xref: /kernel/linux/linux-5.10/fs/ext2/super.c (revision 8c2ecf20)
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 *  linux/fs/ext2/super.c
4 *
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
9 *
10 *  from
11 *
12 *  linux/fs/minix/inode.c
13 *
14 *  Copyright (C) 1991, 1992  Linus Torvalds
15 *
16 *  Big-endian to little-endian byte-swapping/bitmaps by
17 *        David S. Miller (davem@caip.rutgers.edu), 1995
18 */
19
20#include <linux/module.h>
21#include <linux/string.h>
22#include <linux/fs.h>
23#include <linux/slab.h>
24#include <linux/init.h>
25#include <linux/blkdev.h>
26#include <linux/parser.h>
27#include <linux/random.h>
28#include <linux/buffer_head.h>
29#include <linux/exportfs.h>
30#include <linux/vfs.h>
31#include <linux/seq_file.h>
32#include <linux/mount.h>
33#include <linux/log2.h>
34#include <linux/quotaops.h>
35#include <linux/uaccess.h>
36#include <linux/dax.h>
37#include <linux/iversion.h>
38#include "ext2.h"
39#include "xattr.h"
40#include "acl.h"
41
42static void ext2_write_super(struct super_block *sb);
43static int ext2_remount (struct super_block * sb, int * flags, char * data);
44static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
45static int ext2_sync_fs(struct super_block *sb, int wait);
46static int ext2_freeze(struct super_block *sb);
47static int ext2_unfreeze(struct super_block *sb);
48
49void ext2_error(struct super_block *sb, const char *function,
50		const char *fmt, ...)
51{
52	struct va_format vaf;
53	va_list args;
54	struct ext2_sb_info *sbi = EXT2_SB(sb);
55	struct ext2_super_block *es = sbi->s_es;
56
57	if (!sb_rdonly(sb)) {
58		spin_lock(&sbi->s_lock);
59		sbi->s_mount_state |= EXT2_ERROR_FS;
60		es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
61		spin_unlock(&sbi->s_lock);
62		ext2_sync_super(sb, es, 1);
63	}
64
65	va_start(args, fmt);
66
67	vaf.fmt = fmt;
68	vaf.va = &args;
69
70	printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
71	       sb->s_id, function, &vaf);
72
73	va_end(args);
74
75	if (test_opt(sb, ERRORS_PANIC))
76		panic("EXT2-fs: panic from previous error\n");
77	if (!sb_rdonly(sb) && test_opt(sb, ERRORS_RO)) {
78		ext2_msg(sb, KERN_CRIT,
79			     "error: remounting filesystem read-only");
80		sb->s_flags |= SB_RDONLY;
81	}
82}
83
84void ext2_msg(struct super_block *sb, const char *prefix,
85		const char *fmt, ...)
86{
87	struct va_format vaf;
88	va_list args;
89
90	va_start(args, fmt);
91
92	vaf.fmt = fmt;
93	vaf.va = &args;
94
95	printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
96
97	va_end(args);
98}
99
100/*
101 * This must be called with sbi->s_lock held.
102 */
103void ext2_update_dynamic_rev(struct super_block *sb)
104{
105	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
106
107	if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
108		return;
109
110	ext2_msg(sb, KERN_WARNING,
111		     "warning: updating to rev %d because of "
112		     "new feature flag, running e2fsck is recommended",
113		     EXT2_DYNAMIC_REV);
114
115	es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
116	es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
117	es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
118	/* leave es->s_feature_*compat flags alone */
119	/* es->s_uuid will be set by e2fsck if empty */
120
121	/*
122	 * The rest of the superblock fields should be zero, and if not it
123	 * means they are likely already in use, so leave them alone.  We
124	 * can leave it up to e2fsck to clean up any inconsistencies there.
125	 */
126}
127
128#ifdef CONFIG_QUOTA
129static int ext2_quota_off(struct super_block *sb, int type);
130
131static void ext2_quota_off_umount(struct super_block *sb)
132{
133	int type;
134
135	for (type = 0; type < MAXQUOTAS; type++)
136		ext2_quota_off(sb, type);
137}
138#else
139static inline void ext2_quota_off_umount(struct super_block *sb)
140{
141}
142#endif
143
144static void ext2_put_super (struct super_block * sb)
145{
146	int db_count;
147	int i;
148	struct ext2_sb_info *sbi = EXT2_SB(sb);
149
150	ext2_quota_off_umount(sb);
151
152	ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
153	sbi->s_ea_block_cache = NULL;
154
155	if (!sb_rdonly(sb)) {
156		struct ext2_super_block *es = sbi->s_es;
157
158		spin_lock(&sbi->s_lock);
159		es->s_state = cpu_to_le16(sbi->s_mount_state);
160		spin_unlock(&sbi->s_lock);
161		ext2_sync_super(sb, es, 1);
162	}
163	db_count = sbi->s_gdb_count;
164	for (i = 0; i < db_count; i++)
165		brelse(sbi->s_group_desc[i]);
166	kfree(sbi->s_group_desc);
167	kfree(sbi->s_debts);
168	percpu_counter_destroy(&sbi->s_freeblocks_counter);
169	percpu_counter_destroy(&sbi->s_freeinodes_counter);
170	percpu_counter_destroy(&sbi->s_dirs_counter);
171	brelse (sbi->s_sbh);
172	sb->s_fs_info = NULL;
173	kfree(sbi->s_blockgroup_lock);
174	fs_put_dax(sbi->s_daxdev);
175	kfree(sbi);
176}
177
178static struct kmem_cache * ext2_inode_cachep;
179
180static struct inode *ext2_alloc_inode(struct super_block *sb)
181{
182	struct ext2_inode_info *ei;
183	ei = kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
184	if (!ei)
185		return NULL;
186	ei->i_block_alloc_info = NULL;
187	inode_set_iversion(&ei->vfs_inode, 1);
188#ifdef CONFIG_QUOTA
189	memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
190#endif
191
192	return &ei->vfs_inode;
193}
194
195static void ext2_free_in_core_inode(struct inode *inode)
196{
197	kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
198}
199
200static void init_once(void *foo)
201{
202	struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
203
204	rwlock_init(&ei->i_meta_lock);
205#ifdef CONFIG_EXT2_FS_XATTR
206	init_rwsem(&ei->xattr_sem);
207#endif
208	mutex_init(&ei->truncate_mutex);
209#ifdef CONFIG_FS_DAX
210	init_rwsem(&ei->dax_sem);
211#endif
212	inode_init_once(&ei->vfs_inode);
213}
214
215static int __init init_inodecache(void)
216{
217	ext2_inode_cachep = kmem_cache_create_usercopy("ext2_inode_cache",
218				sizeof(struct ext2_inode_info), 0,
219				(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
220					SLAB_ACCOUNT),
221				offsetof(struct ext2_inode_info, i_data),
222				sizeof_field(struct ext2_inode_info, i_data),
223				init_once);
224	if (ext2_inode_cachep == NULL)
225		return -ENOMEM;
226	return 0;
227}
228
229static void destroy_inodecache(void)
230{
231	/*
232	 * Make sure all delayed rcu free inodes are flushed before we
233	 * destroy cache.
234	 */
235	rcu_barrier();
236	kmem_cache_destroy(ext2_inode_cachep);
237}
238
239static int ext2_show_options(struct seq_file *seq, struct dentry *root)
240{
241	struct super_block *sb = root->d_sb;
242	struct ext2_sb_info *sbi = EXT2_SB(sb);
243	struct ext2_super_block *es = sbi->s_es;
244	unsigned long def_mount_opts;
245
246	spin_lock(&sbi->s_lock);
247	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
248
249	if (sbi->s_sb_block != 1)
250		seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
251	if (test_opt(sb, MINIX_DF))
252		seq_puts(seq, ",minixdf");
253	if (test_opt(sb, GRPID))
254		seq_puts(seq, ",grpid");
255	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
256		seq_puts(seq, ",nogrpid");
257	if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
258	    le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
259		seq_printf(seq, ",resuid=%u",
260				from_kuid_munged(&init_user_ns, sbi->s_resuid));
261	}
262	if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
263	    le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
264		seq_printf(seq, ",resgid=%u",
265				from_kgid_munged(&init_user_ns, sbi->s_resgid));
266	}
267	if (test_opt(sb, ERRORS_RO)) {
268		int def_errors = le16_to_cpu(es->s_errors);
269
270		if (def_errors == EXT2_ERRORS_PANIC ||
271		    def_errors == EXT2_ERRORS_CONTINUE) {
272			seq_puts(seq, ",errors=remount-ro");
273		}
274	}
275	if (test_opt(sb, ERRORS_CONT))
276		seq_puts(seq, ",errors=continue");
277	if (test_opt(sb, ERRORS_PANIC))
278		seq_puts(seq, ",errors=panic");
279	if (test_opt(sb, NO_UID32))
280		seq_puts(seq, ",nouid32");
281	if (test_opt(sb, DEBUG))
282		seq_puts(seq, ",debug");
283	if (test_opt(sb, OLDALLOC))
284		seq_puts(seq, ",oldalloc");
285
286#ifdef CONFIG_EXT2_FS_XATTR
287	if (test_opt(sb, XATTR_USER))
288		seq_puts(seq, ",user_xattr");
289	if (!test_opt(sb, XATTR_USER) &&
290	    (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
291		seq_puts(seq, ",nouser_xattr");
292	}
293#endif
294
295#ifdef CONFIG_EXT2_FS_POSIX_ACL
296	if (test_opt(sb, POSIX_ACL))
297		seq_puts(seq, ",acl");
298	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
299		seq_puts(seq, ",noacl");
300#endif
301
302	if (test_opt(sb, NOBH))
303		seq_puts(seq, ",nobh");
304
305	if (test_opt(sb, USRQUOTA))
306		seq_puts(seq, ",usrquota");
307
308	if (test_opt(sb, GRPQUOTA))
309		seq_puts(seq, ",grpquota");
310
311	if (test_opt(sb, XIP))
312		seq_puts(seq, ",xip");
313
314	if (test_opt(sb, DAX))
315		seq_puts(seq, ",dax");
316
317	if (!test_opt(sb, RESERVATION))
318		seq_puts(seq, ",noreservation");
319
320	spin_unlock(&sbi->s_lock);
321	return 0;
322}
323
324#ifdef CONFIG_QUOTA
325static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
326static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
327static int ext2_quota_on(struct super_block *sb, int type, int format_id,
328			 const struct path *path);
329static struct dquot **ext2_get_dquots(struct inode *inode)
330{
331	return EXT2_I(inode)->i_dquot;
332}
333
334static const struct quotactl_ops ext2_quotactl_ops = {
335	.quota_on	= ext2_quota_on,
336	.quota_off	= ext2_quota_off,
337	.quota_sync	= dquot_quota_sync,
338	.get_state	= dquot_get_state,
339	.set_info	= dquot_set_dqinfo,
340	.get_dqblk	= dquot_get_dqblk,
341	.set_dqblk	= dquot_set_dqblk,
342	.get_nextdqblk	= dquot_get_next_dqblk,
343};
344#endif
345
346static const struct super_operations ext2_sops = {
347	.alloc_inode	= ext2_alloc_inode,
348	.free_inode	= ext2_free_in_core_inode,
349	.write_inode	= ext2_write_inode,
350	.evict_inode	= ext2_evict_inode,
351	.put_super	= ext2_put_super,
352	.sync_fs	= ext2_sync_fs,
353	.freeze_fs	= ext2_freeze,
354	.unfreeze_fs	= ext2_unfreeze,
355	.statfs		= ext2_statfs,
356	.remount_fs	= ext2_remount,
357	.show_options	= ext2_show_options,
358#ifdef CONFIG_QUOTA
359	.quota_read	= ext2_quota_read,
360	.quota_write	= ext2_quota_write,
361	.get_dquots	= ext2_get_dquots,
362#endif
363};
364
365static struct inode *ext2_nfs_get_inode(struct super_block *sb,
366		u64 ino, u32 generation)
367{
368	struct inode *inode;
369
370	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
371		return ERR_PTR(-ESTALE);
372	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
373		return ERR_PTR(-ESTALE);
374
375	/*
376	 * ext2_iget isn't quite right if the inode is currently unallocated!
377	 * However ext2_iget currently does appropriate checks to handle stale
378	 * inodes so everything is OK.
379	 */
380	inode = ext2_iget(sb, ino);
381	if (IS_ERR(inode))
382		return ERR_CAST(inode);
383	if (generation && inode->i_generation != generation) {
384		/* we didn't find the right inode.. */
385		iput(inode);
386		return ERR_PTR(-ESTALE);
387	}
388	return inode;
389}
390
391static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
392		int fh_len, int fh_type)
393{
394	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
395				    ext2_nfs_get_inode);
396}
397
398static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
399		int fh_len, int fh_type)
400{
401	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
402				    ext2_nfs_get_inode);
403}
404
405static const struct export_operations ext2_export_ops = {
406	.fh_to_dentry = ext2_fh_to_dentry,
407	.fh_to_parent = ext2_fh_to_parent,
408	.get_parent = ext2_get_parent,
409};
410
411static unsigned long get_sb_block(void **data)
412{
413	unsigned long 	sb_block;
414	char 		*options = (char *) *data;
415
416	if (!options || strncmp(options, "sb=", 3) != 0)
417		return 1;	/* Default location */
418	options += 3;
419	sb_block = simple_strtoul(options, &options, 0);
420	if (*options && *options != ',') {
421		printk("EXT2-fs: Invalid sb specification: %s\n",
422		       (char *) *data);
423		return 1;
424	}
425	if (*options == ',')
426		options++;
427	*data = (void *) options;
428	return sb_block;
429}
430
431enum {
432	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
433	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
434	Opt_err_ro, Opt_nouid32, Opt_debug,
435	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
436	Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
437	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
438};
439
440static const match_table_t tokens = {
441	{Opt_bsd_df, "bsddf"},
442	{Opt_minix_df, "minixdf"},
443	{Opt_grpid, "grpid"},
444	{Opt_grpid, "bsdgroups"},
445	{Opt_nogrpid, "nogrpid"},
446	{Opt_nogrpid, "sysvgroups"},
447	{Opt_resgid, "resgid=%u"},
448	{Opt_resuid, "resuid=%u"},
449	{Opt_sb, "sb=%u"},
450	{Opt_err_cont, "errors=continue"},
451	{Opt_err_panic, "errors=panic"},
452	{Opt_err_ro, "errors=remount-ro"},
453	{Opt_nouid32, "nouid32"},
454	{Opt_debug, "debug"},
455	{Opt_oldalloc, "oldalloc"},
456	{Opt_orlov, "orlov"},
457	{Opt_nobh, "nobh"},
458	{Opt_user_xattr, "user_xattr"},
459	{Opt_nouser_xattr, "nouser_xattr"},
460	{Opt_acl, "acl"},
461	{Opt_noacl, "noacl"},
462	{Opt_xip, "xip"},
463	{Opt_dax, "dax"},
464	{Opt_grpquota, "grpquota"},
465	{Opt_ignore, "noquota"},
466	{Opt_quota, "quota"},
467	{Opt_usrquota, "usrquota"},
468	{Opt_reservation, "reservation"},
469	{Opt_noreservation, "noreservation"},
470	{Opt_err, NULL}
471};
472
473static int parse_options(char *options, struct super_block *sb,
474			 struct ext2_mount_options *opts)
475{
476	char *p;
477	substring_t args[MAX_OPT_ARGS];
478	int option;
479	kuid_t uid;
480	kgid_t gid;
481
482	if (!options)
483		return 1;
484
485	while ((p = strsep (&options, ",")) != NULL) {
486		int token;
487		if (!*p)
488			continue;
489
490		token = match_token(p, tokens, args);
491		switch (token) {
492		case Opt_bsd_df:
493			clear_opt (opts->s_mount_opt, MINIX_DF);
494			break;
495		case Opt_minix_df:
496			set_opt (opts->s_mount_opt, MINIX_DF);
497			break;
498		case Opt_grpid:
499			set_opt (opts->s_mount_opt, GRPID);
500			break;
501		case Opt_nogrpid:
502			clear_opt (opts->s_mount_opt, GRPID);
503			break;
504		case Opt_resuid:
505			if (match_int(&args[0], &option))
506				return 0;
507			uid = make_kuid(current_user_ns(), option);
508			if (!uid_valid(uid)) {
509				ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
510				return 0;
511
512			}
513			opts->s_resuid = uid;
514			break;
515		case Opt_resgid:
516			if (match_int(&args[0], &option))
517				return 0;
518			gid = make_kgid(current_user_ns(), option);
519			if (!gid_valid(gid)) {
520				ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
521				return 0;
522			}
523			opts->s_resgid = gid;
524			break;
525		case Opt_sb:
526			/* handled by get_sb_block() instead of here */
527			/* *sb_block = match_int(&args[0]); */
528			break;
529		case Opt_err_panic:
530			clear_opt (opts->s_mount_opt, ERRORS_CONT);
531			clear_opt (opts->s_mount_opt, ERRORS_RO);
532			set_opt (opts->s_mount_opt, ERRORS_PANIC);
533			break;
534		case Opt_err_ro:
535			clear_opt (opts->s_mount_opt, ERRORS_CONT);
536			clear_opt (opts->s_mount_opt, ERRORS_PANIC);
537			set_opt (opts->s_mount_opt, ERRORS_RO);
538			break;
539		case Opt_err_cont:
540			clear_opt (opts->s_mount_opt, ERRORS_RO);
541			clear_opt (opts->s_mount_opt, ERRORS_PANIC);
542			set_opt (opts->s_mount_opt, ERRORS_CONT);
543			break;
544		case Opt_nouid32:
545			set_opt (opts->s_mount_opt, NO_UID32);
546			break;
547		case Opt_debug:
548			set_opt (opts->s_mount_opt, DEBUG);
549			break;
550		case Opt_oldalloc:
551			set_opt (opts->s_mount_opt, OLDALLOC);
552			break;
553		case Opt_orlov:
554			clear_opt (opts->s_mount_opt, OLDALLOC);
555			break;
556		case Opt_nobh:
557			set_opt (opts->s_mount_opt, NOBH);
558			break;
559#ifdef CONFIG_EXT2_FS_XATTR
560		case Opt_user_xattr:
561			set_opt (opts->s_mount_opt, XATTR_USER);
562			break;
563		case Opt_nouser_xattr:
564			clear_opt (opts->s_mount_opt, XATTR_USER);
565			break;
566#else
567		case Opt_user_xattr:
568		case Opt_nouser_xattr:
569			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
570				"not supported");
571			break;
572#endif
573#ifdef CONFIG_EXT2_FS_POSIX_ACL
574		case Opt_acl:
575			set_opt(opts->s_mount_opt, POSIX_ACL);
576			break;
577		case Opt_noacl:
578			clear_opt(opts->s_mount_opt, POSIX_ACL);
579			break;
580#else
581		case Opt_acl:
582		case Opt_noacl:
583			ext2_msg(sb, KERN_INFO,
584				"(no)acl options not supported");
585			break;
586#endif
587		case Opt_xip:
588			ext2_msg(sb, KERN_INFO, "use dax instead of xip");
589			set_opt(opts->s_mount_opt, XIP);
590			fallthrough;
591		case Opt_dax:
592#ifdef CONFIG_FS_DAX
593			ext2_msg(sb, KERN_WARNING,
594		"DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
595			set_opt(opts->s_mount_opt, DAX);
596#else
597			ext2_msg(sb, KERN_INFO, "dax option not supported");
598#endif
599			break;
600
601#if defined(CONFIG_QUOTA)
602		case Opt_quota:
603		case Opt_usrquota:
604			set_opt(opts->s_mount_opt, USRQUOTA);
605			break;
606
607		case Opt_grpquota:
608			set_opt(opts->s_mount_opt, GRPQUOTA);
609			break;
610#else
611		case Opt_quota:
612		case Opt_usrquota:
613		case Opt_grpquota:
614			ext2_msg(sb, KERN_INFO,
615				"quota operations not supported");
616			break;
617#endif
618
619		case Opt_reservation:
620			set_opt(opts->s_mount_opt, RESERVATION);
621			ext2_msg(sb, KERN_INFO, "reservations ON");
622			break;
623		case Opt_noreservation:
624			clear_opt(opts->s_mount_opt, RESERVATION);
625			ext2_msg(sb, KERN_INFO, "reservations OFF");
626			break;
627		case Opt_ignore:
628			break;
629		default:
630			return 0;
631		}
632	}
633	return 1;
634}
635
636static int ext2_setup_super (struct super_block * sb,
637			      struct ext2_super_block * es,
638			      int read_only)
639{
640	int res = 0;
641	struct ext2_sb_info *sbi = EXT2_SB(sb);
642
643	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
644		ext2_msg(sb, KERN_ERR,
645			"error: revision level too high, "
646			"forcing read-only mode");
647		res = SB_RDONLY;
648	}
649	if (read_only)
650		return res;
651	if (!(sbi->s_mount_state & EXT2_VALID_FS))
652		ext2_msg(sb, KERN_WARNING,
653			"warning: mounting unchecked fs, "
654			"running e2fsck is recommended");
655	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
656		ext2_msg(sb, KERN_WARNING,
657			"warning: mounting fs with errors, "
658			"running e2fsck is recommended");
659	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
660		 le16_to_cpu(es->s_mnt_count) >=
661		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
662		ext2_msg(sb, KERN_WARNING,
663			"warning: maximal mount count reached, "
664			"running e2fsck is recommended");
665	else if (le32_to_cpu(es->s_checkinterval) &&
666		(le32_to_cpu(es->s_lastcheck) +
667			le32_to_cpu(es->s_checkinterval) <=
668			ktime_get_real_seconds()))
669		ext2_msg(sb, KERN_WARNING,
670			"warning: checktime reached, "
671			"running e2fsck is recommended");
672	if (!le16_to_cpu(es->s_max_mnt_count))
673		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
674	le16_add_cpu(&es->s_mnt_count, 1);
675	if (test_opt (sb, DEBUG))
676		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, gc=%lu, "
677			"bpg=%lu, ipg=%lu, mo=%04lx]",
678			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
679			sbi->s_groups_count,
680			EXT2_BLOCKS_PER_GROUP(sb),
681			EXT2_INODES_PER_GROUP(sb),
682			sbi->s_mount_opt);
683	return res;
684}
685
686static int ext2_check_descriptors(struct super_block *sb)
687{
688	int i;
689	struct ext2_sb_info *sbi = EXT2_SB(sb);
690
691	ext2_debug ("Checking group descriptors");
692
693	for (i = 0; i < sbi->s_groups_count; i++) {
694		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
695		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
696		ext2_fsblk_t last_block = ext2_group_last_block_no(sb, i);
697
698		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
699		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
700		{
701			ext2_error (sb, "ext2_check_descriptors",
702				    "Block bitmap for group %d"
703				    " not in group (block %lu)!",
704				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
705			return 0;
706		}
707		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
708		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
709		{
710			ext2_error (sb, "ext2_check_descriptors",
711				    "Inode bitmap for group %d"
712				    " not in group (block %lu)!",
713				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
714			return 0;
715		}
716		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
717		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
718		    last_block)
719		{
720			ext2_error (sb, "ext2_check_descriptors",
721				    "Inode table for group %d"
722				    " not in group (block %lu)!",
723				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
724			return 0;
725		}
726	}
727	return 1;
728}
729
730/*
731 * Maximal file size.  There is a direct, and {,double-,triple-}indirect
732 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
733 * We need to be 1 filesystem block less than the 2^32 sector limit.
734 */
735static loff_t ext2_max_size(int bits)
736{
737	loff_t res = EXT2_NDIR_BLOCKS;
738	int meta_blocks;
739	unsigned int upper_limit;
740	unsigned int ppb = 1 << (bits-2);
741
742	/* This is calculated to be the largest file size for a
743	 * dense, file such that the total number of
744	 * sectors in the file, including data and all indirect blocks,
745	 * does not exceed 2^32 -1
746	 * __u32 i_blocks representing the total number of
747	 * 512 bytes blocks of the file
748	 */
749	upper_limit = (1LL << 32) - 1;
750
751	/* total blocks in file system block size */
752	upper_limit >>= (bits - 9);
753
754	/* Compute how many blocks we can address by block tree */
755	res += 1LL << (bits-2);
756	res += 1LL << (2*(bits-2));
757	res += 1LL << (3*(bits-2));
758	/* Compute how many metadata blocks are needed */
759	meta_blocks = 1;
760	meta_blocks += 1 + ppb;
761	meta_blocks += 1 + ppb + ppb * ppb;
762	/* Does block tree limit file size? */
763	if (res + meta_blocks <= upper_limit)
764		goto check_lfs;
765
766	res = upper_limit;
767	/* How many metadata blocks are needed for addressing upper_limit? */
768	upper_limit -= EXT2_NDIR_BLOCKS;
769	/* indirect blocks */
770	meta_blocks = 1;
771	upper_limit -= ppb;
772	/* double indirect blocks */
773	if (upper_limit < ppb * ppb) {
774		meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
775		res -= meta_blocks;
776		goto check_lfs;
777	}
778	meta_blocks += 1 + ppb;
779	upper_limit -= ppb * ppb;
780	/* tripple indirect blocks for the rest */
781	meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
782		DIV_ROUND_UP(upper_limit, ppb*ppb);
783	res -= meta_blocks;
784check_lfs:
785	res <<= bits;
786	if (res > MAX_LFS_FILESIZE)
787		res = MAX_LFS_FILESIZE;
788
789	return res;
790}
791
792static unsigned long descriptor_loc(struct super_block *sb,
793				    unsigned long logic_sb_block,
794				    int nr)
795{
796	struct ext2_sb_info *sbi = EXT2_SB(sb);
797	unsigned long bg, first_meta_bg;
798
799	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
800
801	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
802	    nr < first_meta_bg)
803		return (logic_sb_block + nr + 1);
804	bg = sbi->s_desc_per_block * nr;
805
806	return ext2_group_first_block_no(sb, bg) + ext2_bg_has_super(sb, bg);
807}
808
809static int ext2_fill_super(struct super_block *sb, void *data, int silent)
810{
811	struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
812	struct buffer_head * bh;
813	struct ext2_sb_info * sbi;
814	struct ext2_super_block * es;
815	struct inode *root;
816	unsigned long block;
817	unsigned long sb_block = get_sb_block(&data);
818	unsigned long logic_sb_block;
819	unsigned long offset = 0;
820	unsigned long def_mount_opts;
821	long ret = -ENOMEM;
822	int blocksize = BLOCK_SIZE;
823	int db_count;
824	int i, j;
825	__le32 features;
826	int err;
827	struct ext2_mount_options opts;
828
829	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
830	if (!sbi)
831		goto failed;
832
833	sbi->s_blockgroup_lock =
834		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
835	if (!sbi->s_blockgroup_lock) {
836		kfree(sbi);
837		goto failed;
838	}
839	sb->s_fs_info = sbi;
840	sbi->s_sb_block = sb_block;
841	sbi->s_daxdev = dax_dev;
842
843	spin_lock_init(&sbi->s_lock);
844	ret = -EINVAL;
845
846	/*
847	 * See what the current blocksize for the device is, and
848	 * use that as the blocksize.  Otherwise (or if the blocksize
849	 * is smaller than the default) use the default.
850	 * This is important for devices that have a hardware
851	 * sectorsize that is larger than the default.
852	 */
853	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
854	if (!blocksize) {
855		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
856		goto failed_sbi;
857	}
858
859	/*
860	 * If the superblock doesn't start on a hardware sector boundary,
861	 * calculate the offset.
862	 */
863	if (blocksize != BLOCK_SIZE) {
864		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
865		offset = (sb_block*BLOCK_SIZE) % blocksize;
866	} else {
867		logic_sb_block = sb_block;
868	}
869
870	if (!(bh = sb_bread(sb, logic_sb_block))) {
871		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
872		goto failed_sbi;
873	}
874	/*
875	 * Note: s_es must be initialized as soon as possible because
876	 *       some ext2 macro-instructions depend on its value
877	 */
878	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
879	sbi->s_es = es;
880	sb->s_magic = le16_to_cpu(es->s_magic);
881
882	if (sb->s_magic != EXT2_SUPER_MAGIC)
883		goto cantfind_ext2;
884
885	opts.s_mount_opt = 0;
886	/* Set defaults before we parse the mount options */
887	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
888	if (def_mount_opts & EXT2_DEFM_DEBUG)
889		set_opt(opts.s_mount_opt, DEBUG);
890	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
891		set_opt(opts.s_mount_opt, GRPID);
892	if (def_mount_opts & EXT2_DEFM_UID16)
893		set_opt(opts.s_mount_opt, NO_UID32);
894#ifdef CONFIG_EXT2_FS_XATTR
895	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
896		set_opt(opts.s_mount_opt, XATTR_USER);
897#endif
898#ifdef CONFIG_EXT2_FS_POSIX_ACL
899	if (def_mount_opts & EXT2_DEFM_ACL)
900		set_opt(opts.s_mount_opt, POSIX_ACL);
901#endif
902
903	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
904		set_opt(opts.s_mount_opt, ERRORS_PANIC);
905	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
906		set_opt(opts.s_mount_opt, ERRORS_CONT);
907	else
908		set_opt(opts.s_mount_opt, ERRORS_RO);
909
910	opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
911	opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
912
913	set_opt(opts.s_mount_opt, RESERVATION);
914
915	if (!parse_options((char *) data, sb, &opts))
916		goto failed_mount;
917
918	sbi->s_mount_opt = opts.s_mount_opt;
919	sbi->s_resuid = opts.s_resuid;
920	sbi->s_resgid = opts.s_resgid;
921
922	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
923		(test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
924	sb->s_iflags |= SB_I_CGROUPWB;
925
926	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
927	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
928	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
929	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
930		ext2_msg(sb, KERN_WARNING,
931			"warning: feature flags set on rev 0 fs, "
932			"running e2fsck is recommended");
933	/*
934	 * Check feature flags regardless of the revision level, since we
935	 * previously didn't change the revision level when setting the flags,
936	 * so there is a chance incompat flags are set on a rev 0 filesystem.
937	 */
938	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
939	if (features) {
940		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
941		       "unsupported optional features (%x)",
942			le32_to_cpu(features));
943		goto failed_mount;
944	}
945	if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
946		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
947		       "unsupported optional features (%x)",
948		       le32_to_cpu(features));
949		goto failed_mount;
950	}
951
952	if (le32_to_cpu(es->s_log_block_size) >
953	    (EXT2_MAX_BLOCK_LOG_SIZE - BLOCK_SIZE_BITS)) {
954		ext2_msg(sb, KERN_ERR,
955			 "Invalid log block size: %u",
956			 le32_to_cpu(es->s_log_block_size));
957		goto failed_mount;
958	}
959	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
960
961	if (test_opt(sb, DAX)) {
962		if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
963			ext2_msg(sb, KERN_ERR,
964				"DAX unsupported by block device. Turning off DAX.");
965			clear_opt(sbi->s_mount_opt, DAX);
966		}
967	}
968
969	/* If the blocksize doesn't match, re-read the thing.. */
970	if (sb->s_blocksize != blocksize) {
971		brelse(bh);
972
973		if (!sb_set_blocksize(sb, blocksize)) {
974			ext2_msg(sb, KERN_ERR,
975				"error: bad blocksize %d", blocksize);
976			goto failed_sbi;
977		}
978
979		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
980		offset = (sb_block*BLOCK_SIZE) % blocksize;
981		bh = sb_bread(sb, logic_sb_block);
982		if(!bh) {
983			ext2_msg(sb, KERN_ERR, "error: couldn't read"
984				"superblock on 2nd try");
985			goto failed_sbi;
986		}
987		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
988		sbi->s_es = es;
989		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
990			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
991			goto failed_mount;
992		}
993	}
994
995	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
996	sb->s_max_links = EXT2_LINK_MAX;
997	sb->s_time_min = S32_MIN;
998	sb->s_time_max = S32_MAX;
999
1000	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
1001		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
1002		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
1003	} else {
1004		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1005		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1006		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1007		    !is_power_of_2(sbi->s_inode_size) ||
1008		    (sbi->s_inode_size > blocksize)) {
1009			ext2_msg(sb, KERN_ERR,
1010				"error: unsupported inode size: %d",
1011				sbi->s_inode_size);
1012			goto failed_mount;
1013		}
1014	}
1015
1016	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1017	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1018
1019	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1020	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1021		goto cantfind_ext2;
1022	sbi->s_itb_per_group = sbi->s_inodes_per_group /
1023					sbi->s_inodes_per_block;
1024	sbi->s_desc_per_block = sb->s_blocksize /
1025					sizeof (struct ext2_group_desc);
1026	sbi->s_sbh = bh;
1027	sbi->s_mount_state = le16_to_cpu(es->s_state);
1028	sbi->s_addr_per_block_bits =
1029		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1030	sbi->s_desc_per_block_bits =
1031		ilog2 (EXT2_DESC_PER_BLOCK(sb));
1032
1033	if (sb->s_magic != EXT2_SUPER_MAGIC)
1034		goto cantfind_ext2;
1035
1036	if (sb->s_blocksize != bh->b_size) {
1037		if (!silent)
1038			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1039		goto failed_mount;
1040	}
1041
1042	if (es->s_log_frag_size != es->s_log_block_size) {
1043		ext2_msg(sb, KERN_ERR,
1044			"error: fragsize log %u != blocksize log %u",
1045			le32_to_cpu(es->s_log_frag_size), sb->s_blocksize_bits);
1046		goto failed_mount;
1047	}
1048
1049	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1050		ext2_msg(sb, KERN_ERR,
1051			"error: #blocks per group too big: %lu",
1052			sbi->s_blocks_per_group);
1053		goto failed_mount;
1054	}
1055	if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
1056	    sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1057		ext2_msg(sb, KERN_ERR,
1058			"error: invalid #inodes per group: %lu",
1059			sbi->s_inodes_per_group);
1060		goto failed_mount;
1061	}
1062
1063	if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1064		goto cantfind_ext2;
1065	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1066				le32_to_cpu(es->s_first_data_block) - 1)
1067					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1068	if ((u64)sbi->s_groups_count * sbi->s_inodes_per_group !=
1069	    le32_to_cpu(es->s_inodes_count)) {
1070		ext2_msg(sb, KERN_ERR, "error: invalid #inodes: %u vs computed %llu",
1071			 le32_to_cpu(es->s_inodes_count),
1072			 (u64)sbi->s_groups_count * sbi->s_inodes_per_group);
1073		goto failed_mount;
1074	}
1075	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1076		   EXT2_DESC_PER_BLOCK(sb);
1077	sbi->s_group_desc = kmalloc_array (db_count,
1078					   sizeof(struct buffer_head *),
1079					   GFP_KERNEL);
1080	if (sbi->s_group_desc == NULL) {
1081		ret = -ENOMEM;
1082		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1083		goto failed_mount;
1084	}
1085	bgl_lock_init(sbi->s_blockgroup_lock);
1086	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1087	if (!sbi->s_debts) {
1088		ret = -ENOMEM;
1089		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1090		goto failed_mount_group_desc;
1091	}
1092	for (i = 0; i < db_count; i++) {
1093		block = descriptor_loc(sb, logic_sb_block, i);
1094		sbi->s_group_desc[i] = sb_bread(sb, block);
1095		if (!sbi->s_group_desc[i]) {
1096			for (j = 0; j < i; j++)
1097				brelse (sbi->s_group_desc[j]);
1098			ext2_msg(sb, KERN_ERR,
1099				"error: unable to read group descriptors");
1100			goto failed_mount_group_desc;
1101		}
1102	}
1103	if (!ext2_check_descriptors (sb)) {
1104		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1105		goto failed_mount2;
1106	}
1107	sbi->s_gdb_count = db_count;
1108	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1109	spin_lock_init(&sbi->s_next_gen_lock);
1110
1111	/* per fileystem reservation list head & lock */
1112	spin_lock_init(&sbi->s_rsv_window_lock);
1113	sbi->s_rsv_window_root = RB_ROOT;
1114	/*
1115	 * Add a single, static dummy reservation to the start of the
1116	 * reservation window list --- it gives us a placeholder for
1117	 * append-at-start-of-list which makes the allocation logic
1118	 * _much_ simpler.
1119	 */
1120	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1121	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1122	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1123	sbi->s_rsv_window_head.rsv_goal_size = 0;
1124	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1125
1126	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1127				ext2_count_free_blocks(sb), GFP_KERNEL);
1128	if (!err) {
1129		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1130				ext2_count_free_inodes(sb), GFP_KERNEL);
1131	}
1132	if (!err) {
1133		err = percpu_counter_init(&sbi->s_dirs_counter,
1134				ext2_count_dirs(sb), GFP_KERNEL);
1135	}
1136	if (err) {
1137		ret = err;
1138		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1139		goto failed_mount3;
1140	}
1141
1142#ifdef CONFIG_EXT2_FS_XATTR
1143	sbi->s_ea_block_cache = ext2_xattr_create_cache();
1144	if (!sbi->s_ea_block_cache) {
1145		ret = -ENOMEM;
1146		ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1147		goto failed_mount3;
1148	}
1149#endif
1150	/*
1151	 * set up enough so that it can read an inode
1152	 */
1153	sb->s_op = &ext2_sops;
1154	sb->s_export_op = &ext2_export_ops;
1155	sb->s_xattr = ext2_xattr_handlers;
1156
1157#ifdef CONFIG_QUOTA
1158	sb->dq_op = &dquot_operations;
1159	sb->s_qcop = &ext2_quotactl_ops;
1160	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1161#endif
1162
1163	root = ext2_iget(sb, EXT2_ROOT_INO);
1164	if (IS_ERR(root)) {
1165		ret = PTR_ERR(root);
1166		goto failed_mount3;
1167	}
1168	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1169		iput(root);
1170		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1171		goto failed_mount3;
1172	}
1173
1174	sb->s_root = d_make_root(root);
1175	if (!sb->s_root) {
1176		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1177		ret = -ENOMEM;
1178		goto failed_mount3;
1179	}
1180	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1181		ext2_msg(sb, KERN_WARNING,
1182			"warning: mounting ext3 filesystem as ext2");
1183	if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1184		sb->s_flags |= SB_RDONLY;
1185	ext2_write_super(sb);
1186	return 0;
1187
1188cantfind_ext2:
1189	if (!silent)
1190		ext2_msg(sb, KERN_ERR,
1191			"error: can't find an ext2 filesystem on dev %s.",
1192			sb->s_id);
1193	goto failed_mount;
1194failed_mount3:
1195	ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1196	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1197	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1198	percpu_counter_destroy(&sbi->s_dirs_counter);
1199failed_mount2:
1200	for (i = 0; i < db_count; i++)
1201		brelse(sbi->s_group_desc[i]);
1202failed_mount_group_desc:
1203	kfree(sbi->s_group_desc);
1204	kfree(sbi->s_debts);
1205failed_mount:
1206	brelse(bh);
1207failed_sbi:
1208	sb->s_fs_info = NULL;
1209	kfree(sbi->s_blockgroup_lock);
1210	kfree(sbi);
1211failed:
1212	fs_put_dax(dax_dev);
1213	return ret;
1214}
1215
1216static void ext2_clear_super_error(struct super_block *sb)
1217{
1218	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1219
1220	if (buffer_write_io_error(sbh)) {
1221		/*
1222		 * Oh, dear.  A previous attempt to write the
1223		 * superblock failed.  This could happen because the
1224		 * USB device was yanked out.  Or it could happen to
1225		 * be a transient write error and maybe the block will
1226		 * be remapped.  Nothing we can do but to retry the
1227		 * write and hope for the best.
1228		 */
1229		ext2_msg(sb, KERN_ERR,
1230		       "previous I/O error to superblock detected");
1231		clear_buffer_write_io_error(sbh);
1232		set_buffer_uptodate(sbh);
1233	}
1234}
1235
1236void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1237		     int wait)
1238{
1239	ext2_clear_super_error(sb);
1240	spin_lock(&EXT2_SB(sb)->s_lock);
1241	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1242	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1243	es->s_wtime = cpu_to_le32(ktime_get_real_seconds());
1244	/* unlock before we do IO */
1245	spin_unlock(&EXT2_SB(sb)->s_lock);
1246	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1247	if (wait)
1248		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1249}
1250
1251/*
1252 * In the second extended file system, it is not necessary to
1253 * write the super block since we use a mapping of the
1254 * disk super block in a buffer.
1255 *
1256 * However, this function is still used to set the fs valid
1257 * flags to 0.  We need to set this flag to 0 since the fs
1258 * may have been checked while mounted and e2fsck may have
1259 * set s_state to EXT2_VALID_FS after some corrections.
1260 */
1261static int ext2_sync_fs(struct super_block *sb, int wait)
1262{
1263	struct ext2_sb_info *sbi = EXT2_SB(sb);
1264	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1265
1266	/*
1267	 * Write quota structures to quota file, sync_blockdev() will write
1268	 * them to disk later
1269	 */
1270	dquot_writeback_dquots(sb, -1);
1271
1272	spin_lock(&sbi->s_lock);
1273	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1274		ext2_debug("setting valid to 0\n");
1275		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1276	}
1277	spin_unlock(&sbi->s_lock);
1278	ext2_sync_super(sb, es, wait);
1279	return 0;
1280}
1281
1282static int ext2_freeze(struct super_block *sb)
1283{
1284	struct ext2_sb_info *sbi = EXT2_SB(sb);
1285
1286	/*
1287	 * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1288	 * because we have unattached inodes and thus filesystem is not fully
1289	 * consistent.
1290	 */
1291	if (atomic_long_read(&sb->s_remove_count)) {
1292		ext2_sync_fs(sb, 1);
1293		return 0;
1294	}
1295	/* Set EXT2_FS_VALID flag */
1296	spin_lock(&sbi->s_lock);
1297	sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1298	spin_unlock(&sbi->s_lock);
1299	ext2_sync_super(sb, sbi->s_es, 1);
1300
1301	return 0;
1302}
1303
1304static int ext2_unfreeze(struct super_block *sb)
1305{
1306	/* Just write sb to clear EXT2_VALID_FS flag */
1307	ext2_write_super(sb);
1308
1309	return 0;
1310}
1311
1312static void ext2_write_super(struct super_block *sb)
1313{
1314	if (!sb_rdonly(sb))
1315		ext2_sync_fs(sb, 1);
1316}
1317
1318static int ext2_remount (struct super_block * sb, int * flags, char * data)
1319{
1320	struct ext2_sb_info * sbi = EXT2_SB(sb);
1321	struct ext2_super_block * es;
1322	struct ext2_mount_options new_opts;
1323	int err;
1324
1325	sync_filesystem(sb);
1326
1327	spin_lock(&sbi->s_lock);
1328	new_opts.s_mount_opt = sbi->s_mount_opt;
1329	new_opts.s_resuid = sbi->s_resuid;
1330	new_opts.s_resgid = sbi->s_resgid;
1331	spin_unlock(&sbi->s_lock);
1332
1333	if (!parse_options(data, sb, &new_opts))
1334		return -EINVAL;
1335
1336	spin_lock(&sbi->s_lock);
1337	es = sbi->s_es;
1338	if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1339		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1340			 "dax flag with busy inodes while remounting");
1341		new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1342	}
1343	if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1344		goto out_set;
1345	if (*flags & SB_RDONLY) {
1346		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1347		    !(sbi->s_mount_state & EXT2_VALID_FS))
1348			goto out_set;
1349
1350		/*
1351		 * OK, we are remounting a valid rw partition rdonly, so set
1352		 * the rdonly flag and then mark the partition as valid again.
1353		 */
1354		es->s_state = cpu_to_le16(sbi->s_mount_state);
1355		es->s_mtime = cpu_to_le32(ktime_get_real_seconds());
1356		spin_unlock(&sbi->s_lock);
1357
1358		err = dquot_suspend(sb, -1);
1359		if (err < 0)
1360			return err;
1361
1362		ext2_sync_super(sb, es, 1);
1363	} else {
1364		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1365					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1366		if (ret) {
1367			spin_unlock(&sbi->s_lock);
1368			ext2_msg(sb, KERN_WARNING,
1369				"warning: couldn't remount RDWR because of "
1370				"unsupported optional features (%x).",
1371				le32_to_cpu(ret));
1372			return -EROFS;
1373		}
1374		/*
1375		 * Mounting a RDONLY partition read-write, so reread and
1376		 * store the current valid flag.  (It may have been changed
1377		 * by e2fsck since we originally mounted the partition.)
1378		 */
1379		sbi->s_mount_state = le16_to_cpu(es->s_state);
1380		if (!ext2_setup_super (sb, es, 0))
1381			sb->s_flags &= ~SB_RDONLY;
1382		spin_unlock(&sbi->s_lock);
1383
1384		ext2_write_super(sb);
1385
1386		dquot_resume(sb, -1);
1387	}
1388
1389	spin_lock(&sbi->s_lock);
1390out_set:
1391	sbi->s_mount_opt = new_opts.s_mount_opt;
1392	sbi->s_resuid = new_opts.s_resuid;
1393	sbi->s_resgid = new_opts.s_resgid;
1394	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1395		(test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
1396	spin_unlock(&sbi->s_lock);
1397
1398	return 0;
1399}
1400
1401static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1402{
1403	struct super_block *sb = dentry->d_sb;
1404	struct ext2_sb_info *sbi = EXT2_SB(sb);
1405	struct ext2_super_block *es = sbi->s_es;
1406	u64 fsid;
1407
1408	spin_lock(&sbi->s_lock);
1409
1410	if (test_opt (sb, MINIX_DF))
1411		sbi->s_overhead_last = 0;
1412	else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1413		unsigned long i, overhead = 0;
1414		smp_rmb();
1415
1416		/*
1417		 * Compute the overhead (FS structures). This is constant
1418		 * for a given filesystem unless the number of block groups
1419		 * changes so we cache the previous value until it does.
1420		 */
1421
1422		/*
1423		 * All of the blocks before first_data_block are
1424		 * overhead
1425		 */
1426		overhead = le32_to_cpu(es->s_first_data_block);
1427
1428		/*
1429		 * Add the overhead attributed to the superblock and
1430		 * block group descriptors.  If the sparse superblocks
1431		 * feature is turned on, then not all groups have this.
1432		 */
1433		for (i = 0; i < sbi->s_groups_count; i++)
1434			overhead += ext2_bg_has_super(sb, i) +
1435				ext2_bg_num_gdb(sb, i);
1436
1437		/*
1438		 * Every block group has an inode bitmap, a block
1439		 * bitmap, and an inode table.
1440		 */
1441		overhead += (sbi->s_groups_count *
1442			     (2 + sbi->s_itb_per_group));
1443		sbi->s_overhead_last = overhead;
1444		smp_wmb();
1445		sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1446	}
1447
1448	buf->f_type = EXT2_SUPER_MAGIC;
1449	buf->f_bsize = sb->s_blocksize;
1450	buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1451	buf->f_bfree = ext2_count_free_blocks(sb);
1452	es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1453	buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1454	if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1455		buf->f_bavail = 0;
1456	buf->f_files = le32_to_cpu(es->s_inodes_count);
1457	buf->f_ffree = ext2_count_free_inodes(sb);
1458	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1459	buf->f_namelen = EXT2_NAME_LEN;
1460	fsid = le64_to_cpup((void *)es->s_uuid) ^
1461	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1462	buf->f_fsid = u64_to_fsid(fsid);
1463	spin_unlock(&sbi->s_lock);
1464	return 0;
1465}
1466
1467static struct dentry *ext2_mount(struct file_system_type *fs_type,
1468	int flags, const char *dev_name, void *data)
1469{
1470	return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1471}
1472
1473#ifdef CONFIG_QUOTA
1474
1475/* Read data from quotafile - avoid pagecache and such because we cannot afford
1476 * acquiring the locks... As quota files are never truncated and quota code
1477 * itself serializes the operations (and no one else should touch the files)
1478 * we don't have to be afraid of races */
1479static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1480			       size_t len, loff_t off)
1481{
1482	struct inode *inode = sb_dqopt(sb)->files[type];
1483	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1484	int err = 0;
1485	int offset = off & (sb->s_blocksize - 1);
1486	int tocopy;
1487	size_t toread;
1488	struct buffer_head tmp_bh;
1489	struct buffer_head *bh;
1490	loff_t i_size = i_size_read(inode);
1491
1492	if (off > i_size)
1493		return 0;
1494	if (off+len > i_size)
1495		len = i_size-off;
1496	toread = len;
1497	while (toread > 0) {
1498		tocopy = sb->s_blocksize - offset < toread ?
1499				sb->s_blocksize - offset : toread;
1500
1501		tmp_bh.b_state = 0;
1502		tmp_bh.b_size = sb->s_blocksize;
1503		err = ext2_get_block(inode, blk, &tmp_bh, 0);
1504		if (err < 0)
1505			return err;
1506		if (!buffer_mapped(&tmp_bh))	/* A hole? */
1507			memset(data, 0, tocopy);
1508		else {
1509			bh = sb_bread(sb, tmp_bh.b_blocknr);
1510			if (!bh)
1511				return -EIO;
1512			memcpy(data, bh->b_data+offset, tocopy);
1513			brelse(bh);
1514		}
1515		offset = 0;
1516		toread -= tocopy;
1517		data += tocopy;
1518		blk++;
1519	}
1520	return len;
1521}
1522
1523/* Write to quotafile */
1524static ssize_t ext2_quota_write(struct super_block *sb, int type,
1525				const char *data, size_t len, loff_t off)
1526{
1527	struct inode *inode = sb_dqopt(sb)->files[type];
1528	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1529	int err = 0;
1530	int offset = off & (sb->s_blocksize - 1);
1531	int tocopy;
1532	size_t towrite = len;
1533	struct buffer_head tmp_bh;
1534	struct buffer_head *bh;
1535
1536	while (towrite > 0) {
1537		tocopy = sb->s_blocksize - offset < towrite ?
1538				sb->s_blocksize - offset : towrite;
1539
1540		tmp_bh.b_state = 0;
1541		tmp_bh.b_size = sb->s_blocksize;
1542		err = ext2_get_block(inode, blk, &tmp_bh, 1);
1543		if (err < 0)
1544			goto out;
1545		if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1546			bh = sb_bread(sb, tmp_bh.b_blocknr);
1547		else
1548			bh = sb_getblk(sb, tmp_bh.b_blocknr);
1549		if (unlikely(!bh)) {
1550			err = -EIO;
1551			goto out;
1552		}
1553		lock_buffer(bh);
1554		memcpy(bh->b_data+offset, data, tocopy);
1555		flush_dcache_page(bh->b_page);
1556		set_buffer_uptodate(bh);
1557		mark_buffer_dirty(bh);
1558		unlock_buffer(bh);
1559		brelse(bh);
1560		offset = 0;
1561		towrite -= tocopy;
1562		data += tocopy;
1563		blk++;
1564	}
1565out:
1566	if (len == towrite)
1567		return err;
1568	if (inode->i_size < off+len-towrite)
1569		i_size_write(inode, off+len-towrite);
1570	inode_inc_iversion(inode);
1571	inode->i_mtime = inode->i_ctime = current_time(inode);
1572	mark_inode_dirty(inode);
1573	return len - towrite;
1574}
1575
1576static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1577			 const struct path *path)
1578{
1579	int err;
1580	struct inode *inode;
1581
1582	err = dquot_quota_on(sb, type, format_id, path);
1583	if (err)
1584		return err;
1585
1586	inode = d_inode(path->dentry);
1587	inode_lock(inode);
1588	EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1589	inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1590			S_NOATIME | S_IMMUTABLE);
1591	inode_unlock(inode);
1592	mark_inode_dirty(inode);
1593
1594	return 0;
1595}
1596
1597static int ext2_quota_off(struct super_block *sb, int type)
1598{
1599	struct inode *inode = sb_dqopt(sb)->files[type];
1600	int err;
1601
1602	if (!inode || !igrab(inode))
1603		goto out;
1604
1605	err = dquot_quota_off(sb, type);
1606	if (err)
1607		goto out_put;
1608
1609	inode_lock(inode);
1610	EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1611	inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1612	inode_unlock(inode);
1613	mark_inode_dirty(inode);
1614out_put:
1615	iput(inode);
1616	return err;
1617out:
1618	return dquot_quota_off(sb, type);
1619}
1620
1621#endif
1622
1623static struct file_system_type ext2_fs_type = {
1624	.owner		= THIS_MODULE,
1625	.name		= "ext2",
1626	.mount		= ext2_mount,
1627	.kill_sb	= kill_block_super,
1628	.fs_flags	= FS_REQUIRES_DEV,
1629};
1630MODULE_ALIAS_FS("ext2");
1631
1632static int __init init_ext2_fs(void)
1633{
1634	int err;
1635
1636	err = init_inodecache();
1637	if (err)
1638		return err;
1639        err = register_filesystem(&ext2_fs_type);
1640	if (err)
1641		goto out;
1642	return 0;
1643out:
1644	destroy_inodecache();
1645	return err;
1646}
1647
1648static void __exit exit_ext2_fs(void)
1649{
1650	unregister_filesystem(&ext2_fs_type);
1651	destroy_inodecache();
1652}
1653
1654MODULE_AUTHOR("Remy Card and others");
1655MODULE_DESCRIPTION("Second Extended Filesystem");
1656MODULE_LICENSE("GPL");
1657module_init(init_ext2_fs)
1658module_exit(exit_ext2_fs)
1659