]> git.proxmox.com Git - mirror_ubuntu-focal-kernel.git/blame - fs/ext2/super.c
Linux 2.6.36-rc6
[mirror_ubuntu-focal-kernel.git] / fs / ext2 / super.c
CommitLineData
1da177e4
LT
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
1da177e4
LT
19#include <linux/module.h>
20#include <linux/string.h>
8fc2751b 21#include <linux/fs.h>
1da177e4
LT
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>
a5694255 28#include <linux/exportfs.h>
1da177e4 29#include <linux/vfs.h>
8fc2751b
MB
30#include <linux/seq_file.h>
31#include <linux/mount.h>
d8ea6cf8 32#include <linux/log2.h>
74abb989 33#include <linux/quotaops.h>
1da177e4
LT
34#include <asm/uaccess.h>
35#include "ext2.h"
36#include "xattr.h"
37#include "acl.h"
6d79125b 38#include "xip.h"
1da177e4
LT
39
40static void ext2_sync_super(struct super_block *sb,
ee6921eb 41 struct ext2_super_block *es, int wait);
1da177e4 42static int ext2_remount (struct super_block * sb, int * flags, char * data);
726c3342 43static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
40f31dd4 44static int ext2_sync_fs(struct super_block *sb, int wait);
1da177e4
LT
45
46void ext2_error (struct super_block * sb, const char * function,
47 const char * fmt, ...)
48{
49 va_list args;
50 struct ext2_sb_info *sbi = EXT2_SB(sb);
51 struct ext2_super_block *es = sbi->s_es;
52
53 if (!(sb->s_flags & MS_RDONLY)) {
c15271f4 54 spin_lock(&sbi->s_lock);
1da177e4 55 sbi->s_mount_state |= EXT2_ERROR_FS;
31f68e13 56 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
c15271f4 57 spin_unlock(&sbi->s_lock);
ee6921eb 58 ext2_sync_super(sb, es, 1);
1da177e4
LT
59 }
60
61 va_start(args, fmt);
2314b07c 62 printk(KERN_CRIT "EXT2-fs (%s): error: %s: ", sb->s_id, function);
1da177e4
LT
63 vprintk(fmt, args);
64 printk("\n");
65 va_end(args);
66
67 if (test_opt(sb, ERRORS_PANIC))
2314b07c 68 panic("EXT2-fs: panic from previous error\n");
1da177e4 69 if (test_opt(sb, ERRORS_RO)) {
2314b07c
OR
70 ext2_msg(sb, KERN_CRIT,
71 "error: remounting filesystem read-only");
1da177e4
LT
72 sb->s_flags |= MS_RDONLY;
73 }
74}
75
2314b07c
OR
76void ext2_msg(struct super_block *sb, const char *prefix,
77 const char *fmt, ...)
1da177e4
LT
78{
79 va_list args;
80
81 va_start(args, fmt);
2314b07c 82 printk("%sEXT2-fs (%s): ", prefix, sb->s_id);
1da177e4
LT
83 vprintk(fmt, args);
84 printk("\n");
85 va_end(args);
86}
87
c15271f4
JB
88/*
89 * This must be called with sbi->s_lock held.
90 */
1da177e4
LT
91void ext2_update_dynamic_rev(struct super_block *sb)
92{
93 struct ext2_super_block *es = EXT2_SB(sb)->s_es;
94
95 if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
96 return;
97
2314b07c
OR
98 ext2_msg(sb, KERN_WARNING,
99 "warning: updating to rev %d because of "
100 "new feature flag, running e2fsck is recommended",
1da177e4
LT
101 EXT2_DYNAMIC_REV);
102
103 es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
104 es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
105 es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
106 /* leave es->s_feature_*compat flags alone */
107 /* es->s_uuid will be set by e2fsck if empty */
108
109 /*
110 * The rest of the superblock fields should be zero, and if not it
111 * means they are likely already in use, so leave them alone. We
112 * can leave it up to e2fsck to clean up any inconsistencies there.
113 */
114}
115
116static void ext2_put_super (struct super_block * sb)
117{
118 int db_count;
119 int i;
120 struct ext2_sb_info *sbi = EXT2_SB(sb);
121
e0ccfd95
CH
122 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
123
8c85e125
CH
124 if (sb->s_dirt)
125 ext2_write_super(sb);
126
1da177e4
LT
127 ext2_xattr_put_super(sb);
128 if (!(sb->s_flags & MS_RDONLY)) {
129 struct ext2_super_block *es = sbi->s_es;
130
c15271f4 131 spin_lock(&sbi->s_lock);
1da177e4 132 es->s_state = cpu_to_le16(sbi->s_mount_state);
c15271f4 133 spin_unlock(&sbi->s_lock);
ee6921eb 134 ext2_sync_super(sb, es, 1);
1da177e4
LT
135 }
136 db_count = sbi->s_gdb_count;
137 for (i = 0; i < db_count; i++)
138 if (sbi->s_group_desc[i])
139 brelse (sbi->s_group_desc[i]);
140 kfree(sbi->s_group_desc);
141 kfree(sbi->s_debts);
142 percpu_counter_destroy(&sbi->s_freeblocks_counter);
143 percpu_counter_destroy(&sbi->s_freeinodes_counter);
144 percpu_counter_destroy(&sbi->s_dirs_counter);
145 brelse (sbi->s_sbh);
146 sb->s_fs_info = NULL;
18a82eb9 147 kfree(sbi->s_blockgroup_lock);
1da177e4 148 kfree(sbi);
1da177e4
LT
149}
150
e18b890b 151static struct kmem_cache * ext2_inode_cachep;
1da177e4
LT
152
153static struct inode *ext2_alloc_inode(struct super_block *sb)
154{
155 struct ext2_inode_info *ei;
e94b1766 156 ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
1da177e4
LT
157 if (!ei)
158 return NULL;
a686cd89 159 ei->i_block_alloc_info = NULL;
1da177e4
LT
160 ei->vfs_inode.i_version = 1;
161 return &ei->vfs_inode;
162}
163
164static void ext2_destroy_inode(struct inode *inode)
165{
166 kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
167}
168
51cc5068 169static void init_once(void *foo)
1da177e4
LT
170{
171 struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
172
a35afb83 173 rwlock_init(&ei->i_meta_lock);
1da177e4 174#ifdef CONFIG_EXT2_FS_XATTR
a35afb83 175 init_rwsem(&ei->xattr_sem);
1da177e4 176#endif
a686cd89 177 mutex_init(&ei->truncate_mutex);
a35afb83 178 inode_init_once(&ei->vfs_inode);
1da177e4 179}
20c2df83 180
1da177e4
LT
181static int init_inodecache(void)
182{
183 ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
184 sizeof(struct ext2_inode_info),
fffb60f9
PJ
185 0, (SLAB_RECLAIM_ACCOUNT|
186 SLAB_MEM_SPREAD),
20c2df83 187 init_once);
1da177e4
LT
188 if (ext2_inode_cachep == NULL)
189 return -ENOMEM;
190 return 0;
191}
192
193static void destroy_inodecache(void)
194{
1a1d92c1 195 kmem_cache_destroy(ext2_inode_cachep);
1da177e4
LT
196}
197
8fc2751b
MB
198static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
199{
93d44cb2
MS
200 struct super_block *sb = vfs->mnt_sb;
201 struct ext2_sb_info *sbi = EXT2_SB(sb);
202 struct ext2_super_block *es = sbi->s_es;
203 unsigned long def_mount_opts;
8fc2751b 204
c15271f4 205 spin_lock(&sbi->s_lock);
93d44cb2
MS
206 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
207
208 if (sbi->s_sb_block != 1)
209 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
210 if (test_opt(sb, MINIX_DF))
211 seq_puts(seq, ",minixdf");
212 if (test_opt(sb, GRPID))
8fc2751b 213 seq_puts(seq, ",grpid");
93d44cb2
MS
214 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
215 seq_puts(seq, ",nogrpid");
216 if (sbi->s_resuid != EXT2_DEF_RESUID ||
217 le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
218 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
219 }
220 if (sbi->s_resgid != EXT2_DEF_RESGID ||
221 le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
222 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
223 }
14e11e10 224 if (test_opt(sb, ERRORS_RO)) {
93d44cb2
MS
225 int def_errors = le16_to_cpu(es->s_errors);
226
227 if (def_errors == EXT2_ERRORS_PANIC ||
14e11e10
AK
228 def_errors == EXT2_ERRORS_CONTINUE) {
229 seq_puts(seq, ",errors=remount-ro");
93d44cb2
MS
230 }
231 }
14e11e10
AK
232 if (test_opt(sb, ERRORS_CONT))
233 seq_puts(seq, ",errors=continue");
93d44cb2
MS
234 if (test_opt(sb, ERRORS_PANIC))
235 seq_puts(seq, ",errors=panic");
236 if (test_opt(sb, NO_UID32))
237 seq_puts(seq, ",nouid32");
238 if (test_opt(sb, DEBUG))
239 seq_puts(seq, ",debug");
240 if (test_opt(sb, OLDALLOC))
241 seq_puts(seq, ",oldalloc");
242
243#ifdef CONFIG_EXT2_FS_XATTR
244 if (test_opt(sb, XATTR_USER))
245 seq_puts(seq, ",user_xattr");
246 if (!test_opt(sb, XATTR_USER) &&
247 (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
248 seq_puts(seq, ",nouser_xattr");
249 }
250#endif
251
252#ifdef CONFIG_EXT2_FS_POSIX_ACL
253 if (test_opt(sb, POSIX_ACL))
254 seq_puts(seq, ",acl");
255 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
256 seq_puts(seq, ",noacl");
257#endif
258
259 if (test_opt(sb, NOBH))
260 seq_puts(seq, ",nobh");
8fc2751b
MB
261
262#if defined(CONFIG_QUOTA)
263 if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
264 seq_puts(seq, ",usrquota");
265
266 if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
267 seq_puts(seq, ",grpquota");
268#endif
269
83541796
CO
270#if defined(CONFIG_EXT2_FS_XIP)
271 if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
272 seq_puts(seq, ",xip");
273#endif
274
35c879dc
MS
275 if (!test_opt(sb, RESERVATION))
276 seq_puts(seq, ",noreservation");
277
c15271f4 278 spin_unlock(&sbi->s_lock);
8fc2751b
MB
279 return 0;
280}
281
1da177e4
LT
282#ifdef CONFIG_QUOTA
283static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
284static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
285#endif
286
ee9b6d61 287static const struct super_operations ext2_sops = {
1da177e4
LT
288 .alloc_inode = ext2_alloc_inode,
289 .destroy_inode = ext2_destroy_inode,
1da177e4 290 .write_inode = ext2_write_inode,
72edc4d0 291 .evict_inode = ext2_evict_inode,
1da177e4
LT
292 .put_super = ext2_put_super,
293 .write_super = ext2_write_super,
40f31dd4 294 .sync_fs = ext2_sync_fs,
1da177e4
LT
295 .statfs = ext2_statfs,
296 .remount_fs = ext2_remount,
8fc2751b 297 .show_options = ext2_show_options,
1da177e4
LT
298#ifdef CONFIG_QUOTA
299 .quota_read = ext2_quota_read,
300 .quota_write = ext2_quota_write,
301#endif
302};
303
2e4c68e3
CH
304static struct inode *ext2_nfs_get_inode(struct super_block *sb,
305 u64 ino, u32 generation)
ecaff756 306{
ecaff756 307 struct inode *inode;
ecaff756
N
308
309 if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
310 return ERR_PTR(-ESTALE);
311 if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
312 return ERR_PTR(-ESTALE);
313
314 /* iget isn't really right if the inode is currently unallocated!!
315 * ext2_read_inode currently does appropriate checks, but
316 * it might be "neater" to call ext2_get_inode first and check
317 * if the inode is valid.....
318 */
52fcf703
DH
319 inode = ext2_iget(sb, ino);
320 if (IS_ERR(inode))
321 return ERR_CAST(inode);
322 if (generation && inode->i_generation != generation) {
ecaff756
N
323 /* we didn't find the right inode.. */
324 iput(inode);
325 return ERR_PTR(-ESTALE);
326 }
2e4c68e3
CH
327 return inode;
328}
329
330static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
331 int fh_len, int fh_type)
332{
333 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
334 ext2_nfs_get_inode);
335}
336
337static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
338 int fh_len, int fh_type)
339{
340 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
341 ext2_nfs_get_inode);
ecaff756
N
342}
343
1da177e4
LT
344/* Yes, most of these are left as NULL!!
345 * A NULL value implies the default, which works with ext2-like file
346 * systems, but can be improved upon.
347 * Currently only get_parent is required.
348 */
39655164 349static const struct export_operations ext2_export_ops = {
2e4c68e3
CH
350 .fh_to_dentry = ext2_fh_to_dentry,
351 .fh_to_parent = ext2_fh_to_parent,
1da177e4
LT
352 .get_parent = ext2_get_parent,
353};
354
355static unsigned long get_sb_block(void **data)
356{
357 unsigned long sb_block;
358 char *options = (char *) *data;
359
360 if (!options || strncmp(options, "sb=", 3) != 0)
361 return 1; /* Default location */
362 options += 3;
363 sb_block = simple_strtoul(options, &options, 0);
364 if (*options && *options != ',') {
365 printk("EXT2-fs: Invalid sb specification: %s\n",
366 (char *) *data);
367 return 1;
368 }
369 if (*options == ',')
370 options++;
371 *data = (void *) options;
372 return sb_block;
373}
374
375enum {
376 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
8fc2751b 377 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
2860b733 378 Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
8fc2751b
MB
379 Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
380 Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
a686cd89 381 Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
1da177e4
LT
382};
383
a447c093 384static const match_table_t tokens = {
1da177e4
LT
385 {Opt_bsd_df, "bsddf"},
386 {Opt_minix_df, "minixdf"},
387 {Opt_grpid, "grpid"},
388 {Opt_grpid, "bsdgroups"},
389 {Opt_nogrpid, "nogrpid"},
390 {Opt_nogrpid, "sysvgroups"},
391 {Opt_resgid, "resgid=%u"},
392 {Opt_resuid, "resuid=%u"},
393 {Opt_sb, "sb=%u"},
394 {Opt_err_cont, "errors=continue"},
395 {Opt_err_panic, "errors=panic"},
396 {Opt_err_ro, "errors=remount-ro"},
397 {Opt_nouid32, "nouid32"},
398 {Opt_nocheck, "check=none"},
399 {Opt_nocheck, "nocheck"},
1da177e4
LT
400 {Opt_debug, "debug"},
401 {Opt_oldalloc, "oldalloc"},
402 {Opt_orlov, "orlov"},
403 {Opt_nobh, "nobh"},
404 {Opt_user_xattr, "user_xattr"},
405 {Opt_nouser_xattr, "nouser_xattr"},
406 {Opt_acl, "acl"},
407 {Opt_noacl, "noacl"},
6d79125b 408 {Opt_xip, "xip"},
8fc2751b 409 {Opt_grpquota, "grpquota"},
1da177e4 410 {Opt_ignore, "noquota"},
8fc2751b
MB
411 {Opt_quota, "quota"},
412 {Opt_usrquota, "usrquota"},
a686cd89
MB
413 {Opt_reservation, "reservation"},
414 {Opt_noreservation, "noreservation"},
1da177e4
LT
415 {Opt_err, NULL}
416};
417
2314b07c 418static int parse_options(char *options, struct super_block *sb)
1da177e4 419{
2314b07c
OR
420 char *p;
421 struct ext2_sb_info *sbi = EXT2_SB(sb);
1da177e4 422 substring_t args[MAX_OPT_ARGS];
1da177e4
LT
423 int option;
424
425 if (!options)
426 return 1;
427
428 while ((p = strsep (&options, ",")) != NULL) {
429 int token;
430 if (!*p)
431 continue;
432
433 token = match_token(p, tokens, args);
434 switch (token) {
435 case Opt_bsd_df:
436 clear_opt (sbi->s_mount_opt, MINIX_DF);
437 break;
438 case Opt_minix_df:
439 set_opt (sbi->s_mount_opt, MINIX_DF);
440 break;
441 case Opt_grpid:
442 set_opt (sbi->s_mount_opt, GRPID);
443 break;
444 case Opt_nogrpid:
445 clear_opt (sbi->s_mount_opt, GRPID);
446 break;
447 case Opt_resuid:
448 if (match_int(&args[0], &option))
449 return 0;
450 sbi->s_resuid = option;
451 break;
452 case Opt_resgid:
453 if (match_int(&args[0], &option))
454 return 0;
455 sbi->s_resgid = option;
456 break;
457 case Opt_sb:
458 /* handled by get_sb_block() instead of here */
459 /* *sb_block = match_int(&args[0]); */
460 break;
461 case Opt_err_panic:
5a2b4062
VA
462 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
463 clear_opt (sbi->s_mount_opt, ERRORS_RO);
464 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1da177e4
LT
465 break;
466 case Opt_err_ro:
5a2b4062
VA
467 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
468 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
469 set_opt (sbi->s_mount_opt, ERRORS_RO);
1da177e4
LT
470 break;
471 case Opt_err_cont:
5a2b4062
VA
472 clear_opt (sbi->s_mount_opt, ERRORS_RO);
473 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
474 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1da177e4
LT
475 break;
476 case Opt_nouid32:
477 set_opt (sbi->s_mount_opt, NO_UID32);
478 break;
1da177e4
LT
479 case Opt_nocheck:
480 clear_opt (sbi->s_mount_opt, CHECK);
481 break;
482 case Opt_debug:
483 set_opt (sbi->s_mount_opt, DEBUG);
484 break;
485 case Opt_oldalloc:
486 set_opt (sbi->s_mount_opt, OLDALLOC);
487 break;
488 case Opt_orlov:
489 clear_opt (sbi->s_mount_opt, OLDALLOC);
490 break;
491 case Opt_nobh:
492 set_opt (sbi->s_mount_opt, NOBH);
493 break;
494#ifdef CONFIG_EXT2_FS_XATTR
495 case Opt_user_xattr:
496 set_opt (sbi->s_mount_opt, XATTR_USER);
497 break;
498 case Opt_nouser_xattr:
499 clear_opt (sbi->s_mount_opt, XATTR_USER);
500 break;
501#else
502 case Opt_user_xattr:
503 case Opt_nouser_xattr:
2314b07c
OR
504 ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
505 "not supported");
1da177e4
LT
506 break;
507#endif
508#ifdef CONFIG_EXT2_FS_POSIX_ACL
509 case Opt_acl:
510 set_opt(sbi->s_mount_opt, POSIX_ACL);
511 break;
512 case Opt_noacl:
513 clear_opt(sbi->s_mount_opt, POSIX_ACL);
514 break;
515#else
516 case Opt_acl:
517 case Opt_noacl:
2314b07c
OR
518 ext2_msg(sb, KERN_INFO,
519 "(no)acl options not supported");
1da177e4
LT
520 break;
521#endif
6d79125b
CO
522 case Opt_xip:
523#ifdef CONFIG_EXT2_FS_XIP
524 set_opt (sbi->s_mount_opt, XIP);
525#else
2314b07c 526 ext2_msg(sb, KERN_INFO, "xip option not supported");
6d79125b
CO
527#endif
528 break;
8fc2751b
MB
529
530#if defined(CONFIG_QUOTA)
531 case Opt_quota:
532 case Opt_usrquota:
533 set_opt(sbi->s_mount_opt, USRQUOTA);
534 break;
535
536 case Opt_grpquota:
537 set_opt(sbi->s_mount_opt, GRPQUOTA);
538 break;
539#else
540 case Opt_quota:
541 case Opt_usrquota:
542 case Opt_grpquota:
2314b07c
OR
543 ext2_msg(sb, KERN_INFO,
544 "quota operations not supported");
8fc2751b
MB
545 break;
546#endif
547
a686cd89
MB
548 case Opt_reservation:
549 set_opt(sbi->s_mount_opt, RESERVATION);
2314b07c 550 ext2_msg(sb, KERN_INFO, "reservations ON");
a686cd89
MB
551 break;
552 case Opt_noreservation:
553 clear_opt(sbi->s_mount_opt, RESERVATION);
2314b07c 554 ext2_msg(sb, KERN_INFO, "reservations OFF");
a686cd89 555 break;
1da177e4
LT
556 case Opt_ignore:
557 break;
558 default:
559 return 0;
560 }
561 }
1da177e4
LT
562 return 1;
563}
564
565static int ext2_setup_super (struct super_block * sb,
566 struct ext2_super_block * es,
567 int read_only)
568{
569 int res = 0;
570 struct ext2_sb_info *sbi = EXT2_SB(sb);
571
572 if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
2314b07c
OR
573 ext2_msg(sb, KERN_ERR,
574 "error: revision level too high, "
575 "forcing read-only mode");
1da177e4
LT
576 res = MS_RDONLY;
577 }
578 if (read_only)
579 return res;
580 if (!(sbi->s_mount_state & EXT2_VALID_FS))
2314b07c
OR
581 ext2_msg(sb, KERN_WARNING,
582 "warning: mounting unchecked fs, "
583 "running e2fsck is recommended");
1da177e4 584 else if ((sbi->s_mount_state & EXT2_ERROR_FS))
2314b07c
OR
585 ext2_msg(sb, KERN_WARNING,
586 "warning: mounting fs with errors, "
587 "running e2fsck is recommended");
1da177e4
LT
588 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
589 le16_to_cpu(es->s_mnt_count) >=
590 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2314b07c
OR
591 ext2_msg(sb, KERN_WARNING,
592 "warning: maximal mount count reached, "
593 "running e2fsck is recommended");
1da177e4 594 else if (le32_to_cpu(es->s_checkinterval) &&
2314b07c
OR
595 (le32_to_cpu(es->s_lastcheck) +
596 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
597 ext2_msg(sb, KERN_WARNING,
598 "warning: checktime reached, "
599 "running e2fsck is recommended");
1da177e4
LT
600 if (!le16_to_cpu(es->s_max_mnt_count))
601 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
fba4d399 602 le16_add_cpu(&es->s_mnt_count, 1);
1da177e4 603 if (test_opt (sb, DEBUG))
2314b07c
OR
604 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
605 "bpg=%lu, ipg=%lu, mo=%04lx]",
1da177e4
LT
606 EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
607 sbi->s_frag_size,
608 sbi->s_groups_count,
609 EXT2_BLOCKS_PER_GROUP(sb),
610 EXT2_INODES_PER_GROUP(sb),
611 sbi->s_mount_opt);
1da177e4
LT
612 return res;
613}
614
197cd65a 615static int ext2_check_descriptors(struct super_block *sb)
1da177e4
LT
616{
617 int i;
1da177e4 618 struct ext2_sb_info *sbi = EXT2_SB(sb);
1da177e4
LT
619
620 ext2_debug ("Checking group descriptors");
621
197cd65a
AM
622 for (i = 0; i < sbi->s_groups_count; i++) {
623 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
24097d12
AM
624 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
625 ext2_fsblk_t last_block;
197cd65a 626
41f04d85
ES
627 if (i == sbi->s_groups_count - 1)
628 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
629 else
630 last_block = first_block +
631 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
632
41f04d85
ES
633 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
634 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1da177e4
LT
635 {
636 ext2_error (sb, "ext2_check_descriptors",
637 "Block bitmap for group %d"
638 " not in group (block %lu)!",
639 i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
640 return 0;
641 }
41f04d85
ES
642 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
643 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1da177e4
LT
644 {
645 ext2_error (sb, "ext2_check_descriptors",
646 "Inode bitmap for group %d"
647 " not in group (block %lu)!",
648 i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
649 return 0;
650 }
41f04d85 651 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
780dcdb2 652 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
41f04d85 653 last_block)
1da177e4
LT
654 {
655 ext2_error (sb, "ext2_check_descriptors",
656 "Inode table for group %d"
657 " not in group (block %lu)!",
658 i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
659 return 0;
660 }
1da177e4
LT
661 }
662 return 1;
663}
664
1da177e4
LT
665/*
666 * Maximal file size. There is a direct, and {,double-,triple-}indirect
667 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
668 * We need to be 1 filesystem block less than the 2^32 sector limit.
669 */
670static loff_t ext2_max_size(int bits)
671{
672 loff_t res = EXT2_NDIR_BLOCKS;
902be4c5
AK
673 int meta_blocks;
674 loff_t upper_limit;
675
676 /* This is calculated to be the largest file size for a
677 * dense, file such that the total number of
1da177e4 678 * sectors in the file, including data and all indirect blocks,
902be4c5
AK
679 * does not exceed 2^32 -1
680 * __u32 i_blocks representing the total number of
681 * 512 bytes blocks of the file
682 */
683 upper_limit = (1LL << 32) - 1;
684
685 /* total blocks in file system block size */
686 upper_limit >>= (bits - 9);
687
688
689 /* indirect blocks */
690 meta_blocks = 1;
691 /* double indirect blocks */
692 meta_blocks += 1 + (1LL << (bits-2));
693 /* tripple indirect blocks */
694 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
695
696 upper_limit -= meta_blocks;
697 upper_limit <<= bits;
1da177e4
LT
698
699 res += 1LL << (bits-2);
700 res += 1LL << (2*(bits-2));
701 res += 1LL << (3*(bits-2));
702 res <<= bits;
703 if (res > upper_limit)
704 res = upper_limit;
902be4c5
AK
705
706 if (res > MAX_LFS_FILESIZE)
707 res = MAX_LFS_FILESIZE;
708
1da177e4
LT
709 return res;
710}
711
712static unsigned long descriptor_loc(struct super_block *sb,
713 unsigned long logic_sb_block,
714 int nr)
715{
716 struct ext2_sb_info *sbi = EXT2_SB(sb);
24097d12 717 unsigned long bg, first_meta_bg;
1da177e4
LT
718 int has_super = 0;
719
1da177e4
LT
720 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
721
722 if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
723 nr < first_meta_bg)
724 return (logic_sb_block + nr + 1);
725 bg = sbi->s_desc_per_block * nr;
726 if (ext2_bg_has_super(sb, bg))
727 has_super = 1;
24097d12
AM
728
729 return ext2_group_first_block_no(sb, bg) + has_super;
1da177e4
LT
730}
731
732static int ext2_fill_super(struct super_block *sb, void *data, int silent)
733{
734 struct buffer_head * bh;
735 struct ext2_sb_info * sbi;
736 struct ext2_super_block * es;
737 struct inode *root;
738 unsigned long block;
739 unsigned long sb_block = get_sb_block(&data);
740 unsigned long logic_sb_block;
741 unsigned long offset = 0;
742 unsigned long def_mount_opts;
52fcf703 743 long ret = -EINVAL;
1da177e4
LT
744 int blocksize = BLOCK_SIZE;
745 int db_count;
746 int i, j;
747 __le32 features;
833f4077 748 int err;
1da177e4 749
f8314dc6 750 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1da177e4
LT
751 if (!sbi)
752 return -ENOMEM;
18a82eb9
PE
753
754 sbi->s_blockgroup_lock =
755 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
756 if (!sbi->s_blockgroup_lock) {
757 kfree(sbi);
758 return -ENOMEM;
759 }
1da177e4 760 sb->s_fs_info = sbi;
93d44cb2 761 sbi->s_sb_block = sb_block;
1da177e4 762
c15271f4
JB
763 spin_lock_init(&sbi->s_lock);
764
1da177e4
LT
765 /*
766 * See what the current blocksize for the device is, and
767 * use that as the blocksize. Otherwise (or if the blocksize
768 * is smaller than the default) use the default.
769 * This is important for devices that have a hardware
770 * sectorsize that is larger than the default.
771 */
772 blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
773 if (!blocksize) {
2314b07c 774 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
1da177e4
LT
775 goto failed_sbi;
776 }
777
778 /*
779 * If the superblock doesn't start on a hardware sector boundary,
780 * calculate the offset.
781 */
782 if (blocksize != BLOCK_SIZE) {
783 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
784 offset = (sb_block*BLOCK_SIZE) % blocksize;
785 } else {
786 logic_sb_block = sb_block;
787 }
788
789 if (!(bh = sb_bread(sb, logic_sb_block))) {
2314b07c 790 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
1da177e4
LT
791 goto failed_sbi;
792 }
793 /*
794 * Note: s_es must be initialized as soon as possible because
795 * some ext2 macro-instructions depend on its value
796 */
797 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
798 sbi->s_es = es;
799 sb->s_magic = le16_to_cpu(es->s_magic);
800
801 if (sb->s_magic != EXT2_SUPER_MAGIC)
802 goto cantfind_ext2;
803
804 /* Set defaults before we parse the mount options */
805 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
806 if (def_mount_opts & EXT2_DEFM_DEBUG)
807 set_opt(sbi->s_mount_opt, DEBUG);
808 if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
809 set_opt(sbi->s_mount_opt, GRPID);
810 if (def_mount_opts & EXT2_DEFM_UID16)
811 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 812#ifdef CONFIG_EXT2_FS_XATTR
1da177e4
LT
813 if (def_mount_opts & EXT2_DEFM_XATTR_USER)
814 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD
815#endif
816#ifdef CONFIG_EXT2_FS_POSIX_ACL
1da177e4
LT
817 if (def_mount_opts & EXT2_DEFM_ACL)
818 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 819#endif
1da177e4
LT
820
821 if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
822 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
14e11e10 823 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
5a2b4062 824 set_opt(sbi->s_mount_opt, ERRORS_CONT);
14e11e10
AK
825 else
826 set_opt(sbi->s_mount_opt, ERRORS_RO);
1da177e4
LT
827
828 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
829 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
830
a686cd89
MB
831 set_opt(sbi->s_mount_opt, RESERVATION);
832
2314b07c 833 if (!parse_options((char *) data, sb))
1da177e4
LT
834 goto failed_mount;
835
836 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
837 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
838 MS_POSIXACL : 0);
839
6d79125b
CO
840 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
841 EXT2_MOUNT_XIP if not */
842
1da177e4
LT
843 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
844 (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
845 EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
846 EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2314b07c
OR
847 ext2_msg(sb, KERN_WARNING,
848 "warning: feature flags set on rev 0 fs, "
849 "running e2fsck is recommended");
1da177e4
LT
850 /*
851 * Check feature flags regardless of the revision level, since we
852 * previously didn't change the revision level when setting the flags,
853 * so there is a chance incompat flags are set on a rev 0 filesystem.
854 */
855 features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
856 if (features) {
2314b07c
OR
857 ext2_msg(sb, KERN_ERR, "error: couldn't mount because of "
858 "unsupported optional features (%x)",
859 le32_to_cpu(features));
1da177e4
LT
860 goto failed_mount;
861 }
862 if (!(sb->s_flags & MS_RDONLY) &&
863 (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
2314b07c
OR
864 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
865 "unsupported optional features (%x)",
866 le32_to_cpu(features));
1da177e4
LT
867 goto failed_mount;
868 }
869
870 blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
871
a8e3eff4 872 if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
6d79125b 873 if (!silent)
2314b07c
OR
874 ext2_msg(sb, KERN_ERR,
875 "error: unsupported blocksize for xip");
6d79125b
CO
876 goto failed_mount;
877 }
878
1da177e4
LT
879 /* If the blocksize doesn't match, re-read the thing.. */
880 if (sb->s_blocksize != blocksize) {
881 brelse(bh);
882
883 if (!sb_set_blocksize(sb, blocksize)) {
2314b07c 884 ext2_msg(sb, KERN_ERR, "error: blocksize is too small");
1da177e4
LT
885 goto failed_sbi;
886 }
887
888 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
889 offset = (sb_block*BLOCK_SIZE) % blocksize;
890 bh = sb_bread(sb, logic_sb_block);
891 if(!bh) {
2314b07c
OR
892 ext2_msg(sb, KERN_ERR, "error: couldn't read"
893 "superblock on 2nd try");
1da177e4
LT
894 goto failed_sbi;
895 }
896 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
897 sbi->s_es = es;
898 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
2314b07c 899 ext2_msg(sb, KERN_ERR, "error: magic mismatch");
1da177e4
LT
900 goto failed_mount;
901 }
902 }
903
904 sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
905
906 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
907 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
908 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
909 } else {
910 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
911 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
912 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
d8ea6cf8 913 !is_power_of_2(sbi->s_inode_size) ||
1da177e4 914 (sbi->s_inode_size > blocksize)) {
2314b07c
OR
915 ext2_msg(sb, KERN_ERR,
916 "error: unsupported inode size: %d",
1da177e4
LT
917 sbi->s_inode_size);
918 goto failed_mount;
919 }
920 }
921
922 sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
923 le32_to_cpu(es->s_log_frag_size);
924 if (sbi->s_frag_size == 0)
925 goto cantfind_ext2;
926 sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
927
928 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
929 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
930 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
931
932 if (EXT2_INODE_SIZE(sb) == 0)
933 goto cantfind_ext2;
934 sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
607eb266 935 if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1da177e4
LT
936 goto cantfind_ext2;
937 sbi->s_itb_per_group = sbi->s_inodes_per_group /
938 sbi->s_inodes_per_block;
939 sbi->s_desc_per_block = sb->s_blocksize /
940 sizeof (struct ext2_group_desc);
941 sbi->s_sbh = bh;
942 sbi->s_mount_state = le16_to_cpu(es->s_state);
943 sbi->s_addr_per_block_bits =
f0d1b0b3 944 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1da177e4 945 sbi->s_desc_per_block_bits =
f0d1b0b3 946 ilog2 (EXT2_DESC_PER_BLOCK(sb));
1da177e4
LT
947
948 if (sb->s_magic != EXT2_SUPER_MAGIC)
949 goto cantfind_ext2;
950
951 if (sb->s_blocksize != bh->b_size) {
952 if (!silent)
2314b07c 953 ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1da177e4
LT
954 goto failed_mount;
955 }
956
957 if (sb->s_blocksize != sbi->s_frag_size) {
2314b07c
OR
958 ext2_msg(sb, KERN_ERR,
959 "error: fragsize %lu != blocksize %lu"
960 "(not supported yet)",
1da177e4
LT
961 sbi->s_frag_size, sb->s_blocksize);
962 goto failed_mount;
963 }
964
965 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
2314b07c
OR
966 ext2_msg(sb, KERN_ERR,
967 "error: #blocks per group too big: %lu",
1da177e4
LT
968 sbi->s_blocks_per_group);
969 goto failed_mount;
970 }
971 if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
2314b07c
OR
972 ext2_msg(sb, KERN_ERR,
973 "error: #fragments per group too big: %lu",
1da177e4
LT
974 sbi->s_frags_per_group);
975 goto failed_mount;
976 }
977 if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
2314b07c
OR
978 ext2_msg(sb, KERN_ERR,
979 "error: #inodes per group too big: %lu",
1da177e4
LT
980 sbi->s_inodes_per_group);
981 goto failed_mount;
982 }
983
984 if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
985 goto cantfind_ext2;
41f04d85
ES
986 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
987 le32_to_cpu(es->s_first_data_block) - 1)
988 / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1da177e4
LT
989 db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
990 EXT2_DESC_PER_BLOCK(sb);
991 sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
992 if (sbi->s_group_desc == NULL) {
2314b07c 993 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1da177e4
LT
994 goto failed_mount;
995 }
18a82eb9 996 bgl_lock_init(sbi->s_blockgroup_lock);
dd00cc48 997 sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1da177e4 998 if (!sbi->s_debts) {
2314b07c 999 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1da177e4
LT
1000 goto failed_mount_group_desc;
1001 }
1da177e4
LT
1002 for (i = 0; i < db_count; i++) {
1003 block = descriptor_loc(sb, logic_sb_block, i);
1004 sbi->s_group_desc[i] = sb_bread(sb, block);
1005 if (!sbi->s_group_desc[i]) {
1006 for (j = 0; j < i; j++)
1007 brelse (sbi->s_group_desc[j]);
2314b07c
OR
1008 ext2_msg(sb, KERN_ERR,
1009 "error: unable to read group descriptors");
1da177e4
LT
1010 goto failed_mount_group_desc;
1011 }
1012 }
1013 if (!ext2_check_descriptors (sb)) {
2314b07c 1014 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1da177e4
LT
1015 goto failed_mount2;
1016 }
1017 sbi->s_gdb_count = db_count;
1018 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1019 spin_lock_init(&sbi->s_next_gen_lock);
0216bfcf 1020
a686cd89
MB
1021 /* per fileystem reservation list head & lock */
1022 spin_lock_init(&sbi->s_rsv_window_lock);
1023 sbi->s_rsv_window_root = RB_ROOT;
1024 /*
1025 * Add a single, static dummy reservation to the start of the
1026 * reservation window list --- it gives us a placeholder for
1027 * append-at-start-of-list which makes the allocation logic
1028 * _much_ simpler.
1029 */
1030 sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1031 sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1032 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1033 sbi->s_rsv_window_head.rsv_goal_size = 0;
1034 ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1035
833f4077 1036 err = percpu_counter_init(&sbi->s_freeblocks_counter,
0216bfcf 1037 ext2_count_free_blocks(sb));
833f4077
PZ
1038 if (!err) {
1039 err = percpu_counter_init(&sbi->s_freeinodes_counter,
0216bfcf 1040 ext2_count_free_inodes(sb));
833f4077
PZ
1041 }
1042 if (!err) {
1043 err = percpu_counter_init(&sbi->s_dirs_counter,
0216bfcf 1044 ext2_count_dirs(sb));
833f4077
PZ
1045 }
1046 if (err) {
2314b07c 1047 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
833f4077
PZ
1048 goto failed_mount3;
1049 }
1da177e4
LT
1050 /*
1051 * set up enough so that it can read an inode
1052 */
1053 sb->s_op = &ext2_sops;
1054 sb->s_export_op = &ext2_export_ops;
1055 sb->s_xattr = ext2_xattr_handlers;
123e9caf
CH
1056
1057#ifdef CONFIG_QUOTA
1058 sb->dq_op = &dquot_operations;
287a8095 1059 sb->s_qcop = &dquot_quotactl_ops;
123e9caf
CH
1060#endif
1061
52fcf703
DH
1062 root = ext2_iget(sb, EXT2_ROOT_INO);
1063 if (IS_ERR(root)) {
1064 ret = PTR_ERR(root);
0216bfcf 1065 goto failed_mount3;
1da177e4
LT
1066 }
1067 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
52fcf703 1068 iput(root);
2314b07c 1069 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
0216bfcf 1070 goto failed_mount3;
1da177e4 1071 }
52fcf703
DH
1072
1073 sb->s_root = d_alloc_root(root);
1074 if (!sb->s_root) {
1075 iput(root);
2314b07c 1076 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
52fcf703
DH
1077 ret = -ENOMEM;
1078 goto failed_mount3;
1079 }
1da177e4 1080 if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
2314b07c
OR
1081 ext2_msg(sb, KERN_WARNING,
1082 "warning: mounting ext3 filesystem as ext2");
4c96a68b
JB
1083 if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1084 sb->s_flags |= MS_RDONLY;
1085 ext2_write_super(sb);
1da177e4
LT
1086 return 0;
1087
1088cantfind_ext2:
1089 if (!silent)
2314b07c
OR
1090 ext2_msg(sb, KERN_ERR,
1091 "error: can't find an ext2 filesystem on dev %s.",
1092 sb->s_id);
1da177e4 1093 goto failed_mount;
0216bfcf
MC
1094failed_mount3:
1095 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1096 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1097 percpu_counter_destroy(&sbi->s_dirs_counter);
1da177e4
LT
1098failed_mount2:
1099 for (i = 0; i < db_count; i++)
1100 brelse(sbi->s_group_desc[i]);
1101failed_mount_group_desc:
1102 kfree(sbi->s_group_desc);
1103 kfree(sbi->s_debts);
1104failed_mount:
1105 brelse(bh);
1106failed_sbi:
1107 sb->s_fs_info = NULL;
0f7ee7c1 1108 kfree(sbi->s_blockgroup_lock);
1da177e4 1109 kfree(sbi);
52fcf703 1110 return ret;
1da177e4
LT
1111}
1112
7bf0dc9b
TT
1113static void ext2_clear_super_error(struct super_block *sb)
1114{
1115 struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1116
1117 if (buffer_write_io_error(sbh)) {
1118 /*
1119 * Oh, dear. A previous attempt to write the
1120 * superblock failed. This could happen because the
1121 * USB device was yanked out. Or it could happen to
1122 * be a transient write error and maybe the block will
1123 * be remapped. Nothing we can do but to retry the
1124 * write and hope for the best.
1125 */
2b8120ef
JB
1126 ext2_msg(sb, KERN_ERR,
1127 "previous I/O error to superblock detected\n");
7bf0dc9b
TT
1128 clear_buffer_write_io_error(sbh);
1129 set_buffer_uptodate(sbh);
1130 }
1131}
1132
ee6921eb
JB
1133static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1134 int wait)
1da177e4 1135{
7bf0dc9b 1136 ext2_clear_super_error(sb);
c15271f4 1137 spin_lock(&EXT2_SB(sb)->s_lock);
1da177e4
LT
1138 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1139 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1140 es->s_wtime = cpu_to_le32(get_seconds());
c15271f4
JB
1141 /* unlock before we do IO */
1142 spin_unlock(&EXT2_SB(sb)->s_lock);
1da177e4 1143 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
ee6921eb
JB
1144 if (wait)
1145 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1da177e4
LT
1146 sb->s_dirt = 0;
1147}
1148
1149/*
1150 * In the second extended file system, it is not necessary to
1151 * write the super block since we use a mapping of the
1152 * disk super block in a buffer.
1153 *
1154 * However, this function is still used to set the fs valid
1155 * flags to 0. We need to set this flag to 0 since the fs
1156 * may have been checked while mounted and e2fsck may have
1157 * set s_state to EXT2_VALID_FS after some corrections.
1158 */
40f31dd4 1159static int ext2_sync_fs(struct super_block *sb, int wait)
1da177e4 1160{
c15271f4 1161 struct ext2_sb_info *sbi = EXT2_SB(sb);
40f31dd4
CH
1162 struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1163
c15271f4 1164 spin_lock(&sbi->s_lock);
40f31dd4
CH
1165 if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1166 ext2_debug("setting valid to 0\n");
1167 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1da177e4 1168 }
c15271f4 1169 spin_unlock(&sbi->s_lock);
ee6921eb 1170 ext2_sync_super(sb, es, wait);
40f31dd4
CH
1171 return 0;
1172}
1173
1174
1175void ext2_write_super(struct super_block *sb)
1176{
1177 if (!(sb->s_flags & MS_RDONLY))
1178 ext2_sync_fs(sb, 1);
1179 else
1180 sb->s_dirt = 0;
1da177e4
LT
1181}
1182
1183static int ext2_remount (struct super_block * sb, int * flags, char * data)
1184{
1185 struct ext2_sb_info * sbi = EXT2_SB(sb);
1186 struct ext2_super_block * es;
6d79125b 1187 unsigned long old_mount_opt = sbi->s_mount_opt;
50a52234
JK
1188 struct ext2_mount_options old_opts;
1189 unsigned long old_sb_flags;
1190 int err;
1191
c15271f4 1192 spin_lock(&sbi->s_lock);
337eb00a 1193
50a52234
JK
1194 /* Store the old options */
1195 old_sb_flags = sb->s_flags;
1196 old_opts.s_mount_opt = sbi->s_mount_opt;
1197 old_opts.s_resuid = sbi->s_resuid;
1198 old_opts.s_resgid = sbi->s_resgid;
1da177e4
LT
1199
1200 /*
1201 * Allow the "check" option to be passed as a remount option.
1202 */
2314b07c 1203 if (!parse_options(data, sb)) {
50a52234
JK
1204 err = -EINVAL;
1205 goto restore_opts;
1206 }
1da177e4
LT
1207
1208 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1209 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1210
266f5aa0
CO
1211 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1212 EXT2_MOUNT_XIP if not */
1213
1214 if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
2314b07c
OR
1215 ext2_msg(sb, KERN_WARNING,
1216 "warning: unsupported blocksize for xip");
ddc80bd7 1217 err = -EINVAL;
266f5aa0
CO
1218 goto restore_opts;
1219 }
1220
1da177e4 1221 es = sbi->s_es;
6d79125b
CO
1222 if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1223 (old_mount_opt & EXT2_MOUNT_XIP)) &&
0e4a9b59 1224 invalidate_inodes(sb)) {
2314b07c
OR
1225 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1226 "xip flag with busy inodes while remounting");
0e4a9b59
CO
1227 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1228 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1229 }
337eb00a 1230 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
c15271f4 1231 spin_unlock(&sbi->s_lock);
1da177e4 1232 return 0;
337eb00a 1233 }
1da177e4
LT
1234 if (*flags & MS_RDONLY) {
1235 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
337eb00a 1236 !(sbi->s_mount_state & EXT2_VALID_FS)) {
c15271f4 1237 spin_unlock(&sbi->s_lock);
1da177e4 1238 return 0;
337eb00a 1239 }
c79d967d 1240
1da177e4
LT
1241 /*
1242 * OK, we are remounting a valid rw partition rdonly, so set
1243 * the rdonly flag and then mark the partition as valid again.
1244 */
1245 es->s_state = cpu_to_le16(sbi->s_mount_state);
1246 es->s_mtime = cpu_to_le32(get_seconds());
c15271f4 1247 spin_unlock(&sbi->s_lock);
c79d967d 1248
0f0dd62f
CH
1249 err = dquot_suspend(sb, -1);
1250 if (err < 0) {
c79d967d
CH
1251 spin_lock(&sbi->s_lock);
1252 goto restore_opts;
1253 }
1254
4c96a68b 1255 ext2_sync_super(sb, es, 1);
1da177e4
LT
1256 } else {
1257 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1258 ~EXT2_FEATURE_RO_COMPAT_SUPP);
1259 if (ret) {
2314b07c
OR
1260 ext2_msg(sb, KERN_WARNING,
1261 "warning: couldn't remount RDWR because of "
1262 "unsupported optional features (%x).",
1263 le32_to_cpu(ret));
50a52234
JK
1264 err = -EROFS;
1265 goto restore_opts;
1da177e4
LT
1266 }
1267 /*
1268 * Mounting a RDONLY partition read-write, so reread and
1269 * store the current valid flag. (It may have been changed
1270 * by e2fsck since we originally mounted the partition.)
1271 */
1272 sbi->s_mount_state = le16_to_cpu(es->s_state);
1273 if (!ext2_setup_super (sb, es, 0))
1274 sb->s_flags &= ~MS_RDONLY;
c15271f4 1275 spin_unlock(&sbi->s_lock);
c79d967d 1276
4c96a68b 1277 ext2_write_super(sb);
c79d967d 1278
0f0dd62f 1279 dquot_resume(sb, -1);
1da177e4 1280 }
c79d967d 1281
1da177e4 1282 return 0;
50a52234
JK
1283restore_opts:
1284 sbi->s_mount_opt = old_opts.s_mount_opt;
1285 sbi->s_resuid = old_opts.s_resuid;
1286 sbi->s_resgid = old_opts.s_resgid;
1287 sb->s_flags = old_sb_flags;
c15271f4 1288 spin_unlock(&sbi->s_lock);
50a52234 1289 return err;
1da177e4
LT
1290}
1291
726c3342 1292static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1da177e4 1293{
726c3342 1294 struct super_block *sb = dentry->d_sb;
1da177e4 1295 struct ext2_sb_info *sbi = EXT2_SB(sb);
e4fca01e 1296 struct ext2_super_block *es = sbi->s_es;
e4fca01e 1297 u64 fsid;
1da177e4 1298
c15271f4
JB
1299 spin_lock(&sbi->s_lock);
1300
1da177e4 1301 if (test_opt (sb, MINIX_DF))
2235219b
BP
1302 sbi->s_overhead_last = 0;
1303 else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1304 unsigned long i, overhead = 0;
1305 smp_rmb();
1306
1da177e4 1307 /*
2235219b
BP
1308 * Compute the overhead (FS structures). This is constant
1309 * for a given filesystem unless the number of block groups
1310 * changes so we cache the previous value until it does.
1da177e4
LT
1311 */
1312
1313 /*
1314 * All of the blocks before first_data_block are
1315 * overhead
1316 */
e4fca01e 1317 overhead = le32_to_cpu(es->s_first_data_block);
1da177e4
LT
1318
1319 /*
1320 * Add the overhead attributed to the superblock and
1321 * block group descriptors. If the sparse superblocks
1322 * feature is turned on, then not all groups have this.
1323 */
1324 for (i = 0; i < sbi->s_groups_count; i++)
1325 overhead += ext2_bg_has_super(sb, i) +
1326 ext2_bg_num_gdb(sb, i);
1327
1328 /*
1329 * Every block group has an inode bitmap, a block
1330 * bitmap, and an inode table.
1331 */
1332 overhead += (sbi->s_groups_count *
1333 (2 + sbi->s_itb_per_group));
2235219b
BP
1334 sbi->s_overhead_last = overhead;
1335 smp_wmb();
1336 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1da177e4
LT
1337 }
1338
1339 buf->f_type = EXT2_SUPER_MAGIC;
1340 buf->f_bsize = sb->s_blocksize;
2235219b 1341 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1da177e4 1342 buf->f_bfree = ext2_count_free_blocks(sb);
2235219b 1343 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
e4fca01e
PE
1344 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1345 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1da177e4 1346 buf->f_bavail = 0;
e4fca01e
PE
1347 buf->f_files = le32_to_cpu(es->s_inodes_count);
1348 buf->f_ffree = ext2_count_free_inodes(sb);
2235219b 1349 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1da177e4 1350 buf->f_namelen = EXT2_NAME_LEN;
e4fca01e
PE
1351 fsid = le64_to_cpup((void *)es->s_uuid) ^
1352 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1353 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1354 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
c15271f4 1355 spin_unlock(&sbi->s_lock);
1da177e4
LT
1356 return 0;
1357}
1358
454e2398
DH
1359static int ext2_get_sb(struct file_system_type *fs_type,
1360 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1da177e4 1361{
454e2398 1362 return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1da177e4
LT
1363}
1364
1365#ifdef CONFIG_QUOTA
1366
1367/* Read data from quotafile - avoid pagecache and such because we cannot afford
1368 * acquiring the locks... As quota files are never truncated and quota code
1369 * itself serializes the operations (and noone else should touch the files)
1370 * we don't have to be afraid of races */
1371static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1372 size_t len, loff_t off)
1373{
1374 struct inode *inode = sb_dqopt(sb)->files[type];
1375 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1376 int err = 0;
1377 int offset = off & (sb->s_blocksize - 1);
1378 int tocopy;
1379 size_t toread;
1380 struct buffer_head tmp_bh;
1381 struct buffer_head *bh;
1382 loff_t i_size = i_size_read(inode);
1383
1384 if (off > i_size)
1385 return 0;
1386 if (off+len > i_size)
1387 len = i_size-off;
1388 toread = len;
1389 while (toread > 0) {
1390 tocopy = sb->s_blocksize - offset < toread ?
1391 sb->s_blocksize - offset : toread;
1392
1393 tmp_bh.b_state = 0;
c16831b4 1394 tmp_bh.b_size = sb->s_blocksize;
1da177e4 1395 err = ext2_get_block(inode, blk, &tmp_bh, 0);
a686cd89 1396 if (err < 0)
1da177e4
LT
1397 return err;
1398 if (!buffer_mapped(&tmp_bh)) /* A hole? */
1399 memset(data, 0, tocopy);
1400 else {
1401 bh = sb_bread(sb, tmp_bh.b_blocknr);
1402 if (!bh)
1403 return -EIO;
1404 memcpy(data, bh->b_data+offset, tocopy);
1405 brelse(bh);
1406 }
1407 offset = 0;
1408 toread -= tocopy;
1409 data += tocopy;
1410 blk++;
1411 }
1412 return len;
1413}
1414
1415/* Write to quotafile */
1416static ssize_t ext2_quota_write(struct super_block *sb, int type,
1417 const char *data, size_t len, loff_t off)
1418{
1419 struct inode *inode = sb_dqopt(sb)->files[type];
1420 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1421 int err = 0;
1422 int offset = off & (sb->s_blocksize - 1);
1423 int tocopy;
1424 size_t towrite = len;
1425 struct buffer_head tmp_bh;
1426 struct buffer_head *bh;
1427
5c81a419 1428 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1da177e4
LT
1429 while (towrite > 0) {
1430 tocopy = sb->s_blocksize - offset < towrite ?
1431 sb->s_blocksize - offset : towrite;
1432
1433 tmp_bh.b_state = 0;
1434 err = ext2_get_block(inode, blk, &tmp_bh, 1);
a686cd89 1435 if (err < 0)
1da177e4
LT
1436 goto out;
1437 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1438 bh = sb_bread(sb, tmp_bh.b_blocknr);
1439 else
1440 bh = sb_getblk(sb, tmp_bh.b_blocknr);
1441 if (!bh) {
1442 err = -EIO;
1443 goto out;
1444 }
1445 lock_buffer(bh);
1446 memcpy(bh->b_data+offset, data, tocopy);
1447 flush_dcache_page(bh->b_page);
1448 set_buffer_uptodate(bh);
1449 mark_buffer_dirty(bh);
1450 unlock_buffer(bh);
1451 brelse(bh);
1452 offset = 0;
1453 towrite -= tocopy;
1454 data += tocopy;
1455 blk++;
1456 }
1457out:
a069e9ce
DC
1458 if (len == towrite) {
1459 mutex_unlock(&inode->i_mutex);
1da177e4 1460 return err;
a069e9ce 1461 }
1da177e4
LT
1462 if (inode->i_size < off+len-towrite)
1463 i_size_write(inode, off+len-towrite);
1464 inode->i_version++;
1465 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1466 mark_inode_dirty(inode);
1b1dcc1b 1467 mutex_unlock(&inode->i_mutex);
1da177e4
LT
1468 return len - towrite;
1469}
1470
1471#endif
1472
1473static struct file_system_type ext2_fs_type = {
1474 .owner = THIS_MODULE,
1475 .name = "ext2",
1476 .get_sb = ext2_get_sb,
1477 .kill_sb = kill_block_super,
1478 .fs_flags = FS_REQUIRES_DEV,
1479};
1480
1481static int __init init_ext2_fs(void)
1482{
1483 int err = init_ext2_xattr();
1484 if (err)
1485 return err;
1486 err = init_inodecache();
1487 if (err)
1488 goto out1;
1489 err = register_filesystem(&ext2_fs_type);
1490 if (err)
1491 goto out;
1492 return 0;
1493out:
1494 destroy_inodecache();
1495out1:
1496 exit_ext2_xattr();
1497 return err;
1498}
1499
1500static void __exit exit_ext2_fs(void)
1501{
1502 unregister_filesystem(&ext2_fs_type);
1503 destroy_inodecache();
1504 exit_ext2_xattr();
1505}
1506
1507module_init(init_ext2_fs)
1508module_exit(exit_ext2_fs)