?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ext4: Enable delalloc by default.
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext4 / super.c
blob4e104dd0ec212124b42da43c5ae08ebe62f663f7
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/log2.h>
38 #include <linux/crc16.h>
39 #include <asm/uaccess.h>
40
41 #include "ext4.h"
42 #include "ext4_jbd2.h"
43 #include "xattr.h"
44 #include "acl.h"
45 #include "namei.h"
46 #include "group.h"
47
48 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
49 unsigned long journal_devnum);
50 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
51 unsigned int);
52 static void ext4_commit_super (struct super_block * sb,
53 struct ext4_super_block * es,
54 int sync);
55 static void ext4_mark_recovery_complete(struct super_block * sb,
56 struct ext4_super_block * es);
57 static void ext4_clear_journal_err(struct super_block * sb,
58 struct ext4_super_block * es);
59 static int ext4_sync_fs(struct super_block *sb, int wait);
60 static const char *ext4_decode_error(struct super_block * sb, int errno,
61 char nbuf[16]);
62 static int ext4_remount (struct super_block * sb, int * flags, char * data);
63 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
64 static void ext4_unlockfs(struct super_block *sb);
65 static void ext4_write_super (struct super_block * sb);
66 static void ext4_write_super_lockfs(struct super_block *sb);
67
68
69 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
70 struct ext4_group_desc *bg)
71 {
72 return le32_to_cpu(bg->bg_block_bitmap_lo) |
73 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
74 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
75 }
76
77 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
78 struct ext4_group_desc *bg)
79 {
80 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
81 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
82 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
83 }
84
85 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
86 struct ext4_group_desc *bg)
87 {
88 return le32_to_cpu(bg->bg_inode_table_lo) |
89 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
90 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
91 }
92
93 void ext4_block_bitmap_set(struct super_block *sb,
94 struct ext4_group_desc *bg, ext4_fsblk_t blk)
95 {
96 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
97 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
98 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
99 }
100
101 void ext4_inode_bitmap_set(struct super_block *sb,
102 struct ext4_group_desc *bg, ext4_fsblk_t blk)
103 {
104 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
105 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
106 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
107 }
108
109 void ext4_inode_table_set(struct super_block *sb,
110 struct ext4_group_desc *bg, ext4_fsblk_t blk)
111 {
112 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
113 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
114 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
115 }
116
117 /*
118 * Wrappers for jbd2_journal_start/end.
119 *
120 * The only special thing we need to do here is to make sure that all
121 * journal_end calls result in the superblock being marked dirty, so
122 * that sync() will call the filesystem's write_super callback if
123 * appropriate.
124 */
125 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
126 {
127 journal_t *journal;
128
129 if (sb->s_flags & MS_RDONLY)
130 return ERR_PTR(-EROFS);
131
132 /* Special case here: if the journal has aborted behind our
133 * backs (eg. EIO in the commit thread), then we still need to
134 * take the FS itself readonly cleanly. */
135 journal = EXT4_SB(sb)->s_journal;
136 if (is_journal_aborted(journal)) {
137 ext4_abort(sb, __func__,
138 "Detected aborted journal");
139 return ERR_PTR(-EROFS);
140 }
141
142 return jbd2_journal_start(journal, nblocks);
143 }
144
145 /*
146 * The only special thing we need to do here is to make sure that all
147 * jbd2_journal_stop calls result in the superblock being marked dirty, so
148 * that sync() will call the filesystem's write_super callback if
149 * appropriate.
150 */
151 int __ext4_journal_stop(const char *where, handle_t *handle)
152 {
153 struct super_block *sb;
154 int err;
155 int rc;
156
157 sb = handle->h_transaction->t_journal->j_private;
158 err = handle->h_err;
159 rc = jbd2_journal_stop(handle);
160
161 if (!err)
162 err = rc;
163 if (err)
164 __ext4_std_error(sb, where, err);
165 return err;
166 }
167
168 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
169 struct buffer_head *bh, handle_t *handle, int err)
170 {
171 char nbuf[16];
172 const char *errstr = ext4_decode_error(NULL, err, nbuf);
173
174 if (bh)
175 BUFFER_TRACE(bh, "abort");
176
177 if (!handle->h_err)
178 handle->h_err = err;
179
180 if (is_handle_aborted(handle))
181 return;
182
183 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
184 caller, errstr, err_fn);
185
186 jbd2_journal_abort_handle(handle);
187 }
188
189 /* Deal with the reporting of failure conditions on a filesystem such as
190 * inconsistencies detected or read IO failures.
191 *
192 * On ext2, we can store the error state of the filesystem in the
193 * superblock. That is not possible on ext4, because we may have other
194 * write ordering constraints on the superblock which prevent us from
195 * writing it out straight away; and given that the journal is about to
196 * be aborted, we can't rely on the current, or future, transactions to
197 * write out the superblock safely.
198 *
199 * We'll just use the jbd2_journal_abort() error code to record an error in
200 * the journal instead. On recovery, the journal will compain about
201 * that error until we've noted it down and cleared it.
202 */
203
204 static void ext4_handle_error(struct super_block *sb)
205 {
206 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
207
208 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
209 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
210
211 if (sb->s_flags & MS_RDONLY)
212 return;
213
214 if (!test_opt (sb, ERRORS_CONT)) {
215 journal_t *journal = EXT4_SB(sb)->s_journal;
216
217 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
218 if (journal)
219 jbd2_journal_abort(journal, -EIO);
220 }
221 if (test_opt (sb, ERRORS_RO)) {
222 printk (KERN_CRIT "Remounting filesystem read-only\n");
223 sb->s_flags |= MS_RDONLY;
224 }
225 ext4_commit_super(sb, es, 1);
226 if (test_opt(sb, ERRORS_PANIC))
227 panic("EXT4-fs (device %s): panic forced after error\n",
228 sb->s_id);
229 }
230
231 void ext4_error (struct super_block * sb, const char * function,
232 const char * fmt, ...)
233 {
234 va_list args;
235
236 va_start(args, fmt);
237 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
238 vprintk(fmt, args);
239 printk("\n");
240 va_end(args);
241
242 ext4_handle_error(sb);
243 }
244
245 static const char *ext4_decode_error(struct super_block * sb, int errno,
246 char nbuf[16])
247 {
248 char *errstr = NULL;
249
250 switch (errno) {
251 case -EIO:
252 errstr = "IO failure";
253 break;
254 case -ENOMEM:
255 errstr = "Out of memory";
256 break;
257 case -EROFS:
258 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
259 errstr = "Journal has aborted";
260 else
261 errstr = "Readonly filesystem";
262 break;
263 default:
264 /* If the caller passed in an extra buffer for unknown
265 * errors, textualise them now. Else we just return
266 * NULL. */
267 if (nbuf) {
268 /* Check for truncated error codes... */
269 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
270 errstr = nbuf;
271 }
272 break;
273 }
274
275 return errstr;
276 }
277
278 /* __ext4_std_error decodes expected errors from journaling functions
279 * automatically and invokes the appropriate error response. */
280
281 void __ext4_std_error (struct super_block * sb, const char * function,
282 int errno)
283 {
284 char nbuf[16];
285 const char *errstr;
286
287 /* Special case: if the error is EROFS, and we're not already
288 * inside a transaction, then there's really no point in logging
289 * an error. */
290 if (errno == -EROFS && journal_current_handle() == NULL &&
291 (sb->s_flags & MS_RDONLY))
292 return;
293
294 errstr = ext4_decode_error(sb, errno, nbuf);
295 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
296 sb->s_id, function, errstr);
297
298 ext4_handle_error(sb);
299 }
300
301 /*
302 * ext4_abort is a much stronger failure handler than ext4_error. The
303 * abort function may be used to deal with unrecoverable failures such
304 * as journal IO errors or ENOMEM at a critical moment in log management.
305 *
306 * We unconditionally force the filesystem into an ABORT|READONLY state,
307 * unless the error response on the fs has been set to panic in which
308 * case we take the easy way out and panic immediately.
309 */
310
311 void ext4_abort (struct super_block * sb, const char * function,
312 const char * fmt, ...)
313 {
314 va_list args;
315
316 printk (KERN_CRIT "ext4_abort called.\n");
317
318 va_start(args, fmt);
319 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
320 vprintk(fmt, args);
321 printk("\n");
322 va_end(args);
323
324 if (test_opt(sb, ERRORS_PANIC))
325 panic("EXT4-fs panic from previous error\n");
326
327 if (sb->s_flags & MS_RDONLY)
328 return;
329
330 printk(KERN_CRIT "Remounting filesystem read-only\n");
331 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
332 sb->s_flags |= MS_RDONLY;
333 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
334 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
335 }
336
337 void ext4_warning (struct super_block * sb, const char * function,
338 const char * fmt, ...)
339 {
340 va_list args;
341
342 va_start(args, fmt);
343 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
344 sb->s_id, function);
345 vprintk(fmt, args);
346 printk("\n");
347 va_end(args);
348 }
349
350 void ext4_update_dynamic_rev(struct super_block *sb)
351 {
352 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
353
354 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
355 return;
356
357 ext4_warning(sb, __func__,
358 "updating to rev %d because of new feature flag, "
359 "running e2fsck is recommended",
360 EXT4_DYNAMIC_REV);
361
362 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
363 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
364 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
365 /* leave es->s_feature_*compat flags alone */
366 /* es->s_uuid will be set by e2fsck if empty */
367
368 /*
369 * The rest of the superblock fields should be zero, and if not it
370 * means they are likely already in use, so leave them alone. We
371 * can leave it up to e2fsck to clean up any inconsistencies there.
372 */
373 }
374
375 int ext4_update_compat_feature(handle_t *handle,
376 struct super_block *sb, __u32 compat)
377 {
378 int err = 0;
379 if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
380 err = ext4_journal_get_write_access(handle,
381 EXT4_SB(sb)->s_sbh);
382 if (err)
383 return err;
384 EXT4_SET_COMPAT_FEATURE(sb, compat);
385 sb->s_dirt = 1;
386 handle->h_sync = 1;
387 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
388 "call ext4_journal_dirty_met adata");
389 err = ext4_journal_dirty_metadata(handle,
390 EXT4_SB(sb)->s_sbh);
391 }
392 return err;
393 }
394
395 int ext4_update_rocompat_feature(handle_t *handle,
396 struct super_block *sb, __u32 rocompat)
397 {
398 int err = 0;
399 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
400 err = ext4_journal_get_write_access(handle,
401 EXT4_SB(sb)->s_sbh);
402 if (err)
403 return err;
404 EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
405 sb->s_dirt = 1;
406 handle->h_sync = 1;
407 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
408 "call ext4_journal_dirty_met adata");
409 err = ext4_journal_dirty_metadata(handle,
410 EXT4_SB(sb)->s_sbh);
411 }
412 return err;
413 }
414
415 int ext4_update_incompat_feature(handle_t *handle,
416 struct super_block *sb, __u32 incompat)
417 {
418 int err = 0;
419 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
420 err = ext4_journal_get_write_access(handle,
421 EXT4_SB(sb)->s_sbh);
422 if (err)
423 return err;
424 EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
425 sb->s_dirt = 1;
426 handle->h_sync = 1;
427 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
428 "call ext4_journal_dirty_met adata");
429 err = ext4_journal_dirty_metadata(handle,
430 EXT4_SB(sb)->s_sbh);
431 }
432 return err;
433 }
434
435 /*
436 * Open the external journal device
437 */
438 static struct block_device *ext4_blkdev_get(dev_t dev)
439 {
440 struct block_device *bdev;
441 char b[BDEVNAME_SIZE];
442
443 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
444 if (IS_ERR(bdev))
445 goto fail;
446 return bdev;
447
448 fail:
449 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
450 __bdevname(dev, b), PTR_ERR(bdev));
451 return NULL;
452 }
453
454 /*
455 * Release the journal device
456 */
457 static int ext4_blkdev_put(struct block_device *bdev)
458 {
459 bd_release(bdev);
460 return blkdev_put(bdev);
461 }
462
463 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
464 {
465 struct block_device *bdev;
466 int ret = -ENODEV;
467
468 bdev = sbi->journal_bdev;
469 if (bdev) {
470 ret = ext4_blkdev_put(bdev);
471 sbi->journal_bdev = NULL;
472 }
473 return ret;
474 }
475
476 static inline struct inode *orphan_list_entry(struct list_head *l)
477 {
478 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
479 }
480
481 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
482 {
483 struct list_head *l;
484
485 printk(KERN_ERR "sb orphan head is %d\n",
486 le32_to_cpu(sbi->s_es->s_last_orphan));
487
488 printk(KERN_ERR "sb_info orphan list:\n");
489 list_for_each(l, &sbi->s_orphan) {
490 struct inode *inode = orphan_list_entry(l);
491 printk(KERN_ERR " "
492 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
493 inode->i_sb->s_id, inode->i_ino, inode,
494 inode->i_mode, inode->i_nlink,
495 NEXT_ORPHAN(inode));
496 }
497 }
498
499 static void ext4_put_super (struct super_block * sb)
500 {
501 struct ext4_sb_info *sbi = EXT4_SB(sb);
502 struct ext4_super_block *es = sbi->s_es;
503 int i;
504
505 ext4_mb_release(sb);
506 ext4_ext_release(sb);
507 ext4_xattr_put_super(sb);
508 jbd2_journal_destroy(sbi->s_journal);
509 sbi->s_journal = NULL;
510 if (!(sb->s_flags & MS_RDONLY)) {
511 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
512 es->s_state = cpu_to_le16(sbi->s_mount_state);
513 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
514 mark_buffer_dirty(sbi->s_sbh);
515 ext4_commit_super(sb, es, 1);
516 }
517
518 for (i = 0; i < sbi->s_gdb_count; i++)
519 brelse(sbi->s_group_desc[i]);
520 kfree(sbi->s_group_desc);
521 kfree(sbi->s_flex_groups);
522 percpu_counter_destroy(&sbi->s_freeblocks_counter);
523 percpu_counter_destroy(&sbi->s_freeinodes_counter);
524 percpu_counter_destroy(&sbi->s_dirs_counter);
525 brelse(sbi->s_sbh);
526 #ifdef CONFIG_QUOTA
527 for (i = 0; i < MAXQUOTAS; i++)
528 kfree(sbi->s_qf_names[i]);
529 #endif
530
531 /* Debugging code just in case the in-memory inode orphan list
532 * isn't empty. The on-disk one can be non-empty if we've
533 * detected an error and taken the fs readonly, but the
534 * in-memory list had better be clean by this point. */
535 if (!list_empty(&sbi->s_orphan))
536 dump_orphan_list(sb, sbi);
537 J_ASSERT(list_empty(&sbi->s_orphan));
538
539 invalidate_bdev(sb->s_bdev);
540 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
541 /*
542 * Invalidate the journal device's buffers. We don't want them
543 * floating about in memory - the physical journal device may
544 * hotswapped, and it breaks the `ro-after' testing code.
545 */
546 sync_blockdev(sbi->journal_bdev);
547 invalidate_bdev(sbi->journal_bdev);
548 ext4_blkdev_remove(sbi);
549 }
550 sb->s_fs_info = NULL;
551 kfree(sbi);
552 return;
553 }
554
555 static struct kmem_cache *ext4_inode_cachep;
556
557 /*
558 * Called inside transaction, so use GFP_NOFS
559 */
560 static struct inode *ext4_alloc_inode(struct super_block *sb)
561 {
562 struct ext4_inode_info *ei;
563
564 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
565 if (!ei)
566 return NULL;
567 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
568 ei->i_acl = EXT4_ACL_NOT_CACHED;
569 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
570 #endif
571 ei->i_block_alloc_info = NULL;
572 ei->vfs_inode.i_version = 1;
573 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
574 INIT_LIST_HEAD(&ei->i_prealloc_list);
575 spin_lock_init(&ei->i_prealloc_lock);
576 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
577 ei->i_reserved_data_blocks = 0;
578 ei->i_reserved_meta_blocks = 0;
579 ei->i_allocated_meta_blocks = 0;
580 ei->i_delalloc_reserved_flag = 0;
581 spin_lock_init(&(ei->i_block_reservation_lock));
582 return &ei->vfs_inode;
583 }
584
585 static void ext4_destroy_inode(struct inode *inode)
586 {
587 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
588 printk("EXT4 Inode %p: orphan list check failed!\n",
589 EXT4_I(inode));
590 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
591 EXT4_I(inode), sizeof(struct ext4_inode_info),
592 true);
593 dump_stack();
594 }
595 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
596 }
597
598 static void init_once(struct kmem_cache *cachep, void *foo)
599 {
600 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
601
602 INIT_LIST_HEAD(&ei->i_orphan);
603 #ifdef CONFIG_EXT4DEV_FS_XATTR
604 init_rwsem(&ei->xattr_sem);
605 #endif
606 init_rwsem(&ei->i_data_sem);
607 inode_init_once(&ei->vfs_inode);
608 }
609
610 static int init_inodecache(void)
611 {
612 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
613 sizeof(struct ext4_inode_info),
614 0, (SLAB_RECLAIM_ACCOUNT|
615 SLAB_MEM_SPREAD),
616 init_once);
617 if (ext4_inode_cachep == NULL)
618 return -ENOMEM;
619 return 0;
620 }
621
622 static void destroy_inodecache(void)
623 {
624 kmem_cache_destroy(ext4_inode_cachep);
625 }
626
627 static void ext4_clear_inode(struct inode *inode)
628 {
629 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
630 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
631 if (EXT4_I(inode)->i_acl &&
632 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
633 posix_acl_release(EXT4_I(inode)->i_acl);
634 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
635 }
636 if (EXT4_I(inode)->i_default_acl &&
637 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
638 posix_acl_release(EXT4_I(inode)->i_default_acl);
639 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
640 }
641 #endif
642 ext4_discard_reservation(inode);
643 EXT4_I(inode)->i_block_alloc_info = NULL;
644 if (unlikely(rsv))
645 kfree(rsv);
646 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
647 &EXT4_I(inode)->jinode);
648 }
649
650 static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
651 {
652 #if defined(CONFIG_QUOTA)
653 struct ext4_sb_info *sbi = EXT4_SB(sb);
654
655 if (sbi->s_jquota_fmt)
656 seq_printf(seq, ",jqfmt=%s",
657 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
658
659 if (sbi->s_qf_names[USRQUOTA])
660 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
661
662 if (sbi->s_qf_names[GRPQUOTA])
663 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
664
665 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
666 seq_puts(seq, ",usrquota");
667
668 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
669 seq_puts(seq, ",grpquota");
670 #endif
671 }
672
673 /*
674 * Show an option if
675 * - it's set to a non-default value OR
676 * - if the per-sb default is different from the global default
677 */
678 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
679 {
680 int def_errors;
681 unsigned long def_mount_opts;
682 struct super_block *sb = vfs->mnt_sb;
683 struct ext4_sb_info *sbi = EXT4_SB(sb);
684 struct ext4_super_block *es = sbi->s_es;
685
686 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
687 def_errors = le16_to_cpu(es->s_errors);
688
689 if (sbi->s_sb_block != 1)
690 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
691 if (test_opt(sb, MINIX_DF))
692 seq_puts(seq, ",minixdf");
693 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
694 seq_puts(seq, ",grpid");
695 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
696 seq_puts(seq, ",nogrpid");
697 if (sbi->s_resuid != EXT4_DEF_RESUID ||
698 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
699 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
700 }
701 if (sbi->s_resgid != EXT4_DEF_RESGID ||
702 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
703 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
704 }
705 if (test_opt(sb, ERRORS_RO)) {
706 if (def_errors == EXT4_ERRORS_PANIC ||
707 def_errors == EXT4_ERRORS_CONTINUE) {
708 seq_puts(seq, ",errors=remount-ro");
709 }
710 }
711 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
712 seq_puts(seq, ",errors=continue");
713 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
714 seq_puts(seq, ",errors=panic");
715 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
716 seq_puts(seq, ",nouid32");
717 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
718 seq_puts(seq, ",debug");
719 if (test_opt(sb, OLDALLOC))
720 seq_puts(seq, ",oldalloc");
721 #ifdef CONFIG_EXT4DEV_FS_XATTR
722 if (test_opt(sb, XATTR_USER) &&
723 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
724 seq_puts(seq, ",user_xattr");
725 if (!test_opt(sb, XATTR_USER) &&
726 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
727 seq_puts(seq, ",nouser_xattr");
728 }
729 #endif
730 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
731 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
732 seq_puts(seq, ",acl");
733 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
734 seq_puts(seq, ",noacl");
735 #endif
736 if (!test_opt(sb, RESERVATION))
737 seq_puts(seq, ",noreservation");
738 if (sbi->s_commit_interval) {
739 seq_printf(seq, ",commit=%u",
740 (unsigned) (sbi->s_commit_interval / HZ));
741 }
742 /*
743 * We're changing the default of barrier mount option, so
744 * let's always display its mount state so it's clear what its
745 * status is.
746 */
747 seq_puts(seq, ",barrier=");
748 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
749 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
750 seq_puts(seq, ",journal_async_commit");
751 if (test_opt(sb, NOBH))
752 seq_puts(seq, ",nobh");
753 if (!test_opt(sb, EXTENTS))
754 seq_puts(seq, ",noextents");
755 if (!test_opt(sb, MBALLOC))
756 seq_puts(seq, ",nomballoc");
757 if (test_opt(sb, I_VERSION))
758 seq_puts(seq, ",i_version");
759 if (!test_opt(sb, DELALLOC))
760 seq_puts(seq, ",nodelalloc");
761
762
763 if (sbi->s_stripe)
764 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
765 /*
766 * journal mode get enabled in different ways
767 * So just print the value even if we didn't specify it
768 */
769 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
770 seq_puts(seq, ",data=journal");
771 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
772 seq_puts(seq, ",data=ordered");
773 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
774 seq_puts(seq, ",data=writeback");
775
776 ext4_show_quota_options(seq, sb);
777 return 0;
778 }
779
780
781 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
782 u64 ino, u32 generation)
783 {
784 struct inode *inode;
785
786 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
787 return ERR_PTR(-ESTALE);
788 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
789 return ERR_PTR(-ESTALE);
790
791 /* iget isn't really right if the inode is currently unallocated!!
792 *
793 * ext4_read_inode will return a bad_inode if the inode had been
794 * deleted, so we should be safe.
795 *
796 * Currently we don't know the generation for parent directory, so
797 * a generation of 0 means "accept any"
798 */
799 inode = ext4_iget(sb, ino);
800 if (IS_ERR(inode))
801 return ERR_CAST(inode);
802 if (generation && inode->i_generation != generation) {
803 iput(inode);
804 return ERR_PTR(-ESTALE);
805 }
806
807 return inode;
808 }
809
810 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
811 int fh_len, int fh_type)
812 {
813 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
814 ext4_nfs_get_inode);
815 }
816
817 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
818 int fh_len, int fh_type)
819 {
820 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
821 ext4_nfs_get_inode);
822 }
823
824 #ifdef CONFIG_QUOTA
825 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
826 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
827
828 static int ext4_dquot_initialize(struct inode *inode, int type);
829 static int ext4_dquot_drop(struct inode *inode);
830 static int ext4_write_dquot(struct dquot *dquot);
831 static int ext4_acquire_dquot(struct dquot *dquot);
832 static int ext4_release_dquot(struct dquot *dquot);
833 static int ext4_mark_dquot_dirty(struct dquot *dquot);
834 static int ext4_write_info(struct super_block *sb, int type);
835 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
836 char *path, int remount);
837 static int ext4_quota_on_mount(struct super_block *sb, int type);
838 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
839 size_t len, loff_t off);
840 static ssize_t ext4_quota_write(struct super_block *sb, int type,
841 const char *data, size_t len, loff_t off);
842
843 static struct dquot_operations ext4_quota_operations = {
844 .initialize = ext4_dquot_initialize,
845 .drop = ext4_dquot_drop,
846 .alloc_space = dquot_alloc_space,
847 .alloc_inode = dquot_alloc_inode,
848 .free_space = dquot_free_space,
849 .free_inode = dquot_free_inode,
850 .transfer = dquot_transfer,
851 .write_dquot = ext4_write_dquot,
852 .acquire_dquot = ext4_acquire_dquot,
853 .release_dquot = ext4_release_dquot,
854 .mark_dirty = ext4_mark_dquot_dirty,
855 .write_info = ext4_write_info
856 };
857
858 static struct quotactl_ops ext4_qctl_operations = {
859 .quota_on = ext4_quota_on,
860 .quota_off = vfs_quota_off,
861 .quota_sync = vfs_quota_sync,
862 .get_info = vfs_get_dqinfo,
863 .set_info = vfs_set_dqinfo,
864 .get_dqblk = vfs_get_dqblk,
865 .set_dqblk = vfs_set_dqblk
866 };
867 #endif
868
869 static const struct super_operations ext4_sops = {
870 .alloc_inode = ext4_alloc_inode,
871 .destroy_inode = ext4_destroy_inode,
872 .write_inode = ext4_write_inode,
873 .dirty_inode = ext4_dirty_inode,
874 .delete_inode = ext4_delete_inode,
875 .put_super = ext4_put_super,
876 .write_super = ext4_write_super,
877 .sync_fs = ext4_sync_fs,
878 .write_super_lockfs = ext4_write_super_lockfs,
879 .unlockfs = ext4_unlockfs,
880 .statfs = ext4_statfs,
881 .remount_fs = ext4_remount,
882 .clear_inode = ext4_clear_inode,
883 .show_options = ext4_show_options,
884 #ifdef CONFIG_QUOTA
885 .quota_read = ext4_quota_read,
886 .quota_write = ext4_quota_write,
887 #endif
888 };
889
890 static const struct export_operations ext4_export_ops = {
891 .fh_to_dentry = ext4_fh_to_dentry,
892 .fh_to_parent = ext4_fh_to_parent,
893 .get_parent = ext4_get_parent,
894 };
895
896 enum {
897 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
898 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
899 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
900 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
901 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
902 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
903 Opt_journal_checksum, Opt_journal_async_commit,
904 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
905 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
906 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
907 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
908 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
909 Opt_mballoc, Opt_nomballoc, Opt_stripe, Opt_delalloc, Opt_nodelalloc,
910 };
911
912 static match_table_t tokens = {
913 {Opt_bsd_df, "bsddf"},
914 {Opt_minix_df, "minixdf"},
915 {Opt_grpid, "grpid"},
916 {Opt_grpid, "bsdgroups"},
917 {Opt_nogrpid, "nogrpid"},
918 {Opt_nogrpid, "sysvgroups"},
919 {Opt_resgid, "resgid=%u"},
920 {Opt_resuid, "resuid=%u"},
921 {Opt_sb, "sb=%u"},
922 {Opt_err_cont, "errors=continue"},
923 {Opt_err_panic, "errors=panic"},
924 {Opt_err_ro, "errors=remount-ro"},
925 {Opt_nouid32, "nouid32"},
926 {Opt_nocheck, "nocheck"},
927 {Opt_nocheck, "check=none"},
928 {Opt_debug, "debug"},
929 {Opt_oldalloc, "oldalloc"},
930 {Opt_orlov, "orlov"},
931 {Opt_user_xattr, "user_xattr"},
932 {Opt_nouser_xattr, "nouser_xattr"},
933 {Opt_acl, "acl"},
934 {Opt_noacl, "noacl"},
935 {Opt_reservation, "reservation"},
936 {Opt_noreservation, "noreservation"},
937 {Opt_noload, "noload"},
938 {Opt_nobh, "nobh"},
939 {Opt_bh, "bh"},
940 {Opt_commit, "commit=%u"},
941 {Opt_journal_update, "journal=update"},
942 {Opt_journal_inum, "journal=%u"},
943 {Opt_journal_dev, "journal_dev=%u"},
944 {Opt_journal_checksum, "journal_checksum"},
945 {Opt_journal_async_commit, "journal_async_commit"},
946 {Opt_abort, "abort"},
947 {Opt_data_journal, "data=journal"},
948 {Opt_data_ordered, "data=ordered"},
949 {Opt_data_writeback, "data=writeback"},
950 {Opt_offusrjquota, "usrjquota="},
951 {Opt_usrjquota, "usrjquota=%s"},
952 {Opt_offgrpjquota, "grpjquota="},
953 {Opt_grpjquota, "grpjquota=%s"},
954 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
955 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
956 {Opt_grpquota, "grpquota"},
957 {Opt_noquota, "noquota"},
958 {Opt_quota, "quota"},
959 {Opt_usrquota, "usrquota"},
960 {Opt_barrier, "barrier=%u"},
961 {Opt_extents, "extents"},
962 {Opt_noextents, "noextents"},
963 {Opt_i_version, "i_version"},
964 {Opt_mballoc, "mballoc"},
965 {Opt_nomballoc, "nomballoc"},
966 {Opt_stripe, "stripe=%u"},
967 {Opt_resize, "resize"},
968 {Opt_delalloc, "delalloc"},
969 {Opt_nodelalloc, "nodelalloc"},
970 {Opt_err, NULL},
971 };
972
973 static ext4_fsblk_t get_sb_block(void **data)
974 {
975 ext4_fsblk_t sb_block;
976 char *options = (char *) *data;
977
978 if (!options || strncmp(options, "sb=", 3) != 0)
979 return 1; /* Default location */
980 options += 3;
981 /*todo: use simple_strtoll with >32bit ext4 */
982 sb_block = simple_strtoul(options, &options, 0);
983 if (*options && *options != ',') {
984 printk("EXT4-fs: Invalid sb specification: %s\n",
985 (char *) *data);
986 return 1;
987 }
988 if (*options == ',')
989 options++;
990 *data = (void *) options;
991 return sb_block;
992 }
993
994 static int parse_options (char *options, struct super_block *sb,
995 unsigned int *inum, unsigned long *journal_devnum,
996 ext4_fsblk_t *n_blocks_count, int is_remount)
997 {
998 struct ext4_sb_info *sbi = EXT4_SB(sb);
999 char * p;
1000 substring_t args[MAX_OPT_ARGS];
1001 int data_opt = 0;
1002 int option;
1003 #ifdef CONFIG_QUOTA
1004 int qtype, qfmt;
1005 char *qname;
1006 #endif
1007
1008 if (!options)
1009 return 1;
1010
1011 while ((p = strsep (&options, ",")) != NULL) {
1012 int token;
1013 if (!*p)
1014 continue;
1015
1016 token = match_token(p, tokens, args);
1017 switch (token) {
1018 case Opt_bsd_df:
1019 clear_opt (sbi->s_mount_opt, MINIX_DF);
1020 break;
1021 case Opt_minix_df:
1022 set_opt (sbi->s_mount_opt, MINIX_DF);
1023 break;
1024 case Opt_grpid:
1025 set_opt (sbi->s_mount_opt, GRPID);
1026 break;
1027 case Opt_nogrpid:
1028 clear_opt (sbi->s_mount_opt, GRPID);
1029 break;
1030 case Opt_resuid:
1031 if (match_int(&args[0], &option))
1032 return 0;
1033 sbi->s_resuid = option;
1034 break;
1035 case Opt_resgid:
1036 if (match_int(&args[0], &option))
1037 return 0;
1038 sbi->s_resgid = option;
1039 break;
1040 case Opt_sb:
1041 /* handled by get_sb_block() instead of here */
1042 /* *sb_block = match_int(&args[0]); */
1043 break;
1044 case Opt_err_panic:
1045 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1046 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1047 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1048 break;
1049 case Opt_err_ro:
1050 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1051 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1052 set_opt (sbi->s_mount_opt, ERRORS_RO);
1053 break;
1054 case Opt_err_cont:
1055 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1056 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1057 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1058 break;
1059 case Opt_nouid32:
1060 set_opt (sbi->s_mount_opt, NO_UID32);
1061 break;
1062 case Opt_nocheck:
1063 clear_opt (sbi->s_mount_opt, CHECK);
1064 break;
1065 case Opt_debug:
1066 set_opt (sbi->s_mount_opt, DEBUG);
1067 break;
1068 case Opt_oldalloc:
1069 set_opt (sbi->s_mount_opt, OLDALLOC);
1070 break;
1071 case Opt_orlov:
1072 clear_opt (sbi->s_mount_opt, OLDALLOC);
1073 break;
1074 #ifdef CONFIG_EXT4DEV_FS_XATTR
1075 case Opt_user_xattr:
1076 set_opt (sbi->s_mount_opt, XATTR_USER);
1077 break;
1078 case Opt_nouser_xattr:
1079 clear_opt (sbi->s_mount_opt, XATTR_USER);
1080 break;
1081 #else
1082 case Opt_user_xattr:
1083 case Opt_nouser_xattr:
1084 printk("EXT4 (no)user_xattr options not supported\n");
1085 break;
1086 #endif
1087 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1088 case Opt_acl:
1089 set_opt(sbi->s_mount_opt, POSIX_ACL);
1090 break;
1091 case Opt_noacl:
1092 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1093 break;
1094 #else
1095 case Opt_acl:
1096 case Opt_noacl:
1097 printk("EXT4 (no)acl options not supported\n");
1098 break;
1099 #endif
1100 case Opt_reservation:
1101 set_opt(sbi->s_mount_opt, RESERVATION);
1102 break;
1103 case Opt_noreservation:
1104 clear_opt(sbi->s_mount_opt, RESERVATION);
1105 break;
1106 case Opt_journal_update:
1107 /* @@@ FIXME */
1108 /* Eventually we will want to be able to create
1109 a journal file here. For now, only allow the
1110 user to specify an existing inode to be the
1111 journal file. */
1112 if (is_remount) {
1113 printk(KERN_ERR "EXT4-fs: cannot specify "
1114 "journal on remount\n");
1115 return 0;
1116 }
1117 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1118 break;
1119 case Opt_journal_inum:
1120 if (is_remount) {
1121 printk(KERN_ERR "EXT4-fs: cannot specify "
1122 "journal on remount\n");
1123 return 0;
1124 }
1125 if (match_int(&args[0], &option))
1126 return 0;
1127 *inum = option;
1128 break;
1129 case Opt_journal_dev:
1130 if (is_remount) {
1131 printk(KERN_ERR "EXT4-fs: cannot specify "
1132 "journal on remount\n");
1133 return 0;
1134 }
1135 if (match_int(&args[0], &option))
1136 return 0;
1137 *journal_devnum = option;
1138 break;
1139 case Opt_journal_checksum:
1140 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1141 break;
1142 case Opt_journal_async_commit:
1143 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1144 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1145 break;
1146 case Opt_noload:
1147 set_opt (sbi->s_mount_opt, NOLOAD);
1148 break;
1149 case Opt_commit:
1150 if (match_int(&args[0], &option))
1151 return 0;
1152 if (option < 0)
1153 return 0;
1154 if (option == 0)
1155 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1156 sbi->s_commit_interval = HZ * option;
1157 break;
1158 case Opt_data_journal:
1159 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1160 goto datacheck;
1161 case Opt_data_ordered:
1162 data_opt = EXT4_MOUNT_ORDERED_DATA;
1163 goto datacheck;
1164 case Opt_data_writeback:
1165 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1166 datacheck:
1167 if (is_remount) {
1168 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1169 != data_opt) {
1170 printk(KERN_ERR
1171 "EXT4-fs: cannot change data "
1172 "mode on remount\n");
1173 return 0;
1174 }
1175 } else {
1176 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1177 sbi->s_mount_opt |= data_opt;
1178 }
1179 break;
1180 #ifdef CONFIG_QUOTA
1181 case Opt_usrjquota:
1182 qtype = USRQUOTA;
1183 goto set_qf_name;
1184 case Opt_grpjquota:
1185 qtype = GRPQUOTA;
1186 set_qf_name:
1187 if ((sb_any_quota_enabled(sb) ||
1188 sb_any_quota_suspended(sb)) &&
1189 !sbi->s_qf_names[qtype]) {
1190 printk(KERN_ERR
1191 "EXT4-fs: Cannot change journaled "
1192 "quota options when quota turned on.\n");
1193 return 0;
1194 }
1195 qname = match_strdup(&args[0]);
1196 if (!qname) {
1197 printk(KERN_ERR
1198 "EXT4-fs: not enough memory for "
1199 "storing quotafile name.\n");
1200 return 0;
1201 }
1202 if (sbi->s_qf_names[qtype] &&
1203 strcmp(sbi->s_qf_names[qtype], qname)) {
1204 printk(KERN_ERR
1205 "EXT4-fs: %s quota file already "
1206 "specified.\n", QTYPE2NAME(qtype));
1207 kfree(qname);
1208 return 0;
1209 }
1210 sbi->s_qf_names[qtype] = qname;
1211 if (strchr(sbi->s_qf_names[qtype], '/')) {
1212 printk(KERN_ERR
1213 "EXT4-fs: quotafile must be on "
1214 "filesystem root.\n");
1215 kfree(sbi->s_qf_names[qtype]);
1216 sbi->s_qf_names[qtype] = NULL;
1217 return 0;
1218 }
1219 set_opt(sbi->s_mount_opt, QUOTA);
1220 break;
1221 case Opt_offusrjquota:
1222 qtype = USRQUOTA;
1223 goto clear_qf_name;
1224 case Opt_offgrpjquota:
1225 qtype = GRPQUOTA;
1226 clear_qf_name:
1227 if ((sb_any_quota_enabled(sb) ||
1228 sb_any_quota_suspended(sb)) &&
1229 sbi->s_qf_names[qtype]) {
1230 printk(KERN_ERR "EXT4-fs: Cannot change "
1231 "journaled quota options when "
1232 "quota turned on.\n");
1233 return 0;
1234 }
1235 /*
1236 * The space will be released later when all options
1237 * are confirmed to be correct
1238 */
1239 sbi->s_qf_names[qtype] = NULL;
1240 break;
1241 case Opt_jqfmt_vfsold:
1242 qfmt = QFMT_VFS_OLD;
1243 goto set_qf_format;
1244 case Opt_jqfmt_vfsv0:
1245 qfmt = QFMT_VFS_V0;
1246 set_qf_format:
1247 if ((sb_any_quota_enabled(sb) ||
1248 sb_any_quota_suspended(sb)) &&
1249 sbi->s_jquota_fmt != qfmt) {
1250 printk(KERN_ERR "EXT4-fs: Cannot change "
1251 "journaled quota options when "
1252 "quota turned on.\n");
1253 return 0;
1254 }
1255 sbi->s_jquota_fmt = qfmt;
1256 break;
1257 case Opt_quota:
1258 case Opt_usrquota:
1259 set_opt(sbi->s_mount_opt, QUOTA);
1260 set_opt(sbi->s_mount_opt, USRQUOTA);
1261 break;
1262 case Opt_grpquota:
1263 set_opt(sbi->s_mount_opt, QUOTA);
1264 set_opt(sbi->s_mount_opt, GRPQUOTA);
1265 break;
1266 case Opt_noquota:
1267 if (sb_any_quota_enabled(sb)) {
1268 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1269 "options when quota turned on.\n");
1270 return 0;
1271 }
1272 clear_opt(sbi->s_mount_opt, QUOTA);
1273 clear_opt(sbi->s_mount_opt, USRQUOTA);
1274 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1275 break;
1276 #else
1277 case Opt_quota:
1278 case Opt_usrquota:
1279 case Opt_grpquota:
1280 printk(KERN_ERR
1281 "EXT4-fs: quota options not supported.\n");
1282 break;
1283 case Opt_usrjquota:
1284 case Opt_grpjquota:
1285 case Opt_offusrjquota:
1286 case Opt_offgrpjquota:
1287 case Opt_jqfmt_vfsold:
1288 case Opt_jqfmt_vfsv0:
1289 printk(KERN_ERR
1290 "EXT4-fs: journaled quota options not "
1291 "supported.\n");
1292 break;
1293 case Opt_noquota:
1294 break;
1295 #endif
1296 case Opt_abort:
1297 set_opt(sbi->s_mount_opt, ABORT);
1298 break;
1299 case Opt_barrier:
1300 if (match_int(&args[0], &option))
1301 return 0;
1302 if (option)
1303 set_opt(sbi->s_mount_opt, BARRIER);
1304 else
1305 clear_opt(sbi->s_mount_opt, BARRIER);
1306 break;
1307 case Opt_ignore:
1308 break;
1309 case Opt_resize:
1310 if (!is_remount) {
1311 printk("EXT4-fs: resize option only available "
1312 "for remount\n");
1313 return 0;
1314 }
1315 if (match_int(&args[0], &option) != 0)
1316 return 0;
1317 *n_blocks_count = option;
1318 break;
1319 case Opt_nobh:
1320 set_opt(sbi->s_mount_opt, NOBH);
1321 break;
1322 case Opt_bh:
1323 clear_opt(sbi->s_mount_opt, NOBH);
1324 break;
1325 case Opt_extents:
1326 set_opt (sbi->s_mount_opt, EXTENTS);
1327 break;
1328 case Opt_noextents:
1329 clear_opt (sbi->s_mount_opt, EXTENTS);
1330 break;
1331 case Opt_i_version:
1332 set_opt(sbi->s_mount_opt, I_VERSION);
1333 sb->s_flags |= MS_I_VERSION;
1334 break;
1335 case Opt_nodelalloc:
1336 clear_opt(sbi->s_mount_opt, DELALLOC);
1337 break;
1338 case Opt_mballoc:
1339 set_opt(sbi->s_mount_opt, MBALLOC);
1340 break;
1341 case Opt_nomballoc:
1342 clear_opt(sbi->s_mount_opt, MBALLOC);
1343 break;
1344 case Opt_stripe:
1345 if (match_int(&args[0], &option))
1346 return 0;
1347 if (option < 0)
1348 return 0;
1349 sbi->s_stripe = option;
1350 break;
1351 case Opt_delalloc:
1352 set_opt(sbi->s_mount_opt, DELALLOC);
1353 break;
1354 default:
1355 printk (KERN_ERR
1356 "EXT4-fs: Unrecognized mount option \"%s\" "
1357 "or missing value\n", p);
1358 return 0;
1359 }
1360 }
1361 #ifdef CONFIG_QUOTA
1362 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1363 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1364 sbi->s_qf_names[USRQUOTA])
1365 clear_opt(sbi->s_mount_opt, USRQUOTA);
1366
1367 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1368 sbi->s_qf_names[GRPQUOTA])
1369 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1370
1371 if ((sbi->s_qf_names[USRQUOTA] &&
1372 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1373 (sbi->s_qf_names[GRPQUOTA] &&
1374 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1375 printk(KERN_ERR "EXT4-fs: old and new quota "
1376 "format mixing.\n");
1377 return 0;
1378 }
1379
1380 if (!sbi->s_jquota_fmt) {
1381 printk(KERN_ERR "EXT4-fs: journaled quota format "
1382 "not specified.\n");
1383 return 0;
1384 }
1385 } else {
1386 if (sbi->s_jquota_fmt) {
1387 printk(KERN_ERR "EXT4-fs: journaled quota format "
1388 "specified with no journaling "
1389 "enabled.\n");
1390 return 0;
1391 }
1392 }
1393 #endif
1394 return 1;
1395 }
1396
1397 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1398 int read_only)
1399 {
1400 struct ext4_sb_info *sbi = EXT4_SB(sb);
1401 int res = 0;
1402
1403 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1404 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1405 "forcing read-only mode\n");
1406 res = MS_RDONLY;
1407 }
1408 if (read_only)
1409 return res;
1410 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1411 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1412 "running e2fsck is recommended\n");
1413 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1414 printk (KERN_WARNING
1415 "EXT4-fs warning: mounting fs with errors, "
1416 "running e2fsck is recommended\n");
1417 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1418 le16_to_cpu(es->s_mnt_count) >=
1419 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1420 printk (KERN_WARNING
1421 "EXT4-fs warning: maximal mount count reached, "
1422 "running e2fsck is recommended\n");
1423 else if (le32_to_cpu(es->s_checkinterval) &&
1424 (le32_to_cpu(es->s_lastcheck) +
1425 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1426 printk (KERN_WARNING
1427 "EXT4-fs warning: checktime reached, "
1428 "running e2fsck is recommended\n");
1429 #if 0
1430 /* @@@ We _will_ want to clear the valid bit if we find
1431 * inconsistencies, to force a fsck at reboot. But for
1432 * a plain journaled filesystem we can keep it set as
1433 * valid forever! :)
1434 */
1435 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1436 #endif
1437 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1438 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1439 le16_add_cpu(&es->s_mnt_count, 1);
1440 es->s_mtime = cpu_to_le32(get_seconds());
1441 ext4_update_dynamic_rev(sb);
1442 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1443
1444 ext4_commit_super(sb, es, 1);
1445 if (test_opt(sb, DEBUG))
1446 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1447 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1448 sb->s_blocksize,
1449 sbi->s_groups_count,
1450 EXT4_BLOCKS_PER_GROUP(sb),
1451 EXT4_INODES_PER_GROUP(sb),
1452 sbi->s_mount_opt);
1453
1454 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1455 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1456 char b[BDEVNAME_SIZE];
1457
1458 printk("external journal on %s\n",
1459 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1460 } else {
1461 printk("internal journal\n");
1462 }
1463 return res;
1464 }
1465
1466 static int ext4_fill_flex_info(struct super_block *sb)
1467 {
1468 struct ext4_sb_info *sbi = EXT4_SB(sb);
1469 struct ext4_group_desc *gdp = NULL;
1470 struct buffer_head *bh;
1471 ext4_group_t flex_group_count;
1472 ext4_group_t flex_group;
1473 int groups_per_flex = 0;
1474 __u64 block_bitmap = 0;
1475 int i;
1476
1477 if (!sbi->s_es->s_log_groups_per_flex) {
1478 sbi->s_log_groups_per_flex = 0;
1479 return 1;
1480 }
1481
1482 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1483 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1484
1485 flex_group_count = (sbi->s_groups_count + groups_per_flex - 1) /
1486 groups_per_flex;
1487 sbi->s_flex_groups = kmalloc(flex_group_count *
1488 sizeof(struct flex_groups), GFP_KERNEL);
1489 if (sbi->s_flex_groups == NULL) {
1490 printk(KERN_ERR "EXT4-fs: not enough memory\n");
1491 goto failed;
1492 }
1493 memset(sbi->s_flex_groups, 0, flex_group_count *
1494 sizeof(struct flex_groups));
1495
1496 gdp = ext4_get_group_desc(sb, 1, &bh);
1497 block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1498
1499 for (i = 0; i < sbi->s_groups_count; i++) {
1500 gdp = ext4_get_group_desc(sb, i, &bh);
1501
1502 flex_group = ext4_flex_group(sbi, i);
1503 sbi->s_flex_groups[flex_group].free_inodes +=
1504 le16_to_cpu(gdp->bg_free_inodes_count);
1505 sbi->s_flex_groups[flex_group].free_blocks +=
1506 le16_to_cpu(gdp->bg_free_blocks_count);
1507 }
1508
1509 return 1;
1510 failed:
1511 return 0;
1512 }
1513
1514 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1515 struct ext4_group_desc *gdp)
1516 {
1517 __u16 crc = 0;
1518
1519 if (sbi->s_es->s_feature_ro_compat &
1520 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1521 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1522 __le32 le_group = cpu_to_le32(block_group);
1523
1524 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1525 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1526 crc = crc16(crc, (__u8 *)gdp, offset);
1527 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1528 /* for checksum of struct ext4_group_desc do the rest...*/
1529 if ((sbi->s_es->s_feature_incompat &
1530 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1531 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1532 crc = crc16(crc, (__u8 *)gdp + offset,
1533 le16_to_cpu(sbi->s_es->s_desc_size) -
1534 offset);
1535 }
1536
1537 return cpu_to_le16(crc);
1538 }
1539
1540 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1541 struct ext4_group_desc *gdp)
1542 {
1543 if ((sbi->s_es->s_feature_ro_compat &
1544 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1545 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1546 return 0;
1547
1548 return 1;
1549 }
1550
1551 /* Called at mount-time, super-block is locked */
1552 static int ext4_check_descriptors(struct super_block *sb)
1553 {
1554 struct ext4_sb_info *sbi = EXT4_SB(sb);
1555 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1556 ext4_fsblk_t last_block;
1557 ext4_fsblk_t block_bitmap;
1558 ext4_fsblk_t inode_bitmap;
1559 ext4_fsblk_t inode_table;
1560 int flexbg_flag = 0;
1561 ext4_group_t i;
1562
1563 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1564 flexbg_flag = 1;
1565
1566 ext4_debug ("Checking group descriptors");
1567
1568 for (i = 0; i < sbi->s_groups_count; i++) {
1569 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1570
1571 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1572 last_block = ext4_blocks_count(sbi->s_es) - 1;
1573 else
1574 last_block = first_block +
1575 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1576
1577 block_bitmap = ext4_block_bitmap(sb, gdp);
1578 if (block_bitmap < first_block || block_bitmap > last_block)
1579 {
1580 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1581 "Block bitmap for group %lu not in group "
1582 "(block %llu)!", i, block_bitmap);
1583 return 0;
1584 }
1585 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1586 if (inode_bitmap < first_block || inode_bitmap > last_block)
1587 {
1588 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1589 "Inode bitmap for group %lu not in group "
1590 "(block %llu)!", i, inode_bitmap);
1591 return 0;
1592 }
1593 inode_table = ext4_inode_table(sb, gdp);
1594 if (inode_table < first_block ||
1595 inode_table + sbi->s_itb_per_group - 1 > last_block)
1596 {
1597 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1598 "Inode table for group %lu not in group "
1599 "(block %llu)!", i, inode_table);
1600 return 0;
1601 }
1602 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1603 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1604 "Checksum for group %lu failed (%u!=%u)\n",
1605 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1606 gdp)), le16_to_cpu(gdp->bg_checksum));
1607 return 0;
1608 }
1609 if (!flexbg_flag)
1610 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1611 }
1612
1613 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1614 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1615 return 1;
1616 }
1617
1618 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1619 * the superblock) which were deleted from all directories, but held open by
1620 * a process at the time of a crash. We walk the list and try to delete these
1621 * inodes at recovery time (only with a read-write filesystem).
1622 *
1623 * In order to keep the orphan inode chain consistent during traversal (in
1624 * case of crash during recovery), we link each inode into the superblock
1625 * orphan list_head and handle it the same way as an inode deletion during
1626 * normal operation (which journals the operations for us).
1627 *
1628 * We only do an iget() and an iput() on each inode, which is very safe if we
1629 * accidentally point at an in-use or already deleted inode. The worst that
1630 * can happen in this case is that we get a "bit already cleared" message from
1631 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1632 * e2fsck was run on this filesystem, and it must have already done the orphan
1633 * inode cleanup for us, so we can safely abort without any further action.
1634 */
1635 static void ext4_orphan_cleanup (struct super_block * sb,
1636 struct ext4_super_block * es)
1637 {
1638 unsigned int s_flags = sb->s_flags;
1639 int nr_orphans = 0, nr_truncates = 0;
1640 #ifdef CONFIG_QUOTA
1641 int i;
1642 #endif
1643 if (!es->s_last_orphan) {
1644 jbd_debug(4, "no orphan inodes to clean up\n");
1645 return;
1646 }
1647
1648 if (bdev_read_only(sb->s_bdev)) {
1649 printk(KERN_ERR "EXT4-fs: write access "
1650 "unavailable, skipping orphan cleanup.\n");
1651 return;
1652 }
1653
1654 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1655 if (es->s_last_orphan)
1656 jbd_debug(1, "Errors on filesystem, "
1657 "clearing orphan list.\n");
1658 es->s_last_orphan = 0;
1659 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1660 return;
1661 }
1662
1663 if (s_flags & MS_RDONLY) {
1664 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1665 sb->s_id);
1666 sb->s_flags &= ~MS_RDONLY;
1667 }
1668 #ifdef CONFIG_QUOTA
1669 /* Needed for iput() to work correctly and not trash data */
1670 sb->s_flags |= MS_ACTIVE;
1671 /* Turn on quotas so that they are updated correctly */
1672 for (i = 0; i < MAXQUOTAS; i++) {
1673 if (EXT4_SB(sb)->s_qf_names[i]) {
1674 int ret = ext4_quota_on_mount(sb, i);
1675 if (ret < 0)
1676 printk(KERN_ERR
1677 "EXT4-fs: Cannot turn on journaled "
1678 "quota: error %d\n", ret);
1679 }
1680 }
1681 #endif
1682
1683 while (es->s_last_orphan) {
1684 struct inode *inode;
1685
1686 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1687 if (IS_ERR(inode)) {
1688 es->s_last_orphan = 0;
1689 break;
1690 }
1691
1692 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1693 DQUOT_INIT(inode);
1694 if (inode->i_nlink) {
1695 printk(KERN_DEBUG
1696 "%s: truncating inode %lu to %Ld bytes\n",
1697 __func__, inode->i_ino, inode->i_size);
1698 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1699 inode->i_ino, inode->i_size);
1700 ext4_truncate(inode);
1701 nr_truncates++;
1702 } else {
1703 printk(KERN_DEBUG
1704 "%s: deleting unreferenced inode %lu\n",
1705 __func__, inode->i_ino);
1706 jbd_debug(2, "deleting unreferenced inode %lu\n",
1707 inode->i_ino);
1708 nr_orphans++;
1709 }
1710 iput(inode); /* The delete magic happens here! */
1711 }
1712
1713 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1714
1715 if (nr_orphans)
1716 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1717 sb->s_id, PLURAL(nr_orphans));
1718 if (nr_truncates)
1719 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1720 sb->s_id, PLURAL(nr_truncates));
1721 #ifdef CONFIG_QUOTA
1722 /* Turn quotas off */
1723 for (i = 0; i < MAXQUOTAS; i++) {
1724 if (sb_dqopt(sb)->files[i])
1725 vfs_quota_off(sb, i, 0);
1726 }
1727 #endif
1728 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1729 }
1730 /*
1731 * Maximal extent format file size.
1732 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1733 * extent format containers, within a sector_t, and within i_blocks
1734 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1735 * so that won't be a limiting factor.
1736 *
1737 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1738 */
1739 static loff_t ext4_max_size(int blkbits)
1740 {
1741 loff_t res;
1742 loff_t upper_limit = MAX_LFS_FILESIZE;
1743
1744 /* small i_blocks in vfs inode? */
1745 if (sizeof(blkcnt_t) < sizeof(u64)) {
1746 /*
1747 * CONFIG_LSF is not enabled implies the inode
1748 * i_block represent total blocks in 512 bytes
1749 * 32 == size of vfs inode i_blocks * 8
1750 */
1751 upper_limit = (1LL << 32) - 1;
1752
1753 /* total blocks in file system block size */
1754 upper_limit >>= (blkbits - 9);
1755 upper_limit <<= blkbits;
1756 }
1757
1758 /* 32-bit extent-start container, ee_block */
1759 res = 1LL << 32;
1760 res <<= blkbits;
1761 res -= 1;
1762
1763 /* Sanity check against vm- & vfs- imposed limits */
1764 if (res > upper_limit)
1765 res = upper_limit;
1766
1767 return res;
1768 }
1769
1770 /*
1771 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1772 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1773 * We need to be 1 filesystem block less than the 2^48 sector limit.
1774 */
1775 static loff_t ext4_max_bitmap_size(int bits)
1776 {
1777 loff_t res = EXT4_NDIR_BLOCKS;
1778 int meta_blocks;
1779 loff_t upper_limit;
1780 /* This is calculated to be the largest file size for a
1781 * dense, bitmapped file such that the total number of
1782 * sectors in the file, including data and all indirect blocks,
1783 * does not exceed 2^48 -1
1784 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1785 * total number of 512 bytes blocks of the file
1786 */
1787
1788 if (sizeof(blkcnt_t) < sizeof(u64)) {
1789 /*
1790 * CONFIG_LSF is not enabled implies the inode
1791 * i_block represent total blocks in 512 bytes
1792 * 32 == size of vfs inode i_blocks * 8
1793 */
1794 upper_limit = (1LL << 32) - 1;
1795
1796 /* total blocks in file system block size */
1797 upper_limit >>= (bits - 9);
1798
1799 } else {
1800 /*
1801 * We use 48 bit ext4_inode i_blocks
1802 * With EXT4_HUGE_FILE_FL set the i_blocks
1803 * represent total number of blocks in
1804 * file system block size
1805 */
1806 upper_limit = (1LL << 48) - 1;
1807
1808 }
1809
1810 /* indirect blocks */
1811 meta_blocks = 1;
1812 /* double indirect blocks */
1813 meta_blocks += 1 + (1LL << (bits-2));
1814 /* tripple indirect blocks */
1815 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1816
1817 upper_limit -= meta_blocks;
1818 upper_limit <<= bits;
1819
1820 res += 1LL << (bits-2);
1821 res += 1LL << (2*(bits-2));
1822 res += 1LL << (3*(bits-2));
1823 res <<= bits;
1824 if (res > upper_limit)
1825 res = upper_limit;
1826
1827 if (res > MAX_LFS_FILESIZE)
1828 res = MAX_LFS_FILESIZE;
1829
1830 return res;
1831 }
1832
1833 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1834 ext4_fsblk_t logical_sb_block, int nr)
1835 {
1836 struct ext4_sb_info *sbi = EXT4_SB(sb);
1837 ext4_group_t bg, first_meta_bg;
1838 int has_super = 0;
1839
1840 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1841
1842 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1843 nr < first_meta_bg)
1844 return logical_sb_block + nr + 1;
1845 bg = sbi->s_desc_per_block * nr;
1846 if (ext4_bg_has_super(sb, bg))
1847 has_super = 1;
1848 return (has_super + ext4_group_first_block_no(sb, bg));
1849 }
1850
1851 /**
1852 * ext4_get_stripe_size: Get the stripe size.
1853 * @sbi: In memory super block info
1854 *
1855 * If we have specified it via mount option, then
1856 * use the mount option value. If the value specified at mount time is
1857 * greater than the blocks per group use the super block value.
1858 * If the super block value is greater than blocks per group return 0.
1859 * Allocator needs it be less than blocks per group.
1860 *
1861 */
1862 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1863 {
1864 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1865 unsigned long stripe_width =
1866 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1867
1868 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1869 return sbi->s_stripe;
1870
1871 if (stripe_width <= sbi->s_blocks_per_group)
1872 return stripe_width;
1873
1874 if (stride <= sbi->s_blocks_per_group)
1875 return stride;
1876
1877 return 0;
1878 }
1879
1880 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1881 __releases(kernel_lock)
1882 __acquires(kernel_lock)
1883
1884 {
1885 struct buffer_head * bh;
1886 struct ext4_super_block *es = NULL;
1887 struct ext4_sb_info *sbi;
1888 ext4_fsblk_t block;
1889 ext4_fsblk_t sb_block = get_sb_block(&data);
1890 ext4_fsblk_t logical_sb_block;
1891 unsigned long offset = 0;
1892 unsigned int journal_inum = 0;
1893 unsigned long journal_devnum = 0;
1894 unsigned long def_mount_opts;
1895 struct inode *root;
1896 int ret = -EINVAL;
1897 int blocksize;
1898 int db_count;
1899 int i;
1900 int needs_recovery;
1901 __le32 features;
1902 __u64 blocks_count;
1903 int err;
1904
1905 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1906 if (!sbi)
1907 return -ENOMEM;
1908 sb->s_fs_info = sbi;
1909 sbi->s_mount_opt = 0;
1910 sbi->s_resuid = EXT4_DEF_RESUID;
1911 sbi->s_resgid = EXT4_DEF_RESGID;
1912 sbi->s_sb_block = sb_block;
1913
1914 unlock_kernel();
1915
1916 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1917 if (!blocksize) {
1918 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1919 goto out_fail;
1920 }
1921
1922 /*
1923 * The ext4 superblock will not be buffer aligned for other than 1kB
1924 * block sizes. We need to calculate the offset from buffer start.
1925 */
1926 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1927 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1928 offset = do_div(logical_sb_block, blocksize);
1929 } else {
1930 logical_sb_block = sb_block;
1931 }
1932
1933 if (!(bh = sb_bread(sb, logical_sb_block))) {
1934 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1935 goto out_fail;
1936 }
1937 /*
1938 * Note: s_es must be initialized as soon as possible because
1939 * some ext4 macro-instructions depend on its value
1940 */
1941 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1942 sbi->s_es = es;
1943 sb->s_magic = le16_to_cpu(es->s_magic);
1944 if (sb->s_magic != EXT4_SUPER_MAGIC)
1945 goto cantfind_ext4;
1946
1947 /* Set defaults before we parse the mount options */
1948 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1949 if (def_mount_opts & EXT4_DEFM_DEBUG)
1950 set_opt(sbi->s_mount_opt, DEBUG);
1951 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1952 set_opt(sbi->s_mount_opt, GRPID);
1953 if (def_mount_opts & EXT4_DEFM_UID16)
1954 set_opt(sbi->s_mount_opt, NO_UID32);
1955 #ifdef CONFIG_EXT4DEV_FS_XATTR
1956 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1957 set_opt(sbi->s_mount_opt, XATTR_USER);
1958 #endif
1959 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1960 if (def_mount_opts & EXT4_DEFM_ACL)
1961 set_opt(sbi->s_mount_opt, POSIX_ACL);
1962 #endif
1963 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1964 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1965 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1966 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1967 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1968 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1969
1970 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1971 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1972 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1973 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1974 else
1975 set_opt(sbi->s_mount_opt, ERRORS_RO);
1976
1977 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1978 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1979
1980 set_opt(sbi->s_mount_opt, RESERVATION);
1981 set_opt(sbi->s_mount_opt, BARRIER);
1982
1983 /*
1984 * turn on extents feature by default in ext4 filesystem
1985 * User -o noextents to turn it off
1986 */
1987 set_opt(sbi->s_mount_opt, EXTENTS);
1988 /*
1989 * turn on mballoc feature by default in ext4 filesystem
1990 * User -o nomballoc to turn it off
1991 */
1992 set_opt(sbi->s_mount_opt, MBALLOC);
1993
1994 /*
1995 * enable delayed allocation by default
1996 * Use -o nodelalloc to turn it off
1997 */
1998 set_opt(sbi->s_mount_opt, DELALLOC);
1999
2000
2001 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
2002 NULL, 0))
2003 goto failed_mount;
2004
2005 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2006 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2007
2008 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2009 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2010 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2011 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2012 printk(KERN_WARNING
2013 "EXT4-fs warning: feature flags set on rev 0 fs, "
2014 "running e2fsck is recommended\n");
2015
2016 /*
2017 * Since ext4 is still considered development code, we require
2018 * that the TEST_FILESYS flag in s->flags be set.
2019 */
2020 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
2021 printk(KERN_WARNING "EXT4-fs: %s: not marked "
2022 "OK to use with test code.\n", sb->s_id);
2023 goto failed_mount;
2024 }
2025
2026 /*
2027 * Check feature flags regardless of the revision level, since we
2028 * previously didn't change the revision level when setting the flags,
2029 * so there is a chance incompat flags are set on a rev 0 filesystem.
2030 */
2031 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2032 if (features) {
2033 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2034 "unsupported optional features (%x).\n",
2035 sb->s_id, le32_to_cpu(features));
2036 goto failed_mount;
2037 }
2038 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2039 if (!(sb->s_flags & MS_RDONLY) && features) {
2040 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2041 "unsupported optional features (%x).\n",
2042 sb->s_id, le32_to_cpu(features));
2043 goto failed_mount;
2044 }
2045 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2046 /*
2047 * Large file size enabled file system can only be
2048 * mount if kernel is build with CONFIG_LSF
2049 */
2050 if (sizeof(root->i_blocks) < sizeof(u64) &&
2051 !(sb->s_flags & MS_RDONLY)) {
2052 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2053 "files cannot be mounted read-write "
2054 "without CONFIG_LSF.\n", sb->s_id);
2055 goto failed_mount;
2056 }
2057 }
2058 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2059
2060 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2061 blocksize > EXT4_MAX_BLOCK_SIZE) {
2062 printk(KERN_ERR
2063 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2064 blocksize, sb->s_id);
2065 goto failed_mount;
2066 }
2067
2068 if (sb->s_blocksize != blocksize) {
2069
2070 /* Validate the filesystem blocksize */
2071 if (!sb_set_blocksize(sb, blocksize)) {
2072 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2073 blocksize);
2074 goto failed_mount;
2075 }
2076
2077 brelse (bh);
2078 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2079 offset = do_div(logical_sb_block, blocksize);
2080 bh = sb_bread(sb, logical_sb_block);
2081 if (!bh) {
2082 printk(KERN_ERR
2083 "EXT4-fs: Can't read superblock on 2nd try.\n");
2084 goto failed_mount;
2085 }
2086 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2087 sbi->s_es = es;
2088 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2089 printk (KERN_ERR
2090 "EXT4-fs: Magic mismatch, very weird !\n");
2091 goto failed_mount;
2092 }
2093 }
2094
2095 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
2096 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
2097
2098 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2099 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2100 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2101 } else {
2102 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2103 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2104 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2105 (!is_power_of_2(sbi->s_inode_size)) ||
2106 (sbi->s_inode_size > blocksize)) {
2107 printk (KERN_ERR
2108 "EXT4-fs: unsupported inode size: %d\n",
2109 sbi->s_inode_size);
2110 goto failed_mount;
2111 }
2112 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2113 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2114 }
2115 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2116 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2117 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2118 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2119 !is_power_of_2(sbi->s_desc_size)) {
2120 printk(KERN_ERR
2121 "EXT4-fs: unsupported descriptor size %lu\n",
2122 sbi->s_desc_size);
2123 goto failed_mount;
2124 }
2125 } else
2126 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2127 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2128 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2129 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2130 goto cantfind_ext4;
2131 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2132 if (sbi->s_inodes_per_block == 0)
2133 goto cantfind_ext4;
2134 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2135 sbi->s_inodes_per_block;
2136 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2137 sbi->s_sbh = bh;
2138 sbi->s_mount_state = le16_to_cpu(es->s_state);
2139 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2140 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2141 for (i=0; i < 4; i++)
2142 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2143 sbi->s_def_hash_version = es->s_def_hash_version;
2144
2145 if (sbi->s_blocks_per_group > blocksize * 8) {
2146 printk (KERN_ERR
2147 "EXT4-fs: #blocks per group too big: %lu\n",
2148 sbi->s_blocks_per_group);
2149 goto failed_mount;
2150 }
2151 if (sbi->s_inodes_per_group > blocksize * 8) {
2152 printk (KERN_ERR
2153 "EXT4-fs: #inodes per group too big: %lu\n",
2154 sbi->s_inodes_per_group);
2155 goto failed_mount;
2156 }
2157
2158 if (ext4_blocks_count(es) >
2159 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2160 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2161 " too large to mount safely\n", sb->s_id);
2162 if (sizeof(sector_t) < 8)
2163 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2164 "enabled\n");
2165 goto failed_mount;
2166 }
2167
2168 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2169 goto cantfind_ext4;
2170
2171 /* ensure blocks_count calculation below doesn't sign-extend */
2172 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2173 le32_to_cpu(es->s_first_data_block) + 1) {
2174 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2175 "first data block %u, blocks per group %lu\n",
2176 ext4_blocks_count(es),
2177 le32_to_cpu(es->s_first_data_block),
2178 EXT4_BLOCKS_PER_GROUP(sb));
2179 goto failed_mount;
2180 }
2181 blocks_count = (ext4_blocks_count(es) -
2182 le32_to_cpu(es->s_first_data_block) +
2183 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2184 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2185 sbi->s_groups_count = blocks_count;
2186 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2187 EXT4_DESC_PER_BLOCK(sb);
2188 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2189 GFP_KERNEL);
2190 if (sbi->s_group_desc == NULL) {
2191 printk (KERN_ERR "EXT4-fs: not enough memory\n");
2192 goto failed_mount;
2193 }
2194
2195 bgl_lock_init(&sbi->s_blockgroup_lock);
2196
2197 for (i = 0; i < db_count; i++) {
2198 block = descriptor_loc(sb, logical_sb_block, i);
2199 sbi->s_group_desc[i] = sb_bread(sb, block);
2200 if (!sbi->s_group_desc[i]) {
2201 printk (KERN_ERR "EXT4-fs: "
2202 "can't read group descriptor %d\n", i);
2203 db_count = i;
2204 goto failed_mount2;
2205 }
2206 }
2207 if (!ext4_check_descriptors (sb)) {
2208 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2209 goto failed_mount2;
2210 }
2211 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2212 if (!ext4_fill_flex_info(sb)) {
2213 printk(KERN_ERR
2214 "EXT4-fs: unable to initialize "
2215 "flex_bg meta info!\n");
2216 goto failed_mount2;
2217 }
2218
2219 sbi->s_gdb_count = db_count;
2220 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2221 spin_lock_init(&sbi->s_next_gen_lock);
2222
2223 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2224 ext4_count_free_blocks(sb));
2225 if (!err) {
2226 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2227 ext4_count_free_inodes(sb));
2228 }
2229 if (!err) {
2230 err = percpu_counter_init(&sbi->s_dirs_counter,
2231 ext4_count_dirs(sb));
2232 }
2233 if (err) {
2234 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2235 goto failed_mount3;
2236 }
2237
2238 /* per fileystem reservation list head & lock */
2239 spin_lock_init(&sbi->s_rsv_window_lock);
2240 sbi->s_rsv_window_root = RB_ROOT;
2241 /* Add a single, static dummy reservation to the start of the
2242 * reservation window list --- it gives us a placeholder for
2243 * append-at-start-of-list which makes the allocation logic
2244 * _much_ simpler. */
2245 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2246 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2247 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2248 sbi->s_rsv_window_head.rsv_goal_size = 0;
2249 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
2250
2251 sbi->s_stripe = ext4_get_stripe_size(sbi);
2252
2253 /*
2254 * set up enough so that it can read an inode
2255 */
2256 sb->s_op = &ext4_sops;
2257 sb->s_export_op = &ext4_export_ops;
2258 sb->s_xattr = ext4_xattr_handlers;
2259 #ifdef CONFIG_QUOTA
2260 sb->s_qcop = &ext4_qctl_operations;
2261 sb->dq_op = &ext4_quota_operations;
2262 #endif
2263 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2264
2265 sb->s_root = NULL;
2266
2267 needs_recovery = (es->s_last_orphan != 0 ||
2268 EXT4_HAS_INCOMPAT_FEATURE(sb,
2269 EXT4_FEATURE_INCOMPAT_RECOVER));
2270
2271 /*
2272 * The first inode we look at is the journal inode. Don't try
2273 * root first: it may be modified in the journal!
2274 */
2275 if (!test_opt(sb, NOLOAD) &&
2276 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2277 if (ext4_load_journal(sb, es, journal_devnum))
2278 goto failed_mount3;
2279 if (!(sb->s_flags & MS_RDONLY) &&
2280 EXT4_SB(sb)->s_journal->j_failed_commit) {
2281 printk(KERN_CRIT "EXT4-fs error (device %s): "
2282 "ext4_fill_super: Journal transaction "
2283 "%u is corrupt\n", sb->s_id,
2284 EXT4_SB(sb)->s_journal->j_failed_commit);
2285 if (test_opt (sb, ERRORS_RO)) {
2286 printk (KERN_CRIT
2287 "Mounting filesystem read-only\n");
2288 sb->s_flags |= MS_RDONLY;
2289 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2290 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2291 }
2292 if (test_opt(sb, ERRORS_PANIC)) {
2293 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2294 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2295 ext4_commit_super(sb, es, 1);
2296 printk(KERN_CRIT
2297 "EXT4-fs (device %s): mount failed\n",
2298 sb->s_id);
2299 goto failed_mount4;
2300 }
2301 }
2302 } else if (journal_inum) {
2303 if (ext4_create_journal(sb, es, journal_inum))
2304 goto failed_mount3;
2305 } else {
2306 if (!silent)
2307 printk (KERN_ERR
2308 "ext4: No journal on filesystem on %s\n",
2309 sb->s_id);
2310 goto failed_mount3;
2311 }
2312
2313 if (ext4_blocks_count(es) > 0xffffffffULL &&
2314 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2315 JBD2_FEATURE_INCOMPAT_64BIT)) {
2316 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2317 goto failed_mount4;
2318 }
2319
2320 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2321 jbd2_journal_set_features(sbi->s_journal,
2322 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2323 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2324 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2325 jbd2_journal_set_features(sbi->s_journal,
2326 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2327 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2328 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2329 } else {
2330 jbd2_journal_clear_features(sbi->s_journal,
2331 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2332 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2333 }
2334
2335 /* We have now updated the journal if required, so we can
2336 * validate the data journaling mode. */
2337 switch (test_opt(sb, DATA_FLAGS)) {
2338 case 0:
2339 /* No mode set, assume a default based on the journal
2340 * capabilities: ORDERED_DATA if the journal can
2341 * cope, else JOURNAL_DATA
2342 */
2343 if (jbd2_journal_check_available_features
2344 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2345 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2346 else
2347 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2348 break;
2349
2350 case EXT4_MOUNT_ORDERED_DATA:
2351 case EXT4_MOUNT_WRITEBACK_DATA:
2352 if (!jbd2_journal_check_available_features
2353 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2354 printk(KERN_ERR "EXT4-fs: Journal does not support "
2355 "requested data journaling mode\n");
2356 goto failed_mount4;
2357 }
2358 default:
2359 break;
2360 }
2361
2362 if (test_opt(sb, NOBH)) {
2363 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2364 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2365 "its supported only with writeback mode\n");
2366 clear_opt(sbi->s_mount_opt, NOBH);
2367 }
2368 }
2369 /*
2370 * The jbd2_journal_load will have done any necessary log recovery,
2371 * so we can safely mount the rest of the filesystem now.
2372 */
2373
2374 root = ext4_iget(sb, EXT4_ROOT_INO);
2375 if (IS_ERR(root)) {
2376 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2377 ret = PTR_ERR(root);
2378 goto failed_mount4;
2379 }
2380 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2381 iput(root);
2382 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2383 goto failed_mount4;
2384 }
2385 sb->s_root = d_alloc_root(root);
2386 if (!sb->s_root) {
2387 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2388 iput(root);
2389 ret = -ENOMEM;
2390 goto failed_mount4;
2391 }
2392
2393 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2394
2395 /* determine the minimum size of new large inodes, if present */
2396 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2397 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2398 EXT4_GOOD_OLD_INODE_SIZE;
2399 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2400 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2401 if (sbi->s_want_extra_isize <
2402 le16_to_cpu(es->s_want_extra_isize))
2403 sbi->s_want_extra_isize =
2404 le16_to_cpu(es->s_want_extra_isize);
2405 if (sbi->s_want_extra_isize <
2406 le16_to_cpu(es->s_min_extra_isize))
2407 sbi->s_want_extra_isize =
2408 le16_to_cpu(es->s_min_extra_isize);
2409 }
2410 }
2411 /* Check if enough inode space is available */
2412 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2413 sbi->s_inode_size) {
2414 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2415 EXT4_GOOD_OLD_INODE_SIZE;
2416 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2417 "available.\n");
2418 }
2419
2420 /*
2421 * akpm: core read_super() calls in here with the superblock locked.
2422 * That deadlocks, because orphan cleanup needs to lock the superblock
2423 * in numerous places. Here we just pop the lock - it's relatively
2424 * harmless, because we are now ready to accept write_super() requests,
2425 * and aviro says that's the only reason for hanging onto the
2426 * superblock lock.
2427 */
2428 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2429 ext4_orphan_cleanup(sb, es);
2430 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2431 if (needs_recovery)
2432 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2433 ext4_mark_recovery_complete(sb, es);
2434 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2435 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2436 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2437 "writeback");
2438
2439 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2440 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2441 "requested data journaling mode\n");
2442 clear_opt(sbi->s_mount_opt, DELALLOC);
2443 } else if (test_opt(sb, DELALLOC))
2444 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2445
2446 ext4_ext_init(sb);
2447 ext4_mb_init(sb, needs_recovery);
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 failed_mount2:
2466 for (i = 0; i < db_count; i++)
2467 brelse(sbi->s_group_desc[i]);
2468 kfree(sbi->s_group_desc);
2469 failed_mount:
2470 #ifdef CONFIG_QUOTA
2471 for (i = 0; i < MAXQUOTAS; i++)
2472 kfree(sbi->s_qf_names[i]);
2473 #endif
2474 ext4_blkdev_remove(sbi);
2475 brelse(bh);
2476 out_fail:
2477 sb->s_fs_info = NULL;
2478 kfree(sbi);
2479 lock_kernel();
2480 return ret;
2481 }
2482
2483 /*
2484 * Setup any per-fs journal parameters now. We'll do this both on
2485 * initial mount, once the journal has been initialised but before we've
2486 * done any recovery; and again on any subsequent remount.
2487 */
2488 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2489 {
2490 struct ext4_sb_info *sbi = EXT4_SB(sb);
2491
2492 if (sbi->s_commit_interval)
2493 journal->j_commit_interval = sbi->s_commit_interval;
2494 /* We could also set up an ext4-specific default for the commit
2495 * interval here, but for now we'll just fall back to the jbd
2496 * default. */
2497
2498 spin_lock(&journal->j_state_lock);
2499 if (test_opt(sb, BARRIER))
2500 journal->j_flags |= JBD2_BARRIER;
2501 else
2502 journal->j_flags &= ~JBD2_BARRIER;
2503 spin_unlock(&journal->j_state_lock);
2504 }
2505
2506 static journal_t *ext4_get_journal(struct super_block *sb,
2507 unsigned int journal_inum)
2508 {
2509 struct inode *journal_inode;
2510 journal_t *journal;
2511
2512 /* First, test for the existence of a valid inode on disk. Bad
2513 * things happen if we iget() an unused inode, as the subsequent
2514 * iput() will try to delete it. */
2515
2516 journal_inode = ext4_iget(sb, journal_inum);
2517 if (IS_ERR(journal_inode)) {
2518 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2519 return NULL;
2520 }
2521 if (!journal_inode->i_nlink) {
2522 make_bad_inode(journal_inode);
2523 iput(journal_inode);
2524 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2525 return NULL;
2526 }
2527
2528 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2529 journal_inode, journal_inode->i_size);
2530 if (!S_ISREG(journal_inode->i_mode)) {
2531 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2532 iput(journal_inode);
2533 return NULL;
2534 }
2535
2536 journal = jbd2_journal_init_inode(journal_inode);
2537 if (!journal) {
2538 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2539 iput(journal_inode);
2540 return NULL;
2541 }
2542 journal->j_private = sb;
2543 ext4_init_journal_params(sb, journal);
2544 return journal;
2545 }
2546
2547 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2548 dev_t j_dev)
2549 {
2550 struct buffer_head * bh;
2551 journal_t *journal;
2552 ext4_fsblk_t start;
2553 ext4_fsblk_t len;
2554 int hblock, blocksize;
2555 ext4_fsblk_t sb_block;
2556 unsigned long offset;
2557 struct ext4_super_block * es;
2558 struct block_device *bdev;
2559
2560 bdev = ext4_blkdev_get(j_dev);
2561 if (bdev == NULL)
2562 return NULL;
2563
2564 if (bd_claim(bdev, sb)) {
2565 printk(KERN_ERR
2566 "EXT4: failed to claim external journal device.\n");
2567 blkdev_put(bdev);
2568 return NULL;
2569 }
2570
2571 blocksize = sb->s_blocksize;
2572 hblock = bdev_hardsect_size(bdev);
2573 if (blocksize < hblock) {
2574 printk(KERN_ERR
2575 "EXT4-fs: blocksize too small for journal device.\n");
2576 goto out_bdev;
2577 }
2578
2579 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2580 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2581 set_blocksize(bdev, blocksize);
2582 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2583 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2584 "external journal\n");
2585 goto out_bdev;
2586 }
2587
2588 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2589 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2590 !(le32_to_cpu(es->s_feature_incompat) &
2591 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2592 printk(KERN_ERR "EXT4-fs: external journal has "
2593 "bad superblock\n");
2594 brelse(bh);
2595 goto out_bdev;
2596 }
2597
2598 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2599 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2600 brelse(bh);
2601 goto out_bdev;
2602 }
2603
2604 len = ext4_blocks_count(es);
2605 start = sb_block + 1;
2606 brelse(bh); /* we're done with the superblock */
2607
2608 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2609 start, len, blocksize);
2610 if (!journal) {
2611 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2612 goto out_bdev;
2613 }
2614 journal->j_private = sb;
2615 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2616 wait_on_buffer(journal->j_sb_buffer);
2617 if (!buffer_uptodate(journal->j_sb_buffer)) {
2618 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2619 goto out_journal;
2620 }
2621 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2622 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2623 "user (unsupported) - %d\n",
2624 be32_to_cpu(journal->j_superblock->s_nr_users));
2625 goto out_journal;
2626 }
2627 EXT4_SB(sb)->journal_bdev = bdev;
2628 ext4_init_journal_params(sb, journal);
2629 return journal;
2630 out_journal:
2631 jbd2_journal_destroy(journal);
2632 out_bdev:
2633 ext4_blkdev_put(bdev);
2634 return NULL;
2635 }
2636
2637 static int ext4_load_journal(struct super_block *sb,
2638 struct ext4_super_block *es,
2639 unsigned long journal_devnum)
2640 {
2641 journal_t *journal;
2642 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2643 dev_t journal_dev;
2644 int err = 0;
2645 int really_read_only;
2646
2647 if (journal_devnum &&
2648 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2649 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2650 "numbers have changed\n");
2651 journal_dev = new_decode_dev(journal_devnum);
2652 } else
2653 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2654
2655 really_read_only = bdev_read_only(sb->s_bdev);
2656
2657 /*
2658 * Are we loading a blank journal or performing recovery after a
2659 * crash? For recovery, we need to check in advance whether we
2660 * can get read-write access to the device.
2661 */
2662
2663 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2664 if (sb->s_flags & MS_RDONLY) {
2665 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2666 "required on readonly filesystem.\n");
2667 if (really_read_only) {
2668 printk(KERN_ERR "EXT4-fs: write access "
2669 "unavailable, cannot proceed.\n");
2670 return -EROFS;
2671 }
2672 printk (KERN_INFO "EXT4-fs: write access will "
2673 "be enabled during recovery.\n");
2674 }
2675 }
2676
2677 if (journal_inum && journal_dev) {
2678 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2679 "and inode journals!\n");
2680 return -EINVAL;
2681 }
2682
2683 if (journal_inum) {
2684 if (!(journal = ext4_get_journal(sb, journal_inum)))
2685 return -EINVAL;
2686 } else {
2687 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2688 return -EINVAL;
2689 }
2690
2691 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2692 err = jbd2_journal_update_format(journal);
2693 if (err) {
2694 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2695 jbd2_journal_destroy(journal);
2696 return err;
2697 }
2698 }
2699
2700 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2701 err = jbd2_journal_wipe(journal, !really_read_only);
2702 if (!err)
2703 err = jbd2_journal_load(journal);
2704
2705 if (err) {
2706 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2707 jbd2_journal_destroy(journal);
2708 return err;
2709 }
2710
2711 EXT4_SB(sb)->s_journal = journal;
2712 ext4_clear_journal_err(sb, es);
2713
2714 if (journal_devnum &&
2715 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2716 es->s_journal_dev = cpu_to_le32(journal_devnum);
2717 sb->s_dirt = 1;
2718
2719 /* Make sure we flush the recovery flag to disk. */
2720 ext4_commit_super(sb, es, 1);
2721 }
2722
2723 return 0;
2724 }
2725
2726 static int ext4_create_journal(struct super_block * sb,
2727 struct ext4_super_block * es,
2728 unsigned int journal_inum)
2729 {
2730 journal_t *journal;
2731 int err;
2732
2733 if (sb->s_flags & MS_RDONLY) {
2734 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2735 "create journal.\n");
2736 return -EROFS;
2737 }
2738
2739 journal = ext4_get_journal(sb, journal_inum);
2740 if (!journal)
2741 return -EINVAL;
2742
2743 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2744 journal_inum);
2745
2746 err = jbd2_journal_create(journal);
2747 if (err) {
2748 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2749 jbd2_journal_destroy(journal);
2750 return -EIO;
2751 }
2752
2753 EXT4_SB(sb)->s_journal = journal;
2754
2755 ext4_update_dynamic_rev(sb);
2756 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2757 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2758
2759 es->s_journal_inum = cpu_to_le32(journal_inum);
2760 sb->s_dirt = 1;
2761
2762 /* Make sure we flush the recovery flag to disk. */
2763 ext4_commit_super(sb, es, 1);
2764
2765 return 0;
2766 }
2767
2768 static void ext4_commit_super (struct super_block * sb,
2769 struct ext4_super_block * es,
2770 int sync)
2771 {
2772 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2773
2774 if (!sbh)
2775 return;
2776 es->s_wtime = cpu_to_le32(get_seconds());
2777 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2778 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2779 BUFFER_TRACE(sbh, "marking dirty");
2780 mark_buffer_dirty(sbh);
2781 if (sync)
2782 sync_dirty_buffer(sbh);
2783 }
2784
2785
2786 /*
2787 * Have we just finished recovery? If so, and if we are mounting (or
2788 * remounting) the filesystem readonly, then we will end up with a
2789 * consistent fs on disk. Record that fact.
2790 */
2791 static void ext4_mark_recovery_complete(struct super_block * sb,
2792 struct ext4_super_block * es)
2793 {
2794 journal_t *journal = EXT4_SB(sb)->s_journal;
2795
2796 jbd2_journal_lock_updates(journal);
2797 jbd2_journal_flush(journal);
2798 lock_super(sb);
2799 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2800 sb->s_flags & MS_RDONLY) {
2801 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2802 sb->s_dirt = 0;
2803 ext4_commit_super(sb, es, 1);
2804 }
2805 unlock_super(sb);
2806 jbd2_journal_unlock_updates(journal);
2807 }
2808
2809 /*
2810 * If we are mounting (or read-write remounting) a filesystem whose journal
2811 * has recorded an error from a previous lifetime, move that error to the
2812 * main filesystem now.
2813 */
2814 static void ext4_clear_journal_err(struct super_block * sb,
2815 struct ext4_super_block * es)
2816 {
2817 journal_t *journal;
2818 int j_errno;
2819 const char *errstr;
2820
2821 journal = EXT4_SB(sb)->s_journal;
2822
2823 /*
2824 * Now check for any error status which may have been recorded in the
2825 * journal by a prior ext4_error() or ext4_abort()
2826 */
2827
2828 j_errno = jbd2_journal_errno(journal);
2829 if (j_errno) {
2830 char nbuf[16];
2831
2832 errstr = ext4_decode_error(sb, j_errno, nbuf);
2833 ext4_warning(sb, __func__, "Filesystem error recorded "
2834 "from previous mount: %s", errstr);
2835 ext4_warning(sb, __func__, "Marking fs in need of "
2836 "filesystem check.");
2837
2838 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2839 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2840 ext4_commit_super (sb, es, 1);
2841
2842 jbd2_journal_clear_err(journal);
2843 }
2844 }
2845
2846 /*
2847 * Force the running and committing transactions to commit,
2848 * and wait on the commit.
2849 */
2850 int ext4_force_commit(struct super_block *sb)
2851 {
2852 journal_t *journal;
2853 int ret;
2854
2855 if (sb->s_flags & MS_RDONLY)
2856 return 0;
2857
2858 journal = EXT4_SB(sb)->s_journal;
2859 sb->s_dirt = 0;
2860 ret = ext4_journal_force_commit(journal);
2861 return ret;
2862 }
2863
2864 /*
2865 * Ext4 always journals updates to the superblock itself, so we don't
2866 * have to propagate any other updates to the superblock on disk at this
2867 * point. Just start an async writeback to get the buffers on their way
2868 * to the disk.
2869 *
2870 * This implicitly triggers the writebehind on sync().
2871 */
2872
2873 static void ext4_write_super (struct super_block * sb)
2874 {
2875 if (mutex_trylock(&sb->s_lock) != 0)
2876 BUG();
2877 sb->s_dirt = 0;
2878 }
2879
2880 static int ext4_sync_fs(struct super_block *sb, int wait)
2881 {
2882 tid_t target;
2883
2884 sb->s_dirt = 0;
2885 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2886 if (wait)
2887 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2888 }
2889 return 0;
2890 }
2891
2892 /*
2893 * LVM calls this function before a (read-only) snapshot is created. This
2894 * gives us a chance to flush the journal completely and mark the fs clean.
2895 */
2896 static void ext4_write_super_lockfs(struct super_block *sb)
2897 {
2898 sb->s_dirt = 0;
2899
2900 if (!(sb->s_flags & MS_RDONLY)) {
2901 journal_t *journal = EXT4_SB(sb)->s_journal;
2902
2903 /* Now we set up the journal barrier. */
2904 jbd2_journal_lock_updates(journal);
2905 jbd2_journal_flush(journal);
2906
2907 /* Journal blocked and flushed, clear needs_recovery flag. */
2908 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2909 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2910 }
2911 }
2912
2913 /*
2914 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2915 * flag here, even though the filesystem is not technically dirty yet.
2916 */
2917 static void ext4_unlockfs(struct super_block *sb)
2918 {
2919 if (!(sb->s_flags & MS_RDONLY)) {
2920 lock_super(sb);
2921 /* Reser the needs_recovery flag before the fs is unlocked. */
2922 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2923 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2924 unlock_super(sb);
2925 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2926 }
2927 }
2928
2929 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2930 {
2931 struct ext4_super_block * es;
2932 struct ext4_sb_info *sbi = EXT4_SB(sb);
2933 ext4_fsblk_t n_blocks_count = 0;
2934 unsigned long old_sb_flags;
2935 struct ext4_mount_options old_opts;
2936 int err;
2937 #ifdef CONFIG_QUOTA
2938 int i;
2939 #endif
2940
2941 /* Store the original options */
2942 old_sb_flags = sb->s_flags;
2943 old_opts.s_mount_opt = sbi->s_mount_opt;
2944 old_opts.s_resuid = sbi->s_resuid;
2945 old_opts.s_resgid = sbi->s_resgid;
2946 old_opts.s_commit_interval = sbi->s_commit_interval;
2947 #ifdef CONFIG_QUOTA
2948 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2949 for (i = 0; i < MAXQUOTAS; i++)
2950 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2951 #endif
2952
2953 /*
2954 * Allow the "check" option to be passed as a remount option.
2955 */
2956 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2957 err = -EINVAL;
2958 goto restore_opts;
2959 }
2960
2961 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2962 ext4_abort(sb, __func__, "Abort forced by user");
2963
2964 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2965 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2966
2967 es = sbi->s_es;
2968
2969 ext4_init_journal_params(sb, sbi->s_journal);
2970
2971 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2972 n_blocks_count > ext4_blocks_count(es)) {
2973 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2974 err = -EROFS;
2975 goto restore_opts;
2976 }
2977
2978 if (*flags & MS_RDONLY) {
2979 /*
2980 * First of all, the unconditional stuff we have to do
2981 * to disable replay of the journal when we next remount
2982 */
2983 sb->s_flags |= MS_RDONLY;
2984
2985 /*
2986 * OK, test if we are remounting a valid rw partition
2987 * readonly, and if so set the rdonly flag and then
2988 * mark the partition as valid again.
2989 */
2990 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2991 (sbi->s_mount_state & EXT4_VALID_FS))
2992 es->s_state = cpu_to_le16(sbi->s_mount_state);
2993
2994 /*
2995 * We have to unlock super so that we can wait for
2996 * transactions.
2997 */
2998 unlock_super(sb);
2999 ext4_mark_recovery_complete(sb, es);
3000 lock_super(sb);
3001 } else {
3002 __le32 ret;
3003 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3004 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3005 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3006 "remount RDWR because of unsupported "
3007 "optional features (%x).\n",
3008 sb->s_id, le32_to_cpu(ret));
3009 err = -EROFS;
3010 goto restore_opts;
3011 }
3012
3013 /*
3014 * If we have an unprocessed orphan list hanging
3015 * around from a previously readonly bdev mount,
3016 * require a full umount/remount for now.
3017 */
3018 if (es->s_last_orphan) {
3019 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3020 "remount RDWR because of unprocessed "
3021 "orphan inode list. Please "
3022 "umount/remount instead.\n",
3023 sb->s_id);
3024 err = -EINVAL;
3025 goto restore_opts;
3026 }
3027
3028 /*
3029 * Mounting a RDONLY partition read-write, so reread
3030 * and store the current valid flag. (It may have
3031 * been changed by e2fsck since we originally mounted
3032 * the partition.)
3033 */
3034 ext4_clear_journal_err(sb, es);
3035 sbi->s_mount_state = le16_to_cpu(es->s_state);
3036 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3037 goto restore_opts;
3038 if (!ext4_setup_super (sb, es, 0))
3039 sb->s_flags &= ~MS_RDONLY;
3040 }
3041 }
3042 #ifdef CONFIG_QUOTA
3043 /* Release old quota file names */
3044 for (i = 0; i < MAXQUOTAS; i++)
3045 if (old_opts.s_qf_names[i] &&
3046 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3047 kfree(old_opts.s_qf_names[i]);
3048 #endif
3049 return 0;
3050 restore_opts:
3051 sb->s_flags = old_sb_flags;
3052 sbi->s_mount_opt = old_opts.s_mount_opt;
3053 sbi->s_resuid = old_opts.s_resuid;
3054 sbi->s_resgid = old_opts.s_resgid;
3055 sbi->s_commit_interval = old_opts.s_commit_interval;
3056 #ifdef CONFIG_QUOTA
3057 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3058 for (i = 0; i < MAXQUOTAS; i++) {
3059 if (sbi->s_qf_names[i] &&
3060 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3061 kfree(sbi->s_qf_names[i]);
3062 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3063 }
3064 #endif
3065 return err;
3066 }
3067
3068 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
3069 {
3070 struct super_block *sb = dentry->d_sb;
3071 struct ext4_sb_info *sbi = EXT4_SB(sb);
3072 struct ext4_super_block *es = sbi->s_es;
3073 u64 fsid;
3074
3075 if (test_opt(sb, MINIX_DF)) {
3076 sbi->s_overhead_last = 0;
3077 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3078 ext4_group_t ngroups = sbi->s_groups_count, i;
3079 ext4_fsblk_t overhead = 0;
3080 smp_rmb();
3081
3082 /*
3083 * Compute the overhead (FS structures). This is constant
3084 * for a given filesystem unless the number of block groups
3085 * changes so we cache the previous value until it does.
3086 */
3087
3088 /*
3089 * All of the blocks before first_data_block are
3090 * overhead
3091 */
3092 overhead = le32_to_cpu(es->s_first_data_block);
3093
3094 /*
3095 * Add the overhead attributed to the superblock and
3096 * block group descriptors. If the sparse superblocks
3097 * feature is turned on, then not all groups have this.
3098 */
3099 for (i = 0; i < ngroups; i++) {
3100 overhead += ext4_bg_has_super(sb, i) +
3101 ext4_bg_num_gdb(sb, i);
3102 cond_resched();
3103 }
3104
3105 /*
3106 * Every block group has an inode bitmap, a block
3107 * bitmap, and an inode table.
3108 */
3109 overhead += ngroups * (2 + sbi->s_itb_per_group);
3110 sbi->s_overhead_last = overhead;
3111 smp_wmb();
3112 sbi->s_blocks_last = ext4_blocks_count(es);
3113 }
3114
3115 buf->f_type = EXT4_SUPER_MAGIC;
3116 buf->f_bsize = sb->s_blocksize;
3117 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3118 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
3119 ext4_free_blocks_count_set(es, buf->f_bfree);
3120 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3121 if (buf->f_bfree < ext4_r_blocks_count(es))
3122 buf->f_bavail = 0;
3123 buf->f_files = le32_to_cpu(es->s_inodes_count);
3124 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3125 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3126 buf->f_namelen = EXT4_NAME_LEN;
3127 fsid = le64_to_cpup((void *)es->s_uuid) ^
3128 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3129 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3130 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3131 return 0;
3132 }
3133
3134 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3135 * is locked for write. Otherwise the are possible deadlocks:
3136 * Process 1 Process 2
3137 * ext4_create() quota_sync()
3138 * jbd2_journal_start() write_dquot()
3139 * DQUOT_INIT() down(dqio_mutex)
3140 * down(dqio_mutex) jbd2_journal_start()
3141 *
3142 */
3143
3144 #ifdef CONFIG_QUOTA
3145
3146 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3147 {
3148 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3149 }
3150
3151 static int ext4_dquot_initialize(struct inode *inode, int type)
3152 {
3153 handle_t *handle;
3154 int ret, err;
3155
3156 /* We may create quota structure so we need to reserve enough blocks */
3157 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3158 if (IS_ERR(handle))
3159 return PTR_ERR(handle);
3160 ret = dquot_initialize(inode, type);
3161 err = ext4_journal_stop(handle);
3162 if (!ret)
3163 ret = err;
3164 return ret;
3165 }
3166
3167 static int ext4_dquot_drop(struct inode *inode)
3168 {
3169 handle_t *handle;
3170 int ret, err;
3171
3172 /* We may delete quota structure so we need to reserve enough blocks */
3173 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3174 if (IS_ERR(handle)) {
3175 /*
3176 * We call dquot_drop() anyway to at least release references
3177 * to quota structures so that umount does not hang.
3178 */
3179 dquot_drop(inode);
3180 return PTR_ERR(handle);
3181 }
3182 ret = dquot_drop(inode);
3183 err = ext4_journal_stop(handle);
3184 if (!ret)
3185 ret = err;
3186 return ret;
3187 }
3188
3189 static int ext4_write_dquot(struct dquot *dquot)
3190 {
3191 int ret, err;
3192 handle_t *handle;
3193 struct inode *inode;
3194
3195 inode = dquot_to_inode(dquot);
3196 handle = ext4_journal_start(inode,
3197 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3198 if (IS_ERR(handle))
3199 return PTR_ERR(handle);
3200 ret = dquot_commit(dquot);
3201 err = ext4_journal_stop(handle);
3202 if (!ret)
3203 ret = err;
3204 return ret;
3205 }
3206
3207 static int ext4_acquire_dquot(struct dquot *dquot)
3208 {
3209 int ret, err;
3210 handle_t *handle;
3211
3212 handle = ext4_journal_start(dquot_to_inode(dquot),
3213 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3214 if (IS_ERR(handle))
3215 return PTR_ERR(handle);
3216 ret = dquot_acquire(dquot);
3217 err = ext4_journal_stop(handle);
3218 if (!ret)
3219 ret = err;
3220 return ret;
3221 }
3222
3223 static int ext4_release_dquot(struct dquot *dquot)
3224 {
3225 int ret, err;
3226 handle_t *handle;
3227
3228 handle = ext4_journal_start(dquot_to_inode(dquot),
3229 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3230 if (IS_ERR(handle)) {
3231 /* Release dquot anyway to avoid endless cycle in dqput() */
3232 dquot_release(dquot);
3233 return PTR_ERR(handle);
3234 }
3235 ret = dquot_release(dquot);
3236 err = ext4_journal_stop(handle);
3237 if (!ret)
3238 ret = err;
3239 return ret;
3240 }
3241
3242 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3243 {
3244 /* Are we journaling quotas? */
3245 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3246 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3247 dquot_mark_dquot_dirty(dquot);
3248 return ext4_write_dquot(dquot);
3249 } else {
3250 return dquot_mark_dquot_dirty(dquot);
3251 }
3252 }
3253
3254 static int ext4_write_info(struct super_block *sb, int type)
3255 {
3256 int ret, err;
3257 handle_t *handle;
3258
3259 /* Data block + inode block */
3260 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3261 if (IS_ERR(handle))
3262 return PTR_ERR(handle);
3263 ret = dquot_commit_info(sb, type);
3264 err = ext4_journal_stop(handle);
3265 if (!ret)
3266 ret = err;
3267 return ret;
3268 }
3269
3270 /*
3271 * Turn on quotas during mount time - we need to find
3272 * the quota file and such...
3273 */
3274 static int ext4_quota_on_mount(struct super_block *sb, int type)
3275 {
3276 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3277 EXT4_SB(sb)->s_jquota_fmt, type);
3278 }
3279
3280 /*
3281 * Standard function to be called on quota_on
3282 */
3283 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3284 char *path, int remount)
3285 {
3286 int err;
3287 struct nameidata nd;
3288
3289 if (!test_opt(sb, QUOTA))
3290 return -EINVAL;
3291 /* When remounting, no checks are needed and in fact, path is NULL */
3292 if (remount)
3293 return vfs_quota_on(sb, type, format_id, path, remount);
3294
3295 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3296 if (err)
3297 return err;
3298
3299 /* Quotafile not on the same filesystem? */
3300 if (nd.path.mnt->mnt_sb != sb) {
3301 path_put(&nd.path);
3302 return -EXDEV;
3303 }
3304 /* Journaling quota? */
3305 if (EXT4_SB(sb)->s_qf_names[type]) {
3306 /* Quotafile not of fs root? */
3307 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3308 printk(KERN_WARNING
3309 "EXT4-fs: Quota file not on filesystem root. "
3310 "Journaled quota will not work.\n");
3311 }
3312
3313 /*
3314 * When we journal data on quota file, we have to flush journal to see
3315 * all updates to the file when we bypass pagecache...
3316 */
3317 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3318 /*
3319 * We don't need to lock updates but journal_flush() could
3320 * otherwise be livelocked...
3321 */
3322 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3323 jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3324 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3325 }
3326
3327 path_put(&nd.path);
3328 return vfs_quota_on(sb, type, format_id, path, remount);
3329 }
3330
3331 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3332 * acquiring the locks... As quota files are never truncated and quota code
3333 * itself serializes the operations (and noone else should touch the files)
3334 * we don't have to be afraid of races */
3335 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3336 size_t len, loff_t off)
3337 {
3338 struct inode *inode = sb_dqopt(sb)->files[type];
3339 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3340 int err = 0;
3341 int offset = off & (sb->s_blocksize - 1);
3342 int tocopy;
3343 size_t toread;
3344 struct buffer_head *bh;
3345 loff_t i_size = i_size_read(inode);
3346
3347 if (off > i_size)
3348 return 0;
3349 if (off+len > i_size)
3350 len = i_size-off;
3351 toread = len;
3352 while (toread > 0) {
3353 tocopy = sb->s_blocksize - offset < toread ?
3354 sb->s_blocksize - offset : toread;
3355 bh = ext4_bread(NULL, inode, blk, 0, &err);
3356 if (err)
3357 return err;
3358 if (!bh) /* A hole? */
3359 memset(data, 0, tocopy);
3360 else
3361 memcpy(data, bh->b_data+offset, tocopy);
3362 brelse(bh);
3363 offset = 0;
3364 toread -= tocopy;
3365 data += tocopy;
3366 blk++;
3367 }
3368 return len;
3369 }
3370
3371 /* Write to quotafile (we know the transaction is already started and has
3372 * enough credits) */
3373 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3374 const char *data, size_t len, loff_t off)
3375 {
3376 struct inode *inode = sb_dqopt(sb)->files[type];
3377 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3378 int err = 0;
3379 int offset = off & (sb->s_blocksize - 1);
3380 int tocopy;
3381 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3382 size_t towrite = len;
3383 struct buffer_head *bh;
3384 handle_t *handle = journal_current_handle();
3385
3386 if (!handle) {
3387 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3388 " cancelled because transaction is not started.\n",
3389 (unsigned long long)off, (unsigned long long)len);
3390 return -EIO;
3391 }
3392 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3393 while (towrite > 0) {
3394 tocopy = sb->s_blocksize - offset < towrite ?
3395 sb->s_blocksize - offset : towrite;
3396 bh = ext4_bread(handle, inode, blk, 1, &err);
3397 if (!bh)
3398 goto out;
3399 if (journal_quota) {
3400 err = ext4_journal_get_write_access(handle, bh);
3401 if (err) {
3402 brelse(bh);
3403 goto out;
3404 }
3405 }
3406 lock_buffer(bh);
3407 memcpy(bh->b_data+offset, data, tocopy);
3408 flush_dcache_page(bh->b_page);
3409 unlock_buffer(bh);
3410 if (journal_quota)
3411 err = ext4_journal_dirty_metadata(handle, bh);
3412 else {
3413 /* Always do at least ordered writes for quotas */
3414 err = ext4_jbd2_file_inode(handle, inode);
3415 mark_buffer_dirty(bh);
3416 }
3417 brelse(bh);
3418 if (err)
3419 goto out;
3420 offset = 0;
3421 towrite -= tocopy;
3422 data += tocopy;
3423 blk++;
3424 }
3425 out:
3426 if (len == towrite) {
3427 mutex_unlock(&inode->i_mutex);
3428 return err;
3429 }
3430 if (inode->i_size < off+len-towrite) {
3431 i_size_write(inode, off+len-towrite);
3432 EXT4_I(inode)->i_disksize = inode->i_size;
3433 }
3434 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3435 ext4_mark_inode_dirty(handle, inode);
3436 mutex_unlock(&inode->i_mutex);
3437 return len - towrite;
3438 }
3439
3440 #endif
3441
3442 static int ext4_get_sb(struct file_system_type *fs_type,
3443 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3444 {
3445 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3446 }
3447
3448 static struct file_system_type ext4dev_fs_type = {
3449 .owner = THIS_MODULE,
3450 .name = "ext4dev",
3451 .get_sb = ext4_get_sb,
3452 .kill_sb = kill_block_super,
3453 .fs_flags = FS_REQUIRES_DEV,
3454 };
3455
3456 static int __init init_ext4_fs(void)
3457 {
3458 int err;
3459
3460 err = init_ext4_mballoc();
3461 if (err)
3462 return err;
3463
3464 err = init_ext4_xattr();
3465 if (err)
3466 goto out2;
3467 err = init_inodecache();
3468 if (err)
3469 goto out1;
3470 err = register_filesystem(&ext4dev_fs_type);
3471 if (err)
3472 goto out;
3473 return 0;
3474 out:
3475 destroy_inodecache();
3476 out1:
3477 exit_ext4_xattr();
3478 out2:
3479 exit_ext4_mballoc();
3480 return err;
3481 }
3482
3483 static void __exit exit_ext4_fs(void)
3484 {
3485 unregister_filesystem(&ext4dev_fs_type);
3486 destroy_inodecache();
3487 exit_ext4_xattr();
3488 exit_ext4_mballoc();
3489 }
3490
3491 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3492 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3493 MODULE_LICENSE("GPL");
3494 module_init(init_ext4_fs)
3495 module_exit(exit_ext4_fs)
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree