BKL: Remove BKL from ext2 filesystem
[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         va_list args;
50         struct ext2_sb_info *sbi = EXT2_SB(sb);
51         struct ext2_super_block *es = sbi->s_es;
52
53         if (!(sb->s_flags & MS_RDONLY)) {
54                 spin_lock(&sbi->s_lock);
55                 sbi->s_mount_state |= EXT2_ERROR_FS;
56                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
57                 spin_unlock(&sbi->s_lock);
58                 ext2_sync_super(sb, es, 1);
59         }
60
61         va_start(args, fmt);
62         printk(KERN_CRIT "EXT2-fs (%s): error: %s: ", sb->s_id, function);
63         vprintk(fmt, args);
64         printk("\n");
65         va_end(args);
66
67         if (test_opt(sb, ERRORS_PANIC))
68                 panic("EXT2-fs: panic from previous error\n");
69         if (test_opt(sb, ERRORS_RO)) {
70                 ext2_msg(sb, KERN_CRIT,
71                              "error: remounting filesystem read-only");
72                 sb->s_flags |= MS_RDONLY;
73         }
74 }
75
76 void ext2_msg(struct super_block *sb, const char *prefix,
77                 const char *fmt, ...)
78 {
79         va_list args;
80
81         va_start(args, fmt);
82         printk("%sEXT2-fs (%s): ", prefix, sb->s_id);
83         vprintk(fmt, args);
84         printk("\n");
85         va_end(args);
86 }
87
88 /*
89  * This must be called with sbi->s_lock held.
90  */
91 void ext2_update_dynamic_rev(struct super_block *sb)
92 {
93         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
94
95         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
96                 return;
97
98         ext2_msg(sb, KERN_WARNING,
99                      "warning: updating to rev %d because of "
100                      "new feature flag, running e2fsck is recommended",
101                      EXT2_DYNAMIC_REV);
102
103         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
104         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
105         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
106         /* leave es->s_feature_*compat flags alone */
107         /* es->s_uuid will be set by e2fsck if empty */
108
109         /*
110          * The rest of the superblock fields should be zero, and if not it
111          * means they are likely already in use, so leave them alone.  We
112          * can leave it up to e2fsck to clean up any inconsistencies there.
113          */
114 }
115
116 static void ext2_put_super (struct super_block * sb)
117 {
118         int db_count;
119         int i;
120         struct ext2_sb_info *sbi = EXT2_SB(sb);
121
122         dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
123
124         if (sb->s_dirt)
125                 ext2_write_super(sb);
126
127         ext2_xattr_put_super(sb);
128         if (!(sb->s_flags & MS_RDONLY)) {
129                 struct ext2_super_block *es = sbi->s_es;
130
131                 spin_lock(&sbi->s_lock);
132                 es->s_state = cpu_to_le16(sbi->s_mount_state);
133                 spin_unlock(&sbi->s_lock);
134                 ext2_sync_super(sb, es, 1);
135         }
136         db_count = sbi->s_gdb_count;
137         for (i = 0; i < db_count; i++)
138                 if (sbi->s_group_desc[i])
139                         brelse (sbi->s_group_desc[i]);
140         kfree(sbi->s_group_desc);
141         kfree(sbi->s_debts);
142         percpu_counter_destroy(&sbi->s_freeblocks_counter);
143         percpu_counter_destroy(&sbi->s_freeinodes_counter);
144         percpu_counter_destroy(&sbi->s_dirs_counter);
145         brelse (sbi->s_sbh);
146         sb->s_fs_info = NULL;
147         kfree(sbi->s_blockgroup_lock);
148         kfree(sbi);
149 }
150
151 static struct kmem_cache * ext2_inode_cachep;
152
153 static struct inode *ext2_alloc_inode(struct super_block *sb)
154 {
155         struct ext2_inode_info *ei;
156         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
157         if (!ei)
158                 return NULL;
159         ei->i_block_alloc_info = NULL;
160         ei->vfs_inode.i_version = 1;
161         return &ei->vfs_inode;
162 }
163
164 static void ext2_destroy_inode(struct inode *inode)
165 {
166         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
167 }
168
169 static void init_once(void *foo)
170 {
171         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
172
173         rwlock_init(&ei->i_meta_lock);
174 #ifdef CONFIG_EXT2_FS_XATTR
175         init_rwsem(&ei->xattr_sem);
176 #endif
177         mutex_init(&ei->truncate_mutex);
178         inode_init_once(&ei->vfs_inode);
179 }
180
181 static int init_inodecache(void)
182 {
183         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
184                                              sizeof(struct ext2_inode_info),
185                                              0, (SLAB_RECLAIM_ACCOUNT|
186                                                 SLAB_MEM_SPREAD),
187                                              init_once);
188         if (ext2_inode_cachep == NULL)
189                 return -ENOMEM;
190         return 0;
191 }
192
193 static void destroy_inodecache(void)
194 {
195         kmem_cache_destroy(ext2_inode_cachep);
196 }
197
198 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
199 {
200         struct super_block *sb = vfs->mnt_sb;
201         struct ext2_sb_info *sbi = EXT2_SB(sb);
202         struct ext2_super_block *es = sbi->s_es;
203         unsigned long def_mount_opts;
204
205         spin_lock(&sbi->s_lock);
206         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
207
208         if (sbi->s_sb_block != 1)
209                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
210         if (test_opt(sb, MINIX_DF))
211                 seq_puts(seq, ",minixdf");
212         if (test_opt(sb, GRPID))
213                 seq_puts(seq, ",grpid");
214         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
215                 seq_puts(seq, ",nogrpid");
216         if (sbi->s_resuid != EXT2_DEF_RESUID ||
217             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
218                 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
219         }
220         if (sbi->s_resgid != EXT2_DEF_RESGID ||
221             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
222                 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
223         }
224         if (test_opt(sb, ERRORS_RO)) {
225                 int def_errors = le16_to_cpu(es->s_errors);
226
227                 if (def_errors == EXT2_ERRORS_PANIC ||
228                     def_errors == EXT2_ERRORS_CONTINUE) {
229                         seq_puts(seq, ",errors=remount-ro");
230                 }
231         }
232         if (test_opt(sb, ERRORS_CONT))
233                 seq_puts(seq, ",errors=continue");
234         if (test_opt(sb, ERRORS_PANIC))
235                 seq_puts(seq, ",errors=panic");
236         if (test_opt(sb, NO_UID32))
237                 seq_puts(seq, ",nouid32");
238         if (test_opt(sb, DEBUG))
239                 seq_puts(seq, ",debug");
240         if (test_opt(sb, OLDALLOC))
241                 seq_puts(seq, ",oldalloc");
242
243 #ifdef CONFIG_EXT2_FS_XATTR
244         if (test_opt(sb, XATTR_USER))
245                 seq_puts(seq, ",user_xattr");
246         if (!test_opt(sb, XATTR_USER) &&
247             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
248                 seq_puts(seq, ",nouser_xattr");
249         }
250 #endif
251
252 #ifdef CONFIG_EXT2_FS_POSIX_ACL
253         if (test_opt(sb, POSIX_ACL))
254                 seq_puts(seq, ",acl");
255         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
256                 seq_puts(seq, ",noacl");
257 #endif
258
259         if (test_opt(sb, NOBH))
260                 seq_puts(seq, ",nobh");
261
262 #if defined(CONFIG_QUOTA)
263         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
264                 seq_puts(seq, ",usrquota");
265
266         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
267                 seq_puts(seq, ",grpquota");
268 #endif
269
270 #if defined(CONFIG_EXT2_FS_XIP)
271         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
272                 seq_puts(seq, ",xip");
273 #endif
274
275         if (!test_opt(sb, RESERVATION))
276                 seq_puts(seq, ",noreservation");
277
278         spin_unlock(&sbi->s_lock);
279         return 0;
280 }
281
282 #ifdef CONFIG_QUOTA
283 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
284 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
285 #endif
286
287 static const struct super_operations ext2_sops = {
288         .alloc_inode    = ext2_alloc_inode,
289         .destroy_inode  = ext2_destroy_inode,
290         .write_inode    = ext2_write_inode,
291         .evict_inode    = ext2_evict_inode,
292         .put_super      = ext2_put_super,
293         .write_super    = ext2_write_super,
294         .sync_fs        = ext2_sync_fs,
295         .statfs         = ext2_statfs,
296         .remount_fs     = ext2_remount,
297         .show_options   = ext2_show_options,
298 #ifdef CONFIG_QUOTA
299         .quota_read     = ext2_quota_read,
300         .quota_write    = ext2_quota_write,
301 #endif
302 };
303
304 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
305                 u64 ino, u32 generation)
306 {
307         struct inode *inode;
308
309         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
310                 return ERR_PTR(-ESTALE);
311         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
312                 return ERR_PTR(-ESTALE);
313
314         /* iget isn't really right if the inode is currently unallocated!!
315          * ext2_read_inode currently does appropriate checks, but
316          * it might be "neater" to call ext2_get_inode first and check
317          * if the inode is valid.....
318          */
319         inode = ext2_iget(sb, ino);
320         if (IS_ERR(inode))
321                 return ERR_CAST(inode);
322         if (generation && inode->i_generation != generation) {
323                 /* we didn't find the right inode.. */
324                 iput(inode);
325                 return ERR_PTR(-ESTALE);
326         }
327         return inode;
328 }
329
330 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
331                 int fh_len, int fh_type)
332 {
333         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
334                                     ext2_nfs_get_inode);
335 }
336
337 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
338                 int fh_len, int fh_type)
339 {
340         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
341                                     ext2_nfs_get_inode);
342 }
343
344 /* Yes, most of these are left as NULL!!
345  * A NULL value implies the default, which works with ext2-like file
346  * systems, but can be improved upon.
347  * Currently only get_parent is required.
348  */
349 static const struct export_operations ext2_export_ops = {
350         .fh_to_dentry = ext2_fh_to_dentry,
351         .fh_to_parent = ext2_fh_to_parent,
352         .get_parent = ext2_get_parent,
353 };
354
355 static unsigned long get_sb_block(void **data)
356 {
357         unsigned long   sb_block;
358         char            *options = (char *) *data;
359
360         if (!options || strncmp(options, "sb=", 3) != 0)
361                 return 1;       /* Default location */
362         options += 3;
363         sb_block = simple_strtoul(options, &options, 0);
364         if (*options && *options != ',') {
365                 printk("EXT2-fs: Invalid sb specification: %s\n",
366                        (char *) *data);
367                 return 1;
368         }
369         if (*options == ',')
370                 options++;
371         *data = (void *) options;
372         return sb_block;
373 }
374
375 enum {
376         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
377         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
378         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
379         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
380         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
381         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
382 };
383
384 static const match_table_t tokens = {
385         {Opt_bsd_df, "bsddf"},
386         {Opt_minix_df, "minixdf"},
387         {Opt_grpid, "grpid"},
388         {Opt_grpid, "bsdgroups"},
389         {Opt_nogrpid, "nogrpid"},
390         {Opt_nogrpid, "sysvgroups"},
391         {Opt_resgid, "resgid=%u"},
392         {Opt_resuid, "resuid=%u"},
393         {Opt_sb, "sb=%u"},
394         {Opt_err_cont, "errors=continue"},
395         {Opt_err_panic, "errors=panic"},
396         {Opt_err_ro, "errors=remount-ro"},
397         {Opt_nouid32, "nouid32"},
398         {Opt_nocheck, "check=none"},
399         {Opt_nocheck, "nocheck"},
400         {Opt_debug, "debug"},
401         {Opt_oldalloc, "oldalloc"},
402         {Opt_orlov, "orlov"},
403         {Opt_nobh, "nobh"},
404         {Opt_user_xattr, "user_xattr"},
405         {Opt_nouser_xattr, "nouser_xattr"},
406         {Opt_acl, "acl"},
407         {Opt_noacl, "noacl"},
408         {Opt_xip, "xip"},
409         {Opt_grpquota, "grpquota"},
410         {Opt_ignore, "noquota"},
411         {Opt_quota, "quota"},
412         {Opt_usrquota, "usrquota"},
413         {Opt_reservation, "reservation"},
414         {Opt_noreservation, "noreservation"},
415         {Opt_err, NULL}
416 };
417
418 static int parse_options(char *options, struct super_block *sb)
419 {
420         char *p;
421         struct ext2_sb_info *sbi = EXT2_SB(sb);
422         substring_t args[MAX_OPT_ARGS];
423         int option;
424
425         if (!options)
426                 return 1;
427
428         while ((p = strsep (&options, ",")) != NULL) {
429                 int token;
430                 if (!*p)
431                         continue;
432
433                 token = match_token(p, tokens, args);
434                 switch (token) {
435                 case Opt_bsd_df:
436                         clear_opt (sbi->s_mount_opt, MINIX_DF);
437                         break;
438                 case Opt_minix_df:
439                         set_opt (sbi->s_mount_opt, MINIX_DF);
440                         break;
441                 case Opt_grpid:
442                         set_opt (sbi->s_mount_opt, GRPID);
443                         break;
444                 case Opt_nogrpid:
445                         clear_opt (sbi->s_mount_opt, GRPID);
446                         break;
447                 case Opt_resuid:
448                         if (match_int(&args[0], &option))
449                                 return 0;
450                         sbi->s_resuid = option;
451                         break;
452                 case Opt_resgid:
453                         if (match_int(&args[0], &option))
454                                 return 0;
455                         sbi->s_resgid = option;
456                         break;
457                 case Opt_sb:
458                         /* handled by get_sb_block() instead of here */
459                         /* *sb_block = match_int(&args[0]); */
460                         break;
461                 case Opt_err_panic:
462                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
463                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
464                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
465                         break;
466                 case Opt_err_ro:
467                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
468                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
469                         set_opt (sbi->s_mount_opt, ERRORS_RO);
470                         break;
471                 case Opt_err_cont:
472                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
473                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
474                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
475                         break;
476                 case Opt_nouid32:
477                         set_opt (sbi->s_mount_opt, NO_UID32);
478                         break;
479                 case Opt_nocheck:
480                         clear_opt (sbi->s_mount_opt, CHECK);
481                         break;
482                 case Opt_debug:
483                         set_opt (sbi->s_mount_opt, DEBUG);
484                         break;
485                 case Opt_oldalloc:
486                         set_opt (sbi->s_mount_opt, OLDALLOC);
487                         break;
488                 case Opt_orlov:
489                         clear_opt (sbi->s_mount_opt, OLDALLOC);
490                         break;
491                 case Opt_nobh:
492                         set_opt (sbi->s_mount_opt, NOBH);
493                         break;
494 #ifdef CONFIG_EXT2_FS_XATTR
495                 case Opt_user_xattr:
496                         set_opt (sbi->s_mount_opt, XATTR_USER);
497                         break;
498                 case Opt_nouser_xattr:
499                         clear_opt (sbi->s_mount_opt, XATTR_USER);
500                         break;
501 #else
502                 case Opt_user_xattr:
503                 case Opt_nouser_xattr:
504                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
505                                 "not supported");
506                         break;
507 #endif
508 #ifdef CONFIG_EXT2_FS_POSIX_ACL
509                 case Opt_acl:
510                         set_opt(sbi->s_mount_opt, POSIX_ACL);
511                         break;
512                 case Opt_noacl:
513                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
514                         break;
515 #else
516                 case Opt_acl:
517                 case Opt_noacl:
518                         ext2_msg(sb, KERN_INFO,
519                                 "(no)acl options not supported");
520                         break;
521 #endif
522                 case Opt_xip:
523 #ifdef CONFIG_EXT2_FS_XIP
524                         set_opt (sbi->s_mount_opt, XIP);
525 #else
526                         ext2_msg(sb, KERN_INFO, "xip option not supported");
527 #endif
528                         break;
529
530 #if defined(CONFIG_QUOTA)
531                 case Opt_quota:
532                 case Opt_usrquota:
533                         set_opt(sbi->s_mount_opt, USRQUOTA);
534                         break;
535
536                 case Opt_grpquota:
537                         set_opt(sbi->s_mount_opt, GRPQUOTA);
538                         break;
539 #else
540                 case Opt_quota:
541                 case Opt_usrquota:
542                 case Opt_grpquota:
543                         ext2_msg(sb, KERN_INFO,
544                                 "quota operations not supported");
545                         break;
546 #endif
547
548                 case Opt_reservation:
549                         set_opt(sbi->s_mount_opt, RESERVATION);
550                         ext2_msg(sb, KERN_INFO, "reservations ON");
551                         break;
552                 case Opt_noreservation:
553                         clear_opt(sbi->s_mount_opt, RESERVATION);
554                         ext2_msg(sb, KERN_INFO, "reservations OFF");
555                         break;
556                 case Opt_ignore:
557                         break;
558                 default:
559                         return 0;
560                 }
561         }
562         return 1;
563 }
564
565 static int ext2_setup_super (struct super_block * sb,
566                               struct ext2_super_block * es,
567                               int read_only)
568 {
569         int res = 0;
570         struct ext2_sb_info *sbi = EXT2_SB(sb);
571
572         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
573                 ext2_msg(sb, KERN_ERR,
574                         "error: revision level too high, "
575                         "forcing read-only mode");
576                 res = MS_RDONLY;
577         }
578         if (read_only)
579                 return res;
580         if (!(sbi->s_mount_state & EXT2_VALID_FS))
581                 ext2_msg(sb, KERN_WARNING,
582                         "warning: mounting unchecked fs, "
583                         "running e2fsck is recommended");
584         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
585                 ext2_msg(sb, KERN_WARNING,
586                         "warning: mounting fs with errors, "
587                         "running e2fsck is recommended");
588         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
589                  le16_to_cpu(es->s_mnt_count) >=
590                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
591                 ext2_msg(sb, KERN_WARNING,
592                         "warning: maximal mount count reached, "
593                         "running e2fsck is recommended");
594         else if (le32_to_cpu(es->s_checkinterval) &&
595                 (le32_to_cpu(es->s_lastcheck) +
596                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
597                 ext2_msg(sb, KERN_WARNING,
598                         "warning: checktime reached, "
599                         "running e2fsck is recommended");
600         if (!le16_to_cpu(es->s_max_mnt_count))
601                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
602         le16_add_cpu(&es->s_mnt_count, 1);
603         if (test_opt (sb, DEBUG))
604                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
605                         "bpg=%lu, ipg=%lu, mo=%04lx]",
606                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
607                         sbi->s_frag_size,
608                         sbi->s_groups_count,
609                         EXT2_BLOCKS_PER_GROUP(sb),
610                         EXT2_INODES_PER_GROUP(sb),
611                         sbi->s_mount_opt);
612         return res;
613 }
614
615 static int ext2_check_descriptors(struct super_block *sb)
616 {
617         int i;
618         struct ext2_sb_info *sbi = EXT2_SB(sb);
619
620         ext2_debug ("Checking group descriptors");
621
622         for (i = 0; i < sbi->s_groups_count; i++) {
623                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
624                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
625                 ext2_fsblk_t last_block;
626
627                 if (i == sbi->s_groups_count - 1)
628                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
629                 else
630                         last_block = first_block +
631                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
632
633                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
634                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
635                 {
636                         ext2_error (sb, "ext2_check_descriptors",
637                                     "Block bitmap for group %d"
638                                     " not in group (block %lu)!",
639                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
640                         return 0;
641                 }
642                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
643                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
644                 {
645                         ext2_error (sb, "ext2_check_descriptors",
646                                     "Inode bitmap for group %d"
647                                     " not in group (block %lu)!",
648                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
649                         return 0;
650                 }
651                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
652                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
653                     last_block)
654                 {
655                         ext2_error (sb, "ext2_check_descriptors",
656                                     "Inode table for group %d"
657                                     " not in group (block %lu)!",
658                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
659                         return 0;
660                 }
661         }
662         return 1;
663 }
664
665 /*
666  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
667  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
668  * We need to be 1 filesystem block less than the 2^32 sector limit.
669  */
670 static loff_t ext2_max_size(int bits)
671 {
672         loff_t res = EXT2_NDIR_BLOCKS;
673         int meta_blocks;
674         loff_t upper_limit;
675
676         /* This is calculated to be the largest file size for a
677          * dense, file such that the total number of
678          * sectors in the file, including data and all indirect blocks,
679          * does not exceed 2^32 -1
680          * __u32 i_blocks representing the total number of
681          * 512 bytes blocks of the file
682          */
683         upper_limit = (1LL << 32) - 1;
684
685         /* total blocks in file system block size */
686         upper_limit >>= (bits - 9);
687
688
689         /* indirect blocks */
690         meta_blocks = 1;
691         /* double indirect blocks */
692         meta_blocks += 1 + (1LL << (bits-2));
693         /* tripple indirect blocks */
694         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
695
696         upper_limit -= meta_blocks;
697         upper_limit <<= bits;
698
699         res += 1LL << (bits-2);
700         res += 1LL << (2*(bits-2));
701         res += 1LL << (3*(bits-2));
702         res <<= bits;
703         if (res > upper_limit)
704                 res = upper_limit;
705
706         if (res > MAX_LFS_FILESIZE)
707                 res = MAX_LFS_FILESIZE;
708
709         return res;
710 }
711
712 static unsigned long descriptor_loc(struct super_block *sb,
713                                     unsigned long logic_sb_block,
714                                     int nr)
715 {
716         struct ext2_sb_info *sbi = EXT2_SB(sb);
717         unsigned long bg, first_meta_bg;
718         int has_super = 0;
719         
720         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
721
722         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
723             nr < first_meta_bg)
724                 return (logic_sb_block + nr + 1);
725         bg = sbi->s_desc_per_block * nr;
726         if (ext2_bg_has_super(sb, bg))
727                 has_super = 1;
728
729         return ext2_group_first_block_no(sb, bg) + has_super;
730 }
731
732 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
733 {
734         struct buffer_head * bh;
735         struct ext2_sb_info * sbi;
736         struct ext2_super_block * es;
737         struct inode *root;
738         unsigned long block;
739         unsigned long sb_block = get_sb_block(&data);
740         unsigned long logic_sb_block;
741         unsigned long offset = 0;
742         unsigned long def_mount_opts;
743         long ret = -EINVAL;
744         int blocksize = BLOCK_SIZE;
745         int db_count;
746         int i, j;
747         __le32 features;
748         int err;
749
750         err = -ENOMEM;
751         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
752         if (!sbi)
753                 goto failed_unlock;
754
755         sbi->s_blockgroup_lock =
756                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
757         if (!sbi->s_blockgroup_lock) {
758                 kfree(sbi);
759                 goto failed_unlock;
760         }
761         sb->s_fs_info = sbi;
762         sbi->s_sb_block = sb_block;
763
764         spin_lock_init(&sbi->s_lock);
765
766         /*
767          * See what the current blocksize for the device is, and
768          * use that as the blocksize.  Otherwise (or if the blocksize
769          * is smaller than the default) use the default.
770          * This is important for devices that have a hardware
771          * sectorsize that is larger than the default.
772          */
773         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
774         if (!blocksize) {
775                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
776                 goto failed_sbi;
777         }
778
779         /*
780          * If the superblock doesn't start on a hardware sector boundary,
781          * calculate the offset.  
782          */
783         if (blocksize != BLOCK_SIZE) {
784                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
785                 offset = (sb_block*BLOCK_SIZE) % blocksize;
786         } else {
787                 logic_sb_block = sb_block;
788         }
789
790         if (!(bh = sb_bread(sb, logic_sb_block))) {
791                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
792                 goto failed_sbi;
793         }
794         /*
795          * Note: s_es must be initialized as soon as possible because
796          *       some ext2 macro-instructions depend on its value
797          */
798         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
799         sbi->s_es = es;
800         sb->s_magic = le16_to_cpu(es->s_magic);
801
802         if (sb->s_magic != EXT2_SUPER_MAGIC)
803                 goto cantfind_ext2;
804
805         /* Set defaults before we parse the mount options */
806         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
807         if (def_mount_opts & EXT2_DEFM_DEBUG)
808                 set_opt(sbi->s_mount_opt, DEBUG);
809         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
810                 set_opt(sbi->s_mount_opt, GRPID);
811         if (def_mount_opts & EXT2_DEFM_UID16)
812                 set_opt(sbi->s_mount_opt, NO_UID32);
813 #ifdef CONFIG_EXT2_FS_XATTR
814         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
815                 set_opt(sbi->s_mount_opt, XATTR_USER);
816 #endif
817 #ifdef CONFIG_EXT2_FS_POSIX_ACL
818         if (def_mount_opts & EXT2_DEFM_ACL)
819                 set_opt(sbi->s_mount_opt, POSIX_ACL);
820 #endif
821         
822         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
823                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
824         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
825                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
826         else
827                 set_opt(sbi->s_mount_opt, ERRORS_RO);
828
829         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
830         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
831         
832         set_opt(sbi->s_mount_opt, RESERVATION);
833
834         if (!parse_options((char *) data, sb))
835                 goto failed_mount;
836
837         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
838                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
839                  MS_POSIXACL : 0);
840
841         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
842                                     EXT2_MOUNT_XIP if not */
843
844         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
845             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
846              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
847              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
848                 ext2_msg(sb, KERN_WARNING,
849                         "warning: feature flags set on rev 0 fs, "
850                         "running e2fsck is recommended");
851         /*
852          * Check feature flags regardless of the revision level, since we
853          * previously didn't change the revision level when setting the flags,
854          * so there is a chance incompat flags are set on a rev 0 filesystem.
855          */
856         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
857         if (features) {
858                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
859                        "unsupported optional features (%x)",
860                         le32_to_cpu(features));
861                 goto failed_mount;
862         }
863         if (!(sb->s_flags & MS_RDONLY) &&
864             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
865                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
866                        "unsupported optional features (%x)",
867                        le32_to_cpu(features));
868                 goto failed_mount;
869         }
870
871         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
872
873         if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
874                 if (!silent)
875                         ext2_msg(sb, KERN_ERR,
876                                 "error: unsupported blocksize for xip");
877                 goto failed_mount;
878         }
879
880         /* If the blocksize doesn't match, re-read the thing.. */
881         if (sb->s_blocksize != blocksize) {
882                 brelse(bh);
883
884                 if (!sb_set_blocksize(sb, blocksize)) {
885                         ext2_msg(sb, KERN_ERR, "error: blocksize is too small");
886                         goto failed_sbi;
887                 }
888
889                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
890                 offset = (sb_block*BLOCK_SIZE) % blocksize;
891                 bh = sb_bread(sb, logic_sb_block);
892                 if(!bh) {
893                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
894                                 "superblock on 2nd try");
895                         goto failed_sbi;
896                 }
897                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
898                 sbi->s_es = es;
899                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
900                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
901                         goto failed_mount;
902                 }
903         }
904
905         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
906
907         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
908                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
909                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
910         } else {
911                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
912                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
913                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
914                     !is_power_of_2(sbi->s_inode_size) ||
915                     (sbi->s_inode_size > blocksize)) {
916                         ext2_msg(sb, KERN_ERR,
917                                 "error: unsupported inode size: %d",
918                                 sbi->s_inode_size);
919                         goto failed_mount;
920                 }
921         }
922
923         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
924                                    le32_to_cpu(es->s_log_frag_size);
925         if (sbi->s_frag_size == 0)
926                 goto cantfind_ext2;
927         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
928
929         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
930         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
931         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
932
933         if (EXT2_INODE_SIZE(sb) == 0)
934                 goto cantfind_ext2;
935         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
936         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
937                 goto cantfind_ext2;
938         sbi->s_itb_per_group = sbi->s_inodes_per_group /
939                                         sbi->s_inodes_per_block;
940         sbi->s_desc_per_block = sb->s_blocksize /
941                                         sizeof (struct ext2_group_desc);
942         sbi->s_sbh = bh;
943         sbi->s_mount_state = le16_to_cpu(es->s_state);
944         sbi->s_addr_per_block_bits =
945                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
946         sbi->s_desc_per_block_bits =
947                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
948
949         if (sb->s_magic != EXT2_SUPER_MAGIC)
950                 goto cantfind_ext2;
951
952         if (sb->s_blocksize != bh->b_size) {
953                 if (!silent)
954                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
955                 goto failed_mount;
956         }
957
958         if (sb->s_blocksize != sbi->s_frag_size) {
959                 ext2_msg(sb, KERN_ERR,
960                         "error: fragsize %lu != blocksize %lu"
961                         "(not supported yet)",
962                         sbi->s_frag_size, sb->s_blocksize);
963                 goto failed_mount;
964         }
965
966         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
967                 ext2_msg(sb, KERN_ERR,
968                         "error: #blocks per group too big: %lu",
969                         sbi->s_blocks_per_group);
970                 goto failed_mount;
971         }
972         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
973                 ext2_msg(sb, KERN_ERR,
974                         "error: #fragments per group too big: %lu",
975                         sbi->s_frags_per_group);
976                 goto failed_mount;
977         }
978         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
979                 ext2_msg(sb, KERN_ERR,
980                         "error: #inodes per group too big: %lu",
981                         sbi->s_inodes_per_group);
982                 goto failed_mount;
983         }
984
985         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
986                 goto cantfind_ext2;
987         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
988                                 le32_to_cpu(es->s_first_data_block) - 1)
989                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
990         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
991                    EXT2_DESC_PER_BLOCK(sb);
992         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
993         if (sbi->s_group_desc == NULL) {
994                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
995                 goto failed_mount;
996         }
997         bgl_lock_init(sbi->s_blockgroup_lock);
998         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
999         if (!sbi->s_debts) {
1000                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1001                 goto failed_mount_group_desc;
1002         }
1003         for (i = 0; i < db_count; i++) {
1004                 block = descriptor_loc(sb, logic_sb_block, i);
1005                 sbi->s_group_desc[i] = sb_bread(sb, block);
1006                 if (!sbi->s_group_desc[i]) {
1007                         for (j = 0; j < i; j++)
1008                                 brelse (sbi->s_group_desc[j]);
1009                         ext2_msg(sb, KERN_ERR,
1010                                 "error: unable to read group descriptors");
1011                         goto failed_mount_group_desc;
1012                 }
1013         }
1014         if (!ext2_check_descriptors (sb)) {
1015                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1016                 goto failed_mount2;
1017         }
1018         sbi->s_gdb_count = db_count;
1019         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1020         spin_lock_init(&sbi->s_next_gen_lock);
1021
1022         /* per fileystem reservation list head & lock */
1023         spin_lock_init(&sbi->s_rsv_window_lock);
1024         sbi->s_rsv_window_root = RB_ROOT;
1025         /*
1026          * Add a single, static dummy reservation to the start of the
1027          * reservation window list --- it gives us a placeholder for
1028          * append-at-start-of-list which makes the allocation logic
1029          * _much_ simpler.
1030          */
1031         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1032         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1033         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1034         sbi->s_rsv_window_head.rsv_goal_size = 0;
1035         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1036
1037         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1038                                 ext2_count_free_blocks(sb));
1039         if (!err) {
1040                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1041                                 ext2_count_free_inodes(sb));
1042         }
1043         if (!err) {
1044                 err = percpu_counter_init(&sbi->s_dirs_counter,
1045                                 ext2_count_dirs(sb));
1046         }
1047         if (err) {
1048                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1049                 goto failed_mount3;
1050         }
1051         /*
1052          * set up enough so that it can read an inode
1053          */
1054         sb->s_op = &ext2_sops;
1055         sb->s_export_op = &ext2_export_ops;
1056         sb->s_xattr = ext2_xattr_handlers;
1057
1058 #ifdef CONFIG_QUOTA
1059         sb->dq_op = &dquot_operations;
1060         sb->s_qcop = &dquot_quotactl_ops;
1061 #endif
1062
1063         root = ext2_iget(sb, EXT2_ROOT_INO);
1064         if (IS_ERR(root)) {
1065                 ret = PTR_ERR(root);
1066                 goto failed_mount3;
1067         }
1068         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1069                 iput(root);
1070                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1071                 goto failed_mount3;
1072         }
1073
1074         sb->s_root = d_alloc_root(root);
1075         if (!sb->s_root) {
1076                 iput(root);
1077                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1078                 ret = -ENOMEM;
1079                 goto failed_mount3;
1080         }
1081         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1082                 ext2_msg(sb, KERN_WARNING,
1083                         "warning: mounting ext3 filesystem as ext2");
1084         if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1085                 sb->s_flags |= MS_RDONLY;
1086         ext2_write_super(sb);
1087         return 0;
1088
1089 cantfind_ext2:
1090         if (!silent)
1091                 ext2_msg(sb, KERN_ERR,
1092                         "error: can't find an ext2 filesystem on dev %s.",
1093                         sb->s_id);
1094         goto failed_mount;
1095 failed_mount3:
1096         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1097         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1098         percpu_counter_destroy(&sbi->s_dirs_counter);
1099 failed_mount2:
1100         for (i = 0; i < db_count; i++)
1101                 brelse(sbi->s_group_desc[i]);
1102 failed_mount_group_desc:
1103         kfree(sbi->s_group_desc);
1104         kfree(sbi->s_debts);
1105 failed_mount:
1106         brelse(bh);
1107 failed_sbi:
1108         sb->s_fs_info = NULL;
1109         kfree(sbi->s_blockgroup_lock);
1110         kfree(sbi);
1111 failed_unlock:
1112         return ret;
1113 }
1114
1115 static void ext2_clear_super_error(struct super_block *sb)
1116 {
1117         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1118
1119         if (buffer_write_io_error(sbh)) {
1120                 /*
1121                  * Oh, dear.  A previous attempt to write the
1122                  * superblock failed.  This could happen because the
1123                  * USB device was yanked out.  Or it could happen to
1124                  * be a transient write error and maybe the block will
1125                  * be remapped.  Nothing we can do but to retry the
1126                  * write and hope for the best.
1127                  */
1128                 ext2_msg(sb, KERN_ERR,
1129                        "previous I/O error to superblock detected\n");
1130                 clear_buffer_write_io_error(sbh);
1131                 set_buffer_uptodate(sbh);
1132         }
1133 }
1134
1135 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1136                             int wait)
1137 {
1138         ext2_clear_super_error(sb);
1139         spin_lock(&EXT2_SB(sb)->s_lock);
1140         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1141         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1142         es->s_wtime = cpu_to_le32(get_seconds());
1143         /* unlock before we do IO */
1144         spin_unlock(&EXT2_SB(sb)->s_lock);
1145         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1146         if (wait)
1147                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1148         sb->s_dirt = 0;
1149 }
1150
1151 /*
1152  * In the second extended file system, it is not necessary to
1153  * write the super block since we use a mapping of the
1154  * disk super block in a buffer.
1155  *
1156  * However, this function is still used to set the fs valid
1157  * flags to 0.  We need to set this flag to 0 since the fs
1158  * may have been checked while mounted and e2fsck may have
1159  * set s_state to EXT2_VALID_FS after some corrections.
1160  */
1161 static int ext2_sync_fs(struct super_block *sb, int wait)
1162 {
1163         struct ext2_sb_info *sbi = EXT2_SB(sb);
1164         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1165
1166         spin_lock(&sbi->s_lock);
1167         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1168                 ext2_debug("setting valid to 0\n");
1169                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1170         }
1171         spin_unlock(&sbi->s_lock);
1172         ext2_sync_super(sb, es, wait);
1173         return 0;
1174 }
1175
1176
1177 void ext2_write_super(struct super_block *sb)
1178 {
1179         if (!(sb->s_flags & MS_RDONLY))
1180                 ext2_sync_fs(sb, 1);
1181         else
1182                 sb->s_dirt = 0;
1183 }
1184
1185 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1186 {
1187         struct ext2_sb_info * sbi = EXT2_SB(sb);
1188         struct ext2_super_block * es;
1189         unsigned long old_mount_opt = sbi->s_mount_opt;
1190         struct ext2_mount_options old_opts;
1191         unsigned long old_sb_flags;
1192         int err;
1193
1194         spin_lock(&sbi->s_lock);
1195
1196         /* Store the old options */
1197         old_sb_flags = sb->s_flags;
1198         old_opts.s_mount_opt = sbi->s_mount_opt;
1199         old_opts.s_resuid = sbi->s_resuid;
1200         old_opts.s_resgid = sbi->s_resgid;
1201
1202         /*
1203          * Allow the "check" option to be passed as a remount option.
1204          */
1205         if (!parse_options(data, sb)) {
1206                 err = -EINVAL;
1207                 goto restore_opts;
1208         }
1209
1210         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1211                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1212
1213         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1214                                     EXT2_MOUNT_XIP if not */
1215
1216         if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1217                 ext2_msg(sb, KERN_WARNING,
1218                         "warning: unsupported blocksize for xip");
1219                 err = -EINVAL;
1220                 goto restore_opts;
1221         }
1222
1223         es = sbi->s_es;
1224         if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1225             (old_mount_opt & EXT2_MOUNT_XIP)) &&
1226             invalidate_inodes(sb)) {
1227                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1228                          "xip flag with busy inodes while remounting");
1229                 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1230                 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1231         }
1232         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1233                 spin_unlock(&sbi->s_lock);
1234                 return 0;
1235         }
1236         if (*flags & MS_RDONLY) {
1237                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1238                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1239                         spin_unlock(&sbi->s_lock);
1240                         return 0;
1241                 }
1242
1243                 /*
1244                  * OK, we are remounting a valid rw partition rdonly, so set
1245                  * the rdonly flag and then mark the partition as valid again.
1246                  */
1247                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1248                 es->s_mtime = cpu_to_le32(get_seconds());
1249                 spin_unlock(&sbi->s_lock);
1250
1251                 err = dquot_suspend(sb, -1);
1252                 if (err < 0) {
1253                         spin_lock(&sbi->s_lock);
1254                         goto restore_opts;
1255                 }
1256
1257                 ext2_sync_super(sb, es, 1);
1258         } else {
1259                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1260                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1261                 if (ret) {
1262                         ext2_msg(sb, KERN_WARNING,
1263                                 "warning: couldn't remount RDWR because of "
1264                                 "unsupported optional features (%x).",
1265                                 le32_to_cpu(ret));
1266                         err = -EROFS;
1267                         goto restore_opts;
1268                 }
1269                 /*
1270                  * Mounting a RDONLY partition read-write, so reread and
1271                  * store the current valid flag.  (It may have been changed
1272                  * by e2fsck since we originally mounted the partition.)
1273                  */
1274                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1275                 if (!ext2_setup_super (sb, es, 0))
1276                         sb->s_flags &= ~MS_RDONLY;
1277                 spin_unlock(&sbi->s_lock);
1278
1279                 ext2_write_super(sb);
1280
1281                 dquot_resume(sb, -1);
1282         }
1283
1284         return 0;
1285 restore_opts:
1286         sbi->s_mount_opt = old_opts.s_mount_opt;
1287         sbi->s_resuid = old_opts.s_resuid;
1288         sbi->s_resgid = old_opts.s_resgid;
1289         sb->s_flags = old_sb_flags;
1290         spin_unlock(&sbi->s_lock);
1291         return err;
1292 }
1293
1294 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1295 {
1296         struct super_block *sb = dentry->d_sb;
1297         struct ext2_sb_info *sbi = EXT2_SB(sb);
1298         struct ext2_super_block *es = sbi->s_es;
1299         u64 fsid;
1300
1301         spin_lock(&sbi->s_lock);
1302
1303         if (test_opt (sb, MINIX_DF))
1304                 sbi->s_overhead_last = 0;
1305         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1306                 unsigned long i, overhead = 0;
1307                 smp_rmb();
1308
1309                 /*
1310                  * Compute the overhead (FS structures). This is constant
1311                  * for a given filesystem unless the number of block groups
1312                  * changes so we cache the previous value until it does.
1313                  */
1314
1315                 /*
1316                  * All of the blocks before first_data_block are
1317                  * overhead
1318                  */
1319                 overhead = le32_to_cpu(es->s_first_data_block);
1320
1321                 /*
1322                  * Add the overhead attributed to the superblock and
1323                  * block group descriptors.  If the sparse superblocks
1324                  * feature is turned on, then not all groups have this.
1325                  */
1326                 for (i = 0; i < sbi->s_groups_count; i++)
1327                         overhead += ext2_bg_has_super(sb, i) +
1328                                 ext2_bg_num_gdb(sb, i);
1329
1330                 /*
1331                  * Every block group has an inode bitmap, a block
1332                  * bitmap, and an inode table.
1333                  */
1334                 overhead += (sbi->s_groups_count *
1335                              (2 + sbi->s_itb_per_group));
1336                 sbi->s_overhead_last = overhead;
1337                 smp_wmb();
1338                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1339         }
1340
1341         buf->f_type = EXT2_SUPER_MAGIC;
1342         buf->f_bsize = sb->s_blocksize;
1343         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1344         buf->f_bfree = ext2_count_free_blocks(sb);
1345         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1346         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1347         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1348                 buf->f_bavail = 0;
1349         buf->f_files = le32_to_cpu(es->s_inodes_count);
1350         buf->f_ffree = ext2_count_free_inodes(sb);
1351         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1352         buf->f_namelen = EXT2_NAME_LEN;
1353         fsid = le64_to_cpup((void *)es->s_uuid) ^
1354                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1355         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1356         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1357         spin_unlock(&sbi->s_lock);
1358         return 0;
1359 }
1360
1361 static int ext2_get_sb(struct file_system_type *fs_type,
1362         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1363 {
1364         return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1365 }
1366
1367 #ifdef CONFIG_QUOTA
1368
1369 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1370  * acquiring the locks... As quota files are never truncated and quota code
1371  * itself serializes the operations (and noone else should touch the files)
1372  * we don't have to be afraid of races */
1373 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1374                                size_t len, loff_t off)
1375 {
1376         struct inode *inode = sb_dqopt(sb)->files[type];
1377         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1378         int err = 0;
1379         int offset = off & (sb->s_blocksize - 1);
1380         int tocopy;
1381         size_t toread;
1382         struct buffer_head tmp_bh;
1383         struct buffer_head *bh;
1384         loff_t i_size = i_size_read(inode);
1385
1386         if (off > i_size)
1387                 return 0;
1388         if (off+len > i_size)
1389                 len = i_size-off;
1390         toread = len;
1391         while (toread > 0) {
1392                 tocopy = sb->s_blocksize - offset < toread ?
1393                                 sb->s_blocksize - offset : toread;
1394
1395                 tmp_bh.b_state = 0;
1396                 tmp_bh.b_size = sb->s_blocksize;
1397                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1398                 if (err < 0)
1399                         return err;
1400                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1401                         memset(data, 0, tocopy);
1402                 else {
1403                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1404                         if (!bh)
1405                                 return -EIO;
1406                         memcpy(data, bh->b_data+offset, tocopy);
1407                         brelse(bh);
1408                 }
1409                 offset = 0;
1410                 toread -= tocopy;
1411                 data += tocopy;
1412                 blk++;
1413         }
1414         return len;
1415 }
1416
1417 /* Write to quotafile */
1418 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1419                                 const char *data, size_t len, loff_t off)
1420 {
1421         struct inode *inode = sb_dqopt(sb)->files[type];
1422         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1423         int err = 0;
1424         int offset = off & (sb->s_blocksize - 1);
1425         int tocopy;
1426         size_t towrite = len;
1427         struct buffer_head tmp_bh;
1428         struct buffer_head *bh;
1429
1430         mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1431         while (towrite > 0) {
1432                 tocopy = sb->s_blocksize - offset < towrite ?
1433                                 sb->s_blocksize - offset : towrite;
1434
1435                 tmp_bh.b_state = 0;
1436                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1437                 if (err < 0)
1438                         goto out;
1439                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1440                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1441                 else
1442                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1443                 if (!bh) {
1444                         err = -EIO;
1445                         goto out;
1446                 }
1447                 lock_buffer(bh);
1448                 memcpy(bh->b_data+offset, data, tocopy);
1449                 flush_dcache_page(bh->b_page);
1450                 set_buffer_uptodate(bh);
1451                 mark_buffer_dirty(bh);
1452                 unlock_buffer(bh);
1453                 brelse(bh);
1454                 offset = 0;
1455                 towrite -= tocopy;
1456                 data += tocopy;
1457                 blk++;
1458         }
1459 out:
1460         if (len == towrite) {
1461                 mutex_unlock(&inode->i_mutex);
1462                 return err;
1463         }
1464         if (inode->i_size < off+len-towrite)
1465                 i_size_write(inode, off+len-towrite);
1466         inode->i_version++;
1467         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1468         mark_inode_dirty(inode);
1469         mutex_unlock(&inode->i_mutex);
1470         return len - towrite;
1471 }
1472
1473 #endif
1474
1475 static struct file_system_type ext2_fs_type = {
1476         .owner          = THIS_MODULE,
1477         .name           = "ext2",
1478         .get_sb         = ext2_get_sb,
1479         .kill_sb        = kill_block_super,
1480         .fs_flags       = FS_REQUIRES_DEV,
1481 };
1482
1483 static int __init init_ext2_fs(void)
1484 {
1485         int err = init_ext2_xattr();
1486         if (err)
1487                 return err;
1488         err = init_inodecache();
1489         if (err)
1490                 goto out1;
1491         err = register_filesystem(&ext2_fs_type);
1492         if (err)
1493                 goto out;
1494         return 0;
1495 out:
1496         destroy_inodecache();
1497 out1:
1498         exit_ext2_xattr();
1499         return err;
1500 }
1501
1502 static void __exit exit_ext2_fs(void)
1503 {
1504         unregister_filesystem(&ext2_fs_type);
1505         destroy_inodecache();
1506         exit_ext2_xattr();
1507 }
1508
1509 module_init(init_ext2_fs)
1510 module_exit(exit_ext2_fs)