ext2 reservations
[safe/jmp/linux-2.6] / 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/smp_lock.h>
30 #include <linux/vfs.h>
31 #include <linux/seq_file.h>
32 #include <linux/mount.h>
33 #include <linux/log2.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);
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
45 void ext2_error (struct super_block * sb, const char * function,
46                  const char * fmt, ...)
47 {
48         va_list args;
49         struct ext2_sb_info *sbi = EXT2_SB(sb);
50         struct ext2_super_block *es = sbi->s_es;
51
52         if (!(sb->s_flags & MS_RDONLY)) {
53                 sbi->s_mount_state |= EXT2_ERROR_FS;
54                 es->s_state =
55                         cpu_to_le16(le16_to_cpu(es->s_state) | EXT2_ERROR_FS);
56                 ext2_sync_super(sb, es);
57         }
58
59         va_start(args, fmt);
60         printk(KERN_CRIT "EXT2-fs error (device %s): %s: ",sb->s_id, function);
61         vprintk(fmt, args);
62         printk("\n");
63         va_end(args);
64
65         if (test_opt(sb, ERRORS_PANIC))
66                 panic("EXT2-fs panic from previous error\n");
67         if (test_opt(sb, ERRORS_RO)) {
68                 printk("Remounting filesystem read-only\n");
69                 sb->s_flags |= MS_RDONLY;
70         }
71 }
72
73 void ext2_warning (struct super_block * sb, const char * function,
74                    const char * fmt, ...)
75 {
76         va_list args;
77
78         va_start(args, fmt);
79         printk(KERN_WARNING "EXT2-fs warning (device %s): %s: ",
80                sb->s_id, function);
81         vprintk(fmt, args);
82         printk("\n");
83         va_end(args);
84 }
85
86 void ext2_update_dynamic_rev(struct super_block *sb)
87 {
88         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
89
90         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
91                 return;
92
93         ext2_warning(sb, __FUNCTION__,
94                      "updating to rev %d because of new feature flag, "
95                      "running e2fsck is recommended",
96                      EXT2_DYNAMIC_REV);
97
98         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
99         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
100         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
101         /* leave es->s_feature_*compat flags alone */
102         /* es->s_uuid will be set by e2fsck if empty */
103
104         /*
105          * The rest of the superblock fields should be zero, and if not it
106          * means they are likely already in use, so leave them alone.  We
107          * can leave it up to e2fsck to clean up any inconsistencies there.
108          */
109 }
110
111 static void ext2_put_super (struct super_block * sb)
112 {
113         int db_count;
114         int i;
115         struct ext2_sb_info *sbi = EXT2_SB(sb);
116
117         ext2_xattr_put_super(sb);
118         if (!(sb->s_flags & MS_RDONLY)) {
119                 struct ext2_super_block *es = sbi->s_es;
120
121                 es->s_state = cpu_to_le16(sbi->s_mount_state);
122                 ext2_sync_super(sb, es);
123         }
124         db_count = sbi->s_gdb_count;
125         for (i = 0; i < db_count; i++)
126                 if (sbi->s_group_desc[i])
127                         brelse (sbi->s_group_desc[i]);
128         kfree(sbi->s_group_desc);
129         kfree(sbi->s_debts);
130         percpu_counter_destroy(&sbi->s_freeblocks_counter);
131         percpu_counter_destroy(&sbi->s_freeinodes_counter);
132         percpu_counter_destroy(&sbi->s_dirs_counter);
133         brelse (sbi->s_sbh);
134         sb->s_fs_info = NULL;
135         kfree(sbi);
136
137         return;
138 }
139
140 static struct kmem_cache * ext2_inode_cachep;
141
142 static struct inode *ext2_alloc_inode(struct super_block *sb)
143 {
144         struct ext2_inode_info *ei;
145         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
146         if (!ei)
147                 return NULL;
148 #ifdef CONFIG_EXT2_FS_POSIX_ACL
149         ei->i_acl = EXT2_ACL_NOT_CACHED;
150         ei->i_default_acl = EXT2_ACL_NOT_CACHED;
151 #endif
152         ei->i_block_alloc_info = NULL;
153         ei->vfs_inode.i_version = 1;
154         return &ei->vfs_inode;
155 }
156
157 static void ext2_destroy_inode(struct inode *inode)
158 {
159         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
160 }
161
162 static void init_once(struct kmem_cache * cachep, void *foo)
163 {
164         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
165
166         rwlock_init(&ei->i_meta_lock);
167 #ifdef CONFIG_EXT2_FS_XATTR
168         init_rwsem(&ei->xattr_sem);
169 #endif
170         mutex_init(&ei->truncate_mutex);
171         inode_init_once(&ei->vfs_inode);
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                                                 SLAB_MEM_SPREAD),
180                                              init_once);
181         if (ext2_inode_cachep == NULL)
182                 return -ENOMEM;
183         return 0;
184 }
185
186 static void destroy_inodecache(void)
187 {
188         kmem_cache_destroy(ext2_inode_cachep);
189 }
190
191 static void ext2_clear_inode(struct inode *inode)
192 {
193         struct ext2_block_alloc_info *rsv = EXT2_I(inode)->i_block_alloc_info;
194 #ifdef CONFIG_EXT2_FS_POSIX_ACL
195         struct ext2_inode_info *ei = EXT2_I(inode);
196
197         if (ei->i_acl && ei->i_acl != EXT2_ACL_NOT_CACHED) {
198                 posix_acl_release(ei->i_acl);
199                 ei->i_acl = EXT2_ACL_NOT_CACHED;
200         }
201         if (ei->i_default_acl && ei->i_default_acl != EXT2_ACL_NOT_CACHED) {
202                 posix_acl_release(ei->i_default_acl);
203                 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
204         }
205 #endif
206         ext2_discard_reservation(inode);
207         EXT2_I(inode)->i_block_alloc_info = NULL;
208         if (unlikely(rsv))
209                 kfree(rsv);
210 }
211
212 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
213 {
214         struct super_block *sb = vfs->mnt_sb;
215         struct ext2_sb_info *sbi = EXT2_SB(sb);
216         struct ext2_super_block *es = sbi->s_es;
217         unsigned long def_mount_opts;
218
219         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
220
221         if (sbi->s_sb_block != 1)
222                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
223         if (test_opt(sb, MINIX_DF))
224                 seq_puts(seq, ",minixdf");
225         if (test_opt(sb, GRPID))
226                 seq_puts(seq, ",grpid");
227         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
228                 seq_puts(seq, ",nogrpid");
229         if (sbi->s_resuid != EXT2_DEF_RESUID ||
230             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
231                 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
232         }
233         if (sbi->s_resgid != EXT2_DEF_RESGID ||
234             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
235                 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
236         }
237         if (test_opt(sb, ERRORS_CONT)) {
238                 int def_errors = le16_to_cpu(es->s_errors);
239
240                 if (def_errors == EXT2_ERRORS_PANIC ||
241                     def_errors == EXT2_ERRORS_RO) {
242                         seq_puts(seq, ",errors=continue");
243                 }
244         }
245         if (test_opt(sb, ERRORS_RO))
246                 seq_puts(seq, ",errors=remount-ro");
247         if (test_opt(sb, ERRORS_PANIC))
248                 seq_puts(seq, ",errors=panic");
249         if (test_opt(sb, NO_UID32))
250                 seq_puts(seq, ",nouid32");
251         if (test_opt(sb, DEBUG))
252                 seq_puts(seq, ",debug");
253         if (test_opt(sb, OLDALLOC))
254                 seq_puts(seq, ",oldalloc");
255
256 #ifdef CONFIG_EXT2_FS_XATTR
257         if (test_opt(sb, XATTR_USER))
258                 seq_puts(seq, ",user_xattr");
259         if (!test_opt(sb, XATTR_USER) &&
260             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
261                 seq_puts(seq, ",nouser_xattr");
262         }
263 #endif
264
265 #ifdef CONFIG_EXT2_FS_POSIX_ACL
266         if (test_opt(sb, POSIX_ACL))
267                 seq_puts(seq, ",acl");
268         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
269                 seq_puts(seq, ",noacl");
270 #endif
271
272         if (test_opt(sb, NOBH))
273                 seq_puts(seq, ",nobh");
274
275 #if defined(CONFIG_QUOTA)
276         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
277                 seq_puts(seq, ",usrquota");
278
279         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
280                 seq_puts(seq, ",grpquota");
281 #endif
282
283 #if defined(CONFIG_EXT2_FS_XIP)
284         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
285                 seq_puts(seq, ",xip");
286 #endif
287
288         return 0;
289 }
290
291 #ifdef CONFIG_QUOTA
292 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
293 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
294 #endif
295
296 static const struct super_operations ext2_sops = {
297         .alloc_inode    = ext2_alloc_inode,
298         .destroy_inode  = ext2_destroy_inode,
299         .read_inode     = ext2_read_inode,
300         .write_inode    = ext2_write_inode,
301         .delete_inode   = ext2_delete_inode,
302         .put_super      = ext2_put_super,
303         .write_super    = ext2_write_super,
304         .statfs         = ext2_statfs,
305         .remount_fs     = ext2_remount,
306         .clear_inode    = ext2_clear_inode,
307         .show_options   = ext2_show_options,
308 #ifdef CONFIG_QUOTA
309         .quota_read     = ext2_quota_read,
310         .quota_write    = ext2_quota_write,
311 #endif
312 };
313
314 static struct dentry *ext2_get_dentry(struct super_block *sb, void *vobjp)
315 {
316         __u32 *objp = vobjp;
317         unsigned long ino = objp[0];
318         __u32 generation = objp[1];
319         struct inode *inode;
320         struct dentry *result;
321
322         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
323                 return ERR_PTR(-ESTALE);
324         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
325                 return ERR_PTR(-ESTALE);
326
327         /* iget isn't really right if the inode is currently unallocated!!
328          * ext2_read_inode currently does appropriate checks, but
329          * it might be "neater" to call ext2_get_inode first and check
330          * if the inode is valid.....
331          */
332         inode = iget(sb, ino);
333         if (inode == NULL)
334                 return ERR_PTR(-ENOMEM);
335         if (is_bad_inode(inode) ||
336             (generation && inode->i_generation != generation)) {
337                 /* we didn't find the right inode.. */
338                 iput(inode);
339                 return ERR_PTR(-ESTALE);
340         }
341         /* now to find a dentry.
342          * If possible, get a well-connected one
343          */
344         result = d_alloc_anon(inode);
345         if (!result) {
346                 iput(inode);
347                 return ERR_PTR(-ENOMEM);
348         }
349         return result;
350 }
351
352 /* Yes, most of these are left as NULL!!
353  * A NULL value implies the default, which works with ext2-like file
354  * systems, but can be improved upon.
355  * Currently only get_parent is required.
356  */
357 static struct export_operations ext2_export_ops = {
358         .get_parent = ext2_get_parent,
359         .get_dentry = ext2_get_dentry,
360 };
361
362 static unsigned long get_sb_block(void **data)
363 {
364         unsigned long   sb_block;
365         char            *options = (char *) *data;
366
367         if (!options || strncmp(options, "sb=", 3) != 0)
368                 return 1;       /* Default location */
369         options += 3;
370         sb_block = simple_strtoul(options, &options, 0);
371         if (*options && *options != ',') {
372                 printk("EXT2-fs: Invalid sb specification: %s\n",
373                        (char *) *data);
374                 return 1;
375         }
376         if (*options == ',')
377                 options++;
378         *data = (void *) options;
379         return sb_block;
380 }
381
382 enum {
383         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
384         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
385         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
386         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
387         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
388         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
389 };
390
391 static match_table_t tokens = {
392         {Opt_bsd_df, "bsddf"},
393         {Opt_minix_df, "minixdf"},
394         {Opt_grpid, "grpid"},
395         {Opt_grpid, "bsdgroups"},
396         {Opt_nogrpid, "nogrpid"},
397         {Opt_nogrpid, "sysvgroups"},
398         {Opt_resgid, "resgid=%u"},
399         {Opt_resuid, "resuid=%u"},
400         {Opt_sb, "sb=%u"},
401         {Opt_err_cont, "errors=continue"},
402         {Opt_err_panic, "errors=panic"},
403         {Opt_err_ro, "errors=remount-ro"},
404         {Opt_nouid32, "nouid32"},
405         {Opt_nocheck, "check=none"},
406         {Opt_nocheck, "nocheck"},
407         {Opt_debug, "debug"},
408         {Opt_oldalloc, "oldalloc"},
409         {Opt_orlov, "orlov"},
410         {Opt_nobh, "nobh"},
411         {Opt_user_xattr, "user_xattr"},
412         {Opt_nouser_xattr, "nouser_xattr"},
413         {Opt_acl, "acl"},
414         {Opt_noacl, "noacl"},
415         {Opt_xip, "xip"},
416         {Opt_grpquota, "grpquota"},
417         {Opt_ignore, "noquota"},
418         {Opt_quota, "quota"},
419         {Opt_usrquota, "usrquota"},
420         {Opt_reservation, "reservation"},
421         {Opt_noreservation, "noreservation"},
422         {Opt_err, NULL}
423 };
424
425 static int parse_options (char * options,
426                           struct ext2_sb_info *sbi)
427 {
428         char * p;
429         substring_t args[MAX_OPT_ARGS];
430         int option;
431
432         if (!options)
433                 return 1;
434
435         while ((p = strsep (&options, ",")) != NULL) {
436                 int token;
437                 if (!*p)
438                         continue;
439
440                 token = match_token(p, tokens, args);
441                 switch (token) {
442                 case Opt_bsd_df:
443                         clear_opt (sbi->s_mount_opt, MINIX_DF);
444                         break;
445                 case Opt_minix_df:
446                         set_opt (sbi->s_mount_opt, MINIX_DF);
447                         break;
448                 case Opt_grpid:
449                         set_opt (sbi->s_mount_opt, GRPID);
450                         break;
451                 case Opt_nogrpid:
452                         clear_opt (sbi->s_mount_opt, GRPID);
453                         break;
454                 case Opt_resuid:
455                         if (match_int(&args[0], &option))
456                                 return 0;
457                         sbi->s_resuid = option;
458                         break;
459                 case Opt_resgid:
460                         if (match_int(&args[0], &option))
461                                 return 0;
462                         sbi->s_resgid = option;
463                         break;
464                 case Opt_sb:
465                         /* handled by get_sb_block() instead of here */
466                         /* *sb_block = match_int(&args[0]); */
467                         break;
468                 case Opt_err_panic:
469                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
470                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
471                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
472                         break;
473                 case Opt_err_ro:
474                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
475                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
476                         set_opt (sbi->s_mount_opt, ERRORS_RO);
477                         break;
478                 case Opt_err_cont:
479                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
480                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
481                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
482                         break;
483                 case Opt_nouid32:
484                         set_opt (sbi->s_mount_opt, NO_UID32);
485                         break;
486                 case Opt_nocheck:
487                         clear_opt (sbi->s_mount_opt, CHECK);
488                         break;
489                 case Opt_debug:
490                         set_opt (sbi->s_mount_opt, DEBUG);
491                         break;
492                 case Opt_oldalloc:
493                         set_opt (sbi->s_mount_opt, OLDALLOC);
494                         break;
495                 case Opt_orlov:
496                         clear_opt (sbi->s_mount_opt, OLDALLOC);
497                         break;
498                 case Opt_nobh:
499                         set_opt (sbi->s_mount_opt, NOBH);
500                         break;
501 #ifdef CONFIG_EXT2_FS_XATTR
502                 case Opt_user_xattr:
503                         set_opt (sbi->s_mount_opt, XATTR_USER);
504                         break;
505                 case Opt_nouser_xattr:
506                         clear_opt (sbi->s_mount_opt, XATTR_USER);
507                         break;
508 #else
509                 case Opt_user_xattr:
510                 case Opt_nouser_xattr:
511                         printk("EXT2 (no)user_xattr options not supported\n");
512                         break;
513 #endif
514 #ifdef CONFIG_EXT2_FS_POSIX_ACL
515                 case Opt_acl:
516                         set_opt(sbi->s_mount_opt, POSIX_ACL);
517                         break;
518                 case Opt_noacl:
519                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
520                         break;
521 #else
522                 case Opt_acl:
523                 case Opt_noacl:
524                         printk("EXT2 (no)acl options not supported\n");
525                         break;
526 #endif
527                 case Opt_xip:
528 #ifdef CONFIG_EXT2_FS_XIP
529                         set_opt (sbi->s_mount_opt, XIP);
530 #else
531                         printk("EXT2 xip option not supported\n");
532 #endif
533                         break;
534
535 #if defined(CONFIG_QUOTA)
536                 case Opt_quota:
537                 case Opt_usrquota:
538                         set_opt(sbi->s_mount_opt, USRQUOTA);
539                         break;
540
541                 case Opt_grpquota:
542                         set_opt(sbi->s_mount_opt, GRPQUOTA);
543                         break;
544 #else
545                 case Opt_quota:
546                 case Opt_usrquota:
547                 case Opt_grpquota:
548                         printk(KERN_ERR
549                                 "EXT2-fs: quota operations not supported.\n");
550
551                         break;
552 #endif
553
554                 case Opt_reservation:
555                         set_opt(sbi->s_mount_opt, RESERVATION);
556                         printk("reservations ON\n");
557                         break;
558                 case Opt_noreservation:
559                         clear_opt(sbi->s_mount_opt, RESERVATION);
560                         printk("reservations OFF\n");
561                         break;
562                 case Opt_ignore:
563                         break;
564                 default:
565                         return 0;
566                 }
567         }
568         return 1;
569 }
570
571 static int ext2_setup_super (struct super_block * sb,
572                               struct ext2_super_block * es,
573                               int read_only)
574 {
575         int res = 0;
576         struct ext2_sb_info *sbi = EXT2_SB(sb);
577
578         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
579                 printk ("EXT2-fs warning: revision level too high, "
580                         "forcing read-only mode\n");
581                 res = MS_RDONLY;
582         }
583         if (read_only)
584                 return res;
585         if (!(sbi->s_mount_state & EXT2_VALID_FS))
586                 printk ("EXT2-fs warning: mounting unchecked fs, "
587                         "running e2fsck is recommended\n");
588         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
589                 printk ("EXT2-fs warning: mounting fs with errors, "
590                         "running e2fsck is recommended\n");
591         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
592                  le16_to_cpu(es->s_mnt_count) >=
593                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
594                 printk ("EXT2-fs warning: maximal mount count reached, "
595                         "running e2fsck is recommended\n");
596         else if (le32_to_cpu(es->s_checkinterval) &&
597                 (le32_to_cpu(es->s_lastcheck) + le32_to_cpu(es->s_checkinterval) <= get_seconds()))
598                 printk ("EXT2-fs warning: checktime reached, "
599                         "running e2fsck is recommended\n");
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         es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
603         ext2_write_super(sb);
604         if (test_opt (sb, DEBUG))
605                 printk ("[EXT II FS %s, %s, bs=%lu, fs=%lu, gc=%lu, "
606                         "bpg=%lu, ipg=%lu, mo=%04lx]\n",
607                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
608                         sbi->s_frag_size,
609                         sbi->s_groups_count,
610                         EXT2_BLOCKS_PER_GROUP(sb),
611                         EXT2_INODES_PER_GROUP(sb),
612                         sbi->s_mount_opt);
613         return res;
614 }
615
616 static int ext2_check_descriptors (struct super_block * sb)
617 {
618         int i;
619         int desc_block = 0;
620         struct ext2_sb_info *sbi = EXT2_SB(sb);
621         unsigned long first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
622         unsigned long last_block;
623         struct ext2_group_desc * gdp = NULL;
624
625         ext2_debug ("Checking group descriptors");
626
627         for (i = 0; i < sbi->s_groups_count; i++)
628         {
629                 if (i == sbi->s_groups_count - 1)
630                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
631                 else
632                         last_block = first_block +
633                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
634
635                 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
636                         gdp = (struct ext2_group_desc *) sbi->s_group_desc[desc_block++]->b_data;
637                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
638                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
639                 {
640                         ext2_error (sb, "ext2_check_descriptors",
641                                     "Block bitmap for group %d"
642                                     " not in group (block %lu)!",
643                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
644                         return 0;
645                 }
646                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
647                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
648                 {
649                         ext2_error (sb, "ext2_check_descriptors",
650                                     "Inode bitmap for group %d"
651                                     " not in group (block %lu)!",
652                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
653                         return 0;
654                 }
655                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
656                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
657                     last_block)
658                 {
659                         ext2_error (sb, "ext2_check_descriptors",
660                                     "Inode table for group %d"
661                                     " not in group (block %lu)!",
662                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
663                         return 0;
664                 }
665                 first_block += EXT2_BLOCKS_PER_GROUP(sb);
666                 gdp++;
667         }
668         return 1;
669 }
670
671 /*
672  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
673  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
674  * We need to be 1 filesystem block less than the 2^32 sector limit.
675  */
676 static loff_t ext2_max_size(int bits)
677 {
678         loff_t res = EXT2_NDIR_BLOCKS;
679         /* This constant is calculated to be the largest file size for a
680          * dense, 4k-blocksize file such that the total number of
681          * sectors in the file, including data and all indirect blocks,
682          * does not exceed 2^32. */
683         const loff_t upper_limit = 0x1ff7fffd000LL;
684
685         res += 1LL << (bits-2);
686         res += 1LL << (2*(bits-2));
687         res += 1LL << (3*(bits-2));
688         res <<= bits;
689         if (res > upper_limit)
690                 res = upper_limit;
691         return res;
692 }
693
694 static unsigned long descriptor_loc(struct super_block *sb,
695                                     unsigned long logic_sb_block,
696                                     int nr)
697 {
698         struct ext2_sb_info *sbi = EXT2_SB(sb);
699         unsigned long bg, first_data_block, first_meta_bg;
700         int has_super = 0;
701         
702         first_data_block = le32_to_cpu(sbi->s_es->s_first_data_block);
703         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
704
705         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
706             nr < first_meta_bg)
707                 return (logic_sb_block + nr + 1);
708         bg = sbi->s_desc_per_block * nr;
709         if (ext2_bg_has_super(sb, bg))
710                 has_super = 1;
711         return (first_data_block + has_super + (bg * sbi->s_blocks_per_group));
712 }
713
714 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
715 {
716         struct buffer_head * bh;
717         struct ext2_sb_info * sbi;
718         struct ext2_super_block * es;
719         struct inode *root;
720         unsigned long block;
721         unsigned long sb_block = get_sb_block(&data);
722         unsigned long logic_sb_block;
723         unsigned long offset = 0;
724         unsigned long def_mount_opts;
725         int blocksize = BLOCK_SIZE;
726         int db_count;
727         int i, j;
728         __le32 features;
729         int err;
730
731         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
732         if (!sbi)
733                 return -ENOMEM;
734         sb->s_fs_info = sbi;
735         sbi->s_sb_block = sb_block;
736
737         /*
738          * See what the current blocksize for the device is, and
739          * use that as the blocksize.  Otherwise (or if the blocksize
740          * is smaller than the default) use the default.
741          * This is important for devices that have a hardware
742          * sectorsize that is larger than the default.
743          */
744         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
745         if (!blocksize) {
746                 printk ("EXT2-fs: unable to set blocksize\n");
747                 goto failed_sbi;
748         }
749
750         /*
751          * If the superblock doesn't start on a hardware sector boundary,
752          * calculate the offset.  
753          */
754         if (blocksize != BLOCK_SIZE) {
755                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
756                 offset = (sb_block*BLOCK_SIZE) % blocksize;
757         } else {
758                 logic_sb_block = sb_block;
759         }
760
761         if (!(bh = sb_bread(sb, logic_sb_block))) {
762                 printk ("EXT2-fs: unable to read superblock\n");
763                 goto failed_sbi;
764         }
765         /*
766          * Note: s_es must be initialized as soon as possible because
767          *       some ext2 macro-instructions depend on its value
768          */
769         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
770         sbi->s_es = es;
771         sb->s_magic = le16_to_cpu(es->s_magic);
772
773         if (sb->s_magic != EXT2_SUPER_MAGIC)
774                 goto cantfind_ext2;
775
776         /* Set defaults before we parse the mount options */
777         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
778         if (def_mount_opts & EXT2_DEFM_DEBUG)
779                 set_opt(sbi->s_mount_opt, DEBUG);
780         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
781                 set_opt(sbi->s_mount_opt, GRPID);
782         if (def_mount_opts & EXT2_DEFM_UID16)
783                 set_opt(sbi->s_mount_opt, NO_UID32);
784 #ifdef CONFIG_EXT2_FS_XATTR
785         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
786                 set_opt(sbi->s_mount_opt, XATTR_USER);
787 #endif
788 #ifdef CONFIG_EXT2_FS_POSIX_ACL
789         if (def_mount_opts & EXT2_DEFM_ACL)
790                 set_opt(sbi->s_mount_opt, POSIX_ACL);
791 #endif
792         
793         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
794                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
795         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_RO)
796                 set_opt(sbi->s_mount_opt, ERRORS_RO);
797         else
798                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
799
800         sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
801         sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
802         
803         set_opt(sbi->s_mount_opt, RESERVATION);
804
805         if (!parse_options ((char *) data, sbi))
806                 goto failed_mount;
807
808         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
809                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
810                  MS_POSIXACL : 0);
811
812         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
813                                     EXT2_MOUNT_XIP if not */
814
815         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
816             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
817              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
818              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
819                 printk("EXT2-fs warning: feature flags set on rev 0 fs, "
820                        "running e2fsck is recommended\n");
821         /*
822          * Check feature flags regardless of the revision level, since we
823          * previously didn't change the revision level when setting the flags,
824          * so there is a chance incompat flags are set on a rev 0 filesystem.
825          */
826         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
827         if (features) {
828                 printk("EXT2-fs: %s: couldn't mount because of "
829                        "unsupported optional features (%x).\n",
830                        sb->s_id, le32_to_cpu(features));
831                 goto failed_mount;
832         }
833         if (!(sb->s_flags & MS_RDONLY) &&
834             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
835                 printk("EXT2-fs: %s: couldn't mount RDWR because of "
836                        "unsupported optional features (%x).\n",
837                        sb->s_id, le32_to_cpu(features));
838                 goto failed_mount;
839         }
840
841         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
842
843         if ((ext2_use_xip(sb)) && ((blocksize != PAGE_SIZE) ||
844                                   (sb->s_blocksize != blocksize))) {
845                 if (!silent)
846                         printk("XIP: Unsupported blocksize\n");
847                 goto failed_mount;
848         }
849
850         /* If the blocksize doesn't match, re-read the thing.. */
851         if (sb->s_blocksize != blocksize) {
852                 brelse(bh);
853
854                 if (!sb_set_blocksize(sb, blocksize)) {
855                         printk(KERN_ERR "EXT2-fs: blocksize too small for device.\n");
856                         goto failed_sbi;
857                 }
858
859                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
860                 offset = (sb_block*BLOCK_SIZE) % blocksize;
861                 bh = sb_bread(sb, logic_sb_block);
862                 if(!bh) {
863                         printk("EXT2-fs: Couldn't read superblock on "
864                                "2nd try.\n");
865                         goto failed_sbi;
866                 }
867                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
868                 sbi->s_es = es;
869                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
870                         printk ("EXT2-fs: Magic mismatch, very weird !\n");
871                         goto failed_mount;
872                 }
873         }
874
875         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
876
877         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
878                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
879                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
880         } else {
881                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
882                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
883                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
884                     !is_power_of_2(sbi->s_inode_size) ||
885                     (sbi->s_inode_size > blocksize)) {
886                         printk ("EXT2-fs: unsupported inode size: %d\n",
887                                 sbi->s_inode_size);
888                         goto failed_mount;
889                 }
890         }
891
892         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
893                                    le32_to_cpu(es->s_log_frag_size);
894         if (sbi->s_frag_size == 0)
895                 goto cantfind_ext2;
896         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
897
898         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
899         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
900         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
901
902         if (EXT2_INODE_SIZE(sb) == 0)
903                 goto cantfind_ext2;
904         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
905         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
906                 goto cantfind_ext2;
907         sbi->s_itb_per_group = sbi->s_inodes_per_group /
908                                         sbi->s_inodes_per_block;
909         sbi->s_desc_per_block = sb->s_blocksize /
910                                         sizeof (struct ext2_group_desc);
911         sbi->s_sbh = bh;
912         sbi->s_mount_state = le16_to_cpu(es->s_state);
913         sbi->s_addr_per_block_bits =
914                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
915         sbi->s_desc_per_block_bits =
916                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
917
918         if (sb->s_magic != EXT2_SUPER_MAGIC)
919                 goto cantfind_ext2;
920
921         if (sb->s_blocksize != bh->b_size) {
922                 if (!silent)
923                         printk ("VFS: Unsupported blocksize on dev "
924                                 "%s.\n", sb->s_id);
925                 goto failed_mount;
926         }
927
928         if (sb->s_blocksize != sbi->s_frag_size) {
929                 printk ("EXT2-fs: fragsize %lu != blocksize %lu (not supported yet)\n",
930                         sbi->s_frag_size, sb->s_blocksize);
931                 goto failed_mount;
932         }
933
934         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
935                 printk ("EXT2-fs: #blocks per group too big: %lu\n",
936                         sbi->s_blocks_per_group);
937                 goto failed_mount;
938         }
939         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
940                 printk ("EXT2-fs: #fragments per group too big: %lu\n",
941                         sbi->s_frags_per_group);
942                 goto failed_mount;
943         }
944         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
945                 printk ("EXT2-fs: #inodes per group too big: %lu\n",
946                         sbi->s_inodes_per_group);
947                 goto failed_mount;
948         }
949
950         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
951                 goto cantfind_ext2;
952         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
953                                 le32_to_cpu(es->s_first_data_block) - 1)
954                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
955         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
956                    EXT2_DESC_PER_BLOCK(sb);
957         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
958         if (sbi->s_group_desc == NULL) {
959                 printk ("EXT2-fs: not enough memory\n");
960                 goto failed_mount;
961         }
962         bgl_lock_init(&sbi->s_blockgroup_lock);
963         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
964         if (!sbi->s_debts) {
965                 printk ("EXT2-fs: not enough memory\n");
966                 goto failed_mount_group_desc;
967         }
968         for (i = 0; i < db_count; i++) {
969                 block = descriptor_loc(sb, logic_sb_block, i);
970                 sbi->s_group_desc[i] = sb_bread(sb, block);
971                 if (!sbi->s_group_desc[i]) {
972                         for (j = 0; j < i; j++)
973                                 brelse (sbi->s_group_desc[j]);
974                         printk ("EXT2-fs: unable to read group descriptors\n");
975                         goto failed_mount_group_desc;
976                 }
977         }
978         if (!ext2_check_descriptors (sb)) {
979                 printk ("EXT2-fs: group descriptors corrupted!\n");
980                 goto failed_mount2;
981         }
982         sbi->s_gdb_count = db_count;
983         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
984         spin_lock_init(&sbi->s_next_gen_lock);
985
986         /* per fileystem reservation list head & lock */
987         spin_lock_init(&sbi->s_rsv_window_lock);
988         sbi->s_rsv_window_root = RB_ROOT;
989         /*
990          * Add a single, static dummy reservation to the start of the
991          * reservation window list --- it gives us a placeholder for
992          * append-at-start-of-list which makes the allocation logic
993          * _much_ simpler.
994          */
995         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
996         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
997         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
998         sbi->s_rsv_window_head.rsv_goal_size = 0;
999         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1000
1001         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1002                                 ext2_count_free_blocks(sb));
1003         if (!err) {
1004                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1005                                 ext2_count_free_inodes(sb));
1006         }
1007         if (!err) {
1008                 err = percpu_counter_init(&sbi->s_dirs_counter,
1009                                 ext2_count_dirs(sb));
1010         }
1011         if (err) {
1012                 printk(KERN_ERR "EXT2-fs: insufficient memory\n");
1013                 goto failed_mount3;
1014         }
1015         /*
1016          * set up enough so that it can read an inode
1017          */
1018         sb->s_op = &ext2_sops;
1019         sb->s_export_op = &ext2_export_ops;
1020         sb->s_xattr = ext2_xattr_handlers;
1021         root = iget(sb, EXT2_ROOT_INO);
1022         sb->s_root = d_alloc_root(root);
1023         if (!sb->s_root) {
1024                 iput(root);
1025                 printk(KERN_ERR "EXT2-fs: get root inode failed\n");
1026                 goto failed_mount3;
1027         }
1028         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1029                 dput(sb->s_root);
1030                 sb->s_root = NULL;
1031                 printk(KERN_ERR "EXT2-fs: corrupt root inode, run e2fsck\n");
1032                 goto failed_mount3;
1033         }
1034         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1035                 ext2_warning(sb, __FUNCTION__,
1036                         "mounting ext3 filesystem as ext2");
1037         ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1038         return 0;
1039
1040 cantfind_ext2:
1041         if (!silent)
1042                 printk("VFS: Can't find an ext2 filesystem on dev %s.\n",
1043                        sb->s_id);
1044         goto failed_mount;
1045 failed_mount3:
1046         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1047         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1048         percpu_counter_destroy(&sbi->s_dirs_counter);
1049 failed_mount2:
1050         for (i = 0; i < db_count; i++)
1051                 brelse(sbi->s_group_desc[i]);
1052 failed_mount_group_desc:
1053         kfree(sbi->s_group_desc);
1054         kfree(sbi->s_debts);
1055 failed_mount:
1056         brelse(bh);
1057 failed_sbi:
1058         sb->s_fs_info = NULL;
1059         kfree(sbi);
1060         return -EINVAL;
1061 }
1062
1063 static void ext2_commit_super (struct super_block * sb,
1064                                struct ext2_super_block * es)
1065 {
1066         es->s_wtime = cpu_to_le32(get_seconds());
1067         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1068         sb->s_dirt = 0;
1069 }
1070
1071 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
1072 {
1073         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1074         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1075         es->s_wtime = cpu_to_le32(get_seconds());
1076         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1077         sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1078         sb->s_dirt = 0;
1079 }
1080
1081 /*
1082  * In the second extended file system, it is not necessary to
1083  * write the super block since we use a mapping of the
1084  * disk super block in a buffer.
1085  *
1086  * However, this function is still used to set the fs valid
1087  * flags to 0.  We need to set this flag to 0 since the fs
1088  * may have been checked while mounted and e2fsck may have
1089  * set s_state to EXT2_VALID_FS after some corrections.
1090  */
1091
1092 void ext2_write_super (struct super_block * sb)
1093 {
1094         struct ext2_super_block * es;
1095         lock_kernel();
1096         if (!(sb->s_flags & MS_RDONLY)) {
1097                 es = EXT2_SB(sb)->s_es;
1098
1099                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS) {
1100                         ext2_debug ("setting valid to 0\n");
1101                         es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) &
1102                                                   ~EXT2_VALID_FS);
1103                         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1104                         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1105                         es->s_mtime = cpu_to_le32(get_seconds());
1106                         ext2_sync_super(sb, es);
1107                 } else
1108                         ext2_commit_super (sb, es);
1109         }
1110         sb->s_dirt = 0;
1111         unlock_kernel();
1112 }
1113
1114 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1115 {
1116         struct ext2_sb_info * sbi = EXT2_SB(sb);
1117         struct ext2_super_block * es;
1118         unsigned long old_mount_opt = sbi->s_mount_opt;
1119         struct ext2_mount_options old_opts;
1120         unsigned long old_sb_flags;
1121         int err;
1122
1123         /* Store the old options */
1124         old_sb_flags = sb->s_flags;
1125         old_opts.s_mount_opt = sbi->s_mount_opt;
1126         old_opts.s_resuid = sbi->s_resuid;
1127         old_opts.s_resgid = sbi->s_resgid;
1128
1129         /*
1130          * Allow the "check" option to be passed as a remount option.
1131          */
1132         if (!parse_options (data, sbi)) {
1133                 err = -EINVAL;
1134                 goto restore_opts;
1135         }
1136
1137         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1138                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1139
1140         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1141                                     EXT2_MOUNT_XIP if not */
1142
1143         if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1144                 printk("XIP: Unsupported blocksize\n");
1145                 err = -EINVAL;
1146                 goto restore_opts;
1147         }
1148
1149         es = sbi->s_es;
1150         if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1151             (old_mount_opt & EXT2_MOUNT_XIP)) &&
1152             invalidate_inodes(sb))
1153                 ext2_warning(sb, __FUNCTION__, "busy inodes while remounting "\
1154                              "xip remain in cache (no functional problem)");
1155         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1156                 return 0;
1157         if (*flags & MS_RDONLY) {
1158                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1159                     !(sbi->s_mount_state & EXT2_VALID_FS))
1160                         return 0;
1161                 /*
1162                  * OK, we are remounting a valid rw partition rdonly, so set
1163                  * the rdonly flag and then mark the partition as valid again.
1164                  */
1165                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1166                 es->s_mtime = cpu_to_le32(get_seconds());
1167         } else {
1168                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1169                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1170                 if (ret) {
1171                         printk("EXT2-fs: %s: couldn't remount RDWR because of "
1172                                "unsupported optional features (%x).\n",
1173                                sb->s_id, le32_to_cpu(ret));
1174                         err = -EROFS;
1175                         goto restore_opts;
1176                 }
1177                 /*
1178                  * Mounting a RDONLY partition read-write, so reread and
1179                  * store the current valid flag.  (It may have been changed
1180                  * by e2fsck since we originally mounted the partition.)
1181                  */
1182                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1183                 if (!ext2_setup_super (sb, es, 0))
1184                         sb->s_flags &= ~MS_RDONLY;
1185         }
1186         ext2_sync_super(sb, es);
1187         return 0;
1188 restore_opts:
1189         sbi->s_mount_opt = old_opts.s_mount_opt;
1190         sbi->s_resuid = old_opts.s_resuid;
1191         sbi->s_resgid = old_opts.s_resgid;
1192         sb->s_flags = old_sb_flags;
1193         return err;
1194 }
1195
1196 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1197 {
1198         struct super_block *sb = dentry->d_sb;
1199         struct ext2_sb_info *sbi = EXT2_SB(sb);
1200         struct ext2_super_block *es = sbi->s_es;
1201         u64 fsid;
1202
1203         if (test_opt (sb, MINIX_DF))
1204                 sbi->s_overhead_last = 0;
1205         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1206                 unsigned long i, overhead = 0;
1207                 smp_rmb();
1208
1209                 /*
1210                  * Compute the overhead (FS structures). This is constant
1211                  * for a given filesystem unless the number of block groups
1212                  * changes so we cache the previous value until it does.
1213                  */
1214
1215                 /*
1216                  * All of the blocks before first_data_block are
1217                  * overhead
1218                  */
1219                 overhead = le32_to_cpu(es->s_first_data_block);
1220
1221                 /*
1222                  * Add the overhead attributed to the superblock and
1223                  * block group descriptors.  If the sparse superblocks
1224                  * feature is turned on, then not all groups have this.
1225                  */
1226                 for (i = 0; i < sbi->s_groups_count; i++)
1227                         overhead += ext2_bg_has_super(sb, i) +
1228                                 ext2_bg_num_gdb(sb, i);
1229
1230                 /*
1231                  * Every block group has an inode bitmap, a block
1232                  * bitmap, and an inode table.
1233                  */
1234                 overhead += (sbi->s_groups_count *
1235                              (2 + sbi->s_itb_per_group));
1236                 sbi->s_overhead_last = overhead;
1237                 smp_wmb();
1238                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1239         }
1240
1241         buf->f_type = EXT2_SUPER_MAGIC;
1242         buf->f_bsize = sb->s_blocksize;
1243         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1244         buf->f_bfree = ext2_count_free_blocks(sb);
1245         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1246         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1247         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1248                 buf->f_bavail = 0;
1249         buf->f_files = le32_to_cpu(es->s_inodes_count);
1250         buf->f_ffree = ext2_count_free_inodes(sb);
1251         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1252         buf->f_namelen = EXT2_NAME_LEN;
1253         fsid = le64_to_cpup((void *)es->s_uuid) ^
1254                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1255         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1256         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1257         return 0;
1258 }
1259
1260 static int ext2_get_sb(struct file_system_type *fs_type,
1261         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1262 {
1263         return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1264 }
1265
1266 #ifdef CONFIG_QUOTA
1267
1268 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1269  * acquiring the locks... As quota files are never truncated and quota code
1270  * itself serializes the operations (and noone else should touch the files)
1271  * we don't have to be afraid of races */
1272 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1273                                size_t len, loff_t off)
1274 {
1275         struct inode *inode = sb_dqopt(sb)->files[type];
1276         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1277         int err = 0;
1278         int offset = off & (sb->s_blocksize - 1);
1279         int tocopy;
1280         size_t toread;
1281         struct buffer_head tmp_bh;
1282         struct buffer_head *bh;
1283         loff_t i_size = i_size_read(inode);
1284
1285         if (off > i_size)
1286                 return 0;
1287         if (off+len > i_size)
1288                 len = i_size-off;
1289         toread = len;
1290         while (toread > 0) {
1291                 tocopy = sb->s_blocksize - offset < toread ?
1292                                 sb->s_blocksize - offset : toread;
1293
1294                 tmp_bh.b_state = 0;
1295                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1296                 if (err < 0)
1297                         return err;
1298                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1299                         memset(data, 0, tocopy);
1300                 else {
1301                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1302                         if (!bh)
1303                                 return -EIO;
1304                         memcpy(data, bh->b_data+offset, tocopy);
1305                         brelse(bh);
1306                 }
1307                 offset = 0;
1308                 toread -= tocopy;
1309                 data += tocopy;
1310                 blk++;
1311         }
1312         return len;
1313 }
1314
1315 /* Write to quotafile */
1316 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1317                                 const char *data, size_t len, loff_t off)
1318 {
1319         struct inode *inode = sb_dqopt(sb)->files[type];
1320         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1321         int err = 0;
1322         int offset = off & (sb->s_blocksize - 1);
1323         int tocopy;
1324         size_t towrite = len;
1325         struct buffer_head tmp_bh;
1326         struct buffer_head *bh;
1327
1328         mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1329         while (towrite > 0) {
1330                 tocopy = sb->s_blocksize - offset < towrite ?
1331                                 sb->s_blocksize - offset : towrite;
1332
1333                 tmp_bh.b_state = 0;
1334                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1335                 if (err < 0)
1336                         goto out;
1337                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1338                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1339                 else
1340                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1341                 if (!bh) {
1342                         err = -EIO;
1343                         goto out;
1344                 }
1345                 lock_buffer(bh);
1346                 memcpy(bh->b_data+offset, data, tocopy);
1347                 flush_dcache_page(bh->b_page);
1348                 set_buffer_uptodate(bh);
1349                 mark_buffer_dirty(bh);
1350                 unlock_buffer(bh);
1351                 brelse(bh);
1352                 offset = 0;
1353                 towrite -= tocopy;
1354                 data += tocopy;
1355                 blk++;
1356         }
1357 out:
1358         if (len == towrite)
1359                 return err;
1360         if (inode->i_size < off+len-towrite)
1361                 i_size_write(inode, off+len-towrite);
1362         inode->i_version++;
1363         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1364         mark_inode_dirty(inode);
1365         mutex_unlock(&inode->i_mutex);
1366         return len - towrite;
1367 }
1368
1369 #endif
1370
1371 static struct file_system_type ext2_fs_type = {
1372         .owner          = THIS_MODULE,
1373         .name           = "ext2",
1374         .get_sb         = ext2_get_sb,
1375         .kill_sb        = kill_block_super,
1376         .fs_flags       = FS_REQUIRES_DEV,
1377 };
1378
1379 static int __init init_ext2_fs(void)
1380 {
1381         int err = init_ext2_xattr();
1382         if (err)
1383                 return err;
1384         err = init_inodecache();
1385         if (err)
1386                 goto out1;
1387         err = register_filesystem(&ext2_fs_type);
1388         if (err)
1389                 goto out;
1390         return 0;
1391 out:
1392         destroy_inodecache();
1393 out1:
1394         exit_ext2_xattr();
1395         return err;
1396 }
1397
1398 static void __exit exit_ext2_fs(void)
1399 {
1400         unregister_filesystem(&ext2_fs_type);
1401         destroy_inodecache();
1402         exit_ext2_xattr();
1403 }
1404
1405 module_init(init_ext2_fs)
1406 module_exit(exit_ext2_fs)