switch open-coded instances of d_make_root() to new helper
[linux-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 <asm/uaccess.h>
35 #include "ext2.h"
36 #include "xattr.h"
37 #include "acl.h"
38 #include "xip.h"
39
40 static void ext2_sync_super(struct super_block *sb,
41                             struct ext2_super_block *es, int wait);
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
46 void ext2_error(struct super_block *sb, const char *function,
47                 const char *fmt, ...)
48 {
49         struct va_format vaf;
50         va_list args;
51         struct ext2_sb_info *sbi = EXT2_SB(sb);
52         struct ext2_super_block *es = sbi->s_es;
53
54         if (!(sb->s_flags & MS_RDONLY)) {
55                 spin_lock(&sbi->s_lock);
56                 sbi->s_mount_state |= EXT2_ERROR_FS;
57                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
58                 spin_unlock(&sbi->s_lock);
59                 ext2_sync_super(sb, es, 1);
60         }
61
62         va_start(args, fmt);
63
64         vaf.fmt = fmt;
65         vaf.va = &args;
66
67         printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
68                sb->s_id, function, &vaf);
69
70         va_end(args);
71
72         if (test_opt(sb, ERRORS_PANIC))
73                 panic("EXT2-fs: panic from previous error\n");
74         if (test_opt(sb, ERRORS_RO)) {
75                 ext2_msg(sb, KERN_CRIT,
76                              "error: remounting filesystem read-only");
77                 sb->s_flags |= MS_RDONLY;
78         }
79 }
80
81 void ext2_msg(struct super_block *sb, const char *prefix,
82                 const char *fmt, ...)
83 {
84         struct va_format vaf;
85         va_list args;
86
87         va_start(args, fmt);
88
89         vaf.fmt = fmt;
90         vaf.va = &args;
91
92         printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
93
94         va_end(args);
95 }
96
97 /*
98  * This must be called with sbi->s_lock held.
99  */
100 void ext2_update_dynamic_rev(struct super_block *sb)
101 {
102         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
103
104         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
105                 return;
106
107         ext2_msg(sb, KERN_WARNING,
108                      "warning: updating to rev %d because of "
109                      "new feature flag, running e2fsck is recommended",
110                      EXT2_DYNAMIC_REV);
111
112         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
113         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
114         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
115         /* leave es->s_feature_*compat flags alone */
116         /* es->s_uuid will be set by e2fsck if empty */
117
118         /*
119          * The rest of the superblock fields should be zero, and if not it
120          * means they are likely already in use, so leave them alone.  We
121          * can leave it up to e2fsck to clean up any inconsistencies there.
122          */
123 }
124
125 static void ext2_put_super (struct super_block * sb)
126 {
127         int db_count;
128         int i;
129         struct ext2_sb_info *sbi = EXT2_SB(sb);
130
131         dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
132
133         if (sb->s_dirt)
134                 ext2_write_super(sb);
135
136         ext2_xattr_put_super(sb);
137         if (!(sb->s_flags & MS_RDONLY)) {
138                 struct ext2_super_block *es = sbi->s_es;
139
140                 spin_lock(&sbi->s_lock);
141                 es->s_state = cpu_to_le16(sbi->s_mount_state);
142                 spin_unlock(&sbi->s_lock);
143                 ext2_sync_super(sb, es, 1);
144         }
145         db_count = sbi->s_gdb_count;
146         for (i = 0; i < db_count; i++)
147                 if (sbi->s_group_desc[i])
148                         brelse (sbi->s_group_desc[i]);
149         kfree(sbi->s_group_desc);
150         kfree(sbi->s_debts);
151         percpu_counter_destroy(&sbi->s_freeblocks_counter);
152         percpu_counter_destroy(&sbi->s_freeinodes_counter);
153         percpu_counter_destroy(&sbi->s_dirs_counter);
154         brelse (sbi->s_sbh);
155         sb->s_fs_info = NULL;
156         kfree(sbi->s_blockgroup_lock);
157         kfree(sbi);
158 }
159
160 static struct kmem_cache * ext2_inode_cachep;
161
162 static struct inode *ext2_alloc_inode(struct super_block *sb)
163 {
164         struct ext2_inode_info *ei;
165         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
166         if (!ei)
167                 return NULL;
168         ei->i_block_alloc_info = NULL;
169         ei->vfs_inode.i_version = 1;
170         return &ei->vfs_inode;
171 }
172
173 static void ext2_i_callback(struct rcu_head *head)
174 {
175         struct inode *inode = container_of(head, struct inode, i_rcu);
176         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
177 }
178
179 static void ext2_destroy_inode(struct inode *inode)
180 {
181         call_rcu(&inode->i_rcu, ext2_i_callback);
182 }
183
184 static void init_once(void *foo)
185 {
186         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
187
188         rwlock_init(&ei->i_meta_lock);
189 #ifdef CONFIG_EXT2_FS_XATTR
190         init_rwsem(&ei->xattr_sem);
191 #endif
192         mutex_init(&ei->truncate_mutex);
193         inode_init_once(&ei->vfs_inode);
194 }
195
196 static int init_inodecache(void)
197 {
198         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
199                                              sizeof(struct ext2_inode_info),
200                                              0, (SLAB_RECLAIM_ACCOUNT|
201                                                 SLAB_MEM_SPREAD),
202                                              init_once);
203         if (ext2_inode_cachep == NULL)
204                 return -ENOMEM;
205         return 0;
206 }
207
208 static void destroy_inodecache(void)
209 {
210         kmem_cache_destroy(ext2_inode_cachep);
211 }
212
213 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
214 {
215         struct super_block *sb = root->d_sb;
216         struct ext2_sb_info *sbi = EXT2_SB(sb);
217         struct ext2_super_block *es = sbi->s_es;
218         unsigned long def_mount_opts;
219
220         spin_lock(&sbi->s_lock);
221         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
222
223         if (sbi->s_sb_block != 1)
224                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
225         if (test_opt(sb, MINIX_DF))
226                 seq_puts(seq, ",minixdf");
227         if (test_opt(sb, GRPID))
228                 seq_puts(seq, ",grpid");
229         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
230                 seq_puts(seq, ",nogrpid");
231         if (sbi->s_resuid != EXT2_DEF_RESUID ||
232             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
233                 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
234         }
235         if (sbi->s_resgid != EXT2_DEF_RESGID ||
236             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
237                 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
238         }
239         if (test_opt(sb, ERRORS_RO)) {
240                 int def_errors = le16_to_cpu(es->s_errors);
241
242                 if (def_errors == EXT2_ERRORS_PANIC ||
243                     def_errors == EXT2_ERRORS_CONTINUE) {
244                         seq_puts(seq, ",errors=remount-ro");
245                 }
246         }
247         if (test_opt(sb, ERRORS_CONT))
248                 seq_puts(seq, ",errors=continue");
249         if (test_opt(sb, ERRORS_PANIC))
250                 seq_puts(seq, ",errors=panic");
251         if (test_opt(sb, NO_UID32))
252                 seq_puts(seq, ",nouid32");
253         if (test_opt(sb, DEBUG))
254                 seq_puts(seq, ",debug");
255         if (test_opt(sb, OLDALLOC))
256                 seq_puts(seq, ",oldalloc");
257
258 #ifdef CONFIG_EXT2_FS_XATTR
259         if (test_opt(sb, XATTR_USER))
260                 seq_puts(seq, ",user_xattr");
261         if (!test_opt(sb, XATTR_USER) &&
262             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
263                 seq_puts(seq, ",nouser_xattr");
264         }
265 #endif
266
267 #ifdef CONFIG_EXT2_FS_POSIX_ACL
268         if (test_opt(sb, POSIX_ACL))
269                 seq_puts(seq, ",acl");
270         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
271                 seq_puts(seq, ",noacl");
272 #endif
273
274         if (test_opt(sb, NOBH))
275                 seq_puts(seq, ",nobh");
276
277 #if defined(CONFIG_QUOTA)
278         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
279                 seq_puts(seq, ",usrquota");
280
281         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
282                 seq_puts(seq, ",grpquota");
283 #endif
284
285 #if defined(CONFIG_EXT2_FS_XIP)
286         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
287                 seq_puts(seq, ",xip");
288 #endif
289
290         if (!test_opt(sb, RESERVATION))
291                 seq_puts(seq, ",noreservation");
292
293         spin_unlock(&sbi->s_lock);
294         return 0;
295 }
296
297 #ifdef CONFIG_QUOTA
298 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
299 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
300 #endif
301
302 static const struct super_operations ext2_sops = {
303         .alloc_inode    = ext2_alloc_inode,
304         .destroy_inode  = ext2_destroy_inode,
305         .write_inode    = ext2_write_inode,
306         .evict_inode    = ext2_evict_inode,
307         .put_super      = ext2_put_super,
308         .write_super    = ext2_write_super,
309         .sync_fs        = ext2_sync_fs,
310         .statfs         = ext2_statfs,
311         .remount_fs     = ext2_remount,
312         .show_options   = ext2_show_options,
313 #ifdef CONFIG_QUOTA
314         .quota_read     = ext2_quota_read,
315         .quota_write    = ext2_quota_write,
316 #endif
317 };
318
319 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
320                 u64 ino, u32 generation)
321 {
322         struct inode *inode;
323
324         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
325                 return ERR_PTR(-ESTALE);
326         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
327                 return ERR_PTR(-ESTALE);
328
329         /*
330          * ext2_iget isn't quite right if the inode is currently unallocated!
331          * However ext2_iget currently does appropriate checks to handle stale
332          * inodes so everything is OK.
333          */
334         inode = ext2_iget(sb, ino);
335         if (IS_ERR(inode))
336                 return ERR_CAST(inode);
337         if (generation && inode->i_generation != generation) {
338                 /* we didn't find the right inode.. */
339                 iput(inode);
340                 return ERR_PTR(-ESTALE);
341         }
342         return inode;
343 }
344
345 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
346                 int fh_len, int fh_type)
347 {
348         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
349                                     ext2_nfs_get_inode);
350 }
351
352 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
353                 int fh_len, int fh_type)
354 {
355         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
356                                     ext2_nfs_get_inode);
357 }
358
359 /* Yes, most of these are left as NULL!!
360  * A NULL value implies the default, which works with ext2-like file
361  * systems, but can be improved upon.
362  * Currently only get_parent is required.
363  */
364 static const struct export_operations ext2_export_ops = {
365         .fh_to_dentry = ext2_fh_to_dentry,
366         .fh_to_parent = ext2_fh_to_parent,
367         .get_parent = ext2_get_parent,
368 };
369
370 static unsigned long get_sb_block(void **data)
371 {
372         unsigned long   sb_block;
373         char            *options = (char *) *data;
374
375         if (!options || strncmp(options, "sb=", 3) != 0)
376                 return 1;       /* Default location */
377         options += 3;
378         sb_block = simple_strtoul(options, &options, 0);
379         if (*options && *options != ',') {
380                 printk("EXT2-fs: Invalid sb specification: %s\n",
381                        (char *) *data);
382                 return 1;
383         }
384         if (*options == ',')
385                 options++;
386         *data = (void *) options;
387         return sb_block;
388 }
389
390 enum {
391         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
392         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
393         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
394         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
395         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
396         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
397 };
398
399 static const match_table_t tokens = {
400         {Opt_bsd_df, "bsddf"},
401         {Opt_minix_df, "minixdf"},
402         {Opt_grpid, "grpid"},
403         {Opt_grpid, "bsdgroups"},
404         {Opt_nogrpid, "nogrpid"},
405         {Opt_nogrpid, "sysvgroups"},
406         {Opt_resgid, "resgid=%u"},
407         {Opt_resuid, "resuid=%u"},
408         {Opt_sb, "sb=%u"},
409         {Opt_err_cont, "errors=continue"},
410         {Opt_err_panic, "errors=panic"},
411         {Opt_err_ro, "errors=remount-ro"},
412         {Opt_nouid32, "nouid32"},
413         {Opt_nocheck, "check=none"},
414         {Opt_nocheck, "nocheck"},
415         {Opt_debug, "debug"},
416         {Opt_oldalloc, "oldalloc"},
417         {Opt_orlov, "orlov"},
418         {Opt_nobh, "nobh"},
419         {Opt_user_xattr, "user_xattr"},
420         {Opt_nouser_xattr, "nouser_xattr"},
421         {Opt_acl, "acl"},
422         {Opt_noacl, "noacl"},
423         {Opt_xip, "xip"},
424         {Opt_grpquota, "grpquota"},
425         {Opt_ignore, "noquota"},
426         {Opt_quota, "quota"},
427         {Opt_usrquota, "usrquota"},
428         {Opt_reservation, "reservation"},
429         {Opt_noreservation, "noreservation"},
430         {Opt_err, NULL}
431 };
432
433 static int parse_options(char *options, struct super_block *sb)
434 {
435         char *p;
436         struct ext2_sb_info *sbi = EXT2_SB(sb);
437         substring_t args[MAX_OPT_ARGS];
438         int option;
439
440         if (!options)
441                 return 1;
442
443         while ((p = strsep (&options, ",")) != NULL) {
444                 int token;
445                 if (!*p)
446                         continue;
447
448                 token = match_token(p, tokens, args);
449                 switch (token) {
450                 case Opt_bsd_df:
451                         clear_opt (sbi->s_mount_opt, MINIX_DF);
452                         break;
453                 case Opt_minix_df:
454                         set_opt (sbi->s_mount_opt, MINIX_DF);
455                         break;
456                 case Opt_grpid:
457                         set_opt (sbi->s_mount_opt, GRPID);
458                         break;
459                 case Opt_nogrpid:
460                         clear_opt (sbi->s_mount_opt, GRPID);
461                         break;
462                 case Opt_resuid:
463                         if (match_int(&args[0], &option))
464                                 return 0;
465                         sbi->s_resuid = option;
466                         break;
467                 case Opt_resgid:
468                         if (match_int(&args[0], &option))
469                                 return 0;
470                         sbi->s_resgid = option;
471                         break;
472                 case Opt_sb:
473                         /* handled by get_sb_block() instead of here */
474                         /* *sb_block = match_int(&args[0]); */
475                         break;
476                 case Opt_err_panic:
477                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
478                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
479                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
480                         break;
481                 case Opt_err_ro:
482                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
483                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
484                         set_opt (sbi->s_mount_opt, ERRORS_RO);
485                         break;
486                 case Opt_err_cont:
487                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
488                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
489                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
490                         break;
491                 case Opt_nouid32:
492                         set_opt (sbi->s_mount_opt, NO_UID32);
493                         break;
494                 case Opt_nocheck:
495                         clear_opt (sbi->s_mount_opt, CHECK);
496                         break;
497                 case Opt_debug:
498                         set_opt (sbi->s_mount_opt, DEBUG);
499                         break;
500                 case Opt_oldalloc:
501                         set_opt (sbi->s_mount_opt, OLDALLOC);
502                         break;
503                 case Opt_orlov:
504                         clear_opt (sbi->s_mount_opt, OLDALLOC);
505                         break;
506                 case Opt_nobh:
507                         set_opt (sbi->s_mount_opt, NOBH);
508                         break;
509 #ifdef CONFIG_EXT2_FS_XATTR
510                 case Opt_user_xattr:
511                         set_opt (sbi->s_mount_opt, XATTR_USER);
512                         break;
513                 case Opt_nouser_xattr:
514                         clear_opt (sbi->s_mount_opt, XATTR_USER);
515                         break;
516 #else
517                 case Opt_user_xattr:
518                 case Opt_nouser_xattr:
519                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
520                                 "not supported");
521                         break;
522 #endif
523 #ifdef CONFIG_EXT2_FS_POSIX_ACL
524                 case Opt_acl:
525                         set_opt(sbi->s_mount_opt, POSIX_ACL);
526                         break;
527                 case Opt_noacl:
528                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
529                         break;
530 #else
531                 case Opt_acl:
532                 case Opt_noacl:
533                         ext2_msg(sb, KERN_INFO,
534                                 "(no)acl options not supported");
535                         break;
536 #endif
537                 case Opt_xip:
538 #ifdef CONFIG_EXT2_FS_XIP
539                         set_opt (sbi->s_mount_opt, XIP);
540 #else
541                         ext2_msg(sb, KERN_INFO, "xip option not supported");
542 #endif
543                         break;
544
545 #if defined(CONFIG_QUOTA)
546                 case Opt_quota:
547                 case Opt_usrquota:
548                         set_opt(sbi->s_mount_opt, USRQUOTA);
549                         break;
550
551                 case Opt_grpquota:
552                         set_opt(sbi->s_mount_opt, GRPQUOTA);
553                         break;
554 #else
555                 case Opt_quota:
556                 case Opt_usrquota:
557                 case Opt_grpquota:
558                         ext2_msg(sb, KERN_INFO,
559                                 "quota operations not supported");
560                         break;
561 #endif
562
563                 case Opt_reservation:
564                         set_opt(sbi->s_mount_opt, RESERVATION);
565                         ext2_msg(sb, KERN_INFO, "reservations ON");
566                         break;
567                 case Opt_noreservation:
568                         clear_opt(sbi->s_mount_opt, RESERVATION);
569                         ext2_msg(sb, KERN_INFO, "reservations OFF");
570                         break;
571                 case Opt_ignore:
572                         break;
573                 default:
574                         return 0;
575                 }
576         }
577         return 1;
578 }
579
580 static int ext2_setup_super (struct super_block * sb,
581                               struct ext2_super_block * es,
582                               int read_only)
583 {
584         int res = 0;
585         struct ext2_sb_info *sbi = EXT2_SB(sb);
586
587         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
588                 ext2_msg(sb, KERN_ERR,
589                         "error: revision level too high, "
590                         "forcing read-only mode");
591                 res = MS_RDONLY;
592         }
593         if (read_only)
594                 return res;
595         if (!(sbi->s_mount_state & EXT2_VALID_FS))
596                 ext2_msg(sb, KERN_WARNING,
597                         "warning: mounting unchecked fs, "
598                         "running e2fsck is recommended");
599         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
600                 ext2_msg(sb, KERN_WARNING,
601                         "warning: mounting fs with errors, "
602                         "running e2fsck is recommended");
603         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
604                  le16_to_cpu(es->s_mnt_count) >=
605                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
606                 ext2_msg(sb, KERN_WARNING,
607                         "warning: maximal mount count reached, "
608                         "running e2fsck is recommended");
609         else if (le32_to_cpu(es->s_checkinterval) &&
610                 (le32_to_cpu(es->s_lastcheck) +
611                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
612                 ext2_msg(sb, KERN_WARNING,
613                         "warning: checktime reached, "
614                         "running e2fsck is recommended");
615         if (!le16_to_cpu(es->s_max_mnt_count))
616                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
617         le16_add_cpu(&es->s_mnt_count, 1);
618         if (test_opt (sb, DEBUG))
619                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
620                         "bpg=%lu, ipg=%lu, mo=%04lx]",
621                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
622                         sbi->s_frag_size,
623                         sbi->s_groups_count,
624                         EXT2_BLOCKS_PER_GROUP(sb),
625                         EXT2_INODES_PER_GROUP(sb),
626                         sbi->s_mount_opt);
627         return res;
628 }
629
630 static int ext2_check_descriptors(struct super_block *sb)
631 {
632         int i;
633         struct ext2_sb_info *sbi = EXT2_SB(sb);
634
635         ext2_debug ("Checking group descriptors");
636
637         for (i = 0; i < sbi->s_groups_count; i++) {
638                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
639                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
640                 ext2_fsblk_t last_block;
641
642                 if (i == sbi->s_groups_count - 1)
643                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
644                 else
645                         last_block = first_block +
646                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
647
648                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
649                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
650                 {
651                         ext2_error (sb, "ext2_check_descriptors",
652                                     "Block bitmap for group %d"
653                                     " not in group (block %lu)!",
654                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
655                         return 0;
656                 }
657                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
658                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
659                 {
660                         ext2_error (sb, "ext2_check_descriptors",
661                                     "Inode bitmap for group %d"
662                                     " not in group (block %lu)!",
663                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
664                         return 0;
665                 }
666                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
667                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
668                     last_block)
669                 {
670                         ext2_error (sb, "ext2_check_descriptors",
671                                     "Inode table for group %d"
672                                     " not in group (block %lu)!",
673                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
674                         return 0;
675                 }
676         }
677         return 1;
678 }
679
680 /*
681  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
682  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
683  * We need to be 1 filesystem block less than the 2^32 sector limit.
684  */
685 static loff_t ext2_max_size(int bits)
686 {
687         loff_t res = EXT2_NDIR_BLOCKS;
688         int meta_blocks;
689         loff_t upper_limit;
690
691         /* This is calculated to be the largest file size for a
692          * dense, file such that the total number of
693          * sectors in the file, including data and all indirect blocks,
694          * does not exceed 2^32 -1
695          * __u32 i_blocks representing the total number of
696          * 512 bytes blocks of the file
697          */
698         upper_limit = (1LL << 32) - 1;
699
700         /* total blocks in file system block size */
701         upper_limit >>= (bits - 9);
702
703
704         /* indirect blocks */
705         meta_blocks = 1;
706         /* double indirect blocks */
707         meta_blocks += 1 + (1LL << (bits-2));
708         /* tripple indirect blocks */
709         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
710
711         upper_limit -= meta_blocks;
712         upper_limit <<= bits;
713
714         res += 1LL << (bits-2);
715         res += 1LL << (2*(bits-2));
716         res += 1LL << (3*(bits-2));
717         res <<= bits;
718         if (res > upper_limit)
719                 res = upper_limit;
720
721         if (res > MAX_LFS_FILESIZE)
722                 res = MAX_LFS_FILESIZE;
723
724         return res;
725 }
726
727 static unsigned long descriptor_loc(struct super_block *sb,
728                                     unsigned long logic_sb_block,
729                                     int nr)
730 {
731         struct ext2_sb_info *sbi = EXT2_SB(sb);
732         unsigned long bg, first_meta_bg;
733         int has_super = 0;
734         
735         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
736
737         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
738             nr < first_meta_bg)
739                 return (logic_sb_block + nr + 1);
740         bg = sbi->s_desc_per_block * nr;
741         if (ext2_bg_has_super(sb, bg))
742                 has_super = 1;
743
744         return ext2_group_first_block_no(sb, bg) + has_super;
745 }
746
747 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
748 {
749         struct buffer_head * bh;
750         struct ext2_sb_info * sbi;
751         struct ext2_super_block * es;
752         struct inode *root;
753         unsigned long block;
754         unsigned long sb_block = get_sb_block(&data);
755         unsigned long logic_sb_block;
756         unsigned long offset = 0;
757         unsigned long def_mount_opts;
758         long ret = -EINVAL;
759         int blocksize = BLOCK_SIZE;
760         int db_count;
761         int i, j;
762         __le32 features;
763         int err;
764
765         err = -ENOMEM;
766         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
767         if (!sbi)
768                 goto failed_unlock;
769
770         sbi->s_blockgroup_lock =
771                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
772         if (!sbi->s_blockgroup_lock) {
773                 kfree(sbi);
774                 goto failed_unlock;
775         }
776         sb->s_fs_info = sbi;
777         sbi->s_sb_block = sb_block;
778
779         spin_lock_init(&sbi->s_lock);
780
781         /*
782          * See what the current blocksize for the device is, and
783          * use that as the blocksize.  Otherwise (or if the blocksize
784          * is smaller than the default) use the default.
785          * This is important for devices that have a hardware
786          * sectorsize that is larger than the default.
787          */
788         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
789         if (!blocksize) {
790                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
791                 goto failed_sbi;
792         }
793
794         /*
795          * If the superblock doesn't start on a hardware sector boundary,
796          * calculate the offset.  
797          */
798         if (blocksize != BLOCK_SIZE) {
799                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
800                 offset = (sb_block*BLOCK_SIZE) % blocksize;
801         } else {
802                 logic_sb_block = sb_block;
803         }
804
805         if (!(bh = sb_bread(sb, logic_sb_block))) {
806                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
807                 goto failed_sbi;
808         }
809         /*
810          * Note: s_es must be initialized as soon as possible because
811          *       some ext2 macro-instructions depend on its value
812          */
813         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
814         sbi->s_es = es;
815         sb->s_magic = le16_to_cpu(es->s_magic);
816
817         if (sb->s_magic != EXT2_SUPER_MAGIC)
818                 goto cantfind_ext2;
819
820         /* Set defaults before we parse the mount options */
821         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
822         if (def_mount_opts & EXT2_DEFM_DEBUG)
823                 set_opt(sbi->s_mount_opt, DEBUG);
824         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
825                 set_opt(sbi->s_mount_opt, GRPID);
826         if (def_mount_opts & EXT2_DEFM_UID16)
827                 set_opt(sbi->s_mount_opt, NO_UID32);
828 #ifdef CONFIG_EXT2_FS_XATTR
829         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
830                 set_opt(sbi->s_mount_opt, XATTR_USER);
831 #endif
832 #ifdef CONFIG_EXT2_FS_POSIX_ACL
833         if (def_mount_opts & EXT2_DEFM_ACL)
834                 set_opt(sbi->s_mount_opt, POSIX_ACL);
835 #endif
836         
837         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
838                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
839         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
840                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
841         else
842                 set_opt(sbi->s_mount_opt, ERRORS_RO);
843
844         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
845         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
846         
847         set_opt(sbi->s_mount_opt, RESERVATION);
848
849         if (!parse_options((char *) data, sb))
850                 goto failed_mount;
851
852         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
853                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
854                  MS_POSIXACL : 0);
855
856         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
857                                     EXT2_MOUNT_XIP if not */
858
859         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
860             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
861              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
862              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
863                 ext2_msg(sb, KERN_WARNING,
864                         "warning: feature flags set on rev 0 fs, "
865                         "running e2fsck is recommended");
866         /*
867          * Check feature flags regardless of the revision level, since we
868          * previously didn't change the revision level when setting the flags,
869          * so there is a chance incompat flags are set on a rev 0 filesystem.
870          */
871         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
872         if (features) {
873                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
874                        "unsupported optional features (%x)",
875                         le32_to_cpu(features));
876                 goto failed_mount;
877         }
878         if (!(sb->s_flags & MS_RDONLY) &&
879             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
880                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
881                        "unsupported optional features (%x)",
882                        le32_to_cpu(features));
883                 goto failed_mount;
884         }
885
886         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
887
888         if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
889                 if (!silent)
890                         ext2_msg(sb, KERN_ERR,
891                                 "error: unsupported blocksize for xip");
892                 goto failed_mount;
893         }
894
895         /* If the blocksize doesn't match, re-read the thing.. */
896         if (sb->s_blocksize != blocksize) {
897                 brelse(bh);
898
899                 if (!sb_set_blocksize(sb, blocksize)) {
900                         ext2_msg(sb, KERN_ERR,
901                                 "error: bad blocksize %d", blocksize);
902                         goto failed_sbi;
903                 }
904
905                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
906                 offset = (sb_block*BLOCK_SIZE) % blocksize;
907                 bh = sb_bread(sb, logic_sb_block);
908                 if(!bh) {
909                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
910                                 "superblock on 2nd try");
911                         goto failed_sbi;
912                 }
913                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
914                 sbi->s_es = es;
915                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
916                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
917                         goto failed_mount;
918                 }
919         }
920
921         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
922         sb->s_max_links = EXT2_LINK_MAX;
923
924         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
925                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
926                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
927         } else {
928                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
929                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
930                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
931                     !is_power_of_2(sbi->s_inode_size) ||
932                     (sbi->s_inode_size > blocksize)) {
933                         ext2_msg(sb, KERN_ERR,
934                                 "error: unsupported inode size: %d",
935                                 sbi->s_inode_size);
936                         goto failed_mount;
937                 }
938         }
939
940         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
941                                    le32_to_cpu(es->s_log_frag_size);
942         if (sbi->s_frag_size == 0)
943                 goto cantfind_ext2;
944         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
945
946         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
947         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
948         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
949
950         if (EXT2_INODE_SIZE(sb) == 0)
951                 goto cantfind_ext2;
952         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
953         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
954                 goto cantfind_ext2;
955         sbi->s_itb_per_group = sbi->s_inodes_per_group /
956                                         sbi->s_inodes_per_block;
957         sbi->s_desc_per_block = sb->s_blocksize /
958                                         sizeof (struct ext2_group_desc);
959         sbi->s_sbh = bh;
960         sbi->s_mount_state = le16_to_cpu(es->s_state);
961         sbi->s_addr_per_block_bits =
962                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
963         sbi->s_desc_per_block_bits =
964                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
965
966         if (sb->s_magic != EXT2_SUPER_MAGIC)
967                 goto cantfind_ext2;
968
969         if (sb->s_blocksize != bh->b_size) {
970                 if (!silent)
971                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
972                 goto failed_mount;
973         }
974
975         if (sb->s_blocksize != sbi->s_frag_size) {
976                 ext2_msg(sb, KERN_ERR,
977                         "error: fragsize %lu != blocksize %lu"
978                         "(not supported yet)",
979                         sbi->s_frag_size, sb->s_blocksize);
980                 goto failed_mount;
981         }
982
983         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
984                 ext2_msg(sb, KERN_ERR,
985                         "error: #blocks per group too big: %lu",
986                         sbi->s_blocks_per_group);
987                 goto failed_mount;
988         }
989         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
990                 ext2_msg(sb, KERN_ERR,
991                         "error: #fragments per group too big: %lu",
992                         sbi->s_frags_per_group);
993                 goto failed_mount;
994         }
995         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
996                 ext2_msg(sb, KERN_ERR,
997                         "error: #inodes per group too big: %lu",
998                         sbi->s_inodes_per_group);
999                 goto failed_mount;
1000         }
1001
1002         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1003                 goto cantfind_ext2;
1004         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1005                                 le32_to_cpu(es->s_first_data_block) - 1)
1006                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1007         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1008                    EXT2_DESC_PER_BLOCK(sb);
1009         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1010         if (sbi->s_group_desc == NULL) {
1011                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1012                 goto failed_mount;
1013         }
1014         bgl_lock_init(sbi->s_blockgroup_lock);
1015         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1016         if (!sbi->s_debts) {
1017                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1018                 goto failed_mount_group_desc;
1019         }
1020         for (i = 0; i < db_count; i++) {
1021                 block = descriptor_loc(sb, logic_sb_block, i);
1022                 sbi->s_group_desc[i] = sb_bread(sb, block);
1023                 if (!sbi->s_group_desc[i]) {
1024                         for (j = 0; j < i; j++)
1025                                 brelse (sbi->s_group_desc[j]);
1026                         ext2_msg(sb, KERN_ERR,
1027                                 "error: unable to read group descriptors");
1028                         goto failed_mount_group_desc;
1029                 }
1030         }
1031         if (!ext2_check_descriptors (sb)) {
1032                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1033                 goto failed_mount2;
1034         }
1035         sbi->s_gdb_count = db_count;
1036         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1037         spin_lock_init(&sbi->s_next_gen_lock);
1038
1039         /* per fileystem reservation list head & lock */
1040         spin_lock_init(&sbi->s_rsv_window_lock);
1041         sbi->s_rsv_window_root = RB_ROOT;
1042         /*
1043          * Add a single, static dummy reservation to the start of the
1044          * reservation window list --- it gives us a placeholder for
1045          * append-at-start-of-list which makes the allocation logic
1046          * _much_ simpler.
1047          */
1048         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1049         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1050         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1051         sbi->s_rsv_window_head.rsv_goal_size = 0;
1052         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1053
1054         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1055                                 ext2_count_free_blocks(sb));
1056         if (!err) {
1057                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1058                                 ext2_count_free_inodes(sb));
1059         }
1060         if (!err) {
1061                 err = percpu_counter_init(&sbi->s_dirs_counter,
1062                                 ext2_count_dirs(sb));
1063         }
1064         if (err) {
1065                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1066                 goto failed_mount3;
1067         }
1068         /*
1069          * set up enough so that it can read an inode
1070          */
1071         sb->s_op = &ext2_sops;
1072         sb->s_export_op = &ext2_export_ops;
1073         sb->s_xattr = ext2_xattr_handlers;
1074
1075 #ifdef CONFIG_QUOTA
1076         sb->dq_op = &dquot_operations;
1077         sb->s_qcop = &dquot_quotactl_ops;
1078 #endif
1079
1080         root = ext2_iget(sb, EXT2_ROOT_INO);
1081         if (IS_ERR(root)) {
1082                 ret = PTR_ERR(root);
1083                 goto failed_mount3;
1084         }
1085         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1086                 iput(root);
1087                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1088                 goto failed_mount3;
1089         }
1090
1091         sb->s_root = d_make_root(root);
1092         if (!sb->s_root) {
1093                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1094                 ret = -ENOMEM;
1095                 goto failed_mount3;
1096         }
1097         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1098                 ext2_msg(sb, KERN_WARNING,
1099                         "warning: mounting ext3 filesystem as ext2");
1100         if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1101                 sb->s_flags |= MS_RDONLY;
1102         ext2_write_super(sb);
1103         return 0;
1104
1105 cantfind_ext2:
1106         if (!silent)
1107                 ext2_msg(sb, KERN_ERR,
1108                         "error: can't find an ext2 filesystem on dev %s.",
1109                         sb->s_id);
1110         goto failed_mount;
1111 failed_mount3:
1112         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1113         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1114         percpu_counter_destroy(&sbi->s_dirs_counter);
1115 failed_mount2:
1116         for (i = 0; i < db_count; i++)
1117                 brelse(sbi->s_group_desc[i]);
1118 failed_mount_group_desc:
1119         kfree(sbi->s_group_desc);
1120         kfree(sbi->s_debts);
1121 failed_mount:
1122         brelse(bh);
1123 failed_sbi:
1124         sb->s_fs_info = NULL;
1125         kfree(sbi->s_blockgroup_lock);
1126         kfree(sbi);
1127 failed_unlock:
1128         return ret;
1129 }
1130
1131 static void ext2_clear_super_error(struct super_block *sb)
1132 {
1133         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1134
1135         if (buffer_write_io_error(sbh)) {
1136                 /*
1137                  * Oh, dear.  A previous attempt to write the
1138                  * superblock failed.  This could happen because the
1139                  * USB device was yanked out.  Or it could happen to
1140                  * be a transient write error and maybe the block will
1141                  * be remapped.  Nothing we can do but to retry the
1142                  * write and hope for the best.
1143                  */
1144                 ext2_msg(sb, KERN_ERR,
1145                        "previous I/O error to superblock detected\n");
1146                 clear_buffer_write_io_error(sbh);
1147                 set_buffer_uptodate(sbh);
1148         }
1149 }
1150
1151 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1152                             int wait)
1153 {
1154         ext2_clear_super_error(sb);
1155         spin_lock(&EXT2_SB(sb)->s_lock);
1156         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1157         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1158         es->s_wtime = cpu_to_le32(get_seconds());
1159         /* unlock before we do IO */
1160         spin_unlock(&EXT2_SB(sb)->s_lock);
1161         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1162         if (wait)
1163                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1164         sb->s_dirt = 0;
1165 }
1166
1167 /*
1168  * In the second extended file system, it is not necessary to
1169  * write the super block since we use a mapping of the
1170  * disk super block in a buffer.
1171  *
1172  * However, this function is still used to set the fs valid
1173  * flags to 0.  We need to set this flag to 0 since the fs
1174  * may have been checked while mounted and e2fsck may have
1175  * set s_state to EXT2_VALID_FS after some corrections.
1176  */
1177 static int ext2_sync_fs(struct super_block *sb, int wait)
1178 {
1179         struct ext2_sb_info *sbi = EXT2_SB(sb);
1180         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1181
1182         spin_lock(&sbi->s_lock);
1183         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1184                 ext2_debug("setting valid to 0\n");
1185                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1186         }
1187         spin_unlock(&sbi->s_lock);
1188         ext2_sync_super(sb, es, wait);
1189         return 0;
1190 }
1191
1192
1193 void ext2_write_super(struct super_block *sb)
1194 {
1195         if (!(sb->s_flags & MS_RDONLY))
1196                 ext2_sync_fs(sb, 1);
1197         else
1198                 sb->s_dirt = 0;
1199 }
1200
1201 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1202 {
1203         struct ext2_sb_info * sbi = EXT2_SB(sb);
1204         struct ext2_super_block * es;
1205         unsigned long old_mount_opt = sbi->s_mount_opt;
1206         struct ext2_mount_options old_opts;
1207         unsigned long old_sb_flags;
1208         int err;
1209
1210         spin_lock(&sbi->s_lock);
1211
1212         /* Store the old options */
1213         old_sb_flags = sb->s_flags;
1214         old_opts.s_mount_opt = sbi->s_mount_opt;
1215         old_opts.s_resuid = sbi->s_resuid;
1216         old_opts.s_resgid = sbi->s_resgid;
1217
1218         /*
1219          * Allow the "check" option to be passed as a remount option.
1220          */
1221         if (!parse_options(data, sb)) {
1222                 err = -EINVAL;
1223                 goto restore_opts;
1224         }
1225
1226         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1227                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1228
1229         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1230                                     EXT2_MOUNT_XIP if not */
1231
1232         if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1233                 ext2_msg(sb, KERN_WARNING,
1234                         "warning: unsupported blocksize for xip");
1235                 err = -EINVAL;
1236                 goto restore_opts;
1237         }
1238
1239         es = sbi->s_es;
1240         if ((sbi->s_mount_opt ^ old_mount_opt) & EXT2_MOUNT_XIP) {
1241                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1242                          "xip flag with busy inodes while remounting");
1243                 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1244                 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1245         }
1246         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1247                 spin_unlock(&sbi->s_lock);
1248                 return 0;
1249         }
1250         if (*flags & MS_RDONLY) {
1251                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1252                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1253                         spin_unlock(&sbi->s_lock);
1254                         return 0;
1255                 }
1256
1257                 /*
1258                  * OK, we are remounting a valid rw partition rdonly, so set
1259                  * the rdonly flag and then mark the partition as valid again.
1260                  */
1261                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1262                 es->s_mtime = cpu_to_le32(get_seconds());
1263                 spin_unlock(&sbi->s_lock);
1264
1265                 err = dquot_suspend(sb, -1);
1266                 if (err < 0) {
1267                         spin_lock(&sbi->s_lock);
1268                         goto restore_opts;
1269                 }
1270
1271                 ext2_sync_super(sb, es, 1);
1272         } else {
1273                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1274                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1275                 if (ret) {
1276                         ext2_msg(sb, KERN_WARNING,
1277                                 "warning: couldn't remount RDWR because of "
1278                                 "unsupported optional features (%x).",
1279                                 le32_to_cpu(ret));
1280                         err = -EROFS;
1281                         goto restore_opts;
1282                 }
1283                 /*
1284                  * Mounting a RDONLY partition read-write, so reread and
1285                  * store the current valid flag.  (It may have been changed
1286                  * by e2fsck since we originally mounted the partition.)
1287                  */
1288                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1289                 if (!ext2_setup_super (sb, es, 0))
1290                         sb->s_flags &= ~MS_RDONLY;
1291                 spin_unlock(&sbi->s_lock);
1292
1293                 ext2_write_super(sb);
1294
1295                 dquot_resume(sb, -1);
1296         }
1297
1298         return 0;
1299 restore_opts:
1300         sbi->s_mount_opt = old_opts.s_mount_opt;
1301         sbi->s_resuid = old_opts.s_resuid;
1302         sbi->s_resgid = old_opts.s_resgid;
1303         sb->s_flags = old_sb_flags;
1304         spin_unlock(&sbi->s_lock);
1305         return err;
1306 }
1307
1308 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1309 {
1310         struct super_block *sb = dentry->d_sb;
1311         struct ext2_sb_info *sbi = EXT2_SB(sb);
1312         struct ext2_super_block *es = sbi->s_es;
1313         u64 fsid;
1314
1315         spin_lock(&sbi->s_lock);
1316
1317         if (test_opt (sb, MINIX_DF))
1318                 sbi->s_overhead_last = 0;
1319         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1320                 unsigned long i, overhead = 0;
1321                 smp_rmb();
1322
1323                 /*
1324                  * Compute the overhead (FS structures). This is constant
1325                  * for a given filesystem unless the number of block groups
1326                  * changes so we cache the previous value until it does.
1327                  */
1328
1329                 /*
1330                  * All of the blocks before first_data_block are
1331                  * overhead
1332                  */
1333                 overhead = le32_to_cpu(es->s_first_data_block);
1334
1335                 /*
1336                  * Add the overhead attributed to the superblock and
1337                  * block group descriptors.  If the sparse superblocks
1338                  * feature is turned on, then not all groups have this.
1339                  */
1340                 for (i = 0; i < sbi->s_groups_count; i++)
1341                         overhead += ext2_bg_has_super(sb, i) +
1342                                 ext2_bg_num_gdb(sb, i);
1343
1344                 /*
1345                  * Every block group has an inode bitmap, a block
1346                  * bitmap, and an inode table.
1347                  */
1348                 overhead += (sbi->s_groups_count *
1349                              (2 + sbi->s_itb_per_group));
1350                 sbi->s_overhead_last = overhead;
1351                 smp_wmb();
1352                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1353         }
1354
1355         buf->f_type = EXT2_SUPER_MAGIC;
1356         buf->f_bsize = sb->s_blocksize;
1357         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1358         buf->f_bfree = ext2_count_free_blocks(sb);
1359         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1360         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1361         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1362                 buf->f_bavail = 0;
1363         buf->f_files = le32_to_cpu(es->s_inodes_count);
1364         buf->f_ffree = ext2_count_free_inodes(sb);
1365         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1366         buf->f_namelen = EXT2_NAME_LEN;
1367         fsid = le64_to_cpup((void *)es->s_uuid) ^
1368                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1369         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1370         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1371         spin_unlock(&sbi->s_lock);
1372         return 0;
1373 }
1374
1375 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1376         int flags, const char *dev_name, void *data)
1377 {
1378         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1379 }
1380
1381 #ifdef CONFIG_QUOTA
1382
1383 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1384  * acquiring the locks... As quota files are never truncated and quota code
1385  * itself serializes the operations (and no one else should touch the files)
1386  * we don't have to be afraid of races */
1387 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1388                                size_t len, loff_t off)
1389 {
1390         struct inode *inode = sb_dqopt(sb)->files[type];
1391         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1392         int err = 0;
1393         int offset = off & (sb->s_blocksize - 1);
1394         int tocopy;
1395         size_t toread;
1396         struct buffer_head tmp_bh;
1397         struct buffer_head *bh;
1398         loff_t i_size = i_size_read(inode);
1399
1400         if (off > i_size)
1401                 return 0;
1402         if (off+len > i_size)
1403                 len = i_size-off;
1404         toread = len;
1405         while (toread > 0) {
1406                 tocopy = sb->s_blocksize - offset < toread ?
1407                                 sb->s_blocksize - offset : toread;
1408
1409                 tmp_bh.b_state = 0;
1410                 tmp_bh.b_size = sb->s_blocksize;
1411                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1412                 if (err < 0)
1413                         return err;
1414                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1415                         memset(data, 0, tocopy);
1416                 else {
1417                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1418                         if (!bh)
1419                                 return -EIO;
1420                         memcpy(data, bh->b_data+offset, tocopy);
1421                         brelse(bh);
1422                 }
1423                 offset = 0;
1424                 toread -= tocopy;
1425                 data += tocopy;
1426                 blk++;
1427         }
1428         return len;
1429 }
1430
1431 /* Write to quotafile */
1432 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1433                                 const char *data, size_t len, loff_t off)
1434 {
1435         struct inode *inode = sb_dqopt(sb)->files[type];
1436         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1437         int err = 0;
1438         int offset = off & (sb->s_blocksize - 1);
1439         int tocopy;
1440         size_t towrite = len;
1441         struct buffer_head tmp_bh;
1442         struct buffer_head *bh;
1443
1444         mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1445         while (towrite > 0) {
1446                 tocopy = sb->s_blocksize - offset < towrite ?
1447                                 sb->s_blocksize - offset : towrite;
1448
1449                 tmp_bh.b_state = 0;
1450                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1451                 if (err < 0)
1452                         goto out;
1453                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1454                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1455                 else
1456                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1457                 if (!bh) {
1458                         err = -EIO;
1459                         goto out;
1460                 }
1461                 lock_buffer(bh);
1462                 memcpy(bh->b_data+offset, data, tocopy);
1463                 flush_dcache_page(bh->b_page);
1464                 set_buffer_uptodate(bh);
1465                 mark_buffer_dirty(bh);
1466                 unlock_buffer(bh);
1467                 brelse(bh);
1468                 offset = 0;
1469                 towrite -= tocopy;
1470                 data += tocopy;
1471                 blk++;
1472         }
1473 out:
1474         if (len == towrite) {
1475                 mutex_unlock(&inode->i_mutex);
1476                 return err;
1477         }
1478         if (inode->i_size < off+len-towrite)
1479                 i_size_write(inode, off+len-towrite);
1480         inode->i_version++;
1481         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1482         mark_inode_dirty(inode);
1483         mutex_unlock(&inode->i_mutex);
1484         return len - towrite;
1485 }
1486
1487 #endif
1488
1489 static struct file_system_type ext2_fs_type = {
1490         .owner          = THIS_MODULE,
1491         .name           = "ext2",
1492         .mount          = ext2_mount,
1493         .kill_sb        = kill_block_super,
1494         .fs_flags       = FS_REQUIRES_DEV,
1495 };
1496
1497 static int __init init_ext2_fs(void)
1498 {
1499         int err = init_ext2_xattr();
1500         if (err)
1501                 return err;
1502         err = init_inodecache();
1503         if (err)
1504                 goto out1;
1505         err = register_filesystem(&ext2_fs_type);
1506         if (err)
1507                 goto out;
1508         return 0;
1509 out:
1510         destroy_inodecache();
1511 out1:
1512         exit_ext2_xattr();
1513         return err;
1514 }
1515
1516 static void __exit exit_ext2_fs(void)
1517 {
1518         unregister_filesystem(&ext2_fs_type);
1519         destroy_inodecache();
1520         exit_ext2_xattr();
1521 }
1522
1523 MODULE_AUTHOR("Remy Card and others");
1524 MODULE_DESCRIPTION("Second Extended Filesystem");
1525 MODULE_LICENSE("GPL");
1526 module_init(init_ext2_fs)
1527 module_exit(exit_ext2_fs)