1// SPDX-License-Identifier: GPL-2.0
2/*
3 *
4 * Copyright (C) 2019-2021 Paragon Software GmbH, All rights reserved.
5 *
6 *
7 *                 terminology
8 *
9 * cluster - allocation unit     - 512,1K,2K,4K,...,2M
10 * vcn - virtual cluster number  - Offset inside the file in clusters.
11 * vbo - virtual byte offset     - Offset inside the file in bytes.
12 * lcn - logical cluster number  - 0 based cluster in clusters heap.
13 * lbo - logical byte offset     - Absolute position inside volume.
14 * run - maps VCN to LCN         - Stored in attributes in packed form.
15 * attr - attribute segment      - std/name/data etc records inside MFT.
16 * mi  - MFT inode               - One MFT record(usually 1024 bytes or 4K), consists of attributes.
17 * ni  - NTFS inode              - Extends linux inode. consists of one or more mft inodes.
18 * index - unit inside directory - 2K, 4K, <=page size, does not depend on cluster size.
19 *
20 * WSL - Windows Subsystem for Linux
21 * https://docs.microsoft.com/en-us/windows/wsl/file-permissions
22 * It stores uid/gid/mode/dev in xattr
23 *
24 * ntfs allows up to 2^64 clusters per volume.
25 * It means you should use 64 bits lcn to operate with ntfs.
26 * Implementation of ntfs.sys uses only 32 bits lcn.
27 * Default ntfs3 uses 32 bits lcn too.
28 * ntfs3 built with CONFIG_NTFS3_64BIT_CLUSTER (ntfs3_64) uses 64 bits per lcn.
29 *
30 *
31 *     ntfs limits, cluster size is 4K (2^12)
32 * -----------------------------------------------------------------------------
33 * | Volume size   | Clusters | ntfs.sys | ntfs3  | ntfs3_64 | mkntfs | chkdsk |
34 * -----------------------------------------------------------------------------
35 * | < 16T, 2^44   |  < 2^32  |  yes     |  yes   |   yes    |  yes   |  yes   |
36 * | > 16T, 2^44   |  > 2^32  |  no      |  no    |   yes    |  yes   |  yes   |
37 * ----------------------------------------------------------|------------------
38 *
39 * To mount large volumes as ntfs one should use large cluster size (up to 2M)
40 * The maximum volume size in this case is 2^32 * 2^21 = 2^53 = 8P
41 *
42 *     ntfs limits, cluster size is 2M (2^21)
43 * -----------------------------------------------------------------------------
44 * | < 8P, 2^53    |  < 2^32  |  yes     |  yes   |   yes    |  yes   |  yes   |
45 * | > 8P, 2^53    |  > 2^32  |  no      |  no    |   yes    |  yes   |  yes   |
46 * ----------------------------------------------------------|------------------
47 *
48 */
49
50#include <linux/blkdev.h>
51#include <linux/buffer_head.h>
52#include <linux/exportfs.h>
53#include <linux/fs.h>
54#include <linux/fs_context.h>
55#include <linux/fs_parser.h>
56#include <linux/log2.h>
57#include <linux/minmax.h>
58#include <linux/module.h>
59#include <linux/nls.h>
60#include <linux/proc_fs.h>
61#include <linux/seq_file.h>
62#include <linux/statfs.h>
63
64#include "debug.h"
65#include "ntfs.h"
66#include "ntfs_fs.h"
67#ifdef CONFIG_NTFS3_LZX_XPRESS
68#include "lib/lib.h"
69#endif
70
71#ifdef CONFIG_PRINTK
72/*
73 * ntfs_printk - Trace warnings/notices/errors.
74 *
75 * Thanks Joe Perches <joe@perches.com> for implementation
76 */
77void ntfs_printk(const struct super_block *sb, const char *fmt, ...)
78{
79	struct va_format vaf;
80	va_list args;
81	int level;
82	struct ntfs_sb_info *sbi = sb->s_fs_info;
83
84	/* Should we use different ratelimits for warnings/notices/errors? */
85	if (!___ratelimit(&sbi->msg_ratelimit, "ntfs3"))
86		return;
87
88	va_start(args, fmt);
89
90	level = printk_get_level(fmt);
91	vaf.fmt = printk_skip_level(fmt);
92	vaf.va = &args;
93	printk("%c%cntfs3: %s: %pV\n", KERN_SOH_ASCII, level, sb->s_id, &vaf);
94
95	va_end(args);
96}
97
98static char s_name_buf[512];
99static atomic_t s_name_buf_cnt = ATOMIC_INIT(1); // 1 means 'free s_name_buf'.
100
101/*
102 * ntfs_inode_printk
103 *
104 * Print warnings/notices/errors about inode using name or inode number.
105 */
106void ntfs_inode_printk(struct inode *inode, const char *fmt, ...)
107{
108	struct super_block *sb = inode->i_sb;
109	struct ntfs_sb_info *sbi = sb->s_fs_info;
110	char *name;
111	va_list args;
112	struct va_format vaf;
113	int level;
114
115	if (!___ratelimit(&sbi->msg_ratelimit, "ntfs3"))
116		return;
117
118	/* Use static allocated buffer, if possible. */
119	name = atomic_dec_and_test(&s_name_buf_cnt) ?
120		       s_name_buf :
121		       kmalloc(sizeof(s_name_buf), GFP_NOFS);
122
123	if (name) {
124		struct dentry *de = d_find_alias(inode);
125
126		if (de) {
127			spin_lock(&de->d_lock);
128			snprintf(name, sizeof(s_name_buf), " \"%s\"",
129				 de->d_name.name);
130			spin_unlock(&de->d_lock);
131		} else {
132			name[0] = 0;
133		}
134		dput(de); /* Cocci warns if placed in branch "if (de)" */
135	}
136
137	va_start(args, fmt);
138
139	level = printk_get_level(fmt);
140	vaf.fmt = printk_skip_level(fmt);
141	vaf.va = &args;
142
143	printk("%c%cntfs3: %s: ino=%lx,%s %pV\n", KERN_SOH_ASCII, level,
144	       sb->s_id, inode->i_ino, name ? name : "", &vaf);
145
146	va_end(args);
147
148	atomic_inc(&s_name_buf_cnt);
149	if (name != s_name_buf)
150		kfree(name);
151}
152#endif
153
154/*
155 * Shared memory struct.
156 *
157 * On-disk ntfs's upcase table is created by ntfs formatter.
158 * 'upcase' table is 128K bytes of memory.
159 * We should read it into memory when mounting.
160 * Several ntfs volumes likely use the same 'upcase' table.
161 * It is good idea to share in-memory 'upcase' table between different volumes.
162 * Unfortunately winxp/vista/win7 use different upcase tables.
163 */
164static DEFINE_SPINLOCK(s_shared_lock);
165
166static struct {
167	void *ptr;
168	u32 len;
169	int cnt;
170} s_shared[8];
171
172/*
173 * ntfs_set_shared
174 *
175 * Return:
176 * * @ptr - If pointer was saved in shared memory.
177 * * NULL - If pointer was not shared.
178 */
179void *ntfs_set_shared(void *ptr, u32 bytes)
180{
181	void *ret = NULL;
182	int i, j = -1;
183
184	spin_lock(&s_shared_lock);
185	for (i = 0; i < ARRAY_SIZE(s_shared); i++) {
186		if (!s_shared[i].cnt) {
187			j = i;
188		} else if (bytes == s_shared[i].len &&
189			   !memcmp(s_shared[i].ptr, ptr, bytes)) {
190			s_shared[i].cnt += 1;
191			ret = s_shared[i].ptr;
192			break;
193		}
194	}
195
196	if (!ret && j != -1) {
197		s_shared[j].ptr = ptr;
198		s_shared[j].len = bytes;
199		s_shared[j].cnt = 1;
200		ret = ptr;
201	}
202	spin_unlock(&s_shared_lock);
203
204	return ret;
205}
206
207/*
208 * ntfs_put_shared
209 *
210 * Return:
211 * * @ptr - If pointer is not shared anymore.
212 * * NULL - If pointer is still shared.
213 */
214void *ntfs_put_shared(void *ptr)
215{
216	void *ret = ptr;
217	int i;
218
219	spin_lock(&s_shared_lock);
220	for (i = 0; i < ARRAY_SIZE(s_shared); i++) {
221		if (s_shared[i].cnt && s_shared[i].ptr == ptr) {
222			if (--s_shared[i].cnt)
223				ret = NULL;
224			break;
225		}
226	}
227	spin_unlock(&s_shared_lock);
228
229	return ret;
230}
231
232static inline void put_mount_options(struct ntfs_mount_options *options)
233{
234	kfree(options->nls_name);
235	unload_nls(options->nls);
236	kfree(options);
237}
238
239enum Opt {
240	Opt_uid,
241	Opt_gid,
242	Opt_umask,
243	Opt_dmask,
244	Opt_fmask,
245	Opt_immutable,
246	Opt_discard,
247	Opt_force,
248	Opt_sparse,
249	Opt_nohidden,
250	Opt_hide_dot_files,
251	Opt_windows_names,
252	Opt_showmeta,
253	Opt_acl,
254	Opt_iocharset,
255	Opt_prealloc,
256	Opt_nocase,
257	Opt_err,
258};
259
260// clang-format off
261static const struct fs_parameter_spec ntfs_fs_parameters[] = {
262	fsparam_u32("uid",			Opt_uid),
263	fsparam_u32("gid",			Opt_gid),
264	fsparam_u32oct("umask",			Opt_umask),
265	fsparam_u32oct("dmask",			Opt_dmask),
266	fsparam_u32oct("fmask",			Opt_fmask),
267	fsparam_flag_no("sys_immutable",	Opt_immutable),
268	fsparam_flag_no("discard",		Opt_discard),
269	fsparam_flag_no("force",		Opt_force),
270	fsparam_flag_no("sparse",		Opt_sparse),
271	fsparam_flag_no("hidden",		Opt_nohidden),
272	fsparam_flag_no("hide_dot_files",	Opt_hide_dot_files),
273	fsparam_flag_no("windows_names",	Opt_windows_names),
274	fsparam_flag_no("showmeta",		Opt_showmeta),
275	fsparam_flag_no("acl",			Opt_acl),
276	fsparam_string("iocharset",		Opt_iocharset),
277	fsparam_flag_no("prealloc",		Opt_prealloc),
278	fsparam_flag_no("nocase",		Opt_nocase),
279	{}
280};
281// clang-format on
282
283/*
284 * Load nls table or if @nls is utf8 then return NULL.
285 *
286 * It is good idea to use here "const char *nls".
287 * But load_nls accepts "char*".
288 */
289static struct nls_table *ntfs_load_nls(char *nls)
290{
291	struct nls_table *ret;
292
293	if (!nls)
294		nls = CONFIG_NLS_DEFAULT;
295
296	if (strcmp(nls, "utf8") == 0)
297		return NULL;
298
299	if (strcmp(nls, CONFIG_NLS_DEFAULT) == 0)
300		return load_nls_default();
301
302	ret = load_nls(nls);
303	if (ret)
304		return ret;
305
306	return ERR_PTR(-EINVAL);
307}
308
309static int ntfs_fs_parse_param(struct fs_context *fc,
310			       struct fs_parameter *param)
311{
312	struct ntfs_mount_options *opts = fc->fs_private;
313	struct fs_parse_result result;
314	int opt;
315
316	opt = fs_parse(fc, ntfs_fs_parameters, param, &result);
317	if (opt < 0)
318		return opt;
319
320	switch (opt) {
321	case Opt_uid:
322		opts->fs_uid = make_kuid(current_user_ns(), result.uint_32);
323		if (!uid_valid(opts->fs_uid))
324			return invalf(fc, "ntfs3: Invalid value for uid.");
325		break;
326	case Opt_gid:
327		opts->fs_gid = make_kgid(current_user_ns(), result.uint_32);
328		if (!gid_valid(opts->fs_gid))
329			return invalf(fc, "ntfs3: Invalid value for gid.");
330		break;
331	case Opt_umask:
332		if (result.uint_32 & ~07777)
333			return invalf(fc, "ntfs3: Invalid value for umask.");
334		opts->fs_fmask_inv = ~result.uint_32;
335		opts->fs_dmask_inv = ~result.uint_32;
336		opts->fmask = 1;
337		opts->dmask = 1;
338		break;
339	case Opt_dmask:
340		if (result.uint_32 & ~07777)
341			return invalf(fc, "ntfs3: Invalid value for dmask.");
342		opts->fs_dmask_inv = ~result.uint_32;
343		opts->dmask = 1;
344		break;
345	case Opt_fmask:
346		if (result.uint_32 & ~07777)
347			return invalf(fc, "ntfs3: Invalid value for fmask.");
348		opts->fs_fmask_inv = ~result.uint_32;
349		opts->fmask = 1;
350		break;
351	case Opt_immutable:
352		opts->sys_immutable = result.negated ? 0 : 1;
353		break;
354	case Opt_discard:
355		opts->discard = result.negated ? 0 : 1;
356		break;
357	case Opt_force:
358		opts->force = result.negated ? 0 : 1;
359		break;
360	case Opt_sparse:
361		opts->sparse = result.negated ? 0 : 1;
362		break;
363	case Opt_nohidden:
364		opts->nohidden = result.negated ? 1 : 0;
365		break;
366	case Opt_hide_dot_files:
367		opts->hide_dot_files = result.negated ? 0 : 1;
368		break;
369	case Opt_windows_names:
370		opts->windows_names = result.negated ? 0 : 1;
371		break;
372	case Opt_showmeta:
373		opts->showmeta = result.negated ? 0 : 1;
374		break;
375	case Opt_acl:
376		if (!result.negated)
377#ifdef CONFIG_NTFS3_FS_POSIX_ACL
378			fc->sb_flags |= SB_POSIXACL;
379#else
380			return invalf(
381				fc, "ntfs3: Support for ACL not compiled in!");
382#endif
383		else
384			fc->sb_flags &= ~SB_POSIXACL;
385		break;
386	case Opt_iocharset:
387		kfree(opts->nls_name);
388		opts->nls_name = param->string;
389		param->string = NULL;
390		break;
391	case Opt_prealloc:
392		opts->prealloc = result.negated ? 0 : 1;
393		break;
394	case Opt_nocase:
395		opts->nocase = result.negated ? 1 : 0;
396		break;
397	default:
398		/* Should not be here unless we forget add case. */
399		return -EINVAL;
400	}
401	return 0;
402}
403
404static int ntfs_fs_reconfigure(struct fs_context *fc)
405{
406	struct super_block *sb = fc->root->d_sb;
407	struct ntfs_sb_info *sbi = sb->s_fs_info;
408	struct ntfs_mount_options *new_opts = fc->fs_private;
409	int ro_rw;
410
411	ro_rw = sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY);
412	if (ro_rw && (sbi->flags & NTFS_FLAGS_NEED_REPLAY)) {
413		errorf(fc,
414		       "ntfs3: Couldn't remount rw because journal is not replayed. Please umount/remount instead\n");
415		return -EINVAL;
416	}
417
418	new_opts->nls = ntfs_load_nls(new_opts->nls_name);
419	if (IS_ERR(new_opts->nls)) {
420		new_opts->nls = NULL;
421		errorf(fc, "ntfs3: Cannot load iocharset %s",
422		       new_opts->nls_name);
423		return -EINVAL;
424	}
425	if (new_opts->nls != sbi->options->nls)
426		return invalf(
427			fc,
428			"ntfs3: Cannot use different iocharset when remounting!");
429
430	sync_filesystem(sb);
431
432	if (ro_rw && (sbi->volume.flags & VOLUME_FLAG_DIRTY) &&
433	    !new_opts->force) {
434		errorf(fc,
435		       "ntfs3: Volume is dirty and \"force\" flag is not set!");
436		return -EINVAL;
437	}
438
439	swap(sbi->options, fc->fs_private);
440
441	return 0;
442}
443
444#ifdef CONFIG_PROC_FS
445static struct proc_dir_entry *proc_info_root;
446
447/*
448 * ntfs3_volinfo:
449 *
450 * The content of /proc/fs/ntfs3/<dev>/volinfo
451 *
452 * ntfs3.1
453 * cluster size
454 * number of clusters
455 * total number of mft records
456 * number of used mft records ~= number of files + folders
457 * real state of ntfs "dirty"/"clean"
458 * current state of ntfs "dirty"/"clean"
459*/
460static int ntfs3_volinfo(struct seq_file *m, void *o)
461{
462	struct super_block *sb = m->private;
463	struct ntfs_sb_info *sbi = sb->s_fs_info;
464
465	seq_printf(m, "ntfs%d.%d\n%u\n%zu\n\%zu\n%zu\n%s\n%s\n",
466		   sbi->volume.major_ver, sbi->volume.minor_ver,
467		   sbi->cluster_size, sbi->used.bitmap.nbits,
468		   sbi->mft.bitmap.nbits,
469		   sbi->mft.bitmap.nbits - wnd_zeroes(&sbi->mft.bitmap),
470		   sbi->volume.real_dirty ? "dirty" : "clean",
471		   (sbi->volume.flags & VOLUME_FLAG_DIRTY) ? "dirty" : "clean");
472
473	return 0;
474}
475
476static int ntfs3_volinfo_open(struct inode *inode, struct file *file)
477{
478	return single_open(file, ntfs3_volinfo, pde_data(inode));
479}
480
481/* read /proc/fs/ntfs3/<dev>/label */
482static int ntfs3_label_show(struct seq_file *m, void *o)
483{
484	struct super_block *sb = m->private;
485	struct ntfs_sb_info *sbi = sb->s_fs_info;
486
487	seq_printf(m, "%s\n", sbi->volume.label);
488
489	return 0;
490}
491
492/* write /proc/fs/ntfs3/<dev>/label */
493static ssize_t ntfs3_label_write(struct file *file, const char __user *buffer,
494				 size_t count, loff_t *ppos)
495{
496	int err;
497	struct super_block *sb = pde_data(file_inode(file));
498	ssize_t ret = count;
499	u8 *label;
500
501	if (sb_rdonly(sb))
502		return -EROFS;
503
504	label = kmalloc(count, GFP_NOFS);
505
506	if (!label)
507		return -ENOMEM;
508
509	if (copy_from_user(label, buffer, ret)) {
510		ret = -EFAULT;
511		goto out;
512	}
513	while (ret > 0 && label[ret - 1] == '\n')
514		ret -= 1;
515
516	err = ntfs_set_label(sb->s_fs_info, label, ret);
517
518	if (err < 0) {
519		ntfs_err(sb, "failed (%d) to write label", err);
520		ret = err;
521		goto out;
522	}
523
524	*ppos += count;
525	ret = count;
526out:
527	kfree(label);
528	return ret;
529}
530
531static int ntfs3_label_open(struct inode *inode, struct file *file)
532{
533	return single_open(file, ntfs3_label_show, pde_data(inode));
534}
535
536static const struct proc_ops ntfs3_volinfo_fops = {
537	.proc_read = seq_read,
538	.proc_lseek = seq_lseek,
539	.proc_release = single_release,
540	.proc_open = ntfs3_volinfo_open,
541};
542
543static const struct proc_ops ntfs3_label_fops = {
544	.proc_read = seq_read,
545	.proc_lseek = seq_lseek,
546	.proc_release = single_release,
547	.proc_open = ntfs3_label_open,
548	.proc_write = ntfs3_label_write,
549};
550
551#endif
552
553static struct kmem_cache *ntfs_inode_cachep;
554
555static struct inode *ntfs_alloc_inode(struct super_block *sb)
556{
557	struct ntfs_inode *ni = alloc_inode_sb(sb, ntfs_inode_cachep, GFP_NOFS);
558
559	if (!ni)
560		return NULL;
561
562	memset(ni, 0, offsetof(struct ntfs_inode, vfs_inode));
563	mutex_init(&ni->ni_lock);
564	return &ni->vfs_inode;
565}
566
567static void ntfs_free_inode(struct inode *inode)
568{
569	struct ntfs_inode *ni = ntfs_i(inode);
570
571	mutex_destroy(&ni->ni_lock);
572	kmem_cache_free(ntfs_inode_cachep, ni);
573}
574
575static void init_once(void *foo)
576{
577	struct ntfs_inode *ni = foo;
578
579	inode_init_once(&ni->vfs_inode);
580}
581
582/*
583 * Noinline to reduce binary size.
584 */
585static noinline void ntfs3_put_sbi(struct ntfs_sb_info *sbi)
586{
587	wnd_close(&sbi->mft.bitmap);
588	wnd_close(&sbi->used.bitmap);
589
590	if (sbi->mft.ni) {
591		iput(&sbi->mft.ni->vfs_inode);
592		sbi->mft.ni = NULL;
593	}
594
595	if (sbi->security.ni) {
596		iput(&sbi->security.ni->vfs_inode);
597		sbi->security.ni = NULL;
598	}
599
600	if (sbi->reparse.ni) {
601		iput(&sbi->reparse.ni->vfs_inode);
602		sbi->reparse.ni = NULL;
603	}
604
605	if (sbi->objid.ni) {
606		iput(&sbi->objid.ni->vfs_inode);
607		sbi->objid.ni = NULL;
608	}
609
610	if (sbi->volume.ni) {
611		iput(&sbi->volume.ni->vfs_inode);
612		sbi->volume.ni = NULL;
613	}
614
615	ntfs_update_mftmirr(sbi, 0);
616
617	indx_clear(&sbi->security.index_sii);
618	indx_clear(&sbi->security.index_sdh);
619	indx_clear(&sbi->reparse.index_r);
620	indx_clear(&sbi->objid.index_o);
621}
622
623static void ntfs3_free_sbi(struct ntfs_sb_info *sbi)
624{
625	kfree(sbi->new_rec);
626	kvfree(ntfs_put_shared(sbi->upcase));
627	kvfree(sbi->def_table);
628	kfree(sbi->compress.lznt);
629#ifdef CONFIG_NTFS3_LZX_XPRESS
630	xpress_free_decompressor(sbi->compress.xpress);
631	lzx_free_decompressor(sbi->compress.lzx);
632#endif
633	kfree(sbi);
634}
635
636static void ntfs_put_super(struct super_block *sb)
637{
638	struct ntfs_sb_info *sbi = sb->s_fs_info;
639
640#ifdef CONFIG_PROC_FS
641	// Remove /proc/fs/ntfs3/..
642	if (sbi->procdir) {
643		remove_proc_entry("label", sbi->procdir);
644		remove_proc_entry("volinfo", sbi->procdir);
645		remove_proc_entry(sb->s_id, proc_info_root);
646		sbi->procdir = NULL;
647	}
648#endif
649
650	/* Mark rw ntfs as clear, if possible. */
651	ntfs_set_state(sbi, NTFS_DIRTY_CLEAR);
652	ntfs3_put_sbi(sbi);
653}
654
655static int ntfs_statfs(struct dentry *dentry, struct kstatfs *buf)
656{
657	struct super_block *sb = dentry->d_sb;
658	struct ntfs_sb_info *sbi = sb->s_fs_info;
659	struct wnd_bitmap *wnd = &sbi->used.bitmap;
660
661	buf->f_type = sb->s_magic;
662	buf->f_bsize = sbi->cluster_size;
663	buf->f_blocks = wnd->nbits;
664
665	buf->f_bfree = buf->f_bavail = wnd_zeroes(wnd);
666	buf->f_fsid.val[0] = sbi->volume.ser_num;
667	buf->f_fsid.val[1] = (sbi->volume.ser_num >> 32);
668	buf->f_namelen = NTFS_NAME_LEN;
669
670	return 0;
671}
672
673static int ntfs_show_options(struct seq_file *m, struct dentry *root)
674{
675	struct super_block *sb = root->d_sb;
676	struct ntfs_sb_info *sbi = sb->s_fs_info;
677	struct ntfs_mount_options *opts = sbi->options;
678	struct user_namespace *user_ns = seq_user_ns(m);
679
680	seq_printf(m, ",uid=%u", from_kuid_munged(user_ns, opts->fs_uid));
681	seq_printf(m, ",gid=%u", from_kgid_munged(user_ns, opts->fs_gid));
682	if (opts->dmask)
683		seq_printf(m, ",dmask=%04o", opts->fs_dmask_inv ^ 0xffff);
684	if (opts->fmask)
685		seq_printf(m, ",fmask=%04o", opts->fs_fmask_inv ^ 0xffff);
686	if (opts->sys_immutable)
687		seq_puts(m, ",sys_immutable");
688	if (opts->discard)
689		seq_puts(m, ",discard");
690	if (opts->force)
691		seq_puts(m, ",force");
692	if (opts->sparse)
693		seq_puts(m, ",sparse");
694	if (opts->nohidden)
695		seq_puts(m, ",nohidden");
696	if (opts->hide_dot_files)
697		seq_puts(m, ",hide_dot_files");
698	if (opts->windows_names)
699		seq_puts(m, ",windows_names");
700	if (opts->showmeta)
701		seq_puts(m, ",showmeta");
702	if (sb->s_flags & SB_POSIXACL)
703		seq_puts(m, ",acl");
704	if (opts->nls)
705		seq_printf(m, ",iocharset=%s", opts->nls->charset);
706	else
707		seq_puts(m, ",iocharset=utf8");
708	if (opts->prealloc)
709		seq_puts(m, ",prealloc");
710	if (opts->nocase)
711		seq_puts(m, ",nocase");
712
713	return 0;
714}
715
716/*
717 * ntfs_shutdown - super_operations::shutdown
718 */
719static void ntfs_shutdown(struct super_block *sb)
720{
721	set_bit(NTFS_FLAGS_SHUTDOWN_BIT, &ntfs_sb(sb)->flags);
722}
723
724/*
725 * ntfs_sync_fs - super_operations::sync_fs
726 */
727static int ntfs_sync_fs(struct super_block *sb, int wait)
728{
729	int err = 0, err2;
730	struct ntfs_sb_info *sbi = sb->s_fs_info;
731	struct ntfs_inode *ni;
732	struct inode *inode;
733
734	if (unlikely(ntfs3_forced_shutdown(sb)))
735		return -EIO;
736
737	ni = sbi->security.ni;
738	if (ni) {
739		inode = &ni->vfs_inode;
740		err2 = _ni_write_inode(inode, wait);
741		if (err2 && !err)
742			err = err2;
743	}
744
745	ni = sbi->objid.ni;
746	if (ni) {
747		inode = &ni->vfs_inode;
748		err2 = _ni_write_inode(inode, wait);
749		if (err2 && !err)
750			err = err2;
751	}
752
753	ni = sbi->reparse.ni;
754	if (ni) {
755		inode = &ni->vfs_inode;
756		err2 = _ni_write_inode(inode, wait);
757		if (err2 && !err)
758			err = err2;
759	}
760
761	if (!err)
762		ntfs_set_state(sbi, NTFS_DIRTY_CLEAR);
763
764	ntfs_update_mftmirr(sbi, wait);
765
766	return err;
767}
768
769static const struct super_operations ntfs_sops = {
770	.alloc_inode = ntfs_alloc_inode,
771	.free_inode = ntfs_free_inode,
772	.evict_inode = ntfs_evict_inode,
773	.put_super = ntfs_put_super,
774	.statfs = ntfs_statfs,
775	.show_options = ntfs_show_options,
776	.shutdown = ntfs_shutdown,
777	.sync_fs = ntfs_sync_fs,
778	.write_inode = ntfs3_write_inode,
779};
780
781static struct inode *ntfs_export_get_inode(struct super_block *sb, u64 ino,
782					   u32 generation)
783{
784	struct MFT_REF ref;
785	struct inode *inode;
786
787	ref.low = cpu_to_le32(ino);
788#ifdef CONFIG_NTFS3_64BIT_CLUSTER
789	ref.high = cpu_to_le16(ino >> 32);
790#else
791	ref.high = 0;
792#endif
793	ref.seq = cpu_to_le16(generation);
794
795	inode = ntfs_iget5(sb, &ref, NULL);
796	if (!IS_ERR(inode) && is_bad_inode(inode)) {
797		iput(inode);
798		inode = ERR_PTR(-ESTALE);
799	}
800
801	return inode;
802}
803
804static struct dentry *ntfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
805					int fh_len, int fh_type)
806{
807	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
808				    ntfs_export_get_inode);
809}
810
811static struct dentry *ntfs_fh_to_parent(struct super_block *sb, struct fid *fid,
812					int fh_len, int fh_type)
813{
814	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
815				    ntfs_export_get_inode);
816}
817
818/* TODO: == ntfs_sync_inode */
819static int ntfs_nfs_commit_metadata(struct inode *inode)
820{
821	return _ni_write_inode(inode, 1);
822}
823
824static const struct export_operations ntfs_export_ops = {
825	.encode_fh = generic_encode_ino32_fh,
826	.fh_to_dentry = ntfs_fh_to_dentry,
827	.fh_to_parent = ntfs_fh_to_parent,
828	.get_parent = ntfs3_get_parent,
829	.commit_metadata = ntfs_nfs_commit_metadata,
830};
831
832/*
833 * format_size_gb - Return Gb,Mb to print with "%u.%02u Gb".
834 */
835static u32 format_size_gb(const u64 bytes, u32 *mb)
836{
837	/* Do simple right 30 bit shift of 64 bit value. */
838	u64 kbytes = bytes >> 10;
839	u32 kbytes32 = kbytes;
840
841	*mb = (100 * (kbytes32 & 0xfffff) + 0x7ffff) >> 20;
842	if (*mb >= 100)
843		*mb = 99;
844
845	return (kbytes32 >> 20) | (((u32)(kbytes >> 32)) << 12);
846}
847
848static u32 true_sectors_per_clst(const struct NTFS_BOOT *boot)
849{
850	if (boot->sectors_per_clusters <= 0x80)
851		return boot->sectors_per_clusters;
852	if (boot->sectors_per_clusters >= 0xf4) /* limit shift to 2MB max */
853		return 1U << (-(s8)boot->sectors_per_clusters);
854	return -EINVAL;
855}
856
857/*
858 * ntfs_init_from_boot - Init internal info from on-disk boot sector.
859 *
860 * NTFS mount begins from boot - special formatted 512 bytes.
861 * There are two boots: the first and the last 512 bytes of volume.
862 * The content of boot is not changed during ntfs life.
863 *
864 * NOTE: ntfs.sys checks only first (primary) boot.
865 * chkdsk checks both boots.
866 */
867static int ntfs_init_from_boot(struct super_block *sb, u32 sector_size,
868			       u64 dev_size, struct NTFS_BOOT **boot2)
869{
870	struct ntfs_sb_info *sbi = sb->s_fs_info;
871	int err;
872	u32 mb, gb, boot_sector_size, sct_per_clst, record_size;
873	u64 sectors, clusters, mlcn, mlcn2, dev_size0;
874	struct NTFS_BOOT *boot;
875	struct buffer_head *bh;
876	struct MFT_REC *rec;
877	u16 fn, ao;
878	u8 cluster_bits;
879	u32 boot_off = 0;
880	sector_t boot_block = 0;
881	const char *hint = "Primary boot";
882
883	/* Save original dev_size. Used with alternative boot. */
884	dev_size0 = dev_size;
885
886	sbi->volume.blocks = dev_size >> PAGE_SHIFT;
887
888read_boot:
889	bh = ntfs_bread(sb, boot_block);
890	if (!bh)
891		return boot_block ? -EINVAL : -EIO;
892
893	err = -EINVAL;
894
895	/* Corrupted image; do not read OOB */
896	if (bh->b_size - sizeof(*boot) < boot_off)
897		goto out;
898
899	boot = (struct NTFS_BOOT *)Add2Ptr(bh->b_data, boot_off);
900
901	if (memcmp(boot->system_id, "NTFS    ", sizeof("NTFS    ") - 1)) {
902		ntfs_err(sb, "%s signature is not NTFS.", hint);
903		goto out;
904	}
905
906	/* 0x55AA is not mandaroty. Thanks Maxim Suhanov*/
907	/*if (0x55 != boot->boot_magic[0] || 0xAA != boot->boot_magic[1])
908	 *	goto out;
909	 */
910
911	boot_sector_size = ((u32)boot->bytes_per_sector[1] << 8) |
912			   boot->bytes_per_sector[0];
913	if (boot_sector_size < SECTOR_SIZE ||
914	    !is_power_of_2(boot_sector_size)) {
915		ntfs_err(sb, "%s: invalid bytes per sector %u.", hint,
916			 boot_sector_size);
917		goto out;
918	}
919
920	/* cluster size: 512, 1K, 2K, 4K, ... 2M */
921	sct_per_clst = true_sectors_per_clst(boot);
922	if ((int)sct_per_clst < 0 || !is_power_of_2(sct_per_clst)) {
923		ntfs_err(sb, "%s: invalid sectors per cluster %u.", hint,
924			 sct_per_clst);
925		goto out;
926	}
927
928	sbi->cluster_size = boot_sector_size * sct_per_clst;
929	sbi->cluster_bits = cluster_bits = blksize_bits(sbi->cluster_size);
930	sbi->cluster_mask = sbi->cluster_size - 1;
931	sbi->cluster_mask_inv = ~(u64)sbi->cluster_mask;
932
933	mlcn = le64_to_cpu(boot->mft_clst);
934	mlcn2 = le64_to_cpu(boot->mft2_clst);
935	sectors = le64_to_cpu(boot->sectors_per_volume);
936
937	if (mlcn * sct_per_clst >= sectors || mlcn2 * sct_per_clst >= sectors) {
938		ntfs_err(
939			sb,
940			"%s: start of MFT 0x%llx (0x%llx) is out of volume 0x%llx.",
941			hint, mlcn, mlcn2, sectors);
942		goto out;
943	}
944
945	if (boot->record_size >= 0) {
946		record_size = (u32)boot->record_size << cluster_bits;
947	} else if (-boot->record_size <= MAXIMUM_SHIFT_BYTES_PER_MFT) {
948		record_size = 1u << (-boot->record_size);
949	} else {
950		ntfs_err(sb, "%s: invalid record size %d.", hint,
951			 boot->record_size);
952		goto out;
953	}
954
955	sbi->record_size = record_size;
956	sbi->record_bits = blksize_bits(record_size);
957	sbi->attr_size_tr = (5 * record_size >> 4); // ~320 bytes
958
959	/* Check MFT record size. */
960	if (record_size < SECTOR_SIZE || !is_power_of_2(record_size)) {
961		ntfs_err(sb, "%s: invalid bytes per MFT record %u (%d).", hint,
962			 record_size, boot->record_size);
963		goto out;
964	}
965
966	if (record_size > MAXIMUM_BYTES_PER_MFT) {
967		ntfs_err(sb, "Unsupported bytes per MFT record %u.",
968			 record_size);
969		goto out;
970	}
971
972	if (boot->index_size >= 0) {
973		sbi->index_size = (u32)boot->index_size << cluster_bits;
974	} else if (-boot->index_size <= MAXIMUM_SHIFT_BYTES_PER_INDEX) {
975		sbi->index_size = 1u << (-boot->index_size);
976	} else {
977		ntfs_err(sb, "%s: invalid index size %d.", hint,
978			 boot->index_size);
979		goto out;
980	}
981
982	/* Check index record size. */
983	if (sbi->index_size < SECTOR_SIZE || !is_power_of_2(sbi->index_size)) {
984		ntfs_err(sb, "%s: invalid bytes per index %u(%d).", hint,
985			 sbi->index_size, boot->index_size);
986		goto out;
987	}
988
989	if (sbi->index_size > MAXIMUM_BYTES_PER_INDEX) {
990		ntfs_err(sb, "%s: unsupported bytes per index %u.", hint,
991			 sbi->index_size);
992		goto out;
993	}
994
995	sbi->volume.size = sectors * boot_sector_size;
996
997	gb = format_size_gb(sbi->volume.size + boot_sector_size, &mb);
998
999	/*
1000	 * - Volume formatted and mounted with the same sector size.
1001	 * - Volume formatted 4K and mounted as 512.
1002	 * - Volume formatted 512 and mounted as 4K.
1003	 */
1004	if (boot_sector_size != sector_size) {
1005		ntfs_warn(
1006			sb,
1007			"Different NTFS sector size (%u) and media sector size (%u).",
1008			boot_sector_size, sector_size);
1009		dev_size += sector_size - 1;
1010	}
1011
1012	sbi->mft.lbo = mlcn << cluster_bits;
1013	sbi->mft.lbo2 = mlcn2 << cluster_bits;
1014
1015	/* Compare boot's cluster and sector. */
1016	if (sbi->cluster_size < boot_sector_size) {
1017		ntfs_err(sb, "%s: invalid bytes per cluster (%u).", hint,
1018			 sbi->cluster_size);
1019		goto out;
1020	}
1021
1022	/* Compare boot's cluster and media sector. */
1023	if (sbi->cluster_size < sector_size) {
1024		/* No way to use ntfs_get_block in this case. */
1025		ntfs_err(
1026			sb,
1027			"Failed to mount 'cause NTFS's cluster size (%u) is less than media sector size (%u).",
1028			sbi->cluster_size, sector_size);
1029		goto out;
1030	}
1031
1032	sbi->max_bytes_per_attr =
1033		record_size - ALIGN(MFTRECORD_FIXUP_OFFSET, 8) -
1034		ALIGN(((record_size >> SECTOR_SHIFT) * sizeof(short)), 8) -
1035		ALIGN(sizeof(enum ATTR_TYPE), 8);
1036
1037	sbi->volume.ser_num = le64_to_cpu(boot->serial_num);
1038
1039	/* Warning if RAW volume. */
1040	if (dev_size < sbi->volume.size + boot_sector_size) {
1041		u32 mb0, gb0;
1042
1043		gb0 = format_size_gb(dev_size, &mb0);
1044		ntfs_warn(
1045			sb,
1046			"RAW NTFS volume: Filesystem size %u.%02u Gb > volume size %u.%02u Gb. Mount in read-only.",
1047			gb, mb, gb0, mb0);
1048		sb->s_flags |= SB_RDONLY;
1049	}
1050
1051	clusters = sbi->volume.size >> cluster_bits;
1052#ifndef CONFIG_NTFS3_64BIT_CLUSTER
1053	/* 32 bits per cluster. */
1054	if (clusters >> 32) {
1055		ntfs_notice(
1056			sb,
1057			"NTFS %u.%02u Gb is too big to use 32 bits per cluster.",
1058			gb, mb);
1059		goto out;
1060	}
1061#elif BITS_PER_LONG < 64
1062#error "CONFIG_NTFS3_64BIT_CLUSTER incompatible in 32 bit OS"
1063#endif
1064
1065	sbi->used.bitmap.nbits = clusters;
1066
1067	rec = kzalloc(record_size, GFP_NOFS);
1068	if (!rec) {
1069		err = -ENOMEM;
1070		goto out;
1071	}
1072
1073	sbi->new_rec = rec;
1074	rec->rhdr.sign = NTFS_FILE_SIGNATURE;
1075	rec->rhdr.fix_off = cpu_to_le16(MFTRECORD_FIXUP_OFFSET);
1076	fn = (sbi->record_size >> SECTOR_SHIFT) + 1;
1077	rec->rhdr.fix_num = cpu_to_le16(fn);
1078	ao = ALIGN(MFTRECORD_FIXUP_OFFSET + sizeof(short) * fn, 8);
1079	rec->attr_off = cpu_to_le16(ao);
1080	rec->used = cpu_to_le32(ao + ALIGN(sizeof(enum ATTR_TYPE), 8));
1081	rec->total = cpu_to_le32(sbi->record_size);
1082	((struct ATTRIB *)Add2Ptr(rec, ao))->type = ATTR_END;
1083
1084	sb_set_blocksize(sb, min_t(u32, sbi->cluster_size, PAGE_SIZE));
1085
1086	sbi->block_mask = sb->s_blocksize - 1;
1087	sbi->blocks_per_cluster = sbi->cluster_size >> sb->s_blocksize_bits;
1088	sbi->volume.blocks = sbi->volume.size >> sb->s_blocksize_bits;
1089
1090	/* Maximum size for normal files. */
1091	sbi->maxbytes = (clusters << cluster_bits) - 1;
1092
1093#ifdef CONFIG_NTFS3_64BIT_CLUSTER
1094	if (clusters >= (1ull << (64 - cluster_bits)))
1095		sbi->maxbytes = -1;
1096	sbi->maxbytes_sparse = -1;
1097	sb->s_maxbytes = MAX_LFS_FILESIZE;
1098#else
1099	/* Maximum size for sparse file. */
1100	sbi->maxbytes_sparse = (1ull << (cluster_bits + 32)) - 1;
1101	sb->s_maxbytes = 0xFFFFFFFFull << cluster_bits;
1102#endif
1103
1104	/*
1105	 * Compute the MFT zone at two steps.
1106	 * It would be nice if we are able to allocate 1/8 of
1107	 * total clusters for MFT but not more then 512 MB.
1108	 */
1109	sbi->zone_max = min_t(CLST, 0x20000000 >> cluster_bits, clusters >> 3);
1110
1111	err = 0;
1112
1113	if (bh->b_blocknr && !sb_rdonly(sb)) {
1114		/*
1115	 	 * Alternative boot is ok but primary is not ok.
1116	 	 * Do not update primary boot here 'cause it may be faked boot.
1117	 	 * Let ntfs to be mounted and update boot later.
1118		 */
1119		*boot2 = kmemdup(boot, sizeof(*boot), GFP_NOFS | __GFP_NOWARN);
1120	}
1121
1122out:
1123	brelse(bh);
1124
1125	if (err == -EINVAL && !boot_block && dev_size0 > PAGE_SHIFT) {
1126		u32 block_size = min_t(u32, sector_size, PAGE_SIZE);
1127		u64 lbo = dev_size0 - sizeof(*boot);
1128
1129		boot_block = lbo >> blksize_bits(block_size);
1130		boot_off = lbo & (block_size - 1);
1131		if (boot_block && block_size >= boot_off + sizeof(*boot)) {
1132			/*
1133			 * Try alternative boot (last sector)
1134			 */
1135			sb_set_blocksize(sb, block_size);
1136			hint = "Alternative boot";
1137			dev_size = dev_size0; /* restore original size. */
1138			goto read_boot;
1139		}
1140	}
1141
1142	return err;
1143}
1144
1145/*
1146 * ntfs_fill_super - Try to mount.
1147 */
1148static int ntfs_fill_super(struct super_block *sb, struct fs_context *fc)
1149{
1150	int err;
1151	struct ntfs_sb_info *sbi = sb->s_fs_info;
1152	struct block_device *bdev = sb->s_bdev;
1153	struct ntfs_mount_options *options;
1154	struct inode *inode;
1155	struct ntfs_inode *ni;
1156	size_t i, tt, bad_len, bad_frags;
1157	CLST vcn, lcn, len;
1158	struct ATTRIB *attr;
1159	const struct VOLUME_INFO *info;
1160	u32 idx, done, bytes;
1161	struct ATTR_DEF_ENTRY *t;
1162	u16 *shared;
1163	struct MFT_REF ref;
1164	bool ro = sb_rdonly(sb);
1165	struct NTFS_BOOT *boot2 = NULL;
1166
1167	ref.high = 0;
1168
1169	sbi->sb = sb;
1170	sbi->options = options = fc->fs_private;
1171	fc->fs_private = NULL;
1172	sb->s_flags |= SB_NODIRATIME;
1173	sb->s_magic = 0x7366746e; // "ntfs"
1174	sb->s_op = &ntfs_sops;
1175	sb->s_export_op = &ntfs_export_ops;
1176	sb->s_time_gran = NTFS_TIME_GRAN; // 100 nsec
1177	sb->s_xattr = ntfs_xattr_handlers;
1178	sb->s_d_op = options->nocase ? &ntfs_dentry_ops : NULL;
1179
1180	options->nls = ntfs_load_nls(options->nls_name);
1181	if (IS_ERR(options->nls)) {
1182		options->nls = NULL;
1183		errorf(fc, "Cannot load nls %s", options->nls_name);
1184		err = -EINVAL;
1185		goto out;
1186	}
1187
1188	if (bdev_max_discard_sectors(bdev) && bdev_discard_granularity(bdev)) {
1189		sbi->discard_granularity = bdev_discard_granularity(bdev);
1190		sbi->discard_granularity_mask_inv =
1191			~(u64)(sbi->discard_granularity - 1);
1192	}
1193
1194	/* Parse boot. */
1195	err = ntfs_init_from_boot(sb, bdev_logical_block_size(bdev),
1196				  bdev_nr_bytes(bdev), &boot2);
1197	if (err)
1198		goto out;
1199
1200	/*
1201	 * Load $Volume. This should be done before $LogFile
1202	 * 'cause 'sbi->volume.ni' is used 'ntfs_set_state'.
1203	 */
1204	ref.low = cpu_to_le32(MFT_REC_VOL);
1205	ref.seq = cpu_to_le16(MFT_REC_VOL);
1206	inode = ntfs_iget5(sb, &ref, &NAME_VOLUME);
1207	if (IS_ERR(inode)) {
1208		err = PTR_ERR(inode);
1209		ntfs_err(sb, "Failed to load $Volume (%d).", err);
1210		goto out;
1211	}
1212
1213	ni = ntfs_i(inode);
1214
1215	/* Load and save label (not necessary). */
1216	attr = ni_find_attr(ni, NULL, NULL, ATTR_LABEL, NULL, 0, NULL, NULL);
1217
1218	if (!attr) {
1219		/* It is ok if no ATTR_LABEL */
1220	} else if (!attr->non_res && !is_attr_ext(attr)) {
1221		/* $AttrDef allows labels to be up to 128 symbols. */
1222		err = utf16s_to_utf8s(resident_data(attr),
1223				      le32_to_cpu(attr->res.data_size) >> 1,
1224				      UTF16_LITTLE_ENDIAN, sbi->volume.label,
1225				      sizeof(sbi->volume.label));
1226		if (err < 0)
1227			sbi->volume.label[0] = 0;
1228	} else {
1229		/* Should we break mounting here? */
1230		//err = -EINVAL;
1231		//goto put_inode_out;
1232	}
1233
1234	attr = ni_find_attr(ni, attr, NULL, ATTR_VOL_INFO, NULL, 0, NULL, NULL);
1235	if (!attr || is_attr_ext(attr) ||
1236	    !(info = resident_data_ex(attr, SIZEOF_ATTRIBUTE_VOLUME_INFO))) {
1237		ntfs_err(sb, "$Volume is corrupted.");
1238		err = -EINVAL;
1239		goto put_inode_out;
1240	}
1241
1242	sbi->volume.major_ver = info->major_ver;
1243	sbi->volume.minor_ver = info->minor_ver;
1244	sbi->volume.flags = info->flags;
1245	sbi->volume.ni = ni;
1246	if (info->flags & VOLUME_FLAG_DIRTY) {
1247		sbi->volume.real_dirty = true;
1248		ntfs_info(sb, "It is recommened to use chkdsk.");
1249	}
1250
1251	/* Load $MFTMirr to estimate recs_mirr. */
1252	ref.low = cpu_to_le32(MFT_REC_MIRR);
1253	ref.seq = cpu_to_le16(MFT_REC_MIRR);
1254	inode = ntfs_iget5(sb, &ref, &NAME_MIRROR);
1255	if (IS_ERR(inode)) {
1256		err = PTR_ERR(inode);
1257		ntfs_err(sb, "Failed to load $MFTMirr (%d).", err);
1258		goto out;
1259	}
1260
1261	sbi->mft.recs_mirr = ntfs_up_cluster(sbi, inode->i_size) >>
1262			     sbi->record_bits;
1263
1264	iput(inode);
1265
1266	/* Load LogFile to replay. */
1267	ref.low = cpu_to_le32(MFT_REC_LOG);
1268	ref.seq = cpu_to_le16(MFT_REC_LOG);
1269	inode = ntfs_iget5(sb, &ref, &NAME_LOGFILE);
1270	if (IS_ERR(inode)) {
1271		err = PTR_ERR(inode);
1272		ntfs_err(sb, "Failed to load \x24LogFile (%d).", err);
1273		goto out;
1274	}
1275
1276	ni = ntfs_i(inode);
1277
1278	err = ntfs_loadlog_and_replay(ni, sbi);
1279	if (err)
1280		goto put_inode_out;
1281
1282	iput(inode);
1283
1284	if ((sbi->flags & NTFS_FLAGS_NEED_REPLAY) && !ro) {
1285		ntfs_warn(sb, "failed to replay log file. Can't mount rw!");
1286		err = -EINVAL;
1287		goto out;
1288	}
1289
1290	if ((sbi->volume.flags & VOLUME_FLAG_DIRTY) && !ro && !options->force) {
1291		ntfs_warn(sb, "volume is dirty and \"force\" flag is not set!");
1292		err = -EINVAL;
1293		goto out;
1294	}
1295
1296	/* Load $MFT. */
1297	ref.low = cpu_to_le32(MFT_REC_MFT);
1298	ref.seq = cpu_to_le16(1);
1299
1300	inode = ntfs_iget5(sb, &ref, &NAME_MFT);
1301	if (IS_ERR(inode)) {
1302		err = PTR_ERR(inode);
1303		ntfs_err(sb, "Failed to load $MFT (%d).", err);
1304		goto out;
1305	}
1306
1307	ni = ntfs_i(inode);
1308
1309	sbi->mft.used = ni->i_valid >> sbi->record_bits;
1310	tt = inode->i_size >> sbi->record_bits;
1311	sbi->mft.next_free = MFT_REC_USER;
1312
1313	err = wnd_init(&sbi->mft.bitmap, sb, tt);
1314	if (err)
1315		goto put_inode_out;
1316
1317	err = ni_load_all_mi(ni);
1318	if (err) {
1319		ntfs_err(sb, "Failed to load $MFT's subrecords (%d).", err);
1320		goto put_inode_out;
1321	}
1322
1323	sbi->mft.ni = ni;
1324
1325	/* Load $Bitmap. */
1326	ref.low = cpu_to_le32(MFT_REC_BITMAP);
1327	ref.seq = cpu_to_le16(MFT_REC_BITMAP);
1328	inode = ntfs_iget5(sb, &ref, &NAME_BITMAP);
1329	if (IS_ERR(inode)) {
1330		err = PTR_ERR(inode);
1331		ntfs_err(sb, "Failed to load $Bitmap (%d).", err);
1332		goto out;
1333	}
1334
1335#ifndef CONFIG_NTFS3_64BIT_CLUSTER
1336	if (inode->i_size >> 32) {
1337		err = -EINVAL;
1338		goto put_inode_out;
1339	}
1340#endif
1341
1342	/* Check bitmap boundary. */
1343	tt = sbi->used.bitmap.nbits;
1344	if (inode->i_size < bitmap_size(tt)) {
1345		ntfs_err(sb, "$Bitmap is corrupted.");
1346		err = -EINVAL;
1347		goto put_inode_out;
1348	}
1349
1350	err = wnd_init(&sbi->used.bitmap, sb, tt);
1351	if (err) {
1352		ntfs_err(sb, "Failed to initialize $Bitmap (%d).", err);
1353		goto put_inode_out;
1354	}
1355
1356	iput(inode);
1357
1358	/* Compute the MFT zone. */
1359	err = ntfs_refresh_zone(sbi);
1360	if (err) {
1361		ntfs_err(sb, "Failed to initialize MFT zone (%d).", err);
1362		goto out;
1363	}
1364
1365	/* Load $BadClus. */
1366	ref.low = cpu_to_le32(MFT_REC_BADCLUST);
1367	ref.seq = cpu_to_le16(MFT_REC_BADCLUST);
1368	inode = ntfs_iget5(sb, &ref, &NAME_BADCLUS);
1369	if (IS_ERR(inode)) {
1370		err = PTR_ERR(inode);
1371		ntfs_err(sb, "Failed to load $BadClus (%d).", err);
1372		goto out;
1373	}
1374
1375	ni = ntfs_i(inode);
1376	bad_len = bad_frags = 0;
1377	for (i = 0; run_get_entry(&ni->file.run, i, &vcn, &lcn, &len); i++) {
1378		if (lcn == SPARSE_LCN)
1379			continue;
1380
1381		bad_len += len;
1382		bad_frags += 1;
1383		if (ro)
1384			continue;
1385
1386		if (wnd_set_used_safe(&sbi->used.bitmap, lcn, len, &tt) || tt) {
1387			/* Bad blocks marked as free in bitmap. */
1388			ntfs_set_state(sbi, NTFS_DIRTY_ERROR);
1389		}
1390	}
1391	if (bad_len) {
1392		/*
1393		 * Notice about bad blocks.
1394		 * In normal cases these blocks are marked as used in bitmap.
1395		 * And we never allocate space in it.
1396		 */
1397		ntfs_notice(sb,
1398			    "Volume contains %zu bad blocks in %zu fragments.",
1399			    bad_len, bad_frags);
1400	}
1401	iput(inode);
1402
1403	/* Load $AttrDef. */
1404	ref.low = cpu_to_le32(MFT_REC_ATTR);
1405	ref.seq = cpu_to_le16(MFT_REC_ATTR);
1406	inode = ntfs_iget5(sb, &ref, &NAME_ATTRDEF);
1407	if (IS_ERR(inode)) {
1408		err = PTR_ERR(inode);
1409		ntfs_err(sb, "Failed to load $AttrDef (%d)", err);
1410		goto out;
1411	}
1412
1413	/*
1414	 * Typical $AttrDef contains up to 20 entries.
1415	 * Check for extremely large/small size.
1416	 */
1417	if (inode->i_size < sizeof(struct ATTR_DEF_ENTRY) ||
1418	    inode->i_size > 100 * sizeof(struct ATTR_DEF_ENTRY)) {
1419		ntfs_err(sb, "Looks like $AttrDef is corrupted (size=%llu).",
1420			 inode->i_size);
1421		err = -EINVAL;
1422		goto put_inode_out;
1423	}
1424
1425	bytes = inode->i_size;
1426	sbi->def_table = t = kvmalloc(bytes, GFP_KERNEL);
1427	if (!t) {
1428		err = -ENOMEM;
1429		goto put_inode_out;
1430	}
1431
1432	for (done = idx = 0; done < bytes; done += PAGE_SIZE, idx++) {
1433		unsigned long tail = bytes - done;
1434		struct page *page = ntfs_map_page(inode->i_mapping, idx);
1435
1436		if (IS_ERR(page)) {
1437			err = PTR_ERR(page);
1438			ntfs_err(sb, "Failed to read $AttrDef (%d).", err);
1439			goto put_inode_out;
1440		}
1441		memcpy(Add2Ptr(t, done), page_address(page),
1442		       min(PAGE_SIZE, tail));
1443		ntfs_unmap_page(page);
1444
1445		if (!idx && ATTR_STD != t->type) {
1446			ntfs_err(sb, "$AttrDef is corrupted.");
1447			err = -EINVAL;
1448			goto put_inode_out;
1449		}
1450	}
1451
1452	t += 1;
1453	sbi->def_entries = 1;
1454	done = sizeof(struct ATTR_DEF_ENTRY);
1455	sbi->reparse.max_size = MAXIMUM_REPARSE_DATA_BUFFER_SIZE;
1456	sbi->ea_max_size = 0x10000; /* default formatter value */
1457
1458	while (done + sizeof(struct ATTR_DEF_ENTRY) <= bytes) {
1459		u32 t32 = le32_to_cpu(t->type);
1460		u64 sz = le64_to_cpu(t->max_sz);
1461
1462		if ((t32 & 0xF) || le32_to_cpu(t[-1].type) >= t32)
1463			break;
1464
1465		if (t->type == ATTR_REPARSE)
1466			sbi->reparse.max_size = sz;
1467		else if (t->type == ATTR_EA)
1468			sbi->ea_max_size = sz;
1469
1470		done += sizeof(struct ATTR_DEF_ENTRY);
1471		t += 1;
1472		sbi->def_entries += 1;
1473	}
1474	iput(inode);
1475
1476	/* Load $UpCase. */
1477	ref.low = cpu_to_le32(MFT_REC_UPCASE);
1478	ref.seq = cpu_to_le16(MFT_REC_UPCASE);
1479	inode = ntfs_iget5(sb, &ref, &NAME_UPCASE);
1480	if (IS_ERR(inode)) {
1481		err = PTR_ERR(inode);
1482		ntfs_err(sb, "Failed to load $UpCase (%d).", err);
1483		goto out;
1484	}
1485
1486	if (inode->i_size != 0x10000 * sizeof(short)) {
1487		err = -EINVAL;
1488		ntfs_err(sb, "$UpCase is corrupted.");
1489		goto put_inode_out;
1490	}
1491
1492	for (idx = 0; idx < (0x10000 * sizeof(short) >> PAGE_SHIFT); idx++) {
1493		const __le16 *src;
1494		u16 *dst = Add2Ptr(sbi->upcase, idx << PAGE_SHIFT);
1495		struct page *page = ntfs_map_page(inode->i_mapping, idx);
1496
1497		if (IS_ERR(page)) {
1498			err = PTR_ERR(page);
1499			ntfs_err(sb, "Failed to read $UpCase (%d).", err);
1500			goto put_inode_out;
1501		}
1502
1503		src = page_address(page);
1504
1505#ifdef __BIG_ENDIAN
1506		for (i = 0; i < PAGE_SIZE / sizeof(u16); i++)
1507			*dst++ = le16_to_cpu(*src++);
1508#else
1509		memcpy(dst, src, PAGE_SIZE);
1510#endif
1511		ntfs_unmap_page(page);
1512	}
1513
1514	shared = ntfs_set_shared(sbi->upcase, 0x10000 * sizeof(short));
1515	if (shared && sbi->upcase != shared) {
1516		kvfree(sbi->upcase);
1517		sbi->upcase = shared;
1518	}
1519
1520	iput(inode);
1521
1522	if (is_ntfs3(sbi)) {
1523		/* Load $Secure. */
1524		err = ntfs_security_init(sbi);
1525		if (err) {
1526			ntfs_err(sb, "Failed to initialize $Secure (%d).", err);
1527			goto out;
1528		}
1529
1530		/* Load $Extend. */
1531		err = ntfs_extend_init(sbi);
1532		if (err) {
1533			ntfs_warn(sb, "Failed to initialize $Extend.");
1534			goto load_root;
1535		}
1536
1537		/* Load $Extend/$Reparse. */
1538		err = ntfs_reparse_init(sbi);
1539		if (err) {
1540			ntfs_warn(sb, "Failed to initialize $Extend/$Reparse.");
1541			goto load_root;
1542		}
1543
1544		/* Load $Extend/$ObjId. */
1545		err = ntfs_objid_init(sbi);
1546		if (err) {
1547			ntfs_warn(sb, "Failed to initialize $Extend/$ObjId.");
1548			goto load_root;
1549		}
1550	}
1551
1552load_root:
1553	/* Load root. */
1554	ref.low = cpu_to_le32(MFT_REC_ROOT);
1555	ref.seq = cpu_to_le16(MFT_REC_ROOT);
1556	inode = ntfs_iget5(sb, &ref, &NAME_ROOT);
1557	if (IS_ERR(inode)) {
1558		err = PTR_ERR(inode);
1559		ntfs_err(sb, "Failed to load root (%d).", err);
1560		goto out;
1561	}
1562
1563	/*
1564	 * Final check. Looks like this case should never occurs.
1565	 */
1566	if (!inode->i_op) {
1567		err = -EINVAL;
1568		ntfs_err(sb, "Failed to load root (%d).", err);
1569		goto put_inode_out;
1570	}
1571
1572	sb->s_root = d_make_root(inode);
1573	if (!sb->s_root) {
1574		err = -ENOMEM;
1575		goto put_inode_out;
1576	}
1577
1578	if (boot2) {
1579		/*
1580	 	 * Alternative boot is ok but primary is not ok.
1581	 	 * Volume is recognized as NTFS. Update primary boot.
1582		 */
1583		struct buffer_head *bh0 = sb_getblk(sb, 0);
1584		if (bh0) {
1585			if (buffer_locked(bh0))
1586				__wait_on_buffer(bh0);
1587
1588			lock_buffer(bh0);
1589			memcpy(bh0->b_data, boot2, sizeof(*boot2));
1590			set_buffer_uptodate(bh0);
1591			mark_buffer_dirty(bh0);
1592			unlock_buffer(bh0);
1593			if (!sync_dirty_buffer(bh0))
1594				ntfs_warn(sb, "primary boot is updated");
1595			put_bh(bh0);
1596		}
1597
1598		kfree(boot2);
1599	}
1600
1601#ifdef CONFIG_PROC_FS
1602	/* Create /proc/fs/ntfs3/.. */
1603	if (proc_info_root) {
1604		struct proc_dir_entry *e = proc_mkdir(sb->s_id, proc_info_root);
1605		static_assert((S_IRUGO | S_IWUSR) == 0644);
1606		if (e) {
1607			proc_create_data("volinfo", S_IRUGO, e,
1608					 &ntfs3_volinfo_fops, sb);
1609			proc_create_data("label", S_IRUGO | S_IWUSR, e,
1610					 &ntfs3_label_fops, sb);
1611			sbi->procdir = e;
1612		}
1613	}
1614#endif
1615
1616	return 0;
1617
1618put_inode_out:
1619	iput(inode);
1620out:
1621	ntfs3_put_sbi(sbi);
1622	kfree(boot2);
1623	ntfs3_put_sbi(sbi);
1624	return err;
1625}
1626
1627void ntfs_unmap_meta(struct super_block *sb, CLST lcn, CLST len)
1628{
1629	struct ntfs_sb_info *sbi = sb->s_fs_info;
1630	struct block_device *bdev = sb->s_bdev;
1631	sector_t devblock = (u64)lcn * sbi->blocks_per_cluster;
1632	unsigned long blocks = (u64)len * sbi->blocks_per_cluster;
1633	unsigned long cnt = 0;
1634	unsigned long limit = global_zone_page_state(NR_FREE_PAGES)
1635			      << (PAGE_SHIFT - sb->s_blocksize_bits);
1636
1637	if (limit >= 0x2000)
1638		limit -= 0x1000;
1639	else if (limit < 32)
1640		limit = 32;
1641	else
1642		limit >>= 1;
1643
1644	while (blocks--) {
1645		clean_bdev_aliases(bdev, devblock++, 1);
1646		if (cnt++ >= limit) {
1647			sync_blockdev(bdev);
1648			cnt = 0;
1649		}
1650	}
1651}
1652
1653/*
1654 * ntfs_discard - Issue a discard request (trim for SSD).
1655 */
1656int ntfs_discard(struct ntfs_sb_info *sbi, CLST lcn, CLST len)
1657{
1658	int err;
1659	u64 lbo, bytes, start, end;
1660	struct super_block *sb;
1661
1662	if (sbi->used.next_free_lcn == lcn + len)
1663		sbi->used.next_free_lcn = lcn;
1664
1665	if (sbi->flags & NTFS_FLAGS_NODISCARD)
1666		return -EOPNOTSUPP;
1667
1668	if (!sbi->options->discard)
1669		return -EOPNOTSUPP;
1670
1671	lbo = (u64)lcn << sbi->cluster_bits;
1672	bytes = (u64)len << sbi->cluster_bits;
1673
1674	/* Align up 'start' on discard_granularity. */
1675	start = (lbo + sbi->discard_granularity - 1) &
1676		sbi->discard_granularity_mask_inv;
1677	/* Align down 'end' on discard_granularity. */
1678	end = (lbo + bytes) & sbi->discard_granularity_mask_inv;
1679
1680	sb = sbi->sb;
1681	if (start >= end)
1682		return 0;
1683
1684	err = blkdev_issue_discard(sb->s_bdev, start >> 9, (end - start) >> 9,
1685				   GFP_NOFS);
1686
1687	if (err == -EOPNOTSUPP)
1688		sbi->flags |= NTFS_FLAGS_NODISCARD;
1689
1690	return err;
1691}
1692
1693static int ntfs_fs_get_tree(struct fs_context *fc)
1694{
1695	return get_tree_bdev(fc, ntfs_fill_super);
1696}
1697
1698/*
1699 * ntfs_fs_free - Free fs_context.
1700 *
1701 * Note that this will be called after fill_super and reconfigure
1702 * even when they pass. So they have to take pointers if they pass.
1703 */
1704static void ntfs_fs_free(struct fs_context *fc)
1705{
1706	struct ntfs_mount_options *opts = fc->fs_private;
1707	struct ntfs_sb_info *sbi = fc->s_fs_info;
1708
1709	if (sbi) {
1710		ntfs3_put_sbi(sbi);
1711		ntfs3_free_sbi(sbi);
1712	}
1713
1714	if (opts)
1715		put_mount_options(opts);
1716}
1717
1718// clang-format off
1719static const struct fs_context_operations ntfs_context_ops = {
1720	.parse_param	= ntfs_fs_parse_param,
1721	.get_tree	= ntfs_fs_get_tree,
1722	.reconfigure	= ntfs_fs_reconfigure,
1723	.free		= ntfs_fs_free,
1724};
1725// clang-format on
1726
1727/*
1728 * ntfs_init_fs_context - Initialize sbi and opts
1729 *
1730 * This will called when mount/remount. We will first initialize
1731 * options so that if remount we can use just that.
1732 */
1733static int ntfs_init_fs_context(struct fs_context *fc)
1734{
1735	struct ntfs_mount_options *opts;
1736	struct ntfs_sb_info *sbi;
1737
1738	opts = kzalloc(sizeof(struct ntfs_mount_options), GFP_NOFS);
1739	if (!opts)
1740		return -ENOMEM;
1741
1742	/* Default options. */
1743	opts->fs_uid = current_uid();
1744	opts->fs_gid = current_gid();
1745	opts->fs_fmask_inv = ~current_umask();
1746	opts->fs_dmask_inv = ~current_umask();
1747
1748	if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE)
1749		goto ok;
1750
1751	sbi = kzalloc(sizeof(struct ntfs_sb_info), GFP_NOFS);
1752	if (!sbi)
1753		goto free_opts;
1754
1755	sbi->upcase = kvmalloc(0x10000 * sizeof(short), GFP_KERNEL);
1756	if (!sbi->upcase)
1757		goto free_sbi;
1758
1759	ratelimit_state_init(&sbi->msg_ratelimit, DEFAULT_RATELIMIT_INTERVAL,
1760			     DEFAULT_RATELIMIT_BURST);
1761
1762	mutex_init(&sbi->compress.mtx_lznt);
1763#ifdef CONFIG_NTFS3_LZX_XPRESS
1764	mutex_init(&sbi->compress.mtx_xpress);
1765	mutex_init(&sbi->compress.mtx_lzx);
1766#endif
1767
1768	fc->s_fs_info = sbi;
1769ok:
1770	fc->fs_private = opts;
1771	fc->ops = &ntfs_context_ops;
1772
1773	return 0;
1774free_sbi:
1775	kfree(sbi);
1776free_opts:
1777	kfree(opts);
1778	return -ENOMEM;
1779}
1780
1781static void ntfs3_kill_sb(struct super_block *sb)
1782{
1783	struct ntfs_sb_info *sbi = sb->s_fs_info;
1784
1785	kill_block_super(sb);
1786
1787	if (sbi->options)
1788		put_mount_options(sbi->options);
1789	ntfs3_free_sbi(sbi);
1790}
1791
1792// clang-format off
1793static struct file_system_type ntfs_fs_type = {
1794	.owner			= THIS_MODULE,
1795	.name			= "ntfs3",
1796	.init_fs_context	= ntfs_init_fs_context,
1797	.parameters		= ntfs_fs_parameters,
1798	.kill_sb		= ntfs3_kill_sb,
1799	.fs_flags		= FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
1800};
1801// clang-format on
1802
1803static int __init init_ntfs_fs(void)
1804{
1805	int err;
1806
1807	pr_info("ntfs3: Max link count %u\n", NTFS_LINK_MAX);
1808
1809	if (IS_ENABLED(CONFIG_NTFS3_FS_POSIX_ACL))
1810		pr_info("ntfs3: Enabled Linux POSIX ACLs support\n");
1811	if (IS_ENABLED(CONFIG_NTFS3_64BIT_CLUSTER))
1812		pr_notice(
1813			"ntfs3: Warning: Activated 64 bits per cluster. Windows does not support this\n");
1814	if (IS_ENABLED(CONFIG_NTFS3_LZX_XPRESS))
1815		pr_info("ntfs3: Read-only LZX/Xpress compression included\n");
1816
1817#ifdef CONFIG_PROC_FS
1818	/* Create "/proc/fs/ntfs3" */
1819	proc_info_root = proc_mkdir("fs/ntfs3", NULL);
1820#endif
1821
1822	err = ntfs3_init_bitmap();
1823	if (err)
1824		return err;
1825
1826	ntfs_inode_cachep = kmem_cache_create(
1827		"ntfs_inode_cache", sizeof(struct ntfs_inode), 0,
1828		(SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT),
1829		init_once);
1830	if (!ntfs_inode_cachep) {
1831		err = -ENOMEM;
1832		goto out1;
1833	}
1834
1835	err = register_filesystem(&ntfs_fs_type);
1836	if (err)
1837		goto out;
1838
1839	return 0;
1840out:
1841	kmem_cache_destroy(ntfs_inode_cachep);
1842out1:
1843	ntfs3_exit_bitmap();
1844	return err;
1845}
1846
1847static void __exit exit_ntfs_fs(void)
1848{
1849	rcu_barrier();
1850	kmem_cache_destroy(ntfs_inode_cachep);
1851	unregister_filesystem(&ntfs_fs_type);
1852	ntfs3_exit_bitmap();
1853
1854#ifdef CONFIG_PROC_FS
1855	if (proc_info_root)
1856		remove_proc_entry("fs/ntfs3", NULL);
1857#endif
1858}
1859
1860MODULE_LICENSE("GPL");
1861MODULE_DESCRIPTION("ntfs3 read/write filesystem");
1862#ifdef CONFIG_NTFS3_FS_POSIX_ACL
1863MODULE_INFO(behaviour, "Enabled Linux POSIX ACLs support");
1864#endif
1865#ifdef CONFIG_NTFS3_64BIT_CLUSTER
1866MODULE_INFO(
1867	cluster,
1868	"Warning: Activated 64 bits per cluster. Windows does not support this");
1869#endif
1870#ifdef CONFIG_NTFS3_LZX_XPRESS
1871MODULE_INFO(compression, "Read-only lzx/xpress compression included");
1872#endif
1873
1874MODULE_AUTHOR("Konstantin Komarov");
1875MODULE_ALIAS_FS("ntfs3");
1876
1877module_init(init_ntfs_fs);
1878module_exit(exit_ntfs_fs);
1879