?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ACPI: thinkpad-acpi: clean up hotkey_notify()
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext4 / super.c
blobe522b8c649d661dd22750ff1172f97cbdc742400
1 /*
2 * linux/fs/ext4/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
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/jbd2.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/smp_lock.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/marker.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <asm/uaccess.h>
42
43 #include "ext4.h"
44 #include "ext4_jbd2.h"
45 #include "xattr.h"
46 #include "acl.h"
47 #include "namei.h"
48 #include "group.h"
49
50 struct proc_dir_entry *ext4_proc_root;
51
52 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
53 unsigned long journal_devnum);
54 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
55 unsigned int);
56 static void ext4_commit_super(struct super_block *sb,
57 struct ext4_super_block *es, int sync);
58 static void ext4_mark_recovery_complete(struct super_block *sb,
59 struct ext4_super_block *es);
60 static void ext4_clear_journal_err(struct super_block *sb,
61 struct ext4_super_block *es);
62 static int ext4_sync_fs(struct super_block *sb, int wait);
63 static const char *ext4_decode_error(struct super_block *sb, int errno,
64 char nbuf[16]);
65 static int ext4_remount(struct super_block *sb, int *flags, char *data);
66 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
67 static void ext4_unlockfs(struct super_block *sb);
68 static void ext4_write_super(struct super_block *sb);
69 static void ext4_write_super_lockfs(struct super_block *sb);
70
71
72 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
73 struct ext4_group_desc *bg)
74 {
75 return le32_to_cpu(bg->bg_block_bitmap_lo) |
76 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
77 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
78 }
79
80 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
81 struct ext4_group_desc *bg)
82 {
83 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
84 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
85 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
86 }
87
88 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
89 struct ext4_group_desc *bg)
90 {
91 return le32_to_cpu(bg->bg_inode_table_lo) |
92 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
93 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
94 }
95
96 void ext4_block_bitmap_set(struct super_block *sb,
97 struct ext4_group_desc *bg, ext4_fsblk_t blk)
98 {
99 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
100 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
101 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
102 }
103
104 void ext4_inode_bitmap_set(struct super_block *sb,
105 struct ext4_group_desc *bg, ext4_fsblk_t blk)
106 {
107 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
108 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
109 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
110 }
111
112 void ext4_inode_table_set(struct super_block *sb,
113 struct ext4_group_desc *bg, ext4_fsblk_t blk)
114 {
115 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
116 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
117 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
118 }
119
120 /*
121 * Wrappers for jbd2_journal_start/end.
122 *
123 * The only special thing we need to do here is to make sure that all
124 * journal_end calls result in the superblock being marked dirty, so
125 * that sync() will call the filesystem's write_super callback if
126 * appropriate.
127 */
128 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
129 {
130 journal_t *journal;
131
132 if (sb->s_flags & MS_RDONLY)
133 return ERR_PTR(-EROFS);
134
135 /* Special case here: if the journal has aborted behind our
136 * backs (eg. EIO in the commit thread), then we still need to
137 * take the FS itself readonly cleanly. */
138 journal = EXT4_SB(sb)->s_journal;
139 if (is_journal_aborted(journal)) {
140 ext4_abort(sb, __func__,
141 "Detected aborted journal");
142 return ERR_PTR(-EROFS);
143 }
144
145 return jbd2_journal_start(journal, nblocks);
146 }
147
148 /*
149 * The only special thing we need to do here is to make sure that all
150 * jbd2_journal_stop calls result in the superblock being marked dirty, so
151 * that sync() will call the filesystem's write_super callback if
152 * appropriate.
153 */
154 int __ext4_journal_stop(const char *where, handle_t *handle)
155 {
156 struct super_block *sb;
157 int err;
158 int rc;
159
160 sb = handle->h_transaction->t_journal->j_private;
161 err = handle->h_err;
162 rc = jbd2_journal_stop(handle);
163
164 if (!err)
165 err = rc;
166 if (err)
167 __ext4_std_error(sb, where, err);
168 return err;
169 }
170
171 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
172 struct buffer_head *bh, handle_t *handle, int err)
173 {
174 char nbuf[16];
175 const char *errstr = ext4_decode_error(NULL, err, nbuf);
176
177 if (bh)
178 BUFFER_TRACE(bh, "abort");
179
180 if (!handle->h_err)
181 handle->h_err = err;
182
183 if (is_handle_aborted(handle))
184 return;
185
186 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
187 caller, errstr, err_fn);
188
189 jbd2_journal_abort_handle(handle);
190 }
191
192 /* Deal with the reporting of failure conditions on a filesystem such as
193 * inconsistencies detected or read IO failures.
194 *
195 * On ext2, we can store the error state of the filesystem in the
196 * superblock. That is not possible on ext4, because we may have other
197 * write ordering constraints on the superblock which prevent us from
198 * writing it out straight away; and given that the journal is about to
199 * be aborted, we can't rely on the current, or future, transactions to
200 * write out the superblock safely.
201 *
202 * We'll just use the jbd2_journal_abort() error code to record an error in
203 * the journal instead. On recovery, the journal will compain about
204 * that error until we've noted it down and cleared it.
205 */
206
207 static void ext4_handle_error(struct super_block *sb)
208 {
209 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
210
211 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
212 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
213
214 if (sb->s_flags & MS_RDONLY)
215 return;
216
217 if (!test_opt(sb, ERRORS_CONT)) {
218 journal_t *journal = EXT4_SB(sb)->s_journal;
219
220 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
221 if (journal)
222 jbd2_journal_abort(journal, -EIO);
223 }
224 if (test_opt(sb, ERRORS_RO)) {
225 printk(KERN_CRIT "Remounting filesystem read-only\n");
226 sb->s_flags |= MS_RDONLY;
227 }
228 ext4_commit_super(sb, es, 1);
229 if (test_opt(sb, ERRORS_PANIC))
230 panic("EXT4-fs (device %s): panic forced after error\n",
231 sb->s_id);
232 }
233
234 void ext4_error(struct super_block *sb, const char *function,
235 const char *fmt, ...)
236 {
237 va_list args;
238
239 va_start(args, fmt);
240 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
241 vprintk(fmt, args);
242 printk("\n");
243 va_end(args);
244
245 ext4_handle_error(sb);
246 }
247
248 static const char *ext4_decode_error(struct super_block *sb, int errno,
249 char nbuf[16])
250 {
251 char *errstr = NULL;
252
253 switch (errno) {
254 case -EIO:
255 errstr = "IO failure";
256 break;
257 case -ENOMEM:
258 errstr = "Out of memory";
259 break;
260 case -EROFS:
261 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
262 errstr = "Journal has aborted";
263 else
264 errstr = "Readonly filesystem";
265 break;
266 default:
267 /* If the caller passed in an extra buffer for unknown
268 * errors, textualise them now. Else we just return
269 * NULL. */
270 if (nbuf) {
271 /* Check for truncated error codes... */
272 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
273 errstr = nbuf;
274 }
275 break;
276 }
277
278 return errstr;
279 }
280
281 /* __ext4_std_error decodes expected errors from journaling functions
282 * automatically and invokes the appropriate error response. */
283
284 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
285 {
286 char nbuf[16];
287 const char *errstr;
288
289 /* Special case: if the error is EROFS, and we're not already
290 * inside a transaction, then there's really no point in logging
291 * an error. */
292 if (errno == -EROFS && journal_current_handle() == NULL &&
293 (sb->s_flags & MS_RDONLY))
294 return;
295
296 errstr = ext4_decode_error(sb, errno, nbuf);
297 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
298 sb->s_id, function, errstr);
299
300 ext4_handle_error(sb);
301 }
302
303 /*
304 * ext4_abort is a much stronger failure handler than ext4_error. The
305 * abort function may be used to deal with unrecoverable failures such
306 * as journal IO errors or ENOMEM at a critical moment in log management.
307 *
308 * We unconditionally force the filesystem into an ABORT|READONLY state,
309 * unless the error response on the fs has been set to panic in which
310 * case we take the easy way out and panic immediately.
311 */
312
313 void ext4_abort(struct super_block *sb, const char *function,
314 const char *fmt, ...)
315 {
316 va_list args;
317
318 printk(KERN_CRIT "ext4_abort called.\n");
319
320 va_start(args, fmt);
321 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
322 vprintk(fmt, args);
323 printk("\n");
324 va_end(args);
325
326 if (test_opt(sb, ERRORS_PANIC))
327 panic("EXT4-fs panic from previous error\n");
328
329 if (sb->s_flags & MS_RDONLY)
330 return;
331
332 printk(KERN_CRIT "Remounting filesystem read-only\n");
333 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
334 sb->s_flags |= MS_RDONLY;
335 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
336 if (EXT4_SB(sb)->s_journal)
337 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
338 }
339
340 void ext4_warning(struct super_block *sb, const char *function,
341 const char *fmt, ...)
342 {
343 va_list args;
344
345 va_start(args, fmt);
346 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
347 sb->s_id, function);
348 vprintk(fmt, args);
349 printk("\n");
350 va_end(args);
351 }
352
353 void ext4_update_dynamic_rev(struct super_block *sb)
354 {
355 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
356
357 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
358 return;
359
360 ext4_warning(sb, __func__,
361 "updating to rev %d because of new feature flag, "
362 "running e2fsck is recommended",
363 EXT4_DYNAMIC_REV);
364
365 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
366 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
367 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
368 /* leave es->s_feature_*compat flags alone */
369 /* es->s_uuid will be set by e2fsck if empty */
370
371 /*
372 * The rest of the superblock fields should be zero, and if not it
373 * means they are likely already in use, so leave them alone. We
374 * can leave it up to e2fsck to clean up any inconsistencies there.
375 */
376 }
377
378 /*
379 * Open the external journal device
380 */
381 static struct block_device *ext4_blkdev_get(dev_t dev)
382 {
383 struct block_device *bdev;
384 char b[BDEVNAME_SIZE];
385
386 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
387 if (IS_ERR(bdev))
388 goto fail;
389 return bdev;
390
391 fail:
392 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
393 __bdevname(dev, b), PTR_ERR(bdev));
394 return NULL;
395 }
396
397 /*
398 * Release the journal device
399 */
400 static int ext4_blkdev_put(struct block_device *bdev)
401 {
402 bd_release(bdev);
403 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
404 }
405
406 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
407 {
408 struct block_device *bdev;
409 int ret = -ENODEV;
410
411 bdev = sbi->journal_bdev;
412 if (bdev) {
413 ret = ext4_blkdev_put(bdev);
414 sbi->journal_bdev = NULL;
415 }
416 return ret;
417 }
418
419 static inline struct inode *orphan_list_entry(struct list_head *l)
420 {
421 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
422 }
423
424 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
425 {
426 struct list_head *l;
427
428 printk(KERN_ERR "sb orphan head is %d\n",
429 le32_to_cpu(sbi->s_es->s_last_orphan));
430
431 printk(KERN_ERR "sb_info orphan list:\n");
432 list_for_each(l, &sbi->s_orphan) {
433 struct inode *inode = orphan_list_entry(l);
434 printk(KERN_ERR " "
435 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
436 inode->i_sb->s_id, inode->i_ino, inode,
437 inode->i_mode, inode->i_nlink,
438 NEXT_ORPHAN(inode));
439 }
440 }
441
442 static void ext4_put_super(struct super_block *sb)
443 {
444 struct ext4_sb_info *sbi = EXT4_SB(sb);
445 struct ext4_super_block *es = sbi->s_es;
446 int i, err;
447
448 ext4_mb_release(sb);
449 ext4_ext_release(sb);
450 ext4_xattr_put_super(sb);
451 err = jbd2_journal_destroy(sbi->s_journal);
452 sbi->s_journal = NULL;
453 if (err < 0)
454 ext4_abort(sb, __func__, "Couldn't clean up the journal");
455
456 if (!(sb->s_flags & MS_RDONLY)) {
457 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
458 es->s_state = cpu_to_le16(sbi->s_mount_state);
459 ext4_commit_super(sb, es, 1);
460 }
461 if (sbi->s_proc) {
462 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
463 remove_proc_entry(sb->s_id, ext4_proc_root);
464 }
465
466 for (i = 0; i < sbi->s_gdb_count; i++)
467 brelse(sbi->s_group_desc[i]);
468 kfree(sbi->s_group_desc);
469 kfree(sbi->s_flex_groups);
470 percpu_counter_destroy(&sbi->s_freeblocks_counter);
471 percpu_counter_destroy(&sbi->s_freeinodes_counter);
472 percpu_counter_destroy(&sbi->s_dirs_counter);
473 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
474 brelse(sbi->s_sbh);
475 #ifdef CONFIG_QUOTA
476 for (i = 0; i < MAXQUOTAS; i++)
477 kfree(sbi->s_qf_names[i]);
478 #endif
479
480 /* Debugging code just in case the in-memory inode orphan list
481 * isn't empty. The on-disk one can be non-empty if we've
482 * detected an error and taken the fs readonly, but the
483 * in-memory list had better be clean by this point. */
484 if (!list_empty(&sbi->s_orphan))
485 dump_orphan_list(sb, sbi);
486 J_ASSERT(list_empty(&sbi->s_orphan));
487
488 invalidate_bdev(sb->s_bdev);
489 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
490 /*
491 * Invalidate the journal device's buffers. We don't want them
492 * floating about in memory - the physical journal device may
493 * hotswapped, and it breaks the `ro-after' testing code.
494 */
495 sync_blockdev(sbi->journal_bdev);
496 invalidate_bdev(sbi->journal_bdev);
497 ext4_blkdev_remove(sbi);
498 }
499 sb->s_fs_info = NULL;
500 kfree(sbi);
501 return;
502 }
503
504 static struct kmem_cache *ext4_inode_cachep;
505
506 /*
507 * Called inside transaction, so use GFP_NOFS
508 */
509 static struct inode *ext4_alloc_inode(struct super_block *sb)
510 {
511 struct ext4_inode_info *ei;
512
513 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
514 if (!ei)
515 return NULL;
516 #ifdef CONFIG_EXT4_FS_POSIX_ACL
517 ei->i_acl = EXT4_ACL_NOT_CACHED;
518 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
519 #endif
520 ei->vfs_inode.i_version = 1;
521 ei->vfs_inode.i_data.writeback_index = 0;
522 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
523 INIT_LIST_HEAD(&ei->i_prealloc_list);
524 spin_lock_init(&ei->i_prealloc_lock);
525 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
526 ei->i_reserved_data_blocks = 0;
527 ei->i_reserved_meta_blocks = 0;
528 ei->i_allocated_meta_blocks = 0;
529 ei->i_delalloc_reserved_flag = 0;
530 spin_lock_init(&(ei->i_block_reservation_lock));
531 return &ei->vfs_inode;
532 }
533
534 static void ext4_destroy_inode(struct inode *inode)
535 {
536 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
537 printk("EXT4 Inode %p: orphan list check failed!\n",
538 EXT4_I(inode));
539 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
540 EXT4_I(inode), sizeof(struct ext4_inode_info),
541 true);
542 dump_stack();
543 }
544 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
545 }
546
547 static void init_once(void *foo)
548 {
549 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
550
551 INIT_LIST_HEAD(&ei->i_orphan);
552 #ifdef CONFIG_EXT4_FS_XATTR
553 init_rwsem(&ei->xattr_sem);
554 #endif
555 init_rwsem(&ei->i_data_sem);
556 inode_init_once(&ei->vfs_inode);
557 }
558
559 static int init_inodecache(void)
560 {
561 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
562 sizeof(struct ext4_inode_info),
563 0, (SLAB_RECLAIM_ACCOUNT|
564 SLAB_MEM_SPREAD),
565 init_once);
566 if (ext4_inode_cachep == NULL)
567 return -ENOMEM;
568 return 0;
569 }
570
571 static void destroy_inodecache(void)
572 {
573 kmem_cache_destroy(ext4_inode_cachep);
574 }
575
576 static void ext4_clear_inode(struct inode *inode)
577 {
578 #ifdef CONFIG_EXT4_FS_POSIX_ACL
579 if (EXT4_I(inode)->i_acl &&
580 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
581 posix_acl_release(EXT4_I(inode)->i_acl);
582 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
583 }
584 if (EXT4_I(inode)->i_default_acl &&
585 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
586 posix_acl_release(EXT4_I(inode)->i_default_acl);
587 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
588 }
589 #endif
590 ext4_discard_preallocations(inode);
591 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
592 &EXT4_I(inode)->jinode);
593 }
594
595 static inline void ext4_show_quota_options(struct seq_file *seq,
596 struct super_block *sb)
597 {
598 #if defined(CONFIG_QUOTA)
599 struct ext4_sb_info *sbi = EXT4_SB(sb);
600
601 if (sbi->s_jquota_fmt)
602 seq_printf(seq, ",jqfmt=%s",
603 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
604
605 if (sbi->s_qf_names[USRQUOTA])
606 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
607
608 if (sbi->s_qf_names[GRPQUOTA])
609 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
610
611 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
612 seq_puts(seq, ",usrquota");
613
614 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
615 seq_puts(seq, ",grpquota");
616 #endif
617 }
618
619 /*
620 * Show an option if
621 * - it's set to a non-default value OR
622 * - if the per-sb default is different from the global default
623 */
624 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
625 {
626 int def_errors;
627 unsigned long def_mount_opts;
628 struct super_block *sb = vfs->mnt_sb;
629 struct ext4_sb_info *sbi = EXT4_SB(sb);
630 struct ext4_super_block *es = sbi->s_es;
631
632 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
633 def_errors = le16_to_cpu(es->s_errors);
634
635 if (sbi->s_sb_block != 1)
636 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
637 if (test_opt(sb, MINIX_DF))
638 seq_puts(seq, ",minixdf");
639 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
640 seq_puts(seq, ",grpid");
641 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
642 seq_puts(seq, ",nogrpid");
643 if (sbi->s_resuid != EXT4_DEF_RESUID ||
644 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
645 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
646 }
647 if (sbi->s_resgid != EXT4_DEF_RESGID ||
648 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
649 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
650 }
651 if (test_opt(sb, ERRORS_RO)) {
652 if (def_errors == EXT4_ERRORS_PANIC ||
653 def_errors == EXT4_ERRORS_CONTINUE) {
654 seq_puts(seq, ",errors=remount-ro");
655 }
656 }
657 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
658 seq_puts(seq, ",errors=continue");
659 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
660 seq_puts(seq, ",errors=panic");
661 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
662 seq_puts(seq, ",nouid32");
663 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
664 seq_puts(seq, ",debug");
665 if (test_opt(sb, OLDALLOC))
666 seq_puts(seq, ",oldalloc");
667 #ifdef CONFIG_EXT4_FS_XATTR
668 if (test_opt(sb, XATTR_USER) &&
669 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
670 seq_puts(seq, ",user_xattr");
671 if (!test_opt(sb, XATTR_USER) &&
672 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
673 seq_puts(seq, ",nouser_xattr");
674 }
675 #endif
676 #ifdef CONFIG_EXT4_FS_POSIX_ACL
677 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
678 seq_puts(seq, ",acl");
679 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
680 seq_puts(seq, ",noacl");
681 #endif
682 if (!test_opt(sb, RESERVATION))
683 seq_puts(seq, ",noreservation");
684 if (sbi->s_commit_interval) {
685 seq_printf(seq, ",commit=%u",
686 (unsigned) (sbi->s_commit_interval / HZ));
687 }
688 /*
689 * We're changing the default of barrier mount option, so
690 * let's always display its mount state so it's clear what its
691 * status is.
692 */
693 seq_puts(seq, ",barrier=");
694 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
695 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
696 seq_puts(seq, ",journal_async_commit");
697 if (test_opt(sb, NOBH))
698 seq_puts(seq, ",nobh");
699 if (!test_opt(sb, EXTENTS))
700 seq_puts(seq, ",noextents");
701 if (test_opt(sb, I_VERSION))
702 seq_puts(seq, ",i_version");
703 if (!test_opt(sb, DELALLOC))
704 seq_puts(seq, ",nodelalloc");
705
706
707 if (sbi->s_stripe)
708 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
709 /*
710 * journal mode get enabled in different ways
711 * So just print the value even if we didn't specify it
712 */
713 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
714 seq_puts(seq, ",data=journal");
715 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
716 seq_puts(seq, ",data=ordered");
717 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
718 seq_puts(seq, ",data=writeback");
719
720 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
721 seq_printf(seq, ",inode_readahead_blks=%u",
722 sbi->s_inode_readahead_blks);
723
724 if (test_opt(sb, DATA_ERR_ABORT))
725 seq_puts(seq, ",data_err=abort");
726
727 ext4_show_quota_options(seq, sb);
728 return 0;
729 }
730
731
732 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
733 u64 ino, u32 generation)
734 {
735 struct inode *inode;
736
737 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
738 return ERR_PTR(-ESTALE);
739 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
740 return ERR_PTR(-ESTALE);
741
742 /* iget isn't really right if the inode is currently unallocated!!
743 *
744 * ext4_read_inode will return a bad_inode if the inode had been
745 * deleted, so we should be safe.
746 *
747 * Currently we don't know the generation for parent directory, so
748 * a generation of 0 means "accept any"
749 */
750 inode = ext4_iget(sb, ino);
751 if (IS_ERR(inode))
752 return ERR_CAST(inode);
753 if (generation && inode->i_generation != generation) {
754 iput(inode);
755 return ERR_PTR(-ESTALE);
756 }
757
758 return inode;
759 }
760
761 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
762 int fh_len, int fh_type)
763 {
764 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
765 ext4_nfs_get_inode);
766 }
767
768 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
769 int fh_len, int fh_type)
770 {
771 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
772 ext4_nfs_get_inode);
773 }
774
775 #ifdef CONFIG_QUOTA
776 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
777 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
778
779 static int ext4_dquot_initialize(struct inode *inode, int type);
780 static int ext4_dquot_drop(struct inode *inode);
781 static int ext4_write_dquot(struct dquot *dquot);
782 static int ext4_acquire_dquot(struct dquot *dquot);
783 static int ext4_release_dquot(struct dquot *dquot);
784 static int ext4_mark_dquot_dirty(struct dquot *dquot);
785 static int ext4_write_info(struct super_block *sb, int type);
786 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
787 char *path, int remount);
788 static int ext4_quota_on_mount(struct super_block *sb, int type);
789 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
790 size_t len, loff_t off);
791 static ssize_t ext4_quota_write(struct super_block *sb, int type,
792 const char *data, size_t len, loff_t off);
793
794 static struct dquot_operations ext4_quota_operations = {
795 .initialize = ext4_dquot_initialize,
796 .drop = ext4_dquot_drop,
797 .alloc_space = dquot_alloc_space,
798 .alloc_inode = dquot_alloc_inode,
799 .free_space = dquot_free_space,
800 .free_inode = dquot_free_inode,
801 .transfer = dquot_transfer,
802 .write_dquot = ext4_write_dquot,
803 .acquire_dquot = ext4_acquire_dquot,
804 .release_dquot = ext4_release_dquot,
805 .mark_dirty = ext4_mark_dquot_dirty,
806 .write_info = ext4_write_info
807 };
808
809 static struct quotactl_ops ext4_qctl_operations = {
810 .quota_on = ext4_quota_on,
811 .quota_off = vfs_quota_off,
812 .quota_sync = vfs_quota_sync,
813 .get_info = vfs_get_dqinfo,
814 .set_info = vfs_set_dqinfo,
815 .get_dqblk = vfs_get_dqblk,
816 .set_dqblk = vfs_set_dqblk
817 };
818 #endif
819
820 static const struct super_operations ext4_sops = {
821 .alloc_inode = ext4_alloc_inode,
822 .destroy_inode = ext4_destroy_inode,
823 .write_inode = ext4_write_inode,
824 .dirty_inode = ext4_dirty_inode,
825 .delete_inode = ext4_delete_inode,
826 .put_super = ext4_put_super,
827 .write_super = ext4_write_super,
828 .sync_fs = ext4_sync_fs,
829 .write_super_lockfs = ext4_write_super_lockfs,
830 .unlockfs = ext4_unlockfs,
831 .statfs = ext4_statfs,
832 .remount_fs = ext4_remount,
833 .clear_inode = ext4_clear_inode,
834 .show_options = ext4_show_options,
835 #ifdef CONFIG_QUOTA
836 .quota_read = ext4_quota_read,
837 .quota_write = ext4_quota_write,
838 #endif
839 };
840
841 static const struct export_operations ext4_export_ops = {
842 .fh_to_dentry = ext4_fh_to_dentry,
843 .fh_to_parent = ext4_fh_to_parent,
844 .get_parent = ext4_get_parent,
845 };
846
847 enum {
848 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
849 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
850 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
851 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
852 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
853 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
854 Opt_journal_checksum, Opt_journal_async_commit,
855 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
856 Opt_data_err_abort, Opt_data_err_ignore,
857 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
858 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
859 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
860 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
861 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
862 Opt_inode_readahead_blks
863 };
864
865 static const match_table_t tokens = {
866 {Opt_bsd_df, "bsddf"},
867 {Opt_minix_df, "minixdf"},
868 {Opt_grpid, "grpid"},
869 {Opt_grpid, "bsdgroups"},
870 {Opt_nogrpid, "nogrpid"},
871 {Opt_nogrpid, "sysvgroups"},
872 {Opt_resgid, "resgid=%u"},
873 {Opt_resuid, "resuid=%u"},
874 {Opt_sb, "sb=%u"},
875 {Opt_err_cont, "errors=continue"},
876 {Opt_err_panic, "errors=panic"},
877 {Opt_err_ro, "errors=remount-ro"},
878 {Opt_nouid32, "nouid32"},
879 {Opt_debug, "debug"},
880 {Opt_oldalloc, "oldalloc"},
881 {Opt_orlov, "orlov"},
882 {Opt_user_xattr, "user_xattr"},
883 {Opt_nouser_xattr, "nouser_xattr"},
884 {Opt_acl, "acl"},
885 {Opt_noacl, "noacl"},
886 {Opt_reservation, "reservation"},
887 {Opt_noreservation, "noreservation"},
888 {Opt_noload, "noload"},
889 {Opt_nobh, "nobh"},
890 {Opt_bh, "bh"},
891 {Opt_commit, "commit=%u"},
892 {Opt_journal_update, "journal=update"},
893 {Opt_journal_inum, "journal=%u"},
894 {Opt_journal_dev, "journal_dev=%u"},
895 {Opt_journal_checksum, "journal_checksum"},
896 {Opt_journal_async_commit, "journal_async_commit"},
897 {Opt_abort, "abort"},
898 {Opt_data_journal, "data=journal"},
899 {Opt_data_ordered, "data=ordered"},
900 {Opt_data_writeback, "data=writeback"},
901 {Opt_data_err_abort, "data_err=abort"},
902 {Opt_data_err_ignore, "data_err=ignore"},
903 {Opt_offusrjquota, "usrjquota="},
904 {Opt_usrjquota, "usrjquota=%s"},
905 {Opt_offgrpjquota, "grpjquota="},
906 {Opt_grpjquota, "grpjquota=%s"},
907 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
908 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
909 {Opt_grpquota, "grpquota"},
910 {Opt_noquota, "noquota"},
911 {Opt_quota, "quota"},
912 {Opt_usrquota, "usrquota"},
913 {Opt_barrier, "barrier=%u"},
914 {Opt_extents, "extents"},
915 {Opt_noextents, "noextents"},
916 {Opt_i_version, "i_version"},
917 {Opt_stripe, "stripe=%u"},
918 {Opt_resize, "resize"},
919 {Opt_delalloc, "delalloc"},
920 {Opt_nodelalloc, "nodelalloc"},
921 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
922 {Opt_err, NULL},
923 };
924
925 static ext4_fsblk_t get_sb_block(void **data)
926 {
927 ext4_fsblk_t sb_block;
928 char *options = (char *) *data;
929
930 if (!options || strncmp(options, "sb=", 3) != 0)
931 return 1; /* Default location */
932 options += 3;
933 /*todo: use simple_strtoll with >32bit ext4 */
934 sb_block = simple_strtoul(options, &options, 0);
935 if (*options && *options != ',') {
936 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
937 (char *) *data);
938 return 1;
939 }
940 if (*options == ',')
941 options++;
942 *data = (void *) options;
943 return sb_block;
944 }
945
946 static int parse_options(char *options, struct super_block *sb,
947 unsigned int *inum, unsigned long *journal_devnum,
948 ext4_fsblk_t *n_blocks_count, int is_remount)
949 {
950 struct ext4_sb_info *sbi = EXT4_SB(sb);
951 char *p;
952 substring_t args[MAX_OPT_ARGS];
953 int data_opt = 0;
954 int option;
955 #ifdef CONFIG_QUOTA
956 int qtype, qfmt;
957 char *qname;
958 #endif
959 ext4_fsblk_t last_block;
960
961 if (!options)
962 return 1;
963
964 while ((p = strsep(&options, ",")) != NULL) {
965 int token;
966 if (!*p)
967 continue;
968
969 token = match_token(p, tokens, args);
970 switch (token) {
971 case Opt_bsd_df:
972 clear_opt(sbi->s_mount_opt, MINIX_DF);
973 break;
974 case Opt_minix_df:
975 set_opt(sbi->s_mount_opt, MINIX_DF);
976 break;
977 case Opt_grpid:
978 set_opt(sbi->s_mount_opt, GRPID);
979 break;
980 case Opt_nogrpid:
981 clear_opt(sbi->s_mount_opt, GRPID);
982 break;
983 case Opt_resuid:
984 if (match_int(&args[0], &option))
985 return 0;
986 sbi->s_resuid = option;
987 break;
988 case Opt_resgid:
989 if (match_int(&args[0], &option))
990 return 0;
991 sbi->s_resgid = option;
992 break;
993 case Opt_sb:
994 /* handled by get_sb_block() instead of here */
995 /* *sb_block = match_int(&args[0]); */
996 break;
997 case Opt_err_panic:
998 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
999 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1000 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1001 break;
1002 case Opt_err_ro:
1003 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1004 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1005 set_opt(sbi->s_mount_opt, ERRORS_RO);
1006 break;
1007 case Opt_err_cont:
1008 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1009 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1010 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1011 break;
1012 case Opt_nouid32:
1013 set_opt(sbi->s_mount_opt, NO_UID32);
1014 break;
1015 case Opt_debug:
1016 set_opt(sbi->s_mount_opt, DEBUG);
1017 break;
1018 case Opt_oldalloc:
1019 set_opt(sbi->s_mount_opt, OLDALLOC);
1020 break;
1021 case Opt_orlov:
1022 clear_opt(sbi->s_mount_opt, OLDALLOC);
1023 break;
1024 #ifdef CONFIG_EXT4_FS_XATTR
1025 case Opt_user_xattr:
1026 set_opt(sbi->s_mount_opt, XATTR_USER);
1027 break;
1028 case Opt_nouser_xattr:
1029 clear_opt(sbi->s_mount_opt, XATTR_USER);
1030 break;
1031 #else
1032 case Opt_user_xattr:
1033 case Opt_nouser_xattr:
1034 printk(KERN_ERR "EXT4 (no)user_xattr options "
1035 "not supported\n");
1036 break;
1037 #endif
1038 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1039 case Opt_acl:
1040 set_opt(sbi->s_mount_opt, POSIX_ACL);
1041 break;
1042 case Opt_noacl:
1043 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1044 break;
1045 #else
1046 case Opt_acl:
1047 case Opt_noacl:
1048 printk(KERN_ERR "EXT4 (no)acl options "
1049 "not supported\n");
1050 break;
1051 #endif
1052 case Opt_reservation:
1053 set_opt(sbi->s_mount_opt, RESERVATION);
1054 break;
1055 case Opt_noreservation:
1056 clear_opt(sbi->s_mount_opt, RESERVATION);
1057 break;
1058 case Opt_journal_update:
1059 /* @@@ FIXME */
1060 /* Eventually we will want to be able to create
1061 a journal file here. For now, only allow the
1062 user to specify an existing inode to be the
1063 journal file. */
1064 if (is_remount) {
1065 printk(KERN_ERR "EXT4-fs: cannot specify "
1066 "journal on remount\n");
1067 return 0;
1068 }
1069 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1070 break;
1071 case Opt_journal_inum:
1072 if (is_remount) {
1073 printk(KERN_ERR "EXT4-fs: cannot specify "
1074 "journal on remount\n");
1075 return 0;
1076 }
1077 if (match_int(&args[0], &option))
1078 return 0;
1079 *inum = option;
1080 break;
1081 case Opt_journal_dev:
1082 if (is_remount) {
1083 printk(KERN_ERR "EXT4-fs: cannot specify "
1084 "journal on remount\n");
1085 return 0;
1086 }
1087 if (match_int(&args[0], &option))
1088 return 0;
1089 *journal_devnum = option;
1090 break;
1091 case Opt_journal_checksum:
1092 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1093 break;
1094 case Opt_journal_async_commit:
1095 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1096 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1097 break;
1098 case Opt_noload:
1099 set_opt(sbi->s_mount_opt, NOLOAD);
1100 break;
1101 case Opt_commit:
1102 if (match_int(&args[0], &option))
1103 return 0;
1104 if (option < 0)
1105 return 0;
1106 if (option == 0)
1107 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1108 sbi->s_commit_interval = HZ * option;
1109 break;
1110 case Opt_data_journal:
1111 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1112 goto datacheck;
1113 case Opt_data_ordered:
1114 data_opt = EXT4_MOUNT_ORDERED_DATA;
1115 goto datacheck;
1116 case Opt_data_writeback:
1117 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1118 datacheck:
1119 if (is_remount) {
1120 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1121 != data_opt) {
1122 printk(KERN_ERR
1123 "EXT4-fs: cannot change data "
1124 "mode on remount\n");
1125 return 0;
1126 }
1127 } else {
1128 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1129 sbi->s_mount_opt |= data_opt;
1130 }
1131 break;
1132 case Opt_data_err_abort:
1133 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1134 break;
1135 case Opt_data_err_ignore:
1136 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1137 break;
1138 #ifdef CONFIG_QUOTA
1139 case Opt_usrjquota:
1140 qtype = USRQUOTA;
1141 goto set_qf_name;
1142 case Opt_grpjquota:
1143 qtype = GRPQUOTA;
1144 set_qf_name:
1145 if ((sb_any_quota_enabled(sb) ||
1146 sb_any_quota_suspended(sb)) &&
1147 !sbi->s_qf_names[qtype]) {
1148 printk(KERN_ERR
1149 "EXT4-fs: Cannot change journaled "
1150 "quota options when quota turned on.\n");
1151 return 0;
1152 }
1153 qname = match_strdup(&args[0]);
1154 if (!qname) {
1155 printk(KERN_ERR
1156 "EXT4-fs: not enough memory for "
1157 "storing quotafile name.\n");
1158 return 0;
1159 }
1160 if (sbi->s_qf_names[qtype] &&
1161 strcmp(sbi->s_qf_names[qtype], qname)) {
1162 printk(KERN_ERR
1163 "EXT4-fs: %s quota file already "
1164 "specified.\n", QTYPE2NAME(qtype));
1165 kfree(qname);
1166 return 0;
1167 }
1168 sbi->s_qf_names[qtype] = qname;
1169 if (strchr(sbi->s_qf_names[qtype], '/')) {
1170 printk(KERN_ERR
1171 "EXT4-fs: quotafile must be on "
1172 "filesystem root.\n");
1173 kfree(sbi->s_qf_names[qtype]);
1174 sbi->s_qf_names[qtype] = NULL;
1175 return 0;
1176 }
1177 set_opt(sbi->s_mount_opt, QUOTA);
1178 break;
1179 case Opt_offusrjquota:
1180 qtype = USRQUOTA;
1181 goto clear_qf_name;
1182 case Opt_offgrpjquota:
1183 qtype = GRPQUOTA;
1184 clear_qf_name:
1185 if ((sb_any_quota_enabled(sb) ||
1186 sb_any_quota_suspended(sb)) &&
1187 sbi->s_qf_names[qtype]) {
1188 printk(KERN_ERR "EXT4-fs: Cannot change "
1189 "journaled quota options when "
1190 "quota turned on.\n");
1191 return 0;
1192 }
1193 /*
1194 * The space will be released later when all options
1195 * are confirmed to be correct
1196 */
1197 sbi->s_qf_names[qtype] = NULL;
1198 break;
1199 case Opt_jqfmt_vfsold:
1200 qfmt = QFMT_VFS_OLD;
1201 goto set_qf_format;
1202 case Opt_jqfmt_vfsv0:
1203 qfmt = QFMT_VFS_V0;
1204 set_qf_format:
1205 if ((sb_any_quota_enabled(sb) ||
1206 sb_any_quota_suspended(sb)) &&
1207 sbi->s_jquota_fmt != qfmt) {
1208 printk(KERN_ERR "EXT4-fs: Cannot change "
1209 "journaled quota options when "
1210 "quota turned on.\n");
1211 return 0;
1212 }
1213 sbi->s_jquota_fmt = qfmt;
1214 break;
1215 case Opt_quota:
1216 case Opt_usrquota:
1217 set_opt(sbi->s_mount_opt, QUOTA);
1218 set_opt(sbi->s_mount_opt, USRQUOTA);
1219 break;
1220 case Opt_grpquota:
1221 set_opt(sbi->s_mount_opt, QUOTA);
1222 set_opt(sbi->s_mount_opt, GRPQUOTA);
1223 break;
1224 case Opt_noquota:
1225 if (sb_any_quota_enabled(sb)) {
1226 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1227 "options when quota turned on.\n");
1228 return 0;
1229 }
1230 clear_opt(sbi->s_mount_opt, QUOTA);
1231 clear_opt(sbi->s_mount_opt, USRQUOTA);
1232 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1233 break;
1234 #else
1235 case Opt_quota:
1236 case Opt_usrquota:
1237 case Opt_grpquota:
1238 printk(KERN_ERR
1239 "EXT4-fs: quota options not supported.\n");
1240 break;
1241 case Opt_usrjquota:
1242 case Opt_grpjquota:
1243 case Opt_offusrjquota:
1244 case Opt_offgrpjquota:
1245 case Opt_jqfmt_vfsold:
1246 case Opt_jqfmt_vfsv0:
1247 printk(KERN_ERR
1248 "EXT4-fs: journaled quota options not "
1249 "supported.\n");
1250 break;
1251 case Opt_noquota:
1252 break;
1253 #endif
1254 case Opt_abort:
1255 set_opt(sbi->s_mount_opt, ABORT);
1256 break;
1257 case Opt_barrier:
1258 if (match_int(&args[0], &option))
1259 return 0;
1260 if (option)
1261 set_opt(sbi->s_mount_opt, BARRIER);
1262 else
1263 clear_opt(sbi->s_mount_opt, BARRIER);
1264 break;
1265 case Opt_ignore:
1266 break;
1267 case Opt_resize:
1268 if (!is_remount) {
1269 printk("EXT4-fs: resize option only available "
1270 "for remount\n");
1271 return 0;
1272 }
1273 if (match_int(&args[0], &option) != 0)
1274 return 0;
1275 *n_blocks_count = option;
1276 break;
1277 case Opt_nobh:
1278 set_opt(sbi->s_mount_opt, NOBH);
1279 break;
1280 case Opt_bh:
1281 clear_opt(sbi->s_mount_opt, NOBH);
1282 break;
1283 case Opt_extents:
1284 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1285 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1286 ext4_warning(sb, __func__,
1287 "extents feature not enabled "
1288 "on this filesystem, use tune2fs\n");
1289 return 0;
1290 }
1291 set_opt(sbi->s_mount_opt, EXTENTS);
1292 break;
1293 case Opt_noextents:
1294 /*
1295 * When e2fsprogs support resizing an already existing
1296 * ext3 file system to greater than 2**32 we need to
1297 * add support to block allocator to handle growing
1298 * already existing block mapped inode so that blocks
1299 * allocated for them fall within 2**32
1300 */
1301 last_block = ext4_blocks_count(sbi->s_es) - 1;
1302 if (last_block > 0xffffffffULL) {
1303 printk(KERN_ERR "EXT4-fs: Filesystem too "
1304 "large to mount with "
1305 "-o noextents options\n");
1306 return 0;
1307 }
1308 clear_opt(sbi->s_mount_opt, EXTENTS);
1309 break;
1310 case Opt_i_version:
1311 set_opt(sbi->s_mount_opt, I_VERSION);
1312 sb->s_flags |= MS_I_VERSION;
1313 break;
1314 case Opt_nodelalloc:
1315 clear_opt(sbi->s_mount_opt, DELALLOC);
1316 break;
1317 case Opt_stripe:
1318 if (match_int(&args[0], &option))
1319 return 0;
1320 if (option < 0)
1321 return 0;
1322 sbi->s_stripe = option;
1323 break;
1324 case Opt_delalloc:
1325 set_opt(sbi->s_mount_opt, DELALLOC);
1326 break;
1327 case Opt_inode_readahead_blks:
1328 if (match_int(&args[0], &option))
1329 return 0;
1330 if (option < 0 || option > (1 << 30))
1331 return 0;
1332 sbi->s_inode_readahead_blks = option;
1333 break;
1334 default:
1335 printk(KERN_ERR
1336 "EXT4-fs: Unrecognized mount option \"%s\" "
1337 "or missing value\n", p);
1338 return 0;
1339 }
1340 }
1341 #ifdef CONFIG_QUOTA
1342 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1343 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1344 sbi->s_qf_names[USRQUOTA])
1345 clear_opt(sbi->s_mount_opt, USRQUOTA);
1346
1347 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1348 sbi->s_qf_names[GRPQUOTA])
1349 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1350
1351 if ((sbi->s_qf_names[USRQUOTA] &&
1352 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1353 (sbi->s_qf_names[GRPQUOTA] &&
1354 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1355 printk(KERN_ERR "EXT4-fs: old and new quota "
1356 "format mixing.\n");
1357 return 0;
1358 }
1359
1360 if (!sbi->s_jquota_fmt) {
1361 printk(KERN_ERR "EXT4-fs: journaled quota format "
1362 "not specified.\n");
1363 return 0;
1364 }
1365 } else {
1366 if (sbi->s_jquota_fmt) {
1367 printk(KERN_ERR "EXT4-fs: journaled quota format "
1368 "specified with no journaling "
1369 "enabled.\n");
1370 return 0;
1371 }
1372 }
1373 #endif
1374 return 1;
1375 }
1376
1377 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1378 int read_only)
1379 {
1380 struct ext4_sb_info *sbi = EXT4_SB(sb);
1381 int res = 0;
1382
1383 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1384 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1385 "forcing read-only mode\n");
1386 res = MS_RDONLY;
1387 }
1388 if (read_only)
1389 return res;
1390 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1391 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1392 "running e2fsck is recommended\n");
1393 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1394 printk(KERN_WARNING
1395 "EXT4-fs warning: mounting fs with errors, "
1396 "running e2fsck is recommended\n");
1397 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1398 le16_to_cpu(es->s_mnt_count) >=
1399 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1400 printk(KERN_WARNING
1401 "EXT4-fs warning: maximal mount count reached, "
1402 "running e2fsck is recommended\n");
1403 else if (le32_to_cpu(es->s_checkinterval) &&
1404 (le32_to_cpu(es->s_lastcheck) +
1405 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1406 printk(KERN_WARNING
1407 "EXT4-fs warning: checktime reached, "
1408 "running e2fsck is recommended\n");
1409 #if 0
1410 /* @@@ We _will_ want to clear the valid bit if we find
1411 * inconsistencies, to force a fsck at reboot. But for
1412 * a plain journaled filesystem we can keep it set as
1413 * valid forever! :)
1414 */
1415 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1416 #endif
1417 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1418 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1419 le16_add_cpu(&es->s_mnt_count, 1);
1420 es->s_mtime = cpu_to_le32(get_seconds());
1421 ext4_update_dynamic_rev(sb);
1422 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1423
1424 ext4_commit_super(sb, es, 1);
1425 if (test_opt(sb, DEBUG))
1426 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1427 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1428 sb->s_blocksize,
1429 sbi->s_groups_count,
1430 EXT4_BLOCKS_PER_GROUP(sb),
1431 EXT4_INODES_PER_GROUP(sb),
1432 sbi->s_mount_opt);
1433
1434 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1435 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1436 "external", EXT4_SB(sb)->s_journal->j_devname);
1437 return res;
1438 }
1439
1440 static int ext4_fill_flex_info(struct super_block *sb)
1441 {
1442 struct ext4_sb_info *sbi = EXT4_SB(sb);
1443 struct ext4_group_desc *gdp = NULL;
1444 struct buffer_head *bh;
1445 ext4_group_t flex_group_count;
1446 ext4_group_t flex_group;
1447 int groups_per_flex = 0;
1448 int i;
1449
1450 if (!sbi->s_es->s_log_groups_per_flex) {
1451 sbi->s_log_groups_per_flex = 0;
1452 return 1;
1453 }
1454
1455 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1456 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1457
1458 /* We allocate both existing and potentially added groups */
1459 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1460 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1461 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1462 sbi->s_flex_groups = kzalloc(flex_group_count *
1463 sizeof(struct flex_groups), GFP_KERNEL);
1464 if (sbi->s_flex_groups == NULL) {
1465 printk(KERN_ERR "EXT4-fs: not enough memory for "
1466 "%lu flex groups\n", flex_group_count);
1467 goto failed;
1468 }
1469
1470 for (i = 0; i < sbi->s_groups_count; i++) {
1471 gdp = ext4_get_group_desc(sb, i, &bh);
1472
1473 flex_group = ext4_flex_group(sbi, i);
1474 sbi->s_flex_groups[flex_group].free_inodes +=
1475 le16_to_cpu(gdp->bg_free_inodes_count);
1476 sbi->s_flex_groups[flex_group].free_blocks +=
1477 le16_to_cpu(gdp->bg_free_blocks_count);
1478 }
1479
1480 return 1;
1481 failed:
1482 return 0;
1483 }
1484
1485 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1486 struct ext4_group_desc *gdp)
1487 {
1488 __u16 crc = 0;
1489
1490 if (sbi->s_es->s_feature_ro_compat &
1491 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1492 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1493 __le32 le_group = cpu_to_le32(block_group);
1494
1495 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1496 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1497 crc = crc16(crc, (__u8 *)gdp, offset);
1498 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1499 /* for checksum of struct ext4_group_desc do the rest...*/
1500 if ((sbi->s_es->s_feature_incompat &
1501 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1502 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1503 crc = crc16(crc, (__u8 *)gdp + offset,
1504 le16_to_cpu(sbi->s_es->s_desc_size) -
1505 offset);
1506 }
1507
1508 return cpu_to_le16(crc);
1509 }
1510
1511 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1512 struct ext4_group_desc *gdp)
1513 {
1514 if ((sbi->s_es->s_feature_ro_compat &
1515 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1516 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1517 return 0;
1518
1519 return 1;
1520 }
1521
1522 /* Called at mount-time, super-block is locked */
1523 static int ext4_check_descriptors(struct super_block *sb)
1524 {
1525 struct ext4_sb_info *sbi = EXT4_SB(sb);
1526 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1527 ext4_fsblk_t last_block;
1528 ext4_fsblk_t block_bitmap;
1529 ext4_fsblk_t inode_bitmap;
1530 ext4_fsblk_t inode_table;
1531 int flexbg_flag = 0;
1532 ext4_group_t i;
1533
1534 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1535 flexbg_flag = 1;
1536
1537 ext4_debug("Checking group descriptors");
1538
1539 for (i = 0; i < sbi->s_groups_count; i++) {
1540 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1541
1542 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1543 last_block = ext4_blocks_count(sbi->s_es) - 1;
1544 else
1545 last_block = first_block +
1546 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1547
1548 block_bitmap = ext4_block_bitmap(sb, gdp);
1549 if (block_bitmap < first_block || block_bitmap > last_block) {
1550 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1551 "Block bitmap for group %lu not in group "
1552 "(block %llu)!\n", i, block_bitmap);
1553 return 0;
1554 }
1555 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1556 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1557 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1558 "Inode bitmap for group %lu not in group "
1559 "(block %llu)!\n", i, inode_bitmap);
1560 return 0;
1561 }
1562 inode_table = ext4_inode_table(sb, gdp);
1563 if (inode_table < first_block ||
1564 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1565 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1566 "Inode table for group %lu not in group "
1567 "(block %llu)!\n", i, inode_table);
1568 return 0;
1569 }
1570 spin_lock(sb_bgl_lock(sbi, i));
1571 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1572 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1573 "Checksum for group %lu failed (%u!=%u)\n",
1574 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1575 gdp)), le16_to_cpu(gdp->bg_checksum));
1576 if (!(sb->s_flags & MS_RDONLY)) {
1577 spin_unlock(sb_bgl_lock(sbi, i));
1578 return 0;
1579 }
1580 }
1581 spin_unlock(sb_bgl_lock(sbi, i));
1582 if (!flexbg_flag)
1583 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1584 }
1585
1586 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1587 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1588 return 1;
1589 }
1590
1591 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1592 * the superblock) which were deleted from all directories, but held open by
1593 * a process at the time of a crash. We walk the list and try to delete these
1594 * inodes at recovery time (only with a read-write filesystem).
1595 *
1596 * In order to keep the orphan inode chain consistent during traversal (in
1597 * case of crash during recovery), we link each inode into the superblock
1598 * orphan list_head and handle it the same way as an inode deletion during
1599 * normal operation (which journals the operations for us).
1600 *
1601 * We only do an iget() and an iput() on each inode, which is very safe if we
1602 * accidentally point at an in-use or already deleted inode. The worst that
1603 * can happen in this case is that we get a "bit already cleared" message from
1604 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1605 * e2fsck was run on this filesystem, and it must have already done the orphan
1606 * inode cleanup for us, so we can safely abort without any further action.
1607 */
1608 static void ext4_orphan_cleanup(struct super_block *sb,
1609 struct ext4_super_block *es)
1610 {
1611 unsigned int s_flags = sb->s_flags;
1612 int nr_orphans = 0, nr_truncates = 0;
1613 #ifdef CONFIG_QUOTA
1614 int i;
1615 #endif
1616 if (!es->s_last_orphan) {
1617 jbd_debug(4, "no orphan inodes to clean up\n");
1618 return;
1619 }
1620
1621 if (bdev_read_only(sb->s_bdev)) {
1622 printk(KERN_ERR "EXT4-fs: write access "
1623 "unavailable, skipping orphan cleanup.\n");
1624 return;
1625 }
1626
1627 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1628 if (es->s_last_orphan)
1629 jbd_debug(1, "Errors on filesystem, "
1630 "clearing orphan list.\n");
1631 es->s_last_orphan = 0;
1632 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1633 return;
1634 }
1635
1636 if (s_flags & MS_RDONLY) {
1637 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1638 sb->s_id);
1639 sb->s_flags &= ~MS_RDONLY;
1640 }
1641 #ifdef CONFIG_QUOTA
1642 /* Needed for iput() to work correctly and not trash data */
1643 sb->s_flags |= MS_ACTIVE;
1644 /* Turn on quotas so that they are updated correctly */
1645 for (i = 0; i < MAXQUOTAS; i++) {
1646 if (EXT4_SB(sb)->s_qf_names[i]) {
1647 int ret = ext4_quota_on_mount(sb, i);
1648 if (ret < 0)
1649 printk(KERN_ERR
1650 "EXT4-fs: Cannot turn on journaled "
1651 "quota: error %d\n", ret);
1652 }
1653 }
1654 #endif
1655
1656 while (es->s_last_orphan) {
1657 struct inode *inode;
1658
1659 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1660 if (IS_ERR(inode)) {
1661 es->s_last_orphan = 0;
1662 break;
1663 }
1664
1665 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1666 DQUOT_INIT(inode);
1667 if (inode->i_nlink) {
1668 printk(KERN_DEBUG
1669 "%s: truncating inode %lu to %lld bytes\n",
1670 __func__, inode->i_ino, inode->i_size);
1671 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1672 inode->i_ino, inode->i_size);
1673 ext4_truncate(inode);
1674 nr_truncates++;
1675 } else {
1676 printk(KERN_DEBUG
1677 "%s: deleting unreferenced inode %lu\n",
1678 __func__, inode->i_ino);
1679 jbd_debug(2, "deleting unreferenced inode %lu\n",
1680 inode->i_ino);
1681 nr_orphans++;
1682 }
1683 iput(inode); /* The delete magic happens here! */
1684 }
1685
1686 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1687
1688 if (nr_orphans)
1689 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1690 sb->s_id, PLURAL(nr_orphans));
1691 if (nr_truncates)
1692 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1693 sb->s_id, PLURAL(nr_truncates));
1694 #ifdef CONFIG_QUOTA
1695 /* Turn quotas off */
1696 for (i = 0; i < MAXQUOTAS; i++) {
1697 if (sb_dqopt(sb)->files[i])
1698 vfs_quota_off(sb, i, 0);
1699 }
1700 #endif
1701 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1702 }
1703 /*
1704 * Maximal extent format file size.
1705 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1706 * extent format containers, within a sector_t, and within i_blocks
1707 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1708 * so that won't be a limiting factor.
1709 *
1710 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1711 */
1712 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1713 {
1714 loff_t res;
1715 loff_t upper_limit = MAX_LFS_FILESIZE;
1716
1717 /* small i_blocks in vfs inode? */
1718 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1719 /*
1720 * CONFIG_LSF is not enabled implies the inode
1721 * i_block represent total blocks in 512 bytes
1722 * 32 == size of vfs inode i_blocks * 8
1723 */
1724 upper_limit = (1LL << 32) - 1;
1725
1726 /* total blocks in file system block size */
1727 upper_limit >>= (blkbits - 9);
1728 upper_limit <<= blkbits;
1729 }
1730
1731 /* 32-bit extent-start container, ee_block */
1732 res = 1LL << 32;
1733 res <<= blkbits;
1734 res -= 1;
1735
1736 /* Sanity check against vm- & vfs- imposed limits */
1737 if (res > upper_limit)
1738 res = upper_limit;
1739
1740 return res;
1741 }
1742
1743 /*
1744 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1745 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1746 * We need to be 1 filesystem block less than the 2^48 sector limit.
1747 */
1748 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1749 {
1750 loff_t res = EXT4_NDIR_BLOCKS;
1751 int meta_blocks;
1752 loff_t upper_limit;
1753 /* This is calculated to be the largest file size for a
1754 * dense, bitmapped file such that the total number of
1755 * sectors in the file, including data and all indirect blocks,
1756 * does not exceed 2^48 -1
1757 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1758 * total number of 512 bytes blocks of the file
1759 */
1760
1761 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1762 /*
1763 * !has_huge_files or CONFIG_LSF is not enabled
1764 * implies the inode i_block represent total blocks in
1765 * 512 bytes 32 == size of vfs inode i_blocks * 8
1766 */
1767 upper_limit = (1LL << 32) - 1;
1768
1769 /* total blocks in file system block size */
1770 upper_limit >>= (bits - 9);
1771
1772 } else {
1773 /*
1774 * We use 48 bit ext4_inode i_blocks
1775 * With EXT4_HUGE_FILE_FL set the i_blocks
1776 * represent total number of blocks in
1777 * file system block size
1778 */
1779 upper_limit = (1LL << 48) - 1;
1780
1781 }
1782
1783 /* indirect blocks */
1784 meta_blocks = 1;
1785 /* double indirect blocks */
1786 meta_blocks += 1 + (1LL << (bits-2));
1787 /* tripple indirect blocks */
1788 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1789
1790 upper_limit -= meta_blocks;
1791 upper_limit <<= bits;
1792
1793 res += 1LL << (bits-2);
1794 res += 1LL << (2*(bits-2));
1795 res += 1LL << (3*(bits-2));
1796 res <<= bits;
1797 if (res > upper_limit)
1798 res = upper_limit;
1799
1800 if (res > MAX_LFS_FILESIZE)
1801 res = MAX_LFS_FILESIZE;
1802
1803 return res;
1804 }
1805
1806 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1807 ext4_fsblk_t logical_sb_block, int nr)
1808 {
1809 struct ext4_sb_info *sbi = EXT4_SB(sb);
1810 ext4_group_t bg, first_meta_bg;
1811 int has_super = 0;
1812
1813 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1814
1815 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1816 nr < first_meta_bg)
1817 return logical_sb_block + nr + 1;
1818 bg = sbi->s_desc_per_block * nr;
1819 if (ext4_bg_has_super(sb, bg))
1820 has_super = 1;
1821 return (has_super + ext4_group_first_block_no(sb, bg));
1822 }
1823
1824 /**
1825 * ext4_get_stripe_size: Get the stripe size.
1826 * @sbi: In memory super block info
1827 *
1828 * If we have specified it via mount option, then
1829 * use the mount option value. If the value specified at mount time is
1830 * greater than the blocks per group use the super block value.
1831 * If the super block value is greater than blocks per group return 0.
1832 * Allocator needs it be less than blocks per group.
1833 *
1834 */
1835 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1836 {
1837 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1838 unsigned long stripe_width =
1839 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1840
1841 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1842 return sbi->s_stripe;
1843
1844 if (stripe_width <= sbi->s_blocks_per_group)
1845 return stripe_width;
1846
1847 if (stride <= sbi->s_blocks_per_group)
1848 return stride;
1849
1850 return 0;
1851 }
1852
1853 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
1854 __releases(kernel_lock)
1855 __acquires(kernel_lock)
1856
1857 {
1858 struct buffer_head *bh;
1859 struct ext4_super_block *es = NULL;
1860 struct ext4_sb_info *sbi;
1861 ext4_fsblk_t block;
1862 ext4_fsblk_t sb_block = get_sb_block(&data);
1863 ext4_fsblk_t logical_sb_block;
1864 unsigned long offset = 0;
1865 unsigned int journal_inum = 0;
1866 unsigned long journal_devnum = 0;
1867 unsigned long def_mount_opts;
1868 struct inode *root;
1869 char *cp;
1870 int ret = -EINVAL;
1871 int blocksize;
1872 unsigned int db_count;
1873 unsigned int i;
1874 int needs_recovery, has_huge_files;
1875 __le32 features;
1876 __u64 blocks_count;
1877 int err;
1878
1879 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1880 if (!sbi)
1881 return -ENOMEM;
1882 sb->s_fs_info = sbi;
1883 sbi->s_mount_opt = 0;
1884 sbi->s_resuid = EXT4_DEF_RESUID;
1885 sbi->s_resgid = EXT4_DEF_RESGID;
1886 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
1887 sbi->s_sb_block = sb_block;
1888
1889 unlock_kernel();
1890
1891 /* Cleanup superblock name */
1892 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
1893 *cp = '!';
1894
1895 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1896 if (!blocksize) {
1897 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1898 goto out_fail;
1899 }
1900
1901 /*
1902 * The ext4 superblock will not be buffer aligned for other than 1kB
1903 * block sizes. We need to calculate the offset from buffer start.
1904 */
1905 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1906 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1907 offset = do_div(logical_sb_block, blocksize);
1908 } else {
1909 logical_sb_block = sb_block;
1910 }
1911
1912 if (!(bh = sb_bread(sb, logical_sb_block))) {
1913 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
1914 goto out_fail;
1915 }
1916 /*
1917 * Note: s_es must be initialized as soon as possible because
1918 * some ext4 macro-instructions depend on its value
1919 */
1920 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1921 sbi->s_es = es;
1922 sb->s_magic = le16_to_cpu(es->s_magic);
1923 if (sb->s_magic != EXT4_SUPER_MAGIC)
1924 goto cantfind_ext4;
1925
1926 /* Set defaults before we parse the mount options */
1927 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1928 if (def_mount_opts & EXT4_DEFM_DEBUG)
1929 set_opt(sbi->s_mount_opt, DEBUG);
1930 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1931 set_opt(sbi->s_mount_opt, GRPID);
1932 if (def_mount_opts & EXT4_DEFM_UID16)
1933 set_opt(sbi->s_mount_opt, NO_UID32);
1934 #ifdef CONFIG_EXT4_FS_XATTR
1935 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1936 set_opt(sbi->s_mount_opt, XATTR_USER);
1937 #endif
1938 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1939 if (def_mount_opts & EXT4_DEFM_ACL)
1940 set_opt(sbi->s_mount_opt, POSIX_ACL);
1941 #endif
1942 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1943 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1944 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1945 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1946 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1947 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1948
1949 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1950 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1951 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1952 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1953 else
1954 set_opt(sbi->s_mount_opt, ERRORS_RO);
1955
1956 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1957 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1958
1959 set_opt(sbi->s_mount_opt, RESERVATION);
1960 set_opt(sbi->s_mount_opt, BARRIER);
1961
1962 /*
1963 * turn on extents feature by default in ext4 filesystem
1964 * only if feature flag already set by mkfs or tune2fs.
1965 * Use -o noextents to turn it off
1966 */
1967 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
1968 set_opt(sbi->s_mount_opt, EXTENTS);
1969 else
1970 ext4_warning(sb, __func__,
1971 "extents feature not enabled on this filesystem, "
1972 "use tune2fs.\n");
1973
1974 /*
1975 * enable delayed allocation by default
1976 * Use -o nodelalloc to turn it off
1977 */
1978 set_opt(sbi->s_mount_opt, DELALLOC);
1979
1980
1981 if (!parse_options((char *) data, sb, &journal_inum, &journal_devnum,
1982 NULL, 0))
1983 goto failed_mount;
1984
1985 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1986 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1987
1988 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1989 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1990 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1991 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1992 printk(KERN_WARNING
1993 "EXT4-fs warning: feature flags set on rev 0 fs, "
1994 "running e2fsck is recommended\n");
1995
1996 /*
1997 * Check feature flags regardless of the revision level, since we
1998 * previously didn't change the revision level when setting the flags,
1999 * so there is a chance incompat flags are set on a rev 0 filesystem.
2000 */
2001 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2002 if (features) {
2003 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2004 "unsupported optional features (%x).\n",
2005 sb->s_id, le32_to_cpu(features));
2006 goto failed_mount;
2007 }
2008 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2009 if (!(sb->s_flags & MS_RDONLY) && features) {
2010 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2011 "unsupported optional features (%x).\n",
2012 sb->s_id, le32_to_cpu(features));
2013 goto failed_mount;
2014 }
2015 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2016 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2017 if (has_huge_files) {
2018 /*
2019 * Large file size enabled file system can only be
2020 * mount if kernel is build with CONFIG_LSF
2021 */
2022 if (sizeof(root->i_blocks) < sizeof(u64) &&
2023 !(sb->s_flags & MS_RDONLY)) {
2024 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2025 "files cannot be mounted read-write "
2026 "without CONFIG_LSF.\n", sb->s_id);
2027 goto failed_mount;
2028 }
2029 }
2030 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2031
2032 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2033 blocksize > EXT4_MAX_BLOCK_SIZE) {
2034 printk(KERN_ERR
2035 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2036 blocksize, sb->s_id);
2037 goto failed_mount;
2038 }
2039
2040 if (sb->s_blocksize != blocksize) {
2041
2042 /* Validate the filesystem blocksize */
2043 if (!sb_set_blocksize(sb, blocksize)) {
2044 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2045 blocksize);
2046 goto failed_mount;
2047 }
2048
2049 brelse(bh);
2050 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2051 offset = do_div(logical_sb_block, blocksize);
2052 bh = sb_bread(sb, logical_sb_block);
2053 if (!bh) {
2054 printk(KERN_ERR
2055 "EXT4-fs: Can't read superblock on 2nd try.\n");
2056 goto failed_mount;
2057 }
2058 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2059 sbi->s_es = es;
2060 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2061 printk(KERN_ERR
2062 "EXT4-fs: Magic mismatch, very weird !\n");
2063 goto failed_mount;
2064 }
2065 }
2066
2067 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2068 has_huge_files);
2069 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2070
2071 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2072 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2073 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2074 } else {
2075 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2076 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2077 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2078 (!is_power_of_2(sbi->s_inode_size)) ||
2079 (sbi->s_inode_size > blocksize)) {
2080 printk(KERN_ERR
2081 "EXT4-fs: unsupported inode size: %d\n",
2082 sbi->s_inode_size);
2083 goto failed_mount;
2084 }
2085 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2086 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2087 }
2088 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2089 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2090 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2091 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2092 !is_power_of_2(sbi->s_desc_size)) {
2093 printk(KERN_ERR
2094 "EXT4-fs: unsupported descriptor size %lu\n",
2095 sbi->s_desc_size);
2096 goto failed_mount;
2097 }
2098 } else
2099 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2100 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2101 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2102 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2103 goto cantfind_ext4;
2104 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2105 if (sbi->s_inodes_per_block == 0)
2106 goto cantfind_ext4;
2107 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2108 sbi->s_inodes_per_block;
2109 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2110 sbi->s_sbh = bh;
2111 sbi->s_mount_state = le16_to_cpu(es->s_state);
2112 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2113 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2114 for (i = 0; i < 4; i++)
2115 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2116 sbi->s_def_hash_version = es->s_def_hash_version;
2117 i = le32_to_cpu(es->s_flags);
2118 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2119 sbi->s_hash_unsigned = 3;
2120 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2121 #ifdef __CHAR_UNSIGNED__
2122 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2123 sbi->s_hash_unsigned = 3;
2124 #else
2125 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2126 #endif
2127 sb->s_dirt = 1;
2128 }
2129
2130 if (sbi->s_blocks_per_group > blocksize * 8) {
2131 printk(KERN_ERR
2132 "EXT4-fs: #blocks per group too big: %lu\n",
2133 sbi->s_blocks_per_group);
2134 goto failed_mount;
2135 }
2136 if (sbi->s_inodes_per_group > blocksize * 8) {
2137 printk(KERN_ERR
2138 "EXT4-fs: #inodes per group too big: %lu\n",
2139 sbi->s_inodes_per_group);
2140 goto failed_mount;
2141 }
2142
2143 if (ext4_blocks_count(es) >
2144 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2145 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2146 " too large to mount safely\n", sb->s_id);
2147 if (sizeof(sector_t) < 8)
2148 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2149 "enabled\n");
2150 goto failed_mount;
2151 }
2152
2153 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2154 goto cantfind_ext4;
2155
2156 /*
2157 * It makes no sense for the first data block to be beyond the end
2158 * of the filesystem.
2159 */
2160 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2161 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2162 "block %u is beyond end of filesystem (%llu)\n",
2163 le32_to_cpu(es->s_first_data_block),
2164 ext4_blocks_count(es));
2165 goto failed_mount;
2166 }
2167 blocks_count = (ext4_blocks_count(es) -
2168 le32_to_cpu(es->s_first_data_block) +
2169 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2170 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2171 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2172 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2173 "(block count %llu, first data block %u, "
2174 "blocks per group %lu)\n", sbi->s_groups_count,
2175 ext4_blocks_count(es),
2176 le32_to_cpu(es->s_first_data_block),
2177 EXT4_BLOCKS_PER_GROUP(sb));
2178 goto failed_mount;
2179 }
2180 sbi->s_groups_count = blocks_count;
2181 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2182 EXT4_DESC_PER_BLOCK(sb);
2183 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2184 GFP_KERNEL);
2185 if (sbi->s_group_desc == NULL) {
2186 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2187 goto failed_mount;
2188 }
2189
2190 #ifdef CONFIG_PROC_FS
2191 if (ext4_proc_root)
2192 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2193
2194 if (sbi->s_proc)
2195 proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2196 &ext4_ui_proc_fops,
2197 &sbi->s_inode_readahead_blks);
2198 #endif
2199
2200 bgl_lock_init(&sbi->s_blockgroup_lock);
2201
2202 for (i = 0; i < db_count; i++) {
2203 block = descriptor_loc(sb, logical_sb_block, i);
2204 sbi->s_group_desc[i] = sb_bread(sb, block);
2205 if (!sbi->s_group_desc[i]) {
2206 printk(KERN_ERR "EXT4-fs: "
2207 "can't read group descriptor %d\n", i);
2208 db_count = i;
2209 goto failed_mount2;
2210 }
2211 }
2212 if (!ext4_check_descriptors(sb)) {
2213 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2214 goto failed_mount2;
2215 }
2216 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2217 if (!ext4_fill_flex_info(sb)) {
2218 printk(KERN_ERR
2219 "EXT4-fs: unable to initialize "
2220 "flex_bg meta info!\n");
2221 goto failed_mount2;
2222 }
2223
2224 sbi->s_gdb_count = db_count;
2225 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2226 spin_lock_init(&sbi->s_next_gen_lock);
2227
2228 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2229 ext4_count_free_blocks(sb));
2230 if (!err) {
2231 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2232 ext4_count_free_inodes(sb));
2233 }
2234 if (!err) {
2235 err = percpu_counter_init(&sbi->s_dirs_counter,
2236 ext4_count_dirs(sb));
2237 }
2238 if (!err) {
2239 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2240 }
2241 if (err) {
2242 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2243 goto failed_mount3;
2244 }
2245
2246 sbi->s_stripe = ext4_get_stripe_size(sbi);
2247
2248 /*
2249 * set up enough so that it can read an inode
2250 */
2251 sb->s_op = &ext4_sops;
2252 sb->s_export_op = &ext4_export_ops;
2253 sb->s_xattr = ext4_xattr_handlers;
2254 #ifdef CONFIG_QUOTA
2255 sb->s_qcop = &ext4_qctl_operations;
2256 sb->dq_op = &ext4_quota_operations;
2257 #endif
2258 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2259
2260 sb->s_root = NULL;
2261
2262 needs_recovery = (es->s_last_orphan != 0 ||
2263 EXT4_HAS_INCOMPAT_FEATURE(sb,
2264 EXT4_FEATURE_INCOMPAT_RECOVER));
2265
2266 /*
2267 * The first inode we look at is the journal inode. Don't try
2268 * root first: it may be modified in the journal!
2269 */
2270 if (!test_opt(sb, NOLOAD) &&
2271 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2272 if (ext4_load_journal(sb, es, journal_devnum))
2273 goto failed_mount3;
2274 if (!(sb->s_flags & MS_RDONLY) &&
2275 EXT4_SB(sb)->s_journal->j_failed_commit) {
2276 printk(KERN_CRIT "EXT4-fs error (device %s): "
2277 "ext4_fill_super: Journal transaction "
2278 "%u is corrupt\n", sb->s_id,
2279 EXT4_SB(sb)->s_journal->j_failed_commit);
2280 if (test_opt(sb, ERRORS_RO)) {
2281 printk(KERN_CRIT
2282 "Mounting filesystem read-only\n");
2283 sb->s_flags |= MS_RDONLY;
2284 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2285 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2286 }
2287 if (test_opt(sb, ERRORS_PANIC)) {
2288 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2289 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2290 ext4_commit_super(sb, es, 1);
2291 printk(KERN_CRIT
2292 "EXT4-fs (device %s): mount failed\n",
2293 sb->s_id);
2294 goto failed_mount4;
2295 }
2296 }
2297 } else if (journal_inum) {
2298 if (ext4_create_journal(sb, es, journal_inum))
2299 goto failed_mount3;
2300 } else {
2301 if (!silent)
2302 printk(KERN_ERR
2303 "ext4: No journal on filesystem on %s\n",
2304 sb->s_id);
2305 goto failed_mount3;
2306 }
2307
2308 if (ext4_blocks_count(es) > 0xffffffffULL &&
2309 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2310 JBD2_FEATURE_INCOMPAT_64BIT)) {
2311 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2312 goto failed_mount4;
2313 }
2314
2315 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2316 jbd2_journal_set_features(sbi->s_journal,
2317 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2318 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2319 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2320 jbd2_journal_set_features(sbi->s_journal,
2321 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2322 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2323 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2324 } else {
2325 jbd2_journal_clear_features(sbi->s_journal,
2326 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2327 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2328 }
2329
2330 /* We have now updated the journal if required, so we can
2331 * validate the data journaling mode. */
2332 switch (test_opt(sb, DATA_FLAGS)) {
2333 case 0:
2334 /* No mode set, assume a default based on the journal
2335 * capabilities: ORDERED_DATA if the journal can
2336 * cope, else JOURNAL_DATA
2337 */
2338 if (jbd2_journal_check_available_features
2339 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2340 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2341 else
2342 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2343 break;
2344
2345 case EXT4_MOUNT_ORDERED_DATA:
2346 case EXT4_MOUNT_WRITEBACK_DATA:
2347 if (!jbd2_journal_check_available_features
2348 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2349 printk(KERN_ERR "EXT4-fs: Journal does not support "
2350 "requested data journaling mode\n");
2351 goto failed_mount4;
2352 }
2353 default:
2354 break;
2355 }
2356
2357 if (test_opt(sb, NOBH)) {
2358 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2359 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2360 "its supported only with writeback mode\n");
2361 clear_opt(sbi->s_mount_opt, NOBH);
2362 }
2363 }
2364 /*
2365 * The jbd2_journal_load will have done any necessary log recovery,
2366 * so we can safely mount the rest of the filesystem now.
2367 */
2368
2369 root = ext4_iget(sb, EXT4_ROOT_INO);
2370 if (IS_ERR(root)) {
2371 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2372 ret = PTR_ERR(root);
2373 goto failed_mount4;
2374 }
2375 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2376 iput(root);
2377 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2378 goto failed_mount4;
2379 }
2380 sb->s_root = d_alloc_root(root);
2381 if (!sb->s_root) {
2382 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2383 iput(root);
2384 ret = -ENOMEM;
2385 goto failed_mount4;
2386 }
2387
2388 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2389
2390 /* determine the minimum size of new large inodes, if present */
2391 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2392 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2393 EXT4_GOOD_OLD_INODE_SIZE;
2394 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2395 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2396 if (sbi->s_want_extra_isize <
2397 le16_to_cpu(es->s_want_extra_isize))
2398 sbi->s_want_extra_isize =
2399 le16_to_cpu(es->s_want_extra_isize);
2400 if (sbi->s_want_extra_isize <
2401 le16_to_cpu(es->s_min_extra_isize))
2402 sbi->s_want_extra_isize =
2403 le16_to_cpu(es->s_min_extra_isize);
2404 }
2405 }
2406 /* Check if enough inode space is available */
2407 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2408 sbi->s_inode_size) {
2409 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2410 EXT4_GOOD_OLD_INODE_SIZE;
2411 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2412 "available.\n");
2413 }
2414
2415 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2416 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2417 "requested data journaling mode\n");
2418 clear_opt(sbi->s_mount_opt, DELALLOC);
2419 } else if (test_opt(sb, DELALLOC))
2420 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2421
2422 ext4_ext_init(sb);
2423 err = ext4_mb_init(sb, needs_recovery);
2424 if (err) {
2425 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2426 err);
2427 goto failed_mount4;
2428 }
2429
2430 /*
2431 * akpm: core read_super() calls in here with the superblock locked.
2432 * That deadlocks, because orphan cleanup needs to lock the superblock
2433 * in numerous places. Here we just pop the lock - it's relatively
2434 * harmless, because we are now ready to accept write_super() requests,
2435 * and aviro says that's the only reason for hanging onto the
2436 * superblock lock.
2437 */
2438 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2439 ext4_orphan_cleanup(sb, es);
2440 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2441 if (needs_recovery)
2442 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2443 ext4_mark_recovery_complete(sb, es);
2444 printk(KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2445 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2446 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2447 "writeback");
2448
2449 lock_kernel();
2450 return 0;
2451
2452 cantfind_ext4:
2453 if (!silent)
2454 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2455 sb->s_id);
2456 goto failed_mount;
2457
2458 failed_mount4:
2459 jbd2_journal_destroy(sbi->s_journal);
2460 sbi->s_journal = NULL;
2461 failed_mount3:
2462 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2463 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2464 percpu_counter_destroy(&sbi->s_dirs_counter);
2465 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2466 failed_mount2:
2467 for (i = 0; i < db_count; i++)
2468 brelse(sbi->s_group_desc[i]);
2469 kfree(sbi->s_group_desc);
2470 failed_mount:
2471 if (sbi->s_proc) {
2472 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
2473 remove_proc_entry(sb->s_id, ext4_proc_root);
2474 }
2475 #ifdef CONFIG_QUOTA
2476 for (i = 0; i < MAXQUOTAS; i++)
2477 kfree(sbi->s_qf_names[i]);
2478 #endif
2479 ext4_blkdev_remove(sbi);
2480 brelse(bh);
2481 out_fail:
2482 sb->s_fs_info = NULL;
2483 kfree(sbi);
2484 lock_kernel();
2485 return ret;
2486 }
2487
2488 /*
2489 * Setup any per-fs journal parameters now. We'll do this both on
2490 * initial mount, once the journal has been initialised but before we've
2491 * done any recovery; and again on any subsequent remount.
2492 */
2493 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2494 {
2495 struct ext4_sb_info *sbi = EXT4_SB(sb);
2496
2497 if (sbi->s_commit_interval)
2498 journal->j_commit_interval = sbi->s_commit_interval;
2499 /* We could also set up an ext4-specific default for the commit
2500 * interval here, but for now we'll just fall back to the jbd
2501 * default. */
2502
2503 spin_lock(&journal->j_state_lock);
2504 if (test_opt(sb, BARRIER))
2505 journal->j_flags |= JBD2_BARRIER;
2506 else
2507 journal->j_flags &= ~JBD2_BARRIER;
2508 if (test_opt(sb, DATA_ERR_ABORT))
2509 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2510 else
2511 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2512 spin_unlock(&journal->j_state_lock);
2513 }
2514
2515 static journal_t *ext4_get_journal(struct super_block *sb,
2516 unsigned int journal_inum)
2517 {
2518 struct inode *journal_inode;
2519 journal_t *journal;
2520
2521 /* First, test for the existence of a valid inode on disk. Bad
2522 * things happen if we iget() an unused inode, as the subsequent
2523 * iput() will try to delete it. */
2524
2525 journal_inode = ext4_iget(sb, journal_inum);
2526 if (IS_ERR(journal_inode)) {
2527 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2528 return NULL;
2529 }
2530 if (!journal_inode->i_nlink) {
2531 make_bad_inode(journal_inode);
2532 iput(journal_inode);
2533 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2534 return NULL;
2535 }
2536
2537 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2538 journal_inode, journal_inode->i_size);
2539 if (!S_ISREG(journal_inode->i_mode)) {
2540 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2541 iput(journal_inode);
2542 return NULL;
2543 }
2544
2545 journal = jbd2_journal_init_inode(journal_inode);
2546 if (!journal) {
2547 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2548 iput(journal_inode);
2549 return NULL;
2550 }
2551 journal->j_private = sb;
2552 ext4_init_journal_params(sb, journal);
2553 return journal;
2554 }
2555
2556 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2557 dev_t j_dev)
2558 {
2559 struct buffer_head *bh;
2560 journal_t *journal;
2561 ext4_fsblk_t start;
2562 ext4_fsblk_t len;
2563 int hblock, blocksize;
2564 ext4_fsblk_t sb_block;
2565 unsigned long offset;
2566 struct ext4_super_block *es;
2567 struct block_device *bdev;
2568
2569 bdev = ext4_blkdev_get(j_dev);
2570 if (bdev == NULL)
2571 return NULL;
2572
2573 if (bd_claim(bdev, sb)) {
2574 printk(KERN_ERR
2575 "EXT4: failed to claim external journal device.\n");
2576 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2577 return NULL;
2578 }
2579
2580 blocksize = sb->s_blocksize;
2581 hblock = bdev_hardsect_size(bdev);
2582 if (blocksize < hblock) {
2583 printk(KERN_ERR
2584 "EXT4-fs: blocksize too small for journal device.\n");
2585 goto out_bdev;
2586 }
2587
2588 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2589 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2590 set_blocksize(bdev, blocksize);
2591 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2592 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2593 "external journal\n");
2594 goto out_bdev;
2595 }
2596
2597 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2598 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2599 !(le32_to_cpu(es->s_feature_incompat) &
2600 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2601 printk(KERN_ERR "EXT4-fs: external journal has "
2602 "bad superblock\n");
2603 brelse(bh);
2604 goto out_bdev;
2605 }
2606
2607 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2608 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2609 brelse(bh);
2610 goto out_bdev;
2611 }
2612
2613 len = ext4_blocks_count(es);
2614 start = sb_block + 1;
2615 brelse(bh); /* we're done with the superblock */
2616
2617 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2618 start, len, blocksize);
2619 if (!journal) {
2620 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2621 goto out_bdev;
2622 }
2623 journal->j_private = sb;
2624 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2625 wait_on_buffer(journal->j_sb_buffer);
2626 if (!buffer_uptodate(journal->j_sb_buffer)) {
2627 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2628 goto out_journal;
2629 }
2630 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2631 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2632 "user (unsupported) - %d\n",
2633 be32_to_cpu(journal->j_superblock->s_nr_users));
2634 goto out_journal;
2635 }
2636 EXT4_SB(sb)->journal_bdev = bdev;
2637 ext4_init_journal_params(sb, journal);
2638 return journal;
2639 out_journal:
2640 jbd2_journal_destroy(journal);
2641 out_bdev:
2642 ext4_blkdev_put(bdev);
2643 return NULL;
2644 }
2645
2646 static int ext4_load_journal(struct super_block *sb,
2647 struct ext4_super_block *es,
2648 unsigned long journal_devnum)
2649 {
2650 journal_t *journal;
2651 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2652 dev_t journal_dev;
2653 int err = 0;
2654 int really_read_only;
2655
2656 if (journal_devnum &&
2657 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2658 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2659 "numbers have changed\n");
2660 journal_dev = new_decode_dev(journal_devnum);
2661 } else
2662 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2663
2664 really_read_only = bdev_read_only(sb->s_bdev);
2665
2666 /*
2667 * Are we loading a blank journal or performing recovery after a
2668 * crash? For recovery, we need to check in advance whether we
2669 * can get read-write access to the device.
2670 */
2671
2672 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2673 if (sb->s_flags & MS_RDONLY) {
2674 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2675 "required on readonly filesystem.\n");
2676 if (really_read_only) {
2677 printk(KERN_ERR "EXT4-fs: write access "
2678 "unavailable, cannot proceed.\n");
2679 return -EROFS;
2680 }
2681 printk(KERN_INFO "EXT4-fs: write access will "
2682 "be enabled during recovery.\n");
2683 }
2684 }
2685
2686 if (journal_inum && journal_dev) {
2687 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2688 "and inode journals!\n");
2689 return -EINVAL;
2690 }
2691
2692 if (journal_inum) {
2693 if (!(journal = ext4_get_journal(sb, journal_inum)))
2694 return -EINVAL;
2695 } else {
2696 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2697 return -EINVAL;
2698 }
2699
2700 if (journal->j_flags & JBD2_BARRIER)
2701 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2702 else
2703 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2704
2705 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2706 err = jbd2_journal_update_format(journal);
2707 if (err) {
2708 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2709 jbd2_journal_destroy(journal);
2710 return err;
2711 }
2712 }
2713
2714 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2715 err = jbd2_journal_wipe(journal, !really_read_only);
2716 if (!err)
2717 err = jbd2_journal_load(journal);
2718
2719 if (err) {
2720 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2721 jbd2_journal_destroy(journal);
2722 return err;
2723 }
2724
2725 EXT4_SB(sb)->s_journal = journal;
2726 ext4_clear_journal_err(sb, es);
2727
2728 if (journal_devnum &&
2729 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2730 es->s_journal_dev = cpu_to_le32(journal_devnum);
2731 sb->s_dirt = 1;
2732
2733 /* Make sure we flush the recovery flag to disk. */
2734 ext4_commit_super(sb, es, 1);
2735 }
2736
2737 return 0;
2738 }
2739
2740 static int ext4_create_journal(struct super_block *sb,
2741 struct ext4_super_block *es,
2742 unsigned int journal_inum)
2743 {
2744 journal_t *journal;
2745 int err;
2746
2747 if (sb->s_flags & MS_RDONLY) {
2748 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2749 "create journal.\n");
2750 return -EROFS;
2751 }
2752
2753 journal = ext4_get_journal(sb, journal_inum);
2754 if (!journal)
2755 return -EINVAL;
2756
2757 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2758 journal_inum);
2759
2760 err = jbd2_journal_create(journal);
2761 if (err) {
2762 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2763 jbd2_journal_destroy(journal);
2764 return -EIO;
2765 }
2766
2767 EXT4_SB(sb)->s_journal = journal;
2768
2769 ext4_update_dynamic_rev(sb);
2770 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2771 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2772
2773 es->s_journal_inum = cpu_to_le32(journal_inum);
2774 sb->s_dirt = 1;
2775
2776 /* Make sure we flush the recovery flag to disk. */
2777 ext4_commit_super(sb, es, 1);
2778
2779 return 0;
2780 }
2781
2782 static void ext4_commit_super(struct super_block *sb,
2783 struct ext4_super_block *es, int sync)
2784 {
2785 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2786
2787 if (!sbh)
2788 return;
2789 if (buffer_write_io_error(sbh)) {
2790 /*
2791 * Oh, dear. A previous attempt to write the
2792 * superblock failed. This could happen because the
2793 * USB device was yanked out. Or it could happen to
2794 * be a transient write error and maybe the block will
2795 * be remapped. Nothing we can do but to retry the
2796 * write and hope for the best.
2797 */
2798 printk(KERN_ERR "ext4: previous I/O error to "
2799 "superblock detected for %s.\n", sb->s_id);
2800 clear_buffer_write_io_error(sbh);
2801 set_buffer_uptodate(sbh);
2802 }
2803 es->s_wtime = cpu_to_le32(get_seconds());
2804 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2805 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2806 BUFFER_TRACE(sbh, "marking dirty");
2807 mark_buffer_dirty(sbh);
2808 if (sync) {
2809 sync_dirty_buffer(sbh);
2810 if (buffer_write_io_error(sbh)) {
2811 printk(KERN_ERR "ext4: I/O error while writing "
2812 "superblock for %s.\n", sb->s_id);
2813 clear_buffer_write_io_error(sbh);
2814 set_buffer_uptodate(sbh);
2815 }
2816 }
2817 }
2818
2819
2820 /*
2821 * Have we just finished recovery? If so, and if we are mounting (or
2822 * remounting) the filesystem readonly, then we will end up with a
2823 * consistent fs on disk. Record that fact.
2824 */
2825 static void ext4_mark_recovery_complete(struct super_block *sb,
2826 struct ext4_super_block *es)
2827 {
2828 journal_t *journal = EXT4_SB(sb)->s_journal;
2829
2830 jbd2_journal_lock_updates(journal);
2831 if (jbd2_journal_flush(journal) < 0)
2832 goto out;
2833
2834 lock_super(sb);
2835 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2836 sb->s_flags & MS_RDONLY) {
2837 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2838 sb->s_dirt = 0;
2839 ext4_commit_super(sb, es, 1);
2840 }
2841 unlock_super(sb);
2842
2843 out:
2844 jbd2_journal_unlock_updates(journal);
2845 }
2846
2847 /*
2848 * If we are mounting (or read-write remounting) a filesystem whose journal
2849 * has recorded an error from a previous lifetime, move that error to the
2850 * main filesystem now.
2851 */
2852 static void ext4_clear_journal_err(struct super_block *sb,
2853 struct ext4_super_block *es)
2854 {
2855 journal_t *journal;
2856 int j_errno;
2857 const char *errstr;
2858
2859 journal = EXT4_SB(sb)->s_journal;
2860
2861 /*
2862 * Now check for any error status which may have been recorded in the
2863 * journal by a prior ext4_error() or ext4_abort()
2864 */
2865
2866 j_errno = jbd2_journal_errno(journal);
2867 if (j_errno) {
2868 char nbuf[16];
2869
2870 errstr = ext4_decode_error(sb, j_errno, nbuf);
2871 ext4_warning(sb, __func__, "Filesystem error recorded "
2872 "from previous mount: %s", errstr);
2873 ext4_warning(sb, __func__, "Marking fs in need of "
2874 "filesystem check.");
2875
2876 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2877 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2878 ext4_commit_super(sb, es, 1);
2879
2880 jbd2_journal_clear_err(journal);
2881 }
2882 }
2883
2884 /*
2885 * Force the running and committing transactions to commit,
2886 * and wait on the commit.
2887 */
2888 int ext4_force_commit(struct super_block *sb)
2889 {
2890 journal_t *journal;
2891 int ret;
2892
2893 if (sb->s_flags & MS_RDONLY)
2894 return 0;
2895
2896 journal = EXT4_SB(sb)->s_journal;
2897 sb->s_dirt = 0;
2898 ret = ext4_journal_force_commit(journal);
2899 return ret;
2900 }
2901
2902 /*
2903 * Ext4 always journals updates to the superblock itself, so we don't
2904 * have to propagate any other updates to the superblock on disk at this
2905 * point. (We can probably nuke this function altogether, and remove
2906 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
2907 */
2908 static void ext4_write_super(struct super_block *sb)
2909 {
2910 if (mutex_trylock(&sb->s_lock) != 0)
2911 BUG();
2912 sb->s_dirt = 0;
2913 }
2914
2915 static int ext4_sync_fs(struct super_block *sb, int wait)
2916 {
2917 tid_t target;
2918
2919 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
2920 sb->s_dirt = 0;
2921 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2922 if (wait)
2923 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2924 }
2925 return 0;
2926 }
2927
2928 /*
2929 * LVM calls this function before a (read-only) snapshot is created. This
2930 * gives us a chance to flush the journal completely and mark the fs clean.
2931 */
2932 static void ext4_write_super_lockfs(struct super_block *sb)
2933 {
2934 sb->s_dirt = 0;
2935
2936 if (!(sb->s_flags & MS_RDONLY)) {
2937 journal_t *journal = EXT4_SB(sb)->s_journal;
2938
2939 /* Now we set up the journal barrier. */
2940 jbd2_journal_lock_updates(journal);
2941
2942 /*
2943 * We don't want to clear needs_recovery flag when we failed
2944 * to flush the journal.
2945 */
2946 if (jbd2_journal_flush(journal) < 0)
2947 return;
2948
2949 /* Journal blocked and flushed, clear needs_recovery flag. */
2950 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2951 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2952 }
2953 }
2954
2955 /*
2956 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2957 * flag here, even though the filesystem is not technically dirty yet.
2958 */
2959 static void ext4_unlockfs(struct super_block *sb)
2960 {
2961 if (!(sb->s_flags & MS_RDONLY)) {
2962 lock_super(sb);
2963 /* Reser the needs_recovery flag before the fs is unlocked. */
2964 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2965 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2966 unlock_super(sb);
2967 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2968 }
2969 }
2970
2971 static int ext4_remount(struct super_block *sb, int *flags, char *data)
2972 {
2973 struct ext4_super_block *es;
2974 struct ext4_sb_info *sbi = EXT4_SB(sb);
2975 ext4_fsblk_t n_blocks_count = 0;
2976 unsigned long old_sb_flags;
2977 struct ext4_mount_options old_opts;
2978 ext4_group_t g;
2979 int err;
2980 #ifdef CONFIG_QUOTA
2981 int i;
2982 #endif
2983
2984 /* Store the original options */
2985 old_sb_flags = sb->s_flags;
2986 old_opts.s_mount_opt = sbi->s_mount_opt;
2987 old_opts.s_resuid = sbi->s_resuid;
2988 old_opts.s_resgid = sbi->s_resgid;
2989 old_opts.s_commit_interval = sbi->s_commit_interval;
2990 #ifdef CONFIG_QUOTA
2991 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2992 for (i = 0; i < MAXQUOTAS; i++)
2993 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2994 #endif
2995
2996 /*
2997 * Allow the "check" option to be passed as a remount option.
2998 */
2999 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
3000 err = -EINVAL;
3001 goto restore_opts;
3002 }
3003
3004 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3005 ext4_abort(sb, __func__, "Abort forced by user");
3006
3007 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3008 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3009
3010 es = sbi->s_es;
3011
3012 ext4_init_journal_params(sb, sbi->s_journal);
3013
3014 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3015 n_blocks_count > ext4_blocks_count(es)) {
3016 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3017 err = -EROFS;
3018 goto restore_opts;
3019 }
3020
3021 if (*flags & MS_RDONLY) {
3022 /*
3023 * First of all, the unconditional stuff we have to do
3024 * to disable replay of the journal when we next remount
3025 */
3026 sb->s_flags |= MS_RDONLY;
3027
3028 /*
3029 * OK, test if we are remounting a valid rw partition
3030 * readonly, and if so set the rdonly flag and then
3031 * mark the partition as valid again.
3032 */
3033 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3034 (sbi->s_mount_state & EXT4_VALID_FS))
3035 es->s_state = cpu_to_le16(sbi->s_mount_state);
3036
3037 /*
3038 * We have to unlock super so that we can wait for
3039 * transactions.
3040 */
3041 unlock_super(sb);
3042 ext4_mark_recovery_complete(sb, es);
3043 lock_super(sb);
3044 } else {
3045 __le32 ret;
3046 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3047 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3048 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3049 "remount RDWR because of unsupported "
3050 "optional features (%x).\n",
3051 sb->s_id, le32_to_cpu(ret));
3052 err = -EROFS;
3053 goto restore_opts;
3054 }
3055
3056 /*
3057 * Make sure the group descriptor checksums
3058 * are sane. If they aren't, refuse to
3059 * remount r/w.
3060 */
3061 for (g = 0; g < sbi->s_groups_count; g++) {
3062 struct ext4_group_desc *gdp =
3063 ext4_get_group_desc(sb, g, NULL);
3064
3065 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3066 printk(KERN_ERR
3067 "EXT4-fs: ext4_remount: "
3068 "Checksum for group %lu failed (%u!=%u)\n",
3069 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3070 le16_to_cpu(gdp->bg_checksum));
3071 err = -EINVAL;
3072 goto restore_opts;
3073 }
3074 }
3075
3076 /*
3077 * If we have an unprocessed orphan list hanging
3078 * around from a previously readonly bdev mount,
3079 * require a full umount/remount for now.
3080 */
3081 if (es->s_last_orphan) {
3082 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3083 "remount RDWR because of unprocessed "
3084 "orphan inode list. Please "
3085 "umount/remount instead.\n",
3086 sb->s_id);
3087 err = -EINVAL;
3088 goto restore_opts;
3089 }
3090
3091 /*
3092 * Mounting a RDONLY partition read-write, so reread
3093 * and store the current valid flag. (It may have
3094 * been changed by e2fsck since we originally mounted
3095 * the partition.)
3096 */
3097 ext4_clear_journal_err(sb, es);
3098 sbi->s_mount_state = le16_to_cpu(es->s_state);
3099 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3100 goto restore_opts;
3101 if (!ext4_setup_super(sb, es, 0))
3102 sb->s_flags &= ~MS_RDONLY;
3103 }
3104 }
3105 #ifdef CONFIG_QUOTA
3106 /* Release old quota file names */
3107 for (i = 0; i < MAXQUOTAS; i++)
3108 if (old_opts.s_qf_names[i] &&
3109 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3110 kfree(old_opts.s_qf_names[i]);
3111 #endif
3112 return 0;
3113 restore_opts:
3114 sb->s_flags = old_sb_flags;
3115 sbi->s_mount_opt = old_opts.s_mount_opt;
3116 sbi->s_resuid = old_opts.s_resuid;
3117 sbi->s_resgid = old_opts.s_resgid;
3118 sbi->s_commit_interval = old_opts.s_commit_interval;
3119 #ifdef CONFIG_QUOTA
3120 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3121 for (i = 0; i < MAXQUOTAS; i++) {
3122 if (sbi->s_qf_names[i] &&
3123 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3124 kfree(sbi->s_qf_names[i]);
3125 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3126 }
3127 #endif
3128 return err;
3129 }
3130
3131 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3132 {
3133 struct super_block *sb = dentry->d_sb;
3134 struct ext4_sb_info *sbi = EXT4_SB(sb);
3135 struct ext4_super_block *es = sbi->s_es;
3136 u64 fsid;
3137
3138 if (test_opt(sb, MINIX_DF)) {
3139 sbi->s_overhead_last = 0;
3140 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3141 ext4_group_t ngroups = sbi->s_groups_count, i;
3142 ext4_fsblk_t overhead = 0;
3143 smp_rmb();
3144
3145 /*
3146 * Compute the overhead (FS structures). This is constant
3147 * for a given filesystem unless the number of block groups
3148 * changes so we cache the previous value until it does.
3149 */
3150
3151 /*
3152 * All of the blocks before first_data_block are
3153 * overhead
3154 */
3155 overhead = le32_to_cpu(es->s_first_data_block);
3156
3157 /*
3158 * Add the overhead attributed to the superblock and
3159 * block group descriptors. If the sparse superblocks
3160 * feature is turned on, then not all groups have this.
3161 */
3162 for (i = 0; i < ngroups; i++) {
3163 overhead += ext4_bg_has_super(sb, i) +
3164 ext4_bg_num_gdb(sb, i);
3165 cond_resched();
3166 }
3167
3168 /*
3169 * Every block group has an inode bitmap, a block
3170 * bitmap, and an inode table.
3171 */
3172 overhead += ngroups * (2 + sbi->s_itb_per_group);
3173 sbi->s_overhead_last = overhead;
3174 smp_wmb();
3175 sbi->s_blocks_last = ext4_blocks_count(es);
3176 }
3177
3178 buf->f_type = EXT4_SUPER_MAGIC;
3179 buf->f_bsize = sb->s_blocksize;
3180 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3181 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3182 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3183 ext4_free_blocks_count_set(es, buf->f_bfree);
3184 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3185 if (buf->f_bfree < ext4_r_blocks_count(es))
3186 buf->f_bavail = 0;
3187 buf->f_files = le32_to_cpu(es->s_inodes_count);
3188 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3189 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3190 buf->f_namelen = EXT4_NAME_LEN;
3191 fsid = le64_to_cpup((void *)es->s_uuid) ^
3192 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3193 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3194 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3195 return 0;
3196 }
3197
3198 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3199 * is locked for write. Otherwise the are possible deadlocks:
3200 * Process 1 Process 2
3201 * ext4_create() quota_sync()
3202 * jbd2_journal_start() write_dquot()
3203 * DQUOT_INIT() down(dqio_mutex)
3204 * down(dqio_mutex) jbd2_journal_start()
3205 *
3206 */
3207
3208 #ifdef CONFIG_QUOTA
3209
3210 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3211 {
3212 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3213 }
3214
3215 static int ext4_dquot_initialize(struct inode *inode, int type)
3216 {
3217 handle_t *handle;
3218 int ret, err;
3219
3220 /* We may create quota structure so we need to reserve enough blocks */
3221 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3222 if (IS_ERR(handle))
3223 return PTR_ERR(handle);
3224 ret = dquot_initialize(inode, type);
3225 err = ext4_journal_stop(handle);
3226 if (!ret)
3227 ret = err;
3228 return ret;
3229 }
3230
3231 static int ext4_dquot_drop(struct inode *inode)
3232 {
3233 handle_t *handle;
3234 int ret, err;
3235
3236 /* We may delete quota structure so we need to reserve enough blocks */
3237 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3238 if (IS_ERR(handle)) {
3239 /*
3240 * We call dquot_drop() anyway to at least release references
3241 * to quota structures so that umount does not hang.
3242 */
3243 dquot_drop(inode);
3244 return PTR_ERR(handle);
3245 }
3246 ret = dquot_drop(inode);
3247 err = ext4_journal_stop(handle);
3248 if (!ret)
3249 ret = err;
3250 return ret;
3251 }
3252
3253 static int ext4_write_dquot(struct dquot *dquot)
3254 {
3255 int ret, err;
3256 handle_t *handle;
3257 struct inode *inode;
3258
3259 inode = dquot_to_inode(dquot);
3260 handle = ext4_journal_start(inode,
3261 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3262 if (IS_ERR(handle))
3263 return PTR_ERR(handle);
3264 ret = dquot_commit(dquot);
3265 err = ext4_journal_stop(handle);
3266 if (!ret)
3267 ret = err;
3268 return ret;
3269 }
3270
3271 static int ext4_acquire_dquot(struct dquot *dquot)
3272 {
3273 int ret, err;
3274 handle_t *handle;
3275
3276 handle = ext4_journal_start(dquot_to_inode(dquot),
3277 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3278 if (IS_ERR(handle))
3279 return PTR_ERR(handle);
3280 ret = dquot_acquire(dquot);
3281 err = ext4_journal_stop(handle);
3282 if (!ret)
3283 ret = err;
3284 return ret;
3285 }
3286
3287 static int ext4_release_dquot(struct dquot *dquot)
3288 {
3289 int ret, err;
3290 handle_t *handle;
3291
3292 handle = ext4_journal_start(dquot_to_inode(dquot),
3293 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3294 if (IS_ERR(handle)) {
3295 /* Release dquot anyway to avoid endless cycle in dqput() */
3296 dquot_release(dquot);
3297 return PTR_ERR(handle);
3298 }
3299 ret = dquot_release(dquot);
3300 err = ext4_journal_stop(handle);
3301 if (!ret)
3302 ret = err;
3303 return ret;
3304 }
3305
3306 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3307 {
3308 /* Are we journaling quotas? */
3309 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3310 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3311 dquot_mark_dquot_dirty(dquot);
3312 return ext4_write_dquot(dquot);
3313 } else {
3314 return dquot_mark_dquot_dirty(dquot);
3315 }
3316 }
3317
3318 static int ext4_write_info(struct super_block *sb, int type)
3319 {
3320 int ret, err;
3321 handle_t *handle;
3322
3323 /* Data block + inode block */
3324 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3325 if (IS_ERR(handle))
3326 return PTR_ERR(handle);
3327 ret = dquot_commit_info(sb, type);
3328 err = ext4_journal_stop(handle);
3329 if (!ret)
3330 ret = err;
3331 return ret;
3332 }
3333
3334 /*
3335 * Turn on quotas during mount time - we need to find
3336 * the quota file and such...
3337 */
3338 static int ext4_quota_on_mount(struct super_block *sb, int type)
3339 {
3340 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3341 EXT4_SB(sb)->s_jquota_fmt, type);
3342 }
3343
3344 /*
3345 * Standard function to be called on quota_on
3346 */
3347 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3348 char *name, int remount)
3349 {
3350 int err;
3351 struct path path;
3352
3353 if (!test_opt(sb, QUOTA))
3354 return -EINVAL;
3355 /* When remounting, no checks are needed and in fact, name is NULL */
3356 if (remount)
3357 return vfs_quota_on(sb, type, format_id, name, remount);
3358
3359 err = kern_path(name, LOOKUP_FOLLOW, &path);
3360 if (err)
3361 return err;
3362
3363 /* Quotafile not on the same filesystem? */
3364 if (path.mnt->mnt_sb != sb) {
3365 path_put(&path);
3366 return -EXDEV;
3367 }
3368 /* Journaling quota? */
3369 if (EXT4_SB(sb)->s_qf_names[type]) {
3370 /* Quotafile not in fs root? */
3371 if (path.dentry->d_parent != sb->s_root)
3372 printk(KERN_WARNING
3373 "EXT4-fs: Quota file not on filesystem root. "
3374 "Journaled quota will not work.\n");
3375 }
3376
3377 /*
3378 * When we journal data on quota file, we have to flush journal to see
3379 * all updates to the file when we bypass pagecache...
3380 */
3381 if (ext4_should_journal_data(path.dentry->d_inode)) {
3382 /*
3383 * We don't need to lock updates but journal_flush() could
3384 * otherwise be livelocked...
3385 */
3386 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3387 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3388 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3389 if (err) {
3390 path_put(&path);
3391 return err;
3392 }
3393 }
3394
3395 err = vfs_quota_on_path(sb, type, format_id, &path);
3396 path_put(&path);
3397 return err;
3398 }
3399
3400 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3401 * acquiring the locks... As quota files are never truncated and quota code
3402 * itself serializes the operations (and noone else should touch the files)
3403 * we don't have to be afraid of races */
3404 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3405 size_t len, loff_t off)
3406 {
3407 struct inode *inode = sb_dqopt(sb)->files[type];
3408 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3409 int err = 0;
3410 int offset = off & (sb->s_blocksize - 1);
3411 int tocopy;
3412 size_t toread;
3413 struct buffer_head *bh;
3414 loff_t i_size = i_size_read(inode);
3415
3416 if (off > i_size)
3417 return 0;
3418 if (off+len > i_size)
3419 len = i_size-off;
3420 toread = len;
3421 while (toread > 0) {
3422 tocopy = sb->s_blocksize - offset < toread ?
3423 sb->s_blocksize - offset : toread;
3424 bh = ext4_bread(NULL, inode, blk, 0, &err);
3425 if (err)
3426 return err;
3427 if (!bh) /* A hole? */
3428 memset(data, 0, tocopy);
3429 else
3430 memcpy(data, bh->b_data+offset, tocopy);
3431 brelse(bh);
3432 offset = 0;
3433 toread -= tocopy;
3434 data += tocopy;
3435 blk++;
3436 }
3437 return len;
3438 }
3439
3440 /* Write to quotafile (we know the transaction is already started and has
3441 * enough credits) */
3442 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3443 const char *data, size_t len, loff_t off)
3444 {
3445 struct inode *inode = sb_dqopt(sb)->files[type];
3446 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3447 int err = 0;
3448 int offset = off & (sb->s_blocksize - 1);
3449 int tocopy;
3450 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3451 size_t towrite = len;
3452 struct buffer_head *bh;
3453 handle_t *handle = journal_current_handle();
3454
3455 if (!handle) {
3456 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3457 " cancelled because transaction is not started.\n",
3458 (unsigned long long)off, (unsigned long long)len);
3459 return -EIO;
3460 }
3461 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3462 while (towrite > 0) {
3463 tocopy = sb->s_blocksize - offset < towrite ?
3464 sb->s_blocksize - offset : towrite;
3465 bh = ext4_bread(handle, inode, blk, 1, &err);
3466 if (!bh)
3467 goto out;
3468 if (journal_quota) {
3469 err = ext4_journal_get_write_access(handle, bh);
3470 if (err) {
3471 brelse(bh);
3472 goto out;
3473 }
3474 }
3475 lock_buffer(bh);
3476 memcpy(bh->b_data+offset, data, tocopy);
3477 flush_dcache_page(bh->b_page);
3478 unlock_buffer(bh);
3479 if (journal_quota)
3480 err = ext4_journal_dirty_metadata(handle, bh);
3481 else {
3482 /* Always do at least ordered writes for quotas */
3483 err = ext4_jbd2_file_inode(handle, inode);
3484 mark_buffer_dirty(bh);
3485 }
3486 brelse(bh);
3487 if (err)
3488 goto out;
3489 offset = 0;
3490 towrite -= tocopy;
3491 data += tocopy;
3492 blk++;
3493 }
3494 out:
3495 if (len == towrite) {
3496 mutex_unlock(&inode->i_mutex);
3497 return err;
3498 }
3499 if (inode->i_size < off+len-towrite) {
3500 i_size_write(inode, off+len-towrite);
3501 EXT4_I(inode)->i_disksize = inode->i_size;
3502 }
3503 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3504 ext4_mark_inode_dirty(handle, inode);
3505 mutex_unlock(&inode->i_mutex);
3506 return len - towrite;
3507 }
3508
3509 #endif
3510
3511 static int ext4_get_sb(struct file_system_type *fs_type,
3512 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3513 {
3514 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3515 }
3516
3517 #ifdef CONFIG_PROC_FS
3518 static int ext4_ui_proc_show(struct seq_file *m, void *v)
3519 {
3520 unsigned int *p = m->private;
3521
3522 seq_printf(m, "%u\n", *p);
3523 return 0;
3524 }
3525
3526 static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3527 {
3528 return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3529 }
3530
3531 static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3532 size_t cnt, loff_t *ppos)
3533 {
3534 unsigned int *p = PDE(file->f_path.dentry->d_inode)->data;
3535 char str[32];
3536 unsigned long value;
3537
3538 if (cnt >= sizeof(str))
3539 return -EINVAL;
3540 if (copy_from_user(str, buf, cnt))
3541 return -EFAULT;
3542 value = simple_strtol(str, NULL, 0);
3543 if (value < 0)
3544 return -ERANGE;
3545 *p = value;
3546 return cnt;
3547 }
3548
3549 const struct file_operations ext4_ui_proc_fops = {
3550 .owner = THIS_MODULE,
3551 .open = ext4_ui_proc_open,
3552 .read = seq_read,
3553 .llseek = seq_lseek,
3554 .release = single_release,
3555 .write = ext4_ui_proc_write,
3556 };
3557 #endif
3558
3559 static struct file_system_type ext4_fs_type = {
3560 .owner = THIS_MODULE,
3561 .name = "ext4",
3562 .get_sb = ext4_get_sb,
3563 .kill_sb = kill_block_super,
3564 .fs_flags = FS_REQUIRES_DEV,
3565 };
3566
3567 #ifdef CONFIG_EXT4DEV_COMPAT
3568 static int ext4dev_get_sb(struct file_system_type *fs_type,
3569 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3570 {
3571 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3572 "to mount using ext4\n");
3573 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3574 "will go away by 2.6.31\n");
3575 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3576 }
3577
3578 static struct file_system_type ext4dev_fs_type = {
3579 .owner = THIS_MODULE,
3580 .name = "ext4dev",
3581 .get_sb = ext4dev_get_sb,
3582 .kill_sb = kill_block_super,
3583 .fs_flags = FS_REQUIRES_DEV,
3584 };
3585 MODULE_ALIAS("ext4dev");
3586 #endif
3587
3588 static int __init init_ext4_fs(void)
3589 {
3590 int err;
3591
3592 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3593 err = init_ext4_mballoc();
3594 if (err)
3595 return err;
3596
3597 err = init_ext4_xattr();
3598 if (err)
3599 goto out2;
3600 err = init_inodecache();
3601 if (err)
3602 goto out1;
3603 err = register_filesystem(&ext4_fs_type);
3604 if (err)
3605 goto out;
3606 #ifdef CONFIG_EXT4DEV_COMPAT
3607 err = register_filesystem(&ext4dev_fs_type);
3608 if (err) {
3609 unregister_filesystem(&ext4_fs_type);
3610 goto out;
3611 }
3612 #endif
3613 return 0;
3614 out:
3615 destroy_inodecache();
3616 out1:
3617 exit_ext4_xattr();
3618 out2:
3619 exit_ext4_mballoc();
3620 return err;
3621 }
3622
3623 static void __exit exit_ext4_fs(void)
3624 {
3625 unregister_filesystem(&ext4_fs_type);
3626 #ifdef CONFIG_EXT4DEV_COMPAT
3627 unregister_filesystem(&ext4dev_fs_type);
3628 #endif
3629 destroy_inodecache();
3630 exit_ext4_xattr();
3631 exit_ext4_mballoc();
3632 remove_proc_entry("fs/ext4", NULL);
3633 }
3634
3635 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3636 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3637 MODULE_LICENSE("GPL");
3638 module_init(init_ext4_fs)
3639 module_exit(exit_ext4_fs)
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree