GNU Linux-libre 5.10.219-gnu1
[releases.git] / fs / ext2 / super.c
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
42 static void ext2_write_super(struct super_block *sb);
43 static int ext2_remount (struct super_block * sb, int * flags, char * data);
44 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
45 static int ext2_sync_fs(struct super_block *sb, int wait);
46 static int ext2_freeze(struct super_block *sb);
47 static int ext2_unfreeze(struct super_block *sb);
48
49 void 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
84 void 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  */
103 void 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
129 static int ext2_quota_off(struct super_block *sb, int type);
130
131 static 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
139 static inline void ext2_quota_off_umount(struct super_block *sb)
140 {
141 }
142 #endif
143
144 static 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
178 static struct kmem_cache * ext2_inode_cachep;
179
180 static 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
195 static void ext2_free_in_core_inode(struct inode *inode)
196 {
197         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
198 }
199
200 static 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
215 static 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
229 static 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
239 static 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
325 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
326 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
327 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
328                          const struct path *path);
329 static struct dquot **ext2_get_dquots(struct inode *inode)
330 {
331         return EXT2_I(inode)->i_dquot;
332 }
333
334 static 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
346 static 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
365 static 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
391 static 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
398 static 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
405 static 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
411 static 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
431 enum {
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
440 static 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
473 static 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
636 static 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
686 static 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  */
735 static 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;
784 check_lfs:
785         res <<= bits;
786         if (res > MAX_LFS_FILESIZE)
787                 res = MAX_LFS_FILESIZE;
788
789         return res;
790 }
791
792 static 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
809 static 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
1188 cantfind_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;
1194 failed_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);
1199 failed_mount2:
1200         for (i = 0; i < db_count; i++)
1201                 brelse(sbi->s_group_desc[i]);
1202 failed_mount_group_desc:
1203         kfree(sbi->s_group_desc);
1204         kfree(sbi->s_debts);
1205 failed_mount:
1206         brelse(bh);
1207 failed_sbi:
1208         sb->s_fs_info = NULL;
1209         kfree(sbi->s_blockgroup_lock);
1210         kfree(sbi);
1211 failed:
1212         fs_put_dax(dax_dev);
1213         return ret;
1214 }
1215
1216 static 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
1236 void 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  */
1261 static 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
1282 static 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
1304 static 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
1312 static void ext2_write_super(struct super_block *sb)
1313 {
1314         if (!sb_rdonly(sb))
1315                 ext2_sync_fs(sb, 1);
1316 }
1317
1318 static 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);
1390 out_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
1401 static 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
1467 static 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 */
1479 static 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 */
1524 static 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         }
1565 out:
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
1576 static 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
1597 static 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);
1614 out_put:
1615         iput(inode);
1616         return err;
1617 out:
1618         return dquot_quota_off(sb, type);
1619 }
1620
1621 #endif
1622
1623 static 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 };
1630 MODULE_ALIAS_FS("ext2");
1631
1632 static 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;
1643 out:
1644         destroy_inodecache();
1645         return err;
1646 }
1647
1648 static void __exit exit_ext2_fs(void)
1649 {
1650         unregister_filesystem(&ext2_fs_type);
1651         destroy_inodecache();
1652 }
1653
1654 MODULE_AUTHOR("Remy Card and others");
1655 MODULE_DESCRIPTION("Second Extended Filesystem");
1656 MODULE_LICENSE("GPL");
1657 module_init(init_ext2_fs)
1658 module_exit(exit_ext2_fs)