projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: record the checksum algorithm in use in the superblock
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
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
d25425f8
DW		 115static int ext4_verify_csum_type(struct super_block *sb,
116 struct ext4_super_block *es)
117{
118 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
119 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
120 return 1;
121
122 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
123}
124
9933fc0a
TT		 125void *ext4_kvmalloc(size_t size, gfp_t flags)
126{
127 void *ret;
128
129 ret = kmalloc(size, flags);
130 if (!ret)
131 ret = __vmalloc(size, flags, PAGE_KERNEL);
132 return ret;
133}
134
135void *ext4_kvzalloc(size_t size, gfp_t flags)
136{
137 void *ret;
138
db9481c0 139 ret = kzalloc(size, flags);
9933fc0a
TT		 140 if (!ret)
141 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
142 return ret;
143}
144
145void ext4_kvfree(void *ptr)
146{
147 if (is_vmalloc_addr(ptr))
148 vfree(ptr);
149 else
150 kfree(ptr);
151
152}
153
8fadc143
AR		 154ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
155 struct ext4_group_desc *bg)
bd81d8ee 156{
3a14589c 157 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 158 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 159 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 160}
161
8fadc143
AR		 162ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
163 struct ext4_group_desc *bg)
bd81d8ee 164{
5272f837 165 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 166 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 167 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 168}
169
8fadc143
AR		 170ext4_fsblk_t ext4_inode_table(struct super_block *sb,
171 struct ext4_group_desc *bg)
bd81d8ee 172{
5272f837 173 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 174 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 175 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 176}
177
021b65bb
TT		 178__u32 ext4_free_group_clusters(struct super_block *sb,
179 struct ext4_group_desc *bg)
560671a0
AK		 180{
181 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
182 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 183 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 184}
185
186__u32 ext4_free_inodes_count(struct super_block *sb,
187 struct ext4_group_desc *bg)
188{
189 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 191 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 192}
193
194__u32 ext4_used_dirs_count(struct super_block *sb,
195 struct ext4_group_desc *bg)
196{
197 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 199 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 200}
201
202__u32 ext4_itable_unused_count(struct super_block *sb,
203 struct ext4_group_desc *bg)
204{
205 return le16_to_cpu(bg->bg_itable_unused_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 207 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 208}
209
8fadc143
AR		 210void ext4_block_bitmap_set(struct super_block *sb,
211 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 212{
3a14589c 213 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 214 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
215 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 216}
217
8fadc143
AR		 218void ext4_inode_bitmap_set(struct super_block *sb,
219 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 220{
5272f837 221 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 222 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
223 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 224}
225
8fadc143
AR		 226void ext4_inode_table_set(struct super_block *sb,
227 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 228{
5272f837 229 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 230 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
231 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 232}
233
021b65bb
TT		 234void ext4_free_group_clusters_set(struct super_block *sb,
235 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 236{
237 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
238 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
239 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
240}
241
242void ext4_free_inodes_set(struct super_block *sb,
243 struct ext4_group_desc *bg, __u32 count)
244{
245 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
248}
249
250void ext4_used_dirs_set(struct super_block *sb,
251 struct ext4_group_desc *bg, __u32 count)
252{
253 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
256}
257
258void ext4_itable_unused_set(struct super_block *sb,
259 struct ext4_group_desc *bg, __u32 count)
260{
261 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
264}
265
d3d1faf6
CW		 266
267/* Just increment the non-pointer handle value */
268static handle_t *ext4_get_nojournal(void)
269{
270 handle_t *handle = current->journal_info;
271 unsigned long ref_cnt = (unsigned long)handle;
272
273 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
274
275 ref_cnt++;
276 handle = (handle_t *)ref_cnt;
277
278 current->journal_info = handle;
279 return handle;
280}
281
282
283/* Decrement the non-pointer handle value */
284static void ext4_put_nojournal(handle_t *handle)
285{
286 unsigned long ref_cnt = (unsigned long)handle;
287
288 BUG_ON(ref_cnt == 0);
289
290 ref_cnt--;
291 handle = (handle_t *)ref_cnt;
292
293 current->journal_info = handle;
294}
295
ac27a0ec 296/*
dab291af 297 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 298 *
299 * The only special thing we need to do here is to make sure that all
300 * journal_end calls result in the superblock being marked dirty, so
301 * that sync() will call the filesystem's write_super callback if
302 * appropriate.
be4f27d3
YY		 303 *
304 * To avoid j_barrier hold in userspace when a user calls freeze(),
305 * ext4 prevents a new handle from being started by s_frozen, which
306 * is in an upper layer.
ac27a0ec 307 */
617ba13b 308handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 309{
310 journal_t *journal;
be4f27d3 311 handle_t *handle;
ac27a0ec 312
12706394 313 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
ac27a0ec
DK		 314 if (sb->s_flags & MS_RDONLY)
315 return ERR_PTR(-EROFS);
316
617ba13b 317 journal = EXT4_SB(sb)->s_journal;
be4f27d3
YY		 318 handle = ext4_journal_current_handle();
319
320 /*
321 * If a handle has been started, it should be allowed to
322 * finish, otherwise deadlock could happen between freeze
323 * and others(e.g. truncate) due to the restart of the
324 * journal handle if the filesystem is forzen and active
325 * handles are not stopped.
326 */
327 if (!handle)
328 vfs_check_frozen(sb, SB_FREEZE_TRANS);
329
330 if (!journal)
331 return ext4_get_nojournal();
332 /*
333 * Special case here: if the journal has aborted behind our
334 * backs (eg. EIO in the commit thread), then we still need to
335 * take the FS itself readonly cleanly.
336 */
337 if (is_journal_aborted(journal)) {
338 ext4_abort(sb, "Detected aborted journal");
339 return ERR_PTR(-EROFS);
ac27a0ec 340 }
be4f27d3 341 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 342}
343
344/*
345 * The only special thing we need to do here is to make sure that all
dab291af 346 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 347 * that sync() will call the filesystem's write_super callback if
348 * appropriate.
349 */
c398eda0 350int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 351{
352 struct super_block *sb;
353 int err;
354 int rc;
355
0390131b 356 if (!ext4_handle_valid(handle)) {
d3d1faf6 357 ext4_put_nojournal(handle);
0390131b
FM		 358 return 0;
359 }
ac27a0ec
DK		 360 sb = handle->h_transaction->t_journal->j_private;
361 err = handle->h_err;
dab291af 362 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 363
364 if (!err)
365 err = rc;
366 if (err)
c398eda0 367 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 368 return err;
369}
370
90c7201b
TT		 371void ext4_journal_abort_handle(const char *caller, unsigned int line,
372 const char *err_fn, struct buffer_head *bh,
373 handle_t *handle, int err)
ac27a0ec
DK		 374{
375 char nbuf[16];
617ba13b 376 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 377
0390131b
FM		 378 BUG_ON(!ext4_handle_valid(handle));
379
ac27a0ec
DK		 380 if (bh)
381 BUFFER_TRACE(bh, "abort");
382
383 if (!handle->h_err)
384 handle->h_err = err;
385
386 if (is_handle_aborted(handle))
387 return;
388
92b97816 389 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
90c7201b 390 caller, line, errstr, err_fn);
ac27a0ec 391
dab291af 392 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 393}
394
1c13d5c0
TT		 395static void __save_error_info(struct super_block *sb, const char *func,
396 unsigned int line)
397{
398 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
399
400 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
401 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
402 es->s_last_error_time = cpu_to_le32(get_seconds());
403 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
404 es->s_last_error_line = cpu_to_le32(line);
405 if (!es->s_first_error_time) {
406 es->s_first_error_time = es->s_last_error_time;
407 strncpy(es->s_first_error_func, func,
408 sizeof(es->s_first_error_func));
409 es->s_first_error_line = cpu_to_le32(line);
410 es->s_first_error_ino = es->s_last_error_ino;
411 es->s_first_error_block = es->s_last_error_block;
412 }
66e61a9e
TT		 413 /*
414 * Start the daily error reporting function if it hasn't been
415 * started already
416 */
417 if (!es->s_error_count)
418 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
1c13d5c0
TT		 419 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
420}
421
422static void save_error_info(struct super_block *sb, const char *func,
423 unsigned int line)
424{
425 __save_error_info(sb, func, line);
426 ext4_commit_super(sb, 1);
427}
428
7c2e7087
TT		 429/*
430 * The del_gendisk() function uninitializes the disk-specific data
431 * structures, including the bdi structure, without telling anyone
432 * else. Once this happens, any attempt to call mark_buffer_dirty()
433 * (for example, by ext4_commit_super), will cause a kernel OOPS.
434 * This is a kludge to prevent these oops until we can put in a proper
435 * hook in del_gendisk() to inform the VFS and file system layers.
436 */
437static int block_device_ejected(struct super_block *sb)
438{
439 struct inode *bd_inode = sb->s_bdev->bd_inode;
440 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
441
442 return bdi->dev == NULL;
443}
444
18aadd47
BJ		 445static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
446{
447 struct super_block *sb = journal->j_private;
448 struct ext4_sb_info *sbi = EXT4_SB(sb);
449 int error = is_journal_aborted(journal);
450 struct ext4_journal_cb_entry *jce, *tmp;
451
452 spin_lock(&sbi->s_md_lock);
453 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
454 list_del_init(&jce->jce_list);
455 spin_unlock(&sbi->s_md_lock);
456 jce->jce_func(sb, jce, error);
457 spin_lock(&sbi->s_md_lock);
458 }
459 spin_unlock(&sbi->s_md_lock);
460}
1c13d5c0 461
ac27a0ec
DK		 462/* Deal with the reporting of failure conditions on a filesystem such as
463 * inconsistencies detected or read IO failures.
464 *
465 * On ext2, we can store the error state of the filesystem in the
617ba13b 466 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 467 * write ordering constraints on the superblock which prevent us from
468 * writing it out straight away; and given that the journal is about to
469 * be aborted, we can't rely on the current, or future, transactions to
470 * write out the superblock safely.
471 *
dab291af 472 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 473 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 474 * that error until we've noted it down and cleared it.
475 */
476
617ba13b 477static void ext4_handle_error(struct super_block *sb)
ac27a0ec 478{
ac27a0ec
DK		 479 if (sb->s_flags & MS_RDONLY)
480 return;
481
2b2d6d01 482 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 483 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 484
4ab2f15b 485 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 486 if (journal)
dab291af 487 jbd2_journal_abort(journal, -EIO);
ac27a0ec 488 }
2b2d6d01 489 if (test_opt(sb, ERRORS_RO)) {
b31e1552 490 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 491 sb->s_flags |= MS_RDONLY;
492 }
ac27a0ec 493 if (test_opt(sb, ERRORS_PANIC))
617ba13b 494 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 495 sb->s_id);
496}
497
12062ddd 498void __ext4_error(struct super_block *sb, const char *function,
c398eda0 499 unsigned int line, const char *fmt, ...)
ac27a0ec 500{
0ff2ea7d 501 struct va_format vaf;
ac27a0ec
DK		 502 va_list args;
503
504 va_start(args, fmt);
0ff2ea7d
JP		 505 vaf.fmt = fmt;
506 vaf.va = &args;
507 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
508 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec
DK		 509 va_end(args);
510
617ba13b 511 ext4_handle_error(sb);
ac27a0ec
DK		 512}
513
c398eda0
TT		 514void ext4_error_inode(struct inode *inode, const char *function,
515 unsigned int line, ext4_fsblk_t block,
273df556
FM		 516 const char *fmt, ...)
517{
518 va_list args;
f7c21177 519 struct va_format vaf;
1c13d5c0 520 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 521
1c13d5c0
TT		 522 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
523 es->s_last_error_block = cpu_to_le64(block);
524 save_error_info(inode->i_sb, function, line);
273df556 525 va_start(args, fmt);
f7c21177
TT		 526 vaf.fmt = fmt;
527 vaf.va = &args;
c398eda0 528 if (block)
d9ee81da
JP		 529 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
530 "inode #%lu: block %llu: comm %s: %pV\n",
531 inode->i_sb->s_id, function, line, inode->i_ino,
532 block, current->comm, &vaf);
533 else
534 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
535 "inode #%lu: comm %s: %pV\n",
536 inode->i_sb->s_id, function, line, inode->i_ino,
537 current->comm, &vaf);
273df556
FM		 538 va_end(args);
539
540 ext4_handle_error(inode->i_sb);
541}
542
c398eda0 543void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 544 unsigned int line, ext4_fsblk_t block,
545 const char *fmt, ...)
273df556
FM		 546{
547 va_list args;
f7c21177 548 struct va_format vaf;
1c13d5c0 549 struct ext4_super_block *es;
273df556
FM		 550 struct inode *inode = file->f_dentry->d_inode;
551 char pathname[80], *path;
552
1c13d5c0
TT		 553 es = EXT4_SB(inode->i_sb)->s_es;
554 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
555 save_error_info(inode->i_sb, function, line);
273df556 556 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 557 if (IS_ERR(path))
273df556 558 path = "(unknown)";
f7c21177
TT		 559 va_start(args, fmt);
560 vaf.fmt = fmt;
561 vaf.va = &args;
d9ee81da
JP		 562 if (block)
563 printk(KERN_CRIT
564 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
565 "block %llu: comm %s: path %s: %pV\n",
566 inode->i_sb->s_id, function, line, inode->i_ino,
567 block, current->comm, path, &vaf);
568 else
569 printk(KERN_CRIT
570 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
571 "comm %s: path %s: %pV\n",
572 inode->i_sb->s_id, function, line, inode->i_ino,
573 current->comm, path, &vaf);
273df556
FM		 574 va_end(args);
575
576 ext4_handle_error(inode->i_sb);
577}
578
2b2d6d01 579static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 580 char nbuf[16])
581{
582 char *errstr = NULL;
583
584 switch (errno) {
585 case -EIO:
586 errstr = "IO failure";
587 break;
588 case -ENOMEM:
589 errstr = "Out of memory";
590 break;
591 case -EROFS:
78f1ddbb
TT		 592 if (!sb || (EXT4_SB(sb)->s_journal &&
593 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 594 errstr = "Journal has aborted";
595 else
596 errstr = "Readonly filesystem";
597 break;
598 default:
599 /* If the caller passed in an extra buffer for unknown
600 * errors, textualise them now. Else we just return
601 * NULL. */
602 if (nbuf) {
603 /* Check for truncated error codes... */
604 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
605 errstr = nbuf;
606 }
607 break;
608 }
609
610 return errstr;
611}
612
617ba13b 613/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 614 * automatically and invokes the appropriate error response. */
615
c398eda0
TT		 616void __ext4_std_error(struct super_block *sb, const char *function,
617 unsigned int line, int errno)
ac27a0ec
DK		 618{
619 char nbuf[16];
620 const char *errstr;
621
622 /* Special case: if the error is EROFS, and we're not already
623 * inside a transaction, then there's really no point in logging
624 * an error. */
625 if (errno == -EROFS && journal_current_handle() == NULL &&
626 (sb->s_flags & MS_RDONLY))
627 return;
628
617ba13b 629 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 630 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
631 sb->s_id, function, line, errstr);
1c13d5c0 632 save_error_info(sb, function, line);
ac27a0ec 633
617ba13b 634 ext4_handle_error(sb);
ac27a0ec
DK		 635}
636
637/*
617ba13b 638 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 639 * abort function may be used to deal with unrecoverable failures such
640 * as journal IO errors or ENOMEM at a critical moment in log management.
641 *
642 * We unconditionally force the filesystem into an ABORT|READONLY state,
643 * unless the error response on the fs has been set to panic in which
644 * case we take the easy way out and panic immediately.
645 */
646
c67d859e 647void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 648 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 649{
650 va_list args;
651
1c13d5c0 652 save_error_info(sb, function, line);
ac27a0ec 653 va_start(args, fmt);
c398eda0
TT		 654 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
655 function, line);
ac27a0ec
DK		 656 vprintk(fmt, args);
657 printk("\n");
658 va_end(args);
659
1c13d5c0
TT		 660 if ((sb->s_flags & MS_RDONLY) == 0) {
661 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
662 sb->s_flags |= MS_RDONLY;
663 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
664 if (EXT4_SB(sb)->s_journal)
665 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
666 save_error_info(sb, function, line);
667 }
ac27a0ec 668 if (test_opt(sb, ERRORS_PANIC))
617ba13b 669 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 670}
671
0ff2ea7d 672void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 673{
0ff2ea7d 674 struct va_format vaf;
b31e1552
ES		 675 va_list args;
676
677 va_start(args, fmt);
0ff2ea7d
JP		 678 vaf.fmt = fmt;
679 vaf.va = &args;
680 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 681 va_end(args);
682}
683
12062ddd 684void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 685 unsigned int line, const char *fmt, ...)
ac27a0ec 686{
0ff2ea7d 687 struct va_format vaf;
ac27a0ec
DK		 688 va_list args;
689
690 va_start(args, fmt);
0ff2ea7d
JP		 691 vaf.fmt = fmt;
692 vaf.va = &args;
693 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
694 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 695 va_end(args);
696}
697
e29136f8
TT		 698void __ext4_grp_locked_error(const char *function, unsigned int line,
699 struct super_block *sb, ext4_group_t grp,
700 unsigned long ino, ext4_fsblk_t block,
701 const char *fmt, ...)
5d1b1b3f
AK		 702__releases(bitlock)
703__acquires(bitlock)
704{
0ff2ea7d 705 struct va_format vaf;
5d1b1b3f
AK		 706 va_list args;
707 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
708
1c13d5c0
TT		 709 es->s_last_error_ino = cpu_to_le32(ino);
710 es->s_last_error_block = cpu_to_le64(block);
711 __save_error_info(sb, function, line);
0ff2ea7d 712
5d1b1b3f 713 va_start(args, fmt);
0ff2ea7d
JP		 714
715 vaf.fmt = fmt;
716 vaf.va = &args;
21149d61 717 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 718 sb->s_id, function, line, grp);
719 if (ino)
0ff2ea7d 720 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 721 if (block)
0ff2ea7d
JP		 722 printk(KERN_CONT "block %llu:", (unsigned long long) block);
723 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 724 va_end(args);
725
726 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 727 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 728 return;
729 }
1c13d5c0 730
5d1b1b3f
AK		 731 ext4_unlock_group(sb, grp);
732 ext4_handle_error(sb);
733 /*
734 * We only get here in the ERRORS_RO case; relocking the group
735 * may be dangerous, but nothing bad will happen since the
736 * filesystem will have already been marked read/only and the
737 * journal has been aborted. We return 1 as a hint to callers
738 * who might what to use the return value from
25985edc 739 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 740 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
741 * aggressively from the ext4 function in question, with a
742 * more appropriate error code.
743 */
744 ext4_lock_group(sb, grp);
745 return;
746}
747
617ba13b 748void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 749{
617ba13b 750 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 751
617ba13b 752 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 753 return;
754
12062ddd 755 ext4_warning(sb,
ac27a0ec
DK		 756 "updating to rev %d because of new feature flag, "
757 "running e2fsck is recommended",
617ba13b 758 EXT4_DYNAMIC_REV);
ac27a0ec 759
617ba13b
MC		 760 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
761 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
762 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 763 /* leave es->s_feature_*compat flags alone */
764 /* es->s_uuid will be set by e2fsck if empty */
765
766 /*
767 * The rest of the superblock fields should be zero, and if not it
768 * means they are likely already in use, so leave them alone. We
769 * can leave it up to e2fsck to clean up any inconsistencies there.
770 */
771}
772
773/*
774 * Open the external journal device
775 */
b31e1552 776static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 777{
778 struct block_device *bdev;
779 char b[BDEVNAME_SIZE];
780
d4d77629 781 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 782 if (IS_ERR(bdev))
783 goto fail;
784 return bdev;
785
786fail:
b31e1552 787 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 788 __bdevname(dev, b), PTR_ERR(bdev));
789 return NULL;
790}
791
792/*
793 * Release the journal device
794 */
617ba13b 795static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 796{
e525fd89 797 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 798}
799
617ba13b 800static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 801{
802 struct block_device *bdev;
803 int ret = -ENODEV;
804
805 bdev = sbi->journal_bdev;
806 if (bdev) {
617ba13b 807 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 808 sbi->journal_bdev = NULL;
809 }
810 return ret;
811}
812
813static inline struct inode *orphan_list_entry(struct list_head *l)
814{
617ba13b 815 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 816}
817
617ba13b 818static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 819{
820 struct list_head *l;
821
b31e1552
ES		 822 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
823 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 824
825 printk(KERN_ERR "sb_info orphan list:\n");
826 list_for_each(l, &sbi->s_orphan) {
827 struct inode *inode = orphan_list_entry(l);
828 printk(KERN_ERR " "
829 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
830 inode->i_sb->s_id, inode->i_ino, inode,
831 inode->i_mode, inode->i_nlink,
832 NEXT_ORPHAN(inode));
833 }
834}
835
2b2d6d01 836static void ext4_put_super(struct super_block *sb)
ac27a0ec 837{
617ba13b
MC		 838 struct ext4_sb_info *sbi = EXT4_SB(sb);
839 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 840 int i, err;
ac27a0ec 841
857ac889 842 ext4_unregister_li_request(sb);
e0ccfd95
CH		 843 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
844
4c0425ff
MC		 845 flush_workqueue(sbi->dio_unwritten_wq);
846 destroy_workqueue(sbi->dio_unwritten_wq);
847
a9e220f8 848 lock_super(sb);
0390131b
FM		 849 if (sbi->s_journal) {
850 err = jbd2_journal_destroy(sbi->s_journal);
851 sbi->s_journal = NULL;
852 if (err < 0)
c67d859e 853 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 854 }
d4edac31 855
a1c6c569 856 del_timer(&sbi->s_err_report);
d4edac31
JB		 857 ext4_release_system_zone(sb);
858 ext4_mb_release(sb);
859 ext4_ext_release(sb);
860 ext4_xattr_put_super(sb);
861
ac27a0ec 862 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 863 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 864 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 865 }
a8e25a83
AB		 866 if (sb->s_dirt || !(sb->s_flags & MS_RDONLY))
867 ext4_commit_super(sb, 1);
868
240799cd 869 if (sbi->s_proc) {
66acdcf4 870 remove_proc_entry("options", sbi->s_proc);
9f6200bb 871 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 872 }
3197ebdb 873 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 874
875 for (i = 0; i < sbi->s_gdb_count; i++)
876 brelse(sbi->s_group_desc[i]);
f18a5f21 877 ext4_kvfree(sbi->s_group_desc);
9933fc0a 878 ext4_kvfree(sbi->s_flex_groups);
57042651 879 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 880 percpu_counter_destroy(&sbi->s_freeinodes_counter);
881 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 882 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 883 brelse(sbi->s_sbh);
884#ifdef CONFIG_QUOTA
885 for (i = 0; i < MAXQUOTAS; i++)
886 kfree(sbi->s_qf_names[i]);
887#endif
888
889 /* Debugging code just in case the in-memory inode orphan list
890 * isn't empty. The on-disk one can be non-empty if we've
891 * detected an error and taken the fs readonly, but the
892 * in-memory list had better be clean by this point. */
893 if (!list_empty(&sbi->s_orphan))
894 dump_orphan_list(sb, sbi);
895 J_ASSERT(list_empty(&sbi->s_orphan));
896
f98393a6 897 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 898 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
899 /*
900 * Invalidate the journal device's buffers. We don't want them
901 * floating about in memory - the physical journal device may
902 * hotswapped, and it breaks the `ro-after' testing code.
903 */
904 sync_blockdev(sbi->journal_bdev);
f98393a6 905 invalidate_bdev(sbi->journal_bdev);
617ba13b 906 ext4_blkdev_remove(sbi);
ac27a0ec 907 }
c5e06d10
JL		 908 if (sbi->s_mmp_tsk)
909 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 910 sb->s_fs_info = NULL;
3197ebdb
TT		 911 /*
912 * Now that we are completely done shutting down the
913 * superblock, we need to actually destroy the kobject.
914 */
3197ebdb
TT		 915 unlock_super(sb);
916 kobject_put(&sbi->s_kobj);
917 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 918 kfree(sbi->s_blockgroup_lock);
ac27a0ec 919 kfree(sbi);
ac27a0ec
DK		 920}
921
e18b890b 922static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 923
924/*
925 * Called inside transaction, so use GFP_NOFS
926 */
617ba13b 927static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 928{
617ba13b 929 struct ext4_inode_info *ei;
ac27a0ec 930
e6b4f8da 931 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 932 if (!ei)
933 return NULL;
0b8e58a1 934
ac27a0ec 935 ei->vfs_inode.i_version = 1;
91246c00 936 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 937 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 938 INIT_LIST_HEAD(&ei->i_prealloc_list);
939 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 940 ei->i_reserved_data_blocks = 0;
941 ei->i_reserved_meta_blocks = 0;
942 ei->i_allocated_meta_blocks = 0;
9d0be502 943 ei->i_da_metadata_calc_len = 0;
d2a17637 944 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 945#ifdef CONFIG_QUOTA
946 ei->i_reserved_quota = 0;
947#endif
8aefcd55 948 ei->jinode = NULL;
c7064ef1 949 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 950 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 951 ei->cur_aio_dio = NULL;
b436b9be
JK		 952 ei->i_sync_tid = 0;
953 ei->i_datasync_tid = 0;
f7ad6d2e 954 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 955 atomic_set(&ei->i_aiodio_unwritten, 0);
0b8e58a1 956
ac27a0ec
DK		 957 return &ei->vfs_inode;
958}
959
7ff9c073
TT		 960static int ext4_drop_inode(struct inode *inode)
961{
962 int drop = generic_drop_inode(inode);
963
964 trace_ext4_drop_inode(inode, drop);
965 return drop;
966}
967
fa0d7e3d
NP		 968static void ext4_i_callback(struct rcu_head *head)
969{
970 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 971 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
972}
973
617ba13b 974static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 975{
9f7dd93d 976 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 977 ext4_msg(inode->i_sb, KERN_ERR,
978 "Inode %lu (%p): orphan list check failed!",
979 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 980 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
981 EXT4_I(inode), sizeof(struct ext4_inode_info),
982 true);
983 dump_stack();
984 }
fa0d7e3d 985 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 986}
987
51cc5068 988static void init_once(void *foo)
ac27a0ec 989{
617ba13b 990 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 991
a35afb83 992 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 993#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 994 init_rwsem(&ei->xattr_sem);
ac27a0ec 995#endif
0e855ac8 996 init_rwsem(&ei->i_data_sem);
a35afb83 997 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 998}
999
1000static int init_inodecache(void)
1001{
617ba13b
MC		 1002 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1003 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 1004 0, (SLAB_RECLAIM_ACCOUNT|
1005 SLAB_MEM_SPREAD),
20c2df83 1006 init_once);
617ba13b 1007 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1008 return -ENOMEM;
1009 return 0;
1010}
1011
1012static void destroy_inodecache(void)
1013{
617ba13b 1014 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1015}
1016
0930fcc1 1017void ext4_clear_inode(struct inode *inode)
ac27a0ec 1018{
0930fcc1
AV		 1019 invalidate_inode_buffers(inode);
1020 end_writeback(inode);
9f754758 1021 dquot_drop(inode);
c2ea3fde 1022 ext4_discard_preallocations(inode);
8aefcd55
TT		 1023 if (EXT4_I(inode)->jinode) {
1024 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1025 EXT4_I(inode)->jinode);
1026 jbd2_free_inode(EXT4_I(inode)->jinode);
1027 EXT4_I(inode)->jinode = NULL;
1028 }
ac27a0ec
DK		 1029}
1030
1b961ac0 1031static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1032 u64 ino, u32 generation)
ac27a0ec 1033{
ac27a0ec 1034 struct inode *inode;
ac27a0ec 1035
617ba13b 1036 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1037 return ERR_PTR(-ESTALE);
617ba13b 1038 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1039 return ERR_PTR(-ESTALE);
1040
1041 /* iget isn't really right if the inode is currently unallocated!!
1042 *
617ba13b 1043 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1044 * deleted, so we should be safe.
1045 *
1046 * Currently we don't know the generation for parent directory, so
1047 * a generation of 0 means "accept any"
1048 */
1d1fe1ee
DH		 1049 inode = ext4_iget(sb, ino);
1050 if (IS_ERR(inode))
1051 return ERR_CAST(inode);
1052 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1053 iput(inode);
1054 return ERR_PTR(-ESTALE);
1055 }
1b961ac0
CH		 1056
1057 return inode;
1058}
1059
1060static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1061 int fh_len, int fh_type)
1b961ac0
CH		 1062{
1063 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1064 ext4_nfs_get_inode);
1065}
1066
1067static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1068 int fh_len, int fh_type)
1b961ac0
CH		 1069{
1070 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1071 ext4_nfs_get_inode);
ac27a0ec
DK		 1072}
1073
c39a7f84
TO		 1074/*
1075 * Try to release metadata pages (indirect blocks, directories) which are
1076 * mapped via the block device. Since these pages could have journal heads
1077 * which would prevent try_to_free_buffers() from freeing them, we must use
1078 * jbd2 layer's try_to_free_buffers() function to release them.
1079 */
0b8e58a1
AD		 1080static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1081 gfp_t wait)
c39a7f84
TO		 1082{
1083 journal_t *journal = EXT4_SB(sb)->s_journal;
1084
1085 WARN_ON(PageChecked(page));
1086 if (!page_has_buffers(page))
1087 return 0;
1088 if (journal)
1089 return jbd2_journal_try_to_free_buffers(journal, page,
1090 wait & ~__GFP_WAIT);
1091 return try_to_free_buffers(page);
1092}
1093
ac27a0ec 1094#ifdef CONFIG_QUOTA
af5bc92d 1095#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1096#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1097
617ba13b
MC		 1098static int ext4_write_dquot(struct dquot *dquot);
1099static int ext4_acquire_dquot(struct dquot *dquot);
1100static int ext4_release_dquot(struct dquot *dquot);
1101static int ext4_mark_dquot_dirty(struct dquot *dquot);
1102static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1103static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1104 struct path *path);
ca0e05e4 1105static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1106static int ext4_quota_on_mount(struct super_block *sb, int type);
1107static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1108 size_t len, loff_t off);
617ba13b 1109static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1110 const char *data, size_t len, loff_t off);
1111
61e225dc 1112static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1113 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1114 .write_dquot = ext4_write_dquot,
1115 .acquire_dquot = ext4_acquire_dquot,
1116 .release_dquot = ext4_release_dquot,
1117 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1118 .write_info = ext4_write_info,
1119 .alloc_dquot = dquot_alloc,
1120 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1121};
1122
0d54b217 1123static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1124 .quota_on = ext4_quota_on,
ca0e05e4 1125 .quota_off = ext4_quota_off,
287a8095
CH		 1126 .quota_sync = dquot_quota_sync,
1127 .get_info = dquot_get_dqinfo,
1128 .set_info = dquot_set_dqinfo,
1129 .get_dqblk = dquot_get_dqblk,
1130 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1131};
1132#endif
1133
ee9b6d61 1134static const struct super_operations ext4_sops = {
617ba13b
MC		 1135 .alloc_inode = ext4_alloc_inode,
1136 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1137 .write_inode = ext4_write_inode,
1138 .dirty_inode = ext4_dirty_inode,
7ff9c073 1139 .drop_inode = ext4_drop_inode,
0930fcc1 1140 .evict_inode = ext4_evict_inode,
617ba13b 1141 .put_super = ext4_put_super,
617ba13b 1142 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1143 .freeze_fs = ext4_freeze,
1144 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1145 .statfs = ext4_statfs,
1146 .remount_fs = ext4_remount,
617ba13b 1147 .show_options = ext4_show_options,
ac27a0ec 1148#ifdef CONFIG_QUOTA
617ba13b
MC		 1149 .quota_read = ext4_quota_read,
1150 .quota_write = ext4_quota_write,
ac27a0ec 1151#endif
c39a7f84 1152 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1153};
1154
9ca92389
TT		 1155static const struct super_operations ext4_nojournal_sops = {
1156 .alloc_inode = ext4_alloc_inode,
1157 .destroy_inode = ext4_destroy_inode,
1158 .write_inode = ext4_write_inode,
1159 .dirty_inode = ext4_dirty_inode,
7ff9c073 1160 .drop_inode = ext4_drop_inode,
0930fcc1 1161 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1162 .write_super = ext4_write_super,
1163 .put_super = ext4_put_super,
1164 .statfs = ext4_statfs,
1165 .remount_fs = ext4_remount,
9ca92389
TT		 1166 .show_options = ext4_show_options,
1167#ifdef CONFIG_QUOTA
1168 .quota_read = ext4_quota_read,
1169 .quota_write = ext4_quota_write,
1170#endif
1171 .bdev_try_to_free_page = bdev_try_to_free_page,
1172};
1173
39655164 1174static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1175 .fh_to_dentry = ext4_fh_to_dentry,
1176 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1177 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1178};
1179
1180enum {
1181 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1182 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1183 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1184 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1185 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1186 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1187 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1188 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1189 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1190 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1191 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1192 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1193 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1194 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1195 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1196 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1197 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1198 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
ac27a0ec
DK		 1199};
1200
a447c093 1201static const match_table_t tokens = {
ac27a0ec
DK		 1202 {Opt_bsd_df, "bsddf"},
1203 {Opt_minix_df, "minixdf"},
1204 {Opt_grpid, "grpid"},
1205 {Opt_grpid, "bsdgroups"},
1206 {Opt_nogrpid, "nogrpid"},
1207 {Opt_nogrpid, "sysvgroups"},
1208 {Opt_resgid, "resgid=%u"},
1209 {Opt_resuid, "resuid=%u"},
1210 {Opt_sb, "sb=%u"},
1211 {Opt_err_cont, "errors=continue"},
1212 {Opt_err_panic, "errors=panic"},
1213 {Opt_err_ro, "errors=remount-ro"},
1214 {Opt_nouid32, "nouid32"},
ac27a0ec 1215 {Opt_debug, "debug"},
72578c33
TT		 1216 {Opt_removed, "oldalloc"},
1217 {Opt_removed, "orlov"},
ac27a0ec
DK		 1218 {Opt_user_xattr, "user_xattr"},
1219 {Opt_nouser_xattr, "nouser_xattr"},
1220 {Opt_acl, "acl"},
1221 {Opt_noacl, "noacl"},
e3bb52ae 1222 {Opt_noload, "norecovery"},
5a916be1 1223 {Opt_noload, "noload"},
72578c33
TT		 1224 {Opt_removed, "nobh"},
1225 {Opt_removed, "bh"},
ac27a0ec 1226 {Opt_commit, "commit=%u"},
30773840
TT		 1227 {Opt_min_batch_time, "min_batch_time=%u"},
1228 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1229 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1230 {Opt_journal_checksum, "journal_checksum"},
1231 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1232 {Opt_abort, "abort"},
1233 {Opt_data_journal, "data=journal"},
1234 {Opt_data_ordered, "data=ordered"},
1235 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1236 {Opt_data_err_abort, "data_err=abort"},
1237 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1238 {Opt_offusrjquota, "usrjquota="},
1239 {Opt_usrjquota, "usrjquota=%s"},
1240 {Opt_offgrpjquota, "grpjquota="},
1241 {Opt_grpjquota, "grpjquota=%s"},
1242 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1243 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1244 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1245 {Opt_grpquota, "grpquota"},
1246 {Opt_noquota, "noquota"},
1247 {Opt_quota, "quota"},
1248 {Opt_usrquota, "usrquota"},
1249 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1250 {Opt_barrier, "barrier"},
1251 {Opt_nobarrier, "nobarrier"},
25ec56b5 1252 {Opt_i_version, "i_version"},
c9de560d 1253 {Opt_stripe, "stripe=%u"},
64769240 1254 {Opt_delalloc, "delalloc"},
dd919b98 1255 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1256 {Opt_mblk_io_submit, "mblk_io_submit"},
1257 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1258 {Opt_block_validity, "block_validity"},
1259 {Opt_noblock_validity, "noblock_validity"},
240799cd 1260 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1261 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1262 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1263 {Opt_auto_da_alloc, "auto_da_alloc"},
1264 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1265 {Opt_dioread_nolock, "dioread_nolock"},
1266 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1267 {Opt_discard, "discard"},
1268 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1269 {Opt_init_itable, "init_itable=%u"},
1270 {Opt_init_itable, "init_itable"},
1271 {Opt_noinit_itable, "noinit_itable"},
c7198b9c
TT		 1272 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1273 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1274 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1275 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1276 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1277 {Opt_err, NULL},
ac27a0ec
DK		 1278};
1279
617ba13b 1280static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1281{
617ba13b 1282 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1283 char *options = (char *) *data;
1284
1285 if (!options || strncmp(options, "sb=", 3) != 0)
1286 return 1; /* Default location */
0b8e58a1 1287
ac27a0ec 1288 options += 3;
0b8e58a1 1289 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1290 sb_block = simple_strtoul(options, &options, 0);
1291 if (*options && *options != ',') {
4776004f 1292 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1293 (char *) *data);
1294 return 1;
1295 }
1296 if (*options == ',')
1297 options++;
1298 *data = (void *) options;
0b8e58a1 1299
ac27a0ec
DK		 1300 return sb_block;
1301}
1302
b3881f74 1303#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1304static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1305 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1306
56c50f11
DM		 1307#ifdef CONFIG_QUOTA
1308static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1309{
1310 struct ext4_sb_info *sbi = EXT4_SB(sb);
1311 char *qname;
1312
1313 if (sb_any_quota_loaded(sb) &&
1314 !sbi->s_qf_names[qtype]) {
1315 ext4_msg(sb, KERN_ERR,
1316 "Cannot change journaled "
1317 "quota options when quota turned on");
57f73c2c 1318 return -1;
56c50f11
DM		 1319 }
1320 qname = match_strdup(args);
1321 if (!qname) {
1322 ext4_msg(sb, KERN_ERR,
1323 "Not enough memory for storing quotafile name");
57f73c2c 1324 return -1;
56c50f11
DM		 1325 }
1326 if (sbi->s_qf_names[qtype] &&
1327 strcmp(sbi->s_qf_names[qtype], qname)) {
1328 ext4_msg(sb, KERN_ERR,
1329 "%s quota file already specified", QTYPE2NAME(qtype));
1330 kfree(qname);
57f73c2c 1331 return -1;
56c50f11
DM		 1332 }
1333 sbi->s_qf_names[qtype] = qname;
1334 if (strchr(sbi->s_qf_names[qtype], '/')) {
1335 ext4_msg(sb, KERN_ERR,
1336 "quotafile must be on filesystem root");
1337 kfree(sbi->s_qf_names[qtype]);
1338 sbi->s_qf_names[qtype] = NULL;
57f73c2c 1339 return -1;
56c50f11 1340 }
fd8c37ec 1341 set_opt(sb, QUOTA);
56c50f11
DM		 1342 return 1;
1343}
1344
1345static int clear_qf_name(struct super_block *sb, int qtype)
1346{
1347
1348 struct ext4_sb_info *sbi = EXT4_SB(sb);
1349
1350 if (sb_any_quota_loaded(sb) &&
1351 sbi->s_qf_names[qtype]) {
1352 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1353 " when quota turned on");
57f73c2c 1354 return -1;
56c50f11
DM		 1355 }
1356 /*
1357 * The space will be released later when all options are confirmed
1358 * to be correct
1359 */
1360 sbi->s_qf_names[qtype] = NULL;
1361 return 1;
1362}
1363#endif
1364
26092bf5
TT		 1365#define MOPT_SET 0x0001
1366#define MOPT_CLEAR 0x0002
1367#define MOPT_NOSUPPORT 0x0004
1368#define MOPT_EXPLICIT 0x0008
1369#define MOPT_CLEAR_ERR 0x0010
1370#define MOPT_GTE0 0x0020
ac27a0ec 1371#ifdef CONFIG_QUOTA
26092bf5
TT		 1372#define MOPT_Q 0
1373#define MOPT_QFMT 0x0040
1374#else
1375#define MOPT_Q MOPT_NOSUPPORT
1376#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1377#endif
26092bf5
TT		 1378#define MOPT_DATAJ 0x0080
1379
1380static const struct mount_opts {
1381 int token;
1382 int mount_opt;
1383 int flags;
1384} ext4_mount_opts[] = {
1385 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1386 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1387 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1388 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1389 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1390 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1391 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1392 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1393 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1394 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1395 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1396 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1397 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1398 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1399 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1400 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1401 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1402 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1403 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1404 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1405 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1406 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1407 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1408 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1409 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1410 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1411 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1412 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1413 {Opt_commit, 0, MOPT_GTE0},
1414 {Opt_max_batch_time, 0, MOPT_GTE0},
1415 {Opt_min_batch_time, 0, MOPT_GTE0},
1416 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1417 {Opt_init_itable, 0, MOPT_GTE0},
1418 {Opt_stripe, 0, MOPT_GTE0},
1419 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1420 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1421 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
03010a33 1422#ifdef CONFIG_EXT4_FS_XATTR
26092bf5
TT		 1423 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1424 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
ac27a0ec 1425#else
26092bf5
TT		 1426 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1427 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
ac27a0ec 1428#endif
03010a33 1429#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1430 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1431 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1432#else
26092bf5
TT		 1433 {Opt_acl, 0, MOPT_NOSUPPORT},
1434 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1435#endif
26092bf5
TT		 1436 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1437 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1438 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1439 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1440 MOPT_SET | MOPT_Q},
1441 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1442 MOPT_SET | MOPT_Q},
1443 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1444 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1445 {Opt_usrjquota, 0, MOPT_Q},
1446 {Opt_grpjquota, 0, MOPT_Q},
1447 {Opt_offusrjquota, 0, MOPT_Q},
1448 {Opt_offgrpjquota, 0, MOPT_Q},
1449 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1450 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1451 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1452 {Opt_err, 0, 0}
1453};
1454
1455static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1456 substring_t *args, unsigned long *journal_devnum,
1457 unsigned int *journal_ioprio, int is_remount)
1458{
1459 struct ext4_sb_info *sbi = EXT4_SB(sb);
1460 const struct mount_opts *m;
1461 int arg = 0;
1462
57f73c2c
TT		 1463#ifdef CONFIG_QUOTA
1464 if (token == Opt_usrjquota)
1465 return set_qf_name(sb, USRQUOTA, &args[0]);
1466 else if (token == Opt_grpjquota)
1467 return set_qf_name(sb, GRPQUOTA, &args[0]);
1468 else if (token == Opt_offusrjquota)
1469 return clear_qf_name(sb, USRQUOTA);
1470 else if (token == Opt_offgrpjquota)
1471 return clear_qf_name(sb, GRPQUOTA);
1472#endif
26092bf5
TT		 1473 if (args->from && match_int(args, &arg))
1474 return -1;
1475 switch (token) {
f7048605
TT		 1476 case Opt_noacl:
1477 case Opt_nouser_xattr:
1478 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1479 break;
26092bf5
TT		 1480 case Opt_sb:
1481 return 1; /* handled by get_sb_block() */
1482 case Opt_removed:
1483 ext4_msg(sb, KERN_WARNING,
1484 "Ignoring removed %s option", opt);
1485 return 1;
1486 case Opt_resuid:
1487 sbi->s_resuid = arg;
1488 return 1;
1489 case Opt_resgid:
1490 sbi->s_resgid = arg;
1491 return 1;
1492 case Opt_abort:
1493 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1494 return 1;
1495 case Opt_i_version:
1496 sb->s_flags |= MS_I_VERSION;
1497 return 1;
1498 case Opt_journal_dev:
1499 if (is_remount) {
1500 ext4_msg(sb, KERN_ERR,
1501 "Cannot specify journal on remount");
1502 return -1;
1503 }
1504 *journal_devnum = arg;
1505 return 1;
1506 case Opt_journal_ioprio:
1507 if (arg < 0 || arg > 7)
1508 return -1;
1509 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1510 return 1;
1511 }
1512
1513 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1514 if (token != m->token)
1515 continue;
1516 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1517 return -1;
1518 if (m->flags & MOPT_EXPLICIT)
1519 set_opt2(sb, EXPLICIT_DELALLOC);
1520 if (m->flags & MOPT_CLEAR_ERR)
1521 clear_opt(sb, ERRORS_MASK);
1522 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1523 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1524 "options when quota turned on");
1525 return -1;
1526 }
1527
1528 if (m->flags & MOPT_NOSUPPORT) {
1529 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1530 } else if (token == Opt_commit) {
1531 if (arg == 0)
1532 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1533 sbi->s_commit_interval = HZ * arg;
1534 } else if (token == Opt_max_batch_time) {
1535 if (arg == 0)
1536 arg = EXT4_DEF_MAX_BATCH_TIME;
1537 sbi->s_max_batch_time = arg;
1538 } else if (token == Opt_min_batch_time) {
1539 sbi->s_min_batch_time = arg;
1540 } else if (token == Opt_inode_readahead_blks) {
1541 if (arg > (1 << 30))
1542 return -1;
1543 if (arg && !is_power_of_2(arg)) {
b31e1552 1544 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1545 "EXT4-fs: inode_readahead_blks"
1546 " must be a power of 2");
1547 return -1;
ac27a0ec 1548 }
26092bf5
TT		 1549 sbi->s_inode_readahead_blks = arg;
1550 } else if (token == Opt_init_itable) {
1551 set_opt(sb, INIT_INODE_TABLE);
1552 if (!args->from)
1553 arg = EXT4_DEF_LI_WAIT_MULT;
1554 sbi->s_li_wait_mult = arg;
1555 } else if (token == Opt_stripe) {
1556 sbi->s_stripe = arg;
1557 } else if (m->flags & MOPT_DATAJ) {
ac27a0ec 1558 if (is_remount) {
eb513689
TT		 1559 if (!sbi->s_journal)
1560 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
26092bf5
TT		 1561 else if (test_opt(sb, DATA_FLAGS) !=
1562 m->mount_opt) {
b31e1552 1563 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1564 "Cannot change data mode on remount");
1565 return -1;
ac27a0ec
DK		 1566 }
1567 } else {
fd8c37ec 1568 clear_opt(sb, DATA_FLAGS);
26092bf5 1569 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1570 }
ac27a0ec 1571#ifdef CONFIG_QUOTA
26092bf5 1572 } else if (m->flags & MOPT_QFMT) {
17bd13b3 1573 if (sb_any_quota_loaded(sb) &&
26092bf5
TT		 1574 sbi->s_jquota_fmt != m->mount_opt) {
1575 ext4_msg(sb, KERN_ERR, "Cannot "
1576 "change journaled quota options "
1577 "when quota turned on");
1578 return -1;
ac27a0ec 1579 }
26092bf5 1580 sbi->s_jquota_fmt = m->mount_opt;
ac27a0ec 1581#endif
26092bf5
TT		 1582 } else {
1583 if (!args->from)
1584 arg = 1;
1585 if (m->flags & MOPT_CLEAR)
1586 arg = !arg;
1587 else if (unlikely(!(m->flags & MOPT_SET))) {
1588 ext4_msg(sb, KERN_WARNING,
1589 "buggy handling of option %s", opt);
1590 WARN_ON(1);
1591 return -1;
3197ebdb 1592 }
26092bf5
TT		 1593 if (arg != 0)
1594 sbi->s_mount_opt |= m->mount_opt;
afd4672d 1595 else
26092bf5 1596 sbi->s_mount_opt &= ~m->mount_opt;
ac27a0ec 1597 }
26092bf5
TT		 1598 return 1;
1599 }
1600 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1601 "or missing value", opt);
1602 return -1;
1603}
1604
1605static int parse_options(char *options, struct super_block *sb,
1606 unsigned long *journal_devnum,
1607 unsigned int *journal_ioprio,
1608 int is_remount)
1609{
db7e5c66 1610#ifdef CONFIG_QUOTA
26092bf5 1611 struct ext4_sb_info *sbi = EXT4_SB(sb);
db7e5c66 1612#endif
26092bf5
TT		 1613 char *p;
1614 substring_t args[MAX_OPT_ARGS];
1615 int token;
1616
1617 if (!options)
1618 return 1;
1619
1620 while ((p = strsep(&options, ",")) != NULL) {
1621 if (!*p)
1622 continue;
1623 /*
1624 * Initialize args struct so we know whether arg was
1625 * found; some options take optional arguments.
1626 */
1627 args[0].to = args[0].from = 0;
1628 token = match_token(p, tokens, args);
1629 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1630 journal_ioprio, is_remount) < 0)
1631 return 0;
ac27a0ec
DK		 1632 }
1633#ifdef CONFIG_QUOTA
1634 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1635 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1636 clear_opt(sb, USRQUOTA);
ac27a0ec 1637
482a7425 1638 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1639 clear_opt(sb, GRPQUOTA);
ac27a0ec 1640
56c50f11 1641 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1642 ext4_msg(sb, KERN_ERR, "old and new quota "
1643 "format mixing");
ac27a0ec
DK		 1644 return 0;
1645 }
1646
1647 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1648 ext4_msg(sb, KERN_ERR, "journaled quota format "
1649 "not specified");
ac27a0ec
DK		 1650 return 0;
1651 }
1652 } else {
1653 if (sbi->s_jquota_fmt) {
b31e1552 1654 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1655 "specified with no journaling "
b31e1552 1656 "enabled");
ac27a0ec
DK		 1657 return 0;
1658 }
1659 }
1660#endif
1661 return 1;
1662}
1663
2adf6da8
TT		 1664static inline void ext4_show_quota_options(struct seq_file *seq,
1665 struct super_block *sb)
1666{
1667#if defined(CONFIG_QUOTA)
1668 struct ext4_sb_info *sbi = EXT4_SB(sb);
1669
1670 if (sbi->s_jquota_fmt) {
1671 char *fmtname = "";
1672
1673 switch (sbi->s_jquota_fmt) {
1674 case QFMT_VFS_OLD:
1675 fmtname = "vfsold";
1676 break;
1677 case QFMT_VFS_V0:
1678 fmtname = "vfsv0";
1679 break;
1680 case QFMT_VFS_V1:
1681 fmtname = "vfsv1";
1682 break;
1683 }
1684 seq_printf(seq, ",jqfmt=%s", fmtname);
1685 }
1686
1687 if (sbi->s_qf_names[USRQUOTA])
1688 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1689
1690 if (sbi->s_qf_names[GRPQUOTA])
1691 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1692
1693 if (test_opt(sb, USRQUOTA))
1694 seq_puts(seq, ",usrquota");
1695
1696 if (test_opt(sb, GRPQUOTA))
1697 seq_puts(seq, ",grpquota");
1698#endif
1699}
1700
5a916be1
TT		 1701static const char *token2str(int token)
1702{
1703 static const struct match_token *t;
1704
1705 for (t = tokens; t->token != Opt_err; t++)
1706 if (t->token == token && !strchr(t->pattern, '='))
1707 break;
1708 return t->pattern;
1709}
1710
2adf6da8
TT		 1711/*
1712 * Show an option if
1713 * - it's set to a non-default value OR
1714 * - if the per-sb default is different from the global default
1715 */
66acdcf4
TT		 1716static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1717 int nodefs)
2adf6da8 1718{
2adf6da8
TT		 1719 struct ext4_sb_info *sbi = EXT4_SB(sb);
1720 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1721 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1722 const struct mount_opts *m;
66acdcf4 1723 char sep = nodefs ? '\n' : ',';
2adf6da8 1724
66acdcf4
TT		 1725#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1726#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1727
1728 if (sbi->s_sb_block != 1)
5a916be1
TT		 1729 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1730
1731 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1732 int want_set = m->flags & MOPT_SET;
1733 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1734 (m->flags & MOPT_CLEAR_ERR))
1735 continue;
66acdcf4 1736 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1737 continue; /* skip if same as the default */
1738 if ((want_set &&
1739 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1740 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1741 continue; /* select Opt_noFoo vs Opt_Foo */
1742 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1743 }
5a916be1 1744
66acdcf4 1745 if (nodefs || sbi->s_resuid != EXT4_DEF_RESUID ||
5a916be1
TT		 1746 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1747 SEQ_OPTS_PRINT("resuid=%u", sbi->s_resuid);
66acdcf4 1748 if (nodefs || sbi->s_resgid != EXT4_DEF_RESGID ||
5a916be1
TT		 1749 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1750 SEQ_OPTS_PRINT("resgid=%u", sbi->s_resgid);
66acdcf4 1751 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1752 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1753 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1754 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1755 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1756 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1757 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1758 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1759 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1760 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1761 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1762 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1763 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1764 if (sb->s_flags & MS_I_VERSION)
5a916be1 1765 SEQ_OPTS_PUTS("i_version");
66acdcf4 1766 if (nodefs || sbi->s_stripe)
5a916be1 1767 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1768 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1769 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1770 SEQ_OPTS_PUTS("data=journal");
1771 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1772 SEQ_OPTS_PUTS("data=ordered");
1773 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1774 SEQ_OPTS_PUTS("data=writeback");
1775 }
66acdcf4
TT		 1776 if (nodefs ||
1777 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1778 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1779 sbi->s_inode_readahead_blks);
2adf6da8 1780
66acdcf4
TT		 1781 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1782 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1783 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2adf6da8
TT		 1784
1785 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1786 return 0;
1787}
1788
66acdcf4
TT		 1789static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1790{
1791 return _ext4_show_options(seq, root->d_sb, 0);
1792}
1793
1794static int options_seq_show(struct seq_file *seq, void *offset)
1795{
1796 struct super_block *sb = seq->private;
1797 int rc;
1798
1799 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1800 rc = _ext4_show_options(seq, sb, 1);
1801 seq_puts(seq, "\n");
1802 return rc;
1803}
1804
1805static int options_open_fs(struct inode *inode, struct file *file)
1806{
1807 return single_open(file, options_seq_show, PDE(inode)->data);
1808}
1809
1810static const struct file_operations ext4_seq_options_fops = {
1811 .owner = THIS_MODULE,
1812 .open = options_open_fs,
1813 .read = seq_read,
1814 .llseek = seq_lseek,
1815 .release = single_release,
1816};
1817
617ba13b 1818static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1819 int read_only)
1820{
617ba13b 1821 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1822 int res = 0;
1823
617ba13b 1824 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1825 ext4_msg(sb, KERN_ERR, "revision level too high, "
1826 "forcing read-only mode");
ac27a0ec
DK		 1827 res = MS_RDONLY;
1828 }
1829 if (read_only)
281b5995 1830 goto done;
617ba13b 1831 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1832 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1833 "running e2fsck is recommended");
617ba13b 1834 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1835 ext4_msg(sb, KERN_WARNING,
1836 "warning: mounting fs with errors, "
1837 "running e2fsck is recommended");
ed3ce80a 1838 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1839 le16_to_cpu(es->s_mnt_count) >=
1840 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1841 ext4_msg(sb, KERN_WARNING,
1842 "warning: maximal mount count reached, "
1843 "running e2fsck is recommended");
ac27a0ec
DK		 1844 else if (le32_to_cpu(es->s_checkinterval) &&
1845 (le32_to_cpu(es->s_lastcheck) +
1846 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1847 ext4_msg(sb, KERN_WARNING,
1848 "warning: checktime reached, "
1849 "running e2fsck is recommended");
0b8e58a1 1850 if (!sbi->s_journal)
0390131b 1851 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1852 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1853 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1854 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1855 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1856 ext4_update_dynamic_rev(sb);
0390131b
FM		 1857 if (sbi->s_journal)
1858 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1859
e2d67052 1860 ext4_commit_super(sb, 1);
281b5995 1861done:
ac27a0ec 1862 if (test_opt(sb, DEBUG))
a9df9a49 1863 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1864 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1865 sb->s_blocksize,
1866 sbi->s_groups_count,
617ba13b
MC		 1867 EXT4_BLOCKS_PER_GROUP(sb),
1868 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1869 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1870
7abc52c2 1871 cleancache_init_fs(sb);
ac27a0ec
DK		 1872 return res;
1873}
1874
772cb7c8
JS		 1875static int ext4_fill_flex_info(struct super_block *sb)
1876{
1877 struct ext4_sb_info *sbi = EXT4_SB(sb);
1878 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1879 ext4_group_t flex_group_count;
1880 ext4_group_t flex_group;
d50f2ab6 1881 unsigned int groups_per_flex = 0;
c5ca7c76 1882 size_t size;
772cb7c8
JS		 1883 int i;
1884
503358ae 1885 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1886 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1887 sbi->s_log_groups_per_flex = 0;
1888 return 1;
1889 }
d50f2ab6 1890 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
772cb7c8 1891
c62a11fd
FB		 1892 /* We allocate both existing and potentially added groups */
1893 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1894 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1895 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76 1896 size = flex_group_count * sizeof(struct flex_groups);
9933fc0a 1897 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
c5ca7c76 1898 if (sbi->s_flex_groups == NULL) {
9933fc0a
TT		 1899 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1900 flex_group_count);
1901 goto failed;
772cb7c8 1902 }
772cb7c8 1903
772cb7c8 1904 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1905 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1906
1907 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1908 atomic_add(ext4_free_inodes_count(sb, gdp),
1909 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 1910 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 1911 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1912 atomic_add(ext4_used_dirs_count(sb, gdp),
1913 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1914 }
1915
1916 return 1;
1917failed:
1918 return 0;
1919}
1920
717d50e4
AD		 1921__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1922 struct ext4_group_desc *gdp)
1923{
1924 __u16 crc = 0;
1925
1926 if (sbi->s_es->s_feature_ro_compat &
1927 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1928 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1929 __le32 le_group = cpu_to_le32(block_group);
1930
1931 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1932 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1933 crc = crc16(crc, (__u8 *)gdp, offset);
1934 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1935 /* for checksum of struct ext4_group_desc do the rest...*/
1936 if ((sbi->s_es->s_feature_incompat &
1937 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1938 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1939 crc = crc16(crc, (__u8 *)gdp + offset,
1940 le16_to_cpu(sbi->s_es->s_desc_size) -
1941 offset);
1942 }
1943
1944 return cpu_to_le16(crc);
1945}
1946
1947int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1948 struct ext4_group_desc *gdp)
1949{
1950 if ((sbi->s_es->s_feature_ro_compat &
1951 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1952 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1953 return 0;
1954
1955 return 1;
1956}
1957
ac27a0ec 1958/* Called at mount-time, super-block is locked */
bfff6873
LC		 1959static int ext4_check_descriptors(struct super_block *sb,
1960 ext4_group_t *first_not_zeroed)
ac27a0ec 1961{
617ba13b
MC		 1962 struct ext4_sb_info *sbi = EXT4_SB(sb);
1963 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1964 ext4_fsblk_t last_block;
bd81d8ee
LV		 1965 ext4_fsblk_t block_bitmap;
1966 ext4_fsblk_t inode_bitmap;
1967 ext4_fsblk_t inode_table;
ce421581 1968 int flexbg_flag = 0;
bfff6873 1969 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 1970
ce421581
JS		 1971 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1972 flexbg_flag = 1;
1973
af5bc92d 1974 ext4_debug("Checking group descriptors");
ac27a0ec 1975
197cd65a
AM		 1976 for (i = 0; i < sbi->s_groups_count; i++) {
1977 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1978
ce421581 1979 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1980 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1981 else
1982 last_block = first_block +
617ba13b 1983 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1984
bfff6873
LC		 1985 if ((grp == sbi->s_groups_count) &&
1986 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
1987 grp = i;
1988
8fadc143 1989 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1990 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 1991 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1992 "Block bitmap for group %u not in group "
b31e1552 1993 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1994 return 0;
1995 }
8fadc143 1996 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1997 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 1998 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1999 "Inode bitmap for group %u not in group "
b31e1552 2000 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2001 return 0;
2002 }
8fadc143 2003 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2004 if (inode_table < first_block ||
2b2d6d01 2005 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2006 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2007 "Inode table for group %u not in group "
b31e1552 2008 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2009 return 0;
2010 }
955ce5f5 2011 ext4_lock_group(sb, i);
717d50e4 2012 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 2013 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2014 "Checksum for group %u failed (%u!=%u)",
2015 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2016 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2017 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2018 ext4_unlock_group(sb, i);
8a266467 2019 return 0;
7ee1ec4c 2020 }
717d50e4 2021 }
955ce5f5 2022 ext4_unlock_group(sb, i);
ce421581
JS		 2023 if (!flexbg_flag)
2024 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2025 }
bfff6873
LC		 2026 if (NULL != first_not_zeroed)
2027 *first_not_zeroed = grp;
ac27a0ec 2028
5dee5437
TT		 2029 ext4_free_blocks_count_set(sbi->s_es,
2030 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2031 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2032 return 1;
2033}
2034
617ba13b 2035/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2036 * the superblock) which were deleted from all directories, but held open by
2037 * a process at the time of a crash. We walk the list and try to delete these
2038 * inodes at recovery time (only with a read-write filesystem).
2039 *
2040 * In order to keep the orphan inode chain consistent during traversal (in
2041 * case of crash during recovery), we link each inode into the superblock
2042 * orphan list_head and handle it the same way as an inode deletion during
2043 * normal operation (which journals the operations for us).
2044 *
2045 * We only do an iget() and an iput() on each inode, which is very safe if we
2046 * accidentally point at an in-use or already deleted inode. The worst that
2047 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2048 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2049 * e2fsck was run on this filesystem, and it must have already done the orphan
2050 * inode cleanup for us, so we can safely abort without any further action.
2051 */
2b2d6d01
TT		 2052static void ext4_orphan_cleanup(struct super_block *sb,
2053 struct ext4_super_block *es)
ac27a0ec
DK		 2054{
2055 unsigned int s_flags = sb->s_flags;
2056 int nr_orphans = 0, nr_truncates = 0;
2057#ifdef CONFIG_QUOTA
2058 int i;
2059#endif
2060 if (!es->s_last_orphan) {
2061 jbd_debug(4, "no orphan inodes to clean up\n");
2062 return;
2063 }
2064
a8f48a95 2065 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2066 ext4_msg(sb, KERN_ERR, "write access "
2067 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2068 return;
2069 }
2070
d39195c3
AG		 2071 /* Check if feature set would not allow a r/w mount */
2072 if (!ext4_feature_set_ok(sb, 0)) {
2073 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2074 "unknown ROCOMPAT features");
2075 return;
2076 }
2077
617ba13b 2078 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2079 if (es->s_last_orphan)
2080 jbd_debug(1, "Errors on filesystem, "
2081 "clearing orphan list.\n");
2082 es->s_last_orphan = 0;
2083 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2084 return;
2085 }
2086
2087 if (s_flags & MS_RDONLY) {
b31e1552 2088 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2089 sb->s_flags &= ~MS_RDONLY;
2090 }
2091#ifdef CONFIG_QUOTA
2092 /* Needed for iput() to work correctly and not trash data */
2093 sb->s_flags |= MS_ACTIVE;
2094 /* Turn on quotas so that they are updated correctly */
2095 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2096 if (EXT4_SB(sb)->s_qf_names[i]) {
2097 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2098 if (ret < 0)
b31e1552
ES		 2099 ext4_msg(sb, KERN_ERR,
2100 "Cannot turn on journaled "
2101 "quota: error %d", ret);
ac27a0ec
DK		 2102 }
2103 }
2104#endif
2105
2106 while (es->s_last_orphan) {
2107 struct inode *inode;
2108
97bd42b9
JB		 2109 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2110 if (IS_ERR(inode)) {
ac27a0ec
DK		 2111 es->s_last_orphan = 0;
2112 break;
2113 }
2114
617ba13b 2115 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2116 dquot_initialize(inode);
ac27a0ec 2117 if (inode->i_nlink) {
b31e1552
ES		 2118 ext4_msg(sb, KERN_DEBUG,
2119 "%s: truncating inode %lu to %lld bytes",
46e665e9 2120 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2121 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2122 inode->i_ino, inode->i_size);
617ba13b 2123 ext4_truncate(inode);
ac27a0ec
DK		 2124 nr_truncates++;
2125 } else {
b31e1552
ES		 2126 ext4_msg(sb, KERN_DEBUG,
2127 "%s: deleting unreferenced inode %lu",
46e665e9 2128 __func__, inode->i_ino);
ac27a0ec
DK		 2129 jbd_debug(2, "deleting unreferenced inode %lu\n",
2130 inode->i_ino);
2131 nr_orphans++;
2132 }
2133 iput(inode); /* The delete magic happens here! */
2134 }
2135
2b2d6d01 2136#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2137
2138 if (nr_orphans)
b31e1552
ES		 2139 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2140 PLURAL(nr_orphans));
ac27a0ec 2141 if (nr_truncates)
b31e1552
ES		 2142 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2143 PLURAL(nr_truncates));
ac27a0ec
DK		 2144#ifdef CONFIG_QUOTA
2145 /* Turn quotas off */
2146 for (i = 0; i < MAXQUOTAS; i++) {
2147 if (sb_dqopt(sb)->files[i])
287a8095 2148 dquot_quota_off(sb, i);
ac27a0ec
DK		 2149 }
2150#endif
2151 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2152}
0b8e58a1 2153
cd2291a4
ES		 2154/*
2155 * Maximal extent format file size.
2156 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2157 * extent format containers, within a sector_t, and within i_blocks
2158 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2159 * so that won't be a limiting factor.
2160 *
f17722f9
LC		 2161 * However there is other limiting factor. We do store extents in the form
2162 * of starting block and length, hence the resulting length of the extent
2163 * covering maximum file size must fit into on-disk format containers as
2164 * well. Given that length is always by 1 unit bigger than max unit (because
2165 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2166 *
cd2291a4
ES		 2167 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2168 */
f287a1a5 2169static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2170{
2171 loff_t res;
2172 loff_t upper_limit = MAX_LFS_FILESIZE;
2173
2174 /* small i_blocks in vfs inode? */
f287a1a5 2175 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2176 /*
90c699a9 2177 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2178 * i_block represent total blocks in 512 bytes
2179 * 32 == size of vfs inode i_blocks * 8
2180 */
2181 upper_limit = (1LL << 32) - 1;
2182
2183 /* total blocks in file system block size */
2184 upper_limit >>= (blkbits - 9);
2185 upper_limit <<= blkbits;
2186 }
2187
f17722f9
LC		 2188 /*
2189 * 32-bit extent-start container, ee_block. We lower the maxbytes
2190 * by one fs block, so ee_len can cover the extent of maximum file
2191 * size
2192 */
2193 res = (1LL << 32) - 1;
cd2291a4 2194 res <<= blkbits;
cd2291a4
ES		 2195
2196 /* Sanity check against vm- & vfs- imposed limits */
2197 if (res > upper_limit)
2198 res = upper_limit;
2199
2200 return res;
2201}
ac27a0ec 2202
ac27a0ec 2203/*
cd2291a4 2204 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2205 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2206 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2207 */
f287a1a5 2208static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2209{
617ba13b 2210 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2211 int meta_blocks;
2212 loff_t upper_limit;
0b8e58a1
AD		 2213 /* This is calculated to be the largest file size for a dense, block
2214 * mapped file such that the file's total number of 512-byte sectors,
2215 * including data and all indirect blocks, does not exceed (2^48 - 1).
2216 *
2217 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2218 * number of 512-byte sectors of the file.
0fc1b451
AK		 2219 */
2220
f287a1a5 2221 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2222 /*
90c699a9 2223 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2224 * the inode i_block field represents total file blocks in
2225 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2226 */
2227 upper_limit = (1LL << 32) - 1;
2228
2229 /* total blocks in file system block size */
2230 upper_limit >>= (bits - 9);
2231
2232 } else {
8180a562
AK		 2233 /*
2234 * We use 48 bit ext4_inode i_blocks
2235 * With EXT4_HUGE_FILE_FL set the i_blocks
2236 * represent total number of blocks in
2237 * file system block size
2238 */
0fc1b451
AK		 2239 upper_limit = (1LL << 48) - 1;
2240
0fc1b451
AK		 2241 }
2242
2243 /* indirect blocks */
2244 meta_blocks = 1;
2245 /* double indirect blocks */
2246 meta_blocks += 1 + (1LL << (bits-2));
2247 /* tripple indirect blocks */
2248 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2249
2250 upper_limit -= meta_blocks;
2251 upper_limit <<= bits;
ac27a0ec
DK		 2252
2253 res += 1LL << (bits-2);
2254 res += 1LL << (2*(bits-2));
2255 res += 1LL << (3*(bits-2));
2256 res <<= bits;
2257 if (res > upper_limit)
2258 res = upper_limit;
0fc1b451
AK		 2259
2260 if (res > MAX_LFS_FILESIZE)
2261 res = MAX_LFS_FILESIZE;
2262
ac27a0ec
DK		 2263 return res;
2264}
2265
617ba13b 2266static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2267 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2268{
617ba13b 2269 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2270 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2271 int has_super = 0;
2272
2273 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2274
617ba13b 2275 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2276 nr < first_meta_bg)
70bbb3e0 2277 return logical_sb_block + nr + 1;
ac27a0ec 2278 bg = sbi->s_desc_per_block * nr;
617ba13b 2279 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2280 has_super = 1;
0b8e58a1 2281
617ba13b 2282 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2283}
2284
c9de560d
AT		 2285/**
2286 * ext4_get_stripe_size: Get the stripe size.
2287 * @sbi: In memory super block info
2288 *
2289 * If we have specified it via mount option, then
2290 * use the mount option value. If the value specified at mount time is
2291 * greater than the blocks per group use the super block value.
2292 * If the super block value is greater than blocks per group return 0.
2293 * Allocator needs it be less than blocks per group.
2294 *
2295 */
2296static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2297{
2298 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2299 unsigned long stripe_width =
2300 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2301 int ret;
c9de560d
AT		 2302
2303 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2304 ret = sbi->s_stripe;
2305 else if (stripe_width <= sbi->s_blocks_per_group)
2306 ret = stripe_width;
2307 else if (stride <= sbi->s_blocks_per_group)
2308 ret = stride;
2309 else
2310 ret = 0;
c9de560d 2311
3eb08658
DE		 2312 /*
2313 * If the stripe width is 1, this makes no sense and
2314 * we set it to 0 to turn off stripe handling code.
2315 */
2316 if (ret <= 1)
2317 ret = 0;
c9de560d 2318
3eb08658 2319 return ret;
c9de560d 2320}
ac27a0ec 2321
3197ebdb
TT		 2322/* sysfs supprt */
2323
2324struct ext4_attr {
2325 struct attribute attr;
2326 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2327 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2328 const char *, size_t);
2329 int offset;
2330};
2331
2332static int parse_strtoul(const char *buf,
2333 unsigned long max, unsigned long *value)
2334{
2335 char *endp;
2336
e7d2860b
AGR		 2337 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2338 endp = skip_spaces(endp);
3197ebdb
TT		 2339 if (*endp || *value > max)
2340 return -EINVAL;
2341
2342 return 0;
2343}
2344
2345static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2346 struct ext4_sb_info *sbi,
2347 char *buf)
2348{
2349 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2350 (s64) EXT4_C2B(sbi,
2351 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2352}
2353
2354static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2355 struct ext4_sb_info *sbi, char *buf)
2356{
2357 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2358
f613dfcb
TT		 2359 if (!sb->s_bdev->bd_part)
2360 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2361 return snprintf(buf, PAGE_SIZE, "%lu\n",
2362 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2363 sbi->s_sectors_written_start) >> 1);
2364}
2365
2366static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2367 struct ext4_sb_info *sbi, char *buf)
2368{
2369 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2370
f613dfcb
TT		 2371 if (!sb->s_bdev->bd_part)
2372 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2373 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2374 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2375 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2376 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2377}
2378
2379static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2380 struct ext4_sb_info *sbi,
2381 const char *buf, size_t count)
2382{
2383 unsigned long t;
2384
2385 if (parse_strtoul(buf, 0x40000000, &t))
2386 return -EINVAL;
2387
5dbd571d 2388 if (t && !is_power_of_2(t))
3197ebdb
TT		 2389 return -EINVAL;
2390
2391 sbi->s_inode_readahead_blks = t;
2392 return count;
2393}
2394
2395static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2396 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2397{
2398 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2399
2400 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2401}
2402
2403static ssize_t sbi_ui_store(struct ext4_attr *a,
2404 struct ext4_sb_info *sbi,
2405 const char *buf, size_t count)
2406{
2407 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2408 unsigned long t;
2409
2410 if (parse_strtoul(buf, 0xffffffff, &t))
2411 return -EINVAL;
2412 *ui = t;
2413 return count;
2414}
2415
2416#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2417static struct ext4_attr ext4_attr_##_name = { \
2418 .attr = {.name = __stringify(_name), .mode = _mode }, \
2419 .show = _show, \
2420 .store = _store, \
2421 .offset = offsetof(struct ext4_sb_info, _elname), \
2422}
2423#define EXT4_ATTR(name, mode, show, store) \
2424static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2425
857ac889 2426#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2427#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2428#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2429#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2430 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2431#define ATTR_LIST(name) &ext4_attr_##name.attr
2432
2433EXT4_RO_ATTR(delayed_allocation_blocks);
2434EXT4_RO_ATTR(session_write_kbytes);
2435EXT4_RO_ATTR(lifetime_write_kbytes);
2436EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2437 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2438EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2439EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2440EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2441EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2442EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2443EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2444EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2445EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2446
2447static struct attribute *ext4_attrs[] = {
2448 ATTR_LIST(delayed_allocation_blocks),
2449 ATTR_LIST(session_write_kbytes),
2450 ATTR_LIST(lifetime_write_kbytes),
2451 ATTR_LIST(inode_readahead_blks),
11013911 2452 ATTR_LIST(inode_goal),
3197ebdb
TT		 2453 ATTR_LIST(mb_stats),
2454 ATTR_LIST(mb_max_to_scan),
2455 ATTR_LIST(mb_min_to_scan),
2456 ATTR_LIST(mb_order2_req),
2457 ATTR_LIST(mb_stream_req),
2458 ATTR_LIST(mb_group_prealloc),
55138e0b 2459 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2460 NULL,
2461};
2462
857ac889
LC		 2463/* Features this copy of ext4 supports */
2464EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2465EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2466
2467static struct attribute *ext4_feat_attrs[] = {
2468 ATTR_LIST(lazy_itable_init),
27ee40df 2469 ATTR_LIST(batched_discard),
857ac889
LC		 2470 NULL,
2471};
2472
3197ebdb
TT		 2473static ssize_t ext4_attr_show(struct kobject *kobj,
2474 struct attribute *attr, char *buf)
2475{
2476 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2477 s_kobj);
2478 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2479
2480 return a->show ? a->show(a, sbi, buf) : 0;
2481}
2482
2483static ssize_t ext4_attr_store(struct kobject *kobj,
2484 struct attribute *attr,
2485 const char *buf, size_t len)
2486{
2487 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2488 s_kobj);
2489 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2490
2491 return a->store ? a->store(a, sbi, buf, len) : 0;
2492}
2493
2494static void ext4_sb_release(struct kobject *kobj)
2495{
2496 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2497 s_kobj);
2498 complete(&sbi->s_kobj_unregister);
2499}
2500
52cf25d0 2501static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2502 .show = ext4_attr_show,
2503 .store = ext4_attr_store,
2504};
2505
2506static struct kobj_type ext4_ktype = {
2507 .default_attrs = ext4_attrs,
2508 .sysfs_ops = &ext4_attr_ops,
2509 .release = ext4_sb_release,
2510};
2511
857ac889
LC		 2512static void ext4_feat_release(struct kobject *kobj)
2513{
2514 complete(&ext4_feat->f_kobj_unregister);
2515}
2516
2517static struct kobj_type ext4_feat_ktype = {
2518 .default_attrs = ext4_feat_attrs,
2519 .sysfs_ops = &ext4_attr_ops,
2520 .release = ext4_feat_release,
2521};
2522
a13fb1a4
ES		 2523/*
2524 * Check whether this filesystem can be mounted based on
2525 * the features present and the RDONLY/RDWR mount requested.
2526 * Returns 1 if this filesystem can be mounted as requested,
2527 * 0 if it cannot be.
2528 */
2529static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2530{
2531 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2532 ext4_msg(sb, KERN_ERR,
2533 "Couldn't mount because of "
2534 "unsupported optional features (%x)",
2535 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2536 ~EXT4_FEATURE_INCOMPAT_SUPP));
2537 return 0;
2538 }
2539
2540 if (readonly)
2541 return 1;
2542
2543 /* Check that feature set is OK for a read-write mount */
2544 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2545 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2546 "unsupported optional features (%x)",
2547 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2548 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2549 return 0;
2550 }
2551 /*
2552 * Large file size enabled file system can only be mounted
2553 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2554 */
2555 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2556 if (sizeof(blkcnt_t) < sizeof(u64)) {
2557 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2558 "cannot be mounted RDWR without "
2559 "CONFIG_LBDAF");
2560 return 0;
2561 }
2562 }
bab08ab9
TT		 2563 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2564 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2565 ext4_msg(sb, KERN_ERR,
2566 "Can't support bigalloc feature without "
2567 "extents feature\n");
2568 return 0;
2569 }
a13fb1a4
ES		 2570 return 1;
2571}
2572
66e61a9e
TT		 2573/*
2574 * This function is called once a day if we have errors logged
2575 * on the file system
2576 */
2577static void print_daily_error_info(unsigned long arg)
2578{
2579 struct super_block *sb = (struct super_block *) arg;
2580 struct ext4_sb_info *sbi;
2581 struct ext4_super_block *es;
2582
2583 sbi = EXT4_SB(sb);
2584 es = sbi->s_es;
2585
2586 if (es->s_error_count)
2587 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2588 le32_to_cpu(es->s_error_count));
2589 if (es->s_first_error_time) {
2590 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2591 sb->s_id, le32_to_cpu(es->s_first_error_time),
2592 (int) sizeof(es->s_first_error_func),
2593 es->s_first_error_func,
2594 le32_to_cpu(es->s_first_error_line));
2595 if (es->s_first_error_ino)
2596 printk(": inode %u",
2597 le32_to_cpu(es->s_first_error_ino));
2598 if (es->s_first_error_block)
2599 printk(": block %llu", (unsigned long long)
2600 le64_to_cpu(es->s_first_error_block));
2601 printk("\n");
2602 }
2603 if (es->s_last_error_time) {
2604 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2605 sb->s_id, le32_to_cpu(es->s_last_error_time),
2606 (int) sizeof(es->s_last_error_func),
2607 es->s_last_error_func,
2608 le32_to_cpu(es->s_last_error_line));
2609 if (es->s_last_error_ino)
2610 printk(": inode %u",
2611 le32_to_cpu(es->s_last_error_ino));
2612 if (es->s_last_error_block)
2613 printk(": block %llu", (unsigned long long)
2614 le64_to_cpu(es->s_last_error_block));
2615 printk("\n");
2616 }
2617 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2618}
2619
bfff6873
LC		 2620/* Find next suitable group and run ext4_init_inode_table */
2621static int ext4_run_li_request(struct ext4_li_request *elr)
2622{
2623 struct ext4_group_desc *gdp = NULL;
2624 ext4_group_t group, ngroups;
2625 struct super_block *sb;
2626 unsigned long timeout = 0;
2627 int ret = 0;
2628
2629 sb = elr->lr_super;
2630 ngroups = EXT4_SB(sb)->s_groups_count;
2631
2632 for (group = elr->lr_next_group; group < ngroups; group++) {
2633 gdp = ext4_get_group_desc(sb, group, NULL);
2634 if (!gdp) {
2635 ret = 1;
2636 break;
2637 }
2638
2639 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2640 break;
2641 }
2642
2643 if (group == ngroups)
2644 ret = 1;
2645
2646 if (!ret) {
2647 timeout = jiffies;
2648 ret = ext4_init_inode_table(sb, group,
2649 elr->lr_timeout ? 0 : 1);
2650 if (elr->lr_timeout == 0) {
51ce6511
LC		 2651 timeout = (jiffies - timeout) *
2652 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2653 elr->lr_timeout = timeout;
2654 }
2655 elr->lr_next_sched = jiffies + elr->lr_timeout;
2656 elr->lr_next_group = group + 1;
2657 }
2658
2659 return ret;
2660}
2661
2662/*
2663 * Remove lr_request from the list_request and free the
4ed5c033 2664 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2665 */
2666static void ext4_remove_li_request(struct ext4_li_request *elr)
2667{
2668 struct ext4_sb_info *sbi;
2669
2670 if (!elr)
2671 return;
2672
2673 sbi = elr->lr_sbi;
2674
2675 list_del(&elr->lr_request);
2676 sbi->s_li_request = NULL;
2677 kfree(elr);
2678}
2679
2680static void ext4_unregister_li_request(struct super_block *sb)
2681{
1bb933fb
LC		 2682 mutex_lock(&ext4_li_mtx);
2683 if (!ext4_li_info) {
2684 mutex_unlock(&ext4_li_mtx);
bfff6873 2685 return;
1bb933fb 2686 }
bfff6873
LC		 2687
2688 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2689 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2690 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2691 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2692}
2693
8f1f7453
ES		 2694static struct task_struct *ext4_lazyinit_task;
2695
bfff6873
LC		 2696/*
2697 * This is the function where ext4lazyinit thread lives. It walks
2698 * through the request list searching for next scheduled filesystem.
2699 * When such a fs is found, run the lazy initialization request
2700 * (ext4_rn_li_request) and keep track of the time spend in this
2701 * function. Based on that time we compute next schedule time of
2702 * the request. When walking through the list is complete, compute
2703 * next waking time and put itself into sleep.
2704 */
2705static int ext4_lazyinit_thread(void *arg)
2706{
2707 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2708 struct list_head *pos, *n;
2709 struct ext4_li_request *elr;
4ed5c033 2710 unsigned long next_wakeup, cur;
bfff6873
LC		 2711
2712 BUG_ON(NULL == eli);
2713
bfff6873
LC		 2714cont_thread:
2715 while (true) {
2716 next_wakeup = MAX_JIFFY_OFFSET;
2717
2718 mutex_lock(&eli->li_list_mtx);
2719 if (list_empty(&eli->li_request_list)) {
2720 mutex_unlock(&eli->li_list_mtx);
2721 goto exit_thread;
2722 }
2723
2724 list_for_each_safe(pos, n, &eli->li_request_list) {
2725 elr = list_entry(pos, struct ext4_li_request,
2726 lr_request);
2727
b2c78cd0
TT		 2728 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2729 if (ext4_run_li_request(elr) != 0) {
2730 /* error, remove the lazy_init job */
2731 ext4_remove_li_request(elr);
2732 continue;
2733 }
bfff6873
LC		 2734 }
2735
2736 if (time_before(elr->lr_next_sched, next_wakeup))
2737 next_wakeup = elr->lr_next_sched;
2738 }
2739 mutex_unlock(&eli->li_list_mtx);
2740
a0acae0e 2741 try_to_freeze();
bfff6873 2742
4ed5c033
LC		 2743 cur = jiffies;
2744 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2745 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2746 cond_resched();
2747 continue;
2748 }
2749
4ed5c033
LC		 2750 schedule_timeout_interruptible(next_wakeup - cur);
2751
8f1f7453
ES		 2752 if (kthread_should_stop()) {
2753 ext4_clear_request_list();
2754 goto exit_thread;
2755 }
bfff6873
LC		 2756 }
2757
2758exit_thread:
2759 /*
2760 * It looks like the request list is empty, but we need
2761 * to check it under the li_list_mtx lock, to prevent any
2762 * additions into it, and of course we should lock ext4_li_mtx
2763 * to atomically free the list and ext4_li_info, because at
2764 * this point another ext4 filesystem could be registering
2765 * new one.
2766 */
2767 mutex_lock(&ext4_li_mtx);
2768 mutex_lock(&eli->li_list_mtx);
2769 if (!list_empty(&eli->li_request_list)) {
2770 mutex_unlock(&eli->li_list_mtx);
2771 mutex_unlock(&ext4_li_mtx);
2772 goto cont_thread;
2773 }
2774 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2775 kfree(ext4_li_info);
2776 ext4_li_info = NULL;
2777 mutex_unlock(&ext4_li_mtx);
2778
2779 return 0;
2780}
2781
2782static void ext4_clear_request_list(void)
2783{
2784 struct list_head *pos, *n;
2785 struct ext4_li_request *elr;
2786
2787 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2788 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2789 elr = list_entry(pos, struct ext4_li_request,
2790 lr_request);
2791 ext4_remove_li_request(elr);
2792 }
2793 mutex_unlock(&ext4_li_info->li_list_mtx);
2794}
2795
2796static int ext4_run_lazyinit_thread(void)
2797{
8f1f7453
ES		 2798 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2799 ext4_li_info, "ext4lazyinit");
2800 if (IS_ERR(ext4_lazyinit_task)) {
2801 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2802 ext4_clear_request_list();
bfff6873
LC		 2803 kfree(ext4_li_info);
2804 ext4_li_info = NULL;
92b97816 2805 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2806 "initialization thread\n",
2807 err);
2808 return err;
2809 }
2810 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2811 return 0;
2812}
2813
2814/*
2815 * Check whether it make sense to run itable init. thread or not.
2816 * If there is at least one uninitialized inode table, return
2817 * corresponding group number, else the loop goes through all
2818 * groups and return total number of groups.
2819 */
2820static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2821{
2822 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2823 struct ext4_group_desc *gdp = NULL;
2824
2825 for (group = 0; group < ngroups; group++) {
2826 gdp = ext4_get_group_desc(sb, group, NULL);
2827 if (!gdp)
2828 continue;
2829
2830 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2831 break;
2832 }
2833
2834 return group;
2835}
2836
2837static int ext4_li_info_new(void)
2838{
2839 struct ext4_lazy_init *eli = NULL;
2840
2841 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2842 if (!eli)
2843 return -ENOMEM;
2844
bfff6873
LC		 2845 INIT_LIST_HEAD(&eli->li_request_list);
2846 mutex_init(&eli->li_list_mtx);
2847
bfff6873
LC		 2848 eli->li_state |= EXT4_LAZYINIT_QUIT;
2849
2850 ext4_li_info = eli;
2851
2852 return 0;
2853}
2854
2855static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2856 ext4_group_t start)
2857{
2858 struct ext4_sb_info *sbi = EXT4_SB(sb);
2859 struct ext4_li_request *elr;
2860 unsigned long rnd;
2861
2862 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2863 if (!elr)
2864 return NULL;
2865
2866 elr->lr_super = sb;
2867 elr->lr_sbi = sbi;
2868 elr->lr_next_group = start;
2869
2870 /*
2871 * Randomize first schedule time of the request to
2872 * spread the inode table initialization requests
2873 * better.
2874 */
2875 get_random_bytes(&rnd, sizeof(rnd));
2876 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2877 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2878
2879 return elr;
2880}
2881
2882static int ext4_register_li_request(struct super_block *sb,
2883 ext4_group_t first_not_zeroed)
2884{
2885 struct ext4_sb_info *sbi = EXT4_SB(sb);
2886 struct ext4_li_request *elr;
2887 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2888 int ret = 0;
bfff6873 2889
51ce6511
LC		 2890 if (sbi->s_li_request != NULL) {
2891 /*
2892 * Reset timeout so it can be computed again, because
2893 * s_li_wait_mult might have changed.
2894 */
2895 sbi->s_li_request->lr_timeout = 0;
beed5ecb 2896 return 0;
51ce6511 2897 }
bfff6873
LC		 2898
2899 if (first_not_zeroed == ngroups ||
2900 (sb->s_flags & MS_RDONLY) ||
55ff3840 2901 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 2902 return 0;
bfff6873
LC		 2903
2904 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 2905 if (!elr)
2906 return -ENOMEM;
bfff6873
LC		 2907
2908 mutex_lock(&ext4_li_mtx);
2909
2910 if (NULL == ext4_li_info) {
2911 ret = ext4_li_info_new();
2912 if (ret)
2913 goto out;
2914 }
2915
2916 mutex_lock(&ext4_li_info->li_list_mtx);
2917 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2918 mutex_unlock(&ext4_li_info->li_list_mtx);
2919
2920 sbi->s_li_request = elr;
46e4690b
TM		 2921 /*
2922 * set elr to NULL here since it has been inserted to
2923 * the request_list and the removal and free of it is
2924 * handled by ext4_clear_request_list from now on.
2925 */
2926 elr = NULL;
bfff6873
LC		 2927
2928 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2929 ret = ext4_run_lazyinit_thread();
2930 if (ret)
2931 goto out;
2932 }
bfff6873 2933out:
beed5ecb
NK		 2934 mutex_unlock(&ext4_li_mtx);
2935 if (ret)
bfff6873 2936 kfree(elr);
bfff6873
LC		 2937 return ret;
2938}
2939
2940/*
2941 * We do not need to lock anything since this is called on
2942 * module unload.
2943 */
2944static void ext4_destroy_lazyinit_thread(void)
2945{
2946 /*
2947 * If thread exited earlier
2948 * there's nothing to be done.
2949 */
8f1f7453 2950 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 2951 return;
2952
8f1f7453 2953 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 2954}
2955
2b2d6d01 2956static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 2957{
d4c402d9 2958 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 2959 struct buffer_head *bh;
617ba13b
MC		 2960 struct ext4_super_block *es = NULL;
2961 struct ext4_sb_info *sbi;
2962 ext4_fsblk_t block;
2963 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2964 ext4_fsblk_t logical_sb_block;
ac27a0ec 2965 unsigned long offset = 0;
ac27a0ec
DK		 2966 unsigned long journal_devnum = 0;
2967 unsigned long def_mount_opts;
2968 struct inode *root;
9f6200bb 2969 char *cp;
0390131b 2970 const char *descr;
dcc7dae3 2971 int ret = -ENOMEM;
281b5995 2972 int blocksize, clustersize;
4ec11028
TT		 2973 unsigned int db_count;
2974 unsigned int i;
281b5995 2975 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 2976 __u64 blocks_count;
833f4077 2977 int err;
b3881f74 2978 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 2979 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 2980
2981 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2982 if (!sbi)
dcc7dae3 2983 goto out_free_orig;
705895b6
PE		 2984
2985 sbi->s_blockgroup_lock =
2986 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2987 if (!sbi->s_blockgroup_lock) {
2988 kfree(sbi);
dcc7dae3 2989 goto out_free_orig;
705895b6 2990 }
ac27a0ec
DK		 2991 sb->s_fs_info = sbi;
2992 sbi->s_mount_opt = 0;
617ba13b
MC		 2993 sbi->s_resuid = EXT4_DEF_RESUID;
2994 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2995 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2996 sbi->s_sb_block = sb_block;
f613dfcb
TT		 2997 if (sb->s_bdev->bd_part)
2998 sbi->s_sectors_written_start =
2999 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3000
9f6200bb
TT		 3001 /* Cleanup superblock name */
3002 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3003 *cp = '!';
3004
dcc7dae3 3005 ret = -EINVAL;
617ba13b 3006 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3007 if (!blocksize) {
b31e1552 3008 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3009 goto out_fail;
3010 }
3011
3012 /*
617ba13b 3013 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3014 * block sizes. We need to calculate the offset from buffer start.
3015 */
617ba13b 3016 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3017 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3018 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3019 } else {
70bbb3e0 3020 logical_sb_block = sb_block;
ac27a0ec
DK		 3021 }
3022
70bbb3e0 3023 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3024 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3025 goto out_fail;
3026 }
3027 /*
3028 * Note: s_es must be initialized as soon as possible because
617ba13b 3029 * some ext4 macro-instructions depend on its value
ac27a0ec 3030 */
617ba13b 3031 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 3032 sbi->s_es = es;
3033 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3034 if (sb->s_magic != EXT4_SUPER_MAGIC)
3035 goto cantfind_ext4;
afc32f7e 3036 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3037
d25425f8
DW		 3038 /* Check for a known checksum algorithm */
3039 if (!ext4_verify_csum_type(sb, es)) {
3040 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3041 "unknown checksum algorithm.");
3042 silent = 1;
3043 goto cantfind_ext4;
3044 }
3045
ac27a0ec
DK		 3046 /* Set defaults before we parse the mount options */
3047 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3048 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3049 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3050 set_opt(sb, DEBUG);
87f26807 3051 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3052 set_opt(sb, GRPID);
617ba13b 3053 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3054 set_opt(sb, NO_UID32);
ea663336 3055 /* xattr user namespace & acls are now defaulted on */
03010a33 3056#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3057 set_opt(sb, XATTR_USER);
2e7842b8 3058#endif
03010a33 3059#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3060 set_opt(sb, POSIX_ACL);
2e7842b8 3061#endif
6fd7a467 3062 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3063 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3064 set_opt(sb, JOURNAL_DATA);
617ba13b 3065 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3066 set_opt(sb, ORDERED_DATA);
617ba13b 3067 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3068 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3069
3070 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3071 set_opt(sb, ERRORS_PANIC);
bb4f397a 3072 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3073 set_opt(sb, ERRORS_CONT);
bb4f397a 3074 else
fd8c37ec 3075 set_opt(sb, ERRORS_RO);
8b67f04a 3076 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3077 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3078 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3079 set_opt(sb, DISCARD);
ac27a0ec
DK		 3080
3081 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3082 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 3083 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3084 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3085 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3086
8b67f04a 3087 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3088 set_opt(sb, BARRIER);
ac27a0ec 3089
dd919b98
AK		 3090 /*
3091 * enable delayed allocation by default
3092 * Use -o nodelalloc to turn it off
3093 */
8b67f04a
TT		 3094 if (!IS_EXT3_SB(sb) &&
3095 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3096 set_opt(sb, DELALLOC);
dd919b98 3097
51ce6511
LC		 3098 /*
3099 * set default s_li_wait_mult for lazyinit, for the case there is
3100 * no mount option specified.
3101 */
3102 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3103
8b67f04a 3104 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3105 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3106 ext4_msg(sb, KERN_WARNING,
3107 "failed to parse options in superblock: %s",
3108 sbi->s_es->s_mount_opts);
3109 }
5a916be1 3110 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3111 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3112 &journal_ioprio, 0))
ac27a0ec
DK		 3113 goto failed_mount;
3114
56889787
TT		 3115 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3116 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3117 "with data=journal disables delayed "
3118 "allocation and O_DIRECT support!\n");
3119 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3120 ext4_msg(sb, KERN_ERR, "can't mount with "
3121 "both data=journal and delalloc");
3122 goto failed_mount;
3123 }
3124 if (test_opt(sb, DIOREAD_NOLOCK)) {
3125 ext4_msg(sb, KERN_ERR, "can't mount with "
3126 "both data=journal and delalloc");
3127 goto failed_mount;
3128 }
3129 if (test_opt(sb, DELALLOC))
3130 clear_opt(sb, DELALLOC);
3131 }
3132
3133 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3134 if (test_opt(sb, DIOREAD_NOLOCK)) {
3135 if (blocksize < PAGE_SIZE) {
3136 ext4_msg(sb, KERN_ERR, "can't mount with "
3137 "dioread_nolock if block size != PAGE_SIZE");
3138 goto failed_mount;
3139 }
3140 }
3141
ac27a0ec 3142 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3143 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3144
617ba13b
MC		 3145 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3146 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3147 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3148 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3149 ext4_msg(sb, KERN_WARNING,
3150 "feature flags set on rev 0 fs, "
3151 "running e2fsck is recommended");
469108ff 3152
2035e776
TT		 3153 if (IS_EXT2_SB(sb)) {
3154 if (ext2_feature_set_ok(sb))
3155 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3156 "using the ext4 subsystem");
3157 else {
3158 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3159 "to feature incompatibilities");
3160 goto failed_mount;
3161 }
3162 }
3163
3164 if (IS_EXT3_SB(sb)) {
3165 if (ext3_feature_set_ok(sb))
3166 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3167 "using the ext4 subsystem");
3168 else {
3169 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3170 "to feature incompatibilities");
3171 goto failed_mount;
3172 }
3173 }
3174
ac27a0ec
DK		 3175 /*
3176 * Check feature flags regardless of the revision level, since we
3177 * previously didn't change the revision level when setting the flags,
3178 * so there is a chance incompat flags are set on a rev 0 filesystem.
3179 */
a13fb1a4 3180 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3181 goto failed_mount;
a13fb1a4 3182
617ba13b
MC		 3183 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3184 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3185 ext4_msg(sb, KERN_ERR,
3186 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3187 goto failed_mount;
3188 }
3189
ac27a0ec 3190 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3191 /* Validate the filesystem blocksize */
3192 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3193 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3194 blocksize);
ac27a0ec
DK		 3195 goto failed_mount;
3196 }
3197
2b2d6d01 3198 brelse(bh);
70bbb3e0
AM		 3199 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3200 offset = do_div(logical_sb_block, blocksize);
3201 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3202 if (!bh) {
b31e1552
ES		 3203 ext4_msg(sb, KERN_ERR,
3204 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3205 goto failed_mount;
3206 }
617ba13b 3207 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 3208 sbi->s_es = es;
617ba13b 3209 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3210 ext4_msg(sb, KERN_ERR,
3211 "Magic mismatch, very weird!");
ac27a0ec
DK		 3212 goto failed_mount;
3213 }
3214 }
3215
a13fb1a4
ES		 3216 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3217 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3218 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3219 has_huge_files);
3220 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3221
617ba13b
MC		 3222 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3223 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3224 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3225 } else {
3226 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3227 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3228 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3229 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3230 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3231 ext4_msg(sb, KERN_ERR,
3232 "unsupported inode size: %d",
2b2d6d01 3233 sbi->s_inode_size);
ac27a0ec
DK		 3234 goto failed_mount;
3235 }
ef7f3835
KS		 3236 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3237 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3238 }
0b8e58a1 3239
0d1ee42f
AR		 3240 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3241 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3242 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3243 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3244 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3245 ext4_msg(sb, KERN_ERR,
3246 "unsupported descriptor size %lu",
0d1ee42f
AR		 3247 sbi->s_desc_size);
3248 goto failed_mount;
3249 }
3250 } else
3251 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3252
ac27a0ec 3253 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3254 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3255 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3256 goto cantfind_ext4;
0b8e58a1 3257
617ba13b 3258 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3259 if (sbi->s_inodes_per_block == 0)
617ba13b 3260 goto cantfind_ext4;
ac27a0ec
DK		 3261 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3262 sbi->s_inodes_per_block;
0d1ee42f 3263 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3264 sbi->s_sbh = bh;
3265 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3266 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3267 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3268
2b2d6d01 3269 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3270 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3271 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3272 i = le32_to_cpu(es->s_flags);
3273 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3274 sbi->s_hash_unsigned = 3;
3275 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3276#ifdef __CHAR_UNSIGNED__
3277 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3278 sbi->s_hash_unsigned = 3;
3279#else
3280 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3281#endif
f99b2589 3282 }
ac27a0ec 3283
281b5995
TT		 3284 /* Handle clustersize */
3285 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3286 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3287 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3288 if (has_bigalloc) {
3289 if (clustersize < blocksize) {
3290 ext4_msg(sb, KERN_ERR,
3291 "cluster size (%d) smaller than "
3292 "block size (%d)", clustersize, blocksize);
3293 goto failed_mount;
3294 }
3295 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3296 le32_to_cpu(es->s_log_block_size);
3297 sbi->s_clusters_per_group =
3298 le32_to_cpu(es->s_clusters_per_group);
3299 if (sbi->s_clusters_per_group > blocksize * 8) {
3300 ext4_msg(sb, KERN_ERR,
3301 "#clusters per group too big: %lu",
3302 sbi->s_clusters_per_group);
3303 goto failed_mount;
3304 }
3305 if (sbi->s_blocks_per_group !=
3306 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3307 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3308 "clusters per group (%lu) inconsistent",
3309 sbi->s_blocks_per_group,
3310 sbi->s_clusters_per_group);
3311 goto failed_mount;
3312 }
3313 } else {
3314 if (clustersize != blocksize) {
3315 ext4_warning(sb, "fragment/cluster size (%d) != "
3316 "block size (%d)", clustersize,
3317 blocksize);
3318 clustersize = blocksize;
3319 }
3320 if (sbi->s_blocks_per_group > blocksize * 8) {
3321 ext4_msg(sb, KERN_ERR,
3322 "#blocks per group too big: %lu",
3323 sbi->s_blocks_per_group);
3324 goto failed_mount;
3325 }
3326 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3327 sbi->s_cluster_bits = 0;
ac27a0ec 3328 }
281b5995
TT		 3329 sbi->s_cluster_ratio = clustersize / blocksize;
3330
ac27a0ec 3331 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3332 ext4_msg(sb, KERN_ERR,
3333 "#inodes per group too big: %lu",
2b2d6d01 3334 sbi->s_inodes_per_group);
ac27a0ec
DK		 3335 goto failed_mount;
3336 }
3337
bf43d84b
ES		 3338 /*
3339 * Test whether we have more sectors than will fit in sector_t,
3340 * and whether the max offset is addressable by the page cache.
3341 */
5a9ae68a 3342 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3343 ext4_blocks_count(es));
5a9ae68a 3344 if (err) {
b31e1552 3345 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3346 " too large to mount safely on this system");
ac27a0ec 3347 if (sizeof(sector_t) < 8)
90c699a9 3348 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3349 ret = err;
ac27a0ec
DK		 3350 goto failed_mount;
3351 }
3352
617ba13b
MC		 3353 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3354 goto cantfind_ext4;
e7c95593 3355
0f2ddca6
FTN		 3356 /* check blocks count against device size */
3357 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3358 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3359 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3360 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3361 ext4_blocks_count(es), blocks_count);
3362 goto failed_mount;
3363 }
3364
0b8e58a1
AD		 3365 /*
3366 * It makes no sense for the first data block to be beyond the end
3367 * of the filesystem.
3368 */
3369 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3370 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3371 "block %u is beyond end of filesystem (%llu)",
3372 le32_to_cpu(es->s_first_data_block),
3373 ext4_blocks_count(es));
e7c95593
ES		 3374 goto failed_mount;
3375 }
bd81d8ee
LV		 3376 blocks_count = (ext4_blocks_count(es) -
3377 le32_to_cpu(es->s_first_data_block) +
3378 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3379 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3380 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3381 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3382 "(block count %llu, first data block %u, "
b31e1552 3383 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3384 ext4_blocks_count(es),
3385 le32_to_cpu(es->s_first_data_block),
3386 EXT4_BLOCKS_PER_GROUP(sb));
3387 goto failed_mount;
3388 }
bd81d8ee 3389 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3390 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3391 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3392 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3393 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3394 sbi->s_group_desc = ext4_kvmalloc(db_count *
3395 sizeof(struct buffer_head *),
3396 GFP_KERNEL);
ac27a0ec 3397 if (sbi->s_group_desc == NULL) {
b31e1552 3398 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 3399 goto failed_mount;
3400 }
3401
9f6200bb
TT		 3402 if (ext4_proc_root)
3403 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3404
66acdcf4
TT		 3405 if (sbi->s_proc)
3406 proc_create_data("options", S_IRUGO, sbi->s_proc,
3407 &ext4_seq_options_fops, sb);
3408
705895b6 3409 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3410
3411 for (i = 0; i < db_count; i++) {
70bbb3e0 3412 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3413 sbi->s_group_desc[i] = sb_bread(sb, block);
3414 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3415 ext4_msg(sb, KERN_ERR,
3416 "can't read group descriptor %d", i);
ac27a0ec
DK		 3417 db_count = i;
3418 goto failed_mount2;
3419 }
3420 }
bfff6873 3421 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3422 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3423 goto failed_mount2;
3424 }
772cb7c8
JS		 3425 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3426 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3427 ext4_msg(sb, KERN_ERR,
3428 "unable to initialize "
3429 "flex_bg meta info!");
772cb7c8
JS		 3430 goto failed_mount2;
3431 }
3432
ac27a0ec
DK		 3433 sbi->s_gdb_count = db_count;
3434 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3435 spin_lock_init(&sbi->s_next_gen_lock);
3436
04496411
TM		 3437 init_timer(&sbi->s_err_report);
3438 sbi->s_err_report.function = print_daily_error_info;
3439 sbi->s_err_report.data = (unsigned long) sb;
3440
57042651 3441 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3442 ext4_count_free_clusters(sb));
ce7e010a
TT		 3443 if (!err) {
3444 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3445 ext4_count_free_inodes(sb));
3446 }
3447 if (!err) {
3448 err = percpu_counter_init(&sbi->s_dirs_counter,
3449 ext4_count_dirs(sb));
3450 }
3451 if (!err) {
57042651 3452 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3453 }
3454 if (err) {
3455 ext4_msg(sb, KERN_ERR, "insufficient memory");
3456 goto failed_mount3;
3457 }
3458
c9de560d 3459 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3460 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3461
ac27a0ec
DK		 3462 /*
3463 * set up enough so that it can read an inode
3464 */
9ca92389
TT		 3465 if (!test_opt(sb, NOLOAD) &&
3466 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3467 sb->s_op = &ext4_sops;
3468 else
3469 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3470 sb->s_export_op = &ext4_export_ops;
3471 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3472#ifdef CONFIG_QUOTA
617ba13b
MC		 3473 sb->s_qcop = &ext4_qctl_operations;
3474 sb->dq_op = &ext4_quota_operations;
ac27a0ec 3475#endif
f2fa2ffc
AK		 3476 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3477
ac27a0ec 3478 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3479 mutex_init(&sbi->s_orphan_lock);
8f82f840 3480 sbi->s_resize_flags = 0;
ac27a0ec
DK		 3481
3482 sb->s_root = NULL;
3483
3484 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3485 EXT4_HAS_INCOMPAT_FEATURE(sb,
3486 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3487
c5e06d10
JL		 3488 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3489 !(sb->s_flags & MS_RDONLY))
3490 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3491 goto failed_mount3;
3492
ac27a0ec
DK		 3493 /*
3494 * The first inode we look at is the journal inode. Don't try
3495 * root first: it may be modified in the journal!
3496 */
3497 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3498 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3499 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3500 goto failed_mount3;
0390131b
FM		 3501 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3502 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3503 ext4_msg(sb, KERN_ERR, "required journal recovery "
3504 "suppressed and not mounted read-only");
744692dc 3505 goto failed_mount_wq;
ac27a0ec 3506 } else {
fd8c37ec 3507 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3508 sbi->s_journal = NULL;
3509 needs_recovery = 0;
3510 goto no_journal;
ac27a0ec
DK		 3511 }
3512
eb40a09c
JS		 3513 if (ext4_blocks_count(es) > 0xffffffffULL &&
3514 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3515 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3516 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3517 goto failed_mount_wq;
eb40a09c
JS		 3518 }
3519
d4da6c9c
LT		 3520 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3521 jbd2_journal_set_features(sbi->s_journal,
3522 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 3523 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3524 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3525 jbd2_journal_set_features(sbi->s_journal,
3526 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 3527 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3528 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3529 } else {
3530 jbd2_journal_clear_features(sbi->s_journal,
3531 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3532 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3533 }
818d276c 3534
ac27a0ec
DK		 3535 /* We have now updated the journal if required, so we can
3536 * validate the data journaling mode. */
3537 switch (test_opt(sb, DATA_FLAGS)) {
3538 case 0:
3539 /* No mode set, assume a default based on the journal
63f57933
AM		 3540 * capabilities: ORDERED_DATA if the journal can
3541 * cope, else JOURNAL_DATA
3542 */
dab291af
MC		 3543 if (jbd2_journal_check_available_features
3544 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3545 set_opt(sb, ORDERED_DATA);
ac27a0ec 3546 else
fd8c37ec 3547 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3548 break;
3549
617ba13b
MC		 3550 case EXT4_MOUNT_ORDERED_DATA:
3551 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3552 if (!jbd2_journal_check_available_features
3553 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3554 ext4_msg(sb, KERN_ERR, "Journal does not support "
3555 "requested data journaling mode");
744692dc 3556 goto failed_mount_wq;
ac27a0ec
DK		 3557 }
3558 default:
3559 break;
3560 }
b3881f74 3561 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3562
18aadd47
BJ		 3563 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3564
ce7e010a
TT		 3565 /*
3566 * The journal may have updated the bg summary counts, so we
3567 * need to update the global counters.
3568 */
57042651 3569 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3570 ext4_count_free_clusters(sb));
ce7e010a
TT		 3571 percpu_counter_set(&sbi->s_freeinodes_counter,
3572 ext4_count_free_inodes(sb));
3573 percpu_counter_set(&sbi->s_dirs_counter,
3574 ext4_count_dirs(sb));
57042651 3575 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3576
ce7e010a 3577no_journal:
fd89d5f2
TH		 3578 /*
3579 * The maximum number of concurrent works can be high and
3580 * concurrency isn't really necessary. Limit it to 1.
3581 */
3582 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3583 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3584 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3585 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3586 goto failed_mount_wq;
3587 }
3588
ac27a0ec 3589 /*
dab291af 3590 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3591 * so we can safely mount the rest of the filesystem now.
3592 */
3593
1d1fe1ee
DH		 3594 root = ext4_iget(sb, EXT4_ROOT_INO);
3595 if (IS_ERR(root)) {
b31e1552 3596 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3597 ret = PTR_ERR(root);
32a9bb57 3598 root = NULL;
ac27a0ec
DK		 3599 goto failed_mount4;
3600 }
3601 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3602 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3603 iput(root);
ac27a0ec
DK		 3604 goto failed_mount4;
3605 }
48fde701 3606 sb->s_root = d_make_root(root);
1d1fe1ee 3607 if (!sb->s_root) {
b31e1552 3608 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3609 ret = -ENOMEM;
3610 goto failed_mount4;
3611 }
ac27a0ec 3612
2b2d6d01 3613 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 3614
3615 /* determine the minimum size of new large inodes, if present */
3616 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3617 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3618 EXT4_GOOD_OLD_INODE_SIZE;
3619 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3620 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3621 if (sbi->s_want_extra_isize <
3622 le16_to_cpu(es->s_want_extra_isize))
3623 sbi->s_want_extra_isize =
3624 le16_to_cpu(es->s_want_extra_isize);
3625 if (sbi->s_want_extra_isize <
3626 le16_to_cpu(es->s_min_extra_isize))
3627 sbi->s_want_extra_isize =
3628 le16_to_cpu(es->s_min_extra_isize);
3629 }
3630 }
3631 /* Check if enough inode space is available */
3632 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3633 sbi->s_inode_size) {
3634 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3635 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3636 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3637 "available");
ef7f3835
KS		 3638 }
3639
6fd058f7
TT		 3640 err = ext4_setup_system_zone(sb);
3641 if (err) {
b31e1552 3642 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3643 "zone (%d)", err);
94bf608a 3644 goto failed_mount4a;
6fd058f7
TT		 3645 }
3646
c2774d84
AK		 3647 ext4_ext_init(sb);
3648 err = ext4_mb_init(sb, needs_recovery);
3649 if (err) {
421f91d2 3650 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3651 err);
dcf2d804 3652 goto failed_mount5;
c2774d84
AK		 3653 }
3654
bfff6873
LC		 3655 err = ext4_register_li_request(sb, first_not_zeroed);
3656 if (err)
dcf2d804 3657 goto failed_mount6;
bfff6873 3658
3197ebdb
TT		 3659 sbi->s_kobj.kset = ext4_kset;
3660 init_completion(&sbi->s_kobj_unregister);
3661 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3662 "%s", sb->s_id);
dcf2d804
TM		 3663 if (err)
3664 goto failed_mount7;
3197ebdb 3665
617ba13b
MC		 3666 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3667 ext4_orphan_cleanup(sb, es);
3668 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3669 if (needs_recovery) {
b31e1552 3670 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3671 ext4_mark_recovery_complete(sb, es);
3672 }
3673 if (EXT4_SB(sb)->s_journal) {
3674 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3675 descr = " journalled data mode";
3676 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3677 descr = " ordered data mode";
3678 else
3679 descr = " writeback data mode";
3680 } else
3681 descr = "out journal";
3682
d4c402d9 3683 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3684 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3685 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3686
66e61a9e
TT		 3687 if (es->s_error_count)
3688 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3689
d4c402d9 3690 kfree(orig_data);
ac27a0ec
DK		 3691 return 0;
3692
617ba13b 3693cantfind_ext4:
ac27a0ec 3694 if (!silent)
b31e1552 3695 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3696 goto failed_mount;
3697
dcf2d804
TM		 3698failed_mount7:
3699 ext4_unregister_li_request(sb);
3700failed_mount6:
dcf2d804 3701 ext4_mb_release(sb);
94bf608a
AV		 3702failed_mount5:
3703 ext4_ext_release(sb);
dcf2d804 3704 ext4_release_system_zone(sb);
94bf608a
AV		 3705failed_mount4a:
3706 dput(sb->s_root);
32a9bb57 3707 sb->s_root = NULL;
94bf608a 3708failed_mount4:
b31e1552 3709 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3710 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3711failed_mount_wq:
0390131b
FM		 3712 if (sbi->s_journal) {
3713 jbd2_journal_destroy(sbi->s_journal);
3714 sbi->s_journal = NULL;
3715 }
ac27a0ec 3716failed_mount3:
04496411 3717 del_timer(&sbi->s_err_report);
9933fc0a
TT		 3718 if (sbi->s_flex_groups)
3719 ext4_kvfree(sbi->s_flex_groups);
57042651 3720 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 3721 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3722 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 3723 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 3724 if (sbi->s_mmp_tsk)
3725 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 3726failed_mount2:
3727 for (i = 0; i < db_count; i++)
3728 brelse(sbi->s_group_desc[i]);
f18a5f21 3729 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 3730failed_mount:
240799cd 3731 if (sbi->s_proc) {
66acdcf4 3732 remove_proc_entry("options", sbi->s_proc);
9f6200bb 3733 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3734 }
ac27a0ec
DK		 3735#ifdef CONFIG_QUOTA
3736 for (i = 0; i < MAXQUOTAS; i++)
3737 kfree(sbi->s_qf_names[i]);
3738#endif
617ba13b 3739 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3740 brelse(bh);
3741out_fail:
3742 sb->s_fs_info = NULL;
f6830165 3743 kfree(sbi->s_blockgroup_lock);
ac27a0ec 3744 kfree(sbi);
dcc7dae3 3745out_free_orig:
d4c402d9 3746 kfree(orig_data);
1d1fe1ee 3747 return ret;
ac27a0ec
DK		 3748}
3749
3750/*
3751 * Setup any per-fs journal parameters now. We'll do this both on
3752 * initial mount, once the journal has been initialised but before we've
3753 * done any recovery; and again on any subsequent remount.
3754 */
617ba13b 3755static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3756{
617ba13b 3757 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3758
30773840
TT		 3759 journal->j_commit_interval = sbi->s_commit_interval;
3760 journal->j_min_batch_time = sbi->s_min_batch_time;
3761 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 3762
a931da6a 3763 write_lock(&journal->j_state_lock);
ac27a0ec 3764 if (test_opt(sb, BARRIER))
dab291af 3765 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3766 else
dab291af 3767 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3768 if (test_opt(sb, DATA_ERR_ABORT))
3769 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3770 else
3771 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 3772 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 3773}
3774
617ba13b 3775static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3776 unsigned int journal_inum)
3777{
3778 struct inode *journal_inode;
3779 journal_t *journal;
3780
0390131b
FM		 3781 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3782
ac27a0ec
DK		 3783 /* First, test for the existence of a valid inode on disk. Bad
3784 * things happen if we iget() an unused inode, as the subsequent
3785 * iput() will try to delete it. */
3786
1d1fe1ee
DH		 3787 journal_inode = ext4_iget(sb, journal_inum);
3788 if (IS_ERR(journal_inode)) {
b31e1552 3789 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3790 return NULL;
3791 }
3792 if (!journal_inode->i_nlink) {
3793 make_bad_inode(journal_inode);
3794 iput(journal_inode);
b31e1552 3795 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3796 return NULL;
3797 }
3798
e5f8eab8 3799 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3800 journal_inode, journal_inode->i_size);
1d1fe1ee 3801 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3802 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3803 iput(journal_inode);
3804 return NULL;
3805 }
3806
dab291af 3807 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3808 if (!journal) {
b31e1552 3809 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3810 iput(journal_inode);
3811 return NULL;
3812 }
3813 journal->j_private = sb;
617ba13b 3814 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3815 return journal;
3816}
3817
617ba13b 3818static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3819 dev_t j_dev)
3820{
2b2d6d01 3821 struct buffer_head *bh;
ac27a0ec 3822 journal_t *journal;
617ba13b
MC		 3823 ext4_fsblk_t start;
3824 ext4_fsblk_t len;
ac27a0ec 3825 int hblock, blocksize;
617ba13b 3826 ext4_fsblk_t sb_block;
ac27a0ec 3827 unsigned long offset;
2b2d6d01 3828 struct ext4_super_block *es;
ac27a0ec
DK		 3829 struct block_device *bdev;
3830
0390131b
FM		 3831 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3832
b31e1552 3833 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3834 if (bdev == NULL)
3835 return NULL;
3836
ac27a0ec 3837 blocksize = sb->s_blocksize;
e1defc4f 3838 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3839 if (blocksize < hblock) {
b31e1552
ES		 3840 ext4_msg(sb, KERN_ERR,
3841 "blocksize too small for journal device");
ac27a0ec
DK		 3842 goto out_bdev;
3843 }
3844
617ba13b
MC		 3845 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3846 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3847 set_blocksize(bdev, blocksize);
3848 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3849 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3850 "external journal");
ac27a0ec
DK		 3851 goto out_bdev;
3852 }
3853
617ba13b
MC		 3854 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3855 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3856 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3857 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3858 ext4_msg(sb, KERN_ERR, "external journal has "
3859 "bad superblock");
ac27a0ec
DK		 3860 brelse(bh);
3861 goto out_bdev;
3862 }
3863
617ba13b 3864 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3865 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3866 brelse(bh);
3867 goto out_bdev;
3868 }
3869
bd81d8ee 3870 len = ext4_blocks_count(es);
ac27a0ec
DK		 3871 start = sb_block + 1;
3872 brelse(bh); /* we're done with the superblock */
3873
dab291af 3874 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3875 start, len, blocksize);
3876 if (!journal) {
b31e1552 3877 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3878 goto out_bdev;
3879 }
3880 journal->j_private = sb;
3881 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3882 wait_on_buffer(journal->j_sb_buffer);
3883 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3884 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3885 goto out_journal;
3886 }
3887 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3888 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3889 "user (unsupported) - %d",
ac27a0ec
DK		 3890 be32_to_cpu(journal->j_superblock->s_nr_users));
3891 goto out_journal;
3892 }
617ba13b
MC		 3893 EXT4_SB(sb)->journal_bdev = bdev;
3894 ext4_init_journal_params(sb, journal);
ac27a0ec 3895 return journal;
0b8e58a1 3896
ac27a0ec 3897out_journal:
dab291af 3898 jbd2_journal_destroy(journal);
ac27a0ec 3899out_bdev:
617ba13b 3900 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3901 return NULL;
3902}
3903
617ba13b
MC		 3904static int ext4_load_journal(struct super_block *sb,
3905 struct ext4_super_block *es,
ac27a0ec
DK		 3906 unsigned long journal_devnum)
3907{
3908 journal_t *journal;
3909 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3910 dev_t journal_dev;
3911 int err = 0;
3912 int really_read_only;
3913
0390131b
FM		 3914 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3915
ac27a0ec
DK		 3916 if (journal_devnum &&
3917 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3918 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3919 "numbers have changed");
ac27a0ec
DK		 3920 journal_dev = new_decode_dev(journal_devnum);
3921 } else
3922 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3923
3924 really_read_only = bdev_read_only(sb->s_bdev);
3925
3926 /*
3927 * Are we loading a blank journal or performing recovery after a
3928 * crash? For recovery, we need to check in advance whether we
3929 * can get read-write access to the device.
3930 */
617ba13b 3931 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3932 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3933 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3934 "required on readonly filesystem");
ac27a0ec 3935 if (really_read_only) {
b31e1552
ES		 3936 ext4_msg(sb, KERN_ERR, "write access "
3937 "unavailable, cannot proceed");
ac27a0ec
DK		 3938 return -EROFS;
3939 }
b31e1552
ES		 3940 ext4_msg(sb, KERN_INFO, "write access will "
3941 "be enabled during recovery");
ac27a0ec
DK		 3942 }
3943 }
3944
3945 if (journal_inum && journal_dev) {
b31e1552
ES		 3946 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3947 "and inode journals!");
ac27a0ec
DK		 3948 return -EINVAL;
3949 }
3950
3951 if (journal_inum) {
617ba13b 3952 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3953 return -EINVAL;
3954 } else {
617ba13b 3955 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3956 return -EINVAL;
3957 }
3958
90576c0b 3959 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 3960 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3961
617ba13b 3962 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3963 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 3964 if (!err) {
3965 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3966 if (save)
3967 memcpy(save, ((char *) es) +
3968 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 3969 err = jbd2_journal_load(journal);
1c13d5c0
TT		 3970 if (save)
3971 memcpy(((char *) es) + EXT4_S_ERR_START,
3972 save, EXT4_S_ERR_LEN);
3973 kfree(save);
3974 }
ac27a0ec
DK		 3975
3976 if (err) {
b31e1552 3977 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3978 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3979 return err;
3980 }
3981
617ba13b
MC		 3982 EXT4_SB(sb)->s_journal = journal;
3983 ext4_clear_journal_err(sb, es);
ac27a0ec 3984
c41303ce 3985 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 3986 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3987 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3988
3989 /* Make sure we flush the recovery flag to disk. */
e2d67052 3990 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3991 }
3992
3993 return 0;
3994}
3995
e2d67052 3996static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3997{
e2d67052 3998 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3999 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4000 int error = 0;
ac27a0ec 4001
7c2e7087 4002 if (!sbh || block_device_ejected(sb))
c4be0c1d 4003 return error;
914258bf
TT		 4004 if (buffer_write_io_error(sbh)) {
4005 /*
4006 * Oh, dear. A previous attempt to write the
4007 * superblock failed. This could happen because the
4008 * USB device was yanked out. Or it could happen to
4009 * be a transient write error and maybe the block will
4010 * be remapped. Nothing we can do but to retry the
4011 * write and hope for the best.
4012 */
b31e1552
ES		 4013 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4014 "superblock detected");
914258bf
TT		 4015 clear_buffer_write_io_error(sbh);
4016 set_buffer_uptodate(sbh);
4017 }
71290b36
TT		 4018 /*
4019 * If the file system is mounted read-only, don't update the
4020 * superblock write time. This avoids updating the superblock
4021 * write time when we are mounting the root file system
4022 * read/only but we need to replay the journal; at that point,
4023 * for people who are east of GMT and who make their clock
4024 * tick in localtime for Windows bug-for-bug compatibility,
4025 * the clock is set in the future, and this will cause e2fsck
4026 * to complain and force a full file system check.
4027 */
4028 if (!(sb->s_flags & MS_RDONLY))
4029 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4030 if (sb->s_bdev->bd_part)
4031 es->s_kbytes_written =
4032 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4033 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4034 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4035 else
4036 es->s_kbytes_written =
4037 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4038 ext4_free_blocks_count_set(es,
4039 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4040 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4041 es->s_free_inodes_count =
4042 cpu_to_le32(percpu_counter_sum_positive(
4043 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 4044 sb->s_dirt = 0;
ac27a0ec
DK		 4045 BUFFER_TRACE(sbh, "marking dirty");
4046 mark_buffer_dirty(sbh);
914258bf 4047 if (sync) {
c4be0c1d
TS		 4048 error = sync_dirty_buffer(sbh);
4049 if (error)
4050 return error;
4051
4052 error = buffer_write_io_error(sbh);
4053 if (error) {
b31e1552
ES		 4054 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4055 "superblock");
914258bf
TT		 4056 clear_buffer_write_io_error(sbh);
4057 set_buffer_uptodate(sbh);
4058 }
4059 }
c4be0c1d 4060 return error;
ac27a0ec
DK		 4061}
4062
ac27a0ec
DK		 4063/*
4064 * Have we just finished recovery? If so, and if we are mounting (or
4065 * remounting) the filesystem readonly, then we will end up with a
4066 * consistent fs on disk. Record that fact.
4067 */
2b2d6d01
TT		 4068static void ext4_mark_recovery_complete(struct super_block *sb,
4069 struct ext4_super_block *es)
ac27a0ec 4070{
617ba13b 4071 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4072
0390131b
FM		 4073 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4074 BUG_ON(journal != NULL);
4075 return;
4076 }
dab291af 4077 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4078 if (jbd2_journal_flush(journal) < 0)
4079 goto out;
4080
617ba13b 4081 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4082 sb->s_flags & MS_RDONLY) {
617ba13b 4083 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4084 ext4_commit_super(sb, 1);
ac27a0ec 4085 }
7ffe1ea8
HK		 4086
4087out:
dab291af 4088 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4089}
4090
4091/*
4092 * If we are mounting (or read-write remounting) a filesystem whose journal
4093 * has recorded an error from a previous lifetime, move that error to the
4094 * main filesystem now.
4095 */
2b2d6d01
TT		 4096static void ext4_clear_journal_err(struct super_block *sb,
4097 struct ext4_super_block *es)
ac27a0ec
DK		 4098{
4099 journal_t *journal;
4100 int j_errno;
4101 const char *errstr;
4102
0390131b
FM		 4103 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4104
617ba13b 4105 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4106
4107 /*
4108 * Now check for any error status which may have been recorded in the
617ba13b 4109 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4110 */
4111
dab291af 4112 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4113 if (j_errno) {
4114 char nbuf[16];
4115
617ba13b 4116 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4117 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4118 "from previous mount: %s", errstr);
12062ddd 4119 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4120
617ba13b
MC		 4121 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4122 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4123 ext4_commit_super(sb, 1);
ac27a0ec 4124
dab291af 4125 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 4126 }
4127}
4128
4129/*
4130 * Force the running and committing transactions to commit,
4131 * and wait on the commit.
4132 */
617ba13b 4133int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4134{
4135 journal_t *journal;
0390131b 4136 int ret = 0;
ac27a0ec
DK		 4137
4138 if (sb->s_flags & MS_RDONLY)
4139 return 0;
4140
617ba13b 4141 journal = EXT4_SB(sb)->s_journal;
6b0310fb 4142 if (journal) {
437f88cc 4143 vfs_check_frozen(sb, SB_FREEZE_TRANS);
0390131b 4144 ret = ext4_journal_force_commit(journal);
6b0310fb 4145 }
0390131b 4146
ac27a0ec
DK		 4147 return ret;
4148}
4149
2b2d6d01 4150static void ext4_write_super(struct super_block *sb)
ac27a0ec 4151{
ebc1ac16 4152 lock_super(sb);
9ca92389 4153 ext4_commit_super(sb, 1);
ebc1ac16 4154 unlock_super(sb);
ac27a0ec
DK		 4155}
4156
617ba13b 4157static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4158{
14ce0cb4 4159 int ret = 0;
9eddacf9 4160 tid_t target;
8d5d02e6 4161 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4162
9bffad1e 4163 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 4164 flush_workqueue(sbi->dio_unwritten_wq);
4165 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4166 if (wait)
8d5d02e6 4167 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4168 }
14ce0cb4 4169 return ret;
ac27a0ec
DK		 4170}
4171
4172/*
4173 * LVM calls this function before a (read-only) snapshot is created. This
4174 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4175 *
4176 * Note that only this function cannot bring a filesystem to be in a clean
4177 * state independently, because ext4 prevents a new handle from being started
4178 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4179 * the upper layer.
ac27a0ec 4180 */
c4be0c1d 4181static int ext4_freeze(struct super_block *sb)
ac27a0ec 4182{
c4be0c1d
TS		 4183 int error = 0;
4184 journal_t *journal;
ac27a0ec 4185
9ca92389
TT		 4186 if (sb->s_flags & MS_RDONLY)
4187 return 0;
ac27a0ec 4188
9ca92389 4189 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4190
9ca92389
TT		 4191 /* Now we set up the journal barrier. */
4192 jbd2_journal_lock_updates(journal);
ac27a0ec 4193
9ca92389
TT		 4194 /*
4195 * Don't clear the needs_recovery flag if we failed to flush
4196 * the journal.
4197 */
4198 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4199 if (error < 0)
4200 goto out;
9ca92389
TT		 4201
4202 /* Journal blocked and flushed, clear needs_recovery flag. */
4203 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4204 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 4205out:
4206 /* we rely on s_frozen to stop further updates */
4207 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4208 return error;
ac27a0ec
DK		 4209}
4210
4211/*
4212 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4213 * flag here, even though the filesystem is not technically dirty yet.
4214 */
c4be0c1d 4215static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4216{
9ca92389
TT		 4217 if (sb->s_flags & MS_RDONLY)
4218 return 0;
4219
4220 lock_super(sb);
4221 /* Reset the needs_recovery flag before the fs is unlocked. */
4222 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4223 ext4_commit_super(sb, 1);
4224 unlock_super(sb);
c4be0c1d 4225 return 0;
ac27a0ec
DK		 4226}
4227
673c6100
TT		 4228/*
4229 * Structure to save mount options for ext4_remount's benefit
4230 */
4231struct ext4_mount_options {
4232 unsigned long s_mount_opt;
a2595b8a 4233 unsigned long s_mount_opt2;
673c6100
TT		 4234 uid_t s_resuid;
4235 gid_t s_resgid;
4236 unsigned long s_commit_interval;
4237 u32 s_min_batch_time, s_max_batch_time;
4238#ifdef CONFIG_QUOTA
4239 int s_jquota_fmt;
4240 char *s_qf_names[MAXQUOTAS];
4241#endif
4242};
4243
2b2d6d01 4244static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4245{
2b2d6d01 4246 struct ext4_super_block *es;
617ba13b 4247 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4248 unsigned long old_sb_flags;
617ba13b 4249 struct ext4_mount_options old_opts;
c79d967d 4250 int enable_quota = 0;
8a266467 4251 ext4_group_t g;
b3881f74 4252 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4253 int err = 0;
ac27a0ec
DK		 4254#ifdef CONFIG_QUOTA
4255 int i;
4256#endif
d4c402d9 4257 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4258
4259 /* Store the original options */
bbd6851a 4260 lock_super(sb);
ac27a0ec
DK		 4261 old_sb_flags = sb->s_flags;
4262 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4263 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4264 old_opts.s_resuid = sbi->s_resuid;
4265 old_opts.s_resgid = sbi->s_resgid;
4266 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4267 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4268 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4269#ifdef CONFIG_QUOTA
4270 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4271 for (i = 0; i < MAXQUOTAS; i++)
4272 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4273#endif
b3881f74
TT		 4274 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4275 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4276
4277 /*
4278 * Allow the "check" option to be passed as a remount option.
4279 */
661aa520 4280 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4281 err = -EINVAL;
4282 goto restore_opts;
4283 }
4284
4ab2f15b 4285 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4286 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4287
4288 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4289 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4290
4291 es = sbi->s_es;
4292
b3881f74 4293 if (sbi->s_journal) {
0390131b 4294 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4295 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4296 }
ac27a0ec 4297
661aa520 4298 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4299 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4300 err = -EROFS;
4301 goto restore_opts;
4302 }
4303
4304 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4305 err = dquot_suspend(sb, -1);
4306 if (err < 0)
c79d967d 4307 goto restore_opts;
c79d967d 4308
ac27a0ec
DK		 4309 /*
4310 * First of all, the unconditional stuff we have to do
4311 * to disable replay of the journal when we next remount
4312 */
4313 sb->s_flags |= MS_RDONLY;
4314
4315 /*
4316 * OK, test if we are remounting a valid rw partition
4317 * readonly, and if so set the rdonly flag and then
4318 * mark the partition as valid again.
4319 */
617ba13b
MC		 4320 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4321 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4322 es->s_state = cpu_to_le16(sbi->s_mount_state);
4323
a63c9eb2 4324 if (sbi->s_journal)
0390131b 4325 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4326 } else {
a13fb1a4
ES		 4327 /* Make sure we can mount this feature set readwrite */
4328 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4329 err = -EROFS;
4330 goto restore_opts;
4331 }
8a266467
TT		 4332 /*
4333 * Make sure the group descriptor checksums
0b8e58a1 4334 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4335 */
4336 for (g = 0; g < sbi->s_groups_count; g++) {
4337 struct ext4_group_desc *gdp =
4338 ext4_get_group_desc(sb, g, NULL);
4339
4340 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 4341 ext4_msg(sb, KERN_ERR,
4342 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4343 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4344 le16_to_cpu(gdp->bg_checksum));
4345 err = -EINVAL;
4346 goto restore_opts;
4347 }
4348 }
4349
ead6596b
ES		 4350 /*
4351 * If we have an unprocessed orphan list hanging
4352 * around from a previously readonly bdev mount,
4353 * require a full umount/remount for now.
4354 */
4355 if (es->s_last_orphan) {
b31e1552 4356 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4357 "remount RDWR because of unprocessed "
4358 "orphan inode list. Please "
b31e1552 4359 "umount/remount instead");
ead6596b
ES		 4360 err = -EINVAL;
4361 goto restore_opts;
4362 }
4363
ac27a0ec
DK		 4364 /*
4365 * Mounting a RDONLY partition read-write, so reread
4366 * and store the current valid flag. (It may have
4367 * been changed by e2fsck since we originally mounted
4368 * the partition.)
4369 */
0390131b
FM		 4370 if (sbi->s_journal)
4371 ext4_clear_journal_err(sb, es);
ac27a0ec 4372 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4373 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4374 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4375 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4376 EXT4_FEATURE_INCOMPAT_MMP))
4377 if (ext4_multi_mount_protect(sb,
4378 le64_to_cpu(es->s_mmp_block))) {
4379 err = -EROFS;
4380 goto restore_opts;
4381 }
c79d967d 4382 enable_quota = 1;
ac27a0ec
DK		 4383 }
4384 }
bfff6873
LC		 4385
4386 /*
4387 * Reinitialize lazy itable initialization thread based on
4388 * current settings
4389 */
4390 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4391 ext4_unregister_li_request(sb);
4392 else {
4393 ext4_group_t first_not_zeroed;
4394 first_not_zeroed = ext4_has_uninit_itable(sb);
4395 ext4_register_li_request(sb, first_not_zeroed);
4396 }
4397
6fd058f7 4398 ext4_setup_system_zone(sb);
0390131b 4399 if (sbi->s_journal == NULL)
e2d67052 4400 ext4_commit_super(sb, 1);
0390131b 4401
ac27a0ec
DK		 4402#ifdef CONFIG_QUOTA
4403 /* Release old quota file names */
4404 for (i = 0; i < MAXQUOTAS; i++)
4405 if (old_opts.s_qf_names[i] &&
4406 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4407 kfree(old_opts.s_qf_names[i]);
4408#endif
bbd6851a 4409 unlock_super(sb);
c79d967d 4410 if (enable_quota)
0f0dd62f 4411 dquot_resume(sb, -1);
d4c402d9
CW		 4412
4413 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4414 kfree(orig_data);
ac27a0ec 4415 return 0;
0b8e58a1 4416
ac27a0ec
DK		 4417restore_opts:
4418 sb->s_flags = old_sb_flags;
4419 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4420 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4421 sbi->s_resuid = old_opts.s_resuid;
4422 sbi->s_resgid = old_opts.s_resgid;
4423 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4424 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4425 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4426#ifdef CONFIG_QUOTA
4427 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4428 for (i = 0; i < MAXQUOTAS; i++) {
4429 if (sbi->s_qf_names[i] &&
4430 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4431 kfree(sbi->s_qf_names[i]);
4432 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4433 }
4434#endif
bbd6851a 4435 unlock_super(sb);
d4c402d9 4436 kfree(orig_data);
ac27a0ec
DK		 4437 return err;
4438}
4439
f975d6bc
TT		 4440/*
4441 * Note: calculating the overhead so we can be compatible with
4442 * historical BSD practice is quite difficult in the face of
4443 * clusters/bigalloc. This is because multiple metadata blocks from
4444 * different block group can end up in the same allocation cluster.
4445 * Calculating the exact overhead in the face of clustered allocation
4446 * requires either O(all block bitmaps) in memory or O(number of block
4447 * groups**2) in time. We will still calculate the superblock for
4448 * older file systems --- and if we come across with a bigalloc file
4449 * system with zero in s_overhead_clusters the estimate will be close to
4450 * correct especially for very large cluster sizes --- but for newer
4451 * file systems, it's better to calculate this figure once at mkfs
4452 * time, and store it in the superblock. If the superblock value is
4453 * present (even for non-bigalloc file systems), we will use it.
4454 */
2b2d6d01 4455static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4456{
4457 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4458 struct ext4_sb_info *sbi = EXT4_SB(sb);
4459 struct ext4_super_block *es = sbi->s_es;
f975d6bc 4460 struct ext4_group_desc *gdp;
960cc398 4461 u64 fsid;
d02a9391 4462 s64 bfree;
ac27a0ec 4463
5e70030d
BP		 4464 if (test_opt(sb, MINIX_DF)) {
4465 sbi->s_overhead_last = 0;
f975d6bc
TT		 4466 } else if (es->s_overhead_clusters) {
4467 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
6bc9feff 4468 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 4469 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 4470 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 4471
4472 /*
5e70030d
BP		 4473 * Compute the overhead (FS structures). This is constant
4474 * for a given filesystem unless the number of block groups
4475 * changes so we cache the previous value until it does.
ac27a0ec
DK		 4476 */
4477
4478 /*
4479 * All of the blocks before first_data_block are
4480 * overhead
4481 */
f975d6bc 4482 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
ac27a0ec
DK		 4483
4484 /*
f975d6bc 4485 * Add the overhead found in each block group
ac27a0ec
DK		 4486 */
4487 for (i = 0; i < ngroups; i++) {
f975d6bc
TT		 4488 gdp = ext4_get_group_desc(sb, i, NULL);
4489 overhead += ext4_num_overhead_clusters(sb, i, gdp);
ac27a0ec
DK		 4490 cond_resched();
4491 }
5e70030d
BP		 4492 sbi->s_overhead_last = overhead;
4493 smp_wmb();
6bc9feff 4494 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 4495 }
4496
617ba13b 4497 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4498 buf->f_bsize = sb->s_blocksize;
f975d6bc
TT		 4499 buf->f_blocks = (ext4_blocks_count(es) -
4500 EXT4_C2B(sbi, sbi->s_overhead_last));
57042651
TT		 4501 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4502 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4503 /* prevent underflow in case that few free space is available */
57042651 4504 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4505 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4506 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4507 buf->f_bavail = 0;
4508 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4509 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4510 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4511 fsid = le64_to_cpup((void *)es->s_uuid) ^
4512 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4513 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4514 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4515
ac27a0ec
DK		 4516 return 0;
4517}
4518
0b8e58a1
AD		 4519/* Helper function for writing quotas on sync - we need to start transaction
4520 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4521 * Process 1 Process 2
617ba13b 4522 * ext4_create() quota_sync()
a269eb18 4523 * jbd2_journal_start() write_dquot()
871a2931 4524 * dquot_initialize() down(dqio_mutex)
dab291af 4525 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4526 *
4527 */
4528
4529#ifdef CONFIG_QUOTA
4530
4531static inline struct inode *dquot_to_inode(struct dquot *dquot)
4532{
4533 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4534}
4535
617ba13b 4536static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4537{
4538 int ret, err;
4539 handle_t *handle;
4540 struct inode *inode;
4541
4542 inode = dquot_to_inode(dquot);
617ba13b 4543 handle = ext4_journal_start(inode,
0b8e58a1 4544 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4545 if (IS_ERR(handle))
4546 return PTR_ERR(handle);
4547 ret = dquot_commit(dquot);
617ba13b 4548 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4549 if (!ret)
4550 ret = err;
4551 return ret;
4552}
4553
617ba13b 4554static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4555{
4556 int ret, err;
4557 handle_t *handle;
4558
617ba13b 4559 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4560 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4561 if (IS_ERR(handle))
4562 return PTR_ERR(handle);
4563 ret = dquot_acquire(dquot);
617ba13b 4564 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4565 if (!ret)
4566 ret = err;
4567 return ret;
4568}
4569
617ba13b 4570static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4571{
4572 int ret, err;
4573 handle_t *handle;
4574
617ba13b 4575 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4576 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4577 if (IS_ERR(handle)) {
4578 /* Release dquot anyway to avoid endless cycle in dqput() */
4579 dquot_release(dquot);
ac27a0ec 4580 return PTR_ERR(handle);
9c3013e9 4581 }
ac27a0ec 4582 ret = dquot_release(dquot);
617ba13b 4583 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4584 if (!ret)
4585 ret = err;
4586 return ret;
4587}
4588
617ba13b 4589static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4590{
2c8be6b2 4591 /* Are we journaling quotas? */
617ba13b
MC		 4592 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4593 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4594 dquot_mark_dquot_dirty(dquot);
617ba13b 4595 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4596 } else {
4597 return dquot_mark_dquot_dirty(dquot);
4598 }
4599}
4600
617ba13b 4601static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4602{
4603 int ret, err;
4604 handle_t *handle;
4605
4606 /* Data block + inode block */
617ba13b 4607 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4608 if (IS_ERR(handle))
4609 return PTR_ERR(handle);
4610 ret = dquot_commit_info(sb, type);
617ba13b 4611 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4612 if (!ret)
4613 ret = err;
4614 return ret;
4615}
4616
4617/*
4618 * Turn on quotas during mount time - we need to find
4619 * the quota file and such...
4620 */
617ba13b 4621static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4622{
287a8095
CH		 4623 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4624 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4625}
4626
4627/*
4628 * Standard function to be called on quota_on
4629 */
617ba13b 4630static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4631 struct path *path)
ac27a0ec
DK		 4632{
4633 int err;
ac27a0ec
DK		 4634
4635 if (!test_opt(sb, QUOTA))
4636 return -EINVAL;
0623543b 4637
ac27a0ec 4638 /* Quotafile not on the same filesystem? */
d8c9584e 4639 if (path->dentry->d_sb != sb)
ac27a0ec 4640 return -EXDEV;
0623543b
JK		 4641 /* Journaling quota? */
4642 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4643 /* Quotafile not in fs root? */
f00c9e44 4644 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4645 ext4_msg(sb, KERN_WARNING,
4646 "Quota file not on filesystem root. "
4647 "Journaled quota will not work");
2b2d6d01 4648 }
0623543b
JK		 4649
4650 /*
4651 * When we journal data on quota file, we have to flush journal to see
4652 * all updates to the file when we bypass pagecache...
4653 */
0390131b 4654 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4655 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4656 /*
4657 * We don't need to lock updates but journal_flush() could
4658 * otherwise be livelocked...
4659 */
4660 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4661 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4662 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4663 if (err)
7ffe1ea8 4664 return err;
0623543b
JK		 4665 }
4666
f00c9e44 4667 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4668}
4669
ca0e05e4
DM		 4670static int ext4_quota_off(struct super_block *sb, int type)
4671{
21f97697
JK		 4672 struct inode *inode = sb_dqopt(sb)->files[type];
4673 handle_t *handle;
4674
87009d86
DM		 4675 /* Force all delayed allocation blocks to be allocated.
4676 * Caller already holds s_umount sem */
4677 if (test_opt(sb, DELALLOC))
ca0e05e4 4678 sync_filesystem(sb);
ca0e05e4 4679
0b268590
AG		 4680 if (!inode)
4681 goto out;
4682
21f97697
JK		 4683 /* Update modification times of quota files when userspace can
4684 * start looking at them */
4685 handle = ext4_journal_start(inode, 1);
4686 if (IS_ERR(handle))
4687 goto out;
4688 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4689 ext4_mark_inode_dirty(handle, inode);
4690 ext4_journal_stop(handle);
4691
4692out:
ca0e05e4
DM		 4693 return dquot_quota_off(sb, type);
4694}
4695
ac27a0ec
DK		 4696/* Read data from quotafile - avoid pagecache and such because we cannot afford
4697 * acquiring the locks... As quota files are never truncated and quota code
25985edc 4698 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 4699 * we don't have to be afraid of races */
617ba13b 4700static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4701 size_t len, loff_t off)
4702{
4703 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4704 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4705 int err = 0;
4706 int offset = off & (sb->s_blocksize - 1);
4707 int tocopy;
4708 size_t toread;
4709 struct buffer_head *bh;
4710 loff_t i_size = i_size_read(inode);
4711
4712 if (off > i_size)
4713 return 0;
4714 if (off+len > i_size)
4715 len = i_size-off;
4716 toread = len;
4717 while (toread > 0) {
4718 tocopy = sb->s_blocksize - offset < toread ?
4719 sb->s_blocksize - offset : toread;
617ba13b 4720 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4721 if (err)
4722 return err;
4723 if (!bh) /* A hole? */
4724 memset(data, 0, tocopy);
4725 else
4726 memcpy(data, bh->b_data+offset, tocopy);
4727 brelse(bh);
4728 offset = 0;
4729 toread -= tocopy;
4730 data += tocopy;
4731 blk++;
4732 }
4733 return len;
4734}
4735
4736/* Write to quotafile (we know the transaction is already started and has
4737 * enough credits) */
617ba13b 4738static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4739 const char *data, size_t len, loff_t off)
4740{
4741 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4742 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4743 int err = 0;
4744 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 4745 struct buffer_head *bh;
4746 handle_t *handle = journal_current_handle();
4747
0390131b 4748 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4749 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4750 " cancelled because transaction is not started",
9c3013e9
JK		 4751 (unsigned long long)off, (unsigned long long)len);
4752 return -EIO;
4753 }
67eeb568
DM		 4754 /*
4755 * Since we account only one data block in transaction credits,
4756 * then it is impossible to cross a block boundary.
4757 */
4758 if (sb->s_blocksize - offset < len) {
4759 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4760 " cancelled because not block aligned",
4761 (unsigned long long)off, (unsigned long long)len);
4762 return -EIO;
4763 }
4764
ac27a0ec 4765 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4766 bh = ext4_bread(handle, inode, blk, 1, &err);
4767 if (!bh)
4768 goto out;
62d2b5f2
JK		 4769 err = ext4_journal_get_write_access(handle, bh);
4770 if (err) {
4771 brelse(bh);
4772 goto out;
ac27a0ec 4773 }
67eeb568
DM		 4774 lock_buffer(bh);
4775 memcpy(bh->b_data+offset, data, len);
4776 flush_dcache_page(bh->b_page);
4777 unlock_buffer(bh);
62d2b5f2 4778 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 4779 brelse(bh);
ac27a0ec 4780out:
67eeb568 4781 if (err) {
4d04e4fb 4782 mutex_unlock(&inode->i_mutex);
ac27a0ec 4783 return err;
4d04e4fb 4784 }
67eeb568
DM		 4785 if (inode->i_size < off + len) {
4786 i_size_write(inode, off + len);
617ba13b 4787 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 4788 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4789 }
ac27a0ec 4790 mutex_unlock(&inode->i_mutex);
67eeb568 4791 return len;
ac27a0ec
DK		 4792}
4793
4794#endif
4795
152a0836
AV		 4796static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4797 const char *dev_name, void *data)
ac27a0ec 4798{
152a0836 4799 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 4800}
4801
37f328eb 4802#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4803static inline void register_as_ext2(void)
4804{
4805 int err = register_filesystem(&ext2_fs_type);
4806 if (err)
4807 printk(KERN_WARNING
4808 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4809}
4810
4811static inline void unregister_as_ext2(void)
4812{
4813 unregister_filesystem(&ext2_fs_type);
4814}
2035e776
TT		 4815
4816static inline int ext2_feature_set_ok(struct super_block *sb)
4817{
4818 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4819 return 0;
4820 if (sb->s_flags & MS_RDONLY)
4821 return 1;
4822 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4823 return 0;
4824 return 1;
4825}
51b7e3c9 4826MODULE_ALIAS("ext2");
24b58424
TT		 4827#else
4828static inline void register_as_ext2(void) { }
4829static inline void unregister_as_ext2(void) { }
2035e776 4830static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4831#endif
4832
37f328eb 4833#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4834static inline void register_as_ext3(void)
4835{
4836 int err = register_filesystem(&ext3_fs_type);
4837 if (err)
4838 printk(KERN_WARNING
4839 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4840}
4841
4842static inline void unregister_as_ext3(void)
4843{
4844 unregister_filesystem(&ext3_fs_type);
4845}
2035e776
TT		 4846
4847static inline int ext3_feature_set_ok(struct super_block *sb)
4848{
4849 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4850 return 0;
4851 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4852 return 0;
4853 if (sb->s_flags & MS_RDONLY)
4854 return 1;
4855 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4856 return 0;
4857 return 1;
4858}
51b7e3c9 4859MODULE_ALIAS("ext3");
24b58424
TT		 4860#else
4861static inline void register_as_ext3(void) { }
4862static inline void unregister_as_ext3(void) { }
2035e776 4863static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4864#endif
4865
03010a33
TT		 4866static struct file_system_type ext4_fs_type = {
4867 .owner = THIS_MODULE,
4868 .name = "ext4",
152a0836 4869 .mount = ext4_mount,
03010a33
TT		 4870 .kill_sb = kill_block_super,
4871 .fs_flags = FS_REQUIRES_DEV,
4872};
4873
8f021222 4874static int __init ext4_init_feat_adverts(void)
857ac889
LC		 4875{
4876 struct ext4_features *ef;
4877 int ret = -ENOMEM;
4878
4879 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4880 if (!ef)
4881 goto out;
4882
4883 ef->f_kobj.kset = ext4_kset;
4884 init_completion(&ef->f_kobj_unregister);
4885 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4886 "features");
4887 if (ret) {
4888 kfree(ef);
4889 goto out;
4890 }
4891
4892 ext4_feat = ef;
4893 ret = 0;
4894out:
4895 return ret;
4896}
4897
8f021222
LC		 4898static void ext4_exit_feat_adverts(void)
4899{
4900 kobject_put(&ext4_feat->f_kobj);
4901 wait_for_completion(&ext4_feat->f_kobj_unregister);
4902 kfree(ext4_feat);
4903}
4904
e9e3bcec
ES		 4905/* Shared across all ext4 file systems */
4906wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4907struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4908
5dabfc78 4909static int __init ext4_init_fs(void)
ac27a0ec 4910{
e9e3bcec 4911 int i, err;
c9de560d 4912
07c0c5d8
AV		 4913 ext4_li_info = NULL;
4914 mutex_init(&ext4_li_mtx);
4915
12e9b892 4916 ext4_check_flag_values();
e9e3bcec
ES		 4917
4918 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4919 mutex_init(&ext4__aio_mutex[i]);
4920 init_waitqueue_head(&ext4__ioend_wq[i]);
4921 }
4922
5dabfc78 4923 err = ext4_init_pageio();
6fd058f7
TT		 4924 if (err)
4925 return err;
5dabfc78 4926 err = ext4_init_system_zone();
bd2d0210 4927 if (err)
d44651d0 4928 goto out6;
3197ebdb
TT		 4929 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4930 if (!ext4_kset)
dd68314c 4931 goto out5;
d44651d0 4932 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 4933
4934 err = ext4_init_feat_adverts();
dd68314c
TT		 4935 if (err)
4936 goto out4;
857ac889 4937
5dabfc78 4938 err = ext4_init_mballoc();
ac27a0ec 4939 if (err)
6fd058f7 4940 goto out3;
c9de560d 4941
5dabfc78 4942 err = ext4_init_xattr();
c9de560d
AT		 4943 if (err)
4944 goto out2;
ac27a0ec
DK		 4945 err = init_inodecache();
4946 if (err)
4947 goto out1;
24b58424 4948 register_as_ext3();
2035e776 4949 register_as_ext2();
03010a33 4950 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 4951 if (err)
4952 goto out;
bfff6873 4953
ac27a0ec
DK		 4954 return 0;
4955out:
24b58424
TT		 4956 unregister_as_ext2();
4957 unregister_as_ext3();
ac27a0ec
DK		 4958 destroy_inodecache();
4959out1:
5dabfc78 4960 ext4_exit_xattr();
c9de560d 4961out2:
5dabfc78 4962 ext4_exit_mballoc();
6fd058f7 4963out3:
8f021222 4964 ext4_exit_feat_adverts();
dd68314c 4965out4:
d44651d0
FJ		 4966 if (ext4_proc_root)
4967 remove_proc_entry("fs/ext4", NULL);
6fd058f7 4968 kset_unregister(ext4_kset);
d44651d0 4969out5:
5dabfc78 4970 ext4_exit_system_zone();
d44651d0 4971out6:
5dabfc78 4972 ext4_exit_pageio();
ac27a0ec
DK		 4973 return err;
4974}
4975
5dabfc78 4976static void __exit ext4_exit_fs(void)
ac27a0ec 4977{
bfff6873 4978 ext4_destroy_lazyinit_thread();
24b58424
TT		 4979 unregister_as_ext2();
4980 unregister_as_ext3();
03010a33 4981 unregister_filesystem(&ext4_fs_type);
ac27a0ec 4982 destroy_inodecache();
5dabfc78
TT		 4983 ext4_exit_xattr();
4984 ext4_exit_mballoc();
8f021222 4985 ext4_exit_feat_adverts();
9f6200bb 4986 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4987 kset_unregister(ext4_kset);
5dabfc78
TT		 4988 ext4_exit_system_zone();
4989 ext4_exit_pageio();
ac27a0ec
DK		 4990}
4991
4992MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4993MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4994MODULE_LICENSE("GPL");
5dabfc78
TT		 4995module_init(ext4_init_fs)
4996module_exit(ext4_exit_fs)
kernel source for telekom puls (alcatel/mediatek)