]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/ext4/super.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
i2o debug output cleanup
[net-next-2.6.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
dab291af 23#include <linux/jbd2.h>
617ba13b 24#include <linux/ext4_fs.h>
dab291af 25#include <linux/ext4_jbd2.h>
ac27a0ec
DK		 26#include <linux/slab.h>
27#include <linux/init.h>
28#include <linux/blkdev.h>
29#include <linux/parser.h>
30#include <linux/smp_lock.h>
31#include <linux/buffer_head.h>
32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
38
39#include <asm/uaccess.h>
40
41#include "xattr.h"
42#include "acl.h"
43#include "namei.h"
44
617ba13b 45static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 46 unsigned long journal_devnum);
617ba13b 47static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 48 unsigned int);
617ba13b
MC		 49static void ext4_commit_super (struct super_block * sb,
50 struct ext4_super_block * es,
ac27a0ec 51 int sync);
617ba13b
MC		 52static void ext4_mark_recovery_complete(struct super_block * sb,
53 struct ext4_super_block * es);
54static void ext4_clear_journal_err(struct super_block * sb,
55 struct ext4_super_block * es);
56static int ext4_sync_fs(struct super_block *sb, int wait);
57static const char *ext4_decode_error(struct super_block * sb, int errno,
ac27a0ec 58 char nbuf[16]);
617ba13b
MC		 59static int ext4_remount (struct super_block * sb, int * flags, char * data);
60static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
61static void ext4_unlockfs(struct super_block *sb);
62static void ext4_write_super (struct super_block * sb);
63static void ext4_write_super_lockfs(struct super_block *sb);
ac27a0ec 64
bd81d8ee 65
8fadc143
AR		 66ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
67 struct ext4_group_desc *bg)
bd81d8ee
LV		 68{
69 return le32_to_cpu(bg->bg_block_bitmap) |
8fadc143
AR		 70 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
71 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 72}
73
8fadc143
AR		 74ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
75 struct ext4_group_desc *bg)
bd81d8ee
LV		 76{
77 return le32_to_cpu(bg->bg_inode_bitmap) |
8fadc143
AR		 78 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
79 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 80}
81
8fadc143
AR		 82ext4_fsblk_t ext4_inode_table(struct super_block *sb,
83 struct ext4_group_desc *bg)
bd81d8ee
LV		 84{
85 return le32_to_cpu(bg->bg_inode_table) |
8fadc143
AR		 86 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
87 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 88}
89
8fadc143
AR		 90void ext4_block_bitmap_set(struct super_block *sb,
91 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee
LV		 92{
93 bg->bg_block_bitmap = cpu_to_le32((u32)blk);
8fadc143
AR		 94 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
95 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 96}
97
8fadc143
AR		 98void ext4_inode_bitmap_set(struct super_block *sb,
99 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee
LV		 100{
101 bg->bg_inode_bitmap = cpu_to_le32((u32)blk);
8fadc143
AR		 102 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
103 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 104}
105
8fadc143
AR		 106void ext4_inode_table_set(struct super_block *sb,
107 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee
LV		 108{
109 bg->bg_inode_table = cpu_to_le32((u32)blk);
8fadc143
AR		 110 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
111 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 112}
113
ac27a0ec 114/*
dab291af 115 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 116 *
117 * The only special thing we need to do here is to make sure that all
118 * journal_end calls result in the superblock being marked dirty, so
119 * that sync() will call the filesystem's write_super callback if
120 * appropriate.
121 */
617ba13b 122handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 123{
124 journal_t *journal;
125
126 if (sb->s_flags & MS_RDONLY)
127 return ERR_PTR(-EROFS);
128
129 /* Special case here: if the journal has aborted behind our
130 * backs (eg. EIO in the commit thread), then we still need to
131 * take the FS itself readonly cleanly. */
617ba13b 132 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 133 if (is_journal_aborted(journal)) {
617ba13b 134 ext4_abort(sb, __FUNCTION__,
ac27a0ec
DK		 135 "Detected aborted journal");
136 return ERR_PTR(-EROFS);
137 }
138
dab291af 139 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 140}
141
142/*
143 * The only special thing we need to do here is to make sure that all
dab291af 144 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 145 * that sync() will call the filesystem's write_super callback if
146 * appropriate.
147 */
617ba13b 148int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 149{
150 struct super_block *sb;
151 int err;
152 int rc;
153
154 sb = handle->h_transaction->t_journal->j_private;
155 err = handle->h_err;
dab291af 156 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 157
158 if (!err)
159 err = rc;
160 if (err)
617ba13b 161 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 162 return err;
163}
164
617ba13b 165void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 166 struct buffer_head *bh, handle_t *handle, int err)
167{
168 char nbuf[16];
617ba13b 169 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec
DK		 170
171 if (bh)
172 BUFFER_TRACE(bh, "abort");
173
174 if (!handle->h_err)
175 handle->h_err = err;
176
177 if (is_handle_aborted(handle))
178 return;
179
180 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
181 caller, errstr, err_fn);
182
dab291af 183 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 184}
185
186/* Deal with the reporting of failure conditions on a filesystem such as
187 * inconsistencies detected or read IO failures.
188 *
189 * On ext2, we can store the error state of the filesystem in the
617ba13b 190 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 191 * write ordering constraints on the superblock which prevent us from
192 * writing it out straight away; and given that the journal is about to
193 * be aborted, we can't rely on the current, or future, transactions to
194 * write out the superblock safely.
195 *
dab291af 196 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 197 * the journal instead. On recovery, the journal will compain about
198 * that error until we've noted it down and cleared it.
199 */
200
617ba13b 201static void ext4_handle_error(struct super_block *sb)
ac27a0ec 202{
617ba13b 203 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 204
617ba13b
MC		 205 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
206 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 207
208 if (sb->s_flags & MS_RDONLY)
209 return;
210
211 if (!test_opt (sb, ERRORS_CONT)) {
617ba13b 212 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 213
617ba13b 214 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ac27a0ec 215 if (journal)
dab291af 216 jbd2_journal_abort(journal, -EIO);
ac27a0ec
DK		 217 }
218 if (test_opt (sb, ERRORS_RO)) {
219 printk (KERN_CRIT "Remounting filesystem read-only\n");
220 sb->s_flags |= MS_RDONLY;
221 }
617ba13b 222 ext4_commit_super(sb, es, 1);
ac27a0ec 223 if (test_opt(sb, ERRORS_PANIC))
617ba13b 224 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 225 sb->s_id);
226}
227
617ba13b 228void ext4_error (struct super_block * sb, const char * function,
ac27a0ec
DK		 229 const char * fmt, ...)
230{
231 va_list args;
232
233 va_start(args, fmt);
617ba13b 234 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
ac27a0ec
DK		 235 vprintk(fmt, args);
236 printk("\n");
237 va_end(args);
238
617ba13b 239 ext4_handle_error(sb);
ac27a0ec
DK		 240}
241
617ba13b 242static const char *ext4_decode_error(struct super_block * sb, int errno,
ac27a0ec
DK		 243 char nbuf[16])
244{
245 char *errstr = NULL;
246
247 switch (errno) {
248 case -EIO:
249 errstr = "IO failure";
250 break;
251 case -ENOMEM:
252 errstr = "Out of memory";
253 break;
254 case -EROFS:
dab291af 255 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
ac27a0ec
DK		 256 errstr = "Journal has aborted";
257 else
258 errstr = "Readonly filesystem";
259 break;
260 default:
261 /* If the caller passed in an extra buffer for unknown
262 * errors, textualise them now. Else we just return
263 * NULL. */
264 if (nbuf) {
265 /* Check for truncated error codes... */
266 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
267 errstr = nbuf;
268 }
269 break;
270 }
271
272 return errstr;
273}
274
617ba13b 275/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 276 * automatically and invokes the appropriate error response. */
277
617ba13b 278void __ext4_std_error (struct super_block * sb, const char * function,
ac27a0ec
DK		 279 int errno)
280{
281 char nbuf[16];
282 const char *errstr;
283
284 /* Special case: if the error is EROFS, and we're not already
285 * inside a transaction, then there's really no point in logging
286 * an error. */
287 if (errno == -EROFS && journal_current_handle() == NULL &&
288 (sb->s_flags & MS_RDONLY))
289 return;
290
617ba13b
MC		 291 errstr = ext4_decode_error(sb, errno, nbuf);
292 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
ac27a0ec
DK		 293 sb->s_id, function, errstr);
294
617ba13b 295 ext4_handle_error(sb);
ac27a0ec
DK		 296}
297
298/*
617ba13b 299 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 300 * abort function may be used to deal with unrecoverable failures such
301 * as journal IO errors or ENOMEM at a critical moment in log management.
302 *
303 * We unconditionally force the filesystem into an ABORT|READONLY state,
304 * unless the error response on the fs has been set to panic in which
305 * case we take the easy way out and panic immediately.
306 */
307
617ba13b 308void ext4_abort (struct super_block * sb, const char * function,
ac27a0ec
DK		 309 const char * fmt, ...)
310{
311 va_list args;
312
617ba13b 313 printk (KERN_CRIT "ext4_abort called.\n");
ac27a0ec
DK		 314
315 va_start(args, fmt);
617ba13b 316 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
ac27a0ec
DK		 317 vprintk(fmt, args);
318 printk("\n");
319 va_end(args);
320
321 if (test_opt(sb, ERRORS_PANIC))
617ba13b 322 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 323
324 if (sb->s_flags & MS_RDONLY)
325 return;
326
327 printk(KERN_CRIT "Remounting filesystem read-only\n");
617ba13b 328 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 329 sb->s_flags |= MS_RDONLY;
617ba13b 330 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
dab291af 331 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 332}
333
617ba13b 334void ext4_warning (struct super_block * sb, const char * function,
ac27a0ec
DK		 335 const char * fmt, ...)
336{
337 va_list args;
338
339 va_start(args, fmt);
617ba13b 340 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 341 sb->s_id, function);
342 vprintk(fmt, args);
343 printk("\n");
344 va_end(args);
345}
346
617ba13b 347void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 348{
617ba13b 349 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 350
617ba13b 351 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 352 return;
353
617ba13b 354 ext4_warning(sb, __FUNCTION__,
ac27a0ec
DK		 355 "updating to rev %d because of new feature flag, "
356 "running e2fsck is recommended",
617ba13b 357 EXT4_DYNAMIC_REV);
ac27a0ec 358
617ba13b
MC		 359 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
360 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
361 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 362 /* leave es->s_feature_*compat flags alone */
363 /* es->s_uuid will be set by e2fsck if empty */
364
365 /*
366 * The rest of the superblock fields should be zero, and if not it
367 * means they are likely already in use, so leave them alone. We
368 * can leave it up to e2fsck to clean up any inconsistencies there.
369 */
370}
371
372/*
373 * Open the external journal device
374 */
617ba13b 375static struct block_device *ext4_blkdev_get(dev_t dev)
ac27a0ec
DK		 376{
377 struct block_device *bdev;
378 char b[BDEVNAME_SIZE];
379
380 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
381 if (IS_ERR(bdev))
382 goto fail;
383 return bdev;
384
385fail:
617ba13b 386 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
ac27a0ec
DK		 387 __bdevname(dev, b), PTR_ERR(bdev));
388 return NULL;
389}
390
391/*
392 * Release the journal device
393 */
617ba13b 394static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 395{
396 bd_release(bdev);
397 return blkdev_put(bdev);
398}
399
617ba13b 400static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 401{
402 struct block_device *bdev;
403 int ret = -ENODEV;
404
405 bdev = sbi->journal_bdev;
406 if (bdev) {
617ba13b 407 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 408 sbi->journal_bdev = NULL;
409 }
410 return ret;
411}
412
413static inline struct inode *orphan_list_entry(struct list_head *l)
414{
617ba13b 415 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 416}
417
617ba13b 418static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 419{
420 struct list_head *l;
421
422 printk(KERN_ERR "sb orphan head is %d\n",
423 le32_to_cpu(sbi->s_es->s_last_orphan));
424
425 printk(KERN_ERR "sb_info orphan list:\n");
426 list_for_each(l, &sbi->s_orphan) {
427 struct inode *inode = orphan_list_entry(l);
428 printk(KERN_ERR " "
429 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
430 inode->i_sb->s_id, inode->i_ino, inode,
431 inode->i_mode, inode->i_nlink,
432 NEXT_ORPHAN(inode));
433 }
434}
435
617ba13b 436static void ext4_put_super (struct super_block * sb)
ac27a0ec 437{
617ba13b
MC		 438 struct ext4_sb_info *sbi = EXT4_SB(sb);
439 struct ext4_super_block *es = sbi->s_es;
ac27a0ec
DK		 440 int i;
441
a86c6181 442 ext4_ext_release(sb);
617ba13b 443 ext4_xattr_put_super(sb);
dab291af 444 jbd2_journal_destroy(sbi->s_journal);
ac27a0ec 445 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 446 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec
DK		 447 es->s_state = cpu_to_le16(sbi->s_mount_state);
448 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
449 mark_buffer_dirty(sbi->s_sbh);
617ba13b 450 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 451 }
452
453 for (i = 0; i < sbi->s_gdb_count; i++)
454 brelse(sbi->s_group_desc[i]);
455 kfree(sbi->s_group_desc);
456 percpu_counter_destroy(&sbi->s_freeblocks_counter);
457 percpu_counter_destroy(&sbi->s_freeinodes_counter);
458 percpu_counter_destroy(&sbi->s_dirs_counter);
459 brelse(sbi->s_sbh);
460#ifdef CONFIG_QUOTA
461 for (i = 0; i < MAXQUOTAS; i++)
462 kfree(sbi->s_qf_names[i]);
463#endif
464
465 /* Debugging code just in case the in-memory inode orphan list
466 * isn't empty. The on-disk one can be non-empty if we've
467 * detected an error and taken the fs readonly, but the
468 * in-memory list had better be clean by this point. */
469 if (!list_empty(&sbi->s_orphan))
470 dump_orphan_list(sb, sbi);
471 J_ASSERT(list_empty(&sbi->s_orphan));
472
f98393a6 473 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 474 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
475 /*
476 * Invalidate the journal device's buffers. We don't want them
477 * floating about in memory - the physical journal device may
478 * hotswapped, and it breaks the `ro-after' testing code.
479 */
480 sync_blockdev(sbi->journal_bdev);
f98393a6 481 invalidate_bdev(sbi->journal_bdev);
617ba13b 482 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 483 }
484 sb->s_fs_info = NULL;
485 kfree(sbi);
486 return;
487}
488
e18b890b 489static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 490
491/*
492 * Called inside transaction, so use GFP_NOFS
493 */
617ba13b 494static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 495{
617ba13b 496 struct ext4_inode_info *ei;
ac27a0ec 497
e6b4f8da 498 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 499 if (!ei)
500 return NULL;
617ba13b
MC		 501#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
502 ei->i_acl = EXT4_ACL_NOT_CACHED;
503 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 504#endif
505 ei->i_block_alloc_info = NULL;
506 ei->vfs_inode.i_version = 1;
a86c6181 507 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
ac27a0ec
DK		 508 return &ei->vfs_inode;
509}
510
617ba13b 511static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 512{
9f7dd93d
VA		 513 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
514 printk("EXT4 Inode %p: orphan list check failed!\n",
515 EXT4_I(inode));
516 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
517 EXT4_I(inode), sizeof(struct ext4_inode_info),
518 true);
519 dump_stack();
520 }
617ba13b 521 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 522}
523
e18b890b 524static void init_once(void * foo, struct kmem_cache * cachep, unsigned long flags)
ac27a0ec 525{
617ba13b 526 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 527
a35afb83 528 INIT_LIST_HEAD(&ei->i_orphan);
617ba13b 529#ifdef CONFIG_EXT4DEV_FS_XATTR
a35afb83 530 init_rwsem(&ei->xattr_sem);
ac27a0ec 531#endif
a35afb83
CL		 532 mutex_init(&ei->truncate_mutex);
533 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 534}
535
536static int init_inodecache(void)
537{
617ba13b
MC		 538 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
539 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 540 0, (SLAB_RECLAIM_ACCOUNT|
541 SLAB_MEM_SPREAD),
542 init_once, NULL);
617ba13b 543 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 544 return -ENOMEM;
545 return 0;
546}
547
548static void destroy_inodecache(void)
549{
617ba13b 550 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 551}
552
617ba13b 553static void ext4_clear_inode(struct inode *inode)
ac27a0ec 554{
617ba13b
MC		 555 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
556#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
557 if (EXT4_I(inode)->i_acl &&
558 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
559 posix_acl_release(EXT4_I(inode)->i_acl);
560 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
561 }
562 if (EXT4_I(inode)->i_default_acl &&
563 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
564 posix_acl_release(EXT4_I(inode)->i_default_acl);
565 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 566 }
567#endif
617ba13b
MC		 568 ext4_discard_reservation(inode);
569 EXT4_I(inode)->i_block_alloc_info = NULL;
ac27a0ec
DK		 570 if (unlikely(rsv))
571 kfree(rsv);
572}
573
617ba13b 574static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
ac27a0ec
DK		 575{
576#if defined(CONFIG_QUOTA)
617ba13b 577 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 578
579 if (sbi->s_jquota_fmt)
580 seq_printf(seq, ",jqfmt=%s",
581 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
582
583 if (sbi->s_qf_names[USRQUOTA])
584 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
585
586 if (sbi->s_qf_names[GRPQUOTA])
587 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
588
617ba13b 589 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 590 seq_puts(seq, ",usrquota");
591
617ba13b 592 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 593 seq_puts(seq, ",grpquota");
594#endif
595}
596
617ba13b 597static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec
DK		 598{
599 struct super_block *sb = vfs->mnt_sb;
600
617ba13b 601 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 602 seq_puts(seq, ",data=journal");
617ba13b 603 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 604 seq_puts(seq, ",data=ordered");
617ba13b 605 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 606 seq_puts(seq, ",data=writeback");
607
617ba13b 608 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 609
610 return 0;
611}
612
613
617ba13b 614static struct dentry *ext4_get_dentry(struct super_block *sb, void *vobjp)
ac27a0ec
DK		 615{
616 __u32 *objp = vobjp;
617 unsigned long ino = objp[0];
618 __u32 generation = objp[1];
619 struct inode *inode;
620 struct dentry *result;
621
617ba13b 622 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 623 return ERR_PTR(-ESTALE);
617ba13b 624 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 625 return ERR_PTR(-ESTALE);
626
627 /* iget isn't really right if the inode is currently unallocated!!
628 *
617ba13b 629 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 630 * deleted, so we should be safe.
631 *
632 * Currently we don't know the generation for parent directory, so
633 * a generation of 0 means "accept any"
634 */
635 inode = iget(sb, ino);
636 if (inode == NULL)
637 return ERR_PTR(-ENOMEM);
638 if (is_bad_inode(inode) ||
639 (generation && inode->i_generation != generation)) {
640 iput(inode);
641 return ERR_PTR(-ESTALE);
642 }
643 /* now to find a dentry.
644 * If possible, get a well-connected one
645 */
646 result = d_alloc_anon(inode);
647 if (!result) {
648 iput(inode);
649 return ERR_PTR(-ENOMEM);
650 }
651 return result;
652}
653
654#ifdef CONFIG_QUOTA
655#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
656#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
657
617ba13b
MC		 658static int ext4_dquot_initialize(struct inode *inode, int type);
659static int ext4_dquot_drop(struct inode *inode);
660static int ext4_write_dquot(struct dquot *dquot);
661static int ext4_acquire_dquot(struct dquot *dquot);
662static int ext4_release_dquot(struct dquot *dquot);
663static int ext4_mark_dquot_dirty(struct dquot *dquot);
664static int ext4_write_info(struct super_block *sb, int type);
665static int ext4_quota_on(struct super_block *sb, int type, int format_id, char *path);
666static int ext4_quota_on_mount(struct super_block *sb, int type);
667static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 668 size_t len, loff_t off);
617ba13b 669static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 670 const char *data, size_t len, loff_t off);
671
617ba13b
MC		 672static struct dquot_operations ext4_quota_operations = {
673 .initialize = ext4_dquot_initialize,
674 .drop = ext4_dquot_drop,
ac27a0ec
DK		 675 .alloc_space = dquot_alloc_space,
676 .alloc_inode = dquot_alloc_inode,
677 .free_space = dquot_free_space,
678 .free_inode = dquot_free_inode,
679 .transfer = dquot_transfer,
617ba13b
MC		 680 .write_dquot = ext4_write_dquot,
681 .acquire_dquot = ext4_acquire_dquot,
682 .release_dquot = ext4_release_dquot,
683 .mark_dirty = ext4_mark_dquot_dirty,
684 .write_info = ext4_write_info
ac27a0ec
DK		 685};
686
617ba13b
MC		 687static struct quotactl_ops ext4_qctl_operations = {
688 .quota_on = ext4_quota_on,
ac27a0ec
DK		 689 .quota_off = vfs_quota_off,
690 .quota_sync = vfs_quota_sync,
691 .get_info = vfs_get_dqinfo,
692 .set_info = vfs_set_dqinfo,
693 .get_dqblk = vfs_get_dqblk,
694 .set_dqblk = vfs_set_dqblk
695};
696#endif
697
ee9b6d61 698static const struct super_operations ext4_sops = {
617ba13b
MC		 699 .alloc_inode = ext4_alloc_inode,
700 .destroy_inode = ext4_destroy_inode,
701 .read_inode = ext4_read_inode,
702 .write_inode = ext4_write_inode,
703 .dirty_inode = ext4_dirty_inode,
704 .delete_inode = ext4_delete_inode,
705 .put_super = ext4_put_super,
706 .write_super = ext4_write_super,
707 .sync_fs = ext4_sync_fs,
708 .write_super_lockfs = ext4_write_super_lockfs,
709 .unlockfs = ext4_unlockfs,
710 .statfs = ext4_statfs,
711 .remount_fs = ext4_remount,
712 .clear_inode = ext4_clear_inode,
713 .show_options = ext4_show_options,
ac27a0ec 714#ifdef CONFIG_QUOTA
617ba13b
MC		 715 .quota_read = ext4_quota_read,
716 .quota_write = ext4_quota_write,
ac27a0ec
DK		 717#endif
718};
719
617ba13b
MC		 720static struct export_operations ext4_export_ops = {
721 .get_parent = ext4_get_parent,
722 .get_dentry = ext4_get_dentry,
ac27a0ec
DK		 723};
724
725enum {
726 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
727 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
728 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
729 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
730 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
731 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
732 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
733 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
734 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
735 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
a86c6181 736 Opt_grpquota, Opt_extents,
ac27a0ec
DK		 737};
738
739static match_table_t tokens = {
740 {Opt_bsd_df, "bsddf"},
741 {Opt_minix_df, "minixdf"},
742 {Opt_grpid, "grpid"},
743 {Opt_grpid, "bsdgroups"},
744 {Opt_nogrpid, "nogrpid"},
745 {Opt_nogrpid, "sysvgroups"},
746 {Opt_resgid, "resgid=%u"},
747 {Opt_resuid, "resuid=%u"},
748 {Opt_sb, "sb=%u"},
749 {Opt_err_cont, "errors=continue"},
750 {Opt_err_panic, "errors=panic"},
751 {Opt_err_ro, "errors=remount-ro"},
752 {Opt_nouid32, "nouid32"},
753 {Opt_nocheck, "nocheck"},
754 {Opt_nocheck, "check=none"},
755 {Opt_debug, "debug"},
756 {Opt_oldalloc, "oldalloc"},
757 {Opt_orlov, "orlov"},
758 {Opt_user_xattr, "user_xattr"},
759 {Opt_nouser_xattr, "nouser_xattr"},
760 {Opt_acl, "acl"},
761 {Opt_noacl, "noacl"},
762 {Opt_reservation, "reservation"},
763 {Opt_noreservation, "noreservation"},
764 {Opt_noload, "noload"},
765 {Opt_nobh, "nobh"},
766 {Opt_bh, "bh"},
767 {Opt_commit, "commit=%u"},
768 {Opt_journal_update, "journal=update"},
769 {Opt_journal_inum, "journal=%u"},
770 {Opt_journal_dev, "journal_dev=%u"},
771 {Opt_abort, "abort"},
772 {Opt_data_journal, "data=journal"},
773 {Opt_data_ordered, "data=ordered"},
774 {Opt_data_writeback, "data=writeback"},
775 {Opt_offusrjquota, "usrjquota="},
776 {Opt_usrjquota, "usrjquota=%s"},
777 {Opt_offgrpjquota, "grpjquota="},
778 {Opt_grpjquota, "grpjquota=%s"},
779 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
780 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
781 {Opt_grpquota, "grpquota"},
782 {Opt_noquota, "noquota"},
783 {Opt_quota, "quota"},
784 {Opt_usrquota, "usrquota"},
785 {Opt_barrier, "barrier=%u"},
a86c6181 786 {Opt_extents, "extents"},
ac27a0ec
DK		 787 {Opt_err, NULL},
788 {Opt_resize, "resize"},
789};
790
617ba13b 791static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 792{
617ba13b 793 ext4_fsblk_t sb_block;
ac27a0ec
DK		 794 char *options = (char *) *data;
795
796 if (!options || strncmp(options, "sb=", 3) != 0)
797 return 1; /* Default location */
798 options += 3;
617ba13b 799 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 800 sb_block = simple_strtoul(options, &options, 0);
801 if (*options && *options != ',') {
617ba13b 802 printk("EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 803 (char *) *data);
804 return 1;
805 }
806 if (*options == ',')
807 options++;
808 *data = (void *) options;
809 return sb_block;
810}
811
812static int parse_options (char *options, struct super_block *sb,
813 unsigned int *inum, unsigned long *journal_devnum,
617ba13b 814 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 815{
617ba13b 816 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 817 char * p;
818 substring_t args[MAX_OPT_ARGS];
819 int data_opt = 0;
820 int option;
821#ifdef CONFIG_QUOTA
822 int qtype;
823 char *qname;
824#endif
825
826 if (!options)
827 return 1;
828
829 while ((p = strsep (&options, ",")) != NULL) {
830 int token;
831 if (!*p)
832 continue;
833
834 token = match_token(p, tokens, args);
835 switch (token) {
836 case Opt_bsd_df:
837 clear_opt (sbi->s_mount_opt, MINIX_DF);
838 break;
839 case Opt_minix_df:
840 set_opt (sbi->s_mount_opt, MINIX_DF);
841 break;
842 case Opt_grpid:
843 set_opt (sbi->s_mount_opt, GRPID);
844 break;
845 case Opt_nogrpid:
846 clear_opt (sbi->s_mount_opt, GRPID);
847 break;
848 case Opt_resuid:
849 if (match_int(&args[0], &option))
850 return 0;
851 sbi->s_resuid = option;
852 break;
853 case Opt_resgid:
854 if (match_int(&args[0], &option))
855 return 0;
856 sbi->s_resgid = option;
857 break;
858 case Opt_sb:
859 /* handled by get_sb_block() instead of here */
860 /* *sb_block = match_int(&args[0]); */
861 break;
862 case Opt_err_panic:
863 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
864 clear_opt (sbi->s_mount_opt, ERRORS_RO);
865 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
866 break;
867 case Opt_err_ro:
868 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
869 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
870 set_opt (sbi->s_mount_opt, ERRORS_RO);
871 break;
872 case Opt_err_cont:
873 clear_opt (sbi->s_mount_opt, ERRORS_RO);
874 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
875 set_opt (sbi->s_mount_opt, ERRORS_CONT);
876 break;
877 case Opt_nouid32:
878 set_opt (sbi->s_mount_opt, NO_UID32);
879 break;
880 case Opt_nocheck:
881 clear_opt (sbi->s_mount_opt, CHECK);
882 break;
883 case Opt_debug:
884 set_opt (sbi->s_mount_opt, DEBUG);
885 break;
886 case Opt_oldalloc:
887 set_opt (sbi->s_mount_opt, OLDALLOC);
888 break;
889 case Opt_orlov:
890 clear_opt (sbi->s_mount_opt, OLDALLOC);
891 break;
617ba13b 892#ifdef CONFIG_EXT4DEV_FS_XATTR
ac27a0ec
DK		 893 case Opt_user_xattr:
894 set_opt (sbi->s_mount_opt, XATTR_USER);
895 break;
896 case Opt_nouser_xattr:
897 clear_opt (sbi->s_mount_opt, XATTR_USER);
898 break;
899#else
900 case Opt_user_xattr:
901 case Opt_nouser_xattr:
617ba13b 902 printk("EXT4 (no)user_xattr options not supported\n");
ac27a0ec
DK		 903 break;
904#endif
617ba13b 905#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
ac27a0ec
DK		 906 case Opt_acl:
907 set_opt(sbi->s_mount_opt, POSIX_ACL);
908 break;
909 case Opt_noacl:
910 clear_opt(sbi->s_mount_opt, POSIX_ACL);
911 break;
912#else
913 case Opt_acl:
914 case Opt_noacl:
617ba13b 915 printk("EXT4 (no)acl options not supported\n");
ac27a0ec
DK		 916 break;
917#endif
918 case Opt_reservation:
919 set_opt(sbi->s_mount_opt, RESERVATION);
920 break;
921 case Opt_noreservation:
922 clear_opt(sbi->s_mount_opt, RESERVATION);
923 break;
924 case Opt_journal_update:
925 /* @@@ FIXME */
926 /* Eventually we will want to be able to create
927 a journal file here. For now, only allow the
928 user to specify an existing inode to be the
929 journal file. */
930 if (is_remount) {
617ba13b 931 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 932 "journal on remount\n");
933 return 0;
934 }
935 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
936 break;
937 case Opt_journal_inum:
938 if (is_remount) {
617ba13b 939 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 940 "journal on remount\n");
941 return 0;
942 }
943 if (match_int(&args[0], &option))
944 return 0;
945 *inum = option;
946 break;
947 case Opt_journal_dev:
948 if (is_remount) {
617ba13b 949 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 950 "journal on remount\n");
951 return 0;
952 }
953 if (match_int(&args[0], &option))
954 return 0;
955 *journal_devnum = option;
956 break;
957 case Opt_noload:
958 set_opt (sbi->s_mount_opt, NOLOAD);
959 break;
960 case Opt_commit:
961 if (match_int(&args[0], &option))
962 return 0;
963 if (option < 0)
964 return 0;
965 if (option == 0)
966 option = JBD_DEFAULT_MAX_COMMIT_AGE;
967 sbi->s_commit_interval = HZ * option;
968 break;
969 case Opt_data_journal:
617ba13b 970 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 971 goto datacheck;
972 case Opt_data_ordered:
617ba13b 973 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 974 goto datacheck;
975 case Opt_data_writeback:
617ba13b 976 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 977 datacheck:
978 if (is_remount) {
617ba13b 979 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 980 != data_opt) {
981 printk(KERN_ERR
617ba13b 982 "EXT4-fs: cannot change data "
ac27a0ec
DK		 983 "mode on remount\n");
984 return 0;
985 }
986 } else {
617ba13b 987 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 988 sbi->s_mount_opt |= data_opt;
989 }
990 break;
991#ifdef CONFIG_QUOTA
992 case Opt_usrjquota:
993 qtype = USRQUOTA;
994 goto set_qf_name;
995 case Opt_grpjquota:
996 qtype = GRPQUOTA;
997set_qf_name:
998 if (sb_any_quota_enabled(sb)) {
999 printk(KERN_ERR
617ba13b 1000 "EXT4-fs: Cannot change journalled "
ac27a0ec
DK		 1001 "quota options when quota turned on.\n");
1002 return 0;
1003 }
1004 qname = match_strdup(&args[0]);
1005 if (!qname) {
1006 printk(KERN_ERR
617ba13b 1007 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1008 "storing quotafile name.\n");
1009 return 0;
1010 }
1011 if (sbi->s_qf_names[qtype] &&
1012 strcmp(sbi->s_qf_names[qtype], qname)) {
1013 printk(KERN_ERR
617ba13b 1014 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1015 "specified.\n", QTYPE2NAME(qtype));
1016 kfree(qname);
1017 return 0;
1018 }
1019 sbi->s_qf_names[qtype] = qname;
1020 if (strchr(sbi->s_qf_names[qtype], '/')) {
1021 printk(KERN_ERR
617ba13b 1022 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1023 "filesystem root.\n");
1024 kfree(sbi->s_qf_names[qtype]);
1025 sbi->s_qf_names[qtype] = NULL;
1026 return 0;
1027 }
1028 set_opt(sbi->s_mount_opt, QUOTA);
1029 break;
1030 case Opt_offusrjquota:
1031 qtype = USRQUOTA;
1032 goto clear_qf_name;
1033 case Opt_offgrpjquota:
1034 qtype = GRPQUOTA;
1035clear_qf_name:
1036 if (sb_any_quota_enabled(sb)) {
617ba13b 1037 printk(KERN_ERR "EXT4-fs: Cannot change "
ac27a0ec
DK		 1038 "journalled quota options when "
1039 "quota turned on.\n");
1040 return 0;
1041 }
1042 /*
1043 * The space will be released later when all options
1044 * are confirmed to be correct
1045 */
1046 sbi->s_qf_names[qtype] = NULL;
1047 break;
1048 case Opt_jqfmt_vfsold:
1049 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1050 break;
1051 case Opt_jqfmt_vfsv0:
1052 sbi->s_jquota_fmt = QFMT_VFS_V0;
1053 break;
1054 case Opt_quota:
1055 case Opt_usrquota:
1056 set_opt(sbi->s_mount_opt, QUOTA);
1057 set_opt(sbi->s_mount_opt, USRQUOTA);
1058 break;
1059 case Opt_grpquota:
1060 set_opt(sbi->s_mount_opt, QUOTA);
1061 set_opt(sbi->s_mount_opt, GRPQUOTA);
1062 break;
1063 case Opt_noquota:
1064 if (sb_any_quota_enabled(sb)) {
617ba13b 1065 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1066 "options when quota turned on.\n");
1067 return 0;
1068 }
1069 clear_opt(sbi->s_mount_opt, QUOTA);
1070 clear_opt(sbi->s_mount_opt, USRQUOTA);
1071 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1072 break;
1073#else
1074 case Opt_quota:
1075 case Opt_usrquota:
1076 case Opt_grpquota:
1077 case Opt_usrjquota:
1078 case Opt_grpjquota:
1079 case Opt_offusrjquota:
1080 case Opt_offgrpjquota:
1081 case Opt_jqfmt_vfsold:
1082 case Opt_jqfmt_vfsv0:
1083 printk(KERN_ERR
617ba13b 1084 "EXT4-fs: journalled quota options not "
ac27a0ec
DK		 1085 "supported.\n");
1086 break;
1087 case Opt_noquota:
1088 break;
1089#endif
1090 case Opt_abort:
1091 set_opt(sbi->s_mount_opt, ABORT);
1092 break;
1093 case Opt_barrier:
1094 if (match_int(&args[0], &option))
1095 return 0;
1096 if (option)
1097 set_opt(sbi->s_mount_opt, BARRIER);
1098 else
1099 clear_opt(sbi->s_mount_opt, BARRIER);
1100 break;
1101 case Opt_ignore:
1102 break;
1103 case Opt_resize:
1104 if (!is_remount) {
617ba13b 1105 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1106 "for remount\n");
1107 return 0;
1108 }
1109 if (match_int(&args[0], &option) != 0)
1110 return 0;
1111 *n_blocks_count = option;
1112 break;
1113 case Opt_nobh:
1114 set_opt(sbi->s_mount_opt, NOBH);
1115 break;
1116 case Opt_bh:
1117 clear_opt(sbi->s_mount_opt, NOBH);
1118 break;
a86c6181
AT		 1119 case Opt_extents:
1120 set_opt (sbi->s_mount_opt, EXTENTS);
1121 break;
ac27a0ec
DK		 1122 default:
1123 printk (KERN_ERR
617ba13b 1124 "EXT4-fs: Unrecognized mount option \"%s\" "
ac27a0ec
DK		 1125 "or missing value\n", p);
1126 return 0;
1127 }
1128 }
1129#ifdef CONFIG_QUOTA
1130 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1131 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1132 sbi->s_qf_names[USRQUOTA])
1133 clear_opt(sbi->s_mount_opt, USRQUOTA);
1134
617ba13b 1135 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1136 sbi->s_qf_names[GRPQUOTA])
1137 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1138
1139 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1140 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1141 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1142 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1143 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1144 "format mixing.\n");
1145 return 0;
1146 }
1147
1148 if (!sbi->s_jquota_fmt) {
617ba13b 1149 printk(KERN_ERR "EXT4-fs: journalled quota format "
ac27a0ec
DK		 1150 "not specified.\n");
1151 return 0;
1152 }
1153 } else {
1154 if (sbi->s_jquota_fmt) {
617ba13b 1155 printk(KERN_ERR "EXT4-fs: journalled quota format "
ac27a0ec
DK		 1156 "specified with no journalling "
1157 "enabled.\n");
1158 return 0;
1159 }
1160 }
1161#endif
1162 return 1;
1163}
1164
617ba13b 1165static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1166 int read_only)
1167{
617ba13b 1168 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1169 int res = 0;
1170
617ba13b
MC		 1171 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1172 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
ac27a0ec
DK		 1173 "forcing read-only mode\n");
1174 res = MS_RDONLY;
1175 }
1176 if (read_only)
1177 return res;
617ba13b
MC		 1178 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1179 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
ac27a0ec 1180 "running e2fsck is recommended\n");
617ba13b 1181 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
ac27a0ec 1182 printk (KERN_WARNING
617ba13b 1183 "EXT4-fs warning: mounting fs with errors, "
ac27a0ec
DK		 1184 "running e2fsck is recommended\n");
1185 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1186 le16_to_cpu(es->s_mnt_count) >=
1187 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1188 printk (KERN_WARNING
617ba13b 1189 "EXT4-fs warning: maximal mount count reached, "
ac27a0ec
DK		 1190 "running e2fsck is recommended\n");
1191 else if (le32_to_cpu(es->s_checkinterval) &&
1192 (le32_to_cpu(es->s_lastcheck) +
1193 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1194 printk (KERN_WARNING
617ba13b 1195 "EXT4-fs warning: checktime reached, "
ac27a0ec
DK		 1196 "running e2fsck is recommended\n");
1197#if 0
1198 /* @@@ We _will_ want to clear the valid bit if we find
63f57933
AM		 1199 * inconsistencies, to force a fsck at reboot. But for
1200 * a plain journaled filesystem we can keep it set as
1201 * valid forever! :)
1202 */
617ba13b 1203 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
ac27a0ec
DK		 1204#endif
1205 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1206 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
ac27a0ec
DK		 1207 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1208 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b
MC		 1209 ext4_update_dynamic_rev(sb);
1210 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1211
617ba13b 1212 ext4_commit_super(sb, es, 1);
ac27a0ec 1213 if (test_opt(sb, DEBUG))
617ba13b 1214 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
ac27a0ec
DK		 1215 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1216 sb->s_blocksize,
1217 sbi->s_groups_count,
617ba13b
MC		 1218 EXT4_BLOCKS_PER_GROUP(sb),
1219 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1220 sbi->s_mount_opt);
1221
617ba13b
MC		 1222 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1223 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
ac27a0ec
DK		 1224 char b[BDEVNAME_SIZE];
1225
1226 printk("external journal on %s\n",
617ba13b 1227 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
ac27a0ec
DK		 1228 } else {
1229 printk("internal journal\n");
1230 }
1231 return res;
1232}
1233
1234/* Called at mount-time, super-block is locked */
617ba13b 1235static int ext4_check_descriptors (struct super_block * sb)
ac27a0ec 1236{
617ba13b
MC		 1237 struct ext4_sb_info *sbi = EXT4_SB(sb);
1238 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1239 ext4_fsblk_t last_block;
bd81d8ee
LV		 1240 ext4_fsblk_t block_bitmap;
1241 ext4_fsblk_t inode_bitmap;
1242 ext4_fsblk_t inode_table;
617ba13b 1243 struct ext4_group_desc * gdp = NULL;
ac27a0ec
DK		 1244 int desc_block = 0;
1245 int i;
1246
617ba13b 1247 ext4_debug ("Checking group descriptors");
ac27a0ec
DK		 1248
1249 for (i = 0; i < sbi->s_groups_count; i++)
1250 {
1251 if (i == sbi->s_groups_count - 1)
bd81d8ee 1252 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1253 else
1254 last_block = first_block +
617ba13b 1255 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1256
617ba13b
MC		 1257 if ((i % EXT4_DESC_PER_BLOCK(sb)) == 0)
1258 gdp = (struct ext4_group_desc *)
ac27a0ec 1259 sbi->s_group_desc[desc_block++]->b_data;
8fadc143 1260 block_bitmap = ext4_block_bitmap(sb, gdp);
bd81d8ee 1261 if (block_bitmap < first_block || block_bitmap > last_block)
ac27a0ec 1262 {
617ba13b 1263 ext4_error (sb, "ext4_check_descriptors",
ac27a0ec 1264 "Block bitmap for group %d"
2ae02107 1265 " not in group (block %llu)!",
bd81d8ee 1266 i, block_bitmap);
ac27a0ec
DK		 1267 return 0;
1268 }
8fadc143 1269 inode_bitmap = ext4_inode_bitmap(sb, gdp);
bd81d8ee 1270 if (inode_bitmap < first_block || inode_bitmap > last_block)
ac27a0ec 1271 {
617ba13b 1272 ext4_error (sb, "ext4_check_descriptors",
ac27a0ec 1273 "Inode bitmap for group %d"
2ae02107 1274 " not in group (block %llu)!",
bd81d8ee 1275 i, inode_bitmap);
ac27a0ec
DK		 1276 return 0;
1277 }
8fadc143 1278 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee
LV		 1279 if (inode_table < first_block ||
1280 inode_table + sbi->s_itb_per_group > last_block)
ac27a0ec 1281 {
617ba13b 1282 ext4_error (sb, "ext4_check_descriptors",
ac27a0ec 1283 "Inode table for group %d"
2ae02107 1284 " not in group (block %llu)!",
bd81d8ee 1285 i, inode_table);
ac27a0ec
DK		 1286 return 0;
1287 }
617ba13b 1288 first_block += EXT4_BLOCKS_PER_GROUP(sb);
0d1ee42f
AR		 1289 gdp = (struct ext4_group_desc *)
1290 ((__u8 *)gdp + EXT4_DESC_SIZE(sb));
ac27a0ec
DK		 1291 }
1292
bd81d8ee 1293 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
617ba13b 1294 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1295 return 1;
1296}
1297
1298
617ba13b 1299/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1300 * the superblock) which were deleted from all directories, but held open by
1301 * a process at the time of a crash. We walk the list and try to delete these
1302 * inodes at recovery time (only with a read-write filesystem).
1303 *
1304 * In order to keep the orphan inode chain consistent during traversal (in
1305 * case of crash during recovery), we link each inode into the superblock
1306 * orphan list_head and handle it the same way as an inode deletion during
1307 * normal operation (which journals the operations for us).
1308 *
1309 * We only do an iget() and an iput() on each inode, which is very safe if we
1310 * accidentally point at an in-use or already deleted inode. The worst that
1311 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1312 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1313 * e2fsck was run on this filesystem, and it must have already done the orphan
1314 * inode cleanup for us, so we can safely abort without any further action.
1315 */
617ba13b
MC		 1316static void ext4_orphan_cleanup (struct super_block * sb,
1317 struct ext4_super_block * es)
ac27a0ec
DK		 1318{
1319 unsigned int s_flags = sb->s_flags;
1320 int nr_orphans = 0, nr_truncates = 0;
1321#ifdef CONFIG_QUOTA
1322 int i;
1323#endif
1324 if (!es->s_last_orphan) {
1325 jbd_debug(4, "no orphan inodes to clean up\n");
1326 return;
1327 }
1328
a8f48a95
ES		 1329 if (bdev_read_only(sb->s_bdev)) {
1330 printk(KERN_ERR "EXT4-fs: write access "
1331 "unavailable, skipping orphan cleanup.\n");
1332 return;
1333 }
1334
617ba13b 1335 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1336 if (es->s_last_orphan)
1337 jbd_debug(1, "Errors on filesystem, "
1338 "clearing orphan list.\n");
1339 es->s_last_orphan = 0;
1340 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1341 return;
1342 }
1343
1344 if (s_flags & MS_RDONLY) {
617ba13b 1345 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1346 sb->s_id);
1347 sb->s_flags &= ~MS_RDONLY;
1348 }
1349#ifdef CONFIG_QUOTA
1350 /* Needed for iput() to work correctly and not trash data */
1351 sb->s_flags |= MS_ACTIVE;
1352 /* Turn on quotas so that they are updated correctly */
1353 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1354 if (EXT4_SB(sb)->s_qf_names[i]) {
1355 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1356 if (ret < 0)
1357 printk(KERN_ERR
617ba13b 1358 "EXT4-fs: Cannot turn on journalled "
ac27a0ec
DK		 1359 "quota: error %d\n", ret);
1360 }
1361 }
1362#endif
1363
1364 while (es->s_last_orphan) {
1365 struct inode *inode;
1366
1367 if (!(inode =
617ba13b 1368 ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
ac27a0ec
DK		 1369 es->s_last_orphan = 0;
1370 break;
1371 }
1372
617ba13b 1373 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1374 DQUOT_INIT(inode);
1375 if (inode->i_nlink) {
1376 printk(KERN_DEBUG
1377 "%s: truncating inode %lu to %Ld bytes\n",
1378 __FUNCTION__, inode->i_ino, inode->i_size);
1379 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1380 inode->i_ino, inode->i_size);
617ba13b 1381 ext4_truncate(inode);
ac27a0ec
DK		 1382 nr_truncates++;
1383 } else {
1384 printk(KERN_DEBUG
1385 "%s: deleting unreferenced inode %lu\n",
1386 __FUNCTION__, inode->i_ino);
1387 jbd_debug(2, "deleting unreferenced inode %lu\n",
1388 inode->i_ino);
1389 nr_orphans++;
1390 }
1391 iput(inode); /* The delete magic happens here! */
1392 }
1393
1394#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1395
1396 if (nr_orphans)
617ba13b 1397 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1398 sb->s_id, PLURAL(nr_orphans));
1399 if (nr_truncates)
617ba13b 1400 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1401 sb->s_id, PLURAL(nr_truncates));
1402#ifdef CONFIG_QUOTA
1403 /* Turn quotas off */
1404 for (i = 0; i < MAXQUOTAS; i++) {
1405 if (sb_dqopt(sb)->files[i])
1406 vfs_quota_off(sb, i);
1407 }
1408#endif
1409 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1410}
1411
1412#define log2(n) ffz(~(n))
1413
1414/*
1415 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1416 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1417 * We need to be 1 filesystem block less than the 2^32 sector limit.
1418 */
617ba13b 1419static loff_t ext4_max_size(int bits)
ac27a0ec 1420{
617ba13b 1421 loff_t res = EXT4_NDIR_BLOCKS;
ac27a0ec
DK		 1422 /* This constant is calculated to be the largest file size for a
1423 * dense, 4k-blocksize file such that the total number of
1424 * sectors in the file, including data and all indirect blocks,
1425 * does not exceed 2^32. */
1426 const loff_t upper_limit = 0x1ff7fffd000LL;
1427
1428 res += 1LL << (bits-2);
1429 res += 1LL << (2*(bits-2));
1430 res += 1LL << (3*(bits-2));
1431 res <<= bits;
1432 if (res > upper_limit)
1433 res = upper_limit;
1434 return res;
1435}
1436
617ba13b 1437static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1438 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1439{
617ba13b 1440 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1441 unsigned long bg, first_meta_bg;
1442 int has_super = 0;
1443
1444 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1445
617ba13b 1446 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1447 nr < first_meta_bg)
70bbb3e0 1448 return logical_sb_block + nr + 1;
ac27a0ec 1449 bg = sbi->s_desc_per_block * nr;
617ba13b 1450 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1451 has_super = 1;
617ba13b 1452 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1453}
1454
1455
617ba13b 1456static int ext4_fill_super (struct super_block *sb, void *data, int silent)
ac27a0ec
DK		 1457{
1458 struct buffer_head * bh;
617ba13b
MC		 1459 struct ext4_super_block *es = NULL;
1460 struct ext4_sb_info *sbi;
1461 ext4_fsblk_t block;
1462 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 1463 ext4_fsblk_t logical_sb_block;
ac27a0ec
DK		 1464 unsigned long offset = 0;
1465 unsigned int journal_inum = 0;
1466 unsigned long journal_devnum = 0;
1467 unsigned long def_mount_opts;
1468 struct inode *root;
1469 int blocksize;
1470 int hblock;
1471 int db_count;
1472 int i;
1473 int needs_recovery;
1474 __le32 features;
bd81d8ee 1475 __u64 blocks_count;
ac27a0ec
DK		 1476
1477 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1478 if (!sbi)
1479 return -ENOMEM;
1480 sb->s_fs_info = sbi;
1481 sbi->s_mount_opt = 0;
617ba13b
MC		 1482 sbi->s_resuid = EXT4_DEF_RESUID;
1483 sbi->s_resgid = EXT4_DEF_RESGID;
ac27a0ec
DK		 1484
1485 unlock_kernel();
1486
617ba13b 1487 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 1488 if (!blocksize) {
617ba13b 1489 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 1490 goto out_fail;
1491 }
1492
1493 /*
617ba13b 1494 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 1495 * block sizes. We need to calculate the offset from buffer start.
1496 */
617ba13b 1497 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 1498 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1499 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 1500 } else {
70bbb3e0 1501 logical_sb_block = sb_block;
ac27a0ec
DK		 1502 }
1503
70bbb3e0 1504 if (!(bh = sb_bread(sb, logical_sb_block))) {
617ba13b 1505 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 1506 goto out_fail;
1507 }
1508 /*
1509 * Note: s_es must be initialized as soon as possible because
617ba13b 1510 * some ext4 macro-instructions depend on its value
ac27a0ec 1511 */
617ba13b 1512 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 1513 sbi->s_es = es;
1514 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 1515 if (sb->s_magic != EXT4_SUPER_MAGIC)
1516 goto cantfind_ext4;
ac27a0ec
DK		 1517
1518 /* Set defaults before we parse the mount options */
1519 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 1520 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 1521 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 1522 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 1523 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 1524 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 1525 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1526#ifdef CONFIG_EXT4DEV_FS_XATTR
617ba13b 1527 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 1528 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1529#endif
1530#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
617ba13b 1531 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 1532 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1533#endif
617ba13b
MC		 1534 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1535 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1536 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1537 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1538 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1539 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1540
1541 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 1542 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
617ba13b 1543 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_RO)
ac27a0ec 1544 set_opt(sbi->s_mount_opt, ERRORS_RO);
ceea16bf
DM		 1545 else
1546 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1547
1548 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1549 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1550
1551 set_opt(sbi->s_mount_opt, RESERVATION);
1552
1553 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1554 NULL, 0))
1555 goto failed_mount;
1556
1557 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 1558 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 1559
617ba13b
MC		 1560 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1561 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1562 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1563 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 1564 printk(KERN_WARNING
617ba13b 1565 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec
DK		 1566 "running e2fsck is recommended\n");
1567 /*
1568 * Check feature flags regardless of the revision level, since we
1569 * previously didn't change the revision level when setting the flags,
1570 * so there is a chance incompat flags are set on a rev 0 filesystem.
1571 */
617ba13b 1572 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 1573 if (features) {
617ba13b 1574 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
ac27a0ec
DK		 1575 "unsupported optional features (%x).\n",
1576 sb->s_id, le32_to_cpu(features));
1577 goto failed_mount;
1578 }
617ba13b 1579 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 1580 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 1581 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
ac27a0ec
DK		 1582 "unsupported optional features (%x).\n",
1583 sb->s_id, le32_to_cpu(features));
1584 goto failed_mount;
1585 }
1586 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1587
617ba13b
MC		 1588 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1589 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 1590 printk(KERN_ERR
617ba13b 1591 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 1592 blocksize, sb->s_id);
1593 goto failed_mount;
1594 }
1595
1596 hblock = bdev_hardsect_size(sb->s_bdev);
1597 if (sb->s_blocksize != blocksize) {
1598 /*
1599 * Make sure the blocksize for the filesystem is larger
1600 * than the hardware sectorsize for the machine.
1601 */
1602 if (blocksize < hblock) {
617ba13b 1603 printk(KERN_ERR "EXT4-fs: blocksize %d too small for "
ac27a0ec
DK		 1604 "device blocksize %d.\n", blocksize, hblock);
1605 goto failed_mount;
1606 }
1607
1608 brelse (bh);
1609 sb_set_blocksize(sb, blocksize);
70bbb3e0
AM		 1610 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1611 offset = do_div(logical_sb_block, blocksize);
1612 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 1613 if (!bh) {
1614 printk(KERN_ERR
617ba13b 1615 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 1616 goto failed_mount;
1617 }
617ba13b 1618 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 1619 sbi->s_es = es;
617ba13b 1620 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
ac27a0ec 1621 printk (KERN_ERR
617ba13b 1622 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 1623 goto failed_mount;
1624 }
1625 }
1626
617ba13b 1627 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
ac27a0ec 1628
617ba13b
MC		 1629 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
1630 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
1631 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 1632 } else {
1633 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1634 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 1635 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
ac27a0ec
DK		 1636 (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
1637 (sbi->s_inode_size > blocksize)) {
1638 printk (KERN_ERR
617ba13b 1639 "EXT4-fs: unsupported inode size: %d\n",
ac27a0ec
DK		 1640 sbi->s_inode_size);
1641 goto failed_mount;
1642 }
1643 }
617ba13b 1644 sbi->s_frag_size = EXT4_MIN_FRAG_SIZE <<
ac27a0ec
DK		 1645 le32_to_cpu(es->s_log_frag_size);
1646 if (blocksize != sbi->s_frag_size) {
1647 printk(KERN_ERR
617ba13b 1648 "EXT4-fs: fragsize %lu != blocksize %u (unsupported)\n",
ac27a0ec
DK		 1649 sbi->s_frag_size, blocksize);
1650 goto failed_mount;
1651 }
0d1ee42f
AR		 1652 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
1653 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 1654 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f
AR		 1655 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
1656 sbi->s_desc_size & (sbi->s_desc_size - 1)) {
1657 printk(KERN_ERR
8fadc143 1658 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 1659 sbi->s_desc_size);
1660 goto failed_mount;
1661 }
1662 } else
1663 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec
DK		 1664 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1665 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1666 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
617ba13b
MC		 1667 if (EXT4_INODE_SIZE(sb) == 0)
1668 goto cantfind_ext4;
1669 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 1670 if (sbi->s_inodes_per_block == 0)
617ba13b 1671 goto cantfind_ext4;
ac27a0ec
DK		 1672 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1673 sbi->s_inodes_per_block;
0d1ee42f 1674 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 1675 sbi->s_sbh = bh;
1676 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b
MC		 1677 sbi->s_addr_per_block_bits = log2(EXT4_ADDR_PER_BLOCK(sb));
1678 sbi->s_desc_per_block_bits = log2(EXT4_DESC_PER_BLOCK(sb));
ac27a0ec
DK		 1679 for (i=0; i < 4; i++)
1680 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1681 sbi->s_def_hash_version = es->s_def_hash_version;
1682
1683 if (sbi->s_blocks_per_group > blocksize * 8) {
1684 printk (KERN_ERR
617ba13b 1685 "EXT4-fs: #blocks per group too big: %lu\n",
ac27a0ec
DK		 1686 sbi->s_blocks_per_group);
1687 goto failed_mount;
1688 }
1689 if (sbi->s_frags_per_group > blocksize * 8) {
1690 printk (KERN_ERR
617ba13b 1691 "EXT4-fs: #fragments per group too big: %lu\n",
ac27a0ec
DK		 1692 sbi->s_frags_per_group);
1693 goto failed_mount;
1694 }
1695 if (sbi->s_inodes_per_group > blocksize * 8) {
1696 printk (KERN_ERR
617ba13b 1697 "EXT4-fs: #inodes per group too big: %lu\n",
ac27a0ec
DK		 1698 sbi->s_inodes_per_group);
1699 goto failed_mount;
1700 }
1701
bd81d8ee 1702 if (ext4_blocks_count(es) >
ac27a0ec 1703 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 1704 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 1705 " too large to mount safely\n", sb->s_id);
1706 if (sizeof(sector_t) < 8)
617ba13b 1707 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 1708 "enabled\n");
1709 goto failed_mount;
1710 }
1711
617ba13b
MC		 1712 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
1713 goto cantfind_ext4;
bd81d8ee
LV		 1714 blocks_count = (ext4_blocks_count(es) -
1715 le32_to_cpu(es->s_first_data_block) +
1716 EXT4_BLOCKS_PER_GROUP(sb) - 1);
1717 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
1718 sbi->s_groups_count = blocks_count;
617ba13b
MC		 1719 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
1720 EXT4_DESC_PER_BLOCK(sb);
ac27a0ec
DK		 1721 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1722 GFP_KERNEL);
1723 if (sbi->s_group_desc == NULL) {
617ba13b 1724 printk (KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 1725 goto failed_mount;
1726 }
1727
1728 bgl_lock_init(&sbi->s_blockgroup_lock);
1729
1730 for (i = 0; i < db_count; i++) {
70bbb3e0 1731 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 1732 sbi->s_group_desc[i] = sb_bread(sb, block);
1733 if (!sbi->s_group_desc[i]) {
617ba13b 1734 printk (KERN_ERR "EXT4-fs: "
ac27a0ec
DK		 1735 "can't read group descriptor %d\n", i);
1736 db_count = i;
1737 goto failed_mount2;
1738 }
1739 }
617ba13b
MC		 1740 if (!ext4_check_descriptors (sb)) {
1741 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 1742 goto failed_mount2;
1743 }
1744 sbi->s_gdb_count = db_count;
1745 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1746 spin_lock_init(&sbi->s_next_gen_lock);
1747
1748 percpu_counter_init(&sbi->s_freeblocks_counter,
617ba13b 1749 ext4_count_free_blocks(sb));
ac27a0ec 1750 percpu_counter_init(&sbi->s_freeinodes_counter,
617ba13b 1751 ext4_count_free_inodes(sb));
ac27a0ec 1752 percpu_counter_init(&sbi->s_dirs_counter,
617ba13b 1753 ext4_count_dirs(sb));
ac27a0ec
DK		 1754
1755 /* per fileystem reservation list head & lock */
1756 spin_lock_init(&sbi->s_rsv_window_lock);
1757 sbi->s_rsv_window_root = RB_ROOT;
1758 /* Add a single, static dummy reservation to the start of the
1759 * reservation window list --- it gives us a placeholder for
1760 * append-at-start-of-list which makes the allocation logic
1761 * _much_ simpler. */
617ba13b
MC		 1762 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1763 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
ac27a0ec
DK		 1764 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1765 sbi->s_rsv_window_head.rsv_goal_size = 0;
617ba13b 1766 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
ac27a0ec
DK		 1767
1768 /*
1769 * set up enough so that it can read an inode
1770 */
617ba13b
MC		 1771 sb->s_op = &ext4_sops;
1772 sb->s_export_op = &ext4_export_ops;
1773 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 1774#ifdef CONFIG_QUOTA
617ba13b
MC		 1775 sb->s_qcop = &ext4_qctl_operations;
1776 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 1777#endif
1778 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1779
1780 sb->s_root = NULL;
1781
1782 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 1783 EXT4_HAS_INCOMPAT_FEATURE(sb,
1784 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 1785
1786 /*
1787 * The first inode we look at is the journal inode. Don't try
1788 * root first: it may be modified in the journal!
1789 */
1790 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 1791 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
1792 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec
DK		 1793 goto failed_mount3;
1794 } else if (journal_inum) {
617ba13b 1795 if (ext4_create_journal(sb, es, journal_inum))
ac27a0ec
DK		 1796 goto failed_mount3;
1797 } else {
1798 if (!silent)
1799 printk (KERN_ERR
617ba13b 1800 "ext4: No journal on filesystem on %s\n",
ac27a0ec
DK		 1801 sb->s_id);
1802 goto failed_mount3;
1803 }
1804
1805 /* We have now updated the journal if required, so we can
1806 * validate the data journaling mode. */
1807 switch (test_opt(sb, DATA_FLAGS)) {
1808 case 0:
1809 /* No mode set, assume a default based on the journal
63f57933
AM		 1810 * capabilities: ORDERED_DATA if the journal can
1811 * cope, else JOURNAL_DATA
1812 */
dab291af
MC		 1813 if (jbd2_journal_check_available_features
1814 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 1815 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1816 else
1817 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1818 break;
1819
617ba13b
MC		 1820 case EXT4_MOUNT_ORDERED_DATA:
1821 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 1822 if (!jbd2_journal_check_available_features
1823 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 1824 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 1825 "requested data journaling mode\n");
1826 goto failed_mount4;
1827 }
1828 default:
1829 break;
1830 }
1831
1832 if (test_opt(sb, NOBH)) {
617ba13b
MC		 1833 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
1834 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 1835 "its supported only with writeback mode\n");
1836 clear_opt(sbi->s_mount_opt, NOBH);
1837 }
1838 }
1839 /*
dab291af 1840 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 1841 * so we can safely mount the rest of the filesystem now.
1842 */
1843
617ba13b 1844 root = iget(sb, EXT4_ROOT_INO);
ac27a0ec
DK		 1845 sb->s_root = d_alloc_root(root);
1846 if (!sb->s_root) {
617ba13b 1847 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
ac27a0ec
DK		 1848 iput(root);
1849 goto failed_mount4;
1850 }
1851 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1852 dput(sb->s_root);
1853 sb->s_root = NULL;
617ba13b 1854 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 1855 goto failed_mount4;
1856 }
1857
617ba13b 1858 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
ac27a0ec
DK		 1859 /*
1860 * akpm: core read_super() calls in here with the superblock locked.
1861 * That deadlocks, because orphan cleanup needs to lock the superblock
1862 * in numerous places. Here we just pop the lock - it's relatively
1863 * harmless, because we are now ready to accept write_super() requests,
1864 * and aviro says that's the only reason for hanging onto the
1865 * superblock lock.
1866 */
617ba13b
MC		 1867 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
1868 ext4_orphan_cleanup(sb, es);
1869 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
ac27a0ec 1870 if (needs_recovery)
617ba13b
MC		 1871 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
1872 ext4_mark_recovery_complete(sb, es);
1873 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
1874 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
1875 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
ac27a0ec
DK		 1876 "writeback");
1877
a86c6181
AT		 1878 ext4_ext_init(sb);
1879
ac27a0ec
DK		 1880 lock_kernel();
1881 return 0;
1882
617ba13b 1883cantfind_ext4:
ac27a0ec 1884 if (!silent)
617ba13b 1885 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 1886 sb->s_id);
1887 goto failed_mount;
1888
1889failed_mount4:
dab291af 1890 jbd2_journal_destroy(sbi->s_journal);
ac27a0ec
DK		 1891failed_mount3:
1892 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1893 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1894 percpu_counter_destroy(&sbi->s_dirs_counter);
1895failed_mount2:
1896 for (i = 0; i < db_count; i++)
1897 brelse(sbi->s_group_desc[i]);
1898 kfree(sbi->s_group_desc);
1899failed_mount:
1900#ifdef CONFIG_QUOTA
1901 for (i = 0; i < MAXQUOTAS; i++)
1902 kfree(sbi->s_qf_names[i]);
1903#endif
617ba13b 1904 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 1905 brelse(bh);
1906out_fail:
1907 sb->s_fs_info = NULL;
1908 kfree(sbi);
1909 lock_kernel();
1910 return -EINVAL;
1911}
1912
1913/*
1914 * Setup any per-fs journal parameters now. We'll do this both on
1915 * initial mount, once the journal has been initialised but before we've
1916 * done any recovery; and again on any subsequent remount.
1917 */
617ba13b 1918static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 1919{
617ba13b 1920 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1921
1922 if (sbi->s_commit_interval)
1923 journal->j_commit_interval = sbi->s_commit_interval;
617ba13b 1924 /* We could also set up an ext4-specific default for the commit
ac27a0ec
DK		 1925 * interval here, but for now we'll just fall back to the jbd
1926 * default. */
1927
1928 spin_lock(&journal->j_state_lock);
1929 if (test_opt(sb, BARRIER))
dab291af 1930 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 1931 else
dab291af 1932 journal->j_flags &= ~JBD2_BARRIER;
ac27a0ec
DK		 1933 spin_unlock(&journal->j_state_lock);
1934}
1935
617ba13b 1936static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 1937 unsigned int journal_inum)
1938{
1939 struct inode *journal_inode;
1940 journal_t *journal;
1941
1942 /* First, test for the existence of a valid inode on disk. Bad
1943 * things happen if we iget() an unused inode, as the subsequent
1944 * iput() will try to delete it. */
1945
1946 journal_inode = iget(sb, journal_inum);
1947 if (!journal_inode) {
617ba13b 1948 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 1949 return NULL;
1950 }
1951 if (!journal_inode->i_nlink) {
1952 make_bad_inode(journal_inode);
1953 iput(journal_inode);
617ba13b 1954 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 1955 return NULL;
1956 }
1957
1958 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1959 journal_inode, journal_inode->i_size);
1960 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
617ba13b 1961 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 1962 iput(journal_inode);
1963 return NULL;
1964 }
1965
dab291af 1966 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 1967 if (!journal) {
617ba13b 1968 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 1969 iput(journal_inode);
1970 return NULL;
1971 }
1972 journal->j_private = sb;
617ba13b 1973 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 1974 return journal;
1975}
1976
617ba13b 1977static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 1978 dev_t j_dev)
1979{
1980 struct buffer_head * bh;
1981 journal_t *journal;
617ba13b
MC		 1982 ext4_fsblk_t start;
1983 ext4_fsblk_t len;
ac27a0ec 1984 int hblock, blocksize;
617ba13b 1985 ext4_fsblk_t sb_block;
ac27a0ec 1986 unsigned long offset;
617ba13b 1987 struct ext4_super_block * es;
ac27a0ec
DK		 1988 struct block_device *bdev;
1989
617ba13b 1990 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 1991 if (bdev == NULL)
1992 return NULL;
1993
1994 if (bd_claim(bdev, sb)) {
1995 printk(KERN_ERR
8c55e204 1996 "EXT4: failed to claim external journal device.\n");
ac27a0ec
DK		 1997 blkdev_put(bdev);
1998 return NULL;
1999 }
2000
2001 blocksize = sb->s_blocksize;
2002 hblock = bdev_hardsect_size(bdev);
2003 if (blocksize < hblock) {
2004 printk(KERN_ERR
617ba13b 2005 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2006 goto out_bdev;
2007 }
2008
617ba13b
MC		 2009 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2010 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2011 set_blocksize(bdev, blocksize);
2012 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2013 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2014 "external journal\n");
2015 goto out_bdev;
2016 }
2017
617ba13b
MC		 2018 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2019 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2020 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2021 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2022 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2023 "bad superblock\n");
2024 brelse(bh);
2025 goto out_bdev;
2026 }
2027
617ba13b
MC		 2028 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2029 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2030 brelse(bh);
2031 goto out_bdev;
2032 }
2033
bd81d8ee 2034 len = ext4_blocks_count(es);
ac27a0ec
DK		 2035 start = sb_block + 1;
2036 brelse(bh); /* we're done with the superblock */
2037
dab291af 2038 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2039 start, len, blocksize);
2040 if (!journal) {
617ba13b 2041 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2042 goto out_bdev;
2043 }
2044 journal->j_private = sb;
2045 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2046 wait_on_buffer(journal->j_sb_buffer);
2047 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2048 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2049 goto out_journal;
2050 }
2051 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2052 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2053 "user (unsupported) - %d\n",
2054 be32_to_cpu(journal->j_superblock->s_nr_users));
2055 goto out_journal;
2056 }
617ba13b
MC		 2057 EXT4_SB(sb)->journal_bdev = bdev;
2058 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2059 return journal;
2060out_journal:
dab291af 2061 jbd2_journal_destroy(journal);
ac27a0ec 2062out_bdev:
617ba13b 2063 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2064 return NULL;
2065}
2066
617ba13b
MC		 2067static int ext4_load_journal(struct super_block *sb,
2068 struct ext4_super_block *es,
ac27a0ec
DK		 2069 unsigned long journal_devnum)
2070{
2071 journal_t *journal;
2072 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2073 dev_t journal_dev;
2074 int err = 0;
2075 int really_read_only;
2076
2077 if (journal_devnum &&
2078 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2079 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2080 "numbers have changed\n");
2081 journal_dev = new_decode_dev(journal_devnum);
2082 } else
2083 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2084
2085 really_read_only = bdev_read_only(sb->s_bdev);
2086
2087 /*
2088 * Are we loading a blank journal or performing recovery after a
2089 * crash? For recovery, we need to check in advance whether we
2090 * can get read-write access to the device.
2091 */
2092
617ba13b 2093 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2094 if (sb->s_flags & MS_RDONLY) {
617ba13b 2095 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2096 "required on readonly filesystem.\n");
2097 if (really_read_only) {
617ba13b 2098 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2099 "unavailable, cannot proceed.\n");
2100 return -EROFS;
2101 }
617ba13b 2102 printk (KERN_INFO "EXT4-fs: write access will "
ac27a0ec
DK		 2103 "be enabled during recovery.\n");
2104 }
2105 }
2106
2107 if (journal_inum && journal_dev) {
617ba13b 2108 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2109 "and inode journals!\n");
2110 return -EINVAL;
2111 }
2112
2113 if (journal_inum) {
617ba13b 2114 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2115 return -EINVAL;
2116 } else {
617ba13b 2117 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2118 return -EINVAL;
2119 }
2120
2121 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2122 err = jbd2_journal_update_format(journal);
ac27a0ec 2123 if (err) {
617ba13b 2124 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2125 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2126 return err;
2127 }
2128 }
2129
617ba13b 2130 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2131 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2132 if (!err)
dab291af 2133 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2134
2135 if (err) {
617ba13b 2136 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2137 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2138 return err;
2139 }
2140
617ba13b
MC		 2141 EXT4_SB(sb)->s_journal = journal;
2142 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2143
2144 if (journal_devnum &&
2145 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2146 es->s_journal_dev = cpu_to_le32(journal_devnum);
2147 sb->s_dirt = 1;
2148
2149 /* Make sure we flush the recovery flag to disk. */
617ba13b 2150 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2151 }
2152
2153 return 0;
2154}
2155
617ba13b
MC		 2156static int ext4_create_journal(struct super_block * sb,
2157 struct ext4_super_block * es,
ac27a0ec
DK		 2158 unsigned int journal_inum)
2159{
2160 journal_t *journal;
6c675bd4 2161 int err;
ac27a0ec
DK		 2162
2163 if (sb->s_flags & MS_RDONLY) {
617ba13b 2164 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
ac27a0ec
DK		 2165 "create journal.\n");
2166 return -EROFS;
2167 }
2168
6c675bd4
BP		 2169 journal = ext4_get_journal(sb, journal_inum);
2170 if (!journal)
ac27a0ec
DK		 2171 return -EINVAL;
2172
617ba13b 2173 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
ac27a0ec
DK		 2174 journal_inum);
2175
6c675bd4
BP		 2176 err = jbd2_journal_create(journal);
2177 if (err) {
617ba13b 2178 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
dab291af 2179 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2180 return -EIO;
2181 }
2182
617ba13b 2183 EXT4_SB(sb)->s_journal = journal;
ac27a0ec 2184
617ba13b
MC		 2185 ext4_update_dynamic_rev(sb);
2186 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2187 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
ac27a0ec
DK		 2188
2189 es->s_journal_inum = cpu_to_le32(journal_inum);
2190 sb->s_dirt = 1;
2191
2192 /* Make sure we flush the recovery flag to disk. */
617ba13b 2193 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2194
2195 return 0;
2196}
2197
617ba13b
MC		 2198static void ext4_commit_super (struct super_block * sb,
2199 struct ext4_super_block * es,
ac27a0ec
DK		 2200 int sync)
2201{
617ba13b 2202 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
ac27a0ec
DK		 2203
2204 if (!sbh)
2205 return;
2206 es->s_wtime = cpu_to_le32(get_seconds());
bd81d8ee 2207 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
617ba13b 2208 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2209 BUFFER_TRACE(sbh, "marking dirty");
2210 mark_buffer_dirty(sbh);
2211 if (sync)
2212 sync_dirty_buffer(sbh);
2213}
2214
2215
2216/*
2217 * Have we just finished recovery? If so, and if we are mounting (or
2218 * remounting) the filesystem readonly, then we will end up with a
2219 * consistent fs on disk. Record that fact.
2220 */
617ba13b
MC		 2221static void ext4_mark_recovery_complete(struct super_block * sb,
2222 struct ext4_super_block * es)
ac27a0ec 2223{
617ba13b 2224 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2225
dab291af
MC		 2226 jbd2_journal_lock_updates(journal);
2227 jbd2_journal_flush(journal);
32c37730 2228 lock_super(sb);
617ba13b 2229 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2230 sb->s_flags & MS_RDONLY) {
617ba13b 2231 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2232 sb->s_dirt = 0;
617ba13b 2233 ext4_commit_super(sb, es, 1);
ac27a0ec 2234 }
32c37730 2235 unlock_super(sb);
dab291af 2236 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2237}
2238
2239/*
2240 * If we are mounting (or read-write remounting) a filesystem whose journal
2241 * has recorded an error from a previous lifetime, move that error to the
2242 * main filesystem now.
2243 */
617ba13b
MC		 2244static void ext4_clear_journal_err(struct super_block * sb,
2245 struct ext4_super_block * es)
ac27a0ec
DK		 2246{
2247 journal_t *journal;
2248 int j_errno;
2249 const char *errstr;
2250
617ba13b 2251 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2252
2253 /*
2254 * Now check for any error status which may have been recorded in the
617ba13b 2255 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2256 */
2257
dab291af 2258 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2259 if (j_errno) {
2260 char nbuf[16];
2261
617ba13b
MC		 2262 errstr = ext4_decode_error(sb, j_errno, nbuf);
2263 ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
ac27a0ec 2264 "from previous mount: %s", errstr);
617ba13b 2265 ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
ac27a0ec
DK		 2266 "filesystem check.");
2267
617ba13b
MC		 2268 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2269 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2270 ext4_commit_super (sb, es, 1);
ac27a0ec 2271
dab291af 2272 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 2273 }
2274}
2275
2276/*
2277 * Force the running and committing transactions to commit,
2278 * and wait on the commit.
2279 */
617ba13b 2280int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 2281{
2282 journal_t *journal;
2283 int ret;
2284
2285 if (sb->s_flags & MS_RDONLY)
2286 return 0;
2287
617ba13b 2288 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2289 sb->s_dirt = 0;
617ba13b 2290 ret = ext4_journal_force_commit(journal);
ac27a0ec
DK		 2291 return ret;
2292}
2293
2294/*
617ba13b 2295 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec
DK		 2296 * have to propagate any other updates to the superblock on disk at this
2297 * point. Just start an async writeback to get the buffers on their way
2298 * to the disk.
2299 *
2300 * This implicitly triggers the writebehind on sync().
2301 */
2302
617ba13b 2303static void ext4_write_super (struct super_block * sb)
ac27a0ec
DK		 2304{
2305 if (mutex_trylock(&sb->s_lock) != 0)
2306 BUG();
2307 sb->s_dirt = 0;
2308}
2309
617ba13b 2310static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec
DK		 2311{
2312 tid_t target;
2313
2314 sb->s_dirt = 0;
dab291af 2315 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
ac27a0ec 2316 if (wait)
dab291af 2317 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
ac27a0ec
DK		 2318 }
2319 return 0;
2320}
2321
2322/*
2323 * LVM calls this function before a (read-only) snapshot is created. This
2324 * gives us a chance to flush the journal completely and mark the fs clean.
2325 */
617ba13b 2326static void ext4_write_super_lockfs(struct super_block *sb)
ac27a0ec
DK		 2327{
2328 sb->s_dirt = 0;
2329
2330 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 2331 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2332
2333 /* Now we set up the journal barrier. */
dab291af
MC		 2334 jbd2_journal_lock_updates(journal);
2335 jbd2_journal_flush(journal);
ac27a0ec
DK		 2336
2337 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 2338 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2339 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec
DK		 2340 }
2341}
2342
2343/*
2344 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2345 * flag here, even though the filesystem is not technically dirty yet.
2346 */
617ba13b 2347static void ext4_unlockfs(struct super_block *sb)
ac27a0ec
DK		 2348{
2349 if (!(sb->s_flags & MS_RDONLY)) {
2350 lock_super(sb);
2351 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 2352 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2353 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 2354 unlock_super(sb);
dab291af 2355 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 2356 }
2357}
2358
617ba13b 2359static int ext4_remount (struct super_block * sb, int * flags, char * data)
ac27a0ec 2360{
617ba13b
MC		 2361 struct ext4_super_block * es;
2362 struct ext4_sb_info *sbi = EXT4_SB(sb);
2363 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 2364 unsigned long old_sb_flags;
617ba13b 2365 struct ext4_mount_options old_opts;
ac27a0ec
DK		 2366 int err;
2367#ifdef CONFIG_QUOTA
2368 int i;
2369#endif
2370
2371 /* Store the original options */
2372 old_sb_flags = sb->s_flags;
2373 old_opts.s_mount_opt = sbi->s_mount_opt;
2374 old_opts.s_resuid = sbi->s_resuid;
2375 old_opts.s_resgid = sbi->s_resgid;
2376 old_opts.s_commit_interval = sbi->s_commit_interval;
2377#ifdef CONFIG_QUOTA
2378 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2379 for (i = 0; i < MAXQUOTAS; i++)
2380 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2381#endif
2382
2383 /*
2384 * Allow the "check" option to be passed as a remount option.
2385 */
2386 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2387 err = -EINVAL;
2388 goto restore_opts;
2389 }
2390
617ba13b
MC		 2391 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2392 ext4_abort(sb, __FUNCTION__, "Abort forced by user");
ac27a0ec
DK		 2393
2394 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2395 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 2396
2397 es = sbi->s_es;
2398
617ba13b 2399 ext4_init_journal_params(sb, sbi->s_journal);
ac27a0ec
DK		 2400
2401 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 2402 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 2403 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 2404 err = -EROFS;
2405 goto restore_opts;
2406 }
2407
2408 if (*flags & MS_RDONLY) {
2409 /*
2410 * First of all, the unconditional stuff we have to do
2411 * to disable replay of the journal when we next remount
2412 */
2413 sb->s_flags |= MS_RDONLY;
2414
2415 /*
2416 * OK, test if we are remounting a valid rw partition
2417 * readonly, and if so set the rdonly flag and then
2418 * mark the partition as valid again.
2419 */
617ba13b
MC		 2420 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2421 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 2422 es->s_state = cpu_to_le16(sbi->s_mount_state);
2423
32c37730
JK		 2424 /*
2425 * We have to unlock super so that we can wait for
2426 * transactions.
2427 */
2428 unlock_super(sb);
617ba13b 2429 ext4_mark_recovery_complete(sb, es);
32c37730 2430 lock_super(sb);
ac27a0ec
DK		 2431 } else {
2432 __le32 ret;
617ba13b
MC		 2433 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2434 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2435 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec
DK		 2436 "remount RDWR because of unsupported "
2437 "optional features (%x).\n",
2438 sb->s_id, le32_to_cpu(ret));
2439 err = -EROFS;
2440 goto restore_opts;
2441 }
ead6596b
ES		 2442
2443 /*
2444 * If we have an unprocessed orphan list hanging
2445 * around from a previously readonly bdev mount,
2446 * require a full umount/remount for now.
2447 */
2448 if (es->s_last_orphan) {
2449 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2450 "remount RDWR because of unprocessed "
2451 "orphan inode list. Please "
2452 "umount/remount instead.\n",
2453 sb->s_id);
2454 err = -EINVAL;
2455 goto restore_opts;
2456 }
2457
ac27a0ec
DK		 2458 /*
2459 * Mounting a RDONLY partition read-write, so reread
2460 * and store the current valid flag. (It may have
2461 * been changed by e2fsck since we originally mounted
2462 * the partition.)
2463 */
617ba13b 2464 ext4_clear_journal_err(sb, es);
ac27a0ec 2465 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 2466 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 2467 goto restore_opts;
617ba13b 2468 if (!ext4_setup_super (sb, es, 0))
ac27a0ec
DK		 2469 sb->s_flags &= ~MS_RDONLY;
2470 }
2471 }
2472#ifdef CONFIG_QUOTA
2473 /* Release old quota file names */
2474 for (i = 0; i < MAXQUOTAS; i++)
2475 if (old_opts.s_qf_names[i] &&
2476 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2477 kfree(old_opts.s_qf_names[i]);
2478#endif
2479 return 0;
2480restore_opts:
2481 sb->s_flags = old_sb_flags;
2482 sbi->s_mount_opt = old_opts.s_mount_opt;
2483 sbi->s_resuid = old_opts.s_resuid;
2484 sbi->s_resgid = old_opts.s_resgid;
2485 sbi->s_commit_interval = old_opts.s_commit_interval;
2486#ifdef CONFIG_QUOTA
2487 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2488 for (i = 0; i < MAXQUOTAS; i++) {
2489 if (sbi->s_qf_names[i] &&
2490 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2491 kfree(sbi->s_qf_names[i]);
2492 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2493 }
2494#endif
2495 return err;
2496}
2497
617ba13b 2498static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
ac27a0ec
DK		 2499{
2500 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 2501 struct ext4_sb_info *sbi = EXT4_SB(sb);
2502 struct ext4_super_block *es = sbi->s_es;
960cc398 2503 u64 fsid;
ac27a0ec 2504
5e70030d
BP		 2505 if (test_opt(sb, MINIX_DF)) {
2506 sbi->s_overhead_last = 0;
2507 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2508 unsigned long ngroups = sbi->s_groups_count, i;
2509 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 2510 smp_rmb();
2511
2512 /*
5e70030d
BP		 2513 * Compute the overhead (FS structures). This is constant
2514 * for a given filesystem unless the number of block groups
2515 * changes so we cache the previous value until it does.
ac27a0ec
DK		 2516 */
2517
2518 /*
2519 * All of the blocks before first_data_block are
2520 * overhead
2521 */
2522 overhead = le32_to_cpu(es->s_first_data_block);
2523
2524 /*
2525 * Add the overhead attributed to the superblock and
2526 * block group descriptors. If the sparse superblocks
2527 * feature is turned on, then not all groups have this.
2528 */
2529 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 2530 overhead += ext4_bg_has_super(sb, i) +
2531 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 2532 cond_resched();
2533 }
2534
2535 /*
2536 * Every block group has an inode bitmap, a block
2537 * bitmap, and an inode table.
2538 */
5e70030d
BP		 2539 overhead += ngroups * (2 + sbi->s_itb_per_group);
2540 sbi->s_overhead_last = overhead;
2541 smp_wmb();
2542 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
ac27a0ec
DK		 2543 }
2544
617ba13b 2545 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 2546 buf->f_bsize = sb->s_blocksize;
5e70030d 2547 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
ac27a0ec 2548 buf->f_bfree = percpu_counter_sum(&sbi->s_freeblocks_counter);
5e70030d 2549 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
bd81d8ee
LV		 2550 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2551 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 2552 buf->f_bavail = 0;
2553 buf->f_files = le32_to_cpu(es->s_inodes_count);
2554 buf->f_ffree = percpu_counter_sum(&sbi->s_freeinodes_counter);
5e70030d 2555 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 2556 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 2557 fsid = le64_to_cpup((void *)es->s_uuid) ^
2558 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2559 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2560 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 2561 return 0;
2562}
2563
2564/* Helper function for writing quotas on sync - we need to start transaction before quota file
2565 * is locked for write. Otherwise the are possible deadlocks:
2566 * Process 1 Process 2
617ba13b 2567 * ext4_create() quota_sync()
dab291af 2568 * jbd2_journal_start() write_dquot()
ac27a0ec 2569 * DQUOT_INIT() down(dqio_mutex)
dab291af 2570 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 2571 *
2572 */
2573
2574#ifdef CONFIG_QUOTA
2575
2576static inline struct inode *dquot_to_inode(struct dquot *dquot)
2577{
2578 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2579}
2580
617ba13b 2581static int ext4_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 2582{
2583 handle_t *handle;
2584 int ret, err;
2585
2586 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 2587 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 2588 if (IS_ERR(handle))
2589 return PTR_ERR(handle);
2590 ret = dquot_initialize(inode, type);
617ba13b 2591 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2592 if (!ret)
2593 ret = err;
2594 return ret;
2595}
2596
617ba13b 2597static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 2598{
2599 handle_t *handle;
2600 int ret, err;
2601
2602 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 2603 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
ac27a0ec
DK		 2604 if (IS_ERR(handle))
2605 return PTR_ERR(handle);
2606 ret = dquot_drop(inode);
617ba13b 2607 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2608 if (!ret)
2609 ret = err;
2610 return ret;
2611}
2612
617ba13b 2613static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 2614{
2615 int ret, err;
2616 handle_t *handle;
2617 struct inode *inode;
2618
2619 inode = dquot_to_inode(dquot);
617ba13b
MC		 2620 handle = ext4_journal_start(inode,
2621 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 2622 if (IS_ERR(handle))
2623 return PTR_ERR(handle);
2624 ret = dquot_commit(dquot);
617ba13b 2625 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2626 if (!ret)
2627 ret = err;
2628 return ret;
2629}
2630
617ba13b 2631static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 2632{
2633 int ret, err;
2634 handle_t *handle;
2635
617ba13b
MC		 2636 handle = ext4_journal_start(dquot_to_inode(dquot),
2637 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 2638 if (IS_ERR(handle))
2639 return PTR_ERR(handle);
2640 ret = dquot_acquire(dquot);
617ba13b 2641 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2642 if (!ret)
2643 ret = err;
2644 return ret;
2645}
2646
617ba13b 2647static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 2648{
2649 int ret, err;
2650 handle_t *handle;
2651
617ba13b
MC		 2652 handle = ext4_journal_start(dquot_to_inode(dquot),
2653 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 2654 if (IS_ERR(handle))
2655 return PTR_ERR(handle);
2656 ret = dquot_release(dquot);
617ba13b 2657 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2658 if (!ret)
2659 ret = err;
2660 return ret;
2661}
2662
617ba13b 2663static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec
DK		 2664{
2665 /* Are we journalling quotas? */
617ba13b
MC		 2666 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2667 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 2668 dquot_mark_dquot_dirty(dquot);
617ba13b 2669 return ext4_write_dquot(dquot);
ac27a0ec
DK		 2670 } else {
2671 return dquot_mark_dquot_dirty(dquot);
2672 }
2673}
2674
617ba13b 2675static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 2676{
2677 int ret, err;
2678 handle_t *handle;
2679
2680 /* Data block + inode block */
617ba13b 2681 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 2682 if (IS_ERR(handle))
2683 return PTR_ERR(handle);
2684 ret = dquot_commit_info(sb, type);
617ba13b 2685 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2686 if (!ret)
2687 ret = err;
2688 return ret;
2689}
2690
2691/*
2692 * Turn on quotas during mount time - we need to find
2693 * the quota file and such...
2694 */
617ba13b 2695static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 2696{
617ba13b
MC		 2697 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
2698 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 2699}
2700
2701/*
2702 * Standard function to be called on quota_on
2703 */
617ba13b 2704static int ext4_quota_on(struct super_block *sb, int type, int format_id,
ac27a0ec
DK		 2705 char *path)
2706{
2707 int err;
2708 struct nameidata nd;
2709
2710 if (!test_opt(sb, QUOTA))
2711 return -EINVAL;
2712 /* Not journalling quota? */
617ba13b
MC		 2713 if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
2714 !EXT4_SB(sb)->s_qf_names[GRPQUOTA])
ac27a0ec
DK		 2715 return vfs_quota_on(sb, type, format_id, path);
2716 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2717 if (err)
2718 return err;
2719 /* Quotafile not on the same filesystem? */
2720 if (nd.mnt->mnt_sb != sb) {
2721 path_release(&nd);
2722 return -EXDEV;
2723 }
2724 /* Quotafile not of fs root? */
2725 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2726 printk(KERN_WARNING
617ba13b 2727 "EXT4-fs: Quota file not on filesystem root. "
ac27a0ec
DK		 2728 "Journalled quota will not work.\n");
2729 path_release(&nd);
2730 return vfs_quota_on(sb, type, format_id, path);
2731}
2732
2733/* Read data from quotafile - avoid pagecache and such because we cannot afford
2734 * acquiring the locks... As quota files are never truncated and quota code
2735 * itself serializes the operations (and noone else should touch the files)
2736 * we don't have to be afraid of races */
617ba13b 2737static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 2738 size_t len, loff_t off)
2739{
2740 struct inode *inode = sb_dqopt(sb)->files[type];
617ba13b 2741 sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 2742 int err = 0;
2743 int offset = off & (sb->s_blocksize - 1);
2744 int tocopy;
2745 size_t toread;
2746 struct buffer_head *bh;
2747 loff_t i_size = i_size_read(inode);
2748
2749 if (off > i_size)
2750 return 0;
2751 if (off+len > i_size)
2752 len = i_size-off;
2753 toread = len;
2754 while (toread > 0) {
2755 tocopy = sb->s_blocksize - offset < toread ?
2756 sb->s_blocksize - offset : toread;
617ba13b 2757 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 2758 if (err)
2759 return err;
2760 if (!bh) /* A hole? */
2761 memset(data, 0, tocopy);
2762 else
2763 memcpy(data, bh->b_data+offset, tocopy);
2764 brelse(bh);
2765 offset = 0;
2766 toread -= tocopy;
2767 data += tocopy;
2768 blk++;
2769 }
2770 return len;
2771}
2772
2773/* Write to quotafile (we know the transaction is already started and has
2774 * enough credits) */
617ba13b 2775static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 2776 const char *data, size_t len, loff_t off)
2777{
2778 struct inode *inode = sb_dqopt(sb)->files[type];
617ba13b 2779 sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 2780 int err = 0;
2781 int offset = off & (sb->s_blocksize - 1);
2782 int tocopy;
617ba13b 2783 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 2784 size_t towrite = len;
2785 struct buffer_head *bh;
2786 handle_t *handle = journal_current_handle();
2787
2788 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2789 while (towrite > 0) {
2790 tocopy = sb->s_blocksize - offset < towrite ?
2791 sb->s_blocksize - offset : towrite;
617ba13b 2792 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 2793 if (!bh)
2794 goto out;
2795 if (journal_quota) {
617ba13b 2796 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 2797 if (err) {
2798 brelse(bh);
2799 goto out;
2800 }
2801 }
2802 lock_buffer(bh);
2803 memcpy(bh->b_data+offset, data, tocopy);
2804 flush_dcache_page(bh->b_page);
2805 unlock_buffer(bh);
2806 if (journal_quota)
617ba13b 2807 err = ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK		 2808 else {
2809 /* Always do at least ordered writes for quotas */
617ba13b 2810 err = ext4_journal_dirty_data(handle, bh);
ac27a0ec
DK		 2811 mark_buffer_dirty(bh);
2812 }
2813 brelse(bh);
2814 if (err)
2815 goto out;
2816 offset = 0;
2817 towrite -= tocopy;
2818 data += tocopy;
2819 blk++;
2820 }
2821out:
2822 if (len == towrite)
2823 return err;
2824 if (inode->i_size < off+len-towrite) {
2825 i_size_write(inode, off+len-towrite);
617ba13b 2826 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec
DK		 2827 }
2828 inode->i_version++;
2829 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 2830 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 2831 mutex_unlock(&inode->i_mutex);
2832 return len - towrite;
2833}
2834
2835#endif
2836
617ba13b 2837static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 2838 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2839{
617ba13b 2840 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 2841}
2842
617ba13b 2843static struct file_system_type ext4dev_fs_type = {
ac27a0ec 2844 .owner = THIS_MODULE,
617ba13b
MC		 2845 .name = "ext4dev",
2846 .get_sb = ext4_get_sb,
ac27a0ec
DK		 2847 .kill_sb = kill_block_super,
2848 .fs_flags = FS_REQUIRES_DEV,
2849};
2850
617ba13b 2851static int __init init_ext4_fs(void)
ac27a0ec 2852{
617ba13b 2853 int err = init_ext4_xattr();
ac27a0ec
DK		 2854 if (err)
2855 return err;
2856 err = init_inodecache();
2857 if (err)
2858 goto out1;
63f57933 2859 err = register_filesystem(&ext4dev_fs_type);
ac27a0ec
DK		 2860 if (err)
2861 goto out;
2862 return 0;
2863out:
2864 destroy_inodecache();
2865out1:
617ba13b 2866 exit_ext4_xattr();
ac27a0ec
DK		 2867 return err;
2868}
2869
617ba13b 2870static void __exit exit_ext4_fs(void)
ac27a0ec 2871{
617ba13b 2872 unregister_filesystem(&ext4dev_fs_type);
ac27a0ec 2873 destroy_inodecache();
617ba13b 2874 exit_ext4_xattr();
ac27a0ec
DK		 2875}
2876
2877MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
617ba13b 2878MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
ac27a0ec 2879MODULE_LICENSE("GPL");
617ba13b
MC		 2880module_init(init_ext4_fs)
2881module_exit(exit_ext4_fs)
Clone of davem's net-next-2.6 tree