projects / GitHub / LineageOS / android_kernel_motorola_exynos9610.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: Use is_power_of_2() for clarity
[GitHub/LineageOS/android_kernel_motorola_exynos9610.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
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);
c4be0c1d 57static int ext4_commit_super(struct super_block *sb,
2b2d6d01
TT		 58 struct ext4_super_block *es, int sync);
59static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
617ba13b 63static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 64static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 65 char nbuf[16]);
2b2d6d01
TT		 66static int ext4_remount(struct super_block *sb, int *flags, char *data);
67static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 68static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 69static void ext4_write_super(struct super_block *sb);
c4be0c1d 70static int ext4_freeze(struct super_block *sb);
ac27a0ec 71
bd81d8ee 72
8fadc143
AR		 73ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
74 struct ext4_group_desc *bg)
bd81d8ee 75{
3a14589c 76 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 77 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
3a14589c 78 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 79}
80
8fadc143
AR		 81ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
82 struct ext4_group_desc *bg)
bd81d8ee 83{
5272f837 84 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 85 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 86 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 87}
88
8fadc143
AR		 89ext4_fsblk_t ext4_inode_table(struct super_block *sb,
90 struct ext4_group_desc *bg)
bd81d8ee 91{
5272f837 92 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 93 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 94 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 95}
96
560671a0
AK		 97__u32 ext4_free_blks_count(struct super_block *sb,
98 struct ext4_group_desc *bg)
99{
100 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
101 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
102 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
103}
104
105__u32 ext4_free_inodes_count(struct super_block *sb,
106 struct ext4_group_desc *bg)
107{
108 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
109 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
110 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
111}
112
113__u32 ext4_used_dirs_count(struct super_block *sb,
114 struct ext4_group_desc *bg)
115{
116 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
117 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
118 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
119}
120
121__u32 ext4_itable_unused_count(struct super_block *sb,
122 struct ext4_group_desc *bg)
123{
124 return le16_to_cpu(bg->bg_itable_unused_lo) |
125 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
126 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
127}
128
8fadc143
AR		 129void ext4_block_bitmap_set(struct super_block *sb,
130 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 131{
3a14589c 132 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 133 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
134 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 135}
136
8fadc143
AR		 137void ext4_inode_bitmap_set(struct super_block *sb,
138 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 139{
5272f837 140 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 141 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
142 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 143}
144
8fadc143
AR		 145void ext4_inode_table_set(struct super_block *sb,
146 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 147{
5272f837 148 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 149 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
150 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 151}
152
560671a0
AK		 153void ext4_free_blks_set(struct super_block *sb,
154 struct ext4_group_desc *bg, __u32 count)
155{
156 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
157 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
158 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
159}
160
161void ext4_free_inodes_set(struct super_block *sb,
162 struct ext4_group_desc *bg, __u32 count)
163{
164 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
165 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
166 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
167}
168
169void ext4_used_dirs_set(struct super_block *sb,
170 struct ext4_group_desc *bg, __u32 count)
171{
172 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
173 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
174 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
175}
176
177void ext4_itable_unused_set(struct super_block *sb,
178 struct ext4_group_desc *bg, __u32 count)
179{
180 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
181 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
182 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
183}
184
ac27a0ec 185/*
dab291af 186 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 187 *
188 * The only special thing we need to do here is to make sure that all
189 * journal_end calls result in the superblock being marked dirty, so
190 * that sync() will call the filesystem's write_super callback if
191 * appropriate.
192 */
617ba13b 193handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 194{
195 journal_t *journal;
196
197 if (sb->s_flags & MS_RDONLY)
198 return ERR_PTR(-EROFS);
199
200 /* Special case here: if the journal has aborted behind our
201 * backs (eg. EIO in the commit thread), then we still need to
202 * take the FS itself readonly cleanly. */
617ba13b 203 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 204 if (journal) {
205 if (is_journal_aborted(journal)) {
206 ext4_abort(sb, __func__,
207 "Detected aborted journal");
208 return ERR_PTR(-EROFS);
209 }
210 return jbd2_journal_start(journal, nblocks);
ac27a0ec 211 }
0390131b
FM		 212 /*
213 * We're not journaling, return the appropriate indication.
214 */
215 current->journal_info = EXT4_NOJOURNAL_HANDLE;
216 return current->journal_info;
ac27a0ec
DK		 217}
218
219/*
220 * The only special thing we need to do here is to make sure that all
dab291af 221 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 222 * that sync() will call the filesystem's write_super callback if
223 * appropriate.
224 */
617ba13b 225int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 226{
227 struct super_block *sb;
228 int err;
229 int rc;
230
0390131b
FM		 231 if (!ext4_handle_valid(handle)) {
232 /*
233 * Do this here since we don't call jbd2_journal_stop() in
234 * no-journal mode.
235 */
236 current->journal_info = NULL;
237 return 0;
238 }
ac27a0ec
DK		 239 sb = handle->h_transaction->t_journal->j_private;
240 err = handle->h_err;
dab291af 241 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 242
243 if (!err)
244 err = rc;
245 if (err)
617ba13b 246 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 247 return err;
248}
249
617ba13b 250void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 251 struct buffer_head *bh, handle_t *handle, int err)
252{
253 char nbuf[16];
617ba13b 254 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 255
0390131b
FM		 256 BUG_ON(!ext4_handle_valid(handle));
257
ac27a0ec
DK		 258 if (bh)
259 BUFFER_TRACE(bh, "abort");
260
261 if (!handle->h_err)
262 handle->h_err = err;
263
264 if (is_handle_aborted(handle))
265 return;
266
267 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
268 caller, errstr, err_fn);
269
dab291af 270 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 271}
272
273/* Deal with the reporting of failure conditions on a filesystem such as
274 * inconsistencies detected or read IO failures.
275 *
276 * On ext2, we can store the error state of the filesystem in the
617ba13b 277 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 278 * write ordering constraints on the superblock which prevent us from
279 * writing it out straight away; and given that the journal is about to
280 * be aborted, we can't rely on the current, or future, transactions to
281 * write out the superblock safely.
282 *
dab291af 283 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 284 * the journal instead. On recovery, the journal will compain about
285 * that error until we've noted it down and cleared it.
286 */
287
617ba13b 288static void ext4_handle_error(struct super_block *sb)
ac27a0ec 289{
617ba13b 290 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 291
617ba13b
MC		 292 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
293 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 294
295 if (sb->s_flags & MS_RDONLY)
296 return;
297
2b2d6d01 298 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 299 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 300
617ba13b 301 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ac27a0ec 302 if (journal)
dab291af 303 jbd2_journal_abort(journal, -EIO);
ac27a0ec 304 }
2b2d6d01
TT		 305 if (test_opt(sb, ERRORS_RO)) {
306 printk(KERN_CRIT "Remounting filesystem read-only\n");
ac27a0ec
DK		 307 sb->s_flags |= MS_RDONLY;
308 }
617ba13b 309 ext4_commit_super(sb, es, 1);
ac27a0ec 310 if (test_opt(sb, ERRORS_PANIC))
617ba13b 311 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 312 sb->s_id);
313}
314
2b2d6d01
TT		 315void ext4_error(struct super_block *sb, const char *function,
316 const char *fmt, ...)
ac27a0ec
DK		 317{
318 va_list args;
319
320 va_start(args, fmt);
2b2d6d01 321 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 322 vprintk(fmt, args);
323 printk("\n");
324 va_end(args);
325
617ba13b 326 ext4_handle_error(sb);
ac27a0ec
DK		 327}
328
2b2d6d01 329static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 330 char nbuf[16])
331{
332 char *errstr = NULL;
333
334 switch (errno) {
335 case -EIO:
336 errstr = "IO failure";
337 break;
338 case -ENOMEM:
339 errstr = "Out of memory";
340 break;
341 case -EROFS:
dab291af 342 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
ac27a0ec
DK		 343 errstr = "Journal has aborted";
344 else
345 errstr = "Readonly filesystem";
346 break;
347 default:
348 /* If the caller passed in an extra buffer for unknown
349 * errors, textualise them now. Else we just return
350 * NULL. */
351 if (nbuf) {
352 /* Check for truncated error codes... */
353 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
354 errstr = nbuf;
355 }
356 break;
357 }
358
359 return errstr;
360}
361
617ba13b 362/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 363 * automatically and invokes the appropriate error response. */
364
2b2d6d01 365void __ext4_std_error(struct super_block *sb, const char *function, int errno)
ac27a0ec
DK		 366{
367 char nbuf[16];
368 const char *errstr;
369
370 /* Special case: if the error is EROFS, and we're not already
371 * inside a transaction, then there's really no point in logging
372 * an error. */
373 if (errno == -EROFS && journal_current_handle() == NULL &&
374 (sb->s_flags & MS_RDONLY))
375 return;
376
617ba13b 377 errstr = ext4_decode_error(sb, errno, nbuf);
2b2d6d01
TT		 378 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
379 sb->s_id, function, errstr);
ac27a0ec 380
617ba13b 381 ext4_handle_error(sb);
ac27a0ec
DK		 382}
383
384/*
617ba13b 385 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 386 * abort function may be used to deal with unrecoverable failures such
387 * as journal IO errors or ENOMEM at a critical moment in log management.
388 *
389 * We unconditionally force the filesystem into an ABORT|READONLY state,
390 * unless the error response on the fs has been set to panic in which
391 * case we take the easy way out and panic immediately.
392 */
393
2b2d6d01
TT		 394void ext4_abort(struct super_block *sb, const char *function,
395 const char *fmt, ...)
ac27a0ec
DK		 396{
397 va_list args;
398
2b2d6d01 399 printk(KERN_CRIT "ext4_abort called.\n");
ac27a0ec
DK		 400
401 va_start(args, fmt);
2b2d6d01 402 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 403 vprintk(fmt, args);
404 printk("\n");
405 va_end(args);
406
407 if (test_opt(sb, ERRORS_PANIC))
617ba13b 408 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 409
410 if (sb->s_flags & MS_RDONLY)
411 return;
412
413 printk(KERN_CRIT "Remounting filesystem read-only\n");
617ba13b 414 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 415 sb->s_flags |= MS_RDONLY;
617ba13b 416 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ef2cabf7
HK		 417 if (EXT4_SB(sb)->s_journal)
418 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 419}
420
2b2d6d01
TT		 421void ext4_warning(struct super_block *sb, const char *function,
422 const char *fmt, ...)
ac27a0ec
DK		 423{
424 va_list args;
425
426 va_start(args, fmt);
617ba13b 427 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 428 sb->s_id, function);
429 vprintk(fmt, args);
430 printk("\n");
431 va_end(args);
432}
433
5d1b1b3f
AK		 434void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
435 const char *function, const char *fmt, ...)
436__releases(bitlock)
437__acquires(bitlock)
438{
439 va_list args;
440 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
441
442 va_start(args, fmt);
443 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
444 vprintk(fmt, args);
445 printk("\n");
446 va_end(args);
447
448 if (test_opt(sb, ERRORS_CONT)) {
449 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
450 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
451 ext4_commit_super(sb, es, 0);
452 return;
453 }
454 ext4_unlock_group(sb, grp);
455 ext4_handle_error(sb);
456 /*
457 * We only get here in the ERRORS_RO case; relocking the group
458 * may be dangerous, but nothing bad will happen since the
459 * filesystem will have already been marked read/only and the
460 * journal has been aborted. We return 1 as a hint to callers
461 * who might what to use the return value from
462 * ext4_grp_locked_error() to distinguish beween the
463 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
464 * aggressively from the ext4 function in question, with a
465 * more appropriate error code.
466 */
467 ext4_lock_group(sb, grp);
468 return;
469}
470
471
617ba13b 472void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 473{
617ba13b 474 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 475
617ba13b 476 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 477 return;
478
46e665e9 479 ext4_warning(sb, __func__,
ac27a0ec
DK		 480 "updating to rev %d because of new feature flag, "
481 "running e2fsck is recommended",
617ba13b 482 EXT4_DYNAMIC_REV);
ac27a0ec 483
617ba13b
MC		 484 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
485 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
486 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 487 /* leave es->s_feature_*compat flags alone */
488 /* es->s_uuid will be set by e2fsck if empty */
489
490 /*
491 * The rest of the superblock fields should be zero, and if not it
492 * means they are likely already in use, so leave them alone. We
493 * can leave it up to e2fsck to clean up any inconsistencies there.
494 */
495}
496
497/*
498 * Open the external journal device
499 */
617ba13b 500static struct block_device *ext4_blkdev_get(dev_t dev)
ac27a0ec
DK		 501{
502 struct block_device *bdev;
503 char b[BDEVNAME_SIZE];
504
505 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
506 if (IS_ERR(bdev))
507 goto fail;
508 return bdev;
509
510fail:
abda1418 511 printk(KERN_ERR "EXT4-fs: failed to open journal device %s: %ld\n",
ac27a0ec
DK		 512 __bdevname(dev, b), PTR_ERR(bdev));
513 return NULL;
514}
515
516/*
517 * Release the journal device
518 */
617ba13b 519static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 520{
521 bd_release(bdev);
9a1c3542 522 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 523}
524
617ba13b 525static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 526{
527 struct block_device *bdev;
528 int ret = -ENODEV;
529
530 bdev = sbi->journal_bdev;
531 if (bdev) {
617ba13b 532 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 533 sbi->journal_bdev = NULL;
534 }
535 return ret;
536}
537
538static inline struct inode *orphan_list_entry(struct list_head *l)
539{
617ba13b 540 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 541}
542
617ba13b 543static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 544{
545 struct list_head *l;
546
547 printk(KERN_ERR "sb orphan head is %d\n",
548 le32_to_cpu(sbi->s_es->s_last_orphan));
549
550 printk(KERN_ERR "sb_info orphan list:\n");
551 list_for_each(l, &sbi->s_orphan) {
552 struct inode *inode = orphan_list_entry(l);
553 printk(KERN_ERR " "
554 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
555 inode->i_sb->s_id, inode->i_ino, inode,
556 inode->i_mode, inode->i_nlink,
557 NEXT_ORPHAN(inode));
558 }
559}
560
2b2d6d01 561static void ext4_put_super(struct super_block *sb)
ac27a0ec 562{
617ba13b
MC		 563 struct ext4_sb_info *sbi = EXT4_SB(sb);
564 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 565 int i, err;
ac27a0ec 566
c9de560d 567 ext4_mb_release(sb);
a86c6181 568 ext4_ext_release(sb);
617ba13b 569 ext4_xattr_put_super(sb);
0390131b
FM		 570 if (sbi->s_journal) {
571 err = jbd2_journal_destroy(sbi->s_journal);
572 sbi->s_journal = NULL;
573 if (err < 0)
574 ext4_abort(sb, __func__,
575 "Couldn't clean up the journal");
576 }
ac27a0ec 577 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 578 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 579 es->s_state = cpu_to_le16(sbi->s_mount_state);
617ba13b 580 ext4_commit_super(sb, es, 1);
ac27a0ec 581 }
240799cd 582 if (sbi->s_proc) {
9f6200bb 583 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 584 }
3197ebdb 585 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 586
587 for (i = 0; i < sbi->s_gdb_count; i++)
588 brelse(sbi->s_group_desc[i]);
589 kfree(sbi->s_group_desc);
c5ca7c76
TT		 590 if (is_vmalloc_addr(sbi->s_flex_groups))
591 vfree(sbi->s_flex_groups);
592 else
593 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 594 percpu_counter_destroy(&sbi->s_freeblocks_counter);
595 percpu_counter_destroy(&sbi->s_freeinodes_counter);
596 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 597 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 598 brelse(sbi->s_sbh);
599#ifdef CONFIG_QUOTA
600 for (i = 0; i < MAXQUOTAS; i++)
601 kfree(sbi->s_qf_names[i]);
602#endif
603
604 /* Debugging code just in case the in-memory inode orphan list
605 * isn't empty. The on-disk one can be non-empty if we've
606 * detected an error and taken the fs readonly, but the
607 * in-memory list had better be clean by this point. */
608 if (!list_empty(&sbi->s_orphan))
609 dump_orphan_list(sb, sbi);
610 J_ASSERT(list_empty(&sbi->s_orphan));
611
f98393a6 612 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 613 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
614 /*
615 * Invalidate the journal device's buffers. We don't want them
616 * floating about in memory - the physical journal device may
617 * hotswapped, and it breaks the `ro-after' testing code.
618 */
619 sync_blockdev(sbi->journal_bdev);
f98393a6 620 invalidate_bdev(sbi->journal_bdev);
617ba13b 621 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 622 }
623 sb->s_fs_info = NULL;
3197ebdb
TT		 624 /*
625 * Now that we are completely done shutting down the
626 * superblock, we need to actually destroy the kobject.
627 */
628 unlock_kernel();
629 unlock_super(sb);
630 kobject_put(&sbi->s_kobj);
631 wait_for_completion(&sbi->s_kobj_unregister);
632 lock_super(sb);
633 lock_kernel();
705895b6 634 kfree(sbi->s_blockgroup_lock);
ac27a0ec
DK		 635 kfree(sbi);
636 return;
637}
638
e18b890b 639static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 640
641/*
642 * Called inside transaction, so use GFP_NOFS
643 */
617ba13b 644static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 645{
617ba13b 646 struct ext4_inode_info *ei;
ac27a0ec 647
e6b4f8da 648 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 649 if (!ei)
650 return NULL;
03010a33 651#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b
MC		 652 ei->i_acl = EXT4_ACL_NOT_CACHED;
653 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec 654#endif
ac27a0ec 655 ei->vfs_inode.i_version = 1;
91246c00 656 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 657 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 658 INIT_LIST_HEAD(&ei->i_prealloc_list);
659 spin_lock_init(&ei->i_prealloc_lock);
0390131b
FM		 660 /*
661 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
662 * therefore it can be null here. Don't check it, just initialize
663 * jinode.
664 */
678aaf48 665 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 666 ei->i_reserved_data_blocks = 0;
667 ei->i_reserved_meta_blocks = 0;
668 ei->i_allocated_meta_blocks = 0;
669 ei->i_delalloc_reserved_flag = 0;
670 spin_lock_init(&(ei->i_block_reservation_lock));
ac27a0ec
DK		 671 return &ei->vfs_inode;
672}
673
617ba13b 674static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 675{
9f7dd93d
VA		 676 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
677 printk("EXT4 Inode %p: orphan list check failed!\n",
678 EXT4_I(inode));
679 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
680 EXT4_I(inode), sizeof(struct ext4_inode_info),
681 true);
682 dump_stack();
683 }
617ba13b 684 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 685}
686
51cc5068 687static void init_once(void *foo)
ac27a0ec 688{
617ba13b 689 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 690
a35afb83 691 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 692#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 693 init_rwsem(&ei->xattr_sem);
ac27a0ec 694#endif
0e855ac8 695 init_rwsem(&ei->i_data_sem);
a35afb83 696 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 697}
698
699static int init_inodecache(void)
700{
617ba13b
MC		 701 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
702 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 703 0, (SLAB_RECLAIM_ACCOUNT|
704 SLAB_MEM_SPREAD),
20c2df83 705 init_once);
617ba13b 706 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 707 return -ENOMEM;
708 return 0;
709}
710
711static void destroy_inodecache(void)
712{
617ba13b 713 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 714}
715
617ba13b 716static void ext4_clear_inode(struct inode *inode)
ac27a0ec 717{
03010a33 718#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b
MC		 719 if (EXT4_I(inode)->i_acl &&
720 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
721 posix_acl_release(EXT4_I(inode)->i_acl);
722 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
723 }
724 if (EXT4_I(inode)->i_default_acl &&
725 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
726 posix_acl_release(EXT4_I(inode)->i_default_acl);
727 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 728 }
729#endif
c2ea3fde 730 ext4_discard_preallocations(inode);
0390131b
FM		 731 if (EXT4_JOURNAL(inode))
732 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
678aaf48 733 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 734}
735
2b2d6d01
TT		 736static inline void ext4_show_quota_options(struct seq_file *seq,
737 struct super_block *sb)
ac27a0ec
DK		 738{
739#if defined(CONFIG_QUOTA)
617ba13b 740 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 741
742 if (sbi->s_jquota_fmt)
743 seq_printf(seq, ",jqfmt=%s",
af5bc92d 744 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
ac27a0ec
DK		 745
746 if (sbi->s_qf_names[USRQUOTA])
747 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
748
749 if (sbi->s_qf_names[GRPQUOTA])
750 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
751
617ba13b 752 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 753 seq_puts(seq, ",usrquota");
754
617ba13b 755 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 756 seq_puts(seq, ",grpquota");
757#endif
758}
759
d9c9bef1
MS		 760/*
761 * Show an option if
762 * - it's set to a non-default value OR
763 * - if the per-sb default is different from the global default
764 */
617ba13b 765static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 766{
aa22df2c
AK		 767 int def_errors;
768 unsigned long def_mount_opts;
ac27a0ec 769 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 770 struct ext4_sb_info *sbi = EXT4_SB(sb);
771 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 772
773 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 774 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 775
776 if (sbi->s_sb_block != 1)
777 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
778 if (test_opt(sb, MINIX_DF))
779 seq_puts(seq, ",minixdf");
aa22df2c 780 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 781 seq_puts(seq, ",grpid");
782 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
783 seq_puts(seq, ",nogrpid");
784 if (sbi->s_resuid != EXT4_DEF_RESUID ||
785 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
786 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
787 }
788 if (sbi->s_resgid != EXT4_DEF_RESGID ||
789 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
790 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
791 }
bb4f397a 792 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 793 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 794 def_errors == EXT4_ERRORS_CONTINUE) {
795 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 796 }
797 }
aa22df2c 798 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 799 seq_puts(seq, ",errors=continue");
aa22df2c 800 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 801 seq_puts(seq, ",errors=panic");
aa22df2c 802 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 803 seq_puts(seq, ",nouid32");
aa22df2c 804 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 805 seq_puts(seq, ",debug");
806 if (test_opt(sb, OLDALLOC))
807 seq_puts(seq, ",oldalloc");
03010a33 808#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 809 if (test_opt(sb, XATTR_USER) &&
810 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 811 seq_puts(seq, ",user_xattr");
812 if (!test_opt(sb, XATTR_USER) &&
813 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
814 seq_puts(seq, ",nouser_xattr");
815 }
816#endif
03010a33 817#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 818 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 819 seq_puts(seq, ",acl");
820 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
821 seq_puts(seq, ",noacl");
822#endif
30773840 823 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 824 seq_printf(seq, ",commit=%u",
825 (unsigned) (sbi->s_commit_interval / HZ));
826 }
30773840
TT		 827 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
828 seq_printf(seq, ",min_batch_time=%u",
829 (unsigned) sbi->s_min_batch_time);
830 }
831 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
832 seq_printf(seq, ",max_batch_time=%u",
833 (unsigned) sbi->s_min_batch_time);
834 }
835
571640ca
ES		 836 /*
837 * We're changing the default of barrier mount option, so
838 * let's always display its mount state so it's clear what its
839 * status is.
840 */
841 seq_puts(seq, ",barrier=");
842 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 843 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
844 seq_puts(seq, ",journal_async_commit");
d9c9bef1
MS		 845 if (test_opt(sb, NOBH))
846 seq_puts(seq, ",nobh");
25ec56b5
JNC		 847 if (test_opt(sb, I_VERSION))
848 seq_puts(seq, ",i_version");
dd919b98
AK		 849 if (!test_opt(sb, DELALLOC))
850 seq_puts(seq, ",nodelalloc");
851
ac27a0ec 852
cb45bbe4
MS		 853 if (sbi->s_stripe)
854 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 855 /*
856 * journal mode get enabled in different ways
857 * So just print the value even if we didn't specify it
858 */
617ba13b 859 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 860 seq_puts(seq, ",data=journal");
617ba13b 861 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 862 seq_puts(seq, ",data=ordered");
617ba13b 863 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 864 seq_puts(seq, ",data=writeback");
865
240799cd
TT		 866 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
867 seq_printf(seq, ",inode_readahead_blks=%u",
868 sbi->s_inode_readahead_blks);
869
5bf5683a
HK		 870 if (test_opt(sb, DATA_ERR_ABORT))
871 seq_puts(seq, ",data_err=abort");
872
afd4672d 873 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 874 seq_puts(seq, ",noauto_da_alloc");
afd4672d 875
617ba13b 876 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 877 return 0;
878}
879
880
1b961ac0
CH		 881static struct inode *ext4_nfs_get_inode(struct super_block *sb,
882 u64 ino, u32 generation)
ac27a0ec 883{
ac27a0ec 884 struct inode *inode;
ac27a0ec 885
617ba13b 886 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 887 return ERR_PTR(-ESTALE);
617ba13b 888 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 889 return ERR_PTR(-ESTALE);
890
891 /* iget isn't really right if the inode is currently unallocated!!
892 *
617ba13b 893 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 894 * deleted, so we should be safe.
895 *
896 * Currently we don't know the generation for parent directory, so
897 * a generation of 0 means "accept any"
898 */
1d1fe1ee
DH		 899 inode = ext4_iget(sb, ino);
900 if (IS_ERR(inode))
901 return ERR_CAST(inode);
902 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 903 iput(inode);
904 return ERR_PTR(-ESTALE);
905 }
1b961ac0
CH		 906
907 return inode;
908}
909
910static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
911 int fh_len, int fh_type)
912{
913 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
914 ext4_nfs_get_inode);
915}
916
917static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
918 int fh_len, int fh_type)
919{
920 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
921 ext4_nfs_get_inode);
ac27a0ec
DK		 922}
923
c39a7f84
TO		 924/*
925 * Try to release metadata pages (indirect blocks, directories) which are
926 * mapped via the block device. Since these pages could have journal heads
927 * which would prevent try_to_free_buffers() from freeing them, we must use
928 * jbd2 layer's try_to_free_buffers() function to release them.
929 */
930static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
931{
932 journal_t *journal = EXT4_SB(sb)->s_journal;
933
934 WARN_ON(PageChecked(page));
935 if (!page_has_buffers(page))
936 return 0;
937 if (journal)
938 return jbd2_journal_try_to_free_buffers(journal, page,
939 wait & ~__GFP_WAIT);
940 return try_to_free_buffers(page);
941}
942
ac27a0ec 943#ifdef CONFIG_QUOTA
af5bc92d 944#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 945#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 946
617ba13b
MC		 947static int ext4_write_dquot(struct dquot *dquot);
948static int ext4_acquire_dquot(struct dquot *dquot);
949static int ext4_release_dquot(struct dquot *dquot);
950static int ext4_mark_dquot_dirty(struct dquot *dquot);
951static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 952static int ext4_quota_on(struct super_block *sb, int type, int format_id,
953 char *path, int remount);
617ba13b
MC		 954static int ext4_quota_on_mount(struct super_block *sb, int type);
955static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 956 size_t len, loff_t off);
617ba13b 957static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 958 const char *data, size_t len, loff_t off);
959
617ba13b 960static struct dquot_operations ext4_quota_operations = {
edf72453
JK		 961 .initialize = dquot_initialize,
962 .drop = dquot_drop,
ac27a0ec 963 .alloc_space = dquot_alloc_space,
60e58e0f
MC		 964 .reserve_space = dquot_reserve_space,
965 .claim_space = dquot_claim_space,
966 .release_rsv = dquot_release_reserved_space,
967 .get_reserved_space = ext4_get_reserved_space,
ac27a0ec
DK		 968 .alloc_inode = dquot_alloc_inode,
969 .free_space = dquot_free_space,
970 .free_inode = dquot_free_inode,
971 .transfer = dquot_transfer,
617ba13b
MC		 972 .write_dquot = ext4_write_dquot,
973 .acquire_dquot = ext4_acquire_dquot,
974 .release_dquot = ext4_release_dquot,
975 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 976 .write_info = ext4_write_info,
977 .alloc_dquot = dquot_alloc,
978 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 979};
980
617ba13b
MC		 981static struct quotactl_ops ext4_qctl_operations = {
982 .quota_on = ext4_quota_on,
ac27a0ec
DK		 983 .quota_off = vfs_quota_off,
984 .quota_sync = vfs_quota_sync,
985 .get_info = vfs_get_dqinfo,
986 .set_info = vfs_set_dqinfo,
987 .get_dqblk = vfs_get_dqblk,
988 .set_dqblk = vfs_set_dqblk
989};
990#endif
991
ee9b6d61 992static const struct super_operations ext4_sops = {
617ba13b
MC		 993 .alloc_inode = ext4_alloc_inode,
994 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 995 .write_inode = ext4_write_inode,
996 .dirty_inode = ext4_dirty_inode,
997 .delete_inode = ext4_delete_inode,
998 .put_super = ext4_put_super,
999 .write_super = ext4_write_super,
1000 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1001 .freeze_fs = ext4_freeze,
1002 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1003 .statfs = ext4_statfs,
1004 .remount_fs = ext4_remount,
1005 .clear_inode = ext4_clear_inode,
1006 .show_options = ext4_show_options,
ac27a0ec 1007#ifdef CONFIG_QUOTA
617ba13b
MC		 1008 .quota_read = ext4_quota_read,
1009 .quota_write = ext4_quota_write,
ac27a0ec 1010#endif
c39a7f84 1011 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1012};
1013
39655164 1014static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1015 .fh_to_dentry = ext4_fh_to_dentry,
1016 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1017 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1018};
1019
1020enum {
1021 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1022 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1023 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1024 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1025 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1026 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1027 Opt_journal_update, Opt_journal_dev,
818d276c 1028 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1029 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
5bf5683a 1030 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec
DK		 1031 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1032 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
06705bff
TT		 1033 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1034 Opt_usrquota, Opt_grpquota, Opt_i_version,
01436ef2 1035 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
b3881f74 1036 Opt_inode_readahead_blks, Opt_journal_ioprio
ac27a0ec
DK		 1037};
1038
a447c093 1039static const match_table_t tokens = {
ac27a0ec
DK		 1040 {Opt_bsd_df, "bsddf"},
1041 {Opt_minix_df, "minixdf"},
1042 {Opt_grpid, "grpid"},
1043 {Opt_grpid, "bsdgroups"},
1044 {Opt_nogrpid, "nogrpid"},
1045 {Opt_nogrpid, "sysvgroups"},
1046 {Opt_resgid, "resgid=%u"},
1047 {Opt_resuid, "resuid=%u"},
1048 {Opt_sb, "sb=%u"},
1049 {Opt_err_cont, "errors=continue"},
1050 {Opt_err_panic, "errors=panic"},
1051 {Opt_err_ro, "errors=remount-ro"},
1052 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1053 {Opt_debug, "debug"},
1054 {Opt_oldalloc, "oldalloc"},
1055 {Opt_orlov, "orlov"},
1056 {Opt_user_xattr, "user_xattr"},
1057 {Opt_nouser_xattr, "nouser_xattr"},
1058 {Opt_acl, "acl"},
1059 {Opt_noacl, "noacl"},
ac27a0ec
DK		 1060 {Opt_noload, "noload"},
1061 {Opt_nobh, "nobh"},
1062 {Opt_bh, "bh"},
1063 {Opt_commit, "commit=%u"},
30773840
TT		 1064 {Opt_min_batch_time, "min_batch_time=%u"},
1065 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1066 {Opt_journal_update, "journal=update"},
ac27a0ec 1067 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1068 {Opt_journal_checksum, "journal_checksum"},
1069 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1070 {Opt_abort, "abort"},
1071 {Opt_data_journal, "data=journal"},
1072 {Opt_data_ordered, "data=ordered"},
1073 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1074 {Opt_data_err_abort, "data_err=abort"},
1075 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1076 {Opt_offusrjquota, "usrjquota="},
1077 {Opt_usrjquota, "usrjquota=%s"},
1078 {Opt_offgrpjquota, "grpjquota="},
1079 {Opt_grpjquota, "grpjquota=%s"},
1080 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1081 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1082 {Opt_grpquota, "grpquota"},
1083 {Opt_noquota, "noquota"},
1084 {Opt_quota, "quota"},
1085 {Opt_usrquota, "usrquota"},
1086 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1087 {Opt_barrier, "barrier"},
1088 {Opt_nobarrier, "nobarrier"},
25ec56b5 1089 {Opt_i_version, "i_version"},
c9de560d 1090 {Opt_stripe, "stripe=%u"},
ac27a0ec 1091 {Opt_resize, "resize"},
64769240 1092 {Opt_delalloc, "delalloc"},
dd919b98 1093 {Opt_nodelalloc, "nodelalloc"},
240799cd 1094 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1095 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1096 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1097 {Opt_auto_da_alloc, "auto_da_alloc"},
1098 {Opt_noauto_da_alloc, "noauto_da_alloc"},
f3f12faa 1099 {Opt_err, NULL},
ac27a0ec
DK		 1100};
1101
617ba13b 1102static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1103{
617ba13b 1104 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1105 char *options = (char *) *data;
1106
1107 if (!options || strncmp(options, "sb=", 3) != 0)
1108 return 1; /* Default location */
1109 options += 3;
617ba13b 1110 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1111 sb_block = simple_strtoul(options, &options, 0);
1112 if (*options && *options != ',') {
4776004f 1113 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1114 (char *) *data);
1115 return 1;
1116 }
1117 if (*options == ',')
1118 options++;
1119 *data = (void *) options;
1120 return sb_block;
1121}
1122
b3881f74
TT		 1123#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1124
2b2d6d01 1125static int parse_options(char *options, struct super_block *sb,
c3191067 1126 unsigned long *journal_devnum,
b3881f74 1127 unsigned int *journal_ioprio,
2b2d6d01 1128 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1129{
617ba13b 1130 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1131 char *p;
ac27a0ec
DK		 1132 substring_t args[MAX_OPT_ARGS];
1133 int data_opt = 0;
1134 int option;
1135#ifdef CONFIG_QUOTA
dfc5d03f 1136 int qtype, qfmt;
ac27a0ec
DK		 1137 char *qname;
1138#endif
1139
1140 if (!options)
1141 return 1;
1142
2b2d6d01 1143 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1144 int token;
1145 if (!*p)
1146 continue;
1147
1148 token = match_token(p, tokens, args);
1149 switch (token) {
1150 case Opt_bsd_df:
2b2d6d01 1151 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1152 break;
1153 case Opt_minix_df:
2b2d6d01 1154 set_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1155 break;
1156 case Opt_grpid:
2b2d6d01 1157 set_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1158 break;
1159 case Opt_nogrpid:
2b2d6d01 1160 clear_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1161 break;
1162 case Opt_resuid:
1163 if (match_int(&args[0], &option))
1164 return 0;
1165 sbi->s_resuid = option;
1166 break;
1167 case Opt_resgid:
1168 if (match_int(&args[0], &option))
1169 return 0;
1170 sbi->s_resgid = option;
1171 break;
1172 case Opt_sb:
1173 /* handled by get_sb_block() instead of here */
1174 /* *sb_block = match_int(&args[0]); */
1175 break;
1176 case Opt_err_panic:
2b2d6d01
TT		 1177 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1178 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1179 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1180 break;
1181 case Opt_err_ro:
2b2d6d01
TT		 1182 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1183 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1184 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1185 break;
1186 case Opt_err_cont:
2b2d6d01
TT		 1187 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1188 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1189 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1190 break;
1191 case Opt_nouid32:
2b2d6d01 1192 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1193 break;
ac27a0ec 1194 case Opt_debug:
2b2d6d01 1195 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1196 break;
1197 case Opt_oldalloc:
2b2d6d01 1198 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1199 break;
1200 case Opt_orlov:
2b2d6d01 1201 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1202 break;
03010a33 1203#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1204 case Opt_user_xattr:
2b2d6d01 1205 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1206 break;
1207 case Opt_nouser_xattr:
2b2d6d01 1208 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1209 break;
1210#else
1211 case Opt_user_xattr:
1212 case Opt_nouser_xattr:
4776004f
TT		 1213 printk(KERN_ERR "EXT4 (no)user_xattr options "
1214 "not supported\n");
ac27a0ec
DK		 1215 break;
1216#endif
03010a33 1217#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1218 case Opt_acl:
1219 set_opt(sbi->s_mount_opt, POSIX_ACL);
1220 break;
1221 case Opt_noacl:
1222 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1223 break;
1224#else
1225 case Opt_acl:
1226 case Opt_noacl:
4776004f
TT		 1227 printk(KERN_ERR "EXT4 (no)acl options "
1228 "not supported\n");
ac27a0ec
DK		 1229 break;
1230#endif
ac27a0ec
DK		 1231 case Opt_journal_update:
1232 /* @@@ FIXME */
1233 /* Eventually we will want to be able to create
1234 a journal file here. For now, only allow the
1235 user to specify an existing inode to be the
1236 journal file. */
1237 if (is_remount) {
617ba13b 1238 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1239 "journal on remount\n");
1240 return 0;
1241 }
2b2d6d01 1242 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1243 break;
ac27a0ec
DK		 1244 case Opt_journal_dev:
1245 if (is_remount) {
617ba13b 1246 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1247 "journal on remount\n");
1248 return 0;
1249 }
1250 if (match_int(&args[0], &option))
1251 return 0;
1252 *journal_devnum = option;
1253 break;
818d276c
GS		 1254 case Opt_journal_checksum:
1255 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1256 break;
1257 case Opt_journal_async_commit:
1258 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1259 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1260 break;
ac27a0ec 1261 case Opt_noload:
2b2d6d01 1262 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1263 break;
1264 case Opt_commit:
1265 if (match_int(&args[0], &option))
1266 return 0;
1267 if (option < 0)
1268 return 0;
1269 if (option == 0)
cd02ff0b 1270 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1271 sbi->s_commit_interval = HZ * option;
1272 break;
30773840
TT		 1273 case Opt_max_batch_time:
1274 if (match_int(&args[0], &option))
1275 return 0;
1276 if (option < 0)
1277 return 0;
1278 if (option == 0)
1279 option = EXT4_DEF_MAX_BATCH_TIME;
1280 sbi->s_max_batch_time = option;
1281 break;
1282 case Opt_min_batch_time:
1283 if (match_int(&args[0], &option))
1284 return 0;
1285 if (option < 0)
1286 return 0;
1287 sbi->s_min_batch_time = option;
1288 break;
ac27a0ec 1289 case Opt_data_journal:
617ba13b 1290 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1291 goto datacheck;
1292 case Opt_data_ordered:
617ba13b 1293 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1294 goto datacheck;
1295 case Opt_data_writeback:
617ba13b 1296 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1297 datacheck:
1298 if (is_remount) {
617ba13b 1299 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1300 != data_opt) {
1301 printk(KERN_ERR
617ba13b 1302 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1303 "mode on remount\n");
1304 return 0;
1305 }
1306 } else {
617ba13b 1307 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1308 sbi->s_mount_opt |= data_opt;
1309 }
1310 break;
5bf5683a
HK		 1311 case Opt_data_err_abort:
1312 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1313 break;
1314 case Opt_data_err_ignore:
1315 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1316 break;
ac27a0ec
DK		 1317#ifdef CONFIG_QUOTA
1318 case Opt_usrjquota:
1319 qtype = USRQUOTA;
1320 goto set_qf_name;
1321 case Opt_grpjquota:
1322 qtype = GRPQUOTA;
1323set_qf_name:
17bd13b3 1324 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1325 !sbi->s_qf_names[qtype]) {
ac27a0ec 1326 printk(KERN_ERR
4776004f
TT		 1327 "EXT4-fs: Cannot change journaled "
1328 "quota options when quota turned on.\n");
ac27a0ec
DK		 1329 return 0;
1330 }
1331 qname = match_strdup(&args[0]);
1332 if (!qname) {
1333 printk(KERN_ERR
617ba13b 1334 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1335 "storing quotafile name.\n");
1336 return 0;
1337 }
1338 if (sbi->s_qf_names[qtype] &&
1339 strcmp(sbi->s_qf_names[qtype], qname)) {
1340 printk(KERN_ERR
617ba13b 1341 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1342 "specified.\n", QTYPE2NAME(qtype));
1343 kfree(qname);
1344 return 0;
1345 }
1346 sbi->s_qf_names[qtype] = qname;
1347 if (strchr(sbi->s_qf_names[qtype], '/')) {
1348 printk(KERN_ERR
617ba13b 1349 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1350 "filesystem root.\n");
1351 kfree(sbi->s_qf_names[qtype]);
1352 sbi->s_qf_names[qtype] = NULL;
1353 return 0;
1354 }
1355 set_opt(sbi->s_mount_opt, QUOTA);
1356 break;
1357 case Opt_offusrjquota:
1358 qtype = USRQUOTA;
1359 goto clear_qf_name;
1360 case Opt_offgrpjquota:
1361 qtype = GRPQUOTA;
1362clear_qf_name:
17bd13b3 1363 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1364 sbi->s_qf_names[qtype]) {
617ba13b 1365 printk(KERN_ERR "EXT4-fs: Cannot change "
2c8be6b2 1366 "journaled quota options when "
ac27a0ec
DK		 1367 "quota turned on.\n");
1368 return 0;
1369 }
1370 /*
1371 * The space will be released later when all options
1372 * are confirmed to be correct
1373 */
1374 sbi->s_qf_names[qtype] = NULL;
1375 break;
1376 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1377 qfmt = QFMT_VFS_OLD;
1378 goto set_qf_format;
ac27a0ec 1379 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1380 qfmt = QFMT_VFS_V0;
1381set_qf_format:
17bd13b3 1382 if (sb_any_quota_loaded(sb) &&
dfc5d03f
JK		 1383 sbi->s_jquota_fmt != qfmt) {
1384 printk(KERN_ERR "EXT4-fs: Cannot change "
1385 "journaled quota options when "
1386 "quota turned on.\n");
1387 return 0;
1388 }
1389 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1390 break;
1391 case Opt_quota:
1392 case Opt_usrquota:
1393 set_opt(sbi->s_mount_opt, QUOTA);
1394 set_opt(sbi->s_mount_opt, USRQUOTA);
1395 break;
1396 case Opt_grpquota:
1397 set_opt(sbi->s_mount_opt, QUOTA);
1398 set_opt(sbi->s_mount_opt, GRPQUOTA);
1399 break;
1400 case Opt_noquota:
17bd13b3 1401 if (sb_any_quota_loaded(sb)) {
617ba13b 1402 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1403 "options when quota turned on.\n");
1404 return 0;
1405 }
1406 clear_opt(sbi->s_mount_opt, QUOTA);
1407 clear_opt(sbi->s_mount_opt, USRQUOTA);
1408 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1409 break;
1410#else
1411 case Opt_quota:
1412 case Opt_usrquota:
1413 case Opt_grpquota:
cd59e7b9
JK		 1414 printk(KERN_ERR
1415 "EXT4-fs: quota options not supported.\n");
1416 break;
ac27a0ec
DK		 1417 case Opt_usrjquota:
1418 case Opt_grpjquota:
1419 case Opt_offusrjquota:
1420 case Opt_offgrpjquota:
1421 case Opt_jqfmt_vfsold:
1422 case Opt_jqfmt_vfsv0:
1423 printk(KERN_ERR
cd59e7b9 1424 "EXT4-fs: journaled quota options not "
ac27a0ec
DK		 1425 "supported.\n");
1426 break;
1427 case Opt_noquota:
1428 break;
1429#endif
1430 case Opt_abort:
1431 set_opt(sbi->s_mount_opt, ABORT);
1432 break;
06705bff
TT		 1433 case Opt_nobarrier:
1434 clear_opt(sbi->s_mount_opt, BARRIER);
1435 break;
ac27a0ec 1436 case Opt_barrier:
06705bff
TT		 1437 if (match_int(&args[0], &option)) {
1438 set_opt(sbi->s_mount_opt, BARRIER);
1439 break;
1440 }
ac27a0ec
DK		 1441 if (option)
1442 set_opt(sbi->s_mount_opt, BARRIER);
1443 else
1444 clear_opt(sbi->s_mount_opt, BARRIER);
1445 break;
1446 case Opt_ignore:
1447 break;
1448 case Opt_resize:
1449 if (!is_remount) {
617ba13b 1450 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1451 "for remount\n");
1452 return 0;
1453 }
1454 if (match_int(&args[0], &option) != 0)
1455 return 0;
1456 *n_blocks_count = option;
1457 break;
1458 case Opt_nobh:
1459 set_opt(sbi->s_mount_opt, NOBH);
1460 break;
1461 case Opt_bh:
1462 clear_opt(sbi->s_mount_opt, NOBH);
1463 break;
25ec56b5
JNC		 1464 case Opt_i_version:
1465 set_opt(sbi->s_mount_opt, I_VERSION);
1466 sb->s_flags |= MS_I_VERSION;
1467 break;
dd919b98
AK		 1468 case Opt_nodelalloc:
1469 clear_opt(sbi->s_mount_opt, DELALLOC);
1470 break;
c9de560d
AT		 1471 case Opt_stripe:
1472 if (match_int(&args[0], &option))
1473 return 0;
1474 if (option < 0)
1475 return 0;
1476 sbi->s_stripe = option;
1477 break;
64769240
AT		 1478 case Opt_delalloc:
1479 set_opt(sbi->s_mount_opt, DELALLOC);
1480 break;
240799cd
TT		 1481 case Opt_inode_readahead_blks:
1482 if (match_int(&args[0], &option))
1483 return 0;
1484 if (option < 0 || option > (1 << 30))
1485 return 0;
f7c43950 1486 if (!is_power_of_2(option)) {
3197ebdb
TT		 1487 printk(KERN_ERR "EXT4-fs: inode_readahead_blks"
1488 " must be a power of 2\n");
1489 return 0;
1490 }
240799cd
TT		 1491 sbi->s_inode_readahead_blks = option;
1492 break;
b3881f74
TT		 1493 case Opt_journal_ioprio:
1494 if (match_int(&args[0], &option))
1495 return 0;
1496 if (option < 0 || option > 7)
1497 break;
1498 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1499 option);
1500 break;
06705bff
TT		 1501 case Opt_noauto_da_alloc:
1502 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1503 break;
afd4672d 1504 case Opt_auto_da_alloc:
06705bff
TT		 1505 if (match_int(&args[0], &option)) {
1506 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1507 break;
1508 }
afd4672d
TT		 1509 if (option)
1510 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1511 else
1512 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1513 break;
ac27a0ec 1514 default:
2b2d6d01
TT		 1515 printk(KERN_ERR
1516 "EXT4-fs: Unrecognized mount option \"%s\" "
1517 "or missing value\n", p);
ac27a0ec
DK		 1518 return 0;
1519 }
1520 }
1521#ifdef CONFIG_QUOTA
1522 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1523 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1524 sbi->s_qf_names[USRQUOTA])
1525 clear_opt(sbi->s_mount_opt, USRQUOTA);
1526
617ba13b 1527 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1528 sbi->s_qf_names[GRPQUOTA])
1529 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1530
1531 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1532 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1533 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1534 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1535 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1536 "format mixing.\n");
1537 return 0;
1538 }
1539
1540 if (!sbi->s_jquota_fmt) {
2c8be6b2 1541 printk(KERN_ERR "EXT4-fs: journaled quota format "
ac27a0ec
DK		 1542 "not specified.\n");
1543 return 0;
1544 }
1545 } else {
1546 if (sbi->s_jquota_fmt) {
2c8be6b2
JK		 1547 printk(KERN_ERR "EXT4-fs: journaled quota format "
1548 "specified with no journaling "
ac27a0ec
DK		 1549 "enabled.\n");
1550 return 0;
1551 }
1552 }
1553#endif
1554 return 1;
1555}
1556
617ba13b 1557static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1558 int read_only)
1559{
617ba13b 1560 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1561 int res = 0;
1562
617ba13b 1563 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2b2d6d01
TT		 1564 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1565 "forcing read-only mode\n");
ac27a0ec
DK		 1566 res = MS_RDONLY;
1567 }
1568 if (read_only)
1569 return res;
617ba13b 1570 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2b2d6d01
TT		 1571 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1572 "running e2fsck is recommended\n");
617ba13b 1573 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
2b2d6d01
TT		 1574 printk(KERN_WARNING
1575 "EXT4-fs warning: mounting fs with errors, "
1576 "running e2fsck is recommended\n");
ac27a0ec
DK		 1577 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1578 le16_to_cpu(es->s_mnt_count) >=
1579 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2b2d6d01
TT		 1580 printk(KERN_WARNING
1581 "EXT4-fs warning: maximal mount count reached, "
1582 "running e2fsck is recommended\n");
ac27a0ec
DK		 1583 else if (le32_to_cpu(es->s_checkinterval) &&
1584 (le32_to_cpu(es->s_lastcheck) +
1585 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2b2d6d01
TT		 1586 printk(KERN_WARNING
1587 "EXT4-fs warning: checktime reached, "
1588 "running e2fsck is recommended\n");
0390131b
FM		 1589 if (!sbi->s_journal)
1590 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1591 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1592 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1593 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1594 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1595 ext4_update_dynamic_rev(sb);
0390131b
FM		 1596 if (sbi->s_journal)
1597 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1598
617ba13b 1599 ext4_commit_super(sb, es, 1);
ac27a0ec 1600 if (test_opt(sb, DEBUG))
a9df9a49 1601 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
ac27a0ec
DK		 1602 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1603 sb->s_blocksize,
1604 sbi->s_groups_count,
617ba13b
MC		 1605 EXT4_BLOCKS_PER_GROUP(sb),
1606 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1607 sbi->s_mount_opt);
1608
0390131b
FM		 1609 if (EXT4_SB(sb)->s_journal) {
1610 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1611 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1612 "external", EXT4_SB(sb)->s_journal->j_devname);
1613 } else {
1614 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1615 }
ac27a0ec
DK		 1616 return res;
1617}
1618
772cb7c8
JS		 1619static int ext4_fill_flex_info(struct super_block *sb)
1620{
1621 struct ext4_sb_info *sbi = EXT4_SB(sb);
1622 struct ext4_group_desc *gdp = NULL;
1623 struct buffer_head *bh;
1624 ext4_group_t flex_group_count;
1625 ext4_group_t flex_group;
1626 int groups_per_flex = 0;
c5ca7c76 1627 size_t size;
772cb7c8
JS		 1628 int i;
1629
1630 if (!sbi->s_es->s_log_groups_per_flex) {
1631 sbi->s_log_groups_per_flex = 0;
1632 return 1;
1633 }
1634
1635 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1636 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1637
c62a11fd
FB		 1638 /* We allocate both existing and potentially added groups */
1639 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1640 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1641 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76
TT		 1642 size = flex_group_count * sizeof(struct flex_groups);
1643 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1644 if (sbi->s_flex_groups == NULL) {
1645 sbi->s_flex_groups = vmalloc(size);
1646 if (sbi->s_flex_groups)
1647 memset(sbi->s_flex_groups, 0, size);
1648 }
772cb7c8 1649 if (sbi->s_flex_groups == NULL) {
ec05e868 1650 printk(KERN_ERR "EXT4-fs: not enough memory for "
a9df9a49 1651 "%u flex groups\n", flex_group_count);
772cb7c8
JS		 1652 goto failed;
1653 }
772cb7c8 1654
772cb7c8
JS		 1655 for (i = 0; i < sbi->s_groups_count; i++) {
1656 gdp = ext4_get_group_desc(sb, i, &bh);
1657
1658 flex_group = ext4_flex_group(sbi, i);
9f24e420
TT		 1659 atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1660 ext4_free_inodes_count(sb, gdp));
1661 atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1662 ext4_free_blks_count(sb, gdp));
7d39db14
TT		 1663 atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1664 ext4_used_dirs_count(sb, gdp));
772cb7c8
JS		 1665 }
1666
1667 return 1;
1668failed:
1669 return 0;
1670}
1671
717d50e4
AD		 1672__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1673 struct ext4_group_desc *gdp)
1674{
1675 __u16 crc = 0;
1676
1677 if (sbi->s_es->s_feature_ro_compat &
1678 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1679 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1680 __le32 le_group = cpu_to_le32(block_group);
1681
1682 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1683 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1684 crc = crc16(crc, (__u8 *)gdp, offset);
1685 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1686 /* for checksum of struct ext4_group_desc do the rest...*/
1687 if ((sbi->s_es->s_feature_incompat &
1688 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1689 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1690 crc = crc16(crc, (__u8 *)gdp + offset,
1691 le16_to_cpu(sbi->s_es->s_desc_size) -
1692 offset);
1693 }
1694
1695 return cpu_to_le16(crc);
1696}
1697
1698int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1699 struct ext4_group_desc *gdp)
1700{
1701 if ((sbi->s_es->s_feature_ro_compat &
1702 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1703 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1704 return 0;
1705
1706 return 1;
1707}
1708
ac27a0ec 1709/* Called at mount-time, super-block is locked */
197cd65a 1710static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1711{
617ba13b
MC		 1712 struct ext4_sb_info *sbi = EXT4_SB(sb);
1713 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1714 ext4_fsblk_t last_block;
bd81d8ee
LV		 1715 ext4_fsblk_t block_bitmap;
1716 ext4_fsblk_t inode_bitmap;
1717 ext4_fsblk_t inode_table;
ce421581 1718 int flexbg_flag = 0;
fd2d4291 1719 ext4_group_t i;
ac27a0ec 1720
ce421581
JS		 1721 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1722 flexbg_flag = 1;
1723
af5bc92d 1724 ext4_debug("Checking group descriptors");
ac27a0ec 1725
197cd65a
AM		 1726 for (i = 0; i < sbi->s_groups_count; i++) {
1727 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1728
ce421581 1729 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1730 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1731 else
1732 last_block = first_block +
617ba13b 1733 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1734
8fadc143 1735 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1736 if (block_bitmap < first_block || block_bitmap > last_block) {
c19204b0 1737 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1738 "Block bitmap for group %u not in group "
5128273a 1739 "(block %llu)!\n", i, block_bitmap);
ac27a0ec
DK		 1740 return 0;
1741 }
8fadc143 1742 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1743 if (inode_bitmap < first_block || inode_bitmap > last_block) {
c19204b0 1744 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1745 "Inode bitmap for group %u not in group "
5128273a 1746 "(block %llu)!\n", i, inode_bitmap);
ac27a0ec
DK		 1747 return 0;
1748 }
8fadc143 1749 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1750 if (inode_table < first_block ||
2b2d6d01 1751 inode_table + sbi->s_itb_per_group - 1 > last_block) {
c19204b0 1752 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1753 "Inode table for group %u not in group "
5128273a 1754 "(block %llu)!\n", i, inode_table);
ac27a0ec
DK		 1755 return 0;
1756 }
b5f10eed 1757 spin_lock(sb_bgl_lock(sbi, i));
717d50e4 1758 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
c19204b0 1759 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1760 "Checksum for group %u failed (%u!=%u)\n",
c19204b0
JB		 1761 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1762 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c
LZ		 1763 if (!(sb->s_flags & MS_RDONLY)) {
1764 spin_unlock(sb_bgl_lock(sbi, i));
8a266467 1765 return 0;
7ee1ec4c 1766 }
717d50e4 1767 }
b5f10eed 1768 spin_unlock(sb_bgl_lock(sbi, i));
ce421581
JS		 1769 if (!flexbg_flag)
1770 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1771 }
1772
bd81d8ee 1773 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2b2d6d01 1774 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1775 return 1;
1776}
1777
617ba13b 1778/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1779 * the superblock) which were deleted from all directories, but held open by
1780 * a process at the time of a crash. We walk the list and try to delete these
1781 * inodes at recovery time (only with a read-write filesystem).
1782 *
1783 * In order to keep the orphan inode chain consistent during traversal (in
1784 * case of crash during recovery), we link each inode into the superblock
1785 * orphan list_head and handle it the same way as an inode deletion during
1786 * normal operation (which journals the operations for us).
1787 *
1788 * We only do an iget() and an iput() on each inode, which is very safe if we
1789 * accidentally point at an in-use or already deleted inode. The worst that
1790 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1791 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1792 * e2fsck was run on this filesystem, and it must have already done the orphan
1793 * inode cleanup for us, so we can safely abort without any further action.
1794 */
2b2d6d01
TT		 1795static void ext4_orphan_cleanup(struct super_block *sb,
1796 struct ext4_super_block *es)
ac27a0ec
DK		 1797{
1798 unsigned int s_flags = sb->s_flags;
1799 int nr_orphans = 0, nr_truncates = 0;
1800#ifdef CONFIG_QUOTA
1801 int i;
1802#endif
1803 if (!es->s_last_orphan) {
1804 jbd_debug(4, "no orphan inodes to clean up\n");
1805 return;
1806 }
1807
a8f48a95
ES		 1808 if (bdev_read_only(sb->s_bdev)) {
1809 printk(KERN_ERR "EXT4-fs: write access "
1810 "unavailable, skipping orphan cleanup.\n");
1811 return;
1812 }
1813
617ba13b 1814 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1815 if (es->s_last_orphan)
1816 jbd_debug(1, "Errors on filesystem, "
1817 "clearing orphan list.\n");
1818 es->s_last_orphan = 0;
1819 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1820 return;
1821 }
1822
1823 if (s_flags & MS_RDONLY) {
617ba13b 1824 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1825 sb->s_id);
1826 sb->s_flags &= ~MS_RDONLY;
1827 }
1828#ifdef CONFIG_QUOTA
1829 /* Needed for iput() to work correctly and not trash data */
1830 sb->s_flags |= MS_ACTIVE;
1831 /* Turn on quotas so that they are updated correctly */
1832 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1833 if (EXT4_SB(sb)->s_qf_names[i]) {
1834 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1835 if (ret < 0)
1836 printk(KERN_ERR
2c8be6b2 1837 "EXT4-fs: Cannot turn on journaled "
ac27a0ec
DK		 1838 "quota: error %d\n", ret);
1839 }
1840 }
1841#endif
1842
1843 while (es->s_last_orphan) {
1844 struct inode *inode;
1845
97bd42b9
JB		 1846 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1847 if (IS_ERR(inode)) {
ac27a0ec
DK		 1848 es->s_last_orphan = 0;
1849 break;
1850 }
1851
617ba13b 1852 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
a269eb18 1853 vfs_dq_init(inode);
ac27a0ec
DK		 1854 if (inode->i_nlink) {
1855 printk(KERN_DEBUG
e5f8eab8 1856 "%s: truncating inode %lu to %lld bytes\n",
46e665e9 1857 __func__, inode->i_ino, inode->i_size);
e5f8eab8 1858 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 1859 inode->i_ino, inode->i_size);
617ba13b 1860 ext4_truncate(inode);
ac27a0ec
DK		 1861 nr_truncates++;
1862 } else {
1863 printk(KERN_DEBUG
1864 "%s: deleting unreferenced inode %lu\n",
46e665e9 1865 __func__, inode->i_ino);
ac27a0ec
DK		 1866 jbd_debug(2, "deleting unreferenced inode %lu\n",
1867 inode->i_ino);
1868 nr_orphans++;
1869 }
1870 iput(inode); /* The delete magic happens here! */
1871 }
1872
2b2d6d01 1873#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 1874
1875 if (nr_orphans)
617ba13b 1876 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1877 sb->s_id, PLURAL(nr_orphans));
1878 if (nr_truncates)
617ba13b 1879 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1880 sb->s_id, PLURAL(nr_truncates));
1881#ifdef CONFIG_QUOTA
1882 /* Turn quotas off */
1883 for (i = 0; i < MAXQUOTAS; i++) {
1884 if (sb_dqopt(sb)->files[i])
6f28e087 1885 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1886 }
1887#endif
1888 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1889}
cd2291a4
ES		 1890/*
1891 * Maximal extent format file size.
1892 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1893 * extent format containers, within a sector_t, and within i_blocks
1894 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1895 * so that won't be a limiting factor.
1896 *
1897 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1898 */
f287a1a5 1899static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 1900{
1901 loff_t res;
1902 loff_t upper_limit = MAX_LFS_FILESIZE;
1903
1904 /* small i_blocks in vfs inode? */
f287a1a5 1905 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 1906 /*
b3a6ffe1 1907 * CONFIG_LBD is not enabled implies the inode
cd2291a4
ES		 1908 * i_block represent total blocks in 512 bytes
1909 * 32 == size of vfs inode i_blocks * 8
1910 */
1911 upper_limit = (1LL << 32) - 1;
1912
1913 /* total blocks in file system block size */
1914 upper_limit >>= (blkbits - 9);
1915 upper_limit <<= blkbits;
1916 }
1917
1918 /* 32-bit extent-start container, ee_block */
1919 res = 1LL << 32;
1920 res <<= blkbits;
1921 res -= 1;
1922
1923 /* Sanity check against vm- & vfs- imposed limits */
1924 if (res > upper_limit)
1925 res = upper_limit;
1926
1927 return res;
1928}
ac27a0ec 1929
ac27a0ec 1930/*
cd2291a4 1931 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1932 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1933 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1934 */
f287a1a5 1935static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 1936{
617ba13b 1937 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1938 int meta_blocks;
1939 loff_t upper_limit;
1940 /* This is calculated to be the largest file size for a
cd2291a4 1941 * dense, bitmapped file such that the total number of
ac27a0ec 1942 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1943 * does not exceed 2^48 -1
1944 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1945 * total number of 512 bytes blocks of the file
1946 */
1947
f287a1a5 1948 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 1949 /*
b3a6ffe1 1950 * !has_huge_files or CONFIG_LBD is not enabled
f287a1a5
TT		 1951 * implies the inode i_block represent total blocks in
1952 * 512 bytes 32 == size of vfs inode i_blocks * 8
0fc1b451
AK		 1953 */
1954 upper_limit = (1LL << 32) - 1;
1955
1956 /* total blocks in file system block size */
1957 upper_limit >>= (bits - 9);
1958
1959 } else {
8180a562
AK		 1960 /*
1961 * We use 48 bit ext4_inode i_blocks
1962 * With EXT4_HUGE_FILE_FL set the i_blocks
1963 * represent total number of blocks in
1964 * file system block size
1965 */
0fc1b451
AK		 1966 upper_limit = (1LL << 48) - 1;
1967
0fc1b451
AK		 1968 }
1969
1970 /* indirect blocks */
1971 meta_blocks = 1;
1972 /* double indirect blocks */
1973 meta_blocks += 1 + (1LL << (bits-2));
1974 /* tripple indirect blocks */
1975 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1976
1977 upper_limit -= meta_blocks;
1978 upper_limit <<= bits;
ac27a0ec
DK		 1979
1980 res += 1LL << (bits-2);
1981 res += 1LL << (2*(bits-2));
1982 res += 1LL << (3*(bits-2));
1983 res <<= bits;
1984 if (res > upper_limit)
1985 res = upper_limit;
0fc1b451
AK		 1986
1987 if (res > MAX_LFS_FILESIZE)
1988 res = MAX_LFS_FILESIZE;
1989
ac27a0ec
DK		 1990 return res;
1991}
1992
617ba13b 1993static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1994 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1995{
617ba13b 1996 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1997 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1998 int has_super = 0;
1999
2000 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2001
617ba13b 2002 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2003 nr < first_meta_bg)
70bbb3e0 2004 return logical_sb_block + nr + 1;
ac27a0ec 2005 bg = sbi->s_desc_per_block * nr;
617ba13b 2006 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2007 has_super = 1;
617ba13b 2008 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2009}
2010
c9de560d
AT		 2011/**
2012 * ext4_get_stripe_size: Get the stripe size.
2013 * @sbi: In memory super block info
2014 *
2015 * If we have specified it via mount option, then
2016 * use the mount option value. If the value specified at mount time is
2017 * greater than the blocks per group use the super block value.
2018 * If the super block value is greater than blocks per group return 0.
2019 * Allocator needs it be less than blocks per group.
2020 *
2021 */
2022static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2023{
2024 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2025 unsigned long stripe_width =
2026 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2027
2028 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2029 return sbi->s_stripe;
2030
2031 if (stripe_width <= sbi->s_blocks_per_group)
2032 return stripe_width;
2033
2034 if (stride <= sbi->s_blocks_per_group)
2035 return stride;
2036
2037 return 0;
2038}
ac27a0ec 2039
3197ebdb
TT		 2040/* sysfs supprt */
2041
2042struct ext4_attr {
2043 struct attribute attr;
2044 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2045 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2046 const char *, size_t);
2047 int offset;
2048};
2049
2050static int parse_strtoul(const char *buf,
2051 unsigned long max, unsigned long *value)
2052{
2053 char *endp;
2054
2055 while (*buf && isspace(*buf))
2056 buf++;
2057 *value = simple_strtoul(buf, &endp, 0);
2058 while (*endp && isspace(*endp))
2059 endp++;
2060 if (*endp || *value > max)
2061 return -EINVAL;
2062
2063 return 0;
2064}
2065
2066static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2067 struct ext4_sb_info *sbi,
2068 char *buf)
2069{
2070 return snprintf(buf, PAGE_SIZE, "%llu\n",
2071 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2072}
2073
2074static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2075 struct ext4_sb_info *sbi, char *buf)
2076{
2077 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2078
2079 return snprintf(buf, PAGE_SIZE, "%lu\n",
2080 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2081 sbi->s_sectors_written_start) >> 1);
2082}
2083
2084static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2085 struct ext4_sb_info *sbi, char *buf)
2086{
2087 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2088
2089 return snprintf(buf, PAGE_SIZE, "%llu\n",
2090 sbi->s_kbytes_written +
2091 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2092 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2093}
2094
2095static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2096 struct ext4_sb_info *sbi,
2097 const char *buf, size_t count)
2098{
2099 unsigned long t;
2100
2101 if (parse_strtoul(buf, 0x40000000, &t))
2102 return -EINVAL;
2103
f7c43950 2104 if (!is_power_of_2(t))
3197ebdb
TT		 2105 return -EINVAL;
2106
2107 sbi->s_inode_readahead_blks = t;
2108 return count;
2109}
2110
2111static ssize_t sbi_ui_show(struct ext4_attr *a,
2112 struct ext4_sb_info *sbi, char *buf)
2113{
2114 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2115
2116 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2117}
2118
2119static ssize_t sbi_ui_store(struct ext4_attr *a,
2120 struct ext4_sb_info *sbi,
2121 const char *buf, size_t count)
2122{
2123 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2124 unsigned long t;
2125
2126 if (parse_strtoul(buf, 0xffffffff, &t))
2127 return -EINVAL;
2128 *ui = t;
2129 return count;
2130}
2131
2132#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2133static struct ext4_attr ext4_attr_##_name = { \
2134 .attr = {.name = __stringify(_name), .mode = _mode }, \
2135 .show = _show, \
2136 .store = _store, \
2137 .offset = offsetof(struct ext4_sb_info, _elname), \
2138}
2139#define EXT4_ATTR(name, mode, show, store) \
2140static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2141
2142#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2143#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2144#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2145 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2146#define ATTR_LIST(name) &ext4_attr_##name.attr
2147
2148EXT4_RO_ATTR(delayed_allocation_blocks);
2149EXT4_RO_ATTR(session_write_kbytes);
2150EXT4_RO_ATTR(lifetime_write_kbytes);
2151EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2152 inode_readahead_blks_store, s_inode_readahead_blks);
2153EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2154EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2155EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2156EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2157EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2158EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2159
2160static struct attribute *ext4_attrs[] = {
2161 ATTR_LIST(delayed_allocation_blocks),
2162 ATTR_LIST(session_write_kbytes),
2163 ATTR_LIST(lifetime_write_kbytes),
2164 ATTR_LIST(inode_readahead_blks),
2165 ATTR_LIST(mb_stats),
2166 ATTR_LIST(mb_max_to_scan),
2167 ATTR_LIST(mb_min_to_scan),
2168 ATTR_LIST(mb_order2_req),
2169 ATTR_LIST(mb_stream_req),
2170 ATTR_LIST(mb_group_prealloc),
2171 NULL,
2172};
2173
2174static ssize_t ext4_attr_show(struct kobject *kobj,
2175 struct attribute *attr, char *buf)
2176{
2177 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2178 s_kobj);
2179 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2180
2181 return a->show ? a->show(a, sbi, buf) : 0;
2182}
2183
2184static ssize_t ext4_attr_store(struct kobject *kobj,
2185 struct attribute *attr,
2186 const char *buf, size_t len)
2187{
2188 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2189 s_kobj);
2190 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2191
2192 return a->store ? a->store(a, sbi, buf, len) : 0;
2193}
2194
2195static void ext4_sb_release(struct kobject *kobj)
2196{
2197 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2198 s_kobj);
2199 complete(&sbi->s_kobj_unregister);
2200}
2201
2202
2203static struct sysfs_ops ext4_attr_ops = {
2204 .show = ext4_attr_show,
2205 .store = ext4_attr_store,
2206};
2207
2208static struct kobj_type ext4_ktype = {
2209 .default_attrs = ext4_attrs,
2210 .sysfs_ops = &ext4_attr_ops,
2211 .release = ext4_sb_release,
2212};
2213
2b2d6d01 2214static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 2215 __releases(kernel_lock)
2216 __acquires(kernel_lock)
1d03ec98 2217
ac27a0ec 2218{
2b2d6d01 2219 struct buffer_head *bh;
617ba13b
MC		 2220 struct ext4_super_block *es = NULL;
2221 struct ext4_sb_info *sbi;
2222 ext4_fsblk_t block;
2223 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2224 ext4_fsblk_t logical_sb_block;
ac27a0ec 2225 unsigned long offset = 0;
ac27a0ec
DK		 2226 unsigned long journal_devnum = 0;
2227 unsigned long def_mount_opts;
2228 struct inode *root;
9f6200bb 2229 char *cp;
0390131b 2230 const char *descr;
1d1fe1ee 2231 int ret = -EINVAL;
ac27a0ec 2232 int blocksize;
4ec11028
TT		 2233 unsigned int db_count;
2234 unsigned int i;
f287a1a5 2235 int needs_recovery, has_huge_files;
3a06d778 2236 int features;
bd81d8ee 2237 __u64 blocks_count;
833f4077 2238 int err;
b3881f74 2239 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 2240
2241 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2242 if (!sbi)
2243 return -ENOMEM;
705895b6
PE		 2244
2245 sbi->s_blockgroup_lock =
2246 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2247 if (!sbi->s_blockgroup_lock) {
2248 kfree(sbi);
2249 return -ENOMEM;
2250 }
ac27a0ec
DK		 2251 sb->s_fs_info = sbi;
2252 sbi->s_mount_opt = 0;
617ba13b
MC		 2253 sbi->s_resuid = EXT4_DEF_RESUID;
2254 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2255 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2256 sbi->s_sb_block = sb_block;
afc32f7e
TT		 2257 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2258 sectors[1]);
ac27a0ec
DK		 2259
2260 unlock_kernel();
2261
9f6200bb
TT		 2262 /* Cleanup superblock name */
2263 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2264 *cp = '!';
2265
617ba13b 2266 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2267 if (!blocksize) {
617ba13b 2268 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 2269 goto out_fail;
2270 }
2271
2272 /*
617ba13b 2273 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 2274 * block sizes. We need to calculate the offset from buffer start.
2275 */
617ba13b 2276 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 2277 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2278 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 2279 } else {
70bbb3e0 2280 logical_sb_block = sb_block;
ac27a0ec
DK		 2281 }
2282
70bbb3e0 2283 if (!(bh = sb_bread(sb, logical_sb_block))) {
2b2d6d01 2284 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 2285 goto out_fail;
2286 }
2287 /*
2288 * Note: s_es must be initialized as soon as possible because
617ba13b 2289 * some ext4 macro-instructions depend on its value
ac27a0ec 2290 */
617ba13b 2291 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 2292 sbi->s_es = es;
2293 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 2294 if (sb->s_magic != EXT4_SUPER_MAGIC)
2295 goto cantfind_ext4;
afc32f7e 2296 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 2297
2298 /* Set defaults before we parse the mount options */
2299 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 2300 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 2301 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 2302 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 2303 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 2304 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 2305 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 2306#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 2307 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 2308 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 2309#endif
03010a33 2310#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2311 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2312 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2313#endif
617ba13b
MC		 2314 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2315 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2316 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2317 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2318 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2319 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2320
2321 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2322 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2323 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2324 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2325 else
2326 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2327
2328 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2329 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 2330 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2331 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2332 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 2333
571640ca 2334 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2335
dd919b98
AK		 2336 /*
2337 * enable delayed allocation by default
2338 * Use -o nodelalloc to turn it off
2339 */
2340 set_opt(sbi->s_mount_opt, DELALLOC);
2341
2342
b3881f74
TT		 2343 if (!parse_options((char *) data, sb, &journal_devnum,
2344 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 2345 goto failed_mount;
2346
2347 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2348 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2349
617ba13b
MC		 2350 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2351 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2352 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2353 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 2354 printk(KERN_WARNING
617ba13b 2355 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec 2356 "running e2fsck is recommended\n");
469108ff 2357
ac27a0ec
DK		 2358 /*
2359 * Check feature flags regardless of the revision level, since we
2360 * previously didn't change the revision level when setting the flags,
2361 * so there is a chance incompat flags are set on a rev 0 filesystem.
2362 */
617ba13b 2363 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 2364 if (features) {
617ba13b 2365 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
3a06d778
AK		 2366 "unsupported optional features (%x).\n", sb->s_id,
2367 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2368 ~EXT4_FEATURE_INCOMPAT_SUPP));
ac27a0ec
DK		 2369 goto failed_mount;
2370 }
617ba13b 2371 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 2372 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 2373 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
3a06d778
AK		 2374 "unsupported optional features (%x).\n", sb->s_id,
2375 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2376 ~EXT4_FEATURE_RO_COMPAT_SUPP));
ac27a0ec
DK		 2377 goto failed_mount;
2378 }
f287a1a5
TT		 2379 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2380 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2381 if (has_huge_files) {
0fc1b451
AK		 2382 /*
2383 * Large file size enabled file system can only be
b3a6ffe1 2384 * mount if kernel is build with CONFIG_LBD
0fc1b451
AK		 2385 */
2386 if (sizeof(root->i_blocks) < sizeof(u64) &&
2387 !(sb->s_flags & MS_RDONLY)) {
2388 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2389 "files cannot be mounted read-write "
b3a6ffe1 2390 "without CONFIG_LBD.\n", sb->s_id);
0fc1b451
AK		 2391 goto failed_mount;
2392 }
2393 }
ac27a0ec
DK		 2394 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2395
617ba13b
MC		 2396 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2397 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 2398 printk(KERN_ERR
617ba13b 2399 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 2400 blocksize, sb->s_id);
2401 goto failed_mount;
2402 }
2403
ac27a0ec 2404 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2405
2406 /* Validate the filesystem blocksize */
2407 if (!sb_set_blocksize(sb, blocksize)) {
2408 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2409 blocksize);
ac27a0ec
DK		 2410 goto failed_mount;
2411 }
2412
2b2d6d01 2413 brelse(bh);
70bbb3e0
AM		 2414 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2415 offset = do_div(logical_sb_block, blocksize);
2416 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 2417 if (!bh) {
2418 printk(KERN_ERR
617ba13b 2419 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 2420 goto failed_mount;
2421 }
617ba13b 2422 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2423 sbi->s_es = es;
617ba13b 2424 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2b2d6d01
TT		 2425 printk(KERN_ERR
2426 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 2427 goto failed_mount;
2428 }
2429 }
2430
f287a1a5
TT		 2431 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2432 has_huge_files);
2433 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2434
617ba13b
MC		 2435 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2436 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2437 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2438 } else {
2439 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2440 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2441 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2442 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2443 (sbi->s_inode_size > blocksize)) {
2b2d6d01
TT		 2444 printk(KERN_ERR
2445 "EXT4-fs: unsupported inode size: %d\n",
2446 sbi->s_inode_size);
ac27a0ec
DK		 2447 goto failed_mount;
2448 }
ef7f3835
KS		 2449 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2450 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2451 }
0d1ee42f
AR		 2452 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2453 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2454 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2455 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2456 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2457 printk(KERN_ERR
8fadc143 2458 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2459 sbi->s_desc_size);
2460 goto failed_mount;
2461 }
2462 } else
2463 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2464 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2465 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2466 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2467 goto cantfind_ext4;
2468 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2469 if (sbi->s_inodes_per_block == 0)
617ba13b 2470 goto cantfind_ext4;
ac27a0ec
DK		 2471 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2472 sbi->s_inodes_per_block;
0d1ee42f 2473 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2474 sbi->s_sbh = bh;
2475 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2476 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2477 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2b2d6d01 2478 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2479 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2480 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2481 i = le32_to_cpu(es->s_flags);
2482 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2483 sbi->s_hash_unsigned = 3;
2484 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2485#ifdef __CHAR_UNSIGNED__
2486 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2487 sbi->s_hash_unsigned = 3;
2488#else
2489 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2490#endif
2491 sb->s_dirt = 1;
2492 }
ac27a0ec
DK		 2493
2494 if (sbi->s_blocks_per_group > blocksize * 8) {
2b2d6d01
TT		 2495 printk(KERN_ERR
2496 "EXT4-fs: #blocks per group too big: %lu\n",
2497 sbi->s_blocks_per_group);
ac27a0ec
DK		 2498 goto failed_mount;
2499 }
ac27a0ec 2500 if (sbi->s_inodes_per_group > blocksize * 8) {
2b2d6d01
TT		 2501 printk(KERN_ERR
2502 "EXT4-fs: #inodes per group too big: %lu\n",
2503 sbi->s_inodes_per_group);
ac27a0ec
DK		 2504 goto failed_mount;
2505 }
2506
bd81d8ee 2507 if (ext4_blocks_count(es) >
ac27a0ec 2508 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2509 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2510 " too large to mount safely\n", sb->s_id);
2511 if (sizeof(sector_t) < 8)
617ba13b 2512 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2513 "enabled\n");
2514 goto failed_mount;
2515 }
2516
617ba13b
MC		 2517 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2518 goto cantfind_ext4;
e7c95593 2519
0f2ddca6
FTN		 2520 /* check blocks count against device size */
2521 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2522 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2523 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu "
2524 "exceeds size of device (%llu blocks)\n",
2525 ext4_blocks_count(es), blocks_count);
2526 goto failed_mount;
2527 }
2528
4ec11028
TT		 2529 /*
2530 * It makes no sense for the first data block to be beyond the end
2531 * of the filesystem.
2532 */
2533 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2534 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2535 "block %u is beyond end of filesystem (%llu)\n",
2536 le32_to_cpu(es->s_first_data_block),
2537 ext4_blocks_count(es));
e7c95593
ES		 2538 goto failed_mount;
2539 }
bd81d8ee
LV		 2540 blocks_count = (ext4_blocks_count(es) -
2541 le32_to_cpu(es->s_first_data_block) +
2542 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2543 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028
TT		 2544 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2545 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2546 "(block count %llu, first data block %u, "
2547 "blocks per group %lu)\n", sbi->s_groups_count,
2548 ext4_blocks_count(es),
2549 le32_to_cpu(es->s_first_data_block),
2550 EXT4_BLOCKS_PER_GROUP(sb));
2551 goto failed_mount;
2552 }
bd81d8ee 2553 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2554 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2555 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2556 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2557 GFP_KERNEL);
2558 if (sbi->s_group_desc == NULL) {
2b2d6d01 2559 printk(KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2560 goto failed_mount;
2561 }
2562
3244fcb1 2563#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2564 if (ext4_proc_root)
2565 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 2566#endif
240799cd 2567
705895b6 2568 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2569
2570 for (i = 0; i < db_count; i++) {
70bbb3e0 2571 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2572 sbi->s_group_desc[i] = sb_bread(sb, block);
2573 if (!sbi->s_group_desc[i]) {
2b2d6d01
TT		 2574 printk(KERN_ERR "EXT4-fs: "
2575 "can't read group descriptor %d\n", i);
ac27a0ec
DK		 2576 db_count = i;
2577 goto failed_mount2;
2578 }
2579 }
2b2d6d01 2580 if (!ext4_check_descriptors(sb)) {
617ba13b 2581 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2582 goto failed_mount2;
2583 }
772cb7c8
JS		 2584 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2585 if (!ext4_fill_flex_info(sb)) {
2586 printk(KERN_ERR
2587 "EXT4-fs: unable to initialize "
2588 "flex_bg meta info!\n");
2589 goto failed_mount2;
2590 }
2591
ac27a0ec
DK		 2592 sbi->s_gdb_count = db_count;
2593 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2594 spin_lock_init(&sbi->s_next_gen_lock);
2595
833f4077
PZ		 2596 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2597 ext4_count_free_blocks(sb));
2598 if (!err) {
2599 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2600 ext4_count_free_inodes(sb));
2601 }
2602 if (!err) {
2603 err = percpu_counter_init(&sbi->s_dirs_counter,
2604 ext4_count_dirs(sb));
2605 }
6bc6e63f
AK		 2606 if (!err) {
2607 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2608 }
833f4077
PZ		 2609 if (err) {
2610 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2611 goto failed_mount3;
2612 }
ac27a0ec 2613
c9de560d
AT		 2614 sbi->s_stripe = ext4_get_stripe_size(sbi);
2615
ac27a0ec
DK		 2616 /*
2617 * set up enough so that it can read an inode
2618 */
617ba13b
MC		 2619 sb->s_op = &ext4_sops;
2620 sb->s_export_op = &ext4_export_ops;
2621 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2622#ifdef CONFIG_QUOTA
617ba13b
MC		 2623 sb->s_qcop = &ext4_qctl_operations;
2624 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2625#endif
2626 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2627
2628 sb->s_root = NULL;
2629
2630 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2631 EXT4_HAS_INCOMPAT_FEATURE(sb,
2632 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2633
2634 /*
2635 * The first inode we look at is the journal inode. Don't try
2636 * root first: it may be modified in the journal!
2637 */
2638 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2639 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2640 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2641 goto failed_mount3;
624080ed
TT		 2642 if (!(sb->s_flags & MS_RDONLY) &&
2643 EXT4_SB(sb)->s_journal->j_failed_commit) {
2644 printk(KERN_CRIT "EXT4-fs error (device %s): "
2645 "ext4_fill_super: Journal transaction "
2b2d6d01 2646 "%u is corrupt\n", sb->s_id,
624080ed 2647 EXT4_SB(sb)->s_journal->j_failed_commit);
2b2d6d01
TT		 2648 if (test_opt(sb, ERRORS_RO)) {
2649 printk(KERN_CRIT
2650 "Mounting filesystem read-only\n");
624080ed
TT		 2651 sb->s_flags |= MS_RDONLY;
2652 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2653 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2654 }
2655 if (test_opt(sb, ERRORS_PANIC)) {
2656 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2657 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2658 ext4_commit_super(sb, es, 1);
624080ed
TT		 2659 goto failed_mount4;
2660 }
2661 }
0390131b
FM		 2662 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2663 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2664 printk(KERN_ERR "EXT4-fs: required journal recovery "
2665 "suppressed and not mounted read-only\n");
2666 goto failed_mount4;
ac27a0ec 2667 } else {
0390131b
FM		 2668 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2669 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2670 sbi->s_journal = NULL;
2671 needs_recovery = 0;
2672 goto no_journal;
ac27a0ec
DK		 2673 }
2674
eb40a09c
JS		 2675 if (ext4_blocks_count(es) > 0xffffffffULL &&
2676 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2677 JBD2_FEATURE_INCOMPAT_64BIT)) {
abda1418 2678 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
eb40a09c
JS		 2679 goto failed_mount4;
2680 }
2681
818d276c
GS		 2682 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2683 jbd2_journal_set_features(sbi->s_journal,
2684 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2685 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2686 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2687 jbd2_journal_set_features(sbi->s_journal,
2688 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2689 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2690 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2691 } else {
2692 jbd2_journal_clear_features(sbi->s_journal,
2693 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2694 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2695 }
2696
ac27a0ec
DK		 2697 /* We have now updated the journal if required, so we can
2698 * validate the data journaling mode. */
2699 switch (test_opt(sb, DATA_FLAGS)) {
2700 case 0:
2701 /* No mode set, assume a default based on the journal
63f57933
AM		 2702 * capabilities: ORDERED_DATA if the journal can
2703 * cope, else JOURNAL_DATA
2704 */
dab291af
MC		 2705 if (jbd2_journal_check_available_features
2706 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2707 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2708 else
2709 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2710 break;
2711
617ba13b
MC		 2712 case EXT4_MOUNT_ORDERED_DATA:
2713 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2714 if (!jbd2_journal_check_available_features
2715 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2716 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2717 "requested data journaling mode\n");
2718 goto failed_mount4;
2719 }
2720 default:
2721 break;
2722 }
b3881f74 2723 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 2724
0390131b 2725no_journal:
ac27a0ec
DK		 2726
2727 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2728 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2729 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2730 "its supported only with writeback mode\n");
2731 clear_opt(sbi->s_mount_opt, NOBH);
2732 }
2733 }
2734 /*
dab291af 2735 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2736 * so we can safely mount the rest of the filesystem now.
2737 */
2738
1d1fe1ee
DH		 2739 root = ext4_iget(sb, EXT4_ROOT_INO);
2740 if (IS_ERR(root)) {
617ba13b 2741 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1d1fe1ee 2742 ret = PTR_ERR(root);
ac27a0ec
DK		 2743 goto failed_mount4;
2744 }
2745 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2746 iput(root);
617ba13b 2747 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2748 goto failed_mount4;
2749 }
1d1fe1ee
DH		 2750 sb->s_root = d_alloc_root(root);
2751 if (!sb->s_root) {
2752 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2753 iput(root);
2754 ret = -ENOMEM;
2755 goto failed_mount4;
2756 }
ac27a0ec 2757
2b2d6d01 2758 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2759
2760 /* determine the minimum size of new large inodes, if present */
2761 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2762 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2763 EXT4_GOOD_OLD_INODE_SIZE;
2764 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2765 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2766 if (sbi->s_want_extra_isize <
2767 le16_to_cpu(es->s_want_extra_isize))
2768 sbi->s_want_extra_isize =
2769 le16_to_cpu(es->s_want_extra_isize);
2770 if (sbi->s_want_extra_isize <
2771 le16_to_cpu(es->s_min_extra_isize))
2772 sbi->s_want_extra_isize =
2773 le16_to_cpu(es->s_min_extra_isize);
2774 }
2775 }
2776 /* Check if enough inode space is available */
2777 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2778 sbi->s_inode_size) {
2779 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2780 EXT4_GOOD_OLD_INODE_SIZE;
2781 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2782 "available.\n");
2783 }
2784
c2774d84
AK		 2785 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2786 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2787 "requested data journaling mode\n");
2788 clear_opt(sbi->s_mount_opt, DELALLOC);
2789 } else if (test_opt(sb, DELALLOC))
2790 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2791
2792 ext4_ext_init(sb);
2793 err = ext4_mb_init(sb, needs_recovery);
2794 if (err) {
2795 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2796 err);
2797 goto failed_mount4;
2798 }
2799
3197ebdb
TT		 2800 sbi->s_kobj.kset = ext4_kset;
2801 init_completion(&sbi->s_kobj_unregister);
2802 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2803 "%s", sb->s_id);
2804 if (err) {
2805 ext4_mb_release(sb);
2806 ext4_ext_release(sb);
2807 goto failed_mount4;
2808 };
2809
ac27a0ec
DK		 2810 /*
2811 * akpm: core read_super() calls in here with the superblock locked.
2812 * That deadlocks, because orphan cleanup needs to lock the superblock
2813 * in numerous places. Here we just pop the lock - it's relatively
2814 * harmless, because we are now ready to accept write_super() requests,
2815 * and aviro says that's the only reason for hanging onto the
2816 * superblock lock.
2817 */
617ba13b
MC		 2818 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2819 ext4_orphan_cleanup(sb, es);
2820 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 2821 if (needs_recovery) {
2b2d6d01 2822 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
0390131b
FM		 2823 ext4_mark_recovery_complete(sb, es);
2824 }
2825 if (EXT4_SB(sb)->s_journal) {
2826 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2827 descr = " journalled data mode";
2828 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2829 descr = " ordered data mode";
2830 else
2831 descr = " writeback data mode";
2832 } else
2833 descr = "out journal";
2834
2835 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2836 sb->s_id, descr);
ac27a0ec
DK		 2837
2838 lock_kernel();
2839 return 0;
2840
617ba13b 2841cantfind_ext4:
ac27a0ec 2842 if (!silent)
617ba13b 2843 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2844 sb->s_id);
2845 goto failed_mount;
2846
2847failed_mount4:
0390131b
FM		 2848 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2849 if (sbi->s_journal) {
2850 jbd2_journal_destroy(sbi->s_journal);
2851 sbi->s_journal = NULL;
2852 }
ac27a0ec 2853failed_mount3:
c5ca7c76
TT		 2854 if (sbi->s_flex_groups) {
2855 if (is_vmalloc_addr(sbi->s_flex_groups))
2856 vfree(sbi->s_flex_groups);
2857 else
2858 kfree(sbi->s_flex_groups);
2859 }
ac27a0ec
DK		 2860 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2861 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2862 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 2863 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 2864failed_mount2:
2865 for (i = 0; i < db_count; i++)
2866 brelse(sbi->s_group_desc[i]);
2867 kfree(sbi->s_group_desc);
2868failed_mount:
240799cd 2869 if (sbi->s_proc) {
9f6200bb 2870 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 2871 }
ac27a0ec
DK		 2872#ifdef CONFIG_QUOTA
2873 for (i = 0; i < MAXQUOTAS; i++)
2874 kfree(sbi->s_qf_names[i]);
2875#endif
617ba13b 2876 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2877 brelse(bh);
2878out_fail:
2879 sb->s_fs_info = NULL;
2880 kfree(sbi);
2881 lock_kernel();
1d1fe1ee 2882 return ret;
ac27a0ec
DK		 2883}
2884
2885/*
2886 * Setup any per-fs journal parameters now. We'll do this both on
2887 * initial mount, once the journal has been initialised but before we've
2888 * done any recovery; and again on any subsequent remount.
2889 */
617ba13b 2890static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2891{
617ba13b 2892 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 2893
30773840
TT		 2894 journal->j_commit_interval = sbi->s_commit_interval;
2895 journal->j_min_batch_time = sbi->s_min_batch_time;
2896 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 2897
2898 spin_lock(&journal->j_state_lock);
2899 if (test_opt(sb, BARRIER))
dab291af 2900 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2901 else
dab291af 2902 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 2903 if (test_opt(sb, DATA_ERR_ABORT))
2904 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2905 else
2906 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 2907 spin_unlock(&journal->j_state_lock);
2908}
2909
617ba13b 2910static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2911 unsigned int journal_inum)
2912{
2913 struct inode *journal_inode;
2914 journal_t *journal;
2915
0390131b
FM		 2916 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2917
ac27a0ec
DK		 2918 /* First, test for the existence of a valid inode on disk. Bad
2919 * things happen if we iget() an unused inode, as the subsequent
2920 * iput() will try to delete it. */
2921
1d1fe1ee
DH		 2922 journal_inode = ext4_iget(sb, journal_inum);
2923 if (IS_ERR(journal_inode)) {
617ba13b 2924 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2925 return NULL;
2926 }
2927 if (!journal_inode->i_nlink) {
2928 make_bad_inode(journal_inode);
2929 iput(journal_inode);
617ba13b 2930 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2931 return NULL;
2932 }
2933
e5f8eab8 2934 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 2935 journal_inode, journal_inode->i_size);
1d1fe1ee 2936 if (!S_ISREG(journal_inode->i_mode)) {
617ba13b 2937 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2938 iput(journal_inode);
2939 return NULL;
2940 }
2941
dab291af 2942 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2943 if (!journal) {
617ba13b 2944 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2945 iput(journal_inode);
2946 return NULL;
2947 }
2948 journal->j_private = sb;
617ba13b 2949 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2950 return journal;
2951}
2952
617ba13b 2953static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2954 dev_t j_dev)
2955{
2b2d6d01 2956 struct buffer_head *bh;
ac27a0ec 2957 journal_t *journal;
617ba13b
MC		 2958 ext4_fsblk_t start;
2959 ext4_fsblk_t len;
ac27a0ec 2960 int hblock, blocksize;
617ba13b 2961 ext4_fsblk_t sb_block;
ac27a0ec 2962 unsigned long offset;
2b2d6d01 2963 struct ext4_super_block *es;
ac27a0ec
DK		 2964 struct block_device *bdev;
2965
0390131b
FM		 2966 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2967
617ba13b 2968 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2969 if (bdev == NULL)
2970 return NULL;
2971
2972 if (bd_claim(bdev, sb)) {
2973 printk(KERN_ERR
abda1418 2974 "EXT4-fs: failed to claim external journal device.\n");
9a1c3542 2975 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 2976 return NULL;
2977 }
2978
2979 blocksize = sb->s_blocksize;
2980 hblock = bdev_hardsect_size(bdev);
2981 if (blocksize < hblock) {
2982 printk(KERN_ERR
617ba13b 2983 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2984 goto out_bdev;
2985 }
2986
617ba13b
MC		 2987 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2988 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2989 set_blocksize(bdev, blocksize);
2990 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2991 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2992 "external journal\n");
2993 goto out_bdev;
2994 }
2995
617ba13b
MC		 2996 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2997 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2998 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2999 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3000 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 3001 "bad superblock\n");
3002 brelse(bh);
3003 goto out_bdev;
3004 }
3005
617ba13b
MC		 3006 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3007 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 3008 brelse(bh);
3009 goto out_bdev;
3010 }
3011
bd81d8ee 3012 len = ext4_blocks_count(es);
ac27a0ec
DK		 3013 start = sb_block + 1;
3014 brelse(bh); /* we're done with the superblock */
3015
dab291af 3016 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3017 start, len, blocksize);
3018 if (!journal) {
617ba13b 3019 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 3020 goto out_bdev;
3021 }
3022 journal->j_private = sb;
3023 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3024 wait_on_buffer(journal->j_sb_buffer);
3025 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 3026 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 3027 goto out_journal;
3028 }
3029 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 3030 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 3031 "user (unsupported) - %d\n",
3032 be32_to_cpu(journal->j_superblock->s_nr_users));
3033 goto out_journal;
3034 }
617ba13b
MC		 3035 EXT4_SB(sb)->journal_bdev = bdev;
3036 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3037 return journal;
3038out_journal:
dab291af 3039 jbd2_journal_destroy(journal);
ac27a0ec 3040out_bdev:
617ba13b 3041 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3042 return NULL;
3043}
3044
617ba13b
MC		 3045static int ext4_load_journal(struct super_block *sb,
3046 struct ext4_super_block *es,
ac27a0ec
DK		 3047 unsigned long journal_devnum)
3048{
3049 journal_t *journal;
3050 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3051 dev_t journal_dev;
3052 int err = 0;
3053 int really_read_only;
3054
0390131b
FM		 3055 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3056
ac27a0ec
DK		 3057 if (journal_devnum &&
3058 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 3059 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 3060 "numbers have changed\n");
3061 journal_dev = new_decode_dev(journal_devnum);
3062 } else
3063 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3064
3065 really_read_only = bdev_read_only(sb->s_bdev);
3066
3067 /*
3068 * Are we loading a blank journal or performing recovery after a
3069 * crash? For recovery, we need to check in advance whether we
3070 * can get read-write access to the device.
3071 */
3072
617ba13b 3073 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3074 if (sb->s_flags & MS_RDONLY) {
617ba13b 3075 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 3076 "required on readonly filesystem.\n");
3077 if (really_read_only) {
617ba13b 3078 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 3079 "unavailable, cannot proceed.\n");
3080 return -EROFS;
3081 }
2b2d6d01
TT		 3082 printk(KERN_INFO "EXT4-fs: write access will "
3083 "be enabled during recovery.\n");
ac27a0ec
DK		 3084 }
3085 }
3086
3087 if (journal_inum && journal_dev) {
617ba13b 3088 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 3089 "and inode journals!\n");
3090 return -EINVAL;
3091 }
3092
3093 if (journal_inum) {
617ba13b 3094 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3095 return -EINVAL;
3096 } else {
617ba13b 3097 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3098 return -EINVAL;
3099 }
3100
4776004f
TT		 3101 if (journal->j_flags & JBD2_BARRIER)
3102 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
3103 else
3104 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
3105
ac27a0ec 3106 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3107 err = jbd2_journal_update_format(journal);
ac27a0ec 3108 if (err) {
617ba13b 3109 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 3110 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3111 return err;
3112 }
3113 }
3114
617ba13b 3115 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3116 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 3117 if (!err)
dab291af 3118 err = jbd2_journal_load(journal);
ac27a0ec
DK		 3119
3120 if (err) {
617ba13b 3121 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 3122 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3123 return err;
3124 }
3125
617ba13b
MC		 3126 EXT4_SB(sb)->s_journal = journal;
3127 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 3128
3129 if (journal_devnum &&
3130 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3131 es->s_journal_dev = cpu_to_le32(journal_devnum);
3132 sb->s_dirt = 1;
3133
3134 /* Make sure we flush the recovery flag to disk. */
617ba13b 3135 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 3136 }
3137
3138 return 0;
3139}
3140
c4be0c1d 3141static int ext4_commit_super(struct super_block *sb,
2b2d6d01 3142 struct ext4_super_block *es, int sync)
ac27a0ec 3143{
617ba13b 3144 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3145 int error = 0;
ac27a0ec
DK		 3146
3147 if (!sbh)
c4be0c1d 3148 return error;
914258bf
TT		 3149 if (buffer_write_io_error(sbh)) {
3150 /*
3151 * Oh, dear. A previous attempt to write the
3152 * superblock failed. This could happen because the
3153 * USB device was yanked out. Or it could happen to
3154 * be a transient write error and maybe the block will
3155 * be remapped. Nothing we can do but to retry the
3156 * write and hope for the best.
3157 */
abda1418 3158 printk(KERN_ERR "EXT4-fs: previous I/O error to "
914258bf
TT		 3159 "superblock detected for %s.\n", sb->s_id);
3160 clear_buffer_write_io_error(sbh);
3161 set_buffer_uptodate(sbh);
3162 }
ac27a0ec 3163 es->s_wtime = cpu_to_le32(get_seconds());
afc32f7e
TT		 3164 es->s_kbytes_written =
3165 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3166 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3167 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5d1b1b3f
AK		 3168 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3169 &EXT4_SB(sb)->s_freeblocks_counter));
3170 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3171 &EXT4_SB(sb)->s_freeinodes_counter));
3172
ac27a0ec
DK		 3173 BUFFER_TRACE(sbh, "marking dirty");
3174 mark_buffer_dirty(sbh);
914258bf 3175 if (sync) {
c4be0c1d
TS		 3176 error = sync_dirty_buffer(sbh);
3177 if (error)
3178 return error;
3179
3180 error = buffer_write_io_error(sbh);
3181 if (error) {
abda1418 3182 printk(KERN_ERR "EXT4-fs: I/O error while writing "
914258bf
TT		 3183 "superblock for %s.\n", sb->s_id);
3184 clear_buffer_write_io_error(sbh);
3185 set_buffer_uptodate(sbh);
3186 }
3187 }
c4be0c1d 3188 return error;
ac27a0ec
DK		 3189}
3190
3191
3192/*
3193 * Have we just finished recovery? If so, and if we are mounting (or
3194 * remounting) the filesystem readonly, then we will end up with a
3195 * consistent fs on disk. Record that fact.
3196 */
2b2d6d01
TT		 3197static void ext4_mark_recovery_complete(struct super_block *sb,
3198 struct ext4_super_block *es)
ac27a0ec 3199{
617ba13b 3200 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3201
0390131b
FM		 3202 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3203 BUG_ON(journal != NULL);
3204 return;
3205 }
dab291af 3206 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 3207 if (jbd2_journal_flush(journal) < 0)
3208 goto out;
3209
32c37730 3210 lock_super(sb);
617ba13b 3211 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 3212 sb->s_flags & MS_RDONLY) {
617ba13b 3213 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 3214 sb->s_dirt = 0;
617ba13b 3215 ext4_commit_super(sb, es, 1);
ac27a0ec 3216 }
32c37730 3217 unlock_super(sb);
7ffe1ea8
HK		 3218
3219out:
dab291af 3220 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 3221}
3222
3223/*
3224 * If we are mounting (or read-write remounting) a filesystem whose journal
3225 * has recorded an error from a previous lifetime, move that error to the
3226 * main filesystem now.
3227 */
2b2d6d01
TT		 3228static void ext4_clear_journal_err(struct super_block *sb,
3229 struct ext4_super_block *es)
ac27a0ec
DK		 3230{
3231 journal_t *journal;
3232 int j_errno;
3233 const char *errstr;
3234
0390131b
FM		 3235 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3236
617ba13b 3237 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 3238
3239 /*
3240 * Now check for any error status which may have been recorded in the
617ba13b 3241 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 3242 */
3243
dab291af 3244 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 3245 if (j_errno) {
3246 char nbuf[16];
3247
617ba13b 3248 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 3249 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 3250 "from previous mount: %s", errstr);
46e665e9 3251 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 3252 "filesystem check.");
3253
617ba13b
MC		 3254 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3255 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2b2d6d01 3256 ext4_commit_super(sb, es, 1);
ac27a0ec 3257
dab291af 3258 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 3259 }
3260}
3261
3262/*
3263 * Force the running and committing transactions to commit,
3264 * and wait on the commit.
3265 */
617ba13b 3266int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 3267{
3268 journal_t *journal;
0390131b 3269 int ret = 0;
ac27a0ec
DK		 3270
3271 if (sb->s_flags & MS_RDONLY)
3272 return 0;
3273
617ba13b 3274 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 3275 if (journal) {
3276 sb->s_dirt = 0;
3277 ret = ext4_journal_force_commit(journal);
3278 }
3279
ac27a0ec
DK		 3280 return ret;
3281}
3282
3283/*
617ba13b 3284 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec 3285 * have to propagate any other updates to the superblock on disk at this
14ce0cb4
TT		 3286 * point. (We can probably nuke this function altogether, and remove
3287 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
ac27a0ec 3288 */
2b2d6d01 3289static void ext4_write_super(struct super_block *sb)
ac27a0ec 3290{
0390131b
FM		 3291 if (EXT4_SB(sb)->s_journal) {
3292 if (mutex_trylock(&sb->s_lock) != 0)
3293 BUG();
3294 sb->s_dirt = 0;
3295 } else {
3296 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3297 }
ac27a0ec
DK		 3298}
3299
617ba13b 3300static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 3301{
14ce0cb4 3302 int ret = 0;
9eddacf9 3303 tid_t target;
ac27a0ec 3304
ede86cc4 3305 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
ac27a0ec 3306 sb->s_dirt = 0;
0390131b 3307 if (EXT4_SB(sb)->s_journal) {
9eddacf9
JK		 3308 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal,
3309 &target)) {
3310 if (wait)
3311 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal,
3312 target);
3313 }
0390131b
FM		 3314 } else {
3315 ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
3316 }
14ce0cb4 3317 return ret;
ac27a0ec
DK		 3318}
3319
3320/*
3321 * LVM calls this function before a (read-only) snapshot is created. This
3322 * gives us a chance to flush the journal completely and mark the fs clean.
3323 */
c4be0c1d 3324static int ext4_freeze(struct super_block *sb)
ac27a0ec 3325{
c4be0c1d
TS		 3326 int error = 0;
3327 journal_t *journal;
ac27a0ec
DK		 3328 sb->s_dirt = 0;
3329
3330 if (!(sb->s_flags & MS_RDONLY)) {
c4be0c1d 3331 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3332
0390131b
FM		 3333 if (journal) {
3334 /* Now we set up the journal barrier. */
3335 jbd2_journal_lock_updates(journal);
7ffe1ea8 3336
0390131b
FM		 3337 /*
3338 * We don't want to clear needs_recovery flag when we
3339 * failed to flush the journal.
3340 */
c4be0c1d
TS		 3341 error = jbd2_journal_flush(journal);
3342 if (error < 0)
3343 goto out;
0390131b 3344 }
ac27a0ec
DK		 3345
3346 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b 3347 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
c4be0c1d
TS		 3348 error = ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3349 if (error)
3350 goto out;
ac27a0ec 3351 }
c4be0c1d
TS		 3352 return 0;
3353out:
3354 jbd2_journal_unlock_updates(journal);
3355 return error;
ac27a0ec
DK		 3356}
3357
3358/*
3359 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3360 * flag here, even though the filesystem is not technically dirty yet.
3361 */
c4be0c1d 3362static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 3363{
0390131b 3364 if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
ac27a0ec
DK		 3365 lock_super(sb);
3366 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 3367 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3368 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 3369 unlock_super(sb);
dab291af 3370 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec 3371 }
c4be0c1d 3372 return 0;
ac27a0ec
DK		 3373}
3374
2b2d6d01 3375static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3376{
2b2d6d01 3377 struct ext4_super_block *es;
617ba13b
MC		 3378 struct ext4_sb_info *sbi = EXT4_SB(sb);
3379 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3380 unsigned long old_sb_flags;
617ba13b 3381 struct ext4_mount_options old_opts;
8a266467 3382 ext4_group_t g;
b3881f74 3383 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 3384 int err;
3385#ifdef CONFIG_QUOTA
3386 int i;
3387#endif
3388
3389 /* Store the original options */
3390 old_sb_flags = sb->s_flags;
3391 old_opts.s_mount_opt = sbi->s_mount_opt;
3392 old_opts.s_resuid = sbi->s_resuid;
3393 old_opts.s_resgid = sbi->s_resgid;
3394 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 3395 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3396 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3397#ifdef CONFIG_QUOTA
3398 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3399 for (i = 0; i < MAXQUOTAS; i++)
3400 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3401#endif
b3881f74
TT		 3402 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3403 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 3404
3405 /*
3406 * Allow the "check" option to be passed as a remount option.
3407 */
b3881f74
TT		 3408 if (!parse_options(data, sb, NULL, &journal_ioprio,
3409 &n_blocks_count, 1)) {
ac27a0ec
DK		 3410 err = -EINVAL;
3411 goto restore_opts;
3412 }
3413
617ba13b 3414 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
46e665e9 3415 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 3416
3417 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 3418 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3419
3420 es = sbi->s_es;
3421
b3881f74 3422 if (sbi->s_journal) {
0390131b 3423 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 3424 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3425 }
ac27a0ec
DK		 3426
3427 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3428 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 3429 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 3430 err = -EROFS;
3431 goto restore_opts;
3432 }
3433
3434 if (*flags & MS_RDONLY) {
3435 /*
3436 * First of all, the unconditional stuff we have to do
3437 * to disable replay of the journal when we next remount
3438 */
3439 sb->s_flags |= MS_RDONLY;
3440
3441 /*
3442 * OK, test if we are remounting a valid rw partition
3443 * readonly, and if so set the rdonly flag and then
3444 * mark the partition as valid again.
3445 */
617ba13b
MC		 3446 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3447 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3448 es->s_state = cpu_to_le16(sbi->s_mount_state);
3449
32c37730
JK		 3450 /*
3451 * We have to unlock super so that we can wait for
3452 * transactions.
3453 */
0390131b
FM		 3454 if (sbi->s_journal) {
3455 unlock_super(sb);
3456 ext4_mark_recovery_complete(sb, es);
3457 lock_super(sb);
3458 }
ac27a0ec 3459 } else {
3a06d778 3460 int ret;
617ba13b
MC		 3461 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3462 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3463 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec 3464 "remount RDWR because of unsupported "
3a06d778
AK		 3465 "optional features (%x).\n", sb->s_id,
3466 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3467 ~EXT4_FEATURE_RO_COMPAT_SUPP));
ac27a0ec
DK		 3468 err = -EROFS;
3469 goto restore_opts;
3470 }
ead6596b 3471
8a266467
TT		 3472 /*
3473 * Make sure the group descriptor checksums
3474 * are sane. If they aren't, refuse to
3475 * remount r/w.
3476 */
3477 for (g = 0; g < sbi->s_groups_count; g++) {
3478 struct ext4_group_desc *gdp =
3479 ext4_get_group_desc(sb, g, NULL);
3480
3481 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3482 printk(KERN_ERR
3483 "EXT4-fs: ext4_remount: "
a9df9a49 3484 "Checksum for group %u failed (%u!=%u)\n",
8a266467
TT		 3485 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3486 le16_to_cpu(gdp->bg_checksum));
3487 err = -EINVAL;
3488 goto restore_opts;
3489 }
3490 }
3491
ead6596b
ES		 3492 /*
3493 * If we have an unprocessed orphan list hanging
3494 * around from a previously readonly bdev mount,
3495 * require a full umount/remount for now.
3496 */
3497 if (es->s_last_orphan) {
3498 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3499 "remount RDWR because of unprocessed "
3500 "orphan inode list. Please "
3501 "umount/remount instead.\n",
3502 sb->s_id);
3503 err = -EINVAL;
3504 goto restore_opts;
3505 }
3506
ac27a0ec
DK		 3507 /*
3508 * Mounting a RDONLY partition read-write, so reread
3509 * and store the current valid flag. (It may have
3510 * been changed by e2fsck since we originally mounted
3511 * the partition.)
3512 */
0390131b
FM		 3513 if (sbi->s_journal)
3514 ext4_clear_journal_err(sb, es);
ac27a0ec 3515 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3516 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3517 goto restore_opts;
2b2d6d01 3518 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3519 sb->s_flags &= ~MS_RDONLY;
3520 }
3521 }
0390131b
FM		 3522 if (sbi->s_journal == NULL)
3523 ext4_commit_super(sb, es, 1);
3524
ac27a0ec
DK		 3525#ifdef CONFIG_QUOTA
3526 /* Release old quota file names */
3527 for (i = 0; i < MAXQUOTAS; i++)
3528 if (old_opts.s_qf_names[i] &&
3529 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3530 kfree(old_opts.s_qf_names[i]);
3531#endif
3532 return 0;
3533restore_opts:
3534 sb->s_flags = old_sb_flags;
3535 sbi->s_mount_opt = old_opts.s_mount_opt;
3536 sbi->s_resuid = old_opts.s_resuid;
3537 sbi->s_resgid = old_opts.s_resgid;
3538 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 3539 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3540 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 3541#ifdef CONFIG_QUOTA
3542 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3543 for (i = 0; i < MAXQUOTAS; i++) {
3544 if (sbi->s_qf_names[i] &&
3545 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3546 kfree(sbi->s_qf_names[i]);
3547 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3548 }
3549#endif
3550 return err;
3551}
3552
2b2d6d01 3553static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3554{
3555 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3556 struct ext4_sb_info *sbi = EXT4_SB(sb);
3557 struct ext4_super_block *es = sbi->s_es;
960cc398 3558 u64 fsid;
ac27a0ec 3559
5e70030d
BP		 3560 if (test_opt(sb, MINIX_DF)) {
3561 sbi->s_overhead_last = 0;
6bc9feff 3562 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 3563 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 3564 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3565 smp_rmb();
3566
3567 /*
5e70030d
BP		 3568 * Compute the overhead (FS structures). This is constant
3569 * for a given filesystem unless the number of block groups
3570 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3571 */
3572
3573 /*
3574 * All of the blocks before first_data_block are
3575 * overhead
3576 */
3577 overhead = le32_to_cpu(es->s_first_data_block);
3578
3579 /*
3580 * Add the overhead attributed to the superblock and
3581 * block group descriptors. If the sparse superblocks
3582 * feature is turned on, then not all groups have this.
3583 */
3584 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3585 overhead += ext4_bg_has_super(sb, i) +
3586 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3587 cond_resched();
3588 }
3589
3590 /*
3591 * Every block group has an inode bitmap, a block
3592 * bitmap, and an inode table.
3593 */
5e70030d
BP		 3594 overhead += ngroups * (2 + sbi->s_itb_per_group);
3595 sbi->s_overhead_last = overhead;
3596 smp_wmb();
6bc9feff 3597 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3598 }
3599
617ba13b 3600 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3601 buf->f_bsize = sb->s_blocksize;
5e70030d 3602 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3603 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3604 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
308ba3ec 3605 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3606 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3607 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3608 buf->f_bavail = 0;
3609 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3610 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3611 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3612 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3613 fsid = le64_to_cpup((void *)es->s_uuid) ^
3614 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3615 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3616 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 3617 return 0;
3618}
3619
3620/* Helper function for writing quotas on sync - we need to start transaction before quota file
3621 * is locked for write. Otherwise the are possible deadlocks:
3622 * Process 1 Process 2
617ba13b 3623 * ext4_create() quota_sync()
a269eb18
JK		 3624 * jbd2_journal_start() write_dquot()
3625 * vfs_dq_init() down(dqio_mutex)
dab291af 3626 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3627 *
3628 */
3629
3630#ifdef CONFIG_QUOTA
3631
3632static inline struct inode *dquot_to_inode(struct dquot *dquot)
3633{
3634 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3635}
3636
617ba13b 3637static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3638{
3639 int ret, err;
3640 handle_t *handle;
3641 struct inode *inode;
3642
3643 inode = dquot_to_inode(dquot);
617ba13b
MC		 3644 handle = ext4_journal_start(inode,
3645 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3646 if (IS_ERR(handle))
3647 return PTR_ERR(handle);
3648 ret = dquot_commit(dquot);
617ba13b 3649 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3650 if (!ret)
3651 ret = err;
3652 return ret;
3653}
3654
617ba13b 3655static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3656{
3657 int ret, err;
3658 handle_t *handle;
3659
617ba13b
MC		 3660 handle = ext4_journal_start(dquot_to_inode(dquot),
3661 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3662 if (IS_ERR(handle))
3663 return PTR_ERR(handle);
3664 ret = dquot_acquire(dquot);
617ba13b 3665 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3666 if (!ret)
3667 ret = err;
3668 return ret;
3669}
3670
617ba13b 3671static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3672{
3673 int ret, err;
3674 handle_t *handle;
3675
617ba13b
MC		 3676 handle = ext4_journal_start(dquot_to_inode(dquot),
3677 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3678 if (IS_ERR(handle)) {
3679 /* Release dquot anyway to avoid endless cycle in dqput() */
3680 dquot_release(dquot);
ac27a0ec 3681 return PTR_ERR(handle);
9c3013e9 3682 }
ac27a0ec 3683 ret = dquot_release(dquot);
617ba13b 3684 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3685 if (!ret)
3686 ret = err;
3687 return ret;
3688}
3689
617ba13b 3690static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3691{
2c8be6b2 3692 /* Are we journaling quotas? */
617ba13b
MC		 3693 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3694 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3695 dquot_mark_dquot_dirty(dquot);
617ba13b 3696 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3697 } else {
3698 return dquot_mark_dquot_dirty(dquot);
3699 }
3700}
3701
617ba13b 3702static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3703{
3704 int ret, err;
3705 handle_t *handle;
3706
3707 /* Data block + inode block */
617ba13b 3708 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3709 if (IS_ERR(handle))
3710 return PTR_ERR(handle);
3711 ret = dquot_commit_info(sb, type);
617ba13b 3712 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3713 if (!ret)
3714 ret = err;
3715 return ret;
3716}
3717
3718/*
3719 * Turn on quotas during mount time - we need to find
3720 * the quota file and such...
3721 */
617ba13b 3722static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3723{
617ba13b
MC		 3724 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3725 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3726}
3727
3728/*
3729 * Standard function to be called on quota_on
3730 */
617ba13b 3731static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8264613d 3732 char *name, int remount)
ac27a0ec
DK		 3733{
3734 int err;
8264613d 3735 struct path path;
ac27a0ec
DK		 3736
3737 if (!test_opt(sb, QUOTA))
3738 return -EINVAL;
8264613d 3739 /* When remounting, no checks are needed and in fact, name is NULL */
0623543b 3740 if (remount)
8264613d 3741 return vfs_quota_on(sb, type, format_id, name, remount);
0623543b 3742
8264613d 3743 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 3744 if (err)
3745 return err;
0623543b 3746
ac27a0ec 3747 /* Quotafile not on the same filesystem? */
8264613d
AV		 3748 if (path.mnt->mnt_sb != sb) {
3749 path_put(&path);
ac27a0ec
DK		 3750 return -EXDEV;
3751 }
0623543b
JK		 3752 /* Journaling quota? */
3753 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3754 /* Quotafile not in fs root? */
8264613d 3755 if (path.dentry->d_parent != sb->s_root)
0623543b
JK		 3756 printk(KERN_WARNING
3757 "EXT4-fs: Quota file not on filesystem root. "
3758 "Journaled quota will not work.\n");
2b2d6d01 3759 }
0623543b
JK		 3760
3761 /*
3762 * When we journal data on quota file, we have to flush journal to see
3763 * all updates to the file when we bypass pagecache...
3764 */
0390131b
FM		 3765 if (EXT4_SB(sb)->s_journal &&
3766 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 3767 /*
3768 * We don't need to lock updates but journal_flush() could
3769 * otherwise be livelocked...
3770 */
3771 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3772 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3773 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3774 if (err) {
8264613d 3775 path_put(&path);
7ffe1ea8
HK		 3776 return err;
3777 }
0623543b
JK		 3778 }
3779
8264613d
AV		 3780 err = vfs_quota_on_path(sb, type, format_id, &path);
3781 path_put(&path);
77e69dac 3782 return err;
ac27a0ec
DK		 3783}
3784
3785/* Read data from quotafile - avoid pagecache and such because we cannot afford
3786 * acquiring the locks... As quota files are never truncated and quota code
3787 * itself serializes the operations (and noone else should touch the files)
3788 * we don't have to be afraid of races */
617ba13b 3789static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3790 size_t len, loff_t off)
3791{
3792 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3793 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3794 int err = 0;
3795 int offset = off & (sb->s_blocksize - 1);
3796 int tocopy;
3797 size_t toread;
3798 struct buffer_head *bh;
3799 loff_t i_size = i_size_read(inode);
3800
3801 if (off > i_size)
3802 return 0;
3803 if (off+len > i_size)
3804 len = i_size-off;
3805 toread = len;
3806 while (toread > 0) {
3807 tocopy = sb->s_blocksize - offset < toread ?
3808 sb->s_blocksize - offset : toread;
617ba13b 3809 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3810 if (err)
3811 return err;
3812 if (!bh) /* A hole? */
3813 memset(data, 0, tocopy);
3814 else
3815 memcpy(data, bh->b_data+offset, tocopy);
3816 brelse(bh);
3817 offset = 0;
3818 toread -= tocopy;
3819 data += tocopy;
3820 blk++;
3821 }
3822 return len;
3823}
3824
3825/* Write to quotafile (we know the transaction is already started and has
3826 * enough credits) */
617ba13b 3827static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3828 const char *data, size_t len, loff_t off)
3829{
3830 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3831 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3832 int err = 0;
3833 int offset = off & (sb->s_blocksize - 1);
3834 int tocopy;
617ba13b 3835 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3836 size_t towrite = len;
3837 struct buffer_head *bh;
3838 handle_t *handle = journal_current_handle();
3839
0390131b 3840 if (EXT4_SB(sb)->s_journal && !handle) {
e5f8eab8 3841 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
9c3013e9
JK		 3842 " cancelled because transaction is not started.\n",
3843 (unsigned long long)off, (unsigned long long)len);
3844 return -EIO;
3845 }
ac27a0ec
DK		 3846 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3847 while (towrite > 0) {
3848 tocopy = sb->s_blocksize - offset < towrite ?
3849 sb->s_blocksize - offset : towrite;
617ba13b 3850 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3851 if (!bh)
3852 goto out;
3853 if (journal_quota) {
617ba13b 3854 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3855 if (err) {
3856 brelse(bh);
3857 goto out;
3858 }
3859 }
3860 lock_buffer(bh);
3861 memcpy(bh->b_data+offset, data, tocopy);
3862 flush_dcache_page(bh->b_page);
3863 unlock_buffer(bh);
3864 if (journal_quota)
0390131b 3865 err = ext4_handle_dirty_metadata(handle, NULL, bh);
ac27a0ec
DK		 3866 else {
3867 /* Always do at least ordered writes for quotas */
678aaf48 3868 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3869 mark_buffer_dirty(bh);
3870 }
3871 brelse(bh);
3872 if (err)
3873 goto out;
3874 offset = 0;
3875 towrite -= tocopy;
3876 data += tocopy;
3877 blk++;
3878 }
3879out:
4d04e4fb
JK		 3880 if (len == towrite) {
3881 mutex_unlock(&inode->i_mutex);
ac27a0ec 3882 return err;
4d04e4fb 3883 }
ac27a0ec
DK		 3884 if (inode->i_size < off+len-towrite) {
3885 i_size_write(inode, off+len-towrite);
617ba13b 3886 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3887 }
ac27a0ec 3888 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3889 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3890 mutex_unlock(&inode->i_mutex);
3891 return len - towrite;
3892}
3893
3894#endif
3895
617ba13b 3896static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3897 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3898{
617ba13b 3899 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3900}
3901
03010a33
TT		 3902static struct file_system_type ext4_fs_type = {
3903 .owner = THIS_MODULE,
3904 .name = "ext4",
3905 .get_sb = ext4_get_sb,
3906 .kill_sb = kill_block_super,
3907 .fs_flags = FS_REQUIRES_DEV,
3908};
3909
3910#ifdef CONFIG_EXT4DEV_COMPAT
3911static int ext4dev_get_sb(struct file_system_type *fs_type,
3912 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3913{
3914 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3915 "to mount using ext4\n");
3916 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3917 "will go away by 2.6.31\n");
3918 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3919}
3920
617ba13b 3921static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3922 .owner = THIS_MODULE,
617ba13b 3923 .name = "ext4dev",
03010a33 3924 .get_sb = ext4dev_get_sb,
ac27a0ec
DK		 3925 .kill_sb = kill_block_super,
3926 .fs_flags = FS_REQUIRES_DEV,
3927};
03010a33
TT		 3928MODULE_ALIAS("ext4dev");
3929#endif
ac27a0ec 3930
617ba13b 3931static int __init init_ext4_fs(void)
ac27a0ec 3932{
c9de560d
AT		 3933 int err;
3934
3197ebdb
TT		 3935 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3936 if (!ext4_kset)
3937 return -ENOMEM;
9f6200bb 3938 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 3939 err = init_ext4_mballoc();
ac27a0ec
DK		 3940 if (err)
3941 return err;
c9de560d
AT		 3942
3943 err = init_ext4_xattr();
3944 if (err)
3945 goto out2;
ac27a0ec
DK		 3946 err = init_inodecache();
3947 if (err)
3948 goto out1;
03010a33 3949 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 3950 if (err)
3951 goto out;
03010a33
TT		 3952#ifdef CONFIG_EXT4DEV_COMPAT
3953 err = register_filesystem(&ext4dev_fs_type);
3954 if (err) {
3955 unregister_filesystem(&ext4_fs_type);
3956 goto out;
3957 }
3958#endif
ac27a0ec
DK		 3959 return 0;
3960out:
3961 destroy_inodecache();
3962out1:
617ba13b 3963 exit_ext4_xattr();
c9de560d
AT		 3964out2:
3965 exit_ext4_mballoc();
ac27a0ec
DK		 3966 return err;
3967}
3968
617ba13b 3969static void __exit exit_ext4_fs(void)
ac27a0ec 3970{
03010a33
TT		 3971 unregister_filesystem(&ext4_fs_type);
3972#ifdef CONFIG_EXT4DEV_COMPAT
617ba13b 3973 unregister_filesystem(&ext4dev_fs_type);
03010a33 3974#endif
ac27a0ec 3975 destroy_inodecache();
617ba13b 3976 exit_ext4_xattr();
c9de560d 3977 exit_ext4_mballoc();
9f6200bb 3978 remove_proc_entry("fs/ext4", NULL);
3197ebdb 3979 kset_unregister(ext4_kset);
ac27a0ec
DK		 3980}
3981
3982MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 3983MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 3984MODULE_LICENSE("GPL");
617ba13b
MC		 3985module_init(init_ext4_fs)
3986module_exit(exit_ext4_fs)