]> 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
ext4: Fix long long checkpatch warnings
[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 != ',') {
4776004f 984 printk(KERN_ERR "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:
4776004f
TT		 1085 printk(KERN_ERR "EXT4 (no)user_xattr options "
1086 "not supported\n");
ac27a0ec
DK		 1087 break;
1088#endif
617ba13b 1089#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
ac27a0ec
DK		 1090 case Opt_acl:
1091 set_opt(sbi->s_mount_opt, POSIX_ACL);
1092 break;
1093 case Opt_noacl:
1094 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1095 break;
1096#else
1097 case Opt_acl:
1098 case Opt_noacl:
4776004f
TT		 1099 printk(KERN_ERR "EXT4 (no)acl options "
1100 "not supported\n");
ac27a0ec
DK		 1101 break;
1102#endif
1103 case Opt_reservation:
1104 set_opt(sbi->s_mount_opt, RESERVATION);
1105 break;
1106 case Opt_noreservation:
1107 clear_opt(sbi->s_mount_opt, RESERVATION);
1108 break;
1109 case Opt_journal_update:
1110 /* @@@ FIXME */
1111 /* Eventually we will want to be able to create
1112 a journal file here. For now, only allow the
1113 user to specify an existing inode to be the
1114 journal file. */
1115 if (is_remount) {
617ba13b 1116 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1117 "journal on remount\n");
1118 return 0;
1119 }
2b2d6d01 1120 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec
DK		 1121 break;
1122 case Opt_journal_inum:
1123 if (is_remount) {
617ba13b 1124 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1125 "journal on remount\n");
1126 return 0;
1127 }
1128 if (match_int(&args[0], &option))
1129 return 0;
1130 *inum = option;
1131 break;
1132 case Opt_journal_dev:
1133 if (is_remount) {
617ba13b 1134 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1135 "journal on remount\n");
1136 return 0;
1137 }
1138 if (match_int(&args[0], &option))
1139 return 0;
1140 *journal_devnum = option;
1141 break;
818d276c
GS		 1142 case Opt_journal_checksum:
1143 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1144 break;
1145 case Opt_journal_async_commit:
1146 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1147 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1148 break;
ac27a0ec 1149 case Opt_noload:
2b2d6d01 1150 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1151 break;
1152 case Opt_commit:
1153 if (match_int(&args[0], &option))
1154 return 0;
1155 if (option < 0)
1156 return 0;
1157 if (option == 0)
cd02ff0b 1158 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1159 sbi->s_commit_interval = HZ * option;
1160 break;
1161 case Opt_data_journal:
617ba13b 1162 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1163 goto datacheck;
1164 case Opt_data_ordered:
617ba13b 1165 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1166 goto datacheck;
1167 case Opt_data_writeback:
617ba13b 1168 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1169 datacheck:
1170 if (is_remount) {
617ba13b 1171 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1172 != data_opt) {
1173 printk(KERN_ERR
617ba13b 1174 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1175 "mode on remount\n");
1176 return 0;
1177 }
1178 } else {
617ba13b 1179 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1180 sbi->s_mount_opt |= data_opt;
1181 }
1182 break;
1183#ifdef CONFIG_QUOTA
1184 case Opt_usrjquota:
1185 qtype = USRQUOTA;
1186 goto set_qf_name;
1187 case Opt_grpjquota:
1188 qtype = GRPQUOTA;
1189set_qf_name:
dfc5d03f
JK		 1190 if ((sb_any_quota_enabled(sb) ||
1191 sb_any_quota_suspended(sb)) &&
1192 !sbi->s_qf_names[qtype]) {
ac27a0ec 1193 printk(KERN_ERR
4776004f
TT		 1194 "EXT4-fs: Cannot change journaled "
1195 "quota options when quota turned on.\n");
ac27a0ec
DK		 1196 return 0;
1197 }
1198 qname = match_strdup(&args[0]);
1199 if (!qname) {
1200 printk(KERN_ERR
617ba13b 1201 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1202 "storing quotafile name.\n");
1203 return 0;
1204 }
1205 if (sbi->s_qf_names[qtype] &&
1206 strcmp(sbi->s_qf_names[qtype], qname)) {
1207 printk(KERN_ERR
617ba13b 1208 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1209 "specified.\n", QTYPE2NAME(qtype));
1210 kfree(qname);
1211 return 0;
1212 }
1213 sbi->s_qf_names[qtype] = qname;
1214 if (strchr(sbi->s_qf_names[qtype], '/')) {
1215 printk(KERN_ERR
617ba13b 1216 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1217 "filesystem root.\n");
1218 kfree(sbi->s_qf_names[qtype]);
1219 sbi->s_qf_names[qtype] = NULL;
1220 return 0;
1221 }
1222 set_opt(sbi->s_mount_opt, QUOTA);
1223 break;
1224 case Opt_offusrjquota:
1225 qtype = USRQUOTA;
1226 goto clear_qf_name;
1227 case Opt_offgrpjquota:
1228 qtype = GRPQUOTA;
1229clear_qf_name:
dfc5d03f
JK		 1230 if ((sb_any_quota_enabled(sb) ||
1231 sb_any_quota_suspended(sb)) &&
1232 sbi->s_qf_names[qtype]) {
617ba13b 1233 printk(KERN_ERR "EXT4-fs: Cannot change "
2c8be6b2 1234 "journaled quota options when "
ac27a0ec
DK		 1235 "quota turned on.\n");
1236 return 0;
1237 }
1238 /*
1239 * The space will be released later when all options
1240 * are confirmed to be correct
1241 */
1242 sbi->s_qf_names[qtype] = NULL;
1243 break;
1244 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1245 qfmt = QFMT_VFS_OLD;
1246 goto set_qf_format;
ac27a0ec 1247 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1248 qfmt = QFMT_VFS_V0;
1249set_qf_format:
1250 if ((sb_any_quota_enabled(sb) ||
1251 sb_any_quota_suspended(sb)) &&
1252 sbi->s_jquota_fmt != qfmt) {
1253 printk(KERN_ERR "EXT4-fs: Cannot change "
1254 "journaled quota options when "
1255 "quota turned on.\n");
1256 return 0;
1257 }
1258 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1259 break;
1260 case Opt_quota:
1261 case Opt_usrquota:
1262 set_opt(sbi->s_mount_opt, QUOTA);
1263 set_opt(sbi->s_mount_opt, USRQUOTA);
1264 break;
1265 case Opt_grpquota:
1266 set_opt(sbi->s_mount_opt, QUOTA);
1267 set_opt(sbi->s_mount_opt, GRPQUOTA);
1268 break;
1269 case Opt_noquota:
1270 if (sb_any_quota_enabled(sb)) {
617ba13b 1271 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1272 "options when quota turned on.\n");
1273 return 0;
1274 }
1275 clear_opt(sbi->s_mount_opt, QUOTA);
1276 clear_opt(sbi->s_mount_opt, USRQUOTA);
1277 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1278 break;
1279#else
1280 case Opt_quota:
1281 case Opt_usrquota:
1282 case Opt_grpquota:
cd59e7b9
JK		 1283 printk(KERN_ERR
1284 "EXT4-fs: quota options not supported.\n");
1285 break;
ac27a0ec
DK		 1286 case Opt_usrjquota:
1287 case Opt_grpjquota:
1288 case Opt_offusrjquota:
1289 case Opt_offgrpjquota:
1290 case Opt_jqfmt_vfsold:
1291 case Opt_jqfmt_vfsv0:
1292 printk(KERN_ERR
cd59e7b9 1293 "EXT4-fs: journaled quota options not "
ac27a0ec
DK		 1294 "supported.\n");
1295 break;
1296 case Opt_noquota:
1297 break;
1298#endif
1299 case Opt_abort:
1300 set_opt(sbi->s_mount_opt, ABORT);
1301 break;
1302 case Opt_barrier:
1303 if (match_int(&args[0], &option))
1304 return 0;
1305 if (option)
1306 set_opt(sbi->s_mount_opt, BARRIER);
1307 else
1308 clear_opt(sbi->s_mount_opt, BARRIER);
1309 break;
1310 case Opt_ignore:
1311 break;
1312 case Opt_resize:
1313 if (!is_remount) {
617ba13b 1314 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1315 "for remount\n");
1316 return 0;
1317 }
1318 if (match_int(&args[0], &option) != 0)
1319 return 0;
1320 *n_blocks_count = option;
1321 break;
1322 case Opt_nobh:
1323 set_opt(sbi->s_mount_opt, NOBH);
1324 break;
1325 case Opt_bh:
1326 clear_opt(sbi->s_mount_opt, NOBH);
1327 break;
a86c6181 1328 case Opt_extents:
e4079a11
ES		 1329 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1330 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1331 ext4_warning(sb, __func__,
1332 "extents feature not enabled "
1333 "on this filesystem, use tune2fs\n");
1334 return 0;
1335 }
2b2d6d01 1336 set_opt(sbi->s_mount_opt, EXTENTS);
a86c6181 1337 break;
1e2462f9 1338 case Opt_noextents:
c07651b5
AK		 1339 /*
1340 * When e2fsprogs support resizing an already existing
1341 * ext3 file system to greater than 2**32 we need to
1342 * add support to block allocator to handle growing
1343 * already existing block mapped inode so that blocks
1344 * allocated for them fall within 2**32
1345 */
1346 last_block = ext4_blocks_count(sbi->s_es) - 1;
1347 if (last_block > 0xffffffffULL) {
1348 printk(KERN_ERR "EXT4-fs: Filesystem too "
1349 "large to mount with "
1350 "-o noextents options\n");
1351 return 0;
1352 }
2b2d6d01 1353 clear_opt(sbi->s_mount_opt, EXTENTS);
1e2462f9 1354 break;
25ec56b5
JNC		 1355 case Opt_i_version:
1356 set_opt(sbi->s_mount_opt, I_VERSION);
1357 sb->s_flags |= MS_I_VERSION;
1358 break;
dd919b98
AK		 1359 case Opt_nodelalloc:
1360 clear_opt(sbi->s_mount_opt, DELALLOC);
1361 break;
c9de560d
AT		 1362 case Opt_mballoc:
1363 set_opt(sbi->s_mount_opt, MBALLOC);
1364 break;
1365 case Opt_nomballoc:
1366 clear_opt(sbi->s_mount_opt, MBALLOC);
1367 break;
1368 case Opt_stripe:
1369 if (match_int(&args[0], &option))
1370 return 0;
1371 if (option < 0)
1372 return 0;
1373 sbi->s_stripe = option;
1374 break;
64769240
AT		 1375 case Opt_delalloc:
1376 set_opt(sbi->s_mount_opt, DELALLOC);
1377 break;
ac27a0ec 1378 default:
2b2d6d01
TT		 1379 printk(KERN_ERR
1380 "EXT4-fs: Unrecognized mount option \"%s\" "
1381 "or missing value\n", p);
ac27a0ec
DK		 1382 return 0;
1383 }
1384 }
1385#ifdef CONFIG_QUOTA
1386 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1387 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1388 sbi->s_qf_names[USRQUOTA])
1389 clear_opt(sbi->s_mount_opt, USRQUOTA);
1390
617ba13b 1391 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1392 sbi->s_qf_names[GRPQUOTA])
1393 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1394
1395 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1396 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1397 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1398 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1399 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1400 "format mixing.\n");
1401 return 0;
1402 }
1403
1404 if (!sbi->s_jquota_fmt) {
2c8be6b2 1405 printk(KERN_ERR "EXT4-fs: journaled quota format "
ac27a0ec
DK		 1406 "not specified.\n");
1407 return 0;
1408 }
1409 } else {
1410 if (sbi->s_jquota_fmt) {
2c8be6b2
JK		 1411 printk(KERN_ERR "EXT4-fs: journaled quota format "
1412 "specified with no journaling "
ac27a0ec
DK		 1413 "enabled.\n");
1414 return 0;
1415 }
1416 }
1417#endif
1418 return 1;
1419}
1420
617ba13b 1421static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1422 int read_only)
1423{
617ba13b 1424 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1425 int res = 0;
1426
617ba13b 1427 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2b2d6d01
TT		 1428 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1429 "forcing read-only mode\n");
ac27a0ec
DK		 1430 res = MS_RDONLY;
1431 }
1432 if (read_only)
1433 return res;
617ba13b 1434 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2b2d6d01
TT		 1435 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1436 "running e2fsck is recommended\n");
617ba13b 1437 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
2b2d6d01
TT		 1438 printk(KERN_WARNING
1439 "EXT4-fs warning: mounting fs with errors, "
1440 "running e2fsck is recommended\n");
ac27a0ec
DK		 1441 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1442 le16_to_cpu(es->s_mnt_count) >=
1443 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2b2d6d01
TT		 1444 printk(KERN_WARNING
1445 "EXT4-fs warning: maximal mount count reached, "
1446 "running e2fsck is recommended\n");
ac27a0ec
DK		 1447 else if (le32_to_cpu(es->s_checkinterval) &&
1448 (le32_to_cpu(es->s_lastcheck) +
1449 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2b2d6d01
TT		 1450 printk(KERN_WARNING
1451 "EXT4-fs warning: checktime reached, "
1452 "running e2fsck is recommended\n");
ac27a0ec
DK		 1453#if 0
1454 /* @@@ We _will_ want to clear the valid bit if we find
63f57933
AM		 1455 * inconsistencies, to force a fsck at reboot. But for
1456 * a plain journaled filesystem we can keep it set as
1457 * valid forever! :)
1458 */
216c34b2 1459 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec
DK		 1460#endif
1461 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1462 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1463 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1464 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b
MC		 1465 ext4_update_dynamic_rev(sb);
1466 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1467
617ba13b 1468 ext4_commit_super(sb, es, 1);
ac27a0ec 1469 if (test_opt(sb, DEBUG))
617ba13b 1470 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
ac27a0ec
DK		 1471 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1472 sb->s_blocksize,
1473 sbi->s_groups_count,
617ba13b
MC		 1474 EXT4_BLOCKS_PER_GROUP(sb),
1475 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1476 sbi->s_mount_opt);
1477
617ba13b 1478 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
ac27a0ec
DK		 1479 char b[BDEVNAME_SIZE];
1480
4776004f
TT		 1481 printk(KERN_INFO "EXT4 FS on %s, external journal on %s\n",
1482 sb->s_id, bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
ac27a0ec 1483 } else {
4776004f
TT		 1484 printk(KERN_INFO "EXT4 FS on %s, internal journal\n",
1485 sb->s_id);
ac27a0ec
DK		 1486 }
1487 return res;
1488}
1489
772cb7c8
JS		 1490static int ext4_fill_flex_info(struct super_block *sb)
1491{
1492 struct ext4_sb_info *sbi = EXT4_SB(sb);
1493 struct ext4_group_desc *gdp = NULL;
1494 struct buffer_head *bh;
1495 ext4_group_t flex_group_count;
1496 ext4_group_t flex_group;
1497 int groups_per_flex = 0;
1498 __u64 block_bitmap = 0;
1499 int i;
1500
1501 if (!sbi->s_es->s_log_groups_per_flex) {
1502 sbi->s_log_groups_per_flex = 0;
1503 return 1;
1504 }
1505
1506 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1507 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1508
1509 flex_group_count = (sbi->s_groups_count + groups_per_flex - 1) /
1510 groups_per_flex;
ec05e868 1511 sbi->s_flex_groups = kzalloc(flex_group_count *
772cb7c8
JS		 1512 sizeof(struct flex_groups), GFP_KERNEL);
1513 if (sbi->s_flex_groups == NULL) {
ec05e868
LZ		 1514 printk(KERN_ERR "EXT4-fs: not enough memory for "
1515 "%lu flex groups\n", flex_group_count);
772cb7c8
JS		 1516 goto failed;
1517 }
772cb7c8
JS		 1518
1519 gdp = ext4_get_group_desc(sb, 1, &bh);
1520 block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1521
1522 for (i = 0; i < sbi->s_groups_count; i++) {
1523 gdp = ext4_get_group_desc(sb, i, &bh);
1524
1525 flex_group = ext4_flex_group(sbi, i);
1526 sbi->s_flex_groups[flex_group].free_inodes +=
1527 le16_to_cpu(gdp->bg_free_inodes_count);
1528 sbi->s_flex_groups[flex_group].free_blocks +=
1529 le16_to_cpu(gdp->bg_free_blocks_count);
1530 }
1531
1532 return 1;
1533failed:
1534 return 0;
1535}
1536
717d50e4
AD		 1537__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1538 struct ext4_group_desc *gdp)
1539{
1540 __u16 crc = 0;
1541
1542 if (sbi->s_es->s_feature_ro_compat &
1543 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1544 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1545 __le32 le_group = cpu_to_le32(block_group);
1546
1547 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1548 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1549 crc = crc16(crc, (__u8 *)gdp, offset);
1550 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1551 /* for checksum of struct ext4_group_desc do the rest...*/
1552 if ((sbi->s_es->s_feature_incompat &
1553 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1554 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1555 crc = crc16(crc, (__u8 *)gdp + offset,
1556 le16_to_cpu(sbi->s_es->s_desc_size) -
1557 offset);
1558 }
1559
1560 return cpu_to_le16(crc);
1561}
1562
1563int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1564 struct ext4_group_desc *gdp)
1565{
1566 if ((sbi->s_es->s_feature_ro_compat &
1567 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1568 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1569 return 0;
1570
1571 return 1;
1572}
1573
ac27a0ec 1574/* Called at mount-time, super-block is locked */
197cd65a 1575static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1576{
617ba13b
MC		 1577 struct ext4_sb_info *sbi = EXT4_SB(sb);
1578 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1579 ext4_fsblk_t last_block;
bd81d8ee
LV		 1580 ext4_fsblk_t block_bitmap;
1581 ext4_fsblk_t inode_bitmap;
1582 ext4_fsblk_t inode_table;
ce421581 1583 int flexbg_flag = 0;
fd2d4291 1584 ext4_group_t i;
ac27a0ec 1585
ce421581
JS		 1586 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1587 flexbg_flag = 1;
1588
617ba13b 1589 ext4_debug ("Checking group descriptors");
ac27a0ec 1590
197cd65a
AM		 1591 for (i = 0; i < sbi->s_groups_count; i++) {
1592 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1593
ce421581 1594 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1595 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1596 else
1597 last_block = first_block +
617ba13b 1598 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1599
8fadc143 1600 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1601 if (block_bitmap < first_block || block_bitmap > last_block) {
c19204b0
JB		 1602 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1603 "Block bitmap for group %lu not in group "
1604 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1605 return 0;
1606 }
8fadc143 1607 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1608 if (inode_bitmap < first_block || inode_bitmap > last_block) {
c19204b0
JB		 1609 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1610 "Inode bitmap for group %lu not in group "
1611 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1612 return 0;
1613 }
8fadc143 1614 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1615 if (inode_table < first_block ||
2b2d6d01 1616 inode_table + sbi->s_itb_per_group - 1 > last_block) {
c19204b0
JB		 1617 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1618 "Inode table for group %lu not in group "
1619 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1620 return 0;
1621 }
b5f10eed 1622 spin_lock(sb_bgl_lock(sbi, i));
717d50e4 1623 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
c19204b0
JB		 1624 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1625 "Checksum for group %lu failed (%u!=%u)\n",
1626 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1627 gdp)), le16_to_cpu(gdp->bg_checksum));
8a266467
TT		 1628 if (!(sb->s_flags & MS_RDONLY))
1629 return 0;
717d50e4 1630 }
b5f10eed 1631 spin_unlock(sb_bgl_lock(sbi, i));
ce421581
JS		 1632 if (!flexbg_flag)
1633 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1634 }
1635
bd81d8ee 1636 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2b2d6d01 1637 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1638 return 1;
1639}
1640
617ba13b 1641/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1642 * the superblock) which were deleted from all directories, but held open by
1643 * a process at the time of a crash. We walk the list and try to delete these
1644 * inodes at recovery time (only with a read-write filesystem).
1645 *
1646 * In order to keep the orphan inode chain consistent during traversal (in
1647 * case of crash during recovery), we link each inode into the superblock
1648 * orphan list_head and handle it the same way as an inode deletion during
1649 * normal operation (which journals the operations for us).
1650 *
1651 * We only do an iget() and an iput() on each inode, which is very safe if we
1652 * accidentally point at an in-use or already deleted inode. The worst that
1653 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1654 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1655 * e2fsck was run on this filesystem, and it must have already done the orphan
1656 * inode cleanup for us, so we can safely abort without any further action.
1657 */
2b2d6d01
TT		 1658static void ext4_orphan_cleanup(struct super_block *sb,
1659 struct ext4_super_block *es)
ac27a0ec
DK		 1660{
1661 unsigned int s_flags = sb->s_flags;
1662 int nr_orphans = 0, nr_truncates = 0;
1663#ifdef CONFIG_QUOTA
1664 int i;
1665#endif
1666 if (!es->s_last_orphan) {
1667 jbd_debug(4, "no orphan inodes to clean up\n");
1668 return;
1669 }
1670
a8f48a95
ES		 1671 if (bdev_read_only(sb->s_bdev)) {
1672 printk(KERN_ERR "EXT4-fs: write access "
1673 "unavailable, skipping orphan cleanup.\n");
1674 return;
1675 }
1676
617ba13b 1677 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1678 if (es->s_last_orphan)
1679 jbd_debug(1, "Errors on filesystem, "
1680 "clearing orphan list.\n");
1681 es->s_last_orphan = 0;
1682 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1683 return;
1684 }
1685
1686 if (s_flags & MS_RDONLY) {
617ba13b 1687 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1688 sb->s_id);
1689 sb->s_flags &= ~MS_RDONLY;
1690 }
1691#ifdef CONFIG_QUOTA
1692 /* Needed for iput() to work correctly and not trash data */
1693 sb->s_flags |= MS_ACTIVE;
1694 /* Turn on quotas so that they are updated correctly */
1695 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1696 if (EXT4_SB(sb)->s_qf_names[i]) {
1697 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1698 if (ret < 0)
1699 printk(KERN_ERR
2c8be6b2 1700 "EXT4-fs: Cannot turn on journaled "
ac27a0ec
DK		 1701 "quota: error %d\n", ret);
1702 }
1703 }
1704#endif
1705
1706 while (es->s_last_orphan) {
1707 struct inode *inode;
1708
97bd42b9
JB		 1709 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1710 if (IS_ERR(inode)) {
ac27a0ec
DK		 1711 es->s_last_orphan = 0;
1712 break;
1713 }
1714
617ba13b 1715 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1716 DQUOT_INIT(inode);
1717 if (inode->i_nlink) {
1718 printk(KERN_DEBUG
e5f8eab8 1719 "%s: truncating inode %lu to %lld bytes\n",
46e665e9 1720 __func__, inode->i_ino, inode->i_size);
e5f8eab8 1721 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 1722 inode->i_ino, inode->i_size);
617ba13b 1723 ext4_truncate(inode);
ac27a0ec
DK		 1724 nr_truncates++;
1725 } else {
1726 printk(KERN_DEBUG
1727 "%s: deleting unreferenced inode %lu\n",
46e665e9 1728 __func__, inode->i_ino);
ac27a0ec
DK		 1729 jbd_debug(2, "deleting unreferenced inode %lu\n",
1730 inode->i_ino);
1731 nr_orphans++;
1732 }
1733 iput(inode); /* The delete magic happens here! */
1734 }
1735
2b2d6d01 1736#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 1737
1738 if (nr_orphans)
617ba13b 1739 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1740 sb->s_id, PLURAL(nr_orphans));
1741 if (nr_truncates)
617ba13b 1742 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1743 sb->s_id, PLURAL(nr_truncates));
1744#ifdef CONFIG_QUOTA
1745 /* Turn quotas off */
1746 for (i = 0; i < MAXQUOTAS; i++) {
1747 if (sb_dqopt(sb)->files[i])
6f28e087 1748 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1749 }
1750#endif
1751 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1752}
cd2291a4
ES		 1753/*
1754 * Maximal extent format file size.
1755 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1756 * extent format containers, within a sector_t, and within i_blocks
1757 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1758 * so that won't be a limiting factor.
1759 *
1760 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1761 */
1762static loff_t ext4_max_size(int blkbits)
1763{
1764 loff_t res;
1765 loff_t upper_limit = MAX_LFS_FILESIZE;
1766
1767 /* small i_blocks in vfs inode? */
1768 if (sizeof(blkcnt_t) < sizeof(u64)) {
1769 /*
1770 * CONFIG_LSF is not enabled implies the inode
1771 * i_block represent total blocks in 512 bytes
1772 * 32 == size of vfs inode i_blocks * 8
1773 */
1774 upper_limit = (1LL << 32) - 1;
1775
1776 /* total blocks in file system block size */
1777 upper_limit >>= (blkbits - 9);
1778 upper_limit <<= blkbits;
1779 }
1780
1781 /* 32-bit extent-start container, ee_block */
1782 res = 1LL << 32;
1783 res <<= blkbits;
1784 res -= 1;
1785
1786 /* Sanity check against vm- & vfs- imposed limits */
1787 if (res > upper_limit)
1788 res = upper_limit;
1789
1790 return res;
1791}
ac27a0ec 1792
ac27a0ec 1793/*
cd2291a4 1794 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1795 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1796 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1797 */
cd2291a4 1798static loff_t ext4_max_bitmap_size(int bits)
ac27a0ec 1799{
617ba13b 1800 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1801 int meta_blocks;
1802 loff_t upper_limit;
1803 /* This is calculated to be the largest file size for a
cd2291a4 1804 * dense, bitmapped file such that the total number of
ac27a0ec 1805 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1806 * does not exceed 2^48 -1
1807 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1808 * total number of 512 bytes blocks of the file
1809 */
1810
1811 if (sizeof(blkcnt_t) < sizeof(u64)) {
1812 /*
1813 * CONFIG_LSF is not enabled implies the inode
1814 * i_block represent total blocks in 512 bytes
1815 * 32 == size of vfs inode i_blocks * 8
1816 */
1817 upper_limit = (1LL << 32) - 1;
1818
1819 /* total blocks in file system block size */
1820 upper_limit >>= (bits - 9);
1821
1822 } else {
8180a562
AK		 1823 /*
1824 * We use 48 bit ext4_inode i_blocks
1825 * With EXT4_HUGE_FILE_FL set the i_blocks
1826 * represent total number of blocks in
1827 * file system block size
1828 */
0fc1b451
AK		 1829 upper_limit = (1LL << 48) - 1;
1830
0fc1b451
AK		 1831 }
1832
1833 /* indirect blocks */
1834 meta_blocks = 1;
1835 /* double indirect blocks */
1836 meta_blocks += 1 + (1LL << (bits-2));
1837 /* tripple indirect blocks */
1838 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1839
1840 upper_limit -= meta_blocks;
1841 upper_limit <<= bits;
ac27a0ec
DK		 1842
1843 res += 1LL << (bits-2);
1844 res += 1LL << (2*(bits-2));
1845 res += 1LL << (3*(bits-2));
1846 res <<= bits;
1847 if (res > upper_limit)
1848 res = upper_limit;
0fc1b451
AK		 1849
1850 if (res > MAX_LFS_FILESIZE)
1851 res = MAX_LFS_FILESIZE;
1852
ac27a0ec
DK		 1853 return res;
1854}
1855
617ba13b 1856static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1857 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1858{
617ba13b 1859 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1860 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1861 int has_super = 0;
1862
1863 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1864
617ba13b 1865 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1866 nr < first_meta_bg)
70bbb3e0 1867 return logical_sb_block + nr + 1;
ac27a0ec 1868 bg = sbi->s_desc_per_block * nr;
617ba13b 1869 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1870 has_super = 1;
617ba13b 1871 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1872}
1873
c9de560d
AT		 1874/**
1875 * ext4_get_stripe_size: Get the stripe size.
1876 * @sbi: In memory super block info
1877 *
1878 * If we have specified it via mount option, then
1879 * use the mount option value. If the value specified at mount time is
1880 * greater than the blocks per group use the super block value.
1881 * If the super block value is greater than blocks per group return 0.
1882 * Allocator needs it be less than blocks per group.
1883 *
1884 */
1885static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1886{
1887 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1888 unsigned long stripe_width =
1889 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1890
1891 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1892 return sbi->s_stripe;
1893
1894 if (stripe_width <= sbi->s_blocks_per_group)
1895 return stripe_width;
1896
1897 if (stride <= sbi->s_blocks_per_group)
1898 return stride;
1899
1900 return 0;
1901}
ac27a0ec 1902
2b2d6d01 1903static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 1904 __releases(kernel_lock)
1905 __acquires(kernel_lock)
1d03ec98 1906
ac27a0ec 1907{
2b2d6d01 1908 struct buffer_head *bh;
617ba13b
MC		 1909 struct ext4_super_block *es = NULL;
1910 struct ext4_sb_info *sbi;
1911 ext4_fsblk_t block;
1912 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 1913 ext4_fsblk_t logical_sb_block;
ac27a0ec
DK		 1914 unsigned long offset = 0;
1915 unsigned int journal_inum = 0;
1916 unsigned long journal_devnum = 0;
1917 unsigned long def_mount_opts;
1918 struct inode *root;
1d1fe1ee 1919 int ret = -EINVAL;
ac27a0ec 1920 int blocksize;
ac27a0ec
DK		 1921 int db_count;
1922 int i;
1923 int needs_recovery;
1924 __le32 features;
bd81d8ee 1925 __u64 blocks_count;
833f4077 1926 int err;
ac27a0ec
DK		 1927
1928 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1929 if (!sbi)
1930 return -ENOMEM;
1931 sb->s_fs_info = sbi;
1932 sbi->s_mount_opt = 0;
617ba13b
MC		 1933 sbi->s_resuid = EXT4_DEF_RESUID;
1934 sbi->s_resgid = EXT4_DEF_RESGID;
d9c9bef1 1935 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 1936
1937 unlock_kernel();
1938
617ba13b 1939 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 1940 if (!blocksize) {
617ba13b 1941 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 1942 goto out_fail;
1943 }
1944
1945 /*
617ba13b 1946 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 1947 * block sizes. We need to calculate the offset from buffer start.
1948 */
617ba13b 1949 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 1950 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1951 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 1952 } else {
70bbb3e0 1953 logical_sb_block = sb_block;
ac27a0ec
DK		 1954 }
1955
70bbb3e0 1956 if (!(bh = sb_bread(sb, logical_sb_block))) {
2b2d6d01 1957 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 1958 goto out_fail;
1959 }
1960 /*
1961 * Note: s_es must be initialized as soon as possible because
617ba13b 1962 * some ext4 macro-instructions depend on its value
ac27a0ec 1963 */
617ba13b 1964 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 1965 sbi->s_es = es;
1966 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 1967 if (sb->s_magic != EXT4_SUPER_MAGIC)
1968 goto cantfind_ext4;
ac27a0ec
DK		 1969
1970 /* Set defaults before we parse the mount options */
1971 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 1972 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 1973 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 1974 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 1975 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 1976 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 1977 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1978#ifdef CONFIG_EXT4DEV_FS_XATTR
617ba13b 1979 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 1980 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1981#endif
1982#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
617ba13b 1983 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 1984 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1985#endif
617ba13b
MC		 1986 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1987 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1988 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1989 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1990 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1991 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1992
1993 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 1994 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 1995 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 1996 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 1997 else
1998 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1999
2000 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2001 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2002
2003 set_opt(sbi->s_mount_opt, RESERVATION);
571640ca 2004 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2005
1e2462f9
MC		 2006 /*
2007 * turn on extents feature by default in ext4 filesystem
e4079a11
ES		 2008 * only if feature flag already set by mkfs or tune2fs.
2009 * Use -o noextents to turn it off
1e2462f9 2010 */
e4079a11
ES		 2011 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
2012 set_opt(sbi->s_mount_opt, EXTENTS);
2013 else
2014 ext4_warning(sb, __func__,
2015 "extents feature not enabled on this filesystem, "
2016 "use tune2fs.\n");
3dbd0ede 2017 /*
e4079a11
ES		 2018 * turn on mballoc code by default in ext4 filesystem
2019 * Use -o nomballoc to turn it off
3dbd0ede
AK		 2020 */
2021 set_opt(sbi->s_mount_opt, MBALLOC);
1e2462f9 2022
dd919b98
AK		 2023 /*
2024 * enable delayed allocation by default
2025 * Use -o nodelalloc to turn it off
2026 */
2027 set_opt(sbi->s_mount_opt, DELALLOC);
2028
2029
2b2d6d01
TT		 2030 if (!parse_options((char *) data, sb, &journal_inum, &journal_devnum,
2031 NULL, 0))
ac27a0ec
DK		 2032 goto failed_mount;
2033
2034 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2035 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2036
617ba13b
MC		 2037 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2038 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2039 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2040 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 2041 printk(KERN_WARNING
617ba13b 2042 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec 2043 "running e2fsck is recommended\n");
469108ff
TT		 2044
2045 /*
2046 * Since ext4 is still considered development code, we require
2047 * that the TEST_FILESYS flag in s->flags be set.
2048 */
2049 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
2050 printk(KERN_WARNING "EXT4-fs: %s: not marked "
2051 "OK to use with test code.\n", sb->s_id);
2052 goto failed_mount;
2053 }
2054
ac27a0ec
DK		 2055 /*
2056 * Check feature flags regardless of the revision level, since we
2057 * previously didn't change the revision level when setting the flags,
2058 * so there is a chance incompat flags are set on a rev 0 filesystem.
2059 */
617ba13b 2060 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 2061 if (features) {
617ba13b 2062 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
ac27a0ec
DK		 2063 "unsupported optional features (%x).\n",
2064 sb->s_id, le32_to_cpu(features));
2065 goto failed_mount;
2066 }
617ba13b 2067 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 2068 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 2069 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
ac27a0ec
DK		 2070 "unsupported optional features (%x).\n",
2071 sb->s_id, le32_to_cpu(features));
2072 goto failed_mount;
2073 }
0fc1b451
AK		 2074 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2075 /*
2076 * Large file size enabled file system can only be
2077 * mount if kernel is build with CONFIG_LSF
2078 */
2079 if (sizeof(root->i_blocks) < sizeof(u64) &&
2080 !(sb->s_flags & MS_RDONLY)) {
2081 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2082 "files cannot be mounted read-write "
2083 "without CONFIG_LSF.\n", sb->s_id);
2084 goto failed_mount;
2085 }
2086 }
ac27a0ec
DK		 2087 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2088
617ba13b
MC		 2089 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2090 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 2091 printk(KERN_ERR
617ba13b 2092 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 2093 blocksize, sb->s_id);
2094 goto failed_mount;
2095 }
2096
ac27a0ec 2097 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2098
2099 /* Validate the filesystem blocksize */
2100 if (!sb_set_blocksize(sb, blocksize)) {
2101 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2102 blocksize);
ac27a0ec
DK		 2103 goto failed_mount;
2104 }
2105
2b2d6d01 2106 brelse(bh);
70bbb3e0
AM		 2107 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2108 offset = do_div(logical_sb_block, blocksize);
2109 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 2110 if (!bh) {
2111 printk(KERN_ERR
617ba13b 2112 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 2113 goto failed_mount;
2114 }
617ba13b 2115 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2116 sbi->s_es = es;
617ba13b 2117 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2b2d6d01
TT		 2118 printk(KERN_ERR
2119 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 2120 goto failed_mount;
2121 }
2122 }
2123
e2b46574 2124 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
617ba13b 2125 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
ac27a0ec 2126
617ba13b
MC		 2127 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2128 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2129 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2130 } else {
2131 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2132 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2133 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2134 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2135 (sbi->s_inode_size > blocksize)) {
2b2d6d01
TT		 2136 printk(KERN_ERR
2137 "EXT4-fs: unsupported inode size: %d\n",
2138 sbi->s_inode_size);
ac27a0ec
DK		 2139 goto failed_mount;
2140 }
ef7f3835
KS		 2141 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2142 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2143 }
0d1ee42f
AR		 2144 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2145 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2146 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2147 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2148 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2149 printk(KERN_ERR
8fadc143 2150 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2151 sbi->s_desc_size);
2152 goto failed_mount;
2153 }
2154 } else
2155 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2156 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2157 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2158 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2159 goto cantfind_ext4;
2160 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2161 if (sbi->s_inodes_per_block == 0)
617ba13b 2162 goto cantfind_ext4;
ac27a0ec
DK		 2163 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2164 sbi->s_inodes_per_block;
0d1ee42f 2165 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2166 sbi->s_sbh = bh;
2167 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2168 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2169 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2b2d6d01 2170 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2171 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2172 sbi->s_def_hash_version = es->s_def_hash_version;
2173
2174 if (sbi->s_blocks_per_group > blocksize * 8) {
2b2d6d01
TT		 2175 printk(KERN_ERR
2176 "EXT4-fs: #blocks per group too big: %lu\n",
2177 sbi->s_blocks_per_group);
ac27a0ec
DK		 2178 goto failed_mount;
2179 }
ac27a0ec 2180 if (sbi->s_inodes_per_group > blocksize * 8) {
2b2d6d01
TT		 2181 printk(KERN_ERR
2182 "EXT4-fs: #inodes per group too big: %lu\n",
2183 sbi->s_inodes_per_group);
ac27a0ec
DK		 2184 goto failed_mount;
2185 }
2186
bd81d8ee 2187 if (ext4_blocks_count(es) >
ac27a0ec 2188 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2189 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2190 " too large to mount safely\n", sb->s_id);
2191 if (sizeof(sector_t) < 8)
617ba13b 2192 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2193 "enabled\n");
2194 goto failed_mount;
2195 }
2196
617ba13b
MC		 2197 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2198 goto cantfind_ext4;
e7c95593
ES		 2199
2200 /* ensure blocks_count calculation below doesn't sign-extend */
2201 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2202 le32_to_cpu(es->s_first_data_block) + 1) {
2203 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2204 "first data block %u, blocks per group %lu\n",
2205 ext4_blocks_count(es),
2206 le32_to_cpu(es->s_first_data_block),
2207 EXT4_BLOCKS_PER_GROUP(sb));
2208 goto failed_mount;
2209 }
bd81d8ee
LV		 2210 blocks_count = (ext4_blocks_count(es) -
2211 le32_to_cpu(es->s_first_data_block) +
2212 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2213 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2214 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2215 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2216 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2217 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2218 GFP_KERNEL);
2219 if (sbi->s_group_desc == NULL) {
2b2d6d01 2220 printk(KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2221 goto failed_mount;
2222 }
2223
2224 bgl_lock_init(&sbi->s_blockgroup_lock);
2225
2226 for (i = 0; i < db_count; i++) {
70bbb3e0 2227 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2228 sbi->s_group_desc[i] = sb_bread(sb, block);
2229 if (!sbi->s_group_desc[i]) {
2b2d6d01
TT		 2230 printk(KERN_ERR "EXT4-fs: "
2231 "can't read group descriptor %d\n", i);
ac27a0ec
DK		 2232 db_count = i;
2233 goto failed_mount2;
2234 }
2235 }
2b2d6d01 2236 if (!ext4_check_descriptors(sb)) {
617ba13b 2237 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2238 goto failed_mount2;
2239 }
772cb7c8
JS		 2240 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2241 if (!ext4_fill_flex_info(sb)) {
2242 printk(KERN_ERR
2243 "EXT4-fs: unable to initialize "
2244 "flex_bg meta info!\n");
2245 goto failed_mount2;
2246 }
2247
ac27a0ec
DK		 2248 sbi->s_gdb_count = db_count;
2249 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2250 spin_lock_init(&sbi->s_next_gen_lock);
2251
833f4077
PZ		 2252 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2253 ext4_count_free_blocks(sb));
2254 if (!err) {
2255 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2256 ext4_count_free_inodes(sb));
2257 }
2258 if (!err) {
2259 err = percpu_counter_init(&sbi->s_dirs_counter,
2260 ext4_count_dirs(sb));
2261 }
2262 if (err) {
2263 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2264 goto failed_mount3;
2265 }
ac27a0ec
DK		 2266
2267 /* per fileystem reservation list head & lock */
2268 spin_lock_init(&sbi->s_rsv_window_lock);
2269 sbi->s_rsv_window_root = RB_ROOT;
2270 /* Add a single, static dummy reservation to the start of the
2271 * reservation window list --- it gives us a placeholder for
2272 * append-at-start-of-list which makes the allocation logic
2273 * _much_ simpler. */
617ba13b
MC		 2274 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2275 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
ac27a0ec
DK		 2276 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2277 sbi->s_rsv_window_head.rsv_goal_size = 0;
617ba13b 2278 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
ac27a0ec 2279
c9de560d
AT		 2280 sbi->s_stripe = ext4_get_stripe_size(sbi);
2281
ac27a0ec
DK		 2282 /*
2283 * set up enough so that it can read an inode
2284 */
617ba13b
MC		 2285 sb->s_op = &ext4_sops;
2286 sb->s_export_op = &ext4_export_ops;
2287 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2288#ifdef CONFIG_QUOTA
617ba13b
MC		 2289 sb->s_qcop = &ext4_qctl_operations;
2290 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2291#endif
2292 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2293
2294 sb->s_root = NULL;
2295
2296 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2297 EXT4_HAS_INCOMPAT_FEATURE(sb,
2298 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2299
2300 /*
2301 * The first inode we look at is the journal inode. Don't try
2302 * root first: it may be modified in the journal!
2303 */
2304 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2305 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2306 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2307 goto failed_mount3;
624080ed
TT		 2308 if (!(sb->s_flags & MS_RDONLY) &&
2309 EXT4_SB(sb)->s_journal->j_failed_commit) {
2310 printk(KERN_CRIT "EXT4-fs error (device %s): "
2311 "ext4_fill_super: Journal transaction "
2b2d6d01 2312 "%u is corrupt\n", sb->s_id,
624080ed 2313 EXT4_SB(sb)->s_journal->j_failed_commit);
2b2d6d01
TT		 2314 if (test_opt(sb, ERRORS_RO)) {
2315 printk(KERN_CRIT
2316 "Mounting filesystem read-only\n");
624080ed
TT		 2317 sb->s_flags |= MS_RDONLY;
2318 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2319 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2320 }
2321 if (test_opt(sb, ERRORS_PANIC)) {
2322 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2323 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2324 ext4_commit_super(sb, es, 1);
2325 printk(KERN_CRIT
2326 "EXT4-fs (device %s): mount failed\n",
2327 sb->s_id);
2328 goto failed_mount4;
2329 }
2330 }
ac27a0ec 2331 } else if (journal_inum) {
617ba13b 2332 if (ext4_create_journal(sb, es, journal_inum))
ac27a0ec
DK		 2333 goto failed_mount3;
2334 } else {
2335 if (!silent)
2b2d6d01
TT		 2336 printk(KERN_ERR
2337 "ext4: No journal on filesystem on %s\n",
2338 sb->s_id);
ac27a0ec
DK		 2339 goto failed_mount3;
2340 }
2341
eb40a09c
JS		 2342 if (ext4_blocks_count(es) > 0xffffffffULL &&
2343 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2344 JBD2_FEATURE_INCOMPAT_64BIT)) {
2345 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2346 goto failed_mount4;
2347 }
2348
818d276c
GS		 2349 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2350 jbd2_journal_set_features(sbi->s_journal,
2351 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2352 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2353 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2354 jbd2_journal_set_features(sbi->s_journal,
2355 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2356 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2357 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2358 } else {
2359 jbd2_journal_clear_features(sbi->s_journal,
2360 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2361 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2362 }
2363
ac27a0ec
DK		 2364 /* We have now updated the journal if required, so we can
2365 * validate the data journaling mode. */
2366 switch (test_opt(sb, DATA_FLAGS)) {
2367 case 0:
2368 /* No mode set, assume a default based on the journal
63f57933
AM		 2369 * capabilities: ORDERED_DATA if the journal can
2370 * cope, else JOURNAL_DATA
2371 */
dab291af
MC		 2372 if (jbd2_journal_check_available_features
2373 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2374 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2375 else
2376 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2377 break;
2378
617ba13b
MC		 2379 case EXT4_MOUNT_ORDERED_DATA:
2380 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2381 if (!jbd2_journal_check_available_features
2382 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2383 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2384 "requested data journaling mode\n");
2385 goto failed_mount4;
2386 }
2387 default:
2388 break;
2389 }
2390
2391 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2392 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2393 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2394 "its supported only with writeback mode\n");
2395 clear_opt(sbi->s_mount_opt, NOBH);
2396 }
2397 }
2398 /*
dab291af 2399 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2400 * so we can safely mount the rest of the filesystem now.
2401 */
2402
1d1fe1ee
DH		 2403 root = ext4_iget(sb, EXT4_ROOT_INO);
2404 if (IS_ERR(root)) {
617ba13b 2405 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1d1fe1ee 2406 ret = PTR_ERR(root);
ac27a0ec
DK		 2407 goto failed_mount4;
2408 }
2409 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2410 iput(root);
617ba13b 2411 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2412 goto failed_mount4;
2413 }
1d1fe1ee
DH		 2414 sb->s_root = d_alloc_root(root);
2415 if (!sb->s_root) {
2416 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2417 iput(root);
2418 ret = -ENOMEM;
2419 goto failed_mount4;
2420 }
ac27a0ec 2421
2b2d6d01 2422 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2423
2424 /* determine the minimum size of new large inodes, if present */
2425 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2426 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2427 EXT4_GOOD_OLD_INODE_SIZE;
2428 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2429 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2430 if (sbi->s_want_extra_isize <
2431 le16_to_cpu(es->s_want_extra_isize))
2432 sbi->s_want_extra_isize =
2433 le16_to_cpu(es->s_want_extra_isize);
2434 if (sbi->s_want_extra_isize <
2435 le16_to_cpu(es->s_min_extra_isize))
2436 sbi->s_want_extra_isize =
2437 le16_to_cpu(es->s_min_extra_isize);
2438 }
2439 }
2440 /* Check if enough inode space is available */
2441 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2442 sbi->s_inode_size) {
2443 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2444 EXT4_GOOD_OLD_INODE_SIZE;
2445 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2446 "available.\n");
2447 }
2448
ac27a0ec
DK		 2449 /*
2450 * akpm: core read_super() calls in here with the superblock locked.
2451 * That deadlocks, because orphan cleanup needs to lock the superblock
2452 * in numerous places. Here we just pop the lock - it's relatively
2453 * harmless, because we are now ready to accept write_super() requests,
2454 * and aviro says that's the only reason for hanging onto the
2455 * superblock lock.
2456 */
617ba13b
MC		 2457 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2458 ext4_orphan_cleanup(sb, es);
2459 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
ac27a0ec 2460 if (needs_recovery)
2b2d6d01 2461 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
617ba13b 2462 ext4_mark_recovery_complete(sb, es);
2b2d6d01
TT		 2463 printk(KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2464 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2465 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2466 "writeback");
ac27a0ec 2467
dd919b98
AK		 2468 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2469 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2470 "requested data journaling mode\n");
2471 clear_opt(sbi->s_mount_opt, DELALLOC);
2472 } else if (test_opt(sb, DELALLOC))
2473 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2474
a86c6181 2475 ext4_ext_init(sb);
c9de560d 2476 ext4_mb_init(sb, needs_recovery);
a86c6181 2477
ac27a0ec
DK		 2478 lock_kernel();
2479 return 0;
2480
617ba13b 2481cantfind_ext4:
ac27a0ec 2482 if (!silent)
617ba13b 2483 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2484 sb->s_id);
2485 goto failed_mount;
2486
2487failed_mount4:
dab291af 2488 jbd2_journal_destroy(sbi->s_journal);
47b4a50b 2489 sbi->s_journal = NULL;
ac27a0ec
DK		 2490failed_mount3:
2491 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2492 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2493 percpu_counter_destroy(&sbi->s_dirs_counter);
2494failed_mount2:
2495 for (i = 0; i < db_count; i++)
2496 brelse(sbi->s_group_desc[i]);
2497 kfree(sbi->s_group_desc);
2498failed_mount:
2499#ifdef CONFIG_QUOTA
2500 for (i = 0; i < MAXQUOTAS; i++)
2501 kfree(sbi->s_qf_names[i]);
2502#endif
617ba13b 2503 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2504 brelse(bh);
2505out_fail:
2506 sb->s_fs_info = NULL;
2507 kfree(sbi);
2508 lock_kernel();
1d1fe1ee 2509 return ret;
ac27a0ec
DK		 2510}
2511
2512/*
2513 * Setup any per-fs journal parameters now. We'll do this both on
2514 * initial mount, once the journal has been initialised but before we've
2515 * done any recovery; and again on any subsequent remount.
2516 */
617ba13b 2517static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2518{
617ba13b 2519 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2520
2521 if (sbi->s_commit_interval)
2522 journal->j_commit_interval = sbi->s_commit_interval;
617ba13b 2523 /* We could also set up an ext4-specific default for the commit
ac27a0ec
DK		 2524 * interval here, but for now we'll just fall back to the jbd
2525 * default. */
2526
2527 spin_lock(&journal->j_state_lock);
2528 if (test_opt(sb, BARRIER))
dab291af 2529 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2530 else
dab291af 2531 journal->j_flags &= ~JBD2_BARRIER;
ac27a0ec
DK		 2532 spin_unlock(&journal->j_state_lock);
2533}
2534
617ba13b 2535static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2536 unsigned int journal_inum)
2537{
2538 struct inode *journal_inode;
2539 journal_t *journal;
2540
2541 /* First, test for the existence of a valid inode on disk. Bad
2542 * things happen if we iget() an unused inode, as the subsequent
2543 * iput() will try to delete it. */
2544
1d1fe1ee
DH		 2545 journal_inode = ext4_iget(sb, journal_inum);
2546 if (IS_ERR(journal_inode)) {
617ba13b 2547 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2548 return NULL;
2549 }
2550 if (!journal_inode->i_nlink) {
2551 make_bad_inode(journal_inode);
2552 iput(journal_inode);
617ba13b 2553 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2554 return NULL;
2555 }
2556
e5f8eab8 2557 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 2558 journal_inode, journal_inode->i_size);
1d1fe1ee 2559 if (!S_ISREG(journal_inode->i_mode)) {
617ba13b 2560 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2561 iput(journal_inode);
2562 return NULL;
2563 }
2564
dab291af 2565 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2566 if (!journal) {
617ba13b 2567 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2568 iput(journal_inode);
2569 return NULL;
2570 }
2571 journal->j_private = sb;
617ba13b 2572 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2573 return journal;
2574}
2575
617ba13b 2576static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2577 dev_t j_dev)
2578{
2b2d6d01 2579 struct buffer_head *bh;
ac27a0ec 2580 journal_t *journal;
617ba13b
MC		 2581 ext4_fsblk_t start;
2582 ext4_fsblk_t len;
ac27a0ec 2583 int hblock, blocksize;
617ba13b 2584 ext4_fsblk_t sb_block;
ac27a0ec 2585 unsigned long offset;
2b2d6d01 2586 struct ext4_super_block *es;
ac27a0ec
DK		 2587 struct block_device *bdev;
2588
617ba13b 2589 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2590 if (bdev == NULL)
2591 return NULL;
2592
2593 if (bd_claim(bdev, sb)) {
2594 printk(KERN_ERR
8c55e204 2595 "EXT4: failed to claim external journal device.\n");
ac27a0ec
DK		 2596 blkdev_put(bdev);
2597 return NULL;
2598 }
2599
2600 blocksize = sb->s_blocksize;
2601 hblock = bdev_hardsect_size(bdev);
2602 if (blocksize < hblock) {
2603 printk(KERN_ERR
617ba13b 2604 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2605 goto out_bdev;
2606 }
2607
617ba13b
MC		 2608 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2609 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2610 set_blocksize(bdev, blocksize);
2611 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2612 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2613 "external journal\n");
2614 goto out_bdev;
2615 }
2616
617ba13b
MC		 2617 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2618 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2619 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2620 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2621 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2622 "bad superblock\n");
2623 brelse(bh);
2624 goto out_bdev;
2625 }
2626
617ba13b
MC		 2627 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2628 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2629 brelse(bh);
2630 goto out_bdev;
2631 }
2632
bd81d8ee 2633 len = ext4_blocks_count(es);
ac27a0ec
DK		 2634 start = sb_block + 1;
2635 brelse(bh); /* we're done with the superblock */
2636
dab291af 2637 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2638 start, len, blocksize);
2639 if (!journal) {
617ba13b 2640 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2641 goto out_bdev;
2642 }
2643 journal->j_private = sb;
2644 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2645 wait_on_buffer(journal->j_sb_buffer);
2646 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2647 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2648 goto out_journal;
2649 }
2650 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2651 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2652 "user (unsupported) - %d\n",
2653 be32_to_cpu(journal->j_superblock->s_nr_users));
2654 goto out_journal;
2655 }
617ba13b
MC		 2656 EXT4_SB(sb)->journal_bdev = bdev;
2657 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2658 return journal;
2659out_journal:
dab291af 2660 jbd2_journal_destroy(journal);
ac27a0ec 2661out_bdev:
617ba13b 2662 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2663 return NULL;
2664}
2665
617ba13b
MC		 2666static int ext4_load_journal(struct super_block *sb,
2667 struct ext4_super_block *es,
ac27a0ec
DK		 2668 unsigned long journal_devnum)
2669{
2670 journal_t *journal;
2671 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2672 dev_t journal_dev;
2673 int err = 0;
2674 int really_read_only;
2675
2676 if (journal_devnum &&
2677 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2678 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2679 "numbers have changed\n");
2680 journal_dev = new_decode_dev(journal_devnum);
2681 } else
2682 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2683
2684 really_read_only = bdev_read_only(sb->s_bdev);
2685
2686 /*
2687 * Are we loading a blank journal or performing recovery after a
2688 * crash? For recovery, we need to check in advance whether we
2689 * can get read-write access to the device.
2690 */
2691
617ba13b 2692 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2693 if (sb->s_flags & MS_RDONLY) {
617ba13b 2694 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2695 "required on readonly filesystem.\n");
2696 if (really_read_only) {
617ba13b 2697 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2698 "unavailable, cannot proceed.\n");
2699 return -EROFS;
2700 }
2b2d6d01
TT		 2701 printk(KERN_INFO "EXT4-fs: write access will "
2702 "be enabled during recovery.\n");
ac27a0ec
DK		 2703 }
2704 }
2705
2706 if (journal_inum && journal_dev) {
617ba13b 2707 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2708 "and inode journals!\n");
2709 return -EINVAL;
2710 }
2711
2712 if (journal_inum) {
617ba13b 2713 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2714 return -EINVAL;
2715 } else {
617ba13b 2716 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2717 return -EINVAL;
2718 }
2719
4776004f
TT		 2720 if (journal->j_flags & JBD2_BARRIER)
2721 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2722 else
2723 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2724
ac27a0ec 2725 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2726 err = jbd2_journal_update_format(journal);
ac27a0ec 2727 if (err) {
617ba13b 2728 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2729 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2730 return err;
2731 }
2732 }
2733
617ba13b 2734 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2735 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2736 if (!err)
dab291af 2737 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2738
2739 if (err) {
617ba13b 2740 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2741 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2742 return err;
2743 }
2744
617ba13b
MC		 2745 EXT4_SB(sb)->s_journal = journal;
2746 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2747
2748 if (journal_devnum &&
2749 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2750 es->s_journal_dev = cpu_to_le32(journal_devnum);
2751 sb->s_dirt = 1;
2752
2753 /* Make sure we flush the recovery flag to disk. */
617ba13b 2754 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2755 }
2756
2757 return 0;
2758}
2759
2b2d6d01
TT		 2760static int ext4_create_journal(struct super_block *sb,
2761 struct ext4_super_block *es,
ac27a0ec
DK		 2762 unsigned int journal_inum)
2763{
2764 journal_t *journal;
6c675bd4 2765 int err;
ac27a0ec
DK		 2766
2767 if (sb->s_flags & MS_RDONLY) {
617ba13b 2768 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
ac27a0ec
DK		 2769 "create journal.\n");
2770 return -EROFS;
2771 }
2772
6c675bd4
BP		 2773 journal = ext4_get_journal(sb, journal_inum);
2774 if (!journal)
ac27a0ec
DK		 2775 return -EINVAL;
2776
617ba13b 2777 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
ac27a0ec
DK		 2778 journal_inum);
2779
6c675bd4
BP		 2780 err = jbd2_journal_create(journal);
2781 if (err) {
617ba13b 2782 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
dab291af 2783 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2784 return -EIO;
2785 }
2786
617ba13b 2787 EXT4_SB(sb)->s_journal = journal;
ac27a0ec 2788
617ba13b
MC		 2789 ext4_update_dynamic_rev(sb);
2790 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2791 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
ac27a0ec
DK		 2792
2793 es->s_journal_inum = cpu_to_le32(journal_inum);
2794 sb->s_dirt = 1;
2795
2796 /* Make sure we flush the recovery flag to disk. */
617ba13b 2797 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2798
2799 return 0;
2800}
2801
2b2d6d01
TT		 2802static void ext4_commit_super(struct super_block *sb,
2803 struct ext4_super_block *es, int sync)
ac27a0ec 2804{
617ba13b 2805 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
ac27a0ec
DK		 2806
2807 if (!sbh)
2808 return;
2809 es->s_wtime = cpu_to_le32(get_seconds());
bd81d8ee 2810 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
617ba13b 2811 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2812 BUFFER_TRACE(sbh, "marking dirty");
2813 mark_buffer_dirty(sbh);
2814 if (sync)
2815 sync_dirty_buffer(sbh);
2816}
2817
2818
2819/*
2820 * Have we just finished recovery? If so, and if we are mounting (or
2821 * remounting) the filesystem readonly, then we will end up with a
2822 * consistent fs on disk. Record that fact.
2823 */
2b2d6d01
TT		 2824static void ext4_mark_recovery_complete(struct super_block *sb,
2825 struct ext4_super_block *es)
ac27a0ec 2826{
617ba13b 2827 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2828
dab291af
MC		 2829 jbd2_journal_lock_updates(journal);
2830 jbd2_journal_flush(journal);
32c37730 2831 lock_super(sb);
617ba13b 2832 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2833 sb->s_flags & MS_RDONLY) {
617ba13b 2834 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2835 sb->s_dirt = 0;
617ba13b 2836 ext4_commit_super(sb, es, 1);
ac27a0ec 2837 }
32c37730 2838 unlock_super(sb);
dab291af 2839 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2840}
2841
2842/*
2843 * If we are mounting (or read-write remounting) a filesystem whose journal
2844 * has recorded an error from a previous lifetime, move that error to the
2845 * main filesystem now.
2846 */
2b2d6d01
TT		 2847static void ext4_clear_journal_err(struct super_block *sb,
2848 struct ext4_super_block *es)
ac27a0ec
DK		 2849{
2850 journal_t *journal;
2851 int j_errno;
2852 const char *errstr;
2853
617ba13b 2854 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2855
2856 /*
2857 * Now check for any error status which may have been recorded in the
617ba13b 2858 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2859 */
2860
dab291af 2861 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2862 if (j_errno) {
2863 char nbuf[16];
2864
617ba13b 2865 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 2866 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 2867 "from previous mount: %s", errstr);
46e665e9 2868 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 2869 "filesystem check.");
2870
617ba13b
MC		 2871 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2872 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2b2d6d01 2873 ext4_commit_super(sb, es, 1);
ac27a0ec 2874
dab291af 2875 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 2876 }
2877}
2878
2879/*
2880 * Force the running and committing transactions to commit,
2881 * and wait on the commit.
2882 */
617ba13b 2883int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 2884{
2885 journal_t *journal;
2886 int ret;
2887
2888 if (sb->s_flags & MS_RDONLY)
2889 return 0;
2890
617ba13b 2891 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2892 sb->s_dirt = 0;
617ba13b 2893 ret = ext4_journal_force_commit(journal);
ac27a0ec
DK		 2894 return ret;
2895}
2896
2897/*
617ba13b 2898 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec
DK		 2899 * have to propagate any other updates to the superblock on disk at this
2900 * point. Just start an async writeback to get the buffers on their way
2901 * to the disk.
2902 *
2903 * This implicitly triggers the writebehind on sync().
2904 */
2905
2b2d6d01 2906static void ext4_write_super(struct super_block *sb)
ac27a0ec
DK		 2907{
2908 if (mutex_trylock(&sb->s_lock) != 0)
2909 BUG();
2910 sb->s_dirt = 0;
2911}
2912
617ba13b 2913static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec
DK		 2914{
2915 tid_t target;
2916
2917 sb->s_dirt = 0;
dab291af 2918 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
ac27a0ec 2919 if (wait)
dab291af 2920 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
ac27a0ec
DK		 2921 }
2922 return 0;
2923}
2924
2925/*
2926 * LVM calls this function before a (read-only) snapshot is created. This
2927 * gives us a chance to flush the journal completely and mark the fs clean.
2928 */
617ba13b 2929static void ext4_write_super_lockfs(struct super_block *sb)
ac27a0ec
DK		 2930{
2931 sb->s_dirt = 0;
2932
2933 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 2934 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2935
2936 /* Now we set up the journal barrier. */
dab291af
MC		 2937 jbd2_journal_lock_updates(journal);
2938 jbd2_journal_flush(journal);
ac27a0ec
DK		 2939
2940 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 2941 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2942 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec
DK		 2943 }
2944}
2945
2946/*
2947 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2948 * flag here, even though the filesystem is not technically dirty yet.
2949 */
617ba13b 2950static void ext4_unlockfs(struct super_block *sb)
ac27a0ec
DK		 2951{
2952 if (!(sb->s_flags & MS_RDONLY)) {
2953 lock_super(sb);
2954 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 2955 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2956 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 2957 unlock_super(sb);
dab291af 2958 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 2959 }
2960}
2961
2b2d6d01 2962static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 2963{
2b2d6d01 2964 struct ext4_super_block *es;
617ba13b
MC		 2965 struct ext4_sb_info *sbi = EXT4_SB(sb);
2966 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 2967 unsigned long old_sb_flags;
617ba13b 2968 struct ext4_mount_options old_opts;
8a266467 2969 ext4_group_t g;
ac27a0ec
DK		 2970 int err;
2971#ifdef CONFIG_QUOTA
2972 int i;
2973#endif
2974
2975 /* Store the original options */
2976 old_sb_flags = sb->s_flags;
2977 old_opts.s_mount_opt = sbi->s_mount_opt;
2978 old_opts.s_resuid = sbi->s_resuid;
2979 old_opts.s_resgid = sbi->s_resgid;
2980 old_opts.s_commit_interval = sbi->s_commit_interval;
2981#ifdef CONFIG_QUOTA
2982 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2983 for (i = 0; i < MAXQUOTAS; i++)
2984 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2985#endif
2986
2987 /*
2988 * Allow the "check" option to be passed as a remount option.
2989 */
2990 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2991 err = -EINVAL;
2992 goto restore_opts;
2993 }
2994
617ba13b 2995 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
46e665e9 2996 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 2997
2998 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2999 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3000
3001 es = sbi->s_es;
3002
617ba13b 3003 ext4_init_journal_params(sb, sbi->s_journal);
ac27a0ec
DK		 3004
3005 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3006 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 3007 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 3008 err = -EROFS;
3009 goto restore_opts;
3010 }
3011
3012 if (*flags & MS_RDONLY) {
3013 /*
3014 * First of all, the unconditional stuff we have to do
3015 * to disable replay of the journal when we next remount
3016 */
3017 sb->s_flags |= MS_RDONLY;
3018
3019 /*
3020 * OK, test if we are remounting a valid rw partition
3021 * readonly, and if so set the rdonly flag and then
3022 * mark the partition as valid again.
3023 */
617ba13b
MC		 3024 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3025 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3026 es->s_state = cpu_to_le16(sbi->s_mount_state);
3027
32c37730
JK		 3028 /*
3029 * We have to unlock super so that we can wait for
3030 * transactions.
3031 */
3032 unlock_super(sb);
617ba13b 3033 ext4_mark_recovery_complete(sb, es);
32c37730 3034 lock_super(sb);
ac27a0ec
DK		 3035 } else {
3036 __le32 ret;
617ba13b
MC		 3037 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3038 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3039 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec
DK		 3040 "remount RDWR because of unsupported "
3041 "optional features (%x).\n",
3042 sb->s_id, le32_to_cpu(ret));
3043 err = -EROFS;
3044 goto restore_opts;
3045 }
ead6596b 3046
8a266467
TT		 3047 /*
3048 * Make sure the group descriptor checksums
3049 * are sane. If they aren't, refuse to
3050 * remount r/w.
3051 */
3052 for (g = 0; g < sbi->s_groups_count; g++) {
3053 struct ext4_group_desc *gdp =
3054 ext4_get_group_desc(sb, g, NULL);
3055
3056 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3057 printk(KERN_ERR
3058 "EXT4-fs: ext4_remount: "
3059 "Checksum for group %lu failed (%u!=%u)\n",
3060 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3061 le16_to_cpu(gdp->bg_checksum));
3062 err = -EINVAL;
3063 goto restore_opts;
3064 }
3065 }
3066
ead6596b
ES		 3067 /*
3068 * If we have an unprocessed orphan list hanging
3069 * around from a previously readonly bdev mount,
3070 * require a full umount/remount for now.
3071 */
3072 if (es->s_last_orphan) {
3073 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3074 "remount RDWR because of unprocessed "
3075 "orphan inode list. Please "
3076 "umount/remount instead.\n",
3077 sb->s_id);
3078 err = -EINVAL;
3079 goto restore_opts;
3080 }
3081
ac27a0ec
DK		 3082 /*
3083 * Mounting a RDONLY partition read-write, so reread
3084 * and store the current valid flag. (It may have
3085 * been changed by e2fsck since we originally mounted
3086 * the partition.)
3087 */
617ba13b 3088 ext4_clear_journal_err(sb, es);
ac27a0ec 3089 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3090 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3091 goto restore_opts;
2b2d6d01 3092 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3093 sb->s_flags &= ~MS_RDONLY;
3094 }
3095 }
3096#ifdef CONFIG_QUOTA
3097 /* Release old quota file names */
3098 for (i = 0; i < MAXQUOTAS; i++)
3099 if (old_opts.s_qf_names[i] &&
3100 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3101 kfree(old_opts.s_qf_names[i]);
3102#endif
3103 return 0;
3104restore_opts:
3105 sb->s_flags = old_sb_flags;
3106 sbi->s_mount_opt = old_opts.s_mount_opt;
3107 sbi->s_resuid = old_opts.s_resuid;
3108 sbi->s_resgid = old_opts.s_resgid;
3109 sbi->s_commit_interval = old_opts.s_commit_interval;
3110#ifdef CONFIG_QUOTA
3111 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3112 for (i = 0; i < MAXQUOTAS; i++) {
3113 if (sbi->s_qf_names[i] &&
3114 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3115 kfree(sbi->s_qf_names[i]);
3116 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3117 }
3118#endif
3119 return err;
3120}
3121
2b2d6d01 3122static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3123{
3124 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3125 struct ext4_sb_info *sbi = EXT4_SB(sb);
3126 struct ext4_super_block *es = sbi->s_es;
960cc398 3127 u64 fsid;
ac27a0ec 3128
5e70030d
BP		 3129 if (test_opt(sb, MINIX_DF)) {
3130 sbi->s_overhead_last = 0;
6bc9feff 3131 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 3132 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 3133 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3134 smp_rmb();
3135
3136 /*
5e70030d
BP		 3137 * Compute the overhead (FS structures). This is constant
3138 * for a given filesystem unless the number of block groups
3139 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3140 */
3141
3142 /*
3143 * All of the blocks before first_data_block are
3144 * overhead
3145 */
3146 overhead = le32_to_cpu(es->s_first_data_block);
3147
3148 /*
3149 * Add the overhead attributed to the superblock and
3150 * block group descriptors. If the sparse superblocks
3151 * feature is turned on, then not all groups have this.
3152 */
3153 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3154 overhead += ext4_bg_has_super(sb, i) +
3155 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3156 cond_resched();
3157 }
3158
3159 /*
3160 * Every block group has an inode bitmap, a block
3161 * bitmap, and an inode table.
3162 */
5e70030d
BP		 3163 overhead += ngroups * (2 + sbi->s_itb_per_group);
3164 sbi->s_overhead_last = overhead;
3165 smp_wmb();
6bc9feff 3166 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3167 }
3168
617ba13b 3169 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3170 buf->f_bsize = sb->s_blocksize;
5e70030d 3171 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
52d9f3b4 3172 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
308ba3ec 3173 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3174 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3175 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3176 buf->f_bavail = 0;
3177 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3178 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3179 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3180 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3181 fsid = le64_to_cpup((void *)es->s_uuid) ^
3182 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3183 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3184 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 3185 return 0;
3186}
3187
3188/* Helper function for writing quotas on sync - we need to start transaction before quota file
3189 * is locked for write. Otherwise the are possible deadlocks:
3190 * Process 1 Process 2
617ba13b 3191 * ext4_create() quota_sync()
dab291af 3192 * jbd2_journal_start() write_dquot()
ac27a0ec 3193 * DQUOT_INIT() down(dqio_mutex)
dab291af 3194 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3195 *
3196 */
3197
3198#ifdef CONFIG_QUOTA
3199
3200static inline struct inode *dquot_to_inode(struct dquot *dquot)
3201{
3202 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3203}
3204
617ba13b 3205static int ext4_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 3206{
3207 handle_t *handle;
3208 int ret, err;
3209
3210 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 3211 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 3212 if (IS_ERR(handle))
3213 return PTR_ERR(handle);
3214 ret = dquot_initialize(inode, type);
617ba13b 3215 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3216 if (!ret)
3217 ret = err;
3218 return ret;
3219}
3220
617ba13b 3221static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 3222{
3223 handle_t *handle;
3224 int ret, err;
3225
3226 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 3227 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2887df13
JK		 3228 if (IS_ERR(handle)) {
3229 /*
3230 * We call dquot_drop() anyway to at least release references
3231 * to quota structures so that umount does not hang.
3232 */
3233 dquot_drop(inode);
ac27a0ec 3234 return PTR_ERR(handle);
2887df13 3235 }
ac27a0ec 3236 ret = dquot_drop(inode);
617ba13b 3237 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3238 if (!ret)
3239 ret = err;
3240 return ret;
3241}
3242
617ba13b 3243static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3244{
3245 int ret, err;
3246 handle_t *handle;
3247 struct inode *inode;
3248
3249 inode = dquot_to_inode(dquot);
617ba13b
MC		 3250 handle = ext4_journal_start(inode,
3251 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3252 if (IS_ERR(handle))
3253 return PTR_ERR(handle);
3254 ret = dquot_commit(dquot);
617ba13b 3255 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3256 if (!ret)
3257 ret = err;
3258 return ret;
3259}
3260
617ba13b 3261static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3262{
3263 int ret, err;
3264 handle_t *handle;
3265
617ba13b
MC		 3266 handle = ext4_journal_start(dquot_to_inode(dquot),
3267 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3268 if (IS_ERR(handle))
3269 return PTR_ERR(handle);
3270 ret = dquot_acquire(dquot);
617ba13b 3271 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3272 if (!ret)
3273 ret = err;
3274 return ret;
3275}
3276
617ba13b 3277static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3278{
3279 int ret, err;
3280 handle_t *handle;
3281
617ba13b
MC		 3282 handle = ext4_journal_start(dquot_to_inode(dquot),
3283 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3284 if (IS_ERR(handle)) {
3285 /* Release dquot anyway to avoid endless cycle in dqput() */
3286 dquot_release(dquot);
ac27a0ec 3287 return PTR_ERR(handle);
9c3013e9 3288 }
ac27a0ec 3289 ret = dquot_release(dquot);
617ba13b 3290 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3291 if (!ret)
3292 ret = err;
3293 return ret;
3294}
3295
617ba13b 3296static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3297{
2c8be6b2 3298 /* Are we journaling quotas? */
617ba13b
MC		 3299 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3300 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3301 dquot_mark_dquot_dirty(dquot);
617ba13b 3302 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3303 } else {
3304 return dquot_mark_dquot_dirty(dquot);
3305 }
3306}
3307
617ba13b 3308static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3309{
3310 int ret, err;
3311 handle_t *handle;
3312
3313 /* Data block + inode block */
617ba13b 3314 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3315 if (IS_ERR(handle))
3316 return PTR_ERR(handle);
3317 ret = dquot_commit_info(sb, type);
617ba13b 3318 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3319 if (!ret)
3320 ret = err;
3321 return ret;
3322}
3323
3324/*
3325 * Turn on quotas during mount time - we need to find
3326 * the quota file and such...
3327 */
617ba13b 3328static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3329{
617ba13b
MC		 3330 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3331 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3332}
3333
3334/*
3335 * Standard function to be called on quota_on
3336 */
617ba13b 3337static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6f28e087 3338 char *path, int remount)
ac27a0ec
DK		 3339{
3340 int err;
3341 struct nameidata nd;
3342
3343 if (!test_opt(sb, QUOTA))
3344 return -EINVAL;
0623543b
JK		 3345 /* When remounting, no checks are needed and in fact, path is NULL */
3346 if (remount)
6f28e087 3347 return vfs_quota_on(sb, type, format_id, path, remount);
0623543b 3348
ac27a0ec
DK		 3349 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3350 if (err)
3351 return err;
0623543b 3352
ac27a0ec 3353 /* Quotafile not on the same filesystem? */
4ac91378 3354 if (nd.path.mnt->mnt_sb != sb) {
1d957f9b 3355 path_put(&nd.path);
ac27a0ec
DK		 3356 return -EXDEV;
3357 }
0623543b
JK		 3358 /* Journaling quota? */
3359 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3360 /* Quotafile not in fs root? */
0623543b
JK		 3361 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3362 printk(KERN_WARNING
3363 "EXT4-fs: Quota file not on filesystem root. "
3364 "Journaled quota will not work.\n");
2b2d6d01 3365 }
0623543b
JK		 3366
3367 /*
3368 * When we journal data on quota file, we have to flush journal to see
3369 * all updates to the file when we bypass pagecache...
3370 */
3371 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3372 /*
3373 * We don't need to lock updates but journal_flush() could
3374 * otherwise be livelocked...
3375 */
3376 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3377 jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3378 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3379 }
3380
77e69dac 3381 err = vfs_quota_on_path(sb, type, format_id, &nd.path);
1d957f9b 3382 path_put(&nd.path);
77e69dac 3383 return err;
ac27a0ec
DK		 3384}
3385
3386/* Read data from quotafile - avoid pagecache and such because we cannot afford
3387 * acquiring the locks... As quota files are never truncated and quota code
3388 * itself serializes the operations (and noone else should touch the files)
3389 * we don't have to be afraid of races */
617ba13b 3390static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3391 size_t len, loff_t off)
3392{
3393 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3394 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3395 int err = 0;
3396 int offset = off & (sb->s_blocksize - 1);
3397 int tocopy;
3398 size_t toread;
3399 struct buffer_head *bh;
3400 loff_t i_size = i_size_read(inode);
3401
3402 if (off > i_size)
3403 return 0;
3404 if (off+len > i_size)
3405 len = i_size-off;
3406 toread = len;
3407 while (toread > 0) {
3408 tocopy = sb->s_blocksize - offset < toread ?
3409 sb->s_blocksize - offset : toread;
617ba13b 3410 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3411 if (err)
3412 return err;
3413 if (!bh) /* A hole? */
3414 memset(data, 0, tocopy);
3415 else
3416 memcpy(data, bh->b_data+offset, tocopy);
3417 brelse(bh);
3418 offset = 0;
3419 toread -= tocopy;
3420 data += tocopy;
3421 blk++;
3422 }
3423 return len;
3424}
3425
3426/* Write to quotafile (we know the transaction is already started and has
3427 * enough credits) */
617ba13b 3428static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3429 const char *data, size_t len, loff_t off)
3430{
3431 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3432 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3433 int err = 0;
3434 int offset = off & (sb->s_blocksize - 1);
3435 int tocopy;
617ba13b 3436 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3437 size_t towrite = len;
3438 struct buffer_head *bh;
3439 handle_t *handle = journal_current_handle();
3440
9c3013e9 3441 if (!handle) {
e5f8eab8 3442 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
9c3013e9
JK		 3443 " cancelled because transaction is not started.\n",
3444 (unsigned long long)off, (unsigned long long)len);
3445 return -EIO;
3446 }
ac27a0ec
DK		 3447 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3448 while (towrite > 0) {
3449 tocopy = sb->s_blocksize - offset < towrite ?
3450 sb->s_blocksize - offset : towrite;
617ba13b 3451 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3452 if (!bh)
3453 goto out;
3454 if (journal_quota) {
617ba13b 3455 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3456 if (err) {
3457 brelse(bh);
3458 goto out;
3459 }
3460 }
3461 lock_buffer(bh);
3462 memcpy(bh->b_data+offset, data, tocopy);
3463 flush_dcache_page(bh->b_page);
3464 unlock_buffer(bh);
3465 if (journal_quota)
617ba13b 3466 err = ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK		 3467 else {
3468 /* Always do at least ordered writes for quotas */
678aaf48 3469 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3470 mark_buffer_dirty(bh);
3471 }
3472 brelse(bh);
3473 if (err)
3474 goto out;
3475 offset = 0;
3476 towrite -= tocopy;
3477 data += tocopy;
3478 blk++;
3479 }
3480out:
4d04e4fb
JK		 3481 if (len == towrite) {
3482 mutex_unlock(&inode->i_mutex);
ac27a0ec 3483 return err;
4d04e4fb 3484 }
ac27a0ec
DK		 3485 if (inode->i_size < off+len-towrite) {
3486 i_size_write(inode, off+len-towrite);
617ba13b 3487 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3488 }
ac27a0ec 3489 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3490 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3491 mutex_unlock(&inode->i_mutex);
3492 return len - towrite;
3493}
3494
3495#endif
3496
617ba13b 3497static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3498 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3499{
617ba13b 3500 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3501}
3502
617ba13b 3503static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3504 .owner = THIS_MODULE,
617ba13b
MC		 3505 .name = "ext4dev",
3506 .get_sb = ext4_get_sb,
ac27a0ec
DK		 3507 .kill_sb = kill_block_super,
3508 .fs_flags = FS_REQUIRES_DEV,
3509};
3510
617ba13b 3511static int __init init_ext4_fs(void)
ac27a0ec 3512{
c9de560d
AT		 3513 int err;
3514
3515 err = init_ext4_mballoc();
ac27a0ec
DK		 3516 if (err)
3517 return err;
c9de560d
AT		 3518
3519 err = init_ext4_xattr();
3520 if (err)
3521 goto out2;
ac27a0ec
DK		 3522 err = init_inodecache();
3523 if (err)
3524 goto out1;
63f57933 3525 err = register_filesystem(&ext4dev_fs_type);
ac27a0ec
DK		 3526 if (err)
3527 goto out;
3528 return 0;
3529out:
3530 destroy_inodecache();
3531out1:
617ba13b 3532 exit_ext4_xattr();
c9de560d
AT		 3533out2:
3534 exit_ext4_mballoc();
ac27a0ec
DK		 3535 return err;
3536}
3537
617ba13b 3538static void __exit exit_ext4_fs(void)
ac27a0ec 3539{
617ba13b 3540 unregister_filesystem(&ext4dev_fs_type);
ac27a0ec 3541 destroy_inodecache();
617ba13b 3542 exit_ext4_xattr();
c9de560d 3543 exit_ext4_mballoc();
ac27a0ec
DK		 3544}
3545
3546MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
617ba13b 3547MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
ac27a0ec 3548MODULE_LICENSE("GPL");
617ba13b
MC		 3549module_init(init_ext4_fs)
3550module_exit(exit_ext4_fs)
Clone of davem's net-next-2.6 tree