projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: do not mark superblock as dirty unnecessarily
[deliverable/linux.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);
8c85e125 839 if (sb->s_dirt)
ebc1ac16 840 ext4_commit_super(sb, 1);
8c85e125 841
0390131b
FM		 842 if (sbi->s_journal) {
843 err = jbd2_journal_destroy(sbi->s_journal);
844 sbi->s_journal = NULL;
845 if (err < 0)
c67d859e 846 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 847 }
d4edac31 848
a1c6c569 849 del_timer(&sbi->s_err_report);
d4edac31
JB		 850 ext4_release_system_zone(sb);
851 ext4_mb_release(sb);
852 ext4_ext_release(sb);
853 ext4_xattr_put_super(sb);
854
ac27a0ec 855 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 856 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 857 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 858 ext4_commit_super(sb, 1);
ac27a0ec 859 }
240799cd 860 if (sbi->s_proc) {
66acdcf4 861 remove_proc_entry("options", sbi->s_proc);
9f6200bb 862 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 863 }
3197ebdb 864 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 865
866 for (i = 0; i < sbi->s_gdb_count; i++)
867 brelse(sbi->s_group_desc[i]);
f18a5f21 868 ext4_kvfree(sbi->s_group_desc);
9933fc0a 869 ext4_kvfree(sbi->s_flex_groups);
57042651 870 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 871 percpu_counter_destroy(&sbi->s_freeinodes_counter);
872 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 873 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 874 brelse(sbi->s_sbh);
875#ifdef CONFIG_QUOTA
876 for (i = 0; i < MAXQUOTAS; i++)
877 kfree(sbi->s_qf_names[i]);
878#endif
879
880 /* Debugging code just in case the in-memory inode orphan list
881 * isn't empty. The on-disk one can be non-empty if we've
882 * detected an error and taken the fs readonly, but the
883 * in-memory list had better be clean by this point. */
884 if (!list_empty(&sbi->s_orphan))
885 dump_orphan_list(sb, sbi);
886 J_ASSERT(list_empty(&sbi->s_orphan));
887
f98393a6 888 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 889 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
890 /*
891 * Invalidate the journal device's buffers. We don't want them
892 * floating about in memory - the physical journal device may
893 * hotswapped, and it breaks the `ro-after' testing code.
894 */
895 sync_blockdev(sbi->journal_bdev);
f98393a6 896 invalidate_bdev(sbi->journal_bdev);
617ba13b 897 ext4_blkdev_remove(sbi);
ac27a0ec 898 }
c5e06d10
JL		 899 if (sbi->s_mmp_tsk)
900 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 901 sb->s_fs_info = NULL;
3197ebdb
TT		 902 /*
903 * Now that we are completely done shutting down the
904 * superblock, we need to actually destroy the kobject.
905 */
3197ebdb
TT		 906 unlock_super(sb);
907 kobject_put(&sbi->s_kobj);
908 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 909 kfree(sbi->s_blockgroup_lock);
ac27a0ec 910 kfree(sbi);
ac27a0ec
DK		 911}
912
e18b890b 913static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 914
915/*
916 * Called inside transaction, so use GFP_NOFS
917 */
617ba13b 918static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 919{
617ba13b 920 struct ext4_inode_info *ei;
ac27a0ec 921
e6b4f8da 922 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 923 if (!ei)
924 return NULL;
0b8e58a1 925
ac27a0ec 926 ei->vfs_inode.i_version = 1;
91246c00 927 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 928 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 929 INIT_LIST_HEAD(&ei->i_prealloc_list);
930 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 931 ei->i_reserved_data_blocks = 0;
932 ei->i_reserved_meta_blocks = 0;
933 ei->i_allocated_meta_blocks = 0;
9d0be502 934 ei->i_da_metadata_calc_len = 0;
d2a17637 935 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 936#ifdef CONFIG_QUOTA
937 ei->i_reserved_quota = 0;
938#endif
8aefcd55 939 ei->jinode = NULL;
c7064ef1 940 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 941 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 942 ei->cur_aio_dio = NULL;
b436b9be
JK		 943 ei->i_sync_tid = 0;
944 ei->i_datasync_tid = 0;
f7ad6d2e 945 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 946 atomic_set(&ei->i_aiodio_unwritten, 0);
0b8e58a1 947
ac27a0ec
DK		 948 return &ei->vfs_inode;
949}
950
7ff9c073
TT		 951static int ext4_drop_inode(struct inode *inode)
952{
953 int drop = generic_drop_inode(inode);
954
955 trace_ext4_drop_inode(inode, drop);
956 return drop;
957}
958
fa0d7e3d
NP		 959static void ext4_i_callback(struct rcu_head *head)
960{
961 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 962 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
963}
964
617ba13b 965static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 966{
9f7dd93d 967 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 968 ext4_msg(inode->i_sb, KERN_ERR,
969 "Inode %lu (%p): orphan list check failed!",
970 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 971 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
972 EXT4_I(inode), sizeof(struct ext4_inode_info),
973 true);
974 dump_stack();
975 }
fa0d7e3d 976 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 977}
978
51cc5068 979static void init_once(void *foo)
ac27a0ec 980{
617ba13b 981 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 982
a35afb83 983 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 984#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 985 init_rwsem(&ei->xattr_sem);
ac27a0ec 986#endif
0e855ac8 987 init_rwsem(&ei->i_data_sem);
a35afb83 988 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 989}
990
991static int init_inodecache(void)
992{
617ba13b
MC		 993 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
994 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 995 0, (SLAB_RECLAIM_ACCOUNT|
996 SLAB_MEM_SPREAD),
20c2df83 997 init_once);
617ba13b 998 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 999 return -ENOMEM;
1000 return 0;
1001}
1002
1003static void destroy_inodecache(void)
1004{
617ba13b 1005 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1006}
1007
0930fcc1 1008void ext4_clear_inode(struct inode *inode)
ac27a0ec 1009{
0930fcc1
AV		 1010 invalidate_inode_buffers(inode);
1011 end_writeback(inode);
9f754758 1012 dquot_drop(inode);
c2ea3fde 1013 ext4_discard_preallocations(inode);
8aefcd55
TT		 1014 if (EXT4_I(inode)->jinode) {
1015 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1016 EXT4_I(inode)->jinode);
1017 jbd2_free_inode(EXT4_I(inode)->jinode);
1018 EXT4_I(inode)->jinode = NULL;
1019 }
ac27a0ec
DK		 1020}
1021
1b961ac0 1022static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1023 u64 ino, u32 generation)
ac27a0ec 1024{
ac27a0ec 1025 struct inode *inode;
ac27a0ec 1026
617ba13b 1027 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1028 return ERR_PTR(-ESTALE);
617ba13b 1029 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1030 return ERR_PTR(-ESTALE);
1031
1032 /* iget isn't really right if the inode is currently unallocated!!
1033 *
617ba13b 1034 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1035 * deleted, so we should be safe.
1036 *
1037 * Currently we don't know the generation for parent directory, so
1038 * a generation of 0 means "accept any"
1039 */
1d1fe1ee
DH		 1040 inode = ext4_iget(sb, ino);
1041 if (IS_ERR(inode))
1042 return ERR_CAST(inode);
1043 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1044 iput(inode);
1045 return ERR_PTR(-ESTALE);
1046 }
1b961ac0
CH		 1047
1048 return inode;
1049}
1050
1051static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1052 int fh_len, int fh_type)
1b961ac0
CH		 1053{
1054 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1055 ext4_nfs_get_inode);
1056}
1057
1058static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1059 int fh_len, int fh_type)
1b961ac0
CH		 1060{
1061 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1062 ext4_nfs_get_inode);
ac27a0ec
DK		 1063}
1064
c39a7f84
TO		 1065/*
1066 * Try to release metadata pages (indirect blocks, directories) which are
1067 * mapped via the block device. Since these pages could have journal heads
1068 * which would prevent try_to_free_buffers() from freeing them, we must use
1069 * jbd2 layer's try_to_free_buffers() function to release them.
1070 */
0b8e58a1
AD		 1071static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1072 gfp_t wait)
c39a7f84
TO		 1073{
1074 journal_t *journal = EXT4_SB(sb)->s_journal;
1075
1076 WARN_ON(PageChecked(page));
1077 if (!page_has_buffers(page))
1078 return 0;
1079 if (journal)
1080 return jbd2_journal_try_to_free_buffers(journal, page,
1081 wait & ~__GFP_WAIT);
1082 return try_to_free_buffers(page);
1083}
1084
ac27a0ec 1085#ifdef CONFIG_QUOTA
af5bc92d 1086#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1087#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1088
617ba13b
MC		 1089static int ext4_write_dquot(struct dquot *dquot);
1090static int ext4_acquire_dquot(struct dquot *dquot);
1091static int ext4_release_dquot(struct dquot *dquot);
1092static int ext4_mark_dquot_dirty(struct dquot *dquot);
1093static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1094static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1095 struct path *path);
ca0e05e4 1096static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1097static int ext4_quota_on_mount(struct super_block *sb, int type);
1098static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1099 size_t len, loff_t off);
617ba13b 1100static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1101 const char *data, size_t len, loff_t off);
1102
61e225dc 1103static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1104 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1105 .write_dquot = ext4_write_dquot,
1106 .acquire_dquot = ext4_acquire_dquot,
1107 .release_dquot = ext4_release_dquot,
1108 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1109 .write_info = ext4_write_info,
1110 .alloc_dquot = dquot_alloc,
1111 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1112};
1113
0d54b217 1114static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1115 .quota_on = ext4_quota_on,
ca0e05e4 1116 .quota_off = ext4_quota_off,
287a8095
CH		 1117 .quota_sync = dquot_quota_sync,
1118 .get_info = dquot_get_dqinfo,
1119 .set_info = dquot_set_dqinfo,
1120 .get_dqblk = dquot_get_dqblk,
1121 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1122};
1123#endif
1124
ee9b6d61 1125static const struct super_operations ext4_sops = {
617ba13b
MC		 1126 .alloc_inode = ext4_alloc_inode,
1127 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1128 .write_inode = ext4_write_inode,
1129 .dirty_inode = ext4_dirty_inode,
7ff9c073 1130 .drop_inode = ext4_drop_inode,
0930fcc1 1131 .evict_inode = ext4_evict_inode,
617ba13b 1132 .put_super = ext4_put_super,
617ba13b 1133 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1134 .freeze_fs = ext4_freeze,
1135 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1136 .statfs = ext4_statfs,
1137 .remount_fs = ext4_remount,
617ba13b 1138 .show_options = ext4_show_options,
ac27a0ec 1139#ifdef CONFIG_QUOTA
617ba13b
MC		 1140 .quota_read = ext4_quota_read,
1141 .quota_write = ext4_quota_write,
ac27a0ec 1142#endif
c39a7f84 1143 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1144};
1145
9ca92389
TT		 1146static const struct super_operations ext4_nojournal_sops = {
1147 .alloc_inode = ext4_alloc_inode,
1148 .destroy_inode = ext4_destroy_inode,
1149 .write_inode = ext4_write_inode,
1150 .dirty_inode = ext4_dirty_inode,
7ff9c073 1151 .drop_inode = ext4_drop_inode,
0930fcc1 1152 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1153 .write_super = ext4_write_super,
1154 .put_super = ext4_put_super,
1155 .statfs = ext4_statfs,
1156 .remount_fs = ext4_remount,
9ca92389
TT		 1157 .show_options = ext4_show_options,
1158#ifdef CONFIG_QUOTA
1159 .quota_read = ext4_quota_read,
1160 .quota_write = ext4_quota_write,
1161#endif
1162 .bdev_try_to_free_page = bdev_try_to_free_page,
1163};
1164
39655164 1165static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1166 .fh_to_dentry = ext4_fh_to_dentry,
1167 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1168 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1169};
1170
1171enum {
1172 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1173 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1174 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1175 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1176 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1177 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1178 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1179 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1180 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1181 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1182 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1183 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1184 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1185 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1186 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1187 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1188 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1189 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
ac27a0ec
DK		 1190};
1191
a447c093 1192static const match_table_t tokens = {
ac27a0ec
DK		 1193 {Opt_bsd_df, "bsddf"},
1194 {Opt_minix_df, "minixdf"},
1195 {Opt_grpid, "grpid"},
1196 {Opt_grpid, "bsdgroups"},
1197 {Opt_nogrpid, "nogrpid"},
1198 {Opt_nogrpid, "sysvgroups"},
1199 {Opt_resgid, "resgid=%u"},
1200 {Opt_resuid, "resuid=%u"},
1201 {Opt_sb, "sb=%u"},
1202 {Opt_err_cont, "errors=continue"},
1203 {Opt_err_panic, "errors=panic"},
1204 {Opt_err_ro, "errors=remount-ro"},
1205 {Opt_nouid32, "nouid32"},
ac27a0ec 1206 {Opt_debug, "debug"},
72578c33
TT		 1207 {Opt_removed, "oldalloc"},
1208 {Opt_removed, "orlov"},
ac27a0ec
DK		 1209 {Opt_user_xattr, "user_xattr"},
1210 {Opt_nouser_xattr, "nouser_xattr"},
1211 {Opt_acl, "acl"},
1212 {Opt_noacl, "noacl"},
e3bb52ae 1213 {Opt_noload, "norecovery"},
5a916be1 1214 {Opt_noload, "noload"},
72578c33
TT		 1215 {Opt_removed, "nobh"},
1216 {Opt_removed, "bh"},
ac27a0ec 1217 {Opt_commit, "commit=%u"},
30773840
TT		 1218 {Opt_min_batch_time, "min_batch_time=%u"},
1219 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1220 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1221 {Opt_journal_checksum, "journal_checksum"},
1222 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1223 {Opt_abort, "abort"},
1224 {Opt_data_journal, "data=journal"},
1225 {Opt_data_ordered, "data=ordered"},
1226 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1227 {Opt_data_err_abort, "data_err=abort"},
1228 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1229 {Opt_offusrjquota, "usrjquota="},
1230 {Opt_usrjquota, "usrjquota=%s"},
1231 {Opt_offgrpjquota, "grpjquota="},
1232 {Opt_grpjquota, "grpjquota=%s"},
1233 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1234 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1235 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1236 {Opt_grpquota, "grpquota"},
1237 {Opt_noquota, "noquota"},
1238 {Opt_quota, "quota"},
1239 {Opt_usrquota, "usrquota"},
1240 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1241 {Opt_barrier, "barrier"},
1242 {Opt_nobarrier, "nobarrier"},
25ec56b5 1243 {Opt_i_version, "i_version"},
c9de560d 1244 {Opt_stripe, "stripe=%u"},
64769240 1245 {Opt_delalloc, "delalloc"},
dd919b98 1246 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1247 {Opt_mblk_io_submit, "mblk_io_submit"},
1248 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1249 {Opt_block_validity, "block_validity"},
1250 {Opt_noblock_validity, "noblock_validity"},
240799cd 1251 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1252 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1253 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1254 {Opt_auto_da_alloc, "auto_da_alloc"},
1255 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1256 {Opt_dioread_nolock, "dioread_nolock"},
1257 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1258 {Opt_discard, "discard"},
1259 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1260 {Opt_init_itable, "init_itable=%u"},
1261 {Opt_init_itable, "init_itable"},
1262 {Opt_noinit_itable, "noinit_itable"},
c7198b9c
TT		 1263 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1264 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1265 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1266 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1267 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1268 {Opt_err, NULL},
ac27a0ec
DK		 1269};
1270
617ba13b 1271static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1272{
617ba13b 1273 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1274 char *options = (char *) *data;
1275
1276 if (!options || strncmp(options, "sb=", 3) != 0)
1277 return 1; /* Default location */
0b8e58a1 1278
ac27a0ec 1279 options += 3;
0b8e58a1 1280 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1281 sb_block = simple_strtoul(options, &options, 0);
1282 if (*options && *options != ',') {
4776004f 1283 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1284 (char *) *data);
1285 return 1;
1286 }
1287 if (*options == ',')
1288 options++;
1289 *data = (void *) options;
0b8e58a1 1290
ac27a0ec
DK		 1291 return sb_block;
1292}
1293
b3881f74 1294#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1295static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1296 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1297
56c50f11
DM		 1298#ifdef CONFIG_QUOTA
1299static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1300{
1301 struct ext4_sb_info *sbi = EXT4_SB(sb);
1302 char *qname;
1303
1304 if (sb_any_quota_loaded(sb) &&
1305 !sbi->s_qf_names[qtype]) {
1306 ext4_msg(sb, KERN_ERR,
1307 "Cannot change journaled "
1308 "quota options when quota turned on");
1309 return 0;
1310 }
1311 qname = match_strdup(args);
1312 if (!qname) {
1313 ext4_msg(sb, KERN_ERR,
1314 "Not enough memory for storing quotafile name");
1315 return 0;
1316 }
1317 if (sbi->s_qf_names[qtype] &&
1318 strcmp(sbi->s_qf_names[qtype], qname)) {
1319 ext4_msg(sb, KERN_ERR,
1320 "%s quota file already specified", QTYPE2NAME(qtype));
1321 kfree(qname);
1322 return 0;
1323 }
1324 sbi->s_qf_names[qtype] = qname;
1325 if (strchr(sbi->s_qf_names[qtype], '/')) {
1326 ext4_msg(sb, KERN_ERR,
1327 "quotafile must be on filesystem root");
1328 kfree(sbi->s_qf_names[qtype]);
1329 sbi->s_qf_names[qtype] = NULL;
1330 return 0;
1331 }
fd8c37ec 1332 set_opt(sb, QUOTA);
56c50f11
DM		 1333 return 1;
1334}
1335
1336static int clear_qf_name(struct super_block *sb, int qtype)
1337{
1338
1339 struct ext4_sb_info *sbi = EXT4_SB(sb);
1340
1341 if (sb_any_quota_loaded(sb) &&
1342 sbi->s_qf_names[qtype]) {
1343 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1344 " when quota turned on");
1345 return 0;
1346 }
1347 /*
1348 * The space will be released later when all options are confirmed
1349 * to be correct
1350 */
1351 sbi->s_qf_names[qtype] = NULL;
1352 return 1;
1353}
1354#endif
1355
26092bf5
TT		 1356#define MOPT_SET 0x0001
1357#define MOPT_CLEAR 0x0002
1358#define MOPT_NOSUPPORT 0x0004
1359#define MOPT_EXPLICIT 0x0008
1360#define MOPT_CLEAR_ERR 0x0010
1361#define MOPT_GTE0 0x0020
ac27a0ec 1362#ifdef CONFIG_QUOTA
26092bf5
TT		 1363#define MOPT_Q 0
1364#define MOPT_QFMT 0x0040
1365#else
1366#define MOPT_Q MOPT_NOSUPPORT
1367#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1368#endif
26092bf5
TT		 1369#define MOPT_DATAJ 0x0080
1370
1371static const struct mount_opts {
1372 int token;
1373 int mount_opt;
1374 int flags;
1375} ext4_mount_opts[] = {
1376 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1377 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1378 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1379 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1380 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1381 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1382 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1383 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1384 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1385 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1386 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1387 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1388 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1389 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1390 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1391 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1392 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1393 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1394 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1395 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1396 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1397 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1398 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1399 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1400 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1401 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1402 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1403 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1404 {Opt_commit, 0, MOPT_GTE0},
1405 {Opt_max_batch_time, 0, MOPT_GTE0},
1406 {Opt_min_batch_time, 0, MOPT_GTE0},
1407 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1408 {Opt_init_itable, 0, MOPT_GTE0},
1409 {Opt_stripe, 0, MOPT_GTE0},
1410 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1411 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1412 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
03010a33 1413#ifdef CONFIG_EXT4_FS_XATTR
26092bf5
TT		 1414 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1415 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
ac27a0ec 1416#else
26092bf5
TT		 1417 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1418 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
ac27a0ec 1419#endif
03010a33 1420#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1421 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1422 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1423#else
26092bf5
TT		 1424 {Opt_acl, 0, MOPT_NOSUPPORT},
1425 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1426#endif
26092bf5
TT		 1427 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1428 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1429 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1430 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1431 MOPT_SET | MOPT_Q},
1432 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1433 MOPT_SET | MOPT_Q},
1434 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1435 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1436 {Opt_usrjquota, 0, MOPT_Q},
1437 {Opt_grpjquota, 0, MOPT_Q},
1438 {Opt_offusrjquota, 0, MOPT_Q},
1439 {Opt_offgrpjquota, 0, MOPT_Q},
1440 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1441 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1442 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1443 {Opt_err, 0, 0}
1444};
1445
1446static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1447 substring_t *args, unsigned long *journal_devnum,
1448 unsigned int *journal_ioprio, int is_remount)
1449{
1450 struct ext4_sb_info *sbi = EXT4_SB(sb);
1451 const struct mount_opts *m;
1452 int arg = 0;
1453
1454 if (args->from && match_int(args, &arg))
1455 return -1;
1456 switch (token) {
f7048605
TT		 1457 case Opt_noacl:
1458 case Opt_nouser_xattr:
1459 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1460 break;
26092bf5
TT		 1461 case Opt_sb:
1462 return 1; /* handled by get_sb_block() */
1463 case Opt_removed:
1464 ext4_msg(sb, KERN_WARNING,
1465 "Ignoring removed %s option", opt);
1466 return 1;
1467 case Opt_resuid:
1468 sbi->s_resuid = arg;
1469 return 1;
1470 case Opt_resgid:
1471 sbi->s_resgid = arg;
1472 return 1;
1473 case Opt_abort:
1474 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1475 return 1;
1476 case Opt_i_version:
1477 sb->s_flags |= MS_I_VERSION;
1478 return 1;
1479 case Opt_journal_dev:
1480 if (is_remount) {
1481 ext4_msg(sb, KERN_ERR,
1482 "Cannot specify journal on remount");
1483 return -1;
1484 }
1485 *journal_devnum = arg;
1486 return 1;
1487 case Opt_journal_ioprio:
1488 if (arg < 0 || arg > 7)
1489 return -1;
1490 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1491 return 1;
1492 }
1493
1494 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1495 if (token != m->token)
1496 continue;
1497 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1498 return -1;
1499 if (m->flags & MOPT_EXPLICIT)
1500 set_opt2(sb, EXPLICIT_DELALLOC);
1501 if (m->flags & MOPT_CLEAR_ERR)
1502 clear_opt(sb, ERRORS_MASK);
1503 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1504 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1505 "options when quota turned on");
1506 return -1;
1507 }
1508
1509 if (m->flags & MOPT_NOSUPPORT) {
1510 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1511 } else if (token == Opt_commit) {
1512 if (arg == 0)
1513 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1514 sbi->s_commit_interval = HZ * arg;
1515 } else if (token == Opt_max_batch_time) {
1516 if (arg == 0)
1517 arg = EXT4_DEF_MAX_BATCH_TIME;
1518 sbi->s_max_batch_time = arg;
1519 } else if (token == Opt_min_batch_time) {
1520 sbi->s_min_batch_time = arg;
1521 } else if (token == Opt_inode_readahead_blks) {
1522 if (arg > (1 << 30))
1523 return -1;
1524 if (arg && !is_power_of_2(arg)) {
b31e1552 1525 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1526 "EXT4-fs: inode_readahead_blks"
1527 " must be a power of 2");
1528 return -1;
ac27a0ec 1529 }
26092bf5
TT		 1530 sbi->s_inode_readahead_blks = arg;
1531 } else if (token == Opt_init_itable) {
1532 set_opt(sb, INIT_INODE_TABLE);
1533 if (!args->from)
1534 arg = EXT4_DEF_LI_WAIT_MULT;
1535 sbi->s_li_wait_mult = arg;
1536 } else if (token == Opt_stripe) {
1537 sbi->s_stripe = arg;
1538 } else if (m->flags & MOPT_DATAJ) {
ac27a0ec 1539 if (is_remount) {
eb513689
TT		 1540 if (!sbi->s_journal)
1541 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
26092bf5
TT		 1542 else if (test_opt(sb, DATA_FLAGS) !=
1543 m->mount_opt) {
b31e1552 1544 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1545 "Cannot change data mode on remount");
1546 return -1;
ac27a0ec
DK		 1547 }
1548 } else {
fd8c37ec 1549 clear_opt(sb, DATA_FLAGS);
26092bf5 1550 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1551 }
ac27a0ec 1552#ifdef CONFIG_QUOTA
26092bf5 1553 } else if (token == Opt_usrjquota) {
56c50f11 1554 if (!set_qf_name(sb, USRQUOTA, &args[0]))
26092bf5
TT		 1555 return -1;
1556 } else if (token == Opt_grpjquota) {
56c50f11 1557 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
26092bf5
TT		 1558 return -1;
1559 } else if (token == Opt_offusrjquota) {
56c50f11 1560 if (!clear_qf_name(sb, USRQUOTA))
26092bf5
TT		 1561 return -1;
1562 } else if (token == Opt_offgrpjquota) {
56c50f11 1563 if (!clear_qf_name(sb, GRPQUOTA))
26092bf5
TT		 1564 return -1;
1565 } else if (m->flags & MOPT_QFMT) {
17bd13b3 1566 if (sb_any_quota_loaded(sb) &&
26092bf5
TT		 1567 sbi->s_jquota_fmt != m->mount_opt) {
1568 ext4_msg(sb, KERN_ERR, "Cannot "
1569 "change journaled quota options "
1570 "when quota turned on");
1571 return -1;
dfc5d03f 1572 }
26092bf5 1573 sbi->s_jquota_fmt = m->mount_opt;
ac27a0ec 1574#endif
26092bf5
TT		 1575 } else {
1576 if (!args->from)
1577 arg = 1;
1578 if (m->flags & MOPT_CLEAR)
1579 arg = !arg;
1580 else if (unlikely(!(m->flags & MOPT_SET))) {
1581 ext4_msg(sb, KERN_WARNING,
1582 "buggy handling of option %s", opt);
1583 WARN_ON(1);
1584 return -1;
3197ebdb 1585 }
26092bf5
TT		 1586 if (arg != 0)
1587 sbi->s_mount_opt |= m->mount_opt;
afd4672d 1588 else
26092bf5 1589 sbi->s_mount_opt &= ~m->mount_opt;
ac27a0ec 1590 }
26092bf5
TT		 1591 return 1;
1592 }
1593 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1594 "or missing value", opt);
1595 return -1;
1596}
1597
1598static int parse_options(char *options, struct super_block *sb,
1599 unsigned long *journal_devnum,
1600 unsigned int *journal_ioprio,
1601 int is_remount)
1602{
1603 struct ext4_sb_info *sbi = EXT4_SB(sb);
1604 char *p;
1605 substring_t args[MAX_OPT_ARGS];
1606 int token;
1607
1608 if (!options)
1609 return 1;
1610
1611 while ((p = strsep(&options, ",")) != NULL) {
1612 if (!*p)
1613 continue;
1614 /*
1615 * Initialize args struct so we know whether arg was
1616 * found; some options take optional arguments.
1617 */
1618 args[0].to = args[0].from = 0;
1619 token = match_token(p, tokens, args);
1620 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1621 journal_ioprio, is_remount) < 0)
1622 return 0;
ac27a0ec
DK		 1623 }
1624#ifdef CONFIG_QUOTA
1625 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1626 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1627 clear_opt(sb, USRQUOTA);
ac27a0ec 1628
482a7425 1629 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1630 clear_opt(sb, GRPQUOTA);
ac27a0ec 1631
56c50f11 1632 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1633 ext4_msg(sb, KERN_ERR, "old and new quota "
1634 "format mixing");
ac27a0ec
DK		 1635 return 0;
1636 }
1637
1638 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1639 ext4_msg(sb, KERN_ERR, "journaled quota format "
1640 "not specified");
ac27a0ec
DK		 1641 return 0;
1642 }
1643 } else {
1644 if (sbi->s_jquota_fmt) {
b31e1552 1645 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1646 "specified with no journaling "
b31e1552 1647 "enabled");
ac27a0ec
DK		 1648 return 0;
1649 }
1650 }
1651#endif
1652 return 1;
1653}
1654
2adf6da8
TT		 1655static inline void ext4_show_quota_options(struct seq_file *seq,
1656 struct super_block *sb)
1657{
1658#if defined(CONFIG_QUOTA)
1659 struct ext4_sb_info *sbi = EXT4_SB(sb);
1660
1661 if (sbi->s_jquota_fmt) {
1662 char *fmtname = "";
1663
1664 switch (sbi->s_jquota_fmt) {
1665 case QFMT_VFS_OLD:
1666 fmtname = "vfsold";
1667 break;
1668 case QFMT_VFS_V0:
1669 fmtname = "vfsv0";
1670 break;
1671 case QFMT_VFS_V1:
1672 fmtname = "vfsv1";
1673 break;
1674 }
1675 seq_printf(seq, ",jqfmt=%s", fmtname);
1676 }
1677
1678 if (sbi->s_qf_names[USRQUOTA])
1679 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1680
1681 if (sbi->s_qf_names[GRPQUOTA])
1682 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1683
1684 if (test_opt(sb, USRQUOTA))
1685 seq_puts(seq, ",usrquota");
1686
1687 if (test_opt(sb, GRPQUOTA))
1688 seq_puts(seq, ",grpquota");
1689#endif
1690}
1691
5a916be1
TT		 1692static const char *token2str(int token)
1693{
1694 static const struct match_token *t;
1695
1696 for (t = tokens; t->token != Opt_err; t++)
1697 if (t->token == token && !strchr(t->pattern, '='))
1698 break;
1699 return t->pattern;
1700}
1701
2adf6da8
TT		 1702/*
1703 * Show an option if
1704 * - it's set to a non-default value OR
1705 * - if the per-sb default is different from the global default
1706 */
66acdcf4
TT		 1707static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1708 int nodefs)
2adf6da8 1709{
2adf6da8
TT		 1710 struct ext4_sb_info *sbi = EXT4_SB(sb);
1711 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1712 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1713 const struct mount_opts *m;
66acdcf4 1714 char sep = nodefs ? '\n' : ',';
2adf6da8 1715
66acdcf4
TT		 1716#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1717#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1718
1719 if (sbi->s_sb_block != 1)
5a916be1
TT		 1720 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1721
1722 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1723 int want_set = m->flags & MOPT_SET;
1724 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1725 (m->flags & MOPT_CLEAR_ERR))
1726 continue;
66acdcf4 1727 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1728 continue; /* skip if same as the default */
1729 if ((want_set &&
1730 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1731 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1732 continue; /* select Opt_noFoo vs Opt_Foo */
1733 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1734 }
5a916be1 1735
66acdcf4 1736 if (nodefs || sbi->s_resuid != EXT4_DEF_RESUID ||
5a916be1
TT		 1737 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1738 SEQ_OPTS_PRINT("resuid=%u", sbi->s_resuid);
66acdcf4 1739 if (nodefs || sbi->s_resgid != EXT4_DEF_RESGID ||
5a916be1
TT		 1740 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1741 SEQ_OPTS_PRINT("resgid=%u", sbi->s_resgid);
66acdcf4 1742 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1743 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1744 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1745 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1746 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1747 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1748 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1749 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1750 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1751 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1752 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1753 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1754 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1755 if (sb->s_flags & MS_I_VERSION)
5a916be1 1756 SEQ_OPTS_PUTS("i_version");
66acdcf4 1757 if (nodefs || sbi->s_stripe)
5a916be1 1758 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1759 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1760 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1761 SEQ_OPTS_PUTS("data=journal");
1762 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1763 SEQ_OPTS_PUTS("data=ordered");
1764 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1765 SEQ_OPTS_PUTS("data=writeback");
1766 }
66acdcf4
TT		 1767 if (nodefs ||
1768 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1769 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1770 sbi->s_inode_readahead_blks);
2adf6da8 1771
66acdcf4
TT		 1772 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1773 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1774 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2adf6da8
TT		 1775
1776 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1777 return 0;
1778}
1779
66acdcf4
TT		 1780static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1781{
1782 return _ext4_show_options(seq, root->d_sb, 0);
1783}
1784
1785static int options_seq_show(struct seq_file *seq, void *offset)
1786{
1787 struct super_block *sb = seq->private;
1788 int rc;
1789
1790 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1791 rc = _ext4_show_options(seq, sb, 1);
1792 seq_puts(seq, "\n");
1793 return rc;
1794}
1795
1796static int options_open_fs(struct inode *inode, struct file *file)
1797{
1798 return single_open(file, options_seq_show, PDE(inode)->data);
1799}
1800
1801static const struct file_operations ext4_seq_options_fops = {
1802 .owner = THIS_MODULE,
1803 .open = options_open_fs,
1804 .read = seq_read,
1805 .llseek = seq_lseek,
1806 .release = single_release,
1807};
1808
617ba13b 1809static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1810 int read_only)
1811{
617ba13b 1812 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1813 int res = 0;
1814
617ba13b 1815 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1816 ext4_msg(sb, KERN_ERR, "revision level too high, "
1817 "forcing read-only mode");
ac27a0ec
DK		 1818 res = MS_RDONLY;
1819 }
1820 if (read_only)
281b5995 1821 goto done;
617ba13b 1822 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1823 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1824 "running e2fsck is recommended");
617ba13b 1825 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1826 ext4_msg(sb, KERN_WARNING,
1827 "warning: mounting fs with errors, "
1828 "running e2fsck is recommended");
ed3ce80a 1829 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1830 le16_to_cpu(es->s_mnt_count) >=
1831 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1832 ext4_msg(sb, KERN_WARNING,
1833 "warning: maximal mount count reached, "
1834 "running e2fsck is recommended");
ac27a0ec
DK		 1835 else if (le32_to_cpu(es->s_checkinterval) &&
1836 (le32_to_cpu(es->s_lastcheck) +
1837 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1838 ext4_msg(sb, KERN_WARNING,
1839 "warning: checktime reached, "
1840 "running e2fsck is recommended");
0b8e58a1 1841 if (!sbi->s_journal)
0390131b 1842 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1843 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1844 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1845 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1846 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1847 ext4_update_dynamic_rev(sb);
0390131b
FM		 1848 if (sbi->s_journal)
1849 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1850
e2d67052 1851 ext4_commit_super(sb, 1);
281b5995 1852done:
ac27a0ec 1853 if (test_opt(sb, DEBUG))
a9df9a49 1854 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1855 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1856 sb->s_blocksize,
1857 sbi->s_groups_count,
617ba13b
MC		 1858 EXT4_BLOCKS_PER_GROUP(sb),
1859 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1860 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1861
7abc52c2 1862 cleancache_init_fs(sb);
ac27a0ec
DK		 1863 return res;
1864}
1865
772cb7c8
JS		 1866static int ext4_fill_flex_info(struct super_block *sb)
1867{
1868 struct ext4_sb_info *sbi = EXT4_SB(sb);
1869 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1870 ext4_group_t flex_group_count;
1871 ext4_group_t flex_group;
d50f2ab6 1872 unsigned int groups_per_flex = 0;
c5ca7c76 1873 size_t size;
772cb7c8
JS		 1874 int i;
1875
503358ae 1876 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1877 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1878 sbi->s_log_groups_per_flex = 0;
1879 return 1;
1880 }
d50f2ab6 1881 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
772cb7c8 1882
c62a11fd
FB		 1883 /* We allocate both existing and potentially added groups */
1884 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1885 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1886 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76 1887 size = flex_group_count * sizeof(struct flex_groups);
9933fc0a 1888 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
c5ca7c76 1889 if (sbi->s_flex_groups == NULL) {
9933fc0a
TT		 1890 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1891 flex_group_count);
1892 goto failed;
772cb7c8 1893 }
772cb7c8 1894
772cb7c8 1895 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1896 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1897
1898 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1899 atomic_add(ext4_free_inodes_count(sb, gdp),
1900 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 1901 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 1902 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1903 atomic_add(ext4_used_dirs_count(sb, gdp),
1904 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1905 }
1906
1907 return 1;
1908failed:
1909 return 0;
1910}
1911
717d50e4
AD		 1912__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1913 struct ext4_group_desc *gdp)
1914{
1915 __u16 crc = 0;
1916
1917 if (sbi->s_es->s_feature_ro_compat &
1918 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1919 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1920 __le32 le_group = cpu_to_le32(block_group);
1921
1922 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1923 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1924 crc = crc16(crc, (__u8 *)gdp, offset);
1925 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1926 /* for checksum of struct ext4_group_desc do the rest...*/
1927 if ((sbi->s_es->s_feature_incompat &
1928 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1929 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1930 crc = crc16(crc, (__u8 *)gdp + offset,
1931 le16_to_cpu(sbi->s_es->s_desc_size) -
1932 offset);
1933 }
1934
1935 return cpu_to_le16(crc);
1936}
1937
1938int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1939 struct ext4_group_desc *gdp)
1940{
1941 if ((sbi->s_es->s_feature_ro_compat &
1942 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1943 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1944 return 0;
1945
1946 return 1;
1947}
1948
ac27a0ec 1949/* Called at mount-time, super-block is locked */
bfff6873
LC		 1950static int ext4_check_descriptors(struct super_block *sb,
1951 ext4_group_t *first_not_zeroed)
ac27a0ec 1952{
617ba13b
MC		 1953 struct ext4_sb_info *sbi = EXT4_SB(sb);
1954 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1955 ext4_fsblk_t last_block;
bd81d8ee
LV		 1956 ext4_fsblk_t block_bitmap;
1957 ext4_fsblk_t inode_bitmap;
1958 ext4_fsblk_t inode_table;
ce421581 1959 int flexbg_flag = 0;
bfff6873 1960 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 1961
ce421581
JS		 1962 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1963 flexbg_flag = 1;
1964
af5bc92d 1965 ext4_debug("Checking group descriptors");
ac27a0ec 1966
197cd65a
AM		 1967 for (i = 0; i < sbi->s_groups_count; i++) {
1968 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1969
ce421581 1970 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1971 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1972 else
1973 last_block = first_block +
617ba13b 1974 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1975
bfff6873
LC		 1976 if ((grp == sbi->s_groups_count) &&
1977 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
1978 grp = i;
1979
8fadc143 1980 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1981 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 1982 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1983 "Block bitmap for group %u not in group "
b31e1552 1984 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1985 return 0;
1986 }
8fadc143 1987 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1988 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 1989 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1990 "Inode bitmap for group %u not in group "
b31e1552 1991 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1992 return 0;
1993 }
8fadc143 1994 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1995 if (inode_table < first_block ||
2b2d6d01 1996 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 1997 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1998 "Inode table for group %u not in group "
b31e1552 1999 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2000 return 0;
2001 }
955ce5f5 2002 ext4_lock_group(sb, i);
717d50e4 2003 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 2004 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2005 "Checksum for group %u failed (%u!=%u)",
2006 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2007 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2008 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2009 ext4_unlock_group(sb, i);
8a266467 2010 return 0;
7ee1ec4c 2011 }
717d50e4 2012 }
955ce5f5 2013 ext4_unlock_group(sb, i);
ce421581
JS		 2014 if (!flexbg_flag)
2015 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2016 }
bfff6873
LC		 2017 if (NULL != first_not_zeroed)
2018 *first_not_zeroed = grp;
ac27a0ec 2019
5dee5437
TT		 2020 ext4_free_blocks_count_set(sbi->s_es,
2021 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2022 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2023 return 1;
2024}
2025
617ba13b 2026/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2027 * the superblock) which were deleted from all directories, but held open by
2028 * a process at the time of a crash. We walk the list and try to delete these
2029 * inodes at recovery time (only with a read-write filesystem).
2030 *
2031 * In order to keep the orphan inode chain consistent during traversal (in
2032 * case of crash during recovery), we link each inode into the superblock
2033 * orphan list_head and handle it the same way as an inode deletion during
2034 * normal operation (which journals the operations for us).
2035 *
2036 * We only do an iget() and an iput() on each inode, which is very safe if we
2037 * accidentally point at an in-use or already deleted inode. The worst that
2038 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2039 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2040 * e2fsck was run on this filesystem, and it must have already done the orphan
2041 * inode cleanup for us, so we can safely abort without any further action.
2042 */
2b2d6d01
TT		 2043static void ext4_orphan_cleanup(struct super_block *sb,
2044 struct ext4_super_block *es)
ac27a0ec
DK		 2045{
2046 unsigned int s_flags = sb->s_flags;
2047 int nr_orphans = 0, nr_truncates = 0;
2048#ifdef CONFIG_QUOTA
2049 int i;
2050#endif
2051 if (!es->s_last_orphan) {
2052 jbd_debug(4, "no orphan inodes to clean up\n");
2053 return;
2054 }
2055
a8f48a95 2056 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2057 ext4_msg(sb, KERN_ERR, "write access "
2058 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2059 return;
2060 }
2061
d39195c3
AG		 2062 /* Check if feature set would not allow a r/w mount */
2063 if (!ext4_feature_set_ok(sb, 0)) {
2064 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2065 "unknown ROCOMPAT features");
2066 return;
2067 }
2068
617ba13b 2069 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2070 if (es->s_last_orphan)
2071 jbd_debug(1, "Errors on filesystem, "
2072 "clearing orphan list.\n");
2073 es->s_last_orphan = 0;
2074 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2075 return;
2076 }
2077
2078 if (s_flags & MS_RDONLY) {
b31e1552 2079 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2080 sb->s_flags &= ~MS_RDONLY;
2081 }
2082#ifdef CONFIG_QUOTA
2083 /* Needed for iput() to work correctly and not trash data */
2084 sb->s_flags |= MS_ACTIVE;
2085 /* Turn on quotas so that they are updated correctly */
2086 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2087 if (EXT4_SB(sb)->s_qf_names[i]) {
2088 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2089 if (ret < 0)
b31e1552
ES		 2090 ext4_msg(sb, KERN_ERR,
2091 "Cannot turn on journaled "
2092 "quota: error %d", ret);
ac27a0ec
DK		 2093 }
2094 }
2095#endif
2096
2097 while (es->s_last_orphan) {
2098 struct inode *inode;
2099
97bd42b9
JB		 2100 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2101 if (IS_ERR(inode)) {
ac27a0ec
DK		 2102 es->s_last_orphan = 0;
2103 break;
2104 }
2105
617ba13b 2106 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2107 dquot_initialize(inode);
ac27a0ec 2108 if (inode->i_nlink) {
b31e1552
ES		 2109 ext4_msg(sb, KERN_DEBUG,
2110 "%s: truncating inode %lu to %lld bytes",
46e665e9 2111 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2112 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2113 inode->i_ino, inode->i_size);
617ba13b 2114 ext4_truncate(inode);
ac27a0ec
DK		 2115 nr_truncates++;
2116 } else {
b31e1552
ES		 2117 ext4_msg(sb, KERN_DEBUG,
2118 "%s: deleting unreferenced inode %lu",
46e665e9 2119 __func__, inode->i_ino);
ac27a0ec
DK		 2120 jbd_debug(2, "deleting unreferenced inode %lu\n",
2121 inode->i_ino);
2122 nr_orphans++;
2123 }
2124 iput(inode); /* The delete magic happens here! */
2125 }
2126
2b2d6d01 2127#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2128
2129 if (nr_orphans)
b31e1552
ES		 2130 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2131 PLURAL(nr_orphans));
ac27a0ec 2132 if (nr_truncates)
b31e1552
ES		 2133 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2134 PLURAL(nr_truncates));
ac27a0ec
DK		 2135#ifdef CONFIG_QUOTA
2136 /* Turn quotas off */
2137 for (i = 0; i < MAXQUOTAS; i++) {
2138 if (sb_dqopt(sb)->files[i])
287a8095 2139 dquot_quota_off(sb, i);
ac27a0ec
DK		 2140 }
2141#endif
2142 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2143}
0b8e58a1 2144
cd2291a4
ES		 2145/*
2146 * Maximal extent format file size.
2147 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2148 * extent format containers, within a sector_t, and within i_blocks
2149 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2150 * so that won't be a limiting factor.
2151 *
f17722f9
LC		 2152 * However there is other limiting factor. We do store extents in the form
2153 * of starting block and length, hence the resulting length of the extent
2154 * covering maximum file size must fit into on-disk format containers as
2155 * well. Given that length is always by 1 unit bigger than max unit (because
2156 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2157 *
cd2291a4
ES		 2158 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2159 */
f287a1a5 2160static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2161{
2162 loff_t res;
2163 loff_t upper_limit = MAX_LFS_FILESIZE;
2164
2165 /* small i_blocks in vfs inode? */
f287a1a5 2166 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2167 /*
90c699a9 2168 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2169 * i_block represent total blocks in 512 bytes
2170 * 32 == size of vfs inode i_blocks * 8
2171 */
2172 upper_limit = (1LL << 32) - 1;
2173
2174 /* total blocks in file system block size */
2175 upper_limit >>= (blkbits - 9);
2176 upper_limit <<= blkbits;
2177 }
2178
f17722f9
LC		 2179 /*
2180 * 32-bit extent-start container, ee_block. We lower the maxbytes
2181 * by one fs block, so ee_len can cover the extent of maximum file
2182 * size
2183 */
2184 res = (1LL << 32) - 1;
cd2291a4 2185 res <<= blkbits;
cd2291a4
ES		 2186
2187 /* Sanity check against vm- & vfs- imposed limits */
2188 if (res > upper_limit)
2189 res = upper_limit;
2190
2191 return res;
2192}
ac27a0ec 2193
ac27a0ec 2194/*
cd2291a4 2195 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2196 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2197 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2198 */
f287a1a5 2199static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2200{
617ba13b 2201 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2202 int meta_blocks;
2203 loff_t upper_limit;
0b8e58a1
AD		 2204 /* This is calculated to be the largest file size for a dense, block
2205 * mapped file such that the file's total number of 512-byte sectors,
2206 * including data and all indirect blocks, does not exceed (2^48 - 1).
2207 *
2208 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2209 * number of 512-byte sectors of the file.
0fc1b451
AK		 2210 */
2211
f287a1a5 2212 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2213 /*
90c699a9 2214 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2215 * the inode i_block field represents total file blocks in
2216 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2217 */
2218 upper_limit = (1LL << 32) - 1;
2219
2220 /* total blocks in file system block size */
2221 upper_limit >>= (bits - 9);
2222
2223 } else {
8180a562
AK		 2224 /*
2225 * We use 48 bit ext4_inode i_blocks
2226 * With EXT4_HUGE_FILE_FL set the i_blocks
2227 * represent total number of blocks in
2228 * file system block size
2229 */
0fc1b451
AK		 2230 upper_limit = (1LL << 48) - 1;
2231
0fc1b451
AK		 2232 }
2233
2234 /* indirect blocks */
2235 meta_blocks = 1;
2236 /* double indirect blocks */
2237 meta_blocks += 1 + (1LL << (bits-2));
2238 /* tripple indirect blocks */
2239 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2240
2241 upper_limit -= meta_blocks;
2242 upper_limit <<= bits;
ac27a0ec
DK		 2243
2244 res += 1LL << (bits-2);
2245 res += 1LL << (2*(bits-2));
2246 res += 1LL << (3*(bits-2));
2247 res <<= bits;
2248 if (res > upper_limit)
2249 res = upper_limit;
0fc1b451
AK		 2250
2251 if (res > MAX_LFS_FILESIZE)
2252 res = MAX_LFS_FILESIZE;
2253
ac27a0ec
DK		 2254 return res;
2255}
2256
617ba13b 2257static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2258 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2259{
617ba13b 2260 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2261 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2262 int has_super = 0;
2263
2264 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2265
617ba13b 2266 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2267 nr < first_meta_bg)
70bbb3e0 2268 return logical_sb_block + nr + 1;
ac27a0ec 2269 bg = sbi->s_desc_per_block * nr;
617ba13b 2270 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2271 has_super = 1;
0b8e58a1 2272
617ba13b 2273 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2274}
2275
c9de560d
AT		 2276/**
2277 * ext4_get_stripe_size: Get the stripe size.
2278 * @sbi: In memory super block info
2279 *
2280 * If we have specified it via mount option, then
2281 * use the mount option value. If the value specified at mount time is
2282 * greater than the blocks per group use the super block value.
2283 * If the super block value is greater than blocks per group return 0.
2284 * Allocator needs it be less than blocks per group.
2285 *
2286 */
2287static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2288{
2289 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2290 unsigned long stripe_width =
2291 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2292 int ret;
c9de560d
AT		 2293
2294 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2295 ret = sbi->s_stripe;
2296 else if (stripe_width <= sbi->s_blocks_per_group)
2297 ret = stripe_width;
2298 else if (stride <= sbi->s_blocks_per_group)
2299 ret = stride;
2300 else
2301 ret = 0;
c9de560d 2302
3eb08658
DE		 2303 /*
2304 * If the stripe width is 1, this makes no sense and
2305 * we set it to 0 to turn off stripe handling code.
2306 */
2307 if (ret <= 1)
2308 ret = 0;
c9de560d 2309
3eb08658 2310 return ret;
c9de560d 2311}
ac27a0ec 2312
3197ebdb
TT		 2313/* sysfs supprt */
2314
2315struct ext4_attr {
2316 struct attribute attr;
2317 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2318 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2319 const char *, size_t);
2320 int offset;
2321};
2322
2323static int parse_strtoul(const char *buf,
2324 unsigned long max, unsigned long *value)
2325{
2326 char *endp;
2327
e7d2860b
AGR		 2328 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2329 endp = skip_spaces(endp);
3197ebdb
TT		 2330 if (*endp || *value > max)
2331 return -EINVAL;
2332
2333 return 0;
2334}
2335
2336static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2337 struct ext4_sb_info *sbi,
2338 char *buf)
2339{
2340 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2341 (s64) EXT4_C2B(sbi,
2342 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2343}
2344
2345static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2346 struct ext4_sb_info *sbi, char *buf)
2347{
2348 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2349
f613dfcb
TT		 2350 if (!sb->s_bdev->bd_part)
2351 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2352 return snprintf(buf, PAGE_SIZE, "%lu\n",
2353 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2354 sbi->s_sectors_written_start) >> 1);
2355}
2356
2357static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2358 struct ext4_sb_info *sbi, char *buf)
2359{
2360 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2361
f613dfcb
TT		 2362 if (!sb->s_bdev->bd_part)
2363 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2364 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2365 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2366 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2367 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2368}
2369
77f4135f
VH		 2370static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2371 struct ext4_sb_info *sbi, char *buf)
2372{
2373 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2374}
2375
2376static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2377 struct ext4_sb_info *sbi, char *buf)
2378{
2379 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2380}
2381
3197ebdb
TT		 2382static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2383 struct ext4_sb_info *sbi,
2384 const char *buf, size_t count)
2385{
2386 unsigned long t;
2387
2388 if (parse_strtoul(buf, 0x40000000, &t))
2389 return -EINVAL;
2390
5dbd571d 2391 if (t && !is_power_of_2(t))
3197ebdb
TT		 2392 return -EINVAL;
2393
2394 sbi->s_inode_readahead_blks = t;
2395 return count;
2396}
2397
2398static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2399 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2400{
2401 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2402
2403 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2404}
2405
2406static ssize_t sbi_ui_store(struct ext4_attr *a,
2407 struct ext4_sb_info *sbi,
2408 const char *buf, size_t count)
2409{
2410 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2411 unsigned long t;
2412
2413 if (parse_strtoul(buf, 0xffffffff, &t))
2414 return -EINVAL;
2415 *ui = t;
2416 return count;
2417}
2418
2419#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2420static struct ext4_attr ext4_attr_##_name = { \
2421 .attr = {.name = __stringify(_name), .mode = _mode }, \
2422 .show = _show, \
2423 .store = _store, \
2424 .offset = offsetof(struct ext4_sb_info, _elname), \
2425}
2426#define EXT4_ATTR(name, mode, show, store) \
2427static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2428
857ac889 2429#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2430#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2431#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2432#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2433 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2434#define ATTR_LIST(name) &ext4_attr_##name.attr
2435
2436EXT4_RO_ATTR(delayed_allocation_blocks);
2437EXT4_RO_ATTR(session_write_kbytes);
2438EXT4_RO_ATTR(lifetime_write_kbytes);
77f4135f
VH		 2439EXT4_RO_ATTR(extent_cache_hits);
2440EXT4_RO_ATTR(extent_cache_misses);
3197ebdb
TT		 2441EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2442 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2443EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2444EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2445EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2446EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2447EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2448EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2449EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2450EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2451
2452static struct attribute *ext4_attrs[] = {
2453 ATTR_LIST(delayed_allocation_blocks),
2454 ATTR_LIST(session_write_kbytes),
2455 ATTR_LIST(lifetime_write_kbytes),
77f4135f
VH		 2456 ATTR_LIST(extent_cache_hits),
2457 ATTR_LIST(extent_cache_misses),
3197ebdb 2458 ATTR_LIST(inode_readahead_blks),
11013911 2459 ATTR_LIST(inode_goal),
3197ebdb
TT		 2460 ATTR_LIST(mb_stats),
2461 ATTR_LIST(mb_max_to_scan),
2462 ATTR_LIST(mb_min_to_scan),
2463 ATTR_LIST(mb_order2_req),
2464 ATTR_LIST(mb_stream_req),
2465 ATTR_LIST(mb_group_prealloc),
55138e0b 2466 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2467 NULL,
2468};
2469
857ac889
LC		 2470/* Features this copy of ext4 supports */
2471EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2472EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2473
2474static struct attribute *ext4_feat_attrs[] = {
2475 ATTR_LIST(lazy_itable_init),
27ee40df 2476 ATTR_LIST(batched_discard),
857ac889
LC		 2477 NULL,
2478};
2479
3197ebdb
TT		 2480static ssize_t ext4_attr_show(struct kobject *kobj,
2481 struct attribute *attr, char *buf)
2482{
2483 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2484 s_kobj);
2485 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2486
2487 return a->show ? a->show(a, sbi, buf) : 0;
2488}
2489
2490static ssize_t ext4_attr_store(struct kobject *kobj,
2491 struct attribute *attr,
2492 const char *buf, size_t len)
2493{
2494 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2495 s_kobj);
2496 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2497
2498 return a->store ? a->store(a, sbi, buf, len) : 0;
2499}
2500
2501static void ext4_sb_release(struct kobject *kobj)
2502{
2503 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2504 s_kobj);
2505 complete(&sbi->s_kobj_unregister);
2506}
2507
52cf25d0 2508static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2509 .show = ext4_attr_show,
2510 .store = ext4_attr_store,
2511};
2512
2513static struct kobj_type ext4_ktype = {
2514 .default_attrs = ext4_attrs,
2515 .sysfs_ops = &ext4_attr_ops,
2516 .release = ext4_sb_release,
2517};
2518
857ac889
LC		 2519static void ext4_feat_release(struct kobject *kobj)
2520{
2521 complete(&ext4_feat->f_kobj_unregister);
2522}
2523
2524static struct kobj_type ext4_feat_ktype = {
2525 .default_attrs = ext4_feat_attrs,
2526 .sysfs_ops = &ext4_attr_ops,
2527 .release = ext4_feat_release,
2528};
2529
a13fb1a4
ES		 2530/*
2531 * Check whether this filesystem can be mounted based on
2532 * the features present and the RDONLY/RDWR mount requested.
2533 * Returns 1 if this filesystem can be mounted as requested,
2534 * 0 if it cannot be.
2535 */
2536static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2537{
2538 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2539 ext4_msg(sb, KERN_ERR,
2540 "Couldn't mount because of "
2541 "unsupported optional features (%x)",
2542 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2543 ~EXT4_FEATURE_INCOMPAT_SUPP));
2544 return 0;
2545 }
2546
2547 if (readonly)
2548 return 1;
2549
2550 /* Check that feature set is OK for a read-write mount */
2551 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2552 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2553 "unsupported optional features (%x)",
2554 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2555 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2556 return 0;
2557 }
2558 /*
2559 * Large file size enabled file system can only be mounted
2560 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2561 */
2562 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2563 if (sizeof(blkcnt_t) < sizeof(u64)) {
2564 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2565 "cannot be mounted RDWR without "
2566 "CONFIG_LBDAF");
2567 return 0;
2568 }
2569 }
bab08ab9
TT		 2570 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2571 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2572 ext4_msg(sb, KERN_ERR,
2573 "Can't support bigalloc feature without "
2574 "extents feature\n");
2575 return 0;
2576 }
a13fb1a4
ES		 2577 return 1;
2578}
2579
66e61a9e
TT		 2580/*
2581 * This function is called once a day if we have errors logged
2582 * on the file system
2583 */
2584static void print_daily_error_info(unsigned long arg)
2585{
2586 struct super_block *sb = (struct super_block *) arg;
2587 struct ext4_sb_info *sbi;
2588 struct ext4_super_block *es;
2589
2590 sbi = EXT4_SB(sb);
2591 es = sbi->s_es;
2592
2593 if (es->s_error_count)
2594 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2595 le32_to_cpu(es->s_error_count));
2596 if (es->s_first_error_time) {
2597 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2598 sb->s_id, le32_to_cpu(es->s_first_error_time),
2599 (int) sizeof(es->s_first_error_func),
2600 es->s_first_error_func,
2601 le32_to_cpu(es->s_first_error_line));
2602 if (es->s_first_error_ino)
2603 printk(": inode %u",
2604 le32_to_cpu(es->s_first_error_ino));
2605 if (es->s_first_error_block)
2606 printk(": block %llu", (unsigned long long)
2607 le64_to_cpu(es->s_first_error_block));
2608 printk("\n");
2609 }
2610 if (es->s_last_error_time) {
2611 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2612 sb->s_id, le32_to_cpu(es->s_last_error_time),
2613 (int) sizeof(es->s_last_error_func),
2614 es->s_last_error_func,
2615 le32_to_cpu(es->s_last_error_line));
2616 if (es->s_last_error_ino)
2617 printk(": inode %u",
2618 le32_to_cpu(es->s_last_error_ino));
2619 if (es->s_last_error_block)
2620 printk(": block %llu", (unsigned long long)
2621 le64_to_cpu(es->s_last_error_block));
2622 printk("\n");
2623 }
2624 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2625}
2626
bfff6873
LC		 2627/* Find next suitable group and run ext4_init_inode_table */
2628static int ext4_run_li_request(struct ext4_li_request *elr)
2629{
2630 struct ext4_group_desc *gdp = NULL;
2631 ext4_group_t group, ngroups;
2632 struct super_block *sb;
2633 unsigned long timeout = 0;
2634 int ret = 0;
2635
2636 sb = elr->lr_super;
2637 ngroups = EXT4_SB(sb)->s_groups_count;
2638
2639 for (group = elr->lr_next_group; group < ngroups; group++) {
2640 gdp = ext4_get_group_desc(sb, group, NULL);
2641 if (!gdp) {
2642 ret = 1;
2643 break;
2644 }
2645
2646 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2647 break;
2648 }
2649
2650 if (group == ngroups)
2651 ret = 1;
2652
2653 if (!ret) {
2654 timeout = jiffies;
2655 ret = ext4_init_inode_table(sb, group,
2656 elr->lr_timeout ? 0 : 1);
2657 if (elr->lr_timeout == 0) {
51ce6511
LC		 2658 timeout = (jiffies - timeout) *
2659 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2660 elr->lr_timeout = timeout;
2661 }
2662 elr->lr_next_sched = jiffies + elr->lr_timeout;
2663 elr->lr_next_group = group + 1;
2664 }
2665
2666 return ret;
2667}
2668
2669/*
2670 * Remove lr_request from the list_request and free the
4ed5c033 2671 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2672 */
2673static void ext4_remove_li_request(struct ext4_li_request *elr)
2674{
2675 struct ext4_sb_info *sbi;
2676
2677 if (!elr)
2678 return;
2679
2680 sbi = elr->lr_sbi;
2681
2682 list_del(&elr->lr_request);
2683 sbi->s_li_request = NULL;
2684 kfree(elr);
2685}
2686
2687static void ext4_unregister_li_request(struct super_block *sb)
2688{
1bb933fb
LC		 2689 mutex_lock(&ext4_li_mtx);
2690 if (!ext4_li_info) {
2691 mutex_unlock(&ext4_li_mtx);
bfff6873 2692 return;
1bb933fb 2693 }
bfff6873
LC		 2694
2695 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2696 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2697 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2698 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2699}
2700
8f1f7453
ES		 2701static struct task_struct *ext4_lazyinit_task;
2702
bfff6873
LC		 2703/*
2704 * This is the function where ext4lazyinit thread lives. It walks
2705 * through the request list searching for next scheduled filesystem.
2706 * When such a fs is found, run the lazy initialization request
2707 * (ext4_rn_li_request) and keep track of the time spend in this
2708 * function. Based on that time we compute next schedule time of
2709 * the request. When walking through the list is complete, compute
2710 * next waking time and put itself into sleep.
2711 */
2712static int ext4_lazyinit_thread(void *arg)
2713{
2714 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2715 struct list_head *pos, *n;
2716 struct ext4_li_request *elr;
4ed5c033 2717 unsigned long next_wakeup, cur;
bfff6873
LC		 2718
2719 BUG_ON(NULL == eli);
2720
bfff6873
LC		 2721cont_thread:
2722 while (true) {
2723 next_wakeup = MAX_JIFFY_OFFSET;
2724
2725 mutex_lock(&eli->li_list_mtx);
2726 if (list_empty(&eli->li_request_list)) {
2727 mutex_unlock(&eli->li_list_mtx);
2728 goto exit_thread;
2729 }
2730
2731 list_for_each_safe(pos, n, &eli->li_request_list) {
2732 elr = list_entry(pos, struct ext4_li_request,
2733 lr_request);
2734
b2c78cd0
TT		 2735 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2736 if (ext4_run_li_request(elr) != 0) {
2737 /* error, remove the lazy_init job */
2738 ext4_remove_li_request(elr);
2739 continue;
2740 }
bfff6873
LC		 2741 }
2742
2743 if (time_before(elr->lr_next_sched, next_wakeup))
2744 next_wakeup = elr->lr_next_sched;
2745 }
2746 mutex_unlock(&eli->li_list_mtx);
2747
a0acae0e 2748 try_to_freeze();
bfff6873 2749
4ed5c033
LC		 2750 cur = jiffies;
2751 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2752 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2753 cond_resched();
2754 continue;
2755 }
2756
4ed5c033
LC		 2757 schedule_timeout_interruptible(next_wakeup - cur);
2758
8f1f7453
ES		 2759 if (kthread_should_stop()) {
2760 ext4_clear_request_list();
2761 goto exit_thread;
2762 }
bfff6873
LC		 2763 }
2764
2765exit_thread:
2766 /*
2767 * It looks like the request list is empty, but we need
2768 * to check it under the li_list_mtx lock, to prevent any
2769 * additions into it, and of course we should lock ext4_li_mtx
2770 * to atomically free the list and ext4_li_info, because at
2771 * this point another ext4 filesystem could be registering
2772 * new one.
2773 */
2774 mutex_lock(&ext4_li_mtx);
2775 mutex_lock(&eli->li_list_mtx);
2776 if (!list_empty(&eli->li_request_list)) {
2777 mutex_unlock(&eli->li_list_mtx);
2778 mutex_unlock(&ext4_li_mtx);
2779 goto cont_thread;
2780 }
2781 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2782 kfree(ext4_li_info);
2783 ext4_li_info = NULL;
2784 mutex_unlock(&ext4_li_mtx);
2785
2786 return 0;
2787}
2788
2789static void ext4_clear_request_list(void)
2790{
2791 struct list_head *pos, *n;
2792 struct ext4_li_request *elr;
2793
2794 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2795 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2796 elr = list_entry(pos, struct ext4_li_request,
2797 lr_request);
2798 ext4_remove_li_request(elr);
2799 }
2800 mutex_unlock(&ext4_li_info->li_list_mtx);
2801}
2802
2803static int ext4_run_lazyinit_thread(void)
2804{
8f1f7453
ES		 2805 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2806 ext4_li_info, "ext4lazyinit");
2807 if (IS_ERR(ext4_lazyinit_task)) {
2808 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2809 ext4_clear_request_list();
bfff6873
LC		 2810 kfree(ext4_li_info);
2811 ext4_li_info = NULL;
92b97816 2812 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2813 "initialization thread\n",
2814 err);
2815 return err;
2816 }
2817 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2818 return 0;
2819}
2820
2821/*
2822 * Check whether it make sense to run itable init. thread or not.
2823 * If there is at least one uninitialized inode table, return
2824 * corresponding group number, else the loop goes through all
2825 * groups and return total number of groups.
2826 */
2827static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2828{
2829 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2830 struct ext4_group_desc *gdp = NULL;
2831
2832 for (group = 0; group < ngroups; group++) {
2833 gdp = ext4_get_group_desc(sb, group, NULL);
2834 if (!gdp)
2835 continue;
2836
2837 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2838 break;
2839 }
2840
2841 return group;
2842}
2843
2844static int ext4_li_info_new(void)
2845{
2846 struct ext4_lazy_init *eli = NULL;
2847
2848 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2849 if (!eli)
2850 return -ENOMEM;
2851
bfff6873
LC		 2852 INIT_LIST_HEAD(&eli->li_request_list);
2853 mutex_init(&eli->li_list_mtx);
2854
bfff6873
LC		 2855 eli->li_state |= EXT4_LAZYINIT_QUIT;
2856
2857 ext4_li_info = eli;
2858
2859 return 0;
2860}
2861
2862static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2863 ext4_group_t start)
2864{
2865 struct ext4_sb_info *sbi = EXT4_SB(sb);
2866 struct ext4_li_request *elr;
2867 unsigned long rnd;
2868
2869 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2870 if (!elr)
2871 return NULL;
2872
2873 elr->lr_super = sb;
2874 elr->lr_sbi = sbi;
2875 elr->lr_next_group = start;
2876
2877 /*
2878 * Randomize first schedule time of the request to
2879 * spread the inode table initialization requests
2880 * better.
2881 */
2882 get_random_bytes(&rnd, sizeof(rnd));
2883 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2884 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2885
2886 return elr;
2887}
2888
2889static int ext4_register_li_request(struct super_block *sb,
2890 ext4_group_t first_not_zeroed)
2891{
2892 struct ext4_sb_info *sbi = EXT4_SB(sb);
2893 struct ext4_li_request *elr;
2894 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2895 int ret = 0;
bfff6873 2896
51ce6511
LC		 2897 if (sbi->s_li_request != NULL) {
2898 /*
2899 * Reset timeout so it can be computed again, because
2900 * s_li_wait_mult might have changed.
2901 */
2902 sbi->s_li_request->lr_timeout = 0;
beed5ecb 2903 return 0;
51ce6511 2904 }
bfff6873
LC		 2905
2906 if (first_not_zeroed == ngroups ||
2907 (sb->s_flags & MS_RDONLY) ||
55ff3840 2908 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 2909 return 0;
bfff6873
LC		 2910
2911 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 2912 if (!elr)
2913 return -ENOMEM;
bfff6873
LC		 2914
2915 mutex_lock(&ext4_li_mtx);
2916
2917 if (NULL == ext4_li_info) {
2918 ret = ext4_li_info_new();
2919 if (ret)
2920 goto out;
2921 }
2922
2923 mutex_lock(&ext4_li_info->li_list_mtx);
2924 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2925 mutex_unlock(&ext4_li_info->li_list_mtx);
2926
2927 sbi->s_li_request = elr;
46e4690b
TM		 2928 /*
2929 * set elr to NULL here since it has been inserted to
2930 * the request_list and the removal and free of it is
2931 * handled by ext4_clear_request_list from now on.
2932 */
2933 elr = NULL;
bfff6873
LC		 2934
2935 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2936 ret = ext4_run_lazyinit_thread();
2937 if (ret)
2938 goto out;
2939 }
bfff6873 2940out:
beed5ecb
NK		 2941 mutex_unlock(&ext4_li_mtx);
2942 if (ret)
bfff6873 2943 kfree(elr);
bfff6873
LC		 2944 return ret;
2945}
2946
2947/*
2948 * We do not need to lock anything since this is called on
2949 * module unload.
2950 */
2951static void ext4_destroy_lazyinit_thread(void)
2952{
2953 /*
2954 * If thread exited earlier
2955 * there's nothing to be done.
2956 */
8f1f7453 2957 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 2958 return;
2959
8f1f7453 2960 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 2961}
2962
2b2d6d01 2963static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 2964{
d4c402d9 2965 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 2966 struct buffer_head *bh;
617ba13b
MC		 2967 struct ext4_super_block *es = NULL;
2968 struct ext4_sb_info *sbi;
2969 ext4_fsblk_t block;
2970 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2971 ext4_fsblk_t logical_sb_block;
ac27a0ec 2972 unsigned long offset = 0;
ac27a0ec
DK		 2973 unsigned long journal_devnum = 0;
2974 unsigned long def_mount_opts;
2975 struct inode *root;
9f6200bb 2976 char *cp;
0390131b 2977 const char *descr;
dcc7dae3 2978 int ret = -ENOMEM;
281b5995 2979 int blocksize, clustersize;
4ec11028
TT		 2980 unsigned int db_count;
2981 unsigned int i;
281b5995 2982 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 2983 __u64 blocks_count;
833f4077 2984 int err;
b3881f74 2985 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 2986 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 2987
2988 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2989 if (!sbi)
dcc7dae3 2990 goto out_free_orig;
705895b6
PE		 2991
2992 sbi->s_blockgroup_lock =
2993 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2994 if (!sbi->s_blockgroup_lock) {
2995 kfree(sbi);
dcc7dae3 2996 goto out_free_orig;
705895b6 2997 }
ac27a0ec
DK		 2998 sb->s_fs_info = sbi;
2999 sbi->s_mount_opt = 0;
617ba13b
MC		 3000 sbi->s_resuid = EXT4_DEF_RESUID;
3001 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 3002 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3003 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3004 if (sb->s_bdev->bd_part)
3005 sbi->s_sectors_written_start =
3006 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3007
9f6200bb
TT		 3008 /* Cleanup superblock name */
3009 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3010 *cp = '!';
3011
dcc7dae3 3012 ret = -EINVAL;
617ba13b 3013 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3014 if (!blocksize) {
b31e1552 3015 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3016 goto out_fail;
3017 }
3018
3019 /*
617ba13b 3020 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3021 * block sizes. We need to calculate the offset from buffer start.
3022 */
617ba13b 3023 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3024 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3025 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3026 } else {
70bbb3e0 3027 logical_sb_block = sb_block;
ac27a0ec
DK		 3028 }
3029
70bbb3e0 3030 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3031 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3032 goto out_fail;
3033 }
3034 /*
3035 * Note: s_es must be initialized as soon as possible because
617ba13b 3036 * some ext4 macro-instructions depend on its value
ac27a0ec 3037 */
617ba13b 3038 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 3039 sbi->s_es = es;
3040 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3041 if (sb->s_magic != EXT4_SUPER_MAGIC)
3042 goto cantfind_ext4;
afc32f7e 3043 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 3044
3045 /* Set defaults before we parse the mount options */
3046 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3047 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3048 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3049 set_opt(sb, DEBUG);
87f26807 3050 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3051 set_opt(sb, GRPID);
617ba13b 3052 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3053 set_opt(sb, NO_UID32);
ea663336 3054 /* xattr user namespace & acls are now defaulted on */
03010a33 3055#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3056 set_opt(sb, XATTR_USER);
2e7842b8 3057#endif
03010a33 3058#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3059 set_opt(sb, POSIX_ACL);
2e7842b8 3060#endif
6fd7a467 3061 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3062 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3063 set_opt(sb, JOURNAL_DATA);
617ba13b 3064 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3065 set_opt(sb, ORDERED_DATA);
617ba13b 3066 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3067 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3068
3069 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3070 set_opt(sb, ERRORS_PANIC);
bb4f397a 3071 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3072 set_opt(sb, ERRORS_CONT);
bb4f397a 3073 else
fd8c37ec 3074 set_opt(sb, ERRORS_RO);
8b67f04a 3075 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3076 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3077 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3078 set_opt(sb, DISCARD);
ac27a0ec
DK		 3079
3080 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3081 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 3082 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3083 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3084 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3085
8b67f04a 3086 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3087 set_opt(sb, BARRIER);
ac27a0ec 3088
dd919b98
AK		 3089 /*
3090 * enable delayed allocation by default
3091 * Use -o nodelalloc to turn it off
3092 */
8b67f04a
TT		 3093 if (!IS_EXT3_SB(sb) &&
3094 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3095 set_opt(sb, DELALLOC);
dd919b98 3096
51ce6511
LC		 3097 /*
3098 * set default s_li_wait_mult for lazyinit, for the case there is
3099 * no mount option specified.
3100 */
3101 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3102
8b67f04a 3103 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3104 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3105 ext4_msg(sb, KERN_WARNING,
3106 "failed to parse options in superblock: %s",
3107 sbi->s_es->s_mount_opts);
3108 }
5a916be1 3109 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3110 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3111 &journal_ioprio, 0))
ac27a0ec
DK		 3112 goto failed_mount;
3113
56889787
TT		 3114 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3115 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3116 "with data=journal disables delayed "
3117 "allocation and O_DIRECT support!\n");
3118 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3119 ext4_msg(sb, KERN_ERR, "can't mount with "
3120 "both data=journal and delalloc");
3121 goto failed_mount;
3122 }
3123 if (test_opt(sb, DIOREAD_NOLOCK)) {
3124 ext4_msg(sb, KERN_ERR, "can't mount with "
3125 "both data=journal and delalloc");
3126 goto failed_mount;
3127 }
3128 if (test_opt(sb, DELALLOC))
3129 clear_opt(sb, DELALLOC);
3130 }
3131
3132 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3133 if (test_opt(sb, DIOREAD_NOLOCK)) {
3134 if (blocksize < PAGE_SIZE) {
3135 ext4_msg(sb, KERN_ERR, "can't mount with "
3136 "dioread_nolock if block size != PAGE_SIZE");
3137 goto failed_mount;
3138 }
3139 }
3140
ac27a0ec 3141 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3142 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3143
617ba13b
MC		 3144 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3145 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3146 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3147 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3148 ext4_msg(sb, KERN_WARNING,
3149 "feature flags set on rev 0 fs, "
3150 "running e2fsck is recommended");
469108ff 3151
2035e776
TT		 3152 if (IS_EXT2_SB(sb)) {
3153 if (ext2_feature_set_ok(sb))
3154 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3155 "using the ext4 subsystem");
3156 else {
3157 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3158 "to feature incompatibilities");
3159 goto failed_mount;
3160 }
3161 }
3162
3163 if (IS_EXT3_SB(sb)) {
3164 if (ext3_feature_set_ok(sb))
3165 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3166 "using the ext4 subsystem");
3167 else {
3168 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3169 "to feature incompatibilities");
3170 goto failed_mount;
3171 }
3172 }
3173
ac27a0ec
DK		 3174 /*
3175 * Check feature flags regardless of the revision level, since we
3176 * previously didn't change the revision level when setting the flags,
3177 * so there is a chance incompat flags are set on a rev 0 filesystem.
3178 */
a13fb1a4 3179 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3180 goto failed_mount;
a13fb1a4 3181
617ba13b
MC		 3182 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3183 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3184 ext4_msg(sb, KERN_ERR,
3185 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3186 goto failed_mount;
3187 }
3188
ac27a0ec 3189 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3190 /* Validate the filesystem blocksize */
3191 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3192 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3193 blocksize);
ac27a0ec
DK		 3194 goto failed_mount;
3195 }
3196
2b2d6d01 3197 brelse(bh);
70bbb3e0
AM		 3198 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3199 offset = do_div(logical_sb_block, blocksize);
3200 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3201 if (!bh) {
b31e1552
ES		 3202 ext4_msg(sb, KERN_ERR,
3203 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3204 goto failed_mount;
3205 }
617ba13b 3206 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 3207 sbi->s_es = es;
617ba13b 3208 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3209 ext4_msg(sb, KERN_ERR,
3210 "Magic mismatch, very weird!");
ac27a0ec
DK		 3211 goto failed_mount;
3212 }
3213 }
3214
a13fb1a4
ES		 3215 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3216 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3217 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3218 has_huge_files);
3219 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3220
617ba13b
MC		 3221 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3222 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3223 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3224 } else {
3225 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3226 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3227 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3228 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3229 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3230 ext4_msg(sb, KERN_ERR,
3231 "unsupported inode size: %d",
2b2d6d01 3232 sbi->s_inode_size);
ac27a0ec
DK		 3233 goto failed_mount;
3234 }
ef7f3835
KS		 3235 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3236 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3237 }
0b8e58a1 3238
0d1ee42f
AR		 3239 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3240 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3241 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3242 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3243 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3244 ext4_msg(sb, KERN_ERR,
3245 "unsupported descriptor size %lu",
0d1ee42f
AR		 3246 sbi->s_desc_size);
3247 goto failed_mount;
3248 }
3249 } else
3250 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3251
ac27a0ec 3252 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3253 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3254 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3255 goto cantfind_ext4;
0b8e58a1 3256
617ba13b 3257 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3258 if (sbi->s_inodes_per_block == 0)
617ba13b 3259 goto cantfind_ext4;
ac27a0ec
DK		 3260 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3261 sbi->s_inodes_per_block;
0d1ee42f 3262 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3263 sbi->s_sbh = bh;
3264 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3265 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3266 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3267
2b2d6d01 3268 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3269 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3270 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3271 i = le32_to_cpu(es->s_flags);
3272 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3273 sbi->s_hash_unsigned = 3;
3274 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3275#ifdef __CHAR_UNSIGNED__
3276 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3277 sbi->s_hash_unsigned = 3;
3278#else
3279 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3280#endif
3281 sb->s_dirt = 1;
3282 }
ac27a0ec 3283
281b5995
TT		 3284 /* Handle clustersize */
3285 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3286 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3287 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3288 if (has_bigalloc) {
3289 if (clustersize < blocksize) {
3290 ext4_msg(sb, KERN_ERR,
3291 "cluster size (%d) smaller than "
3292 "block size (%d)", clustersize, blocksize);
3293 goto failed_mount;
3294 }
3295 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3296 le32_to_cpu(es->s_log_block_size);
3297 sbi->s_clusters_per_group =
3298 le32_to_cpu(es->s_clusters_per_group);
3299 if (sbi->s_clusters_per_group > blocksize * 8) {
3300 ext4_msg(sb, KERN_ERR,
3301 "#clusters per group too big: %lu",
3302 sbi->s_clusters_per_group);
3303 goto failed_mount;
3304 }
3305 if (sbi->s_blocks_per_group !=
3306 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3307 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3308 "clusters per group (%lu) inconsistent",
3309 sbi->s_blocks_per_group,
3310 sbi->s_clusters_per_group);
3311 goto failed_mount;
3312 }
3313 } else {
3314 if (clustersize != blocksize) {
3315 ext4_warning(sb, "fragment/cluster size (%d) != "
3316 "block size (%d)", clustersize,
3317 blocksize);
3318 clustersize = blocksize;
3319 }
3320 if (sbi->s_blocks_per_group > blocksize * 8) {
3321 ext4_msg(sb, KERN_ERR,
3322 "#blocks per group too big: %lu",
3323 sbi->s_blocks_per_group);
3324 goto failed_mount;
3325 }
3326 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3327 sbi->s_cluster_bits = 0;
ac27a0ec 3328 }
281b5995
TT		 3329 sbi->s_cluster_ratio = clustersize / blocksize;
3330
ac27a0ec 3331 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3332 ext4_msg(sb, KERN_ERR,
3333 "#inodes per group too big: %lu",
2b2d6d01 3334 sbi->s_inodes_per_group);
ac27a0ec
DK		 3335 goto failed_mount;
3336 }
3337
bf43d84b
ES		 3338 /*
3339 * Test whether we have more sectors than will fit in sector_t,
3340 * and whether the max offset is addressable by the page cache.
3341 */
5a9ae68a 3342 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3343 ext4_blocks_count(es));
5a9ae68a 3344 if (err) {
b31e1552 3345 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3346 " too large to mount safely on this system");
ac27a0ec 3347 if (sizeof(sector_t) < 8)
90c699a9 3348 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3349 ret = err;
ac27a0ec
DK		 3350 goto failed_mount;
3351 }
3352
617ba13b
MC		 3353 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3354 goto cantfind_ext4;
e7c95593 3355
0f2ddca6
FTN		 3356 /* check blocks count against device size */
3357 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3358 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3359 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3360 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3361 ext4_blocks_count(es), blocks_count);
3362 goto failed_mount;
3363 }
3364
0b8e58a1
AD		 3365 /*
3366 * It makes no sense for the first data block to be beyond the end
3367 * of the filesystem.
3368 */
3369 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3370 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3371 "block %u is beyond end of filesystem (%llu)",
3372 le32_to_cpu(es->s_first_data_block),
3373 ext4_blocks_count(es));
e7c95593
ES		 3374 goto failed_mount;
3375 }
bd81d8ee
LV		 3376 blocks_count = (ext4_blocks_count(es) -
3377 le32_to_cpu(es->s_first_data_block) +
3378 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3379 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3380 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3381 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3382 "(block count %llu, first data block %u, "
b31e1552 3383 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3384 ext4_blocks_count(es),
3385 le32_to_cpu(es->s_first_data_block),
3386 EXT4_BLOCKS_PER_GROUP(sb));
3387 goto failed_mount;
3388 }
bd81d8ee 3389 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3390 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3391 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3392 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3393 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3394 sbi->s_group_desc = ext4_kvmalloc(db_count *
3395 sizeof(struct buffer_head *),
3396 GFP_KERNEL);
ac27a0ec 3397 if (sbi->s_group_desc == NULL) {
b31e1552 3398 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 3399 goto failed_mount;
3400 }
3401
9f6200bb
TT		 3402 if (ext4_proc_root)
3403 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3404
66acdcf4
TT		 3405 if (sbi->s_proc)
3406 proc_create_data("options", S_IRUGO, sbi->s_proc,
3407 &ext4_seq_options_fops, sb);
3408
705895b6 3409 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3410
3411 for (i = 0; i < db_count; i++) {
70bbb3e0 3412 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3413 sbi->s_group_desc[i] = sb_bread(sb, block);
3414 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3415 ext4_msg(sb, KERN_ERR,
3416 "can't read group descriptor %d", i);
ac27a0ec
DK		 3417 db_count = i;
3418 goto failed_mount2;
3419 }
3420 }
bfff6873 3421 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3422 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3423 goto failed_mount2;
3424 }
772cb7c8
JS		 3425 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3426 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3427 ext4_msg(sb, KERN_ERR,
3428 "unable to initialize "
3429 "flex_bg meta info!");
772cb7c8
JS		 3430 goto failed_mount2;
3431 }
3432
ac27a0ec
DK		 3433 sbi->s_gdb_count = db_count;
3434 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3435 spin_lock_init(&sbi->s_next_gen_lock);
3436
04496411
TM		 3437 init_timer(&sbi->s_err_report);
3438 sbi->s_err_report.function = print_daily_error_info;
3439 sbi->s_err_report.data = (unsigned long) sb;
3440
57042651 3441 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3442 ext4_count_free_clusters(sb));
ce7e010a
TT		 3443 if (!err) {
3444 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3445 ext4_count_free_inodes(sb));
3446 }
3447 if (!err) {
3448 err = percpu_counter_init(&sbi->s_dirs_counter,
3449 ext4_count_dirs(sb));
3450 }
3451 if (!err) {
57042651 3452 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3453 }
3454 if (err) {
3455 ext4_msg(sb, KERN_ERR, "insufficient memory");
3456 goto failed_mount3;
3457 }
3458
c9de560d 3459 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3460 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3461
ac27a0ec
DK		 3462 /*
3463 * set up enough so that it can read an inode
3464 */
9ca92389
TT		 3465 if (!test_opt(sb, NOLOAD) &&
3466 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3467 sb->s_op = &ext4_sops;
3468 else
3469 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3470 sb->s_export_op = &ext4_export_ops;
3471 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3472#ifdef CONFIG_QUOTA
617ba13b
MC		 3473 sb->s_qcop = &ext4_qctl_operations;
3474 sb->dq_op = &ext4_quota_operations;
ac27a0ec 3475#endif
f2fa2ffc
AK		 3476 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3477
ac27a0ec 3478 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3479 mutex_init(&sbi->s_orphan_lock);
8f82f840 3480 sbi->s_resize_flags = 0;
ac27a0ec
DK		 3481
3482 sb->s_root = NULL;
3483
3484 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3485 EXT4_HAS_INCOMPAT_FEATURE(sb,
3486 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3487
c5e06d10
JL		 3488 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3489 !(sb->s_flags & MS_RDONLY))
3490 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3491 goto failed_mount3;
3492
ac27a0ec
DK		 3493 /*
3494 * The first inode we look at is the journal inode. Don't try
3495 * root first: it may be modified in the journal!
3496 */
3497 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3498 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3499 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3500 goto failed_mount3;
0390131b
FM		 3501 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3502 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3503 ext4_msg(sb, KERN_ERR, "required journal recovery "
3504 "suppressed and not mounted read-only");
744692dc 3505 goto failed_mount_wq;
ac27a0ec 3506 } else {
fd8c37ec 3507 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3508 sbi->s_journal = NULL;
3509 needs_recovery = 0;
3510 goto no_journal;
ac27a0ec
DK		 3511 }
3512
eb40a09c
JS		 3513 if (ext4_blocks_count(es) > 0xffffffffULL &&
3514 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3515 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3516 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3517 goto failed_mount_wq;
eb40a09c
JS		 3518 }
3519
d4da6c9c
LT		 3520 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3521 jbd2_journal_set_features(sbi->s_journal,
3522 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 3523 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3524 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3525 jbd2_journal_set_features(sbi->s_journal,
3526 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 3527 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3528 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3529 } else {
3530 jbd2_journal_clear_features(sbi->s_journal,
3531 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3532 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3533 }
818d276c 3534
ac27a0ec
DK		 3535 /* We have now updated the journal if required, so we can
3536 * validate the data journaling mode. */
3537 switch (test_opt(sb, DATA_FLAGS)) {
3538 case 0:
3539 /* No mode set, assume a default based on the journal
63f57933
AM		 3540 * capabilities: ORDERED_DATA if the journal can
3541 * cope, else JOURNAL_DATA
3542 */
dab291af
MC		 3543 if (jbd2_journal_check_available_features
3544 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3545 set_opt(sb, ORDERED_DATA);
ac27a0ec 3546 else
fd8c37ec 3547 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3548 break;
3549
617ba13b
MC		 3550 case EXT4_MOUNT_ORDERED_DATA:
3551 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3552 if (!jbd2_journal_check_available_features
3553 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3554 ext4_msg(sb, KERN_ERR, "Journal does not support "
3555 "requested data journaling mode");
744692dc 3556 goto failed_mount_wq;
ac27a0ec
DK		 3557 }
3558 default:
3559 break;
3560 }
b3881f74 3561 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3562
18aadd47
BJ		 3563 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3564
ce7e010a
TT		 3565 /*
3566 * The journal may have updated the bg summary counts, so we
3567 * need to update the global counters.
3568 */
57042651 3569 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3570 ext4_count_free_clusters(sb));
ce7e010a
TT		 3571 percpu_counter_set(&sbi->s_freeinodes_counter,
3572 ext4_count_free_inodes(sb));
3573 percpu_counter_set(&sbi->s_dirs_counter,
3574 ext4_count_dirs(sb));
57042651 3575 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3576
ce7e010a 3577no_journal:
fd89d5f2
TH		 3578 /*
3579 * The maximum number of concurrent works can be high and
3580 * concurrency isn't really necessary. Limit it to 1.
3581 */
3582 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3583 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3584 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3585 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3586 goto failed_mount_wq;
3587 }
3588
ac27a0ec 3589 /*
dab291af 3590 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3591 * so we can safely mount the rest of the filesystem now.
3592 */
3593
1d1fe1ee
DH		 3594 root = ext4_iget(sb, EXT4_ROOT_INO);
3595 if (IS_ERR(root)) {
b31e1552 3596 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3597 ret = PTR_ERR(root);
32a9bb57 3598 root = NULL;
ac27a0ec
DK		 3599 goto failed_mount4;
3600 }
3601 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3602 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3603 iput(root);
ac27a0ec
DK		 3604 goto failed_mount4;
3605 }
1d1fe1ee
DH		 3606 sb->s_root = d_alloc_root(root);
3607 if (!sb->s_root) {
94bf608a 3608 iput(root);
b31e1552 3609 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3610 ret = -ENOMEM;
3611 goto failed_mount4;
3612 }
ac27a0ec 3613
2b2d6d01 3614 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 3615
3616 /* determine the minimum size of new large inodes, if present */
3617 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3618 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3619 EXT4_GOOD_OLD_INODE_SIZE;
3620 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3621 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3622 if (sbi->s_want_extra_isize <
3623 le16_to_cpu(es->s_want_extra_isize))
3624 sbi->s_want_extra_isize =
3625 le16_to_cpu(es->s_want_extra_isize);
3626 if (sbi->s_want_extra_isize <
3627 le16_to_cpu(es->s_min_extra_isize))
3628 sbi->s_want_extra_isize =
3629 le16_to_cpu(es->s_min_extra_isize);
3630 }
3631 }
3632 /* Check if enough inode space is available */
3633 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3634 sbi->s_inode_size) {
3635 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3636 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3637 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3638 "available");
ef7f3835
KS		 3639 }
3640
6fd058f7
TT		 3641 err = ext4_setup_system_zone(sb);
3642 if (err) {
b31e1552 3643 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3644 "zone (%d)", err);
94bf608a 3645 goto failed_mount4a;
6fd058f7
TT		 3646 }
3647
c2774d84
AK		 3648 ext4_ext_init(sb);
3649 err = ext4_mb_init(sb, needs_recovery);
3650 if (err) {
421f91d2 3651 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3652 err);
dcf2d804 3653 goto failed_mount5;
c2774d84
AK		 3654 }
3655
bfff6873
LC		 3656 err = ext4_register_li_request(sb, first_not_zeroed);
3657 if (err)
dcf2d804 3658 goto failed_mount6;
bfff6873 3659
3197ebdb
TT		 3660 sbi->s_kobj.kset = ext4_kset;
3661 init_completion(&sbi->s_kobj_unregister);
3662 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3663 "%s", sb->s_id);
dcf2d804
TM		 3664 if (err)
3665 goto failed_mount7;
3197ebdb 3666
617ba13b
MC		 3667 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3668 ext4_orphan_cleanup(sb, es);
3669 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3670 if (needs_recovery) {
b31e1552 3671 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3672 ext4_mark_recovery_complete(sb, es);
3673 }
3674 if (EXT4_SB(sb)->s_journal) {
3675 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3676 descr = " journalled data mode";
3677 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3678 descr = " ordered data mode";
3679 else
3680 descr = " writeback data mode";
3681 } else
3682 descr = "out journal";
3683
d4c402d9 3684 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3685 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3686 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3687
66e61a9e
TT		 3688 if (es->s_error_count)
3689 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3690
d4c402d9 3691 kfree(orig_data);
ac27a0ec
DK		 3692 return 0;
3693
617ba13b 3694cantfind_ext4:
ac27a0ec 3695 if (!silent)
b31e1552 3696 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3697 goto failed_mount;
3698
dcf2d804
TM		 3699failed_mount7:
3700 ext4_unregister_li_request(sb);
3701failed_mount6:
dcf2d804 3702 ext4_mb_release(sb);
94bf608a
AV		 3703failed_mount5:
3704 ext4_ext_release(sb);
dcf2d804 3705 ext4_release_system_zone(sb);
94bf608a
AV		 3706failed_mount4a:
3707 dput(sb->s_root);
32a9bb57 3708 sb->s_root = NULL;
94bf608a 3709failed_mount4:
b31e1552 3710 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3711 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3712failed_mount_wq:
0390131b
FM		 3713 if (sbi->s_journal) {
3714 jbd2_journal_destroy(sbi->s_journal);
3715 sbi->s_journal = NULL;
3716 }
ac27a0ec 3717failed_mount3:
04496411 3718 del_timer(&sbi->s_err_report);
9933fc0a
TT		 3719 if (sbi->s_flex_groups)
3720 ext4_kvfree(sbi->s_flex_groups);
57042651 3721 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 3722 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3723 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 3724 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 3725 if (sbi->s_mmp_tsk)
3726 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 3727failed_mount2:
3728 for (i = 0; i < db_count; i++)
3729 brelse(sbi->s_group_desc[i]);
f18a5f21 3730 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 3731failed_mount:
240799cd 3732 if (sbi->s_proc) {
66acdcf4 3733 remove_proc_entry("options", sbi->s_proc);
9f6200bb 3734 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3735 }
ac27a0ec
DK		 3736#ifdef CONFIG_QUOTA
3737 for (i = 0; i < MAXQUOTAS; i++)
3738 kfree(sbi->s_qf_names[i]);
3739#endif
617ba13b 3740 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3741 brelse(bh);
3742out_fail:
3743 sb->s_fs_info = NULL;
f6830165 3744 kfree(sbi->s_blockgroup_lock);
ac27a0ec 3745 kfree(sbi);
dcc7dae3 3746out_free_orig:
d4c402d9 3747 kfree(orig_data);
1d1fe1ee 3748 return ret;
ac27a0ec
DK		 3749}
3750
3751/*
3752 * Setup any per-fs journal parameters now. We'll do this both on
3753 * initial mount, once the journal has been initialised but before we've
3754 * done any recovery; and again on any subsequent remount.
3755 */
617ba13b 3756static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3757{
617ba13b 3758 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3759
30773840
TT		 3760 journal->j_commit_interval = sbi->s_commit_interval;
3761 journal->j_min_batch_time = sbi->s_min_batch_time;
3762 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 3763
a931da6a 3764 write_lock(&journal->j_state_lock);
ac27a0ec 3765 if (test_opt(sb, BARRIER))
dab291af 3766 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3767 else
dab291af 3768 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3769 if (test_opt(sb, DATA_ERR_ABORT))
3770 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3771 else
3772 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 3773 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 3774}
3775
617ba13b 3776static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3777 unsigned int journal_inum)
3778{
3779 struct inode *journal_inode;
3780 journal_t *journal;
3781
0390131b
FM		 3782 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3783
ac27a0ec
DK		 3784 /* First, test for the existence of a valid inode on disk. Bad
3785 * things happen if we iget() an unused inode, as the subsequent
3786 * iput() will try to delete it. */
3787
1d1fe1ee
DH		 3788 journal_inode = ext4_iget(sb, journal_inum);
3789 if (IS_ERR(journal_inode)) {
b31e1552 3790 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3791 return NULL;
3792 }
3793 if (!journal_inode->i_nlink) {
3794 make_bad_inode(journal_inode);
3795 iput(journal_inode);
b31e1552 3796 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3797 return NULL;
3798 }
3799
e5f8eab8 3800 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3801 journal_inode, journal_inode->i_size);
1d1fe1ee 3802 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3803 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3804 iput(journal_inode);
3805 return NULL;
3806 }
3807
dab291af 3808 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3809 if (!journal) {
b31e1552 3810 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3811 iput(journal_inode);
3812 return NULL;
3813 }
3814 journal->j_private = sb;
617ba13b 3815 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3816 return journal;
3817}
3818
617ba13b 3819static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3820 dev_t j_dev)
3821{
2b2d6d01 3822 struct buffer_head *bh;
ac27a0ec 3823 journal_t *journal;
617ba13b
MC		 3824 ext4_fsblk_t start;
3825 ext4_fsblk_t len;
ac27a0ec 3826 int hblock, blocksize;
617ba13b 3827 ext4_fsblk_t sb_block;
ac27a0ec 3828 unsigned long offset;
2b2d6d01 3829 struct ext4_super_block *es;
ac27a0ec
DK		 3830 struct block_device *bdev;
3831
0390131b
FM		 3832 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3833
b31e1552 3834 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3835 if (bdev == NULL)
3836 return NULL;
3837
ac27a0ec 3838 blocksize = sb->s_blocksize;
e1defc4f 3839 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3840 if (blocksize < hblock) {
b31e1552
ES		 3841 ext4_msg(sb, KERN_ERR,
3842 "blocksize too small for journal device");
ac27a0ec
DK		 3843 goto out_bdev;
3844 }
3845
617ba13b
MC		 3846 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3847 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3848 set_blocksize(bdev, blocksize);
3849 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3850 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3851 "external journal");
ac27a0ec
DK		 3852 goto out_bdev;
3853 }
3854
617ba13b
MC		 3855 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3856 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3857 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3858 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3859 ext4_msg(sb, KERN_ERR, "external journal has "
3860 "bad superblock");
ac27a0ec
DK		 3861 brelse(bh);
3862 goto out_bdev;
3863 }
3864
617ba13b 3865 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3866 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3867 brelse(bh);
3868 goto out_bdev;
3869 }
3870
bd81d8ee 3871 len = ext4_blocks_count(es);
ac27a0ec
DK		 3872 start = sb_block + 1;
3873 brelse(bh); /* we're done with the superblock */
3874
dab291af 3875 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3876 start, len, blocksize);
3877 if (!journal) {
b31e1552 3878 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3879 goto out_bdev;
3880 }
3881 journal->j_private = sb;
3882 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3883 wait_on_buffer(journal->j_sb_buffer);
3884 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3885 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3886 goto out_journal;
3887 }
3888 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3889 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3890 "user (unsupported) - %d",
ac27a0ec
DK		 3891 be32_to_cpu(journal->j_superblock->s_nr_users));
3892 goto out_journal;
3893 }
617ba13b
MC		 3894 EXT4_SB(sb)->journal_bdev = bdev;
3895 ext4_init_journal_params(sb, journal);
ac27a0ec 3896 return journal;
0b8e58a1 3897
ac27a0ec 3898out_journal:
dab291af 3899 jbd2_journal_destroy(journal);
ac27a0ec 3900out_bdev:
617ba13b 3901 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3902 return NULL;
3903}
3904
617ba13b
MC		 3905static int ext4_load_journal(struct super_block *sb,
3906 struct ext4_super_block *es,
ac27a0ec
DK		 3907 unsigned long journal_devnum)
3908{
3909 journal_t *journal;
3910 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3911 dev_t journal_dev;
3912 int err = 0;
3913 int really_read_only;
3914
0390131b
FM		 3915 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3916
ac27a0ec
DK		 3917 if (journal_devnum &&
3918 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3919 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3920 "numbers have changed");
ac27a0ec
DK		 3921 journal_dev = new_decode_dev(journal_devnum);
3922 } else
3923 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3924
3925 really_read_only = bdev_read_only(sb->s_bdev);
3926
3927 /*
3928 * Are we loading a blank journal or performing recovery after a
3929 * crash? For recovery, we need to check in advance whether we
3930 * can get read-write access to the device.
3931 */
617ba13b 3932 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3933 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3934 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3935 "required on readonly filesystem");
ac27a0ec 3936 if (really_read_only) {
b31e1552
ES		 3937 ext4_msg(sb, KERN_ERR, "write access "
3938 "unavailable, cannot proceed");
ac27a0ec
DK		 3939 return -EROFS;
3940 }
b31e1552
ES		 3941 ext4_msg(sb, KERN_INFO, "write access will "
3942 "be enabled during recovery");
ac27a0ec
DK		 3943 }
3944 }
3945
3946 if (journal_inum && journal_dev) {
b31e1552
ES		 3947 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3948 "and inode journals!");
ac27a0ec
DK		 3949 return -EINVAL;
3950 }
3951
3952 if (journal_inum) {
617ba13b 3953 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3954 return -EINVAL;
3955 } else {
617ba13b 3956 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3957 return -EINVAL;
3958 }
3959
90576c0b 3960 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 3961 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3962
617ba13b 3963 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3964 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 3965 if (!err) {
3966 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3967 if (save)
3968 memcpy(save, ((char *) es) +
3969 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 3970 err = jbd2_journal_load(journal);
1c13d5c0
TT		 3971 if (save)
3972 memcpy(((char *) es) + EXT4_S_ERR_START,
3973 save, EXT4_S_ERR_LEN);
3974 kfree(save);
3975 }
ac27a0ec
DK		 3976
3977 if (err) {
b31e1552 3978 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3979 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3980 return err;
3981 }
3982
617ba13b
MC		 3983 EXT4_SB(sb)->s_journal = journal;
3984 ext4_clear_journal_err(sb, es);
ac27a0ec 3985
c41303ce 3986 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 3987 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3988 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3989
3990 /* Make sure we flush the recovery flag to disk. */
e2d67052 3991 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3992 }
3993
3994 return 0;
3995}
3996
e2d67052 3997static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3998{
e2d67052 3999 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4000 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4001 int error = 0;
ac27a0ec 4002
7c2e7087 4003 if (!sbh || block_device_ejected(sb))
c4be0c1d 4004 return error;
914258bf
TT		 4005 if (buffer_write_io_error(sbh)) {
4006 /*
4007 * Oh, dear. A previous attempt to write the
4008 * superblock failed. This could happen because the
4009 * USB device was yanked out. Or it could happen to
4010 * be a transient write error and maybe the block will
4011 * be remapped. Nothing we can do but to retry the
4012 * write and hope for the best.
4013 */
b31e1552
ES		 4014 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4015 "superblock detected");
914258bf
TT		 4016 clear_buffer_write_io_error(sbh);
4017 set_buffer_uptodate(sbh);
4018 }
71290b36
TT		 4019 /*
4020 * If the file system is mounted read-only, don't update the
4021 * superblock write time. This avoids updating the superblock
4022 * write time when we are mounting the root file system
4023 * read/only but we need to replay the journal; at that point,
4024 * for people who are east of GMT and who make their clock
4025 * tick in localtime for Windows bug-for-bug compatibility,
4026 * the clock is set in the future, and this will cause e2fsck
4027 * to complain and force a full file system check.
4028 */
4029 if (!(sb->s_flags & MS_RDONLY))
4030 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4031 if (sb->s_bdev->bd_part)
4032 es->s_kbytes_written =
4033 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4034 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4035 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4036 else
4037 es->s_kbytes_written =
4038 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4039 ext4_free_blocks_count_set(es,
4040 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4041 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4042 es->s_free_inodes_count =
4043 cpu_to_le32(percpu_counter_sum_positive(
4044 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 4045 sb->s_dirt = 0;
ac27a0ec
DK		 4046 BUFFER_TRACE(sbh, "marking dirty");
4047 mark_buffer_dirty(sbh);
914258bf 4048 if (sync) {
c4be0c1d
TS		 4049 error = sync_dirty_buffer(sbh);
4050 if (error)
4051 return error;
4052
4053 error = buffer_write_io_error(sbh);
4054 if (error) {
b31e1552
ES		 4055 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4056 "superblock");
914258bf
TT		 4057 clear_buffer_write_io_error(sbh);
4058 set_buffer_uptodate(sbh);
4059 }
4060 }
c4be0c1d 4061 return error;
ac27a0ec
DK		 4062}
4063
ac27a0ec
DK		 4064/*
4065 * Have we just finished recovery? If so, and if we are mounting (or
4066 * remounting) the filesystem readonly, then we will end up with a
4067 * consistent fs on disk. Record that fact.
4068 */
2b2d6d01
TT		 4069static void ext4_mark_recovery_complete(struct super_block *sb,
4070 struct ext4_super_block *es)
ac27a0ec 4071{
617ba13b 4072 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4073
0390131b
FM		 4074 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4075 BUG_ON(journal != NULL);
4076 return;
4077 }
dab291af 4078 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4079 if (jbd2_journal_flush(journal) < 0)
4080 goto out;
4081
617ba13b 4082 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4083 sb->s_flags & MS_RDONLY) {
617ba13b 4084 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4085 ext4_commit_super(sb, 1);
ac27a0ec 4086 }
7ffe1ea8
HK		 4087
4088out:
dab291af 4089 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4090}
4091
4092/*
4093 * If we are mounting (or read-write remounting) a filesystem whose journal
4094 * has recorded an error from a previous lifetime, move that error to the
4095 * main filesystem now.
4096 */
2b2d6d01
TT		 4097static void ext4_clear_journal_err(struct super_block *sb,
4098 struct ext4_super_block *es)
ac27a0ec
DK		 4099{
4100 journal_t *journal;
4101 int j_errno;
4102 const char *errstr;
4103
0390131b
FM		 4104 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4105
617ba13b 4106 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4107
4108 /*
4109 * Now check for any error status which may have been recorded in the
617ba13b 4110 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4111 */
4112
dab291af 4113 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4114 if (j_errno) {
4115 char nbuf[16];
4116
617ba13b 4117 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4118 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4119 "from previous mount: %s", errstr);
12062ddd 4120 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4121
617ba13b
MC		 4122 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4123 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4124 ext4_commit_super(sb, 1);
ac27a0ec 4125
dab291af 4126 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 4127 }
4128}
4129
4130/*
4131 * Force the running and committing transactions to commit,
4132 * and wait on the commit.
4133 */
617ba13b 4134int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4135{
4136 journal_t *journal;
0390131b 4137 int ret = 0;
ac27a0ec
DK		 4138
4139 if (sb->s_flags & MS_RDONLY)
4140 return 0;
4141
617ba13b 4142 journal = EXT4_SB(sb)->s_journal;
6b0310fb 4143 if (journal) {
437f88cc 4144 vfs_check_frozen(sb, SB_FREEZE_TRANS);
0390131b 4145 ret = ext4_journal_force_commit(journal);
6b0310fb 4146 }
0390131b 4147
ac27a0ec
DK		 4148 return ret;
4149}
4150
2b2d6d01 4151static void ext4_write_super(struct super_block *sb)
ac27a0ec 4152{
ebc1ac16 4153 lock_super(sb);
9ca92389 4154 ext4_commit_super(sb, 1);
ebc1ac16 4155 unlock_super(sb);
ac27a0ec
DK		 4156}
4157
617ba13b 4158static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4159{
14ce0cb4 4160 int ret = 0;
9eddacf9 4161 tid_t target;
8d5d02e6 4162 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4163
9bffad1e 4164 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 4165 flush_workqueue(sbi->dio_unwritten_wq);
4166 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4167 if (wait)
8d5d02e6 4168 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4169 }
14ce0cb4 4170 return ret;
ac27a0ec
DK		 4171}
4172
4173/*
4174 * LVM calls this function before a (read-only) snapshot is created. This
4175 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4176 *
4177 * Note that only this function cannot bring a filesystem to be in a clean
4178 * state independently, because ext4 prevents a new handle from being started
4179 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4180 * the upper layer.
ac27a0ec 4181 */
c4be0c1d 4182static int ext4_freeze(struct super_block *sb)
ac27a0ec 4183{
c4be0c1d
TS		 4184 int error = 0;
4185 journal_t *journal;
ac27a0ec 4186
9ca92389
TT		 4187 if (sb->s_flags & MS_RDONLY)
4188 return 0;
ac27a0ec 4189
9ca92389 4190 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4191
9ca92389
TT		 4192 /* Now we set up the journal barrier. */
4193 jbd2_journal_lock_updates(journal);
ac27a0ec 4194
9ca92389
TT		 4195 /*
4196 * Don't clear the needs_recovery flag if we failed to flush
4197 * the journal.
4198 */
4199 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4200 if (error < 0)
4201 goto out;
9ca92389
TT		 4202
4203 /* Journal blocked and flushed, clear needs_recovery flag. */
4204 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4205 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 4206out:
4207 /* we rely on s_frozen to stop further updates */
4208 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4209 return error;
ac27a0ec
DK		 4210}
4211
4212/*
4213 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4214 * flag here, even though the filesystem is not technically dirty yet.
4215 */
c4be0c1d 4216static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4217{
9ca92389
TT		 4218 if (sb->s_flags & MS_RDONLY)
4219 return 0;
4220
4221 lock_super(sb);
4222 /* Reset the needs_recovery flag before the fs is unlocked. */
4223 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4224 ext4_commit_super(sb, 1);
4225 unlock_super(sb);
c4be0c1d 4226 return 0;
ac27a0ec
DK		 4227}
4228
673c6100
TT		 4229/*
4230 * Structure to save mount options for ext4_remount's benefit
4231 */
4232struct ext4_mount_options {
4233 unsigned long s_mount_opt;
a2595b8a 4234 unsigned long s_mount_opt2;
673c6100
TT		 4235 uid_t s_resuid;
4236 gid_t s_resgid;
4237 unsigned long s_commit_interval;
4238 u32 s_min_batch_time, s_max_batch_time;
4239#ifdef CONFIG_QUOTA
4240 int s_jquota_fmt;
4241 char *s_qf_names[MAXQUOTAS];
4242#endif
4243};
4244
2b2d6d01 4245static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4246{
2b2d6d01 4247 struct ext4_super_block *es;
617ba13b 4248 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4249 unsigned long old_sb_flags;
617ba13b 4250 struct ext4_mount_options old_opts;
c79d967d 4251 int enable_quota = 0;
8a266467 4252 ext4_group_t g;
b3881f74 4253 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4254 int err = 0;
ac27a0ec
DK		 4255#ifdef CONFIG_QUOTA
4256 int i;
4257#endif
d4c402d9 4258 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4259
4260 /* Store the original options */
bbd6851a 4261 lock_super(sb);
ac27a0ec
DK		 4262 old_sb_flags = sb->s_flags;
4263 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4264 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4265 old_opts.s_resuid = sbi->s_resuid;
4266 old_opts.s_resgid = sbi->s_resgid;
4267 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4268 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4269 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4270#ifdef CONFIG_QUOTA
4271 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4272 for (i = 0; i < MAXQUOTAS; i++)
4273 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4274#endif
b3881f74
TT		 4275 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4276 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4277
4278 /*
4279 * Allow the "check" option to be passed as a remount option.
4280 */
661aa520 4281 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4282 err = -EINVAL;
4283 goto restore_opts;
4284 }
4285
4ab2f15b 4286 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4287 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4288
4289 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4290 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4291
4292 es = sbi->s_es;
4293
b3881f74 4294 if (sbi->s_journal) {
0390131b 4295 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4296 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4297 }
ac27a0ec 4298
661aa520 4299 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4300 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4301 err = -EROFS;
4302 goto restore_opts;
4303 }
4304
4305 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4306 err = dquot_suspend(sb, -1);
4307 if (err < 0)
c79d967d 4308 goto restore_opts;
c79d967d 4309
ac27a0ec
DK		 4310 /*
4311 * First of all, the unconditional stuff we have to do
4312 * to disable replay of the journal when we next remount
4313 */
4314 sb->s_flags |= MS_RDONLY;
4315
4316 /*
4317 * OK, test if we are remounting a valid rw partition
4318 * readonly, and if so set the rdonly flag and then
4319 * mark the partition as valid again.
4320 */
617ba13b
MC		 4321 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4322 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4323 es->s_state = cpu_to_le16(sbi->s_mount_state);
4324
a63c9eb2 4325 if (sbi->s_journal)
0390131b 4326 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4327 } else {
a13fb1a4
ES		 4328 /* Make sure we can mount this feature set readwrite */
4329 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4330 err = -EROFS;
4331 goto restore_opts;
4332 }
8a266467
TT		 4333 /*
4334 * Make sure the group descriptor checksums
0b8e58a1 4335 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4336 */
4337 for (g = 0; g < sbi->s_groups_count; g++) {
4338 struct ext4_group_desc *gdp =
4339 ext4_get_group_desc(sb, g, NULL);
4340
4341 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 4342 ext4_msg(sb, KERN_ERR,
4343 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4344 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4345 le16_to_cpu(gdp->bg_checksum));
4346 err = -EINVAL;
4347 goto restore_opts;
4348 }
4349 }
4350
ead6596b
ES		 4351 /*
4352 * If we have an unprocessed orphan list hanging
4353 * around from a previously readonly bdev mount,
4354 * require a full umount/remount for now.
4355 */
4356 if (es->s_last_orphan) {
b31e1552 4357 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4358 "remount RDWR because of unprocessed "
4359 "orphan inode list. Please "
b31e1552 4360 "umount/remount instead");
ead6596b
ES		 4361 err = -EINVAL;
4362 goto restore_opts;
4363 }
4364
ac27a0ec
DK		 4365 /*
4366 * Mounting a RDONLY partition read-write, so reread
4367 * and store the current valid flag. (It may have
4368 * been changed by e2fsck since we originally mounted
4369 * the partition.)
4370 */
0390131b
FM		 4371 if (sbi->s_journal)
4372 ext4_clear_journal_err(sb, es);
ac27a0ec 4373 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4374 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4375 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4376 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4377 EXT4_FEATURE_INCOMPAT_MMP))
4378 if (ext4_multi_mount_protect(sb,
4379 le64_to_cpu(es->s_mmp_block))) {
4380 err = -EROFS;
4381 goto restore_opts;
4382 }
c79d967d 4383 enable_quota = 1;
ac27a0ec
DK		 4384 }
4385 }
bfff6873
LC		 4386
4387 /*
4388 * Reinitialize lazy itable initialization thread based on
4389 * current settings
4390 */
4391 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4392 ext4_unregister_li_request(sb);
4393 else {
4394 ext4_group_t first_not_zeroed;
4395 first_not_zeroed = ext4_has_uninit_itable(sb);
4396 ext4_register_li_request(sb, first_not_zeroed);
4397 }
4398
6fd058f7 4399 ext4_setup_system_zone(sb);
0390131b 4400 if (sbi->s_journal == NULL)
e2d67052 4401 ext4_commit_super(sb, 1);
0390131b 4402
ac27a0ec
DK		 4403#ifdef CONFIG_QUOTA
4404 /* Release old quota file names */
4405 for (i = 0; i < MAXQUOTAS; i++)
4406 if (old_opts.s_qf_names[i] &&
4407 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4408 kfree(old_opts.s_qf_names[i]);
4409#endif
bbd6851a 4410 unlock_super(sb);
c79d967d 4411 if (enable_quota)
0f0dd62f 4412 dquot_resume(sb, -1);
d4c402d9
CW		 4413
4414 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4415 kfree(orig_data);
ac27a0ec 4416 return 0;
0b8e58a1 4417
ac27a0ec
DK		 4418restore_opts:
4419 sb->s_flags = old_sb_flags;
4420 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4421 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4422 sbi->s_resuid = old_opts.s_resuid;
4423 sbi->s_resgid = old_opts.s_resgid;
4424 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4425 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4426 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4427#ifdef CONFIG_QUOTA
4428 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4429 for (i = 0; i < MAXQUOTAS; i++) {
4430 if (sbi->s_qf_names[i] &&
4431 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4432 kfree(sbi->s_qf_names[i]);
4433 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4434 }
4435#endif
bbd6851a 4436 unlock_super(sb);
d4c402d9 4437 kfree(orig_data);
ac27a0ec
DK		 4438 return err;
4439}
4440
f975d6bc
TT		 4441/*
4442 * Note: calculating the overhead so we can be compatible with
4443 * historical BSD practice is quite difficult in the face of
4444 * clusters/bigalloc. This is because multiple metadata blocks from
4445 * different block group can end up in the same allocation cluster.
4446 * Calculating the exact overhead in the face of clustered allocation
4447 * requires either O(all block bitmaps) in memory or O(number of block
4448 * groups**2) in time. We will still calculate the superblock for
4449 * older file systems --- and if we come across with a bigalloc file
4450 * system with zero in s_overhead_clusters the estimate will be close to
4451 * correct especially for very large cluster sizes --- but for newer
4452 * file systems, it's better to calculate this figure once at mkfs
4453 * time, and store it in the superblock. If the superblock value is
4454 * present (even for non-bigalloc file systems), we will use it.
4455 */
2b2d6d01 4456static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4457{
4458 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4459 struct ext4_sb_info *sbi = EXT4_SB(sb);
4460 struct ext4_super_block *es = sbi->s_es;
f975d6bc 4461 struct ext4_group_desc *gdp;
960cc398 4462 u64 fsid;
d02a9391 4463 s64 bfree;
ac27a0ec 4464
5e70030d
BP		 4465 if (test_opt(sb, MINIX_DF)) {
4466 sbi->s_overhead_last = 0;
f975d6bc
TT		 4467 } else if (es->s_overhead_clusters) {
4468 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
6bc9feff 4469 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 4470 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 4471 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 4472
4473 /*
5e70030d
BP		 4474 * Compute the overhead (FS structures). This is constant
4475 * for a given filesystem unless the number of block groups
4476 * changes so we cache the previous value until it does.
ac27a0ec
DK		 4477 */
4478
4479 /*
4480 * All of the blocks before first_data_block are
4481 * overhead
4482 */
f975d6bc 4483 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
ac27a0ec
DK		 4484
4485 /*
f975d6bc 4486 * Add the overhead found in each block group
ac27a0ec
DK		 4487 */
4488 for (i = 0; i < ngroups; i++) {
f975d6bc
TT		 4489 gdp = ext4_get_group_desc(sb, i, NULL);
4490 overhead += ext4_num_overhead_clusters(sb, i, gdp);
ac27a0ec
DK		 4491 cond_resched();
4492 }
5e70030d
BP		 4493 sbi->s_overhead_last = overhead;
4494 smp_wmb();
6bc9feff 4495 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 4496 }
4497
617ba13b 4498 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4499 buf->f_bsize = sb->s_blocksize;
f975d6bc
TT		 4500 buf->f_blocks = (ext4_blocks_count(es) -
4501 EXT4_C2B(sbi, sbi->s_overhead_last));
57042651
TT		 4502 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4503 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4504 /* prevent underflow in case that few free space is available */
57042651 4505 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4506 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4507 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4508 buf->f_bavail = 0;
4509 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4510 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4511 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4512 fsid = le64_to_cpup((void *)es->s_uuid) ^
4513 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4514 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4515 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4516
ac27a0ec
DK		 4517 return 0;
4518}
4519
0b8e58a1
AD		 4520/* Helper function for writing quotas on sync - we need to start transaction
4521 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4522 * Process 1 Process 2
617ba13b 4523 * ext4_create() quota_sync()
a269eb18 4524 * jbd2_journal_start() write_dquot()
871a2931 4525 * dquot_initialize() down(dqio_mutex)
dab291af 4526 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4527 *
4528 */
4529
4530#ifdef CONFIG_QUOTA
4531
4532static inline struct inode *dquot_to_inode(struct dquot *dquot)
4533{
4534 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4535}
4536
617ba13b 4537static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4538{
4539 int ret, err;
4540 handle_t *handle;
4541 struct inode *inode;
4542
4543 inode = dquot_to_inode(dquot);
617ba13b 4544 handle = ext4_journal_start(inode,
0b8e58a1 4545 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4546 if (IS_ERR(handle))
4547 return PTR_ERR(handle);
4548 ret = dquot_commit(dquot);
617ba13b 4549 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4550 if (!ret)
4551 ret = err;
4552 return ret;
4553}
4554
617ba13b 4555static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4556{
4557 int ret, err;
4558 handle_t *handle;
4559
617ba13b 4560 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4561 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4562 if (IS_ERR(handle))
4563 return PTR_ERR(handle);
4564 ret = dquot_acquire(dquot);
617ba13b 4565 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4566 if (!ret)
4567 ret = err;
4568 return ret;
4569}
4570
617ba13b 4571static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4572{
4573 int ret, err;
4574 handle_t *handle;
4575
617ba13b 4576 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4577 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4578 if (IS_ERR(handle)) {
4579 /* Release dquot anyway to avoid endless cycle in dqput() */
4580 dquot_release(dquot);
ac27a0ec 4581 return PTR_ERR(handle);
9c3013e9 4582 }
ac27a0ec 4583 ret = dquot_release(dquot);
617ba13b 4584 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4585 if (!ret)
4586 ret = err;
4587 return ret;
4588}
4589
617ba13b 4590static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4591{
2c8be6b2 4592 /* Are we journaling quotas? */
617ba13b
MC		 4593 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4594 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4595 dquot_mark_dquot_dirty(dquot);
617ba13b 4596 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4597 } else {
4598 return dquot_mark_dquot_dirty(dquot);
4599 }
4600}
4601
617ba13b 4602static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4603{
4604 int ret, err;
4605 handle_t *handle;
4606
4607 /* Data block + inode block */
617ba13b 4608 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4609 if (IS_ERR(handle))
4610 return PTR_ERR(handle);
4611 ret = dquot_commit_info(sb, type);
617ba13b 4612 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4613 if (!ret)
4614 ret = err;
4615 return ret;
4616}
4617
4618/*
4619 * Turn on quotas during mount time - we need to find
4620 * the quota file and such...
4621 */
617ba13b 4622static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4623{
287a8095
CH		 4624 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4625 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4626}
4627
4628/*
4629 * Standard function to be called on quota_on
4630 */
617ba13b 4631static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4632 struct path *path)
ac27a0ec
DK		 4633{
4634 int err;
ac27a0ec
DK		 4635
4636 if (!test_opt(sb, QUOTA))
4637 return -EINVAL;
0623543b 4638
ac27a0ec 4639 /* Quotafile not on the same filesystem? */
d8c9584e 4640 if (path->dentry->d_sb != sb)
ac27a0ec 4641 return -EXDEV;
0623543b
JK		 4642 /* Journaling quota? */
4643 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4644 /* Quotafile not in fs root? */
f00c9e44 4645 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4646 ext4_msg(sb, KERN_WARNING,
4647 "Quota file not on filesystem root. "
4648 "Journaled quota will not work");
2b2d6d01 4649 }
0623543b
JK		 4650
4651 /*
4652 * When we journal data on quota file, we have to flush journal to see
4653 * all updates to the file when we bypass pagecache...
4654 */
0390131b 4655 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4656 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4657 /*
4658 * We don't need to lock updates but journal_flush() could
4659 * otherwise be livelocked...
4660 */
4661 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4662 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4663 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4664 if (err)
7ffe1ea8 4665 return err;
0623543b
JK		 4666 }
4667
f00c9e44 4668 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4669}
4670
ca0e05e4
DM		 4671static int ext4_quota_off(struct super_block *sb, int type)
4672{
21f97697
JK		 4673 struct inode *inode = sb_dqopt(sb)->files[type];
4674 handle_t *handle;
4675
87009d86
DM		 4676 /* Force all delayed allocation blocks to be allocated.
4677 * Caller already holds s_umount sem */
4678 if (test_opt(sb, DELALLOC))
ca0e05e4 4679 sync_filesystem(sb);
ca0e05e4 4680
0b268590
AG		 4681 if (!inode)
4682 goto out;
4683
21f97697
JK		 4684 /* Update modification times of quota files when userspace can
4685 * start looking at them */
4686 handle = ext4_journal_start(inode, 1);
4687 if (IS_ERR(handle))
4688 goto out;
4689 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4690 ext4_mark_inode_dirty(handle, inode);
4691 ext4_journal_stop(handle);
4692
4693out:
ca0e05e4
DM		 4694 return dquot_quota_off(sb, type);
4695}
4696
ac27a0ec
DK		 4697/* Read data from quotafile - avoid pagecache and such because we cannot afford
4698 * acquiring the locks... As quota files are never truncated and quota code
25985edc 4699 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 4700 * we don't have to be afraid of races */
617ba13b 4701static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4702 size_t len, loff_t off)
4703{
4704 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4705 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4706 int err = 0;
4707 int offset = off & (sb->s_blocksize - 1);
4708 int tocopy;
4709 size_t toread;
4710 struct buffer_head *bh;
4711 loff_t i_size = i_size_read(inode);
4712
4713 if (off > i_size)
4714 return 0;
4715 if (off+len > i_size)
4716 len = i_size-off;
4717 toread = len;
4718 while (toread > 0) {
4719 tocopy = sb->s_blocksize - offset < toread ?
4720 sb->s_blocksize - offset : toread;
617ba13b 4721 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4722 if (err)
4723 return err;
4724 if (!bh) /* A hole? */
4725 memset(data, 0, tocopy);
4726 else
4727 memcpy(data, bh->b_data+offset, tocopy);
4728 brelse(bh);
4729 offset = 0;
4730 toread -= tocopy;
4731 data += tocopy;
4732 blk++;
4733 }
4734 return len;
4735}
4736
4737/* Write to quotafile (we know the transaction is already started and has
4738 * enough credits) */
617ba13b 4739static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4740 const char *data, size_t len, loff_t off)
4741{
4742 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4743 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4744 int err = 0;
4745 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 4746 struct buffer_head *bh;
4747 handle_t *handle = journal_current_handle();
4748
0390131b 4749 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4750 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4751 " cancelled because transaction is not started",
9c3013e9
JK		 4752 (unsigned long long)off, (unsigned long long)len);
4753 return -EIO;
4754 }
67eeb568
DM		 4755 /*
4756 * Since we account only one data block in transaction credits,
4757 * then it is impossible to cross a block boundary.
4758 */
4759 if (sb->s_blocksize - offset < len) {
4760 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4761 " cancelled because not block aligned",
4762 (unsigned long long)off, (unsigned long long)len);
4763 return -EIO;
4764 }
4765
ac27a0ec 4766 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4767 bh = ext4_bread(handle, inode, blk, 1, &err);
4768 if (!bh)
4769 goto out;
62d2b5f2
JK		 4770 err = ext4_journal_get_write_access(handle, bh);
4771 if (err) {
4772 brelse(bh);
4773 goto out;
ac27a0ec 4774 }
67eeb568
DM		 4775 lock_buffer(bh);
4776 memcpy(bh->b_data+offset, data, len);
4777 flush_dcache_page(bh->b_page);
4778 unlock_buffer(bh);
62d2b5f2 4779 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 4780 brelse(bh);
ac27a0ec 4781out:
67eeb568 4782 if (err) {
4d04e4fb 4783 mutex_unlock(&inode->i_mutex);
ac27a0ec 4784 return err;
4d04e4fb 4785 }
67eeb568
DM		 4786 if (inode->i_size < off + len) {
4787 i_size_write(inode, off + len);
617ba13b 4788 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 4789 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4790 }
ac27a0ec 4791 mutex_unlock(&inode->i_mutex);
67eeb568 4792 return len;
ac27a0ec
DK		 4793}
4794
4795#endif
4796
152a0836
AV		 4797static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4798 const char *dev_name, void *data)
ac27a0ec 4799{
152a0836 4800 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 4801}
4802
37f328eb 4803#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4804static inline void register_as_ext2(void)
4805{
4806 int err = register_filesystem(&ext2_fs_type);
4807 if (err)
4808 printk(KERN_WARNING
4809 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4810}
4811
4812static inline void unregister_as_ext2(void)
4813{
4814 unregister_filesystem(&ext2_fs_type);
4815}
2035e776
TT		 4816
4817static inline int ext2_feature_set_ok(struct super_block *sb)
4818{
4819 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4820 return 0;
4821 if (sb->s_flags & MS_RDONLY)
4822 return 1;
4823 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4824 return 0;
4825 return 1;
4826}
51b7e3c9 4827MODULE_ALIAS("ext2");
24b58424
TT		 4828#else
4829static inline void register_as_ext2(void) { }
4830static inline void unregister_as_ext2(void) { }
2035e776 4831static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4832#endif
4833
37f328eb 4834#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4835static inline void register_as_ext3(void)
4836{
4837 int err = register_filesystem(&ext3_fs_type);
4838 if (err)
4839 printk(KERN_WARNING
4840 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4841}
4842
4843static inline void unregister_as_ext3(void)
4844{
4845 unregister_filesystem(&ext3_fs_type);
4846}
2035e776
TT		 4847
4848static inline int ext3_feature_set_ok(struct super_block *sb)
4849{
4850 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4851 return 0;
4852 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4853 return 0;
4854 if (sb->s_flags & MS_RDONLY)
4855 return 1;
4856 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4857 return 0;
4858 return 1;
4859}
51b7e3c9 4860MODULE_ALIAS("ext3");
24b58424
TT		 4861#else
4862static inline void register_as_ext3(void) { }
4863static inline void unregister_as_ext3(void) { }
2035e776 4864static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4865#endif
4866
03010a33
TT		 4867static struct file_system_type ext4_fs_type = {
4868 .owner = THIS_MODULE,
4869 .name = "ext4",
152a0836 4870 .mount = ext4_mount,
03010a33
TT		 4871 .kill_sb = kill_block_super,
4872 .fs_flags = FS_REQUIRES_DEV,
4873};
4874
8f021222 4875static int __init ext4_init_feat_adverts(void)
857ac889
LC		 4876{
4877 struct ext4_features *ef;
4878 int ret = -ENOMEM;
4879
4880 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4881 if (!ef)
4882 goto out;
4883
4884 ef->f_kobj.kset = ext4_kset;
4885 init_completion(&ef->f_kobj_unregister);
4886 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4887 "features");
4888 if (ret) {
4889 kfree(ef);
4890 goto out;
4891 }
4892
4893 ext4_feat = ef;
4894 ret = 0;
4895out:
4896 return ret;
4897}
4898
8f021222
LC		 4899static void ext4_exit_feat_adverts(void)
4900{
4901 kobject_put(&ext4_feat->f_kobj);
4902 wait_for_completion(&ext4_feat->f_kobj_unregister);
4903 kfree(ext4_feat);
4904}
4905
e9e3bcec
ES		 4906/* Shared across all ext4 file systems */
4907wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4908struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4909
5dabfc78 4910static int __init ext4_init_fs(void)
ac27a0ec 4911{
e9e3bcec 4912 int i, err;
c9de560d 4913
12e9b892 4914 ext4_check_flag_values();
e9e3bcec
ES		 4915
4916 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4917 mutex_init(&ext4__aio_mutex[i]);
4918 init_waitqueue_head(&ext4__ioend_wq[i]);
4919 }
4920
5dabfc78 4921 err = ext4_init_pageio();
6fd058f7
TT		 4922 if (err)
4923 return err;
5dabfc78 4924 err = ext4_init_system_zone();
bd2d0210 4925 if (err)
d44651d0 4926 goto out6;
3197ebdb
TT		 4927 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4928 if (!ext4_kset)
dd68314c 4929 goto out5;
d44651d0 4930 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 4931
4932 err = ext4_init_feat_adverts();
dd68314c
TT		 4933 if (err)
4934 goto out4;
857ac889 4935
5dabfc78 4936 err = ext4_init_mballoc();
ac27a0ec 4937 if (err)
6fd058f7 4938 goto out3;
c9de560d 4939
5dabfc78 4940 err = ext4_init_xattr();
c9de560d
AT		 4941 if (err)
4942 goto out2;
ac27a0ec
DK		 4943 err = init_inodecache();
4944 if (err)
4945 goto out1;
24b58424 4946 register_as_ext3();
2035e776 4947 register_as_ext2();
03010a33 4948 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 4949 if (err)
4950 goto out;
bfff6873
LC		 4951
4952 ext4_li_info = NULL;
4953 mutex_init(&ext4_li_mtx);
ac27a0ec
DK		 4954 return 0;
4955out:
24b58424
TT		 4956 unregister_as_ext2();
4957 unregister_as_ext3();
ac27a0ec
DK		 4958 destroy_inodecache();
4959out1:
5dabfc78 4960 ext4_exit_xattr();
c9de560d 4961out2:
5dabfc78 4962 ext4_exit_mballoc();
6fd058f7 4963out3:
8f021222 4964 ext4_exit_feat_adverts();
dd68314c 4965out4:
d44651d0
FJ		 4966 if (ext4_proc_root)
4967 remove_proc_entry("fs/ext4", NULL);
6fd058f7 4968 kset_unregister(ext4_kset);
d44651d0 4969out5:
5dabfc78 4970 ext4_exit_system_zone();
d44651d0 4971out6:
5dabfc78 4972 ext4_exit_pageio();
ac27a0ec
DK		 4973 return err;
4974}
4975
5dabfc78 4976static void __exit ext4_exit_fs(void)
ac27a0ec 4977{
bfff6873 4978 ext4_destroy_lazyinit_thread();
24b58424
TT		 4979 unregister_as_ext2();
4980 unregister_as_ext3();
03010a33 4981 unregister_filesystem(&ext4_fs_type);
ac27a0ec 4982 destroy_inodecache();
5dabfc78
TT		 4983 ext4_exit_xattr();
4984 ext4_exit_mballoc();
8f021222 4985 ext4_exit_feat_adverts();
9f6200bb 4986 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4987 kset_unregister(ext4_kset);
5dabfc78
TT		 4988 ext4_exit_system_zone();
4989 ext4_exit_pageio();
ac27a0ec
DK		 4990}
4991
4992MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4993MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4994MODULE_LICENSE("GPL");
5dabfc78
TT		 4995module_init(ext4_init_fs)
4996module_exit(ext4_exit_fs)
Linux kernel - Deliverables
This page took 0.843608 seconds and 5 git commands to generate.