projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
userns: Convert ext3 to use kuid/kgid where appropriate
[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);
0390131b
FM		 839 if (sbi->s_journal) {
840 err = jbd2_journal_destroy(sbi->s_journal);
841 sbi->s_journal = NULL;
842 if (err < 0)
c67d859e 843 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 844 }
d4edac31 845
a1c6c569 846 del_timer(&sbi->s_err_report);
d4edac31
JB		 847 ext4_release_system_zone(sb);
848 ext4_mb_release(sb);
849 ext4_ext_release(sb);
850 ext4_xattr_put_super(sb);
851
ac27a0ec 852 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 853 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 854 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 855 }
a8e25a83
AB		 856 if (sb->s_dirt || !(sb->s_flags & MS_RDONLY))
857 ext4_commit_super(sb, 1);
858
240799cd 859 if (sbi->s_proc) {
66acdcf4 860 remove_proc_entry("options", sbi->s_proc);
9f6200bb 861 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 862 }
3197ebdb 863 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 864
865 for (i = 0; i < sbi->s_gdb_count; i++)
866 brelse(sbi->s_group_desc[i]);
f18a5f21 867 ext4_kvfree(sbi->s_group_desc);
9933fc0a 868 ext4_kvfree(sbi->s_flex_groups);
57042651 869 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 870 percpu_counter_destroy(&sbi->s_freeinodes_counter);
871 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 872 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 873 brelse(sbi->s_sbh);
874#ifdef CONFIG_QUOTA
875 for (i = 0; i < MAXQUOTAS; i++)
876 kfree(sbi->s_qf_names[i]);
877#endif
878
879 /* Debugging code just in case the in-memory inode orphan list
880 * isn't empty. The on-disk one can be non-empty if we've
881 * detected an error and taken the fs readonly, but the
882 * in-memory list had better be clean by this point. */
883 if (!list_empty(&sbi->s_orphan))
884 dump_orphan_list(sb, sbi);
885 J_ASSERT(list_empty(&sbi->s_orphan));
886
f98393a6 887 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 888 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
889 /*
890 * Invalidate the journal device's buffers. We don't want them
891 * floating about in memory - the physical journal device may
892 * hotswapped, and it breaks the `ro-after' testing code.
893 */
894 sync_blockdev(sbi->journal_bdev);
f98393a6 895 invalidate_bdev(sbi->journal_bdev);
617ba13b 896 ext4_blkdev_remove(sbi);
ac27a0ec 897 }
c5e06d10
JL		 898 if (sbi->s_mmp_tsk)
899 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 900 sb->s_fs_info = NULL;
3197ebdb
TT		 901 /*
902 * Now that we are completely done shutting down the
903 * superblock, we need to actually destroy the kobject.
904 */
3197ebdb
TT		 905 unlock_super(sb);
906 kobject_put(&sbi->s_kobj);
907 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 908 kfree(sbi->s_blockgroup_lock);
ac27a0ec 909 kfree(sbi);
ac27a0ec
DK		 910}
911
e18b890b 912static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 913
914/*
915 * Called inside transaction, so use GFP_NOFS
916 */
617ba13b 917static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 918{
617ba13b 919 struct ext4_inode_info *ei;
ac27a0ec 920
e6b4f8da 921 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 922 if (!ei)
923 return NULL;
0b8e58a1 924
ac27a0ec 925 ei->vfs_inode.i_version = 1;
91246c00 926 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 927 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 928 INIT_LIST_HEAD(&ei->i_prealloc_list);
929 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 930 ei->i_reserved_data_blocks = 0;
931 ei->i_reserved_meta_blocks = 0;
932 ei->i_allocated_meta_blocks = 0;
9d0be502 933 ei->i_da_metadata_calc_len = 0;
d2a17637 934 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 935#ifdef CONFIG_QUOTA
936 ei->i_reserved_quota = 0;
937#endif
8aefcd55 938 ei->jinode = NULL;
c7064ef1 939 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 940 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 941 ei->cur_aio_dio = NULL;
b436b9be
JK		 942 ei->i_sync_tid = 0;
943 ei->i_datasync_tid = 0;
f7ad6d2e 944 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 945 atomic_set(&ei->i_aiodio_unwritten, 0);
0b8e58a1 946
ac27a0ec
DK		 947 return &ei->vfs_inode;
948}
949
7ff9c073
TT		 950static int ext4_drop_inode(struct inode *inode)
951{
952 int drop = generic_drop_inode(inode);
953
954 trace_ext4_drop_inode(inode, drop);
955 return drop;
956}
957
fa0d7e3d
NP		 958static void ext4_i_callback(struct rcu_head *head)
959{
960 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 961 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
962}
963
617ba13b 964static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 965{
9f7dd93d 966 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 967 ext4_msg(inode->i_sb, KERN_ERR,
968 "Inode %lu (%p): orphan list check failed!",
969 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 970 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
971 EXT4_I(inode), sizeof(struct ext4_inode_info),
972 true);
973 dump_stack();
974 }
fa0d7e3d 975 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 976}
977
51cc5068 978static void init_once(void *foo)
ac27a0ec 979{
617ba13b 980 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 981
a35afb83 982 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 983#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 984 init_rwsem(&ei->xattr_sem);
ac27a0ec 985#endif
0e855ac8 986 init_rwsem(&ei->i_data_sem);
a35afb83 987 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 988}
989
990static int init_inodecache(void)
991{
617ba13b
MC		 992 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
993 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 994 0, (SLAB_RECLAIM_ACCOUNT|
995 SLAB_MEM_SPREAD),
20c2df83 996 init_once);
617ba13b 997 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 998 return -ENOMEM;
999 return 0;
1000}
1001
1002static void destroy_inodecache(void)
1003{
617ba13b 1004 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1005}
1006
0930fcc1 1007void ext4_clear_inode(struct inode *inode)
ac27a0ec 1008{
0930fcc1
AV		 1009 invalidate_inode_buffers(inode);
1010 end_writeback(inode);
9f754758 1011 dquot_drop(inode);
c2ea3fde 1012 ext4_discard_preallocations(inode);
8aefcd55
TT		 1013 if (EXT4_I(inode)->jinode) {
1014 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1015 EXT4_I(inode)->jinode);
1016 jbd2_free_inode(EXT4_I(inode)->jinode);
1017 EXT4_I(inode)->jinode = NULL;
1018 }
ac27a0ec
DK		 1019}
1020
1b961ac0 1021static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1022 u64 ino, u32 generation)
ac27a0ec 1023{
ac27a0ec 1024 struct inode *inode;
ac27a0ec 1025
617ba13b 1026 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1027 return ERR_PTR(-ESTALE);
617ba13b 1028 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1029 return ERR_PTR(-ESTALE);
1030
1031 /* iget isn't really right if the inode is currently unallocated!!
1032 *
617ba13b 1033 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1034 * deleted, so we should be safe.
1035 *
1036 * Currently we don't know the generation for parent directory, so
1037 * a generation of 0 means "accept any"
1038 */
1d1fe1ee
DH		 1039 inode = ext4_iget(sb, ino);
1040 if (IS_ERR(inode))
1041 return ERR_CAST(inode);
1042 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1043 iput(inode);
1044 return ERR_PTR(-ESTALE);
1045 }
1b961ac0
CH		 1046
1047 return inode;
1048}
1049
1050static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1051 int fh_len, int fh_type)
1b961ac0
CH		 1052{
1053 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1054 ext4_nfs_get_inode);
1055}
1056
1057static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1058 int fh_len, int fh_type)
1b961ac0
CH		 1059{
1060 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1061 ext4_nfs_get_inode);
ac27a0ec
DK		 1062}
1063
c39a7f84
TO		 1064/*
1065 * Try to release metadata pages (indirect blocks, directories) which are
1066 * mapped via the block device. Since these pages could have journal heads
1067 * which would prevent try_to_free_buffers() from freeing them, we must use
1068 * jbd2 layer's try_to_free_buffers() function to release them.
1069 */
0b8e58a1
AD		 1070static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1071 gfp_t wait)
c39a7f84
TO		 1072{
1073 journal_t *journal = EXT4_SB(sb)->s_journal;
1074
1075 WARN_ON(PageChecked(page));
1076 if (!page_has_buffers(page))
1077 return 0;
1078 if (journal)
1079 return jbd2_journal_try_to_free_buffers(journal, page,
1080 wait & ~__GFP_WAIT);
1081 return try_to_free_buffers(page);
1082}
1083
ac27a0ec 1084#ifdef CONFIG_QUOTA
af5bc92d 1085#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1086#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1087
617ba13b
MC		 1088static int ext4_write_dquot(struct dquot *dquot);
1089static int ext4_acquire_dquot(struct dquot *dquot);
1090static int ext4_release_dquot(struct dquot *dquot);
1091static int ext4_mark_dquot_dirty(struct dquot *dquot);
1092static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1093static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1094 struct path *path);
ca0e05e4 1095static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1096static int ext4_quota_on_mount(struct super_block *sb, int type);
1097static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1098 size_t len, loff_t off);
617ba13b 1099static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1100 const char *data, size_t len, loff_t off);
1101
61e225dc 1102static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1103 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1104 .write_dquot = ext4_write_dquot,
1105 .acquire_dquot = ext4_acquire_dquot,
1106 .release_dquot = ext4_release_dquot,
1107 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1108 .write_info = ext4_write_info,
1109 .alloc_dquot = dquot_alloc,
1110 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1111};
1112
0d54b217 1113static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1114 .quota_on = ext4_quota_on,
ca0e05e4 1115 .quota_off = ext4_quota_off,
287a8095
CH		 1116 .quota_sync = dquot_quota_sync,
1117 .get_info = dquot_get_dqinfo,
1118 .set_info = dquot_set_dqinfo,
1119 .get_dqblk = dquot_get_dqblk,
1120 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1121};
1122#endif
1123
ee9b6d61 1124static const struct super_operations ext4_sops = {
617ba13b
MC		 1125 .alloc_inode = ext4_alloc_inode,
1126 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1127 .write_inode = ext4_write_inode,
1128 .dirty_inode = ext4_dirty_inode,
7ff9c073 1129 .drop_inode = ext4_drop_inode,
0930fcc1 1130 .evict_inode = ext4_evict_inode,
617ba13b 1131 .put_super = ext4_put_super,
617ba13b 1132 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1133 .freeze_fs = ext4_freeze,
1134 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1135 .statfs = ext4_statfs,
1136 .remount_fs = ext4_remount,
617ba13b 1137 .show_options = ext4_show_options,
ac27a0ec 1138#ifdef CONFIG_QUOTA
617ba13b
MC		 1139 .quota_read = ext4_quota_read,
1140 .quota_write = ext4_quota_write,
ac27a0ec 1141#endif
c39a7f84 1142 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1143};
1144
9ca92389
TT		 1145static const struct super_operations ext4_nojournal_sops = {
1146 .alloc_inode = ext4_alloc_inode,
1147 .destroy_inode = ext4_destroy_inode,
1148 .write_inode = ext4_write_inode,
1149 .dirty_inode = ext4_dirty_inode,
7ff9c073 1150 .drop_inode = ext4_drop_inode,
0930fcc1 1151 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1152 .write_super = ext4_write_super,
1153 .put_super = ext4_put_super,
1154 .statfs = ext4_statfs,
1155 .remount_fs = ext4_remount,
9ca92389
TT		 1156 .show_options = ext4_show_options,
1157#ifdef CONFIG_QUOTA
1158 .quota_read = ext4_quota_read,
1159 .quota_write = ext4_quota_write,
1160#endif
1161 .bdev_try_to_free_page = bdev_try_to_free_page,
1162};
1163
39655164 1164static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1165 .fh_to_dentry = ext4_fh_to_dentry,
1166 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1167 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1168};
1169
1170enum {
1171 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1172 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1173 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1174 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1175 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1176 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1177 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1178 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1179 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1180 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1181 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1182 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1183 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1184 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1185 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1186 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1187 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1188 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
ac27a0ec
DK		 1189};
1190
a447c093 1191static const match_table_t tokens = {
ac27a0ec
DK		 1192 {Opt_bsd_df, "bsddf"},
1193 {Opt_minix_df, "minixdf"},
1194 {Opt_grpid, "grpid"},
1195 {Opt_grpid, "bsdgroups"},
1196 {Opt_nogrpid, "nogrpid"},
1197 {Opt_nogrpid, "sysvgroups"},
1198 {Opt_resgid, "resgid=%u"},
1199 {Opt_resuid, "resuid=%u"},
1200 {Opt_sb, "sb=%u"},
1201 {Opt_err_cont, "errors=continue"},
1202 {Opt_err_panic, "errors=panic"},
1203 {Opt_err_ro, "errors=remount-ro"},
1204 {Opt_nouid32, "nouid32"},
ac27a0ec 1205 {Opt_debug, "debug"},
72578c33
TT		 1206 {Opt_removed, "oldalloc"},
1207 {Opt_removed, "orlov"},
ac27a0ec
DK		 1208 {Opt_user_xattr, "user_xattr"},
1209 {Opt_nouser_xattr, "nouser_xattr"},
1210 {Opt_acl, "acl"},
1211 {Opt_noacl, "noacl"},
e3bb52ae 1212 {Opt_noload, "norecovery"},
5a916be1 1213 {Opt_noload, "noload"},
72578c33
TT		 1214 {Opt_removed, "nobh"},
1215 {Opt_removed, "bh"},
ac27a0ec 1216 {Opt_commit, "commit=%u"},
30773840
TT		 1217 {Opt_min_batch_time, "min_batch_time=%u"},
1218 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1219 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1220 {Opt_journal_checksum, "journal_checksum"},
1221 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1222 {Opt_abort, "abort"},
1223 {Opt_data_journal, "data=journal"},
1224 {Opt_data_ordered, "data=ordered"},
1225 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1226 {Opt_data_err_abort, "data_err=abort"},
1227 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1228 {Opt_offusrjquota, "usrjquota="},
1229 {Opt_usrjquota, "usrjquota=%s"},
1230 {Opt_offgrpjquota, "grpjquota="},
1231 {Opt_grpjquota, "grpjquota=%s"},
1232 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1233 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1234 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1235 {Opt_grpquota, "grpquota"},
1236 {Opt_noquota, "noquota"},
1237 {Opt_quota, "quota"},
1238 {Opt_usrquota, "usrquota"},
1239 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1240 {Opt_barrier, "barrier"},
1241 {Opt_nobarrier, "nobarrier"},
25ec56b5 1242 {Opt_i_version, "i_version"},
c9de560d 1243 {Opt_stripe, "stripe=%u"},
64769240 1244 {Opt_delalloc, "delalloc"},
dd919b98 1245 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1246 {Opt_mblk_io_submit, "mblk_io_submit"},
1247 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1248 {Opt_block_validity, "block_validity"},
1249 {Opt_noblock_validity, "noblock_validity"},
240799cd 1250 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1251 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1252 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1253 {Opt_auto_da_alloc, "auto_da_alloc"},
1254 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1255 {Opt_dioread_nolock, "dioread_nolock"},
1256 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1257 {Opt_discard, "discard"},
1258 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1259 {Opt_init_itable, "init_itable=%u"},
1260 {Opt_init_itable, "init_itable"},
1261 {Opt_noinit_itable, "noinit_itable"},
c7198b9c
TT		 1262 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1263 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1264 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1265 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1266 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1267 {Opt_err, NULL},
ac27a0ec
DK		 1268};
1269
617ba13b 1270static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1271{
617ba13b 1272 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1273 char *options = (char *) *data;
1274
1275 if (!options || strncmp(options, "sb=", 3) != 0)
1276 return 1; /* Default location */
0b8e58a1 1277
ac27a0ec 1278 options += 3;
0b8e58a1 1279 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1280 sb_block = simple_strtoul(options, &options, 0);
1281 if (*options && *options != ',') {
4776004f 1282 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1283 (char *) *data);
1284 return 1;
1285 }
1286 if (*options == ',')
1287 options++;
1288 *data = (void *) options;
0b8e58a1 1289
ac27a0ec
DK		 1290 return sb_block;
1291}
1292
b3881f74 1293#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1294static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1295 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1296
56c50f11
DM		 1297#ifdef CONFIG_QUOTA
1298static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1299{
1300 struct ext4_sb_info *sbi = EXT4_SB(sb);
1301 char *qname;
1302
1303 if (sb_any_quota_loaded(sb) &&
1304 !sbi->s_qf_names[qtype]) {
1305 ext4_msg(sb, KERN_ERR,
1306 "Cannot change journaled "
1307 "quota options when quota turned on");
1308 return 0;
1309 }
1310 qname = match_strdup(args);
1311 if (!qname) {
1312 ext4_msg(sb, KERN_ERR,
1313 "Not enough memory for storing quotafile name");
1314 return 0;
1315 }
1316 if (sbi->s_qf_names[qtype] &&
1317 strcmp(sbi->s_qf_names[qtype], qname)) {
1318 ext4_msg(sb, KERN_ERR,
1319 "%s quota file already specified", QTYPE2NAME(qtype));
1320 kfree(qname);
1321 return 0;
1322 }
1323 sbi->s_qf_names[qtype] = qname;
1324 if (strchr(sbi->s_qf_names[qtype], '/')) {
1325 ext4_msg(sb, KERN_ERR,
1326 "quotafile must be on filesystem root");
1327 kfree(sbi->s_qf_names[qtype]);
1328 sbi->s_qf_names[qtype] = NULL;
1329 return 0;
1330 }
fd8c37ec 1331 set_opt(sb, QUOTA);
56c50f11
DM		 1332 return 1;
1333}
1334
1335static int clear_qf_name(struct super_block *sb, int qtype)
1336{
1337
1338 struct ext4_sb_info *sbi = EXT4_SB(sb);
1339
1340 if (sb_any_quota_loaded(sb) &&
1341 sbi->s_qf_names[qtype]) {
1342 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1343 " when quota turned on");
1344 return 0;
1345 }
1346 /*
1347 * The space will be released later when all options are confirmed
1348 * to be correct
1349 */
1350 sbi->s_qf_names[qtype] = NULL;
1351 return 1;
1352}
1353#endif
1354
26092bf5
TT		 1355#define MOPT_SET 0x0001
1356#define MOPT_CLEAR 0x0002
1357#define MOPT_NOSUPPORT 0x0004
1358#define MOPT_EXPLICIT 0x0008
1359#define MOPT_CLEAR_ERR 0x0010
1360#define MOPT_GTE0 0x0020
ac27a0ec 1361#ifdef CONFIG_QUOTA
26092bf5
TT		 1362#define MOPT_Q 0
1363#define MOPT_QFMT 0x0040
1364#else
1365#define MOPT_Q MOPT_NOSUPPORT
1366#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1367#endif
26092bf5
TT		 1368#define MOPT_DATAJ 0x0080
1369
1370static const struct mount_opts {
1371 int token;
1372 int mount_opt;
1373 int flags;
1374} ext4_mount_opts[] = {
1375 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1376 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1377 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1378 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1379 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1380 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1381 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1382 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1383 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1384 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1385 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1386 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1387 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1388 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1389 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1390 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1391 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1392 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1393 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1394 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1395 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1396 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1397 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1398 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1399 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1400 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1401 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1402 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1403 {Opt_commit, 0, MOPT_GTE0},
1404 {Opt_max_batch_time, 0, MOPT_GTE0},
1405 {Opt_min_batch_time, 0, MOPT_GTE0},
1406 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1407 {Opt_init_itable, 0, MOPT_GTE0},
1408 {Opt_stripe, 0, MOPT_GTE0},
1409 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1410 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1411 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
03010a33 1412#ifdef CONFIG_EXT4_FS_XATTR
26092bf5
TT		 1413 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1414 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
ac27a0ec 1415#else
26092bf5
TT		 1416 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1417 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
ac27a0ec 1418#endif
03010a33 1419#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1420 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1421 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1422#else
26092bf5
TT		 1423 {Opt_acl, 0, MOPT_NOSUPPORT},
1424 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1425#endif
26092bf5
TT		 1426 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1427 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1428 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1429 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1430 MOPT_SET | MOPT_Q},
1431 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1432 MOPT_SET | MOPT_Q},
1433 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1434 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1435 {Opt_usrjquota, 0, MOPT_Q},
1436 {Opt_grpjquota, 0, MOPT_Q},
1437 {Opt_offusrjquota, 0, MOPT_Q},
1438 {Opt_offgrpjquota, 0, MOPT_Q},
1439 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1440 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1441 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1442 {Opt_err, 0, 0}
1443};
1444
1445static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1446 substring_t *args, unsigned long *journal_devnum,
1447 unsigned int *journal_ioprio, int is_remount)
1448{
1449 struct ext4_sb_info *sbi = EXT4_SB(sb);
1450 const struct mount_opts *m;
1451 int arg = 0;
1452
1453 if (args->from && match_int(args, &arg))
1454 return -1;
1455 switch (token) {
f7048605
TT		 1456 case Opt_noacl:
1457 case Opt_nouser_xattr:
1458 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1459 break;
26092bf5
TT		 1460 case Opt_sb:
1461 return 1; /* handled by get_sb_block() */
1462 case Opt_removed:
1463 ext4_msg(sb, KERN_WARNING,
1464 "Ignoring removed %s option", opt);
1465 return 1;
1466 case Opt_resuid:
1467 sbi->s_resuid = arg;
1468 return 1;
1469 case Opt_resgid:
1470 sbi->s_resgid = arg;
1471 return 1;
1472 case Opt_abort:
1473 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1474 return 1;
1475 case Opt_i_version:
1476 sb->s_flags |= MS_I_VERSION;
1477 return 1;
1478 case Opt_journal_dev:
1479 if (is_remount) {
1480 ext4_msg(sb, KERN_ERR,
1481 "Cannot specify journal on remount");
1482 return -1;
1483 }
1484 *journal_devnum = arg;
1485 return 1;
1486 case Opt_journal_ioprio:
1487 if (arg < 0 || arg > 7)
1488 return -1;
1489 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1490 return 1;
1491 }
1492
1493 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1494 if (token != m->token)
1495 continue;
1496 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1497 return -1;
1498 if (m->flags & MOPT_EXPLICIT)
1499 set_opt2(sb, EXPLICIT_DELALLOC);
1500 if (m->flags & MOPT_CLEAR_ERR)
1501 clear_opt(sb, ERRORS_MASK);
1502 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1503 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1504 "options when quota turned on");
1505 return -1;
1506 }
1507
1508 if (m->flags & MOPT_NOSUPPORT) {
1509 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1510 } else if (token == Opt_commit) {
1511 if (arg == 0)
1512 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1513 sbi->s_commit_interval = HZ * arg;
1514 } else if (token == Opt_max_batch_time) {
1515 if (arg == 0)
1516 arg = EXT4_DEF_MAX_BATCH_TIME;
1517 sbi->s_max_batch_time = arg;
1518 } else if (token == Opt_min_batch_time) {
1519 sbi->s_min_batch_time = arg;
1520 } else if (token == Opt_inode_readahead_blks) {
1521 if (arg > (1 << 30))
1522 return -1;
1523 if (arg && !is_power_of_2(arg)) {
b31e1552 1524 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1525 "EXT4-fs: inode_readahead_blks"
1526 " must be a power of 2");
1527 return -1;
ac27a0ec 1528 }
26092bf5
TT		 1529 sbi->s_inode_readahead_blks = arg;
1530 } else if (token == Opt_init_itable) {
1531 set_opt(sb, INIT_INODE_TABLE);
1532 if (!args->from)
1533 arg = EXT4_DEF_LI_WAIT_MULT;
1534 sbi->s_li_wait_mult = arg;
1535 } else if (token == Opt_stripe) {
1536 sbi->s_stripe = arg;
1537 } else if (m->flags & MOPT_DATAJ) {
ac27a0ec 1538 if (is_remount) {
eb513689
TT		 1539 if (!sbi->s_journal)
1540 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
26092bf5
TT		 1541 else if (test_opt(sb, DATA_FLAGS) !=
1542 m->mount_opt) {
b31e1552 1543 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1544 "Cannot change data mode on remount");
1545 return -1;
ac27a0ec
DK		 1546 }
1547 } else {
fd8c37ec 1548 clear_opt(sb, DATA_FLAGS);
26092bf5 1549 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1550 }
ac27a0ec 1551#ifdef CONFIG_QUOTA
26092bf5 1552 } else if (token == Opt_usrjquota) {
56c50f11 1553 if (!set_qf_name(sb, USRQUOTA, &args[0]))
26092bf5
TT		 1554 return -1;
1555 } else if (token == Opt_grpjquota) {
56c50f11 1556 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
26092bf5
TT		 1557 return -1;
1558 } else if (token == Opt_offusrjquota) {
56c50f11 1559 if (!clear_qf_name(sb, USRQUOTA))
26092bf5
TT		 1560 return -1;
1561 } else if (token == Opt_offgrpjquota) {
56c50f11 1562 if (!clear_qf_name(sb, GRPQUOTA))
26092bf5
TT		 1563 return -1;
1564 } else if (m->flags & MOPT_QFMT) {
17bd13b3 1565 if (sb_any_quota_loaded(sb) &&
26092bf5
TT		 1566 sbi->s_jquota_fmt != m->mount_opt) {
1567 ext4_msg(sb, KERN_ERR, "Cannot "
1568 "change journaled quota options "
1569 "when quota turned on");
1570 return -1;
ac27a0ec 1571 }
26092bf5 1572 sbi->s_jquota_fmt = m->mount_opt;
ac27a0ec 1573#endif
26092bf5
TT		 1574 } else {
1575 if (!args->from)
1576 arg = 1;
1577 if (m->flags & MOPT_CLEAR)
1578 arg = !arg;
1579 else if (unlikely(!(m->flags & MOPT_SET))) {
1580 ext4_msg(sb, KERN_WARNING,
1581 "buggy handling of option %s", opt);
1582 WARN_ON(1);
1583 return -1;
3197ebdb 1584 }
26092bf5
TT		 1585 if (arg != 0)
1586 sbi->s_mount_opt |= m->mount_opt;
afd4672d 1587 else
26092bf5 1588 sbi->s_mount_opt &= ~m->mount_opt;
ac27a0ec 1589 }
26092bf5
TT		 1590 return 1;
1591 }
1592 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1593 "or missing value", opt);
1594 return -1;
1595}
1596
1597static int parse_options(char *options, struct super_block *sb,
1598 unsigned long *journal_devnum,
1599 unsigned int *journal_ioprio,
1600 int is_remount)
1601{
1602 struct ext4_sb_info *sbi = EXT4_SB(sb);
1603 char *p;
1604 substring_t args[MAX_OPT_ARGS];
1605 int token;
1606
1607 if (!options)
1608 return 1;
1609
1610 while ((p = strsep(&options, ",")) != NULL) {
1611 if (!*p)
1612 continue;
1613 /*
1614 * Initialize args struct so we know whether arg was
1615 * found; some options take optional arguments.
1616 */
1617 args[0].to = args[0].from = 0;
1618 token = match_token(p, tokens, args);
1619 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1620 journal_ioprio, is_remount) < 0)
1621 return 0;
ac27a0ec
DK		 1622 }
1623#ifdef CONFIG_QUOTA
1624 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1625 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1626 clear_opt(sb, USRQUOTA);
ac27a0ec 1627
482a7425 1628 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1629 clear_opt(sb, GRPQUOTA);
ac27a0ec 1630
56c50f11 1631 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1632 ext4_msg(sb, KERN_ERR, "old and new quota "
1633 "format mixing");
ac27a0ec
DK		 1634 return 0;
1635 }
1636
1637 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1638 ext4_msg(sb, KERN_ERR, "journaled quota format "
1639 "not specified");
ac27a0ec
DK		 1640 return 0;
1641 }
1642 } else {
1643 if (sbi->s_jquota_fmt) {
b31e1552 1644 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1645 "specified with no journaling "
b31e1552 1646 "enabled");
ac27a0ec
DK		 1647 return 0;
1648 }
1649 }
1650#endif
1651 return 1;
1652}
1653
2adf6da8
TT		 1654static inline void ext4_show_quota_options(struct seq_file *seq,
1655 struct super_block *sb)
1656{
1657#if defined(CONFIG_QUOTA)
1658 struct ext4_sb_info *sbi = EXT4_SB(sb);
1659
1660 if (sbi->s_jquota_fmt) {
1661 char *fmtname = "";
1662
1663 switch (sbi->s_jquota_fmt) {
1664 case QFMT_VFS_OLD:
1665 fmtname = "vfsold";
1666 break;
1667 case QFMT_VFS_V0:
1668 fmtname = "vfsv0";
1669 break;
1670 case QFMT_VFS_V1:
1671 fmtname = "vfsv1";
1672 break;
1673 }
1674 seq_printf(seq, ",jqfmt=%s", fmtname);
1675 }
1676
1677 if (sbi->s_qf_names[USRQUOTA])
1678 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1679
1680 if (sbi->s_qf_names[GRPQUOTA])
1681 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1682
1683 if (test_opt(sb, USRQUOTA))
1684 seq_puts(seq, ",usrquota");
1685
1686 if (test_opt(sb, GRPQUOTA))
1687 seq_puts(seq, ",grpquota");
1688#endif
1689}
1690
5a916be1
TT		 1691static const char *token2str(int token)
1692{
1693 static const struct match_token *t;
1694
1695 for (t = tokens; t->token != Opt_err; t++)
1696 if (t->token == token && !strchr(t->pattern, '='))
1697 break;
1698 return t->pattern;
1699}
1700
2adf6da8
TT		 1701/*
1702 * Show an option if
1703 * - it's set to a non-default value OR
1704 * - if the per-sb default is different from the global default
1705 */
66acdcf4
TT		 1706static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1707 int nodefs)
2adf6da8 1708{
2adf6da8
TT		 1709 struct ext4_sb_info *sbi = EXT4_SB(sb);
1710 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1711 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1712 const struct mount_opts *m;
66acdcf4 1713 char sep = nodefs ? '\n' : ',';
2adf6da8 1714
66acdcf4
TT		 1715#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1716#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1717
1718 if (sbi->s_sb_block != 1)
5a916be1
TT		 1719 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1720
1721 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1722 int want_set = m->flags & MOPT_SET;
1723 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1724 (m->flags & MOPT_CLEAR_ERR))
1725 continue;
66acdcf4 1726 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1727 continue; /* skip if same as the default */
1728 if ((want_set &&
1729 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1730 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1731 continue; /* select Opt_noFoo vs Opt_Foo */
1732 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1733 }
5a916be1 1734
66acdcf4 1735 if (nodefs || sbi->s_resuid != EXT4_DEF_RESUID ||
5a916be1
TT		 1736 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1737 SEQ_OPTS_PRINT("resuid=%u", sbi->s_resuid);
66acdcf4 1738 if (nodefs || sbi->s_resgid != EXT4_DEF_RESGID ||
5a916be1
TT		 1739 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1740 SEQ_OPTS_PRINT("resgid=%u", sbi->s_resgid);
66acdcf4 1741 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1742 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1743 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1744 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1745 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1746 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1747 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1748 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1749 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1750 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1751 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1752 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1753 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1754 if (sb->s_flags & MS_I_VERSION)
5a916be1 1755 SEQ_OPTS_PUTS("i_version");
66acdcf4 1756 if (nodefs || sbi->s_stripe)
5a916be1 1757 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1758 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1759 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1760 SEQ_OPTS_PUTS("data=journal");
1761 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1762 SEQ_OPTS_PUTS("data=ordered");
1763 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1764 SEQ_OPTS_PUTS("data=writeback");
1765 }
66acdcf4
TT		 1766 if (nodefs ||
1767 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1768 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1769 sbi->s_inode_readahead_blks);
2adf6da8 1770
66acdcf4
TT		 1771 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1772 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1773 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2adf6da8
TT		 1774
1775 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1776 return 0;
1777}
1778
66acdcf4
TT		 1779static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1780{
1781 return _ext4_show_options(seq, root->d_sb, 0);
1782}
1783
1784static int options_seq_show(struct seq_file *seq, void *offset)
1785{
1786 struct super_block *sb = seq->private;
1787 int rc;
1788
1789 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1790 rc = _ext4_show_options(seq, sb, 1);
1791 seq_puts(seq, "\n");
1792 return rc;
1793}
1794
1795static int options_open_fs(struct inode *inode, struct file *file)
1796{
1797 return single_open(file, options_seq_show, PDE(inode)->data);
1798}
1799
1800static const struct file_operations ext4_seq_options_fops = {
1801 .owner = THIS_MODULE,
1802 .open = options_open_fs,
1803 .read = seq_read,
1804 .llseek = seq_lseek,
1805 .release = single_release,
1806};
1807
617ba13b 1808static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1809 int read_only)
1810{
617ba13b 1811 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1812 int res = 0;
1813
617ba13b 1814 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1815 ext4_msg(sb, KERN_ERR, "revision level too high, "
1816 "forcing read-only mode");
ac27a0ec
DK		 1817 res = MS_RDONLY;
1818 }
1819 if (read_only)
281b5995 1820 goto done;
617ba13b 1821 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1822 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1823 "running e2fsck is recommended");
617ba13b 1824 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1825 ext4_msg(sb, KERN_WARNING,
1826 "warning: mounting fs with errors, "
1827 "running e2fsck is recommended");
ed3ce80a 1828 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1829 le16_to_cpu(es->s_mnt_count) >=
1830 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1831 ext4_msg(sb, KERN_WARNING,
1832 "warning: maximal mount count reached, "
1833 "running e2fsck is recommended");
ac27a0ec
DK		 1834 else if (le32_to_cpu(es->s_checkinterval) &&
1835 (le32_to_cpu(es->s_lastcheck) +
1836 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1837 ext4_msg(sb, KERN_WARNING,
1838 "warning: checktime reached, "
1839 "running e2fsck is recommended");
0b8e58a1 1840 if (!sbi->s_journal)
0390131b 1841 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1842 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1843 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1844 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1845 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1846 ext4_update_dynamic_rev(sb);
0390131b
FM		 1847 if (sbi->s_journal)
1848 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1849
e2d67052 1850 ext4_commit_super(sb, 1);
281b5995 1851done:
ac27a0ec 1852 if (test_opt(sb, DEBUG))
a9df9a49 1853 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1854 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1855 sb->s_blocksize,
1856 sbi->s_groups_count,
617ba13b
MC		 1857 EXT4_BLOCKS_PER_GROUP(sb),
1858 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1859 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1860
7abc52c2 1861 cleancache_init_fs(sb);
ac27a0ec
DK		 1862 return res;
1863}
1864
772cb7c8
JS		 1865static int ext4_fill_flex_info(struct super_block *sb)
1866{
1867 struct ext4_sb_info *sbi = EXT4_SB(sb);
1868 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1869 ext4_group_t flex_group_count;
1870 ext4_group_t flex_group;
d50f2ab6 1871 unsigned int groups_per_flex = 0;
c5ca7c76 1872 size_t size;
772cb7c8
JS		 1873 int i;
1874
503358ae 1875 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1876 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1877 sbi->s_log_groups_per_flex = 0;
1878 return 1;
1879 }
d50f2ab6 1880 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
772cb7c8 1881
c62a11fd
FB		 1882 /* We allocate both existing and potentially added groups */
1883 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1884 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1885 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76 1886 size = flex_group_count * sizeof(struct flex_groups);
9933fc0a 1887 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
c5ca7c76 1888 if (sbi->s_flex_groups == NULL) {
9933fc0a
TT		 1889 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1890 flex_group_count);
1891 goto failed;
772cb7c8 1892 }
772cb7c8 1893
772cb7c8 1894 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1895 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1896
1897 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1898 atomic_add(ext4_free_inodes_count(sb, gdp),
1899 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 1900 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 1901 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1902 atomic_add(ext4_used_dirs_count(sb, gdp),
1903 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1904 }
1905
1906 return 1;
1907failed:
1908 return 0;
1909}
1910
717d50e4
AD		 1911__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1912 struct ext4_group_desc *gdp)
1913{
1914 __u16 crc = 0;
1915
1916 if (sbi->s_es->s_feature_ro_compat &
1917 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1918 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1919 __le32 le_group = cpu_to_le32(block_group);
1920
1921 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1922 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1923 crc = crc16(crc, (__u8 *)gdp, offset);
1924 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1925 /* for checksum of struct ext4_group_desc do the rest...*/
1926 if ((sbi->s_es->s_feature_incompat &
1927 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1928 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1929 crc = crc16(crc, (__u8 *)gdp + offset,
1930 le16_to_cpu(sbi->s_es->s_desc_size) -
1931 offset);
1932 }
1933
1934 return cpu_to_le16(crc);
1935}
1936
1937int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1938 struct ext4_group_desc *gdp)
1939{
1940 if ((sbi->s_es->s_feature_ro_compat &
1941 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1942 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1943 return 0;
1944
1945 return 1;
1946}
1947
ac27a0ec 1948/* Called at mount-time, super-block is locked */
bfff6873
LC		 1949static int ext4_check_descriptors(struct super_block *sb,
1950 ext4_group_t *first_not_zeroed)
ac27a0ec 1951{
617ba13b
MC		 1952 struct ext4_sb_info *sbi = EXT4_SB(sb);
1953 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1954 ext4_fsblk_t last_block;
bd81d8ee
LV		 1955 ext4_fsblk_t block_bitmap;
1956 ext4_fsblk_t inode_bitmap;
1957 ext4_fsblk_t inode_table;
ce421581 1958 int flexbg_flag = 0;
bfff6873 1959 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 1960
ce421581
JS		 1961 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1962 flexbg_flag = 1;
1963
af5bc92d 1964 ext4_debug("Checking group descriptors");
ac27a0ec 1965
197cd65a
AM		 1966 for (i = 0; i < sbi->s_groups_count; i++) {
1967 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1968
ce421581 1969 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1970 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1971 else
1972 last_block = first_block +
617ba13b 1973 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1974
bfff6873
LC		 1975 if ((grp == sbi->s_groups_count) &&
1976 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
1977 grp = i;
1978
8fadc143 1979 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1980 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 1981 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1982 "Block bitmap for group %u not in group "
b31e1552 1983 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1984 return 0;
1985 }
8fadc143 1986 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1987 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 1988 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1989 "Inode bitmap for group %u not in group "
b31e1552 1990 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1991 return 0;
1992 }
8fadc143 1993 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1994 if (inode_table < first_block ||
2b2d6d01 1995 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 1996 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1997 "Inode table for group %u not in group "
b31e1552 1998 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1999 return 0;
2000 }
955ce5f5 2001 ext4_lock_group(sb, i);
717d50e4 2002 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 2003 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2004 "Checksum for group %u failed (%u!=%u)",
2005 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2006 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2007 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2008 ext4_unlock_group(sb, i);
8a266467 2009 return 0;
7ee1ec4c 2010 }
717d50e4 2011 }
955ce5f5 2012 ext4_unlock_group(sb, i);
ce421581
JS		 2013 if (!flexbg_flag)
2014 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2015 }
bfff6873
LC		 2016 if (NULL != first_not_zeroed)
2017 *first_not_zeroed = grp;
ac27a0ec 2018
5dee5437
TT		 2019 ext4_free_blocks_count_set(sbi->s_es,
2020 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2021 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2022 return 1;
2023}
2024
617ba13b 2025/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2026 * the superblock) which were deleted from all directories, but held open by
2027 * a process at the time of a crash. We walk the list and try to delete these
2028 * inodes at recovery time (only with a read-write filesystem).
2029 *
2030 * In order to keep the orphan inode chain consistent during traversal (in
2031 * case of crash during recovery), we link each inode into the superblock
2032 * orphan list_head and handle it the same way as an inode deletion during
2033 * normal operation (which journals the operations for us).
2034 *
2035 * We only do an iget() and an iput() on each inode, which is very safe if we
2036 * accidentally point at an in-use or already deleted inode. The worst that
2037 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2038 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2039 * e2fsck was run on this filesystem, and it must have already done the orphan
2040 * inode cleanup for us, so we can safely abort without any further action.
2041 */
2b2d6d01
TT		 2042static void ext4_orphan_cleanup(struct super_block *sb,
2043 struct ext4_super_block *es)
ac27a0ec
DK		 2044{
2045 unsigned int s_flags = sb->s_flags;
2046 int nr_orphans = 0, nr_truncates = 0;
2047#ifdef CONFIG_QUOTA
2048 int i;
2049#endif
2050 if (!es->s_last_orphan) {
2051 jbd_debug(4, "no orphan inodes to clean up\n");
2052 return;
2053 }
2054
a8f48a95 2055 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2056 ext4_msg(sb, KERN_ERR, "write access "
2057 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2058 return;
2059 }
2060
d39195c3
AG		 2061 /* Check if feature set would not allow a r/w mount */
2062 if (!ext4_feature_set_ok(sb, 0)) {
2063 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2064 "unknown ROCOMPAT features");
2065 return;
2066 }
2067
617ba13b 2068 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2069 if (es->s_last_orphan)
2070 jbd_debug(1, "Errors on filesystem, "
2071 "clearing orphan list.\n");
2072 es->s_last_orphan = 0;
2073 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2074 return;
2075 }
2076
2077 if (s_flags & MS_RDONLY) {
b31e1552 2078 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2079 sb->s_flags &= ~MS_RDONLY;
2080 }
2081#ifdef CONFIG_QUOTA
2082 /* Needed for iput() to work correctly and not trash data */
2083 sb->s_flags |= MS_ACTIVE;
2084 /* Turn on quotas so that they are updated correctly */
2085 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2086 if (EXT4_SB(sb)->s_qf_names[i]) {
2087 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2088 if (ret < 0)
b31e1552
ES		 2089 ext4_msg(sb, KERN_ERR,
2090 "Cannot turn on journaled "
2091 "quota: error %d", ret);
ac27a0ec
DK		 2092 }
2093 }
2094#endif
2095
2096 while (es->s_last_orphan) {
2097 struct inode *inode;
2098
97bd42b9
JB		 2099 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2100 if (IS_ERR(inode)) {
ac27a0ec
DK		 2101 es->s_last_orphan = 0;
2102 break;
2103 }
2104
617ba13b 2105 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2106 dquot_initialize(inode);
ac27a0ec 2107 if (inode->i_nlink) {
b31e1552
ES		 2108 ext4_msg(sb, KERN_DEBUG,
2109 "%s: truncating inode %lu to %lld bytes",
46e665e9 2110 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2111 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2112 inode->i_ino, inode->i_size);
617ba13b 2113 ext4_truncate(inode);
ac27a0ec
DK		 2114 nr_truncates++;
2115 } else {
b31e1552
ES		 2116 ext4_msg(sb, KERN_DEBUG,
2117 "%s: deleting unreferenced inode %lu",
46e665e9 2118 __func__, inode->i_ino);
ac27a0ec
DK		 2119 jbd_debug(2, "deleting unreferenced inode %lu\n",
2120 inode->i_ino);
2121 nr_orphans++;
2122 }
2123 iput(inode); /* The delete magic happens here! */
2124 }
2125
2b2d6d01 2126#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2127
2128 if (nr_orphans)
b31e1552
ES		 2129 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2130 PLURAL(nr_orphans));
ac27a0ec 2131 if (nr_truncates)
b31e1552
ES		 2132 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2133 PLURAL(nr_truncates));
ac27a0ec
DK		 2134#ifdef CONFIG_QUOTA
2135 /* Turn quotas off */
2136 for (i = 0; i < MAXQUOTAS; i++) {
2137 if (sb_dqopt(sb)->files[i])
287a8095 2138 dquot_quota_off(sb, i);
ac27a0ec
DK		 2139 }
2140#endif
2141 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2142}
0b8e58a1 2143
cd2291a4
ES		 2144/*
2145 * Maximal extent format file size.
2146 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2147 * extent format containers, within a sector_t, and within i_blocks
2148 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2149 * so that won't be a limiting factor.
2150 *
f17722f9
LC		 2151 * However there is other limiting factor. We do store extents in the form
2152 * of starting block and length, hence the resulting length of the extent
2153 * covering maximum file size must fit into on-disk format containers as
2154 * well. Given that length is always by 1 unit bigger than max unit (because
2155 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2156 *
cd2291a4
ES		 2157 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2158 */
f287a1a5 2159static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2160{
2161 loff_t res;
2162 loff_t upper_limit = MAX_LFS_FILESIZE;
2163
2164 /* small i_blocks in vfs inode? */
f287a1a5 2165 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2166 /*
90c699a9 2167 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2168 * i_block represent total blocks in 512 bytes
2169 * 32 == size of vfs inode i_blocks * 8
2170 */
2171 upper_limit = (1LL << 32) - 1;
2172
2173 /* total blocks in file system block size */
2174 upper_limit >>= (blkbits - 9);
2175 upper_limit <<= blkbits;
2176 }
2177
f17722f9
LC		 2178 /*
2179 * 32-bit extent-start container, ee_block. We lower the maxbytes
2180 * by one fs block, so ee_len can cover the extent of maximum file
2181 * size
2182 */
2183 res = (1LL << 32) - 1;
cd2291a4 2184 res <<= blkbits;
cd2291a4
ES		 2185
2186 /* Sanity check against vm- & vfs- imposed limits */
2187 if (res > upper_limit)
2188 res = upper_limit;
2189
2190 return res;
2191}
ac27a0ec 2192
ac27a0ec 2193/*
cd2291a4 2194 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2195 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2196 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2197 */
f287a1a5 2198static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2199{
617ba13b 2200 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2201 int meta_blocks;
2202 loff_t upper_limit;
0b8e58a1
AD		 2203 /* This is calculated to be the largest file size for a dense, block
2204 * mapped file such that the file's total number of 512-byte sectors,
2205 * including data and all indirect blocks, does not exceed (2^48 - 1).
2206 *
2207 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2208 * number of 512-byte sectors of the file.
0fc1b451
AK		 2209 */
2210
f287a1a5 2211 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2212 /*
90c699a9 2213 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2214 * the inode i_block field represents total file blocks in
2215 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2216 */
2217 upper_limit = (1LL << 32) - 1;
2218
2219 /* total blocks in file system block size */
2220 upper_limit >>= (bits - 9);
2221
2222 } else {
8180a562
AK		 2223 /*
2224 * We use 48 bit ext4_inode i_blocks
2225 * With EXT4_HUGE_FILE_FL set the i_blocks
2226 * represent total number of blocks in
2227 * file system block size
2228 */
0fc1b451
AK		 2229 upper_limit = (1LL << 48) - 1;
2230
0fc1b451
AK		 2231 }
2232
2233 /* indirect blocks */
2234 meta_blocks = 1;
2235 /* double indirect blocks */
2236 meta_blocks += 1 + (1LL << (bits-2));
2237 /* tripple indirect blocks */
2238 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2239
2240 upper_limit -= meta_blocks;
2241 upper_limit <<= bits;
ac27a0ec
DK		 2242
2243 res += 1LL << (bits-2);
2244 res += 1LL << (2*(bits-2));
2245 res += 1LL << (3*(bits-2));
2246 res <<= bits;
2247 if (res > upper_limit)
2248 res = upper_limit;
0fc1b451
AK		 2249
2250 if (res > MAX_LFS_FILESIZE)
2251 res = MAX_LFS_FILESIZE;
2252
ac27a0ec
DK		 2253 return res;
2254}
2255
617ba13b 2256static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2257 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2258{
617ba13b 2259 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2260 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2261 int has_super = 0;
2262
2263 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2264
617ba13b 2265 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2266 nr < first_meta_bg)
70bbb3e0 2267 return logical_sb_block + nr + 1;
ac27a0ec 2268 bg = sbi->s_desc_per_block * nr;
617ba13b 2269 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2270 has_super = 1;
0b8e58a1 2271
617ba13b 2272 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2273}
2274
c9de560d
AT		 2275/**
2276 * ext4_get_stripe_size: Get the stripe size.
2277 * @sbi: In memory super block info
2278 *
2279 * If we have specified it via mount option, then
2280 * use the mount option value. If the value specified at mount time is
2281 * greater than the blocks per group use the super block value.
2282 * If the super block value is greater than blocks per group return 0.
2283 * Allocator needs it be less than blocks per group.
2284 *
2285 */
2286static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2287{
2288 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2289 unsigned long stripe_width =
2290 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2291 int ret;
c9de560d
AT		 2292
2293 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2294 ret = sbi->s_stripe;
2295 else if (stripe_width <= sbi->s_blocks_per_group)
2296 ret = stripe_width;
2297 else if (stride <= sbi->s_blocks_per_group)
2298 ret = stride;
2299 else
2300 ret = 0;
c9de560d 2301
3eb08658
DE		 2302 /*
2303 * If the stripe width is 1, this makes no sense and
2304 * we set it to 0 to turn off stripe handling code.
2305 */
2306 if (ret <= 1)
2307 ret = 0;
c9de560d 2308
3eb08658 2309 return ret;
c9de560d 2310}
ac27a0ec 2311
3197ebdb
TT		 2312/* sysfs supprt */
2313
2314struct ext4_attr {
2315 struct attribute attr;
2316 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2317 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2318 const char *, size_t);
2319 int offset;
2320};
2321
2322static int parse_strtoul(const char *buf,
2323 unsigned long max, unsigned long *value)
2324{
2325 char *endp;
2326
e7d2860b
AGR		 2327 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2328 endp = skip_spaces(endp);
3197ebdb
TT		 2329 if (*endp || *value > max)
2330 return -EINVAL;
2331
2332 return 0;
2333}
2334
2335static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2336 struct ext4_sb_info *sbi,
2337 char *buf)
2338{
2339 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2340 (s64) EXT4_C2B(sbi,
2341 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2342}
2343
2344static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2345 struct ext4_sb_info *sbi, char *buf)
2346{
2347 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2348
f613dfcb
TT		 2349 if (!sb->s_bdev->bd_part)
2350 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2351 return snprintf(buf, PAGE_SIZE, "%lu\n",
2352 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2353 sbi->s_sectors_written_start) >> 1);
2354}
2355
2356static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2357 struct ext4_sb_info *sbi, char *buf)
2358{
2359 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2360
f613dfcb
TT		 2361 if (!sb->s_bdev->bd_part)
2362 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2363 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2364 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2365 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2366 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2367}
2368
77f4135f
VH		 2369static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2370 struct ext4_sb_info *sbi, char *buf)
2371{
2372 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2373}
2374
2375static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2376 struct ext4_sb_info *sbi, char *buf)
2377{
2378 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2379}
2380
3197ebdb
TT		 2381static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2382 struct ext4_sb_info *sbi,
2383 const char *buf, size_t count)
2384{
2385 unsigned long t;
2386
2387 if (parse_strtoul(buf, 0x40000000, &t))
2388 return -EINVAL;
2389
5dbd571d 2390 if (t && !is_power_of_2(t))
3197ebdb
TT		 2391 return -EINVAL;
2392
2393 sbi->s_inode_readahead_blks = t;
2394 return count;
2395}
2396
2397static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2398 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2399{
2400 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2401
2402 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2403}
2404
2405static ssize_t sbi_ui_store(struct ext4_attr *a,
2406 struct ext4_sb_info *sbi,
2407 const char *buf, size_t count)
2408{
2409 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2410 unsigned long t;
2411
2412 if (parse_strtoul(buf, 0xffffffff, &t))
2413 return -EINVAL;
2414 *ui = t;
2415 return count;
2416}
2417
2418#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2419static struct ext4_attr ext4_attr_##_name = { \
2420 .attr = {.name = __stringify(_name), .mode = _mode }, \
2421 .show = _show, \
2422 .store = _store, \
2423 .offset = offsetof(struct ext4_sb_info, _elname), \
2424}
2425#define EXT4_ATTR(name, mode, show, store) \
2426static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2427
857ac889 2428#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2429#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2430#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2431#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2432 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2433#define ATTR_LIST(name) &ext4_attr_##name.attr
2434
2435EXT4_RO_ATTR(delayed_allocation_blocks);
2436EXT4_RO_ATTR(session_write_kbytes);
2437EXT4_RO_ATTR(lifetime_write_kbytes);
77f4135f
VH		 2438EXT4_RO_ATTR(extent_cache_hits);
2439EXT4_RO_ATTR(extent_cache_misses);
3197ebdb
TT		 2440EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2441 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2442EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2443EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2444EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2445EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2446EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2447EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2448EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2449EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2450
2451static struct attribute *ext4_attrs[] = {
2452 ATTR_LIST(delayed_allocation_blocks),
2453 ATTR_LIST(session_write_kbytes),
2454 ATTR_LIST(lifetime_write_kbytes),
77f4135f
VH		 2455 ATTR_LIST(extent_cache_hits),
2456 ATTR_LIST(extent_cache_misses),
3197ebdb 2457 ATTR_LIST(inode_readahead_blks),
11013911 2458 ATTR_LIST(inode_goal),
3197ebdb
TT		 2459 ATTR_LIST(mb_stats),
2460 ATTR_LIST(mb_max_to_scan),
2461 ATTR_LIST(mb_min_to_scan),
2462 ATTR_LIST(mb_order2_req),
2463 ATTR_LIST(mb_stream_req),
2464 ATTR_LIST(mb_group_prealloc),
55138e0b 2465 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2466 NULL,
2467};
2468
857ac889
LC		 2469/* Features this copy of ext4 supports */
2470EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2471EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2472
2473static struct attribute *ext4_feat_attrs[] = {
2474 ATTR_LIST(lazy_itable_init),
27ee40df 2475 ATTR_LIST(batched_discard),
857ac889
LC		 2476 NULL,
2477};
2478
3197ebdb
TT		 2479static ssize_t ext4_attr_show(struct kobject *kobj,
2480 struct attribute *attr, char *buf)
2481{
2482 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2483 s_kobj);
2484 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2485
2486 return a->show ? a->show(a, sbi, buf) : 0;
2487}
2488
2489static ssize_t ext4_attr_store(struct kobject *kobj,
2490 struct attribute *attr,
2491 const char *buf, size_t len)
2492{
2493 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2494 s_kobj);
2495 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2496
2497 return a->store ? a->store(a, sbi, buf, len) : 0;
2498}
2499
2500static void ext4_sb_release(struct kobject *kobj)
2501{
2502 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2503 s_kobj);
2504 complete(&sbi->s_kobj_unregister);
2505}
2506
52cf25d0 2507static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2508 .show = ext4_attr_show,
2509 .store = ext4_attr_store,
2510};
2511
2512static struct kobj_type ext4_ktype = {
2513 .default_attrs = ext4_attrs,
2514 .sysfs_ops = &ext4_attr_ops,
2515 .release = ext4_sb_release,
2516};
2517
857ac889
LC		 2518static void ext4_feat_release(struct kobject *kobj)
2519{
2520 complete(&ext4_feat->f_kobj_unregister);
2521}
2522
2523static struct kobj_type ext4_feat_ktype = {
2524 .default_attrs = ext4_feat_attrs,
2525 .sysfs_ops = &ext4_attr_ops,
2526 .release = ext4_feat_release,
2527};
2528
a13fb1a4
ES		 2529/*
2530 * Check whether this filesystem can be mounted based on
2531 * the features present and the RDONLY/RDWR mount requested.
2532 * Returns 1 if this filesystem can be mounted as requested,
2533 * 0 if it cannot be.
2534 */
2535static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2536{
2537 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2538 ext4_msg(sb, KERN_ERR,
2539 "Couldn't mount because of "
2540 "unsupported optional features (%x)",
2541 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2542 ~EXT4_FEATURE_INCOMPAT_SUPP));
2543 return 0;
2544 }
2545
2546 if (readonly)
2547 return 1;
2548
2549 /* Check that feature set is OK for a read-write mount */
2550 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2551 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2552 "unsupported optional features (%x)",
2553 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2554 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2555 return 0;
2556 }
2557 /*
2558 * Large file size enabled file system can only be mounted
2559 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2560 */
2561 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2562 if (sizeof(blkcnt_t) < sizeof(u64)) {
2563 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2564 "cannot be mounted RDWR without "
2565 "CONFIG_LBDAF");
2566 return 0;
2567 }
2568 }
bab08ab9
TT		 2569 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2570 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2571 ext4_msg(sb, KERN_ERR,
2572 "Can't support bigalloc feature without "
2573 "extents feature\n");
2574 return 0;
2575 }
a13fb1a4
ES		 2576 return 1;
2577}
2578
66e61a9e
TT		 2579/*
2580 * This function is called once a day if we have errors logged
2581 * on the file system
2582 */
2583static void print_daily_error_info(unsigned long arg)
2584{
2585 struct super_block *sb = (struct super_block *) arg;
2586 struct ext4_sb_info *sbi;
2587 struct ext4_super_block *es;
2588
2589 sbi = EXT4_SB(sb);
2590 es = sbi->s_es;
2591
2592 if (es->s_error_count)
2593 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2594 le32_to_cpu(es->s_error_count));
2595 if (es->s_first_error_time) {
2596 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2597 sb->s_id, le32_to_cpu(es->s_first_error_time),
2598 (int) sizeof(es->s_first_error_func),
2599 es->s_first_error_func,
2600 le32_to_cpu(es->s_first_error_line));
2601 if (es->s_first_error_ino)
2602 printk(": inode %u",
2603 le32_to_cpu(es->s_first_error_ino));
2604 if (es->s_first_error_block)
2605 printk(": block %llu", (unsigned long long)
2606 le64_to_cpu(es->s_first_error_block));
2607 printk("\n");
2608 }
2609 if (es->s_last_error_time) {
2610 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2611 sb->s_id, le32_to_cpu(es->s_last_error_time),
2612 (int) sizeof(es->s_last_error_func),
2613 es->s_last_error_func,
2614 le32_to_cpu(es->s_last_error_line));
2615 if (es->s_last_error_ino)
2616 printk(": inode %u",
2617 le32_to_cpu(es->s_last_error_ino));
2618 if (es->s_last_error_block)
2619 printk(": block %llu", (unsigned long long)
2620 le64_to_cpu(es->s_last_error_block));
2621 printk("\n");
2622 }
2623 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2624}
2625
bfff6873
LC		 2626/* Find next suitable group and run ext4_init_inode_table */
2627static int ext4_run_li_request(struct ext4_li_request *elr)
2628{
2629 struct ext4_group_desc *gdp = NULL;
2630 ext4_group_t group, ngroups;
2631 struct super_block *sb;
2632 unsigned long timeout = 0;
2633 int ret = 0;
2634
2635 sb = elr->lr_super;
2636 ngroups = EXT4_SB(sb)->s_groups_count;
2637
2638 for (group = elr->lr_next_group; group < ngroups; group++) {
2639 gdp = ext4_get_group_desc(sb, group, NULL);
2640 if (!gdp) {
2641 ret = 1;
2642 break;
2643 }
2644
2645 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2646 break;
2647 }
2648
2649 if (group == ngroups)
2650 ret = 1;
2651
2652 if (!ret) {
2653 timeout = jiffies;
2654 ret = ext4_init_inode_table(sb, group,
2655 elr->lr_timeout ? 0 : 1);
2656 if (elr->lr_timeout == 0) {
51ce6511
LC		 2657 timeout = (jiffies - timeout) *
2658 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2659 elr->lr_timeout = timeout;
2660 }
2661 elr->lr_next_sched = jiffies + elr->lr_timeout;
2662 elr->lr_next_group = group + 1;
2663 }
2664
2665 return ret;
2666}
2667
2668/*
2669 * Remove lr_request from the list_request and free the
4ed5c033 2670 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2671 */
2672static void ext4_remove_li_request(struct ext4_li_request *elr)
2673{
2674 struct ext4_sb_info *sbi;
2675
2676 if (!elr)
2677 return;
2678
2679 sbi = elr->lr_sbi;
2680
2681 list_del(&elr->lr_request);
2682 sbi->s_li_request = NULL;
2683 kfree(elr);
2684}
2685
2686static void ext4_unregister_li_request(struct super_block *sb)
2687{
1bb933fb
LC		 2688 mutex_lock(&ext4_li_mtx);
2689 if (!ext4_li_info) {
2690 mutex_unlock(&ext4_li_mtx);
bfff6873 2691 return;
1bb933fb 2692 }
bfff6873
LC		 2693
2694 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2695 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2696 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2697 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2698}
2699
8f1f7453
ES		 2700static struct task_struct *ext4_lazyinit_task;
2701
bfff6873
LC		 2702/*
2703 * This is the function where ext4lazyinit thread lives. It walks
2704 * through the request list searching for next scheduled filesystem.
2705 * When such a fs is found, run the lazy initialization request
2706 * (ext4_rn_li_request) and keep track of the time spend in this
2707 * function. Based on that time we compute next schedule time of
2708 * the request. When walking through the list is complete, compute
2709 * next waking time and put itself into sleep.
2710 */
2711static int ext4_lazyinit_thread(void *arg)
2712{
2713 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2714 struct list_head *pos, *n;
2715 struct ext4_li_request *elr;
4ed5c033 2716 unsigned long next_wakeup, cur;
bfff6873
LC		 2717
2718 BUG_ON(NULL == eli);
2719
bfff6873
LC		 2720cont_thread:
2721 while (true) {
2722 next_wakeup = MAX_JIFFY_OFFSET;
2723
2724 mutex_lock(&eli->li_list_mtx);
2725 if (list_empty(&eli->li_request_list)) {
2726 mutex_unlock(&eli->li_list_mtx);
2727 goto exit_thread;
2728 }
2729
2730 list_for_each_safe(pos, n, &eli->li_request_list) {
2731 elr = list_entry(pos, struct ext4_li_request,
2732 lr_request);
2733
b2c78cd0
TT		 2734 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2735 if (ext4_run_li_request(elr) != 0) {
2736 /* error, remove the lazy_init job */
2737 ext4_remove_li_request(elr);
2738 continue;
2739 }
bfff6873
LC		 2740 }
2741
2742 if (time_before(elr->lr_next_sched, next_wakeup))
2743 next_wakeup = elr->lr_next_sched;
2744 }
2745 mutex_unlock(&eli->li_list_mtx);
2746
a0acae0e 2747 try_to_freeze();
bfff6873 2748
4ed5c033
LC		 2749 cur = jiffies;
2750 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2751 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2752 cond_resched();
2753 continue;
2754 }
2755
4ed5c033
LC		 2756 schedule_timeout_interruptible(next_wakeup - cur);
2757
8f1f7453
ES		 2758 if (kthread_should_stop()) {
2759 ext4_clear_request_list();
2760 goto exit_thread;
2761 }
bfff6873
LC		 2762 }
2763
2764exit_thread:
2765 /*
2766 * It looks like the request list is empty, but we need
2767 * to check it under the li_list_mtx lock, to prevent any
2768 * additions into it, and of course we should lock ext4_li_mtx
2769 * to atomically free the list and ext4_li_info, because at
2770 * this point another ext4 filesystem could be registering
2771 * new one.
2772 */
2773 mutex_lock(&ext4_li_mtx);
2774 mutex_lock(&eli->li_list_mtx);
2775 if (!list_empty(&eli->li_request_list)) {
2776 mutex_unlock(&eli->li_list_mtx);
2777 mutex_unlock(&ext4_li_mtx);
2778 goto cont_thread;
2779 }
2780 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2781 kfree(ext4_li_info);
2782 ext4_li_info = NULL;
2783 mutex_unlock(&ext4_li_mtx);
2784
2785 return 0;
2786}
2787
2788static void ext4_clear_request_list(void)
2789{
2790 struct list_head *pos, *n;
2791 struct ext4_li_request *elr;
2792
2793 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2794 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2795 elr = list_entry(pos, struct ext4_li_request,
2796 lr_request);
2797 ext4_remove_li_request(elr);
2798 }
2799 mutex_unlock(&ext4_li_info->li_list_mtx);
2800}
2801
2802static int ext4_run_lazyinit_thread(void)
2803{
8f1f7453
ES		 2804 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2805 ext4_li_info, "ext4lazyinit");
2806 if (IS_ERR(ext4_lazyinit_task)) {
2807 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2808 ext4_clear_request_list();
bfff6873
LC		 2809 kfree(ext4_li_info);
2810 ext4_li_info = NULL;
92b97816 2811 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2812 "initialization thread\n",
2813 err);
2814 return err;
2815 }
2816 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2817 return 0;
2818}
2819
2820/*
2821 * Check whether it make sense to run itable init. thread or not.
2822 * If there is at least one uninitialized inode table, return
2823 * corresponding group number, else the loop goes through all
2824 * groups and return total number of groups.
2825 */
2826static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2827{
2828 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2829 struct ext4_group_desc *gdp = NULL;
2830
2831 for (group = 0; group < ngroups; group++) {
2832 gdp = ext4_get_group_desc(sb, group, NULL);
2833 if (!gdp)
2834 continue;
2835
2836 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2837 break;
2838 }
2839
2840 return group;
2841}
2842
2843static int ext4_li_info_new(void)
2844{
2845 struct ext4_lazy_init *eli = NULL;
2846
2847 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2848 if (!eli)
2849 return -ENOMEM;
2850
bfff6873
LC		 2851 INIT_LIST_HEAD(&eli->li_request_list);
2852 mutex_init(&eli->li_list_mtx);
2853
bfff6873
LC		 2854 eli->li_state |= EXT4_LAZYINIT_QUIT;
2855
2856 ext4_li_info = eli;
2857
2858 return 0;
2859}
2860
2861static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2862 ext4_group_t start)
2863{
2864 struct ext4_sb_info *sbi = EXT4_SB(sb);
2865 struct ext4_li_request *elr;
2866 unsigned long rnd;
2867
2868 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2869 if (!elr)
2870 return NULL;
2871
2872 elr->lr_super = sb;
2873 elr->lr_sbi = sbi;
2874 elr->lr_next_group = start;
2875
2876 /*
2877 * Randomize first schedule time of the request to
2878 * spread the inode table initialization requests
2879 * better.
2880 */
2881 get_random_bytes(&rnd, sizeof(rnd));
2882 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2883 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2884
2885 return elr;
2886}
2887
2888static int ext4_register_li_request(struct super_block *sb,
2889 ext4_group_t first_not_zeroed)
2890{
2891 struct ext4_sb_info *sbi = EXT4_SB(sb);
2892 struct ext4_li_request *elr;
2893 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2894 int ret = 0;
bfff6873 2895
51ce6511
LC		 2896 if (sbi->s_li_request != NULL) {
2897 /*
2898 * Reset timeout so it can be computed again, because
2899 * s_li_wait_mult might have changed.
2900 */
2901 sbi->s_li_request->lr_timeout = 0;
beed5ecb 2902 return 0;
51ce6511 2903 }
bfff6873
LC		 2904
2905 if (first_not_zeroed == ngroups ||
2906 (sb->s_flags & MS_RDONLY) ||
55ff3840 2907 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 2908 return 0;
bfff6873
LC		 2909
2910 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 2911 if (!elr)
2912 return -ENOMEM;
bfff6873
LC		 2913
2914 mutex_lock(&ext4_li_mtx);
2915
2916 if (NULL == ext4_li_info) {
2917 ret = ext4_li_info_new();
2918 if (ret)
2919 goto out;
2920 }
2921
2922 mutex_lock(&ext4_li_info->li_list_mtx);
2923 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2924 mutex_unlock(&ext4_li_info->li_list_mtx);
2925
2926 sbi->s_li_request = elr;
46e4690b
TM		 2927 /*
2928 * set elr to NULL here since it has been inserted to
2929 * the request_list and the removal and free of it is
2930 * handled by ext4_clear_request_list from now on.
2931 */
2932 elr = NULL;
bfff6873
LC		 2933
2934 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2935 ret = ext4_run_lazyinit_thread();
2936 if (ret)
2937 goto out;
2938 }
bfff6873 2939out:
beed5ecb
NK		 2940 mutex_unlock(&ext4_li_mtx);
2941 if (ret)
bfff6873 2942 kfree(elr);
bfff6873
LC		 2943 return ret;
2944}
2945
2946/*
2947 * We do not need to lock anything since this is called on
2948 * module unload.
2949 */
2950static void ext4_destroy_lazyinit_thread(void)
2951{
2952 /*
2953 * If thread exited earlier
2954 * there's nothing to be done.
2955 */
8f1f7453 2956 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 2957 return;
2958
8f1f7453 2959 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 2960}
2961
2b2d6d01 2962static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 2963{
d4c402d9 2964 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 2965 struct buffer_head *bh;
617ba13b
MC		 2966 struct ext4_super_block *es = NULL;
2967 struct ext4_sb_info *sbi;
2968 ext4_fsblk_t block;
2969 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2970 ext4_fsblk_t logical_sb_block;
ac27a0ec 2971 unsigned long offset = 0;
ac27a0ec
DK		 2972 unsigned long journal_devnum = 0;
2973 unsigned long def_mount_opts;
2974 struct inode *root;
9f6200bb 2975 char *cp;
0390131b 2976 const char *descr;
dcc7dae3 2977 int ret = -ENOMEM;
281b5995 2978 int blocksize, clustersize;
4ec11028
TT		 2979 unsigned int db_count;
2980 unsigned int i;
281b5995 2981 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 2982 __u64 blocks_count;
833f4077 2983 int err;
b3881f74 2984 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 2985 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 2986
2987 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2988 if (!sbi)
dcc7dae3 2989 goto out_free_orig;
705895b6
PE		 2990
2991 sbi->s_blockgroup_lock =
2992 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2993 if (!sbi->s_blockgroup_lock) {
2994 kfree(sbi);
dcc7dae3 2995 goto out_free_orig;
705895b6 2996 }
ac27a0ec
DK		 2997 sb->s_fs_info = sbi;
2998 sbi->s_mount_opt = 0;
617ba13b
MC		 2999 sbi->s_resuid = EXT4_DEF_RESUID;
3000 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 3001 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3002 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3003 if (sb->s_bdev->bd_part)
3004 sbi->s_sectors_written_start =
3005 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3006
9f6200bb
TT		 3007 /* Cleanup superblock name */
3008 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3009 *cp = '!';
3010
dcc7dae3 3011 ret = -EINVAL;
617ba13b 3012 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3013 if (!blocksize) {
b31e1552 3014 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3015 goto out_fail;
3016 }
3017
3018 /*
617ba13b 3019 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3020 * block sizes. We need to calculate the offset from buffer start.
3021 */
617ba13b 3022 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3023 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3024 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3025 } else {
70bbb3e0 3026 logical_sb_block = sb_block;
ac27a0ec
DK		 3027 }
3028
70bbb3e0 3029 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3030 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3031 goto out_fail;
3032 }
3033 /*
3034 * Note: s_es must be initialized as soon as possible because
617ba13b 3035 * some ext4 macro-instructions depend on its value
ac27a0ec 3036 */
617ba13b 3037 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 3038 sbi->s_es = es;
3039 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3040 if (sb->s_magic != EXT4_SUPER_MAGIC)
3041 goto cantfind_ext4;
afc32f7e 3042 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 3043
3044 /* Set defaults before we parse the mount options */
3045 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3046 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3047 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3048 set_opt(sb, DEBUG);
87f26807 3049 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3050 set_opt(sb, GRPID);
617ba13b 3051 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3052 set_opt(sb, NO_UID32);
ea663336 3053 /* xattr user namespace & acls are now defaulted on */
03010a33 3054#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3055 set_opt(sb, XATTR_USER);
2e7842b8 3056#endif
03010a33 3057#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3058 set_opt(sb, POSIX_ACL);
2e7842b8 3059#endif
6fd7a467 3060 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3061 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3062 set_opt(sb, JOURNAL_DATA);
617ba13b 3063 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3064 set_opt(sb, ORDERED_DATA);
617ba13b 3065 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3066 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3067
3068 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3069 set_opt(sb, ERRORS_PANIC);
bb4f397a 3070 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3071 set_opt(sb, ERRORS_CONT);
bb4f397a 3072 else
fd8c37ec 3073 set_opt(sb, ERRORS_RO);
8b67f04a 3074 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3075 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3076 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3077 set_opt(sb, DISCARD);
ac27a0ec
DK		 3078
3079 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3080 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 3081 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3082 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3083 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3084
8b67f04a 3085 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3086 set_opt(sb, BARRIER);
ac27a0ec 3087
dd919b98
AK		 3088 /*
3089 * enable delayed allocation by default
3090 * Use -o nodelalloc to turn it off
3091 */
8b67f04a
TT		 3092 if (!IS_EXT3_SB(sb) &&
3093 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3094 set_opt(sb, DELALLOC);
dd919b98 3095
51ce6511
LC		 3096 /*
3097 * set default s_li_wait_mult for lazyinit, for the case there is
3098 * no mount option specified.
3099 */
3100 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3101
8b67f04a 3102 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3103 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3104 ext4_msg(sb, KERN_WARNING,
3105 "failed to parse options in superblock: %s",
3106 sbi->s_es->s_mount_opts);
3107 }
5a916be1 3108 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3109 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3110 &journal_ioprio, 0))
ac27a0ec
DK		 3111 goto failed_mount;
3112
56889787
TT		 3113 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3114 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3115 "with data=journal disables delayed "
3116 "allocation and O_DIRECT support!\n");
3117 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3118 ext4_msg(sb, KERN_ERR, "can't mount with "
3119 "both data=journal and delalloc");
3120 goto failed_mount;
3121 }
3122 if (test_opt(sb, DIOREAD_NOLOCK)) {
3123 ext4_msg(sb, KERN_ERR, "can't mount with "
3124 "both data=journal and delalloc");
3125 goto failed_mount;
3126 }
3127 if (test_opt(sb, DELALLOC))
3128 clear_opt(sb, DELALLOC);
3129 }
3130
3131 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3132 if (test_opt(sb, DIOREAD_NOLOCK)) {
3133 if (blocksize < PAGE_SIZE) {
3134 ext4_msg(sb, KERN_ERR, "can't mount with "
3135 "dioread_nolock if block size != PAGE_SIZE");
3136 goto failed_mount;
3137 }
3138 }
3139
ac27a0ec 3140 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3141 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3142
617ba13b
MC		 3143 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3144 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3145 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3146 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3147 ext4_msg(sb, KERN_WARNING,
3148 "feature flags set on rev 0 fs, "
3149 "running e2fsck is recommended");
469108ff 3150
2035e776
TT		 3151 if (IS_EXT2_SB(sb)) {
3152 if (ext2_feature_set_ok(sb))
3153 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3154 "using the ext4 subsystem");
3155 else {
3156 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3157 "to feature incompatibilities");
3158 goto failed_mount;
3159 }
3160 }
3161
3162 if (IS_EXT3_SB(sb)) {
3163 if (ext3_feature_set_ok(sb))
3164 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3165 "using the ext4 subsystem");
3166 else {
3167 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3168 "to feature incompatibilities");
3169 goto failed_mount;
3170 }
3171 }
3172
ac27a0ec
DK		 3173 /*
3174 * Check feature flags regardless of the revision level, since we
3175 * previously didn't change the revision level when setting the flags,
3176 * so there is a chance incompat flags are set on a rev 0 filesystem.
3177 */
a13fb1a4 3178 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3179 goto failed_mount;
a13fb1a4 3180
617ba13b
MC		 3181 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3182 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3183 ext4_msg(sb, KERN_ERR,
3184 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3185 goto failed_mount;
3186 }
3187
ac27a0ec 3188 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3189 /* Validate the filesystem blocksize */
3190 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3191 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3192 blocksize);
ac27a0ec
DK		 3193 goto failed_mount;
3194 }
3195
2b2d6d01 3196 brelse(bh);
70bbb3e0
AM		 3197 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3198 offset = do_div(logical_sb_block, blocksize);
3199 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3200 if (!bh) {
b31e1552
ES		 3201 ext4_msg(sb, KERN_ERR,
3202 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3203 goto failed_mount;
3204 }
617ba13b 3205 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 3206 sbi->s_es = es;
617ba13b 3207 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3208 ext4_msg(sb, KERN_ERR,
3209 "Magic mismatch, very weird!");
ac27a0ec
DK		 3210 goto failed_mount;
3211 }
3212 }
3213
a13fb1a4
ES		 3214 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3215 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3216 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3217 has_huge_files);
3218 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3219
617ba13b
MC		 3220 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3221 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3222 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3223 } else {
3224 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3225 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3226 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3227 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3228 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3229 ext4_msg(sb, KERN_ERR,
3230 "unsupported inode size: %d",
2b2d6d01 3231 sbi->s_inode_size);
ac27a0ec
DK		 3232 goto failed_mount;
3233 }
ef7f3835
KS		 3234 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3235 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3236 }
0b8e58a1 3237
0d1ee42f
AR		 3238 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3239 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3240 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3241 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3242 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3243 ext4_msg(sb, KERN_ERR,
3244 "unsupported descriptor size %lu",
0d1ee42f
AR		 3245 sbi->s_desc_size);
3246 goto failed_mount;
3247 }
3248 } else
3249 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3250
ac27a0ec 3251 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3252 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3253 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3254 goto cantfind_ext4;
0b8e58a1 3255
617ba13b 3256 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3257 if (sbi->s_inodes_per_block == 0)
617ba13b 3258 goto cantfind_ext4;
ac27a0ec
DK		 3259 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3260 sbi->s_inodes_per_block;
0d1ee42f 3261 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3262 sbi->s_sbh = bh;
3263 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3264 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3265 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3266
2b2d6d01 3267 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3268 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3269 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3270 i = le32_to_cpu(es->s_flags);
3271 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3272 sbi->s_hash_unsigned = 3;
3273 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3274#ifdef __CHAR_UNSIGNED__
3275 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3276 sbi->s_hash_unsigned = 3;
3277#else
3278 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3279#endif
f99b2589 3280 }
ac27a0ec 3281
281b5995
TT		 3282 /* Handle clustersize */
3283 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3284 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3285 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3286 if (has_bigalloc) {
3287 if (clustersize < blocksize) {
3288 ext4_msg(sb, KERN_ERR,
3289 "cluster size (%d) smaller than "
3290 "block size (%d)", clustersize, blocksize);
3291 goto failed_mount;
3292 }
3293 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3294 le32_to_cpu(es->s_log_block_size);
3295 sbi->s_clusters_per_group =
3296 le32_to_cpu(es->s_clusters_per_group);
3297 if (sbi->s_clusters_per_group > blocksize * 8) {
3298 ext4_msg(sb, KERN_ERR,
3299 "#clusters per group too big: %lu",
3300 sbi->s_clusters_per_group);
3301 goto failed_mount;
3302 }
3303 if (sbi->s_blocks_per_group !=
3304 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3305 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3306 "clusters per group (%lu) inconsistent",
3307 sbi->s_blocks_per_group,
3308 sbi->s_clusters_per_group);
3309 goto failed_mount;
3310 }
3311 } else {
3312 if (clustersize != blocksize) {
3313 ext4_warning(sb, "fragment/cluster size (%d) != "
3314 "block size (%d)", clustersize,
3315 blocksize);
3316 clustersize = blocksize;
3317 }
3318 if (sbi->s_blocks_per_group > blocksize * 8) {
3319 ext4_msg(sb, KERN_ERR,
3320 "#blocks per group too big: %lu",
3321 sbi->s_blocks_per_group);
3322 goto failed_mount;
3323 }
3324 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3325 sbi->s_cluster_bits = 0;
ac27a0ec 3326 }
281b5995
TT		 3327 sbi->s_cluster_ratio = clustersize / blocksize;
3328
ac27a0ec 3329 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3330 ext4_msg(sb, KERN_ERR,
3331 "#inodes per group too big: %lu",
2b2d6d01 3332 sbi->s_inodes_per_group);
ac27a0ec
DK		 3333 goto failed_mount;
3334 }
3335
bf43d84b
ES		 3336 /*
3337 * Test whether we have more sectors than will fit in sector_t,
3338 * and whether the max offset is addressable by the page cache.
3339 */
5a9ae68a 3340 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3341 ext4_blocks_count(es));
5a9ae68a 3342 if (err) {
b31e1552 3343 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3344 " too large to mount safely on this system");
ac27a0ec 3345 if (sizeof(sector_t) < 8)
90c699a9 3346 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3347 ret = err;
ac27a0ec
DK		 3348 goto failed_mount;
3349 }
3350
617ba13b
MC		 3351 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3352 goto cantfind_ext4;
e7c95593 3353
0f2ddca6
FTN		 3354 /* check blocks count against device size */
3355 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3356 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3357 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3358 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3359 ext4_blocks_count(es), blocks_count);
3360 goto failed_mount;
3361 }
3362
0b8e58a1
AD		 3363 /*
3364 * It makes no sense for the first data block to be beyond the end
3365 * of the filesystem.
3366 */
3367 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3368 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3369 "block %u is beyond end of filesystem (%llu)",
3370 le32_to_cpu(es->s_first_data_block),
3371 ext4_blocks_count(es));
e7c95593
ES		 3372 goto failed_mount;
3373 }
bd81d8ee
LV		 3374 blocks_count = (ext4_blocks_count(es) -
3375 le32_to_cpu(es->s_first_data_block) +
3376 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3377 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3378 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3379 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3380 "(block count %llu, first data block %u, "
b31e1552 3381 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3382 ext4_blocks_count(es),
3383 le32_to_cpu(es->s_first_data_block),
3384 EXT4_BLOCKS_PER_GROUP(sb));
3385 goto failed_mount;
3386 }
bd81d8ee 3387 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3388 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3389 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3390 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3391 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3392 sbi->s_group_desc = ext4_kvmalloc(db_count *
3393 sizeof(struct buffer_head *),
3394 GFP_KERNEL);
ac27a0ec 3395 if (sbi->s_group_desc == NULL) {
b31e1552 3396 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 3397 goto failed_mount;
3398 }
3399
9f6200bb
TT		 3400 if (ext4_proc_root)
3401 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3402
66acdcf4
TT		 3403 if (sbi->s_proc)
3404 proc_create_data("options", S_IRUGO, sbi->s_proc,
3405 &ext4_seq_options_fops, sb);
3406
705895b6 3407 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3408
3409 for (i = 0; i < db_count; i++) {
70bbb3e0 3410 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3411 sbi->s_group_desc[i] = sb_bread(sb, block);
3412 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3413 ext4_msg(sb, KERN_ERR,
3414 "can't read group descriptor %d", i);
ac27a0ec
DK		 3415 db_count = i;
3416 goto failed_mount2;
3417 }
3418 }
bfff6873 3419 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3420 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3421 goto failed_mount2;
3422 }
772cb7c8
JS		 3423 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3424 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3425 ext4_msg(sb, KERN_ERR,
3426 "unable to initialize "
3427 "flex_bg meta info!");
772cb7c8
JS		 3428 goto failed_mount2;
3429 }
3430
ac27a0ec
DK		 3431 sbi->s_gdb_count = db_count;
3432 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3433 spin_lock_init(&sbi->s_next_gen_lock);
3434
04496411
TM		 3435 init_timer(&sbi->s_err_report);
3436 sbi->s_err_report.function = print_daily_error_info;
3437 sbi->s_err_report.data = (unsigned long) sb;
3438
57042651 3439 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3440 ext4_count_free_clusters(sb));
ce7e010a
TT		 3441 if (!err) {
3442 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3443 ext4_count_free_inodes(sb));
3444 }
3445 if (!err) {
3446 err = percpu_counter_init(&sbi->s_dirs_counter,
3447 ext4_count_dirs(sb));
3448 }
3449 if (!err) {
57042651 3450 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3451 }
3452 if (err) {
3453 ext4_msg(sb, KERN_ERR, "insufficient memory");
3454 goto failed_mount3;
3455 }
3456
c9de560d 3457 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3458 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3459
ac27a0ec
DK		 3460 /*
3461 * set up enough so that it can read an inode
3462 */
9ca92389
TT		 3463 if (!test_opt(sb, NOLOAD) &&
3464 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3465 sb->s_op = &ext4_sops;
3466 else
3467 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3468 sb->s_export_op = &ext4_export_ops;
3469 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3470#ifdef CONFIG_QUOTA
617ba13b
MC		 3471 sb->s_qcop = &ext4_qctl_operations;
3472 sb->dq_op = &ext4_quota_operations;
ac27a0ec 3473#endif
f2fa2ffc
AK		 3474 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3475
ac27a0ec 3476 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3477 mutex_init(&sbi->s_orphan_lock);
8f82f840 3478 sbi->s_resize_flags = 0;
ac27a0ec
DK		 3479
3480 sb->s_root = NULL;
3481
3482 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3483 EXT4_HAS_INCOMPAT_FEATURE(sb,
3484 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3485
c5e06d10
JL		 3486 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3487 !(sb->s_flags & MS_RDONLY))
3488 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3489 goto failed_mount3;
3490
ac27a0ec
DK		 3491 /*
3492 * The first inode we look at is the journal inode. Don't try
3493 * root first: it may be modified in the journal!
3494 */
3495 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3496 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3497 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3498 goto failed_mount3;
0390131b
FM		 3499 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3500 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3501 ext4_msg(sb, KERN_ERR, "required journal recovery "
3502 "suppressed and not mounted read-only");
744692dc 3503 goto failed_mount_wq;
ac27a0ec 3504 } else {
fd8c37ec 3505 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3506 sbi->s_journal = NULL;
3507 needs_recovery = 0;
3508 goto no_journal;
ac27a0ec
DK		 3509 }
3510
eb40a09c
JS		 3511 if (ext4_blocks_count(es) > 0xffffffffULL &&
3512 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3513 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3514 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3515 goto failed_mount_wq;
eb40a09c
JS		 3516 }
3517
d4da6c9c
LT		 3518 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3519 jbd2_journal_set_features(sbi->s_journal,
3520 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 3521 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3522 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3523 jbd2_journal_set_features(sbi->s_journal,
3524 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 3525 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3526 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3527 } else {
3528 jbd2_journal_clear_features(sbi->s_journal,
3529 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3530 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3531 }
818d276c 3532
ac27a0ec
DK		 3533 /* We have now updated the journal if required, so we can
3534 * validate the data journaling mode. */
3535 switch (test_opt(sb, DATA_FLAGS)) {
3536 case 0:
3537 /* No mode set, assume a default based on the journal
63f57933
AM		 3538 * capabilities: ORDERED_DATA if the journal can
3539 * cope, else JOURNAL_DATA
3540 */
dab291af
MC		 3541 if (jbd2_journal_check_available_features
3542 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3543 set_opt(sb, ORDERED_DATA);
ac27a0ec 3544 else
fd8c37ec 3545 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3546 break;
3547
617ba13b
MC		 3548 case EXT4_MOUNT_ORDERED_DATA:
3549 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3550 if (!jbd2_journal_check_available_features
3551 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3552 ext4_msg(sb, KERN_ERR, "Journal does not support "
3553 "requested data journaling mode");
744692dc 3554 goto failed_mount_wq;
ac27a0ec
DK		 3555 }
3556 default:
3557 break;
3558 }
b3881f74 3559 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3560
18aadd47
BJ		 3561 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3562
ce7e010a
TT		 3563 /*
3564 * The journal may have updated the bg summary counts, so we
3565 * need to update the global counters.
3566 */
57042651 3567 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3568 ext4_count_free_clusters(sb));
ce7e010a
TT		 3569 percpu_counter_set(&sbi->s_freeinodes_counter,
3570 ext4_count_free_inodes(sb));
3571 percpu_counter_set(&sbi->s_dirs_counter,
3572 ext4_count_dirs(sb));
57042651 3573 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3574
ce7e010a 3575no_journal:
fd89d5f2
TH		 3576 /*
3577 * The maximum number of concurrent works can be high and
3578 * concurrency isn't really necessary. Limit it to 1.
3579 */
3580 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3581 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3582 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3583 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3584 goto failed_mount_wq;
3585 }
3586
ac27a0ec 3587 /*
dab291af 3588 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3589 * so we can safely mount the rest of the filesystem now.
3590 */
3591
1d1fe1ee
DH		 3592 root = ext4_iget(sb, EXT4_ROOT_INO);
3593 if (IS_ERR(root)) {
b31e1552 3594 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3595 ret = PTR_ERR(root);
32a9bb57 3596 root = NULL;
ac27a0ec
DK		 3597 goto failed_mount4;
3598 }
3599 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3600 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3601 iput(root);
ac27a0ec
DK		 3602 goto failed_mount4;
3603 }
48fde701 3604 sb->s_root = d_make_root(root);
1d1fe1ee 3605 if (!sb->s_root) {
b31e1552 3606 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3607 ret = -ENOMEM;
3608 goto failed_mount4;
3609 }
ac27a0ec 3610
2b2d6d01 3611 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 3612
3613 /* determine the minimum size of new large inodes, if present */
3614 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3615 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3616 EXT4_GOOD_OLD_INODE_SIZE;
3617 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3618 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3619 if (sbi->s_want_extra_isize <
3620 le16_to_cpu(es->s_want_extra_isize))
3621 sbi->s_want_extra_isize =
3622 le16_to_cpu(es->s_want_extra_isize);
3623 if (sbi->s_want_extra_isize <
3624 le16_to_cpu(es->s_min_extra_isize))
3625 sbi->s_want_extra_isize =
3626 le16_to_cpu(es->s_min_extra_isize);
3627 }
3628 }
3629 /* Check if enough inode space is available */
3630 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3631 sbi->s_inode_size) {
3632 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3633 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3634 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3635 "available");
ef7f3835
KS		 3636 }
3637
6fd058f7
TT		 3638 err = ext4_setup_system_zone(sb);
3639 if (err) {
b31e1552 3640 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3641 "zone (%d)", err);
94bf608a 3642 goto failed_mount4a;
6fd058f7
TT		 3643 }
3644
c2774d84
AK		 3645 ext4_ext_init(sb);
3646 err = ext4_mb_init(sb, needs_recovery);
3647 if (err) {
421f91d2 3648 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3649 err);
dcf2d804 3650 goto failed_mount5;
c2774d84
AK		 3651 }
3652
bfff6873
LC		 3653 err = ext4_register_li_request(sb, first_not_zeroed);
3654 if (err)
dcf2d804 3655 goto failed_mount6;
bfff6873 3656
3197ebdb
TT		 3657 sbi->s_kobj.kset = ext4_kset;
3658 init_completion(&sbi->s_kobj_unregister);
3659 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3660 "%s", sb->s_id);
dcf2d804
TM		 3661 if (err)
3662 goto failed_mount7;
3197ebdb 3663
617ba13b
MC		 3664 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3665 ext4_orphan_cleanup(sb, es);
3666 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3667 if (needs_recovery) {
b31e1552 3668 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3669 ext4_mark_recovery_complete(sb, es);
3670 }
3671 if (EXT4_SB(sb)->s_journal) {
3672 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3673 descr = " journalled data mode";
3674 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3675 descr = " ordered data mode";
3676 else
3677 descr = " writeback data mode";
3678 } else
3679 descr = "out journal";
3680
d4c402d9 3681 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3682 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3683 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3684
66e61a9e
TT		 3685 if (es->s_error_count)
3686 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3687
d4c402d9 3688 kfree(orig_data);
ac27a0ec
DK		 3689 return 0;
3690
617ba13b 3691cantfind_ext4:
ac27a0ec 3692 if (!silent)
b31e1552 3693 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3694 goto failed_mount;
3695
dcf2d804
TM		 3696failed_mount7:
3697 ext4_unregister_li_request(sb);
3698failed_mount6:
dcf2d804 3699 ext4_mb_release(sb);
94bf608a
AV		 3700failed_mount5:
3701 ext4_ext_release(sb);
dcf2d804 3702 ext4_release_system_zone(sb);
94bf608a
AV		 3703failed_mount4a:
3704 dput(sb->s_root);
32a9bb57 3705 sb->s_root = NULL;
94bf608a 3706failed_mount4:
b31e1552 3707 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3708 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3709failed_mount_wq:
0390131b
FM		 3710 if (sbi->s_journal) {
3711 jbd2_journal_destroy(sbi->s_journal);
3712 sbi->s_journal = NULL;
3713 }
ac27a0ec 3714failed_mount3:
04496411 3715 del_timer(&sbi->s_err_report);
9933fc0a
TT		 3716 if (sbi->s_flex_groups)
3717 ext4_kvfree(sbi->s_flex_groups);
57042651 3718 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 3719 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3720 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 3721 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 3722 if (sbi->s_mmp_tsk)
3723 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 3724failed_mount2:
3725 for (i = 0; i < db_count; i++)
3726 brelse(sbi->s_group_desc[i]);
f18a5f21 3727 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 3728failed_mount:
240799cd 3729 if (sbi->s_proc) {
66acdcf4 3730 remove_proc_entry("options", sbi->s_proc);
9f6200bb 3731 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3732 }
ac27a0ec
DK		 3733#ifdef CONFIG_QUOTA
3734 for (i = 0; i < MAXQUOTAS; i++)
3735 kfree(sbi->s_qf_names[i]);
3736#endif
617ba13b 3737 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3738 brelse(bh);
3739out_fail:
3740 sb->s_fs_info = NULL;
f6830165 3741 kfree(sbi->s_blockgroup_lock);
ac27a0ec 3742 kfree(sbi);
dcc7dae3 3743out_free_orig:
d4c402d9 3744 kfree(orig_data);
1d1fe1ee 3745 return ret;
ac27a0ec
DK		 3746}
3747
3748/*
3749 * Setup any per-fs journal parameters now. We'll do this both on
3750 * initial mount, once the journal has been initialised but before we've
3751 * done any recovery; and again on any subsequent remount.
3752 */
617ba13b 3753static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3754{
617ba13b 3755 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3756
30773840
TT		 3757 journal->j_commit_interval = sbi->s_commit_interval;
3758 journal->j_min_batch_time = sbi->s_min_batch_time;
3759 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 3760
a931da6a 3761 write_lock(&journal->j_state_lock);
ac27a0ec 3762 if (test_opt(sb, BARRIER))
dab291af 3763 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3764 else
dab291af 3765 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3766 if (test_opt(sb, DATA_ERR_ABORT))
3767 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3768 else
3769 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 3770 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 3771}
3772
617ba13b 3773static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3774 unsigned int journal_inum)
3775{
3776 struct inode *journal_inode;
3777 journal_t *journal;
3778
0390131b
FM		 3779 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3780
ac27a0ec
DK		 3781 /* First, test for the existence of a valid inode on disk. Bad
3782 * things happen if we iget() an unused inode, as the subsequent
3783 * iput() will try to delete it. */
3784
1d1fe1ee
DH		 3785 journal_inode = ext4_iget(sb, journal_inum);
3786 if (IS_ERR(journal_inode)) {
b31e1552 3787 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3788 return NULL;
3789 }
3790 if (!journal_inode->i_nlink) {
3791 make_bad_inode(journal_inode);
3792 iput(journal_inode);
b31e1552 3793 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3794 return NULL;
3795 }
3796
e5f8eab8 3797 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3798 journal_inode, journal_inode->i_size);
1d1fe1ee 3799 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3800 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3801 iput(journal_inode);
3802 return NULL;
3803 }
3804
dab291af 3805 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3806 if (!journal) {
b31e1552 3807 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3808 iput(journal_inode);
3809 return NULL;
3810 }
3811 journal->j_private = sb;
617ba13b 3812 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3813 return journal;
3814}
3815
617ba13b 3816static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3817 dev_t j_dev)
3818{
2b2d6d01 3819 struct buffer_head *bh;
ac27a0ec 3820 journal_t *journal;
617ba13b
MC		 3821 ext4_fsblk_t start;
3822 ext4_fsblk_t len;
ac27a0ec 3823 int hblock, blocksize;
617ba13b 3824 ext4_fsblk_t sb_block;
ac27a0ec 3825 unsigned long offset;
2b2d6d01 3826 struct ext4_super_block *es;
ac27a0ec
DK		 3827 struct block_device *bdev;
3828
0390131b
FM		 3829 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3830
b31e1552 3831 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3832 if (bdev == NULL)
3833 return NULL;
3834
ac27a0ec 3835 blocksize = sb->s_blocksize;
e1defc4f 3836 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3837 if (blocksize < hblock) {
b31e1552
ES		 3838 ext4_msg(sb, KERN_ERR,
3839 "blocksize too small for journal device");
ac27a0ec
DK		 3840 goto out_bdev;
3841 }
3842
617ba13b
MC		 3843 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3844 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3845 set_blocksize(bdev, blocksize);
3846 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3847 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3848 "external journal");
ac27a0ec
DK		 3849 goto out_bdev;
3850 }
3851
617ba13b
MC		 3852 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3853 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3854 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3855 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3856 ext4_msg(sb, KERN_ERR, "external journal has "
3857 "bad superblock");
ac27a0ec
DK		 3858 brelse(bh);
3859 goto out_bdev;
3860 }
3861
617ba13b 3862 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3863 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3864 brelse(bh);
3865 goto out_bdev;
3866 }
3867
bd81d8ee 3868 len = ext4_blocks_count(es);
ac27a0ec
DK		 3869 start = sb_block + 1;
3870 brelse(bh); /* we're done with the superblock */
3871
dab291af 3872 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3873 start, len, blocksize);
3874 if (!journal) {
b31e1552 3875 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3876 goto out_bdev;
3877 }
3878 journal->j_private = sb;
3879 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3880 wait_on_buffer(journal->j_sb_buffer);
3881 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3882 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3883 goto out_journal;
3884 }
3885 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3886 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3887 "user (unsupported) - %d",
ac27a0ec
DK		 3888 be32_to_cpu(journal->j_superblock->s_nr_users));
3889 goto out_journal;
3890 }
617ba13b
MC		 3891 EXT4_SB(sb)->journal_bdev = bdev;
3892 ext4_init_journal_params(sb, journal);
ac27a0ec 3893 return journal;
0b8e58a1 3894
ac27a0ec 3895out_journal:
dab291af 3896 jbd2_journal_destroy(journal);
ac27a0ec 3897out_bdev:
617ba13b 3898 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3899 return NULL;
3900}
3901
617ba13b
MC		 3902static int ext4_load_journal(struct super_block *sb,
3903 struct ext4_super_block *es,
ac27a0ec
DK		 3904 unsigned long journal_devnum)
3905{
3906 journal_t *journal;
3907 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3908 dev_t journal_dev;
3909 int err = 0;
3910 int really_read_only;
3911
0390131b
FM		 3912 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3913
ac27a0ec
DK		 3914 if (journal_devnum &&
3915 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3916 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3917 "numbers have changed");
ac27a0ec
DK		 3918 journal_dev = new_decode_dev(journal_devnum);
3919 } else
3920 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3921
3922 really_read_only = bdev_read_only(sb->s_bdev);
3923
3924 /*
3925 * Are we loading a blank journal or performing recovery after a
3926 * crash? For recovery, we need to check in advance whether we
3927 * can get read-write access to the device.
3928 */
617ba13b 3929 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3930 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3931 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3932 "required on readonly filesystem");
ac27a0ec 3933 if (really_read_only) {
b31e1552
ES		 3934 ext4_msg(sb, KERN_ERR, "write access "
3935 "unavailable, cannot proceed");
ac27a0ec
DK		 3936 return -EROFS;
3937 }
b31e1552
ES		 3938 ext4_msg(sb, KERN_INFO, "write access will "
3939 "be enabled during recovery");
ac27a0ec
DK		 3940 }
3941 }
3942
3943 if (journal_inum && journal_dev) {
b31e1552
ES		 3944 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3945 "and inode journals!");
ac27a0ec
DK		 3946 return -EINVAL;
3947 }
3948
3949 if (journal_inum) {
617ba13b 3950 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3951 return -EINVAL;
3952 } else {
617ba13b 3953 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3954 return -EINVAL;
3955 }
3956
90576c0b 3957 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 3958 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3959
617ba13b 3960 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3961 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 3962 if (!err) {
3963 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3964 if (save)
3965 memcpy(save, ((char *) es) +
3966 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 3967 err = jbd2_journal_load(journal);
1c13d5c0
TT		 3968 if (save)
3969 memcpy(((char *) es) + EXT4_S_ERR_START,
3970 save, EXT4_S_ERR_LEN);
3971 kfree(save);
3972 }
ac27a0ec
DK		 3973
3974 if (err) {
b31e1552 3975 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3976 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3977 return err;
3978 }
3979
617ba13b
MC		 3980 EXT4_SB(sb)->s_journal = journal;
3981 ext4_clear_journal_err(sb, es);
ac27a0ec 3982
c41303ce 3983 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 3984 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3985 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3986
3987 /* Make sure we flush the recovery flag to disk. */
e2d67052 3988 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3989 }
3990
3991 return 0;
3992}
3993
e2d67052 3994static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3995{
e2d67052 3996 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3997 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3998 int error = 0;
ac27a0ec 3999
7c2e7087 4000 if (!sbh || block_device_ejected(sb))
c4be0c1d 4001 return error;
914258bf
TT		 4002 if (buffer_write_io_error(sbh)) {
4003 /*
4004 * Oh, dear. A previous attempt to write the
4005 * superblock failed. This could happen because the
4006 * USB device was yanked out. Or it could happen to
4007 * be a transient write error and maybe the block will
4008 * be remapped. Nothing we can do but to retry the
4009 * write and hope for the best.
4010 */
b31e1552
ES		 4011 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4012 "superblock detected");
914258bf
TT		 4013 clear_buffer_write_io_error(sbh);
4014 set_buffer_uptodate(sbh);
4015 }
71290b36
TT		 4016 /*
4017 * If the file system is mounted read-only, don't update the
4018 * superblock write time. This avoids updating the superblock
4019 * write time when we are mounting the root file system
4020 * read/only but we need to replay the journal; at that point,
4021 * for people who are east of GMT and who make their clock
4022 * tick in localtime for Windows bug-for-bug compatibility,
4023 * the clock is set in the future, and this will cause e2fsck
4024 * to complain and force a full file system check.
4025 */
4026 if (!(sb->s_flags & MS_RDONLY))
4027 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4028 if (sb->s_bdev->bd_part)
4029 es->s_kbytes_written =
4030 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4031 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4032 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4033 else
4034 es->s_kbytes_written =
4035 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4036 ext4_free_blocks_count_set(es,
4037 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4038 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4039 es->s_free_inodes_count =
4040 cpu_to_le32(percpu_counter_sum_positive(
4041 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 4042 sb->s_dirt = 0;
ac27a0ec
DK		 4043 BUFFER_TRACE(sbh, "marking dirty");
4044 mark_buffer_dirty(sbh);
914258bf 4045 if (sync) {
c4be0c1d
TS		 4046 error = sync_dirty_buffer(sbh);
4047 if (error)
4048 return error;
4049
4050 error = buffer_write_io_error(sbh);
4051 if (error) {
b31e1552
ES		 4052 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4053 "superblock");
914258bf
TT		 4054 clear_buffer_write_io_error(sbh);
4055 set_buffer_uptodate(sbh);
4056 }
4057 }
c4be0c1d 4058 return error;
ac27a0ec
DK		 4059}
4060
ac27a0ec
DK		 4061/*
4062 * Have we just finished recovery? If so, and if we are mounting (or
4063 * remounting) the filesystem readonly, then we will end up with a
4064 * consistent fs on disk. Record that fact.
4065 */
2b2d6d01
TT		 4066static void ext4_mark_recovery_complete(struct super_block *sb,
4067 struct ext4_super_block *es)
ac27a0ec 4068{
617ba13b 4069 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4070
0390131b
FM		 4071 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4072 BUG_ON(journal != NULL);
4073 return;
4074 }
dab291af 4075 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4076 if (jbd2_journal_flush(journal) < 0)
4077 goto out;
4078
617ba13b 4079 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4080 sb->s_flags & MS_RDONLY) {
617ba13b 4081 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4082 ext4_commit_super(sb, 1);
ac27a0ec 4083 }
7ffe1ea8
HK		 4084
4085out:
dab291af 4086 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4087}
4088
4089/*
4090 * If we are mounting (or read-write remounting) a filesystem whose journal
4091 * has recorded an error from a previous lifetime, move that error to the
4092 * main filesystem now.
4093 */
2b2d6d01
TT		 4094static void ext4_clear_journal_err(struct super_block *sb,
4095 struct ext4_super_block *es)
ac27a0ec
DK		 4096{
4097 journal_t *journal;
4098 int j_errno;
4099 const char *errstr;
4100
0390131b
FM		 4101 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4102
617ba13b 4103 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4104
4105 /*
4106 * Now check for any error status which may have been recorded in the
617ba13b 4107 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4108 */
4109
dab291af 4110 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4111 if (j_errno) {
4112 char nbuf[16];
4113
617ba13b 4114 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4115 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4116 "from previous mount: %s", errstr);
12062ddd 4117 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4118
617ba13b
MC		 4119 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4120 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4121 ext4_commit_super(sb, 1);
ac27a0ec 4122
dab291af 4123 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 4124 }
4125}
4126
4127/*
4128 * Force the running and committing transactions to commit,
4129 * and wait on the commit.
4130 */
617ba13b 4131int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4132{
4133 journal_t *journal;
0390131b 4134 int ret = 0;
ac27a0ec
DK		 4135
4136 if (sb->s_flags & MS_RDONLY)
4137 return 0;
4138
617ba13b 4139 journal = EXT4_SB(sb)->s_journal;
6b0310fb 4140 if (journal) {
437f88cc 4141 vfs_check_frozen(sb, SB_FREEZE_TRANS);
0390131b 4142 ret = ext4_journal_force_commit(journal);
6b0310fb 4143 }
0390131b 4144
ac27a0ec
DK		 4145 return ret;
4146}
4147
2b2d6d01 4148static void ext4_write_super(struct super_block *sb)
ac27a0ec 4149{
ebc1ac16 4150 lock_super(sb);
9ca92389 4151 ext4_commit_super(sb, 1);
ebc1ac16 4152 unlock_super(sb);
ac27a0ec
DK		 4153}
4154
617ba13b 4155static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4156{
14ce0cb4 4157 int ret = 0;
9eddacf9 4158 tid_t target;
8d5d02e6 4159 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4160
9bffad1e 4161 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 4162 flush_workqueue(sbi->dio_unwritten_wq);
4163 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4164 if (wait)
8d5d02e6 4165 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4166 }
14ce0cb4 4167 return ret;
ac27a0ec
DK		 4168}
4169
4170/*
4171 * LVM calls this function before a (read-only) snapshot is created. This
4172 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4173 *
4174 * Note that only this function cannot bring a filesystem to be in a clean
4175 * state independently, because ext4 prevents a new handle from being started
4176 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4177 * the upper layer.
ac27a0ec 4178 */
c4be0c1d 4179static int ext4_freeze(struct super_block *sb)
ac27a0ec 4180{
c4be0c1d
TS		 4181 int error = 0;
4182 journal_t *journal;
ac27a0ec 4183
9ca92389
TT		 4184 if (sb->s_flags & MS_RDONLY)
4185 return 0;
ac27a0ec 4186
9ca92389 4187 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4188
9ca92389
TT		 4189 /* Now we set up the journal barrier. */
4190 jbd2_journal_lock_updates(journal);
ac27a0ec 4191
9ca92389
TT		 4192 /*
4193 * Don't clear the needs_recovery flag if we failed to flush
4194 * the journal.
4195 */
4196 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4197 if (error < 0)
4198 goto out;
9ca92389
TT		 4199
4200 /* Journal blocked and flushed, clear needs_recovery flag. */
4201 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4202 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 4203out:
4204 /* we rely on s_frozen to stop further updates */
4205 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4206 return error;
ac27a0ec
DK		 4207}
4208
4209/*
4210 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4211 * flag here, even though the filesystem is not technically dirty yet.
4212 */
c4be0c1d 4213static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4214{
9ca92389
TT		 4215 if (sb->s_flags & MS_RDONLY)
4216 return 0;
4217
4218 lock_super(sb);
4219 /* Reset the needs_recovery flag before the fs is unlocked. */
4220 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4221 ext4_commit_super(sb, 1);
4222 unlock_super(sb);
c4be0c1d 4223 return 0;
ac27a0ec
DK		 4224}
4225
673c6100
TT		 4226/*
4227 * Structure to save mount options for ext4_remount's benefit
4228 */
4229struct ext4_mount_options {
4230 unsigned long s_mount_opt;
a2595b8a 4231 unsigned long s_mount_opt2;
673c6100
TT		 4232 uid_t s_resuid;
4233 gid_t s_resgid;
4234 unsigned long s_commit_interval;
4235 u32 s_min_batch_time, s_max_batch_time;
4236#ifdef CONFIG_QUOTA
4237 int s_jquota_fmt;
4238 char *s_qf_names[MAXQUOTAS];
4239#endif
4240};
4241
2b2d6d01 4242static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4243{
2b2d6d01 4244 struct ext4_super_block *es;
617ba13b 4245 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4246 unsigned long old_sb_flags;
617ba13b 4247 struct ext4_mount_options old_opts;
c79d967d 4248 int enable_quota = 0;
8a266467 4249 ext4_group_t g;
b3881f74 4250 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4251 int err = 0;
ac27a0ec
DK		 4252#ifdef CONFIG_QUOTA
4253 int i;
4254#endif
d4c402d9 4255 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4256
4257 /* Store the original options */
bbd6851a 4258 lock_super(sb);
ac27a0ec
DK		 4259 old_sb_flags = sb->s_flags;
4260 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4261 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4262 old_opts.s_resuid = sbi->s_resuid;
4263 old_opts.s_resgid = sbi->s_resgid;
4264 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4265 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4266 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4267#ifdef CONFIG_QUOTA
4268 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4269 for (i = 0; i < MAXQUOTAS; i++)
4270 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4271#endif
b3881f74
TT		 4272 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4273 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4274
4275 /*
4276 * Allow the "check" option to be passed as a remount option.
4277 */
661aa520 4278 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4279 err = -EINVAL;
4280 goto restore_opts;
4281 }
4282
4ab2f15b 4283 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4284 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4285
4286 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4287 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4288
4289 es = sbi->s_es;
4290
b3881f74 4291 if (sbi->s_journal) {
0390131b 4292 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4293 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4294 }
ac27a0ec 4295
661aa520 4296 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4297 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4298 err = -EROFS;
4299 goto restore_opts;
4300 }
4301
4302 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4303 err = dquot_suspend(sb, -1);
4304 if (err < 0)
c79d967d 4305 goto restore_opts;
c79d967d 4306
ac27a0ec
DK		 4307 /*
4308 * First of all, the unconditional stuff we have to do
4309 * to disable replay of the journal when we next remount
4310 */
4311 sb->s_flags |= MS_RDONLY;
4312
4313 /*
4314 * OK, test if we are remounting a valid rw partition
4315 * readonly, and if so set the rdonly flag and then
4316 * mark the partition as valid again.
4317 */
617ba13b
MC		 4318 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4319 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4320 es->s_state = cpu_to_le16(sbi->s_mount_state);
4321
a63c9eb2 4322 if (sbi->s_journal)
0390131b 4323 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4324 } else {
a13fb1a4
ES		 4325 /* Make sure we can mount this feature set readwrite */
4326 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4327 err = -EROFS;
4328 goto restore_opts;
4329 }
8a266467
TT		 4330 /*
4331 * Make sure the group descriptor checksums
0b8e58a1 4332 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4333 */
4334 for (g = 0; g < sbi->s_groups_count; g++) {
4335 struct ext4_group_desc *gdp =
4336 ext4_get_group_desc(sb, g, NULL);
4337
4338 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 4339 ext4_msg(sb, KERN_ERR,
4340 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4341 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4342 le16_to_cpu(gdp->bg_checksum));
4343 err = -EINVAL;
4344 goto restore_opts;
4345 }
4346 }
4347
ead6596b
ES		 4348 /*
4349 * If we have an unprocessed orphan list hanging
4350 * around from a previously readonly bdev mount,
4351 * require a full umount/remount for now.
4352 */
4353 if (es->s_last_orphan) {
b31e1552 4354 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4355 "remount RDWR because of unprocessed "
4356 "orphan inode list. Please "
b31e1552 4357 "umount/remount instead");
ead6596b
ES		 4358 err = -EINVAL;
4359 goto restore_opts;
4360 }
4361
ac27a0ec
DK		 4362 /*
4363 * Mounting a RDONLY partition read-write, so reread
4364 * and store the current valid flag. (It may have
4365 * been changed by e2fsck since we originally mounted
4366 * the partition.)
4367 */
0390131b
FM		 4368 if (sbi->s_journal)
4369 ext4_clear_journal_err(sb, es);
ac27a0ec 4370 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4371 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4372 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4373 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4374 EXT4_FEATURE_INCOMPAT_MMP))
4375 if (ext4_multi_mount_protect(sb,
4376 le64_to_cpu(es->s_mmp_block))) {
4377 err = -EROFS;
4378 goto restore_opts;
4379 }
c79d967d 4380 enable_quota = 1;
ac27a0ec
DK		 4381 }
4382 }
bfff6873
LC		 4383
4384 /*
4385 * Reinitialize lazy itable initialization thread based on
4386 * current settings
4387 */
4388 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4389 ext4_unregister_li_request(sb);
4390 else {
4391 ext4_group_t first_not_zeroed;
4392 first_not_zeroed = ext4_has_uninit_itable(sb);
4393 ext4_register_li_request(sb, first_not_zeroed);
4394 }
4395
6fd058f7 4396 ext4_setup_system_zone(sb);
0390131b 4397 if (sbi->s_journal == NULL)
e2d67052 4398 ext4_commit_super(sb, 1);
0390131b 4399
ac27a0ec
DK		 4400#ifdef CONFIG_QUOTA
4401 /* Release old quota file names */
4402 for (i = 0; i < MAXQUOTAS; i++)
4403 if (old_opts.s_qf_names[i] &&
4404 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4405 kfree(old_opts.s_qf_names[i]);
4406#endif
bbd6851a 4407 unlock_super(sb);
c79d967d 4408 if (enable_quota)
0f0dd62f 4409 dquot_resume(sb, -1);
d4c402d9
CW		 4410
4411 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4412 kfree(orig_data);
ac27a0ec 4413 return 0;
0b8e58a1 4414
ac27a0ec
DK		 4415restore_opts:
4416 sb->s_flags = old_sb_flags;
4417 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4418 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4419 sbi->s_resuid = old_opts.s_resuid;
4420 sbi->s_resgid = old_opts.s_resgid;
4421 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4422 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4423 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4424#ifdef CONFIG_QUOTA
4425 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4426 for (i = 0; i < MAXQUOTAS; i++) {
4427 if (sbi->s_qf_names[i] &&
4428 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4429 kfree(sbi->s_qf_names[i]);
4430 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4431 }
4432#endif
bbd6851a 4433 unlock_super(sb);
d4c402d9 4434 kfree(orig_data);
ac27a0ec
DK		 4435 return err;
4436}
4437
f975d6bc
TT		 4438/*
4439 * Note: calculating the overhead so we can be compatible with
4440 * historical BSD practice is quite difficult in the face of
4441 * clusters/bigalloc. This is because multiple metadata blocks from
4442 * different block group can end up in the same allocation cluster.
4443 * Calculating the exact overhead in the face of clustered allocation
4444 * requires either O(all block bitmaps) in memory or O(number of block
4445 * groups**2) in time. We will still calculate the superblock for
4446 * older file systems --- and if we come across with a bigalloc file
4447 * system with zero in s_overhead_clusters the estimate will be close to
4448 * correct especially for very large cluster sizes --- but for newer
4449 * file systems, it's better to calculate this figure once at mkfs
4450 * time, and store it in the superblock. If the superblock value is
4451 * present (even for non-bigalloc file systems), we will use it.
4452 */
2b2d6d01 4453static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4454{
4455 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4456 struct ext4_sb_info *sbi = EXT4_SB(sb);
4457 struct ext4_super_block *es = sbi->s_es;
f975d6bc 4458 struct ext4_group_desc *gdp;
960cc398 4459 u64 fsid;
d02a9391 4460 s64 bfree;
ac27a0ec 4461
5e70030d
BP		 4462 if (test_opt(sb, MINIX_DF)) {
4463 sbi->s_overhead_last = 0;
f975d6bc
TT		 4464 } else if (es->s_overhead_clusters) {
4465 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
6bc9feff 4466 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 4467 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 4468 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 4469
4470 /*
5e70030d
BP		 4471 * Compute the overhead (FS structures). This is constant
4472 * for a given filesystem unless the number of block groups
4473 * changes so we cache the previous value until it does.
ac27a0ec
DK		 4474 */
4475
4476 /*
4477 * All of the blocks before first_data_block are
4478 * overhead
4479 */
f975d6bc 4480 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
ac27a0ec
DK		 4481
4482 /*
f975d6bc 4483 * Add the overhead found in each block group
ac27a0ec
DK		 4484 */
4485 for (i = 0; i < ngroups; i++) {
f975d6bc
TT		 4486 gdp = ext4_get_group_desc(sb, i, NULL);
4487 overhead += ext4_num_overhead_clusters(sb, i, gdp);
ac27a0ec
DK		 4488 cond_resched();
4489 }
5e70030d
BP		 4490 sbi->s_overhead_last = overhead;
4491 smp_wmb();
6bc9feff 4492 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 4493 }
4494
617ba13b 4495 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4496 buf->f_bsize = sb->s_blocksize;
f975d6bc
TT		 4497 buf->f_blocks = (ext4_blocks_count(es) -
4498 EXT4_C2B(sbi, sbi->s_overhead_last));
57042651
TT		 4499 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4500 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4501 /* prevent underflow in case that few free space is available */
57042651 4502 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4503 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4504 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4505 buf->f_bavail = 0;
4506 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4507 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4508 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4509 fsid = le64_to_cpup((void *)es->s_uuid) ^
4510 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4511 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4512 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4513
ac27a0ec
DK		 4514 return 0;
4515}
4516
0b8e58a1
AD		 4517/* Helper function for writing quotas on sync - we need to start transaction
4518 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4519 * Process 1 Process 2
617ba13b 4520 * ext4_create() quota_sync()
a269eb18 4521 * jbd2_journal_start() write_dquot()
871a2931 4522 * dquot_initialize() down(dqio_mutex)
dab291af 4523 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4524 *
4525 */
4526
4527#ifdef CONFIG_QUOTA
4528
4529static inline struct inode *dquot_to_inode(struct dquot *dquot)
4530{
4531 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4532}
4533
617ba13b 4534static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4535{
4536 int ret, err;
4537 handle_t *handle;
4538 struct inode *inode;
4539
4540 inode = dquot_to_inode(dquot);
617ba13b 4541 handle = ext4_journal_start(inode,
0b8e58a1 4542 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4543 if (IS_ERR(handle))
4544 return PTR_ERR(handle);
4545 ret = dquot_commit(dquot);
617ba13b 4546 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4547 if (!ret)
4548 ret = err;
4549 return ret;
4550}
4551
617ba13b 4552static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4553{
4554 int ret, err;
4555 handle_t *handle;
4556
617ba13b 4557 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4558 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4559 if (IS_ERR(handle))
4560 return PTR_ERR(handle);
4561 ret = dquot_acquire(dquot);
617ba13b 4562 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4563 if (!ret)
4564 ret = err;
4565 return ret;
4566}
4567
617ba13b 4568static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4569{
4570 int ret, err;
4571 handle_t *handle;
4572
617ba13b 4573 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4574 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4575 if (IS_ERR(handle)) {
4576 /* Release dquot anyway to avoid endless cycle in dqput() */
4577 dquot_release(dquot);
ac27a0ec 4578 return PTR_ERR(handle);
9c3013e9 4579 }
ac27a0ec 4580 ret = dquot_release(dquot);
617ba13b 4581 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4582 if (!ret)
4583 ret = err;
4584 return ret;
4585}
4586
617ba13b 4587static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4588{
2c8be6b2 4589 /* Are we journaling quotas? */
617ba13b
MC		 4590 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4591 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4592 dquot_mark_dquot_dirty(dquot);
617ba13b 4593 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4594 } else {
4595 return dquot_mark_dquot_dirty(dquot);
4596 }
4597}
4598
617ba13b 4599static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4600{
4601 int ret, err;
4602 handle_t *handle;
4603
4604 /* Data block + inode block */
617ba13b 4605 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4606 if (IS_ERR(handle))
4607 return PTR_ERR(handle);
4608 ret = dquot_commit_info(sb, type);
617ba13b 4609 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4610 if (!ret)
4611 ret = err;
4612 return ret;
4613}
4614
4615/*
4616 * Turn on quotas during mount time - we need to find
4617 * the quota file and such...
4618 */
617ba13b 4619static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4620{
287a8095
CH		 4621 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4622 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4623}
4624
4625/*
4626 * Standard function to be called on quota_on
4627 */
617ba13b 4628static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4629 struct path *path)
ac27a0ec
DK		 4630{
4631 int err;
ac27a0ec
DK		 4632
4633 if (!test_opt(sb, QUOTA))
4634 return -EINVAL;
0623543b 4635
ac27a0ec 4636 /* Quotafile not on the same filesystem? */
d8c9584e 4637 if (path->dentry->d_sb != sb)
ac27a0ec 4638 return -EXDEV;
0623543b
JK		 4639 /* Journaling quota? */
4640 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4641 /* Quotafile not in fs root? */
f00c9e44 4642 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4643 ext4_msg(sb, KERN_WARNING,
4644 "Quota file not on filesystem root. "
4645 "Journaled quota will not work");
2b2d6d01 4646 }
0623543b
JK		 4647
4648 /*
4649 * When we journal data on quota file, we have to flush journal to see
4650 * all updates to the file when we bypass pagecache...
4651 */
0390131b 4652 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4653 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4654 /*
4655 * We don't need to lock updates but journal_flush() could
4656 * otherwise be livelocked...
4657 */
4658 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4659 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4660 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4661 if (err)
7ffe1ea8 4662 return err;
0623543b
JK		 4663 }
4664
f00c9e44 4665 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4666}
4667
ca0e05e4
DM		 4668static int ext4_quota_off(struct super_block *sb, int type)
4669{
21f97697
JK		 4670 struct inode *inode = sb_dqopt(sb)->files[type];
4671 handle_t *handle;
4672
87009d86
DM		 4673 /* Force all delayed allocation blocks to be allocated.
4674 * Caller already holds s_umount sem */
4675 if (test_opt(sb, DELALLOC))
ca0e05e4 4676 sync_filesystem(sb);
ca0e05e4 4677
0b268590
AG		 4678 if (!inode)
4679 goto out;
4680
21f97697
JK		 4681 /* Update modification times of quota files when userspace can
4682 * start looking at them */
4683 handle = ext4_journal_start(inode, 1);
4684 if (IS_ERR(handle))
4685 goto out;
4686 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4687 ext4_mark_inode_dirty(handle, inode);
4688 ext4_journal_stop(handle);
4689
4690out:
ca0e05e4
DM		 4691 return dquot_quota_off(sb, type);
4692}
4693
ac27a0ec
DK		 4694/* Read data from quotafile - avoid pagecache and such because we cannot afford
4695 * acquiring the locks... As quota files are never truncated and quota code
25985edc 4696 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 4697 * we don't have to be afraid of races */
617ba13b 4698static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4699 size_t len, loff_t off)
4700{
4701 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4702 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4703 int err = 0;
4704 int offset = off & (sb->s_blocksize - 1);
4705 int tocopy;
4706 size_t toread;
4707 struct buffer_head *bh;
4708 loff_t i_size = i_size_read(inode);
4709
4710 if (off > i_size)
4711 return 0;
4712 if (off+len > i_size)
4713 len = i_size-off;
4714 toread = len;
4715 while (toread > 0) {
4716 tocopy = sb->s_blocksize - offset < toread ?
4717 sb->s_blocksize - offset : toread;
617ba13b 4718 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4719 if (err)
4720 return err;
4721 if (!bh) /* A hole? */
4722 memset(data, 0, tocopy);
4723 else
4724 memcpy(data, bh->b_data+offset, tocopy);
4725 brelse(bh);
4726 offset = 0;
4727 toread -= tocopy;
4728 data += tocopy;
4729 blk++;
4730 }
4731 return len;
4732}
4733
4734/* Write to quotafile (we know the transaction is already started and has
4735 * enough credits) */
617ba13b 4736static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4737 const char *data, size_t len, loff_t off)
4738{
4739 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4740 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4741 int err = 0;
4742 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 4743 struct buffer_head *bh;
4744 handle_t *handle = journal_current_handle();
4745
0390131b 4746 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4747 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4748 " cancelled because transaction is not started",
9c3013e9
JK		 4749 (unsigned long long)off, (unsigned long long)len);
4750 return -EIO;
4751 }
67eeb568
DM		 4752 /*
4753 * Since we account only one data block in transaction credits,
4754 * then it is impossible to cross a block boundary.
4755 */
4756 if (sb->s_blocksize - offset < len) {
4757 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4758 " cancelled because not block aligned",
4759 (unsigned long long)off, (unsigned long long)len);
4760 return -EIO;
4761 }
4762
ac27a0ec 4763 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4764 bh = ext4_bread(handle, inode, blk, 1, &err);
4765 if (!bh)
4766 goto out;
62d2b5f2
JK		 4767 err = ext4_journal_get_write_access(handle, bh);
4768 if (err) {
4769 brelse(bh);
4770 goto out;
ac27a0ec 4771 }
67eeb568
DM		 4772 lock_buffer(bh);
4773 memcpy(bh->b_data+offset, data, len);
4774 flush_dcache_page(bh->b_page);
4775 unlock_buffer(bh);
62d2b5f2 4776 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 4777 brelse(bh);
ac27a0ec 4778out:
67eeb568 4779 if (err) {
4d04e4fb 4780 mutex_unlock(&inode->i_mutex);
ac27a0ec 4781 return err;
4d04e4fb 4782 }
67eeb568
DM		 4783 if (inode->i_size < off + len) {
4784 i_size_write(inode, off + len);
617ba13b 4785 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 4786 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4787 }
ac27a0ec 4788 mutex_unlock(&inode->i_mutex);
67eeb568 4789 return len;
ac27a0ec
DK		 4790}
4791
4792#endif
4793
152a0836
AV		 4794static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4795 const char *dev_name, void *data)
ac27a0ec 4796{
152a0836 4797 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 4798}
4799
37f328eb 4800#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4801static inline void register_as_ext2(void)
4802{
4803 int err = register_filesystem(&ext2_fs_type);
4804 if (err)
4805 printk(KERN_WARNING
4806 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4807}
4808
4809static inline void unregister_as_ext2(void)
4810{
4811 unregister_filesystem(&ext2_fs_type);
4812}
2035e776
TT		 4813
4814static inline int ext2_feature_set_ok(struct super_block *sb)
4815{
4816 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4817 return 0;
4818 if (sb->s_flags & MS_RDONLY)
4819 return 1;
4820 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4821 return 0;
4822 return 1;
4823}
51b7e3c9 4824MODULE_ALIAS("ext2");
24b58424
TT		 4825#else
4826static inline void register_as_ext2(void) { }
4827static inline void unregister_as_ext2(void) { }
2035e776 4828static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4829#endif
4830
37f328eb 4831#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4832static inline void register_as_ext3(void)
4833{
4834 int err = register_filesystem(&ext3_fs_type);
4835 if (err)
4836 printk(KERN_WARNING
4837 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4838}
4839
4840static inline void unregister_as_ext3(void)
4841{
4842 unregister_filesystem(&ext3_fs_type);
4843}
2035e776
TT		 4844
4845static inline int ext3_feature_set_ok(struct super_block *sb)
4846{
4847 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4848 return 0;
4849 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4850 return 0;
4851 if (sb->s_flags & MS_RDONLY)
4852 return 1;
4853 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4854 return 0;
4855 return 1;
4856}
51b7e3c9 4857MODULE_ALIAS("ext3");
24b58424
TT		 4858#else
4859static inline void register_as_ext3(void) { }
4860static inline void unregister_as_ext3(void) { }
2035e776 4861static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4862#endif
4863
03010a33
TT		 4864static struct file_system_type ext4_fs_type = {
4865 .owner = THIS_MODULE,
4866 .name = "ext4",
152a0836 4867 .mount = ext4_mount,
03010a33
TT		 4868 .kill_sb = kill_block_super,
4869 .fs_flags = FS_REQUIRES_DEV,
4870};
4871
8f021222 4872static int __init ext4_init_feat_adverts(void)
857ac889
LC		 4873{
4874 struct ext4_features *ef;
4875 int ret = -ENOMEM;
4876
4877 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4878 if (!ef)
4879 goto out;
4880
4881 ef->f_kobj.kset = ext4_kset;
4882 init_completion(&ef->f_kobj_unregister);
4883 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4884 "features");
4885 if (ret) {
4886 kfree(ef);
4887 goto out;
4888 }
4889
4890 ext4_feat = ef;
4891 ret = 0;
4892out:
4893 return ret;
4894}
4895
8f021222
LC		 4896static void ext4_exit_feat_adverts(void)
4897{
4898 kobject_put(&ext4_feat->f_kobj);
4899 wait_for_completion(&ext4_feat->f_kobj_unregister);
4900 kfree(ext4_feat);
4901}
4902
e9e3bcec
ES		 4903/* Shared across all ext4 file systems */
4904wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4905struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4906
5dabfc78 4907static int __init ext4_init_fs(void)
ac27a0ec 4908{
e9e3bcec 4909 int i, err;
c9de560d 4910
07c0c5d8
AV		 4911 ext4_li_info = NULL;
4912 mutex_init(&ext4_li_mtx);
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 4951
ac27a0ec
DK		 4952 return 0;
4953out:
24b58424
TT		 4954 unregister_as_ext2();
4955 unregister_as_ext3();
ac27a0ec
DK		 4956 destroy_inodecache();
4957out1:
5dabfc78 4958 ext4_exit_xattr();
c9de560d 4959out2:
5dabfc78 4960 ext4_exit_mballoc();
6fd058f7 4961out3:
8f021222 4962 ext4_exit_feat_adverts();
dd68314c 4963out4:
d44651d0
FJ		 4964 if (ext4_proc_root)
4965 remove_proc_entry("fs/ext4", NULL);
6fd058f7 4966 kset_unregister(ext4_kset);
d44651d0 4967out5:
5dabfc78 4968 ext4_exit_system_zone();
d44651d0 4969out6:
5dabfc78 4970 ext4_exit_pageio();
ac27a0ec
DK		 4971 return err;
4972}
4973
5dabfc78 4974static void __exit ext4_exit_fs(void)
ac27a0ec 4975{
bfff6873 4976 ext4_destroy_lazyinit_thread();
24b58424
TT		 4977 unregister_as_ext2();
4978 unregister_as_ext3();
03010a33 4979 unregister_filesystem(&ext4_fs_type);
ac27a0ec 4980 destroy_inodecache();
5dabfc78
TT		 4981 ext4_exit_xattr();
4982 ext4_exit_mballoc();
8f021222 4983 ext4_exit_feat_adverts();
9f6200bb 4984 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4985 kset_unregister(ext4_kset);
5dabfc78
TT		 4986 ext4_exit_system_zone();
4987 ext4_exit_pageio();
ac27a0ec
DK		 4988}
4989
4990MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4991MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4992MODULE_LICENSE("GPL");
5dabfc78
TT		 4993module_init(ext4_init_fs)
4994module_exit(ext4_exit_fs)
Linux kernel - Deliverables
This page took 1.049518 seconds and 5 git commands to generate.