projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: Pass line numbers to ext4_error() and friends
[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>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
29#include <linux/smp_lock.h>
30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
9f6200bb 38#include <linux/proc_fs.h>
3197ebdb 39#include <linux/ctype.h>
1330593e 40#include <linux/log2.h>
717d50e4 41#include <linux/crc16.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
3dcf5451
CH		 44#include "ext4.h"
45#include "ext4_jbd2.h"
ac27a0ec
DK		 46#include "xattr.h"
47#include "acl.h"
3661d286 48#include "mballoc.h"
ac27a0ec 49
9bffad1e
TT		 50#define CREATE_TRACE_POINTS
51#include <trace/events/ext4.h>
52
9f6200bb 53struct proc_dir_entry *ext4_proc_root;
3197ebdb 54static struct kset *ext4_kset;
9f6200bb 55
617ba13b 56static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 57 unsigned long journal_devnum);
e2d67052 58static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 59static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
617ba13b 63static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 64static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 65 char nbuf[16]);
2b2d6d01
TT		 66static int ext4_remount(struct super_block *sb, int *flags, char *data);
67static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 68static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 69static void ext4_write_super(struct super_block *sb);
c4be0c1d 70static int ext4_freeze(struct super_block *sb);
ba69f9ab
JK		 71static int ext4_get_sb(struct file_system_type *fs_type, int flags,
72 const char *dev_name, void *data, struct vfsmount *mnt);
ac27a0ec 73
ba69f9ab
JK		 74#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
75static struct file_system_type ext3_fs_type = {
76 .owner = THIS_MODULE,
77 .name = "ext3",
78 .get_sb = ext4_get_sb,
79 .kill_sb = kill_block_super,
80 .fs_flags = FS_REQUIRES_DEV,
81};
82#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
83#else
84#define IS_EXT3_SB(sb) (0)
85#endif
bd81d8ee 86
8fadc143
AR		 87ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
88 struct ext4_group_desc *bg)
bd81d8ee 89{
3a14589c 90 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 91 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 92 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 93}
94
8fadc143
AR		 95ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
96 struct ext4_group_desc *bg)
bd81d8ee 97{
5272f837 98 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 99 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 100 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 101}
102
8fadc143
AR		 103ext4_fsblk_t ext4_inode_table(struct super_block *sb,
104 struct ext4_group_desc *bg)
bd81d8ee 105{
5272f837 106 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 107 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 108 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 109}
110
560671a0
AK		 111__u32 ext4_free_blks_count(struct super_block *sb,
112 struct ext4_group_desc *bg)
113{
114 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
115 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 116 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 117}
118
119__u32 ext4_free_inodes_count(struct super_block *sb,
120 struct ext4_group_desc *bg)
121{
122 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
123 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 124 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 125}
126
127__u32 ext4_used_dirs_count(struct super_block *sb,
128 struct ext4_group_desc *bg)
129{
130 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
131 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 132 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 133}
134
135__u32 ext4_itable_unused_count(struct super_block *sb,
136 struct ext4_group_desc *bg)
137{
138 return le16_to_cpu(bg->bg_itable_unused_lo) |
139 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 140 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 141}
142
8fadc143
AR		 143void ext4_block_bitmap_set(struct super_block *sb,
144 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 145{
3a14589c 146 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 147 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
148 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 149}
150
8fadc143
AR		 151void ext4_inode_bitmap_set(struct super_block *sb,
152 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 153{
5272f837 154 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 155 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
156 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 157}
158
8fadc143
AR		 159void ext4_inode_table_set(struct super_block *sb,
160 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 161{
5272f837 162 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 163 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
164 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 165}
166
560671a0
AK		 167void ext4_free_blks_set(struct super_block *sb,
168 struct ext4_group_desc *bg, __u32 count)
169{
170 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
171 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
172 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
173}
174
175void ext4_free_inodes_set(struct super_block *sb,
176 struct ext4_group_desc *bg, __u32 count)
177{
178 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
179 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
180 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
181}
182
183void ext4_used_dirs_set(struct super_block *sb,
184 struct ext4_group_desc *bg, __u32 count)
185{
186 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
187 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
188 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
189}
190
191void ext4_itable_unused_set(struct super_block *sb,
192 struct ext4_group_desc *bg, __u32 count)
193{
194 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
195 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
196 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
197}
198
d3d1faf6
CW		 199
200/* Just increment the non-pointer handle value */
201static handle_t *ext4_get_nojournal(void)
202{
203 handle_t *handle = current->journal_info;
204 unsigned long ref_cnt = (unsigned long)handle;
205
206 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
207
208 ref_cnt++;
209 handle = (handle_t *)ref_cnt;
210
211 current->journal_info = handle;
212 return handle;
213}
214
215
216/* Decrement the non-pointer handle value */
217static void ext4_put_nojournal(handle_t *handle)
218{
219 unsigned long ref_cnt = (unsigned long)handle;
220
221 BUG_ON(ref_cnt == 0);
222
223 ref_cnt--;
224 handle = (handle_t *)ref_cnt;
225
226 current->journal_info = handle;
227}
228
ac27a0ec 229/*
dab291af 230 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 231 *
232 * The only special thing we need to do here is to make sure that all
233 * journal_end calls result in the superblock being marked dirty, so
234 * that sync() will call the filesystem's write_super callback if
235 * appropriate.
236 */
617ba13b 237handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 238{
239 journal_t *journal;
240
241 if (sb->s_flags & MS_RDONLY)
242 return ERR_PTR(-EROFS);
243
6b0310fb 244 vfs_check_frozen(sb, SB_FREEZE_WRITE);
ac27a0ec
DK		 245 /* Special case here: if the journal has aborted behind our
246 * backs (eg. EIO in the commit thread), then we still need to
247 * take the FS itself readonly cleanly. */
617ba13b 248 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 249 if (journal) {
250 if (is_journal_aborted(journal)) {
c67d859e 251 ext4_abort(sb, "Detected aborted journal");
0390131b
FM		 252 return ERR_PTR(-EROFS);
253 }
254 return jbd2_journal_start(journal, nblocks);
ac27a0ec 255 }
d3d1faf6 256 return ext4_get_nojournal();
ac27a0ec
DK		 257}
258
259/*
260 * The only special thing we need to do here is to make sure that all
dab291af 261 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 262 * that sync() will call the filesystem's write_super callback if
263 * appropriate.
264 */
c398eda0 265int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 266{
267 struct super_block *sb;
268 int err;
269 int rc;
270
0390131b 271 if (!ext4_handle_valid(handle)) {
d3d1faf6 272 ext4_put_nojournal(handle);
0390131b
FM		 273 return 0;
274 }
ac27a0ec
DK		 275 sb = handle->h_transaction->t_journal->j_private;
276 err = handle->h_err;
dab291af 277 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 278
279 if (!err)
280 err = rc;
281 if (err)
c398eda0 282 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 283 return err;
284}
285
90c7201b
TT		 286void ext4_journal_abort_handle(const char *caller, unsigned int line,
287 const char *err_fn, struct buffer_head *bh,
288 handle_t *handle, int err)
ac27a0ec
DK		 289{
290 char nbuf[16];
617ba13b 291 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 292
0390131b
FM		 293 BUG_ON(!ext4_handle_valid(handle));
294
ac27a0ec
DK		 295 if (bh)
296 BUFFER_TRACE(bh, "abort");
297
298 if (!handle->h_err)
299 handle->h_err = err;
300
301 if (is_handle_aborted(handle))
302 return;
303
90c7201b
TT		 304 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
305 caller, line, errstr, err_fn);
ac27a0ec 306
dab291af 307 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 308}
309
310/* Deal with the reporting of failure conditions on a filesystem such as
311 * inconsistencies detected or read IO failures.
312 *
313 * On ext2, we can store the error state of the filesystem in the
617ba13b 314 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 315 * write ordering constraints on the superblock which prevent us from
316 * writing it out straight away; and given that the journal is about to
317 * be aborted, we can't rely on the current, or future, transactions to
318 * write out the superblock safely.
319 *
dab291af 320 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 321 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 322 * that error until we've noted it down and cleared it.
323 */
324
617ba13b 325static void ext4_handle_error(struct super_block *sb)
ac27a0ec 326{
617ba13b 327 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 328
617ba13b
MC		 329 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
330 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 331
332 if (sb->s_flags & MS_RDONLY)
333 return;
334
2b2d6d01 335 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 336 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 337
4ab2f15b 338 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 339 if (journal)
dab291af 340 jbd2_journal_abort(journal, -EIO);
ac27a0ec 341 }
2b2d6d01 342 if (test_opt(sb, ERRORS_RO)) {
b31e1552 343 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 344 sb->s_flags |= MS_RDONLY;
345 }
e2d67052 346 ext4_commit_super(sb, 1);
ac27a0ec 347 if (test_opt(sb, ERRORS_PANIC))
617ba13b 348 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 349 sb->s_id);
350}
351
12062ddd 352void __ext4_error(struct super_block *sb, const char *function,
c398eda0 353 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 354{
355 va_list args;
356
357 va_start(args, fmt);
c398eda0
TT		 358 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: ",
359 sb->s_id, function, line, current->comm);
ac27a0ec
DK		 360 vprintk(fmt, args);
361 printk("\n");
362 va_end(args);
363
617ba13b 364 ext4_handle_error(sb);
ac27a0ec
DK		 365}
366
c398eda0
TT		 367void ext4_error_inode(struct inode *inode, const char *function,
368 unsigned int line, ext4_fsblk_t block,
273df556
FM		 369 const char *fmt, ...)
370{
371 va_list args;
372
373 va_start(args, fmt);
c398eda0
TT		 374 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
375 inode->i_sb->s_id, function, line, inode->i_ino);
376 if (block)
377 printk("block %llu: ", block);
378 printk("comm %s: ", current->comm);
273df556
FM		 379 vprintk(fmt, args);
380 printk("\n");
381 va_end(args);
382
383 ext4_handle_error(inode->i_sb);
384}
385
c398eda0
TT		 386void ext4_error_file(struct file *file, const char *function,
387 unsigned int line, const char *fmt, ...)
273df556
FM		 388{
389 va_list args;
390 struct inode *inode = file->f_dentry->d_inode;
391 char pathname[80], *path;
392
393 va_start(args, fmt);
394 path = d_path(&(file->f_path), pathname, sizeof(pathname));
395 if (!path)
396 path = "(unknown)";
397 printk(KERN_CRIT
c398eda0
TT		 398 "EXT4-fs error (device %s): %s:%d: inode #%lu "
399 "(comm %s path %s): ",
400 inode->i_sb->s_id, function, line, inode->i_ino,
401 current->comm, path);
273df556
FM		 402 vprintk(fmt, args);
403 printk("\n");
404 va_end(args);
405
406 ext4_handle_error(inode->i_sb);
407}
408
2b2d6d01 409static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 410 char nbuf[16])
411{
412 char *errstr = NULL;
413
414 switch (errno) {
415 case -EIO:
416 errstr = "IO failure";
417 break;
418 case -ENOMEM:
419 errstr = "Out of memory";
420 break;
421 case -EROFS:
78f1ddbb
TT		 422 if (!sb || (EXT4_SB(sb)->s_journal &&
423 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 424 errstr = "Journal has aborted";
425 else
426 errstr = "Readonly filesystem";
427 break;
428 default:
429 /* If the caller passed in an extra buffer for unknown
430 * errors, textualise them now. Else we just return
431 * NULL. */
432 if (nbuf) {
433 /* Check for truncated error codes... */
434 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
435 errstr = nbuf;
436 }
437 break;
438 }
439
440 return errstr;
441}
442
617ba13b 443/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 444 * automatically and invokes the appropriate error response. */
445
c398eda0
TT		 446void __ext4_std_error(struct super_block *sb, const char *function,
447 unsigned int line, int errno)
ac27a0ec
DK		 448{
449 char nbuf[16];
450 const char *errstr;
451
452 /* Special case: if the error is EROFS, and we're not already
453 * inside a transaction, then there's really no point in logging
454 * an error. */
455 if (errno == -EROFS && journal_current_handle() == NULL &&
456 (sb->s_flags & MS_RDONLY))
457 return;
458
617ba13b 459 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 460 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
461 sb->s_id, function, line, errstr);
ac27a0ec 462
617ba13b 463 ext4_handle_error(sb);
ac27a0ec
DK		 464}
465
466/*
617ba13b 467 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 468 * abort function may be used to deal with unrecoverable failures such
469 * as journal IO errors or ENOMEM at a critical moment in log management.
470 *
471 * We unconditionally force the filesystem into an ABORT|READONLY state,
472 * unless the error response on the fs has been set to panic in which
473 * case we take the easy way out and panic immediately.
474 */
475
c67d859e 476void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 477 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 478{
479 va_list args;
480
ac27a0ec 481 va_start(args, fmt);
c398eda0
TT		 482 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
483 function, line);
ac27a0ec
DK		 484 vprintk(fmt, args);
485 printk("\n");
486 va_end(args);
487
488 if (test_opt(sb, ERRORS_PANIC))
617ba13b 489 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 490
491 if (sb->s_flags & MS_RDONLY)
492 return;
493
b31e1552 494 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
617ba13b 495 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 496 sb->s_flags |= MS_RDONLY;
4ab2f15b 497 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ef2cabf7
HK		 498 if (EXT4_SB(sb)->s_journal)
499 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 500}
501
b31e1552
ES		 502void ext4_msg (struct super_block * sb, const char *prefix,
503 const char *fmt, ...)
504{
505 va_list args;
506
507 va_start(args, fmt);
508 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
509 vprintk(fmt, args);
510 printk("\n");
511 va_end(args);
512}
513
12062ddd 514void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 515 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 516{
517 va_list args;
518
519 va_start(args, fmt);
c398eda0
TT		 520 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: ",
521 sb->s_id, function, line);
ac27a0ec
DK		 522 vprintk(fmt, args);
523 printk("\n");
524 va_end(args);
525}
526
e29136f8
TT		 527void __ext4_grp_locked_error(const char *function, unsigned int line,
528 struct super_block *sb, ext4_group_t grp,
529 unsigned long ino, ext4_fsblk_t block,
530 const char *fmt, ...)
5d1b1b3f
AK		 531__releases(bitlock)
532__acquires(bitlock)
533{
534 va_list args;
535 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
536
537 va_start(args, fmt);
e29136f8
TT		 538 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
539 sb->s_id, function, line, grp);
540 if (ino)
541 printk("inode %lu: ", ino);
542 if (block)
543 printk("block %llu:", (unsigned long long) block);
5d1b1b3f
AK		 544 vprintk(fmt, args);
545 printk("\n");
546 va_end(args);
547
548 if (test_opt(sb, ERRORS_CONT)) {
549 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
550 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 551 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 552 return;
553 }
554 ext4_unlock_group(sb, grp);
555 ext4_handle_error(sb);
556 /*
557 * We only get here in the ERRORS_RO case; relocking the group
558 * may be dangerous, but nothing bad will happen since the
559 * filesystem will have already been marked read/only and the
560 * journal has been aborted. We return 1 as a hint to callers
561 * who might what to use the return value from
562 * ext4_grp_locked_error() to distinguish beween the
563 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
564 * aggressively from the ext4 function in question, with a
565 * more appropriate error code.
566 */
567 ext4_lock_group(sb, grp);
568 return;
569}
570
617ba13b 571void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 572{
617ba13b 573 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 574
617ba13b 575 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 576 return;
577
12062ddd 578 ext4_warning(sb,
ac27a0ec
DK		 579 "updating to rev %d because of new feature flag, "
580 "running e2fsck is recommended",
617ba13b 581 EXT4_DYNAMIC_REV);
ac27a0ec 582
617ba13b
MC		 583 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
584 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
585 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 586 /* leave es->s_feature_*compat flags alone */
587 /* es->s_uuid will be set by e2fsck if empty */
588
589 /*
590 * The rest of the superblock fields should be zero, and if not it
591 * means they are likely already in use, so leave them alone. We
592 * can leave it up to e2fsck to clean up any inconsistencies there.
593 */
594}
595
596/*
597 * Open the external journal device
598 */
b31e1552 599static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 600{
601 struct block_device *bdev;
602 char b[BDEVNAME_SIZE];
603
604 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
605 if (IS_ERR(bdev))
606 goto fail;
607 return bdev;
608
609fail:
b31e1552 610 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 611 __bdevname(dev, b), PTR_ERR(bdev));
612 return NULL;
613}
614
615/*
616 * Release the journal device
617 */
617ba13b 618static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 619{
620 bd_release(bdev);
9a1c3542 621 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 622}
623
617ba13b 624static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 625{
626 struct block_device *bdev;
627 int ret = -ENODEV;
628
629 bdev = sbi->journal_bdev;
630 if (bdev) {
617ba13b 631 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 632 sbi->journal_bdev = NULL;
633 }
634 return ret;
635}
636
637static inline struct inode *orphan_list_entry(struct list_head *l)
638{
617ba13b 639 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 640}
641
617ba13b 642static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 643{
644 struct list_head *l;
645
b31e1552
ES		 646 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
647 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 648
649 printk(KERN_ERR "sb_info orphan list:\n");
650 list_for_each(l, &sbi->s_orphan) {
651 struct inode *inode = orphan_list_entry(l);
652 printk(KERN_ERR " "
653 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
654 inode->i_sb->s_id, inode->i_ino, inode,
655 inode->i_mode, inode->i_nlink,
656 NEXT_ORPHAN(inode));
657 }
658}
659
2b2d6d01 660static void ext4_put_super(struct super_block *sb)
ac27a0ec 661{
617ba13b
MC		 662 struct ext4_sb_info *sbi = EXT4_SB(sb);
663 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 664 int i, err;
ac27a0ec 665
e0ccfd95
CH		 666 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
667
4c0425ff
MC		 668 flush_workqueue(sbi->dio_unwritten_wq);
669 destroy_workqueue(sbi->dio_unwritten_wq);
670
a9e220f8 671 lock_super(sb);
6cfd0148 672 lock_kernel();
8c85e125 673 if (sb->s_dirt)
ebc1ac16 674 ext4_commit_super(sb, 1);
8c85e125 675
0390131b
FM		 676 if (sbi->s_journal) {
677 err = jbd2_journal_destroy(sbi->s_journal);
678 sbi->s_journal = NULL;
679 if (err < 0)
c67d859e 680 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 681 }
d4edac31
JB		 682
683 ext4_release_system_zone(sb);
684 ext4_mb_release(sb);
685 ext4_ext_release(sb);
686 ext4_xattr_put_super(sb);
687
ac27a0ec 688 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 689 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 690 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 691 ext4_commit_super(sb, 1);
ac27a0ec 692 }
240799cd 693 if (sbi->s_proc) {
9f6200bb 694 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 695 }
3197ebdb 696 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 697
698 for (i = 0; i < sbi->s_gdb_count; i++)
699 brelse(sbi->s_group_desc[i]);
700 kfree(sbi->s_group_desc);
c5ca7c76
TT		 701 if (is_vmalloc_addr(sbi->s_flex_groups))
702 vfree(sbi->s_flex_groups);
703 else
704 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 705 percpu_counter_destroy(&sbi->s_freeblocks_counter);
706 percpu_counter_destroy(&sbi->s_freeinodes_counter);
707 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 708 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 709 brelse(sbi->s_sbh);
710#ifdef CONFIG_QUOTA
711 for (i = 0; i < MAXQUOTAS; i++)
712 kfree(sbi->s_qf_names[i]);
713#endif
714
715 /* Debugging code just in case the in-memory inode orphan list
716 * isn't empty. The on-disk one can be non-empty if we've
717 * detected an error and taken the fs readonly, but the
718 * in-memory list had better be clean by this point. */
719 if (!list_empty(&sbi->s_orphan))
720 dump_orphan_list(sb, sbi);
721 J_ASSERT(list_empty(&sbi->s_orphan));
722
f98393a6 723 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 724 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
725 /*
726 * Invalidate the journal device's buffers. We don't want them
727 * floating about in memory - the physical journal device may
728 * hotswapped, and it breaks the `ro-after' testing code.
729 */
730 sync_blockdev(sbi->journal_bdev);
f98393a6 731 invalidate_bdev(sbi->journal_bdev);
617ba13b 732 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 733 }
734 sb->s_fs_info = NULL;
3197ebdb
TT		 735 /*
736 * Now that we are completely done shutting down the
737 * superblock, we need to actually destroy the kobject.
738 */
739 unlock_kernel();
740 unlock_super(sb);
741 kobject_put(&sbi->s_kobj);
742 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 743 kfree(sbi->s_blockgroup_lock);
ac27a0ec 744 kfree(sbi);
ac27a0ec
DK		 745}
746
e18b890b 747static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 748
749/*
750 * Called inside transaction, so use GFP_NOFS
751 */
617ba13b 752static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 753{
617ba13b 754 struct ext4_inode_info *ei;
ac27a0ec 755
e6b4f8da 756 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 757 if (!ei)
758 return NULL;
0b8e58a1 759
ac27a0ec 760 ei->vfs_inode.i_version = 1;
91246c00 761 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 762 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 763 INIT_LIST_HEAD(&ei->i_prealloc_list);
764 spin_lock_init(&ei->i_prealloc_lock);
0390131b
FM		 765 /*
766 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
767 * therefore it can be null here. Don't check it, just initialize
768 * jinode.
769 */
678aaf48 770 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 771 ei->i_reserved_data_blocks = 0;
772 ei->i_reserved_meta_blocks = 0;
773 ei->i_allocated_meta_blocks = 0;
9d0be502 774 ei->i_da_metadata_calc_len = 0;
d2a17637
MC		 775 ei->i_delalloc_reserved_flag = 0;
776 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 777#ifdef CONFIG_QUOTA
778 ei->i_reserved_quota = 0;
779#endif
c7064ef1 780 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 781 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 782 ei->cur_aio_dio = NULL;
b436b9be
JK		 783 ei->i_sync_tid = 0;
784 ei->i_datasync_tid = 0;
0b8e58a1 785
ac27a0ec
DK		 786 return &ei->vfs_inode;
787}
788
617ba13b 789static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 790{
9f7dd93d 791 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 792 ext4_msg(inode->i_sb, KERN_ERR,
793 "Inode %lu (%p): orphan list check failed!",
794 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 795 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
796 EXT4_I(inode), sizeof(struct ext4_inode_info),
797 true);
798 dump_stack();
799 }
617ba13b 800 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 801}
802
51cc5068 803static void init_once(void *foo)
ac27a0ec 804{
617ba13b 805 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 806
a35afb83 807 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 808#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 809 init_rwsem(&ei->xattr_sem);
ac27a0ec 810#endif
0e855ac8 811 init_rwsem(&ei->i_data_sem);
a35afb83 812 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 813}
814
815static int init_inodecache(void)
816{
617ba13b
MC		 817 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
818 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 819 0, (SLAB_RECLAIM_ACCOUNT|
820 SLAB_MEM_SPREAD),
20c2df83 821 init_once);
617ba13b 822 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 823 return -ENOMEM;
824 return 0;
825}
826
827static void destroy_inodecache(void)
828{
617ba13b 829 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 830}
831
617ba13b 832static void ext4_clear_inode(struct inode *inode)
ac27a0ec 833{
9f754758 834 dquot_drop(inode);
c2ea3fde 835 ext4_discard_preallocations(inode);
0390131b
FM		 836 if (EXT4_JOURNAL(inode))
837 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
678aaf48 838 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 839}
840
2b2d6d01
TT		 841static inline void ext4_show_quota_options(struct seq_file *seq,
842 struct super_block *sb)
ac27a0ec
DK		 843{
844#if defined(CONFIG_QUOTA)
617ba13b 845 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 846
5a20bdfc
JK		 847 if (sbi->s_jquota_fmt) {
848 char *fmtname = "";
849
850 switch (sbi->s_jquota_fmt) {
851 case QFMT_VFS_OLD:
852 fmtname = "vfsold";
853 break;
854 case QFMT_VFS_V0:
855 fmtname = "vfsv0";
856 break;
857 case QFMT_VFS_V1:
858 fmtname = "vfsv1";
859 break;
860 }
861 seq_printf(seq, ",jqfmt=%s", fmtname);
862 }
ac27a0ec
DK		 863
864 if (sbi->s_qf_names[USRQUOTA])
865 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
866
867 if (sbi->s_qf_names[GRPQUOTA])
868 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
869
482a7425 870 if (test_opt(sb, USRQUOTA))
ac27a0ec
DK		 871 seq_puts(seq, ",usrquota");
872
482a7425 873 if (test_opt(sb, GRPQUOTA))
ac27a0ec
DK		 874 seq_puts(seq, ",grpquota");
875#endif
876}
877
d9c9bef1
MS		 878/*
879 * Show an option if
880 * - it's set to a non-default value OR
881 * - if the per-sb default is different from the global default
882 */
617ba13b 883static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 884{
aa22df2c
AK		 885 int def_errors;
886 unsigned long def_mount_opts;
ac27a0ec 887 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 888 struct ext4_sb_info *sbi = EXT4_SB(sb);
889 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 890
891 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 892 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 893
894 if (sbi->s_sb_block != 1)
895 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
896 if (test_opt(sb, MINIX_DF))
897 seq_puts(seq, ",minixdf");
aa22df2c 898 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 899 seq_puts(seq, ",grpid");
900 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
901 seq_puts(seq, ",nogrpid");
902 if (sbi->s_resuid != EXT4_DEF_RESUID ||
903 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
904 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
905 }
906 if (sbi->s_resgid != EXT4_DEF_RESGID ||
907 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
908 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
909 }
bb4f397a 910 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 911 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 912 def_errors == EXT4_ERRORS_CONTINUE) {
913 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 914 }
915 }
aa22df2c 916 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 917 seq_puts(seq, ",errors=continue");
aa22df2c 918 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 919 seq_puts(seq, ",errors=panic");
aa22df2c 920 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 921 seq_puts(seq, ",nouid32");
aa22df2c 922 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 923 seq_puts(seq, ",debug");
924 if (test_opt(sb, OLDALLOC))
925 seq_puts(seq, ",oldalloc");
03010a33 926#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 927 if (test_opt(sb, XATTR_USER) &&
928 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 929 seq_puts(seq, ",user_xattr");
930 if (!test_opt(sb, XATTR_USER) &&
931 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
932 seq_puts(seq, ",nouser_xattr");
933 }
934#endif
03010a33 935#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 936 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 937 seq_puts(seq, ",acl");
938 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
939 seq_puts(seq, ",noacl");
940#endif
30773840 941 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 942 seq_printf(seq, ",commit=%u",
943 (unsigned) (sbi->s_commit_interval / HZ));
944 }
30773840
TT		 945 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
946 seq_printf(seq, ",min_batch_time=%u",
947 (unsigned) sbi->s_min_batch_time);
948 }
949 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
950 seq_printf(seq, ",max_batch_time=%u",
951 (unsigned) sbi->s_min_batch_time);
952 }
953
571640ca
ES		 954 /*
955 * We're changing the default of barrier mount option, so
956 * let's always display its mount state so it's clear what its
957 * status is.
958 */
959 seq_puts(seq, ",barrier=");
960 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 961 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
962 seq_puts(seq, ",journal_async_commit");
39a4bade
JK		 963 else if (test_opt(sb, JOURNAL_CHECKSUM))
964 seq_puts(seq, ",journal_checksum");
25ec56b5
JNC		 965 if (test_opt(sb, I_VERSION))
966 seq_puts(seq, ",i_version");
dd919b98
AK		 967 if (!test_opt(sb, DELALLOC))
968 seq_puts(seq, ",nodelalloc");
969
ac27a0ec 970
cb45bbe4
MS		 971 if (sbi->s_stripe)
972 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 973 /*
974 * journal mode get enabled in different ways
975 * So just print the value even if we didn't specify it
976 */
617ba13b 977 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 978 seq_puts(seq, ",data=journal");
617ba13b 979 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 980 seq_puts(seq, ",data=ordered");
617ba13b 981 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 982 seq_puts(seq, ",data=writeback");
983
240799cd
TT		 984 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
985 seq_printf(seq, ",inode_readahead_blks=%u",
986 sbi->s_inode_readahead_blks);
987
5bf5683a
HK		 988 if (test_opt(sb, DATA_ERR_ABORT))
989 seq_puts(seq, ",data_err=abort");
990
afd4672d 991 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 992 seq_puts(seq, ",noauto_da_alloc");
afd4672d 993
5328e635
ES		 994 if (test_opt(sb, DISCARD))
995 seq_puts(seq, ",discard");
996
e3bb52ae
ES		 997 if (test_opt(sb, NOLOAD))
998 seq_puts(seq, ",norecovery");
999
744692dc
JZ		 1000 if (test_opt(sb, DIOREAD_NOLOCK))
1001 seq_puts(seq, ",dioread_nolock");
1002
617ba13b 1003 ext4_show_quota_options(seq, sb);
0b8e58a1 1004
ac27a0ec
DK		 1005 return 0;
1006}
1007
1b961ac0 1008static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1009 u64 ino, u32 generation)
ac27a0ec 1010{
ac27a0ec 1011 struct inode *inode;
ac27a0ec 1012
617ba13b 1013 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1014 return ERR_PTR(-ESTALE);
617ba13b 1015 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1016 return ERR_PTR(-ESTALE);
1017
1018 /* iget isn't really right if the inode is currently unallocated!!
1019 *
617ba13b 1020 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1021 * deleted, so we should be safe.
1022 *
1023 * Currently we don't know the generation for parent directory, so
1024 * a generation of 0 means "accept any"
1025 */
1d1fe1ee
DH		 1026 inode = ext4_iget(sb, ino);
1027 if (IS_ERR(inode))
1028 return ERR_CAST(inode);
1029 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1030 iput(inode);
1031 return ERR_PTR(-ESTALE);
1032 }
1b961ac0
CH		 1033
1034 return inode;
1035}
1036
1037static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1038 int fh_len, int fh_type)
1b961ac0
CH		 1039{
1040 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1041 ext4_nfs_get_inode);
1042}
1043
1044static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1045 int fh_len, int fh_type)
1b961ac0
CH		 1046{
1047 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1048 ext4_nfs_get_inode);
ac27a0ec
DK		 1049}
1050
c39a7f84
TO		 1051/*
1052 * Try to release metadata pages (indirect blocks, directories) which are
1053 * mapped via the block device. Since these pages could have journal heads
1054 * which would prevent try_to_free_buffers() from freeing them, we must use
1055 * jbd2 layer's try_to_free_buffers() function to release them.
1056 */
0b8e58a1
AD		 1057static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1058 gfp_t wait)
c39a7f84
TO		 1059{
1060 journal_t *journal = EXT4_SB(sb)->s_journal;
1061
1062 WARN_ON(PageChecked(page));
1063 if (!page_has_buffers(page))
1064 return 0;
1065 if (journal)
1066 return jbd2_journal_try_to_free_buffers(journal, page,
1067 wait & ~__GFP_WAIT);
1068 return try_to_free_buffers(page);
1069}
1070
ac27a0ec 1071#ifdef CONFIG_QUOTA
af5bc92d 1072#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1073#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1074
617ba13b
MC		 1075static int ext4_write_dquot(struct dquot *dquot);
1076static int ext4_acquire_dquot(struct dquot *dquot);
1077static int ext4_release_dquot(struct dquot *dquot);
1078static int ext4_mark_dquot_dirty(struct dquot *dquot);
1079static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1080static int ext4_quota_on(struct super_block *sb, int type, int format_id,
307ae18a 1081 char *path);
617ba13b
MC		 1082static int ext4_quota_on_mount(struct super_block *sb, int type);
1083static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1084 size_t len, loff_t off);
617ba13b 1085static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1086 const char *data, size_t len, loff_t off);
1087
61e225dc 1088static const struct dquot_operations ext4_quota_operations = {
a9e7f447 1089#ifdef CONFIG_QUOTA
60e58e0f 1090 .get_reserved_space = ext4_get_reserved_space,
a9e7f447 1091#endif
617ba13b
MC		 1092 .write_dquot = ext4_write_dquot,
1093 .acquire_dquot = ext4_acquire_dquot,
1094 .release_dquot = ext4_release_dquot,
1095 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1096 .write_info = ext4_write_info,
1097 .alloc_dquot = dquot_alloc,
1098 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1099};
1100
0d54b217 1101static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1102 .quota_on = ext4_quota_on,
287a8095
CH		 1103 .quota_off = dquot_quota_off,
1104 .quota_sync = dquot_quota_sync,
1105 .get_info = dquot_get_dqinfo,
1106 .set_info = dquot_set_dqinfo,
1107 .get_dqblk = dquot_get_dqblk,
1108 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1109};
1110#endif
1111
ee9b6d61 1112static const struct super_operations ext4_sops = {
617ba13b
MC		 1113 .alloc_inode = ext4_alloc_inode,
1114 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1115 .write_inode = ext4_write_inode,
1116 .dirty_inode = ext4_dirty_inode,
1117 .delete_inode = ext4_delete_inode,
1118 .put_super = ext4_put_super,
617ba13b 1119 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1120 .freeze_fs = ext4_freeze,
1121 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1122 .statfs = ext4_statfs,
1123 .remount_fs = ext4_remount,
1124 .clear_inode = ext4_clear_inode,
1125 .show_options = ext4_show_options,
ac27a0ec 1126#ifdef CONFIG_QUOTA
617ba13b
MC		 1127 .quota_read = ext4_quota_read,
1128 .quota_write = ext4_quota_write,
ac27a0ec 1129#endif
c39a7f84 1130 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1131};
1132
9ca92389
TT		 1133static const struct super_operations ext4_nojournal_sops = {
1134 .alloc_inode = ext4_alloc_inode,
1135 .destroy_inode = ext4_destroy_inode,
1136 .write_inode = ext4_write_inode,
1137 .dirty_inode = ext4_dirty_inode,
1138 .delete_inode = ext4_delete_inode,
1139 .write_super = ext4_write_super,
1140 .put_super = ext4_put_super,
1141 .statfs = ext4_statfs,
1142 .remount_fs = ext4_remount,
1143 .clear_inode = ext4_clear_inode,
1144 .show_options = ext4_show_options,
1145#ifdef CONFIG_QUOTA
1146 .quota_read = ext4_quota_read,
1147 .quota_write = ext4_quota_write,
1148#endif
1149 .bdev_try_to_free_page = bdev_try_to_free_page,
1150};
1151
39655164 1152static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1153 .fh_to_dentry = ext4_fh_to_dentry,
1154 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1155 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1156};
1157
1158enum {
1159 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1160 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1161 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1162 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1163 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1164 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1165 Opt_journal_update, Opt_journal_dev,
818d276c 1166 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1167 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1168 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1169 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc
JK		 1170 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1171 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1172 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
01436ef2 1173 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
6fd058f7 1174 Opt_block_validity, Opt_noblock_validity,
5328e635 1175 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1176 Opt_dioread_nolock, Opt_dioread_lock,
5328e635 1177 Opt_discard, Opt_nodiscard,
ac27a0ec
DK		 1178};
1179
a447c093 1180static const match_table_t tokens = {
ac27a0ec
DK		 1181 {Opt_bsd_df, "bsddf"},
1182 {Opt_minix_df, "minixdf"},
1183 {Opt_grpid, "grpid"},
1184 {Opt_grpid, "bsdgroups"},
1185 {Opt_nogrpid, "nogrpid"},
1186 {Opt_nogrpid, "sysvgroups"},
1187 {Opt_resgid, "resgid=%u"},
1188 {Opt_resuid, "resuid=%u"},
1189 {Opt_sb, "sb=%u"},
1190 {Opt_err_cont, "errors=continue"},
1191 {Opt_err_panic, "errors=panic"},
1192 {Opt_err_ro, "errors=remount-ro"},
1193 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1194 {Opt_debug, "debug"},
1195 {Opt_oldalloc, "oldalloc"},
1196 {Opt_orlov, "orlov"},
1197 {Opt_user_xattr, "user_xattr"},
1198 {Opt_nouser_xattr, "nouser_xattr"},
1199 {Opt_acl, "acl"},
1200 {Opt_noacl, "noacl"},
ac27a0ec 1201 {Opt_noload, "noload"},
e3bb52ae 1202 {Opt_noload, "norecovery"},
ac27a0ec
DK		 1203 {Opt_nobh, "nobh"},
1204 {Opt_bh, "bh"},
1205 {Opt_commit, "commit=%u"},
30773840
TT		 1206 {Opt_min_batch_time, "min_batch_time=%u"},
1207 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1208 {Opt_journal_update, "journal=update"},
ac27a0ec 1209 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1210 {Opt_journal_checksum, "journal_checksum"},
1211 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1212 {Opt_abort, "abort"},
1213 {Opt_data_journal, "data=journal"},
1214 {Opt_data_ordered, "data=ordered"},
1215 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1216 {Opt_data_err_abort, "data_err=abort"},
1217 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1218 {Opt_offusrjquota, "usrjquota="},
1219 {Opt_usrjquota, "usrjquota=%s"},
1220 {Opt_offgrpjquota, "grpjquota="},
1221 {Opt_grpjquota, "grpjquota=%s"},
1222 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1223 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1224 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1225 {Opt_grpquota, "grpquota"},
1226 {Opt_noquota, "noquota"},
1227 {Opt_quota, "quota"},
1228 {Opt_usrquota, "usrquota"},
1229 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1230 {Opt_barrier, "barrier"},
1231 {Opt_nobarrier, "nobarrier"},
25ec56b5 1232 {Opt_i_version, "i_version"},
c9de560d 1233 {Opt_stripe, "stripe=%u"},
ac27a0ec 1234 {Opt_resize, "resize"},
64769240 1235 {Opt_delalloc, "delalloc"},
dd919b98 1236 {Opt_nodelalloc, "nodelalloc"},
6fd058f7
TT		 1237 {Opt_block_validity, "block_validity"},
1238 {Opt_noblock_validity, "noblock_validity"},
240799cd 1239 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1240 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1241 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1242 {Opt_auto_da_alloc, "auto_da_alloc"},
1243 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1244 {Opt_dioread_nolock, "dioread_nolock"},
1245 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1246 {Opt_discard, "discard"},
1247 {Opt_nodiscard, "nodiscard"},
f3f12faa 1248 {Opt_err, NULL},
ac27a0ec
DK		 1249};
1250
617ba13b 1251static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1252{
617ba13b 1253 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1254 char *options = (char *) *data;
1255
1256 if (!options || strncmp(options, "sb=", 3) != 0)
1257 return 1; /* Default location */
0b8e58a1 1258
ac27a0ec 1259 options += 3;
0b8e58a1 1260 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1261 sb_block = simple_strtoul(options, &options, 0);
1262 if (*options && *options != ',') {
4776004f 1263 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1264 (char *) *data);
1265 return 1;
1266 }
1267 if (*options == ',')
1268 options++;
1269 *data = (void *) options;
0b8e58a1 1270
ac27a0ec
DK		 1271 return sb_block;
1272}
1273
b3881f74 1274#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1275static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1276 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1277
56c50f11
DM		 1278#ifdef CONFIG_QUOTA
1279static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1280{
1281 struct ext4_sb_info *sbi = EXT4_SB(sb);
1282 char *qname;
1283
1284 if (sb_any_quota_loaded(sb) &&
1285 !sbi->s_qf_names[qtype]) {
1286 ext4_msg(sb, KERN_ERR,
1287 "Cannot change journaled "
1288 "quota options when quota turned on");
1289 return 0;
1290 }
1291 qname = match_strdup(args);
1292 if (!qname) {
1293 ext4_msg(sb, KERN_ERR,
1294 "Not enough memory for storing quotafile name");
1295 return 0;
1296 }
1297 if (sbi->s_qf_names[qtype] &&
1298 strcmp(sbi->s_qf_names[qtype], qname)) {
1299 ext4_msg(sb, KERN_ERR,
1300 "%s quota file already specified", QTYPE2NAME(qtype));
1301 kfree(qname);
1302 return 0;
1303 }
1304 sbi->s_qf_names[qtype] = qname;
1305 if (strchr(sbi->s_qf_names[qtype], '/')) {
1306 ext4_msg(sb, KERN_ERR,
1307 "quotafile must be on filesystem root");
1308 kfree(sbi->s_qf_names[qtype]);
1309 sbi->s_qf_names[qtype] = NULL;
1310 return 0;
1311 }
1312 set_opt(sbi->s_mount_opt, QUOTA);
1313 return 1;
1314}
1315
1316static int clear_qf_name(struct super_block *sb, int qtype)
1317{
1318
1319 struct ext4_sb_info *sbi = EXT4_SB(sb);
1320
1321 if (sb_any_quota_loaded(sb) &&
1322 sbi->s_qf_names[qtype]) {
1323 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1324 " when quota turned on");
1325 return 0;
1326 }
1327 /*
1328 * The space will be released later when all options are confirmed
1329 * to be correct
1330 */
1331 sbi->s_qf_names[qtype] = NULL;
1332 return 1;
1333}
1334#endif
1335
2b2d6d01 1336static int parse_options(char *options, struct super_block *sb,
c3191067 1337 unsigned long *journal_devnum,
b3881f74 1338 unsigned int *journal_ioprio,
2b2d6d01 1339 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1340{
617ba13b 1341 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1342 char *p;
ac27a0ec
DK		 1343 substring_t args[MAX_OPT_ARGS];
1344 int data_opt = 0;
1345 int option;
1346#ifdef CONFIG_QUOTA
56c50f11 1347 int qfmt;
ac27a0ec
DK		 1348#endif
1349
1350 if (!options)
1351 return 1;
1352
2b2d6d01 1353 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1354 int token;
1355 if (!*p)
1356 continue;
1357
15121c18
ES		 1358 /*
1359 * Initialize args struct so we know whether arg was
1360 * found; some options take optional arguments.
1361 */
1362 args[0].to = args[0].from = 0;
ac27a0ec
DK		 1363 token = match_token(p, tokens, args);
1364 switch (token) {
1365 case Opt_bsd_df:
437ca0fd 1366 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1367 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1368 break;
1369 case Opt_minix_df:
437ca0fd 1370 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1371 set_opt(sbi->s_mount_opt, MINIX_DF);
437ca0fd 1372
ac27a0ec
DK		 1373 break;
1374 case Opt_grpid:
437ca0fd 1375 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1376 set_opt(sbi->s_mount_opt, GRPID);
437ca0fd 1377
ac27a0ec
DK		 1378 break;
1379 case Opt_nogrpid:
437ca0fd 1380 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1381 clear_opt(sbi->s_mount_opt, GRPID);
437ca0fd 1382
ac27a0ec
DK		 1383 break;
1384 case Opt_resuid:
1385 if (match_int(&args[0], &option))
1386 return 0;
1387 sbi->s_resuid = option;
1388 break;
1389 case Opt_resgid:
1390 if (match_int(&args[0], &option))
1391 return 0;
1392 sbi->s_resgid = option;
1393 break;
1394 case Opt_sb:
1395 /* handled by get_sb_block() instead of here */
1396 /* *sb_block = match_int(&args[0]); */
1397 break;
1398 case Opt_err_panic:
2b2d6d01
TT		 1399 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1400 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1401 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1402 break;
1403 case Opt_err_ro:
2b2d6d01
TT		 1404 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1405 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1406 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1407 break;
1408 case Opt_err_cont:
2b2d6d01
TT		 1409 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1410 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1411 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1412 break;
1413 case Opt_nouid32:
2b2d6d01 1414 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1415 break;
ac27a0ec 1416 case Opt_debug:
2b2d6d01 1417 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1418 break;
1419 case Opt_oldalloc:
2b2d6d01 1420 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1421 break;
1422 case Opt_orlov:
2b2d6d01 1423 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1424 break;
03010a33 1425#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1426 case Opt_user_xattr:
2b2d6d01 1427 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1428 break;
1429 case Opt_nouser_xattr:
2b2d6d01 1430 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1431 break;
1432#else
1433 case Opt_user_xattr:
1434 case Opt_nouser_xattr:
b31e1552 1435 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1436 break;
1437#endif
03010a33 1438#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1439 case Opt_acl:
1440 set_opt(sbi->s_mount_opt, POSIX_ACL);
1441 break;
1442 case Opt_noacl:
1443 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1444 break;
1445#else
1446 case Opt_acl:
1447 case Opt_noacl:
b31e1552 1448 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1449 break;
1450#endif
ac27a0ec
DK		 1451 case Opt_journal_update:
1452 /* @@@ FIXME */
1453 /* Eventually we will want to be able to create
1454 a journal file here. For now, only allow the
1455 user to specify an existing inode to be the
1456 journal file. */
1457 if (is_remount) {
b31e1552
ES		 1458 ext4_msg(sb, KERN_ERR,
1459 "Cannot specify journal on remount");
ac27a0ec
DK		 1460 return 0;
1461 }
2b2d6d01 1462 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1463 break;
ac27a0ec
DK		 1464 case Opt_journal_dev:
1465 if (is_remount) {
b31e1552
ES		 1466 ext4_msg(sb, KERN_ERR,
1467 "Cannot specify journal on remount");
ac27a0ec
DK		 1468 return 0;
1469 }
1470 if (match_int(&args[0], &option))
1471 return 0;
1472 *journal_devnum = option;
1473 break;
818d276c 1474 case Opt_journal_checksum:
d4da6c9c
LT		 1475 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1476 break;
818d276c
GS		 1477 case Opt_journal_async_commit:
1478 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
d4da6c9c 1479 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
818d276c 1480 break;
ac27a0ec 1481 case Opt_noload:
2b2d6d01 1482 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1483 break;
1484 case Opt_commit:
1485 if (match_int(&args[0], &option))
1486 return 0;
1487 if (option < 0)
1488 return 0;
1489 if (option == 0)
cd02ff0b 1490 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1491 sbi->s_commit_interval = HZ * option;
1492 break;
30773840
TT		 1493 case Opt_max_batch_time:
1494 if (match_int(&args[0], &option))
1495 return 0;
1496 if (option < 0)
1497 return 0;
1498 if (option == 0)
1499 option = EXT4_DEF_MAX_BATCH_TIME;
1500 sbi->s_max_batch_time = option;
1501 break;
1502 case Opt_min_batch_time:
1503 if (match_int(&args[0], &option))
1504 return 0;
1505 if (option < 0)
1506 return 0;
1507 sbi->s_min_batch_time = option;
1508 break;
ac27a0ec 1509 case Opt_data_journal:
617ba13b 1510 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1511 goto datacheck;
1512 case Opt_data_ordered:
617ba13b 1513 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1514 goto datacheck;
1515 case Opt_data_writeback:
617ba13b 1516 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1517 datacheck:
1518 if (is_remount) {
482a7425 1519 if (test_opt(sb, DATA_FLAGS) != data_opt) {
b31e1552
ES		 1520 ext4_msg(sb, KERN_ERR,
1521 "Cannot change data mode on remount");
ac27a0ec
DK		 1522 return 0;
1523 }
1524 } else {
482a7425 1525 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
ac27a0ec
DK		 1526 sbi->s_mount_opt |= data_opt;
1527 }
1528 break;
5bf5683a
HK		 1529 case Opt_data_err_abort:
1530 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1531 break;
1532 case Opt_data_err_ignore:
1533 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1534 break;
ac27a0ec
DK		 1535#ifdef CONFIG_QUOTA
1536 case Opt_usrjquota:
56c50f11 1537 if (!set_qf_name(sb, USRQUOTA, &args[0]))
ac27a0ec 1538 return 0;
56c50f11
DM		 1539 break;
1540 case Opt_grpjquota:
1541 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
ac27a0ec 1542 return 0;
ac27a0ec
DK		 1543 break;
1544 case Opt_offusrjquota:
56c50f11
DM		 1545 if (!clear_qf_name(sb, USRQUOTA))
1546 return 0;
1547 break;
ac27a0ec 1548 case Opt_offgrpjquota:
56c50f11 1549 if (!clear_qf_name(sb, GRPQUOTA))
ac27a0ec 1550 return 0;
ac27a0ec 1551 break;
56c50f11 1552
ac27a0ec 1553 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1554 qfmt = QFMT_VFS_OLD;
1555 goto set_qf_format;
ac27a0ec 1556 case Opt_jqfmt_vfsv0:
dfc5d03f 1557 qfmt = QFMT_VFS_V0;
5a20bdfc
JK		 1558 goto set_qf_format;
1559 case Opt_jqfmt_vfsv1:
1560 qfmt = QFMT_VFS_V1;
dfc5d03f 1561set_qf_format:
17bd13b3 1562 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1563 sbi->s_jquota_fmt != qfmt) {
b31e1552 1564 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1565 "journaled quota options when "
b31e1552 1566 "quota turned on");
dfc5d03f
JK		 1567 return 0;
1568 }
1569 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1570 break;
1571 case Opt_quota:
1572 case Opt_usrquota:
1573 set_opt(sbi->s_mount_opt, QUOTA);
1574 set_opt(sbi->s_mount_opt, USRQUOTA);
1575 break;
1576 case Opt_grpquota:
1577 set_opt(sbi->s_mount_opt, QUOTA);
1578 set_opt(sbi->s_mount_opt, GRPQUOTA);
1579 break;
1580 case Opt_noquota:
17bd13b3 1581 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1582 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1583 "options when quota turned on");
ac27a0ec
DK		 1584 return 0;
1585 }
1586 clear_opt(sbi->s_mount_opt, QUOTA);
1587 clear_opt(sbi->s_mount_opt, USRQUOTA);
1588 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1589 break;
1590#else
1591 case Opt_quota:
1592 case Opt_usrquota:
1593 case Opt_grpquota:
b31e1552
ES		 1594 ext4_msg(sb, KERN_ERR,
1595 "quota options not supported");
cd59e7b9 1596 break;
ac27a0ec
DK		 1597 case Opt_usrjquota:
1598 case Opt_grpjquota:
1599 case Opt_offusrjquota:
1600 case Opt_offgrpjquota:
1601 case Opt_jqfmt_vfsold:
1602 case Opt_jqfmt_vfsv0:
5a20bdfc 1603 case Opt_jqfmt_vfsv1:
b31e1552
ES		 1604 ext4_msg(sb, KERN_ERR,
1605 "journaled quota options not supported");
ac27a0ec
DK		 1606 break;
1607 case Opt_noquota:
1608 break;
1609#endif
1610 case Opt_abort:
4ab2f15b 1611 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1612 break;
06705bff
TT		 1613 case Opt_nobarrier:
1614 clear_opt(sbi->s_mount_opt, BARRIER);
1615 break;
ac27a0ec 1616 case Opt_barrier:
15121c18
ES		 1617 if (args[0].from) {
1618 if (match_int(&args[0], &option))
1619 return 0;
1620 } else
1621 option = 1; /* No argument, default to 1 */
ac27a0ec
DK		 1622 if (option)
1623 set_opt(sbi->s_mount_opt, BARRIER);
1624 else
1625 clear_opt(sbi->s_mount_opt, BARRIER);
1626 break;
1627 case Opt_ignore:
1628 break;
1629 case Opt_resize:
1630 if (!is_remount) {
b31e1552
ES		 1631 ext4_msg(sb, KERN_ERR,
1632 "resize option only available "
1633 "for remount");
ac27a0ec
DK		 1634 return 0;
1635 }
1636 if (match_int(&args[0], &option) != 0)
1637 return 0;
1638 *n_blocks_count = option;
1639 break;
1640 case Opt_nobh:
206f7ab4
CH		 1641 ext4_msg(sb, KERN_WARNING,
1642 "Ignoring deprecated nobh option");
ac27a0ec
DK		 1643 break;
1644 case Opt_bh:
206f7ab4
CH		 1645 ext4_msg(sb, KERN_WARNING,
1646 "Ignoring deprecated bh option");
ac27a0ec 1647 break;
25ec56b5
JNC		 1648 case Opt_i_version:
1649 set_opt(sbi->s_mount_opt, I_VERSION);
1650 sb->s_flags |= MS_I_VERSION;
1651 break;
dd919b98
AK		 1652 case Opt_nodelalloc:
1653 clear_opt(sbi->s_mount_opt, DELALLOC);
1654 break;
c9de560d
AT		 1655 case Opt_stripe:
1656 if (match_int(&args[0], &option))
1657 return 0;
1658 if (option < 0)
1659 return 0;
1660 sbi->s_stripe = option;
1661 break;
64769240
AT		 1662 case Opt_delalloc:
1663 set_opt(sbi->s_mount_opt, DELALLOC);
1664 break;
6fd058f7
TT		 1665 case Opt_block_validity:
1666 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1667 break;
1668 case Opt_noblock_validity:
1669 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1670 break;
240799cd
TT		 1671 case Opt_inode_readahead_blks:
1672 if (match_int(&args[0], &option))
1673 return 0;
1674 if (option < 0 || option > (1 << 30))
1675 return 0;
f7c43950 1676 if (!is_power_of_2(option)) {
b31e1552
ES		 1677 ext4_msg(sb, KERN_ERR,
1678 "EXT4-fs: inode_readahead_blks"
1679 " must be a power of 2");
3197ebdb
TT		 1680 return 0;
1681 }
240799cd
TT		 1682 sbi->s_inode_readahead_blks = option;
1683 break;
b3881f74
TT		 1684 case Opt_journal_ioprio:
1685 if (match_int(&args[0], &option))
1686 return 0;
1687 if (option < 0 || option > 7)
1688 break;
1689 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1690 option);
1691 break;
06705bff
TT		 1692 case Opt_noauto_da_alloc:
1693 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1694 break;
afd4672d 1695 case Opt_auto_da_alloc:
15121c18
ES		 1696 if (args[0].from) {
1697 if (match_int(&args[0], &option))
1698 return 0;
1699 } else
1700 option = 1; /* No argument, default to 1 */
afd4672d
TT		 1701 if (option)
1702 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1703 else
1704 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1705 break;
5328e635
ES		 1706 case Opt_discard:
1707 set_opt(sbi->s_mount_opt, DISCARD);
1708 break;
1709 case Opt_nodiscard:
1710 clear_opt(sbi->s_mount_opt, DISCARD);
1711 break;
744692dc
JZ		 1712 case Opt_dioread_nolock:
1713 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1714 break;
1715 case Opt_dioread_lock:
1716 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1717 break;
ac27a0ec 1718 default:
b31e1552
ES		 1719 ext4_msg(sb, KERN_ERR,
1720 "Unrecognized mount option \"%s\" "
1721 "or missing value", p);
ac27a0ec
DK		 1722 return 0;
1723 }
1724 }
1725#ifdef CONFIG_QUOTA
1726 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1727 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
ac27a0ec
DK		 1728 clear_opt(sbi->s_mount_opt, USRQUOTA);
1729
482a7425 1730 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
ac27a0ec
DK		 1731 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1732
56c50f11 1733 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1734 ext4_msg(sb, KERN_ERR, "old and new quota "
1735 "format mixing");
ac27a0ec
DK		 1736 return 0;
1737 }
1738
1739 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1740 ext4_msg(sb, KERN_ERR, "journaled quota format "
1741 "not specified");
ac27a0ec
DK		 1742 return 0;
1743 }
1744 } else {
1745 if (sbi->s_jquota_fmt) {
b31e1552 1746 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1747 "specified with no journaling "
b31e1552 1748 "enabled");
ac27a0ec
DK		 1749 return 0;
1750 }
1751 }
1752#endif
1753 return 1;
1754}
1755
617ba13b 1756static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1757 int read_only)
1758{
617ba13b 1759 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1760 int res = 0;
1761
617ba13b 1762 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1763 ext4_msg(sb, KERN_ERR, "revision level too high, "
1764 "forcing read-only mode");
ac27a0ec
DK		 1765 res = MS_RDONLY;
1766 }
1767 if (read_only)
1768 return res;
617ba13b 1769 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1770 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1771 "running e2fsck is recommended");
617ba13b 1772 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1773 ext4_msg(sb, KERN_WARNING,
1774 "warning: mounting fs with errors, "
1775 "running e2fsck is recommended");
ac27a0ec
DK		 1776 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1777 le16_to_cpu(es->s_mnt_count) >=
1778 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1779 ext4_msg(sb, KERN_WARNING,
1780 "warning: maximal mount count reached, "
1781 "running e2fsck is recommended");
ac27a0ec
DK		 1782 else if (le32_to_cpu(es->s_checkinterval) &&
1783 (le32_to_cpu(es->s_lastcheck) +
1784 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1785 ext4_msg(sb, KERN_WARNING,
1786 "warning: checktime reached, "
1787 "running e2fsck is recommended");
0b8e58a1 1788 if (!sbi->s_journal)
0390131b 1789 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1790 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1791 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1792 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1793 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1794 ext4_update_dynamic_rev(sb);
0390131b
FM		 1795 if (sbi->s_journal)
1796 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1797
e2d67052 1798 ext4_commit_super(sb, 1);
ac27a0ec 1799 if (test_opt(sb, DEBUG))
a9df9a49 1800 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
7f4520cc 1801 "bpg=%lu, ipg=%lu, mo=%04x]\n",
ac27a0ec
DK		 1802 sb->s_blocksize,
1803 sbi->s_groups_count,
617ba13b
MC		 1804 EXT4_BLOCKS_PER_GROUP(sb),
1805 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1806 sbi->s_mount_opt);
1807
ac27a0ec
DK		 1808 return res;
1809}
1810
772cb7c8
JS		 1811static int ext4_fill_flex_info(struct super_block *sb)
1812{
1813 struct ext4_sb_info *sbi = EXT4_SB(sb);
1814 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1815 ext4_group_t flex_group_count;
1816 ext4_group_t flex_group;
1817 int groups_per_flex = 0;
c5ca7c76 1818 size_t size;
772cb7c8
JS		 1819 int i;
1820
503358ae
TT		 1821 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1822 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1823
1824 if (groups_per_flex < 2) {
772cb7c8
JS		 1825 sbi->s_log_groups_per_flex = 0;
1826 return 1;
1827 }
1828
c62a11fd
FB		 1829 /* We allocate both existing and potentially added groups */
1830 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1831 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1832 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76
TT		 1833 size = flex_group_count * sizeof(struct flex_groups);
1834 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1835 if (sbi->s_flex_groups == NULL) {
1836 sbi->s_flex_groups = vmalloc(size);
1837 if (sbi->s_flex_groups)
1838 memset(sbi->s_flex_groups, 0, size);
1839 }
772cb7c8 1840 if (sbi->s_flex_groups == NULL) {
b31e1552
ES		 1841 ext4_msg(sb, KERN_ERR, "not enough memory for "
1842 "%u flex groups", flex_group_count);
772cb7c8
JS		 1843 goto failed;
1844 }
772cb7c8 1845
772cb7c8 1846 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1847 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1848
1849 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1850 atomic_add(ext4_free_inodes_count(sb, gdp),
1851 &sbi->s_flex_groups[flex_group].free_inodes);
1852 atomic_add(ext4_free_blks_count(sb, gdp),
1853 &sbi->s_flex_groups[flex_group].free_blocks);
1854 atomic_add(ext4_used_dirs_count(sb, gdp),
1855 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1856 }
1857
1858 return 1;
1859failed:
1860 return 0;
1861}
1862
717d50e4
AD		 1863__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1864 struct ext4_group_desc *gdp)
1865{
1866 __u16 crc = 0;
1867
1868 if (sbi->s_es->s_feature_ro_compat &
1869 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1870 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1871 __le32 le_group = cpu_to_le32(block_group);
1872
1873 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1874 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1875 crc = crc16(crc, (__u8 *)gdp, offset);
1876 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1877 /* for checksum of struct ext4_group_desc do the rest...*/
1878 if ((sbi->s_es->s_feature_incompat &
1879 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1880 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1881 crc = crc16(crc, (__u8 *)gdp + offset,
1882 le16_to_cpu(sbi->s_es->s_desc_size) -
1883 offset);
1884 }
1885
1886 return cpu_to_le16(crc);
1887}
1888
1889int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1890 struct ext4_group_desc *gdp)
1891{
1892 if ((sbi->s_es->s_feature_ro_compat &
1893 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1894 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1895 return 0;
1896
1897 return 1;
1898}
1899
ac27a0ec 1900/* Called at mount-time, super-block is locked */
197cd65a 1901static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1902{
617ba13b
MC		 1903 struct ext4_sb_info *sbi = EXT4_SB(sb);
1904 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1905 ext4_fsblk_t last_block;
bd81d8ee
LV		 1906 ext4_fsblk_t block_bitmap;
1907 ext4_fsblk_t inode_bitmap;
1908 ext4_fsblk_t inode_table;
ce421581 1909 int flexbg_flag = 0;
fd2d4291 1910 ext4_group_t i;
ac27a0ec 1911
ce421581
JS		 1912 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1913 flexbg_flag = 1;
1914
af5bc92d 1915 ext4_debug("Checking group descriptors");
ac27a0ec 1916
197cd65a
AM		 1917 for (i = 0; i < sbi->s_groups_count; i++) {
1918 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1919
ce421581 1920 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1921 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1922 else
1923 last_block = first_block +
617ba13b 1924 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1925
8fadc143 1926 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1927 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 1928 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1929 "Block bitmap for group %u not in group "
b31e1552 1930 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1931 return 0;
1932 }
8fadc143 1933 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1934 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 1935 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1936 "Inode bitmap for group %u not in group "
b31e1552 1937 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1938 return 0;
1939 }
8fadc143 1940 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1941 if (inode_table < first_block ||
2b2d6d01 1942 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 1943 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1944 "Inode table for group %u not in group "
b31e1552 1945 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1946 return 0;
1947 }
955ce5f5 1948 ext4_lock_group(sb, i);
717d50e4 1949 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 1950 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1951 "Checksum for group %u failed (%u!=%u)",
1952 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1953 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 1954 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 1955 ext4_unlock_group(sb, i);
8a266467 1956 return 0;
7ee1ec4c 1957 }
717d50e4 1958 }
955ce5f5 1959 ext4_unlock_group(sb, i);
ce421581
JS		 1960 if (!flexbg_flag)
1961 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1962 }
1963
bd81d8ee 1964 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
0b8e58a1 1965 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1966 return 1;
1967}
1968
617ba13b 1969/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1970 * the superblock) which were deleted from all directories, but held open by
1971 * a process at the time of a crash. We walk the list and try to delete these
1972 * inodes at recovery time (only with a read-write filesystem).
1973 *
1974 * In order to keep the orphan inode chain consistent during traversal (in
1975 * case of crash during recovery), we link each inode into the superblock
1976 * orphan list_head and handle it the same way as an inode deletion during
1977 * normal operation (which journals the operations for us).
1978 *
1979 * We only do an iget() and an iput() on each inode, which is very safe if we
1980 * accidentally point at an in-use or already deleted inode. The worst that
1981 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1982 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1983 * e2fsck was run on this filesystem, and it must have already done the orphan
1984 * inode cleanup for us, so we can safely abort without any further action.
1985 */
2b2d6d01
TT		 1986static void ext4_orphan_cleanup(struct super_block *sb,
1987 struct ext4_super_block *es)
ac27a0ec
DK		 1988{
1989 unsigned int s_flags = sb->s_flags;
1990 int nr_orphans = 0, nr_truncates = 0;
1991#ifdef CONFIG_QUOTA
1992 int i;
1993#endif
1994 if (!es->s_last_orphan) {
1995 jbd_debug(4, "no orphan inodes to clean up\n");
1996 return;
1997 }
1998
a8f48a95 1999 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2000 ext4_msg(sb, KERN_ERR, "write access "
2001 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2002 return;
2003 }
2004
617ba13b 2005 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2006 if (es->s_last_orphan)
2007 jbd_debug(1, "Errors on filesystem, "
2008 "clearing orphan list.\n");
2009 es->s_last_orphan = 0;
2010 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2011 return;
2012 }
2013
2014 if (s_flags & MS_RDONLY) {
b31e1552 2015 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2016 sb->s_flags &= ~MS_RDONLY;
2017 }
2018#ifdef CONFIG_QUOTA
2019 /* Needed for iput() to work correctly and not trash data */
2020 sb->s_flags |= MS_ACTIVE;
2021 /* Turn on quotas so that they are updated correctly */
2022 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2023 if (EXT4_SB(sb)->s_qf_names[i]) {
2024 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2025 if (ret < 0)
b31e1552
ES		 2026 ext4_msg(sb, KERN_ERR,
2027 "Cannot turn on journaled "
2028 "quota: error %d", ret);
ac27a0ec
DK		 2029 }
2030 }
2031#endif
2032
2033 while (es->s_last_orphan) {
2034 struct inode *inode;
2035
97bd42b9
JB		 2036 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2037 if (IS_ERR(inode)) {
ac27a0ec
DK		 2038 es->s_last_orphan = 0;
2039 break;
2040 }
2041
617ba13b 2042 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2043 dquot_initialize(inode);
ac27a0ec 2044 if (inode->i_nlink) {
b31e1552
ES		 2045 ext4_msg(sb, KERN_DEBUG,
2046 "%s: truncating inode %lu to %lld bytes",
46e665e9 2047 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2048 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2049 inode->i_ino, inode->i_size);
617ba13b 2050 ext4_truncate(inode);
ac27a0ec
DK		 2051 nr_truncates++;
2052 } else {
b31e1552
ES		 2053 ext4_msg(sb, KERN_DEBUG,
2054 "%s: deleting unreferenced inode %lu",
46e665e9 2055 __func__, inode->i_ino);
ac27a0ec
DK		 2056 jbd_debug(2, "deleting unreferenced inode %lu\n",
2057 inode->i_ino);
2058 nr_orphans++;
2059 }
2060 iput(inode); /* The delete magic happens here! */
2061 }
2062
2b2d6d01 2063#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2064
2065 if (nr_orphans)
b31e1552
ES		 2066 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2067 PLURAL(nr_orphans));
ac27a0ec 2068 if (nr_truncates)
b31e1552
ES		 2069 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2070 PLURAL(nr_truncates));
ac27a0ec
DK		 2071#ifdef CONFIG_QUOTA
2072 /* Turn quotas off */
2073 for (i = 0; i < MAXQUOTAS; i++) {
2074 if (sb_dqopt(sb)->files[i])
287a8095 2075 dquot_quota_off(sb, i);
ac27a0ec
DK		 2076 }
2077#endif
2078 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2079}
0b8e58a1 2080
cd2291a4
ES		 2081/*
2082 * Maximal extent format file size.
2083 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2084 * extent format containers, within a sector_t, and within i_blocks
2085 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2086 * so that won't be a limiting factor.
2087 *
2088 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2089 */
f287a1a5 2090static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2091{
2092 loff_t res;
2093 loff_t upper_limit = MAX_LFS_FILESIZE;
2094
2095 /* small i_blocks in vfs inode? */
f287a1a5 2096 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2097 /*
90c699a9 2098 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2099 * i_block represent total blocks in 512 bytes
2100 * 32 == size of vfs inode i_blocks * 8
2101 */
2102 upper_limit = (1LL << 32) - 1;
2103
2104 /* total blocks in file system block size */
2105 upper_limit >>= (blkbits - 9);
2106 upper_limit <<= blkbits;
2107 }
2108
2109 /* 32-bit extent-start container, ee_block */
2110 res = 1LL << 32;
2111 res <<= blkbits;
2112 res -= 1;
2113
2114 /* Sanity check against vm- & vfs- imposed limits */
2115 if (res > upper_limit)
2116 res = upper_limit;
2117
2118 return res;
2119}
ac27a0ec 2120
ac27a0ec 2121/*
cd2291a4 2122 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2123 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2124 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2125 */
f287a1a5 2126static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2127{
617ba13b 2128 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2129 int meta_blocks;
2130 loff_t upper_limit;
0b8e58a1
AD		 2131 /* This is calculated to be the largest file size for a dense, block
2132 * mapped file such that the file's total number of 512-byte sectors,
2133 * including data and all indirect blocks, does not exceed (2^48 - 1).
2134 *
2135 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2136 * number of 512-byte sectors of the file.
0fc1b451
AK		 2137 */
2138
f287a1a5 2139 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2140 /*
90c699a9 2141 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2142 * the inode i_block field represents total file blocks in
2143 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2144 */
2145 upper_limit = (1LL << 32) - 1;
2146
2147 /* total blocks in file system block size */
2148 upper_limit >>= (bits - 9);
2149
2150 } else {
8180a562
AK		 2151 /*
2152 * We use 48 bit ext4_inode i_blocks
2153 * With EXT4_HUGE_FILE_FL set the i_blocks
2154 * represent total number of blocks in
2155 * file system block size
2156 */
0fc1b451
AK		 2157 upper_limit = (1LL << 48) - 1;
2158
0fc1b451
AK		 2159 }
2160
2161 /* indirect blocks */
2162 meta_blocks = 1;
2163 /* double indirect blocks */
2164 meta_blocks += 1 + (1LL << (bits-2));
2165 /* tripple indirect blocks */
2166 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2167
2168 upper_limit -= meta_blocks;
2169 upper_limit <<= bits;
ac27a0ec
DK		 2170
2171 res += 1LL << (bits-2);
2172 res += 1LL << (2*(bits-2));
2173 res += 1LL << (3*(bits-2));
2174 res <<= bits;
2175 if (res > upper_limit)
2176 res = upper_limit;
0fc1b451
AK		 2177
2178 if (res > MAX_LFS_FILESIZE)
2179 res = MAX_LFS_FILESIZE;
2180
ac27a0ec
DK		 2181 return res;
2182}
2183
617ba13b 2184static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2185 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2186{
617ba13b 2187 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2188 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2189 int has_super = 0;
2190
2191 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2192
617ba13b 2193 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2194 nr < first_meta_bg)
70bbb3e0 2195 return logical_sb_block + nr + 1;
ac27a0ec 2196 bg = sbi->s_desc_per_block * nr;
617ba13b 2197 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2198 has_super = 1;
0b8e58a1 2199
617ba13b 2200 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2201}
2202
c9de560d
AT		 2203/**
2204 * ext4_get_stripe_size: Get the stripe size.
2205 * @sbi: In memory super block info
2206 *
2207 * If we have specified it via mount option, then
2208 * use the mount option value. If the value specified at mount time is
2209 * greater than the blocks per group use the super block value.
2210 * If the super block value is greater than blocks per group return 0.
2211 * Allocator needs it be less than blocks per group.
2212 *
2213 */
2214static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2215{
2216 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2217 unsigned long stripe_width =
2218 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2219
2220 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2221 return sbi->s_stripe;
2222
2223 if (stripe_width <= sbi->s_blocks_per_group)
2224 return stripe_width;
2225
2226 if (stride <= sbi->s_blocks_per_group)
2227 return stride;
2228
2229 return 0;
2230}
ac27a0ec 2231
3197ebdb
TT		 2232/* sysfs supprt */
2233
2234struct ext4_attr {
2235 struct attribute attr;
2236 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2237 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2238 const char *, size_t);
2239 int offset;
2240};
2241
2242static int parse_strtoul(const char *buf,
2243 unsigned long max, unsigned long *value)
2244{
2245 char *endp;
2246
e7d2860b
AGR		 2247 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2248 endp = skip_spaces(endp);
3197ebdb
TT		 2249 if (*endp || *value > max)
2250 return -EINVAL;
2251
2252 return 0;
2253}
2254
2255static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2256 struct ext4_sb_info *sbi,
2257 char *buf)
2258{
2259 return snprintf(buf, PAGE_SIZE, "%llu\n",
2260 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2261}
2262
2263static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2264 struct ext4_sb_info *sbi, char *buf)
2265{
2266 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2267
2268 return snprintf(buf, PAGE_SIZE, "%lu\n",
2269 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2270 sbi->s_sectors_written_start) >> 1);
2271}
2272
2273static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2274 struct ext4_sb_info *sbi, char *buf)
2275{
2276 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2277
2278 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2279 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2280 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2281 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2282}
2283
2284static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2285 struct ext4_sb_info *sbi,
2286 const char *buf, size_t count)
2287{
2288 unsigned long t;
2289
2290 if (parse_strtoul(buf, 0x40000000, &t))
2291 return -EINVAL;
2292
f7c43950 2293 if (!is_power_of_2(t))
3197ebdb
TT		 2294 return -EINVAL;
2295
2296 sbi->s_inode_readahead_blks = t;
2297 return count;
2298}
2299
2300static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2301 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2302{
2303 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2304
2305 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2306}
2307
2308static ssize_t sbi_ui_store(struct ext4_attr *a,
2309 struct ext4_sb_info *sbi,
2310 const char *buf, size_t count)
2311{
2312 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2313 unsigned long t;
2314
2315 if (parse_strtoul(buf, 0xffffffff, &t))
2316 return -EINVAL;
2317 *ui = t;
2318 return count;
2319}
2320
2321#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2322static struct ext4_attr ext4_attr_##_name = { \
2323 .attr = {.name = __stringify(_name), .mode = _mode }, \
2324 .show = _show, \
2325 .store = _store, \
2326 .offset = offsetof(struct ext4_sb_info, _elname), \
2327}
2328#define EXT4_ATTR(name, mode, show, store) \
2329static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2330
2331#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2332#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2333#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2334 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2335#define ATTR_LIST(name) &ext4_attr_##name.attr
2336
2337EXT4_RO_ATTR(delayed_allocation_blocks);
2338EXT4_RO_ATTR(session_write_kbytes);
2339EXT4_RO_ATTR(lifetime_write_kbytes);
2340EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2341 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2342EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2343EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2344EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2345EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2346EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2347EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2348EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2349EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2350
2351static struct attribute *ext4_attrs[] = {
2352 ATTR_LIST(delayed_allocation_blocks),
2353 ATTR_LIST(session_write_kbytes),
2354 ATTR_LIST(lifetime_write_kbytes),
2355 ATTR_LIST(inode_readahead_blks),
11013911 2356 ATTR_LIST(inode_goal),
3197ebdb
TT		 2357 ATTR_LIST(mb_stats),
2358 ATTR_LIST(mb_max_to_scan),
2359 ATTR_LIST(mb_min_to_scan),
2360 ATTR_LIST(mb_order2_req),
2361 ATTR_LIST(mb_stream_req),
2362 ATTR_LIST(mb_group_prealloc),
55138e0b 2363 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2364 NULL,
2365};
2366
2367static ssize_t ext4_attr_show(struct kobject *kobj,
2368 struct attribute *attr, char *buf)
2369{
2370 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2371 s_kobj);
2372 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2373
2374 return a->show ? a->show(a, sbi, buf) : 0;
2375}
2376
2377static ssize_t ext4_attr_store(struct kobject *kobj,
2378 struct attribute *attr,
2379 const char *buf, size_t len)
2380{
2381 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2382 s_kobj);
2383 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2384
2385 return a->store ? a->store(a, sbi, buf, len) : 0;
2386}
2387
2388static void ext4_sb_release(struct kobject *kobj)
2389{
2390 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2391 s_kobj);
2392 complete(&sbi->s_kobj_unregister);
2393}
2394
2395
52cf25d0 2396static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2397 .show = ext4_attr_show,
2398 .store = ext4_attr_store,
2399};
2400
2401static struct kobj_type ext4_ktype = {
2402 .default_attrs = ext4_attrs,
2403 .sysfs_ops = &ext4_attr_ops,
2404 .release = ext4_sb_release,
2405};
2406
a13fb1a4
ES		 2407/*
2408 * Check whether this filesystem can be mounted based on
2409 * the features present and the RDONLY/RDWR mount requested.
2410 * Returns 1 if this filesystem can be mounted as requested,
2411 * 0 if it cannot be.
2412 */
2413static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2414{
2415 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2416 ext4_msg(sb, KERN_ERR,
2417 "Couldn't mount because of "
2418 "unsupported optional features (%x)",
2419 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2420 ~EXT4_FEATURE_INCOMPAT_SUPP));
2421 return 0;
2422 }
2423
2424 if (readonly)
2425 return 1;
2426
2427 /* Check that feature set is OK for a read-write mount */
2428 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2429 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2430 "unsupported optional features (%x)",
2431 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2432 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2433 return 0;
2434 }
2435 /*
2436 * Large file size enabled file system can only be mounted
2437 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2438 */
2439 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2440 if (sizeof(blkcnt_t) < sizeof(u64)) {
2441 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2442 "cannot be mounted RDWR without "
2443 "CONFIG_LBDAF");
2444 return 0;
2445 }
2446 }
2447 return 1;
2448}
2449
2b2d6d01 2450static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 2451 __releases(kernel_lock)
2452 __acquires(kernel_lock)
ac27a0ec 2453{
d4c402d9 2454 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 2455 struct buffer_head *bh;
617ba13b
MC		 2456 struct ext4_super_block *es = NULL;
2457 struct ext4_sb_info *sbi;
2458 ext4_fsblk_t block;
2459 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2460 ext4_fsblk_t logical_sb_block;
ac27a0ec 2461 unsigned long offset = 0;
ac27a0ec
DK		 2462 unsigned long journal_devnum = 0;
2463 unsigned long def_mount_opts;
2464 struct inode *root;
9f6200bb 2465 char *cp;
0390131b 2466 const char *descr;
1d1fe1ee 2467 int ret = -EINVAL;
ac27a0ec 2468 int blocksize;
4ec11028
TT		 2469 unsigned int db_count;
2470 unsigned int i;
f287a1a5 2471 int needs_recovery, has_huge_files;
bd81d8ee 2472 __u64 blocks_count;
833f4077 2473 int err;
b3881f74 2474 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 2475
2476 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2477 if (!sbi)
2478 return -ENOMEM;
705895b6
PE		 2479
2480 sbi->s_blockgroup_lock =
2481 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2482 if (!sbi->s_blockgroup_lock) {
2483 kfree(sbi);
2484 return -ENOMEM;
2485 }
ac27a0ec
DK		 2486 sb->s_fs_info = sbi;
2487 sbi->s_mount_opt = 0;
617ba13b
MC		 2488 sbi->s_resuid = EXT4_DEF_RESUID;
2489 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2490 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2491 sbi->s_sb_block = sb_block;
afc32f7e
TT		 2492 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2493 sectors[1]);
ac27a0ec
DK		 2494
2495 unlock_kernel();
2496
9f6200bb
TT		 2497 /* Cleanup superblock name */
2498 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2499 *cp = '!';
2500
617ba13b 2501 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2502 if (!blocksize) {
b31e1552 2503 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 2504 goto out_fail;
2505 }
2506
2507 /*
617ba13b 2508 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 2509 * block sizes. We need to calculate the offset from buffer start.
2510 */
617ba13b 2511 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 2512 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2513 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 2514 } else {
70bbb3e0 2515 logical_sb_block = sb_block;
ac27a0ec
DK		 2516 }
2517
70bbb3e0 2518 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 2519 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 2520 goto out_fail;
2521 }
2522 /*
2523 * Note: s_es must be initialized as soon as possible because
617ba13b 2524 * some ext4 macro-instructions depend on its value
ac27a0ec 2525 */
617ba13b 2526 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 2527 sbi->s_es = es;
2528 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 2529 if (sb->s_magic != EXT4_SUPER_MAGIC)
2530 goto cantfind_ext4;
afc32f7e 2531 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 2532
2533 /* Set defaults before we parse the mount options */
2534 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 2535 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 2536 set_opt(sbi->s_mount_opt, DEBUG);
437ca0fd
DM		 2537 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2538 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2539 "2.6.38");
ac27a0ec 2540 set_opt(sbi->s_mount_opt, GRPID);
437ca0fd 2541 }
617ba13b 2542 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 2543 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 2544#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 2545 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 2546 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 2547#endif
03010a33 2548#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2549 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2550 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2551#endif
617ba13b 2552 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
482a7425 2553 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
617ba13b 2554 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
482a7425 2555 set_opt(sbi->s_mount_opt, ORDERED_DATA);
617ba13b 2556 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
482a7425 2557 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
617ba13b
MC		 2558
2559 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2560 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2561 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2562 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2563 else
2564 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2565
2566 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2567 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 2568 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2569 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2570 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 2571
571640ca 2572 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2573
dd919b98
AK		 2574 /*
2575 * enable delayed allocation by default
2576 * Use -o nodelalloc to turn it off
2577 */
ba69f9ab
JK		 2578 if (!IS_EXT3_SB(sb))
2579 set_opt(sbi->s_mount_opt, DELALLOC);
dd919b98 2580
b3881f74
TT		 2581 if (!parse_options((char *) data, sb, &journal_devnum,
2582 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 2583 goto failed_mount;
2584
2585 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 2586 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2587
617ba13b
MC		 2588 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2589 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2590 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2591 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 2592 ext4_msg(sb, KERN_WARNING,
2593 "feature flags set on rev 0 fs, "
2594 "running e2fsck is recommended");
469108ff 2595
ac27a0ec
DK		 2596 /*
2597 * Check feature flags regardless of the revision level, since we
2598 * previously didn't change the revision level when setting the flags,
2599 * so there is a chance incompat flags are set on a rev 0 filesystem.
2600 */
a13fb1a4 2601 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 2602 goto failed_mount;
a13fb1a4 2603
ac27a0ec
DK		 2604 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2605
617ba13b
MC		 2606 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2607 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 2608 ext4_msg(sb, KERN_ERR,
2609 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 2610 goto failed_mount;
2611 }
2612
ac27a0ec 2613 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2614 /* Validate the filesystem blocksize */
2615 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 2616 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 2617 blocksize);
ac27a0ec
DK		 2618 goto failed_mount;
2619 }
2620
2b2d6d01 2621 brelse(bh);
70bbb3e0
AM		 2622 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2623 offset = do_div(logical_sb_block, blocksize);
2624 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 2625 if (!bh) {
b31e1552
ES		 2626 ext4_msg(sb, KERN_ERR,
2627 "Can't read superblock on 2nd try");
ac27a0ec
DK		 2628 goto failed_mount;
2629 }
617ba13b 2630 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2631 sbi->s_es = es;
617ba13b 2632 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 2633 ext4_msg(sb, KERN_ERR,
2634 "Magic mismatch, very weird!");
ac27a0ec
DK		 2635 goto failed_mount;
2636 }
2637 }
2638
a13fb1a4
ES		 2639 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2640 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 2641 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2642 has_huge_files);
2643 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2644
617ba13b
MC		 2645 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2646 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2647 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2648 } else {
2649 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2650 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2651 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2652 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2653 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 2654 ext4_msg(sb, KERN_ERR,
2655 "unsupported inode size: %d",
2b2d6d01 2656 sbi->s_inode_size);
ac27a0ec
DK		 2657 goto failed_mount;
2658 }
ef7f3835
KS		 2659 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2660 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2661 }
0b8e58a1 2662
0d1ee42f
AR		 2663 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2664 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2665 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2666 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2667 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 2668 ext4_msg(sb, KERN_ERR,
2669 "unsupported descriptor size %lu",
0d1ee42f
AR		 2670 sbi->s_desc_size);
2671 goto failed_mount;
2672 }
2673 } else
2674 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 2675
ac27a0ec 2676 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2677 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2678 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 2679 goto cantfind_ext4;
0b8e58a1 2680
617ba13b 2681 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2682 if (sbi->s_inodes_per_block == 0)
617ba13b 2683 goto cantfind_ext4;
ac27a0ec
DK		 2684 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2685 sbi->s_inodes_per_block;
0d1ee42f 2686 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2687 sbi->s_sbh = bh;
2688 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2689 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2690 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 2691
2b2d6d01 2692 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2693 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2694 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2695 i = le32_to_cpu(es->s_flags);
2696 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2697 sbi->s_hash_unsigned = 3;
2698 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2699#ifdef __CHAR_UNSIGNED__
2700 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2701 sbi->s_hash_unsigned = 3;
2702#else
2703 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2704#endif
2705 sb->s_dirt = 1;
2706 }
ac27a0ec
DK		 2707
2708 if (sbi->s_blocks_per_group > blocksize * 8) {
b31e1552
ES		 2709 ext4_msg(sb, KERN_ERR,
2710 "#blocks per group too big: %lu",
2b2d6d01 2711 sbi->s_blocks_per_group);
ac27a0ec
DK		 2712 goto failed_mount;
2713 }
ac27a0ec 2714 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 2715 ext4_msg(sb, KERN_ERR,
2716 "#inodes per group too big: %lu",
2b2d6d01 2717 sbi->s_inodes_per_group);
ac27a0ec
DK		 2718 goto failed_mount;
2719 }
2720
bf43d84b
ES		 2721 /*
2722 * Test whether we have more sectors than will fit in sector_t,
2723 * and whether the max offset is addressable by the page cache.
2724 */
2725 if ((ext4_blocks_count(es) >
2726 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2727 (ext4_blocks_count(es) >
2728 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
b31e1552 2729 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 2730 " too large to mount safely on this system");
ac27a0ec 2731 if (sizeof(sector_t) < 8)
90c699a9 2732 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
bf43d84b 2733 ret = -EFBIG;
ac27a0ec
DK		 2734 goto failed_mount;
2735 }
2736
617ba13b
MC		 2737 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2738 goto cantfind_ext4;
e7c95593 2739
0f2ddca6
FTN		 2740 /* check blocks count against device size */
2741 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2742 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 2743 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2744 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 2745 ext4_blocks_count(es), blocks_count);
2746 goto failed_mount;
2747 }
2748
0b8e58a1
AD		 2749 /*
2750 * It makes no sense for the first data block to be beyond the end
2751 * of the filesystem.
2752 */
2753 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 2754 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2755 "block %u is beyond end of filesystem (%llu)",
2756 le32_to_cpu(es->s_first_data_block),
2757 ext4_blocks_count(es));
e7c95593
ES		 2758 goto failed_mount;
2759 }
bd81d8ee
LV		 2760 blocks_count = (ext4_blocks_count(es) -
2761 le32_to_cpu(es->s_first_data_block) +
2762 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2763 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 2764 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 2765 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 2766 "(block count %llu, first data block %u, "
b31e1552 2767 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 2768 ext4_blocks_count(es),
2769 le32_to_cpu(es->s_first_data_block),
2770 EXT4_BLOCKS_PER_GROUP(sb));
2771 goto failed_mount;
2772 }
bd81d8ee 2773 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 2774 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2775 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 2776 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2777 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2778 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2779 GFP_KERNEL);
2780 if (sbi->s_group_desc == NULL) {
b31e1552 2781 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 2782 goto failed_mount;
2783 }
2784
3244fcb1 2785#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2786 if (ext4_proc_root)
2787 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 2788#endif
240799cd 2789
705895b6 2790 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2791
2792 for (i = 0; i < db_count; i++) {
70bbb3e0 2793 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2794 sbi->s_group_desc[i] = sb_bread(sb, block);
2795 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 2796 ext4_msg(sb, KERN_ERR,
2797 "can't read group descriptor %d", i);
ac27a0ec
DK		 2798 db_count = i;
2799 goto failed_mount2;
2800 }
2801 }
2b2d6d01 2802 if (!ext4_check_descriptors(sb)) {
b31e1552 2803 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 2804 goto failed_mount2;
2805 }
772cb7c8
JS		 2806 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2807 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 2808 ext4_msg(sb, KERN_ERR,
2809 "unable to initialize "
2810 "flex_bg meta info!");
772cb7c8
JS		 2811 goto failed_mount2;
2812 }
2813
ac27a0ec
DK		 2814 sbi->s_gdb_count = db_count;
2815 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2816 spin_lock_init(&sbi->s_next_gen_lock);
2817
c9de560d 2818 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 2819 sbi->s_max_writeback_mb_bump = 128;
c9de560d 2820
ac27a0ec
DK		 2821 /*
2822 * set up enough so that it can read an inode
2823 */
9ca92389
TT		 2824 if (!test_opt(sb, NOLOAD) &&
2825 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2826 sb->s_op = &ext4_sops;
2827 else
2828 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 2829 sb->s_export_op = &ext4_export_ops;
2830 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2831#ifdef CONFIG_QUOTA
617ba13b
MC		 2832 sb->s_qcop = &ext4_qctl_operations;
2833 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2834#endif
2835 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 2836 mutex_init(&sbi->s_orphan_lock);
32ed5058 2837 mutex_init(&sbi->s_resize_lock);
ac27a0ec
DK		 2838
2839 sb->s_root = NULL;
2840
2841 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2842 EXT4_HAS_INCOMPAT_FEATURE(sb,
2843 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2844
2845 /*
2846 * The first inode we look at is the journal inode. Don't try
2847 * root first: it may be modified in the journal!
2848 */
2849 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2850 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2851 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2852 goto failed_mount3;
0390131b
FM		 2853 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2854 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 2855 ext4_msg(sb, KERN_ERR, "required journal recovery "
2856 "suppressed and not mounted read-only");
744692dc 2857 goto failed_mount_wq;
ac27a0ec 2858 } else {
0390131b
FM		 2859 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2860 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2861 sbi->s_journal = NULL;
2862 needs_recovery = 0;
2863 goto no_journal;
ac27a0ec
DK		 2864 }
2865
eb40a09c
JS		 2866 if (ext4_blocks_count(es) > 0xffffffffULL &&
2867 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2868 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 2869 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 2870 goto failed_mount_wq;
eb40a09c
JS		 2871 }
2872
d4da6c9c
LT		 2873 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2874 jbd2_journal_set_features(sbi->s_journal,
2875 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 2876 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 2877 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2878 jbd2_journal_set_features(sbi->s_journal,
2879 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 2880 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2881 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 2882 } else {
2883 jbd2_journal_clear_features(sbi->s_journal,
2884 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2885 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2886 }
818d276c 2887
ac27a0ec
DK		 2888 /* We have now updated the journal if required, so we can
2889 * validate the data journaling mode. */
2890 switch (test_opt(sb, DATA_FLAGS)) {
2891 case 0:
2892 /* No mode set, assume a default based on the journal
63f57933
AM		 2893 * capabilities: ORDERED_DATA if the journal can
2894 * cope, else JOURNAL_DATA
2895 */
dab291af
MC		 2896 if (jbd2_journal_check_available_features
2897 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2898 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2899 else
2900 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2901 break;
2902
617ba13b
MC		 2903 case EXT4_MOUNT_ORDERED_DATA:
2904 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2905 if (!jbd2_journal_check_available_features
2906 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 2907 ext4_msg(sb, KERN_ERR, "Journal does not support "
2908 "requested data journaling mode");
744692dc 2909 goto failed_mount_wq;
ac27a0ec
DK		 2910 }
2911 default:
2912 break;
2913 }
b3881f74 2914 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 2915
0390131b 2916no_journal:
84061e07
DM		 2917 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2918 ext4_count_free_blocks(sb));
2919 if (!err)
2920 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2921 ext4_count_free_inodes(sb));
2922 if (!err)
2923 err = percpu_counter_init(&sbi->s_dirs_counter,
2924 ext4_count_dirs(sb));
2925 if (!err)
2926 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2927 if (err) {
2928 ext4_msg(sb, KERN_ERR, "insufficient memory");
2929 goto failed_mount_wq;
2930 }
206f7ab4 2931
4c0425ff
MC		 2932 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2933 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2934 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2935 goto failed_mount_wq;
2936 }
2937
ac27a0ec 2938 /*
dab291af 2939 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2940 * so we can safely mount the rest of the filesystem now.
2941 */
2942
1d1fe1ee
DH		 2943 root = ext4_iget(sb, EXT4_ROOT_INO);
2944 if (IS_ERR(root)) {
b31e1552 2945 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 2946 ret = PTR_ERR(root);
ac27a0ec
DK		 2947 goto failed_mount4;
2948 }
2949 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2950 iput(root);
b31e1552 2951 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 2952 goto failed_mount4;
2953 }
1d1fe1ee
DH		 2954 sb->s_root = d_alloc_root(root);
2955 if (!sb->s_root) {
b31e1552 2956 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 2957 iput(root);
2958 ret = -ENOMEM;
2959 goto failed_mount4;
2960 }
ac27a0ec 2961
2b2d6d01 2962 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2963
2964 /* determine the minimum size of new large inodes, if present */
2965 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2966 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2967 EXT4_GOOD_OLD_INODE_SIZE;
2968 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2969 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2970 if (sbi->s_want_extra_isize <
2971 le16_to_cpu(es->s_want_extra_isize))
2972 sbi->s_want_extra_isize =
2973 le16_to_cpu(es->s_want_extra_isize);
2974 if (sbi->s_want_extra_isize <
2975 le16_to_cpu(es->s_min_extra_isize))
2976 sbi->s_want_extra_isize =
2977 le16_to_cpu(es->s_min_extra_isize);
2978 }
2979 }
2980 /* Check if enough inode space is available */
2981 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2982 sbi->s_inode_size) {
2983 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2984 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 2985 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2986 "available");
ef7f3835
KS		 2987 }
2988
90576c0b
TT		 2989 if (test_opt(sb, DELALLOC) &&
2990 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
b31e1552
ES		 2991 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2992 "requested data journaling mode");
c2774d84 2993 clear_opt(sbi->s_mount_opt, DELALLOC);
90576c0b 2994 }
744692dc
JZ		 2995 if (test_opt(sb, DIOREAD_NOLOCK)) {
2996 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2997 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
2998 "option - requested data journaling mode");
2999 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3000 }
3001 if (sb->s_blocksize < PAGE_SIZE) {
3002 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3003 "option - block size is too small");
3004 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3005 }
3006 }
c2774d84 3007
6fd058f7
TT		 3008 err = ext4_setup_system_zone(sb);
3009 if (err) {
b31e1552 3010 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3011 "zone (%d)", err);
6fd058f7
TT		 3012 goto failed_mount4;
3013 }
3014
c2774d84
AK		 3015 ext4_ext_init(sb);
3016 err = ext4_mb_init(sb, needs_recovery);
3017 if (err) {
b31e1552
ES		 3018 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
3019 err);
c2774d84
AK		 3020 goto failed_mount4;
3021 }
3022
3197ebdb
TT		 3023 sbi->s_kobj.kset = ext4_kset;
3024 init_completion(&sbi->s_kobj_unregister);
3025 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3026 "%s", sb->s_id);
3027 if (err) {
3028 ext4_mb_release(sb);
3029 ext4_ext_release(sb);
3030 goto failed_mount4;
3031 };
3032
617ba13b
MC		 3033 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3034 ext4_orphan_cleanup(sb, es);
3035 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3036 if (needs_recovery) {
b31e1552 3037 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3038 ext4_mark_recovery_complete(sb, es);
3039 }
3040 if (EXT4_SB(sb)->s_journal) {
3041 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3042 descr = " journalled data mode";
3043 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3044 descr = " ordered data mode";
3045 else
3046 descr = " writeback data mode";
3047 } else
3048 descr = "out journal";
3049
d4c402d9
CW		 3050 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3051 "Opts: %s", descr, orig_data);
ac27a0ec
DK		 3052
3053 lock_kernel();
d4c402d9 3054 kfree(orig_data);
ac27a0ec
DK		 3055 return 0;
3056
617ba13b 3057cantfind_ext4:
ac27a0ec 3058 if (!silent)
b31e1552 3059 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3060 goto failed_mount;
3061
3062failed_mount4:
b31e1552 3063 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3064 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3065failed_mount_wq:
6fd058f7 3066 ext4_release_system_zone(sb);
0390131b
FM		 3067 if (sbi->s_journal) {
3068 jbd2_journal_destroy(sbi->s_journal);
3069 sbi->s_journal = NULL;
3070 }
84061e07
DM		 3071 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3072 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3073 percpu_counter_destroy(&sbi->s_dirs_counter);
3074 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec 3075failed_mount3:
c5ca7c76
TT		 3076 if (sbi->s_flex_groups) {
3077 if (is_vmalloc_addr(sbi->s_flex_groups))
3078 vfree(sbi->s_flex_groups);
3079 else
3080 kfree(sbi->s_flex_groups);
3081 }
ac27a0ec
DK		 3082failed_mount2:
3083 for (i = 0; i < db_count; i++)
3084 brelse(sbi->s_group_desc[i]);
3085 kfree(sbi->s_group_desc);
3086failed_mount:
240799cd 3087 if (sbi->s_proc) {
9f6200bb 3088 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3089 }
ac27a0ec
DK		 3090#ifdef CONFIG_QUOTA
3091 for (i = 0; i < MAXQUOTAS; i++)
3092 kfree(sbi->s_qf_names[i]);
3093#endif
617ba13b 3094 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3095 brelse(bh);
3096out_fail:
3097 sb->s_fs_info = NULL;
f6830165 3098 kfree(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3099 kfree(sbi);
3100 lock_kernel();
d4c402d9 3101 kfree(orig_data);
1d1fe1ee 3102 return ret;
ac27a0ec
DK		 3103}
3104
3105/*
3106 * Setup any per-fs journal parameters now. We'll do this both on
3107 * initial mount, once the journal has been initialised but before we've
3108 * done any recovery; and again on any subsequent remount.
3109 */
617ba13b 3110static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3111{
617ba13b 3112 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3113
30773840
TT		 3114 journal->j_commit_interval = sbi->s_commit_interval;
3115 journal->j_min_batch_time = sbi->s_min_batch_time;
3116 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3117
3118 spin_lock(&journal->j_state_lock);
3119 if (test_opt(sb, BARRIER))
dab291af 3120 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3121 else
dab291af 3122 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3123 if (test_opt(sb, DATA_ERR_ABORT))
3124 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3125 else
3126 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 3127 spin_unlock(&journal->j_state_lock);
3128}
3129
617ba13b 3130static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3131 unsigned int journal_inum)
3132{
3133 struct inode *journal_inode;
3134 journal_t *journal;
3135
0390131b
FM		 3136 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3137
ac27a0ec
DK		 3138 /* First, test for the existence of a valid inode on disk. Bad
3139 * things happen if we iget() an unused inode, as the subsequent
3140 * iput() will try to delete it. */
3141
1d1fe1ee
DH		 3142 journal_inode = ext4_iget(sb, journal_inum);
3143 if (IS_ERR(journal_inode)) {
b31e1552 3144 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3145 return NULL;
3146 }
3147 if (!journal_inode->i_nlink) {
3148 make_bad_inode(journal_inode);
3149 iput(journal_inode);
b31e1552 3150 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3151 return NULL;
3152 }
3153
e5f8eab8 3154 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3155 journal_inode, journal_inode->i_size);
1d1fe1ee 3156 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3157 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3158 iput(journal_inode);
3159 return NULL;
3160 }
3161
dab291af 3162 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3163 if (!journal) {
b31e1552 3164 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3165 iput(journal_inode);
3166 return NULL;
3167 }
3168 journal->j_private = sb;
617ba13b 3169 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3170 return journal;
3171}
3172
617ba13b 3173static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3174 dev_t j_dev)
3175{
2b2d6d01 3176 struct buffer_head *bh;
ac27a0ec 3177 journal_t *journal;
617ba13b
MC		 3178 ext4_fsblk_t start;
3179 ext4_fsblk_t len;
ac27a0ec 3180 int hblock, blocksize;
617ba13b 3181 ext4_fsblk_t sb_block;
ac27a0ec 3182 unsigned long offset;
2b2d6d01 3183 struct ext4_super_block *es;
ac27a0ec
DK		 3184 struct block_device *bdev;
3185
0390131b
FM		 3186 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3187
b31e1552 3188 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3189 if (bdev == NULL)
3190 return NULL;
3191
3192 if (bd_claim(bdev, sb)) {
b31e1552
ES		 3193 ext4_msg(sb, KERN_ERR,
3194 "failed to claim external journal device");
9a1c3542 3195 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 3196 return NULL;
3197 }
3198
3199 blocksize = sb->s_blocksize;
e1defc4f 3200 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3201 if (blocksize < hblock) {
b31e1552
ES		 3202 ext4_msg(sb, KERN_ERR,
3203 "blocksize too small for journal device");
ac27a0ec
DK		 3204 goto out_bdev;
3205 }
3206
617ba13b
MC		 3207 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3208 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3209 set_blocksize(bdev, blocksize);
3210 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3211 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3212 "external journal");
ac27a0ec
DK		 3213 goto out_bdev;
3214 }
3215
617ba13b
MC		 3216 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3217 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3218 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3219 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3220 ext4_msg(sb, KERN_ERR, "external journal has "
3221 "bad superblock");
ac27a0ec
DK		 3222 brelse(bh);
3223 goto out_bdev;
3224 }
3225
617ba13b 3226 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3227 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3228 brelse(bh);
3229 goto out_bdev;
3230 }
3231
bd81d8ee 3232 len = ext4_blocks_count(es);
ac27a0ec
DK		 3233 start = sb_block + 1;
3234 brelse(bh); /* we're done with the superblock */
3235
dab291af 3236 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3237 start, len, blocksize);
3238 if (!journal) {
b31e1552 3239 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3240 goto out_bdev;
3241 }
3242 journal->j_private = sb;
3243 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3244 wait_on_buffer(journal->j_sb_buffer);
3245 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3246 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3247 goto out_journal;
3248 }
3249 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3250 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3251 "user (unsupported) - %d",
ac27a0ec
DK		 3252 be32_to_cpu(journal->j_superblock->s_nr_users));
3253 goto out_journal;
3254 }
617ba13b
MC		 3255 EXT4_SB(sb)->journal_bdev = bdev;
3256 ext4_init_journal_params(sb, journal);
ac27a0ec 3257 return journal;
0b8e58a1 3258
ac27a0ec 3259out_journal:
dab291af 3260 jbd2_journal_destroy(journal);
ac27a0ec 3261out_bdev:
617ba13b 3262 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3263 return NULL;
3264}
3265
617ba13b
MC		 3266static int ext4_load_journal(struct super_block *sb,
3267 struct ext4_super_block *es,
ac27a0ec
DK		 3268 unsigned long journal_devnum)
3269{
3270 journal_t *journal;
3271 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3272 dev_t journal_dev;
3273 int err = 0;
3274 int really_read_only;
3275
0390131b
FM		 3276 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3277
ac27a0ec
DK		 3278 if (journal_devnum &&
3279 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3280 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3281 "numbers have changed");
ac27a0ec
DK		 3282 journal_dev = new_decode_dev(journal_devnum);
3283 } else
3284 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3285
3286 really_read_only = bdev_read_only(sb->s_bdev);
3287
3288 /*
3289 * Are we loading a blank journal or performing recovery after a
3290 * crash? For recovery, we need to check in advance whether we
3291 * can get read-write access to the device.
3292 */
617ba13b 3293 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3294 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3295 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3296 "required on readonly filesystem");
ac27a0ec 3297 if (really_read_only) {
b31e1552
ES		 3298 ext4_msg(sb, KERN_ERR, "write access "
3299 "unavailable, cannot proceed");
ac27a0ec
DK		 3300 return -EROFS;
3301 }
b31e1552
ES		 3302 ext4_msg(sb, KERN_INFO, "write access will "
3303 "be enabled during recovery");
ac27a0ec
DK		 3304 }
3305 }
3306
3307 if (journal_inum && journal_dev) {
b31e1552
ES		 3308 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3309 "and inode journals!");
ac27a0ec
DK		 3310 return -EINVAL;
3311 }
3312
3313 if (journal_inum) {
617ba13b 3314 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3315 return -EINVAL;
3316 } else {
617ba13b 3317 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3318 return -EINVAL;
3319 }
3320
90576c0b 3321 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 3322 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3323
ac27a0ec 3324 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3325 err = jbd2_journal_update_format(journal);
ac27a0ec 3326 if (err) {
b31e1552 3327 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 3328 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3329 return err;
3330 }
3331 }
3332
617ba13b 3333 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3334 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 3335 if (!err)
dab291af 3336 err = jbd2_journal_load(journal);
ac27a0ec
DK		 3337
3338 if (err) {
b31e1552 3339 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3340 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3341 return err;
3342 }
3343
617ba13b
MC		 3344 EXT4_SB(sb)->s_journal = journal;
3345 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 3346
3347 if (journal_devnum &&
3348 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3349 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3350
3351 /* Make sure we flush the recovery flag to disk. */
e2d67052 3352 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3353 }
3354
3355 return 0;
3356}
3357
e2d67052 3358static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3359{
e2d67052 3360 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3361 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3362 int error = 0;
ac27a0ec
DK		 3363
3364 if (!sbh)
c4be0c1d 3365 return error;
914258bf
TT		 3366 if (buffer_write_io_error(sbh)) {
3367 /*
3368 * Oh, dear. A previous attempt to write the
3369 * superblock failed. This could happen because the
3370 * USB device was yanked out. Or it could happen to
3371 * be a transient write error and maybe the block will
3372 * be remapped. Nothing we can do but to retry the
3373 * write and hope for the best.
3374 */
b31e1552
ES		 3375 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3376 "superblock detected");
914258bf
TT		 3377 clear_buffer_write_io_error(sbh);
3378 set_buffer_uptodate(sbh);
3379 }
71290b36
TT		 3380 /*
3381 * If the file system is mounted read-only, don't update the
3382 * superblock write time. This avoids updating the superblock
3383 * write time when we are mounting the root file system
3384 * read/only but we need to replay the journal; at that point,
3385 * for people who are east of GMT and who make their clock
3386 * tick in localtime for Windows bug-for-bug compatibility,
3387 * the clock is set in the future, and this will cause e2fsck
3388 * to complain and force a full file system check.
3389 */
3390 if (!(sb->s_flags & MS_RDONLY))
3391 es->s_wtime = cpu_to_le32(get_seconds());
afc32f7e 3392 es->s_kbytes_written =
60e6679e 3393 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 3394 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3395 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5d1b1b3f
AK		 3396 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3397 &EXT4_SB(sb)->s_freeblocks_counter));
3398 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3399 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 3400 sb->s_dirt = 0;
ac27a0ec
DK		 3401 BUFFER_TRACE(sbh, "marking dirty");
3402 mark_buffer_dirty(sbh);
914258bf 3403 if (sync) {
c4be0c1d
TS		 3404 error = sync_dirty_buffer(sbh);
3405 if (error)
3406 return error;
3407
3408 error = buffer_write_io_error(sbh);
3409 if (error) {
b31e1552
ES		 3410 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3411 "superblock");
914258bf
TT		 3412 clear_buffer_write_io_error(sbh);
3413 set_buffer_uptodate(sbh);
3414 }
3415 }
c4be0c1d 3416 return error;
ac27a0ec
DK		 3417}
3418
ac27a0ec
DK		 3419/*
3420 * Have we just finished recovery? If so, and if we are mounting (or
3421 * remounting) the filesystem readonly, then we will end up with a
3422 * consistent fs on disk. Record that fact.
3423 */
2b2d6d01
TT		 3424static void ext4_mark_recovery_complete(struct super_block *sb,
3425 struct ext4_super_block *es)
ac27a0ec 3426{
617ba13b 3427 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3428
0390131b
FM		 3429 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3430 BUG_ON(journal != NULL);
3431 return;
3432 }
dab291af 3433 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 3434 if (jbd2_journal_flush(journal) < 0)
3435 goto out;
3436
617ba13b 3437 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 3438 sb->s_flags & MS_RDONLY) {
617ba13b 3439 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 3440 ext4_commit_super(sb, 1);
ac27a0ec 3441 }
7ffe1ea8
HK		 3442
3443out:
dab291af 3444 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 3445}
3446
3447/*
3448 * If we are mounting (or read-write remounting) a filesystem whose journal
3449 * has recorded an error from a previous lifetime, move that error to the
3450 * main filesystem now.
3451 */
2b2d6d01
TT		 3452static void ext4_clear_journal_err(struct super_block *sb,
3453 struct ext4_super_block *es)
ac27a0ec
DK		 3454{
3455 journal_t *journal;
3456 int j_errno;
3457 const char *errstr;
3458
0390131b
FM		 3459 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3460
617ba13b 3461 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 3462
3463 /*
3464 * Now check for any error status which may have been recorded in the
617ba13b 3465 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 3466 */
3467
dab291af 3468 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 3469 if (j_errno) {
3470 char nbuf[16];
3471
617ba13b 3472 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 3473 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 3474 "from previous mount: %s", errstr);
12062ddd 3475 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 3476
617ba13b
MC		 3477 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3478 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 3479 ext4_commit_super(sb, 1);
ac27a0ec 3480
dab291af 3481 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 3482 }
3483}
3484
3485/*
3486 * Force the running and committing transactions to commit,
3487 * and wait on the commit.
3488 */
617ba13b 3489int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 3490{
3491 journal_t *journal;
0390131b 3492 int ret = 0;
ac27a0ec
DK		 3493
3494 if (sb->s_flags & MS_RDONLY)
3495 return 0;
3496
617ba13b 3497 journal = EXT4_SB(sb)->s_journal;
6b0310fb
ES		 3498 if (journal) {
3499 vfs_check_frozen(sb, SB_FREEZE_WRITE);
0390131b 3500 ret = ext4_journal_force_commit(journal);
6b0310fb 3501 }
0390131b 3502
ac27a0ec
DK		 3503 return ret;
3504}
3505
2b2d6d01 3506static void ext4_write_super(struct super_block *sb)
ac27a0ec 3507{
ebc1ac16 3508 lock_super(sb);
9ca92389 3509 ext4_commit_super(sb, 1);
ebc1ac16 3510 unlock_super(sb);
ac27a0ec
DK		 3511}
3512
617ba13b 3513static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 3514{
14ce0cb4 3515 int ret = 0;
9eddacf9 3516 tid_t target;
8d5d02e6 3517 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3518
9bffad1e 3519 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 3520 flush_workqueue(sbi->dio_unwritten_wq);
3521 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 3522 if (wait)
8d5d02e6 3523 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 3524 }
14ce0cb4 3525 return ret;
ac27a0ec
DK		 3526}
3527
3528/*
3529 * LVM calls this function before a (read-only) snapshot is created. This
3530 * gives us a chance to flush the journal completely and mark the fs clean.
3531 */
c4be0c1d 3532static int ext4_freeze(struct super_block *sb)
ac27a0ec 3533{
c4be0c1d
TS		 3534 int error = 0;
3535 journal_t *journal;
ac27a0ec 3536
9ca92389
TT		 3537 if (sb->s_flags & MS_RDONLY)
3538 return 0;
ac27a0ec 3539
9ca92389 3540 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 3541
9ca92389
TT		 3542 /* Now we set up the journal barrier. */
3543 jbd2_journal_lock_updates(journal);
ac27a0ec 3544
9ca92389
TT		 3545 /*
3546 * Don't clear the needs_recovery flag if we failed to flush
3547 * the journal.
3548 */
3549 error = jbd2_journal_flush(journal);
6b0310fb
ES		 3550 if (error < 0)
3551 goto out;
9ca92389
TT		 3552
3553 /* Journal blocked and flushed, clear needs_recovery flag. */
3554 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3555 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 3556out:
3557 /* we rely on s_frozen to stop further updates */
3558 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3559 return error;
ac27a0ec
DK		 3560}
3561
3562/*
3563 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3564 * flag here, even though the filesystem is not technically dirty yet.
3565 */
c4be0c1d 3566static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 3567{
9ca92389
TT		 3568 if (sb->s_flags & MS_RDONLY)
3569 return 0;
3570
3571 lock_super(sb);
3572 /* Reset the needs_recovery flag before the fs is unlocked. */
3573 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3574 ext4_commit_super(sb, 1);
3575 unlock_super(sb);
c4be0c1d 3576 return 0;
ac27a0ec
DK		 3577}
3578
2b2d6d01 3579static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3580{
2b2d6d01 3581 struct ext4_super_block *es;
617ba13b
MC		 3582 struct ext4_sb_info *sbi = EXT4_SB(sb);
3583 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3584 unsigned long old_sb_flags;
617ba13b 3585 struct ext4_mount_options old_opts;
c79d967d 3586 int enable_quota = 0;
8a266467 3587 ext4_group_t g;
b3881f74 3588 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 3589 int err;
3590#ifdef CONFIG_QUOTA
3591 int i;
3592#endif
d4c402d9 3593 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec 3594
337eb00a
AIB		 3595 lock_kernel();
3596
ac27a0ec 3597 /* Store the original options */
bbd6851a 3598 lock_super(sb);
ac27a0ec
DK		 3599 old_sb_flags = sb->s_flags;
3600 old_opts.s_mount_opt = sbi->s_mount_opt;
3601 old_opts.s_resuid = sbi->s_resuid;
3602 old_opts.s_resgid = sbi->s_resgid;
3603 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 3604 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3605 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3606#ifdef CONFIG_QUOTA
3607 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3608 for (i = 0; i < MAXQUOTAS; i++)
3609 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3610#endif
b3881f74
TT		 3611 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3612 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 3613
3614 /*
3615 * Allow the "check" option to be passed as a remount option.
3616 */
b3881f74
TT		 3617 if (!parse_options(data, sb, NULL, &journal_ioprio,
3618 &n_blocks_count, 1)) {
ac27a0ec
DK		 3619 err = -EINVAL;
3620 goto restore_opts;
3621 }
3622
4ab2f15b 3623 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 3624 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 3625
3626 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3627 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3628
3629 es = sbi->s_es;
3630
b3881f74 3631 if (sbi->s_journal) {
0390131b 3632 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 3633 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3634 }
ac27a0ec
DK		 3635
3636 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3637 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 3638 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 3639 err = -EROFS;
3640 goto restore_opts;
3641 }
3642
3643 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 3644 err = dquot_suspend(sb, -1);
3645 if (err < 0)
c79d967d 3646 goto restore_opts;
c79d967d 3647
ac27a0ec
DK		 3648 /*
3649 * First of all, the unconditional stuff we have to do
3650 * to disable replay of the journal when we next remount
3651 */
3652 sb->s_flags |= MS_RDONLY;
3653
3654 /*
3655 * OK, test if we are remounting a valid rw partition
3656 * readonly, and if so set the rdonly flag and then
3657 * mark the partition as valid again.
3658 */
617ba13b
MC		 3659 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3660 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3661 es->s_state = cpu_to_le16(sbi->s_mount_state);
3662
a63c9eb2 3663 if (sbi->s_journal)
0390131b 3664 ext4_mark_recovery_complete(sb, es);
ac27a0ec 3665 } else {
a13fb1a4
ES		 3666 /* Make sure we can mount this feature set readwrite */
3667 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 3668 err = -EROFS;
3669 goto restore_opts;
3670 }
8a266467
TT		 3671 /*
3672 * Make sure the group descriptor checksums
0b8e58a1 3673 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 3674 */
3675 for (g = 0; g < sbi->s_groups_count; g++) {
3676 struct ext4_group_desc *gdp =
3677 ext4_get_group_desc(sb, g, NULL);
3678
3679 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 3680 ext4_msg(sb, KERN_ERR,
3681 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 3682 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3683 le16_to_cpu(gdp->bg_checksum));
3684 err = -EINVAL;
3685 goto restore_opts;
3686 }
3687 }
3688
ead6596b
ES		 3689 /*
3690 * If we have an unprocessed orphan list hanging
3691 * around from a previously readonly bdev mount,
3692 * require a full umount/remount for now.
3693 */
3694 if (es->s_last_orphan) {
b31e1552 3695 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 3696 "remount RDWR because of unprocessed "
3697 "orphan inode list. Please "
b31e1552 3698 "umount/remount instead");
ead6596b
ES		 3699 err = -EINVAL;
3700 goto restore_opts;
3701 }
3702
ac27a0ec
DK		 3703 /*
3704 * Mounting a RDONLY partition read-write, so reread
3705 * and store the current valid flag. (It may have
3706 * been changed by e2fsck since we originally mounted
3707 * the partition.)
3708 */
0390131b
FM		 3709 if (sbi->s_journal)
3710 ext4_clear_journal_err(sb, es);
ac27a0ec 3711 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3712 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3713 goto restore_opts;
2b2d6d01 3714 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 3715 sb->s_flags &= ~MS_RDONLY;
c79d967d 3716 enable_quota = 1;
ac27a0ec
DK		 3717 }
3718 }
6fd058f7 3719 ext4_setup_system_zone(sb);
0390131b 3720 if (sbi->s_journal == NULL)
e2d67052 3721 ext4_commit_super(sb, 1);
0390131b 3722
ac27a0ec
DK		 3723#ifdef CONFIG_QUOTA
3724 /* Release old quota file names */
3725 for (i = 0; i < MAXQUOTAS; i++)
3726 if (old_opts.s_qf_names[i] &&
3727 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3728 kfree(old_opts.s_qf_names[i]);
3729#endif
bbd6851a 3730 unlock_super(sb);
337eb00a 3731 unlock_kernel();
c79d967d 3732 if (enable_quota)
0f0dd62f 3733 dquot_resume(sb, -1);
d4c402d9
CW		 3734
3735 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
3736 kfree(orig_data);
ac27a0ec 3737 return 0;
0b8e58a1 3738
ac27a0ec
DK		 3739restore_opts:
3740 sb->s_flags = old_sb_flags;
3741 sbi->s_mount_opt = old_opts.s_mount_opt;
3742 sbi->s_resuid = old_opts.s_resuid;
3743 sbi->s_resgid = old_opts.s_resgid;
3744 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 3745 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3746 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 3747#ifdef CONFIG_QUOTA
3748 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3749 for (i = 0; i < MAXQUOTAS; i++) {
3750 if (sbi->s_qf_names[i] &&
3751 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3752 kfree(sbi->s_qf_names[i]);
3753 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3754 }
3755#endif
bbd6851a 3756 unlock_super(sb);
337eb00a 3757 unlock_kernel();
d4c402d9 3758 kfree(orig_data);
ac27a0ec
DK		 3759 return err;
3760}
3761
2b2d6d01 3762static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3763{
3764 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3765 struct ext4_sb_info *sbi = EXT4_SB(sb);
3766 struct ext4_super_block *es = sbi->s_es;
960cc398 3767 u64 fsid;
ac27a0ec 3768
5e70030d
BP		 3769 if (test_opt(sb, MINIX_DF)) {
3770 sbi->s_overhead_last = 0;
6bc9feff 3771 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 3772 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 3773 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3774
3775 /*
5e70030d
BP		 3776 * Compute the overhead (FS structures). This is constant
3777 * for a given filesystem unless the number of block groups
3778 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3779 */
3780
3781 /*
3782 * All of the blocks before first_data_block are
3783 * overhead
3784 */
3785 overhead = le32_to_cpu(es->s_first_data_block);
3786
3787 /*
3788 * Add the overhead attributed to the superblock and
3789 * block group descriptors. If the sparse superblocks
3790 * feature is turned on, then not all groups have this.
3791 */
3792 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3793 overhead += ext4_bg_has_super(sb, i) +
3794 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3795 cond_resched();
3796 }
3797
3798 /*
3799 * Every block group has an inode bitmap, a block
3800 * bitmap, and an inode table.
3801 */
5e70030d
BP		 3802 overhead += ngroups * (2 + sbi->s_itb_per_group);
3803 sbi->s_overhead_last = overhead;
3804 smp_wmb();
6bc9feff 3805 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3806 }
3807
617ba13b 3808 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3809 buf->f_bsize = sb->s_blocksize;
5e70030d 3810 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3811 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3812 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
bd81d8ee
LV		 3813 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3814 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3815 buf->f_bavail = 0;
3816 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3817 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 3818 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3819 fsid = le64_to_cpup((void *)es->s_uuid) ^
3820 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3821 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3822 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 3823
ac27a0ec
DK		 3824 return 0;
3825}
3826
0b8e58a1
AD		 3827/* Helper function for writing quotas on sync - we need to start transaction
3828 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 3829 * Process 1 Process 2
617ba13b 3830 * ext4_create() quota_sync()
a269eb18 3831 * jbd2_journal_start() write_dquot()
871a2931 3832 * dquot_initialize() down(dqio_mutex)
dab291af 3833 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3834 *
3835 */
3836
3837#ifdef CONFIG_QUOTA
3838
3839static inline struct inode *dquot_to_inode(struct dquot *dquot)
3840{
3841 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3842}
3843
617ba13b 3844static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3845{
3846 int ret, err;
3847 handle_t *handle;
3848 struct inode *inode;
3849
3850 inode = dquot_to_inode(dquot);
617ba13b 3851 handle = ext4_journal_start(inode,
0b8e58a1 3852 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3853 if (IS_ERR(handle))
3854 return PTR_ERR(handle);
3855 ret = dquot_commit(dquot);
617ba13b 3856 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3857 if (!ret)
3858 ret = err;
3859 return ret;
3860}
3861
617ba13b 3862static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3863{
3864 int ret, err;
3865 handle_t *handle;
3866
617ba13b 3867 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3868 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3869 if (IS_ERR(handle))
3870 return PTR_ERR(handle);
3871 ret = dquot_acquire(dquot);
617ba13b 3872 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3873 if (!ret)
3874 ret = err;
3875 return ret;
3876}
3877
617ba13b 3878static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3879{
3880 int ret, err;
3881 handle_t *handle;
3882
617ba13b 3883 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3884 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3885 if (IS_ERR(handle)) {
3886 /* Release dquot anyway to avoid endless cycle in dqput() */
3887 dquot_release(dquot);
ac27a0ec 3888 return PTR_ERR(handle);
9c3013e9 3889 }
ac27a0ec 3890 ret = dquot_release(dquot);
617ba13b 3891 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3892 if (!ret)
3893 ret = err;
3894 return ret;
3895}
3896
617ba13b 3897static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3898{
2c8be6b2 3899 /* Are we journaling quotas? */
617ba13b
MC		 3900 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3901 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3902 dquot_mark_dquot_dirty(dquot);
617ba13b 3903 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3904 } else {
3905 return dquot_mark_dquot_dirty(dquot);
3906 }
3907}
3908
617ba13b 3909static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3910{
3911 int ret, err;
3912 handle_t *handle;
3913
3914 /* Data block + inode block */
617ba13b 3915 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3916 if (IS_ERR(handle))
3917 return PTR_ERR(handle);
3918 ret = dquot_commit_info(sb, type);
617ba13b 3919 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3920 if (!ret)
3921 ret = err;
3922 return ret;
3923}
3924
3925/*
3926 * Turn on quotas during mount time - we need to find
3927 * the quota file and such...
3928 */
617ba13b 3929static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3930{
287a8095
CH		 3931 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3932 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3933}
3934
3935/*
3936 * Standard function to be called on quota_on
3937 */
617ba13b 3938static int ext4_quota_on(struct super_block *sb, int type, int format_id,
307ae18a 3939 char *name)
ac27a0ec
DK		 3940{
3941 int err;
8264613d 3942 struct path path;
ac27a0ec
DK		 3943
3944 if (!test_opt(sb, QUOTA))
3945 return -EINVAL;
0623543b 3946
8264613d 3947 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 3948 if (err)
3949 return err;
0623543b 3950
ac27a0ec 3951 /* Quotafile not on the same filesystem? */
8264613d
AV		 3952 if (path.mnt->mnt_sb != sb) {
3953 path_put(&path);
ac27a0ec
DK		 3954 return -EXDEV;
3955 }
0623543b
JK		 3956 /* Journaling quota? */
3957 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3958 /* Quotafile not in fs root? */
8264613d 3959 if (path.dentry->d_parent != sb->s_root)
b31e1552
ES		 3960 ext4_msg(sb, KERN_WARNING,
3961 "Quota file not on filesystem root. "
3962 "Journaled quota will not work");
2b2d6d01 3963 }
0623543b
JK		 3964
3965 /*
3966 * When we journal data on quota file, we have to flush journal to see
3967 * all updates to the file when we bypass pagecache...
3968 */
0390131b
FM		 3969 if (EXT4_SB(sb)->s_journal &&
3970 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 3971 /*
3972 * We don't need to lock updates but journal_flush() could
3973 * otherwise be livelocked...
3974 */
3975 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3976 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3977 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3978 if (err) {
8264613d 3979 path_put(&path);
7ffe1ea8
HK		 3980 return err;
3981 }
0623543b
JK		 3982 }
3983
287a8095 3984 err = dquot_quota_on_path(sb, type, format_id, &path);
8264613d 3985 path_put(&path);
77e69dac 3986 return err;
ac27a0ec
DK		 3987}
3988
3989/* Read data from quotafile - avoid pagecache and such because we cannot afford
3990 * acquiring the locks... As quota files are never truncated and quota code
3991 * itself serializes the operations (and noone else should touch the files)
3992 * we don't have to be afraid of races */
617ba13b 3993static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3994 size_t len, loff_t off)
3995{
3996 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3997 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3998 int err = 0;
3999 int offset = off & (sb->s_blocksize - 1);
4000 int tocopy;
4001 size_t toread;
4002 struct buffer_head *bh;
4003 loff_t i_size = i_size_read(inode);
4004
4005 if (off > i_size)
4006 return 0;
4007 if (off+len > i_size)
4008 len = i_size-off;
4009 toread = len;
4010 while (toread > 0) {
4011 tocopy = sb->s_blocksize - offset < toread ?
4012 sb->s_blocksize - offset : toread;
617ba13b 4013 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4014 if (err)
4015 return err;
4016 if (!bh) /* A hole? */
4017 memset(data, 0, tocopy);
4018 else
4019 memcpy(data, bh->b_data+offset, tocopy);
4020 brelse(bh);
4021 offset = 0;
4022 toread -= tocopy;
4023 data += tocopy;
4024 blk++;
4025 }
4026 return len;
4027}
4028
4029/* Write to quotafile (we know the transaction is already started and has
4030 * enough credits) */
617ba13b 4031static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4032 const char *data, size_t len, loff_t off)
4033{
4034 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4035 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4036 int err = 0;
4037 int offset = off & (sb->s_blocksize - 1);
617ba13b 4038 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 4039 struct buffer_head *bh;
4040 handle_t *handle = journal_current_handle();
4041
0390131b 4042 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4043 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4044 " cancelled because transaction is not started",
9c3013e9
JK		 4045 (unsigned long long)off, (unsigned long long)len);
4046 return -EIO;
4047 }
67eeb568
DM		 4048 /*
4049 * Since we account only one data block in transaction credits,
4050 * then it is impossible to cross a block boundary.
4051 */
4052 if (sb->s_blocksize - offset < len) {
4053 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4054 " cancelled because not block aligned",
4055 (unsigned long long)off, (unsigned long long)len);
4056 return -EIO;
4057 }
4058
ac27a0ec 4059 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4060 bh = ext4_bread(handle, inode, blk, 1, &err);
4061 if (!bh)
4062 goto out;
4063 if (journal_quota) {
4064 err = ext4_journal_get_write_access(handle, bh);
4065 if (err) {
4066 brelse(bh);
ac27a0ec 4067 goto out;
ac27a0ec 4068 }
ac27a0ec 4069 }
67eeb568
DM		 4070 lock_buffer(bh);
4071 memcpy(bh->b_data+offset, data, len);
4072 flush_dcache_page(bh->b_page);
4073 unlock_buffer(bh);
4074 if (journal_quota)
4075 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4076 else {
4077 /* Always do at least ordered writes for quotas */
4078 err = ext4_jbd2_file_inode(handle, inode);
4079 mark_buffer_dirty(bh);
4080 }
4081 brelse(bh);
ac27a0ec 4082out:
67eeb568 4083 if (err) {
4d04e4fb 4084 mutex_unlock(&inode->i_mutex);
ac27a0ec 4085 return err;
4d04e4fb 4086 }
67eeb568
DM		 4087 if (inode->i_size < off + len) {
4088 i_size_write(inode, off + len);
617ba13b 4089 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 4090 }
ac27a0ec 4091 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 4092 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4093 mutex_unlock(&inode->i_mutex);
67eeb568 4094 return len;
ac27a0ec
DK		 4095}
4096
4097#endif
4098
0b8e58a1
AD		 4099static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4100 const char *dev_name, void *data, struct vfsmount *mnt)
ac27a0ec 4101{
0b8e58a1 4102 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
ac27a0ec
DK		 4103}
4104
37f328eb 4105#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4106static struct file_system_type ext2_fs_type = {
4107 .owner = THIS_MODULE,
4108 .name = "ext2",
4109 .get_sb = ext4_get_sb,
4110 .kill_sb = kill_block_super,
4111 .fs_flags = FS_REQUIRES_DEV,
4112};
4113
4114static inline void register_as_ext2(void)
4115{
4116 int err = register_filesystem(&ext2_fs_type);
4117 if (err)
4118 printk(KERN_WARNING
4119 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4120}
4121
4122static inline void unregister_as_ext2(void)
4123{
4124 unregister_filesystem(&ext2_fs_type);
4125}
51b7e3c9 4126MODULE_ALIAS("ext2");
24b58424
TT		 4127#else
4128static inline void register_as_ext2(void) { }
4129static inline void unregister_as_ext2(void) { }
4130#endif
4131
37f328eb 4132#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4133static inline void register_as_ext3(void)
4134{
4135 int err = register_filesystem(&ext3_fs_type);
4136 if (err)
4137 printk(KERN_WARNING
4138 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4139}
4140
4141static inline void unregister_as_ext3(void)
4142{
4143 unregister_filesystem(&ext3_fs_type);
4144}
51b7e3c9 4145MODULE_ALIAS("ext3");
24b58424
TT		 4146#else
4147static inline void register_as_ext3(void) { }
4148static inline void unregister_as_ext3(void) { }
4149#endif
4150
03010a33
TT		 4151static struct file_system_type ext4_fs_type = {
4152 .owner = THIS_MODULE,
4153 .name = "ext4",
4154 .get_sb = ext4_get_sb,
4155 .kill_sb = kill_block_super,
4156 .fs_flags = FS_REQUIRES_DEV,
4157};
4158
617ba13b 4159static int __init init_ext4_fs(void)
ac27a0ec 4160{
c9de560d
AT		 4161 int err;
4162
12e9b892 4163 ext4_check_flag_values();
6fd058f7
TT		 4164 err = init_ext4_system_zone();
4165 if (err)
4166 return err;
3197ebdb
TT		 4167 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4168 if (!ext4_kset)
6fd058f7 4169 goto out4;
9f6200bb 4170 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 4171 err = init_ext4_mballoc();
ac27a0ec 4172 if (err)
6fd058f7 4173 goto out3;
c9de560d
AT		 4174
4175 err = init_ext4_xattr();
4176 if (err)
4177 goto out2;
ac27a0ec
DK		 4178 err = init_inodecache();
4179 if (err)
4180 goto out1;
24b58424
TT		 4181 register_as_ext2();
4182 register_as_ext3();
03010a33 4183 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 4184 if (err)
4185 goto out;
4186 return 0;
4187out:
24b58424
TT		 4188 unregister_as_ext2();
4189 unregister_as_ext3();
ac27a0ec
DK		 4190 destroy_inodecache();
4191out1:
617ba13b 4192 exit_ext4_xattr();
c9de560d
AT		 4193out2:
4194 exit_ext4_mballoc();
6fd058f7
TT		 4195out3:
4196 remove_proc_entry("fs/ext4", NULL);
4197 kset_unregister(ext4_kset);
4198out4:
4199 exit_ext4_system_zone();
ac27a0ec
DK		 4200 return err;
4201}
4202
617ba13b 4203static void __exit exit_ext4_fs(void)
ac27a0ec 4204{
24b58424
TT		 4205 unregister_as_ext2();
4206 unregister_as_ext3();
03010a33 4207 unregister_filesystem(&ext4_fs_type);
ac27a0ec 4208 destroy_inodecache();
617ba13b 4209 exit_ext4_xattr();
c9de560d 4210 exit_ext4_mballoc();
9f6200bb 4211 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4212 kset_unregister(ext4_kset);
6fd058f7 4213 exit_ext4_system_zone();
ac27a0ec
DK		 4214}
4215
4216MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4217MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4218MODULE_LICENSE("GPL");
617ba13b
MC		 4219module_init(init_ext4_fs)
4220module_exit(exit_ext4_fs)
kernel source for telekom puls (alcatel/mediatek)