projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
[PATCH] struct path: convert zorro
[linux-2.6-block.git] / fs / ext3 / super.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/ext3/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>
21#include <linux/fs.h>
22#include <linux/time.h>
23#include <linux/jbd.h>
24#include <linux/ext3_fs.h>
25#include <linux/ext3_jbd.h>
26#include <linux/slab.h>
27#include <linux/init.h>
28#include <linux/blkdev.h>
29#include <linux/parser.h>
30#include <linux/smp_lock.h>
31#include <linux/buffer_head.h>
32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
8fc2751b 37#include <linux/seq_file.h>
381be254 38
1da177e4 39#include <asm/uaccess.h>
381be254 40
1da177e4
LT		 41#include "xattr.h"
42#include "acl.h"
381be254 43#include "namei.h"
1da177e4 44
71b96257
JL		 45static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
46 unsigned long journal_devnum);
1da177e4 47static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
eee194e7 48 unsigned int);
1da177e4
LT		 49static void ext3_commit_super (struct super_block * sb,
50 struct ext3_super_block * es,
51 int sync);
52static void ext3_mark_recovery_complete(struct super_block * sb,
53 struct ext3_super_block * es);
54static void ext3_clear_journal_err(struct super_block * sb,
55 struct ext3_super_block * es);
56static int ext3_sync_fs(struct super_block *sb, int wait);
57static const char *ext3_decode_error(struct super_block * sb, int errno,
58 char nbuf[16]);
59static int ext3_remount (struct super_block * sb, int * flags, char * data);
726c3342 60static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
1da177e4
LT		 61static void ext3_unlockfs(struct super_block *sb);
62static void ext3_write_super (struct super_block * sb);
63static void ext3_write_super_lockfs(struct super_block *sb);
64
ae6ddcc5 65/*
1da177e4
LT		 66 * Wrappers for journal_start/end.
67 *
68 * The only special thing we need to do here is to make sure that all
69 * journal_end calls result in the superblock being marked dirty, so
70 * that sync() will call the filesystem's write_super callback if
ae6ddcc5 71 * appropriate.
1da177e4
LT		 72 */
73handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
74{
75 journal_t *journal;
76
77 if (sb->s_flags & MS_RDONLY)
78 return ERR_PTR(-EROFS);
79
80 /* Special case here: if the journal has aborted behind our
81 * backs (eg. EIO in the commit thread), then we still need to
82 * take the FS itself readonly cleanly. */
83 journal = EXT3_SB(sb)->s_journal;
84 if (is_journal_aborted(journal)) {
85 ext3_abort(sb, __FUNCTION__,
86 "Detected aborted journal");
87 return ERR_PTR(-EROFS);
88 }
89
90 return journal_start(journal, nblocks);
91}
92
ae6ddcc5 93/*
1da177e4
LT		 94 * The only special thing we need to do here is to make sure that all
95 * journal_stop calls result in the superblock being marked dirty, so
96 * that sync() will call the filesystem's write_super callback if
ae6ddcc5 97 * appropriate.
1da177e4
LT		 98 */
99int __ext3_journal_stop(const char *where, handle_t *handle)
100{
101 struct super_block *sb;
102 int err;
103 int rc;
104
105 sb = handle->h_transaction->t_journal->j_private;
106 err = handle->h_err;
107 rc = journal_stop(handle);
108
109 if (!err)
110 err = rc;
111 if (err)
112 __ext3_std_error(sb, where, err);
113 return err;
114}
115
116void ext3_journal_abort_handle(const char *caller, const char *err_fn,
117 struct buffer_head *bh, handle_t *handle, int err)
118{
119 char nbuf[16];
120 const char *errstr = ext3_decode_error(NULL, err, nbuf);
121
122 if (bh)
123 BUFFER_TRACE(bh, "abort");
124
125 if (!handle->h_err)
126 handle->h_err = err;
127
128 if (is_handle_aborted(handle))
129 return;
130
131 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
132 caller, errstr, err_fn);
133
134 journal_abort_handle(handle);
135}
136
137/* Deal with the reporting of failure conditions on a filesystem such as
138 * inconsistencies detected or read IO failures.
139 *
140 * On ext2, we can store the error state of the filesystem in the
141 * superblock. That is not possible on ext3, because we may have other
142 * write ordering constraints on the superblock which prevent us from
143 * writing it out straight away; and given that the journal is about to
144 * be aborted, we can't rely on the current, or future, transactions to
145 * write out the superblock safely.
146 *
147 * We'll just use the journal_abort() error code to record an error in
148 * the journal instead. On recovery, the journal will compain about
149 * that error until we've noted it down and cleared it.
150 */
151
152static void ext3_handle_error(struct super_block *sb)
153{
154 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
155
156 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
157 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
158
159 if (sb->s_flags & MS_RDONLY)
160 return;
161
7543fc7b 162 if (!test_opt (sb, ERRORS_CONT)) {
1da177e4
LT		 163 journal_t *journal = EXT3_SB(sb)->s_journal;
164
165 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
166 if (journal)
167 journal_abort(journal, -EIO);
168 }
7543fc7b
VA		 169 if (test_opt (sb, ERRORS_RO)) {
170 printk (KERN_CRIT "Remounting filesystem read-only\n");
171 sb->s_flags |= MS_RDONLY;
172 }
173 ext3_commit_super(sb, es, 1);
1da177e4
LT		 174 if (test_opt(sb, ERRORS_PANIC))
175 panic("EXT3-fs (device %s): panic forced after error\n",
176 sb->s_id);
1da177e4
LT		 177}
178
179void ext3_error (struct super_block * sb, const char * function,
180 const char * fmt, ...)
181{
182 va_list args;
183
184 va_start(args, fmt);
185 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
186 vprintk(fmt, args);
187 printk("\n");
188 va_end(args);
189
190 ext3_handle_error(sb);
191}
192
193static const char *ext3_decode_error(struct super_block * sb, int errno,
194 char nbuf[16])
195{
196 char *errstr = NULL;
197
198 switch (errno) {
199 case -EIO:
200 errstr = "IO failure";
201 break;
202 case -ENOMEM:
203 errstr = "Out of memory";
204 break;
205 case -EROFS:
206 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
207 errstr = "Journal has aborted";
208 else
209 errstr = "Readonly filesystem";
210 break;
211 default:
212 /* If the caller passed in an extra buffer for unknown
213 * errors, textualise them now. Else we just return
214 * NULL. */
215 if (nbuf) {
216 /* Check for truncated error codes... */
217 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
218 errstr = nbuf;
219 }
220 break;
221 }
222
223 return errstr;
224}
225
226/* __ext3_std_error decodes expected errors from journaling functions
227 * automatically and invokes the appropriate error response. */
228
229void __ext3_std_error (struct super_block * sb, const char * function,
230 int errno)
231{
232 char nbuf[16];
30121624 233 const char *errstr;
1da177e4 234
30121624
ST		 235 /* Special case: if the error is EROFS, and we're not already
236 * inside a transaction, then there's really no point in logging
237 * an error. */
238 if (errno == -EROFS && journal_current_handle() == NULL &&
239 (sb->s_flags & MS_RDONLY))
240 return;
241
242 errstr = ext3_decode_error(sb, errno, nbuf);
1da177e4
LT		 243 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
244 sb->s_id, function, errstr);
245
246 ext3_handle_error(sb);
247}
248
249/*
250 * ext3_abort is a much stronger failure handler than ext3_error. The
251 * abort function may be used to deal with unrecoverable failures such
252 * as journal IO errors or ENOMEM at a critical moment in log management.
253 *
254 * We unconditionally force the filesystem into an ABORT|READONLY state,
255 * unless the error response on the fs has been set to panic in which
256 * case we take the easy way out and panic immediately.
257 */
258
259void ext3_abort (struct super_block * sb, const char * function,
260 const char * fmt, ...)
261{
262 va_list args;
263
264 printk (KERN_CRIT "ext3_abort called.\n");
265
266 va_start(args, fmt);
267 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
268 vprintk(fmt, args);
269 printk("\n");
270 va_end(args);
271
272 if (test_opt(sb, ERRORS_PANIC))
273 panic("EXT3-fs panic from previous error\n");
274
275 if (sb->s_flags & MS_RDONLY)
276 return;
277
278 printk(KERN_CRIT "Remounting filesystem read-only\n");
279 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
280 sb->s_flags |= MS_RDONLY;
281 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
282 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
283}
284
285void ext3_warning (struct super_block * sb, const char * function,
286 const char * fmt, ...)
287{
288 va_list args;
289
290 va_start(args, fmt);
291 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
292 sb->s_id, function);
293 vprintk(fmt, args);
294 printk("\n");
295 va_end(args);
296}
297
298void ext3_update_dynamic_rev(struct super_block *sb)
299{
300 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
301
302 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
303 return;
304
305 ext3_warning(sb, __FUNCTION__,
306 "updating to rev %d because of new feature flag, "
307 "running e2fsck is recommended",
308 EXT3_DYNAMIC_REV);
309
310 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
311 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
312 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
313 /* leave es->s_feature_*compat flags alone */
314 /* es->s_uuid will be set by e2fsck if empty */
315
316 /*
317 * The rest of the superblock fields should be zero, and if not it
318 * means they are likely already in use, so leave them alone. We
319 * can leave it up to e2fsck to clean up any inconsistencies there.
320 */
321}
322
323/*
324 * Open the external journal device
325 */
326static struct block_device *ext3_blkdev_get(dev_t dev)
327{
328 struct block_device *bdev;
329 char b[BDEVNAME_SIZE];
330
331 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
332 if (IS_ERR(bdev))
333 goto fail;
334 return bdev;
335
336fail:
337 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
338 __bdevname(dev, b), PTR_ERR(bdev));
339 return NULL;
340}
341
342/*
343 * Release the journal device
344 */
345static int ext3_blkdev_put(struct block_device *bdev)
346{
347 bd_release(bdev);
348 return blkdev_put(bdev);
349}
350
351static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
352{
353 struct block_device *bdev;
354 int ret = -ENODEV;
355
356 bdev = sbi->journal_bdev;
357 if (bdev) {
358 ret = ext3_blkdev_put(bdev);
359 sbi->journal_bdev = NULL;
360 }
361 return ret;
362}
363
364static inline struct inode *orphan_list_entry(struct list_head *l)
365{
366 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
367}
368
369static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
370{
371 struct list_head *l;
372
ae6ddcc5 373 printk(KERN_ERR "sb orphan head is %d\n",
1da177e4
LT		 374 le32_to_cpu(sbi->s_es->s_last_orphan));
375
376 printk(KERN_ERR "sb_info orphan list:\n");
377 list_for_each(l, &sbi->s_orphan) {
378 struct inode *inode = orphan_list_entry(l);
379 printk(KERN_ERR " "
eee194e7 380 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1da177e4 381 inode->i_sb->s_id, inode->i_ino, inode,
ae6ddcc5 382 inode->i_mode, inode->i_nlink,
1da177e4
LT		 383 NEXT_ORPHAN(inode));
384 }
385}
386
387static void ext3_put_super (struct super_block * sb)
388{
389 struct ext3_sb_info *sbi = EXT3_SB(sb);
390 struct ext3_super_block *es = sbi->s_es;
391 int i;
392
393 ext3_xattr_put_super(sb);
394 journal_destroy(sbi->s_journal);
395 if (!(sb->s_flags & MS_RDONLY)) {
396 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
397 es->s_state = cpu_to_le16(sbi->s_mount_state);
398 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
399 mark_buffer_dirty(sbi->s_sbh);
400 ext3_commit_super(sb, es, 1);
401 }
402
403 for (i = 0; i < sbi->s_gdb_count; i++)
404 brelse(sbi->s_group_desc[i]);
405 kfree(sbi->s_group_desc);
406 percpu_counter_destroy(&sbi->s_freeblocks_counter);
407 percpu_counter_destroy(&sbi->s_freeinodes_counter);
408 percpu_counter_destroy(&sbi->s_dirs_counter);
409 brelse(sbi->s_sbh);
410#ifdef CONFIG_QUOTA
411 for (i = 0; i < MAXQUOTAS; i++)
412 kfree(sbi->s_qf_names[i]);
413#endif
414
415 /* Debugging code just in case the in-memory inode orphan list
416 * isn't empty. The on-disk one can be non-empty if we've
417 * detected an error and taken the fs readonly, but the
418 * in-memory list had better be clean by this point. */
419 if (!list_empty(&sbi->s_orphan))
420 dump_orphan_list(sb, sbi);
421 J_ASSERT(list_empty(&sbi->s_orphan));
422
423 invalidate_bdev(sb->s_bdev, 0);
424 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
425 /*
426 * Invalidate the journal device's buffers. We don't want them
427 * floating about in memory - the physical journal device may
428 * hotswapped, and it breaks the `ro-after' testing code.
429 */
430 sync_blockdev(sbi->journal_bdev);
431 invalidate_bdev(sbi->journal_bdev, 0);
432 ext3_blkdev_remove(sbi);
433 }
434 sb->s_fs_info = NULL;
435 kfree(sbi);
436 return;
437}
438
e18b890b 439static struct kmem_cache *ext3_inode_cachep;
1da177e4
LT		 440
441/*
442 * Called inside transaction, so use GFP_NOFS
443 */
444static struct inode *ext3_alloc_inode(struct super_block *sb)
445{
446 struct ext3_inode_info *ei;
447
e6b4f8da 448 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
1da177e4
LT		 449 if (!ei)
450 return NULL;
451#ifdef CONFIG_EXT3_FS_POSIX_ACL
452 ei->i_acl = EXT3_ACL_NOT_CACHED;
453 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
454#endif
455 ei->i_block_alloc_info = NULL;
456 ei->vfs_inode.i_version = 1;
457 return &ei->vfs_inode;
458}
459
460static void ext3_destroy_inode(struct inode *inode)
461{
462 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
463}
464
e18b890b 465static void init_once(void * foo, struct kmem_cache * cachep, unsigned long flags)
1da177e4
LT		 466{
467 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
468
469 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
470 SLAB_CTOR_CONSTRUCTOR) {
471 INIT_LIST_HEAD(&ei->i_orphan);
472#ifdef CONFIG_EXT3_FS_XATTR
473 init_rwsem(&ei->xattr_sem);
474#endif
97461518 475 mutex_init(&ei->truncate_mutex);
1da177e4
LT		 476 inode_init_once(&ei->vfs_inode);
477 }
478}
ae6ddcc5 479
1da177e4
LT		 480static int init_inodecache(void)
481{
482 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
483 sizeof(struct ext3_inode_info),
fffb60f9
PJ		 484 0, (SLAB_RECLAIM_ACCOUNT|
485 SLAB_MEM_SPREAD),
1da177e4
LT		 486 init_once, NULL);
487 if (ext3_inode_cachep == NULL)
488 return -ENOMEM;
489 return 0;
490}
491
492static void destroy_inodecache(void)
493{
1a1d92c1 494 kmem_cache_destroy(ext3_inode_cachep);
1da177e4
LT		 495}
496
497static void ext3_clear_inode(struct inode *inode)
498{
499 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
500#ifdef CONFIG_EXT3_FS_POSIX_ACL
e6022603
AM		 501 if (EXT3_I(inode)->i_acl &&
502 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
503 posix_acl_release(EXT3_I(inode)->i_acl);
504 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
505 }
506 if (EXT3_I(inode)->i_default_acl &&
507 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
508 posix_acl_release(EXT3_I(inode)->i_default_acl);
509 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
510 }
1da177e4
LT		 511#endif
512 ext3_discard_reservation(inode);
513 EXT3_I(inode)->i_block_alloc_info = NULL;
e6022603
AM		 514 if (unlikely(rsv))
515 kfree(rsv);
1da177e4
LT		 516}
517
6cd37cda 518static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
8fc2751b 519{
6cd37cda 520#if defined(CONFIG_QUOTA)
275abf5b 521 struct ext3_sb_info *sbi = EXT3_SB(sb);
8fc2751b 522
8fc2751b
MB		 523 if (sbi->s_jquota_fmt)
524 seq_printf(seq, ",jqfmt=%s",
525 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
526
527 if (sbi->s_qf_names[USRQUOTA])
528 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
529
530 if (sbi->s_qf_names[GRPQUOTA])
531 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
532
533 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
534 seq_puts(seq, ",usrquota");
535
536 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
537 seq_puts(seq, ",grpquota");
538#endif
6cd37cda
PO		 539}
540
541static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
542{
543 struct super_block *sb = vfs->mnt_sb;
544
545 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
546 seq_puts(seq, ",data=journal");
547 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
548 seq_puts(seq, ",data=ordered");
549 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
550 seq_puts(seq, ",data=writeback");
551
552 ext3_show_quota_options(seq, sb);
8fc2751b
MB		 553
554 return 0;
555}
1da177e4 556
fdb36673
N		 557
558static struct dentry *ext3_get_dentry(struct super_block *sb, void *vobjp)
559{
560 __u32 *objp = vobjp;
561 unsigned long ino = objp[0];
562 __u32 generation = objp[1];
563 struct inode *inode;
564 struct dentry *result;
565
566 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
567 return ERR_PTR(-ESTALE);
568 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
569 return ERR_PTR(-ESTALE);
570
571 /* iget isn't really right if the inode is currently unallocated!!
572 *
573 * ext3_read_inode will return a bad_inode if the inode had been
574 * deleted, so we should be safe.
575 *
576 * Currently we don't know the generation for parent directory, so
577 * a generation of 0 means "accept any"
578 */
579 inode = iget(sb, ino);
580 if (inode == NULL)
581 return ERR_PTR(-ENOMEM);
582 if (is_bad_inode(inode) ||
583 (generation && inode->i_generation != generation)) {
584 iput(inode);
585 return ERR_PTR(-ESTALE);
586 }
587 /* now to find a dentry.
588 * If possible, get a well-connected one
589 */
590 result = d_alloc_anon(inode);
591 if (!result) {
592 iput(inode);
593 return ERR_PTR(-ENOMEM);
594 }
595 return result;
596}
597
8fc2751b 598#ifdef CONFIG_QUOTA
1da177e4
LT		 599#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
600#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
601
602static int ext3_dquot_initialize(struct inode *inode, int type);
603static int ext3_dquot_drop(struct inode *inode);
604static int ext3_write_dquot(struct dquot *dquot);
605static int ext3_acquire_dquot(struct dquot *dquot);
606static int ext3_release_dquot(struct dquot *dquot);
607static int ext3_mark_dquot_dirty(struct dquot *dquot);
608static int ext3_write_info(struct super_block *sb, int type);
609static int ext3_quota_on(struct super_block *sb, int type, int format_id, char *path);
610static int ext3_quota_on_mount(struct super_block *sb, int type);
611static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
612 size_t len, loff_t off);
613static ssize_t ext3_quota_write(struct super_block *sb, int type,
614 const char *data, size_t len, loff_t off);
615
616static struct dquot_operations ext3_quota_operations = {
617 .initialize = ext3_dquot_initialize,
618 .drop = ext3_dquot_drop,
619 .alloc_space = dquot_alloc_space,
620 .alloc_inode = dquot_alloc_inode,
621 .free_space = dquot_free_space,
622 .free_inode = dquot_free_inode,
623 .transfer = dquot_transfer,
624 .write_dquot = ext3_write_dquot,
625 .acquire_dquot = ext3_acquire_dquot,
626 .release_dquot = ext3_release_dquot,
627 .mark_dirty = ext3_mark_dquot_dirty,
628 .write_info = ext3_write_info
629};
630
631static struct quotactl_ops ext3_qctl_operations = {
632 .quota_on = ext3_quota_on,
633 .quota_off = vfs_quota_off,
634 .quota_sync = vfs_quota_sync,
635 .get_info = vfs_get_dqinfo,
636 .set_info = vfs_set_dqinfo,
637 .get_dqblk = vfs_get_dqblk,
638 .set_dqblk = vfs_set_dqblk
639};
640#endif
641
642static struct super_operations ext3_sops = {
643 .alloc_inode = ext3_alloc_inode,
644 .destroy_inode = ext3_destroy_inode,
645 .read_inode = ext3_read_inode,
646 .write_inode = ext3_write_inode,
647 .dirty_inode = ext3_dirty_inode,
648 .delete_inode = ext3_delete_inode,
649 .put_super = ext3_put_super,
650 .write_super = ext3_write_super,
651 .sync_fs = ext3_sync_fs,
652 .write_super_lockfs = ext3_write_super_lockfs,
653 .unlockfs = ext3_unlockfs,
654 .statfs = ext3_statfs,
655 .remount_fs = ext3_remount,
656 .clear_inode = ext3_clear_inode,
8fc2751b 657 .show_options = ext3_show_options,
1da177e4
LT		 658#ifdef CONFIG_QUOTA
659 .quota_read = ext3_quota_read,
660 .quota_write = ext3_quota_write,
661#endif
662};
663
1da177e4
LT		 664static struct export_operations ext3_export_ops = {
665 .get_parent = ext3_get_parent,
fdb36673 666 .get_dentry = ext3_get_dentry,
1da177e4
LT		 667};
668
669enum {
670 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
671 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
2860b733 672 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
1da177e4 673 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
ade1a29e 674 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
71b96257 675 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
1da177e4
LT		 676 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
677 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1f54587b 678 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
8fc2751b
MB		 679 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
680 Opt_grpquota
1da177e4
LT		 681};
682
683static match_table_t tokens = {
684 {Opt_bsd_df, "bsddf"},
685 {Opt_minix_df, "minixdf"},
686 {Opt_grpid, "grpid"},
687 {Opt_grpid, "bsdgroups"},
688 {Opt_nogrpid, "nogrpid"},
689 {Opt_nogrpid, "sysvgroups"},
690 {Opt_resgid, "resgid=%u"},
691 {Opt_resuid, "resuid=%u"},
692 {Opt_sb, "sb=%u"},
693 {Opt_err_cont, "errors=continue"},
694 {Opt_err_panic, "errors=panic"},
695 {Opt_err_ro, "errors=remount-ro"},
696 {Opt_nouid32, "nouid32"},
697 {Opt_nocheck, "nocheck"},
698 {Opt_nocheck, "check=none"},
1da177e4
LT		 699 {Opt_debug, "debug"},
700 {Opt_oldalloc, "oldalloc"},
701 {Opt_orlov, "orlov"},
702 {Opt_user_xattr, "user_xattr"},
703 {Opt_nouser_xattr, "nouser_xattr"},
704 {Opt_acl, "acl"},
705 {Opt_noacl, "noacl"},
706 {Opt_reservation, "reservation"},
707 {Opt_noreservation, "noreservation"},
708 {Opt_noload, "noload"},
709 {Opt_nobh, "nobh"},
ade1a29e 710 {Opt_bh, "bh"},
1da177e4
LT		 711 {Opt_commit, "commit=%u"},
712 {Opt_journal_update, "journal=update"},
713 {Opt_journal_inum, "journal=%u"},
71b96257 714 {Opt_journal_dev, "journal_dev=%u"},
1da177e4
LT		 715 {Opt_abort, "abort"},
716 {Opt_data_journal, "data=journal"},
717 {Opt_data_ordered, "data=ordered"},
718 {Opt_data_writeback, "data=writeback"},
719 {Opt_offusrjquota, "usrjquota="},
720 {Opt_usrjquota, "usrjquota=%s"},
721 {Opt_offgrpjquota, "grpjquota="},
722 {Opt_grpjquota, "grpjquota=%s"},
723 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
724 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
8fc2751b 725 {Opt_grpquota, "grpquota"},
1f54587b
JK		 726 {Opt_noquota, "noquota"},
727 {Opt_quota, "quota"},
8fc2751b 728 {Opt_usrquota, "usrquota"},
1da177e4
LT		 729 {Opt_barrier, "barrier=%u"},
730 {Opt_err, NULL},
731 {Opt_resize, "resize"},
732};
733
43d23f90 734static ext3_fsblk_t get_sb_block(void **data)
1da177e4 735{
e9ad5620
DK		 736 ext3_fsblk_t sb_block;
737 char *options = (char *) *data;
1da177e4
LT		 738
739 if (!options || strncmp(options, "sb=", 3) != 0)
740 return 1; /* Default location */
741 options += 3;
43d23f90 742 /*todo: use simple_strtoll with >32bit ext3 */
1da177e4
LT		 743 sb_block = simple_strtoul(options, &options, 0);
744 if (*options && *options != ',') {
745 printk("EXT3-fs: Invalid sb specification: %s\n",
746 (char *) *data);
747 return 1;
748 }
749 if (*options == ',')
750 options++;
751 *data = (void *) options;
752 return sb_block;
753}
754
71b96257 755static int parse_options (char *options, struct super_block *sb,
eee194e7 756 unsigned int *inum, unsigned long *journal_devnum,
43d23f90 757 ext3_fsblk_t *n_blocks_count, int is_remount)
1da177e4
LT		 758{
759 struct ext3_sb_info *sbi = EXT3_SB(sb);
760 char * p;
761 substring_t args[MAX_OPT_ARGS];
762 int data_opt = 0;
763 int option;
764#ifdef CONFIG_QUOTA
765 int qtype;
766 char *qname;
767#endif
768
769 if (!options)
770 return 1;
771
772 while ((p = strsep (&options, ",")) != NULL) {
773 int token;
774 if (!*p)
775 continue;
776
777 token = match_token(p, tokens, args);
778 switch (token) {
779 case Opt_bsd_df:
780 clear_opt (sbi->s_mount_opt, MINIX_DF);
781 break;
782 case Opt_minix_df:
783 set_opt (sbi->s_mount_opt, MINIX_DF);
784 break;
785 case Opt_grpid:
786 set_opt (sbi->s_mount_opt, GRPID);
787 break;
788 case Opt_nogrpid:
789 clear_opt (sbi->s_mount_opt, GRPID);
790 break;
791 case Opt_resuid:
792 if (match_int(&args[0], &option))
793 return 0;
794 sbi->s_resuid = option;
795 break;
796 case Opt_resgid:
797 if (match_int(&args[0], &option))
798 return 0;
799 sbi->s_resgid = option;
800 break;
801 case Opt_sb:
802 /* handled by get_sb_block() instead of here */
803 /* *sb_block = match_int(&args[0]); */
804 break;
805 case Opt_err_panic:
806 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
807 clear_opt (sbi->s_mount_opt, ERRORS_RO);
808 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
809 break;
810 case Opt_err_ro:
811 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
812 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
813 set_opt (sbi->s_mount_opt, ERRORS_RO);
814 break;
815 case Opt_err_cont:
816 clear_opt (sbi->s_mount_opt, ERRORS_RO);
817 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
818 set_opt (sbi->s_mount_opt, ERRORS_CONT);
819 break;
820 case Opt_nouid32:
821 set_opt (sbi->s_mount_opt, NO_UID32);
822 break;
1da177e4
LT		 823 case Opt_nocheck:
824 clear_opt (sbi->s_mount_opt, CHECK);
825 break;
826 case Opt_debug:
827 set_opt (sbi->s_mount_opt, DEBUG);
828 break;
829 case Opt_oldalloc:
830 set_opt (sbi->s_mount_opt, OLDALLOC);
831 break;
832 case Opt_orlov:
833 clear_opt (sbi->s_mount_opt, OLDALLOC);
834 break;
835#ifdef CONFIG_EXT3_FS_XATTR
836 case Opt_user_xattr:
837 set_opt (sbi->s_mount_opt, XATTR_USER);
838 break;
839 case Opt_nouser_xattr:
840 clear_opt (sbi->s_mount_opt, XATTR_USER);
841 break;
842#else
843 case Opt_user_xattr:
844 case Opt_nouser_xattr:
845 printk("EXT3 (no)user_xattr options not supported\n");
846 break;
847#endif
848#ifdef CONFIG_EXT3_FS_POSIX_ACL
849 case Opt_acl:
850 set_opt(sbi->s_mount_opt, POSIX_ACL);
851 break;
852 case Opt_noacl:
853 clear_opt(sbi->s_mount_opt, POSIX_ACL);
854 break;
855#else
856 case Opt_acl:
857 case Opt_noacl:
858 printk("EXT3 (no)acl options not supported\n");
859 break;
860#endif
861 case Opt_reservation:
862 set_opt(sbi->s_mount_opt, RESERVATION);
863 break;
864 case Opt_noreservation:
865 clear_opt(sbi->s_mount_opt, RESERVATION);
866 break;
867 case Opt_journal_update:
868 /* @@@ FIXME */
869 /* Eventually we will want to be able to create
870 a journal file here. For now, only allow the
871 user to specify an existing inode to be the
872 journal file. */
873 if (is_remount) {
874 printk(KERN_ERR "EXT3-fs: cannot specify "
875 "journal on remount\n");
876 return 0;
877 }
878 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
879 break;
880 case Opt_journal_inum:
881 if (is_remount) {
882 printk(KERN_ERR "EXT3-fs: cannot specify "
883 "journal on remount\n");
884 return 0;
885 }
886 if (match_int(&args[0], &option))
887 return 0;
888 *inum = option;
889 break;
71b96257
JL		 890 case Opt_journal_dev:
891 if (is_remount) {
892 printk(KERN_ERR "EXT3-fs: cannot specify "
893 "journal on remount\n");
894 return 0;
895 }
896 if (match_int(&args[0], &option))
897 return 0;
898 *journal_devnum = option;
899 break;
1da177e4
LT		 900 case Opt_noload:
901 set_opt (sbi->s_mount_opt, NOLOAD);
902 break;
903 case Opt_commit:
904 if (match_int(&args[0], &option))
905 return 0;
906 if (option < 0)
907 return 0;
908 if (option == 0)
909 option = JBD_DEFAULT_MAX_COMMIT_AGE;
910 sbi->s_commit_interval = HZ * option;
911 break;
912 case Opt_data_journal:
913 data_opt = EXT3_MOUNT_JOURNAL_DATA;
914 goto datacheck;
915 case Opt_data_ordered:
916 data_opt = EXT3_MOUNT_ORDERED_DATA;
917 goto datacheck;
918 case Opt_data_writeback:
919 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
920 datacheck:
921 if (is_remount) {
922 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
923 != data_opt) {
924 printk(KERN_ERR
925 "EXT3-fs: cannot change data "
926 "mode on remount\n");
927 return 0;
928 }
929 } else {
930 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
931 sbi->s_mount_opt |= data_opt;
932 }
933 break;
934#ifdef CONFIG_QUOTA
935 case Opt_usrjquota:
936 qtype = USRQUOTA;
937 goto set_qf_name;
938 case Opt_grpjquota:
939 qtype = GRPQUOTA;
940set_qf_name:
941 if (sb_any_quota_enabled(sb)) {
942 printk(KERN_ERR
943 "EXT3-fs: Cannot change journalled "
944 "quota options when quota turned on.\n");
945 return 0;
946 }
947 qname = match_strdup(&args[0]);
948 if (!qname) {
949 printk(KERN_ERR
950 "EXT3-fs: not enough memory for "
951 "storing quotafile name.\n");
952 return 0;
953 }
954 if (sbi->s_qf_names[qtype] &&
955 strcmp(sbi->s_qf_names[qtype], qname)) {
956 printk(KERN_ERR
957 "EXT3-fs: %s quota file already "
958 "specified.\n", QTYPE2NAME(qtype));
959 kfree(qname);
960 return 0;
961 }
962 sbi->s_qf_names[qtype] = qname;
963 if (strchr(sbi->s_qf_names[qtype], '/')) {
964 printk(KERN_ERR
965 "EXT3-fs: quotafile must be on "
966 "filesystem root.\n");
967 kfree(sbi->s_qf_names[qtype]);
968 sbi->s_qf_names[qtype] = NULL;
969 return 0;
970 }
1f54587b 971 set_opt(sbi->s_mount_opt, QUOTA);
1da177e4
LT		 972 break;
973 case Opt_offusrjquota:
974 qtype = USRQUOTA;
975 goto clear_qf_name;
976 case Opt_offgrpjquota:
977 qtype = GRPQUOTA;
978clear_qf_name:
979 if (sb_any_quota_enabled(sb)) {
980 printk(KERN_ERR "EXT3-fs: Cannot change "
981 "journalled quota options when "
982 "quota turned on.\n");
983 return 0;
984 }
08c6a96f
JK		 985 /*
986 * The space will be released later when all options
987 * are confirmed to be correct
988 */
1da177e4
LT		 989 sbi->s_qf_names[qtype] = NULL;
990 break;
991 case Opt_jqfmt_vfsold:
992 sbi->s_jquota_fmt = QFMT_VFS_OLD;
993 break;
994 case Opt_jqfmt_vfsv0:
995 sbi->s_jquota_fmt = QFMT_VFS_V0;
996 break;
1f54587b 997 case Opt_quota:
8fc2751b
MB		 998 case Opt_usrquota:
999 set_opt(sbi->s_mount_opt, QUOTA);
1000 set_opt(sbi->s_mount_opt, USRQUOTA);
1001 break;
1002 case Opt_grpquota:
1f54587b 1003 set_opt(sbi->s_mount_opt, QUOTA);
8fc2751b 1004 set_opt(sbi->s_mount_opt, GRPQUOTA);
1f54587b
JK		 1005 break;
1006 case Opt_noquota:
1007 if (sb_any_quota_enabled(sb)) {
1008 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1009 "options when quota turned on.\n");
1010 return 0;
1011 }
1012 clear_opt(sbi->s_mount_opt, QUOTA);
8fc2751b
MB		 1013 clear_opt(sbi->s_mount_opt, USRQUOTA);
1014 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1f54587b 1015 break;
1da177e4 1016#else
8fc2751b
MB		 1017 case Opt_quota:
1018 case Opt_usrquota:
1019 case Opt_grpquota:
1da177e4
LT		 1020 case Opt_usrjquota:
1021 case Opt_grpjquota:
1022 case Opt_offusrjquota:
1023 case Opt_offgrpjquota:
1024 case Opt_jqfmt_vfsold:
1025 case Opt_jqfmt_vfsv0:
1026 printk(KERN_ERR
1027 "EXT3-fs: journalled quota options not "
1028 "supported.\n");
1029 break;
1f54587b
JK		 1030 case Opt_noquota:
1031 break;
1da177e4
LT		 1032#endif
1033 case Opt_abort:
1034 set_opt(sbi->s_mount_opt, ABORT);
1035 break;
1036 case Opt_barrier:
1037 if (match_int(&args[0], &option))
1038 return 0;
1039 if (option)
1040 set_opt(sbi->s_mount_opt, BARRIER);
1041 else
1042 clear_opt(sbi->s_mount_opt, BARRIER);
1043 break;
1044 case Opt_ignore:
1045 break;
1046 case Opt_resize:
08c6a96f 1047 if (!is_remount) {
1da177e4
LT		 1048 printk("EXT3-fs: resize option only available "
1049 "for remount\n");
1050 return 0;
1051 }
c7f1721e
KZ		 1052 if (match_int(&args[0], &option) != 0)
1053 return 0;
1da177e4
LT		 1054 *n_blocks_count = option;
1055 break;
1056 case Opt_nobh:
1057 set_opt(sbi->s_mount_opt, NOBH);
1058 break;
ade1a29e
BP		 1059 case Opt_bh:
1060 clear_opt(sbi->s_mount_opt, NOBH);
1061 break;
1da177e4
LT		 1062 default:
1063 printk (KERN_ERR
1064 "EXT3-fs: Unrecognized mount option \"%s\" "
1065 "or missing value\n", p);
1066 return 0;
1067 }
1068 }
1069#ifdef CONFIG_QUOTA
8fc2751b
MB		 1070 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1071 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1072 sbi->s_qf_names[USRQUOTA])
1073 clear_opt(sbi->s_mount_opt, USRQUOTA);
1074
1075 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1076 sbi->s_qf_names[GRPQUOTA])
1077 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1078
1079 if ((sbi->s_qf_names[USRQUOTA] &&
1080 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1081 (sbi->s_qf_names[GRPQUOTA] &&
1082 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1083 printk(KERN_ERR "EXT3-fs: old and new quota "
1084 "format mixing.\n");
1085 return 0;
1086 }
1087
1088 if (!sbi->s_jquota_fmt) {
1089 printk(KERN_ERR "EXT3-fs: journalled quota format "
1090 "not specified.\n");
1091 return 0;
1092 }
1093 } else {
1094 if (sbi->s_jquota_fmt) {
1095 printk(KERN_ERR "EXT3-fs: journalled quota format "
1096 "specified with no journalling "
1097 "enabled.\n");
1098 return 0;
1099 }
1da177e4
LT		 1100 }
1101#endif
1da177e4
LT		 1102 return 1;
1103}
1104
1105static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1106 int read_only)
1107{
1108 struct ext3_sb_info *sbi = EXT3_SB(sb);
1109 int res = 0;
1110
1111 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1112 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1113 "forcing read-only mode\n");
1114 res = MS_RDONLY;
1115 }
1116 if (read_only)
1117 return res;
1118 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1119 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1120 "running e2fsck is recommended\n");
1121 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1122 printk (KERN_WARNING
1123 "EXT3-fs warning: mounting fs with errors, "
1124 "running e2fsck is recommended\n");
1125 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1126 le16_to_cpu(es->s_mnt_count) >=
1127 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1128 printk (KERN_WARNING
1129 "EXT3-fs warning: maximal mount count reached, "
1130 "running e2fsck is recommended\n");
1131 else if (le32_to_cpu(es->s_checkinterval) &&
1132 (le32_to_cpu(es->s_lastcheck) +
1133 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1134 printk (KERN_WARNING
1135 "EXT3-fs warning: checktime reached, "
1136 "running e2fsck is recommended\n");
1137#if 0
1138 /* @@@ We _will_ want to clear the valid bit if we find
1139 inconsistencies, to force a fsck at reboot. But for
1140 a plain journaled filesystem we can keep it set as
1141 valid forever! :) */
1142 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT3_VALID_FS);
1143#endif
1144 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1145 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1146 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1147 es->s_mtime = cpu_to_le32(get_seconds());
1148 ext3_update_dynamic_rev(sb);
1149 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1150
1151 ext3_commit_super(sb, es, 1);
1152 if (test_opt(sb, DEBUG))
1153 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1154 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1155 sb->s_blocksize,
1156 sbi->s_groups_count,
1157 EXT3_BLOCKS_PER_GROUP(sb),
1158 EXT3_INODES_PER_GROUP(sb),
1159 sbi->s_mount_opt);
1160
1161 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1162 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1163 char b[BDEVNAME_SIZE];
1164
1165 printk("external journal on %s\n",
1166 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1167 } else {
1168 printk("internal journal\n");
1169 }
1da177e4
LT		 1170 return res;
1171}
1172
1173/* Called at mount-time, super-block is locked */
1174static int ext3_check_descriptors (struct super_block * sb)
1175{
1176 struct ext3_sb_info *sbi = EXT3_SB(sb);
855565e8
ES		 1177 ext3_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1178 ext3_fsblk_t last_block;
1da177e4
LT		 1179 struct ext3_group_desc * gdp = NULL;
1180 int desc_block = 0;
1181 int i;
1182
1183 ext3_debug ("Checking group descriptors");
1184
1185 for (i = 0; i < sbi->s_groups_count; i++)
1186 {
855565e8
ES		 1187 if (i == sbi->s_groups_count - 1)
1188 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1189 else
1190 last_block = first_block +
1191 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1192
1da177e4
LT		 1193 if ((i % EXT3_DESC_PER_BLOCK(sb)) == 0)
1194 gdp = (struct ext3_group_desc *)
1195 sbi->s_group_desc[desc_block++]->b_data;
855565e8
ES		 1196 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1197 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1da177e4
LT		 1198 {
1199 ext3_error (sb, "ext3_check_descriptors",
1200 "Block bitmap for group %d"
1201 " not in group (block %lu)!",
1202 i, (unsigned long)
1203 le32_to_cpu(gdp->bg_block_bitmap));
1204 return 0;
1205 }
855565e8
ES		 1206 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1207 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1da177e4
LT		 1208 {
1209 ext3_error (sb, "ext3_check_descriptors",
1210 "Inode bitmap for group %d"
1211 " not in group (block %lu)!",
1212 i, (unsigned long)
1213 le32_to_cpu(gdp->bg_inode_bitmap));
1214 return 0;
1215 }
855565e8
ES		 1216 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1217 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group >
1218 last_block)
1da177e4
LT		 1219 {
1220 ext3_error (sb, "ext3_check_descriptors",
1221 "Inode table for group %d"
1222 " not in group (block %lu)!",
1223 i, (unsigned long)
1224 le32_to_cpu(gdp->bg_inode_table));
1225 return 0;
1226 }
855565e8 1227 first_block += EXT3_BLOCKS_PER_GROUP(sb);
1da177e4
LT		 1228 gdp++;
1229 }
1230
1231 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1232 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1233 return 1;
1234}
1235
1236
1237/* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1238 * the superblock) which were deleted from all directories, but held open by
1239 * a process at the time of a crash. We walk the list and try to delete these
1240 * inodes at recovery time (only with a read-write filesystem).
1241 *
1242 * In order to keep the orphan inode chain consistent during traversal (in
1243 * case of crash during recovery), we link each inode into the superblock
1244 * orphan list_head and handle it the same way as an inode deletion during
1245 * normal operation (which journals the operations for us).
1246 *
1247 * We only do an iget() and an iput() on each inode, which is very safe if we
1248 * accidentally point at an in-use or already deleted inode. The worst that
1249 * can happen in this case is that we get a "bit already cleared" message from
1250 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1251 * e2fsck was run on this filesystem, and it must have already done the orphan
1252 * inode cleanup for us, so we can safely abort without any further action.
1253 */
1254static void ext3_orphan_cleanup (struct super_block * sb,
1255 struct ext3_super_block * es)
1256{
1257 unsigned int s_flags = sb->s_flags;
1258 int nr_orphans = 0, nr_truncates = 0;
1259#ifdef CONFIG_QUOTA
1260 int i;
1261#endif
1262 if (!es->s_last_orphan) {
1263 jbd_debug(4, "no orphan inodes to clean up\n");
1264 return;
1265 }
1266
a8f48a95
ES		 1267 if (bdev_read_only(sb->s_bdev)) {
1268 printk(KERN_ERR "EXT3-fs: write access "
1269 "unavailable, skipping orphan cleanup.\n");
1270 return;
1271 }
1272
1da177e4
LT		 1273 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1274 if (es->s_last_orphan)
1275 jbd_debug(1, "Errors on filesystem, "
1276 "clearing orphan list.\n");
1277 es->s_last_orphan = 0;
1278 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1279 return;
1280 }
1281
1282 if (s_flags & MS_RDONLY) {
1283 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1284 sb->s_id);
1285 sb->s_flags &= ~MS_RDONLY;
1286 }
1287#ifdef CONFIG_QUOTA
1288 /* Needed for iput() to work correctly and not trash data */
1289 sb->s_flags |= MS_ACTIVE;
1290 /* Turn on quotas so that they are updated correctly */
1291 for (i = 0; i < MAXQUOTAS; i++) {
1292 if (EXT3_SB(sb)->s_qf_names[i]) {
1293 int ret = ext3_quota_on_mount(sb, i);
1294 if (ret < 0)
1295 printk(KERN_ERR
1296 "EXT3-fs: Cannot turn on journalled "
1297 "quota: error %d\n", ret);
1298 }
1299 }
1300#endif
1301
1302 while (es->s_last_orphan) {
1303 struct inode *inode;
1304
1305 if (!(inode =
1306 ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1307 es->s_last_orphan = 0;
1308 break;
1309 }
1310
1311 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1312 DQUOT_INIT(inode);
1313 if (inode->i_nlink) {
1314 printk(KERN_DEBUG
eee194e7 1315 "%s: truncating inode %lu to %Ld bytes\n",
1da177e4 1316 __FUNCTION__, inode->i_ino, inode->i_size);
eee194e7 1317 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1da177e4
LT		 1318 inode->i_ino, inode->i_size);
1319 ext3_truncate(inode);
1320 nr_truncates++;
1321 } else {
1322 printk(KERN_DEBUG
eee194e7 1323 "%s: deleting unreferenced inode %lu\n",
1da177e4 1324 __FUNCTION__, inode->i_ino);
eee194e7 1325 jbd_debug(2, "deleting unreferenced inode %lu\n",
1da177e4
LT		 1326 inode->i_ino);
1327 nr_orphans++;
1328 }
1329 iput(inode); /* The delete magic happens here! */
1330 }
1331
1332#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1333
1334 if (nr_orphans)
1335 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1336 sb->s_id, PLURAL(nr_orphans));
1337 if (nr_truncates)
1338 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1339 sb->s_id, PLURAL(nr_truncates));
1340#ifdef CONFIG_QUOTA
1341 /* Turn quotas off */
1342 for (i = 0; i < MAXQUOTAS; i++) {
1343 if (sb_dqopt(sb)->files[i])
1344 vfs_quota_off(sb, i);
1345 }
1346#endif
1347 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1348}
1349
1350#define log2(n) ffz(~(n))
1351
1352/*
1353 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1354 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1355 * We need to be 1 filesystem block less than the 2^32 sector limit.
1356 */
1357static loff_t ext3_max_size(int bits)
1358{
1359 loff_t res = EXT3_NDIR_BLOCKS;
1360 /* This constant is calculated to be the largest file size for a
1361 * dense, 4k-blocksize file such that the total number of
1362 * sectors in the file, including data and all indirect blocks,
1363 * does not exceed 2^32. */
1364 const loff_t upper_limit = 0x1ff7fffd000LL;
1365
1366 res += 1LL << (bits-2);
1367 res += 1LL << (2*(bits-2));
1368 res += 1LL << (3*(bits-2));
1369 res <<= bits;
1370 if (res > upper_limit)
1371 res = upper_limit;
1372 return res;
1373}
1374
43d23f90
MC		 1375static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1376 ext3_fsblk_t logic_sb_block,
1da177e4
LT		 1377 int nr)
1378{
1379 struct ext3_sb_info *sbi = EXT3_SB(sb);
43d23f90 1380 unsigned long bg, first_meta_bg;
1da177e4
LT		 1381 int has_super = 0;
1382
1da177e4
LT		 1383 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1384
1385 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1386 nr < first_meta_bg)
1387 return (logic_sb_block + nr + 1);
1388 bg = sbi->s_desc_per_block * nr;
1389 if (ext3_bg_has_super(sb, bg))
1390 has_super = 1;
43d23f90 1391 return (has_super + ext3_group_first_block_no(sb, bg));
1da177e4
LT		 1392}
1393
1394
1395static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1396{
1397 struct buffer_head * bh;
1398 struct ext3_super_block *es = NULL;
1399 struct ext3_sb_info *sbi;
43d23f90
MC		 1400 ext3_fsblk_t block;
1401 ext3_fsblk_t sb_block = get_sb_block(&data);
1402 ext3_fsblk_t logic_sb_block;
1da177e4 1403 unsigned long offset = 0;
eee194e7 1404 unsigned int journal_inum = 0;
71b96257 1405 unsigned long journal_devnum = 0;
1da177e4
LT		 1406 unsigned long def_mount_opts;
1407 struct inode *root;
1408 int blocksize;
1409 int hblock;
1410 int db_count;
1411 int i;
1412 int needs_recovery;
1413 __le32 features;
1414
f8314dc6 1415 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1da177e4
LT		 1416 if (!sbi)
1417 return -ENOMEM;
1418 sb->s_fs_info = sbi;
1da177e4
LT		 1419 sbi->s_mount_opt = 0;
1420 sbi->s_resuid = EXT3_DEF_RESUID;
1421 sbi->s_resgid = EXT3_DEF_RESGID;
1422
1423 unlock_kernel();
1424
1425 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1426 if (!blocksize) {
1427 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1428 goto out_fail;
1429 }
1430
1431 /*
1432 * The ext3 superblock will not be buffer aligned for other than 1kB
1433 * block sizes. We need to calculate the offset from buffer start.
1434 */
1435 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1436 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1437 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1438 } else {
1439 logic_sb_block = sb_block;
1440 }
1441
1442 if (!(bh = sb_bread(sb, logic_sb_block))) {
1443 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1444 goto out_fail;
1445 }
1446 /*
1447 * Note: s_es must be initialized as soon as possible because
1448 * some ext3 macro-instructions depend on its value
1449 */
1450 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1451 sbi->s_es = es;
1452 sb->s_magic = le16_to_cpu(es->s_magic);
1453 if (sb->s_magic != EXT3_SUPER_MAGIC)
1454 goto cantfind_ext3;
1455
1456 /* Set defaults before we parse the mount options */
1457 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1458 if (def_mount_opts & EXT3_DEFM_DEBUG)
1459 set_opt(sbi->s_mount_opt, DEBUG);
1460 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1461 set_opt(sbi->s_mount_opt, GRPID);
1462 if (def_mount_opts & EXT3_DEFM_UID16)
1463 set_opt(sbi->s_mount_opt, NO_UID32);
1464 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1465 set_opt(sbi->s_mount_opt, XATTR_USER);
1466 if (def_mount_opts & EXT3_DEFM_ACL)
1467 set_opt(sbi->s_mount_opt, POSIX_ACL);
1468 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1469 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1470 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1471 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1472 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1473 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1474
1475 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1476 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1477 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_RO)
1478 set_opt(sbi->s_mount_opt, ERRORS_RO);
2245d7c2
DM		 1479 else
1480 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1da177e4
LT		 1481
1482 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1483 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1484
1485 set_opt(sbi->s_mount_opt, RESERVATION);
1486
71b96257
JL		 1487 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1488 NULL, 0))
1da177e4
LT		 1489 goto failed_mount;
1490
1491 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1492 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1493
1494 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1495 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1496 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1497 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ae6ddcc5 1498 printk(KERN_WARNING
1da177e4
LT		 1499 "EXT3-fs warning: feature flags set on rev 0 fs, "
1500 "running e2fsck is recommended\n");
1501 /*
1502 * Check feature flags regardless of the revision level, since we
1503 * previously didn't change the revision level when setting the flags,
1504 * so there is a chance incompat flags are set on a rev 0 filesystem.
1505 */
1506 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1507 if (features) {
1508 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1509 "unsupported optional features (%x).\n",
1510 sb->s_id, le32_to_cpu(features));
1511 goto failed_mount;
1512 }
1513 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1514 if (!(sb->s_flags & MS_RDONLY) && features) {
1515 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1516 "unsupported optional features (%x).\n",
1517 sb->s_id, le32_to_cpu(features));
1518 goto failed_mount;
1519 }
1520 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1521
1522 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1523 blocksize > EXT3_MAX_BLOCK_SIZE) {
ae6ddcc5 1524 printk(KERN_ERR
1da177e4
LT		 1525 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1526 blocksize, sb->s_id);
1527 goto failed_mount;
1528 }
1529
1530 hblock = bdev_hardsect_size(sb->s_bdev);
1531 if (sb->s_blocksize != blocksize) {
1532 /*
1533 * Make sure the blocksize for the filesystem is larger
1534 * than the hardware sectorsize for the machine.
1535 */
1536 if (blocksize < hblock) {
1537 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1538 "device blocksize %d.\n", blocksize, hblock);
1539 goto failed_mount;
1540 }
1541
1542 brelse (bh);
1543 sb_set_blocksize(sb, blocksize);
1544 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1545 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1546 bh = sb_bread(sb, logic_sb_block);
1547 if (!bh) {
ae6ddcc5 1548 printk(KERN_ERR
1da177e4
LT		 1549 "EXT3-fs: Can't read superblock on 2nd try.\n");
1550 goto failed_mount;
1551 }
1552 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1553 sbi->s_es = es;
1554 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
ae6ddcc5 1555 printk (KERN_ERR
1da177e4
LT		 1556 "EXT3-fs: Magic mismatch, very weird !\n");
1557 goto failed_mount;
1558 }
1559 }
1560
1561 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1562
1563 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1564 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1565 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1566 } else {
1567 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1568 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1569 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1570 (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
1571 (sbi->s_inode_size > blocksize)) {
1572 printk (KERN_ERR
1573 "EXT3-fs: unsupported inode size: %d\n",
1574 sbi->s_inode_size);
1575 goto failed_mount;
1576 }
1577 }
1578 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1579 le32_to_cpu(es->s_log_frag_size);
1580 if (blocksize != sbi->s_frag_size) {
1581 printk(KERN_ERR
1582 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1583 sbi->s_frag_size, blocksize);
1584 goto failed_mount;
1585 }
1586 sbi->s_frags_per_block = 1;
1587 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1588 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1589 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1590 if (EXT3_INODE_SIZE(sb) == 0)
1591 goto cantfind_ext3;
1592 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1593 if (sbi->s_inodes_per_block == 0)
1594 goto cantfind_ext3;
1595 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1596 sbi->s_inodes_per_block;
1597 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1598 sbi->s_sbh = bh;
1599 sbi->s_mount_state = le16_to_cpu(es->s_state);
1600 sbi->s_addr_per_block_bits = log2(EXT3_ADDR_PER_BLOCK(sb));
1601 sbi->s_desc_per_block_bits = log2(EXT3_DESC_PER_BLOCK(sb));
1602 for (i=0; i < 4; i++)
1603 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1604 sbi->s_def_hash_version = es->s_def_hash_version;
1605
1606 if (sbi->s_blocks_per_group > blocksize * 8) {
1607 printk (KERN_ERR
1608 "EXT3-fs: #blocks per group too big: %lu\n",
1609 sbi->s_blocks_per_group);
1610 goto failed_mount;
1611 }
1612 if (sbi->s_frags_per_group > blocksize * 8) {
1613 printk (KERN_ERR
1614 "EXT3-fs: #fragments per group too big: %lu\n",
1615 sbi->s_frags_per_group);
1616 goto failed_mount;
1617 }
1618 if (sbi->s_inodes_per_group > blocksize * 8) {
1619 printk (KERN_ERR
1620 "EXT3-fs: #inodes per group too big: %lu\n",
1621 sbi->s_inodes_per_group);
1622 goto failed_mount;
1623 }
1624
fcd5df35
MC		 1625 if (le32_to_cpu(es->s_blocks_count) >
1626 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1627 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1628 " too large to mount safely\n", sb->s_id);
1629 if (sizeof(sector_t) < 8)
1630 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1631 "enabled\n");
1632 goto failed_mount;
1633 }
1634
1da177e4
LT		 1635 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1636 goto cantfind_ext3;
855565e8
ES		 1637 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1638 le32_to_cpu(es->s_first_data_block) - 1)
1639 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1da177e4
LT		 1640 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1641 EXT3_DESC_PER_BLOCK(sb);
1642 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1643 GFP_KERNEL);
1644 if (sbi->s_group_desc == NULL) {
1645 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1646 goto failed_mount;
1647 }
1648
1da177e4
LT		 1649 bgl_lock_init(&sbi->s_blockgroup_lock);
1650
1651 for (i = 0; i < db_count; i++) {
1652 block = descriptor_loc(sb, logic_sb_block, i);
1653 sbi->s_group_desc[i] = sb_bread(sb, block);
1654 if (!sbi->s_group_desc[i]) {
1655 printk (KERN_ERR "EXT3-fs: "
1656 "can't read group descriptor %d\n", i);
1657 db_count = i;
1658 goto failed_mount2;
1659 }
1660 }
1661 if (!ext3_check_descriptors (sb)) {
d2e5b13c 1662 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1da177e4
LT		 1663 goto failed_mount2;
1664 }
1665 sbi->s_gdb_count = db_count;
1666 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1667 spin_lock_init(&sbi->s_next_gen_lock);
0216bfcf
MC		 1668
1669 percpu_counter_init(&sbi->s_freeblocks_counter,
1670 ext3_count_free_blocks(sb));
1671 percpu_counter_init(&sbi->s_freeinodes_counter,
1672 ext3_count_free_inodes(sb));
1673 percpu_counter_init(&sbi->s_dirs_counter,
1674 ext3_count_dirs(sb));
1675
1da177e4
LT		 1676 /* per fileystem reservation list head & lock */
1677 spin_lock_init(&sbi->s_rsv_window_lock);
1678 sbi->s_rsv_window_root = RB_ROOT;
1679 /* Add a single, static dummy reservation to the start of the
1680 * reservation window list --- it gives us a placeholder for
1681 * append-at-start-of-list which makes the allocation logic
1682 * _much_ simpler. */
1683 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1684 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1685 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1686 sbi->s_rsv_window_head.rsv_goal_size = 0;
1687 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1688
1689 /*
1690 * set up enough so that it can read an inode
1691 */
1692 sb->s_op = &ext3_sops;
1693 sb->s_export_op = &ext3_export_ops;
1694 sb->s_xattr = ext3_xattr_handlers;
1695#ifdef CONFIG_QUOTA
1696 sb->s_qcop = &ext3_qctl_operations;
1697 sb->dq_op = &ext3_quota_operations;
1698#endif
1699 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1700
1701 sb->s_root = NULL;
1702
1703 needs_recovery = (es->s_last_orphan != 0 ||
1704 EXT3_HAS_INCOMPAT_FEATURE(sb,
1705 EXT3_FEATURE_INCOMPAT_RECOVER));
1706
1707 /*
1708 * The first inode we look at is the journal inode. Don't try
1709 * root first: it may be modified in the journal!
1710 */
1711 if (!test_opt(sb, NOLOAD) &&
1712 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
71b96257 1713 if (ext3_load_journal(sb, es, journal_devnum))
0216bfcf 1714 goto failed_mount3;
1da177e4
LT		 1715 } else if (journal_inum) {
1716 if (ext3_create_journal(sb, es, journal_inum))
0216bfcf 1717 goto failed_mount3;
1da177e4
LT		 1718 } else {
1719 if (!silent)
1720 printk (KERN_ERR
1721 "ext3: No journal on filesystem on %s\n",
1722 sb->s_id);
0216bfcf 1723 goto failed_mount3;
1da177e4
LT		 1724 }
1725
1726 /* We have now updated the journal if required, so we can
1727 * validate the data journaling mode. */
1728 switch (test_opt(sb, DATA_FLAGS)) {
1729 case 0:
1730 /* No mode set, assume a default based on the journal
1731 capabilities: ORDERED_DATA if the journal can
1732 cope, else JOURNAL_DATA */
1733 if (journal_check_available_features
1734 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1735 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1736 else
1737 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1738 break;
1739
1740 case EXT3_MOUNT_ORDERED_DATA:
1741 case EXT3_MOUNT_WRITEBACK_DATA:
1742 if (!journal_check_available_features
1743 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1744 printk(KERN_ERR "EXT3-fs: Journal does not support "
1745 "requested data journaling mode\n");
0216bfcf 1746 goto failed_mount4;
1da177e4
LT		 1747 }
1748 default:
1749 break;
1750 }
1751
1752 if (test_opt(sb, NOBH)) {
1da177e4
LT		 1753 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1754 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1755 "its supported only with writeback mode\n");
1756 clear_opt(sbi->s_mount_opt, NOBH);
1757 }
1758 }
1759 /*
1760 * The journal_load will have done any necessary log recovery,
1761 * so we can safely mount the rest of the filesystem now.
1762 */
1763
1764 root = iget(sb, EXT3_ROOT_INO);
1765 sb->s_root = d_alloc_root(root);
1766 if (!sb->s_root) {
1767 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1768 iput(root);
0216bfcf 1769 goto failed_mount4;
1da177e4
LT		 1770 }
1771 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1772 dput(sb->s_root);
1773 sb->s_root = NULL;
1774 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
0216bfcf 1775 goto failed_mount4;
1da177e4
LT		 1776 }
1777
1778 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1779 /*
1780 * akpm: core read_super() calls in here with the superblock locked.
1781 * That deadlocks, because orphan cleanup needs to lock the superblock
1782 * in numerous places. Here we just pop the lock - it's relatively
1783 * harmless, because we are now ready to accept write_super() requests,
1784 * and aviro says that's the only reason for hanging onto the
1785 * superblock lock.
1786 */
1787 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1788 ext3_orphan_cleanup(sb, es);
1789 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1790 if (needs_recovery)
1791 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1792 ext3_mark_recovery_complete(sb, es);
1793 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1794 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1795 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1796 "writeback");
1797
1da177e4
LT		 1798 lock_kernel();
1799 return 0;
1800
1801cantfind_ext3:
1802 if (!silent)
1803 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1804 sb->s_id);
1805 goto failed_mount;
1806
0216bfcf 1807failed_mount4:
1da177e4 1808 journal_destroy(sbi->s_journal);
0216bfcf
MC		 1809failed_mount3:
1810 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1811 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1812 percpu_counter_destroy(&sbi->s_dirs_counter);
1da177e4
LT		 1813failed_mount2:
1814 for (i = 0; i < db_count; i++)
1815 brelse(sbi->s_group_desc[i]);
1816 kfree(sbi->s_group_desc);
1817failed_mount:
1818#ifdef CONFIG_QUOTA
1819 for (i = 0; i < MAXQUOTAS; i++)
1820 kfree(sbi->s_qf_names[i]);
1821#endif
1822 ext3_blkdev_remove(sbi);
1823 brelse(bh);
1824out_fail:
1825 sb->s_fs_info = NULL;
1826 kfree(sbi);
1827 lock_kernel();
1828 return -EINVAL;
1829}
1830
1831/*
1832 * Setup any per-fs journal parameters now. We'll do this both on
1833 * initial mount, once the journal has been initialised but before we've
ae6ddcc5 1834 * done any recovery; and again on any subsequent remount.
1da177e4
LT		 1835 */
1836static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1837{
1838 struct ext3_sb_info *sbi = EXT3_SB(sb);
1839
1840 if (sbi->s_commit_interval)
1841 journal->j_commit_interval = sbi->s_commit_interval;
1842 /* We could also set up an ext3-specific default for the commit
1843 * interval here, but for now we'll just fall back to the jbd
1844 * default. */
1845
1846 spin_lock(&journal->j_state_lock);
1847 if (test_opt(sb, BARRIER))
1848 journal->j_flags |= JFS_BARRIER;
1849 else
1850 journal->j_flags &= ~JFS_BARRIER;
1851 spin_unlock(&journal->j_state_lock);
1852}
1853
eee194e7
ES		 1854static journal_t *ext3_get_journal(struct super_block *sb,
1855 unsigned int journal_inum)
1da177e4
LT		 1856{
1857 struct inode *journal_inode;
1858 journal_t *journal;
1859
1860 /* First, test for the existence of a valid inode on disk. Bad
1861 * things happen if we iget() an unused inode, as the subsequent
1862 * iput() will try to delete it. */
1863
1864 journal_inode = iget(sb, journal_inum);
1865 if (!journal_inode) {
1866 printk(KERN_ERR "EXT3-fs: no journal found.\n");
1867 return NULL;
1868 }
1869 if (!journal_inode->i_nlink) {
1870 make_bad_inode(journal_inode);
1871 iput(journal_inode);
1872 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
1873 return NULL;
1874 }
1875
1876 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1877 journal_inode, journal_inode->i_size);
1878 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
1879 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
1880 iput(journal_inode);
1881 return NULL;
1882 }
1883
1884 journal = journal_init_inode(journal_inode);
1885 if (!journal) {
1886 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
1887 iput(journal_inode);
1888 return NULL;
1889 }
1890 journal->j_private = sb;
1891 ext3_init_journal_params(sb, journal);
1892 return journal;
1893}
1894
1895static journal_t *ext3_get_dev_journal(struct super_block *sb,
1896 dev_t j_dev)
1897{
1898 struct buffer_head * bh;
1899 journal_t *journal;
43d23f90 1900 ext3_fsblk_t start;
1c2bf374 1901 ext3_fsblk_t len;
1da177e4 1902 int hblock, blocksize;
43d23f90 1903 ext3_fsblk_t sb_block;
1da177e4
LT		 1904 unsigned long offset;
1905 struct ext3_super_block * es;
1906 struct block_device *bdev;
1907
1908 bdev = ext3_blkdev_get(j_dev);
1909 if (bdev == NULL)
1910 return NULL;
1911
1912 if (bd_claim(bdev, sb)) {
1913 printk(KERN_ERR
1914 "EXT3: failed to claim external journal device.\n");
1915 blkdev_put(bdev);
1916 return NULL;
1917 }
1918
1919 blocksize = sb->s_blocksize;
1920 hblock = bdev_hardsect_size(bdev);
1921 if (blocksize < hblock) {
1922 printk(KERN_ERR
1923 "EXT3-fs: blocksize too small for journal device.\n");
1924 goto out_bdev;
1925 }
1926
1927 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
1928 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
1929 set_blocksize(bdev, blocksize);
1930 if (!(bh = __bread(bdev, sb_block, blocksize))) {
1931 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
1932 "external journal\n");
1933 goto out_bdev;
1934 }
1935
1936 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1937 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
1938 !(le32_to_cpu(es->s_feature_incompat) &
1939 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
1940 printk(KERN_ERR "EXT3-fs: external journal has "
1941 "bad superblock\n");
1942 brelse(bh);
1943 goto out_bdev;
1944 }
1945
1946 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
1947 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
1948 brelse(bh);
1949 goto out_bdev;
1950 }
1951
1952 len = le32_to_cpu(es->s_blocks_count);
1953 start = sb_block + 1;
1954 brelse(bh); /* we're done with the superblock */
1955
1956 journal = journal_init_dev(bdev, sb->s_bdev,
1957 start, len, blocksize);
1958 if (!journal) {
1959 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
1960 goto out_bdev;
1961 }
1962 journal->j_private = sb;
1963 ll_rw_block(READ, 1, &journal->j_sb_buffer);
1964 wait_on_buffer(journal->j_sb_buffer);
1965 if (!buffer_uptodate(journal->j_sb_buffer)) {
1966 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
1967 goto out_journal;
1968 }
1969 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
1970 printk(KERN_ERR "EXT3-fs: External journal has more than one "
1971 "user (unsupported) - %d\n",
1972 be32_to_cpu(journal->j_superblock->s_nr_users));
1973 goto out_journal;
1974 }
1975 EXT3_SB(sb)->journal_bdev = bdev;
1976 ext3_init_journal_params(sb, journal);
1977 return journal;
1978out_journal:
1979 journal_destroy(journal);
1980out_bdev:
1981 ext3_blkdev_put(bdev);
1982 return NULL;
1983}
1984
71b96257
JL		 1985static int ext3_load_journal(struct super_block *sb,
1986 struct ext3_super_block *es,
1987 unsigned long journal_devnum)
1da177e4
LT		 1988{
1989 journal_t *journal;
eee194e7 1990 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
71b96257 1991 dev_t journal_dev;
1da177e4
LT		 1992 int err = 0;
1993 int really_read_only;
1994
71b96257
JL		 1995 if (journal_devnum &&
1996 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
1997 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
1998 "numbers have changed\n");
1999 journal_dev = new_decode_dev(journal_devnum);
2000 } else
2001 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2002
1da177e4
LT		 2003 really_read_only = bdev_read_only(sb->s_bdev);
2004
2005 /*
2006 * Are we loading a blank journal or performing recovery after a
2007 * crash? For recovery, we need to check in advance whether we
2008 * can get read-write access to the device.
2009 */
2010
2011 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2012 if (sb->s_flags & MS_RDONLY) {
2013 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2014 "required on readonly filesystem.\n");
2015 if (really_read_only) {
2016 printk(KERN_ERR "EXT3-fs: write access "
2017 "unavailable, cannot proceed.\n");
2018 return -EROFS;
2019 }
2020 printk (KERN_INFO "EXT3-fs: write access will "
2021 "be enabled during recovery.\n");
2022 }
2023 }
2024
2025 if (journal_inum && journal_dev) {
2026 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2027 "and inode journals!\n");
2028 return -EINVAL;
2029 }
2030
2031 if (journal_inum) {
2032 if (!(journal = ext3_get_journal(sb, journal_inum)))
2033 return -EINVAL;
2034 } else {
2035 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2036 return -EINVAL;
2037 }
2038
2039 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2040 err = journal_update_format(journal);
2041 if (err) {
2042 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2043 journal_destroy(journal);
2044 return err;
2045 }
2046 }
2047
2048 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2049 err = journal_wipe(journal, !really_read_only);
2050 if (!err)
2051 err = journal_load(journal);
2052
2053 if (err) {
2054 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2055 journal_destroy(journal);
2056 return err;
2057 }
2058
2059 EXT3_SB(sb)->s_journal = journal;
2060 ext3_clear_journal_err(sb, es);
71b96257
JL		 2061
2062 if (journal_devnum &&
2063 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2064 es->s_journal_dev = cpu_to_le32(journal_devnum);
2065 sb->s_dirt = 1;
2066
2067 /* Make sure we flush the recovery flag to disk. */
2068 ext3_commit_super(sb, es, 1);
2069 }
2070
1da177e4
LT		 2071 return 0;
2072}
2073
2074static int ext3_create_journal(struct super_block * sb,
2075 struct ext3_super_block * es,
eee194e7 2076 unsigned int journal_inum)
1da177e4
LT		 2077{
2078 journal_t *journal;
2079
2080 if (sb->s_flags & MS_RDONLY) {
2081 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2082 "create journal.\n");
2083 return -EROFS;
2084 }
2085
2086 if (!(journal = ext3_get_journal(sb, journal_inum)))
2087 return -EINVAL;
2088
eee194e7 2089 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
1da177e4
LT		 2090 journal_inum);
2091
2092 if (journal_create(journal)) {
2093 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2094 journal_destroy(journal);
2095 return -EIO;
2096 }
2097
2098 EXT3_SB(sb)->s_journal = journal;
2099
2100 ext3_update_dynamic_rev(sb);
2101 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2102 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2103
2104 es->s_journal_inum = cpu_to_le32(journal_inum);
2105 sb->s_dirt = 1;
2106
2107 /* Make sure we flush the recovery flag to disk. */
2108 ext3_commit_super(sb, es, 1);
2109
2110 return 0;
2111}
2112
2113static void ext3_commit_super (struct super_block * sb,
2114 struct ext3_super_block * es,
2115 int sync)
2116{
2117 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2118
2119 if (!sbh)
2120 return;
2121 es->s_wtime = cpu_to_le32(get_seconds());
2122 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2123 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2124 BUFFER_TRACE(sbh, "marking dirty");
2125 mark_buffer_dirty(sbh);
2126 if (sync)
2127 sync_dirty_buffer(sbh);
2128}
2129
2130
2131/*
2132 * Have we just finished recovery? If so, and if we are mounting (or
2133 * remounting) the filesystem readonly, then we will end up with a
2134 * consistent fs on disk. Record that fact.
2135 */
2136static void ext3_mark_recovery_complete(struct super_block * sb,
2137 struct ext3_super_block * es)
2138{
2139 journal_t *journal = EXT3_SB(sb)->s_journal;
2140
2141 journal_lock_updates(journal);
2142 journal_flush(journal);
2143 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2144 sb->s_flags & MS_RDONLY) {
2145 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2146 sb->s_dirt = 0;
2147 ext3_commit_super(sb, es, 1);
2148 }
2149 journal_unlock_updates(journal);
2150}
2151
2152/*
2153 * If we are mounting (or read-write remounting) a filesystem whose journal
2154 * has recorded an error from a previous lifetime, move that error to the
2155 * main filesystem now.
2156 */
2157static void ext3_clear_journal_err(struct super_block * sb,
2158 struct ext3_super_block * es)
2159{
2160 journal_t *journal;
2161 int j_errno;
2162 const char *errstr;
2163
2164 journal = EXT3_SB(sb)->s_journal;
2165
2166 /*
2167 * Now check for any error status which may have been recorded in the
2168 * journal by a prior ext3_error() or ext3_abort()
2169 */
2170
2171 j_errno = journal_errno(journal);
2172 if (j_errno) {
2173 char nbuf[16];
2174
2175 errstr = ext3_decode_error(sb, j_errno, nbuf);
2176 ext3_warning(sb, __FUNCTION__, "Filesystem error recorded "
2177 "from previous mount: %s", errstr);
2178 ext3_warning(sb, __FUNCTION__, "Marking fs in need of "
2179 "filesystem check.");
2180
2181 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2182 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2183 ext3_commit_super (sb, es, 1);
2184
2185 journal_clear_err(journal);
2186 }
2187}
2188
2189/*
2190 * Force the running and committing transactions to commit,
2191 * and wait on the commit.
2192 */
2193int ext3_force_commit(struct super_block *sb)
2194{
2195 journal_t *journal;
2196 int ret;
2197
2198 if (sb->s_flags & MS_RDONLY)
2199 return 0;
2200
2201 journal = EXT3_SB(sb)->s_journal;
2202 sb->s_dirt = 0;
2203 ret = ext3_journal_force_commit(journal);
2204 return ret;
2205}
2206
2207/*
2208 * Ext3 always journals updates to the superblock itself, so we don't
2209 * have to propagate any other updates to the superblock on disk at this
2210 * point. Just start an async writeback to get the buffers on their way
2211 * to the disk.
2212 *
2213 * This implicitly triggers the writebehind on sync().
2214 */
2215
2216static void ext3_write_super (struct super_block * sb)
2217{
7892f2f4 2218 if (mutex_trylock(&sb->s_lock) != 0)
1da177e4
LT		 2219 BUG();
2220 sb->s_dirt = 0;
2221}
2222
2223static int ext3_sync_fs(struct super_block *sb, int wait)
2224{
2225 tid_t target;
2226
2227 sb->s_dirt = 0;
2228 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2229 if (wait)
2230 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2231 }
2232 return 0;
2233}
2234
2235/*
2236 * LVM calls this function before a (read-only) snapshot is created. This
2237 * gives us a chance to flush the journal completely and mark the fs clean.
2238 */
2239static void ext3_write_super_lockfs(struct super_block *sb)
2240{
2241 sb->s_dirt = 0;
2242
2243 if (!(sb->s_flags & MS_RDONLY)) {
2244 journal_t *journal = EXT3_SB(sb)->s_journal;
2245
2246 /* Now we set up the journal barrier. */
2247 journal_lock_updates(journal);
2248 journal_flush(journal);
2249
2250 /* Journal blocked and flushed, clear needs_recovery flag. */
2251 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2252 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2253 }
2254}
2255
2256/*
2257 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2258 * flag here, even though the filesystem is not technically dirty yet.
2259 */
2260static void ext3_unlockfs(struct super_block *sb)
2261{
2262 if (!(sb->s_flags & MS_RDONLY)) {
2263 lock_super(sb);
2264 /* Reser the needs_recovery flag before the fs is unlocked. */
2265 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2266 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2267 unlock_super(sb);
2268 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2269 }
2270}
2271
2272static int ext3_remount (struct super_block * sb, int * flags, char * data)
2273{
2274 struct ext3_super_block * es;
2275 struct ext3_sb_info *sbi = EXT3_SB(sb);
43d23f90 2276 ext3_fsblk_t n_blocks_count = 0;
08c6a96f
JK		 2277 unsigned long old_sb_flags;
2278 struct ext3_mount_options old_opts;
2279 int err;
2280#ifdef CONFIG_QUOTA
2281 int i;
2282#endif
2283
2284 /* Store the original options */
2285 old_sb_flags = sb->s_flags;
2286 old_opts.s_mount_opt = sbi->s_mount_opt;
2287 old_opts.s_resuid = sbi->s_resuid;
2288 old_opts.s_resgid = sbi->s_resgid;
2289 old_opts.s_commit_interval = sbi->s_commit_interval;
2290#ifdef CONFIG_QUOTA
2291 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2292 for (i = 0; i < MAXQUOTAS; i++)
2293 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2294#endif
1da177e4
LT		 2295
2296 /*
2297 * Allow the "check" option to be passed as a remount option.
2298 */
71b96257 2299 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
08c6a96f
JK		 2300 err = -EINVAL;
2301 goto restore_opts;
2302 }
1da177e4
LT		 2303
2304 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2305 ext3_abort(sb, __FUNCTION__, "Abort forced by user");
2306
2307 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2308 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2309
2310 es = sbi->s_es;
2311
2312 ext3_init_journal_params(sb, sbi->s_journal);
2313
2314 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2315 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
08c6a96f
JK		 2316 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2317 err = -EROFS;
2318 goto restore_opts;
2319 }
1da177e4
LT		 2320
2321 if (*flags & MS_RDONLY) {
2322 /*
2323 * First of all, the unconditional stuff we have to do
2324 * to disable replay of the journal when we next remount
2325 */
2326 sb->s_flags |= MS_RDONLY;
2327
2328 /*
2329 * OK, test if we are remounting a valid rw partition
2330 * readonly, and if so set the rdonly flag and then
2331 * mark the partition as valid again.
2332 */
2333 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2334 (sbi->s_mount_state & EXT3_VALID_FS))
2335 es->s_state = cpu_to_le16(sbi->s_mount_state);
2336
2337 ext3_mark_recovery_complete(sb, es);
2338 } else {
2339 __le32 ret;
2340 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2341 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2342 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2343 "remount RDWR because of unsupported "
2344 "optional features (%x).\n",
2345 sb->s_id, le32_to_cpu(ret));
08c6a96f
JK		 2346 err = -EROFS;
2347 goto restore_opts;
1da177e4
LT		 2348 }
2349 /*
2350 * Mounting a RDONLY partition read-write, so reread
2351 * and store the current valid flag. (It may have
2352 * been changed by e2fsck since we originally mounted
2353 * the partition.)
2354 */
2355 ext3_clear_journal_err(sb, es);
2356 sbi->s_mount_state = le16_to_cpu(es->s_state);
a4e4de36 2357 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
08c6a96f 2358 goto restore_opts;
1da177e4
LT		 2359 if (!ext3_setup_super (sb, es, 0))
2360 sb->s_flags &= ~MS_RDONLY;
2361 }
2362 }
08c6a96f
JK		 2363#ifdef CONFIG_QUOTA
2364 /* Release old quota file names */
2365 for (i = 0; i < MAXQUOTAS; i++)
2366 if (old_opts.s_qf_names[i] &&
2367 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2368 kfree(old_opts.s_qf_names[i]);
2369#endif
1da177e4 2370 return 0;
08c6a96f
JK		 2371restore_opts:
2372 sb->s_flags = old_sb_flags;
2373 sbi->s_mount_opt = old_opts.s_mount_opt;
2374 sbi->s_resuid = old_opts.s_resuid;
2375 sbi->s_resgid = old_opts.s_resgid;
2376 sbi->s_commit_interval = old_opts.s_commit_interval;
2377#ifdef CONFIG_QUOTA
2378 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2379 for (i = 0; i < MAXQUOTAS; i++) {
2380 if (sbi->s_qf_names[i] &&
2381 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2382 kfree(sbi->s_qf_names[i]);
2383 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2384 }
2385#endif
2386 return err;
1da177e4
LT		 2387}
2388
726c3342 2389static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
1da177e4 2390{
726c3342 2391 struct super_block *sb = dentry->d_sb;
09fe316a
AT		 2392 struct ext3_sb_info *sbi = EXT3_SB(sb);
2393 struct ext3_super_block *es = sbi->s_es;
43d23f90 2394 ext3_fsblk_t overhead;
1da177e4 2395 int i;
50ee0a32 2396 u64 fsid;
1da177e4
LT		 2397
2398 if (test_opt (sb, MINIX_DF))
2399 overhead = 0;
2400 else {
2401 unsigned long ngroups;
2402 ngroups = EXT3_SB(sb)->s_groups_count;
2403 smp_rmb();
2404
2405 /*
2406 * Compute the overhead (FS structures)
2407 */
2408
2409 /*
2410 * All of the blocks before first_data_block are
2411 * overhead
2412 */
2413 overhead = le32_to_cpu(es->s_first_data_block);
2414
2415 /*
2416 * Add the overhead attributed to the superblock and
2417 * block group descriptors. If the sparse superblocks
2418 * feature is turned on, then not all groups have this.
2419 */
2420 for (i = 0; i < ngroups; i++) {
2421 overhead += ext3_bg_has_super(sb, i) +
2422 ext3_bg_num_gdb(sb, i);
2423 cond_resched();
2424 }
2425
2426 /*
2427 * Every block group has an inode bitmap, a block
2428 * bitmap, and an inode table.
2429 */
2430 overhead += (ngroups * (2 + EXT3_SB(sb)->s_itb_per_group));
2431 }
2432
2433 buf->f_type = EXT3_SUPER_MAGIC;
2434 buf->f_bsize = sb->s_blocksize;
2435 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - overhead;
09fe316a 2436 buf->f_bfree = percpu_counter_sum(&sbi->s_freeblocks_counter);
1da177e4
LT		 2437 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2438 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2439 buf->f_bavail = 0;
2440 buf->f_files = le32_to_cpu(es->s_inodes_count);
09fe316a 2441 buf->f_ffree = percpu_counter_sum(&sbi->s_freeinodes_counter);
1da177e4 2442 buf->f_namelen = EXT3_NAME_LEN;
50ee0a32
PE		 2443 fsid = le64_to_cpup((void *)es->s_uuid) ^
2444 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2445 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2446 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1da177e4
LT		 2447 return 0;
2448}
2449
2450/* Helper function for writing quotas on sync - we need to start transaction before quota file
2451 * is locked for write. Otherwise the are possible deadlocks:
2452 * Process 1 Process 2
2453 * ext3_create() quota_sync()
2454 * journal_start() write_dquot()
d3be915f
IM		 2455 * DQUOT_INIT() down(dqio_mutex)
2456 * down(dqio_mutex) journal_start()
1da177e4
LT		 2457 *
2458 */
2459
2460#ifdef CONFIG_QUOTA
2461
2462static inline struct inode *dquot_to_inode(struct dquot *dquot)
2463{
2464 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2465}
2466
2467static int ext3_dquot_initialize(struct inode *inode, int type)
2468{
2469 handle_t *handle;
2470 int ret, err;
2471
2472 /* We may create quota structure so we need to reserve enough blocks */
1f54587b 2473 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
1da177e4
LT		 2474 if (IS_ERR(handle))
2475 return PTR_ERR(handle);
2476 ret = dquot_initialize(inode, type);
2477 err = ext3_journal_stop(handle);
2478 if (!ret)
2479 ret = err;
2480 return ret;
2481}
2482
2483static int ext3_dquot_drop(struct inode *inode)
2484{
2485 handle_t *handle;
2486 int ret, err;
2487
2488 /* We may delete quota structure so we need to reserve enough blocks */
1f54587b 2489 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
1da177e4
LT		 2490 if (IS_ERR(handle))
2491 return PTR_ERR(handle);
2492 ret = dquot_drop(inode);
2493 err = ext3_journal_stop(handle);
2494 if (!ret)
2495 ret = err;
2496 return ret;
2497}
2498
2499static int ext3_write_dquot(struct dquot *dquot)
2500{
2501 int ret, err;
2502 handle_t *handle;
2503 struct inode *inode;
2504
2505 inode = dquot_to_inode(dquot);
2506 handle = ext3_journal_start(inode,
1f54587b 2507 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2508 if (IS_ERR(handle))
2509 return PTR_ERR(handle);
2510 ret = dquot_commit(dquot);
2511 err = ext3_journal_stop(handle);
2512 if (!ret)
2513 ret = err;
2514 return ret;
2515}
2516
2517static int ext3_acquire_dquot(struct dquot *dquot)
2518{
2519 int ret, err;
2520 handle_t *handle;
2521
2522 handle = ext3_journal_start(dquot_to_inode(dquot),
1f54587b 2523 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2524 if (IS_ERR(handle))
2525 return PTR_ERR(handle);
2526 ret = dquot_acquire(dquot);
2527 err = ext3_journal_stop(handle);
2528 if (!ret)
2529 ret = err;
2530 return ret;
2531}
2532
2533static int ext3_release_dquot(struct dquot *dquot)
2534{
2535 int ret, err;
2536 handle_t *handle;
2537
2538 handle = ext3_journal_start(dquot_to_inode(dquot),
1f54587b 2539 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2540 if (IS_ERR(handle))
2541 return PTR_ERR(handle);
2542 ret = dquot_release(dquot);
2543 err = ext3_journal_stop(handle);
2544 if (!ret)
2545 ret = err;
2546 return ret;
2547}
2548
2549static int ext3_mark_dquot_dirty(struct dquot *dquot)
2550{
2551 /* Are we journalling quotas? */
2552 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2553 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2554 dquot_mark_dquot_dirty(dquot);
2555 return ext3_write_dquot(dquot);
2556 } else {
2557 return dquot_mark_dquot_dirty(dquot);
2558 }
2559}
2560
2561static int ext3_write_info(struct super_block *sb, int type)
2562{
2563 int ret, err;
2564 handle_t *handle;
2565
2566 /* Data block + inode block */
2567 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2568 if (IS_ERR(handle))
2569 return PTR_ERR(handle);
2570 ret = dquot_commit_info(sb, type);
2571 err = ext3_journal_stop(handle);
2572 if (!ret)
2573 ret = err;
2574 return ret;
2575}
2576
2577/*
2578 * Turn on quotas during mount time - we need to find
2579 * the quota file and such...
2580 */
2581static int ext3_quota_on_mount(struct super_block *sb, int type)
2582{
84de856e
CH		 2583 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2584 EXT3_SB(sb)->s_jquota_fmt, type);
1da177e4
LT		 2585}
2586
2587/*
2588 * Standard function to be called on quota_on
2589 */
2590static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2591 char *path)
2592{
2593 int err;
2594 struct nameidata nd;
2595
1f54587b
JK		 2596 if (!test_opt(sb, QUOTA))
2597 return -EINVAL;
1da177e4
LT		 2598 /* Not journalling quota? */
2599 if (!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
2600 !EXT3_SB(sb)->s_qf_names[GRPQUOTA])
2601 return vfs_quota_on(sb, type, format_id, path);
2602 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2603 if (err)
2604 return err;
2605 /* Quotafile not on the same filesystem? */
2606 if (nd.mnt->mnt_sb != sb) {
2607 path_release(&nd);
2608 return -EXDEV;
2609 }
2610 /* Quotafile not of fs root? */
2611 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2612 printk(KERN_WARNING
2613 "EXT3-fs: Quota file not on filesystem root. "
2614 "Journalled quota will not work.\n");
2615 path_release(&nd);
2616 return vfs_quota_on(sb, type, format_id, path);
2617}
2618
2619/* Read data from quotafile - avoid pagecache and such because we cannot afford
2620 * acquiring the locks... As quota files are never truncated and quota code
2621 * itself serializes the operations (and noone else should touch the files)
2622 * we don't have to be afraid of races */
2623static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2624 size_t len, loff_t off)
2625{
2626 struct inode *inode = sb_dqopt(sb)->files[type];
2627 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2628 int err = 0;
2629 int offset = off & (sb->s_blocksize - 1);
2630 int tocopy;
2631 size_t toread;
2632 struct buffer_head *bh;
2633 loff_t i_size = i_size_read(inode);
2634
2635 if (off > i_size)
2636 return 0;
2637 if (off+len > i_size)
2638 len = i_size-off;
2639 toread = len;
2640 while (toread > 0) {
2641 tocopy = sb->s_blocksize - offset < toread ?
2642 sb->s_blocksize - offset : toread;
2643 bh = ext3_bread(NULL, inode, blk, 0, &err);
2644 if (err)
2645 return err;
2646 if (!bh) /* A hole? */
2647 memset(data, 0, tocopy);
2648 else
2649 memcpy(data, bh->b_data+offset, tocopy);
2650 brelse(bh);
2651 offset = 0;
2652 toread -= tocopy;
2653 data += tocopy;
2654 blk++;
2655 }
2656 return len;
2657}
2658
2659/* Write to quotafile (we know the transaction is already started and has
2660 * enough credits) */
2661static ssize_t ext3_quota_write(struct super_block *sb, int type,
2662 const char *data, size_t len, loff_t off)
2663{
2664 struct inode *inode = sb_dqopt(sb)->files[type];
2665 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2666 int err = 0;
2667 int offset = off & (sb->s_blocksize - 1);
2668 int tocopy;
2669 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2670 size_t towrite = len;
2671 struct buffer_head *bh;
2672 handle_t *handle = journal_current_handle();
2673
5c81a419 2674 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1da177e4
LT		 2675 while (towrite > 0) {
2676 tocopy = sb->s_blocksize - offset < towrite ?
2677 sb->s_blocksize - offset : towrite;
2678 bh = ext3_bread(handle, inode, blk, 1, &err);
2679 if (!bh)
2680 goto out;
2681 if (journal_quota) {
2682 err = ext3_journal_get_write_access(handle, bh);
2683 if (err) {
2684 brelse(bh);
2685 goto out;
2686 }
2687 }
2688 lock_buffer(bh);
2689 memcpy(bh->b_data+offset, data, tocopy);
2690 flush_dcache_page(bh->b_page);
2691 unlock_buffer(bh);
2692 if (journal_quota)
2693 err = ext3_journal_dirty_metadata(handle, bh);
2694 else {
2695 /* Always do at least ordered writes for quotas */
2696 err = ext3_journal_dirty_data(handle, bh);
2697 mark_buffer_dirty(bh);
2698 }
2699 brelse(bh);
2700 if (err)
2701 goto out;
2702 offset = 0;
2703 towrite -= tocopy;
2704 data += tocopy;
2705 blk++;
2706 }
2707out:
2708 if (len == towrite)
2709 return err;
2710 if (inode->i_size < off+len-towrite) {
2711 i_size_write(inode, off+len-towrite);
2712 EXT3_I(inode)->i_disksize = inode->i_size;
2713 }
2714 inode->i_version++;
2715 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2716 ext3_mark_inode_dirty(handle, inode);
1b1dcc1b 2717 mutex_unlock(&inode->i_mutex);
1da177e4
LT		 2718 return len - towrite;
2719}
2720
2721#endif
2722
454e2398
DH		 2723static int ext3_get_sb(struct file_system_type *fs_type,
2724 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1da177e4 2725{
454e2398 2726 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
1da177e4
LT		 2727}
2728
2729static struct file_system_type ext3_fs_type = {
2730 .owner = THIS_MODULE,
2731 .name = "ext3",
2732 .get_sb = ext3_get_sb,
2733 .kill_sb = kill_block_super,
2734 .fs_flags = FS_REQUIRES_DEV,
2735};
2736
2737static int __init init_ext3_fs(void)
2738{
2739 int err = init_ext3_xattr();
2740 if (err)
2741 return err;
2742 err = init_inodecache();
2743 if (err)
2744 goto out1;
2745 err = register_filesystem(&ext3_fs_type);
2746 if (err)
2747 goto out;
2748 return 0;
2749out:
2750 destroy_inodecache();
2751out1:
e9ad5620 2752 exit_ext3_xattr();
1da177e4
LT		 2753 return err;
2754}
2755
2756static void __exit exit_ext3_fs(void)
2757{
2758 unregister_filesystem(&ext3_fs_type);
2759 destroy_inodecache();
2760 exit_ext3_xattr();
2761}
2762
2763MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2764MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2765MODULE_LICENSE("GPL");
2766module_init(init_ext3_fs)
2767module_exit(exit_ext3_fs)
Linux 6.x block layer and io_uring tree(s)