projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'fixes-for-v3.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/balbi...
[linux-2.6-block.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>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
7abc52c2 41#include <linux/cleancache.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
6f91bc5f 48#include "ext4_extents.h"
3dcf5451 49#include "ext4_jbd2.h"
ac27a0ec
DK		 50#include "xattr.h"
51#include "acl.h"
3661d286 52#include "mballoc.h"
ac27a0ec 53
9bffad1e
TT		 54#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
1f109d5a 57static struct proc_dir_entry *ext4_proc_root;
3197ebdb 58static struct kset *ext4_kset;
0b75a840
LC		 59static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
2adf6da8 65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 66static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 67static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
617ba13b 71static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 72static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 73 char nbuf[16]);
2b2d6d01
TT		 74static int ext4_remount(struct super_block *sb, int *flags, char *data);
75static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 76static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 77static void ext4_write_super(struct super_block *sb);
c4be0c1d 78static int ext4_freeze(struct super_block *sb);
152a0836
AV		 79static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
80 const char *dev_name, void *data);
2035e776
TT		 81static inline int ext2_feature_set_ok(struct super_block *sb);
82static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 83static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 84static void ext4_destroy_lazyinit_thread(void);
85static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 86static void ext4_clear_request_list(void);
ac27a0ec 87
2035e776
TT		 88#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
89static struct file_system_type ext2_fs_type = {
90 .owner = THIS_MODULE,
91 .name = "ext2",
92 .mount = ext4_mount,
93 .kill_sb = kill_block_super,
94 .fs_flags = FS_REQUIRES_DEV,
95};
96#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97#else
98#define IS_EXT2_SB(sb) (0)
99#endif
100
101
ba69f9ab
JK		 102#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
105 .name = "ext3",
152a0836 106 .mount = ext4_mount,
ba69f9ab
JK		 107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
109};
110#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
111#else
112#define IS_EXT3_SB(sb) (0)
113#endif
bd81d8ee 114
9933fc0a
TT		 115void *ext4_kvmalloc(size_t size, gfp_t flags)
116{
117 void *ret;
118
119 ret = kmalloc(size, flags);
120 if (!ret)
121 ret = __vmalloc(size, flags, PAGE_KERNEL);
122 return ret;
123}
124
125void *ext4_kvzalloc(size_t size, gfp_t flags)
126{
127 void *ret;
128
db9481c0 129 ret = kzalloc(size, flags);
9933fc0a
TT		 130 if (!ret)
131 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
132 return ret;
133}
134
135void ext4_kvfree(void *ptr)
136{
137 if (is_vmalloc_addr(ptr))
138 vfree(ptr);
139 else
140 kfree(ptr);
141
142}
143
8fadc143
AR		 144ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
145 struct ext4_group_desc *bg)
bd81d8ee 146{
3a14589c 147 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 148 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 149 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 150}
151
8fadc143
AR		 152ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
153 struct ext4_group_desc *bg)
bd81d8ee 154{
5272f837 155 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 156 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 157 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 158}
159
8fadc143
AR		 160ext4_fsblk_t ext4_inode_table(struct super_block *sb,
161 struct ext4_group_desc *bg)
bd81d8ee 162{
5272f837 163 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 164 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 165 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 166}
167
021b65bb
TT		 168__u32 ext4_free_group_clusters(struct super_block *sb,
169 struct ext4_group_desc *bg)
560671a0
AK		 170{
171 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
172 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 173 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 174}
175
176__u32 ext4_free_inodes_count(struct super_block *sb,
177 struct ext4_group_desc *bg)
178{
179 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
180 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 181 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 182}
183
184__u32 ext4_used_dirs_count(struct super_block *sb,
185 struct ext4_group_desc *bg)
186{
187 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
188 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 189 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 190}
191
192__u32 ext4_itable_unused_count(struct super_block *sb,
193 struct ext4_group_desc *bg)
194{
195 return le16_to_cpu(bg->bg_itable_unused_lo) |
196 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 197 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 198}
199
8fadc143
AR		 200void ext4_block_bitmap_set(struct super_block *sb,
201 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 202{
3a14589c 203 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 204 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
205 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 206}
207
8fadc143
AR		 208void ext4_inode_bitmap_set(struct super_block *sb,
209 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 210{
5272f837 211 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 212 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
213 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 214}
215
8fadc143
AR		 216void ext4_inode_table_set(struct super_block *sb,
217 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 218{
5272f837 219 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 220 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
221 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 222}
223
021b65bb
TT		 224void ext4_free_group_clusters_set(struct super_block *sb,
225 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 226{
227 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
228 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
229 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
230}
231
232void ext4_free_inodes_set(struct super_block *sb,
233 struct ext4_group_desc *bg, __u32 count)
234{
235 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
236 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
237 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
238}
239
240void ext4_used_dirs_set(struct super_block *sb,
241 struct ext4_group_desc *bg, __u32 count)
242{
243 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
244 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
245 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
246}
247
248void ext4_itable_unused_set(struct super_block *sb,
249 struct ext4_group_desc *bg, __u32 count)
250{
251 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
252 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
253 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
254}
255
d3d1faf6
CW		 256
257/* Just increment the non-pointer handle value */
258static handle_t *ext4_get_nojournal(void)
259{
260 handle_t *handle = current->journal_info;
261 unsigned long ref_cnt = (unsigned long)handle;
262
263 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
264
265 ref_cnt++;
266 handle = (handle_t *)ref_cnt;
267
268 current->journal_info = handle;
269 return handle;
270}
271
272
273/* Decrement the non-pointer handle value */
274static void ext4_put_nojournal(handle_t *handle)
275{
276 unsigned long ref_cnt = (unsigned long)handle;
277
278 BUG_ON(ref_cnt == 0);
279
280 ref_cnt--;
281 handle = (handle_t *)ref_cnt;
282
283 current->journal_info = handle;
284}
285
ac27a0ec 286/*
dab291af 287 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 288 *
289 * The only special thing we need to do here is to make sure that all
290 * journal_end calls result in the superblock being marked dirty, so
291 * that sync() will call the filesystem's write_super callback if
292 * appropriate.
be4f27d3
YY		 293 *
294 * To avoid j_barrier hold in userspace when a user calls freeze(),
295 * ext4 prevents a new handle from being started by s_frozen, which
296 * is in an upper layer.
ac27a0ec 297 */
617ba13b 298handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 299{
300 journal_t *journal;
be4f27d3 301 handle_t *handle;
ac27a0ec 302
12706394 303 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
ac27a0ec
DK		 304 if (sb->s_flags & MS_RDONLY)
305 return ERR_PTR(-EROFS);
306
617ba13b 307 journal = EXT4_SB(sb)->s_journal;
be4f27d3
YY		 308 handle = ext4_journal_current_handle();
309
310 /*
311 * If a handle has been started, it should be allowed to
312 * finish, otherwise deadlock could happen between freeze
313 * and others(e.g. truncate) due to the restart of the
314 * journal handle if the filesystem is forzen and active
315 * handles are not stopped.
316 */
317 if (!handle)
318 vfs_check_frozen(sb, SB_FREEZE_TRANS);
319
320 if (!journal)
321 return ext4_get_nojournal();
322 /*
323 * Special case here: if the journal has aborted behind our
324 * backs (eg. EIO in the commit thread), then we still need to
325 * take the FS itself readonly cleanly.
326 */
327 if (is_journal_aborted(journal)) {
328 ext4_abort(sb, "Detected aborted journal");
329 return ERR_PTR(-EROFS);
ac27a0ec 330 }
be4f27d3 331 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 332}
333
334/*
335 * The only special thing we need to do here is to make sure that all
dab291af 336 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 337 * that sync() will call the filesystem's write_super callback if
338 * appropriate.
339 */
c398eda0 340int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 341{
342 struct super_block *sb;
343 int err;
344 int rc;
345
0390131b 346 if (!ext4_handle_valid(handle)) {
d3d1faf6 347 ext4_put_nojournal(handle);
0390131b
FM		 348 return 0;
349 }
ac27a0ec
DK		 350 sb = handle->h_transaction->t_journal->j_private;
351 err = handle->h_err;
dab291af 352 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 353
354 if (!err)
355 err = rc;
356 if (err)
c398eda0 357 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 358 return err;
359}
360
90c7201b
TT		 361void ext4_journal_abort_handle(const char *caller, unsigned int line,
362 const char *err_fn, struct buffer_head *bh,
363 handle_t *handle, int err)
ac27a0ec
DK		 364{
365 char nbuf[16];
617ba13b 366 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 367
0390131b
FM		 368 BUG_ON(!ext4_handle_valid(handle));
369
ac27a0ec
DK		 370 if (bh)
371 BUFFER_TRACE(bh, "abort");
372
373 if (!handle->h_err)
374 handle->h_err = err;
375
376 if (is_handle_aborted(handle))
377 return;
378
92b97816 379 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
90c7201b 380 caller, line, errstr, err_fn);
ac27a0ec 381
dab291af 382 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 383}
384
1c13d5c0
TT		 385static void __save_error_info(struct super_block *sb, const char *func,
386 unsigned int line)
387{
388 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
389
390 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
391 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
392 es->s_last_error_time = cpu_to_le32(get_seconds());
393 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
394 es->s_last_error_line = cpu_to_le32(line);
395 if (!es->s_first_error_time) {
396 es->s_first_error_time = es->s_last_error_time;
397 strncpy(es->s_first_error_func, func,
398 sizeof(es->s_first_error_func));
399 es->s_first_error_line = cpu_to_le32(line);
400 es->s_first_error_ino = es->s_last_error_ino;
401 es->s_first_error_block = es->s_last_error_block;
402 }
66e61a9e
TT		 403 /*
404 * Start the daily error reporting function if it hasn't been
405 * started already
406 */
407 if (!es->s_error_count)
408 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
1c13d5c0
TT		 409 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
410}
411
412static void save_error_info(struct super_block *sb, const char *func,
413 unsigned int line)
414{
415 __save_error_info(sb, func, line);
416 ext4_commit_super(sb, 1);
417}
418
7c2e7087
TT		 419/*
420 * The del_gendisk() function uninitializes the disk-specific data
421 * structures, including the bdi structure, without telling anyone
422 * else. Once this happens, any attempt to call mark_buffer_dirty()
423 * (for example, by ext4_commit_super), will cause a kernel OOPS.
424 * This is a kludge to prevent these oops until we can put in a proper
425 * hook in del_gendisk() to inform the VFS and file system layers.
426 */
427static int block_device_ejected(struct super_block *sb)
428{
429 struct inode *bd_inode = sb->s_bdev->bd_inode;
430 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
431
432 return bdi->dev == NULL;
433}
434
18aadd47
BJ		 435static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
436{
437 struct super_block *sb = journal->j_private;
438 struct ext4_sb_info *sbi = EXT4_SB(sb);
439 int error = is_journal_aborted(journal);
440 struct ext4_journal_cb_entry *jce, *tmp;
441
442 spin_lock(&sbi->s_md_lock);
443 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
444 list_del_init(&jce->jce_list);
445 spin_unlock(&sbi->s_md_lock);
446 jce->jce_func(sb, jce, error);
447 spin_lock(&sbi->s_md_lock);
448 }
449 spin_unlock(&sbi->s_md_lock);
450}
1c13d5c0 451
ac27a0ec
DK		 452/* Deal with the reporting of failure conditions on a filesystem such as
453 * inconsistencies detected or read IO failures.
454 *
455 * On ext2, we can store the error state of the filesystem in the
617ba13b 456 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 457 * write ordering constraints on the superblock which prevent us from
458 * writing it out straight away; and given that the journal is about to
459 * be aborted, we can't rely on the current, or future, transactions to
460 * write out the superblock safely.
461 *
dab291af 462 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 463 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 464 * that error until we've noted it down and cleared it.
465 */
466
617ba13b 467static void ext4_handle_error(struct super_block *sb)
ac27a0ec 468{
ac27a0ec
DK		 469 if (sb->s_flags & MS_RDONLY)
470 return;
471
2b2d6d01 472 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 473 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 474
4ab2f15b 475 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 476 if (journal)
dab291af 477 jbd2_journal_abort(journal, -EIO);
ac27a0ec 478 }
2b2d6d01 479 if (test_opt(sb, ERRORS_RO)) {
b31e1552 480 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 481 sb->s_flags |= MS_RDONLY;
482 }
ac27a0ec 483 if (test_opt(sb, ERRORS_PANIC))
617ba13b 484 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 485 sb->s_id);
486}
487
12062ddd 488void __ext4_error(struct super_block *sb, const char *function,
c398eda0 489 unsigned int line, const char *fmt, ...)
ac27a0ec 490{
0ff2ea7d 491 struct va_format vaf;
ac27a0ec
DK		 492 va_list args;
493
494 va_start(args, fmt);
0ff2ea7d
JP		 495 vaf.fmt = fmt;
496 vaf.va = &args;
497 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
498 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec
DK		 499 va_end(args);
500
617ba13b 501 ext4_handle_error(sb);
ac27a0ec
DK		 502}
503
c398eda0
TT		 504void ext4_error_inode(struct inode *inode, const char *function,
505 unsigned int line, ext4_fsblk_t block,
273df556
FM		 506 const char *fmt, ...)
507{
508 va_list args;
f7c21177 509 struct va_format vaf;
1c13d5c0 510 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 511
1c13d5c0
TT		 512 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
513 es->s_last_error_block = cpu_to_le64(block);
514 save_error_info(inode->i_sb, function, line);
273df556 515 va_start(args, fmt);
f7c21177
TT		 516 vaf.fmt = fmt;
517 vaf.va = &args;
c398eda0 518 if (block)
d9ee81da
JP		 519 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
520 "inode #%lu: block %llu: comm %s: %pV\n",
521 inode->i_sb->s_id, function, line, inode->i_ino,
522 block, current->comm, &vaf);
523 else
524 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
525 "inode #%lu: comm %s: %pV\n",
526 inode->i_sb->s_id, function, line, inode->i_ino,
527 current->comm, &vaf);
273df556
FM		 528 va_end(args);
529
530 ext4_handle_error(inode->i_sb);
531}
532
c398eda0 533void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 534 unsigned int line, ext4_fsblk_t block,
535 const char *fmt, ...)
273df556
FM		 536{
537 va_list args;
f7c21177 538 struct va_format vaf;
1c13d5c0 539 struct ext4_super_block *es;
273df556
FM		 540 struct inode *inode = file->f_dentry->d_inode;
541 char pathname[80], *path;
542
1c13d5c0
TT		 543 es = EXT4_SB(inode->i_sb)->s_es;
544 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
545 save_error_info(inode->i_sb, function, line);
273df556 546 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 547 if (IS_ERR(path))
273df556 548 path = "(unknown)";
f7c21177
TT		 549 va_start(args, fmt);
550 vaf.fmt = fmt;
551 vaf.va = &args;
d9ee81da
JP		 552 if (block)
553 printk(KERN_CRIT
554 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
555 "block %llu: comm %s: path %s: %pV\n",
556 inode->i_sb->s_id, function, line, inode->i_ino,
557 block, current->comm, path, &vaf);
558 else
559 printk(KERN_CRIT
560 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
561 "comm %s: path %s: %pV\n",
562 inode->i_sb->s_id, function, line, inode->i_ino,
563 current->comm, path, &vaf);
273df556
FM		 564 va_end(args);
565
566 ext4_handle_error(inode->i_sb);
567}
568
2b2d6d01 569static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 570 char nbuf[16])
571{
572 char *errstr = NULL;
573
574 switch (errno) {
575 case -EIO:
576 errstr = "IO failure";
577 break;
578 case -ENOMEM:
579 errstr = "Out of memory";
580 break;
581 case -EROFS:
78f1ddbb
TT		 582 if (!sb || (EXT4_SB(sb)->s_journal &&
583 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 584 errstr = "Journal has aborted";
585 else
586 errstr = "Readonly filesystem";
587 break;
588 default:
589 /* If the caller passed in an extra buffer for unknown
590 * errors, textualise them now. Else we just return
591 * NULL. */
592 if (nbuf) {
593 /* Check for truncated error codes... */
594 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
595 errstr = nbuf;
596 }
597 break;
598 }
599
600 return errstr;
601}
602
617ba13b 603/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 604 * automatically and invokes the appropriate error response. */
605
c398eda0
TT		 606void __ext4_std_error(struct super_block *sb, const char *function,
607 unsigned int line, int errno)
ac27a0ec
DK		 608{
609 char nbuf[16];
610 const char *errstr;
611
612 /* Special case: if the error is EROFS, and we're not already
613 * inside a transaction, then there's really no point in logging
614 * an error. */
615 if (errno == -EROFS && journal_current_handle() == NULL &&
616 (sb->s_flags & MS_RDONLY))
617 return;
618
617ba13b 619 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 620 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
621 sb->s_id, function, line, errstr);
1c13d5c0 622 save_error_info(sb, function, line);
ac27a0ec 623
617ba13b 624 ext4_handle_error(sb);
ac27a0ec
DK		 625}
626
627/*
617ba13b 628 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 629 * abort function may be used to deal with unrecoverable failures such
630 * as journal IO errors or ENOMEM at a critical moment in log management.
631 *
632 * We unconditionally force the filesystem into an ABORT|READONLY state,
633 * unless the error response on the fs has been set to panic in which
634 * case we take the easy way out and panic immediately.
635 */
636
c67d859e 637void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 638 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 639{
640 va_list args;
641
1c13d5c0 642 save_error_info(sb, function, line);
ac27a0ec 643 va_start(args, fmt);
c398eda0
TT		 644 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
645 function, line);
ac27a0ec
DK		 646 vprintk(fmt, args);
647 printk("\n");
648 va_end(args);
649
1c13d5c0
TT		 650 if ((sb->s_flags & MS_RDONLY) == 0) {
651 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
652 sb->s_flags |= MS_RDONLY;
653 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
654 if (EXT4_SB(sb)->s_journal)
655 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
656 save_error_info(sb, function, line);
657 }
ac27a0ec 658 if (test_opt(sb, ERRORS_PANIC))
617ba13b 659 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 660}
661
0ff2ea7d 662void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 663{
0ff2ea7d 664 struct va_format vaf;
b31e1552
ES		 665 va_list args;
666
667 va_start(args, fmt);
0ff2ea7d
JP		 668 vaf.fmt = fmt;
669 vaf.va = &args;
670 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 671 va_end(args);
672}
673
12062ddd 674void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 675 unsigned int line, const char *fmt, ...)
ac27a0ec 676{
0ff2ea7d 677 struct va_format vaf;
ac27a0ec
DK		 678 va_list args;
679
680 va_start(args, fmt);
0ff2ea7d
JP		 681 vaf.fmt = fmt;
682 vaf.va = &args;
683 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
684 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 685 va_end(args);
686}
687
e29136f8
TT		 688void __ext4_grp_locked_error(const char *function, unsigned int line,
689 struct super_block *sb, ext4_group_t grp,
690 unsigned long ino, ext4_fsblk_t block,
691 const char *fmt, ...)
5d1b1b3f
AK		 692__releases(bitlock)
693__acquires(bitlock)
694{
0ff2ea7d 695 struct va_format vaf;
5d1b1b3f
AK		 696 va_list args;
697 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
698
1c13d5c0
TT		 699 es->s_last_error_ino = cpu_to_le32(ino);
700 es->s_last_error_block = cpu_to_le64(block);
701 __save_error_info(sb, function, line);
0ff2ea7d 702
5d1b1b3f 703 va_start(args, fmt);
0ff2ea7d
JP		 704
705 vaf.fmt = fmt;
706 vaf.va = &args;
21149d61 707 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 708 sb->s_id, function, line, grp);
709 if (ino)
0ff2ea7d 710 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 711 if (block)
0ff2ea7d
JP		 712 printk(KERN_CONT "block %llu:", (unsigned long long) block);
713 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 714 va_end(args);
715
716 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 717 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 718 return;
719 }
1c13d5c0 720
5d1b1b3f
AK		 721 ext4_unlock_group(sb, grp);
722 ext4_handle_error(sb);
723 /*
724 * We only get here in the ERRORS_RO case; relocking the group
725 * may be dangerous, but nothing bad will happen since the
726 * filesystem will have already been marked read/only and the
727 * journal has been aborted. We return 1 as a hint to callers
728 * who might what to use the return value from
25985edc 729 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 730 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
731 * aggressively from the ext4 function in question, with a
732 * more appropriate error code.
733 */
734 ext4_lock_group(sb, grp);
735 return;
736}
737
617ba13b 738void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 739{
617ba13b 740 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 741
617ba13b 742 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 743 return;
744
12062ddd 745 ext4_warning(sb,
ac27a0ec
DK		 746 "updating to rev %d because of new feature flag, "
747 "running e2fsck is recommended",
617ba13b 748 EXT4_DYNAMIC_REV);
ac27a0ec 749
617ba13b
MC		 750 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
751 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
752 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 753 /* leave es->s_feature_*compat flags alone */
754 /* es->s_uuid will be set by e2fsck if empty */
755
756 /*
757 * The rest of the superblock fields should be zero, and if not it
758 * means they are likely already in use, so leave them alone. We
759 * can leave it up to e2fsck to clean up any inconsistencies there.
760 */
761}
762
763/*
764 * Open the external journal device
765 */
b31e1552 766static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 767{
768 struct block_device *bdev;
769 char b[BDEVNAME_SIZE];
770
d4d77629 771 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 772 if (IS_ERR(bdev))
773 goto fail;
774 return bdev;
775
776fail:
b31e1552 777 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 778 __bdevname(dev, b), PTR_ERR(bdev));
779 return NULL;
780}
781
782/*
783 * Release the journal device
784 */
617ba13b 785static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 786{
e525fd89 787 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 788}
789
617ba13b 790static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 791{
792 struct block_device *bdev;
793 int ret = -ENODEV;
794
795 bdev = sbi->journal_bdev;
796 if (bdev) {
617ba13b 797 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 798 sbi->journal_bdev = NULL;
799 }
800 return ret;
801}
802
803static inline struct inode *orphan_list_entry(struct list_head *l)
804{
617ba13b 805 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 806}
807
617ba13b 808static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 809{
810 struct list_head *l;
811
b31e1552
ES		 812 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
813 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 814
815 printk(KERN_ERR "sb_info orphan list:\n");
816 list_for_each(l, &sbi->s_orphan) {
817 struct inode *inode = orphan_list_entry(l);
818 printk(KERN_ERR " "
819 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
820 inode->i_sb->s_id, inode->i_ino, inode,
821 inode->i_mode, inode->i_nlink,
822 NEXT_ORPHAN(inode));
823 }
824}
825
2b2d6d01 826static void ext4_put_super(struct super_block *sb)
ac27a0ec 827{
617ba13b
MC		 828 struct ext4_sb_info *sbi = EXT4_SB(sb);
829 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 830 int i, err;
ac27a0ec 831
857ac889 832 ext4_unregister_li_request(sb);
e0ccfd95
CH		 833 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
834
4c0425ff
MC		 835 flush_workqueue(sbi->dio_unwritten_wq);
836 destroy_workqueue(sbi->dio_unwritten_wq);
837
a9e220f8 838 lock_super(sb);
0390131b
FM		 839 if (sbi->s_journal) {
840 err = jbd2_journal_destroy(sbi->s_journal);
841 sbi->s_journal = NULL;
842 if (err < 0)
c67d859e 843 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 844 }
d4edac31 845
a1c6c569 846 del_timer(&sbi->s_err_report);
d4edac31
JB		 847 ext4_release_system_zone(sb);
848 ext4_mb_release(sb);
849 ext4_ext_release(sb);
850 ext4_xattr_put_super(sb);
851
ac27a0ec 852 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 853 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 854 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 855 }
a8e25a83
AB		 856 if (sb->s_dirt || !(sb->s_flags & MS_RDONLY))
857 ext4_commit_super(sb, 1);
858
240799cd 859 if (sbi->s_proc) {
66acdcf4 860 remove_proc_entry("options", sbi->s_proc);
9f6200bb 861 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 862 }
3197ebdb 863 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 864
865 for (i = 0; i < sbi->s_gdb_count; i++)
866 brelse(sbi->s_group_desc[i]);
f18a5f21 867 ext4_kvfree(sbi->s_group_desc);
9933fc0a 868 ext4_kvfree(sbi->s_flex_groups);
57042651 869 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 870 percpu_counter_destroy(&sbi->s_freeinodes_counter);
871 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 872 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 873 brelse(sbi->s_sbh);
874#ifdef CONFIG_QUOTA
875 for (i = 0; i < MAXQUOTAS; i++)
876 kfree(sbi->s_qf_names[i]);
877#endif
878
879 /* Debugging code just in case the in-memory inode orphan list
880 * isn't empty. The on-disk one can be non-empty if we've
881 * detected an error and taken the fs readonly, but the
882 * in-memory list had better be clean by this point. */
883 if (!list_empty(&sbi->s_orphan))
884 dump_orphan_list(sb, sbi);
885 J_ASSERT(list_empty(&sbi->s_orphan));
886
f98393a6 887 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 888 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
889 /*
890 * Invalidate the journal device's buffers. We don't want them
891 * floating about in memory - the physical journal device may
892 * hotswapped, and it breaks the `ro-after' testing code.
893 */
894 sync_blockdev(sbi->journal_bdev);
f98393a6 895 invalidate_bdev(sbi->journal_bdev);
617ba13b 896 ext4_blkdev_remove(sbi);
ac27a0ec 897 }
c5e06d10
JL		 898 if (sbi->s_mmp_tsk)
899 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 900 sb->s_fs_info = NULL;
3197ebdb
TT		 901 /*
902 * Now that we are completely done shutting down the
903 * superblock, we need to actually destroy the kobject.
904 */
3197ebdb
TT		 905 unlock_super(sb);
906 kobject_put(&sbi->s_kobj);
907 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 908 kfree(sbi->s_blockgroup_lock);
ac27a0ec 909 kfree(sbi);
ac27a0ec
DK		 910}
911
e18b890b 912static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 913
914/*
915 * Called inside transaction, so use GFP_NOFS
916 */
617ba13b 917static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 918{
617ba13b 919 struct ext4_inode_info *ei;
ac27a0ec 920
e6b4f8da 921 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 922 if (!ei)
923 return NULL;
0b8e58a1 924
ac27a0ec 925 ei->vfs_inode.i_version = 1;
91246c00 926 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 927 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 928 INIT_LIST_HEAD(&ei->i_prealloc_list);
929 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 930 ei->i_reserved_data_blocks = 0;
931 ei->i_reserved_meta_blocks = 0;
932 ei->i_allocated_meta_blocks = 0;
9d0be502 933 ei->i_da_metadata_calc_len = 0;
d2a17637 934 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 935#ifdef CONFIG_QUOTA
936 ei->i_reserved_quota = 0;
937#endif
8aefcd55 938 ei->jinode = NULL;
c7064ef1 939 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 940 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 941 ei->cur_aio_dio = NULL;
b436b9be
JK		 942 ei->i_sync_tid = 0;
943 ei->i_datasync_tid = 0;
f7ad6d2e 944 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 945 atomic_set(&ei->i_aiodio_unwritten, 0);
0b8e58a1 946
ac27a0ec
DK		 947 return &ei->vfs_inode;
948}
949
7ff9c073
TT		 950static int ext4_drop_inode(struct inode *inode)
951{
952 int drop = generic_drop_inode(inode);
953
954 trace_ext4_drop_inode(inode, drop);
955 return drop;
956}
957
fa0d7e3d
NP		 958static void ext4_i_callback(struct rcu_head *head)
959{
960 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 961 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
962}
963
617ba13b 964static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 965{
9f7dd93d 966 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 967 ext4_msg(inode->i_sb, KERN_ERR,
968 "Inode %lu (%p): orphan list check failed!",
969 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 970 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
971 EXT4_I(inode), sizeof(struct ext4_inode_info),
972 true);
973 dump_stack();
974 }
fa0d7e3d 975 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 976}
977
51cc5068 978static void init_once(void *foo)
ac27a0ec 979{
617ba13b 980 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 981
a35afb83 982 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 983#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 984 init_rwsem(&ei->xattr_sem);
ac27a0ec 985#endif
0e855ac8 986 init_rwsem(&ei->i_data_sem);
a35afb83 987 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 988}
989
990static int init_inodecache(void)
991{
617ba13b
MC		 992 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
993 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 994 0, (SLAB_RECLAIM_ACCOUNT|
995 SLAB_MEM_SPREAD),
20c2df83 996 init_once);
617ba13b 997 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 998 return -ENOMEM;
999 return 0;
1000}
1001
1002static void destroy_inodecache(void)
1003{
617ba13b 1004 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1005}
1006
0930fcc1 1007void ext4_clear_inode(struct inode *inode)
ac27a0ec 1008{
0930fcc1
AV		 1009 invalidate_inode_buffers(inode);
1010 end_writeback(inode);
9f754758 1011 dquot_drop(inode);
c2ea3fde 1012 ext4_discard_preallocations(inode);
8aefcd55
TT		 1013 if (EXT4_I(inode)->jinode) {
1014 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1015 EXT4_I(inode)->jinode);
1016 jbd2_free_inode(EXT4_I(inode)->jinode);
1017 EXT4_I(inode)->jinode = NULL;
1018 }
ac27a0ec
DK		 1019}
1020
1b961ac0 1021static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1022 u64 ino, u32 generation)
ac27a0ec 1023{
ac27a0ec 1024 struct inode *inode;
ac27a0ec 1025
617ba13b 1026 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1027 return ERR_PTR(-ESTALE);
617ba13b 1028 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1029 return ERR_PTR(-ESTALE);
1030
1031 /* iget isn't really right if the inode is currently unallocated!!
1032 *
617ba13b 1033 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1034 * deleted, so we should be safe.
1035 *
1036 * Currently we don't know the generation for parent directory, so
1037 * a generation of 0 means "accept any"
1038 */
1d1fe1ee
DH		 1039 inode = ext4_iget(sb, ino);
1040 if (IS_ERR(inode))
1041 return ERR_CAST(inode);
1042 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1043 iput(inode);
1044 return ERR_PTR(-ESTALE);
1045 }
1b961ac0
CH		 1046
1047 return inode;
1048}
1049
1050static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1051 int fh_len, int fh_type)
1b961ac0
CH		 1052{
1053 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1054 ext4_nfs_get_inode);
1055}
1056
1057static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1058 int fh_len, int fh_type)
1b961ac0
CH		 1059{
1060 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1061 ext4_nfs_get_inode);
ac27a0ec
DK		 1062}
1063
c39a7f84
TO		 1064/*
1065 * Try to release metadata pages (indirect blocks, directories) which are
1066 * mapped via the block device. Since these pages could have journal heads
1067 * which would prevent try_to_free_buffers() from freeing them, we must use
1068 * jbd2 layer's try_to_free_buffers() function to release them.
1069 */
0b8e58a1
AD		 1070static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1071 gfp_t wait)
c39a7f84
TO		 1072{
1073 journal_t *journal = EXT4_SB(sb)->s_journal;
1074
1075 WARN_ON(PageChecked(page));
1076 if (!page_has_buffers(page))
1077 return 0;
1078 if (journal)
1079 return jbd2_journal_try_to_free_buffers(journal, page,
1080 wait & ~__GFP_WAIT);
1081 return try_to_free_buffers(page);
1082}
1083
ac27a0ec 1084#ifdef CONFIG_QUOTA
af5bc92d 1085#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1086#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1087
617ba13b
MC		 1088static int ext4_write_dquot(struct dquot *dquot);
1089static int ext4_acquire_dquot(struct dquot *dquot);
1090static int ext4_release_dquot(struct dquot *dquot);
1091static int ext4_mark_dquot_dirty(struct dquot *dquot);
1092static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1093static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1094 struct path *path);
ca0e05e4 1095static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1096static int ext4_quota_on_mount(struct super_block *sb, int type);
1097static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1098 size_t len, loff_t off);
617ba13b 1099static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1100 const char *data, size_t len, loff_t off);
1101
61e225dc 1102static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1103 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1104 .write_dquot = ext4_write_dquot,
1105 .acquire_dquot = ext4_acquire_dquot,
1106 .release_dquot = ext4_release_dquot,
1107 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1108 .write_info = ext4_write_info,
1109 .alloc_dquot = dquot_alloc,
1110 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1111};
1112
0d54b217 1113static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1114 .quota_on = ext4_quota_on,
ca0e05e4 1115 .quota_off = ext4_quota_off,
287a8095
CH		 1116 .quota_sync = dquot_quota_sync,
1117 .get_info = dquot_get_dqinfo,
1118 .set_info = dquot_set_dqinfo,
1119 .get_dqblk = dquot_get_dqblk,
1120 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1121};
1122#endif
1123
ee9b6d61 1124static const struct super_operations ext4_sops = {
617ba13b
MC		 1125 .alloc_inode = ext4_alloc_inode,
1126 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1127 .write_inode = ext4_write_inode,
1128 .dirty_inode = ext4_dirty_inode,
7ff9c073 1129 .drop_inode = ext4_drop_inode,
0930fcc1 1130 .evict_inode = ext4_evict_inode,
617ba13b 1131 .put_super = ext4_put_super,
617ba13b 1132 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1133 .freeze_fs = ext4_freeze,
1134 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1135 .statfs = ext4_statfs,
1136 .remount_fs = ext4_remount,
617ba13b 1137 .show_options = ext4_show_options,
ac27a0ec 1138#ifdef CONFIG_QUOTA
617ba13b
MC		 1139 .quota_read = ext4_quota_read,
1140 .quota_write = ext4_quota_write,
ac27a0ec 1141#endif
c39a7f84 1142 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1143};
1144
9ca92389
TT		 1145static const struct super_operations ext4_nojournal_sops = {
1146 .alloc_inode = ext4_alloc_inode,
1147 .destroy_inode = ext4_destroy_inode,
1148 .write_inode = ext4_write_inode,
1149 .dirty_inode = ext4_dirty_inode,
7ff9c073 1150 .drop_inode = ext4_drop_inode,
0930fcc1 1151 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1152 .write_super = ext4_write_super,
1153 .put_super = ext4_put_super,
1154 .statfs = ext4_statfs,
1155 .remount_fs = ext4_remount,
9ca92389
TT		 1156 .show_options = ext4_show_options,
1157#ifdef CONFIG_QUOTA
1158 .quota_read = ext4_quota_read,
1159 .quota_write = ext4_quota_write,
1160#endif
1161 .bdev_try_to_free_page = bdev_try_to_free_page,
1162};
1163
39655164 1164static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1165 .fh_to_dentry = ext4_fh_to_dentry,
1166 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1167 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1168};
1169
1170enum {
1171 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1172 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1173 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1174 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1175 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1176 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1177 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1178 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1179 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1180 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1181 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1182 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1183 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1184 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1185 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1186 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1187 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1188 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
ac27a0ec
DK		 1189};
1190
a447c093 1191static const match_table_t tokens = {
ac27a0ec
DK		 1192 {Opt_bsd_df, "bsddf"},
1193 {Opt_minix_df, "minixdf"},
1194 {Opt_grpid, "grpid"},
1195 {Opt_grpid, "bsdgroups"},
1196 {Opt_nogrpid, "nogrpid"},
1197 {Opt_nogrpid, "sysvgroups"},
1198 {Opt_resgid, "resgid=%u"},
1199 {Opt_resuid, "resuid=%u"},
1200 {Opt_sb, "sb=%u"},
1201 {Opt_err_cont, "errors=continue"},
1202 {Opt_err_panic, "errors=panic"},
1203 {Opt_err_ro, "errors=remount-ro"},
1204 {Opt_nouid32, "nouid32"},
ac27a0ec 1205 {Opt_debug, "debug"},
72578c33
TT		 1206 {Opt_removed, "oldalloc"},
1207 {Opt_removed, "orlov"},
ac27a0ec
DK		 1208 {Opt_user_xattr, "user_xattr"},
1209 {Opt_nouser_xattr, "nouser_xattr"},
1210 {Opt_acl, "acl"},
1211 {Opt_noacl, "noacl"},
e3bb52ae 1212 {Opt_noload, "norecovery"},
5a916be1 1213 {Opt_noload, "noload"},
72578c33
TT		 1214 {Opt_removed, "nobh"},
1215 {Opt_removed, "bh"},
ac27a0ec 1216 {Opt_commit, "commit=%u"},
30773840
TT		 1217 {Opt_min_batch_time, "min_batch_time=%u"},
1218 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1219 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1220 {Opt_journal_checksum, "journal_checksum"},
1221 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1222 {Opt_abort, "abort"},
1223 {Opt_data_journal, "data=journal"},
1224 {Opt_data_ordered, "data=ordered"},
1225 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1226 {Opt_data_err_abort, "data_err=abort"},
1227 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1228 {Opt_offusrjquota, "usrjquota="},
1229 {Opt_usrjquota, "usrjquota=%s"},
1230 {Opt_offgrpjquota, "grpjquota="},
1231 {Opt_grpjquota, "grpjquota=%s"},
1232 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1233 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1234 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1235 {Opt_grpquota, "grpquota"},
1236 {Opt_noquota, "noquota"},
1237 {Opt_quota, "quota"},
1238 {Opt_usrquota, "usrquota"},
1239 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1240 {Opt_barrier, "barrier"},
1241 {Opt_nobarrier, "nobarrier"},
25ec56b5 1242 {Opt_i_version, "i_version"},
c9de560d 1243 {Opt_stripe, "stripe=%u"},
64769240 1244 {Opt_delalloc, "delalloc"},
dd919b98 1245 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1246 {Opt_mblk_io_submit, "mblk_io_submit"},
1247 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1248 {Opt_block_validity, "block_validity"},
1249 {Opt_noblock_validity, "noblock_validity"},
240799cd 1250 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1251 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1252 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1253 {Opt_auto_da_alloc, "auto_da_alloc"},
1254 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1255 {Opt_dioread_nolock, "dioread_nolock"},
1256 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1257 {Opt_discard, "discard"},
1258 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1259 {Opt_init_itable, "init_itable=%u"},
1260 {Opt_init_itable, "init_itable"},
1261 {Opt_noinit_itable, "noinit_itable"},
c7198b9c
TT		 1262 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1263 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1264 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1265 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1266 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1267 {Opt_err, NULL},
ac27a0ec
DK		 1268};
1269
617ba13b 1270static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1271{
617ba13b 1272 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1273 char *options = (char *) *data;
1274
1275 if (!options || strncmp(options, "sb=", 3) != 0)
1276 return 1; /* Default location */
0b8e58a1 1277
ac27a0ec 1278 options += 3;
0b8e58a1 1279 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1280 sb_block = simple_strtoul(options, &options, 0);
1281 if (*options && *options != ',') {
4776004f 1282 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1283 (char *) *data);
1284 return 1;
1285 }
1286 if (*options == ',')
1287 options++;
1288 *data = (void *) options;
0b8e58a1 1289
ac27a0ec
DK		 1290 return sb_block;
1291}
1292
b3881f74 1293#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1294static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1295 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1296
56c50f11
DM		 1297#ifdef CONFIG_QUOTA
1298static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1299{
1300 struct ext4_sb_info *sbi = EXT4_SB(sb);
1301 char *qname;
1302
1303 if (sb_any_quota_loaded(sb) &&
1304 !sbi->s_qf_names[qtype]) {
1305 ext4_msg(sb, KERN_ERR,
1306 "Cannot change journaled "
1307 "quota options when quota turned on");
57f73c2c 1308 return -1;
56c50f11
DM		 1309 }
1310 qname = match_strdup(args);
1311 if (!qname) {
1312 ext4_msg(sb, KERN_ERR,
1313 "Not enough memory for storing quotafile name");
57f73c2c 1314 return -1;
56c50f11
DM		 1315 }
1316 if (sbi->s_qf_names[qtype] &&
1317 strcmp(sbi->s_qf_names[qtype], qname)) {
1318 ext4_msg(sb, KERN_ERR,
1319 "%s quota file already specified", QTYPE2NAME(qtype));
1320 kfree(qname);
57f73c2c 1321 return -1;
56c50f11
DM		 1322 }
1323 sbi->s_qf_names[qtype] = qname;
1324 if (strchr(sbi->s_qf_names[qtype], '/')) {
1325 ext4_msg(sb, KERN_ERR,
1326 "quotafile must be on filesystem root");
1327 kfree(sbi->s_qf_names[qtype]);
1328 sbi->s_qf_names[qtype] = NULL;
57f73c2c 1329 return -1;
56c50f11 1330 }
fd8c37ec 1331 set_opt(sb, QUOTA);
56c50f11
DM		 1332 return 1;
1333}
1334
1335static int clear_qf_name(struct super_block *sb, int qtype)
1336{
1337
1338 struct ext4_sb_info *sbi = EXT4_SB(sb);
1339
1340 if (sb_any_quota_loaded(sb) &&
1341 sbi->s_qf_names[qtype]) {
1342 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1343 " when quota turned on");
57f73c2c 1344 return -1;
56c50f11
DM		 1345 }
1346 /*
1347 * The space will be released later when all options are confirmed
1348 * to be correct
1349 */
1350 sbi->s_qf_names[qtype] = NULL;
1351 return 1;
1352}
1353#endif
1354
26092bf5
TT		 1355#define MOPT_SET 0x0001
1356#define MOPT_CLEAR 0x0002
1357#define MOPT_NOSUPPORT 0x0004
1358#define MOPT_EXPLICIT 0x0008
1359#define MOPT_CLEAR_ERR 0x0010
1360#define MOPT_GTE0 0x0020
ac27a0ec 1361#ifdef CONFIG_QUOTA
26092bf5
TT		 1362#define MOPT_Q 0
1363#define MOPT_QFMT 0x0040
1364#else
1365#define MOPT_Q MOPT_NOSUPPORT
1366#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1367#endif
26092bf5
TT		 1368#define MOPT_DATAJ 0x0080
1369
1370static const struct mount_opts {
1371 int token;
1372 int mount_opt;
1373 int flags;
1374} ext4_mount_opts[] = {
1375 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1376 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1377 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1378 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1379 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1380 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1381 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1382 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1383 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1384 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1385 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1386 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1387 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1388 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1389 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1390 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1391 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1392 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1393 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1394 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1395 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1396 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1397 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1398 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1399 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1400 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1401 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1402 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1403 {Opt_commit, 0, MOPT_GTE0},
1404 {Opt_max_batch_time, 0, MOPT_GTE0},
1405 {Opt_min_batch_time, 0, MOPT_GTE0},
1406 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1407 {Opt_init_itable, 0, MOPT_GTE0},
1408 {Opt_stripe, 0, MOPT_GTE0},
1409 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1410 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1411 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
03010a33 1412#ifdef CONFIG_EXT4_FS_XATTR
26092bf5
TT		 1413 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1414 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
ac27a0ec 1415#else
26092bf5
TT		 1416 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1417 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
ac27a0ec 1418#endif
03010a33 1419#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1420 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1421 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1422#else
26092bf5
TT		 1423 {Opt_acl, 0, MOPT_NOSUPPORT},
1424 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1425#endif
26092bf5
TT		 1426 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1427 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1428 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1429 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1430 MOPT_SET | MOPT_Q},
1431 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1432 MOPT_SET | MOPT_Q},
1433 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1434 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1435 {Opt_usrjquota, 0, MOPT_Q},
1436 {Opt_grpjquota, 0, MOPT_Q},
1437 {Opt_offusrjquota, 0, MOPT_Q},
1438 {Opt_offgrpjquota, 0, MOPT_Q},
1439 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1440 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1441 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1442 {Opt_err, 0, 0}
1443};
1444
1445static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1446 substring_t *args, unsigned long *journal_devnum,
1447 unsigned int *journal_ioprio, int is_remount)
1448{
1449 struct ext4_sb_info *sbi = EXT4_SB(sb);
1450 const struct mount_opts *m;
1451 int arg = 0;
1452
57f73c2c
TT		 1453#ifdef CONFIG_QUOTA
1454 if (token == Opt_usrjquota)
1455 return set_qf_name(sb, USRQUOTA, &args[0]);
1456 else if (token == Opt_grpjquota)
1457 return set_qf_name(sb, GRPQUOTA, &args[0]);
1458 else if (token == Opt_offusrjquota)
1459 return clear_qf_name(sb, USRQUOTA);
1460 else if (token == Opt_offgrpjquota)
1461 return clear_qf_name(sb, GRPQUOTA);
1462#endif
26092bf5
TT		 1463 if (args->from && match_int(args, &arg))
1464 return -1;
1465 switch (token) {
f7048605
TT		 1466 case Opt_noacl:
1467 case Opt_nouser_xattr:
1468 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1469 break;
26092bf5
TT		 1470 case Opt_sb:
1471 return 1; /* handled by get_sb_block() */
1472 case Opt_removed:
1473 ext4_msg(sb, KERN_WARNING,
1474 "Ignoring removed %s option", opt);
1475 return 1;
1476 case Opt_resuid:
1477 sbi->s_resuid = arg;
1478 return 1;
1479 case Opt_resgid:
1480 sbi->s_resgid = arg;
1481 return 1;
1482 case Opt_abort:
1483 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1484 return 1;
1485 case Opt_i_version:
1486 sb->s_flags |= MS_I_VERSION;
1487 return 1;
1488 case Opt_journal_dev:
1489 if (is_remount) {
1490 ext4_msg(sb, KERN_ERR,
1491 "Cannot specify journal on remount");
1492 return -1;
1493 }
1494 *journal_devnum = arg;
1495 return 1;
1496 case Opt_journal_ioprio:
1497 if (arg < 0 || arg > 7)
1498 return -1;
1499 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1500 return 1;
1501 }
1502
1503 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1504 if (token != m->token)
1505 continue;
1506 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1507 return -1;
1508 if (m->flags & MOPT_EXPLICIT)
1509 set_opt2(sb, EXPLICIT_DELALLOC);
1510 if (m->flags & MOPT_CLEAR_ERR)
1511 clear_opt(sb, ERRORS_MASK);
1512 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1513 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1514 "options when quota turned on");
1515 return -1;
1516 }
1517
1518 if (m->flags & MOPT_NOSUPPORT) {
1519 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1520 } else if (token == Opt_commit) {
1521 if (arg == 0)
1522 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1523 sbi->s_commit_interval = HZ * arg;
1524 } else if (token == Opt_max_batch_time) {
1525 if (arg == 0)
1526 arg = EXT4_DEF_MAX_BATCH_TIME;
1527 sbi->s_max_batch_time = arg;
1528 } else if (token == Opt_min_batch_time) {
1529 sbi->s_min_batch_time = arg;
1530 } else if (token == Opt_inode_readahead_blks) {
1531 if (arg > (1 << 30))
1532 return -1;
1533 if (arg && !is_power_of_2(arg)) {
b31e1552 1534 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1535 "EXT4-fs: inode_readahead_blks"
1536 " must be a power of 2");
1537 return -1;
ac27a0ec 1538 }
26092bf5
TT		 1539 sbi->s_inode_readahead_blks = arg;
1540 } else if (token == Opt_init_itable) {
1541 set_opt(sb, INIT_INODE_TABLE);
1542 if (!args->from)
1543 arg = EXT4_DEF_LI_WAIT_MULT;
1544 sbi->s_li_wait_mult = arg;
1545 } else if (token == Opt_stripe) {
1546 sbi->s_stripe = arg;
1547 } else if (m->flags & MOPT_DATAJ) {
ac27a0ec 1548 if (is_remount) {
eb513689
TT		 1549 if (!sbi->s_journal)
1550 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
26092bf5
TT		 1551 else if (test_opt(sb, DATA_FLAGS) !=
1552 m->mount_opt) {
b31e1552 1553 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1554 "Cannot change data mode on remount");
1555 return -1;
ac27a0ec
DK		 1556 }
1557 } else {
fd8c37ec 1558 clear_opt(sb, DATA_FLAGS);
26092bf5 1559 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1560 }
ac27a0ec 1561#ifdef CONFIG_QUOTA
26092bf5 1562 } else if (m->flags & MOPT_QFMT) {
17bd13b3 1563 if (sb_any_quota_loaded(sb) &&
26092bf5
TT		 1564 sbi->s_jquota_fmt != m->mount_opt) {
1565 ext4_msg(sb, KERN_ERR, "Cannot "
1566 "change journaled quota options "
1567 "when quota turned on");
1568 return -1;
ac27a0ec 1569 }
26092bf5 1570 sbi->s_jquota_fmt = m->mount_opt;
ac27a0ec 1571#endif
26092bf5
TT		 1572 } else {
1573 if (!args->from)
1574 arg = 1;
1575 if (m->flags & MOPT_CLEAR)
1576 arg = !arg;
1577 else if (unlikely(!(m->flags & MOPT_SET))) {
1578 ext4_msg(sb, KERN_WARNING,
1579 "buggy handling of option %s", opt);
1580 WARN_ON(1);
1581 return -1;
3197ebdb 1582 }
26092bf5
TT		 1583 if (arg != 0)
1584 sbi->s_mount_opt |= m->mount_opt;
afd4672d 1585 else
26092bf5 1586 sbi->s_mount_opt &= ~m->mount_opt;
ac27a0ec 1587 }
26092bf5
TT		 1588 return 1;
1589 }
1590 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1591 "or missing value", opt);
1592 return -1;
1593}
1594
1595static int parse_options(char *options, struct super_block *sb,
1596 unsigned long *journal_devnum,
1597 unsigned int *journal_ioprio,
1598 int is_remount)
1599{
1600 struct ext4_sb_info *sbi = EXT4_SB(sb);
1601 char *p;
1602 substring_t args[MAX_OPT_ARGS];
1603 int token;
1604
1605 if (!options)
1606 return 1;
1607
1608 while ((p = strsep(&options, ",")) != NULL) {
1609 if (!*p)
1610 continue;
1611 /*
1612 * Initialize args struct so we know whether arg was
1613 * found; some options take optional arguments.
1614 */
1615 args[0].to = args[0].from = 0;
1616 token = match_token(p, tokens, args);
1617 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1618 journal_ioprio, is_remount) < 0)
1619 return 0;
ac27a0ec
DK		 1620 }
1621#ifdef CONFIG_QUOTA
1622 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1623 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1624 clear_opt(sb, USRQUOTA);
ac27a0ec 1625
482a7425 1626 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1627 clear_opt(sb, GRPQUOTA);
ac27a0ec 1628
56c50f11 1629 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1630 ext4_msg(sb, KERN_ERR, "old and new quota "
1631 "format mixing");
ac27a0ec
DK		 1632 return 0;
1633 }
1634
1635 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1636 ext4_msg(sb, KERN_ERR, "journaled quota format "
1637 "not specified");
ac27a0ec
DK		 1638 return 0;
1639 }
1640 } else {
1641 if (sbi->s_jquota_fmt) {
b31e1552 1642 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1643 "specified with no journaling "
b31e1552 1644 "enabled");
ac27a0ec
DK		 1645 return 0;
1646 }
1647 }
1648#endif
1649 return 1;
1650}
1651
2adf6da8
TT		 1652static inline void ext4_show_quota_options(struct seq_file *seq,
1653 struct super_block *sb)
1654{
1655#if defined(CONFIG_QUOTA)
1656 struct ext4_sb_info *sbi = EXT4_SB(sb);
1657
1658 if (sbi->s_jquota_fmt) {
1659 char *fmtname = "";
1660
1661 switch (sbi->s_jquota_fmt) {
1662 case QFMT_VFS_OLD:
1663 fmtname = "vfsold";
1664 break;
1665 case QFMT_VFS_V0:
1666 fmtname = "vfsv0";
1667 break;
1668 case QFMT_VFS_V1:
1669 fmtname = "vfsv1";
1670 break;
1671 }
1672 seq_printf(seq, ",jqfmt=%s", fmtname);
1673 }
1674
1675 if (sbi->s_qf_names[USRQUOTA])
1676 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1677
1678 if (sbi->s_qf_names[GRPQUOTA])
1679 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1680
1681 if (test_opt(sb, USRQUOTA))
1682 seq_puts(seq, ",usrquota");
1683
1684 if (test_opt(sb, GRPQUOTA))
1685 seq_puts(seq, ",grpquota");
1686#endif
1687}
1688
5a916be1
TT		 1689static const char *token2str(int token)
1690{
1691 static const struct match_token *t;
1692
1693 for (t = tokens; t->token != Opt_err; t++)
1694 if (t->token == token && !strchr(t->pattern, '='))
1695 break;
1696 return t->pattern;
1697}
1698
2adf6da8
TT		 1699/*
1700 * Show an option if
1701 * - it's set to a non-default value OR
1702 * - if the per-sb default is different from the global default
1703 */
66acdcf4
TT		 1704static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1705 int nodefs)
2adf6da8 1706{
2adf6da8
TT		 1707 struct ext4_sb_info *sbi = EXT4_SB(sb);
1708 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1709 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1710 const struct mount_opts *m;
66acdcf4 1711 char sep = nodefs ? '\n' : ',';
2adf6da8 1712
66acdcf4
TT		 1713#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1714#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1715
1716 if (sbi->s_sb_block != 1)
5a916be1
TT		 1717 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1718
1719 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1720 int want_set = m->flags & MOPT_SET;
1721 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1722 (m->flags & MOPT_CLEAR_ERR))
1723 continue;
66acdcf4 1724 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1725 continue; /* skip if same as the default */
1726 if ((want_set &&
1727 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1728 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1729 continue; /* select Opt_noFoo vs Opt_Foo */
1730 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1731 }
5a916be1 1732
66acdcf4 1733 if (nodefs || sbi->s_resuid != EXT4_DEF_RESUID ||
5a916be1
TT		 1734 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1735 SEQ_OPTS_PRINT("resuid=%u", sbi->s_resuid);
66acdcf4 1736 if (nodefs || sbi->s_resgid != EXT4_DEF_RESGID ||
5a916be1
TT		 1737 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1738 SEQ_OPTS_PRINT("resgid=%u", sbi->s_resgid);
66acdcf4 1739 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1740 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1741 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1742 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1743 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1744 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1745 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1746 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1747 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1748 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1749 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1750 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1751 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1752 if (sb->s_flags & MS_I_VERSION)
5a916be1 1753 SEQ_OPTS_PUTS("i_version");
66acdcf4 1754 if (nodefs || sbi->s_stripe)
5a916be1 1755 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1756 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1757 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1758 SEQ_OPTS_PUTS("data=journal");
1759 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1760 SEQ_OPTS_PUTS("data=ordered");
1761 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1762 SEQ_OPTS_PUTS("data=writeback");
1763 }
66acdcf4
TT		 1764 if (nodefs ||
1765 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1766 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1767 sbi->s_inode_readahead_blks);
2adf6da8 1768
66acdcf4
TT		 1769 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1770 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1771 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2adf6da8
TT		 1772
1773 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1774 return 0;
1775}
1776
66acdcf4
TT		 1777static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1778{
1779 return _ext4_show_options(seq, root->d_sb, 0);
1780}
1781
1782static int options_seq_show(struct seq_file *seq, void *offset)
1783{
1784 struct super_block *sb = seq->private;
1785 int rc;
1786
1787 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1788 rc = _ext4_show_options(seq, sb, 1);
1789 seq_puts(seq, "\n");
1790 return rc;
1791}
1792
1793static int options_open_fs(struct inode *inode, struct file *file)
1794{
1795 return single_open(file, options_seq_show, PDE(inode)->data);
1796}
1797
1798static const struct file_operations ext4_seq_options_fops = {
1799 .owner = THIS_MODULE,
1800 .open = options_open_fs,
1801 .read = seq_read,
1802 .llseek = seq_lseek,
1803 .release = single_release,
1804};
1805
617ba13b 1806static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1807 int read_only)
1808{
617ba13b 1809 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1810 int res = 0;
1811
617ba13b 1812 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1813 ext4_msg(sb, KERN_ERR, "revision level too high, "
1814 "forcing read-only mode");
ac27a0ec
DK		 1815 res = MS_RDONLY;
1816 }
1817 if (read_only)
281b5995 1818 goto done;
617ba13b 1819 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1820 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1821 "running e2fsck is recommended");
617ba13b 1822 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1823 ext4_msg(sb, KERN_WARNING,
1824 "warning: mounting fs with errors, "
1825 "running e2fsck is recommended");
ed3ce80a 1826 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1827 le16_to_cpu(es->s_mnt_count) >=
1828 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1829 ext4_msg(sb, KERN_WARNING,
1830 "warning: maximal mount count reached, "
1831 "running e2fsck is recommended");
ac27a0ec
DK		 1832 else if (le32_to_cpu(es->s_checkinterval) &&
1833 (le32_to_cpu(es->s_lastcheck) +
1834 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1835 ext4_msg(sb, KERN_WARNING,
1836 "warning: checktime reached, "
1837 "running e2fsck is recommended");
0b8e58a1 1838 if (!sbi->s_journal)
0390131b 1839 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1840 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1841 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1842 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1843 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1844 ext4_update_dynamic_rev(sb);
0390131b
FM		 1845 if (sbi->s_journal)
1846 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1847
e2d67052 1848 ext4_commit_super(sb, 1);
281b5995 1849done:
ac27a0ec 1850 if (test_opt(sb, DEBUG))
a9df9a49 1851 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1852 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1853 sb->s_blocksize,
1854 sbi->s_groups_count,
617ba13b
MC		 1855 EXT4_BLOCKS_PER_GROUP(sb),
1856 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1857 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1858
7abc52c2 1859 cleancache_init_fs(sb);
ac27a0ec
DK		 1860 return res;
1861}
1862
772cb7c8
JS		 1863static int ext4_fill_flex_info(struct super_block *sb)
1864{
1865 struct ext4_sb_info *sbi = EXT4_SB(sb);
1866 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1867 ext4_group_t flex_group_count;
1868 ext4_group_t flex_group;
d50f2ab6 1869 unsigned int groups_per_flex = 0;
c5ca7c76 1870 size_t size;
772cb7c8
JS		 1871 int i;
1872
503358ae 1873 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1874 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1875 sbi->s_log_groups_per_flex = 0;
1876 return 1;
1877 }
d50f2ab6 1878 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
772cb7c8 1879
c62a11fd
FB		 1880 /* We allocate both existing and potentially added groups */
1881 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1882 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1883 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76 1884 size = flex_group_count * sizeof(struct flex_groups);
9933fc0a 1885 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
c5ca7c76 1886 if (sbi->s_flex_groups == NULL) {
9933fc0a
TT		 1887 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1888 flex_group_count);
1889 goto failed;
772cb7c8 1890 }
772cb7c8 1891
772cb7c8 1892 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1893 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1894
1895 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1896 atomic_add(ext4_free_inodes_count(sb, gdp),
1897 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 1898 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 1899 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1900 atomic_add(ext4_used_dirs_count(sb, gdp),
1901 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1902 }
1903
1904 return 1;
1905failed:
1906 return 0;
1907}
1908
717d50e4
AD		 1909__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1910 struct ext4_group_desc *gdp)
1911{
1912 __u16 crc = 0;
1913
1914 if (sbi->s_es->s_feature_ro_compat &
1915 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1916 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1917 __le32 le_group = cpu_to_le32(block_group);
1918
1919 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1920 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1921 crc = crc16(crc, (__u8 *)gdp, offset);
1922 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1923 /* for checksum of struct ext4_group_desc do the rest...*/
1924 if ((sbi->s_es->s_feature_incompat &
1925 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1926 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1927 crc = crc16(crc, (__u8 *)gdp + offset,
1928 le16_to_cpu(sbi->s_es->s_desc_size) -
1929 offset);
1930 }
1931
1932 return cpu_to_le16(crc);
1933}
1934
1935int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1936 struct ext4_group_desc *gdp)
1937{
1938 if ((sbi->s_es->s_feature_ro_compat &
1939 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1940 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1941 return 0;
1942
1943 return 1;
1944}
1945
ac27a0ec 1946/* Called at mount-time, super-block is locked */
bfff6873
LC		 1947static int ext4_check_descriptors(struct super_block *sb,
1948 ext4_group_t *first_not_zeroed)
ac27a0ec 1949{
617ba13b
MC		 1950 struct ext4_sb_info *sbi = EXT4_SB(sb);
1951 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1952 ext4_fsblk_t last_block;
bd81d8ee
LV		 1953 ext4_fsblk_t block_bitmap;
1954 ext4_fsblk_t inode_bitmap;
1955 ext4_fsblk_t inode_table;
ce421581 1956 int flexbg_flag = 0;
bfff6873 1957 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 1958
ce421581
JS		 1959 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1960 flexbg_flag = 1;
1961
af5bc92d 1962 ext4_debug("Checking group descriptors");
ac27a0ec 1963
197cd65a
AM		 1964 for (i = 0; i < sbi->s_groups_count; i++) {
1965 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1966
ce421581 1967 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1968 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1969 else
1970 last_block = first_block +
617ba13b 1971 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1972
bfff6873
LC		 1973 if ((grp == sbi->s_groups_count) &&
1974 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
1975 grp = i;
1976
8fadc143 1977 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1978 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 1979 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1980 "Block bitmap for group %u not in group "
b31e1552 1981 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1982 return 0;
1983 }
8fadc143 1984 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1985 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 1986 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1987 "Inode bitmap for group %u not in group "
b31e1552 1988 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1989 return 0;
1990 }
8fadc143 1991 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1992 if (inode_table < first_block ||
2b2d6d01 1993 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 1994 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1995 "Inode table for group %u not in group "
b31e1552 1996 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1997 return 0;
1998 }
955ce5f5 1999 ext4_lock_group(sb, i);
717d50e4 2000 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 2001 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2002 "Checksum for group %u failed (%u!=%u)",
2003 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2004 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2005 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2006 ext4_unlock_group(sb, i);
8a266467 2007 return 0;
7ee1ec4c 2008 }
717d50e4 2009 }
955ce5f5 2010 ext4_unlock_group(sb, i);
ce421581
JS		 2011 if (!flexbg_flag)
2012 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2013 }
bfff6873
LC		 2014 if (NULL != first_not_zeroed)
2015 *first_not_zeroed = grp;
ac27a0ec 2016
5dee5437
TT		 2017 ext4_free_blocks_count_set(sbi->s_es,
2018 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2019 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2020 return 1;
2021}
2022
617ba13b 2023/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2024 * the superblock) which were deleted from all directories, but held open by
2025 * a process at the time of a crash. We walk the list and try to delete these
2026 * inodes at recovery time (only with a read-write filesystem).
2027 *
2028 * In order to keep the orphan inode chain consistent during traversal (in
2029 * case of crash during recovery), we link each inode into the superblock
2030 * orphan list_head and handle it the same way as an inode deletion during
2031 * normal operation (which journals the operations for us).
2032 *
2033 * We only do an iget() and an iput() on each inode, which is very safe if we
2034 * accidentally point at an in-use or already deleted inode. The worst that
2035 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2036 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2037 * e2fsck was run on this filesystem, and it must have already done the orphan
2038 * inode cleanup for us, so we can safely abort without any further action.
2039 */
2b2d6d01
TT		 2040static void ext4_orphan_cleanup(struct super_block *sb,
2041 struct ext4_super_block *es)
ac27a0ec
DK		 2042{
2043 unsigned int s_flags = sb->s_flags;
2044 int nr_orphans = 0, nr_truncates = 0;
2045#ifdef CONFIG_QUOTA
2046 int i;
2047#endif
2048 if (!es->s_last_orphan) {
2049 jbd_debug(4, "no orphan inodes to clean up\n");
2050 return;
2051 }
2052
a8f48a95 2053 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2054 ext4_msg(sb, KERN_ERR, "write access "
2055 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2056 return;
2057 }
2058
d39195c3
AG		 2059 /* Check if feature set would not allow a r/w mount */
2060 if (!ext4_feature_set_ok(sb, 0)) {
2061 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2062 "unknown ROCOMPAT features");
2063 return;
2064 }
2065
617ba13b 2066 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2067 if (es->s_last_orphan)
2068 jbd_debug(1, "Errors on filesystem, "
2069 "clearing orphan list.\n");
2070 es->s_last_orphan = 0;
2071 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2072 return;
2073 }
2074
2075 if (s_flags & MS_RDONLY) {
b31e1552 2076 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2077 sb->s_flags &= ~MS_RDONLY;
2078 }
2079#ifdef CONFIG_QUOTA
2080 /* Needed for iput() to work correctly and not trash data */
2081 sb->s_flags |= MS_ACTIVE;
2082 /* Turn on quotas so that they are updated correctly */
2083 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2084 if (EXT4_SB(sb)->s_qf_names[i]) {
2085 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2086 if (ret < 0)
b31e1552
ES		 2087 ext4_msg(sb, KERN_ERR,
2088 "Cannot turn on journaled "
2089 "quota: error %d", ret);
ac27a0ec
DK		 2090 }
2091 }
2092#endif
2093
2094 while (es->s_last_orphan) {
2095 struct inode *inode;
2096
97bd42b9
JB		 2097 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2098 if (IS_ERR(inode)) {
ac27a0ec
DK		 2099 es->s_last_orphan = 0;
2100 break;
2101 }
2102
617ba13b 2103 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2104 dquot_initialize(inode);
ac27a0ec 2105 if (inode->i_nlink) {
b31e1552
ES		 2106 ext4_msg(sb, KERN_DEBUG,
2107 "%s: truncating inode %lu to %lld bytes",
46e665e9 2108 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2109 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2110 inode->i_ino, inode->i_size);
617ba13b 2111 ext4_truncate(inode);
ac27a0ec
DK		 2112 nr_truncates++;
2113 } else {
b31e1552
ES		 2114 ext4_msg(sb, KERN_DEBUG,
2115 "%s: deleting unreferenced inode %lu",
46e665e9 2116 __func__, inode->i_ino);
ac27a0ec
DK		 2117 jbd_debug(2, "deleting unreferenced inode %lu\n",
2118 inode->i_ino);
2119 nr_orphans++;
2120 }
2121 iput(inode); /* The delete magic happens here! */
2122 }
2123
2b2d6d01 2124#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2125
2126 if (nr_orphans)
b31e1552
ES		 2127 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2128 PLURAL(nr_orphans));
ac27a0ec 2129 if (nr_truncates)
b31e1552
ES		 2130 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2131 PLURAL(nr_truncates));
ac27a0ec
DK		 2132#ifdef CONFIG_QUOTA
2133 /* Turn quotas off */
2134 for (i = 0; i < MAXQUOTAS; i++) {
2135 if (sb_dqopt(sb)->files[i])
287a8095 2136 dquot_quota_off(sb, i);
ac27a0ec
DK		 2137 }
2138#endif
2139 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2140}
0b8e58a1 2141
cd2291a4
ES		 2142/*
2143 * Maximal extent format file size.
2144 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2145 * extent format containers, within a sector_t, and within i_blocks
2146 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2147 * so that won't be a limiting factor.
2148 *
f17722f9
LC		 2149 * However there is other limiting factor. We do store extents in the form
2150 * of starting block and length, hence the resulting length of the extent
2151 * covering maximum file size must fit into on-disk format containers as
2152 * well. Given that length is always by 1 unit bigger than max unit (because
2153 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2154 *
cd2291a4
ES		 2155 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2156 */
f287a1a5 2157static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2158{
2159 loff_t res;
2160 loff_t upper_limit = MAX_LFS_FILESIZE;
2161
2162 /* small i_blocks in vfs inode? */
f287a1a5 2163 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2164 /*
90c699a9 2165 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2166 * i_block represent total blocks in 512 bytes
2167 * 32 == size of vfs inode i_blocks * 8
2168 */
2169 upper_limit = (1LL << 32) - 1;
2170
2171 /* total blocks in file system block size */
2172 upper_limit >>= (blkbits - 9);
2173 upper_limit <<= blkbits;
2174 }
2175
f17722f9
LC		 2176 /*
2177 * 32-bit extent-start container, ee_block. We lower the maxbytes
2178 * by one fs block, so ee_len can cover the extent of maximum file
2179 * size
2180 */
2181 res = (1LL << 32) - 1;
cd2291a4 2182 res <<= blkbits;
cd2291a4
ES		 2183
2184 /* Sanity check against vm- & vfs- imposed limits */
2185 if (res > upper_limit)
2186 res = upper_limit;
2187
2188 return res;
2189}
ac27a0ec 2190
ac27a0ec 2191/*
cd2291a4 2192 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2193 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2194 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2195 */
f287a1a5 2196static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2197{
617ba13b 2198 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2199 int meta_blocks;
2200 loff_t upper_limit;
0b8e58a1
AD		 2201 /* This is calculated to be the largest file size for a dense, block
2202 * mapped file such that the file's total number of 512-byte sectors,
2203 * including data and all indirect blocks, does not exceed (2^48 - 1).
2204 *
2205 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2206 * number of 512-byte sectors of the file.
0fc1b451
AK		 2207 */
2208
f287a1a5 2209 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2210 /*
90c699a9 2211 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2212 * the inode i_block field represents total file blocks in
2213 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2214 */
2215 upper_limit = (1LL << 32) - 1;
2216
2217 /* total blocks in file system block size */
2218 upper_limit >>= (bits - 9);
2219
2220 } else {
8180a562
AK		 2221 /*
2222 * We use 48 bit ext4_inode i_blocks
2223 * With EXT4_HUGE_FILE_FL set the i_blocks
2224 * represent total number of blocks in
2225 * file system block size
2226 */
0fc1b451
AK		 2227 upper_limit = (1LL << 48) - 1;
2228
0fc1b451
AK		 2229 }
2230
2231 /* indirect blocks */
2232 meta_blocks = 1;
2233 /* double indirect blocks */
2234 meta_blocks += 1 + (1LL << (bits-2));
2235 /* tripple indirect blocks */
2236 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2237
2238 upper_limit -= meta_blocks;
2239 upper_limit <<= bits;
ac27a0ec
DK		 2240
2241 res += 1LL << (bits-2);
2242 res += 1LL << (2*(bits-2));
2243 res += 1LL << (3*(bits-2));
2244 res <<= bits;
2245 if (res > upper_limit)
2246 res = upper_limit;
0fc1b451
AK		 2247
2248 if (res > MAX_LFS_FILESIZE)
2249 res = MAX_LFS_FILESIZE;
2250
ac27a0ec
DK		 2251 return res;
2252}
2253
617ba13b 2254static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2255 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2256{
617ba13b 2257 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2258 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2259 int has_super = 0;
2260
2261 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2262
617ba13b 2263 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2264 nr < first_meta_bg)
70bbb3e0 2265 return logical_sb_block + nr + 1;
ac27a0ec 2266 bg = sbi->s_desc_per_block * nr;
617ba13b 2267 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2268 has_super = 1;
0b8e58a1 2269
617ba13b 2270 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2271}
2272
c9de560d
AT		 2273/**
2274 * ext4_get_stripe_size: Get the stripe size.
2275 * @sbi: In memory super block info
2276 *
2277 * If we have specified it via mount option, then
2278 * use the mount option value. If the value specified at mount time is
2279 * greater than the blocks per group use the super block value.
2280 * If the super block value is greater than blocks per group return 0.
2281 * Allocator needs it be less than blocks per group.
2282 *
2283 */
2284static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2285{
2286 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2287 unsigned long stripe_width =
2288 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2289 int ret;
c9de560d
AT		 2290
2291 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2292 ret = sbi->s_stripe;
2293 else if (stripe_width <= sbi->s_blocks_per_group)
2294 ret = stripe_width;
2295 else if (stride <= sbi->s_blocks_per_group)
2296 ret = stride;
2297 else
2298 ret = 0;
c9de560d 2299
3eb08658
DE		 2300 /*
2301 * If the stripe width is 1, this makes no sense and
2302 * we set it to 0 to turn off stripe handling code.
2303 */
2304 if (ret <= 1)
2305 ret = 0;
c9de560d 2306
3eb08658 2307 return ret;
c9de560d 2308}
ac27a0ec 2309
3197ebdb
TT		 2310/* sysfs supprt */
2311
2312struct ext4_attr {
2313 struct attribute attr;
2314 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2315 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2316 const char *, size_t);
2317 int offset;
2318};
2319
2320static int parse_strtoul(const char *buf,
2321 unsigned long max, unsigned long *value)
2322{
2323 char *endp;
2324
e7d2860b
AGR		 2325 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2326 endp = skip_spaces(endp);
3197ebdb
TT		 2327 if (*endp || *value > max)
2328 return -EINVAL;
2329
2330 return 0;
2331}
2332
2333static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2334 struct ext4_sb_info *sbi,
2335 char *buf)
2336{
2337 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2338 (s64) EXT4_C2B(sbi,
2339 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2340}
2341
2342static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2343 struct ext4_sb_info *sbi, char *buf)
2344{
2345 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2346
f613dfcb
TT		 2347 if (!sb->s_bdev->bd_part)
2348 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2349 return snprintf(buf, PAGE_SIZE, "%lu\n",
2350 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2351 sbi->s_sectors_written_start) >> 1);
2352}
2353
2354static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2355 struct ext4_sb_info *sbi, char *buf)
2356{
2357 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2358
f613dfcb
TT		 2359 if (!sb->s_bdev->bd_part)
2360 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2361 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2362 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2363 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2364 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2365}
2366
2367static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2368 struct ext4_sb_info *sbi,
2369 const char *buf, size_t count)
2370{
2371 unsigned long t;
2372
2373 if (parse_strtoul(buf, 0x40000000, &t))
2374 return -EINVAL;
2375
5dbd571d 2376 if (t && !is_power_of_2(t))
3197ebdb
TT		 2377 return -EINVAL;
2378
2379 sbi->s_inode_readahead_blks = t;
2380 return count;
2381}
2382
2383static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2384 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2385{
2386 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2387
2388 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2389}
2390
2391static ssize_t sbi_ui_store(struct ext4_attr *a,
2392 struct ext4_sb_info *sbi,
2393 const char *buf, size_t count)
2394{
2395 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2396 unsigned long t;
2397
2398 if (parse_strtoul(buf, 0xffffffff, &t))
2399 return -EINVAL;
2400 *ui = t;
2401 return count;
2402}
2403
2404#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2405static struct ext4_attr ext4_attr_##_name = { \
2406 .attr = {.name = __stringify(_name), .mode = _mode }, \
2407 .show = _show, \
2408 .store = _store, \
2409 .offset = offsetof(struct ext4_sb_info, _elname), \
2410}
2411#define EXT4_ATTR(name, mode, show, store) \
2412static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2413
857ac889 2414#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2415#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2416#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2417#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2418 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2419#define ATTR_LIST(name) &ext4_attr_##name.attr
2420
2421EXT4_RO_ATTR(delayed_allocation_blocks);
2422EXT4_RO_ATTR(session_write_kbytes);
2423EXT4_RO_ATTR(lifetime_write_kbytes);
2424EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2425 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2426EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2427EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2428EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2429EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2430EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2431EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2432EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2433EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2434
2435static struct attribute *ext4_attrs[] = {
2436 ATTR_LIST(delayed_allocation_blocks),
2437 ATTR_LIST(session_write_kbytes),
2438 ATTR_LIST(lifetime_write_kbytes),
2439 ATTR_LIST(inode_readahead_blks),
11013911 2440 ATTR_LIST(inode_goal),
3197ebdb
TT		 2441 ATTR_LIST(mb_stats),
2442 ATTR_LIST(mb_max_to_scan),
2443 ATTR_LIST(mb_min_to_scan),
2444 ATTR_LIST(mb_order2_req),
2445 ATTR_LIST(mb_stream_req),
2446 ATTR_LIST(mb_group_prealloc),
55138e0b 2447 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2448 NULL,
2449};
2450
857ac889
LC		 2451/* Features this copy of ext4 supports */
2452EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2453EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2454
2455static struct attribute *ext4_feat_attrs[] = {
2456 ATTR_LIST(lazy_itable_init),
27ee40df 2457 ATTR_LIST(batched_discard),
857ac889
LC		 2458 NULL,
2459};
2460
3197ebdb
TT		 2461static ssize_t ext4_attr_show(struct kobject *kobj,
2462 struct attribute *attr, char *buf)
2463{
2464 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2465 s_kobj);
2466 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2467
2468 return a->show ? a->show(a, sbi, buf) : 0;
2469}
2470
2471static ssize_t ext4_attr_store(struct kobject *kobj,
2472 struct attribute *attr,
2473 const char *buf, size_t len)
2474{
2475 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2476 s_kobj);
2477 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2478
2479 return a->store ? a->store(a, sbi, buf, len) : 0;
2480}
2481
2482static void ext4_sb_release(struct kobject *kobj)
2483{
2484 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2485 s_kobj);
2486 complete(&sbi->s_kobj_unregister);
2487}
2488
52cf25d0 2489static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2490 .show = ext4_attr_show,
2491 .store = ext4_attr_store,
2492};
2493
2494static struct kobj_type ext4_ktype = {
2495 .default_attrs = ext4_attrs,
2496 .sysfs_ops = &ext4_attr_ops,
2497 .release = ext4_sb_release,
2498};
2499
857ac889
LC		 2500static void ext4_feat_release(struct kobject *kobj)
2501{
2502 complete(&ext4_feat->f_kobj_unregister);
2503}
2504
2505static struct kobj_type ext4_feat_ktype = {
2506 .default_attrs = ext4_feat_attrs,
2507 .sysfs_ops = &ext4_attr_ops,
2508 .release = ext4_feat_release,
2509};
2510
a13fb1a4
ES		 2511/*
2512 * Check whether this filesystem can be mounted based on
2513 * the features present and the RDONLY/RDWR mount requested.
2514 * Returns 1 if this filesystem can be mounted as requested,
2515 * 0 if it cannot be.
2516 */
2517static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2518{
2519 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2520 ext4_msg(sb, KERN_ERR,
2521 "Couldn't mount because of "
2522 "unsupported optional features (%x)",
2523 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2524 ~EXT4_FEATURE_INCOMPAT_SUPP));
2525 return 0;
2526 }
2527
2528 if (readonly)
2529 return 1;
2530
2531 /* Check that feature set is OK for a read-write mount */
2532 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2533 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2534 "unsupported optional features (%x)",
2535 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2536 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2537 return 0;
2538 }
2539 /*
2540 * Large file size enabled file system can only be mounted
2541 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2542 */
2543 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2544 if (sizeof(blkcnt_t) < sizeof(u64)) {
2545 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2546 "cannot be mounted RDWR without "
2547 "CONFIG_LBDAF");
2548 return 0;
2549 }
2550 }
bab08ab9
TT		 2551 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2552 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2553 ext4_msg(sb, KERN_ERR,
2554 "Can't support bigalloc feature without "
2555 "extents feature\n");
2556 return 0;
2557 }
a13fb1a4
ES		 2558 return 1;
2559}
2560
66e61a9e
TT		 2561/*
2562 * This function is called once a day if we have errors logged
2563 * on the file system
2564 */
2565static void print_daily_error_info(unsigned long arg)
2566{
2567 struct super_block *sb = (struct super_block *) arg;
2568 struct ext4_sb_info *sbi;
2569 struct ext4_super_block *es;
2570
2571 sbi = EXT4_SB(sb);
2572 es = sbi->s_es;
2573
2574 if (es->s_error_count)
2575 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2576 le32_to_cpu(es->s_error_count));
2577 if (es->s_first_error_time) {
2578 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2579 sb->s_id, le32_to_cpu(es->s_first_error_time),
2580 (int) sizeof(es->s_first_error_func),
2581 es->s_first_error_func,
2582 le32_to_cpu(es->s_first_error_line));
2583 if (es->s_first_error_ino)
2584 printk(": inode %u",
2585 le32_to_cpu(es->s_first_error_ino));
2586 if (es->s_first_error_block)
2587 printk(": block %llu", (unsigned long long)
2588 le64_to_cpu(es->s_first_error_block));
2589 printk("\n");
2590 }
2591 if (es->s_last_error_time) {
2592 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2593 sb->s_id, le32_to_cpu(es->s_last_error_time),
2594 (int) sizeof(es->s_last_error_func),
2595 es->s_last_error_func,
2596 le32_to_cpu(es->s_last_error_line));
2597 if (es->s_last_error_ino)
2598 printk(": inode %u",
2599 le32_to_cpu(es->s_last_error_ino));
2600 if (es->s_last_error_block)
2601 printk(": block %llu", (unsigned long long)
2602 le64_to_cpu(es->s_last_error_block));
2603 printk("\n");
2604 }
2605 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2606}
2607
bfff6873
LC		 2608/* Find next suitable group and run ext4_init_inode_table */
2609static int ext4_run_li_request(struct ext4_li_request *elr)
2610{
2611 struct ext4_group_desc *gdp = NULL;
2612 ext4_group_t group, ngroups;
2613 struct super_block *sb;
2614 unsigned long timeout = 0;
2615 int ret = 0;
2616
2617 sb = elr->lr_super;
2618 ngroups = EXT4_SB(sb)->s_groups_count;
2619
2620 for (group = elr->lr_next_group; group < ngroups; group++) {
2621 gdp = ext4_get_group_desc(sb, group, NULL);
2622 if (!gdp) {
2623 ret = 1;
2624 break;
2625 }
2626
2627 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2628 break;
2629 }
2630
2631 if (group == ngroups)
2632 ret = 1;
2633
2634 if (!ret) {
2635 timeout = jiffies;
2636 ret = ext4_init_inode_table(sb, group,
2637 elr->lr_timeout ? 0 : 1);
2638 if (elr->lr_timeout == 0) {
51ce6511
LC		 2639 timeout = (jiffies - timeout) *
2640 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2641 elr->lr_timeout = timeout;
2642 }
2643 elr->lr_next_sched = jiffies + elr->lr_timeout;
2644 elr->lr_next_group = group + 1;
2645 }
2646
2647 return ret;
2648}
2649
2650/*
2651 * Remove lr_request from the list_request and free the
4ed5c033 2652 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2653 */
2654static void ext4_remove_li_request(struct ext4_li_request *elr)
2655{
2656 struct ext4_sb_info *sbi;
2657
2658 if (!elr)
2659 return;
2660
2661 sbi = elr->lr_sbi;
2662
2663 list_del(&elr->lr_request);
2664 sbi->s_li_request = NULL;
2665 kfree(elr);
2666}
2667
2668static void ext4_unregister_li_request(struct super_block *sb)
2669{
1bb933fb
LC		 2670 mutex_lock(&ext4_li_mtx);
2671 if (!ext4_li_info) {
2672 mutex_unlock(&ext4_li_mtx);
bfff6873 2673 return;
1bb933fb 2674 }
bfff6873
LC		 2675
2676 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2677 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2678 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2679 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2680}
2681
8f1f7453
ES		 2682static struct task_struct *ext4_lazyinit_task;
2683
bfff6873
LC		 2684/*
2685 * This is the function where ext4lazyinit thread lives. It walks
2686 * through the request list searching for next scheduled filesystem.
2687 * When such a fs is found, run the lazy initialization request
2688 * (ext4_rn_li_request) and keep track of the time spend in this
2689 * function. Based on that time we compute next schedule time of
2690 * the request. When walking through the list is complete, compute
2691 * next waking time and put itself into sleep.
2692 */
2693static int ext4_lazyinit_thread(void *arg)
2694{
2695 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2696 struct list_head *pos, *n;
2697 struct ext4_li_request *elr;
4ed5c033 2698 unsigned long next_wakeup, cur;
bfff6873
LC		 2699
2700 BUG_ON(NULL == eli);
2701
bfff6873
LC		 2702cont_thread:
2703 while (true) {
2704 next_wakeup = MAX_JIFFY_OFFSET;
2705
2706 mutex_lock(&eli->li_list_mtx);
2707 if (list_empty(&eli->li_request_list)) {
2708 mutex_unlock(&eli->li_list_mtx);
2709 goto exit_thread;
2710 }
2711
2712 list_for_each_safe(pos, n, &eli->li_request_list) {
2713 elr = list_entry(pos, struct ext4_li_request,
2714 lr_request);
2715
b2c78cd0
TT		 2716 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2717 if (ext4_run_li_request(elr) != 0) {
2718 /* error, remove the lazy_init job */
2719 ext4_remove_li_request(elr);
2720 continue;
2721 }
bfff6873
LC		 2722 }
2723
2724 if (time_before(elr->lr_next_sched, next_wakeup))
2725 next_wakeup = elr->lr_next_sched;
2726 }
2727 mutex_unlock(&eli->li_list_mtx);
2728
a0acae0e 2729 try_to_freeze();
bfff6873 2730
4ed5c033
LC		 2731 cur = jiffies;
2732 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2733 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2734 cond_resched();
2735 continue;
2736 }
2737
4ed5c033
LC		 2738 schedule_timeout_interruptible(next_wakeup - cur);
2739
8f1f7453
ES		 2740 if (kthread_should_stop()) {
2741 ext4_clear_request_list();
2742 goto exit_thread;
2743 }
bfff6873
LC		 2744 }
2745
2746exit_thread:
2747 /*
2748 * It looks like the request list is empty, but we need
2749 * to check it under the li_list_mtx lock, to prevent any
2750 * additions into it, and of course we should lock ext4_li_mtx
2751 * to atomically free the list and ext4_li_info, because at
2752 * this point another ext4 filesystem could be registering
2753 * new one.
2754 */
2755 mutex_lock(&ext4_li_mtx);
2756 mutex_lock(&eli->li_list_mtx);
2757 if (!list_empty(&eli->li_request_list)) {
2758 mutex_unlock(&eli->li_list_mtx);
2759 mutex_unlock(&ext4_li_mtx);
2760 goto cont_thread;
2761 }
2762 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2763 kfree(ext4_li_info);
2764 ext4_li_info = NULL;
2765 mutex_unlock(&ext4_li_mtx);
2766
2767 return 0;
2768}
2769
2770static void ext4_clear_request_list(void)
2771{
2772 struct list_head *pos, *n;
2773 struct ext4_li_request *elr;
2774
2775 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2776 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2777 elr = list_entry(pos, struct ext4_li_request,
2778 lr_request);
2779 ext4_remove_li_request(elr);
2780 }
2781 mutex_unlock(&ext4_li_info->li_list_mtx);
2782}
2783
2784static int ext4_run_lazyinit_thread(void)
2785{
8f1f7453
ES		 2786 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2787 ext4_li_info, "ext4lazyinit");
2788 if (IS_ERR(ext4_lazyinit_task)) {
2789 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2790 ext4_clear_request_list();
bfff6873
LC		 2791 kfree(ext4_li_info);
2792 ext4_li_info = NULL;
92b97816 2793 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2794 "initialization thread\n",
2795 err);
2796 return err;
2797 }
2798 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2799 return 0;
2800}
2801
2802/*
2803 * Check whether it make sense to run itable init. thread or not.
2804 * If there is at least one uninitialized inode table, return
2805 * corresponding group number, else the loop goes through all
2806 * groups and return total number of groups.
2807 */
2808static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2809{
2810 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2811 struct ext4_group_desc *gdp = NULL;
2812
2813 for (group = 0; group < ngroups; group++) {
2814 gdp = ext4_get_group_desc(sb, group, NULL);
2815 if (!gdp)
2816 continue;
2817
2818 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2819 break;
2820 }
2821
2822 return group;
2823}
2824
2825static int ext4_li_info_new(void)
2826{
2827 struct ext4_lazy_init *eli = NULL;
2828
2829 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2830 if (!eli)
2831 return -ENOMEM;
2832
bfff6873
LC		 2833 INIT_LIST_HEAD(&eli->li_request_list);
2834 mutex_init(&eli->li_list_mtx);
2835
bfff6873
LC		 2836 eli->li_state |= EXT4_LAZYINIT_QUIT;
2837
2838 ext4_li_info = eli;
2839
2840 return 0;
2841}
2842
2843static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2844 ext4_group_t start)
2845{
2846 struct ext4_sb_info *sbi = EXT4_SB(sb);
2847 struct ext4_li_request *elr;
2848 unsigned long rnd;
2849
2850 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2851 if (!elr)
2852 return NULL;
2853
2854 elr->lr_super = sb;
2855 elr->lr_sbi = sbi;
2856 elr->lr_next_group = start;
2857
2858 /*
2859 * Randomize first schedule time of the request to
2860 * spread the inode table initialization requests
2861 * better.
2862 */
2863 get_random_bytes(&rnd, sizeof(rnd));
2864 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2865 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2866
2867 return elr;
2868}
2869
2870static int ext4_register_li_request(struct super_block *sb,
2871 ext4_group_t first_not_zeroed)
2872{
2873 struct ext4_sb_info *sbi = EXT4_SB(sb);
2874 struct ext4_li_request *elr;
2875 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2876 int ret = 0;
bfff6873 2877
51ce6511
LC		 2878 if (sbi->s_li_request != NULL) {
2879 /*
2880 * Reset timeout so it can be computed again, because
2881 * s_li_wait_mult might have changed.
2882 */
2883 sbi->s_li_request->lr_timeout = 0;
beed5ecb 2884 return 0;
51ce6511 2885 }
bfff6873
LC		 2886
2887 if (first_not_zeroed == ngroups ||
2888 (sb->s_flags & MS_RDONLY) ||
55ff3840 2889 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 2890 return 0;
bfff6873
LC		 2891
2892 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 2893 if (!elr)
2894 return -ENOMEM;
bfff6873
LC		 2895
2896 mutex_lock(&ext4_li_mtx);
2897
2898 if (NULL == ext4_li_info) {
2899 ret = ext4_li_info_new();
2900 if (ret)
2901 goto out;
2902 }
2903
2904 mutex_lock(&ext4_li_info->li_list_mtx);
2905 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2906 mutex_unlock(&ext4_li_info->li_list_mtx);
2907
2908 sbi->s_li_request = elr;
46e4690b
TM		 2909 /*
2910 * set elr to NULL here since it has been inserted to
2911 * the request_list and the removal and free of it is
2912 * handled by ext4_clear_request_list from now on.
2913 */
2914 elr = NULL;
bfff6873
LC		 2915
2916 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2917 ret = ext4_run_lazyinit_thread();
2918 if (ret)
2919 goto out;
2920 }
bfff6873 2921out:
beed5ecb
NK		 2922 mutex_unlock(&ext4_li_mtx);
2923 if (ret)
bfff6873 2924 kfree(elr);
bfff6873
LC		 2925 return ret;
2926}
2927
2928/*
2929 * We do not need to lock anything since this is called on
2930 * module unload.
2931 */
2932static void ext4_destroy_lazyinit_thread(void)
2933{
2934 /*
2935 * If thread exited earlier
2936 * there's nothing to be done.
2937 */
8f1f7453 2938 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 2939 return;
2940
8f1f7453 2941 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 2942}
2943
2b2d6d01 2944static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 2945{
d4c402d9 2946 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 2947 struct buffer_head *bh;
617ba13b
MC		 2948 struct ext4_super_block *es = NULL;
2949 struct ext4_sb_info *sbi;
2950 ext4_fsblk_t block;
2951 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2952 ext4_fsblk_t logical_sb_block;
ac27a0ec 2953 unsigned long offset = 0;
ac27a0ec
DK		 2954 unsigned long journal_devnum = 0;
2955 unsigned long def_mount_opts;
2956 struct inode *root;
9f6200bb 2957 char *cp;
0390131b 2958 const char *descr;
dcc7dae3 2959 int ret = -ENOMEM;
281b5995 2960 int blocksize, clustersize;
4ec11028
TT		 2961 unsigned int db_count;
2962 unsigned int i;
281b5995 2963 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 2964 __u64 blocks_count;
833f4077 2965 int err;
b3881f74 2966 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 2967 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 2968
2969 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2970 if (!sbi)
dcc7dae3 2971 goto out_free_orig;
705895b6
PE		 2972
2973 sbi->s_blockgroup_lock =
2974 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2975 if (!sbi->s_blockgroup_lock) {
2976 kfree(sbi);
dcc7dae3 2977 goto out_free_orig;
705895b6 2978 }
ac27a0ec
DK		 2979 sb->s_fs_info = sbi;
2980 sbi->s_mount_opt = 0;
617ba13b
MC		 2981 sbi->s_resuid = EXT4_DEF_RESUID;
2982 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2983 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2984 sbi->s_sb_block = sb_block;
f613dfcb
TT		 2985 if (sb->s_bdev->bd_part)
2986 sbi->s_sectors_written_start =
2987 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 2988
9f6200bb
TT		 2989 /* Cleanup superblock name */
2990 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2991 *cp = '!';
2992
dcc7dae3 2993 ret = -EINVAL;
617ba13b 2994 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2995 if (!blocksize) {
b31e1552 2996 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 2997 goto out_fail;
2998 }
2999
3000 /*
617ba13b 3001 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3002 * block sizes. We need to calculate the offset from buffer start.
3003 */
617ba13b 3004 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3005 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3006 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3007 } else {
70bbb3e0 3008 logical_sb_block = sb_block;
ac27a0ec
DK		 3009 }
3010
70bbb3e0 3011 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3012 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3013 goto out_fail;
3014 }
3015 /*
3016 * Note: s_es must be initialized as soon as possible because
617ba13b 3017 * some ext4 macro-instructions depend on its value
ac27a0ec 3018 */
617ba13b 3019 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 3020 sbi->s_es = es;
3021 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3022 if (sb->s_magic != EXT4_SUPER_MAGIC)
3023 goto cantfind_ext4;
afc32f7e 3024 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 3025
3026 /* Set defaults before we parse the mount options */
3027 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3028 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3029 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3030 set_opt(sb, DEBUG);
87f26807 3031 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3032 set_opt(sb, GRPID);
617ba13b 3033 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3034 set_opt(sb, NO_UID32);
ea663336 3035 /* xattr user namespace & acls are now defaulted on */
03010a33 3036#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3037 set_opt(sb, XATTR_USER);
2e7842b8 3038#endif
03010a33 3039#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3040 set_opt(sb, POSIX_ACL);
2e7842b8 3041#endif
6fd7a467 3042 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3043 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3044 set_opt(sb, JOURNAL_DATA);
617ba13b 3045 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3046 set_opt(sb, ORDERED_DATA);
617ba13b 3047 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3048 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3049
3050 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3051 set_opt(sb, ERRORS_PANIC);
bb4f397a 3052 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3053 set_opt(sb, ERRORS_CONT);
bb4f397a 3054 else
fd8c37ec 3055 set_opt(sb, ERRORS_RO);
8b67f04a 3056 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3057 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3058 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3059 set_opt(sb, DISCARD);
ac27a0ec
DK		 3060
3061 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3062 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 3063 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3064 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3065 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3066
8b67f04a 3067 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3068 set_opt(sb, BARRIER);
ac27a0ec 3069
dd919b98
AK		 3070 /*
3071 * enable delayed allocation by default
3072 * Use -o nodelalloc to turn it off
3073 */
8b67f04a
TT		 3074 if (!IS_EXT3_SB(sb) &&
3075 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3076 set_opt(sb, DELALLOC);
dd919b98 3077
51ce6511
LC		 3078 /*
3079 * set default s_li_wait_mult for lazyinit, for the case there is
3080 * no mount option specified.
3081 */
3082 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3083
8b67f04a 3084 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3085 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3086 ext4_msg(sb, KERN_WARNING,
3087 "failed to parse options in superblock: %s",
3088 sbi->s_es->s_mount_opts);
3089 }
5a916be1 3090 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3091 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3092 &journal_ioprio, 0))
ac27a0ec
DK		 3093 goto failed_mount;
3094
56889787
TT		 3095 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3096 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3097 "with data=journal disables delayed "
3098 "allocation and O_DIRECT support!\n");
3099 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3100 ext4_msg(sb, KERN_ERR, "can't mount with "
3101 "both data=journal and delalloc");
3102 goto failed_mount;
3103 }
3104 if (test_opt(sb, DIOREAD_NOLOCK)) {
3105 ext4_msg(sb, KERN_ERR, "can't mount with "
3106 "both data=journal and delalloc");
3107 goto failed_mount;
3108 }
3109 if (test_opt(sb, DELALLOC))
3110 clear_opt(sb, DELALLOC);
3111 }
3112
3113 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3114 if (test_opt(sb, DIOREAD_NOLOCK)) {
3115 if (blocksize < PAGE_SIZE) {
3116 ext4_msg(sb, KERN_ERR, "can't mount with "
3117 "dioread_nolock if block size != PAGE_SIZE");
3118 goto failed_mount;
3119 }
3120 }
3121
ac27a0ec 3122 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3123 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3124
617ba13b
MC		 3125 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3126 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3127 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3128 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3129 ext4_msg(sb, KERN_WARNING,
3130 "feature flags set on rev 0 fs, "
3131 "running e2fsck is recommended");
469108ff 3132
2035e776
TT		 3133 if (IS_EXT2_SB(sb)) {
3134 if (ext2_feature_set_ok(sb))
3135 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3136 "using the ext4 subsystem");
3137 else {
3138 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3139 "to feature incompatibilities");
3140 goto failed_mount;
3141 }
3142 }
3143
3144 if (IS_EXT3_SB(sb)) {
3145 if (ext3_feature_set_ok(sb))
3146 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3147 "using the ext4 subsystem");
3148 else {
3149 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3150 "to feature incompatibilities");
3151 goto failed_mount;
3152 }
3153 }
3154
ac27a0ec
DK		 3155 /*
3156 * Check feature flags regardless of the revision level, since we
3157 * previously didn't change the revision level when setting the flags,
3158 * so there is a chance incompat flags are set on a rev 0 filesystem.
3159 */
a13fb1a4 3160 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3161 goto failed_mount;
a13fb1a4 3162
617ba13b
MC		 3163 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3164 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3165 ext4_msg(sb, KERN_ERR,
3166 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3167 goto failed_mount;
3168 }
3169
ac27a0ec 3170 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3171 /* Validate the filesystem blocksize */
3172 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3173 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3174 blocksize);
ac27a0ec
DK		 3175 goto failed_mount;
3176 }
3177
2b2d6d01 3178 brelse(bh);
70bbb3e0
AM		 3179 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3180 offset = do_div(logical_sb_block, blocksize);
3181 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3182 if (!bh) {
b31e1552
ES		 3183 ext4_msg(sb, KERN_ERR,
3184 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3185 goto failed_mount;
3186 }
617ba13b 3187 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 3188 sbi->s_es = es;
617ba13b 3189 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3190 ext4_msg(sb, KERN_ERR,
3191 "Magic mismatch, very weird!");
ac27a0ec
DK		 3192 goto failed_mount;
3193 }
3194 }
3195
a13fb1a4
ES		 3196 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3197 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3198 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3199 has_huge_files);
3200 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3201
617ba13b
MC		 3202 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3203 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3204 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3205 } else {
3206 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3207 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3208 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3209 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3210 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3211 ext4_msg(sb, KERN_ERR,
3212 "unsupported inode size: %d",
2b2d6d01 3213 sbi->s_inode_size);
ac27a0ec
DK		 3214 goto failed_mount;
3215 }
ef7f3835
KS		 3216 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3217 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3218 }
0b8e58a1 3219
0d1ee42f
AR		 3220 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3221 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3222 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3223 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3224 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3225 ext4_msg(sb, KERN_ERR,
3226 "unsupported descriptor size %lu",
0d1ee42f
AR		 3227 sbi->s_desc_size);
3228 goto failed_mount;
3229 }
3230 } else
3231 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3232
ac27a0ec 3233 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3234 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3235 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3236 goto cantfind_ext4;
0b8e58a1 3237
617ba13b 3238 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3239 if (sbi->s_inodes_per_block == 0)
617ba13b 3240 goto cantfind_ext4;
ac27a0ec
DK		 3241 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3242 sbi->s_inodes_per_block;
0d1ee42f 3243 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3244 sbi->s_sbh = bh;
3245 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3246 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3247 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3248
2b2d6d01 3249 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3250 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3251 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3252 i = le32_to_cpu(es->s_flags);
3253 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3254 sbi->s_hash_unsigned = 3;
3255 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3256#ifdef __CHAR_UNSIGNED__
3257 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3258 sbi->s_hash_unsigned = 3;
3259#else
3260 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3261#endif
f99b2589 3262 }
ac27a0ec 3263
281b5995
TT		 3264 /* Handle clustersize */
3265 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3266 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3267 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3268 if (has_bigalloc) {
3269 if (clustersize < blocksize) {
3270 ext4_msg(sb, KERN_ERR,
3271 "cluster size (%d) smaller than "
3272 "block size (%d)", clustersize, blocksize);
3273 goto failed_mount;
3274 }
3275 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3276 le32_to_cpu(es->s_log_block_size);
3277 sbi->s_clusters_per_group =
3278 le32_to_cpu(es->s_clusters_per_group);
3279 if (sbi->s_clusters_per_group > blocksize * 8) {
3280 ext4_msg(sb, KERN_ERR,
3281 "#clusters per group too big: %lu",
3282 sbi->s_clusters_per_group);
3283 goto failed_mount;
3284 }
3285 if (sbi->s_blocks_per_group !=
3286 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3287 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3288 "clusters per group (%lu) inconsistent",
3289 sbi->s_blocks_per_group,
3290 sbi->s_clusters_per_group);
3291 goto failed_mount;
3292 }
3293 } else {
3294 if (clustersize != blocksize) {
3295 ext4_warning(sb, "fragment/cluster size (%d) != "
3296 "block size (%d)", clustersize,
3297 blocksize);
3298 clustersize = blocksize;
3299 }
3300 if (sbi->s_blocks_per_group > blocksize * 8) {
3301 ext4_msg(sb, KERN_ERR,
3302 "#blocks per group too big: %lu",
3303 sbi->s_blocks_per_group);
3304 goto failed_mount;
3305 }
3306 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3307 sbi->s_cluster_bits = 0;
ac27a0ec 3308 }
281b5995
TT		 3309 sbi->s_cluster_ratio = clustersize / blocksize;
3310
ac27a0ec 3311 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3312 ext4_msg(sb, KERN_ERR,
3313 "#inodes per group too big: %lu",
2b2d6d01 3314 sbi->s_inodes_per_group);
ac27a0ec
DK		 3315 goto failed_mount;
3316 }
3317
bf43d84b
ES		 3318 /*
3319 * Test whether we have more sectors than will fit in sector_t,
3320 * and whether the max offset is addressable by the page cache.
3321 */
5a9ae68a 3322 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3323 ext4_blocks_count(es));
5a9ae68a 3324 if (err) {
b31e1552 3325 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3326 " too large to mount safely on this system");
ac27a0ec 3327 if (sizeof(sector_t) < 8)
90c699a9 3328 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3329 ret = err;
ac27a0ec
DK		 3330 goto failed_mount;
3331 }
3332
617ba13b
MC		 3333 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3334 goto cantfind_ext4;
e7c95593 3335
0f2ddca6
FTN		 3336 /* check blocks count against device size */
3337 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3338 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3339 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3340 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3341 ext4_blocks_count(es), blocks_count);
3342 goto failed_mount;
3343 }
3344
0b8e58a1
AD		 3345 /*
3346 * It makes no sense for the first data block to be beyond the end
3347 * of the filesystem.
3348 */
3349 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3350 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3351 "block %u is beyond end of filesystem (%llu)",
3352 le32_to_cpu(es->s_first_data_block),
3353 ext4_blocks_count(es));
e7c95593
ES		 3354 goto failed_mount;
3355 }
bd81d8ee
LV		 3356 blocks_count = (ext4_blocks_count(es) -
3357 le32_to_cpu(es->s_first_data_block) +
3358 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3359 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3360 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3361 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3362 "(block count %llu, first data block %u, "
b31e1552 3363 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3364 ext4_blocks_count(es),
3365 le32_to_cpu(es->s_first_data_block),
3366 EXT4_BLOCKS_PER_GROUP(sb));
3367 goto failed_mount;
3368 }
bd81d8ee 3369 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3370 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3371 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3372 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3373 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3374 sbi->s_group_desc = ext4_kvmalloc(db_count *
3375 sizeof(struct buffer_head *),
3376 GFP_KERNEL);
ac27a0ec 3377 if (sbi->s_group_desc == NULL) {
b31e1552 3378 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 3379 goto failed_mount;
3380 }
3381
9f6200bb
TT		 3382 if (ext4_proc_root)
3383 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3384
66acdcf4
TT		 3385 if (sbi->s_proc)
3386 proc_create_data("options", S_IRUGO, sbi->s_proc,
3387 &ext4_seq_options_fops, sb);
3388
705895b6 3389 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3390
3391 for (i = 0; i < db_count; i++) {
70bbb3e0 3392 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3393 sbi->s_group_desc[i] = sb_bread(sb, block);
3394 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3395 ext4_msg(sb, KERN_ERR,
3396 "can't read group descriptor %d", i);
ac27a0ec
DK		 3397 db_count = i;
3398 goto failed_mount2;
3399 }
3400 }
bfff6873 3401 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3402 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3403 goto failed_mount2;
3404 }
772cb7c8
JS		 3405 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3406 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3407 ext4_msg(sb, KERN_ERR,
3408 "unable to initialize "
3409 "flex_bg meta info!");
772cb7c8
JS		 3410 goto failed_mount2;
3411 }
3412
ac27a0ec
DK		 3413 sbi->s_gdb_count = db_count;
3414 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3415 spin_lock_init(&sbi->s_next_gen_lock);
3416
04496411
TM		 3417 init_timer(&sbi->s_err_report);
3418 sbi->s_err_report.function = print_daily_error_info;
3419 sbi->s_err_report.data = (unsigned long) sb;
3420
57042651 3421 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3422 ext4_count_free_clusters(sb));
ce7e010a
TT		 3423 if (!err) {
3424 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3425 ext4_count_free_inodes(sb));
3426 }
3427 if (!err) {
3428 err = percpu_counter_init(&sbi->s_dirs_counter,
3429 ext4_count_dirs(sb));
3430 }
3431 if (!err) {
57042651 3432 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3433 }
3434 if (err) {
3435 ext4_msg(sb, KERN_ERR, "insufficient memory");
3436 goto failed_mount3;
3437 }
3438
c9de560d 3439 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3440 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3441
ac27a0ec
DK		 3442 /*
3443 * set up enough so that it can read an inode
3444 */
9ca92389
TT		 3445 if (!test_opt(sb, NOLOAD) &&
3446 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3447 sb->s_op = &ext4_sops;
3448 else
3449 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3450 sb->s_export_op = &ext4_export_ops;
3451 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3452#ifdef CONFIG_QUOTA
617ba13b
MC		 3453 sb->s_qcop = &ext4_qctl_operations;
3454 sb->dq_op = &ext4_quota_operations;
ac27a0ec 3455#endif
f2fa2ffc
AK		 3456 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3457
ac27a0ec 3458 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3459 mutex_init(&sbi->s_orphan_lock);
8f82f840 3460 sbi->s_resize_flags = 0;
ac27a0ec
DK		 3461
3462 sb->s_root = NULL;
3463
3464 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3465 EXT4_HAS_INCOMPAT_FEATURE(sb,
3466 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3467
c5e06d10
JL		 3468 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3469 !(sb->s_flags & MS_RDONLY))
3470 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3471 goto failed_mount3;
3472
ac27a0ec
DK		 3473 /*
3474 * The first inode we look at is the journal inode. Don't try
3475 * root first: it may be modified in the journal!
3476 */
3477 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3478 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3479 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3480 goto failed_mount3;
0390131b
FM		 3481 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3482 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3483 ext4_msg(sb, KERN_ERR, "required journal recovery "
3484 "suppressed and not mounted read-only");
744692dc 3485 goto failed_mount_wq;
ac27a0ec 3486 } else {
fd8c37ec 3487 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3488 sbi->s_journal = NULL;
3489 needs_recovery = 0;
3490 goto no_journal;
ac27a0ec
DK		 3491 }
3492
eb40a09c
JS		 3493 if (ext4_blocks_count(es) > 0xffffffffULL &&
3494 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3495 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3496 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3497 goto failed_mount_wq;
eb40a09c
JS		 3498 }
3499
d4da6c9c
LT		 3500 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3501 jbd2_journal_set_features(sbi->s_journal,
3502 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 3503 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3504 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3505 jbd2_journal_set_features(sbi->s_journal,
3506 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 3507 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3508 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3509 } else {
3510 jbd2_journal_clear_features(sbi->s_journal,
3511 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3512 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3513 }
818d276c 3514
ac27a0ec
DK		 3515 /* We have now updated the journal if required, so we can
3516 * validate the data journaling mode. */
3517 switch (test_opt(sb, DATA_FLAGS)) {
3518 case 0:
3519 /* No mode set, assume a default based on the journal
63f57933
AM		 3520 * capabilities: ORDERED_DATA if the journal can
3521 * cope, else JOURNAL_DATA
3522 */
dab291af
MC		 3523 if (jbd2_journal_check_available_features
3524 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3525 set_opt(sb, ORDERED_DATA);
ac27a0ec 3526 else
fd8c37ec 3527 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3528 break;
3529
617ba13b
MC		 3530 case EXT4_MOUNT_ORDERED_DATA:
3531 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3532 if (!jbd2_journal_check_available_features
3533 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3534 ext4_msg(sb, KERN_ERR, "Journal does not support "
3535 "requested data journaling mode");
744692dc 3536 goto failed_mount_wq;
ac27a0ec
DK		 3537 }
3538 default:
3539 break;
3540 }
b3881f74 3541 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3542
18aadd47
BJ		 3543 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3544
ce7e010a
TT		 3545 /*
3546 * The journal may have updated the bg summary counts, so we
3547 * need to update the global counters.
3548 */
57042651 3549 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3550 ext4_count_free_clusters(sb));
ce7e010a
TT		 3551 percpu_counter_set(&sbi->s_freeinodes_counter,
3552 ext4_count_free_inodes(sb));
3553 percpu_counter_set(&sbi->s_dirs_counter,
3554 ext4_count_dirs(sb));
57042651 3555 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3556
ce7e010a 3557no_journal:
fd89d5f2
TH		 3558 /*
3559 * The maximum number of concurrent works can be high and
3560 * concurrency isn't really necessary. Limit it to 1.
3561 */
3562 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3563 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3564 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3565 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3566 goto failed_mount_wq;
3567 }
3568
ac27a0ec 3569 /*
dab291af 3570 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3571 * so we can safely mount the rest of the filesystem now.
3572 */
3573
1d1fe1ee
DH		 3574 root = ext4_iget(sb, EXT4_ROOT_INO);
3575 if (IS_ERR(root)) {
b31e1552 3576 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3577 ret = PTR_ERR(root);
32a9bb57 3578 root = NULL;
ac27a0ec
DK		 3579 goto failed_mount4;
3580 }
3581 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3582 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3583 iput(root);
ac27a0ec
DK		 3584 goto failed_mount4;
3585 }
48fde701 3586 sb->s_root = d_make_root(root);
1d1fe1ee 3587 if (!sb->s_root) {
b31e1552 3588 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3589 ret = -ENOMEM;
3590 goto failed_mount4;
3591 }
ac27a0ec 3592
2b2d6d01 3593 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 3594
3595 /* determine the minimum size of new large inodes, if present */
3596 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3597 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3598 EXT4_GOOD_OLD_INODE_SIZE;
3599 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3600 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3601 if (sbi->s_want_extra_isize <
3602 le16_to_cpu(es->s_want_extra_isize))
3603 sbi->s_want_extra_isize =
3604 le16_to_cpu(es->s_want_extra_isize);
3605 if (sbi->s_want_extra_isize <
3606 le16_to_cpu(es->s_min_extra_isize))
3607 sbi->s_want_extra_isize =
3608 le16_to_cpu(es->s_min_extra_isize);
3609 }
3610 }
3611 /* Check if enough inode space is available */
3612 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3613 sbi->s_inode_size) {
3614 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3615 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3616 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3617 "available");
ef7f3835
KS		 3618 }
3619
6fd058f7
TT		 3620 err = ext4_setup_system_zone(sb);
3621 if (err) {
b31e1552 3622 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3623 "zone (%d)", err);
94bf608a 3624 goto failed_mount4a;
6fd058f7
TT		 3625 }
3626
c2774d84
AK		 3627 ext4_ext_init(sb);
3628 err = ext4_mb_init(sb, needs_recovery);
3629 if (err) {
421f91d2 3630 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3631 err);
dcf2d804 3632 goto failed_mount5;
c2774d84
AK		 3633 }
3634
bfff6873
LC		 3635 err = ext4_register_li_request(sb, first_not_zeroed);
3636 if (err)
dcf2d804 3637 goto failed_mount6;
bfff6873 3638
3197ebdb
TT		 3639 sbi->s_kobj.kset = ext4_kset;
3640 init_completion(&sbi->s_kobj_unregister);
3641 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3642 "%s", sb->s_id);
dcf2d804
TM		 3643 if (err)
3644 goto failed_mount7;
3197ebdb 3645
617ba13b
MC		 3646 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3647 ext4_orphan_cleanup(sb, es);
3648 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3649 if (needs_recovery) {
b31e1552 3650 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3651 ext4_mark_recovery_complete(sb, es);
3652 }
3653 if (EXT4_SB(sb)->s_journal) {
3654 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3655 descr = " journalled data mode";
3656 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3657 descr = " ordered data mode";
3658 else
3659 descr = " writeback data mode";
3660 } else
3661 descr = "out journal";
3662
d4c402d9 3663 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3664 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3665 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3666
66e61a9e
TT		 3667 if (es->s_error_count)
3668 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3669
d4c402d9 3670 kfree(orig_data);
ac27a0ec
DK		 3671 return 0;
3672
617ba13b 3673cantfind_ext4:
ac27a0ec 3674 if (!silent)
b31e1552 3675 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3676 goto failed_mount;
3677
dcf2d804
TM		 3678failed_mount7:
3679 ext4_unregister_li_request(sb);
3680failed_mount6:
dcf2d804 3681 ext4_mb_release(sb);
94bf608a
AV		 3682failed_mount5:
3683 ext4_ext_release(sb);
dcf2d804 3684 ext4_release_system_zone(sb);
94bf608a
AV		 3685failed_mount4a:
3686 dput(sb->s_root);
32a9bb57 3687 sb->s_root = NULL;
94bf608a 3688failed_mount4:
b31e1552 3689 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3690 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3691failed_mount_wq:
0390131b
FM		 3692 if (sbi->s_journal) {
3693 jbd2_journal_destroy(sbi->s_journal);
3694 sbi->s_journal = NULL;
3695 }
ac27a0ec 3696failed_mount3:
04496411 3697 del_timer(&sbi->s_err_report);
9933fc0a
TT		 3698 if (sbi->s_flex_groups)
3699 ext4_kvfree(sbi->s_flex_groups);
57042651 3700 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 3701 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3702 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 3703 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 3704 if (sbi->s_mmp_tsk)
3705 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 3706failed_mount2:
3707 for (i = 0; i < db_count; i++)
3708 brelse(sbi->s_group_desc[i]);
f18a5f21 3709 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 3710failed_mount:
240799cd 3711 if (sbi->s_proc) {
66acdcf4 3712 remove_proc_entry("options", sbi->s_proc);
9f6200bb 3713 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3714 }
ac27a0ec
DK		 3715#ifdef CONFIG_QUOTA
3716 for (i = 0; i < MAXQUOTAS; i++)
3717 kfree(sbi->s_qf_names[i]);
3718#endif
617ba13b 3719 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3720 brelse(bh);
3721out_fail:
3722 sb->s_fs_info = NULL;
f6830165 3723 kfree(sbi->s_blockgroup_lock);
ac27a0ec 3724 kfree(sbi);
dcc7dae3 3725out_free_orig:
d4c402d9 3726 kfree(orig_data);
1d1fe1ee 3727 return ret;
ac27a0ec
DK		 3728}
3729
3730/*
3731 * Setup any per-fs journal parameters now. We'll do this both on
3732 * initial mount, once the journal has been initialised but before we've
3733 * done any recovery; and again on any subsequent remount.
3734 */
617ba13b 3735static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3736{
617ba13b 3737 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3738
30773840
TT		 3739 journal->j_commit_interval = sbi->s_commit_interval;
3740 journal->j_min_batch_time = sbi->s_min_batch_time;
3741 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 3742
a931da6a 3743 write_lock(&journal->j_state_lock);
ac27a0ec 3744 if (test_opt(sb, BARRIER))
dab291af 3745 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3746 else
dab291af 3747 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3748 if (test_opt(sb, DATA_ERR_ABORT))
3749 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3750 else
3751 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 3752 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 3753}
3754
617ba13b 3755static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3756 unsigned int journal_inum)
3757{
3758 struct inode *journal_inode;
3759 journal_t *journal;
3760
0390131b
FM		 3761 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3762
ac27a0ec
DK		 3763 /* First, test for the existence of a valid inode on disk. Bad
3764 * things happen if we iget() an unused inode, as the subsequent
3765 * iput() will try to delete it. */
3766
1d1fe1ee
DH		 3767 journal_inode = ext4_iget(sb, journal_inum);
3768 if (IS_ERR(journal_inode)) {
b31e1552 3769 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3770 return NULL;
3771 }
3772 if (!journal_inode->i_nlink) {
3773 make_bad_inode(journal_inode);
3774 iput(journal_inode);
b31e1552 3775 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3776 return NULL;
3777 }
3778
e5f8eab8 3779 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3780 journal_inode, journal_inode->i_size);
1d1fe1ee 3781 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3782 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3783 iput(journal_inode);
3784 return NULL;
3785 }
3786
dab291af 3787 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3788 if (!journal) {
b31e1552 3789 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3790 iput(journal_inode);
3791 return NULL;
3792 }
3793 journal->j_private = sb;
617ba13b 3794 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3795 return journal;
3796}
3797
617ba13b 3798static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3799 dev_t j_dev)
3800{
2b2d6d01 3801 struct buffer_head *bh;
ac27a0ec 3802 journal_t *journal;
617ba13b
MC		 3803 ext4_fsblk_t start;
3804 ext4_fsblk_t len;
ac27a0ec 3805 int hblock, blocksize;
617ba13b 3806 ext4_fsblk_t sb_block;
ac27a0ec 3807 unsigned long offset;
2b2d6d01 3808 struct ext4_super_block *es;
ac27a0ec
DK		 3809 struct block_device *bdev;
3810
0390131b
FM		 3811 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3812
b31e1552 3813 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3814 if (bdev == NULL)
3815 return NULL;
3816
ac27a0ec 3817 blocksize = sb->s_blocksize;
e1defc4f 3818 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3819 if (blocksize < hblock) {
b31e1552
ES		 3820 ext4_msg(sb, KERN_ERR,
3821 "blocksize too small for journal device");
ac27a0ec
DK		 3822 goto out_bdev;
3823 }
3824
617ba13b
MC		 3825 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3826 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3827 set_blocksize(bdev, blocksize);
3828 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3829 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3830 "external journal");
ac27a0ec
DK		 3831 goto out_bdev;
3832 }
3833
617ba13b
MC		 3834 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3835 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3836 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3837 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3838 ext4_msg(sb, KERN_ERR, "external journal has "
3839 "bad superblock");
ac27a0ec
DK		 3840 brelse(bh);
3841 goto out_bdev;
3842 }
3843
617ba13b 3844 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3845 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3846 brelse(bh);
3847 goto out_bdev;
3848 }
3849
bd81d8ee 3850 len = ext4_blocks_count(es);
ac27a0ec
DK		 3851 start = sb_block + 1;
3852 brelse(bh); /* we're done with the superblock */
3853
dab291af 3854 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3855 start, len, blocksize);
3856 if (!journal) {
b31e1552 3857 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3858 goto out_bdev;
3859 }
3860 journal->j_private = sb;
3861 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3862 wait_on_buffer(journal->j_sb_buffer);
3863 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3864 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3865 goto out_journal;
3866 }
3867 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3868 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3869 "user (unsupported) - %d",
ac27a0ec
DK		 3870 be32_to_cpu(journal->j_superblock->s_nr_users));
3871 goto out_journal;
3872 }
617ba13b
MC		 3873 EXT4_SB(sb)->journal_bdev = bdev;
3874 ext4_init_journal_params(sb, journal);
ac27a0ec 3875 return journal;
0b8e58a1 3876
ac27a0ec 3877out_journal:
dab291af 3878 jbd2_journal_destroy(journal);
ac27a0ec 3879out_bdev:
617ba13b 3880 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3881 return NULL;
3882}
3883
617ba13b
MC		 3884static int ext4_load_journal(struct super_block *sb,
3885 struct ext4_super_block *es,
ac27a0ec
DK		 3886 unsigned long journal_devnum)
3887{
3888 journal_t *journal;
3889 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3890 dev_t journal_dev;
3891 int err = 0;
3892 int really_read_only;
3893
0390131b
FM		 3894 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3895
ac27a0ec
DK		 3896 if (journal_devnum &&
3897 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3898 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3899 "numbers have changed");
ac27a0ec
DK		 3900 journal_dev = new_decode_dev(journal_devnum);
3901 } else
3902 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3903
3904 really_read_only = bdev_read_only(sb->s_bdev);
3905
3906 /*
3907 * Are we loading a blank journal or performing recovery after a
3908 * crash? For recovery, we need to check in advance whether we
3909 * can get read-write access to the device.
3910 */
617ba13b 3911 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3912 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3913 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3914 "required on readonly filesystem");
ac27a0ec 3915 if (really_read_only) {
b31e1552
ES		 3916 ext4_msg(sb, KERN_ERR, "write access "
3917 "unavailable, cannot proceed");
ac27a0ec
DK		 3918 return -EROFS;
3919 }
b31e1552
ES		 3920 ext4_msg(sb, KERN_INFO, "write access will "
3921 "be enabled during recovery");
ac27a0ec
DK		 3922 }
3923 }
3924
3925 if (journal_inum && journal_dev) {
b31e1552
ES		 3926 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3927 "and inode journals!");
ac27a0ec
DK		 3928 return -EINVAL;
3929 }
3930
3931 if (journal_inum) {
617ba13b 3932 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3933 return -EINVAL;
3934 } else {
617ba13b 3935 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3936 return -EINVAL;
3937 }
3938
90576c0b 3939 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 3940 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3941
617ba13b 3942 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3943 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 3944 if (!err) {
3945 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3946 if (save)
3947 memcpy(save, ((char *) es) +
3948 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 3949 err = jbd2_journal_load(journal);
1c13d5c0
TT		 3950 if (save)
3951 memcpy(((char *) es) + EXT4_S_ERR_START,
3952 save, EXT4_S_ERR_LEN);
3953 kfree(save);
3954 }
ac27a0ec
DK		 3955
3956 if (err) {
b31e1552 3957 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3958 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3959 return err;
3960 }
3961
617ba13b
MC		 3962 EXT4_SB(sb)->s_journal = journal;
3963 ext4_clear_journal_err(sb, es);
ac27a0ec 3964
c41303ce 3965 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 3966 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3967 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3968
3969 /* Make sure we flush the recovery flag to disk. */
e2d67052 3970 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3971 }
3972
3973 return 0;
3974}
3975
e2d67052 3976static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3977{
e2d67052 3978 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3979 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3980 int error = 0;
ac27a0ec 3981
7c2e7087 3982 if (!sbh || block_device_ejected(sb))
c4be0c1d 3983 return error;
914258bf
TT		 3984 if (buffer_write_io_error(sbh)) {
3985 /*
3986 * Oh, dear. A previous attempt to write the
3987 * superblock failed. This could happen because the
3988 * USB device was yanked out. Or it could happen to
3989 * be a transient write error and maybe the block will
3990 * be remapped. Nothing we can do but to retry the
3991 * write and hope for the best.
3992 */
b31e1552
ES		 3993 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3994 "superblock detected");
914258bf
TT		 3995 clear_buffer_write_io_error(sbh);
3996 set_buffer_uptodate(sbh);
3997 }
71290b36
TT		 3998 /*
3999 * If the file system is mounted read-only, don't update the
4000 * superblock write time. This avoids updating the superblock
4001 * write time when we are mounting the root file system
4002 * read/only but we need to replay the journal; at that point,
4003 * for people who are east of GMT and who make their clock
4004 * tick in localtime for Windows bug-for-bug compatibility,
4005 * the clock is set in the future, and this will cause e2fsck
4006 * to complain and force a full file system check.
4007 */
4008 if (!(sb->s_flags & MS_RDONLY))
4009 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4010 if (sb->s_bdev->bd_part)
4011 es->s_kbytes_written =
4012 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4013 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4014 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4015 else
4016 es->s_kbytes_written =
4017 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4018 ext4_free_blocks_count_set(es,
4019 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4020 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4021 es->s_free_inodes_count =
4022 cpu_to_le32(percpu_counter_sum_positive(
4023 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 4024 sb->s_dirt = 0;
ac27a0ec
DK		 4025 BUFFER_TRACE(sbh, "marking dirty");
4026 mark_buffer_dirty(sbh);
914258bf 4027 if (sync) {
c4be0c1d
TS		 4028 error = sync_dirty_buffer(sbh);
4029 if (error)
4030 return error;
4031
4032 error = buffer_write_io_error(sbh);
4033 if (error) {
b31e1552
ES		 4034 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4035 "superblock");
914258bf
TT		 4036 clear_buffer_write_io_error(sbh);
4037 set_buffer_uptodate(sbh);
4038 }
4039 }
c4be0c1d 4040 return error;
ac27a0ec
DK		 4041}
4042
ac27a0ec
DK		 4043/*
4044 * Have we just finished recovery? If so, and if we are mounting (or
4045 * remounting) the filesystem readonly, then we will end up with a
4046 * consistent fs on disk. Record that fact.
4047 */
2b2d6d01
TT		 4048static void ext4_mark_recovery_complete(struct super_block *sb,
4049 struct ext4_super_block *es)
ac27a0ec 4050{
617ba13b 4051 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4052
0390131b
FM		 4053 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4054 BUG_ON(journal != NULL);
4055 return;
4056 }
dab291af 4057 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4058 if (jbd2_journal_flush(journal) < 0)
4059 goto out;
4060
617ba13b 4061 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4062 sb->s_flags & MS_RDONLY) {
617ba13b 4063 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4064 ext4_commit_super(sb, 1);
ac27a0ec 4065 }
7ffe1ea8
HK		 4066
4067out:
dab291af 4068 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4069}
4070
4071/*
4072 * If we are mounting (or read-write remounting) a filesystem whose journal
4073 * has recorded an error from a previous lifetime, move that error to the
4074 * main filesystem now.
4075 */
2b2d6d01
TT		 4076static void ext4_clear_journal_err(struct super_block *sb,
4077 struct ext4_super_block *es)
ac27a0ec
DK		 4078{
4079 journal_t *journal;
4080 int j_errno;
4081 const char *errstr;
4082
0390131b
FM		 4083 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4084
617ba13b 4085 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4086
4087 /*
4088 * Now check for any error status which may have been recorded in the
617ba13b 4089 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4090 */
4091
dab291af 4092 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4093 if (j_errno) {
4094 char nbuf[16];
4095
617ba13b 4096 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4097 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4098 "from previous mount: %s", errstr);
12062ddd 4099 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4100
617ba13b
MC		 4101 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4102 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4103 ext4_commit_super(sb, 1);
ac27a0ec 4104
dab291af 4105 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 4106 }
4107}
4108
4109/*
4110 * Force the running and committing transactions to commit,
4111 * and wait on the commit.
4112 */
617ba13b 4113int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4114{
4115 journal_t *journal;
0390131b 4116 int ret = 0;
ac27a0ec
DK		 4117
4118 if (sb->s_flags & MS_RDONLY)
4119 return 0;
4120
617ba13b 4121 journal = EXT4_SB(sb)->s_journal;
6b0310fb 4122 if (journal) {
437f88cc 4123 vfs_check_frozen(sb, SB_FREEZE_TRANS);
0390131b 4124 ret = ext4_journal_force_commit(journal);
6b0310fb 4125 }
0390131b 4126
ac27a0ec
DK		 4127 return ret;
4128}
4129
2b2d6d01 4130static void ext4_write_super(struct super_block *sb)
ac27a0ec 4131{
ebc1ac16 4132 lock_super(sb);
9ca92389 4133 ext4_commit_super(sb, 1);
ebc1ac16 4134 unlock_super(sb);
ac27a0ec
DK		 4135}
4136
617ba13b 4137static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4138{
14ce0cb4 4139 int ret = 0;
9eddacf9 4140 tid_t target;
8d5d02e6 4141 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4142
9bffad1e 4143 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 4144 flush_workqueue(sbi->dio_unwritten_wq);
4145 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4146 if (wait)
8d5d02e6 4147 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4148 }
14ce0cb4 4149 return ret;
ac27a0ec
DK		 4150}
4151
4152/*
4153 * LVM calls this function before a (read-only) snapshot is created. This
4154 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4155 *
4156 * Note that only this function cannot bring a filesystem to be in a clean
4157 * state independently, because ext4 prevents a new handle from being started
4158 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4159 * the upper layer.
ac27a0ec 4160 */
c4be0c1d 4161static int ext4_freeze(struct super_block *sb)
ac27a0ec 4162{
c4be0c1d
TS		 4163 int error = 0;
4164 journal_t *journal;
ac27a0ec 4165
9ca92389
TT		 4166 if (sb->s_flags & MS_RDONLY)
4167 return 0;
ac27a0ec 4168
9ca92389 4169 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4170
9ca92389
TT		 4171 /* Now we set up the journal barrier. */
4172 jbd2_journal_lock_updates(journal);
ac27a0ec 4173
9ca92389
TT		 4174 /*
4175 * Don't clear the needs_recovery flag if we failed to flush
4176 * the journal.
4177 */
4178 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4179 if (error < 0)
4180 goto out;
9ca92389
TT		 4181
4182 /* Journal blocked and flushed, clear needs_recovery flag. */
4183 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4184 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 4185out:
4186 /* we rely on s_frozen to stop further updates */
4187 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4188 return error;
ac27a0ec
DK		 4189}
4190
4191/*
4192 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4193 * flag here, even though the filesystem is not technically dirty yet.
4194 */
c4be0c1d 4195static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4196{
9ca92389
TT		 4197 if (sb->s_flags & MS_RDONLY)
4198 return 0;
4199
4200 lock_super(sb);
4201 /* Reset the needs_recovery flag before the fs is unlocked. */
4202 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4203 ext4_commit_super(sb, 1);
4204 unlock_super(sb);
c4be0c1d 4205 return 0;
ac27a0ec
DK		 4206}
4207
673c6100
TT		 4208/*
4209 * Structure to save mount options for ext4_remount's benefit
4210 */
4211struct ext4_mount_options {
4212 unsigned long s_mount_opt;
a2595b8a 4213 unsigned long s_mount_opt2;
673c6100
TT		 4214 uid_t s_resuid;
4215 gid_t s_resgid;
4216 unsigned long s_commit_interval;
4217 u32 s_min_batch_time, s_max_batch_time;
4218#ifdef CONFIG_QUOTA
4219 int s_jquota_fmt;
4220 char *s_qf_names[MAXQUOTAS];
4221#endif
4222};
4223
2b2d6d01 4224static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4225{
2b2d6d01 4226 struct ext4_super_block *es;
617ba13b 4227 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4228 unsigned long old_sb_flags;
617ba13b 4229 struct ext4_mount_options old_opts;
c79d967d 4230 int enable_quota = 0;
8a266467 4231 ext4_group_t g;
b3881f74 4232 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4233 int err = 0;
ac27a0ec
DK		 4234#ifdef CONFIG_QUOTA
4235 int i;
4236#endif
d4c402d9 4237 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4238
4239 /* Store the original options */
bbd6851a 4240 lock_super(sb);
ac27a0ec
DK		 4241 old_sb_flags = sb->s_flags;
4242 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4243 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4244 old_opts.s_resuid = sbi->s_resuid;
4245 old_opts.s_resgid = sbi->s_resgid;
4246 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4247 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4248 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4249#ifdef CONFIG_QUOTA
4250 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4251 for (i = 0; i < MAXQUOTAS; i++)
4252 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4253#endif
b3881f74
TT		 4254 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4255 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4256
4257 /*
4258 * Allow the "check" option to be passed as a remount option.
4259 */
661aa520 4260 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4261 err = -EINVAL;
4262 goto restore_opts;
4263 }
4264
4ab2f15b 4265 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4266 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4267
4268 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4269 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4270
4271 es = sbi->s_es;
4272
b3881f74 4273 if (sbi->s_journal) {
0390131b 4274 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4275 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4276 }
ac27a0ec 4277
661aa520 4278 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4279 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4280 err = -EROFS;
4281 goto restore_opts;
4282 }
4283
4284 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4285 err = dquot_suspend(sb, -1);
4286 if (err < 0)
c79d967d 4287 goto restore_opts;
c79d967d 4288
ac27a0ec
DK		 4289 /*
4290 * First of all, the unconditional stuff we have to do
4291 * to disable replay of the journal when we next remount
4292 */
4293 sb->s_flags |= MS_RDONLY;
4294
4295 /*
4296 * OK, test if we are remounting a valid rw partition
4297 * readonly, and if so set the rdonly flag and then
4298 * mark the partition as valid again.
4299 */
617ba13b
MC		 4300 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4301 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4302 es->s_state = cpu_to_le16(sbi->s_mount_state);
4303
a63c9eb2 4304 if (sbi->s_journal)
0390131b 4305 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4306 } else {
a13fb1a4
ES		 4307 /* Make sure we can mount this feature set readwrite */
4308 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4309 err = -EROFS;
4310 goto restore_opts;
4311 }
8a266467
TT		 4312 /*
4313 * Make sure the group descriptor checksums
0b8e58a1 4314 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4315 */
4316 for (g = 0; g < sbi->s_groups_count; g++) {
4317 struct ext4_group_desc *gdp =
4318 ext4_get_group_desc(sb, g, NULL);
4319
4320 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 4321 ext4_msg(sb, KERN_ERR,
4322 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4323 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4324 le16_to_cpu(gdp->bg_checksum));
4325 err = -EINVAL;
4326 goto restore_opts;
4327 }
4328 }
4329
ead6596b
ES		 4330 /*
4331 * If we have an unprocessed orphan list hanging
4332 * around from a previously readonly bdev mount,
4333 * require a full umount/remount for now.
4334 */
4335 if (es->s_last_orphan) {
b31e1552 4336 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4337 "remount RDWR because of unprocessed "
4338 "orphan inode list. Please "
b31e1552 4339 "umount/remount instead");
ead6596b
ES		 4340 err = -EINVAL;
4341 goto restore_opts;
4342 }
4343
ac27a0ec
DK		 4344 /*
4345 * Mounting a RDONLY partition read-write, so reread
4346 * and store the current valid flag. (It may have
4347 * been changed by e2fsck since we originally mounted
4348 * the partition.)
4349 */
0390131b
FM		 4350 if (sbi->s_journal)
4351 ext4_clear_journal_err(sb, es);
ac27a0ec 4352 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4353 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4354 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4355 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4356 EXT4_FEATURE_INCOMPAT_MMP))
4357 if (ext4_multi_mount_protect(sb,
4358 le64_to_cpu(es->s_mmp_block))) {
4359 err = -EROFS;
4360 goto restore_opts;
4361 }
c79d967d 4362 enable_quota = 1;
ac27a0ec
DK		 4363 }
4364 }
bfff6873
LC		 4365
4366 /*
4367 * Reinitialize lazy itable initialization thread based on
4368 * current settings
4369 */
4370 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4371 ext4_unregister_li_request(sb);
4372 else {
4373 ext4_group_t first_not_zeroed;
4374 first_not_zeroed = ext4_has_uninit_itable(sb);
4375 ext4_register_li_request(sb, first_not_zeroed);
4376 }
4377
6fd058f7 4378 ext4_setup_system_zone(sb);
0390131b 4379 if (sbi->s_journal == NULL)
e2d67052 4380 ext4_commit_super(sb, 1);
0390131b 4381
ac27a0ec
DK		 4382#ifdef CONFIG_QUOTA
4383 /* Release old quota file names */
4384 for (i = 0; i < MAXQUOTAS; i++)
4385 if (old_opts.s_qf_names[i] &&
4386 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4387 kfree(old_opts.s_qf_names[i]);
4388#endif
bbd6851a 4389 unlock_super(sb);
c79d967d 4390 if (enable_quota)
0f0dd62f 4391 dquot_resume(sb, -1);
d4c402d9
CW		 4392
4393 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4394 kfree(orig_data);
ac27a0ec 4395 return 0;
0b8e58a1 4396
ac27a0ec
DK		 4397restore_opts:
4398 sb->s_flags = old_sb_flags;
4399 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4400 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4401 sbi->s_resuid = old_opts.s_resuid;
4402 sbi->s_resgid = old_opts.s_resgid;
4403 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4404 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4405 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4406#ifdef CONFIG_QUOTA
4407 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4408 for (i = 0; i < MAXQUOTAS; i++) {
4409 if (sbi->s_qf_names[i] &&
4410 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4411 kfree(sbi->s_qf_names[i]);
4412 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4413 }
4414#endif
bbd6851a 4415 unlock_super(sb);
d4c402d9 4416 kfree(orig_data);
ac27a0ec
DK		 4417 return err;
4418}
4419
f975d6bc
TT		 4420/*
4421 * Note: calculating the overhead so we can be compatible with
4422 * historical BSD practice is quite difficult in the face of
4423 * clusters/bigalloc. This is because multiple metadata blocks from
4424 * different block group can end up in the same allocation cluster.
4425 * Calculating the exact overhead in the face of clustered allocation
4426 * requires either O(all block bitmaps) in memory or O(number of block
4427 * groups**2) in time. We will still calculate the superblock for
4428 * older file systems --- and if we come across with a bigalloc file
4429 * system with zero in s_overhead_clusters the estimate will be close to
4430 * correct especially for very large cluster sizes --- but for newer
4431 * file systems, it's better to calculate this figure once at mkfs
4432 * time, and store it in the superblock. If the superblock value is
4433 * present (even for non-bigalloc file systems), we will use it.
4434 */
2b2d6d01 4435static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4436{
4437 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4438 struct ext4_sb_info *sbi = EXT4_SB(sb);
4439 struct ext4_super_block *es = sbi->s_es;
f975d6bc 4440 struct ext4_group_desc *gdp;
960cc398 4441 u64 fsid;
d02a9391 4442 s64 bfree;
ac27a0ec 4443
5e70030d
BP		 4444 if (test_opt(sb, MINIX_DF)) {
4445 sbi->s_overhead_last = 0;
f975d6bc
TT		 4446 } else if (es->s_overhead_clusters) {
4447 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
6bc9feff 4448 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 4449 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 4450 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 4451
4452 /*
5e70030d
BP		 4453 * Compute the overhead (FS structures). This is constant
4454 * for a given filesystem unless the number of block groups
4455 * changes so we cache the previous value until it does.
ac27a0ec
DK		 4456 */
4457
4458 /*
4459 * All of the blocks before first_data_block are
4460 * overhead
4461 */
f975d6bc 4462 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
ac27a0ec
DK		 4463
4464 /*
f975d6bc 4465 * Add the overhead found in each block group
ac27a0ec
DK		 4466 */
4467 for (i = 0; i < ngroups; i++) {
f975d6bc
TT		 4468 gdp = ext4_get_group_desc(sb, i, NULL);
4469 overhead += ext4_num_overhead_clusters(sb, i, gdp);
ac27a0ec
DK		 4470 cond_resched();
4471 }
5e70030d
BP		 4472 sbi->s_overhead_last = overhead;
4473 smp_wmb();
6bc9feff 4474 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 4475 }
4476
617ba13b 4477 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4478 buf->f_bsize = sb->s_blocksize;
f975d6bc
TT		 4479 buf->f_blocks = (ext4_blocks_count(es) -
4480 EXT4_C2B(sbi, sbi->s_overhead_last));
57042651
TT		 4481 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4482 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4483 /* prevent underflow in case that few free space is available */
57042651 4484 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4485 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4486 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4487 buf->f_bavail = 0;
4488 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4489 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4490 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4491 fsid = le64_to_cpup((void *)es->s_uuid) ^
4492 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4493 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4494 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4495
ac27a0ec
DK		 4496 return 0;
4497}
4498
0b8e58a1
AD		 4499/* Helper function for writing quotas on sync - we need to start transaction
4500 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4501 * Process 1 Process 2
617ba13b 4502 * ext4_create() quota_sync()
a269eb18 4503 * jbd2_journal_start() write_dquot()
871a2931 4504 * dquot_initialize() down(dqio_mutex)
dab291af 4505 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4506 *
4507 */
4508
4509#ifdef CONFIG_QUOTA
4510
4511static inline struct inode *dquot_to_inode(struct dquot *dquot)
4512{
4513 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4514}
4515
617ba13b 4516static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4517{
4518 int ret, err;
4519 handle_t *handle;
4520 struct inode *inode;
4521
4522 inode = dquot_to_inode(dquot);
617ba13b 4523 handle = ext4_journal_start(inode,
0b8e58a1 4524 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4525 if (IS_ERR(handle))
4526 return PTR_ERR(handle);
4527 ret = dquot_commit(dquot);
617ba13b 4528 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4529 if (!ret)
4530 ret = err;
4531 return ret;
4532}
4533
617ba13b 4534static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4535{
4536 int ret, err;
4537 handle_t *handle;
4538
617ba13b 4539 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4540 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4541 if (IS_ERR(handle))
4542 return PTR_ERR(handle);
4543 ret = dquot_acquire(dquot);
617ba13b 4544 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4545 if (!ret)
4546 ret = err;
4547 return ret;
4548}
4549
617ba13b 4550static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4551{
4552 int ret, err;
4553 handle_t *handle;
4554
617ba13b 4555 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4556 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4557 if (IS_ERR(handle)) {
4558 /* Release dquot anyway to avoid endless cycle in dqput() */
4559 dquot_release(dquot);
ac27a0ec 4560 return PTR_ERR(handle);
9c3013e9 4561 }
ac27a0ec 4562 ret = dquot_release(dquot);
617ba13b 4563 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4564 if (!ret)
4565 ret = err;
4566 return ret;
4567}
4568
617ba13b 4569static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4570{
2c8be6b2 4571 /* Are we journaling quotas? */
617ba13b
MC		 4572 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4573 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4574 dquot_mark_dquot_dirty(dquot);
617ba13b 4575 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4576 } else {
4577 return dquot_mark_dquot_dirty(dquot);
4578 }
4579}
4580
617ba13b 4581static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4582{
4583 int ret, err;
4584 handle_t *handle;
4585
4586 /* Data block + inode block */
617ba13b 4587 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4588 if (IS_ERR(handle))
4589 return PTR_ERR(handle);
4590 ret = dquot_commit_info(sb, type);
617ba13b 4591 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4592 if (!ret)
4593 ret = err;
4594 return ret;
4595}
4596
4597/*
4598 * Turn on quotas during mount time - we need to find
4599 * the quota file and such...
4600 */
617ba13b 4601static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4602{
287a8095
CH		 4603 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4604 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4605}
4606
4607/*
4608 * Standard function to be called on quota_on
4609 */
617ba13b 4610static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4611 struct path *path)
ac27a0ec
DK		 4612{
4613 int err;
ac27a0ec
DK		 4614
4615 if (!test_opt(sb, QUOTA))
4616 return -EINVAL;
0623543b 4617
ac27a0ec 4618 /* Quotafile not on the same filesystem? */
d8c9584e 4619 if (path->dentry->d_sb != sb)
ac27a0ec 4620 return -EXDEV;
0623543b
JK		 4621 /* Journaling quota? */
4622 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4623 /* Quotafile not in fs root? */
f00c9e44 4624 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4625 ext4_msg(sb, KERN_WARNING,
4626 "Quota file not on filesystem root. "
4627 "Journaled quota will not work");
2b2d6d01 4628 }
0623543b
JK		 4629
4630 /*
4631 * When we journal data on quota file, we have to flush journal to see
4632 * all updates to the file when we bypass pagecache...
4633 */
0390131b 4634 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4635 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4636 /*
4637 * We don't need to lock updates but journal_flush() could
4638 * otherwise be livelocked...
4639 */
4640 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4641 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4642 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4643 if (err)
7ffe1ea8 4644 return err;
0623543b
JK		 4645 }
4646
f00c9e44 4647 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4648}
4649
ca0e05e4
DM		 4650static int ext4_quota_off(struct super_block *sb, int type)
4651{
21f97697
JK		 4652 struct inode *inode = sb_dqopt(sb)->files[type];
4653 handle_t *handle;
4654
87009d86
DM		 4655 /* Force all delayed allocation blocks to be allocated.
4656 * Caller already holds s_umount sem */
4657 if (test_opt(sb, DELALLOC))
ca0e05e4 4658 sync_filesystem(sb);
ca0e05e4 4659
0b268590
AG		 4660 if (!inode)
4661 goto out;
4662
21f97697
JK		 4663 /* Update modification times of quota files when userspace can
4664 * start looking at them */
4665 handle = ext4_journal_start(inode, 1);
4666 if (IS_ERR(handle))
4667 goto out;
4668 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4669 ext4_mark_inode_dirty(handle, inode);
4670 ext4_journal_stop(handle);
4671
4672out:
ca0e05e4
DM		 4673 return dquot_quota_off(sb, type);
4674}
4675
ac27a0ec
DK		 4676/* Read data from quotafile - avoid pagecache and such because we cannot afford
4677 * acquiring the locks... As quota files are never truncated and quota code
25985edc 4678 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 4679 * we don't have to be afraid of races */
617ba13b 4680static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4681 size_t len, loff_t off)
4682{
4683 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4684 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4685 int err = 0;
4686 int offset = off & (sb->s_blocksize - 1);
4687 int tocopy;
4688 size_t toread;
4689 struct buffer_head *bh;
4690 loff_t i_size = i_size_read(inode);
4691
4692 if (off > i_size)
4693 return 0;
4694 if (off+len > i_size)
4695 len = i_size-off;
4696 toread = len;
4697 while (toread > 0) {
4698 tocopy = sb->s_blocksize - offset < toread ?
4699 sb->s_blocksize - offset : toread;
617ba13b 4700 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4701 if (err)
4702 return err;
4703 if (!bh) /* A hole? */
4704 memset(data, 0, tocopy);
4705 else
4706 memcpy(data, bh->b_data+offset, tocopy);
4707 brelse(bh);
4708 offset = 0;
4709 toread -= tocopy;
4710 data += tocopy;
4711 blk++;
4712 }
4713 return len;
4714}
4715
4716/* Write to quotafile (we know the transaction is already started and has
4717 * enough credits) */
617ba13b 4718static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4719 const char *data, size_t len, loff_t off)
4720{
4721 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4722 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4723 int err = 0;
4724 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 4725 struct buffer_head *bh;
4726 handle_t *handle = journal_current_handle();
4727
0390131b 4728 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4729 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4730 " cancelled because transaction is not started",
9c3013e9
JK		 4731 (unsigned long long)off, (unsigned long long)len);
4732 return -EIO;
4733 }
67eeb568
DM		 4734 /*
4735 * Since we account only one data block in transaction credits,
4736 * then it is impossible to cross a block boundary.
4737 */
4738 if (sb->s_blocksize - offset < len) {
4739 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4740 " cancelled because not block aligned",
4741 (unsigned long long)off, (unsigned long long)len);
4742 return -EIO;
4743 }
4744
ac27a0ec 4745 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4746 bh = ext4_bread(handle, inode, blk, 1, &err);
4747 if (!bh)
4748 goto out;
62d2b5f2
JK		 4749 err = ext4_journal_get_write_access(handle, bh);
4750 if (err) {
4751 brelse(bh);
4752 goto out;
ac27a0ec 4753 }
67eeb568
DM		 4754 lock_buffer(bh);
4755 memcpy(bh->b_data+offset, data, len);
4756 flush_dcache_page(bh->b_page);
4757 unlock_buffer(bh);
62d2b5f2 4758 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 4759 brelse(bh);
ac27a0ec 4760out:
67eeb568 4761 if (err) {
4d04e4fb 4762 mutex_unlock(&inode->i_mutex);
ac27a0ec 4763 return err;
4d04e4fb 4764 }
67eeb568
DM		 4765 if (inode->i_size < off + len) {
4766 i_size_write(inode, off + len);
617ba13b 4767 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 4768 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4769 }
ac27a0ec 4770 mutex_unlock(&inode->i_mutex);
67eeb568 4771 return len;
ac27a0ec
DK		 4772}
4773
4774#endif
4775
152a0836
AV		 4776static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4777 const char *dev_name, void *data)
ac27a0ec 4778{
152a0836 4779 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 4780}
4781
37f328eb 4782#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4783static inline void register_as_ext2(void)
4784{
4785 int err = register_filesystem(&ext2_fs_type);
4786 if (err)
4787 printk(KERN_WARNING
4788 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4789}
4790
4791static inline void unregister_as_ext2(void)
4792{
4793 unregister_filesystem(&ext2_fs_type);
4794}
2035e776
TT		 4795
4796static inline int ext2_feature_set_ok(struct super_block *sb)
4797{
4798 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4799 return 0;
4800 if (sb->s_flags & MS_RDONLY)
4801 return 1;
4802 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4803 return 0;
4804 return 1;
4805}
51b7e3c9 4806MODULE_ALIAS("ext2");
24b58424
TT		 4807#else
4808static inline void register_as_ext2(void) { }
4809static inline void unregister_as_ext2(void) { }
2035e776 4810static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4811#endif
4812
37f328eb 4813#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4814static inline void register_as_ext3(void)
4815{
4816 int err = register_filesystem(&ext3_fs_type);
4817 if (err)
4818 printk(KERN_WARNING
4819 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4820}
4821
4822static inline void unregister_as_ext3(void)
4823{
4824 unregister_filesystem(&ext3_fs_type);
4825}
2035e776
TT		 4826
4827static inline int ext3_feature_set_ok(struct super_block *sb)
4828{
4829 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4830 return 0;
4831 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4832 return 0;
4833 if (sb->s_flags & MS_RDONLY)
4834 return 1;
4835 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4836 return 0;
4837 return 1;
4838}
51b7e3c9 4839MODULE_ALIAS("ext3");
24b58424
TT		 4840#else
4841static inline void register_as_ext3(void) { }
4842static inline void unregister_as_ext3(void) { }
2035e776 4843static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4844#endif
4845
03010a33
TT		 4846static struct file_system_type ext4_fs_type = {
4847 .owner = THIS_MODULE,
4848 .name = "ext4",
152a0836 4849 .mount = ext4_mount,
03010a33
TT		 4850 .kill_sb = kill_block_super,
4851 .fs_flags = FS_REQUIRES_DEV,
4852};
4853
8f021222 4854static int __init ext4_init_feat_adverts(void)
857ac889
LC		 4855{
4856 struct ext4_features *ef;
4857 int ret = -ENOMEM;
4858
4859 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4860 if (!ef)
4861 goto out;
4862
4863 ef->f_kobj.kset = ext4_kset;
4864 init_completion(&ef->f_kobj_unregister);
4865 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4866 "features");
4867 if (ret) {
4868 kfree(ef);
4869 goto out;
4870 }
4871
4872 ext4_feat = ef;
4873 ret = 0;
4874out:
4875 return ret;
4876}
4877
8f021222
LC		 4878static void ext4_exit_feat_adverts(void)
4879{
4880 kobject_put(&ext4_feat->f_kobj);
4881 wait_for_completion(&ext4_feat->f_kobj_unregister);
4882 kfree(ext4_feat);
4883}
4884
e9e3bcec
ES		 4885/* Shared across all ext4 file systems */
4886wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4887struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4888
5dabfc78 4889static int __init ext4_init_fs(void)
ac27a0ec 4890{
e9e3bcec 4891 int i, err;
c9de560d 4892
07c0c5d8
AV		 4893 ext4_li_info = NULL;
4894 mutex_init(&ext4_li_mtx);
4895
12e9b892 4896 ext4_check_flag_values();
e9e3bcec
ES		 4897
4898 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4899 mutex_init(&ext4__aio_mutex[i]);
4900 init_waitqueue_head(&ext4__ioend_wq[i]);
4901 }
4902
5dabfc78 4903 err = ext4_init_pageio();
6fd058f7
TT		 4904 if (err)
4905 return err;
5dabfc78 4906 err = ext4_init_system_zone();
bd2d0210 4907 if (err)
d44651d0 4908 goto out6;
3197ebdb
TT		 4909 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4910 if (!ext4_kset)
dd68314c 4911 goto out5;
d44651d0 4912 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 4913
4914 err = ext4_init_feat_adverts();
dd68314c
TT		 4915 if (err)
4916 goto out4;
857ac889 4917
5dabfc78 4918 err = ext4_init_mballoc();
ac27a0ec 4919 if (err)
6fd058f7 4920 goto out3;
c9de560d 4921
5dabfc78 4922 err = ext4_init_xattr();
c9de560d
AT		 4923 if (err)
4924 goto out2;
ac27a0ec
DK		 4925 err = init_inodecache();
4926 if (err)
4927 goto out1;
24b58424 4928 register_as_ext3();
2035e776 4929 register_as_ext2();
03010a33 4930 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 4931 if (err)
4932 goto out;
bfff6873 4933
ac27a0ec
DK		 4934 return 0;
4935out:
24b58424
TT		 4936 unregister_as_ext2();
4937 unregister_as_ext3();
ac27a0ec
DK		 4938 destroy_inodecache();
4939out1:
5dabfc78 4940 ext4_exit_xattr();
c9de560d 4941out2:
5dabfc78 4942 ext4_exit_mballoc();
6fd058f7 4943out3:
8f021222 4944 ext4_exit_feat_adverts();
dd68314c 4945out4:
d44651d0
FJ		 4946 if (ext4_proc_root)
4947 remove_proc_entry("fs/ext4", NULL);
6fd058f7 4948 kset_unregister(ext4_kset);
d44651d0 4949out5:
5dabfc78 4950 ext4_exit_system_zone();
d44651d0 4951out6:
5dabfc78 4952 ext4_exit_pageio();
ac27a0ec
DK		 4953 return err;
4954}
4955
5dabfc78 4956static void __exit ext4_exit_fs(void)
ac27a0ec 4957{
bfff6873 4958 ext4_destroy_lazyinit_thread();
24b58424
TT		 4959 unregister_as_ext2();
4960 unregister_as_ext3();
03010a33 4961 unregister_filesystem(&ext4_fs_type);
ac27a0ec 4962 destroy_inodecache();
5dabfc78
TT		 4963 ext4_exit_xattr();
4964 ext4_exit_mballoc();
8f021222 4965 ext4_exit_feat_adverts();
9f6200bb 4966 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4967 kset_unregister(ext4_kset);
5dabfc78
TT		 4968 ext4_exit_system_zone();
4969 ext4_exit_pageio();
ac27a0ec
DK		 4970}
4971
4972MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4973MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4974MODULE_LICENSE("GPL");
5dabfc78
TT		 4975module_init(ext4_init_fs)
4976module_exit(ext4_exit_fs)
Linux 6.x block layer and io_uring tree(s)