]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-jammy-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: correct mount option parsing to detect when quota options can be changed
[mirror_ubuntu-jammy-kernel.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
dab291af 23#include <linux/jbd2.h>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
27#include <linux/parser.h>
28#include <linux/smp_lock.h>
29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
1330593e 37#include <linux/log2.h>
717d50e4 38#include <linux/crc16.h>
ac27a0ec
DK		 39#include <asm/uaccess.h>
40
3dcf5451
CH		 41#include "ext4.h"
42#include "ext4_jbd2.h"
ac27a0ec
DK		 43#include "xattr.h"
44#include "acl.h"
45#include "namei.h"
717d50e4 46#include "group.h"
ac27a0ec 47
617ba13b 48static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 49 unsigned long journal_devnum);
617ba13b 50static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 51 unsigned int);
617ba13b
MC		 52static void ext4_commit_super (struct super_block * sb,
53 struct ext4_super_block * es,
ac27a0ec 54 int sync);
617ba13b
MC		 55static void ext4_mark_recovery_complete(struct super_block * sb,
56 struct ext4_super_block * es);
57static void ext4_clear_journal_err(struct super_block * sb,
58 struct ext4_super_block * es);
59static int ext4_sync_fs(struct super_block *sb, int wait);
60static const char *ext4_decode_error(struct super_block * sb, int errno,
ac27a0ec 61 char nbuf[16]);
617ba13b
MC		 62static int ext4_remount (struct super_block * sb, int * flags, char * data);
63static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
64static void ext4_unlockfs(struct super_block *sb);
65static void ext4_write_super (struct super_block * sb);
66static void ext4_write_super_lockfs(struct super_block *sb);
ac27a0ec 67
bd81d8ee 68
8fadc143
AR		 69ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
70 struct ext4_group_desc *bg)
bd81d8ee 71{
3a14589c 72 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 73 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
3a14589c 74 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 75}
76
8fadc143
AR		 77ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
78 struct ext4_group_desc *bg)
bd81d8ee 79{
5272f837 80 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 81 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 82 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 83}
84
8fadc143
AR		 85ext4_fsblk_t ext4_inode_table(struct super_block *sb,
86 struct ext4_group_desc *bg)
bd81d8ee 87{
5272f837 88 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 89 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 90 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 91}
92
8fadc143
AR		 93void ext4_block_bitmap_set(struct super_block *sb,
94 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 95{
3a14589c 96 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 97 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
98 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 99}
100
8fadc143
AR		 101void ext4_inode_bitmap_set(struct super_block *sb,
102 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 103{
5272f837 104 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 105 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
106 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 107}
108
8fadc143
AR		 109void ext4_inode_table_set(struct super_block *sb,
110 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 111{
5272f837 112 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 113 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
114 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 115}
116
ac27a0ec 117/*
dab291af 118 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 119 *
120 * The only special thing we need to do here is to make sure that all
121 * journal_end calls result in the superblock being marked dirty, so
122 * that sync() will call the filesystem's write_super callback if
123 * appropriate.
124 */
617ba13b 125handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 126{
127 journal_t *journal;
128
129 if (sb->s_flags & MS_RDONLY)
130 return ERR_PTR(-EROFS);
131
132 /* Special case here: if the journal has aborted behind our
133 * backs (eg. EIO in the commit thread), then we still need to
134 * take the FS itself readonly cleanly. */
617ba13b 135 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 136 if (is_journal_aborted(journal)) {
46e665e9 137 ext4_abort(sb, __func__,
ac27a0ec
DK		 138 "Detected aborted journal");
139 return ERR_PTR(-EROFS);
140 }
141
dab291af 142 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 143}
144
145/*
146 * The only special thing we need to do here is to make sure that all
dab291af 147 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 148 * that sync() will call the filesystem's write_super callback if
149 * appropriate.
150 */
617ba13b 151int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 152{
153 struct super_block *sb;
154 int err;
155 int rc;
156
157 sb = handle->h_transaction->t_journal->j_private;
158 err = handle->h_err;
dab291af 159 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 160
161 if (!err)
162 err = rc;
163 if (err)
617ba13b 164 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 165 return err;
166}
167
617ba13b 168void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 169 struct buffer_head *bh, handle_t *handle, int err)
170{
171 char nbuf[16];
617ba13b 172 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec
DK		 173
174 if (bh)
175 BUFFER_TRACE(bh, "abort");
176
177 if (!handle->h_err)
178 handle->h_err = err;
179
180 if (is_handle_aborted(handle))
181 return;
182
183 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
184 caller, errstr, err_fn);
185
dab291af 186 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 187}
188
189/* Deal with the reporting of failure conditions on a filesystem such as
190 * inconsistencies detected or read IO failures.
191 *
192 * On ext2, we can store the error state of the filesystem in the
617ba13b 193 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 194 * write ordering constraints on the superblock which prevent us from
195 * writing it out straight away; and given that the journal is about to
196 * be aborted, we can't rely on the current, or future, transactions to
197 * write out the superblock safely.
198 *
dab291af 199 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 200 * the journal instead. On recovery, the journal will compain about
201 * that error until we've noted it down and cleared it.
202 */
203
617ba13b 204static void ext4_handle_error(struct super_block *sb)
ac27a0ec 205{
617ba13b 206 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 207
617ba13b
MC		 208 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
209 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 210
211 if (sb->s_flags & MS_RDONLY)
212 return;
213
214 if (!test_opt (sb, ERRORS_CONT)) {
617ba13b 215 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 216
617ba13b 217 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ac27a0ec 218 if (journal)
dab291af 219 jbd2_journal_abort(journal, -EIO);
ac27a0ec
DK		 220 }
221 if (test_opt (sb, ERRORS_RO)) {
222 printk (KERN_CRIT "Remounting filesystem read-only\n");
223 sb->s_flags |= MS_RDONLY;
224 }
617ba13b 225 ext4_commit_super(sb, es, 1);
ac27a0ec 226 if (test_opt(sb, ERRORS_PANIC))
617ba13b 227 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 228 sb->s_id);
229}
230
617ba13b 231void ext4_error (struct super_block * sb, const char * function,
ac27a0ec
DK		 232 const char * fmt, ...)
233{
234 va_list args;
235
236 va_start(args, fmt);
617ba13b 237 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
ac27a0ec
DK		 238 vprintk(fmt, args);
239 printk("\n");
240 va_end(args);
241
617ba13b 242 ext4_handle_error(sb);
ac27a0ec
DK		 243}
244
617ba13b 245static const char *ext4_decode_error(struct super_block * sb, int errno,
ac27a0ec
DK		 246 char nbuf[16])
247{
248 char *errstr = NULL;
249
250 switch (errno) {
251 case -EIO:
252 errstr = "IO failure";
253 break;
254 case -ENOMEM:
255 errstr = "Out of memory";
256 break;
257 case -EROFS:
dab291af 258 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
ac27a0ec
DK		 259 errstr = "Journal has aborted";
260 else
261 errstr = "Readonly filesystem";
262 break;
263 default:
264 /* If the caller passed in an extra buffer for unknown
265 * errors, textualise them now. Else we just return
266 * NULL. */
267 if (nbuf) {
268 /* Check for truncated error codes... */
269 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
270 errstr = nbuf;
271 }
272 break;
273 }
274
275 return errstr;
276}
277
617ba13b 278/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 279 * automatically and invokes the appropriate error response. */
280
617ba13b 281void __ext4_std_error (struct super_block * sb, const char * function,
ac27a0ec
DK		 282 int errno)
283{
284 char nbuf[16];
285 const char *errstr;
286
287 /* Special case: if the error is EROFS, and we're not already
288 * inside a transaction, then there's really no point in logging
289 * an error. */
290 if (errno == -EROFS && journal_current_handle() == NULL &&
291 (sb->s_flags & MS_RDONLY))
292 return;
293
617ba13b
MC		 294 errstr = ext4_decode_error(sb, errno, nbuf);
295 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
ac27a0ec
DK		 296 sb->s_id, function, errstr);
297
617ba13b 298 ext4_handle_error(sb);
ac27a0ec
DK		 299}
300
301/*
617ba13b 302 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 303 * abort function may be used to deal with unrecoverable failures such
304 * as journal IO errors or ENOMEM at a critical moment in log management.
305 *
306 * We unconditionally force the filesystem into an ABORT|READONLY state,
307 * unless the error response on the fs has been set to panic in which
308 * case we take the easy way out and panic immediately.
309 */
310
617ba13b 311void ext4_abort (struct super_block * sb, const char * function,
ac27a0ec
DK		 312 const char * fmt, ...)
313{
314 va_list args;
315
617ba13b 316 printk (KERN_CRIT "ext4_abort called.\n");
ac27a0ec
DK		 317
318 va_start(args, fmt);
617ba13b 319 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
ac27a0ec
DK		 320 vprintk(fmt, args);
321 printk("\n");
322 va_end(args);
323
324 if (test_opt(sb, ERRORS_PANIC))
617ba13b 325 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 326
327 if (sb->s_flags & MS_RDONLY)
328 return;
329
330 printk(KERN_CRIT "Remounting filesystem read-only\n");
617ba13b 331 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 332 sb->s_flags |= MS_RDONLY;
617ba13b 333 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
dab291af 334 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 335}
336
617ba13b 337void ext4_warning (struct super_block * sb, const char * function,
ac27a0ec
DK		 338 const char * fmt, ...)
339{
340 va_list args;
341
342 va_start(args, fmt);
617ba13b 343 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 344 sb->s_id, function);
345 vprintk(fmt, args);
346 printk("\n");
347 va_end(args);
348}
349
617ba13b 350void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 351{
617ba13b 352 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 353
617ba13b 354 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 355 return;
356
46e665e9 357 ext4_warning(sb, __func__,
ac27a0ec
DK		 358 "updating to rev %d because of new feature flag, "
359 "running e2fsck is recommended",
617ba13b 360 EXT4_DYNAMIC_REV);
ac27a0ec 361
617ba13b
MC		 362 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
363 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
364 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 365 /* leave es->s_feature_*compat flags alone */
366 /* es->s_uuid will be set by e2fsck if empty */
367
368 /*
369 * The rest of the superblock fields should be zero, and if not it
370 * means they are likely already in use, so leave them alone. We
371 * can leave it up to e2fsck to clean up any inconsistencies there.
372 */
373}
374
99e6f829
AK		 375int ext4_update_compat_feature(handle_t *handle,
376 struct super_block *sb, __u32 compat)
377{
378 int err = 0;
379 if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
380 err = ext4_journal_get_write_access(handle,
381 EXT4_SB(sb)->s_sbh);
382 if (err)
383 return err;
384 EXT4_SET_COMPAT_FEATURE(sb, compat);
385 sb->s_dirt = 1;
386 handle->h_sync = 1;
387 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
388 "call ext4_journal_dirty_met adata");
389 err = ext4_journal_dirty_metadata(handle,
390 EXT4_SB(sb)->s_sbh);
391 }
392 return err;
393}
394
395int ext4_update_rocompat_feature(handle_t *handle,
396 struct super_block *sb, __u32 rocompat)
397{
398 int err = 0;
399 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
400 err = ext4_journal_get_write_access(handle,
401 EXT4_SB(sb)->s_sbh);
402 if (err)
403 return err;
404 EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
405 sb->s_dirt = 1;
406 handle->h_sync = 1;
407 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
408 "call ext4_journal_dirty_met adata");
409 err = ext4_journal_dirty_metadata(handle,
410 EXT4_SB(sb)->s_sbh);
411 }
412 return err;
413}
414
415int ext4_update_incompat_feature(handle_t *handle,
416 struct super_block *sb, __u32 incompat)
417{
418 int err = 0;
419 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
420 err = ext4_journal_get_write_access(handle,
421 EXT4_SB(sb)->s_sbh);
422 if (err)
423 return err;
424 EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
425 sb->s_dirt = 1;
426 handle->h_sync = 1;
427 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
428 "call ext4_journal_dirty_met adata");
429 err = ext4_journal_dirty_metadata(handle,
430 EXT4_SB(sb)->s_sbh);
431 }
432 return err;
433}
434
ac27a0ec
DK		 435/*
436 * Open the external journal device
437 */
617ba13b 438static struct block_device *ext4_blkdev_get(dev_t dev)
ac27a0ec
DK		 439{
440 struct block_device *bdev;
441 char b[BDEVNAME_SIZE];
442
443 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
444 if (IS_ERR(bdev))
445 goto fail;
446 return bdev;
447
448fail:
617ba13b 449 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
ac27a0ec
DK		 450 __bdevname(dev, b), PTR_ERR(bdev));
451 return NULL;
452}
453
454/*
455 * Release the journal device
456 */
617ba13b 457static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 458{
459 bd_release(bdev);
460 return blkdev_put(bdev);
461}
462
617ba13b 463static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 464{
465 struct block_device *bdev;
466 int ret = -ENODEV;
467
468 bdev = sbi->journal_bdev;
469 if (bdev) {
617ba13b 470 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 471 sbi->journal_bdev = NULL;
472 }
473 return ret;
474}
475
476static inline struct inode *orphan_list_entry(struct list_head *l)
477{
617ba13b 478 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 479}
480
617ba13b 481static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 482{
483 struct list_head *l;
484
485 printk(KERN_ERR "sb orphan head is %d\n",
486 le32_to_cpu(sbi->s_es->s_last_orphan));
487
488 printk(KERN_ERR "sb_info orphan list:\n");
489 list_for_each(l, &sbi->s_orphan) {
490 struct inode *inode = orphan_list_entry(l);
491 printk(KERN_ERR " "
492 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
493 inode->i_sb->s_id, inode->i_ino, inode,
494 inode->i_mode, inode->i_nlink,
495 NEXT_ORPHAN(inode));
496 }
497}
498
617ba13b 499static void ext4_put_super (struct super_block * sb)
ac27a0ec 500{
617ba13b
MC		 501 struct ext4_sb_info *sbi = EXT4_SB(sb);
502 struct ext4_super_block *es = sbi->s_es;
ac27a0ec
DK		 503 int i;
504
c9de560d 505 ext4_mb_release(sb);
a86c6181 506 ext4_ext_release(sb);
617ba13b 507 ext4_xattr_put_super(sb);
dab291af 508 jbd2_journal_destroy(sbi->s_journal);
ac27a0ec 509 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 510 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec
DK		 511 es->s_state = cpu_to_le16(sbi->s_mount_state);
512 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
513 mark_buffer_dirty(sbi->s_sbh);
617ba13b 514 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 515 }
516
517 for (i = 0; i < sbi->s_gdb_count; i++)
518 brelse(sbi->s_group_desc[i]);
519 kfree(sbi->s_group_desc);
520 percpu_counter_destroy(&sbi->s_freeblocks_counter);
521 percpu_counter_destroy(&sbi->s_freeinodes_counter);
522 percpu_counter_destroy(&sbi->s_dirs_counter);
523 brelse(sbi->s_sbh);
524#ifdef CONFIG_QUOTA
525 for (i = 0; i < MAXQUOTAS; i++)
526 kfree(sbi->s_qf_names[i]);
527#endif
528
529 /* Debugging code just in case the in-memory inode orphan list
530 * isn't empty. The on-disk one can be non-empty if we've
531 * detected an error and taken the fs readonly, but the
532 * in-memory list had better be clean by this point. */
533 if (!list_empty(&sbi->s_orphan))
534 dump_orphan_list(sb, sbi);
535 J_ASSERT(list_empty(&sbi->s_orphan));
536
f98393a6 537 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 538 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
539 /*
540 * Invalidate the journal device's buffers. We don't want them
541 * floating about in memory - the physical journal device may
542 * hotswapped, and it breaks the `ro-after' testing code.
543 */
544 sync_blockdev(sbi->journal_bdev);
f98393a6 545 invalidate_bdev(sbi->journal_bdev);
617ba13b 546 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 547 }
548 sb->s_fs_info = NULL;
549 kfree(sbi);
550 return;
551}
552
e18b890b 553static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 554
555/*
556 * Called inside transaction, so use GFP_NOFS
557 */
617ba13b 558static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 559{
617ba13b 560 struct ext4_inode_info *ei;
ac27a0ec 561
e6b4f8da 562 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 563 if (!ei)
564 return NULL;
617ba13b
MC		 565#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
566 ei->i_acl = EXT4_ACL_NOT_CACHED;
567 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 568#endif
569 ei->i_block_alloc_info = NULL;
570 ei->vfs_inode.i_version = 1;
a86c6181 571 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 572 INIT_LIST_HEAD(&ei->i_prealloc_list);
573 spin_lock_init(&ei->i_prealloc_lock);
ac27a0ec
DK		 574 return &ei->vfs_inode;
575}
576
617ba13b 577static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 578{
9f7dd93d
VA		 579 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
580 printk("EXT4 Inode %p: orphan list check failed!\n",
581 EXT4_I(inode));
582 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
583 EXT4_I(inode), sizeof(struct ext4_inode_info),
584 true);
585 dump_stack();
586 }
617ba13b 587 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 588}
589
4ba9b9d0 590static void init_once(struct kmem_cache *cachep, void *foo)
ac27a0ec 591{
617ba13b 592 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 593
a35afb83 594 INIT_LIST_HEAD(&ei->i_orphan);
617ba13b 595#ifdef CONFIG_EXT4DEV_FS_XATTR
a35afb83 596 init_rwsem(&ei->xattr_sem);
ac27a0ec 597#endif
0e855ac8 598 init_rwsem(&ei->i_data_sem);
a35afb83 599 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 600}
601
602static int init_inodecache(void)
603{
617ba13b
MC		 604 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
605 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 606 0, (SLAB_RECLAIM_ACCOUNT|
607 SLAB_MEM_SPREAD),
20c2df83 608 init_once);
617ba13b 609 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 610 return -ENOMEM;
611 return 0;
612}
613
614static void destroy_inodecache(void)
615{
617ba13b 616 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 617}
618
617ba13b 619static void ext4_clear_inode(struct inode *inode)
ac27a0ec 620{
617ba13b
MC		 621 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
622#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
623 if (EXT4_I(inode)->i_acl &&
624 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
625 posix_acl_release(EXT4_I(inode)->i_acl);
626 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
627 }
628 if (EXT4_I(inode)->i_default_acl &&
629 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
630 posix_acl_release(EXT4_I(inode)->i_default_acl);
631 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 632 }
633#endif
617ba13b
MC		 634 ext4_discard_reservation(inode);
635 EXT4_I(inode)->i_block_alloc_info = NULL;
ac27a0ec
DK		 636 if (unlikely(rsv))
637 kfree(rsv);
638}
639
617ba13b 640static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
ac27a0ec
DK		 641{
642#if defined(CONFIG_QUOTA)
617ba13b 643 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 644
645 if (sbi->s_jquota_fmt)
646 seq_printf(seq, ",jqfmt=%s",
647 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
648
649 if (sbi->s_qf_names[USRQUOTA])
650 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
651
652 if (sbi->s_qf_names[GRPQUOTA])
653 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
654
617ba13b 655 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 656 seq_puts(seq, ",usrquota");
657
617ba13b 658 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 659 seq_puts(seq, ",grpquota");
660#endif
661}
662
d9c9bef1
MS		 663/*
664 * Show an option if
665 * - it's set to a non-default value OR
666 * - if the per-sb default is different from the global default
667 */
617ba13b 668static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 669{
aa22df2c
AK		 670 int def_errors;
671 unsigned long def_mount_opts;
ac27a0ec 672 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 673 struct ext4_sb_info *sbi = EXT4_SB(sb);
674 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 675
676 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 677 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 678
679 if (sbi->s_sb_block != 1)
680 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
681 if (test_opt(sb, MINIX_DF))
682 seq_puts(seq, ",minixdf");
aa22df2c 683 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 684 seq_puts(seq, ",grpid");
685 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
686 seq_puts(seq, ",nogrpid");
687 if (sbi->s_resuid != EXT4_DEF_RESUID ||
688 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
689 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
690 }
691 if (sbi->s_resgid != EXT4_DEF_RESGID ||
692 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
693 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
694 }
bb4f397a 695 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 696 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 697 def_errors == EXT4_ERRORS_CONTINUE) {
698 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 699 }
700 }
aa22df2c 701 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 702 seq_puts(seq, ",errors=continue");
aa22df2c 703 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 704 seq_puts(seq, ",errors=panic");
aa22df2c 705 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 706 seq_puts(seq, ",nouid32");
aa22df2c 707 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 708 seq_puts(seq, ",debug");
709 if (test_opt(sb, OLDALLOC))
710 seq_puts(seq, ",oldalloc");
07620f69 711#ifdef CONFIG_EXT4DEV_FS_XATTR
aa22df2c
AK		 712 if (test_opt(sb, XATTR_USER) &&
713 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 714 seq_puts(seq, ",user_xattr");
715 if (!test_opt(sb, XATTR_USER) &&
716 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
717 seq_puts(seq, ",nouser_xattr");
718 }
719#endif
07620f69 720#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
aa22df2c 721 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 722 seq_puts(seq, ",acl");
723 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
724 seq_puts(seq, ",noacl");
725#endif
726 if (!test_opt(sb, RESERVATION))
727 seq_puts(seq, ",noreservation");
728 if (sbi->s_commit_interval) {
729 seq_printf(seq, ",commit=%u",
730 (unsigned) (sbi->s_commit_interval / HZ));
731 }
732 if (test_opt(sb, BARRIER))
733 seq_puts(seq, ",barrier=1");
734 if (test_opt(sb, NOBH))
735 seq_puts(seq, ",nobh");
736 if (!test_opt(sb, EXTENTS))
737 seq_puts(seq, ",noextents");
3dbd0ede
AK		 738 if (!test_opt(sb, MBALLOC))
739 seq_puts(seq, ",nomballoc");
25ec56b5
JNC		 740 if (test_opt(sb, I_VERSION))
741 seq_puts(seq, ",i_version");
ac27a0ec 742
cb45bbe4
MS		 743 if (sbi->s_stripe)
744 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 745 /*
746 * journal mode get enabled in different ways
747 * So just print the value even if we didn't specify it
748 */
617ba13b 749 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 750 seq_puts(seq, ",data=journal");
617ba13b 751 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 752 seq_puts(seq, ",data=ordered");
617ba13b 753 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 754 seq_puts(seq, ",data=writeback");
755
617ba13b 756 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 757 return 0;
758}
759
760
1b961ac0
CH		 761static struct inode *ext4_nfs_get_inode(struct super_block *sb,
762 u64 ino, u32 generation)
ac27a0ec 763{
ac27a0ec 764 struct inode *inode;
ac27a0ec 765
617ba13b 766 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 767 return ERR_PTR(-ESTALE);
617ba13b 768 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 769 return ERR_PTR(-ESTALE);
770
771 /* iget isn't really right if the inode is currently unallocated!!
772 *
617ba13b 773 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 774 * deleted, so we should be safe.
775 *
776 * Currently we don't know the generation for parent directory, so
777 * a generation of 0 means "accept any"
778 */
1d1fe1ee
DH		 779 inode = ext4_iget(sb, ino);
780 if (IS_ERR(inode))
781 return ERR_CAST(inode);
782 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 783 iput(inode);
784 return ERR_PTR(-ESTALE);
785 }
1b961ac0
CH		 786
787 return inode;
788}
789
790static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
791 int fh_len, int fh_type)
792{
793 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
794 ext4_nfs_get_inode);
795}
796
797static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
798 int fh_len, int fh_type)
799{
800 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
801 ext4_nfs_get_inode);
ac27a0ec
DK		 802}
803
804#ifdef CONFIG_QUOTA
805#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
806#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
807
617ba13b
MC		 808static int ext4_dquot_initialize(struct inode *inode, int type);
809static int ext4_dquot_drop(struct inode *inode);
810static int ext4_write_dquot(struct dquot *dquot);
811static int ext4_acquire_dquot(struct dquot *dquot);
812static int ext4_release_dquot(struct dquot *dquot);
813static int ext4_mark_dquot_dirty(struct dquot *dquot);
814static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 815static int ext4_quota_on(struct super_block *sb, int type, int format_id,
816 char *path, int remount);
617ba13b
MC		 817static int ext4_quota_on_mount(struct super_block *sb, int type);
818static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 819 size_t len, loff_t off);
617ba13b 820static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 821 const char *data, size_t len, loff_t off);
822
617ba13b
MC		 823static struct dquot_operations ext4_quota_operations = {
824 .initialize = ext4_dquot_initialize,
825 .drop = ext4_dquot_drop,
ac27a0ec
DK		 826 .alloc_space = dquot_alloc_space,
827 .alloc_inode = dquot_alloc_inode,
828 .free_space = dquot_free_space,
829 .free_inode = dquot_free_inode,
830 .transfer = dquot_transfer,
617ba13b
MC		 831 .write_dquot = ext4_write_dquot,
832 .acquire_dquot = ext4_acquire_dquot,
833 .release_dquot = ext4_release_dquot,
834 .mark_dirty = ext4_mark_dquot_dirty,
835 .write_info = ext4_write_info
ac27a0ec
DK		 836};
837
617ba13b
MC		 838static struct quotactl_ops ext4_qctl_operations = {
839 .quota_on = ext4_quota_on,
ac27a0ec
DK		 840 .quota_off = vfs_quota_off,
841 .quota_sync = vfs_quota_sync,
842 .get_info = vfs_get_dqinfo,
843 .set_info = vfs_set_dqinfo,
844 .get_dqblk = vfs_get_dqblk,
845 .set_dqblk = vfs_set_dqblk
846};
847#endif
848
ee9b6d61 849static const struct super_operations ext4_sops = {
617ba13b
MC		 850 .alloc_inode = ext4_alloc_inode,
851 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 852 .write_inode = ext4_write_inode,
853 .dirty_inode = ext4_dirty_inode,
854 .delete_inode = ext4_delete_inode,
855 .put_super = ext4_put_super,
856 .write_super = ext4_write_super,
857 .sync_fs = ext4_sync_fs,
858 .write_super_lockfs = ext4_write_super_lockfs,
859 .unlockfs = ext4_unlockfs,
860 .statfs = ext4_statfs,
861 .remount_fs = ext4_remount,
862 .clear_inode = ext4_clear_inode,
863 .show_options = ext4_show_options,
ac27a0ec 864#ifdef CONFIG_QUOTA
617ba13b
MC		 865 .quota_read = ext4_quota_read,
866 .quota_write = ext4_quota_write,
ac27a0ec
DK		 867#endif
868};
869
39655164 870static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 871 .fh_to_dentry = ext4_fh_to_dentry,
872 .fh_to_parent = ext4_fh_to_parent,
617ba13b 873 .get_parent = ext4_get_parent,
ac27a0ec
DK		 874};
875
876enum {
877 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
878 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
879 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
880 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
881 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
882 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
818d276c 883 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec
DK		 884 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
885 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
886 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
887 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
25ec56b5 888 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
c9de560d 889 Opt_mballoc, Opt_nomballoc, Opt_stripe,
ac27a0ec
DK		 890};
891
892static match_table_t tokens = {
893 {Opt_bsd_df, "bsddf"},
894 {Opt_minix_df, "minixdf"},
895 {Opt_grpid, "grpid"},
896 {Opt_grpid, "bsdgroups"},
897 {Opt_nogrpid, "nogrpid"},
898 {Opt_nogrpid, "sysvgroups"},
899 {Opt_resgid, "resgid=%u"},
900 {Opt_resuid, "resuid=%u"},
901 {Opt_sb, "sb=%u"},
902 {Opt_err_cont, "errors=continue"},
903 {Opt_err_panic, "errors=panic"},
904 {Opt_err_ro, "errors=remount-ro"},
905 {Opt_nouid32, "nouid32"},
906 {Opt_nocheck, "nocheck"},
907 {Opt_nocheck, "check=none"},
908 {Opt_debug, "debug"},
909 {Opt_oldalloc, "oldalloc"},
910 {Opt_orlov, "orlov"},
911 {Opt_user_xattr, "user_xattr"},
912 {Opt_nouser_xattr, "nouser_xattr"},
913 {Opt_acl, "acl"},
914 {Opt_noacl, "noacl"},
915 {Opt_reservation, "reservation"},
916 {Opt_noreservation, "noreservation"},
917 {Opt_noload, "noload"},
918 {Opt_nobh, "nobh"},
919 {Opt_bh, "bh"},
920 {Opt_commit, "commit=%u"},
921 {Opt_journal_update, "journal=update"},
922 {Opt_journal_inum, "journal=%u"},
923 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 924 {Opt_journal_checksum, "journal_checksum"},
925 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 926 {Opt_abort, "abort"},
927 {Opt_data_journal, "data=journal"},
928 {Opt_data_ordered, "data=ordered"},
929 {Opt_data_writeback, "data=writeback"},
930 {Opt_offusrjquota, "usrjquota="},
931 {Opt_usrjquota, "usrjquota=%s"},
932 {Opt_offgrpjquota, "grpjquota="},
933 {Opt_grpjquota, "grpjquota=%s"},
934 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
935 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
936 {Opt_grpquota, "grpquota"},
937 {Opt_noquota, "noquota"},
938 {Opt_quota, "quota"},
939 {Opt_usrquota, "usrquota"},
940 {Opt_barrier, "barrier=%u"},
a86c6181 941 {Opt_extents, "extents"},
1e2462f9 942 {Opt_noextents, "noextents"},
25ec56b5 943 {Opt_i_version, "i_version"},
c9de560d
AT		 944 {Opt_mballoc, "mballoc"},
945 {Opt_nomballoc, "nomballoc"},
946 {Opt_stripe, "stripe=%u"},
ac27a0ec 947 {Opt_resize, "resize"},
f3f12faa 948 {Opt_err, NULL},
ac27a0ec
DK		 949};
950
617ba13b 951static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 952{
617ba13b 953 ext4_fsblk_t sb_block;
ac27a0ec
DK		 954 char *options = (char *) *data;
955
956 if (!options || strncmp(options, "sb=", 3) != 0)
957 return 1; /* Default location */
958 options += 3;
617ba13b 959 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 960 sb_block = simple_strtoul(options, &options, 0);
961 if (*options && *options != ',') {
617ba13b 962 printk("EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 963 (char *) *data);
964 return 1;
965 }
966 if (*options == ',')
967 options++;
968 *data = (void *) options;
969 return sb_block;
970}
971
972static int parse_options (char *options, struct super_block *sb,
973 unsigned int *inum, unsigned long *journal_devnum,
617ba13b 974 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 975{
617ba13b 976 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 977 char * p;
978 substring_t args[MAX_OPT_ARGS];
979 int data_opt = 0;
980 int option;
981#ifdef CONFIG_QUOTA
dfc5d03f 982 int qtype, qfmt;
ac27a0ec
DK		 983 char *qname;
984#endif
985
986 if (!options)
987 return 1;
988
989 while ((p = strsep (&options, ",")) != NULL) {
990 int token;
991 if (!*p)
992 continue;
993
994 token = match_token(p, tokens, args);
995 switch (token) {
996 case Opt_bsd_df:
997 clear_opt (sbi->s_mount_opt, MINIX_DF);
998 break;
999 case Opt_minix_df:
1000 set_opt (sbi->s_mount_opt, MINIX_DF);
1001 break;
1002 case Opt_grpid:
1003 set_opt (sbi->s_mount_opt, GRPID);
1004 break;
1005 case Opt_nogrpid:
1006 clear_opt (sbi->s_mount_opt, GRPID);
1007 break;
1008 case Opt_resuid:
1009 if (match_int(&args[0], &option))
1010 return 0;
1011 sbi->s_resuid = option;
1012 break;
1013 case Opt_resgid:
1014 if (match_int(&args[0], &option))
1015 return 0;
1016 sbi->s_resgid = option;
1017 break;
1018 case Opt_sb:
1019 /* handled by get_sb_block() instead of here */
1020 /* *sb_block = match_int(&args[0]); */
1021 break;
1022 case Opt_err_panic:
1023 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1024 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1025 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1026 break;
1027 case Opt_err_ro:
1028 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1029 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1030 set_opt (sbi->s_mount_opt, ERRORS_RO);
1031 break;
1032 case Opt_err_cont:
1033 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1034 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1035 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1036 break;
1037 case Opt_nouid32:
1038 set_opt (sbi->s_mount_opt, NO_UID32);
1039 break;
1040 case Opt_nocheck:
1041 clear_opt (sbi->s_mount_opt, CHECK);
1042 break;
1043 case Opt_debug:
1044 set_opt (sbi->s_mount_opt, DEBUG);
1045 break;
1046 case Opt_oldalloc:
1047 set_opt (sbi->s_mount_opt, OLDALLOC);
1048 break;
1049 case Opt_orlov:
1050 clear_opt (sbi->s_mount_opt, OLDALLOC);
1051 break;
617ba13b 1052#ifdef CONFIG_EXT4DEV_FS_XATTR
ac27a0ec
DK		 1053 case Opt_user_xattr:
1054 set_opt (sbi->s_mount_opt, XATTR_USER);
1055 break;
1056 case Opt_nouser_xattr:
1057 clear_opt (sbi->s_mount_opt, XATTR_USER);
1058 break;
1059#else
1060 case Opt_user_xattr:
1061 case Opt_nouser_xattr:
617ba13b 1062 printk("EXT4 (no)user_xattr options not supported\n");
ac27a0ec
DK		 1063 break;
1064#endif
617ba13b 1065#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
ac27a0ec
DK		 1066 case Opt_acl:
1067 set_opt(sbi->s_mount_opt, POSIX_ACL);
1068 break;
1069 case Opt_noacl:
1070 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1071 break;
1072#else
1073 case Opt_acl:
1074 case Opt_noacl:
617ba13b 1075 printk("EXT4 (no)acl options not supported\n");
ac27a0ec
DK		 1076 break;
1077#endif
1078 case Opt_reservation:
1079 set_opt(sbi->s_mount_opt, RESERVATION);
1080 break;
1081 case Opt_noreservation:
1082 clear_opt(sbi->s_mount_opt, RESERVATION);
1083 break;
1084 case Opt_journal_update:
1085 /* @@@ FIXME */
1086 /* Eventually we will want to be able to create
1087 a journal file here. For now, only allow the
1088 user to specify an existing inode to be the
1089 journal file. */
1090 if (is_remount) {
617ba13b 1091 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1092 "journal on remount\n");
1093 return 0;
1094 }
1095 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1096 break;
1097 case Opt_journal_inum:
1098 if (is_remount) {
617ba13b 1099 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1100 "journal on remount\n");
1101 return 0;
1102 }
1103 if (match_int(&args[0], &option))
1104 return 0;
1105 *inum = option;
1106 break;
1107 case Opt_journal_dev:
1108 if (is_remount) {
617ba13b 1109 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1110 "journal on remount\n");
1111 return 0;
1112 }
1113 if (match_int(&args[0], &option))
1114 return 0;
1115 *journal_devnum = option;
1116 break;
818d276c
GS		 1117 case Opt_journal_checksum:
1118 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1119 break;
1120 case Opt_journal_async_commit:
1121 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1122 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1123 break;
ac27a0ec
DK		 1124 case Opt_noload:
1125 set_opt (sbi->s_mount_opt, NOLOAD);
1126 break;
1127 case Opt_commit:
1128 if (match_int(&args[0], &option))
1129 return 0;
1130 if (option < 0)
1131 return 0;
1132 if (option == 0)
cd02ff0b 1133 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1134 sbi->s_commit_interval = HZ * option;
1135 break;
1136 case Opt_data_journal:
617ba13b 1137 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1138 goto datacheck;
1139 case Opt_data_ordered:
617ba13b 1140 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1141 goto datacheck;
1142 case Opt_data_writeback:
617ba13b 1143 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1144 datacheck:
1145 if (is_remount) {
617ba13b 1146 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1147 != data_opt) {
1148 printk(KERN_ERR
617ba13b 1149 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1150 "mode on remount\n");
1151 return 0;
1152 }
1153 } else {
617ba13b 1154 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1155 sbi->s_mount_opt |= data_opt;
1156 }
1157 break;
1158#ifdef CONFIG_QUOTA
1159 case Opt_usrjquota:
1160 qtype = USRQUOTA;
1161 goto set_qf_name;
1162 case Opt_grpjquota:
1163 qtype = GRPQUOTA;
1164set_qf_name:
dfc5d03f
JK		 1165 if ((sb_any_quota_enabled(sb) ||
1166 sb_any_quota_suspended(sb)) &&
1167 !sbi->s_qf_names[qtype]) {
ac27a0ec 1168 printk(KERN_ERR
617ba13b 1169 "EXT4-fs: Cannot change journalled "
ac27a0ec
DK		 1170 "quota options when quota turned on.\n");
1171 return 0;
1172 }
1173 qname = match_strdup(&args[0]);
1174 if (!qname) {
1175 printk(KERN_ERR
617ba13b 1176 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1177 "storing quotafile name.\n");
1178 return 0;
1179 }
1180 if (sbi->s_qf_names[qtype] &&
1181 strcmp(sbi->s_qf_names[qtype], qname)) {
1182 printk(KERN_ERR
617ba13b 1183 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1184 "specified.\n", QTYPE2NAME(qtype));
1185 kfree(qname);
1186 return 0;
1187 }
1188 sbi->s_qf_names[qtype] = qname;
1189 if (strchr(sbi->s_qf_names[qtype], '/')) {
1190 printk(KERN_ERR
617ba13b 1191 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1192 "filesystem root.\n");
1193 kfree(sbi->s_qf_names[qtype]);
1194 sbi->s_qf_names[qtype] = NULL;
1195 return 0;
1196 }
1197 set_opt(sbi->s_mount_opt, QUOTA);
1198 break;
1199 case Opt_offusrjquota:
1200 qtype = USRQUOTA;
1201 goto clear_qf_name;
1202 case Opt_offgrpjquota:
1203 qtype = GRPQUOTA;
1204clear_qf_name:
dfc5d03f
JK		 1205 if ((sb_any_quota_enabled(sb) ||
1206 sb_any_quota_suspended(sb)) &&
1207 sbi->s_qf_names[qtype]) {
617ba13b 1208 printk(KERN_ERR "EXT4-fs: Cannot change "
ac27a0ec
DK		 1209 "journalled quota options when "
1210 "quota turned on.\n");
1211 return 0;
1212 }
1213 /*
1214 * The space will be released later when all options
1215 * are confirmed to be correct
1216 */
1217 sbi->s_qf_names[qtype] = NULL;
1218 break;
1219 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1220 qfmt = QFMT_VFS_OLD;
1221 goto set_qf_format;
ac27a0ec 1222 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1223 qfmt = QFMT_VFS_V0;
1224set_qf_format:
1225 if ((sb_any_quota_enabled(sb) ||
1226 sb_any_quota_suspended(sb)) &&
1227 sbi->s_jquota_fmt != qfmt) {
1228 printk(KERN_ERR "EXT4-fs: Cannot change "
1229 "journaled quota options when "
1230 "quota turned on.\n");
1231 return 0;
1232 }
1233 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1234 break;
1235 case Opt_quota:
1236 case Opt_usrquota:
1237 set_opt(sbi->s_mount_opt, QUOTA);
1238 set_opt(sbi->s_mount_opt, USRQUOTA);
1239 break;
1240 case Opt_grpquota:
1241 set_opt(sbi->s_mount_opt, QUOTA);
1242 set_opt(sbi->s_mount_opt, GRPQUOTA);
1243 break;
1244 case Opt_noquota:
1245 if (sb_any_quota_enabled(sb)) {
617ba13b 1246 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1247 "options when quota turned on.\n");
1248 return 0;
1249 }
1250 clear_opt(sbi->s_mount_opt, QUOTA);
1251 clear_opt(sbi->s_mount_opt, USRQUOTA);
1252 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1253 break;
1254#else
1255 case Opt_quota:
1256 case Opt_usrquota:
1257 case Opt_grpquota:
1258 case Opt_usrjquota:
1259 case Opt_grpjquota:
1260 case Opt_offusrjquota:
1261 case Opt_offgrpjquota:
1262 case Opt_jqfmt_vfsold:
1263 case Opt_jqfmt_vfsv0:
1264 printk(KERN_ERR
617ba13b 1265 "EXT4-fs: journalled quota options not "
ac27a0ec
DK		 1266 "supported.\n");
1267 break;
1268 case Opt_noquota:
1269 break;
1270#endif
1271 case Opt_abort:
1272 set_opt(sbi->s_mount_opt, ABORT);
1273 break;
1274 case Opt_barrier:
1275 if (match_int(&args[0], &option))
1276 return 0;
1277 if (option)
1278 set_opt(sbi->s_mount_opt, BARRIER);
1279 else
1280 clear_opt(sbi->s_mount_opt, BARRIER);
1281 break;
1282 case Opt_ignore:
1283 break;
1284 case Opt_resize:
1285 if (!is_remount) {
617ba13b 1286 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1287 "for remount\n");
1288 return 0;
1289 }
1290 if (match_int(&args[0], &option) != 0)
1291 return 0;
1292 *n_blocks_count = option;
1293 break;
1294 case Opt_nobh:
1295 set_opt(sbi->s_mount_opt, NOBH);
1296 break;
1297 case Opt_bh:
1298 clear_opt(sbi->s_mount_opt, NOBH);
1299 break;
a86c6181
AT		 1300 case Opt_extents:
1301 set_opt (sbi->s_mount_opt, EXTENTS);
1302 break;
1e2462f9
MC		 1303 case Opt_noextents:
1304 clear_opt (sbi->s_mount_opt, EXTENTS);
1305 break;
25ec56b5
JNC		 1306 case Opt_i_version:
1307 set_opt(sbi->s_mount_opt, I_VERSION);
1308 sb->s_flags |= MS_I_VERSION;
1309 break;
c9de560d
AT		 1310 case Opt_mballoc:
1311 set_opt(sbi->s_mount_opt, MBALLOC);
1312 break;
1313 case Opt_nomballoc:
1314 clear_opt(sbi->s_mount_opt, MBALLOC);
1315 break;
1316 case Opt_stripe:
1317 if (match_int(&args[0], &option))
1318 return 0;
1319 if (option < 0)
1320 return 0;
1321 sbi->s_stripe = option;
1322 break;
ac27a0ec
DK		 1323 default:
1324 printk (KERN_ERR
617ba13b 1325 "EXT4-fs: Unrecognized mount option \"%s\" "
ac27a0ec
DK		 1326 "or missing value\n", p);
1327 return 0;
1328 }
1329 }
1330#ifdef CONFIG_QUOTA
1331 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1332 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1333 sbi->s_qf_names[USRQUOTA])
1334 clear_opt(sbi->s_mount_opt, USRQUOTA);
1335
617ba13b 1336 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1337 sbi->s_qf_names[GRPQUOTA])
1338 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1339
1340 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1341 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1342 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1343 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1344 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1345 "format mixing.\n");
1346 return 0;
1347 }
1348
1349 if (!sbi->s_jquota_fmt) {
617ba13b 1350 printk(KERN_ERR "EXT4-fs: journalled quota format "
ac27a0ec
DK		 1351 "not specified.\n");
1352 return 0;
1353 }
1354 } else {
1355 if (sbi->s_jquota_fmt) {
617ba13b 1356 printk(KERN_ERR "EXT4-fs: journalled quota format "
ac27a0ec
DK		 1357 "specified with no journalling "
1358 "enabled.\n");
1359 return 0;
1360 }
1361 }
1362#endif
1363 return 1;
1364}
1365
617ba13b 1366static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1367 int read_only)
1368{
617ba13b 1369 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1370 int res = 0;
1371
617ba13b
MC		 1372 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1373 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
ac27a0ec
DK		 1374 "forcing read-only mode\n");
1375 res = MS_RDONLY;
1376 }
1377 if (read_only)
1378 return res;
617ba13b
MC		 1379 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1380 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
ac27a0ec 1381 "running e2fsck is recommended\n");
617ba13b 1382 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
ac27a0ec 1383 printk (KERN_WARNING
617ba13b 1384 "EXT4-fs warning: mounting fs with errors, "
ac27a0ec
DK		 1385 "running e2fsck is recommended\n");
1386 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1387 le16_to_cpu(es->s_mnt_count) >=
1388 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1389 printk (KERN_WARNING
617ba13b 1390 "EXT4-fs warning: maximal mount count reached, "
ac27a0ec
DK		 1391 "running e2fsck is recommended\n");
1392 else if (le32_to_cpu(es->s_checkinterval) &&
1393 (le32_to_cpu(es->s_lastcheck) +
1394 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1395 printk (KERN_WARNING
617ba13b 1396 "EXT4-fs warning: checktime reached, "
ac27a0ec
DK		 1397 "running e2fsck is recommended\n");
1398#if 0
1399 /* @@@ We _will_ want to clear the valid bit if we find
63f57933
AM		 1400 * inconsistencies, to force a fsck at reboot. But for
1401 * a plain journaled filesystem we can keep it set as
1402 * valid forever! :)
1403 */
216c34b2 1404 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec
DK		 1405#endif
1406 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1407 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1408 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1409 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b
MC		 1410 ext4_update_dynamic_rev(sb);
1411 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1412
617ba13b 1413 ext4_commit_super(sb, es, 1);
ac27a0ec 1414 if (test_opt(sb, DEBUG))
617ba13b 1415 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
ac27a0ec
DK		 1416 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1417 sb->s_blocksize,
1418 sbi->s_groups_count,
617ba13b
MC		 1419 EXT4_BLOCKS_PER_GROUP(sb),
1420 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1421 sbi->s_mount_opt);
1422
617ba13b
MC		 1423 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1424 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
ac27a0ec
DK		 1425 char b[BDEVNAME_SIZE];
1426
1427 printk("external journal on %s\n",
617ba13b 1428 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
ac27a0ec
DK		 1429 } else {
1430 printk("internal journal\n");
1431 }
1432 return res;
1433}
1434
717d50e4
AD		 1435__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1436 struct ext4_group_desc *gdp)
1437{
1438 __u16 crc = 0;
1439
1440 if (sbi->s_es->s_feature_ro_compat &
1441 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1442 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1443 __le32 le_group = cpu_to_le32(block_group);
1444
1445 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1446 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1447 crc = crc16(crc, (__u8 *)gdp, offset);
1448 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1449 /* for checksum of struct ext4_group_desc do the rest...*/
1450 if ((sbi->s_es->s_feature_incompat &
1451 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1452 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1453 crc = crc16(crc, (__u8 *)gdp + offset,
1454 le16_to_cpu(sbi->s_es->s_desc_size) -
1455 offset);
1456 }
1457
1458 return cpu_to_le16(crc);
1459}
1460
1461int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1462 struct ext4_group_desc *gdp)
1463{
1464 if ((sbi->s_es->s_feature_ro_compat &
1465 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1466 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1467 return 0;
1468
1469 return 1;
1470}
1471
ac27a0ec 1472/* Called at mount-time, super-block is locked */
197cd65a 1473static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1474{
617ba13b
MC		 1475 struct ext4_sb_info *sbi = EXT4_SB(sb);
1476 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1477 ext4_fsblk_t last_block;
bd81d8ee
LV		 1478 ext4_fsblk_t block_bitmap;
1479 ext4_fsblk_t inode_bitmap;
1480 ext4_fsblk_t inode_table;
ce421581 1481 int flexbg_flag = 0;
fd2d4291 1482 ext4_group_t i;
ac27a0ec 1483
ce421581
JS		 1484 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1485 flexbg_flag = 1;
1486
617ba13b 1487 ext4_debug ("Checking group descriptors");
ac27a0ec 1488
197cd65a
AM		 1489 for (i = 0; i < sbi->s_groups_count; i++) {
1490 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1491
ce421581 1492 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1493 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1494 else
1495 last_block = first_block +
617ba13b 1496 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1497
8fadc143 1498 block_bitmap = ext4_block_bitmap(sb, gdp);
bd81d8ee 1499 if (block_bitmap < first_block || block_bitmap > last_block)
ac27a0ec 1500 {
c19204b0
JB		 1501 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1502 "Block bitmap for group %lu not in group "
1503 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1504 return 0;
1505 }
8fadc143 1506 inode_bitmap = ext4_inode_bitmap(sb, gdp);
bd81d8ee 1507 if (inode_bitmap < first_block || inode_bitmap > last_block)
ac27a0ec 1508 {
c19204b0
JB		 1509 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1510 "Inode bitmap for group %lu not in group "
1511 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1512 return 0;
1513 }
8fadc143 1514 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1515 if (inode_table < first_block ||
780dcdb2 1516 inode_table + sbi->s_itb_per_group - 1 > last_block)
ac27a0ec 1517 {
c19204b0
JB		 1518 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1519 "Inode table for group %lu not in group "
1520 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1521 return 0;
1522 }
717d50e4 1523 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
c19204b0
JB		 1524 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1525 "Checksum for group %lu failed (%u!=%u)\n",
1526 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1527 gdp)), le16_to_cpu(gdp->bg_checksum));
717d50e4
AD		 1528 return 0;
1529 }
ce421581
JS		 1530 if (!flexbg_flag)
1531 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1532 }
1533
bd81d8ee 1534 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
617ba13b 1535 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1536 return 1;
1537}
1538
617ba13b 1539/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1540 * the superblock) which were deleted from all directories, but held open by
1541 * a process at the time of a crash. We walk the list and try to delete these
1542 * inodes at recovery time (only with a read-write filesystem).
1543 *
1544 * In order to keep the orphan inode chain consistent during traversal (in
1545 * case of crash during recovery), we link each inode into the superblock
1546 * orphan list_head and handle it the same way as an inode deletion during
1547 * normal operation (which journals the operations for us).
1548 *
1549 * We only do an iget() and an iput() on each inode, which is very safe if we
1550 * accidentally point at an in-use or already deleted inode. The worst that
1551 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1552 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1553 * e2fsck was run on this filesystem, and it must have already done the orphan
1554 * inode cleanup for us, so we can safely abort without any further action.
1555 */
617ba13b
MC		 1556static void ext4_orphan_cleanup (struct super_block * sb,
1557 struct ext4_super_block * es)
ac27a0ec
DK		 1558{
1559 unsigned int s_flags = sb->s_flags;
1560 int nr_orphans = 0, nr_truncates = 0;
1561#ifdef CONFIG_QUOTA
1562 int i;
1563#endif
1564 if (!es->s_last_orphan) {
1565 jbd_debug(4, "no orphan inodes to clean up\n");
1566 return;
1567 }
1568
a8f48a95
ES		 1569 if (bdev_read_only(sb->s_bdev)) {
1570 printk(KERN_ERR "EXT4-fs: write access "
1571 "unavailable, skipping orphan cleanup.\n");
1572 return;
1573 }
1574
617ba13b 1575 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1576 if (es->s_last_orphan)
1577 jbd_debug(1, "Errors on filesystem, "
1578 "clearing orphan list.\n");
1579 es->s_last_orphan = 0;
1580 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1581 return;
1582 }
1583
1584 if (s_flags & MS_RDONLY) {
617ba13b 1585 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1586 sb->s_id);
1587 sb->s_flags &= ~MS_RDONLY;
1588 }
1589#ifdef CONFIG_QUOTA
1590 /* Needed for iput() to work correctly and not trash data */
1591 sb->s_flags |= MS_ACTIVE;
1592 /* Turn on quotas so that they are updated correctly */
1593 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1594 if (EXT4_SB(sb)->s_qf_names[i]) {
1595 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1596 if (ret < 0)
1597 printk(KERN_ERR
617ba13b 1598 "EXT4-fs: Cannot turn on journalled "
ac27a0ec
DK		 1599 "quota: error %d\n", ret);
1600 }
1601 }
1602#endif
1603
1604 while (es->s_last_orphan) {
1605 struct inode *inode;
1606
97bd42b9
JB		 1607 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1608 if (IS_ERR(inode)) {
ac27a0ec
DK		 1609 es->s_last_orphan = 0;
1610 break;
1611 }
1612
617ba13b 1613 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1614 DQUOT_INIT(inode);
1615 if (inode->i_nlink) {
1616 printk(KERN_DEBUG
1617 "%s: truncating inode %lu to %Ld bytes\n",
46e665e9 1618 __func__, inode->i_ino, inode->i_size);
ac27a0ec
DK		 1619 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1620 inode->i_ino, inode->i_size);
617ba13b 1621 ext4_truncate(inode);
ac27a0ec
DK		 1622 nr_truncates++;
1623 } else {
1624 printk(KERN_DEBUG
1625 "%s: deleting unreferenced inode %lu\n",
46e665e9 1626 __func__, inode->i_ino);
ac27a0ec
DK		 1627 jbd_debug(2, "deleting unreferenced inode %lu\n",
1628 inode->i_ino);
1629 nr_orphans++;
1630 }
1631 iput(inode); /* The delete magic happens here! */
1632 }
1633
1634#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1635
1636 if (nr_orphans)
617ba13b 1637 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1638 sb->s_id, PLURAL(nr_orphans));
1639 if (nr_truncates)
617ba13b 1640 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1641 sb->s_id, PLURAL(nr_truncates));
1642#ifdef CONFIG_QUOTA
1643 /* Turn quotas off */
1644 for (i = 0; i < MAXQUOTAS; i++) {
1645 if (sb_dqopt(sb)->files[i])
6f28e087 1646 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1647 }
1648#endif
1649 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1650}
cd2291a4
ES		 1651/*
1652 * Maximal extent format file size.
1653 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1654 * extent format containers, within a sector_t, and within i_blocks
1655 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1656 * so that won't be a limiting factor.
1657 *
1658 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1659 */
1660static loff_t ext4_max_size(int blkbits)
1661{
1662 loff_t res;
1663 loff_t upper_limit = MAX_LFS_FILESIZE;
1664
1665 /* small i_blocks in vfs inode? */
1666 if (sizeof(blkcnt_t) < sizeof(u64)) {
1667 /*
1668 * CONFIG_LSF is not enabled implies the inode
1669 * i_block represent total blocks in 512 bytes
1670 * 32 == size of vfs inode i_blocks * 8
1671 */
1672 upper_limit = (1LL << 32) - 1;
1673
1674 /* total blocks in file system block size */
1675 upper_limit >>= (blkbits - 9);
1676 upper_limit <<= blkbits;
1677 }
1678
1679 /* 32-bit extent-start container, ee_block */
1680 res = 1LL << 32;
1681 res <<= blkbits;
1682 res -= 1;
1683
1684 /* Sanity check against vm- & vfs- imposed limits */
1685 if (res > upper_limit)
1686 res = upper_limit;
1687
1688 return res;
1689}
ac27a0ec 1690
ac27a0ec 1691/*
cd2291a4 1692 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1693 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1694 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1695 */
cd2291a4 1696static loff_t ext4_max_bitmap_size(int bits)
ac27a0ec 1697{
617ba13b 1698 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1699 int meta_blocks;
1700 loff_t upper_limit;
1701 /* This is calculated to be the largest file size for a
cd2291a4 1702 * dense, bitmapped file such that the total number of
ac27a0ec 1703 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1704 * does not exceed 2^48 -1
1705 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1706 * total number of 512 bytes blocks of the file
1707 */
1708
1709 if (sizeof(blkcnt_t) < sizeof(u64)) {
1710 /*
1711 * CONFIG_LSF is not enabled implies the inode
1712 * i_block represent total blocks in 512 bytes
1713 * 32 == size of vfs inode i_blocks * 8
1714 */
1715 upper_limit = (1LL << 32) - 1;
1716
1717 /* total blocks in file system block size */
1718 upper_limit >>= (bits - 9);
1719
1720 } else {
8180a562
AK		 1721 /*
1722 * We use 48 bit ext4_inode i_blocks
1723 * With EXT4_HUGE_FILE_FL set the i_blocks
1724 * represent total number of blocks in
1725 * file system block size
1726 */
0fc1b451
AK		 1727 upper_limit = (1LL << 48) - 1;
1728
0fc1b451
AK		 1729 }
1730
1731 /* indirect blocks */
1732 meta_blocks = 1;
1733 /* double indirect blocks */
1734 meta_blocks += 1 + (1LL << (bits-2));
1735 /* tripple indirect blocks */
1736 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1737
1738 upper_limit -= meta_blocks;
1739 upper_limit <<= bits;
ac27a0ec
DK		 1740
1741 res += 1LL << (bits-2);
1742 res += 1LL << (2*(bits-2));
1743 res += 1LL << (3*(bits-2));
1744 res <<= bits;
1745 if (res > upper_limit)
1746 res = upper_limit;
0fc1b451
AK		 1747
1748 if (res > MAX_LFS_FILESIZE)
1749 res = MAX_LFS_FILESIZE;
1750
ac27a0ec
DK		 1751 return res;
1752}
1753
617ba13b 1754static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1755 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1756{
617ba13b 1757 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1758 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1759 int has_super = 0;
1760
1761 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1762
617ba13b 1763 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1764 nr < first_meta_bg)
70bbb3e0 1765 return logical_sb_block + nr + 1;
ac27a0ec 1766 bg = sbi->s_desc_per_block * nr;
617ba13b 1767 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1768 has_super = 1;
617ba13b 1769 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1770}
1771
c9de560d
AT		 1772/**
1773 * ext4_get_stripe_size: Get the stripe size.
1774 * @sbi: In memory super block info
1775 *
1776 * If we have specified it via mount option, then
1777 * use the mount option value. If the value specified at mount time is
1778 * greater than the blocks per group use the super block value.
1779 * If the super block value is greater than blocks per group return 0.
1780 * Allocator needs it be less than blocks per group.
1781 *
1782 */
1783static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1784{
1785 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1786 unsigned long stripe_width =
1787 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1788
1789 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1790 return sbi->s_stripe;
1791
1792 if (stripe_width <= sbi->s_blocks_per_group)
1793 return stripe_width;
1794
1795 if (stride <= sbi->s_blocks_per_group)
1796 return stride;
1797
1798 return 0;
1799}
ac27a0ec 1800
617ba13b 1801static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1d03ec98
AK		 1802 __releases(kernel_sem)
1803 __acquires(kernel_sem)
1804
ac27a0ec
DK		 1805{
1806 struct buffer_head * bh;
617ba13b
MC		 1807 struct ext4_super_block *es = NULL;
1808 struct ext4_sb_info *sbi;
1809 ext4_fsblk_t block;
1810 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 1811 ext4_fsblk_t logical_sb_block;
ac27a0ec
DK		 1812 unsigned long offset = 0;
1813 unsigned int journal_inum = 0;
1814 unsigned long journal_devnum = 0;
1815 unsigned long def_mount_opts;
1816 struct inode *root;
1d1fe1ee 1817 int ret = -EINVAL;
ac27a0ec 1818 int blocksize;
ac27a0ec
DK		 1819 int db_count;
1820 int i;
1821 int needs_recovery;
1822 __le32 features;
bd81d8ee 1823 __u64 blocks_count;
833f4077 1824 int err;
ac27a0ec
DK		 1825
1826 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1827 if (!sbi)
1828 return -ENOMEM;
1829 sb->s_fs_info = sbi;
1830 sbi->s_mount_opt = 0;
617ba13b
MC		 1831 sbi->s_resuid = EXT4_DEF_RESUID;
1832 sbi->s_resgid = EXT4_DEF_RESGID;
d9c9bef1 1833 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 1834
1835 unlock_kernel();
1836
617ba13b 1837 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 1838 if (!blocksize) {
617ba13b 1839 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 1840 goto out_fail;
1841 }
1842
afc7cbca
TS		 1843 if (!sb_set_blocksize(sb, blocksize)) {
1844 printk(KERN_ERR "EXT4-fs: bad blocksize %d.\n", blocksize);
1845 goto out_fail;
1846 }
1847
ac27a0ec 1848 /*
617ba13b 1849 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 1850 * block sizes. We need to calculate the offset from buffer start.
1851 */
617ba13b 1852 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 1853 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1854 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 1855 } else {
70bbb3e0 1856 logical_sb_block = sb_block;
ac27a0ec
DK		 1857 }
1858
70bbb3e0 1859 if (!(bh = sb_bread(sb, logical_sb_block))) {
617ba13b 1860 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 1861 goto out_fail;
1862 }
1863 /*
1864 * Note: s_es must be initialized as soon as possible because
617ba13b 1865 * some ext4 macro-instructions depend on its value
ac27a0ec 1866 */
617ba13b 1867 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 1868 sbi->s_es = es;
1869 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 1870 if (sb->s_magic != EXT4_SUPER_MAGIC)
1871 goto cantfind_ext4;
ac27a0ec
DK		 1872
1873 /* Set defaults before we parse the mount options */
1874 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 1875 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 1876 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 1877 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 1878 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 1879 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 1880 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1881#ifdef CONFIG_EXT4DEV_FS_XATTR
617ba13b 1882 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 1883 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1884#endif
1885#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
617ba13b 1886 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 1887 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1888#endif
617ba13b
MC		 1889 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1890 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1891 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1892 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1893 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1894 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1895
1896 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 1897 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 1898 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 1899 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 1900 else
1901 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1902
1903 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1904 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1905
1906 set_opt(sbi->s_mount_opt, RESERVATION);
1907
1e2462f9
MC		 1908 /*
1909 * turn on extents feature by default in ext4 filesystem
1910 * User -o noextents to turn it off
1911 */
1912 set_opt(sbi->s_mount_opt, EXTENTS);
3dbd0ede
AK		 1913 /*
1914 * turn on mballoc feature by default in ext4 filesystem
1915 * User -o nomballoc to turn it off
1916 */
1917 set_opt(sbi->s_mount_opt, MBALLOC);
1e2462f9 1918
ac27a0ec
DK		 1919 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1920 NULL, 0))
1921 goto failed_mount;
1922
1923 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 1924 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 1925
617ba13b
MC		 1926 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1927 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1928 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1929 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 1930 printk(KERN_WARNING
617ba13b 1931 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec 1932 "running e2fsck is recommended\n");
469108ff
TT		 1933
1934 /*
1935 * Since ext4 is still considered development code, we require
1936 * that the TEST_FILESYS flag in s->flags be set.
1937 */
1938 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
1939 printk(KERN_WARNING "EXT4-fs: %s: not marked "
1940 "OK to use with test code.\n", sb->s_id);
1941 goto failed_mount;
1942 }
1943
ac27a0ec
DK		 1944 /*
1945 * Check feature flags regardless of the revision level, since we
1946 * previously didn't change the revision level when setting the flags,
1947 * so there is a chance incompat flags are set on a rev 0 filesystem.
1948 */
617ba13b 1949 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 1950 if (features) {
617ba13b 1951 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
ac27a0ec
DK		 1952 "unsupported optional features (%x).\n",
1953 sb->s_id, le32_to_cpu(features));
1954 goto failed_mount;
1955 }
617ba13b 1956 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 1957 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 1958 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
ac27a0ec
DK		 1959 "unsupported optional features (%x).\n",
1960 sb->s_id, le32_to_cpu(features));
1961 goto failed_mount;
1962 }
0fc1b451
AK		 1963 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
1964 /*
1965 * Large file size enabled file system can only be
1966 * mount if kernel is build with CONFIG_LSF
1967 */
1968 if (sizeof(root->i_blocks) < sizeof(u64) &&
1969 !(sb->s_flags & MS_RDONLY)) {
1970 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
1971 "files cannot be mounted read-write "
1972 "without CONFIG_LSF.\n", sb->s_id);
1973 goto failed_mount;
1974 }
1975 }
ac27a0ec
DK		 1976 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1977
617ba13b
MC		 1978 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1979 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 1980 printk(KERN_ERR
617ba13b 1981 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 1982 blocksize, sb->s_id);
1983 goto failed_mount;
1984 }
1985
ac27a0ec 1986 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 1987
1988 /* Validate the filesystem blocksize */
1989 if (!sb_set_blocksize(sb, blocksize)) {
1990 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
1991 blocksize);
ac27a0ec
DK		 1992 goto failed_mount;
1993 }
1994
1995 brelse (bh);
70bbb3e0
AM		 1996 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1997 offset = do_div(logical_sb_block, blocksize);
1998 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 1999 if (!bh) {
2000 printk(KERN_ERR
617ba13b 2001 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 2002 goto failed_mount;
2003 }
617ba13b 2004 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2005 sbi->s_es = es;
617ba13b 2006 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
ac27a0ec 2007 printk (KERN_ERR
617ba13b 2008 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 2009 goto failed_mount;
2010 }
2011 }
2012
e2b46574 2013 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
617ba13b 2014 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
ac27a0ec 2015
617ba13b
MC		 2016 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2017 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2018 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2019 } else {
2020 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2021 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2022 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2023 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec
DK		 2024 (sbi->s_inode_size > blocksize)) {
2025 printk (KERN_ERR
617ba13b 2026 "EXT4-fs: unsupported inode size: %d\n",
ac27a0ec
DK		 2027 sbi->s_inode_size);
2028 goto failed_mount;
2029 }
ef7f3835
KS		 2030 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2031 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2032 }
0d1ee42f
AR		 2033 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2034 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2035 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2036 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2037 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2038 printk(KERN_ERR
8fadc143 2039 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2040 sbi->s_desc_size);
2041 goto failed_mount;
2042 }
2043 } else
2044 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2045 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2046 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2047 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2048 goto cantfind_ext4;
2049 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2050 if (sbi->s_inodes_per_block == 0)
617ba13b 2051 goto cantfind_ext4;
ac27a0ec
DK		 2052 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2053 sbi->s_inodes_per_block;
0d1ee42f 2054 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2055 sbi->s_sbh = bh;
2056 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2057 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2058 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
ac27a0ec
DK		 2059 for (i=0; i < 4; i++)
2060 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2061 sbi->s_def_hash_version = es->s_def_hash_version;
2062
2063 if (sbi->s_blocks_per_group > blocksize * 8) {
2064 printk (KERN_ERR
617ba13b 2065 "EXT4-fs: #blocks per group too big: %lu\n",
ac27a0ec
DK		 2066 sbi->s_blocks_per_group);
2067 goto failed_mount;
2068 }
ac27a0ec
DK		 2069 if (sbi->s_inodes_per_group > blocksize * 8) {
2070 printk (KERN_ERR
617ba13b 2071 "EXT4-fs: #inodes per group too big: %lu\n",
ac27a0ec
DK		 2072 sbi->s_inodes_per_group);
2073 goto failed_mount;
2074 }
2075
bd81d8ee 2076 if (ext4_blocks_count(es) >
ac27a0ec 2077 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2078 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2079 " too large to mount safely\n", sb->s_id);
2080 if (sizeof(sector_t) < 8)
617ba13b 2081 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2082 "enabled\n");
2083 goto failed_mount;
2084 }
2085
617ba13b
MC		 2086 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2087 goto cantfind_ext4;
e7c95593
ES		 2088
2089 /* ensure blocks_count calculation below doesn't sign-extend */
2090 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2091 le32_to_cpu(es->s_first_data_block) + 1) {
2092 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2093 "first data block %u, blocks per group %lu\n",
2094 ext4_blocks_count(es),
2095 le32_to_cpu(es->s_first_data_block),
2096 EXT4_BLOCKS_PER_GROUP(sb));
2097 goto failed_mount;
2098 }
bd81d8ee
LV		 2099 blocks_count = (ext4_blocks_count(es) -
2100 le32_to_cpu(es->s_first_data_block) +
2101 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2102 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2103 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2104 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2105 EXT4_DESC_PER_BLOCK(sb);
ac27a0ec
DK		 2106 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2107 GFP_KERNEL);
2108 if (sbi->s_group_desc == NULL) {
617ba13b 2109 printk (KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2110 goto failed_mount;
2111 }
2112
2113 bgl_lock_init(&sbi->s_blockgroup_lock);
2114
2115 for (i = 0; i < db_count; i++) {
70bbb3e0 2116 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2117 sbi->s_group_desc[i] = sb_bread(sb, block);
2118 if (!sbi->s_group_desc[i]) {
617ba13b 2119 printk (KERN_ERR "EXT4-fs: "
ac27a0ec
DK		 2120 "can't read group descriptor %d\n", i);
2121 db_count = i;
2122 goto failed_mount2;
2123 }
2124 }
617ba13b
MC		 2125 if (!ext4_check_descriptors (sb)) {
2126 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2127 goto failed_mount2;
2128 }
2129 sbi->s_gdb_count = db_count;
2130 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2131 spin_lock_init(&sbi->s_next_gen_lock);
2132
833f4077
PZ		 2133 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2134 ext4_count_free_blocks(sb));
2135 if (!err) {
2136 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2137 ext4_count_free_inodes(sb));
2138 }
2139 if (!err) {
2140 err = percpu_counter_init(&sbi->s_dirs_counter,
2141 ext4_count_dirs(sb));
2142 }
2143 if (err) {
2144 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2145 goto failed_mount3;
2146 }
ac27a0ec
DK		 2147
2148 /* per fileystem reservation list head & lock */
2149 spin_lock_init(&sbi->s_rsv_window_lock);
2150 sbi->s_rsv_window_root = RB_ROOT;
2151 /* Add a single, static dummy reservation to the start of the
2152 * reservation window list --- it gives us a placeholder for
2153 * append-at-start-of-list which makes the allocation logic
2154 * _much_ simpler. */
617ba13b
MC		 2155 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2156 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
ac27a0ec
DK		 2157 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2158 sbi->s_rsv_window_head.rsv_goal_size = 0;
617ba13b 2159 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
ac27a0ec 2160
c9de560d
AT		 2161 sbi->s_stripe = ext4_get_stripe_size(sbi);
2162
ac27a0ec
DK		 2163 /*
2164 * set up enough so that it can read an inode
2165 */
617ba13b
MC		 2166 sb->s_op = &ext4_sops;
2167 sb->s_export_op = &ext4_export_ops;
2168 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2169#ifdef CONFIG_QUOTA
617ba13b
MC		 2170 sb->s_qcop = &ext4_qctl_operations;
2171 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2172#endif
2173 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2174
2175 sb->s_root = NULL;
2176
2177 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2178 EXT4_HAS_INCOMPAT_FEATURE(sb,
2179 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2180
2181 /*
2182 * The first inode we look at is the journal inode. Don't try
2183 * root first: it may be modified in the journal!
2184 */
2185 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2186 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2187 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec
DK		 2188 goto failed_mount3;
2189 } else if (journal_inum) {
617ba13b 2190 if (ext4_create_journal(sb, es, journal_inum))
ac27a0ec
DK		 2191 goto failed_mount3;
2192 } else {
2193 if (!silent)
2194 printk (KERN_ERR
617ba13b 2195 "ext4: No journal on filesystem on %s\n",
ac27a0ec
DK		 2196 sb->s_id);
2197 goto failed_mount3;
2198 }
2199
eb40a09c
JS		 2200 if (ext4_blocks_count(es) > 0xffffffffULL &&
2201 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2202 JBD2_FEATURE_INCOMPAT_64BIT)) {
2203 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2204 goto failed_mount4;
2205 }
2206
818d276c
GS		 2207 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2208 jbd2_journal_set_features(sbi->s_journal,
2209 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2210 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2211 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2212 jbd2_journal_set_features(sbi->s_journal,
2213 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2214 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2215 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2216 } else {
2217 jbd2_journal_clear_features(sbi->s_journal,
2218 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2219 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2220 }
2221
ac27a0ec
DK		 2222 /* We have now updated the journal if required, so we can
2223 * validate the data journaling mode. */
2224 switch (test_opt(sb, DATA_FLAGS)) {
2225 case 0:
2226 /* No mode set, assume a default based on the journal
63f57933
AM		 2227 * capabilities: ORDERED_DATA if the journal can
2228 * cope, else JOURNAL_DATA
2229 */
dab291af
MC		 2230 if (jbd2_journal_check_available_features
2231 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2232 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2233 else
2234 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2235 break;
2236
617ba13b
MC		 2237 case EXT4_MOUNT_ORDERED_DATA:
2238 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2239 if (!jbd2_journal_check_available_features
2240 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2241 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2242 "requested data journaling mode\n");
2243 goto failed_mount4;
2244 }
2245 default:
2246 break;
2247 }
2248
2249 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2250 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2251 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2252 "its supported only with writeback mode\n");
2253 clear_opt(sbi->s_mount_opt, NOBH);
2254 }
2255 }
2256 /*
dab291af 2257 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2258 * so we can safely mount the rest of the filesystem now.
2259 */
2260
1d1fe1ee
DH		 2261 root = ext4_iget(sb, EXT4_ROOT_INO);
2262 if (IS_ERR(root)) {
617ba13b 2263 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1d1fe1ee 2264 ret = PTR_ERR(root);
ac27a0ec
DK		 2265 goto failed_mount4;
2266 }
2267 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2268 iput(root);
617ba13b 2269 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2270 goto failed_mount4;
2271 }
1d1fe1ee
DH		 2272 sb->s_root = d_alloc_root(root);
2273 if (!sb->s_root) {
2274 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2275 iput(root);
2276 ret = -ENOMEM;
2277 goto failed_mount4;
2278 }
ac27a0ec 2279
617ba13b 2280 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2281
2282 /* determine the minimum size of new large inodes, if present */
2283 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2284 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2285 EXT4_GOOD_OLD_INODE_SIZE;
2286 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2287 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2288 if (sbi->s_want_extra_isize <
2289 le16_to_cpu(es->s_want_extra_isize))
2290 sbi->s_want_extra_isize =
2291 le16_to_cpu(es->s_want_extra_isize);
2292 if (sbi->s_want_extra_isize <
2293 le16_to_cpu(es->s_min_extra_isize))
2294 sbi->s_want_extra_isize =
2295 le16_to_cpu(es->s_min_extra_isize);
2296 }
2297 }
2298 /* Check if enough inode space is available */
2299 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2300 sbi->s_inode_size) {
2301 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2302 EXT4_GOOD_OLD_INODE_SIZE;
2303 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2304 "available.\n");
2305 }
2306
ac27a0ec
DK		 2307 /*
2308 * akpm: core read_super() calls in here with the superblock locked.
2309 * That deadlocks, because orphan cleanup needs to lock the superblock
2310 * in numerous places. Here we just pop the lock - it's relatively
2311 * harmless, because we are now ready to accept write_super() requests,
2312 * and aviro says that's the only reason for hanging onto the
2313 * superblock lock.
2314 */
617ba13b
MC		 2315 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2316 ext4_orphan_cleanup(sb, es);
2317 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
ac27a0ec 2318 if (needs_recovery)
617ba13b
MC		 2319 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2320 ext4_mark_recovery_complete(sb, es);
2321 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2322 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2323 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
ac27a0ec
DK		 2324 "writeback");
2325
a86c6181 2326 ext4_ext_init(sb);
c9de560d 2327 ext4_mb_init(sb, needs_recovery);
a86c6181 2328
ac27a0ec
DK		 2329 lock_kernel();
2330 return 0;
2331
617ba13b 2332cantfind_ext4:
ac27a0ec 2333 if (!silent)
617ba13b 2334 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2335 sb->s_id);
2336 goto failed_mount;
2337
2338failed_mount4:
dab291af 2339 jbd2_journal_destroy(sbi->s_journal);
ac27a0ec
DK		 2340failed_mount3:
2341 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2342 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2343 percpu_counter_destroy(&sbi->s_dirs_counter);
2344failed_mount2:
2345 for (i = 0; i < db_count; i++)
2346 brelse(sbi->s_group_desc[i]);
2347 kfree(sbi->s_group_desc);
2348failed_mount:
2349#ifdef CONFIG_QUOTA
2350 for (i = 0; i < MAXQUOTAS; i++)
2351 kfree(sbi->s_qf_names[i]);
2352#endif
617ba13b 2353 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2354 brelse(bh);
2355out_fail:
2356 sb->s_fs_info = NULL;
2357 kfree(sbi);
2358 lock_kernel();
1d1fe1ee 2359 return ret;
ac27a0ec
DK		 2360}
2361
2362/*
2363 * Setup any per-fs journal parameters now. We'll do this both on
2364 * initial mount, once the journal has been initialised but before we've
2365 * done any recovery; and again on any subsequent remount.
2366 */
617ba13b 2367static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2368{
617ba13b 2369 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2370
2371 if (sbi->s_commit_interval)
2372 journal->j_commit_interval = sbi->s_commit_interval;
617ba13b 2373 /* We could also set up an ext4-specific default for the commit
ac27a0ec
DK		 2374 * interval here, but for now we'll just fall back to the jbd
2375 * default. */
2376
2377 spin_lock(&journal->j_state_lock);
2378 if (test_opt(sb, BARRIER))
dab291af 2379 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2380 else
dab291af 2381 journal->j_flags &= ~JBD2_BARRIER;
ac27a0ec
DK		 2382 spin_unlock(&journal->j_state_lock);
2383}
2384
617ba13b 2385static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2386 unsigned int journal_inum)
2387{
2388 struct inode *journal_inode;
2389 journal_t *journal;
2390
2391 /* First, test for the existence of a valid inode on disk. Bad
2392 * things happen if we iget() an unused inode, as the subsequent
2393 * iput() will try to delete it. */
2394
1d1fe1ee
DH		 2395 journal_inode = ext4_iget(sb, journal_inum);
2396 if (IS_ERR(journal_inode)) {
617ba13b 2397 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2398 return NULL;
2399 }
2400 if (!journal_inode->i_nlink) {
2401 make_bad_inode(journal_inode);
2402 iput(journal_inode);
617ba13b 2403 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2404 return NULL;
2405 }
2406
2407 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2408 journal_inode, journal_inode->i_size);
1d1fe1ee 2409 if (!S_ISREG(journal_inode->i_mode)) {
617ba13b 2410 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2411 iput(journal_inode);
2412 return NULL;
2413 }
2414
dab291af 2415 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2416 if (!journal) {
617ba13b 2417 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2418 iput(journal_inode);
2419 return NULL;
2420 }
2421 journal->j_private = sb;
617ba13b 2422 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2423 return journal;
2424}
2425
617ba13b 2426static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2427 dev_t j_dev)
2428{
2429 struct buffer_head * bh;
2430 journal_t *journal;
617ba13b
MC		 2431 ext4_fsblk_t start;
2432 ext4_fsblk_t len;
ac27a0ec 2433 int hblock, blocksize;
617ba13b 2434 ext4_fsblk_t sb_block;
ac27a0ec 2435 unsigned long offset;
617ba13b 2436 struct ext4_super_block * es;
ac27a0ec
DK		 2437 struct block_device *bdev;
2438
617ba13b 2439 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2440 if (bdev == NULL)
2441 return NULL;
2442
2443 if (bd_claim(bdev, sb)) {
2444 printk(KERN_ERR
8c55e204 2445 "EXT4: failed to claim external journal device.\n");
ac27a0ec
DK		 2446 blkdev_put(bdev);
2447 return NULL;
2448 }
2449
2450 blocksize = sb->s_blocksize;
2451 hblock = bdev_hardsect_size(bdev);
2452 if (blocksize < hblock) {
2453 printk(KERN_ERR
617ba13b 2454 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2455 goto out_bdev;
2456 }
2457
617ba13b
MC		 2458 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2459 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2460 set_blocksize(bdev, blocksize);
2461 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2462 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2463 "external journal\n");
2464 goto out_bdev;
2465 }
2466
617ba13b
MC		 2467 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2468 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2469 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2470 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2471 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2472 "bad superblock\n");
2473 brelse(bh);
2474 goto out_bdev;
2475 }
2476
617ba13b
MC		 2477 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2478 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2479 brelse(bh);
2480 goto out_bdev;
2481 }
2482
bd81d8ee 2483 len = ext4_blocks_count(es);
ac27a0ec
DK		 2484 start = sb_block + 1;
2485 brelse(bh); /* we're done with the superblock */
2486
dab291af 2487 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2488 start, len, blocksize);
2489 if (!journal) {
617ba13b 2490 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2491 goto out_bdev;
2492 }
2493 journal->j_private = sb;
2494 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2495 wait_on_buffer(journal->j_sb_buffer);
2496 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2497 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2498 goto out_journal;
2499 }
2500 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2501 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2502 "user (unsupported) - %d\n",
2503 be32_to_cpu(journal->j_superblock->s_nr_users));
2504 goto out_journal;
2505 }
617ba13b
MC		 2506 EXT4_SB(sb)->journal_bdev = bdev;
2507 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2508 return journal;
2509out_journal:
dab291af 2510 jbd2_journal_destroy(journal);
ac27a0ec 2511out_bdev:
617ba13b 2512 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2513 return NULL;
2514}
2515
617ba13b
MC		 2516static int ext4_load_journal(struct super_block *sb,
2517 struct ext4_super_block *es,
ac27a0ec
DK		 2518 unsigned long journal_devnum)
2519{
2520 journal_t *journal;
2521 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2522 dev_t journal_dev;
2523 int err = 0;
2524 int really_read_only;
2525
2526 if (journal_devnum &&
2527 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2528 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2529 "numbers have changed\n");
2530 journal_dev = new_decode_dev(journal_devnum);
2531 } else
2532 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2533
2534 really_read_only = bdev_read_only(sb->s_bdev);
2535
2536 /*
2537 * Are we loading a blank journal or performing recovery after a
2538 * crash? For recovery, we need to check in advance whether we
2539 * can get read-write access to the device.
2540 */
2541
617ba13b 2542 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2543 if (sb->s_flags & MS_RDONLY) {
617ba13b 2544 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2545 "required on readonly filesystem.\n");
2546 if (really_read_only) {
617ba13b 2547 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2548 "unavailable, cannot proceed.\n");
2549 return -EROFS;
2550 }
617ba13b 2551 printk (KERN_INFO "EXT4-fs: write access will "
ac27a0ec
DK		 2552 "be enabled during recovery.\n");
2553 }
2554 }
2555
2556 if (journal_inum && journal_dev) {
617ba13b 2557 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2558 "and inode journals!\n");
2559 return -EINVAL;
2560 }
2561
2562 if (journal_inum) {
617ba13b 2563 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2564 return -EINVAL;
2565 } else {
617ba13b 2566 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2567 return -EINVAL;
2568 }
2569
2570 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2571 err = jbd2_journal_update_format(journal);
ac27a0ec 2572 if (err) {
617ba13b 2573 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2574 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2575 return err;
2576 }
2577 }
2578
617ba13b 2579 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2580 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2581 if (!err)
dab291af 2582 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2583
2584 if (err) {
617ba13b 2585 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2586 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2587 return err;
2588 }
2589
617ba13b
MC		 2590 EXT4_SB(sb)->s_journal = journal;
2591 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2592
2593 if (journal_devnum &&
2594 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2595 es->s_journal_dev = cpu_to_le32(journal_devnum);
2596 sb->s_dirt = 1;
2597
2598 /* Make sure we flush the recovery flag to disk. */
617ba13b 2599 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2600 }
2601
2602 return 0;
2603}
2604
617ba13b
MC		 2605static int ext4_create_journal(struct super_block * sb,
2606 struct ext4_super_block * es,
ac27a0ec
DK		 2607 unsigned int journal_inum)
2608{
2609 journal_t *journal;
6c675bd4 2610 int err;
ac27a0ec
DK		 2611
2612 if (sb->s_flags & MS_RDONLY) {
617ba13b 2613 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
ac27a0ec
DK		 2614 "create journal.\n");
2615 return -EROFS;
2616 }
2617
6c675bd4
BP		 2618 journal = ext4_get_journal(sb, journal_inum);
2619 if (!journal)
ac27a0ec
DK		 2620 return -EINVAL;
2621
617ba13b 2622 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
ac27a0ec
DK		 2623 journal_inum);
2624
6c675bd4
BP		 2625 err = jbd2_journal_create(journal);
2626 if (err) {
617ba13b 2627 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
dab291af 2628 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2629 return -EIO;
2630 }
2631
617ba13b 2632 EXT4_SB(sb)->s_journal = journal;
ac27a0ec 2633
617ba13b
MC		 2634 ext4_update_dynamic_rev(sb);
2635 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2636 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
ac27a0ec
DK		 2637
2638 es->s_journal_inum = cpu_to_le32(journal_inum);
2639 sb->s_dirt = 1;
2640
2641 /* Make sure we flush the recovery flag to disk. */
617ba13b 2642 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2643
2644 return 0;
2645}
2646
617ba13b
MC		 2647static void ext4_commit_super (struct super_block * sb,
2648 struct ext4_super_block * es,
ac27a0ec
DK		 2649 int sync)
2650{
617ba13b 2651 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
ac27a0ec
DK		 2652
2653 if (!sbh)
2654 return;
2655 es->s_wtime = cpu_to_le32(get_seconds());
bd81d8ee 2656 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
617ba13b 2657 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2658 BUFFER_TRACE(sbh, "marking dirty");
2659 mark_buffer_dirty(sbh);
2660 if (sync)
2661 sync_dirty_buffer(sbh);
2662}
2663
2664
2665/*
2666 * Have we just finished recovery? If so, and if we are mounting (or
2667 * remounting) the filesystem readonly, then we will end up with a
2668 * consistent fs on disk. Record that fact.
2669 */
617ba13b
MC		 2670static void ext4_mark_recovery_complete(struct super_block * sb,
2671 struct ext4_super_block * es)
ac27a0ec 2672{
617ba13b 2673 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2674
dab291af
MC		 2675 jbd2_journal_lock_updates(journal);
2676 jbd2_journal_flush(journal);
32c37730 2677 lock_super(sb);
617ba13b 2678 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2679 sb->s_flags & MS_RDONLY) {
617ba13b 2680 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2681 sb->s_dirt = 0;
617ba13b 2682 ext4_commit_super(sb, es, 1);
ac27a0ec 2683 }
32c37730 2684 unlock_super(sb);
dab291af 2685 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2686}
2687
2688/*
2689 * If we are mounting (or read-write remounting) a filesystem whose journal
2690 * has recorded an error from a previous lifetime, move that error to the
2691 * main filesystem now.
2692 */
617ba13b
MC		 2693static void ext4_clear_journal_err(struct super_block * sb,
2694 struct ext4_super_block * es)
ac27a0ec
DK		 2695{
2696 journal_t *journal;
2697 int j_errno;
2698 const char *errstr;
2699
617ba13b 2700 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2701
2702 /*
2703 * Now check for any error status which may have been recorded in the
617ba13b 2704 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2705 */
2706
dab291af 2707 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2708 if (j_errno) {
2709 char nbuf[16];
2710
617ba13b 2711 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 2712 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 2713 "from previous mount: %s", errstr);
46e665e9 2714 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 2715 "filesystem check.");
2716
617ba13b
MC		 2717 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2718 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2719 ext4_commit_super (sb, es, 1);
ac27a0ec 2720
dab291af 2721 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 2722 }
2723}
2724
2725/*
2726 * Force the running and committing transactions to commit,
2727 * and wait on the commit.
2728 */
617ba13b 2729int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 2730{
2731 journal_t *journal;
2732 int ret;
2733
2734 if (sb->s_flags & MS_RDONLY)
2735 return 0;
2736
617ba13b 2737 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2738 sb->s_dirt = 0;
617ba13b 2739 ret = ext4_journal_force_commit(journal);
ac27a0ec
DK		 2740 return ret;
2741}
2742
2743/*
617ba13b 2744 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec
DK		 2745 * have to propagate any other updates to the superblock on disk at this
2746 * point. Just start an async writeback to get the buffers on their way
2747 * to the disk.
2748 *
2749 * This implicitly triggers the writebehind on sync().
2750 */
2751
617ba13b 2752static void ext4_write_super (struct super_block * sb)
ac27a0ec
DK		 2753{
2754 if (mutex_trylock(&sb->s_lock) != 0)
2755 BUG();
2756 sb->s_dirt = 0;
2757}
2758
617ba13b 2759static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec
DK		 2760{
2761 tid_t target;
2762
2763 sb->s_dirt = 0;
dab291af 2764 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
ac27a0ec 2765 if (wait)
dab291af 2766 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
ac27a0ec
DK		 2767 }
2768 return 0;
2769}
2770
2771/*
2772 * LVM calls this function before a (read-only) snapshot is created. This
2773 * gives us a chance to flush the journal completely and mark the fs clean.
2774 */
617ba13b 2775static void ext4_write_super_lockfs(struct super_block *sb)
ac27a0ec
DK		 2776{
2777 sb->s_dirt = 0;
2778
2779 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 2780 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2781
2782 /* Now we set up the journal barrier. */
dab291af
MC		 2783 jbd2_journal_lock_updates(journal);
2784 jbd2_journal_flush(journal);
ac27a0ec
DK		 2785
2786 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 2787 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2788 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec
DK		 2789 }
2790}
2791
2792/*
2793 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2794 * flag here, even though the filesystem is not technically dirty yet.
2795 */
617ba13b 2796static void ext4_unlockfs(struct super_block *sb)
ac27a0ec
DK		 2797{
2798 if (!(sb->s_flags & MS_RDONLY)) {
2799 lock_super(sb);
2800 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 2801 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2802 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 2803 unlock_super(sb);
dab291af 2804 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 2805 }
2806}
2807
617ba13b 2808static int ext4_remount (struct super_block * sb, int * flags, char * data)
ac27a0ec 2809{
617ba13b
MC		 2810 struct ext4_super_block * es;
2811 struct ext4_sb_info *sbi = EXT4_SB(sb);
2812 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 2813 unsigned long old_sb_flags;
617ba13b 2814 struct ext4_mount_options old_opts;
ac27a0ec
DK		 2815 int err;
2816#ifdef CONFIG_QUOTA
2817 int i;
2818#endif
2819
2820 /* Store the original options */
2821 old_sb_flags = sb->s_flags;
2822 old_opts.s_mount_opt = sbi->s_mount_opt;
2823 old_opts.s_resuid = sbi->s_resuid;
2824 old_opts.s_resgid = sbi->s_resgid;
2825 old_opts.s_commit_interval = sbi->s_commit_interval;
2826#ifdef CONFIG_QUOTA
2827 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2828 for (i = 0; i < MAXQUOTAS; i++)
2829 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2830#endif
2831
2832 /*
2833 * Allow the "check" option to be passed as a remount option.
2834 */
2835 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2836 err = -EINVAL;
2837 goto restore_opts;
2838 }
2839
617ba13b 2840 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
46e665e9 2841 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 2842
2843 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2844 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 2845
2846 es = sbi->s_es;
2847
617ba13b 2848 ext4_init_journal_params(sb, sbi->s_journal);
ac27a0ec
DK		 2849
2850 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 2851 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 2852 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 2853 err = -EROFS;
2854 goto restore_opts;
2855 }
2856
2857 if (*flags & MS_RDONLY) {
2858 /*
2859 * First of all, the unconditional stuff we have to do
2860 * to disable replay of the journal when we next remount
2861 */
2862 sb->s_flags |= MS_RDONLY;
2863
2864 /*
2865 * OK, test if we are remounting a valid rw partition
2866 * readonly, and if so set the rdonly flag and then
2867 * mark the partition as valid again.
2868 */
617ba13b
MC		 2869 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2870 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 2871 es->s_state = cpu_to_le16(sbi->s_mount_state);
2872
32c37730
JK		 2873 /*
2874 * We have to unlock super so that we can wait for
2875 * transactions.
2876 */
2877 unlock_super(sb);
617ba13b 2878 ext4_mark_recovery_complete(sb, es);
32c37730 2879 lock_super(sb);
ac27a0ec
DK		 2880 } else {
2881 __le32 ret;
617ba13b
MC		 2882 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2883 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2884 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec
DK		 2885 "remount RDWR because of unsupported "
2886 "optional features (%x).\n",
2887 sb->s_id, le32_to_cpu(ret));
2888 err = -EROFS;
2889 goto restore_opts;
2890 }
ead6596b
ES		 2891
2892 /*
2893 * If we have an unprocessed orphan list hanging
2894 * around from a previously readonly bdev mount,
2895 * require a full umount/remount for now.
2896 */
2897 if (es->s_last_orphan) {
2898 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2899 "remount RDWR because of unprocessed "
2900 "orphan inode list. Please "
2901 "umount/remount instead.\n",
2902 sb->s_id);
2903 err = -EINVAL;
2904 goto restore_opts;
2905 }
2906
ac27a0ec
DK		 2907 /*
2908 * Mounting a RDONLY partition read-write, so reread
2909 * and store the current valid flag. (It may have
2910 * been changed by e2fsck since we originally mounted
2911 * the partition.)
2912 */
617ba13b 2913 ext4_clear_journal_err(sb, es);
ac27a0ec 2914 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 2915 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 2916 goto restore_opts;
617ba13b 2917 if (!ext4_setup_super (sb, es, 0))
ac27a0ec
DK		 2918 sb->s_flags &= ~MS_RDONLY;
2919 }
2920 }
2921#ifdef CONFIG_QUOTA
2922 /* Release old quota file names */
2923 for (i = 0; i < MAXQUOTAS; i++)
2924 if (old_opts.s_qf_names[i] &&
2925 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2926 kfree(old_opts.s_qf_names[i]);
2927#endif
2928 return 0;
2929restore_opts:
2930 sb->s_flags = old_sb_flags;
2931 sbi->s_mount_opt = old_opts.s_mount_opt;
2932 sbi->s_resuid = old_opts.s_resuid;
2933 sbi->s_resgid = old_opts.s_resgid;
2934 sbi->s_commit_interval = old_opts.s_commit_interval;
2935#ifdef CONFIG_QUOTA
2936 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2937 for (i = 0; i < MAXQUOTAS; i++) {
2938 if (sbi->s_qf_names[i] &&
2939 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2940 kfree(sbi->s_qf_names[i]);
2941 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2942 }
2943#endif
2944 return err;
2945}
2946
617ba13b 2947static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
ac27a0ec
DK		 2948{
2949 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 2950 struct ext4_sb_info *sbi = EXT4_SB(sb);
2951 struct ext4_super_block *es = sbi->s_es;
960cc398 2952 u64 fsid;
ac27a0ec 2953
5e70030d
BP		 2954 if (test_opt(sb, MINIX_DF)) {
2955 sbi->s_overhead_last = 0;
6bc9feff 2956 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 2957 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 2958 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 2959 smp_rmb();
2960
2961 /*
5e70030d
BP		 2962 * Compute the overhead (FS structures). This is constant
2963 * for a given filesystem unless the number of block groups
2964 * changes so we cache the previous value until it does.
ac27a0ec
DK		 2965 */
2966
2967 /*
2968 * All of the blocks before first_data_block are
2969 * overhead
2970 */
2971 overhead = le32_to_cpu(es->s_first_data_block);
2972
2973 /*
2974 * Add the overhead attributed to the superblock and
2975 * block group descriptors. If the sparse superblocks
2976 * feature is turned on, then not all groups have this.
2977 */
2978 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 2979 overhead += ext4_bg_has_super(sb, i) +
2980 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 2981 cond_resched();
2982 }
2983
2984 /*
2985 * Every block group has an inode bitmap, a block
2986 * bitmap, and an inode table.
2987 */
5e70030d
BP		 2988 overhead += ngroups * (2 + sbi->s_itb_per_group);
2989 sbi->s_overhead_last = overhead;
2990 smp_wmb();
6bc9feff 2991 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 2992 }
2993
617ba13b 2994 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 2995 buf->f_bsize = sb->s_blocksize;
5e70030d 2996 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
52d9f3b4 2997 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
308ba3ec 2998 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 2999 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3000 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3001 buf->f_bavail = 0;
3002 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3003 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3004 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3005 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3006 fsid = le64_to_cpup((void *)es->s_uuid) ^
3007 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3008 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3009 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 3010 return 0;
3011}
3012
3013/* Helper function for writing quotas on sync - we need to start transaction before quota file
3014 * is locked for write. Otherwise the are possible deadlocks:
3015 * Process 1 Process 2
617ba13b 3016 * ext4_create() quota_sync()
dab291af 3017 * jbd2_journal_start() write_dquot()
ac27a0ec 3018 * DQUOT_INIT() down(dqio_mutex)
dab291af 3019 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3020 *
3021 */
3022
3023#ifdef CONFIG_QUOTA
3024
3025static inline struct inode *dquot_to_inode(struct dquot *dquot)
3026{
3027 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3028}
3029
617ba13b 3030static int ext4_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 3031{
3032 handle_t *handle;
3033 int ret, err;
3034
3035 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 3036 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 3037 if (IS_ERR(handle))
3038 return PTR_ERR(handle);
3039 ret = dquot_initialize(inode, type);
617ba13b 3040 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3041 if (!ret)
3042 ret = err;
3043 return ret;
3044}
3045
617ba13b 3046static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 3047{
3048 handle_t *handle;
3049 int ret, err;
3050
3051 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 3052 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2887df13
JK		 3053 if (IS_ERR(handle)) {
3054 /*
3055 * We call dquot_drop() anyway to at least release references
3056 * to quota structures so that umount does not hang.
3057 */
3058 dquot_drop(inode);
ac27a0ec 3059 return PTR_ERR(handle);
2887df13 3060 }
ac27a0ec 3061 ret = dquot_drop(inode);
617ba13b 3062 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3063 if (!ret)
3064 ret = err;
3065 return ret;
3066}
3067
617ba13b 3068static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3069{
3070 int ret, err;
3071 handle_t *handle;
3072 struct inode *inode;
3073
3074 inode = dquot_to_inode(dquot);
617ba13b
MC		 3075 handle = ext4_journal_start(inode,
3076 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3077 if (IS_ERR(handle))
3078 return PTR_ERR(handle);
3079 ret = dquot_commit(dquot);
617ba13b 3080 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3081 if (!ret)
3082 ret = err;
3083 return ret;
3084}
3085
617ba13b 3086static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3087{
3088 int ret, err;
3089 handle_t *handle;
3090
617ba13b
MC		 3091 handle = ext4_journal_start(dquot_to_inode(dquot),
3092 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3093 if (IS_ERR(handle))
3094 return PTR_ERR(handle);
3095 ret = dquot_acquire(dquot);
617ba13b 3096 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3097 if (!ret)
3098 ret = err;
3099 return ret;
3100}
3101
617ba13b 3102static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3103{
3104 int ret, err;
3105 handle_t *handle;
3106
617ba13b
MC		 3107 handle = ext4_journal_start(dquot_to_inode(dquot),
3108 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3109 if (IS_ERR(handle)) {
3110 /* Release dquot anyway to avoid endless cycle in dqput() */
3111 dquot_release(dquot);
ac27a0ec 3112 return PTR_ERR(handle);
9c3013e9 3113 }
ac27a0ec 3114 ret = dquot_release(dquot);
617ba13b 3115 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3116 if (!ret)
3117 ret = err;
3118 return ret;
3119}
3120
617ba13b 3121static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec
DK		 3122{
3123 /* Are we journalling quotas? */
617ba13b
MC		 3124 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3125 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3126 dquot_mark_dquot_dirty(dquot);
617ba13b 3127 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3128 } else {
3129 return dquot_mark_dquot_dirty(dquot);
3130 }
3131}
3132
617ba13b 3133static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3134{
3135 int ret, err;
3136 handle_t *handle;
3137
3138 /* Data block + inode block */
617ba13b 3139 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3140 if (IS_ERR(handle))
3141 return PTR_ERR(handle);
3142 ret = dquot_commit_info(sb, type);
617ba13b 3143 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3144 if (!ret)
3145 ret = err;
3146 return ret;
3147}
3148
3149/*
3150 * Turn on quotas during mount time - we need to find
3151 * the quota file and such...
3152 */
617ba13b 3153static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3154{
617ba13b
MC		 3155 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3156 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3157}
3158
3159/*
3160 * Standard function to be called on quota_on
3161 */
617ba13b 3162static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6f28e087 3163 char *path, int remount)
ac27a0ec
DK		 3164{
3165 int err;
3166 struct nameidata nd;
3167
3168 if (!test_opt(sb, QUOTA))
3169 return -EINVAL;
3170 /* Not journalling quota? */
6f28e087
JK		 3171 if ((!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
3172 !EXT4_SB(sb)->s_qf_names[GRPQUOTA]) || remount)
3173 return vfs_quota_on(sb, type, format_id, path, remount);
ac27a0ec
DK		 3174 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3175 if (err)
3176 return err;
3177 /* Quotafile not on the same filesystem? */
4ac91378 3178 if (nd.path.mnt->mnt_sb != sb) {
1d957f9b 3179 path_put(&nd.path);
ac27a0ec
DK		 3180 return -EXDEV;
3181 }
3182 /* Quotafile not of fs root? */
4ac91378 3183 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
ac27a0ec 3184 printk(KERN_WARNING
617ba13b 3185 "EXT4-fs: Quota file not on filesystem root. "
ac27a0ec 3186 "Journalled quota will not work.\n");
1d957f9b 3187 path_put(&nd.path);
6f28e087 3188 return vfs_quota_on(sb, type, format_id, path, remount);
ac27a0ec
DK		 3189}
3190
3191/* Read data from quotafile - avoid pagecache and such because we cannot afford
3192 * acquiring the locks... As quota files are never truncated and quota code
3193 * itself serializes the operations (and noone else should touch the files)
3194 * we don't have to be afraid of races */
617ba13b 3195static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3196 size_t len, loff_t off)
3197{
3198 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3199 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3200 int err = 0;
3201 int offset = off & (sb->s_blocksize - 1);
3202 int tocopy;
3203 size_t toread;
3204 struct buffer_head *bh;
3205 loff_t i_size = i_size_read(inode);
3206
3207 if (off > i_size)
3208 return 0;
3209 if (off+len > i_size)
3210 len = i_size-off;
3211 toread = len;
3212 while (toread > 0) {
3213 tocopy = sb->s_blocksize - offset < toread ?
3214 sb->s_blocksize - offset : toread;
617ba13b 3215 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3216 if (err)
3217 return err;
3218 if (!bh) /* A hole? */
3219 memset(data, 0, tocopy);
3220 else
3221 memcpy(data, bh->b_data+offset, tocopy);
3222 brelse(bh);
3223 offset = 0;
3224 toread -= tocopy;
3225 data += tocopy;
3226 blk++;
3227 }
3228 return len;
3229}
3230
3231/* Write to quotafile (we know the transaction is already started and has
3232 * enough credits) */
617ba13b 3233static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3234 const char *data, size_t len, loff_t off)
3235{
3236 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3237 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3238 int err = 0;
3239 int offset = off & (sb->s_blocksize - 1);
3240 int tocopy;
617ba13b 3241 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3242 size_t towrite = len;
3243 struct buffer_head *bh;
3244 handle_t *handle = journal_current_handle();
3245
9c3013e9
JK		 3246 if (!handle) {
3247 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3248 " cancelled because transaction is not started.\n",
3249 (unsigned long long)off, (unsigned long long)len);
3250 return -EIO;
3251 }
ac27a0ec
DK		 3252 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3253 while (towrite > 0) {
3254 tocopy = sb->s_blocksize - offset < towrite ?
3255 sb->s_blocksize - offset : towrite;
617ba13b 3256 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3257 if (!bh)
3258 goto out;
3259 if (journal_quota) {
617ba13b 3260 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3261 if (err) {
3262 brelse(bh);
3263 goto out;
3264 }
3265 }
3266 lock_buffer(bh);
3267 memcpy(bh->b_data+offset, data, tocopy);
3268 flush_dcache_page(bh->b_page);
3269 unlock_buffer(bh);
3270 if (journal_quota)
617ba13b 3271 err = ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK		 3272 else {
3273 /* Always do at least ordered writes for quotas */
617ba13b 3274 err = ext4_journal_dirty_data(handle, bh);
ac27a0ec
DK		 3275 mark_buffer_dirty(bh);
3276 }
3277 brelse(bh);
3278 if (err)
3279 goto out;
3280 offset = 0;
3281 towrite -= tocopy;
3282 data += tocopy;
3283 blk++;
3284 }
3285out:
3286 if (len == towrite)
3287 return err;
3288 if (inode->i_size < off+len-towrite) {
3289 i_size_write(inode, off+len-towrite);
617ba13b 3290 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3291 }
ac27a0ec 3292 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3293 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3294 mutex_unlock(&inode->i_mutex);
3295 return len - towrite;
3296}
3297
3298#endif
3299
617ba13b 3300static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3301 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3302{
617ba13b 3303 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3304}
3305
617ba13b 3306static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3307 .owner = THIS_MODULE,
617ba13b
MC		 3308 .name = "ext4dev",
3309 .get_sb = ext4_get_sb,
ac27a0ec
DK		 3310 .kill_sb = kill_block_super,
3311 .fs_flags = FS_REQUIRES_DEV,
3312};
3313
617ba13b 3314static int __init init_ext4_fs(void)
ac27a0ec 3315{
c9de560d
AT		 3316 int err;
3317
3318 err = init_ext4_mballoc();
ac27a0ec
DK		 3319 if (err)
3320 return err;
c9de560d
AT		 3321
3322 err = init_ext4_xattr();
3323 if (err)
3324 goto out2;
ac27a0ec
DK		 3325 err = init_inodecache();
3326 if (err)
3327 goto out1;
63f57933 3328 err = register_filesystem(&ext4dev_fs_type);
ac27a0ec
DK		 3329 if (err)
3330 goto out;
3331 return 0;
3332out:
3333 destroy_inodecache();
3334out1:
617ba13b 3335 exit_ext4_xattr();
c9de560d
AT		 3336out2:
3337 exit_ext4_mballoc();
ac27a0ec
DK		 3338 return err;
3339}
3340
617ba13b 3341static void __exit exit_ext4_fs(void)
ac27a0ec 3342{
617ba13b 3343 unregister_filesystem(&ext4dev_fs_type);
ac27a0ec 3344 destroy_inodecache();
617ba13b 3345 exit_ext4_xattr();
c9de560d 3346 exit_ext4_mballoc();
ac27a0ec
DK		 3347}
3348
3349MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
617ba13b 3350MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
ac27a0ec 3351MODULE_LICENSE("GPL");
617ba13b
MC		 3352module_init(init_ext4_fs)
3353module_exit(exit_ext4_fs)
Ubuntu Jammy Linux kernel mirror