blob: 883ff965d9844206b1049c9279e12cf9239b9b85 [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>
Christoph Hellwiga5694252007-07-17 04:04:28 -070032#include <linux/exportfs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#include <linux/vfs.h>
34#include <linux/random.h>
35#include <linux/mount.h>
36#include <linux/namei.h>
37#include <linux/quotaops.h>
Mark Bellon8fc27512005-09-06 15:16:54 -070038#include <linux/seq_file.h>
vignesh babu3fc74262007-07-15 23:41:17 -070039#include <linux/log2.h>
Ben Dooks381be252005-10-30 15:02:56 -080040
Linus Torvalds1da177e2005-04-16 15:20:36 -070041#include <asm/uaccess.h>
Ben Dooks381be252005-10-30 15:02:56 -080042
Linus Torvalds1da177e2005-04-16 15:20:36 -070043#include "xattr.h"
44#include "acl.h"
Ben Dooks381be252005-10-30 15:02:56 -080045#include "namei.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070046
Johann Lombardi71b96252006-01-08 01:03:20 -080047static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
48 unsigned long journal_devnum);
Linus Torvalds1da177e2005-04-16 15:20:36 -070049static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
Eric Sandeeneee194e2006-09-27 01:49:30 -070050 unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -070051static void ext3_commit_super (struct super_block * sb,
52 struct ext3_super_block * es,
53 int sync);
54static void ext3_mark_recovery_complete(struct super_block * sb,
55 struct ext3_super_block * es);
56static void ext3_clear_journal_err(struct super_block * sb,
57 struct ext3_super_block * es);
58static int ext3_sync_fs(struct super_block *sb, int wait);
59static const char *ext3_decode_error(struct super_block * sb, int errno,
60 char nbuf[16]);
61static int ext3_remount (struct super_block * sb, int * flags, char * data);
David Howells726c3342006-06-23 02:02:58 -070062static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -070063static void ext3_unlockfs(struct super_block *sb);
64static void ext3_write_super (struct super_block * sb);
65static void ext3_write_super_lockfs(struct super_block *sb);
66
Mingming Caoae6ddcc2006-09-27 01:49:27 -070067/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 * Wrappers for journal_start/end.
69 *
70 * The only special thing we need to do here is to make sure that all
71 * journal_end calls result in the superblock being marked dirty, so
72 * that sync() will call the filesystem's write_super callback if
Mingming Caoae6ddcc2006-09-27 01:49:27 -070073 * appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -070074 */
75handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
76{
77 journal_t *journal;
78
79 if (sb->s_flags & MS_RDONLY)
80 return ERR_PTR(-EROFS);
81
82 /* Special case here: if the journal has aborted behind our
83 * backs (eg. EIO in the commit thread), then we still need to
84 * take the FS itself readonly cleanly. */
85 journal = EXT3_SB(sb)->s_journal;
86 if (is_journal_aborted(journal)) {
87 ext3_abort(sb, __FUNCTION__,
88 "Detected aborted journal");
89 return ERR_PTR(-EROFS);
90 }
91
92 return journal_start(journal, nblocks);
93}
94
Mingming Caoae6ddcc2006-09-27 01:49:27 -070095/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070096 * The only special thing we need to do here is to make sure that all
97 * journal_stop calls result in the superblock being marked dirty, so
98 * that sync() will call the filesystem's write_super callback if
Mingming Caoae6ddcc2006-09-27 01:49:27 -070099 * appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 */
101int __ext3_journal_stop(const char *where, handle_t *handle)
102{
103 struct super_block *sb;
104 int err;
105 int rc;
106
107 sb = handle->h_transaction->t_journal->j_private;
108 err = handle->h_err;
109 rc = journal_stop(handle);
110
111 if (!err)
112 err = rc;
113 if (err)
114 __ext3_std_error(sb, where, err);
115 return err;
116}
117
118void ext3_journal_abort_handle(const char *caller, const char *err_fn,
119 struct buffer_head *bh, handle_t *handle, int err)
120{
121 char nbuf[16];
122 const char *errstr = ext3_decode_error(NULL, err, nbuf);
123
124 if (bh)
125 BUFFER_TRACE(bh, "abort");
126
127 if (!handle->h_err)
128 handle->h_err = err;
129
130 if (is_handle_aborted(handle))
131 return;
132
133 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
134 caller, errstr, err_fn);
135
136 journal_abort_handle(handle);
137}
138
139/* Deal with the reporting of failure conditions on a filesystem such as
140 * inconsistencies detected or read IO failures.
141 *
142 * On ext2, we can store the error state of the filesystem in the
143 * superblock. That is not possible on ext3, because we may have other
144 * write ordering constraints on the superblock which prevent us from
145 * writing it out straight away; and given that the journal is about to
146 * be aborted, we can't rely on the current, or future, transactions to
147 * write out the superblock safely.
148 *
149 * We'll just use the journal_abort() error code to record an error in
150 * the journal instead. On recovery, the journal will compain about
151 * that error until we've noted it down and cleared it.
152 */
153
154static void ext3_handle_error(struct super_block *sb)
155{
156 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
157
158 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
159 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
160
161 if (sb->s_flags & MS_RDONLY)
162 return;
163
Vasily Averin7543fc72006-09-27 01:49:33 -0700164 if (!test_opt (sb, ERRORS_CONT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165 journal_t *journal = EXT3_SB(sb)->s_journal;
166
167 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
168 if (journal)
169 journal_abort(journal, -EIO);
170 }
Vasily Averin7543fc72006-09-27 01:49:33 -0700171 if (test_opt (sb, ERRORS_RO)) {
172 printk (KERN_CRIT "Remounting filesystem read-only\n");
173 sb->s_flags |= MS_RDONLY;
174 }
175 ext3_commit_super(sb, es, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176 if (test_opt(sb, ERRORS_PANIC))
177 panic("EXT3-fs (device %s): panic forced after error\n",
178 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179}
180
181void ext3_error (struct super_block * sb, const char * function,
182 const char * fmt, ...)
183{
184 va_list args;
185
186 va_start(args, fmt);
187 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
188 vprintk(fmt, args);
189 printk("\n");
190 va_end(args);
191
192 ext3_handle_error(sb);
193}
194
195static const char *ext3_decode_error(struct super_block * sb, int errno,
196 char nbuf[16])
197{
198 char *errstr = NULL;
199
200 switch (errno) {
201 case -EIO:
202 errstr = "IO failure";
203 break;
204 case -ENOMEM:
205 errstr = "Out of memory";
206 break;
207 case -EROFS:
208 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
209 errstr = "Journal has aborted";
210 else
211 errstr = "Readonly filesystem";
212 break;
213 default:
214 /* If the caller passed in an extra buffer for unknown
215 * errors, textualise them now. Else we just return
216 * NULL. */
217 if (nbuf) {
218 /* Check for truncated error codes... */
219 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
220 errstr = nbuf;
221 }
222 break;
223 }
224
225 return errstr;
226}
227
228/* __ext3_std_error decodes expected errors from journaling functions
229 * automatically and invokes the appropriate error response. */
230
231void __ext3_std_error (struct super_block * sb, const char * function,
232 int errno)
233{
234 char nbuf[16];
Stephen Tweedie30121622005-05-18 11:47:17 -0400235 const char *errstr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236
Stephen Tweedie30121622005-05-18 11:47:17 -0400237 /* Special case: if the error is EROFS, and we're not already
238 * inside a transaction, then there's really no point in logging
239 * an error. */
240 if (errno == -EROFS && journal_current_handle() == NULL &&
241 (sb->s_flags & MS_RDONLY))
242 return;
243
244 errstr = ext3_decode_error(sb, errno, nbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
246 sb->s_id, function, errstr);
247
248 ext3_handle_error(sb);
249}
250
251/*
252 * ext3_abort is a much stronger failure handler than ext3_error. The
253 * abort function may be used to deal with unrecoverable failures such
254 * as journal IO errors or ENOMEM at a critical moment in log management.
255 *
256 * We unconditionally force the filesystem into an ABORT|READONLY state,
257 * unless the error response on the fs has been set to panic in which
258 * case we take the easy way out and panic immediately.
259 */
260
261void ext3_abort (struct super_block * sb, const char * function,
262 const char * fmt, ...)
263{
264 va_list args;
265
266 printk (KERN_CRIT "ext3_abort called.\n");
267
268 va_start(args, fmt);
269 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
270 vprintk(fmt, args);
271 printk("\n");
272 va_end(args);
273
274 if (test_opt(sb, ERRORS_PANIC))
275 panic("EXT3-fs panic from previous error\n");
276
277 if (sb->s_flags & MS_RDONLY)
278 return;
279
280 printk(KERN_CRIT "Remounting filesystem read-only\n");
281 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
282 sb->s_flags |= MS_RDONLY;
283 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
284 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
285}
286
287void ext3_warning (struct super_block * sb, const char * function,
288 const char * fmt, ...)
289{
290 va_list args;
291
292 va_start(args, fmt);
293 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
294 sb->s_id, function);
295 vprintk(fmt, args);
296 printk("\n");
297 va_end(args);
298}
299
300void ext3_update_dynamic_rev(struct super_block *sb)
301{
302 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
303
304 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
305 return;
306
307 ext3_warning(sb, __FUNCTION__,
308 "updating to rev %d because of new feature flag, "
309 "running e2fsck is recommended",
310 EXT3_DYNAMIC_REV);
311
312 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
313 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
314 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
315 /* leave es->s_feature_*compat flags alone */
316 /* es->s_uuid will be set by e2fsck if empty */
317
318 /*
319 * The rest of the superblock fields should be zero, and if not it
320 * means they are likely already in use, so leave them alone. We
321 * can leave it up to e2fsck to clean up any inconsistencies there.
322 */
323}
324
325/*
326 * Open the external journal device
327 */
328static struct block_device *ext3_blkdev_get(dev_t dev)
329{
330 struct block_device *bdev;
331 char b[BDEVNAME_SIZE];
332
333 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
334 if (IS_ERR(bdev))
335 goto fail;
336 return bdev;
337
338fail:
339 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
340 __bdevname(dev, b), PTR_ERR(bdev));
341 return NULL;
342}
343
344/*
345 * Release the journal device
346 */
347static int ext3_blkdev_put(struct block_device *bdev)
348{
349 bd_release(bdev);
350 return blkdev_put(bdev);
351}
352
353static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
354{
355 struct block_device *bdev;
356 int ret = -ENODEV;
357
358 bdev = sbi->journal_bdev;
359 if (bdev) {
360 ret = ext3_blkdev_put(bdev);
361 sbi->journal_bdev = NULL;
362 }
363 return ret;
364}
365
366static inline struct inode *orphan_list_entry(struct list_head *l)
367{
368 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
369}
370
371static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
372{
373 struct list_head *l;
374
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700375 printk(KERN_ERR "sb orphan head is %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 le32_to_cpu(sbi->s_es->s_last_orphan));
377
378 printk(KERN_ERR "sb_info orphan list:\n");
379 list_for_each(l, &sbi->s_orphan) {
380 struct inode *inode = orphan_list_entry(l);
381 printk(KERN_ERR " "
Eric Sandeeneee194e2006-09-27 01:49:30 -0700382 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 inode->i_sb->s_id, inode->i_ino, inode,
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700384 inode->i_mode, inode->i_nlink,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 NEXT_ORPHAN(inode));
386 }
387}
388
389static void ext3_put_super (struct super_block * sb)
390{
391 struct ext3_sb_info *sbi = EXT3_SB(sb);
392 struct ext3_super_block *es = sbi->s_es;
393 int i;
394
395 ext3_xattr_put_super(sb);
396 journal_destroy(sbi->s_journal);
397 if (!(sb->s_flags & MS_RDONLY)) {
398 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
399 es->s_state = cpu_to_le16(sbi->s_mount_state);
400 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
401 mark_buffer_dirty(sbi->s_sbh);
402 ext3_commit_super(sb, es, 1);
403 }
404
405 for (i = 0; i < sbi->s_gdb_count; i++)
406 brelse(sbi->s_group_desc[i]);
407 kfree(sbi->s_group_desc);
408 percpu_counter_destroy(&sbi->s_freeblocks_counter);
409 percpu_counter_destroy(&sbi->s_freeinodes_counter);
410 percpu_counter_destroy(&sbi->s_dirs_counter);
411 brelse(sbi->s_sbh);
412#ifdef CONFIG_QUOTA
413 for (i = 0; i < MAXQUOTAS; i++)
414 kfree(sbi->s_qf_names[i]);
415#endif
416
417 /* Debugging code just in case the in-memory inode orphan list
418 * isn't empty. The on-disk one can be non-empty if we've
419 * detected an error and taken the fs readonly, but the
420 * in-memory list had better be clean by this point. */
421 if (!list_empty(&sbi->s_orphan))
422 dump_orphan_list(sb, sbi);
423 J_ASSERT(list_empty(&sbi->s_orphan));
424
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700425 invalidate_bdev(sb->s_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
427 /*
428 * Invalidate the journal device's buffers. We don't want them
429 * floating about in memory - the physical journal device may
430 * hotswapped, and it breaks the `ro-after' testing code.
431 */
432 sync_blockdev(sbi->journal_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700433 invalidate_bdev(sbi->journal_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 ext3_blkdev_remove(sbi);
435 }
436 sb->s_fs_info = NULL;
437 kfree(sbi);
438 return;
439}
440
Christoph Lametere18b8902006-12-06 20:33:20 -0800441static struct kmem_cache *ext3_inode_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442
443/*
444 * Called inside transaction, so use GFP_NOFS
445 */
446static struct inode *ext3_alloc_inode(struct super_block *sb)
447{
448 struct ext3_inode_info *ei;
449
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800450 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 if (!ei)
452 return NULL;
453#ifdef CONFIG_EXT3_FS_POSIX_ACL
454 ei->i_acl = EXT3_ACL_NOT_CACHED;
455 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
456#endif
457 ei->i_block_alloc_info = NULL;
458 ei->vfs_inode.i_version = 1;
459 return &ei->vfs_inode;
460}
461
462static void ext3_destroy_inode(struct inode *inode)
463{
Vasily Averin9f7dd932007-07-15 23:40:45 -0700464 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
465 printk("EXT3 Inode %p: orphan list check failed!\n",
466 EXT3_I(inode));
467 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
468 EXT3_I(inode), sizeof(struct ext3_inode_info),
469 false);
470 dump_stack();
471 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
473}
474
Christoph Lameter4ba9b9d2007-10-16 23:25:51 -0700475static void init_once(struct kmem_cache * cachep, void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476{
477 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
478
Christoph Lametera35afb82007-05-16 22:10:57 -0700479 INIT_LIST_HEAD(&ei->i_orphan);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480#ifdef CONFIG_EXT3_FS_XATTR
Christoph Lametera35afb82007-05-16 22:10:57 -0700481 init_rwsem(&ei->xattr_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482#endif
Christoph Lametera35afb82007-05-16 22:10:57 -0700483 mutex_init(&ei->truncate_mutex);
484 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485}
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700486
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487static int init_inodecache(void)
488{
489 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
490 sizeof(struct ext3_inode_info),
Paul Jacksonfffb60f2006-03-24 03:16:06 -0800491 0, (SLAB_RECLAIM_ACCOUNT|
492 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900493 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 if (ext3_inode_cachep == NULL)
495 return -ENOMEM;
496 return 0;
497}
498
499static void destroy_inodecache(void)
500{
Alexey Dobriyan1a1d92c2006-09-27 01:49:40 -0700501 kmem_cache_destroy(ext3_inode_cachep);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502}
503
504static void ext3_clear_inode(struct inode *inode)
505{
506 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
507#ifdef CONFIG_EXT3_FS_POSIX_ACL
Andrew Mortone6022602006-06-23 02:05:32 -0700508 if (EXT3_I(inode)->i_acl &&
509 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
510 posix_acl_release(EXT3_I(inode)->i_acl);
511 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
512 }
513 if (EXT3_I(inode)->i_default_acl &&
514 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
515 posix_acl_release(EXT3_I(inode)->i_default_acl);
516 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
517 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518#endif
519 ext3_discard_reservation(inode);
520 EXT3_I(inode)->i_block_alloc_info = NULL;
Andrew Mortone6022602006-06-23 02:05:32 -0700521 if (unlikely(rsv))
522 kfree(rsv);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523}
524
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200525static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
Mark Bellon8fc27512005-09-06 15:16:54 -0700526{
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200527#if defined(CONFIG_QUOTA)
OGAWA Hirofumi275abf52005-09-22 21:44:03 -0700528 struct ext3_sb_info *sbi = EXT3_SB(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529
Mark Bellon8fc27512005-09-06 15:16:54 -0700530 if (sbi->s_jquota_fmt)
531 seq_printf(seq, ",jqfmt=%s",
532 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
533
534 if (sbi->s_qf_names[USRQUOTA])
535 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
536
537 if (sbi->s_qf_names[GRPQUOTA])
538 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
539
540 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
541 seq_puts(seq, ",usrquota");
542
543 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
544 seq_puts(seq, ",grpquota");
545#endif
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200546}
547
Miklos Szeredi571beed2007-10-16 23:26:26 -0700548/*
549 * Show an option if
550 * - it's set to a non-default value OR
551 * - if the per-sb default is different from the global default
552 */
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200553static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
554{
555 struct super_block *sb = vfs->mnt_sb;
Miklos Szeredi571beed2007-10-16 23:26:26 -0700556 struct ext3_sb_info *sbi = EXT3_SB(sb);
557 struct ext3_super_block *es = sbi->s_es;
558 unsigned long def_mount_opts;
559
560 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
561
562 if (sbi->s_sb_block != 1)
563 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
564 if (test_opt(sb, MINIX_DF))
565 seq_puts(seq, ",minixdf");
566 if (test_opt(sb, GRPID))
567 seq_puts(seq, ",grpid");
568 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
569 seq_puts(seq, ",nogrpid");
570 if (sbi->s_resuid != EXT3_DEF_RESUID ||
571 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
572 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
573 }
574 if (sbi->s_resgid != EXT3_DEF_RESGID ||
575 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
576 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
577 }
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -0800578 if (test_opt(sb, ERRORS_RO)) {
Miklos Szeredi571beed2007-10-16 23:26:26 -0700579 int def_errors = le16_to_cpu(es->s_errors);
580
581 if (def_errors == EXT3_ERRORS_PANIC ||
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -0800582 def_errors == EXT3_ERRORS_CONTINUE) {
583 seq_puts(seq, ",errors=remount-ro");
Miklos Szeredi571beed2007-10-16 23:26:26 -0700584 }
585 }
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -0800586 if (test_opt(sb, ERRORS_CONT))
587 seq_puts(seq, ",errors=continue");
Miklos Szeredi571beed2007-10-16 23:26:26 -0700588 if (test_opt(sb, ERRORS_PANIC))
589 seq_puts(seq, ",errors=panic");
590 if (test_opt(sb, NO_UID32))
591 seq_puts(seq, ",nouid32");
592 if (test_opt(sb, DEBUG))
593 seq_puts(seq, ",debug");
594 if (test_opt(sb, OLDALLOC))
595 seq_puts(seq, ",oldalloc");
596#ifdef CONFIG_EXT3_FS_XATTR
597 if (test_opt(sb, XATTR_USER))
598 seq_puts(seq, ",user_xattr");
599 if (!test_opt(sb, XATTR_USER) &&
600 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
601 seq_puts(seq, ",nouser_xattr");
602 }
603#endif
604#ifdef CONFIG_EXT3_FS_POSIX_ACL
605 if (test_opt(sb, POSIX_ACL))
606 seq_puts(seq, ",acl");
607 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
608 seq_puts(seq, ",noacl");
609#endif
610 if (!test_opt(sb, RESERVATION))
611 seq_puts(seq, ",noreservation");
612 if (sbi->s_commit_interval) {
613 seq_printf(seq, ",commit=%u",
614 (unsigned) (sbi->s_commit_interval / HZ));
615 }
616 if (test_opt(sb, BARRIER))
617 seq_puts(seq, ",barrier=1");
618 if (test_opt(sb, NOBH))
619 seq_puts(seq, ",nobh");
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200620
621 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
622 seq_puts(seq, ",data=journal");
623 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
624 seq_puts(seq, ",data=ordered");
625 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
626 seq_puts(seq, ",data=writeback");
627
628 ext3_show_quota_options(seq, sb);
Mark Bellon8fc27512005-09-06 15:16:54 -0700629
630 return 0;
631}
632
NeilBrownfdb36672006-09-16 12:15:38 -0700633
Christoph Hellwig74af0ba2007-10-21 16:42:07 -0700634static struct inode *ext3_nfs_get_inode(struct super_block *sb,
635 u64 ino, u32 generation)
NeilBrownfdb36672006-09-16 12:15:38 -0700636{
NeilBrownfdb36672006-09-16 12:15:38 -0700637 struct inode *inode;
NeilBrownfdb36672006-09-16 12:15:38 -0700638
639 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
640 return ERR_PTR(-ESTALE);
641 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
642 return ERR_PTR(-ESTALE);
643
644 /* iget isn't really right if the inode is currently unallocated!!
645 *
646 * ext3_read_inode will return a bad_inode if the inode had been
647 * deleted, so we should be safe.
648 *
649 * Currently we don't know the generation for parent directory, so
650 * a generation of 0 means "accept any"
651 */
David Howells473043d2008-02-07 00:15:36 -0800652 inode = ext3_iget(sb, ino);
653 if (IS_ERR(inode))
654 return ERR_CAST(inode);
655 if (generation && inode->i_generation != generation) {
NeilBrownfdb36672006-09-16 12:15:38 -0700656 iput(inode);
657 return ERR_PTR(-ESTALE);
658 }
Christoph Hellwig74af0ba2007-10-21 16:42:07 -0700659
660 return inode;
661}
662
663static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
664 int fh_len, int fh_type)
665{
666 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
667 ext3_nfs_get_inode);
668}
669
670static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
671 int fh_len, int fh_type)
672{
673 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
674 ext3_nfs_get_inode);
NeilBrownfdb36672006-09-16 12:15:38 -0700675}
676
Mark Bellon8fc27512005-09-06 15:16:54 -0700677#ifdef CONFIG_QUOTA
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
679#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
680
681static int ext3_dquot_initialize(struct inode *inode, int type);
682static int ext3_dquot_drop(struct inode *inode);
683static int ext3_write_dquot(struct dquot *dquot);
684static int ext3_acquire_dquot(struct dquot *dquot);
685static int ext3_release_dquot(struct dquot *dquot);
686static int ext3_mark_dquot_dirty(struct dquot *dquot);
687static int ext3_write_info(struct super_block *sb, int type);
Jan Kara2fd83a42008-04-28 02:14:34 -0700688static int ext3_quota_on(struct super_block *sb, int type, int format_id,
689 char *path, int remount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690static int ext3_quota_on_mount(struct super_block *sb, int type);
691static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
692 size_t len, loff_t off);
693static ssize_t ext3_quota_write(struct super_block *sb, int type,
694 const char *data, size_t len, loff_t off);
695
696static struct dquot_operations ext3_quota_operations = {
697 .initialize = ext3_dquot_initialize,
698 .drop = ext3_dquot_drop,
699 .alloc_space = dquot_alloc_space,
700 .alloc_inode = dquot_alloc_inode,
701 .free_space = dquot_free_space,
702 .free_inode = dquot_free_inode,
703 .transfer = dquot_transfer,
704 .write_dquot = ext3_write_dquot,
705 .acquire_dquot = ext3_acquire_dquot,
706 .release_dquot = ext3_release_dquot,
707 .mark_dirty = ext3_mark_dquot_dirty,
708 .write_info = ext3_write_info
709};
710
711static struct quotactl_ops ext3_qctl_operations = {
712 .quota_on = ext3_quota_on,
713 .quota_off = vfs_quota_off,
714 .quota_sync = vfs_quota_sync,
715 .get_info = vfs_get_dqinfo,
716 .set_info = vfs_set_dqinfo,
717 .get_dqblk = vfs_get_dqblk,
718 .set_dqblk = vfs_set_dqblk
719};
720#endif
721
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800722static const struct super_operations ext3_sops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 .alloc_inode = ext3_alloc_inode,
724 .destroy_inode = ext3_destroy_inode,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725 .write_inode = ext3_write_inode,
726 .dirty_inode = ext3_dirty_inode,
727 .delete_inode = ext3_delete_inode,
728 .put_super = ext3_put_super,
729 .write_super = ext3_write_super,
730 .sync_fs = ext3_sync_fs,
731 .write_super_lockfs = ext3_write_super_lockfs,
732 .unlockfs = ext3_unlockfs,
733 .statfs = ext3_statfs,
734 .remount_fs = ext3_remount,
735 .clear_inode = ext3_clear_inode,
Mark Bellon8fc27512005-09-06 15:16:54 -0700736 .show_options = ext3_show_options,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737#ifdef CONFIG_QUOTA
738 .quota_read = ext3_quota_read,
739 .quota_write = ext3_quota_write,
740#endif
741};
742
Christoph Hellwig39655162007-10-21 16:42:17 -0700743static const struct export_operations ext3_export_ops = {
Christoph Hellwig74af0ba2007-10-21 16:42:07 -0700744 .fh_to_dentry = ext3_fh_to_dentry,
745 .fh_to_parent = ext3_fh_to_parent,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 .get_parent = ext3_get_parent,
747};
748
749enum {
750 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
751 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
Adrian Bunk2860b732005-11-08 21:34:59 -0800752 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
Badari Pulavartyade1a292006-06-26 00:25:04 -0700754 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
Johann Lombardi71b96252006-01-08 01:03:20 -0800755 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
757 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
Jan Kara1f545872005-06-23 22:01:04 -0700758 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
Mark Bellon8fc27512005-09-06 15:16:54 -0700759 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
760 Opt_grpquota
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761};
762
763static match_table_t tokens = {
764 {Opt_bsd_df, "bsddf"},
765 {Opt_minix_df, "minixdf"},
766 {Opt_grpid, "grpid"},
767 {Opt_grpid, "bsdgroups"},
768 {Opt_nogrpid, "nogrpid"},
769 {Opt_nogrpid, "sysvgroups"},
770 {Opt_resgid, "resgid=%u"},
771 {Opt_resuid, "resuid=%u"},
772 {Opt_sb, "sb=%u"},
773 {Opt_err_cont, "errors=continue"},
774 {Opt_err_panic, "errors=panic"},
775 {Opt_err_ro, "errors=remount-ro"},
776 {Opt_nouid32, "nouid32"},
777 {Opt_nocheck, "nocheck"},
778 {Opt_nocheck, "check=none"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 {Opt_debug, "debug"},
780 {Opt_oldalloc, "oldalloc"},
781 {Opt_orlov, "orlov"},
782 {Opt_user_xattr, "user_xattr"},
783 {Opt_nouser_xattr, "nouser_xattr"},
784 {Opt_acl, "acl"},
785 {Opt_noacl, "noacl"},
786 {Opt_reservation, "reservation"},
787 {Opt_noreservation, "noreservation"},
788 {Opt_noload, "noload"},
789 {Opt_nobh, "nobh"},
Badari Pulavartyade1a292006-06-26 00:25:04 -0700790 {Opt_bh, "bh"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 {Opt_commit, "commit=%u"},
792 {Opt_journal_update, "journal=update"},
793 {Opt_journal_inum, "journal=%u"},
Johann Lombardi71b96252006-01-08 01:03:20 -0800794 {Opt_journal_dev, "journal_dev=%u"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 {Opt_abort, "abort"},
796 {Opt_data_journal, "data=journal"},
797 {Opt_data_ordered, "data=ordered"},
798 {Opt_data_writeback, "data=writeback"},
799 {Opt_offusrjquota, "usrjquota="},
800 {Opt_usrjquota, "usrjquota=%s"},
801 {Opt_offgrpjquota, "grpjquota="},
802 {Opt_grpjquota, "grpjquota=%s"},
803 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
804 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
Mark Bellon8fc27512005-09-06 15:16:54 -0700805 {Opt_grpquota, "grpquota"},
Jan Kara1f545872005-06-23 22:01:04 -0700806 {Opt_noquota, "noquota"},
807 {Opt_quota, "quota"},
Mark Bellon8fc27512005-09-06 15:16:54 -0700808 {Opt_usrquota, "usrquota"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809 {Opt_barrier, "barrier=%u"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 {Opt_resize, "resize"},
Josef Bacik92587212008-03-04 14:29:43 -0800811 {Opt_err, NULL},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812};
813
Mingming Cao43d23f92006-06-25 05:48:07 -0700814static ext3_fsblk_t get_sb_block(void **data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815{
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700816 ext3_fsblk_t sb_block;
817 char *options = (char *) *data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818
819 if (!options || strncmp(options, "sb=", 3) != 0)
820 return 1; /* Default location */
821 options += 3;
Mingming Cao43d23f92006-06-25 05:48:07 -0700822 /*todo: use simple_strtoll with >32bit ext3 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 sb_block = simple_strtoul(options, &options, 0);
824 if (*options && *options != ',') {
825 printk("EXT3-fs: Invalid sb specification: %s\n",
826 (char *) *data);
827 return 1;
828 }
829 if (*options == ',')
830 options++;
831 *data = (void *) options;
832 return sb_block;
833}
834
Johann Lombardi71b96252006-01-08 01:03:20 -0800835static int parse_options (char *options, struct super_block *sb,
Eric Sandeeneee194e2006-09-27 01:49:30 -0700836 unsigned int *inum, unsigned long *journal_devnum,
Mingming Cao43d23f92006-06-25 05:48:07 -0700837 ext3_fsblk_t *n_blocks_count, int is_remount)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838{
839 struct ext3_sb_info *sbi = EXT3_SB(sb);
840 char * p;
841 substring_t args[MAX_OPT_ARGS];
842 int data_opt = 0;
843 int option;
844#ifdef CONFIG_QUOTA
845 int qtype;
846 char *qname;
847#endif
848
849 if (!options)
850 return 1;
851
852 while ((p = strsep (&options, ",")) != NULL) {
853 int token;
854 if (!*p)
855 continue;
856
857 token = match_token(p, tokens, args);
858 switch (token) {
859 case Opt_bsd_df:
860 clear_opt (sbi->s_mount_opt, MINIX_DF);
861 break;
862 case Opt_minix_df:
863 set_opt (sbi->s_mount_opt, MINIX_DF);
864 break;
865 case Opt_grpid:
866 set_opt (sbi->s_mount_opt, GRPID);
867 break;
868 case Opt_nogrpid:
869 clear_opt (sbi->s_mount_opt, GRPID);
870 break;
871 case Opt_resuid:
872 if (match_int(&args[0], &option))
873 return 0;
874 sbi->s_resuid = option;
875 break;
876 case Opt_resgid:
877 if (match_int(&args[0], &option))
878 return 0;
879 sbi->s_resgid = option;
880 break;
881 case Opt_sb:
882 /* handled by get_sb_block() instead of here */
883 /* *sb_block = match_int(&args[0]); */
884 break;
885 case Opt_err_panic:
886 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
887 clear_opt (sbi->s_mount_opt, ERRORS_RO);
888 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
889 break;
890 case Opt_err_ro:
891 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
892 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
893 set_opt (sbi->s_mount_opt, ERRORS_RO);
894 break;
895 case Opt_err_cont:
896 clear_opt (sbi->s_mount_opt, ERRORS_RO);
897 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
898 set_opt (sbi->s_mount_opt, ERRORS_CONT);
899 break;
900 case Opt_nouid32:
901 set_opt (sbi->s_mount_opt, NO_UID32);
902 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 case Opt_nocheck:
904 clear_opt (sbi->s_mount_opt, CHECK);
905 break;
906 case Opt_debug:
907 set_opt (sbi->s_mount_opt, DEBUG);
908 break;
909 case Opt_oldalloc:
910 set_opt (sbi->s_mount_opt, OLDALLOC);
911 break;
912 case Opt_orlov:
913 clear_opt (sbi->s_mount_opt, OLDALLOC);
914 break;
915#ifdef CONFIG_EXT3_FS_XATTR
916 case Opt_user_xattr:
917 set_opt (sbi->s_mount_opt, XATTR_USER);
918 break;
919 case Opt_nouser_xattr:
920 clear_opt (sbi->s_mount_opt, XATTR_USER);
921 break;
922#else
923 case Opt_user_xattr:
924 case Opt_nouser_xattr:
925 printk("EXT3 (no)user_xattr options not supported\n");
926 break;
927#endif
928#ifdef CONFIG_EXT3_FS_POSIX_ACL
929 case Opt_acl:
930 set_opt(sbi->s_mount_opt, POSIX_ACL);
931 break;
932 case Opt_noacl:
933 clear_opt(sbi->s_mount_opt, POSIX_ACL);
934 break;
935#else
936 case Opt_acl:
937 case Opt_noacl:
938 printk("EXT3 (no)acl options not supported\n");
939 break;
940#endif
941 case Opt_reservation:
942 set_opt(sbi->s_mount_opt, RESERVATION);
943 break;
944 case Opt_noreservation:
945 clear_opt(sbi->s_mount_opt, RESERVATION);
946 break;
947 case Opt_journal_update:
948 /* @@@ FIXME */
949 /* Eventually we will want to be able to create
950 a journal file here. For now, only allow the
951 user to specify an existing inode to be the
952 journal file. */
953 if (is_remount) {
954 printk(KERN_ERR "EXT3-fs: cannot specify "
955 "journal on remount\n");
956 return 0;
957 }
958 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
959 break;
960 case Opt_journal_inum:
961 if (is_remount) {
962 printk(KERN_ERR "EXT3-fs: cannot specify "
963 "journal on remount\n");
964 return 0;
965 }
966 if (match_int(&args[0], &option))
967 return 0;
968 *inum = option;
969 break;
Johann Lombardi71b96252006-01-08 01:03:20 -0800970 case Opt_journal_dev:
971 if (is_remount) {
972 printk(KERN_ERR "EXT3-fs: cannot specify "
973 "journal on remount\n");
974 return 0;
975 }
976 if (match_int(&args[0], &option))
977 return 0;
978 *journal_devnum = option;
979 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 case Opt_noload:
981 set_opt (sbi->s_mount_opt, NOLOAD);
982 break;
983 case Opt_commit:
984 if (match_int(&args[0], &option))
985 return 0;
986 if (option < 0)
987 return 0;
988 if (option == 0)
989 option = JBD_DEFAULT_MAX_COMMIT_AGE;
990 sbi->s_commit_interval = HZ * option;
991 break;
992 case Opt_data_journal:
993 data_opt = EXT3_MOUNT_JOURNAL_DATA;
994 goto datacheck;
995 case Opt_data_ordered:
996 data_opt = EXT3_MOUNT_ORDERED_DATA;
997 goto datacheck;
998 case Opt_data_writeback:
999 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1000 datacheck:
1001 if (is_remount) {
1002 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
1003 != data_opt) {
1004 printk(KERN_ERR
1005 "EXT3-fs: cannot change data "
1006 "mode on remount\n");
1007 return 0;
1008 }
1009 } else {
1010 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
1011 sbi->s_mount_opt |= data_opt;
1012 }
1013 break;
1014#ifdef CONFIG_QUOTA
1015 case Opt_usrjquota:
1016 qtype = USRQUOTA;
1017 goto set_qf_name;
1018 case Opt_grpjquota:
1019 qtype = GRPQUOTA;
1020set_qf_name:
1021 if (sb_any_quota_enabled(sb)) {
1022 printk(KERN_ERR
1023 "EXT3-fs: Cannot change journalled "
1024 "quota options when quota turned on.\n");
1025 return 0;
1026 }
1027 qname = match_strdup(&args[0]);
1028 if (!qname) {
1029 printk(KERN_ERR
1030 "EXT3-fs: not enough memory for "
1031 "storing quotafile name.\n");
1032 return 0;
1033 }
1034 if (sbi->s_qf_names[qtype] &&
1035 strcmp(sbi->s_qf_names[qtype], qname)) {
1036 printk(KERN_ERR
1037 "EXT3-fs: %s quota file already "
1038 "specified.\n", QTYPE2NAME(qtype));
1039 kfree(qname);
1040 return 0;
1041 }
1042 sbi->s_qf_names[qtype] = qname;
1043 if (strchr(sbi->s_qf_names[qtype], '/')) {
1044 printk(KERN_ERR
1045 "EXT3-fs: quotafile must be on "
1046 "filesystem root.\n");
1047 kfree(sbi->s_qf_names[qtype]);
1048 sbi->s_qf_names[qtype] = NULL;
1049 return 0;
1050 }
Jan Kara1f545872005-06-23 22:01:04 -07001051 set_opt(sbi->s_mount_opt, QUOTA);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 break;
1053 case Opt_offusrjquota:
1054 qtype = USRQUOTA;
1055 goto clear_qf_name;
1056 case Opt_offgrpjquota:
1057 qtype = GRPQUOTA;
1058clear_qf_name:
1059 if (sb_any_quota_enabled(sb)) {
1060 printk(KERN_ERR "EXT3-fs: Cannot change "
1061 "journalled quota options when "
1062 "quota turned on.\n");
1063 return 0;
1064 }
Jan Kara08c6a962005-07-12 13:58:28 -07001065 /*
1066 * The space will be released later when all options
1067 * are confirmed to be correct
1068 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 sbi->s_qf_names[qtype] = NULL;
1070 break;
1071 case Opt_jqfmt_vfsold:
1072 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1073 break;
1074 case Opt_jqfmt_vfsv0:
1075 sbi->s_jquota_fmt = QFMT_VFS_V0;
1076 break;
Jan Kara1f545872005-06-23 22:01:04 -07001077 case Opt_quota:
Mark Bellon8fc27512005-09-06 15:16:54 -07001078 case Opt_usrquota:
Jan Kara1f545872005-06-23 22:01:04 -07001079 set_opt(sbi->s_mount_opt, QUOTA);
Mark Bellon8fc27512005-09-06 15:16:54 -07001080 set_opt(sbi->s_mount_opt, USRQUOTA);
1081 break;
1082 case Opt_grpquota:
1083 set_opt(sbi->s_mount_opt, QUOTA);
1084 set_opt(sbi->s_mount_opt, GRPQUOTA);
Jan Kara1f545872005-06-23 22:01:04 -07001085 break;
1086 case Opt_noquota:
1087 if (sb_any_quota_enabled(sb)) {
1088 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1089 "options when quota turned on.\n");
1090 return 0;
1091 }
1092 clear_opt(sbi->s_mount_opt, QUOTA);
Mark Bellon8fc27512005-09-06 15:16:54 -07001093 clear_opt(sbi->s_mount_opt, USRQUOTA);
1094 clear_opt(sbi->s_mount_opt, GRPQUOTA);
Jan Kara1f545872005-06-23 22:01:04 -07001095 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096#else
Mark Bellon8fc27512005-09-06 15:16:54 -07001097 case Opt_quota:
1098 case Opt_usrquota:
1099 case Opt_grpquota:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 case Opt_usrjquota:
1101 case Opt_grpjquota:
1102 case Opt_offusrjquota:
1103 case Opt_offgrpjquota:
1104 case Opt_jqfmt_vfsold:
1105 case Opt_jqfmt_vfsv0:
1106 printk(KERN_ERR
1107 "EXT3-fs: journalled quota options not "
1108 "supported.\n");
1109 break;
Jan Kara1f545872005-06-23 22:01:04 -07001110 case Opt_noquota:
1111 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112#endif
1113 case Opt_abort:
1114 set_opt(sbi->s_mount_opt, ABORT);
1115 break;
1116 case Opt_barrier:
1117 if (match_int(&args[0], &option))
1118 return 0;
1119 if (option)
1120 set_opt(sbi->s_mount_opt, BARRIER);
1121 else
1122 clear_opt(sbi->s_mount_opt, BARRIER);
1123 break;
1124 case Opt_ignore:
1125 break;
1126 case Opt_resize:
Jan Kara08c6a962005-07-12 13:58:28 -07001127 if (!is_remount) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 printk("EXT3-fs: resize option only available "
1129 "for remount\n");
1130 return 0;
1131 }
KAMBAROV, ZAURc7f17212005-06-28 20:45:11 -07001132 if (match_int(&args[0], &option) != 0)
1133 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134 *n_blocks_count = option;
1135 break;
1136 case Opt_nobh:
1137 set_opt(sbi->s_mount_opt, NOBH);
1138 break;
Badari Pulavartyade1a292006-06-26 00:25:04 -07001139 case Opt_bh:
1140 clear_opt(sbi->s_mount_opt, NOBH);
1141 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142 default:
1143 printk (KERN_ERR
1144 "EXT3-fs: Unrecognized mount option \"%s\" "
1145 "or missing value\n", p);
1146 return 0;
1147 }
1148 }
1149#ifdef CONFIG_QUOTA
Mark Bellon8fc27512005-09-06 15:16:54 -07001150 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1151 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1152 sbi->s_qf_names[USRQUOTA])
1153 clear_opt(sbi->s_mount_opt, USRQUOTA);
1154
1155 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1156 sbi->s_qf_names[GRPQUOTA])
1157 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1158
1159 if ((sbi->s_qf_names[USRQUOTA] &&
1160 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1161 (sbi->s_qf_names[GRPQUOTA] &&
1162 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1163 printk(KERN_ERR "EXT3-fs: old and new quota "
1164 "format mixing.\n");
1165 return 0;
1166 }
1167
1168 if (!sbi->s_jquota_fmt) {
1169 printk(KERN_ERR "EXT3-fs: journalled quota format "
1170 "not specified.\n");
1171 return 0;
1172 }
1173 } else {
1174 if (sbi->s_jquota_fmt) {
1175 printk(KERN_ERR "EXT3-fs: journalled quota format "
1176 "specified with no journalling "
1177 "enabled.\n");
1178 return 0;
1179 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 }
1181#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 return 1;
1183}
1184
1185static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1186 int read_only)
1187{
1188 struct ext3_sb_info *sbi = EXT3_SB(sb);
1189 int res = 0;
1190
1191 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1192 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1193 "forcing read-only mode\n");
1194 res = MS_RDONLY;
1195 }
1196 if (read_only)
1197 return res;
1198 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1199 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1200 "running e2fsck is recommended\n");
1201 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1202 printk (KERN_WARNING
1203 "EXT3-fs warning: mounting fs with errors, "
1204 "running e2fsck is recommended\n");
1205 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1206 le16_to_cpu(es->s_mnt_count) >=
1207 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1208 printk (KERN_WARNING
1209 "EXT3-fs warning: maximal mount count reached, "
1210 "running e2fsck is recommended\n");
1211 else if (le32_to_cpu(es->s_checkinterval) &&
1212 (le32_to_cpu(es->s_lastcheck) +
1213 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1214 printk (KERN_WARNING
1215 "EXT3-fs warning: checktime reached, "
1216 "running e2fsck is recommended\n");
1217#if 0
1218 /* @@@ We _will_ want to clear the valid bit if we find
1219 inconsistencies, to force a fsck at reboot. But for
1220 a plain journaled filesystem we can keep it set as
1221 valid forever! :) */
1222 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT3_VALID_FS);
1223#endif
1224 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1225 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
Marcin Slusarz50e8a282008-02-08 04:20:13 -08001226 le16_add_cpu(&es->s_mnt_count, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227 es->s_mtime = cpu_to_le32(get_seconds());
1228 ext3_update_dynamic_rev(sb);
1229 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1230
1231 ext3_commit_super(sb, es, 1);
1232 if (test_opt(sb, DEBUG))
1233 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1234 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1235 sb->s_blocksize,
1236 sbi->s_groups_count,
1237 EXT3_BLOCKS_PER_GROUP(sb),
1238 EXT3_INODES_PER_GROUP(sb),
1239 sbi->s_mount_opt);
1240
1241 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1242 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1243 char b[BDEVNAME_SIZE];
1244
1245 printk("external journal on %s\n",
1246 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1247 } else {
1248 printk("internal journal\n");
1249 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 return res;
1251}
1252
1253/* Called at mount-time, super-block is locked */
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001254static int ext3_check_descriptors(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255{
1256 struct ext3_sb_info *sbi = EXT3_SB(sb);
Eric Sandeen855565e2006-09-27 01:49:29 -07001257 ext3_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1258 ext3_fsblk_t last_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259 int i;
1260
1261 ext3_debug ("Checking group descriptors");
1262
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001263 for (i = 0; i < sbi->s_groups_count; i++) {
1264 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1265
Eric Sandeen855565e2006-09-27 01:49:29 -07001266 if (i == sbi->s_groups_count - 1)
1267 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1268 else
1269 last_block = first_block +
1270 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1271
Eric Sandeen855565e2006-09-27 01:49:29 -07001272 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1273 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 {
1275 ext3_error (sb, "ext3_check_descriptors",
1276 "Block bitmap for group %d"
1277 " not in group (block %lu)!",
1278 i, (unsigned long)
1279 le32_to_cpu(gdp->bg_block_bitmap));
1280 return 0;
1281 }
Eric Sandeen855565e2006-09-27 01:49:29 -07001282 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1283 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 {
1285 ext3_error (sb, "ext3_check_descriptors",
1286 "Inode bitmap for group %d"
1287 " not in group (block %lu)!",
1288 i, (unsigned long)
1289 le32_to_cpu(gdp->bg_inode_bitmap));
1290 return 0;
1291 }
Eric Sandeen855565e2006-09-27 01:49:29 -07001292 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
Eric Sandeen780dcdb2007-07-26 10:41:11 -07001293 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
Eric Sandeen855565e2006-09-27 01:49:29 -07001294 last_block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295 {
1296 ext3_error (sb, "ext3_check_descriptors",
1297 "Inode table for group %d"
1298 " not in group (block %lu)!",
1299 i, (unsigned long)
1300 le32_to_cpu(gdp->bg_inode_table));
1301 return 0;
1302 }
Eric Sandeen855565e2006-09-27 01:49:29 -07001303 first_block += EXT3_BLOCKS_PER_GROUP(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304 }
1305
1306 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1307 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1308 return 1;
1309}
1310
1311
1312/* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1313 * the superblock) which were deleted from all directories, but held open by
1314 * a process at the time of a crash. We walk the list and try to delete these
1315 * inodes at recovery time (only with a read-write filesystem).
1316 *
1317 * In order to keep the orphan inode chain consistent during traversal (in
1318 * case of crash during recovery), we link each inode into the superblock
1319 * orphan list_head and handle it the same way as an inode deletion during
1320 * normal operation (which journals the operations for us).
1321 *
1322 * We only do an iget() and an iput() on each inode, which is very safe if we
1323 * accidentally point at an in-use or already deleted inode. The worst that
1324 * can happen in this case is that we get a "bit already cleared" message from
1325 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1326 * e2fsck was run on this filesystem, and it must have already done the orphan
1327 * inode cleanup for us, so we can safely abort without any further action.
1328 */
1329static void ext3_orphan_cleanup (struct super_block * sb,
1330 struct ext3_super_block * es)
1331{
1332 unsigned int s_flags = sb->s_flags;
1333 int nr_orphans = 0, nr_truncates = 0;
1334#ifdef CONFIG_QUOTA
1335 int i;
1336#endif
1337 if (!es->s_last_orphan) {
1338 jbd_debug(4, "no orphan inodes to clean up\n");
1339 return;
1340 }
1341
Eric Sandeena8f48a92006-12-06 20:40:13 -08001342 if (bdev_read_only(sb->s_bdev)) {
1343 printk(KERN_ERR "EXT3-fs: write access "
1344 "unavailable, skipping orphan cleanup.\n");
1345 return;
1346 }
1347
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1349 if (es->s_last_orphan)
1350 jbd_debug(1, "Errors on filesystem, "
1351 "clearing orphan list.\n");
1352 es->s_last_orphan = 0;
1353 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1354 return;
1355 }
1356
1357 if (s_flags & MS_RDONLY) {
1358 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1359 sb->s_id);
1360 sb->s_flags &= ~MS_RDONLY;
1361 }
1362#ifdef CONFIG_QUOTA
1363 /* Needed for iput() to work correctly and not trash data */
1364 sb->s_flags |= MS_ACTIVE;
1365 /* Turn on quotas so that they are updated correctly */
1366 for (i = 0; i < MAXQUOTAS; i++) {
1367 if (EXT3_SB(sb)->s_qf_names[i]) {
1368 int ret = ext3_quota_on_mount(sb, i);
1369 if (ret < 0)
1370 printk(KERN_ERR
1371 "EXT3-fs: Cannot turn on journalled "
1372 "quota: error %d\n", ret);
1373 }
1374 }
1375#endif
1376
1377 while (es->s_last_orphan) {
1378 struct inode *inode;
1379
David Howells473043d2008-02-07 00:15:36 -08001380 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1381 if (IS_ERR(inode)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382 es->s_last_orphan = 0;
1383 break;
1384 }
1385
1386 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1387 DQUOT_INIT(inode);
1388 if (inode->i_nlink) {
1389 printk(KERN_DEBUG
Eric Sandeeneee194e2006-09-27 01:49:30 -07001390 "%s: truncating inode %lu to %Ld bytes\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391 __FUNCTION__, inode->i_ino, inode->i_size);
Eric Sandeeneee194e2006-09-27 01:49:30 -07001392 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393 inode->i_ino, inode->i_size);
1394 ext3_truncate(inode);
1395 nr_truncates++;
1396 } else {
1397 printk(KERN_DEBUG
Eric Sandeeneee194e2006-09-27 01:49:30 -07001398 "%s: deleting unreferenced inode %lu\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 __FUNCTION__, inode->i_ino);
Eric Sandeeneee194e2006-09-27 01:49:30 -07001400 jbd_debug(2, "deleting unreferenced inode %lu\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401 inode->i_ino);
1402 nr_orphans++;
1403 }
1404 iput(inode); /* The delete magic happens here! */
1405 }
1406
1407#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1408
1409 if (nr_orphans)
1410 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1411 sb->s_id, PLURAL(nr_orphans));
1412 if (nr_truncates)
1413 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1414 sb->s_id, PLURAL(nr_truncates));
1415#ifdef CONFIG_QUOTA
1416 /* Turn quotas off */
1417 for (i = 0; i < MAXQUOTAS; i++) {
1418 if (sb_dqopt(sb)->files[i])
Jan Kara2fd83a42008-04-28 02:14:34 -07001419 vfs_quota_off(sb, i, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 }
1421#endif
1422 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1423}
1424
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425/*
1426 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1427 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1428 * We need to be 1 filesystem block less than the 2^32 sector limit.
1429 */
1430static loff_t ext3_max_size(int bits)
1431{
1432 loff_t res = EXT3_NDIR_BLOCKS;
Aneesh Kumar K.Vfe7fdc32008-01-28 23:58:26 -05001433 int meta_blocks;
1434 loff_t upper_limit;
1435
1436 /* This is calculated to be the largest file size for a
1437 * dense, file such that the total number of
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438 * sectors in the file, including data and all indirect blocks,
Aneesh Kumar K.Vfe7fdc32008-01-28 23:58:26 -05001439 * does not exceed 2^32 -1
1440 * __u32 i_blocks representing the total number of
1441 * 512 bytes blocks of the file
1442 */
1443 upper_limit = (1LL << 32) - 1;
1444
1445 /* total blocks in file system block size */
1446 upper_limit >>= (bits - 9);
1447
1448
1449 /* indirect blocks */
1450 meta_blocks = 1;
1451 /* double indirect blocks */
1452 meta_blocks += 1 + (1LL << (bits-2));
1453 /* tripple indirect blocks */
1454 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1455
1456 upper_limit -= meta_blocks;
1457 upper_limit <<= bits;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458
1459 res += 1LL << (bits-2);
1460 res += 1LL << (2*(bits-2));
1461 res += 1LL << (3*(bits-2));
1462 res <<= bits;
1463 if (res > upper_limit)
1464 res = upper_limit;
Aneesh Kumar K.Vfe7fdc32008-01-28 23:58:26 -05001465
1466 if (res > MAX_LFS_FILESIZE)
1467 res = MAX_LFS_FILESIZE;
1468
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 return res;
1470}
1471
Mingming Cao43d23f92006-06-25 05:48:07 -07001472static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1473 ext3_fsblk_t logic_sb_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 int nr)
1475{
1476 struct ext3_sb_info *sbi = EXT3_SB(sb);
Mingming Cao43d23f92006-06-25 05:48:07 -07001477 unsigned long bg, first_meta_bg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 int has_super = 0;
1479
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1481
1482 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1483 nr < first_meta_bg)
1484 return (logic_sb_block + nr + 1);
1485 bg = sbi->s_desc_per_block * nr;
1486 if (ext3_bg_has_super(sb, bg))
1487 has_super = 1;
Mingming Cao43d23f92006-06-25 05:48:07 -07001488 return (has_super + ext3_group_first_block_no(sb, bg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489}
1490
1491
1492static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1493{
1494 struct buffer_head * bh;
1495 struct ext3_super_block *es = NULL;
1496 struct ext3_sb_info *sbi;
Mingming Cao43d23f92006-06-25 05:48:07 -07001497 ext3_fsblk_t block;
1498 ext3_fsblk_t sb_block = get_sb_block(&data);
1499 ext3_fsblk_t logic_sb_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500 unsigned long offset = 0;
Eric Sandeeneee194e2006-09-27 01:49:30 -07001501 unsigned int journal_inum = 0;
Johann Lombardi71b96252006-01-08 01:03:20 -08001502 unsigned long journal_devnum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503 unsigned long def_mount_opts;
1504 struct inode *root;
1505 int blocksize;
1506 int hblock;
1507 int db_count;
1508 int i;
1509 int needs_recovery;
David Howells473043d2008-02-07 00:15:36 -08001510 int ret = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 __le32 features;
Peter Zijlstra833f4072007-10-16 23:25:45 -07001512 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513
Panagiotis Issarisf8314dc2006-09-27 01:49:37 -07001514 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515 if (!sbi)
1516 return -ENOMEM;
1517 sb->s_fs_info = sbi;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518 sbi->s_mount_opt = 0;
1519 sbi->s_resuid = EXT3_DEF_RESUID;
1520 sbi->s_resgid = EXT3_DEF_RESGID;
Miklos Szeredi571beed2007-10-16 23:26:26 -07001521 sbi->s_sb_block = sb_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522
1523 unlock_kernel();
1524
1525 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1526 if (!blocksize) {
1527 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1528 goto out_fail;
1529 }
1530
1531 /*
1532 * The ext3 superblock will not be buffer aligned for other than 1kB
1533 * block sizes. We need to calculate the offset from buffer start.
1534 */
1535 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1536 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1537 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1538 } else {
1539 logic_sb_block = sb_block;
1540 }
1541
1542 if (!(bh = sb_bread(sb, logic_sb_block))) {
1543 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1544 goto out_fail;
1545 }
1546 /*
1547 * Note: s_es must be initialized as soon as possible because
1548 * some ext3 macro-instructions depend on its value
1549 */
1550 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1551 sbi->s_es = es;
1552 sb->s_magic = le16_to_cpu(es->s_magic);
1553 if (sb->s_magic != EXT3_SUPER_MAGIC)
1554 goto cantfind_ext3;
1555
1556 /* Set defaults before we parse the mount options */
1557 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1558 if (def_mount_opts & EXT3_DEFM_DEBUG)
1559 set_opt(sbi->s_mount_opt, DEBUG);
1560 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1561 set_opt(sbi->s_mount_opt, GRPID);
1562 if (def_mount_opts & EXT3_DEFM_UID16)
1563 set_opt(sbi->s_mount_opt, NO_UID32);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08001564#ifdef CONFIG_EXT3_FS_XATTR
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1566 set_opt(sbi->s_mount_opt, XATTR_USER);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08001567#endif
1568#ifdef CONFIG_EXT3_FS_POSIX_ACL
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569 if (def_mount_opts & EXT3_DEFM_ACL)
1570 set_opt(sbi->s_mount_opt, POSIX_ACL);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08001571#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1573 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1574 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1575 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1576 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1577 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1578
1579 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1580 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -08001581 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
Dmitry Mishin2245d7c2006-10-11 01:21:49 -07001582 set_opt(sbi->s_mount_opt, ERRORS_CONT);
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -08001583 else
1584 set_opt(sbi->s_mount_opt, ERRORS_RO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585
1586 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1587 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1588
1589 set_opt(sbi->s_mount_opt, RESERVATION);
1590
Johann Lombardi71b96252006-01-08 01:03:20 -08001591 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1592 NULL, 0))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593 goto failed_mount;
1594
1595 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1596 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1597
1598 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1599 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1600 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1601 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001602 printk(KERN_WARNING
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603 "EXT3-fs warning: feature flags set on rev 0 fs, "
1604 "running e2fsck is recommended\n");
1605 /*
1606 * Check feature flags regardless of the revision level, since we
1607 * previously didn't change the revision level when setting the flags,
1608 * so there is a chance incompat flags are set on a rev 0 filesystem.
1609 */
1610 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1611 if (features) {
1612 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1613 "unsupported optional features (%x).\n",
1614 sb->s_id, le32_to_cpu(features));
1615 goto failed_mount;
1616 }
1617 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1618 if (!(sb->s_flags & MS_RDONLY) && features) {
1619 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1620 "unsupported optional features (%x).\n",
1621 sb->s_id, le32_to_cpu(features));
1622 goto failed_mount;
1623 }
1624 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1625
1626 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1627 blocksize > EXT3_MAX_BLOCK_SIZE) {
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001628 printk(KERN_ERR
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1630 blocksize, sb->s_id);
1631 goto failed_mount;
1632 }
1633
1634 hblock = bdev_hardsect_size(sb->s_bdev);
1635 if (sb->s_blocksize != blocksize) {
1636 /*
1637 * Make sure the blocksize for the filesystem is larger
1638 * than the hardware sectorsize for the machine.
1639 */
1640 if (blocksize < hblock) {
1641 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1642 "device blocksize %d.\n", blocksize, hblock);
1643 goto failed_mount;
1644 }
1645
1646 brelse (bh);
Takashi Sato0f0a89e2007-10-18 03:06:56 -07001647 if (!sb_set_blocksize(sb, blocksize)) {
1648 printk(KERN_ERR "EXT3-fs: bad blocksize %d.\n",
1649 blocksize);
1650 goto out_fail;
1651 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1653 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1654 bh = sb_bread(sb, logic_sb_block);
1655 if (!bh) {
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001656 printk(KERN_ERR
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657 "EXT3-fs: Can't read superblock on 2nd try.\n");
1658 goto failed_mount;
1659 }
1660 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1661 sbi->s_es = es;
1662 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001663 printk (KERN_ERR
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664 "EXT3-fs: Magic mismatch, very weird !\n");
1665 goto failed_mount;
1666 }
1667 }
1668
1669 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1670
1671 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1672 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1673 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1674 } else {
1675 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1676 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1677 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
vignesh babu3fc74262007-07-15 23:41:17 -07001678 (!is_power_of_2(sbi->s_inode_size)) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679 (sbi->s_inode_size > blocksize)) {
1680 printk (KERN_ERR
1681 "EXT3-fs: unsupported inode size: %d\n",
1682 sbi->s_inode_size);
1683 goto failed_mount;
1684 }
1685 }
1686 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1687 le32_to_cpu(es->s_log_frag_size);
1688 if (blocksize != sbi->s_frag_size) {
1689 printk(KERN_ERR
1690 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1691 sbi->s_frag_size, blocksize);
1692 goto failed_mount;
1693 }
1694 sbi->s_frags_per_block = 1;
1695 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1696 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1697 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Andries E. Brouwerb47b6f32007-12-17 16:19:55 -08001698 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699 goto cantfind_ext3;
1700 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1701 if (sbi->s_inodes_per_block == 0)
1702 goto cantfind_ext3;
1703 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1704 sbi->s_inodes_per_block;
1705 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1706 sbi->s_sbh = bh;
1707 sbi->s_mount_state = le16_to_cpu(es->s_state);
David Howellsf0d1b0b2006-12-08 02:37:49 -08001708 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1709 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 for (i=0; i < 4; i++)
1711 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1712 sbi->s_def_hash_version = es->s_def_hash_version;
1713
1714 if (sbi->s_blocks_per_group > blocksize * 8) {
1715 printk (KERN_ERR
1716 "EXT3-fs: #blocks per group too big: %lu\n",
1717 sbi->s_blocks_per_group);
1718 goto failed_mount;
1719 }
1720 if (sbi->s_frags_per_group > blocksize * 8) {
1721 printk (KERN_ERR
1722 "EXT3-fs: #fragments per group too big: %lu\n",
1723 sbi->s_frags_per_group);
1724 goto failed_mount;
1725 }
1726 if (sbi->s_inodes_per_group > blocksize * 8) {
1727 printk (KERN_ERR
1728 "EXT3-fs: #inodes per group too big: %lu\n",
1729 sbi->s_inodes_per_group);
1730 goto failed_mount;
1731 }
1732
Mingming Caofcd5df32006-06-25 05:47:50 -07001733 if (le32_to_cpu(es->s_blocks_count) >
1734 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1735 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1736 " too large to mount safely\n", sb->s_id);
1737 if (sizeof(sector_t) < 8)
1738 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1739 "enabled\n");
1740 goto failed_mount;
1741 }
1742
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1744 goto cantfind_ext3;
Eric Sandeen855565e2006-09-27 01:49:29 -07001745 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1746 le32_to_cpu(es->s_first_data_block) - 1)
1747 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1749 EXT3_DESC_PER_BLOCK(sb);
1750 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1751 GFP_KERNEL);
1752 if (sbi->s_group_desc == NULL) {
1753 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1754 goto failed_mount;
1755 }
1756
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 bgl_lock_init(&sbi->s_blockgroup_lock);
1758
1759 for (i = 0; i < db_count; i++) {
1760 block = descriptor_loc(sb, logic_sb_block, i);
1761 sbi->s_group_desc[i] = sb_bread(sb, block);
1762 if (!sbi->s_group_desc[i]) {
1763 printk (KERN_ERR "EXT3-fs: "
1764 "can't read group descriptor %d\n", i);
1765 db_count = i;
1766 goto failed_mount2;
1767 }
1768 }
1769 if (!ext3_check_descriptors (sb)) {
Theodore Ts'od2e5b132006-06-25 05:47:52 -07001770 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771 goto failed_mount2;
1772 }
1773 sbi->s_gdb_count = db_count;
1774 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1775 spin_lock_init(&sbi->s_next_gen_lock);
Mingming Cao0216bfc2006-06-23 02:05:41 -07001776
Peter Zijlstra833f4072007-10-16 23:25:45 -07001777 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1778 ext3_count_free_blocks(sb));
1779 if (!err) {
1780 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1781 ext3_count_free_inodes(sb));
1782 }
1783 if (!err) {
1784 err = percpu_counter_init(&sbi->s_dirs_counter,
1785 ext3_count_dirs(sb));
1786 }
1787 if (err) {
1788 printk(KERN_ERR "EXT3-fs: insufficient memory\n");
1789 goto failed_mount3;
1790 }
Mingming Cao0216bfc2006-06-23 02:05:41 -07001791
Linus Torvalds1da177e2005-04-16 15:20:36 -07001792 /* per fileystem reservation list head & lock */
1793 spin_lock_init(&sbi->s_rsv_window_lock);
1794 sbi->s_rsv_window_root = RB_ROOT;
1795 /* Add a single, static dummy reservation to the start of the
1796 * reservation window list --- it gives us a placeholder for
1797 * append-at-start-of-list which makes the allocation logic
1798 * _much_ simpler. */
1799 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1800 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1801 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1802 sbi->s_rsv_window_head.rsv_goal_size = 0;
1803 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1804
1805 /*
1806 * set up enough so that it can read an inode
1807 */
1808 sb->s_op = &ext3_sops;
1809 sb->s_export_op = &ext3_export_ops;
1810 sb->s_xattr = ext3_xattr_handlers;
1811#ifdef CONFIG_QUOTA
1812 sb->s_qcop = &ext3_qctl_operations;
1813 sb->dq_op = &ext3_quota_operations;
1814#endif
1815 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1816
1817 sb->s_root = NULL;
1818
1819 needs_recovery = (es->s_last_orphan != 0 ||
1820 EXT3_HAS_INCOMPAT_FEATURE(sb,
1821 EXT3_FEATURE_INCOMPAT_RECOVER));
1822
1823 /*
1824 * The first inode we look at is the journal inode. Don't try
1825 * root first: it may be modified in the journal!
1826 */
1827 if (!test_opt(sb, NOLOAD) &&
1828 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
Johann Lombardi71b96252006-01-08 01:03:20 -08001829 if (ext3_load_journal(sb, es, journal_devnum))
Mingming Cao0216bfc2006-06-23 02:05:41 -07001830 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001831 } else if (journal_inum) {
1832 if (ext3_create_journal(sb, es, journal_inum))
Mingming Cao0216bfc2006-06-23 02:05:41 -07001833 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834 } else {
1835 if (!silent)
1836 printk (KERN_ERR
1837 "ext3: No journal on filesystem on %s\n",
1838 sb->s_id);
Mingming Cao0216bfc2006-06-23 02:05:41 -07001839 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840 }
1841
1842 /* We have now updated the journal if required, so we can
1843 * validate the data journaling mode. */
1844 switch (test_opt(sb, DATA_FLAGS)) {
1845 case 0:
1846 /* No mode set, assume a default based on the journal
1847 capabilities: ORDERED_DATA if the journal can
1848 cope, else JOURNAL_DATA */
1849 if (journal_check_available_features
1850 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1851 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1852 else
1853 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1854 break;
1855
1856 case EXT3_MOUNT_ORDERED_DATA:
1857 case EXT3_MOUNT_WRITEBACK_DATA:
1858 if (!journal_check_available_features
1859 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1860 printk(KERN_ERR "EXT3-fs: Journal does not support "
1861 "requested data journaling mode\n");
Mingming Cao0216bfc2006-06-23 02:05:41 -07001862 goto failed_mount4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 }
1864 default:
1865 break;
1866 }
1867
1868 if (test_opt(sb, NOBH)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1870 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1871 "its supported only with writeback mode\n");
1872 clear_opt(sbi->s_mount_opt, NOBH);
1873 }
1874 }
1875 /*
1876 * The journal_load will have done any necessary log recovery,
1877 * so we can safely mount the rest of the filesystem now.
1878 */
1879
David Howells473043d2008-02-07 00:15:36 -08001880 root = ext3_iget(sb, EXT3_ROOT_INO);
1881 if (IS_ERR(root)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
David Howells473043d2008-02-07 00:15:36 -08001883 ret = PTR_ERR(root);
Mingming Cao0216bfc2006-06-23 02:05:41 -07001884 goto failed_mount4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885 }
1886 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
David Howells473043d2008-02-07 00:15:36 -08001887 iput(root);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
Mingming Cao0216bfc2006-06-23 02:05:41 -07001889 goto failed_mount4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001890 }
David Howells473043d2008-02-07 00:15:36 -08001891 sb->s_root = d_alloc_root(root);
1892 if (!sb->s_root) {
1893 printk(KERN_ERR "EXT3-fs: get root dentry failed\n");
1894 iput(root);
1895 ret = -ENOMEM;
1896 goto failed_mount4;
1897 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001898
1899 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1900 /*
1901 * akpm: core read_super() calls in here with the superblock locked.
1902 * That deadlocks, because orphan cleanup needs to lock the superblock
1903 * in numerous places. Here we just pop the lock - it's relatively
1904 * harmless, because we are now ready to accept write_super() requests,
1905 * and aviro says that's the only reason for hanging onto the
1906 * superblock lock.
1907 */
1908 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1909 ext3_orphan_cleanup(sb, es);
1910 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1911 if (needs_recovery)
1912 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1913 ext3_mark_recovery_complete(sb, es);
1914 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1915 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1916 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1917 "writeback");
1918
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919 lock_kernel();
1920 return 0;
1921
1922cantfind_ext3:
1923 if (!silent)
1924 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1925 sb->s_id);
1926 goto failed_mount;
1927
Mingming Cao0216bfc2006-06-23 02:05:41 -07001928failed_mount4:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929 journal_destroy(sbi->s_journal);
Mingming Cao0216bfc2006-06-23 02:05:41 -07001930failed_mount3:
1931 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1932 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1933 percpu_counter_destroy(&sbi->s_dirs_counter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934failed_mount2:
1935 for (i = 0; i < db_count; i++)
1936 brelse(sbi->s_group_desc[i]);
1937 kfree(sbi->s_group_desc);
1938failed_mount:
1939#ifdef CONFIG_QUOTA
1940 for (i = 0; i < MAXQUOTAS; i++)
1941 kfree(sbi->s_qf_names[i]);
1942#endif
1943 ext3_blkdev_remove(sbi);
1944 brelse(bh);
1945out_fail:
1946 sb->s_fs_info = NULL;
1947 kfree(sbi);
1948 lock_kernel();
David Howells473043d2008-02-07 00:15:36 -08001949 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001950}
1951
1952/*
1953 * Setup any per-fs journal parameters now. We'll do this both on
1954 * initial mount, once the journal has been initialised but before we've
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001955 * done any recovery; and again on any subsequent remount.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956 */
1957static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1958{
1959 struct ext3_sb_info *sbi = EXT3_SB(sb);
1960
1961 if (sbi->s_commit_interval)
1962 journal->j_commit_interval = sbi->s_commit_interval;
1963 /* We could also set up an ext3-specific default for the commit
1964 * interval here, but for now we'll just fall back to the jbd
1965 * default. */
1966
1967 spin_lock(&journal->j_state_lock);
1968 if (test_opt(sb, BARRIER))
1969 journal->j_flags |= JFS_BARRIER;
1970 else
1971 journal->j_flags &= ~JFS_BARRIER;
1972 spin_unlock(&journal->j_state_lock);
1973}
1974
Eric Sandeeneee194e2006-09-27 01:49:30 -07001975static journal_t *ext3_get_journal(struct super_block *sb,
1976 unsigned int journal_inum)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977{
1978 struct inode *journal_inode;
1979 journal_t *journal;
1980
1981 /* First, test for the existence of a valid inode on disk. Bad
1982 * things happen if we iget() an unused inode, as the subsequent
1983 * iput() will try to delete it. */
1984
David Howells473043d2008-02-07 00:15:36 -08001985 journal_inode = ext3_iget(sb, journal_inum);
1986 if (IS_ERR(journal_inode)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987 printk(KERN_ERR "EXT3-fs: no journal found.\n");
1988 return NULL;
1989 }
1990 if (!journal_inode->i_nlink) {
1991 make_bad_inode(journal_inode);
1992 iput(journal_inode);
1993 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
1994 return NULL;
1995 }
1996
1997 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1998 journal_inode, journal_inode->i_size);
David Howells473043d2008-02-07 00:15:36 -08001999 if (!S_ISREG(journal_inode->i_mode)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
2001 iput(journal_inode);
2002 return NULL;
2003 }
2004
2005 journal = journal_init_inode(journal_inode);
2006 if (!journal) {
2007 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
2008 iput(journal_inode);
2009 return NULL;
2010 }
2011 journal->j_private = sb;
2012 ext3_init_journal_params(sb, journal);
2013 return journal;
2014}
2015
2016static journal_t *ext3_get_dev_journal(struct super_block *sb,
2017 dev_t j_dev)
2018{
2019 struct buffer_head * bh;
2020 journal_t *journal;
Mingming Cao43d23f92006-06-25 05:48:07 -07002021 ext3_fsblk_t start;
Mingming Cao1c2bf372006-06-25 05:48:06 -07002022 ext3_fsblk_t len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023 int hblock, blocksize;
Mingming Cao43d23f92006-06-25 05:48:07 -07002024 ext3_fsblk_t sb_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025 unsigned long offset;
2026 struct ext3_super_block * es;
2027 struct block_device *bdev;
2028
2029 bdev = ext3_blkdev_get(j_dev);
2030 if (bdev == NULL)
2031 return NULL;
2032
2033 if (bd_claim(bdev, sb)) {
2034 printk(KERN_ERR
2035 "EXT3: failed to claim external journal device.\n");
2036 blkdev_put(bdev);
2037 return NULL;
2038 }
2039
2040 blocksize = sb->s_blocksize;
2041 hblock = bdev_hardsect_size(bdev);
2042 if (blocksize < hblock) {
2043 printk(KERN_ERR
2044 "EXT3-fs: blocksize too small for journal device.\n");
2045 goto out_bdev;
2046 }
2047
2048 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2049 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2050 set_blocksize(bdev, blocksize);
2051 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2052 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
2053 "external journal\n");
2054 goto out_bdev;
2055 }
2056
2057 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2058 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2059 !(le32_to_cpu(es->s_feature_incompat) &
2060 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2061 printk(KERN_ERR "EXT3-fs: external journal has "
2062 "bad superblock\n");
2063 brelse(bh);
2064 goto out_bdev;
2065 }
2066
2067 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2068 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
2069 brelse(bh);
2070 goto out_bdev;
2071 }
2072
2073 len = le32_to_cpu(es->s_blocks_count);
2074 start = sb_block + 1;
2075 brelse(bh); /* we're done with the superblock */
2076
2077 journal = journal_init_dev(bdev, sb->s_bdev,
2078 start, len, blocksize);
2079 if (!journal) {
2080 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
2081 goto out_bdev;
2082 }
2083 journal->j_private = sb;
2084 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2085 wait_on_buffer(journal->j_sb_buffer);
2086 if (!buffer_uptodate(journal->j_sb_buffer)) {
2087 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
2088 goto out_journal;
2089 }
2090 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2091 printk(KERN_ERR "EXT3-fs: External journal has more than one "
2092 "user (unsupported) - %d\n",
2093 be32_to_cpu(journal->j_superblock->s_nr_users));
2094 goto out_journal;
2095 }
2096 EXT3_SB(sb)->journal_bdev = bdev;
2097 ext3_init_journal_params(sb, journal);
2098 return journal;
2099out_journal:
2100 journal_destroy(journal);
2101out_bdev:
2102 ext3_blkdev_put(bdev);
2103 return NULL;
2104}
2105
Johann Lombardi71b96252006-01-08 01:03:20 -08002106static int ext3_load_journal(struct super_block *sb,
2107 struct ext3_super_block *es,
2108 unsigned long journal_devnum)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109{
2110 journal_t *journal;
Eric Sandeeneee194e2006-09-27 01:49:30 -07002111 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
Johann Lombardi71b96252006-01-08 01:03:20 -08002112 dev_t journal_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002113 int err = 0;
2114 int really_read_only;
2115
Johann Lombardi71b96252006-01-08 01:03:20 -08002116 if (journal_devnum &&
2117 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2118 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2119 "numbers have changed\n");
2120 journal_dev = new_decode_dev(journal_devnum);
2121 } else
2122 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2123
Linus Torvalds1da177e2005-04-16 15:20:36 -07002124 really_read_only = bdev_read_only(sb->s_bdev);
2125
2126 /*
2127 * Are we loading a blank journal or performing recovery after a
2128 * crash? For recovery, we need to check in advance whether we
2129 * can get read-write access to the device.
2130 */
2131
2132 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2133 if (sb->s_flags & MS_RDONLY) {
2134 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2135 "required on readonly filesystem.\n");
2136 if (really_read_only) {
2137 printk(KERN_ERR "EXT3-fs: write access "
2138 "unavailable, cannot proceed.\n");
2139 return -EROFS;
2140 }
2141 printk (KERN_INFO "EXT3-fs: write access will "
2142 "be enabled during recovery.\n");
2143 }
2144 }
2145
2146 if (journal_inum && journal_dev) {
2147 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2148 "and inode journals!\n");
2149 return -EINVAL;
2150 }
2151
2152 if (journal_inum) {
2153 if (!(journal = ext3_get_journal(sb, journal_inum)))
2154 return -EINVAL;
2155 } else {
2156 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2157 return -EINVAL;
2158 }
2159
2160 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2161 err = journal_update_format(journal);
2162 if (err) {
2163 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2164 journal_destroy(journal);
2165 return err;
2166 }
2167 }
2168
2169 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2170 err = journal_wipe(journal, !really_read_only);
2171 if (!err)
2172 err = journal_load(journal);
2173
2174 if (err) {
2175 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2176 journal_destroy(journal);
2177 return err;
2178 }
2179
2180 EXT3_SB(sb)->s_journal = journal;
2181 ext3_clear_journal_err(sb, es);
Johann Lombardi71b96252006-01-08 01:03:20 -08002182
2183 if (journal_devnum &&
2184 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2185 es->s_journal_dev = cpu_to_le32(journal_devnum);
2186 sb->s_dirt = 1;
2187
2188 /* Make sure we flush the recovery flag to disk. */
2189 ext3_commit_super(sb, es, 1);
2190 }
2191
Linus Torvalds1da177e2005-04-16 15:20:36 -07002192 return 0;
2193}
2194
2195static int ext3_create_journal(struct super_block * sb,
2196 struct ext3_super_block * es,
Eric Sandeeneee194e2006-09-27 01:49:30 -07002197 unsigned int journal_inum)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002198{
2199 journal_t *journal;
Borislav Petkov952d9de2007-07-15 23:41:45 -07002200 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002201
2202 if (sb->s_flags & MS_RDONLY) {
2203 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2204 "create journal.\n");
2205 return -EROFS;
2206 }
2207
Borislav Petkov952d9de2007-07-15 23:41:45 -07002208 journal = ext3_get_journal(sb, journal_inum);
2209 if (!journal)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002210 return -EINVAL;
2211
Eric Sandeeneee194e2006-09-27 01:49:30 -07002212 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002213 journal_inum);
2214
Borislav Petkov952d9de2007-07-15 23:41:45 -07002215 err = journal_create(journal);
2216 if (err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2218 journal_destroy(journal);
2219 return -EIO;
2220 }
2221
2222 EXT3_SB(sb)->s_journal = journal;
2223
2224 ext3_update_dynamic_rev(sb);
2225 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2226 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2227
2228 es->s_journal_inum = cpu_to_le32(journal_inum);
2229 sb->s_dirt = 1;
2230
2231 /* Make sure we flush the recovery flag to disk. */
2232 ext3_commit_super(sb, es, 1);
2233
2234 return 0;
2235}
2236
2237static void ext3_commit_super (struct super_block * sb,
2238 struct ext3_super_block * es,
2239 int sync)
2240{
2241 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2242
2243 if (!sbh)
2244 return;
2245 es->s_wtime = cpu_to_le32(get_seconds());
2246 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2247 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2248 BUFFER_TRACE(sbh, "marking dirty");
2249 mark_buffer_dirty(sbh);
2250 if (sync)
2251 sync_dirty_buffer(sbh);
2252}
2253
2254
2255/*
2256 * Have we just finished recovery? If so, and if we are mounting (or
2257 * remounting) the filesystem readonly, then we will end up with a
2258 * consistent fs on disk. Record that fact.
2259 */
2260static void ext3_mark_recovery_complete(struct super_block * sb,
2261 struct ext3_super_block * es)
2262{
2263 journal_t *journal = EXT3_SB(sb)->s_journal;
2264
2265 journal_lock_updates(journal);
2266 journal_flush(journal);
Jan Kara030703e2007-07-15 23:41:08 -07002267 lock_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002268 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2269 sb->s_flags & MS_RDONLY) {
2270 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2271 sb->s_dirt = 0;
2272 ext3_commit_super(sb, es, 1);
2273 }
Jan Kara030703e2007-07-15 23:41:08 -07002274 unlock_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002275 journal_unlock_updates(journal);
2276}
2277
2278/*
2279 * If we are mounting (or read-write remounting) a filesystem whose journal
2280 * has recorded an error from a previous lifetime, move that error to the
2281 * main filesystem now.
2282 */
2283static void ext3_clear_journal_err(struct super_block * sb,
2284 struct ext3_super_block * es)
2285{
2286 journal_t *journal;
2287 int j_errno;
2288 const char *errstr;
2289
2290 journal = EXT3_SB(sb)->s_journal;
2291
2292 /*
2293 * Now check for any error status which may have been recorded in the
2294 * journal by a prior ext3_error() or ext3_abort()
2295 */
2296
2297 j_errno = journal_errno(journal);
2298 if (j_errno) {
2299 char nbuf[16];
2300
2301 errstr = ext3_decode_error(sb, j_errno, nbuf);
2302 ext3_warning(sb, __FUNCTION__, "Filesystem error recorded "
2303 "from previous mount: %s", errstr);
2304 ext3_warning(sb, __FUNCTION__, "Marking fs in need of "
2305 "filesystem check.");
2306
2307 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2308 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2309 ext3_commit_super (sb, es, 1);
2310
2311 journal_clear_err(journal);
2312 }
2313}
2314
2315/*
2316 * Force the running and committing transactions to commit,
2317 * and wait on the commit.
2318 */
2319int ext3_force_commit(struct super_block *sb)
2320{
2321 journal_t *journal;
2322 int ret;
2323
2324 if (sb->s_flags & MS_RDONLY)
2325 return 0;
2326
2327 journal = EXT3_SB(sb)->s_journal;
2328 sb->s_dirt = 0;
2329 ret = ext3_journal_force_commit(journal);
2330 return ret;
2331}
2332
2333/*
2334 * Ext3 always journals updates to the superblock itself, so we don't
2335 * have to propagate any other updates to the superblock on disk at this
2336 * point. Just start an async writeback to get the buffers on their way
2337 * to the disk.
2338 *
2339 * This implicitly triggers the writebehind on sync().
2340 */
2341
2342static void ext3_write_super (struct super_block * sb)
2343{
Ingo Molnar7892f2f2006-01-09 15:59:25 -08002344 if (mutex_trylock(&sb->s_lock) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002345 BUG();
2346 sb->s_dirt = 0;
2347}
2348
2349static int ext3_sync_fs(struct super_block *sb, int wait)
2350{
2351 tid_t target;
2352
2353 sb->s_dirt = 0;
2354 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2355 if (wait)
2356 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2357 }
2358 return 0;
2359}
2360
2361/*
2362 * LVM calls this function before a (read-only) snapshot is created. This
2363 * gives us a chance to flush the journal completely and mark the fs clean.
2364 */
2365static void ext3_write_super_lockfs(struct super_block *sb)
2366{
2367 sb->s_dirt = 0;
2368
2369 if (!(sb->s_flags & MS_RDONLY)) {
2370 journal_t *journal = EXT3_SB(sb)->s_journal;
2371
2372 /* Now we set up the journal barrier. */
2373 journal_lock_updates(journal);
2374 journal_flush(journal);
2375
2376 /* Journal blocked and flushed, clear needs_recovery flag. */
2377 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2378 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2379 }
2380}
2381
2382/*
2383 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2384 * flag here, even though the filesystem is not technically dirty yet.
2385 */
2386static void ext3_unlockfs(struct super_block *sb)
2387{
2388 if (!(sb->s_flags & MS_RDONLY)) {
2389 lock_super(sb);
2390 /* Reser the needs_recovery flag before the fs is unlocked. */
2391 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2392 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2393 unlock_super(sb);
2394 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2395 }
2396}
2397
2398static int ext3_remount (struct super_block * sb, int * flags, char * data)
2399{
2400 struct ext3_super_block * es;
2401 struct ext3_sb_info *sbi = EXT3_SB(sb);
Mingming Cao43d23f92006-06-25 05:48:07 -07002402 ext3_fsblk_t n_blocks_count = 0;
Jan Kara08c6a962005-07-12 13:58:28 -07002403 unsigned long old_sb_flags;
2404 struct ext3_mount_options old_opts;
2405 int err;
2406#ifdef CONFIG_QUOTA
2407 int i;
2408#endif
2409
2410 /* Store the original options */
2411 old_sb_flags = sb->s_flags;
2412 old_opts.s_mount_opt = sbi->s_mount_opt;
2413 old_opts.s_resuid = sbi->s_resuid;
2414 old_opts.s_resgid = sbi->s_resgid;
2415 old_opts.s_commit_interval = sbi->s_commit_interval;
2416#ifdef CONFIG_QUOTA
2417 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2418 for (i = 0; i < MAXQUOTAS; i++)
2419 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2420#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002421
2422 /*
2423 * Allow the "check" option to be passed as a remount option.
2424 */
Johann Lombardi71b96252006-01-08 01:03:20 -08002425 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
Jan Kara08c6a962005-07-12 13:58:28 -07002426 err = -EINVAL;
2427 goto restore_opts;
2428 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002429
2430 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2431 ext3_abort(sb, __FUNCTION__, "Abort forced by user");
2432
2433 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2434 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2435
2436 es = sbi->s_es;
2437
2438 ext3_init_journal_params(sb, sbi->s_journal);
2439
2440 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2441 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
Jan Kara08c6a962005-07-12 13:58:28 -07002442 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2443 err = -EROFS;
2444 goto restore_opts;
2445 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002446
2447 if (*flags & MS_RDONLY) {
2448 /*
2449 * First of all, the unconditional stuff we have to do
2450 * to disable replay of the journal when we next remount
2451 */
2452 sb->s_flags |= MS_RDONLY;
2453
2454 /*
2455 * OK, test if we are remounting a valid rw partition
2456 * readonly, and if so set the rdonly flag and then
2457 * mark the partition as valid again.
2458 */
2459 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2460 (sbi->s_mount_state & EXT3_VALID_FS))
2461 es->s_state = cpu_to_le16(sbi->s_mount_state);
2462
Jan Kara030703e2007-07-15 23:41:08 -07002463 /*
2464 * We have to unlock super so that we can wait for
2465 * transactions.
2466 */
2467 unlock_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002468 ext3_mark_recovery_complete(sb, es);
Jan Kara030703e2007-07-15 23:41:08 -07002469 lock_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002470 } else {
2471 __le32 ret;
2472 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2473 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2474 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2475 "remount RDWR because of unsupported "
2476 "optional features (%x).\n",
2477 sb->s_id, le32_to_cpu(ret));
Jan Kara08c6a962005-07-12 13:58:28 -07002478 err = -EROFS;
2479 goto restore_opts;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002480 }
Eric Sandeenea9a05a2007-02-10 01:46:07 -08002481
2482 /*
2483 * If we have an unprocessed orphan list hanging
2484 * around from a previously readonly bdev mount,
2485 * require a full umount/remount for now.
2486 */
2487 if (es->s_last_orphan) {
2488 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2489 "remount RDWR because of unprocessed "
2490 "orphan inode list. Please "
2491 "umount/remount instead.\n",
2492 sb->s_id);
2493 err = -EINVAL;
2494 goto restore_opts;
2495 }
2496
Linus Torvalds1da177e2005-04-16 15:20:36 -07002497 /*
2498 * Mounting a RDONLY partition read-write, so reread
2499 * and store the current valid flag. (It may have
2500 * been changed by e2fsck since we originally mounted
2501 * the partition.)
2502 */
2503 ext3_clear_journal_err(sb, es);
2504 sbi->s_mount_state = le16_to_cpu(es->s_state);
Dave Kleikampa4e4de32006-09-27 01:49:36 -07002505 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
Jan Kara08c6a962005-07-12 13:58:28 -07002506 goto restore_opts;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002507 if (!ext3_setup_super (sb, es, 0))
2508 sb->s_flags &= ~MS_RDONLY;
2509 }
2510 }
Jan Kara08c6a962005-07-12 13:58:28 -07002511#ifdef CONFIG_QUOTA
2512 /* Release old quota file names */
2513 for (i = 0; i < MAXQUOTAS; i++)
2514 if (old_opts.s_qf_names[i] &&
2515 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2516 kfree(old_opts.s_qf_names[i]);
2517#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002518 return 0;
Jan Kara08c6a962005-07-12 13:58:28 -07002519restore_opts:
2520 sb->s_flags = old_sb_flags;
2521 sbi->s_mount_opt = old_opts.s_mount_opt;
2522 sbi->s_resuid = old_opts.s_resuid;
2523 sbi->s_resgid = old_opts.s_resgid;
2524 sbi->s_commit_interval = old_opts.s_commit_interval;
2525#ifdef CONFIG_QUOTA
2526 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2527 for (i = 0; i < MAXQUOTAS; i++) {
2528 if (sbi->s_qf_names[i] &&
2529 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2530 kfree(sbi->s_qf_names[i]);
2531 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2532 }
2533#endif
2534 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002535}
2536
David Howells726c3342006-06-23 02:02:58 -07002537static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002538{
David Howells726c3342006-06-23 02:02:58 -07002539 struct super_block *sb = dentry->d_sb;
Alex Tomas09fe3162006-03-24 03:16:16 -08002540 struct ext3_sb_info *sbi = EXT3_SB(sb);
2541 struct ext3_super_block *es = sbi->s_es;
Pekka Enberg50ee0a32006-12-06 20:35:28 -08002542 u64 fsid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002543
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002544 if (test_opt(sb, MINIX_DF)) {
2545 sbi->s_overhead_last = 0;
2546 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2547 unsigned long ngroups = sbi->s_groups_count, i;
2548 ext3_fsblk_t overhead = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002549 smp_rmb();
2550
2551 /*
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002552 * Compute the overhead (FS structures). This is constant
2553 * for a given filesystem unless the number of block groups
2554 * changes so we cache the previous value until it does.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002555 */
2556
2557 /*
2558 * All of the blocks before first_data_block are
2559 * overhead
2560 */
2561 overhead = le32_to_cpu(es->s_first_data_block);
2562
2563 /*
2564 * Add the overhead attributed to the superblock and
2565 * block group descriptors. If the sparse superblocks
2566 * feature is turned on, then not all groups have this.
2567 */
2568 for (i = 0; i < ngroups; i++) {
2569 overhead += ext3_bg_has_super(sb, i) +
2570 ext3_bg_num_gdb(sb, i);
2571 cond_resched();
2572 }
2573
2574 /*
2575 * Every block group has an inode bitmap, a block
2576 * bitmap, and an inode table.
2577 */
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002578 overhead += ngroups * (2 + sbi->s_itb_per_group);
2579 sbi->s_overhead_last = overhead;
2580 smp_wmb();
2581 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002582 }
2583
2584 buf->f_type = EXT3_SUPER_MAGIC;
2585 buf->f_bsize = sb->s_blocksize;
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002586 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07002587 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002588 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002589 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2590 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2591 buf->f_bavail = 0;
2592 buf->f_files = le32_to_cpu(es->s_inodes_count);
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07002593 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002594 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002595 buf->f_namelen = EXT3_NAME_LEN;
Pekka Enberg50ee0a32006-12-06 20:35:28 -08002596 fsid = le64_to_cpup((void *)es->s_uuid) ^
2597 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2598 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2599 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002600 return 0;
2601}
2602
2603/* Helper function for writing quotas on sync - we need to start transaction before quota file
2604 * is locked for write. Otherwise the are possible deadlocks:
2605 * Process 1 Process 2
2606 * ext3_create() quota_sync()
2607 * journal_start() write_dquot()
Ingo Molnard3be9152006-03-23 03:00:29 -08002608 * DQUOT_INIT() down(dqio_mutex)
2609 * down(dqio_mutex) journal_start()
Linus Torvalds1da177e2005-04-16 15:20:36 -07002610 *
2611 */
2612
2613#ifdef CONFIG_QUOTA
2614
2615static inline struct inode *dquot_to_inode(struct dquot *dquot)
2616{
2617 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2618}
2619
2620static int ext3_dquot_initialize(struct inode *inode, int type)
2621{
2622 handle_t *handle;
2623 int ret, err;
2624
2625 /* We may create quota structure so we need to reserve enough blocks */
Jan Kara1f545872005-06-23 22:01:04 -07002626 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002627 if (IS_ERR(handle))
2628 return PTR_ERR(handle);
2629 ret = dquot_initialize(inode, type);
2630 err = ext3_journal_stop(handle);
2631 if (!ret)
2632 ret = err;
2633 return ret;
2634}
2635
2636static int ext3_dquot_drop(struct inode *inode)
2637{
2638 handle_t *handle;
2639 int ret, err;
2640
2641 /* We may delete quota structure so we need to reserve enough blocks */
Jan Kara1f545872005-06-23 22:01:04 -07002642 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002643 if (IS_ERR(handle))
2644 return PTR_ERR(handle);
2645 ret = dquot_drop(inode);
2646 err = ext3_journal_stop(handle);
2647 if (!ret)
2648 ret = err;
2649 return ret;
2650}
2651
2652static int ext3_write_dquot(struct dquot *dquot)
2653{
2654 int ret, err;
2655 handle_t *handle;
2656 struct inode *inode;
2657
2658 inode = dquot_to_inode(dquot);
2659 handle = ext3_journal_start(inode,
Jan Kara1f545872005-06-23 22:01:04 -07002660 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002661 if (IS_ERR(handle))
2662 return PTR_ERR(handle);
2663 ret = dquot_commit(dquot);
2664 err = ext3_journal_stop(handle);
2665 if (!ret)
2666 ret = err;
2667 return ret;
2668}
2669
2670static int ext3_acquire_dquot(struct dquot *dquot)
2671{
2672 int ret, err;
2673 handle_t *handle;
2674
2675 handle = ext3_journal_start(dquot_to_inode(dquot),
Jan Kara1f545872005-06-23 22:01:04 -07002676 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002677 if (IS_ERR(handle))
2678 return PTR_ERR(handle);
2679 ret = dquot_acquire(dquot);
2680 err = ext3_journal_stop(handle);
2681 if (!ret)
2682 ret = err;
2683 return ret;
2684}
2685
2686static int ext3_release_dquot(struct dquot *dquot)
2687{
2688 int ret, err;
2689 handle_t *handle;
2690
2691 handle = ext3_journal_start(dquot_to_inode(dquot),
Jan Kara1f545872005-06-23 22:01:04 -07002692 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Jan Kara9c3013e2007-09-11 15:23:29 -07002693 if (IS_ERR(handle)) {
2694 /* Release dquot anyway to avoid endless cycle in dqput() */
2695 dquot_release(dquot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002696 return PTR_ERR(handle);
Jan Kara9c3013e2007-09-11 15:23:29 -07002697 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002698 ret = dquot_release(dquot);
2699 err = ext3_journal_stop(handle);
2700 if (!ret)
2701 ret = err;
2702 return ret;
2703}
2704
2705static int ext3_mark_dquot_dirty(struct dquot *dquot)
2706{
2707 /* Are we journalling quotas? */
2708 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2709 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2710 dquot_mark_dquot_dirty(dquot);
2711 return ext3_write_dquot(dquot);
2712 } else {
2713 return dquot_mark_dquot_dirty(dquot);
2714 }
2715}
2716
2717static int ext3_write_info(struct super_block *sb, int type)
2718{
2719 int ret, err;
2720 handle_t *handle;
2721
2722 /* Data block + inode block */
2723 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2724 if (IS_ERR(handle))
2725 return PTR_ERR(handle);
2726 ret = dquot_commit_info(sb, type);
2727 err = ext3_journal_stop(handle);
2728 if (!ret)
2729 ret = err;
2730 return ret;
2731}
2732
2733/*
2734 * Turn on quotas during mount time - we need to find
2735 * the quota file and such...
2736 */
2737static int ext3_quota_on_mount(struct super_block *sb, int type)
2738{
Christoph Hellwig84de8562005-06-23 00:09:16 -07002739 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2740 EXT3_SB(sb)->s_jquota_fmt, type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002741}
2742
2743/*
2744 * Standard function to be called on quota_on
2745 */
2746static int ext3_quota_on(struct super_block *sb, int type, int format_id,
Jan Kara2fd83a42008-04-28 02:14:34 -07002747 char *path, int remount)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002748{
2749 int err;
2750 struct nameidata nd;
2751
Jan Kara1f545872005-06-23 22:01:04 -07002752 if (!test_opt(sb, QUOTA))
2753 return -EINVAL;
Jan Kara2fd83a42008-04-28 02:14:34 -07002754 /* Not journalling quota or remount? */
2755 if ((!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
2756 !EXT3_SB(sb)->s_qf_names[GRPQUOTA]) || remount)
2757 return vfs_quota_on(sb, type, format_id, path, remount);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002758 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2759 if (err)
2760 return err;
2761 /* Quotafile not on the same filesystem? */
Jan Blunck4ac91372008-02-14 19:34:32 -08002762 if (nd.path.mnt->mnt_sb != sb) {
Jan Blunck1d957f92008-02-14 19:34:35 -08002763 path_put(&nd.path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002764 return -EXDEV;
2765 }
Jan Kara2fd83a42008-04-28 02:14:34 -07002766 /* Quotafile not in fs root? */
Jan Blunck4ac91372008-02-14 19:34:32 -08002767 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002768 printk(KERN_WARNING
2769 "EXT3-fs: Quota file not on filesystem root. "
2770 "Journalled quota will not work.\n");
Jan Blunck1d957f92008-02-14 19:34:35 -08002771 path_put(&nd.path);
Jan Kara2fd83a42008-04-28 02:14:34 -07002772 return vfs_quota_on(sb, type, format_id, path, remount);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002773}
2774
2775/* Read data from quotafile - avoid pagecache and such because we cannot afford
2776 * acquiring the locks... As quota files are never truncated and quota code
2777 * itself serializes the operations (and noone else should touch the files)
2778 * we don't have to be afraid of races */
2779static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2780 size_t len, loff_t off)
2781{
2782 struct inode *inode = sb_dqopt(sb)->files[type];
2783 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2784 int err = 0;
2785 int offset = off & (sb->s_blocksize - 1);
2786 int tocopy;
2787 size_t toread;
2788 struct buffer_head *bh;
2789 loff_t i_size = i_size_read(inode);
2790
2791 if (off > i_size)
2792 return 0;
2793 if (off+len > i_size)
2794 len = i_size-off;
2795 toread = len;
2796 while (toread > 0) {
2797 tocopy = sb->s_blocksize - offset < toread ?
2798 sb->s_blocksize - offset : toread;
2799 bh = ext3_bread(NULL, inode, blk, 0, &err);
2800 if (err)
2801 return err;
2802 if (!bh) /* A hole? */
2803 memset(data, 0, tocopy);
2804 else
2805 memcpy(data, bh->b_data+offset, tocopy);
2806 brelse(bh);
2807 offset = 0;
2808 toread -= tocopy;
2809 data += tocopy;
2810 blk++;
2811 }
2812 return len;
2813}
2814
2815/* Write to quotafile (we know the transaction is already started and has
2816 * enough credits) */
2817static ssize_t ext3_quota_write(struct super_block *sb, int type,
2818 const char *data, size_t len, loff_t off)
2819{
2820 struct inode *inode = sb_dqopt(sb)->files[type];
2821 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2822 int err = 0;
2823 int offset = off & (sb->s_blocksize - 1);
2824 int tocopy;
2825 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2826 size_t towrite = len;
2827 struct buffer_head *bh;
2828 handle_t *handle = journal_current_handle();
2829
Jan Kara9c3013e2007-09-11 15:23:29 -07002830 if (!handle) {
2831 printk(KERN_WARNING "EXT3-fs: Quota write (off=%Lu, len=%Lu)"
2832 " cancelled because transaction is not started.\n",
2833 (unsigned long long)off, (unsigned long long)len);
2834 return -EIO;
2835 }
Arjan van de Ven5c81a412006-07-03 00:25:20 -07002836 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002837 while (towrite > 0) {
2838 tocopy = sb->s_blocksize - offset < towrite ?
2839 sb->s_blocksize - offset : towrite;
2840 bh = ext3_bread(handle, inode, blk, 1, &err);
2841 if (!bh)
2842 goto out;
2843 if (journal_quota) {
2844 err = ext3_journal_get_write_access(handle, bh);
2845 if (err) {
2846 brelse(bh);
2847 goto out;
2848 }
2849 }
2850 lock_buffer(bh);
2851 memcpy(bh->b_data+offset, data, tocopy);
2852 flush_dcache_page(bh->b_page);
2853 unlock_buffer(bh);
2854 if (journal_quota)
2855 err = ext3_journal_dirty_metadata(handle, bh);
2856 else {
2857 /* Always do at least ordered writes for quotas */
2858 err = ext3_journal_dirty_data(handle, bh);
2859 mark_buffer_dirty(bh);
2860 }
2861 brelse(bh);
2862 if (err)
2863 goto out;
2864 offset = 0;
2865 towrite -= tocopy;
2866 data += tocopy;
2867 blk++;
2868 }
2869out:
2870 if (len == towrite)
2871 return err;
2872 if (inode->i_size < off+len-towrite) {
2873 i_size_write(inode, off+len-towrite);
2874 EXT3_I(inode)->i_disksize = inode->i_size;
2875 }
2876 inode->i_version++;
2877 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2878 ext3_mark_inode_dirty(handle, inode);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002879 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002880 return len - towrite;
2881}
2882
2883#endif
2884
David Howells454e2392006-06-23 02:02:57 -07002885static int ext3_get_sb(struct file_system_type *fs_type,
2886 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002887{
David Howells454e2392006-06-23 02:02:57 -07002888 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002889}
2890
2891static struct file_system_type ext3_fs_type = {
2892 .owner = THIS_MODULE,
2893 .name = "ext3",
2894 .get_sb = ext3_get_sb,
2895 .kill_sb = kill_block_super,
2896 .fs_flags = FS_REQUIRES_DEV,
2897};
2898
2899static int __init init_ext3_fs(void)
2900{
2901 int err = init_ext3_xattr();
2902 if (err)
2903 return err;
2904 err = init_inodecache();
2905 if (err)
2906 goto out1;
2907 err = register_filesystem(&ext3_fs_type);
2908 if (err)
2909 goto out;
2910 return 0;
2911out:
2912 destroy_inodecache();
2913out1:
Dave Kleikampe9ad5622006-09-27 01:49:35 -07002914 exit_ext3_xattr();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002915 return err;
2916}
2917
2918static void __exit exit_ext3_fs(void)
2919{
2920 unregister_filesystem(&ext3_fs_type);
2921 destroy_inodecache();
2922 exit_ext3_xattr();
2923}
2924
2925MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2926MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2927MODULE_LICENSE("GPL");
2928module_init(init_ext3_fs)
2929module_exit(exit_ext3_fs)