	ctx = ntfs_attr_get_search_ctx(base_ni, m);
1209 	if (!ctx) {
1210 		err = -ENOMEM;
1211 		goto unm_err_out;
1212 	}
1213 	/* Find the attribute. */
1214 	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
1215 			CASE_SENSITIVE, 0, NULL, 0, ctx);
1216 	if (unlikely(err))
1217 		goto unm_err_out;
1218 	a = ctx->attr;
1219 	if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1220 		if (a->flags & ATTR_COMPRESSION_MASK) {
1221 			NInoSetCompressed(ni);
1222 			if ((ni->type != AT_DATA) || (ni->type == AT_DATA &&
1223 					ni->name_len)) {
1224 				ntfs_error(vi->i_sb, "Found compressed "
1225 						"non-data or named data "
1226 						"attribute.  Please report "
1227 						"you saw this message to "
1228 						"linux-ntfs-dev@lists."
1229 						"sourceforge.net");
1230 				goto unm_err_out;
1231 			}
1232 			if (vol->cluster_size > 4096) {
1233 				ntfs_error(vi->i_sb, "Found compressed "
1234 						"attribute but compression is "
1235 						"disabled due to cluster size "
1236 						"(%i) > 4kiB.",
1237 						vol->cluster_size);
1238 				goto unm_err_out;
1239 			}
1240 			if ((a->flags & ATTR_COMPRESSION_MASK) !=
1241 					ATTR_IS_COMPRESSED) {
1242 				ntfs_error(vi->i_sb, "Found unknown "
1243 						"compression method.");
1244 				goto unm_err_out;
1245 			}
1246 		}
1247 		/*
1248 		 * The compressed/sparse flag set in an index root just means
1249 		 * to compress all files.
1250 		 */
1251 		if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1252 			ntfs_error(vi->i_sb, "Found mst protected attribute "
1253 					"but the attribute is %s.  Please "
1254 					"report you saw this message to "
1255 					"linux-ntfs-dev@lists.sourceforge.net",
1256 					NInoCompressed(ni) ? "compressed" :
1257 					"sparse");
1258 			goto unm_err_out;
1259 		}
1260 		if (a->flags & ATTR_IS_SPARSE)
1261 			NInoSetSparse(ni);
1262 	}
1263 	if (a->flags & ATTR_IS_ENCRYPTED) {
1264 		if (NInoCompressed(ni)) {
1265 			ntfs_error(vi->i_sb, "Found encrypted and compressed "
1266 					"data.");
1267 			goto unm_err_out;
1268 		}
1269 		/*
1270 		 * The encryption flag set in an index root just means to
1271 		 * encrypt all files.
1272 		 */
1273 		if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1274 			ntfs_error(vi->i_sb, "Found mst protected attribute "
1275 					"but the attribute is encrypted.  "
1276 					"Please report you saw this message "
1277 					"to linux-ntfs-dev@lists.sourceforge."
1278 					"net");
1279 			goto unm_err_out;
1280 		}
1281 		if (ni->type != AT_DATA) {
1282 			ntfs_error(vi->i_sb, "Found encrypted non-data "
1283 					"attribute.");
1284 			goto unm_err_out;
1285 		}
1286 		NInoSetEncrypted(ni);
1287 	}
1288 	if (!a->non_resident) {
1289 		/* Ensure the attribute name is placed before the value. */
1290 		if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1291 				le16_to_cpu(a->data.resident.value_offset)))) {
1292 			ntfs_error(vol->sb, "Attribute name is placed after "
1293 					"the attribute value.");
1294 			goto unm_err_out;
1295 		}
1296 		if (NInoMstProtected(ni)) {
1297 			ntfs_error(vi->i_sb, "Found mst protected attribute "
1298 					"but the attribute is resident.  "
1299 					"Please report you saw this message to "
1300 					"linux-ntfs-dev@lists.sourceforge.net");
1301 			goto unm_err_out;
1302 		}
1303 		vi->i_size = ni->initialized_size = le32_to_cpu(
1304 				a->data.resident.value_length);
1305 		ni->allocated_size = le32_to_cpu(a->length) -
1306 				le16_to_cpu(a->data.resident.value_offset);
1307 		if (vi->i_size > ni->allocated_size) {
1308 			ntfs_error(vi->i_sb, "Resident attribute is corrupt "
1309 					"(size exceeds allocation).");
1310 			goto unm_err_out;
1311 		}
1312 	} else {
1313 		NInoSetNonResident(ni);
1314 		/*
1315 		 * Ensure the attribute name is placed before the mapping pairs
1316 		 * array.
1317 		 */
1318 		if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1319 				le16_to_cpu(
1320 				a->data.non_resident.mapping_pairs_offset)))) {
1321 			ntfs_error(vol->sb, "Attribute name is placed after "
1322 					"the mapping pairs array.");
1323 			goto unm_err_out;
1324 		}
1325 		if (NInoCompressed(ni) || NInoSparse(ni)) {
1326 			if (a->data.non_resident.compression_unit != 4) {
1327 				ntfs_error(vi->i_sb, "Found nonstandard "
1328 						"compression unit (%u instead "
1329 						"of 4).  Cannot handle this.",
1330 						a->data.non_resident.
1331 						compression_unit);
1332 				err = -EOPNOTSUPP;
1333 				goto unm_err_out;
1334 			}
1335 			ni->itype.compressed.block_clusters = 1U <<
1336 					a->data.non_resident.compression_unit;
1337 			ni->itype.compressed.block_size = 1U << (
1338 					a->data.non_resident.compression_unit +
1339 					vol->cluster_size_bits);
1340 			ni->itype.compressed.block_size_bits = ffs(
1341 					ni->itype.compressed.block_size) - 1;
1342 			ni->itype.compressed.size = sle64_to_cpu(
1343 					a->data.non_resident.compressed_size);
1344 		}
1345 		if (a->data.non_resident.lowest_vcn) {
1346 			ntfs_error(vi->i_sb, "First extent of attribute has "
1347 					"non-zero lowest_vcn.");
1348 			goto unm_err_out;
1349 		}
1350 		vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1351 		ni->initialized_size = sle64_to_cpu(
1352 				a->data.non_resident.initialized_size);
1353 		ni->allocated_size = sle64_to_cpu(
1354 				a->data.non_resident.allocated_size);
1355 	}
1356 	/* Setup the operations for this attribute inode. */
1357 	vi->i_op = NULL;
1358 	vi->i_fop = NULL;
1359 	if (NInoMstProtected(ni))
1360 		vi->i_mapping->a_ops = &ntfs_mst_aops;
1361 	else
1362 		vi->i_mapping->a_ops = &ntfs_aops;
1363 	if ((NInoCompressed(ni) || NInoSparse(ni)) && ni->type != AT_INDEX_ROOT)
1364 		vi->i_blocks = ni->itype.compressed.size >> 9;
1365 	else
1366 		vi->i_blocks = ni->allocated_size >> 9;
1367 	/*
1368 	 * Make sure the base inode does not go away and attach it to the
1369 	 * attribute inode.
1370 	 */
1371 	igrab(base_vi);
1372 	ni->ext.base_ntfs_ino = base_ni;
1373 	ni->nr_extents = -1;
1374 
1375 	ntfs_attr_put_search_ctx(ctx);
1376 	unmap_mft_record(base_ni);
1377 
1378 	ntfs_debug("Done.");
1379 	return 0;
1380 
1381 unm_err_out:
1382 	if (!err)
1383 		err = -EIO;
1384 	if (ctx)
1385 		ntfs_attr_put_search_ctx(ctx);
1386 	unmap_mft_record(base_ni);
1387 err_out:
1388 	ntfs_error(vol->sb, "Failed with error code %i while reading attribute "
1389 			"inode (mft_no 0x%lx, type 0x%x, name_len %i).  "
1390 			"Marking corrupt inode and base inode 0x%lx as bad.  "
1391 			"Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len,
1392 			base_vi->i_ino);
1393 	make_bad_inode(vi);
1394 	make_bad_inode(base_vi);
1395 	if (err != -ENOMEM)
1396 		NVolSetErrors(vol);
1397 	return err;
1398 }
1399 
1400 /**
1401  * ntfs_read_locked_index_inode - read an index inode from its base inode
1402  * @base_vi:	base inode
1403  * @vi:		index inode to read
1404  *
1405  * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the
1406  * index inode described by @vi into memory from the base mft record described
1407  * by @base_ni.
1408  *
1409  * ntfs_read_locked_index_inode() maps, pins and locks the base inode for
1410  * reading and looks up the attributes relating to the index described by @vi
1411  * before setting up the necessary fields in @vi as well as initializing the
1412  * ntfs inode.
1413  *
1414  * Note, index inodes are essentially attribute inodes (NInoAttr() is true)
1415  * with the attribute type set to AT_INDEX_ALLOCATION.  Apart from that, they
1416  * are setup like directory inodes since directories are a special case of
1417  * indices ao they need to be treated in much the same way.  Most importantly,
1418  * for small indices the index allocation attribute might not actually exist.
1419  * However, the index root attribute always exists but this does not need to
1420  * have an inode associated with it and this is why we define a new inode type
1421  * index.  Also, like for directories, we need to have an attribute inode for
1422  * the bitmap attribute corresponding to the index allocation attribute and we
1423  * can store this in the appropriate field of the inode, just like we do for
1424  * normal directory inodes.
1425  *
1426  * Q: What locks are held when the function is called?
1427  * A: i_state has I_LOCK set, hence the inode is locked, also
1428  *    i_count is set to 1, so it is not going to go away
1429  *
1430  * Return 0 on success and -errno on error.  In the error case, the inode will
1431  * have had make_bad_inode() executed on it.
1432  */
1433 static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi)
1434 {
1435 	loff_t bvi_size;
1436 	ntfs_volume *vol = NTFS_SB(vi->i_sb);
1437 	ntfs_inode *ni, *base_ni, *bni;
1438 	struct inode *bvi;
1439 	MFT_RECORD *m;
1440 	ATTR_RECORD *a;
1441 	ntfs_attr_search_ctx *ctx;
1442 	INDEX_ROOT *ir;
1443 	u8 *ir_end, *index_end;
1444 	int err = 0;
1445 
1446 	ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1447 	ntfs_init_big_inode(vi);
1448 	ni	= NTFS_I(vi);
1449 	base_ni = NTFS_I(base_vi);
1450 	/* Just mirror the values from the base inode. */
1451 	vi->i_blksize	= base_vi->i_blksize;
1452 	vi->i_version	= base_vi->i_version;
1453 	vi->i_uid	= base_vi->i_uid;
1454 	vi->i_gid	= base_vi->i_gid;
1455 	vi->i_nlink	= base_vi->i_nlink;
1456 	vi->i_mtime	= base_vi->i_mtime;
1457 	vi->i_ctime	= base_vi->i_ctime;
1458 	vi->i_atime	= base_vi->i_atime;
1459 	vi->i_generation = ni->seq_no = base_ni->seq_no;
1460 	/* Set inode type to zero but preserve permissions. */
1461 	vi->i_mode	= base_vi->i_mode & ~S_IFMT;
1462 	/* Map the mft record for the base inode. */
1463 	m = map_mft_record(base_ni);
1464 	if (IS_ERR(m)) {
1465 		err = PTR_ERR(m);
1466 		goto err_out;
1467 	}
1468 	ctx = ntfs_attr_get_search_ctx(base_ni, m);
1469 	if (!ctx) {
1470 		err = -ENOMEM;
1471 		goto unm_err_out;
1472 	}
1473 	/* Find the index root attribute. */
1474 	err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len,
1475 			CASE_SENSITIVE, 0, NULL, 0, ctx);
1476 	if (unlikely(err)) {
1477 		if (err == -ENOENT)
1478 			ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
1479 					"missing.");
1480 		goto unm_err_out;
1481 	}
1482 	a = ctx->attr;
1483 	/* Set up the state. */
1484 	if (unlikely(a->non_resident)) {
1485 		ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident.");
1486 		goto unm_err_out;
1487 	}
1488 	/* Ensure the attribute name is placed before the value. */
1489 	if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1490 			le16_to_cpu(a->data.resident.value_offset)))) {
1491 		ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed "
1492 				"after the attribute value.");
1493 		goto unm_err_out;
1494 	}
1495 	/*
1496 	 * Compressed/encrypted/sparse index root is not allowed, except for
1497 	 * directories of course but those are not dealt with here.
1498 	 */
1499 	if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED |
1500 			ATTR_IS_SPARSE)) {
1501 		ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index "
1502 				"root attribute.");
1503 		goto unm_err_out;
1504 	}
1505 	ir = (INDEX_ROOT*)((u8*)a + le16_to_cpu(a->data.resident.value_offset));
1506 	ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
1507 	if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
1508 		ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt.");
1509 		goto unm_err_out;
1510 	}
1511 	index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
1512 	if (index_end > ir_end) {
1513 		ntfs_error(vi->i_sb, "Index is corrupt.");
1514 		goto unm_err_out;
1515 	}
1516 	if (ir->type) {
1517 		ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).",
1518 				le32_to_cpu(ir->type));
1519 		goto unm_err_out;
1520 	}
1521 	ni->itype.index.collation_rule = ir->collation_rule;
1522 	ntfs_debug("Index collation rule is 0x%x.",
1523 			le32_to_cpu(ir->collation_rule));
1524 	ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
1525 	if (ni->itype.index.block_size & (ni->itype.index.block_size - 1)) {
1526 		ntfs_error(vi->i_sb, "Index block size (%u) is not a power of "
1527 				"two.", ni->itype.index.block_size);
1528 		goto unm_err_out;
1529 	}
1530 	if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
1531 		ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_CACHE_SIZE "
1532 				"(%ld) is not supported.  Sorry.",
1533 				ni->itype.index.block_size, PAGE_CACHE_SIZE);
1534 		err = -EOPNOTSUPP;
1535 		goto unm_err_out;
1536 	}
1537 	if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
1538 		ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE "
1539 				"(%i) is not supported.  Sorry.",
1540 				ni->itype.index.block_size, NTFS_BLOCK_SIZE);
1541 		err = -EOPNOTSUPP;
1542 		goto unm_err_out;
1543 	}
1544 	ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1;
1545 	/* Determine the size of a vcn in the index. */
1546 	if (vol->cluster_size <= ni->itype.index.block_size) {
1547 		ni->itype.index.vcn_size = vol->cluster_size;
1548 		ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
1549 	} else {
1550 		ni->itype.index.vcn_size = vol->sector_size;
1551 		ni->itype.index.vcn_size_bits = vol->sector_size_bits;
1552 	}
1553 	/* Check for presence of index allocation attribute. */
1554 	if (!(ir->index.flags & LARGE_INDEX)) {
1555 		/* No index allocation. */
1556 		vi->i_size = ni->initialized_size = ni->allocated_size = 0;
1557 		/* We are done with the mft record, so we release it. */
1558 		ntfs_attr_put_search_ctx(ctx);
1559 		unmap_mft_record(base_ni);
1560 		m = NULL;
1561 		ctx = NULL;
1562 		goto skip_large_index_stuff;
1563 	} /* LARGE_INDEX:  Index allocation present.  Setup state. */
1564 	NInoSetIndexAllocPresent(ni);
1565 	/* Find index allocation attribute. */
1566 	ntfs_attr_reinit_search_ctx(ctx);
1567 	err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len,
1568 			CASE_SENSITIVE, 0, NULL, 0, ctx);
1569 	if (unlikely(err)) {
1570 		if (err == -ENOENT)
1571 			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1572 					"not present but $INDEX_ROOT "
1573 					"indicated it is.");
1574 		else
1575 			ntfs_error(vi->i_sb, "Failed to lookup "
1576 					"$INDEX_ALLOCATION attribute.");
1577 		goto unm_err_out;
1578 	}
1579 	if (!a->non_resident) {
1580 		ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1581 				"resident.");
1582 		goto unm_err_out;
1583 	}
1584 	/*
1585 	 * Ensure the attribute name is placed before the mapping pairs array.
1586 	 */
1587 	if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1588 			le16_to_cpu(
1589 			a->data.non_resident.mapping_pairs_offset)))) {
1590 		ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is "
1591 				"placed after the mapping pairs array.");
1592 		goto unm_err_out;
1593 	}
1594 	if (a->flags & ATTR_IS_ENCRYPTED) {
1595 		ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1596 				"encrypted.");
1597 		goto unm_err_out;
1598 	}
1599 	if (a->flags & ATTR_IS_SPARSE) {
1600 		ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse.");
1601 		goto unm_err_out;
1602 	}
1603 	if (a->flags & ATTR_COMPRESSION_MASK) {
1604 		ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1605 				"compressed.");
1606 		goto unm_err_out;
1607 	}
1608 	if (a->data.non_resident.lowest_vcn) {
1609 		ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION "
1610 				"attribute has non zero lowest_vcn.");
1611 		goto unm_err_out;
1612 	}
1613 	vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1614 	ni->initialized_size = sle64_to_cpu(
1615 			a->data.non_resident.initialized_size);
1616 	ni->allocated_size = sle64_to_cpu(a->data.non_resident.allocated_size);
1617 	/*
1618 	 * We are done with the mft record, so we release it.  Otherwise
1619 	 * we would deadlock in ntfs_attr_iget().
1620 	 */
1621 	ntfs_attr_put_search_ctx(ctx);
1622 	unmap_mft_record(base_ni);
1623 	m = NULL;
1624 	ctx = NULL;
1625 	/* Get the index bitmap attribute inode. */
1626 	bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len);
1627 	if (IS_ERR(bvi)) {
1628 		ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
1629 		err = PTR_ERR(bvi);
1630 		goto unm_err_out;
1631 	}
1632 	bni = NTFS_I(bvi);
1633 	if (NInoCompressed(bni) || NInoEncrypted(bni) ||
1634 			NInoSparse(bni)) {
1635 		ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or "
1636 				"encrypted and/or sparse.");
1637 		goto iput_unm_err_out;
1638 	}
1639 	/* Consistency check bitmap size vs. index allocation size. */
1640 	bvi_size = i_size_read(bvi);
1641 	if ((bvi_size << 3) < (vi->i_size >> ni->itype.index.block_size_bits)) {
1642 		ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for "
1643 				"index allocation (0x%llx).", bvi_size << 3,
1644 				vi->i_size);
1645 		goto iput_unm_err_out;
1646 	}
1647 	ni->itype.index.bmp_ino = bvi;
1648 skip_large_index_stuff:
1649 	/* Setup the operations for this index inode. */
1650 	vi->i_op = NULL;
1651 	vi->i_fop = NULL;
1652 	vi->i_mapping->a_ops = &ntfs_mst_aops;
1653 	vi->i_blocks = ni->allocated_size >> 9;
1654 	/*
1655 	 * Make sure the base inode doesn't go away and attach it to the
1656 	 * index inode.
1657 	 */
1658 	igrab(base_vi);
1659 	ni->ext.base_ntfs_ino = base_ni;
1660 	ni->nr_extents = -1;
1661 
1662 	ntfs_debug("Done.");
1663 	return 0;
1664 
1665 iput_unm_err_out:
1666 	iput(bvi);
1667 unm_err_out:
1668 	if (!err)
1669 		err = -EIO;
1670 	if (ctx)
1671 		ntfs_attr_put_search_ctx(ctx);
1672 	if (m)
1673 		unmap_mft_record(base_ni);
1674 err_out:
1675 	ntfs_error(vi->i_sb, "Failed with error code %i while reading index "
1676 			"inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino,
1677 			ni->name_len);
1678 	make_bad_inode(vi);
1679 	if (err != -EOPNOTSUPP && err != -ENOMEM)
1680 		NVolSetErrors(vol);
1681 	return err;
1682 }
1683 
1684 /**
1685  * ntfs_read_inode_mount - special read_inode for mount time use only
1686  * @vi:		inode to read
1687  *
1688  * Read inode FILE_MFT at mount time, only called with super_block lock
1689  * held from within the read_super() code path.
1690  *
1691  * This function exists because when it is called the page cache for $MFT/$DATA
1692  * is not initialized and hence we cannot get at the contents of mft records
1693  * by calling map_mft_record*().
1694  *
1695  * Further it needs to cope with the circular references problem, i.e. cannot
1696  * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because
1697  * we do not know where the other extent mft records are yet and again, because
1698  * we cannot call map_mft_record*() yet.  Obviously this applies only when an
1699  * attribute list is actually present in $MFT inode.
1700  *
1701  * We solve these problems by starting with the $DATA attribute before anything
1702  * else and iterating using ntfs_attr_lookup($DATA) over all extents.  As each
1703  * extent is found, we ntfs_mapping_pairs_decompress() including the implied
1704  * ntfs_runlists_merge().  Each step of the iteration necessarily provides
1705  * sufficient information for the next step to complete.
1706  *
1707  * This should work but there are two possible pit falls (see inline comments
1708  * below), but only time will tell if they are real pits or just smoke...
1709  */
1710 int ntfs_read_inode_mount(struct inode *vi)
1711 {
1712 	VCN next_vcn, last_vcn, highest_vcn;
1713 	s64 block;
1714 	struct super_block *sb = vi->i_sb;
1715 	ntfs_volume *vol = NTFS_SB(sb);
1716 	struct buffer_head *bh;
1717 	ntfs_inode *ni;
1718 	MFT_RECORD *m = NULL;
1719 	ATTR_RECORD *a;
1720 	ntfs_attr_search_ctx *ctx;
1721 	unsigned int i, nr_blocks;
1722 	int err;
1723 
1724 	ntfs_debug("Entering.");
1725 
1726 	/* Initialize the ntfs specific part of @vi. */
1727 	ntfs_init_big_inode(vi);
1728 
1729 	ni = NTFS_I(vi);
1730 
1731 	/* Setup the data attribute. It is special as it is mst protected. */
1732 	NInoSetNonResident(ni);
1733 	NInoSetMstProtected(ni);
1734 	NInoSetSparseDisabled(ni);
1735 	ni->type = AT_DATA;
1736 	ni->name = NULL;
1737 	ni->name_len = 0;
1738 	/*
1739 	 * This sets up our little cheat allowing us to reuse the async read io
1740 	 * completion handler for directories.
1741 	 */
1742 	ni->itype.index.block_size = vol->mft_record_size;
1743 	ni->itype.index.block_size_bits = vol->mft_record_size_bits;
1744 
1745 	/* Very important! Needed to be able to call map_mft_record*(). */
1746 	vol->mft_ino = vi;
1747 
1748 	/* Allocate enough memory to read the first mft record. */
1749 	if (vol->mft_record_size > 64 * 1024) {
1750 		ntfs_error(sb, "Unsupported mft record size %i (max 64kiB).",
1751 				vol->mft_record_size);
1752 		goto err_out;
1753 	}
1754 	i = vol->mft_record_size;
1755 	if (i < sb->s_blocksize)
1756 		i = sb->s_blocksize;
1757 	m = (MFT_RECORD*)ntfs_malloc_nofs(i);
1758 	if (!m) {
1759 		ntfs_error(sb, "Failed to allocate buffer for $MFT record 0.");
1760 		goto err_out;
1761 	}
1762 
1763 	/* Determine the first block of the $MFT/$DATA attribute. */
1764 	block = vol->mft_lcn << vol->cluster_size_bits >>
1765 			sb->s_blocksize_bits;
1766 	nr_blocks = vol->mft_record_size >> sb->s_blocksize_bits;
1767 	if (!nr_blocks)
1768 		nr_blocks = 1;
1769 
1770 	/* Load $MFT/$DATA's first mft record. */
1771 	for (i = 0; i < nr_blocks; i++) {
1772 		bh = sb_bread(sb, block++);
1773 		if (!bh) {
1774 			ntfs_error(sb, "Device read failed.");
1775 			goto err_out;
1776 		}
1777 		memcpy((char*)m + (i << sb->s_blocksize_bits), bh->b_data,
1778 				sb->s_blocksize);
1779 		brelse(bh);
1780 	}
1781 
1782 	/* Apply the mst fixups. */
1783 	if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) {
1784 		/* FIXME: Try to use the $MFTMirr now. */
1785 		ntfs_error(sb, "MST fixup failed. $MFT is corrupt.");
1786 		goto err_out;
1787 	}
1788 
1789 	/* Need this to sanity check attribute list references to $MFT. */
1790 	vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
1791 
1792 	/* Provides readpage() and sync_page() for map_mft_record(). */
1793 	vi->i_mapping->a_ops = &ntfs_mst_aops;
1794 
1795 	ctx = ntfs_attr_get_search_ctx(ni, m);
1796 	if (!ctx) {
1797 		err = -ENOMEM;
1798 		goto err_out;
1799 	}
1800 
1801 	/* Find the attribute list attribute if present. */
1802 	err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
1803 	if (err) {
1804 		if (unlikely(err != -ENOENT)) {
1805 			ntfs_error(sb, "Failed to lookup attribute list "
1806 					"attribute. You should run chkdsk.");
1807 			goto put_err_out;
1808 		}
1809 	} else /* if (!err) */ {
1810 		ATTR_LIST_ENTRY *al_entry, *next_al_entry;
1811 		u8 *al_end;
1812 
1813 		ntfs_debug("Attribute list attribute found in $MFT.");
1814 		NInoSetAttrList(ni);
1815 		a = ctx->attr;
1816 		if (a->flags & ATTR_IS_ENCRYPTED ||
1817 				a->flags & ATTR_COMPRESSION_MASK ||
1818 				a->flags & ATTR_IS_SPARSE) {
1819 			ntfs_error(sb, "Attribute list attribute is "
1820 					"compressed/encrypted/sparse. Not "
1821 					"allowed. $MFT is corrupt. You should "
1822 					"run chkdsk.");
1823 			goto put_err_out;
1824 		}
1825 		/* Now allocate memory for the attribute list. */
1826 		ni->attr_list_size = (u32)ntfs_attr_size(a);
1827 		ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
1828 		if (!ni->attr_list) {
1829 			ntfs_error(sb, "Not enough memory to allocate buffer "
1830 					"for attribute list.");
1831 			goto put_err_out;
1832 		}
1833 		if (a->non_resident) {
1834 			NInoSetAttrListNonResident(ni);
1835 			if (a->data.non_resident.lowest_vcn) {
1836 				ntfs_error(sb, "Attribute list has non zero "
1837 						"lowest_vcn. $MFT is corrupt. "
1838 						"You should run chkdsk.");
1839 				goto put_err_out;
1840 			}
1841 			/* Setup the runlist. */
1842 			ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
1843 					a, NULL);
1844 			if (IS_ERR(ni->attr_list_rl.rl)) {
1845 				err = PTR_ERR(ni->attr_list_rl.rl);
1846 				ni->attr_list_rl.rl = NULL;
1847 				ntfs_error(sb, "Mapping pairs decompression "
1848 						"failed with error code %i.",
1849 						-err);
1850 				goto put_err_out;
1851 			}
1852 			/* Now load the attribute list. */
1853 			if ((err = load_attribute_list(vol, &ni->attr_list_rl,
1854 					ni->attr_list, ni->attr_list_size,
1855 					sle64_to_cpu(a->data.
1856 					non_resident.initialized_size)))) {
1857 				ntfs_error(sb, "Failed to load attribute list "
1858 						"attribute with error code %i.",
1859 						-err);
1860 				goto put_err_out;
1861 			}
1862 		} else /* if (!ctx.attr->non_resident) */ {
1863 			if ((u8*)a + le16_to_cpu(
1864 					a->data.resident.value_offset) +
1865 					le32_to_cpu(
1866 					a->data.resident.value_length) >
1867 					(u8*)ctx->mrec + vol->mft_record_size) {
1868 				ntfs_error(sb, "Corrupt attribute list "
1869 						"attribute.");
1870 				goto put_err_out;
1871 			}
1872 			/* Now copy the attribute list. */
1873 			memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
1874 					a->data.resident.value_offset),
1875 					le32_to_cpu(
1876 					a->data.resident.value_length));
1877 		}
1878 		/* The attribute list is now setup in memory. */
1879 		/*
1880 		 * FIXME: I don't know if this case is actually possible.
1881 		 * According to logic it is not possible but I have seen too
1882 		 * many weird things in MS software to rely on logic... Thus we
1883 		 * perform a manual search and make sure the first $MFT/$DATA
1884 		 * extent is in the base inode. If it is not we abort with an
1885 		 * error and if we ever see a report of this error we will need
1886 		 * to do some magic in order to have the necessary mft record
1887 		 * loaded and in the right place in the page cache. But
1888 		 * hopefully logic will prevail and this never happens...
1889 		 */
1890 		al_entry = (ATTR_LIST_ENTRY*)ni->attr_list;
1891 		al_end = (u8*)al_entry + ni->attr_list_size;
1892 		for (;; al_entry = next_al_entry) {
1893 			/* Out of bounds check. */
1894 			if ((u8*)al_entry < ni->attr_list ||
1895 					(u8*)al_entry > al_end)
1896 				goto em_put_err_out;
1897 			/* Catch the end of the attribute list. */
1898 			if ((u8*)al_entry == al_end)
1899 				goto em_put_err_out;
1900 			if (!al_entry->length)
1901 				goto em_put_err_out;
1902 			if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
1903 					le16_to_cpu(al_entry->length) > al_end)
1904 				goto em_put_err_out;
1905 			next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
1906 					le16_to_cpu(al_entry->length));
1907 			if (le32_to_cpu(al_entry->type) >
1908 					const_le32_to_cpu(AT_DATA))
1909 				goto em_put_err_out;
1910 			if (AT_DATA != al_entry->type)
1911 				continue;
1912 			/* We want an unnamed attribute. */
1913 			if (al_entry->name_length)
1914 				goto em_put_err_out;
1915 			/* Want the first entry, i.e. lowest_vcn == 0. */
1916 			if (al_entry->lowest_vcn)
1917 				goto em_put_err_out;
1918 			/* First entry has to be in the base mft record. */
1919 			if (MREF_LE(al_entry->mft_reference) != vi->i_ino) {
1920 				/* MFT references do not match, logic fails. */
1921 				ntfs_error(sb, "BUG: The first $DATA extent "
1922 						"of $MFT is not in the base "
1923 						"mft record. Please report "
1924 						"you saw this message to "
1925 						"linux-ntfs-dev@lists."
1926 						"sourceforge.net");
1927 				goto put_err_out;
1928 			} else {
1929 				/* Sequence numbers must match. */
1930 				if (MSEQNO_LE(al_entry->mft_reference) !=
1931 						ni->seq_no)
1932 					goto em_put_err_out;
1933 				/* Got it. All is ok. We can stop now. */
1934 				break;
1935 			}
1936 		}
1937 	}
1938 
1939 	ntfs_attr_reinit_search_ctx(ctx);
1940 
1941 	/* Now load all attribute extents. */
1942 	a = NULL;
1943 	next_vcn = last_vcn = highest_vcn = 0;
1944 	while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0,
1945 			ctx))) {
1946 		runlist_element *nrl;
1947 
1948 		/* Cache the current attribute. */
1949 		a = ctx->attr;
1950 		/* $MFT must be non-resident. */
1951 		if (!a->non_resident) {
1952 			ntfs_error(sb, "$MFT must be non-resident but a "
1953 					"resident extent was found. $MFT is "
1954 					"corrupt. Run chkdsk.");
1955 			goto put_err_out;
1956 		}
1957 		/* $MFT must be uncompressed and unencrypted. */
1958 		if (a->flags & ATTR_COMPRESSION_MASK ||
1959 				a->flags & ATTR_IS_ENCRYPTED ||
1960 				a->flags & ATTR_IS_SPARSE) {
1961 			ntfs_error(sb, "$MFT must be uncompressed, "
1962 					"non-sparse, and unencrypted but a "
1963 					"compressed/sparse/encrypted extent "
1964 					"was found. $MFT is corrupt. Run "
1965 					"chkdsk.");
1966 			goto put_err_out;
1967 		}
1968 		/*
1969 		 * Decompress the mapping pairs array of this extent and merge
1970 		 * the result into the existing runlist. No need for locking
1971 		 * as we have exclusive access to the inode at this time and we
1972 		 * are a mount in progress task, too.
1973 		 */
1974 		nrl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl);
1975 		if (IS_ERR(nrl)) {
1976 			ntfs_error(sb, "ntfs_mapping_pairs_decompress() "
1977 					"failed with error code %ld.  $MFT is "
1978 					"corrupt.", PTR_ERR(nrl));
1979 			goto put_err_out;
1980 		}
1981 		ni->runlist.rl = nrl;
1982 
1983 		/* Are we in the first extent? */
1984 		if (!next_vcn) {
1985 			if (a->data.non_resident.lowest_vcn) {
1986 				ntfs_error(sb, "First extent of $DATA "
1987 						"attribute has non zero "
1988 						"lowest_vcn. $MFT is corrupt. "
1989 						"You should run chkdsk.");
1990 				goto put_err_out;
1991 			}
1992 			/* Get the last vcn in the $DATA attribute. */
1993 			last_vcn = sle64_to_cpu(
1994 					a->data.non_resident.allocated_size)
1995 					>> vol->cluster_size_bits;
1996 			/* Fill in the inode size. */
1997 			vi->i_size = sle64_to_cpu(
1998 					a->data.non_resident.data_size);
1999 			ni->initialized_size = sle64_to_cpu(
2000 					a->data.non_resident.initialized_size);
2001 			ni->allocated_size = sle64_to_cpu(
2002 					a->data.non_resident.allocated_size);
2003 			/*
2004 			 * Verify the number of mft records does not exceed
2005 			 * 2^32 - 1.
2006 			 */
2007 			if ((vi->i_size >> vol->mft_record_size_bits) >=
2008 					(1ULL << 32)) {
2009 				ntfs_error(sb, "$MFT is too big! Aborting.");
2010 				goto put_err_out;
2011 			}
2012 			/*
2013 			 * We have got the first extent of the runlist for
2014 			 * $MFT which means it is now relatively safe to call
2015 			 * the normal ntfs_read_inode() function.
2016 			 * Complete reading the inode, this will actually
2017 			 * re-read the mft record for $MFT, this time entering
2018 			 * it into the page cache with which we complete the
2019 			 * kick start of the volume. It should be safe to do
2020 			 * this now as the first extent of $MFT/$DATA is
2021 			 * already known and we would hope that we don't need
2022 			 * further extents in order to find the other
2023 			 * attributes belonging to $MFT. Only time will tell if
2024 			 * this is really the case. If not we will have to play
2025 			 * magic at this point, possibly duplicating a lot of
2026 			 * ntfs_read_inode() at this point. We will need to
2027 			 * ensure we do enough of its work to be able to call
2028 			 * ntfs_read_inode() on extents of $MFT/$DATA. But lets
2029 			 * hope this never happens...
2030 			 */
2031 			ntfs_read_locked_inode(vi);
2032 			if (is_bad_inode(vi)) {
2033 				ntfs_error(sb, "ntfs_read_inode() of $MFT "
2034 						"failed. BUG or corrupt $MFT. "
2035 						"Run chkdsk and if no errors "
2036 						"are found, please report you "
2037 						"saw this message to "
2038 						"linux-ntfs-dev@lists."
2039 						"sourceforge.net");
2040 				ntfs_attr_put_search_ctx(ctx);
2041 				/* Revert to the safe super operations. */
2042 				ntfs_free(m);
2043 				return -1;
2044 			}
2045 			/*
2046 			 * Re-initialize some specifics about $MFT's inode as
2047 			 * ntfs_read_inode() will have set up the default ones.
2048 			 */
2049 			/* Set uid and gid to root. */
2050 			vi->i_uid = vi->i_gid = 0;
2051 			/* Regular file. No access for anyone. */
2052 			vi->i_mode = S_IFREG;
2053 			/* No VFS initiated operations allowed for $MFT. */
2054 			vi->i_op = &ntfs_empty_inode_ops;
2055 			vi->i_fop = &ntfs_empty_file_ops;
2056 		}
2057 
2058 		/* Get the lowest vcn for the next extent. */
2059 		highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2060 		next_vcn = highest_vcn + 1;
2061 
2062 		/* Only one extent or error, which we catch below. */
2063 		if (next_vcn <= 0)
2064 			break;
2065 
2066 		/* Avoid endless loops due to corruption. */
2067 		if (next_vcn < sle64_to_cpu(
2068 				a->data.non_resident.lowest_vcn)) {
2069 			ntfs_error(sb, "$MFT has corrupt attribute list "
2070 					"attribute. Run chkdsk.");
2071 			goto put_err_out;
2072 		}
2073 	}
2074 	if (err != -ENOENT) {
2075 		ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. "
2076 				"$MFT is corrupt. Run chkdsk.");
2077 		goto put_err_out;
2078 	}
2079 	if (!a) {
2080 		ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is "
2081 				"corrupt. Run chkdsk.");
2082 		goto put_err_out;
2083 	}
2084 	if (highest_vcn && highest_vcn != last_vcn - 1) {
2085 		ntfs_error(sb, "Failed to load the complete runlist for "
2086 				"$MFT/$DATA. Driver bug or corrupt $MFT. "
2087 				"Run chkdsk.");
2088 		ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx",
2089 				(unsigned long long)highest_vcn,
2090 				(unsigned long long)last_vcn - 1);
2091 		goto put_err_out;
2092 	}
2093 	ntfs_attr_put_search_ctx(ctx);
2094 	ntfs_debug("Done.");
2095 	ntfs_free(m);
2096 	return 0;
2097 
2098 em_put_err_out:
2099 	ntfs_error(sb, "Couldn't find first extent of $DATA attribute in "
2100 			"attribute list. $MFT is corrupt. Run chkdsk.");
2101 put_err_out:
2102 	ntfs_attr_put_search_ctx(ctx);
2103 err_out:
2104 	ntfs_error(sb, "Failed. Marking inode as bad.");
2105 	make_bad_inode(vi);
2106 	ntfs_free(m);
2107 	return -1;
2108 }
2109 
2110 /**
2111  * ntfs_put_inode - handler for when the inode reference count is decremented
2112  * @vi:		vfs inode
2113  *
2114  * The VFS calls ntfs_put_inode() every time the inode reference count (i_count)
2115  * is about to be decremented (but before the decrement itself.
2116  *
2117  * If the inode @vi is a directory with two references, one of which is being
2118  * dropped, we need to put the attribute inode for the directory index bitmap,
2119  * if it is present, otherwise the directory inode would remain pinned for
2120  * ever.
2121  */
2122 void ntfs_put_inode(struct inode *vi)
2123 {
2124 	if (S_ISDIR(vi->i_mode) && atomic_read(&vi->i_count) == 2) {
2125 		ntfs_inode *ni = NTFS_I(vi);
2126 		if (NInoIndexAllocPresent(ni)) {
2127 			struct inode *bvi = NULL;
2128 			down(&vi->i_sem);
2129 			if (atomic_read(&vi->i_count) == 2) {
2130 				bvi = ni->itype.index.bmp_ino;
2131 				if (bvi)
2132 					ni->itype.index.bmp_ino = NULL;
2133 			}
2134 			up(&vi->i_sem);
2135 			if (bvi)
2136 				iput(bvi);
2137 		}
2138 	}
2139 }
2140 
2141 static void __ntfs_clear_inode(ntfs_inode *ni)
2142 {
2143 	/* Free all alocated memory. */
2144 	down_write(&ni->runlist.lock);
2145 	if (ni->runlist.rl) {
2146 		ntfs_free(ni->runlist.rl);
2147 		ni->runlist.rl = NULL;
2148 	}
2149 	up_write(&ni->runlist.lock);
2150 
2151 	if (ni->attr_list) {
2152 		ntfs_free(ni->attr_list);
2153 		ni->attr_list = NULL;
2154 	}
2155 
2156 	down_write(&ni->attr_list_rl.lock);
2157 	if (ni->attr_list_rl.rl) {
2158 		ntfs_free(ni->attr_list_rl.rl);
2159 		ni->attr_list_rl.rl = NULL;
2160 	}
2161 	up_write(&ni->attr_list_rl.lock);
2162 
2163 	if (ni->name_len && ni->name != I30) {
2164 		/* Catch bugs... */
2165 		BUG_ON(!ni->name);
2166 		kfree(ni->name);
2167 	}
2168 }
2169 
2170 void ntfs_clear_extent_inode(ntfs_inode *ni)
2171 {
2172 	ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
2173 
2174 	BUG_ON(NInoAttr(ni));
2175 	BUG_ON(ni->nr_extents != -1);
2176 
2177 #ifdef NTFS_RW
2178 	if (NInoDirty(ni)) {
2179 		if (!is_bad_inode(VFS_I(ni->ext.base_ntfs_ino)))
2180 			ntfs_error(ni->vol->sb, "Clearing dirty extent inode!  "
2181 					"Losing data!  This is a BUG!!!");
2182 		// FIXME:  Do something!!!
2183 	}
2184 #endif /* NTFS_RW */
2185 
2186 	__ntfs_clear_inode(ni);
2187 
2188 	/* Bye, bye... */
2189 	ntfs_destroy_extent_inode(ni);
2190 }
2191 
2192 /**
2193  * ntfs_clear_big_inode - clean up the ntfs specific part of an inode
2194  * @vi:		vfs inode pending annihilation
2195  *
2196  * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode()
2197  * is called, which deallocates all memory belonging to the NTFS specific part
2198  * of the inode and returns.
2199  *
2200  * If the MFT record is dirty, we commit it before doing anything else.
2201  */
2202 void ntfs_clear_big_inode(struct inode *vi)
2203 {
2204 	ntfs_inode *ni = NTFS_I(vi);
2205 
2206 	/*
2207 	 * If the inode @vi is an index inode we need to put the attribute
2208 	 * inode for the index bitmap, if it is present, otherwise the index
2209 	 * inode would disappear and the attribute inode for the index bitmap
2210 	 * would no longer be referenced from anywhere and thus it would remain
2211 	 * pinned for ever.
2212 	 */
2213 	if (NInoAttr(ni) && (ni->type == AT_INDEX_ALLOCATION) &&
2214 			NInoIndexAllocPresent(ni) && ni->itype.index.bmp_ino) {
2215 		iput(ni->itype.index.bmp_ino);
2216 		ni->itype.index.bmp_ino = NULL;
2217 	}
2218 #ifdef NTFS_RW
2219 	if (NInoDirty(ni)) {
2220 		BOOL was_bad = (is_bad_inode(vi));
2221 
2222 		/* Committing the inode also commits all extent inodes. */
2223 		ntfs_commit_inode(vi);
2224 
2225 		if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) {
2226 			ntfs_error(vi->i_sb, "Failed to commit dirty inode "
2227 					"0x%lx.  Losing data!", vi->i_ino);
2228 			// FIXME:  Do something!!!
2229 		}
2230 	}
2231 #endif /* NTFS_RW */
2232 
2233 	/* No need to lock at this stage as no one else has a reference. */
2234 	if (ni->nr_extents > 0) {
2235 		int i;
2236 
2237 		for (i = 0; i < ni->nr_extents; i++)
2238 			ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]);
2239 		kfree(ni->ext.extent_ntfs_inos);
2240 	}
2241 
2242 	__ntfs_clear_inode(ni);
2243 
2244 	if (NInoAttr(ni)) {
2245 		/* Release the base inode if we are holding it. */
2246 		if (ni->nr_extents == -1) {
2247 			iput(VFS_I(ni->ext.base_ntfs_ino));
2248 			ni->nr_extents = 0;
2249 			ni->ext.base_ntfs_ino = NULL;
2250 		}
2251 	}
2252 	return;
2253 }
2254 
2255 /**
2256  * ntfs_show_options - show mount options in /proc/mounts
2257  * @sf:		seq_file in which to write our mount options
2258  * @mnt:	vfs mount whose mount options to display
2259  *
2260  * Called by the VFS once for each mounted ntfs volume when someone reads
2261  * /proc/mounts in order to display the NTFS specific mount options of each
2262  * mount. The mount options of the vfs mount @mnt are written to the seq file
2263  * @sf and success is returned.
2264  */
2265 int ntfs_show_options(struct seq_file *sf, struct vfsmount *mnt)
2266 {
2267 	ntfs_volume *vol = NTFS_SB(mnt->mnt_sb);
2268 	int i;
2269 
2270 	seq_printf(sf, ",uid=%i", vol->uid);
2271 	seq_printf(sf, ",gid=%i", vol->gid);
2272 	if (vol->fmask == vol->dmask)
2273 		seq_printf(sf, ",umask=0%o", vol->fmask);
2274 	else {
2275 		seq_printf(sf, ",fmask=0%o", vol->fmask);
2276 		seq_printf(sf, ",dmask=0%o", vol->dmask);
2277 	}
2278 	seq_printf(sf, ",nls=%s", vol->nls_map->charset);
2279 	if (NVolCaseSensitive(vol))
2280 		seq_printf(sf, ",case_sensitive");
2281 	if (NVolShowSystemFiles(vol))
2282 		seq_printf(sf, ",show_sys_files");
2283 	if (!NVolSparseEnabled(vol))
2284 		seq_printf(sf, ",disable_sparse");
2285 	for (i = 0; on_errors_arr[i].val; i++) {
2286 		if (on_errors_arr[i].val & vol->on_errors)
2287 			seq_printf(sf, ",errors=%s", on_errors_arr[i].str);
2288 	}
2289 	seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier);
2290 	return 0;
2291 }
2292 
2293 #ifdef NTFS_RW
2294 
2295 static const char *es = "  Leaving inconsistent metadata.  Unmount and run "
2296 		"chkdsk.";
2297 
2298 /**
2299  * ntfs_truncate - called when the i_size of an ntfs inode is changed
2300  * @vi:		inode for which the i_size was changed
2301  *
2302  * We only support i_size changes for normal files at present, i.e. not
2303  * compressed and not encrypted.  This is enforced in ntfs_setattr(), see
2304  * below.
2305  *
2306  * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and
2307  * that the change is allowed.
2308  *
2309  * This implies for us that @vi is a file inode rather than a directory, index,
2310  * or attribute inode as well as that @vi is a base inode.
2311  *
2312  * Returns 0 on success or -errno on error.
2313  *
2314  * Called with ->i_sem held.  In all but one case ->i_alloc_sem is held for
2315  * writing.  The only case in the kernel where ->i_alloc_sem is not held is
2316  * mm/filemap.c::generic_file_buffered_write() where vmtruncate() is called
2317  * with the current i_size as the offset.  The analogous place in NTFS is in
2318  * fs/ntfs/file.c::ntfs_file_buffered_write() where we call vmtruncate() again
2319  * without holding ->i_alloc_sem.
2320  */
2321 int ntfs_truncate(struct inode *vi)
2322 {
2323 	s64 new_size, old_size, nr_freed, new_alloc_size, old_alloc_size;
2324 	VCN highest_vcn;
2325 	unsigned long flags;
2326 	ntfs_inode *base_ni, *ni = NTFS_I(vi);
2327 	ntfs_volume *vol = ni->vol;
2328 	ntfs_attr_search_ctx *ctx;
2329 	MFT_RECORD *m;
2330 	ATTR_RECORD *a;
2331 	const char *te = "  Leaving file length out of sync with i_size.";
2332 	int err, mp_size, size_change, alloc_change;
2333 	u32 attr_len;
2334 
2335 	ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
2336 	BUG_ON(NInoAttr(ni));
2337 	BUG_ON(S_ISDIR(vi->i_mode));
2338 	BUG_ON(NInoMstProtected(ni));
2339 	BUG_ON(ni->nr_extents < 0);
2340 retry_truncate:
2341 	/*
2342 	 * Lock the runlist for writing and map the mft record to ensure it is
2343 	 * safe to mess with the attribute runlist and sizes.
2344 	 */
2345 	down_write(&ni->runlist.lock);
2346 	if (!NInoAttr(ni))
2347 		base_ni = ni;
2348 	else
2349 		base_ni = ni->ext.base_ntfs_ino;
2350 	m = map_mft_record(base_ni);
2351 	if (IS_ERR(m)) {
2352 		err = PTR_ERR(m);
2353 		ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx "
2354 				"(error code %d).%s", vi->i_ino, err, te);
2355 		ctx = NULL;
2356 		m = NULL;
2357 		goto old_bad_out;
2358 	}
2359 	ctx = ntfs_attr_get_search_ctx(base_ni, m);
2360 	if (unlikely(!ctx)) {
2361 		ntfs_error(vi->i_sb, "Failed to allocate a search context for "
2362 				"inode 0x%lx (not enough memory).%s",
2363 				vi->i_ino, te);
2364 		err = -ENOMEM;
2365 		goto old_bad_out;
2366 	}
2367 	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2368 			CASE_SENSITIVE, 0, NULL, 0, ctx);
2369 	if (unlikely(err)) {
2370 		if (err == -ENOENT) {
2371 			ntfs_error(vi->i_sb, "Open attribute is missing from "
2372 					"mft record.  Inode 0x%lx is corrupt.  "
2373 					"Run chkdsk.%s", vi->i_ino, te);
2374 			err = -EIO;
2375 		} else
2376 			ntfs_error(vi->i_sb, "Failed to lookup attribute in "
2377 					"inode 0x%lx (error code %d).%s",
2378 					vi->i_ino, err, te);
2379 		goto old_bad_out;
2380 	}
2381 	m = ctx->mrec;
2382 	a = ctx->attr;
2383 	/*
2384 	 * The i_size of the vfs inode is the new size for the attribute value.
2385 	 */
2386 	new_size = i_size_read(vi);
2387 	/* The current size of the attribute value is the old size. */
2388 	old_size = ntfs_attr_size(a);
2389 	/* Calculate the new allocated size. */
2390 	if (NInoNonResident(ni))
2391 		new_alloc_size = (new_size + vol->cluster_size - 1) &
2392 				~(s64)vol->cluster_size_mask;
2393 	else
2394 		new_alloc_size = (new_size + 7) & ~7;
2395 	/* The current allocated size is the old allocated size. */
2396 	read_lock_irqsave(&ni->size_lock, flags);
2397 	old_alloc_size = ni->allocated_size;
2398 	read_unlock_irqrestore(&ni->size_lock, flags);
2399 	/*
2400 	 * The change in the file size.  This will be 0 if no change, >0 if the
2401 	 * size is growing, and <0 if the size is shrinking.
2402 	 */
2403 	size_change = -1;
2404 	if (new_size - old_size >= 0) {
2405 		size_change = 1;
2406 		if (new_size == old_size)
2407 			size_change = 0;
2408 	}
2409 	/* As above for the allocated size. */
2410 	alloc_change = -1;
2411 	if (new_alloc_size - old_alloc_size >= 0) {
2412 		alloc_change = 1;
2413 		if (new_alloc_size == old_alloc_size)
2414 			alloc_change = 0;
2415 	}
2416 	/*
2417 	 * If neither the size nor the allocation are being changed there is
2418 	 * nothing to do.
2419 	 */
2420 	if (!size_change && !alloc_change)
2421 		goto unm_done;
2422 	/* If the size is changing, check if new size is allowed in $AttrDef. */
2423 	if (size_change) {
2424 		err = ntfs_attr_size_bounds_check(vol, ni->type, new_size);
2425 		if (unlikely(err)) {
2426 			if (err == -ERANGE) {
2427 				ntfs_error(vol->sb, "Truncate would cause the "
2428 						"inode 0x%lx to %simum size "
2429 						"for its attribute type "
2430 						"(0x%x).  Aborting truncate.",
2431 						vi->i_ino,
2432 						new_size > old_size ? "exceed "
2433 						"the max" : "go under the min",
2434 						le32_to_cpu(ni->type));
2435 				err = -EFBIG;
2436 			} else {
2437 				ntfs_error(vol->sb, "Inode 0x%lx has unknown "
2438 						"attribute type 0x%x.  "
2439 						"Aborting truncate.",
2440 						vi->i_ino,
2441 						le32_to_cpu(ni->type));
2442 				err = -EIO;
2443 			}
2444 			/* Reset the vfs inode size to the old size. */
2445 			i_size_write(vi, old_size);
2446 			goto err_out;
2447 		}
2448 	}
2449 	if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2450 		ntfs_warning(vi->i_sb, "Changes in inode size are not "
2451 				"supported yet for %s files, ignoring.",
2452 				NInoCompressed(ni) ? "compressed" :
2453 				"encrypted");
2454 		err = -EOPNOTSUPP;
2455 		goto bad_out;
2456 	}
2457 	if (a->non_resident)
2458 		goto do_non_resident_truncate;
2459 	BUG_ON(NInoNonResident(ni));
2460 	/* Resize the attribute record to best fit the new attribute size. */
2461 	if (new_size < vol->mft_record_size &&
2462 			!ntfs_resident_attr_value_resize(m, a, new_size)) {
2463 		unsigned long flags;
2464 
2465 		/* The resize succeeded! */
2466 		flush_dcache_mft_record_page(ctx->ntfs_ino);
2467 		mark_mft_record_dirty(ctx->ntfs_ino);
2468 		write_lock_irqsave(&ni->size_lock, flags);
2469 		/* Update the sizes in the ntfs inode and all is done. */
2470 		ni->allocated_size = le32_to_cpu(a->length) -
2471 				le16_to_cpu(a->data.resident.value_offset);
2472 		/*
2473 		 * Note ntfs_resident_attr_value_resize() has already done any
2474 		 * necessary data clearing in the attribute record.  When the
2475 		 * file is being shrunk vmtruncate() will already have cleared
2476 		 * the top part of the last partial page, i.e. since this is
2477 		 * the resident case this is the page with index 0.  However,
2478 		 * when the file is being expanded, the page cache page data
2479 		 * between the old data_size, i.e. old_size, and the new_size
2480 		 * has not been zeroed.  Fortunately, we do not need to zero it
2481 		 * either since on one hand it will either already be zero due
2482 		 * to both readpage and writepage clearing partial page data
2483 		 * beyond i_size in which case there is nothing to do or in the
2484 		 * case of the file being mmap()ped at the same time, POSIX
2485 		 * specifies that the behaviour is unspecified thus we do not
2486 		 * have to do anything.  This means that in our implementation
2487 		 * in the rare case that the file is mmap()ped and a write
2488 		 * occured into the mmap()ped region just beyond the file size
2489 		 * and writepage has not yet been called to write out the page
2490 		 * (which would clear the area beyond the file size) and we now
2491 		 * extend the file size to incorporate this dirty region
2492 		 * outside the file size, a write of the page would result in
2493 		 * this data being written to disk instead of being cleared.
2494 		 * Given both POSIX and the Linux mmap(2) man page specify that
2495 		 * this corner case is undefined, we choose to leave it like
2496 		 * that as this is much simpler for us as we cannot lock the
2497 		 * relevant page now since we are holding too many ntfs locks
2498 		 * which would result in a lock reversal deadlock.
2499 		 */
2500 		ni->initialized_size = new_size;
2501 		write_unlock_irqrestore(&ni->size_lock, flags);
2502 		goto unm_done;
2503 	}
2504 	/* If the above resize failed, this must be an attribute extension. */
2505 	BUG_ON(size_change < 0);
2506 	/*
2507 	 * We have to drop all the locks so we can call
2508 	 * ntfs_attr_make_non_resident().  This could be optimised by try-
2509 	 * locking the first page cache page and only if that fails dropping
2510 	 * the locks, locking the page, and redoing all the locking and
2511 	 * lookups.  While this would be a huge optimisation, it is not worth
2512 	 * it as this is definitely a slow code path as it only ever can happen
2513 	 * once for any given file.
2514 	 */
2515 	ntfs_attr_put_search_ctx(ctx);
2516 	unmap_mft_record(base_ni);
2517 	up_write(&ni->runlist.lock);
2518 	/*
2519 	 * Not enough space in the mft record, try to make the attribute
2520 	 * non-resident and if successful restart the truncation process.
2521 	 */
2522 	err = ntfs_attr_make_non_resident(ni, old_size);
2523 	if (likely(!err))
2524 		goto retry_truncate;
2525 	/*
2526 	 * Could not make non-resident.  If this is due to this not being
2527 	 * permitted for this attribute type or there not being enough space,
2528 	 * try to make other attributes non-resident.  Otherwise fail.
2529 	 */
2530 	if (unlikely(err != -EPERM && err != -ENOSPC)) {
2531 		ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, attribute "
2532 				"type 0x%x, because the conversion from "
2533 				"resident to non-resident attribute failed "
2534 				"with error code %i.", vi->i_ino,
2535 				(unsigned)le32_to_cpu(ni->type), err);
2536 		if (err != -ENOMEM)
2537 			err = -EIO;
2538 		goto conv_err_out;
2539 	}
2540 	/* TODO: Not implemented from here, abort. */
2541 	if (err == -ENOSPC)
2542 		ntfs_error(vol->sb, "Not enough space in the mft record/on "
2543 				"disk for the non-resident attribute value.  "
2544 				"This case is not implemented yet.");
2545 	else /* if (err == -EPERM) */
2546 		ntfs_error(vol->sb, "This attribute type may not be "
2547 				"non-resident.  This case is not implemented "
2548 				"yet.");
2549 	err = -EOPNOTSUPP;
2550 	goto conv_err_out;
2551 #if 0
2552 	// TODO: Attempt to make other attributes non-resident.
2553 	if (!err)
2554 		goto do_resident_extend;
2555 	/*
2556 	 * Both the attribute list attribute and the standard information
2557 	 * attribute must remain in the base inode.  Thus, if this is one of
2558 	 * these attributes, we have to try to move other attributes out into
2559 	 * extent mft records instead.
2560 	 */
2561 	if (ni->type == AT_ATTRIBUTE_LIST ||
2562 			ni->type == AT_STANDARD_INFORMATION) {
2563 		// TODO: Attempt to move other attributes into extent mft
2564 		// records.
2565 		err = -EOPNOTSUPP;
2566 		if (!err)
2567 			goto do_resident_extend;
2568 		goto err_out;
2569 	}
2570 	// TODO: Attempt to move this attribute to an extent mft record, but
2571 	// only if it is not already the only attribute in an mft record in
2572 	// which case there would be nothing to gain.
2573 	err = -EOPNOTSUPP;
2574 	if (!err)
2575 		goto do_resident_extend;
2576 	/* There is nothing we can do to make enough space. )-: */
2577 	goto err_out;
2578 #endif
2579 do_non_resident_truncate:
2580 	BUG_ON(!NInoNonResident(ni));
2581 	if (alloc_change < 0) {
2582 		highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2583 		if (highest_vcn > 0 &&
2584 				old_alloc_size >> vol->cluster_size_bits >
2585 				highest_vcn + 1) {
2586 			/*
2587 			 * This attribute has multiple extents.  Not yet
2588 			 * supported.
2589 			 */
2590 			ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, "
2591 					"attribute type 0x%x, because the "
2592 					"attribute is highly fragmented (it "
2593 					"consists of multiple extents) and "
2594 					"this case is not implemented yet.",
2595 					vi->i_ino,
2596 					(unsigned)le32_to_cpu(ni->type));
2597 			err = -EOPNOTSUPP;
2598 			goto bad_out;
2599 		}
2600 	}
2601 	/*
2602 	 * If the size is shrinking, need to reduce the initialized_size and
2603 	 * the data_size before reducing the allocation.
2604 	 */
2605 	if (size_change < 0) {
2606 		/*
2607 		 * Make the valid size smaller (i_size is already up-to-date).
2608 		 */
2609 		write_lock_irqsave(&ni->size_lock, flags);
2610 		if (new_size < ni->initialized_size) {
2611 			ni->initialized_size = new_size;
2612 			a->data.non_resident.initialized_size =
2613 					cpu_to_sle64(new_size);
2614 		}
2615 		a->data.non_resident.data_size = cpu_to_sle64(new_size);
2616 		write_unlock_irqrestore(&ni->size_lock, flags);
2617 		flush_dcache_mft_record_page(ctx->ntfs_ino);
2618 		mark_mft_record_dirty(ctx->ntfs_ino);
2619 		/* If the allocated size is not changing, we are done. */
2620 		if (!alloc_change)
2621 			goto unm_done;
2622 		/*
2623 		 * If the size is shrinking it makes no sense for the
2624 		 * allocation to be growing.
2625 		 */
2626 		BUG_ON(alloc_change > 0);
2627 	} else /* if (size_change >= 0) */ {
2628 		/*
2629 		 * The file size is growing or staying the same but the
2630 		 * allocation can be shrinking, growing or staying the same.
2631 		 */
2632 		if (alloc_change > 0) {
2633 			/*
2634 			 * We need to extend the allocation and possibly update
2635 			 * the data size.  If we are updating the data size,
2636 			 * since we are not touching the initialized_size we do
2637 			 * not need to worry about the actual data on disk.
2638 			 * And as far as the page cache is concerned, there
2639 			 * will be no pages beyond the old data size and any
2640 			 * partial region in the last page between the old and
2641 			 * new data size (or the end of the page if the new
2642 			 * data size is outside the page) does not need to be
2643 			 * modified as explained above for the resident
2644 			 * attribute truncate case.  To do this, we simply drop
2645 			 * the locks we hold and leave all the work to our
2646 			 * friendly helper ntfs_attr_extend_allocation().
2647 			 */
2648 			ntfs_attr_put_search_ctx(ctx);
2649 			unmap_mft_record(base_ni);
2650 			up_write(&ni->runlist.lock);
2651 			err = ntfs_attr_extend_allocation(ni, new_size,
2652 					size_change > 0 ? new_size : -1, -1);
2653 			/*
2654 			 * ntfs_attr_extend_allocation() will have done error
2655 			 * output already.
2656 			 */
2657 			goto done;
2658 		}
2659 		if (!alloc_change)
2660 			goto alloc_done;
2661 	}
2662 	/* alloc_change < 0 */
2663 	/* Free the clusters. */
2664 	nr_freed = ntfs_cluster_free(ni, new_alloc_size >>
2665 			vol->cluster_size_bits, -1, ctx);
2666 	m = ctx->mrec;
2667 	a = ctx->attr;
2668 	if (unlikely(nr_freed < 0)) {
2669 		ntfs_error(vol->sb, "Failed to release cluster(s) (error code "
2670 				"%lli).  Unmount and run chkdsk to recover "
2671 				"the lost cluster(s).", (long long)nr_freed);
2672 		NVolSetErrors(vol);
2673 		nr_freed = 0;
2674 	}
2675 	/* Truncate the runlist. */
2676 	err = ntfs_rl_truncate_nolock(vol, &ni->runlist,
2677 			new_alloc_size >> vol->cluster_size_bits);
2678 	/*
2679 	 * If the runlist truncation failed and/or the search context is no
2680 	 * longer valid, we cannot resize the attribute record or build the
2681 	 * mapping pairs array thus we mark the inode bad so that no access to
2682 	 * the freed clusters can happen.
2683 	 */
2684 	if (unlikely(err || IS_ERR(m))) {
2685 		ntfs_error(vol->sb, "Failed to %s (error code %li).%s",
2686 				IS_ERR(m) ?
2687 				"restore attribute search context" :
2688 				"truncate attribute runlist",
2689 				IS_ERR(m) ? PTR_ERR(m) : err, es);
2690 		err = -EIO;
2691 		goto bad_out;
2692 	}
2693 	/* Get the size for the shrunk mapping pairs array for the runlist. */
2694 	mp_size = ntfs_get_size_for_mapping_pairs(vol, ni->runlist.rl, 0, -1);
2695 	if (unlikely(mp_size <= 0)) {
2696 		ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2697 				"attribute type 0x%x, because determining the "
2698 				"size for the mapping pairs failed with error "
2699 				"code %i.%s", vi->i_ino,
2700 				(unsigned)le32_to_cpu(ni->type), mp_size, es);
2701 		err = -EIO;
2702 		goto bad_out;
2703 	}
2704 	/*
2705 	 * Shrink the attribute record for the new mapping pairs array.  Note,
2706 	 * this cannot fail since we are making the attribute smaller thus by
2707 	 * definition there is enough space to do so.
2708 	 */
2709 	attr_len = le32_to_cpu(a->length);
2710 	err = ntfs_attr_record_resize(m, a, mp_size +
2711 			le16_to_cpu(a->data.non_resident.mapping_pairs_offset));
2712 	BUG_ON(err);
2713 	/*
2714 	 * Generate the mapping pairs array directly into the attribute record.
2715 	 */
2716 	err = ntfs_mapping_pairs_build(vol, (u8*)a +
2717 			le16_to_cpu(a->data.non_resident.mapping_pairs_offset),
2718 			mp_size, ni->runlist.rl, 0, -1, NULL);
2719 	if (unlikely(err)) {
2720 		ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2721 				"attribute type 0x%x, because building the "
2722 				"mapping pairs failed with error code %i.%s",
2723 				vi->i_ino, (unsigned)le32_to_cpu(ni->type),
2724 				err, es);
2725 		err = -EIO;
2726 		goto bad_out;
2727 	}
2728 	/* Update the allocated/compressed size as well as the highest vcn. */
2729 	a->data.non_resident.highest_vcn = cpu_to_sle64((new_alloc_size >>
2730 			vol->cluster_size_bits) - 1);
2731 	write_lock_irqsave(&ni->size_lock, flags);
2732 	ni->allocated_size = new_alloc_size;
2733 	a->data.non_resident.allocated_size = cpu_to_sle64(new_alloc_size);
2734 	if (NInoSparse(ni) || NInoCompressed(ni)) {
2735 		if (nr_freed) {
2736 			ni->itype.compressed.size -= nr_freed <<
2737 					vol->cluster_size_bits;
2738 			BUG_ON(ni->itype.compressed.size < 0);
2739 			a->data.non_resident.compressed_size = cpu_to_sle64(
2740 					ni->itype.compressed.size);
2741 			vi->i_blocks = ni->itype.compressed.size >> 9;
2742 		}
2743 	} else
2744 		vi->i_blocks = new_alloc_size >> 9;
2745 	write_unlock_irqrestore(&ni->size_lock, flags);
2746 	/*
2747 	 * We have shrunk the allocation.  If this is a shrinking truncate we
2748 	 * have already dealt with the initialized_size and the data_size above
2749 	 * and we are done.  If the truncate is only changing the allocation
2750 	 * and not the data_size, we are also done.  If this is an extending
2751 	 * truncate, need to extend the data_size now which is ensured by the
2752 	 * fact that @size_change is positive.
2753 	 */
2754 alloc_done:
2755 	/*
2756 	 * If the size is growing, need to update it now.  If it is shrinking,
2757 	 * we have already updated it above (before the allocation change).
2758 	 */
2759 	if (size_change > 0)
2760 		a->data.non_resident.data_size = cpu_to_sle64(new_size);
2761 	/* Ensure the modified mft record is written out. */
2762 	flush_dcache_mft_record_page(ctx->ntfs_ino);
2763 	mark_mft_record_dirty(ctx->ntfs_ino);
2764 unm_done:
2765 	ntfs_attr_put_search_ctx(ctx);
2766 	unmap_mft_record(base_ni);
2767 	up_write(&ni->runlist.lock);
2768 done:
2769 	/* Update the mtime and ctime on the base inode. */
2770 	inode_update_time(VFS_I(base_ni), 1);
2771 	if (likely(!err)) {
2772 		NInoClearTruncateFailed(ni);
2773 		ntfs_debug("Done.");
2774 	}
2775 	return err;
2776 old_bad_out:
2777 	old_size = -1;
2778 bad_out:
2779 	if (err != -ENOMEM && err != -EOPNOTSUPP) {
2780 		make_bad_inode(vi);
2781 		make_bad_inode(VFS_I(base_ni));
2782 		NVolSetErrors(vol);
2783 	}
2784 	if (err != -EOPNOTSUPP)
2785 		NInoSetTruncateFailed(ni);
2786 	else if (old_size >= 0)
2787 		i_size_write(vi, old_size);
2788 err_out:
2789 	if (ctx)
2790 		ntfs_attr_put_search_ctx(ctx);
2791 	if (m)
2792 		unmap_mft_record(base_ni);
2793 	up_write(&ni->runlist.lock);
2794 out:
2795 	ntfs_debug("Failed.  Returning error code %i.", err);
2796 	return err;
2797 conv_err_out:
2798 	if (err != -ENOMEM && err != -EOPNOTSUPP) {
2799 		make_bad_inode(vi);
2800 		make_bad_inode(VFS_I(base_ni));
2801 		NVolSetErrors(vol);
2802 	}
2803 	if (err != -EOPNOTSUPP)
2804 		NInoSetTruncateFailed(ni);
2805 	else
2806 		i_size_write(vi, old_size);
2807 	goto out;
2808 }
2809 
2810 /**
2811  * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value
2812  * @vi:		inode for which the i_size was changed
2813  *
2814  * Wrapper for ntfs_truncate() that has no return value.
2815  *
2816  * See ntfs_truncate() description above for details.
2817  */
2818 void ntfs_truncate_vfs(struct inode *vi) {
2819 	ntfs_truncate(vi);
2820 }
2821 
2822 /**
2823  * ntfs_setattr - called from notify_change() when an attribute is being changed
2824  * @dentry:	dentry whose attributes to change
2825  * @attr:	structure describing the attributes and the changes
2826  *
2827  * We have to trap VFS attempts to truncate the file described by @dentry as
2828  * soon as possible, because we do not implement changes in i_size yet.  So we
2829  * abort all i_size changes here.
2830  *
2831  * We also abort all changes of user, group, and mode as we do not implement
2832  * the NTFS ACLs yet.
2833  *
2834  * Called with ->i_sem held.  For the ATTR_SIZE (i.e. ->truncate) case, also
2835  * called with ->i_alloc_sem held for writing.
2836  *
2837  * Basically this is a copy of generic notify_change() and inode_setattr()
2838  * functionality, except we intercept and abort changes in i_size.
2839  */
2840 int ntfs_setattr(struct dentry *dentry, struct iattr *attr)
2841 {
2842 	struct inode *vi = dentry->d_inode;
2843 	int err;
2844 	unsigned int ia_valid = attr->ia_valid;
2845 
2846 	err = inode_change_ok(vi, attr);
2847 	if (err)
2848 		goto out;
2849 	/* We do not support NTFS ACLs yet. */
2850 	if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) {
2851 		ntfs_warning(vi->i_sb, "Changes in user/group/mode are not "
2852 				"supported yet, ignoring.");
2853 		err = -EOPNOTSUPP;
2854 		goto out;
2855 	}
2856 	if (ia_valid & ATTR_SIZE) {
2857 		if (attr->ia_size != i_size_read(vi)) {
2858 			ntfs_inode *ni = NTFS_I(vi);
2859 			/*
2860 			 * FIXME: For now we do not support resizing of
2861 			 * compressed or encrypted files yet.
2862 			 */
2863 			if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2864 				ntfs_warning(vi->i_sb, "Changes in inode size "
2865 						"are not supported yet for "
2866 						"%s files, ignoring.",
2867 						NInoCompressed(ni) ?
2868 						"compressed" : "encrypted");
2869 				err = -EOPNOTSUPP;
2870 			} else
2871 				err = vmtruncate(vi, attr->ia_size);
2872 			if (err || ia_valid == ATTR_SIZE)
2873 				goto out;
2874 		} else {
2875 			/*
2876 			 * We skipped the truncate but must still update
2877 			 * timestamps.
2878 			 */
2879 			ia_valid |= ATTR_MTIME | ATTR_CTIME;
2880 		}
2881 	}
2882 	if (ia_valid & ATTR_ATIME)
2883 		vi->i_atime = timespec_trunc(attr->ia_atime,
2884 				vi->i_sb->s_time_gran);
2885 	if (ia_valid & ATTR_MTIME)
2886 		vi->i_mtime = timespec_trunc(attr->ia_mtime,
2887 				vi->i_sb->s_time_gran);
2888 	if (ia_valid & ATTR_CTIME)
2889 		vi->i_ctime = timespec_trunc(attr->ia_ctime,
2890 				vi->i_sb->s_time_gran);
2891 	mark_inode_dirty(vi);
2892 out:
2893 	return err;
2894 }
2895 
2896 /**
2897  * ntfs_write_inode - write out a dirty inode
2898  * @vi:		inode to write out
2899  * @sync:	if true, write out synchronously
2900  *
2901  * Write out a dirty inode to disk including any extent inodes if present.
2902  *
2903  * If @sync is true, commit the inode to disk and wait for io completion.  This
2904  * is done using write_mft_record().
2905  *
2906  * If @sync is false, just schedule the write to happen but do not wait for i/o
2907  * completion.  In 2.6 kernels, scheduling usually happens just by virtue of
2908  * marking the page (and in this case mft record) dirty but we do not implement
2909  * this yet as write_mft_record() largely ignores the @sync parameter and
2910  * always performs synchronous writes.
2911  *
2912  * Return 0 on success and -errno on error.
2913  */
2914 int ntfs_write_inode(struct inode *vi, int sync)
2915 {
2916 	sle64 nt;
2917 	ntfs_inode *ni = NTFS_I(vi);
2918 	ntfs_attr_search_ctx *ctx;
2919 	MFT_RECORD *m;
2920 	STANDARD_INFORMATION *si;
2921 	int err = 0;
2922 	BOOL modified = FALSE;
2923 
2924 	ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "",
2925 			vi->i_ino);
2926 	/*
2927 	 * Dirty attribute inodes are written via their real inodes so just
2928 	 * clean them here.  Access time updates are taken care off when the
2929 	 * real inode is written.
2930 	 */
2931 	if (NInoAttr(ni)) {
2932 		NInoClearDirty(ni);
2933 		ntfs_debug("Done.");
2934 		return 0;
2935 	}
2936 	/* Map, pin, and lock the mft record belonging to the inode. */
2937 	m = map_mft_record(ni);
2938 	if (IS_ERR(m)) {
2939 		err = PTR_ERR(m);
2940 		goto err_out;
2941 	}
2942 	/* Update the access times in the standard information attribute. */
2943 	ctx = ntfs_attr_get_search_ctx(ni, m);
2944 	if (unlikely(!ctx)) {
2945 		err = -ENOMEM;
2946 		goto unm_err_out;
2947 	}
2948 	err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0,
2949 			CASE_SENSITIVE, 0, NULL, 0, ctx);
2950 	if (unlikely(err)) {
2951 		ntfs_attr_put_search_ctx(ctx);
2952 		goto unm_err_out;
2953 	}
2954 	si = (STANDARD_INFORMATION*)((u8*)ctx->attr +
2955 			le16_to_cpu(ctx->attr->data.resident.value_offset));
2956 	/* Update the access times if they have changed. */
2957 	nt = utc2ntfs(vi->i_mtime);
2958 	if (si->last_data_change_time != nt) {
2959 		ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, "
2960 				"new = 0x%llx", vi->i_ino, (long long)
2961 				sle64_to_cpu(si->last_data_change_time),
2962 				(long long)sle64_to_cpu(nt));
2963 		si->last_data_change_time = nt;
2964 		modified = TRUE;
2965 	}
2966 	nt = utc2ntfs(vi->i_ctime);
2967 	if (si->last_mft_change_time != nt) {
2968 		ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, "
2969 				"new = 0x%llx", vi->i_ino, (long long)
2970 				sle64_to_cpu(si->last_mft_change_time),
2971 				(long long)sle64_to_cpu(nt));
2972 		si->last_mft_change_time = nt;
2973 		modified = TRUE;
2974 	}
2975 	nt = utc2ntfs(vi->i_atime);
2976 	if (si->last_access_time != nt) {
2977 		ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, "
2978 				"new = 0x%llx", vi->i_ino,
2979 				(long long)sle64_to_cpu(si->last_access_time),
2980 				(long long)sle64_to_cpu(nt));
2981 		si->last_access_time = nt;
2982 		modified = TRUE;
2983 	}
2984 	/*
2985 	 * If we just modified the standard information attribute we need to
2986 	 * mark the mft record it is in dirty.  We do this manually so that
2987 	 * mark_inode_dirty() is not called which would redirty the inode and
2988 	 * hence result in an infinite loop of trying to write the inode.
2989 	 * There is no need to mark the base inode nor the base mft record
2990 	 * dirty, since we are going to write this mft record below in any case
2991 	 * and the base mft record may actually not have been modified so it
2992 	 * might not need to be written out.
2993 	 * NOTE: It is not a problem when the inode for $MFT itself is being
2994 	 * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES
2995 	 * on the $MFT inode and hence ntfs_write_inode() will not be
2996 	 * re-invoked because of it which in turn is ok since the dirtied mft
2997 	 * record will be cleaned and written out to disk below, i.e. before
2998 	 * this function returns.
2999 	 */
3000 	if (modified && !NInoTestSetDirty(ctx->ntfs_ino))
3001 		mark_ntfs_record_dirty(ctx->ntfs_ino->page,
3002 				ctx->ntfs_ino->page_ofs);
3003 	ntfs_attr_put_search_ctx(ctx);
3004 	/* Now the access times are updated, write the base mft record. */
3005 	if (NInoDirty(ni))
3006 		err = write_mft_record(ni, m, sync);
3007 	/* Write all attached extent mft records. */
3008 	down(&ni->extent_lock);
3009 	if (ni->nr_extents > 0) {
3010 		ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos;
3011 		int i;
3012 
3013 		ntfs_debug("Writing %i extent inodes.", ni->nr_extents);
3014 		for (i = 0; i < ni->nr_extents; i++) {
3015 			ntfs_inode *tni = extent_nis[i];
3016 
3017 			if (NInoDirty(tni)) {
3018 				MFT_RECORD *tm = map_mft_record(tni);
3019 				int ret;
3020 
3021 				if (IS_ERR(tm)) {
3022 					if (!err || err == -ENOMEM)
3023 						err = PTR_ERR(tm);
3024 					continue;
3025 				}
3026 				ret = write_mft_record(tni, tm, sync);
3027 				unmap_mft_record(tni);
3028 				if (unlikely(ret)) {
3029 					if (!err || err == -ENOMEM)
3030 						err = ret;
3031 				}
3032 			}
3033 		}
3034 	}
3035 	up(&ni->extent_lock);
3036 	unmap_mft_record(ni);
3037 	if (unlikely(err))
3038 		goto err_out;
3039 	ntfs_debug("Done.");
3040 	return 0;
3041 unm_err_out:
3042 	unmap_mft_record(ni);
3043 err_out:
3044 	if (err == -ENOMEM) {
3045 		ntfs_warning(vi->i_sb, "Not enough memory to write inode.  "
3046 				"Marking the inode dirty again, so the VFS "
3047 				"retries later.");
3048 		mark_inode_dirty(vi);
3049 	} else {
3050 		ntfs_error(vi->i_sb, "Failed (error code %i):  Marking inode "
3051 				"as bad.  You should run chkdsk.", -err);
3052 		make_bad_inode(vi);
3053 		NVolSetErrors(ni->vol);
3054 	}
3055 	return err;
3056 }
3057 
3058 #endif /* NTFS_RW */
3059