]> 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: delayed allocation i_blocks fix for stat
[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);
617ba13b
MC		 52static void ext4_commit_super (struct super_block * sb,
53 struct ext4_super_block * es,
ac27a0ec 54 int sync);
617ba13b
MC		 55static void ext4_mark_recovery_complete(struct super_block * sb,
56 struct ext4_super_block * es);
57static void ext4_clear_journal_err(struct super_block * sb,
58 struct ext4_super_block * es);
59static int ext4_sync_fs(struct super_block *sb, int wait);
60static const char *ext4_decode_error(struct super_block * sb, int errno,
ac27a0ec 61 char nbuf[16]);
617ba13b
MC		 62static int ext4_remount (struct super_block * sb, int * flags, char * data);
63static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
64static void ext4_unlockfs(struct super_block *sb);
65static void ext4_write_super (struct super_block * sb);
66static void ext4_write_super_lockfs(struct super_block *sb);
ac27a0ec 67
bd81d8ee 68
8fadc143
AR		 69ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
70 struct ext4_group_desc *bg)
bd81d8ee 71{
3a14589c 72 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 73 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
3a14589c 74 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 75}
76
8fadc143
AR		 77ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
78 struct ext4_group_desc *bg)
bd81d8ee 79{
5272f837 80 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 81 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 82 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 83}
84
8fadc143
AR		 85ext4_fsblk_t ext4_inode_table(struct super_block *sb,
86 struct ext4_group_desc *bg)
bd81d8ee 87{
5272f837 88 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 89 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 90 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 91}
92
8fadc143
AR		 93void ext4_block_bitmap_set(struct super_block *sb,
94 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 95{
3a14589c 96 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 97 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
98 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 99}
100
8fadc143
AR		 101void ext4_inode_bitmap_set(struct super_block *sb,
102 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 103{
5272f837 104 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 105 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
106 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 107}
108
8fadc143
AR		 109void ext4_inode_table_set(struct super_block *sb,
110 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 111{
5272f837 112 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 113 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
114 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 115}
116
ac27a0ec 117/*
dab291af 118 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 119 *
120 * The only special thing we need to do here is to make sure that all
121 * journal_end calls result in the superblock being marked dirty, so
122 * that sync() will call the filesystem's write_super callback if
123 * appropriate.
124 */
617ba13b 125handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 126{
127 journal_t *journal;
128
129 if (sb->s_flags & MS_RDONLY)
130 return ERR_PTR(-EROFS);
131
132 /* Special case here: if the journal has aborted behind our
133 * backs (eg. EIO in the commit thread), then we still need to
134 * take the FS itself readonly cleanly. */
617ba13b 135 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 136 if (is_journal_aborted(journal)) {
46e665e9 137 ext4_abort(sb, __func__,
ac27a0ec
DK		 138 "Detected aborted journal");
139 return ERR_PTR(-EROFS);
140 }
141
dab291af 142 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 143}
144
145/*
146 * The only special thing we need to do here is to make sure that all
dab291af 147 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 148 * that sync() will call the filesystem's write_super callback if
149 * appropriate.
150 */
617ba13b 151int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 152{
153 struct super_block *sb;
154 int err;
155 int rc;
156
157 sb = handle->h_transaction->t_journal->j_private;
158 err = handle->h_err;
dab291af 159 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 160
161 if (!err)
162 err = rc;
163 if (err)
617ba13b 164 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 165 return err;
166}
167
617ba13b 168void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 169 struct buffer_head *bh, handle_t *handle, int err)
170{
171 char nbuf[16];
617ba13b 172 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec
DK		 173
174 if (bh)
175 BUFFER_TRACE(bh, "abort");
176
177 if (!handle->h_err)
178 handle->h_err = err;
179
180 if (is_handle_aborted(handle))
181 return;
182
183 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
184 caller, errstr, err_fn);
185
dab291af 186 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 187}
188
189/* Deal with the reporting of failure conditions on a filesystem such as
190 * inconsistencies detected or read IO failures.
191 *
192 * On ext2, we can store the error state of the filesystem in the
617ba13b 193 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 194 * write ordering constraints on the superblock which prevent us from
195 * writing it out straight away; and given that the journal is about to
196 * be aborted, we can't rely on the current, or future, transactions to
197 * write out the superblock safely.
198 *
dab291af 199 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 200 * the journal instead. On recovery, the journal will compain about
201 * that error until we've noted it down and cleared it.
202 */
203
617ba13b 204static void ext4_handle_error(struct super_block *sb)
ac27a0ec 205{
617ba13b 206 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 207
617ba13b
MC		 208 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
209 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 210
211 if (sb->s_flags & MS_RDONLY)
212 return;
213
214 if (!test_opt (sb, ERRORS_CONT)) {
617ba13b 215 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 216
617ba13b 217 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ac27a0ec 218 if (journal)
dab291af 219 jbd2_journal_abort(journal, -EIO);
ac27a0ec
DK		 220 }
221 if (test_opt (sb, ERRORS_RO)) {
222 printk (KERN_CRIT "Remounting filesystem read-only\n");
223 sb->s_flags |= MS_RDONLY;
224 }
617ba13b 225 ext4_commit_super(sb, es, 1);
ac27a0ec 226 if (test_opt(sb, ERRORS_PANIC))
617ba13b 227 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 228 sb->s_id);
229}
230
617ba13b 231void ext4_error (struct super_block * sb, const char * function,
ac27a0ec
DK		 232 const char * fmt, ...)
233{
234 va_list args;
235
236 va_start(args, fmt);
617ba13b 237 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
ac27a0ec
DK		 238 vprintk(fmt, args);
239 printk("\n");
240 va_end(args);
241
617ba13b 242 ext4_handle_error(sb);
ac27a0ec
DK		 243}
244
617ba13b 245static const char *ext4_decode_error(struct super_block * sb, int errno,
ac27a0ec
DK		 246 char nbuf[16])
247{
248 char *errstr = NULL;
249
250 switch (errno) {
251 case -EIO:
252 errstr = "IO failure";
253 break;
254 case -ENOMEM:
255 errstr = "Out of memory";
256 break;
257 case -EROFS:
dab291af 258 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
ac27a0ec
DK		 259 errstr = "Journal has aborted";
260 else
261 errstr = "Readonly filesystem";
262 break;
263 default:
264 /* If the caller passed in an extra buffer for unknown
265 * errors, textualise them now. Else we just return
266 * NULL. */
267 if (nbuf) {
268 /* Check for truncated error codes... */
269 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
270 errstr = nbuf;
271 }
272 break;
273 }
274
275 return errstr;
276}
277
617ba13b 278/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 279 * automatically and invokes the appropriate error response. */
280
617ba13b 281void __ext4_std_error (struct super_block * sb, const char * function,
ac27a0ec
DK		 282 int errno)
283{
284 char nbuf[16];
285 const char *errstr;
286
287 /* Special case: if the error is EROFS, and we're not already
288 * inside a transaction, then there's really no point in logging
289 * an error. */
290 if (errno == -EROFS && journal_current_handle() == NULL &&
291 (sb->s_flags & MS_RDONLY))
292 return;
293
617ba13b
MC		 294 errstr = ext4_decode_error(sb, errno, nbuf);
295 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
ac27a0ec
DK		 296 sb->s_id, function, errstr);
297
617ba13b 298 ext4_handle_error(sb);
ac27a0ec
DK		 299}
300
301/*
617ba13b 302 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 303 * abort function may be used to deal with unrecoverable failures such
304 * as journal IO errors or ENOMEM at a critical moment in log management.
305 *
306 * We unconditionally force the filesystem into an ABORT|READONLY state,
307 * unless the error response on the fs has been set to panic in which
308 * case we take the easy way out and panic immediately.
309 */
310
617ba13b 311void ext4_abort (struct super_block * sb, const char * function,
ac27a0ec
DK		 312 const char * fmt, ...)
313{
314 va_list args;
315
617ba13b 316 printk (KERN_CRIT "ext4_abort called.\n");
ac27a0ec
DK		 317
318 va_start(args, fmt);
617ba13b 319 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
ac27a0ec
DK		 320 vprintk(fmt, args);
321 printk("\n");
322 va_end(args);
323
324 if (test_opt(sb, ERRORS_PANIC))
617ba13b 325 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 326
327 if (sb->s_flags & MS_RDONLY)
328 return;
329
330 printk(KERN_CRIT "Remounting filesystem read-only\n");
617ba13b 331 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 332 sb->s_flags |= MS_RDONLY;
617ba13b 333 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
dab291af 334 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 335}
336
617ba13b 337void ext4_warning (struct super_block * sb, const char * function,
ac27a0ec
DK		 338 const char * fmt, ...)
339{
340 va_list args;
341
342 va_start(args, fmt);
617ba13b 343 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 344 sb->s_id, function);
345 vprintk(fmt, args);
346 printk("\n");
347 va_end(args);
348}
349
617ba13b 350void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 351{
617ba13b 352 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 353
617ba13b 354 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 355 return;
356
46e665e9 357 ext4_warning(sb, __func__,
ac27a0ec
DK		 358 "updating to rev %d because of new feature flag, "
359 "running e2fsck is recommended",
617ba13b 360 EXT4_DYNAMIC_REV);
ac27a0ec 361
617ba13b
MC		 362 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
363 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
364 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 365 /* leave es->s_feature_*compat flags alone */
366 /* es->s_uuid will be set by e2fsck if empty */
367
368 /*
369 * The rest of the superblock fields should be zero, and if not it
370 * means they are likely already in use, so leave them alone. We
371 * can leave it up to e2fsck to clean up any inconsistencies there.
372 */
373}
374
99e6f829
AK		 375int ext4_update_compat_feature(handle_t *handle,
376 struct super_block *sb, __u32 compat)
377{
378 int err = 0;
379 if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
380 err = ext4_journal_get_write_access(handle,
381 EXT4_SB(sb)->s_sbh);
382 if (err)
383 return err;
384 EXT4_SET_COMPAT_FEATURE(sb, compat);
385 sb->s_dirt = 1;
386 handle->h_sync = 1;
387 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
388 "call ext4_journal_dirty_met adata");
389 err = ext4_journal_dirty_metadata(handle,
390 EXT4_SB(sb)->s_sbh);
391 }
392 return err;
393}
394
395int ext4_update_rocompat_feature(handle_t *handle,
396 struct super_block *sb, __u32 rocompat)
397{
398 int err = 0;
399 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
400 err = ext4_journal_get_write_access(handle,
401 EXT4_SB(sb)->s_sbh);
402 if (err)
403 return err;
404 EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
405 sb->s_dirt = 1;
406 handle->h_sync = 1;
407 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
408 "call ext4_journal_dirty_met adata");
409 err = ext4_journal_dirty_metadata(handle,
410 EXT4_SB(sb)->s_sbh);
411 }
412 return err;
413}
414
415int ext4_update_incompat_feature(handle_t *handle,
416 struct super_block *sb, __u32 incompat)
417{
418 int err = 0;
419 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
420 err = ext4_journal_get_write_access(handle,
421 EXT4_SB(sb)->s_sbh);
422 if (err)
423 return err;
424 EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
425 sb->s_dirt = 1;
426 handle->h_sync = 1;
427 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
428 "call ext4_journal_dirty_met adata");
429 err = ext4_journal_dirty_metadata(handle,
430 EXT4_SB(sb)->s_sbh);
431 }
432 return err;
433}
434
ac27a0ec
DK		 435/*
436 * Open the external journal device
437 */
617ba13b 438static struct block_device *ext4_blkdev_get(dev_t dev)
ac27a0ec
DK		 439{
440 struct block_device *bdev;
441 char b[BDEVNAME_SIZE];
442
443 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
444 if (IS_ERR(bdev))
445 goto fail;
446 return bdev;
447
448fail:
617ba13b 449 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
ac27a0ec
DK		 450 __bdevname(dev, b), PTR_ERR(bdev));
451 return NULL;
452}
453
454/*
455 * Release the journal device
456 */
617ba13b 457static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 458{
459 bd_release(bdev);
460 return blkdev_put(bdev);
461}
462
617ba13b 463static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 464{
465 struct block_device *bdev;
466 int ret = -ENODEV;
467
468 bdev = sbi->journal_bdev;
469 if (bdev) {
617ba13b 470 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 471 sbi->journal_bdev = NULL;
472 }
473 return ret;
474}
475
476static inline struct inode *orphan_list_entry(struct list_head *l)
477{
617ba13b 478 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 479}
480
617ba13b 481static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 482{
483 struct list_head *l;
484
485 printk(KERN_ERR "sb orphan head is %d\n",
486 le32_to_cpu(sbi->s_es->s_last_orphan));
487
488 printk(KERN_ERR "sb_info orphan list:\n");
489 list_for_each(l, &sbi->s_orphan) {
490 struct inode *inode = orphan_list_entry(l);
491 printk(KERN_ERR " "
492 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
493 inode->i_sb->s_id, inode->i_ino, inode,
494 inode->i_mode, inode->i_nlink,
495 NEXT_ORPHAN(inode));
496 }
497}
498
617ba13b 499static void ext4_put_super (struct super_block * sb)
ac27a0ec 500{
617ba13b
MC		 501 struct ext4_sb_info *sbi = EXT4_SB(sb);
502 struct ext4_super_block *es = sbi->s_es;
ac27a0ec
DK		 503 int i;
504
c9de560d 505 ext4_mb_release(sb);
a86c6181 506 ext4_ext_release(sb);
617ba13b 507 ext4_xattr_put_super(sb);
dab291af 508 jbd2_journal_destroy(sbi->s_journal);
47b4a50b 509 sbi->s_journal = NULL;
ac27a0ec 510 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 511 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec
DK		 512 es->s_state = cpu_to_le16(sbi->s_mount_state);
513 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
514 mark_buffer_dirty(sbi->s_sbh);
617ba13b 515 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 516 }
517
518 for (i = 0; i < sbi->s_gdb_count; i++)
519 brelse(sbi->s_group_desc[i]);
520 kfree(sbi->s_group_desc);
772cb7c8 521 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 522 percpu_counter_destroy(&sbi->s_freeblocks_counter);
523 percpu_counter_destroy(&sbi->s_freeinodes_counter);
524 percpu_counter_destroy(&sbi->s_dirs_counter);
525 brelse(sbi->s_sbh);
526#ifdef CONFIG_QUOTA
527 for (i = 0; i < MAXQUOTAS; i++)
528 kfree(sbi->s_qf_names[i]);
529#endif
530
531 /* Debugging code just in case the in-memory inode orphan list
532 * isn't empty. The on-disk one can be non-empty if we've
533 * detected an error and taken the fs readonly, but the
534 * in-memory list had better be clean by this point. */
535 if (!list_empty(&sbi->s_orphan))
536 dump_orphan_list(sb, sbi);
537 J_ASSERT(list_empty(&sbi->s_orphan));
538
f98393a6 539 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 540 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
541 /*
542 * Invalidate the journal device's buffers. We don't want them
543 * floating about in memory - the physical journal device may
544 * hotswapped, and it breaks the `ro-after' testing code.
545 */
546 sync_blockdev(sbi->journal_bdev);
f98393a6 547 invalidate_bdev(sbi->journal_bdev);
617ba13b 548 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 549 }
550 sb->s_fs_info = NULL;
551 kfree(sbi);
552 return;
553}
554
e18b890b 555static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 556
557/*
558 * Called inside transaction, so use GFP_NOFS
559 */
617ba13b 560static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 561{
617ba13b 562 struct ext4_inode_info *ei;
ac27a0ec 563
e6b4f8da 564 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 565 if (!ei)
566 return NULL;
617ba13b
MC		 567#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
568 ei->i_acl = EXT4_ACL_NOT_CACHED;
569 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 570#endif
571 ei->i_block_alloc_info = NULL;
572 ei->vfs_inode.i_version = 1;
a86c6181 573 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 574 INIT_LIST_HEAD(&ei->i_prealloc_list);
575 spin_lock_init(&ei->i_prealloc_lock);
678aaf48 576 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 577 ei->i_reserved_data_blocks = 0;
578 ei->i_reserved_meta_blocks = 0;
579 ei->i_allocated_meta_blocks = 0;
580 ei->i_delalloc_reserved_flag = 0;
581 spin_lock_init(&(ei->i_block_reservation_lock));
ac27a0ec
DK		 582 return &ei->vfs_inode;
583}
584
617ba13b 585static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 586{
9f7dd93d
VA		 587 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
588 printk("EXT4 Inode %p: orphan list check failed!\n",
589 EXT4_I(inode));
590 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
591 EXT4_I(inode), sizeof(struct ext4_inode_info),
592 true);
593 dump_stack();
594 }
617ba13b 595 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 596}
597
4ba9b9d0 598static void init_once(struct kmem_cache *cachep, void *foo)
ac27a0ec 599{
617ba13b 600 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 601
a35afb83 602 INIT_LIST_HEAD(&ei->i_orphan);
617ba13b 603#ifdef CONFIG_EXT4DEV_FS_XATTR
a35afb83 604 init_rwsem(&ei->xattr_sem);
ac27a0ec 605#endif
0e855ac8 606 init_rwsem(&ei->i_data_sem);
a35afb83 607 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 608}
609
610static int init_inodecache(void)
611{
617ba13b
MC		 612 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
613 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 614 0, (SLAB_RECLAIM_ACCOUNT|
615 SLAB_MEM_SPREAD),
20c2df83 616 init_once);
617ba13b 617 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 618 return -ENOMEM;
619 return 0;
620}
621
622static void destroy_inodecache(void)
623{
617ba13b 624 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 625}
626
617ba13b 627static void ext4_clear_inode(struct inode *inode)
ac27a0ec 628{
617ba13b
MC		 629 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
630#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
631 if (EXT4_I(inode)->i_acl &&
632 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
633 posix_acl_release(EXT4_I(inode)->i_acl);
634 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
635 }
636 if (EXT4_I(inode)->i_default_acl &&
637 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
638 posix_acl_release(EXT4_I(inode)->i_default_acl);
639 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 640 }
641#endif
617ba13b
MC		 642 ext4_discard_reservation(inode);
643 EXT4_I(inode)->i_block_alloc_info = NULL;
ac27a0ec
DK		 644 if (unlikely(rsv))
645 kfree(rsv);
678aaf48
JK		 646 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
647 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 648}
649
617ba13b 650static inline void ext4_show_quota_options(struct seq_file *seq, 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");
ac27a0ec 759
cb45bbe4
MS		 760 if (sbi->s_stripe)
761 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 762 /*
763 * journal mode get enabled in different ways
764 * So just print the value even if we didn't specify it
765 */
617ba13b 766 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 767 seq_puts(seq, ",data=journal");
617ba13b 768 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 769 seq_puts(seq, ",data=ordered");
617ba13b 770 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 771 seq_puts(seq, ",data=writeback");
772
617ba13b 773 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 774 return 0;
775}
776
777
1b961ac0
CH		 778static struct inode *ext4_nfs_get_inode(struct super_block *sb,
779 u64 ino, u32 generation)
ac27a0ec 780{
ac27a0ec 781 struct inode *inode;
ac27a0ec 782
617ba13b 783 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 784 return ERR_PTR(-ESTALE);
617ba13b 785 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 786 return ERR_PTR(-ESTALE);
787
788 /* iget isn't really right if the inode is currently unallocated!!
789 *
617ba13b 790 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 791 * deleted, so we should be safe.
792 *
793 * Currently we don't know the generation for parent directory, so
794 * a generation of 0 means "accept any"
795 */
1d1fe1ee
DH		 796 inode = ext4_iget(sb, ino);
797 if (IS_ERR(inode))
798 return ERR_CAST(inode);
799 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 800 iput(inode);
801 return ERR_PTR(-ESTALE);
802 }
1b961ac0
CH		 803
804 return inode;
805}
806
807static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
808 int fh_len, int fh_type)
809{
810 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
811 ext4_nfs_get_inode);
812}
813
814static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
815 int fh_len, int fh_type)
816{
817 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
818 ext4_nfs_get_inode);
ac27a0ec
DK		 819}
820
821#ifdef CONFIG_QUOTA
822#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
823#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
824
617ba13b
MC		 825static int ext4_dquot_initialize(struct inode *inode, int type);
826static int ext4_dquot_drop(struct inode *inode);
827static int ext4_write_dquot(struct dquot *dquot);
828static int ext4_acquire_dquot(struct dquot *dquot);
829static int ext4_release_dquot(struct dquot *dquot);
830static int ext4_mark_dquot_dirty(struct dquot *dquot);
831static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 832static int ext4_quota_on(struct super_block *sb, int type, int format_id,
833 char *path, int remount);
617ba13b
MC		 834static int ext4_quota_on_mount(struct super_block *sb, int type);
835static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 836 size_t len, loff_t off);
617ba13b 837static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 838 const char *data, size_t len, loff_t off);
839
617ba13b
MC		 840static struct dquot_operations ext4_quota_operations = {
841 .initialize = ext4_dquot_initialize,
842 .drop = ext4_dquot_drop,
ac27a0ec
DK		 843 .alloc_space = dquot_alloc_space,
844 .alloc_inode = dquot_alloc_inode,
845 .free_space = dquot_free_space,
846 .free_inode = dquot_free_inode,
847 .transfer = dquot_transfer,
617ba13b
MC		 848 .write_dquot = ext4_write_dquot,
849 .acquire_dquot = ext4_acquire_dquot,
850 .release_dquot = ext4_release_dquot,
851 .mark_dirty = ext4_mark_dquot_dirty,
852 .write_info = ext4_write_info
ac27a0ec
DK		 853};
854
617ba13b
MC		 855static struct quotactl_ops ext4_qctl_operations = {
856 .quota_on = ext4_quota_on,
ac27a0ec
DK		 857 .quota_off = vfs_quota_off,
858 .quota_sync = vfs_quota_sync,
859 .get_info = vfs_get_dqinfo,
860 .set_info = vfs_set_dqinfo,
861 .get_dqblk = vfs_get_dqblk,
862 .set_dqblk = vfs_set_dqblk
863};
864#endif
865
ee9b6d61 866static const struct super_operations ext4_sops = {
617ba13b
MC		 867 .alloc_inode = ext4_alloc_inode,
868 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 869 .write_inode = ext4_write_inode,
870 .dirty_inode = ext4_dirty_inode,
871 .delete_inode = ext4_delete_inode,
872 .put_super = ext4_put_super,
873 .write_super = ext4_write_super,
874 .sync_fs = ext4_sync_fs,
875 .write_super_lockfs = ext4_write_super_lockfs,
876 .unlockfs = ext4_unlockfs,
877 .statfs = ext4_statfs,
878 .remount_fs = ext4_remount,
879 .clear_inode = ext4_clear_inode,
880 .show_options = ext4_show_options,
ac27a0ec 881#ifdef CONFIG_QUOTA
617ba13b
MC		 882 .quota_read = ext4_quota_read,
883 .quota_write = ext4_quota_write,
ac27a0ec
DK		 884#endif
885};
886
39655164 887static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 888 .fh_to_dentry = ext4_fh_to_dentry,
889 .fh_to_parent = ext4_fh_to_parent,
617ba13b 890 .get_parent = ext4_get_parent,
ac27a0ec
DK		 891};
892
893enum {
894 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
895 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
896 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
897 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
898 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
899 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
818d276c 900 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec
DK		 901 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
902 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
903 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
904 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
25ec56b5 905 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
64769240 906 Opt_mballoc, Opt_nomballoc, Opt_stripe, Opt_delalloc,
ac27a0ec
DK		 907};
908
909static match_table_t tokens = {
910 {Opt_bsd_df, "bsddf"},
911 {Opt_minix_df, "minixdf"},
912 {Opt_grpid, "grpid"},
913 {Opt_grpid, "bsdgroups"},
914 {Opt_nogrpid, "nogrpid"},
915 {Opt_nogrpid, "sysvgroups"},
916 {Opt_resgid, "resgid=%u"},
917 {Opt_resuid, "resuid=%u"},
918 {Opt_sb, "sb=%u"},
919 {Opt_err_cont, "errors=continue"},
920 {Opt_err_panic, "errors=panic"},
921 {Opt_err_ro, "errors=remount-ro"},
922 {Opt_nouid32, "nouid32"},
923 {Opt_nocheck, "nocheck"},
924 {Opt_nocheck, "check=none"},
925 {Opt_debug, "debug"},
926 {Opt_oldalloc, "oldalloc"},
927 {Opt_orlov, "orlov"},
928 {Opt_user_xattr, "user_xattr"},
929 {Opt_nouser_xattr, "nouser_xattr"},
930 {Opt_acl, "acl"},
931 {Opt_noacl, "noacl"},
932 {Opt_reservation, "reservation"},
933 {Opt_noreservation, "noreservation"},
934 {Opt_noload, "noload"},
935 {Opt_nobh, "nobh"},
936 {Opt_bh, "bh"},
937 {Opt_commit, "commit=%u"},
938 {Opt_journal_update, "journal=update"},
939 {Opt_journal_inum, "journal=%u"},
940 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 941 {Opt_journal_checksum, "journal_checksum"},
942 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 943 {Opt_abort, "abort"},
944 {Opt_data_journal, "data=journal"},
945 {Opt_data_ordered, "data=ordered"},
946 {Opt_data_writeback, "data=writeback"},
947 {Opt_offusrjquota, "usrjquota="},
948 {Opt_usrjquota, "usrjquota=%s"},
949 {Opt_offgrpjquota, "grpjquota="},
950 {Opt_grpjquota, "grpjquota=%s"},
951 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
952 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
953 {Opt_grpquota, "grpquota"},
954 {Opt_noquota, "noquota"},
955 {Opt_quota, "quota"},
956 {Opt_usrquota, "usrquota"},
957 {Opt_barrier, "barrier=%u"},
a86c6181 958 {Opt_extents, "extents"},
1e2462f9 959 {Opt_noextents, "noextents"},
25ec56b5 960 {Opt_i_version, "i_version"},
c9de560d
AT		 961 {Opt_mballoc, "mballoc"},
962 {Opt_nomballoc, "nomballoc"},
963 {Opt_stripe, "stripe=%u"},
ac27a0ec 964 {Opt_resize, "resize"},
64769240 965 {Opt_delalloc, "delalloc"},
f3f12faa 966 {Opt_err, NULL},
ac27a0ec
DK		 967};
968
617ba13b 969static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 970{
617ba13b 971 ext4_fsblk_t sb_block;
ac27a0ec
DK		 972 char *options = (char *) *data;
973
974 if (!options || strncmp(options, "sb=", 3) != 0)
975 return 1; /* Default location */
976 options += 3;
617ba13b 977 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 978 sb_block = simple_strtoul(options, &options, 0);
979 if (*options && *options != ',') {
617ba13b 980 printk("EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 981 (char *) *data);
982 return 1;
983 }
984 if (*options == ',')
985 options++;
986 *data = (void *) options;
987 return sb_block;
988}
989
990static int parse_options (char *options, struct super_block *sb,
991 unsigned int *inum, unsigned long *journal_devnum,
617ba13b 992 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 993{
617ba13b 994 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 995 char * p;
996 substring_t args[MAX_OPT_ARGS];
997 int data_opt = 0;
998 int option;
999#ifdef CONFIG_QUOTA
dfc5d03f 1000 int qtype, qfmt;
ac27a0ec
DK		 1001 char *qname;
1002#endif
1003
1004 if (!options)
1005 return 1;
1006
1007 while ((p = strsep (&options, ",")) != NULL) {
1008 int token;
1009 if (!*p)
1010 continue;
1011
1012 token = match_token(p, tokens, args);
1013 switch (token) {
1014 case Opt_bsd_df:
1015 clear_opt (sbi->s_mount_opt, MINIX_DF);
1016 break;
1017 case Opt_minix_df:
1018 set_opt (sbi->s_mount_opt, MINIX_DF);
1019 break;
1020 case Opt_grpid:
1021 set_opt (sbi->s_mount_opt, GRPID);
1022 break;
1023 case Opt_nogrpid:
1024 clear_opt (sbi->s_mount_opt, GRPID);
1025 break;
1026 case Opt_resuid:
1027 if (match_int(&args[0], &option))
1028 return 0;
1029 sbi->s_resuid = option;
1030 break;
1031 case Opt_resgid:
1032 if (match_int(&args[0], &option))
1033 return 0;
1034 sbi->s_resgid = option;
1035 break;
1036 case Opt_sb:
1037 /* handled by get_sb_block() instead of here */
1038 /* *sb_block = match_int(&args[0]); */
1039 break;
1040 case Opt_err_panic:
1041 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1042 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1043 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1044 break;
1045 case Opt_err_ro:
1046 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1047 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1048 set_opt (sbi->s_mount_opt, ERRORS_RO);
1049 break;
1050 case Opt_err_cont:
1051 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1052 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1053 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1054 break;
1055 case Opt_nouid32:
1056 set_opt (sbi->s_mount_opt, NO_UID32);
1057 break;
1058 case Opt_nocheck:
1059 clear_opt (sbi->s_mount_opt, CHECK);
1060 break;
1061 case Opt_debug:
1062 set_opt (sbi->s_mount_opt, DEBUG);
1063 break;
1064 case Opt_oldalloc:
1065 set_opt (sbi->s_mount_opt, OLDALLOC);
1066 break;
1067 case Opt_orlov:
1068 clear_opt (sbi->s_mount_opt, OLDALLOC);
1069 break;
617ba13b 1070#ifdef CONFIG_EXT4DEV_FS_XATTR
ac27a0ec
DK		 1071 case Opt_user_xattr:
1072 set_opt (sbi->s_mount_opt, XATTR_USER);
1073 break;
1074 case Opt_nouser_xattr:
1075 clear_opt (sbi->s_mount_opt, XATTR_USER);
1076 break;
1077#else
1078 case Opt_user_xattr:
1079 case Opt_nouser_xattr:
617ba13b 1080 printk("EXT4 (no)user_xattr options not supported\n");
ac27a0ec
DK		 1081 break;
1082#endif
617ba13b 1083#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
ac27a0ec
DK		 1084 case Opt_acl:
1085 set_opt(sbi->s_mount_opt, POSIX_ACL);
1086 break;
1087 case Opt_noacl:
1088 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1089 break;
1090#else
1091 case Opt_acl:
1092 case Opt_noacl:
617ba13b 1093 printk("EXT4 (no)acl options not supported\n");
ac27a0ec
DK		 1094 break;
1095#endif
1096 case Opt_reservation:
1097 set_opt(sbi->s_mount_opt, RESERVATION);
1098 break;
1099 case Opt_noreservation:
1100 clear_opt(sbi->s_mount_opt, RESERVATION);
1101 break;
1102 case Opt_journal_update:
1103 /* @@@ FIXME */
1104 /* Eventually we will want to be able to create
1105 a journal file here. For now, only allow the
1106 user to specify an existing inode to be the
1107 journal file. */
1108 if (is_remount) {
617ba13b 1109 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1110 "journal on remount\n");
1111 return 0;
1112 }
1113 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1114 break;
1115 case Opt_journal_inum:
1116 if (is_remount) {
617ba13b 1117 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1118 "journal on remount\n");
1119 return 0;
1120 }
1121 if (match_int(&args[0], &option))
1122 return 0;
1123 *inum = option;
1124 break;
1125 case Opt_journal_dev:
1126 if (is_remount) {
617ba13b 1127 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1128 "journal on remount\n");
1129 return 0;
1130 }
1131 if (match_int(&args[0], &option))
1132 return 0;
1133 *journal_devnum = option;
1134 break;
818d276c
GS		 1135 case Opt_journal_checksum:
1136 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1137 break;
1138 case Opt_journal_async_commit:
1139 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1140 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1141 break;
ac27a0ec
DK		 1142 case Opt_noload:
1143 set_opt (sbi->s_mount_opt, NOLOAD);
1144 break;
1145 case Opt_commit:
1146 if (match_int(&args[0], &option))
1147 return 0;
1148 if (option < 0)
1149 return 0;
1150 if (option == 0)
cd02ff0b 1151 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1152 sbi->s_commit_interval = HZ * option;
1153 break;
1154 case Opt_data_journal:
617ba13b 1155 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1156 goto datacheck;
1157 case Opt_data_ordered:
617ba13b 1158 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1159 goto datacheck;
1160 case Opt_data_writeback:
617ba13b 1161 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1162 datacheck:
1163 if (is_remount) {
617ba13b 1164 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1165 != data_opt) {
1166 printk(KERN_ERR
617ba13b 1167 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1168 "mode on remount\n");
1169 return 0;
1170 }
1171 } else {
617ba13b 1172 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1173 sbi->s_mount_opt |= data_opt;
1174 }
1175 break;
1176#ifdef CONFIG_QUOTA
1177 case Opt_usrjquota:
1178 qtype = USRQUOTA;
1179 goto set_qf_name;
1180 case Opt_grpjquota:
1181 qtype = GRPQUOTA;
1182set_qf_name:
dfc5d03f
JK		 1183 if ((sb_any_quota_enabled(sb) ||
1184 sb_any_quota_suspended(sb)) &&
1185 !sbi->s_qf_names[qtype]) {
ac27a0ec 1186 printk(KERN_ERR
2c8be6b2 1187 "EXT4-fs: Cannot change journaled "
ac27a0ec
DK		 1188 "quota options when quota turned on.\n");
1189 return 0;
1190 }
1191 qname = match_strdup(&args[0]);
1192 if (!qname) {
1193 printk(KERN_ERR
617ba13b 1194 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1195 "storing quotafile name.\n");
1196 return 0;
1197 }
1198 if (sbi->s_qf_names[qtype] &&
1199 strcmp(sbi->s_qf_names[qtype], qname)) {
1200 printk(KERN_ERR
617ba13b 1201 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1202 "specified.\n", QTYPE2NAME(qtype));
1203 kfree(qname);
1204 return 0;
1205 }
1206 sbi->s_qf_names[qtype] = qname;
1207 if (strchr(sbi->s_qf_names[qtype], '/')) {
1208 printk(KERN_ERR
617ba13b 1209 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1210 "filesystem root.\n");
1211 kfree(sbi->s_qf_names[qtype]);
1212 sbi->s_qf_names[qtype] = NULL;
1213 return 0;
1214 }
1215 set_opt(sbi->s_mount_opt, QUOTA);
1216 break;
1217 case Opt_offusrjquota:
1218 qtype = USRQUOTA;
1219 goto clear_qf_name;
1220 case Opt_offgrpjquota:
1221 qtype = GRPQUOTA;
1222clear_qf_name:
dfc5d03f
JK		 1223 if ((sb_any_quota_enabled(sb) ||
1224 sb_any_quota_suspended(sb)) &&
1225 sbi->s_qf_names[qtype]) {
617ba13b 1226 printk(KERN_ERR "EXT4-fs: Cannot change "
2c8be6b2 1227 "journaled quota options when "
ac27a0ec
DK		 1228 "quota turned on.\n");
1229 return 0;
1230 }
1231 /*
1232 * The space will be released later when all options
1233 * are confirmed to be correct
1234 */
1235 sbi->s_qf_names[qtype] = NULL;
1236 break;
1237 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1238 qfmt = QFMT_VFS_OLD;
1239 goto set_qf_format;
ac27a0ec 1240 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1241 qfmt = QFMT_VFS_V0;
1242set_qf_format:
1243 if ((sb_any_quota_enabled(sb) ||
1244 sb_any_quota_suspended(sb)) &&
1245 sbi->s_jquota_fmt != qfmt) {
1246 printk(KERN_ERR "EXT4-fs: Cannot change "
1247 "journaled quota options when "
1248 "quota turned on.\n");
1249 return 0;
1250 }
1251 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1252 break;
1253 case Opt_quota:
1254 case Opt_usrquota:
1255 set_opt(sbi->s_mount_opt, QUOTA);
1256 set_opt(sbi->s_mount_opt, USRQUOTA);
1257 break;
1258 case Opt_grpquota:
1259 set_opt(sbi->s_mount_opt, QUOTA);
1260 set_opt(sbi->s_mount_opt, GRPQUOTA);
1261 break;
1262 case Opt_noquota:
1263 if (sb_any_quota_enabled(sb)) {
617ba13b 1264 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1265 "options when quota turned on.\n");
1266 return 0;
1267 }
1268 clear_opt(sbi->s_mount_opt, QUOTA);
1269 clear_opt(sbi->s_mount_opt, USRQUOTA);
1270 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1271 break;
1272#else
1273 case Opt_quota:
1274 case Opt_usrquota:
1275 case Opt_grpquota:
cd59e7b9
JK		 1276 printk(KERN_ERR
1277 "EXT4-fs: quota options not supported.\n");
1278 break;
ac27a0ec
DK		 1279 case Opt_usrjquota:
1280 case Opt_grpjquota:
1281 case Opt_offusrjquota:
1282 case Opt_offgrpjquota:
1283 case Opt_jqfmt_vfsold:
1284 case Opt_jqfmt_vfsv0:
1285 printk(KERN_ERR
cd59e7b9 1286 "EXT4-fs: journaled quota options not "
ac27a0ec
DK		 1287 "supported.\n");
1288 break;
1289 case Opt_noquota:
1290 break;
1291#endif
1292 case Opt_abort:
1293 set_opt(sbi->s_mount_opt, ABORT);
1294 break;
1295 case Opt_barrier:
1296 if (match_int(&args[0], &option))
1297 return 0;
1298 if (option)
1299 set_opt(sbi->s_mount_opt, BARRIER);
1300 else
1301 clear_opt(sbi->s_mount_opt, BARRIER);
1302 break;
1303 case Opt_ignore:
1304 break;
1305 case Opt_resize:
1306 if (!is_remount) {
617ba13b 1307 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1308 "for remount\n");
1309 return 0;
1310 }
1311 if (match_int(&args[0], &option) != 0)
1312 return 0;
1313 *n_blocks_count = option;
1314 break;
1315 case Opt_nobh:
1316 set_opt(sbi->s_mount_opt, NOBH);
1317 break;
1318 case Opt_bh:
1319 clear_opt(sbi->s_mount_opt, NOBH);
1320 break;
a86c6181
AT		 1321 case Opt_extents:
1322 set_opt (sbi->s_mount_opt, EXTENTS);
1323 break;
1e2462f9
MC		 1324 case Opt_noextents:
1325 clear_opt (sbi->s_mount_opt, EXTENTS);
1326 break;
25ec56b5
JNC		 1327 case Opt_i_version:
1328 set_opt(sbi->s_mount_opt, I_VERSION);
1329 sb->s_flags |= MS_I_VERSION;
1330 break;
c9de560d
AT		 1331 case Opt_mballoc:
1332 set_opt(sbi->s_mount_opt, MBALLOC);
1333 break;
1334 case Opt_nomballoc:
1335 clear_opt(sbi->s_mount_opt, MBALLOC);
1336 break;
1337 case Opt_stripe:
1338 if (match_int(&args[0], &option))
1339 return 0;
1340 if (option < 0)
1341 return 0;
1342 sbi->s_stripe = option;
1343 break;
64769240
AT		 1344 case Opt_delalloc:
1345 set_opt(sbi->s_mount_opt, DELALLOC);
1346 break;
ac27a0ec
DK		 1347 default:
1348 printk (KERN_ERR
617ba13b 1349 "EXT4-fs: Unrecognized mount option \"%s\" "
ac27a0ec
DK		 1350 "or missing value\n", p);
1351 return 0;
1352 }
1353 }
1354#ifdef CONFIG_QUOTA
1355 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1356 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1357 sbi->s_qf_names[USRQUOTA])
1358 clear_opt(sbi->s_mount_opt, USRQUOTA);
1359
617ba13b 1360 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1361 sbi->s_qf_names[GRPQUOTA])
1362 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1363
1364 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1365 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1366 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1367 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1368 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1369 "format mixing.\n");
1370 return 0;
1371 }
1372
1373 if (!sbi->s_jquota_fmt) {
2c8be6b2 1374 printk(KERN_ERR "EXT4-fs: journaled quota format "
ac27a0ec
DK		 1375 "not specified.\n");
1376 return 0;
1377 }
1378 } else {
1379 if (sbi->s_jquota_fmt) {
2c8be6b2
JK		 1380 printk(KERN_ERR "EXT4-fs: journaled quota format "
1381 "specified with no journaling "
ac27a0ec
DK		 1382 "enabled.\n");
1383 return 0;
1384 }
1385 }
1386#endif
1387 return 1;
1388}
1389
617ba13b 1390static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1391 int read_only)
1392{
617ba13b 1393 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1394 int res = 0;
1395
617ba13b
MC		 1396 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1397 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
ac27a0ec
DK		 1398 "forcing read-only mode\n");
1399 res = MS_RDONLY;
1400 }
1401 if (read_only)
1402 return res;
617ba13b
MC		 1403 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1404 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
ac27a0ec 1405 "running e2fsck is recommended\n");
617ba13b 1406 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
ac27a0ec 1407 printk (KERN_WARNING
617ba13b 1408 "EXT4-fs warning: mounting fs with errors, "
ac27a0ec
DK		 1409 "running e2fsck is recommended\n");
1410 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1411 le16_to_cpu(es->s_mnt_count) >=
1412 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1413 printk (KERN_WARNING
617ba13b 1414 "EXT4-fs warning: maximal mount count reached, "
ac27a0ec
DK		 1415 "running e2fsck is recommended\n");
1416 else if (le32_to_cpu(es->s_checkinterval) &&
1417 (le32_to_cpu(es->s_lastcheck) +
1418 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1419 printk (KERN_WARNING
617ba13b 1420 "EXT4-fs warning: checktime reached, "
ac27a0ec
DK		 1421 "running e2fsck is recommended\n");
1422#if 0
1423 /* @@@ We _will_ want to clear the valid bit if we find
63f57933
AM		 1424 * inconsistencies, to force a fsck at reboot. But for
1425 * a plain journaled filesystem we can keep it set as
1426 * valid forever! :)
1427 */
216c34b2 1428 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec
DK		 1429#endif
1430 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1431 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1432 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1433 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b
MC		 1434 ext4_update_dynamic_rev(sb);
1435 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1436
617ba13b 1437 ext4_commit_super(sb, es, 1);
ac27a0ec 1438 if (test_opt(sb, DEBUG))
617ba13b 1439 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
ac27a0ec
DK		 1440 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1441 sb->s_blocksize,
1442 sbi->s_groups_count,
617ba13b
MC		 1443 EXT4_BLOCKS_PER_GROUP(sb),
1444 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1445 sbi->s_mount_opt);
1446
617ba13b
MC		 1447 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1448 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
ac27a0ec
DK		 1449 char b[BDEVNAME_SIZE];
1450
1451 printk("external journal on %s\n",
617ba13b 1452 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
ac27a0ec
DK		 1453 } else {
1454 printk("internal journal\n");
1455 }
1456 return res;
1457}
1458
772cb7c8
JS		 1459static int ext4_fill_flex_info(struct super_block *sb)
1460{
1461 struct ext4_sb_info *sbi = EXT4_SB(sb);
1462 struct ext4_group_desc *gdp = NULL;
1463 struct buffer_head *bh;
1464 ext4_group_t flex_group_count;
1465 ext4_group_t flex_group;
1466 int groups_per_flex = 0;
1467 __u64 block_bitmap = 0;
1468 int i;
1469
1470 if (!sbi->s_es->s_log_groups_per_flex) {
1471 sbi->s_log_groups_per_flex = 0;
1472 return 1;
1473 }
1474
1475 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1476 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1477
1478 flex_group_count = (sbi->s_groups_count + groups_per_flex - 1) /
1479 groups_per_flex;
1480 sbi->s_flex_groups = kmalloc(flex_group_count *
1481 sizeof(struct flex_groups), GFP_KERNEL);
1482 if (sbi->s_flex_groups == NULL) {
1483 printk(KERN_ERR "EXT4-fs: not enough memory\n");
1484 goto failed;
1485 }
1486 memset(sbi->s_flex_groups, 0, flex_group_count *
1487 sizeof(struct flex_groups));
1488
1489 gdp = ext4_get_group_desc(sb, 1, &bh);
1490 block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1491
1492 for (i = 0; i < sbi->s_groups_count; i++) {
1493 gdp = ext4_get_group_desc(sb, i, &bh);
1494
1495 flex_group = ext4_flex_group(sbi, i);
1496 sbi->s_flex_groups[flex_group].free_inodes +=
1497 le16_to_cpu(gdp->bg_free_inodes_count);
1498 sbi->s_flex_groups[flex_group].free_blocks +=
1499 le16_to_cpu(gdp->bg_free_blocks_count);
1500 }
1501
1502 return 1;
1503failed:
1504 return 0;
1505}
1506
717d50e4
AD		 1507__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1508 struct ext4_group_desc *gdp)
1509{
1510 __u16 crc = 0;
1511
1512 if (sbi->s_es->s_feature_ro_compat &
1513 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1514 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1515 __le32 le_group = cpu_to_le32(block_group);
1516
1517 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1518 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1519 crc = crc16(crc, (__u8 *)gdp, offset);
1520 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1521 /* for checksum of struct ext4_group_desc do the rest...*/
1522 if ((sbi->s_es->s_feature_incompat &
1523 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1524 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1525 crc = crc16(crc, (__u8 *)gdp + offset,
1526 le16_to_cpu(sbi->s_es->s_desc_size) -
1527 offset);
1528 }
1529
1530 return cpu_to_le16(crc);
1531}
1532
1533int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1534 struct ext4_group_desc *gdp)
1535{
1536 if ((sbi->s_es->s_feature_ro_compat &
1537 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1538 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1539 return 0;
1540
1541 return 1;
1542}
1543
ac27a0ec 1544/* Called at mount-time, super-block is locked */
197cd65a 1545static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1546{
617ba13b
MC		 1547 struct ext4_sb_info *sbi = EXT4_SB(sb);
1548 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1549 ext4_fsblk_t last_block;
bd81d8ee
LV		 1550 ext4_fsblk_t block_bitmap;
1551 ext4_fsblk_t inode_bitmap;
1552 ext4_fsblk_t inode_table;
ce421581 1553 int flexbg_flag = 0;
fd2d4291 1554 ext4_group_t i;
ac27a0ec 1555
ce421581
JS		 1556 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1557 flexbg_flag = 1;
1558
617ba13b 1559 ext4_debug ("Checking group descriptors");
ac27a0ec 1560
197cd65a
AM		 1561 for (i = 0; i < sbi->s_groups_count; i++) {
1562 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1563
ce421581 1564 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1565 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1566 else
1567 last_block = first_block +
617ba13b 1568 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1569
8fadc143 1570 block_bitmap = ext4_block_bitmap(sb, gdp);
bd81d8ee 1571 if (block_bitmap < first_block || block_bitmap > last_block)
ac27a0ec 1572 {
c19204b0
JB		 1573 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1574 "Block bitmap for group %lu not in group "
1575 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1576 return 0;
1577 }
8fadc143 1578 inode_bitmap = ext4_inode_bitmap(sb, gdp);
bd81d8ee 1579 if (inode_bitmap < first_block || inode_bitmap > last_block)
ac27a0ec 1580 {
c19204b0
JB		 1581 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1582 "Inode bitmap for group %lu not in group "
1583 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1584 return 0;
1585 }
8fadc143 1586 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1587 if (inode_table < first_block ||
780dcdb2 1588 inode_table + sbi->s_itb_per_group - 1 > last_block)
ac27a0ec 1589 {
c19204b0
JB		 1590 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1591 "Inode table for group %lu not in group "
1592 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1593 return 0;
1594 }
717d50e4 1595 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
c19204b0
JB		 1596 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1597 "Checksum for group %lu failed (%u!=%u)\n",
1598 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1599 gdp)), le16_to_cpu(gdp->bg_checksum));
717d50e4
AD		 1600 return 0;
1601 }
ce421581
JS		 1602 if (!flexbg_flag)
1603 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1604 }
1605
bd81d8ee 1606 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
617ba13b 1607 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1608 return 1;
1609}
1610
617ba13b 1611/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1612 * the superblock) which were deleted from all directories, but held open by
1613 * a process at the time of a crash. We walk the list and try to delete these
1614 * inodes at recovery time (only with a read-write filesystem).
1615 *
1616 * In order to keep the orphan inode chain consistent during traversal (in
1617 * case of crash during recovery), we link each inode into the superblock
1618 * orphan list_head and handle it the same way as an inode deletion during
1619 * normal operation (which journals the operations for us).
1620 *
1621 * We only do an iget() and an iput() on each inode, which is very safe if we
1622 * accidentally point at an in-use or already deleted inode. The worst that
1623 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1624 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1625 * e2fsck was run on this filesystem, and it must have already done the orphan
1626 * inode cleanup for us, so we can safely abort without any further action.
1627 */
617ba13b
MC		 1628static void ext4_orphan_cleanup (struct super_block * sb,
1629 struct ext4_super_block * es)
ac27a0ec
DK		 1630{
1631 unsigned int s_flags = sb->s_flags;
1632 int nr_orphans = 0, nr_truncates = 0;
1633#ifdef CONFIG_QUOTA
1634 int i;
1635#endif
1636 if (!es->s_last_orphan) {
1637 jbd_debug(4, "no orphan inodes to clean up\n");
1638 return;
1639 }
1640
a8f48a95
ES		 1641 if (bdev_read_only(sb->s_bdev)) {
1642 printk(KERN_ERR "EXT4-fs: write access "
1643 "unavailable, skipping orphan cleanup.\n");
1644 return;
1645 }
1646
617ba13b 1647 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1648 if (es->s_last_orphan)
1649 jbd_debug(1, "Errors on filesystem, "
1650 "clearing orphan list.\n");
1651 es->s_last_orphan = 0;
1652 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1653 return;
1654 }
1655
1656 if (s_flags & MS_RDONLY) {
617ba13b 1657 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1658 sb->s_id);
1659 sb->s_flags &= ~MS_RDONLY;
1660 }
1661#ifdef CONFIG_QUOTA
1662 /* Needed for iput() to work correctly and not trash data */
1663 sb->s_flags |= MS_ACTIVE;
1664 /* Turn on quotas so that they are updated correctly */
1665 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1666 if (EXT4_SB(sb)->s_qf_names[i]) {
1667 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1668 if (ret < 0)
1669 printk(KERN_ERR
2c8be6b2 1670 "EXT4-fs: Cannot turn on journaled "
ac27a0ec
DK		 1671 "quota: error %d\n", ret);
1672 }
1673 }
1674#endif
1675
1676 while (es->s_last_orphan) {
1677 struct inode *inode;
1678
97bd42b9
JB		 1679 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1680 if (IS_ERR(inode)) {
ac27a0ec
DK		 1681 es->s_last_orphan = 0;
1682 break;
1683 }
1684
617ba13b 1685 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1686 DQUOT_INIT(inode);
1687 if (inode->i_nlink) {
1688 printk(KERN_DEBUG
1689 "%s: truncating inode %lu to %Ld bytes\n",
46e665e9 1690 __func__, inode->i_ino, inode->i_size);
ac27a0ec
DK		 1691 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1692 inode->i_ino, inode->i_size);
617ba13b 1693 ext4_truncate(inode);
ac27a0ec
DK		 1694 nr_truncates++;
1695 } else {
1696 printk(KERN_DEBUG
1697 "%s: deleting unreferenced inode %lu\n",
46e665e9 1698 __func__, inode->i_ino);
ac27a0ec
DK		 1699 jbd_debug(2, "deleting unreferenced inode %lu\n",
1700 inode->i_ino);
1701 nr_orphans++;
1702 }
1703 iput(inode); /* The delete magic happens here! */
1704 }
1705
1706#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1707
1708 if (nr_orphans)
617ba13b 1709 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1710 sb->s_id, PLURAL(nr_orphans));
1711 if (nr_truncates)
617ba13b 1712 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1713 sb->s_id, PLURAL(nr_truncates));
1714#ifdef CONFIG_QUOTA
1715 /* Turn quotas off */
1716 for (i = 0; i < MAXQUOTAS; i++) {
1717 if (sb_dqopt(sb)->files[i])
6f28e087 1718 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1719 }
1720#endif
1721 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1722}
cd2291a4
ES		 1723/*
1724 * Maximal extent format file size.
1725 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1726 * extent format containers, within a sector_t, and within i_blocks
1727 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1728 * so that won't be a limiting factor.
1729 *
1730 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1731 */
1732static loff_t ext4_max_size(int blkbits)
1733{
1734 loff_t res;
1735 loff_t upper_limit = MAX_LFS_FILESIZE;
1736
1737 /* small i_blocks in vfs inode? */
1738 if (sizeof(blkcnt_t) < sizeof(u64)) {
1739 /*
1740 * CONFIG_LSF is not enabled implies the inode
1741 * i_block represent total blocks in 512 bytes
1742 * 32 == size of vfs inode i_blocks * 8
1743 */
1744 upper_limit = (1LL << 32) - 1;
1745
1746 /* total blocks in file system block size */
1747 upper_limit >>= (blkbits - 9);
1748 upper_limit <<= blkbits;
1749 }
1750
1751 /* 32-bit extent-start container, ee_block */
1752 res = 1LL << 32;
1753 res <<= blkbits;
1754 res -= 1;
1755
1756 /* Sanity check against vm- & vfs- imposed limits */
1757 if (res > upper_limit)
1758 res = upper_limit;
1759
1760 return res;
1761}
ac27a0ec 1762
ac27a0ec 1763/*
cd2291a4 1764 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1765 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1766 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1767 */
cd2291a4 1768static loff_t ext4_max_bitmap_size(int bits)
ac27a0ec 1769{
617ba13b 1770 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1771 int meta_blocks;
1772 loff_t upper_limit;
1773 /* This is calculated to be the largest file size for a
cd2291a4 1774 * dense, bitmapped file such that the total number of
ac27a0ec 1775 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1776 * does not exceed 2^48 -1
1777 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1778 * total number of 512 bytes blocks of the file
1779 */
1780
1781 if (sizeof(blkcnt_t) < sizeof(u64)) {
1782 /*
1783 * CONFIG_LSF is not enabled implies the inode
1784 * i_block represent total blocks in 512 bytes
1785 * 32 == size of vfs inode i_blocks * 8
1786 */
1787 upper_limit = (1LL << 32) - 1;
1788
1789 /* total blocks in file system block size */
1790 upper_limit >>= (bits - 9);
1791
1792 } else {
8180a562
AK		 1793 /*
1794 * We use 48 bit ext4_inode i_blocks
1795 * With EXT4_HUGE_FILE_FL set the i_blocks
1796 * represent total number of blocks in
1797 * file system block size
1798 */
0fc1b451
AK		 1799 upper_limit = (1LL << 48) - 1;
1800
0fc1b451
AK		 1801 }
1802
1803 /* indirect blocks */
1804 meta_blocks = 1;
1805 /* double indirect blocks */
1806 meta_blocks += 1 + (1LL << (bits-2));
1807 /* tripple indirect blocks */
1808 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1809
1810 upper_limit -= meta_blocks;
1811 upper_limit <<= bits;
ac27a0ec
DK		 1812
1813 res += 1LL << (bits-2);
1814 res += 1LL << (2*(bits-2));
1815 res += 1LL << (3*(bits-2));
1816 res <<= bits;
1817 if (res > upper_limit)
1818 res = upper_limit;
0fc1b451
AK		 1819
1820 if (res > MAX_LFS_FILESIZE)
1821 res = MAX_LFS_FILESIZE;
1822
ac27a0ec
DK		 1823 return res;
1824}
1825
617ba13b 1826static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1827 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1828{
617ba13b 1829 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1830 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1831 int has_super = 0;
1832
1833 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1834
617ba13b 1835 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1836 nr < first_meta_bg)
70bbb3e0 1837 return logical_sb_block + nr + 1;
ac27a0ec 1838 bg = sbi->s_desc_per_block * nr;
617ba13b 1839 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1840 has_super = 1;
617ba13b 1841 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1842}
1843
c9de560d
AT		 1844/**
1845 * ext4_get_stripe_size: Get the stripe size.
1846 * @sbi: In memory super block info
1847 *
1848 * If we have specified it via mount option, then
1849 * use the mount option value. If the value specified at mount time is
1850 * greater than the blocks per group use the super block value.
1851 * If the super block value is greater than blocks per group return 0.
1852 * Allocator needs it be less than blocks per group.
1853 *
1854 */
1855static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1856{
1857 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1858 unsigned long stripe_width =
1859 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1860
1861 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1862 return sbi->s_stripe;
1863
1864 if (stripe_width <= sbi->s_blocks_per_group)
1865 return stripe_width;
1866
1867 if (stride <= sbi->s_blocks_per_group)
1868 return stride;
1869
1870 return 0;
1871}
ac27a0ec 1872
617ba13b 1873static int ext4_fill_super (struct super_block *sb, void *data, int silent)
7477827f
AK		 1874 __releases(kernel_lock)
1875 __acquires(kernel_lock)
1d03ec98 1876
ac27a0ec
DK		 1877{
1878 struct buffer_head * bh;
617ba13b
MC		 1879 struct ext4_super_block *es = NULL;
1880 struct ext4_sb_info *sbi;
1881 ext4_fsblk_t block;
1882 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 1883 ext4_fsblk_t logical_sb_block;
ac27a0ec
DK		 1884 unsigned long offset = 0;
1885 unsigned int journal_inum = 0;
1886 unsigned long journal_devnum = 0;
1887 unsigned long def_mount_opts;
1888 struct inode *root;
1d1fe1ee 1889 int ret = -EINVAL;
ac27a0ec 1890 int blocksize;
ac27a0ec
DK		 1891 int db_count;
1892 int i;
1893 int needs_recovery;
1894 __le32 features;
bd81d8ee 1895 __u64 blocks_count;
833f4077 1896 int err;
ac27a0ec
DK		 1897
1898 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1899 if (!sbi)
1900 return -ENOMEM;
1901 sb->s_fs_info = sbi;
1902 sbi->s_mount_opt = 0;
617ba13b
MC		 1903 sbi->s_resuid = EXT4_DEF_RESUID;
1904 sbi->s_resgid = EXT4_DEF_RESGID;
d9c9bef1 1905 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 1906
1907 unlock_kernel();
1908
617ba13b 1909 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 1910 if (!blocksize) {
617ba13b 1911 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 1912 goto out_fail;
1913 }
1914
1915 /*
617ba13b 1916 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 1917 * block sizes. We need to calculate the offset from buffer start.
1918 */
617ba13b 1919 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 1920 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1921 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 1922 } else {
70bbb3e0 1923 logical_sb_block = sb_block;
ac27a0ec
DK		 1924 }
1925
70bbb3e0 1926 if (!(bh = sb_bread(sb, logical_sb_block))) {
617ba13b 1927 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 1928 goto out_fail;
1929 }
1930 /*
1931 * Note: s_es must be initialized as soon as possible because
617ba13b 1932 * some ext4 macro-instructions depend on its value
ac27a0ec 1933 */
617ba13b 1934 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 1935 sbi->s_es = es;
1936 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 1937 if (sb->s_magic != EXT4_SUPER_MAGIC)
1938 goto cantfind_ext4;
ac27a0ec
DK		 1939
1940 /* Set defaults before we parse the mount options */
1941 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 1942 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 1943 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 1944 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 1945 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 1946 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 1947 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1948#ifdef CONFIG_EXT4DEV_FS_XATTR
617ba13b 1949 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 1950 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1951#endif
1952#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
617ba13b 1953 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 1954 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1955#endif
617ba13b
MC		 1956 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1957 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1958 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1959 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1960 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1961 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1962
1963 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 1964 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 1965 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 1966 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 1967 else
1968 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1969
1970 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1971 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1972
1973 set_opt(sbi->s_mount_opt, RESERVATION);
571640ca 1974 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 1975
1e2462f9
MC		 1976 /*
1977 * turn on extents feature by default in ext4 filesystem
1978 * User -o noextents to turn it off
1979 */
1980 set_opt(sbi->s_mount_opt, EXTENTS);
3dbd0ede
AK		 1981 /*
1982 * turn on mballoc feature by default in ext4 filesystem
1983 * User -o nomballoc to turn it off
1984 */
1985 set_opt(sbi->s_mount_opt, MBALLOC);
1e2462f9 1986
ac27a0ec
DK		 1987 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1988 NULL, 0))
1989 goto failed_mount;
1990
1991 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 1992 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 1993
617ba13b
MC		 1994 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1995 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1996 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1997 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 1998 printk(KERN_WARNING
617ba13b 1999 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec 2000 "running e2fsck is recommended\n");
469108ff
TT		 2001
2002 /*
2003 * Since ext4 is still considered development code, we require
2004 * that the TEST_FILESYS flag in s->flags be set.
2005 */
2006 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
2007 printk(KERN_WARNING "EXT4-fs: %s: not marked "
2008 "OK to use with test code.\n", sb->s_id);
2009 goto failed_mount;
2010 }
2011
ac27a0ec
DK		 2012 /*
2013 * Check feature flags regardless of the revision level, since we
2014 * previously didn't change the revision level when setting the flags,
2015 * so there is a chance incompat flags are set on a rev 0 filesystem.
2016 */
617ba13b 2017 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 2018 if (features) {
617ba13b 2019 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
ac27a0ec
DK		 2020 "unsupported optional features (%x).\n",
2021 sb->s_id, le32_to_cpu(features));
2022 goto failed_mount;
2023 }
617ba13b 2024 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 2025 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 2026 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
ac27a0ec
DK		 2027 "unsupported optional features (%x).\n",
2028 sb->s_id, le32_to_cpu(features));
2029 goto failed_mount;
2030 }
0fc1b451
AK		 2031 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2032 /*
2033 * Large file size enabled file system can only be
2034 * mount if kernel is build with CONFIG_LSF
2035 */
2036 if (sizeof(root->i_blocks) < sizeof(u64) &&
2037 !(sb->s_flags & MS_RDONLY)) {
2038 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2039 "files cannot be mounted read-write "
2040 "without CONFIG_LSF.\n", sb->s_id);
2041 goto failed_mount;
2042 }
2043 }
ac27a0ec
DK		 2044 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2045
617ba13b
MC		 2046 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2047 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 2048 printk(KERN_ERR
617ba13b 2049 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 2050 blocksize, sb->s_id);
2051 goto failed_mount;
2052 }
2053
ac27a0ec 2054 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2055
2056 /* Validate the filesystem blocksize */
2057 if (!sb_set_blocksize(sb, blocksize)) {
2058 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2059 blocksize);
ac27a0ec
DK		 2060 goto failed_mount;
2061 }
2062
2063 brelse (bh);
70bbb3e0
AM		 2064 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2065 offset = do_div(logical_sb_block, blocksize);
2066 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 2067 if (!bh) {
2068 printk(KERN_ERR
617ba13b 2069 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 2070 goto failed_mount;
2071 }
617ba13b 2072 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2073 sbi->s_es = es;
617ba13b 2074 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
ac27a0ec 2075 printk (KERN_ERR
617ba13b 2076 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 2077 goto failed_mount;
2078 }
2079 }
2080
e2b46574 2081 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
617ba13b 2082 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
ac27a0ec 2083
617ba13b
MC		 2084 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2085 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2086 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2087 } else {
2088 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2089 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2090 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2091 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec
DK		 2092 (sbi->s_inode_size > blocksize)) {
2093 printk (KERN_ERR
617ba13b 2094 "EXT4-fs: unsupported inode size: %d\n",
ac27a0ec
DK		 2095 sbi->s_inode_size);
2096 goto failed_mount;
2097 }
ef7f3835
KS		 2098 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2099 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2100 }
0d1ee42f
AR		 2101 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2102 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2103 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2104 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2105 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2106 printk(KERN_ERR
8fadc143 2107 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2108 sbi->s_desc_size);
2109 goto failed_mount;
2110 }
2111 } else
2112 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2113 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2114 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2115 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2116 goto cantfind_ext4;
2117 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2118 if (sbi->s_inodes_per_block == 0)
617ba13b 2119 goto cantfind_ext4;
ac27a0ec
DK		 2120 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2121 sbi->s_inodes_per_block;
0d1ee42f 2122 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2123 sbi->s_sbh = bh;
2124 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2125 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2126 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
ac27a0ec
DK		 2127 for (i=0; i < 4; i++)
2128 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2129 sbi->s_def_hash_version = es->s_def_hash_version;
2130
2131 if (sbi->s_blocks_per_group > blocksize * 8) {
2132 printk (KERN_ERR
617ba13b 2133 "EXT4-fs: #blocks per group too big: %lu\n",
ac27a0ec
DK		 2134 sbi->s_blocks_per_group);
2135 goto failed_mount;
2136 }
ac27a0ec
DK		 2137 if (sbi->s_inodes_per_group > blocksize * 8) {
2138 printk (KERN_ERR
617ba13b 2139 "EXT4-fs: #inodes per group too big: %lu\n",
ac27a0ec
DK		 2140 sbi->s_inodes_per_group);
2141 goto failed_mount;
2142 }
2143
bd81d8ee 2144 if (ext4_blocks_count(es) >
ac27a0ec 2145 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2146 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2147 " too large to mount safely\n", sb->s_id);
2148 if (sizeof(sector_t) < 8)
617ba13b 2149 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2150 "enabled\n");
2151 goto failed_mount;
2152 }
2153
617ba13b
MC		 2154 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2155 goto cantfind_ext4;
e7c95593
ES		 2156
2157 /* ensure blocks_count calculation below doesn't sign-extend */
2158 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2159 le32_to_cpu(es->s_first_data_block) + 1) {
2160 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2161 "first data block %u, blocks per group %lu\n",
2162 ext4_blocks_count(es),
2163 le32_to_cpu(es->s_first_data_block),
2164 EXT4_BLOCKS_PER_GROUP(sb));
2165 goto failed_mount;
2166 }
bd81d8ee
LV		 2167 blocks_count = (ext4_blocks_count(es) -
2168 le32_to_cpu(es->s_first_data_block) +
2169 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2170 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2171 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2172 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2173 EXT4_DESC_PER_BLOCK(sb);
ac27a0ec
DK		 2174 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2175 GFP_KERNEL);
2176 if (sbi->s_group_desc == NULL) {
617ba13b 2177 printk (KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2178 goto failed_mount;
2179 }
2180
2181 bgl_lock_init(&sbi->s_blockgroup_lock);
2182
2183 for (i = 0; i < db_count; i++) {
70bbb3e0 2184 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2185 sbi->s_group_desc[i] = sb_bread(sb, block);
2186 if (!sbi->s_group_desc[i]) {
617ba13b 2187 printk (KERN_ERR "EXT4-fs: "
ac27a0ec
DK		 2188 "can't read group descriptor %d\n", i);
2189 db_count = i;
2190 goto failed_mount2;
2191 }
2192 }
617ba13b
MC		 2193 if (!ext4_check_descriptors (sb)) {
2194 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2195 goto failed_mount2;
2196 }
772cb7c8
JS		 2197 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2198 if (!ext4_fill_flex_info(sb)) {
2199 printk(KERN_ERR
2200 "EXT4-fs: unable to initialize "
2201 "flex_bg meta info!\n");
2202 goto failed_mount2;
2203 }
2204
ac27a0ec
DK		 2205 sbi->s_gdb_count = db_count;
2206 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2207 spin_lock_init(&sbi->s_next_gen_lock);
2208
833f4077
PZ		 2209 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2210 ext4_count_free_blocks(sb));
2211 if (!err) {
2212 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2213 ext4_count_free_inodes(sb));
2214 }
2215 if (!err) {
2216 err = percpu_counter_init(&sbi->s_dirs_counter,
2217 ext4_count_dirs(sb));
2218 }
2219 if (err) {
2220 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2221 goto failed_mount3;
2222 }
ac27a0ec
DK		 2223
2224 /* per fileystem reservation list head & lock */
2225 spin_lock_init(&sbi->s_rsv_window_lock);
2226 sbi->s_rsv_window_root = RB_ROOT;
2227 /* Add a single, static dummy reservation to the start of the
2228 * reservation window list --- it gives us a placeholder for
2229 * append-at-start-of-list which makes the allocation logic
2230 * _much_ simpler. */
617ba13b
MC		 2231 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2232 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
ac27a0ec
DK		 2233 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2234 sbi->s_rsv_window_head.rsv_goal_size = 0;
617ba13b 2235 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
ac27a0ec 2236
c9de560d
AT		 2237 sbi->s_stripe = ext4_get_stripe_size(sbi);
2238
ac27a0ec
DK		 2239 /*
2240 * set up enough so that it can read an inode
2241 */
617ba13b
MC		 2242 sb->s_op = &ext4_sops;
2243 sb->s_export_op = &ext4_export_ops;
2244 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2245#ifdef CONFIG_QUOTA
617ba13b
MC		 2246 sb->s_qcop = &ext4_qctl_operations;
2247 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2248#endif
2249 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2250
2251 sb->s_root = NULL;
2252
2253 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2254 EXT4_HAS_INCOMPAT_FEATURE(sb,
2255 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2256
2257 /*
2258 * The first inode we look at is the journal inode. Don't try
2259 * root first: it may be modified in the journal!
2260 */
2261 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2262 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2263 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2264 goto failed_mount3;
624080ed
TT		 2265 if (!(sb->s_flags & MS_RDONLY) &&
2266 EXT4_SB(sb)->s_journal->j_failed_commit) {
2267 printk(KERN_CRIT "EXT4-fs error (device %s): "
2268 "ext4_fill_super: Journal transaction "
2269 "%u is corrupt\n", sb->s_id,
2270 EXT4_SB(sb)->s_journal->j_failed_commit);
2271 if (test_opt (sb, ERRORS_RO)) {
2272 printk (KERN_CRIT
2273 "Mounting filesystem read-only\n");
2274 sb->s_flags |= MS_RDONLY;
2275 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2276 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2277 }
2278 if (test_opt(sb, ERRORS_PANIC)) {
2279 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2280 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2281 ext4_commit_super(sb, es, 1);
2282 printk(KERN_CRIT
2283 "EXT4-fs (device %s): mount failed\n",
2284 sb->s_id);
2285 goto failed_mount4;
2286 }
2287 }
ac27a0ec 2288 } else if (journal_inum) {
617ba13b 2289 if (ext4_create_journal(sb, es, journal_inum))
ac27a0ec
DK		 2290 goto failed_mount3;
2291 } else {
2292 if (!silent)
2293 printk (KERN_ERR
617ba13b 2294 "ext4: No journal on filesystem on %s\n",
ac27a0ec
DK		 2295 sb->s_id);
2296 goto failed_mount3;
2297 }
2298
eb40a09c
JS		 2299 if (ext4_blocks_count(es) > 0xffffffffULL &&
2300 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2301 JBD2_FEATURE_INCOMPAT_64BIT)) {
2302 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2303 goto failed_mount4;
2304 }
2305
818d276c
GS		 2306 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2307 jbd2_journal_set_features(sbi->s_journal,
2308 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2309 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2310 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2311 jbd2_journal_set_features(sbi->s_journal,
2312 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2313 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2314 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2315 } else {
2316 jbd2_journal_clear_features(sbi->s_journal,
2317 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2318 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2319 }
2320
ac27a0ec
DK		 2321 /* We have now updated the journal if required, so we can
2322 * validate the data journaling mode. */
2323 switch (test_opt(sb, DATA_FLAGS)) {
2324 case 0:
2325 /* No mode set, assume a default based on the journal
63f57933
AM		 2326 * capabilities: ORDERED_DATA if the journal can
2327 * cope, else JOURNAL_DATA
2328 */
dab291af
MC		 2329 if (jbd2_journal_check_available_features
2330 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2331 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2332 else
2333 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2334 break;
2335
617ba13b
MC		 2336 case EXT4_MOUNT_ORDERED_DATA:
2337 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2338 if (!jbd2_journal_check_available_features
2339 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2340 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2341 "requested data journaling mode\n");
2342 goto failed_mount4;
2343 }
2344 default:
2345 break;
2346 }
2347
2348 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2349 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2350 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2351 "its supported only with writeback mode\n");
2352 clear_opt(sbi->s_mount_opt, NOBH);
2353 }
2354 }
2355 /*
dab291af 2356 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2357 * so we can safely mount the rest of the filesystem now.
2358 */
2359
1d1fe1ee
DH		 2360 root = ext4_iget(sb, EXT4_ROOT_INO);
2361 if (IS_ERR(root)) {
617ba13b 2362 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1d1fe1ee 2363 ret = PTR_ERR(root);
ac27a0ec
DK		 2364 goto failed_mount4;
2365 }
2366 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2367 iput(root);
617ba13b 2368 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2369 goto failed_mount4;
2370 }
1d1fe1ee
DH		 2371 sb->s_root = d_alloc_root(root);
2372 if (!sb->s_root) {
2373 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2374 iput(root);
2375 ret = -ENOMEM;
2376 goto failed_mount4;
2377 }
ac27a0ec 2378
617ba13b 2379 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2380
2381 /* determine the minimum size of new large inodes, if present */
2382 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2383 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2384 EXT4_GOOD_OLD_INODE_SIZE;
2385 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2386 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2387 if (sbi->s_want_extra_isize <
2388 le16_to_cpu(es->s_want_extra_isize))
2389 sbi->s_want_extra_isize =
2390 le16_to_cpu(es->s_want_extra_isize);
2391 if (sbi->s_want_extra_isize <
2392 le16_to_cpu(es->s_min_extra_isize))
2393 sbi->s_want_extra_isize =
2394 le16_to_cpu(es->s_min_extra_isize);
2395 }
2396 }
2397 /* Check if enough inode space is available */
2398 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2399 sbi->s_inode_size) {
2400 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2401 EXT4_GOOD_OLD_INODE_SIZE;
2402 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2403 "available.\n");
2404 }
2405
ac27a0ec
DK		 2406 /*
2407 * akpm: core read_super() calls in here with the superblock locked.
2408 * That deadlocks, because orphan cleanup needs to lock the superblock
2409 * in numerous places. Here we just pop the lock - it's relatively
2410 * harmless, because we are now ready to accept write_super() requests,
2411 * and aviro says that's the only reason for hanging onto the
2412 * superblock lock.
2413 */
617ba13b
MC		 2414 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2415 ext4_orphan_cleanup(sb, es);
2416 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
ac27a0ec 2417 if (needs_recovery)
617ba13b
MC		 2418 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2419 ext4_mark_recovery_complete(sb, es);
2420 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2421 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2422 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
ac27a0ec
DK		 2423 "writeback");
2424
a86c6181 2425 ext4_ext_init(sb);
c9de560d 2426 ext4_mb_init(sb, needs_recovery);
a86c6181 2427
ac27a0ec
DK		 2428 lock_kernel();
2429 return 0;
2430
617ba13b 2431cantfind_ext4:
ac27a0ec 2432 if (!silent)
617ba13b 2433 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2434 sb->s_id);
2435 goto failed_mount;
2436
2437failed_mount4:
dab291af 2438 jbd2_journal_destroy(sbi->s_journal);
47b4a50b 2439 sbi->s_journal = NULL;
ac27a0ec
DK		 2440failed_mount3:
2441 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2442 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2443 percpu_counter_destroy(&sbi->s_dirs_counter);
2444failed_mount2:
2445 for (i = 0; i < db_count; i++)
2446 brelse(sbi->s_group_desc[i]);
2447 kfree(sbi->s_group_desc);
2448failed_mount:
2449#ifdef CONFIG_QUOTA
2450 for (i = 0; i < MAXQUOTAS; i++)
2451 kfree(sbi->s_qf_names[i]);
2452#endif
617ba13b 2453 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2454 brelse(bh);
2455out_fail:
2456 sb->s_fs_info = NULL;
2457 kfree(sbi);
2458 lock_kernel();
1d1fe1ee 2459 return ret;
ac27a0ec
DK		 2460}
2461
2462/*
2463 * Setup any per-fs journal parameters now. We'll do this both on
2464 * initial mount, once the journal has been initialised but before we've
2465 * done any recovery; and again on any subsequent remount.
2466 */
617ba13b 2467static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2468{
617ba13b 2469 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2470
2471 if (sbi->s_commit_interval)
2472 journal->j_commit_interval = sbi->s_commit_interval;
617ba13b 2473 /* We could also set up an ext4-specific default for the commit
ac27a0ec
DK		 2474 * interval here, but for now we'll just fall back to the jbd
2475 * default. */
2476
2477 spin_lock(&journal->j_state_lock);
2478 if (test_opt(sb, BARRIER))
dab291af 2479 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2480 else
dab291af 2481 journal->j_flags &= ~JBD2_BARRIER;
ac27a0ec
DK		 2482 spin_unlock(&journal->j_state_lock);
2483}
2484
617ba13b 2485static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2486 unsigned int journal_inum)
2487{
2488 struct inode *journal_inode;
2489 journal_t *journal;
2490
2491 /* First, test for the existence of a valid inode on disk. Bad
2492 * things happen if we iget() an unused inode, as the subsequent
2493 * iput() will try to delete it. */
2494
1d1fe1ee
DH		 2495 journal_inode = ext4_iget(sb, journal_inum);
2496 if (IS_ERR(journal_inode)) {
617ba13b 2497 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2498 return NULL;
2499 }
2500 if (!journal_inode->i_nlink) {
2501 make_bad_inode(journal_inode);
2502 iput(journal_inode);
617ba13b 2503 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2504 return NULL;
2505 }
2506
2507 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2508 journal_inode, journal_inode->i_size);
1d1fe1ee 2509 if (!S_ISREG(journal_inode->i_mode)) {
617ba13b 2510 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2511 iput(journal_inode);
2512 return NULL;
2513 }
2514
dab291af 2515 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2516 if (!journal) {
617ba13b 2517 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2518 iput(journal_inode);
2519 return NULL;
2520 }
2521 journal->j_private = sb;
617ba13b 2522 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2523 return journal;
2524}
2525
617ba13b 2526static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2527 dev_t j_dev)
2528{
2529 struct buffer_head * bh;
2530 journal_t *journal;
617ba13b
MC		 2531 ext4_fsblk_t start;
2532 ext4_fsblk_t len;
ac27a0ec 2533 int hblock, blocksize;
617ba13b 2534 ext4_fsblk_t sb_block;
ac27a0ec 2535 unsigned long offset;
617ba13b 2536 struct ext4_super_block * es;
ac27a0ec
DK		 2537 struct block_device *bdev;
2538
617ba13b 2539 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2540 if (bdev == NULL)
2541 return NULL;
2542
2543 if (bd_claim(bdev, sb)) {
2544 printk(KERN_ERR
8c55e204 2545 "EXT4: failed to claim external journal device.\n");
ac27a0ec
DK		 2546 blkdev_put(bdev);
2547 return NULL;
2548 }
2549
2550 blocksize = sb->s_blocksize;
2551 hblock = bdev_hardsect_size(bdev);
2552 if (blocksize < hblock) {
2553 printk(KERN_ERR
617ba13b 2554 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2555 goto out_bdev;
2556 }
2557
617ba13b
MC		 2558 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2559 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2560 set_blocksize(bdev, blocksize);
2561 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2562 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2563 "external journal\n");
2564 goto out_bdev;
2565 }
2566
617ba13b
MC		 2567 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2568 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2569 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2570 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2571 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2572 "bad superblock\n");
2573 brelse(bh);
2574 goto out_bdev;
2575 }
2576
617ba13b
MC		 2577 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2578 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2579 brelse(bh);
2580 goto out_bdev;
2581 }
2582
bd81d8ee 2583 len = ext4_blocks_count(es);
ac27a0ec
DK		 2584 start = sb_block + 1;
2585 brelse(bh); /* we're done with the superblock */
2586
dab291af 2587 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2588 start, len, blocksize);
2589 if (!journal) {
617ba13b 2590 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2591 goto out_bdev;
2592 }
2593 journal->j_private = sb;
2594 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2595 wait_on_buffer(journal->j_sb_buffer);
2596 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2597 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2598 goto out_journal;
2599 }
2600 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2601 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2602 "user (unsupported) - %d\n",
2603 be32_to_cpu(journal->j_superblock->s_nr_users));
2604 goto out_journal;
2605 }
617ba13b
MC		 2606 EXT4_SB(sb)->journal_bdev = bdev;
2607 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2608 return journal;
2609out_journal:
dab291af 2610 jbd2_journal_destroy(journal);
ac27a0ec 2611out_bdev:
617ba13b 2612 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2613 return NULL;
2614}
2615
617ba13b
MC		 2616static int ext4_load_journal(struct super_block *sb,
2617 struct ext4_super_block *es,
ac27a0ec
DK		 2618 unsigned long journal_devnum)
2619{
2620 journal_t *journal;
2621 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2622 dev_t journal_dev;
2623 int err = 0;
2624 int really_read_only;
2625
2626 if (journal_devnum &&
2627 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2628 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2629 "numbers have changed\n");
2630 journal_dev = new_decode_dev(journal_devnum);
2631 } else
2632 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2633
2634 really_read_only = bdev_read_only(sb->s_bdev);
2635
2636 /*
2637 * Are we loading a blank journal or performing recovery after a
2638 * crash? For recovery, we need to check in advance whether we
2639 * can get read-write access to the device.
2640 */
2641
617ba13b 2642 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2643 if (sb->s_flags & MS_RDONLY) {
617ba13b 2644 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2645 "required on readonly filesystem.\n");
2646 if (really_read_only) {
617ba13b 2647 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2648 "unavailable, cannot proceed.\n");
2649 return -EROFS;
2650 }
617ba13b 2651 printk (KERN_INFO "EXT4-fs: write access will "
ac27a0ec
DK		 2652 "be enabled during recovery.\n");
2653 }
2654 }
2655
2656 if (journal_inum && journal_dev) {
617ba13b 2657 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2658 "and inode journals!\n");
2659 return -EINVAL;
2660 }
2661
2662 if (journal_inum) {
617ba13b 2663 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2664 return -EINVAL;
2665 } else {
617ba13b 2666 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2667 return -EINVAL;
2668 }
2669
2670 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2671 err = jbd2_journal_update_format(journal);
ac27a0ec 2672 if (err) {
617ba13b 2673 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2674 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2675 return err;
2676 }
2677 }
2678
617ba13b 2679 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2680 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2681 if (!err)
dab291af 2682 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2683
2684 if (err) {
617ba13b 2685 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2686 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2687 return err;
2688 }
2689
617ba13b
MC		 2690 EXT4_SB(sb)->s_journal = journal;
2691 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2692
2693 if (journal_devnum &&
2694 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2695 es->s_journal_dev = cpu_to_le32(journal_devnum);
2696 sb->s_dirt = 1;
2697
2698 /* Make sure we flush the recovery flag to disk. */
617ba13b 2699 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2700 }
2701
2702 return 0;
2703}
2704
617ba13b
MC		 2705static int ext4_create_journal(struct super_block * sb,
2706 struct ext4_super_block * es,
ac27a0ec
DK		 2707 unsigned int journal_inum)
2708{
2709 journal_t *journal;
6c675bd4 2710 int err;
ac27a0ec
DK		 2711
2712 if (sb->s_flags & MS_RDONLY) {
617ba13b 2713 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
ac27a0ec
DK		 2714 "create journal.\n");
2715 return -EROFS;
2716 }
2717
6c675bd4
BP		 2718 journal = ext4_get_journal(sb, journal_inum);
2719 if (!journal)
ac27a0ec
DK		 2720 return -EINVAL;
2721
617ba13b 2722 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
ac27a0ec
DK		 2723 journal_inum);
2724
6c675bd4
BP		 2725 err = jbd2_journal_create(journal);
2726 if (err) {
617ba13b 2727 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
dab291af 2728 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2729 return -EIO;
2730 }
2731
617ba13b 2732 EXT4_SB(sb)->s_journal = journal;
ac27a0ec 2733
617ba13b
MC		 2734 ext4_update_dynamic_rev(sb);
2735 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2736 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
ac27a0ec
DK		 2737
2738 es->s_journal_inum = cpu_to_le32(journal_inum);
2739 sb->s_dirt = 1;
2740
2741 /* Make sure we flush the recovery flag to disk. */
617ba13b 2742 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2743
2744 return 0;
2745}
2746
617ba13b
MC		 2747static void ext4_commit_super (struct super_block * sb,
2748 struct ext4_super_block * es,
ac27a0ec
DK		 2749 int sync)
2750{
617ba13b 2751 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
ac27a0ec
DK		 2752
2753 if (!sbh)
2754 return;
2755 es->s_wtime = cpu_to_le32(get_seconds());
bd81d8ee 2756 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
617ba13b 2757 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2758 BUFFER_TRACE(sbh, "marking dirty");
2759 mark_buffer_dirty(sbh);
2760 if (sync)
2761 sync_dirty_buffer(sbh);
2762}
2763
2764
2765/*
2766 * Have we just finished recovery? If so, and if we are mounting (or
2767 * remounting) the filesystem readonly, then we will end up with a
2768 * consistent fs on disk. Record that fact.
2769 */
617ba13b
MC		 2770static void ext4_mark_recovery_complete(struct super_block * sb,
2771 struct ext4_super_block * es)
ac27a0ec 2772{
617ba13b 2773 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2774
dab291af
MC		 2775 jbd2_journal_lock_updates(journal);
2776 jbd2_journal_flush(journal);
32c37730 2777 lock_super(sb);
617ba13b 2778 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2779 sb->s_flags & MS_RDONLY) {
617ba13b 2780 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2781 sb->s_dirt = 0;
617ba13b 2782 ext4_commit_super(sb, es, 1);
ac27a0ec 2783 }
32c37730 2784 unlock_super(sb);
dab291af 2785 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2786}
2787
2788/*
2789 * If we are mounting (or read-write remounting) a filesystem whose journal
2790 * has recorded an error from a previous lifetime, move that error to the
2791 * main filesystem now.
2792 */
617ba13b
MC		 2793static void ext4_clear_journal_err(struct super_block * sb,
2794 struct ext4_super_block * es)
ac27a0ec
DK		 2795{
2796 journal_t *journal;
2797 int j_errno;
2798 const char *errstr;
2799
617ba13b 2800 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2801
2802 /*
2803 * Now check for any error status which may have been recorded in the
617ba13b 2804 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2805 */
2806
dab291af 2807 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2808 if (j_errno) {
2809 char nbuf[16];
2810
617ba13b 2811 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 2812 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 2813 "from previous mount: %s", errstr);
46e665e9 2814 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 2815 "filesystem check.");
2816
617ba13b
MC		 2817 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2818 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2819 ext4_commit_super (sb, es, 1);
ac27a0ec 2820
dab291af 2821 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 2822 }
2823}
2824
2825/*
2826 * Force the running and committing transactions to commit,
2827 * and wait on the commit.
2828 */
617ba13b 2829int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 2830{
2831 journal_t *journal;
2832 int ret;
2833
2834 if (sb->s_flags & MS_RDONLY)
2835 return 0;
2836
617ba13b 2837 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2838 sb->s_dirt = 0;
617ba13b 2839 ret = ext4_journal_force_commit(journal);
ac27a0ec
DK		 2840 return ret;
2841}
2842
2843/*
617ba13b 2844 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec
DK		 2845 * have to propagate any other updates to the superblock on disk at this
2846 * point. Just start an async writeback to get the buffers on their way
2847 * to the disk.
2848 *
2849 * This implicitly triggers the writebehind on sync().
2850 */
2851
617ba13b 2852static void ext4_write_super (struct super_block * sb)
ac27a0ec
DK		 2853{
2854 if (mutex_trylock(&sb->s_lock) != 0)
2855 BUG();
2856 sb->s_dirt = 0;
2857}
2858
617ba13b 2859static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec
DK		 2860{
2861 tid_t target;
2862
2863 sb->s_dirt = 0;
dab291af 2864 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
ac27a0ec 2865 if (wait)
dab291af 2866 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
ac27a0ec
DK		 2867 }
2868 return 0;
2869}
2870
2871/*
2872 * LVM calls this function before a (read-only) snapshot is created. This
2873 * gives us a chance to flush the journal completely and mark the fs clean.
2874 */
617ba13b 2875static void ext4_write_super_lockfs(struct super_block *sb)
ac27a0ec
DK		 2876{
2877 sb->s_dirt = 0;
2878
2879 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 2880 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2881
2882 /* Now we set up the journal barrier. */
dab291af
MC		 2883 jbd2_journal_lock_updates(journal);
2884 jbd2_journal_flush(journal);
ac27a0ec
DK		 2885
2886 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 2887 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2888 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec
DK		 2889 }
2890}
2891
2892/*
2893 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2894 * flag here, even though the filesystem is not technically dirty yet.
2895 */
617ba13b 2896static void ext4_unlockfs(struct super_block *sb)
ac27a0ec
DK		 2897{
2898 if (!(sb->s_flags & MS_RDONLY)) {
2899 lock_super(sb);
2900 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 2901 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2902 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 2903 unlock_super(sb);
dab291af 2904 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 2905 }
2906}
2907
617ba13b 2908static int ext4_remount (struct super_block * sb, int * flags, char * data)
ac27a0ec 2909{
617ba13b
MC		 2910 struct ext4_super_block * es;
2911 struct ext4_sb_info *sbi = EXT4_SB(sb);
2912 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 2913 unsigned long old_sb_flags;
617ba13b 2914 struct ext4_mount_options old_opts;
ac27a0ec
DK		 2915 int err;
2916#ifdef CONFIG_QUOTA
2917 int i;
2918#endif
2919
2920 /* Store the original options */
2921 old_sb_flags = sb->s_flags;
2922 old_opts.s_mount_opt = sbi->s_mount_opt;
2923 old_opts.s_resuid = sbi->s_resuid;
2924 old_opts.s_resgid = sbi->s_resgid;
2925 old_opts.s_commit_interval = sbi->s_commit_interval;
2926#ifdef CONFIG_QUOTA
2927 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2928 for (i = 0; i < MAXQUOTAS; i++)
2929 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2930#endif
2931
2932 /*
2933 * Allow the "check" option to be passed as a remount option.
2934 */
2935 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2936 err = -EINVAL;
2937 goto restore_opts;
2938 }
2939
617ba13b 2940 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
46e665e9 2941 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 2942
2943 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2944 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 2945
2946 es = sbi->s_es;
2947
617ba13b 2948 ext4_init_journal_params(sb, sbi->s_journal);
ac27a0ec
DK		 2949
2950 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 2951 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 2952 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 2953 err = -EROFS;
2954 goto restore_opts;
2955 }
2956
2957 if (*flags & MS_RDONLY) {
2958 /*
2959 * First of all, the unconditional stuff we have to do
2960 * to disable replay of the journal when we next remount
2961 */
2962 sb->s_flags |= MS_RDONLY;
2963
2964 /*
2965 * OK, test if we are remounting a valid rw partition
2966 * readonly, and if so set the rdonly flag and then
2967 * mark the partition as valid again.
2968 */
617ba13b
MC		 2969 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2970 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 2971 es->s_state = cpu_to_le16(sbi->s_mount_state);
2972
32c37730
JK		 2973 /*
2974 * We have to unlock super so that we can wait for
2975 * transactions.
2976 */
2977 unlock_super(sb);
617ba13b 2978 ext4_mark_recovery_complete(sb, es);
32c37730 2979 lock_super(sb);
ac27a0ec
DK		 2980 } else {
2981 __le32 ret;
617ba13b
MC		 2982 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2983 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2984 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec
DK		 2985 "remount RDWR because of unsupported "
2986 "optional features (%x).\n",
2987 sb->s_id, le32_to_cpu(ret));
2988 err = -EROFS;
2989 goto restore_opts;
2990 }
ead6596b
ES		 2991
2992 /*
2993 * If we have an unprocessed orphan list hanging
2994 * around from a previously readonly bdev mount,
2995 * require a full umount/remount for now.
2996 */
2997 if (es->s_last_orphan) {
2998 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2999 "remount RDWR because of unprocessed "
3000 "orphan inode list. Please "
3001 "umount/remount instead.\n",
3002 sb->s_id);
3003 err = -EINVAL;
3004 goto restore_opts;
3005 }
3006
ac27a0ec
DK		 3007 /*
3008 * Mounting a RDONLY partition read-write, so reread
3009 * and store the current valid flag. (It may have
3010 * been changed by e2fsck since we originally mounted
3011 * the partition.)
3012 */
617ba13b 3013 ext4_clear_journal_err(sb, es);
ac27a0ec 3014 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3015 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3016 goto restore_opts;
617ba13b 3017 if (!ext4_setup_super (sb, es, 0))
ac27a0ec
DK		 3018 sb->s_flags &= ~MS_RDONLY;
3019 }
3020 }
3021#ifdef CONFIG_QUOTA
3022 /* Release old quota file names */
3023 for (i = 0; i < MAXQUOTAS; i++)
3024 if (old_opts.s_qf_names[i] &&
3025 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3026 kfree(old_opts.s_qf_names[i]);
3027#endif
3028 return 0;
3029restore_opts:
3030 sb->s_flags = old_sb_flags;
3031 sbi->s_mount_opt = old_opts.s_mount_opt;
3032 sbi->s_resuid = old_opts.s_resuid;
3033 sbi->s_resgid = old_opts.s_resgid;
3034 sbi->s_commit_interval = old_opts.s_commit_interval;
3035#ifdef CONFIG_QUOTA
3036 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3037 for (i = 0; i < MAXQUOTAS; i++) {
3038 if (sbi->s_qf_names[i] &&
3039 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3040 kfree(sbi->s_qf_names[i]);
3041 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3042 }
3043#endif
3044 return err;
3045}
3046
617ba13b 3047static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
ac27a0ec
DK		 3048{
3049 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3050 struct ext4_sb_info *sbi = EXT4_SB(sb);
3051 struct ext4_super_block *es = sbi->s_es;
960cc398 3052 u64 fsid;
ac27a0ec 3053
5e70030d
BP		 3054 if (test_opt(sb, MINIX_DF)) {
3055 sbi->s_overhead_last = 0;
6bc9feff 3056 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 3057 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 3058 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3059 smp_rmb();
3060
3061 /*
5e70030d
BP		 3062 * Compute the overhead (FS structures). This is constant
3063 * for a given filesystem unless the number of block groups
3064 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3065 */
3066
3067 /*
3068 * All of the blocks before first_data_block are
3069 * overhead
3070 */
3071 overhead = le32_to_cpu(es->s_first_data_block);
3072
3073 /*
3074 * Add the overhead attributed to the superblock and
3075 * block group descriptors. If the sparse superblocks
3076 * feature is turned on, then not all groups have this.
3077 */
3078 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3079 overhead += ext4_bg_has_super(sb, i) +
3080 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3081 cond_resched();
3082 }
3083
3084 /*
3085 * Every block group has an inode bitmap, a block
3086 * bitmap, and an inode table.
3087 */
5e70030d
BP		 3088 overhead += ngroups * (2 + sbi->s_itb_per_group);
3089 sbi->s_overhead_last = overhead;
3090 smp_wmb();
6bc9feff 3091 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3092 }
3093
617ba13b 3094 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3095 buf->f_bsize = sb->s_blocksize;
5e70030d 3096 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
52d9f3b4 3097 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
308ba3ec 3098 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3099 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3100 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3101 buf->f_bavail = 0;
3102 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3103 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3104 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3105 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3106 fsid = le64_to_cpup((void *)es->s_uuid) ^
3107 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3108 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3109 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 3110 return 0;
3111}
3112
3113/* Helper function for writing quotas on sync - we need to start transaction before quota file
3114 * is locked for write. Otherwise the are possible deadlocks:
3115 * Process 1 Process 2
617ba13b 3116 * ext4_create() quota_sync()
dab291af 3117 * jbd2_journal_start() write_dquot()
ac27a0ec 3118 * DQUOT_INIT() down(dqio_mutex)
dab291af 3119 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3120 *
3121 */
3122
3123#ifdef CONFIG_QUOTA
3124
3125static inline struct inode *dquot_to_inode(struct dquot *dquot)
3126{
3127 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3128}
3129
617ba13b 3130static int ext4_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 3131{
3132 handle_t *handle;
3133 int ret, err;
3134
3135 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 3136 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 3137 if (IS_ERR(handle))
3138 return PTR_ERR(handle);
3139 ret = dquot_initialize(inode, type);
617ba13b 3140 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3141 if (!ret)
3142 ret = err;
3143 return ret;
3144}
3145
617ba13b 3146static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 3147{
3148 handle_t *handle;
3149 int ret, err;
3150
3151 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 3152 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2887df13
JK		 3153 if (IS_ERR(handle)) {
3154 /*
3155 * We call dquot_drop() anyway to at least release references
3156 * to quota structures so that umount does not hang.
3157 */
3158 dquot_drop(inode);
ac27a0ec 3159 return PTR_ERR(handle);
2887df13 3160 }
ac27a0ec 3161 ret = dquot_drop(inode);
617ba13b 3162 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3163 if (!ret)
3164 ret = err;
3165 return ret;
3166}
3167
617ba13b 3168static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3169{
3170 int ret, err;
3171 handle_t *handle;
3172 struct inode *inode;
3173
3174 inode = dquot_to_inode(dquot);
617ba13b
MC		 3175 handle = ext4_journal_start(inode,
3176 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3177 if (IS_ERR(handle))
3178 return PTR_ERR(handle);
3179 ret = dquot_commit(dquot);
617ba13b 3180 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3181 if (!ret)
3182 ret = err;
3183 return ret;
3184}
3185
617ba13b 3186static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3187{
3188 int ret, err;
3189 handle_t *handle;
3190
617ba13b
MC		 3191 handle = ext4_journal_start(dquot_to_inode(dquot),
3192 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3193 if (IS_ERR(handle))
3194 return PTR_ERR(handle);
3195 ret = dquot_acquire(dquot);
617ba13b 3196 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3197 if (!ret)
3198 ret = err;
3199 return ret;
3200}
3201
617ba13b 3202static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3203{
3204 int ret, err;
3205 handle_t *handle;
3206
617ba13b
MC		 3207 handle = ext4_journal_start(dquot_to_inode(dquot),
3208 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3209 if (IS_ERR(handle)) {
3210 /* Release dquot anyway to avoid endless cycle in dqput() */
3211 dquot_release(dquot);
ac27a0ec 3212 return PTR_ERR(handle);
9c3013e9 3213 }
ac27a0ec 3214 ret = dquot_release(dquot);
617ba13b 3215 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3216 if (!ret)
3217 ret = err;
3218 return ret;
3219}
3220
617ba13b 3221static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3222{
2c8be6b2 3223 /* Are we journaling quotas? */
617ba13b
MC		 3224 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3225 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3226 dquot_mark_dquot_dirty(dquot);
617ba13b 3227 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3228 } else {
3229 return dquot_mark_dquot_dirty(dquot);
3230 }
3231}
3232
617ba13b 3233static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3234{
3235 int ret, err;
3236 handle_t *handle;
3237
3238 /* Data block + inode block */
617ba13b 3239 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3240 if (IS_ERR(handle))
3241 return PTR_ERR(handle);
3242 ret = dquot_commit_info(sb, type);
617ba13b 3243 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3244 if (!ret)
3245 ret = err;
3246 return ret;
3247}
3248
3249/*
3250 * Turn on quotas during mount time - we need to find
3251 * the quota file and such...
3252 */
617ba13b 3253static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3254{
617ba13b
MC		 3255 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3256 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3257}
3258
3259/*
3260 * Standard function to be called on quota_on
3261 */
617ba13b 3262static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6f28e087 3263 char *path, int remount)
ac27a0ec
DK		 3264{
3265 int err;
3266 struct nameidata nd;
3267
3268 if (!test_opt(sb, QUOTA))
3269 return -EINVAL;
0623543b
JK		 3270 /* When remounting, no checks are needed and in fact, path is NULL */
3271 if (remount)
6f28e087 3272 return vfs_quota_on(sb, type, format_id, path, remount);
0623543b 3273
ac27a0ec
DK		 3274 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3275 if (err)
3276 return err;
0623543b 3277
ac27a0ec 3278 /* Quotafile not on the same filesystem? */
4ac91378 3279 if (nd.path.mnt->mnt_sb != sb) {
1d957f9b 3280 path_put(&nd.path);
ac27a0ec
DK		 3281 return -EXDEV;
3282 }
0623543b
JK		 3283 /* Journaling quota? */
3284 if (EXT4_SB(sb)->s_qf_names[type]) {
3285 /* Quotafile not of fs root? */
3286 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3287 printk(KERN_WARNING
3288 "EXT4-fs: Quota file not on filesystem root. "
3289 "Journaled quota will not work.\n");
3290 }
3291
3292 /*
3293 * When we journal data on quota file, we have to flush journal to see
3294 * all updates to the file when we bypass pagecache...
3295 */
3296 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3297 /*
3298 * We don't need to lock updates but journal_flush() could
3299 * otherwise be livelocked...
3300 */
3301 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3302 jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3303 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3304 }
3305
1d957f9b 3306 path_put(&nd.path);
6f28e087 3307 return vfs_quota_on(sb, type, format_id, path, remount);
ac27a0ec
DK		 3308}
3309
3310/* Read data from quotafile - avoid pagecache and such because we cannot afford
3311 * acquiring the locks... As quota files are never truncated and quota code
3312 * itself serializes the operations (and noone else should touch the files)
3313 * we don't have to be afraid of races */
617ba13b 3314static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3315 size_t len, loff_t off)
3316{
3317 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3318 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3319 int err = 0;
3320 int offset = off & (sb->s_blocksize - 1);
3321 int tocopy;
3322 size_t toread;
3323 struct buffer_head *bh;
3324 loff_t i_size = i_size_read(inode);
3325
3326 if (off > i_size)
3327 return 0;
3328 if (off+len > i_size)
3329 len = i_size-off;
3330 toread = len;
3331 while (toread > 0) {
3332 tocopy = sb->s_blocksize - offset < toread ?
3333 sb->s_blocksize - offset : toread;
617ba13b 3334 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3335 if (err)
3336 return err;
3337 if (!bh) /* A hole? */
3338 memset(data, 0, tocopy);
3339 else
3340 memcpy(data, bh->b_data+offset, tocopy);
3341 brelse(bh);
3342 offset = 0;
3343 toread -= tocopy;
3344 data += tocopy;
3345 blk++;
3346 }
3347 return len;
3348}
3349
3350/* Write to quotafile (we know the transaction is already started and has
3351 * enough credits) */
617ba13b 3352static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3353 const char *data, size_t len, loff_t off)
3354{
3355 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3356 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3357 int err = 0;
3358 int offset = off & (sb->s_blocksize - 1);
3359 int tocopy;
617ba13b 3360 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3361 size_t towrite = len;
3362 struct buffer_head *bh;
3363 handle_t *handle = journal_current_handle();
3364
9c3013e9
JK		 3365 if (!handle) {
3366 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3367 " cancelled because transaction is not started.\n",
3368 (unsigned long long)off, (unsigned long long)len);
3369 return -EIO;
3370 }
ac27a0ec
DK		 3371 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3372 while (towrite > 0) {
3373 tocopy = sb->s_blocksize - offset < towrite ?
3374 sb->s_blocksize - offset : towrite;
617ba13b 3375 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3376 if (!bh)
3377 goto out;
3378 if (journal_quota) {
617ba13b 3379 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3380 if (err) {
3381 brelse(bh);
3382 goto out;
3383 }
3384 }
3385 lock_buffer(bh);
3386 memcpy(bh->b_data+offset, data, tocopy);
3387 flush_dcache_page(bh->b_page);
3388 unlock_buffer(bh);
3389 if (journal_quota)
617ba13b 3390 err = ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK		 3391 else {
3392 /* Always do at least ordered writes for quotas */
678aaf48 3393 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3394 mark_buffer_dirty(bh);
3395 }
3396 brelse(bh);
3397 if (err)
3398 goto out;
3399 offset = 0;
3400 towrite -= tocopy;
3401 data += tocopy;
3402 blk++;
3403 }
3404out:
4d04e4fb
JK		 3405 if (len == towrite) {
3406 mutex_unlock(&inode->i_mutex);
ac27a0ec 3407 return err;
4d04e4fb 3408 }
ac27a0ec
DK		 3409 if (inode->i_size < off+len-towrite) {
3410 i_size_write(inode, off+len-towrite);
617ba13b 3411 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3412 }
ac27a0ec 3413 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3414 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3415 mutex_unlock(&inode->i_mutex);
3416 return len - towrite;
3417}
3418
3419#endif
3420
617ba13b 3421static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3422 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3423{
617ba13b 3424 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3425}
3426
617ba13b 3427static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3428 .owner = THIS_MODULE,
617ba13b
MC		 3429 .name = "ext4dev",
3430 .get_sb = ext4_get_sb,
ac27a0ec
DK		 3431 .kill_sb = kill_block_super,
3432 .fs_flags = FS_REQUIRES_DEV,
3433};
3434
617ba13b 3435static int __init init_ext4_fs(void)
ac27a0ec 3436{
c9de560d
AT		 3437 int err;
3438
3439 err = init_ext4_mballoc();
ac27a0ec
DK		 3440 if (err)
3441 return err;
c9de560d
AT		 3442
3443 err = init_ext4_xattr();
3444 if (err)
3445 goto out2;
ac27a0ec
DK		 3446 err = init_inodecache();
3447 if (err)
3448 goto out1;
63f57933 3449 err = register_filesystem(&ext4dev_fs_type);
ac27a0ec
DK		 3450 if (err)
3451 goto out;
3452 return 0;
3453out:
3454 destroy_inodecache();
3455out1:
617ba13b 3456 exit_ext4_xattr();
c9de560d
AT		 3457out2:
3458 exit_ext4_mballoc();
ac27a0ec
DK		 3459 return err;
3460}
3461
617ba13b 3462static void __exit exit_ext4_fs(void)
ac27a0ec 3463{
617ba13b 3464 unregister_filesystem(&ext4dev_fs_type);
ac27a0ec 3465 destroy_inodecache();
617ba13b 3466 exit_ext4_xattr();
c9de560d 3467 exit_ext4_mballoc();
ac27a0ec
DK		 3468}
3469
3470MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
617ba13b 3471MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
ac27a0ec 3472MODULE_LICENSE("GPL");
617ba13b
MC		 3473module_init(init_ext4_fs)
3474module_exit(exit_ext4_fs)
Clone of davem's net-next-2.6 tree