]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: Use struct flex_groups to calculate get_orlov_stats()
[mirror_ubuntu-artful-kernel.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 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>
dab291af 23#include <linux/jbd2.h>
ac27a0ec
DK		 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>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 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>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
ede86cc4 39#include <linux/marker.h>
1330593e 40#include <linux/log2.h>
717d50e4 41#include <linux/crc16.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
3dcf5451
CH		 44#include "ext4.h"
45#include "ext4_jbd2.h"
ac27a0ec
DK		 46#include "xattr.h"
47#include "acl.h"
48#include "namei.h"
717d50e4 49#include "group.h"
ac27a0ec 50
9f6200bb 51struct proc_dir_entry *ext4_proc_root;
3197ebdb 52static struct kset *ext4_kset;
9f6200bb 53
617ba13b 54static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 55 unsigned long journal_devnum);
c4be0c1d 56static int ext4_commit_super(struct super_block *sb,
2b2d6d01
TT		 57 struct ext4_super_block *es, int sync);
58static void ext4_mark_recovery_complete(struct super_block *sb,
59 struct ext4_super_block *es);
60static void ext4_clear_journal_err(struct super_block *sb,
61 struct ext4_super_block *es);
617ba13b 62static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 63static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 64 char nbuf[16]);
2b2d6d01
TT		 65static int ext4_remount(struct super_block *sb, int *flags, char *data);
66static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 67static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 68static void ext4_write_super(struct super_block *sb);
c4be0c1d 69static int ext4_freeze(struct super_block *sb);
ac27a0ec 70
bd81d8ee 71
8fadc143
AR		 72ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
73 struct ext4_group_desc *bg)
bd81d8ee 74{
3a14589c 75 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 76 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
3a14589c 77 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 78}
79
8fadc143
AR		 80ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
81 struct ext4_group_desc *bg)
bd81d8ee 82{
5272f837 83 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 84 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 85 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 86}
87
8fadc143
AR		 88ext4_fsblk_t ext4_inode_table(struct super_block *sb,
89 struct ext4_group_desc *bg)
bd81d8ee 90{
5272f837 91 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 92 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 93 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 94}
95
560671a0
AK		 96__u32 ext4_free_blks_count(struct super_block *sb,
97 struct ext4_group_desc *bg)
98{
99 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
100 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
101 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
102}
103
104__u32 ext4_free_inodes_count(struct super_block *sb,
105 struct ext4_group_desc *bg)
106{
107 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
108 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
109 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
110}
111
112__u32 ext4_used_dirs_count(struct super_block *sb,
113 struct ext4_group_desc *bg)
114{
115 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
116 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
117 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
118}
119
120__u32 ext4_itable_unused_count(struct super_block *sb,
121 struct ext4_group_desc *bg)
122{
123 return le16_to_cpu(bg->bg_itable_unused_lo) |
124 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
125 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
126}
127
8fadc143
AR		 128void ext4_block_bitmap_set(struct super_block *sb,
129 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 130{
3a14589c 131 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 132 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
133 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 134}
135
8fadc143
AR		 136void ext4_inode_bitmap_set(struct super_block *sb,
137 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 138{
5272f837 139 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 140 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
141 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 142}
143
8fadc143
AR		 144void ext4_inode_table_set(struct super_block *sb,
145 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 146{
5272f837 147 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 148 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
149 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 150}
151
560671a0
AK		 152void ext4_free_blks_set(struct super_block *sb,
153 struct ext4_group_desc *bg, __u32 count)
154{
155 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
156 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
157 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
158}
159
160void ext4_free_inodes_set(struct super_block *sb,
161 struct ext4_group_desc *bg, __u32 count)
162{
163 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
164 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
165 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
166}
167
168void ext4_used_dirs_set(struct super_block *sb,
169 struct ext4_group_desc *bg, __u32 count)
170{
171 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
172 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
173 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
174}
175
176void ext4_itable_unused_set(struct super_block *sb,
177 struct ext4_group_desc *bg, __u32 count)
178{
179 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
180 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
181 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
182}
183
ac27a0ec 184/*
dab291af 185 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 186 *
187 * The only special thing we need to do here is to make sure that all
188 * journal_end calls result in the superblock being marked dirty, so
189 * that sync() will call the filesystem's write_super callback if
190 * appropriate.
191 */
617ba13b 192handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 193{
194 journal_t *journal;
195
196 if (sb->s_flags & MS_RDONLY)
197 return ERR_PTR(-EROFS);
198
199 /* Special case here: if the journal has aborted behind our
200 * backs (eg. EIO in the commit thread), then we still need to
201 * take the FS itself readonly cleanly. */
617ba13b 202 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 203 if (journal) {
204 if (is_journal_aborted(journal)) {
205 ext4_abort(sb, __func__,
206 "Detected aborted journal");
207 return ERR_PTR(-EROFS);
208 }
209 return jbd2_journal_start(journal, nblocks);
ac27a0ec 210 }
0390131b
FM		 211 /*
212 * We're not journaling, return the appropriate indication.
213 */
214 current->journal_info = EXT4_NOJOURNAL_HANDLE;
215 return current->journal_info;
ac27a0ec
DK		 216}
217
218/*
219 * The only special thing we need to do here is to make sure that all
dab291af 220 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 221 * that sync() will call the filesystem's write_super callback if
222 * appropriate.
223 */
617ba13b 224int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 225{
226 struct super_block *sb;
227 int err;
228 int rc;
229
0390131b
FM		 230 if (!ext4_handle_valid(handle)) {
231 /*
232 * Do this here since we don't call jbd2_journal_stop() in
233 * no-journal mode.
234 */
235 current->journal_info = NULL;
236 return 0;
237 }
ac27a0ec
DK		 238 sb = handle->h_transaction->t_journal->j_private;
239 err = handle->h_err;
dab291af 240 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 241
242 if (!err)
243 err = rc;
244 if (err)
617ba13b 245 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 246 return err;
247}
248
617ba13b 249void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 250 struct buffer_head *bh, handle_t *handle, int err)
251{
252 char nbuf[16];
617ba13b 253 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 254
0390131b
FM		 255 BUG_ON(!ext4_handle_valid(handle));
256
ac27a0ec
DK		 257 if (bh)
258 BUFFER_TRACE(bh, "abort");
259
260 if (!handle->h_err)
261 handle->h_err = err;
262
263 if (is_handle_aborted(handle))
264 return;
265
266 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
267 caller, errstr, err_fn);
268
dab291af 269 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 270}
271
272/* Deal with the reporting of failure conditions on a filesystem such as
273 * inconsistencies detected or read IO failures.
274 *
275 * On ext2, we can store the error state of the filesystem in the
617ba13b 276 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 277 * write ordering constraints on the superblock which prevent us from
278 * writing it out straight away; and given that the journal is about to
279 * be aborted, we can't rely on the current, or future, transactions to
280 * write out the superblock safely.
281 *
dab291af 282 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 283 * the journal instead. On recovery, the journal will compain about
284 * that error until we've noted it down and cleared it.
285 */
286
617ba13b 287static void ext4_handle_error(struct super_block *sb)
ac27a0ec 288{
617ba13b 289 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 290
617ba13b
MC		 291 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
292 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 293
294 if (sb->s_flags & MS_RDONLY)
295 return;
296
2b2d6d01 297 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 298 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 299
617ba13b 300 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ac27a0ec 301 if (journal)
dab291af 302 jbd2_journal_abort(journal, -EIO);
ac27a0ec 303 }
2b2d6d01
TT		 304 if (test_opt(sb, ERRORS_RO)) {
305 printk(KERN_CRIT "Remounting filesystem read-only\n");
ac27a0ec
DK		 306 sb->s_flags |= MS_RDONLY;
307 }
617ba13b 308 ext4_commit_super(sb, es, 1);
ac27a0ec 309 if (test_opt(sb, ERRORS_PANIC))
617ba13b 310 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 311 sb->s_id);
312}
313
2b2d6d01
TT		 314void ext4_error(struct super_block *sb, const char *function,
315 const char *fmt, ...)
ac27a0ec
DK		 316{
317 va_list args;
318
319 va_start(args, fmt);
2b2d6d01 320 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 321 vprintk(fmt, args);
322 printk("\n");
323 va_end(args);
324
617ba13b 325 ext4_handle_error(sb);
ac27a0ec
DK		 326}
327
2b2d6d01 328static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 329 char nbuf[16])
330{
331 char *errstr = NULL;
332
333 switch (errno) {
334 case -EIO:
335 errstr = "IO failure";
336 break;
337 case -ENOMEM:
338 errstr = "Out of memory";
339 break;
340 case -EROFS:
dab291af 341 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
ac27a0ec
DK		 342 errstr = "Journal has aborted";
343 else
344 errstr = "Readonly filesystem";
345 break;
346 default:
347 /* If the caller passed in an extra buffer for unknown
348 * errors, textualise them now. Else we just return
349 * NULL. */
350 if (nbuf) {
351 /* Check for truncated error codes... */
352 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
353 errstr = nbuf;
354 }
355 break;
356 }
357
358 return errstr;
359}
360
617ba13b 361/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 362 * automatically and invokes the appropriate error response. */
363
2b2d6d01 364void __ext4_std_error(struct super_block *sb, const char *function, int errno)
ac27a0ec
DK		 365{
366 char nbuf[16];
367 const char *errstr;
368
369 /* Special case: if the error is EROFS, and we're not already
370 * inside a transaction, then there's really no point in logging
371 * an error. */
372 if (errno == -EROFS && journal_current_handle() == NULL &&
373 (sb->s_flags & MS_RDONLY))
374 return;
375
617ba13b 376 errstr = ext4_decode_error(sb, errno, nbuf);
2b2d6d01
TT		 377 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
378 sb->s_id, function, errstr);
ac27a0ec 379
617ba13b 380 ext4_handle_error(sb);
ac27a0ec
DK		 381}
382
383/*
617ba13b 384 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 385 * abort function may be used to deal with unrecoverable failures such
386 * as journal IO errors or ENOMEM at a critical moment in log management.
387 *
388 * We unconditionally force the filesystem into an ABORT|READONLY state,
389 * unless the error response on the fs has been set to panic in which
390 * case we take the easy way out and panic immediately.
391 */
392
2b2d6d01
TT		 393void ext4_abort(struct super_block *sb, const char *function,
394 const char *fmt, ...)
ac27a0ec
DK		 395{
396 va_list args;
397
2b2d6d01 398 printk(KERN_CRIT "ext4_abort called.\n");
ac27a0ec
DK		 399
400 va_start(args, fmt);
2b2d6d01 401 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 402 vprintk(fmt, args);
403 printk("\n");
404 va_end(args);
405
406 if (test_opt(sb, ERRORS_PANIC))
617ba13b 407 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 408
409 if (sb->s_flags & MS_RDONLY)
410 return;
411
412 printk(KERN_CRIT "Remounting filesystem read-only\n");
617ba13b 413 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 414 sb->s_flags |= MS_RDONLY;
617ba13b 415 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ef2cabf7
HK		 416 if (EXT4_SB(sb)->s_journal)
417 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 418}
419
2b2d6d01
TT		 420void ext4_warning(struct super_block *sb, const char *function,
421 const char *fmt, ...)
ac27a0ec
DK		 422{
423 va_list args;
424
425 va_start(args, fmt);
617ba13b 426 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 427 sb->s_id, function);
428 vprintk(fmt, args);
429 printk("\n");
430 va_end(args);
431}
432
5d1b1b3f
AK		 433void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
434 const char *function, const char *fmt, ...)
435__releases(bitlock)
436__acquires(bitlock)
437{
438 va_list args;
439 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
440
441 va_start(args, fmt);
442 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
443 vprintk(fmt, args);
444 printk("\n");
445 va_end(args);
446
447 if (test_opt(sb, ERRORS_CONT)) {
448 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
449 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
450 ext4_commit_super(sb, es, 0);
451 return;
452 }
453 ext4_unlock_group(sb, grp);
454 ext4_handle_error(sb);
455 /*
456 * We only get here in the ERRORS_RO case; relocking the group
457 * may be dangerous, but nothing bad will happen since the
458 * filesystem will have already been marked read/only and the
459 * journal has been aborted. We return 1 as a hint to callers
460 * who might what to use the return value from
461 * ext4_grp_locked_error() to distinguish beween the
462 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
463 * aggressively from the ext4 function in question, with a
464 * more appropriate error code.
465 */
466 ext4_lock_group(sb, grp);
467 return;
468}
469
470
617ba13b 471void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 472{
617ba13b 473 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 474
617ba13b 475 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 476 return;
477
46e665e9 478 ext4_warning(sb, __func__,
ac27a0ec
DK		 479 "updating to rev %d because of new feature flag, "
480 "running e2fsck is recommended",
617ba13b 481 EXT4_DYNAMIC_REV);
ac27a0ec 482
617ba13b
MC		 483 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
484 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
485 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 486 /* leave es->s_feature_*compat flags alone */
487 /* es->s_uuid will be set by e2fsck if empty */
488
489 /*
490 * The rest of the superblock fields should be zero, and if not it
491 * means they are likely already in use, so leave them alone. We
492 * can leave it up to e2fsck to clean up any inconsistencies there.
493 */
494}
495
496/*
497 * Open the external journal device
498 */
617ba13b 499static struct block_device *ext4_blkdev_get(dev_t dev)
ac27a0ec
DK		 500{
501 struct block_device *bdev;
502 char b[BDEVNAME_SIZE];
503
504 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
505 if (IS_ERR(bdev))
506 goto fail;
507 return bdev;
508
509fail:
abda1418 510 printk(KERN_ERR "EXT4-fs: failed to open journal device %s: %ld\n",
ac27a0ec
DK		 511 __bdevname(dev, b), PTR_ERR(bdev));
512 return NULL;
513}
514
515/*
516 * Release the journal device
517 */
617ba13b 518static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 519{
520 bd_release(bdev);
9a1c3542 521 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 522}
523
617ba13b 524static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 525{
526 struct block_device *bdev;
527 int ret = -ENODEV;
528
529 bdev = sbi->journal_bdev;
530 if (bdev) {
617ba13b 531 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 532 sbi->journal_bdev = NULL;
533 }
534 return ret;
535}
536
537static inline struct inode *orphan_list_entry(struct list_head *l)
538{
617ba13b 539 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 540}
541
617ba13b 542static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 543{
544 struct list_head *l;
545
546 printk(KERN_ERR "sb orphan head is %d\n",
547 le32_to_cpu(sbi->s_es->s_last_orphan));
548
549 printk(KERN_ERR "sb_info orphan list:\n");
550 list_for_each(l, &sbi->s_orphan) {
551 struct inode *inode = orphan_list_entry(l);
552 printk(KERN_ERR " "
553 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
554 inode->i_sb->s_id, inode->i_ino, inode,
555 inode->i_mode, inode->i_nlink,
556 NEXT_ORPHAN(inode));
557 }
558}
559
2b2d6d01 560static void ext4_put_super(struct super_block *sb)
ac27a0ec 561{
617ba13b
MC		 562 struct ext4_sb_info *sbi = EXT4_SB(sb);
563 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 564 int i, err;
ac27a0ec 565
c9de560d 566 ext4_mb_release(sb);
a86c6181 567 ext4_ext_release(sb);
617ba13b 568 ext4_xattr_put_super(sb);
0390131b
FM		 569 if (sbi->s_journal) {
570 err = jbd2_journal_destroy(sbi->s_journal);
571 sbi->s_journal = NULL;
572 if (err < 0)
573 ext4_abort(sb, __func__,
574 "Couldn't clean up the journal");
575 }
ac27a0ec 576 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 577 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 578 es->s_state = cpu_to_le16(sbi->s_mount_state);
617ba13b 579 ext4_commit_super(sb, es, 1);
ac27a0ec 580 }
240799cd 581 if (sbi->s_proc) {
9f6200bb 582 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 583 }
3197ebdb 584 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 585
586 for (i = 0; i < sbi->s_gdb_count; i++)
587 brelse(sbi->s_group_desc[i]);
588 kfree(sbi->s_group_desc);
772cb7c8 589 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 590 percpu_counter_destroy(&sbi->s_freeblocks_counter);
591 percpu_counter_destroy(&sbi->s_freeinodes_counter);
592 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 593 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 594 brelse(sbi->s_sbh);
595#ifdef CONFIG_QUOTA
596 for (i = 0; i < MAXQUOTAS; i++)
597 kfree(sbi->s_qf_names[i]);
598#endif
599
600 /* Debugging code just in case the in-memory inode orphan list
601 * isn't empty. The on-disk one can be non-empty if we've
602 * detected an error and taken the fs readonly, but the
603 * in-memory list had better be clean by this point. */
604 if (!list_empty(&sbi->s_orphan))
605 dump_orphan_list(sb, sbi);
606 J_ASSERT(list_empty(&sbi->s_orphan));
607
f98393a6 608 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 609 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
610 /*
611 * Invalidate the journal device's buffers. We don't want them
612 * floating about in memory - the physical journal device may
613 * hotswapped, and it breaks the `ro-after' testing code.
614 */
615 sync_blockdev(sbi->journal_bdev);
f98393a6 616 invalidate_bdev(sbi->journal_bdev);
617ba13b 617 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 618 }
619 sb->s_fs_info = NULL;
3197ebdb
TT		 620 /*
621 * Now that we are completely done shutting down the
622 * superblock, we need to actually destroy the kobject.
623 */
624 unlock_kernel();
625 unlock_super(sb);
626 kobject_put(&sbi->s_kobj);
627 wait_for_completion(&sbi->s_kobj_unregister);
628 lock_super(sb);
629 lock_kernel();
705895b6 630 kfree(sbi->s_blockgroup_lock);
ac27a0ec
DK		 631 kfree(sbi);
632 return;
633}
634
e18b890b 635static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 636
637/*
638 * Called inside transaction, so use GFP_NOFS
639 */
617ba13b 640static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 641{
617ba13b 642 struct ext4_inode_info *ei;
ac27a0ec 643
e6b4f8da 644 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 645 if (!ei)
646 return NULL;
03010a33 647#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b
MC		 648 ei->i_acl = EXT4_ACL_NOT_CACHED;
649 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec 650#endif
ac27a0ec 651 ei->vfs_inode.i_version = 1;
91246c00 652 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 653 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 654 INIT_LIST_HEAD(&ei->i_prealloc_list);
655 spin_lock_init(&ei->i_prealloc_lock);
0390131b
FM		 656 /*
657 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
658 * therefore it can be null here. Don't check it, just initialize
659 * jinode.
660 */
678aaf48 661 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 662 ei->i_reserved_data_blocks = 0;
663 ei->i_reserved_meta_blocks = 0;
664 ei->i_allocated_meta_blocks = 0;
665 ei->i_delalloc_reserved_flag = 0;
666 spin_lock_init(&(ei->i_block_reservation_lock));
ac27a0ec
DK		 667 return &ei->vfs_inode;
668}
669
617ba13b 670static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 671{
9f7dd93d
VA		 672 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
673 printk("EXT4 Inode %p: orphan list check failed!\n",
674 EXT4_I(inode));
675 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
676 EXT4_I(inode), sizeof(struct ext4_inode_info),
677 true);
678 dump_stack();
679 }
617ba13b 680 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 681}
682
51cc5068 683static void init_once(void *foo)
ac27a0ec 684{
617ba13b 685 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 686
a35afb83 687 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 688#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 689 init_rwsem(&ei->xattr_sem);
ac27a0ec 690#endif
0e855ac8 691 init_rwsem(&ei->i_data_sem);
a35afb83 692 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 693}
694
695static int init_inodecache(void)
696{
617ba13b
MC		 697 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
698 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 699 0, (SLAB_RECLAIM_ACCOUNT|
700 SLAB_MEM_SPREAD),
20c2df83 701 init_once);
617ba13b 702 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 703 return -ENOMEM;
704 return 0;
705}
706
707static void destroy_inodecache(void)
708{
617ba13b 709 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 710}
711
617ba13b 712static void ext4_clear_inode(struct inode *inode)
ac27a0ec 713{
03010a33 714#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b
MC		 715 if (EXT4_I(inode)->i_acl &&
716 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
717 posix_acl_release(EXT4_I(inode)->i_acl);
718 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
719 }
720 if (EXT4_I(inode)->i_default_acl &&
721 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
722 posix_acl_release(EXT4_I(inode)->i_default_acl);
723 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 724 }
725#endif
c2ea3fde 726 ext4_discard_preallocations(inode);
0390131b
FM		 727 if (EXT4_JOURNAL(inode))
728 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
678aaf48 729 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 730}
731
2b2d6d01
TT		 732static inline void ext4_show_quota_options(struct seq_file *seq,
733 struct super_block *sb)
ac27a0ec
DK		 734{
735#if defined(CONFIG_QUOTA)
617ba13b 736 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 737
738 if (sbi->s_jquota_fmt)
739 seq_printf(seq, ",jqfmt=%s",
af5bc92d 740 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
ac27a0ec
DK		 741
742 if (sbi->s_qf_names[USRQUOTA])
743 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
744
745 if (sbi->s_qf_names[GRPQUOTA])
746 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
747
617ba13b 748 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 749 seq_puts(seq, ",usrquota");
750
617ba13b 751 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 752 seq_puts(seq, ",grpquota");
753#endif
754}
755
d9c9bef1
MS		 756/*
757 * Show an option if
758 * - it's set to a non-default value OR
759 * - if the per-sb default is different from the global default
760 */
617ba13b 761static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 762{
aa22df2c
AK		 763 int def_errors;
764 unsigned long def_mount_opts;
ac27a0ec 765 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 766 struct ext4_sb_info *sbi = EXT4_SB(sb);
767 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 768
769 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 770 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 771
772 if (sbi->s_sb_block != 1)
773 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
774 if (test_opt(sb, MINIX_DF))
775 seq_puts(seq, ",minixdf");
aa22df2c 776 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 777 seq_puts(seq, ",grpid");
778 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
779 seq_puts(seq, ",nogrpid");
780 if (sbi->s_resuid != EXT4_DEF_RESUID ||
781 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
782 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
783 }
784 if (sbi->s_resgid != EXT4_DEF_RESGID ||
785 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
786 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
787 }
bb4f397a 788 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 789 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 790 def_errors == EXT4_ERRORS_CONTINUE) {
791 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 792 }
793 }
aa22df2c 794 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 795 seq_puts(seq, ",errors=continue");
aa22df2c 796 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 797 seq_puts(seq, ",errors=panic");
aa22df2c 798 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 799 seq_puts(seq, ",nouid32");
aa22df2c 800 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 801 seq_puts(seq, ",debug");
802 if (test_opt(sb, OLDALLOC))
803 seq_puts(seq, ",oldalloc");
03010a33 804#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 805 if (test_opt(sb, XATTR_USER) &&
806 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 807 seq_puts(seq, ",user_xattr");
808 if (!test_opt(sb, XATTR_USER) &&
809 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
810 seq_puts(seq, ",nouser_xattr");
811 }
812#endif
03010a33 813#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 814 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 815 seq_puts(seq, ",acl");
816 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
817 seq_puts(seq, ",noacl");
818#endif
819 if (!test_opt(sb, RESERVATION))
820 seq_puts(seq, ",noreservation");
30773840 821 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 822 seq_printf(seq, ",commit=%u",
823 (unsigned) (sbi->s_commit_interval / HZ));
824 }
30773840
TT		 825 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
826 seq_printf(seq, ",min_batch_time=%u",
827 (unsigned) sbi->s_min_batch_time);
828 }
829 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
830 seq_printf(seq, ",max_batch_time=%u",
831 (unsigned) sbi->s_min_batch_time);
832 }
833
571640ca
ES		 834 /*
835 * We're changing the default of barrier mount option, so
836 * let's always display its mount state so it's clear what its
837 * status is.
838 */
839 seq_puts(seq, ",barrier=");
840 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 841 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
842 seq_puts(seq, ",journal_async_commit");
d9c9bef1
MS		 843 if (test_opt(sb, NOBH))
844 seq_puts(seq, ",nobh");
25ec56b5
JNC		 845 if (test_opt(sb, I_VERSION))
846 seq_puts(seq, ",i_version");
dd919b98
AK		 847 if (!test_opt(sb, DELALLOC))
848 seq_puts(seq, ",nodelalloc");
849
ac27a0ec 850
cb45bbe4
MS		 851 if (sbi->s_stripe)
852 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 853 /*
854 * journal mode get enabled in different ways
855 * So just print the value even if we didn't specify it
856 */
617ba13b 857 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 858 seq_puts(seq, ",data=journal");
617ba13b 859 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 860 seq_puts(seq, ",data=ordered");
617ba13b 861 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 862 seq_puts(seq, ",data=writeback");
863
240799cd
TT		 864 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
865 seq_printf(seq, ",inode_readahead_blks=%u",
866 sbi->s_inode_readahead_blks);
867
5bf5683a
HK		 868 if (test_opt(sb, DATA_ERR_ABORT))
869 seq_puts(seq, ",data_err=abort");
870
617ba13b 871 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 872 return 0;
873}
874
875
1b961ac0
CH		 876static struct inode *ext4_nfs_get_inode(struct super_block *sb,
877 u64 ino, u32 generation)
ac27a0ec 878{
ac27a0ec 879 struct inode *inode;
ac27a0ec 880
617ba13b 881 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 882 return ERR_PTR(-ESTALE);
617ba13b 883 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 884 return ERR_PTR(-ESTALE);
885
886 /* iget isn't really right if the inode is currently unallocated!!
887 *
617ba13b 888 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 889 * deleted, so we should be safe.
890 *
891 * Currently we don't know the generation for parent directory, so
892 * a generation of 0 means "accept any"
893 */
1d1fe1ee
DH		 894 inode = ext4_iget(sb, ino);
895 if (IS_ERR(inode))
896 return ERR_CAST(inode);
897 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 898 iput(inode);
899 return ERR_PTR(-ESTALE);
900 }
1b961ac0
CH		 901
902 return inode;
903}
904
905static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
906 int fh_len, int fh_type)
907{
908 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
909 ext4_nfs_get_inode);
910}
911
912static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
913 int fh_len, int fh_type)
914{
915 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
916 ext4_nfs_get_inode);
ac27a0ec
DK		 917}
918
c39a7f84
TO		 919/*
920 * Try to release metadata pages (indirect blocks, directories) which are
921 * mapped via the block device. Since these pages could have journal heads
922 * which would prevent try_to_free_buffers() from freeing them, we must use
923 * jbd2 layer's try_to_free_buffers() function to release them.
924 */
925static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
926{
927 journal_t *journal = EXT4_SB(sb)->s_journal;
928
929 WARN_ON(PageChecked(page));
930 if (!page_has_buffers(page))
931 return 0;
932 if (journal)
933 return jbd2_journal_try_to_free_buffers(journal, page,
934 wait & ~__GFP_WAIT);
935 return try_to_free_buffers(page);
936}
937
ac27a0ec 938#ifdef CONFIG_QUOTA
af5bc92d 939#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 940#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 941
617ba13b
MC		 942static int ext4_write_dquot(struct dquot *dquot);
943static int ext4_acquire_dquot(struct dquot *dquot);
944static int ext4_release_dquot(struct dquot *dquot);
945static int ext4_mark_dquot_dirty(struct dquot *dquot);
946static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 947static int ext4_quota_on(struct super_block *sb, int type, int format_id,
948 char *path, int remount);
617ba13b
MC		 949static int ext4_quota_on_mount(struct super_block *sb, int type);
950static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 951 size_t len, loff_t off);
617ba13b 952static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 953 const char *data, size_t len, loff_t off);
954
617ba13b 955static struct dquot_operations ext4_quota_operations = {
edf72453
JK		 956 .initialize = dquot_initialize,
957 .drop = dquot_drop,
ac27a0ec 958 .alloc_space = dquot_alloc_space,
60e58e0f
MC		 959 .reserve_space = dquot_reserve_space,
960 .claim_space = dquot_claim_space,
961 .release_rsv = dquot_release_reserved_space,
962 .get_reserved_space = ext4_get_reserved_space,
ac27a0ec
DK		 963 .alloc_inode = dquot_alloc_inode,
964 .free_space = dquot_free_space,
965 .free_inode = dquot_free_inode,
966 .transfer = dquot_transfer,
617ba13b
MC		 967 .write_dquot = ext4_write_dquot,
968 .acquire_dquot = ext4_acquire_dquot,
969 .release_dquot = ext4_release_dquot,
970 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 971 .write_info = ext4_write_info,
972 .alloc_dquot = dquot_alloc,
973 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 974};
975
617ba13b
MC		 976static struct quotactl_ops ext4_qctl_operations = {
977 .quota_on = ext4_quota_on,
ac27a0ec
DK		 978 .quota_off = vfs_quota_off,
979 .quota_sync = vfs_quota_sync,
980 .get_info = vfs_get_dqinfo,
981 .set_info = vfs_set_dqinfo,
982 .get_dqblk = vfs_get_dqblk,
983 .set_dqblk = vfs_set_dqblk
984};
985#endif
986
ee9b6d61 987static const struct super_operations ext4_sops = {
617ba13b
MC		 988 .alloc_inode = ext4_alloc_inode,
989 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 990 .write_inode = ext4_write_inode,
991 .dirty_inode = ext4_dirty_inode,
992 .delete_inode = ext4_delete_inode,
993 .put_super = ext4_put_super,
994 .write_super = ext4_write_super,
995 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 996 .freeze_fs = ext4_freeze,
997 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 998 .statfs = ext4_statfs,
999 .remount_fs = ext4_remount,
1000 .clear_inode = ext4_clear_inode,
1001 .show_options = ext4_show_options,
ac27a0ec 1002#ifdef CONFIG_QUOTA
617ba13b
MC		 1003 .quota_read = ext4_quota_read,
1004 .quota_write = ext4_quota_write,
ac27a0ec 1005#endif
c39a7f84 1006 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1007};
1008
39655164 1009static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1010 .fh_to_dentry = ext4_fh_to_dentry,
1011 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1012 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1013};
1014
1015enum {
1016 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1017 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1018 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec
DK		 1019 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1020 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
30773840 1021 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1022 Opt_journal_update, Opt_journal_dev,
818d276c 1023 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1024 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
5bf5683a 1025 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec
DK		 1026 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1027 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1028 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
83982b6f 1029 Opt_grpquota, Opt_i_version,
01436ef2 1030 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
b3881f74 1031 Opt_inode_readahead_blks, Opt_journal_ioprio
ac27a0ec
DK		 1032};
1033
a447c093 1034static const match_table_t tokens = {
ac27a0ec
DK		 1035 {Opt_bsd_df, "bsddf"},
1036 {Opt_minix_df, "minixdf"},
1037 {Opt_grpid, "grpid"},
1038 {Opt_grpid, "bsdgroups"},
1039 {Opt_nogrpid, "nogrpid"},
1040 {Opt_nogrpid, "sysvgroups"},
1041 {Opt_resgid, "resgid=%u"},
1042 {Opt_resuid, "resuid=%u"},
1043 {Opt_sb, "sb=%u"},
1044 {Opt_err_cont, "errors=continue"},
1045 {Opt_err_panic, "errors=panic"},
1046 {Opt_err_ro, "errors=remount-ro"},
1047 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1048 {Opt_debug, "debug"},
1049 {Opt_oldalloc, "oldalloc"},
1050 {Opt_orlov, "orlov"},
1051 {Opt_user_xattr, "user_xattr"},
1052 {Opt_nouser_xattr, "nouser_xattr"},
1053 {Opt_acl, "acl"},
1054 {Opt_noacl, "noacl"},
1055 {Opt_reservation, "reservation"},
1056 {Opt_noreservation, "noreservation"},
1057 {Opt_noload, "noload"},
1058 {Opt_nobh, "nobh"},
1059 {Opt_bh, "bh"},
1060 {Opt_commit, "commit=%u"},
30773840
TT		 1061 {Opt_min_batch_time, "min_batch_time=%u"},
1062 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1063 {Opt_journal_update, "journal=update"},
ac27a0ec 1064 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1065 {Opt_journal_checksum, "journal_checksum"},
1066 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1067 {Opt_abort, "abort"},
1068 {Opt_data_journal, "data=journal"},
1069 {Opt_data_ordered, "data=ordered"},
1070 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1071 {Opt_data_err_abort, "data_err=abort"},
1072 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1073 {Opt_offusrjquota, "usrjquota="},
1074 {Opt_usrjquota, "usrjquota=%s"},
1075 {Opt_offgrpjquota, "grpjquota="},
1076 {Opt_grpjquota, "grpjquota=%s"},
1077 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1078 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1079 {Opt_grpquota, "grpquota"},
1080 {Opt_noquota, "noquota"},
1081 {Opt_quota, "quota"},
1082 {Opt_usrquota, "usrquota"},
1083 {Opt_barrier, "barrier=%u"},
25ec56b5 1084 {Opt_i_version, "i_version"},
c9de560d 1085 {Opt_stripe, "stripe=%u"},
ac27a0ec 1086 {Opt_resize, "resize"},
64769240 1087 {Opt_delalloc, "delalloc"},
dd919b98 1088 {Opt_nodelalloc, "nodelalloc"},
240799cd 1089 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1090 {Opt_journal_ioprio, "journal_ioprio=%u"},
f3f12faa 1091 {Opt_err, NULL},
ac27a0ec
DK		 1092};
1093
617ba13b 1094static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1095{
617ba13b 1096 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1097 char *options = (char *) *data;
1098
1099 if (!options || strncmp(options, "sb=", 3) != 0)
1100 return 1; /* Default location */
1101 options += 3;
617ba13b 1102 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1103 sb_block = simple_strtoul(options, &options, 0);
1104 if (*options && *options != ',') {
4776004f 1105 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1106 (char *) *data);
1107 return 1;
1108 }
1109 if (*options == ',')
1110 options++;
1111 *data = (void *) options;
1112 return sb_block;
1113}
1114
b3881f74
TT		 1115#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1116
2b2d6d01 1117static int parse_options(char *options, struct super_block *sb,
c3191067 1118 unsigned long *journal_devnum,
b3881f74 1119 unsigned int *journal_ioprio,
2b2d6d01 1120 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1121{
617ba13b 1122 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1123 char *p;
ac27a0ec
DK		 1124 substring_t args[MAX_OPT_ARGS];
1125 int data_opt = 0;
1126 int option;
1127#ifdef CONFIG_QUOTA
dfc5d03f 1128 int qtype, qfmt;
ac27a0ec
DK		 1129 char *qname;
1130#endif
1131
1132 if (!options)
1133 return 1;
1134
2b2d6d01 1135 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1136 int token;
1137 if (!*p)
1138 continue;
1139
1140 token = match_token(p, tokens, args);
1141 switch (token) {
1142 case Opt_bsd_df:
2b2d6d01 1143 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1144 break;
1145 case Opt_minix_df:
2b2d6d01 1146 set_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1147 break;
1148 case Opt_grpid:
2b2d6d01 1149 set_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1150 break;
1151 case Opt_nogrpid:
2b2d6d01 1152 clear_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1153 break;
1154 case Opt_resuid:
1155 if (match_int(&args[0], &option))
1156 return 0;
1157 sbi->s_resuid = option;
1158 break;
1159 case Opt_resgid:
1160 if (match_int(&args[0], &option))
1161 return 0;
1162 sbi->s_resgid = option;
1163 break;
1164 case Opt_sb:
1165 /* handled by get_sb_block() instead of here */
1166 /* *sb_block = match_int(&args[0]); */
1167 break;
1168 case Opt_err_panic:
2b2d6d01
TT		 1169 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1170 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1171 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1172 break;
1173 case Opt_err_ro:
2b2d6d01
TT		 1174 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1175 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1176 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1177 break;
1178 case Opt_err_cont:
2b2d6d01
TT		 1179 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1180 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1181 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1182 break;
1183 case Opt_nouid32:
2b2d6d01 1184 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1185 break;
ac27a0ec 1186 case Opt_debug:
2b2d6d01 1187 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1188 break;
1189 case Opt_oldalloc:
2b2d6d01 1190 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1191 break;
1192 case Opt_orlov:
2b2d6d01 1193 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1194 break;
03010a33 1195#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1196 case Opt_user_xattr:
2b2d6d01 1197 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1198 break;
1199 case Opt_nouser_xattr:
2b2d6d01 1200 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1201 break;
1202#else
1203 case Opt_user_xattr:
1204 case Opt_nouser_xattr:
4776004f
TT		 1205 printk(KERN_ERR "EXT4 (no)user_xattr options "
1206 "not supported\n");
ac27a0ec
DK		 1207 break;
1208#endif
03010a33 1209#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1210 case Opt_acl:
1211 set_opt(sbi->s_mount_opt, POSIX_ACL);
1212 break;
1213 case Opt_noacl:
1214 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1215 break;
1216#else
1217 case Opt_acl:
1218 case Opt_noacl:
4776004f
TT		 1219 printk(KERN_ERR "EXT4 (no)acl options "
1220 "not supported\n");
ac27a0ec
DK		 1221 break;
1222#endif
1223 case Opt_reservation:
1224 set_opt(sbi->s_mount_opt, RESERVATION);
1225 break;
1226 case Opt_noreservation:
1227 clear_opt(sbi->s_mount_opt, RESERVATION);
1228 break;
1229 case Opt_journal_update:
1230 /* @@@ FIXME */
1231 /* Eventually we will want to be able to create
1232 a journal file here. For now, only allow the
1233 user to specify an existing inode to be the
1234 journal file. */
1235 if (is_remount) {
617ba13b 1236 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1237 "journal on remount\n");
1238 return 0;
1239 }
2b2d6d01 1240 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1241 break;
ac27a0ec
DK		 1242 case Opt_journal_dev:
1243 if (is_remount) {
617ba13b 1244 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1245 "journal on remount\n");
1246 return 0;
1247 }
1248 if (match_int(&args[0], &option))
1249 return 0;
1250 *journal_devnum = option;
1251 break;
818d276c
GS		 1252 case Opt_journal_checksum:
1253 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1254 break;
1255 case Opt_journal_async_commit:
1256 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1257 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1258 break;
ac27a0ec 1259 case Opt_noload:
2b2d6d01 1260 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1261 break;
1262 case Opt_commit:
1263 if (match_int(&args[0], &option))
1264 return 0;
1265 if (option < 0)
1266 return 0;
1267 if (option == 0)
cd02ff0b 1268 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1269 sbi->s_commit_interval = HZ * option;
1270 break;
30773840
TT		 1271 case Opt_max_batch_time:
1272 if (match_int(&args[0], &option))
1273 return 0;
1274 if (option < 0)
1275 return 0;
1276 if (option == 0)
1277 option = EXT4_DEF_MAX_BATCH_TIME;
1278 sbi->s_max_batch_time = option;
1279 break;
1280 case Opt_min_batch_time:
1281 if (match_int(&args[0], &option))
1282 return 0;
1283 if (option < 0)
1284 return 0;
1285 sbi->s_min_batch_time = option;
1286 break;
ac27a0ec 1287 case Opt_data_journal:
617ba13b 1288 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1289 goto datacheck;
1290 case Opt_data_ordered:
617ba13b 1291 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1292 goto datacheck;
1293 case Opt_data_writeback:
617ba13b 1294 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1295 datacheck:
1296 if (is_remount) {
617ba13b 1297 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1298 != data_opt) {
1299 printk(KERN_ERR
617ba13b 1300 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1301 "mode on remount\n");
1302 return 0;
1303 }
1304 } else {
617ba13b 1305 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1306 sbi->s_mount_opt |= data_opt;
1307 }
1308 break;
5bf5683a
HK		 1309 case Opt_data_err_abort:
1310 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1311 break;
1312 case Opt_data_err_ignore:
1313 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1314 break;
ac27a0ec
DK		 1315#ifdef CONFIG_QUOTA
1316 case Opt_usrjquota:
1317 qtype = USRQUOTA;
1318 goto set_qf_name;
1319 case Opt_grpjquota:
1320 qtype = GRPQUOTA;
1321set_qf_name:
17bd13b3 1322 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1323 !sbi->s_qf_names[qtype]) {
ac27a0ec 1324 printk(KERN_ERR
4776004f
TT		 1325 "EXT4-fs: Cannot change journaled "
1326 "quota options when quota turned on.\n");
ac27a0ec
DK		 1327 return 0;
1328 }
1329 qname = match_strdup(&args[0]);
1330 if (!qname) {
1331 printk(KERN_ERR
617ba13b 1332 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1333 "storing quotafile name.\n");
1334 return 0;
1335 }
1336 if (sbi->s_qf_names[qtype] &&
1337 strcmp(sbi->s_qf_names[qtype], qname)) {
1338 printk(KERN_ERR
617ba13b 1339 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1340 "specified.\n", QTYPE2NAME(qtype));
1341 kfree(qname);
1342 return 0;
1343 }
1344 sbi->s_qf_names[qtype] = qname;
1345 if (strchr(sbi->s_qf_names[qtype], '/')) {
1346 printk(KERN_ERR
617ba13b 1347 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1348 "filesystem root.\n");
1349 kfree(sbi->s_qf_names[qtype]);
1350 sbi->s_qf_names[qtype] = NULL;
1351 return 0;
1352 }
1353 set_opt(sbi->s_mount_opt, QUOTA);
1354 break;
1355 case Opt_offusrjquota:
1356 qtype = USRQUOTA;
1357 goto clear_qf_name;
1358 case Opt_offgrpjquota:
1359 qtype = GRPQUOTA;
1360clear_qf_name:
17bd13b3 1361 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1362 sbi->s_qf_names[qtype]) {
617ba13b 1363 printk(KERN_ERR "EXT4-fs: Cannot change "
2c8be6b2 1364 "journaled quota options when "
ac27a0ec
DK		 1365 "quota turned on.\n");
1366 return 0;
1367 }
1368 /*
1369 * The space will be released later when all options
1370 * are confirmed to be correct
1371 */
1372 sbi->s_qf_names[qtype] = NULL;
1373 break;
1374 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1375 qfmt = QFMT_VFS_OLD;
1376 goto set_qf_format;
ac27a0ec 1377 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1378 qfmt = QFMT_VFS_V0;
1379set_qf_format:
17bd13b3 1380 if (sb_any_quota_loaded(sb) &&
dfc5d03f
JK		 1381 sbi->s_jquota_fmt != qfmt) {
1382 printk(KERN_ERR "EXT4-fs: Cannot change "
1383 "journaled quota options when "
1384 "quota turned on.\n");
1385 return 0;
1386 }
1387 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1388 break;
1389 case Opt_quota:
1390 case Opt_usrquota:
1391 set_opt(sbi->s_mount_opt, QUOTA);
1392 set_opt(sbi->s_mount_opt, USRQUOTA);
1393 break;
1394 case Opt_grpquota:
1395 set_opt(sbi->s_mount_opt, QUOTA);
1396 set_opt(sbi->s_mount_opt, GRPQUOTA);
1397 break;
1398 case Opt_noquota:
17bd13b3 1399 if (sb_any_quota_loaded(sb)) {
617ba13b 1400 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1401 "options when quota turned on.\n");
1402 return 0;
1403 }
1404 clear_opt(sbi->s_mount_opt, QUOTA);
1405 clear_opt(sbi->s_mount_opt, USRQUOTA);
1406 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1407 break;
1408#else
1409 case Opt_quota:
1410 case Opt_usrquota:
1411 case Opt_grpquota:
cd59e7b9
JK		 1412 printk(KERN_ERR
1413 "EXT4-fs: quota options not supported.\n");
1414 break;
ac27a0ec
DK		 1415 case Opt_usrjquota:
1416 case Opt_grpjquota:
1417 case Opt_offusrjquota:
1418 case Opt_offgrpjquota:
1419 case Opt_jqfmt_vfsold:
1420 case Opt_jqfmt_vfsv0:
1421 printk(KERN_ERR
cd59e7b9 1422 "EXT4-fs: journaled quota options not "
ac27a0ec
DK		 1423 "supported.\n");
1424 break;
1425 case Opt_noquota:
1426 break;
1427#endif
1428 case Opt_abort:
1429 set_opt(sbi->s_mount_opt, ABORT);
1430 break;
1431 case Opt_barrier:
1432 if (match_int(&args[0], &option))
1433 return 0;
1434 if (option)
1435 set_opt(sbi->s_mount_opt, BARRIER);
1436 else
1437 clear_opt(sbi->s_mount_opt, BARRIER);
1438 break;
1439 case Opt_ignore:
1440 break;
1441 case Opt_resize:
1442 if (!is_remount) {
617ba13b 1443 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1444 "for remount\n");
1445 return 0;
1446 }
1447 if (match_int(&args[0], &option) != 0)
1448 return 0;
1449 *n_blocks_count = option;
1450 break;
1451 case Opt_nobh:
1452 set_opt(sbi->s_mount_opt, NOBH);
1453 break;
1454 case Opt_bh:
1455 clear_opt(sbi->s_mount_opt, NOBH);
1456 break;
25ec56b5
JNC		 1457 case Opt_i_version:
1458 set_opt(sbi->s_mount_opt, I_VERSION);
1459 sb->s_flags |= MS_I_VERSION;
1460 break;
dd919b98
AK		 1461 case Opt_nodelalloc:
1462 clear_opt(sbi->s_mount_opt, DELALLOC);
1463 break;
c9de560d
AT		 1464 case Opt_stripe:
1465 if (match_int(&args[0], &option))
1466 return 0;
1467 if (option < 0)
1468 return 0;
1469 sbi->s_stripe = option;
1470 break;
64769240
AT		 1471 case Opt_delalloc:
1472 set_opt(sbi->s_mount_opt, DELALLOC);
1473 break;
240799cd
TT		 1474 case Opt_inode_readahead_blks:
1475 if (match_int(&args[0], &option))
1476 return 0;
1477 if (option < 0 || option > (1 << 30))
1478 return 0;
3197ebdb
TT		 1479 if (option & (option - 1)) {
1480 printk(KERN_ERR "EXT4-fs: inode_readahead_blks"
1481 " must be a power of 2\n");
1482 return 0;
1483 }
240799cd
TT		 1484 sbi->s_inode_readahead_blks = option;
1485 break;
b3881f74
TT		 1486 case Opt_journal_ioprio:
1487 if (match_int(&args[0], &option))
1488 return 0;
1489 if (option < 0 || option > 7)
1490 break;
1491 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1492 option);
1493 break;
ac27a0ec 1494 default:
2b2d6d01
TT		 1495 printk(KERN_ERR
1496 "EXT4-fs: Unrecognized mount option \"%s\" "
1497 "or missing value\n", p);
ac27a0ec
DK		 1498 return 0;
1499 }
1500 }
1501#ifdef CONFIG_QUOTA
1502 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1503 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1504 sbi->s_qf_names[USRQUOTA])
1505 clear_opt(sbi->s_mount_opt, USRQUOTA);
1506
617ba13b 1507 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1508 sbi->s_qf_names[GRPQUOTA])
1509 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1510
1511 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1512 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1513 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1514 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1515 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1516 "format mixing.\n");
1517 return 0;
1518 }
1519
1520 if (!sbi->s_jquota_fmt) {
2c8be6b2 1521 printk(KERN_ERR "EXT4-fs: journaled quota format "
ac27a0ec
DK		 1522 "not specified.\n");
1523 return 0;
1524 }
1525 } else {
1526 if (sbi->s_jquota_fmt) {
2c8be6b2
JK		 1527 printk(KERN_ERR "EXT4-fs: journaled quota format "
1528 "specified with no journaling "
ac27a0ec
DK		 1529 "enabled.\n");
1530 return 0;
1531 }
1532 }
1533#endif
1534 return 1;
1535}
1536
617ba13b 1537static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1538 int read_only)
1539{
617ba13b 1540 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1541 int res = 0;
1542
617ba13b 1543 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2b2d6d01
TT		 1544 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1545 "forcing read-only mode\n");
ac27a0ec
DK		 1546 res = MS_RDONLY;
1547 }
1548 if (read_only)
1549 return res;
617ba13b 1550 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2b2d6d01
TT		 1551 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1552 "running e2fsck is recommended\n");
617ba13b 1553 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
2b2d6d01
TT		 1554 printk(KERN_WARNING
1555 "EXT4-fs warning: mounting fs with errors, "
1556 "running e2fsck is recommended\n");
ac27a0ec
DK		 1557 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1558 le16_to_cpu(es->s_mnt_count) >=
1559 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2b2d6d01
TT		 1560 printk(KERN_WARNING
1561 "EXT4-fs warning: maximal mount count reached, "
1562 "running e2fsck is recommended\n");
ac27a0ec
DK		 1563 else if (le32_to_cpu(es->s_checkinterval) &&
1564 (le32_to_cpu(es->s_lastcheck) +
1565 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2b2d6d01
TT		 1566 printk(KERN_WARNING
1567 "EXT4-fs warning: checktime reached, "
1568 "running e2fsck is recommended\n");
0390131b
FM		 1569 if (!sbi->s_journal)
1570 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1571 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1572 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1573 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1574 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1575 ext4_update_dynamic_rev(sb);
0390131b
FM		 1576 if (sbi->s_journal)
1577 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1578
617ba13b 1579 ext4_commit_super(sb, es, 1);
ac27a0ec 1580 if (test_opt(sb, DEBUG))
a9df9a49 1581 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
ac27a0ec
DK		 1582 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1583 sb->s_blocksize,
1584 sbi->s_groups_count,
617ba13b
MC		 1585 EXT4_BLOCKS_PER_GROUP(sb),
1586 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1587 sbi->s_mount_opt);
1588
0390131b
FM		 1589 if (EXT4_SB(sb)->s_journal) {
1590 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1591 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1592 "external", EXT4_SB(sb)->s_journal->j_devname);
1593 } else {
1594 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1595 }
ac27a0ec
DK		 1596 return res;
1597}
1598
772cb7c8
JS		 1599static int ext4_fill_flex_info(struct super_block *sb)
1600{
1601 struct ext4_sb_info *sbi = EXT4_SB(sb);
1602 struct ext4_group_desc *gdp = NULL;
1603 struct buffer_head *bh;
1604 ext4_group_t flex_group_count;
1605 ext4_group_t flex_group;
1606 int groups_per_flex = 0;
772cb7c8
JS		 1607 int i;
1608
1609 if (!sbi->s_es->s_log_groups_per_flex) {
1610 sbi->s_log_groups_per_flex = 0;
1611 return 1;
1612 }
1613
1614 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1615 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1616
c62a11fd
FB		 1617 /* We allocate both existing and potentially added groups */
1618 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1619 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1620 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
ec05e868 1621 sbi->s_flex_groups = kzalloc(flex_group_count *
772cb7c8
JS		 1622 sizeof(struct flex_groups), GFP_KERNEL);
1623 if (sbi->s_flex_groups == NULL) {
ec05e868 1624 printk(KERN_ERR "EXT4-fs: not enough memory for "
a9df9a49 1625 "%u flex groups\n", flex_group_count);
772cb7c8
JS		 1626 goto failed;
1627 }
772cb7c8 1628
772cb7c8
JS		 1629 for (i = 0; i < sbi->s_groups_count; i++) {
1630 gdp = ext4_get_group_desc(sb, i, &bh);
1631
1632 flex_group = ext4_flex_group(sbi, i);
9f24e420
TT		 1633 atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1634 ext4_free_inodes_count(sb, gdp));
1635 atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1636 ext4_free_blks_count(sb, gdp));
7d39db14
TT		 1637 atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1638 ext4_used_dirs_count(sb, gdp));
772cb7c8
JS		 1639 }
1640
1641 return 1;
1642failed:
1643 return 0;
1644}
1645
717d50e4
AD		 1646__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1647 struct ext4_group_desc *gdp)
1648{
1649 __u16 crc = 0;
1650
1651 if (sbi->s_es->s_feature_ro_compat &
1652 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1653 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1654 __le32 le_group = cpu_to_le32(block_group);
1655
1656 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1657 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1658 crc = crc16(crc, (__u8 *)gdp, offset);
1659 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1660 /* for checksum of struct ext4_group_desc do the rest...*/
1661 if ((sbi->s_es->s_feature_incompat &
1662 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1663 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1664 crc = crc16(crc, (__u8 *)gdp + offset,
1665 le16_to_cpu(sbi->s_es->s_desc_size) -
1666 offset);
1667 }
1668
1669 return cpu_to_le16(crc);
1670}
1671
1672int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1673 struct ext4_group_desc *gdp)
1674{
1675 if ((sbi->s_es->s_feature_ro_compat &
1676 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1677 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1678 return 0;
1679
1680 return 1;
1681}
1682
ac27a0ec 1683/* Called at mount-time, super-block is locked */
197cd65a 1684static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1685{
617ba13b
MC		 1686 struct ext4_sb_info *sbi = EXT4_SB(sb);
1687 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1688 ext4_fsblk_t last_block;
bd81d8ee
LV		 1689 ext4_fsblk_t block_bitmap;
1690 ext4_fsblk_t inode_bitmap;
1691 ext4_fsblk_t inode_table;
ce421581 1692 int flexbg_flag = 0;
fd2d4291 1693 ext4_group_t i;
ac27a0ec 1694
ce421581
JS		 1695 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1696 flexbg_flag = 1;
1697
af5bc92d 1698 ext4_debug("Checking group descriptors");
ac27a0ec 1699
197cd65a
AM		 1700 for (i = 0; i < sbi->s_groups_count; i++) {
1701 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1702
ce421581 1703 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1704 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1705 else
1706 last_block = first_block +
617ba13b 1707 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1708
8fadc143 1709 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1710 if (block_bitmap < first_block || block_bitmap > last_block) {
c19204b0 1711 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1712 "Block bitmap for group %u not in group "
5128273a 1713 "(block %llu)!\n", i, block_bitmap);
ac27a0ec
DK		 1714 return 0;
1715 }
8fadc143 1716 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1717 if (inode_bitmap < first_block || inode_bitmap > last_block) {
c19204b0 1718 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1719 "Inode bitmap for group %u not in group "
5128273a 1720 "(block %llu)!\n", i, inode_bitmap);
ac27a0ec
DK		 1721 return 0;
1722 }
8fadc143 1723 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1724 if (inode_table < first_block ||
2b2d6d01 1725 inode_table + sbi->s_itb_per_group - 1 > last_block) {
c19204b0 1726 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1727 "Inode table for group %u not in group "
5128273a 1728 "(block %llu)!\n", i, inode_table);
ac27a0ec
DK		 1729 return 0;
1730 }
b5f10eed 1731 spin_lock(sb_bgl_lock(sbi, i));
717d50e4 1732 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
c19204b0 1733 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1734 "Checksum for group %u failed (%u!=%u)\n",
c19204b0
JB		 1735 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1736 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c
LZ		 1737 if (!(sb->s_flags & MS_RDONLY)) {
1738 spin_unlock(sb_bgl_lock(sbi, i));
8a266467 1739 return 0;
7ee1ec4c 1740 }
717d50e4 1741 }
b5f10eed 1742 spin_unlock(sb_bgl_lock(sbi, i));
ce421581
JS		 1743 if (!flexbg_flag)
1744 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1745 }
1746
bd81d8ee 1747 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2b2d6d01 1748 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1749 return 1;
1750}
1751
617ba13b 1752/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1753 * the superblock) which were deleted from all directories, but held open by
1754 * a process at the time of a crash. We walk the list and try to delete these
1755 * inodes at recovery time (only with a read-write filesystem).
1756 *
1757 * In order to keep the orphan inode chain consistent during traversal (in
1758 * case of crash during recovery), we link each inode into the superblock
1759 * orphan list_head and handle it the same way as an inode deletion during
1760 * normal operation (which journals the operations for us).
1761 *
1762 * We only do an iget() and an iput() on each inode, which is very safe if we
1763 * accidentally point at an in-use or already deleted inode. The worst that
1764 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1765 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1766 * e2fsck was run on this filesystem, and it must have already done the orphan
1767 * inode cleanup for us, so we can safely abort without any further action.
1768 */
2b2d6d01
TT		 1769static void ext4_orphan_cleanup(struct super_block *sb,
1770 struct ext4_super_block *es)
ac27a0ec
DK		 1771{
1772 unsigned int s_flags = sb->s_flags;
1773 int nr_orphans = 0, nr_truncates = 0;
1774#ifdef CONFIG_QUOTA
1775 int i;
1776#endif
1777 if (!es->s_last_orphan) {
1778 jbd_debug(4, "no orphan inodes to clean up\n");
1779 return;
1780 }
1781
a8f48a95
ES		 1782 if (bdev_read_only(sb->s_bdev)) {
1783 printk(KERN_ERR "EXT4-fs: write access "
1784 "unavailable, skipping orphan cleanup.\n");
1785 return;
1786 }
1787
617ba13b 1788 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1789 if (es->s_last_orphan)
1790 jbd_debug(1, "Errors on filesystem, "
1791 "clearing orphan list.\n");
1792 es->s_last_orphan = 0;
1793 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1794 return;
1795 }
1796
1797 if (s_flags & MS_RDONLY) {
617ba13b 1798 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1799 sb->s_id);
1800 sb->s_flags &= ~MS_RDONLY;
1801 }
1802#ifdef CONFIG_QUOTA
1803 /* Needed for iput() to work correctly and not trash data */
1804 sb->s_flags |= MS_ACTIVE;
1805 /* Turn on quotas so that they are updated correctly */
1806 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1807 if (EXT4_SB(sb)->s_qf_names[i]) {
1808 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1809 if (ret < 0)
1810 printk(KERN_ERR
2c8be6b2 1811 "EXT4-fs: Cannot turn on journaled "
ac27a0ec
DK		 1812 "quota: error %d\n", ret);
1813 }
1814 }
1815#endif
1816
1817 while (es->s_last_orphan) {
1818 struct inode *inode;
1819
97bd42b9
JB		 1820 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1821 if (IS_ERR(inode)) {
ac27a0ec
DK		 1822 es->s_last_orphan = 0;
1823 break;
1824 }
1825
617ba13b 1826 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
a269eb18 1827 vfs_dq_init(inode);
ac27a0ec
DK		 1828 if (inode->i_nlink) {
1829 printk(KERN_DEBUG
e5f8eab8 1830 "%s: truncating inode %lu to %lld bytes\n",
46e665e9 1831 __func__, inode->i_ino, inode->i_size);
e5f8eab8 1832 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 1833 inode->i_ino, inode->i_size);
617ba13b 1834 ext4_truncate(inode);
ac27a0ec
DK		 1835 nr_truncates++;
1836 } else {
1837 printk(KERN_DEBUG
1838 "%s: deleting unreferenced inode %lu\n",
46e665e9 1839 __func__, inode->i_ino);
ac27a0ec
DK		 1840 jbd_debug(2, "deleting unreferenced inode %lu\n",
1841 inode->i_ino);
1842 nr_orphans++;
1843 }
1844 iput(inode); /* The delete magic happens here! */
1845 }
1846
2b2d6d01 1847#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 1848
1849 if (nr_orphans)
617ba13b 1850 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1851 sb->s_id, PLURAL(nr_orphans));
1852 if (nr_truncates)
617ba13b 1853 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1854 sb->s_id, PLURAL(nr_truncates));
1855#ifdef CONFIG_QUOTA
1856 /* Turn quotas off */
1857 for (i = 0; i < MAXQUOTAS; i++) {
1858 if (sb_dqopt(sb)->files[i])
6f28e087 1859 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1860 }
1861#endif
1862 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1863}
cd2291a4
ES		 1864/*
1865 * Maximal extent format file size.
1866 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1867 * extent format containers, within a sector_t, and within i_blocks
1868 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1869 * so that won't be a limiting factor.
1870 *
1871 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1872 */
f287a1a5 1873static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 1874{
1875 loff_t res;
1876 loff_t upper_limit = MAX_LFS_FILESIZE;
1877
1878 /* small i_blocks in vfs inode? */
f287a1a5 1879 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 1880 /*
b3a6ffe1 1881 * CONFIG_LBD is not enabled implies the inode
cd2291a4
ES		 1882 * i_block represent total blocks in 512 bytes
1883 * 32 == size of vfs inode i_blocks * 8
1884 */
1885 upper_limit = (1LL << 32) - 1;
1886
1887 /* total blocks in file system block size */
1888 upper_limit >>= (blkbits - 9);
1889 upper_limit <<= blkbits;
1890 }
1891
1892 /* 32-bit extent-start container, ee_block */
1893 res = 1LL << 32;
1894 res <<= blkbits;
1895 res -= 1;
1896
1897 /* Sanity check against vm- & vfs- imposed limits */
1898 if (res > upper_limit)
1899 res = upper_limit;
1900
1901 return res;
1902}
ac27a0ec 1903
ac27a0ec 1904/*
cd2291a4 1905 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1906 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1907 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1908 */
f287a1a5 1909static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 1910{
617ba13b 1911 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1912 int meta_blocks;
1913 loff_t upper_limit;
1914 /* This is calculated to be the largest file size for a
cd2291a4 1915 * dense, bitmapped file such that the total number of
ac27a0ec 1916 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1917 * does not exceed 2^48 -1
1918 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1919 * total number of 512 bytes blocks of the file
1920 */
1921
f287a1a5 1922 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 1923 /*
b3a6ffe1 1924 * !has_huge_files or CONFIG_LBD is not enabled
f287a1a5
TT		 1925 * implies the inode i_block represent total blocks in
1926 * 512 bytes 32 == size of vfs inode i_blocks * 8
0fc1b451
AK		 1927 */
1928 upper_limit = (1LL << 32) - 1;
1929
1930 /* total blocks in file system block size */
1931 upper_limit >>= (bits - 9);
1932
1933 } else {
8180a562
AK		 1934 /*
1935 * We use 48 bit ext4_inode i_blocks
1936 * With EXT4_HUGE_FILE_FL set the i_blocks
1937 * represent total number of blocks in
1938 * file system block size
1939 */
0fc1b451
AK		 1940 upper_limit = (1LL << 48) - 1;
1941
0fc1b451
AK		 1942 }
1943
1944 /* indirect blocks */
1945 meta_blocks = 1;
1946 /* double indirect blocks */
1947 meta_blocks += 1 + (1LL << (bits-2));
1948 /* tripple indirect blocks */
1949 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1950
1951 upper_limit -= meta_blocks;
1952 upper_limit <<= bits;
ac27a0ec
DK		 1953
1954 res += 1LL << (bits-2);
1955 res += 1LL << (2*(bits-2));
1956 res += 1LL << (3*(bits-2));
1957 res <<= bits;
1958 if (res > upper_limit)
1959 res = upper_limit;
0fc1b451
AK		 1960
1961 if (res > MAX_LFS_FILESIZE)
1962 res = MAX_LFS_FILESIZE;
1963
ac27a0ec
DK		 1964 return res;
1965}
1966
617ba13b 1967static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1968 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1969{
617ba13b 1970 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1971 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1972 int has_super = 0;
1973
1974 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1975
617ba13b 1976 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1977 nr < first_meta_bg)
70bbb3e0 1978 return logical_sb_block + nr + 1;
ac27a0ec 1979 bg = sbi->s_desc_per_block * nr;
617ba13b 1980 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1981 has_super = 1;
617ba13b 1982 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1983}
1984
c9de560d
AT		 1985/**
1986 * ext4_get_stripe_size: Get the stripe size.
1987 * @sbi: In memory super block info
1988 *
1989 * If we have specified it via mount option, then
1990 * use the mount option value. If the value specified at mount time is
1991 * greater than the blocks per group use the super block value.
1992 * If the super block value is greater than blocks per group return 0.
1993 * Allocator needs it be less than blocks per group.
1994 *
1995 */
1996static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1997{
1998 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1999 unsigned long stripe_width =
2000 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2001
2002 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2003 return sbi->s_stripe;
2004
2005 if (stripe_width <= sbi->s_blocks_per_group)
2006 return stripe_width;
2007
2008 if (stride <= sbi->s_blocks_per_group)
2009 return stride;
2010
2011 return 0;
2012}
ac27a0ec 2013
3197ebdb
TT		 2014/* sysfs supprt */
2015
2016struct ext4_attr {
2017 struct attribute attr;
2018 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2019 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2020 const char *, size_t);
2021 int offset;
2022};
2023
2024static int parse_strtoul(const char *buf,
2025 unsigned long max, unsigned long *value)
2026{
2027 char *endp;
2028
2029 while (*buf && isspace(*buf))
2030 buf++;
2031 *value = simple_strtoul(buf, &endp, 0);
2032 while (*endp && isspace(*endp))
2033 endp++;
2034 if (*endp || *value > max)
2035 return -EINVAL;
2036
2037 return 0;
2038}
2039
2040static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2041 struct ext4_sb_info *sbi,
2042 char *buf)
2043{
2044 return snprintf(buf, PAGE_SIZE, "%llu\n",
2045 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2046}
2047
2048static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2049 struct ext4_sb_info *sbi, char *buf)
2050{
2051 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2052
2053 return snprintf(buf, PAGE_SIZE, "%lu\n",
2054 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2055 sbi->s_sectors_written_start) >> 1);
2056}
2057
2058static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2059 struct ext4_sb_info *sbi, char *buf)
2060{
2061 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2062
2063 return snprintf(buf, PAGE_SIZE, "%llu\n",
2064 sbi->s_kbytes_written +
2065 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2066 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2067}
2068
2069static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2070 struct ext4_sb_info *sbi,
2071 const char *buf, size_t count)
2072{
2073 unsigned long t;
2074
2075 if (parse_strtoul(buf, 0x40000000, &t))
2076 return -EINVAL;
2077
2078 /* inode_readahead_blks must be a power of 2 */
2079 if (t & (t-1))
2080 return -EINVAL;
2081
2082 sbi->s_inode_readahead_blks = t;
2083 return count;
2084}
2085
2086static ssize_t sbi_ui_show(struct ext4_attr *a,
2087 struct ext4_sb_info *sbi, char *buf)
2088{
2089 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2090
2091 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2092}
2093
2094static ssize_t sbi_ui_store(struct ext4_attr *a,
2095 struct ext4_sb_info *sbi,
2096 const char *buf, size_t count)
2097{
2098 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2099 unsigned long t;
2100
2101 if (parse_strtoul(buf, 0xffffffff, &t))
2102 return -EINVAL;
2103 *ui = t;
2104 return count;
2105}
2106
2107#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2108static struct ext4_attr ext4_attr_##_name = { \
2109 .attr = {.name = __stringify(_name), .mode = _mode }, \
2110 .show = _show, \
2111 .store = _store, \
2112 .offset = offsetof(struct ext4_sb_info, _elname), \
2113}
2114#define EXT4_ATTR(name, mode, show, store) \
2115static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2116
2117#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2118#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2119#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2120 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2121#define ATTR_LIST(name) &ext4_attr_##name.attr
2122
2123EXT4_RO_ATTR(delayed_allocation_blocks);
2124EXT4_RO_ATTR(session_write_kbytes);
2125EXT4_RO_ATTR(lifetime_write_kbytes);
2126EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2127 inode_readahead_blks_store, s_inode_readahead_blks);
2128EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2129EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2130EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2131EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2132EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2133EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2134
2135static struct attribute *ext4_attrs[] = {
2136 ATTR_LIST(delayed_allocation_blocks),
2137 ATTR_LIST(session_write_kbytes),
2138 ATTR_LIST(lifetime_write_kbytes),
2139 ATTR_LIST(inode_readahead_blks),
2140 ATTR_LIST(mb_stats),
2141 ATTR_LIST(mb_max_to_scan),
2142 ATTR_LIST(mb_min_to_scan),
2143 ATTR_LIST(mb_order2_req),
2144 ATTR_LIST(mb_stream_req),
2145 ATTR_LIST(mb_group_prealloc),
2146 NULL,
2147};
2148
2149static ssize_t ext4_attr_show(struct kobject *kobj,
2150 struct attribute *attr, char *buf)
2151{
2152 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2153 s_kobj);
2154 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2155
2156 return a->show ? a->show(a, sbi, buf) : 0;
2157}
2158
2159static ssize_t ext4_attr_store(struct kobject *kobj,
2160 struct attribute *attr,
2161 const char *buf, size_t len)
2162{
2163 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2164 s_kobj);
2165 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2166
2167 return a->store ? a->store(a, sbi, buf, len) : 0;
2168}
2169
2170static void ext4_sb_release(struct kobject *kobj)
2171{
2172 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2173 s_kobj);
2174 complete(&sbi->s_kobj_unregister);
2175}
2176
2177
2178static struct sysfs_ops ext4_attr_ops = {
2179 .show = ext4_attr_show,
2180 .store = ext4_attr_store,
2181};
2182
2183static struct kobj_type ext4_ktype = {
2184 .default_attrs = ext4_attrs,
2185 .sysfs_ops = &ext4_attr_ops,
2186 .release = ext4_sb_release,
2187};
2188
2b2d6d01 2189static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 2190 __releases(kernel_lock)
2191 __acquires(kernel_lock)
1d03ec98 2192
ac27a0ec 2193{
2b2d6d01 2194 struct buffer_head *bh;
617ba13b
MC		 2195 struct ext4_super_block *es = NULL;
2196 struct ext4_sb_info *sbi;
2197 ext4_fsblk_t block;
2198 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2199 ext4_fsblk_t logical_sb_block;
ac27a0ec 2200 unsigned long offset = 0;
ac27a0ec
DK		 2201 unsigned long journal_devnum = 0;
2202 unsigned long def_mount_opts;
2203 struct inode *root;
9f6200bb 2204 char *cp;
0390131b 2205 const char *descr;
1d1fe1ee 2206 int ret = -EINVAL;
ac27a0ec 2207 int blocksize;
4ec11028
TT		 2208 unsigned int db_count;
2209 unsigned int i;
f287a1a5 2210 int needs_recovery, has_huge_files;
3a06d778 2211 int features;
bd81d8ee 2212 __u64 blocks_count;
833f4077 2213 int err;
b3881f74 2214 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 2215
2216 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2217 if (!sbi)
2218 return -ENOMEM;
705895b6
PE		 2219
2220 sbi->s_blockgroup_lock =
2221 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2222 if (!sbi->s_blockgroup_lock) {
2223 kfree(sbi);
2224 return -ENOMEM;
2225 }
ac27a0ec
DK		 2226 sb->s_fs_info = sbi;
2227 sbi->s_mount_opt = 0;
617ba13b
MC		 2228 sbi->s_resuid = EXT4_DEF_RESUID;
2229 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2230 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2231 sbi->s_sb_block = sb_block;
afc32f7e
TT		 2232 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2233 sectors[1]);
ac27a0ec
DK		 2234
2235 unlock_kernel();
2236
9f6200bb
TT		 2237 /* Cleanup superblock name */
2238 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2239 *cp = '!';
2240
617ba13b 2241 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2242 if (!blocksize) {
617ba13b 2243 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 2244 goto out_fail;
2245 }
2246
2247 /*
617ba13b 2248 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 2249 * block sizes. We need to calculate the offset from buffer start.
2250 */
617ba13b 2251 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 2252 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2253 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 2254 } else {
70bbb3e0 2255 logical_sb_block = sb_block;
ac27a0ec
DK		 2256 }
2257
70bbb3e0 2258 if (!(bh = sb_bread(sb, logical_sb_block))) {
2b2d6d01 2259 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 2260 goto out_fail;
2261 }
2262 /*
2263 * Note: s_es must be initialized as soon as possible because
617ba13b 2264 * some ext4 macro-instructions depend on its value
ac27a0ec 2265 */
617ba13b 2266 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 2267 sbi->s_es = es;
2268 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 2269 if (sb->s_magic != EXT4_SUPER_MAGIC)
2270 goto cantfind_ext4;
afc32f7e 2271 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 2272
2273 /* Set defaults before we parse the mount options */
2274 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 2275 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 2276 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 2277 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 2278 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 2279 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 2280 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 2281#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 2282 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 2283 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 2284#endif
03010a33 2285#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2286 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2287 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2288#endif
617ba13b
MC		 2289 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2290 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2291 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2292 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2293 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2294 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2295
2296 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2297 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2298 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2299 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2300 else
2301 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2302
2303 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2304 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 2305 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2306 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2307 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec
DK		 2308
2309 set_opt(sbi->s_mount_opt, RESERVATION);
571640ca 2310 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2311
dd919b98
AK		 2312 /*
2313 * enable delayed allocation by default
2314 * Use -o nodelalloc to turn it off
2315 */
2316 set_opt(sbi->s_mount_opt, DELALLOC);
2317
2318
b3881f74
TT		 2319 if (!parse_options((char *) data, sb, &journal_devnum,
2320 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 2321 goto failed_mount;
2322
2323 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2324 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2325
617ba13b
MC		 2326 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2327 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2328 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2329 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 2330 printk(KERN_WARNING
617ba13b 2331 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec 2332 "running e2fsck is recommended\n");
469108ff 2333
ac27a0ec
DK		 2334 /*
2335 * Check feature flags regardless of the revision level, since we
2336 * previously didn't change the revision level when setting the flags,
2337 * so there is a chance incompat flags are set on a rev 0 filesystem.
2338 */
617ba13b 2339 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 2340 if (features) {
617ba13b 2341 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
3a06d778
AK		 2342 "unsupported optional features (%x).\n", sb->s_id,
2343 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2344 ~EXT4_FEATURE_INCOMPAT_SUPP));
ac27a0ec
DK		 2345 goto failed_mount;
2346 }
617ba13b 2347 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 2348 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 2349 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
3a06d778
AK		 2350 "unsupported optional features (%x).\n", sb->s_id,
2351 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2352 ~EXT4_FEATURE_RO_COMPAT_SUPP));
ac27a0ec
DK		 2353 goto failed_mount;
2354 }
f287a1a5
TT		 2355 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2356 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2357 if (has_huge_files) {
0fc1b451
AK		 2358 /*
2359 * Large file size enabled file system can only be
b3a6ffe1 2360 * mount if kernel is build with CONFIG_LBD
0fc1b451
AK		 2361 */
2362 if (sizeof(root->i_blocks) < sizeof(u64) &&
2363 !(sb->s_flags & MS_RDONLY)) {
2364 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2365 "files cannot be mounted read-write "
b3a6ffe1 2366 "without CONFIG_LBD.\n", sb->s_id);
0fc1b451
AK		 2367 goto failed_mount;
2368 }
2369 }
ac27a0ec
DK		 2370 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2371
617ba13b
MC		 2372 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2373 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 2374 printk(KERN_ERR
617ba13b 2375 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 2376 blocksize, sb->s_id);
2377 goto failed_mount;
2378 }
2379
ac27a0ec 2380 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2381
2382 /* Validate the filesystem blocksize */
2383 if (!sb_set_blocksize(sb, blocksize)) {
2384 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2385 blocksize);
ac27a0ec
DK		 2386 goto failed_mount;
2387 }
2388
2b2d6d01 2389 brelse(bh);
70bbb3e0
AM		 2390 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2391 offset = do_div(logical_sb_block, blocksize);
2392 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 2393 if (!bh) {
2394 printk(KERN_ERR
617ba13b 2395 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 2396 goto failed_mount;
2397 }
617ba13b 2398 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2399 sbi->s_es = es;
617ba13b 2400 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2b2d6d01
TT		 2401 printk(KERN_ERR
2402 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 2403 goto failed_mount;
2404 }
2405 }
2406
f287a1a5
TT		 2407 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2408 has_huge_files);
2409 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2410
617ba13b
MC		 2411 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2412 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2413 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2414 } else {
2415 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2416 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2417 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2418 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2419 (sbi->s_inode_size > blocksize)) {
2b2d6d01
TT		 2420 printk(KERN_ERR
2421 "EXT4-fs: unsupported inode size: %d\n",
2422 sbi->s_inode_size);
ac27a0ec
DK		 2423 goto failed_mount;
2424 }
ef7f3835
KS		 2425 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2426 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2427 }
0d1ee42f
AR		 2428 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2429 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2430 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2431 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2432 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2433 printk(KERN_ERR
8fadc143 2434 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2435 sbi->s_desc_size);
2436 goto failed_mount;
2437 }
2438 } else
2439 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2440 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2441 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2442 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2443 goto cantfind_ext4;
2444 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2445 if (sbi->s_inodes_per_block == 0)
617ba13b 2446 goto cantfind_ext4;
ac27a0ec
DK		 2447 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2448 sbi->s_inodes_per_block;
0d1ee42f 2449 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2450 sbi->s_sbh = bh;
2451 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2452 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2453 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2b2d6d01 2454 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2455 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2456 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2457 i = le32_to_cpu(es->s_flags);
2458 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2459 sbi->s_hash_unsigned = 3;
2460 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2461#ifdef __CHAR_UNSIGNED__
2462 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2463 sbi->s_hash_unsigned = 3;
2464#else
2465 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2466#endif
2467 sb->s_dirt = 1;
2468 }
ac27a0ec
DK		 2469
2470 if (sbi->s_blocks_per_group > blocksize * 8) {
2b2d6d01
TT		 2471 printk(KERN_ERR
2472 "EXT4-fs: #blocks per group too big: %lu\n",
2473 sbi->s_blocks_per_group);
ac27a0ec
DK		 2474 goto failed_mount;
2475 }
ac27a0ec 2476 if (sbi->s_inodes_per_group > blocksize * 8) {
2b2d6d01
TT		 2477 printk(KERN_ERR
2478 "EXT4-fs: #inodes per group too big: %lu\n",
2479 sbi->s_inodes_per_group);
ac27a0ec
DK		 2480 goto failed_mount;
2481 }
2482
bd81d8ee 2483 if (ext4_blocks_count(es) >
ac27a0ec 2484 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2485 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2486 " too large to mount safely\n", sb->s_id);
2487 if (sizeof(sector_t) < 8)
617ba13b 2488 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2489 "enabled\n");
2490 goto failed_mount;
2491 }
2492
617ba13b
MC		 2493 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2494 goto cantfind_ext4;
e7c95593 2495
4ec11028
TT		 2496 /*
2497 * It makes no sense for the first data block to be beyond the end
2498 * of the filesystem.
2499 */
2500 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2501 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2502 "block %u is beyond end of filesystem (%llu)\n",
2503 le32_to_cpu(es->s_first_data_block),
2504 ext4_blocks_count(es));
e7c95593
ES		 2505 goto failed_mount;
2506 }
bd81d8ee
LV		 2507 blocks_count = (ext4_blocks_count(es) -
2508 le32_to_cpu(es->s_first_data_block) +
2509 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2510 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028
TT		 2511 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2512 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2513 "(block count %llu, first data block %u, "
2514 "blocks per group %lu)\n", sbi->s_groups_count,
2515 ext4_blocks_count(es),
2516 le32_to_cpu(es->s_first_data_block),
2517 EXT4_BLOCKS_PER_GROUP(sb));
2518 goto failed_mount;
2519 }
bd81d8ee 2520 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2521 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2522 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2523 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2524 GFP_KERNEL);
2525 if (sbi->s_group_desc == NULL) {
2b2d6d01 2526 printk(KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2527 goto failed_mount;
2528 }
2529
3244fcb1 2530#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2531 if (ext4_proc_root)
2532 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 2533#endif
240799cd 2534
705895b6 2535 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2536
2537 for (i = 0; i < db_count; i++) {
70bbb3e0 2538 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2539 sbi->s_group_desc[i] = sb_bread(sb, block);
2540 if (!sbi->s_group_desc[i]) {
2b2d6d01
TT		 2541 printk(KERN_ERR "EXT4-fs: "
2542 "can't read group descriptor %d\n", i);
ac27a0ec
DK		 2543 db_count = i;
2544 goto failed_mount2;
2545 }
2546 }
2b2d6d01 2547 if (!ext4_check_descriptors(sb)) {
617ba13b 2548 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2549 goto failed_mount2;
2550 }
772cb7c8
JS		 2551 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2552 if (!ext4_fill_flex_info(sb)) {
2553 printk(KERN_ERR
2554 "EXT4-fs: unable to initialize "
2555 "flex_bg meta info!\n");
2556 goto failed_mount2;
2557 }
2558
ac27a0ec
DK		 2559 sbi->s_gdb_count = db_count;
2560 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2561 spin_lock_init(&sbi->s_next_gen_lock);
2562
833f4077
PZ		 2563 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2564 ext4_count_free_blocks(sb));
2565 if (!err) {
2566 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2567 ext4_count_free_inodes(sb));
2568 }
2569 if (!err) {
2570 err = percpu_counter_init(&sbi->s_dirs_counter,
2571 ext4_count_dirs(sb));
2572 }
6bc6e63f
AK		 2573 if (!err) {
2574 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2575 }
833f4077
PZ		 2576 if (err) {
2577 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2578 goto failed_mount3;
2579 }
ac27a0ec 2580
c9de560d
AT		 2581 sbi->s_stripe = ext4_get_stripe_size(sbi);
2582
ac27a0ec
DK		 2583 /*
2584 * set up enough so that it can read an inode
2585 */
617ba13b
MC		 2586 sb->s_op = &ext4_sops;
2587 sb->s_export_op = &ext4_export_ops;
2588 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2589#ifdef CONFIG_QUOTA
617ba13b
MC		 2590 sb->s_qcop = &ext4_qctl_operations;
2591 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2592#endif
2593 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2594
2595 sb->s_root = NULL;
2596
2597 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2598 EXT4_HAS_INCOMPAT_FEATURE(sb,
2599 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2600
2601 /*
2602 * The first inode we look at is the journal inode. Don't try
2603 * root first: it may be modified in the journal!
2604 */
2605 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2606 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2607 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2608 goto failed_mount3;
624080ed
TT		 2609 if (!(sb->s_flags & MS_RDONLY) &&
2610 EXT4_SB(sb)->s_journal->j_failed_commit) {
2611 printk(KERN_CRIT "EXT4-fs error (device %s): "
2612 "ext4_fill_super: Journal transaction "
2b2d6d01 2613 "%u is corrupt\n", sb->s_id,
624080ed 2614 EXT4_SB(sb)->s_journal->j_failed_commit);
2b2d6d01
TT		 2615 if (test_opt(sb, ERRORS_RO)) {
2616 printk(KERN_CRIT
2617 "Mounting filesystem read-only\n");
624080ed
TT		 2618 sb->s_flags |= MS_RDONLY;
2619 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2620 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2621 }
2622 if (test_opt(sb, ERRORS_PANIC)) {
2623 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2624 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2625 ext4_commit_super(sb, es, 1);
624080ed
TT		 2626 goto failed_mount4;
2627 }
2628 }
0390131b
FM		 2629 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2630 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2631 printk(KERN_ERR "EXT4-fs: required journal recovery "
2632 "suppressed and not mounted read-only\n");
2633 goto failed_mount4;
ac27a0ec 2634 } else {
0390131b
FM		 2635 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2636 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2637 sbi->s_journal = NULL;
2638 needs_recovery = 0;
2639 goto no_journal;
ac27a0ec
DK		 2640 }
2641
eb40a09c
JS		 2642 if (ext4_blocks_count(es) > 0xffffffffULL &&
2643 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2644 JBD2_FEATURE_INCOMPAT_64BIT)) {
abda1418 2645 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
eb40a09c
JS		 2646 goto failed_mount4;
2647 }
2648
818d276c
GS		 2649 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2650 jbd2_journal_set_features(sbi->s_journal,
2651 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2652 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2653 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2654 jbd2_journal_set_features(sbi->s_journal,
2655 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2656 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2657 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2658 } else {
2659 jbd2_journal_clear_features(sbi->s_journal,
2660 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2661 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2662 }
2663
ac27a0ec
DK		 2664 /* We have now updated the journal if required, so we can
2665 * validate the data journaling mode. */
2666 switch (test_opt(sb, DATA_FLAGS)) {
2667 case 0:
2668 /* No mode set, assume a default based on the journal
63f57933
AM		 2669 * capabilities: ORDERED_DATA if the journal can
2670 * cope, else JOURNAL_DATA
2671 */
dab291af
MC		 2672 if (jbd2_journal_check_available_features
2673 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2674 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2675 else
2676 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2677 break;
2678
617ba13b
MC		 2679 case EXT4_MOUNT_ORDERED_DATA:
2680 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2681 if (!jbd2_journal_check_available_features
2682 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2683 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2684 "requested data journaling mode\n");
2685 goto failed_mount4;
2686 }
2687 default:
2688 break;
2689 }
b3881f74 2690 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 2691
0390131b 2692no_journal:
ac27a0ec
DK		 2693
2694 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2695 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2696 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2697 "its supported only with writeback mode\n");
2698 clear_opt(sbi->s_mount_opt, NOBH);
2699 }
2700 }
2701 /*
dab291af 2702 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2703 * so we can safely mount the rest of the filesystem now.
2704 */
2705
1d1fe1ee
DH		 2706 root = ext4_iget(sb, EXT4_ROOT_INO);
2707 if (IS_ERR(root)) {
617ba13b 2708 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1d1fe1ee 2709 ret = PTR_ERR(root);
ac27a0ec
DK		 2710 goto failed_mount4;
2711 }
2712 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2713 iput(root);
617ba13b 2714 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2715 goto failed_mount4;
2716 }
1d1fe1ee
DH		 2717 sb->s_root = d_alloc_root(root);
2718 if (!sb->s_root) {
2719 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2720 iput(root);
2721 ret = -ENOMEM;
2722 goto failed_mount4;
2723 }
ac27a0ec 2724
2b2d6d01 2725 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2726
2727 /* determine the minimum size of new large inodes, if present */
2728 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2729 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2730 EXT4_GOOD_OLD_INODE_SIZE;
2731 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2732 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2733 if (sbi->s_want_extra_isize <
2734 le16_to_cpu(es->s_want_extra_isize))
2735 sbi->s_want_extra_isize =
2736 le16_to_cpu(es->s_want_extra_isize);
2737 if (sbi->s_want_extra_isize <
2738 le16_to_cpu(es->s_min_extra_isize))
2739 sbi->s_want_extra_isize =
2740 le16_to_cpu(es->s_min_extra_isize);
2741 }
2742 }
2743 /* Check if enough inode space is available */
2744 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2745 sbi->s_inode_size) {
2746 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2747 EXT4_GOOD_OLD_INODE_SIZE;
2748 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2749 "available.\n");
2750 }
2751
c2774d84
AK		 2752 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2753 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2754 "requested data journaling mode\n");
2755 clear_opt(sbi->s_mount_opt, DELALLOC);
2756 } else if (test_opt(sb, DELALLOC))
2757 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2758
2759 ext4_ext_init(sb);
2760 err = ext4_mb_init(sb, needs_recovery);
2761 if (err) {
2762 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2763 err);
2764 goto failed_mount4;
2765 }
2766
3197ebdb
TT		 2767 sbi->s_kobj.kset = ext4_kset;
2768 init_completion(&sbi->s_kobj_unregister);
2769 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2770 "%s", sb->s_id);
2771 if (err) {
2772 ext4_mb_release(sb);
2773 ext4_ext_release(sb);
2774 goto failed_mount4;
2775 };
2776
ac27a0ec
DK		 2777 /*
2778 * akpm: core read_super() calls in here with the superblock locked.
2779 * That deadlocks, because orphan cleanup needs to lock the superblock
2780 * in numerous places. Here we just pop the lock - it's relatively
2781 * harmless, because we are now ready to accept write_super() requests,
2782 * and aviro says that's the only reason for hanging onto the
2783 * superblock lock.
2784 */
617ba13b
MC		 2785 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2786 ext4_orphan_cleanup(sb, es);
2787 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 2788 if (needs_recovery) {
2b2d6d01 2789 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
0390131b
FM		 2790 ext4_mark_recovery_complete(sb, es);
2791 }
2792 if (EXT4_SB(sb)->s_journal) {
2793 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2794 descr = " journalled data mode";
2795 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2796 descr = " ordered data mode";
2797 else
2798 descr = " writeback data mode";
2799 } else
2800 descr = "out journal";
2801
2802 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2803 sb->s_id, descr);
ac27a0ec
DK		 2804
2805 lock_kernel();
2806 return 0;
2807
617ba13b 2808cantfind_ext4:
ac27a0ec 2809 if (!silent)
617ba13b 2810 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2811 sb->s_id);
2812 goto failed_mount;
2813
2814failed_mount4:
0390131b
FM		 2815 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2816 if (sbi->s_journal) {
2817 jbd2_journal_destroy(sbi->s_journal);
2818 sbi->s_journal = NULL;
2819 }
ac27a0ec
DK		 2820failed_mount3:
2821 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2822 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2823 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 2824 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 2825failed_mount2:
2826 for (i = 0; i < db_count; i++)
2827 brelse(sbi->s_group_desc[i]);
2828 kfree(sbi->s_group_desc);
2829failed_mount:
240799cd 2830 if (sbi->s_proc) {
9f6200bb 2831 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 2832 }
ac27a0ec
DK		 2833#ifdef CONFIG_QUOTA
2834 for (i = 0; i < MAXQUOTAS; i++)
2835 kfree(sbi->s_qf_names[i]);
2836#endif
617ba13b 2837 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2838 brelse(bh);
2839out_fail:
2840 sb->s_fs_info = NULL;
2841 kfree(sbi);
2842 lock_kernel();
1d1fe1ee 2843 return ret;
ac27a0ec
DK		 2844}
2845
2846/*
2847 * Setup any per-fs journal parameters now. We'll do this both on
2848 * initial mount, once the journal has been initialised but before we've
2849 * done any recovery; and again on any subsequent remount.
2850 */
617ba13b 2851static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2852{
617ba13b 2853 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 2854
30773840
TT		 2855 journal->j_commit_interval = sbi->s_commit_interval;
2856 journal->j_min_batch_time = sbi->s_min_batch_time;
2857 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 2858
2859 spin_lock(&journal->j_state_lock);
2860 if (test_opt(sb, BARRIER))
dab291af 2861 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2862 else
dab291af 2863 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 2864 if (test_opt(sb, DATA_ERR_ABORT))
2865 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2866 else
2867 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 2868 spin_unlock(&journal->j_state_lock);
2869}
2870
617ba13b 2871static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2872 unsigned int journal_inum)
2873{
2874 struct inode *journal_inode;
2875 journal_t *journal;
2876
0390131b
FM		 2877 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2878
ac27a0ec
DK		 2879 /* First, test for the existence of a valid inode on disk. Bad
2880 * things happen if we iget() an unused inode, as the subsequent
2881 * iput() will try to delete it. */
2882
1d1fe1ee
DH		 2883 journal_inode = ext4_iget(sb, journal_inum);
2884 if (IS_ERR(journal_inode)) {
617ba13b 2885 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2886 return NULL;
2887 }
2888 if (!journal_inode->i_nlink) {
2889 make_bad_inode(journal_inode);
2890 iput(journal_inode);
617ba13b 2891 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2892 return NULL;
2893 }
2894
e5f8eab8 2895 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 2896 journal_inode, journal_inode->i_size);
1d1fe1ee 2897 if (!S_ISREG(journal_inode->i_mode)) {
617ba13b 2898 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2899 iput(journal_inode);
2900 return NULL;
2901 }
2902
dab291af 2903 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2904 if (!journal) {
617ba13b 2905 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2906 iput(journal_inode);
2907 return NULL;
2908 }
2909 journal->j_private = sb;
617ba13b 2910 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2911 return journal;
2912}
2913
617ba13b 2914static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2915 dev_t j_dev)
2916{
2b2d6d01 2917 struct buffer_head *bh;
ac27a0ec 2918 journal_t *journal;
617ba13b
MC		 2919 ext4_fsblk_t start;
2920 ext4_fsblk_t len;
ac27a0ec 2921 int hblock, blocksize;
617ba13b 2922 ext4_fsblk_t sb_block;
ac27a0ec 2923 unsigned long offset;
2b2d6d01 2924 struct ext4_super_block *es;
ac27a0ec
DK		 2925 struct block_device *bdev;
2926
0390131b
FM		 2927 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2928
617ba13b 2929 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2930 if (bdev == NULL)
2931 return NULL;
2932
2933 if (bd_claim(bdev, sb)) {
2934 printk(KERN_ERR
abda1418 2935 "EXT4-fs: failed to claim external journal device.\n");
9a1c3542 2936 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 2937 return NULL;
2938 }
2939
2940 blocksize = sb->s_blocksize;
2941 hblock = bdev_hardsect_size(bdev);
2942 if (blocksize < hblock) {
2943 printk(KERN_ERR
617ba13b 2944 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2945 goto out_bdev;
2946 }
2947
617ba13b
MC		 2948 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2949 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2950 set_blocksize(bdev, blocksize);
2951 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2952 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2953 "external journal\n");
2954 goto out_bdev;
2955 }
2956
617ba13b
MC		 2957 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2958 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2959 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2960 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2961 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2962 "bad superblock\n");
2963 brelse(bh);
2964 goto out_bdev;
2965 }
2966
617ba13b
MC		 2967 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2968 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2969 brelse(bh);
2970 goto out_bdev;
2971 }
2972
bd81d8ee 2973 len = ext4_blocks_count(es);
ac27a0ec
DK		 2974 start = sb_block + 1;
2975 brelse(bh); /* we're done with the superblock */
2976
dab291af 2977 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2978 start, len, blocksize);
2979 if (!journal) {
617ba13b 2980 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2981 goto out_bdev;
2982 }
2983 journal->j_private = sb;
2984 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2985 wait_on_buffer(journal->j_sb_buffer);
2986 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2987 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2988 goto out_journal;
2989 }
2990 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2991 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2992 "user (unsupported) - %d\n",
2993 be32_to_cpu(journal->j_superblock->s_nr_users));
2994 goto out_journal;
2995 }
617ba13b
MC		 2996 EXT4_SB(sb)->journal_bdev = bdev;
2997 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2998 return journal;
2999out_journal:
dab291af 3000 jbd2_journal_destroy(journal);
ac27a0ec 3001out_bdev:
617ba13b 3002 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3003 return NULL;
3004}
3005
617ba13b
MC		 3006static int ext4_load_journal(struct super_block *sb,
3007 struct ext4_super_block *es,
ac27a0ec
DK		 3008 unsigned long journal_devnum)
3009{
3010 journal_t *journal;
3011 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3012 dev_t journal_dev;
3013 int err = 0;
3014 int really_read_only;
3015
0390131b
FM		 3016 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3017
ac27a0ec
DK		 3018 if (journal_devnum &&
3019 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 3020 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 3021 "numbers have changed\n");
3022 journal_dev = new_decode_dev(journal_devnum);
3023 } else
3024 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3025
3026 really_read_only = bdev_read_only(sb->s_bdev);
3027
3028 /*
3029 * Are we loading a blank journal or performing recovery after a
3030 * crash? For recovery, we need to check in advance whether we
3031 * can get read-write access to the device.
3032 */
3033
617ba13b 3034 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3035 if (sb->s_flags & MS_RDONLY) {
617ba13b 3036 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 3037 "required on readonly filesystem.\n");
3038 if (really_read_only) {
617ba13b 3039 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 3040 "unavailable, cannot proceed.\n");
3041 return -EROFS;
3042 }
2b2d6d01
TT		 3043 printk(KERN_INFO "EXT4-fs: write access will "
3044 "be enabled during recovery.\n");
ac27a0ec
DK		 3045 }
3046 }
3047
3048 if (journal_inum && journal_dev) {
617ba13b 3049 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 3050 "and inode journals!\n");
3051 return -EINVAL;
3052 }
3053
3054 if (journal_inum) {
617ba13b 3055 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3056 return -EINVAL;
3057 } else {
617ba13b 3058 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3059 return -EINVAL;
3060 }
3061
4776004f
TT		 3062 if (journal->j_flags & JBD2_BARRIER)
3063 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
3064 else
3065 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
3066
ac27a0ec 3067 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3068 err = jbd2_journal_update_format(journal);
ac27a0ec 3069 if (err) {
617ba13b 3070 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 3071 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3072 return err;
3073 }
3074 }
3075
617ba13b 3076 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3077 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 3078 if (!err)
dab291af 3079 err = jbd2_journal_load(journal);
ac27a0ec
DK		 3080
3081 if (err) {
617ba13b 3082 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 3083 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3084 return err;
3085 }
3086
617ba13b
MC		 3087 EXT4_SB(sb)->s_journal = journal;
3088 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 3089
3090 if (journal_devnum &&
3091 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3092 es->s_journal_dev = cpu_to_le32(journal_devnum);
3093 sb->s_dirt = 1;
3094
3095 /* Make sure we flush the recovery flag to disk. */
617ba13b 3096 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 3097 }
3098
3099 return 0;
3100}
3101
c4be0c1d 3102static int ext4_commit_super(struct super_block *sb,
2b2d6d01 3103 struct ext4_super_block *es, int sync)
ac27a0ec 3104{
617ba13b 3105 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3106 int error = 0;
ac27a0ec
DK		 3107
3108 if (!sbh)
c4be0c1d 3109 return error;
914258bf
TT		 3110 if (buffer_write_io_error(sbh)) {
3111 /*
3112 * Oh, dear. A previous attempt to write the
3113 * superblock failed. This could happen because the
3114 * USB device was yanked out. Or it could happen to
3115 * be a transient write error and maybe the block will
3116 * be remapped. Nothing we can do but to retry the
3117 * write and hope for the best.
3118 */
abda1418 3119 printk(KERN_ERR "EXT4-fs: previous I/O error to "
914258bf
TT		 3120 "superblock detected for %s.\n", sb->s_id);
3121 clear_buffer_write_io_error(sbh);
3122 set_buffer_uptodate(sbh);
3123 }
ac27a0ec 3124 es->s_wtime = cpu_to_le32(get_seconds());
afc32f7e
TT		 3125 es->s_kbytes_written =
3126 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3127 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3128 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5d1b1b3f
AK		 3129 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3130 &EXT4_SB(sb)->s_freeblocks_counter));
3131 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3132 &EXT4_SB(sb)->s_freeinodes_counter));
3133
ac27a0ec
DK		 3134 BUFFER_TRACE(sbh, "marking dirty");
3135 mark_buffer_dirty(sbh);
914258bf 3136 if (sync) {
c4be0c1d
TS		 3137 error = sync_dirty_buffer(sbh);
3138 if (error)
3139 return error;
3140
3141 error = buffer_write_io_error(sbh);
3142 if (error) {
abda1418 3143 printk(KERN_ERR "EXT4-fs: I/O error while writing "
914258bf
TT		 3144 "superblock for %s.\n", sb->s_id);
3145 clear_buffer_write_io_error(sbh);
3146 set_buffer_uptodate(sbh);
3147 }
3148 }
c4be0c1d 3149 return error;
ac27a0ec
DK		 3150}
3151
3152
3153/*
3154 * Have we just finished recovery? If so, and if we are mounting (or
3155 * remounting) the filesystem readonly, then we will end up with a
3156 * consistent fs on disk. Record that fact.
3157 */
2b2d6d01
TT		 3158static void ext4_mark_recovery_complete(struct super_block *sb,
3159 struct ext4_super_block *es)
ac27a0ec 3160{
617ba13b 3161 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3162
0390131b
FM		 3163 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3164 BUG_ON(journal != NULL);
3165 return;
3166 }
dab291af 3167 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 3168 if (jbd2_journal_flush(journal) < 0)
3169 goto out;
3170
32c37730 3171 lock_super(sb);
617ba13b 3172 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 3173 sb->s_flags & MS_RDONLY) {
617ba13b 3174 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 3175 sb->s_dirt = 0;
617ba13b 3176 ext4_commit_super(sb, es, 1);
ac27a0ec 3177 }
32c37730 3178 unlock_super(sb);
7ffe1ea8
HK		 3179
3180out:
dab291af 3181 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 3182}
3183
3184/*
3185 * If we are mounting (or read-write remounting) a filesystem whose journal
3186 * has recorded an error from a previous lifetime, move that error to the
3187 * main filesystem now.
3188 */
2b2d6d01
TT		 3189static void ext4_clear_journal_err(struct super_block *sb,
3190 struct ext4_super_block *es)
ac27a0ec
DK		 3191{
3192 journal_t *journal;
3193 int j_errno;
3194 const char *errstr;
3195
0390131b
FM		 3196 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3197
617ba13b 3198 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 3199
3200 /*
3201 * Now check for any error status which may have been recorded in the
617ba13b 3202 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 3203 */
3204
dab291af 3205 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 3206 if (j_errno) {
3207 char nbuf[16];
3208
617ba13b 3209 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 3210 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 3211 "from previous mount: %s", errstr);
46e665e9 3212 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 3213 "filesystem check.");
3214
617ba13b
MC		 3215 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3216 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2b2d6d01 3217 ext4_commit_super(sb, es, 1);
ac27a0ec 3218
dab291af 3219 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 3220 }
3221}
3222
3223/*
3224 * Force the running and committing transactions to commit,
3225 * and wait on the commit.
3226 */
617ba13b 3227int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 3228{
3229 journal_t *journal;
0390131b 3230 int ret = 0;
ac27a0ec
DK		 3231
3232 if (sb->s_flags & MS_RDONLY)
3233 return 0;
3234
617ba13b 3235 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 3236 if (journal) {
3237 sb->s_dirt = 0;
3238 ret = ext4_journal_force_commit(journal);
3239 }
3240
ac27a0ec
DK		 3241 return ret;
3242}
3243
3244/*
617ba13b 3245 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec 3246 * have to propagate any other updates to the superblock on disk at this
14ce0cb4
TT		 3247 * point. (We can probably nuke this function altogether, and remove
3248 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
ac27a0ec 3249 */
2b2d6d01 3250static void ext4_write_super(struct super_block *sb)
ac27a0ec 3251{
0390131b
FM		 3252 if (EXT4_SB(sb)->s_journal) {
3253 if (mutex_trylock(&sb->s_lock) != 0)
3254 BUG();
3255 sb->s_dirt = 0;
3256 } else {
3257 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3258 }
ac27a0ec
DK		 3259}
3260
617ba13b 3261static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 3262{
14ce0cb4 3263 int ret = 0;
9eddacf9 3264 tid_t target;
ac27a0ec 3265
ede86cc4 3266 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
ac27a0ec 3267 sb->s_dirt = 0;
0390131b 3268 if (EXT4_SB(sb)->s_journal) {
9eddacf9
JK		 3269 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal,
3270 &target)) {
3271 if (wait)
3272 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal,
3273 target);
3274 }
0390131b
FM		 3275 } else {
3276 ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
3277 }
14ce0cb4 3278 return ret;
ac27a0ec
DK		 3279}
3280
3281/*
3282 * LVM calls this function before a (read-only) snapshot is created. This
3283 * gives us a chance to flush the journal completely and mark the fs clean.
3284 */
c4be0c1d 3285static int ext4_freeze(struct super_block *sb)
ac27a0ec 3286{
c4be0c1d
TS		 3287 int error = 0;
3288 journal_t *journal;
ac27a0ec
DK		 3289 sb->s_dirt = 0;
3290
3291 if (!(sb->s_flags & MS_RDONLY)) {
c4be0c1d 3292 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3293
0390131b
FM		 3294 if (journal) {
3295 /* Now we set up the journal barrier. */
3296 jbd2_journal_lock_updates(journal);
7ffe1ea8 3297
0390131b
FM		 3298 /*
3299 * We don't want to clear needs_recovery flag when we
3300 * failed to flush the journal.
3301 */
c4be0c1d
TS		 3302 error = jbd2_journal_flush(journal);
3303 if (error < 0)
3304 goto out;
0390131b 3305 }
ac27a0ec
DK		 3306
3307 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b 3308 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
c4be0c1d
TS		 3309 error = ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3310 if (error)
3311 goto out;
ac27a0ec 3312 }
c4be0c1d
TS		 3313 return 0;
3314out:
3315 jbd2_journal_unlock_updates(journal);
3316 return error;
ac27a0ec
DK		 3317}
3318
3319/*
3320 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3321 * flag here, even though the filesystem is not technically dirty yet.
3322 */
c4be0c1d 3323static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 3324{
0390131b 3325 if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
ac27a0ec
DK		 3326 lock_super(sb);
3327 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 3328 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3329 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 3330 unlock_super(sb);
dab291af 3331 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec 3332 }
c4be0c1d 3333 return 0;
ac27a0ec
DK		 3334}
3335
2b2d6d01 3336static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3337{
2b2d6d01 3338 struct ext4_super_block *es;
617ba13b
MC		 3339 struct ext4_sb_info *sbi = EXT4_SB(sb);
3340 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3341 unsigned long old_sb_flags;
617ba13b 3342 struct ext4_mount_options old_opts;
8a266467 3343 ext4_group_t g;
b3881f74 3344 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 3345 int err;
3346#ifdef CONFIG_QUOTA
3347 int i;
3348#endif
3349
3350 /* Store the original options */
3351 old_sb_flags = sb->s_flags;
3352 old_opts.s_mount_opt = sbi->s_mount_opt;
3353 old_opts.s_resuid = sbi->s_resuid;
3354 old_opts.s_resgid = sbi->s_resgid;
3355 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 3356 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3357 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3358#ifdef CONFIG_QUOTA
3359 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3360 for (i = 0; i < MAXQUOTAS; i++)
3361 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3362#endif
b3881f74
TT		 3363 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3364 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 3365
3366 /*
3367 * Allow the "check" option to be passed as a remount option.
3368 */
b3881f74
TT		 3369 if (!parse_options(data, sb, NULL, &journal_ioprio,
3370 &n_blocks_count, 1)) {
ac27a0ec
DK		 3371 err = -EINVAL;
3372 goto restore_opts;
3373 }
3374
617ba13b 3375 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
46e665e9 3376 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 3377
3378 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 3379 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3380
3381 es = sbi->s_es;
3382
b3881f74 3383 if (sbi->s_journal) {
0390131b 3384 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 3385 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3386 }
ac27a0ec
DK		 3387
3388 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3389 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 3390 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 3391 err = -EROFS;
3392 goto restore_opts;
3393 }
3394
3395 if (*flags & MS_RDONLY) {
3396 /*
3397 * First of all, the unconditional stuff we have to do
3398 * to disable replay of the journal when we next remount
3399 */
3400 sb->s_flags |= MS_RDONLY;
3401
3402 /*
3403 * OK, test if we are remounting a valid rw partition
3404 * readonly, and if so set the rdonly flag and then
3405 * mark the partition as valid again.
3406 */
617ba13b
MC		 3407 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3408 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3409 es->s_state = cpu_to_le16(sbi->s_mount_state);
3410
32c37730
JK		 3411 /*
3412 * We have to unlock super so that we can wait for
3413 * transactions.
3414 */
0390131b
FM		 3415 if (sbi->s_journal) {
3416 unlock_super(sb);
3417 ext4_mark_recovery_complete(sb, es);
3418 lock_super(sb);
3419 }
ac27a0ec 3420 } else {
3a06d778 3421 int ret;
617ba13b
MC		 3422 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3423 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3424 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec 3425 "remount RDWR because of unsupported "
3a06d778
AK		 3426 "optional features (%x).\n", sb->s_id,
3427 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3428 ~EXT4_FEATURE_RO_COMPAT_SUPP));
ac27a0ec
DK		 3429 err = -EROFS;
3430 goto restore_opts;
3431 }
ead6596b 3432
8a266467
TT		 3433 /*
3434 * Make sure the group descriptor checksums
3435 * are sane. If they aren't, refuse to
3436 * remount r/w.
3437 */
3438 for (g = 0; g < sbi->s_groups_count; g++) {
3439 struct ext4_group_desc *gdp =
3440 ext4_get_group_desc(sb, g, NULL);
3441
3442 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3443 printk(KERN_ERR
3444 "EXT4-fs: ext4_remount: "
a9df9a49 3445 "Checksum for group %u failed (%u!=%u)\n",
8a266467
TT		 3446 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3447 le16_to_cpu(gdp->bg_checksum));
3448 err = -EINVAL;
3449 goto restore_opts;
3450 }
3451 }
3452
ead6596b
ES		 3453 /*
3454 * If we have an unprocessed orphan list hanging
3455 * around from a previously readonly bdev mount,
3456 * require a full umount/remount for now.
3457 */
3458 if (es->s_last_orphan) {
3459 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3460 "remount RDWR because of unprocessed "
3461 "orphan inode list. Please "
3462 "umount/remount instead.\n",
3463 sb->s_id);
3464 err = -EINVAL;
3465 goto restore_opts;
3466 }
3467
ac27a0ec
DK		 3468 /*
3469 * Mounting a RDONLY partition read-write, so reread
3470 * and store the current valid flag. (It may have
3471 * been changed by e2fsck since we originally mounted
3472 * the partition.)
3473 */
0390131b
FM		 3474 if (sbi->s_journal)
3475 ext4_clear_journal_err(sb, es);
ac27a0ec 3476 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3477 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3478 goto restore_opts;
2b2d6d01 3479 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3480 sb->s_flags &= ~MS_RDONLY;
3481 }
3482 }
0390131b
FM		 3483 if (sbi->s_journal == NULL)
3484 ext4_commit_super(sb, es, 1);
3485
ac27a0ec
DK		 3486#ifdef CONFIG_QUOTA
3487 /* Release old quota file names */
3488 for (i = 0; i < MAXQUOTAS; i++)
3489 if (old_opts.s_qf_names[i] &&
3490 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3491 kfree(old_opts.s_qf_names[i]);
3492#endif
3493 return 0;
3494restore_opts:
3495 sb->s_flags = old_sb_flags;
3496 sbi->s_mount_opt = old_opts.s_mount_opt;
3497 sbi->s_resuid = old_opts.s_resuid;
3498 sbi->s_resgid = old_opts.s_resgid;
3499 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 3500 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3501 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 3502#ifdef CONFIG_QUOTA
3503 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3504 for (i = 0; i < MAXQUOTAS; i++) {
3505 if (sbi->s_qf_names[i] &&
3506 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3507 kfree(sbi->s_qf_names[i]);
3508 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3509 }
3510#endif
3511 return err;
3512}
3513
2b2d6d01 3514static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3515{
3516 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3517 struct ext4_sb_info *sbi = EXT4_SB(sb);
3518 struct ext4_super_block *es = sbi->s_es;
960cc398 3519 u64 fsid;
ac27a0ec 3520
5e70030d
BP		 3521 if (test_opt(sb, MINIX_DF)) {
3522 sbi->s_overhead_last = 0;
6bc9feff 3523 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 3524 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 3525 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3526 smp_rmb();
3527
3528 /*
5e70030d
BP		 3529 * Compute the overhead (FS structures). This is constant
3530 * for a given filesystem unless the number of block groups
3531 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3532 */
3533
3534 /*
3535 * All of the blocks before first_data_block are
3536 * overhead
3537 */
3538 overhead = le32_to_cpu(es->s_first_data_block);
3539
3540 /*
3541 * Add the overhead attributed to the superblock and
3542 * block group descriptors. If the sparse superblocks
3543 * feature is turned on, then not all groups have this.
3544 */
3545 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3546 overhead += ext4_bg_has_super(sb, i) +
3547 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3548 cond_resched();
3549 }
3550
3551 /*
3552 * Every block group has an inode bitmap, a block
3553 * bitmap, and an inode table.
3554 */
5e70030d
BP		 3555 overhead += ngroups * (2 + sbi->s_itb_per_group);
3556 sbi->s_overhead_last = overhead;
3557 smp_wmb();
6bc9feff 3558 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3559 }
3560
617ba13b 3561 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3562 buf->f_bsize = sb->s_blocksize;
5e70030d 3563 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3564 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3565 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
308ba3ec 3566 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3567 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3568 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3569 buf->f_bavail = 0;
3570 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3571 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3572 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3573 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3574 fsid = le64_to_cpup((void *)es->s_uuid) ^
3575 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3576 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3577 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 3578 return 0;
3579}
3580
3581/* Helper function for writing quotas on sync - we need to start transaction before quota file
3582 * is locked for write. Otherwise the are possible deadlocks:
3583 * Process 1 Process 2
617ba13b 3584 * ext4_create() quota_sync()
a269eb18
JK		 3585 * jbd2_journal_start() write_dquot()
3586 * vfs_dq_init() down(dqio_mutex)
dab291af 3587 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3588 *
3589 */
3590
3591#ifdef CONFIG_QUOTA
3592
3593static inline struct inode *dquot_to_inode(struct dquot *dquot)
3594{
3595 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3596}
3597
617ba13b 3598static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3599{
3600 int ret, err;
3601 handle_t *handle;
3602 struct inode *inode;
3603
3604 inode = dquot_to_inode(dquot);
617ba13b
MC		 3605 handle = ext4_journal_start(inode,
3606 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3607 if (IS_ERR(handle))
3608 return PTR_ERR(handle);
3609 ret = dquot_commit(dquot);
617ba13b 3610 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3611 if (!ret)
3612 ret = err;
3613 return ret;
3614}
3615
617ba13b 3616static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3617{
3618 int ret, err;
3619 handle_t *handle;
3620
617ba13b
MC		 3621 handle = ext4_journal_start(dquot_to_inode(dquot),
3622 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3623 if (IS_ERR(handle))
3624 return PTR_ERR(handle);
3625 ret = dquot_acquire(dquot);
617ba13b 3626 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3627 if (!ret)
3628 ret = err;
3629 return ret;
3630}
3631
617ba13b 3632static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3633{
3634 int ret, err;
3635 handle_t *handle;
3636
617ba13b
MC		 3637 handle = ext4_journal_start(dquot_to_inode(dquot),
3638 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3639 if (IS_ERR(handle)) {
3640 /* Release dquot anyway to avoid endless cycle in dqput() */
3641 dquot_release(dquot);
ac27a0ec 3642 return PTR_ERR(handle);
9c3013e9 3643 }
ac27a0ec 3644 ret = dquot_release(dquot);
617ba13b 3645 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3646 if (!ret)
3647 ret = err;
3648 return ret;
3649}
3650
617ba13b 3651static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3652{
2c8be6b2 3653 /* Are we journaling quotas? */
617ba13b
MC		 3654 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3655 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3656 dquot_mark_dquot_dirty(dquot);
617ba13b 3657 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3658 } else {
3659 return dquot_mark_dquot_dirty(dquot);
3660 }
3661}
3662
617ba13b 3663static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3664{
3665 int ret, err;
3666 handle_t *handle;
3667
3668 /* Data block + inode block */
617ba13b 3669 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3670 if (IS_ERR(handle))
3671 return PTR_ERR(handle);
3672 ret = dquot_commit_info(sb, type);
617ba13b 3673 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3674 if (!ret)
3675 ret = err;
3676 return ret;
3677}
3678
3679/*
3680 * Turn on quotas during mount time - we need to find
3681 * the quota file and such...
3682 */
617ba13b 3683static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3684{
617ba13b
MC		 3685 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3686 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3687}
3688
3689/*
3690 * Standard function to be called on quota_on
3691 */
617ba13b 3692static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8264613d 3693 char *name, int remount)
ac27a0ec
DK		 3694{
3695 int err;
8264613d 3696 struct path path;
ac27a0ec
DK		 3697
3698 if (!test_opt(sb, QUOTA))
3699 return -EINVAL;
8264613d 3700 /* When remounting, no checks are needed and in fact, name is NULL */
0623543b 3701 if (remount)
8264613d 3702 return vfs_quota_on(sb, type, format_id, name, remount);
0623543b 3703
8264613d 3704 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 3705 if (err)
3706 return err;
0623543b 3707
ac27a0ec 3708 /* Quotafile not on the same filesystem? */
8264613d
AV		 3709 if (path.mnt->mnt_sb != sb) {
3710 path_put(&path);
ac27a0ec
DK		 3711 return -EXDEV;
3712 }
0623543b
JK		 3713 /* Journaling quota? */
3714 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3715 /* Quotafile not in fs root? */
8264613d 3716 if (path.dentry->d_parent != sb->s_root)
0623543b
JK		 3717 printk(KERN_WARNING
3718 "EXT4-fs: Quota file not on filesystem root. "
3719 "Journaled quota will not work.\n");
2b2d6d01 3720 }
0623543b
JK		 3721
3722 /*
3723 * When we journal data on quota file, we have to flush journal to see
3724 * all updates to the file when we bypass pagecache...
3725 */
0390131b
FM		 3726 if (EXT4_SB(sb)->s_journal &&
3727 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 3728 /*
3729 * We don't need to lock updates but journal_flush() could
3730 * otherwise be livelocked...
3731 */
3732 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3733 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3734 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3735 if (err) {
8264613d 3736 path_put(&path);
7ffe1ea8
HK		 3737 return err;
3738 }
0623543b
JK		 3739 }
3740
8264613d
AV		 3741 err = vfs_quota_on_path(sb, type, format_id, &path);
3742 path_put(&path);
77e69dac 3743 return err;
ac27a0ec
DK		 3744}
3745
3746/* Read data from quotafile - avoid pagecache and such because we cannot afford
3747 * acquiring the locks... As quota files are never truncated and quota code
3748 * itself serializes the operations (and noone else should touch the files)
3749 * we don't have to be afraid of races */
617ba13b 3750static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3751 size_t len, loff_t off)
3752{
3753 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3754 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3755 int err = 0;
3756 int offset = off & (sb->s_blocksize - 1);
3757 int tocopy;
3758 size_t toread;
3759 struct buffer_head *bh;
3760 loff_t i_size = i_size_read(inode);
3761
3762 if (off > i_size)
3763 return 0;
3764 if (off+len > i_size)
3765 len = i_size-off;
3766 toread = len;
3767 while (toread > 0) {
3768 tocopy = sb->s_blocksize - offset < toread ?
3769 sb->s_blocksize - offset : toread;
617ba13b 3770 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3771 if (err)
3772 return err;
3773 if (!bh) /* A hole? */
3774 memset(data, 0, tocopy);
3775 else
3776 memcpy(data, bh->b_data+offset, tocopy);
3777 brelse(bh);
3778 offset = 0;
3779 toread -= tocopy;
3780 data += tocopy;
3781 blk++;
3782 }
3783 return len;
3784}
3785
3786/* Write to quotafile (we know the transaction is already started and has
3787 * enough credits) */
617ba13b 3788static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3789 const char *data, size_t len, loff_t off)
3790{
3791 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3792 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3793 int err = 0;
3794 int offset = off & (sb->s_blocksize - 1);
3795 int tocopy;
617ba13b 3796 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3797 size_t towrite = len;
3798 struct buffer_head *bh;
3799 handle_t *handle = journal_current_handle();
3800
0390131b 3801 if (EXT4_SB(sb)->s_journal && !handle) {
e5f8eab8 3802 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
9c3013e9
JK		 3803 " cancelled because transaction is not started.\n",
3804 (unsigned long long)off, (unsigned long long)len);
3805 return -EIO;
3806 }
ac27a0ec
DK		 3807 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3808 while (towrite > 0) {
3809 tocopy = sb->s_blocksize - offset < towrite ?
3810 sb->s_blocksize - offset : towrite;
617ba13b 3811 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3812 if (!bh)
3813 goto out;
3814 if (journal_quota) {
617ba13b 3815 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3816 if (err) {
3817 brelse(bh);
3818 goto out;
3819 }
3820 }
3821 lock_buffer(bh);
3822 memcpy(bh->b_data+offset, data, tocopy);
3823 flush_dcache_page(bh->b_page);
3824 unlock_buffer(bh);
3825 if (journal_quota)
0390131b 3826 err = ext4_handle_dirty_metadata(handle, NULL, bh);
ac27a0ec
DK		 3827 else {
3828 /* Always do at least ordered writes for quotas */
678aaf48 3829 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3830 mark_buffer_dirty(bh);
3831 }
3832 brelse(bh);
3833 if (err)
3834 goto out;
3835 offset = 0;
3836 towrite -= tocopy;
3837 data += tocopy;
3838 blk++;
3839 }
3840out:
4d04e4fb
JK		 3841 if (len == towrite) {
3842 mutex_unlock(&inode->i_mutex);
ac27a0ec 3843 return err;
4d04e4fb 3844 }
ac27a0ec
DK		 3845 if (inode->i_size < off+len-towrite) {
3846 i_size_write(inode, off+len-towrite);
617ba13b 3847 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3848 }
ac27a0ec 3849 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3850 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3851 mutex_unlock(&inode->i_mutex);
3852 return len - towrite;
3853}
3854
3855#endif
3856
617ba13b 3857static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3858 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3859{
617ba13b 3860 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3861}
3862
03010a33
TT		 3863static struct file_system_type ext4_fs_type = {
3864 .owner = THIS_MODULE,
3865 .name = "ext4",
3866 .get_sb = ext4_get_sb,
3867 .kill_sb = kill_block_super,
3868 .fs_flags = FS_REQUIRES_DEV,
3869};
3870
3871#ifdef CONFIG_EXT4DEV_COMPAT
3872static int ext4dev_get_sb(struct file_system_type *fs_type,
3873 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3874{
3875 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3876 "to mount using ext4\n");
3877 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3878 "will go away by 2.6.31\n");
3879 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3880}
3881
617ba13b 3882static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3883 .owner = THIS_MODULE,
617ba13b 3884 .name = "ext4dev",
03010a33 3885 .get_sb = ext4dev_get_sb,
ac27a0ec
DK		 3886 .kill_sb = kill_block_super,
3887 .fs_flags = FS_REQUIRES_DEV,
3888};
03010a33
TT		 3889MODULE_ALIAS("ext4dev");
3890#endif
ac27a0ec 3891
617ba13b 3892static int __init init_ext4_fs(void)
ac27a0ec 3893{
c9de560d
AT		 3894 int err;
3895
3197ebdb
TT		 3896 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3897 if (!ext4_kset)
3898 return -ENOMEM;
9f6200bb 3899 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 3900 err = init_ext4_mballoc();
ac27a0ec
DK		 3901 if (err)
3902 return err;
c9de560d
AT		 3903
3904 err = init_ext4_xattr();
3905 if (err)
3906 goto out2;
ac27a0ec
DK		 3907 err = init_inodecache();
3908 if (err)
3909 goto out1;
03010a33 3910 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 3911 if (err)
3912 goto out;
03010a33
TT		 3913#ifdef CONFIG_EXT4DEV_COMPAT
3914 err = register_filesystem(&ext4dev_fs_type);
3915 if (err) {
3916 unregister_filesystem(&ext4_fs_type);
3917 goto out;
3918 }
3919#endif
ac27a0ec
DK		 3920 return 0;
3921out:
3922 destroy_inodecache();
3923out1:
617ba13b 3924 exit_ext4_xattr();
c9de560d
AT		 3925out2:
3926 exit_ext4_mballoc();
ac27a0ec
DK		 3927 return err;
3928}
3929
617ba13b 3930static void __exit exit_ext4_fs(void)
ac27a0ec 3931{
03010a33
TT		 3932 unregister_filesystem(&ext4_fs_type);
3933#ifdef CONFIG_EXT4DEV_COMPAT
617ba13b 3934 unregister_filesystem(&ext4dev_fs_type);
03010a33 3935#endif
ac27a0ec 3936 destroy_inodecache();
617ba13b 3937 exit_ext4_xattr();
c9de560d 3938 exit_ext4_mballoc();
9f6200bb 3939 remove_proc_entry("fs/ext4", NULL);
3197ebdb 3940 kset_unregister(ext4_kset);
ac27a0ec
DK		 3941}
3942
3943MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 3944MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 3945MODULE_LICENSE("GPL");
617ba13b
MC		 3946module_init(init_ext4_fs)
3947module_exit(exit_ext4_fs)
Ubuntu Artful kernel mirror