projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: fix synchronization of quota files in journal=data mode
[GitHub/mt8127/android_kernel_alcatel_ttab.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);
ac27a0ec 509 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 510 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec
DK		 511 es->s_state = cpu_to_le16(sbi->s_mount_state);
512 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
513 mark_buffer_dirty(sbi->s_sbh);
617ba13b 514 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 515 }
516
517 for (i = 0; i < sbi->s_gdb_count; i++)
518 brelse(sbi->s_group_desc[i]);
519 kfree(sbi->s_group_desc);
520 percpu_counter_destroy(&sbi->s_freeblocks_counter);
521 percpu_counter_destroy(&sbi->s_freeinodes_counter);
522 percpu_counter_destroy(&sbi->s_dirs_counter);
523 brelse(sbi->s_sbh);
524#ifdef CONFIG_QUOTA
525 for (i = 0; i < MAXQUOTAS; i++)
526 kfree(sbi->s_qf_names[i]);
527#endif
528
529 /* Debugging code just in case the in-memory inode orphan list
530 * isn't empty. The on-disk one can be non-empty if we've
531 * detected an error and taken the fs readonly, but the
532 * in-memory list had better be clean by this point. */
533 if (!list_empty(&sbi->s_orphan))
534 dump_orphan_list(sb, sbi);
535 J_ASSERT(list_empty(&sbi->s_orphan));
536
f98393a6 537 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 538 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
539 /*
540 * Invalidate the journal device's buffers. We don't want them
541 * floating about in memory - the physical journal device may
542 * hotswapped, and it breaks the `ro-after' testing code.
543 */
544 sync_blockdev(sbi->journal_bdev);
f98393a6 545 invalidate_bdev(sbi->journal_bdev);
617ba13b 546 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 547 }
548 sb->s_fs_info = NULL;
549 kfree(sbi);
550 return;
551}
552
e18b890b 553static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 554
555/*
556 * Called inside transaction, so use GFP_NOFS
557 */
617ba13b 558static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 559{
617ba13b 560 struct ext4_inode_info *ei;
ac27a0ec 561
e6b4f8da 562 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 563 if (!ei)
564 return NULL;
617ba13b
MC		 565#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
566 ei->i_acl = EXT4_ACL_NOT_CACHED;
567 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 568#endif
569 ei->i_block_alloc_info = NULL;
570 ei->vfs_inode.i_version = 1;
a86c6181 571 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 572 INIT_LIST_HEAD(&ei->i_prealloc_list);
573 spin_lock_init(&ei->i_prealloc_lock);
ac27a0ec
DK		 574 return &ei->vfs_inode;
575}
576
617ba13b 577static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 578{
9f7dd93d
VA		 579 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
580 printk("EXT4 Inode %p: orphan list check failed!\n",
581 EXT4_I(inode));
582 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
583 EXT4_I(inode), sizeof(struct ext4_inode_info),
584 true);
585 dump_stack();
586 }
617ba13b 587 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 588}
589
4ba9b9d0 590static void init_once(struct kmem_cache *cachep, void *foo)
ac27a0ec 591{
617ba13b 592 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 593
a35afb83 594 INIT_LIST_HEAD(&ei->i_orphan);
617ba13b 595#ifdef CONFIG_EXT4DEV_FS_XATTR
a35afb83 596 init_rwsem(&ei->xattr_sem);
ac27a0ec 597#endif
0e855ac8 598 init_rwsem(&ei->i_data_sem);
a35afb83 599 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 600}
601
602static int init_inodecache(void)
603{
617ba13b
MC		 604 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
605 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 606 0, (SLAB_RECLAIM_ACCOUNT|
607 SLAB_MEM_SPREAD),
20c2df83 608 init_once);
617ba13b 609 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 610 return -ENOMEM;
611 return 0;
612}
613
614static void destroy_inodecache(void)
615{
617ba13b 616 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 617}
618
617ba13b 619static void ext4_clear_inode(struct inode *inode)
ac27a0ec 620{
617ba13b
MC		 621 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
622#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
623 if (EXT4_I(inode)->i_acl &&
624 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
625 posix_acl_release(EXT4_I(inode)->i_acl);
626 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
627 }
628 if (EXT4_I(inode)->i_default_acl &&
629 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
630 posix_acl_release(EXT4_I(inode)->i_default_acl);
631 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 632 }
633#endif
617ba13b
MC		 634 ext4_discard_reservation(inode);
635 EXT4_I(inode)->i_block_alloc_info = NULL;
ac27a0ec
DK		 636 if (unlikely(rsv))
637 kfree(rsv);
638}
639
617ba13b 640static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
ac27a0ec
DK		 641{
642#if defined(CONFIG_QUOTA)
617ba13b 643 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 644
645 if (sbi->s_jquota_fmt)
646 seq_printf(seq, ",jqfmt=%s",
647 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
648
649 if (sbi->s_qf_names[USRQUOTA])
650 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
651
652 if (sbi->s_qf_names[GRPQUOTA])
653 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
654
617ba13b 655 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 656 seq_puts(seq, ",usrquota");
657
617ba13b 658 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 659 seq_puts(seq, ",grpquota");
660#endif
661}
662
d9c9bef1
MS		 663/*
664 * Show an option if
665 * - it's set to a non-default value OR
666 * - if the per-sb default is different from the global default
667 */
617ba13b 668static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 669{
aa22df2c
AK		 670 int def_errors;
671 unsigned long def_mount_opts;
ac27a0ec 672 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 673 struct ext4_sb_info *sbi = EXT4_SB(sb);
674 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 675
676 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 677 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 678
679 if (sbi->s_sb_block != 1)
680 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
681 if (test_opt(sb, MINIX_DF))
682 seq_puts(seq, ",minixdf");
aa22df2c 683 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 684 seq_puts(seq, ",grpid");
685 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
686 seq_puts(seq, ",nogrpid");
687 if (sbi->s_resuid != EXT4_DEF_RESUID ||
688 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
689 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
690 }
691 if (sbi->s_resgid != EXT4_DEF_RESGID ||
692 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
693 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
694 }
bb4f397a 695 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 696 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 697 def_errors == EXT4_ERRORS_CONTINUE) {
698 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 699 }
700 }
aa22df2c 701 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 702 seq_puts(seq, ",errors=continue");
aa22df2c 703 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 704 seq_puts(seq, ",errors=panic");
aa22df2c 705 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 706 seq_puts(seq, ",nouid32");
aa22df2c 707 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 708 seq_puts(seq, ",debug");
709 if (test_opt(sb, OLDALLOC))
710 seq_puts(seq, ",oldalloc");
07620f69 711#ifdef CONFIG_EXT4DEV_FS_XATTR
aa22df2c
AK		 712 if (test_opt(sb, XATTR_USER) &&
713 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 714 seq_puts(seq, ",user_xattr");
715 if (!test_opt(sb, XATTR_USER) &&
716 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
717 seq_puts(seq, ",nouser_xattr");
718 }
719#endif
07620f69 720#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
aa22df2c 721 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 722 seq_puts(seq, ",acl");
723 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
724 seq_puts(seq, ",noacl");
725#endif
726 if (!test_opt(sb, RESERVATION))
727 seq_puts(seq, ",noreservation");
728 if (sbi->s_commit_interval) {
729 seq_printf(seq, ",commit=%u",
730 (unsigned) (sbi->s_commit_interval / HZ));
731 }
732 if (test_opt(sb, BARRIER))
733 seq_puts(seq, ",barrier=1");
734 if (test_opt(sb, NOBH))
735 seq_puts(seq, ",nobh");
736 if (!test_opt(sb, EXTENTS))
737 seq_puts(seq, ",noextents");
3dbd0ede
AK		 738 if (!test_opt(sb, MBALLOC))
739 seq_puts(seq, ",nomballoc");
25ec56b5
JNC		 740 if (test_opt(sb, I_VERSION))
741 seq_puts(seq, ",i_version");
ac27a0ec 742
cb45bbe4
MS		 743 if (sbi->s_stripe)
744 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 745 /*
746 * journal mode get enabled in different ways
747 * So just print the value even if we didn't specify it
748 */
617ba13b 749 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 750 seq_puts(seq, ",data=journal");
617ba13b 751 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 752 seq_puts(seq, ",data=ordered");
617ba13b 753 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 754 seq_puts(seq, ",data=writeback");
755
617ba13b 756 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 757 return 0;
758}
759
760
1b961ac0
CH		 761static struct inode *ext4_nfs_get_inode(struct super_block *sb,
762 u64 ino, u32 generation)
ac27a0ec 763{
ac27a0ec 764 struct inode *inode;
ac27a0ec 765
617ba13b 766 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 767 return ERR_PTR(-ESTALE);
617ba13b 768 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 769 return ERR_PTR(-ESTALE);
770
771 /* iget isn't really right if the inode is currently unallocated!!
772 *
617ba13b 773 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 774 * deleted, so we should be safe.
775 *
776 * Currently we don't know the generation for parent directory, so
777 * a generation of 0 means "accept any"
778 */
1d1fe1ee
DH		 779 inode = ext4_iget(sb, ino);
780 if (IS_ERR(inode))
781 return ERR_CAST(inode);
782 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 783 iput(inode);
784 return ERR_PTR(-ESTALE);
785 }
1b961ac0
CH		 786
787 return inode;
788}
789
790static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
791 int fh_len, int fh_type)
792{
793 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
794 ext4_nfs_get_inode);
795}
796
797static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
798 int fh_len, int fh_type)
799{
800 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
801 ext4_nfs_get_inode);
ac27a0ec
DK		 802}
803
804#ifdef CONFIG_QUOTA
805#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
806#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
807
617ba13b
MC		 808static int ext4_dquot_initialize(struct inode *inode, int type);
809static int ext4_dquot_drop(struct inode *inode);
810static int ext4_write_dquot(struct dquot *dquot);
811static int ext4_acquire_dquot(struct dquot *dquot);
812static int ext4_release_dquot(struct dquot *dquot);
813static int ext4_mark_dquot_dirty(struct dquot *dquot);
814static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 815static int ext4_quota_on(struct super_block *sb, int type, int format_id,
816 char *path, int remount);
617ba13b
MC		 817static int ext4_quota_on_mount(struct super_block *sb, int type);
818static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 819 size_t len, loff_t off);
617ba13b 820static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 821 const char *data, size_t len, loff_t off);
822
617ba13b
MC		 823static struct dquot_operations ext4_quota_operations = {
824 .initialize = ext4_dquot_initialize,
825 .drop = ext4_dquot_drop,
ac27a0ec
DK		 826 .alloc_space = dquot_alloc_space,
827 .alloc_inode = dquot_alloc_inode,
828 .free_space = dquot_free_space,
829 .free_inode = dquot_free_inode,
830 .transfer = dquot_transfer,
617ba13b
MC		 831 .write_dquot = ext4_write_dquot,
832 .acquire_dquot = ext4_acquire_dquot,
833 .release_dquot = ext4_release_dquot,
834 .mark_dirty = ext4_mark_dquot_dirty,
835 .write_info = ext4_write_info
ac27a0ec
DK		 836};
837
617ba13b
MC		 838static struct quotactl_ops ext4_qctl_operations = {
839 .quota_on = ext4_quota_on,
ac27a0ec
DK		 840 .quota_off = vfs_quota_off,
841 .quota_sync = vfs_quota_sync,
842 .get_info = vfs_get_dqinfo,
843 .set_info = vfs_set_dqinfo,
844 .get_dqblk = vfs_get_dqblk,
845 .set_dqblk = vfs_set_dqblk
846};
847#endif
848
ee9b6d61 849static const struct super_operations ext4_sops = {
617ba13b
MC		 850 .alloc_inode = ext4_alloc_inode,
851 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 852 .write_inode = ext4_write_inode,
853 .dirty_inode = ext4_dirty_inode,
854 .delete_inode = ext4_delete_inode,
855 .put_super = ext4_put_super,
856 .write_super = ext4_write_super,
857 .sync_fs = ext4_sync_fs,
858 .write_super_lockfs = ext4_write_super_lockfs,
859 .unlockfs = ext4_unlockfs,
860 .statfs = ext4_statfs,
861 .remount_fs = ext4_remount,
862 .clear_inode = ext4_clear_inode,
863 .show_options = ext4_show_options,
ac27a0ec 864#ifdef CONFIG_QUOTA
617ba13b
MC		 865 .quota_read = ext4_quota_read,
866 .quota_write = ext4_quota_write,
ac27a0ec
DK		 867#endif
868};
869
39655164 870static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 871 .fh_to_dentry = ext4_fh_to_dentry,
872 .fh_to_parent = ext4_fh_to_parent,
617ba13b 873 .get_parent = ext4_get_parent,
ac27a0ec
DK		 874};
875
876enum {
877 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
878 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
879 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
880 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
881 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
882 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
818d276c 883 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec
DK		 884 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
885 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
886 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
887 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
25ec56b5 888 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
c9de560d 889 Opt_mballoc, Opt_nomballoc, Opt_stripe,
ac27a0ec
DK		 890};
891
892static match_table_t tokens = {
893 {Opt_bsd_df, "bsddf"},
894 {Opt_minix_df, "minixdf"},
895 {Opt_grpid, "grpid"},
896 {Opt_grpid, "bsdgroups"},
897 {Opt_nogrpid, "nogrpid"},
898 {Opt_nogrpid, "sysvgroups"},
899 {Opt_resgid, "resgid=%u"},
900 {Opt_resuid, "resuid=%u"},
901 {Opt_sb, "sb=%u"},
902 {Opt_err_cont, "errors=continue"},
903 {Opt_err_panic, "errors=panic"},
904 {Opt_err_ro, "errors=remount-ro"},
905 {Opt_nouid32, "nouid32"},
906 {Opt_nocheck, "nocheck"},
907 {Opt_nocheck, "check=none"},
908 {Opt_debug, "debug"},
909 {Opt_oldalloc, "oldalloc"},
910 {Opt_orlov, "orlov"},
911 {Opt_user_xattr, "user_xattr"},
912 {Opt_nouser_xattr, "nouser_xattr"},
913 {Opt_acl, "acl"},
914 {Opt_noacl, "noacl"},
915 {Opt_reservation, "reservation"},
916 {Opt_noreservation, "noreservation"},
917 {Opt_noload, "noload"},
918 {Opt_nobh, "nobh"},
919 {Opt_bh, "bh"},
920 {Opt_commit, "commit=%u"},
921 {Opt_journal_update, "journal=update"},
922 {Opt_journal_inum, "journal=%u"},
923 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 924 {Opt_journal_checksum, "journal_checksum"},
925 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 926 {Opt_abort, "abort"},
927 {Opt_data_journal, "data=journal"},
928 {Opt_data_ordered, "data=ordered"},
929 {Opt_data_writeback, "data=writeback"},
930 {Opt_offusrjquota, "usrjquota="},
931 {Opt_usrjquota, "usrjquota=%s"},
932 {Opt_offgrpjquota, "grpjquota="},
933 {Opt_grpjquota, "grpjquota=%s"},
934 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
935 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
936 {Opt_grpquota, "grpquota"},
937 {Opt_noquota, "noquota"},
938 {Opt_quota, "quota"},
939 {Opt_usrquota, "usrquota"},
940 {Opt_barrier, "barrier=%u"},
a86c6181 941 {Opt_extents, "extents"},
1e2462f9 942 {Opt_noextents, "noextents"},
25ec56b5 943 {Opt_i_version, "i_version"},
c9de560d
AT		 944 {Opt_mballoc, "mballoc"},
945 {Opt_nomballoc, "nomballoc"},
946 {Opt_stripe, "stripe=%u"},
ac27a0ec 947 {Opt_resize, "resize"},
f3f12faa 948 {Opt_err, NULL},
ac27a0ec
DK		 949};
950
617ba13b 951static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 952{
617ba13b 953 ext4_fsblk_t sb_block;
ac27a0ec
DK		 954 char *options = (char *) *data;
955
956 if (!options || strncmp(options, "sb=", 3) != 0)
957 return 1; /* Default location */
958 options += 3;
617ba13b 959 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 960 sb_block = simple_strtoul(options, &options, 0);
961 if (*options && *options != ',') {
617ba13b 962 printk("EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 963 (char *) *data);
964 return 1;
965 }
966 if (*options == ',')
967 options++;
968 *data = (void *) options;
969 return sb_block;
970}
971
972static int parse_options (char *options, struct super_block *sb,
973 unsigned int *inum, unsigned long *journal_devnum,
617ba13b 974 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 975{
617ba13b 976 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 977 char * p;
978 substring_t args[MAX_OPT_ARGS];
979 int data_opt = 0;
980 int option;
981#ifdef CONFIG_QUOTA
dfc5d03f 982 int qtype, qfmt;
ac27a0ec
DK		 983 char *qname;
984#endif
985
986 if (!options)
987 return 1;
988
989 while ((p = strsep (&options, ",")) != NULL) {
990 int token;
991 if (!*p)
992 continue;
993
994 token = match_token(p, tokens, args);
995 switch (token) {
996 case Opt_bsd_df:
997 clear_opt (sbi->s_mount_opt, MINIX_DF);
998 break;
999 case Opt_minix_df:
1000 set_opt (sbi->s_mount_opt, MINIX_DF);
1001 break;
1002 case Opt_grpid:
1003 set_opt (sbi->s_mount_opt, GRPID);
1004 break;
1005 case Opt_nogrpid:
1006 clear_opt (sbi->s_mount_opt, GRPID);
1007 break;
1008 case Opt_resuid:
1009 if (match_int(&args[0], &option))
1010 return 0;
1011 sbi->s_resuid = option;
1012 break;
1013 case Opt_resgid:
1014 if (match_int(&args[0], &option))
1015 return 0;
1016 sbi->s_resgid = option;
1017 break;
1018 case Opt_sb:
1019 /* handled by get_sb_block() instead of here */
1020 /* *sb_block = match_int(&args[0]); */
1021 break;
1022 case Opt_err_panic:
1023 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1024 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1025 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1026 break;
1027 case Opt_err_ro:
1028 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1029 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1030 set_opt (sbi->s_mount_opt, ERRORS_RO);
1031 break;
1032 case Opt_err_cont:
1033 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1034 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1035 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1036 break;
1037 case Opt_nouid32:
1038 set_opt (sbi->s_mount_opt, NO_UID32);
1039 break;
1040 case Opt_nocheck:
1041 clear_opt (sbi->s_mount_opt, CHECK);
1042 break;
1043 case Opt_debug:
1044 set_opt (sbi->s_mount_opt, DEBUG);
1045 break;
1046 case Opt_oldalloc:
1047 set_opt (sbi->s_mount_opt, OLDALLOC);
1048 break;
1049 case Opt_orlov:
1050 clear_opt (sbi->s_mount_opt, OLDALLOC);
1051 break;
617ba13b 1052#ifdef CONFIG_EXT4DEV_FS_XATTR
ac27a0ec
DK		 1053 case Opt_user_xattr:
1054 set_opt (sbi->s_mount_opt, XATTR_USER);
1055 break;
1056 case Opt_nouser_xattr:
1057 clear_opt (sbi->s_mount_opt, XATTR_USER);
1058 break;
1059#else
1060 case Opt_user_xattr:
1061 case Opt_nouser_xattr:
617ba13b 1062 printk("EXT4 (no)user_xattr options not supported\n");
ac27a0ec
DK		 1063 break;
1064#endif
617ba13b 1065#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
ac27a0ec
DK		 1066 case Opt_acl:
1067 set_opt(sbi->s_mount_opt, POSIX_ACL);
1068 break;
1069 case Opt_noacl:
1070 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1071 break;
1072#else
1073 case Opt_acl:
1074 case Opt_noacl:
617ba13b 1075 printk("EXT4 (no)acl options not supported\n");
ac27a0ec
DK		 1076 break;
1077#endif
1078 case Opt_reservation:
1079 set_opt(sbi->s_mount_opt, RESERVATION);
1080 break;
1081 case Opt_noreservation:
1082 clear_opt(sbi->s_mount_opt, RESERVATION);
1083 break;
1084 case Opt_journal_update:
1085 /* @@@ FIXME */
1086 /* Eventually we will want to be able to create
1087 a journal file here. For now, only allow the
1088 user to specify an existing inode to be the
1089 journal file. */
1090 if (is_remount) {
617ba13b 1091 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1092 "journal on remount\n");
1093 return 0;
1094 }
1095 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1096 break;
1097 case Opt_journal_inum:
1098 if (is_remount) {
617ba13b 1099 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1100 "journal on remount\n");
1101 return 0;
1102 }
1103 if (match_int(&args[0], &option))
1104 return 0;
1105 *inum = option;
1106 break;
1107 case Opt_journal_dev:
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 if (match_int(&args[0], &option))
1114 return 0;
1115 *journal_devnum = option;
1116 break;
818d276c
GS		 1117 case Opt_journal_checksum:
1118 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1119 break;
1120 case Opt_journal_async_commit:
1121 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1122 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1123 break;
ac27a0ec
DK		 1124 case Opt_noload:
1125 set_opt (sbi->s_mount_opt, NOLOAD);
1126 break;
1127 case Opt_commit:
1128 if (match_int(&args[0], &option))
1129 return 0;
1130 if (option < 0)
1131 return 0;
1132 if (option == 0)
cd02ff0b 1133 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1134 sbi->s_commit_interval = HZ * option;
1135 break;
1136 case Opt_data_journal:
617ba13b 1137 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1138 goto datacheck;
1139 case Opt_data_ordered:
617ba13b 1140 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1141 goto datacheck;
1142 case Opt_data_writeback:
617ba13b 1143 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1144 datacheck:
1145 if (is_remount) {
617ba13b 1146 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1147 != data_opt) {
1148 printk(KERN_ERR
617ba13b 1149 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1150 "mode on remount\n");
1151 return 0;
1152 }
1153 } else {
617ba13b 1154 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1155 sbi->s_mount_opt |= data_opt;
1156 }
1157 break;
1158#ifdef CONFIG_QUOTA
1159 case Opt_usrjquota:
1160 qtype = USRQUOTA;
1161 goto set_qf_name;
1162 case Opt_grpjquota:
1163 qtype = GRPQUOTA;
1164set_qf_name:
dfc5d03f
JK		 1165 if ((sb_any_quota_enabled(sb) ||
1166 sb_any_quota_suspended(sb)) &&
1167 !sbi->s_qf_names[qtype]) {
ac27a0ec 1168 printk(KERN_ERR
617ba13b 1169 "EXT4-fs: Cannot change journalled "
ac27a0ec
DK		 1170 "quota options when quota turned on.\n");
1171 return 0;
1172 }
1173 qname = match_strdup(&args[0]);
1174 if (!qname) {
1175 printk(KERN_ERR
617ba13b 1176 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1177 "storing quotafile name.\n");
1178 return 0;
1179 }
1180 if (sbi->s_qf_names[qtype] &&
1181 strcmp(sbi->s_qf_names[qtype], qname)) {
1182 printk(KERN_ERR
617ba13b 1183 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1184 "specified.\n", QTYPE2NAME(qtype));
1185 kfree(qname);
1186 return 0;
1187 }
1188 sbi->s_qf_names[qtype] = qname;
1189 if (strchr(sbi->s_qf_names[qtype], '/')) {
1190 printk(KERN_ERR
617ba13b 1191 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1192 "filesystem root.\n");
1193 kfree(sbi->s_qf_names[qtype]);
1194 sbi->s_qf_names[qtype] = NULL;
1195 return 0;
1196 }
1197 set_opt(sbi->s_mount_opt, QUOTA);
1198 break;
1199 case Opt_offusrjquota:
1200 qtype = USRQUOTA;
1201 goto clear_qf_name;
1202 case Opt_offgrpjquota:
1203 qtype = GRPQUOTA;
1204clear_qf_name:
dfc5d03f
JK		 1205 if ((sb_any_quota_enabled(sb) ||
1206 sb_any_quota_suspended(sb)) &&
1207 sbi->s_qf_names[qtype]) {
617ba13b 1208 printk(KERN_ERR "EXT4-fs: Cannot change "
ac27a0ec
DK		 1209 "journalled quota options when "
1210 "quota turned on.\n");
1211 return 0;
1212 }
1213 /*
1214 * The space will be released later when all options
1215 * are confirmed to be correct
1216 */
1217 sbi->s_qf_names[qtype] = NULL;
1218 break;
1219 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1220 qfmt = QFMT_VFS_OLD;
1221 goto set_qf_format;
ac27a0ec 1222 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1223 qfmt = QFMT_VFS_V0;
1224set_qf_format:
1225 if ((sb_any_quota_enabled(sb) ||
1226 sb_any_quota_suspended(sb)) &&
1227 sbi->s_jquota_fmt != qfmt) {
1228 printk(KERN_ERR "EXT4-fs: Cannot change "
1229 "journaled quota options when "
1230 "quota turned on.\n");
1231 return 0;
1232 }
1233 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1234 break;
1235 case Opt_quota:
1236 case Opt_usrquota:
1237 set_opt(sbi->s_mount_opt, QUOTA);
1238 set_opt(sbi->s_mount_opt, USRQUOTA);
1239 break;
1240 case Opt_grpquota:
1241 set_opt(sbi->s_mount_opt, QUOTA);
1242 set_opt(sbi->s_mount_opt, GRPQUOTA);
1243 break;
1244 case Opt_noquota:
1245 if (sb_any_quota_enabled(sb)) {
617ba13b 1246 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1247 "options when quota turned on.\n");
1248 return 0;
1249 }
1250 clear_opt(sbi->s_mount_opt, QUOTA);
1251 clear_opt(sbi->s_mount_opt, USRQUOTA);
1252 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1253 break;
1254#else
1255 case Opt_quota:
1256 case Opt_usrquota:
1257 case Opt_grpquota:
cd59e7b9
JK		 1258 printk(KERN_ERR
1259 "EXT4-fs: quota options not supported.\n");
1260 break;
ac27a0ec
DK		 1261 case Opt_usrjquota:
1262 case Opt_grpjquota:
1263 case Opt_offusrjquota:
1264 case Opt_offgrpjquota:
1265 case Opt_jqfmt_vfsold:
1266 case Opt_jqfmt_vfsv0:
1267 printk(KERN_ERR
cd59e7b9 1268 "EXT4-fs: journaled quota options not "
ac27a0ec
DK		 1269 "supported.\n");
1270 break;
1271 case Opt_noquota:
1272 break;
1273#endif
1274 case Opt_abort:
1275 set_opt(sbi->s_mount_opt, ABORT);
1276 break;
1277 case Opt_barrier:
1278 if (match_int(&args[0], &option))
1279 return 0;
1280 if (option)
1281 set_opt(sbi->s_mount_opt, BARRIER);
1282 else
1283 clear_opt(sbi->s_mount_opt, BARRIER);
1284 break;
1285 case Opt_ignore:
1286 break;
1287 case Opt_resize:
1288 if (!is_remount) {
617ba13b 1289 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1290 "for remount\n");
1291 return 0;
1292 }
1293 if (match_int(&args[0], &option) != 0)
1294 return 0;
1295 *n_blocks_count = option;
1296 break;
1297 case Opt_nobh:
1298 set_opt(sbi->s_mount_opt, NOBH);
1299 break;
1300 case Opt_bh:
1301 clear_opt(sbi->s_mount_opt, NOBH);
1302 break;
a86c6181
AT		 1303 case Opt_extents:
1304 set_opt (sbi->s_mount_opt, EXTENTS);
1305 break;
1e2462f9
MC		 1306 case Opt_noextents:
1307 clear_opt (sbi->s_mount_opt, EXTENTS);
1308 break;
25ec56b5
JNC		 1309 case Opt_i_version:
1310 set_opt(sbi->s_mount_opt, I_VERSION);
1311 sb->s_flags |= MS_I_VERSION;
1312 break;
c9de560d
AT		 1313 case Opt_mballoc:
1314 set_opt(sbi->s_mount_opt, MBALLOC);
1315 break;
1316 case Opt_nomballoc:
1317 clear_opt(sbi->s_mount_opt, MBALLOC);
1318 break;
1319 case Opt_stripe:
1320 if (match_int(&args[0], &option))
1321 return 0;
1322 if (option < 0)
1323 return 0;
1324 sbi->s_stripe = option;
1325 break;
ac27a0ec
DK		 1326 default:
1327 printk (KERN_ERR
617ba13b 1328 "EXT4-fs: Unrecognized mount option \"%s\" "
ac27a0ec
DK		 1329 "or missing value\n", p);
1330 return 0;
1331 }
1332 }
1333#ifdef CONFIG_QUOTA
1334 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1335 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1336 sbi->s_qf_names[USRQUOTA])
1337 clear_opt(sbi->s_mount_opt, USRQUOTA);
1338
617ba13b 1339 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1340 sbi->s_qf_names[GRPQUOTA])
1341 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1342
1343 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1344 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1345 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1346 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1347 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1348 "format mixing.\n");
1349 return 0;
1350 }
1351
1352 if (!sbi->s_jquota_fmt) {
617ba13b 1353 printk(KERN_ERR "EXT4-fs: journalled quota format "
ac27a0ec
DK		 1354 "not specified.\n");
1355 return 0;
1356 }
1357 } else {
1358 if (sbi->s_jquota_fmt) {
617ba13b 1359 printk(KERN_ERR "EXT4-fs: journalled quota format "
ac27a0ec
DK		 1360 "specified with no journalling "
1361 "enabled.\n");
1362 return 0;
1363 }
1364 }
1365#endif
1366 return 1;
1367}
1368
617ba13b 1369static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1370 int read_only)
1371{
617ba13b 1372 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1373 int res = 0;
1374
617ba13b
MC		 1375 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1376 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
ac27a0ec
DK		 1377 "forcing read-only mode\n");
1378 res = MS_RDONLY;
1379 }
1380 if (read_only)
1381 return res;
617ba13b
MC		 1382 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1383 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
ac27a0ec 1384 "running e2fsck is recommended\n");
617ba13b 1385 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
ac27a0ec 1386 printk (KERN_WARNING
617ba13b 1387 "EXT4-fs warning: mounting fs with errors, "
ac27a0ec
DK		 1388 "running e2fsck is recommended\n");
1389 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1390 le16_to_cpu(es->s_mnt_count) >=
1391 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1392 printk (KERN_WARNING
617ba13b 1393 "EXT4-fs warning: maximal mount count reached, "
ac27a0ec
DK		 1394 "running e2fsck is recommended\n");
1395 else if (le32_to_cpu(es->s_checkinterval) &&
1396 (le32_to_cpu(es->s_lastcheck) +
1397 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1398 printk (KERN_WARNING
617ba13b 1399 "EXT4-fs warning: checktime reached, "
ac27a0ec
DK		 1400 "running e2fsck is recommended\n");
1401#if 0
1402 /* @@@ We _will_ want to clear the valid bit if we find
63f57933
AM		 1403 * inconsistencies, to force a fsck at reboot. But for
1404 * a plain journaled filesystem we can keep it set as
1405 * valid forever! :)
1406 */
216c34b2 1407 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec
DK		 1408#endif
1409 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1410 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1411 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1412 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b
MC		 1413 ext4_update_dynamic_rev(sb);
1414 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1415
617ba13b 1416 ext4_commit_super(sb, es, 1);
ac27a0ec 1417 if (test_opt(sb, DEBUG))
617ba13b 1418 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
ac27a0ec
DK		 1419 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1420 sb->s_blocksize,
1421 sbi->s_groups_count,
617ba13b
MC		 1422 EXT4_BLOCKS_PER_GROUP(sb),
1423 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1424 sbi->s_mount_opt);
1425
617ba13b
MC		 1426 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1427 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
ac27a0ec
DK		 1428 char b[BDEVNAME_SIZE];
1429
1430 printk("external journal on %s\n",
617ba13b 1431 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
ac27a0ec
DK		 1432 } else {
1433 printk("internal journal\n");
1434 }
1435 return res;
1436}
1437
717d50e4
AD		 1438__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1439 struct ext4_group_desc *gdp)
1440{
1441 __u16 crc = 0;
1442
1443 if (sbi->s_es->s_feature_ro_compat &
1444 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1445 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1446 __le32 le_group = cpu_to_le32(block_group);
1447
1448 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1449 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1450 crc = crc16(crc, (__u8 *)gdp, offset);
1451 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1452 /* for checksum of struct ext4_group_desc do the rest...*/
1453 if ((sbi->s_es->s_feature_incompat &
1454 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1455 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1456 crc = crc16(crc, (__u8 *)gdp + offset,
1457 le16_to_cpu(sbi->s_es->s_desc_size) -
1458 offset);
1459 }
1460
1461 return cpu_to_le16(crc);
1462}
1463
1464int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1465 struct ext4_group_desc *gdp)
1466{
1467 if ((sbi->s_es->s_feature_ro_compat &
1468 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1469 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1470 return 0;
1471
1472 return 1;
1473}
1474
ac27a0ec 1475/* Called at mount-time, super-block is locked */
197cd65a 1476static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1477{
617ba13b
MC		 1478 struct ext4_sb_info *sbi = EXT4_SB(sb);
1479 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1480 ext4_fsblk_t last_block;
bd81d8ee
LV		 1481 ext4_fsblk_t block_bitmap;
1482 ext4_fsblk_t inode_bitmap;
1483 ext4_fsblk_t inode_table;
ce421581 1484 int flexbg_flag = 0;
fd2d4291 1485 ext4_group_t i;
ac27a0ec 1486
ce421581
JS		 1487 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1488 flexbg_flag = 1;
1489
617ba13b 1490 ext4_debug ("Checking group descriptors");
ac27a0ec 1491
197cd65a
AM		 1492 for (i = 0; i < sbi->s_groups_count; i++) {
1493 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1494
ce421581 1495 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1496 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1497 else
1498 last_block = first_block +
617ba13b 1499 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1500
8fadc143 1501 block_bitmap = ext4_block_bitmap(sb, gdp);
bd81d8ee 1502 if (block_bitmap < first_block || block_bitmap > last_block)
ac27a0ec 1503 {
c19204b0
JB		 1504 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1505 "Block bitmap for group %lu not in group "
1506 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1507 return 0;
1508 }
8fadc143 1509 inode_bitmap = ext4_inode_bitmap(sb, gdp);
bd81d8ee 1510 if (inode_bitmap < first_block || inode_bitmap > last_block)
ac27a0ec 1511 {
c19204b0
JB		 1512 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1513 "Inode bitmap for group %lu not in group "
1514 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1515 return 0;
1516 }
8fadc143 1517 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1518 if (inode_table < first_block ||
780dcdb2 1519 inode_table + sbi->s_itb_per_group - 1 > last_block)
ac27a0ec 1520 {
c19204b0
JB		 1521 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1522 "Inode table for group %lu not in group "
1523 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1524 return 0;
1525 }
717d50e4 1526 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
c19204b0
JB		 1527 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1528 "Checksum for group %lu failed (%u!=%u)\n",
1529 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1530 gdp)), le16_to_cpu(gdp->bg_checksum));
717d50e4
AD		 1531 return 0;
1532 }
ce421581
JS		 1533 if (!flexbg_flag)
1534 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1535 }
1536
bd81d8ee 1537 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
617ba13b 1538 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1539 return 1;
1540}
1541
617ba13b 1542/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1543 * the superblock) which were deleted from all directories, but held open by
1544 * a process at the time of a crash. We walk the list and try to delete these
1545 * inodes at recovery time (only with a read-write filesystem).
1546 *
1547 * In order to keep the orphan inode chain consistent during traversal (in
1548 * case of crash during recovery), we link each inode into the superblock
1549 * orphan list_head and handle it the same way as an inode deletion during
1550 * normal operation (which journals the operations for us).
1551 *
1552 * We only do an iget() and an iput() on each inode, which is very safe if we
1553 * accidentally point at an in-use or already deleted inode. The worst that
1554 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1555 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1556 * e2fsck was run on this filesystem, and it must have already done the orphan
1557 * inode cleanup for us, so we can safely abort without any further action.
1558 */
617ba13b
MC		 1559static void ext4_orphan_cleanup (struct super_block * sb,
1560 struct ext4_super_block * es)
ac27a0ec
DK		 1561{
1562 unsigned int s_flags = sb->s_flags;
1563 int nr_orphans = 0, nr_truncates = 0;
1564#ifdef CONFIG_QUOTA
1565 int i;
1566#endif
1567 if (!es->s_last_orphan) {
1568 jbd_debug(4, "no orphan inodes to clean up\n");
1569 return;
1570 }
1571
a8f48a95
ES		 1572 if (bdev_read_only(sb->s_bdev)) {
1573 printk(KERN_ERR "EXT4-fs: write access "
1574 "unavailable, skipping orphan cleanup.\n");
1575 return;
1576 }
1577
617ba13b 1578 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1579 if (es->s_last_orphan)
1580 jbd_debug(1, "Errors on filesystem, "
1581 "clearing orphan list.\n");
1582 es->s_last_orphan = 0;
1583 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1584 return;
1585 }
1586
1587 if (s_flags & MS_RDONLY) {
617ba13b 1588 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1589 sb->s_id);
1590 sb->s_flags &= ~MS_RDONLY;
1591 }
1592#ifdef CONFIG_QUOTA
1593 /* Needed for iput() to work correctly and not trash data */
1594 sb->s_flags |= MS_ACTIVE;
1595 /* Turn on quotas so that they are updated correctly */
1596 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1597 if (EXT4_SB(sb)->s_qf_names[i]) {
1598 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1599 if (ret < 0)
1600 printk(KERN_ERR
617ba13b 1601 "EXT4-fs: Cannot turn on journalled "
ac27a0ec
DK		 1602 "quota: error %d\n", ret);
1603 }
1604 }
1605#endif
1606
1607 while (es->s_last_orphan) {
1608 struct inode *inode;
1609
97bd42b9
JB		 1610 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1611 if (IS_ERR(inode)) {
ac27a0ec
DK		 1612 es->s_last_orphan = 0;
1613 break;
1614 }
1615
617ba13b 1616 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1617 DQUOT_INIT(inode);
1618 if (inode->i_nlink) {
1619 printk(KERN_DEBUG
1620 "%s: truncating inode %lu to %Ld bytes\n",
46e665e9 1621 __func__, inode->i_ino, inode->i_size);
ac27a0ec
DK		 1622 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1623 inode->i_ino, inode->i_size);
617ba13b 1624 ext4_truncate(inode);
ac27a0ec
DK		 1625 nr_truncates++;
1626 } else {
1627 printk(KERN_DEBUG
1628 "%s: deleting unreferenced inode %lu\n",
46e665e9 1629 __func__, inode->i_ino);
ac27a0ec
DK		 1630 jbd_debug(2, "deleting unreferenced inode %lu\n",
1631 inode->i_ino);
1632 nr_orphans++;
1633 }
1634 iput(inode); /* The delete magic happens here! */
1635 }
1636
1637#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1638
1639 if (nr_orphans)
617ba13b 1640 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1641 sb->s_id, PLURAL(nr_orphans));
1642 if (nr_truncates)
617ba13b 1643 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1644 sb->s_id, PLURAL(nr_truncates));
1645#ifdef CONFIG_QUOTA
1646 /* Turn quotas off */
1647 for (i = 0; i < MAXQUOTAS; i++) {
1648 if (sb_dqopt(sb)->files[i])
6f28e087 1649 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1650 }
1651#endif
1652 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1653}
cd2291a4
ES		 1654/*
1655 * Maximal extent format file size.
1656 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1657 * extent format containers, within a sector_t, and within i_blocks
1658 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1659 * so that won't be a limiting factor.
1660 *
1661 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1662 */
1663static loff_t ext4_max_size(int blkbits)
1664{
1665 loff_t res;
1666 loff_t upper_limit = MAX_LFS_FILESIZE;
1667
1668 /* small i_blocks in vfs inode? */
1669 if (sizeof(blkcnt_t) < sizeof(u64)) {
1670 /*
1671 * CONFIG_LSF is not enabled implies the inode
1672 * i_block represent total blocks in 512 bytes
1673 * 32 == size of vfs inode i_blocks * 8
1674 */
1675 upper_limit = (1LL << 32) - 1;
1676
1677 /* total blocks in file system block size */
1678 upper_limit >>= (blkbits - 9);
1679 upper_limit <<= blkbits;
1680 }
1681
1682 /* 32-bit extent-start container, ee_block */
1683 res = 1LL << 32;
1684 res <<= blkbits;
1685 res -= 1;
1686
1687 /* Sanity check against vm- & vfs- imposed limits */
1688 if (res > upper_limit)
1689 res = upper_limit;
1690
1691 return res;
1692}
ac27a0ec 1693
ac27a0ec 1694/*
cd2291a4 1695 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1696 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1697 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1698 */
cd2291a4 1699static loff_t ext4_max_bitmap_size(int bits)
ac27a0ec 1700{
617ba13b 1701 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1702 int meta_blocks;
1703 loff_t upper_limit;
1704 /* This is calculated to be the largest file size for a
cd2291a4 1705 * dense, bitmapped file such that the total number of
ac27a0ec 1706 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1707 * does not exceed 2^48 -1
1708 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1709 * total number of 512 bytes blocks of the file
1710 */
1711
1712 if (sizeof(blkcnt_t) < sizeof(u64)) {
1713 /*
1714 * CONFIG_LSF is not enabled implies the inode
1715 * i_block represent total blocks in 512 bytes
1716 * 32 == size of vfs inode i_blocks * 8
1717 */
1718 upper_limit = (1LL << 32) - 1;
1719
1720 /* total blocks in file system block size */
1721 upper_limit >>= (bits - 9);
1722
1723 } else {
8180a562
AK		 1724 /*
1725 * We use 48 bit ext4_inode i_blocks
1726 * With EXT4_HUGE_FILE_FL set the i_blocks
1727 * represent total number of blocks in
1728 * file system block size
1729 */
0fc1b451
AK		 1730 upper_limit = (1LL << 48) - 1;
1731
0fc1b451
AK		 1732 }
1733
1734 /* indirect blocks */
1735 meta_blocks = 1;
1736 /* double indirect blocks */
1737 meta_blocks += 1 + (1LL << (bits-2));
1738 /* tripple indirect blocks */
1739 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1740
1741 upper_limit -= meta_blocks;
1742 upper_limit <<= bits;
ac27a0ec
DK		 1743
1744 res += 1LL << (bits-2);
1745 res += 1LL << (2*(bits-2));
1746 res += 1LL << (3*(bits-2));
1747 res <<= bits;
1748 if (res > upper_limit)
1749 res = upper_limit;
0fc1b451
AK		 1750
1751 if (res > MAX_LFS_FILESIZE)
1752 res = MAX_LFS_FILESIZE;
1753
ac27a0ec
DK		 1754 return res;
1755}
1756
617ba13b 1757static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1758 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1759{
617ba13b 1760 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1761 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1762 int has_super = 0;
1763
1764 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1765
617ba13b 1766 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1767 nr < first_meta_bg)
70bbb3e0 1768 return logical_sb_block + nr + 1;
ac27a0ec 1769 bg = sbi->s_desc_per_block * nr;
617ba13b 1770 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1771 has_super = 1;
617ba13b 1772 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1773}
1774
c9de560d
AT		 1775/**
1776 * ext4_get_stripe_size: Get the stripe size.
1777 * @sbi: In memory super block info
1778 *
1779 * If we have specified it via mount option, then
1780 * use the mount option value. If the value specified at mount time is
1781 * greater than the blocks per group use the super block value.
1782 * If the super block value is greater than blocks per group return 0.
1783 * Allocator needs it be less than blocks per group.
1784 *
1785 */
1786static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1787{
1788 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1789 unsigned long stripe_width =
1790 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1791
1792 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1793 return sbi->s_stripe;
1794
1795 if (stripe_width <= sbi->s_blocks_per_group)
1796 return stripe_width;
1797
1798 if (stride <= sbi->s_blocks_per_group)
1799 return stride;
1800
1801 return 0;
1802}
ac27a0ec 1803
617ba13b 1804static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1d03ec98
AK		 1805 __releases(kernel_sem)
1806 __acquires(kernel_sem)
1807
ac27a0ec
DK		 1808{
1809 struct buffer_head * bh;
617ba13b
MC		 1810 struct ext4_super_block *es = NULL;
1811 struct ext4_sb_info *sbi;
1812 ext4_fsblk_t block;
1813 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 1814 ext4_fsblk_t logical_sb_block;
ac27a0ec
DK		 1815 unsigned long offset = 0;
1816 unsigned int journal_inum = 0;
1817 unsigned long journal_devnum = 0;
1818 unsigned long def_mount_opts;
1819 struct inode *root;
1d1fe1ee 1820 int ret = -EINVAL;
ac27a0ec 1821 int blocksize;
ac27a0ec
DK		 1822 int db_count;
1823 int i;
1824 int needs_recovery;
1825 __le32 features;
bd81d8ee 1826 __u64 blocks_count;
833f4077 1827 int err;
ac27a0ec
DK		 1828
1829 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1830 if (!sbi)
1831 return -ENOMEM;
1832 sb->s_fs_info = sbi;
1833 sbi->s_mount_opt = 0;
617ba13b
MC		 1834 sbi->s_resuid = EXT4_DEF_RESUID;
1835 sbi->s_resgid = EXT4_DEF_RESGID;
d9c9bef1 1836 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 1837
1838 unlock_kernel();
1839
617ba13b 1840 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 1841 if (!blocksize) {
617ba13b 1842 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 1843 goto out_fail;
1844 }
1845
afc7cbca
TS		 1846 if (!sb_set_blocksize(sb, blocksize)) {
1847 printk(KERN_ERR "EXT4-fs: bad blocksize %d.\n", blocksize);
1848 goto out_fail;
1849 }
1850
ac27a0ec 1851 /*
617ba13b 1852 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 1853 * block sizes. We need to calculate the offset from buffer start.
1854 */
617ba13b 1855 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 1856 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1857 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 1858 } else {
70bbb3e0 1859 logical_sb_block = sb_block;
ac27a0ec
DK		 1860 }
1861
70bbb3e0 1862 if (!(bh = sb_bread(sb, logical_sb_block))) {
617ba13b 1863 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 1864 goto out_fail;
1865 }
1866 /*
1867 * Note: s_es must be initialized as soon as possible because
617ba13b 1868 * some ext4 macro-instructions depend on its value
ac27a0ec 1869 */
617ba13b 1870 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 1871 sbi->s_es = es;
1872 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 1873 if (sb->s_magic != EXT4_SUPER_MAGIC)
1874 goto cantfind_ext4;
ac27a0ec
DK		 1875
1876 /* Set defaults before we parse the mount options */
1877 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 1878 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 1879 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 1880 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 1881 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 1882 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 1883 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1884#ifdef CONFIG_EXT4DEV_FS_XATTR
617ba13b 1885 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 1886 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1887#endif
1888#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
617ba13b 1889 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 1890 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1891#endif
617ba13b
MC		 1892 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1893 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1894 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1895 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1896 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1897 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1898
1899 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 1900 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 1901 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 1902 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 1903 else
1904 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1905
1906 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1907 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1908
1909 set_opt(sbi->s_mount_opt, RESERVATION);
1910
1e2462f9
MC		 1911 /*
1912 * turn on extents feature by default in ext4 filesystem
1913 * User -o noextents to turn it off
1914 */
1915 set_opt(sbi->s_mount_opt, EXTENTS);
3dbd0ede
AK		 1916 /*
1917 * turn on mballoc feature by default in ext4 filesystem
1918 * User -o nomballoc to turn it off
1919 */
1920 set_opt(sbi->s_mount_opt, MBALLOC);
1e2462f9 1921
ac27a0ec
DK		 1922 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1923 NULL, 0))
1924 goto failed_mount;
1925
1926 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 1927 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 1928
617ba13b
MC		 1929 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1930 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1931 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1932 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 1933 printk(KERN_WARNING
617ba13b 1934 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec 1935 "running e2fsck is recommended\n");
469108ff
TT		 1936
1937 /*
1938 * Since ext4 is still considered development code, we require
1939 * that the TEST_FILESYS flag in s->flags be set.
1940 */
1941 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
1942 printk(KERN_WARNING "EXT4-fs: %s: not marked "
1943 "OK to use with test code.\n", sb->s_id);
1944 goto failed_mount;
1945 }
1946
ac27a0ec
DK		 1947 /*
1948 * Check feature flags regardless of the revision level, since we
1949 * previously didn't change the revision level when setting the flags,
1950 * so there is a chance incompat flags are set on a rev 0 filesystem.
1951 */
617ba13b 1952 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 1953 if (features) {
617ba13b 1954 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
ac27a0ec
DK		 1955 "unsupported optional features (%x).\n",
1956 sb->s_id, le32_to_cpu(features));
1957 goto failed_mount;
1958 }
617ba13b 1959 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 1960 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 1961 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
ac27a0ec
DK		 1962 "unsupported optional features (%x).\n",
1963 sb->s_id, le32_to_cpu(features));
1964 goto failed_mount;
1965 }
0fc1b451
AK		 1966 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
1967 /*
1968 * Large file size enabled file system can only be
1969 * mount if kernel is build with CONFIG_LSF
1970 */
1971 if (sizeof(root->i_blocks) < sizeof(u64) &&
1972 !(sb->s_flags & MS_RDONLY)) {
1973 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
1974 "files cannot be mounted read-write "
1975 "without CONFIG_LSF.\n", sb->s_id);
1976 goto failed_mount;
1977 }
1978 }
ac27a0ec
DK		 1979 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1980
617ba13b
MC		 1981 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1982 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 1983 printk(KERN_ERR
617ba13b 1984 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 1985 blocksize, sb->s_id);
1986 goto failed_mount;
1987 }
1988
ac27a0ec 1989 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 1990
1991 /* Validate the filesystem blocksize */
1992 if (!sb_set_blocksize(sb, blocksize)) {
1993 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
1994 blocksize);
ac27a0ec
DK		 1995 goto failed_mount;
1996 }
1997
1998 brelse (bh);
70bbb3e0
AM		 1999 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2000 offset = do_div(logical_sb_block, blocksize);
2001 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 2002 if (!bh) {
2003 printk(KERN_ERR
617ba13b 2004 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 2005 goto failed_mount;
2006 }
617ba13b 2007 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2008 sbi->s_es = es;
617ba13b 2009 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
ac27a0ec 2010 printk (KERN_ERR
617ba13b 2011 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 2012 goto failed_mount;
2013 }
2014 }
2015
e2b46574 2016 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
617ba13b 2017 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
ac27a0ec 2018
617ba13b
MC		 2019 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2020 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2021 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2022 } else {
2023 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2024 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2025 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2026 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec
DK		 2027 (sbi->s_inode_size > blocksize)) {
2028 printk (KERN_ERR
617ba13b 2029 "EXT4-fs: unsupported inode size: %d\n",
ac27a0ec
DK		 2030 sbi->s_inode_size);
2031 goto failed_mount;
2032 }
ef7f3835
KS		 2033 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2034 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2035 }
0d1ee42f
AR		 2036 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2037 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2038 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2039 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2040 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2041 printk(KERN_ERR
8fadc143 2042 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2043 sbi->s_desc_size);
2044 goto failed_mount;
2045 }
2046 } else
2047 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2048 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2049 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2050 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2051 goto cantfind_ext4;
2052 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2053 if (sbi->s_inodes_per_block == 0)
617ba13b 2054 goto cantfind_ext4;
ac27a0ec
DK		 2055 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2056 sbi->s_inodes_per_block;
0d1ee42f 2057 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2058 sbi->s_sbh = bh;
2059 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2060 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2061 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
ac27a0ec
DK		 2062 for (i=0; i < 4; i++)
2063 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2064 sbi->s_def_hash_version = es->s_def_hash_version;
2065
2066 if (sbi->s_blocks_per_group > blocksize * 8) {
2067 printk (KERN_ERR
617ba13b 2068 "EXT4-fs: #blocks per group too big: %lu\n",
ac27a0ec
DK		 2069 sbi->s_blocks_per_group);
2070 goto failed_mount;
2071 }
ac27a0ec
DK		 2072 if (sbi->s_inodes_per_group > blocksize * 8) {
2073 printk (KERN_ERR
617ba13b 2074 "EXT4-fs: #inodes per group too big: %lu\n",
ac27a0ec
DK		 2075 sbi->s_inodes_per_group);
2076 goto failed_mount;
2077 }
2078
bd81d8ee 2079 if (ext4_blocks_count(es) >
ac27a0ec 2080 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2081 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2082 " too large to mount safely\n", sb->s_id);
2083 if (sizeof(sector_t) < 8)
617ba13b 2084 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2085 "enabled\n");
2086 goto failed_mount;
2087 }
2088
617ba13b
MC		 2089 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2090 goto cantfind_ext4;
e7c95593
ES		 2091
2092 /* ensure blocks_count calculation below doesn't sign-extend */
2093 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2094 le32_to_cpu(es->s_first_data_block) + 1) {
2095 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2096 "first data block %u, blocks per group %lu\n",
2097 ext4_blocks_count(es),
2098 le32_to_cpu(es->s_first_data_block),
2099 EXT4_BLOCKS_PER_GROUP(sb));
2100 goto failed_mount;
2101 }
bd81d8ee
LV		 2102 blocks_count = (ext4_blocks_count(es) -
2103 le32_to_cpu(es->s_first_data_block) +
2104 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2105 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2106 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2107 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2108 EXT4_DESC_PER_BLOCK(sb);
ac27a0ec
DK		 2109 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2110 GFP_KERNEL);
2111 if (sbi->s_group_desc == NULL) {
617ba13b 2112 printk (KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2113 goto failed_mount;
2114 }
2115
2116 bgl_lock_init(&sbi->s_blockgroup_lock);
2117
2118 for (i = 0; i < db_count; i++) {
70bbb3e0 2119 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2120 sbi->s_group_desc[i] = sb_bread(sb, block);
2121 if (!sbi->s_group_desc[i]) {
617ba13b 2122 printk (KERN_ERR "EXT4-fs: "
ac27a0ec
DK		 2123 "can't read group descriptor %d\n", i);
2124 db_count = i;
2125 goto failed_mount2;
2126 }
2127 }
617ba13b
MC		 2128 if (!ext4_check_descriptors (sb)) {
2129 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2130 goto failed_mount2;
2131 }
2132 sbi->s_gdb_count = db_count;
2133 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2134 spin_lock_init(&sbi->s_next_gen_lock);
2135
833f4077
PZ		 2136 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2137 ext4_count_free_blocks(sb));
2138 if (!err) {
2139 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2140 ext4_count_free_inodes(sb));
2141 }
2142 if (!err) {
2143 err = percpu_counter_init(&sbi->s_dirs_counter,
2144 ext4_count_dirs(sb));
2145 }
2146 if (err) {
2147 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2148 goto failed_mount3;
2149 }
ac27a0ec
DK		 2150
2151 /* per fileystem reservation list head & lock */
2152 spin_lock_init(&sbi->s_rsv_window_lock);
2153 sbi->s_rsv_window_root = RB_ROOT;
2154 /* Add a single, static dummy reservation to the start of the
2155 * reservation window list --- it gives us a placeholder for
2156 * append-at-start-of-list which makes the allocation logic
2157 * _much_ simpler. */
617ba13b
MC		 2158 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2159 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
ac27a0ec
DK		 2160 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2161 sbi->s_rsv_window_head.rsv_goal_size = 0;
617ba13b 2162 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
ac27a0ec 2163
c9de560d
AT		 2164 sbi->s_stripe = ext4_get_stripe_size(sbi);
2165
ac27a0ec
DK		 2166 /*
2167 * set up enough so that it can read an inode
2168 */
617ba13b
MC		 2169 sb->s_op = &ext4_sops;
2170 sb->s_export_op = &ext4_export_ops;
2171 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2172#ifdef CONFIG_QUOTA
617ba13b
MC		 2173 sb->s_qcop = &ext4_qctl_operations;
2174 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2175#endif
2176 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2177
2178 sb->s_root = NULL;
2179
2180 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2181 EXT4_HAS_INCOMPAT_FEATURE(sb,
2182 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2183
2184 /*
2185 * The first inode we look at is the journal inode. Don't try
2186 * root first: it may be modified in the journal!
2187 */
2188 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2189 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2190 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec
DK		 2191 goto failed_mount3;
2192 } else if (journal_inum) {
617ba13b 2193 if (ext4_create_journal(sb, es, journal_inum))
ac27a0ec
DK		 2194 goto failed_mount3;
2195 } else {
2196 if (!silent)
2197 printk (KERN_ERR
617ba13b 2198 "ext4: No journal on filesystem on %s\n",
ac27a0ec
DK		 2199 sb->s_id);
2200 goto failed_mount3;
2201 }
2202
eb40a09c
JS		 2203 if (ext4_blocks_count(es) > 0xffffffffULL &&
2204 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2205 JBD2_FEATURE_INCOMPAT_64BIT)) {
2206 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2207 goto failed_mount4;
2208 }
2209
818d276c
GS		 2210 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2211 jbd2_journal_set_features(sbi->s_journal,
2212 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2213 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2214 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2215 jbd2_journal_set_features(sbi->s_journal,
2216 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2217 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2218 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2219 } else {
2220 jbd2_journal_clear_features(sbi->s_journal,
2221 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2222 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2223 }
2224
ac27a0ec
DK		 2225 /* We have now updated the journal if required, so we can
2226 * validate the data journaling mode. */
2227 switch (test_opt(sb, DATA_FLAGS)) {
2228 case 0:
2229 /* No mode set, assume a default based on the journal
63f57933
AM		 2230 * capabilities: ORDERED_DATA if the journal can
2231 * cope, else JOURNAL_DATA
2232 */
dab291af
MC		 2233 if (jbd2_journal_check_available_features
2234 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2235 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2236 else
2237 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2238 break;
2239
617ba13b
MC		 2240 case EXT4_MOUNT_ORDERED_DATA:
2241 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2242 if (!jbd2_journal_check_available_features
2243 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2244 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2245 "requested data journaling mode\n");
2246 goto failed_mount4;
2247 }
2248 default:
2249 break;
2250 }
2251
2252 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2253 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2254 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2255 "its supported only with writeback mode\n");
2256 clear_opt(sbi->s_mount_opt, NOBH);
2257 }
2258 }
2259 /*
dab291af 2260 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2261 * so we can safely mount the rest of the filesystem now.
2262 */
2263
1d1fe1ee
DH		 2264 root = ext4_iget(sb, EXT4_ROOT_INO);
2265 if (IS_ERR(root)) {
617ba13b 2266 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1d1fe1ee 2267 ret = PTR_ERR(root);
ac27a0ec
DK		 2268 goto failed_mount4;
2269 }
2270 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2271 iput(root);
617ba13b 2272 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2273 goto failed_mount4;
2274 }
1d1fe1ee
DH		 2275 sb->s_root = d_alloc_root(root);
2276 if (!sb->s_root) {
2277 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2278 iput(root);
2279 ret = -ENOMEM;
2280 goto failed_mount4;
2281 }
ac27a0ec 2282
617ba13b 2283 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2284
2285 /* determine the minimum size of new large inodes, if present */
2286 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2287 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2288 EXT4_GOOD_OLD_INODE_SIZE;
2289 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2290 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2291 if (sbi->s_want_extra_isize <
2292 le16_to_cpu(es->s_want_extra_isize))
2293 sbi->s_want_extra_isize =
2294 le16_to_cpu(es->s_want_extra_isize);
2295 if (sbi->s_want_extra_isize <
2296 le16_to_cpu(es->s_min_extra_isize))
2297 sbi->s_want_extra_isize =
2298 le16_to_cpu(es->s_min_extra_isize);
2299 }
2300 }
2301 /* Check if enough inode space is available */
2302 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2303 sbi->s_inode_size) {
2304 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2305 EXT4_GOOD_OLD_INODE_SIZE;
2306 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2307 "available.\n");
2308 }
2309
ac27a0ec
DK		 2310 /*
2311 * akpm: core read_super() calls in here with the superblock locked.
2312 * That deadlocks, because orphan cleanup needs to lock the superblock
2313 * in numerous places. Here we just pop the lock - it's relatively
2314 * harmless, because we are now ready to accept write_super() requests,
2315 * and aviro says that's the only reason for hanging onto the
2316 * superblock lock.
2317 */
617ba13b
MC		 2318 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2319 ext4_orphan_cleanup(sb, es);
2320 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
ac27a0ec 2321 if (needs_recovery)
617ba13b
MC		 2322 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2323 ext4_mark_recovery_complete(sb, es);
2324 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2325 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2326 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
ac27a0ec
DK		 2327 "writeback");
2328
a86c6181 2329 ext4_ext_init(sb);
c9de560d 2330 ext4_mb_init(sb, needs_recovery);
a86c6181 2331
ac27a0ec
DK		 2332 lock_kernel();
2333 return 0;
2334
617ba13b 2335cantfind_ext4:
ac27a0ec 2336 if (!silent)
617ba13b 2337 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2338 sb->s_id);
2339 goto failed_mount;
2340
2341failed_mount4:
dab291af 2342 jbd2_journal_destroy(sbi->s_journal);
ac27a0ec
DK		 2343failed_mount3:
2344 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2345 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2346 percpu_counter_destroy(&sbi->s_dirs_counter);
2347failed_mount2:
2348 for (i = 0; i < db_count; i++)
2349 brelse(sbi->s_group_desc[i]);
2350 kfree(sbi->s_group_desc);
2351failed_mount:
2352#ifdef CONFIG_QUOTA
2353 for (i = 0; i < MAXQUOTAS; i++)
2354 kfree(sbi->s_qf_names[i]);
2355#endif
617ba13b 2356 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2357 brelse(bh);
2358out_fail:
2359 sb->s_fs_info = NULL;
2360 kfree(sbi);
2361 lock_kernel();
1d1fe1ee 2362 return ret;
ac27a0ec
DK		 2363}
2364
2365/*
2366 * Setup any per-fs journal parameters now. We'll do this both on
2367 * initial mount, once the journal has been initialised but before we've
2368 * done any recovery; and again on any subsequent remount.
2369 */
617ba13b 2370static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2371{
617ba13b 2372 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2373
2374 if (sbi->s_commit_interval)
2375 journal->j_commit_interval = sbi->s_commit_interval;
617ba13b 2376 /* We could also set up an ext4-specific default for the commit
ac27a0ec
DK		 2377 * interval here, but for now we'll just fall back to the jbd
2378 * default. */
2379
2380 spin_lock(&journal->j_state_lock);
2381 if (test_opt(sb, BARRIER))
dab291af 2382 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2383 else
dab291af 2384 journal->j_flags &= ~JBD2_BARRIER;
ac27a0ec
DK		 2385 spin_unlock(&journal->j_state_lock);
2386}
2387
617ba13b 2388static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2389 unsigned int journal_inum)
2390{
2391 struct inode *journal_inode;
2392 journal_t *journal;
2393
2394 /* First, test for the existence of a valid inode on disk. Bad
2395 * things happen if we iget() an unused inode, as the subsequent
2396 * iput() will try to delete it. */
2397
1d1fe1ee
DH		 2398 journal_inode = ext4_iget(sb, journal_inum);
2399 if (IS_ERR(journal_inode)) {
617ba13b 2400 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2401 return NULL;
2402 }
2403 if (!journal_inode->i_nlink) {
2404 make_bad_inode(journal_inode);
2405 iput(journal_inode);
617ba13b 2406 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2407 return NULL;
2408 }
2409
2410 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2411 journal_inode, journal_inode->i_size);
1d1fe1ee 2412 if (!S_ISREG(journal_inode->i_mode)) {
617ba13b 2413 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2414 iput(journal_inode);
2415 return NULL;
2416 }
2417
dab291af 2418 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2419 if (!journal) {
617ba13b 2420 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2421 iput(journal_inode);
2422 return NULL;
2423 }
2424 journal->j_private = sb;
617ba13b 2425 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2426 return journal;
2427}
2428
617ba13b 2429static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2430 dev_t j_dev)
2431{
2432 struct buffer_head * bh;
2433 journal_t *journal;
617ba13b
MC		 2434 ext4_fsblk_t start;
2435 ext4_fsblk_t len;
ac27a0ec 2436 int hblock, blocksize;
617ba13b 2437 ext4_fsblk_t sb_block;
ac27a0ec 2438 unsigned long offset;
617ba13b 2439 struct ext4_super_block * es;
ac27a0ec
DK		 2440 struct block_device *bdev;
2441
617ba13b 2442 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2443 if (bdev == NULL)
2444 return NULL;
2445
2446 if (bd_claim(bdev, sb)) {
2447 printk(KERN_ERR
8c55e204 2448 "EXT4: failed to claim external journal device.\n");
ac27a0ec
DK		 2449 blkdev_put(bdev);
2450 return NULL;
2451 }
2452
2453 blocksize = sb->s_blocksize;
2454 hblock = bdev_hardsect_size(bdev);
2455 if (blocksize < hblock) {
2456 printk(KERN_ERR
617ba13b 2457 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2458 goto out_bdev;
2459 }
2460
617ba13b
MC		 2461 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2462 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2463 set_blocksize(bdev, blocksize);
2464 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2465 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2466 "external journal\n");
2467 goto out_bdev;
2468 }
2469
617ba13b
MC		 2470 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2471 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2472 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2473 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2474 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2475 "bad superblock\n");
2476 brelse(bh);
2477 goto out_bdev;
2478 }
2479
617ba13b
MC		 2480 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2481 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2482 brelse(bh);
2483 goto out_bdev;
2484 }
2485
bd81d8ee 2486 len = ext4_blocks_count(es);
ac27a0ec
DK		 2487 start = sb_block + 1;
2488 brelse(bh); /* we're done with the superblock */
2489
dab291af 2490 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2491 start, len, blocksize);
2492 if (!journal) {
617ba13b 2493 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2494 goto out_bdev;
2495 }
2496 journal->j_private = sb;
2497 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2498 wait_on_buffer(journal->j_sb_buffer);
2499 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2500 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2501 goto out_journal;
2502 }
2503 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2504 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2505 "user (unsupported) - %d\n",
2506 be32_to_cpu(journal->j_superblock->s_nr_users));
2507 goto out_journal;
2508 }
617ba13b
MC		 2509 EXT4_SB(sb)->journal_bdev = bdev;
2510 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2511 return journal;
2512out_journal:
dab291af 2513 jbd2_journal_destroy(journal);
ac27a0ec 2514out_bdev:
617ba13b 2515 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2516 return NULL;
2517}
2518
617ba13b
MC		 2519static int ext4_load_journal(struct super_block *sb,
2520 struct ext4_super_block *es,
ac27a0ec
DK		 2521 unsigned long journal_devnum)
2522{
2523 journal_t *journal;
2524 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2525 dev_t journal_dev;
2526 int err = 0;
2527 int really_read_only;
2528
2529 if (journal_devnum &&
2530 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2531 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2532 "numbers have changed\n");
2533 journal_dev = new_decode_dev(journal_devnum);
2534 } else
2535 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2536
2537 really_read_only = bdev_read_only(sb->s_bdev);
2538
2539 /*
2540 * Are we loading a blank journal or performing recovery after a
2541 * crash? For recovery, we need to check in advance whether we
2542 * can get read-write access to the device.
2543 */
2544
617ba13b 2545 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2546 if (sb->s_flags & MS_RDONLY) {
617ba13b 2547 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2548 "required on readonly filesystem.\n");
2549 if (really_read_only) {
617ba13b 2550 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2551 "unavailable, cannot proceed.\n");
2552 return -EROFS;
2553 }
617ba13b 2554 printk (KERN_INFO "EXT4-fs: write access will "
ac27a0ec
DK		 2555 "be enabled during recovery.\n");
2556 }
2557 }
2558
2559 if (journal_inum && journal_dev) {
617ba13b 2560 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2561 "and inode journals!\n");
2562 return -EINVAL;
2563 }
2564
2565 if (journal_inum) {
617ba13b 2566 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2567 return -EINVAL;
2568 } else {
617ba13b 2569 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2570 return -EINVAL;
2571 }
2572
2573 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2574 err = jbd2_journal_update_format(journal);
ac27a0ec 2575 if (err) {
617ba13b 2576 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2577 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2578 return err;
2579 }
2580 }
2581
617ba13b 2582 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2583 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2584 if (!err)
dab291af 2585 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2586
2587 if (err) {
617ba13b 2588 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2589 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2590 return err;
2591 }
2592
617ba13b
MC		 2593 EXT4_SB(sb)->s_journal = journal;
2594 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2595
2596 if (journal_devnum &&
2597 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2598 es->s_journal_dev = cpu_to_le32(journal_devnum);
2599 sb->s_dirt = 1;
2600
2601 /* Make sure we flush the recovery flag to disk. */
617ba13b 2602 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2603 }
2604
2605 return 0;
2606}
2607
617ba13b
MC		 2608static int ext4_create_journal(struct super_block * sb,
2609 struct ext4_super_block * es,
ac27a0ec
DK		 2610 unsigned int journal_inum)
2611{
2612 journal_t *journal;
6c675bd4 2613 int err;
ac27a0ec
DK		 2614
2615 if (sb->s_flags & MS_RDONLY) {
617ba13b 2616 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
ac27a0ec
DK		 2617 "create journal.\n");
2618 return -EROFS;
2619 }
2620
6c675bd4
BP		 2621 journal = ext4_get_journal(sb, journal_inum);
2622 if (!journal)
ac27a0ec
DK		 2623 return -EINVAL;
2624
617ba13b 2625 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
ac27a0ec
DK		 2626 journal_inum);
2627
6c675bd4
BP		 2628 err = jbd2_journal_create(journal);
2629 if (err) {
617ba13b 2630 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
dab291af 2631 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2632 return -EIO;
2633 }
2634
617ba13b 2635 EXT4_SB(sb)->s_journal = journal;
ac27a0ec 2636
617ba13b
MC		 2637 ext4_update_dynamic_rev(sb);
2638 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2639 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
ac27a0ec
DK		 2640
2641 es->s_journal_inum = cpu_to_le32(journal_inum);
2642 sb->s_dirt = 1;
2643
2644 /* Make sure we flush the recovery flag to disk. */
617ba13b 2645 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2646
2647 return 0;
2648}
2649
617ba13b
MC		 2650static void ext4_commit_super (struct super_block * sb,
2651 struct ext4_super_block * es,
ac27a0ec
DK		 2652 int sync)
2653{
617ba13b 2654 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
ac27a0ec
DK		 2655
2656 if (!sbh)
2657 return;
2658 es->s_wtime = cpu_to_le32(get_seconds());
bd81d8ee 2659 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
617ba13b 2660 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2661 BUFFER_TRACE(sbh, "marking dirty");
2662 mark_buffer_dirty(sbh);
2663 if (sync)
2664 sync_dirty_buffer(sbh);
2665}
2666
2667
2668/*
2669 * Have we just finished recovery? If so, and if we are mounting (or
2670 * remounting) the filesystem readonly, then we will end up with a
2671 * consistent fs on disk. Record that fact.
2672 */
617ba13b
MC		 2673static void ext4_mark_recovery_complete(struct super_block * sb,
2674 struct ext4_super_block * es)
ac27a0ec 2675{
617ba13b 2676 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2677
dab291af
MC		 2678 jbd2_journal_lock_updates(journal);
2679 jbd2_journal_flush(journal);
32c37730 2680 lock_super(sb);
617ba13b 2681 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2682 sb->s_flags & MS_RDONLY) {
617ba13b 2683 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2684 sb->s_dirt = 0;
617ba13b 2685 ext4_commit_super(sb, es, 1);
ac27a0ec 2686 }
32c37730 2687 unlock_super(sb);
dab291af 2688 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2689}
2690
2691/*
2692 * If we are mounting (or read-write remounting) a filesystem whose journal
2693 * has recorded an error from a previous lifetime, move that error to the
2694 * main filesystem now.
2695 */
617ba13b
MC		 2696static void ext4_clear_journal_err(struct super_block * sb,
2697 struct ext4_super_block * es)
ac27a0ec
DK		 2698{
2699 journal_t *journal;
2700 int j_errno;
2701 const char *errstr;
2702
617ba13b 2703 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2704
2705 /*
2706 * Now check for any error status which may have been recorded in the
617ba13b 2707 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2708 */
2709
dab291af 2710 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2711 if (j_errno) {
2712 char nbuf[16];
2713
617ba13b 2714 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 2715 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 2716 "from previous mount: %s", errstr);
46e665e9 2717 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 2718 "filesystem check.");
2719
617ba13b
MC		 2720 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2721 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2722 ext4_commit_super (sb, es, 1);
ac27a0ec 2723
dab291af 2724 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 2725 }
2726}
2727
2728/*
2729 * Force the running and committing transactions to commit,
2730 * and wait on the commit.
2731 */
617ba13b 2732int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 2733{
2734 journal_t *journal;
2735 int ret;
2736
2737 if (sb->s_flags & MS_RDONLY)
2738 return 0;
2739
617ba13b 2740 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2741 sb->s_dirt = 0;
617ba13b 2742 ret = ext4_journal_force_commit(journal);
ac27a0ec
DK		 2743 return ret;
2744}
2745
2746/*
617ba13b 2747 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec
DK		 2748 * have to propagate any other updates to the superblock on disk at this
2749 * point. Just start an async writeback to get the buffers on their way
2750 * to the disk.
2751 *
2752 * This implicitly triggers the writebehind on sync().
2753 */
2754
617ba13b 2755static void ext4_write_super (struct super_block * sb)
ac27a0ec
DK		 2756{
2757 if (mutex_trylock(&sb->s_lock) != 0)
2758 BUG();
2759 sb->s_dirt = 0;
2760}
2761
617ba13b 2762static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec
DK		 2763{
2764 tid_t target;
2765
2766 sb->s_dirt = 0;
dab291af 2767 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
ac27a0ec 2768 if (wait)
dab291af 2769 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
ac27a0ec
DK		 2770 }
2771 return 0;
2772}
2773
2774/*
2775 * LVM calls this function before a (read-only) snapshot is created. This
2776 * gives us a chance to flush the journal completely and mark the fs clean.
2777 */
617ba13b 2778static void ext4_write_super_lockfs(struct super_block *sb)
ac27a0ec
DK		 2779{
2780 sb->s_dirt = 0;
2781
2782 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 2783 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2784
2785 /* Now we set up the journal barrier. */
dab291af
MC		 2786 jbd2_journal_lock_updates(journal);
2787 jbd2_journal_flush(journal);
ac27a0ec
DK		 2788
2789 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 2790 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2791 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec
DK		 2792 }
2793}
2794
2795/*
2796 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2797 * flag here, even though the filesystem is not technically dirty yet.
2798 */
617ba13b 2799static void ext4_unlockfs(struct super_block *sb)
ac27a0ec
DK		 2800{
2801 if (!(sb->s_flags & MS_RDONLY)) {
2802 lock_super(sb);
2803 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 2804 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2805 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 2806 unlock_super(sb);
dab291af 2807 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 2808 }
2809}
2810
617ba13b 2811static int ext4_remount (struct super_block * sb, int * flags, char * data)
ac27a0ec 2812{
617ba13b
MC		 2813 struct ext4_super_block * es;
2814 struct ext4_sb_info *sbi = EXT4_SB(sb);
2815 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 2816 unsigned long old_sb_flags;
617ba13b 2817 struct ext4_mount_options old_opts;
ac27a0ec
DK		 2818 int err;
2819#ifdef CONFIG_QUOTA
2820 int i;
2821#endif
2822
2823 /* Store the original options */
2824 old_sb_flags = sb->s_flags;
2825 old_opts.s_mount_opt = sbi->s_mount_opt;
2826 old_opts.s_resuid = sbi->s_resuid;
2827 old_opts.s_resgid = sbi->s_resgid;
2828 old_opts.s_commit_interval = sbi->s_commit_interval;
2829#ifdef CONFIG_QUOTA
2830 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2831 for (i = 0; i < MAXQUOTAS; i++)
2832 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2833#endif
2834
2835 /*
2836 * Allow the "check" option to be passed as a remount option.
2837 */
2838 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2839 err = -EINVAL;
2840 goto restore_opts;
2841 }
2842
617ba13b 2843 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
46e665e9 2844 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 2845
2846 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2847 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 2848
2849 es = sbi->s_es;
2850
617ba13b 2851 ext4_init_journal_params(sb, sbi->s_journal);
ac27a0ec
DK		 2852
2853 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 2854 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 2855 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 2856 err = -EROFS;
2857 goto restore_opts;
2858 }
2859
2860 if (*flags & MS_RDONLY) {
2861 /*
2862 * First of all, the unconditional stuff we have to do
2863 * to disable replay of the journal when we next remount
2864 */
2865 sb->s_flags |= MS_RDONLY;
2866
2867 /*
2868 * OK, test if we are remounting a valid rw partition
2869 * readonly, and if so set the rdonly flag and then
2870 * mark the partition as valid again.
2871 */
617ba13b
MC		 2872 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2873 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 2874 es->s_state = cpu_to_le16(sbi->s_mount_state);
2875
32c37730
JK		 2876 /*
2877 * We have to unlock super so that we can wait for
2878 * transactions.
2879 */
2880 unlock_super(sb);
617ba13b 2881 ext4_mark_recovery_complete(sb, es);
32c37730 2882 lock_super(sb);
ac27a0ec
DK		 2883 } else {
2884 __le32 ret;
617ba13b
MC		 2885 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2886 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2887 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec
DK		 2888 "remount RDWR because of unsupported "
2889 "optional features (%x).\n",
2890 sb->s_id, le32_to_cpu(ret));
2891 err = -EROFS;
2892 goto restore_opts;
2893 }
ead6596b
ES		 2894
2895 /*
2896 * If we have an unprocessed orphan list hanging
2897 * around from a previously readonly bdev mount,
2898 * require a full umount/remount for now.
2899 */
2900 if (es->s_last_orphan) {
2901 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2902 "remount RDWR because of unprocessed "
2903 "orphan inode list. Please "
2904 "umount/remount instead.\n",
2905 sb->s_id);
2906 err = -EINVAL;
2907 goto restore_opts;
2908 }
2909
ac27a0ec
DK		 2910 /*
2911 * Mounting a RDONLY partition read-write, so reread
2912 * and store the current valid flag. (It may have
2913 * been changed by e2fsck since we originally mounted
2914 * the partition.)
2915 */
617ba13b 2916 ext4_clear_journal_err(sb, es);
ac27a0ec 2917 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 2918 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 2919 goto restore_opts;
617ba13b 2920 if (!ext4_setup_super (sb, es, 0))
ac27a0ec
DK		 2921 sb->s_flags &= ~MS_RDONLY;
2922 }
2923 }
2924#ifdef CONFIG_QUOTA
2925 /* Release old quota file names */
2926 for (i = 0; i < MAXQUOTAS; i++)
2927 if (old_opts.s_qf_names[i] &&
2928 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2929 kfree(old_opts.s_qf_names[i]);
2930#endif
2931 return 0;
2932restore_opts:
2933 sb->s_flags = old_sb_flags;
2934 sbi->s_mount_opt = old_opts.s_mount_opt;
2935 sbi->s_resuid = old_opts.s_resuid;
2936 sbi->s_resgid = old_opts.s_resgid;
2937 sbi->s_commit_interval = old_opts.s_commit_interval;
2938#ifdef CONFIG_QUOTA
2939 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2940 for (i = 0; i < MAXQUOTAS; i++) {
2941 if (sbi->s_qf_names[i] &&
2942 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2943 kfree(sbi->s_qf_names[i]);
2944 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2945 }
2946#endif
2947 return err;
2948}
2949
617ba13b 2950static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
ac27a0ec
DK		 2951{
2952 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 2953 struct ext4_sb_info *sbi = EXT4_SB(sb);
2954 struct ext4_super_block *es = sbi->s_es;
960cc398 2955 u64 fsid;
ac27a0ec 2956
5e70030d
BP		 2957 if (test_opt(sb, MINIX_DF)) {
2958 sbi->s_overhead_last = 0;
6bc9feff 2959 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 2960 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 2961 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 2962 smp_rmb();
2963
2964 /*
5e70030d
BP		 2965 * Compute the overhead (FS structures). This is constant
2966 * for a given filesystem unless the number of block groups
2967 * changes so we cache the previous value until it does.
ac27a0ec
DK		 2968 */
2969
2970 /*
2971 * All of the blocks before first_data_block are
2972 * overhead
2973 */
2974 overhead = le32_to_cpu(es->s_first_data_block);
2975
2976 /*
2977 * Add the overhead attributed to the superblock and
2978 * block group descriptors. If the sparse superblocks
2979 * feature is turned on, then not all groups have this.
2980 */
2981 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 2982 overhead += ext4_bg_has_super(sb, i) +
2983 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 2984 cond_resched();
2985 }
2986
2987 /*
2988 * Every block group has an inode bitmap, a block
2989 * bitmap, and an inode table.
2990 */
5e70030d
BP		 2991 overhead += ngroups * (2 + sbi->s_itb_per_group);
2992 sbi->s_overhead_last = overhead;
2993 smp_wmb();
6bc9feff 2994 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 2995 }
2996
617ba13b 2997 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 2998 buf->f_bsize = sb->s_blocksize;
5e70030d 2999 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
52d9f3b4 3000 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
308ba3ec 3001 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3002 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3003 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3004 buf->f_bavail = 0;
3005 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3006 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3007 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3008 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3009 fsid = le64_to_cpup((void *)es->s_uuid) ^
3010 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3011 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3012 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 3013 return 0;
3014}
3015
3016/* Helper function for writing quotas on sync - we need to start transaction before quota file
3017 * is locked for write. Otherwise the are possible deadlocks:
3018 * Process 1 Process 2
617ba13b 3019 * ext4_create() quota_sync()
dab291af 3020 * jbd2_journal_start() write_dquot()
ac27a0ec 3021 * DQUOT_INIT() down(dqio_mutex)
dab291af 3022 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3023 *
3024 */
3025
3026#ifdef CONFIG_QUOTA
3027
3028static inline struct inode *dquot_to_inode(struct dquot *dquot)
3029{
3030 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3031}
3032
617ba13b 3033static int ext4_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 3034{
3035 handle_t *handle;
3036 int ret, err;
3037
3038 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 3039 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 3040 if (IS_ERR(handle))
3041 return PTR_ERR(handle);
3042 ret = dquot_initialize(inode, type);
617ba13b 3043 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3044 if (!ret)
3045 ret = err;
3046 return ret;
3047}
3048
617ba13b 3049static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 3050{
3051 handle_t *handle;
3052 int ret, err;
3053
3054 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 3055 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2887df13
JK		 3056 if (IS_ERR(handle)) {
3057 /*
3058 * We call dquot_drop() anyway to at least release references
3059 * to quota structures so that umount does not hang.
3060 */
3061 dquot_drop(inode);
ac27a0ec 3062 return PTR_ERR(handle);
2887df13 3063 }
ac27a0ec 3064 ret = dquot_drop(inode);
617ba13b 3065 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3066 if (!ret)
3067 ret = err;
3068 return ret;
3069}
3070
617ba13b 3071static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3072{
3073 int ret, err;
3074 handle_t *handle;
3075 struct inode *inode;
3076
3077 inode = dquot_to_inode(dquot);
617ba13b
MC		 3078 handle = ext4_journal_start(inode,
3079 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3080 if (IS_ERR(handle))
3081 return PTR_ERR(handle);
3082 ret = dquot_commit(dquot);
617ba13b 3083 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3084 if (!ret)
3085 ret = err;
3086 return ret;
3087}
3088
617ba13b 3089static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3090{
3091 int ret, err;
3092 handle_t *handle;
3093
617ba13b
MC		 3094 handle = ext4_journal_start(dquot_to_inode(dquot),
3095 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3096 if (IS_ERR(handle))
3097 return PTR_ERR(handle);
3098 ret = dquot_acquire(dquot);
617ba13b 3099 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3100 if (!ret)
3101 ret = err;
3102 return ret;
3103}
3104
617ba13b 3105static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3106{
3107 int ret, err;
3108 handle_t *handle;
3109
617ba13b
MC		 3110 handle = ext4_journal_start(dquot_to_inode(dquot),
3111 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3112 if (IS_ERR(handle)) {
3113 /* Release dquot anyway to avoid endless cycle in dqput() */
3114 dquot_release(dquot);
ac27a0ec 3115 return PTR_ERR(handle);
9c3013e9 3116 }
ac27a0ec 3117 ret = dquot_release(dquot);
617ba13b 3118 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3119 if (!ret)
3120 ret = err;
3121 return ret;
3122}
3123
617ba13b 3124static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec
DK		 3125{
3126 /* Are we journalling quotas? */
617ba13b
MC		 3127 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3128 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3129 dquot_mark_dquot_dirty(dquot);
617ba13b 3130 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3131 } else {
3132 return dquot_mark_dquot_dirty(dquot);
3133 }
3134}
3135
617ba13b 3136static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3137{
3138 int ret, err;
3139 handle_t *handle;
3140
3141 /* Data block + inode block */
617ba13b 3142 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3143 if (IS_ERR(handle))
3144 return PTR_ERR(handle);
3145 ret = dquot_commit_info(sb, type);
617ba13b 3146 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3147 if (!ret)
3148 ret = err;
3149 return ret;
3150}
3151
3152/*
3153 * Turn on quotas during mount time - we need to find
3154 * the quota file and such...
3155 */
617ba13b 3156static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3157{
617ba13b
MC		 3158 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3159 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3160}
3161
3162/*
3163 * Standard function to be called on quota_on
3164 */
617ba13b 3165static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6f28e087 3166 char *path, int remount)
ac27a0ec
DK		 3167{
3168 int err;
3169 struct nameidata nd;
3170
3171 if (!test_opt(sb, QUOTA))
3172 return -EINVAL;
0623543b
JK		 3173 /* When remounting, no checks are needed and in fact, path is NULL */
3174 if (remount)
6f28e087 3175 return vfs_quota_on(sb, type, format_id, path, remount);
0623543b 3176
ac27a0ec
DK		 3177 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3178 if (err)
3179 return err;
0623543b 3180
ac27a0ec 3181 /* Quotafile not on the same filesystem? */
4ac91378 3182 if (nd.path.mnt->mnt_sb != sb) {
1d957f9b 3183 path_put(&nd.path);
ac27a0ec
DK		 3184 return -EXDEV;
3185 }
0623543b
JK		 3186 /* Journaling quota? */
3187 if (EXT4_SB(sb)->s_qf_names[type]) {
3188 /* Quotafile not of fs root? */
3189 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3190 printk(KERN_WARNING
3191 "EXT4-fs: Quota file not on filesystem root. "
3192 "Journaled quota will not work.\n");
3193 }
3194
3195 /*
3196 * When we journal data on quota file, we have to flush journal to see
3197 * all updates to the file when we bypass pagecache...
3198 */
3199 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3200 /*
3201 * We don't need to lock updates but journal_flush() could
3202 * otherwise be livelocked...
3203 */
3204 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3205 jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3206 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3207 }
3208
1d957f9b 3209 path_put(&nd.path);
6f28e087 3210 return vfs_quota_on(sb, type, format_id, path, remount);
ac27a0ec
DK		 3211}
3212
3213/* Read data from quotafile - avoid pagecache and such because we cannot afford
3214 * acquiring the locks... As quota files are never truncated and quota code
3215 * itself serializes the operations (and noone else should touch the files)
3216 * we don't have to be afraid of races */
617ba13b 3217static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3218 size_t len, loff_t off)
3219{
3220 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3221 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3222 int err = 0;
3223 int offset = off & (sb->s_blocksize - 1);
3224 int tocopy;
3225 size_t toread;
3226 struct buffer_head *bh;
3227 loff_t i_size = i_size_read(inode);
3228
3229 if (off > i_size)
3230 return 0;
3231 if (off+len > i_size)
3232 len = i_size-off;
3233 toread = len;
3234 while (toread > 0) {
3235 tocopy = sb->s_blocksize - offset < toread ?
3236 sb->s_blocksize - offset : toread;
617ba13b 3237 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3238 if (err)
3239 return err;
3240 if (!bh) /* A hole? */
3241 memset(data, 0, tocopy);
3242 else
3243 memcpy(data, bh->b_data+offset, tocopy);
3244 brelse(bh);
3245 offset = 0;
3246 toread -= tocopy;
3247 data += tocopy;
3248 blk++;
3249 }
3250 return len;
3251}
3252
3253/* Write to quotafile (we know the transaction is already started and has
3254 * enough credits) */
617ba13b 3255static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3256 const char *data, size_t len, loff_t off)
3257{
3258 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3259 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3260 int err = 0;
3261 int offset = off & (sb->s_blocksize - 1);
3262 int tocopy;
617ba13b 3263 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3264 size_t towrite = len;
3265 struct buffer_head *bh;
3266 handle_t *handle = journal_current_handle();
3267
9c3013e9
JK		 3268 if (!handle) {
3269 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3270 " cancelled because transaction is not started.\n",
3271 (unsigned long long)off, (unsigned long long)len);
3272 return -EIO;
3273 }
ac27a0ec
DK		 3274 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3275 while (towrite > 0) {
3276 tocopy = sb->s_blocksize - offset < towrite ?
3277 sb->s_blocksize - offset : towrite;
617ba13b 3278 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3279 if (!bh)
3280 goto out;
3281 if (journal_quota) {
617ba13b 3282 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3283 if (err) {
3284 brelse(bh);
3285 goto out;
3286 }
3287 }
3288 lock_buffer(bh);
3289 memcpy(bh->b_data+offset, data, tocopy);
3290 flush_dcache_page(bh->b_page);
3291 unlock_buffer(bh);
3292 if (journal_quota)
617ba13b 3293 err = ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK		 3294 else {
3295 /* Always do at least ordered writes for quotas */
617ba13b 3296 err = ext4_journal_dirty_data(handle, bh);
ac27a0ec
DK		 3297 mark_buffer_dirty(bh);
3298 }
3299 brelse(bh);
3300 if (err)
3301 goto out;
3302 offset = 0;
3303 towrite -= tocopy;
3304 data += tocopy;
3305 blk++;
3306 }
3307out:
3308 if (len == towrite)
3309 return err;
3310 if (inode->i_size < off+len-towrite) {
3311 i_size_write(inode, off+len-towrite);
617ba13b 3312 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3313 }
ac27a0ec 3314 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3315 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3316 mutex_unlock(&inode->i_mutex);
3317 return len - towrite;
3318}
3319
3320#endif
3321
617ba13b 3322static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3323 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3324{
617ba13b 3325 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3326}
3327
617ba13b 3328static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3329 .owner = THIS_MODULE,
617ba13b
MC		 3330 .name = "ext4dev",
3331 .get_sb = ext4_get_sb,
ac27a0ec
DK		 3332 .kill_sb = kill_block_super,
3333 .fs_flags = FS_REQUIRES_DEV,
3334};
3335
617ba13b 3336static int __init init_ext4_fs(void)
ac27a0ec 3337{
c9de560d
AT		 3338 int err;
3339
3340 err = init_ext4_mballoc();
ac27a0ec
DK		 3341 if (err)
3342 return err;
c9de560d
AT		 3343
3344 err = init_ext4_xattr();
3345 if (err)
3346 goto out2;
ac27a0ec
DK		 3347 err = init_inodecache();
3348 if (err)
3349 goto out1;
63f57933 3350 err = register_filesystem(&ext4dev_fs_type);
ac27a0ec
DK		 3351 if (err)
3352 goto out;
3353 return 0;
3354out:
3355 destroy_inodecache();
3356out1:
617ba13b 3357 exit_ext4_xattr();
c9de560d
AT		 3358out2:
3359 exit_ext4_mballoc();
ac27a0ec
DK		 3360 return err;
3361}
3362
617ba13b 3363static void __exit exit_ext4_fs(void)
ac27a0ec 3364{
617ba13b 3365 unregister_filesystem(&ext4dev_fs_type);
ac27a0ec 3366 destroy_inodecache();
617ba13b 3367 exit_ext4_xattr();
c9de560d 3368 exit_ext4_mballoc();
ac27a0ec
DK		 3369}
3370
3371MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
617ba13b 3372MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
ac27a0ec 3373MODULE_LICENSE("GPL");
617ba13b
MC		 3374module_init(init_ext4_fs)
3375module_exit(exit_ext4_fs)
kernel source for telekom puls (alcatel/mediatek)