]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - fs/ext3/super.c
projects / mirror_ubuntu-zesty-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext3: NULL dereference in ext3_evict_inode()
[mirror_ubuntu-zesty-kernel.git] / fs / ext3 / super.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/ext3/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
1da177e4
LT		 19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
23#include <linux/jbd.h>
24#include <linux/ext3_fs.h>
25#include <linux/ext3_jbd.h>
26#include <linux/slab.h>
27#include <linux/init.h>
28#include <linux/blkdev.h>
29#include <linux/parser.h>
1da177e4 30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
1da177e4
LT		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
8fc2751b 37#include <linux/seq_file.h>
3fc74269 38#include <linux/log2.h>
d71bc6db 39#include <linux/cleancache.h>
381be254 40
1da177e4 41#include <asm/uaccess.h>
381be254 42
1da177e4
LT		 43#include "xattr.h"
44#include "acl.h"
381be254 45#include "namei.h"
1da177e4 46
785c4bcc
LC		 47#define CREATE_TRACE_POINTS
48#include <trace/events/ext3.h>
49
bbae8bcc
LT		 50#ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
51 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
52#else
53 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
54#endif
55
71b96257
JL		 56static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
57 unsigned long journal_devnum);
1da177e4 58static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
eee194e7 59 unsigned int);
c4be0c1d
TS		 60static int ext3_commit_super(struct super_block *sb,
61 struct ext3_super_block *es,
1da177e4
LT		 62 int sync);
63static void ext3_mark_recovery_complete(struct super_block * sb,
64 struct ext3_super_block * es);
65static void ext3_clear_journal_err(struct super_block * sb,
66 struct ext3_super_block * es);
67static int ext3_sync_fs(struct super_block *sb, int wait);
68static const char *ext3_decode_error(struct super_block * sb, int errno,
69 char nbuf[16]);
70static int ext3_remount (struct super_block * sb, int * flags, char * data);
726c3342 71static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
c4be0c1d 72static int ext3_unfreeze(struct super_block *sb);
c4be0c1d 73static int ext3_freeze(struct super_block *sb);
1da177e4 74
ae6ddcc5 75/*
1da177e4
LT		 76 * Wrappers for journal_start/end.
77 *
78 * The only special thing we need to do here is to make sure that all
79 * journal_end calls result in the superblock being marked dirty, so
80 * that sync() will call the filesystem's write_super callback if
ae6ddcc5 81 * appropriate.
1da177e4
LT		 82 */
83handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
84{
85 journal_t *journal;
86
87 if (sb->s_flags & MS_RDONLY)
88 return ERR_PTR(-EROFS);
89
90 /* Special case here: if the journal has aborted behind our
91 * backs (eg. EIO in the commit thread), then we still need to
92 * take the FS itself readonly cleanly. */
93 journal = EXT3_SB(sb)->s_journal;
94 if (is_journal_aborted(journal)) {
e05b6b52 95 ext3_abort(sb, __func__,
1da177e4
LT		 96 "Detected aborted journal");
97 return ERR_PTR(-EROFS);
98 }
99
100 return journal_start(journal, nblocks);
101}
102
ae6ddcc5 103/*
1da177e4
LT		 104 * The only special thing we need to do here is to make sure that all
105 * journal_stop calls result in the superblock being marked dirty, so
106 * that sync() will call the filesystem's write_super callback if
ae6ddcc5 107 * appropriate.
1da177e4
LT		 108 */
109int __ext3_journal_stop(const char *where, handle_t *handle)
110{
111 struct super_block *sb;
112 int err;
113 int rc;
114
115 sb = handle->h_transaction->t_journal->j_private;
116 err = handle->h_err;
117 rc = journal_stop(handle);
118
119 if (!err)
120 err = rc;
121 if (err)
122 __ext3_std_error(sb, where, err);
123 return err;
124}
125
126void ext3_journal_abort_handle(const char *caller, const char *err_fn,
127 struct buffer_head *bh, handle_t *handle, int err)
128{
129 char nbuf[16];
130 const char *errstr = ext3_decode_error(NULL, err, nbuf);
131
132 if (bh)
133 BUFFER_TRACE(bh, "abort");
134
135 if (!handle->h_err)
136 handle->h_err = err;
137
138 if (is_handle_aborted(handle))
139 return;
140
4cf46b67
AF		 141 printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n",
142 caller, errstr, err_fn);
1da177e4
LT		 143
144 journal_abort_handle(handle);
145}
146
4cf46b67
AF		 147void ext3_msg(struct super_block *sb, const char *prefix,
148 const char *fmt, ...)
149{
99fbb1e2 150 struct va_format vaf;
4cf46b67
AF		 151 va_list args;
152
153 va_start(args, fmt);
99fbb1e2
JP		 154
155 vaf.fmt = fmt;
156 vaf.va = &args;
157
158 printk("%sEXT3-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
159
4cf46b67
AF		 160 va_end(args);
161}
162
1da177e4
LT		 163/* Deal with the reporting of failure conditions on a filesystem such as
164 * inconsistencies detected or read IO failures.
165 *
166 * On ext2, we can store the error state of the filesystem in the
167 * superblock. That is not possible on ext3, because we may have other
168 * write ordering constraints on the superblock which prevent us from
169 * writing it out straight away; and given that the journal is about to
170 * be aborted, we can't rely on the current, or future, transactions to
171 * write out the superblock safely.
172 *
173 * We'll just use the journal_abort() error code to record an error in
d6b198bc 174 * the journal instead. On recovery, the journal will complain about
1da177e4
LT		 175 * that error until we've noted it down and cleared it.
176 */
177
178static void ext3_handle_error(struct super_block *sb)
179{
180 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
181
182 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
183 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
184
185 if (sb->s_flags & MS_RDONLY)
186 return;
187
7543fc7b 188 if (!test_opt (sb, ERRORS_CONT)) {
1da177e4
LT		 189 journal_t *journal = EXT3_SB(sb)->s_journal;
190
e3c96435 191 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
1da177e4
LT		 192 if (journal)
193 journal_abort(journal, -EIO);
194 }
7543fc7b 195 if (test_opt (sb, ERRORS_RO)) {
4cf46b67
AF		 196 ext3_msg(sb, KERN_CRIT,
197 "error: remounting filesystem read-only");
7543fc7b
VA		 198 sb->s_flags |= MS_RDONLY;
199 }
200 ext3_commit_super(sb, es, 1);
1da177e4 201 if (test_opt(sb, ERRORS_PANIC))
4cf46b67 202 panic("EXT3-fs (%s): panic forced after error\n",
1da177e4 203 sb->s_id);
1da177e4
LT		 204}
205
99fbb1e2
JP		 206void ext3_error(struct super_block *sb, const char *function,
207 const char *fmt, ...)
1da177e4 208{
99fbb1e2 209 struct va_format vaf;
1da177e4
LT		 210 va_list args;
211
212 va_start(args, fmt);
99fbb1e2
JP		 213
214 vaf.fmt = fmt;
215 vaf.va = &args;
216
217 printk(KERN_CRIT "EXT3-fs error (device %s): %s: %pV\n",
218 sb->s_id, function, &vaf);
219
1da177e4
LT		 220 va_end(args);
221
222 ext3_handle_error(sb);
223}
224
225static const char *ext3_decode_error(struct super_block * sb, int errno,
226 char nbuf[16])
227{
228 char *errstr = NULL;
229
230 switch (errno) {
231 case -EIO:
232 errstr = "IO failure";
233 break;
234 case -ENOMEM:
235 errstr = "Out of memory";
236 break;
237 case -EROFS:
238 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
239 errstr = "Journal has aborted";
240 else
241 errstr = "Readonly filesystem";
242 break;
243 default:
244 /* If the caller passed in an extra buffer for unknown
245 * errors, textualise them now. Else we just return
246 * NULL. */
247 if (nbuf) {
248 /* Check for truncated error codes... */
249 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
250 errstr = nbuf;
251 }
252 break;
253 }
254
255 return errstr;
256}
257
258/* __ext3_std_error decodes expected errors from journaling functions
259 * automatically and invokes the appropriate error response. */
260
261void __ext3_std_error (struct super_block * sb, const char * function,
262 int errno)
263{
264 char nbuf[16];
30121624 265 const char *errstr;
1da177e4 266
30121624
ST		 267 /* Special case: if the error is EROFS, and we're not already
268 * inside a transaction, then there's really no point in logging
269 * an error. */
270 if (errno == -EROFS && journal_current_handle() == NULL &&
271 (sb->s_flags & MS_RDONLY))
272 return;
273
274 errstr = ext3_decode_error(sb, errno, nbuf);
4cf46b67 275 ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
1da177e4
LT		 276
277 ext3_handle_error(sb);
278}
279
280/*
281 * ext3_abort is a much stronger failure handler than ext3_error. The
282 * abort function may be used to deal with unrecoverable failures such
283 * as journal IO errors or ENOMEM at a critical moment in log management.
284 *
285 * We unconditionally force the filesystem into an ABORT|READONLY state,
286 * unless the error response on the fs has been set to panic in which
287 * case we take the easy way out and panic immediately.
288 */
289
99fbb1e2
JP		 290void ext3_abort(struct super_block *sb, const char *function,
291 const char *fmt, ...)
1da177e4 292{
99fbb1e2 293 struct va_format vaf;
1da177e4
LT		 294 va_list args;
295
1da177e4 296 va_start(args, fmt);
99fbb1e2
JP		 297
298 vaf.fmt = fmt;
299 vaf.va = &args;
300
301 printk(KERN_CRIT "EXT3-fs (%s): error: %s: %pV\n",
302 sb->s_id, function, &vaf);
303
1da177e4
LT		 304 va_end(args);
305
306 if (test_opt(sb, ERRORS_PANIC))
4cf46b67 307 panic("EXT3-fs: panic from previous error\n");
1da177e4
LT		 308
309 if (sb->s_flags & MS_RDONLY)
310 return;
311
4cf46b67
AF		 312 ext3_msg(sb, KERN_CRIT,
313 "error: remounting filesystem read-only");
1da177e4
LT		 314 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
315 sb->s_flags |= MS_RDONLY;
e3c96435 316 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
44d6f787
HK		 317 if (EXT3_SB(sb)->s_journal)
318 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
1da177e4
LT		 319}
320
99fbb1e2
JP		 321void ext3_warning(struct super_block *sb, const char *function,
322 const char *fmt, ...)
1da177e4 323{
99fbb1e2 324 struct va_format vaf;
1da177e4
LT		 325 va_list args;
326
327 va_start(args, fmt);
99fbb1e2
JP		 328
329 vaf.fmt = fmt;
330 vaf.va = &args;
331
332 printk(KERN_WARNING "EXT3-fs (%s): warning: %s: %pV\n",
333 sb->s_id, function, &vaf);
334
1da177e4
LT		 335 va_end(args);
336}
337
338void ext3_update_dynamic_rev(struct super_block *sb)
339{
340 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
341
342 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
343 return;
344
4cf46b67
AF		 345 ext3_msg(sb, KERN_WARNING,
346 "warning: updating to rev %d because of "
347 "new feature flag, running e2fsck is recommended",
348 EXT3_DYNAMIC_REV);
1da177e4
LT		 349
350 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
351 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
352 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
353 /* leave es->s_feature_*compat flags alone */
354 /* es->s_uuid will be set by e2fsck if empty */
355
356 /*
357 * The rest of the superblock fields should be zero, and if not it
358 * means they are likely already in use, so leave them alone. We
359 * can leave it up to e2fsck to clean up any inconsistencies there.
360 */
361}
362
363/*
364 * Open the external journal device
365 */
4cf46b67 366static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
1da177e4
LT		 367{
368 struct block_device *bdev;
369 char b[BDEVNAME_SIZE];
370
d4d77629 371 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1da177e4
LT		 372 if (IS_ERR(bdev))
373 goto fail;
374 return bdev;
375
376fail:
4cf46b67
AF		 377 ext3_msg(sb, "error: failed to open journal device %s: %ld",
378 __bdevname(dev, b), PTR_ERR(bdev));
379
1da177e4
LT		 380 return NULL;
381}
382
383/*
384 * Release the journal device
385 */
386static int ext3_blkdev_put(struct block_device *bdev)
387{
e525fd89 388 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1da177e4
LT		 389}
390
391static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
392{
393 struct block_device *bdev;
394 int ret = -ENODEV;
395
396 bdev = sbi->journal_bdev;
397 if (bdev) {
398 ret = ext3_blkdev_put(bdev);
399 sbi->journal_bdev = NULL;
400 }
401 return ret;
402}
403
404static inline struct inode *orphan_list_entry(struct list_head *l)
405{
406 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
407}
408
409static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
410{
411 struct list_head *l;
412
4cf46b67 413 ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
1da177e4
LT		 414 le32_to_cpu(sbi->s_es->s_last_orphan));
415
4cf46b67 416 ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
1da177e4
LT		 417 list_for_each(l, &sbi->s_orphan) {
418 struct inode *inode = orphan_list_entry(l);
4cf46b67 419 ext3_msg(sb, KERN_ERR, " "
eee194e7 420 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1da177e4 421 inode->i_sb->s_id, inode->i_ino, inode,
ae6ddcc5 422 inode->i_mode, inode->i_nlink,
1da177e4
LT		 423 NEXT_ORPHAN(inode));
424 }
425}
426
427static void ext3_put_super (struct super_block * sb)
428{
429 struct ext3_sb_info *sbi = EXT3_SB(sb);
430 struct ext3_super_block *es = sbi->s_es;
44d6f787 431 int i, err;
1da177e4 432
e0ccfd95 433 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
1da177e4 434 ext3_xattr_put_super(sb);
44d6f787
HK		 435 err = journal_destroy(sbi->s_journal);
436 sbi->s_journal = NULL;
437 if (err < 0)
2d7c820e 438 ext3_abort(sb, __func__, "Couldn't clean up the journal");
44d6f787 439
1da177e4
LT		 440 if (!(sb->s_flags & MS_RDONLY)) {
441 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
442 es->s_state = cpu_to_le16(sbi->s_mount_state);
443 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
444 mark_buffer_dirty(sbi->s_sbh);
445 ext3_commit_super(sb, es, 1);
446 }
447
448 for (i = 0; i < sbi->s_gdb_count; i++)
449 brelse(sbi->s_group_desc[i]);
450 kfree(sbi->s_group_desc);
451 percpu_counter_destroy(&sbi->s_freeblocks_counter);
452 percpu_counter_destroy(&sbi->s_freeinodes_counter);
453 percpu_counter_destroy(&sbi->s_dirs_counter);
454 brelse(sbi->s_sbh);
455#ifdef CONFIG_QUOTA
456 for (i = 0; i < MAXQUOTAS; i++)
457 kfree(sbi->s_qf_names[i]);
458#endif
459
460 /* Debugging code just in case the in-memory inode orphan list
461 * isn't empty. The on-disk one can be non-empty if we've
462 * detected an error and taken the fs readonly, but the
463 * in-memory list had better be clean by this point. */
464 if (!list_empty(&sbi->s_orphan))
465 dump_orphan_list(sb, sbi);
466 J_ASSERT(list_empty(&sbi->s_orphan));
467
f98393a6 468 invalidate_bdev(sb->s_bdev);
1da177e4
LT		 469 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
470 /*
471 * Invalidate the journal device's buffers. We don't want them
472 * floating about in memory - the physical journal device may
473 * hotswapped, and it breaks the `ro-after' testing code.
474 */
475 sync_blockdev(sbi->journal_bdev);
f98393a6 476 invalidate_bdev(sbi->journal_bdev);
1da177e4
LT		 477 ext3_blkdev_remove(sbi);
478 }
479 sb->s_fs_info = NULL;
5df096d6 480 kfree(sbi->s_blockgroup_lock);
1da177e4 481 kfree(sbi);
1da177e4
LT		 482}
483
e18b890b 484static struct kmem_cache *ext3_inode_cachep;
1da177e4
LT		 485
486/*
487 * Called inside transaction, so use GFP_NOFS
488 */
489static struct inode *ext3_alloc_inode(struct super_block *sb)
490{
491 struct ext3_inode_info *ei;
492
e6b4f8da 493 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
1da177e4
LT		 494 if (!ei)
495 return NULL;
1da177e4
LT		 496 ei->i_block_alloc_info = NULL;
497 ei->vfs_inode.i_version = 1;
fe8bc91c
JK		 498 atomic_set(&ei->i_datasync_tid, 0);
499 atomic_set(&ei->i_sync_tid, 0);
1da177e4
LT		 500 return &ei->vfs_inode;
501}
502
785c4bcc
LC		 503static int ext3_drop_inode(struct inode *inode)
504{
505 int drop = generic_drop_inode(inode);
506
507 trace_ext3_drop_inode(inode, drop);
508 return drop;
509}
510
fa0d7e3d
NP		 511static void ext3_i_callback(struct rcu_head *head)
512{
513 struct inode *inode = container_of(head, struct inode, i_rcu);
514 INIT_LIST_HEAD(&inode->i_dentry);
515 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
516}
517
1da177e4
LT		 518static void ext3_destroy_inode(struct inode *inode)
519{
9f7dd93d
VA		 520 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
521 printk("EXT3 Inode %p: orphan list check failed!\n",
522 EXT3_I(inode));
523 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
524 EXT3_I(inode), sizeof(struct ext3_inode_info),
525 false);
526 dump_stack();
527 }
fa0d7e3d 528 call_rcu(&inode->i_rcu, ext3_i_callback);
1da177e4
LT		 529}
530
51cc5068 531static void init_once(void *foo)
1da177e4
LT		 532{
533 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
534
a35afb83 535 INIT_LIST_HEAD(&ei->i_orphan);
1da177e4 536#ifdef CONFIG_EXT3_FS_XATTR
a35afb83 537 init_rwsem(&ei->xattr_sem);
1da177e4 538#endif
a35afb83
CL		 539 mutex_init(&ei->truncate_mutex);
540 inode_init_once(&ei->vfs_inode);
1da177e4 541}
ae6ddcc5 542
1da177e4
LT		 543static int init_inodecache(void)
544{
545 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
546 sizeof(struct ext3_inode_info),
fffb60f9
PJ		 547 0, (SLAB_RECLAIM_ACCOUNT|
548 SLAB_MEM_SPREAD),
20c2df83 549 init_once);
1da177e4
LT		 550 if (ext3_inode_cachep == NULL)
551 return -ENOMEM;
552 return 0;
553}
554
555static void destroy_inodecache(void)
556{
1a1d92c1 557 kmem_cache_destroy(ext3_inode_cachep);
1da177e4
LT		 558}
559
6cd37cda 560static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
8fc2751b 561{
6cd37cda 562#if defined(CONFIG_QUOTA)
275abf5b 563 struct ext3_sb_info *sbi = EXT3_SB(sb);
8fc2751b 564
1aeec434
JK		 565 if (sbi->s_jquota_fmt) {
566 char *fmtname = "";
567
568 switch (sbi->s_jquota_fmt) {
569 case QFMT_VFS_OLD:
570 fmtname = "vfsold";
571 break;
572 case QFMT_VFS_V0:
573 fmtname = "vfsv0";
574 break;
575 case QFMT_VFS_V1:
576 fmtname = "vfsv1";
577 break;
578 }
579 seq_printf(seq, ",jqfmt=%s", fmtname);
580 }
8fc2751b
MB		 581
582 if (sbi->s_qf_names[USRQUOTA])
583 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
584
585 if (sbi->s_qf_names[GRPQUOTA])
586 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
587
e3c96435 588 if (test_opt(sb, USRQUOTA))
8fc2751b
MB		 589 seq_puts(seq, ",usrquota");
590
e3c96435 591 if (test_opt(sb, GRPQUOTA))
8fc2751b
MB		 592 seq_puts(seq, ",grpquota");
593#endif
6cd37cda
PO		 594}
595
3c4cec65
JK		 596static char *data_mode_string(unsigned long mode)
597{
598 switch (mode) {
599 case EXT3_MOUNT_JOURNAL_DATA:
600 return "journal";
601 case EXT3_MOUNT_ORDERED_DATA:
602 return "ordered";
603 case EXT3_MOUNT_WRITEBACK_DATA:
604 return "writeback";
605 }
606 return "unknown";
607}
608
571beed8
MS		 609/*
610 * Show an option if
611 * - it's set to a non-default value OR
612 * - if the per-sb default is different from the global default
613 */
6cd37cda
PO		 614static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
615{
616 struct super_block *sb = vfs->mnt_sb;
571beed8
MS		 617 struct ext3_sb_info *sbi = EXT3_SB(sb);
618 struct ext3_super_block *es = sbi->s_es;
619 unsigned long def_mount_opts;
620
621 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
622
623 if (sbi->s_sb_block != 1)
624 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
625 if (test_opt(sb, MINIX_DF))
626 seq_puts(seq, ",minixdf");
627 if (test_opt(sb, GRPID))
628 seq_puts(seq, ",grpid");
629 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
630 seq_puts(seq, ",nogrpid");
631 if (sbi->s_resuid != EXT3_DEF_RESUID ||
632 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
633 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
634 }
635 if (sbi->s_resgid != EXT3_DEF_RESGID ||
636 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
637 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
638 }
1eca93f9 639 if (test_opt(sb, ERRORS_RO)) {
571beed8
MS		 640 int def_errors = le16_to_cpu(es->s_errors);
641
642 if (def_errors == EXT3_ERRORS_PANIC ||
1eca93f9
AK		 643 def_errors == EXT3_ERRORS_CONTINUE) {
644 seq_puts(seq, ",errors=remount-ro");
571beed8
MS		 645 }
646 }
1eca93f9
AK		 647 if (test_opt(sb, ERRORS_CONT))
648 seq_puts(seq, ",errors=continue");
571beed8
MS		 649 if (test_opt(sb, ERRORS_PANIC))
650 seq_puts(seq, ",errors=panic");
651 if (test_opt(sb, NO_UID32))
652 seq_puts(seq, ",nouid32");
653 if (test_opt(sb, DEBUG))
654 seq_puts(seq, ",debug");
571beed8
MS		 655#ifdef CONFIG_EXT3_FS_XATTR
656 if (test_opt(sb, XATTR_USER))
657 seq_puts(seq, ",user_xattr");
658 if (!test_opt(sb, XATTR_USER) &&
659 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
660 seq_puts(seq, ",nouser_xattr");
661 }
662#endif
663#ifdef CONFIG_EXT3_FS_POSIX_ACL
664 if (test_opt(sb, POSIX_ACL))
665 seq_puts(seq, ",acl");
666 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
667 seq_puts(seq, ",noacl");
668#endif
669 if (!test_opt(sb, RESERVATION))
670 seq_puts(seq, ",noreservation");
671 if (sbi->s_commit_interval) {
672 seq_printf(seq, ",commit=%u",
673 (unsigned) (sbi->s_commit_interval / HZ));
674 }
0636c73e
ES		 675
676 /*
677 * Always display barrier state so it's clear what the status is.
678 */
679 seq_puts(seq, ",barrier=");
680 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
e3c96435 681 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
0e4fb5e2
HK		 682 if (test_opt(sb, DATA_ERR_ABORT))
683 seq_puts(seq, ",data_err=abort");
684
dee1d3b6
ES		 685 if (test_opt(sb, NOLOAD))
686 seq_puts(seq, ",norecovery");
687
6cd37cda 688 ext3_show_quota_options(seq, sb);
8fc2751b
MB		 689
690 return 0;
691}
1da177e4 692
fdb36673 693
74af0baa
CH		 694static struct inode *ext3_nfs_get_inode(struct super_block *sb,
695 u64 ino, u32 generation)
fdb36673 696{
fdb36673 697 struct inode *inode;
fdb36673
N		 698
699 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
700 return ERR_PTR(-ESTALE);
701 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
702 return ERR_PTR(-ESTALE);
703
704 /* iget isn't really right if the inode is currently unallocated!!
705 *
706 * ext3_read_inode will return a bad_inode if the inode had been
707 * deleted, so we should be safe.
708 *
709 * Currently we don't know the generation for parent directory, so
710 * a generation of 0 means "accept any"
711 */
473043dc
DH		 712 inode = ext3_iget(sb, ino);
713 if (IS_ERR(inode))
714 return ERR_CAST(inode);
715 if (generation && inode->i_generation != generation) {
fdb36673
N		 716 iput(inode);
717 return ERR_PTR(-ESTALE);
718 }
74af0baa
CH		 719
720 return inode;
721}
722
723static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
724 int fh_len, int fh_type)
725{
726 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
727 ext3_nfs_get_inode);
728}
729
730static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
731 int fh_len, int fh_type)
732{
733 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
734 ext3_nfs_get_inode);
fdb36673
N		 735}
736
6b082b53
TO		 737/*
738 * Try to release metadata pages (indirect blocks, directories) which are
739 * mapped via the block device. Since these pages could have journal heads
740 * which would prevent try_to_free_buffers() from freeing them, we must use
741 * jbd layer's try_to_free_buffers() function to release them.
742 */
743static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
744 gfp_t wait)
745{
746 journal_t *journal = EXT3_SB(sb)->s_journal;
747
748 WARN_ON(PageChecked(page));
749 if (!page_has_buffers(page))
750 return 0;
751 if (journal)
752 return journal_try_to_free_buffers(journal, page,
753 wait & ~__GFP_WAIT);
754 return try_to_free_buffers(page);
755}
756
8fc2751b 757#ifdef CONFIG_QUOTA
1da177e4
LT		 758#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
759#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
760
1da177e4
LT		 761static int ext3_write_dquot(struct dquot *dquot);
762static int ext3_acquire_dquot(struct dquot *dquot);
763static int ext3_release_dquot(struct dquot *dquot);
764static int ext3_mark_dquot_dirty(struct dquot *dquot);
765static int ext3_write_info(struct super_block *sb, int type);
2fd83a4f 766static int ext3_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 767 struct path *path);
1da177e4
LT		 768static int ext3_quota_on_mount(struct super_block *sb, int type);
769static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
770 size_t len, loff_t off);
771static ssize_t ext3_quota_write(struct super_block *sb, int type,
772 const char *data, size_t len, loff_t off);
773
61e225dc 774static const struct dquot_operations ext3_quota_operations = {
1da177e4
LT		 775 .write_dquot = ext3_write_dquot,
776 .acquire_dquot = ext3_acquire_dquot,
777 .release_dquot = ext3_release_dquot,
778 .mark_dirty = ext3_mark_dquot_dirty,
157091a2
JK		 779 .write_info = ext3_write_info,
780 .alloc_dquot = dquot_alloc,
781 .destroy_dquot = dquot_destroy,
1da177e4
LT		 782};
783
0d54b217 784static const struct quotactl_ops ext3_qctl_operations = {
1da177e4 785 .quota_on = ext3_quota_on,
287a8095
CH		 786 .quota_off = dquot_quota_off,
787 .quota_sync = dquot_quota_sync,
788 .get_info = dquot_get_dqinfo,
789 .set_info = dquot_set_dqinfo,
790 .get_dqblk = dquot_get_dqblk,
791 .set_dqblk = dquot_set_dqblk
1da177e4
LT		 792};
793#endif
794
ee9b6d61 795static const struct super_operations ext3_sops = {
1da177e4
LT		 796 .alloc_inode = ext3_alloc_inode,
797 .destroy_inode = ext3_destroy_inode,
1da177e4
LT		 798 .write_inode = ext3_write_inode,
799 .dirty_inode = ext3_dirty_inode,
785c4bcc 800 .drop_inode = ext3_drop_inode,
ac14a95b 801 .evict_inode = ext3_evict_inode,
1da177e4 802 .put_super = ext3_put_super,
1da177e4 803 .sync_fs = ext3_sync_fs,
c4be0c1d
TS		 804 .freeze_fs = ext3_freeze,
805 .unfreeze_fs = ext3_unfreeze,
1da177e4
LT		 806 .statfs = ext3_statfs,
807 .remount_fs = ext3_remount,
8fc2751b 808 .show_options = ext3_show_options,
1da177e4
LT		 809#ifdef CONFIG_QUOTA
810 .quota_read = ext3_quota_read,
811 .quota_write = ext3_quota_write,
812#endif
6b082b53 813 .bdev_try_to_free_page = bdev_try_to_free_page,
1da177e4
LT		 814};
815
39655164 816static const struct export_operations ext3_export_ops = {
74af0baa
CH		 817 .fh_to_dentry = ext3_fh_to_dentry,
818 .fh_to_parent = ext3_fh_to_parent,
1da177e4
LT		 819 .get_parent = ext3_get_parent,
820};
821
822enum {
823 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
824 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
2860b733 825 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
1da177e4 826 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
ade1a29e 827 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
71b96257 828 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
1da177e4 829 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
0e4fb5e2 830 Opt_data_err_abort, Opt_data_err_ignore,
1da177e4 831 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1aeec434 832 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
0636c73e
ES		 833 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
834 Opt_resize, Opt_usrquota, Opt_grpquota
1da177e4
LT		 835};
836
a447c093 837static const match_table_t tokens = {
1da177e4
LT		 838 {Opt_bsd_df, "bsddf"},
839 {Opt_minix_df, "minixdf"},
840 {Opt_grpid, "grpid"},
841 {Opt_grpid, "bsdgroups"},
842 {Opt_nogrpid, "nogrpid"},
843 {Opt_nogrpid, "sysvgroups"},
844 {Opt_resgid, "resgid=%u"},
845 {Opt_resuid, "resuid=%u"},
846 {Opt_sb, "sb=%u"},
847 {Opt_err_cont, "errors=continue"},
848 {Opt_err_panic, "errors=panic"},
849 {Opt_err_ro, "errors=remount-ro"},
850 {Opt_nouid32, "nouid32"},
851 {Opt_nocheck, "nocheck"},
852 {Opt_nocheck, "check=none"},
1da177e4
LT		 853 {Opt_debug, "debug"},
854 {Opt_oldalloc, "oldalloc"},
855 {Opt_orlov, "orlov"},
856 {Opt_user_xattr, "user_xattr"},
857 {Opt_nouser_xattr, "nouser_xattr"},
858 {Opt_acl, "acl"},
859 {Opt_noacl, "noacl"},
860 {Opt_reservation, "reservation"},
861 {Opt_noreservation, "noreservation"},
862 {Opt_noload, "noload"},
dee1d3b6 863 {Opt_noload, "norecovery"},
1da177e4 864 {Opt_nobh, "nobh"},
ade1a29e 865 {Opt_bh, "bh"},
1da177e4
LT		 866 {Opt_commit, "commit=%u"},
867 {Opt_journal_update, "journal=update"},
868 {Opt_journal_inum, "journal=%u"},
71b96257 869 {Opt_journal_dev, "journal_dev=%u"},
1da177e4
LT		 870 {Opt_abort, "abort"},
871 {Opt_data_journal, "data=journal"},
872 {Opt_data_ordered, "data=ordered"},
873 {Opt_data_writeback, "data=writeback"},
0e4fb5e2
HK		 874 {Opt_data_err_abort, "data_err=abort"},
875 {Opt_data_err_ignore, "data_err=ignore"},
1da177e4
LT		 876 {Opt_offusrjquota, "usrjquota="},
877 {Opt_usrjquota, "usrjquota=%s"},
878 {Opt_offgrpjquota, "grpjquota="},
879 {Opt_grpjquota, "grpjquota=%s"},
880 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
881 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1aeec434 882 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
8fc2751b 883 {Opt_grpquota, "grpquota"},
1f54587b
JK		 884 {Opt_noquota, "noquota"},
885 {Opt_quota, "quota"},
8fc2751b 886 {Opt_usrquota, "usrquota"},
1da177e4 887 {Opt_barrier, "barrier=%u"},
0636c73e
ES		 888 {Opt_barrier, "barrier"},
889 {Opt_nobarrier, "nobarrier"},
1da177e4 890 {Opt_resize, "resize"},
92587216 891 {Opt_err, NULL},
1da177e4
LT		 892};
893
4cf46b67 894static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
1da177e4 895{
e9ad5620
DK		 896 ext3_fsblk_t sb_block;
897 char *options = (char *) *data;
1da177e4
LT		 898
899 if (!options || strncmp(options, "sb=", 3) != 0)
900 return 1; /* Default location */
901 options += 3;
43d23f90 902 /*todo: use simple_strtoll with >32bit ext3 */
1da177e4
LT		 903 sb_block = simple_strtoul(options, &options, 0);
904 if (*options && *options != ',') {
4cf46b67 905 ext3_msg(sb, "error: invalid sb specification: %s",
1da177e4
LT		 906 (char *) *data);
907 return 1;
908 }
909 if (*options == ',')
910 options++;
911 *data = (void *) options;
912 return sb_block;
913}
914
e1f5c67a
DM		 915#ifdef CONFIG_QUOTA
916static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
917{
918 struct ext3_sb_info *sbi = EXT3_SB(sb);
919 char *qname;
920
921 if (sb_any_quota_loaded(sb) &&
922 !sbi->s_qf_names[qtype]) {
923 ext3_msg(sb, KERN_ERR,
924 "Cannot change journaled "
925 "quota options when quota turned on");
926 return 0;
927 }
928 qname = match_strdup(args);
929 if (!qname) {
930 ext3_msg(sb, KERN_ERR,
931 "Not enough memory for storing quotafile name");
932 return 0;
933 }
934 if (sbi->s_qf_names[qtype] &&
935 strcmp(sbi->s_qf_names[qtype], qname)) {
936 ext3_msg(sb, KERN_ERR,
937 "%s quota file already specified", QTYPE2NAME(qtype));
938 kfree(qname);
939 return 0;
940 }
941 sbi->s_qf_names[qtype] = qname;
942 if (strchr(sbi->s_qf_names[qtype], '/')) {
943 ext3_msg(sb, KERN_ERR,
944 "quotafile must be on filesystem root");
945 kfree(sbi->s_qf_names[qtype]);
946 sbi->s_qf_names[qtype] = NULL;
947 return 0;
948 }
949 set_opt(sbi->s_mount_opt, QUOTA);
950 return 1;
951}
952
953static int clear_qf_name(struct super_block *sb, int qtype) {
954
955 struct ext3_sb_info *sbi = EXT3_SB(sb);
956
957 if (sb_any_quota_loaded(sb) &&
958 sbi->s_qf_names[qtype]) {
959 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
960 " when quota turned on");
961 return 0;
962 }
963 /*
964 * The space will be released later when all options are confirmed
965 * to be correct
966 */
967 sbi->s_qf_names[qtype] = NULL;
968 return 1;
969}
970#endif
971
71b96257 972static int parse_options (char *options, struct super_block *sb,
eee194e7 973 unsigned int *inum, unsigned long *journal_devnum,
43d23f90 974 ext3_fsblk_t *n_blocks_count, int is_remount)
1da177e4
LT		 975{
976 struct ext3_sb_info *sbi = EXT3_SB(sb);
977 char * p;
978 substring_t args[MAX_OPT_ARGS];
979 int data_opt = 0;
980 int option;
981#ifdef CONFIG_QUOTA
e1f5c67a 982 int qfmt;
1da177e4
LT		 983#endif
984
985 if (!options)
986 return 1;
987
988 while ((p = strsep (&options, ",")) != NULL) {
989 int token;
990 if (!*p)
991 continue;
0636c73e
ES		 992 /*
993 * Initialize args struct so we know whether arg was
994 * found; some options take optional arguments.
995 */
996 args[0].to = args[0].from = 0;
1da177e4
LT		 997 token = match_token(p, tokens, args);
998 switch (token) {
999 case Opt_bsd_df:
1000 clear_opt (sbi->s_mount_opt, MINIX_DF);
1001 break;
1002 case Opt_minix_df:
1003 set_opt (sbi->s_mount_opt, MINIX_DF);
1004 break;
1005 case Opt_grpid:
1006 set_opt (sbi->s_mount_opt, GRPID);
1007 break;
1008 case Opt_nogrpid:
1009 clear_opt (sbi->s_mount_opt, GRPID);
1010 break;
1011 case Opt_resuid:
1012 if (match_int(&args[0], &option))
1013 return 0;
1014 sbi->s_resuid = option;
1015 break;
1016 case Opt_resgid:
1017 if (match_int(&args[0], &option))
1018 return 0;
1019 sbi->s_resgid = option;
1020 break;
1021 case Opt_sb:
1022 /* handled by get_sb_block() instead of here */
1023 /* *sb_block = match_int(&args[0]); */
1024 break;
1025 case Opt_err_panic:
1026 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1027 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1028 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1029 break;
1030 case Opt_err_ro:
1031 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1032 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1033 set_opt (sbi->s_mount_opt, ERRORS_RO);
1034 break;
1035 case Opt_err_cont:
1036 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1037 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1038 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1039 break;
1040 case Opt_nouid32:
1041 set_opt (sbi->s_mount_opt, NO_UID32);
1042 break;
1da177e4
LT		 1043 case Opt_nocheck:
1044 clear_opt (sbi->s_mount_opt, CHECK);
1045 break;
1046 case Opt_debug:
1047 set_opt (sbi->s_mount_opt, DEBUG);
1048 break;
1049 case Opt_oldalloc:
fbc85402
LC		 1050 ext3_msg(sb, KERN_WARNING,
1051 "Ignoring deprecated oldalloc option");
1da177e4
LT		 1052 break;
1053 case Opt_orlov:
fbc85402
LC		 1054 ext3_msg(sb, KERN_WARNING,
1055 "Ignoring deprecated orlov option");
1da177e4
LT		 1056 break;
1057#ifdef CONFIG_EXT3_FS_XATTR
1058 case Opt_user_xattr:
1059 set_opt (sbi->s_mount_opt, XATTR_USER);
1060 break;
1061 case Opt_nouser_xattr:
1062 clear_opt (sbi->s_mount_opt, XATTR_USER);
1063 break;
1064#else
1065 case Opt_user_xattr:
1066 case Opt_nouser_xattr:
4cf46b67
AF		 1067 ext3_msg(sb, KERN_INFO,
1068 "(no)user_xattr options not supported");
1da177e4
LT		 1069 break;
1070#endif
1071#ifdef CONFIG_EXT3_FS_POSIX_ACL
1072 case Opt_acl:
1073 set_opt(sbi->s_mount_opt, POSIX_ACL);
1074 break;
1075 case Opt_noacl:
1076 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1077 break;
1078#else
1079 case Opt_acl:
1080 case Opt_noacl:
4cf46b67
AF		 1081 ext3_msg(sb, KERN_INFO,
1082 "(no)acl options not supported");
1da177e4
LT		 1083 break;
1084#endif
1085 case Opt_reservation:
1086 set_opt(sbi->s_mount_opt, RESERVATION);
1087 break;
1088 case Opt_noreservation:
1089 clear_opt(sbi->s_mount_opt, RESERVATION);
1090 break;
1091 case Opt_journal_update:
1092 /* @@@ FIXME */
1093 /* Eventually we will want to be able to create
1094 a journal file here. For now, only allow the
1095 user to specify an existing inode to be the
1096 journal file. */
1097 if (is_remount) {
4cf46b67
AF		 1098 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1099 "journal on remount");
1da177e4
LT		 1100 return 0;
1101 }
1102 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1103 break;
1104 case Opt_journal_inum:
1105 if (is_remount) {
4cf46b67
AF		 1106 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1107 "journal on remount");
1da177e4
LT		 1108 return 0;
1109 }
1110 if (match_int(&args[0], &option))
1111 return 0;
1112 *inum = option;
1113 break;
71b96257
JL		 1114 case Opt_journal_dev:
1115 if (is_remount) {
4cf46b67
AF		 1116 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1117 "journal on remount");
71b96257
JL		 1118 return 0;
1119 }
1120 if (match_int(&args[0], &option))
1121 return 0;
1122 *journal_devnum = option;
1123 break;
1da177e4
LT		 1124 case Opt_noload:
1125 set_opt (sbi->s_mount_opt, NOLOAD);
1126 break;
1127 case Opt_commit:
1128 if (match_int(&args[0], &option))
1129 return 0;
1130 if (option < 0)
1131 return 0;
1132 if (option == 0)
1133 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1134 sbi->s_commit_interval = HZ * option;
1135 break;
1136 case Opt_data_journal:
1137 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1138 goto datacheck;
1139 case Opt_data_ordered:
1140 data_opt = EXT3_MOUNT_ORDERED_DATA;
1141 goto datacheck;
1142 case Opt_data_writeback:
1143 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1144 datacheck:
1145 if (is_remount) {
e3c96435 1146 if (test_opt(sb, DATA_FLAGS) == data_opt)
3c4cec65 1147 break;
4cf46b67
AF		 1148 ext3_msg(sb, KERN_ERR,
1149 "error: cannot change "
3c4cec65
JK		 1150 "data mode on remount. The filesystem "
1151 "is mounted in data=%s mode and you "
4cf46b67 1152 "try to remount it in data=%s mode.",
e3c96435
DM		 1153 data_mode_string(test_opt(sb,
1154 DATA_FLAGS)),
3c4cec65
JK		 1155 data_mode_string(data_opt));
1156 return 0;
1da177e4 1157 } else {
e3c96435 1158 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1da177e4
LT		 1159 sbi->s_mount_opt |= data_opt;
1160 }
1161 break;
0e4fb5e2
HK		 1162 case Opt_data_err_abort:
1163 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1164 break;
1165 case Opt_data_err_ignore:
1166 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1167 break;
1da177e4
LT		 1168#ifdef CONFIG_QUOTA
1169 case Opt_usrjquota:
e1f5c67a 1170 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1da177e4 1171 return 0;
e1f5c67a
DM		 1172 break;
1173 case Opt_grpjquota:
1174 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1da177e4 1175 return 0;
1da177e4
LT		 1176 break;
1177 case Opt_offusrjquota:
e1f5c67a
DM		 1178 if (!clear_qf_name(sb, USRQUOTA))
1179 return 0;
1180 break;
1da177e4 1181 case Opt_offgrpjquota:
e1f5c67a 1182 if (!clear_qf_name(sb, GRPQUOTA))
1da177e4 1183 return 0;
1da177e4
LT		 1184 break;
1185 case Opt_jqfmt_vfsold:
d06bf1d2
JK		 1186 qfmt = QFMT_VFS_OLD;
1187 goto set_qf_format;
1da177e4 1188 case Opt_jqfmt_vfsv0:
d06bf1d2 1189 qfmt = QFMT_VFS_V0;
1aeec434
JK		 1190 goto set_qf_format;
1191 case Opt_jqfmt_vfsv1:
1192 qfmt = QFMT_VFS_V1;
d06bf1d2 1193set_qf_format:
ee0d5ffe 1194 if (sb_any_quota_loaded(sb) &&
d06bf1d2 1195 sbi->s_jquota_fmt != qfmt) {
4cf46b67 1196 ext3_msg(sb, KERN_ERR, "error: cannot change "
d06bf1d2 1197 "journaled quota options when "
4cf46b67 1198 "quota turned on.");
d06bf1d2
JK		 1199 return 0;
1200 }
1201 sbi->s_jquota_fmt = qfmt;
1da177e4 1202 break;
1f54587b 1203 case Opt_quota:
8fc2751b
MB		 1204 case Opt_usrquota:
1205 set_opt(sbi->s_mount_opt, QUOTA);
1206 set_opt(sbi->s_mount_opt, USRQUOTA);
1207 break;
1208 case Opt_grpquota:
1f54587b 1209 set_opt(sbi->s_mount_opt, QUOTA);
8fc2751b 1210 set_opt(sbi->s_mount_opt, GRPQUOTA);
1f54587b
JK		 1211 break;
1212 case Opt_noquota:
ee0d5ffe 1213 if (sb_any_quota_loaded(sb)) {
4cf46b67
AF		 1214 ext3_msg(sb, KERN_ERR, "error: cannot change "
1215 "quota options when quota turned on.");
1f54587b
JK		 1216 return 0;
1217 }
1218 clear_opt(sbi->s_mount_opt, QUOTA);
8fc2751b
MB		 1219 clear_opt(sbi->s_mount_opt, USRQUOTA);
1220 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1f54587b 1221 break;
1da177e4 1222#else
8fc2751b
MB		 1223 case Opt_quota:
1224 case Opt_usrquota:
1225 case Opt_grpquota:
4cf46b67
AF		 1226 ext3_msg(sb, KERN_ERR,
1227 "error: quota options not supported.");
fa1ff1e0 1228 break;
1da177e4
LT		 1229 case Opt_usrjquota:
1230 case Opt_grpjquota:
1231 case Opt_offusrjquota:
1232 case Opt_offgrpjquota:
1233 case Opt_jqfmt_vfsold:
1234 case Opt_jqfmt_vfsv0:
1aeec434 1235 case Opt_jqfmt_vfsv1:
4cf46b67
AF		 1236 ext3_msg(sb, KERN_ERR,
1237 "error: journaled quota options not "
1238 "supported.");
1da177e4 1239 break;
1f54587b
JK		 1240 case Opt_noquota:
1241 break;
1da177e4
LT		 1242#endif
1243 case Opt_abort:
1244 set_opt(sbi->s_mount_opt, ABORT);
1245 break;
0636c73e
ES		 1246 case Opt_nobarrier:
1247 clear_opt(sbi->s_mount_opt, BARRIER);
1248 break;
1da177e4 1249 case Opt_barrier:
0636c73e
ES		 1250 if (args[0].from) {
1251 if (match_int(&args[0], &option))
1252 return 0;
1253 } else
1254 option = 1; /* No argument, default to 1 */
1da177e4
LT		 1255 if (option)
1256 set_opt(sbi->s_mount_opt, BARRIER);
1257 else
1258 clear_opt(sbi->s_mount_opt, BARRIER);
1259 break;
1260 case Opt_ignore:
1261 break;
1262 case Opt_resize:
08c6a96f 1263 if (!is_remount) {
4cf46b67
AF		 1264 ext3_msg(sb, KERN_ERR,
1265 "error: resize option only available "
1266 "for remount");
1da177e4
LT		 1267 return 0;
1268 }
c7f1721e
KZ		 1269 if (match_int(&args[0], &option) != 0)
1270 return 0;
1da177e4
LT		 1271 *n_blocks_count = option;
1272 break;
1273 case Opt_nobh:
4c4d3901
CH		 1274 ext3_msg(sb, KERN_WARNING,
1275 "warning: ignoring deprecated nobh option");
1da177e4 1276 break;
ade1a29e 1277 case Opt_bh:
4c4d3901
CH		 1278 ext3_msg(sb, KERN_WARNING,
1279 "warning: ignoring deprecated bh option");
ade1a29e 1280 break;
1da177e4 1281 default:
4cf46b67
AF		 1282 ext3_msg(sb, KERN_ERR,
1283 "error: unrecognized mount option \"%s\" "
1284 "or missing value", p);
1da177e4
LT		 1285 return 0;
1286 }
1287 }
1288#ifdef CONFIG_QUOTA
8fc2751b 1289 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
e3c96435 1290 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
8fc2751b 1291 clear_opt(sbi->s_mount_opt, USRQUOTA);
e3c96435 1292 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
8fc2751b
MB		 1293 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1294
e1f5c67a 1295 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
4cf46b67
AF		 1296 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1297 "format mixing.");
8fc2751b
MB		 1298 return 0;
1299 }
1300
1301 if (!sbi->s_jquota_fmt) {
4cf46b67
AF		 1302 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1303 "not specified.");
8fc2751b
MB		 1304 return 0;
1305 }
1306 } else {
1307 if (sbi->s_jquota_fmt) {
4cf46b67 1308 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
99aeaf63 1309 "specified with no journaling "
4cf46b67 1310 "enabled.");
8fc2751b
MB		 1311 return 0;
1312 }
1da177e4
LT		 1313 }
1314#endif
1da177e4
LT		 1315 return 1;
1316}
1317
1318static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1319 int read_only)
1320{
1321 struct ext3_sb_info *sbi = EXT3_SB(sb);
1322 int res = 0;
1323
1324 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
4cf46b67
AF		 1325 ext3_msg(sb, KERN_ERR,
1326 "error: revision level too high, "
1327 "forcing read-only mode");
1da177e4
LT		 1328 res = MS_RDONLY;
1329 }
1330 if (read_only)
1331 return res;
1332 if (!(sbi->s_mount_state & EXT3_VALID_FS))
4cf46b67
AF		 1333 ext3_msg(sb, KERN_WARNING,
1334 "warning: mounting unchecked fs, "
1335 "running e2fsck is recommended");
1da177e4 1336 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
4cf46b67
AF		 1337 ext3_msg(sb, KERN_WARNING,
1338 "warning: mounting fs with errors, "
1339 "running e2fsck is recommended");
df0d6b8f 1340 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1da177e4 1341 le16_to_cpu(es->s_mnt_count) >=
df0d6b8f 1342 le16_to_cpu(es->s_max_mnt_count))
4cf46b67
AF		 1343 ext3_msg(sb, KERN_WARNING,
1344 "warning: maximal mount count reached, "
1345 "running e2fsck is recommended");
1da177e4
LT		 1346 else if (le32_to_cpu(es->s_checkinterval) &&
1347 (le32_to_cpu(es->s_lastcheck) +
1348 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
4cf46b67
AF		 1349 ext3_msg(sb, KERN_WARNING,
1350 "warning: checktime reached, "
1351 "running e2fsck is recommended");
1da177e4
LT		 1352#if 0
1353 /* @@@ We _will_ want to clear the valid bit if we find
1354 inconsistencies, to force a fsck at reboot. But for
1355 a plain journaled filesystem we can keep it set as
1356 valid forever! :) */
e7f23ebd 1357 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1da177e4 1358#endif
df0d6b8f 1359 if (!le16_to_cpu(es->s_max_mnt_count))
1da177e4 1360 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
50e8a289 1361 le16_add_cpu(&es->s_mnt_count, 1);
1da177e4
LT		 1362 es->s_mtime = cpu_to_le32(get_seconds());
1363 ext3_update_dynamic_rev(sb);
1364 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1365
1366 ext3_commit_super(sb, es, 1);
1367 if (test_opt(sb, DEBUG))
4cf46b67
AF		 1368 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1369 "bpg=%lu, ipg=%lu, mo=%04lx]",
1da177e4
LT		 1370 sb->s_blocksize,
1371 sbi->s_groups_count,
1372 EXT3_BLOCKS_PER_GROUP(sb),
1373 EXT3_INODES_PER_GROUP(sb),
1374 sbi->s_mount_opt);
1375
1da177e4
LT		 1376 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1377 char b[BDEVNAME_SIZE];
4cf46b67 1378 ext3_msg(sb, KERN_INFO, "using external journal on %s",
1da177e4
LT		 1379 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1380 } else {
4cf46b67 1381 ext3_msg(sb, KERN_INFO, "using internal journal");
1da177e4 1382 }
d71bc6db 1383 cleancache_init_fs(sb);
1da177e4
LT		 1384 return res;
1385}
1386
1387/* Called at mount-time, super-block is locked */
197cd65a 1388static int ext3_check_descriptors(struct super_block *sb)
1da177e4
LT		 1389{
1390 struct ext3_sb_info *sbi = EXT3_SB(sb);
1da177e4
LT		 1391 int i;
1392
1393 ext3_debug ("Checking group descriptors");
1394
197cd65a
AM		 1395 for (i = 0; i < sbi->s_groups_count; i++) {
1396 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1eaafeae
AM		 1397 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1398 ext3_fsblk_t last_block;
197cd65a 1399
855565e8
ES		 1400 if (i == sbi->s_groups_count - 1)
1401 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1402 else
1403 last_block = first_block +
1404 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1405
855565e8
ES		 1406 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1407 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1da177e4
LT		 1408 {
1409 ext3_error (sb, "ext3_check_descriptors",
1410 "Block bitmap for group %d"
1411 " not in group (block %lu)!",
1412 i, (unsigned long)
1413 le32_to_cpu(gdp->bg_block_bitmap));
1414 return 0;
1415 }
855565e8
ES		 1416 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1417 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1da177e4
LT		 1418 {
1419 ext3_error (sb, "ext3_check_descriptors",
1420 "Inode bitmap for group %d"
1421 " not in group (block %lu)!",
1422 i, (unsigned long)
1423 le32_to_cpu(gdp->bg_inode_bitmap));
1424 return 0;
1425 }
855565e8 1426 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
780dcdb2 1427 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
855565e8 1428 last_block)
1da177e4
LT		 1429 {
1430 ext3_error (sb, "ext3_check_descriptors",
1431 "Inode table for group %d"
1432 " not in group (block %lu)!",
1433 i, (unsigned long)
1434 le32_to_cpu(gdp->bg_inode_table));
1435 return 0;
1436 }
1da177e4
LT		 1437 }
1438
1439 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1440 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1441 return 1;
1442}
1443
1444
1445/* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1446 * the superblock) which were deleted from all directories, but held open by
1447 * a process at the time of a crash. We walk the list and try to delete these
1448 * inodes at recovery time (only with a read-write filesystem).
1449 *
1450 * In order to keep the orphan inode chain consistent during traversal (in
1451 * case of crash during recovery), we link each inode into the superblock
1452 * orphan list_head and handle it the same way as an inode deletion during
1453 * normal operation (which journals the operations for us).
1454 *
1455 * We only do an iget() and an iput() on each inode, which is very safe if we
1456 * accidentally point at an in-use or already deleted inode. The worst that
1457 * can happen in this case is that we get a "bit already cleared" message from
1458 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1459 * e2fsck was run on this filesystem, and it must have already done the orphan
1460 * inode cleanup for us, so we can safely abort without any further action.
1461 */
1462static void ext3_orphan_cleanup (struct super_block * sb,
1463 struct ext3_super_block * es)
1464{
1465 unsigned int s_flags = sb->s_flags;
1466 int nr_orphans = 0, nr_truncates = 0;
1467#ifdef CONFIG_QUOTA
1468 int i;
1469#endif
1470 if (!es->s_last_orphan) {
1471 jbd_debug(4, "no orphan inodes to clean up\n");
1472 return;
1473 }
1474
a8f48a95 1475 if (bdev_read_only(sb->s_bdev)) {
4cf46b67
AF		 1476 ext3_msg(sb, KERN_ERR, "error: write access "
1477 "unavailable, skipping orphan cleanup.");
a8f48a95
ES		 1478 return;
1479 }
1480
ce654b37
AG		 1481 /* Check if feature set allows readwrite operations */
1482 if (EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP)) {
1483 ext3_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
1484 "unknown ROCOMPAT features");
1485 return;
a8f48a95
ES		 1486 }
1487
1da177e4
LT		 1488 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1489 if (es->s_last_orphan)
1490 jbd_debug(1, "Errors on filesystem, "
1491 "clearing orphan list.\n");
1492 es->s_last_orphan = 0;
1493 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1494 return;
1495 }
1496
1497 if (s_flags & MS_RDONLY) {
4cf46b67 1498 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1da177e4
LT		 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)
4cf46b67
AF		 1509 ext3_msg(sb, KERN_ERR,
1510 "error: cannot turn on journaled "
1511 "quota: %d", ret);
1da177e4
LT		 1512 }
1513 }
1514#endif
1515
1516 while (es->s_last_orphan) {
1517 struct inode *inode;
1518
473043dc
DH		 1519 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1520 if (IS_ERR(inode)) {
1da177e4
LT		 1521 es->s_last_orphan = 0;
1522 break;
1523 }
1524
1525 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
871a2931 1526 dquot_initialize(inode);
1da177e4
LT		 1527 if (inode->i_nlink) {
1528 printk(KERN_DEBUG
eee194e7 1529 "%s: truncating inode %lu to %Ld bytes\n",
e05b6b52 1530 __func__, inode->i_ino, inode->i_size);
eee194e7 1531 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1da177e4
LT		 1532 inode->i_ino, inode->i_size);
1533 ext3_truncate(inode);
1534 nr_truncates++;
1535 } else {
1536 printk(KERN_DEBUG
eee194e7 1537 "%s: deleting unreferenced inode %lu\n",
e05b6b52 1538 __func__, inode->i_ino);
eee194e7 1539 jbd_debug(2, "deleting unreferenced inode %lu\n",
1da177e4
LT		 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)
4cf46b67
AF		 1549 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1550 PLURAL(nr_orphans));
1da177e4 1551 if (nr_truncates)
4cf46b67
AF		 1552 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1553 PLURAL(nr_truncates));
1da177e4
LT		 1554#ifdef CONFIG_QUOTA
1555 /* Turn quotas off */
1556 for (i = 0; i < MAXQUOTAS; i++) {
1557 if (sb_dqopt(sb)->files[i])
287a8095 1558 dquot_quota_off(sb, i);
1da177e4
LT		 1559 }
1560#endif
1561 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1562}
1563
1da177e4
LT		 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 */
1569static loff_t ext3_max_size(int bits)
1570{
1571 loff_t res = EXT3_NDIR_BLOCKS;
fe7fdc37
AK		 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
1da177e4 1577 * sectors in the file, including data and all indirect blocks,
fe7fdc37
AK		 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;
1da177e4
LT		 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;
fe7fdc37
AK		 1604
1605 if (res > MAX_LFS_FILESIZE)
1606 res = MAX_LFS_FILESIZE;
1607
1da177e4
LT		 1608 return res;
1609}
1610
43d23f90
MC		 1611static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1612 ext3_fsblk_t logic_sb_block,
1da177e4
LT		 1613 int nr)
1614{
1615 struct ext3_sb_info *sbi = EXT3_SB(sb);
43d23f90 1616 unsigned long bg, first_meta_bg;
1da177e4
LT		 1617 int has_super = 0;
1618
1da177e4
LT		 1619 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1620
1621 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
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;
43d23f90 1627 return (has_super + ext3_group_first_block_no(sb, bg));
1da177e4
LT		 1628}
1629
1630
1631static 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;
43d23f90 1636 ext3_fsblk_t block;
4cf46b67 1637 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
43d23f90 1638 ext3_fsblk_t logic_sb_block;
1da177e4 1639 unsigned long offset = 0;
eee194e7 1640 unsigned int journal_inum = 0;
71b96257 1641 unsigned long journal_devnum = 0;
1da177e4
LT		 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;
473043dc 1649 int ret = -EINVAL;
1da177e4 1650 __le32 features;
833f4077 1651 int err;
1da177e4 1652
f8314dc6 1653 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1da177e4
LT		 1654 if (!sbi)
1655 return -ENOMEM;
5df096d6
PE		 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 }
1da177e4 1663 sb->s_fs_info = sbi;
1da177e4
LT		 1664 sbi->s_mount_opt = 0;
1665 sbi->s_resuid = EXT3_DEF_RESUID;
1666 sbi->s_resgid = EXT3_DEF_RESGID;
571beed8 1667 sbi->s_sb_block = sb_block;
1da177e4 1668
1da177e4
LT		 1669 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1670 if (!blocksize) {
4cf46b67 1671 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1da177e4
LT		 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))) {
4cf46b67 1687 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1da177e4
LT		 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 */
57e94d86 1694 es = (struct ext3_super_block *) (bh->b_data + offset);
1da177e4
LT		 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);
2e7842b8 1708#ifdef CONFIG_EXT3_FS_XATTR
1da177e4
LT		 1709 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1710 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1711#endif
1712#ifdef CONFIG_EXT3_FS_POSIX_ACL
1da177e4
LT		 1713 if (def_mount_opts & EXT3_DEFM_ACL)
1714 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1715#endif
1da177e4 1716 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
e3c96435 1717 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1da177e4 1718 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
e3c96435 1719 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1da177e4 1720 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
e3c96435 1721 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1da177e4
LT		 1722
1723 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1724 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1eca93f9 1725 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
2245d7c2 1726 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1eca93f9
AK		 1727 else
1728 set_opt(sbi->s_mount_opt, ERRORS_RO);
1da177e4
LT		 1729
1730 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1731 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1732
00eacd66
CH		 1733 /* enable barriers by default */
1734 set_opt(sbi->s_mount_opt, BARRIER);
1da177e4
LT		 1735 set_opt(sbi->s_mount_opt, RESERVATION);
1736
71b96257
JL		 1737 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1738 NULL, 0))
1da177e4
LT		 1739 goto failed_mount;
1740
1741 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
e3c96435 1742 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1da177e4
LT		 1743
1744 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1745 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1746 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1747 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
4cf46b67
AF		 1748 ext3_msg(sb, KERN_WARNING,
1749 "warning: feature flags set on rev 0 fs, "
1750 "running e2fsck is recommended");
1da177e4
LT		 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 */
1756 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1757 if (features) {
4cf46b67
AF		 1758 ext3_msg(sb, KERN_ERR,
1759 "error: couldn't mount because of unsupported "
1760 "optional features (%x)", le32_to_cpu(features));
1da177e4
LT		 1761 goto failed_mount;
1762 }
1763 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1764 if (!(sb->s_flags & MS_RDONLY) && features) {
4cf46b67
AF		 1765 ext3_msg(sb, KERN_ERR,
1766 "error: couldn't mount RDWR because of unsupported "
1767 "optional features (%x)", le32_to_cpu(features));
1da177e4
LT		 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) {
4cf46b67
AF		 1774 ext3_msg(sb, KERN_ERR,
1775 "error: couldn't mount because of unsupported "
1776 "filesystem blocksize %d", blocksize);
1da177e4
LT		 1777 goto failed_mount;
1778 }
1779
e1defc4f 1780 hblock = bdev_logical_block_size(sb->s_bdev);
1da177e4
LT		 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) {
4cf46b67
AF		 1787 ext3_msg(sb, KERN_ERR,
1788 "error: fsblocksize %d too small for "
1789 "hardware sectorsize %d", blocksize, hblock);
1da177e4
LT		 1790 goto failed_mount;
1791 }
1792
1793 brelse (bh);
0f0a89eb 1794 if (!sb_set_blocksize(sb, blocksize)) {
4cf46b67
AF		 1795 ext3_msg(sb, KERN_ERR,
1796 "error: bad blocksize %d", blocksize);
0f0a89eb
TS		 1797 goto out_fail;
1798 }
1da177e4
LT		 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) {
4cf46b67
AF		 1803 ext3_msg(sb, KERN_ERR,
1804 "error: can't read superblock on 2nd try");
1da177e4
LT		 1805 goto failed_mount;
1806 }
57e94d86 1807 es = (struct ext3_super_block *)(bh->b_data + offset);
1da177e4
LT		 1808 sbi->s_es = es;
1809 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
4cf46b67
AF		 1810 ext3_msg(sb, KERN_ERR,
1811 "error: magic mismatch");
1da177e4
LT		 1812 goto failed_mount;
1813 }
1814 }
1815
1816 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1817
1818 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1819 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1820 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1821 } else {
1822 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1823 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1824 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
3fc74269 1825 (!is_power_of_2(sbi->s_inode_size)) ||
1da177e4 1826 (sbi->s_inode_size > blocksize)) {
4cf46b67
AF		 1827 ext3_msg(sb, KERN_ERR,
1828 "error: unsupported inode size: %d",
1da177e4
LT		 1829 sbi->s_inode_size);
1830 goto failed_mount;
1831 }
1832 }
1833 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1834 le32_to_cpu(es->s_log_frag_size);
1835 if (blocksize != sbi->s_frag_size) {
4cf46b67
AF		 1836 ext3_msg(sb, KERN_ERR,
1837 "error: fragsize %lu != blocksize %u (unsupported)",
1da177e4
LT		 1838 sbi->s_frag_size, blocksize);
1839 goto failed_mount;
1840 }
1841 sbi->s_frags_per_block = 1;
1842 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1843 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1844 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 1845 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1da177e4
LT		 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);
f0d1b0b3
DH		 1855 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1856 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1da177e4
LT		 1857 for (i=0; i < 4; i++)
1858 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1859 sbi->s_def_hash_version = es->s_def_hash_version;
5e1f8c9e
TT		 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
5e1f8c9e 1870 }
1da177e4
LT		 1871
1872 if (sbi->s_blocks_per_group > blocksize * 8) {
4cf46b67
AF		 1873 ext3_msg(sb, KERN_ERR,
1874 "#blocks per group too big: %lu",
1da177e4
LT		 1875 sbi->s_blocks_per_group);
1876 goto failed_mount;
1877 }
1878 if (sbi->s_frags_per_group > blocksize * 8) {
4cf46b67
AF		 1879 ext3_msg(sb, KERN_ERR,
1880 "error: #fragments per group too big: %lu",
1da177e4
LT		 1881 sbi->s_frags_per_group);
1882 goto failed_mount;
1883 }
1884 if (sbi->s_inodes_per_group > blocksize * 8) {
4cf46b67
AF		 1885 ext3_msg(sb, KERN_ERR,
1886 "error: #inodes per group too big: %lu",
1da177e4
LT		 1887 sbi->s_inodes_per_group);
1888 goto failed_mount;
1889 }
1890
ad692bf3
DW		 1891 err = generic_check_addressable(sb->s_blocksize_bits,
1892 le32_to_cpu(es->s_blocks_count));
1893 if (err) {
4cf46b67
AF		 1894 ext3_msg(sb, KERN_ERR,
1895 "error: filesystem is too large to mount safely");
fcd5df35 1896 if (sizeof(sector_t) < 8)
4cf46b67
AF		 1897 ext3_msg(sb, KERN_ERR,
1898 "error: CONFIG_LBDAF not enabled");
ad692bf3 1899 ret = err;
fcd5df35
MC		 1900 goto failed_mount;
1901 }
1902
1da177e4
LT		 1903 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1904 goto cantfind_ext3;
855565e8
ES		 1905 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1906 le32_to_cpu(es->s_first_data_block) - 1)
1907 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
81a4e320 1908 db_count = DIV_ROUND_UP(sbi->s_groups_count, EXT3_DESC_PER_BLOCK(sb));
1da177e4
LT		 1909 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1910 GFP_KERNEL);
1911 if (sbi->s_group_desc == NULL) {
4cf46b67
AF		 1912 ext3_msg(sb, KERN_ERR,
1913 "error: not enough memory");
4569cd1b 1914 ret = -ENOMEM;
1da177e4
LT		 1915 goto failed_mount;
1916 }
1917
5df096d6 1918 bgl_lock_init(sbi->s_blockgroup_lock);
1da177e4
LT		 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]) {
4cf46b67
AF		 1924 ext3_msg(sb, KERN_ERR,
1925 "error: can't read group descriptor %d", i);
1da177e4
LT		 1926 db_count = i;
1927 goto failed_mount2;
1928 }
1929 }
1930 if (!ext3_check_descriptors (sb)) {
4cf46b67
AF		 1931 ext3_msg(sb, KERN_ERR,
1932 "error: group descriptors corrupted");
1da177e4
LT		 1933 goto failed_mount2;
1934 }
1935 sbi->s_gdb_count = db_count;
1936 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1937 spin_lock_init(&sbi->s_next_gen_lock);
0216bfcf 1938
1da177e4
LT		 1939 /* per fileystem reservation list head & lock */
1940 spin_lock_init(&sbi->s_rsv_window_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. */
1946 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1947 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1948 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1949 sbi->s_rsv_window_head.rsv_goal_size = 0;
1950 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
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;
1957 sb->s_xattr = ext3_xattr_handlers;
1958#ifdef CONFIG_QUOTA
1959 sb->s_qcop = &ext3_qctl_operations;
1960 sb->dq_op = &ext3_quota_operations;
1961#endif
03cb5f03 1962 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
1da177e4 1963 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
b8a052d0 1964 mutex_init(&sbi->s_orphan_lock);
96d2a495 1965 mutex_init(&sbi->s_resize_lock);
1da177e4
LT		 1966
1967 sb->s_root = NULL;
1968
1969 needs_recovery = (es->s_last_orphan != 0 ||
1970 EXT3_HAS_INCOMPAT_FEATURE(sb,
1971 EXT3_FEATURE_INCOMPAT_RECOVER));
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) &&
1978 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
71b96257 1979 if (ext3_load_journal(sb, es, journal_devnum))
41d1a636 1980 goto failed_mount2;
1da177e4
LT		 1981 } else if (journal_inum) {
1982 if (ext3_create_journal(sb, es, journal_inum))
41d1a636 1983 goto failed_mount2;
1da177e4
LT		 1984 } else {
1985 if (!silent)
4cf46b67
AF		 1986 ext3_msg(sb, KERN_ERR,
1987 "error: no journal found. "
1988 "mounting ext3 over ext2?");
41d1a636
DM		 1989 goto failed_mount2;
1990 }
1991 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1992 ext3_count_free_blocks(sb));
1993 if (!err) {
1994 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1995 ext3_count_free_inodes(sb));
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");
4569cd1b 2003 ret = err;
0216bfcf 2004 goto failed_mount3;
1da177e4
LT		 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 */
2014 if (journal_check_available_features
2015 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
bbae8bcc 2016 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1da177e4
LT		 2017 else
2018 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2019 break;
2020
2021 case EXT3_MOUNT_ORDERED_DATA:
2022 case EXT3_MOUNT_WRITEBACK_DATA:
2023 if (!journal_check_available_features
2024 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
4cf46b67
AF		 2025 ext3_msg(sb, KERN_ERR,
2026 "error: journal does not support "
2027 "requested data journaling mode");
41d1a636 2028 goto failed_mount3;
1da177e4
LT		 2029 }
2030 default:
2031 break;
2032 }
2033
1da177e4
LT		 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
473043dc
DH		 2039 root = ext3_iget(sb, EXT3_ROOT_INO);
2040 if (IS_ERR(root)) {
4cf46b67 2041 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
473043dc 2042 ret = PTR_ERR(root);
41d1a636 2043 goto failed_mount3;
1da177e4
LT		 2044 }
2045 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
473043dc 2046 iput(root);
4cf46b67 2047 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
41d1a636 2048 goto failed_mount3;
1da177e4 2049 }
473043dc
DH		 2050 sb->s_root = d_alloc_root(root);
2051 if (!sb->s_root) {
4cf46b67 2052 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
473043dc
DH		 2053 iput(root);
2054 ret = -ENOMEM;
41d1a636 2055 goto failed_mount3;
473043dc 2056 }
1da177e4
LT		 2057
2058 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
ed505ee4 2059
1da177e4
LT		 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;
63894ab9
EG		 2063 if (needs_recovery) {
2064 ext3_mark_recovery_complete(sb, es);
4cf46b67 2065 ext3_msg(sb, KERN_INFO, "recovery complete");
63894ab9 2066 }
4cf46b67 2067 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
1da177e4
LT		 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
1da177e4
LT		 2072 return 0;
2073
2074cantfind_ext3:
2075 if (!silent)
4cf46b67
AF		 2076 ext3_msg(sb, KERN_INFO,
2077 "error: can't find ext3 filesystem on dev %s.",
1da177e4
LT		 2078 sb->s_id);
2079 goto failed_mount;
2080
0216bfcf
MC		 2081failed_mount3:
2082 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2083 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2084 percpu_counter_destroy(&sbi->s_dirs_counter);
41d1a636 2085 journal_destroy(sbi->s_journal);
1da177e4
LT		 2086failed_mount2:
2087 for (i = 0; i < db_count; i++)
2088 brelse(sbi->s_group_desc[i]);
2089 kfree(sbi->s_group_desc);
2090failed_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);
2097out_fail:
2098 sb->s_fs_info = NULL;
de5ce037 2099 kfree(sbi->s_blockgroup_lock);
1da177e4 2100 kfree(sbi);
473043dc 2101 return ret;
1da177e4
LT		 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
ae6ddcc5 2107 * done any recovery; and again on any subsequent remount.
1da177e4
LT		 2108 */
2109static 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;
0e4fb5e2
HK		 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;
1da177e4
LT		 2128 spin_unlock(&journal->j_state_lock);
2129}
2130
eee194e7
ES		 2131static journal_t *ext3_get_journal(struct super_block *sb,
2132 unsigned int journal_inum)
1da177e4
LT		 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
473043dc
DH		 2141 journal_inode = ext3_iget(sb, journal_inum);
2142 if (IS_ERR(journal_inode)) {
4cf46b67 2143 ext3_msg(sb, KERN_ERR, "error: no journal found");
1da177e4
LT		 2144 return NULL;
2145 }
2146 if (!journal_inode->i_nlink) {
2147 make_bad_inode(journal_inode);
2148 iput(journal_inode);
4cf46b67 2149 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
1da177e4
LT		 2150 return NULL;
2151 }
2152
2153 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2154 journal_inode, journal_inode->i_size);
473043dc 2155 if (!S_ISREG(journal_inode->i_mode)) {
4cf46b67 2156 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
1da177e4
LT		 2157 iput(journal_inode);
2158 return NULL;
2159 }
2160
2161 journal = journal_init_inode(journal_inode);
2162 if (!journal) {
4cf46b67 2163 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
1da177e4
LT		 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
2172static 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;
43d23f90 2177 ext3_fsblk_t start;
1c2bf374 2178 ext3_fsblk_t len;
1da177e4 2179 int hblock, blocksize;
43d23f90 2180 ext3_fsblk_t sb_block;
1da177e4
LT		 2181 unsigned long offset;
2182 struct ext3_super_block * es;
2183 struct block_device *bdev;
2184
4cf46b67 2185 bdev = ext3_blkdev_get(j_dev, sb);
1da177e4
LT		 2186 if (bdev == NULL)
2187 return NULL;
2188
1da177e4 2189 blocksize = sb->s_blocksize;
e1defc4f 2190 hblock = bdev_logical_block_size(bdev);
1da177e4 2191 if (blocksize < hblock) {
4cf46b67
AF		 2192 ext3_msg(sb, KERN_ERR,
2193 "error: blocksize too small for journal device");
1da177e4
LT		 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))) {
4cf46b67
AF		 2201 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2202 "external journal");
1da177e4
LT		 2203 goto out_bdev;
2204 }
2205
57e94d86 2206 es = (struct ext3_super_block *) (bh->b_data + offset);
1da177e4
LT		 2207 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2208 !(le32_to_cpu(es->s_feature_incompat) &
2209 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4cf46b67
AF		 2210 ext3_msg(sb, KERN_ERR, "error: external journal has "
2211 "bad superblock");
1da177e4
LT		 2212 brelse(bh);
2213 goto out_bdev;
2214 }
2215
2216 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4cf46b67 2217 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
1da177e4
LT		 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) {
4cf46b67
AF		 2229 ext3_msg(sb, KERN_ERR,
2230 "error: failed to create device journal");
1da177e4
LT		 2231 goto out_bdev;
2232 }
2233 journal->j_private = sb;
2234 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2235 wait_on_buffer(journal->j_sb_buffer);
2236 if (!buffer_uptodate(journal->j_sb_buffer)) {
4cf46b67 2237 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
1da177e4
LT		 2238 goto out_journal;
2239 }
2240 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4cf46b67
AF		 2241 ext3_msg(sb, KERN_ERR,
2242 "error: external journal has more than one "
2243 "user (unsupported) - %d",
1da177e4
LT		 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;
2250out_journal:
2251 journal_destroy(journal);
2252out_bdev:
2253 ext3_blkdev_put(bdev);
2254 return NULL;
2255}
2256
71b96257
JL		 2257static int ext3_load_journal(struct super_block *sb,
2258 struct ext3_super_block *es,
2259 unsigned long journal_devnum)
1da177e4
LT		 2260{
2261 journal_t *journal;
eee194e7 2262 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
71b96257 2263 dev_t journal_dev;
1da177e4
LT		 2264 int err = 0;
2265 int really_read_only;
2266
71b96257
JL		 2267 if (journal_devnum &&
2268 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4cf46b67
AF		 2269 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2270 "numbers have changed");
71b96257
JL		 2271 journal_dev = new_decode_dev(journal_devnum);
2272 } else
2273 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2274
1da177e4
LT		 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
2283 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2284 if (sb->s_flags & MS_RDONLY) {
4cf46b67
AF		 2285 ext3_msg(sb, KERN_INFO,
2286 "recovery required on readonly filesystem");
1da177e4 2287 if (really_read_only) {
4cf46b67
AF		 2288 ext3_msg(sb, KERN_ERR, "error: write access "
2289 "unavailable, cannot proceed");
1da177e4
LT		 2290 return -EROFS;
2291 }
4cf46b67
AF		 2292 ext3_msg(sb, KERN_INFO,
2293 "write access will be enabled during recovery");
1da177e4
LT		 2294 }
2295 }
2296
2297 if (journal_inum && journal_dev) {
4cf46b67
AF		 2298 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2299 "and inode journals");
1da177e4
LT		 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
0636c73e
ES		 2311 if (!(journal->j_flags & JFS_BARRIER))
2312 printk(KERN_INFO "EXT3-fs: barriers not enabled\n");
2313
1da177e4
LT		 2314 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2315 err = journal_update_format(journal);
2316 if (err) {
4cf46b67 2317 ext3_msg(sb, KERN_ERR, "error updating journal");
1da177e4
LT		 2318 journal_destroy(journal);
2319 return err;
2320 }
2321 }
2322
2323 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2324 err = journal_wipe(journal, !really_read_only);
2325 if (!err)
2326 err = journal_load(journal);
2327
2328 if (err) {
4cf46b67 2329 ext3_msg(sb, KERN_ERR, "error loading journal");
1da177e4
LT		 2330 journal_destroy(journal);
2331 return err;
2332 }
2333
2334 EXT3_SB(sb)->s_journal = journal;
2335 ext3_clear_journal_err(sb, es);
71b96257 2336
31d710a7 2337 if (!really_read_only && journal_devnum &&
71b96257
JL		 2338 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2339 es->s_journal_dev = cpu_to_le32(journal_devnum);
71b96257
JL		 2340
2341 /* Make sure we flush the recovery flag to disk. */
2342 ext3_commit_super(sb, es, 1);
2343 }
2344
1da177e4
LT		 2345 return 0;
2346}
2347
4cf46b67
AF		 2348static int ext3_create_journal(struct super_block *sb,
2349 struct ext3_super_block *es,
eee194e7 2350 unsigned int journal_inum)
1da177e4
LT		 2351{
2352 journal_t *journal;
952d9de1 2353 int err;
1da177e4
LT		 2354
2355 if (sb->s_flags & MS_RDONLY) {
4cf46b67
AF		 2356 ext3_msg(sb, KERN_ERR,
2357 "error: readonly filesystem when trying to "
2358 "create journal");
1da177e4
LT		 2359 return -EROFS;
2360 }
2361
952d9de1
BP		 2362 journal = ext3_get_journal(sb, journal_inum);
2363 if (!journal)
1da177e4
LT		 2364 return -EINVAL;
2365
4cf46b67 2366 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
1da177e4
LT		 2367 journal_inum);
2368
952d9de1
BP		 2369 err = journal_create(journal);
2370 if (err) {
4cf46b67 2371 ext3_msg(sb, KERN_ERR, "error creating journal");
1da177e4
LT		 2372 journal_destroy(journal);
2373 return -EIO;
2374 }
2375
2376 EXT3_SB(sb)->s_journal = journal;
2377
2378 ext3_update_dynamic_rev(sb);
2379 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2380 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2381
2382 es->s_journal_inum = cpu_to_le32(journal_inum);
1da177e4
LT		 2383
2384 /* Make sure we flush the recovery flag to disk. */
2385 ext3_commit_super(sb, es, 1);
2386
2387 return 0;
2388}
2389
c4be0c1d
TS		 2390static int ext3_commit_super(struct super_block *sb,
2391 struct ext3_super_block *es,
1da177e4
LT		 2392 int sync)
2393{
2394 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
c4be0c1d 2395 int error = 0;
1da177e4
LT		 2396
2397 if (!sbh)
c4be0c1d 2398 return error;
dff6825e
DW		 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 }
96ec2e0a
TT		 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());
1da177e4
LT		 2426 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2427 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2428 BUFFER_TRACE(sbh, "marking dirty");
2429 mark_buffer_dirty(sbh);
dff6825e 2430 if (sync) {
c4be0c1d 2431 error = sync_dirty_buffer(sbh);
dff6825e
DW		 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 }
c4be0c1d 2439 return error;
1da177e4
LT		 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 */
2448static 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);
2d7c820e
HK		 2454 if (journal_flush(journal) < 0)
2455 goto out;
2456
1da177e4
LT		 2457 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2458 sb->s_flags & MS_RDONLY) {
2459 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1da177e4
LT		 2460 ext3_commit_super(sb, es, 1);
2461 }
2d7c820e
HK		 2462
2463out:
1da177e4
LT		 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 */
4cf46b67
AF		 2472static void ext3_clear_journal_err(struct super_block *sb,
2473 struct ext3_super_block *es)
1da177e4
LT		 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);
e05b6b52 2491 ext3_warning(sb, __func__, "Filesystem error recorded "
1da177e4 2492 "from previous mount: %s", errstr);
e05b6b52 2493 ext3_warning(sb, __func__, "Marking fs in need of "
1da177e4
LT		 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 */
2508int ext3_force_commit(struct super_block *sb)
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;
1da177e4
LT		 2517 ret = ext3_journal_force_commit(journal);
2518 return ret;
2519}
2520
1da177e4
LT		 2521static int ext3_sync_fs(struct super_block *sb, int wait)
2522{
02ac597c 2523 tid_t target;
c87591b7 2524
785c4bcc 2525 trace_ext3_sync_fs(sb, wait);
02ac597c
JK		 2526 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2527 if (wait)
2528 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2529 }
1da177e4
LT		 2530 return 0;
2531}
2532
2533/*
2534 * LVM calls this function before a (read-only) snapshot is created. This
2535 * gives us a chance to flush the journal completely and mark the fs clean.
2536 */
c4be0c1d 2537static int ext3_freeze(struct super_block *sb)
1da177e4 2538{
c4be0c1d
TS		 2539 int error = 0;
2540 journal_t *journal;
1da177e4
LT		 2541
2542 if (!(sb->s_flags & MS_RDONLY)) {
c4be0c1d 2543 journal = EXT3_SB(sb)->s_journal;
1da177e4
LT		 2544
2545 /* Now we set up the journal barrier. */
2546 journal_lock_updates(journal);
2d7c820e
HK		 2547
2548 /*
2549 * We don't want to clear needs_recovery flag when we failed
2550 * to flush the journal.
2551 */
c4be0c1d
TS		 2552 error = journal_flush(journal);
2553 if (error < 0)
2554 goto out;
1da177e4
LT		 2555
2556 /* Journal blocked and flushed, clear needs_recovery flag. */
2557 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
c4be0c1d
TS		 2558 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2559 if (error)
2560 goto out;
1da177e4 2561 }
c4be0c1d
TS		 2562 return 0;
2563
2564out:
2565 journal_unlock_updates(journal);
2566 return error;
1da177e4
LT		 2567}
2568
2569/*
2570 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2571 * flag here, even though the filesystem is not technically dirty yet.
2572 */
c4be0c1d 2573static int ext3_unfreeze(struct super_block *sb)
1da177e4
LT		 2574{
2575 if (!(sb->s_flags & MS_RDONLY)) {
2576 lock_super(sb);
2577 /* Reser the needs_recovery flag before the fs is unlocked. */
2578 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2579 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2580 unlock_super(sb);
2581 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2582 }
c4be0c1d 2583 return 0;
1da177e4
LT		 2584}
2585
2586static int ext3_remount (struct super_block * sb, int * flags, char * data)
2587{
2588 struct ext3_super_block * es;
2589 struct ext3_sb_info *sbi = EXT3_SB(sb);
43d23f90 2590 ext3_fsblk_t n_blocks_count = 0;
08c6a96f
JK		 2591 unsigned long old_sb_flags;
2592 struct ext3_mount_options old_opts;
c79d967d 2593 int enable_quota = 0;
08c6a96f
JK		 2594 int err;
2595#ifdef CONFIG_QUOTA
2596 int i;
2597#endif
2598
2599 /* Store the original options */
bbd6851a 2600 lock_super(sb);
08c6a96f
JK		 2601 old_sb_flags = sb->s_flags;
2602 old_opts.s_mount_opt = sbi->s_mount_opt;
2603 old_opts.s_resuid = sbi->s_resuid;
2604 old_opts.s_resgid = sbi->s_resgid;
2605 old_opts.s_commit_interval = sbi->s_commit_interval;
2606#ifdef CONFIG_QUOTA
2607 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2608 for (i = 0; i < MAXQUOTAS; i++)
2609 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2610#endif
1da177e4
LT		 2611
2612 /*
2613 * Allow the "check" option to be passed as a remount option.
2614 */
71b96257 2615 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
08c6a96f
JK		 2616 err = -EINVAL;
2617 goto restore_opts;
2618 }
1da177e4 2619
e3c96435 2620 if (test_opt(sb, ABORT))
e05b6b52 2621 ext3_abort(sb, __func__, "Abort forced by user");
1da177e4
LT		 2622
2623 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
e3c96435 2624 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1da177e4
LT		 2625
2626 es = sbi->s_es;
2627
2628 ext3_init_journal_params(sb, sbi->s_journal);
2629
2630 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2631 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
e3c96435 2632 if (test_opt(sb, ABORT)) {
08c6a96f
JK		 2633 err = -EROFS;
2634 goto restore_opts;
2635 }
1da177e4
LT		 2636
2637 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 2638 err = dquot_suspend(sb, -1);
2639 if (err < 0)
c79d967d 2640 goto restore_opts;
c79d967d 2641
1da177e4
LT		 2642 /*
2643 * First of all, the unconditional stuff we have to do
2644 * to disable replay of the journal when we next remount
2645 */
2646 sb->s_flags |= MS_RDONLY;
2647
2648 /*
2649 * OK, test if we are remounting a valid rw partition
2650 * readonly, and if so set the rdonly flag and then
2651 * mark the partition as valid again.
2652 */
2653 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2654 (sbi->s_mount_state & EXT3_VALID_FS))
2655 es->s_state = cpu_to_le16(sbi->s_mount_state);
2656
2657 ext3_mark_recovery_complete(sb, es);
2658 } else {
2659 __le32 ret;
2660 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2661 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
4cf46b67
AF		 2662 ext3_msg(sb, KERN_WARNING,
2663 "warning: couldn't remount RDWR "
2664 "because of unsupported optional "
2665 "features (%x)", le32_to_cpu(ret));
08c6a96f
JK		 2666 err = -EROFS;
2667 goto restore_opts;
1da177e4 2668 }
ea9a05a1
ES		 2669
2670 /*
2671 * If we have an unprocessed orphan list hanging
2672 * around from a previously readonly bdev mount,
1cde201d 2673 * require a full umount & mount for now.
ea9a05a1
ES		 2674 */
2675 if (es->s_last_orphan) {
4cf46b67 2676 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
ea9a05a1
ES		 2677 "remount RDWR because of unprocessed "
2678 "orphan inode list. Please "
1cde201d 2679 "umount & mount instead.");
ea9a05a1
ES		 2680 err = -EINVAL;
2681 goto restore_opts;
2682 }
2683
1da177e4
LT		 2684 /*
2685 * Mounting a RDONLY partition read-write, so reread
2686 * and store the current valid flag. (It may have
2687 * been changed by e2fsck since we originally mounted
2688 * the partition.)
2689 */
2690 ext3_clear_journal_err(sb, es);
2691 sbi->s_mount_state = le16_to_cpu(es->s_state);
a4e4de36 2692 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
08c6a96f 2693 goto restore_opts;
1da177e4
LT		 2694 if (!ext3_setup_super (sb, es, 0))
2695 sb->s_flags &= ~MS_RDONLY;
c79d967d 2696 enable_quota = 1;
1da177e4
LT		 2697 }
2698 }
08c6a96f
JK		 2699#ifdef CONFIG_QUOTA
2700 /* Release old quota file names */
2701 for (i = 0; i < MAXQUOTAS; i++)
2702 if (old_opts.s_qf_names[i] &&
2703 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2704 kfree(old_opts.s_qf_names[i]);
2705#endif
bbd6851a 2706 unlock_super(sb);
c79d967d
CH		 2707
2708 if (enable_quota)
0f0dd62f 2709 dquot_resume(sb, -1);
1da177e4 2710 return 0;
08c6a96f
JK		 2711restore_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
bbd6851a 2726 unlock_super(sb);
08c6a96f 2727 return err;
1da177e4
LT		 2728}
2729
726c3342 2730static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
1da177e4 2731{
726c3342 2732 struct super_block *sb = dentry->d_sb;
09fe316a
AT		 2733 struct ext3_sb_info *sbi = EXT3_SB(sb);
2734 struct ext3_super_block *es = sbi->s_es;
50ee0a32 2735 u64 fsid;
1da177e4 2736
a71ce8c6
BP		 2737 if (test_opt(sb, MINIX_DF)) {
2738 sbi->s_overhead_last = 0;
2739 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2740 unsigned long ngroups = sbi->s_groups_count, i;
2741 ext3_fsblk_t overhead = 0;
1da177e4
LT		 2742 smp_rmb();
2743
2744 /*
a71ce8c6
BP		 2745 * Compute the overhead (FS structures). This is constant
2746 * for a given filesystem unless the number of block groups
2747 * changes so we cache the previous value until it does.
1da177e4
LT		 2748 */
2749
2750 /*
2751 * All of the blocks before first_data_block are
2752 * overhead
2753 */
2754 overhead = le32_to_cpu(es->s_first_data_block);
2755
2756 /*
2757 * Add the overhead attributed to the superblock and
2758 * block group descriptors. If the sparse superblocks
2759 * feature is turned on, then not all groups have this.
2760 */
2761 for (i = 0; i < ngroups; i++) {
2762 overhead += ext3_bg_has_super(sb, i) +
2763 ext3_bg_num_gdb(sb, i);
2764 cond_resched();
2765 }
2766
2767 /*
2768 * Every block group has an inode bitmap, a block
2769 * bitmap, and an inode table.
2770 */
a71ce8c6
BP		 2771 overhead += ngroups * (2 + sbi->s_itb_per_group);
2772 sbi->s_overhead_last = overhead;
2773 smp_wmb();
2774 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1da177e4
LT		 2775 }
2776
2777 buf->f_type = EXT3_SUPER_MAGIC;
2778 buf->f_bsize = sb->s_blocksize;
a71ce8c6 2779 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
52d9f3b4 2780 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
1da177e4
LT		 2781 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2782 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2783 buf->f_bavail = 0;
2784 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 2785 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
1da177e4 2786 buf->f_namelen = EXT3_NAME_LEN;
50ee0a32
PE		 2787 fsid = le64_to_cpup((void *)es->s_uuid) ^
2788 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2789 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2790 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1da177e4
LT		 2791 return 0;
2792}
2793
2794/* Helper function for writing quotas on sync - we need to start transaction before quota file
2795 * is locked for write. Otherwise the are possible deadlocks:
2796 * Process 1 Process 2
2797 * ext3_create() quota_sync()
2798 * journal_start() write_dquot()
871a2931 2799 * dquot_initialize() down(dqio_mutex)
d3be915f 2800 * down(dqio_mutex) journal_start()
1da177e4
LT		 2801 *
2802 */
2803
2804#ifdef CONFIG_QUOTA
2805
2806static inline struct inode *dquot_to_inode(struct dquot *dquot)
2807{
2808 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2809}
2810
1da177e4
LT		 2811static int ext3_write_dquot(struct dquot *dquot)
2812{
2813 int ret, err;
2814 handle_t *handle;
2815 struct inode *inode;
2816
2817 inode = dquot_to_inode(dquot);
2818 handle = ext3_journal_start(inode,
1f54587b 2819 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2820 if (IS_ERR(handle))
2821 return PTR_ERR(handle);
2822 ret = dquot_commit(dquot);
2823 err = ext3_journal_stop(handle);
2824 if (!ret)
2825 ret = err;
2826 return ret;
2827}
2828
2829static int ext3_acquire_dquot(struct dquot *dquot)
2830{
2831 int ret, err;
2832 handle_t *handle;
2833
2834 handle = ext3_journal_start(dquot_to_inode(dquot),
1f54587b 2835 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
1da177e4
LT		 2836 if (IS_ERR(handle))
2837 return PTR_ERR(handle);
2838 ret = dquot_acquire(dquot);
2839 err = ext3_journal_stop(handle);
2840 if (!ret)
2841 ret = err;
2842 return ret;
2843}
2844
2845static int ext3_release_dquot(struct dquot *dquot)
2846{
2847 int ret, err;
2848 handle_t *handle;
2849
2850 handle = ext3_journal_start(dquot_to_inode(dquot),
1f54587b 2851 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 2852 if (IS_ERR(handle)) {
2853 /* Release dquot anyway to avoid endless cycle in dqput() */
2854 dquot_release(dquot);
1da177e4 2855 return PTR_ERR(handle);
9c3013e9 2856 }
1da177e4
LT		 2857 ret = dquot_release(dquot);
2858 err = ext3_journal_stop(handle);
2859 if (!ret)
2860 ret = err;
2861 return ret;
2862}
2863
2864static int ext3_mark_dquot_dirty(struct dquot *dquot)
2865{
99aeaf63 2866 /* Are we journaling quotas? */
1da177e4
LT		 2867 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2868 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2869 dquot_mark_dquot_dirty(dquot);
2870 return ext3_write_dquot(dquot);
2871 } else {
2872 return dquot_mark_dquot_dirty(dquot);
2873 }
2874}
2875
2876static int ext3_write_info(struct super_block *sb, int type)
2877{
2878 int ret, err;
2879 handle_t *handle;
2880
2881 /* Data block + inode block */
2882 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2883 if (IS_ERR(handle))
2884 return PTR_ERR(handle);
2885 ret = dquot_commit_info(sb, type);
2886 err = ext3_journal_stop(handle);
2887 if (!ret)
2888 ret = err;
2889 return ret;
2890}
2891
2892/*
2893 * Turn on quotas during mount time - we need to find
2894 * the quota file and such...
2895 */
2896static int ext3_quota_on_mount(struct super_block *sb, int type)
2897{
287a8095
CH		 2898 return dquot_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2899 EXT3_SB(sb)->s_jquota_fmt, type);
1da177e4
LT		 2900}
2901
2902/*
2903 * Standard function to be called on quota_on
2904 */
2905static int ext3_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 2906 struct path *path)
1da177e4
LT		 2907{
2908 int err;
1da177e4 2909
1f54587b
JK		 2910 if (!test_opt(sb, QUOTA))
2911 return -EINVAL;
9cfe7b90 2912
1da177e4 2913 /* Quotafile not on the same filesystem? */
f00c9e44 2914 if (path->mnt->mnt_sb != sb)
1da177e4 2915 return -EXDEV;
9cfe7b90
JK		 2916 /* Journaling quota? */
2917 if (EXT3_SB(sb)->s_qf_names[type]) {
2918 /* Quotafile not of fs root? */
f00c9e44 2919 if (path->dentry->d_parent != sb->s_root)
4cf46b67
AF		 2920 ext3_msg(sb, KERN_WARNING,
2921 "warning: Quota file not on filesystem root. "
2922 "Journaled quota will not work.");
9cfe7b90
JK		 2923 }
2924
2925 /*
2926 * When we journal data on quota file, we have to flush journal to see
2927 * all updates to the file when we bypass pagecache...
2928 */
f00c9e44 2929 if (ext3_should_journal_data(path->dentry->d_inode)) {
9cfe7b90
JK		 2930 /*
2931 * We don't need to lock updates but journal_flush() could
2932 * otherwise be livelocked...
2933 */
2934 journal_lock_updates(EXT3_SB(sb)->s_journal);
2d7c820e 2935 err = journal_flush(EXT3_SB(sb)->s_journal);
9cfe7b90 2936 journal_unlock_updates(EXT3_SB(sb)->s_journal);
f00c9e44 2937 if (err)
2d7c820e 2938 return err;
9cfe7b90
JK		 2939 }
2940
f00c9e44 2941 return dquot_quota_on(sb, type, format_id, path);
1da177e4
LT		 2942}
2943
2944/* Read data from quotafile - avoid pagecache and such because we cannot afford
2945 * acquiring the locks... As quota files are never truncated and quota code
25985edc 2946 * itself serializes the operations (and no one else should touch the files)
1da177e4
LT		 2947 * we don't have to be afraid of races */
2948static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2949 size_t len, loff_t off)
2950{
2951 struct inode *inode = sb_dqopt(sb)->files[type];
2952 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2953 int err = 0;
2954 int offset = off & (sb->s_blocksize - 1);
2955 int tocopy;
2956 size_t toread;
2957 struct buffer_head *bh;
2958 loff_t i_size = i_size_read(inode);
2959
2960 if (off > i_size)
2961 return 0;
2962 if (off+len > i_size)
2963 len = i_size-off;
2964 toread = len;
2965 while (toread > 0) {
2966 tocopy = sb->s_blocksize - offset < toread ?
2967 sb->s_blocksize - offset : toread;
2968 bh = ext3_bread(NULL, inode, blk, 0, &err);
2969 if (err)
2970 return err;
2971 if (!bh) /* A hole? */
2972 memset(data, 0, tocopy);
2973 else
2974 memcpy(data, bh->b_data+offset, tocopy);
2975 brelse(bh);
2976 offset = 0;
2977 toread -= tocopy;
2978 data += tocopy;
2979 blk++;
2980 }
2981 return len;
2982}
2983
2984/* Write to quotafile (we know the transaction is already started and has
2985 * enough credits) */
2986static ssize_t ext3_quota_write(struct super_block *sb, int type,
2987 const char *data, size_t len, loff_t off)
2988{
2989 struct inode *inode = sb_dqopt(sb)->files[type];
2990 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2991 int err = 0;
2992 int offset = off & (sb->s_blocksize - 1);
1da177e4 2993 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
1da177e4
LT		 2994 struct buffer_head *bh;
2995 handle_t *handle = journal_current_handle();
2996
9c3013e9 2997 if (!handle) {
4cf46b67
AF		 2998 ext3_msg(sb, KERN_WARNING,
2999 "warning: quota write (off=%llu, len=%llu)"
3000 " cancelled because transaction is not started.",
9c3013e9
JK		 3001 (unsigned long long)off, (unsigned long long)len);
3002 return -EIO;
3003 }
e5472147
DM		 3004
3005 /*
3006 * Since we account only one data block in transaction credits,
3007 * then it is impossible to cross a block boundary.
3008 */
3009 if (sb->s_blocksize - offset < len) {
3010 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3011 " cancelled because not block aligned",
3012 (unsigned long long)off, (unsigned long long)len);
3013 return -EIO;
3014 }
5c81a419 3015 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
e5472147
DM		 3016 bh = ext3_bread(handle, inode, blk, 1, &err);
3017 if (!bh)
3018 goto out;
3019 if (journal_quota) {
3020 err = ext3_journal_get_write_access(handle, bh);
3021 if (err) {
3022 brelse(bh);
1da177e4 3023 goto out;
1da177e4 3024 }
1da177e4 3025 }
e5472147
DM		 3026 lock_buffer(bh);
3027 memcpy(bh->b_data+offset, data, len);
3028 flush_dcache_page(bh->b_page);
3029 unlock_buffer(bh);
3030 if (journal_quota)
3031 err = ext3_journal_dirty_metadata(handle, bh);
3032 else {
3033 /* Always do at least ordered writes for quotas */
3034 err = ext3_journal_dirty_data(handle, bh);
3035 mark_buffer_dirty(bh);
3036 }
3037 brelse(bh);
1da177e4 3038out:
e5472147 3039 if (err) {
f5c8f7da 3040 mutex_unlock(&inode->i_mutex);
1da177e4 3041 return err;
f5c8f7da 3042 }
e5472147
DM		 3043 if (inode->i_size < off + len) {
3044 i_size_write(inode, off + len);
1da177e4
LT		 3045 EXT3_I(inode)->i_disksize = inode->i_size;
3046 }
3047 inode->i_version++;
3048 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3049 ext3_mark_inode_dirty(handle, inode);
1b1dcc1b 3050 mutex_unlock(&inode->i_mutex);
e5472147 3051 return len;
1da177e4
LT		 3052}
3053
3054#endif
3055
152a0836
AV		 3056static struct dentry *ext3_mount(struct file_system_type *fs_type,
3057 int flags, const char *dev_name, void *data)
1da177e4 3058{
152a0836 3059 return mount_bdev(fs_type, flags, dev_name, data, ext3_fill_super);
1da177e4
LT		 3060}
3061
3062static struct file_system_type ext3_fs_type = {
3063 .owner = THIS_MODULE,
3064 .name = "ext3",
152a0836 3065 .mount = ext3_mount,
1da177e4
LT		 3066 .kill_sb = kill_block_super,
3067 .fs_flags = FS_REQUIRES_DEV,
3068};
3069
3070static int __init init_ext3_fs(void)
3071{
3072 int err = init_ext3_xattr();
3073 if (err)
3074 return err;
3075 err = init_inodecache();
3076 if (err)
3077 goto out1;
3078 err = register_filesystem(&ext3_fs_type);
3079 if (err)
3080 goto out;
3081 return 0;
3082out:
3083 destroy_inodecache();
3084out1:
e9ad5620 3085 exit_ext3_xattr();
1da177e4
LT		 3086 return err;
3087}
3088
3089static void __exit exit_ext3_fs(void)
3090{
3091 unregister_filesystem(&ext3_fs_type);
3092 destroy_inodecache();
3093 exit_ext3_xattr();
3094}
3095
3096MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3097MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3098MODULE_LICENSE("GPL");
3099module_init(init_ext3_fs)
3100module_exit(exit_ext3_fs)
ubuntu-zesty-kernel mirror