projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: Replace hackish ext4_mb_poll_new_transaction with commit callback
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
dab291af 23#include <linux/jbd2.h>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
27#include <linux/parser.h>
28#include <linux/smp_lock.h>
29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
ede86cc4 38#include <linux/marker.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
ac27a0ec
DK		 41#include <asm/uaccess.h>
42
3dcf5451
CH		 43#include "ext4.h"
44#include "ext4_jbd2.h"
ac27a0ec
DK		 45#include "xattr.h"
46#include "acl.h"
47#include "namei.h"
717d50e4 48#include "group.h"
ac27a0ec 49
9f6200bb
TT		 50struct proc_dir_entry *ext4_proc_root;
51
617ba13b 52static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 53 unsigned long journal_devnum);
617ba13b 54static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 55 unsigned int);
2b2d6d01
TT		 56static void ext4_commit_super(struct super_block *sb,
57 struct ext4_super_block *es, int sync);
58static void ext4_mark_recovery_complete(struct super_block *sb,
59 struct ext4_super_block *es);
60static void ext4_clear_journal_err(struct super_block *sb,
61 struct ext4_super_block *es);
617ba13b 62static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 63static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 64 char nbuf[16]);
2b2d6d01
TT		 65static int ext4_remount(struct super_block *sb, int *flags, char *data);
66static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
617ba13b 67static void ext4_unlockfs(struct super_block *sb);
2b2d6d01 68static void ext4_write_super(struct super_block *sb);
617ba13b 69static void ext4_write_super_lockfs(struct super_block *sb);
ac27a0ec 70
bd81d8ee 71
8fadc143
AR		 72ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
73 struct ext4_group_desc *bg)
bd81d8ee 74{
3a14589c 75 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 76 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
3a14589c 77 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 78}
79
8fadc143
AR		 80ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
81 struct ext4_group_desc *bg)
bd81d8ee 82{
5272f837 83 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 84 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 85 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 86}
87
8fadc143
AR		 88ext4_fsblk_t ext4_inode_table(struct super_block *sb,
89 struct ext4_group_desc *bg)
bd81d8ee 90{
5272f837 91 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 92 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 93 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 94}
95
8fadc143
AR		 96void ext4_block_bitmap_set(struct super_block *sb,
97 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 98{
3a14589c 99 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 100 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
101 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 102}
103
8fadc143
AR		 104void ext4_inode_bitmap_set(struct super_block *sb,
105 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 106{
5272f837 107 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 108 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
109 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 110}
111
8fadc143
AR		 112void ext4_inode_table_set(struct super_block *sb,
113 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 114{
5272f837 115 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 116 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
117 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 118}
119
ac27a0ec 120/*
dab291af 121 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 122 *
123 * The only special thing we need to do here is to make sure that all
124 * journal_end calls result in the superblock being marked dirty, so
125 * that sync() will call the filesystem's write_super callback if
126 * appropriate.
127 */
617ba13b 128handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 129{
130 journal_t *journal;
131
132 if (sb->s_flags & MS_RDONLY)
133 return ERR_PTR(-EROFS);
134
135 /* Special case here: if the journal has aborted behind our
136 * backs (eg. EIO in the commit thread), then we still need to
137 * take the FS itself readonly cleanly. */
617ba13b 138 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 139 if (is_journal_aborted(journal)) {
46e665e9 140 ext4_abort(sb, __func__,
ac27a0ec
DK		 141 "Detected aborted journal");
142 return ERR_PTR(-EROFS);
143 }
144
dab291af 145 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 146}
147
148/*
149 * The only special thing we need to do here is to make sure that all
dab291af 150 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 151 * that sync() will call the filesystem's write_super callback if
152 * appropriate.
153 */
617ba13b 154int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 155{
156 struct super_block *sb;
157 int err;
158 int rc;
159
160 sb = handle->h_transaction->t_journal->j_private;
161 err = handle->h_err;
dab291af 162 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 163
164 if (!err)
165 err = rc;
166 if (err)
617ba13b 167 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 168 return err;
169}
170
617ba13b 171void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 172 struct buffer_head *bh, handle_t *handle, int err)
173{
174 char nbuf[16];
617ba13b 175 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec
DK		 176
177 if (bh)
178 BUFFER_TRACE(bh, "abort");
179
180 if (!handle->h_err)
181 handle->h_err = err;
182
183 if (is_handle_aborted(handle))
184 return;
185
186 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
187 caller, errstr, err_fn);
188
dab291af 189 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 190}
191
192/* Deal with the reporting of failure conditions on a filesystem such as
193 * inconsistencies detected or read IO failures.
194 *
195 * On ext2, we can store the error state of the filesystem in the
617ba13b 196 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 197 * write ordering constraints on the superblock which prevent us from
198 * writing it out straight away; and given that the journal is about to
199 * be aborted, we can't rely on the current, or future, transactions to
200 * write out the superblock safely.
201 *
dab291af 202 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 203 * the journal instead. On recovery, the journal will compain about
204 * that error until we've noted it down and cleared it.
205 */
206
617ba13b 207static void ext4_handle_error(struct super_block *sb)
ac27a0ec 208{
617ba13b 209 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 210
617ba13b
MC		 211 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
212 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 213
214 if (sb->s_flags & MS_RDONLY)
215 return;
216
2b2d6d01 217 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 218 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 219
617ba13b 220 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ac27a0ec 221 if (journal)
dab291af 222 jbd2_journal_abort(journal, -EIO);
ac27a0ec 223 }
2b2d6d01
TT		 224 if (test_opt(sb, ERRORS_RO)) {
225 printk(KERN_CRIT "Remounting filesystem read-only\n");
ac27a0ec
DK		 226 sb->s_flags |= MS_RDONLY;
227 }
617ba13b 228 ext4_commit_super(sb, es, 1);
ac27a0ec 229 if (test_opt(sb, ERRORS_PANIC))
617ba13b 230 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 231 sb->s_id);
232}
233
2b2d6d01
TT		 234void ext4_error(struct super_block *sb, const char *function,
235 const char *fmt, ...)
ac27a0ec
DK		 236{
237 va_list args;
238
239 va_start(args, fmt);
2b2d6d01 240 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 241 vprintk(fmt, args);
242 printk("\n");
243 va_end(args);
244
617ba13b 245 ext4_handle_error(sb);
ac27a0ec
DK		 246}
247
2b2d6d01 248static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 249 char nbuf[16])
250{
251 char *errstr = NULL;
252
253 switch (errno) {
254 case -EIO:
255 errstr = "IO failure";
256 break;
257 case -ENOMEM:
258 errstr = "Out of memory";
259 break;
260 case -EROFS:
dab291af 261 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
ac27a0ec
DK		 262 errstr = "Journal has aborted";
263 else
264 errstr = "Readonly filesystem";
265 break;
266 default:
267 /* If the caller passed in an extra buffer for unknown
268 * errors, textualise them now. Else we just return
269 * NULL. */
270 if (nbuf) {
271 /* Check for truncated error codes... */
272 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
273 errstr = nbuf;
274 }
275 break;
276 }
277
278 return errstr;
279}
280
617ba13b 281/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 282 * automatically and invokes the appropriate error response. */
283
2b2d6d01 284void __ext4_std_error(struct super_block *sb, const char *function, int errno)
ac27a0ec
DK		 285{
286 char nbuf[16];
287 const char *errstr;
288
289 /* Special case: if the error is EROFS, and we're not already
290 * inside a transaction, then there's really no point in logging
291 * an error. */
292 if (errno == -EROFS && journal_current_handle() == NULL &&
293 (sb->s_flags & MS_RDONLY))
294 return;
295
617ba13b 296 errstr = ext4_decode_error(sb, errno, nbuf);
2b2d6d01
TT		 297 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
298 sb->s_id, function, errstr);
ac27a0ec 299
617ba13b 300 ext4_handle_error(sb);
ac27a0ec
DK		 301}
302
303/*
617ba13b 304 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 305 * abort function may be used to deal with unrecoverable failures such
306 * as journal IO errors or ENOMEM at a critical moment in log management.
307 *
308 * We unconditionally force the filesystem into an ABORT|READONLY state,
309 * unless the error response on the fs has been set to panic in which
310 * case we take the easy way out and panic immediately.
311 */
312
2b2d6d01
TT		 313void ext4_abort(struct super_block *sb, const char *function,
314 const char *fmt, ...)
ac27a0ec
DK		 315{
316 va_list args;
317
2b2d6d01 318 printk(KERN_CRIT "ext4_abort called.\n");
ac27a0ec
DK		 319
320 va_start(args, fmt);
2b2d6d01 321 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 322 vprintk(fmt, args);
323 printk("\n");
324 va_end(args);
325
326 if (test_opt(sb, ERRORS_PANIC))
617ba13b 327 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 328
329 if (sb->s_flags & MS_RDONLY)
330 return;
331
332 printk(KERN_CRIT "Remounting filesystem read-only\n");
617ba13b 333 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 334 sb->s_flags |= MS_RDONLY;
617ba13b 335 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
dab291af 336 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 337}
338
2b2d6d01
TT		 339void ext4_warning(struct super_block *sb, const char *function,
340 const char *fmt, ...)
ac27a0ec
DK		 341{
342 va_list args;
343
344 va_start(args, fmt);
617ba13b 345 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 346 sb->s_id, function);
347 vprintk(fmt, args);
348 printk("\n");
349 va_end(args);
350}
351
617ba13b 352void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 353{
617ba13b 354 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 355
617ba13b 356 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 357 return;
358
46e665e9 359 ext4_warning(sb, __func__,
ac27a0ec
DK		 360 "updating to rev %d because of new feature flag, "
361 "running e2fsck is recommended",
617ba13b 362 EXT4_DYNAMIC_REV);
ac27a0ec 363
617ba13b
MC		 364 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
365 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
366 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 367 /* leave es->s_feature_*compat flags alone */
368 /* es->s_uuid will be set by e2fsck if empty */
369
370 /*
371 * The rest of the superblock fields should be zero, and if not it
372 * means they are likely already in use, so leave them alone. We
373 * can leave it up to e2fsck to clean up any inconsistencies there.
374 */
375}
376
99e6f829
AK		 377int ext4_update_compat_feature(handle_t *handle,
378 struct super_block *sb, __u32 compat)
379{
380 int err = 0;
381 if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
382 err = ext4_journal_get_write_access(handle,
383 EXT4_SB(sb)->s_sbh);
384 if (err)
385 return err;
386 EXT4_SET_COMPAT_FEATURE(sb, compat);
387 sb->s_dirt = 1;
388 handle->h_sync = 1;
389 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
390 "call ext4_journal_dirty_met adata");
391 err = ext4_journal_dirty_metadata(handle,
392 EXT4_SB(sb)->s_sbh);
393 }
394 return err;
395}
396
397int ext4_update_rocompat_feature(handle_t *handle,
398 struct super_block *sb, __u32 rocompat)
399{
400 int err = 0;
401 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
402 err = ext4_journal_get_write_access(handle,
403 EXT4_SB(sb)->s_sbh);
404 if (err)
405 return err;
406 EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
407 sb->s_dirt = 1;
408 handle->h_sync = 1;
409 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
410 "call ext4_journal_dirty_met adata");
411 err = ext4_journal_dirty_metadata(handle,
412 EXT4_SB(sb)->s_sbh);
413 }
414 return err;
415}
416
417int ext4_update_incompat_feature(handle_t *handle,
418 struct super_block *sb, __u32 incompat)
419{
420 int err = 0;
421 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
422 err = ext4_journal_get_write_access(handle,
423 EXT4_SB(sb)->s_sbh);
424 if (err)
425 return err;
426 EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
427 sb->s_dirt = 1;
428 handle->h_sync = 1;
429 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
430 "call ext4_journal_dirty_met adata");
431 err = ext4_journal_dirty_metadata(handle,
432 EXT4_SB(sb)->s_sbh);
433 }
434 return err;
435}
436
ac27a0ec
DK		 437/*
438 * Open the external journal device
439 */
617ba13b 440static struct block_device *ext4_blkdev_get(dev_t dev)
ac27a0ec
DK		 441{
442 struct block_device *bdev;
443 char b[BDEVNAME_SIZE];
444
445 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
446 if (IS_ERR(bdev))
447 goto fail;
448 return bdev;
449
450fail:
617ba13b 451 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
ac27a0ec
DK		 452 __bdevname(dev, b), PTR_ERR(bdev));
453 return NULL;
454}
455
456/*
457 * Release the journal device
458 */
617ba13b 459static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 460{
461 bd_release(bdev);
462 return blkdev_put(bdev);
463}
464
617ba13b 465static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 466{
467 struct block_device *bdev;
468 int ret = -ENODEV;
469
470 bdev = sbi->journal_bdev;
471 if (bdev) {
617ba13b 472 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 473 sbi->journal_bdev = NULL;
474 }
475 return ret;
476}
477
478static inline struct inode *orphan_list_entry(struct list_head *l)
479{
617ba13b 480 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 481}
482
617ba13b 483static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 484{
485 struct list_head *l;
486
487 printk(KERN_ERR "sb orphan head is %d\n",
488 le32_to_cpu(sbi->s_es->s_last_orphan));
489
490 printk(KERN_ERR "sb_info orphan list:\n");
491 list_for_each(l, &sbi->s_orphan) {
492 struct inode *inode = orphan_list_entry(l);
493 printk(KERN_ERR " "
494 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
495 inode->i_sb->s_id, inode->i_ino, inode,
496 inode->i_mode, inode->i_nlink,
497 NEXT_ORPHAN(inode));
498 }
499}
500
2b2d6d01 501static void ext4_put_super(struct super_block *sb)
ac27a0ec 502{
617ba13b
MC		 503 struct ext4_sb_info *sbi = EXT4_SB(sb);
504 struct ext4_super_block *es = sbi->s_es;
ac27a0ec
DK		 505 int i;
506
c9de560d 507 ext4_mb_release(sb);
a86c6181 508 ext4_ext_release(sb);
617ba13b 509 ext4_xattr_put_super(sb);
7ffe1ea8
HK		 510 if (jbd2_journal_destroy(sbi->s_journal) < 0)
511 ext4_abort(sb, __func__, "Couldn't clean up the journal");
47b4a50b 512 sbi->s_journal = NULL;
ac27a0ec 513 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 514 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 515 es->s_state = cpu_to_le16(sbi->s_mount_state);
617ba13b 516 ext4_commit_super(sb, es, 1);
ac27a0ec 517 }
240799cd
TT		 518 if (sbi->s_proc) {
519 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
9f6200bb 520 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 521 }
ac27a0ec
DK		 522
523 for (i = 0; i < sbi->s_gdb_count; i++)
524 brelse(sbi->s_group_desc[i]);
525 kfree(sbi->s_group_desc);
772cb7c8 526 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 527 percpu_counter_destroy(&sbi->s_freeblocks_counter);
528 percpu_counter_destroy(&sbi->s_freeinodes_counter);
529 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 530 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 531 brelse(sbi->s_sbh);
532#ifdef CONFIG_QUOTA
533 for (i = 0; i < MAXQUOTAS; i++)
534 kfree(sbi->s_qf_names[i]);
535#endif
536
537 /* Debugging code just in case the in-memory inode orphan list
538 * isn't empty. The on-disk one can be non-empty if we've
539 * detected an error and taken the fs readonly, but the
540 * in-memory list had better be clean by this point. */
541 if (!list_empty(&sbi->s_orphan))
542 dump_orphan_list(sb, sbi);
543 J_ASSERT(list_empty(&sbi->s_orphan));
544
f98393a6 545 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 546 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
547 /*
548 * Invalidate the journal device's buffers. We don't want them
549 * floating about in memory - the physical journal device may
550 * hotswapped, and it breaks the `ro-after' testing code.
551 */
552 sync_blockdev(sbi->journal_bdev);
f98393a6 553 invalidate_bdev(sbi->journal_bdev);
617ba13b 554 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 555 }
556 sb->s_fs_info = NULL;
557 kfree(sbi);
558 return;
559}
560
e18b890b 561static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 562
563/*
564 * Called inside transaction, so use GFP_NOFS
565 */
617ba13b 566static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 567{
617ba13b 568 struct ext4_inode_info *ei;
ac27a0ec 569
e6b4f8da 570 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 571 if (!ei)
572 return NULL;
03010a33 573#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b
MC		 574 ei->i_acl = EXT4_ACL_NOT_CACHED;
575 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec 576#endif
ac27a0ec 577 ei->vfs_inode.i_version = 1;
91246c00 578 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 579 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 580 INIT_LIST_HEAD(&ei->i_prealloc_list);
581 spin_lock_init(&ei->i_prealloc_lock);
678aaf48 582 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 583 ei->i_reserved_data_blocks = 0;
584 ei->i_reserved_meta_blocks = 0;
585 ei->i_allocated_meta_blocks = 0;
586 ei->i_delalloc_reserved_flag = 0;
587 spin_lock_init(&(ei->i_block_reservation_lock));
ac27a0ec
DK		 588 return &ei->vfs_inode;
589}
590
617ba13b 591static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 592{
9f7dd93d
VA		 593 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
594 printk("EXT4 Inode %p: orphan list check failed!\n",
595 EXT4_I(inode));
596 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
597 EXT4_I(inode), sizeof(struct ext4_inode_info),
598 true);
599 dump_stack();
600 }
617ba13b 601 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 602}
603
51cc5068 604static void init_once(void *foo)
ac27a0ec 605{
617ba13b 606 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 607
a35afb83 608 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 609#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 610 init_rwsem(&ei->xattr_sem);
ac27a0ec 611#endif
0e855ac8 612 init_rwsem(&ei->i_data_sem);
a35afb83 613 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 614}
615
616static int init_inodecache(void)
617{
617ba13b
MC		 618 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
619 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 620 0, (SLAB_RECLAIM_ACCOUNT|
621 SLAB_MEM_SPREAD),
20c2df83 622 init_once);
617ba13b 623 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 624 return -ENOMEM;
625 return 0;
626}
627
628static void destroy_inodecache(void)
629{
617ba13b 630 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 631}
632
617ba13b 633static void ext4_clear_inode(struct inode *inode)
ac27a0ec 634{
03010a33 635#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b
MC		 636 if (EXT4_I(inode)->i_acl &&
637 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
638 posix_acl_release(EXT4_I(inode)->i_acl);
639 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
640 }
641 if (EXT4_I(inode)->i_default_acl &&
642 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
643 posix_acl_release(EXT4_I(inode)->i_default_acl);
644 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 645 }
646#endif
c2ea3fde 647 ext4_discard_preallocations(inode);
678aaf48
JK		 648 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
649 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 650}
651
2b2d6d01
TT		 652static inline void ext4_show_quota_options(struct seq_file *seq,
653 struct super_block *sb)
ac27a0ec
DK		 654{
655#if defined(CONFIG_QUOTA)
617ba13b 656 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 657
658 if (sbi->s_jquota_fmt)
659 seq_printf(seq, ",jqfmt=%s",
af5bc92d 660 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
ac27a0ec
DK		 661
662 if (sbi->s_qf_names[USRQUOTA])
663 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
664
665 if (sbi->s_qf_names[GRPQUOTA])
666 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
667
617ba13b 668 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 669 seq_puts(seq, ",usrquota");
670
617ba13b 671 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 672 seq_puts(seq, ",grpquota");
673#endif
674}
675
d9c9bef1
MS		 676/*
677 * Show an option if
678 * - it's set to a non-default value OR
679 * - if the per-sb default is different from the global default
680 */
617ba13b 681static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 682{
aa22df2c
AK		 683 int def_errors;
684 unsigned long def_mount_opts;
ac27a0ec 685 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 686 struct ext4_sb_info *sbi = EXT4_SB(sb);
687 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 688
689 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 690 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 691
692 if (sbi->s_sb_block != 1)
693 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
694 if (test_opt(sb, MINIX_DF))
695 seq_puts(seq, ",minixdf");
aa22df2c 696 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 697 seq_puts(seq, ",grpid");
698 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
699 seq_puts(seq, ",nogrpid");
700 if (sbi->s_resuid != EXT4_DEF_RESUID ||
701 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
702 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
703 }
704 if (sbi->s_resgid != EXT4_DEF_RESGID ||
705 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
706 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
707 }
bb4f397a 708 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 709 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 710 def_errors == EXT4_ERRORS_CONTINUE) {
711 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 712 }
713 }
aa22df2c 714 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 715 seq_puts(seq, ",errors=continue");
aa22df2c 716 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 717 seq_puts(seq, ",errors=panic");
aa22df2c 718 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 719 seq_puts(seq, ",nouid32");
aa22df2c 720 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 721 seq_puts(seq, ",debug");
722 if (test_opt(sb, OLDALLOC))
723 seq_puts(seq, ",oldalloc");
03010a33 724#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 725 if (test_opt(sb, XATTR_USER) &&
726 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 727 seq_puts(seq, ",user_xattr");
728 if (!test_opt(sb, XATTR_USER) &&
729 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
730 seq_puts(seq, ",nouser_xattr");
731 }
732#endif
03010a33 733#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 734 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 735 seq_puts(seq, ",acl");
736 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
737 seq_puts(seq, ",noacl");
738#endif
739 if (!test_opt(sb, RESERVATION))
740 seq_puts(seq, ",noreservation");
741 if (sbi->s_commit_interval) {
742 seq_printf(seq, ",commit=%u",
743 (unsigned) (sbi->s_commit_interval / HZ));
744 }
571640ca
ES		 745 /*
746 * We're changing the default of barrier mount option, so
747 * let's always display its mount state so it's clear what its
748 * status is.
749 */
750 seq_puts(seq, ",barrier=");
751 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 752 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
753 seq_puts(seq, ",journal_async_commit");
d9c9bef1
MS		 754 if (test_opt(sb, NOBH))
755 seq_puts(seq, ",nobh");
756 if (!test_opt(sb, EXTENTS))
757 seq_puts(seq, ",noextents");
25ec56b5
JNC		 758 if (test_opt(sb, I_VERSION))
759 seq_puts(seq, ",i_version");
dd919b98
AK		 760 if (!test_opt(sb, DELALLOC))
761 seq_puts(seq, ",nodelalloc");
762
ac27a0ec 763
cb45bbe4
MS		 764 if (sbi->s_stripe)
765 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 766 /*
767 * journal mode get enabled in different ways
768 * So just print the value even if we didn't specify it
769 */
617ba13b 770 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 771 seq_puts(seq, ",data=journal");
617ba13b 772 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 773 seq_puts(seq, ",data=ordered");
617ba13b 774 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 775 seq_puts(seq, ",data=writeback");
776
240799cd
TT		 777 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
778 seq_printf(seq, ",inode_readahead_blks=%u",
779 sbi->s_inode_readahead_blks);
780
5bf5683a
HK		 781 if (test_opt(sb, DATA_ERR_ABORT))
782 seq_puts(seq, ",data_err=abort");
783
617ba13b 784 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 785 return 0;
786}
787
788
1b961ac0
CH		 789static struct inode *ext4_nfs_get_inode(struct super_block *sb,
790 u64 ino, u32 generation)
ac27a0ec 791{
ac27a0ec 792 struct inode *inode;
ac27a0ec 793
617ba13b 794 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 795 return ERR_PTR(-ESTALE);
617ba13b 796 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 797 return ERR_PTR(-ESTALE);
798
799 /* iget isn't really right if the inode is currently unallocated!!
800 *
617ba13b 801 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 802 * deleted, so we should be safe.
803 *
804 * Currently we don't know the generation for parent directory, so
805 * a generation of 0 means "accept any"
806 */
1d1fe1ee
DH		 807 inode = ext4_iget(sb, ino);
808 if (IS_ERR(inode))
809 return ERR_CAST(inode);
810 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 811 iput(inode);
812 return ERR_PTR(-ESTALE);
813 }
1b961ac0
CH		 814
815 return inode;
816}
817
818static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
819 int fh_len, int fh_type)
820{
821 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
822 ext4_nfs_get_inode);
823}
824
825static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
826 int fh_len, int fh_type)
827{
828 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
829 ext4_nfs_get_inode);
ac27a0ec
DK		 830}
831
832#ifdef CONFIG_QUOTA
af5bc92d 833#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 834#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 835
617ba13b
MC		 836static int ext4_dquot_initialize(struct inode *inode, int type);
837static int ext4_dquot_drop(struct inode *inode);
838static int ext4_write_dquot(struct dquot *dquot);
839static int ext4_acquire_dquot(struct dquot *dquot);
840static int ext4_release_dquot(struct dquot *dquot);
841static int ext4_mark_dquot_dirty(struct dquot *dquot);
842static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 843static int ext4_quota_on(struct super_block *sb, int type, int format_id,
844 char *path, int remount);
617ba13b
MC		 845static int ext4_quota_on_mount(struct super_block *sb, int type);
846static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 847 size_t len, loff_t off);
617ba13b 848static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 849 const char *data, size_t len, loff_t off);
850
617ba13b
MC		 851static struct dquot_operations ext4_quota_operations = {
852 .initialize = ext4_dquot_initialize,
853 .drop = ext4_dquot_drop,
ac27a0ec
DK		 854 .alloc_space = dquot_alloc_space,
855 .alloc_inode = dquot_alloc_inode,
856 .free_space = dquot_free_space,
857 .free_inode = dquot_free_inode,
858 .transfer = dquot_transfer,
617ba13b
MC		 859 .write_dquot = ext4_write_dquot,
860 .acquire_dquot = ext4_acquire_dquot,
861 .release_dquot = ext4_release_dquot,
862 .mark_dirty = ext4_mark_dquot_dirty,
863 .write_info = ext4_write_info
ac27a0ec
DK		 864};
865
617ba13b
MC		 866static struct quotactl_ops ext4_qctl_operations = {
867 .quota_on = ext4_quota_on,
ac27a0ec
DK		 868 .quota_off = vfs_quota_off,
869 .quota_sync = vfs_quota_sync,
870 .get_info = vfs_get_dqinfo,
871 .set_info = vfs_set_dqinfo,
872 .get_dqblk = vfs_get_dqblk,
873 .set_dqblk = vfs_set_dqblk
874};
875#endif
876
ee9b6d61 877static const struct super_operations ext4_sops = {
617ba13b
MC		 878 .alloc_inode = ext4_alloc_inode,
879 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 880 .write_inode = ext4_write_inode,
881 .dirty_inode = ext4_dirty_inode,
882 .delete_inode = ext4_delete_inode,
883 .put_super = ext4_put_super,
884 .write_super = ext4_write_super,
885 .sync_fs = ext4_sync_fs,
886 .write_super_lockfs = ext4_write_super_lockfs,
887 .unlockfs = ext4_unlockfs,
888 .statfs = ext4_statfs,
889 .remount_fs = ext4_remount,
890 .clear_inode = ext4_clear_inode,
891 .show_options = ext4_show_options,
ac27a0ec 892#ifdef CONFIG_QUOTA
617ba13b
MC		 893 .quota_read = ext4_quota_read,
894 .quota_write = ext4_quota_write,
ac27a0ec
DK		 895#endif
896};
897
39655164 898static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 899 .fh_to_dentry = ext4_fh_to_dentry,
900 .fh_to_parent = ext4_fh_to_parent,
617ba13b 901 .get_parent = ext4_get_parent,
ac27a0ec
DK		 902};
903
904enum {
905 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
906 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 907 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec
DK		 908 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
909 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
910 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
818d276c 911 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 912 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
5bf5683a 913 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec
DK		 914 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
915 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
916 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
25ec56b5 917 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
01436ef2 918 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
240799cd 919 Opt_inode_readahead_blks
ac27a0ec
DK		 920};
921
a447c093 922static const match_table_t tokens = {
ac27a0ec
DK		 923 {Opt_bsd_df, "bsddf"},
924 {Opt_minix_df, "minixdf"},
925 {Opt_grpid, "grpid"},
926 {Opt_grpid, "bsdgroups"},
927 {Opt_nogrpid, "nogrpid"},
928 {Opt_nogrpid, "sysvgroups"},
929 {Opt_resgid, "resgid=%u"},
930 {Opt_resuid, "resuid=%u"},
931 {Opt_sb, "sb=%u"},
932 {Opt_err_cont, "errors=continue"},
933 {Opt_err_panic, "errors=panic"},
934 {Opt_err_ro, "errors=remount-ro"},
935 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 936 {Opt_debug, "debug"},
937 {Opt_oldalloc, "oldalloc"},
938 {Opt_orlov, "orlov"},
939 {Opt_user_xattr, "user_xattr"},
940 {Opt_nouser_xattr, "nouser_xattr"},
941 {Opt_acl, "acl"},
942 {Opt_noacl, "noacl"},
943 {Opt_reservation, "reservation"},
944 {Opt_noreservation, "noreservation"},
945 {Opt_noload, "noload"},
946 {Opt_nobh, "nobh"},
947 {Opt_bh, "bh"},
948 {Opt_commit, "commit=%u"},
949 {Opt_journal_update, "journal=update"},
950 {Opt_journal_inum, "journal=%u"},
951 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 952 {Opt_journal_checksum, "journal_checksum"},
953 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 954 {Opt_abort, "abort"},
955 {Opt_data_journal, "data=journal"},
956 {Opt_data_ordered, "data=ordered"},
957 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 958 {Opt_data_err_abort, "data_err=abort"},
959 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 960 {Opt_offusrjquota, "usrjquota="},
961 {Opt_usrjquota, "usrjquota=%s"},
962 {Opt_offgrpjquota, "grpjquota="},
963 {Opt_grpjquota, "grpjquota=%s"},
964 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
965 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
966 {Opt_grpquota, "grpquota"},
967 {Opt_noquota, "noquota"},
968 {Opt_quota, "quota"},
969 {Opt_usrquota, "usrquota"},
970 {Opt_barrier, "barrier=%u"},
a86c6181 971 {Opt_extents, "extents"},
1e2462f9 972 {Opt_noextents, "noextents"},
25ec56b5 973 {Opt_i_version, "i_version"},
c9de560d 974 {Opt_stripe, "stripe=%u"},
ac27a0ec 975 {Opt_resize, "resize"},
64769240 976 {Opt_delalloc, "delalloc"},
dd919b98 977 {Opt_nodelalloc, "nodelalloc"},
240799cd 978 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
f3f12faa 979 {Opt_err, NULL},
ac27a0ec
DK		 980};
981
617ba13b 982static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 983{
617ba13b 984 ext4_fsblk_t sb_block;
ac27a0ec
DK		 985 char *options = (char *) *data;
986
987 if (!options || strncmp(options, "sb=", 3) != 0)
988 return 1; /* Default location */
989 options += 3;
617ba13b 990 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 991 sb_block = simple_strtoul(options, &options, 0);
992 if (*options && *options != ',') {
4776004f 993 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 994 (char *) *data);
995 return 1;
996 }
997 if (*options == ',')
998 options++;
999 *data = (void *) options;
1000 return sb_block;
1001}
1002
2b2d6d01
TT		 1003static int parse_options(char *options, struct super_block *sb,
1004 unsigned int *inum, unsigned long *journal_devnum,
1005 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1006{
617ba13b 1007 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1008 char *p;
ac27a0ec
DK		 1009 substring_t args[MAX_OPT_ARGS];
1010 int data_opt = 0;
1011 int option;
1012#ifdef CONFIG_QUOTA
dfc5d03f 1013 int qtype, qfmt;
ac27a0ec
DK		 1014 char *qname;
1015#endif
c07651b5 1016 ext4_fsblk_t last_block;
ac27a0ec
DK		 1017
1018 if (!options)
1019 return 1;
1020
2b2d6d01 1021 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1022 int token;
1023 if (!*p)
1024 continue;
1025
1026 token = match_token(p, tokens, args);
1027 switch (token) {
1028 case Opt_bsd_df:
2b2d6d01 1029 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1030 break;
1031 case Opt_minix_df:
2b2d6d01 1032 set_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1033 break;
1034 case Opt_grpid:
2b2d6d01 1035 set_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1036 break;
1037 case Opt_nogrpid:
2b2d6d01 1038 clear_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1039 break;
1040 case Opt_resuid:
1041 if (match_int(&args[0], &option))
1042 return 0;
1043 sbi->s_resuid = option;
1044 break;
1045 case Opt_resgid:
1046 if (match_int(&args[0], &option))
1047 return 0;
1048 sbi->s_resgid = option;
1049 break;
1050 case Opt_sb:
1051 /* handled by get_sb_block() instead of here */
1052 /* *sb_block = match_int(&args[0]); */
1053 break;
1054 case Opt_err_panic:
2b2d6d01
TT		 1055 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1056 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1057 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1058 break;
1059 case Opt_err_ro:
2b2d6d01
TT		 1060 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1061 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1062 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1063 break;
1064 case Opt_err_cont:
2b2d6d01
TT		 1065 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1066 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1067 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1068 break;
1069 case Opt_nouid32:
2b2d6d01 1070 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1071 break;
ac27a0ec 1072 case Opt_debug:
2b2d6d01 1073 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1074 break;
1075 case Opt_oldalloc:
2b2d6d01 1076 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1077 break;
1078 case Opt_orlov:
2b2d6d01 1079 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1080 break;
03010a33 1081#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1082 case Opt_user_xattr:
2b2d6d01 1083 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1084 break;
1085 case Opt_nouser_xattr:
2b2d6d01 1086 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1087 break;
1088#else
1089 case Opt_user_xattr:
1090 case Opt_nouser_xattr:
4776004f
TT		 1091 printk(KERN_ERR "EXT4 (no)user_xattr options "
1092 "not supported\n");
ac27a0ec
DK		 1093 break;
1094#endif
03010a33 1095#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1096 case Opt_acl:
1097 set_opt(sbi->s_mount_opt, POSIX_ACL);
1098 break;
1099 case Opt_noacl:
1100 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1101 break;
1102#else
1103 case Opt_acl:
1104 case Opt_noacl:
4776004f
TT		 1105 printk(KERN_ERR "EXT4 (no)acl options "
1106 "not supported\n");
ac27a0ec
DK		 1107 break;
1108#endif
1109 case Opt_reservation:
1110 set_opt(sbi->s_mount_opt, RESERVATION);
1111 break;
1112 case Opt_noreservation:
1113 clear_opt(sbi->s_mount_opt, RESERVATION);
1114 break;
1115 case Opt_journal_update:
1116 /* @@@ FIXME */
1117 /* Eventually we will want to be able to create
1118 a journal file here. For now, only allow the
1119 user to specify an existing inode to be the
1120 journal file. */
1121 if (is_remount) {
617ba13b 1122 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1123 "journal on remount\n");
1124 return 0;
1125 }
2b2d6d01 1126 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec
DK		 1127 break;
1128 case Opt_journal_inum:
1129 if (is_remount) {
617ba13b 1130 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1131 "journal on remount\n");
1132 return 0;
1133 }
1134 if (match_int(&args[0], &option))
1135 return 0;
1136 *inum = option;
1137 break;
1138 case Opt_journal_dev:
1139 if (is_remount) {
617ba13b 1140 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1141 "journal on remount\n");
1142 return 0;
1143 }
1144 if (match_int(&args[0], &option))
1145 return 0;
1146 *journal_devnum = option;
1147 break;
818d276c
GS		 1148 case Opt_journal_checksum:
1149 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1150 break;
1151 case Opt_journal_async_commit:
1152 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1153 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1154 break;
ac27a0ec 1155 case Opt_noload:
2b2d6d01 1156 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1157 break;
1158 case Opt_commit:
1159 if (match_int(&args[0], &option))
1160 return 0;
1161 if (option < 0)
1162 return 0;
1163 if (option == 0)
cd02ff0b 1164 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1165 sbi->s_commit_interval = HZ * option;
1166 break;
1167 case Opt_data_journal:
617ba13b 1168 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1169 goto datacheck;
1170 case Opt_data_ordered:
617ba13b 1171 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1172 goto datacheck;
1173 case Opt_data_writeback:
617ba13b 1174 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1175 datacheck:
1176 if (is_remount) {
617ba13b 1177 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1178 != data_opt) {
1179 printk(KERN_ERR
617ba13b 1180 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1181 "mode on remount\n");
1182 return 0;
1183 }
1184 } else {
617ba13b 1185 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1186 sbi->s_mount_opt |= data_opt;
1187 }
1188 break;
5bf5683a
HK		 1189 case Opt_data_err_abort:
1190 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1191 break;
1192 case Opt_data_err_ignore:
1193 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1194 break;
ac27a0ec
DK		 1195#ifdef CONFIG_QUOTA
1196 case Opt_usrjquota:
1197 qtype = USRQUOTA;
1198 goto set_qf_name;
1199 case Opt_grpjquota:
1200 qtype = GRPQUOTA;
1201set_qf_name:
dfc5d03f
JK		 1202 if ((sb_any_quota_enabled(sb) ||
1203 sb_any_quota_suspended(sb)) &&
1204 !sbi->s_qf_names[qtype]) {
ac27a0ec 1205 printk(KERN_ERR
4776004f
TT		 1206 "EXT4-fs: Cannot change journaled "
1207 "quota options when quota turned on.\n");
ac27a0ec
DK		 1208 return 0;
1209 }
1210 qname = match_strdup(&args[0]);
1211 if (!qname) {
1212 printk(KERN_ERR
617ba13b 1213 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1214 "storing quotafile name.\n");
1215 return 0;
1216 }
1217 if (sbi->s_qf_names[qtype] &&
1218 strcmp(sbi->s_qf_names[qtype], qname)) {
1219 printk(KERN_ERR
617ba13b 1220 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1221 "specified.\n", QTYPE2NAME(qtype));
1222 kfree(qname);
1223 return 0;
1224 }
1225 sbi->s_qf_names[qtype] = qname;
1226 if (strchr(sbi->s_qf_names[qtype], '/')) {
1227 printk(KERN_ERR
617ba13b 1228 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1229 "filesystem root.\n");
1230 kfree(sbi->s_qf_names[qtype]);
1231 sbi->s_qf_names[qtype] = NULL;
1232 return 0;
1233 }
1234 set_opt(sbi->s_mount_opt, QUOTA);
1235 break;
1236 case Opt_offusrjquota:
1237 qtype = USRQUOTA;
1238 goto clear_qf_name;
1239 case Opt_offgrpjquota:
1240 qtype = GRPQUOTA;
1241clear_qf_name:
dfc5d03f
JK		 1242 if ((sb_any_quota_enabled(sb) ||
1243 sb_any_quota_suspended(sb)) &&
1244 sbi->s_qf_names[qtype]) {
617ba13b 1245 printk(KERN_ERR "EXT4-fs: Cannot change "
2c8be6b2 1246 "journaled quota options when "
ac27a0ec
DK		 1247 "quota turned on.\n");
1248 return 0;
1249 }
1250 /*
1251 * The space will be released later when all options
1252 * are confirmed to be correct
1253 */
1254 sbi->s_qf_names[qtype] = NULL;
1255 break;
1256 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1257 qfmt = QFMT_VFS_OLD;
1258 goto set_qf_format;
ac27a0ec 1259 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1260 qfmt = QFMT_VFS_V0;
1261set_qf_format:
1262 if ((sb_any_quota_enabled(sb) ||
1263 sb_any_quota_suspended(sb)) &&
1264 sbi->s_jquota_fmt != qfmt) {
1265 printk(KERN_ERR "EXT4-fs: Cannot change "
1266 "journaled quota options when "
1267 "quota turned on.\n");
1268 return 0;
1269 }
1270 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1271 break;
1272 case Opt_quota:
1273 case Opt_usrquota:
1274 set_opt(sbi->s_mount_opt, QUOTA);
1275 set_opt(sbi->s_mount_opt, USRQUOTA);
1276 break;
1277 case Opt_grpquota:
1278 set_opt(sbi->s_mount_opt, QUOTA);
1279 set_opt(sbi->s_mount_opt, GRPQUOTA);
1280 break;
1281 case Opt_noquota:
1282 if (sb_any_quota_enabled(sb)) {
617ba13b 1283 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1284 "options when quota turned on.\n");
1285 return 0;
1286 }
1287 clear_opt(sbi->s_mount_opt, QUOTA);
1288 clear_opt(sbi->s_mount_opt, USRQUOTA);
1289 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1290 break;
1291#else
1292 case Opt_quota:
1293 case Opt_usrquota:
1294 case Opt_grpquota:
cd59e7b9
JK		 1295 printk(KERN_ERR
1296 "EXT4-fs: quota options not supported.\n");
1297 break;
ac27a0ec
DK		 1298 case Opt_usrjquota:
1299 case Opt_grpjquota:
1300 case Opt_offusrjquota:
1301 case Opt_offgrpjquota:
1302 case Opt_jqfmt_vfsold:
1303 case Opt_jqfmt_vfsv0:
1304 printk(KERN_ERR
cd59e7b9 1305 "EXT4-fs: journaled quota options not "
ac27a0ec
DK		 1306 "supported.\n");
1307 break;
1308 case Opt_noquota:
1309 break;
1310#endif
1311 case Opt_abort:
1312 set_opt(sbi->s_mount_opt, ABORT);
1313 break;
1314 case Opt_barrier:
1315 if (match_int(&args[0], &option))
1316 return 0;
1317 if (option)
1318 set_opt(sbi->s_mount_opt, BARRIER);
1319 else
1320 clear_opt(sbi->s_mount_opt, BARRIER);
1321 break;
1322 case Opt_ignore:
1323 break;
1324 case Opt_resize:
1325 if (!is_remount) {
617ba13b 1326 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1327 "for remount\n");
1328 return 0;
1329 }
1330 if (match_int(&args[0], &option) != 0)
1331 return 0;
1332 *n_blocks_count = option;
1333 break;
1334 case Opt_nobh:
1335 set_opt(sbi->s_mount_opt, NOBH);
1336 break;
1337 case Opt_bh:
1338 clear_opt(sbi->s_mount_opt, NOBH);
1339 break;
a86c6181 1340 case Opt_extents:
e4079a11
ES		 1341 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1342 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1343 ext4_warning(sb, __func__,
1344 "extents feature not enabled "
1345 "on this filesystem, use tune2fs\n");
1346 return 0;
1347 }
2b2d6d01 1348 set_opt(sbi->s_mount_opt, EXTENTS);
a86c6181 1349 break;
1e2462f9 1350 case Opt_noextents:
c07651b5
AK		 1351 /*
1352 * When e2fsprogs support resizing an already existing
1353 * ext3 file system to greater than 2**32 we need to
1354 * add support to block allocator to handle growing
1355 * already existing block mapped inode so that blocks
1356 * allocated for them fall within 2**32
1357 */
1358 last_block = ext4_blocks_count(sbi->s_es) - 1;
1359 if (last_block > 0xffffffffULL) {
1360 printk(KERN_ERR "EXT4-fs: Filesystem too "
1361 "large to mount with "
1362 "-o noextents options\n");
1363 return 0;
1364 }
2b2d6d01 1365 clear_opt(sbi->s_mount_opt, EXTENTS);
1e2462f9 1366 break;
25ec56b5
JNC		 1367 case Opt_i_version:
1368 set_opt(sbi->s_mount_opt, I_VERSION);
1369 sb->s_flags |= MS_I_VERSION;
1370 break;
dd919b98
AK		 1371 case Opt_nodelalloc:
1372 clear_opt(sbi->s_mount_opt, DELALLOC);
1373 break;
c9de560d
AT		 1374 case Opt_stripe:
1375 if (match_int(&args[0], &option))
1376 return 0;
1377 if (option < 0)
1378 return 0;
1379 sbi->s_stripe = option;
1380 break;
64769240
AT		 1381 case Opt_delalloc:
1382 set_opt(sbi->s_mount_opt, DELALLOC);
1383 break;
240799cd
TT		 1384 case Opt_inode_readahead_blks:
1385 if (match_int(&args[0], &option))
1386 return 0;
1387 if (option < 0 || option > (1 << 30))
1388 return 0;
1389 sbi->s_inode_readahead_blks = option;
1390 break;
ac27a0ec 1391 default:
2b2d6d01
TT		 1392 printk(KERN_ERR
1393 "EXT4-fs: Unrecognized mount option \"%s\" "
1394 "or missing value\n", p);
ac27a0ec
DK		 1395 return 0;
1396 }
1397 }
1398#ifdef CONFIG_QUOTA
1399 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1400 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1401 sbi->s_qf_names[USRQUOTA])
1402 clear_opt(sbi->s_mount_opt, USRQUOTA);
1403
617ba13b 1404 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1405 sbi->s_qf_names[GRPQUOTA])
1406 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1407
1408 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1409 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1410 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1411 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1412 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1413 "format mixing.\n");
1414 return 0;
1415 }
1416
1417 if (!sbi->s_jquota_fmt) {
2c8be6b2 1418 printk(KERN_ERR "EXT4-fs: journaled quota format "
ac27a0ec
DK		 1419 "not specified.\n");
1420 return 0;
1421 }
1422 } else {
1423 if (sbi->s_jquota_fmt) {
2c8be6b2
JK		 1424 printk(KERN_ERR "EXT4-fs: journaled quota format "
1425 "specified with no journaling "
ac27a0ec
DK		 1426 "enabled.\n");
1427 return 0;
1428 }
1429 }
1430#endif
1431 return 1;
1432}
1433
617ba13b 1434static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1435 int read_only)
1436{
617ba13b 1437 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1438 int res = 0;
1439
617ba13b 1440 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2b2d6d01
TT		 1441 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1442 "forcing read-only mode\n");
ac27a0ec
DK		 1443 res = MS_RDONLY;
1444 }
1445 if (read_only)
1446 return res;
617ba13b 1447 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2b2d6d01
TT		 1448 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1449 "running e2fsck is recommended\n");
617ba13b 1450 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
2b2d6d01
TT		 1451 printk(KERN_WARNING
1452 "EXT4-fs warning: mounting fs with errors, "
1453 "running e2fsck is recommended\n");
ac27a0ec
DK		 1454 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1455 le16_to_cpu(es->s_mnt_count) >=
1456 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2b2d6d01
TT		 1457 printk(KERN_WARNING
1458 "EXT4-fs warning: maximal mount count reached, "
1459 "running e2fsck is recommended\n");
ac27a0ec
DK		 1460 else if (le32_to_cpu(es->s_checkinterval) &&
1461 (le32_to_cpu(es->s_lastcheck) +
1462 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2b2d6d01
TT		 1463 printk(KERN_WARNING
1464 "EXT4-fs warning: checktime reached, "
1465 "running e2fsck is recommended\n");
ac27a0ec
DK		 1466#if 0
1467 /* @@@ We _will_ want to clear the valid bit if we find
63f57933
AM		 1468 * inconsistencies, to force a fsck at reboot. But for
1469 * a plain journaled filesystem we can keep it set as
1470 * valid forever! :)
1471 */
216c34b2 1472 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec
DK		 1473#endif
1474 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1475 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1476 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1477 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b
MC		 1478 ext4_update_dynamic_rev(sb);
1479 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1480
617ba13b 1481 ext4_commit_super(sb, es, 1);
ac27a0ec 1482 if (test_opt(sb, DEBUG))
617ba13b 1483 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
ac27a0ec
DK		 1484 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1485 sb->s_blocksize,
1486 sbi->s_groups_count,
617ba13b
MC		 1487 EXT4_BLOCKS_PER_GROUP(sb),
1488 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1489 sbi->s_mount_opt);
1490
05496769
TT		 1491 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1492 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1493 "external", EXT4_SB(sb)->s_journal->j_devname);
ac27a0ec
DK		 1494 return res;
1495}
1496
772cb7c8
JS		 1497static int ext4_fill_flex_info(struct super_block *sb)
1498{
1499 struct ext4_sb_info *sbi = EXT4_SB(sb);
1500 struct ext4_group_desc *gdp = NULL;
1501 struct buffer_head *bh;
1502 ext4_group_t flex_group_count;
1503 ext4_group_t flex_group;
1504 int groups_per_flex = 0;
1505 __u64 block_bitmap = 0;
1506 int i;
1507
1508 if (!sbi->s_es->s_log_groups_per_flex) {
1509 sbi->s_log_groups_per_flex = 0;
1510 return 1;
1511 }
1512
1513 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1514 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1515
c62a11fd
FB		 1516 /* We allocate both existing and potentially added groups */
1517 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1518 ((sbi->s_es->s_reserved_gdt_blocks +1 ) <<
1519 EXT4_DESC_PER_BLOCK_BITS(sb))) /
1520 groups_per_flex;
ec05e868 1521 sbi->s_flex_groups = kzalloc(flex_group_count *
772cb7c8
JS		 1522 sizeof(struct flex_groups), GFP_KERNEL);
1523 if (sbi->s_flex_groups == NULL) {
ec05e868
LZ		 1524 printk(KERN_ERR "EXT4-fs: not enough memory for "
1525 "%lu flex groups\n", flex_group_count);
772cb7c8
JS		 1526 goto failed;
1527 }
772cb7c8
JS		 1528
1529 gdp = ext4_get_group_desc(sb, 1, &bh);
1530 block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1531
1532 for (i = 0; i < sbi->s_groups_count; i++) {
1533 gdp = ext4_get_group_desc(sb, i, &bh);
1534
1535 flex_group = ext4_flex_group(sbi, i);
1536 sbi->s_flex_groups[flex_group].free_inodes +=
1537 le16_to_cpu(gdp->bg_free_inodes_count);
1538 sbi->s_flex_groups[flex_group].free_blocks +=
1539 le16_to_cpu(gdp->bg_free_blocks_count);
1540 }
1541
1542 return 1;
1543failed:
1544 return 0;
1545}
1546
717d50e4
AD		 1547__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1548 struct ext4_group_desc *gdp)
1549{
1550 __u16 crc = 0;
1551
1552 if (sbi->s_es->s_feature_ro_compat &
1553 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1554 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1555 __le32 le_group = cpu_to_le32(block_group);
1556
1557 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1558 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1559 crc = crc16(crc, (__u8 *)gdp, offset);
1560 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1561 /* for checksum of struct ext4_group_desc do the rest...*/
1562 if ((sbi->s_es->s_feature_incompat &
1563 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1564 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1565 crc = crc16(crc, (__u8 *)gdp + offset,
1566 le16_to_cpu(sbi->s_es->s_desc_size) -
1567 offset);
1568 }
1569
1570 return cpu_to_le16(crc);
1571}
1572
1573int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1574 struct ext4_group_desc *gdp)
1575{
1576 if ((sbi->s_es->s_feature_ro_compat &
1577 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1578 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1579 return 0;
1580
1581 return 1;
1582}
1583
ac27a0ec 1584/* Called at mount-time, super-block is locked */
197cd65a 1585static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1586{
617ba13b
MC		 1587 struct ext4_sb_info *sbi = EXT4_SB(sb);
1588 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1589 ext4_fsblk_t last_block;
bd81d8ee
LV		 1590 ext4_fsblk_t block_bitmap;
1591 ext4_fsblk_t inode_bitmap;
1592 ext4_fsblk_t inode_table;
ce421581 1593 int flexbg_flag = 0;
fd2d4291 1594 ext4_group_t i;
ac27a0ec 1595
ce421581
JS		 1596 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1597 flexbg_flag = 1;
1598
af5bc92d 1599 ext4_debug("Checking group descriptors");
ac27a0ec 1600
197cd65a
AM		 1601 for (i = 0; i < sbi->s_groups_count; i++) {
1602 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1603
ce421581 1604 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1605 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1606 else
1607 last_block = first_block +
617ba13b 1608 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1609
8fadc143 1610 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1611 if (block_bitmap < first_block || block_bitmap > last_block) {
c19204b0
JB		 1612 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1613 "Block bitmap for group %lu not in group "
5128273a 1614 "(block %llu)!\n", i, block_bitmap);
ac27a0ec
DK		 1615 return 0;
1616 }
8fadc143 1617 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1618 if (inode_bitmap < first_block || inode_bitmap > last_block) {
c19204b0
JB		 1619 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1620 "Inode bitmap for group %lu not in group "
5128273a 1621 "(block %llu)!\n", i, inode_bitmap);
ac27a0ec
DK		 1622 return 0;
1623 }
8fadc143 1624 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1625 if (inode_table < first_block ||
2b2d6d01 1626 inode_table + sbi->s_itb_per_group - 1 > last_block) {
c19204b0
JB		 1627 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1628 "Inode table for group %lu not in group "
5128273a 1629 "(block %llu)!\n", i, inode_table);
ac27a0ec
DK		 1630 return 0;
1631 }
b5f10eed 1632 spin_lock(sb_bgl_lock(sbi, i));
717d50e4 1633 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
c19204b0
JB		 1634 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1635 "Checksum for group %lu failed (%u!=%u)\n",
1636 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1637 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c
LZ		 1638 if (!(sb->s_flags & MS_RDONLY)) {
1639 spin_unlock(sb_bgl_lock(sbi, i));
8a266467 1640 return 0;
7ee1ec4c 1641 }
717d50e4 1642 }
b5f10eed 1643 spin_unlock(sb_bgl_lock(sbi, i));
ce421581
JS		 1644 if (!flexbg_flag)
1645 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1646 }
1647
bd81d8ee 1648 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2b2d6d01 1649 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1650 return 1;
1651}
1652
617ba13b 1653/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1654 * the superblock) which were deleted from all directories, but held open by
1655 * a process at the time of a crash. We walk the list and try to delete these
1656 * inodes at recovery time (only with a read-write filesystem).
1657 *
1658 * In order to keep the orphan inode chain consistent during traversal (in
1659 * case of crash during recovery), we link each inode into the superblock
1660 * orphan list_head and handle it the same way as an inode deletion during
1661 * normal operation (which journals the operations for us).
1662 *
1663 * We only do an iget() and an iput() on each inode, which is very safe if we
1664 * accidentally point at an in-use or already deleted inode. The worst that
1665 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1666 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1667 * e2fsck was run on this filesystem, and it must have already done the orphan
1668 * inode cleanup for us, so we can safely abort without any further action.
1669 */
2b2d6d01
TT		 1670static void ext4_orphan_cleanup(struct super_block *sb,
1671 struct ext4_super_block *es)
ac27a0ec
DK		 1672{
1673 unsigned int s_flags = sb->s_flags;
1674 int nr_orphans = 0, nr_truncates = 0;
1675#ifdef CONFIG_QUOTA
1676 int i;
1677#endif
1678 if (!es->s_last_orphan) {
1679 jbd_debug(4, "no orphan inodes to clean up\n");
1680 return;
1681 }
1682
a8f48a95
ES		 1683 if (bdev_read_only(sb->s_bdev)) {
1684 printk(KERN_ERR "EXT4-fs: write access "
1685 "unavailable, skipping orphan cleanup.\n");
1686 return;
1687 }
1688
617ba13b 1689 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1690 if (es->s_last_orphan)
1691 jbd_debug(1, "Errors on filesystem, "
1692 "clearing orphan list.\n");
1693 es->s_last_orphan = 0;
1694 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1695 return;
1696 }
1697
1698 if (s_flags & MS_RDONLY) {
617ba13b 1699 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1700 sb->s_id);
1701 sb->s_flags &= ~MS_RDONLY;
1702 }
1703#ifdef CONFIG_QUOTA
1704 /* Needed for iput() to work correctly and not trash data */
1705 sb->s_flags |= MS_ACTIVE;
1706 /* Turn on quotas so that they are updated correctly */
1707 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1708 if (EXT4_SB(sb)->s_qf_names[i]) {
1709 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1710 if (ret < 0)
1711 printk(KERN_ERR
2c8be6b2 1712 "EXT4-fs: Cannot turn on journaled "
ac27a0ec
DK		 1713 "quota: error %d\n", ret);
1714 }
1715 }
1716#endif
1717
1718 while (es->s_last_orphan) {
1719 struct inode *inode;
1720
97bd42b9
JB		 1721 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1722 if (IS_ERR(inode)) {
ac27a0ec
DK		 1723 es->s_last_orphan = 0;
1724 break;
1725 }
1726
617ba13b 1727 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1728 DQUOT_INIT(inode);
1729 if (inode->i_nlink) {
1730 printk(KERN_DEBUG
e5f8eab8 1731 "%s: truncating inode %lu to %lld bytes\n",
46e665e9 1732 __func__, inode->i_ino, inode->i_size);
e5f8eab8 1733 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 1734 inode->i_ino, inode->i_size);
617ba13b 1735 ext4_truncate(inode);
ac27a0ec
DK		 1736 nr_truncates++;
1737 } else {
1738 printk(KERN_DEBUG
1739 "%s: deleting unreferenced inode %lu\n",
46e665e9 1740 __func__, inode->i_ino);
ac27a0ec
DK		 1741 jbd_debug(2, "deleting unreferenced inode %lu\n",
1742 inode->i_ino);
1743 nr_orphans++;
1744 }
1745 iput(inode); /* The delete magic happens here! */
1746 }
1747
2b2d6d01 1748#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 1749
1750 if (nr_orphans)
617ba13b 1751 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1752 sb->s_id, PLURAL(nr_orphans));
1753 if (nr_truncates)
617ba13b 1754 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1755 sb->s_id, PLURAL(nr_truncates));
1756#ifdef CONFIG_QUOTA
1757 /* Turn quotas off */
1758 for (i = 0; i < MAXQUOTAS; i++) {
1759 if (sb_dqopt(sb)->files[i])
6f28e087 1760 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1761 }
1762#endif
1763 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1764}
cd2291a4
ES		 1765/*
1766 * Maximal extent format file size.
1767 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1768 * extent format containers, within a sector_t, and within i_blocks
1769 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1770 * so that won't be a limiting factor.
1771 *
1772 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1773 */
1774static loff_t ext4_max_size(int blkbits)
1775{
1776 loff_t res;
1777 loff_t upper_limit = MAX_LFS_FILESIZE;
1778
1779 /* small i_blocks in vfs inode? */
1780 if (sizeof(blkcnt_t) < sizeof(u64)) {
1781 /*
1782 * CONFIG_LSF is not enabled implies the inode
1783 * i_block represent total blocks in 512 bytes
1784 * 32 == size of vfs inode i_blocks * 8
1785 */
1786 upper_limit = (1LL << 32) - 1;
1787
1788 /* total blocks in file system block size */
1789 upper_limit >>= (blkbits - 9);
1790 upper_limit <<= blkbits;
1791 }
1792
1793 /* 32-bit extent-start container, ee_block */
1794 res = 1LL << 32;
1795 res <<= blkbits;
1796 res -= 1;
1797
1798 /* Sanity check against vm- & vfs- imposed limits */
1799 if (res > upper_limit)
1800 res = upper_limit;
1801
1802 return res;
1803}
ac27a0ec 1804
ac27a0ec 1805/*
cd2291a4 1806 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1807 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1808 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1809 */
cd2291a4 1810static loff_t ext4_max_bitmap_size(int bits)
ac27a0ec 1811{
617ba13b 1812 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1813 int meta_blocks;
1814 loff_t upper_limit;
1815 /* This is calculated to be the largest file size for a
cd2291a4 1816 * dense, bitmapped file such that the total number of
ac27a0ec 1817 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1818 * does not exceed 2^48 -1
1819 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1820 * total number of 512 bytes blocks of the file
1821 */
1822
1823 if (sizeof(blkcnt_t) < sizeof(u64)) {
1824 /*
1825 * CONFIG_LSF is not enabled implies the inode
1826 * i_block represent total blocks in 512 bytes
1827 * 32 == size of vfs inode i_blocks * 8
1828 */
1829 upper_limit = (1LL << 32) - 1;
1830
1831 /* total blocks in file system block size */
1832 upper_limit >>= (bits - 9);
1833
1834 } else {
8180a562
AK		 1835 /*
1836 * We use 48 bit ext4_inode i_blocks
1837 * With EXT4_HUGE_FILE_FL set the i_blocks
1838 * represent total number of blocks in
1839 * file system block size
1840 */
0fc1b451
AK		 1841 upper_limit = (1LL << 48) - 1;
1842
0fc1b451
AK		 1843 }
1844
1845 /* indirect blocks */
1846 meta_blocks = 1;
1847 /* double indirect blocks */
1848 meta_blocks += 1 + (1LL << (bits-2));
1849 /* tripple indirect blocks */
1850 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1851
1852 upper_limit -= meta_blocks;
1853 upper_limit <<= bits;
ac27a0ec
DK		 1854
1855 res += 1LL << (bits-2);
1856 res += 1LL << (2*(bits-2));
1857 res += 1LL << (3*(bits-2));
1858 res <<= bits;
1859 if (res > upper_limit)
1860 res = upper_limit;
0fc1b451
AK		 1861
1862 if (res > MAX_LFS_FILESIZE)
1863 res = MAX_LFS_FILESIZE;
1864
ac27a0ec
DK		 1865 return res;
1866}
1867
617ba13b 1868static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1869 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1870{
617ba13b 1871 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1872 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1873 int has_super = 0;
1874
1875 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1876
617ba13b 1877 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1878 nr < first_meta_bg)
70bbb3e0 1879 return logical_sb_block + nr + 1;
ac27a0ec 1880 bg = sbi->s_desc_per_block * nr;
617ba13b 1881 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1882 has_super = 1;
617ba13b 1883 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1884}
1885
c9de560d
AT		 1886/**
1887 * ext4_get_stripe_size: Get the stripe size.
1888 * @sbi: In memory super block info
1889 *
1890 * If we have specified it via mount option, then
1891 * use the mount option value. If the value specified at mount time is
1892 * greater than the blocks per group use the super block value.
1893 * If the super block value is greater than blocks per group return 0.
1894 * Allocator needs it be less than blocks per group.
1895 *
1896 */
1897static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1898{
1899 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1900 unsigned long stripe_width =
1901 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1902
1903 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1904 return sbi->s_stripe;
1905
1906 if (stripe_width <= sbi->s_blocks_per_group)
1907 return stripe_width;
1908
1909 if (stride <= sbi->s_blocks_per_group)
1910 return stride;
1911
1912 return 0;
1913}
ac27a0ec 1914
2b2d6d01 1915static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 1916 __releases(kernel_lock)
1917 __acquires(kernel_lock)
1d03ec98 1918
ac27a0ec 1919{
2b2d6d01 1920 struct buffer_head *bh;
617ba13b
MC		 1921 struct ext4_super_block *es = NULL;
1922 struct ext4_sb_info *sbi;
1923 ext4_fsblk_t block;
1924 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 1925 ext4_fsblk_t logical_sb_block;
ac27a0ec
DK		 1926 unsigned long offset = 0;
1927 unsigned int journal_inum = 0;
1928 unsigned long journal_devnum = 0;
1929 unsigned long def_mount_opts;
1930 struct inode *root;
9f6200bb 1931 char *cp;
1d1fe1ee 1932 int ret = -EINVAL;
ac27a0ec 1933 int blocksize;
ac27a0ec
DK		 1934 int db_count;
1935 int i;
1936 int needs_recovery;
1937 __le32 features;
bd81d8ee 1938 __u64 blocks_count;
833f4077 1939 int err;
ac27a0ec
DK		 1940
1941 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1942 if (!sbi)
1943 return -ENOMEM;
1944 sb->s_fs_info = sbi;
1945 sbi->s_mount_opt = 0;
617ba13b
MC		 1946 sbi->s_resuid = EXT4_DEF_RESUID;
1947 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 1948 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 1949 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 1950
1951 unlock_kernel();
1952
9f6200bb
TT		 1953 /* Cleanup superblock name */
1954 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
1955 *cp = '!';
1956
617ba13b 1957 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 1958 if (!blocksize) {
617ba13b 1959 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 1960 goto out_fail;
1961 }
1962
1963 /*
617ba13b 1964 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 1965 * block sizes. We need to calculate the offset from buffer start.
1966 */
617ba13b 1967 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 1968 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1969 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 1970 } else {
70bbb3e0 1971 logical_sb_block = sb_block;
ac27a0ec
DK		 1972 }
1973
70bbb3e0 1974 if (!(bh = sb_bread(sb, logical_sb_block))) {
2b2d6d01 1975 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 1976 goto out_fail;
1977 }
1978 /*
1979 * Note: s_es must be initialized as soon as possible because
617ba13b 1980 * some ext4 macro-instructions depend on its value
ac27a0ec 1981 */
617ba13b 1982 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 1983 sbi->s_es = es;
1984 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 1985 if (sb->s_magic != EXT4_SUPER_MAGIC)
1986 goto cantfind_ext4;
ac27a0ec
DK		 1987
1988 /* Set defaults before we parse the mount options */
1989 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 1990 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 1991 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 1992 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 1993 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 1994 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 1995 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 1996#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 1997 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 1998 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 1999#endif
03010a33 2000#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2001 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2002 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2003#endif
617ba13b
MC		 2004 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2005 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2006 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2007 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2008 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2009 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2010
2011 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2012 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2013 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2014 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2015 else
2016 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2017
2018 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2019 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2020
2021 set_opt(sbi->s_mount_opt, RESERVATION);
571640ca 2022 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2023
1e2462f9
MC		 2024 /*
2025 * turn on extents feature by default in ext4 filesystem
e4079a11
ES		 2026 * only if feature flag already set by mkfs or tune2fs.
2027 * Use -o noextents to turn it off
1e2462f9 2028 */
e4079a11
ES		 2029 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
2030 set_opt(sbi->s_mount_opt, EXTENTS);
2031 else
2032 ext4_warning(sb, __func__,
2033 "extents feature not enabled on this filesystem, "
2034 "use tune2fs.\n");
1e2462f9 2035
dd919b98
AK		 2036 /*
2037 * enable delayed allocation by default
2038 * Use -o nodelalloc to turn it off
2039 */
2040 set_opt(sbi->s_mount_opt, DELALLOC);
2041
2042
2b2d6d01
TT		 2043 if (!parse_options((char *) data, sb, &journal_inum, &journal_devnum,
2044 NULL, 0))
ac27a0ec
DK		 2045 goto failed_mount;
2046
2047 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2048 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2049
617ba13b
MC		 2050 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2051 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2052 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2053 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 2054 printk(KERN_WARNING
617ba13b 2055 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec 2056 "running e2fsck is recommended\n");
469108ff 2057
ac27a0ec
DK		 2058 /*
2059 * Check feature flags regardless of the revision level, since we
2060 * previously didn't change the revision level when setting the flags,
2061 * so there is a chance incompat flags are set on a rev 0 filesystem.
2062 */
617ba13b 2063 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 2064 if (features) {
617ba13b 2065 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
ac27a0ec
DK		 2066 "unsupported optional features (%x).\n",
2067 sb->s_id, le32_to_cpu(features));
2068 goto failed_mount;
2069 }
617ba13b 2070 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 2071 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 2072 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
ac27a0ec
DK		 2073 "unsupported optional features (%x).\n",
2074 sb->s_id, le32_to_cpu(features));
2075 goto failed_mount;
2076 }
0fc1b451
AK		 2077 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2078 /*
2079 * Large file size enabled file system can only be
2080 * mount if kernel is build with CONFIG_LSF
2081 */
2082 if (sizeof(root->i_blocks) < sizeof(u64) &&
2083 !(sb->s_flags & MS_RDONLY)) {
2084 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2085 "files cannot be mounted read-write "
2086 "without CONFIG_LSF.\n", sb->s_id);
2087 goto failed_mount;
2088 }
2089 }
ac27a0ec
DK		 2090 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2091
617ba13b
MC		 2092 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2093 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 2094 printk(KERN_ERR
617ba13b 2095 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 2096 blocksize, sb->s_id);
2097 goto failed_mount;
2098 }
2099
ac27a0ec 2100 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2101
2102 /* Validate the filesystem blocksize */
2103 if (!sb_set_blocksize(sb, blocksize)) {
2104 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2105 blocksize);
ac27a0ec
DK		 2106 goto failed_mount;
2107 }
2108
2b2d6d01 2109 brelse(bh);
70bbb3e0
AM		 2110 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2111 offset = do_div(logical_sb_block, blocksize);
2112 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 2113 if (!bh) {
2114 printk(KERN_ERR
617ba13b 2115 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 2116 goto failed_mount;
2117 }
617ba13b 2118 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2119 sbi->s_es = es;
617ba13b 2120 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2b2d6d01
TT		 2121 printk(KERN_ERR
2122 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 2123 goto failed_mount;
2124 }
2125 }
2126
e2b46574 2127 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
617ba13b 2128 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
ac27a0ec 2129
617ba13b
MC		 2130 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2131 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2132 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2133 } else {
2134 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2135 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2136 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2137 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2138 (sbi->s_inode_size > blocksize)) {
2b2d6d01
TT		 2139 printk(KERN_ERR
2140 "EXT4-fs: unsupported inode size: %d\n",
2141 sbi->s_inode_size);
ac27a0ec
DK		 2142 goto failed_mount;
2143 }
ef7f3835
KS		 2144 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2145 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2146 }
0d1ee42f
AR		 2147 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2148 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2149 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2150 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2151 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2152 printk(KERN_ERR
8fadc143 2153 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2154 sbi->s_desc_size);
2155 goto failed_mount;
2156 }
2157 } else
2158 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2159 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2160 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2161 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2162 goto cantfind_ext4;
2163 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2164 if (sbi->s_inodes_per_block == 0)
617ba13b 2165 goto cantfind_ext4;
ac27a0ec
DK		 2166 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2167 sbi->s_inodes_per_block;
0d1ee42f 2168 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2169 sbi->s_sbh = bh;
2170 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2171 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2172 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2b2d6d01 2173 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2174 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2175 sbi->s_def_hash_version = es->s_def_hash_version;
2176
2177 if (sbi->s_blocks_per_group > blocksize * 8) {
2b2d6d01
TT		 2178 printk(KERN_ERR
2179 "EXT4-fs: #blocks per group too big: %lu\n",
2180 sbi->s_blocks_per_group);
ac27a0ec
DK		 2181 goto failed_mount;
2182 }
ac27a0ec 2183 if (sbi->s_inodes_per_group > blocksize * 8) {
2b2d6d01
TT		 2184 printk(KERN_ERR
2185 "EXT4-fs: #inodes per group too big: %lu\n",
2186 sbi->s_inodes_per_group);
ac27a0ec
DK		 2187 goto failed_mount;
2188 }
2189
bd81d8ee 2190 if (ext4_blocks_count(es) >
ac27a0ec 2191 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2192 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2193 " too large to mount safely\n", sb->s_id);
2194 if (sizeof(sector_t) < 8)
617ba13b 2195 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2196 "enabled\n");
2197 goto failed_mount;
2198 }
2199
617ba13b
MC		 2200 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2201 goto cantfind_ext4;
e7c95593
ES		 2202
2203 /* ensure blocks_count calculation below doesn't sign-extend */
2204 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2205 le32_to_cpu(es->s_first_data_block) + 1) {
2206 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2207 "first data block %u, blocks per group %lu\n",
2208 ext4_blocks_count(es),
2209 le32_to_cpu(es->s_first_data_block),
2210 EXT4_BLOCKS_PER_GROUP(sb));
2211 goto failed_mount;
2212 }
bd81d8ee
LV		 2213 blocks_count = (ext4_blocks_count(es) -
2214 le32_to_cpu(es->s_first_data_block) +
2215 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2216 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2217 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2218 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2219 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2220 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2221 GFP_KERNEL);
2222 if (sbi->s_group_desc == NULL) {
2b2d6d01 2223 printk(KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2224 goto failed_mount;
2225 }
2226
3244fcb1 2227#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2228 if (ext4_proc_root)
2229 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2230
240799cd
TT		 2231 if (sbi->s_proc)
2232 proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2233 &ext4_ui_proc_fops,
2234 &sbi->s_inode_readahead_blks);
3244fcb1 2235#endif
240799cd 2236
ac27a0ec
DK		 2237 bgl_lock_init(&sbi->s_blockgroup_lock);
2238
2239 for (i = 0; i < db_count; i++) {
70bbb3e0 2240 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2241 sbi->s_group_desc[i] = sb_bread(sb, block);
2242 if (!sbi->s_group_desc[i]) {
2b2d6d01
TT		 2243 printk(KERN_ERR "EXT4-fs: "
2244 "can't read group descriptor %d\n", i);
ac27a0ec
DK		 2245 db_count = i;
2246 goto failed_mount2;
2247 }
2248 }
2b2d6d01 2249 if (!ext4_check_descriptors(sb)) {
617ba13b 2250 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2251 goto failed_mount2;
2252 }
772cb7c8
JS		 2253 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2254 if (!ext4_fill_flex_info(sb)) {
2255 printk(KERN_ERR
2256 "EXT4-fs: unable to initialize "
2257 "flex_bg meta info!\n");
2258 goto failed_mount2;
2259 }
2260
ac27a0ec
DK		 2261 sbi->s_gdb_count = db_count;
2262 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2263 spin_lock_init(&sbi->s_next_gen_lock);
2264
833f4077
PZ		 2265 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2266 ext4_count_free_blocks(sb));
2267 if (!err) {
2268 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2269 ext4_count_free_inodes(sb));
2270 }
2271 if (!err) {
2272 err = percpu_counter_init(&sbi->s_dirs_counter,
2273 ext4_count_dirs(sb));
2274 }
6bc6e63f
AK		 2275 if (!err) {
2276 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2277 }
833f4077
PZ		 2278 if (err) {
2279 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2280 goto failed_mount3;
2281 }
ac27a0ec 2282
c9de560d
AT		 2283 sbi->s_stripe = ext4_get_stripe_size(sbi);
2284
ac27a0ec
DK		 2285 /*
2286 * set up enough so that it can read an inode
2287 */
617ba13b
MC		 2288 sb->s_op = &ext4_sops;
2289 sb->s_export_op = &ext4_export_ops;
2290 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2291#ifdef CONFIG_QUOTA
617ba13b
MC		 2292 sb->s_qcop = &ext4_qctl_operations;
2293 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2294#endif
2295 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2296
2297 sb->s_root = NULL;
2298
2299 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2300 EXT4_HAS_INCOMPAT_FEATURE(sb,
2301 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2302
2303 /*
2304 * The first inode we look at is the journal inode. Don't try
2305 * root first: it may be modified in the journal!
2306 */
2307 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2308 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2309 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2310 goto failed_mount3;
624080ed
TT		 2311 if (!(sb->s_flags & MS_RDONLY) &&
2312 EXT4_SB(sb)->s_journal->j_failed_commit) {
2313 printk(KERN_CRIT "EXT4-fs error (device %s): "
2314 "ext4_fill_super: Journal transaction "
2b2d6d01 2315 "%u is corrupt\n", sb->s_id,
624080ed 2316 EXT4_SB(sb)->s_journal->j_failed_commit);
2b2d6d01
TT		 2317 if (test_opt(sb, ERRORS_RO)) {
2318 printk(KERN_CRIT
2319 "Mounting filesystem read-only\n");
624080ed
TT		 2320 sb->s_flags |= MS_RDONLY;
2321 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2322 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2323 }
2324 if (test_opt(sb, ERRORS_PANIC)) {
2325 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2326 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2327 ext4_commit_super(sb, es, 1);
2328 printk(KERN_CRIT
2329 "EXT4-fs (device %s): mount failed\n",
2330 sb->s_id);
2331 goto failed_mount4;
2332 }
2333 }
ac27a0ec 2334 } else if (journal_inum) {
617ba13b 2335 if (ext4_create_journal(sb, es, journal_inum))
ac27a0ec
DK		 2336 goto failed_mount3;
2337 } else {
2338 if (!silent)
2b2d6d01
TT		 2339 printk(KERN_ERR
2340 "ext4: No journal on filesystem on %s\n",
2341 sb->s_id);
ac27a0ec
DK		 2342 goto failed_mount3;
2343 }
2344
eb40a09c
JS		 2345 if (ext4_blocks_count(es) > 0xffffffffULL &&
2346 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2347 JBD2_FEATURE_INCOMPAT_64BIT)) {
2348 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2349 goto failed_mount4;
2350 }
2351
818d276c
GS		 2352 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2353 jbd2_journal_set_features(sbi->s_journal,
2354 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2355 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2356 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2357 jbd2_journal_set_features(sbi->s_journal,
2358 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2359 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2360 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2361 } else {
2362 jbd2_journal_clear_features(sbi->s_journal,
2363 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2364 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2365 }
2366
ac27a0ec
DK		 2367 /* We have now updated the journal if required, so we can
2368 * validate the data journaling mode. */
2369 switch (test_opt(sb, DATA_FLAGS)) {
2370 case 0:
2371 /* No mode set, assume a default based on the journal
63f57933
AM		 2372 * capabilities: ORDERED_DATA if the journal can
2373 * cope, else JOURNAL_DATA
2374 */
dab291af
MC		 2375 if (jbd2_journal_check_available_features
2376 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2377 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2378 else
2379 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2380 break;
2381
617ba13b
MC		 2382 case EXT4_MOUNT_ORDERED_DATA:
2383 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2384 if (!jbd2_journal_check_available_features
2385 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2386 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2387 "requested data journaling mode\n");
2388 goto failed_mount4;
2389 }
2390 default:
2391 break;
2392 }
2393
2394 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2395 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2396 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2397 "its supported only with writeback mode\n");
2398 clear_opt(sbi->s_mount_opt, NOBH);
2399 }
2400 }
2401 /*
dab291af 2402 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2403 * so we can safely mount the rest of the filesystem now.
2404 */
2405
1d1fe1ee
DH		 2406 root = ext4_iget(sb, EXT4_ROOT_INO);
2407 if (IS_ERR(root)) {
617ba13b 2408 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1d1fe1ee 2409 ret = PTR_ERR(root);
ac27a0ec
DK		 2410 goto failed_mount4;
2411 }
2412 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2413 iput(root);
617ba13b 2414 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2415 goto failed_mount4;
2416 }
1d1fe1ee
DH		 2417 sb->s_root = d_alloc_root(root);
2418 if (!sb->s_root) {
2419 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2420 iput(root);
2421 ret = -ENOMEM;
2422 goto failed_mount4;
2423 }
ac27a0ec 2424
2b2d6d01 2425 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2426
2427 /* determine the minimum size of new large inodes, if present */
2428 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2429 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2430 EXT4_GOOD_OLD_INODE_SIZE;
2431 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2432 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2433 if (sbi->s_want_extra_isize <
2434 le16_to_cpu(es->s_want_extra_isize))
2435 sbi->s_want_extra_isize =
2436 le16_to_cpu(es->s_want_extra_isize);
2437 if (sbi->s_want_extra_isize <
2438 le16_to_cpu(es->s_min_extra_isize))
2439 sbi->s_want_extra_isize =
2440 le16_to_cpu(es->s_min_extra_isize);
2441 }
2442 }
2443 /* Check if enough inode space is available */
2444 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2445 sbi->s_inode_size) {
2446 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2447 EXT4_GOOD_OLD_INODE_SIZE;
2448 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2449 "available.\n");
2450 }
2451
c2774d84
AK		 2452 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2453 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2454 "requested data journaling mode\n");
2455 clear_opt(sbi->s_mount_opt, DELALLOC);
2456 } else if (test_opt(sb, DELALLOC))
2457 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2458
2459 ext4_ext_init(sb);
2460 err = ext4_mb_init(sb, needs_recovery);
2461 if (err) {
2462 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2463 err);
2464 goto failed_mount4;
2465 }
2466
ac27a0ec
DK		 2467 /*
2468 * akpm: core read_super() calls in here with the superblock locked.
2469 * That deadlocks, because orphan cleanup needs to lock the superblock
2470 * in numerous places. Here we just pop the lock - it's relatively
2471 * harmless, because we are now ready to accept write_super() requests,
2472 * and aviro says that's the only reason for hanging onto the
2473 * superblock lock.
2474 */
617ba13b
MC		 2475 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2476 ext4_orphan_cleanup(sb, es);
2477 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
ac27a0ec 2478 if (needs_recovery)
2b2d6d01 2479 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
617ba13b 2480 ext4_mark_recovery_complete(sb, es);
2b2d6d01
TT		 2481 printk(KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2482 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2483 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2484 "writeback");
ac27a0ec
DK		 2485
2486 lock_kernel();
2487 return 0;
2488
617ba13b 2489cantfind_ext4:
ac27a0ec 2490 if (!silent)
617ba13b 2491 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2492 sb->s_id);
2493 goto failed_mount;
2494
2495failed_mount4:
dab291af 2496 jbd2_journal_destroy(sbi->s_journal);
47b4a50b 2497 sbi->s_journal = NULL;
ac27a0ec
DK		 2498failed_mount3:
2499 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2500 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2501 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 2502 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 2503failed_mount2:
2504 for (i = 0; i < db_count; i++)
2505 brelse(sbi->s_group_desc[i]);
2506 kfree(sbi->s_group_desc);
2507failed_mount:
240799cd
TT		 2508 if (sbi->s_proc) {
2509 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
9f6200bb 2510 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 2511 }
ac27a0ec
DK		 2512#ifdef CONFIG_QUOTA
2513 for (i = 0; i < MAXQUOTAS; i++)
2514 kfree(sbi->s_qf_names[i]);
2515#endif
617ba13b 2516 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2517 brelse(bh);
2518out_fail:
2519 sb->s_fs_info = NULL;
2520 kfree(sbi);
2521 lock_kernel();
1d1fe1ee 2522 return ret;
ac27a0ec
DK		 2523}
2524
2525/*
2526 * Setup any per-fs journal parameters now. We'll do this both on
2527 * initial mount, once the journal has been initialised but before we've
2528 * done any recovery; and again on any subsequent remount.
2529 */
617ba13b 2530static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2531{
617ba13b 2532 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2533
2534 if (sbi->s_commit_interval)
2535 journal->j_commit_interval = sbi->s_commit_interval;
617ba13b 2536 /* We could also set up an ext4-specific default for the commit
ac27a0ec
DK		 2537 * interval here, but for now we'll just fall back to the jbd
2538 * default. */
2539
2540 spin_lock(&journal->j_state_lock);
2541 if (test_opt(sb, BARRIER))
dab291af 2542 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2543 else
dab291af 2544 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 2545 if (test_opt(sb, DATA_ERR_ABORT))
2546 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2547 else
2548 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 2549 spin_unlock(&journal->j_state_lock);
2550}
2551
617ba13b 2552static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2553 unsigned int journal_inum)
2554{
2555 struct inode *journal_inode;
2556 journal_t *journal;
2557
2558 /* First, test for the existence of a valid inode on disk. Bad
2559 * things happen if we iget() an unused inode, as the subsequent
2560 * iput() will try to delete it. */
2561
1d1fe1ee
DH		 2562 journal_inode = ext4_iget(sb, journal_inum);
2563 if (IS_ERR(journal_inode)) {
617ba13b 2564 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2565 return NULL;
2566 }
2567 if (!journal_inode->i_nlink) {
2568 make_bad_inode(journal_inode);
2569 iput(journal_inode);
617ba13b 2570 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2571 return NULL;
2572 }
2573
e5f8eab8 2574 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 2575 journal_inode, journal_inode->i_size);
1d1fe1ee 2576 if (!S_ISREG(journal_inode->i_mode)) {
617ba13b 2577 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2578 iput(journal_inode);
2579 return NULL;
2580 }
2581
dab291af 2582 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2583 if (!journal) {
617ba13b 2584 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2585 iput(journal_inode);
2586 return NULL;
2587 }
2588 journal->j_private = sb;
617ba13b 2589 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2590 return journal;
2591}
2592
617ba13b 2593static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2594 dev_t j_dev)
2595{
2b2d6d01 2596 struct buffer_head *bh;
ac27a0ec 2597 journal_t *journal;
617ba13b
MC		 2598 ext4_fsblk_t start;
2599 ext4_fsblk_t len;
ac27a0ec 2600 int hblock, blocksize;
617ba13b 2601 ext4_fsblk_t sb_block;
ac27a0ec 2602 unsigned long offset;
2b2d6d01 2603 struct ext4_super_block *es;
ac27a0ec
DK		 2604 struct block_device *bdev;
2605
617ba13b 2606 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2607 if (bdev == NULL)
2608 return NULL;
2609
2610 if (bd_claim(bdev, sb)) {
2611 printk(KERN_ERR
8c55e204 2612 "EXT4: failed to claim external journal device.\n");
ac27a0ec
DK		 2613 blkdev_put(bdev);
2614 return NULL;
2615 }
2616
2617 blocksize = sb->s_blocksize;
2618 hblock = bdev_hardsect_size(bdev);
2619 if (blocksize < hblock) {
2620 printk(KERN_ERR
617ba13b 2621 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2622 goto out_bdev;
2623 }
2624
617ba13b
MC		 2625 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2626 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2627 set_blocksize(bdev, blocksize);
2628 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2629 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2630 "external journal\n");
2631 goto out_bdev;
2632 }
2633
617ba13b
MC		 2634 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2635 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2636 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2637 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2638 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2639 "bad superblock\n");
2640 brelse(bh);
2641 goto out_bdev;
2642 }
2643
617ba13b
MC		 2644 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2645 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2646 brelse(bh);
2647 goto out_bdev;
2648 }
2649
bd81d8ee 2650 len = ext4_blocks_count(es);
ac27a0ec
DK		 2651 start = sb_block + 1;
2652 brelse(bh); /* we're done with the superblock */
2653
dab291af 2654 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2655 start, len, blocksize);
2656 if (!journal) {
617ba13b 2657 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2658 goto out_bdev;
2659 }
2660 journal->j_private = sb;
2661 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2662 wait_on_buffer(journal->j_sb_buffer);
2663 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2664 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2665 goto out_journal;
2666 }
2667 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2668 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2669 "user (unsupported) - %d\n",
2670 be32_to_cpu(journal->j_superblock->s_nr_users));
2671 goto out_journal;
2672 }
617ba13b
MC		 2673 EXT4_SB(sb)->journal_bdev = bdev;
2674 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2675 return journal;
2676out_journal:
dab291af 2677 jbd2_journal_destroy(journal);
ac27a0ec 2678out_bdev:
617ba13b 2679 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2680 return NULL;
2681}
2682
617ba13b
MC		 2683static int ext4_load_journal(struct super_block *sb,
2684 struct ext4_super_block *es,
ac27a0ec
DK		 2685 unsigned long journal_devnum)
2686{
2687 journal_t *journal;
2688 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2689 dev_t journal_dev;
2690 int err = 0;
2691 int really_read_only;
2692
2693 if (journal_devnum &&
2694 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2695 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2696 "numbers have changed\n");
2697 journal_dev = new_decode_dev(journal_devnum);
2698 } else
2699 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2700
2701 really_read_only = bdev_read_only(sb->s_bdev);
2702
2703 /*
2704 * Are we loading a blank journal or performing recovery after a
2705 * crash? For recovery, we need to check in advance whether we
2706 * can get read-write access to the device.
2707 */
2708
617ba13b 2709 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2710 if (sb->s_flags & MS_RDONLY) {
617ba13b 2711 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2712 "required on readonly filesystem.\n");
2713 if (really_read_only) {
617ba13b 2714 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2715 "unavailable, cannot proceed.\n");
2716 return -EROFS;
2717 }
2b2d6d01
TT		 2718 printk(KERN_INFO "EXT4-fs: write access will "
2719 "be enabled during recovery.\n");
ac27a0ec
DK		 2720 }
2721 }
2722
2723 if (journal_inum && journal_dev) {
617ba13b 2724 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2725 "and inode journals!\n");
2726 return -EINVAL;
2727 }
2728
2729 if (journal_inum) {
617ba13b 2730 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2731 return -EINVAL;
2732 } else {
617ba13b 2733 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2734 return -EINVAL;
2735 }
2736
4776004f
TT		 2737 if (journal->j_flags & JBD2_BARRIER)
2738 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2739 else
2740 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2741
ac27a0ec 2742 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2743 err = jbd2_journal_update_format(journal);
ac27a0ec 2744 if (err) {
617ba13b 2745 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2746 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2747 return err;
2748 }
2749 }
2750
617ba13b 2751 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2752 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2753 if (!err)
dab291af 2754 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2755
2756 if (err) {
617ba13b 2757 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2758 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2759 return err;
2760 }
2761
617ba13b
MC		 2762 EXT4_SB(sb)->s_journal = journal;
2763 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2764
2765 if (journal_devnum &&
2766 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2767 es->s_journal_dev = cpu_to_le32(journal_devnum);
2768 sb->s_dirt = 1;
2769
2770 /* Make sure we flush the recovery flag to disk. */
617ba13b 2771 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2772 }
2773
2774 return 0;
2775}
2776
2b2d6d01
TT		 2777static int ext4_create_journal(struct super_block *sb,
2778 struct ext4_super_block *es,
ac27a0ec
DK		 2779 unsigned int journal_inum)
2780{
2781 journal_t *journal;
6c675bd4 2782 int err;
ac27a0ec
DK		 2783
2784 if (sb->s_flags & MS_RDONLY) {
617ba13b 2785 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
ac27a0ec
DK		 2786 "create journal.\n");
2787 return -EROFS;
2788 }
2789
6c675bd4
BP		 2790 journal = ext4_get_journal(sb, journal_inum);
2791 if (!journal)
ac27a0ec
DK		 2792 return -EINVAL;
2793
617ba13b 2794 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
ac27a0ec
DK		 2795 journal_inum);
2796
6c675bd4
BP		 2797 err = jbd2_journal_create(journal);
2798 if (err) {
617ba13b 2799 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
dab291af 2800 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2801 return -EIO;
2802 }
2803
617ba13b 2804 EXT4_SB(sb)->s_journal = journal;
ac27a0ec 2805
617ba13b
MC		 2806 ext4_update_dynamic_rev(sb);
2807 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2808 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
ac27a0ec
DK		 2809
2810 es->s_journal_inum = cpu_to_le32(journal_inum);
2811 sb->s_dirt = 1;
2812
2813 /* Make sure we flush the recovery flag to disk. */
617ba13b 2814 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2815
2816 return 0;
2817}
2818
2b2d6d01
TT		 2819static void ext4_commit_super(struct super_block *sb,
2820 struct ext4_super_block *es, int sync)
ac27a0ec 2821{
617ba13b 2822 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
ac27a0ec
DK		 2823
2824 if (!sbh)
2825 return;
914258bf
TT		 2826 if (buffer_write_io_error(sbh)) {
2827 /*
2828 * Oh, dear. A previous attempt to write the
2829 * superblock failed. This could happen because the
2830 * USB device was yanked out. Or it could happen to
2831 * be a transient write error and maybe the block will
2832 * be remapped. Nothing we can do but to retry the
2833 * write and hope for the best.
2834 */
2835 printk(KERN_ERR "ext4: previous I/O error to "
2836 "superblock detected for %s.\n", sb->s_id);
2837 clear_buffer_write_io_error(sbh);
2838 set_buffer_uptodate(sbh);
2839 }
ac27a0ec 2840 es->s_wtime = cpu_to_le32(get_seconds());
bd81d8ee 2841 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
617ba13b 2842 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2843 BUFFER_TRACE(sbh, "marking dirty");
2844 mark_buffer_dirty(sbh);
914258bf 2845 if (sync) {
ac27a0ec 2846 sync_dirty_buffer(sbh);
914258bf
TT		 2847 if (buffer_write_io_error(sbh)) {
2848 printk(KERN_ERR "ext4: I/O error while writing "
2849 "superblock for %s.\n", sb->s_id);
2850 clear_buffer_write_io_error(sbh);
2851 set_buffer_uptodate(sbh);
2852 }
2853 }
ac27a0ec
DK		 2854}
2855
2856
2857/*
2858 * Have we just finished recovery? If so, and if we are mounting (or
2859 * remounting) the filesystem readonly, then we will end up with a
2860 * consistent fs on disk. Record that fact.
2861 */
2b2d6d01
TT		 2862static void ext4_mark_recovery_complete(struct super_block *sb,
2863 struct ext4_super_block *es)
ac27a0ec 2864{
617ba13b 2865 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2866
dab291af 2867 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 2868 if (jbd2_journal_flush(journal) < 0)
2869 goto out;
2870
32c37730 2871 lock_super(sb);
617ba13b 2872 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2873 sb->s_flags & MS_RDONLY) {
617ba13b 2874 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2875 sb->s_dirt = 0;
617ba13b 2876 ext4_commit_super(sb, es, 1);
ac27a0ec 2877 }
32c37730 2878 unlock_super(sb);
7ffe1ea8
HK		 2879
2880out:
dab291af 2881 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2882}
2883
2884/*
2885 * If we are mounting (or read-write remounting) a filesystem whose journal
2886 * has recorded an error from a previous lifetime, move that error to the
2887 * main filesystem now.
2888 */
2b2d6d01
TT		 2889static void ext4_clear_journal_err(struct super_block *sb,
2890 struct ext4_super_block *es)
ac27a0ec
DK		 2891{
2892 journal_t *journal;
2893 int j_errno;
2894 const char *errstr;
2895
617ba13b 2896 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2897
2898 /*
2899 * Now check for any error status which may have been recorded in the
617ba13b 2900 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2901 */
2902
dab291af 2903 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2904 if (j_errno) {
2905 char nbuf[16];
2906
617ba13b 2907 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 2908 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 2909 "from previous mount: %s", errstr);
46e665e9 2910 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 2911 "filesystem check.");
2912
617ba13b
MC		 2913 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2914 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2b2d6d01 2915 ext4_commit_super(sb, es, 1);
ac27a0ec 2916
dab291af 2917 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 2918 }
2919}
2920
2921/*
2922 * Force the running and committing transactions to commit,
2923 * and wait on the commit.
2924 */
617ba13b 2925int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 2926{
2927 journal_t *journal;
2928 int ret;
2929
2930 if (sb->s_flags & MS_RDONLY)
2931 return 0;
2932
617ba13b 2933 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2934 sb->s_dirt = 0;
617ba13b 2935 ret = ext4_journal_force_commit(journal);
ac27a0ec
DK		 2936 return ret;
2937}
2938
2939/*
617ba13b 2940 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec
DK		 2941 * have to propagate any other updates to the superblock on disk at this
2942 * point. Just start an async writeback to get the buffers on their way
2943 * to the disk.
2944 *
2945 * This implicitly triggers the writebehind on sync().
2946 */
2947
2b2d6d01 2948static void ext4_write_super(struct super_block *sb)
ac27a0ec
DK		 2949{
2950 if (mutex_trylock(&sb->s_lock) != 0)
2951 BUG();
2952 sb->s_dirt = 0;
2953}
2954
617ba13b 2955static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec
DK		 2956{
2957 tid_t target;
2958
ede86cc4 2959 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
ac27a0ec 2960 sb->s_dirt = 0;
dab291af 2961 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
ac27a0ec 2962 if (wait)
dab291af 2963 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
ac27a0ec
DK		 2964 }
2965 return 0;
2966}
2967
2968/*
2969 * LVM calls this function before a (read-only) snapshot is created. This
2970 * gives us a chance to flush the journal completely and mark the fs clean.
2971 */
617ba13b 2972static void ext4_write_super_lockfs(struct super_block *sb)
ac27a0ec
DK		 2973{
2974 sb->s_dirt = 0;
2975
2976 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 2977 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2978
2979 /* Now we set up the journal barrier. */
dab291af 2980 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 2981
2982 /*
2983 * We don't want to clear needs_recovery flag when we failed
2984 * to flush the journal.
2985 */
2986 if (jbd2_journal_flush(journal) < 0)
2987 return;
ac27a0ec
DK		 2988
2989 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 2990 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2991 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec
DK		 2992 }
2993}
2994
2995/*
2996 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2997 * flag here, even though the filesystem is not technically dirty yet.
2998 */
617ba13b 2999static void ext4_unlockfs(struct super_block *sb)
ac27a0ec
DK		 3000{
3001 if (!(sb->s_flags & MS_RDONLY)) {
3002 lock_super(sb);
3003 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 3004 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3005 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 3006 unlock_super(sb);
dab291af 3007 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 3008 }
3009}
3010
2b2d6d01 3011static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3012{
2b2d6d01 3013 struct ext4_super_block *es;
617ba13b
MC		 3014 struct ext4_sb_info *sbi = EXT4_SB(sb);
3015 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3016 unsigned long old_sb_flags;
617ba13b 3017 struct ext4_mount_options old_opts;
8a266467 3018 ext4_group_t g;
ac27a0ec
DK		 3019 int err;
3020#ifdef CONFIG_QUOTA
3021 int i;
3022#endif
3023
3024 /* Store the original options */
3025 old_sb_flags = sb->s_flags;
3026 old_opts.s_mount_opt = sbi->s_mount_opt;
3027 old_opts.s_resuid = sbi->s_resuid;
3028 old_opts.s_resgid = sbi->s_resgid;
3029 old_opts.s_commit_interval = sbi->s_commit_interval;
3030#ifdef CONFIG_QUOTA
3031 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3032 for (i = 0; i < MAXQUOTAS; i++)
3033 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3034#endif
3035
3036 /*
3037 * Allow the "check" option to be passed as a remount option.
3038 */
3039 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
3040 err = -EINVAL;
3041 goto restore_opts;
3042 }
3043
617ba13b 3044 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
46e665e9 3045 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 3046
3047 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 3048 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3049
3050 es = sbi->s_es;
3051
617ba13b 3052 ext4_init_journal_params(sb, sbi->s_journal);
ac27a0ec
DK		 3053
3054 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3055 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 3056 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 3057 err = -EROFS;
3058 goto restore_opts;
3059 }
3060
3061 if (*flags & MS_RDONLY) {
3062 /*
3063 * First of all, the unconditional stuff we have to do
3064 * to disable replay of the journal when we next remount
3065 */
3066 sb->s_flags |= MS_RDONLY;
3067
3068 /*
3069 * OK, test if we are remounting a valid rw partition
3070 * readonly, and if so set the rdonly flag and then
3071 * mark the partition as valid again.
3072 */
617ba13b
MC		 3073 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3074 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3075 es->s_state = cpu_to_le16(sbi->s_mount_state);
3076
32c37730
JK		 3077 /*
3078 * We have to unlock super so that we can wait for
3079 * transactions.
3080 */
3081 unlock_super(sb);
617ba13b 3082 ext4_mark_recovery_complete(sb, es);
32c37730 3083 lock_super(sb);
ac27a0ec
DK		 3084 } else {
3085 __le32 ret;
617ba13b
MC		 3086 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3087 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3088 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec
DK		 3089 "remount RDWR because of unsupported "
3090 "optional features (%x).\n",
3091 sb->s_id, le32_to_cpu(ret));
3092 err = -EROFS;
3093 goto restore_opts;
3094 }
ead6596b 3095
8a266467
TT		 3096 /*
3097 * Make sure the group descriptor checksums
3098 * are sane. If they aren't, refuse to
3099 * remount r/w.
3100 */
3101 for (g = 0; g < sbi->s_groups_count; g++) {
3102 struct ext4_group_desc *gdp =
3103 ext4_get_group_desc(sb, g, NULL);
3104
3105 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3106 printk(KERN_ERR
3107 "EXT4-fs: ext4_remount: "
3108 "Checksum for group %lu failed (%u!=%u)\n",
3109 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3110 le16_to_cpu(gdp->bg_checksum));
3111 err = -EINVAL;
3112 goto restore_opts;
3113 }
3114 }
3115
ead6596b
ES		 3116 /*
3117 * If we have an unprocessed orphan list hanging
3118 * around from a previously readonly bdev mount,
3119 * require a full umount/remount for now.
3120 */
3121 if (es->s_last_orphan) {
3122 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3123 "remount RDWR because of unprocessed "
3124 "orphan inode list. Please "
3125 "umount/remount instead.\n",
3126 sb->s_id);
3127 err = -EINVAL;
3128 goto restore_opts;
3129 }
3130
ac27a0ec
DK		 3131 /*
3132 * Mounting a RDONLY partition read-write, so reread
3133 * and store the current valid flag. (It may have
3134 * been changed by e2fsck since we originally mounted
3135 * the partition.)
3136 */
617ba13b 3137 ext4_clear_journal_err(sb, es);
ac27a0ec 3138 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3139 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3140 goto restore_opts;
2b2d6d01 3141 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3142 sb->s_flags &= ~MS_RDONLY;
3143 }
3144 }
3145#ifdef CONFIG_QUOTA
3146 /* Release old quota file names */
3147 for (i = 0; i < MAXQUOTAS; i++)
3148 if (old_opts.s_qf_names[i] &&
3149 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3150 kfree(old_opts.s_qf_names[i]);
3151#endif
3152 return 0;
3153restore_opts:
3154 sb->s_flags = old_sb_flags;
3155 sbi->s_mount_opt = old_opts.s_mount_opt;
3156 sbi->s_resuid = old_opts.s_resuid;
3157 sbi->s_resgid = old_opts.s_resgid;
3158 sbi->s_commit_interval = old_opts.s_commit_interval;
3159#ifdef CONFIG_QUOTA
3160 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3161 for (i = 0; i < MAXQUOTAS; i++) {
3162 if (sbi->s_qf_names[i] &&
3163 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3164 kfree(sbi->s_qf_names[i]);
3165 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3166 }
3167#endif
3168 return err;
3169}
3170
2b2d6d01 3171static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3172{
3173 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3174 struct ext4_sb_info *sbi = EXT4_SB(sb);
3175 struct ext4_super_block *es = sbi->s_es;
960cc398 3176 u64 fsid;
ac27a0ec 3177
5e70030d
BP		 3178 if (test_opt(sb, MINIX_DF)) {
3179 sbi->s_overhead_last = 0;
6bc9feff 3180 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 3181 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 3182 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3183 smp_rmb();
3184
3185 /*
5e70030d
BP		 3186 * Compute the overhead (FS structures). This is constant
3187 * for a given filesystem unless the number of block groups
3188 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3189 */
3190
3191 /*
3192 * All of the blocks before first_data_block are
3193 * overhead
3194 */
3195 overhead = le32_to_cpu(es->s_first_data_block);
3196
3197 /*
3198 * Add the overhead attributed to the superblock and
3199 * block group descriptors. If the sparse superblocks
3200 * feature is turned on, then not all groups have this.
3201 */
3202 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3203 overhead += ext4_bg_has_super(sb, i) +
3204 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3205 cond_resched();
3206 }
3207
3208 /*
3209 * Every block group has an inode bitmap, a block
3210 * bitmap, and an inode table.
3211 */
5e70030d
BP		 3212 overhead += ngroups * (2 + sbi->s_itb_per_group);
3213 sbi->s_overhead_last = overhead;
3214 smp_wmb();
6bc9feff 3215 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3216 }
3217
617ba13b 3218 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3219 buf->f_bsize = sb->s_blocksize;
5e70030d 3220 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3221 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3222 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
308ba3ec 3223 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3224 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3225 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3226 buf->f_bavail = 0;
3227 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3228 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3229 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3230 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3231 fsid = le64_to_cpup((void *)es->s_uuid) ^
3232 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3233 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3234 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 3235 return 0;
3236}
3237
3238/* Helper function for writing quotas on sync - we need to start transaction before quota file
3239 * is locked for write. Otherwise the are possible deadlocks:
3240 * Process 1 Process 2
617ba13b 3241 * ext4_create() quota_sync()
dab291af 3242 * jbd2_journal_start() write_dquot()
ac27a0ec 3243 * DQUOT_INIT() down(dqio_mutex)
dab291af 3244 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3245 *
3246 */
3247
3248#ifdef CONFIG_QUOTA
3249
3250static inline struct inode *dquot_to_inode(struct dquot *dquot)
3251{
3252 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3253}
3254
617ba13b 3255static int ext4_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 3256{
3257 handle_t *handle;
3258 int ret, err;
3259
3260 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 3261 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 3262 if (IS_ERR(handle))
3263 return PTR_ERR(handle);
3264 ret = dquot_initialize(inode, type);
617ba13b 3265 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3266 if (!ret)
3267 ret = err;
3268 return ret;
3269}
3270
617ba13b 3271static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 3272{
3273 handle_t *handle;
3274 int ret, err;
3275
3276 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 3277 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2887df13
JK		 3278 if (IS_ERR(handle)) {
3279 /*
3280 * We call dquot_drop() anyway to at least release references
3281 * to quota structures so that umount does not hang.
3282 */
3283 dquot_drop(inode);
ac27a0ec 3284 return PTR_ERR(handle);
2887df13 3285 }
ac27a0ec 3286 ret = dquot_drop(inode);
617ba13b 3287 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3288 if (!ret)
3289 ret = err;
3290 return ret;
3291}
3292
617ba13b 3293static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3294{
3295 int ret, err;
3296 handle_t *handle;
3297 struct inode *inode;
3298
3299 inode = dquot_to_inode(dquot);
617ba13b
MC		 3300 handle = ext4_journal_start(inode,
3301 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3302 if (IS_ERR(handle))
3303 return PTR_ERR(handle);
3304 ret = dquot_commit(dquot);
617ba13b 3305 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3306 if (!ret)
3307 ret = err;
3308 return ret;
3309}
3310
617ba13b 3311static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3312{
3313 int ret, err;
3314 handle_t *handle;
3315
617ba13b
MC		 3316 handle = ext4_journal_start(dquot_to_inode(dquot),
3317 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3318 if (IS_ERR(handle))
3319 return PTR_ERR(handle);
3320 ret = dquot_acquire(dquot);
617ba13b 3321 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3322 if (!ret)
3323 ret = err;
3324 return ret;
3325}
3326
617ba13b 3327static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3328{
3329 int ret, err;
3330 handle_t *handle;
3331
617ba13b
MC		 3332 handle = ext4_journal_start(dquot_to_inode(dquot),
3333 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3334 if (IS_ERR(handle)) {
3335 /* Release dquot anyway to avoid endless cycle in dqput() */
3336 dquot_release(dquot);
ac27a0ec 3337 return PTR_ERR(handle);
9c3013e9 3338 }
ac27a0ec 3339 ret = dquot_release(dquot);
617ba13b 3340 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3341 if (!ret)
3342 ret = err;
3343 return ret;
3344}
3345
617ba13b 3346static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3347{
2c8be6b2 3348 /* Are we journaling quotas? */
617ba13b
MC		 3349 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3350 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3351 dquot_mark_dquot_dirty(dquot);
617ba13b 3352 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3353 } else {
3354 return dquot_mark_dquot_dirty(dquot);
3355 }
3356}
3357
617ba13b 3358static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3359{
3360 int ret, err;
3361 handle_t *handle;
3362
3363 /* Data block + inode block */
617ba13b 3364 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3365 if (IS_ERR(handle))
3366 return PTR_ERR(handle);
3367 ret = dquot_commit_info(sb, type);
617ba13b 3368 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3369 if (!ret)
3370 ret = err;
3371 return ret;
3372}
3373
3374/*
3375 * Turn on quotas during mount time - we need to find
3376 * the quota file and such...
3377 */
617ba13b 3378static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3379{
617ba13b
MC		 3380 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3381 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3382}
3383
3384/*
3385 * Standard function to be called on quota_on
3386 */
617ba13b 3387static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6f28e087 3388 char *path, int remount)
ac27a0ec
DK		 3389{
3390 int err;
3391 struct nameidata nd;
3392
3393 if (!test_opt(sb, QUOTA))
3394 return -EINVAL;
0623543b
JK		 3395 /* When remounting, no checks are needed and in fact, path is NULL */
3396 if (remount)
6f28e087 3397 return vfs_quota_on(sb, type, format_id, path, remount);
0623543b 3398
ac27a0ec
DK		 3399 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3400 if (err)
3401 return err;
0623543b 3402
ac27a0ec 3403 /* Quotafile not on the same filesystem? */
4ac91378 3404 if (nd.path.mnt->mnt_sb != sb) {
1d957f9b 3405 path_put(&nd.path);
ac27a0ec
DK		 3406 return -EXDEV;
3407 }
0623543b
JK		 3408 /* Journaling quota? */
3409 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3410 /* Quotafile not in fs root? */
0623543b
JK		 3411 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3412 printk(KERN_WARNING
3413 "EXT4-fs: Quota file not on filesystem root. "
3414 "Journaled quota will not work.\n");
2b2d6d01 3415 }
0623543b
JK		 3416
3417 /*
3418 * When we journal data on quota file, we have to flush journal to see
3419 * all updates to the file when we bypass pagecache...
3420 */
3421 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3422 /*
3423 * We don't need to lock updates but journal_flush() could
3424 * otherwise be livelocked...
3425 */
3426 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3427 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3428 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8
HK		 3429 if (err) {
3430 path_put(&nd.path);
3431 return err;
3432 }
0623543b
JK		 3433 }
3434
77e69dac 3435 err = vfs_quota_on_path(sb, type, format_id, &nd.path);
1d957f9b 3436 path_put(&nd.path);
77e69dac 3437 return err;
ac27a0ec
DK		 3438}
3439
3440/* Read data from quotafile - avoid pagecache and such because we cannot afford
3441 * acquiring the locks... As quota files are never truncated and quota code
3442 * itself serializes the operations (and noone else should touch the files)
3443 * we don't have to be afraid of races */
617ba13b 3444static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3445 size_t len, loff_t off)
3446{
3447 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3448 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3449 int err = 0;
3450 int offset = off & (sb->s_blocksize - 1);
3451 int tocopy;
3452 size_t toread;
3453 struct buffer_head *bh;
3454 loff_t i_size = i_size_read(inode);
3455
3456 if (off > i_size)
3457 return 0;
3458 if (off+len > i_size)
3459 len = i_size-off;
3460 toread = len;
3461 while (toread > 0) {
3462 tocopy = sb->s_blocksize - offset < toread ?
3463 sb->s_blocksize - offset : toread;
617ba13b 3464 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3465 if (err)
3466 return err;
3467 if (!bh) /* A hole? */
3468 memset(data, 0, tocopy);
3469 else
3470 memcpy(data, bh->b_data+offset, tocopy);
3471 brelse(bh);
3472 offset = 0;
3473 toread -= tocopy;
3474 data += tocopy;
3475 blk++;
3476 }
3477 return len;
3478}
3479
3480/* Write to quotafile (we know the transaction is already started and has
3481 * enough credits) */
617ba13b 3482static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3483 const char *data, size_t len, loff_t off)
3484{
3485 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3486 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3487 int err = 0;
3488 int offset = off & (sb->s_blocksize - 1);
3489 int tocopy;
617ba13b 3490 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3491 size_t towrite = len;
3492 struct buffer_head *bh;
3493 handle_t *handle = journal_current_handle();
3494
9c3013e9 3495 if (!handle) {
e5f8eab8 3496 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
9c3013e9
JK		 3497 " cancelled because transaction is not started.\n",
3498 (unsigned long long)off, (unsigned long long)len);
3499 return -EIO;
3500 }
ac27a0ec
DK		 3501 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3502 while (towrite > 0) {
3503 tocopy = sb->s_blocksize - offset < towrite ?
3504 sb->s_blocksize - offset : towrite;
617ba13b 3505 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3506 if (!bh)
3507 goto out;
3508 if (journal_quota) {
617ba13b 3509 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3510 if (err) {
3511 brelse(bh);
3512 goto out;
3513 }
3514 }
3515 lock_buffer(bh);
3516 memcpy(bh->b_data+offset, data, tocopy);
3517 flush_dcache_page(bh->b_page);
3518 unlock_buffer(bh);
3519 if (journal_quota)
617ba13b 3520 err = ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK		 3521 else {
3522 /* Always do at least ordered writes for quotas */
678aaf48 3523 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3524 mark_buffer_dirty(bh);
3525 }
3526 brelse(bh);
3527 if (err)
3528 goto out;
3529 offset = 0;
3530 towrite -= tocopy;
3531 data += tocopy;
3532 blk++;
3533 }
3534out:
4d04e4fb
JK		 3535 if (len == towrite) {
3536 mutex_unlock(&inode->i_mutex);
ac27a0ec 3537 return err;
4d04e4fb 3538 }
ac27a0ec
DK		 3539 if (inode->i_size < off+len-towrite) {
3540 i_size_write(inode, off+len-towrite);
617ba13b 3541 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3542 }
ac27a0ec 3543 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3544 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3545 mutex_unlock(&inode->i_mutex);
3546 return len - towrite;
3547}
3548
3549#endif
3550
617ba13b 3551static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3552 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3553{
617ba13b 3554 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3555}
3556
5e8814f2
TT		 3557#ifdef CONFIG_PROC_FS
3558static int ext4_ui_proc_show(struct seq_file *m, void *v)
3559{
3560 unsigned int *p = m->private;
3561
3562 seq_printf(m, "%u\n", *p);
3563 return 0;
3564}
3565
3566static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3567{
3568 return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3569}
3570
3571static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3572 size_t cnt, loff_t *ppos)
3573{
3574 unsigned int *p = PDE(file->f_path.dentry->d_inode)->data;
3575 char str[32];
3576 unsigned long value;
3577
3578 if (cnt >= sizeof(str))
3579 return -EINVAL;
3580 if (copy_from_user(str, buf, cnt))
3581 return -EFAULT;
3582 value = simple_strtol(str, NULL, 0);
3583 if (value < 0)
3584 return -ERANGE;
3585 *p = value;
3586 return cnt;
3587}
3588
3589const struct file_operations ext4_ui_proc_fops = {
3590 .owner = THIS_MODULE,
3591 .open = ext4_ui_proc_open,
3592 .read = seq_read,
3593 .llseek = seq_lseek,
3594 .release = single_release,
3595 .write = ext4_ui_proc_write,
3596};
3597#endif
3598
03010a33
TT		 3599static struct file_system_type ext4_fs_type = {
3600 .owner = THIS_MODULE,
3601 .name = "ext4",
3602 .get_sb = ext4_get_sb,
3603 .kill_sb = kill_block_super,
3604 .fs_flags = FS_REQUIRES_DEV,
3605};
3606
3607#ifdef CONFIG_EXT4DEV_COMPAT
3608static int ext4dev_get_sb(struct file_system_type *fs_type,
3609 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3610{
3611 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3612 "to mount using ext4\n");
3613 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3614 "will go away by 2.6.31\n");
3615 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3616}
3617
617ba13b 3618static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3619 .owner = THIS_MODULE,
617ba13b 3620 .name = "ext4dev",
03010a33 3621 .get_sb = ext4dev_get_sb,
ac27a0ec
DK		 3622 .kill_sb = kill_block_super,
3623 .fs_flags = FS_REQUIRES_DEV,
3624};
03010a33
TT		 3625MODULE_ALIAS("ext4dev");
3626#endif
ac27a0ec 3627
617ba13b 3628static int __init init_ext4_fs(void)
ac27a0ec 3629{
c9de560d
AT		 3630 int err;
3631
9f6200bb 3632 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 3633 err = init_ext4_mballoc();
ac27a0ec
DK		 3634 if (err)
3635 return err;
c9de560d
AT		 3636
3637 err = init_ext4_xattr();
3638 if (err)
3639 goto out2;
ac27a0ec
DK		 3640 err = init_inodecache();
3641 if (err)
3642 goto out1;
03010a33 3643 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 3644 if (err)
3645 goto out;
03010a33
TT		 3646#ifdef CONFIG_EXT4DEV_COMPAT
3647 err = register_filesystem(&ext4dev_fs_type);
3648 if (err) {
3649 unregister_filesystem(&ext4_fs_type);
3650 goto out;
3651 }
3652#endif
ac27a0ec
DK		 3653 return 0;
3654out:
3655 destroy_inodecache();
3656out1:
617ba13b 3657 exit_ext4_xattr();
c9de560d
AT		 3658out2:
3659 exit_ext4_mballoc();
ac27a0ec
DK		 3660 return err;
3661}
3662
617ba13b 3663static void __exit exit_ext4_fs(void)
ac27a0ec 3664{
03010a33
TT		 3665 unregister_filesystem(&ext4_fs_type);
3666#ifdef CONFIG_EXT4DEV_COMPAT
617ba13b 3667 unregister_filesystem(&ext4dev_fs_type);
03010a33 3668#endif
ac27a0ec 3669 destroy_inodecache();
617ba13b 3670 exit_ext4_xattr();
c9de560d 3671 exit_ext4_mballoc();
9f6200bb 3672 remove_proc_entry("fs/ext4", NULL);
ac27a0ec
DK		 3673}
3674
3675MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
617ba13b 3676MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
ac27a0ec 3677MODULE_LICENSE("GPL");
617ba13b
MC		 3678module_init(init_ext4_fs)
3679module_exit(exit_ext4_fs)
kernel source for telekom puls (alcatel/mediatek)