Linux Kernel  3.7.1
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
super.c
Go to the documentation of this file.
1 /*
2  * linux/fs/ext3/super.c
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card ([email protected])
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 ([email protected]), 1995
17  */
18 
19 #include <linux/module.h>
20 #include <linux/blkdev.h>
21 #include <linux/parser.h>
22 #include <linux/exportfs.h>
23 #include <linux/statfs.h>
24 #include <linux/random.h>
25 #include <linux/mount.h>
26 #include <linux/quotaops.h>
27 #include <linux/seq_file.h>
28 #include <linux/log2.h>
29 #include <linux/cleancache.h>
30 
31 #include <asm/uaccess.h>
32 
33 #define CREATE_TRACE_POINTS
34 
35 #include "ext3.h"
36 #include "xattr.h"
37 #include "acl.h"
38 #include "namei.h"
39 
40 #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
41  #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
42 #else
43  #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
44 #endif
45 
46 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
47  unsigned long journal_devnum);
48 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
49  unsigned int);
50 static int ext3_commit_super(struct super_block *sb,
51  struct ext3_super_block *es,
52  int sync);
53 static void ext3_mark_recovery_complete(struct super_block * sb,
54  struct ext3_super_block * es);
55 static void ext3_clear_journal_err(struct super_block * sb,
56  struct ext3_super_block * es);
57 static int ext3_sync_fs(struct super_block *sb, int wait);
58 static const char *ext3_decode_error(struct super_block * sb, int errno,
59  char nbuf[16]);
60 static int ext3_remount (struct super_block * sb, int * flags, char * data);
61 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
62 static int ext3_unfreeze(struct super_block *sb);
63 static int ext3_freeze(struct super_block *sb);
64 
65 /*
66  * Wrappers for journal_start/end.
67  */
68 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
69 {
70  journal_t *journal;
71 
72  if (sb->s_flags & MS_RDONLY)
73  return ERR_PTR(-EROFS);
74 
75  /* Special case here: if the journal has aborted behind our
76  * backs (eg. EIO in the commit thread), then we still need to
77  * take the FS itself readonly cleanly. */
78  journal = EXT3_SB(sb)->s_journal;
79  if (is_journal_aborted(journal)) {
80  ext3_abort(sb, __func__,
81  "Detected aborted journal");
82  return ERR_PTR(-EROFS);
83  }
84 
85  return journal_start(journal, nblocks);
86 }
87 
88 int __ext3_journal_stop(const char *where, handle_t *handle)
89 {
90  struct super_block *sb;
91  int err;
92  int rc;
93 
94  sb = handle->h_transaction->t_journal->j_private;
95  err = handle->h_err;
96  rc = journal_stop(handle);
97 
98  if (!err)
99  err = rc;
100  if (err)
101  __ext3_std_error(sb, where, err);
102  return err;
103 }
104 
105 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
106  struct buffer_head *bh, handle_t *handle, int err)
107 {
108  char nbuf[16];
109  const char *errstr = ext3_decode_error(NULL, err, nbuf);
110 
111  if (bh)
112  BUFFER_TRACE(bh, "abort");
113 
114  if (!handle->h_err)
115  handle->h_err = err;
116 
117  if (is_handle_aborted(handle))
118  return;
119 
120  printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n",
121  caller, errstr, err_fn);
122 
123  journal_abort_handle(handle);
124 }
125 
126 void ext3_msg(struct super_block *sb, const char *prefix,
127  const char *fmt, ...)
128 {
129  struct va_format vaf;
130  va_list args;
131 
132  va_start(args, fmt);
133 
134  vaf.fmt = fmt;
135  vaf.va = &args;
136 
137  printk("%sEXT3-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
138 
139  va_end(args);
140 }
141 
142 /* Deal with the reporting of failure conditions on a filesystem such as
143  * inconsistencies detected or read IO failures.
144  *
145  * On ext2, we can store the error state of the filesystem in the
146  * superblock. That is not possible on ext3, because we may have other
147  * write ordering constraints on the superblock which prevent us from
148  * writing it out straight away; and given that the journal is about to
149  * be aborted, we can't rely on the current, or future, transactions to
150  * write out the superblock safely.
151  *
152  * We'll just use the journal_abort() error code to record an error in
153  * the journal instead. On recovery, the journal will complain about
154  * that error until we've noted it down and cleared it.
155  */
156 
157 static void ext3_handle_error(struct super_block *sb)
158 {
159  struct ext3_super_block *es = EXT3_SB(sb)->s_es;
160 
161  EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
163 
164  if (sb->s_flags & MS_RDONLY)
165  return;
166 
167  if (!test_opt (sb, ERRORS_CONT)) {
168  journal_t *journal = EXT3_SB(sb)->s_journal;
169 
170  set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
171  if (journal)
172  journal_abort(journal, -EIO);
173  }
174  if (test_opt (sb, ERRORS_RO)) {
175  ext3_msg(sb, KERN_CRIT,
176  "error: remounting filesystem read-only");
177  sb->s_flags |= MS_RDONLY;
178  }
179  ext3_commit_super(sb, es, 1);
180  if (test_opt(sb, ERRORS_PANIC))
181  panic("EXT3-fs (%s): panic forced after error\n",
182  sb->s_id);
183 }
184 
185 void ext3_error(struct super_block *sb, const char *function,
186  const char *fmt, ...)
187 {
188  struct va_format vaf;
189  va_list args;
190 
191  va_start(args, fmt);
192 
193  vaf.fmt = fmt;
194  vaf.va = &args;
195 
196  printk(KERN_CRIT "EXT3-fs error (device %s): %s: %pV\n",
197  sb->s_id, function, &vaf);
198 
199  va_end(args);
200 
201  ext3_handle_error(sb);
202 }
203 
204 static const char *ext3_decode_error(struct super_block * sb, int errno,
205  char nbuf[16])
206 {
207  char *errstr = NULL;
208 
209  switch (errno) {
210  case -EIO:
211  errstr = "IO failure";
212  break;
213  case -ENOMEM:
214  errstr = "Out of memory";
215  break;
216  case -EROFS:
217  if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
218  errstr = "Journal has aborted";
219  else
220  errstr = "Readonly filesystem";
221  break;
222  default:
223  /* If the caller passed in an extra buffer for unknown
224  * errors, textualise them now. Else we just return
225  * NULL. */
226  if (nbuf) {
227  /* Check for truncated error codes... */
228  if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
229  errstr = nbuf;
230  }
231  break;
232  }
233 
234  return errstr;
235 }
236 
237 /* __ext3_std_error decodes expected errors from journaling functions
238  * automatically and invokes the appropriate error response. */
239 
240 void __ext3_std_error (struct super_block * sb, const char * function,
241  int errno)
242 {
243  char nbuf[16];
244  const char *errstr;
245 
246  /* Special case: if the error is EROFS, and we're not already
247  * inside a transaction, then there's really no point in logging
248  * an error. */
249  if (errno == -EROFS && journal_current_handle() == NULL &&
250  (sb->s_flags & MS_RDONLY))
251  return;
252 
253  errstr = ext3_decode_error(sb, errno, nbuf);
254  ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
255 
256  ext3_handle_error(sb);
257 }
258 
259 /*
260  * ext3_abort is a much stronger failure handler than ext3_error. The
261  * abort function may be used to deal with unrecoverable failures such
262  * as journal IO errors or ENOMEM at a critical moment in log management.
263  *
264  * We unconditionally force the filesystem into an ABORT|READONLY state,
265  * unless the error response on the fs has been set to panic in which
266  * case we take the easy way out and panic immediately.
267  */
268 
269 void ext3_abort(struct super_block *sb, const char *function,
270  const char *fmt, ...)
271 {
272  struct va_format vaf;
273  va_list args;
274 
275  va_start(args, fmt);
276 
277  vaf.fmt = fmt;
278  vaf.va = &args;
279 
280  printk(KERN_CRIT "EXT3-fs (%s): error: %s: %pV\n",
281  sb->s_id, function, &vaf);
282 
283  va_end(args);
284 
285  if (test_opt(sb, ERRORS_PANIC))
286  panic("EXT3-fs: panic from previous error\n");
287 
288  if (sb->s_flags & MS_RDONLY)
289  return;
290 
291  ext3_msg(sb, KERN_CRIT,
292  "error: remounting filesystem read-only");
293  EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
294  sb->s_flags |= MS_RDONLY;
295  set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
296  if (EXT3_SB(sb)->s_journal)
297  journal_abort(EXT3_SB(sb)->s_journal, -EIO);
298 }
299 
300 void ext3_warning(struct super_block *sb, const char *function,
301  const char *fmt, ...)
302 {
303  struct va_format vaf;
304  va_list args;
305 
306  va_start(args, fmt);
307 
308  vaf.fmt = fmt;
309  vaf.va = &args;
310 
311  printk(KERN_WARNING "EXT3-fs (%s): warning: %s: %pV\n",
312  sb->s_id, function, &vaf);
313 
314  va_end(args);
315 }
316 
318 {
319  struct ext3_super_block *es = EXT3_SB(sb)->s_es;
320 
322  return;
323 
325  "warning: updating to rev %d because of "
326  "new feature flag, running e2fsck is recommended",
328 
332  /* leave es->s_feature_*compat flags alone */
333  /* es->s_uuid will be set by e2fsck if empty */
334 
335  /*
336  * The rest of the superblock fields should be zero, and if not it
337  * means they are likely already in use, so leave them alone. We
338  * can leave it up to e2fsck to clean up any inconsistencies there.
339  */
340 }
341 
342 /*
343  * Open the external journal device
344  */
345 static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
346 {
347  struct block_device *bdev;
348  char b[BDEVNAME_SIZE];
349 
351  if (IS_ERR(bdev))
352  goto fail;
353  return bdev;
354 
355 fail:
356  ext3_msg(sb, "error: failed to open journal device %s: %ld",
357  __bdevname(dev, b), PTR_ERR(bdev));
358 
359  return NULL;
360 }
361 
362 /*
363  * Release the journal device
364  */
365 static int ext3_blkdev_put(struct block_device *bdev)
366 {
368 }
369 
370 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
371 {
372  struct block_device *bdev;
373  int ret = -ENODEV;
374 
375  bdev = sbi->journal_bdev;
376  if (bdev) {
377  ret = ext3_blkdev_put(bdev);
378  sbi->journal_bdev = NULL;
379  }
380  return ret;
381 }
382 
383 static inline struct inode *orphan_list_entry(struct list_head *l)
384 {
385  return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
386 }
387 
388 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
389 {
390  struct list_head *l;
391 
392  ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
393  le32_to_cpu(sbi->s_es->s_last_orphan));
394 
395  ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
396  list_for_each(l, &sbi->s_orphan) {
397  struct inode *inode = orphan_list_entry(l);
398  ext3_msg(sb, KERN_ERR, " "
399  "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
400  inode->i_sb->s_id, inode->i_ino, inode,
401  inode->i_mode, inode->i_nlink,
402  NEXT_ORPHAN(inode));
403  }
404 }
405 
406 static void ext3_put_super (struct super_block * sb)
407 {
408  struct ext3_sb_info *sbi = EXT3_SB(sb);
409  struct ext3_super_block *es = sbi->s_es;
410  int i, err;
411 
414  err = journal_destroy(sbi->s_journal);
415  sbi->s_journal = NULL;
416  if (err < 0)
417  ext3_abort(sb, __func__, "Couldn't clean up the journal");
418 
419  if (!(sb->s_flags & MS_RDONLY)) {
421  es->s_state = cpu_to_le16(sbi->s_mount_state);
422  BUFFER_TRACE(sbi->s_sbh, "marking dirty");
423  mark_buffer_dirty(sbi->s_sbh);
424  ext3_commit_super(sb, es, 1);
425  }
426 
427  for (i = 0; i < sbi->s_gdb_count; i++)
428  brelse(sbi->s_group_desc[i]);
429  kfree(sbi->s_group_desc);
433  brelse(sbi->s_sbh);
434 #ifdef CONFIG_QUOTA
435  for (i = 0; i < MAXQUOTAS; i++)
436  kfree(sbi->s_qf_names[i]);
437 #endif
438 
439  /* Debugging code just in case the in-memory inode orphan list
440  * isn't empty. The on-disk one can be non-empty if we've
441  * detected an error and taken the fs readonly, but the
442  * in-memory list had better be clean by this point. */
443  if (!list_empty(&sbi->s_orphan))
444  dump_orphan_list(sb, sbi);
445  J_ASSERT(list_empty(&sbi->s_orphan));
446 
447  invalidate_bdev(sb->s_bdev);
448  if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
449  /*
450  * Invalidate the journal device's buffers. We don't want them
451  * floating about in memory - the physical journal device may
452  * hotswapped, and it breaks the `ro-after' testing code.
453  */
456  ext3_blkdev_remove(sbi);
457  }
458  sb->s_fs_info = NULL;
459  kfree(sbi->s_blockgroup_lock);
460  kfree(sbi);
461 }
462 
463 static struct kmem_cache *ext3_inode_cachep;
464 
465 /*
466  * Called inside transaction, so use GFP_NOFS
467  */
468 static struct inode *ext3_alloc_inode(struct super_block *sb)
469 {
470  struct ext3_inode_info *ei;
471 
472  ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
473  if (!ei)
474  return NULL;
475  ei->i_block_alloc_info = NULL;
476  ei->vfs_inode.i_version = 1;
477  atomic_set(&ei->i_datasync_tid, 0);
478  atomic_set(&ei->i_sync_tid, 0);
479  return &ei->vfs_inode;
480 }
481 
482 static int ext3_drop_inode(struct inode *inode)
483 {
484  int drop = generic_drop_inode(inode);
485 
486  trace_ext3_drop_inode(inode, drop);
487  return drop;
488 }
489 
490 static void ext3_i_callback(struct rcu_head *head)
491 {
492  struct inode *inode = container_of(head, struct inode, i_rcu);
493  kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
494 }
495 
496 static void ext3_destroy_inode(struct inode *inode)
497 {
498  if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
499  printk("EXT3 Inode %p: orphan list check failed!\n",
500  EXT3_I(inode));
501  print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
502  EXT3_I(inode), sizeof(struct ext3_inode_info),
503  false);
504  dump_stack();
505  }
506  call_rcu(&inode->i_rcu, ext3_i_callback);
507 }
508 
509 static void init_once(void *foo)
510 {
511  struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
512 
513  INIT_LIST_HEAD(&ei->i_orphan);
514 #ifdef CONFIG_EXT3_FS_XATTR
515  init_rwsem(&ei->xattr_sem);
516 #endif
519 }
520 
521 static int init_inodecache(void)
522 {
523  ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
524  sizeof(struct ext3_inode_info),
527  init_once);
528  if (ext3_inode_cachep == NULL)
529  return -ENOMEM;
530  return 0;
531 }
532 
533 static void destroy_inodecache(void)
534 {
535  /*
536  * Make sure all delayed rcu free inodes are flushed before we
537  * destroy cache.
538  */
539  rcu_barrier();
540  kmem_cache_destroy(ext3_inode_cachep);
541 }
542 
543 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
544 {
545 #if defined(CONFIG_QUOTA)
546  struct ext3_sb_info *sbi = EXT3_SB(sb);
547 
548  if (sbi->s_jquota_fmt) {
549  char *fmtname = "";
550 
551  switch (sbi->s_jquota_fmt) {
552  case QFMT_VFS_OLD:
553  fmtname = "vfsold";
554  break;
555  case QFMT_VFS_V0:
556  fmtname = "vfsv0";
557  break;
558  case QFMT_VFS_V1:
559  fmtname = "vfsv1";
560  break;
561  }
562  seq_printf(seq, ",jqfmt=%s", fmtname);
563  }
564 
565  if (sbi->s_qf_names[USRQUOTA])
566  seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
567 
568  if (sbi->s_qf_names[GRPQUOTA])
569  seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
570 
571  if (test_opt(sb, USRQUOTA))
572  seq_puts(seq, ",usrquota");
573 
574  if (test_opt(sb, GRPQUOTA))
575  seq_puts(seq, ",grpquota");
576 #endif
577 }
578 
579 static char *data_mode_string(unsigned long mode)
580 {
581  switch (mode) {
583  return "journal";
585  return "ordered";
587  return "writeback";
588  }
589  return "unknown";
590 }
591 
592 /*
593  * Show an option if
594  * - it's set to a non-default value OR
595  * - if the per-sb default is different from the global default
596  */
597 static int ext3_show_options(struct seq_file *seq, struct dentry *root)
598 {
599  struct super_block *sb = root->d_sb;
600  struct ext3_sb_info *sbi = EXT3_SB(sb);
601  struct ext3_super_block *es = sbi->s_es;
602  unsigned long def_mount_opts;
603 
604  def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
605 
606  if (sbi->s_sb_block != 1)
607  seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
608  if (test_opt(sb, MINIX_DF))
609  seq_puts(seq, ",minixdf");
610  if (test_opt(sb, GRPID))
611  seq_puts(seq, ",grpid");
612  if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
613  seq_puts(seq, ",nogrpid");
614  if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT3_DEF_RESUID)) ||
616  seq_printf(seq, ",resuid=%u",
618  }
619  if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT3_DEF_RESGID)) ||
621  seq_printf(seq, ",resgid=%u",
623  }
624  if (test_opt(sb, ERRORS_RO)) {
625  int def_errors = le16_to_cpu(es->s_errors);
626 
627  if (def_errors == EXT3_ERRORS_PANIC ||
628  def_errors == EXT3_ERRORS_CONTINUE) {
629  seq_puts(seq, ",errors=remount-ro");
630  }
631  }
632  if (test_opt(sb, ERRORS_CONT))
633  seq_puts(seq, ",errors=continue");
634  if (test_opt(sb, ERRORS_PANIC))
635  seq_puts(seq, ",errors=panic");
636  if (test_opt(sb, NO_UID32))
637  seq_puts(seq, ",nouid32");
638  if (test_opt(sb, DEBUG))
639  seq_puts(seq, ",debug");
640 #ifdef CONFIG_EXT3_FS_XATTR
641  if (test_opt(sb, XATTR_USER))
642  seq_puts(seq, ",user_xattr");
643  if (!test_opt(sb, XATTR_USER) &&
644  (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
645  seq_puts(seq, ",nouser_xattr");
646  }
647 #endif
648 #ifdef CONFIG_EXT3_FS_POSIX_ACL
649  if (test_opt(sb, POSIX_ACL))
650  seq_puts(seq, ",acl");
651  if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
652  seq_puts(seq, ",noacl");
653 #endif
654  if (!test_opt(sb, RESERVATION))
655  seq_puts(seq, ",noreservation");
656  if (sbi->s_commit_interval) {
657  seq_printf(seq, ",commit=%u",
658  (unsigned) (sbi->s_commit_interval / HZ));
659  }
660 
661  /*
662  * Always display barrier state so it's clear what the status is.
663  */
664  seq_puts(seq, ",barrier=");
665  seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
666  seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
667  if (test_opt(sb, DATA_ERR_ABORT))
668  seq_puts(seq, ",data_err=abort");
669 
670  if (test_opt(sb, NOLOAD))
671  seq_puts(seq, ",norecovery");
672 
673  ext3_show_quota_options(seq, sb);
674 
675  return 0;
676 }
677 
678 
679 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
681 {
682  struct inode *inode;
683 
684  if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
685  return ERR_PTR(-ESTALE);
686  if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
687  return ERR_PTR(-ESTALE);
688 
689  /* iget isn't really right if the inode is currently unallocated!!
690  *
691  * ext3_read_inode will return a bad_inode if the inode had been
692  * deleted, so we should be safe.
693  *
694  * Currently we don't know the generation for parent directory, so
695  * a generation of 0 means "accept any"
696  */
697  inode = ext3_iget(sb, ino);
698  if (IS_ERR(inode))
699  return ERR_CAST(inode);
700  if (generation && inode->i_generation != generation) {
701  iput(inode);
702  return ERR_PTR(-ESTALE);
703  }
704 
705  return inode;
706 }
707 
708 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
709  int fh_len, int fh_type)
710 {
711  return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
712  ext3_nfs_get_inode);
713 }
714 
715 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
716  int fh_len, int fh_type)
717 {
718  return generic_fh_to_parent(sb, fid, fh_len, fh_type,
719  ext3_nfs_get_inode);
720 }
721 
722 /*
723  * Try to release metadata pages (indirect blocks, directories) which are
724  * mapped via the block device. Since these pages could have journal heads
725  * which would prevent try_to_free_buffers() from freeing them, we must use
726  * jbd layer's try_to_free_buffers() function to release them.
727  */
728 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
729  gfp_t wait)
730 {
731  journal_t *journal = EXT3_SB(sb)->s_journal;
732 
733  WARN_ON(PageChecked(page));
734  if (!page_has_buffers(page))
735  return 0;
736  if (journal)
737  return journal_try_to_free_buffers(journal, page,
738  wait & ~__GFP_WAIT);
739  return try_to_free_buffers(page);
740 }
741 
742 #ifdef CONFIG_QUOTA
743 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
744 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
745 
746 static int ext3_write_dquot(struct dquot *dquot);
747 static int ext3_acquire_dquot(struct dquot *dquot);
748 static int ext3_release_dquot(struct dquot *dquot);
749 static int ext3_mark_dquot_dirty(struct dquot *dquot);
750 static int ext3_write_info(struct super_block *sb, int type);
751 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
752  struct path *path);
753 static int ext3_quota_on_mount(struct super_block *sb, int type);
754 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
755  size_t len, loff_t off);
756 static ssize_t ext3_quota_write(struct super_block *sb, int type,
757  const char *data, size_t len, loff_t off);
758 
759 static const struct dquot_operations ext3_quota_operations = {
760  .write_dquot = ext3_write_dquot,
761  .acquire_dquot = ext3_acquire_dquot,
762  .release_dquot = ext3_release_dquot,
763  .mark_dirty = ext3_mark_dquot_dirty,
764  .write_info = ext3_write_info,
765  .alloc_dquot = dquot_alloc,
766  .destroy_dquot = dquot_destroy,
767 };
768 
769 static const struct quotactl_ops ext3_qctl_operations = {
770  .quota_on = ext3_quota_on,
771  .quota_off = dquot_quota_off,
772  .quota_sync = dquot_quota_sync,
773  .get_info = dquot_get_dqinfo,
774  .set_info = dquot_set_dqinfo,
775  .get_dqblk = dquot_get_dqblk,
776  .set_dqblk = dquot_set_dqblk
777 };
778 #endif
779 
780 static const struct super_operations ext3_sops = {
781  .alloc_inode = ext3_alloc_inode,
782  .destroy_inode = ext3_destroy_inode,
783  .write_inode = ext3_write_inode,
784  .dirty_inode = ext3_dirty_inode,
785  .drop_inode = ext3_drop_inode,
786  .evict_inode = ext3_evict_inode,
787  .put_super = ext3_put_super,
788  .sync_fs = ext3_sync_fs,
789  .freeze_fs = ext3_freeze,
790  .unfreeze_fs = ext3_unfreeze,
791  .statfs = ext3_statfs,
792  .remount_fs = ext3_remount,
793  .show_options = ext3_show_options,
794 #ifdef CONFIG_QUOTA
795  .quota_read = ext3_quota_read,
796  .quota_write = ext3_quota_write,
797 #endif
798  .bdev_try_to_free_page = bdev_try_to_free_page,
799 };
800 
801 static const struct export_operations ext3_export_ops = {
802  .fh_to_dentry = ext3_fh_to_dentry,
803  .fh_to_parent = ext3_fh_to_parent,
804  .get_parent = ext3_get_parent,
805 };
806 
807 enum {
820 };
821 
822 static const match_table_t tokens = {
823  {Opt_bsd_df, "bsddf"},
824  {Opt_minix_df, "minixdf"},
825  {Opt_grpid, "grpid"},
826  {Opt_grpid, "bsdgroups"},
827  {Opt_nogrpid, "nogrpid"},
828  {Opt_nogrpid, "sysvgroups"},
829  {Opt_resgid, "resgid=%u"},
830  {Opt_resuid, "resuid=%u"},
831  {Opt_sb, "sb=%u"},
832  {Opt_err_cont, "errors=continue"},
833  {Opt_err_panic, "errors=panic"},
834  {Opt_err_ro, "errors=remount-ro"},
835  {Opt_nouid32, "nouid32"},
836  {Opt_nocheck, "nocheck"},
837  {Opt_nocheck, "check=none"},
838  {Opt_debug, "debug"},
839  {Opt_oldalloc, "oldalloc"},
840  {Opt_orlov, "orlov"},
841  {Opt_user_xattr, "user_xattr"},
842  {Opt_nouser_xattr, "nouser_xattr"},
843  {Opt_acl, "acl"},
844  {Opt_noacl, "noacl"},
845  {Opt_reservation, "reservation"},
846  {Opt_noreservation, "noreservation"},
847  {Opt_noload, "noload"},
848  {Opt_noload, "norecovery"},
849  {Opt_nobh, "nobh"},
850  {Opt_bh, "bh"},
851  {Opt_commit, "commit=%u"},
852  {Opt_journal_update, "journal=update"},
853  {Opt_journal_inum, "journal=%u"},
854  {Opt_journal_dev, "journal_dev=%u"},
855  {Opt_abort, "abort"},
856  {Opt_data_journal, "data=journal"},
857  {Opt_data_ordered, "data=ordered"},
858  {Opt_data_writeback, "data=writeback"},
859  {Opt_data_err_abort, "data_err=abort"},
860  {Opt_data_err_ignore, "data_err=ignore"},
861  {Opt_offusrjquota, "usrjquota="},
862  {Opt_usrjquota, "usrjquota=%s"},
863  {Opt_offgrpjquota, "grpjquota="},
864  {Opt_grpjquota, "grpjquota=%s"},
865  {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
866  {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
867  {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
868  {Opt_grpquota, "grpquota"},
869  {Opt_noquota, "noquota"},
870  {Opt_quota, "quota"},
871  {Opt_usrquota, "usrquota"},
872  {Opt_barrier, "barrier=%u"},
873  {Opt_barrier, "barrier"},
874  {Opt_nobarrier, "nobarrier"},
875  {Opt_resize, "resize"},
876  {Opt_err, NULL},
877 };
878 
879 static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
880 {
881  ext3_fsblk_t sb_block;
882  char *options = (char *) *data;
883 
884  if (!options || strncmp(options, "sb=", 3) != 0)
885  return 1; /* Default location */
886  options += 3;
887  /*todo: use simple_strtoll with >32bit ext3 */
888  sb_block = simple_strtoul(options, &options, 0);
889  if (*options && *options != ',') {
890  ext3_msg(sb, "error: invalid sb specification: %s",
891  (char *) *data);
892  return 1;
893  }
894  if (*options == ',')
895  options++;
896  *data = (void *) options;
897  return sb_block;
898 }
899 
900 #ifdef CONFIG_QUOTA
901 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
902 {
903  struct ext3_sb_info *sbi = EXT3_SB(sb);
904  char *qname;
905 
906  if (sb_any_quota_loaded(sb) &&
907  !sbi->s_qf_names[qtype]) {
908  ext3_msg(sb, KERN_ERR,
909  "Cannot change journaled "
910  "quota options when quota turned on");
911  return 0;
912  }
913  qname = match_strdup(args);
914  if (!qname) {
915  ext3_msg(sb, KERN_ERR,
916  "Not enough memory for storing quotafile name");
917  return 0;
918  }
919  if (sbi->s_qf_names[qtype] &&
920  strcmp(sbi->s_qf_names[qtype], qname)) {
921  ext3_msg(sb, KERN_ERR,
922  "%s quota file already specified", QTYPE2NAME(qtype));
923  kfree(qname);
924  return 0;
925  }
926  sbi->s_qf_names[qtype] = qname;
927  if (strchr(sbi->s_qf_names[qtype], '/')) {
928  ext3_msg(sb, KERN_ERR,
929  "quotafile must be on filesystem root");
930  kfree(sbi->s_qf_names[qtype]);
931  sbi->s_qf_names[qtype] = NULL;
932  return 0;
933  }
934  set_opt(sbi->s_mount_opt, QUOTA);
935  return 1;
936 }
937 
938 static int clear_qf_name(struct super_block *sb, int qtype) {
939 
940  struct ext3_sb_info *sbi = EXT3_SB(sb);
941 
942  if (sb_any_quota_loaded(sb) &&
943  sbi->s_qf_names[qtype]) {
944  ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
945  " when quota turned on");
946  return 0;
947  }
948  /*
949  * The space will be released later when all options are confirmed
950  * to be correct
951  */
952  sbi->s_qf_names[qtype] = NULL;
953  return 1;
954 }
955 #endif
956 
957 static int parse_options (char *options, struct super_block *sb,
958  unsigned int *inum, unsigned long *journal_devnum,
959  ext3_fsblk_t *n_blocks_count, int is_remount)
960 {
961  struct ext3_sb_info *sbi = EXT3_SB(sb);
962  char * p;
964  int data_opt = 0;
965  int option;
966  kuid_t uid;
967  kgid_t gid;
968 #ifdef CONFIG_QUOTA
969  int qfmt;
970 #endif
971 
972  if (!options)
973  return 1;
974 
975  while ((p = strsep (&options, ",")) != NULL) {
976  int token;
977  if (!*p)
978  continue;
979  /*
980  * Initialize args struct so we know whether arg was
981  * found; some options take optional arguments.
982  */
983  args[0].to = args[0].from = NULL;
984  token = match_token(p, tokens, args);
985  switch (token) {
986  case Opt_bsd_df:
987  clear_opt (sbi->s_mount_opt, MINIX_DF);
988  break;
989  case Opt_minix_df:
990  set_opt (sbi->s_mount_opt, MINIX_DF);
991  break;
992  case Opt_grpid:
993  set_opt (sbi->s_mount_opt, GRPID);
994  break;
995  case Opt_nogrpid:
996  clear_opt (sbi->s_mount_opt, GRPID);
997  break;
998  case Opt_resuid:
999  if (match_int(&args[0], &option))
1000  return 0;
1001  uid = make_kuid(current_user_ns(), option);
1002  if (!uid_valid(uid)) {
1003  ext3_msg(sb, KERN_ERR, "Invalid uid value %d", option);
1004  return 0;
1005 
1006  }
1007  sbi->s_resuid = uid;
1008  break;
1009  case Opt_resgid:
1010  if (match_int(&args[0], &option))
1011  return 0;
1012  gid = make_kgid(current_user_ns(), option);
1013  if (!gid_valid(gid)) {
1014  ext3_msg(sb, KERN_ERR, "Invalid gid value %d", option);
1015  return 0;
1016  }
1017  sbi->s_resgid = gid;
1018  break;
1019  case Opt_sb:
1020  /* handled by get_sb_block() instead of here */
1021  /* *sb_block = match_int(&args[0]); */
1022  break;
1023  case Opt_err_panic:
1024  clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1025  clear_opt (sbi->s_mount_opt, ERRORS_RO);
1026  set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1027  break;
1028  case Opt_err_ro:
1029  clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1030  clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1031  set_opt (sbi->s_mount_opt, ERRORS_RO);
1032  break;
1033  case Opt_err_cont:
1034  clear_opt (sbi->s_mount_opt, ERRORS_RO);
1035  clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1036  set_opt (sbi->s_mount_opt, ERRORS_CONT);
1037  break;
1038  case Opt_nouid32:
1039  set_opt (sbi->s_mount_opt, NO_UID32);
1040  break;
1041  case Opt_nocheck:
1042  clear_opt (sbi->s_mount_opt, CHECK);
1043  break;
1044  case Opt_debug:
1045  set_opt (sbi->s_mount_opt, DEBUG);
1046  break;
1047  case Opt_oldalloc:
1048  ext3_msg(sb, KERN_WARNING,
1049  "Ignoring deprecated oldalloc option");
1050  break;
1051  case Opt_orlov:
1052  ext3_msg(sb, KERN_WARNING,
1053  "Ignoring deprecated orlov option");
1054  break;
1055 #ifdef CONFIG_EXT3_FS_XATTR
1056  case Opt_user_xattr:
1057  set_opt (sbi->s_mount_opt, XATTR_USER);
1058  break;
1059  case Opt_nouser_xattr:
1060  clear_opt (sbi->s_mount_opt, XATTR_USER);
1061  break;
1062 #else
1063  case Opt_user_xattr:
1064  case Opt_nouser_xattr:
1065  ext3_msg(sb, KERN_INFO,
1066  "(no)user_xattr options not supported");
1067  break;
1068 #endif
1069 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1070  case Opt_acl:
1071  set_opt(sbi->s_mount_opt, POSIX_ACL);
1072  break;
1073  case Opt_noacl:
1074  clear_opt(sbi->s_mount_opt, POSIX_ACL);
1075  break;
1076 #else
1077  case Opt_acl:
1078  case Opt_noacl:
1079  ext3_msg(sb, KERN_INFO,
1080  "(no)acl options not supported");
1081  break;
1082 #endif
1083  case Opt_reservation:
1084  set_opt(sbi->s_mount_opt, RESERVATION);
1085  break;
1086  case Opt_noreservation:
1087  clear_opt(sbi->s_mount_opt, RESERVATION);
1088  break;
1089  case Opt_journal_update:
1090  /* @@@ FIXME */
1091  /* Eventually we will want to be able to create
1092  a journal file here. For now, only allow the
1093  user to specify an existing inode to be the
1094  journal file. */
1095  if (is_remount) {
1096  ext3_msg(sb, KERN_ERR, "error: cannot specify "
1097  "journal on remount");
1098  return 0;
1099  }
1100  set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1101  break;
1102  case Opt_journal_inum:
1103  if (is_remount) {
1104  ext3_msg(sb, KERN_ERR, "error: cannot specify "
1105  "journal on remount");
1106  return 0;
1107  }
1108  if (match_int(&args[0], &option))
1109  return 0;
1110  *inum = option;
1111  break;
1112  case Opt_journal_dev:
1113  if (is_remount) {
1114  ext3_msg(sb, KERN_ERR, "error: cannot specify "
1115  "journal on remount");
1116  return 0;
1117  }
1118  if (match_int(&args[0], &option))
1119  return 0;
1120  *journal_devnum = option;
1121  break;
1122  case Opt_noload:
1123  set_opt (sbi->s_mount_opt, NOLOAD);
1124  break;
1125  case Opt_commit:
1126  if (match_int(&args[0], &option))
1127  return 0;
1128  if (option < 0)
1129  return 0;
1130  if (option == 0)
1131  option = JBD_DEFAULT_MAX_COMMIT_AGE;
1132  sbi->s_commit_interval = HZ * option;
1133  break;
1134  case Opt_data_journal:
1135  data_opt = EXT3_MOUNT_JOURNAL_DATA;
1136  goto datacheck;
1137  case Opt_data_ordered:
1138  data_opt = EXT3_MOUNT_ORDERED_DATA;
1139  goto datacheck;
1140  case Opt_data_writeback:
1141  data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1142  datacheck:
1143  if (is_remount) {
1144  if (test_opt(sb, DATA_FLAGS) == data_opt)
1145  break;
1146  ext3_msg(sb, KERN_ERR,
1147  "error: cannot change "
1148  "data mode on remount. The filesystem "
1149  "is mounted in data=%s mode and you "
1150  "try to remount it in data=%s mode.",
1151  data_mode_string(test_opt(sb,
1152  DATA_FLAGS)),
1153  data_mode_string(data_opt));
1154  return 0;
1155  } else {
1156  clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1157  sbi->s_mount_opt |= data_opt;
1158  }
1159  break;
1160  case Opt_data_err_abort:
1161  set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1162  break;
1163  case Opt_data_err_ignore:
1164  clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1165  break;
1166 #ifdef CONFIG_QUOTA
1167  case Opt_usrjquota:
1168  if (!set_qf_name(sb, USRQUOTA, &args[0]))
1169  return 0;
1170  break;
1171  case Opt_grpjquota:
1172  if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1173  return 0;
1174  break;
1175  case Opt_offusrjquota:
1176  if (!clear_qf_name(sb, USRQUOTA))
1177  return 0;
1178  break;
1179  case Opt_offgrpjquota:
1180  if (!clear_qf_name(sb, GRPQUOTA))
1181  return 0;
1182  break;
1183  case Opt_jqfmt_vfsold:
1184  qfmt = QFMT_VFS_OLD;
1185  goto set_qf_format;
1186  case Opt_jqfmt_vfsv0:
1187  qfmt = QFMT_VFS_V0;
1188  goto set_qf_format;
1189  case Opt_jqfmt_vfsv1:
1190  qfmt = QFMT_VFS_V1;
1191 set_qf_format:
1192  if (sb_any_quota_loaded(sb) &&
1193  sbi->s_jquota_fmt != qfmt) {
1194  ext3_msg(sb, KERN_ERR, "error: cannot change "
1195  "journaled quota options when "
1196  "quota turned on.");
1197  return 0;
1198  }
1199  sbi->s_jquota_fmt = qfmt;
1200  break;
1201  case Opt_quota:
1202  case Opt_usrquota:
1203  set_opt(sbi->s_mount_opt, QUOTA);
1204  set_opt(sbi->s_mount_opt, USRQUOTA);
1205  break;
1206  case Opt_grpquota:
1207  set_opt(sbi->s_mount_opt, QUOTA);
1208  set_opt(sbi->s_mount_opt, GRPQUOTA);
1209  break;
1210  case Opt_noquota:
1211  if (sb_any_quota_loaded(sb)) {
1212  ext3_msg(sb, KERN_ERR, "error: cannot change "
1213  "quota options when quota turned on.");
1214  return 0;
1215  }
1216  clear_opt(sbi->s_mount_opt, QUOTA);
1219  break;
1220 #else
1221  case Opt_quota:
1222  case Opt_usrquota:
1223  case Opt_grpquota:
1224  ext3_msg(sb, KERN_ERR,
1225  "error: quota options not supported.");
1226  break;
1227  case Opt_usrjquota:
1228  case Opt_grpjquota:
1229  case Opt_offusrjquota:
1230  case Opt_offgrpjquota:
1231  case Opt_jqfmt_vfsold:
1232  case Opt_jqfmt_vfsv0:
1233  case Opt_jqfmt_vfsv1:
1234  ext3_msg(sb, KERN_ERR,
1235  "error: journaled quota options not "
1236  "supported.");
1237  break;
1238  case Opt_noquota:
1239  break;
1240 #endif
1241  case Opt_abort:
1242  set_opt(sbi->s_mount_opt, ABORT);
1243  break;
1244  case Opt_nobarrier:
1245  clear_opt(sbi->s_mount_opt, BARRIER);
1246  break;
1247  case Opt_barrier:
1248  if (args[0].from) {
1249  if (match_int(&args[0], &option))
1250  return 0;
1251  } else
1252  option = 1; /* No argument, default to 1 */
1253  if (option)
1254  set_opt(sbi->s_mount_opt, BARRIER);
1255  else
1256  clear_opt(sbi->s_mount_opt, BARRIER);
1257  break;
1258  case Opt_ignore:
1259  break;
1260  case Opt_resize:
1261  if (!is_remount) {
1262  ext3_msg(sb, KERN_ERR,
1263  "error: resize option only available "
1264  "for remount");
1265  return 0;
1266  }
1267  if (match_int(&args[0], &option) != 0)
1268  return 0;
1269  *n_blocks_count = option;
1270  break;
1271  case Opt_nobh:
1272  ext3_msg(sb, KERN_WARNING,
1273  "warning: ignoring deprecated nobh option");
1274  break;
1275  case Opt_bh:
1276  ext3_msg(sb, KERN_WARNING,
1277  "warning: ignoring deprecated bh option");
1278  break;
1279  default:
1280  ext3_msg(sb, KERN_ERR,
1281  "error: unrecognized mount option \"%s\" "
1282  "or missing value", p);
1283  return 0;
1284  }
1285  }
1286 #ifdef CONFIG_QUOTA
1287  if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1288  if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1290  if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1292 
1293  if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1294  ext3_msg(sb, KERN_ERR, "error: old and new quota "
1295  "format mixing.");
1296  return 0;
1297  }
1298 
1299  if (!sbi->s_jquota_fmt) {
1300  ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1301  "not specified.");
1302  return 0;
1303  }
1304  } else {
1305  if (sbi->s_jquota_fmt) {
1306  ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1307  "specified with no journaling "
1308  "enabled.");
1309  return 0;
1310  }
1311  }
1312 #endif
1313  return 1;
1314 }
1315 
1316 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1317  int read_only)
1318 {
1319  struct ext3_sb_info *sbi = EXT3_SB(sb);
1320  int res = 0;
1321 
1323  ext3_msg(sb, KERN_ERR,
1324  "error: revision level too high, "
1325  "forcing read-only mode");
1326  res = MS_RDONLY;
1327  }
1328  if (read_only)
1329  return res;
1330  if (!(sbi->s_mount_state & EXT3_VALID_FS))
1331  ext3_msg(sb, KERN_WARNING,
1332  "warning: mounting unchecked fs, "
1333  "running e2fsck is recommended");
1334  else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1335  ext3_msg(sb, KERN_WARNING,
1336  "warning: mounting fs with errors, "
1337  "running e2fsck is recommended");
1338  else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1339  le16_to_cpu(es->s_mnt_count) >=
1341  ext3_msg(sb, KERN_WARNING,
1342  "warning: maximal mount count reached, "
1343  "running e2fsck is recommended");
1344  else if (le32_to_cpu(es->s_checkinterval) &&
1345  (le32_to_cpu(es->s_lastcheck) +
1347  ext3_msg(sb, KERN_WARNING,
1348  "warning: checktime reached, "
1349  "running e2fsck is recommended");
1350 #if 0
1351  /* @@@ We _will_ want to clear the valid bit if we find
1352  inconsistencies, to force a fsck at reboot. But for
1353  a plain journaled filesystem we can keep it set as
1354  valid forever! :) */
1356 #endif
1357  if (!le16_to_cpu(es->s_max_mnt_count))
1359  le16_add_cpu(&es->s_mnt_count, 1);
1360  es->s_mtime = cpu_to_le32(get_seconds());
1363 
1364  ext3_commit_super(sb, es, 1);
1365  if (test_opt(sb, DEBUG))
1366  ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1367  "bpg=%lu, ipg=%lu, mo=%04lx]",
1368  sb->s_blocksize,
1369  sbi->s_groups_count,
1372  sbi->s_mount_opt);
1373 
1374  if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1375  char b[BDEVNAME_SIZE];
1376  ext3_msg(sb, KERN_INFO, "using external journal on %s",
1377  bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1378  } else {
1379  ext3_msg(sb, KERN_INFO, "using internal journal");
1380  }
1381  cleancache_init_fs(sb);
1382  return res;
1383 }
1384 
1385 /* Called at mount-time, super-block is locked */
1386 static int ext3_check_descriptors(struct super_block *sb)
1387 {
1388  struct ext3_sb_info *sbi = EXT3_SB(sb);
1389  int i;
1390 
1391  ext3_debug ("Checking group descriptors");
1392 
1393  for (i = 0; i < sbi->s_groups_count; i++) {
1394  struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1395  ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1396  ext3_fsblk_t last_block;
1397 
1398  if (i == sbi->s_groups_count - 1)
1399  last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1400  else
1401  last_block = first_block +
1402  (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1403 
1404  if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1405  le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1406  {
1407  ext3_error (sb, "ext3_check_descriptors",
1408  "Block bitmap for group %d"
1409  " not in group (block %lu)!",
1410  i, (unsigned long)
1412  return 0;
1413  }
1414  if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1415  le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1416  {
1417  ext3_error (sb, "ext3_check_descriptors",
1418  "Inode bitmap for group %d"
1419  " not in group (block %lu)!",
1420  i, (unsigned long)
1422  return 0;
1423  }
1424  if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1425  le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1426  last_block)
1427  {
1428  ext3_error (sb, "ext3_check_descriptors",
1429  "Inode table for group %d"
1430  " not in group (block %lu)!",
1431  i, (unsigned long)
1432  le32_to_cpu(gdp->bg_inode_table));
1433  return 0;
1434  }
1435  }
1436 
1437  sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1438  sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1439  return 1;
1440 }
1441 
1442 
1443 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1444  * the superblock) which were deleted from all directories, but held open by
1445  * a process at the time of a crash. We walk the list and try to delete these
1446  * inodes at recovery time (only with a read-write filesystem).
1447  *
1448  * In order to keep the orphan inode chain consistent during traversal (in
1449  * case of crash during recovery), we link each inode into the superblock
1450  * orphan list_head and handle it the same way as an inode deletion during
1451  * normal operation (which journals the operations for us).
1452  *
1453  * We only do an iget() and an iput() on each inode, which is very safe if we
1454  * accidentally point at an in-use or already deleted inode. The worst that
1455  * can happen in this case is that we get a "bit already cleared" message from
1456  * ext3_free_inode(). The only reason we would point at a wrong inode is if
1457  * e2fsck was run on this filesystem, and it must have already done the orphan
1458  * inode cleanup for us, so we can safely abort without any further action.
1459  */
1460 static void ext3_orphan_cleanup (struct super_block * sb,
1461  struct ext3_super_block * es)
1462 {
1463  unsigned int s_flags = sb->s_flags;
1464  int nr_orphans = 0, nr_truncates = 0;
1465 #ifdef CONFIG_QUOTA
1466  int i;
1467 #endif
1468  if (!es->s_last_orphan) {
1469  jbd_debug(4, "no orphan inodes to clean up\n");
1470  return;
1471  }
1472 
1473  if (bdev_read_only(sb->s_bdev)) {
1474  ext3_msg(sb, KERN_ERR, "error: write access "
1475  "unavailable, skipping orphan cleanup.");
1476  return;
1477  }
1478 
1479  /* Check if feature set allows readwrite operations */
1481  ext3_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
1482  "unknown ROCOMPAT features");
1483  return;
1484  }
1485 
1486  if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1487  /* don't clear list on RO mount w/ errors */
1488  if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
1489  jbd_debug(1, "Errors on filesystem, "
1490  "clearing orphan list.\n");
1491  es->s_last_orphan = 0;
1492  }
1493  jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1494  return;
1495  }
1496 
1497  if (s_flags & MS_RDONLY) {
1498  ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1499  sb->s_flags &= ~MS_RDONLY;
1500  }
1501 #ifdef CONFIG_QUOTA
1502  /* Needed for iput() to work correctly and not trash data */
1503  sb->s_flags |= MS_ACTIVE;
1504  /* Turn on quotas so that they are updated correctly */
1505  for (i = 0; i < MAXQUOTAS; i++) {
1506  if (EXT3_SB(sb)->s_qf_names[i]) {
1507  int ret = ext3_quota_on_mount(sb, i);
1508  if (ret < 0)
1509  ext3_msg(sb, KERN_ERR,
1510  "error: cannot turn on journaled "
1511  "quota: %d", ret);
1512  }
1513  }
1514 #endif
1515 
1516  while (es->s_last_orphan) {
1517  struct inode *inode;
1518 
1519  inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1520  if (IS_ERR(inode)) {
1521  es->s_last_orphan = 0;
1522  break;
1523  }
1524 
1525  list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1526  dquot_initialize(inode);
1527  if (inode->i_nlink) {
1529  "%s: truncating inode %lu to %Ld bytes\n",
1530  __func__, inode->i_ino, inode->i_size);
1531  jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1532  inode->i_ino, inode->i_size);
1533  ext3_truncate(inode);
1534  nr_truncates++;
1535  } else {
1537  "%s: deleting unreferenced inode %lu\n",
1538  __func__, inode->i_ino);
1539  jbd_debug(2, "deleting unreferenced inode %lu\n",
1540  inode->i_ino);
1541  nr_orphans++;
1542  }
1543  iput(inode); /* The delete magic happens here! */
1544  }
1545 
1546 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1547 
1548  if (nr_orphans)
1549  ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1550  PLURAL(nr_orphans));
1551  if (nr_truncates)
1552  ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1553  PLURAL(nr_truncates));
1554 #ifdef CONFIG_QUOTA
1555  /* Turn quotas off */
1556  for (i = 0; i < MAXQUOTAS; i++) {
1557  if (sb_dqopt(sb)->files[i])
1558  dquot_quota_off(sb, i);
1559  }
1560 #endif
1561  sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1562 }
1563 
1564 /*
1565  * Maximal file size. There is a direct, and {,double-,triple-}indirect
1566  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1567  * We need to be 1 filesystem block less than the 2^32 sector limit.
1568  */
1569 static loff_t ext3_max_size(int bits)
1570 {
1571  loff_t res = EXT3_NDIR_BLOCKS;
1572  int meta_blocks;
1573  loff_t upper_limit;
1574 
1575  /* This is calculated to be the largest file size for a
1576  * dense, file such that the total number of
1577  * sectors in the file, including data and all indirect blocks,
1578  * does not exceed 2^32 -1
1579  * __u32 i_blocks representing the total number of
1580  * 512 bytes blocks of the file
1581  */
1582  upper_limit = (1LL << 32) - 1;
1583 
1584  /* total blocks in file system block size */
1585  upper_limit >>= (bits - 9);
1586 
1587 
1588  /* indirect blocks */
1589  meta_blocks = 1;
1590  /* double indirect blocks */
1591  meta_blocks += 1 + (1LL << (bits-2));
1592  /* tripple indirect blocks */
1593  meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1594 
1595  upper_limit -= meta_blocks;
1596  upper_limit <<= bits;
1597 
1598  res += 1LL << (bits-2);
1599  res += 1LL << (2*(bits-2));
1600  res += 1LL << (3*(bits-2));
1601  res <<= bits;
1602  if (res > upper_limit)
1603  res = upper_limit;
1604 
1605  if (res > MAX_LFS_FILESIZE)
1606  res = MAX_LFS_FILESIZE;
1607 
1608  return res;
1609 }
1610 
1611 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1612  ext3_fsblk_t logic_sb_block,
1613  int nr)
1614 {
1615  struct ext3_sb_info *sbi = EXT3_SB(sb);
1616  unsigned long bg, first_meta_bg;
1617  int has_super = 0;
1618 
1619  first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1620 
1622  nr < first_meta_bg)
1623  return (logic_sb_block + nr + 1);
1624  bg = sbi->s_desc_per_block * nr;
1625  if (ext3_bg_has_super(sb, bg))
1626  has_super = 1;
1627  return (has_super + ext3_group_first_block_no(sb, bg));
1628 }
1629 
1630 
1631 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1632 {
1633  struct buffer_head * bh;
1634  struct ext3_super_block *es = NULL;
1635  struct ext3_sb_info *sbi;
1637  ext3_fsblk_t sb_block = get_sb_block(&data, sb);
1638  ext3_fsblk_t logic_sb_block;
1639  unsigned long offset = 0;
1640  unsigned int journal_inum = 0;
1641  unsigned long journal_devnum = 0;
1642  unsigned long def_mount_opts;
1643  struct inode *root;
1644  int blocksize;
1645  int hblock;
1646  int db_count;
1647  int i;
1648  int needs_recovery;
1649  int ret = -EINVAL;
1650  __le32 features;
1651  int err;
1652 
1653  sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1654  if (!sbi)
1655  return -ENOMEM;
1656 
1657  sbi->s_blockgroup_lock =
1658  kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1659  if (!sbi->s_blockgroup_lock) {
1660  kfree(sbi);
1661  return -ENOMEM;
1662  }
1663  sb->s_fs_info = sbi;
1664  sbi->s_mount_opt = 0;
1667  sbi->s_sb_block = sb_block;
1668 
1669  blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1670  if (!blocksize) {
1671  ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1672  goto out_fail;
1673  }
1674 
1675  /*
1676  * The ext3 superblock will not be buffer aligned for other than 1kB
1677  * block sizes. We need to calculate the offset from buffer start.
1678  */
1679  if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1680  logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1681  offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1682  } else {
1683  logic_sb_block = sb_block;
1684  }
1685 
1686  if (!(bh = sb_bread(sb, logic_sb_block))) {
1687  ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1688  goto out_fail;
1689  }
1690  /*
1691  * Note: s_es must be initialized as soon as possible because
1692  * some ext3 macro-instructions depend on its value
1693  */
1694  es = (struct ext3_super_block *) (bh->b_data + offset);
1695  sbi->s_es = es;
1696  sb->s_magic = le16_to_cpu(es->s_magic);
1697  if (sb->s_magic != EXT3_SUPER_MAGIC)
1698  goto cantfind_ext3;
1699 
1700  /* Set defaults before we parse the mount options */
1701  def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1702  if (def_mount_opts & EXT3_DEFM_DEBUG)
1703  set_opt(sbi->s_mount_opt, DEBUG);
1704  if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1705  set_opt(sbi->s_mount_opt, GRPID);
1706  if (def_mount_opts & EXT3_DEFM_UID16)
1707  set_opt(sbi->s_mount_opt, NO_UID32);
1708 #ifdef CONFIG_EXT3_FS_XATTR
1709  if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1710  set_opt(sbi->s_mount_opt, XATTR_USER);
1711 #endif
1712 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1713  if (def_mount_opts & EXT3_DEFM_ACL)
1714  set_opt(sbi->s_mount_opt, POSIX_ACL);
1715 #endif
1716  if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1717  set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1718  else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1719  set_opt(sbi->s_mount_opt, ORDERED_DATA);
1720  else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1721  set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1722 
1723  if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1724  set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1725  else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1726  set_opt(sbi->s_mount_opt, ERRORS_CONT);
1727  else
1728  set_opt(sbi->s_mount_opt, ERRORS_RO);
1729 
1732 
1733  /* enable barriers by default */
1734  set_opt(sbi->s_mount_opt, BARRIER);
1735  set_opt(sbi->s_mount_opt, RESERVATION);
1736 
1737  if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1738  NULL, 0))
1739  goto failed_mount;
1740 
1741  sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1742  (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1743 
1745  (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1746  EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1747  EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1748  ext3_msg(sb, KERN_WARNING,
1749  "warning: feature flags set on rev 0 fs, "
1750  "running e2fsck is recommended");
1751  /*
1752  * Check feature flags regardless of the revision level, since we
1753  * previously didn't change the revision level when setting the flags,
1754  * so there is a chance incompat flags are set on a rev 0 filesystem.
1755  */
1757  if (features) {
1758  ext3_msg(sb, KERN_ERR,
1759  "error: couldn't mount because of unsupported "
1760  "optional features (%x)", le32_to_cpu(features));
1761  goto failed_mount;
1762  }
1764  if (!(sb->s_flags & MS_RDONLY) && features) {
1765  ext3_msg(sb, KERN_ERR,
1766  "error: couldn't mount RDWR because of unsupported "
1767  "optional features (%x)", le32_to_cpu(features));
1768  goto failed_mount;
1769  }
1770  blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1771 
1772  if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1773  blocksize > EXT3_MAX_BLOCK_SIZE) {
1774  ext3_msg(sb, KERN_ERR,
1775  "error: couldn't mount because of unsupported "
1776  "filesystem blocksize %d", blocksize);
1777  goto failed_mount;
1778  }
1779 
1780  hblock = bdev_logical_block_size(sb->s_bdev);
1781  if (sb->s_blocksize != blocksize) {
1782  /*
1783  * Make sure the blocksize for the filesystem is larger
1784  * than the hardware sectorsize for the machine.
1785  */
1786  if (blocksize < hblock) {
1787  ext3_msg(sb, KERN_ERR,
1788  "error: fsblocksize %d too small for "
1789  "hardware sectorsize %d", blocksize, hblock);
1790  goto failed_mount;
1791  }
1792 
1793  brelse (bh);
1794  if (!sb_set_blocksize(sb, blocksize)) {
1795  ext3_msg(sb, KERN_ERR,
1796  "error: bad blocksize %d", blocksize);
1797  goto out_fail;
1798  }
1799  logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1800  offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1801  bh = sb_bread(sb, logic_sb_block);
1802  if (!bh) {
1803  ext3_msg(sb, KERN_ERR,
1804  "error: can't read superblock on 2nd try");
1805  goto failed_mount;
1806  }
1807  es = (struct ext3_super_block *)(bh->b_data + offset);
1808  sbi->s_es = es;
1809  if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1810  ext3_msg(sb, KERN_ERR,
1811  "error: magic mismatch");
1812  goto failed_mount;
1813  }
1814  }
1815 
1816  sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1817 
1821  } else {
1823  sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1824  if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1825  (!is_power_of_2(sbi->s_inode_size)) ||
1826  (sbi->s_inode_size > blocksize)) {
1827  ext3_msg(sb, KERN_ERR,
1828  "error: unsupported inode size: %d",
1829  sbi->s_inode_size);
1830  goto failed_mount;
1831  }
1832  }
1835  if (blocksize != sbi->s_frag_size) {
1836  ext3_msg(sb, KERN_ERR,
1837  "error: fragsize %lu != blocksize %u (unsupported)",
1838  sbi->s_frag_size, blocksize);
1839  goto failed_mount;
1840  }
1841  sbi->s_frags_per_block = 1;
1845  if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1846  goto cantfind_ext3;
1847  sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1848  if (sbi->s_inodes_per_block == 0)
1849  goto cantfind_ext3;
1850  sbi->s_itb_per_group = sbi->s_inodes_per_group /
1851  sbi->s_inodes_per_block;
1852  sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1853  sbi->s_sbh = bh;
1854  sbi->s_mount_state = le16_to_cpu(es->s_state);
1857  for (i=0; i < 4; i++)
1858  sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1860  i = le32_to_cpu(es->s_flags);
1861  if (i & EXT2_FLAGS_UNSIGNED_HASH)
1862  sbi->s_hash_unsigned = 3;
1863  else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1864 #ifdef __CHAR_UNSIGNED__
1865  es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1866  sbi->s_hash_unsigned = 3;
1867 #else
1868  es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1869 #endif
1870  }
1871 
1872  if (sbi->s_blocks_per_group > blocksize * 8) {
1873  ext3_msg(sb, KERN_ERR,
1874  "#blocks per group too big: %lu",
1875  sbi->s_blocks_per_group);
1876  goto failed_mount;
1877  }
1878  if (sbi->s_frags_per_group > blocksize * 8) {
1879  ext3_msg(sb, KERN_ERR,
1880  "error: #fragments per group too big: %lu",
1881  sbi->s_frags_per_group);
1882  goto failed_mount;
1883  }
1884  if (sbi->s_inodes_per_group > blocksize * 8) {
1885  ext3_msg(sb, KERN_ERR,
1886  "error: #inodes per group too big: %lu",
1887  sbi->s_inodes_per_group);
1888  goto failed_mount;
1889  }
1890 
1893  if (err) {
1894  ext3_msg(sb, KERN_ERR,
1895  "error: filesystem is too large to mount safely");
1896  if (sizeof(sector_t) < 8)
1897  ext3_msg(sb, KERN_ERR,
1898  "error: CONFIG_LBDAF not enabled");
1899  ret = err;
1900  goto failed_mount;
1901  }
1902 
1903  if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1904  goto cantfind_ext3;
1905  sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1907  / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1908  db_count = DIV_ROUND_UP(sbi->s_groups_count, EXT3_DESC_PER_BLOCK(sb));
1909  sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1910  GFP_KERNEL);
1911  if (sbi->s_group_desc == NULL) {
1912  ext3_msg(sb, KERN_ERR,
1913  "error: not enough memory");
1914  ret = -ENOMEM;
1915  goto failed_mount;
1916  }
1917 
1918  bgl_lock_init(sbi->s_blockgroup_lock);
1919 
1920  for (i = 0; i < db_count; i++) {
1921  block = descriptor_loc(sb, logic_sb_block, i);
1922  sbi->s_group_desc[i] = sb_bread(sb, block);
1923  if (!sbi->s_group_desc[i]) {
1924  ext3_msg(sb, KERN_ERR,
1925  "error: can't read group descriptor %d", i);
1926  db_count = i;
1927  goto failed_mount2;
1928  }
1929  }
1930  if (!ext3_check_descriptors (sb)) {
1931  ext3_msg(sb, KERN_ERR,
1932  "error: group descriptors corrupted");
1933  goto failed_mount2;
1934  }
1935  sbi->s_gdb_count = db_count;
1936  get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1938 
1939  /* per fileystem reservation list head & lock */
1941  sbi->s_rsv_window_root = RB_ROOT;
1942  /* Add a single, static dummy reservation to the start of the
1943  * reservation window list --- it gives us a placeholder for
1944  * append-at-start-of-list which makes the allocation logic
1945  * _much_ simpler. */
1948  sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1949  sbi->s_rsv_window_head.rsv_goal_size = 0;
1951 
1952  /*
1953  * set up enough so that it can read an inode
1954  */
1955  sb->s_op = &ext3_sops;
1956  sb->s_export_op = &ext3_export_ops;
1958 #ifdef CONFIG_QUOTA
1959  sb->s_qcop = &ext3_qctl_operations;
1960  sb->dq_op = &ext3_quota_operations;
1961 #endif
1962  memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
1963  INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1964  mutex_init(&sbi->s_orphan_lock);
1965  mutex_init(&sbi->s_resize_lock);
1966 
1967  sb->s_root = NULL;
1968 
1969  needs_recovery = (es->s_last_orphan != 0 ||
1972 
1973  /*
1974  * The first inode we look at is the journal inode. Don't try
1975  * root first: it may be modified in the journal!
1976  */
1977  if (!test_opt(sb, NOLOAD) &&
1979  if (ext3_load_journal(sb, es, journal_devnum))
1980  goto failed_mount2;
1981  } else if (journal_inum) {
1982  if (ext3_create_journal(sb, es, journal_inum))
1983  goto failed_mount2;
1984  } else {
1985  if (!silent)
1986  ext3_msg(sb, KERN_ERR,
1987  "error: no journal found. "
1988  "mounting ext3 over ext2?");
1989  goto failed_mount2;
1990  }
1991  err = percpu_counter_init(&sbi->s_freeblocks_counter,
1993  if (!err) {
1994  err = percpu_counter_init(&sbi->s_freeinodes_counter,
1996  }
1997  if (!err) {
1998  err = percpu_counter_init(&sbi->s_dirs_counter,
1999  ext3_count_dirs(sb));
2000  }
2001  if (err) {
2002  ext3_msg(sb, KERN_ERR, "error: insufficient memory");
2003  ret = err;
2004  goto failed_mount3;
2005  }
2006 
2007  /* We have now updated the journal if required, so we can
2008  * validate the data journaling mode. */
2009  switch (test_opt(sb, DATA_FLAGS)) {
2010  case 0:
2011  /* No mode set, assume a default based on the journal
2012  capabilities: ORDERED_DATA if the journal can
2013  cope, else JOURNAL_DATA */
2015  (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
2016  set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
2017  else
2018  set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2019  break;
2020 
2024  (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
2025  ext3_msg(sb, KERN_ERR,
2026  "error: journal does not support "
2027  "requested data journaling mode");
2028  goto failed_mount3;
2029  }
2030  default:
2031  break;
2032  }
2033 
2034  /*
2035  * The journal_load will have done any necessary log recovery,
2036  * so we can safely mount the rest of the filesystem now.
2037  */
2038 
2039  root = ext3_iget(sb, EXT3_ROOT_INO);
2040  if (IS_ERR(root)) {
2041  ext3_msg(sb, KERN_ERR, "error: get root inode failed");
2042  ret = PTR_ERR(root);
2043  goto failed_mount3;
2044  }
2045  if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2046  iput(root);
2047  ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
2048  goto failed_mount3;
2049  }
2050  sb->s_root = d_make_root(root);
2051  if (!sb->s_root) {
2052  ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
2053  ret = -ENOMEM;
2054  goto failed_mount3;
2055  }
2056 
2057  if (ext3_setup_super(sb, es, sb->s_flags & MS_RDONLY))
2058  sb->s_flags |= MS_RDONLY;
2059 
2060  EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2061  ext3_orphan_cleanup(sb, es);
2062  EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
2063  if (needs_recovery) {
2064  ext3_mark_recovery_complete(sb, es);
2065  ext3_msg(sb, KERN_INFO, "recovery complete");
2066  }
2067  ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
2068  test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2069  test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2070  "writeback");
2071 
2072  return 0;
2073 
2074 cantfind_ext3:
2075  if (!silent)
2076  ext3_msg(sb, KERN_INFO,
2077  "error: can't find ext3 filesystem on dev %s.",
2078  sb->s_id);
2079  goto failed_mount;
2080 
2081 failed_mount3:
2085  journal_destroy(sbi->s_journal);
2086 failed_mount2:
2087  for (i = 0; i < db_count; i++)
2088  brelse(sbi->s_group_desc[i]);
2089  kfree(sbi->s_group_desc);
2090 failed_mount:
2091 #ifdef CONFIG_QUOTA
2092  for (i = 0; i < MAXQUOTAS; i++)
2093  kfree(sbi->s_qf_names[i]);
2094 #endif
2095  ext3_blkdev_remove(sbi);
2096  brelse(bh);
2097 out_fail:
2098  sb->s_fs_info = NULL;
2099  kfree(sbi->s_blockgroup_lock);
2100  kfree(sbi);
2101  return ret;
2102 }
2103 
2104 /*
2105  * Setup any per-fs journal parameters now. We'll do this both on
2106  * initial mount, once the journal has been initialised but before we've
2107  * done any recovery; and again on any subsequent remount.
2108  */
2109 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2110 {
2111  struct ext3_sb_info *sbi = EXT3_SB(sb);
2112 
2113  if (sbi->s_commit_interval)
2114  journal->j_commit_interval = sbi->s_commit_interval;
2115  /* We could also set up an ext3-specific default for the commit
2116  * interval here, but for now we'll just fall back to the jbd
2117  * default. */
2118 
2119  spin_lock(&journal->j_state_lock);
2120  if (test_opt(sb, BARRIER))
2121  journal->j_flags |= JFS_BARRIER;
2122  else
2123  journal->j_flags &= ~JFS_BARRIER;
2124  if (test_opt(sb, DATA_ERR_ABORT))
2125  journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2126  else
2127  journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2128  spin_unlock(&journal->j_state_lock);
2129 }
2130 
2131 static journal_t *ext3_get_journal(struct super_block *sb,
2132  unsigned int journal_inum)
2133 {
2134  struct inode *journal_inode;
2135  journal_t *journal;
2136 
2137  /* First, test for the existence of a valid inode on disk. Bad
2138  * things happen if we iget() an unused inode, as the subsequent
2139  * iput() will try to delete it. */
2140 
2141  journal_inode = ext3_iget(sb, journal_inum);
2142  if (IS_ERR(journal_inode)) {
2143  ext3_msg(sb, KERN_ERR, "error: no journal found");
2144  return NULL;
2145  }
2146  if (!journal_inode->i_nlink) {
2147  make_bad_inode(journal_inode);
2148  iput(journal_inode);
2149  ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
2150  return NULL;
2151  }
2152 
2153  jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2154  journal_inode, journal_inode->i_size);
2155  if (!S_ISREG(journal_inode->i_mode)) {
2156  ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
2157  iput(journal_inode);
2158  return NULL;
2159  }
2160 
2161  journal = journal_init_inode(journal_inode);
2162  if (!journal) {
2163  ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
2164  iput(journal_inode);
2165  return NULL;
2166  }
2167  journal->j_private = sb;
2168  ext3_init_journal_params(sb, journal);
2169  return journal;
2170 }
2171 
2172 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2173  dev_t j_dev)
2174 {
2175  struct buffer_head * bh;
2176  journal_t *journal;
2178  ext3_fsblk_t len;
2179  int hblock, blocksize;
2180  ext3_fsblk_t sb_block;
2181  unsigned long offset;
2182  struct ext3_super_block * es;
2183  struct block_device *bdev;
2184 
2185  bdev = ext3_blkdev_get(j_dev, sb);
2186  if (bdev == NULL)
2187  return NULL;
2188 
2189  blocksize = sb->s_blocksize;
2190  hblock = bdev_logical_block_size(bdev);
2191  if (blocksize < hblock) {
2192  ext3_msg(sb, KERN_ERR,
2193  "error: blocksize too small for journal device");
2194  goto out_bdev;
2195  }
2196 
2197  sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2198  offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2199  set_blocksize(bdev, blocksize);
2200  if (!(bh = __bread(bdev, sb_block, blocksize))) {
2201  ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2202  "external journal");
2203  goto out_bdev;
2204  }
2205 
2206  es = (struct ext3_super_block *) (bh->b_data + offset);
2207  if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2210  ext3_msg(sb, KERN_ERR, "error: external journal has "
2211  "bad superblock");
2212  brelse(bh);
2213  goto out_bdev;
2214  }
2215 
2216  if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2217  ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
2218  brelse(bh);
2219  goto out_bdev;
2220  }
2221 
2222  len = le32_to_cpu(es->s_blocks_count);
2223  start = sb_block + 1;
2224  brelse(bh); /* we're done with the superblock */
2225 
2226  journal = journal_init_dev(bdev, sb->s_bdev,
2227  start, len, blocksize);
2228  if (!journal) {
2229  ext3_msg(sb, KERN_ERR,
2230  "error: failed to create device journal");
2231  goto out_bdev;
2232  }
2233  journal->j_private = sb;
2234  if (!bh_uptodate_or_lock(journal->j_sb_buffer)) {
2235  if (bh_submit_read(journal->j_sb_buffer)) {
2236  ext3_msg(sb, KERN_ERR, "I/O error on journal device");
2237  goto out_journal;
2238  }
2239  }
2240  if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2241  ext3_msg(sb, KERN_ERR,
2242  "error: external journal has more than one "
2243  "user (unsupported) - %d",
2244  be32_to_cpu(journal->j_superblock->s_nr_users));
2245  goto out_journal;
2246  }
2247  EXT3_SB(sb)->journal_bdev = bdev;
2248  ext3_init_journal_params(sb, journal);
2249  return journal;
2250 out_journal:
2251  journal_destroy(journal);
2252 out_bdev:
2253  ext3_blkdev_put(bdev);
2254  return NULL;
2255 }
2256 
2257 static int ext3_load_journal(struct super_block *sb,
2258  struct ext3_super_block *es,
2259  unsigned long journal_devnum)
2260 {
2261  journal_t *journal;
2262  unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2263  dev_t journal_dev;
2264  int err = 0;
2265  int really_read_only;
2266 
2267  if (journal_devnum &&
2268  journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2269  ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2270  "numbers have changed");
2271  journal_dev = new_decode_dev(journal_devnum);
2272  } else
2273  journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2274 
2275  really_read_only = bdev_read_only(sb->s_bdev);
2276 
2277  /*
2278  * Are we loading a blank journal or performing recovery after a
2279  * crash? For recovery, we need to check in advance whether we
2280  * can get read-write access to the device.
2281  */
2282 
2284  if (sb->s_flags & MS_RDONLY) {
2285  ext3_msg(sb, KERN_INFO,
2286  "recovery required on readonly filesystem");
2287  if (really_read_only) {
2288  ext3_msg(sb, KERN_ERR, "error: write access "
2289  "unavailable, cannot proceed");
2290  return -EROFS;
2291  }
2292  ext3_msg(sb, KERN_INFO,
2293  "write access will be enabled during recovery");
2294  }
2295  }
2296 
2297  if (journal_inum && journal_dev) {
2298  ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2299  "and inode journals");
2300  return -EINVAL;
2301  }
2302 
2303  if (journal_inum) {
2304  if (!(journal = ext3_get_journal(sb, journal_inum)))
2305  return -EINVAL;
2306  } else {
2307  if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2308  return -EINVAL;
2309  }
2310 
2311  if (!(journal->j_flags & JFS_BARRIER))
2312  printk(KERN_INFO "EXT3-fs: barriers not enabled\n");
2313 
2314  if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2315  err = journal_update_format(journal);
2316  if (err) {
2317  ext3_msg(sb, KERN_ERR, "error updating journal");
2318  journal_destroy(journal);
2319  return err;
2320  }
2321  }
2322 
2324  err = journal_wipe(journal, !really_read_only);
2325  if (!err)
2326  err = journal_load(journal);
2327 
2328  if (err) {
2329  ext3_msg(sb, KERN_ERR, "error loading journal");
2330  journal_destroy(journal);
2331  return err;
2332  }
2333 
2334  EXT3_SB(sb)->s_journal = journal;
2335  ext3_clear_journal_err(sb, es);
2336 
2337  if (!really_read_only && journal_devnum &&
2338  journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2339  es->s_journal_dev = cpu_to_le32(journal_devnum);
2340 
2341  /* Make sure we flush the recovery flag to disk. */
2342  ext3_commit_super(sb, es, 1);
2343  }
2344 
2345  return 0;
2346 }
2347 
2348 static int ext3_create_journal(struct super_block *sb,
2349  struct ext3_super_block *es,
2350  unsigned int journal_inum)
2351 {
2352  journal_t *journal;
2353  int err;
2354 
2355  if (sb->s_flags & MS_RDONLY) {
2356  ext3_msg(sb, KERN_ERR,
2357  "error: readonly filesystem when trying to "
2358  "create journal");
2359  return -EROFS;
2360  }
2361 
2362  journal = ext3_get_journal(sb, journal_inum);
2363  if (!journal)
2364  return -EINVAL;
2365 
2366  ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
2367  journal_inum);
2368 
2369  err = journal_create(journal);
2370  if (err) {
2371  ext3_msg(sb, KERN_ERR, "error creating journal");
2372  journal_destroy(journal);
2373  return -EIO;
2374  }
2375 
2376  EXT3_SB(sb)->s_journal = journal;
2377 
2381 
2382  es->s_journal_inum = cpu_to_le32(journal_inum);
2383 
2384  /* Make sure we flush the recovery flag to disk. */
2385  ext3_commit_super(sb, es, 1);
2386 
2387  return 0;
2388 }
2389 
2390 static int ext3_commit_super(struct super_block *sb,
2391  struct ext3_super_block *es,
2392  int sync)
2393 {
2394  struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2395  int error = 0;
2396 
2397  if (!sbh)
2398  return error;
2399 
2400  if (buffer_write_io_error(sbh)) {
2401  /*
2402  * Oh, dear. A previous attempt to write the
2403  * superblock failed. This could happen because the
2404  * USB device was yanked out. Or it could happen to
2405  * be a transient write error and maybe the block will
2406  * be remapped. Nothing we can do but to retry the
2407  * write and hope for the best.
2408  */
2409  ext3_msg(sb, KERN_ERR, "previous I/O error to "
2410  "superblock detected");
2411  clear_buffer_write_io_error(sbh);
2412  set_buffer_uptodate(sbh);
2413  }
2414  /*
2415  * If the file system is mounted read-only, don't update the
2416  * superblock write time. This avoids updating the superblock
2417  * write time when we are mounting the root file system
2418  * read/only but we need to replay the journal; at that point,
2419  * for people who are east of GMT and who make their clock
2420  * tick in localtime for Windows bug-for-bug compatibility,
2421  * the clock is set in the future, and this will cause e2fsck
2422  * to complain and force a full file system check.
2423  */
2424  if (!(sb->s_flags & MS_RDONLY))
2425  es->s_wtime = cpu_to_le32(get_seconds());
2428  BUFFER_TRACE(sbh, "marking dirty");
2429  mark_buffer_dirty(sbh);
2430  if (sync) {
2431  error = sync_dirty_buffer(sbh);
2432  if (buffer_write_io_error(sbh)) {
2433  ext3_msg(sb, KERN_ERR, "I/O error while writing "
2434  "superblock");
2435  clear_buffer_write_io_error(sbh);
2436  set_buffer_uptodate(sbh);
2437  }
2438  }
2439  return error;
2440 }
2441 
2442 
2443 /*
2444  * Have we just finished recovery? If so, and if we are mounting (or
2445  * remounting) the filesystem readonly, then we will end up with a
2446  * consistent fs on disk. Record that fact.
2447  */
2448 static void ext3_mark_recovery_complete(struct super_block * sb,
2449  struct ext3_super_block * es)
2450 {
2451  journal_t *journal = EXT3_SB(sb)->s_journal;
2452 
2453  journal_lock_updates(journal);
2454  if (journal_flush(journal) < 0)
2455  goto out;
2456 
2458  sb->s_flags & MS_RDONLY) {
2460  ext3_commit_super(sb, es, 1);
2461  }
2462 
2463 out:
2464  journal_unlock_updates(journal);
2465 }
2466 
2467 /*
2468  * If we are mounting (or read-write remounting) a filesystem whose journal
2469  * has recorded an error from a previous lifetime, move that error to the
2470  * main filesystem now.
2471  */
2472 static void ext3_clear_journal_err(struct super_block *sb,
2473  struct ext3_super_block *es)
2474 {
2475  journal_t *journal;
2476  int j_errno;
2477  const char *errstr;
2478 
2479  journal = EXT3_SB(sb)->s_journal;
2480 
2481  /*
2482  * Now check for any error status which may have been recorded in the
2483  * journal by a prior ext3_error() or ext3_abort()
2484  */
2485 
2486  j_errno = journal_errno(journal);
2487  if (j_errno) {
2488  char nbuf[16];
2489 
2490  errstr = ext3_decode_error(sb, j_errno, nbuf);
2491  ext3_warning(sb, __func__, "Filesystem error recorded "
2492  "from previous mount: %s", errstr);
2493  ext3_warning(sb, __func__, "Marking fs in need of "
2494  "filesystem check.");
2495 
2496  EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2497  es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2498  ext3_commit_super (sb, es, 1);
2499 
2500  journal_clear_err(journal);
2501  }
2502 }
2503 
2504 /*
2505  * Force the running and committing transactions to commit,
2506  * and wait on the commit.
2507  */
2509 {
2510  journal_t *journal;
2511  int ret;
2512 
2513  if (sb->s_flags & MS_RDONLY)
2514  return 0;
2515 
2516  journal = EXT3_SB(sb)->s_journal;
2517  ret = ext3_journal_force_commit(journal);
2518  return ret;
2519 }
2520 
2521 static int ext3_sync_fs(struct super_block *sb, int wait)
2522 {
2523  tid_t target;
2524 
2525  trace_ext3_sync_fs(sb, wait);
2526  /*
2527  * Writeback quota in non-journalled quota case - journalled quota has
2528  * no dirty dquots
2529  */
2530  dquot_writeback_dquots(sb, -1);
2531  if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2532  if (wait)
2533  log_wait_commit(EXT3_SB(sb)->s_journal, target);
2534  }
2535  return 0;
2536 }
2537 
2538 /*
2539  * LVM calls this function before a (read-only) snapshot is created. This
2540  * gives us a chance to flush the journal completely and mark the fs clean.
2541  */
2542 static int ext3_freeze(struct super_block *sb)
2543 {
2544  int error = 0;
2545  journal_t *journal;
2546 
2547  if (!(sb->s_flags & MS_RDONLY)) {
2548  journal = EXT3_SB(sb)->s_journal;
2549 
2550  /* Now we set up the journal barrier. */
2551  journal_lock_updates(journal);
2552 
2553  /*
2554  * We don't want to clear needs_recovery flag when we failed
2555  * to flush the journal.
2556  */
2557  error = journal_flush(journal);
2558  if (error < 0)
2559  goto out;
2560 
2561  /* Journal blocked and flushed, clear needs_recovery flag. */
2563  error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2564  if (error)
2565  goto out;
2566  }
2567  return 0;
2568 
2569 out:
2570  journal_unlock_updates(journal);
2571  return error;
2572 }
2573 
2574 /*
2575  * Called by LVM after the snapshot is done. We need to reset the RECOVER
2576  * flag here, even though the filesystem is not technically dirty yet.
2577  */
2578 static int ext3_unfreeze(struct super_block *sb)
2579 {
2580  if (!(sb->s_flags & MS_RDONLY)) {
2581  /* Reser the needs_recovery flag before the fs is unlocked. */
2583  ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2584  journal_unlock_updates(EXT3_SB(sb)->s_journal);
2585  }
2586  return 0;
2587 }
2588 
2589 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2590 {
2591  struct ext3_super_block * es;
2592  struct ext3_sb_info *sbi = EXT3_SB(sb);
2593  ext3_fsblk_t n_blocks_count = 0;
2594  unsigned long old_sb_flags;
2595  struct ext3_mount_options old_opts;
2596  int enable_quota = 0;
2597  int err;
2598 #ifdef CONFIG_QUOTA
2599  int i;
2600 #endif
2601 
2602  /* Store the original options */
2603  old_sb_flags = sb->s_flags;
2604  old_opts.s_mount_opt = sbi->s_mount_opt;
2605  old_opts.s_resuid = sbi->s_resuid;
2606  old_opts.s_resgid = sbi->s_resgid;
2607  old_opts.s_commit_interval = sbi->s_commit_interval;
2608 #ifdef CONFIG_QUOTA
2609  old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2610  for (i = 0; i < MAXQUOTAS; i++)
2611  old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2612 #endif
2613 
2614  /*
2615  * Allow the "check" option to be passed as a remount option.
2616  */
2617  if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2618  err = -EINVAL;
2619  goto restore_opts;
2620  }
2621 
2622  if (test_opt(sb, ABORT))
2623  ext3_abort(sb, __func__, "Abort forced by user");
2624 
2625  sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2626  (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2627 
2628  es = sbi->s_es;
2629 
2630  ext3_init_journal_params(sb, sbi->s_journal);
2631 
2632  if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2633  n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2634  if (test_opt(sb, ABORT)) {
2635  err = -EROFS;
2636  goto restore_opts;
2637  }
2638 
2639  if (*flags & MS_RDONLY) {
2640  err = dquot_suspend(sb, -1);
2641  if (err < 0)
2642  goto restore_opts;
2643 
2644  /*
2645  * First of all, the unconditional stuff we have to do
2646  * to disable replay of the journal when we next remount
2647  */
2648  sb->s_flags |= MS_RDONLY;
2649 
2650  /*
2651  * OK, test if we are remounting a valid rw partition
2652  * readonly, and if so set the rdonly flag and then
2653  * mark the partition as valid again.
2654  */
2655  if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2656  (sbi->s_mount_state & EXT3_VALID_FS))
2657  es->s_state = cpu_to_le16(sbi->s_mount_state);
2658 
2659  ext3_mark_recovery_complete(sb, es);
2660  } else {
2661  __le32 ret;
2662  if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2664  ext3_msg(sb, KERN_WARNING,
2665  "warning: couldn't remount RDWR "
2666  "because of unsupported optional "
2667  "features (%x)", le32_to_cpu(ret));
2668  err = -EROFS;
2669  goto restore_opts;
2670  }
2671 
2672  /*
2673  * If we have an unprocessed orphan list hanging
2674  * around from a previously readonly bdev mount,
2675  * require a full umount & mount for now.
2676  */
2677  if (es->s_last_orphan) {
2678  ext3_msg(sb, KERN_WARNING, "warning: couldn't "
2679  "remount RDWR because of unprocessed "
2680  "orphan inode list. Please "
2681  "umount & mount instead.");
2682  err = -EINVAL;
2683  goto restore_opts;
2684  }
2685 
2686  /*
2687  * Mounting a RDONLY partition read-write, so reread
2688  * and store the current valid flag. (It may have
2689  * been changed by e2fsck since we originally mounted
2690  * the partition.)
2691  */
2692  ext3_clear_journal_err(sb, es);
2693  sbi->s_mount_state = le16_to_cpu(es->s_state);
2694  if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2695  goto restore_opts;
2696  if (!ext3_setup_super (sb, es, 0))
2697  sb->s_flags &= ~MS_RDONLY;
2698  enable_quota = 1;
2699  }
2700  }
2701 #ifdef CONFIG_QUOTA
2702  /* Release old quota file names */
2703  for (i = 0; i < MAXQUOTAS; i++)
2704  if (old_opts.s_qf_names[i] &&
2705  old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2706  kfree(old_opts.s_qf_names[i]);
2707 #endif
2708  if (enable_quota)
2709  dquot_resume(sb, -1);
2710  return 0;
2711 restore_opts:
2712  sb->s_flags = old_sb_flags;
2713  sbi->s_mount_opt = old_opts.s_mount_opt;
2714  sbi->s_resuid = old_opts.s_resuid;
2715  sbi->s_resgid = old_opts.s_resgid;
2716  sbi->s_commit_interval = old_opts.s_commit_interval;
2717 #ifdef CONFIG_QUOTA
2718  sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2719  for (i = 0; i < MAXQUOTAS; i++) {
2720  if (sbi->s_qf_names[i] &&
2721  old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2722  kfree(sbi->s_qf_names[i]);
2723  sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2724  }
2725 #endif
2726  return err;
2727 }
2728 
2729 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2730 {
2731  struct super_block *sb = dentry->d_sb;
2732  struct ext3_sb_info *sbi = EXT3_SB(sb);
2733  struct ext3_super_block *es = sbi->s_es;
2734  u64 fsid;
2735 
2736  if (test_opt(sb, MINIX_DF)) {
2737  sbi->s_overhead_last = 0;
2738  } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2739  unsigned long ngroups = sbi->s_groups_count, i;
2740  ext3_fsblk_t overhead = 0;
2741  smp_rmb();
2742 
2743  /*
2744  * Compute the overhead (FS structures). This is constant
2745  * for a given filesystem unless the number of block groups
2746  * changes so we cache the previous value until it does.
2747  */
2748 
2749  /*
2750  * All of the blocks before first_data_block are
2751  * overhead
2752  */
2753  overhead = le32_to_cpu(es->s_first_data_block);
2754 
2755  /*
2756  * Add the overhead attributed to the superblock and
2757  * block group descriptors. If the sparse superblocks
2758  * feature is turned on, then not all groups have this.
2759  */
2760  for (i = 0; i < ngroups; i++) {
2761  overhead += ext3_bg_has_super(sb, i) +
2762  ext3_bg_num_gdb(sb, i);
2763  cond_resched();
2764  }
2765 
2766  /*
2767  * Every block group has an inode bitmap, a block
2768  * bitmap, and an inode table.
2769  */
2770  overhead += ngroups * (2 + sbi->s_itb_per_group);
2771  sbi->s_overhead_last = overhead;
2772  smp_wmb();
2774  }
2775 
2776  buf->f_type = EXT3_SUPER_MAGIC;
2777  buf->f_bsize = sb->s_blocksize;
2779  buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2780  buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2781  if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2782  buf->f_bavail = 0;
2783  buf->f_files = le32_to_cpu(es->s_inodes_count);
2784  buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2785  buf->f_namelen = EXT3_NAME_LEN;
2786  fsid = le64_to_cpup((void *)es->s_uuid) ^
2787  le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2788  buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2789  buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2790  return 0;
2791 }
2792 
2793 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2794  * is locked for write. Otherwise the are possible deadlocks:
2795  * Process 1 Process 2
2796  * ext3_create() quota_sync()
2797  * journal_start() write_dquot()
2798  * dquot_initialize() down(dqio_mutex)
2799  * down(dqio_mutex) journal_start()
2800  *
2801  */
2802 
2803 #ifdef CONFIG_QUOTA
2804 
2805 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2806 {
2807  return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
2808 }
2809 
2810 static int ext3_write_dquot(struct dquot *dquot)
2811 {
2812  int ret, err;
2813  handle_t *handle;
2814  struct inode *inode;
2815 
2816  inode = dquot_to_inode(dquot);
2817  handle = ext3_journal_start(inode,
2818  EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2819  if (IS_ERR(handle))
2820  return PTR_ERR(handle);
2821  ret = dquot_commit(dquot);
2822  err = ext3_journal_stop(handle);
2823  if (!ret)
2824  ret = err;
2825  return ret;
2826 }
2827 
2828 static int ext3_acquire_dquot(struct dquot *dquot)
2829 {
2830  int ret, err;
2831  handle_t *handle;
2832 
2833  handle = ext3_journal_start(dquot_to_inode(dquot),
2834  EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2835  if (IS_ERR(handle))
2836  return PTR_ERR(handle);
2837  ret = dquot_acquire(dquot);
2838  err = ext3_journal_stop(handle);
2839  if (!ret)
2840  ret = err;
2841  return ret;
2842 }
2843 
2844 static int ext3_release_dquot(struct dquot *dquot)
2845 {
2846  int ret, err;
2847  handle_t *handle;
2848 
2849  handle = ext3_journal_start(dquot_to_inode(dquot),
2850  EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2851  if (IS_ERR(handle)) {
2852  /* Release dquot anyway to avoid endless cycle in dqput() */
2853  dquot_release(dquot);
2854  return PTR_ERR(handle);
2855  }
2856  ret = dquot_release(dquot);
2857  err = ext3_journal_stop(handle);
2858  if (!ret)
2859  ret = err;
2860  return ret;
2861 }
2862 
2863 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2864 {
2865  /* Are we journaling quotas? */
2866  if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2867  EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2868  dquot_mark_dquot_dirty(dquot);
2869  return ext3_write_dquot(dquot);
2870  } else {
2871  return dquot_mark_dquot_dirty(dquot);
2872  }
2873 }
2874 
2875 static int ext3_write_info(struct super_block *sb, int type)
2876 {
2877  int ret, err;
2878  handle_t *handle;
2879 
2880  /* Data block + inode block */
2881  handle = ext3_journal_start(sb->s_root->d_inode, 2);
2882  if (IS_ERR(handle))
2883  return PTR_ERR(handle);
2884  ret = dquot_commit_info(sb, type);
2885  err = ext3_journal_stop(handle);
2886  if (!ret)
2887  ret = err;
2888  return ret;
2889 }
2890 
2891 /*
2892  * Turn on quotas during mount time - we need to find
2893  * the quota file and such...
2894  */
2895 static int ext3_quota_on_mount(struct super_block *sb, int type)
2896 {
2897  return dquot_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2898  EXT3_SB(sb)->s_jquota_fmt, type);
2899 }
2900 
2901 /*
2902  * Standard function to be called on quota_on
2903  */
2904 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2905  struct path *path)
2906 {
2907  int err;
2908 
2909  if (!test_opt(sb, QUOTA))
2910  return -EINVAL;
2911 
2912  /* Quotafile not on the same filesystem? */
2913  if (path->dentry->d_sb != sb)
2914  return -EXDEV;
2915  /* Journaling quota? */
2916  if (EXT3_SB(sb)->s_qf_names[type]) {
2917  /* Quotafile not of fs root? */
2918  if (path->dentry->d_parent != sb->s_root)
2919  ext3_msg(sb, KERN_WARNING,
2920  "warning: Quota file not on filesystem root. "
2921  "Journaled quota will not work.");
2922  }
2923 
2924  /*
2925  * When we journal data on quota file, we have to flush journal to see
2926  * all updates to the file when we bypass pagecache...
2927  */
2928  if (ext3_should_journal_data(path->dentry->d_inode)) {
2929  /*
2930  * We don't need to lock updates but journal_flush() could
2931  * otherwise be livelocked...
2932  */
2933  journal_lock_updates(EXT3_SB(sb)->s_journal);
2934  err = journal_flush(EXT3_SB(sb)->s_journal);
2935  journal_unlock_updates(EXT3_SB(sb)->s_journal);
2936  if (err)
2937  return err;
2938  }
2939 
2940  return dquot_quota_on(sb, type, format_id, path);
2941 }
2942 
2943 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2944  * acquiring the locks... As quota files are never truncated and quota code
2945  * itself serializes the operations (and no one else should touch the files)
2946  * we don't have to be afraid of races */
2947 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2948  size_t len, loff_t off)
2949 {
2950  struct inode *inode = sb_dqopt(sb)->files[type];
2951  sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2952  int err = 0;
2953  int offset = off & (sb->s_blocksize - 1);
2954  int tocopy;
2955  size_t toread;
2956  struct buffer_head *bh;
2957  loff_t i_size = i_size_read(inode);
2958 
2959  if (off > i_size)
2960  return 0;
2961  if (off+len > i_size)
2962  len = i_size-off;
2963  toread = len;
2964  while (toread > 0) {
2965  tocopy = sb->s_blocksize - offset < toread ?
2966  sb->s_blocksize - offset : toread;
2967  bh = ext3_bread(NULL, inode, blk, 0, &err);
2968  if (err)
2969  return err;
2970  if (!bh) /* A hole? */
2971  memset(data, 0, tocopy);
2972  else
2973  memcpy(data, bh->b_data+offset, tocopy);
2974  brelse(bh);
2975  offset = 0;
2976  toread -= tocopy;
2977  data += tocopy;
2978  blk++;
2979  }
2980  return len;
2981 }
2982 
2983 /* Write to quotafile (we know the transaction is already started and has
2984  * enough credits) */
2985 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2986  const char *data, size_t len, loff_t off)
2987 {
2988  struct inode *inode = sb_dqopt(sb)->files[type];
2989  sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2990  int err = 0;
2991  int offset = off & (sb->s_blocksize - 1);
2992  int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2993  struct buffer_head *bh;
2994  handle_t *handle = journal_current_handle();
2995 
2996  if (!handle) {
2997  ext3_msg(sb, KERN_WARNING,
2998  "warning: quota write (off=%llu, len=%llu)"
2999  " cancelled because transaction is not started.",
3000  (unsigned long long)off, (unsigned long long)len);
3001  return -EIO;
3002  }
3003 
3004  /*
3005  * Since we account only one data block in transaction credits,
3006  * then it is impossible to cross a block boundary.
3007  */
3008  if (sb->s_blocksize - offset < len) {
3009  ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3010  " cancelled because not block aligned",
3011  (unsigned long long)off, (unsigned long long)len);
3012  return -EIO;
3013  }
3014  bh = ext3_bread(handle, inode, blk, 1, &err);
3015  if (!bh)
3016  goto out;
3017  if (journal_quota) {
3018  err = ext3_journal_get_write_access(handle, bh);
3019  if (err) {
3020  brelse(bh);
3021  goto out;
3022  }
3023  }
3024  lock_buffer(bh);
3025  memcpy(bh->b_data+offset, data, len);
3026  flush_dcache_page(bh->b_page);
3027  unlock_buffer(bh);
3028  if (journal_quota)
3029  err = ext3_journal_dirty_metadata(handle, bh);
3030  else {
3031  /* Always do at least ordered writes for quotas */
3032  err = ext3_journal_dirty_data(handle, bh);
3033  mark_buffer_dirty(bh);
3034  }
3035  brelse(bh);
3036 out:
3037  if (err)
3038  return err;
3039  if (inode->i_size < off + len) {
3040  i_size_write(inode, off + len);
3041  EXT3_I(inode)->i_disksize = inode->i_size;
3042  }
3043  inode->i_version++;
3044  inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3045  ext3_mark_inode_dirty(handle, inode);
3046  return len;
3047 }
3048 
3049 #endif
3050 
3051 static struct dentry *ext3_mount(struct file_system_type *fs_type,
3052  int flags, const char *dev_name, void *data)
3053 {
3054  return mount_bdev(fs_type, flags, dev_name, data, ext3_fill_super);
3055 }
3056 
3057 static struct file_system_type ext3_fs_type = {
3058  .owner = THIS_MODULE,
3059  .name = "ext3",
3060  .mount = ext3_mount,
3061  .kill_sb = kill_block_super,
3062  .fs_flags = FS_REQUIRES_DEV,
3063 };
3064 
3065 static int __init init_ext3_fs(void)
3066 {
3067  int err = init_ext3_xattr();
3068  if (err)
3069  return err;
3070  err = init_inodecache();
3071  if (err)
3072  goto out1;
3073  err = register_filesystem(&ext3_fs_type);
3074  if (err)
3075  goto out;
3076  return 0;
3077 out:
3078  destroy_inodecache();
3079 out1:
3080  exit_ext3_xattr();
3081  return err;
3082 }
3083 
3084 static void __exit exit_ext3_fs(void)
3085 {
3086  unregister_filesystem(&ext3_fs_type);
3087  destroy_inodecache();
3088  exit_ext3_xattr();
3089 }
3090 
3091 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3092 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3093 MODULE_LICENSE("GPL");
3094 module_init(init_ext3_fs)
3095 module_exit(exit_ext3_fs)