blob: 09b4b313ca074bd93f19bebc7ac2c225e3e50862 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#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>
Mark Bellon8fc27512005-09-06 15:16:54 -070037#include <linux/seq_file.h>
Ben Dooks381be252005-10-30 15:02:56 -080038
Linus Torvalds1da177e2005-04-16 15:20:36 -070039#include <asm/uaccess.h>
Ben Dooks381be252005-10-30 15:02:56 -080040
Linus Torvalds1da177e2005-04-16 15:20:36 -070041#include "xattr.h"
42#include "acl.h"
Ben Dooks381be252005-10-30 15:02:56 -080043#include "namei.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070044
Johann Lombardi71b96252006-01-08 01:03:20 -080045static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
46 unsigned long journal_devnum);
Linus Torvalds1da177e2005-04-16 15:20:36 -070047static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
Eric Sandeeneee194e2006-09-27 01:49:30 -070048 unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -070049static 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);
David Howells726c3342006-06-23 02:02:58 -070060static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -070061static 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
Mingming Caoae6ddcc2006-09-27 01:49:27 -070065/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070066 * 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
Mingming Caoae6ddcc2006-09-27 01:49:27 -070071 * appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 */
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
Mingming Caoae6ddcc2006-09-27 01:49:27 -070093/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070094 * 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
Mingming Caoae6ddcc2006-09-27 01:49:27 -070097 * appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -070098 */
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
Vasily Averin7543fc72006-09-27 01:49:33 -0700162 if (!test_opt (sb, ERRORS_CONT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163 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 }
Vasily Averin7543fc72006-09-27 01:49:33 -0700169 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);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 if (test_opt(sb, ERRORS_PANIC))
175 panic("EXT3-fs (device %s): panic forced after error\n",
176 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177}
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];
Stephen Tweedie30121622005-05-18 11:47:17 -0400233 const char *errstr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234
Stephen Tweedie30121622005-05-18 11:47:17 -0400235 /* 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);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243 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
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700373 printk(KERN_ERR "sb orphan head is %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 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 " "
Eric Sandeeneee194e2006-09-27 01:49:30 -0700380 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 inode->i_sb->s_id, inode->i_ino, inode,
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700382 inode->i_mode, inode->i_nlink,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 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
439static kmem_cache_t *ext3_inode_cachep;
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
448 ei = kmem_cache_alloc(ext3_inode_cachep, SLAB_NOFS);
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
465static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
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
Arjan van de Ven97461512006-03-23 03:00:42 -0800475 mutex_init(&ei->truncate_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 inode_init_once(&ei->vfs_inode);
477 }
478}
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700479
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480static int init_inodecache(void)
481{
482 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
483 sizeof(struct ext3_inode_info),
Paul Jacksonfffb60f2006-03-24 03:16:06 -0800484 0, (SLAB_RECLAIM_ACCOUNT|
485 SLAB_MEM_SPREAD),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 init_once, NULL);
487 if (ext3_inode_cachep == NULL)
488 return -ENOMEM;
489 return 0;
490}
491
492static void destroy_inodecache(void)
493{
494 if (kmem_cache_destroy(ext3_inode_cachep))
495 printk(KERN_INFO "ext3_inode_cache: not all structures were freed\n");
496}
497
498static void ext3_clear_inode(struct inode *inode)
499{
500 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
501#ifdef CONFIG_EXT3_FS_POSIX_ACL
Andrew Mortone6022602006-06-23 02:05:32 -0700502 if (EXT3_I(inode)->i_acl &&
503 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
504 posix_acl_release(EXT3_I(inode)->i_acl);
505 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
506 }
507 if (EXT3_I(inode)->i_default_acl &&
508 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
509 posix_acl_release(EXT3_I(inode)->i_default_acl);
510 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
511 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512#endif
513 ext3_discard_reservation(inode);
514 EXT3_I(inode)->i_block_alloc_info = NULL;
Andrew Mortone6022602006-06-23 02:05:32 -0700515 if (unlikely(rsv))
516 kfree(rsv);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517}
518
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200519static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
Mark Bellon8fc27512005-09-06 15:16:54 -0700520{
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200521#if defined(CONFIG_QUOTA)
OGAWA Hirofumi275abf52005-09-22 21:44:03 -0700522 struct ext3_sb_info *sbi = EXT3_SB(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523
Mark Bellon8fc27512005-09-06 15:16:54 -0700524 if (sbi->s_jquota_fmt)
525 seq_printf(seq, ",jqfmt=%s",
526 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
527
528 if (sbi->s_qf_names[USRQUOTA])
529 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
530
531 if (sbi->s_qf_names[GRPQUOTA])
532 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
533
534 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
535 seq_puts(seq, ",usrquota");
536
537 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
538 seq_puts(seq, ",grpquota");
539#endif
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200540}
541
542static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
543{
544 struct super_block *sb = vfs->mnt_sb;
545
546 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
547 seq_puts(seq, ",data=journal");
548 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
549 seq_puts(seq, ",data=ordered");
550 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
551 seq_puts(seq, ",data=writeback");
552
553 ext3_show_quota_options(seq, sb);
Mark Bellon8fc27512005-09-06 15:16:54 -0700554
555 return 0;
556}
557
NeilBrownfdb36672006-09-16 12:15:38 -0700558
559static struct dentry *ext3_get_dentry(struct super_block *sb, void *vobjp)
560{
561 __u32 *objp = vobjp;
562 unsigned long ino = objp[0];
563 __u32 generation = objp[1];
564 struct inode *inode;
565 struct dentry *result;
566
567 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
568 return ERR_PTR(-ESTALE);
569 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
570 return ERR_PTR(-ESTALE);
571
572 /* iget isn't really right if the inode is currently unallocated!!
573 *
574 * ext3_read_inode will return a bad_inode if the inode had been
575 * deleted, so we should be safe.
576 *
577 * Currently we don't know the generation for parent directory, so
578 * a generation of 0 means "accept any"
579 */
580 inode = iget(sb, ino);
581 if (inode == NULL)
582 return ERR_PTR(-ENOMEM);
583 if (is_bad_inode(inode) ||
584 (generation && inode->i_generation != generation)) {
585 iput(inode);
586 return ERR_PTR(-ESTALE);
587 }
588 /* now to find a dentry.
589 * If possible, get a well-connected one
590 */
591 result = d_alloc_anon(inode);
592 if (!result) {
593 iput(inode);
594 return ERR_PTR(-ENOMEM);
595 }
596 return result;
597}
598
Mark Bellon8fc27512005-09-06 15:16:54 -0700599#ifdef CONFIG_QUOTA
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
601#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
602
603static int ext3_dquot_initialize(struct inode *inode, int type);
604static int ext3_dquot_drop(struct inode *inode);
605static int ext3_write_dquot(struct dquot *dquot);
606static int ext3_acquire_dquot(struct dquot *dquot);
607static int ext3_release_dquot(struct dquot *dquot);
608static int ext3_mark_dquot_dirty(struct dquot *dquot);
609static int ext3_write_info(struct super_block *sb, int type);
610static int ext3_quota_on(struct super_block *sb, int type, int format_id, char *path);
611static int ext3_quota_on_mount(struct super_block *sb, int type);
612static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
613 size_t len, loff_t off);
614static ssize_t ext3_quota_write(struct super_block *sb, int type,
615 const char *data, size_t len, loff_t off);
616
617static struct dquot_operations ext3_quota_operations = {
618 .initialize = ext3_dquot_initialize,
619 .drop = ext3_dquot_drop,
620 .alloc_space = dquot_alloc_space,
621 .alloc_inode = dquot_alloc_inode,
622 .free_space = dquot_free_space,
623 .free_inode = dquot_free_inode,
624 .transfer = dquot_transfer,
625 .write_dquot = ext3_write_dquot,
626 .acquire_dquot = ext3_acquire_dquot,
627 .release_dquot = ext3_release_dquot,
628 .mark_dirty = ext3_mark_dquot_dirty,
629 .write_info = ext3_write_info
630};
631
632static struct quotactl_ops ext3_qctl_operations = {
633 .quota_on = ext3_quota_on,
634 .quota_off = vfs_quota_off,
635 .quota_sync = vfs_quota_sync,
636 .get_info = vfs_get_dqinfo,
637 .set_info = vfs_set_dqinfo,
638 .get_dqblk = vfs_get_dqblk,
639 .set_dqblk = vfs_set_dqblk
640};
641#endif
642
643static struct super_operations ext3_sops = {
644 .alloc_inode = ext3_alloc_inode,
645 .destroy_inode = ext3_destroy_inode,
646 .read_inode = ext3_read_inode,
647 .write_inode = ext3_write_inode,
648 .dirty_inode = ext3_dirty_inode,
649 .delete_inode = ext3_delete_inode,
650 .put_super = ext3_put_super,
651 .write_super = ext3_write_super,
652 .sync_fs = ext3_sync_fs,
653 .write_super_lockfs = ext3_write_super_lockfs,
654 .unlockfs = ext3_unlockfs,
655 .statfs = ext3_statfs,
656 .remount_fs = ext3_remount,
657 .clear_inode = ext3_clear_inode,
Mark Bellon8fc27512005-09-06 15:16:54 -0700658 .show_options = ext3_show_options,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659#ifdef CONFIG_QUOTA
660 .quota_read = ext3_quota_read,
661 .quota_write = ext3_quota_write,
662#endif
663};
664
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665static struct export_operations ext3_export_ops = {
666 .get_parent = ext3_get_parent,
NeilBrownfdb36672006-09-16 12:15:38 -0700667 .get_dentry = ext3_get_dentry,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668};
669
670enum {
671 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
672 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
Adrian Bunk2860b732005-11-08 21:34:59 -0800673 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
Badari Pulavartyade1a292006-06-26 00:25:04 -0700675 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
Johann Lombardi71b96252006-01-08 01:03:20 -0800676 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
678 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
Jan Kara1f545872005-06-23 22:01:04 -0700679 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
Mark Bellon8fc27512005-09-06 15:16:54 -0700680 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
681 Opt_grpquota
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682};
683
684static match_table_t tokens = {
685 {Opt_bsd_df, "bsddf"},
686 {Opt_minix_df, "minixdf"},
687 {Opt_grpid, "grpid"},
688 {Opt_grpid, "bsdgroups"},
689 {Opt_nogrpid, "nogrpid"},
690 {Opt_nogrpid, "sysvgroups"},
691 {Opt_resgid, "resgid=%u"},
692 {Opt_resuid, "resuid=%u"},
693 {Opt_sb, "sb=%u"},
694 {Opt_err_cont, "errors=continue"},
695 {Opt_err_panic, "errors=panic"},
696 {Opt_err_ro, "errors=remount-ro"},
697 {Opt_nouid32, "nouid32"},
698 {Opt_nocheck, "nocheck"},
699 {Opt_nocheck, "check=none"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700 {Opt_debug, "debug"},
701 {Opt_oldalloc, "oldalloc"},
702 {Opt_orlov, "orlov"},
703 {Opt_user_xattr, "user_xattr"},
704 {Opt_nouser_xattr, "nouser_xattr"},
705 {Opt_acl, "acl"},
706 {Opt_noacl, "noacl"},
707 {Opt_reservation, "reservation"},
708 {Opt_noreservation, "noreservation"},
709 {Opt_noload, "noload"},
710 {Opt_nobh, "nobh"},
Badari Pulavartyade1a292006-06-26 00:25:04 -0700711 {Opt_bh, "bh"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712 {Opt_commit, "commit=%u"},
713 {Opt_journal_update, "journal=update"},
714 {Opt_journal_inum, "journal=%u"},
Johann Lombardi71b96252006-01-08 01:03:20 -0800715 {Opt_journal_dev, "journal_dev=%u"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 {Opt_abort, "abort"},
717 {Opt_data_journal, "data=journal"},
718 {Opt_data_ordered, "data=ordered"},
719 {Opt_data_writeback, "data=writeback"},
720 {Opt_offusrjquota, "usrjquota="},
721 {Opt_usrjquota, "usrjquota=%s"},
722 {Opt_offgrpjquota, "grpjquota="},
723 {Opt_grpjquota, "grpjquota=%s"},
724 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
725 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
Mark Bellon8fc27512005-09-06 15:16:54 -0700726 {Opt_grpquota, "grpquota"},
Jan Kara1f545872005-06-23 22:01:04 -0700727 {Opt_noquota, "noquota"},
728 {Opt_quota, "quota"},
Mark Bellon8fc27512005-09-06 15:16:54 -0700729 {Opt_usrquota, "usrquota"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 {Opt_barrier, "barrier=%u"},
731 {Opt_err, NULL},
732 {Opt_resize, "resize"},
733};
734
Mingming Cao43d23f92006-06-25 05:48:07 -0700735static ext3_fsblk_t get_sb_block(void **data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736{
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700737 ext3_fsblk_t sb_block;
738 char *options = (char *) *data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739
740 if (!options || strncmp(options, "sb=", 3) != 0)
741 return 1; /* Default location */
742 options += 3;
Mingming Cao43d23f92006-06-25 05:48:07 -0700743 /*todo: use simple_strtoll with >32bit ext3 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 sb_block = simple_strtoul(options, &options, 0);
745 if (*options && *options != ',') {
746 printk("EXT3-fs: Invalid sb specification: %s\n",
747 (char *) *data);
748 return 1;
749 }
750 if (*options == ',')
751 options++;
752 *data = (void *) options;
753 return sb_block;
754}
755
Johann Lombardi71b96252006-01-08 01:03:20 -0800756static int parse_options (char *options, struct super_block *sb,
Eric Sandeeneee194e2006-09-27 01:49:30 -0700757 unsigned int *inum, unsigned long *journal_devnum,
Mingming Cao43d23f92006-06-25 05:48:07 -0700758 ext3_fsblk_t *n_blocks_count, int is_remount)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759{
760 struct ext3_sb_info *sbi = EXT3_SB(sb);
761 char * p;
762 substring_t args[MAX_OPT_ARGS];
763 int data_opt = 0;
764 int option;
765#ifdef CONFIG_QUOTA
766 int qtype;
767 char *qname;
768#endif
769
770 if (!options)
771 return 1;
772
773 while ((p = strsep (&options, ",")) != NULL) {
774 int token;
775 if (!*p)
776 continue;
777
778 token = match_token(p, tokens, args);
779 switch (token) {
780 case Opt_bsd_df:
781 clear_opt (sbi->s_mount_opt, MINIX_DF);
782 break;
783 case Opt_minix_df:
784 set_opt (sbi->s_mount_opt, MINIX_DF);
785 break;
786 case Opt_grpid:
787 set_opt (sbi->s_mount_opt, GRPID);
788 break;
789 case Opt_nogrpid:
790 clear_opt (sbi->s_mount_opt, GRPID);
791 break;
792 case Opt_resuid:
793 if (match_int(&args[0], &option))
794 return 0;
795 sbi->s_resuid = option;
796 break;
797 case Opt_resgid:
798 if (match_int(&args[0], &option))
799 return 0;
800 sbi->s_resgid = option;
801 break;
802 case Opt_sb:
803 /* handled by get_sb_block() instead of here */
804 /* *sb_block = match_int(&args[0]); */
805 break;
806 case Opt_err_panic:
807 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
808 clear_opt (sbi->s_mount_opt, ERRORS_RO);
809 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
810 break;
811 case Opt_err_ro:
812 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
813 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
814 set_opt (sbi->s_mount_opt, ERRORS_RO);
815 break;
816 case Opt_err_cont:
817 clear_opt (sbi->s_mount_opt, ERRORS_RO);
818 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
819 set_opt (sbi->s_mount_opt, ERRORS_CONT);
820 break;
821 case Opt_nouid32:
822 set_opt (sbi->s_mount_opt, NO_UID32);
823 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 case Opt_nocheck:
825 clear_opt (sbi->s_mount_opt, CHECK);
826 break;
827 case Opt_debug:
828 set_opt (sbi->s_mount_opt, DEBUG);
829 break;
830 case Opt_oldalloc:
831 set_opt (sbi->s_mount_opt, OLDALLOC);
832 break;
833 case Opt_orlov:
834 clear_opt (sbi->s_mount_opt, OLDALLOC);
835 break;
836#ifdef CONFIG_EXT3_FS_XATTR
837 case Opt_user_xattr:
838 set_opt (sbi->s_mount_opt, XATTR_USER);
839 break;
840 case Opt_nouser_xattr:
841 clear_opt (sbi->s_mount_opt, XATTR_USER);
842 break;
843#else
844 case Opt_user_xattr:
845 case Opt_nouser_xattr:
846 printk("EXT3 (no)user_xattr options not supported\n");
847 break;
848#endif
849#ifdef CONFIG_EXT3_FS_POSIX_ACL
850 case Opt_acl:
851 set_opt(sbi->s_mount_opt, POSIX_ACL);
852 break;
853 case Opt_noacl:
854 clear_opt(sbi->s_mount_opt, POSIX_ACL);
855 break;
856#else
857 case Opt_acl:
858 case Opt_noacl:
859 printk("EXT3 (no)acl options not supported\n");
860 break;
861#endif
862 case Opt_reservation:
863 set_opt(sbi->s_mount_opt, RESERVATION);
864 break;
865 case Opt_noreservation:
866 clear_opt(sbi->s_mount_opt, RESERVATION);
867 break;
868 case Opt_journal_update:
869 /* @@@ FIXME */
870 /* Eventually we will want to be able to create
871 a journal file here. For now, only allow the
872 user to specify an existing inode to be the
873 journal file. */
874 if (is_remount) {
875 printk(KERN_ERR "EXT3-fs: cannot specify "
876 "journal on remount\n");
877 return 0;
878 }
879 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
880 break;
881 case Opt_journal_inum:
882 if (is_remount) {
883 printk(KERN_ERR "EXT3-fs: cannot specify "
884 "journal on remount\n");
885 return 0;
886 }
887 if (match_int(&args[0], &option))
888 return 0;
889 *inum = option;
890 break;
Johann Lombardi71b96252006-01-08 01:03:20 -0800891 case Opt_journal_dev:
892 if (is_remount) {
893 printk(KERN_ERR "EXT3-fs: cannot specify "
894 "journal on remount\n");
895 return 0;
896 }
897 if (match_int(&args[0], &option))
898 return 0;
899 *journal_devnum = option;
900 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901 case Opt_noload:
902 set_opt (sbi->s_mount_opt, NOLOAD);
903 break;
904 case Opt_commit:
905 if (match_int(&args[0], &option))
906 return 0;
907 if (option < 0)
908 return 0;
909 if (option == 0)
910 option = JBD_DEFAULT_MAX_COMMIT_AGE;
911 sbi->s_commit_interval = HZ * option;
912 break;
913 case Opt_data_journal:
914 data_opt = EXT3_MOUNT_JOURNAL_DATA;
915 goto datacheck;
916 case Opt_data_ordered:
917 data_opt = EXT3_MOUNT_ORDERED_DATA;
918 goto datacheck;
919 case Opt_data_writeback:
920 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
921 datacheck:
922 if (is_remount) {
923 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
924 != data_opt) {
925 printk(KERN_ERR
926 "EXT3-fs: cannot change data "
927 "mode on remount\n");
928 return 0;
929 }
930 } else {
931 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
932 sbi->s_mount_opt |= data_opt;
933 }
934 break;
935#ifdef CONFIG_QUOTA
936 case Opt_usrjquota:
937 qtype = USRQUOTA;
938 goto set_qf_name;
939 case Opt_grpjquota:
940 qtype = GRPQUOTA;
941set_qf_name:
942 if (sb_any_quota_enabled(sb)) {
943 printk(KERN_ERR
944 "EXT3-fs: Cannot change journalled "
945 "quota options when quota turned on.\n");
946 return 0;
947 }
948 qname = match_strdup(&args[0]);
949 if (!qname) {
950 printk(KERN_ERR
951 "EXT3-fs: not enough memory for "
952 "storing quotafile name.\n");
953 return 0;
954 }
955 if (sbi->s_qf_names[qtype] &&
956 strcmp(sbi->s_qf_names[qtype], qname)) {
957 printk(KERN_ERR
958 "EXT3-fs: %s quota file already "
959 "specified.\n", QTYPE2NAME(qtype));
960 kfree(qname);
961 return 0;
962 }
963 sbi->s_qf_names[qtype] = qname;
964 if (strchr(sbi->s_qf_names[qtype], '/')) {
965 printk(KERN_ERR
966 "EXT3-fs: quotafile must be on "
967 "filesystem root.\n");
968 kfree(sbi->s_qf_names[qtype]);
969 sbi->s_qf_names[qtype] = NULL;
970 return 0;
971 }
Jan Kara1f545872005-06-23 22:01:04 -0700972 set_opt(sbi->s_mount_opt, QUOTA);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 break;
974 case Opt_offusrjquota:
975 qtype = USRQUOTA;
976 goto clear_qf_name;
977 case Opt_offgrpjquota:
978 qtype = GRPQUOTA;
979clear_qf_name:
980 if (sb_any_quota_enabled(sb)) {
981 printk(KERN_ERR "EXT3-fs: Cannot change "
982 "journalled quota options when "
983 "quota turned on.\n");
984 return 0;
985 }
Jan Kara08c6a962005-07-12 13:58:28 -0700986 /*
987 * The space will be released later when all options
988 * are confirmed to be correct
989 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 sbi->s_qf_names[qtype] = NULL;
991 break;
992 case Opt_jqfmt_vfsold:
993 sbi->s_jquota_fmt = QFMT_VFS_OLD;
994 break;
995 case Opt_jqfmt_vfsv0:
996 sbi->s_jquota_fmt = QFMT_VFS_V0;
997 break;
Jan Kara1f545872005-06-23 22:01:04 -0700998 case Opt_quota:
Mark Bellon8fc27512005-09-06 15:16:54 -0700999 case Opt_usrquota:
Jan Kara1f545872005-06-23 22:01:04 -07001000 set_opt(sbi->s_mount_opt, QUOTA);
Mark Bellon8fc27512005-09-06 15:16:54 -07001001 set_opt(sbi->s_mount_opt, USRQUOTA);
1002 break;
1003 case Opt_grpquota:
1004 set_opt(sbi->s_mount_opt, QUOTA);
1005 set_opt(sbi->s_mount_opt, GRPQUOTA);
Jan Kara1f545872005-06-23 22:01:04 -07001006 break;
1007 case Opt_noquota:
1008 if (sb_any_quota_enabled(sb)) {
1009 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1010 "options when quota turned on.\n");
1011 return 0;
1012 }
1013 clear_opt(sbi->s_mount_opt, QUOTA);
Mark Bellon8fc27512005-09-06 15:16:54 -07001014 clear_opt(sbi->s_mount_opt, USRQUOTA);
1015 clear_opt(sbi->s_mount_opt, GRPQUOTA);
Jan Kara1f545872005-06-23 22:01:04 -07001016 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017#else
Mark Bellon8fc27512005-09-06 15:16:54 -07001018 case Opt_quota:
1019 case Opt_usrquota:
1020 case Opt_grpquota:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 case Opt_usrjquota:
1022 case Opt_grpjquota:
1023 case Opt_offusrjquota:
1024 case Opt_offgrpjquota:
1025 case Opt_jqfmt_vfsold:
1026 case Opt_jqfmt_vfsv0:
1027 printk(KERN_ERR
1028 "EXT3-fs: journalled quota options not "
1029 "supported.\n");
1030 break;
Jan Kara1f545872005-06-23 22:01:04 -07001031 case Opt_noquota:
1032 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033#endif
1034 case Opt_abort:
1035 set_opt(sbi->s_mount_opt, ABORT);
1036 break;
1037 case Opt_barrier:
1038 if (match_int(&args[0], &option))
1039 return 0;
1040 if (option)
1041 set_opt(sbi->s_mount_opt, BARRIER);
1042 else
1043 clear_opt(sbi->s_mount_opt, BARRIER);
1044 break;
1045 case Opt_ignore:
1046 break;
1047 case Opt_resize:
Jan Kara08c6a962005-07-12 13:58:28 -07001048 if (!is_remount) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 printk("EXT3-fs: resize option only available "
1050 "for remount\n");
1051 return 0;
1052 }
KAMBAROV, ZAURc7f17212005-06-28 20:45:11 -07001053 if (match_int(&args[0], &option) != 0)
1054 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 *n_blocks_count = option;
1056 break;
1057 case Opt_nobh:
1058 set_opt(sbi->s_mount_opt, NOBH);
1059 break;
Badari Pulavartyade1a292006-06-26 00:25:04 -07001060 case Opt_bh:
1061 clear_opt(sbi->s_mount_opt, NOBH);
1062 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 default:
1064 printk (KERN_ERR
1065 "EXT3-fs: Unrecognized mount option \"%s\" "
1066 "or missing value\n", p);
1067 return 0;
1068 }
1069 }
1070#ifdef CONFIG_QUOTA
Mark Bellon8fc27512005-09-06 15:16:54 -07001071 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1072 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1073 sbi->s_qf_names[USRQUOTA])
1074 clear_opt(sbi->s_mount_opt, USRQUOTA);
1075
1076 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1077 sbi->s_qf_names[GRPQUOTA])
1078 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1079
1080 if ((sbi->s_qf_names[USRQUOTA] &&
1081 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1082 (sbi->s_qf_names[GRPQUOTA] &&
1083 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1084 printk(KERN_ERR "EXT3-fs: old and new quota "
1085 "format mixing.\n");
1086 return 0;
1087 }
1088
1089 if (!sbi->s_jquota_fmt) {
1090 printk(KERN_ERR "EXT3-fs: journalled quota format "
1091 "not specified.\n");
1092 return 0;
1093 }
1094 } else {
1095 if (sbi->s_jquota_fmt) {
1096 printk(KERN_ERR "EXT3-fs: journalled quota format "
1097 "specified with no journalling "
1098 "enabled.\n");
1099 return 0;
1100 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 }
1102#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 return 1;
1104}
1105
1106static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1107 int read_only)
1108{
1109 struct ext3_sb_info *sbi = EXT3_SB(sb);
1110 int res = 0;
1111
1112 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1113 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1114 "forcing read-only mode\n");
1115 res = MS_RDONLY;
1116 }
1117 if (read_only)
1118 return res;
1119 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1120 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1121 "running e2fsck is recommended\n");
1122 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1123 printk (KERN_WARNING
1124 "EXT3-fs warning: mounting fs with errors, "
1125 "running e2fsck is recommended\n");
1126 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1127 le16_to_cpu(es->s_mnt_count) >=
1128 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1129 printk (KERN_WARNING
1130 "EXT3-fs warning: maximal mount count reached, "
1131 "running e2fsck is recommended\n");
1132 else if (le32_to_cpu(es->s_checkinterval) &&
1133 (le32_to_cpu(es->s_lastcheck) +
1134 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1135 printk (KERN_WARNING
1136 "EXT3-fs warning: checktime reached, "
1137 "running e2fsck is recommended\n");
1138#if 0
1139 /* @@@ We _will_ want to clear the valid bit if we find
1140 inconsistencies, to force a fsck at reboot. But for
1141 a plain journaled filesystem we can keep it set as
1142 valid forever! :) */
1143 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT3_VALID_FS);
1144#endif
1145 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1146 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1147 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1148 es->s_mtime = cpu_to_le32(get_seconds());
1149 ext3_update_dynamic_rev(sb);
1150 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1151
1152 ext3_commit_super(sb, es, 1);
1153 if (test_opt(sb, DEBUG))
1154 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1155 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1156 sb->s_blocksize,
1157 sbi->s_groups_count,
1158 EXT3_BLOCKS_PER_GROUP(sb),
1159 EXT3_INODES_PER_GROUP(sb),
1160 sbi->s_mount_opt);
1161
1162 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1163 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1164 char b[BDEVNAME_SIZE];
1165
1166 printk("external journal on %s\n",
1167 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1168 } else {
1169 printk("internal journal\n");
1170 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171 return res;
1172}
1173
1174/* Called at mount-time, super-block is locked */
1175static int ext3_check_descriptors (struct super_block * sb)
1176{
1177 struct ext3_sb_info *sbi = EXT3_SB(sb);
Eric Sandeen855565e2006-09-27 01:49:29 -07001178 ext3_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1179 ext3_fsblk_t last_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 struct ext3_group_desc * gdp = NULL;
1181 int desc_block = 0;
1182 int i;
1183
1184 ext3_debug ("Checking group descriptors");
1185
1186 for (i = 0; i < sbi->s_groups_count; i++)
1187 {
Eric Sandeen855565e2006-09-27 01:49:29 -07001188 if (i == sbi->s_groups_count - 1)
1189 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1190 else
1191 last_block = first_block +
1192 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1193
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194 if ((i % EXT3_DESC_PER_BLOCK(sb)) == 0)
1195 gdp = (struct ext3_group_desc *)
1196 sbi->s_group_desc[desc_block++]->b_data;
Eric Sandeen855565e2006-09-27 01:49:29 -07001197 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1198 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199 {
1200 ext3_error (sb, "ext3_check_descriptors",
1201 "Block bitmap for group %d"
1202 " not in group (block %lu)!",
1203 i, (unsigned long)
1204 le32_to_cpu(gdp->bg_block_bitmap));
1205 return 0;
1206 }
Eric Sandeen855565e2006-09-27 01:49:29 -07001207 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1208 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209 {
1210 ext3_error (sb, "ext3_check_descriptors",
1211 "Inode bitmap for group %d"
1212 " not in group (block %lu)!",
1213 i, (unsigned long)
1214 le32_to_cpu(gdp->bg_inode_bitmap));
1215 return 0;
1216 }
Eric Sandeen855565e2006-09-27 01:49:29 -07001217 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1218 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group >
1219 last_block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220 {
1221 ext3_error (sb, "ext3_check_descriptors",
1222 "Inode table for group %d"
1223 " not in group (block %lu)!",
1224 i, (unsigned long)
1225 le32_to_cpu(gdp->bg_inode_table));
1226 return 0;
1227 }
Eric Sandeen855565e2006-09-27 01:49:29 -07001228 first_block += EXT3_BLOCKS_PER_GROUP(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 gdp++;
1230 }
1231
1232 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1233 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1234 return 1;
1235}
1236
1237
1238/* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1239 * the superblock) which were deleted from all directories, but held open by
1240 * a process at the time of a crash. We walk the list and try to delete these
1241 * inodes at recovery time (only with a read-write filesystem).
1242 *
1243 * In order to keep the orphan inode chain consistent during traversal (in
1244 * case of crash during recovery), we link each inode into the superblock
1245 * orphan list_head and handle it the same way as an inode deletion during
1246 * normal operation (which journals the operations for us).
1247 *
1248 * We only do an iget() and an iput() on each inode, which is very safe if we
1249 * accidentally point at an in-use or already deleted inode. The worst that
1250 * can happen in this case is that we get a "bit already cleared" message from
1251 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1252 * e2fsck was run on this filesystem, and it must have already done the orphan
1253 * inode cleanup for us, so we can safely abort without any further action.
1254 */
1255static void ext3_orphan_cleanup (struct super_block * sb,
1256 struct ext3_super_block * es)
1257{
1258 unsigned int s_flags = sb->s_flags;
1259 int nr_orphans = 0, nr_truncates = 0;
1260#ifdef CONFIG_QUOTA
1261 int i;
1262#endif
1263 if (!es->s_last_orphan) {
1264 jbd_debug(4, "no orphan inodes to clean up\n");
1265 return;
1266 }
1267
1268 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1269 if (es->s_last_orphan)
1270 jbd_debug(1, "Errors on filesystem, "
1271 "clearing orphan list.\n");
1272 es->s_last_orphan = 0;
1273 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1274 return;
1275 }
1276
1277 if (s_flags & MS_RDONLY) {
1278 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1279 sb->s_id);
1280 sb->s_flags &= ~MS_RDONLY;
1281 }
1282#ifdef CONFIG_QUOTA
1283 /* Needed for iput() to work correctly and not trash data */
1284 sb->s_flags |= MS_ACTIVE;
1285 /* Turn on quotas so that they are updated correctly */
1286 for (i = 0; i < MAXQUOTAS; i++) {
1287 if (EXT3_SB(sb)->s_qf_names[i]) {
1288 int ret = ext3_quota_on_mount(sb, i);
1289 if (ret < 0)
1290 printk(KERN_ERR
1291 "EXT3-fs: Cannot turn on journalled "
1292 "quota: error %d\n", ret);
1293 }
1294 }
1295#endif
1296
1297 while (es->s_last_orphan) {
1298 struct inode *inode;
1299
1300 if (!(inode =
1301 ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1302 es->s_last_orphan = 0;
1303 break;
1304 }
1305
1306 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1307 DQUOT_INIT(inode);
1308 if (inode->i_nlink) {
1309 printk(KERN_DEBUG
Eric Sandeeneee194e2006-09-27 01:49:30 -07001310 "%s: truncating inode %lu to %Ld bytes\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311 __FUNCTION__, inode->i_ino, inode->i_size);
Eric Sandeeneee194e2006-09-27 01:49:30 -07001312 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313 inode->i_ino, inode->i_size);
1314 ext3_truncate(inode);
1315 nr_truncates++;
1316 } else {
1317 printk(KERN_DEBUG
Eric Sandeeneee194e2006-09-27 01:49:30 -07001318 "%s: deleting unreferenced inode %lu\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319 __FUNCTION__, inode->i_ino);
Eric Sandeeneee194e2006-09-27 01:49:30 -07001320 jbd_debug(2, "deleting unreferenced inode %lu\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321 inode->i_ino);
1322 nr_orphans++;
1323 }
1324 iput(inode); /* The delete magic happens here! */
1325 }
1326
1327#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1328
1329 if (nr_orphans)
1330 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1331 sb->s_id, PLURAL(nr_orphans));
1332 if (nr_truncates)
1333 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1334 sb->s_id, PLURAL(nr_truncates));
1335#ifdef CONFIG_QUOTA
1336 /* Turn quotas off */
1337 for (i = 0; i < MAXQUOTAS; i++) {
1338 if (sb_dqopt(sb)->files[i])
1339 vfs_quota_off(sb, i);
1340 }
1341#endif
1342 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1343}
1344
1345#define log2(n) ffz(~(n))
1346
1347/*
1348 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1349 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1350 * We need to be 1 filesystem block less than the 2^32 sector limit.
1351 */
1352static loff_t ext3_max_size(int bits)
1353{
1354 loff_t res = EXT3_NDIR_BLOCKS;
1355 /* This constant is calculated to be the largest file size for a
1356 * dense, 4k-blocksize file such that the total number of
1357 * sectors in the file, including data and all indirect blocks,
1358 * does not exceed 2^32. */
1359 const loff_t upper_limit = 0x1ff7fffd000LL;
1360
1361 res += 1LL << (bits-2);
1362 res += 1LL << (2*(bits-2));
1363 res += 1LL << (3*(bits-2));
1364 res <<= bits;
1365 if (res > upper_limit)
1366 res = upper_limit;
1367 return res;
1368}
1369
Mingming Cao43d23f92006-06-25 05:48:07 -07001370static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1371 ext3_fsblk_t logic_sb_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 int nr)
1373{
1374 struct ext3_sb_info *sbi = EXT3_SB(sb);
Mingming Cao43d23f92006-06-25 05:48:07 -07001375 unsigned long bg, first_meta_bg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 int has_super = 0;
1377
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1379
1380 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1381 nr < first_meta_bg)
1382 return (logic_sb_block + nr + 1);
1383 bg = sbi->s_desc_per_block * nr;
1384 if (ext3_bg_has_super(sb, bg))
1385 has_super = 1;
Mingming Cao43d23f92006-06-25 05:48:07 -07001386 return (has_super + ext3_group_first_block_no(sb, bg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387}
1388
1389
1390static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1391{
1392 struct buffer_head * bh;
1393 struct ext3_super_block *es = NULL;
1394 struct ext3_sb_info *sbi;
Mingming Cao43d23f92006-06-25 05:48:07 -07001395 ext3_fsblk_t block;
1396 ext3_fsblk_t sb_block = get_sb_block(&data);
1397 ext3_fsblk_t logic_sb_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398 unsigned long offset = 0;
Eric Sandeeneee194e2006-09-27 01:49:30 -07001399 unsigned int journal_inum = 0;
Johann Lombardi71b96252006-01-08 01:03:20 -08001400 unsigned long journal_devnum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401 unsigned long def_mount_opts;
1402 struct inode *root;
1403 int blocksize;
1404 int hblock;
1405 int db_count;
1406 int i;
1407 int needs_recovery;
1408 __le32 features;
1409
Panagiotis Issarisf8314dc2006-09-27 01:49:37 -07001410 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411 if (!sbi)
1412 return -ENOMEM;
1413 sb->s_fs_info = sbi;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 sbi->s_mount_opt = 0;
1415 sbi->s_resuid = EXT3_DEF_RESUID;
1416 sbi->s_resgid = EXT3_DEF_RESGID;
1417
1418 unlock_kernel();
1419
1420 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1421 if (!blocksize) {
1422 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1423 goto out_fail;
1424 }
1425
1426 /*
1427 * The ext3 superblock will not be buffer aligned for other than 1kB
1428 * block sizes. We need to calculate the offset from buffer start.
1429 */
1430 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1431 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1432 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1433 } else {
1434 logic_sb_block = sb_block;
1435 }
1436
1437 if (!(bh = sb_bread(sb, logic_sb_block))) {
1438 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1439 goto out_fail;
1440 }
1441 /*
1442 * Note: s_es must be initialized as soon as possible because
1443 * some ext3 macro-instructions depend on its value
1444 */
1445 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1446 sbi->s_es = es;
1447 sb->s_magic = le16_to_cpu(es->s_magic);
1448 if (sb->s_magic != EXT3_SUPER_MAGIC)
1449 goto cantfind_ext3;
1450
1451 /* Set defaults before we parse the mount options */
1452 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1453 if (def_mount_opts & EXT3_DEFM_DEBUG)
1454 set_opt(sbi->s_mount_opt, DEBUG);
1455 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1456 set_opt(sbi->s_mount_opt, GRPID);
1457 if (def_mount_opts & EXT3_DEFM_UID16)
1458 set_opt(sbi->s_mount_opt, NO_UID32);
1459 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1460 set_opt(sbi->s_mount_opt, XATTR_USER);
1461 if (def_mount_opts & EXT3_DEFM_ACL)
1462 set_opt(sbi->s_mount_opt, POSIX_ACL);
1463 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1464 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1465 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1466 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1467 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1468 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1469
1470 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1471 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1472 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_RO)
1473 set_opt(sbi->s_mount_opt, ERRORS_RO);
1474
1475 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1476 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1477
1478 set_opt(sbi->s_mount_opt, RESERVATION);
1479
Johann Lombardi71b96252006-01-08 01:03:20 -08001480 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1481 NULL, 0))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482 goto failed_mount;
1483
1484 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1485 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1486
1487 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1488 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1489 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1490 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001491 printk(KERN_WARNING
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492 "EXT3-fs warning: feature flags set on rev 0 fs, "
1493 "running e2fsck is recommended\n");
1494 /*
1495 * Check feature flags regardless of the revision level, since we
1496 * previously didn't change the revision level when setting the flags,
1497 * so there is a chance incompat flags are set on a rev 0 filesystem.
1498 */
1499 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1500 if (features) {
1501 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1502 "unsupported optional features (%x).\n",
1503 sb->s_id, le32_to_cpu(features));
1504 goto failed_mount;
1505 }
1506 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1507 if (!(sb->s_flags & MS_RDONLY) && features) {
1508 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1509 "unsupported optional features (%x).\n",
1510 sb->s_id, le32_to_cpu(features));
1511 goto failed_mount;
1512 }
1513 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1514
1515 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1516 blocksize > EXT3_MAX_BLOCK_SIZE) {
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001517 printk(KERN_ERR
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1519 blocksize, sb->s_id);
1520 goto failed_mount;
1521 }
1522
1523 hblock = bdev_hardsect_size(sb->s_bdev);
1524 if (sb->s_blocksize != blocksize) {
1525 /*
1526 * Make sure the blocksize for the filesystem is larger
1527 * than the hardware sectorsize for the machine.
1528 */
1529 if (blocksize < hblock) {
1530 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1531 "device blocksize %d.\n", blocksize, hblock);
1532 goto failed_mount;
1533 }
1534
1535 brelse (bh);
1536 sb_set_blocksize(sb, blocksize);
1537 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1538 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1539 bh = sb_bread(sb, logic_sb_block);
1540 if (!bh) {
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001541 printk(KERN_ERR
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542 "EXT3-fs: Can't read superblock on 2nd try.\n");
1543 goto failed_mount;
1544 }
1545 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1546 sbi->s_es = es;
1547 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001548 printk (KERN_ERR
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 "EXT3-fs: Magic mismatch, very weird !\n");
1550 goto failed_mount;
1551 }
1552 }
1553
1554 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1555
1556 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1557 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1558 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1559 } else {
1560 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1561 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1562 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1563 (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
1564 (sbi->s_inode_size > blocksize)) {
1565 printk (KERN_ERR
1566 "EXT3-fs: unsupported inode size: %d\n",
1567 sbi->s_inode_size);
1568 goto failed_mount;
1569 }
1570 }
1571 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1572 le32_to_cpu(es->s_log_frag_size);
1573 if (blocksize != sbi->s_frag_size) {
1574 printk(KERN_ERR
1575 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1576 sbi->s_frag_size, blocksize);
1577 goto failed_mount;
1578 }
1579 sbi->s_frags_per_block = 1;
1580 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1581 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1582 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1583 if (EXT3_INODE_SIZE(sb) == 0)
1584 goto cantfind_ext3;
1585 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1586 if (sbi->s_inodes_per_block == 0)
1587 goto cantfind_ext3;
1588 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1589 sbi->s_inodes_per_block;
1590 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1591 sbi->s_sbh = bh;
1592 sbi->s_mount_state = le16_to_cpu(es->s_state);
1593 sbi->s_addr_per_block_bits = log2(EXT3_ADDR_PER_BLOCK(sb));
1594 sbi->s_desc_per_block_bits = log2(EXT3_DESC_PER_BLOCK(sb));
1595 for (i=0; i < 4; i++)
1596 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1597 sbi->s_def_hash_version = es->s_def_hash_version;
1598
1599 if (sbi->s_blocks_per_group > blocksize * 8) {
1600 printk (KERN_ERR
1601 "EXT3-fs: #blocks per group too big: %lu\n",
1602 sbi->s_blocks_per_group);
1603 goto failed_mount;
1604 }
1605 if (sbi->s_frags_per_group > blocksize * 8) {
1606 printk (KERN_ERR
1607 "EXT3-fs: #fragments per group too big: %lu\n",
1608 sbi->s_frags_per_group);
1609 goto failed_mount;
1610 }
1611 if (sbi->s_inodes_per_group > blocksize * 8) {
1612 printk (KERN_ERR
1613 "EXT3-fs: #inodes per group too big: %lu\n",
1614 sbi->s_inodes_per_group);
1615 goto failed_mount;
1616 }
1617
Mingming Caofcd5df32006-06-25 05:47:50 -07001618 if (le32_to_cpu(es->s_blocks_count) >
1619 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1620 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1621 " too large to mount safely\n", sb->s_id);
1622 if (sizeof(sector_t) < 8)
1623 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1624 "enabled\n");
1625 goto failed_mount;
1626 }
1627
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1629 goto cantfind_ext3;
Eric Sandeen855565e2006-09-27 01:49:29 -07001630 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1631 le32_to_cpu(es->s_first_data_block) - 1)
1632 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1634 EXT3_DESC_PER_BLOCK(sb);
1635 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1636 GFP_KERNEL);
1637 if (sbi->s_group_desc == NULL) {
1638 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1639 goto failed_mount;
1640 }
1641
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 bgl_lock_init(&sbi->s_blockgroup_lock);
1643
1644 for (i = 0; i < db_count; i++) {
1645 block = descriptor_loc(sb, logic_sb_block, i);
1646 sbi->s_group_desc[i] = sb_bread(sb, block);
1647 if (!sbi->s_group_desc[i]) {
1648 printk (KERN_ERR "EXT3-fs: "
1649 "can't read group descriptor %d\n", i);
1650 db_count = i;
1651 goto failed_mount2;
1652 }
1653 }
1654 if (!ext3_check_descriptors (sb)) {
Theodore Ts'od2e5b132006-06-25 05:47:52 -07001655 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656 goto failed_mount2;
1657 }
1658 sbi->s_gdb_count = db_count;
1659 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1660 spin_lock_init(&sbi->s_next_gen_lock);
Mingming Cao0216bfc2006-06-23 02:05:41 -07001661
1662 percpu_counter_init(&sbi->s_freeblocks_counter,
1663 ext3_count_free_blocks(sb));
1664 percpu_counter_init(&sbi->s_freeinodes_counter,
1665 ext3_count_free_inodes(sb));
1666 percpu_counter_init(&sbi->s_dirs_counter,
1667 ext3_count_dirs(sb));
1668
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 /* per fileystem reservation list head & lock */
1670 spin_lock_init(&sbi->s_rsv_window_lock);
1671 sbi->s_rsv_window_root = RB_ROOT;
1672 /* Add a single, static dummy reservation to the start of the
1673 * reservation window list --- it gives us a placeholder for
1674 * append-at-start-of-list which makes the allocation logic
1675 * _much_ simpler. */
1676 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1677 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1678 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1679 sbi->s_rsv_window_head.rsv_goal_size = 0;
1680 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1681
1682 /*
1683 * set up enough so that it can read an inode
1684 */
1685 sb->s_op = &ext3_sops;
1686 sb->s_export_op = &ext3_export_ops;
1687 sb->s_xattr = ext3_xattr_handlers;
1688#ifdef CONFIG_QUOTA
1689 sb->s_qcop = &ext3_qctl_operations;
1690 sb->dq_op = &ext3_quota_operations;
1691#endif
1692 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1693
1694 sb->s_root = NULL;
1695
1696 needs_recovery = (es->s_last_orphan != 0 ||
1697 EXT3_HAS_INCOMPAT_FEATURE(sb,
1698 EXT3_FEATURE_INCOMPAT_RECOVER));
1699
1700 /*
1701 * The first inode we look at is the journal inode. Don't try
1702 * root first: it may be modified in the journal!
1703 */
1704 if (!test_opt(sb, NOLOAD) &&
1705 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
Johann Lombardi71b96252006-01-08 01:03:20 -08001706 if (ext3_load_journal(sb, es, journal_devnum))
Mingming Cao0216bfc2006-06-23 02:05:41 -07001707 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708 } else if (journal_inum) {
1709 if (ext3_create_journal(sb, es, journal_inum))
Mingming Cao0216bfc2006-06-23 02:05:41 -07001710 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711 } else {
1712 if (!silent)
1713 printk (KERN_ERR
1714 "ext3: No journal on filesystem on %s\n",
1715 sb->s_id);
Mingming Cao0216bfc2006-06-23 02:05:41 -07001716 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717 }
1718
1719 /* We have now updated the journal if required, so we can
1720 * validate the data journaling mode. */
1721 switch (test_opt(sb, DATA_FLAGS)) {
1722 case 0:
1723 /* No mode set, assume a default based on the journal
1724 capabilities: ORDERED_DATA if the journal can
1725 cope, else JOURNAL_DATA */
1726 if (journal_check_available_features
1727 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1728 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1729 else
1730 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1731 break;
1732
1733 case EXT3_MOUNT_ORDERED_DATA:
1734 case EXT3_MOUNT_WRITEBACK_DATA:
1735 if (!journal_check_available_features
1736 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1737 printk(KERN_ERR "EXT3-fs: Journal does not support "
1738 "requested data journaling mode\n");
Mingming Cao0216bfc2006-06-23 02:05:41 -07001739 goto failed_mount4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740 }
1741 default:
1742 break;
1743 }
1744
1745 if (test_opt(sb, NOBH)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1747 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1748 "its supported only with writeback mode\n");
1749 clear_opt(sbi->s_mount_opt, NOBH);
1750 }
1751 }
1752 /*
1753 * The journal_load will have done any necessary log recovery,
1754 * so we can safely mount the rest of the filesystem now.
1755 */
1756
1757 root = iget(sb, EXT3_ROOT_INO);
1758 sb->s_root = d_alloc_root(root);
1759 if (!sb->s_root) {
1760 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1761 iput(root);
Mingming Cao0216bfc2006-06-23 02:05:41 -07001762 goto failed_mount4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763 }
1764 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1765 dput(sb->s_root);
1766 sb->s_root = NULL;
1767 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
Mingming Cao0216bfc2006-06-23 02:05:41 -07001768 goto failed_mount4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769 }
1770
1771 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1772 /*
1773 * akpm: core read_super() calls in here with the superblock locked.
1774 * That deadlocks, because orphan cleanup needs to lock the superblock
1775 * in numerous places. Here we just pop the lock - it's relatively
1776 * harmless, because we are now ready to accept write_super() requests,
1777 * and aviro says that's the only reason for hanging onto the
1778 * superblock lock.
1779 */
1780 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1781 ext3_orphan_cleanup(sb, es);
1782 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1783 if (needs_recovery)
1784 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1785 ext3_mark_recovery_complete(sb, es);
1786 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1787 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1788 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1789 "writeback");
1790
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791 lock_kernel();
1792 return 0;
1793
1794cantfind_ext3:
1795 if (!silent)
1796 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1797 sb->s_id);
1798 goto failed_mount;
1799
Mingming Cao0216bfc2006-06-23 02:05:41 -07001800failed_mount4:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801 journal_destroy(sbi->s_journal);
Mingming Cao0216bfc2006-06-23 02:05:41 -07001802failed_mount3:
1803 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1804 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1805 percpu_counter_destroy(&sbi->s_dirs_counter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806failed_mount2:
1807 for (i = 0; i < db_count; i++)
1808 brelse(sbi->s_group_desc[i]);
1809 kfree(sbi->s_group_desc);
1810failed_mount:
1811#ifdef CONFIG_QUOTA
1812 for (i = 0; i < MAXQUOTAS; i++)
1813 kfree(sbi->s_qf_names[i]);
1814#endif
1815 ext3_blkdev_remove(sbi);
1816 brelse(bh);
1817out_fail:
1818 sb->s_fs_info = NULL;
1819 kfree(sbi);
1820 lock_kernel();
1821 return -EINVAL;
1822}
1823
1824/*
1825 * Setup any per-fs journal parameters now. We'll do this both on
1826 * initial mount, once the journal has been initialised but before we've
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001827 * done any recovery; and again on any subsequent remount.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828 */
1829static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1830{
1831 struct ext3_sb_info *sbi = EXT3_SB(sb);
1832
1833 if (sbi->s_commit_interval)
1834 journal->j_commit_interval = sbi->s_commit_interval;
1835 /* We could also set up an ext3-specific default for the commit
1836 * interval here, but for now we'll just fall back to the jbd
1837 * default. */
1838
1839 spin_lock(&journal->j_state_lock);
1840 if (test_opt(sb, BARRIER))
1841 journal->j_flags |= JFS_BARRIER;
1842 else
1843 journal->j_flags &= ~JFS_BARRIER;
1844 spin_unlock(&journal->j_state_lock);
1845}
1846
Eric Sandeeneee194e2006-09-27 01:49:30 -07001847static journal_t *ext3_get_journal(struct super_block *sb,
1848 unsigned int journal_inum)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001849{
1850 struct inode *journal_inode;
1851 journal_t *journal;
1852
1853 /* First, test for the existence of a valid inode on disk. Bad
1854 * things happen if we iget() an unused inode, as the subsequent
1855 * iput() will try to delete it. */
1856
1857 journal_inode = iget(sb, journal_inum);
1858 if (!journal_inode) {
1859 printk(KERN_ERR "EXT3-fs: no journal found.\n");
1860 return NULL;
1861 }
1862 if (!journal_inode->i_nlink) {
1863 make_bad_inode(journal_inode);
1864 iput(journal_inode);
1865 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
1866 return NULL;
1867 }
1868
1869 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1870 journal_inode, journal_inode->i_size);
1871 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
1872 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
1873 iput(journal_inode);
1874 return NULL;
1875 }
1876
1877 journal = journal_init_inode(journal_inode);
1878 if (!journal) {
1879 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
1880 iput(journal_inode);
1881 return NULL;
1882 }
1883 journal->j_private = sb;
1884 ext3_init_journal_params(sb, journal);
1885 return journal;
1886}
1887
1888static journal_t *ext3_get_dev_journal(struct super_block *sb,
1889 dev_t j_dev)
1890{
1891 struct buffer_head * bh;
1892 journal_t *journal;
Mingming Cao43d23f92006-06-25 05:48:07 -07001893 ext3_fsblk_t start;
Mingming Cao1c2bf372006-06-25 05:48:06 -07001894 ext3_fsblk_t len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895 int hblock, blocksize;
Mingming Cao43d23f92006-06-25 05:48:07 -07001896 ext3_fsblk_t sb_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001897 unsigned long offset;
1898 struct ext3_super_block * es;
1899 struct block_device *bdev;
1900
1901 bdev = ext3_blkdev_get(j_dev);
1902 if (bdev == NULL)
1903 return NULL;
1904
1905 if (bd_claim(bdev, sb)) {
1906 printk(KERN_ERR
1907 "EXT3: failed to claim external journal device.\n");
1908 blkdev_put(bdev);
1909 return NULL;
1910 }
1911
1912 blocksize = sb->s_blocksize;
1913 hblock = bdev_hardsect_size(bdev);
1914 if (blocksize < hblock) {
1915 printk(KERN_ERR
1916 "EXT3-fs: blocksize too small for journal device.\n");
1917 goto out_bdev;
1918 }
1919
1920 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
1921 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
1922 set_blocksize(bdev, blocksize);
1923 if (!(bh = __bread(bdev, sb_block, blocksize))) {
1924 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
1925 "external journal\n");
1926 goto out_bdev;
1927 }
1928
1929 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1930 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
1931 !(le32_to_cpu(es->s_feature_incompat) &
1932 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
1933 printk(KERN_ERR "EXT3-fs: external journal has "
1934 "bad superblock\n");
1935 brelse(bh);
1936 goto out_bdev;
1937 }
1938
1939 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
1940 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
1941 brelse(bh);
1942 goto out_bdev;
1943 }
1944
1945 len = le32_to_cpu(es->s_blocks_count);
1946 start = sb_block + 1;
1947 brelse(bh); /* we're done with the superblock */
1948
1949 journal = journal_init_dev(bdev, sb->s_bdev,
1950 start, len, blocksize);
1951 if (!journal) {
1952 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
1953 goto out_bdev;
1954 }
1955 journal->j_private = sb;
1956 ll_rw_block(READ, 1, &journal->j_sb_buffer);
1957 wait_on_buffer(journal->j_sb_buffer);
1958 if (!buffer_uptodate(journal->j_sb_buffer)) {
1959 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
1960 goto out_journal;
1961 }
1962 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
1963 printk(KERN_ERR "EXT3-fs: External journal has more than one "
1964 "user (unsupported) - %d\n",
1965 be32_to_cpu(journal->j_superblock->s_nr_users));
1966 goto out_journal;
1967 }
1968 EXT3_SB(sb)->journal_bdev = bdev;
1969 ext3_init_journal_params(sb, journal);
1970 return journal;
1971out_journal:
1972 journal_destroy(journal);
1973out_bdev:
1974 ext3_blkdev_put(bdev);
1975 return NULL;
1976}
1977
Johann Lombardi71b96252006-01-08 01:03:20 -08001978static int ext3_load_journal(struct super_block *sb,
1979 struct ext3_super_block *es,
1980 unsigned long journal_devnum)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981{
1982 journal_t *journal;
Eric Sandeeneee194e2006-09-27 01:49:30 -07001983 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
Johann Lombardi71b96252006-01-08 01:03:20 -08001984 dev_t journal_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001985 int err = 0;
1986 int really_read_only;
1987
Johann Lombardi71b96252006-01-08 01:03:20 -08001988 if (journal_devnum &&
1989 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
1990 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
1991 "numbers have changed\n");
1992 journal_dev = new_decode_dev(journal_devnum);
1993 } else
1994 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
1995
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996 really_read_only = bdev_read_only(sb->s_bdev);
1997
1998 /*
1999 * Are we loading a blank journal or performing recovery after a
2000 * crash? For recovery, we need to check in advance whether we
2001 * can get read-write access to the device.
2002 */
2003
2004 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2005 if (sb->s_flags & MS_RDONLY) {
2006 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2007 "required on readonly filesystem.\n");
2008 if (really_read_only) {
2009 printk(KERN_ERR "EXT3-fs: write access "
2010 "unavailable, cannot proceed.\n");
2011 return -EROFS;
2012 }
2013 printk (KERN_INFO "EXT3-fs: write access will "
2014 "be enabled during recovery.\n");
2015 }
2016 }
2017
2018 if (journal_inum && journal_dev) {
2019 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2020 "and inode journals!\n");
2021 return -EINVAL;
2022 }
2023
2024 if (journal_inum) {
2025 if (!(journal = ext3_get_journal(sb, journal_inum)))
2026 return -EINVAL;
2027 } else {
2028 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2029 return -EINVAL;
2030 }
2031
2032 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2033 err = journal_update_format(journal);
2034 if (err) {
2035 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2036 journal_destroy(journal);
2037 return err;
2038 }
2039 }
2040
2041 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2042 err = journal_wipe(journal, !really_read_only);
2043 if (!err)
2044 err = journal_load(journal);
2045
2046 if (err) {
2047 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2048 journal_destroy(journal);
2049 return err;
2050 }
2051
2052 EXT3_SB(sb)->s_journal = journal;
2053 ext3_clear_journal_err(sb, es);
Johann Lombardi71b96252006-01-08 01:03:20 -08002054
2055 if (journal_devnum &&
2056 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2057 es->s_journal_dev = cpu_to_le32(journal_devnum);
2058 sb->s_dirt = 1;
2059
2060 /* Make sure we flush the recovery flag to disk. */
2061 ext3_commit_super(sb, es, 1);
2062 }
2063
Linus Torvalds1da177e2005-04-16 15:20:36 -07002064 return 0;
2065}
2066
2067static int ext3_create_journal(struct super_block * sb,
2068 struct ext3_super_block * es,
Eric Sandeeneee194e2006-09-27 01:49:30 -07002069 unsigned int journal_inum)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002070{
2071 journal_t *journal;
2072
2073 if (sb->s_flags & MS_RDONLY) {
2074 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2075 "create journal.\n");
2076 return -EROFS;
2077 }
2078
2079 if (!(journal = ext3_get_journal(sb, journal_inum)))
2080 return -EINVAL;
2081
Eric Sandeeneee194e2006-09-27 01:49:30 -07002082 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002083 journal_inum);
2084
2085 if (journal_create(journal)) {
2086 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2087 journal_destroy(journal);
2088 return -EIO;
2089 }
2090
2091 EXT3_SB(sb)->s_journal = journal;
2092
2093 ext3_update_dynamic_rev(sb);
2094 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2095 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2096
2097 es->s_journal_inum = cpu_to_le32(journal_inum);
2098 sb->s_dirt = 1;
2099
2100 /* Make sure we flush the recovery flag to disk. */
2101 ext3_commit_super(sb, es, 1);
2102
2103 return 0;
2104}
2105
2106static void ext3_commit_super (struct super_block * sb,
2107 struct ext3_super_block * es,
2108 int sync)
2109{
2110 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2111
2112 if (!sbh)
2113 return;
2114 es->s_wtime = cpu_to_le32(get_seconds());
2115 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2116 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2117 BUFFER_TRACE(sbh, "marking dirty");
2118 mark_buffer_dirty(sbh);
2119 if (sync)
2120 sync_dirty_buffer(sbh);
2121}
2122
2123
2124/*
2125 * Have we just finished recovery? If so, and if we are mounting (or
2126 * remounting) the filesystem readonly, then we will end up with a
2127 * consistent fs on disk. Record that fact.
2128 */
2129static void ext3_mark_recovery_complete(struct super_block * sb,
2130 struct ext3_super_block * es)
2131{
2132 journal_t *journal = EXT3_SB(sb)->s_journal;
2133
2134 journal_lock_updates(journal);
2135 journal_flush(journal);
2136 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2137 sb->s_flags & MS_RDONLY) {
2138 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2139 sb->s_dirt = 0;
2140 ext3_commit_super(sb, es, 1);
2141 }
2142 journal_unlock_updates(journal);
2143}
2144
2145/*
2146 * If we are mounting (or read-write remounting) a filesystem whose journal
2147 * has recorded an error from a previous lifetime, move that error to the
2148 * main filesystem now.
2149 */
2150static void ext3_clear_journal_err(struct super_block * sb,
2151 struct ext3_super_block * es)
2152{
2153 journal_t *journal;
2154 int j_errno;
2155 const char *errstr;
2156
2157 journal = EXT3_SB(sb)->s_journal;
2158
2159 /*
2160 * Now check for any error status which may have been recorded in the
2161 * journal by a prior ext3_error() or ext3_abort()
2162 */
2163
2164 j_errno = journal_errno(journal);
2165 if (j_errno) {
2166 char nbuf[16];
2167
2168 errstr = ext3_decode_error(sb, j_errno, nbuf);
2169 ext3_warning(sb, __FUNCTION__, "Filesystem error recorded "
2170 "from previous mount: %s", errstr);
2171 ext3_warning(sb, __FUNCTION__, "Marking fs in need of "
2172 "filesystem check.");
2173
2174 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2175 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2176 ext3_commit_super (sb, es, 1);
2177
2178 journal_clear_err(journal);
2179 }
2180}
2181
2182/*
2183 * Force the running and committing transactions to commit,
2184 * and wait on the commit.
2185 */
2186int ext3_force_commit(struct super_block *sb)
2187{
2188 journal_t *journal;
2189 int ret;
2190
2191 if (sb->s_flags & MS_RDONLY)
2192 return 0;
2193
2194 journal = EXT3_SB(sb)->s_journal;
2195 sb->s_dirt = 0;
2196 ret = ext3_journal_force_commit(journal);
2197 return ret;
2198}
2199
2200/*
2201 * Ext3 always journals updates to the superblock itself, so we don't
2202 * have to propagate any other updates to the superblock on disk at this
2203 * point. Just start an async writeback to get the buffers on their way
2204 * to the disk.
2205 *
2206 * This implicitly triggers the writebehind on sync().
2207 */
2208
2209static void ext3_write_super (struct super_block * sb)
2210{
Ingo Molnar7892f2f2006-01-09 15:59:25 -08002211 if (mutex_trylock(&sb->s_lock) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002212 BUG();
2213 sb->s_dirt = 0;
2214}
2215
2216static int ext3_sync_fs(struct super_block *sb, int wait)
2217{
2218 tid_t target;
2219
2220 sb->s_dirt = 0;
2221 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2222 if (wait)
2223 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2224 }
2225 return 0;
2226}
2227
2228/*
2229 * LVM calls this function before a (read-only) snapshot is created. This
2230 * gives us a chance to flush the journal completely and mark the fs clean.
2231 */
2232static void ext3_write_super_lockfs(struct super_block *sb)
2233{
2234 sb->s_dirt = 0;
2235
2236 if (!(sb->s_flags & MS_RDONLY)) {
2237 journal_t *journal = EXT3_SB(sb)->s_journal;
2238
2239 /* Now we set up the journal barrier. */
2240 journal_lock_updates(journal);
2241 journal_flush(journal);
2242
2243 /* Journal blocked and flushed, clear needs_recovery flag. */
2244 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2245 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2246 }
2247}
2248
2249/*
2250 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2251 * flag here, even though the filesystem is not technically dirty yet.
2252 */
2253static void ext3_unlockfs(struct super_block *sb)
2254{
2255 if (!(sb->s_flags & MS_RDONLY)) {
2256 lock_super(sb);
2257 /* Reser the needs_recovery flag before the fs is unlocked. */
2258 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2259 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2260 unlock_super(sb);
2261 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2262 }
2263}
2264
2265static int ext3_remount (struct super_block * sb, int * flags, char * data)
2266{
2267 struct ext3_super_block * es;
2268 struct ext3_sb_info *sbi = EXT3_SB(sb);
Mingming Cao43d23f92006-06-25 05:48:07 -07002269 ext3_fsblk_t n_blocks_count = 0;
Jan Kara08c6a962005-07-12 13:58:28 -07002270 unsigned long old_sb_flags;
2271 struct ext3_mount_options old_opts;
2272 int err;
2273#ifdef CONFIG_QUOTA
2274 int i;
2275#endif
2276
2277 /* Store the original options */
2278 old_sb_flags = sb->s_flags;
2279 old_opts.s_mount_opt = sbi->s_mount_opt;
2280 old_opts.s_resuid = sbi->s_resuid;
2281 old_opts.s_resgid = sbi->s_resgid;
2282 old_opts.s_commit_interval = sbi->s_commit_interval;
2283#ifdef CONFIG_QUOTA
2284 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2285 for (i = 0; i < MAXQUOTAS; i++)
2286 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2287#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002288
2289 /*
2290 * Allow the "check" option to be passed as a remount option.
2291 */
Johann Lombardi71b96252006-01-08 01:03:20 -08002292 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
Jan Kara08c6a962005-07-12 13:58:28 -07002293 err = -EINVAL;
2294 goto restore_opts;
2295 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002296
2297 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2298 ext3_abort(sb, __FUNCTION__, "Abort forced by user");
2299
2300 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2301 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2302
2303 es = sbi->s_es;
2304
2305 ext3_init_journal_params(sb, sbi->s_journal);
2306
2307 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2308 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
Jan Kara08c6a962005-07-12 13:58:28 -07002309 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2310 err = -EROFS;
2311 goto restore_opts;
2312 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002313
2314 if (*flags & MS_RDONLY) {
2315 /*
2316 * First of all, the unconditional stuff we have to do
2317 * to disable replay of the journal when we next remount
2318 */
2319 sb->s_flags |= MS_RDONLY;
2320
2321 /*
2322 * OK, test if we are remounting a valid rw partition
2323 * readonly, and if so set the rdonly flag and then
2324 * mark the partition as valid again.
2325 */
2326 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2327 (sbi->s_mount_state & EXT3_VALID_FS))
2328 es->s_state = cpu_to_le16(sbi->s_mount_state);
2329
2330 ext3_mark_recovery_complete(sb, es);
2331 } else {
2332 __le32 ret;
2333 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2334 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2335 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2336 "remount RDWR because of unsupported "
2337 "optional features (%x).\n",
2338 sb->s_id, le32_to_cpu(ret));
Jan Kara08c6a962005-07-12 13:58:28 -07002339 err = -EROFS;
2340 goto restore_opts;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002341 }
2342 /*
2343 * Mounting a RDONLY partition read-write, so reread
2344 * and store the current valid flag. (It may have
2345 * been changed by e2fsck since we originally mounted
2346 * the partition.)
2347 */
2348 ext3_clear_journal_err(sb, es);
2349 sbi->s_mount_state = le16_to_cpu(es->s_state);
Dave Kleikampa4e4de32006-09-27 01:49:36 -07002350 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
Jan Kara08c6a962005-07-12 13:58:28 -07002351 goto restore_opts;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002352 if (!ext3_setup_super (sb, es, 0))
2353 sb->s_flags &= ~MS_RDONLY;
2354 }
2355 }
Jan Kara08c6a962005-07-12 13:58:28 -07002356#ifdef CONFIG_QUOTA
2357 /* Release old quota file names */
2358 for (i = 0; i < MAXQUOTAS; i++)
2359 if (old_opts.s_qf_names[i] &&
2360 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2361 kfree(old_opts.s_qf_names[i]);
2362#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002363 return 0;
Jan Kara08c6a962005-07-12 13:58:28 -07002364restore_opts:
2365 sb->s_flags = old_sb_flags;
2366 sbi->s_mount_opt = old_opts.s_mount_opt;
2367 sbi->s_resuid = old_opts.s_resuid;
2368 sbi->s_resgid = old_opts.s_resgid;
2369 sbi->s_commit_interval = old_opts.s_commit_interval;
2370#ifdef CONFIG_QUOTA
2371 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2372 for (i = 0; i < MAXQUOTAS; i++) {
2373 if (sbi->s_qf_names[i] &&
2374 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2375 kfree(sbi->s_qf_names[i]);
2376 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2377 }
2378#endif
2379 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002380}
2381
David Howells726c3342006-06-23 02:02:58 -07002382static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002383{
David Howells726c3342006-06-23 02:02:58 -07002384 struct super_block *sb = dentry->d_sb;
Alex Tomas09fe3162006-03-24 03:16:16 -08002385 struct ext3_sb_info *sbi = EXT3_SB(sb);
2386 struct ext3_super_block *es = sbi->s_es;
Mingming Cao43d23f92006-06-25 05:48:07 -07002387 ext3_fsblk_t overhead;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002388 int i;
2389
2390 if (test_opt (sb, MINIX_DF))
2391 overhead = 0;
2392 else {
2393 unsigned long ngroups;
2394 ngroups = EXT3_SB(sb)->s_groups_count;
2395 smp_rmb();
2396
2397 /*
2398 * Compute the overhead (FS structures)
2399 */
2400
2401 /*
2402 * All of the blocks before first_data_block are
2403 * overhead
2404 */
2405 overhead = le32_to_cpu(es->s_first_data_block);
2406
2407 /*
2408 * Add the overhead attributed to the superblock and
2409 * block group descriptors. If the sparse superblocks
2410 * feature is turned on, then not all groups have this.
2411 */
2412 for (i = 0; i < ngroups; i++) {
2413 overhead += ext3_bg_has_super(sb, i) +
2414 ext3_bg_num_gdb(sb, i);
2415 cond_resched();
2416 }
2417
2418 /*
2419 * Every block group has an inode bitmap, a block
2420 * bitmap, and an inode table.
2421 */
2422 overhead += (ngroups * (2 + EXT3_SB(sb)->s_itb_per_group));
2423 }
2424
2425 buf->f_type = EXT3_SUPER_MAGIC;
2426 buf->f_bsize = sb->s_blocksize;
2427 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - overhead;
Alex Tomas09fe3162006-03-24 03:16:16 -08002428 buf->f_bfree = percpu_counter_sum(&sbi->s_freeblocks_counter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002429 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2430 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2431 buf->f_bavail = 0;
2432 buf->f_files = le32_to_cpu(es->s_inodes_count);
Alex Tomas09fe3162006-03-24 03:16:16 -08002433 buf->f_ffree = percpu_counter_sum(&sbi->s_freeinodes_counter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434 buf->f_namelen = EXT3_NAME_LEN;
2435 return 0;
2436}
2437
2438/* Helper function for writing quotas on sync - we need to start transaction before quota file
2439 * is locked for write. Otherwise the are possible deadlocks:
2440 * Process 1 Process 2
2441 * ext3_create() quota_sync()
2442 * journal_start() write_dquot()
Ingo Molnard3be9152006-03-23 03:00:29 -08002443 * DQUOT_INIT() down(dqio_mutex)
2444 * down(dqio_mutex) journal_start()
Linus Torvalds1da177e2005-04-16 15:20:36 -07002445 *
2446 */
2447
2448#ifdef CONFIG_QUOTA
2449
2450static inline struct inode *dquot_to_inode(struct dquot *dquot)
2451{
2452 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2453}
2454
2455static int ext3_dquot_initialize(struct inode *inode, int type)
2456{
2457 handle_t *handle;
2458 int ret, err;
2459
2460 /* We may create quota structure so we need to reserve enough blocks */
Jan Kara1f545872005-06-23 22:01:04 -07002461 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002462 if (IS_ERR(handle))
2463 return PTR_ERR(handle);
2464 ret = dquot_initialize(inode, type);
2465 err = ext3_journal_stop(handle);
2466 if (!ret)
2467 ret = err;
2468 return ret;
2469}
2470
2471static int ext3_dquot_drop(struct inode *inode)
2472{
2473 handle_t *handle;
2474 int ret, err;
2475
2476 /* We may delete quota structure so we need to reserve enough blocks */
Jan Kara1f545872005-06-23 22:01:04 -07002477 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002478 if (IS_ERR(handle))
2479 return PTR_ERR(handle);
2480 ret = dquot_drop(inode);
2481 err = ext3_journal_stop(handle);
2482 if (!ret)
2483 ret = err;
2484 return ret;
2485}
2486
2487static int ext3_write_dquot(struct dquot *dquot)
2488{
2489 int ret, err;
2490 handle_t *handle;
2491 struct inode *inode;
2492
2493 inode = dquot_to_inode(dquot);
2494 handle = ext3_journal_start(inode,
Jan Kara1f545872005-06-23 22:01:04 -07002495 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002496 if (IS_ERR(handle))
2497 return PTR_ERR(handle);
2498 ret = dquot_commit(dquot);
2499 err = ext3_journal_stop(handle);
2500 if (!ret)
2501 ret = err;
2502 return ret;
2503}
2504
2505static int ext3_acquire_dquot(struct dquot *dquot)
2506{
2507 int ret, err;
2508 handle_t *handle;
2509
2510 handle = ext3_journal_start(dquot_to_inode(dquot),
Jan Kara1f545872005-06-23 22:01:04 -07002511 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002512 if (IS_ERR(handle))
2513 return PTR_ERR(handle);
2514 ret = dquot_acquire(dquot);
2515 err = ext3_journal_stop(handle);
2516 if (!ret)
2517 ret = err;
2518 return ret;
2519}
2520
2521static int ext3_release_dquot(struct dquot *dquot)
2522{
2523 int ret, err;
2524 handle_t *handle;
2525
2526 handle = ext3_journal_start(dquot_to_inode(dquot),
Jan Kara1f545872005-06-23 22:01:04 -07002527 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002528 if (IS_ERR(handle))
2529 return PTR_ERR(handle);
2530 ret = dquot_release(dquot);
2531 err = ext3_journal_stop(handle);
2532 if (!ret)
2533 ret = err;
2534 return ret;
2535}
2536
2537static int ext3_mark_dquot_dirty(struct dquot *dquot)
2538{
2539 /* Are we journalling quotas? */
2540 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2541 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2542 dquot_mark_dquot_dirty(dquot);
2543 return ext3_write_dquot(dquot);
2544 } else {
2545 return dquot_mark_dquot_dirty(dquot);
2546 }
2547}
2548
2549static int ext3_write_info(struct super_block *sb, int type)
2550{
2551 int ret, err;
2552 handle_t *handle;
2553
2554 /* Data block + inode block */
2555 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2556 if (IS_ERR(handle))
2557 return PTR_ERR(handle);
2558 ret = dquot_commit_info(sb, type);
2559 err = ext3_journal_stop(handle);
2560 if (!ret)
2561 ret = err;
2562 return ret;
2563}
2564
2565/*
2566 * Turn on quotas during mount time - we need to find
2567 * the quota file and such...
2568 */
2569static int ext3_quota_on_mount(struct super_block *sb, int type)
2570{
Christoph Hellwig84de8562005-06-23 00:09:16 -07002571 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2572 EXT3_SB(sb)->s_jquota_fmt, type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002573}
2574
2575/*
2576 * Standard function to be called on quota_on
2577 */
2578static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2579 char *path)
2580{
2581 int err;
2582 struct nameidata nd;
2583
Jan Kara1f545872005-06-23 22:01:04 -07002584 if (!test_opt(sb, QUOTA))
2585 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002586 /* Not journalling quota? */
2587 if (!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
2588 !EXT3_SB(sb)->s_qf_names[GRPQUOTA])
2589 return vfs_quota_on(sb, type, format_id, path);
2590 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2591 if (err)
2592 return err;
2593 /* Quotafile not on the same filesystem? */
2594 if (nd.mnt->mnt_sb != sb) {
2595 path_release(&nd);
2596 return -EXDEV;
2597 }
2598 /* Quotafile not of fs root? */
2599 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2600 printk(KERN_WARNING
2601 "EXT3-fs: Quota file not on filesystem root. "
2602 "Journalled quota will not work.\n");
2603 path_release(&nd);
2604 return vfs_quota_on(sb, type, format_id, path);
2605}
2606
2607/* Read data from quotafile - avoid pagecache and such because we cannot afford
2608 * acquiring the locks... As quota files are never truncated and quota code
2609 * itself serializes the operations (and noone else should touch the files)
2610 * we don't have to be afraid of races */
2611static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2612 size_t len, loff_t off)
2613{
2614 struct inode *inode = sb_dqopt(sb)->files[type];
2615 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2616 int err = 0;
2617 int offset = off & (sb->s_blocksize - 1);
2618 int tocopy;
2619 size_t toread;
2620 struct buffer_head *bh;
2621 loff_t i_size = i_size_read(inode);
2622
2623 if (off > i_size)
2624 return 0;
2625 if (off+len > i_size)
2626 len = i_size-off;
2627 toread = len;
2628 while (toread > 0) {
2629 tocopy = sb->s_blocksize - offset < toread ?
2630 sb->s_blocksize - offset : toread;
2631 bh = ext3_bread(NULL, inode, blk, 0, &err);
2632 if (err)
2633 return err;
2634 if (!bh) /* A hole? */
2635 memset(data, 0, tocopy);
2636 else
2637 memcpy(data, bh->b_data+offset, tocopy);
2638 brelse(bh);
2639 offset = 0;
2640 toread -= tocopy;
2641 data += tocopy;
2642 blk++;
2643 }
2644 return len;
2645}
2646
2647/* Write to quotafile (we know the transaction is already started and has
2648 * enough credits) */
2649static ssize_t ext3_quota_write(struct super_block *sb, int type,
2650 const char *data, size_t len, loff_t off)
2651{
2652 struct inode *inode = sb_dqopt(sb)->files[type];
2653 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2654 int err = 0;
2655 int offset = off & (sb->s_blocksize - 1);
2656 int tocopy;
2657 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2658 size_t towrite = len;
2659 struct buffer_head *bh;
2660 handle_t *handle = journal_current_handle();
2661
Arjan van de Ven5c81a412006-07-03 00:25:20 -07002662 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002663 while (towrite > 0) {
2664 tocopy = sb->s_blocksize - offset < towrite ?
2665 sb->s_blocksize - offset : towrite;
2666 bh = ext3_bread(handle, inode, blk, 1, &err);
2667 if (!bh)
2668 goto out;
2669 if (journal_quota) {
2670 err = ext3_journal_get_write_access(handle, bh);
2671 if (err) {
2672 brelse(bh);
2673 goto out;
2674 }
2675 }
2676 lock_buffer(bh);
2677 memcpy(bh->b_data+offset, data, tocopy);
2678 flush_dcache_page(bh->b_page);
2679 unlock_buffer(bh);
2680 if (journal_quota)
2681 err = ext3_journal_dirty_metadata(handle, bh);
2682 else {
2683 /* Always do at least ordered writes for quotas */
2684 err = ext3_journal_dirty_data(handle, bh);
2685 mark_buffer_dirty(bh);
2686 }
2687 brelse(bh);
2688 if (err)
2689 goto out;
2690 offset = 0;
2691 towrite -= tocopy;
2692 data += tocopy;
2693 blk++;
2694 }
2695out:
2696 if (len == towrite)
2697 return err;
2698 if (inode->i_size < off+len-towrite) {
2699 i_size_write(inode, off+len-towrite);
2700 EXT3_I(inode)->i_disksize = inode->i_size;
2701 }
2702 inode->i_version++;
2703 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2704 ext3_mark_inode_dirty(handle, inode);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002705 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002706 return len - towrite;
2707}
2708
2709#endif
2710
David Howells454e2392006-06-23 02:02:57 -07002711static int ext3_get_sb(struct file_system_type *fs_type,
2712 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002713{
David Howells454e2392006-06-23 02:02:57 -07002714 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002715}
2716
2717static struct file_system_type ext3_fs_type = {
2718 .owner = THIS_MODULE,
2719 .name = "ext3",
2720 .get_sb = ext3_get_sb,
2721 .kill_sb = kill_block_super,
2722 .fs_flags = FS_REQUIRES_DEV,
2723};
2724
2725static int __init init_ext3_fs(void)
2726{
2727 int err = init_ext3_xattr();
2728 if (err)
2729 return err;
2730 err = init_inodecache();
2731 if (err)
2732 goto out1;
2733 err = register_filesystem(&ext3_fs_type);
2734 if (err)
2735 goto out;
2736 return 0;
2737out:
2738 destroy_inodecache();
2739out1:
Dave Kleikampe9ad5622006-09-27 01:49:35 -07002740 exit_ext3_xattr();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002741 return err;
2742}
2743
2744static void __exit exit_ext3_fs(void)
2745{
2746 unregister_filesystem(&ext3_fs_type);
2747 destroy_inodecache();
2748 exit_ext3_xattr();
2749}
2750
2751MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2752MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2753MODULE_LICENSE("GPL");
2754module_init(init_ext3_fs)
2755module_exit(exit_ext3_fs)