]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/ext4/super.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
percpu counter: clean up percpu_counter_sum_and_set()
[net-next-2.6.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
dab291af 23#include <linux/jbd2.h>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
27#include <linux/parser.h>
28#include <linux/smp_lock.h>
29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
1330593e 37#include <linux/log2.h>
717d50e4 38#include <linux/crc16.h>
ac27a0ec
DK		 39#include <asm/uaccess.h>
40
3dcf5451
CH		 41#include "ext4.h"
42#include "ext4_jbd2.h"
ac27a0ec
DK		 43#include "xattr.h"
44#include "acl.h"
45#include "namei.h"
717d50e4 46#include "group.h"
ac27a0ec 47
617ba13b 48static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 49 unsigned long journal_devnum);
617ba13b 50static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 51 unsigned int);
2b2d6d01
TT		 52static void ext4_commit_super(struct super_block *sb,
53 struct ext4_super_block *es, int sync);
54static void ext4_mark_recovery_complete(struct super_block *sb,
55 struct ext4_super_block *es);
56static void ext4_clear_journal_err(struct super_block *sb,
57 struct ext4_super_block *es);
617ba13b 58static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 59static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 60 char nbuf[16]);
2b2d6d01
TT		 61static int ext4_remount(struct super_block *sb, int *flags, char *data);
62static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
617ba13b 63static void ext4_unlockfs(struct super_block *sb);
2b2d6d01 64static void ext4_write_super(struct super_block *sb);
617ba13b 65static void ext4_write_super_lockfs(struct super_block *sb);
ac27a0ec 66
bd81d8ee 67
8fadc143
AR		 68ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
69 struct ext4_group_desc *bg)
bd81d8ee 70{
3a14589c 71 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 72 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
3a14589c 73 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 74}
75
8fadc143
AR		 76ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
77 struct ext4_group_desc *bg)
bd81d8ee 78{
5272f837 79 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 80 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 81 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 82}
83
8fadc143
AR		 84ext4_fsblk_t ext4_inode_table(struct super_block *sb,
85 struct ext4_group_desc *bg)
bd81d8ee 86{
5272f837 87 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 88 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 89 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 90}
91
8fadc143
AR		 92void ext4_block_bitmap_set(struct super_block *sb,
93 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 94{
3a14589c 95 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 96 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
97 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 98}
99
8fadc143
AR		 100void ext4_inode_bitmap_set(struct super_block *sb,
101 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 102{
5272f837 103 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 104 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
105 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 106}
107
8fadc143
AR		 108void ext4_inode_table_set(struct super_block *sb,
109 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 110{
5272f837 111 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 112 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
113 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 114}
115
ac27a0ec 116/*
dab291af 117 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 118 *
119 * The only special thing we need to do here is to make sure that all
120 * journal_end calls result in the superblock being marked dirty, so
121 * that sync() will call the filesystem's write_super callback if
122 * appropriate.
123 */
617ba13b 124handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 125{
126 journal_t *journal;
127
128 if (sb->s_flags & MS_RDONLY)
129 return ERR_PTR(-EROFS);
130
131 /* Special case here: if the journal has aborted behind our
132 * backs (eg. EIO in the commit thread), then we still need to
133 * take the FS itself readonly cleanly. */
617ba13b 134 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 135 if (is_journal_aborted(journal)) {
46e665e9 136 ext4_abort(sb, __func__,
ac27a0ec
DK		 137 "Detected aborted journal");
138 return ERR_PTR(-EROFS);
139 }
140
dab291af 141 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 142}
143
144/*
145 * The only special thing we need to do here is to make sure that all
dab291af 146 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 147 * that sync() will call the filesystem's write_super callback if
148 * appropriate.
149 */
617ba13b 150int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 151{
152 struct super_block *sb;
153 int err;
154 int rc;
155
156 sb = handle->h_transaction->t_journal->j_private;
157 err = handle->h_err;
dab291af 158 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 159
160 if (!err)
161 err = rc;
162 if (err)
617ba13b 163 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 164 return err;
165}
166
617ba13b 167void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 168 struct buffer_head *bh, handle_t *handle, int err)
169{
170 char nbuf[16];
617ba13b 171 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec
DK		 172
173 if (bh)
174 BUFFER_TRACE(bh, "abort");
175
176 if (!handle->h_err)
177 handle->h_err = err;
178
179 if (is_handle_aborted(handle))
180 return;
181
182 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
183 caller, errstr, err_fn);
184
dab291af 185 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 186}
187
188/* Deal with the reporting of failure conditions on a filesystem such as
189 * inconsistencies detected or read IO failures.
190 *
191 * On ext2, we can store the error state of the filesystem in the
617ba13b 192 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 193 * write ordering constraints on the superblock which prevent us from
194 * writing it out straight away; and given that the journal is about to
195 * be aborted, we can't rely on the current, or future, transactions to
196 * write out the superblock safely.
197 *
dab291af 198 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 199 * the journal instead. On recovery, the journal will compain about
200 * that error until we've noted it down and cleared it.
201 */
202
617ba13b 203static void ext4_handle_error(struct super_block *sb)
ac27a0ec 204{
617ba13b 205 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 206
617ba13b
MC		 207 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
208 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 209
210 if (sb->s_flags & MS_RDONLY)
211 return;
212
2b2d6d01 213 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 214 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 215
617ba13b 216 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ac27a0ec 217 if (journal)
dab291af 218 jbd2_journal_abort(journal, -EIO);
ac27a0ec 219 }
2b2d6d01
TT		 220 if (test_opt(sb, ERRORS_RO)) {
221 printk(KERN_CRIT "Remounting filesystem read-only\n");
ac27a0ec
DK		 222 sb->s_flags |= MS_RDONLY;
223 }
617ba13b 224 ext4_commit_super(sb, es, 1);
ac27a0ec 225 if (test_opt(sb, ERRORS_PANIC))
617ba13b 226 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 227 sb->s_id);
228}
229
2b2d6d01
TT		 230void ext4_error(struct super_block *sb, const char *function,
231 const char *fmt, ...)
ac27a0ec
DK		 232{
233 va_list args;
234
235 va_start(args, fmt);
2b2d6d01 236 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 237 vprintk(fmt, args);
238 printk("\n");
239 va_end(args);
240
617ba13b 241 ext4_handle_error(sb);
ac27a0ec
DK		 242}
243
2b2d6d01 244static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 245 char nbuf[16])
246{
247 char *errstr = NULL;
248
249 switch (errno) {
250 case -EIO:
251 errstr = "IO failure";
252 break;
253 case -ENOMEM:
254 errstr = "Out of memory";
255 break;
256 case -EROFS:
dab291af 257 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
ac27a0ec
DK		 258 errstr = "Journal has aborted";
259 else
260 errstr = "Readonly filesystem";
261 break;
262 default:
263 /* If the caller passed in an extra buffer for unknown
264 * errors, textualise them now. Else we just return
265 * NULL. */
266 if (nbuf) {
267 /* Check for truncated error codes... */
268 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
269 errstr = nbuf;
270 }
271 break;
272 }
273
274 return errstr;
275}
276
617ba13b 277/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 278 * automatically and invokes the appropriate error response. */
279
2b2d6d01 280void __ext4_std_error(struct super_block *sb, const char *function, int errno)
ac27a0ec
DK		 281{
282 char nbuf[16];
283 const char *errstr;
284
285 /* Special case: if the error is EROFS, and we're not already
286 * inside a transaction, then there's really no point in logging
287 * an error. */
288 if (errno == -EROFS && journal_current_handle() == NULL &&
289 (sb->s_flags & MS_RDONLY))
290 return;
291
617ba13b 292 errstr = ext4_decode_error(sb, errno, nbuf);
2b2d6d01
TT		 293 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
294 sb->s_id, function, errstr);
ac27a0ec 295
617ba13b 296 ext4_handle_error(sb);
ac27a0ec
DK		 297}
298
299/*
617ba13b 300 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 301 * abort function may be used to deal with unrecoverable failures such
302 * as journal IO errors or ENOMEM at a critical moment in log management.
303 *
304 * We unconditionally force the filesystem into an ABORT|READONLY state,
305 * unless the error response on the fs has been set to panic in which
306 * case we take the easy way out and panic immediately.
307 */
308
2b2d6d01
TT		 309void ext4_abort(struct super_block *sb, const char *function,
310 const char *fmt, ...)
ac27a0ec
DK		 311{
312 va_list args;
313
2b2d6d01 314 printk(KERN_CRIT "ext4_abort called.\n");
ac27a0ec
DK		 315
316 va_start(args, fmt);
2b2d6d01 317 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 318 vprintk(fmt, args);
319 printk("\n");
320 va_end(args);
321
322 if (test_opt(sb, ERRORS_PANIC))
617ba13b 323 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 324
325 if (sb->s_flags & MS_RDONLY)
326 return;
327
328 printk(KERN_CRIT "Remounting filesystem read-only\n");
617ba13b 329 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 330 sb->s_flags |= MS_RDONLY;
617ba13b 331 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
dab291af 332 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 333}
334
2b2d6d01
TT		 335void ext4_warning(struct super_block *sb, const char *function,
336 const char *fmt, ...)
ac27a0ec
DK		 337{
338 va_list args;
339
340 va_start(args, fmt);
617ba13b 341 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 342 sb->s_id, function);
343 vprintk(fmt, args);
344 printk("\n");
345 va_end(args);
346}
347
617ba13b 348void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 349{
617ba13b 350 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 351
617ba13b 352 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 353 return;
354
46e665e9 355 ext4_warning(sb, __func__,
ac27a0ec
DK		 356 "updating to rev %d because of new feature flag, "
357 "running e2fsck is recommended",
617ba13b 358 EXT4_DYNAMIC_REV);
ac27a0ec 359
617ba13b
MC		 360 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
361 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
362 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 363 /* leave es->s_feature_*compat flags alone */
364 /* es->s_uuid will be set by e2fsck if empty */
365
366 /*
367 * The rest of the superblock fields should be zero, and if not it
368 * means they are likely already in use, so leave them alone. We
369 * can leave it up to e2fsck to clean up any inconsistencies there.
370 */
371}
372
99e6f829
AK		 373int ext4_update_compat_feature(handle_t *handle,
374 struct super_block *sb, __u32 compat)
375{
376 int err = 0;
377 if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
378 err = ext4_journal_get_write_access(handle,
379 EXT4_SB(sb)->s_sbh);
380 if (err)
381 return err;
382 EXT4_SET_COMPAT_FEATURE(sb, compat);
383 sb->s_dirt = 1;
384 handle->h_sync = 1;
385 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
386 "call ext4_journal_dirty_met adata");
387 err = ext4_journal_dirty_metadata(handle,
388 EXT4_SB(sb)->s_sbh);
389 }
390 return err;
391}
392
393int ext4_update_rocompat_feature(handle_t *handle,
394 struct super_block *sb, __u32 rocompat)
395{
396 int err = 0;
397 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
398 err = ext4_journal_get_write_access(handle,
399 EXT4_SB(sb)->s_sbh);
400 if (err)
401 return err;
402 EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
403 sb->s_dirt = 1;
404 handle->h_sync = 1;
405 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
406 "call ext4_journal_dirty_met adata");
407 err = ext4_journal_dirty_metadata(handle,
408 EXT4_SB(sb)->s_sbh);
409 }
410 return err;
411}
412
413int ext4_update_incompat_feature(handle_t *handle,
414 struct super_block *sb, __u32 incompat)
415{
416 int err = 0;
417 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
418 err = ext4_journal_get_write_access(handle,
419 EXT4_SB(sb)->s_sbh);
420 if (err)
421 return err;
422 EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
423 sb->s_dirt = 1;
424 handle->h_sync = 1;
425 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
426 "call ext4_journal_dirty_met adata");
427 err = ext4_journal_dirty_metadata(handle,
428 EXT4_SB(sb)->s_sbh);
429 }
430 return err;
431}
432
ac27a0ec
DK		 433/*
434 * Open the external journal device
435 */
617ba13b 436static struct block_device *ext4_blkdev_get(dev_t dev)
ac27a0ec
DK		 437{
438 struct block_device *bdev;
439 char b[BDEVNAME_SIZE];
440
441 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
442 if (IS_ERR(bdev))
443 goto fail;
444 return bdev;
445
446fail:
617ba13b 447 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
ac27a0ec
DK		 448 __bdevname(dev, b), PTR_ERR(bdev));
449 return NULL;
450}
451
452/*
453 * Release the journal device
454 */
617ba13b 455static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 456{
457 bd_release(bdev);
458 return blkdev_put(bdev);
459}
460
617ba13b 461static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 462{
463 struct block_device *bdev;
464 int ret = -ENODEV;
465
466 bdev = sbi->journal_bdev;
467 if (bdev) {
617ba13b 468 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 469 sbi->journal_bdev = NULL;
470 }
471 return ret;
472}
473
474static inline struct inode *orphan_list_entry(struct list_head *l)
475{
617ba13b 476 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 477}
478
617ba13b 479static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 480{
481 struct list_head *l;
482
483 printk(KERN_ERR "sb orphan head is %d\n",
484 le32_to_cpu(sbi->s_es->s_last_orphan));
485
486 printk(KERN_ERR "sb_info orphan list:\n");
487 list_for_each(l, &sbi->s_orphan) {
488 struct inode *inode = orphan_list_entry(l);
489 printk(KERN_ERR " "
490 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
491 inode->i_sb->s_id, inode->i_ino, inode,
492 inode->i_mode, inode->i_nlink,
493 NEXT_ORPHAN(inode));
494 }
495}
496
2b2d6d01 497static void ext4_put_super(struct super_block *sb)
ac27a0ec 498{
617ba13b
MC		 499 struct ext4_sb_info *sbi = EXT4_SB(sb);
500 struct ext4_super_block *es = sbi->s_es;
ac27a0ec
DK		 501 int i;
502
c9de560d 503 ext4_mb_release(sb);
a86c6181 504 ext4_ext_release(sb);
617ba13b 505 ext4_xattr_put_super(sb);
dab291af 506 jbd2_journal_destroy(sbi->s_journal);
47b4a50b 507 sbi->s_journal = NULL;
ac27a0ec 508 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 509 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec
DK		 510 es->s_state = cpu_to_le16(sbi->s_mount_state);
511 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
512 mark_buffer_dirty(sbi->s_sbh);
617ba13b 513 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 514 }
515
516 for (i = 0; i < sbi->s_gdb_count; i++)
517 brelse(sbi->s_group_desc[i]);
518 kfree(sbi->s_group_desc);
772cb7c8 519 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 520 percpu_counter_destroy(&sbi->s_freeblocks_counter);
521 percpu_counter_destroy(&sbi->s_freeinodes_counter);
522 percpu_counter_destroy(&sbi->s_dirs_counter);
523 brelse(sbi->s_sbh);
524#ifdef CONFIG_QUOTA
525 for (i = 0; i < MAXQUOTAS; i++)
526 kfree(sbi->s_qf_names[i]);
527#endif
528
529 /* Debugging code just in case the in-memory inode orphan list
530 * isn't empty. The on-disk one can be non-empty if we've
531 * detected an error and taken the fs readonly, but the
532 * in-memory list had better be clean by this point. */
533 if (!list_empty(&sbi->s_orphan))
534 dump_orphan_list(sb, sbi);
535 J_ASSERT(list_empty(&sbi->s_orphan));
536
f98393a6 537 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 538 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
539 /*
540 * Invalidate the journal device's buffers. We don't want them
541 * floating about in memory - the physical journal device may
542 * hotswapped, and it breaks the `ro-after' testing code.
543 */
544 sync_blockdev(sbi->journal_bdev);
f98393a6 545 invalidate_bdev(sbi->journal_bdev);
617ba13b 546 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 547 }
548 sb->s_fs_info = NULL;
549 kfree(sbi);
550 return;
551}
552
e18b890b 553static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 554
555/*
556 * Called inside transaction, so use GFP_NOFS
557 */
617ba13b 558static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 559{
617ba13b 560 struct ext4_inode_info *ei;
ac27a0ec 561
e6b4f8da 562 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 563 if (!ei)
564 return NULL;
617ba13b
MC		 565#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
566 ei->i_acl = EXT4_ACL_NOT_CACHED;
567 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 568#endif
569 ei->i_block_alloc_info = NULL;
570 ei->vfs_inode.i_version = 1;
91246c00 571 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 572 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 573 INIT_LIST_HEAD(&ei->i_prealloc_list);
574 spin_lock_init(&ei->i_prealloc_lock);
678aaf48 575 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 576 ei->i_reserved_data_blocks = 0;
577 ei->i_reserved_meta_blocks = 0;
578 ei->i_allocated_meta_blocks = 0;
579 ei->i_delalloc_reserved_flag = 0;
580 spin_lock_init(&(ei->i_block_reservation_lock));
ac27a0ec
DK		 581 return &ei->vfs_inode;
582}
583
617ba13b 584static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 585{
9f7dd93d
VA		 586 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
587 printk("EXT4 Inode %p: orphan list check failed!\n",
588 EXT4_I(inode));
589 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
590 EXT4_I(inode), sizeof(struct ext4_inode_info),
591 true);
592 dump_stack();
593 }
617ba13b 594 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 595}
596
51cc5068 597static void init_once(void *foo)
ac27a0ec 598{
617ba13b 599 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 600
a35afb83 601 INIT_LIST_HEAD(&ei->i_orphan);
617ba13b 602#ifdef CONFIG_EXT4DEV_FS_XATTR
a35afb83 603 init_rwsem(&ei->xattr_sem);
ac27a0ec 604#endif
0e855ac8 605 init_rwsem(&ei->i_data_sem);
a35afb83 606 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 607}
608
609static int init_inodecache(void)
610{
617ba13b
MC		 611 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
612 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 613 0, (SLAB_RECLAIM_ACCOUNT|
614 SLAB_MEM_SPREAD),
20c2df83 615 init_once);
617ba13b 616 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 617 return -ENOMEM;
618 return 0;
619}
620
621static void destroy_inodecache(void)
622{
617ba13b 623 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 624}
625
617ba13b 626static void ext4_clear_inode(struct inode *inode)
ac27a0ec 627{
617ba13b
MC		 628 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
629#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
630 if (EXT4_I(inode)->i_acl &&
631 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
632 posix_acl_release(EXT4_I(inode)->i_acl);
633 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
634 }
635 if (EXT4_I(inode)->i_default_acl &&
636 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
637 posix_acl_release(EXT4_I(inode)->i_default_acl);
638 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 639 }
640#endif
617ba13b
MC		 641 ext4_discard_reservation(inode);
642 EXT4_I(inode)->i_block_alloc_info = NULL;
ac27a0ec
DK		 643 if (unlikely(rsv))
644 kfree(rsv);
678aaf48
JK		 645 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
646 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 647}
648
2b2d6d01
TT		 649static inline void ext4_show_quota_options(struct seq_file *seq,
650 struct super_block *sb)
ac27a0ec
DK		 651{
652#if defined(CONFIG_QUOTA)
617ba13b 653 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 654
655 if (sbi->s_jquota_fmt)
656 seq_printf(seq, ",jqfmt=%s",
657 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
658
659 if (sbi->s_qf_names[USRQUOTA])
660 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
661
662 if (sbi->s_qf_names[GRPQUOTA])
663 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
664
617ba13b 665 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 666 seq_puts(seq, ",usrquota");
667
617ba13b 668 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 669 seq_puts(seq, ",grpquota");
670#endif
671}
672
d9c9bef1
MS		 673/*
674 * Show an option if
675 * - it's set to a non-default value OR
676 * - if the per-sb default is different from the global default
677 */
617ba13b 678static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 679{
aa22df2c
AK		 680 int def_errors;
681 unsigned long def_mount_opts;
ac27a0ec 682 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 683 struct ext4_sb_info *sbi = EXT4_SB(sb);
684 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 685
686 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 687 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 688
689 if (sbi->s_sb_block != 1)
690 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
691 if (test_opt(sb, MINIX_DF))
692 seq_puts(seq, ",minixdf");
aa22df2c 693 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 694 seq_puts(seq, ",grpid");
695 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
696 seq_puts(seq, ",nogrpid");
697 if (sbi->s_resuid != EXT4_DEF_RESUID ||
698 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
699 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
700 }
701 if (sbi->s_resgid != EXT4_DEF_RESGID ||
702 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
703 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
704 }
bb4f397a 705 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 706 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 707 def_errors == EXT4_ERRORS_CONTINUE) {
708 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 709 }
710 }
aa22df2c 711 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 712 seq_puts(seq, ",errors=continue");
aa22df2c 713 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 714 seq_puts(seq, ",errors=panic");
aa22df2c 715 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 716 seq_puts(seq, ",nouid32");
aa22df2c 717 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 718 seq_puts(seq, ",debug");
719 if (test_opt(sb, OLDALLOC))
720 seq_puts(seq, ",oldalloc");
07620f69 721#ifdef CONFIG_EXT4DEV_FS_XATTR
aa22df2c
AK		 722 if (test_opt(sb, XATTR_USER) &&
723 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 724 seq_puts(seq, ",user_xattr");
725 if (!test_opt(sb, XATTR_USER) &&
726 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
727 seq_puts(seq, ",nouser_xattr");
728 }
729#endif
07620f69 730#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
aa22df2c 731 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 732 seq_puts(seq, ",acl");
733 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
734 seq_puts(seq, ",noacl");
735#endif
736 if (!test_opt(sb, RESERVATION))
737 seq_puts(seq, ",noreservation");
738 if (sbi->s_commit_interval) {
739 seq_printf(seq, ",commit=%u",
740 (unsigned) (sbi->s_commit_interval / HZ));
741 }
571640ca
ES		 742 /*
743 * We're changing the default of barrier mount option, so
744 * let's always display its mount state so it's clear what its
745 * status is.
746 */
747 seq_puts(seq, ",barrier=");
748 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 749 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
750 seq_puts(seq, ",journal_async_commit");
d9c9bef1
MS		 751 if (test_opt(sb, NOBH))
752 seq_puts(seq, ",nobh");
753 if (!test_opt(sb, EXTENTS))
754 seq_puts(seq, ",noextents");
3dbd0ede
AK		 755 if (!test_opt(sb, MBALLOC))
756 seq_puts(seq, ",nomballoc");
25ec56b5
JNC		 757 if (test_opt(sb, I_VERSION))
758 seq_puts(seq, ",i_version");
dd919b98
AK		 759 if (!test_opt(sb, DELALLOC))
760 seq_puts(seq, ",nodelalloc");
761
ac27a0ec 762
cb45bbe4
MS		 763 if (sbi->s_stripe)
764 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 765 /*
766 * journal mode get enabled in different ways
767 * So just print the value even if we didn't specify it
768 */
617ba13b 769 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 770 seq_puts(seq, ",data=journal");
617ba13b 771 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 772 seq_puts(seq, ",data=ordered");
617ba13b 773 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 774 seq_puts(seq, ",data=writeback");
775
617ba13b 776 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 777 return 0;
778}
779
780
1b961ac0
CH		 781static struct inode *ext4_nfs_get_inode(struct super_block *sb,
782 u64 ino, u32 generation)
ac27a0ec 783{
ac27a0ec 784 struct inode *inode;
ac27a0ec 785
617ba13b 786 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 787 return ERR_PTR(-ESTALE);
617ba13b 788 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 789 return ERR_PTR(-ESTALE);
790
791 /* iget isn't really right if the inode is currently unallocated!!
792 *
617ba13b 793 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 794 * deleted, so we should be safe.
795 *
796 * Currently we don't know the generation for parent directory, so
797 * a generation of 0 means "accept any"
798 */
1d1fe1ee
DH		 799 inode = ext4_iget(sb, ino);
800 if (IS_ERR(inode))
801 return ERR_CAST(inode);
802 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 803 iput(inode);
804 return ERR_PTR(-ESTALE);
805 }
1b961ac0
CH		 806
807 return inode;
808}
809
810static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
811 int fh_len, int fh_type)
812{
813 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
814 ext4_nfs_get_inode);
815}
816
817static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
818 int fh_len, int fh_type)
819{
820 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
821 ext4_nfs_get_inode);
ac27a0ec
DK		 822}
823
824#ifdef CONFIG_QUOTA
2b2d6d01
TT		 825#define QTYPE2NAME(t) ((t) == USRQUOTA?"user":"group")
826#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 827
617ba13b
MC		 828static int ext4_dquot_initialize(struct inode *inode, int type);
829static int ext4_dquot_drop(struct inode *inode);
830static int ext4_write_dquot(struct dquot *dquot);
831static int ext4_acquire_dquot(struct dquot *dquot);
832static int ext4_release_dquot(struct dquot *dquot);
833static int ext4_mark_dquot_dirty(struct dquot *dquot);
834static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 835static int ext4_quota_on(struct super_block *sb, int type, int format_id,
836 char *path, int remount);
617ba13b
MC		 837static int ext4_quota_on_mount(struct super_block *sb, int type);
838static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 839 size_t len, loff_t off);
617ba13b 840static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 841 const char *data, size_t len, loff_t off);
842
617ba13b
MC		 843static struct dquot_operations ext4_quota_operations = {
844 .initialize = ext4_dquot_initialize,
845 .drop = ext4_dquot_drop,
ac27a0ec
DK		 846 .alloc_space = dquot_alloc_space,
847 .alloc_inode = dquot_alloc_inode,
848 .free_space = dquot_free_space,
849 .free_inode = dquot_free_inode,
850 .transfer = dquot_transfer,
617ba13b
MC		 851 .write_dquot = ext4_write_dquot,
852 .acquire_dquot = ext4_acquire_dquot,
853 .release_dquot = ext4_release_dquot,
854 .mark_dirty = ext4_mark_dquot_dirty,
855 .write_info = ext4_write_info
ac27a0ec
DK		 856};
857
617ba13b
MC		 858static struct quotactl_ops ext4_qctl_operations = {
859 .quota_on = ext4_quota_on,
ac27a0ec
DK		 860 .quota_off = vfs_quota_off,
861 .quota_sync = vfs_quota_sync,
862 .get_info = vfs_get_dqinfo,
863 .set_info = vfs_set_dqinfo,
864 .get_dqblk = vfs_get_dqblk,
865 .set_dqblk = vfs_set_dqblk
866};
867#endif
868
ee9b6d61 869static const struct super_operations ext4_sops = {
617ba13b
MC		 870 .alloc_inode = ext4_alloc_inode,
871 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 872 .write_inode = ext4_write_inode,
873 .dirty_inode = ext4_dirty_inode,
874 .delete_inode = ext4_delete_inode,
875 .put_super = ext4_put_super,
876 .write_super = ext4_write_super,
877 .sync_fs = ext4_sync_fs,
878 .write_super_lockfs = ext4_write_super_lockfs,
879 .unlockfs = ext4_unlockfs,
880 .statfs = ext4_statfs,
881 .remount_fs = ext4_remount,
882 .clear_inode = ext4_clear_inode,
883 .show_options = ext4_show_options,
ac27a0ec 884#ifdef CONFIG_QUOTA
617ba13b
MC		 885 .quota_read = ext4_quota_read,
886 .quota_write = ext4_quota_write,
ac27a0ec
DK		 887#endif
888};
889
39655164 890static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 891 .fh_to_dentry = ext4_fh_to_dentry,
892 .fh_to_parent = ext4_fh_to_parent,
617ba13b 893 .get_parent = ext4_get_parent,
ac27a0ec
DK		 894};
895
896enum {
897 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
898 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
899 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
900 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
901 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
902 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
818d276c 903 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec
DK		 904 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
905 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
906 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
907 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
25ec56b5 908 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
dd919b98 909 Opt_mballoc, Opt_nomballoc, Opt_stripe, Opt_delalloc, Opt_nodelalloc,
ac27a0ec
DK		 910};
911
912static match_table_t tokens = {
913 {Opt_bsd_df, "bsddf"},
914 {Opt_minix_df, "minixdf"},
915 {Opt_grpid, "grpid"},
916 {Opt_grpid, "bsdgroups"},
917 {Opt_nogrpid, "nogrpid"},
918 {Opt_nogrpid, "sysvgroups"},
919 {Opt_resgid, "resgid=%u"},
920 {Opt_resuid, "resuid=%u"},
921 {Opt_sb, "sb=%u"},
922 {Opt_err_cont, "errors=continue"},
923 {Opt_err_panic, "errors=panic"},
924 {Opt_err_ro, "errors=remount-ro"},
925 {Opt_nouid32, "nouid32"},
926 {Opt_nocheck, "nocheck"},
927 {Opt_nocheck, "check=none"},
928 {Opt_debug, "debug"},
929 {Opt_oldalloc, "oldalloc"},
930 {Opt_orlov, "orlov"},
931 {Opt_user_xattr, "user_xattr"},
932 {Opt_nouser_xattr, "nouser_xattr"},
933 {Opt_acl, "acl"},
934 {Opt_noacl, "noacl"},
935 {Opt_reservation, "reservation"},
936 {Opt_noreservation, "noreservation"},
937 {Opt_noload, "noload"},
938 {Opt_nobh, "nobh"},
939 {Opt_bh, "bh"},
940 {Opt_commit, "commit=%u"},
941 {Opt_journal_update, "journal=update"},
942 {Opt_journal_inum, "journal=%u"},
943 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 944 {Opt_journal_checksum, "journal_checksum"},
945 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 946 {Opt_abort, "abort"},
947 {Opt_data_journal, "data=journal"},
948 {Opt_data_ordered, "data=ordered"},
949 {Opt_data_writeback, "data=writeback"},
950 {Opt_offusrjquota, "usrjquota="},
951 {Opt_usrjquota, "usrjquota=%s"},
952 {Opt_offgrpjquota, "grpjquota="},
953 {Opt_grpjquota, "grpjquota=%s"},
954 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
955 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
956 {Opt_grpquota, "grpquota"},
957 {Opt_noquota, "noquota"},
958 {Opt_quota, "quota"},
959 {Opt_usrquota, "usrquota"},
960 {Opt_barrier, "barrier=%u"},
a86c6181 961 {Opt_extents, "extents"},
1e2462f9 962 {Opt_noextents, "noextents"},
25ec56b5 963 {Opt_i_version, "i_version"},
c9de560d
AT		 964 {Opt_mballoc, "mballoc"},
965 {Opt_nomballoc, "nomballoc"},
966 {Opt_stripe, "stripe=%u"},
ac27a0ec 967 {Opt_resize, "resize"},
64769240 968 {Opt_delalloc, "delalloc"},
dd919b98 969 {Opt_nodelalloc, "nodelalloc"},
f3f12faa 970 {Opt_err, NULL},
ac27a0ec
DK		 971};
972
617ba13b 973static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 974{
617ba13b 975 ext4_fsblk_t sb_block;
ac27a0ec
DK		 976 char *options = (char *) *data;
977
978 if (!options || strncmp(options, "sb=", 3) != 0)
979 return 1; /* Default location */
980 options += 3;
617ba13b 981 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 982 sb_block = simple_strtoul(options, &options, 0);
983 if (*options && *options != ',') {
617ba13b 984 printk("EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 985 (char *) *data);
986 return 1;
987 }
988 if (*options == ',')
989 options++;
990 *data = (void *) options;
991 return sb_block;
992}
993
2b2d6d01
TT		 994static int parse_options(char *options, struct super_block *sb,
995 unsigned int *inum, unsigned long *journal_devnum,
996 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 997{
617ba13b 998 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 999 char *p;
ac27a0ec
DK		 1000 substring_t args[MAX_OPT_ARGS];
1001 int data_opt = 0;
1002 int option;
1003#ifdef CONFIG_QUOTA
dfc5d03f 1004 int qtype, qfmt;
ac27a0ec
DK		 1005 char *qname;
1006#endif
c07651b5 1007 ext4_fsblk_t last_block;
ac27a0ec
DK		 1008
1009 if (!options)
1010 return 1;
1011
2b2d6d01 1012 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1013 int token;
1014 if (!*p)
1015 continue;
1016
1017 token = match_token(p, tokens, args);
1018 switch (token) {
1019 case Opt_bsd_df:
2b2d6d01 1020 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1021 break;
1022 case Opt_minix_df:
2b2d6d01 1023 set_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1024 break;
1025 case Opt_grpid:
2b2d6d01 1026 set_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1027 break;
1028 case Opt_nogrpid:
2b2d6d01 1029 clear_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1030 break;
1031 case Opt_resuid:
1032 if (match_int(&args[0], &option))
1033 return 0;
1034 sbi->s_resuid = option;
1035 break;
1036 case Opt_resgid:
1037 if (match_int(&args[0], &option))
1038 return 0;
1039 sbi->s_resgid = option;
1040 break;
1041 case Opt_sb:
1042 /* handled by get_sb_block() instead of here */
1043 /* *sb_block = match_int(&args[0]); */
1044 break;
1045 case Opt_err_panic:
2b2d6d01
TT		 1046 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1047 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1048 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1049 break;
1050 case Opt_err_ro:
2b2d6d01
TT		 1051 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1052 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1053 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1054 break;
1055 case Opt_err_cont:
2b2d6d01
TT		 1056 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1057 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1058 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1059 break;
1060 case Opt_nouid32:
2b2d6d01 1061 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec
DK		 1062 break;
1063 case Opt_nocheck:
2b2d6d01 1064 clear_opt(sbi->s_mount_opt, CHECK);
ac27a0ec
DK		 1065 break;
1066 case Opt_debug:
2b2d6d01 1067 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1068 break;
1069 case Opt_oldalloc:
2b2d6d01 1070 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1071 break;
1072 case Opt_orlov:
2b2d6d01 1073 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1074 break;
617ba13b 1075#ifdef CONFIG_EXT4DEV_FS_XATTR
ac27a0ec 1076 case Opt_user_xattr:
2b2d6d01 1077 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1078 break;
1079 case Opt_nouser_xattr:
2b2d6d01 1080 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1081 break;
1082#else
1083 case Opt_user_xattr:
1084 case Opt_nouser_xattr:
617ba13b 1085 printk("EXT4 (no)user_xattr options not supported\n");
ac27a0ec
DK		 1086 break;
1087#endif
617ba13b 1088#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
ac27a0ec
DK		 1089 case Opt_acl:
1090 set_opt(sbi->s_mount_opt, POSIX_ACL);
1091 break;
1092 case Opt_noacl:
1093 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1094 break;
1095#else
1096 case Opt_acl:
1097 case Opt_noacl:
617ba13b 1098 printk("EXT4 (no)acl options not supported\n");
ac27a0ec
DK		 1099 break;
1100#endif
1101 case Opt_reservation:
1102 set_opt(sbi->s_mount_opt, RESERVATION);
1103 break;
1104 case Opt_noreservation:
1105 clear_opt(sbi->s_mount_opt, RESERVATION);
1106 break;
1107 case Opt_journal_update:
1108 /* @@@ FIXME */
1109 /* Eventually we will want to be able to create
1110 a journal file here. For now, only allow the
1111 user to specify an existing inode to be the
1112 journal file. */
1113 if (is_remount) {
617ba13b 1114 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1115 "journal on remount\n");
1116 return 0;
1117 }
2b2d6d01 1118 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec
DK		 1119 break;
1120 case Opt_journal_inum:
1121 if (is_remount) {
617ba13b 1122 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1123 "journal on remount\n");
1124 return 0;
1125 }
1126 if (match_int(&args[0], &option))
1127 return 0;
1128 *inum = option;
1129 break;
1130 case Opt_journal_dev:
1131 if (is_remount) {
617ba13b 1132 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1133 "journal on remount\n");
1134 return 0;
1135 }
1136 if (match_int(&args[0], &option))
1137 return 0;
1138 *journal_devnum = option;
1139 break;
818d276c
GS		 1140 case Opt_journal_checksum:
1141 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1142 break;
1143 case Opt_journal_async_commit:
1144 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1145 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1146 break;
ac27a0ec 1147 case Opt_noload:
2b2d6d01 1148 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1149 break;
1150 case Opt_commit:
1151 if (match_int(&args[0], &option))
1152 return 0;
1153 if (option < 0)
1154 return 0;
1155 if (option == 0)
cd02ff0b 1156 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1157 sbi->s_commit_interval = HZ * option;
1158 break;
1159 case Opt_data_journal:
617ba13b 1160 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1161 goto datacheck;
1162 case Opt_data_ordered:
617ba13b 1163 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1164 goto datacheck;
1165 case Opt_data_writeback:
617ba13b 1166 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1167 datacheck:
1168 if (is_remount) {
617ba13b 1169 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1170 != data_opt) {
1171 printk(KERN_ERR
617ba13b 1172 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1173 "mode on remount\n");
1174 return 0;
1175 }
1176 } else {
617ba13b 1177 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1178 sbi->s_mount_opt |= data_opt;
1179 }
1180 break;
1181#ifdef CONFIG_QUOTA
1182 case Opt_usrjquota:
1183 qtype = USRQUOTA;
1184 goto set_qf_name;
1185 case Opt_grpjquota:
1186 qtype = GRPQUOTA;
1187set_qf_name:
dfc5d03f
JK		 1188 if ((sb_any_quota_enabled(sb) ||
1189 sb_any_quota_suspended(sb)) &&
1190 !sbi->s_qf_names[qtype]) {
ac27a0ec 1191 printk(KERN_ERR
2c8be6b2 1192 "EXT4-fs: Cannot change journaled "
ac27a0ec
DK		 1193 "quota options when quota turned on.\n");
1194 return 0;
1195 }
1196 qname = match_strdup(&args[0]);
1197 if (!qname) {
1198 printk(KERN_ERR
617ba13b 1199 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1200 "storing quotafile name.\n");
1201 return 0;
1202 }
1203 if (sbi->s_qf_names[qtype] &&
1204 strcmp(sbi->s_qf_names[qtype], qname)) {
1205 printk(KERN_ERR
617ba13b 1206 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1207 "specified.\n", QTYPE2NAME(qtype));
1208 kfree(qname);
1209 return 0;
1210 }
1211 sbi->s_qf_names[qtype] = qname;
1212 if (strchr(sbi->s_qf_names[qtype], '/')) {
1213 printk(KERN_ERR
617ba13b 1214 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1215 "filesystem root.\n");
1216 kfree(sbi->s_qf_names[qtype]);
1217 sbi->s_qf_names[qtype] = NULL;
1218 return 0;
1219 }
1220 set_opt(sbi->s_mount_opt, QUOTA);
1221 break;
1222 case Opt_offusrjquota:
1223 qtype = USRQUOTA;
1224 goto clear_qf_name;
1225 case Opt_offgrpjquota:
1226 qtype = GRPQUOTA;
1227clear_qf_name:
dfc5d03f
JK		 1228 if ((sb_any_quota_enabled(sb) ||
1229 sb_any_quota_suspended(sb)) &&
1230 sbi->s_qf_names[qtype]) {
617ba13b 1231 printk(KERN_ERR "EXT4-fs: Cannot change "
2c8be6b2 1232 "journaled quota options when "
ac27a0ec
DK		 1233 "quota turned on.\n");
1234 return 0;
1235 }
1236 /*
1237 * The space will be released later when all options
1238 * are confirmed to be correct
1239 */
1240 sbi->s_qf_names[qtype] = NULL;
1241 break;
1242 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1243 qfmt = QFMT_VFS_OLD;
1244 goto set_qf_format;
ac27a0ec 1245 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1246 qfmt = QFMT_VFS_V0;
1247set_qf_format:
1248 if ((sb_any_quota_enabled(sb) ||
1249 sb_any_quota_suspended(sb)) &&
1250 sbi->s_jquota_fmt != qfmt) {
1251 printk(KERN_ERR "EXT4-fs: Cannot change "
1252 "journaled quota options when "
1253 "quota turned on.\n");
1254 return 0;
1255 }
1256 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1257 break;
1258 case Opt_quota:
1259 case Opt_usrquota:
1260 set_opt(sbi->s_mount_opt, QUOTA);
1261 set_opt(sbi->s_mount_opt, USRQUOTA);
1262 break;
1263 case Opt_grpquota:
1264 set_opt(sbi->s_mount_opt, QUOTA);
1265 set_opt(sbi->s_mount_opt, GRPQUOTA);
1266 break;
1267 case Opt_noquota:
1268 if (sb_any_quota_enabled(sb)) {
617ba13b 1269 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1270 "options when quota turned on.\n");
1271 return 0;
1272 }
1273 clear_opt(sbi->s_mount_opt, QUOTA);
1274 clear_opt(sbi->s_mount_opt, USRQUOTA);
1275 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1276 break;
1277#else
1278 case Opt_quota:
1279 case Opt_usrquota:
1280 case Opt_grpquota:
cd59e7b9
JK		 1281 printk(KERN_ERR
1282 "EXT4-fs: quota options not supported.\n");
1283 break;
ac27a0ec
DK		 1284 case Opt_usrjquota:
1285 case Opt_grpjquota:
1286 case Opt_offusrjquota:
1287 case Opt_offgrpjquota:
1288 case Opt_jqfmt_vfsold:
1289 case Opt_jqfmt_vfsv0:
1290 printk(KERN_ERR
cd59e7b9 1291 "EXT4-fs: journaled quota options not "
ac27a0ec
DK		 1292 "supported.\n");
1293 break;
1294 case Opt_noquota:
1295 break;
1296#endif
1297 case Opt_abort:
1298 set_opt(sbi->s_mount_opt, ABORT);
1299 break;
1300 case Opt_barrier:
1301 if (match_int(&args[0], &option))
1302 return 0;
1303 if (option)
1304 set_opt(sbi->s_mount_opt, BARRIER);
1305 else
1306 clear_opt(sbi->s_mount_opt, BARRIER);
1307 break;
1308 case Opt_ignore:
1309 break;
1310 case Opt_resize:
1311 if (!is_remount) {
617ba13b 1312 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1313 "for remount\n");
1314 return 0;
1315 }
1316 if (match_int(&args[0], &option) != 0)
1317 return 0;
1318 *n_blocks_count = option;
1319 break;
1320 case Opt_nobh:
1321 set_opt(sbi->s_mount_opt, NOBH);
1322 break;
1323 case Opt_bh:
1324 clear_opt(sbi->s_mount_opt, NOBH);
1325 break;
a86c6181 1326 case Opt_extents:
e4079a11
ES		 1327 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1328 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1329 ext4_warning(sb, __func__,
1330 "extents feature not enabled "
1331 "on this filesystem, use tune2fs\n");
1332 return 0;
1333 }
2b2d6d01 1334 set_opt(sbi->s_mount_opt, EXTENTS);
a86c6181 1335 break;
1e2462f9 1336 case Opt_noextents:
c07651b5
AK		 1337 /*
1338 * When e2fsprogs support resizing an already existing
1339 * ext3 file system to greater than 2**32 we need to
1340 * add support to block allocator to handle growing
1341 * already existing block mapped inode so that blocks
1342 * allocated for them fall within 2**32
1343 */
1344 last_block = ext4_blocks_count(sbi->s_es) - 1;
1345 if (last_block > 0xffffffffULL) {
1346 printk(KERN_ERR "EXT4-fs: Filesystem too "
1347 "large to mount with "
1348 "-o noextents options\n");
1349 return 0;
1350 }
2b2d6d01 1351 clear_opt(sbi->s_mount_opt, EXTENTS);
1e2462f9 1352 break;
25ec56b5
JNC		 1353 case Opt_i_version:
1354 set_opt(sbi->s_mount_opt, I_VERSION);
1355 sb->s_flags |= MS_I_VERSION;
1356 break;
dd919b98
AK		 1357 case Opt_nodelalloc:
1358 clear_opt(sbi->s_mount_opt, DELALLOC);
1359 break;
c9de560d
AT		 1360 case Opt_mballoc:
1361 set_opt(sbi->s_mount_opt, MBALLOC);
1362 break;
1363 case Opt_nomballoc:
1364 clear_opt(sbi->s_mount_opt, MBALLOC);
1365 break;
1366 case Opt_stripe:
1367 if (match_int(&args[0], &option))
1368 return 0;
1369 if (option < 0)
1370 return 0;
1371 sbi->s_stripe = option;
1372 break;
64769240
AT		 1373 case Opt_delalloc:
1374 set_opt(sbi->s_mount_opt, DELALLOC);
1375 break;
ac27a0ec 1376 default:
2b2d6d01
TT		 1377 printk(KERN_ERR
1378 "EXT4-fs: Unrecognized mount option \"%s\" "
1379 "or missing value\n", p);
ac27a0ec
DK		 1380 return 0;
1381 }
1382 }
1383#ifdef CONFIG_QUOTA
1384 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1385 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1386 sbi->s_qf_names[USRQUOTA])
1387 clear_opt(sbi->s_mount_opt, USRQUOTA);
1388
617ba13b 1389 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1390 sbi->s_qf_names[GRPQUOTA])
1391 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1392
1393 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1394 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1395 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1396 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1397 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1398 "format mixing.\n");
1399 return 0;
1400 }
1401
1402 if (!sbi->s_jquota_fmt) {
2c8be6b2 1403 printk(KERN_ERR "EXT4-fs: journaled quota format "
ac27a0ec
DK		 1404 "not specified.\n");
1405 return 0;
1406 }
1407 } else {
1408 if (sbi->s_jquota_fmt) {
2c8be6b2
JK		 1409 printk(KERN_ERR "EXT4-fs: journaled quota format "
1410 "specified with no journaling "
ac27a0ec
DK		 1411 "enabled.\n");
1412 return 0;
1413 }
1414 }
1415#endif
1416 return 1;
1417}
1418
617ba13b 1419static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1420 int read_only)
1421{
617ba13b 1422 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1423 int res = 0;
1424
617ba13b 1425 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2b2d6d01
TT		 1426 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1427 "forcing read-only mode\n");
ac27a0ec
DK		 1428 res = MS_RDONLY;
1429 }
1430 if (read_only)
1431 return res;
617ba13b 1432 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2b2d6d01
TT		 1433 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1434 "running e2fsck is recommended\n");
617ba13b 1435 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
2b2d6d01
TT		 1436 printk(KERN_WARNING
1437 "EXT4-fs warning: mounting fs with errors, "
1438 "running e2fsck is recommended\n");
ac27a0ec
DK		 1439 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1440 le16_to_cpu(es->s_mnt_count) >=
1441 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2b2d6d01
TT		 1442 printk(KERN_WARNING
1443 "EXT4-fs warning: maximal mount count reached, "
1444 "running e2fsck is recommended\n");
ac27a0ec
DK		 1445 else if (le32_to_cpu(es->s_checkinterval) &&
1446 (le32_to_cpu(es->s_lastcheck) +
1447 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2b2d6d01
TT		 1448 printk(KERN_WARNING
1449 "EXT4-fs warning: checktime reached, "
1450 "running e2fsck is recommended\n");
ac27a0ec
DK		 1451#if 0
1452 /* @@@ We _will_ want to clear the valid bit if we find
63f57933
AM		 1453 * inconsistencies, to force a fsck at reboot. But for
1454 * a plain journaled filesystem we can keep it set as
1455 * valid forever! :)
1456 */
216c34b2 1457 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec
DK		 1458#endif
1459 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1460 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1461 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1462 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b
MC		 1463 ext4_update_dynamic_rev(sb);
1464 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1465
617ba13b 1466 ext4_commit_super(sb, es, 1);
ac27a0ec 1467 if (test_opt(sb, DEBUG))
617ba13b 1468 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
ac27a0ec
DK		 1469 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1470 sb->s_blocksize,
1471 sbi->s_groups_count,
617ba13b
MC		 1472 EXT4_BLOCKS_PER_GROUP(sb),
1473 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1474 sbi->s_mount_opt);
1475
617ba13b
MC		 1476 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1477 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
ac27a0ec
DK		 1478 char b[BDEVNAME_SIZE];
1479
1480 printk("external journal on %s\n",
617ba13b 1481 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
ac27a0ec
DK		 1482 } else {
1483 printk("internal journal\n");
1484 }
1485 return res;
1486}
1487
772cb7c8
JS		 1488static int ext4_fill_flex_info(struct super_block *sb)
1489{
1490 struct ext4_sb_info *sbi = EXT4_SB(sb);
1491 struct ext4_group_desc *gdp = NULL;
1492 struct buffer_head *bh;
1493 ext4_group_t flex_group_count;
1494 ext4_group_t flex_group;
1495 int groups_per_flex = 0;
1496 __u64 block_bitmap = 0;
1497 int i;
1498
1499 if (!sbi->s_es->s_log_groups_per_flex) {
1500 sbi->s_log_groups_per_flex = 0;
1501 return 1;
1502 }
1503
1504 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1505 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1506
1507 flex_group_count = (sbi->s_groups_count + groups_per_flex - 1) /
1508 groups_per_flex;
ec05e868 1509 sbi->s_flex_groups = kzalloc(flex_group_count *
772cb7c8
JS		 1510 sizeof(struct flex_groups), GFP_KERNEL);
1511 if (sbi->s_flex_groups == NULL) {
ec05e868
LZ		 1512 printk(KERN_ERR "EXT4-fs: not enough memory for "
1513 "%lu flex groups\n", flex_group_count);
772cb7c8
JS		 1514 goto failed;
1515 }
772cb7c8
JS		 1516
1517 gdp = ext4_get_group_desc(sb, 1, &bh);
1518 block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1519
1520 for (i = 0; i < sbi->s_groups_count; i++) {
1521 gdp = ext4_get_group_desc(sb, i, &bh);
1522
1523 flex_group = ext4_flex_group(sbi, i);
1524 sbi->s_flex_groups[flex_group].free_inodes +=
1525 le16_to_cpu(gdp->bg_free_inodes_count);
1526 sbi->s_flex_groups[flex_group].free_blocks +=
1527 le16_to_cpu(gdp->bg_free_blocks_count);
1528 }
1529
1530 return 1;
1531failed:
1532 return 0;
1533}
1534
717d50e4
AD		 1535__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1536 struct ext4_group_desc *gdp)
1537{
1538 __u16 crc = 0;
1539
1540 if (sbi->s_es->s_feature_ro_compat &
1541 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1542 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1543 __le32 le_group = cpu_to_le32(block_group);
1544
1545 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1546 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1547 crc = crc16(crc, (__u8 *)gdp, offset);
1548 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1549 /* for checksum of struct ext4_group_desc do the rest...*/
1550 if ((sbi->s_es->s_feature_incompat &
1551 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1552 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1553 crc = crc16(crc, (__u8 *)gdp + offset,
1554 le16_to_cpu(sbi->s_es->s_desc_size) -
1555 offset);
1556 }
1557
1558 return cpu_to_le16(crc);
1559}
1560
1561int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1562 struct ext4_group_desc *gdp)
1563{
1564 if ((sbi->s_es->s_feature_ro_compat &
1565 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1566 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1567 return 0;
1568
1569 return 1;
1570}
1571
ac27a0ec 1572/* Called at mount-time, super-block is locked */
197cd65a 1573static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1574{
617ba13b
MC		 1575 struct ext4_sb_info *sbi = EXT4_SB(sb);
1576 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1577 ext4_fsblk_t last_block;
bd81d8ee
LV		 1578 ext4_fsblk_t block_bitmap;
1579 ext4_fsblk_t inode_bitmap;
1580 ext4_fsblk_t inode_table;
ce421581 1581 int flexbg_flag = 0;
fd2d4291 1582 ext4_group_t i;
ac27a0ec 1583
ce421581
JS		 1584 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1585 flexbg_flag = 1;
1586
617ba13b 1587 ext4_debug ("Checking group descriptors");
ac27a0ec 1588
197cd65a
AM		 1589 for (i = 0; i < sbi->s_groups_count; i++) {
1590 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1591
ce421581 1592 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1593 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1594 else
1595 last_block = first_block +
617ba13b 1596 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1597
8fadc143 1598 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1599 if (block_bitmap < first_block || block_bitmap > last_block) {
c19204b0
JB		 1600 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1601 "Block bitmap for group %lu not in group "
1602 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1603 return 0;
1604 }
8fadc143 1605 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1606 if (inode_bitmap < first_block || inode_bitmap > last_block) {
c19204b0
JB		 1607 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1608 "Inode bitmap for group %lu not in group "
1609 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1610 return 0;
1611 }
8fadc143 1612 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1613 if (inode_table < first_block ||
2b2d6d01 1614 inode_table + sbi->s_itb_per_group - 1 > last_block) {
c19204b0
JB		 1615 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1616 "Inode table for group %lu not in group "
1617 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1618 return 0;
1619 }
b5f10eed 1620 spin_lock(sb_bgl_lock(sbi, i));
717d50e4 1621 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
c19204b0
JB		 1622 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1623 "Checksum for group %lu failed (%u!=%u)\n",
1624 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1625 gdp)), le16_to_cpu(gdp->bg_checksum));
8a266467
TT		 1626 if (!(sb->s_flags & MS_RDONLY))
1627 return 0;
717d50e4 1628 }
b5f10eed 1629 spin_unlock(sb_bgl_lock(sbi, i));
ce421581
JS		 1630 if (!flexbg_flag)
1631 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1632 }
1633
bd81d8ee 1634 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2b2d6d01 1635 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1636 return 1;
1637}
1638
617ba13b 1639/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1640 * the superblock) which were deleted from all directories, but held open by
1641 * a process at the time of a crash. We walk the list and try to delete these
1642 * inodes at recovery time (only with a read-write filesystem).
1643 *
1644 * In order to keep the orphan inode chain consistent during traversal (in
1645 * case of crash during recovery), we link each inode into the superblock
1646 * orphan list_head and handle it the same way as an inode deletion during
1647 * normal operation (which journals the operations for us).
1648 *
1649 * We only do an iget() and an iput() on each inode, which is very safe if we
1650 * accidentally point at an in-use or already deleted inode. The worst that
1651 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1652 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1653 * e2fsck was run on this filesystem, and it must have already done the orphan
1654 * inode cleanup for us, so we can safely abort without any further action.
1655 */
2b2d6d01
TT		 1656static void ext4_orphan_cleanup(struct super_block *sb,
1657 struct ext4_super_block *es)
ac27a0ec
DK		 1658{
1659 unsigned int s_flags = sb->s_flags;
1660 int nr_orphans = 0, nr_truncates = 0;
1661#ifdef CONFIG_QUOTA
1662 int i;
1663#endif
1664 if (!es->s_last_orphan) {
1665 jbd_debug(4, "no orphan inodes to clean up\n");
1666 return;
1667 }
1668
a8f48a95
ES		 1669 if (bdev_read_only(sb->s_bdev)) {
1670 printk(KERN_ERR "EXT4-fs: write access "
1671 "unavailable, skipping orphan cleanup.\n");
1672 return;
1673 }
1674
617ba13b 1675 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1676 if (es->s_last_orphan)
1677 jbd_debug(1, "Errors on filesystem, "
1678 "clearing orphan list.\n");
1679 es->s_last_orphan = 0;
1680 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1681 return;
1682 }
1683
1684 if (s_flags & MS_RDONLY) {
617ba13b 1685 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1686 sb->s_id);
1687 sb->s_flags &= ~MS_RDONLY;
1688 }
1689#ifdef CONFIG_QUOTA
1690 /* Needed for iput() to work correctly and not trash data */
1691 sb->s_flags |= MS_ACTIVE;
1692 /* Turn on quotas so that they are updated correctly */
1693 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1694 if (EXT4_SB(sb)->s_qf_names[i]) {
1695 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1696 if (ret < 0)
1697 printk(KERN_ERR
2c8be6b2 1698 "EXT4-fs: Cannot turn on journaled "
ac27a0ec
DK		 1699 "quota: error %d\n", ret);
1700 }
1701 }
1702#endif
1703
1704 while (es->s_last_orphan) {
1705 struct inode *inode;
1706
97bd42b9
JB		 1707 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1708 if (IS_ERR(inode)) {
ac27a0ec
DK		 1709 es->s_last_orphan = 0;
1710 break;
1711 }
1712
617ba13b 1713 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1714 DQUOT_INIT(inode);
1715 if (inode->i_nlink) {
1716 printk(KERN_DEBUG
1717 "%s: truncating inode %lu to %Ld bytes\n",
46e665e9 1718 __func__, inode->i_ino, inode->i_size);
ac27a0ec
DK		 1719 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1720 inode->i_ino, inode->i_size);
617ba13b 1721 ext4_truncate(inode);
ac27a0ec
DK		 1722 nr_truncates++;
1723 } else {
1724 printk(KERN_DEBUG
1725 "%s: deleting unreferenced inode %lu\n",
46e665e9 1726 __func__, inode->i_ino);
ac27a0ec
DK		 1727 jbd_debug(2, "deleting unreferenced inode %lu\n",
1728 inode->i_ino);
1729 nr_orphans++;
1730 }
1731 iput(inode); /* The delete magic happens here! */
1732 }
1733
2b2d6d01 1734#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 1735
1736 if (nr_orphans)
617ba13b 1737 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1738 sb->s_id, PLURAL(nr_orphans));
1739 if (nr_truncates)
617ba13b 1740 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1741 sb->s_id, PLURAL(nr_truncates));
1742#ifdef CONFIG_QUOTA
1743 /* Turn quotas off */
1744 for (i = 0; i < MAXQUOTAS; i++) {
1745 if (sb_dqopt(sb)->files[i])
6f28e087 1746 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1747 }
1748#endif
1749 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1750}
cd2291a4
ES		 1751/*
1752 * Maximal extent format file size.
1753 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1754 * extent format containers, within a sector_t, and within i_blocks
1755 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1756 * so that won't be a limiting factor.
1757 *
1758 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1759 */
1760static loff_t ext4_max_size(int blkbits)
1761{
1762 loff_t res;
1763 loff_t upper_limit = MAX_LFS_FILESIZE;
1764
1765 /* small i_blocks in vfs inode? */
1766 if (sizeof(blkcnt_t) < sizeof(u64)) {
1767 /*
1768 * CONFIG_LSF is not enabled implies the inode
1769 * i_block represent total blocks in 512 bytes
1770 * 32 == size of vfs inode i_blocks * 8
1771 */
1772 upper_limit = (1LL << 32) - 1;
1773
1774 /* total blocks in file system block size */
1775 upper_limit >>= (blkbits - 9);
1776 upper_limit <<= blkbits;
1777 }
1778
1779 /* 32-bit extent-start container, ee_block */
1780 res = 1LL << 32;
1781 res <<= blkbits;
1782 res -= 1;
1783
1784 /* Sanity check against vm- & vfs- imposed limits */
1785 if (res > upper_limit)
1786 res = upper_limit;
1787
1788 return res;
1789}
ac27a0ec 1790
ac27a0ec 1791/*
cd2291a4 1792 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1793 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1794 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1795 */
cd2291a4 1796static loff_t ext4_max_bitmap_size(int bits)
ac27a0ec 1797{
617ba13b 1798 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1799 int meta_blocks;
1800 loff_t upper_limit;
1801 /* This is calculated to be the largest file size for a
cd2291a4 1802 * dense, bitmapped file such that the total number of
ac27a0ec 1803 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1804 * does not exceed 2^48 -1
1805 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1806 * total number of 512 bytes blocks of the file
1807 */
1808
1809 if (sizeof(blkcnt_t) < sizeof(u64)) {
1810 /*
1811 * CONFIG_LSF is not enabled implies the inode
1812 * i_block represent total blocks in 512 bytes
1813 * 32 == size of vfs inode i_blocks * 8
1814 */
1815 upper_limit = (1LL << 32) - 1;
1816
1817 /* total blocks in file system block size */
1818 upper_limit >>= (bits - 9);
1819
1820 } else {
8180a562
AK		 1821 /*
1822 * We use 48 bit ext4_inode i_blocks
1823 * With EXT4_HUGE_FILE_FL set the i_blocks
1824 * represent total number of blocks in
1825 * file system block size
1826 */
0fc1b451
AK		 1827 upper_limit = (1LL << 48) - 1;
1828
0fc1b451
AK		 1829 }
1830
1831 /* indirect blocks */
1832 meta_blocks = 1;
1833 /* double indirect blocks */
1834 meta_blocks += 1 + (1LL << (bits-2));
1835 /* tripple indirect blocks */
1836 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1837
1838 upper_limit -= meta_blocks;
1839 upper_limit <<= bits;
ac27a0ec
DK		 1840
1841 res += 1LL << (bits-2);
1842 res += 1LL << (2*(bits-2));
1843 res += 1LL << (3*(bits-2));
1844 res <<= bits;
1845 if (res > upper_limit)
1846 res = upper_limit;
0fc1b451
AK		 1847
1848 if (res > MAX_LFS_FILESIZE)
1849 res = MAX_LFS_FILESIZE;
1850
ac27a0ec
DK		 1851 return res;
1852}
1853
617ba13b 1854static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1855 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1856{
617ba13b 1857 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1858 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1859 int has_super = 0;
1860
1861 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1862
617ba13b 1863 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1864 nr < first_meta_bg)
70bbb3e0 1865 return logical_sb_block + nr + 1;
ac27a0ec 1866 bg = sbi->s_desc_per_block * nr;
617ba13b 1867 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1868 has_super = 1;
617ba13b 1869 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1870}
1871
c9de560d
AT		 1872/**
1873 * ext4_get_stripe_size: Get the stripe size.
1874 * @sbi: In memory super block info
1875 *
1876 * If we have specified it via mount option, then
1877 * use the mount option value. If the value specified at mount time is
1878 * greater than the blocks per group use the super block value.
1879 * If the super block value is greater than blocks per group return 0.
1880 * Allocator needs it be less than blocks per group.
1881 *
1882 */
1883static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1884{
1885 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1886 unsigned long stripe_width =
1887 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1888
1889 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1890 return sbi->s_stripe;
1891
1892 if (stripe_width <= sbi->s_blocks_per_group)
1893 return stripe_width;
1894
1895 if (stride <= sbi->s_blocks_per_group)
1896 return stride;
1897
1898 return 0;
1899}
ac27a0ec 1900
2b2d6d01 1901static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 1902 __releases(kernel_lock)
1903 __acquires(kernel_lock)
1d03ec98 1904
ac27a0ec 1905{
2b2d6d01 1906 struct buffer_head *bh;
617ba13b
MC		 1907 struct ext4_super_block *es = NULL;
1908 struct ext4_sb_info *sbi;
1909 ext4_fsblk_t block;
1910 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 1911 ext4_fsblk_t logical_sb_block;
ac27a0ec
DK		 1912 unsigned long offset = 0;
1913 unsigned int journal_inum = 0;
1914 unsigned long journal_devnum = 0;
1915 unsigned long def_mount_opts;
1916 struct inode *root;
1d1fe1ee 1917 int ret = -EINVAL;
ac27a0ec 1918 int blocksize;
ac27a0ec
DK		 1919 int db_count;
1920 int i;
1921 int needs_recovery;
1922 __le32 features;
bd81d8ee 1923 __u64 blocks_count;
833f4077 1924 int err;
ac27a0ec
DK		 1925
1926 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1927 if (!sbi)
1928 return -ENOMEM;
1929 sb->s_fs_info = sbi;
1930 sbi->s_mount_opt = 0;
617ba13b
MC		 1931 sbi->s_resuid = EXT4_DEF_RESUID;
1932 sbi->s_resgid = EXT4_DEF_RESGID;
d9c9bef1 1933 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 1934
1935 unlock_kernel();
1936
617ba13b 1937 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 1938 if (!blocksize) {
617ba13b 1939 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 1940 goto out_fail;
1941 }
1942
1943 /*
617ba13b 1944 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 1945 * block sizes. We need to calculate the offset from buffer start.
1946 */
617ba13b 1947 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 1948 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1949 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 1950 } else {
70bbb3e0 1951 logical_sb_block = sb_block;
ac27a0ec
DK		 1952 }
1953
70bbb3e0 1954 if (!(bh = sb_bread(sb, logical_sb_block))) {
2b2d6d01 1955 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 1956 goto out_fail;
1957 }
1958 /*
1959 * Note: s_es must be initialized as soon as possible because
617ba13b 1960 * some ext4 macro-instructions depend on its value
ac27a0ec 1961 */
617ba13b 1962 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 1963 sbi->s_es = es;
1964 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 1965 if (sb->s_magic != EXT4_SUPER_MAGIC)
1966 goto cantfind_ext4;
ac27a0ec
DK		 1967
1968 /* Set defaults before we parse the mount options */
1969 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 1970 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 1971 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 1972 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 1973 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 1974 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 1975 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1976#ifdef CONFIG_EXT4DEV_FS_XATTR
617ba13b 1977 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 1978 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1979#endif
1980#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
617ba13b 1981 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 1982 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1983#endif
617ba13b
MC		 1984 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1985 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1986 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1987 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1988 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1989 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1990
1991 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 1992 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 1993 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 1994 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 1995 else
1996 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1997
1998 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1999 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2000
2001 set_opt(sbi->s_mount_opt, RESERVATION);
571640ca 2002 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2003
1e2462f9
MC		 2004 /*
2005 * turn on extents feature by default in ext4 filesystem
e4079a11
ES		 2006 * only if feature flag already set by mkfs or tune2fs.
2007 * Use -o noextents to turn it off
1e2462f9 2008 */
e4079a11
ES		 2009 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
2010 set_opt(sbi->s_mount_opt, EXTENTS);
2011 else
2012 ext4_warning(sb, __func__,
2013 "extents feature not enabled on this filesystem, "
2014 "use tune2fs.\n");
3dbd0ede 2015 /*
e4079a11
ES		 2016 * turn on mballoc code by default in ext4 filesystem
2017 * Use -o nomballoc to turn it off
3dbd0ede
AK		 2018 */
2019 set_opt(sbi->s_mount_opt, MBALLOC);
1e2462f9 2020
dd919b98
AK		 2021 /*
2022 * enable delayed allocation by default
2023 * Use -o nodelalloc to turn it off
2024 */
2025 set_opt(sbi->s_mount_opt, DELALLOC);
2026
2027
2b2d6d01
TT		 2028 if (!parse_options((char *) data, sb, &journal_inum, &journal_devnum,
2029 NULL, 0))
ac27a0ec
DK		 2030 goto failed_mount;
2031
2032 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2033 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2034
617ba13b
MC		 2035 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2036 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2037 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2038 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 2039 printk(KERN_WARNING
617ba13b 2040 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec 2041 "running e2fsck is recommended\n");
469108ff
TT		 2042
2043 /*
2044 * Since ext4 is still considered development code, we require
2045 * that the TEST_FILESYS flag in s->flags be set.
2046 */
2047 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
2048 printk(KERN_WARNING "EXT4-fs: %s: not marked "
2049 "OK to use with test code.\n", sb->s_id);
2050 goto failed_mount;
2051 }
2052
ac27a0ec
DK		 2053 /*
2054 * Check feature flags regardless of the revision level, since we
2055 * previously didn't change the revision level when setting the flags,
2056 * so there is a chance incompat flags are set on a rev 0 filesystem.
2057 */
617ba13b 2058 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 2059 if (features) {
617ba13b 2060 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
ac27a0ec
DK		 2061 "unsupported optional features (%x).\n",
2062 sb->s_id, le32_to_cpu(features));
2063 goto failed_mount;
2064 }
617ba13b 2065 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 2066 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 2067 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
ac27a0ec
DK		 2068 "unsupported optional features (%x).\n",
2069 sb->s_id, le32_to_cpu(features));
2070 goto failed_mount;
2071 }
0fc1b451
AK		 2072 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2073 /*
2074 * Large file size enabled file system can only be
2075 * mount if kernel is build with CONFIG_LSF
2076 */
2077 if (sizeof(root->i_blocks) < sizeof(u64) &&
2078 !(sb->s_flags & MS_RDONLY)) {
2079 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2080 "files cannot be mounted read-write "
2081 "without CONFIG_LSF.\n", sb->s_id);
2082 goto failed_mount;
2083 }
2084 }
ac27a0ec
DK		 2085 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2086
617ba13b
MC		 2087 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2088 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 2089 printk(KERN_ERR
617ba13b 2090 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 2091 blocksize, sb->s_id);
2092 goto failed_mount;
2093 }
2094
ac27a0ec 2095 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2096
2097 /* Validate the filesystem blocksize */
2098 if (!sb_set_blocksize(sb, blocksize)) {
2099 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2100 blocksize);
ac27a0ec
DK		 2101 goto failed_mount;
2102 }
2103
2b2d6d01 2104 brelse(bh);
70bbb3e0
AM		 2105 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2106 offset = do_div(logical_sb_block, blocksize);
2107 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 2108 if (!bh) {
2109 printk(KERN_ERR
617ba13b 2110 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 2111 goto failed_mount;
2112 }
617ba13b 2113 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2114 sbi->s_es = es;
617ba13b 2115 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2b2d6d01
TT		 2116 printk(KERN_ERR
2117 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 2118 goto failed_mount;
2119 }
2120 }
2121
e2b46574 2122 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
617ba13b 2123 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
ac27a0ec 2124
617ba13b
MC		 2125 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2126 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2127 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2128 } else {
2129 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2130 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2131 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2132 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2133 (sbi->s_inode_size > blocksize)) {
2b2d6d01
TT		 2134 printk(KERN_ERR
2135 "EXT4-fs: unsupported inode size: %d\n",
2136 sbi->s_inode_size);
ac27a0ec
DK		 2137 goto failed_mount;
2138 }
ef7f3835
KS		 2139 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2140 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2141 }
0d1ee42f
AR		 2142 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2143 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2144 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2145 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2146 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2147 printk(KERN_ERR
8fadc143 2148 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2149 sbi->s_desc_size);
2150 goto failed_mount;
2151 }
2152 } else
2153 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2154 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2155 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2156 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2157 goto cantfind_ext4;
2158 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2159 if (sbi->s_inodes_per_block == 0)
617ba13b 2160 goto cantfind_ext4;
ac27a0ec
DK		 2161 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2162 sbi->s_inodes_per_block;
0d1ee42f 2163 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2164 sbi->s_sbh = bh;
2165 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2166 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2167 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2b2d6d01 2168 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2169 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2170 sbi->s_def_hash_version = es->s_def_hash_version;
2171
2172 if (sbi->s_blocks_per_group > blocksize * 8) {
2b2d6d01
TT		 2173 printk(KERN_ERR
2174 "EXT4-fs: #blocks per group too big: %lu\n",
2175 sbi->s_blocks_per_group);
ac27a0ec
DK		 2176 goto failed_mount;
2177 }
ac27a0ec 2178 if (sbi->s_inodes_per_group > blocksize * 8) {
2b2d6d01
TT		 2179 printk(KERN_ERR
2180 "EXT4-fs: #inodes per group too big: %lu\n",
2181 sbi->s_inodes_per_group);
ac27a0ec
DK		 2182 goto failed_mount;
2183 }
2184
bd81d8ee 2185 if (ext4_blocks_count(es) >
ac27a0ec 2186 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2187 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2188 " too large to mount safely\n", sb->s_id);
2189 if (sizeof(sector_t) < 8)
617ba13b 2190 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2191 "enabled\n");
2192 goto failed_mount;
2193 }
2194
617ba13b
MC		 2195 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2196 goto cantfind_ext4;
e7c95593
ES		 2197
2198 /* ensure blocks_count calculation below doesn't sign-extend */
2199 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2200 le32_to_cpu(es->s_first_data_block) + 1) {
2201 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2202 "first data block %u, blocks per group %lu\n",
2203 ext4_blocks_count(es),
2204 le32_to_cpu(es->s_first_data_block),
2205 EXT4_BLOCKS_PER_GROUP(sb));
2206 goto failed_mount;
2207 }
bd81d8ee
LV		 2208 blocks_count = (ext4_blocks_count(es) -
2209 le32_to_cpu(es->s_first_data_block) +
2210 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2211 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2212 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2213 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2214 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2215 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2216 GFP_KERNEL);
2217 if (sbi->s_group_desc == NULL) {
2b2d6d01 2218 printk(KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2219 goto failed_mount;
2220 }
2221
2222 bgl_lock_init(&sbi->s_blockgroup_lock);
2223
2224 for (i = 0; i < db_count; i++) {
70bbb3e0 2225 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2226 sbi->s_group_desc[i] = sb_bread(sb, block);
2227 if (!sbi->s_group_desc[i]) {
2b2d6d01
TT		 2228 printk(KERN_ERR "EXT4-fs: "
2229 "can't read group descriptor %d\n", i);
ac27a0ec
DK		 2230 db_count = i;
2231 goto failed_mount2;
2232 }
2233 }
2b2d6d01 2234 if (!ext4_check_descriptors(sb)) {
617ba13b 2235 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2236 goto failed_mount2;
2237 }
772cb7c8
JS		 2238 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2239 if (!ext4_fill_flex_info(sb)) {
2240 printk(KERN_ERR
2241 "EXT4-fs: unable to initialize "
2242 "flex_bg meta info!\n");
2243 goto failed_mount2;
2244 }
2245
ac27a0ec
DK		 2246 sbi->s_gdb_count = db_count;
2247 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2248 spin_lock_init(&sbi->s_next_gen_lock);
2249
833f4077
PZ		 2250 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2251 ext4_count_free_blocks(sb));
2252 if (!err) {
2253 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2254 ext4_count_free_inodes(sb));
2255 }
2256 if (!err) {
2257 err = percpu_counter_init(&sbi->s_dirs_counter,
2258 ext4_count_dirs(sb));
2259 }
2260 if (err) {
2261 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2262 goto failed_mount3;
2263 }
ac27a0ec
DK		 2264
2265 /* per fileystem reservation list head & lock */
2266 spin_lock_init(&sbi->s_rsv_window_lock);
2267 sbi->s_rsv_window_root = RB_ROOT;
2268 /* Add a single, static dummy reservation to the start of the
2269 * reservation window list --- it gives us a placeholder for
2270 * append-at-start-of-list which makes the allocation logic
2271 * _much_ simpler. */
617ba13b
MC		 2272 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2273 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
ac27a0ec
DK		 2274 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2275 sbi->s_rsv_window_head.rsv_goal_size = 0;
617ba13b 2276 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
ac27a0ec 2277
c9de560d
AT		 2278 sbi->s_stripe = ext4_get_stripe_size(sbi);
2279
ac27a0ec
DK		 2280 /*
2281 * set up enough so that it can read an inode
2282 */
617ba13b
MC		 2283 sb->s_op = &ext4_sops;
2284 sb->s_export_op = &ext4_export_ops;
2285 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2286#ifdef CONFIG_QUOTA
617ba13b
MC		 2287 sb->s_qcop = &ext4_qctl_operations;
2288 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2289#endif
2290 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2291
2292 sb->s_root = NULL;
2293
2294 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2295 EXT4_HAS_INCOMPAT_FEATURE(sb,
2296 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2297
2298 /*
2299 * The first inode we look at is the journal inode. Don't try
2300 * root first: it may be modified in the journal!
2301 */
2302 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2303 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2304 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2305 goto failed_mount3;
624080ed
TT		 2306 if (!(sb->s_flags & MS_RDONLY) &&
2307 EXT4_SB(sb)->s_journal->j_failed_commit) {
2308 printk(KERN_CRIT "EXT4-fs error (device %s): "
2309 "ext4_fill_super: Journal transaction "
2b2d6d01 2310 "%u is corrupt\n", sb->s_id,
624080ed 2311 EXT4_SB(sb)->s_journal->j_failed_commit);
2b2d6d01
TT		 2312 if (test_opt(sb, ERRORS_RO)) {
2313 printk(KERN_CRIT
2314 "Mounting filesystem read-only\n");
624080ed
TT		 2315 sb->s_flags |= MS_RDONLY;
2316 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2317 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2318 }
2319 if (test_opt(sb, ERRORS_PANIC)) {
2320 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2321 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2322 ext4_commit_super(sb, es, 1);
2323 printk(KERN_CRIT
2324 "EXT4-fs (device %s): mount failed\n",
2325 sb->s_id);
2326 goto failed_mount4;
2327 }
2328 }
ac27a0ec 2329 } else if (journal_inum) {
617ba13b 2330 if (ext4_create_journal(sb, es, journal_inum))
ac27a0ec
DK		 2331 goto failed_mount3;
2332 } else {
2333 if (!silent)
2b2d6d01
TT		 2334 printk(KERN_ERR
2335 "ext4: No journal on filesystem on %s\n",
2336 sb->s_id);
ac27a0ec
DK		 2337 goto failed_mount3;
2338 }
2339
eb40a09c
JS		 2340 if (ext4_blocks_count(es) > 0xffffffffULL &&
2341 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2342 JBD2_FEATURE_INCOMPAT_64BIT)) {
2343 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2344 goto failed_mount4;
2345 }
2346
818d276c
GS		 2347 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2348 jbd2_journal_set_features(sbi->s_journal,
2349 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2350 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2351 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2352 jbd2_journal_set_features(sbi->s_journal,
2353 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2354 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2355 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2356 } else {
2357 jbd2_journal_clear_features(sbi->s_journal,
2358 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2359 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2360 }
2361
ac27a0ec
DK		 2362 /* We have now updated the journal if required, so we can
2363 * validate the data journaling mode. */
2364 switch (test_opt(sb, DATA_FLAGS)) {
2365 case 0:
2366 /* No mode set, assume a default based on the journal
63f57933
AM		 2367 * capabilities: ORDERED_DATA if the journal can
2368 * cope, else JOURNAL_DATA
2369 */
dab291af
MC		 2370 if (jbd2_journal_check_available_features
2371 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2372 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2373 else
2374 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2375 break;
2376
617ba13b
MC		 2377 case EXT4_MOUNT_ORDERED_DATA:
2378 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2379 if (!jbd2_journal_check_available_features
2380 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2381 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2382 "requested data journaling mode\n");
2383 goto failed_mount4;
2384 }
2385 default:
2386 break;
2387 }
2388
2389 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2390 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2391 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2392 "its supported only with writeback mode\n");
2393 clear_opt(sbi->s_mount_opt, NOBH);
2394 }
2395 }
2396 /*
dab291af 2397 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2398 * so we can safely mount the rest of the filesystem now.
2399 */
2400
1d1fe1ee
DH		 2401 root = ext4_iget(sb, EXT4_ROOT_INO);
2402 if (IS_ERR(root)) {
617ba13b 2403 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1d1fe1ee 2404 ret = PTR_ERR(root);
ac27a0ec
DK		 2405 goto failed_mount4;
2406 }
2407 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2408 iput(root);
617ba13b 2409 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2410 goto failed_mount4;
2411 }
1d1fe1ee
DH		 2412 sb->s_root = d_alloc_root(root);
2413 if (!sb->s_root) {
2414 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2415 iput(root);
2416 ret = -ENOMEM;
2417 goto failed_mount4;
2418 }
ac27a0ec 2419
2b2d6d01 2420 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2421
2422 /* determine the minimum size of new large inodes, if present */
2423 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2424 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2425 EXT4_GOOD_OLD_INODE_SIZE;
2426 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2427 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2428 if (sbi->s_want_extra_isize <
2429 le16_to_cpu(es->s_want_extra_isize))
2430 sbi->s_want_extra_isize =
2431 le16_to_cpu(es->s_want_extra_isize);
2432 if (sbi->s_want_extra_isize <
2433 le16_to_cpu(es->s_min_extra_isize))
2434 sbi->s_want_extra_isize =
2435 le16_to_cpu(es->s_min_extra_isize);
2436 }
2437 }
2438 /* Check if enough inode space is available */
2439 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2440 sbi->s_inode_size) {
2441 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2442 EXT4_GOOD_OLD_INODE_SIZE;
2443 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2444 "available.\n");
2445 }
2446
ac27a0ec
DK		 2447 /*
2448 * akpm: core read_super() calls in here with the superblock locked.
2449 * That deadlocks, because orphan cleanup needs to lock the superblock
2450 * in numerous places. Here we just pop the lock - it's relatively
2451 * harmless, because we are now ready to accept write_super() requests,
2452 * and aviro says that's the only reason for hanging onto the
2453 * superblock lock.
2454 */
617ba13b
MC		 2455 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2456 ext4_orphan_cleanup(sb, es);
2457 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
ac27a0ec 2458 if (needs_recovery)
2b2d6d01 2459 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
617ba13b 2460 ext4_mark_recovery_complete(sb, es);
2b2d6d01
TT		 2461 printk(KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2462 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2463 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2464 "writeback");
ac27a0ec 2465
dd919b98
AK		 2466 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2467 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2468 "requested data journaling mode\n");
2469 clear_opt(sbi->s_mount_opt, DELALLOC);
2470 } else if (test_opt(sb, DELALLOC))
2471 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2472
a86c6181 2473 ext4_ext_init(sb);
c9de560d 2474 ext4_mb_init(sb, needs_recovery);
a86c6181 2475
ac27a0ec
DK		 2476 lock_kernel();
2477 return 0;
2478
617ba13b 2479cantfind_ext4:
ac27a0ec 2480 if (!silent)
617ba13b 2481 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2482 sb->s_id);
2483 goto failed_mount;
2484
2485failed_mount4:
dab291af 2486 jbd2_journal_destroy(sbi->s_journal);
47b4a50b 2487 sbi->s_journal = NULL;
ac27a0ec
DK		 2488failed_mount3:
2489 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2490 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2491 percpu_counter_destroy(&sbi->s_dirs_counter);
2492failed_mount2:
2493 for (i = 0; i < db_count; i++)
2494 brelse(sbi->s_group_desc[i]);
2495 kfree(sbi->s_group_desc);
2496failed_mount:
2497#ifdef CONFIG_QUOTA
2498 for (i = 0; i < MAXQUOTAS; i++)
2499 kfree(sbi->s_qf_names[i]);
2500#endif
617ba13b 2501 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2502 brelse(bh);
2503out_fail:
2504 sb->s_fs_info = NULL;
2505 kfree(sbi);
2506 lock_kernel();
1d1fe1ee 2507 return ret;
ac27a0ec
DK		 2508}
2509
2510/*
2511 * Setup any per-fs journal parameters now. We'll do this both on
2512 * initial mount, once the journal has been initialised but before we've
2513 * done any recovery; and again on any subsequent remount.
2514 */
617ba13b 2515static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2516{
617ba13b 2517 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2518
2519 if (sbi->s_commit_interval)
2520 journal->j_commit_interval = sbi->s_commit_interval;
617ba13b 2521 /* We could also set up an ext4-specific default for the commit
ac27a0ec
DK		 2522 * interval here, but for now we'll just fall back to the jbd
2523 * default. */
2524
2525 spin_lock(&journal->j_state_lock);
2526 if (test_opt(sb, BARRIER))
dab291af 2527 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2528 else
dab291af 2529 journal->j_flags &= ~JBD2_BARRIER;
ac27a0ec
DK		 2530 spin_unlock(&journal->j_state_lock);
2531}
2532
617ba13b 2533static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2534 unsigned int journal_inum)
2535{
2536 struct inode *journal_inode;
2537 journal_t *journal;
2538
2539 /* First, test for the existence of a valid inode on disk. Bad
2540 * things happen if we iget() an unused inode, as the subsequent
2541 * iput() will try to delete it. */
2542
1d1fe1ee
DH		 2543 journal_inode = ext4_iget(sb, journal_inum);
2544 if (IS_ERR(journal_inode)) {
617ba13b 2545 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2546 return NULL;
2547 }
2548 if (!journal_inode->i_nlink) {
2549 make_bad_inode(journal_inode);
2550 iput(journal_inode);
617ba13b 2551 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2552 return NULL;
2553 }
2554
2555 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2556 journal_inode, journal_inode->i_size);
1d1fe1ee 2557 if (!S_ISREG(journal_inode->i_mode)) {
617ba13b 2558 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2559 iput(journal_inode);
2560 return NULL;
2561 }
2562
dab291af 2563 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2564 if (!journal) {
617ba13b 2565 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2566 iput(journal_inode);
2567 return NULL;
2568 }
2569 journal->j_private = sb;
617ba13b 2570 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2571 return journal;
2572}
2573
617ba13b 2574static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2575 dev_t j_dev)
2576{
2b2d6d01 2577 struct buffer_head *bh;
ac27a0ec 2578 journal_t *journal;
617ba13b
MC		 2579 ext4_fsblk_t start;
2580 ext4_fsblk_t len;
ac27a0ec 2581 int hblock, blocksize;
617ba13b 2582 ext4_fsblk_t sb_block;
ac27a0ec 2583 unsigned long offset;
2b2d6d01 2584 struct ext4_super_block *es;
ac27a0ec
DK		 2585 struct block_device *bdev;
2586
617ba13b 2587 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2588 if (bdev == NULL)
2589 return NULL;
2590
2591 if (bd_claim(bdev, sb)) {
2592 printk(KERN_ERR
8c55e204 2593 "EXT4: failed to claim external journal device.\n");
ac27a0ec
DK		 2594 blkdev_put(bdev);
2595 return NULL;
2596 }
2597
2598 blocksize = sb->s_blocksize;
2599 hblock = bdev_hardsect_size(bdev);
2600 if (blocksize < hblock) {
2601 printk(KERN_ERR
617ba13b 2602 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2603 goto out_bdev;
2604 }
2605
617ba13b
MC		 2606 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2607 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2608 set_blocksize(bdev, blocksize);
2609 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2610 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2611 "external journal\n");
2612 goto out_bdev;
2613 }
2614
617ba13b
MC		 2615 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2616 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2617 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2618 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2619 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2620 "bad superblock\n");
2621 brelse(bh);
2622 goto out_bdev;
2623 }
2624
617ba13b
MC		 2625 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2626 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2627 brelse(bh);
2628 goto out_bdev;
2629 }
2630
bd81d8ee 2631 len = ext4_blocks_count(es);
ac27a0ec
DK		 2632 start = sb_block + 1;
2633 brelse(bh); /* we're done with the superblock */
2634
dab291af 2635 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2636 start, len, blocksize);
2637 if (!journal) {
617ba13b 2638 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2639 goto out_bdev;
2640 }
2641 journal->j_private = sb;
2642 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2643 wait_on_buffer(journal->j_sb_buffer);
2644 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2645 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2646 goto out_journal;
2647 }
2648 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2649 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2650 "user (unsupported) - %d\n",
2651 be32_to_cpu(journal->j_superblock->s_nr_users));
2652 goto out_journal;
2653 }
617ba13b
MC		 2654 EXT4_SB(sb)->journal_bdev = bdev;
2655 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2656 return journal;
2657out_journal:
dab291af 2658 jbd2_journal_destroy(journal);
ac27a0ec 2659out_bdev:
617ba13b 2660 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2661 return NULL;
2662}
2663
617ba13b
MC		 2664static int ext4_load_journal(struct super_block *sb,
2665 struct ext4_super_block *es,
ac27a0ec
DK		 2666 unsigned long journal_devnum)
2667{
2668 journal_t *journal;
2669 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2670 dev_t journal_dev;
2671 int err = 0;
2672 int really_read_only;
2673
2674 if (journal_devnum &&
2675 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2676 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2677 "numbers have changed\n");
2678 journal_dev = new_decode_dev(journal_devnum);
2679 } else
2680 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2681
2682 really_read_only = bdev_read_only(sb->s_bdev);
2683
2684 /*
2685 * Are we loading a blank journal or performing recovery after a
2686 * crash? For recovery, we need to check in advance whether we
2687 * can get read-write access to the device.
2688 */
2689
617ba13b 2690 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2691 if (sb->s_flags & MS_RDONLY) {
617ba13b 2692 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2693 "required on readonly filesystem.\n");
2694 if (really_read_only) {
617ba13b 2695 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2696 "unavailable, cannot proceed.\n");
2697 return -EROFS;
2698 }
2b2d6d01
TT		 2699 printk(KERN_INFO "EXT4-fs: write access will "
2700 "be enabled during recovery.\n");
ac27a0ec
DK		 2701 }
2702 }
2703
2704 if (journal_inum && journal_dev) {
617ba13b 2705 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2706 "and inode journals!\n");
2707 return -EINVAL;
2708 }
2709
2710 if (journal_inum) {
617ba13b 2711 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2712 return -EINVAL;
2713 } else {
617ba13b 2714 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2715 return -EINVAL;
2716 }
2717
2718 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2719 err = jbd2_journal_update_format(journal);
ac27a0ec 2720 if (err) {
617ba13b 2721 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2722 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2723 return err;
2724 }
2725 }
2726
617ba13b 2727 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2728 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2729 if (!err)
dab291af 2730 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2731
2732 if (err) {
617ba13b 2733 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2734 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2735 return err;
2736 }
2737
617ba13b
MC		 2738 EXT4_SB(sb)->s_journal = journal;
2739 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2740
2741 if (journal_devnum &&
2742 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2743 es->s_journal_dev = cpu_to_le32(journal_devnum);
2744 sb->s_dirt = 1;
2745
2746 /* Make sure we flush the recovery flag to disk. */
617ba13b 2747 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2748 }
2749
2750 return 0;
2751}
2752
2b2d6d01
TT		 2753static int ext4_create_journal(struct super_block *sb,
2754 struct ext4_super_block *es,
ac27a0ec
DK		 2755 unsigned int journal_inum)
2756{
2757 journal_t *journal;
6c675bd4 2758 int err;
ac27a0ec
DK		 2759
2760 if (sb->s_flags & MS_RDONLY) {
617ba13b 2761 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
ac27a0ec
DK		 2762 "create journal.\n");
2763 return -EROFS;
2764 }
2765
6c675bd4
BP		 2766 journal = ext4_get_journal(sb, journal_inum);
2767 if (!journal)
ac27a0ec
DK		 2768 return -EINVAL;
2769
617ba13b 2770 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
ac27a0ec
DK		 2771 journal_inum);
2772
6c675bd4
BP		 2773 err = jbd2_journal_create(journal);
2774 if (err) {
617ba13b 2775 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
dab291af 2776 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2777 return -EIO;
2778 }
2779
617ba13b 2780 EXT4_SB(sb)->s_journal = journal;
ac27a0ec 2781
617ba13b
MC		 2782 ext4_update_dynamic_rev(sb);
2783 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2784 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
ac27a0ec
DK		 2785
2786 es->s_journal_inum = cpu_to_le32(journal_inum);
2787 sb->s_dirt = 1;
2788
2789 /* Make sure we flush the recovery flag to disk. */
617ba13b 2790 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2791
2792 return 0;
2793}
2794
2b2d6d01
TT		 2795static void ext4_commit_super(struct super_block *sb,
2796 struct ext4_super_block *es, int sync)
ac27a0ec 2797{
617ba13b 2798 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
ac27a0ec
DK		 2799
2800 if (!sbh)
2801 return;
2802 es->s_wtime = cpu_to_le32(get_seconds());
bd81d8ee 2803 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
617ba13b 2804 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2805 BUFFER_TRACE(sbh, "marking dirty");
2806 mark_buffer_dirty(sbh);
2807 if (sync)
2808 sync_dirty_buffer(sbh);
2809}
2810
2811
2812/*
2813 * Have we just finished recovery? If so, and if we are mounting (or
2814 * remounting) the filesystem readonly, then we will end up with a
2815 * consistent fs on disk. Record that fact.
2816 */
2b2d6d01
TT		 2817static void ext4_mark_recovery_complete(struct super_block *sb,
2818 struct ext4_super_block *es)
ac27a0ec 2819{
617ba13b 2820 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2821
dab291af
MC		 2822 jbd2_journal_lock_updates(journal);
2823 jbd2_journal_flush(journal);
32c37730 2824 lock_super(sb);
617ba13b 2825 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2826 sb->s_flags & MS_RDONLY) {
617ba13b 2827 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2828 sb->s_dirt = 0;
617ba13b 2829 ext4_commit_super(sb, es, 1);
ac27a0ec 2830 }
32c37730 2831 unlock_super(sb);
dab291af 2832 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2833}
2834
2835/*
2836 * If we are mounting (or read-write remounting) a filesystem whose journal
2837 * has recorded an error from a previous lifetime, move that error to the
2838 * main filesystem now.
2839 */
2b2d6d01
TT		 2840static void ext4_clear_journal_err(struct super_block *sb,
2841 struct ext4_super_block *es)
ac27a0ec
DK		 2842{
2843 journal_t *journal;
2844 int j_errno;
2845 const char *errstr;
2846
617ba13b 2847 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2848
2849 /*
2850 * Now check for any error status which may have been recorded in the
617ba13b 2851 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2852 */
2853
dab291af 2854 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2855 if (j_errno) {
2856 char nbuf[16];
2857
617ba13b 2858 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 2859 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 2860 "from previous mount: %s", errstr);
46e665e9 2861 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 2862 "filesystem check.");
2863
617ba13b
MC		 2864 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2865 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2b2d6d01 2866 ext4_commit_super(sb, es, 1);
ac27a0ec 2867
dab291af 2868 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 2869 }
2870}
2871
2872/*
2873 * Force the running and committing transactions to commit,
2874 * and wait on the commit.
2875 */
617ba13b 2876int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 2877{
2878 journal_t *journal;
2879 int ret;
2880
2881 if (sb->s_flags & MS_RDONLY)
2882 return 0;
2883
617ba13b 2884 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2885 sb->s_dirt = 0;
617ba13b 2886 ret = ext4_journal_force_commit(journal);
ac27a0ec
DK		 2887 return ret;
2888}
2889
2890/*
617ba13b 2891 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec
DK		 2892 * have to propagate any other updates to the superblock on disk at this
2893 * point. Just start an async writeback to get the buffers on their way
2894 * to the disk.
2895 *
2896 * This implicitly triggers the writebehind on sync().
2897 */
2898
2b2d6d01 2899static void ext4_write_super(struct super_block *sb)
ac27a0ec
DK		 2900{
2901 if (mutex_trylock(&sb->s_lock) != 0)
2902 BUG();
2903 sb->s_dirt = 0;
2904}
2905
617ba13b 2906static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec
DK		 2907{
2908 tid_t target;
2909
2910 sb->s_dirt = 0;
dab291af 2911 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
ac27a0ec 2912 if (wait)
dab291af 2913 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
ac27a0ec
DK		 2914 }
2915 return 0;
2916}
2917
2918/*
2919 * LVM calls this function before a (read-only) snapshot is created. This
2920 * gives us a chance to flush the journal completely and mark the fs clean.
2921 */
617ba13b 2922static void ext4_write_super_lockfs(struct super_block *sb)
ac27a0ec
DK		 2923{
2924 sb->s_dirt = 0;
2925
2926 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 2927 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2928
2929 /* Now we set up the journal barrier. */
dab291af
MC		 2930 jbd2_journal_lock_updates(journal);
2931 jbd2_journal_flush(journal);
ac27a0ec
DK		 2932
2933 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 2934 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2935 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec
DK		 2936 }
2937}
2938
2939/*
2940 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2941 * flag here, even though the filesystem is not technically dirty yet.
2942 */
617ba13b 2943static void ext4_unlockfs(struct super_block *sb)
ac27a0ec
DK		 2944{
2945 if (!(sb->s_flags & MS_RDONLY)) {
2946 lock_super(sb);
2947 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 2948 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2949 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 2950 unlock_super(sb);
dab291af 2951 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 2952 }
2953}
2954
2b2d6d01 2955static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 2956{
2b2d6d01 2957 struct ext4_super_block *es;
617ba13b
MC		 2958 struct ext4_sb_info *sbi = EXT4_SB(sb);
2959 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 2960 unsigned long old_sb_flags;
617ba13b 2961 struct ext4_mount_options old_opts;
8a266467 2962 ext4_group_t g;
ac27a0ec
DK		 2963 int err;
2964#ifdef CONFIG_QUOTA
2965 int i;
2966#endif
2967
2968 /* Store the original options */
2969 old_sb_flags = sb->s_flags;
2970 old_opts.s_mount_opt = sbi->s_mount_opt;
2971 old_opts.s_resuid = sbi->s_resuid;
2972 old_opts.s_resgid = sbi->s_resgid;
2973 old_opts.s_commit_interval = sbi->s_commit_interval;
2974#ifdef CONFIG_QUOTA
2975 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2976 for (i = 0; i < MAXQUOTAS; i++)
2977 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2978#endif
2979
2980 /*
2981 * Allow the "check" option to be passed as a remount option.
2982 */
2983 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2984 err = -EINVAL;
2985 goto restore_opts;
2986 }
2987
617ba13b 2988 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
46e665e9 2989 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 2990
2991 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2992 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 2993
2994 es = sbi->s_es;
2995
617ba13b 2996 ext4_init_journal_params(sb, sbi->s_journal);
ac27a0ec
DK		 2997
2998 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 2999 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 3000 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 3001 err = -EROFS;
3002 goto restore_opts;
3003 }
3004
3005 if (*flags & MS_RDONLY) {
3006 /*
3007 * First of all, the unconditional stuff we have to do
3008 * to disable replay of the journal when we next remount
3009 */
3010 sb->s_flags |= MS_RDONLY;
3011
3012 /*
3013 * OK, test if we are remounting a valid rw partition
3014 * readonly, and if so set the rdonly flag and then
3015 * mark the partition as valid again.
3016 */
617ba13b
MC		 3017 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3018 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3019 es->s_state = cpu_to_le16(sbi->s_mount_state);
3020
32c37730
JK		 3021 /*
3022 * We have to unlock super so that we can wait for
3023 * transactions.
3024 */
3025 unlock_super(sb);
617ba13b 3026 ext4_mark_recovery_complete(sb, es);
32c37730 3027 lock_super(sb);
ac27a0ec
DK		 3028 } else {
3029 __le32 ret;
617ba13b
MC		 3030 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3031 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3032 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec
DK		 3033 "remount RDWR because of unsupported "
3034 "optional features (%x).\n",
3035 sb->s_id, le32_to_cpu(ret));
3036 err = -EROFS;
3037 goto restore_opts;
3038 }
ead6596b 3039
8a266467
TT		 3040 /*
3041 * Make sure the group descriptor checksums
3042 * are sane. If they aren't, refuse to
3043 * remount r/w.
3044 */
3045 for (g = 0; g < sbi->s_groups_count; g++) {
3046 struct ext4_group_desc *gdp =
3047 ext4_get_group_desc(sb, g, NULL);
3048
3049 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3050 printk(KERN_ERR
3051 "EXT4-fs: ext4_remount: "
3052 "Checksum for group %lu failed (%u!=%u)\n",
3053 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3054 le16_to_cpu(gdp->bg_checksum));
3055 err = -EINVAL;
3056 goto restore_opts;
3057 }
3058 }
3059
ead6596b
ES		 3060 /*
3061 * If we have an unprocessed orphan list hanging
3062 * around from a previously readonly bdev mount,
3063 * require a full umount/remount for now.
3064 */
3065 if (es->s_last_orphan) {
3066 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3067 "remount RDWR because of unprocessed "
3068 "orphan inode list. Please "
3069 "umount/remount instead.\n",
3070 sb->s_id);
3071 err = -EINVAL;
3072 goto restore_opts;
3073 }
3074
ac27a0ec
DK		 3075 /*
3076 * Mounting a RDONLY partition read-write, so reread
3077 * and store the current valid flag. (It may have
3078 * been changed by e2fsck since we originally mounted
3079 * the partition.)
3080 */
617ba13b 3081 ext4_clear_journal_err(sb, es);
ac27a0ec 3082 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3083 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3084 goto restore_opts;
2b2d6d01 3085 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3086 sb->s_flags &= ~MS_RDONLY;
3087 }
3088 }
3089#ifdef CONFIG_QUOTA
3090 /* Release old quota file names */
3091 for (i = 0; i < MAXQUOTAS; i++)
3092 if (old_opts.s_qf_names[i] &&
3093 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3094 kfree(old_opts.s_qf_names[i]);
3095#endif
3096 return 0;
3097restore_opts:
3098 sb->s_flags = old_sb_flags;
3099 sbi->s_mount_opt = old_opts.s_mount_opt;
3100 sbi->s_resuid = old_opts.s_resuid;
3101 sbi->s_resgid = old_opts.s_resgid;
3102 sbi->s_commit_interval = old_opts.s_commit_interval;
3103#ifdef CONFIG_QUOTA
3104 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3105 for (i = 0; i < MAXQUOTAS; i++) {
3106 if (sbi->s_qf_names[i] &&
3107 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3108 kfree(sbi->s_qf_names[i]);
3109 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3110 }
3111#endif
3112 return err;
3113}
3114
2b2d6d01 3115static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3116{
3117 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3118 struct ext4_sb_info *sbi = EXT4_SB(sb);
3119 struct ext4_super_block *es = sbi->s_es;
960cc398 3120 u64 fsid;
ac27a0ec 3121
5e70030d
BP		 3122 if (test_opt(sb, MINIX_DF)) {
3123 sbi->s_overhead_last = 0;
6bc9feff 3124 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 3125 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 3126 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3127 smp_rmb();
3128
3129 /*
5e70030d
BP		 3130 * Compute the overhead (FS structures). This is constant
3131 * for a given filesystem unless the number of block groups
3132 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3133 */
3134
3135 /*
3136 * All of the blocks before first_data_block are
3137 * overhead
3138 */
3139 overhead = le32_to_cpu(es->s_first_data_block);
3140
3141 /*
3142 * Add the overhead attributed to the superblock and
3143 * block group descriptors. If the sparse superblocks
3144 * feature is turned on, then not all groups have this.
3145 */
3146 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3147 overhead += ext4_bg_has_super(sb, i) +
3148 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3149 cond_resched();
3150 }
3151
3152 /*
3153 * Every block group has an inode bitmap, a block
3154 * bitmap, and an inode table.
3155 */
5e70030d
BP		 3156 overhead += ngroups * (2 + sbi->s_itb_per_group);
3157 sbi->s_overhead_last = overhead;
3158 smp_wmb();
6bc9feff 3159 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3160 }
3161
617ba13b 3162 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3163 buf->f_bsize = sb->s_blocksize;
5e70030d 3164 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
52d9f3b4 3165 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
308ba3ec 3166 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3167 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3168 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3169 buf->f_bavail = 0;
3170 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3171 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3172 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3173 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3174 fsid = le64_to_cpup((void *)es->s_uuid) ^
3175 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3176 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3177 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 3178 return 0;
3179}
3180
3181/* Helper function for writing quotas on sync - we need to start transaction before quota file
3182 * is locked for write. Otherwise the are possible deadlocks:
3183 * Process 1 Process 2
617ba13b 3184 * ext4_create() quota_sync()
dab291af 3185 * jbd2_journal_start() write_dquot()
ac27a0ec 3186 * DQUOT_INIT() down(dqio_mutex)
dab291af 3187 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3188 *
3189 */
3190
3191#ifdef CONFIG_QUOTA
3192
3193static inline struct inode *dquot_to_inode(struct dquot *dquot)
3194{
3195 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3196}
3197
617ba13b 3198static int ext4_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 3199{
3200 handle_t *handle;
3201 int ret, err;
3202
3203 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 3204 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 3205 if (IS_ERR(handle))
3206 return PTR_ERR(handle);
3207 ret = dquot_initialize(inode, type);
617ba13b 3208 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3209 if (!ret)
3210 ret = err;
3211 return ret;
3212}
3213
617ba13b 3214static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 3215{
3216 handle_t *handle;
3217 int ret, err;
3218
3219 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 3220 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2887df13
JK		 3221 if (IS_ERR(handle)) {
3222 /*
3223 * We call dquot_drop() anyway to at least release references
3224 * to quota structures so that umount does not hang.
3225 */
3226 dquot_drop(inode);
ac27a0ec 3227 return PTR_ERR(handle);
2887df13 3228 }
ac27a0ec 3229 ret = dquot_drop(inode);
617ba13b 3230 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3231 if (!ret)
3232 ret = err;
3233 return ret;
3234}
3235
617ba13b 3236static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3237{
3238 int ret, err;
3239 handle_t *handle;
3240 struct inode *inode;
3241
3242 inode = dquot_to_inode(dquot);
617ba13b
MC		 3243 handle = ext4_journal_start(inode,
3244 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3245 if (IS_ERR(handle))
3246 return PTR_ERR(handle);
3247 ret = dquot_commit(dquot);
617ba13b 3248 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3249 if (!ret)
3250 ret = err;
3251 return ret;
3252}
3253
617ba13b 3254static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3255{
3256 int ret, err;
3257 handle_t *handle;
3258
617ba13b
MC		 3259 handle = ext4_journal_start(dquot_to_inode(dquot),
3260 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3261 if (IS_ERR(handle))
3262 return PTR_ERR(handle);
3263 ret = dquot_acquire(dquot);
617ba13b 3264 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3265 if (!ret)
3266 ret = err;
3267 return ret;
3268}
3269
617ba13b 3270static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3271{
3272 int ret, err;
3273 handle_t *handle;
3274
617ba13b
MC		 3275 handle = ext4_journal_start(dquot_to_inode(dquot),
3276 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3277 if (IS_ERR(handle)) {
3278 /* Release dquot anyway to avoid endless cycle in dqput() */
3279 dquot_release(dquot);
ac27a0ec 3280 return PTR_ERR(handle);
9c3013e9 3281 }
ac27a0ec 3282 ret = dquot_release(dquot);
617ba13b 3283 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3284 if (!ret)
3285 ret = err;
3286 return ret;
3287}
3288
617ba13b 3289static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3290{
2c8be6b2 3291 /* Are we journaling quotas? */
617ba13b
MC		 3292 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3293 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3294 dquot_mark_dquot_dirty(dquot);
617ba13b 3295 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3296 } else {
3297 return dquot_mark_dquot_dirty(dquot);
3298 }
3299}
3300
617ba13b 3301static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3302{
3303 int ret, err;
3304 handle_t *handle;
3305
3306 /* Data block + inode block */
617ba13b 3307 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3308 if (IS_ERR(handle))
3309 return PTR_ERR(handle);
3310 ret = dquot_commit_info(sb, type);
617ba13b 3311 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3312 if (!ret)
3313 ret = err;
3314 return ret;
3315}
3316
3317/*
3318 * Turn on quotas during mount time - we need to find
3319 * the quota file and such...
3320 */
617ba13b 3321static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3322{
617ba13b
MC		 3323 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3324 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3325}
3326
3327/*
3328 * Standard function to be called on quota_on
3329 */
617ba13b 3330static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6f28e087 3331 char *path, int remount)
ac27a0ec
DK		 3332{
3333 int err;
3334 struct nameidata nd;
3335
3336 if (!test_opt(sb, QUOTA))
3337 return -EINVAL;
0623543b
JK		 3338 /* When remounting, no checks are needed and in fact, path is NULL */
3339 if (remount)
6f28e087 3340 return vfs_quota_on(sb, type, format_id, path, remount);
0623543b 3341
ac27a0ec
DK		 3342 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3343 if (err)
3344 return err;
0623543b 3345
ac27a0ec 3346 /* Quotafile not on the same filesystem? */
4ac91378 3347 if (nd.path.mnt->mnt_sb != sb) {
1d957f9b 3348 path_put(&nd.path);
ac27a0ec
DK		 3349 return -EXDEV;
3350 }
0623543b
JK		 3351 /* Journaling quota? */
3352 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3353 /* Quotafile not in fs root? */
0623543b
JK		 3354 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3355 printk(KERN_WARNING
3356 "EXT4-fs: Quota file not on filesystem root. "
3357 "Journaled quota will not work.\n");
2b2d6d01 3358 }
0623543b
JK		 3359
3360 /*
3361 * When we journal data on quota file, we have to flush journal to see
3362 * all updates to the file when we bypass pagecache...
3363 */
3364 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3365 /*
3366 * We don't need to lock updates but journal_flush() could
3367 * otherwise be livelocked...
3368 */
3369 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3370 jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3371 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3372 }
3373
77e69dac 3374 err = vfs_quota_on_path(sb, type, format_id, &nd.path);
1d957f9b 3375 path_put(&nd.path);
77e69dac 3376 return err;
ac27a0ec
DK		 3377}
3378
3379/* Read data from quotafile - avoid pagecache and such because we cannot afford
3380 * acquiring the locks... As quota files are never truncated and quota code
3381 * itself serializes the operations (and noone else should touch the files)
3382 * we don't have to be afraid of races */
617ba13b 3383static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3384 size_t len, loff_t off)
3385{
3386 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3387 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3388 int err = 0;
3389 int offset = off & (sb->s_blocksize - 1);
3390 int tocopy;
3391 size_t toread;
3392 struct buffer_head *bh;
3393 loff_t i_size = i_size_read(inode);
3394
3395 if (off > i_size)
3396 return 0;
3397 if (off+len > i_size)
3398 len = i_size-off;
3399 toread = len;
3400 while (toread > 0) {
3401 tocopy = sb->s_blocksize - offset < toread ?
3402 sb->s_blocksize - offset : toread;
617ba13b 3403 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3404 if (err)
3405 return err;
3406 if (!bh) /* A hole? */
3407 memset(data, 0, tocopy);
3408 else
3409 memcpy(data, bh->b_data+offset, tocopy);
3410 brelse(bh);
3411 offset = 0;
3412 toread -= tocopy;
3413 data += tocopy;
3414 blk++;
3415 }
3416 return len;
3417}
3418
3419/* Write to quotafile (we know the transaction is already started and has
3420 * enough credits) */
617ba13b 3421static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3422 const char *data, size_t len, loff_t off)
3423{
3424 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3425 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3426 int err = 0;
3427 int offset = off & (sb->s_blocksize - 1);
3428 int tocopy;
617ba13b 3429 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3430 size_t towrite = len;
3431 struct buffer_head *bh;
3432 handle_t *handle = journal_current_handle();
3433
9c3013e9
JK		 3434 if (!handle) {
3435 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3436 " cancelled because transaction is not started.\n",
3437 (unsigned long long)off, (unsigned long long)len);
3438 return -EIO;
3439 }
ac27a0ec
DK		 3440 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3441 while (towrite > 0) {
3442 tocopy = sb->s_blocksize - offset < towrite ?
3443 sb->s_blocksize - offset : towrite;
617ba13b 3444 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3445 if (!bh)
3446 goto out;
3447 if (journal_quota) {
617ba13b 3448 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3449 if (err) {
3450 brelse(bh);
3451 goto out;
3452 }
3453 }
3454 lock_buffer(bh);
3455 memcpy(bh->b_data+offset, data, tocopy);
3456 flush_dcache_page(bh->b_page);
3457 unlock_buffer(bh);
3458 if (journal_quota)
617ba13b 3459 err = ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK		 3460 else {
3461 /* Always do at least ordered writes for quotas */
678aaf48 3462 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3463 mark_buffer_dirty(bh);
3464 }
3465 brelse(bh);
3466 if (err)
3467 goto out;
3468 offset = 0;
3469 towrite -= tocopy;
3470 data += tocopy;
3471 blk++;
3472 }
3473out:
4d04e4fb
JK		 3474 if (len == towrite) {
3475 mutex_unlock(&inode->i_mutex);
ac27a0ec 3476 return err;
4d04e4fb 3477 }
ac27a0ec
DK		 3478 if (inode->i_size < off+len-towrite) {
3479 i_size_write(inode, off+len-towrite);
617ba13b 3480 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3481 }
ac27a0ec 3482 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3483 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3484 mutex_unlock(&inode->i_mutex);
3485 return len - towrite;
3486}
3487
3488#endif
3489
617ba13b 3490static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3491 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3492{
617ba13b 3493 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3494}
3495
617ba13b 3496static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3497 .owner = THIS_MODULE,
617ba13b
MC		 3498 .name = "ext4dev",
3499 .get_sb = ext4_get_sb,
ac27a0ec
DK		 3500 .kill_sb = kill_block_super,
3501 .fs_flags = FS_REQUIRES_DEV,
3502};
3503
617ba13b 3504static int __init init_ext4_fs(void)
ac27a0ec 3505{
c9de560d
AT		 3506 int err;
3507
3508 err = init_ext4_mballoc();
ac27a0ec
DK		 3509 if (err)
3510 return err;
c9de560d
AT		 3511
3512 err = init_ext4_xattr();
3513 if (err)
3514 goto out2;
ac27a0ec
DK		 3515 err = init_inodecache();
3516 if (err)
3517 goto out1;
63f57933 3518 err = register_filesystem(&ext4dev_fs_type);
ac27a0ec
DK		 3519 if (err)
3520 goto out;
3521 return 0;
3522out:
3523 destroy_inodecache();
3524out1:
617ba13b 3525 exit_ext4_xattr();
c9de560d
AT		 3526out2:
3527 exit_ext4_mballoc();
ac27a0ec
DK		 3528 return err;
3529}
3530
617ba13b 3531static void __exit exit_ext4_fs(void)
ac27a0ec 3532{
617ba13b 3533 unregister_filesystem(&ext4dev_fs_type);
ac27a0ec 3534 destroy_inodecache();
617ba13b 3535 exit_ext4_xattr();
c9de560d 3536 exit_ext4_mballoc();
ac27a0ec
DK		 3537}
3538
3539MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
617ba13b 3540MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
ac27a0ec 3541MODULE_LICENSE("GPL");
617ba13b
MC		 3542module_init(init_ext4_fs)
3543module_exit(exit_ext4_fs)
Clone of davem's net-next-2.6 tree