?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
H8/300: Fix misnamed "CONFIG_BLKDEV_RESERVE_ADDRESS" Kconfig variable
[firewire-audio.git] / fs / ext4 / super.c
blob3c1397fa83df075a6c3e6cfb7df3929a7ebbd563
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/ext4_fs.h>
25 #include <linux/ext4_jbd2.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/exportfs.h>
33 #include <linux/vfs.h>
34 #include <linux/random.h>
35 #include <linux/mount.h>
36 #include <linux/namei.h>
37 #include <linux/quotaops.h>
38 #include <linux/seq_file.h>
39 #include <linux/log2.h>
40
41 #include <asm/uaccess.h>
42
43 #include "xattr.h"
44 #include "acl.h"
45 #include "namei.h"
46
47 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
48 unsigned long journal_devnum);
49 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
50 unsigned int);
51 static void ext4_commit_super (struct super_block * sb,
52 struct ext4_super_block * es,
53 int sync);
54 static void ext4_mark_recovery_complete(struct super_block * sb,
55 struct ext4_super_block * es);
56 static void ext4_clear_journal_err(struct super_block * sb,
57 struct ext4_super_block * es);
58 static int ext4_sync_fs(struct super_block *sb, int wait);
59 static const char *ext4_decode_error(struct super_block * sb, int errno,
60 char nbuf[16]);
61 static int ext4_remount (struct super_block * sb, int * flags, char * data);
62 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
63 static void ext4_unlockfs(struct super_block *sb);
64 static void ext4_write_super (struct super_block * sb);
65 static void ext4_write_super_lockfs(struct super_block *sb);
66
67
68 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
69 struct ext4_group_desc *bg)
70 {
71 return le32_to_cpu(bg->bg_block_bitmap) |
72 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
73 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
74 }
75
76 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
77 struct ext4_group_desc *bg)
78 {
79 return le32_to_cpu(bg->bg_inode_bitmap) |
80 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
81 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
82 }
83
84 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
85 struct ext4_group_desc *bg)
86 {
87 return le32_to_cpu(bg->bg_inode_table) |
88 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
89 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
90 }
91
92 void ext4_block_bitmap_set(struct super_block *sb,
93 struct ext4_group_desc *bg, ext4_fsblk_t blk)
94 {
95 bg->bg_block_bitmap = cpu_to_le32((u32)blk);
96 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
97 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
98 }
99
100 void ext4_inode_bitmap_set(struct super_block *sb,
101 struct ext4_group_desc *bg, ext4_fsblk_t blk)
102 {
103 bg->bg_inode_bitmap = cpu_to_le32((u32)blk);
104 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
105 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
106 }
107
108 void ext4_inode_table_set(struct super_block *sb,
109 struct ext4_group_desc *bg, ext4_fsblk_t blk)
110 {
111 bg->bg_inode_table = cpu_to_le32((u32)blk);
112 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
113 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
114 }
115
116 /*
117 * Wrappers for jbd2_journal_start/end.
118 *
119 * The only special thing we need to do here is to make sure that all
120 * journal_end calls result in the superblock being marked dirty, so
121 * that sync() will call the filesystem's write_super callback if
122 * appropriate.
123 */
124 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
125 {
126 journal_t *journal;
127
128 if (sb->s_flags & MS_RDONLY)
129 return ERR_PTR(-EROFS);
130
131 /* Special case here: if the journal has aborted behind our
132 * backs (eg. EIO in the commit thread), then we still need to
133 * take the FS itself readonly cleanly. */
134 journal = EXT4_SB(sb)->s_journal;
135 if (is_journal_aborted(journal)) {
136 ext4_abort(sb, __FUNCTION__,
137 "Detected aborted journal");
138 return ERR_PTR(-EROFS);
139 }
140
141 return jbd2_journal_start(journal, nblocks);
142 }
143
144 /*
145 * The only special thing we need to do here is to make sure that all
146 * jbd2_journal_stop calls result in the superblock being marked dirty, so
147 * that sync() will call the filesystem's write_super callback if
148 * appropriate.
149 */
150 int __ext4_journal_stop(const char *where, handle_t *handle)
151 {
152 struct super_block *sb;
153 int err;
154 int rc;
155
156 sb = handle->h_transaction->t_journal->j_private;
157 err = handle->h_err;
158 rc = jbd2_journal_stop(handle);
159
160 if (!err)
161 err = rc;
162 if (err)
163 __ext4_std_error(sb, where, err);
164 return err;
165 }
166
167 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
168 struct buffer_head *bh, handle_t *handle, int err)
169 {
170 char nbuf[16];
171 const char *errstr = ext4_decode_error(NULL, err, nbuf);
172
173 if (bh)
174 BUFFER_TRACE(bh, "abort");
175
176 if (!handle->h_err)
177 handle->h_err = err;
178
179 if (is_handle_aborted(handle))
180 return;
181
182 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
183 caller, errstr, err_fn);
184
185 jbd2_journal_abort_handle(handle);
186 }
187
188 /* Deal with the reporting of failure conditions on a filesystem such as
189 * inconsistencies detected or read IO failures.
190 *
191 * On ext2, we can store the error state of the filesystem in the
192 * superblock. That is not possible on ext4, because we may have other
193 * write ordering constraints on the superblock which prevent us from
194 * writing it out straight away; and given that the journal is about to
195 * be aborted, we can't rely on the current, or future, transactions to
196 * write out the superblock safely.
197 *
198 * We'll just use the jbd2_journal_abort() error code to record an error in
199 * the journal instead. On recovery, the journal will compain about
200 * that error until we've noted it down and cleared it.
201 */
202
203 static void ext4_handle_error(struct super_block *sb)
204 {
205 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
206
207 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
208 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
209
210 if (sb->s_flags & MS_RDONLY)
211 return;
212
213 if (!test_opt (sb, ERRORS_CONT)) {
214 journal_t *journal = EXT4_SB(sb)->s_journal;
215
216 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
217 if (journal)
218 jbd2_journal_abort(journal, -EIO);
219 }
220 if (test_opt (sb, ERRORS_RO)) {
221 printk (KERN_CRIT "Remounting filesystem read-only\n");
222 sb->s_flags |= MS_RDONLY;
223 }
224 ext4_commit_super(sb, es, 1);
225 if (test_opt(sb, ERRORS_PANIC))
226 panic("EXT4-fs (device %s): panic forced after error\n",
227 sb->s_id);
228 }
229
230 void ext4_error (struct super_block * sb, const char * function,
231 const char * fmt, ...)
232 {
233 va_list args;
234
235 va_start(args, fmt);
236 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
237 vprintk(fmt, args);
238 printk("\n");
239 va_end(args);
240
241 ext4_handle_error(sb);
242 }
243
244 static const char *ext4_decode_error(struct super_block * sb, int errno,
245 char nbuf[16])
246 {
247 char *errstr = NULL;
248
249 switch (errno) {
250 case -EIO:
251 errstr = "IO failure";
252 break;
253 case -ENOMEM:
254 errstr = "Out of memory";
255 break;
256 case -EROFS:
257 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
258 errstr = "Journal has aborted";
259 else
260 errstr = "Readonly filesystem";
261 break;
262 default:
263 /* If the caller passed in an extra buffer for unknown
264 * errors, textualise them now. Else we just return
265 * NULL. */
266 if (nbuf) {
267 /* Check for truncated error codes... */
268 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
269 errstr = nbuf;
270 }
271 break;
272 }
273
274 return errstr;
275 }
276
277 /* __ext4_std_error decodes expected errors from journaling functions
278 * automatically and invokes the appropriate error response. */
279
280 void __ext4_std_error (struct super_block * sb, const char * function,
281 int errno)
282 {
283 char nbuf[16];
284 const char *errstr;
285
286 /* Special case: if the error is EROFS, and we're not already
287 * inside a transaction, then there's really no point in logging
288 * an error. */
289 if (errno == -EROFS && journal_current_handle() == NULL &&
290 (sb->s_flags & MS_RDONLY))
291 return;
292
293 errstr = ext4_decode_error(sb, errno, nbuf);
294 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
295 sb->s_id, function, errstr);
296
297 ext4_handle_error(sb);
298 }
299
300 /*
301 * ext4_abort is a much stronger failure handler than ext4_error. The
302 * abort function may be used to deal with unrecoverable failures such
303 * as journal IO errors or ENOMEM at a critical moment in log management.
304 *
305 * We unconditionally force the filesystem into an ABORT|READONLY state,
306 * unless the error response on the fs has been set to panic in which
307 * case we take the easy way out and panic immediately.
308 */
309
310 void ext4_abort (struct super_block * sb, const char * function,
311 const char * fmt, ...)
312 {
313 va_list args;
314
315 printk (KERN_CRIT "ext4_abort called.\n");
316
317 va_start(args, fmt);
318 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
319 vprintk(fmt, args);
320 printk("\n");
321 va_end(args);
322
323 if (test_opt(sb, ERRORS_PANIC))
324 panic("EXT4-fs panic from previous error\n");
325
326 if (sb->s_flags & MS_RDONLY)
327 return;
328
329 printk(KERN_CRIT "Remounting filesystem read-only\n");
330 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
331 sb->s_flags |= MS_RDONLY;
332 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
333 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
334 }
335
336 void ext4_warning (struct super_block * sb, const char * function,
337 const char * fmt, ...)
338 {
339 va_list args;
340
341 va_start(args, fmt);
342 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
343 sb->s_id, function);
344 vprintk(fmt, args);
345 printk("\n");
346 va_end(args);
347 }
348
349 void ext4_update_dynamic_rev(struct super_block *sb)
350 {
351 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
352
353 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
354 return;
355
356 ext4_warning(sb, __FUNCTION__,
357 "updating to rev %d because of new feature flag, "
358 "running e2fsck is recommended",
359 EXT4_DYNAMIC_REV);
360
361 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
362 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
363 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
364 /* leave es->s_feature_*compat flags alone */
365 /* es->s_uuid will be set by e2fsck if empty */
366
367 /*
368 * The rest of the superblock fields should be zero, and if not it
369 * means they are likely already in use, so leave them alone. We
370 * can leave it up to e2fsck to clean up any inconsistencies there.
371 */
372 }
373
374 /*
375 * Open the external journal device
376 */
377 static struct block_device *ext4_blkdev_get(dev_t dev)
378 {
379 struct block_device *bdev;
380 char b[BDEVNAME_SIZE];
381
382 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
383 if (IS_ERR(bdev))
384 goto fail;
385 return bdev;
386
387 fail:
388 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
389 __bdevname(dev, b), PTR_ERR(bdev));
390 return NULL;
391 }
392
393 /*
394 * Release the journal device
395 */
396 static int ext4_blkdev_put(struct block_device *bdev)
397 {
398 bd_release(bdev);
399 return blkdev_put(bdev);
400 }
401
402 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
403 {
404 struct block_device *bdev;
405 int ret = -ENODEV;
406
407 bdev = sbi->journal_bdev;
408 if (bdev) {
409 ret = ext4_blkdev_put(bdev);
410 sbi->journal_bdev = NULL;
411 }
412 return ret;
413 }
414
415 static inline struct inode *orphan_list_entry(struct list_head *l)
416 {
417 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
418 }
419
420 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
421 {
422 struct list_head *l;
423
424 printk(KERN_ERR "sb orphan head is %d\n",
425 le32_to_cpu(sbi->s_es->s_last_orphan));
426
427 printk(KERN_ERR "sb_info orphan list:\n");
428 list_for_each(l, &sbi->s_orphan) {
429 struct inode *inode = orphan_list_entry(l);
430 printk(KERN_ERR " "
431 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
432 inode->i_sb->s_id, inode->i_ino, inode,
433 inode->i_mode, inode->i_nlink,
434 NEXT_ORPHAN(inode));
435 }
436 }
437
438 static void ext4_put_super (struct super_block * sb)
439 {
440 struct ext4_sb_info *sbi = EXT4_SB(sb);
441 struct ext4_super_block *es = sbi->s_es;
442 int i;
443
444 ext4_ext_release(sb);
445 ext4_xattr_put_super(sb);
446 jbd2_journal_destroy(sbi->s_journal);
447 if (!(sb->s_flags & MS_RDONLY)) {
448 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
449 es->s_state = cpu_to_le16(sbi->s_mount_state);
450 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
451 mark_buffer_dirty(sbi->s_sbh);
452 ext4_commit_super(sb, es, 1);
453 }
454
455 for (i = 0; i < sbi->s_gdb_count; i++)
456 brelse(sbi->s_group_desc[i]);
457 kfree(sbi->s_group_desc);
458 percpu_counter_destroy(&sbi->s_freeblocks_counter);
459 percpu_counter_destroy(&sbi->s_freeinodes_counter);
460 percpu_counter_destroy(&sbi->s_dirs_counter);
461 brelse(sbi->s_sbh);
462 #ifdef CONFIG_QUOTA
463 for (i = 0; i < MAXQUOTAS; i++)
464 kfree(sbi->s_qf_names[i]);
465 #endif
466
467 /* Debugging code just in case the in-memory inode orphan list
468 * isn't empty. The on-disk one can be non-empty if we've
469 * detected an error and taken the fs readonly, but the
470 * in-memory list had better be clean by this point. */
471 if (!list_empty(&sbi->s_orphan))
472 dump_orphan_list(sb, sbi);
473 J_ASSERT(list_empty(&sbi->s_orphan));
474
475 invalidate_bdev(sb->s_bdev);
476 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
477 /*
478 * Invalidate the journal device's buffers. We don't want them
479 * floating about in memory - the physical journal device may
480 * hotswapped, and it breaks the `ro-after' testing code.
481 */
482 sync_blockdev(sbi->journal_bdev);
483 invalidate_bdev(sbi->journal_bdev);
484 ext4_blkdev_remove(sbi);
485 }
486 sb->s_fs_info = NULL;
487 kfree(sbi);
488 return;
489 }
490
491 static struct kmem_cache *ext4_inode_cachep;
492
493 /*
494 * Called inside transaction, so use GFP_NOFS
495 */
496 static struct inode *ext4_alloc_inode(struct super_block *sb)
497 {
498 struct ext4_inode_info *ei;
499
500 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
501 if (!ei)
502 return NULL;
503 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
504 ei->i_acl = EXT4_ACL_NOT_CACHED;
505 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
506 #endif
507 ei->i_block_alloc_info = NULL;
508 ei->vfs_inode.i_version = 1;
509 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
510 return &ei->vfs_inode;
511 }
512
513 static void ext4_destroy_inode(struct inode *inode)
514 {
515 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
516 printk("EXT4 Inode %p: orphan list check failed!\n",
517 EXT4_I(inode));
518 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
519 EXT4_I(inode), sizeof(struct ext4_inode_info),
520 true);
521 dump_stack();
522 }
523 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
524 }
525
526 static void init_once(void * foo, struct kmem_cache * cachep, unsigned long flags)
527 {
528 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
529
530 INIT_LIST_HEAD(&ei->i_orphan);
531 #ifdef CONFIG_EXT4DEV_FS_XATTR
532 init_rwsem(&ei->xattr_sem);
533 #endif
534 mutex_init(&ei->truncate_mutex);
535 inode_init_once(&ei->vfs_inode);
536 }
537
538 static int init_inodecache(void)
539 {
540 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
541 sizeof(struct ext4_inode_info),
542 0, (SLAB_RECLAIM_ACCOUNT|
543 SLAB_MEM_SPREAD),
544 init_once);
545 if (ext4_inode_cachep == NULL)
546 return -ENOMEM;
547 return 0;
548 }
549
550 static void destroy_inodecache(void)
551 {
552 kmem_cache_destroy(ext4_inode_cachep);
553 }
554
555 static void ext4_clear_inode(struct inode *inode)
556 {
557 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
558 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
559 if (EXT4_I(inode)->i_acl &&
560 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
561 posix_acl_release(EXT4_I(inode)->i_acl);
562 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
563 }
564 if (EXT4_I(inode)->i_default_acl &&
565 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
566 posix_acl_release(EXT4_I(inode)->i_default_acl);
567 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
568 }
569 #endif
570 ext4_discard_reservation(inode);
571 EXT4_I(inode)->i_block_alloc_info = NULL;
572 if (unlikely(rsv))
573 kfree(rsv);
574 }
575
576 static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
577 {
578 #if defined(CONFIG_QUOTA)
579 struct ext4_sb_info *sbi = EXT4_SB(sb);
580
581 if (sbi->s_jquota_fmt)
582 seq_printf(seq, ",jqfmt=%s",
583 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
584
585 if (sbi->s_qf_names[USRQUOTA])
586 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
587
588 if (sbi->s_qf_names[GRPQUOTA])
589 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
590
591 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
592 seq_puts(seq, ",usrquota");
593
594 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
595 seq_puts(seq, ",grpquota");
596 #endif
597 }
598
599 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
600 {
601 struct super_block *sb = vfs->mnt_sb;
602
603 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
604 seq_puts(seq, ",data=journal");
605 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
606 seq_puts(seq, ",data=ordered");
607 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
608 seq_puts(seq, ",data=writeback");
609
610 ext4_show_quota_options(seq, sb);
611
612 return 0;
613 }
614
615
616 static struct dentry *ext4_get_dentry(struct super_block *sb, void *vobjp)
617 {
618 __u32 *objp = vobjp;
619 unsigned long ino = objp[0];
620 __u32 generation = objp[1];
621 struct inode *inode;
622 struct dentry *result;
623
624 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
625 return ERR_PTR(-ESTALE);
626 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
627 return ERR_PTR(-ESTALE);
628
629 /* iget isn't really right if the inode is currently unallocated!!
630 *
631 * ext4_read_inode will return a bad_inode if the inode had been
632 * deleted, so we should be safe.
633 *
634 * Currently we don't know the generation for parent directory, so
635 * a generation of 0 means "accept any"
636 */
637 inode = iget(sb, ino);
638 if (inode == NULL)
639 return ERR_PTR(-ENOMEM);
640 if (is_bad_inode(inode) ||
641 (generation && inode->i_generation != generation)) {
642 iput(inode);
643 return ERR_PTR(-ESTALE);
644 }
645 /* now to find a dentry.
646 * If possible, get a well-connected one
647 */
648 result = d_alloc_anon(inode);
649 if (!result) {
650 iput(inode);
651 return ERR_PTR(-ENOMEM);
652 }
653 return result;
654 }
655
656 #ifdef CONFIG_QUOTA
657 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
658 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
659
660 static int ext4_dquot_initialize(struct inode *inode, int type);
661 static int ext4_dquot_drop(struct inode *inode);
662 static int ext4_write_dquot(struct dquot *dquot);
663 static int ext4_acquire_dquot(struct dquot *dquot);
664 static int ext4_release_dquot(struct dquot *dquot);
665 static int ext4_mark_dquot_dirty(struct dquot *dquot);
666 static int ext4_write_info(struct super_block *sb, int type);
667 static int ext4_quota_on(struct super_block *sb, int type, int format_id, char *path);
668 static int ext4_quota_on_mount(struct super_block *sb, int type);
669 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
670 size_t len, loff_t off);
671 static ssize_t ext4_quota_write(struct super_block *sb, int type,
672 const char *data, size_t len, loff_t off);
673
674 static struct dquot_operations ext4_quota_operations = {
675 .initialize = ext4_dquot_initialize,
676 .drop = ext4_dquot_drop,
677 .alloc_space = dquot_alloc_space,
678 .alloc_inode = dquot_alloc_inode,
679 .free_space = dquot_free_space,
680 .free_inode = dquot_free_inode,
681 .transfer = dquot_transfer,
682 .write_dquot = ext4_write_dquot,
683 .acquire_dquot = ext4_acquire_dquot,
684 .release_dquot = ext4_release_dquot,
685 .mark_dirty = ext4_mark_dquot_dirty,
686 .write_info = ext4_write_info
687 };
688
689 static struct quotactl_ops ext4_qctl_operations = {
690 .quota_on = ext4_quota_on,
691 .quota_off = vfs_quota_off,
692 .quota_sync = vfs_quota_sync,
693 .get_info = vfs_get_dqinfo,
694 .set_info = vfs_set_dqinfo,
695 .get_dqblk = vfs_get_dqblk,
696 .set_dqblk = vfs_set_dqblk
697 };
698 #endif
699
700 static const struct super_operations ext4_sops = {
701 .alloc_inode = ext4_alloc_inode,
702 .destroy_inode = ext4_destroy_inode,
703 .read_inode = ext4_read_inode,
704 .write_inode = ext4_write_inode,
705 .dirty_inode = ext4_dirty_inode,
706 .delete_inode = ext4_delete_inode,
707 .put_super = ext4_put_super,
708 .write_super = ext4_write_super,
709 .sync_fs = ext4_sync_fs,
710 .write_super_lockfs = ext4_write_super_lockfs,
711 .unlockfs = ext4_unlockfs,
712 .statfs = ext4_statfs,
713 .remount_fs = ext4_remount,
714 .clear_inode = ext4_clear_inode,
715 .show_options = ext4_show_options,
716 #ifdef CONFIG_QUOTA
717 .quota_read = ext4_quota_read,
718 .quota_write = ext4_quota_write,
719 #endif
720 };
721
722 static struct export_operations ext4_export_ops = {
723 .get_parent = ext4_get_parent,
724 .get_dentry = ext4_get_dentry,
725 };
726
727 enum {
728 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
729 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
730 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
731 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
732 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
733 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
734 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
735 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
736 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
737 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
738 Opt_grpquota, Opt_extents, Opt_noextents,
739 };
740
741 static match_table_t tokens = {
742 {Opt_bsd_df, "bsddf"},
743 {Opt_minix_df, "minixdf"},
744 {Opt_grpid, "grpid"},
745 {Opt_grpid, "bsdgroups"},
746 {Opt_nogrpid, "nogrpid"},
747 {Opt_nogrpid, "sysvgroups"},
748 {Opt_resgid, "resgid=%u"},
749 {Opt_resuid, "resuid=%u"},
750 {Opt_sb, "sb=%u"},
751 {Opt_err_cont, "errors=continue"},
752 {Opt_err_panic, "errors=panic"},
753 {Opt_err_ro, "errors=remount-ro"},
754 {Opt_nouid32, "nouid32"},
755 {Opt_nocheck, "nocheck"},
756 {Opt_nocheck, "check=none"},
757 {Opt_debug, "debug"},
758 {Opt_oldalloc, "oldalloc"},
759 {Opt_orlov, "orlov"},
760 {Opt_user_xattr, "user_xattr"},
761 {Opt_nouser_xattr, "nouser_xattr"},
762 {Opt_acl, "acl"},
763 {Opt_noacl, "noacl"},
764 {Opt_reservation, "reservation"},
765 {Opt_noreservation, "noreservation"},
766 {Opt_noload, "noload"},
767 {Opt_nobh, "nobh"},
768 {Opt_bh, "bh"},
769 {Opt_commit, "commit=%u"},
770 {Opt_journal_update, "journal=update"},
771 {Opt_journal_inum, "journal=%u"},
772 {Opt_journal_dev, "journal_dev=%u"},
773 {Opt_abort, "abort"},
774 {Opt_data_journal, "data=journal"},
775 {Opt_data_ordered, "data=ordered"},
776 {Opt_data_writeback, "data=writeback"},
777 {Opt_offusrjquota, "usrjquota="},
778 {Opt_usrjquota, "usrjquota=%s"},
779 {Opt_offgrpjquota, "grpjquota="},
780 {Opt_grpjquota, "grpjquota=%s"},
781 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
782 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
783 {Opt_grpquota, "grpquota"},
784 {Opt_noquota, "noquota"},
785 {Opt_quota, "quota"},
786 {Opt_usrquota, "usrquota"},
787 {Opt_barrier, "barrier=%u"},
788 {Opt_extents, "extents"},
789 {Opt_noextents, "noextents"},
790 {Opt_err, NULL},
791 {Opt_resize, "resize"},
792 };
793
794 static ext4_fsblk_t get_sb_block(void **data)
795 {
796 ext4_fsblk_t sb_block;
797 char *options = (char *) *data;
798
799 if (!options || strncmp(options, "sb=", 3) != 0)
800 return 1; /* Default location */
801 options += 3;
802 /*todo: use simple_strtoll with >32bit ext4 */
803 sb_block = simple_strtoul(options, &options, 0);
804 if (*options && *options != ',') {
805 printk("EXT4-fs: Invalid sb specification: %s\n",
806 (char *) *data);
807 return 1;
808 }
809 if (*options == ',')
810 options++;
811 *data = (void *) options;
812 return sb_block;
813 }
814
815 static int parse_options (char *options, struct super_block *sb,
816 unsigned int *inum, unsigned long *journal_devnum,
817 ext4_fsblk_t *n_blocks_count, int is_remount)
818 {
819 struct ext4_sb_info *sbi = EXT4_SB(sb);
820 char * p;
821 substring_t args[MAX_OPT_ARGS];
822 int data_opt = 0;
823 int option;
824 #ifdef CONFIG_QUOTA
825 int qtype;
826 char *qname;
827 #endif
828
829 if (!options)
830 return 1;
831
832 while ((p = strsep (&options, ",")) != NULL) {
833 int token;
834 if (!*p)
835 continue;
836
837 token = match_token(p, tokens, args);
838 switch (token) {
839 case Opt_bsd_df:
840 clear_opt (sbi->s_mount_opt, MINIX_DF);
841 break;
842 case Opt_minix_df:
843 set_opt (sbi->s_mount_opt, MINIX_DF);
844 break;
845 case Opt_grpid:
846 set_opt (sbi->s_mount_opt, GRPID);
847 break;
848 case Opt_nogrpid:
849 clear_opt (sbi->s_mount_opt, GRPID);
850 break;
851 case Opt_resuid:
852 if (match_int(&args[0], &option))
853 return 0;
854 sbi->s_resuid = option;
855 break;
856 case Opt_resgid:
857 if (match_int(&args[0], &option))
858 return 0;
859 sbi->s_resgid = option;
860 break;
861 case Opt_sb:
862 /* handled by get_sb_block() instead of here */
863 /* *sb_block = match_int(&args[0]); */
864 break;
865 case Opt_err_panic:
866 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
867 clear_opt (sbi->s_mount_opt, ERRORS_RO);
868 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
869 break;
870 case Opt_err_ro:
871 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
872 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
873 set_opt (sbi->s_mount_opt, ERRORS_RO);
874 break;
875 case Opt_err_cont:
876 clear_opt (sbi->s_mount_opt, ERRORS_RO);
877 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
878 set_opt (sbi->s_mount_opt, ERRORS_CONT);
879 break;
880 case Opt_nouid32:
881 set_opt (sbi->s_mount_opt, NO_UID32);
882 break;
883 case Opt_nocheck:
884 clear_opt (sbi->s_mount_opt, CHECK);
885 break;
886 case Opt_debug:
887 set_opt (sbi->s_mount_opt, DEBUG);
888 break;
889 case Opt_oldalloc:
890 set_opt (sbi->s_mount_opt, OLDALLOC);
891 break;
892 case Opt_orlov:
893 clear_opt (sbi->s_mount_opt, OLDALLOC);
894 break;
895 #ifdef CONFIG_EXT4DEV_FS_XATTR
896 case Opt_user_xattr:
897 set_opt (sbi->s_mount_opt, XATTR_USER);
898 break;
899 case Opt_nouser_xattr:
900 clear_opt (sbi->s_mount_opt, XATTR_USER);
901 break;
902 #else
903 case Opt_user_xattr:
904 case Opt_nouser_xattr:
905 printk("EXT4 (no)user_xattr options not supported\n");
906 break;
907 #endif
908 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
909 case Opt_acl:
910 set_opt(sbi->s_mount_opt, POSIX_ACL);
911 break;
912 case Opt_noacl:
913 clear_opt(sbi->s_mount_opt, POSIX_ACL);
914 break;
915 #else
916 case Opt_acl:
917 case Opt_noacl:
918 printk("EXT4 (no)acl options not supported\n");
919 break;
920 #endif
921 case Opt_reservation:
922 set_opt(sbi->s_mount_opt, RESERVATION);
923 break;
924 case Opt_noreservation:
925 clear_opt(sbi->s_mount_opt, RESERVATION);
926 break;
927 case Opt_journal_update:
928 /* @@@ FIXME */
929 /* Eventually we will want to be able to create
930 a journal file here. For now, only allow the
931 user to specify an existing inode to be the
932 journal file. */
933 if (is_remount) {
934 printk(KERN_ERR "EXT4-fs: cannot specify "
935 "journal on remount\n");
936 return 0;
937 }
938 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
939 break;
940 case Opt_journal_inum:
941 if (is_remount) {
942 printk(KERN_ERR "EXT4-fs: cannot specify "
943 "journal on remount\n");
944 return 0;
945 }
946 if (match_int(&args[0], &option))
947 return 0;
948 *inum = option;
949 break;
950 case Opt_journal_dev:
951 if (is_remount) {
952 printk(KERN_ERR "EXT4-fs: cannot specify "
953 "journal on remount\n");
954 return 0;
955 }
956 if (match_int(&args[0], &option))
957 return 0;
958 *journal_devnum = option;
959 break;
960 case Opt_noload:
961 set_opt (sbi->s_mount_opt, NOLOAD);
962 break;
963 case Opt_commit:
964 if (match_int(&args[0], &option))
965 return 0;
966 if (option < 0)
967 return 0;
968 if (option == 0)
969 option = JBD_DEFAULT_MAX_COMMIT_AGE;
970 sbi->s_commit_interval = HZ * option;
971 break;
972 case Opt_data_journal:
973 data_opt = EXT4_MOUNT_JOURNAL_DATA;
974 goto datacheck;
975 case Opt_data_ordered:
976 data_opt = EXT4_MOUNT_ORDERED_DATA;
977 goto datacheck;
978 case Opt_data_writeback:
979 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
980 datacheck:
981 if (is_remount) {
982 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
983 != data_opt) {
984 printk(KERN_ERR
985 "EXT4-fs: cannot change data "
986 "mode on remount\n");
987 return 0;
988 }
989 } else {
990 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
991 sbi->s_mount_opt |= data_opt;
992 }
993 break;
994 #ifdef CONFIG_QUOTA
995 case Opt_usrjquota:
996 qtype = USRQUOTA;
997 goto set_qf_name;
998 case Opt_grpjquota:
999 qtype = GRPQUOTA;
1000 set_qf_name:
1001 if (sb_any_quota_enabled(sb)) {
1002 printk(KERN_ERR
1003 "EXT4-fs: Cannot change journalled "
1004 "quota options when quota turned on.\n");
1005 return 0;
1006 }
1007 qname = match_strdup(&args[0]);
1008 if (!qname) {
1009 printk(KERN_ERR
1010 "EXT4-fs: not enough memory for "
1011 "storing quotafile name.\n");
1012 return 0;
1013 }
1014 if (sbi->s_qf_names[qtype] &&
1015 strcmp(sbi->s_qf_names[qtype], qname)) {
1016 printk(KERN_ERR
1017 "EXT4-fs: %s quota file already "
1018 "specified.\n", QTYPE2NAME(qtype));
1019 kfree(qname);
1020 return 0;
1021 }
1022 sbi->s_qf_names[qtype] = qname;
1023 if (strchr(sbi->s_qf_names[qtype], '/')) {
1024 printk(KERN_ERR
1025 "EXT4-fs: quotafile must be on "
1026 "filesystem root.\n");
1027 kfree(sbi->s_qf_names[qtype]);
1028 sbi->s_qf_names[qtype] = NULL;
1029 return 0;
1030 }
1031 set_opt(sbi->s_mount_opt, QUOTA);
1032 break;
1033 case Opt_offusrjquota:
1034 qtype = USRQUOTA;
1035 goto clear_qf_name;
1036 case Opt_offgrpjquota:
1037 qtype = GRPQUOTA;
1038 clear_qf_name:
1039 if (sb_any_quota_enabled(sb)) {
1040 printk(KERN_ERR "EXT4-fs: Cannot change "
1041 "journalled quota options when "
1042 "quota turned on.\n");
1043 return 0;
1044 }
1045 /*
1046 * The space will be released later when all options
1047 * are confirmed to be correct
1048 */
1049 sbi->s_qf_names[qtype] = NULL;
1050 break;
1051 case Opt_jqfmt_vfsold:
1052 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1053 break;
1054 case Opt_jqfmt_vfsv0:
1055 sbi->s_jquota_fmt = QFMT_VFS_V0;
1056 break;
1057 case Opt_quota:
1058 case Opt_usrquota:
1059 set_opt(sbi->s_mount_opt, QUOTA);
1060 set_opt(sbi->s_mount_opt, USRQUOTA);
1061 break;
1062 case Opt_grpquota:
1063 set_opt(sbi->s_mount_opt, QUOTA);
1064 set_opt(sbi->s_mount_opt, GRPQUOTA);
1065 break;
1066 case Opt_noquota:
1067 if (sb_any_quota_enabled(sb)) {
1068 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1069 "options when quota turned on.\n");
1070 return 0;
1071 }
1072 clear_opt(sbi->s_mount_opt, QUOTA);
1073 clear_opt(sbi->s_mount_opt, USRQUOTA);
1074 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1075 break;
1076 #else
1077 case Opt_quota:
1078 case Opt_usrquota:
1079 case Opt_grpquota:
1080 case Opt_usrjquota:
1081 case Opt_grpjquota:
1082 case Opt_offusrjquota:
1083 case Opt_offgrpjquota:
1084 case Opt_jqfmt_vfsold:
1085 case Opt_jqfmt_vfsv0:
1086 printk(KERN_ERR
1087 "EXT4-fs: journalled quota options not "
1088 "supported.\n");
1089 break;
1090 case Opt_noquota:
1091 break;
1092 #endif
1093 case Opt_abort:
1094 set_opt(sbi->s_mount_opt, ABORT);
1095 break;
1096 case Opt_barrier:
1097 if (match_int(&args[0], &option))
1098 return 0;
1099 if (option)
1100 set_opt(sbi->s_mount_opt, BARRIER);
1101 else
1102 clear_opt(sbi->s_mount_opt, BARRIER);
1103 break;
1104 case Opt_ignore:
1105 break;
1106 case Opt_resize:
1107 if (!is_remount) {
1108 printk("EXT4-fs: resize option only available "
1109 "for remount\n");
1110 return 0;
1111 }
1112 if (match_int(&args[0], &option) != 0)
1113 return 0;
1114 *n_blocks_count = option;
1115 break;
1116 case Opt_nobh:
1117 set_opt(sbi->s_mount_opt, NOBH);
1118 break;
1119 case Opt_bh:
1120 clear_opt(sbi->s_mount_opt, NOBH);
1121 break;
1122 case Opt_extents:
1123 set_opt (sbi->s_mount_opt, EXTENTS);
1124 break;
1125 case Opt_noextents:
1126 clear_opt (sbi->s_mount_opt, EXTENTS);
1127 break;
1128 default:
1129 printk (KERN_ERR
1130 "EXT4-fs: Unrecognized mount option \"%s\" "
1131 "or missing value\n", p);
1132 return 0;
1133 }
1134 }
1135 #ifdef CONFIG_QUOTA
1136 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1137 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1138 sbi->s_qf_names[USRQUOTA])
1139 clear_opt(sbi->s_mount_opt, USRQUOTA);
1140
1141 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1142 sbi->s_qf_names[GRPQUOTA])
1143 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1144
1145 if ((sbi->s_qf_names[USRQUOTA] &&
1146 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1147 (sbi->s_qf_names[GRPQUOTA] &&
1148 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1149 printk(KERN_ERR "EXT4-fs: old and new quota "
1150 "format mixing.\n");
1151 return 0;
1152 }
1153
1154 if (!sbi->s_jquota_fmt) {
1155 printk(KERN_ERR "EXT4-fs: journalled quota format "
1156 "not specified.\n");
1157 return 0;
1158 }
1159 } else {
1160 if (sbi->s_jquota_fmt) {
1161 printk(KERN_ERR "EXT4-fs: journalled quota format "
1162 "specified with no journalling "
1163 "enabled.\n");
1164 return 0;
1165 }
1166 }
1167 #endif
1168 return 1;
1169 }
1170
1171 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1172 int read_only)
1173 {
1174 struct ext4_sb_info *sbi = EXT4_SB(sb);
1175 int res = 0;
1176
1177 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1178 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1179 "forcing read-only mode\n");
1180 res = MS_RDONLY;
1181 }
1182 if (read_only)
1183 return res;
1184 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1185 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1186 "running e2fsck is recommended\n");
1187 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1188 printk (KERN_WARNING
1189 "EXT4-fs warning: mounting fs with errors, "
1190 "running e2fsck is recommended\n");
1191 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1192 le16_to_cpu(es->s_mnt_count) >=
1193 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1194 printk (KERN_WARNING
1195 "EXT4-fs warning: maximal mount count reached, "
1196 "running e2fsck is recommended\n");
1197 else if (le32_to_cpu(es->s_checkinterval) &&
1198 (le32_to_cpu(es->s_lastcheck) +
1199 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1200 printk (KERN_WARNING
1201 "EXT4-fs warning: checktime reached, "
1202 "running e2fsck is recommended\n");
1203 #if 0
1204 /* @@@ We _will_ want to clear the valid bit if we find
1205 * inconsistencies, to force a fsck at reboot. But for
1206 * a plain journaled filesystem we can keep it set as
1207 * valid forever! :)
1208 */
1209 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
1210 #endif
1211 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1212 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1213 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1214 es->s_mtime = cpu_to_le32(get_seconds());
1215 ext4_update_dynamic_rev(sb);
1216 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1217
1218 ext4_commit_super(sb, es, 1);
1219 if (test_opt(sb, DEBUG))
1220 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1221 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1222 sb->s_blocksize,
1223 sbi->s_groups_count,
1224 EXT4_BLOCKS_PER_GROUP(sb),
1225 EXT4_INODES_PER_GROUP(sb),
1226 sbi->s_mount_opt);
1227
1228 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1229 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1230 char b[BDEVNAME_SIZE];
1231
1232 printk("external journal on %s\n",
1233 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1234 } else {
1235 printk("internal journal\n");
1236 }
1237 return res;
1238 }
1239
1240 /* Called at mount-time, super-block is locked */
1241 static int ext4_check_descriptors (struct super_block * sb)
1242 {
1243 struct ext4_sb_info *sbi = EXT4_SB(sb);
1244 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1245 ext4_fsblk_t last_block;
1246 ext4_fsblk_t block_bitmap;
1247 ext4_fsblk_t inode_bitmap;
1248 ext4_fsblk_t inode_table;
1249 struct ext4_group_desc * gdp = NULL;
1250 int desc_block = 0;
1251 int i;
1252
1253 ext4_debug ("Checking group descriptors");
1254
1255 for (i = 0; i < sbi->s_groups_count; i++)
1256 {
1257 if (i == sbi->s_groups_count - 1)
1258 last_block = ext4_blocks_count(sbi->s_es) - 1;
1259 else
1260 last_block = first_block +
1261 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1262
1263 if ((i % EXT4_DESC_PER_BLOCK(sb)) == 0)
1264 gdp = (struct ext4_group_desc *)
1265 sbi->s_group_desc[desc_block++]->b_data;
1266 block_bitmap = ext4_block_bitmap(sb, gdp);
1267 if (block_bitmap < first_block || block_bitmap > last_block)
1268 {
1269 ext4_error (sb, "ext4_check_descriptors",
1270 "Block bitmap for group %d"
1271 " not in group (block %llu)!",
1272 i, block_bitmap);
1273 return 0;
1274 }
1275 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1276 if (inode_bitmap < first_block || inode_bitmap > last_block)
1277 {
1278 ext4_error (sb, "ext4_check_descriptors",
1279 "Inode bitmap for group %d"
1280 " not in group (block %llu)!",
1281 i, inode_bitmap);
1282 return 0;
1283 }
1284 inode_table = ext4_inode_table(sb, gdp);
1285 if (inode_table < first_block ||
1286 inode_table + sbi->s_itb_per_group - 1 > last_block)
1287 {
1288 ext4_error (sb, "ext4_check_descriptors",
1289 "Inode table for group %d"
1290 " not in group (block %llu)!",
1291 i, inode_table);
1292 return 0;
1293 }
1294 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1295 gdp = (struct ext4_group_desc *)
1296 ((__u8 *)gdp + EXT4_DESC_SIZE(sb));
1297 }
1298
1299 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1300 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1301 return 1;
1302 }
1303
1304
1305 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1306 * the superblock) which were deleted from all directories, but held open by
1307 * a process at the time of a crash. We walk the list and try to delete these
1308 * inodes at recovery time (only with a read-write filesystem).
1309 *
1310 * In order to keep the orphan inode chain consistent during traversal (in
1311 * case of crash during recovery), we link each inode into the superblock
1312 * orphan list_head and handle it the same way as an inode deletion during
1313 * normal operation (which journals the operations for us).
1314 *
1315 * We only do an iget() and an iput() on each inode, which is very safe if we
1316 * accidentally point at an in-use or already deleted inode. The worst that
1317 * can happen in this case is that we get a "bit already cleared" message from
1318 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1319 * e2fsck was run on this filesystem, and it must have already done the orphan
1320 * inode cleanup for us, so we can safely abort without any further action.
1321 */
1322 static void ext4_orphan_cleanup (struct super_block * sb,
1323 struct ext4_super_block * es)
1324 {
1325 unsigned int s_flags = sb->s_flags;
1326 int nr_orphans = 0, nr_truncates = 0;
1327 #ifdef CONFIG_QUOTA
1328 int i;
1329 #endif
1330 if (!es->s_last_orphan) {
1331 jbd_debug(4, "no orphan inodes to clean up\n");
1332 return;
1333 }
1334
1335 if (bdev_read_only(sb->s_bdev)) {
1336 printk(KERN_ERR "EXT4-fs: write access "
1337 "unavailable, skipping orphan cleanup.\n");
1338 return;
1339 }
1340
1341 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1342 if (es->s_last_orphan)
1343 jbd_debug(1, "Errors on filesystem, "
1344 "clearing orphan list.\n");
1345 es->s_last_orphan = 0;
1346 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1347 return;
1348 }
1349
1350 if (s_flags & MS_RDONLY) {
1351 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1352 sb->s_id);
1353 sb->s_flags &= ~MS_RDONLY;
1354 }
1355 #ifdef CONFIG_QUOTA
1356 /* Needed for iput() to work correctly and not trash data */
1357 sb->s_flags |= MS_ACTIVE;
1358 /* Turn on quotas so that they are updated correctly */
1359 for (i = 0; i < MAXQUOTAS; i++) {
1360 if (EXT4_SB(sb)->s_qf_names[i]) {
1361 int ret = ext4_quota_on_mount(sb, i);
1362 if (ret < 0)
1363 printk(KERN_ERR
1364 "EXT4-fs: Cannot turn on journalled "
1365 "quota: error %d\n", ret);
1366 }
1367 }
1368 #endif
1369
1370 while (es->s_last_orphan) {
1371 struct inode *inode;
1372
1373 if (!(inode =
1374 ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1375 es->s_last_orphan = 0;
1376 break;
1377 }
1378
1379 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1380 DQUOT_INIT(inode);
1381 if (inode->i_nlink) {
1382 printk(KERN_DEBUG
1383 "%s: truncating inode %lu to %Ld bytes\n",
1384 __FUNCTION__, inode->i_ino, inode->i_size);
1385 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1386 inode->i_ino, inode->i_size);
1387 ext4_truncate(inode);
1388 nr_truncates++;
1389 } else {
1390 printk(KERN_DEBUG
1391 "%s: deleting unreferenced inode %lu\n",
1392 __FUNCTION__, inode->i_ino);
1393 jbd_debug(2, "deleting unreferenced inode %lu\n",
1394 inode->i_ino);
1395 nr_orphans++;
1396 }
1397 iput(inode); /* The delete magic happens here! */
1398 }
1399
1400 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1401
1402 if (nr_orphans)
1403 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1404 sb->s_id, PLURAL(nr_orphans));
1405 if (nr_truncates)
1406 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1407 sb->s_id, PLURAL(nr_truncates));
1408 #ifdef CONFIG_QUOTA
1409 /* Turn quotas off */
1410 for (i = 0; i < MAXQUOTAS; i++) {
1411 if (sb_dqopt(sb)->files[i])
1412 vfs_quota_off(sb, i);
1413 }
1414 #endif
1415 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1416 }
1417
1418 #define log2(n) ffz(~(n))
1419
1420 /*
1421 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1422 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1423 * We need to be 1 filesystem block less than the 2^32 sector limit.
1424 */
1425 static loff_t ext4_max_size(int bits)
1426 {
1427 loff_t res = EXT4_NDIR_BLOCKS;
1428 /* This constant is calculated to be the largest file size for a
1429 * dense, 4k-blocksize file such that the total number of
1430 * sectors in the file, including data and all indirect blocks,
1431 * does not exceed 2^32. */
1432 const loff_t upper_limit = 0x1ff7fffd000LL;
1433
1434 res += 1LL << (bits-2);
1435 res += 1LL << (2*(bits-2));
1436 res += 1LL << (3*(bits-2));
1437 res <<= bits;
1438 if (res > upper_limit)
1439 res = upper_limit;
1440 return res;
1441 }
1442
1443 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1444 ext4_fsblk_t logical_sb_block, int nr)
1445 {
1446 struct ext4_sb_info *sbi = EXT4_SB(sb);
1447 unsigned long bg, first_meta_bg;
1448 int has_super = 0;
1449
1450 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1451
1452 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1453 nr < first_meta_bg)
1454 return logical_sb_block + nr + 1;
1455 bg = sbi->s_desc_per_block * nr;
1456 if (ext4_bg_has_super(sb, bg))
1457 has_super = 1;
1458 return (has_super + ext4_group_first_block_no(sb, bg));
1459 }
1460
1461
1462 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1463 {
1464 struct buffer_head * bh;
1465 struct ext4_super_block *es = NULL;
1466 struct ext4_sb_info *sbi;
1467 ext4_fsblk_t block;
1468 ext4_fsblk_t sb_block = get_sb_block(&data);
1469 ext4_fsblk_t logical_sb_block;
1470 unsigned long offset = 0;
1471 unsigned int journal_inum = 0;
1472 unsigned long journal_devnum = 0;
1473 unsigned long def_mount_opts;
1474 struct inode *root;
1475 int blocksize;
1476 int hblock;
1477 int db_count;
1478 int i;
1479 int needs_recovery;
1480 __le32 features;
1481 __u64 blocks_count;
1482
1483 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1484 if (!sbi)
1485 return -ENOMEM;
1486 sb->s_fs_info = sbi;
1487 sbi->s_mount_opt = 0;
1488 sbi->s_resuid = EXT4_DEF_RESUID;
1489 sbi->s_resgid = EXT4_DEF_RESGID;
1490
1491 unlock_kernel();
1492
1493 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1494 if (!blocksize) {
1495 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1496 goto out_fail;
1497 }
1498
1499 /*
1500 * The ext4 superblock will not be buffer aligned for other than 1kB
1501 * block sizes. We need to calculate the offset from buffer start.
1502 */
1503 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1504 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1505 offset = do_div(logical_sb_block, blocksize);
1506 } else {
1507 logical_sb_block = sb_block;
1508 }
1509
1510 if (!(bh = sb_bread(sb, logical_sb_block))) {
1511 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1512 goto out_fail;
1513 }
1514 /*
1515 * Note: s_es must be initialized as soon as possible because
1516 * some ext4 macro-instructions depend on its value
1517 */
1518 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1519 sbi->s_es = es;
1520 sb->s_magic = le16_to_cpu(es->s_magic);
1521 if (sb->s_magic != EXT4_SUPER_MAGIC)
1522 goto cantfind_ext4;
1523
1524 /* Set defaults before we parse the mount options */
1525 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1526 if (def_mount_opts & EXT4_DEFM_DEBUG)
1527 set_opt(sbi->s_mount_opt, DEBUG);
1528 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1529 set_opt(sbi->s_mount_opt, GRPID);
1530 if (def_mount_opts & EXT4_DEFM_UID16)
1531 set_opt(sbi->s_mount_opt, NO_UID32);
1532 #ifdef CONFIG_EXT4DEV_FS_XATTR
1533 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1534 set_opt(sbi->s_mount_opt, XATTR_USER);
1535 #endif
1536 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1537 if (def_mount_opts & EXT4_DEFM_ACL)
1538 set_opt(sbi->s_mount_opt, POSIX_ACL);
1539 #endif
1540 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1541 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1542 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1543 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1544 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1545 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1546
1547 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1548 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1549 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_RO)
1550 set_opt(sbi->s_mount_opt, ERRORS_RO);
1551 else
1552 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1553
1554 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1555 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1556
1557 set_opt(sbi->s_mount_opt, RESERVATION);
1558
1559 /*
1560 * turn on extents feature by default in ext4 filesystem
1561 * User -o noextents to turn it off
1562 */
1563 set_opt(sbi->s_mount_opt, EXTENTS);
1564
1565 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1566 NULL, 0))
1567 goto failed_mount;
1568
1569 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1570 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1571
1572 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1573 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1574 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1575 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1576 printk(KERN_WARNING
1577 "EXT4-fs warning: feature flags set on rev 0 fs, "
1578 "running e2fsck is recommended\n");
1579 /*
1580 * Check feature flags regardless of the revision level, since we
1581 * previously didn't change the revision level when setting the flags,
1582 * so there is a chance incompat flags are set on a rev 0 filesystem.
1583 */
1584 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
1585 if (features) {
1586 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
1587 "unsupported optional features (%x).\n",
1588 sb->s_id, le32_to_cpu(features));
1589 goto failed_mount;
1590 }
1591 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
1592 if (!(sb->s_flags & MS_RDONLY) && features) {
1593 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
1594 "unsupported optional features (%x).\n",
1595 sb->s_id, le32_to_cpu(features));
1596 goto failed_mount;
1597 }
1598 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1599
1600 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1601 blocksize > EXT4_MAX_BLOCK_SIZE) {
1602 printk(KERN_ERR
1603 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
1604 blocksize, sb->s_id);
1605 goto failed_mount;
1606 }
1607
1608 hblock = bdev_hardsect_size(sb->s_bdev);
1609 if (sb->s_blocksize != blocksize) {
1610 /*
1611 * Make sure the blocksize for the filesystem is larger
1612 * than the hardware sectorsize for the machine.
1613 */
1614 if (blocksize < hblock) {
1615 printk(KERN_ERR "EXT4-fs: blocksize %d too small for "
1616 "device blocksize %d.\n", blocksize, hblock);
1617 goto failed_mount;
1618 }
1619
1620 brelse (bh);
1621 sb_set_blocksize(sb, blocksize);
1622 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1623 offset = do_div(logical_sb_block, blocksize);
1624 bh = sb_bread(sb, logical_sb_block);
1625 if (!bh) {
1626 printk(KERN_ERR
1627 "EXT4-fs: Can't read superblock on 2nd try.\n");
1628 goto failed_mount;
1629 }
1630 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
1631 sbi->s_es = es;
1632 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
1633 printk (KERN_ERR
1634 "EXT4-fs: Magic mismatch, very weird !\n");
1635 goto failed_mount;
1636 }
1637 }
1638
1639 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
1640
1641 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
1642 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
1643 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
1644 } else {
1645 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1646 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1647 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1648 (!is_power_of_2(sbi->s_inode_size)) ||
1649 (sbi->s_inode_size > blocksize)) {
1650 printk (KERN_ERR
1651 "EXT4-fs: unsupported inode size: %d\n",
1652 sbi->s_inode_size);
1653 goto failed_mount;
1654 }
1655 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
1656 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
1657 }
1658 sbi->s_frag_size = EXT4_MIN_FRAG_SIZE <<
1659 le32_to_cpu(es->s_log_frag_size);
1660 if (blocksize != sbi->s_frag_size) {
1661 printk(KERN_ERR
1662 "EXT4-fs: fragsize %lu != blocksize %u (unsupported)\n",
1663 sbi->s_frag_size, blocksize);
1664 goto failed_mount;
1665 }
1666 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
1667 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
1668 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
1669 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
1670 sbi->s_desc_size & (sbi->s_desc_size - 1)) {
1671 printk(KERN_ERR
1672 "EXT4-fs: unsupported descriptor size %lu\n",
1673 sbi->s_desc_size);
1674 goto failed_mount;
1675 }
1676 } else
1677 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
1678 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1679 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1680 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1681 if (EXT4_INODE_SIZE(sb) == 0)
1682 goto cantfind_ext4;
1683 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
1684 if (sbi->s_inodes_per_block == 0)
1685 goto cantfind_ext4;
1686 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1687 sbi->s_inodes_per_block;
1688 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
1689 sbi->s_sbh = bh;
1690 sbi->s_mount_state = le16_to_cpu(es->s_state);
1691 sbi->s_addr_per_block_bits = log2(EXT4_ADDR_PER_BLOCK(sb));
1692 sbi->s_desc_per_block_bits = log2(EXT4_DESC_PER_BLOCK(sb));
1693 for (i=0; i < 4; i++)
1694 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1695 sbi->s_def_hash_version = es->s_def_hash_version;
1696
1697 if (sbi->s_blocks_per_group > blocksize * 8) {
1698 printk (KERN_ERR
1699 "EXT4-fs: #blocks per group too big: %lu\n",
1700 sbi->s_blocks_per_group);
1701 goto failed_mount;
1702 }
1703 if (sbi->s_frags_per_group > blocksize * 8) {
1704 printk (KERN_ERR
1705 "EXT4-fs: #fragments per group too big: %lu\n",
1706 sbi->s_frags_per_group);
1707 goto failed_mount;
1708 }
1709 if (sbi->s_inodes_per_group > blocksize * 8) {
1710 printk (KERN_ERR
1711 "EXT4-fs: #inodes per group too big: %lu\n",
1712 sbi->s_inodes_per_group);
1713 goto failed_mount;
1714 }
1715
1716 if (ext4_blocks_count(es) >
1717 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1718 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
1719 " too large to mount safely\n", sb->s_id);
1720 if (sizeof(sector_t) < 8)
1721 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
1722 "enabled\n");
1723 goto failed_mount;
1724 }
1725
1726 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
1727 goto cantfind_ext4;
1728 blocks_count = (ext4_blocks_count(es) -
1729 le32_to_cpu(es->s_first_data_block) +
1730 EXT4_BLOCKS_PER_GROUP(sb) - 1);
1731 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
1732 sbi->s_groups_count = blocks_count;
1733 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
1734 EXT4_DESC_PER_BLOCK(sb);
1735 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1736 GFP_KERNEL);
1737 if (sbi->s_group_desc == NULL) {
1738 printk (KERN_ERR "EXT4-fs: not enough memory\n");
1739 goto failed_mount;
1740 }
1741
1742 bgl_lock_init(&sbi->s_blockgroup_lock);
1743
1744 for (i = 0; i < db_count; i++) {
1745 block = descriptor_loc(sb, logical_sb_block, i);
1746 sbi->s_group_desc[i] = sb_bread(sb, block);
1747 if (!sbi->s_group_desc[i]) {
1748 printk (KERN_ERR "EXT4-fs: "
1749 "can't read group descriptor %d\n", i);
1750 db_count = i;
1751 goto failed_mount2;
1752 }
1753 }
1754 if (!ext4_check_descriptors (sb)) {
1755 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
1756 goto failed_mount2;
1757 }
1758 sbi->s_gdb_count = db_count;
1759 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1760 spin_lock_init(&sbi->s_next_gen_lock);
1761
1762 percpu_counter_init(&sbi->s_freeblocks_counter,
1763 ext4_count_free_blocks(sb));
1764 percpu_counter_init(&sbi->s_freeinodes_counter,
1765 ext4_count_free_inodes(sb));
1766 percpu_counter_init(&sbi->s_dirs_counter,
1767 ext4_count_dirs(sb));
1768
1769 /* per fileystem reservation list head & lock */
1770 spin_lock_init(&sbi->s_rsv_window_lock);
1771 sbi->s_rsv_window_root = RB_ROOT;
1772 /* Add a single, static dummy reservation to the start of the
1773 * reservation window list --- it gives us a placeholder for
1774 * append-at-start-of-list which makes the allocation logic
1775 * _much_ simpler. */
1776 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1777 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1778 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1779 sbi->s_rsv_window_head.rsv_goal_size = 0;
1780 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
1781
1782 /*
1783 * set up enough so that it can read an inode
1784 */
1785 sb->s_op = &ext4_sops;
1786 sb->s_export_op = &ext4_export_ops;
1787 sb->s_xattr = ext4_xattr_handlers;
1788 #ifdef CONFIG_QUOTA
1789 sb->s_qcop = &ext4_qctl_operations;
1790 sb->dq_op = &ext4_quota_operations;
1791 #endif
1792 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1793
1794 sb->s_root = NULL;
1795
1796 needs_recovery = (es->s_last_orphan != 0 ||
1797 EXT4_HAS_INCOMPAT_FEATURE(sb,
1798 EXT4_FEATURE_INCOMPAT_RECOVER));
1799
1800 /*
1801 * The first inode we look at is the journal inode. Don't try
1802 * root first: it may be modified in the journal!
1803 */
1804 if (!test_opt(sb, NOLOAD) &&
1805 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
1806 if (ext4_load_journal(sb, es, journal_devnum))
1807 goto failed_mount3;
1808 } else if (journal_inum) {
1809 if (ext4_create_journal(sb, es, journal_inum))
1810 goto failed_mount3;
1811 } else {
1812 if (!silent)
1813 printk (KERN_ERR
1814 "ext4: No journal on filesystem on %s\n",
1815 sb->s_id);
1816 goto failed_mount3;
1817 }
1818
1819 if (ext4_blocks_count(es) > 0xffffffffULL &&
1820 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
1821 JBD2_FEATURE_INCOMPAT_64BIT)) {
1822 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
1823 goto failed_mount4;
1824 }
1825
1826 /* We have now updated the journal if required, so we can
1827 * validate the data journaling mode. */
1828 switch (test_opt(sb, DATA_FLAGS)) {
1829 case 0:
1830 /* No mode set, assume a default based on the journal
1831 * capabilities: ORDERED_DATA if the journal can
1832 * cope, else JOURNAL_DATA
1833 */
1834 if (jbd2_journal_check_available_features
1835 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
1836 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1837 else
1838 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1839 break;
1840
1841 case EXT4_MOUNT_ORDERED_DATA:
1842 case EXT4_MOUNT_WRITEBACK_DATA:
1843 if (!jbd2_journal_check_available_features
1844 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
1845 printk(KERN_ERR "EXT4-fs: Journal does not support "
1846 "requested data journaling mode\n");
1847 goto failed_mount4;
1848 }
1849 default:
1850 break;
1851 }
1852
1853 if (test_opt(sb, NOBH)) {
1854 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
1855 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
1856 "its supported only with writeback mode\n");
1857 clear_opt(sbi->s_mount_opt, NOBH);
1858 }
1859 }
1860 /*
1861 * The jbd2_journal_load will have done any necessary log recovery,
1862 * so we can safely mount the rest of the filesystem now.
1863 */
1864
1865 root = iget(sb, EXT4_ROOT_INO);
1866 sb->s_root = d_alloc_root(root);
1867 if (!sb->s_root) {
1868 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1869 iput(root);
1870 goto failed_mount4;
1871 }
1872 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1873 dput(sb->s_root);
1874 sb->s_root = NULL;
1875 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
1876 goto failed_mount4;
1877 }
1878
1879 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1880
1881 /* determine the minimum size of new large inodes, if present */
1882 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
1883 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
1884 EXT4_GOOD_OLD_INODE_SIZE;
1885 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1886 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
1887 if (sbi->s_want_extra_isize <
1888 le16_to_cpu(es->s_want_extra_isize))
1889 sbi->s_want_extra_isize =
1890 le16_to_cpu(es->s_want_extra_isize);
1891 if (sbi->s_want_extra_isize <
1892 le16_to_cpu(es->s_min_extra_isize))
1893 sbi->s_want_extra_isize =
1894 le16_to_cpu(es->s_min_extra_isize);
1895 }
1896 }
1897 /* Check if enough inode space is available */
1898 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
1899 sbi->s_inode_size) {
1900 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
1901 EXT4_GOOD_OLD_INODE_SIZE;
1902 printk(KERN_INFO "EXT4-fs: required extra inode space not"
1903 "available.\n");
1904 }
1905
1906 /*
1907 * akpm: core read_super() calls in here with the superblock locked.
1908 * That deadlocks, because orphan cleanup needs to lock the superblock
1909 * in numerous places. Here we just pop the lock - it's relatively
1910 * harmless, because we are now ready to accept write_super() requests,
1911 * and aviro says that's the only reason for hanging onto the
1912 * superblock lock.
1913 */
1914 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
1915 ext4_orphan_cleanup(sb, es);
1916 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
1917 if (needs_recovery)
1918 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
1919 ext4_mark_recovery_complete(sb, es);
1920 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
1921 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
1922 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
1923 "writeback");
1924
1925 ext4_ext_init(sb);
1926
1927 lock_kernel();
1928 return 0;
1929
1930 cantfind_ext4:
1931 if (!silent)
1932 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
1933 sb->s_id);
1934 goto failed_mount;
1935
1936 failed_mount4:
1937 jbd2_journal_destroy(sbi->s_journal);
1938 failed_mount3:
1939 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1940 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1941 percpu_counter_destroy(&sbi->s_dirs_counter);
1942 failed_mount2:
1943 for (i = 0; i < db_count; i++)
1944 brelse(sbi->s_group_desc[i]);
1945 kfree(sbi->s_group_desc);
1946 failed_mount:
1947 #ifdef CONFIG_QUOTA
1948 for (i = 0; i < MAXQUOTAS; i++)
1949 kfree(sbi->s_qf_names[i]);
1950 #endif
1951 ext4_blkdev_remove(sbi);
1952 brelse(bh);
1953 out_fail:
1954 sb->s_fs_info = NULL;
1955 kfree(sbi);
1956 lock_kernel();
1957 return -EINVAL;
1958 }
1959
1960 /*
1961 * Setup any per-fs journal parameters now. We'll do this both on
1962 * initial mount, once the journal has been initialised but before we've
1963 * done any recovery; and again on any subsequent remount.
1964 */
1965 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
1966 {
1967 struct ext4_sb_info *sbi = EXT4_SB(sb);
1968
1969 if (sbi->s_commit_interval)
1970 journal->j_commit_interval = sbi->s_commit_interval;
1971 /* We could also set up an ext4-specific default for the commit
1972 * interval here, but for now we'll just fall back to the jbd
1973 * default. */
1974
1975 spin_lock(&journal->j_state_lock);
1976 if (test_opt(sb, BARRIER))
1977 journal->j_flags |= JBD2_BARRIER;
1978 else
1979 journal->j_flags &= ~JBD2_BARRIER;
1980 spin_unlock(&journal->j_state_lock);
1981 }
1982
1983 static journal_t *ext4_get_journal(struct super_block *sb,
1984 unsigned int journal_inum)
1985 {
1986 struct inode *journal_inode;
1987 journal_t *journal;
1988
1989 /* First, test for the existence of a valid inode on disk. Bad
1990 * things happen if we iget() an unused inode, as the subsequent
1991 * iput() will try to delete it. */
1992
1993 journal_inode = iget(sb, journal_inum);
1994 if (!journal_inode) {
1995 printk(KERN_ERR "EXT4-fs: no journal found.\n");
1996 return NULL;
1997 }
1998 if (!journal_inode->i_nlink) {
1999 make_bad_inode(journal_inode);
2000 iput(journal_inode);
2001 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2002 return NULL;
2003 }
2004
2005 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2006 journal_inode, journal_inode->i_size);
2007 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
2008 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2009 iput(journal_inode);
2010 return NULL;
2011 }
2012
2013 journal = jbd2_journal_init_inode(journal_inode);
2014 if (!journal) {
2015 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2016 iput(journal_inode);
2017 return NULL;
2018 }
2019 journal->j_private = sb;
2020 ext4_init_journal_params(sb, journal);
2021 return journal;
2022 }
2023
2024 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2025 dev_t j_dev)
2026 {
2027 struct buffer_head * bh;
2028 journal_t *journal;
2029 ext4_fsblk_t start;
2030 ext4_fsblk_t len;
2031 int hblock, blocksize;
2032 ext4_fsblk_t sb_block;
2033 unsigned long offset;
2034 struct ext4_super_block * es;
2035 struct block_device *bdev;
2036
2037 bdev = ext4_blkdev_get(j_dev);
2038 if (bdev == NULL)
2039 return NULL;
2040
2041 if (bd_claim(bdev, sb)) {
2042 printk(KERN_ERR
2043 "EXT4: failed to claim external journal device.\n");
2044 blkdev_put(bdev);
2045 return NULL;
2046 }
2047
2048 blocksize = sb->s_blocksize;
2049 hblock = bdev_hardsect_size(bdev);
2050 if (blocksize < hblock) {
2051 printk(KERN_ERR
2052 "EXT4-fs: blocksize too small for journal device.\n");
2053 goto out_bdev;
2054 }
2055
2056 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2057 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2058 set_blocksize(bdev, blocksize);
2059 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2060 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2061 "external journal\n");
2062 goto out_bdev;
2063 }
2064
2065 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2066 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2067 !(le32_to_cpu(es->s_feature_incompat) &
2068 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2069 printk(KERN_ERR "EXT4-fs: external journal has "
2070 "bad superblock\n");
2071 brelse(bh);
2072 goto out_bdev;
2073 }
2074
2075 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2076 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2077 brelse(bh);
2078 goto out_bdev;
2079 }
2080
2081 len = ext4_blocks_count(es);
2082 start = sb_block + 1;
2083 brelse(bh); /* we're done with the superblock */
2084
2085 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2086 start, len, blocksize);
2087 if (!journal) {
2088 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2089 goto out_bdev;
2090 }
2091 journal->j_private = sb;
2092 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2093 wait_on_buffer(journal->j_sb_buffer);
2094 if (!buffer_uptodate(journal->j_sb_buffer)) {
2095 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2096 goto out_journal;
2097 }
2098 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2099 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2100 "user (unsupported) - %d\n",
2101 be32_to_cpu(journal->j_superblock->s_nr_users));
2102 goto out_journal;
2103 }
2104 EXT4_SB(sb)->journal_bdev = bdev;
2105 ext4_init_journal_params(sb, journal);
2106 return journal;
2107 out_journal:
2108 jbd2_journal_destroy(journal);
2109 out_bdev:
2110 ext4_blkdev_put(bdev);
2111 return NULL;
2112 }
2113
2114 static int ext4_load_journal(struct super_block *sb,
2115 struct ext4_super_block *es,
2116 unsigned long journal_devnum)
2117 {
2118 journal_t *journal;
2119 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2120 dev_t journal_dev;
2121 int err = 0;
2122 int really_read_only;
2123
2124 if (journal_devnum &&
2125 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2126 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2127 "numbers have changed\n");
2128 journal_dev = new_decode_dev(journal_devnum);
2129 } else
2130 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2131
2132 really_read_only = bdev_read_only(sb->s_bdev);
2133
2134 /*
2135 * Are we loading a blank journal or performing recovery after a
2136 * crash? For recovery, we need to check in advance whether we
2137 * can get read-write access to the device.
2138 */
2139
2140 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2141 if (sb->s_flags & MS_RDONLY) {
2142 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2143 "required on readonly filesystem.\n");
2144 if (really_read_only) {
2145 printk(KERN_ERR "EXT4-fs: write access "
2146 "unavailable, cannot proceed.\n");
2147 return -EROFS;
2148 }
2149 printk (KERN_INFO "EXT4-fs: write access will "
2150 "be enabled during recovery.\n");
2151 }
2152 }
2153
2154 if (journal_inum && journal_dev) {
2155 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2156 "and inode journals!\n");
2157 return -EINVAL;
2158 }
2159
2160 if (journal_inum) {
2161 if (!(journal = ext4_get_journal(sb, journal_inum)))
2162 return -EINVAL;
2163 } else {
2164 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2165 return -EINVAL;
2166 }
2167
2168 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2169 err = jbd2_journal_update_format(journal);
2170 if (err) {
2171 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2172 jbd2_journal_destroy(journal);
2173 return err;
2174 }
2175 }
2176
2177 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2178 err = jbd2_journal_wipe(journal, !really_read_only);
2179 if (!err)
2180 err = jbd2_journal_load(journal);
2181
2182 if (err) {
2183 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2184 jbd2_journal_destroy(journal);
2185 return err;
2186 }
2187
2188 EXT4_SB(sb)->s_journal = journal;
2189 ext4_clear_journal_err(sb, es);
2190
2191 if (journal_devnum &&
2192 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2193 es->s_journal_dev = cpu_to_le32(journal_devnum);
2194 sb->s_dirt = 1;
2195
2196 /* Make sure we flush the recovery flag to disk. */
2197 ext4_commit_super(sb, es, 1);
2198 }
2199
2200 return 0;
2201 }
2202
2203 static int ext4_create_journal(struct super_block * sb,
2204 struct ext4_super_block * es,
2205 unsigned int journal_inum)
2206 {
2207 journal_t *journal;
2208 int err;
2209
2210 if (sb->s_flags & MS_RDONLY) {
2211 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2212 "create journal.\n");
2213 return -EROFS;
2214 }
2215
2216 journal = ext4_get_journal(sb, journal_inum);
2217 if (!journal)
2218 return -EINVAL;
2219
2220 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2221 journal_inum);
2222
2223 err = jbd2_journal_create(journal);
2224 if (err) {
2225 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2226 jbd2_journal_destroy(journal);
2227 return -EIO;
2228 }
2229
2230 EXT4_SB(sb)->s_journal = journal;
2231
2232 ext4_update_dynamic_rev(sb);
2233 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2234 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2235
2236 es->s_journal_inum = cpu_to_le32(journal_inum);
2237 sb->s_dirt = 1;
2238
2239 /* Make sure we flush the recovery flag to disk. */
2240 ext4_commit_super(sb, es, 1);
2241
2242 return 0;
2243 }
2244
2245 static void ext4_commit_super (struct super_block * sb,
2246 struct ext4_super_block * es,
2247 int sync)
2248 {
2249 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2250
2251 if (!sbh)
2252 return;
2253 es->s_wtime = cpu_to_le32(get_seconds());
2254 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2255 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2256 BUFFER_TRACE(sbh, "marking dirty");
2257 mark_buffer_dirty(sbh);
2258 if (sync)
2259 sync_dirty_buffer(sbh);
2260 }
2261
2262
2263 /*
2264 * Have we just finished recovery? If so, and if we are mounting (or
2265 * remounting) the filesystem readonly, then we will end up with a
2266 * consistent fs on disk. Record that fact.
2267 */
2268 static void ext4_mark_recovery_complete(struct super_block * sb,
2269 struct ext4_super_block * es)
2270 {
2271 journal_t *journal = EXT4_SB(sb)->s_journal;
2272
2273 jbd2_journal_lock_updates(journal);
2274 jbd2_journal_flush(journal);
2275 lock_super(sb);
2276 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2277 sb->s_flags & MS_RDONLY) {
2278 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2279 sb->s_dirt = 0;
2280 ext4_commit_super(sb, es, 1);
2281 }
2282 unlock_super(sb);
2283 jbd2_journal_unlock_updates(journal);
2284 }
2285
2286 /*
2287 * If we are mounting (or read-write remounting) a filesystem whose journal
2288 * has recorded an error from a previous lifetime, move that error to the
2289 * main filesystem now.
2290 */
2291 static void ext4_clear_journal_err(struct super_block * sb,
2292 struct ext4_super_block * es)
2293 {
2294 journal_t *journal;
2295 int j_errno;
2296 const char *errstr;
2297
2298 journal = EXT4_SB(sb)->s_journal;
2299
2300 /*
2301 * Now check for any error status which may have been recorded in the
2302 * journal by a prior ext4_error() or ext4_abort()
2303 */
2304
2305 j_errno = jbd2_journal_errno(journal);
2306 if (j_errno) {
2307 char nbuf[16];
2308
2309 errstr = ext4_decode_error(sb, j_errno, nbuf);
2310 ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
2311 "from previous mount: %s", errstr);
2312 ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
2313 "filesystem check.");
2314
2315 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2316 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2317 ext4_commit_super (sb, es, 1);
2318
2319 jbd2_journal_clear_err(journal);
2320 }
2321 }
2322
2323 /*
2324 * Force the running and committing transactions to commit,
2325 * and wait on the commit.
2326 */
2327 int ext4_force_commit(struct super_block *sb)
2328 {
2329 journal_t *journal;
2330 int ret;
2331
2332 if (sb->s_flags & MS_RDONLY)
2333 return 0;
2334
2335 journal = EXT4_SB(sb)->s_journal;
2336 sb->s_dirt = 0;
2337 ret = ext4_journal_force_commit(journal);
2338 return ret;
2339 }
2340
2341 /*
2342 * Ext4 always journals updates to the superblock itself, so we don't
2343 * have to propagate any other updates to the superblock on disk at this
2344 * point. Just start an async writeback to get the buffers on their way
2345 * to the disk.
2346 *
2347 * This implicitly triggers the writebehind on sync().
2348 */
2349
2350 static void ext4_write_super (struct super_block * sb)
2351 {
2352 if (mutex_trylock(&sb->s_lock) != 0)
2353 BUG();
2354 sb->s_dirt = 0;
2355 }
2356
2357 static int ext4_sync_fs(struct super_block *sb, int wait)
2358 {
2359 tid_t target;
2360
2361 sb->s_dirt = 0;
2362 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2363 if (wait)
2364 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2365 }
2366 return 0;
2367 }
2368
2369 /*
2370 * LVM calls this function before a (read-only) snapshot is created. This
2371 * gives us a chance to flush the journal completely and mark the fs clean.
2372 */
2373 static void ext4_write_super_lockfs(struct super_block *sb)
2374 {
2375 sb->s_dirt = 0;
2376
2377 if (!(sb->s_flags & MS_RDONLY)) {
2378 journal_t *journal = EXT4_SB(sb)->s_journal;
2379
2380 /* Now we set up the journal barrier. */
2381 jbd2_journal_lock_updates(journal);
2382 jbd2_journal_flush(journal);
2383
2384 /* Journal blocked and flushed, clear needs_recovery flag. */
2385 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2386 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2387 }
2388 }
2389
2390 /*
2391 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2392 * flag here, even though the filesystem is not technically dirty yet.
2393 */
2394 static void ext4_unlockfs(struct super_block *sb)
2395 {
2396 if (!(sb->s_flags & MS_RDONLY)) {
2397 lock_super(sb);
2398 /* Reser the needs_recovery flag before the fs is unlocked. */
2399 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2400 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2401 unlock_super(sb);
2402 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2403 }
2404 }
2405
2406 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2407 {
2408 struct ext4_super_block * es;
2409 struct ext4_sb_info *sbi = EXT4_SB(sb);
2410 ext4_fsblk_t n_blocks_count = 0;
2411 unsigned long old_sb_flags;
2412 struct ext4_mount_options old_opts;
2413 int err;
2414 #ifdef CONFIG_QUOTA
2415 int i;
2416 #endif
2417
2418 /* Store the original options */
2419 old_sb_flags = sb->s_flags;
2420 old_opts.s_mount_opt = sbi->s_mount_opt;
2421 old_opts.s_resuid = sbi->s_resuid;
2422 old_opts.s_resgid = sbi->s_resgid;
2423 old_opts.s_commit_interval = sbi->s_commit_interval;
2424 #ifdef CONFIG_QUOTA
2425 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2426 for (i = 0; i < MAXQUOTAS; i++)
2427 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2428 #endif
2429
2430 /*
2431 * Allow the "check" option to be passed as a remount option.
2432 */
2433 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2434 err = -EINVAL;
2435 goto restore_opts;
2436 }
2437
2438 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2439 ext4_abort(sb, __FUNCTION__, "Abort forced by user");
2440
2441 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2442 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2443
2444 es = sbi->s_es;
2445
2446 ext4_init_journal_params(sb, sbi->s_journal);
2447
2448 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2449 n_blocks_count > ext4_blocks_count(es)) {
2450 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2451 err = -EROFS;
2452 goto restore_opts;
2453 }
2454
2455 if (*flags & MS_RDONLY) {
2456 /*
2457 * First of all, the unconditional stuff we have to do
2458 * to disable replay of the journal when we next remount
2459 */
2460 sb->s_flags |= MS_RDONLY;
2461
2462 /*
2463 * OK, test if we are remounting a valid rw partition
2464 * readonly, and if so set the rdonly flag and then
2465 * mark the partition as valid again.
2466 */
2467 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2468 (sbi->s_mount_state & EXT4_VALID_FS))
2469 es->s_state = cpu_to_le16(sbi->s_mount_state);
2470
2471 /*
2472 * We have to unlock super so that we can wait for
2473 * transactions.
2474 */
2475 unlock_super(sb);
2476 ext4_mark_recovery_complete(sb, es);
2477 lock_super(sb);
2478 } else {
2479 __le32 ret;
2480 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2481 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2482 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2483 "remount RDWR because of unsupported "
2484 "optional features (%x).\n",
2485 sb->s_id, le32_to_cpu(ret));
2486 err = -EROFS;
2487 goto restore_opts;
2488 }
2489
2490 /*
2491 * If we have an unprocessed orphan list hanging
2492 * around from a previously readonly bdev mount,
2493 * require a full umount/remount for now.
2494 */
2495 if (es->s_last_orphan) {
2496 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2497 "remount RDWR because of unprocessed "
2498 "orphan inode list. Please "
2499 "umount/remount instead.\n",
2500 sb->s_id);
2501 err = -EINVAL;
2502 goto restore_opts;
2503 }
2504
2505 /*
2506 * Mounting a RDONLY partition read-write, so reread
2507 * and store the current valid flag. (It may have
2508 * been changed by e2fsck since we originally mounted
2509 * the partition.)
2510 */
2511 ext4_clear_journal_err(sb, es);
2512 sbi->s_mount_state = le16_to_cpu(es->s_state);
2513 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
2514 goto restore_opts;
2515 if (!ext4_setup_super (sb, es, 0))
2516 sb->s_flags &= ~MS_RDONLY;
2517 }
2518 }
2519 #ifdef CONFIG_QUOTA
2520 /* Release old quota file names */
2521 for (i = 0; i < MAXQUOTAS; i++)
2522 if (old_opts.s_qf_names[i] &&
2523 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2524 kfree(old_opts.s_qf_names[i]);
2525 #endif
2526 return 0;
2527 restore_opts:
2528 sb->s_flags = old_sb_flags;
2529 sbi->s_mount_opt = old_opts.s_mount_opt;
2530 sbi->s_resuid = old_opts.s_resuid;
2531 sbi->s_resgid = old_opts.s_resgid;
2532 sbi->s_commit_interval = old_opts.s_commit_interval;
2533 #ifdef CONFIG_QUOTA
2534 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2535 for (i = 0; i < MAXQUOTAS; i++) {
2536 if (sbi->s_qf_names[i] &&
2537 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2538 kfree(sbi->s_qf_names[i]);
2539 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2540 }
2541 #endif
2542 return err;
2543 }
2544
2545 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
2546 {
2547 struct super_block *sb = dentry->d_sb;
2548 struct ext4_sb_info *sbi = EXT4_SB(sb);
2549 struct ext4_super_block *es = sbi->s_es;
2550 u64 fsid;
2551
2552 if (test_opt(sb, MINIX_DF)) {
2553 sbi->s_overhead_last = 0;
2554 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2555 unsigned long ngroups = sbi->s_groups_count, i;
2556 ext4_fsblk_t overhead = 0;
2557 smp_rmb();
2558
2559 /*
2560 * Compute the overhead (FS structures). This is constant
2561 * for a given filesystem unless the number of block groups
2562 * changes so we cache the previous value until it does.
2563 */
2564
2565 /*
2566 * All of the blocks before first_data_block are
2567 * overhead
2568 */
2569 overhead = le32_to_cpu(es->s_first_data_block);
2570
2571 /*
2572 * Add the overhead attributed to the superblock and
2573 * block group descriptors. If the sparse superblocks
2574 * feature is turned on, then not all groups have this.
2575 */
2576 for (i = 0; i < ngroups; i++) {
2577 overhead += ext4_bg_has_super(sb, i) +
2578 ext4_bg_num_gdb(sb, i);
2579 cond_resched();
2580 }
2581
2582 /*
2583 * Every block group has an inode bitmap, a block
2584 * bitmap, and an inode table.
2585 */
2586 overhead += ngroups * (2 + sbi->s_itb_per_group);
2587 sbi->s_overhead_last = overhead;
2588 smp_wmb();
2589 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2590 }
2591
2592 buf->f_type = EXT4_SUPER_MAGIC;
2593 buf->f_bsize = sb->s_blocksize;
2594 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
2595 buf->f_bfree = percpu_counter_sum(&sbi->s_freeblocks_counter);
2596 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
2597 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2598 if (buf->f_bfree < ext4_r_blocks_count(es))
2599 buf->f_bavail = 0;
2600 buf->f_files = le32_to_cpu(es->s_inodes_count);
2601 buf->f_ffree = percpu_counter_sum(&sbi->s_freeinodes_counter);
2602 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2603 buf->f_namelen = EXT4_NAME_LEN;
2604 fsid = le64_to_cpup((void *)es->s_uuid) ^
2605 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2606 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2607 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2608 return 0;
2609 }
2610
2611 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2612 * is locked for write. Otherwise the are possible deadlocks:
2613 * Process 1 Process 2
2614 * ext4_create() quota_sync()
2615 * jbd2_journal_start() write_dquot()
2616 * DQUOT_INIT() down(dqio_mutex)
2617 * down(dqio_mutex) jbd2_journal_start()
2618 *
2619 */
2620
2621 #ifdef CONFIG_QUOTA
2622
2623 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2624 {
2625 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2626 }
2627
2628 static int ext4_dquot_initialize(struct inode *inode, int type)
2629 {
2630 handle_t *handle;
2631 int ret, err;
2632
2633 /* We may create quota structure so we need to reserve enough blocks */
2634 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
2635 if (IS_ERR(handle))
2636 return PTR_ERR(handle);
2637 ret = dquot_initialize(inode, type);
2638 err = ext4_journal_stop(handle);
2639 if (!ret)
2640 ret = err;
2641 return ret;
2642 }
2643
2644 static int ext4_dquot_drop(struct inode *inode)
2645 {
2646 handle_t *handle;
2647 int ret, err;
2648
2649 /* We may delete quota structure so we need to reserve enough blocks */
2650 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2651 if (IS_ERR(handle))
2652 return PTR_ERR(handle);
2653 ret = dquot_drop(inode);
2654 err = ext4_journal_stop(handle);
2655 if (!ret)
2656 ret = err;
2657 return ret;
2658 }
2659
2660 static int ext4_write_dquot(struct dquot *dquot)
2661 {
2662 int ret, err;
2663 handle_t *handle;
2664 struct inode *inode;
2665
2666 inode = dquot_to_inode(dquot);
2667 handle = ext4_journal_start(inode,
2668 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2669 if (IS_ERR(handle))
2670 return PTR_ERR(handle);
2671 ret = dquot_commit(dquot);
2672 err = ext4_journal_stop(handle);
2673 if (!ret)
2674 ret = err;
2675 return ret;
2676 }
2677
2678 static int ext4_acquire_dquot(struct dquot *dquot)
2679 {
2680 int ret, err;
2681 handle_t *handle;
2682
2683 handle = ext4_journal_start(dquot_to_inode(dquot),
2684 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2685 if (IS_ERR(handle))
2686 return PTR_ERR(handle);
2687 ret = dquot_acquire(dquot);
2688 err = ext4_journal_stop(handle);
2689 if (!ret)
2690 ret = err;
2691 return ret;
2692 }
2693
2694 static int ext4_release_dquot(struct dquot *dquot)
2695 {
2696 int ret, err;
2697 handle_t *handle;
2698
2699 handle = ext4_journal_start(dquot_to_inode(dquot),
2700 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2701 if (IS_ERR(handle)) {
2702 /* Release dquot anyway to avoid endless cycle in dqput() */
2703 dquot_release(dquot);
2704 return PTR_ERR(handle);
2705 }
2706 ret = dquot_release(dquot);
2707 err = ext4_journal_stop(handle);
2708 if (!ret)
2709 ret = err;
2710 return ret;
2711 }
2712
2713 static int ext4_mark_dquot_dirty(struct dquot *dquot)
2714 {
2715 /* Are we journalling quotas? */
2716 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2717 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2718 dquot_mark_dquot_dirty(dquot);
2719 return ext4_write_dquot(dquot);
2720 } else {
2721 return dquot_mark_dquot_dirty(dquot);
2722 }
2723 }
2724
2725 static int ext4_write_info(struct super_block *sb, int type)
2726 {
2727 int ret, err;
2728 handle_t *handle;
2729
2730 /* Data block + inode block */
2731 handle = ext4_journal_start(sb->s_root->d_inode, 2);
2732 if (IS_ERR(handle))
2733 return PTR_ERR(handle);
2734 ret = dquot_commit_info(sb, type);
2735 err = ext4_journal_stop(handle);
2736 if (!ret)
2737 ret = err;
2738 return ret;
2739 }
2740
2741 /*
2742 * Turn on quotas during mount time - we need to find
2743 * the quota file and such...
2744 */
2745 static int ext4_quota_on_mount(struct super_block *sb, int type)
2746 {
2747 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
2748 EXT4_SB(sb)->s_jquota_fmt, type);
2749 }
2750
2751 /*
2752 * Standard function to be called on quota_on
2753 */
2754 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
2755 char *path)
2756 {
2757 int err;
2758 struct nameidata nd;
2759
2760 if (!test_opt(sb, QUOTA))
2761 return -EINVAL;
2762 /* Not journalling quota? */
2763 if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
2764 !EXT4_SB(sb)->s_qf_names[GRPQUOTA])
2765 return vfs_quota_on(sb, type, format_id, path);
2766 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2767 if (err)
2768 return err;
2769 /* Quotafile not on the same filesystem? */
2770 if (nd.mnt->mnt_sb != sb) {
2771 path_release(&nd);
2772 return -EXDEV;
2773 }
2774 /* Quotafile not of fs root? */
2775 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2776 printk(KERN_WARNING
2777 "EXT4-fs: Quota file not on filesystem root. "
2778 "Journalled quota will not work.\n");
2779 path_release(&nd);
2780 return vfs_quota_on(sb, type, format_id, path);
2781 }
2782
2783 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2784 * acquiring the locks... As quota files are never truncated and quota code
2785 * itself serializes the operations (and noone else should touch the files)
2786 * we don't have to be afraid of races */
2787 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
2788 size_t len, loff_t off)
2789 {
2790 struct inode *inode = sb_dqopt(sb)->files[type];
2791 sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
2792 int err = 0;
2793 int offset = off & (sb->s_blocksize - 1);
2794 int tocopy;
2795 size_t toread;
2796 struct buffer_head *bh;
2797 loff_t i_size = i_size_read(inode);
2798
2799 if (off > i_size)
2800 return 0;
2801 if (off+len > i_size)
2802 len = i_size-off;
2803 toread = len;
2804 while (toread > 0) {
2805 tocopy = sb->s_blocksize - offset < toread ?
2806 sb->s_blocksize - offset : toread;
2807 bh = ext4_bread(NULL, inode, blk, 0, &err);
2808 if (err)
2809 return err;
2810 if (!bh) /* A hole? */
2811 memset(data, 0, tocopy);
2812 else
2813 memcpy(data, bh->b_data+offset, tocopy);
2814 brelse(bh);
2815 offset = 0;
2816 toread -= tocopy;
2817 data += tocopy;
2818 blk++;
2819 }
2820 return len;
2821 }
2822
2823 /* Write to quotafile (we know the transaction is already started and has
2824 * enough credits) */
2825 static ssize_t ext4_quota_write(struct super_block *sb, int type,
2826 const char *data, size_t len, loff_t off)
2827 {
2828 struct inode *inode = sb_dqopt(sb)->files[type];
2829 sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
2830 int err = 0;
2831 int offset = off & (sb->s_blocksize - 1);
2832 int tocopy;
2833 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
2834 size_t towrite = len;
2835 struct buffer_head *bh;
2836 handle_t *handle = journal_current_handle();
2837
2838 if (!handle) {
2839 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
2840 " cancelled because transaction is not started.\n",
2841 (unsigned long long)off, (unsigned long long)len);
2842 return -EIO;
2843 }
2844 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2845 while (towrite > 0) {
2846 tocopy = sb->s_blocksize - offset < towrite ?
2847 sb->s_blocksize - offset : towrite;
2848 bh = ext4_bread(handle, inode, blk, 1, &err);
2849 if (!bh)
2850 goto out;
2851 if (journal_quota) {
2852 err = ext4_journal_get_write_access(handle, bh);
2853 if (err) {
2854 brelse(bh);
2855 goto out;
2856 }
2857 }
2858 lock_buffer(bh);
2859 memcpy(bh->b_data+offset, data, tocopy);
2860 flush_dcache_page(bh->b_page);
2861 unlock_buffer(bh);
2862 if (journal_quota)
2863 err = ext4_journal_dirty_metadata(handle, bh);
2864 else {
2865 /* Always do at least ordered writes for quotas */
2866 err = ext4_journal_dirty_data(handle, bh);
2867 mark_buffer_dirty(bh);
2868 }
2869 brelse(bh);
2870 if (err)
2871 goto out;
2872 offset = 0;
2873 towrite -= tocopy;
2874 data += tocopy;
2875 blk++;
2876 }
2877 out:
2878 if (len == towrite)
2879 return err;
2880 if (inode->i_size < off+len-towrite) {
2881 i_size_write(inode, off+len-towrite);
2882 EXT4_I(inode)->i_disksize = inode->i_size;
2883 }
2884 inode->i_version++;
2885 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2886 ext4_mark_inode_dirty(handle, inode);
2887 mutex_unlock(&inode->i_mutex);
2888 return len - towrite;
2889 }
2890
2891 #endif
2892
2893 static int ext4_get_sb(struct file_system_type *fs_type,
2894 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2895 {
2896 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
2897 }
2898
2899 static struct file_system_type ext4dev_fs_type = {
2900 .owner = THIS_MODULE,
2901 .name = "ext4dev",
2902 .get_sb = ext4_get_sb,
2903 .kill_sb = kill_block_super,
2904 .fs_flags = FS_REQUIRES_DEV,
2905 };
2906
2907 static int __init init_ext4_fs(void)
2908 {
2909 int err = init_ext4_xattr();
2910 if (err)
2911 return err;
2912 err = init_inodecache();
2913 if (err)
2914 goto out1;
2915 err = register_filesystem(&ext4dev_fs_type);
2916 if (err)
2917 goto out;
2918 return 0;
2919 out:
2920 destroy_inodecache();
2921 out1:
2922 exit_ext4_xattr();
2923 return err;
2924 }
2925
2926 static void __exit exit_ext4_fs(void)
2927 {
2928 unregister_filesystem(&ext4dev_fs_type);
2929 destroy_inodecache();
2930 exit_ext4_xattr();
2931 }
2932
2933 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2934 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
2935 MODULE_LICENSE("GPL");
2936 module_init(init_ext4_fs)
2937 module_exit(exit_ext4_fs)
AMDTP streaming driver for the Linux FireWire stack (Juju)