linux 2.6.16.38 w/ vs2.0.3-rc1
[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/config.h>
20 #include <linux/module.h>
21 #include <linux/string.h>
22 #include <linux/fs.h>
23 #include <linux/slab.h>
24 #include <linux/init.h>
25 #include <linux/blkdev.h>
26 #include <linux/parser.h>
27 #include <linux/random.h>
28 #include <linux/buffer_head.h>
29 #include <linux/smp_lock.h>
30 #include <linux/vfs.h>
31 #include <linux/seq_file.h>
32 #include <linux/mount.h>
33 #include <asm/uaccess.h>
34 #include "ext2.h"
35 #include "xattr.h"
36 #include "acl.h"
37 #include "xip.h"
38
39 static void ext2_sync_super(struct super_block *sb,
40                             struct ext2_super_block *es);
41 static int ext2_remount (struct super_block * sb, int * flags, char * data);
42 static int ext2_statfs (struct super_block * sb, struct kstatfs * buf);
43
44 void ext2_error (struct super_block * sb, const char * function,
45                  const char * fmt, ...)
46 {
47         va_list args;
48         struct ext2_sb_info *sbi = EXT2_SB(sb);
49         struct ext2_super_block *es = sbi->s_es;
50
51         if (!(sb->s_flags & MS_RDONLY)) {
52                 sbi->s_mount_state |= EXT2_ERROR_FS;
53                 es->s_state =
54                         cpu_to_le16(le16_to_cpu(es->s_state) | EXT2_ERROR_FS);
55                 ext2_sync_super(sb, es);
56         }
57
58         va_start(args, fmt);
59         printk(KERN_CRIT "EXT2-fs error (device %s): %s: ",sb->s_id, function);
60         vprintk(fmt, args);
61         printk("\n");
62         va_end(args);
63
64         if (test_opt(sb, ERRORS_PANIC))
65                 panic("EXT2-fs panic from previous error\n");
66         if (test_opt(sb, ERRORS_RO)) {
67                 printk("Remounting filesystem read-only\n");
68                 sb->s_flags |= MS_RDONLY;
69         }
70 }
71
72 void ext2_warning (struct super_block * sb, const char * function,
73                    const char * fmt, ...)
74 {
75         va_list args;
76
77         va_start(args, fmt);
78         printk(KERN_WARNING "EXT2-fs warning (device %s): %s: ",
79                sb->s_id, function);
80         vprintk(fmt, args);
81         printk("\n");
82         va_end(args);
83 }
84
85 void ext2_update_dynamic_rev(struct super_block *sb)
86 {
87         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
88
89         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
90                 return;
91
92         ext2_warning(sb, __FUNCTION__,
93                      "updating to rev %d because of new feature flag, "
94                      "running e2fsck is recommended",
95                      EXT2_DYNAMIC_REV);
96
97         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
98         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
99         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
100         /* leave es->s_feature_*compat flags alone */
101         /* es->s_uuid will be set by e2fsck if empty */
102
103         /*
104          * The rest of the superblock fields should be zero, and if not it
105          * means they are likely already in use, so leave them alone.  We
106          * can leave it up to e2fsck to clean up any inconsistencies there.
107          */
108 }
109
110 static void ext2_put_super (struct super_block * sb)
111 {
112         int db_count;
113         int i;
114         struct ext2_sb_info *sbi = EXT2_SB(sb);
115
116         ext2_xattr_put_super(sb);
117         if (!(sb->s_flags & MS_RDONLY)) {
118                 struct ext2_super_block *es = sbi->s_es;
119
120                 es->s_state = cpu_to_le16(sbi->s_mount_state);
121                 ext2_sync_super(sb, es);
122         }
123         db_count = sbi->s_gdb_count;
124         for (i = 0; i < db_count; i++)
125                 if (sbi->s_group_desc[i])
126                         brelse (sbi->s_group_desc[i]);
127         kfree(sbi->s_group_desc);
128         kfree(sbi->s_debts);
129         percpu_counter_destroy(&sbi->s_freeblocks_counter);
130         percpu_counter_destroy(&sbi->s_freeinodes_counter);
131         percpu_counter_destroy(&sbi->s_dirs_counter);
132         brelse (sbi->s_sbh);
133         sb->s_fs_info = NULL;
134         kfree(sbi);
135
136         return;
137 }
138
139 static kmem_cache_t * ext2_inode_cachep;
140
141 static struct inode *ext2_alloc_inode(struct super_block *sb)
142 {
143         struct ext2_inode_info *ei;
144         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, SLAB_KERNEL);
145         if (!ei)
146                 return NULL;
147 #ifdef CONFIG_EXT2_FS_POSIX_ACL
148         ei->i_acl = EXT2_ACL_NOT_CACHED;
149         ei->i_default_acl = EXT2_ACL_NOT_CACHED;
150 #endif
151         ei->vfs_inode.i_version = 1;
152         return &ei->vfs_inode;
153 }
154
155 static void ext2_destroy_inode(struct inode *inode)
156 {
157         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
158 }
159
160 static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
161 {
162         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
163
164         if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
165             SLAB_CTOR_CONSTRUCTOR) {
166                 rwlock_init(&ei->i_meta_lock);
167 #ifdef CONFIG_EXT2_FS_XATTR
168                 init_rwsem(&ei->xattr_sem);
169 #endif
170                 inode_init_once(&ei->vfs_inode);
171         }
172 }
173  
174 static int init_inodecache(void)
175 {
176         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
177                                              sizeof(struct ext2_inode_info),
178                                              0, SLAB_RECLAIM_ACCOUNT,
179                                              init_once, NULL);
180         if (ext2_inode_cachep == NULL)
181                 return -ENOMEM;
182         return 0;
183 }
184
185 static void destroy_inodecache(void)
186 {
187         if (kmem_cache_destroy(ext2_inode_cachep))
188                 printk(KERN_INFO "ext2_inode_cache: not all structures were freed\n");
189 }
190
191 static void ext2_clear_inode(struct inode *inode)
192 {
193 #ifdef CONFIG_EXT2_FS_POSIX_ACL
194         struct ext2_inode_info *ei = EXT2_I(inode);
195
196         if (ei->i_acl && ei->i_acl != EXT2_ACL_NOT_CACHED) {
197                 posix_acl_release(ei->i_acl);
198                 ei->i_acl = EXT2_ACL_NOT_CACHED;
199         }
200         if (ei->i_default_acl && ei->i_default_acl != EXT2_ACL_NOT_CACHED) {
201                 posix_acl_release(ei->i_default_acl);
202                 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
203         }
204 #endif
205 }
206
207 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
208 {
209         struct ext2_sb_info *sbi = EXT2_SB(vfs->mnt_sb);
210
211         if (sbi->s_mount_opt & EXT2_MOUNT_GRPID)
212                 seq_puts(seq, ",grpid");
213         else
214                 seq_puts(seq, ",nogrpid");
215
216 #if defined(CONFIG_QUOTA)
217         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
218                 seq_puts(seq, ",usrquota");
219
220         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
221                 seq_puts(seq, ",grpquota");
222 #endif
223
224 #if defined(CONFIG_EXT2_FS_XIP)
225         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
226                 seq_puts(seq, ",xip");
227 #endif
228
229         return 0;
230 }
231
232 #ifdef CONFIG_QUOTA
233 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
234 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
235 #endif
236
237 static struct super_operations ext2_sops = {
238         .alloc_inode    = ext2_alloc_inode,
239         .destroy_inode  = ext2_destroy_inode,
240         .read_inode     = ext2_read_inode,
241         .write_inode    = ext2_write_inode,
242         .put_inode      = ext2_put_inode,
243         .delete_inode   = ext2_delete_inode,
244         .put_super      = ext2_put_super,
245         .write_super    = ext2_write_super,
246         .statfs         = ext2_statfs,
247         .remount_fs     = ext2_remount,
248         .clear_inode    = ext2_clear_inode,
249         .show_options   = ext2_show_options,
250 #ifdef CONFIG_QUOTA
251         .quota_read     = ext2_quota_read,
252         .quota_write    = ext2_quota_write,
253 #endif
254 };
255
256 static struct dentry *ext2_get_dentry(struct super_block *sb, void *vobjp)
257 {
258         __u32 *objp = vobjp;
259         unsigned long ino = objp[0];
260         __u32 generation = objp[1];
261         struct inode *inode;
262         struct dentry *result;
263
264         if (ino != EXT2_ROOT_INO && ino < EXT2_FIRST_INO(sb))
265                 return ERR_PTR(-ESTALE);
266         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
267                 return ERR_PTR(-ESTALE);
268
269         /* iget isn't really right if the inode is currently unallocated!!
270          * ext2_read_inode currently does appropriate checks, but
271          * it might be "neater" to call ext2_get_inode first and check
272          * if the inode is valid.....
273          */
274         inode = iget(sb, ino);
275         if (inode == NULL)
276                 return ERR_PTR(-ENOMEM);
277         if (is_bad_inode(inode)
278             || (generation && inode->i_generation != generation)
279                 ) {
280                 /* we didn't find the right inode.. */
281                 iput(inode);
282                 return ERR_PTR(-ESTALE);
283         }
284         /* now to find a dentry.
285          * If possible, get a well-connected one
286          */
287         result = d_alloc_anon(inode);
288         if (!result) {
289                 iput(inode);
290                 return ERR_PTR(-ENOMEM);
291         }
292         return result;
293 }
294
295
296 /* Yes, most of these are left as NULL!!
297  * A NULL value implies the default, which works with ext2-like file
298  * systems, but can be improved upon.
299  * Currently only get_parent is required.
300  */
301 struct dentry *ext2_get_parent(struct dentry *child);
302 static struct export_operations ext2_export_ops = {
303         .get_parent = ext2_get_parent,
304         .get_dentry = ext2_get_dentry,
305 };
306
307 static unsigned long get_sb_block(void **data)
308 {
309         unsigned long   sb_block;
310         char            *options = (char *) *data;
311
312         if (!options || strncmp(options, "sb=", 3) != 0)
313                 return 1;       /* Default location */
314         options += 3;
315         sb_block = simple_strtoul(options, &options, 0);
316         if (*options && *options != ',') {
317                 printk("EXT2-fs: Invalid sb specification: %s\n",
318                        (char *) *data);
319                 return 1;
320         }
321         if (*options == ',')
322                 options++;
323         *data = (void *) options;
324         return sb_block;
325 }
326
327 enum {
328         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
329         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
330         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
331         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
332         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
333         Opt_usrquota, Opt_grpquota, Opt_tagxid
334 };
335
336 static match_table_t tokens = {
337         {Opt_bsd_df, "bsddf"},
338         {Opt_minix_df, "minixdf"},
339         {Opt_grpid, "grpid"},
340         {Opt_grpid, "bsdgroups"},
341         {Opt_nogrpid, "nogrpid"},
342         {Opt_nogrpid, "sysvgroups"},
343         {Opt_resgid, "resgid=%u"},
344         {Opt_resuid, "resuid=%u"},
345         {Opt_sb, "sb=%u"},
346         {Opt_err_cont, "errors=continue"},
347         {Opt_err_panic, "errors=panic"},
348         {Opt_err_ro, "errors=remount-ro"},
349         {Opt_nouid32, "nouid32"},
350         {Opt_nocheck, "check=none"},
351         {Opt_nocheck, "nocheck"},
352         {Opt_debug, "debug"},
353         {Opt_oldalloc, "oldalloc"},
354         {Opt_orlov, "orlov"},
355         {Opt_nobh, "nobh"},
356         {Opt_user_xattr, "user_xattr"},
357         {Opt_nouser_xattr, "nouser_xattr"},
358         {Opt_acl, "acl"},
359         {Opt_noacl, "noacl"},
360         {Opt_xip, "xip"},
361         {Opt_tagxid, "tagxid"},
362         {Opt_grpquota, "grpquota"},
363         {Opt_ignore, "noquota"},
364         {Opt_quota, "quota"},
365         {Opt_usrquota, "usrquota"},
366         {Opt_err, NULL}
367 };
368
369 static int parse_options (char * options,
370                           struct ext2_sb_info *sbi)
371 {
372         char * p;
373         substring_t args[MAX_OPT_ARGS];
374         unsigned long kind = EXT2_MOUNT_ERRORS_CONT;
375         int option;
376
377         if (!options)
378                 return 1;
379
380         while ((p = strsep (&options, ",")) != NULL) {
381                 int token;
382                 if (!*p)
383                         continue;
384
385                 token = match_token(p, tokens, args);
386                 switch (token) {
387                 case Opt_bsd_df:
388                         clear_opt (sbi->s_mount_opt, MINIX_DF);
389                         break;
390                 case Opt_minix_df:
391                         set_opt (sbi->s_mount_opt, MINIX_DF);
392                         break;
393                 case Opt_grpid:
394                         set_opt (sbi->s_mount_opt, GRPID);
395                         break;
396                 case Opt_nogrpid:
397                         clear_opt (sbi->s_mount_opt, GRPID);
398                         break;
399                 case Opt_resuid:
400                         if (match_int(&args[0], &option))
401                                 return 0;
402                         sbi->s_resuid = option;
403                         break;
404                 case Opt_resgid:
405                         if (match_int(&args[0], &option))
406                                 return 0;
407                         sbi->s_resgid = option;
408                         break;
409                 case Opt_sb:
410                         /* handled by get_sb_block() instead of here */
411                         /* *sb_block = match_int(&args[0]); */
412                         break;
413                 case Opt_err_panic:
414                         kind = EXT2_MOUNT_ERRORS_PANIC;
415                         break;
416                 case Opt_err_ro:
417                         kind = EXT2_MOUNT_ERRORS_RO;
418                         break;
419                 case Opt_err_cont:
420                         kind = EXT2_MOUNT_ERRORS_CONT;
421                         break;
422                 case Opt_nouid32:
423                         set_opt (sbi->s_mount_opt, NO_UID32);
424                         break;
425 #ifndef CONFIG_INOXID_NONE
426                 case Opt_tagxid:
427                         set_opt (sbi->s_mount_opt, TAGXID);
428                         break;
429 #endif
430                 case Opt_nocheck:
431                         clear_opt (sbi->s_mount_opt, CHECK);
432                         break;
433                 case Opt_debug:
434                         set_opt (sbi->s_mount_opt, DEBUG);
435                         break;
436                 case Opt_oldalloc:
437                         set_opt (sbi->s_mount_opt, OLDALLOC);
438                         break;
439                 case Opt_orlov:
440                         clear_opt (sbi->s_mount_opt, OLDALLOC);
441                         break;
442                 case Opt_nobh:
443                         set_opt (sbi->s_mount_opt, NOBH);
444                         break;
445 #ifdef CONFIG_EXT2_FS_XATTR
446                 case Opt_user_xattr:
447                         set_opt (sbi->s_mount_opt, XATTR_USER);
448                         break;
449                 case Opt_nouser_xattr:
450                         clear_opt (sbi->s_mount_opt, XATTR_USER);
451                         break;
452 #else
453                 case Opt_user_xattr:
454                 case Opt_nouser_xattr:
455                         printk("EXT2 (no)user_xattr options not supported\n");
456                         break;
457 #endif
458 #ifdef CONFIG_EXT2_FS_POSIX_ACL
459                 case Opt_acl:
460                         set_opt(sbi->s_mount_opt, POSIX_ACL);
461                         break;
462                 case Opt_noacl:
463                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
464                         break;
465 #else
466                 case Opt_acl:
467                 case Opt_noacl:
468                         printk("EXT2 (no)acl options not supported\n");
469                         break;
470 #endif
471                 case Opt_xip:
472 #ifdef CONFIG_EXT2_FS_XIP
473                         set_opt (sbi->s_mount_opt, XIP);
474 #else
475                         printk("EXT2 xip option not supported\n");
476 #endif
477                         break;
478
479 #if defined(CONFIG_QUOTA)
480                 case Opt_quota:
481                 case Opt_usrquota:
482                         set_opt(sbi->s_mount_opt, USRQUOTA);
483                         break;
484
485                 case Opt_grpquota:
486                         set_opt(sbi->s_mount_opt, GRPQUOTA);
487                         break;
488 #else
489                 case Opt_quota:
490                 case Opt_usrquota:
491                 case Opt_grpquota:
492                         printk(KERN_ERR
493                                 "EXT2-fs: quota operations not supported.\n");
494
495                         break;
496 #endif
497
498                 case Opt_ignore:
499                         break;
500                 default:
501                         return 0;
502                 }
503         }
504         sbi->s_mount_opt |= kind;
505         return 1;
506 }
507
508 static int ext2_setup_super (struct super_block * sb,
509                               struct ext2_super_block * es,
510                               int read_only)
511 {
512         int res = 0;
513         struct ext2_sb_info *sbi = EXT2_SB(sb);
514
515         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
516                 printk ("EXT2-fs warning: revision level too high, "
517                         "forcing read-only mode\n");
518                 res = MS_RDONLY;
519         }
520         if (read_only)
521                 return res;
522         if (!(sbi->s_mount_state & EXT2_VALID_FS))
523                 printk ("EXT2-fs warning: mounting unchecked fs, "
524                         "running e2fsck is recommended\n");
525         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
526                 printk ("EXT2-fs warning: mounting fs with errors, "
527                         "running e2fsck is recommended\n");
528         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
529                  le16_to_cpu(es->s_mnt_count) >=
530                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
531                 printk ("EXT2-fs warning: maximal mount count reached, "
532                         "running e2fsck is recommended\n");
533         else if (le32_to_cpu(es->s_checkinterval) &&
534                 (le32_to_cpu(es->s_lastcheck) + le32_to_cpu(es->s_checkinterval) <= get_seconds()))
535                 printk ("EXT2-fs warning: checktime reached, "
536                         "running e2fsck is recommended\n");
537         if (!le16_to_cpu(es->s_max_mnt_count))
538                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
539         es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
540         ext2_write_super(sb);
541         if (test_opt (sb, DEBUG))
542                 printk ("[EXT II FS %s, %s, bs=%lu, fs=%lu, gc=%lu, "
543                         "bpg=%lu, ipg=%lu, mo=%04lx]\n",
544                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
545                         sbi->s_frag_size,
546                         sbi->s_groups_count,
547                         EXT2_BLOCKS_PER_GROUP(sb),
548                         EXT2_INODES_PER_GROUP(sb),
549                         sbi->s_mount_opt);
550         return res;
551 }
552
553 static int ext2_check_descriptors (struct super_block * sb)
554 {
555         int i;
556         int desc_block = 0;
557         struct ext2_sb_info *sbi = EXT2_SB(sb);
558         unsigned long block = le32_to_cpu(sbi->s_es->s_first_data_block);
559         struct ext2_group_desc * gdp = NULL;
560
561         ext2_debug ("Checking group descriptors");
562
563         for (i = 0; i < sbi->s_groups_count; i++)
564         {
565                 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
566                         gdp = (struct ext2_group_desc *) sbi->s_group_desc[desc_block++]->b_data;
567                 if (le32_to_cpu(gdp->bg_block_bitmap) < block ||
568                     le32_to_cpu(gdp->bg_block_bitmap) >= block + EXT2_BLOCKS_PER_GROUP(sb))
569                 {
570                         ext2_error (sb, "ext2_check_descriptors",
571                                     "Block bitmap for group %d"
572                                     " not in group (block %lu)!",
573                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
574                         return 0;
575                 }
576                 if (le32_to_cpu(gdp->bg_inode_bitmap) < block ||
577                     le32_to_cpu(gdp->bg_inode_bitmap) >= block + EXT2_BLOCKS_PER_GROUP(sb))
578                 {
579                         ext2_error (sb, "ext2_check_descriptors",
580                                     "Inode bitmap for group %d"
581                                     " not in group (block %lu)!",
582                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
583                         return 0;
584                 }
585                 if (le32_to_cpu(gdp->bg_inode_table) < block ||
586                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group >=
587                     block + EXT2_BLOCKS_PER_GROUP(sb))
588                 {
589                         ext2_error (sb, "ext2_check_descriptors",
590                                     "Inode table for group %d"
591                                     " not in group (block %lu)!",
592                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
593                         return 0;
594                 }
595                 block += EXT2_BLOCKS_PER_GROUP(sb);
596                 gdp++;
597         }
598         return 1;
599 }
600
601 #define log2(n) ffz(~(n))
602  
603 /*
604  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
605  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
606  * We need to be 1 filesystem block less than the 2^32 sector limit.
607  */
608 static loff_t ext2_max_size(int bits)
609 {
610         loff_t res = EXT2_NDIR_BLOCKS;
611         /* This constant is calculated to be the largest file size for a
612          * dense, 4k-blocksize file such that the total number of
613          * sectors in the file, including data and all indirect blocks,
614          * does not exceed 2^32. */
615         const loff_t upper_limit = 0x1ff7fffd000LL;
616
617         res += 1LL << (bits-2);
618         res += 1LL << (2*(bits-2));
619         res += 1LL << (3*(bits-2));
620         res <<= bits;
621         if (res > upper_limit)
622                 res = upper_limit;
623         return res;
624 }
625
626 static unsigned long descriptor_loc(struct super_block *sb,
627                                     unsigned long logic_sb_block,
628                                     int nr)
629 {
630         struct ext2_sb_info *sbi = EXT2_SB(sb);
631         unsigned long bg, first_data_block, first_meta_bg;
632         int has_super = 0;
633         
634         first_data_block = le32_to_cpu(sbi->s_es->s_first_data_block);
635         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
636
637         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
638             nr < first_meta_bg)
639                 return (logic_sb_block + nr + 1);
640         bg = sbi->s_desc_per_block * nr;
641         if (ext2_bg_has_super(sb, bg))
642                 has_super = 1;
643         return (first_data_block + has_super + (bg * sbi->s_blocks_per_group));
644 }
645
646 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
647 {
648         struct buffer_head * bh;
649         struct ext2_sb_info * sbi;
650         struct ext2_super_block * es;
651         struct inode *root;
652         unsigned long block;
653         unsigned long sb_block = get_sb_block(&data);
654         unsigned long logic_sb_block;
655         unsigned long offset = 0;
656         unsigned long def_mount_opts;
657         int blocksize = BLOCK_SIZE;
658         int db_count;
659         int i, j;
660         __le32 features;
661
662         sbi = kmalloc(sizeof(*sbi), GFP_KERNEL);
663         if (!sbi)
664                 return -ENOMEM;
665         sb->s_fs_info = sbi;
666         memset(sbi, 0, sizeof(*sbi));
667
668         /*
669          * See what the current blocksize for the device is, and
670          * use that as the blocksize.  Otherwise (or if the blocksize
671          * is smaller than the default) use the default.
672          * This is important for devices that have a hardware
673          * sectorsize that is larger than the default.
674          */
675         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
676         if (!blocksize) {
677                 printk ("EXT2-fs: unable to set blocksize\n");
678                 goto failed_sbi;
679         }
680
681         /*
682          * If the superblock doesn't start on a hardware sector boundary,
683          * calculate the offset.  
684          */
685         if (blocksize != BLOCK_SIZE) {
686                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
687                 offset = (sb_block*BLOCK_SIZE) % blocksize;
688         } else {
689                 logic_sb_block = sb_block;
690         }
691
692         if (!(bh = sb_bread(sb, logic_sb_block))) {
693                 printk ("EXT2-fs: unable to read superblock\n");
694                 goto failed_sbi;
695         }
696         /*
697          * Note: s_es must be initialized as soon as possible because
698          *       some ext2 macro-instructions depend on its value
699          */
700         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
701         sbi->s_es = es;
702         sb->s_magic = le16_to_cpu(es->s_magic);
703
704         if (sb->s_magic != EXT2_SUPER_MAGIC)
705                 goto cantfind_ext2;
706
707         /* Set defaults before we parse the mount options */
708         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
709         if (def_mount_opts & EXT2_DEFM_DEBUG)
710                 set_opt(sbi->s_mount_opt, DEBUG);
711         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
712                 set_opt(sbi->s_mount_opt, GRPID);
713         if (def_mount_opts & EXT2_DEFM_UID16)
714                 set_opt(sbi->s_mount_opt, NO_UID32);
715         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
716                 set_opt(sbi->s_mount_opt, XATTR_USER);
717         if (def_mount_opts & EXT2_DEFM_ACL)
718                 set_opt(sbi->s_mount_opt, POSIX_ACL);
719         
720         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
721                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
722         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_RO)
723                 set_opt(sbi->s_mount_opt, ERRORS_RO);
724
725         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
726         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
727         
728         if (!parse_options ((char *) data, sbi))
729                 goto failed_mount;
730
731         if (EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_TAGXID)
732                 sb->s_flags |= MS_TAGXID;
733         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
734                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
735                  MS_POSIXACL : 0);
736
737         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
738                                     EXT2_MOUNT_XIP if not */
739
740         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
741             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
742              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
743              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
744                 printk("EXT2-fs warning: feature flags set on rev 0 fs, "
745                        "running e2fsck is recommended\n");
746         /*
747          * Check feature flags regardless of the revision level, since we
748          * previously didn't change the revision level when setting the flags,
749          * so there is a chance incompat flags are set on a rev 0 filesystem.
750          */
751         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
752         if (features) {
753                 printk("EXT2-fs: %s: couldn't mount because of "
754                        "unsupported optional features (%x).\n",
755                        sb->s_id, le32_to_cpu(features));
756                 goto failed_mount;
757         }
758         if (!(sb->s_flags & MS_RDONLY) &&
759             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
760                 printk("EXT2-fs: %s: couldn't mount RDWR because of "
761                        "unsupported optional features (%x).\n",
762                        sb->s_id, le32_to_cpu(features));
763                 goto failed_mount;
764         }
765
766         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
767
768         if ((ext2_use_xip(sb)) && ((blocksize != PAGE_SIZE) ||
769                                   (sb->s_blocksize != blocksize))) {
770                 if (!silent)
771                         printk("XIP: Unsupported blocksize\n");
772                 goto failed_mount;
773         }
774
775         /* If the blocksize doesn't match, re-read the thing.. */
776         if (sb->s_blocksize != blocksize) {
777                 brelse(bh);
778
779                 if (!sb_set_blocksize(sb, blocksize)) {
780                         printk(KERN_ERR "EXT2-fs: blocksize too small for device.\n");
781                         goto failed_sbi;
782                 }
783
784                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
785                 offset = (sb_block*BLOCK_SIZE) % blocksize;
786                 bh = sb_bread(sb, logic_sb_block);
787                 if(!bh) {
788                         printk("EXT2-fs: Couldn't read superblock on "
789                                "2nd try.\n");
790                         goto failed_sbi;
791                 }
792                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
793                 sbi->s_es = es;
794                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
795                         printk ("EXT2-fs: Magic mismatch, very weird !\n");
796                         goto failed_mount;
797                 }
798         }
799
800         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
801
802         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
803                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
804                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
805         } else {
806                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
807                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
808                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
809                     (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
810                     (sbi->s_inode_size > blocksize)) {
811                         printk ("EXT2-fs: unsupported inode size: %d\n",
812                                 sbi->s_inode_size);
813                         goto failed_mount;
814                 }
815         }
816
817         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
818                                    le32_to_cpu(es->s_log_frag_size);
819         if (sbi->s_frag_size == 0)
820                 goto cantfind_ext2;
821         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
822
823         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
824         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
825         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
826
827         if (EXT2_INODE_SIZE(sb) == 0)
828                 goto cantfind_ext2;
829         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
830         if (sbi->s_inodes_per_block == 0)
831                 goto cantfind_ext2;
832         sbi->s_itb_per_group = sbi->s_inodes_per_group /
833                                         sbi->s_inodes_per_block;
834         sbi->s_desc_per_block = sb->s_blocksize /
835                                         sizeof (struct ext2_group_desc);
836         sbi->s_sbh = bh;
837         sbi->s_mount_state = le16_to_cpu(es->s_state);
838         sbi->s_addr_per_block_bits =
839                 log2 (EXT2_ADDR_PER_BLOCK(sb));
840         sbi->s_desc_per_block_bits =
841                 log2 (EXT2_DESC_PER_BLOCK(sb));
842
843         if (sb->s_magic != EXT2_SUPER_MAGIC)
844                 goto cantfind_ext2;
845
846         if (sb->s_blocksize != bh->b_size) {
847                 if (!silent)
848                         printk ("VFS: Unsupported blocksize on dev "
849                                 "%s.\n", sb->s_id);
850                 goto failed_mount;
851         }
852
853         if (sb->s_blocksize != sbi->s_frag_size) {
854                 printk ("EXT2-fs: fragsize %lu != blocksize %lu (not supported yet)\n",
855                         sbi->s_frag_size, sb->s_blocksize);
856                 goto failed_mount;
857         }
858
859         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
860                 printk ("EXT2-fs: #blocks per group too big: %lu\n",
861                         sbi->s_blocks_per_group);
862                 goto failed_mount;
863         }
864         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
865                 printk ("EXT2-fs: #fragments per group too big: %lu\n",
866                         sbi->s_frags_per_group);
867                 goto failed_mount;
868         }
869         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
870                 printk ("EXT2-fs: #inodes per group too big: %lu\n",
871                         sbi->s_inodes_per_group);
872                 goto failed_mount;
873         }
874
875         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
876                 goto cantfind_ext2;
877         sbi->s_groups_count = (le32_to_cpu(es->s_blocks_count) -
878                                         le32_to_cpu(es->s_first_data_block) +
879                                        EXT2_BLOCKS_PER_GROUP(sb) - 1) /
880                                        EXT2_BLOCKS_PER_GROUP(sb);
881         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
882                    EXT2_DESC_PER_BLOCK(sb);
883         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
884         if (sbi->s_group_desc == NULL) {
885                 printk ("EXT2-fs: not enough memory\n");
886                 goto failed_mount;
887         }
888         percpu_counter_init(&sbi->s_freeblocks_counter);
889         percpu_counter_init(&sbi->s_freeinodes_counter);
890         percpu_counter_init(&sbi->s_dirs_counter);
891         bgl_lock_init(&sbi->s_blockgroup_lock);
892         sbi->s_debts = kmalloc(sbi->s_groups_count * sizeof(*sbi->s_debts),
893                                GFP_KERNEL);
894         if (!sbi->s_debts) {
895                 printk ("EXT2-fs: not enough memory\n");
896                 goto failed_mount_group_desc;
897         }
898         memset(sbi->s_debts, 0, sbi->s_groups_count * sizeof(*sbi->s_debts));
899         for (i = 0; i < db_count; i++) {
900                 block = descriptor_loc(sb, logic_sb_block, i);
901                 sbi->s_group_desc[i] = sb_bread(sb, block);
902                 if (!sbi->s_group_desc[i]) {
903                         for (j = 0; j < i; j++)
904                                 brelse (sbi->s_group_desc[j]);
905                         printk ("EXT2-fs: unable to read group descriptors\n");
906                         goto failed_mount_group_desc;
907                 }
908         }
909         if (!ext2_check_descriptors (sb)) {
910                 printk ("EXT2-fs: group descriptors corrupted!\n");
911                 db_count = i;
912                 goto failed_mount2;
913         }
914         sbi->s_gdb_count = db_count;
915         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
916         spin_lock_init(&sbi->s_next_gen_lock);
917         /*
918          * set up enough so that it can read an inode
919          */
920         sb->s_op = &ext2_sops;
921         sb->s_export_op = &ext2_export_ops;
922         sb->s_xattr = ext2_xattr_handlers;
923         root = iget(sb, EXT2_ROOT_INO);
924         sb->s_root = d_alloc_root(root);
925         if (!sb->s_root) {
926                 iput(root);
927                 printk(KERN_ERR "EXT2-fs: get root inode failed\n");
928                 goto failed_mount2;
929         }
930         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
931                 dput(sb->s_root);
932                 sb->s_root = NULL;
933                 printk(KERN_ERR "EXT2-fs: corrupt root inode, run e2fsck\n");
934                 goto failed_mount2;
935         }
936         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
937                 ext2_warning(sb, __FUNCTION__,
938                         "mounting ext3 filesystem as ext2");
939         ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
940         percpu_counter_mod(&sbi->s_freeblocks_counter,
941                                 ext2_count_free_blocks(sb));
942         percpu_counter_mod(&sbi->s_freeinodes_counter,
943                                 ext2_count_free_inodes(sb));
944         percpu_counter_mod(&sbi->s_dirs_counter,
945                                 ext2_count_dirs(sb));
946         return 0;
947
948 cantfind_ext2:
949         if (!silent)
950                 printk("VFS: Can't find an ext2 filesystem on dev %s.\n",
951                        sb->s_id);
952         goto failed_mount;
953
954 failed_mount2:
955         for (i = 0; i < db_count; i++)
956                 brelse(sbi->s_group_desc[i]);
957 failed_mount_group_desc:
958         kfree(sbi->s_group_desc);
959         kfree(sbi->s_debts);
960 failed_mount:
961         brelse(bh);
962 failed_sbi:
963         sb->s_fs_info = NULL;
964         kfree(sbi);
965         return -EINVAL;
966 }
967
968 static void ext2_commit_super (struct super_block * sb,
969                                struct ext2_super_block * es)
970 {
971         es->s_wtime = cpu_to_le32(get_seconds());
972         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
973         sb->s_dirt = 0;
974 }
975
976 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
977 {
978         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
979         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
980         es->s_wtime = cpu_to_le32(get_seconds());
981         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
982         sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
983         sb->s_dirt = 0;
984 }
985
986 /*
987  * In the second extended file system, it is not necessary to
988  * write the super block since we use a mapping of the
989  * disk super block in a buffer.
990  *
991  * However, this function is still used to set the fs valid
992  * flags to 0.  We need to set this flag to 0 since the fs
993  * may have been checked while mounted and e2fsck may have
994  * set s_state to EXT2_VALID_FS after some corrections.
995  */
996
997 void ext2_write_super (struct super_block * sb)
998 {
999         struct ext2_super_block * es;
1000         lock_kernel();
1001         if (!(sb->s_flags & MS_RDONLY)) {
1002                 es = EXT2_SB(sb)->s_es;
1003
1004                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS) {
1005                         ext2_debug ("setting valid to 0\n");
1006                         es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) &
1007                                                   ~EXT2_VALID_FS);
1008                         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1009                         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1010                         es->s_mtime = cpu_to_le32(get_seconds());
1011                         ext2_sync_super(sb, es);
1012                 } else
1013                         ext2_commit_super (sb, es);
1014         }
1015         sb->s_dirt = 0;
1016         unlock_kernel();
1017 }
1018
1019 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1020 {
1021         struct ext2_sb_info * sbi = EXT2_SB(sb);
1022         struct ext2_super_block * es;
1023         unsigned long old_mount_opt = sbi->s_mount_opt;
1024         struct ext2_mount_options old_opts;
1025         unsigned long old_sb_flags;
1026         int err;
1027
1028         /* Store the old options */
1029         old_sb_flags = sb->s_flags;
1030         old_opts.s_mount_opt = sbi->s_mount_opt;
1031         old_opts.s_resuid = sbi->s_resuid;
1032         old_opts.s_resgid = sbi->s_resgid;
1033
1034         /*
1035          * Allow the "check" option to be passed as a remount option.
1036          */
1037         if (!parse_options (data, sbi)) {
1038                 err = -EINVAL;
1039                 goto restore_opts;
1040         }
1041
1042         if ((sbi->s_mount_opt & EXT2_MOUNT_TAGXID) &&
1043                 !(sb->s_flags & MS_TAGXID)) {
1044                 printk("EXT2-fs: %s: tagxid not permitted on remount.\n",
1045                        sb->s_id);
1046                 return -EINVAL;
1047         }
1048
1049         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1050                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1051
1052         es = sbi->s_es;
1053         if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1054             (old_mount_opt & EXT2_MOUNT_XIP)) &&
1055             invalidate_inodes(sb))
1056                 ext2_warning(sb, __FUNCTION__, "busy inodes while remounting "\
1057                              "xip remain in cache (no functional problem)");
1058         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1059                 return 0;
1060         if (*flags & MS_RDONLY) {
1061                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1062                     !(sbi->s_mount_state & EXT2_VALID_FS))
1063                         return 0;
1064                 /*
1065                  * OK, we are remounting a valid rw partition rdonly, so set
1066                  * the rdonly flag and then mark the partition as valid again.
1067                  */
1068                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1069                 es->s_mtime = cpu_to_le32(get_seconds());
1070         } else {
1071                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1072                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1073                 if (ret) {
1074                         printk("EXT2-fs: %s: couldn't remount RDWR because of "
1075                                "unsupported optional features (%x).\n",
1076                                sb->s_id, le32_to_cpu(ret));
1077                         err = -EROFS;
1078                         goto restore_opts;
1079                 }
1080                 /*
1081                  * Mounting a RDONLY partition read-write, so reread and
1082                  * store the current valid flag.  (It may have been changed
1083                  * by e2fsck since we originally mounted the partition.)
1084                  */
1085                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1086                 if (!ext2_setup_super (sb, es, 0))
1087                         sb->s_flags &= ~MS_RDONLY;
1088         }
1089         ext2_sync_super(sb, es);
1090         return 0;
1091 restore_opts:
1092         sbi->s_mount_opt = old_opts.s_mount_opt;
1093         sbi->s_resuid = old_opts.s_resuid;
1094         sbi->s_resgid = old_opts.s_resgid;
1095         sb->s_flags = old_sb_flags;
1096         return err;
1097 }
1098
1099 static int ext2_statfs (struct super_block * sb, struct kstatfs * buf)
1100 {
1101         struct ext2_sb_info *sbi = EXT2_SB(sb);
1102         unsigned long overhead;
1103         int i;
1104
1105         if (test_opt (sb, MINIX_DF))
1106                 overhead = 0;
1107         else {
1108                 /*
1109                  * Compute the overhead (FS structures)
1110                  */
1111
1112                 /*
1113                  * All of the blocks before first_data_block are
1114                  * overhead
1115                  */
1116                 overhead = le32_to_cpu(sbi->s_es->s_first_data_block);
1117
1118                 /*
1119                  * Add the overhead attributed to the superblock and
1120                  * block group descriptors.  If the sparse superblocks
1121                  * feature is turned on, then not all groups have this.
1122                  */
1123                 for (i = 0; i < sbi->s_groups_count; i++)
1124                         overhead += ext2_bg_has_super(sb, i) +
1125                                 ext2_bg_num_gdb(sb, i);
1126
1127                 /*
1128                  * Every block group has an inode bitmap, a block
1129                  * bitmap, and an inode table.
1130                  */
1131                 overhead += (sbi->s_groups_count *
1132                              (2 + sbi->s_itb_per_group));
1133         }
1134
1135         buf->f_type = EXT2_SUPER_MAGIC;
1136         buf->f_bsize = sb->s_blocksize;
1137         buf->f_blocks = le32_to_cpu(sbi->s_es->s_blocks_count) - overhead;
1138         buf->f_bfree = ext2_count_free_blocks(sb);
1139         buf->f_bavail = buf->f_bfree - le32_to_cpu(sbi->s_es->s_r_blocks_count);
1140         if (buf->f_bfree < le32_to_cpu(sbi->s_es->s_r_blocks_count))
1141                 buf->f_bavail = 0;
1142         buf->f_files = le32_to_cpu(sbi->s_es->s_inodes_count);
1143         buf->f_ffree = ext2_count_free_inodes (sb);
1144         buf->f_namelen = EXT2_NAME_LEN;
1145         return 0;
1146 }
1147
1148 static struct super_block *ext2_get_sb(struct file_system_type *fs_type,
1149         int flags, const char *dev_name, void *data)
1150 {
1151         return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1152 }
1153
1154 #ifdef CONFIG_QUOTA
1155
1156 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1157  * acquiring the locks... As quota files are never truncated and quota code
1158  * itself serializes the operations (and noone else should touch the files)
1159  * we don't have to be afraid of races */
1160 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1161                                size_t len, loff_t off)
1162 {
1163         struct inode *inode = sb_dqopt(sb)->files[type];
1164         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1165         int err = 0;
1166         int offset = off & (sb->s_blocksize - 1);
1167         int tocopy;
1168         size_t toread;
1169         struct buffer_head tmp_bh;
1170         struct buffer_head *bh;
1171         loff_t i_size = i_size_read(inode);
1172
1173         if (off > i_size)
1174                 return 0;
1175         if (off+len > i_size)
1176                 len = i_size-off;
1177         toread = len;
1178         while (toread > 0) {
1179                 tocopy = sb->s_blocksize - offset < toread ?
1180                                 sb->s_blocksize - offset : toread;
1181
1182                 tmp_bh.b_state = 0;
1183                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1184                 if (err)
1185                         return err;
1186                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1187                         memset(data, 0, tocopy);
1188                 else {
1189                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1190                         if (!bh)
1191                                 return -EIO;
1192                         memcpy(data, bh->b_data+offset, tocopy);
1193                         brelse(bh);
1194                 }
1195                 offset = 0;
1196                 toread -= tocopy;
1197                 data += tocopy;
1198                 blk++;
1199         }
1200         return len;
1201 }
1202
1203 /* Write to quotafile */
1204 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1205                                 const char *data, size_t len, loff_t off)
1206 {
1207         struct inode *inode = sb_dqopt(sb)->files[type];
1208         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1209         int err = 0;
1210         int offset = off & (sb->s_blocksize - 1);
1211         int tocopy;
1212         size_t towrite = len;
1213         struct buffer_head tmp_bh;
1214         struct buffer_head *bh;
1215
1216         mutex_lock(&inode->i_mutex);
1217         while (towrite > 0) {
1218                 tocopy = sb->s_blocksize - offset < towrite ?
1219                                 sb->s_blocksize - offset : towrite;
1220
1221                 tmp_bh.b_state = 0;
1222                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1223                 if (err)
1224                         goto out;
1225                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1226                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1227                 else
1228                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1229                 if (!bh) {
1230                         err = -EIO;
1231                         goto out;
1232                 }
1233                 lock_buffer(bh);
1234                 memcpy(bh->b_data+offset, data, tocopy);
1235                 flush_dcache_page(bh->b_page);
1236                 set_buffer_uptodate(bh);
1237                 mark_buffer_dirty(bh);
1238                 unlock_buffer(bh);
1239                 brelse(bh);
1240                 offset = 0;
1241                 towrite -= tocopy;
1242                 data += tocopy;
1243                 blk++;
1244         }
1245 out:
1246         if (len == towrite)
1247                 return err;
1248         if (inode->i_size < off+len-towrite)
1249                 i_size_write(inode, off+len-towrite);
1250         inode->i_version++;
1251         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1252         mark_inode_dirty(inode);
1253         mutex_unlock(&inode->i_mutex);
1254         return len - towrite;
1255 }
1256
1257 #endif
1258
1259 static struct file_system_type ext2_fs_type = {
1260         .owner          = THIS_MODULE,
1261         .name           = "ext2",
1262         .get_sb         = ext2_get_sb,
1263         .kill_sb        = kill_block_super,
1264         .fs_flags       = FS_REQUIRES_DEV,
1265 };
1266
1267 static int __init init_ext2_fs(void)
1268 {
1269         int err = init_ext2_xattr();
1270         if (err)
1271                 return err;
1272         err = init_inodecache();
1273         if (err)
1274                 goto out1;
1275         err = register_filesystem(&ext2_fs_type);
1276         if (err)
1277                 goto out;
1278         return 0;
1279 out:
1280         destroy_inodecache();
1281 out1:
1282         exit_ext2_xattr();
1283         return err;
1284 }
1285
1286 static void __exit exit_ext2_fs(void)
1287 {
1288         unregister_filesystem(&ext2_fs_type);
1289         destroy_inodecache();
1290         exit_ext2_xattr();
1291 }
1292
1293 module_init(init_ext2_fs)
1294 module_exit(exit_ext2_fs)