xref: /linux/fs/ntfs3/fsntfs.c (revision a544684b790f3e9f75173b3b42d7dad1c89dd237)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *
4  * Copyright (C) 2019-2021 Paragon Software GmbH, All rights reserved.
5  *
6  */
7 
8 #include <linux/blkdev.h>
9 #include <linux/buffer_head.h>
10 #include <linux/fs.h>
11 #include <linux/kernel.h>
12 
13 #include "debug.h"
14 #include "ntfs.h"
15 #include "ntfs_fs.h"
16 
17 // clang-format off
18 const struct cpu_str NAME_MFT = {
19 	4, 0, { '$', 'M', 'F', 'T' },
20 };
21 const struct cpu_str NAME_MIRROR = {
22 	8, 0, { '$', 'M', 'F', 'T', 'M', 'i', 'r', 'r' },
23 };
24 const struct cpu_str NAME_LOGFILE = {
25 	8, 0, { '$', 'L', 'o', 'g', 'F', 'i', 'l', 'e' },
26 };
27 const struct cpu_str NAME_VOLUME = {
28 	7, 0, { '$', 'V', 'o', 'l', 'u', 'm', 'e' },
29 };
30 const struct cpu_str NAME_ATTRDEF = {
31 	8, 0, { '$', 'A', 't', 't', 'r', 'D', 'e', 'f' },
32 };
33 const struct cpu_str NAME_ROOT = {
34 	1, 0, { '.' },
35 };
36 const struct cpu_str NAME_BITMAP = {
37 	7, 0, { '$', 'B', 'i', 't', 'm', 'a', 'p' },
38 };
39 const struct cpu_str NAME_BOOT = {
40 	5, 0, { '$', 'B', 'o', 'o', 't' },
41 };
42 const struct cpu_str NAME_BADCLUS = {
43 	8, 0, { '$', 'B', 'a', 'd', 'C', 'l', 'u', 's' },
44 };
45 const struct cpu_str NAME_QUOTA = {
46 	6, 0, { '$', 'Q', 'u', 'o', 't', 'a' },
47 };
48 const struct cpu_str NAME_SECURE = {
49 	7, 0, { '$', 'S', 'e', 'c', 'u', 'r', 'e' },
50 };
51 const struct cpu_str NAME_UPCASE = {
52 	7, 0, { '$', 'U', 'p', 'C', 'a', 's', 'e' },
53 };
54 const struct cpu_str NAME_EXTEND = {
55 	7, 0, { '$', 'E', 'x', 't', 'e', 'n', 'd' },
56 };
57 const struct cpu_str NAME_OBJID = {
58 	6, 0, { '$', 'O', 'b', 'j', 'I', 'd' },
59 };
60 const struct cpu_str NAME_REPARSE = {
61 	8, 0, { '$', 'R', 'e', 'p', 'a', 'r', 's', 'e' },
62 };
63 const struct cpu_str NAME_USNJRNL = {
64 	8, 0, { '$', 'U', 's', 'n', 'J', 'r', 'n', 'l' },
65 };
66 const __le16 BAD_NAME[4] = {
67 	cpu_to_le16('$'), cpu_to_le16('B'), cpu_to_le16('a'), cpu_to_le16('d'),
68 };
69 const __le16 I30_NAME[4] = {
70 	cpu_to_le16('$'), cpu_to_le16('I'), cpu_to_le16('3'), cpu_to_le16('0'),
71 };
72 const __le16 SII_NAME[4] = {
73 	cpu_to_le16('$'), cpu_to_le16('S'), cpu_to_le16('I'), cpu_to_le16('I'),
74 };
75 const __le16 SDH_NAME[4] = {
76 	cpu_to_le16('$'), cpu_to_le16('S'), cpu_to_le16('D'), cpu_to_le16('H'),
77 };
78 const __le16 SDS_NAME[4] = {
79 	cpu_to_le16('$'), cpu_to_le16('S'), cpu_to_le16('D'), cpu_to_le16('S'),
80 };
81 const __le16 SO_NAME[2] = {
82 	cpu_to_le16('$'), cpu_to_le16('O'),
83 };
84 const __le16 SQ_NAME[2] = {
85 	cpu_to_le16('$'), cpu_to_le16('Q'),
86 };
87 const __le16 SR_NAME[2] = {
88 	cpu_to_le16('$'), cpu_to_le16('R'),
89 };
90 
91 #ifdef CONFIG_NTFS3_LZX_XPRESS
92 const __le16 WOF_NAME[17] = {
93 	cpu_to_le16('W'), cpu_to_le16('o'), cpu_to_le16('f'), cpu_to_le16('C'),
94 	cpu_to_le16('o'), cpu_to_le16('m'), cpu_to_le16('p'), cpu_to_le16('r'),
95 	cpu_to_le16('e'), cpu_to_le16('s'), cpu_to_le16('s'), cpu_to_le16('e'),
96 	cpu_to_le16('d'), cpu_to_le16('D'), cpu_to_le16('a'), cpu_to_le16('t'),
97 	cpu_to_le16('a'),
98 };
99 #endif
100 
101 // clang-format on
102 
103 /*
104  * ntfs_fix_pre_write - Insert fixups into @rhdr before writing to disk.
105  */
106 bool ntfs_fix_pre_write(struct NTFS_RECORD_HEADER *rhdr, size_t bytes)
107 {
108 	u16 *fixup, *ptr;
109 	u16 sample;
110 	u16 fo = le16_to_cpu(rhdr->fix_off);
111 	u16 fn = le16_to_cpu(rhdr->fix_num);
112 
113 	if ((fo & 1) || fo + fn * sizeof(short) > SECTOR_SIZE || !fn-- ||
114 	    fn * SECTOR_SIZE > bytes) {
115 		return false;
116 	}
117 
118 	/* Get fixup pointer. */
119 	fixup = Add2Ptr(rhdr, fo);
120 
121 	if (*fixup >= 0x7FFF)
122 		*fixup = 1;
123 	else
124 		*fixup += 1;
125 
126 	sample = *fixup;
127 
128 	ptr = Add2Ptr(rhdr, SECTOR_SIZE - sizeof(short));
129 
130 	while (fn--) {
131 		*++fixup = *ptr;
132 		*ptr = sample;
133 		ptr += SECTOR_SIZE / sizeof(short);
134 	}
135 	return true;
136 }
137 
138 /*
139  * ntfs_fix_post_read - Remove fixups after reading from disk.
140  *
141  * Return: < 0 if error, 0 if ok, 1 if need to update fixups.
142  */
143 int ntfs_fix_post_read(struct NTFS_RECORD_HEADER *rhdr, size_t bytes,
144 		       bool simple)
145 {
146 	int ret;
147 	u16 *fixup, *ptr;
148 	u16 sample, fo, fn;
149 
150 	fo = le16_to_cpu(rhdr->fix_off);
151 	fn = simple ? ((bytes >> SECTOR_SHIFT) + 1)
152 		    : le16_to_cpu(rhdr->fix_num);
153 
154 	/* Check errors. */
155 	if ((fo & 1) || fo + fn * sizeof(short) > SECTOR_SIZE || !fn-- ||
156 	    fn * SECTOR_SIZE > bytes) {
157 		return -EINVAL; /* Native chkntfs returns ok! */
158 	}
159 
160 	/* Get fixup pointer. */
161 	fixup = Add2Ptr(rhdr, fo);
162 	sample = *fixup;
163 	ptr = Add2Ptr(rhdr, SECTOR_SIZE - sizeof(short));
164 	ret = 0;
165 
166 	while (fn--) {
167 		/* Test current word. */
168 		if (*ptr != sample) {
169 			/* Fixup does not match! Is it serious error? */
170 			ret = -E_NTFS_FIXUP;
171 		}
172 
173 		/* Replace fixup. */
174 		*ptr = *++fixup;
175 		ptr += SECTOR_SIZE / sizeof(short);
176 	}
177 
178 	return ret;
179 }
180 
181 /*
182  * ntfs_extend_init - Load $Extend file.
183  */
184 int ntfs_extend_init(struct ntfs_sb_info *sbi)
185 {
186 	int err;
187 	struct super_block *sb = sbi->sb;
188 	struct inode *inode, *inode2;
189 	struct MFT_REF ref;
190 
191 	if (sbi->volume.major_ver < 3) {
192 		ntfs_notice(sb, "Skip $Extend 'cause NTFS version");
193 		return 0;
194 	}
195 
196 	ref.low = cpu_to_le32(MFT_REC_EXTEND);
197 	ref.high = 0;
198 	ref.seq = cpu_to_le16(MFT_REC_EXTEND);
199 	inode = ntfs_iget5(sb, &ref, &NAME_EXTEND);
200 	if (IS_ERR(inode)) {
201 		err = PTR_ERR(inode);
202 		ntfs_err(sb, "Failed to load $Extend.");
203 		inode = NULL;
204 		goto out;
205 	}
206 
207 	/* If ntfs_iget5() reads from disk it never returns bad inode. */
208 	if (!S_ISDIR(inode->i_mode)) {
209 		err = -EINVAL;
210 		goto out;
211 	}
212 
213 	/* Try to find $ObjId */
214 	inode2 = dir_search_u(inode, &NAME_OBJID, NULL);
215 	if (inode2 && !IS_ERR(inode2)) {
216 		if (is_bad_inode(inode2)) {
217 			iput(inode2);
218 		} else {
219 			sbi->objid.ni = ntfs_i(inode2);
220 			sbi->objid_no = inode2->i_ino;
221 		}
222 	}
223 
224 	/* Try to find $Quota */
225 	inode2 = dir_search_u(inode, &NAME_QUOTA, NULL);
226 	if (inode2 && !IS_ERR(inode2)) {
227 		sbi->quota_no = inode2->i_ino;
228 		iput(inode2);
229 	}
230 
231 	/* Try to find $Reparse */
232 	inode2 = dir_search_u(inode, &NAME_REPARSE, NULL);
233 	if (inode2 && !IS_ERR(inode2)) {
234 		sbi->reparse.ni = ntfs_i(inode2);
235 		sbi->reparse_no = inode2->i_ino;
236 	}
237 
238 	/* Try to find $UsnJrnl */
239 	inode2 = dir_search_u(inode, &NAME_USNJRNL, NULL);
240 	if (inode2 && !IS_ERR(inode2)) {
241 		sbi->usn_jrnl_no = inode2->i_ino;
242 		iput(inode2);
243 	}
244 
245 	err = 0;
246 out:
247 	iput(inode);
248 	return err;
249 }
250 
251 int ntfs_loadlog_and_replay(struct ntfs_inode *ni, struct ntfs_sb_info *sbi)
252 {
253 	int err = 0;
254 	struct super_block *sb = sbi->sb;
255 	bool initialized = false;
256 	struct MFT_REF ref;
257 	struct inode *inode;
258 
259 	/* Check for 4GB. */
260 	if (ni->vfs_inode.i_size >= 0x100000000ull) {
261 		ntfs_err(sb, "\x24LogFile is too big");
262 		err = -EINVAL;
263 		goto out;
264 	}
265 
266 	sbi->flags |= NTFS_FLAGS_LOG_REPLAYING;
267 
268 	ref.low = cpu_to_le32(MFT_REC_MFT);
269 	ref.high = 0;
270 	ref.seq = cpu_to_le16(1);
271 
272 	inode = ntfs_iget5(sb, &ref, NULL);
273 
274 	if (IS_ERR(inode))
275 		inode = NULL;
276 
277 	if (!inode) {
278 		/* Try to use MFT copy. */
279 		u64 t64 = sbi->mft.lbo;
280 
281 		sbi->mft.lbo = sbi->mft.lbo2;
282 		inode = ntfs_iget5(sb, &ref, NULL);
283 		sbi->mft.lbo = t64;
284 		if (IS_ERR(inode))
285 			inode = NULL;
286 	}
287 
288 	if (!inode) {
289 		err = -EINVAL;
290 		ntfs_err(sb, "Failed to load $MFT.");
291 		goto out;
292 	}
293 
294 	sbi->mft.ni = ntfs_i(inode);
295 
296 	/* LogFile should not contains attribute list. */
297 	err = ni_load_all_mi(sbi->mft.ni);
298 	if (!err)
299 		err = log_replay(ni, &initialized);
300 
301 	iput(inode);
302 	sbi->mft.ni = NULL;
303 
304 	sync_blockdev(sb->s_bdev);
305 	invalidate_bdev(sb->s_bdev);
306 
307 	if (sbi->flags & NTFS_FLAGS_NEED_REPLAY) {
308 		err = 0;
309 		goto out;
310 	}
311 
312 	if (sb_rdonly(sb) || !initialized)
313 		goto out;
314 
315 	/* Fill LogFile by '-1' if it is initialized. */
316 	err = ntfs_bio_fill_1(sbi, &ni->file.run);
317 
318 out:
319 	sbi->flags &= ~NTFS_FLAGS_LOG_REPLAYING;
320 
321 	return err;
322 }
323 
324 /*
325  * ntfs_query_def
326  *
327  * Return: Current ATTR_DEF_ENTRY for given attribute type.
328  */
329 const struct ATTR_DEF_ENTRY *ntfs_query_def(struct ntfs_sb_info *sbi,
330 					    enum ATTR_TYPE type)
331 {
332 	int type_in = le32_to_cpu(type);
333 	size_t min_idx = 0;
334 	size_t max_idx = sbi->def_entries - 1;
335 
336 	while (min_idx <= max_idx) {
337 		size_t i = min_idx + ((max_idx - min_idx) >> 1);
338 		const struct ATTR_DEF_ENTRY *entry = sbi->def_table + i;
339 		int diff = le32_to_cpu(entry->type) - type_in;
340 
341 		if (!diff)
342 			return entry;
343 		if (diff < 0)
344 			min_idx = i + 1;
345 		else if (i)
346 			max_idx = i - 1;
347 		else
348 			return NULL;
349 	}
350 	return NULL;
351 }
352 
353 /*
354  * ntfs_look_for_free_space - Look for a free space in bitmap.
355  */
356 int ntfs_look_for_free_space(struct ntfs_sb_info *sbi, CLST lcn, CLST len,
357 			     CLST *new_lcn, CLST *new_len,
358 			     enum ALLOCATE_OPT opt)
359 {
360 	int err;
361 	CLST alen;
362 	struct super_block *sb = sbi->sb;
363 	size_t alcn, zlen, zeroes, zlcn, zlen2, ztrim, new_zlen;
364 	struct wnd_bitmap *wnd = &sbi->used.bitmap;
365 
366 	down_write_nested(&wnd->rw_lock, BITMAP_MUTEX_CLUSTERS);
367 	if (opt & ALLOCATE_MFT) {
368 		zlen = wnd_zone_len(wnd);
369 
370 		if (!zlen) {
371 			err = ntfs_refresh_zone(sbi);
372 			if (err)
373 				goto up_write;
374 
375 			zlen = wnd_zone_len(wnd);
376 		}
377 
378 		if (!zlen) {
379 			ntfs_err(sbi->sb, "no free space to extend mft");
380 			err = -ENOSPC;
381 			goto up_write;
382 		}
383 
384 		lcn = wnd_zone_bit(wnd);
385 		alen = min_t(CLST, len, zlen);
386 
387 		wnd_zone_set(wnd, lcn + alen, zlen - alen);
388 
389 		err = wnd_set_used(wnd, lcn, alen);
390 		if (err)
391 			goto up_write;
392 
393 		alcn = lcn;
394 		goto space_found;
395 	}
396 	/*
397 	 * 'Cause cluster 0 is always used this value means that we should use
398 	 * cached value of 'next_free_lcn' to improve performance.
399 	 */
400 	if (!lcn)
401 		lcn = sbi->used.next_free_lcn;
402 
403 	if (lcn >= wnd->nbits)
404 		lcn = 0;
405 
406 	alen = wnd_find(wnd, len, lcn, BITMAP_FIND_MARK_AS_USED, &alcn);
407 	if (alen)
408 		goto space_found;
409 
410 	/* Try to use clusters from MftZone. */
411 	zlen = wnd_zone_len(wnd);
412 	zeroes = wnd_zeroes(wnd);
413 
414 	/* Check too big request */
415 	if (len > zeroes + zlen || zlen <= NTFS_MIN_MFT_ZONE) {
416 		err = -ENOSPC;
417 		goto up_write;
418 	}
419 
420 	/* How many clusters to cat from zone. */
421 	zlcn = wnd_zone_bit(wnd);
422 	zlen2 = zlen >> 1;
423 	ztrim = clamp_val(len, zlen2, zlen);
424 	new_zlen = max_t(size_t, zlen - ztrim, NTFS_MIN_MFT_ZONE);
425 
426 	wnd_zone_set(wnd, zlcn, new_zlen);
427 
428 	/* Allocate continues clusters. */
429 	alen = wnd_find(wnd, len, 0,
430 			BITMAP_FIND_MARK_AS_USED | BITMAP_FIND_FULL, &alcn);
431 	if (!alen) {
432 		err = -ENOSPC;
433 		goto up_write;
434 	}
435 
436 space_found:
437 	err = 0;
438 	*new_len = alen;
439 	*new_lcn = alcn;
440 
441 	ntfs_unmap_meta(sb, alcn, alen);
442 
443 	/* Set hint for next requests. */
444 	if (!(opt & ALLOCATE_MFT))
445 		sbi->used.next_free_lcn = alcn + alen;
446 up_write:
447 	up_write(&wnd->rw_lock);
448 	return err;
449 }
450 
451 /*
452  * ntfs_extend_mft - Allocate additional MFT records.
453  *
454  * sbi->mft.bitmap is locked for write.
455  *
456  * NOTE: recursive:
457  *	ntfs_look_free_mft ->
458  *	ntfs_extend_mft ->
459  *	attr_set_size ->
460  *	ni_insert_nonresident ->
461  *	ni_insert_attr ->
462  *	ni_ins_attr_ext ->
463  *	ntfs_look_free_mft ->
464  *	ntfs_extend_mft
465  *
466  * To avoid recursive always allocate space for two new MFT records
467  * see attrib.c: "at least two MFT to avoid recursive loop".
468  */
469 static int ntfs_extend_mft(struct ntfs_sb_info *sbi)
470 {
471 	int err;
472 	struct ntfs_inode *ni = sbi->mft.ni;
473 	size_t new_mft_total;
474 	u64 new_mft_bytes, new_bitmap_bytes;
475 	struct ATTRIB *attr;
476 	struct wnd_bitmap *wnd = &sbi->mft.bitmap;
477 
478 	new_mft_total = (wnd->nbits + MFT_INCREASE_CHUNK + 127) & (CLST)~127;
479 	new_mft_bytes = (u64)new_mft_total << sbi->record_bits;
480 
481 	/* Step 1: Resize $MFT::DATA. */
482 	down_write(&ni->file.run_lock);
483 	err = attr_set_size(ni, ATTR_DATA, NULL, 0, &ni->file.run,
484 			    new_mft_bytes, NULL, false, &attr);
485 
486 	if (err) {
487 		up_write(&ni->file.run_lock);
488 		goto out;
489 	}
490 
491 	attr->nres.valid_size = attr->nres.data_size;
492 	new_mft_total = le64_to_cpu(attr->nres.alloc_size) >> sbi->record_bits;
493 	ni->mi.dirty = true;
494 
495 	/* Step 2: Resize $MFT::BITMAP. */
496 	new_bitmap_bytes = bitmap_size(new_mft_total);
497 
498 	err = attr_set_size(ni, ATTR_BITMAP, NULL, 0, &sbi->mft.bitmap.run,
499 			    new_bitmap_bytes, &new_bitmap_bytes, true, NULL);
500 
501 	/* Refresh MFT Zone if necessary. */
502 	down_write_nested(&sbi->used.bitmap.rw_lock, BITMAP_MUTEX_CLUSTERS);
503 
504 	ntfs_refresh_zone(sbi);
505 
506 	up_write(&sbi->used.bitmap.rw_lock);
507 	up_write(&ni->file.run_lock);
508 
509 	if (err)
510 		goto out;
511 
512 	err = wnd_extend(wnd, new_mft_total);
513 
514 	if (err)
515 		goto out;
516 
517 	ntfs_clear_mft_tail(sbi, sbi->mft.used, new_mft_total);
518 
519 	err = _ni_write_inode(&ni->vfs_inode, 0);
520 out:
521 	return err;
522 }
523 
524 /*
525  * ntfs_look_free_mft - Look for a free MFT record.
526  */
527 int ntfs_look_free_mft(struct ntfs_sb_info *sbi, CLST *rno, bool mft,
528 		       struct ntfs_inode *ni, struct mft_inode **mi)
529 {
530 	int err = 0;
531 	size_t zbit, zlen, from, to, fr;
532 	size_t mft_total;
533 	struct MFT_REF ref;
534 	struct super_block *sb = sbi->sb;
535 	struct wnd_bitmap *wnd = &sbi->mft.bitmap;
536 	u32 ir;
537 
538 	static_assert(sizeof(sbi->mft.reserved_bitmap) * 8 >=
539 		      MFT_REC_FREE - MFT_REC_RESERVED);
540 
541 	if (!mft)
542 		down_write_nested(&wnd->rw_lock, BITMAP_MUTEX_MFT);
543 
544 	zlen = wnd_zone_len(wnd);
545 
546 	/* Always reserve space for MFT. */
547 	if (zlen) {
548 		if (mft) {
549 			zbit = wnd_zone_bit(wnd);
550 			*rno = zbit;
551 			wnd_zone_set(wnd, zbit + 1, zlen - 1);
552 		}
553 		goto found;
554 	}
555 
556 	/* No MFT zone. Find the nearest to '0' free MFT. */
557 	if (!wnd_find(wnd, 1, MFT_REC_FREE, 0, &zbit)) {
558 		/* Resize MFT */
559 		mft_total = wnd->nbits;
560 
561 		err = ntfs_extend_mft(sbi);
562 		if (!err) {
563 			zbit = mft_total;
564 			goto reserve_mft;
565 		}
566 
567 		if (!mft || MFT_REC_FREE == sbi->mft.next_reserved)
568 			goto out;
569 
570 		err = 0;
571 
572 		/*
573 		 * Look for free record reserved area [11-16) ==
574 		 * [MFT_REC_RESERVED, MFT_REC_FREE ) MFT bitmap always
575 		 * marks it as used.
576 		 */
577 		if (!sbi->mft.reserved_bitmap) {
578 			/* Once per session create internal bitmap for 5 bits. */
579 			sbi->mft.reserved_bitmap = 0xFF;
580 
581 			ref.high = 0;
582 			for (ir = MFT_REC_RESERVED; ir < MFT_REC_FREE; ir++) {
583 				struct inode *i;
584 				struct ntfs_inode *ni;
585 				struct MFT_REC *mrec;
586 
587 				ref.low = cpu_to_le32(ir);
588 				ref.seq = cpu_to_le16(ir);
589 
590 				i = ntfs_iget5(sb, &ref, NULL);
591 				if (IS_ERR(i)) {
592 next:
593 					ntfs_notice(
594 						sb,
595 						"Invalid reserved record %x",
596 						ref.low);
597 					continue;
598 				}
599 				if (is_bad_inode(i)) {
600 					iput(i);
601 					goto next;
602 				}
603 
604 				ni = ntfs_i(i);
605 
606 				mrec = ni->mi.mrec;
607 
608 				if (!is_rec_base(mrec))
609 					goto next;
610 
611 				if (mrec->hard_links)
612 					goto next;
613 
614 				if (!ni_std(ni))
615 					goto next;
616 
617 				if (ni_find_attr(ni, NULL, NULL, ATTR_NAME,
618 						 NULL, 0, NULL, NULL))
619 					goto next;
620 
621 				__clear_bit(ir - MFT_REC_RESERVED,
622 					    &sbi->mft.reserved_bitmap);
623 			}
624 		}
625 
626 		/* Scan 5 bits for zero. Bit 0 == MFT_REC_RESERVED */
627 		zbit = find_next_zero_bit(&sbi->mft.reserved_bitmap,
628 					  MFT_REC_FREE, MFT_REC_RESERVED);
629 		if (zbit >= MFT_REC_FREE) {
630 			sbi->mft.next_reserved = MFT_REC_FREE;
631 			goto out;
632 		}
633 
634 		zlen = 1;
635 		sbi->mft.next_reserved = zbit;
636 	} else {
637 reserve_mft:
638 		zlen = zbit == MFT_REC_FREE ? (MFT_REC_USER - MFT_REC_FREE) : 4;
639 		if (zbit + zlen > wnd->nbits)
640 			zlen = wnd->nbits - zbit;
641 
642 		while (zlen > 1 && !wnd_is_free(wnd, zbit, zlen))
643 			zlen -= 1;
644 
645 		/* [zbit, zbit + zlen) will be used for MFT itself. */
646 		from = sbi->mft.used;
647 		if (from < zbit)
648 			from = zbit;
649 		to = zbit + zlen;
650 		if (from < to) {
651 			ntfs_clear_mft_tail(sbi, from, to);
652 			sbi->mft.used = to;
653 		}
654 	}
655 
656 	if (mft) {
657 		*rno = zbit;
658 		zbit += 1;
659 		zlen -= 1;
660 	}
661 
662 	wnd_zone_set(wnd, zbit, zlen);
663 
664 found:
665 	if (!mft) {
666 		/* The request to get record for general purpose. */
667 		if (sbi->mft.next_free < MFT_REC_USER)
668 			sbi->mft.next_free = MFT_REC_USER;
669 
670 		for (;;) {
671 			if (sbi->mft.next_free >= sbi->mft.bitmap.nbits) {
672 			} else if (!wnd_find(wnd, 1, MFT_REC_USER, 0, &fr)) {
673 				sbi->mft.next_free = sbi->mft.bitmap.nbits;
674 			} else {
675 				*rno = fr;
676 				sbi->mft.next_free = *rno + 1;
677 				break;
678 			}
679 
680 			err = ntfs_extend_mft(sbi);
681 			if (err)
682 				goto out;
683 		}
684 	}
685 
686 	if (ni && !ni_add_subrecord(ni, *rno, mi)) {
687 		err = -ENOMEM;
688 		goto out;
689 	}
690 
691 	/* We have found a record that are not reserved for next MFT. */
692 	if (*rno >= MFT_REC_FREE)
693 		wnd_set_used(wnd, *rno, 1);
694 	else if (*rno >= MFT_REC_RESERVED && sbi->mft.reserved_bitmap_inited)
695 		__set_bit(*rno - MFT_REC_RESERVED, &sbi->mft.reserved_bitmap);
696 
697 out:
698 	if (!mft)
699 		up_write(&wnd->rw_lock);
700 
701 	return err;
702 }
703 
704 /*
705  * ntfs_mark_rec_free - Mark record as free.
706  */
707 void ntfs_mark_rec_free(struct ntfs_sb_info *sbi, CLST rno)
708 {
709 	struct wnd_bitmap *wnd = &sbi->mft.bitmap;
710 
711 	down_write_nested(&wnd->rw_lock, BITMAP_MUTEX_MFT);
712 	if (rno >= wnd->nbits)
713 		goto out;
714 
715 	if (rno >= MFT_REC_FREE) {
716 		if (!wnd_is_used(wnd, rno, 1))
717 			ntfs_set_state(sbi, NTFS_DIRTY_ERROR);
718 		else
719 			wnd_set_free(wnd, rno, 1);
720 	} else if (rno >= MFT_REC_RESERVED && sbi->mft.reserved_bitmap_inited) {
721 		__clear_bit(rno - MFT_REC_RESERVED, &sbi->mft.reserved_bitmap);
722 	}
723 
724 	if (rno < wnd_zone_bit(wnd))
725 		wnd_zone_set(wnd, rno, 1);
726 	else if (rno < sbi->mft.next_free && rno >= MFT_REC_USER)
727 		sbi->mft.next_free = rno;
728 
729 out:
730 	up_write(&wnd->rw_lock);
731 }
732 
733 /*
734  * ntfs_clear_mft_tail - Format empty records [from, to).
735  *
736  * sbi->mft.bitmap is locked for write.
737  */
738 int ntfs_clear_mft_tail(struct ntfs_sb_info *sbi, size_t from, size_t to)
739 {
740 	int err;
741 	u32 rs;
742 	u64 vbo;
743 	struct runs_tree *run;
744 	struct ntfs_inode *ni;
745 
746 	if (from >= to)
747 		return 0;
748 
749 	rs = sbi->record_size;
750 	ni = sbi->mft.ni;
751 	run = &ni->file.run;
752 
753 	down_read(&ni->file.run_lock);
754 	vbo = (u64)from * rs;
755 	for (; from < to; from++, vbo += rs) {
756 		struct ntfs_buffers nb;
757 
758 		err = ntfs_get_bh(sbi, run, vbo, rs, &nb);
759 		if (err)
760 			goto out;
761 
762 		err = ntfs_write_bh(sbi, &sbi->new_rec->rhdr, &nb, 0);
763 		nb_put(&nb);
764 		if (err)
765 			goto out;
766 	}
767 
768 out:
769 	sbi->mft.used = from;
770 	up_read(&ni->file.run_lock);
771 	return err;
772 }
773 
774 /*
775  * ntfs_refresh_zone - Refresh MFT zone.
776  *
777  * sbi->used.bitmap is locked for rw.
778  * sbi->mft.bitmap is locked for write.
779  * sbi->mft.ni->file.run_lock for write.
780  */
781 int ntfs_refresh_zone(struct ntfs_sb_info *sbi)
782 {
783 	CLST zone_limit, zone_max, lcn, vcn, len;
784 	size_t lcn_s, zlen;
785 	struct wnd_bitmap *wnd = &sbi->used.bitmap;
786 	struct ntfs_inode *ni = sbi->mft.ni;
787 
788 	/* Do not change anything unless we have non empty MFT zone. */
789 	if (wnd_zone_len(wnd))
790 		return 0;
791 
792 	/*
793 	 * Compute the MFT zone at two steps.
794 	 * It would be nice if we are able to allocate 1/8 of
795 	 * total clusters for MFT but not more then 512 MB.
796 	 */
797 	zone_limit = (512 * 1024 * 1024) >> sbi->cluster_bits;
798 	zone_max = wnd->nbits >> 3;
799 	if (zone_max > zone_limit)
800 		zone_max = zone_limit;
801 
802 	vcn = bytes_to_cluster(sbi,
803 			       (u64)sbi->mft.bitmap.nbits << sbi->record_bits);
804 
805 	if (!run_lookup_entry(&ni->file.run, vcn - 1, &lcn, &len, NULL))
806 		lcn = SPARSE_LCN;
807 
808 	/* We should always find Last Lcn for MFT. */
809 	if (lcn == SPARSE_LCN)
810 		return -EINVAL;
811 
812 	lcn_s = lcn + 1;
813 
814 	/* Try to allocate clusters after last MFT run. */
815 	zlen = wnd_find(wnd, zone_max, lcn_s, 0, &lcn_s);
816 	if (!zlen) {
817 		ntfs_notice(sbi->sb, "MftZone: unavailable");
818 		return 0;
819 	}
820 
821 	/* Truncate too large zone. */
822 	wnd_zone_set(wnd, lcn_s, zlen);
823 
824 	return 0;
825 }
826 
827 /*
828  * ntfs_update_mftmirr - Update $MFTMirr data.
829  */
830 int ntfs_update_mftmirr(struct ntfs_sb_info *sbi, int wait)
831 {
832 	int err;
833 	struct super_block *sb = sbi->sb;
834 	u32 blocksize = sb->s_blocksize;
835 	sector_t block1, block2;
836 	u32 bytes;
837 
838 	if (!(sbi->flags & NTFS_FLAGS_MFTMIRR))
839 		return 0;
840 
841 	err = 0;
842 	bytes = sbi->mft.recs_mirr << sbi->record_bits;
843 	block1 = sbi->mft.lbo >> sb->s_blocksize_bits;
844 	block2 = sbi->mft.lbo2 >> sb->s_blocksize_bits;
845 
846 	for (; bytes >= blocksize; bytes -= blocksize) {
847 		struct buffer_head *bh1, *bh2;
848 
849 		bh1 = sb_bread(sb, block1++);
850 		if (!bh1) {
851 			err = -EIO;
852 			goto out;
853 		}
854 
855 		bh2 = sb_getblk(sb, block2++);
856 		if (!bh2) {
857 			put_bh(bh1);
858 			err = -EIO;
859 			goto out;
860 		}
861 
862 		if (buffer_locked(bh2))
863 			__wait_on_buffer(bh2);
864 
865 		lock_buffer(bh2);
866 		memcpy(bh2->b_data, bh1->b_data, blocksize);
867 		set_buffer_uptodate(bh2);
868 		mark_buffer_dirty(bh2);
869 		unlock_buffer(bh2);
870 
871 		put_bh(bh1);
872 		bh1 = NULL;
873 
874 		if (wait)
875 			err = sync_dirty_buffer(bh2);
876 
877 		put_bh(bh2);
878 		if (err)
879 			goto out;
880 	}
881 
882 	sbi->flags &= ~NTFS_FLAGS_MFTMIRR;
883 
884 out:
885 	return err;
886 }
887 
888 /*
889  * ntfs_set_state
890  *
891  * Mount: ntfs_set_state(NTFS_DIRTY_DIRTY)
892  * Umount: ntfs_set_state(NTFS_DIRTY_CLEAR)
893  * NTFS error: ntfs_set_state(NTFS_DIRTY_ERROR)
894  */
895 int ntfs_set_state(struct ntfs_sb_info *sbi, enum NTFS_DIRTY_FLAGS dirty)
896 {
897 	int err;
898 	struct ATTRIB *attr;
899 	struct VOLUME_INFO *info;
900 	struct mft_inode *mi;
901 	struct ntfs_inode *ni;
902 
903 	/*
904 	 * Do not change state if fs was real_dirty.
905 	 * Do not change state if fs already dirty(clear).
906 	 * Do not change any thing if mounted read only.
907 	 */
908 	if (sbi->volume.real_dirty || sb_rdonly(sbi->sb))
909 		return 0;
910 
911 	/* Check cached value. */
912 	if ((dirty == NTFS_DIRTY_CLEAR ? 0 : VOLUME_FLAG_DIRTY) ==
913 	    (sbi->volume.flags & VOLUME_FLAG_DIRTY))
914 		return 0;
915 
916 	ni = sbi->volume.ni;
917 	if (!ni)
918 		return -EINVAL;
919 
920 	mutex_lock_nested(&ni->ni_lock, NTFS_INODE_MUTEX_DIRTY);
921 
922 	attr = ni_find_attr(ni, NULL, NULL, ATTR_VOL_INFO, NULL, 0, NULL, &mi);
923 	if (!attr) {
924 		err = -EINVAL;
925 		goto out;
926 	}
927 
928 	info = resident_data_ex(attr, SIZEOF_ATTRIBUTE_VOLUME_INFO);
929 	if (!info) {
930 		err = -EINVAL;
931 		goto out;
932 	}
933 
934 	switch (dirty) {
935 	case NTFS_DIRTY_ERROR:
936 		ntfs_notice(sbi->sb, "Mark volume as dirty due to NTFS errors");
937 		sbi->volume.real_dirty = true;
938 		fallthrough;
939 	case NTFS_DIRTY_DIRTY:
940 		info->flags |= VOLUME_FLAG_DIRTY;
941 		break;
942 	case NTFS_DIRTY_CLEAR:
943 		info->flags &= ~VOLUME_FLAG_DIRTY;
944 		break;
945 	}
946 	/* Cache current volume flags. */
947 	sbi->volume.flags = info->flags;
948 	mi->dirty = true;
949 	err = 0;
950 
951 out:
952 	ni_unlock(ni);
953 	if (err)
954 		return err;
955 
956 	mark_inode_dirty(&ni->vfs_inode);
957 	/* verify(!ntfs_update_mftmirr()); */
958 
959 	/*
960 	 * If we used wait=1, sync_inode_metadata waits for the io for the
961 	 * inode to finish. It hangs when media is removed.
962 	 * So wait=0 is sent down to sync_inode_metadata
963 	 * and filemap_fdatawrite is used for the data blocks.
964 	 */
965 	err = sync_inode_metadata(&ni->vfs_inode, 0);
966 	if (!err)
967 		err = filemap_fdatawrite(ni->vfs_inode.i_mapping);
968 
969 	return err;
970 }
971 
972 /*
973  * security_hash - Calculates a hash of security descriptor.
974  */
975 static inline __le32 security_hash(const void *sd, size_t bytes)
976 {
977 	u32 hash = 0;
978 	const __le32 *ptr = sd;
979 
980 	bytes >>= 2;
981 	while (bytes--)
982 		hash = ((hash >> 0x1D) | (hash << 3)) + le32_to_cpu(*ptr++);
983 	return cpu_to_le32(hash);
984 }
985 
986 int ntfs_sb_read(struct super_block *sb, u64 lbo, size_t bytes, void *buffer)
987 {
988 	struct block_device *bdev = sb->s_bdev;
989 	u32 blocksize = sb->s_blocksize;
990 	u64 block = lbo >> sb->s_blocksize_bits;
991 	u32 off = lbo & (blocksize - 1);
992 	u32 op = blocksize - off;
993 
994 	for (; bytes; block += 1, off = 0, op = blocksize) {
995 		struct buffer_head *bh = __bread(bdev, block, blocksize);
996 
997 		if (!bh)
998 			return -EIO;
999 
1000 		if (op > bytes)
1001 			op = bytes;
1002 
1003 		memcpy(buffer, bh->b_data + off, op);
1004 
1005 		put_bh(bh);
1006 
1007 		bytes -= op;
1008 		buffer = Add2Ptr(buffer, op);
1009 	}
1010 
1011 	return 0;
1012 }
1013 
1014 int ntfs_sb_write(struct super_block *sb, u64 lbo, size_t bytes,
1015 		  const void *buf, int wait)
1016 {
1017 	u32 blocksize = sb->s_blocksize;
1018 	struct block_device *bdev = sb->s_bdev;
1019 	sector_t block = lbo >> sb->s_blocksize_bits;
1020 	u32 off = lbo & (blocksize - 1);
1021 	u32 op = blocksize - off;
1022 	struct buffer_head *bh;
1023 
1024 	if (!wait && (sb->s_flags & SB_SYNCHRONOUS))
1025 		wait = 1;
1026 
1027 	for (; bytes; block += 1, off = 0, op = blocksize) {
1028 		if (op > bytes)
1029 			op = bytes;
1030 
1031 		if (op < blocksize) {
1032 			bh = __bread(bdev, block, blocksize);
1033 			if (!bh) {
1034 				ntfs_err(sb, "failed to read block %llx",
1035 					 (u64)block);
1036 				return -EIO;
1037 			}
1038 		} else {
1039 			bh = __getblk(bdev, block, blocksize);
1040 			if (!bh)
1041 				return -ENOMEM;
1042 		}
1043 
1044 		if (buffer_locked(bh))
1045 			__wait_on_buffer(bh);
1046 
1047 		lock_buffer(bh);
1048 		if (buf) {
1049 			memcpy(bh->b_data + off, buf, op);
1050 			buf = Add2Ptr(buf, op);
1051 		} else {
1052 			memset(bh->b_data + off, -1, op);
1053 		}
1054 
1055 		set_buffer_uptodate(bh);
1056 		mark_buffer_dirty(bh);
1057 		unlock_buffer(bh);
1058 
1059 		if (wait) {
1060 			int err = sync_dirty_buffer(bh);
1061 
1062 			if (err) {
1063 				ntfs_err(
1064 					sb,
1065 					"failed to sync buffer at block %llx, error %d",
1066 					(u64)block, err);
1067 				put_bh(bh);
1068 				return err;
1069 			}
1070 		}
1071 
1072 		put_bh(bh);
1073 
1074 		bytes -= op;
1075 	}
1076 	return 0;
1077 }
1078 
1079 int ntfs_sb_write_run(struct ntfs_sb_info *sbi, const struct runs_tree *run,
1080 		      u64 vbo, const void *buf, size_t bytes, int sync)
1081 {
1082 	struct super_block *sb = sbi->sb;
1083 	u8 cluster_bits = sbi->cluster_bits;
1084 	u32 off = vbo & sbi->cluster_mask;
1085 	CLST lcn, clen, vcn = vbo >> cluster_bits, vcn_next;
1086 	u64 lbo, len;
1087 	size_t idx;
1088 
1089 	if (!run_lookup_entry(run, vcn, &lcn, &clen, &idx))
1090 		return -ENOENT;
1091 
1092 	if (lcn == SPARSE_LCN)
1093 		return -EINVAL;
1094 
1095 	lbo = ((u64)lcn << cluster_bits) + off;
1096 	len = ((u64)clen << cluster_bits) - off;
1097 
1098 	for (;;) {
1099 		u32 op = min_t(u64, len, bytes);
1100 		int err = ntfs_sb_write(sb, lbo, op, buf, sync);
1101 
1102 		if (err)
1103 			return err;
1104 
1105 		bytes -= op;
1106 		if (!bytes)
1107 			break;
1108 
1109 		vcn_next = vcn + clen;
1110 		if (!run_get_entry(run, ++idx, &vcn, &lcn, &clen) ||
1111 		    vcn != vcn_next)
1112 			return -ENOENT;
1113 
1114 		if (lcn == SPARSE_LCN)
1115 			return -EINVAL;
1116 
1117 		if (buf)
1118 			buf = Add2Ptr(buf, op);
1119 
1120 		lbo = ((u64)lcn << cluster_bits);
1121 		len = ((u64)clen << cluster_bits);
1122 	}
1123 
1124 	return 0;
1125 }
1126 
1127 struct buffer_head *ntfs_bread_run(struct ntfs_sb_info *sbi,
1128 				   const struct runs_tree *run, u64 vbo)
1129 {
1130 	struct super_block *sb = sbi->sb;
1131 	u8 cluster_bits = sbi->cluster_bits;
1132 	CLST lcn;
1133 	u64 lbo;
1134 
1135 	if (!run_lookup_entry(run, vbo >> cluster_bits, &lcn, NULL, NULL))
1136 		return ERR_PTR(-ENOENT);
1137 
1138 	lbo = ((u64)lcn << cluster_bits) + (vbo & sbi->cluster_mask);
1139 
1140 	return ntfs_bread(sb, lbo >> sb->s_blocksize_bits);
1141 }
1142 
1143 int ntfs_read_run_nb(struct ntfs_sb_info *sbi, const struct runs_tree *run,
1144 		     u64 vbo, void *buf, u32 bytes, struct ntfs_buffers *nb)
1145 {
1146 	int err;
1147 	struct super_block *sb = sbi->sb;
1148 	u32 blocksize = sb->s_blocksize;
1149 	u8 cluster_bits = sbi->cluster_bits;
1150 	u32 off = vbo & sbi->cluster_mask;
1151 	u32 nbh = 0;
1152 	CLST vcn_next, vcn = vbo >> cluster_bits;
1153 	CLST lcn, clen;
1154 	u64 lbo, len;
1155 	size_t idx;
1156 	struct buffer_head *bh;
1157 
1158 	if (!run) {
1159 		/* First reading of $Volume + $MFTMirr + $LogFile goes here. */
1160 		if (vbo > MFT_REC_VOL * sbi->record_size) {
1161 			err = -ENOENT;
1162 			goto out;
1163 		}
1164 
1165 		/* Use absolute boot's 'MFTCluster' to read record. */
1166 		lbo = vbo + sbi->mft.lbo;
1167 		len = sbi->record_size;
1168 	} else if (!run_lookup_entry(run, vcn, &lcn, &clen, &idx)) {
1169 		err = -ENOENT;
1170 		goto out;
1171 	} else {
1172 		if (lcn == SPARSE_LCN) {
1173 			err = -EINVAL;
1174 			goto out;
1175 		}
1176 
1177 		lbo = ((u64)lcn << cluster_bits) + off;
1178 		len = ((u64)clen << cluster_bits) - off;
1179 	}
1180 
1181 	off = lbo & (blocksize - 1);
1182 	if (nb) {
1183 		nb->off = off;
1184 		nb->bytes = bytes;
1185 	}
1186 
1187 	for (;;) {
1188 		u32 len32 = len >= bytes ? bytes : len;
1189 		sector_t block = lbo >> sb->s_blocksize_bits;
1190 
1191 		do {
1192 			u32 op = blocksize - off;
1193 
1194 			if (op > len32)
1195 				op = len32;
1196 
1197 			bh = ntfs_bread(sb, block);
1198 			if (!bh) {
1199 				err = -EIO;
1200 				goto out;
1201 			}
1202 
1203 			if (buf) {
1204 				memcpy(buf, bh->b_data + off, op);
1205 				buf = Add2Ptr(buf, op);
1206 			}
1207 
1208 			if (!nb) {
1209 				put_bh(bh);
1210 			} else if (nbh >= ARRAY_SIZE(nb->bh)) {
1211 				err = -EINVAL;
1212 				goto out;
1213 			} else {
1214 				nb->bh[nbh++] = bh;
1215 				nb->nbufs = nbh;
1216 			}
1217 
1218 			bytes -= op;
1219 			if (!bytes)
1220 				return 0;
1221 			len32 -= op;
1222 			block += 1;
1223 			off = 0;
1224 
1225 		} while (len32);
1226 
1227 		vcn_next = vcn + clen;
1228 		if (!run_get_entry(run, ++idx, &vcn, &lcn, &clen) ||
1229 		    vcn != vcn_next) {
1230 			err = -ENOENT;
1231 			goto out;
1232 		}
1233 
1234 		if (lcn == SPARSE_LCN) {
1235 			err = -EINVAL;
1236 			goto out;
1237 		}
1238 
1239 		lbo = ((u64)lcn << cluster_bits);
1240 		len = ((u64)clen << cluster_bits);
1241 	}
1242 
1243 out:
1244 	if (!nbh)
1245 		return err;
1246 
1247 	while (nbh) {
1248 		put_bh(nb->bh[--nbh]);
1249 		nb->bh[nbh] = NULL;
1250 	}
1251 
1252 	nb->nbufs = 0;
1253 	return err;
1254 }
1255 
1256 /*
1257  * ntfs_read_bh
1258  *
1259  * Return: < 0 if error, 0 if ok, -E_NTFS_FIXUP if need to update fixups.
1260  */
1261 int ntfs_read_bh(struct ntfs_sb_info *sbi, const struct runs_tree *run, u64 vbo,
1262 		 struct NTFS_RECORD_HEADER *rhdr, u32 bytes,
1263 		 struct ntfs_buffers *nb)
1264 {
1265 	int err = ntfs_read_run_nb(sbi, run, vbo, rhdr, bytes, nb);
1266 
1267 	if (err)
1268 		return err;
1269 	return ntfs_fix_post_read(rhdr, nb->bytes, true);
1270 }
1271 
1272 int ntfs_get_bh(struct ntfs_sb_info *sbi, const struct runs_tree *run, u64 vbo,
1273 		u32 bytes, struct ntfs_buffers *nb)
1274 {
1275 	int err = 0;
1276 	struct super_block *sb = sbi->sb;
1277 	u32 blocksize = sb->s_blocksize;
1278 	u8 cluster_bits = sbi->cluster_bits;
1279 	CLST vcn_next, vcn = vbo >> cluster_bits;
1280 	u32 off;
1281 	u32 nbh = 0;
1282 	CLST lcn, clen;
1283 	u64 lbo, len;
1284 	size_t idx;
1285 
1286 	nb->bytes = bytes;
1287 
1288 	if (!run_lookup_entry(run, vcn, &lcn, &clen, &idx)) {
1289 		err = -ENOENT;
1290 		goto out;
1291 	}
1292 
1293 	off = vbo & sbi->cluster_mask;
1294 	lbo = ((u64)lcn << cluster_bits) + off;
1295 	len = ((u64)clen << cluster_bits) - off;
1296 
1297 	nb->off = off = lbo & (blocksize - 1);
1298 
1299 	for (;;) {
1300 		u32 len32 = min_t(u64, len, bytes);
1301 		sector_t block = lbo >> sb->s_blocksize_bits;
1302 
1303 		do {
1304 			u32 op;
1305 			struct buffer_head *bh;
1306 
1307 			if (nbh >= ARRAY_SIZE(nb->bh)) {
1308 				err = -EINVAL;
1309 				goto out;
1310 			}
1311 
1312 			op = blocksize - off;
1313 			if (op > len32)
1314 				op = len32;
1315 
1316 			if (op == blocksize) {
1317 				bh = sb_getblk(sb, block);
1318 				if (!bh) {
1319 					err = -ENOMEM;
1320 					goto out;
1321 				}
1322 				if (buffer_locked(bh))
1323 					__wait_on_buffer(bh);
1324 				set_buffer_uptodate(bh);
1325 			} else {
1326 				bh = ntfs_bread(sb, block);
1327 				if (!bh) {
1328 					err = -EIO;
1329 					goto out;
1330 				}
1331 			}
1332 
1333 			nb->bh[nbh++] = bh;
1334 			bytes -= op;
1335 			if (!bytes) {
1336 				nb->nbufs = nbh;
1337 				return 0;
1338 			}
1339 
1340 			block += 1;
1341 			len32 -= op;
1342 			off = 0;
1343 		} while (len32);
1344 
1345 		vcn_next = vcn + clen;
1346 		if (!run_get_entry(run, ++idx, &vcn, &lcn, &clen) ||
1347 		    vcn != vcn_next) {
1348 			err = -ENOENT;
1349 			goto out;
1350 		}
1351 
1352 		lbo = ((u64)lcn << cluster_bits);
1353 		len = ((u64)clen << cluster_bits);
1354 	}
1355 
1356 out:
1357 	while (nbh) {
1358 		put_bh(nb->bh[--nbh]);
1359 		nb->bh[nbh] = NULL;
1360 	}
1361 
1362 	nb->nbufs = 0;
1363 
1364 	return err;
1365 }
1366 
1367 int ntfs_write_bh(struct ntfs_sb_info *sbi, struct NTFS_RECORD_HEADER *rhdr,
1368 		  struct ntfs_buffers *nb, int sync)
1369 {
1370 	int err = 0;
1371 	struct super_block *sb = sbi->sb;
1372 	u32 block_size = sb->s_blocksize;
1373 	u32 bytes = nb->bytes;
1374 	u32 off = nb->off;
1375 	u16 fo = le16_to_cpu(rhdr->fix_off);
1376 	u16 fn = le16_to_cpu(rhdr->fix_num);
1377 	u32 idx;
1378 	__le16 *fixup;
1379 	__le16 sample;
1380 
1381 	if ((fo & 1) || fo + fn * sizeof(short) > SECTOR_SIZE || !fn-- ||
1382 	    fn * SECTOR_SIZE > bytes) {
1383 		return -EINVAL;
1384 	}
1385 
1386 	for (idx = 0; bytes && idx < nb->nbufs; idx += 1, off = 0) {
1387 		u32 op = block_size - off;
1388 		char *bh_data;
1389 		struct buffer_head *bh = nb->bh[idx];
1390 		__le16 *ptr, *end_data;
1391 
1392 		if (op > bytes)
1393 			op = bytes;
1394 
1395 		if (buffer_locked(bh))
1396 			__wait_on_buffer(bh);
1397 
1398 		lock_buffer(nb->bh[idx]);
1399 
1400 		bh_data = bh->b_data + off;
1401 		end_data = Add2Ptr(bh_data, op);
1402 		memcpy(bh_data, rhdr, op);
1403 
1404 		if (!idx) {
1405 			u16 t16;
1406 
1407 			fixup = Add2Ptr(bh_data, fo);
1408 			sample = *fixup;
1409 			t16 = le16_to_cpu(sample);
1410 			if (t16 >= 0x7FFF) {
1411 				sample = *fixup = cpu_to_le16(1);
1412 			} else {
1413 				sample = cpu_to_le16(t16 + 1);
1414 				*fixup = sample;
1415 			}
1416 
1417 			*(__le16 *)Add2Ptr(rhdr, fo) = sample;
1418 		}
1419 
1420 		ptr = Add2Ptr(bh_data, SECTOR_SIZE - sizeof(short));
1421 
1422 		do {
1423 			*++fixup = *ptr;
1424 			*ptr = sample;
1425 			ptr += SECTOR_SIZE / sizeof(short);
1426 		} while (ptr < end_data);
1427 
1428 		set_buffer_uptodate(bh);
1429 		mark_buffer_dirty(bh);
1430 		unlock_buffer(bh);
1431 
1432 		if (sync) {
1433 			int err2 = sync_dirty_buffer(bh);
1434 
1435 			if (!err && err2)
1436 				err = err2;
1437 		}
1438 
1439 		bytes -= op;
1440 		rhdr = Add2Ptr(rhdr, op);
1441 	}
1442 
1443 	return err;
1444 }
1445 
1446 static inline struct bio *ntfs_alloc_bio(u32 nr_vecs)
1447 {
1448 	struct bio *bio = bio_alloc(GFP_NOFS | __GFP_HIGH, nr_vecs);
1449 
1450 	if (!bio && (current->flags & PF_MEMALLOC)) {
1451 		while (!bio && (nr_vecs /= 2))
1452 			bio = bio_alloc(GFP_NOFS | __GFP_HIGH, nr_vecs);
1453 	}
1454 	return bio;
1455 }
1456 
1457 /*
1458  * ntfs_bio_pages - Read/write pages from/to disk.
1459  */
1460 int ntfs_bio_pages(struct ntfs_sb_info *sbi, const struct runs_tree *run,
1461 		   struct page **pages, u32 nr_pages, u64 vbo, u32 bytes,
1462 		   u32 op)
1463 {
1464 	int err = 0;
1465 	struct bio *new, *bio = NULL;
1466 	struct super_block *sb = sbi->sb;
1467 	struct block_device *bdev = sb->s_bdev;
1468 	struct page *page;
1469 	u8 cluster_bits = sbi->cluster_bits;
1470 	CLST lcn, clen, vcn, vcn_next;
1471 	u32 add, off, page_idx;
1472 	u64 lbo, len;
1473 	size_t run_idx;
1474 	struct blk_plug plug;
1475 
1476 	if (!bytes)
1477 		return 0;
1478 
1479 	blk_start_plug(&plug);
1480 
1481 	/* Align vbo and bytes to be 512 bytes aligned. */
1482 	lbo = (vbo + bytes + 511) & ~511ull;
1483 	vbo = vbo & ~511ull;
1484 	bytes = lbo - vbo;
1485 
1486 	vcn = vbo >> cluster_bits;
1487 	if (!run_lookup_entry(run, vcn, &lcn, &clen, &run_idx)) {
1488 		err = -ENOENT;
1489 		goto out;
1490 	}
1491 	off = vbo & sbi->cluster_mask;
1492 	page_idx = 0;
1493 	page = pages[0];
1494 
1495 	for (;;) {
1496 		lbo = ((u64)lcn << cluster_bits) + off;
1497 		len = ((u64)clen << cluster_bits) - off;
1498 new_bio:
1499 		new = ntfs_alloc_bio(nr_pages - page_idx);
1500 		if (!new) {
1501 			err = -ENOMEM;
1502 			goto out;
1503 		}
1504 		if (bio) {
1505 			bio_chain(bio, new);
1506 			submit_bio(bio);
1507 		}
1508 		bio = new;
1509 		bio_set_dev(bio, bdev);
1510 		bio->bi_iter.bi_sector = lbo >> 9;
1511 		bio->bi_opf = op;
1512 
1513 		while (len) {
1514 			off = vbo & (PAGE_SIZE - 1);
1515 			add = off + len > PAGE_SIZE ? (PAGE_SIZE - off) : len;
1516 
1517 			if (bio_add_page(bio, page, add, off) < add)
1518 				goto new_bio;
1519 
1520 			if (bytes <= add)
1521 				goto out;
1522 			bytes -= add;
1523 			vbo += add;
1524 
1525 			if (add + off == PAGE_SIZE) {
1526 				page_idx += 1;
1527 				if (WARN_ON(page_idx >= nr_pages)) {
1528 					err = -EINVAL;
1529 					goto out;
1530 				}
1531 				page = pages[page_idx];
1532 			}
1533 
1534 			if (len <= add)
1535 				break;
1536 			len -= add;
1537 			lbo += add;
1538 		}
1539 
1540 		vcn_next = vcn + clen;
1541 		if (!run_get_entry(run, ++run_idx, &vcn, &lcn, &clen) ||
1542 		    vcn != vcn_next) {
1543 			err = -ENOENT;
1544 			goto out;
1545 		}
1546 		off = 0;
1547 	}
1548 out:
1549 	if (bio) {
1550 		if (!err)
1551 			err = submit_bio_wait(bio);
1552 		bio_put(bio);
1553 	}
1554 	blk_finish_plug(&plug);
1555 
1556 	return err;
1557 }
1558 
1559 /*
1560  * ntfs_bio_fill_1 - Helper for ntfs_loadlog_and_replay().
1561  *
1562  * Fill on-disk logfile range by (-1)
1563  * this means empty logfile.
1564  */
1565 int ntfs_bio_fill_1(struct ntfs_sb_info *sbi, const struct runs_tree *run)
1566 {
1567 	int err = 0;
1568 	struct super_block *sb = sbi->sb;
1569 	struct block_device *bdev = sb->s_bdev;
1570 	u8 cluster_bits = sbi->cluster_bits;
1571 	struct bio *new, *bio = NULL;
1572 	CLST lcn, clen;
1573 	u64 lbo, len;
1574 	size_t run_idx;
1575 	struct page *fill;
1576 	void *kaddr;
1577 	struct blk_plug plug;
1578 
1579 	fill = alloc_page(GFP_KERNEL);
1580 	if (!fill)
1581 		return -ENOMEM;
1582 
1583 	kaddr = kmap_atomic(fill);
1584 	memset(kaddr, -1, PAGE_SIZE);
1585 	kunmap_atomic(kaddr);
1586 	flush_dcache_page(fill);
1587 	lock_page(fill);
1588 
1589 	if (!run_lookup_entry(run, 0, &lcn, &clen, &run_idx)) {
1590 		err = -ENOENT;
1591 		goto out;
1592 	}
1593 
1594 	/*
1595 	 * TODO: Try blkdev_issue_write_same.
1596 	 */
1597 	blk_start_plug(&plug);
1598 	do {
1599 		lbo = (u64)lcn << cluster_bits;
1600 		len = (u64)clen << cluster_bits;
1601 new_bio:
1602 		new = ntfs_alloc_bio(BIO_MAX_VECS);
1603 		if (!new) {
1604 			err = -ENOMEM;
1605 			break;
1606 		}
1607 		if (bio) {
1608 			bio_chain(bio, new);
1609 			submit_bio(bio);
1610 		}
1611 		bio = new;
1612 		bio_set_dev(bio, bdev);
1613 		bio->bi_opf = REQ_OP_WRITE;
1614 		bio->bi_iter.bi_sector = lbo >> 9;
1615 
1616 		for (;;) {
1617 			u32 add = len > PAGE_SIZE ? PAGE_SIZE : len;
1618 
1619 			if (bio_add_page(bio, fill, add, 0) < add)
1620 				goto new_bio;
1621 
1622 			lbo += add;
1623 			if (len <= add)
1624 				break;
1625 			len -= add;
1626 		}
1627 	} while (run_get_entry(run, ++run_idx, NULL, &lcn, &clen));
1628 
1629 	if (bio) {
1630 		if (!err)
1631 			err = submit_bio_wait(bio);
1632 		bio_put(bio);
1633 	}
1634 	blk_finish_plug(&plug);
1635 out:
1636 	unlock_page(fill);
1637 	put_page(fill);
1638 
1639 	return err;
1640 }
1641 
1642 int ntfs_vbo_to_lbo(struct ntfs_sb_info *sbi, const struct runs_tree *run,
1643 		    u64 vbo, u64 *lbo, u64 *bytes)
1644 {
1645 	u32 off;
1646 	CLST lcn, len;
1647 	u8 cluster_bits = sbi->cluster_bits;
1648 
1649 	if (!run_lookup_entry(run, vbo >> cluster_bits, &lcn, &len, NULL))
1650 		return -ENOENT;
1651 
1652 	off = vbo & sbi->cluster_mask;
1653 	*lbo = lcn == SPARSE_LCN ? -1 : (((u64)lcn << cluster_bits) + off);
1654 	*bytes = ((u64)len << cluster_bits) - off;
1655 
1656 	return 0;
1657 }
1658 
1659 struct ntfs_inode *ntfs_new_inode(struct ntfs_sb_info *sbi, CLST rno, bool dir)
1660 {
1661 	int err = 0;
1662 	struct super_block *sb = sbi->sb;
1663 	struct inode *inode = new_inode(sb);
1664 	struct ntfs_inode *ni;
1665 
1666 	if (!inode)
1667 		return ERR_PTR(-ENOMEM);
1668 
1669 	ni = ntfs_i(inode);
1670 
1671 	err = mi_format_new(&ni->mi, sbi, rno, dir ? RECORD_FLAG_DIR : 0,
1672 			    false);
1673 	if (err)
1674 		goto out;
1675 
1676 	inode->i_ino = rno;
1677 	if (insert_inode_locked(inode) < 0) {
1678 		err = -EIO;
1679 		goto out;
1680 	}
1681 
1682 out:
1683 	if (err) {
1684 		iput(inode);
1685 		ni = ERR_PTR(err);
1686 	}
1687 	return ni;
1688 }
1689 
1690 /*
1691  * O:BAG:BAD:(A;OICI;FA;;;WD)
1692  * Owner S-1-5-32-544 (Administrators)
1693  * Group S-1-5-32-544 (Administrators)
1694  * ACE: allow S-1-1-0 (Everyone) with FILE_ALL_ACCESS
1695  */
1696 const u8 s_default_security[] __aligned(8) = {
1697 	0x01, 0x00, 0x04, 0x80, 0x30, 0x00, 0x00, 0x00, 0x40, 0x00, 0x00, 0x00,
1698 	0x00, 0x00, 0x00, 0x00, 0x14, 0x00, 0x00, 0x00, 0x02, 0x00, 0x1C, 0x00,
1699 	0x01, 0x00, 0x00, 0x00, 0x00, 0x03, 0x14, 0x00, 0xFF, 0x01, 0x1F, 0x00,
1700 	0x01, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00,
1701 	0x01, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x20, 0x00, 0x00, 0x00,
1702 	0x20, 0x02, 0x00, 0x00, 0x01, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05,
1703 	0x20, 0x00, 0x00, 0x00, 0x20, 0x02, 0x00, 0x00,
1704 };
1705 
1706 static_assert(sizeof(s_default_security) == 0x50);
1707 
1708 static inline u32 sid_length(const struct SID *sid)
1709 {
1710 	return struct_size(sid, SubAuthority, sid->SubAuthorityCount);
1711 }
1712 
1713 /*
1714  * is_acl_valid
1715  *
1716  * Thanks Mark Harmstone for idea.
1717  */
1718 static bool is_acl_valid(const struct ACL *acl, u32 len)
1719 {
1720 	const struct ACE_HEADER *ace;
1721 	u32 i;
1722 	u16 ace_count, ace_size;
1723 
1724 	if (acl->AclRevision != ACL_REVISION &&
1725 	    acl->AclRevision != ACL_REVISION_DS) {
1726 		/*
1727 		 * This value should be ACL_REVISION, unless the ACL contains an
1728 		 * object-specific ACE, in which case this value must be ACL_REVISION_DS.
1729 		 * All ACEs in an ACL must be at the same revision level.
1730 		 */
1731 		return false;
1732 	}
1733 
1734 	if (acl->Sbz1)
1735 		return false;
1736 
1737 	if (le16_to_cpu(acl->AclSize) > len)
1738 		return false;
1739 
1740 	if (acl->Sbz2)
1741 		return false;
1742 
1743 	len -= sizeof(struct ACL);
1744 	ace = (struct ACE_HEADER *)&acl[1];
1745 	ace_count = le16_to_cpu(acl->AceCount);
1746 
1747 	for (i = 0; i < ace_count; i++) {
1748 		if (len < sizeof(struct ACE_HEADER))
1749 			return false;
1750 
1751 		ace_size = le16_to_cpu(ace->AceSize);
1752 		if (len < ace_size)
1753 			return false;
1754 
1755 		len -= ace_size;
1756 		ace = Add2Ptr(ace, ace_size);
1757 	}
1758 
1759 	return true;
1760 }
1761 
1762 bool is_sd_valid(const struct SECURITY_DESCRIPTOR_RELATIVE *sd, u32 len)
1763 {
1764 	u32 sd_owner, sd_group, sd_sacl, sd_dacl;
1765 
1766 	if (len < sizeof(struct SECURITY_DESCRIPTOR_RELATIVE))
1767 		return false;
1768 
1769 	if (sd->Revision != 1)
1770 		return false;
1771 
1772 	if (sd->Sbz1)
1773 		return false;
1774 
1775 	if (!(sd->Control & SE_SELF_RELATIVE))
1776 		return false;
1777 
1778 	sd_owner = le32_to_cpu(sd->Owner);
1779 	if (sd_owner) {
1780 		const struct SID *owner = Add2Ptr(sd, sd_owner);
1781 
1782 		if (sd_owner + offsetof(struct SID, SubAuthority) > len)
1783 			return false;
1784 
1785 		if (owner->Revision != 1)
1786 			return false;
1787 
1788 		if (sd_owner + sid_length(owner) > len)
1789 			return false;
1790 	}
1791 
1792 	sd_group = le32_to_cpu(sd->Group);
1793 	if (sd_group) {
1794 		const struct SID *group = Add2Ptr(sd, sd_group);
1795 
1796 		if (sd_group + offsetof(struct SID, SubAuthority) > len)
1797 			return false;
1798 
1799 		if (group->Revision != 1)
1800 			return false;
1801 
1802 		if (sd_group + sid_length(group) > len)
1803 			return false;
1804 	}
1805 
1806 	sd_sacl = le32_to_cpu(sd->Sacl);
1807 	if (sd_sacl) {
1808 		const struct ACL *sacl = Add2Ptr(sd, sd_sacl);
1809 
1810 		if (sd_sacl + sizeof(struct ACL) > len)
1811 			return false;
1812 
1813 		if (!is_acl_valid(sacl, len - sd_sacl))
1814 			return false;
1815 	}
1816 
1817 	sd_dacl = le32_to_cpu(sd->Dacl);
1818 	if (sd_dacl) {
1819 		const struct ACL *dacl = Add2Ptr(sd, sd_dacl);
1820 
1821 		if (sd_dacl + sizeof(struct ACL) > len)
1822 			return false;
1823 
1824 		if (!is_acl_valid(dacl, len - sd_dacl))
1825 			return false;
1826 	}
1827 
1828 	return true;
1829 }
1830 
1831 /*
1832  * ntfs_security_init - Load and parse $Secure.
1833  */
1834 int ntfs_security_init(struct ntfs_sb_info *sbi)
1835 {
1836 	int err;
1837 	struct super_block *sb = sbi->sb;
1838 	struct inode *inode;
1839 	struct ntfs_inode *ni;
1840 	struct MFT_REF ref;
1841 	struct ATTRIB *attr;
1842 	struct ATTR_LIST_ENTRY *le;
1843 	u64 sds_size;
1844 	size_t off;
1845 	struct NTFS_DE *ne;
1846 	struct NTFS_DE_SII *sii_e;
1847 	struct ntfs_fnd *fnd_sii = NULL;
1848 	const struct INDEX_ROOT *root_sii;
1849 	const struct INDEX_ROOT *root_sdh;
1850 	struct ntfs_index *indx_sdh = &sbi->security.index_sdh;
1851 	struct ntfs_index *indx_sii = &sbi->security.index_sii;
1852 
1853 	ref.low = cpu_to_le32(MFT_REC_SECURE);
1854 	ref.high = 0;
1855 	ref.seq = cpu_to_le16(MFT_REC_SECURE);
1856 
1857 	inode = ntfs_iget5(sb, &ref, &NAME_SECURE);
1858 	if (IS_ERR(inode)) {
1859 		err = PTR_ERR(inode);
1860 		ntfs_err(sb, "Failed to load $Secure.");
1861 		inode = NULL;
1862 		goto out;
1863 	}
1864 
1865 	ni = ntfs_i(inode);
1866 
1867 	le = NULL;
1868 
1869 	attr = ni_find_attr(ni, NULL, &le, ATTR_ROOT, SDH_NAME,
1870 			    ARRAY_SIZE(SDH_NAME), NULL, NULL);
1871 	if (!attr) {
1872 		err = -EINVAL;
1873 		goto out;
1874 	}
1875 
1876 	root_sdh = resident_data(attr);
1877 	if (root_sdh->type != ATTR_ZERO ||
1878 	    root_sdh->rule != NTFS_COLLATION_TYPE_SECURITY_HASH) {
1879 		err = -EINVAL;
1880 		goto out;
1881 	}
1882 
1883 	err = indx_init(indx_sdh, sbi, attr, INDEX_MUTEX_SDH);
1884 	if (err)
1885 		goto out;
1886 
1887 	attr = ni_find_attr(ni, attr, &le, ATTR_ROOT, SII_NAME,
1888 			    ARRAY_SIZE(SII_NAME), NULL, NULL);
1889 	if (!attr) {
1890 		err = -EINVAL;
1891 		goto out;
1892 	}
1893 
1894 	root_sii = resident_data(attr);
1895 	if (root_sii->type != ATTR_ZERO ||
1896 	    root_sii->rule != NTFS_COLLATION_TYPE_UINT) {
1897 		err = -EINVAL;
1898 		goto out;
1899 	}
1900 
1901 	err = indx_init(indx_sii, sbi, attr, INDEX_MUTEX_SII);
1902 	if (err)
1903 		goto out;
1904 
1905 	fnd_sii = fnd_get();
1906 	if (!fnd_sii) {
1907 		err = -ENOMEM;
1908 		goto out;
1909 	}
1910 
1911 	sds_size = inode->i_size;
1912 
1913 	/* Find the last valid Id. */
1914 	sbi->security.next_id = SECURITY_ID_FIRST;
1915 	/* Always write new security at the end of bucket. */
1916 	sbi->security.next_off =
1917 		ALIGN(sds_size - SecurityDescriptorsBlockSize, 16);
1918 
1919 	off = 0;
1920 	ne = NULL;
1921 
1922 	for (;;) {
1923 		u32 next_id;
1924 
1925 		err = indx_find_raw(indx_sii, ni, root_sii, &ne, &off, fnd_sii);
1926 		if (err || !ne)
1927 			break;
1928 
1929 		sii_e = (struct NTFS_DE_SII *)ne;
1930 		if (le16_to_cpu(ne->view.data_size) < SIZEOF_SECURITY_HDR)
1931 			continue;
1932 
1933 		next_id = le32_to_cpu(sii_e->sec_id) + 1;
1934 		if (next_id >= sbi->security.next_id)
1935 			sbi->security.next_id = next_id;
1936 	}
1937 
1938 	sbi->security.ni = ni;
1939 	inode = NULL;
1940 out:
1941 	iput(inode);
1942 	fnd_put(fnd_sii);
1943 
1944 	return err;
1945 }
1946 
1947 /*
1948  * ntfs_get_security_by_id - Read security descriptor by id.
1949  */
1950 int ntfs_get_security_by_id(struct ntfs_sb_info *sbi, __le32 security_id,
1951 			    struct SECURITY_DESCRIPTOR_RELATIVE **sd,
1952 			    size_t *size)
1953 {
1954 	int err;
1955 	int diff;
1956 	struct ntfs_inode *ni = sbi->security.ni;
1957 	struct ntfs_index *indx = &sbi->security.index_sii;
1958 	void *p = NULL;
1959 	struct NTFS_DE_SII *sii_e;
1960 	struct ntfs_fnd *fnd_sii;
1961 	struct SECURITY_HDR d_security;
1962 	const struct INDEX_ROOT *root_sii;
1963 	u32 t32;
1964 
1965 	*sd = NULL;
1966 
1967 	mutex_lock_nested(&ni->ni_lock, NTFS_INODE_MUTEX_SECURITY);
1968 
1969 	fnd_sii = fnd_get();
1970 	if (!fnd_sii) {
1971 		err = -ENOMEM;
1972 		goto out;
1973 	}
1974 
1975 	root_sii = indx_get_root(indx, ni, NULL, NULL);
1976 	if (!root_sii) {
1977 		err = -EINVAL;
1978 		goto out;
1979 	}
1980 
1981 	/* Try to find this SECURITY descriptor in SII indexes. */
1982 	err = indx_find(indx, ni, root_sii, &security_id, sizeof(security_id),
1983 			NULL, &diff, (struct NTFS_DE **)&sii_e, fnd_sii);
1984 	if (err)
1985 		goto out;
1986 
1987 	if (diff)
1988 		goto out;
1989 
1990 	t32 = le32_to_cpu(sii_e->sec_hdr.size);
1991 	if (t32 < SIZEOF_SECURITY_HDR) {
1992 		err = -EINVAL;
1993 		goto out;
1994 	}
1995 
1996 	if (t32 > SIZEOF_SECURITY_HDR + 0x10000) {
1997 		/* Looks like too big security. 0x10000 - is arbitrary big number. */
1998 		err = -EFBIG;
1999 		goto out;
2000 	}
2001 
2002 	*size = t32 - SIZEOF_SECURITY_HDR;
2003 
2004 	p = kmalloc(*size, GFP_NOFS);
2005 	if (!p) {
2006 		err = -ENOMEM;
2007 		goto out;
2008 	}
2009 
2010 	err = ntfs_read_run_nb(sbi, &ni->file.run,
2011 			       le64_to_cpu(sii_e->sec_hdr.off), &d_security,
2012 			       sizeof(d_security), NULL);
2013 	if (err)
2014 		goto out;
2015 
2016 	if (memcmp(&d_security, &sii_e->sec_hdr, SIZEOF_SECURITY_HDR)) {
2017 		err = -EINVAL;
2018 		goto out;
2019 	}
2020 
2021 	err = ntfs_read_run_nb(sbi, &ni->file.run,
2022 			       le64_to_cpu(sii_e->sec_hdr.off) +
2023 				       SIZEOF_SECURITY_HDR,
2024 			       p, *size, NULL);
2025 	if (err)
2026 		goto out;
2027 
2028 	*sd = p;
2029 	p = NULL;
2030 
2031 out:
2032 	kfree(p);
2033 	fnd_put(fnd_sii);
2034 	ni_unlock(ni);
2035 
2036 	return err;
2037 }
2038 
2039 /*
2040  * ntfs_insert_security - Insert security descriptor into $Secure::SDS.
2041  *
2042  * SECURITY Descriptor Stream data is organized into chunks of 256K bytes
2043  * and it contains a mirror copy of each security descriptor.  When writing
2044  * to a security descriptor at location X, another copy will be written at
2045  * location (X+256K).
2046  * When writing a security descriptor that will cross the 256K boundary,
2047  * the pointer will be advanced by 256K to skip
2048  * over the mirror portion.
2049  */
2050 int ntfs_insert_security(struct ntfs_sb_info *sbi,
2051 			 const struct SECURITY_DESCRIPTOR_RELATIVE *sd,
2052 			 u32 size_sd, __le32 *security_id, bool *inserted)
2053 {
2054 	int err, diff;
2055 	struct ntfs_inode *ni = sbi->security.ni;
2056 	struct ntfs_index *indx_sdh = &sbi->security.index_sdh;
2057 	struct ntfs_index *indx_sii = &sbi->security.index_sii;
2058 	struct NTFS_DE_SDH *e;
2059 	struct NTFS_DE_SDH sdh_e;
2060 	struct NTFS_DE_SII sii_e;
2061 	struct SECURITY_HDR *d_security;
2062 	u32 new_sec_size = size_sd + SIZEOF_SECURITY_HDR;
2063 	u32 aligned_sec_size = ALIGN(new_sec_size, 16);
2064 	struct SECURITY_KEY hash_key;
2065 	struct ntfs_fnd *fnd_sdh = NULL;
2066 	const struct INDEX_ROOT *root_sdh;
2067 	const struct INDEX_ROOT *root_sii;
2068 	u64 mirr_off, new_sds_size;
2069 	u32 next, left;
2070 
2071 	static_assert((1 << Log2OfSecurityDescriptorsBlockSize) ==
2072 		      SecurityDescriptorsBlockSize);
2073 
2074 	hash_key.hash = security_hash(sd, size_sd);
2075 	hash_key.sec_id = SECURITY_ID_INVALID;
2076 
2077 	if (inserted)
2078 		*inserted = false;
2079 	*security_id = SECURITY_ID_INVALID;
2080 
2081 	/* Allocate a temporal buffer. */
2082 	d_security = kzalloc(aligned_sec_size, GFP_NOFS);
2083 	if (!d_security)
2084 		return -ENOMEM;
2085 
2086 	mutex_lock_nested(&ni->ni_lock, NTFS_INODE_MUTEX_SECURITY);
2087 
2088 	fnd_sdh = fnd_get();
2089 	if (!fnd_sdh) {
2090 		err = -ENOMEM;
2091 		goto out;
2092 	}
2093 
2094 	root_sdh = indx_get_root(indx_sdh, ni, NULL, NULL);
2095 	if (!root_sdh) {
2096 		err = -EINVAL;
2097 		goto out;
2098 	}
2099 
2100 	root_sii = indx_get_root(indx_sii, ni, NULL, NULL);
2101 	if (!root_sii) {
2102 		err = -EINVAL;
2103 		goto out;
2104 	}
2105 
2106 	/*
2107 	 * Check if such security already exists.
2108 	 * Use "SDH" and hash -> to get the offset in "SDS".
2109 	 */
2110 	err = indx_find(indx_sdh, ni, root_sdh, &hash_key, sizeof(hash_key),
2111 			&d_security->key.sec_id, &diff, (struct NTFS_DE **)&e,
2112 			fnd_sdh);
2113 	if (err)
2114 		goto out;
2115 
2116 	while (e) {
2117 		if (le32_to_cpu(e->sec_hdr.size) == new_sec_size) {
2118 			err = ntfs_read_run_nb(sbi, &ni->file.run,
2119 					       le64_to_cpu(e->sec_hdr.off),
2120 					       d_security, new_sec_size, NULL);
2121 			if (err)
2122 				goto out;
2123 
2124 			if (le32_to_cpu(d_security->size) == new_sec_size &&
2125 			    d_security->key.hash == hash_key.hash &&
2126 			    !memcmp(d_security + 1, sd, size_sd)) {
2127 				*security_id = d_security->key.sec_id;
2128 				/* Such security already exists. */
2129 				err = 0;
2130 				goto out;
2131 			}
2132 		}
2133 
2134 		err = indx_find_sort(indx_sdh, ni, root_sdh,
2135 				     (struct NTFS_DE **)&e, fnd_sdh);
2136 		if (err)
2137 			goto out;
2138 
2139 		if (!e || e->key.hash != hash_key.hash)
2140 			break;
2141 	}
2142 
2143 	/* Zero unused space. */
2144 	next = sbi->security.next_off & (SecurityDescriptorsBlockSize - 1);
2145 	left = SecurityDescriptorsBlockSize - next;
2146 
2147 	/* Zero gap until SecurityDescriptorsBlockSize. */
2148 	if (left < new_sec_size) {
2149 		/* Zero "left" bytes from sbi->security.next_off. */
2150 		sbi->security.next_off += SecurityDescriptorsBlockSize + left;
2151 	}
2152 
2153 	/* Zero tail of previous security. */
2154 	//used = ni->vfs_inode.i_size & (SecurityDescriptorsBlockSize - 1);
2155 
2156 	/*
2157 	 * Example:
2158 	 * 0x40438 == ni->vfs_inode.i_size
2159 	 * 0x00440 == sbi->security.next_off
2160 	 * need to zero [0x438-0x440)
2161 	 * if (next > used) {
2162 	 *  u32 tozero = next - used;
2163 	 *  zero "tozero" bytes from sbi->security.next_off - tozero
2164 	 */
2165 
2166 	/* Format new security descriptor. */
2167 	d_security->key.hash = hash_key.hash;
2168 	d_security->key.sec_id = cpu_to_le32(sbi->security.next_id);
2169 	d_security->off = cpu_to_le64(sbi->security.next_off);
2170 	d_security->size = cpu_to_le32(new_sec_size);
2171 	memcpy(d_security + 1, sd, size_sd);
2172 
2173 	/* Write main SDS bucket. */
2174 	err = ntfs_sb_write_run(sbi, &ni->file.run, sbi->security.next_off,
2175 				d_security, aligned_sec_size, 0);
2176 
2177 	if (err)
2178 		goto out;
2179 
2180 	mirr_off = sbi->security.next_off + SecurityDescriptorsBlockSize;
2181 	new_sds_size = mirr_off + aligned_sec_size;
2182 
2183 	if (new_sds_size > ni->vfs_inode.i_size) {
2184 		err = attr_set_size(ni, ATTR_DATA, SDS_NAME,
2185 				    ARRAY_SIZE(SDS_NAME), &ni->file.run,
2186 				    new_sds_size, &new_sds_size, false, NULL);
2187 		if (err)
2188 			goto out;
2189 	}
2190 
2191 	/* Write copy SDS bucket. */
2192 	err = ntfs_sb_write_run(sbi, &ni->file.run, mirr_off, d_security,
2193 				aligned_sec_size, 0);
2194 	if (err)
2195 		goto out;
2196 
2197 	/* Fill SII entry. */
2198 	sii_e.de.view.data_off =
2199 		cpu_to_le16(offsetof(struct NTFS_DE_SII, sec_hdr));
2200 	sii_e.de.view.data_size = cpu_to_le16(SIZEOF_SECURITY_HDR);
2201 	sii_e.de.view.res = 0;
2202 	sii_e.de.size = cpu_to_le16(SIZEOF_SII_DIRENTRY);
2203 	sii_e.de.key_size = cpu_to_le16(sizeof(d_security->key.sec_id));
2204 	sii_e.de.flags = 0;
2205 	sii_e.de.res = 0;
2206 	sii_e.sec_id = d_security->key.sec_id;
2207 	memcpy(&sii_e.sec_hdr, d_security, SIZEOF_SECURITY_HDR);
2208 
2209 	err = indx_insert_entry(indx_sii, ni, &sii_e.de, NULL, NULL, 0);
2210 	if (err)
2211 		goto out;
2212 
2213 	/* Fill SDH entry. */
2214 	sdh_e.de.view.data_off =
2215 		cpu_to_le16(offsetof(struct NTFS_DE_SDH, sec_hdr));
2216 	sdh_e.de.view.data_size = cpu_to_le16(SIZEOF_SECURITY_HDR);
2217 	sdh_e.de.view.res = 0;
2218 	sdh_e.de.size = cpu_to_le16(SIZEOF_SDH_DIRENTRY);
2219 	sdh_e.de.key_size = cpu_to_le16(sizeof(sdh_e.key));
2220 	sdh_e.de.flags = 0;
2221 	sdh_e.de.res = 0;
2222 	sdh_e.key.hash = d_security->key.hash;
2223 	sdh_e.key.sec_id = d_security->key.sec_id;
2224 	memcpy(&sdh_e.sec_hdr, d_security, SIZEOF_SECURITY_HDR);
2225 	sdh_e.magic[0] = cpu_to_le16('I');
2226 	sdh_e.magic[1] = cpu_to_le16('I');
2227 
2228 	fnd_clear(fnd_sdh);
2229 	err = indx_insert_entry(indx_sdh, ni, &sdh_e.de, (void *)(size_t)1,
2230 				fnd_sdh, 0);
2231 	if (err)
2232 		goto out;
2233 
2234 	*security_id = d_security->key.sec_id;
2235 	if (inserted)
2236 		*inserted = true;
2237 
2238 	/* Update Id and offset for next descriptor. */
2239 	sbi->security.next_id += 1;
2240 	sbi->security.next_off += aligned_sec_size;
2241 
2242 out:
2243 	fnd_put(fnd_sdh);
2244 	mark_inode_dirty(&ni->vfs_inode);
2245 	ni_unlock(ni);
2246 	kfree(d_security);
2247 
2248 	return err;
2249 }
2250 
2251 /*
2252  * ntfs_reparse_init - Load and parse $Extend/$Reparse.
2253  */
2254 int ntfs_reparse_init(struct ntfs_sb_info *sbi)
2255 {
2256 	int err;
2257 	struct ntfs_inode *ni = sbi->reparse.ni;
2258 	struct ntfs_index *indx = &sbi->reparse.index_r;
2259 	struct ATTRIB *attr;
2260 	struct ATTR_LIST_ENTRY *le;
2261 	const struct INDEX_ROOT *root_r;
2262 
2263 	if (!ni)
2264 		return 0;
2265 
2266 	le = NULL;
2267 	attr = ni_find_attr(ni, NULL, &le, ATTR_ROOT, SR_NAME,
2268 			    ARRAY_SIZE(SR_NAME), NULL, NULL);
2269 	if (!attr) {
2270 		err = -EINVAL;
2271 		goto out;
2272 	}
2273 
2274 	root_r = resident_data(attr);
2275 	if (root_r->type != ATTR_ZERO ||
2276 	    root_r->rule != NTFS_COLLATION_TYPE_UINTS) {
2277 		err = -EINVAL;
2278 		goto out;
2279 	}
2280 
2281 	err = indx_init(indx, sbi, attr, INDEX_MUTEX_SR);
2282 	if (err)
2283 		goto out;
2284 
2285 out:
2286 	return err;
2287 }
2288 
2289 /*
2290  * ntfs_objid_init - Load and parse $Extend/$ObjId.
2291  */
2292 int ntfs_objid_init(struct ntfs_sb_info *sbi)
2293 {
2294 	int err;
2295 	struct ntfs_inode *ni = sbi->objid.ni;
2296 	struct ntfs_index *indx = &sbi->objid.index_o;
2297 	struct ATTRIB *attr;
2298 	struct ATTR_LIST_ENTRY *le;
2299 	const struct INDEX_ROOT *root;
2300 
2301 	if (!ni)
2302 		return 0;
2303 
2304 	le = NULL;
2305 	attr = ni_find_attr(ni, NULL, &le, ATTR_ROOT, SO_NAME,
2306 			    ARRAY_SIZE(SO_NAME), NULL, NULL);
2307 	if (!attr) {
2308 		err = -EINVAL;
2309 		goto out;
2310 	}
2311 
2312 	root = resident_data(attr);
2313 	if (root->type != ATTR_ZERO ||
2314 	    root->rule != NTFS_COLLATION_TYPE_UINTS) {
2315 		err = -EINVAL;
2316 		goto out;
2317 	}
2318 
2319 	err = indx_init(indx, sbi, attr, INDEX_MUTEX_SO);
2320 	if (err)
2321 		goto out;
2322 
2323 out:
2324 	return err;
2325 }
2326 
2327 int ntfs_objid_remove(struct ntfs_sb_info *sbi, struct GUID *guid)
2328 {
2329 	int err;
2330 	struct ntfs_inode *ni = sbi->objid.ni;
2331 	struct ntfs_index *indx = &sbi->objid.index_o;
2332 
2333 	if (!ni)
2334 		return -EINVAL;
2335 
2336 	mutex_lock_nested(&ni->ni_lock, NTFS_INODE_MUTEX_OBJID);
2337 
2338 	err = indx_delete_entry(indx, ni, guid, sizeof(*guid), NULL);
2339 
2340 	mark_inode_dirty(&ni->vfs_inode);
2341 	ni_unlock(ni);
2342 
2343 	return err;
2344 }
2345 
2346 int ntfs_insert_reparse(struct ntfs_sb_info *sbi, __le32 rtag,
2347 			const struct MFT_REF *ref)
2348 {
2349 	int err;
2350 	struct ntfs_inode *ni = sbi->reparse.ni;
2351 	struct ntfs_index *indx = &sbi->reparse.index_r;
2352 	struct NTFS_DE_R re;
2353 
2354 	if (!ni)
2355 		return -EINVAL;
2356 
2357 	memset(&re, 0, sizeof(re));
2358 
2359 	re.de.view.data_off = cpu_to_le16(offsetof(struct NTFS_DE_R, zero));
2360 	re.de.size = cpu_to_le16(sizeof(struct NTFS_DE_R));
2361 	re.de.key_size = cpu_to_le16(sizeof(re.key));
2362 
2363 	re.key.ReparseTag = rtag;
2364 	memcpy(&re.key.ref, ref, sizeof(*ref));
2365 
2366 	mutex_lock_nested(&ni->ni_lock, NTFS_INODE_MUTEX_REPARSE);
2367 
2368 	err = indx_insert_entry(indx, ni, &re.de, NULL, NULL, 0);
2369 
2370 	mark_inode_dirty(&ni->vfs_inode);
2371 	ni_unlock(ni);
2372 
2373 	return err;
2374 }
2375 
2376 int ntfs_remove_reparse(struct ntfs_sb_info *sbi, __le32 rtag,
2377 			const struct MFT_REF *ref)
2378 {
2379 	int err, diff;
2380 	struct ntfs_inode *ni = sbi->reparse.ni;
2381 	struct ntfs_index *indx = &sbi->reparse.index_r;
2382 	struct ntfs_fnd *fnd = NULL;
2383 	struct REPARSE_KEY rkey;
2384 	struct NTFS_DE_R *re;
2385 	struct INDEX_ROOT *root_r;
2386 
2387 	if (!ni)
2388 		return -EINVAL;
2389 
2390 	rkey.ReparseTag = rtag;
2391 	rkey.ref = *ref;
2392 
2393 	mutex_lock_nested(&ni->ni_lock, NTFS_INODE_MUTEX_REPARSE);
2394 
2395 	if (rtag) {
2396 		err = indx_delete_entry(indx, ni, &rkey, sizeof(rkey), NULL);
2397 		goto out1;
2398 	}
2399 
2400 	fnd = fnd_get();
2401 	if (!fnd) {
2402 		err = -ENOMEM;
2403 		goto out1;
2404 	}
2405 
2406 	root_r = indx_get_root(indx, ni, NULL, NULL);
2407 	if (!root_r) {
2408 		err = -EINVAL;
2409 		goto out;
2410 	}
2411 
2412 	/* 1 - forces to ignore rkey.ReparseTag when comparing keys. */
2413 	err = indx_find(indx, ni, root_r, &rkey, sizeof(rkey), (void *)1, &diff,
2414 			(struct NTFS_DE **)&re, fnd);
2415 	if (err)
2416 		goto out;
2417 
2418 	if (memcmp(&re->key.ref, ref, sizeof(*ref))) {
2419 		/* Impossible. Looks like volume corrupt? */
2420 		goto out;
2421 	}
2422 
2423 	memcpy(&rkey, &re->key, sizeof(rkey));
2424 
2425 	fnd_put(fnd);
2426 	fnd = NULL;
2427 
2428 	err = indx_delete_entry(indx, ni, &rkey, sizeof(rkey), NULL);
2429 	if (err)
2430 		goto out;
2431 
2432 out:
2433 	fnd_put(fnd);
2434 
2435 out1:
2436 	mark_inode_dirty(&ni->vfs_inode);
2437 	ni_unlock(ni);
2438 
2439 	return err;
2440 }
2441 
2442 static inline void ntfs_unmap_and_discard(struct ntfs_sb_info *sbi, CLST lcn,
2443 					  CLST len)
2444 {
2445 	ntfs_unmap_meta(sbi->sb, lcn, len);
2446 	ntfs_discard(sbi, lcn, len);
2447 }
2448 
2449 void mark_as_free_ex(struct ntfs_sb_info *sbi, CLST lcn, CLST len, bool trim)
2450 {
2451 	CLST end, i;
2452 	struct wnd_bitmap *wnd = &sbi->used.bitmap;
2453 
2454 	down_write_nested(&wnd->rw_lock, BITMAP_MUTEX_CLUSTERS);
2455 	if (!wnd_is_used(wnd, lcn, len)) {
2456 		ntfs_set_state(sbi, NTFS_DIRTY_ERROR);
2457 
2458 		end = lcn + len;
2459 		len = 0;
2460 		for (i = lcn; i < end; i++) {
2461 			if (wnd_is_used(wnd, i, 1)) {
2462 				if (!len)
2463 					lcn = i;
2464 				len += 1;
2465 				continue;
2466 			}
2467 
2468 			if (!len)
2469 				continue;
2470 
2471 			if (trim)
2472 				ntfs_unmap_and_discard(sbi, lcn, len);
2473 
2474 			wnd_set_free(wnd, lcn, len);
2475 			len = 0;
2476 		}
2477 
2478 		if (!len)
2479 			goto out;
2480 	}
2481 
2482 	if (trim)
2483 		ntfs_unmap_and_discard(sbi, lcn, len);
2484 	wnd_set_free(wnd, lcn, len);
2485 
2486 out:
2487 	up_write(&wnd->rw_lock);
2488 }
2489 
2490 /*
2491  * run_deallocate - Deallocate clusters.
2492  */
2493 int run_deallocate(struct ntfs_sb_info *sbi, struct runs_tree *run, bool trim)
2494 {
2495 	CLST lcn, len;
2496 	size_t idx = 0;
2497 
2498 	while (run_get_entry(run, idx++, NULL, &lcn, &len)) {
2499 		if (lcn == SPARSE_LCN)
2500 			continue;
2501 
2502 		mark_as_free_ex(sbi, lcn, len, trim);
2503 	}
2504 
2505 	return 0;
2506 }
2507