]> Git Repo - linux.git/blame - fs/ext4/super.c
projects / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: Replace lock/unlock_super() with an explicit lock for the orphan list
[linux.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 ([email protected])
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 ([email protected]), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
29#include <linux/smp_lock.h>
30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
9f6200bb 38#include <linux/proc_fs.h>
3197ebdb 39#include <linux/ctype.h>
ede86cc4 40#include <linux/marker.h>
1330593e 41#include <linux/log2.h>
717d50e4 42#include <linux/crc16.h>
ac27a0ec
DK		 43#include <asm/uaccess.h>
44
3dcf5451
CH		 45#include "ext4.h"
46#include "ext4_jbd2.h"
ac27a0ec
DK		 47#include "xattr.h"
48#include "acl.h"
49#include "namei.h"
717d50e4 50#include "group.h"
ac27a0ec 51
9f6200bb 52struct proc_dir_entry *ext4_proc_root;
3197ebdb 53static struct kset *ext4_kset;
9f6200bb 54
617ba13b 55static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 56 unsigned long journal_devnum);
e2d67052 57static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 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 }
e2d67052 308 ext4_commit_super(sb, 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);
e2d67052 450 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 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);
e2d67052 579 ext4_commit_super(sb, 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);
c5ca7c76
TT		 589 if (is_vmalloc_addr(sbi->s_flex_groups))
590 vfree(sbi->s_flex_groups);
591 else
592 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 593 percpu_counter_destroy(&sbi->s_freeblocks_counter);
594 percpu_counter_destroy(&sbi->s_freeinodes_counter);
595 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 596 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 597 brelse(sbi->s_sbh);
598#ifdef CONFIG_QUOTA
599 for (i = 0; i < MAXQUOTAS; i++)
600 kfree(sbi->s_qf_names[i]);
601#endif
602
603 /* Debugging code just in case the in-memory inode orphan list
604 * isn't empty. The on-disk one can be non-empty if we've
605 * detected an error and taken the fs readonly, but the
606 * in-memory list had better be clean by this point. */
607 if (!list_empty(&sbi->s_orphan))
608 dump_orphan_list(sb, sbi);
609 J_ASSERT(list_empty(&sbi->s_orphan));
610
f98393a6 611 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 612 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
613 /*
614 * Invalidate the journal device's buffers. We don't want them
615 * floating about in memory - the physical journal device may
616 * hotswapped, and it breaks the `ro-after' testing code.
617 */
618 sync_blockdev(sbi->journal_bdev);
f98393a6 619 invalidate_bdev(sbi->journal_bdev);
617ba13b 620 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 621 }
622 sb->s_fs_info = NULL;
3197ebdb
TT		 623 /*
624 * Now that we are completely done shutting down the
625 * superblock, we need to actually destroy the kobject.
626 */
627 unlock_kernel();
628 unlock_super(sb);
629 kobject_put(&sbi->s_kobj);
630 wait_for_completion(&sbi->s_kobj_unregister);
631 lock_super(sb);
632 lock_kernel();
705895b6 633 kfree(sbi->s_blockgroup_lock);
ac27a0ec
DK		 634 kfree(sbi);
635 return;
636}
637
e18b890b 638static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 639
640/*
641 * Called inside transaction, so use GFP_NOFS
642 */
617ba13b 643static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 644{
617ba13b 645 struct ext4_inode_info *ei;
ac27a0ec 646
e6b4f8da 647 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 648 if (!ei)
649 return NULL;
03010a33 650#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b
MC		 651 ei->i_acl = EXT4_ACL_NOT_CACHED;
652 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec 653#endif
ac27a0ec 654 ei->vfs_inode.i_version = 1;
91246c00 655 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 656 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 657 INIT_LIST_HEAD(&ei->i_prealloc_list);
658 spin_lock_init(&ei->i_prealloc_lock);
0390131b
FM		 659 /*
660 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
661 * therefore it can be null here. Don't check it, just initialize
662 * jinode.
663 */
678aaf48 664 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 665 ei->i_reserved_data_blocks = 0;
666 ei->i_reserved_meta_blocks = 0;
667 ei->i_allocated_meta_blocks = 0;
668 ei->i_delalloc_reserved_flag = 0;
669 spin_lock_init(&(ei->i_block_reservation_lock));
ac27a0ec
DK		 670 return &ei->vfs_inode;
671}
672
617ba13b 673static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 674{
9f7dd93d
VA		 675 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
676 printk("EXT4 Inode %p: orphan list check failed!\n",
677 EXT4_I(inode));
678 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
679 EXT4_I(inode), sizeof(struct ext4_inode_info),
680 true);
681 dump_stack();
682 }
617ba13b 683 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 684}
685
51cc5068 686static void init_once(void *foo)
ac27a0ec 687{
617ba13b 688 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 689
a35afb83 690 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 691#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 692 init_rwsem(&ei->xattr_sem);
ac27a0ec 693#endif
0e855ac8 694 init_rwsem(&ei->i_data_sem);
a35afb83 695 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 696}
697
698static int init_inodecache(void)
699{
617ba13b
MC		 700 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
701 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 702 0, (SLAB_RECLAIM_ACCOUNT|
703 SLAB_MEM_SPREAD),
20c2df83 704 init_once);
617ba13b 705 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 706 return -ENOMEM;
707 return 0;
708}
709
710static void destroy_inodecache(void)
711{
617ba13b 712 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 713}
714
617ba13b 715static void ext4_clear_inode(struct inode *inode)
ac27a0ec 716{
03010a33 717#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b
MC		 718 if (EXT4_I(inode)->i_acl &&
719 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
720 posix_acl_release(EXT4_I(inode)->i_acl);
721 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
722 }
723 if (EXT4_I(inode)->i_default_acl &&
724 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
725 posix_acl_release(EXT4_I(inode)->i_default_acl);
726 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 727 }
728#endif
c2ea3fde 729 ext4_discard_preallocations(inode);
0390131b
FM		 730 if (EXT4_JOURNAL(inode))
731 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
678aaf48 732 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 733}
734
2b2d6d01
TT		 735static inline void ext4_show_quota_options(struct seq_file *seq,
736 struct super_block *sb)
ac27a0ec
DK		 737{
738#if defined(CONFIG_QUOTA)
617ba13b 739 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 740
741 if (sbi->s_jquota_fmt)
742 seq_printf(seq, ",jqfmt=%s",
af5bc92d 743 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
ac27a0ec
DK		 744
745 if (sbi->s_qf_names[USRQUOTA])
746 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
747
748 if (sbi->s_qf_names[GRPQUOTA])
749 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
750
617ba13b 751 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 752 seq_puts(seq, ",usrquota");
753
617ba13b 754 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 755 seq_puts(seq, ",grpquota");
756#endif
757}
758
d9c9bef1
MS		 759/*
760 * Show an option if
761 * - it's set to a non-default value OR
762 * - if the per-sb default is different from the global default
763 */
617ba13b 764static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 765{
aa22df2c
AK		 766 int def_errors;
767 unsigned long def_mount_opts;
ac27a0ec 768 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 769 struct ext4_sb_info *sbi = EXT4_SB(sb);
770 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 771
772 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 773 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 774
775 if (sbi->s_sb_block != 1)
776 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
777 if (test_opt(sb, MINIX_DF))
778 seq_puts(seq, ",minixdf");
aa22df2c 779 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 780 seq_puts(seq, ",grpid");
781 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
782 seq_puts(seq, ",nogrpid");
783 if (sbi->s_resuid != EXT4_DEF_RESUID ||
784 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
785 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
786 }
787 if (sbi->s_resgid != EXT4_DEF_RESGID ||
788 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
789 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
790 }
bb4f397a 791 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 792 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 793 def_errors == EXT4_ERRORS_CONTINUE) {
794 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 795 }
796 }
aa22df2c 797 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 798 seq_puts(seq, ",errors=continue");
aa22df2c 799 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 800 seq_puts(seq, ",errors=panic");
aa22df2c 801 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 802 seq_puts(seq, ",nouid32");
aa22df2c 803 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 804 seq_puts(seq, ",debug");
805 if (test_opt(sb, OLDALLOC))
806 seq_puts(seq, ",oldalloc");
03010a33 807#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 808 if (test_opt(sb, XATTR_USER) &&
809 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 810 seq_puts(seq, ",user_xattr");
811 if (!test_opt(sb, XATTR_USER) &&
812 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
813 seq_puts(seq, ",nouser_xattr");
814 }
815#endif
03010a33 816#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 817 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 818 seq_puts(seq, ",acl");
819 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
820 seq_puts(seq, ",noacl");
821#endif
30773840 822 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 823 seq_printf(seq, ",commit=%u",
824 (unsigned) (sbi->s_commit_interval / HZ));
825 }
30773840
TT		 826 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
827 seq_printf(seq, ",min_batch_time=%u",
828 (unsigned) sbi->s_min_batch_time);
829 }
830 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
831 seq_printf(seq, ",max_batch_time=%u",
832 (unsigned) sbi->s_min_batch_time);
833 }
834
571640ca
ES		 835 /*
836 * We're changing the default of barrier mount option, so
837 * let's always display its mount state so it's clear what its
838 * status is.
839 */
840 seq_puts(seq, ",barrier=");
841 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 842 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
843 seq_puts(seq, ",journal_async_commit");
d9c9bef1
MS		 844 if (test_opt(sb, NOBH))
845 seq_puts(seq, ",nobh");
25ec56b5
JNC		 846 if (test_opt(sb, I_VERSION))
847 seq_puts(seq, ",i_version");
dd919b98
AK		 848 if (!test_opt(sb, DELALLOC))
849 seq_puts(seq, ",nodelalloc");
850
ac27a0ec 851
cb45bbe4
MS		 852 if (sbi->s_stripe)
853 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 854 /*
855 * journal mode get enabled in different ways
856 * So just print the value even if we didn't specify it
857 */
617ba13b 858 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 859 seq_puts(seq, ",data=journal");
617ba13b 860 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 861 seq_puts(seq, ",data=ordered");
617ba13b 862 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 863 seq_puts(seq, ",data=writeback");
864
240799cd
TT		 865 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
866 seq_printf(seq, ",inode_readahead_blks=%u",
867 sbi->s_inode_readahead_blks);
868
5bf5683a
HK		 869 if (test_opt(sb, DATA_ERR_ABORT))
870 seq_puts(seq, ",data_err=abort");
871
afd4672d 872 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 873 seq_puts(seq, ",noauto_da_alloc");
afd4672d 874
617ba13b 875 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 876 return 0;
877}
878
879
1b961ac0
CH		 880static struct inode *ext4_nfs_get_inode(struct super_block *sb,
881 u64 ino, u32 generation)
ac27a0ec 882{
ac27a0ec 883 struct inode *inode;
ac27a0ec 884
617ba13b 885 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 886 return ERR_PTR(-ESTALE);
617ba13b 887 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 888 return ERR_PTR(-ESTALE);
889
890 /* iget isn't really right if the inode is currently unallocated!!
891 *
617ba13b 892 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 893 * deleted, so we should be safe.
894 *
895 * Currently we don't know the generation for parent directory, so
896 * a generation of 0 means "accept any"
897 */
1d1fe1ee
DH		 898 inode = ext4_iget(sb, ino);
899 if (IS_ERR(inode))
900 return ERR_CAST(inode);
901 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 902 iput(inode);
903 return ERR_PTR(-ESTALE);
904 }
1b961ac0
CH		 905
906 return inode;
907}
908
909static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
910 int fh_len, int fh_type)
911{
912 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
913 ext4_nfs_get_inode);
914}
915
916static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
917 int fh_len, int fh_type)
918{
919 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
920 ext4_nfs_get_inode);
ac27a0ec
DK		 921}
922
c39a7f84
TO		 923/*
924 * Try to release metadata pages (indirect blocks, directories) which are
925 * mapped via the block device. Since these pages could have journal heads
926 * which would prevent try_to_free_buffers() from freeing them, we must use
927 * jbd2 layer's try_to_free_buffers() function to release them.
928 */
929static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
930{
931 journal_t *journal = EXT4_SB(sb)->s_journal;
932
933 WARN_ON(PageChecked(page));
934 if (!page_has_buffers(page))
935 return 0;
936 if (journal)
937 return jbd2_journal_try_to_free_buffers(journal, page,
938 wait & ~__GFP_WAIT);
939 return try_to_free_buffers(page);
940}
941
ac27a0ec 942#ifdef CONFIG_QUOTA
af5bc92d 943#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 944#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 945
617ba13b
MC		 946static int ext4_write_dquot(struct dquot *dquot);
947static int ext4_acquire_dquot(struct dquot *dquot);
948static int ext4_release_dquot(struct dquot *dquot);
949static int ext4_mark_dquot_dirty(struct dquot *dquot);
950static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 951static int ext4_quota_on(struct super_block *sb, int type, int format_id,
952 char *path, int remount);
617ba13b
MC		 953static int ext4_quota_on_mount(struct super_block *sb, int type);
954static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 955 size_t len, loff_t off);
617ba13b 956static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 957 const char *data, size_t len, loff_t off);
958
617ba13b 959static struct dquot_operations ext4_quota_operations = {
edf72453
JK		 960 .initialize = dquot_initialize,
961 .drop = dquot_drop,
ac27a0ec 962 .alloc_space = dquot_alloc_space,
60e58e0f
MC		 963 .reserve_space = dquot_reserve_space,
964 .claim_space = dquot_claim_space,
965 .release_rsv = dquot_release_reserved_space,
966 .get_reserved_space = ext4_get_reserved_space,
ac27a0ec
DK		 967 .alloc_inode = dquot_alloc_inode,
968 .free_space = dquot_free_space,
969 .free_inode = dquot_free_inode,
970 .transfer = dquot_transfer,
617ba13b
MC		 971 .write_dquot = ext4_write_dquot,
972 .acquire_dquot = ext4_acquire_dquot,
973 .release_dquot = ext4_release_dquot,
974 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 975 .write_info = ext4_write_info,
976 .alloc_dquot = dquot_alloc,
977 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 978};
979
617ba13b
MC		 980static struct quotactl_ops ext4_qctl_operations = {
981 .quota_on = ext4_quota_on,
ac27a0ec
DK		 982 .quota_off = vfs_quota_off,
983 .quota_sync = vfs_quota_sync,
984 .get_info = vfs_get_dqinfo,
985 .set_info = vfs_set_dqinfo,
986 .get_dqblk = vfs_get_dqblk,
987 .set_dqblk = vfs_set_dqblk
988};
989#endif
990
ee9b6d61 991static const struct super_operations ext4_sops = {
617ba13b
MC		 992 .alloc_inode = ext4_alloc_inode,
993 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 994 .write_inode = ext4_write_inode,
995 .dirty_inode = ext4_dirty_inode,
996 .delete_inode = ext4_delete_inode,
997 .put_super = ext4_put_super,
617ba13b 998 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 999 .freeze_fs = ext4_freeze,
1000 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1001 .statfs = ext4_statfs,
1002 .remount_fs = ext4_remount,
1003 .clear_inode = ext4_clear_inode,
1004 .show_options = ext4_show_options,
ac27a0ec 1005#ifdef CONFIG_QUOTA
617ba13b
MC		 1006 .quota_read = ext4_quota_read,
1007 .quota_write = ext4_quota_write,
ac27a0ec 1008#endif
c39a7f84 1009 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1010};
1011
9ca92389
TT		 1012static const struct super_operations ext4_nojournal_sops = {
1013 .alloc_inode = ext4_alloc_inode,
1014 .destroy_inode = ext4_destroy_inode,
1015 .write_inode = ext4_write_inode,
1016 .dirty_inode = ext4_dirty_inode,
1017 .delete_inode = ext4_delete_inode,
1018 .write_super = ext4_write_super,
1019 .put_super = ext4_put_super,
1020 .statfs = ext4_statfs,
1021 .remount_fs = ext4_remount,
1022 .clear_inode = ext4_clear_inode,
1023 .show_options = ext4_show_options,
1024#ifdef CONFIG_QUOTA
1025 .quota_read = ext4_quota_read,
1026 .quota_write = ext4_quota_write,
1027#endif
1028 .bdev_try_to_free_page = bdev_try_to_free_page,
1029};
1030
39655164 1031static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1032 .fh_to_dentry = ext4_fh_to_dentry,
1033 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1034 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1035};
1036
1037enum {
1038 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1039 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1040 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1041 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1042 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1043 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1044 Opt_journal_update, Opt_journal_dev,
818d276c 1045 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1046 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
5bf5683a 1047 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec
DK		 1048 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1049 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
06705bff
TT		 1050 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1051 Opt_usrquota, Opt_grpquota, Opt_i_version,
01436ef2 1052 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
b3881f74 1053 Opt_inode_readahead_blks, Opt_journal_ioprio
ac27a0ec
DK		 1054};
1055
a447c093 1056static const match_table_t tokens = {
ac27a0ec
DK		 1057 {Opt_bsd_df, "bsddf"},
1058 {Opt_minix_df, "minixdf"},
1059 {Opt_grpid, "grpid"},
1060 {Opt_grpid, "bsdgroups"},
1061 {Opt_nogrpid, "nogrpid"},
1062 {Opt_nogrpid, "sysvgroups"},
1063 {Opt_resgid, "resgid=%u"},
1064 {Opt_resuid, "resuid=%u"},
1065 {Opt_sb, "sb=%u"},
1066 {Opt_err_cont, "errors=continue"},
1067 {Opt_err_panic, "errors=panic"},
1068 {Opt_err_ro, "errors=remount-ro"},
1069 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1070 {Opt_debug, "debug"},
1071 {Opt_oldalloc, "oldalloc"},
1072 {Opt_orlov, "orlov"},
1073 {Opt_user_xattr, "user_xattr"},
1074 {Opt_nouser_xattr, "nouser_xattr"},
1075 {Opt_acl, "acl"},
1076 {Opt_noacl, "noacl"},
ac27a0ec
DK		 1077 {Opt_noload, "noload"},
1078 {Opt_nobh, "nobh"},
1079 {Opt_bh, "bh"},
1080 {Opt_commit, "commit=%u"},
30773840
TT		 1081 {Opt_min_batch_time, "min_batch_time=%u"},
1082 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1083 {Opt_journal_update, "journal=update"},
ac27a0ec 1084 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1085 {Opt_journal_checksum, "journal_checksum"},
1086 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1087 {Opt_abort, "abort"},
1088 {Opt_data_journal, "data=journal"},
1089 {Opt_data_ordered, "data=ordered"},
1090 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1091 {Opt_data_err_abort, "data_err=abort"},
1092 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1093 {Opt_offusrjquota, "usrjquota="},
1094 {Opt_usrjquota, "usrjquota=%s"},
1095 {Opt_offgrpjquota, "grpjquota="},
1096 {Opt_grpjquota, "grpjquota=%s"},
1097 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1098 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1099 {Opt_grpquota, "grpquota"},
1100 {Opt_noquota, "noquota"},
1101 {Opt_quota, "quota"},
1102 {Opt_usrquota, "usrquota"},
1103 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1104 {Opt_barrier, "barrier"},
1105 {Opt_nobarrier, "nobarrier"},
25ec56b5 1106 {Opt_i_version, "i_version"},
c9de560d 1107 {Opt_stripe, "stripe=%u"},
ac27a0ec 1108 {Opt_resize, "resize"},
64769240 1109 {Opt_delalloc, "delalloc"},
dd919b98 1110 {Opt_nodelalloc, "nodelalloc"},
240799cd 1111 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1112 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1113 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1114 {Opt_auto_da_alloc, "auto_da_alloc"},
1115 {Opt_noauto_da_alloc, "noauto_da_alloc"},
f3f12faa 1116 {Opt_err, NULL},
ac27a0ec
DK		 1117};
1118
617ba13b 1119static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1120{
617ba13b 1121 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1122 char *options = (char *) *data;
1123
1124 if (!options || strncmp(options, "sb=", 3) != 0)
1125 return 1; /* Default location */
1126 options += 3;
617ba13b 1127 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1128 sb_block = simple_strtoul(options, &options, 0);
1129 if (*options && *options != ',') {
4776004f 1130 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1131 (char *) *data);
1132 return 1;
1133 }
1134 if (*options == ',')
1135 options++;
1136 *data = (void *) options;
1137 return sb_block;
1138}
1139
b3881f74
TT		 1140#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1141
2b2d6d01 1142static int parse_options(char *options, struct super_block *sb,
c3191067 1143 unsigned long *journal_devnum,
b3881f74 1144 unsigned int *journal_ioprio,
2b2d6d01 1145 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1146{
617ba13b 1147 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1148 char *p;
ac27a0ec
DK		 1149 substring_t args[MAX_OPT_ARGS];
1150 int data_opt = 0;
1151 int option;
1152#ifdef CONFIG_QUOTA
dfc5d03f 1153 int qtype, qfmt;
ac27a0ec
DK		 1154 char *qname;
1155#endif
1156
1157 if (!options)
1158 return 1;
1159
2b2d6d01 1160 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1161 int token;
1162 if (!*p)
1163 continue;
1164
1165 token = match_token(p, tokens, args);
1166 switch (token) {
1167 case Opt_bsd_df:
2b2d6d01 1168 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1169 break;
1170 case Opt_minix_df:
2b2d6d01 1171 set_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1172 break;
1173 case Opt_grpid:
2b2d6d01 1174 set_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1175 break;
1176 case Opt_nogrpid:
2b2d6d01 1177 clear_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1178 break;
1179 case Opt_resuid:
1180 if (match_int(&args[0], &option))
1181 return 0;
1182 sbi->s_resuid = option;
1183 break;
1184 case Opt_resgid:
1185 if (match_int(&args[0], &option))
1186 return 0;
1187 sbi->s_resgid = option;
1188 break;
1189 case Opt_sb:
1190 /* handled by get_sb_block() instead of here */
1191 /* *sb_block = match_int(&args[0]); */
1192 break;
1193 case Opt_err_panic:
2b2d6d01
TT		 1194 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1195 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1196 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1197 break;
1198 case Opt_err_ro:
2b2d6d01
TT		 1199 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1200 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1201 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1202 break;
1203 case Opt_err_cont:
2b2d6d01
TT		 1204 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1205 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1206 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1207 break;
1208 case Opt_nouid32:
2b2d6d01 1209 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1210 break;
ac27a0ec 1211 case Opt_debug:
2b2d6d01 1212 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1213 break;
1214 case Opt_oldalloc:
2b2d6d01 1215 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1216 break;
1217 case Opt_orlov:
2b2d6d01 1218 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1219 break;
03010a33 1220#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1221 case Opt_user_xattr:
2b2d6d01 1222 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1223 break;
1224 case Opt_nouser_xattr:
2b2d6d01 1225 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1226 break;
1227#else
1228 case Opt_user_xattr:
1229 case Opt_nouser_xattr:
4776004f
TT		 1230 printk(KERN_ERR "EXT4 (no)user_xattr options "
1231 "not supported\n");
ac27a0ec
DK		 1232 break;
1233#endif
03010a33 1234#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1235 case Opt_acl:
1236 set_opt(sbi->s_mount_opt, POSIX_ACL);
1237 break;
1238 case Opt_noacl:
1239 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1240 break;
1241#else
1242 case Opt_acl:
1243 case Opt_noacl:
4776004f
TT		 1244 printk(KERN_ERR "EXT4 (no)acl options "
1245 "not supported\n");
ac27a0ec
DK		 1246 break;
1247#endif
ac27a0ec
DK		 1248 case Opt_journal_update:
1249 /* @@@ FIXME */
1250 /* Eventually we will want to be able to create
1251 a journal file here. For now, only allow the
1252 user to specify an existing inode to be the
1253 journal file. */
1254 if (is_remount) {
617ba13b 1255 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1256 "journal on remount\n");
1257 return 0;
1258 }
2b2d6d01 1259 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1260 break;
ac27a0ec
DK		 1261 case Opt_journal_dev:
1262 if (is_remount) {
617ba13b 1263 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1264 "journal on remount\n");
1265 return 0;
1266 }
1267 if (match_int(&args[0], &option))
1268 return 0;
1269 *journal_devnum = option;
1270 break;
818d276c
GS		 1271 case Opt_journal_checksum:
1272 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1273 break;
1274 case Opt_journal_async_commit:
1275 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1276 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1277 break;
ac27a0ec 1278 case Opt_noload:
2b2d6d01 1279 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1280 break;
1281 case Opt_commit:
1282 if (match_int(&args[0], &option))
1283 return 0;
1284 if (option < 0)
1285 return 0;
1286 if (option == 0)
cd02ff0b 1287 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1288 sbi->s_commit_interval = HZ * option;
1289 break;
30773840
TT		 1290 case Opt_max_batch_time:
1291 if (match_int(&args[0], &option))
1292 return 0;
1293 if (option < 0)
1294 return 0;
1295 if (option == 0)
1296 option = EXT4_DEF_MAX_BATCH_TIME;
1297 sbi->s_max_batch_time = option;
1298 break;
1299 case Opt_min_batch_time:
1300 if (match_int(&args[0], &option))
1301 return 0;
1302 if (option < 0)
1303 return 0;
1304 sbi->s_min_batch_time = option;
1305 break;
ac27a0ec 1306 case Opt_data_journal:
617ba13b 1307 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1308 goto datacheck;
1309 case Opt_data_ordered:
617ba13b 1310 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1311 goto datacheck;
1312 case Opt_data_writeback:
617ba13b 1313 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1314 datacheck:
1315 if (is_remount) {
617ba13b 1316 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1317 != data_opt) {
1318 printk(KERN_ERR
617ba13b 1319 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1320 "mode on remount\n");
1321 return 0;
1322 }
1323 } else {
617ba13b 1324 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1325 sbi->s_mount_opt |= data_opt;
1326 }
1327 break;
5bf5683a
HK		 1328 case Opt_data_err_abort:
1329 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1330 break;
1331 case Opt_data_err_ignore:
1332 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1333 break;
ac27a0ec
DK		 1334#ifdef CONFIG_QUOTA
1335 case Opt_usrjquota:
1336 qtype = USRQUOTA;
1337 goto set_qf_name;
1338 case Opt_grpjquota:
1339 qtype = GRPQUOTA;
1340set_qf_name:
17bd13b3 1341 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1342 !sbi->s_qf_names[qtype]) {
ac27a0ec 1343 printk(KERN_ERR
4776004f
TT		 1344 "EXT4-fs: Cannot change journaled "
1345 "quota options when quota turned on.\n");
ac27a0ec
DK		 1346 return 0;
1347 }
1348 qname = match_strdup(&args[0]);
1349 if (!qname) {
1350 printk(KERN_ERR
617ba13b 1351 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1352 "storing quotafile name.\n");
1353 return 0;
1354 }
1355 if (sbi->s_qf_names[qtype] &&
1356 strcmp(sbi->s_qf_names[qtype], qname)) {
1357 printk(KERN_ERR
617ba13b 1358 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1359 "specified.\n", QTYPE2NAME(qtype));
1360 kfree(qname);
1361 return 0;
1362 }
1363 sbi->s_qf_names[qtype] = qname;
1364 if (strchr(sbi->s_qf_names[qtype], '/')) {
1365 printk(KERN_ERR
617ba13b 1366 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1367 "filesystem root.\n");
1368 kfree(sbi->s_qf_names[qtype]);
1369 sbi->s_qf_names[qtype] = NULL;
1370 return 0;
1371 }
1372 set_opt(sbi->s_mount_opt, QUOTA);
1373 break;
1374 case Opt_offusrjquota:
1375 qtype = USRQUOTA;
1376 goto clear_qf_name;
1377 case Opt_offgrpjquota:
1378 qtype = GRPQUOTA;
1379clear_qf_name:
17bd13b3 1380 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1381 sbi->s_qf_names[qtype]) {
617ba13b 1382 printk(KERN_ERR "EXT4-fs: Cannot change "
2c8be6b2 1383 "journaled quota options when "
ac27a0ec
DK		 1384 "quota turned on.\n");
1385 return 0;
1386 }
1387 /*
1388 * The space will be released later when all options
1389 * are confirmed to be correct
1390 */
1391 sbi->s_qf_names[qtype] = NULL;
1392 break;
1393 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1394 qfmt = QFMT_VFS_OLD;
1395 goto set_qf_format;
ac27a0ec 1396 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1397 qfmt = QFMT_VFS_V0;
1398set_qf_format:
17bd13b3 1399 if (sb_any_quota_loaded(sb) &&
dfc5d03f
JK		 1400 sbi->s_jquota_fmt != qfmt) {
1401 printk(KERN_ERR "EXT4-fs: Cannot change "
1402 "journaled quota options when "
1403 "quota turned on.\n");
1404 return 0;
1405 }
1406 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1407 break;
1408 case Opt_quota:
1409 case Opt_usrquota:
1410 set_opt(sbi->s_mount_opt, QUOTA);
1411 set_opt(sbi->s_mount_opt, USRQUOTA);
1412 break;
1413 case Opt_grpquota:
1414 set_opt(sbi->s_mount_opt, QUOTA);
1415 set_opt(sbi->s_mount_opt, GRPQUOTA);
1416 break;
1417 case Opt_noquota:
17bd13b3 1418 if (sb_any_quota_loaded(sb)) {
617ba13b 1419 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1420 "options when quota turned on.\n");
1421 return 0;
1422 }
1423 clear_opt(sbi->s_mount_opt, QUOTA);
1424 clear_opt(sbi->s_mount_opt, USRQUOTA);
1425 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1426 break;
1427#else
1428 case Opt_quota:
1429 case Opt_usrquota:
1430 case Opt_grpquota:
cd59e7b9
JK		 1431 printk(KERN_ERR
1432 "EXT4-fs: quota options not supported.\n");
1433 break;
ac27a0ec
DK		 1434 case Opt_usrjquota:
1435 case Opt_grpjquota:
1436 case Opt_offusrjquota:
1437 case Opt_offgrpjquota:
1438 case Opt_jqfmt_vfsold:
1439 case Opt_jqfmt_vfsv0:
1440 printk(KERN_ERR
cd59e7b9 1441 "EXT4-fs: journaled quota options not "
ac27a0ec
DK		 1442 "supported.\n");
1443 break;
1444 case Opt_noquota:
1445 break;
1446#endif
1447 case Opt_abort:
1448 set_opt(sbi->s_mount_opt, ABORT);
1449 break;
06705bff
TT		 1450 case Opt_nobarrier:
1451 clear_opt(sbi->s_mount_opt, BARRIER);
1452 break;
ac27a0ec 1453 case Opt_barrier:
06705bff
TT		 1454 if (match_int(&args[0], &option)) {
1455 set_opt(sbi->s_mount_opt, BARRIER);
1456 break;
1457 }
ac27a0ec
DK		 1458 if (option)
1459 set_opt(sbi->s_mount_opt, BARRIER);
1460 else
1461 clear_opt(sbi->s_mount_opt, BARRIER);
1462 break;
1463 case Opt_ignore:
1464 break;
1465 case Opt_resize:
1466 if (!is_remount) {
617ba13b 1467 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1468 "for remount\n");
1469 return 0;
1470 }
1471 if (match_int(&args[0], &option) != 0)
1472 return 0;
1473 *n_blocks_count = option;
1474 break;
1475 case Opt_nobh:
1476 set_opt(sbi->s_mount_opt, NOBH);
1477 break;
1478 case Opt_bh:
1479 clear_opt(sbi->s_mount_opt, NOBH);
1480 break;
25ec56b5
JNC		 1481 case Opt_i_version:
1482 set_opt(sbi->s_mount_opt, I_VERSION);
1483 sb->s_flags |= MS_I_VERSION;
1484 break;
dd919b98
AK		 1485 case Opt_nodelalloc:
1486 clear_opt(sbi->s_mount_opt, DELALLOC);
1487 break;
c9de560d
AT		 1488 case Opt_stripe:
1489 if (match_int(&args[0], &option))
1490 return 0;
1491 if (option < 0)
1492 return 0;
1493 sbi->s_stripe = option;
1494 break;
64769240
AT		 1495 case Opt_delalloc:
1496 set_opt(sbi->s_mount_opt, DELALLOC);
1497 break;
240799cd
TT		 1498 case Opt_inode_readahead_blks:
1499 if (match_int(&args[0], &option))
1500 return 0;
1501 if (option < 0 || option > (1 << 30))
1502 return 0;
f7c43950 1503 if (!is_power_of_2(option)) {
3197ebdb
TT		 1504 printk(KERN_ERR "EXT4-fs: inode_readahead_blks"
1505 " must be a power of 2\n");
1506 return 0;
1507 }
240799cd
TT		 1508 sbi->s_inode_readahead_blks = option;
1509 break;
b3881f74
TT		 1510 case Opt_journal_ioprio:
1511 if (match_int(&args[0], &option))
1512 return 0;
1513 if (option < 0 || option > 7)
1514 break;
1515 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1516 option);
1517 break;
06705bff
TT		 1518 case Opt_noauto_da_alloc:
1519 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1520 break;
afd4672d 1521 case Opt_auto_da_alloc:
06705bff
TT		 1522 if (match_int(&args[0], &option)) {
1523 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1524 break;
1525 }
afd4672d
TT		 1526 if (option)
1527 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1528 else
1529 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1530 break;
ac27a0ec 1531 default:
2b2d6d01
TT		 1532 printk(KERN_ERR
1533 "EXT4-fs: Unrecognized mount option \"%s\" "
1534 "or missing value\n", p);
ac27a0ec
DK		 1535 return 0;
1536 }
1537 }
1538#ifdef CONFIG_QUOTA
1539 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1540 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1541 sbi->s_qf_names[USRQUOTA])
1542 clear_opt(sbi->s_mount_opt, USRQUOTA);
1543
617ba13b 1544 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1545 sbi->s_qf_names[GRPQUOTA])
1546 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1547
1548 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1549 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1550 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1551 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1552 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1553 "format mixing.\n");
1554 return 0;
1555 }
1556
1557 if (!sbi->s_jquota_fmt) {
2c8be6b2 1558 printk(KERN_ERR "EXT4-fs: journaled quota format "
ac27a0ec
DK		 1559 "not specified.\n");
1560 return 0;
1561 }
1562 } else {
1563 if (sbi->s_jquota_fmt) {
2c8be6b2
JK		 1564 printk(KERN_ERR "EXT4-fs: journaled quota format "
1565 "specified with no journaling "
ac27a0ec
DK		 1566 "enabled.\n");
1567 return 0;
1568 }
1569 }
1570#endif
1571 return 1;
1572}
1573
617ba13b 1574static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1575 int read_only)
1576{
617ba13b 1577 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1578 int res = 0;
1579
617ba13b 1580 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2b2d6d01
TT		 1581 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1582 "forcing read-only mode\n");
ac27a0ec
DK		 1583 res = MS_RDONLY;
1584 }
1585 if (read_only)
1586 return res;
617ba13b 1587 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2b2d6d01
TT		 1588 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1589 "running e2fsck is recommended\n");
617ba13b 1590 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
2b2d6d01
TT		 1591 printk(KERN_WARNING
1592 "EXT4-fs warning: mounting fs with errors, "
1593 "running e2fsck is recommended\n");
ac27a0ec
DK		 1594 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1595 le16_to_cpu(es->s_mnt_count) >=
1596 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2b2d6d01
TT		 1597 printk(KERN_WARNING
1598 "EXT4-fs warning: maximal mount count reached, "
1599 "running e2fsck is recommended\n");
ac27a0ec
DK		 1600 else if (le32_to_cpu(es->s_checkinterval) &&
1601 (le32_to_cpu(es->s_lastcheck) +
1602 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2b2d6d01
TT		 1603 printk(KERN_WARNING
1604 "EXT4-fs warning: checktime reached, "
1605 "running e2fsck is recommended\n");
0390131b
FM		 1606 if (!sbi->s_journal)
1607 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1608 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1609 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1610 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1611 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1612 ext4_update_dynamic_rev(sb);
0390131b
FM		 1613 if (sbi->s_journal)
1614 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1615
e2d67052 1616 ext4_commit_super(sb, 1);
ac27a0ec 1617 if (test_opt(sb, DEBUG))
a9df9a49 1618 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
ac27a0ec
DK		 1619 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1620 sb->s_blocksize,
1621 sbi->s_groups_count,
617ba13b
MC		 1622 EXT4_BLOCKS_PER_GROUP(sb),
1623 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1624 sbi->s_mount_opt);
1625
0390131b
FM		 1626 if (EXT4_SB(sb)->s_journal) {
1627 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1628 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1629 "external", EXT4_SB(sb)->s_journal->j_devname);
1630 } else {
1631 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1632 }
ac27a0ec
DK		 1633 return res;
1634}
1635
772cb7c8
JS		 1636static int ext4_fill_flex_info(struct super_block *sb)
1637{
1638 struct ext4_sb_info *sbi = EXT4_SB(sb);
1639 struct ext4_group_desc *gdp = NULL;
1640 struct buffer_head *bh;
1641 ext4_group_t flex_group_count;
1642 ext4_group_t flex_group;
1643 int groups_per_flex = 0;
c5ca7c76 1644 size_t size;
772cb7c8
JS		 1645 int i;
1646
1647 if (!sbi->s_es->s_log_groups_per_flex) {
1648 sbi->s_log_groups_per_flex = 0;
1649 return 1;
1650 }
1651
1652 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1653 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1654
c62a11fd
FB		 1655 /* We allocate both existing and potentially added groups */
1656 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1657 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1658 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76
TT		 1659 size = flex_group_count * sizeof(struct flex_groups);
1660 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1661 if (sbi->s_flex_groups == NULL) {
1662 sbi->s_flex_groups = vmalloc(size);
1663 if (sbi->s_flex_groups)
1664 memset(sbi->s_flex_groups, 0, size);
1665 }
772cb7c8 1666 if (sbi->s_flex_groups == NULL) {
ec05e868 1667 printk(KERN_ERR "EXT4-fs: not enough memory for "
a9df9a49 1668 "%u flex groups\n", flex_group_count);
772cb7c8
JS		 1669 goto failed;
1670 }
772cb7c8 1671
772cb7c8
JS		 1672 for (i = 0; i < sbi->s_groups_count; i++) {
1673 gdp = ext4_get_group_desc(sb, i, &bh);
1674
1675 flex_group = ext4_flex_group(sbi, i);
9f24e420
TT		 1676 atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1677 ext4_free_inodes_count(sb, gdp));
1678 atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1679 ext4_free_blks_count(sb, gdp));
7d39db14
TT		 1680 atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1681 ext4_used_dirs_count(sb, gdp));
772cb7c8
JS		 1682 }
1683
1684 return 1;
1685failed:
1686 return 0;
1687}
1688
717d50e4
AD		 1689__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1690 struct ext4_group_desc *gdp)
1691{
1692 __u16 crc = 0;
1693
1694 if (sbi->s_es->s_feature_ro_compat &
1695 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1696 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1697 __le32 le_group = cpu_to_le32(block_group);
1698
1699 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1700 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1701 crc = crc16(crc, (__u8 *)gdp, offset);
1702 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1703 /* for checksum of struct ext4_group_desc do the rest...*/
1704 if ((sbi->s_es->s_feature_incompat &
1705 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1706 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1707 crc = crc16(crc, (__u8 *)gdp + offset,
1708 le16_to_cpu(sbi->s_es->s_desc_size) -
1709 offset);
1710 }
1711
1712 return cpu_to_le16(crc);
1713}
1714
1715int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1716 struct ext4_group_desc *gdp)
1717{
1718 if ((sbi->s_es->s_feature_ro_compat &
1719 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1720 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1721 return 0;
1722
1723 return 1;
1724}
1725
ac27a0ec 1726/* Called at mount-time, super-block is locked */
197cd65a 1727static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1728{
617ba13b
MC		 1729 struct ext4_sb_info *sbi = EXT4_SB(sb);
1730 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1731 ext4_fsblk_t last_block;
bd81d8ee
LV		 1732 ext4_fsblk_t block_bitmap;
1733 ext4_fsblk_t inode_bitmap;
1734 ext4_fsblk_t inode_table;
ce421581 1735 int flexbg_flag = 0;
fd2d4291 1736 ext4_group_t i;
ac27a0ec 1737
ce421581
JS		 1738 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1739 flexbg_flag = 1;
1740
af5bc92d 1741 ext4_debug("Checking group descriptors");
ac27a0ec 1742
197cd65a
AM		 1743 for (i = 0; i < sbi->s_groups_count; i++) {
1744 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1745
ce421581 1746 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1747 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1748 else
1749 last_block = first_block +
617ba13b 1750 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1751
8fadc143 1752 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1753 if (block_bitmap < first_block || block_bitmap > last_block) {
c19204b0 1754 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1755 "Block bitmap for group %u not in group "
5128273a 1756 "(block %llu)!\n", i, block_bitmap);
ac27a0ec
DK		 1757 return 0;
1758 }
8fadc143 1759 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1760 if (inode_bitmap < first_block || inode_bitmap > last_block) {
c19204b0 1761 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1762 "Inode bitmap for group %u not in group "
5128273a 1763 "(block %llu)!\n", i, inode_bitmap);
ac27a0ec
DK		 1764 return 0;
1765 }
8fadc143 1766 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1767 if (inode_table < first_block ||
2b2d6d01 1768 inode_table + sbi->s_itb_per_group - 1 > last_block) {
c19204b0 1769 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1770 "Inode table for group %u not in group "
5128273a 1771 "(block %llu)!\n", i, inode_table);
ac27a0ec
DK		 1772 return 0;
1773 }
b5f10eed 1774 spin_lock(sb_bgl_lock(sbi, i));
717d50e4 1775 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
c19204b0 1776 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1777 "Checksum for group %u failed (%u!=%u)\n",
c19204b0
JB		 1778 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1779 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c
LZ		 1780 if (!(sb->s_flags & MS_RDONLY)) {
1781 spin_unlock(sb_bgl_lock(sbi, i));
8a266467 1782 return 0;
7ee1ec4c 1783 }
717d50e4 1784 }
b5f10eed 1785 spin_unlock(sb_bgl_lock(sbi, i));
ce421581
JS		 1786 if (!flexbg_flag)
1787 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1788 }
1789
bd81d8ee 1790 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2b2d6d01 1791 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1792 return 1;
1793}
1794
617ba13b 1795/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1796 * the superblock) which were deleted from all directories, but held open by
1797 * a process at the time of a crash. We walk the list and try to delete these
1798 * inodes at recovery time (only with a read-write filesystem).
1799 *
1800 * In order to keep the orphan inode chain consistent during traversal (in
1801 * case of crash during recovery), we link each inode into the superblock
1802 * orphan list_head and handle it the same way as an inode deletion during
1803 * normal operation (which journals the operations for us).
1804 *
1805 * We only do an iget() and an iput() on each inode, which is very safe if we
1806 * accidentally point at an in-use or already deleted inode. The worst that
1807 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1808 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1809 * e2fsck was run on this filesystem, and it must have already done the orphan
1810 * inode cleanup for us, so we can safely abort without any further action.
1811 */
2b2d6d01
TT		 1812static void ext4_orphan_cleanup(struct super_block *sb,
1813 struct ext4_super_block *es)
ac27a0ec
DK		 1814{
1815 unsigned int s_flags = sb->s_flags;
1816 int nr_orphans = 0, nr_truncates = 0;
1817#ifdef CONFIG_QUOTA
1818 int i;
1819#endif
1820 if (!es->s_last_orphan) {
1821 jbd_debug(4, "no orphan inodes to clean up\n");
1822 return;
1823 }
1824
a8f48a95
ES		 1825 if (bdev_read_only(sb->s_bdev)) {
1826 printk(KERN_ERR "EXT4-fs: write access "
1827 "unavailable, skipping orphan cleanup.\n");
1828 return;
1829 }
1830
617ba13b 1831 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1832 if (es->s_last_orphan)
1833 jbd_debug(1, "Errors on filesystem, "
1834 "clearing orphan list.\n");
1835 es->s_last_orphan = 0;
1836 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1837 return;
1838 }
1839
1840 if (s_flags & MS_RDONLY) {
617ba13b 1841 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1842 sb->s_id);
1843 sb->s_flags &= ~MS_RDONLY;
1844 }
1845#ifdef CONFIG_QUOTA
1846 /* Needed for iput() to work correctly and not trash data */
1847 sb->s_flags |= MS_ACTIVE;
1848 /* Turn on quotas so that they are updated correctly */
1849 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1850 if (EXT4_SB(sb)->s_qf_names[i]) {
1851 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1852 if (ret < 0)
1853 printk(KERN_ERR
2c8be6b2 1854 "EXT4-fs: Cannot turn on journaled "
ac27a0ec
DK		 1855 "quota: error %d\n", ret);
1856 }
1857 }
1858#endif
1859
1860 while (es->s_last_orphan) {
1861 struct inode *inode;
1862
97bd42b9
JB		 1863 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1864 if (IS_ERR(inode)) {
ac27a0ec
DK		 1865 es->s_last_orphan = 0;
1866 break;
1867 }
1868
617ba13b 1869 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
a269eb18 1870 vfs_dq_init(inode);
ac27a0ec
DK		 1871 if (inode->i_nlink) {
1872 printk(KERN_DEBUG
e5f8eab8 1873 "%s: truncating inode %lu to %lld bytes\n",
46e665e9 1874 __func__, inode->i_ino, inode->i_size);
e5f8eab8 1875 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 1876 inode->i_ino, inode->i_size);
617ba13b 1877 ext4_truncate(inode);
ac27a0ec
DK		 1878 nr_truncates++;
1879 } else {
1880 printk(KERN_DEBUG
1881 "%s: deleting unreferenced inode %lu\n",
46e665e9 1882 __func__, inode->i_ino);
ac27a0ec
DK		 1883 jbd_debug(2, "deleting unreferenced inode %lu\n",
1884 inode->i_ino);
1885 nr_orphans++;
1886 }
1887 iput(inode); /* The delete magic happens here! */
1888 }
1889
2b2d6d01 1890#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 1891
1892 if (nr_orphans)
617ba13b 1893 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1894 sb->s_id, PLURAL(nr_orphans));
1895 if (nr_truncates)
617ba13b 1896 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1897 sb->s_id, PLURAL(nr_truncates));
1898#ifdef CONFIG_QUOTA
1899 /* Turn quotas off */
1900 for (i = 0; i < MAXQUOTAS; i++) {
1901 if (sb_dqopt(sb)->files[i])
6f28e087 1902 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1903 }
1904#endif
1905 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1906}
cd2291a4
ES		 1907/*
1908 * Maximal extent format file size.
1909 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1910 * extent format containers, within a sector_t, and within i_blocks
1911 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1912 * so that won't be a limiting factor.
1913 *
1914 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1915 */
f287a1a5 1916static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 1917{
1918 loff_t res;
1919 loff_t upper_limit = MAX_LFS_FILESIZE;
1920
1921 /* small i_blocks in vfs inode? */
f287a1a5 1922 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 1923 /*
b3a6ffe1 1924 * CONFIG_LBD is not enabled implies the inode
cd2291a4
ES		 1925 * i_block represent total blocks in 512 bytes
1926 * 32 == size of vfs inode i_blocks * 8
1927 */
1928 upper_limit = (1LL << 32) - 1;
1929
1930 /* total blocks in file system block size */
1931 upper_limit >>= (blkbits - 9);
1932 upper_limit <<= blkbits;
1933 }
1934
1935 /* 32-bit extent-start container, ee_block */
1936 res = 1LL << 32;
1937 res <<= blkbits;
1938 res -= 1;
1939
1940 /* Sanity check against vm- & vfs- imposed limits */
1941 if (res > upper_limit)
1942 res = upper_limit;
1943
1944 return res;
1945}
ac27a0ec 1946
ac27a0ec 1947/*
cd2291a4 1948 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1949 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1950 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1951 */
f287a1a5 1952static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 1953{
617ba13b 1954 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1955 int meta_blocks;
1956 loff_t upper_limit;
1957 /* This is calculated to be the largest file size for a
cd2291a4 1958 * dense, bitmapped file such that the total number of
ac27a0ec 1959 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1960 * does not exceed 2^48 -1
1961 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1962 * total number of 512 bytes blocks of the file
1963 */
1964
f287a1a5 1965 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 1966 /*
b3a6ffe1 1967 * !has_huge_files or CONFIG_LBD is not enabled
f287a1a5
TT		 1968 * implies the inode i_block represent total blocks in
1969 * 512 bytes 32 == size of vfs inode i_blocks * 8
0fc1b451
AK		 1970 */
1971 upper_limit = (1LL << 32) - 1;
1972
1973 /* total blocks in file system block size */
1974 upper_limit >>= (bits - 9);
1975
1976 } else {
8180a562
AK		 1977 /*
1978 * We use 48 bit ext4_inode i_blocks
1979 * With EXT4_HUGE_FILE_FL set the i_blocks
1980 * represent total number of blocks in
1981 * file system block size
1982 */
0fc1b451
AK		 1983 upper_limit = (1LL << 48) - 1;
1984
0fc1b451
AK		 1985 }
1986
1987 /* indirect blocks */
1988 meta_blocks = 1;
1989 /* double indirect blocks */
1990 meta_blocks += 1 + (1LL << (bits-2));
1991 /* tripple indirect blocks */
1992 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1993
1994 upper_limit -= meta_blocks;
1995 upper_limit <<= bits;
ac27a0ec
DK		 1996
1997 res += 1LL << (bits-2);
1998 res += 1LL << (2*(bits-2));
1999 res += 1LL << (3*(bits-2));
2000 res <<= bits;
2001 if (res > upper_limit)
2002 res = upper_limit;
0fc1b451
AK		 2003
2004 if (res > MAX_LFS_FILESIZE)
2005 res = MAX_LFS_FILESIZE;
2006
ac27a0ec
DK		 2007 return res;
2008}
2009
617ba13b 2010static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 2011 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2012{
617ba13b 2013 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2014 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2015 int has_super = 0;
2016
2017 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2018
617ba13b 2019 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2020 nr < first_meta_bg)
70bbb3e0 2021 return logical_sb_block + nr + 1;
ac27a0ec 2022 bg = sbi->s_desc_per_block * nr;
617ba13b 2023 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2024 has_super = 1;
617ba13b 2025 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2026}
2027
c9de560d
AT		 2028/**
2029 * ext4_get_stripe_size: Get the stripe size.
2030 * @sbi: In memory super block info
2031 *
2032 * If we have specified it via mount option, then
2033 * use the mount option value. If the value specified at mount time is
2034 * greater than the blocks per group use the super block value.
2035 * If the super block value is greater than blocks per group return 0.
2036 * Allocator needs it be less than blocks per group.
2037 *
2038 */
2039static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2040{
2041 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2042 unsigned long stripe_width =
2043 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2044
2045 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2046 return sbi->s_stripe;
2047
2048 if (stripe_width <= sbi->s_blocks_per_group)
2049 return stripe_width;
2050
2051 if (stride <= sbi->s_blocks_per_group)
2052 return stride;
2053
2054 return 0;
2055}
ac27a0ec 2056
3197ebdb
TT		 2057/* sysfs supprt */
2058
2059struct ext4_attr {
2060 struct attribute attr;
2061 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2062 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2063 const char *, size_t);
2064 int offset;
2065};
2066
2067static int parse_strtoul(const char *buf,
2068 unsigned long max, unsigned long *value)
2069{
2070 char *endp;
2071
2072 while (*buf && isspace(*buf))
2073 buf++;
2074 *value = simple_strtoul(buf, &endp, 0);
2075 while (*endp && isspace(*endp))
2076 endp++;
2077 if (*endp || *value > max)
2078 return -EINVAL;
2079
2080 return 0;
2081}
2082
2083static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2084 struct ext4_sb_info *sbi,
2085 char *buf)
2086{
2087 return snprintf(buf, PAGE_SIZE, "%llu\n",
2088 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2089}
2090
2091static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2092 struct ext4_sb_info *sbi, char *buf)
2093{
2094 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2095
2096 return snprintf(buf, PAGE_SIZE, "%lu\n",
2097 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2098 sbi->s_sectors_written_start) >> 1);
2099}
2100
2101static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2102 struct ext4_sb_info *sbi, char *buf)
2103{
2104 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2105
2106 return snprintf(buf, PAGE_SIZE, "%llu\n",
2107 sbi->s_kbytes_written +
2108 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2109 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2110}
2111
2112static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2113 struct ext4_sb_info *sbi,
2114 const char *buf, size_t count)
2115{
2116 unsigned long t;
2117
2118 if (parse_strtoul(buf, 0x40000000, &t))
2119 return -EINVAL;
2120
f7c43950 2121 if (!is_power_of_2(t))
3197ebdb
TT		 2122 return -EINVAL;
2123
2124 sbi->s_inode_readahead_blks = t;
2125 return count;
2126}
2127
2128static ssize_t sbi_ui_show(struct ext4_attr *a,
2129 struct ext4_sb_info *sbi, char *buf)
2130{
2131 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2132
2133 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2134}
2135
2136static ssize_t sbi_ui_store(struct ext4_attr *a,
2137 struct ext4_sb_info *sbi,
2138 const char *buf, size_t count)
2139{
2140 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2141 unsigned long t;
2142
2143 if (parse_strtoul(buf, 0xffffffff, &t))
2144 return -EINVAL;
2145 *ui = t;
2146 return count;
2147}
2148
2149#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2150static struct ext4_attr ext4_attr_##_name = { \
2151 .attr = {.name = __stringify(_name), .mode = _mode }, \
2152 .show = _show, \
2153 .store = _store, \
2154 .offset = offsetof(struct ext4_sb_info, _elname), \
2155}
2156#define EXT4_ATTR(name, mode, show, store) \
2157static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2158
2159#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2160#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2161#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2162 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2163#define ATTR_LIST(name) &ext4_attr_##name.attr
2164
2165EXT4_RO_ATTR(delayed_allocation_blocks);
2166EXT4_RO_ATTR(session_write_kbytes);
2167EXT4_RO_ATTR(lifetime_write_kbytes);
2168EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2169 inode_readahead_blks_store, s_inode_readahead_blks);
2170EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2171EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2172EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2173EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2174EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2175EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2176
2177static struct attribute *ext4_attrs[] = {
2178 ATTR_LIST(delayed_allocation_blocks),
2179 ATTR_LIST(session_write_kbytes),
2180 ATTR_LIST(lifetime_write_kbytes),
2181 ATTR_LIST(inode_readahead_blks),
2182 ATTR_LIST(mb_stats),
2183 ATTR_LIST(mb_max_to_scan),
2184 ATTR_LIST(mb_min_to_scan),
2185 ATTR_LIST(mb_order2_req),
2186 ATTR_LIST(mb_stream_req),
2187 ATTR_LIST(mb_group_prealloc),
2188 NULL,
2189};
2190
2191static ssize_t ext4_attr_show(struct kobject *kobj,
2192 struct attribute *attr, char *buf)
2193{
2194 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2195 s_kobj);
2196 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2197
2198 return a->show ? a->show(a, sbi, buf) : 0;
2199}
2200
2201static ssize_t ext4_attr_store(struct kobject *kobj,
2202 struct attribute *attr,
2203 const char *buf, size_t len)
2204{
2205 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2206 s_kobj);
2207 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2208
2209 return a->store ? a->store(a, sbi, buf, len) : 0;
2210}
2211
2212static void ext4_sb_release(struct kobject *kobj)
2213{
2214 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2215 s_kobj);
2216 complete(&sbi->s_kobj_unregister);
2217}
2218
2219
2220static struct sysfs_ops ext4_attr_ops = {
2221 .show = ext4_attr_show,
2222 .store = ext4_attr_store,
2223};
2224
2225static struct kobj_type ext4_ktype = {
2226 .default_attrs = ext4_attrs,
2227 .sysfs_ops = &ext4_attr_ops,
2228 .release = ext4_sb_release,
2229};
2230
2b2d6d01 2231static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 2232 __releases(kernel_lock)
2233 __acquires(kernel_lock)
1d03ec98 2234
ac27a0ec 2235{
2b2d6d01 2236 struct buffer_head *bh;
617ba13b
MC		 2237 struct ext4_super_block *es = NULL;
2238 struct ext4_sb_info *sbi;
2239 ext4_fsblk_t block;
2240 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2241 ext4_fsblk_t logical_sb_block;
ac27a0ec 2242 unsigned long offset = 0;
ac27a0ec
DK		 2243 unsigned long journal_devnum = 0;
2244 unsigned long def_mount_opts;
2245 struct inode *root;
9f6200bb 2246 char *cp;
0390131b 2247 const char *descr;
1d1fe1ee 2248 int ret = -EINVAL;
ac27a0ec 2249 int blocksize;
4ec11028
TT		 2250 unsigned int db_count;
2251 unsigned int i;
f287a1a5 2252 int needs_recovery, has_huge_files;
3a06d778 2253 int features;
bd81d8ee 2254 __u64 blocks_count;
833f4077 2255 int err;
b3881f74 2256 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 2257
2258 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2259 if (!sbi)
2260 return -ENOMEM;
705895b6
PE		 2261
2262 sbi->s_blockgroup_lock =
2263 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2264 if (!sbi->s_blockgroup_lock) {
2265 kfree(sbi);
2266 return -ENOMEM;
2267 }
ac27a0ec
DK		 2268 sb->s_fs_info = sbi;
2269 sbi->s_mount_opt = 0;
617ba13b
MC		 2270 sbi->s_resuid = EXT4_DEF_RESUID;
2271 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2272 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2273 sbi->s_sb_block = sb_block;
afc32f7e
TT		 2274 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2275 sectors[1]);
ac27a0ec
DK		 2276
2277 unlock_kernel();
2278
9f6200bb
TT		 2279 /* Cleanup superblock name */
2280 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2281 *cp = '!';
2282
617ba13b 2283 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2284 if (!blocksize) {
617ba13b 2285 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 2286 goto out_fail;
2287 }
2288
2289 /*
617ba13b 2290 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 2291 * block sizes. We need to calculate the offset from buffer start.
2292 */
617ba13b 2293 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 2294 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2295 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 2296 } else {
70bbb3e0 2297 logical_sb_block = sb_block;
ac27a0ec
DK		 2298 }
2299
70bbb3e0 2300 if (!(bh = sb_bread(sb, logical_sb_block))) {
2b2d6d01 2301 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 2302 goto out_fail;
2303 }
2304 /*
2305 * Note: s_es must be initialized as soon as possible because
617ba13b 2306 * some ext4 macro-instructions depend on its value
ac27a0ec 2307 */
617ba13b 2308 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 2309 sbi->s_es = es;
2310 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 2311 if (sb->s_magic != EXT4_SUPER_MAGIC)
2312 goto cantfind_ext4;
afc32f7e 2313 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 2314
2315 /* Set defaults before we parse the mount options */
2316 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 2317 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 2318 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 2319 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 2320 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 2321 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 2322 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 2323#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 2324 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 2325 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 2326#endif
03010a33 2327#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2328 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2329 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2330#endif
617ba13b
MC		 2331 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2332 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2333 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2334 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2335 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2336 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2337
2338 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2339 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2340 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2341 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2342 else
2343 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2344
2345 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2346 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 2347 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2348 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2349 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 2350
571640ca 2351 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2352
dd919b98
AK		 2353 /*
2354 * enable delayed allocation by default
2355 * Use -o nodelalloc to turn it off
2356 */
2357 set_opt(sbi->s_mount_opt, DELALLOC);
2358
2359
b3881f74
TT		 2360 if (!parse_options((char *) data, sb, &journal_devnum,
2361 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 2362 goto failed_mount;
2363
2364 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2365 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2366
617ba13b
MC		 2367 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2368 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2369 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2370 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 2371 printk(KERN_WARNING
617ba13b 2372 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec 2373 "running e2fsck is recommended\n");
469108ff 2374
ac27a0ec
DK		 2375 /*
2376 * Check feature flags regardless of the revision level, since we
2377 * previously didn't change the revision level when setting the flags,
2378 * so there is a chance incompat flags are set on a rev 0 filesystem.
2379 */
617ba13b 2380 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 2381 if (features) {
617ba13b 2382 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
3a06d778
AK		 2383 "unsupported optional features (%x).\n", sb->s_id,
2384 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2385 ~EXT4_FEATURE_INCOMPAT_SUPP));
ac27a0ec
DK		 2386 goto failed_mount;
2387 }
617ba13b 2388 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 2389 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 2390 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
3a06d778
AK		 2391 "unsupported optional features (%x).\n", sb->s_id,
2392 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2393 ~EXT4_FEATURE_RO_COMPAT_SUPP));
ac27a0ec
DK		 2394 goto failed_mount;
2395 }
f287a1a5
TT		 2396 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2397 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2398 if (has_huge_files) {
0fc1b451
AK		 2399 /*
2400 * Large file size enabled file system can only be
b3a6ffe1 2401 * mount if kernel is build with CONFIG_LBD
0fc1b451
AK		 2402 */
2403 if (sizeof(root->i_blocks) < sizeof(u64) &&
2404 !(sb->s_flags & MS_RDONLY)) {
2405 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2406 "files cannot be mounted read-write "
b3a6ffe1 2407 "without CONFIG_LBD.\n", sb->s_id);
0fc1b451
AK		 2408 goto failed_mount;
2409 }
2410 }
ac27a0ec
DK		 2411 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2412
617ba13b
MC		 2413 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2414 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 2415 printk(KERN_ERR
617ba13b 2416 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 2417 blocksize, sb->s_id);
2418 goto failed_mount;
2419 }
2420
ac27a0ec 2421 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2422
2423 /* Validate the filesystem blocksize */
2424 if (!sb_set_blocksize(sb, blocksize)) {
2425 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2426 blocksize);
ac27a0ec
DK		 2427 goto failed_mount;
2428 }
2429
2b2d6d01 2430 brelse(bh);
70bbb3e0
AM		 2431 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2432 offset = do_div(logical_sb_block, blocksize);
2433 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 2434 if (!bh) {
2435 printk(KERN_ERR
617ba13b 2436 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 2437 goto failed_mount;
2438 }
617ba13b 2439 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2440 sbi->s_es = es;
617ba13b 2441 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2b2d6d01
TT		 2442 printk(KERN_ERR
2443 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 2444 goto failed_mount;
2445 }
2446 }
2447
f287a1a5
TT		 2448 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2449 has_huge_files);
2450 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2451
617ba13b
MC		 2452 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2453 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2454 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2455 } else {
2456 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2457 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2458 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2459 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2460 (sbi->s_inode_size > blocksize)) {
2b2d6d01
TT		 2461 printk(KERN_ERR
2462 "EXT4-fs: unsupported inode size: %d\n",
2463 sbi->s_inode_size);
ac27a0ec
DK		 2464 goto failed_mount;
2465 }
ef7f3835
KS		 2466 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2467 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2468 }
0d1ee42f
AR		 2469 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2470 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2471 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2472 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2473 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2474 printk(KERN_ERR
8fadc143 2475 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2476 sbi->s_desc_size);
2477 goto failed_mount;
2478 }
2479 } else
2480 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2481 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2482 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2483 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2484 goto cantfind_ext4;
2485 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2486 if (sbi->s_inodes_per_block == 0)
617ba13b 2487 goto cantfind_ext4;
ac27a0ec
DK		 2488 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2489 sbi->s_inodes_per_block;
0d1ee42f 2490 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2491 sbi->s_sbh = bh;
2492 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2493 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2494 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2b2d6d01 2495 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2496 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2497 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2498 i = le32_to_cpu(es->s_flags);
2499 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2500 sbi->s_hash_unsigned = 3;
2501 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2502#ifdef __CHAR_UNSIGNED__
2503 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2504 sbi->s_hash_unsigned = 3;
2505#else
2506 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2507#endif
2508 sb->s_dirt = 1;
2509 }
ac27a0ec
DK		 2510
2511 if (sbi->s_blocks_per_group > blocksize * 8) {
2b2d6d01
TT		 2512 printk(KERN_ERR
2513 "EXT4-fs: #blocks per group too big: %lu\n",
2514 sbi->s_blocks_per_group);
ac27a0ec
DK		 2515 goto failed_mount;
2516 }
ac27a0ec 2517 if (sbi->s_inodes_per_group > blocksize * 8) {
2b2d6d01
TT		 2518 printk(KERN_ERR
2519 "EXT4-fs: #inodes per group too big: %lu\n",
2520 sbi->s_inodes_per_group);
ac27a0ec
DK		 2521 goto failed_mount;
2522 }
2523
bd81d8ee 2524 if (ext4_blocks_count(es) >
ac27a0ec 2525 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2526 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2527 " too large to mount safely\n", sb->s_id);
2528 if (sizeof(sector_t) < 8)
617ba13b 2529 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2530 "enabled\n");
2531 goto failed_mount;
2532 }
2533
617ba13b
MC		 2534 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2535 goto cantfind_ext4;
e7c95593 2536
0f2ddca6
FTN		 2537 /* check blocks count against device size */
2538 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2539 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2540 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu "
2541 "exceeds size of device (%llu blocks)\n",
2542 ext4_blocks_count(es), blocks_count);
2543 goto failed_mount;
2544 }
2545
4ec11028
TT		 2546 /*
2547 * It makes no sense for the first data block to be beyond the end
2548 * of the filesystem.
2549 */
2550 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2551 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2552 "block %u is beyond end of filesystem (%llu)\n",
2553 le32_to_cpu(es->s_first_data_block),
2554 ext4_blocks_count(es));
e7c95593
ES		 2555 goto failed_mount;
2556 }
bd81d8ee
LV		 2557 blocks_count = (ext4_blocks_count(es) -
2558 le32_to_cpu(es->s_first_data_block) +
2559 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2560 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028
TT		 2561 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2562 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2563 "(block count %llu, first data block %u, "
2564 "blocks per group %lu)\n", sbi->s_groups_count,
2565 ext4_blocks_count(es),
2566 le32_to_cpu(es->s_first_data_block),
2567 EXT4_BLOCKS_PER_GROUP(sb));
2568 goto failed_mount;
2569 }
bd81d8ee 2570 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2571 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2572 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2573 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2574 GFP_KERNEL);
2575 if (sbi->s_group_desc == NULL) {
2b2d6d01 2576 printk(KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2577 goto failed_mount;
2578 }
2579
3244fcb1 2580#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2581 if (ext4_proc_root)
2582 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 2583#endif
240799cd 2584
705895b6 2585 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2586
2587 for (i = 0; i < db_count; i++) {
70bbb3e0 2588 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2589 sbi->s_group_desc[i] = sb_bread(sb, block);
2590 if (!sbi->s_group_desc[i]) {
2b2d6d01
TT		 2591 printk(KERN_ERR "EXT4-fs: "
2592 "can't read group descriptor %d\n", i);
ac27a0ec
DK		 2593 db_count = i;
2594 goto failed_mount2;
2595 }
2596 }
2b2d6d01 2597 if (!ext4_check_descriptors(sb)) {
617ba13b 2598 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2599 goto failed_mount2;
2600 }
772cb7c8
JS		 2601 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2602 if (!ext4_fill_flex_info(sb)) {
2603 printk(KERN_ERR
2604 "EXT4-fs: unable to initialize "
2605 "flex_bg meta info!\n");
2606 goto failed_mount2;
2607 }
2608
ac27a0ec
DK		 2609 sbi->s_gdb_count = db_count;
2610 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2611 spin_lock_init(&sbi->s_next_gen_lock);
2612
833f4077
PZ		 2613 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2614 ext4_count_free_blocks(sb));
2615 if (!err) {
2616 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2617 ext4_count_free_inodes(sb));
2618 }
2619 if (!err) {
2620 err = percpu_counter_init(&sbi->s_dirs_counter,
2621 ext4_count_dirs(sb));
2622 }
6bc6e63f
AK		 2623 if (!err) {
2624 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2625 }
833f4077
PZ		 2626 if (err) {
2627 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2628 goto failed_mount3;
2629 }
ac27a0ec 2630
c9de560d
AT		 2631 sbi->s_stripe = ext4_get_stripe_size(sbi);
2632
ac27a0ec
DK		 2633 /*
2634 * set up enough so that it can read an inode
2635 */
9ca92389
TT		 2636 if (!test_opt(sb, NOLOAD) &&
2637 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2638 sb->s_op = &ext4_sops;
2639 else
2640 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 2641 sb->s_export_op = &ext4_export_ops;
2642 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2643#ifdef CONFIG_QUOTA
617ba13b
MC		 2644 sb->s_qcop = &ext4_qctl_operations;
2645 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2646#endif
2647 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 2648 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 2649
2650 sb->s_root = NULL;
2651
2652 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2653 EXT4_HAS_INCOMPAT_FEATURE(sb,
2654 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2655
2656 /*
2657 * The first inode we look at is the journal inode. Don't try
2658 * root first: it may be modified in the journal!
2659 */
2660 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2661 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2662 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2663 goto failed_mount3;
624080ed
TT		 2664 if (!(sb->s_flags & MS_RDONLY) &&
2665 EXT4_SB(sb)->s_journal->j_failed_commit) {
2666 printk(KERN_CRIT "EXT4-fs error (device %s): "
2667 "ext4_fill_super: Journal transaction "
2b2d6d01 2668 "%u is corrupt\n", sb->s_id,
624080ed 2669 EXT4_SB(sb)->s_journal->j_failed_commit);
2b2d6d01
TT		 2670 if (test_opt(sb, ERRORS_RO)) {
2671 printk(KERN_CRIT
2672 "Mounting filesystem read-only\n");
624080ed
TT		 2673 sb->s_flags |= MS_RDONLY;
2674 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2675 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2676 }
2677 if (test_opt(sb, ERRORS_PANIC)) {
2678 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2679 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 2680 ext4_commit_super(sb, 1);
624080ed
TT		 2681 goto failed_mount4;
2682 }
2683 }
0390131b
FM		 2684 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2685 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2686 printk(KERN_ERR "EXT4-fs: required journal recovery "
2687 "suppressed and not mounted read-only\n");
2688 goto failed_mount4;
ac27a0ec 2689 } else {
0390131b
FM		 2690 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2691 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2692 sbi->s_journal = NULL;
2693 needs_recovery = 0;
2694 goto no_journal;
ac27a0ec
DK		 2695 }
2696
eb40a09c
JS		 2697 if (ext4_blocks_count(es) > 0xffffffffULL &&
2698 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2699 JBD2_FEATURE_INCOMPAT_64BIT)) {
abda1418 2700 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
eb40a09c
JS		 2701 goto failed_mount4;
2702 }
2703
818d276c
GS		 2704 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2705 jbd2_journal_set_features(sbi->s_journal,
2706 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2707 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2708 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2709 jbd2_journal_set_features(sbi->s_journal,
2710 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2711 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2712 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2713 } else {
2714 jbd2_journal_clear_features(sbi->s_journal,
2715 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2716 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2717 }
2718
ac27a0ec
DK		 2719 /* We have now updated the journal if required, so we can
2720 * validate the data journaling mode. */
2721 switch (test_opt(sb, DATA_FLAGS)) {
2722 case 0:
2723 /* No mode set, assume a default based on the journal
63f57933
AM		 2724 * capabilities: ORDERED_DATA if the journal can
2725 * cope, else JOURNAL_DATA
2726 */
dab291af
MC		 2727 if (jbd2_journal_check_available_features
2728 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2729 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2730 else
2731 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2732 break;
2733
617ba13b
MC		 2734 case EXT4_MOUNT_ORDERED_DATA:
2735 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2736 if (!jbd2_journal_check_available_features
2737 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2738 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2739 "requested data journaling mode\n");
2740 goto failed_mount4;
2741 }
2742 default:
2743 break;
2744 }
b3881f74 2745 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 2746
0390131b 2747no_journal:
ac27a0ec
DK		 2748
2749 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2750 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2751 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2752 "its supported only with writeback mode\n");
2753 clear_opt(sbi->s_mount_opt, NOBH);
2754 }
2755 }
2756 /*
dab291af 2757 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2758 * so we can safely mount the rest of the filesystem now.
2759 */
2760
1d1fe1ee
DH		 2761 root = ext4_iget(sb, EXT4_ROOT_INO);
2762 if (IS_ERR(root)) {
617ba13b 2763 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1d1fe1ee 2764 ret = PTR_ERR(root);
ac27a0ec
DK		 2765 goto failed_mount4;
2766 }
2767 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2768 iput(root);
617ba13b 2769 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2770 goto failed_mount4;
2771 }
1d1fe1ee
DH		 2772 sb->s_root = d_alloc_root(root);
2773 if (!sb->s_root) {
2774 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2775 iput(root);
2776 ret = -ENOMEM;
2777 goto failed_mount4;
2778 }
ac27a0ec 2779
2b2d6d01 2780 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2781
2782 /* determine the minimum size of new large inodes, if present */
2783 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2784 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2785 EXT4_GOOD_OLD_INODE_SIZE;
2786 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2787 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2788 if (sbi->s_want_extra_isize <
2789 le16_to_cpu(es->s_want_extra_isize))
2790 sbi->s_want_extra_isize =
2791 le16_to_cpu(es->s_want_extra_isize);
2792 if (sbi->s_want_extra_isize <
2793 le16_to_cpu(es->s_min_extra_isize))
2794 sbi->s_want_extra_isize =
2795 le16_to_cpu(es->s_min_extra_isize);
2796 }
2797 }
2798 /* Check if enough inode space is available */
2799 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2800 sbi->s_inode_size) {
2801 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2802 EXT4_GOOD_OLD_INODE_SIZE;
2803 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2804 "available.\n");
2805 }
2806
c2774d84
AK		 2807 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2808 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2809 "requested data journaling mode\n");
2810 clear_opt(sbi->s_mount_opt, DELALLOC);
2811 } else if (test_opt(sb, DELALLOC))
2812 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2813
2814 ext4_ext_init(sb);
2815 err = ext4_mb_init(sb, needs_recovery);
2816 if (err) {
2817 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2818 err);
2819 goto failed_mount4;
2820 }
2821
3197ebdb
TT		 2822 sbi->s_kobj.kset = ext4_kset;
2823 init_completion(&sbi->s_kobj_unregister);
2824 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2825 "%s", sb->s_id);
2826 if (err) {
2827 ext4_mb_release(sb);
2828 ext4_ext_release(sb);
2829 goto failed_mount4;
2830 };
2831
617ba13b
MC		 2832 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2833 ext4_orphan_cleanup(sb, es);
2834 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 2835 if (needs_recovery) {
2b2d6d01 2836 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
0390131b
FM		 2837 ext4_mark_recovery_complete(sb, es);
2838 }
2839 if (EXT4_SB(sb)->s_journal) {
2840 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2841 descr = " journalled data mode";
2842 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2843 descr = " ordered data mode";
2844 else
2845 descr = " writeback data mode";
2846 } else
2847 descr = "out journal";
2848
2849 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2850 sb->s_id, descr);
ac27a0ec
DK		 2851
2852 lock_kernel();
2853 return 0;
2854
617ba13b 2855cantfind_ext4:
ac27a0ec 2856 if (!silent)
617ba13b 2857 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2858 sb->s_id);
2859 goto failed_mount;
2860
2861failed_mount4:
0390131b
FM		 2862 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2863 if (sbi->s_journal) {
2864 jbd2_journal_destroy(sbi->s_journal);
2865 sbi->s_journal = NULL;
2866 }
ac27a0ec 2867failed_mount3:
c5ca7c76
TT		 2868 if (sbi->s_flex_groups) {
2869 if (is_vmalloc_addr(sbi->s_flex_groups))
2870 vfree(sbi->s_flex_groups);
2871 else
2872 kfree(sbi->s_flex_groups);
2873 }
ac27a0ec
DK		 2874 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2875 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2876 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 2877 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 2878failed_mount2:
2879 for (i = 0; i < db_count; i++)
2880 brelse(sbi->s_group_desc[i]);
2881 kfree(sbi->s_group_desc);
2882failed_mount:
240799cd 2883 if (sbi->s_proc) {
9f6200bb 2884 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 2885 }
ac27a0ec
DK		 2886#ifdef CONFIG_QUOTA
2887 for (i = 0; i < MAXQUOTAS; i++)
2888 kfree(sbi->s_qf_names[i]);
2889#endif
617ba13b 2890 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2891 brelse(bh);
2892out_fail:
2893 sb->s_fs_info = NULL;
2894 kfree(sbi);
2895 lock_kernel();
1d1fe1ee 2896 return ret;
ac27a0ec
DK		 2897}
2898
2899/*
2900 * Setup any per-fs journal parameters now. We'll do this both on
2901 * initial mount, once the journal has been initialised but before we've
2902 * done any recovery; and again on any subsequent remount.
2903 */
617ba13b 2904static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2905{
617ba13b 2906 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 2907
30773840
TT		 2908 journal->j_commit_interval = sbi->s_commit_interval;
2909 journal->j_min_batch_time = sbi->s_min_batch_time;
2910 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 2911
2912 spin_lock(&journal->j_state_lock);
2913 if (test_opt(sb, BARRIER))
dab291af 2914 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2915 else
dab291af 2916 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 2917 if (test_opt(sb, DATA_ERR_ABORT))
2918 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2919 else
2920 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 2921 spin_unlock(&journal->j_state_lock);
2922}
2923
617ba13b 2924static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2925 unsigned int journal_inum)
2926{
2927 struct inode *journal_inode;
2928 journal_t *journal;
2929
0390131b
FM		 2930 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2931
ac27a0ec
DK		 2932 /* First, test for the existence of a valid inode on disk. Bad
2933 * things happen if we iget() an unused inode, as the subsequent
2934 * iput() will try to delete it. */
2935
1d1fe1ee
DH		 2936 journal_inode = ext4_iget(sb, journal_inum);
2937 if (IS_ERR(journal_inode)) {
617ba13b 2938 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2939 return NULL;
2940 }
2941 if (!journal_inode->i_nlink) {
2942 make_bad_inode(journal_inode);
2943 iput(journal_inode);
617ba13b 2944 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2945 return NULL;
2946 }
2947
e5f8eab8 2948 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 2949 journal_inode, journal_inode->i_size);
1d1fe1ee 2950 if (!S_ISREG(journal_inode->i_mode)) {
617ba13b 2951 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2952 iput(journal_inode);
2953 return NULL;
2954 }
2955
dab291af 2956 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2957 if (!journal) {
617ba13b 2958 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2959 iput(journal_inode);
2960 return NULL;
2961 }
2962 journal->j_private = sb;
617ba13b 2963 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2964 return journal;
2965}
2966
617ba13b 2967static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2968 dev_t j_dev)
2969{
2b2d6d01 2970 struct buffer_head *bh;
ac27a0ec 2971 journal_t *journal;
617ba13b
MC		 2972 ext4_fsblk_t start;
2973 ext4_fsblk_t len;
ac27a0ec 2974 int hblock, blocksize;
617ba13b 2975 ext4_fsblk_t sb_block;
ac27a0ec 2976 unsigned long offset;
2b2d6d01 2977 struct ext4_super_block *es;
ac27a0ec
DK		 2978 struct block_device *bdev;
2979
0390131b
FM		 2980 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2981
617ba13b 2982 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2983 if (bdev == NULL)
2984 return NULL;
2985
2986 if (bd_claim(bdev, sb)) {
2987 printk(KERN_ERR
abda1418 2988 "EXT4-fs: failed to claim external journal device.\n");
9a1c3542 2989 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 2990 return NULL;
2991 }
2992
2993 blocksize = sb->s_blocksize;
2994 hblock = bdev_hardsect_size(bdev);
2995 if (blocksize < hblock) {
2996 printk(KERN_ERR
617ba13b 2997 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2998 goto out_bdev;
2999 }
3000
617ba13b
MC		 3001 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3002 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3003 set_blocksize(bdev, blocksize);
3004 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 3005 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 3006 "external journal\n");
3007 goto out_bdev;
3008 }
3009
617ba13b
MC		 3010 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3011 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3012 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 3013 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3014 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 3015 "bad superblock\n");
3016 brelse(bh);
3017 goto out_bdev;
3018 }
3019
617ba13b
MC		 3020 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3021 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 3022 brelse(bh);
3023 goto out_bdev;
3024 }
3025
bd81d8ee 3026 len = ext4_blocks_count(es);
ac27a0ec
DK		 3027 start = sb_block + 1;
3028 brelse(bh); /* we're done with the superblock */
3029
dab291af 3030 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3031 start, len, blocksize);
3032 if (!journal) {
617ba13b 3033 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 3034 goto out_bdev;
3035 }
3036 journal->j_private = sb;
3037 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3038 wait_on_buffer(journal->j_sb_buffer);
3039 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 3040 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 3041 goto out_journal;
3042 }
3043 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 3044 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 3045 "user (unsupported) - %d\n",
3046 be32_to_cpu(journal->j_superblock->s_nr_users));
3047 goto out_journal;
3048 }
617ba13b
MC		 3049 EXT4_SB(sb)->journal_bdev = bdev;
3050 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3051 return journal;
3052out_journal:
dab291af 3053 jbd2_journal_destroy(journal);
ac27a0ec 3054out_bdev:
617ba13b 3055 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3056 return NULL;
3057}
3058
617ba13b
MC		 3059static int ext4_load_journal(struct super_block *sb,
3060 struct ext4_super_block *es,
ac27a0ec
DK		 3061 unsigned long journal_devnum)
3062{
3063 journal_t *journal;
3064 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3065 dev_t journal_dev;
3066 int err = 0;
3067 int really_read_only;
3068
0390131b
FM		 3069 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3070
ac27a0ec
DK		 3071 if (journal_devnum &&
3072 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 3073 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 3074 "numbers have changed\n");
3075 journal_dev = new_decode_dev(journal_devnum);
3076 } else
3077 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3078
3079 really_read_only = bdev_read_only(sb->s_bdev);
3080
3081 /*
3082 * Are we loading a blank journal or performing recovery after a
3083 * crash? For recovery, we need to check in advance whether we
3084 * can get read-write access to the device.
3085 */
3086
617ba13b 3087 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3088 if (sb->s_flags & MS_RDONLY) {
617ba13b 3089 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 3090 "required on readonly filesystem.\n");
3091 if (really_read_only) {
617ba13b 3092 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 3093 "unavailable, cannot proceed.\n");
3094 return -EROFS;
3095 }
2b2d6d01
TT		 3096 printk(KERN_INFO "EXT4-fs: write access will "
3097 "be enabled during recovery.\n");
ac27a0ec
DK		 3098 }
3099 }
3100
3101 if (journal_inum && journal_dev) {
617ba13b 3102 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 3103 "and inode journals!\n");
3104 return -EINVAL;
3105 }
3106
3107 if (journal_inum) {
617ba13b 3108 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3109 return -EINVAL;
3110 } else {
617ba13b 3111 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3112 return -EINVAL;
3113 }
3114
4776004f
TT		 3115 if (journal->j_flags & JBD2_BARRIER)
3116 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
3117 else
3118 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
3119
ac27a0ec 3120 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3121 err = jbd2_journal_update_format(journal);
ac27a0ec 3122 if (err) {
617ba13b 3123 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 3124 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3125 return err;
3126 }
3127 }
3128
617ba13b 3129 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3130 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 3131 if (!err)
dab291af 3132 err = jbd2_journal_load(journal);
ac27a0ec
DK		 3133
3134 if (err) {
617ba13b 3135 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 3136 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3137 return err;
3138 }
3139
617ba13b
MC		 3140 EXT4_SB(sb)->s_journal = journal;
3141 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 3142
3143 if (journal_devnum &&
3144 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3145 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3146
3147 /* Make sure we flush the recovery flag to disk. */
e2d67052 3148 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3149 }
3150
3151 return 0;
3152}
3153
e2d67052 3154static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3155{
e2d67052 3156 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3157 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3158 int error = 0;
ac27a0ec
DK		 3159
3160 if (!sbh)
c4be0c1d 3161 return error;
914258bf
TT		 3162 if (buffer_write_io_error(sbh)) {
3163 /*
3164 * Oh, dear. A previous attempt to write the
3165 * superblock failed. This could happen because the
3166 * USB device was yanked out. Or it could happen to
3167 * be a transient write error and maybe the block will
3168 * be remapped. Nothing we can do but to retry the
3169 * write and hope for the best.
3170 */
abda1418 3171 printk(KERN_ERR "EXT4-fs: previous I/O error to "
914258bf
TT		 3172 "superblock detected for %s.\n", sb->s_id);
3173 clear_buffer_write_io_error(sbh);
3174 set_buffer_uptodate(sbh);
3175 }
ac27a0ec 3176 es->s_wtime = cpu_to_le32(get_seconds());
afc32f7e
TT		 3177 es->s_kbytes_written =
3178 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3179 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3180 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5d1b1b3f
AK		 3181 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3182 &EXT4_SB(sb)->s_freeblocks_counter));
3183 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3184 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 3185 sb->s_dirt = 0;
ac27a0ec
DK		 3186 BUFFER_TRACE(sbh, "marking dirty");
3187 mark_buffer_dirty(sbh);
914258bf 3188 if (sync) {
c4be0c1d
TS		 3189 error = sync_dirty_buffer(sbh);
3190 if (error)
3191 return error;
3192
3193 error = buffer_write_io_error(sbh);
3194 if (error) {
abda1418 3195 printk(KERN_ERR "EXT4-fs: I/O error while writing "
914258bf
TT		 3196 "superblock for %s.\n", sb->s_id);
3197 clear_buffer_write_io_error(sbh);
3198 set_buffer_uptodate(sbh);
3199 }
3200 }
c4be0c1d 3201 return error;
ac27a0ec
DK		 3202}
3203
3204
3205/*
3206 * Have we just finished recovery? If so, and if we are mounting (or
3207 * remounting) the filesystem readonly, then we will end up with a
3208 * consistent fs on disk. Record that fact.
3209 */
2b2d6d01
TT		 3210static void ext4_mark_recovery_complete(struct super_block *sb,
3211 struct ext4_super_block *es)
ac27a0ec 3212{
617ba13b 3213 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3214
0390131b
FM		 3215 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3216 BUG_ON(journal != NULL);
3217 return;
3218 }
dab291af 3219 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 3220 if (jbd2_journal_flush(journal) < 0)
3221 goto out;
3222
617ba13b 3223 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 3224 sb->s_flags & MS_RDONLY) {
617ba13b 3225 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 3226 ext4_commit_super(sb, 1);
ac27a0ec 3227 }
7ffe1ea8
HK		 3228
3229out:
dab291af 3230 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 3231}
3232
3233/*
3234 * If we are mounting (or read-write remounting) a filesystem whose journal
3235 * has recorded an error from a previous lifetime, move that error to the
3236 * main filesystem now.
3237 */
2b2d6d01
TT		 3238static void ext4_clear_journal_err(struct super_block *sb,
3239 struct ext4_super_block *es)
ac27a0ec
DK		 3240{
3241 journal_t *journal;
3242 int j_errno;
3243 const char *errstr;
3244
0390131b
FM		 3245 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3246
617ba13b 3247 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 3248
3249 /*
3250 * Now check for any error status which may have been recorded in the
617ba13b 3251 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 3252 */
3253
dab291af 3254 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 3255 if (j_errno) {
3256 char nbuf[16];
3257
617ba13b 3258 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 3259 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 3260 "from previous mount: %s", errstr);
46e665e9 3261 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 3262 "filesystem check.");
3263
617ba13b
MC		 3264 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3265 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 3266 ext4_commit_super(sb, 1);
ac27a0ec 3267
dab291af 3268 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 3269 }
3270}
3271
3272/*
3273 * Force the running and committing transactions to commit,
3274 * and wait on the commit.
3275 */
617ba13b 3276int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 3277{
3278 journal_t *journal;
0390131b 3279 int ret = 0;
ac27a0ec
DK		 3280
3281 if (sb->s_flags & MS_RDONLY)
3282 return 0;
3283
617ba13b 3284 journal = EXT4_SB(sb)->s_journal;
7234ab2a 3285 if (journal)
0390131b 3286 ret = ext4_journal_force_commit(journal);
0390131b 3287
ac27a0ec
DK		 3288 return ret;
3289}
3290
2b2d6d01 3291static void ext4_write_super(struct super_block *sb)
ac27a0ec 3292{
9ca92389 3293 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3294}
3295
617ba13b 3296static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 3297{
14ce0cb4 3298 int ret = 0;
9eddacf9 3299 tid_t target;
ac27a0ec 3300
ede86cc4 3301 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
9ca92389
TT		 3302 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3303 if (wait)
3304 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
0390131b 3305 }
14ce0cb4 3306 return ret;
ac27a0ec
DK		 3307}
3308
3309/*
3310 * LVM calls this function before a (read-only) snapshot is created. This
3311 * gives us a chance to flush the journal completely and mark the fs clean.
3312 */
c4be0c1d 3313static int ext4_freeze(struct super_block *sb)
ac27a0ec 3314{
c4be0c1d
TS		 3315 int error = 0;
3316 journal_t *journal;
ac27a0ec 3317
9ca92389
TT		 3318 if (sb->s_flags & MS_RDONLY)
3319 return 0;
ac27a0ec 3320
9ca92389 3321 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 3322
9ca92389
TT		 3323 /* Now we set up the journal barrier. */
3324 jbd2_journal_lock_updates(journal);
ac27a0ec 3325
9ca92389
TT		 3326 /*
3327 * Don't clear the needs_recovery flag if we failed to flush
3328 * the journal.
3329 */
3330 error = jbd2_journal_flush(journal);
3331 if (error < 0) {
3332 out:
3333 jbd2_journal_unlock_updates(journal);
3334 return error;
ac27a0ec 3335 }
9ca92389
TT		 3336
3337 /* Journal blocked and flushed, clear needs_recovery flag. */
3338 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3339 error = ext4_commit_super(sb, 1);
3340 if (error)
3341 goto out;
c4be0c1d 3342 return 0;
ac27a0ec
DK		 3343}
3344
3345/*
3346 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3347 * flag here, even though the filesystem is not technically dirty yet.
3348 */
c4be0c1d 3349static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 3350{
9ca92389
TT		 3351 if (sb->s_flags & MS_RDONLY)
3352 return 0;
3353
3354 lock_super(sb);
3355 /* Reset the needs_recovery flag before the fs is unlocked. */
3356 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3357 ext4_commit_super(sb, 1);
3358 unlock_super(sb);
3359 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
c4be0c1d 3360 return 0;
ac27a0ec
DK		 3361}
3362
2b2d6d01 3363static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3364{
2b2d6d01 3365 struct ext4_super_block *es;
617ba13b
MC		 3366 struct ext4_sb_info *sbi = EXT4_SB(sb);
3367 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3368 unsigned long old_sb_flags;
617ba13b 3369 struct ext4_mount_options old_opts;
8a266467 3370 ext4_group_t g;
b3881f74 3371 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 3372 int err;
3373#ifdef CONFIG_QUOTA
3374 int i;
3375#endif
3376
3377 /* Store the original options */
3378 old_sb_flags = sb->s_flags;
3379 old_opts.s_mount_opt = sbi->s_mount_opt;
3380 old_opts.s_resuid = sbi->s_resuid;
3381 old_opts.s_resgid = sbi->s_resgid;
3382 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 3383 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3384 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3385#ifdef CONFIG_QUOTA
3386 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3387 for (i = 0; i < MAXQUOTAS; i++)
3388 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3389#endif
b3881f74
TT		 3390 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3391 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 3392
3393 /*
3394 * Allow the "check" option to be passed as a remount option.
3395 */
b3881f74
TT		 3396 if (!parse_options(data, sb, NULL, &journal_ioprio,
3397 &n_blocks_count, 1)) {
ac27a0ec
DK		 3398 err = -EINVAL;
3399 goto restore_opts;
3400 }
3401
617ba13b 3402 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
46e665e9 3403 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 3404
3405 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 3406 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3407
3408 es = sbi->s_es;
3409
b3881f74 3410 if (sbi->s_journal) {
0390131b 3411 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 3412 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3413 }
ac27a0ec
DK		 3414
3415 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3416 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 3417 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 3418 err = -EROFS;
3419 goto restore_opts;
3420 }
3421
3422 if (*flags & MS_RDONLY) {
3423 /*
3424 * First of all, the unconditional stuff we have to do
3425 * to disable replay of the journal when we next remount
3426 */
3427 sb->s_flags |= MS_RDONLY;
3428
3429 /*
3430 * OK, test if we are remounting a valid rw partition
3431 * readonly, and if so set the rdonly flag and then
3432 * mark the partition as valid again.
3433 */
617ba13b
MC		 3434 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3435 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3436 es->s_state = cpu_to_le16(sbi->s_mount_state);
3437
a63c9eb2 3438 if (sbi->s_journal)
0390131b 3439 ext4_mark_recovery_complete(sb, es);
ac27a0ec 3440 } else {
3a06d778 3441 int ret;
617ba13b
MC		 3442 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3443 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3444 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec 3445 "remount RDWR because of unsupported "
3a06d778
AK		 3446 "optional features (%x).\n", sb->s_id,
3447 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3448 ~EXT4_FEATURE_RO_COMPAT_SUPP));
ac27a0ec
DK		 3449 err = -EROFS;
3450 goto restore_opts;
3451 }
ead6596b 3452
8a266467
TT		 3453 /*
3454 * Make sure the group descriptor checksums
3455 * are sane. If they aren't, refuse to
3456 * remount r/w.
3457 */
3458 for (g = 0; g < sbi->s_groups_count; g++) {
3459 struct ext4_group_desc *gdp =
3460 ext4_get_group_desc(sb, g, NULL);
3461
3462 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3463 printk(KERN_ERR
3464 "EXT4-fs: ext4_remount: "
a9df9a49 3465 "Checksum for group %u failed (%u!=%u)\n",
8a266467
TT		 3466 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3467 le16_to_cpu(gdp->bg_checksum));
3468 err = -EINVAL;
3469 goto restore_opts;
3470 }
3471 }
3472
ead6596b
ES		 3473 /*
3474 * If we have an unprocessed orphan list hanging
3475 * around from a previously readonly bdev mount,
3476 * require a full umount/remount for now.
3477 */
3478 if (es->s_last_orphan) {
3479 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3480 "remount RDWR because of unprocessed "
3481 "orphan inode list. Please "
3482 "umount/remount instead.\n",
3483 sb->s_id);
3484 err = -EINVAL;
3485 goto restore_opts;
3486 }
3487
ac27a0ec
DK		 3488 /*
3489 * Mounting a RDONLY partition read-write, so reread
3490 * and store the current valid flag. (It may have
3491 * been changed by e2fsck since we originally mounted
3492 * the partition.)
3493 */
0390131b
FM		 3494 if (sbi->s_journal)
3495 ext4_clear_journal_err(sb, es);
ac27a0ec 3496 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3497 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3498 goto restore_opts;
2b2d6d01 3499 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3500 sb->s_flags &= ~MS_RDONLY;
3501 }
3502 }
0390131b 3503 if (sbi->s_journal == NULL)
e2d67052 3504 ext4_commit_super(sb, 1);
0390131b 3505
ac27a0ec
DK		 3506#ifdef CONFIG_QUOTA
3507 /* Release old quota file names */
3508 for (i = 0; i < MAXQUOTAS; i++)
3509 if (old_opts.s_qf_names[i] &&
3510 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3511 kfree(old_opts.s_qf_names[i]);
3512#endif
3513 return 0;
3514restore_opts:
3515 sb->s_flags = old_sb_flags;
3516 sbi->s_mount_opt = old_opts.s_mount_opt;
3517 sbi->s_resuid = old_opts.s_resuid;
3518 sbi->s_resgid = old_opts.s_resgid;
3519 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 3520 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3521 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 3522#ifdef CONFIG_QUOTA
3523 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3524 for (i = 0; i < MAXQUOTAS; i++) {
3525 if (sbi->s_qf_names[i] &&
3526 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3527 kfree(sbi->s_qf_names[i]);
3528 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3529 }
3530#endif
3531 return err;
3532}
3533
2b2d6d01 3534static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3535{
3536 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3537 struct ext4_sb_info *sbi = EXT4_SB(sb);
3538 struct ext4_super_block *es = sbi->s_es;
960cc398 3539 u64 fsid;
ac27a0ec 3540
5e70030d
BP		 3541 if (test_opt(sb, MINIX_DF)) {
3542 sbi->s_overhead_last = 0;
6bc9feff 3543 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 3544 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 3545 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3546
3547 /*
5e70030d
BP		 3548 * Compute the overhead (FS structures). This is constant
3549 * for a given filesystem unless the number of block groups
3550 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3551 */
3552
3553 /*
3554 * All of the blocks before first_data_block are
3555 * overhead
3556 */
3557 overhead = le32_to_cpu(es->s_first_data_block);
3558
3559 /*
3560 * Add the overhead attributed to the superblock and
3561 * block group descriptors. If the sparse superblocks
3562 * feature is turned on, then not all groups have this.
3563 */
3564 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3565 overhead += ext4_bg_has_super(sb, i) +
3566 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3567 cond_resched();
3568 }
3569
3570 /*
3571 * Every block group has an inode bitmap, a block
3572 * bitmap, and an inode table.
3573 */
5e70030d
BP		 3574 overhead += ngroups * (2 + sbi->s_itb_per_group);
3575 sbi->s_overhead_last = overhead;
3576 smp_wmb();
6bc9feff 3577 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3578 }
3579
617ba13b 3580 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3581 buf->f_bsize = sb->s_blocksize;
5e70030d 3582 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3583 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3584 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
308ba3ec 3585 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3586 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3587 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3588 buf->f_bavail = 0;
3589 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3590 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3591 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3592 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3593 fsid = le64_to_cpup((void *)es->s_uuid) ^
3594 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3595 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3596 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 3597 return 0;
3598}
3599
3600/* Helper function for writing quotas on sync - we need to start transaction before quota file
3601 * is locked for write. Otherwise the are possible deadlocks:
3602 * Process 1 Process 2
617ba13b 3603 * ext4_create() quota_sync()
a269eb18
JK		 3604 * jbd2_journal_start() write_dquot()
3605 * vfs_dq_init() down(dqio_mutex)
dab291af 3606 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3607 *
3608 */
3609
3610#ifdef CONFIG_QUOTA
3611
3612static inline struct inode *dquot_to_inode(struct dquot *dquot)
3613{
3614 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3615}
3616
617ba13b 3617static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3618{
3619 int ret, err;
3620 handle_t *handle;
3621 struct inode *inode;
3622
3623 inode = dquot_to_inode(dquot);
617ba13b
MC		 3624 handle = ext4_journal_start(inode,
3625 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3626 if (IS_ERR(handle))
3627 return PTR_ERR(handle);
3628 ret = dquot_commit(dquot);
617ba13b 3629 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3630 if (!ret)
3631 ret = err;
3632 return ret;
3633}
3634
617ba13b 3635static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3636{
3637 int ret, err;
3638 handle_t *handle;
3639
617ba13b
MC		 3640 handle = ext4_journal_start(dquot_to_inode(dquot),
3641 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3642 if (IS_ERR(handle))
3643 return PTR_ERR(handle);
3644 ret = dquot_acquire(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_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3652{
3653 int ret, err;
3654 handle_t *handle;
3655
617ba13b
MC		 3656 handle = ext4_journal_start(dquot_to_inode(dquot),
3657 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3658 if (IS_ERR(handle)) {
3659 /* Release dquot anyway to avoid endless cycle in dqput() */
3660 dquot_release(dquot);
ac27a0ec 3661 return PTR_ERR(handle);
9c3013e9 3662 }
ac27a0ec 3663 ret = dquot_release(dquot);
617ba13b 3664 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3665 if (!ret)
3666 ret = err;
3667 return ret;
3668}
3669
617ba13b 3670static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3671{
2c8be6b2 3672 /* Are we journaling quotas? */
617ba13b
MC		 3673 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3674 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3675 dquot_mark_dquot_dirty(dquot);
617ba13b 3676 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3677 } else {
3678 return dquot_mark_dquot_dirty(dquot);
3679 }
3680}
3681
617ba13b 3682static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3683{
3684 int ret, err;
3685 handle_t *handle;
3686
3687 /* Data block + inode block */
617ba13b 3688 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3689 if (IS_ERR(handle))
3690 return PTR_ERR(handle);
3691 ret = dquot_commit_info(sb, type);
617ba13b 3692 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3693 if (!ret)
3694 ret = err;
3695 return ret;
3696}
3697
3698/*
3699 * Turn on quotas during mount time - we need to find
3700 * the quota file and such...
3701 */
617ba13b 3702static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3703{
617ba13b
MC		 3704 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3705 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3706}
3707
3708/*
3709 * Standard function to be called on quota_on
3710 */
617ba13b 3711static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8264613d 3712 char *name, int remount)
ac27a0ec
DK		 3713{
3714 int err;
8264613d 3715 struct path path;
ac27a0ec
DK		 3716
3717 if (!test_opt(sb, QUOTA))
3718 return -EINVAL;
8264613d 3719 /* When remounting, no checks are needed and in fact, name is NULL */
0623543b 3720 if (remount)
8264613d 3721 return vfs_quota_on(sb, type, format_id, name, remount);
0623543b 3722
8264613d 3723 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 3724 if (err)
3725 return err;
0623543b 3726
ac27a0ec 3727 /* Quotafile not on the same filesystem? */
8264613d
AV		 3728 if (path.mnt->mnt_sb != sb) {
3729 path_put(&path);
ac27a0ec
DK		 3730 return -EXDEV;
3731 }
0623543b
JK		 3732 /* Journaling quota? */
3733 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3734 /* Quotafile not in fs root? */
8264613d 3735 if (path.dentry->d_parent != sb->s_root)
0623543b
JK		 3736 printk(KERN_WARNING
3737 "EXT4-fs: Quota file not on filesystem root. "
3738 "Journaled quota will not work.\n");
2b2d6d01 3739 }
0623543b
JK		 3740
3741 /*
3742 * When we journal data on quota file, we have to flush journal to see
3743 * all updates to the file when we bypass pagecache...
3744 */
0390131b
FM		 3745 if (EXT4_SB(sb)->s_journal &&
3746 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 3747 /*
3748 * We don't need to lock updates but journal_flush() could
3749 * otherwise be livelocked...
3750 */
3751 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3752 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3753 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3754 if (err) {
8264613d 3755 path_put(&path);
7ffe1ea8
HK		 3756 return err;
3757 }
0623543b
JK		 3758 }
3759
8264613d
AV		 3760 err = vfs_quota_on_path(sb, type, format_id, &path);
3761 path_put(&path);
77e69dac 3762 return err;
ac27a0ec
DK		 3763}
3764
3765/* Read data from quotafile - avoid pagecache and such because we cannot afford
3766 * acquiring the locks... As quota files are never truncated and quota code
3767 * itself serializes the operations (and noone else should touch the files)
3768 * we don't have to be afraid of races */
617ba13b 3769static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3770 size_t len, loff_t off)
3771{
3772 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3773 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3774 int err = 0;
3775 int offset = off & (sb->s_blocksize - 1);
3776 int tocopy;
3777 size_t toread;
3778 struct buffer_head *bh;
3779 loff_t i_size = i_size_read(inode);
3780
3781 if (off > i_size)
3782 return 0;
3783 if (off+len > i_size)
3784 len = i_size-off;
3785 toread = len;
3786 while (toread > 0) {
3787 tocopy = sb->s_blocksize - offset < toread ?
3788 sb->s_blocksize - offset : toread;
617ba13b 3789 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3790 if (err)
3791 return err;
3792 if (!bh) /* A hole? */
3793 memset(data, 0, tocopy);
3794 else
3795 memcpy(data, bh->b_data+offset, tocopy);
3796 brelse(bh);
3797 offset = 0;
3798 toread -= tocopy;
3799 data += tocopy;
3800 blk++;
3801 }
3802 return len;
3803}
3804
3805/* Write to quotafile (we know the transaction is already started and has
3806 * enough credits) */
617ba13b 3807static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3808 const char *data, size_t len, loff_t off)
3809{
3810 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3811 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3812 int err = 0;
3813 int offset = off & (sb->s_blocksize - 1);
3814 int tocopy;
617ba13b 3815 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3816 size_t towrite = len;
3817 struct buffer_head *bh;
3818 handle_t *handle = journal_current_handle();
3819
0390131b 3820 if (EXT4_SB(sb)->s_journal && !handle) {
e5f8eab8 3821 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
9c3013e9
JK		 3822 " cancelled because transaction is not started.\n",
3823 (unsigned long long)off, (unsigned long long)len);
3824 return -EIO;
3825 }
ac27a0ec
DK		 3826 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3827 while (towrite > 0) {
3828 tocopy = sb->s_blocksize - offset < towrite ?
3829 sb->s_blocksize - offset : towrite;
617ba13b 3830 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3831 if (!bh)
3832 goto out;
3833 if (journal_quota) {
617ba13b 3834 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3835 if (err) {
3836 brelse(bh);
3837 goto out;
3838 }
3839 }
3840 lock_buffer(bh);
3841 memcpy(bh->b_data+offset, data, tocopy);
3842 flush_dcache_page(bh->b_page);
3843 unlock_buffer(bh);
3844 if (journal_quota)
0390131b 3845 err = ext4_handle_dirty_metadata(handle, NULL, bh);
ac27a0ec
DK		 3846 else {
3847 /* Always do at least ordered writes for quotas */
678aaf48 3848 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3849 mark_buffer_dirty(bh);
3850 }
3851 brelse(bh);
3852 if (err)
3853 goto out;
3854 offset = 0;
3855 towrite -= tocopy;
3856 data += tocopy;
3857 blk++;
3858 }
3859out:
4d04e4fb
JK		 3860 if (len == towrite) {
3861 mutex_unlock(&inode->i_mutex);
ac27a0ec 3862 return err;
4d04e4fb 3863 }
ac27a0ec
DK		 3864 if (inode->i_size < off+len-towrite) {
3865 i_size_write(inode, off+len-towrite);
617ba13b 3866 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3867 }
ac27a0ec 3868 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3869 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3870 mutex_unlock(&inode->i_mutex);
3871 return len - towrite;
3872}
3873
3874#endif
3875
617ba13b 3876static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3877 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3878{
617ba13b 3879 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3880}
3881
03010a33
TT		 3882static struct file_system_type ext4_fs_type = {
3883 .owner = THIS_MODULE,
3884 .name = "ext4",
3885 .get_sb = ext4_get_sb,
3886 .kill_sb = kill_block_super,
3887 .fs_flags = FS_REQUIRES_DEV,
3888};
3889
3890#ifdef CONFIG_EXT4DEV_COMPAT
3891static int ext4dev_get_sb(struct file_system_type *fs_type,
3892 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3893{
3894 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3895 "to mount using ext4\n");
3896 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3897 "will go away by 2.6.31\n");
3898 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3899}
3900
617ba13b 3901static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3902 .owner = THIS_MODULE,
617ba13b 3903 .name = "ext4dev",
03010a33 3904 .get_sb = ext4dev_get_sb,
ac27a0ec
DK		 3905 .kill_sb = kill_block_super,
3906 .fs_flags = FS_REQUIRES_DEV,
3907};
03010a33
TT		 3908MODULE_ALIAS("ext4dev");
3909#endif
ac27a0ec 3910
617ba13b 3911static int __init init_ext4_fs(void)
ac27a0ec 3912{
c9de560d
AT		 3913 int err;
3914
3197ebdb
TT		 3915 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3916 if (!ext4_kset)
3917 return -ENOMEM;
9f6200bb 3918 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 3919 err = init_ext4_mballoc();
ac27a0ec
DK		 3920 if (err)
3921 return err;
c9de560d
AT		 3922
3923 err = init_ext4_xattr();
3924 if (err)
3925 goto out2;
ac27a0ec
DK		 3926 err = init_inodecache();
3927 if (err)
3928 goto out1;
03010a33 3929 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 3930 if (err)
3931 goto out;
03010a33
TT		 3932#ifdef CONFIG_EXT4DEV_COMPAT
3933 err = register_filesystem(&ext4dev_fs_type);
3934 if (err) {
3935 unregister_filesystem(&ext4_fs_type);
3936 goto out;
3937 }
3938#endif
ac27a0ec
DK		 3939 return 0;
3940out:
3941 destroy_inodecache();
3942out1:
617ba13b 3943 exit_ext4_xattr();
c9de560d
AT		 3944out2:
3945 exit_ext4_mballoc();
ac27a0ec
DK		 3946 return err;
3947}
3948
617ba13b 3949static void __exit exit_ext4_fs(void)
ac27a0ec 3950{
03010a33
TT		 3951 unregister_filesystem(&ext4_fs_type);
3952#ifdef CONFIG_EXT4DEV_COMPAT
617ba13b 3953 unregister_filesystem(&ext4dev_fs_type);
03010a33 3954#endif
ac27a0ec 3955 destroy_inodecache();
617ba13b 3956 exit_ext4_xattr();
c9de560d 3957 exit_ext4_mballoc();
9f6200bb 3958 remove_proc_entry("fs/ext4", NULL);
3197ebdb 3959 kset_unregister(ext4_kset);
ac27a0ec
DK		 3960}
3961
3962MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 3963MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 3964MODULE_LICENSE("GPL");
617ba13b
MC		 3965module_init(init_ext4_fs)
3966module_exit(exit_ext4_fs)
Empty description
This page took 0.977139 seconds and 4 git commands to generate.