ext2: switch to ->free_inode()
[muen/linux.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 <linux/iversion.h>
37 #include "ext2.h"
38 #include "xattr.h"
39 #include "acl.h"
40
41 static void ext2_write_super(struct super_block *sb);
42 static int ext2_remount (struct super_block * sb, int * flags, char * data);
43 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
44 static int ext2_sync_fs(struct super_block *sb, int wait);
45 static int ext2_freeze(struct super_block *sb);
46 static int ext2_unfreeze(struct super_block *sb);
47
48 void ext2_error(struct super_block *sb, const char *function,
49                 const char *fmt, ...)
50 {
51         struct va_format vaf;
52         va_list args;
53         struct ext2_sb_info *sbi = EXT2_SB(sb);
54         struct ext2_super_block *es = sbi->s_es;
55
56         if (!sb_rdonly(sb)) {
57                 spin_lock(&sbi->s_lock);
58                 sbi->s_mount_state |= EXT2_ERROR_FS;
59                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
60                 spin_unlock(&sbi->s_lock);
61                 ext2_sync_super(sb, es, 1);
62         }
63
64         va_start(args, fmt);
65
66         vaf.fmt = fmt;
67         vaf.va = &args;
68
69         printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
70                sb->s_id, function, &vaf);
71
72         va_end(args);
73
74         if (test_opt(sb, ERRORS_PANIC))
75                 panic("EXT2-fs: panic from previous error\n");
76         if (!sb_rdonly(sb) && test_opt(sb, ERRORS_RO)) {
77                 ext2_msg(sb, KERN_CRIT,
78                              "error: remounting filesystem read-only");
79                 sb->s_flags |= SB_RDONLY;
80         }
81 }
82
83 void ext2_msg(struct super_block *sb, const char *prefix,
84                 const char *fmt, ...)
85 {
86         struct va_format vaf;
87         va_list args;
88
89         va_start(args, fmt);
90
91         vaf.fmt = fmt;
92         vaf.va = &args;
93
94         printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
95
96         va_end(args);
97 }
98
99 /*
100  * This must be called with sbi->s_lock held.
101  */
102 void ext2_update_dynamic_rev(struct super_block *sb)
103 {
104         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
105
106         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
107                 return;
108
109         ext2_msg(sb, KERN_WARNING,
110                      "warning: updating to rev %d because of "
111                      "new feature flag, running e2fsck is recommended",
112                      EXT2_DYNAMIC_REV);
113
114         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
115         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
116         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
117         /* leave es->s_feature_*compat flags alone */
118         /* es->s_uuid will be set by e2fsck if empty */
119
120         /*
121          * The rest of the superblock fields should be zero, and if not it
122          * means they are likely already in use, so leave them alone.  We
123          * can leave it up to e2fsck to clean up any inconsistencies there.
124          */
125 }
126
127 #ifdef CONFIG_QUOTA
128 static int ext2_quota_off(struct super_block *sb, int type);
129
130 static void ext2_quota_off_umount(struct super_block *sb)
131 {
132         int type;
133
134         for (type = 0; type < MAXQUOTAS; type++)
135                 ext2_quota_off(sb, type);
136 }
137 #else
138 static inline void ext2_quota_off_umount(struct super_block *sb)
139 {
140 }
141 #endif
142
143 static void ext2_put_super (struct super_block * sb)
144 {
145         int db_count;
146         int i;
147         struct ext2_sb_info *sbi = EXT2_SB(sb);
148
149         ext2_quota_off_umount(sb);
150
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         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 (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
306                 seq_puts(seq, ",usrquota");
307
308         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
309                 seq_puts(seq, ",grpquota");
310
311         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
312                 seq_puts(seq, ",xip");
313
314         if (sbi->s_mount_opt & EXT2_MOUNT_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_nocheck, 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_nocheck, "check=none"},
455         {Opt_nocheck, "nocheck"},
456         {Opt_debug, "debug"},
457         {Opt_oldalloc, "oldalloc"},
458         {Opt_orlov, "orlov"},
459         {Opt_nobh, "nobh"},
460         {Opt_user_xattr, "user_xattr"},
461         {Opt_nouser_xattr, "nouser_xattr"},
462         {Opt_acl, "acl"},
463         {Opt_noacl, "noacl"},
464         {Opt_xip, "xip"},
465         {Opt_dax, "dax"},
466         {Opt_grpquota, "grpquota"},
467         {Opt_ignore, "noquota"},
468         {Opt_quota, "quota"},
469         {Opt_usrquota, "usrquota"},
470         {Opt_reservation, "reservation"},
471         {Opt_noreservation, "noreservation"},
472         {Opt_err, NULL}
473 };
474
475 static int parse_options(char *options, struct super_block *sb,
476                          struct ext2_mount_options *opts)
477 {
478         char *p;
479         substring_t args[MAX_OPT_ARGS];
480         int option;
481         kuid_t uid;
482         kgid_t gid;
483
484         if (!options)
485                 return 1;
486
487         while ((p = strsep (&options, ",")) != NULL) {
488                 int token;
489                 if (!*p)
490                         continue;
491
492                 token = match_token(p, tokens, args);
493                 switch (token) {
494                 case Opt_bsd_df:
495                         clear_opt (opts->s_mount_opt, MINIX_DF);
496                         break;
497                 case Opt_minix_df:
498                         set_opt (opts->s_mount_opt, MINIX_DF);
499                         break;
500                 case Opt_grpid:
501                         set_opt (opts->s_mount_opt, GRPID);
502                         break;
503                 case Opt_nogrpid:
504                         clear_opt (opts->s_mount_opt, GRPID);
505                         break;
506                 case Opt_resuid:
507                         if (match_int(&args[0], &option))
508                                 return 0;
509                         uid = make_kuid(current_user_ns(), option);
510                         if (!uid_valid(uid)) {
511                                 ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
512                                 return 0;
513
514                         }
515                         opts->s_resuid = uid;
516                         break;
517                 case Opt_resgid:
518                         if (match_int(&args[0], &option))
519                                 return 0;
520                         gid = make_kgid(current_user_ns(), option);
521                         if (!gid_valid(gid)) {
522                                 ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
523                                 return 0;
524                         }
525                         opts->s_resgid = gid;
526                         break;
527                 case Opt_sb:
528                         /* handled by get_sb_block() instead of here */
529                         /* *sb_block = match_int(&args[0]); */
530                         break;
531                 case Opt_err_panic:
532                         clear_opt (opts->s_mount_opt, ERRORS_CONT);
533                         clear_opt (opts->s_mount_opt, ERRORS_RO);
534                         set_opt (opts->s_mount_opt, ERRORS_PANIC);
535                         break;
536                 case Opt_err_ro:
537                         clear_opt (opts->s_mount_opt, ERRORS_CONT);
538                         clear_opt (opts->s_mount_opt, ERRORS_PANIC);
539                         set_opt (opts->s_mount_opt, ERRORS_RO);
540                         break;
541                 case Opt_err_cont:
542                         clear_opt (opts->s_mount_opt, ERRORS_RO);
543                         clear_opt (opts->s_mount_opt, ERRORS_PANIC);
544                         set_opt (opts->s_mount_opt, ERRORS_CONT);
545                         break;
546                 case Opt_nouid32:
547                         set_opt (opts->s_mount_opt, NO_UID32);
548                         break;
549                 case Opt_nocheck:
550                         ext2_msg(sb, KERN_WARNING,
551                                 "Option nocheck/check=none is deprecated and"
552                                 " will be removed in June 2020.");
553                         clear_opt (opts->s_mount_opt, CHECK);
554                         break;
555                 case Opt_debug:
556                         set_opt (opts->s_mount_opt, DEBUG);
557                         break;
558                 case Opt_oldalloc:
559                         set_opt (opts->s_mount_opt, OLDALLOC);
560                         break;
561                 case Opt_orlov:
562                         clear_opt (opts->s_mount_opt, OLDALLOC);
563                         break;
564                 case Opt_nobh:
565                         set_opt (opts->s_mount_opt, NOBH);
566                         break;
567 #ifdef CONFIG_EXT2_FS_XATTR
568                 case Opt_user_xattr:
569                         set_opt (opts->s_mount_opt, XATTR_USER);
570                         break;
571                 case Opt_nouser_xattr:
572                         clear_opt (opts->s_mount_opt, XATTR_USER);
573                         break;
574 #else
575                 case Opt_user_xattr:
576                 case Opt_nouser_xattr:
577                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
578                                 "not supported");
579                         break;
580 #endif
581 #ifdef CONFIG_EXT2_FS_POSIX_ACL
582                 case Opt_acl:
583                         set_opt(opts->s_mount_opt, POSIX_ACL);
584                         break;
585                 case Opt_noacl:
586                         clear_opt(opts->s_mount_opt, POSIX_ACL);
587                         break;
588 #else
589                 case Opt_acl:
590                 case Opt_noacl:
591                         ext2_msg(sb, KERN_INFO,
592                                 "(no)acl options not supported");
593                         break;
594 #endif
595                 case Opt_xip:
596                         ext2_msg(sb, KERN_INFO, "use dax instead of xip");
597                         set_opt(opts->s_mount_opt, XIP);
598                         /* Fall through */
599                 case Opt_dax:
600 #ifdef CONFIG_FS_DAX
601                         ext2_msg(sb, KERN_WARNING,
602                 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
603                         set_opt(opts->s_mount_opt, DAX);
604 #else
605                         ext2_msg(sb, KERN_INFO, "dax option not supported");
606 #endif
607                         break;
608
609 #if defined(CONFIG_QUOTA)
610                 case Opt_quota:
611                 case Opt_usrquota:
612                         set_opt(opts->s_mount_opt, USRQUOTA);
613                         break;
614
615                 case Opt_grpquota:
616                         set_opt(opts->s_mount_opt, GRPQUOTA);
617                         break;
618 #else
619                 case Opt_quota:
620                 case Opt_usrquota:
621                 case Opt_grpquota:
622                         ext2_msg(sb, KERN_INFO,
623                                 "quota operations not supported");
624                         break;
625 #endif
626
627                 case Opt_reservation:
628                         set_opt(opts->s_mount_opt, RESERVATION);
629                         ext2_msg(sb, KERN_INFO, "reservations ON");
630                         break;
631                 case Opt_noreservation:
632                         clear_opt(opts->s_mount_opt, RESERVATION);
633                         ext2_msg(sb, KERN_INFO, "reservations OFF");
634                         break;
635                 case Opt_ignore:
636                         break;
637                 default:
638                         return 0;
639                 }
640         }
641         return 1;
642 }
643
644 static int ext2_setup_super (struct super_block * sb,
645                               struct ext2_super_block * es,
646                               int read_only)
647 {
648         int res = 0;
649         struct ext2_sb_info *sbi = EXT2_SB(sb);
650
651         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
652                 ext2_msg(sb, KERN_ERR,
653                         "error: revision level too high, "
654                         "forcing read-only mode");
655                 res = SB_RDONLY;
656         }
657         if (read_only)
658                 return res;
659         if (!(sbi->s_mount_state & EXT2_VALID_FS))
660                 ext2_msg(sb, KERN_WARNING,
661                         "warning: mounting unchecked fs, "
662                         "running e2fsck is recommended");
663         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
664                 ext2_msg(sb, KERN_WARNING,
665                         "warning: mounting fs with errors, "
666                         "running e2fsck is recommended");
667         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
668                  le16_to_cpu(es->s_mnt_count) >=
669                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
670                 ext2_msg(sb, KERN_WARNING,
671                         "warning: maximal mount count reached, "
672                         "running e2fsck is recommended");
673         else if (le32_to_cpu(es->s_checkinterval) &&
674                 (le32_to_cpu(es->s_lastcheck) +
675                         le32_to_cpu(es->s_checkinterval) <=
676                         ktime_get_real_seconds()))
677                 ext2_msg(sb, KERN_WARNING,
678                         "warning: checktime reached, "
679                         "running e2fsck is recommended");
680         if (!le16_to_cpu(es->s_max_mnt_count))
681                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
682         le16_add_cpu(&es->s_mnt_count, 1);
683         if (test_opt (sb, DEBUG))
684                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
685                         "bpg=%lu, ipg=%lu, mo=%04lx]",
686                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
687                         sbi->s_frag_size,
688                         sbi->s_groups_count,
689                         EXT2_BLOCKS_PER_GROUP(sb),
690                         EXT2_INODES_PER_GROUP(sb),
691                         sbi->s_mount_opt);
692         return res;
693 }
694
695 static int ext2_check_descriptors(struct super_block *sb)
696 {
697         int i;
698         struct ext2_sb_info *sbi = EXT2_SB(sb);
699
700         ext2_debug ("Checking group descriptors");
701
702         for (i = 0; i < sbi->s_groups_count; i++) {
703                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
704                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
705                 ext2_fsblk_t last_block;
706
707                 if (i == sbi->s_groups_count - 1)
708                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
709                 else
710                         last_block = first_block +
711                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
712
713                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
714                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
715                 {
716                         ext2_error (sb, "ext2_check_descriptors",
717                                     "Block bitmap for group %d"
718                                     " not in group (block %lu)!",
719                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
720                         return 0;
721                 }
722                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
723                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
724                 {
725                         ext2_error (sb, "ext2_check_descriptors",
726                                     "Inode bitmap for group %d"
727                                     " not in group (block %lu)!",
728                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
729                         return 0;
730                 }
731                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
732                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
733                     last_block)
734                 {
735                         ext2_error (sb, "ext2_check_descriptors",
736                                     "Inode table for group %d"
737                                     " not in group (block %lu)!",
738                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
739                         return 0;
740                 }
741         }
742         return 1;
743 }
744
745 /*
746  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
747  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
748  * We need to be 1 filesystem block less than the 2^32 sector limit.
749  */
750 static loff_t ext2_max_size(int bits)
751 {
752         loff_t res = EXT2_NDIR_BLOCKS;
753         int meta_blocks;
754         unsigned int upper_limit;
755         unsigned int ppb = 1 << (bits-2);
756
757         /* This is calculated to be the largest file size for a
758          * dense, file such that the total number of
759          * sectors in the file, including data and all indirect blocks,
760          * does not exceed 2^32 -1
761          * __u32 i_blocks representing the total number of
762          * 512 bytes blocks of the file
763          */
764         upper_limit = (1LL << 32) - 1;
765
766         /* total blocks in file system block size */
767         upper_limit >>= (bits - 9);
768
769         /* Compute how many blocks we can address by block tree */
770         res += 1LL << (bits-2);
771         res += 1LL << (2*(bits-2));
772         res += 1LL << (3*(bits-2));
773         /* Does block tree limit file size? */
774         if (res < upper_limit)
775                 goto check_lfs;
776
777         res = upper_limit;
778         /* How many metadata blocks are needed for addressing upper_limit? */
779         upper_limit -= EXT2_NDIR_BLOCKS;
780         /* indirect blocks */
781         meta_blocks = 1;
782         upper_limit -= ppb;
783         /* double indirect blocks */
784         if (upper_limit < ppb * ppb) {
785                 meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
786                 res -= meta_blocks;
787                 goto check_lfs;
788         }
789         meta_blocks += 1 + ppb;
790         upper_limit -= ppb * ppb;
791         /* tripple indirect blocks for the rest */
792         meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
793                 DIV_ROUND_UP(upper_limit, ppb*ppb);
794         res -= meta_blocks;
795 check_lfs:
796         res <<= bits;
797         if (res > MAX_LFS_FILESIZE)
798                 res = MAX_LFS_FILESIZE;
799
800         return res;
801 }
802
803 static unsigned long descriptor_loc(struct super_block *sb,
804                                     unsigned long logic_sb_block,
805                                     int nr)
806 {
807         struct ext2_sb_info *sbi = EXT2_SB(sb);
808         unsigned long bg, first_meta_bg;
809         int has_super = 0;
810         
811         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
812
813         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
814             nr < first_meta_bg)
815                 return (logic_sb_block + nr + 1);
816         bg = sbi->s_desc_per_block * nr;
817         if (ext2_bg_has_super(sb, bg))
818                 has_super = 1;
819
820         return ext2_group_first_block_no(sb, bg) + has_super;
821 }
822
823 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
824 {
825         struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
826         struct buffer_head * bh;
827         struct ext2_sb_info * sbi;
828         struct ext2_super_block * es;
829         struct inode *root;
830         unsigned long block;
831         unsigned long sb_block = get_sb_block(&data);
832         unsigned long logic_sb_block;
833         unsigned long offset = 0;
834         unsigned long def_mount_opts;
835         long ret = -ENOMEM;
836         int blocksize = BLOCK_SIZE;
837         int db_count;
838         int i, j;
839         __le32 features;
840         int err;
841         struct ext2_mount_options opts;
842
843         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
844         if (!sbi)
845                 goto failed;
846
847         sbi->s_blockgroup_lock =
848                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
849         if (!sbi->s_blockgroup_lock) {
850                 kfree(sbi);
851                 goto failed;
852         }
853         sb->s_fs_info = sbi;
854         sbi->s_sb_block = sb_block;
855         sbi->s_daxdev = dax_dev;
856
857         spin_lock_init(&sbi->s_lock);
858         ret = -EINVAL;
859
860         /*
861          * See what the current blocksize for the device is, and
862          * use that as the blocksize.  Otherwise (or if the blocksize
863          * is smaller than the default) use the default.
864          * This is important for devices that have a hardware
865          * sectorsize that is larger than the default.
866          */
867         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
868         if (!blocksize) {
869                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
870                 goto failed_sbi;
871         }
872
873         /*
874          * If the superblock doesn't start on a hardware sector boundary,
875          * calculate the offset.  
876          */
877         if (blocksize != BLOCK_SIZE) {
878                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
879                 offset = (sb_block*BLOCK_SIZE) % blocksize;
880         } else {
881                 logic_sb_block = sb_block;
882         }
883
884         if (!(bh = sb_bread(sb, logic_sb_block))) {
885                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
886                 goto failed_sbi;
887         }
888         /*
889          * Note: s_es must be initialized as soon as possible because
890          *       some ext2 macro-instructions depend on its value
891          */
892         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
893         sbi->s_es = es;
894         sb->s_magic = le16_to_cpu(es->s_magic);
895
896         if (sb->s_magic != EXT2_SUPER_MAGIC)
897                 goto cantfind_ext2;
898
899         opts.s_mount_opt = 0;
900         /* Set defaults before we parse the mount options */
901         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
902         if (def_mount_opts & EXT2_DEFM_DEBUG)
903                 set_opt(opts.s_mount_opt, DEBUG);
904         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
905                 set_opt(opts.s_mount_opt, GRPID);
906         if (def_mount_opts & EXT2_DEFM_UID16)
907                 set_opt(opts.s_mount_opt, NO_UID32);
908 #ifdef CONFIG_EXT2_FS_XATTR
909         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
910                 set_opt(opts.s_mount_opt, XATTR_USER);
911 #endif
912 #ifdef CONFIG_EXT2_FS_POSIX_ACL
913         if (def_mount_opts & EXT2_DEFM_ACL)
914                 set_opt(opts.s_mount_opt, POSIX_ACL);
915 #endif
916         
917         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
918                 set_opt(opts.s_mount_opt, ERRORS_PANIC);
919         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
920                 set_opt(opts.s_mount_opt, ERRORS_CONT);
921         else
922                 set_opt(opts.s_mount_opt, ERRORS_RO);
923
924         opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
925         opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
926         
927         set_opt(opts.s_mount_opt, RESERVATION);
928
929         if (!parse_options((char *) data, sb, &opts))
930                 goto failed_mount;
931
932         sbi->s_mount_opt = opts.s_mount_opt;
933         sbi->s_resuid = opts.s_resuid;
934         sbi->s_resgid = opts.s_resgid;
935
936         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
937                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
938                  SB_POSIXACL : 0);
939         sb->s_iflags |= SB_I_CGROUPWB;
940
941         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
942             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
943              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
944              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
945                 ext2_msg(sb, KERN_WARNING,
946                         "warning: feature flags set on rev 0 fs, "
947                         "running e2fsck is recommended");
948         /*
949          * Check feature flags regardless of the revision level, since we
950          * previously didn't change the revision level when setting the flags,
951          * so there is a chance incompat flags are set on a rev 0 filesystem.
952          */
953         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
954         if (features) {
955                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
956                        "unsupported optional features (%x)",
957                         le32_to_cpu(features));
958                 goto failed_mount;
959         }
960         if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
961                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
962                        "unsupported optional features (%x)",
963                        le32_to_cpu(features));
964                 goto failed_mount;
965         }
966
967         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
968
969         if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
970                 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
971                         ext2_msg(sb, KERN_ERR,
972                                 "DAX unsupported by block device. Turning off DAX.");
973                         sbi->s_mount_opt &= ~EXT2_MOUNT_DAX;
974                 }
975         }
976
977         /* If the blocksize doesn't match, re-read the thing.. */
978         if (sb->s_blocksize != blocksize) {
979                 brelse(bh);
980
981                 if (!sb_set_blocksize(sb, blocksize)) {
982                         ext2_msg(sb, KERN_ERR,
983                                 "error: bad blocksize %d", blocksize);
984                         goto failed_sbi;
985                 }
986
987                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
988                 offset = (sb_block*BLOCK_SIZE) % blocksize;
989                 bh = sb_bread(sb, logic_sb_block);
990                 if(!bh) {
991                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
992                                 "superblock on 2nd try");
993                         goto failed_sbi;
994                 }
995                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
996                 sbi->s_es = es;
997                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
998                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
999                         goto failed_mount;
1000                 }
1001         }
1002
1003         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
1004         sb->s_max_links = EXT2_LINK_MAX;
1005
1006         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
1007                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
1008                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
1009         } else {
1010                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1011                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1012                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1013                     !is_power_of_2(sbi->s_inode_size) ||
1014                     (sbi->s_inode_size > blocksize)) {
1015                         ext2_msg(sb, KERN_ERR,
1016                                 "error: unsupported inode size: %d",
1017                                 sbi->s_inode_size);
1018                         goto failed_mount;
1019                 }
1020         }
1021
1022         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1023                                    le32_to_cpu(es->s_log_frag_size);
1024         if (sbi->s_frag_size == 0)
1025                 goto cantfind_ext2;
1026         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1027
1028         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1029         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1030         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1031
1032         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1033         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1034                 goto cantfind_ext2;
1035         sbi->s_itb_per_group = sbi->s_inodes_per_group /
1036                                         sbi->s_inodes_per_block;
1037         sbi->s_desc_per_block = sb->s_blocksize /
1038                                         sizeof (struct ext2_group_desc);
1039         sbi->s_sbh = bh;
1040         sbi->s_mount_state = le16_to_cpu(es->s_state);
1041         sbi->s_addr_per_block_bits =
1042                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1043         sbi->s_desc_per_block_bits =
1044                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
1045
1046         if (sb->s_magic != EXT2_SUPER_MAGIC)
1047                 goto cantfind_ext2;
1048
1049         if (sb->s_blocksize != bh->b_size) {
1050                 if (!silent)
1051                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1052                 goto failed_mount;
1053         }
1054
1055         if (sb->s_blocksize != sbi->s_frag_size) {
1056                 ext2_msg(sb, KERN_ERR,
1057                         "error: fragsize %lu != blocksize %lu"
1058                         "(not supported yet)",
1059                         sbi->s_frag_size, sb->s_blocksize);
1060                 goto failed_mount;
1061         }
1062
1063         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1064                 ext2_msg(sb, KERN_ERR,
1065                         "error: #blocks per group too big: %lu",
1066                         sbi->s_blocks_per_group);
1067                 goto failed_mount;
1068         }
1069         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1070                 ext2_msg(sb, KERN_ERR,
1071                         "error: #fragments per group too big: %lu",
1072                         sbi->s_frags_per_group);
1073                 goto failed_mount;
1074         }
1075         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1076                 ext2_msg(sb, KERN_ERR,
1077                         "error: #inodes per group too big: %lu",
1078                         sbi->s_inodes_per_group);
1079                 goto failed_mount;
1080         }
1081
1082         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1083                 goto cantfind_ext2;
1084         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1085                                 le32_to_cpu(es->s_first_data_block) - 1)
1086                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1087         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1088                    EXT2_DESC_PER_BLOCK(sb);
1089         sbi->s_group_desc = kmalloc_array (db_count,
1090                                            sizeof(struct buffer_head *),
1091                                            GFP_KERNEL);
1092         if (sbi->s_group_desc == NULL) {
1093                 ret = -ENOMEM;
1094                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1095                 goto failed_mount;
1096         }
1097         bgl_lock_init(sbi->s_blockgroup_lock);
1098         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1099         if (!sbi->s_debts) {
1100                 ret = -ENOMEM;
1101                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1102                 goto failed_mount_group_desc;
1103         }
1104         for (i = 0; i < db_count; i++) {
1105                 block = descriptor_loc(sb, logic_sb_block, i);
1106                 sbi->s_group_desc[i] = sb_bread(sb, block);
1107                 if (!sbi->s_group_desc[i]) {
1108                         for (j = 0; j < i; j++)
1109                                 brelse (sbi->s_group_desc[j]);
1110                         ext2_msg(sb, KERN_ERR,
1111                                 "error: unable to read group descriptors");
1112                         goto failed_mount_group_desc;
1113                 }
1114         }
1115         if (!ext2_check_descriptors (sb)) {
1116                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1117                 goto failed_mount2;
1118         }
1119         sbi->s_gdb_count = db_count;
1120         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1121         spin_lock_init(&sbi->s_next_gen_lock);
1122
1123         /* per fileystem reservation list head & lock */
1124         spin_lock_init(&sbi->s_rsv_window_lock);
1125         sbi->s_rsv_window_root = RB_ROOT;
1126         /*
1127          * Add a single, static dummy reservation to the start of the
1128          * reservation window list --- it gives us a placeholder for
1129          * append-at-start-of-list which makes the allocation logic
1130          * _much_ simpler.
1131          */
1132         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1133         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1134         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1135         sbi->s_rsv_window_head.rsv_goal_size = 0;
1136         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1137
1138         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1139                                 ext2_count_free_blocks(sb), GFP_KERNEL);
1140         if (!err) {
1141                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1142                                 ext2_count_free_inodes(sb), GFP_KERNEL);
1143         }
1144         if (!err) {
1145                 err = percpu_counter_init(&sbi->s_dirs_counter,
1146                                 ext2_count_dirs(sb), GFP_KERNEL);
1147         }
1148         if (err) {
1149                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1150                 goto failed_mount3;
1151         }
1152
1153 #ifdef CONFIG_EXT2_FS_XATTR
1154         sbi->s_ea_block_cache = ext2_xattr_create_cache();
1155         if (!sbi->s_ea_block_cache) {
1156                 ret = -ENOMEM;
1157                 ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1158                 goto failed_mount3;
1159         }
1160 #endif
1161         /*
1162          * set up enough so that it can read an inode
1163          */
1164         sb->s_op = &ext2_sops;
1165         sb->s_export_op = &ext2_export_ops;
1166         sb->s_xattr = ext2_xattr_handlers;
1167
1168 #ifdef CONFIG_QUOTA
1169         sb->dq_op = &dquot_operations;
1170         sb->s_qcop = &ext2_quotactl_ops;
1171         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1172 #endif
1173
1174         root = ext2_iget(sb, EXT2_ROOT_INO);
1175         if (IS_ERR(root)) {
1176                 ret = PTR_ERR(root);
1177                 goto failed_mount3;
1178         }
1179         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1180                 iput(root);
1181                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1182                 goto failed_mount3;
1183         }
1184
1185         sb->s_root = d_make_root(root);
1186         if (!sb->s_root) {
1187                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1188                 ret = -ENOMEM;
1189                 goto failed_mount3;
1190         }
1191         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1192                 ext2_msg(sb, KERN_WARNING,
1193                         "warning: mounting ext3 filesystem as ext2");
1194         if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1195                 sb->s_flags |= SB_RDONLY;
1196         ext2_write_super(sb);
1197         return 0;
1198
1199 cantfind_ext2:
1200         if (!silent)
1201                 ext2_msg(sb, KERN_ERR,
1202                         "error: can't find an ext2 filesystem on dev %s.",
1203                         sb->s_id);
1204         goto failed_mount;
1205 failed_mount3:
1206         ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1207         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1208         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1209         percpu_counter_destroy(&sbi->s_dirs_counter);
1210 failed_mount2:
1211         for (i = 0; i < db_count; i++)
1212                 brelse(sbi->s_group_desc[i]);
1213 failed_mount_group_desc:
1214         kfree(sbi->s_group_desc);
1215         kfree(sbi->s_debts);
1216 failed_mount:
1217         brelse(bh);
1218 failed_sbi:
1219         sb->s_fs_info = NULL;
1220         kfree(sbi->s_blockgroup_lock);
1221         kfree(sbi);
1222 failed:
1223         fs_put_dax(dax_dev);
1224         return ret;
1225 }
1226
1227 static void ext2_clear_super_error(struct super_block *sb)
1228 {
1229         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1230
1231         if (buffer_write_io_error(sbh)) {
1232                 /*
1233                  * Oh, dear.  A previous attempt to write the
1234                  * superblock failed.  This could happen because the
1235                  * USB device was yanked out.  Or it could happen to
1236                  * be a transient write error and maybe the block will
1237                  * be remapped.  Nothing we can do but to retry the
1238                  * write and hope for the best.
1239                  */
1240                 ext2_msg(sb, KERN_ERR,
1241                        "previous I/O error to superblock detected");
1242                 clear_buffer_write_io_error(sbh);
1243                 set_buffer_uptodate(sbh);
1244         }
1245 }
1246
1247 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1248                      int wait)
1249 {
1250         ext2_clear_super_error(sb);
1251         spin_lock(&EXT2_SB(sb)->s_lock);
1252         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1253         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1254         es->s_wtime = cpu_to_le32(ktime_get_real_seconds());
1255         /* unlock before we do IO */
1256         spin_unlock(&EXT2_SB(sb)->s_lock);
1257         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1258         if (wait)
1259                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1260 }
1261
1262 /*
1263  * In the second extended file system, it is not necessary to
1264  * write the super block since we use a mapping of the
1265  * disk super block in a buffer.
1266  *
1267  * However, this function is still used to set the fs valid
1268  * flags to 0.  We need to set this flag to 0 since the fs
1269  * may have been checked while mounted and e2fsck may have
1270  * set s_state to EXT2_VALID_FS after some corrections.
1271  */
1272 static int ext2_sync_fs(struct super_block *sb, int wait)
1273 {
1274         struct ext2_sb_info *sbi = EXT2_SB(sb);
1275         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1276
1277         /*
1278          * Write quota structures to quota file, sync_blockdev() will write
1279          * them to disk later
1280          */
1281         dquot_writeback_dquots(sb, -1);
1282
1283         spin_lock(&sbi->s_lock);
1284         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1285                 ext2_debug("setting valid to 0\n");
1286                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1287         }
1288         spin_unlock(&sbi->s_lock);
1289         ext2_sync_super(sb, es, wait);
1290         return 0;
1291 }
1292
1293 static int ext2_freeze(struct super_block *sb)
1294 {
1295         struct ext2_sb_info *sbi = EXT2_SB(sb);
1296
1297         /*
1298          * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1299          * because we have unattached inodes and thus filesystem is not fully
1300          * consistent.
1301          */
1302         if (atomic_long_read(&sb->s_remove_count)) {
1303                 ext2_sync_fs(sb, 1);
1304                 return 0;
1305         }
1306         /* Set EXT2_FS_VALID flag */
1307         spin_lock(&sbi->s_lock);
1308         sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1309         spin_unlock(&sbi->s_lock);
1310         ext2_sync_super(sb, sbi->s_es, 1);
1311
1312         return 0;
1313 }
1314
1315 static int ext2_unfreeze(struct super_block *sb)
1316 {
1317         /* Just write sb to clear EXT2_VALID_FS flag */
1318         ext2_write_super(sb);
1319
1320         return 0;
1321 }
1322
1323 static void ext2_write_super(struct super_block *sb)
1324 {
1325         if (!sb_rdonly(sb))
1326                 ext2_sync_fs(sb, 1);
1327 }
1328
1329 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1330 {
1331         struct ext2_sb_info * sbi = EXT2_SB(sb);
1332         struct ext2_super_block * es;
1333         struct ext2_mount_options new_opts;
1334         int err;
1335
1336         sync_filesystem(sb);
1337
1338         spin_lock(&sbi->s_lock);
1339         new_opts.s_mount_opt = sbi->s_mount_opt;
1340         new_opts.s_resuid = sbi->s_resuid;
1341         new_opts.s_resgid = sbi->s_resgid;
1342         spin_unlock(&sbi->s_lock);
1343
1344         if (!parse_options(data, sb, &new_opts))
1345                 return -EINVAL;
1346
1347         spin_lock(&sbi->s_lock);
1348         es = sbi->s_es;
1349         if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1350                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1351                          "dax flag with busy inodes while remounting");
1352                 new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1353         }
1354         if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1355                 goto out_set;
1356         if (*flags & SB_RDONLY) {
1357                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1358                     !(sbi->s_mount_state & EXT2_VALID_FS))
1359                         goto out_set;
1360
1361                 /*
1362                  * OK, we are remounting a valid rw partition rdonly, so set
1363                  * the rdonly flag and then mark the partition as valid again.
1364                  */
1365                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1366                 es->s_mtime = cpu_to_le32(ktime_get_real_seconds());
1367                 spin_unlock(&sbi->s_lock);
1368
1369                 err = dquot_suspend(sb, -1);
1370                 if (err < 0)
1371                         return err;
1372
1373                 ext2_sync_super(sb, es, 1);
1374         } else {
1375                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1376                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1377                 if (ret) {
1378                         spin_unlock(&sbi->s_lock);
1379                         ext2_msg(sb, KERN_WARNING,
1380                                 "warning: couldn't remount RDWR because of "
1381                                 "unsupported optional features (%x).",
1382                                 le32_to_cpu(ret));
1383                         return -EROFS;
1384                 }
1385                 /*
1386                  * Mounting a RDONLY partition read-write, so reread and
1387                  * store the current valid flag.  (It may have been changed
1388                  * by e2fsck since we originally mounted the partition.)
1389                  */
1390                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1391                 if (!ext2_setup_super (sb, es, 0))
1392                         sb->s_flags &= ~SB_RDONLY;
1393                 spin_unlock(&sbi->s_lock);
1394
1395                 ext2_write_super(sb);
1396
1397                 dquot_resume(sb, -1);
1398         }
1399
1400         spin_lock(&sbi->s_lock);
1401 out_set:
1402         sbi->s_mount_opt = new_opts.s_mount_opt;
1403         sbi->s_resuid = new_opts.s_resuid;
1404         sbi->s_resgid = new_opts.s_resgid;
1405         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1406                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1407         spin_unlock(&sbi->s_lock);
1408
1409         return 0;
1410 }
1411
1412 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1413 {
1414         struct super_block *sb = dentry->d_sb;
1415         struct ext2_sb_info *sbi = EXT2_SB(sb);
1416         struct ext2_super_block *es = sbi->s_es;
1417         u64 fsid;
1418
1419         spin_lock(&sbi->s_lock);
1420
1421         if (test_opt (sb, MINIX_DF))
1422                 sbi->s_overhead_last = 0;
1423         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1424                 unsigned long i, overhead = 0;
1425                 smp_rmb();
1426
1427                 /*
1428                  * Compute the overhead (FS structures). This is constant
1429                  * for a given filesystem unless the number of block groups
1430                  * changes so we cache the previous value until it does.
1431                  */
1432
1433                 /*
1434                  * All of the blocks before first_data_block are
1435                  * overhead
1436                  */
1437                 overhead = le32_to_cpu(es->s_first_data_block);
1438
1439                 /*
1440                  * Add the overhead attributed to the superblock and
1441                  * block group descriptors.  If the sparse superblocks
1442                  * feature is turned on, then not all groups have this.
1443                  */
1444                 for (i = 0; i < sbi->s_groups_count; i++)
1445                         overhead += ext2_bg_has_super(sb, i) +
1446                                 ext2_bg_num_gdb(sb, i);
1447
1448                 /*
1449                  * Every block group has an inode bitmap, a block
1450                  * bitmap, and an inode table.
1451                  */
1452                 overhead += (sbi->s_groups_count *
1453                              (2 + sbi->s_itb_per_group));
1454                 sbi->s_overhead_last = overhead;
1455                 smp_wmb();
1456                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1457         }
1458
1459         buf->f_type = EXT2_SUPER_MAGIC;
1460         buf->f_bsize = sb->s_blocksize;
1461         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1462         buf->f_bfree = ext2_count_free_blocks(sb);
1463         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1464         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1465         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1466                 buf->f_bavail = 0;
1467         buf->f_files = le32_to_cpu(es->s_inodes_count);
1468         buf->f_ffree = ext2_count_free_inodes(sb);
1469         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1470         buf->f_namelen = EXT2_NAME_LEN;
1471         fsid = le64_to_cpup((void *)es->s_uuid) ^
1472                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1473         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1474         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1475         spin_unlock(&sbi->s_lock);
1476         return 0;
1477 }
1478
1479 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1480         int flags, const char *dev_name, void *data)
1481 {
1482         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1483 }
1484
1485 #ifdef CONFIG_QUOTA
1486
1487 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1488  * acquiring the locks... As quota files are never truncated and quota code
1489  * itself serializes the operations (and no one else should touch the files)
1490  * we don't have to be afraid of races */
1491 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1492                                size_t len, loff_t off)
1493 {
1494         struct inode *inode = sb_dqopt(sb)->files[type];
1495         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1496         int err = 0;
1497         int offset = off & (sb->s_blocksize - 1);
1498         int tocopy;
1499         size_t toread;
1500         struct buffer_head tmp_bh;
1501         struct buffer_head *bh;
1502         loff_t i_size = i_size_read(inode);
1503
1504         if (off > i_size)
1505                 return 0;
1506         if (off+len > i_size)
1507                 len = i_size-off;
1508         toread = len;
1509         while (toread > 0) {
1510                 tocopy = sb->s_blocksize - offset < toread ?
1511                                 sb->s_blocksize - offset : toread;
1512
1513                 tmp_bh.b_state = 0;
1514                 tmp_bh.b_size = sb->s_blocksize;
1515                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1516                 if (err < 0)
1517                         return err;
1518                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1519                         memset(data, 0, tocopy);
1520                 else {
1521                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1522                         if (!bh)
1523                                 return -EIO;
1524                         memcpy(data, bh->b_data+offset, tocopy);
1525                         brelse(bh);
1526                 }
1527                 offset = 0;
1528                 toread -= tocopy;
1529                 data += tocopy;
1530                 blk++;
1531         }
1532         return len;
1533 }
1534
1535 /* Write to quotafile */
1536 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1537                                 const char *data, size_t len, loff_t off)
1538 {
1539         struct inode *inode = sb_dqopt(sb)->files[type];
1540         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1541         int err = 0;
1542         int offset = off & (sb->s_blocksize - 1);
1543         int tocopy;
1544         size_t towrite = len;
1545         struct buffer_head tmp_bh;
1546         struct buffer_head *bh;
1547
1548         while (towrite > 0) {
1549                 tocopy = sb->s_blocksize - offset < towrite ?
1550                                 sb->s_blocksize - offset : towrite;
1551
1552                 tmp_bh.b_state = 0;
1553                 tmp_bh.b_size = sb->s_blocksize;
1554                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1555                 if (err < 0)
1556                         goto out;
1557                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1558                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1559                 else
1560                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1561                 if (unlikely(!bh)) {
1562                         err = -EIO;
1563                         goto out;
1564                 }
1565                 lock_buffer(bh);
1566                 memcpy(bh->b_data+offset, data, tocopy);
1567                 flush_dcache_page(bh->b_page);
1568                 set_buffer_uptodate(bh);
1569                 mark_buffer_dirty(bh);
1570                 unlock_buffer(bh);
1571                 brelse(bh);
1572                 offset = 0;
1573                 towrite -= tocopy;
1574                 data += tocopy;
1575                 blk++;
1576         }
1577 out:
1578         if (len == towrite)
1579                 return err;
1580         if (inode->i_size < off+len-towrite)
1581                 i_size_write(inode, off+len-towrite);
1582         inode_inc_iversion(inode);
1583         inode->i_mtime = inode->i_ctime = current_time(inode);
1584         mark_inode_dirty(inode);
1585         return len - towrite;
1586 }
1587
1588 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1589                          const struct path *path)
1590 {
1591         int err;
1592         struct inode *inode;
1593
1594         err = dquot_quota_on(sb, type, format_id, path);
1595         if (err)
1596                 return err;
1597
1598         inode = d_inode(path->dentry);
1599         inode_lock(inode);
1600         EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1601         inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1602                         S_NOATIME | S_IMMUTABLE);
1603         inode_unlock(inode);
1604         mark_inode_dirty(inode);
1605
1606         return 0;
1607 }
1608
1609 static int ext2_quota_off(struct super_block *sb, int type)
1610 {
1611         struct inode *inode = sb_dqopt(sb)->files[type];
1612         int err;
1613
1614         if (!inode || !igrab(inode))
1615                 goto out;
1616
1617         err = dquot_quota_off(sb, type);
1618         if (err)
1619                 goto out_put;
1620
1621         inode_lock(inode);
1622         EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1623         inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1624         inode_unlock(inode);
1625         mark_inode_dirty(inode);
1626 out_put:
1627         iput(inode);
1628         return err;
1629 out:
1630         return dquot_quota_off(sb, type);
1631 }
1632
1633 #endif
1634
1635 static struct file_system_type ext2_fs_type = {
1636         .owner          = THIS_MODULE,
1637         .name           = "ext2",
1638         .mount          = ext2_mount,
1639         .kill_sb        = kill_block_super,
1640         .fs_flags       = FS_REQUIRES_DEV,
1641 };
1642 MODULE_ALIAS_FS("ext2");
1643
1644 static int __init init_ext2_fs(void)
1645 {
1646         int err;
1647
1648         err = init_inodecache();
1649         if (err)
1650                 return err;
1651         err = register_filesystem(&ext2_fs_type);
1652         if (err)
1653                 goto out;
1654         return 0;
1655 out:
1656         destroy_inodecache();
1657         return err;
1658 }
1659
1660 static void __exit exit_ext2_fs(void)
1661 {
1662         unregister_filesystem(&ext2_fs_type);
1663         destroy_inodecache();
1664 }
1665
1666 MODULE_AUTHOR("Remy Card and others");
1667 MODULE_DESCRIPTION("Second Extended Filesystem");
1668 MODULE_LICENSE("GPL");
1669 module_init(init_ext2_fs)
1670 module_exit(exit_ext2_fs)