1458706bd2ec6ac5ed9c0acb6c150d661bd14595
[sfrench/cifs-2.6.git] / fs / ext2 / super.c
1 /*
2  *  linux/fs/ext2/super.c
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/fs/minix/inode.c
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  *
15  *  Big-endian to little-endian byte-swapping/bitmaps by
16  *        David S. Miller (davem@caip.rutgers.edu), 1995
17  */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/slab.h>
23 #include <linux/init.h>
24 #include <linux/blkdev.h>
25 #include <linux/parser.h>
26 #include <linux/random.h>
27 #include <linux/buffer_head.h>
28 #include <linux/exportfs.h>
29 #include <linux/vfs.h>
30 #include <linux/seq_file.h>
31 #include <linux/mount.h>
32 #include <linux/log2.h>
33 #include <linux/quotaops.h>
34 #include <linux/uaccess.h>
35 #include <linux/dax.h>
36 #include "ext2.h"
37 #include "xattr.h"
38 #include "acl.h"
39
40 static void ext2_write_super(struct super_block *sb);
41 static int ext2_remount (struct super_block * sb, int * flags, char * data);
42 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
43 static int ext2_sync_fs(struct super_block *sb, int wait);
44 static int ext2_freeze(struct super_block *sb);
45 static int ext2_unfreeze(struct super_block *sb);
46
47 void ext2_error(struct super_block *sb, const char *function,
48                 const char *fmt, ...)
49 {
50         struct va_format vaf;
51         va_list args;
52         struct ext2_sb_info *sbi = EXT2_SB(sb);
53         struct ext2_super_block *es = sbi->s_es;
54
55         if (!sb_rdonly(sb)) {
56                 spin_lock(&sbi->s_lock);
57                 sbi->s_mount_state |= EXT2_ERROR_FS;
58                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
59                 spin_unlock(&sbi->s_lock);
60                 ext2_sync_super(sb, es, 1);
61         }
62
63         va_start(args, fmt);
64
65         vaf.fmt = fmt;
66         vaf.va = &args;
67
68         printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
69                sb->s_id, function, &vaf);
70
71         va_end(args);
72
73         if (test_opt(sb, ERRORS_PANIC))
74                 panic("EXT2-fs: panic from previous error\n");
75         if (test_opt(sb, ERRORS_RO)) {
76                 ext2_msg(sb, KERN_CRIT,
77                              "error: remounting filesystem read-only");
78                 sb->s_flags |= MS_RDONLY;
79         }
80 }
81
82 void ext2_msg(struct super_block *sb, const char *prefix,
83                 const char *fmt, ...)
84 {
85         struct va_format vaf;
86         va_list args;
87
88         va_start(args, fmt);
89
90         vaf.fmt = fmt;
91         vaf.va = &args;
92
93         printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
94
95         va_end(args);
96 }
97
98 /*
99  * This must be called with sbi->s_lock held.
100  */
101 void ext2_update_dynamic_rev(struct super_block *sb)
102 {
103         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
104
105         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
106                 return;
107
108         ext2_msg(sb, KERN_WARNING,
109                      "warning: updating to rev %d because of "
110                      "new feature flag, running e2fsck is recommended",
111                      EXT2_DYNAMIC_REV);
112
113         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
114         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
115         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
116         /* leave es->s_feature_*compat flags alone */
117         /* es->s_uuid will be set by e2fsck if empty */
118
119         /*
120          * The rest of the superblock fields should be zero, and if not it
121          * means they are likely already in use, so leave them alone.  We
122          * can leave it up to e2fsck to clean up any inconsistencies there.
123          */
124 }
125
126 #ifdef CONFIG_QUOTA
127 static int ext2_quota_off(struct super_block *sb, int type);
128
129 static void ext2_quota_off_umount(struct super_block *sb)
130 {
131         int type;
132
133         for (type = 0; type < MAXQUOTAS; type++)
134                 ext2_quota_off(sb, type);
135 }
136 #else
137 static inline void ext2_quota_off_umount(struct super_block *sb)
138 {
139 }
140 #endif
141
142 static void ext2_put_super (struct super_block * sb)
143 {
144         int db_count;
145         int i;
146         struct ext2_sb_info *sbi = EXT2_SB(sb);
147
148         ext2_quota_off_umount(sb);
149
150         if (sbi->s_ea_block_cache) {
151                 ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
152                 sbi->s_ea_block_cache = NULL;
153         }
154         if (!sb_rdonly(sb)) {
155                 struct ext2_super_block *es = sbi->s_es;
156
157                 spin_lock(&sbi->s_lock);
158                 es->s_state = cpu_to_le16(sbi->s_mount_state);
159                 spin_unlock(&sbi->s_lock);
160                 ext2_sync_super(sb, es, 1);
161         }
162         db_count = sbi->s_gdb_count;
163         for (i = 0; i < db_count; i++)
164                 if (sbi->s_group_desc[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         ei->vfs_inode.i_version = 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_i_callback(struct rcu_head *head)
196 {
197         struct inode *inode = container_of(head, struct inode, i_rcu);
198         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
199 }
200
201 static void ext2_destroy_inode(struct inode *inode)
202 {
203         call_rcu(&inode->i_rcu, ext2_i_callback);
204 }
205
206 static void init_once(void *foo)
207 {
208         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
209
210         rwlock_init(&ei->i_meta_lock);
211 #ifdef CONFIG_EXT2_FS_XATTR
212         init_rwsem(&ei->xattr_sem);
213 #endif
214         mutex_init(&ei->truncate_mutex);
215 #ifdef CONFIG_FS_DAX
216         init_rwsem(&ei->dax_sem);
217 #endif
218         inode_init_once(&ei->vfs_inode);
219 }
220
221 static int __init init_inodecache(void)
222 {
223         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
224                                              sizeof(struct ext2_inode_info),
225                                              0, (SLAB_RECLAIM_ACCOUNT|
226                                                 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
227                                              init_once);
228         if (ext2_inode_cachep == NULL)
229                 return -ENOMEM;
230         return 0;
231 }
232
233 static void destroy_inodecache(void)
234 {
235         /*
236          * Make sure all delayed rcu free inodes are flushed before we
237          * destroy cache.
238          */
239         rcu_barrier();
240         kmem_cache_destroy(ext2_inode_cachep);
241 }
242
243 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
244 {
245         struct super_block *sb = root->d_sb;
246         struct ext2_sb_info *sbi = EXT2_SB(sb);
247         struct ext2_super_block *es = sbi->s_es;
248         unsigned long def_mount_opts;
249
250         spin_lock(&sbi->s_lock);
251         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
252
253         if (sbi->s_sb_block != 1)
254                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
255         if (test_opt(sb, MINIX_DF))
256                 seq_puts(seq, ",minixdf");
257         if (test_opt(sb, GRPID))
258                 seq_puts(seq, ",grpid");
259         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
260                 seq_puts(seq, ",nogrpid");
261         if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
262             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
263                 seq_printf(seq, ",resuid=%u",
264                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
265         }
266         if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
267             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
268                 seq_printf(seq, ",resgid=%u",
269                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
270         }
271         if (test_opt(sb, ERRORS_RO)) {
272                 int def_errors = le16_to_cpu(es->s_errors);
273
274                 if (def_errors == EXT2_ERRORS_PANIC ||
275                     def_errors == EXT2_ERRORS_CONTINUE) {
276                         seq_puts(seq, ",errors=remount-ro");
277                 }
278         }
279         if (test_opt(sb, ERRORS_CONT))
280                 seq_puts(seq, ",errors=continue");
281         if (test_opt(sb, ERRORS_PANIC))
282                 seq_puts(seq, ",errors=panic");
283         if (test_opt(sb, NO_UID32))
284                 seq_puts(seq, ",nouid32");
285         if (test_opt(sb, DEBUG))
286                 seq_puts(seq, ",debug");
287         if (test_opt(sb, OLDALLOC))
288                 seq_puts(seq, ",oldalloc");
289
290 #ifdef CONFIG_EXT2_FS_XATTR
291         if (test_opt(sb, XATTR_USER))
292                 seq_puts(seq, ",user_xattr");
293         if (!test_opt(sb, XATTR_USER) &&
294             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
295                 seq_puts(seq, ",nouser_xattr");
296         }
297 #endif
298
299 #ifdef CONFIG_EXT2_FS_POSIX_ACL
300         if (test_opt(sb, POSIX_ACL))
301                 seq_puts(seq, ",acl");
302         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
303                 seq_puts(seq, ",noacl");
304 #endif
305
306         if (test_opt(sb, NOBH))
307                 seq_puts(seq, ",nobh");
308
309 #if defined(CONFIG_QUOTA)
310         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
311                 seq_puts(seq, ",usrquota");
312
313         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
314                 seq_puts(seq, ",grpquota");
315 #endif
316
317 #ifdef CONFIG_FS_DAX
318         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
319                 seq_puts(seq, ",xip");
320         if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
321                 seq_puts(seq, ",dax");
322 #endif
323
324         if (!test_opt(sb, RESERVATION))
325                 seq_puts(seq, ",noreservation");
326
327         spin_unlock(&sbi->s_lock);
328         return 0;
329 }
330
331 #ifdef CONFIG_QUOTA
332 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
333 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
334 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
335                          const struct path *path);
336 static struct dquot **ext2_get_dquots(struct inode *inode)
337 {
338         return EXT2_I(inode)->i_dquot;
339 }
340
341 static const struct quotactl_ops ext2_quotactl_ops = {
342         .quota_on       = ext2_quota_on,
343         .quota_off      = ext2_quota_off,
344         .quota_sync     = dquot_quota_sync,
345         .get_state      = dquot_get_state,
346         .set_info       = dquot_set_dqinfo,
347         .get_dqblk      = dquot_get_dqblk,
348         .set_dqblk      = dquot_set_dqblk,
349         .get_nextdqblk  = dquot_get_next_dqblk,
350 };
351 #endif
352
353 static const struct super_operations ext2_sops = {
354         .alloc_inode    = ext2_alloc_inode,
355         .destroy_inode  = ext2_destroy_inode,
356         .write_inode    = ext2_write_inode,
357         .evict_inode    = ext2_evict_inode,
358         .put_super      = ext2_put_super,
359         .sync_fs        = ext2_sync_fs,
360         .freeze_fs      = ext2_freeze,
361         .unfreeze_fs    = ext2_unfreeze,
362         .statfs         = ext2_statfs,
363         .remount_fs     = ext2_remount,
364         .show_options   = ext2_show_options,
365 #ifdef CONFIG_QUOTA
366         .quota_read     = ext2_quota_read,
367         .quota_write    = ext2_quota_write,
368         .get_dquots     = ext2_get_dquots,
369 #endif
370 };
371
372 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
373                 u64 ino, u32 generation)
374 {
375         struct inode *inode;
376
377         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
378                 return ERR_PTR(-ESTALE);
379         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
380                 return ERR_PTR(-ESTALE);
381
382         /*
383          * ext2_iget isn't quite right if the inode is currently unallocated!
384          * However ext2_iget currently does appropriate checks to handle stale
385          * inodes so everything is OK.
386          */
387         inode = ext2_iget(sb, ino);
388         if (IS_ERR(inode))
389                 return ERR_CAST(inode);
390         if (generation && inode->i_generation != generation) {
391                 /* we didn't find the right inode.. */
392                 iput(inode);
393                 return ERR_PTR(-ESTALE);
394         }
395         return inode;
396 }
397
398 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
399                 int fh_len, int fh_type)
400 {
401         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
402                                     ext2_nfs_get_inode);
403 }
404
405 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
406                 int fh_len, int fh_type)
407 {
408         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
409                                     ext2_nfs_get_inode);
410 }
411
412 static const struct export_operations ext2_export_ops = {
413         .fh_to_dentry = ext2_fh_to_dentry,
414         .fh_to_parent = ext2_fh_to_parent,
415         .get_parent = ext2_get_parent,
416 };
417
418 static unsigned long get_sb_block(void **data)
419 {
420         unsigned long   sb_block;
421         char            *options = (char *) *data;
422
423         if (!options || strncmp(options, "sb=", 3) != 0)
424                 return 1;       /* Default location */
425         options += 3;
426         sb_block = simple_strtoul(options, &options, 0);
427         if (*options && *options != ',') {
428                 printk("EXT2-fs: Invalid sb specification: %s\n",
429                        (char *) *data);
430                 return 1;
431         }
432         if (*options == ',')
433                 options++;
434         *data = (void *) options;
435         return sb_block;
436 }
437
438 enum {
439         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
440         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
441         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
442         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
443         Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
444         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
445 };
446
447 static const match_table_t tokens = {
448         {Opt_bsd_df, "bsddf"},
449         {Opt_minix_df, "minixdf"},
450         {Opt_grpid, "grpid"},
451         {Opt_grpid, "bsdgroups"},
452         {Opt_nogrpid, "nogrpid"},
453         {Opt_nogrpid, "sysvgroups"},
454         {Opt_resgid, "resgid=%u"},
455         {Opt_resuid, "resuid=%u"},
456         {Opt_sb, "sb=%u"},
457         {Opt_err_cont, "errors=continue"},
458         {Opt_err_panic, "errors=panic"},
459         {Opt_err_ro, "errors=remount-ro"},
460         {Opt_nouid32, "nouid32"},
461         {Opt_nocheck, "check=none"},
462         {Opt_nocheck, "nocheck"},
463         {Opt_debug, "debug"},
464         {Opt_oldalloc, "oldalloc"},
465         {Opt_orlov, "orlov"},
466         {Opt_nobh, "nobh"},
467         {Opt_user_xattr, "user_xattr"},
468         {Opt_nouser_xattr, "nouser_xattr"},
469         {Opt_acl, "acl"},
470         {Opt_noacl, "noacl"},
471         {Opt_xip, "xip"},
472         {Opt_dax, "dax"},
473         {Opt_grpquota, "grpquota"},
474         {Opt_ignore, "noquota"},
475         {Opt_quota, "quota"},
476         {Opt_usrquota, "usrquota"},
477         {Opt_reservation, "reservation"},
478         {Opt_noreservation, "noreservation"},
479         {Opt_err, NULL}
480 };
481
482 static int parse_options(char *options, struct super_block *sb)
483 {
484         char *p;
485         struct ext2_sb_info *sbi = EXT2_SB(sb);
486         substring_t args[MAX_OPT_ARGS];
487         int option;
488         kuid_t uid;
489         kgid_t gid;
490
491         if (!options)
492                 return 1;
493
494         while ((p = strsep (&options, ",")) != NULL) {
495                 int token;
496                 if (!*p)
497                         continue;
498
499                 token = match_token(p, tokens, args);
500                 switch (token) {
501                 case Opt_bsd_df:
502                         clear_opt (sbi->s_mount_opt, MINIX_DF);
503                         break;
504                 case Opt_minix_df:
505                         set_opt (sbi->s_mount_opt, MINIX_DF);
506                         break;
507                 case Opt_grpid:
508                         set_opt (sbi->s_mount_opt, GRPID);
509                         break;
510                 case Opt_nogrpid:
511                         clear_opt (sbi->s_mount_opt, GRPID);
512                         break;
513                 case Opt_resuid:
514                         if (match_int(&args[0], &option))
515                                 return 0;
516                         uid = make_kuid(current_user_ns(), option);
517                         if (!uid_valid(uid)) {
518                                 ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
519                                 return 0;
520
521                         }
522                         sbi->s_resuid = uid;
523                         break;
524                 case Opt_resgid:
525                         if (match_int(&args[0], &option))
526                                 return 0;
527                         gid = make_kgid(current_user_ns(), option);
528                         if (!gid_valid(gid)) {
529                                 ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
530                                 return 0;
531                         }
532                         sbi->s_resgid = gid;
533                         break;
534                 case Opt_sb:
535                         /* handled by get_sb_block() instead of here */
536                         /* *sb_block = match_int(&args[0]); */
537                         break;
538                 case Opt_err_panic:
539                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
540                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
541                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
542                         break;
543                 case Opt_err_ro:
544                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
545                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
546                         set_opt (sbi->s_mount_opt, ERRORS_RO);
547                         break;
548                 case Opt_err_cont:
549                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
550                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
551                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
552                         break;
553                 case Opt_nouid32:
554                         set_opt (sbi->s_mount_opt, NO_UID32);
555                         break;
556                 case Opt_nocheck:
557                         clear_opt (sbi->s_mount_opt, CHECK);
558                         break;
559                 case Opt_debug:
560                         set_opt (sbi->s_mount_opt, DEBUG);
561                         break;
562                 case Opt_oldalloc:
563                         set_opt (sbi->s_mount_opt, OLDALLOC);
564                         break;
565                 case Opt_orlov:
566                         clear_opt (sbi->s_mount_opt, OLDALLOC);
567                         break;
568                 case Opt_nobh:
569                         set_opt (sbi->s_mount_opt, NOBH);
570                         break;
571 #ifdef CONFIG_EXT2_FS_XATTR
572                 case Opt_user_xattr:
573                         set_opt (sbi->s_mount_opt, XATTR_USER);
574                         break;
575                 case Opt_nouser_xattr:
576                         clear_opt (sbi->s_mount_opt, XATTR_USER);
577                         break;
578 #else
579                 case Opt_user_xattr:
580                 case Opt_nouser_xattr:
581                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
582                                 "not supported");
583                         break;
584 #endif
585 #ifdef CONFIG_EXT2_FS_POSIX_ACL
586                 case Opt_acl:
587                         set_opt(sbi->s_mount_opt, POSIX_ACL);
588                         break;
589                 case Opt_noacl:
590                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
591                         break;
592 #else
593                 case Opt_acl:
594                 case Opt_noacl:
595                         ext2_msg(sb, KERN_INFO,
596                                 "(no)acl options not supported");
597                         break;
598 #endif
599                 case Opt_xip:
600                         ext2_msg(sb, KERN_INFO, "use dax instead of xip");
601                         set_opt(sbi->s_mount_opt, XIP);
602                         /* Fall through */
603                 case Opt_dax:
604 #ifdef CONFIG_FS_DAX
605                         ext2_msg(sb, KERN_WARNING,
606                 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
607                         set_opt(sbi->s_mount_opt, DAX);
608 #else
609                         ext2_msg(sb, KERN_INFO, "dax option not supported");
610 #endif
611                         break;
612
613 #if defined(CONFIG_QUOTA)
614                 case Opt_quota:
615                 case Opt_usrquota:
616                         set_opt(sbi->s_mount_opt, USRQUOTA);
617                         break;
618
619                 case Opt_grpquota:
620                         set_opt(sbi->s_mount_opt, GRPQUOTA);
621                         break;
622 #else
623                 case Opt_quota:
624                 case Opt_usrquota:
625                 case Opt_grpquota:
626                         ext2_msg(sb, KERN_INFO,
627                                 "quota operations not supported");
628                         break;
629 #endif
630
631                 case Opt_reservation:
632                         set_opt(sbi->s_mount_opt, RESERVATION);
633                         ext2_msg(sb, KERN_INFO, "reservations ON");
634                         break;
635                 case Opt_noreservation:
636                         clear_opt(sbi->s_mount_opt, RESERVATION);
637                         ext2_msg(sb, KERN_INFO, "reservations OFF");
638                         break;
639                 case Opt_ignore:
640                         break;
641                 default:
642                         return 0;
643                 }
644         }
645         return 1;
646 }
647
648 static int ext2_setup_super (struct super_block * sb,
649                               struct ext2_super_block * es,
650                               int read_only)
651 {
652         int res = 0;
653         struct ext2_sb_info *sbi = EXT2_SB(sb);
654
655         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
656                 ext2_msg(sb, KERN_ERR,
657                         "error: revision level too high, "
658                         "forcing read-only mode");
659                 res = MS_RDONLY;
660         }
661         if (read_only)
662                 return res;
663         if (!(sbi->s_mount_state & EXT2_VALID_FS))
664                 ext2_msg(sb, KERN_WARNING,
665                         "warning: mounting unchecked fs, "
666                         "running e2fsck is recommended");
667         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
668                 ext2_msg(sb, KERN_WARNING,
669                         "warning: mounting fs with errors, "
670                         "running e2fsck is recommended");
671         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
672                  le16_to_cpu(es->s_mnt_count) >=
673                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
674                 ext2_msg(sb, KERN_WARNING,
675                         "warning: maximal mount count reached, "
676                         "running e2fsck is recommended");
677         else if (le32_to_cpu(es->s_checkinterval) &&
678                 (le32_to_cpu(es->s_lastcheck) +
679                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
680                 ext2_msg(sb, KERN_WARNING,
681                         "warning: checktime reached, "
682                         "running e2fsck is recommended");
683         if (!le16_to_cpu(es->s_max_mnt_count))
684                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
685         le16_add_cpu(&es->s_mnt_count, 1);
686         if (test_opt (sb, DEBUG))
687                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
688                         "bpg=%lu, ipg=%lu, mo=%04lx]",
689                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
690                         sbi->s_frag_size,
691                         sbi->s_groups_count,
692                         EXT2_BLOCKS_PER_GROUP(sb),
693                         EXT2_INODES_PER_GROUP(sb),
694                         sbi->s_mount_opt);
695         return res;
696 }
697
698 static int ext2_check_descriptors(struct super_block *sb)
699 {
700         int i;
701         struct ext2_sb_info *sbi = EXT2_SB(sb);
702
703         ext2_debug ("Checking group descriptors");
704
705         for (i = 0; i < sbi->s_groups_count; i++) {
706                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
707                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
708                 ext2_fsblk_t last_block;
709
710                 if (i == sbi->s_groups_count - 1)
711                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
712                 else
713                         last_block = first_block +
714                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
715
716                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
717                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
718                 {
719                         ext2_error (sb, "ext2_check_descriptors",
720                                     "Block bitmap for group %d"
721                                     " not in group (block %lu)!",
722                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
723                         return 0;
724                 }
725                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
726                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
727                 {
728                         ext2_error (sb, "ext2_check_descriptors",
729                                     "Inode bitmap for group %d"
730                                     " not in group (block %lu)!",
731                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
732                         return 0;
733                 }
734                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
735                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
736                     last_block)
737                 {
738                         ext2_error (sb, "ext2_check_descriptors",
739                                     "Inode table for group %d"
740                                     " not in group (block %lu)!",
741                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
742                         return 0;
743                 }
744         }
745         return 1;
746 }
747
748 /*
749  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
750  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
751  * We need to be 1 filesystem block less than the 2^32 sector limit.
752  */
753 static loff_t ext2_max_size(int bits)
754 {
755         loff_t res = EXT2_NDIR_BLOCKS;
756         int meta_blocks;
757         loff_t upper_limit;
758
759         /* This is calculated to be the largest file size for a
760          * dense, file such that the total number of
761          * sectors in the file, including data and all indirect blocks,
762          * does not exceed 2^32 -1
763          * __u32 i_blocks representing the total number of
764          * 512 bytes blocks of the file
765          */
766         upper_limit = (1LL << 32) - 1;
767
768         /* total blocks in file system block size */
769         upper_limit >>= (bits - 9);
770
771
772         /* indirect blocks */
773         meta_blocks = 1;
774         /* double indirect blocks */
775         meta_blocks += 1 + (1LL << (bits-2));
776         /* tripple indirect blocks */
777         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
778
779         upper_limit -= meta_blocks;
780         upper_limit <<= bits;
781
782         res += 1LL << (bits-2);
783         res += 1LL << (2*(bits-2));
784         res += 1LL << (3*(bits-2));
785         res <<= bits;
786         if (res > upper_limit)
787                 res = upper_limit;
788
789         if (res > MAX_LFS_FILESIZE)
790                 res = MAX_LFS_FILESIZE;
791
792         return res;
793 }
794
795 static unsigned long descriptor_loc(struct super_block *sb,
796                                     unsigned long logic_sb_block,
797                                     int nr)
798 {
799         struct ext2_sb_info *sbi = EXT2_SB(sb);
800         unsigned long bg, first_meta_bg;
801         int has_super = 0;
802         
803         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
804
805         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
806             nr < first_meta_bg)
807                 return (logic_sb_block + nr + 1);
808         bg = sbi->s_desc_per_block * nr;
809         if (ext2_bg_has_super(sb, bg))
810                 has_super = 1;
811
812         return ext2_group_first_block_no(sb, bg) + has_super;
813 }
814
815 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
816 {
817         struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
818         struct buffer_head * bh;
819         struct ext2_sb_info * sbi;
820         struct ext2_super_block * es;
821         struct inode *root;
822         unsigned long block;
823         unsigned long sb_block = get_sb_block(&data);
824         unsigned long logic_sb_block;
825         unsigned long offset = 0;
826         unsigned long def_mount_opts;
827         long ret = -EINVAL;
828         int blocksize = BLOCK_SIZE;
829         int db_count;
830         int i, j;
831         __le32 features;
832         int err;
833
834         err = -ENOMEM;
835         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
836         if (!sbi)
837                 goto failed;
838
839         sbi->s_blockgroup_lock =
840                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
841         if (!sbi->s_blockgroup_lock) {
842                 kfree(sbi);
843                 goto failed;
844         }
845         sb->s_fs_info = sbi;
846         sbi->s_sb_block = sb_block;
847         sbi->s_daxdev = dax_dev;
848
849         spin_lock_init(&sbi->s_lock);
850
851         /*
852          * See what the current blocksize for the device is, and
853          * use that as the blocksize.  Otherwise (or if the blocksize
854          * is smaller than the default) use the default.
855          * This is important for devices that have a hardware
856          * sectorsize that is larger than the default.
857          */
858         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
859         if (!blocksize) {
860                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
861                 goto failed_sbi;
862         }
863
864         /*
865          * If the superblock doesn't start on a hardware sector boundary,
866          * calculate the offset.  
867          */
868         if (blocksize != BLOCK_SIZE) {
869                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
870                 offset = (sb_block*BLOCK_SIZE) % blocksize;
871         } else {
872                 logic_sb_block = sb_block;
873         }
874
875         if (!(bh = sb_bread(sb, logic_sb_block))) {
876                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
877                 goto failed_sbi;
878         }
879         /*
880          * Note: s_es must be initialized as soon as possible because
881          *       some ext2 macro-instructions depend on its value
882          */
883         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
884         sbi->s_es = es;
885         sb->s_magic = le16_to_cpu(es->s_magic);
886
887         if (sb->s_magic != EXT2_SUPER_MAGIC)
888                 goto cantfind_ext2;
889
890         /* Set defaults before we parse the mount options */
891         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
892         if (def_mount_opts & EXT2_DEFM_DEBUG)
893                 set_opt(sbi->s_mount_opt, DEBUG);
894         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
895                 set_opt(sbi->s_mount_opt, GRPID);
896         if (def_mount_opts & EXT2_DEFM_UID16)
897                 set_opt(sbi->s_mount_opt, NO_UID32);
898 #ifdef CONFIG_EXT2_FS_XATTR
899         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
900                 set_opt(sbi->s_mount_opt, XATTR_USER);
901 #endif
902 #ifdef CONFIG_EXT2_FS_POSIX_ACL
903         if (def_mount_opts & EXT2_DEFM_ACL)
904                 set_opt(sbi->s_mount_opt, POSIX_ACL);
905 #endif
906         
907         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
908                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
909         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
910                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
911         else
912                 set_opt(sbi->s_mount_opt, ERRORS_RO);
913
914         sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
915         sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
916         
917         set_opt(sbi->s_mount_opt, RESERVATION);
918
919         if (!parse_options((char *) data, sb))
920                 goto failed_mount;
921
922         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
923                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
924                  MS_POSIXACL : 0);
925         sb->s_iflags |= SB_I_CGROUPWB;
926
927         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
928             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
929              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
930              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
931                 ext2_msg(sb, KERN_WARNING,
932                         "warning: feature flags set on rev 0 fs, "
933                         "running e2fsck is recommended");
934         /*
935          * Check feature flags regardless of the revision level, since we
936          * previously didn't change the revision level when setting the flags,
937          * so there is a chance incompat flags are set on a rev 0 filesystem.
938          */
939         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
940         if (features) {
941                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
942                        "unsupported optional features (%x)",
943                         le32_to_cpu(features));
944                 goto failed_mount;
945         }
946         if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
947                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
948                        "unsupported optional features (%x)",
949                        le32_to_cpu(features));
950                 goto failed_mount;
951         }
952
953         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
954
955         if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
956                 err = bdev_dax_supported(sb, blocksize);
957                 if (err)
958                         goto failed_mount;
959         }
960
961         /* If the blocksize doesn't match, re-read the thing.. */
962         if (sb->s_blocksize != blocksize) {
963                 brelse(bh);
964
965                 if (!sb_set_blocksize(sb, blocksize)) {
966                         ext2_msg(sb, KERN_ERR,
967                                 "error: bad blocksize %d", blocksize);
968                         goto failed_sbi;
969                 }
970
971                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
972                 offset = (sb_block*BLOCK_SIZE) % blocksize;
973                 bh = sb_bread(sb, logic_sb_block);
974                 if(!bh) {
975                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
976                                 "superblock on 2nd try");
977                         goto failed_sbi;
978                 }
979                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
980                 sbi->s_es = es;
981                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
982                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
983                         goto failed_mount;
984                 }
985         }
986
987         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
988         sb->s_max_links = EXT2_LINK_MAX;
989
990         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
991                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
992                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
993         } else {
994                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
995                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
996                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
997                     !is_power_of_2(sbi->s_inode_size) ||
998                     (sbi->s_inode_size > blocksize)) {
999                         ext2_msg(sb, KERN_ERR,
1000                                 "error: unsupported inode size: %d",
1001                                 sbi->s_inode_size);
1002                         goto failed_mount;
1003                 }
1004         }
1005
1006         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1007                                    le32_to_cpu(es->s_log_frag_size);
1008         if (sbi->s_frag_size == 0)
1009                 goto cantfind_ext2;
1010         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1011
1012         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1013         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1014         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1015
1016         if (EXT2_INODE_SIZE(sb) == 0)
1017                 goto cantfind_ext2;
1018         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1019         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1020                 goto cantfind_ext2;
1021         sbi->s_itb_per_group = sbi->s_inodes_per_group /
1022                                         sbi->s_inodes_per_block;
1023         sbi->s_desc_per_block = sb->s_blocksize /
1024                                         sizeof (struct ext2_group_desc);
1025         sbi->s_sbh = bh;
1026         sbi->s_mount_state = le16_to_cpu(es->s_state);
1027         sbi->s_addr_per_block_bits =
1028                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1029         sbi->s_desc_per_block_bits =
1030                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
1031
1032         if (sb->s_magic != EXT2_SUPER_MAGIC)
1033                 goto cantfind_ext2;
1034
1035         if (sb->s_blocksize != bh->b_size) {
1036                 if (!silent)
1037                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1038                 goto failed_mount;
1039         }
1040
1041         if (sb->s_blocksize != sbi->s_frag_size) {
1042                 ext2_msg(sb, KERN_ERR,
1043                         "error: fragsize %lu != blocksize %lu"
1044                         "(not supported yet)",
1045                         sbi->s_frag_size, sb->s_blocksize);
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_frags_per_group > sb->s_blocksize * 8) {
1056                 ext2_msg(sb, KERN_ERR,
1057                         "error: #fragments per group too big: %lu",
1058                         sbi->s_frags_per_group);
1059                 goto failed_mount;
1060         }
1061         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1062                 ext2_msg(sb, KERN_ERR,
1063                         "error: #inodes per group too big: %lu",
1064                         sbi->s_inodes_per_group);
1065                 goto failed_mount;
1066         }
1067
1068         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1069                 goto cantfind_ext2;
1070         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1071                                 le32_to_cpu(es->s_first_data_block) - 1)
1072                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1073         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1074                    EXT2_DESC_PER_BLOCK(sb);
1075         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1076         if (sbi->s_group_desc == NULL) {
1077                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1078                 goto failed_mount;
1079         }
1080         bgl_lock_init(sbi->s_blockgroup_lock);
1081         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1082         if (!sbi->s_debts) {
1083                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1084                 goto failed_mount_group_desc;
1085         }
1086         for (i = 0; i < db_count; i++) {
1087                 block = descriptor_loc(sb, logic_sb_block, i);
1088                 sbi->s_group_desc[i] = sb_bread(sb, block);
1089                 if (!sbi->s_group_desc[i]) {
1090                         for (j = 0; j < i; j++)
1091                                 brelse (sbi->s_group_desc[j]);
1092                         ext2_msg(sb, KERN_ERR,
1093                                 "error: unable to read group descriptors");
1094                         goto failed_mount_group_desc;
1095                 }
1096         }
1097         if (!ext2_check_descriptors (sb)) {
1098                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1099                 goto failed_mount2;
1100         }
1101         sbi->s_gdb_count = db_count;
1102         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1103         spin_lock_init(&sbi->s_next_gen_lock);
1104
1105         /* per fileystem reservation list head & lock */
1106         spin_lock_init(&sbi->s_rsv_window_lock);
1107         sbi->s_rsv_window_root = RB_ROOT;
1108         /*
1109          * Add a single, static dummy reservation to the start of the
1110          * reservation window list --- it gives us a placeholder for
1111          * append-at-start-of-list which makes the allocation logic
1112          * _much_ simpler.
1113          */
1114         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1115         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1116         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1117         sbi->s_rsv_window_head.rsv_goal_size = 0;
1118         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1119
1120         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1121                                 ext2_count_free_blocks(sb), GFP_KERNEL);
1122         if (!err) {
1123                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1124                                 ext2_count_free_inodes(sb), GFP_KERNEL);
1125         }
1126         if (!err) {
1127                 err = percpu_counter_init(&sbi->s_dirs_counter,
1128                                 ext2_count_dirs(sb), GFP_KERNEL);
1129         }
1130         if (err) {
1131                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1132                 goto failed_mount3;
1133         }
1134
1135 #ifdef CONFIG_EXT2_FS_XATTR
1136         sbi->s_ea_block_cache = ext2_xattr_create_cache();
1137         if (!sbi->s_ea_block_cache) {
1138                 ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1139                 goto failed_mount3;
1140         }
1141 #endif
1142         /*
1143          * set up enough so that it can read an inode
1144          */
1145         sb->s_op = &ext2_sops;
1146         sb->s_export_op = &ext2_export_ops;
1147         sb->s_xattr = ext2_xattr_handlers;
1148
1149 #ifdef CONFIG_QUOTA
1150         sb->dq_op = &dquot_operations;
1151         sb->s_qcop = &ext2_quotactl_ops;
1152         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1153 #endif
1154
1155         root = ext2_iget(sb, EXT2_ROOT_INO);
1156         if (IS_ERR(root)) {
1157                 ret = PTR_ERR(root);
1158                 goto failed_mount3;
1159         }
1160         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1161                 iput(root);
1162                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1163                 goto failed_mount3;
1164         }
1165
1166         sb->s_root = d_make_root(root);
1167         if (!sb->s_root) {
1168                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1169                 ret = -ENOMEM;
1170                 goto failed_mount3;
1171         }
1172         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1173                 ext2_msg(sb, KERN_WARNING,
1174                         "warning: mounting ext3 filesystem as ext2");
1175         if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1176                 sb->s_flags |= MS_RDONLY;
1177         ext2_write_super(sb);
1178         return 0;
1179
1180 cantfind_ext2:
1181         if (!silent)
1182                 ext2_msg(sb, KERN_ERR,
1183                         "error: can't find an ext2 filesystem on dev %s.",
1184                         sb->s_id);
1185         goto failed_mount;
1186 failed_mount3:
1187         if (sbi->s_ea_block_cache)
1188                 ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1189         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1190         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1191         percpu_counter_destroy(&sbi->s_dirs_counter);
1192 failed_mount2:
1193         for (i = 0; i < db_count; i++)
1194                 brelse(sbi->s_group_desc[i]);
1195 failed_mount_group_desc:
1196         kfree(sbi->s_group_desc);
1197         kfree(sbi->s_debts);
1198 failed_mount:
1199         brelse(bh);
1200 failed_sbi:
1201         sb->s_fs_info = NULL;
1202         kfree(sbi->s_blockgroup_lock);
1203         kfree(sbi);
1204 failed:
1205         fs_put_dax(dax_dev);
1206         return ret;
1207 }
1208
1209 static void ext2_clear_super_error(struct super_block *sb)
1210 {
1211         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1212
1213         if (buffer_write_io_error(sbh)) {
1214                 /*
1215                  * Oh, dear.  A previous attempt to write the
1216                  * superblock failed.  This could happen because the
1217                  * USB device was yanked out.  Or it could happen to
1218                  * be a transient write error and maybe the block will
1219                  * be remapped.  Nothing we can do but to retry the
1220                  * write and hope for the best.
1221                  */
1222                 ext2_msg(sb, KERN_ERR,
1223                        "previous I/O error to superblock detected\n");
1224                 clear_buffer_write_io_error(sbh);
1225                 set_buffer_uptodate(sbh);
1226         }
1227 }
1228
1229 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1230                      int wait)
1231 {
1232         ext2_clear_super_error(sb);
1233         spin_lock(&EXT2_SB(sb)->s_lock);
1234         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1235         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1236         es->s_wtime = cpu_to_le32(get_seconds());
1237         /* unlock before we do IO */
1238         spin_unlock(&EXT2_SB(sb)->s_lock);
1239         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1240         if (wait)
1241                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1242 }
1243
1244 /*
1245  * In the second extended file system, it is not necessary to
1246  * write the super block since we use a mapping of the
1247  * disk super block in a buffer.
1248  *
1249  * However, this function is still used to set the fs valid
1250  * flags to 0.  We need to set this flag to 0 since the fs
1251  * may have been checked while mounted and e2fsck may have
1252  * set s_state to EXT2_VALID_FS after some corrections.
1253  */
1254 static int ext2_sync_fs(struct super_block *sb, int wait)
1255 {
1256         struct ext2_sb_info *sbi = EXT2_SB(sb);
1257         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1258
1259         /*
1260          * Write quota structures to quota file, sync_blockdev() will write
1261          * them to disk later
1262          */
1263         dquot_writeback_dquots(sb, -1);
1264
1265         spin_lock(&sbi->s_lock);
1266         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1267                 ext2_debug("setting valid to 0\n");
1268                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1269         }
1270         spin_unlock(&sbi->s_lock);
1271         ext2_sync_super(sb, es, wait);
1272         return 0;
1273 }
1274
1275 static int ext2_freeze(struct super_block *sb)
1276 {
1277         struct ext2_sb_info *sbi = EXT2_SB(sb);
1278
1279         /*
1280          * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1281          * because we have unattached inodes and thus filesystem is not fully
1282          * consistent.
1283          */
1284         if (atomic_long_read(&sb->s_remove_count)) {
1285                 ext2_sync_fs(sb, 1);
1286                 return 0;
1287         }
1288         /* Set EXT2_FS_VALID flag */
1289         spin_lock(&sbi->s_lock);
1290         sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1291         spin_unlock(&sbi->s_lock);
1292         ext2_sync_super(sb, sbi->s_es, 1);
1293
1294         return 0;
1295 }
1296
1297 static int ext2_unfreeze(struct super_block *sb)
1298 {
1299         /* Just write sb to clear EXT2_VALID_FS flag */
1300         ext2_write_super(sb);
1301
1302         return 0;
1303 }
1304
1305 static void ext2_write_super(struct super_block *sb)
1306 {
1307         if (!sb_rdonly(sb))
1308                 ext2_sync_fs(sb, 1);
1309 }
1310
1311 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1312 {
1313         struct ext2_sb_info * sbi = EXT2_SB(sb);
1314         struct ext2_super_block * es;
1315         struct ext2_mount_options old_opts;
1316         unsigned long old_sb_flags;
1317         int err;
1318
1319         sync_filesystem(sb);
1320         spin_lock(&sbi->s_lock);
1321
1322         /* Store the old options */
1323         old_sb_flags = sb->s_flags;
1324         old_opts.s_mount_opt = sbi->s_mount_opt;
1325         old_opts.s_resuid = sbi->s_resuid;
1326         old_opts.s_resgid = sbi->s_resgid;
1327
1328         /*
1329          * Allow the "check" option to be passed as a remount option.
1330          */
1331         if (!parse_options(data, sb)) {
1332                 err = -EINVAL;
1333                 goto restore_opts;
1334         }
1335
1336         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1337                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1338
1339         es = sbi->s_es;
1340         if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1341                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1342                          "dax flag with busy inodes while remounting");
1343                 sbi->s_mount_opt ^= EXT2_MOUNT_DAX;
1344         }
1345         if ((bool)(*flags & MS_RDONLY) == sb_rdonly(sb)) {
1346                 spin_unlock(&sbi->s_lock);
1347                 return 0;
1348         }
1349         if (*flags & MS_RDONLY) {
1350                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1351                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1352                         spin_unlock(&sbi->s_lock);
1353                         return 0;
1354                 }
1355
1356                 /*
1357                  * OK, we are remounting a valid rw partition rdonly, so set
1358                  * the rdonly flag and then mark the partition as valid again.
1359                  */
1360                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1361                 es->s_mtime = cpu_to_le32(get_seconds());
1362                 spin_unlock(&sbi->s_lock);
1363
1364                 err = dquot_suspend(sb, -1);
1365                 if (err < 0) {
1366                         spin_lock(&sbi->s_lock);
1367                         goto restore_opts;
1368                 }
1369
1370                 ext2_sync_super(sb, es, 1);
1371         } else {
1372                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1373                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1374                 if (ret) {
1375                         ext2_msg(sb, KERN_WARNING,
1376                                 "warning: couldn't remount RDWR because of "
1377                                 "unsupported optional features (%x).",
1378                                 le32_to_cpu(ret));
1379                         err = -EROFS;
1380                         goto restore_opts;
1381                 }
1382                 /*
1383                  * Mounting a RDONLY partition read-write, so reread and
1384                  * store the current valid flag.  (It may have been changed
1385                  * by e2fsck since we originally mounted the partition.)
1386                  */
1387                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1388                 if (!ext2_setup_super (sb, es, 0))
1389                         sb->s_flags &= ~MS_RDONLY;
1390                 spin_unlock(&sbi->s_lock);
1391
1392                 ext2_write_super(sb);
1393
1394                 dquot_resume(sb, -1);
1395         }
1396
1397         return 0;
1398 restore_opts:
1399         sbi->s_mount_opt = old_opts.s_mount_opt;
1400         sbi->s_resuid = old_opts.s_resuid;
1401         sbi->s_resgid = old_opts.s_resgid;
1402         sb->s_flags = old_sb_flags;
1403         spin_unlock(&sbi->s_lock);
1404         return err;
1405 }
1406
1407 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1408 {
1409         struct super_block *sb = dentry->d_sb;
1410         struct ext2_sb_info *sbi = EXT2_SB(sb);
1411         struct ext2_super_block *es = sbi->s_es;
1412         u64 fsid;
1413
1414         spin_lock(&sbi->s_lock);
1415
1416         if (test_opt (sb, MINIX_DF))
1417                 sbi->s_overhead_last = 0;
1418         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1419                 unsigned long i, overhead = 0;
1420                 smp_rmb();
1421
1422                 /*
1423                  * Compute the overhead (FS structures). This is constant
1424                  * for a given filesystem unless the number of block groups
1425                  * changes so we cache the previous value until it does.
1426                  */
1427
1428                 /*
1429                  * All of the blocks before first_data_block are
1430                  * overhead
1431                  */
1432                 overhead = le32_to_cpu(es->s_first_data_block);
1433
1434                 /*
1435                  * Add the overhead attributed to the superblock and
1436                  * block group descriptors.  If the sparse superblocks
1437                  * feature is turned on, then not all groups have this.
1438                  */
1439                 for (i = 0; i < sbi->s_groups_count; i++)
1440                         overhead += ext2_bg_has_super(sb, i) +
1441                                 ext2_bg_num_gdb(sb, i);
1442
1443                 /*
1444                  * Every block group has an inode bitmap, a block
1445                  * bitmap, and an inode table.
1446                  */
1447                 overhead += (sbi->s_groups_count *
1448                              (2 + sbi->s_itb_per_group));
1449                 sbi->s_overhead_last = overhead;
1450                 smp_wmb();
1451                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1452         }
1453
1454         buf->f_type = EXT2_SUPER_MAGIC;
1455         buf->f_bsize = sb->s_blocksize;
1456         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1457         buf->f_bfree = ext2_count_free_blocks(sb);
1458         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1459         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1460         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1461                 buf->f_bavail = 0;
1462         buf->f_files = le32_to_cpu(es->s_inodes_count);
1463         buf->f_ffree = ext2_count_free_inodes(sb);
1464         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1465         buf->f_namelen = EXT2_NAME_LEN;
1466         fsid = le64_to_cpup((void *)es->s_uuid) ^
1467                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1468         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1469         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1470         spin_unlock(&sbi->s_lock);
1471         return 0;
1472 }
1473
1474 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1475         int flags, const char *dev_name, void *data)
1476 {
1477         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1478 }
1479
1480 #ifdef CONFIG_QUOTA
1481
1482 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1483  * acquiring the locks... As quota files are never truncated and quota code
1484  * itself serializes the operations (and no one else should touch the files)
1485  * we don't have to be afraid of races */
1486 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1487                                size_t len, loff_t off)
1488 {
1489         struct inode *inode = sb_dqopt(sb)->files[type];
1490         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1491         int err = 0;
1492         int offset = off & (sb->s_blocksize - 1);
1493         int tocopy;
1494         size_t toread;
1495         struct buffer_head tmp_bh;
1496         struct buffer_head *bh;
1497         loff_t i_size = i_size_read(inode);
1498
1499         if (off > i_size)
1500                 return 0;
1501         if (off+len > i_size)
1502                 len = i_size-off;
1503         toread = len;
1504         while (toread > 0) {
1505                 tocopy = sb->s_blocksize - offset < toread ?
1506                                 sb->s_blocksize - offset : toread;
1507
1508                 tmp_bh.b_state = 0;
1509                 tmp_bh.b_size = sb->s_blocksize;
1510                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1511                 if (err < 0)
1512                         return err;
1513                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1514                         memset(data, 0, tocopy);
1515                 else {
1516                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1517                         if (!bh)
1518                                 return -EIO;
1519                         memcpy(data, bh->b_data+offset, tocopy);
1520                         brelse(bh);
1521                 }
1522                 offset = 0;
1523                 toread -= tocopy;
1524                 data += tocopy;
1525                 blk++;
1526         }
1527         return len;
1528 }
1529
1530 /* Write to quotafile */
1531 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1532                                 const char *data, size_t len, loff_t off)
1533 {
1534         struct inode *inode = sb_dqopt(sb)->files[type];
1535         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1536         int err = 0;
1537         int offset = off & (sb->s_blocksize - 1);
1538         int tocopy;
1539         size_t towrite = len;
1540         struct buffer_head tmp_bh;
1541         struct buffer_head *bh;
1542
1543         while (towrite > 0) {
1544                 tocopy = sb->s_blocksize - offset < towrite ?
1545                                 sb->s_blocksize - offset : towrite;
1546
1547                 tmp_bh.b_state = 0;
1548                 tmp_bh.b_size = sb->s_blocksize;
1549                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1550                 if (err < 0)
1551                         goto out;
1552                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1553                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1554                 else
1555                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1556                 if (unlikely(!bh)) {
1557                         err = -EIO;
1558                         goto out;
1559                 }
1560                 lock_buffer(bh);
1561                 memcpy(bh->b_data+offset, data, tocopy);
1562                 flush_dcache_page(bh->b_page);
1563                 set_buffer_uptodate(bh);
1564                 mark_buffer_dirty(bh);
1565                 unlock_buffer(bh);
1566                 brelse(bh);
1567                 offset = 0;
1568                 towrite -= tocopy;
1569                 data += tocopy;
1570                 blk++;
1571         }
1572 out:
1573         if (len == towrite)
1574                 return err;
1575         if (inode->i_size < off+len-towrite)
1576                 i_size_write(inode, off+len-towrite);
1577         inode->i_version++;
1578         inode->i_mtime = inode->i_ctime = current_time(inode);
1579         mark_inode_dirty(inode);
1580         return len - towrite;
1581 }
1582
1583 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1584                          const struct path *path)
1585 {
1586         int err;
1587         struct inode *inode;
1588
1589         err = dquot_quota_on(sb, type, format_id, path);
1590         if (err)
1591                 return err;
1592
1593         inode = d_inode(path->dentry);
1594         inode_lock(inode);
1595         EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1596         inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1597                         S_NOATIME | S_IMMUTABLE);
1598         inode_unlock(inode);
1599         mark_inode_dirty(inode);
1600
1601         return 0;
1602 }
1603
1604 static int ext2_quota_off(struct super_block *sb, int type)
1605 {
1606         struct inode *inode = sb_dqopt(sb)->files[type];
1607         int err;
1608
1609         if (!inode || !igrab(inode))
1610                 goto out;
1611
1612         err = dquot_quota_off(sb, type);
1613         if (err)
1614                 goto out_put;
1615
1616         inode_lock(inode);
1617         EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1618         inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1619         inode_unlock(inode);
1620         mark_inode_dirty(inode);
1621 out_put:
1622         iput(inode);
1623         return err;
1624 out:
1625         return dquot_quota_off(sb, type);
1626 }
1627
1628 #endif
1629
1630 static struct file_system_type ext2_fs_type = {
1631         .owner          = THIS_MODULE,
1632         .name           = "ext2",
1633         .mount          = ext2_mount,
1634         .kill_sb        = kill_block_super,
1635         .fs_flags       = FS_REQUIRES_DEV,
1636 };
1637 MODULE_ALIAS_FS("ext2");
1638
1639 static int __init init_ext2_fs(void)
1640 {
1641         int err;
1642
1643         err = init_inodecache();
1644         if (err)
1645                 return err;
1646         err = register_filesystem(&ext2_fs_type);
1647         if (err)
1648                 goto out;
1649         return 0;
1650 out:
1651         destroy_inodecache();
1652         return err;
1653 }
1654
1655 static void __exit exit_ext2_fs(void)
1656 {
1657         unregister_filesystem(&ext2_fs_type);
1658         destroy_inodecache();
1659 }
1660
1661 MODULE_AUTHOR("Remy Card and others");
1662 MODULE_DESCRIPTION("Second Extended Filesystem");
1663 MODULE_LICENSE("GPL");
1664 module_init(init_ext2_fs)
1665 module_exit(exit_ext2_fs)