]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-bionic-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: send parallel discards on commit completions
[mirror_ubuntu-bionic-kernel.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>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
66114cad 27#include <linux/backing-dev.h>
ac27a0ec 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>
3197ebdb 37#include <linux/ctype.h>
1330593e 38#include <linux/log2.h>
717d50e4 39#include <linux/crc16.h>
ef510424 40#include <linux/dax.h>
7abc52c2 41#include <linux/cleancache.h>
7c0f6ba6 42#include <linux/uaccess.h>
ac27a0ec 43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
4a092d73 48#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 49#include "ext4_jbd2.h"
ac27a0ec
DK		 50#include "xattr.h"
51#include "acl.h"
3661d286 52#include "mballoc.h"
0c9ec4be 53#include "fsmap.h"
ac27a0ec 54
9bffad1e
TT		 55#define CREATE_TRACE_POINTS
56#include <trace/events/ext4.h>
57
0b75a840
LC		 58static struct ext4_lazy_init *ext4_li_info;
59static struct mutex ext4_li_mtx;
e294a537 60static struct ratelimit_state ext4_mount_msg_ratelimit;
9f6200bb 61
617ba13b 62static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 63 unsigned long journal_devnum);
2adf6da8 64static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 65static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 66static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
617ba13b 70static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 71static int ext4_remount(struct super_block *sb, int *flags, char *data);
72static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 73static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 74static int ext4_freeze(struct super_block *sb);
152a0836
AV		 75static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
76 const char *dev_name, void *data);
2035e776
TT		 77static inline int ext2_feature_set_ok(struct super_block *sb);
78static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 79static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 80static void ext4_destroy_lazyinit_thread(void);
81static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 82static void ext4_clear_request_list(void);
c6cb7e77
EW		 83static struct inode *ext4_get_journal_inode(struct super_block *sb,
84 unsigned int journal_inum);
ac27a0ec 85
e74031fd
JK		 86/*
87 * Lock ordering
88 *
89 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
90 * i_mmap_rwsem (inode->i_mmap_rwsem)!
91 *
92 * page fault path:
93 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
94 * page lock -> i_data_sem (rw)
95 *
96 * buffered write path:
97 * sb_start_write -> i_mutex -> mmap_sem
98 * sb_start_write -> i_mutex -> transaction start -> page lock ->
99 * i_data_sem (rw)
100 *
101 * truncate:
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * i_mmap_rwsem (w) -> page lock
104 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
105 * transaction start -> i_data_sem (rw)
106 *
107 * direct IO:
108 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
109 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
110 * transaction start -> i_data_sem (rw)
111 *
112 * writepages:
113 * transaction start -> page lock(s) -> i_data_sem (rw)
114 */
115
c290ea01 116#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
2035e776
TT		 117static struct file_system_type ext2_fs_type = {
118 .owner = THIS_MODULE,
119 .name = "ext2",
120 .mount = ext4_mount,
121 .kill_sb = kill_block_super,
122 .fs_flags = FS_REQUIRES_DEV,
123};
7f78e035 124MODULE_ALIAS_FS("ext2");
fa7614dd 125MODULE_ALIAS("ext2");
2035e776
TT		 126#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
127#else
128#define IS_EXT2_SB(sb) (0)
129#endif
130
131
ba69f9ab
JK		 132static struct file_system_type ext3_fs_type = {
133 .owner = THIS_MODULE,
134 .name = "ext3",
152a0836 135 .mount = ext4_mount,
ba69f9ab
JK		 136 .kill_sb = kill_block_super,
137 .fs_flags = FS_REQUIRES_DEV,
138};
7f78e035 139MODULE_ALIAS_FS("ext3");
fa7614dd 140MODULE_ALIAS("ext3");
ba69f9ab 141#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
bd81d8ee 142
d25425f8
DW		 143static int ext4_verify_csum_type(struct super_block *sb,
144 struct ext4_super_block *es)
145{
e2b911c5 146 if (!ext4_has_feature_metadata_csum(sb))
d25425f8
DW		 147 return 1;
148
149 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
150}
151
a9c47317
DW		 152static __le32 ext4_superblock_csum(struct super_block *sb,
153 struct ext4_super_block *es)
154{
155 struct ext4_sb_info *sbi = EXT4_SB(sb);
156 int offset = offsetof(struct ext4_super_block, s_checksum);
157 __u32 csum;
158
159 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
160
161 return cpu_to_le32(csum);
162}
163
c197855e
SH		 164static int ext4_superblock_csum_verify(struct super_block *sb,
165 struct ext4_super_block *es)
a9c47317 166{
9aa5d32b 167 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 168 return 1;
169
170 return es->s_checksum == ext4_superblock_csum(sb, es);
171}
172
06db49e6 173void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 174{
06db49e6
TT		 175 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
176
9aa5d32b 177 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 178 return;
179
180 es->s_checksum = ext4_superblock_csum(sb, es);
181}
182
9933fc0a
TT		 183void *ext4_kvmalloc(size_t size, gfp_t flags)
184{
185 void *ret;
186
8be04b93 187 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 188 if (!ret)
189 ret = __vmalloc(size, flags, PAGE_KERNEL);
190 return ret;
191}
192
193void *ext4_kvzalloc(size_t size, gfp_t flags)
194{
195 void *ret;
196
8be04b93 197 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 198 if (!ret)
199 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
200 return ret;
201}
202
8fadc143
AR		 203ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
204 struct ext4_group_desc *bg)
bd81d8ee 205{
3a14589c 206 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 208 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 209}
210
8fadc143
AR		 211ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
212 struct ext4_group_desc *bg)
bd81d8ee 213{
5272f837 214 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 216 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 217}
218
8fadc143
AR		 219ext4_fsblk_t ext4_inode_table(struct super_block *sb,
220 struct ext4_group_desc *bg)
bd81d8ee 221{
5272f837 222 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 224 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 225}
226
021b65bb
TT		 227__u32 ext4_free_group_clusters(struct super_block *sb,
228 struct ext4_group_desc *bg)
560671a0
AK		 229{
230 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 232 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 233}
234
235__u32 ext4_free_inodes_count(struct super_block *sb,
236 struct ext4_group_desc *bg)
237{
238 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
239 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 240 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 241}
242
243__u32 ext4_used_dirs_count(struct super_block *sb,
244 struct ext4_group_desc *bg)
245{
246 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
247 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 248 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 249}
250
251__u32 ext4_itable_unused_count(struct super_block *sb,
252 struct ext4_group_desc *bg)
253{
254 return le16_to_cpu(bg->bg_itable_unused_lo) |
255 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 256 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 257}
258
8fadc143
AR		 259void ext4_block_bitmap_set(struct super_block *sb,
260 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 261{
3a14589c 262 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 265}
266
8fadc143
AR		 267void ext4_inode_bitmap_set(struct super_block *sb,
268 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 269{
5272f837 270 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 273}
274
8fadc143
AR		 275void ext4_inode_table_set(struct super_block *sb,
276 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 277{
5272f837 278 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 281}
282
021b65bb
TT		 283void ext4_free_group_clusters_set(struct super_block *sb,
284 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 285{
286 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
289}
290
291void ext4_free_inodes_set(struct super_block *sb,
292 struct ext4_group_desc *bg, __u32 count)
293{
294 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
295 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
296 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
297}
298
299void ext4_used_dirs_set(struct super_block *sb,
300 struct ext4_group_desc *bg, __u32 count)
301{
302 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
303 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
304 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
305}
306
307void ext4_itable_unused_set(struct super_block *sb,
308 struct ext4_group_desc *bg, __u32 count)
309{
310 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
311 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
312 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
313}
314
d3d1faf6 315
1c13d5c0
TT		 316static void __save_error_info(struct super_block *sb, const char *func,
317 unsigned int line)
318{
319 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
320
321 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1b46617b
TT		 322 if (bdev_read_only(sb->s_bdev))
323 return;
1c13d5c0
TT		 324 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
325 es->s_last_error_time = cpu_to_le32(get_seconds());
326 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
327 es->s_last_error_line = cpu_to_le32(line);
328 if (!es->s_first_error_time) {
329 es->s_first_error_time = es->s_last_error_time;
330 strncpy(es->s_first_error_func, func,
331 sizeof(es->s_first_error_func));
332 es->s_first_error_line = cpu_to_le32(line);
333 es->s_first_error_ino = es->s_last_error_ino;
334 es->s_first_error_block = es->s_last_error_block;
335 }
66e61a9e
TT		 336 /*
337 * Start the daily error reporting function if it hasn't been
338 * started already
339 */
340 if (!es->s_error_count)
341 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 342 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 343}
344
345static void save_error_info(struct super_block *sb, const char *func,
346 unsigned int line)
347{
348 __save_error_info(sb, func, line);
349 ext4_commit_super(sb, 1);
350}
351
bdfe0cbd
TT		 352/*
353 * The del_gendisk() function uninitializes the disk-specific data
354 * structures, including the bdi structure, without telling anyone
355 * else. Once this happens, any attempt to call mark_buffer_dirty()
356 * (for example, by ext4_commit_super), will cause a kernel OOPS.
357 * This is a kludge to prevent these oops until we can put in a proper
358 * hook in del_gendisk() to inform the VFS and file system layers.
359 */
360static int block_device_ejected(struct super_block *sb)
361{
362 struct inode *bd_inode = sb->s_bdev->bd_inode;
363 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
364
365 return bdi->dev == NULL;
366}
367
18aadd47
BJ		 368static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
369{
370 struct super_block *sb = journal->j_private;
371 struct ext4_sb_info *sbi = EXT4_SB(sb);
372 int error = is_journal_aborted(journal);
5d3ee208 373 struct ext4_journal_cb_entry *jce;
18aadd47 374
5d3ee208 375 BUG_ON(txn->t_state == T_FINISHED);
a0154344
DJ		 376
377 ext4_process_freed_data(sb, txn->t_tid);
378
18aadd47 379 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 380 while (!list_empty(&txn->t_private_list)) {
381 jce = list_entry(txn->t_private_list.next,
382 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 383 list_del_init(&jce->jce_list);
384 spin_unlock(&sbi->s_md_lock);
385 jce->jce_func(sb, jce, error);
386 spin_lock(&sbi->s_md_lock);
387 }
388 spin_unlock(&sbi->s_md_lock);
389}
1c13d5c0 390
ac27a0ec
DK		 391/* Deal with the reporting of failure conditions on a filesystem such as
392 * inconsistencies detected or read IO failures.
393 *
394 * On ext2, we can store the error state of the filesystem in the
617ba13b 395 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 396 * write ordering constraints on the superblock which prevent us from
397 * writing it out straight away; and given that the journal is about to
398 * be aborted, we can't rely on the current, or future, transactions to
399 * write out the superblock safely.
400 *
dab291af 401 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 402 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 403 * that error until we've noted it down and cleared it.
404 */
405
617ba13b 406static void ext4_handle_error(struct super_block *sb)
ac27a0ec 407{
ac27a0ec
DK		 408 if (sb->s_flags & MS_RDONLY)
409 return;
410
2b2d6d01 411 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 412 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 413
4ab2f15b 414 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 415 if (journal)
dab291af 416 jbd2_journal_abort(journal, -EIO);
ac27a0ec 417 }
2b2d6d01 418 if (test_opt(sb, ERRORS_RO)) {
b31e1552 419 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 420 /*
421 * Make sure updated value of ->s_mount_flags will be visible
422 * before ->s_flags update
423 */
424 smp_wmb();
ac27a0ec
DK		 425 sb->s_flags |= MS_RDONLY;
426 }
4327ba52
DJ		 427 if (test_opt(sb, ERRORS_PANIC)) {
428 if (EXT4_SB(sb)->s_journal &&
429 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
430 return;
617ba13b 431 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec 432 sb->s_id);
4327ba52 433 }
ac27a0ec
DK		 434}
435
efbed4dc
TT		 436#define ext4_error_ratelimit(sb) \
437 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
438 "EXT4-fs error")
439
12062ddd 440void __ext4_error(struct super_block *sb, const char *function,
c398eda0 441 unsigned int line, const char *fmt, ...)
ac27a0ec 442{
0ff2ea7d 443 struct va_format vaf;
ac27a0ec
DK		 444 va_list args;
445
0db1ff22
TT		 446 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
447 return;
448
efbed4dc
TT		 449 if (ext4_error_ratelimit(sb)) {
450 va_start(args, fmt);
451 vaf.fmt = fmt;
452 vaf.va = &args;
453 printk(KERN_CRIT
454 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
455 sb->s_id, function, line, current->comm, &vaf);
456 va_end(args);
457 }
f3fc0210 458 save_error_info(sb, function, line);
617ba13b 459 ext4_handle_error(sb);
ac27a0ec
DK		 460}
461
e7c96e8e
JP		 462void __ext4_error_inode(struct inode *inode, const char *function,
463 unsigned int line, ext4_fsblk_t block,
464 const char *fmt, ...)
273df556
FM		 465{
466 va_list args;
f7c21177 467 struct va_format vaf;
1c13d5c0 468 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 469
0db1ff22
TT		 470 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
471 return;
472
1c13d5c0
TT		 473 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
474 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 475 if (ext4_error_ratelimit(inode->i_sb)) {
476 va_start(args, fmt);
477 vaf.fmt = fmt;
478 vaf.va = &args;
479 if (block)
480 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
481 "inode #%lu: block %llu: comm %s: %pV\n",
482 inode->i_sb->s_id, function, line, inode->i_ino,
483 block, current->comm, &vaf);
484 else
485 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
486 "inode #%lu: comm %s: %pV\n",
487 inode->i_sb->s_id, function, line, inode->i_ino,
488 current->comm, &vaf);
489 va_end(args);
490 }
1c13d5c0 491 save_error_info(inode->i_sb, function, line);
273df556
FM		 492 ext4_handle_error(inode->i_sb);
493}
494
e7c96e8e
JP		 495void __ext4_error_file(struct file *file, const char *function,
496 unsigned int line, ext4_fsblk_t block,
497 const char *fmt, ...)
273df556
FM		 498{
499 va_list args;
f7c21177 500 struct va_format vaf;
1c13d5c0 501 struct ext4_super_block *es;
496ad9aa 502 struct inode *inode = file_inode(file);
273df556
FM		 503 char pathname[80], *path;
504
0db1ff22
TT		 505 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
506 return;
507
1c13d5c0
TT		 508 es = EXT4_SB(inode->i_sb)->s_es;
509 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 510 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 511 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 512 if (IS_ERR(path))
513 path = "(unknown)";
514 va_start(args, fmt);
515 vaf.fmt = fmt;
516 vaf.va = &args;
517 if (block)
518 printk(KERN_CRIT
519 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
520 "block %llu: comm %s: path %s: %pV\n",
521 inode->i_sb->s_id, function, line, inode->i_ino,
522 block, current->comm, path, &vaf);
523 else
524 printk(KERN_CRIT
525 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
526 "comm %s: path %s: %pV\n",
527 inode->i_sb->s_id, function, line, inode->i_ino,
528 current->comm, path, &vaf);
529 va_end(args);
530 }
1c13d5c0 531 save_error_info(inode->i_sb, function, line);
273df556
FM		 532 ext4_handle_error(inode->i_sb);
533}
534
722887dd
TT		 535const char *ext4_decode_error(struct super_block *sb, int errno,
536 char nbuf[16])
ac27a0ec
DK		 537{
538 char *errstr = NULL;
539
540 switch (errno) {
6a797d27
DW		 541 case -EFSCORRUPTED:
542 errstr = "Corrupt filesystem";
543 break;
544 case -EFSBADCRC:
545 errstr = "Filesystem failed CRC";
546 break;
ac27a0ec
DK		 547 case -EIO:
548 errstr = "IO failure";
549 break;
550 case -ENOMEM:
551 errstr = "Out of memory";
552 break;
553 case -EROFS:
78f1ddbb
TT		 554 if (!sb || (EXT4_SB(sb)->s_journal &&
555 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 556 errstr = "Journal has aborted";
557 else
558 errstr = "Readonly filesystem";
559 break;
560 default:
561 /* If the caller passed in an extra buffer for unknown
562 * errors, textualise them now. Else we just return
563 * NULL. */
564 if (nbuf) {
565 /* Check for truncated error codes... */
566 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
567 errstr = nbuf;
568 }
569 break;
570 }
571
572 return errstr;
573}
574
617ba13b 575/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 576 * automatically and invokes the appropriate error response. */
577
c398eda0
TT		 578void __ext4_std_error(struct super_block *sb, const char *function,
579 unsigned int line, int errno)
ac27a0ec
DK		 580{
581 char nbuf[16];
582 const char *errstr;
583
0db1ff22
TT		 584 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
585 return;
586
ac27a0ec
DK		 587 /* Special case: if the error is EROFS, and we're not already
588 * inside a transaction, then there's really no point in logging
589 * an error. */
590 if (errno == -EROFS && journal_current_handle() == NULL &&
591 (sb->s_flags & MS_RDONLY))
592 return;
593
efbed4dc
TT		 594 if (ext4_error_ratelimit(sb)) {
595 errstr = ext4_decode_error(sb, errno, nbuf);
596 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
597 sb->s_id, function, line, errstr);
598 }
ac27a0ec 599
efbed4dc 600 save_error_info(sb, function, line);
617ba13b 601 ext4_handle_error(sb);
ac27a0ec
DK		 602}
603
604/*
617ba13b 605 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 606 * abort function may be used to deal with unrecoverable failures such
607 * as journal IO errors or ENOMEM at a critical moment in log management.
608 *
609 * We unconditionally force the filesystem into an ABORT|READONLY state,
610 * unless the error response on the fs has been set to panic in which
611 * case we take the easy way out and panic immediately.
612 */
613
c67d859e 614void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 615 unsigned int line, const char *fmt, ...)
ac27a0ec 616{
651e1c3b 617 struct va_format vaf;
ac27a0ec
DK		 618 va_list args;
619
0db1ff22
TT		 620 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
621 return;
622
1c13d5c0 623 save_error_info(sb, function, line);
ac27a0ec 624 va_start(args, fmt);
651e1c3b
JP		 625 vaf.fmt = fmt;
626 vaf.va = &args;
627 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
628 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 629 va_end(args);
630
1c13d5c0
TT		 631 if ((sb->s_flags & MS_RDONLY) == 0) {
632 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 633 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 634 /*
635 * Make sure updated value of ->s_mount_flags will be visible
636 * before ->s_flags update
637 */
638 smp_wmb();
639 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 640 if (EXT4_SB(sb)->s_journal)
641 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
642 save_error_info(sb, function, line);
643 }
4327ba52
DJ		 644 if (test_opt(sb, ERRORS_PANIC)) {
645 if (EXT4_SB(sb)->s_journal &&
646 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
647 return;
617ba13b 648 panic("EXT4-fs panic from previous error\n");
4327ba52 649 }
ac27a0ec
DK		 650}
651
e7c96e8e
JP		 652void __ext4_msg(struct super_block *sb,
653 const char *prefix, const char *fmt, ...)
b31e1552 654{
0ff2ea7d 655 struct va_format vaf;
b31e1552
ES		 656 va_list args;
657
efbed4dc
TT		 658 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
659 return;
660
b31e1552 661 va_start(args, fmt);
0ff2ea7d
JP		 662 vaf.fmt = fmt;
663 vaf.va = &args;
664 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 665 va_end(args);
666}
667
b03a2f7e
AD		 668#define ext4_warning_ratelimit(sb) \
669 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
670 "EXT4-fs warning")
671
12062ddd 672void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 673 unsigned int line, const char *fmt, ...)
ac27a0ec 674{
0ff2ea7d 675 struct va_format vaf;
ac27a0ec
DK		 676 va_list args;
677
b03a2f7e 678 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 679 return;
680
ac27a0ec 681 va_start(args, fmt);
0ff2ea7d
JP		 682 vaf.fmt = fmt;
683 vaf.va = &args;
684 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
685 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 686 va_end(args);
687}
688
b03a2f7e
AD		 689void __ext4_warning_inode(const struct inode *inode, const char *function,
690 unsigned int line, const char *fmt, ...)
691{
692 struct va_format vaf;
693 va_list args;
694
695 if (!ext4_warning_ratelimit(inode->i_sb))
696 return;
697
698 va_start(args, fmt);
699 vaf.fmt = fmt;
700 vaf.va = &args;
701 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
702 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
703 function, line, inode->i_ino, current->comm, &vaf);
704 va_end(args);
705}
706
e29136f8
TT		 707void __ext4_grp_locked_error(const char *function, unsigned int line,
708 struct super_block *sb, ext4_group_t grp,
709 unsigned long ino, ext4_fsblk_t block,
710 const char *fmt, ...)
5d1b1b3f
AK		 711__releases(bitlock)
712__acquires(bitlock)
713{
0ff2ea7d 714 struct va_format vaf;
5d1b1b3f
AK		 715 va_list args;
716 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
717
0db1ff22
TT		 718 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
719 return;
720
1c13d5c0
TT		 721 es->s_last_error_ino = cpu_to_le32(ino);
722 es->s_last_error_block = cpu_to_le64(block);
723 __save_error_info(sb, function, line);
0ff2ea7d 724
efbed4dc
TT		 725 if (ext4_error_ratelimit(sb)) {
726 va_start(args, fmt);
727 vaf.fmt = fmt;
728 vaf.va = &args;
729 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
730 sb->s_id, function, line, grp);
731 if (ino)
732 printk(KERN_CONT "inode %lu: ", ino);
733 if (block)
734 printk(KERN_CONT "block %llu:",
735 (unsigned long long) block);
736 printk(KERN_CONT "%pV\n", &vaf);
737 va_end(args);
738 }
5d1b1b3f
AK		 739
740 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 741 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 742 return;
743 }
1c13d5c0 744
5d1b1b3f
AK		 745 ext4_unlock_group(sb, grp);
746 ext4_handle_error(sb);
747 /*
748 * We only get here in the ERRORS_RO case; relocking the group
749 * may be dangerous, but nothing bad will happen since the
750 * filesystem will have already been marked read/only and the
751 * journal has been aborted. We return 1 as a hint to callers
752 * who might what to use the return value from
25985edc 753 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 754 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
755 * aggressively from the ext4 function in question, with a
756 * more appropriate error code.
757 */
758 ext4_lock_group(sb, grp);
759 return;
760}
761
617ba13b 762void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 763{
617ba13b 764 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 765
617ba13b 766 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 767 return;
768
12062ddd 769 ext4_warning(sb,
ac27a0ec
DK		 770 "updating to rev %d because of new feature flag, "
771 "running e2fsck is recommended",
617ba13b 772 EXT4_DYNAMIC_REV);
ac27a0ec 773
617ba13b
MC		 774 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
775 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
776 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 777 /* leave es->s_feature_*compat flags alone */
778 /* es->s_uuid will be set by e2fsck if empty */
779
780 /*
781 * The rest of the superblock fields should be zero, and if not it
782 * means they are likely already in use, so leave them alone. We
783 * can leave it up to e2fsck to clean up any inconsistencies there.
784 */
785}
786
787/*
788 * Open the external journal device
789 */
b31e1552 790static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 791{
792 struct block_device *bdev;
793 char b[BDEVNAME_SIZE];
794
d4d77629 795 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 796 if (IS_ERR(bdev))
797 goto fail;
798 return bdev;
799
800fail:
b31e1552 801 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 802 __bdevname(dev, b), PTR_ERR(bdev));
803 return NULL;
804}
805
806/*
807 * Release the journal device
808 */
4385bab1 809static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 810{
4385bab1 811 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 812}
813
4385bab1 814static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 815{
816 struct block_device *bdev;
ac27a0ec
DK		 817 bdev = sbi->journal_bdev;
818 if (bdev) {
4385bab1 819 ext4_blkdev_put(bdev);
ac27a0ec
DK		 820 sbi->journal_bdev = NULL;
821 }
ac27a0ec
DK		 822}
823
824static inline struct inode *orphan_list_entry(struct list_head *l)
825{
617ba13b 826 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 827}
828
617ba13b 829static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 830{
831 struct list_head *l;
832
b31e1552
ES		 833 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
834 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 835
836 printk(KERN_ERR "sb_info orphan list:\n");
837 list_for_each(l, &sbi->s_orphan) {
838 struct inode *inode = orphan_list_entry(l);
839 printk(KERN_ERR " "
840 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
841 inode->i_sb->s_id, inode->i_ino, inode,
842 inode->i_mode, inode->i_nlink,
843 NEXT_ORPHAN(inode));
844 }
845}
846
957153fc
JK		 847#ifdef CONFIG_QUOTA
848static int ext4_quota_off(struct super_block *sb, int type);
849
850static inline void ext4_quota_off_umount(struct super_block *sb)
851{
852 int type;
853
964edf66
JK		 854 /* Use our quota_off function to clear inode flags etc. */
855 for (type = 0; type < EXT4_MAXQUOTAS; type++)
856 ext4_quota_off(sb, type);
957153fc
JK		 857}
858#else
859static inline void ext4_quota_off_umount(struct super_block *sb)
860{
861}
862#endif
863
2b2d6d01 864static void ext4_put_super(struct super_block *sb)
ac27a0ec 865{
617ba13b
MC		 866 struct ext4_sb_info *sbi = EXT4_SB(sb);
867 struct ext4_super_block *es = sbi->s_es;
97abd7d4 868 int aborted = 0;
ef2cabf7 869 int i, err;
ac27a0ec 870
857ac889 871 ext4_unregister_li_request(sb);
957153fc 872 ext4_quota_off_umount(sb);
e0ccfd95 873
2e8fa54e 874 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 875 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 876
0390131b 877 if (sbi->s_journal) {
97abd7d4 878 aborted = is_journal_aborted(sbi->s_journal);
0390131b
FM		 879 err = jbd2_journal_destroy(sbi->s_journal);
880 sbi->s_journal = NULL;
97abd7d4 881 if ((err < 0) && !aborted)
c67d859e 882 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 883 }
d4edac31 884
ebd173be 885 ext4_unregister_sysfs(sb);
d3922a77 886 ext4_es_unregister_shrinker(sbi);
9105bb14 887 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 888 ext4_release_system_zone(sb);
889 ext4_mb_release(sb);
890 ext4_ext_release(sb);
d4edac31 891
97abd7d4 892 if (!(sb->s_flags & MS_RDONLY) && !aborted) {
e2b911c5 893 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 894 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 895 }
58c5873a 896 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 897 ext4_commit_super(sb, 1);
898
ac27a0ec
DK		 899 for (i = 0; i < sbi->s_gdb_count; i++)
900 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 901 kvfree(sbi->s_group_desc);
902 kvfree(sbi->s_flex_groups);
57042651 903 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 904 percpu_counter_destroy(&sbi->s_freeinodes_counter);
905 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 906 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c8585c6f 907 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
ac27a0ec 908#ifdef CONFIG_QUOTA
a2d4a646 909 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 910 kfree(sbi->s_qf_names[i]);
911#endif
912
913 /* Debugging code just in case the in-memory inode orphan list
914 * isn't empty. The on-disk one can be non-empty if we've
915 * detected an error and taken the fs readonly, but the
916 * in-memory list had better be clean by this point. */
917 if (!list_empty(&sbi->s_orphan))
918 dump_orphan_list(sb, sbi);
919 J_ASSERT(list_empty(&sbi->s_orphan));
920
89d96a6f 921 sync_blockdev(sb->s_bdev);
f98393a6 922 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 923 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
924 /*
925 * Invalidate the journal device's buffers. We don't want them
926 * floating about in memory - the physical journal device may
927 * hotswapped, and it breaks the `ro-after' testing code.
928 */
929 sync_blockdev(sbi->journal_bdev);
f98393a6 930 invalidate_bdev(sbi->journal_bdev);
617ba13b 931 ext4_blkdev_remove(sbi);
ac27a0ec 932 }
dec214d0
TE		 933 if (sbi->s_ea_inode_cache) {
934 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
935 sbi->s_ea_inode_cache = NULL;
936 }
47387409
TE		 937 if (sbi->s_ea_block_cache) {
938 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
939 sbi->s_ea_block_cache = NULL;
9c191f70 940 }
c5e06d10
JL		 941 if (sbi->s_mmp_tsk)
942 kthread_stop(sbi->s_mmp_tsk);
9060dd2c 943 brelse(sbi->s_sbh);
ac27a0ec 944 sb->s_fs_info = NULL;
3197ebdb
TT		 945 /*
946 * Now that we are completely done shutting down the
947 * superblock, we need to actually destroy the kobject.
948 */
3197ebdb
TT		 949 kobject_put(&sbi->s_kobj);
950 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 951 if (sbi->s_chksum_driver)
952 crypto_free_shash(sbi->s_chksum_driver);
705895b6 953 kfree(sbi->s_blockgroup_lock);
ac27a0ec 954 kfree(sbi);
ac27a0ec
DK		 955}
956
e18b890b 957static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 958
959/*
960 * Called inside transaction, so use GFP_NOFS
961 */
617ba13b 962static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 963{
617ba13b 964 struct ext4_inode_info *ei;
ac27a0ec 965
e6b4f8da 966 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 967 if (!ei)
968 return NULL;
0b8e58a1 969
ac27a0ec 970 ei->vfs_inode.i_version = 1;
202ee5df 971 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 972 INIT_LIST_HEAD(&ei->i_prealloc_list);
973 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 974 ext4_es_init_tree(&ei->i_es_tree);
975 rwlock_init(&ei->i_es_lock);
edaa53ca 976 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 977 ei->i_es_all_nr = 0;
edaa53ca 978 ei->i_es_shk_nr = 0;
dd475925 979 ei->i_es_shrink_lblk = 0;
d2a17637
MC		 980 ei->i_reserved_data_blocks = 0;
981 ei->i_reserved_meta_blocks = 0;
982 ei->i_allocated_meta_blocks = 0;
9d0be502 983 ei->i_da_metadata_calc_len = 0;
7e731bc9 984 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 985 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 986#ifdef CONFIG_QUOTA
987 ei->i_reserved_quota = 0;
96c7e0d9 988 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 989#endif
8aefcd55 990 ei->jinode = NULL;
2e8fa54e 991 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 992 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 993 ei->i_sync_tid = 0;
994 ei->i_datasync_tid = 0;
e27f41e1 995 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 996 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
ac27a0ec
DK		 997 return &ei->vfs_inode;
998}
999
7ff9c073
TT		 1000static int ext4_drop_inode(struct inode *inode)
1001{
1002 int drop = generic_drop_inode(inode);
1003
1004 trace_ext4_drop_inode(inode, drop);
1005 return drop;
1006}
1007
fa0d7e3d
NP		 1008static void ext4_i_callback(struct rcu_head *head)
1009{
1010 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 1011 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1012}
1013
617ba13b 1014static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1015{
9f7dd93d 1016 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1017 ext4_msg(inode->i_sb, KERN_ERR,
1018 "Inode %lu (%p): orphan list check failed!",
1019 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1020 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1021 EXT4_I(inode), sizeof(struct ext4_inode_info),
1022 true);
1023 dump_stack();
1024 }
fa0d7e3d 1025 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 1026}
1027
51cc5068 1028static void init_once(void *foo)
ac27a0ec 1029{
617ba13b 1030 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1031
a35afb83 1032 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1033 init_rwsem(&ei->xattr_sem);
0e855ac8 1034 init_rwsem(&ei->i_data_sem);
ea3d7209 1035 init_rwsem(&ei->i_mmap_sem);
a35afb83 1036 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1037}
1038
e67bc2b3 1039static int __init init_inodecache(void)
ac27a0ec 1040{
617ba13b
MC		 1041 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1042 sizeof(struct ext4_inode_info),
ac27a0ec 1043 0, (SLAB_RECLAIM_ACCOUNT|
5d097056 1044 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
20c2df83 1045 init_once);
617ba13b 1046 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1047 return -ENOMEM;
1048 return 0;
1049}
1050
1051static void destroy_inodecache(void)
1052{
8c0a8537
KS		 1053 /*
1054 * Make sure all delayed rcu free inodes are flushed before we
1055 * destroy cache.
1056 */
1057 rcu_barrier();
617ba13b 1058 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1059}
1060
0930fcc1 1061void ext4_clear_inode(struct inode *inode)
ac27a0ec 1062{
0930fcc1 1063 invalidate_inode_buffers(inode);
dbd5768f 1064 clear_inode(inode);
9f754758 1065 dquot_drop(inode);
c2ea3fde 1066 ext4_discard_preallocations(inode);
51865fda 1067 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1068 if (EXT4_I(inode)->jinode) {
1069 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1070 EXT4_I(inode)->jinode);
1071 jbd2_free_inode(EXT4_I(inode)->jinode);
1072 EXT4_I(inode)->jinode = NULL;
1073 }
b7236e21 1074#ifdef CONFIG_EXT4_FS_ENCRYPTION
a7550b30 1075 fscrypt_put_encryption_info(inode, NULL);
b7236e21 1076#endif
ac27a0ec
DK		 1077}
1078
1b961ac0 1079static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1080 u64 ino, u32 generation)
ac27a0ec 1081{
ac27a0ec 1082 struct inode *inode;
ac27a0ec 1083
617ba13b 1084 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1085 return ERR_PTR(-ESTALE);
617ba13b 1086 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1087 return ERR_PTR(-ESTALE);
1088
1089 /* iget isn't really right if the inode is currently unallocated!!
1090 *
617ba13b 1091 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1092 * deleted, so we should be safe.
1093 *
1094 * Currently we don't know the generation for parent directory, so
1095 * a generation of 0 means "accept any"
1096 */
f4bb2981 1097 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1098 if (IS_ERR(inode))
1099 return ERR_CAST(inode);
1100 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1101 iput(inode);
1102 return ERR_PTR(-ESTALE);
1103 }
1b961ac0
CH		 1104
1105 return inode;
1106}
1107
1108static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1109 int fh_len, int fh_type)
1b961ac0
CH		 1110{
1111 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1112 ext4_nfs_get_inode);
1113}
1114
1115static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1116 int fh_len, int fh_type)
1b961ac0
CH		 1117{
1118 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1119 ext4_nfs_get_inode);
ac27a0ec
DK		 1120}
1121
c39a7f84
TO		 1122/*
1123 * Try to release metadata pages (indirect blocks, directories) which are
1124 * mapped via the block device. Since these pages could have journal heads
1125 * which would prevent try_to_free_buffers() from freeing them, we must use
1126 * jbd2 layer's try_to_free_buffers() function to release them.
1127 */
0b8e58a1
AD		 1128static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1129 gfp_t wait)
c39a7f84
TO		 1130{
1131 journal_t *journal = EXT4_SB(sb)->s_journal;
1132
1133 WARN_ON(PageChecked(page));
1134 if (!page_has_buffers(page))
1135 return 0;
1136 if (journal)
1137 return jbd2_journal_try_to_free_buffers(journal, page,
d0164adc 1138 wait & ~__GFP_DIRECT_RECLAIM);
c39a7f84
TO		 1139 return try_to_free_buffers(page);
1140}
1141
a7550b30
JK		 1142#ifdef CONFIG_EXT4_FS_ENCRYPTION
1143static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1144{
1145 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1146 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1147}
1148
a7550b30
JK		 1149static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1150 void *fs_data)
1151{
2f8f5e76 1152 handle_t *handle = fs_data;
c1a5d5f6 1153 int res, res2, credits, retries = 0;
2f8f5e76 1154
94840e3c
EB		 1155 res = ext4_convert_inline_data(inode);
1156 if (res)
1157 return res;
1158
2f8f5e76
EB		 1159 /*
1160 * If a journal handle was specified, then the encryption context is
1161 * being set on a new inode via inheritance and is part of a larger
1162 * transaction to create the inode. Otherwise the encryption context is
1163 * being set on an existing inode in its own transaction. Only in the
1164 * latter case should the "retry on ENOSPC" logic be used.
1165 */
a7550b30 1166
2f8f5e76
EB		 1167 if (handle) {
1168 res = ext4_xattr_set_handle(handle, inode,
1169 EXT4_XATTR_INDEX_ENCRYPTION,
1170 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1171 ctx, len, 0);
a7550b30
JK		 1172 if (!res) {
1173 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1174 ext4_clear_inode_state(inode,
1175 EXT4_STATE_MAY_INLINE_DATA);
a3caa24b
JK		 1176 /*
1177 * Update inode->i_flags - e.g. S_DAX may get disabled
1178 */
1179 ext4_set_inode_flags(inode);
a7550b30
JK		 1180 }
1181 return res;
1182 }
1183
b8cb5a54
TE		 1184 res = dquot_initialize(inode);
1185 if (res)
1186 return res;
2f8f5e76 1187retry:
dec214d0
TE		 1188 res = ext4_xattr_set_credits(inode, len, &credits);
1189 if (res)
1190 return res;
1191
c1a5d5f6 1192 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
a7550b30
JK		 1193 if (IS_ERR(handle))
1194 return PTR_ERR(handle);
1195
2f8f5e76
EB		 1196 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1197 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1198 ctx, len, 0);
a7550b30
JK		 1199 if (!res) {
1200 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
a3caa24b
JK		 1201 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1202 ext4_set_inode_flags(inode);
a7550b30
JK		 1203 res = ext4_mark_inode_dirty(handle, inode);
1204 if (res)
1205 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1206 }
1207 res2 = ext4_journal_stop(handle);
2f8f5e76
EB		 1208
1209 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1210 goto retry;
a7550b30
JK		 1211 if (!res)
1212 res = res2;
1213 return res;
1214}
1215
1216static int ext4_dummy_context(struct inode *inode)
1217{
1218 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1219}
1220
1221static unsigned ext4_max_namelen(struct inode *inode)
1222{
1223 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1224 EXT4_NAME_LEN;
1225}
1226
6f69f0ed 1227static const struct fscrypt_operations ext4_cryptops = {
a5d431ef 1228 .key_prefix = "ext4:",
a7550b30 1229 .get_context = ext4_get_context,
a7550b30
JK		 1230 .set_context = ext4_set_context,
1231 .dummy_context = ext4_dummy_context,
1232 .is_encrypted = ext4_encrypted_inode,
1233 .empty_dir = ext4_empty_dir,
1234 .max_namelen = ext4_max_namelen,
1235};
1236#else
6f69f0ed 1237static const struct fscrypt_operations ext4_cryptops = {
a7550b30
JK		 1238 .is_encrypted = ext4_encrypted_inode,
1239};
1240#endif
1241
ac27a0ec 1242#ifdef CONFIG_QUOTA
d6006186 1243static const char * const quotatypes[] = INITQFNAMES;
689c958c 1244#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1245
617ba13b
MC		 1246static int ext4_write_dquot(struct dquot *dquot);
1247static int ext4_acquire_dquot(struct dquot *dquot);
1248static int ext4_release_dquot(struct dquot *dquot);
1249static int ext4_mark_dquot_dirty(struct dquot *dquot);
1250static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1251static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1252 const struct path *path);
617ba13b
MC		 1253static int ext4_quota_on_mount(struct super_block *sb, int type);
1254static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1255 size_t len, loff_t off);
617ba13b 1256static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1257 const char *data, size_t len, loff_t off);
7c319d32
AK		 1258static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1259 unsigned int flags);
1260static int ext4_enable_quotas(struct super_block *sb);
8f0e8746 1261static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
ac27a0ec 1262
96c7e0d9
JK		 1263static struct dquot **ext4_get_dquots(struct inode *inode)
1264{
1265 return EXT4_I(inode)->i_dquot;
1266}
1267
61e225dc 1268static const struct dquot_operations ext4_quota_operations = {
7a9ca53a
TE		 1269 .get_reserved_space = ext4_get_reserved_space,
1270 .write_dquot = ext4_write_dquot,
1271 .acquire_dquot = ext4_acquire_dquot,
1272 .release_dquot = ext4_release_dquot,
1273 .mark_dirty = ext4_mark_dquot_dirty,
1274 .write_info = ext4_write_info,
1275 .alloc_dquot = dquot_alloc,
1276 .destroy_dquot = dquot_destroy,
1277 .get_projid = ext4_get_projid,
1278 .get_inode_usage = ext4_get_inode_usage,
1279 .get_next_id = ext4_get_next_id,
ac27a0ec
DK		 1280};
1281
0d54b217 1282static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1283 .quota_on = ext4_quota_on,
ca0e05e4 1284 .quota_off = ext4_quota_off,
287a8095 1285 .quota_sync = dquot_quota_sync,
0a240339 1286 .get_state = dquot_get_state,
287a8095
CH		 1287 .set_info = dquot_set_dqinfo,
1288 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1289 .set_dqblk = dquot_set_dqblk,
1290 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1291};
1292#endif
1293
ee9b6d61 1294static const struct super_operations ext4_sops = {
617ba13b
MC		 1295 .alloc_inode = ext4_alloc_inode,
1296 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1297 .write_inode = ext4_write_inode,
1298 .dirty_inode = ext4_dirty_inode,
7ff9c073 1299 .drop_inode = ext4_drop_inode,
0930fcc1 1300 .evict_inode = ext4_evict_inode,
617ba13b 1301 .put_super = ext4_put_super,
617ba13b 1302 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1303 .freeze_fs = ext4_freeze,
1304 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1305 .statfs = ext4_statfs,
1306 .remount_fs = ext4_remount,
617ba13b 1307 .show_options = ext4_show_options,
ac27a0ec 1308#ifdef CONFIG_QUOTA
617ba13b
MC		 1309 .quota_read = ext4_quota_read,
1310 .quota_write = ext4_quota_write,
96c7e0d9 1311 .get_dquots = ext4_get_dquots,
ac27a0ec 1312#endif
c39a7f84 1313 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1314};
1315
39655164 1316static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1317 .fh_to_dentry = ext4_fh_to_dentry,
1318 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1319 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1320};
1321
1322enum {
1323 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1324 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1325 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1326 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1327 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1328 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1329 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1330 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1331 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1332 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1333 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1334 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
49da9392 1335 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1449032b 1336 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
670e9875 1337 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1449032b 1338 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1339 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1340 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1341 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1342 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
ac27a0ec
DK		 1343};
1344
a447c093 1345static const match_table_t tokens = {
ac27a0ec
DK		 1346 {Opt_bsd_df, "bsddf"},
1347 {Opt_minix_df, "minixdf"},
1348 {Opt_grpid, "grpid"},
1349 {Opt_grpid, "bsdgroups"},
1350 {Opt_nogrpid, "nogrpid"},
1351 {Opt_nogrpid, "sysvgroups"},
1352 {Opt_resgid, "resgid=%u"},
1353 {Opt_resuid, "resuid=%u"},
1354 {Opt_sb, "sb=%u"},
1355 {Opt_err_cont, "errors=continue"},
1356 {Opt_err_panic, "errors=panic"},
1357 {Opt_err_ro, "errors=remount-ro"},
1358 {Opt_nouid32, "nouid32"},
ac27a0ec 1359 {Opt_debug, "debug"},
72578c33
TT		 1360 {Opt_removed, "oldalloc"},
1361 {Opt_removed, "orlov"},
ac27a0ec
DK		 1362 {Opt_user_xattr, "user_xattr"},
1363 {Opt_nouser_xattr, "nouser_xattr"},
1364 {Opt_acl, "acl"},
1365 {Opt_noacl, "noacl"},
e3bb52ae 1366 {Opt_noload, "norecovery"},
5a916be1 1367 {Opt_noload, "noload"},
72578c33
TT		 1368 {Opt_removed, "nobh"},
1369 {Opt_removed, "bh"},
ac27a0ec 1370 {Opt_commit, "commit=%u"},
30773840
TT		 1371 {Opt_min_batch_time, "min_batch_time=%u"},
1372 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1373 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1374 {Opt_journal_path, "journal_path=%s"},
818d276c 1375 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1376 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1377 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1378 {Opt_abort, "abort"},
1379 {Opt_data_journal, "data=journal"},
1380 {Opt_data_ordered, "data=ordered"},
1381 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1382 {Opt_data_err_abort, "data_err=abort"},
1383 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1384 {Opt_offusrjquota, "usrjquota="},
1385 {Opt_usrjquota, "usrjquota=%s"},
1386 {Opt_offgrpjquota, "grpjquota="},
1387 {Opt_grpjquota, "grpjquota=%s"},
1388 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1389 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1390 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1391 {Opt_grpquota, "grpquota"},
1392 {Opt_noquota, "noquota"},
1393 {Opt_quota, "quota"},
1394 {Opt_usrquota, "usrquota"},
49da9392 1395 {Opt_prjquota, "prjquota"},
ac27a0ec 1396 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1397 {Opt_barrier, "barrier"},
1398 {Opt_nobarrier, "nobarrier"},
25ec56b5 1399 {Opt_i_version, "i_version"},
923ae0ff 1400 {Opt_dax, "dax"},
c9de560d 1401 {Opt_stripe, "stripe=%u"},
64769240 1402 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1403 {Opt_lazytime, "lazytime"},
1404 {Opt_nolazytime, "nolazytime"},
670e9875 1405 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
dd919b98 1406 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1407 {Opt_removed, "mblk_io_submit"},
1408 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1409 {Opt_block_validity, "block_validity"},
1410 {Opt_noblock_validity, "noblock_validity"},
240799cd 1411 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1412 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1413 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1414 {Opt_auto_da_alloc, "auto_da_alloc"},
1415 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1416 {Opt_dioread_nolock, "dioread_nolock"},
1417 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1418 {Opt_discard, "discard"},
1419 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1420 {Opt_init_itable, "init_itable=%u"},
1421 {Opt_init_itable, "init_itable"},
1422 {Opt_noinit_itable, "noinit_itable"},
df981d03 1423 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1424 {Opt_test_dummy_encryption, "test_dummy_encryption"},
cdb7ee4c
TE		 1425 {Opt_nombcache, "nombcache"},
1426 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
c7198b9c
TT		 1427 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1428 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1429 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1430 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1431 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1432 {Opt_err, NULL},
ac27a0ec
DK		 1433};
1434
617ba13b 1435static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1436{
617ba13b 1437 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1438 char *options = (char *) *data;
1439
1440 if (!options || strncmp(options, "sb=", 3) != 0)
1441 return 1; /* Default location */
0b8e58a1 1442
ac27a0ec 1443 options += 3;
0b8e58a1 1444 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1445 sb_block = simple_strtoul(options, &options, 0);
1446 if (*options && *options != ',') {
4776004f 1447 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1448 (char *) *data);
1449 return 1;
1450 }
1451 if (*options == ',')
1452 options++;
1453 *data = (void *) options;
0b8e58a1 1454
ac27a0ec
DK		 1455 return sb_block;
1456}
1457
b3881f74 1458#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
d6006186
EB		 1459static const char deprecated_msg[] =
1460 "Mount option \"%s\" will be removed by %s\n"
437ca0fd 1461 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1462
56c50f11
DM		 1463#ifdef CONFIG_QUOTA
1464static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1465{
1466 struct ext4_sb_info *sbi = EXT4_SB(sb);
1467 char *qname;
03dafb5f 1468 int ret = -1;
56c50f11
DM		 1469
1470 if (sb_any_quota_loaded(sb) &&
1471 !sbi->s_qf_names[qtype]) {
1472 ext4_msg(sb, KERN_ERR,
1473 "Cannot change journaled "
1474 "quota options when quota turned on");
57f73c2c 1475 return -1;
56c50f11 1476 }
e2b911c5 1477 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1478 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1479 "ignored when QUOTA feature is enabled");
1480 return 1;
262b4662 1481 }
56c50f11
DM		 1482 qname = match_strdup(args);
1483 if (!qname) {
1484 ext4_msg(sb, KERN_ERR,
1485 "Not enough memory for storing quotafile name");
57f73c2c 1486 return -1;
56c50f11 1487 }
03dafb5f
CG		 1488 if (sbi->s_qf_names[qtype]) {
1489 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1490 ret = 1;
1491 else
1492 ext4_msg(sb, KERN_ERR,
1493 "%s quota file already specified",
1494 QTYPE2NAME(qtype));
1495 goto errout;
56c50f11 1496 }
03dafb5f 1497 if (strchr(qname, '/')) {
56c50f11
DM		 1498 ext4_msg(sb, KERN_ERR,
1499 "quotafile must be on filesystem root");
03dafb5f 1500 goto errout;
56c50f11 1501 }
03dafb5f 1502 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1503 set_opt(sb, QUOTA);
56c50f11 1504 return 1;
03dafb5f
CG		 1505errout:
1506 kfree(qname);
1507 return ret;
56c50f11
DM		 1508}
1509
1510static int clear_qf_name(struct super_block *sb, int qtype)
1511{
1512
1513 struct ext4_sb_info *sbi = EXT4_SB(sb);
1514
1515 if (sb_any_quota_loaded(sb) &&
1516 sbi->s_qf_names[qtype]) {
1517 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1518 " when quota turned on");
57f73c2c 1519 return -1;
56c50f11 1520 }
03dafb5f 1521 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1522 sbi->s_qf_names[qtype] = NULL;
1523 return 1;
1524}
1525#endif
1526
26092bf5
TT		 1527#define MOPT_SET 0x0001
1528#define MOPT_CLEAR 0x0002
1529#define MOPT_NOSUPPORT 0x0004
1530#define MOPT_EXPLICIT 0x0008
1531#define MOPT_CLEAR_ERR 0x0010
1532#define MOPT_GTE0 0x0020
ac27a0ec 1533#ifdef CONFIG_QUOTA
26092bf5
TT		 1534#define MOPT_Q 0
1535#define MOPT_QFMT 0x0040
1536#else
1537#define MOPT_Q MOPT_NOSUPPORT
1538#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1539#endif
26092bf5 1540#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1541#define MOPT_NO_EXT2 0x0100
1542#define MOPT_NO_EXT3 0x0200
1543#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1544#define MOPT_STRING 0x0400
26092bf5
TT		 1545
1546static const struct mount_opts {
1547 int token;
1548 int mount_opt;
1549 int flags;
1550} ext4_mount_opts[] = {
1551 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1552 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1553 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1554 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1555 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1556 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1557 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1558 MOPT_EXT4_ONLY | MOPT_SET},
1559 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1560 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1561 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1562 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1563 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1564 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1565 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1566 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1567 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1568 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1569 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1570 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1571 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1572 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1573 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1574 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1575 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1576 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1577 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1578 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1579 MOPT_NO_EXT2},
8dc0aa8c 1580 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1581 MOPT_NO_EXT2},
26092bf5
TT		 1582 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1583 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1584 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1585 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1586 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1587 {Opt_commit, 0, MOPT_GTE0},
1588 {Opt_max_batch_time, 0, MOPT_GTE0},
1589 {Opt_min_batch_time, 0, MOPT_GTE0},
1590 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1591 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1592 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1593 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1594 {Opt_resuid, 0, MOPT_GTE0},
1595 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1596 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1597 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1598 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1599 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1600 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1601 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1602 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1603 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1604 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1605#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1606 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1607 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1608#else
26092bf5
TT		 1609 {Opt_acl, 0, MOPT_NOSUPPORT},
1610 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1611#endif
26092bf5
TT		 1612 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1613 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
670e9875 1614 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
26092bf5
TT		 1615 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1616 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1617 MOPT_SET | MOPT_Q},
1618 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1619 MOPT_SET | MOPT_Q},
49da9392
JK		 1620 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1621 MOPT_SET | MOPT_Q},
26092bf5 1622 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1623 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1624 MOPT_CLEAR | MOPT_Q},
26092bf5
TT		 1625 {Opt_usrjquota, 0, MOPT_Q},
1626 {Opt_grpjquota, 0, MOPT_Q},
1627 {Opt_offusrjquota, 0, MOPT_Q},
1628 {Opt_offgrpjquota, 0, MOPT_Q},
1629 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1630 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1631 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1632 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1633 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
cdb7ee4c 1634 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
26092bf5
TT		 1635 {Opt_err, 0, 0}
1636};
1637
1638static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1639 substring_t *args, unsigned long *journal_devnum,
1640 unsigned int *journal_ioprio, int is_remount)
1641{
1642 struct ext4_sb_info *sbi = EXT4_SB(sb);
1643 const struct mount_opts *m;
08cefc7a
EB		 1644 kuid_t uid;
1645 kgid_t gid;
26092bf5
TT		 1646 int arg = 0;
1647
57f73c2c
TT		 1648#ifdef CONFIG_QUOTA
1649 if (token == Opt_usrjquota)
1650 return set_qf_name(sb, USRQUOTA, &args[0]);
1651 else if (token == Opt_grpjquota)
1652 return set_qf_name(sb, GRPQUOTA, &args[0]);
1653 else if (token == Opt_offusrjquota)
1654 return clear_qf_name(sb, USRQUOTA);
1655 else if (token == Opt_offgrpjquota)
1656 return clear_qf_name(sb, GRPQUOTA);
1657#endif
26092bf5 1658 switch (token) {
f7048605
TT		 1659 case Opt_noacl:
1660 case Opt_nouser_xattr:
1661 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1662 break;
26092bf5
TT		 1663 case Opt_sb:
1664 return 1; /* handled by get_sb_block() */
1665 case Opt_removed:
5f3633e3 1666 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1667 return 1;
26092bf5
TT		 1668 case Opt_abort:
1669 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1670 return 1;
1671 case Opt_i_version:
1672 sb->s_flags |= MS_I_VERSION;
1673 return 1;
a26f4992
TT		 1674 case Opt_lazytime:
1675 sb->s_flags |= MS_LAZYTIME;
1676 return 1;
1677 case Opt_nolazytime:
1678 sb->s_flags &= ~MS_LAZYTIME;
1679 return 1;
26092bf5
TT		 1680 }
1681
5f3633e3
JK		 1682 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1683 if (token == m->token)
1684 break;
1685
1686 if (m->token == Opt_err) {
1687 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1688 "or missing value", opt);
1689 return -1;
1690 }
1691
8dc0aa8c
TT		 1692 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1693 ext4_msg(sb, KERN_ERR,
1694 "Mount option \"%s\" incompatible with ext2", opt);
1695 return -1;
1696 }
1697 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1698 ext4_msg(sb, KERN_ERR,
1699 "Mount option \"%s\" incompatible with ext3", opt);
1700 return -1;
1701 }
1702
ad4eec61 1703 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1704 return -1;
1705 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1706 return -1;
c93cf2d7
DM		 1707 if (m->flags & MOPT_EXPLICIT) {
1708 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1709 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1710 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1711 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1712 } else
1713 return -1;
1714 }
5f3633e3
JK		 1715 if (m->flags & MOPT_CLEAR_ERR)
1716 clear_opt(sb, ERRORS_MASK);
1717 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1718 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1719 "options when quota turned on");
1720 return -1;
1721 }
1722
1723 if (m->flags & MOPT_NOSUPPORT) {
1724 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1725 } else if (token == Opt_commit) {
1726 if (arg == 0)
1727 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1728 sbi->s_commit_interval = HZ * arg;
670e9875
TT		 1729 } else if (token == Opt_debug_want_extra_isize) {
1730 sbi->s_want_extra_isize = arg;
5f3633e3 1731 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1732 sbi->s_max_batch_time = arg;
1733 } else if (token == Opt_min_batch_time) {
1734 sbi->s_min_batch_time = arg;
1735 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1736 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1737 ext4_msg(sb, KERN_ERR,
1738 "EXT4-fs: inode_readahead_blks must be "
1739 "0 or a power of 2 smaller than 2^31");
26092bf5 1740 return -1;
5f3633e3
JK		 1741 }
1742 sbi->s_inode_readahead_blks = arg;
1743 } else if (token == Opt_init_itable) {
1744 set_opt(sb, INIT_INODE_TABLE);
1745 if (!args->from)
1746 arg = EXT4_DEF_LI_WAIT_MULT;
1747 sbi->s_li_wait_mult = arg;
1748 } else if (token == Opt_max_dir_size_kb) {
1749 sbi->s_max_dir_size_kb = arg;
1750 } else if (token == Opt_stripe) {
1751 sbi->s_stripe = arg;
1752 } else if (token == Opt_resuid) {
1753 uid = make_kuid(current_user_ns(), arg);
1754 if (!uid_valid(uid)) {
1755 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1756 return -1;
1757 }
5f3633e3
JK		 1758 sbi->s_resuid = uid;
1759 } else if (token == Opt_resgid) {
1760 gid = make_kgid(current_user_ns(), arg);
1761 if (!gid_valid(gid)) {
1762 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1763 return -1;
1764 }
1765 sbi->s_resgid = gid;
1766 } else if (token == Opt_journal_dev) {
1767 if (is_remount) {
1768 ext4_msg(sb, KERN_ERR,
1769 "Cannot specify journal on remount");
1770 return -1;
1771 }
1772 *journal_devnum = arg;
ad4eec61
ES		 1773 } else if (token == Opt_journal_path) {
1774 char *journal_path;
1775 struct inode *journal_inode;
1776 struct path path;
1777 int error;
1778
1779 if (is_remount) {
1780 ext4_msg(sb, KERN_ERR,
1781 "Cannot specify journal on remount");
1782 return -1;
1783 }
1784 journal_path = match_strdup(&args[0]);
1785 if (!journal_path) {
1786 ext4_msg(sb, KERN_ERR, "error: could not dup "
1787 "journal device string");
1788 return -1;
1789 }
1790
1791 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1792 if (error) {
1793 ext4_msg(sb, KERN_ERR, "error: could not find "
1794 "journal device path: error %d", error);
1795 kfree(journal_path);
1796 return -1;
1797 }
1798
2b0143b5 1799 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1800 if (!S_ISBLK(journal_inode->i_mode)) {
1801 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1802 "is not a block device", journal_path);
1803 path_put(&path);
1804 kfree(journal_path);
1805 return -1;
1806 }
1807
1808 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1809 path_put(&path);
1810 kfree(journal_path);
5f3633e3
JK		 1811 } else if (token == Opt_journal_ioprio) {
1812 if (arg > 7) {
1813 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1814 " (must be 0-7)");
1815 return -1;
1816 }
1817 *journal_ioprio =
1818 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1819 } else if (token == Opt_test_dummy_encryption) {
1820#ifdef CONFIG_EXT4_FS_ENCRYPTION
1821 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1822 ext4_msg(sb, KERN_WARNING,
1823 "Test dummy encryption mode enabled");
1824#else
1825 ext4_msg(sb, KERN_WARNING,
1826 "Test dummy encryption mount option ignored");
1827#endif
5f3633e3
JK		 1828 } else if (m->flags & MOPT_DATAJ) {
1829 if (is_remount) {
1830 if (!sbi->s_journal)
1831 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1832 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1833 ext4_msg(sb, KERN_ERR,
26092bf5 1834 "Cannot change data mode on remount");
26092bf5 1835 return -1;
ac27a0ec 1836 }
26092bf5 1837 } else {
5f3633e3
JK		 1838 clear_opt(sb, DATA_FLAGS);
1839 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1840 }
5f3633e3
JK		 1841#ifdef CONFIG_QUOTA
1842 } else if (m->flags & MOPT_QFMT) {
1843 if (sb_any_quota_loaded(sb) &&
1844 sbi->s_jquota_fmt != m->mount_opt) {
1845 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1846 "quota options when quota turned on");
1847 return -1;
1848 }
e2b911c5 1849 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1850 ext4_msg(sb, KERN_INFO,
1851 "Quota format mount options ignored "
262b4662 1852 "when QUOTA feature is enabled");
c325a67c 1853 return 1;
262b4662 1854 }
5f3633e3 1855 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 1856#endif
923ae0ff 1857 } else if (token == Opt_dax) {
ef83b6e8
DW		 1858#ifdef CONFIG_FS_DAX
1859 ext4_msg(sb, KERN_WARNING,
1860 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1861 sbi->s_mount_opt |= m->mount_opt;
1862#else
923ae0ff
RZ		 1863 ext4_msg(sb, KERN_INFO, "dax option not supported");
1864 return -1;
5f3633e3 1865#endif
7915a861
AN		 1866 } else if (token == Opt_data_err_abort) {
1867 sbi->s_mount_opt |= m->mount_opt;
1868 } else if (token == Opt_data_err_ignore) {
1869 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 1870 } else {
1871 if (!args->from)
1872 arg = 1;
1873 if (m->flags & MOPT_CLEAR)
1874 arg = !arg;
1875 else if (unlikely(!(m->flags & MOPT_SET))) {
1876 ext4_msg(sb, KERN_WARNING,
1877 "buggy handling of option %s", opt);
1878 WARN_ON(1);
1879 return -1;
1880 }
1881 if (arg != 0)
1882 sbi->s_mount_opt |= m->mount_opt;
1883 else
1884 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1885 }
5f3633e3 1886 return 1;
26092bf5
TT		 1887}
1888
1889static int parse_options(char *options, struct super_block *sb,
1890 unsigned long *journal_devnum,
1891 unsigned int *journal_ioprio,
1892 int is_remount)
1893{
1894 struct ext4_sb_info *sbi = EXT4_SB(sb);
1895 char *p;
1896 substring_t args[MAX_OPT_ARGS];
1897 int token;
1898
1899 if (!options)
1900 return 1;
1901
1902 while ((p = strsep(&options, ",")) != NULL) {
1903 if (!*p)
1904 continue;
1905 /*
1906 * Initialize args struct so we know whether arg was
1907 * found; some options take optional arguments.
1908 */
caecd0af 1909 args[0].to = args[0].from = NULL;
26092bf5
TT		 1910 token = match_token(p, tokens, args);
1911 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1912 journal_ioprio, is_remount) < 0)
1913 return 0;
ac27a0ec
DK		 1914 }
1915#ifdef CONFIG_QUOTA
49da9392
JK		 1916 /*
1917 * We do the test below only for project quotas. 'usrquota' and
1918 * 'grpquota' mount options are allowed even without quota feature
1919 * to support legacy quotas in quota files.
1920 */
1921 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1922 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1923 "Cannot enable project quota enforcement.");
1924 return 0;
1925 }
1926 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1927 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1928 clear_opt(sb, USRQUOTA);
ac27a0ec 1929
482a7425 1930 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1931 clear_opt(sb, GRPQUOTA);
ac27a0ec 1932
56c50f11 1933 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1934 ext4_msg(sb, KERN_ERR, "old and new quota "
1935 "format mixing");
ac27a0ec
DK		 1936 return 0;
1937 }
1938
1939 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1940 ext4_msg(sb, KERN_ERR, "journaled quota format "
1941 "not specified");
ac27a0ec
DK		 1942 return 0;
1943 }
ac27a0ec
DK		 1944 }
1945#endif
261cb20c
JK		 1946 if (test_opt(sb, DIOREAD_NOLOCK)) {
1947 int blocksize =
1948 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1949
09cbfeaf 1950 if (blocksize < PAGE_SIZE) {
261cb20c
JK		 1951 ext4_msg(sb, KERN_ERR, "can't mount with "
1952 "dioread_nolock if block size != PAGE_SIZE");
1953 return 0;
1954 }
1955 }
ac27a0ec
DK		 1956 return 1;
1957}
1958
2adf6da8
TT		 1959static inline void ext4_show_quota_options(struct seq_file *seq,
1960 struct super_block *sb)
1961{
1962#if defined(CONFIG_QUOTA)
1963 struct ext4_sb_info *sbi = EXT4_SB(sb);
1964
1965 if (sbi->s_jquota_fmt) {
1966 char *fmtname = "";
1967
1968 switch (sbi->s_jquota_fmt) {
1969 case QFMT_VFS_OLD:
1970 fmtname = "vfsold";
1971 break;
1972 case QFMT_VFS_V0:
1973 fmtname = "vfsv0";
1974 break;
1975 case QFMT_VFS_V1:
1976 fmtname = "vfsv1";
1977 break;
1978 }
1979 seq_printf(seq, ",jqfmt=%s", fmtname);
1980 }
1981
1982 if (sbi->s_qf_names[USRQUOTA])
a068acf2 1983 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 1984
1985 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 1986 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1987#endif
1988}
1989
5a916be1
TT		 1990static const char *token2str(int token)
1991{
50df9fd5 1992 const struct match_token *t;
5a916be1
TT		 1993
1994 for (t = tokens; t->token != Opt_err; t++)
1995 if (t->token == token && !strchr(t->pattern, '='))
1996 break;
1997 return t->pattern;
1998}
1999
2adf6da8
TT		 2000/*
2001 * Show an option if
2002 * - it's set to a non-default value OR
2003 * - if the per-sb default is different from the global default
2004 */
66acdcf4
TT		 2005static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2006 int nodefs)
2adf6da8 2007{
2adf6da8
TT		 2008 struct ext4_sb_info *sbi = EXT4_SB(sb);
2009 struct ext4_super_block *es = sbi->s_es;
66acdcf4 2010 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 2011 const struct mount_opts *m;
66acdcf4 2012 char sep = nodefs ? '\n' : ',';
2adf6da8 2013
66acdcf4
TT		 2014#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2015#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2016
2017 if (sbi->s_sb_block != 1)
5a916be1
TT		 2018 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2019
2020 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2021 int want_set = m->flags & MOPT_SET;
2022 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2023 (m->flags & MOPT_CLEAR_ERR))
2024 continue;
66acdcf4 2025 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 2026 continue; /* skip if same as the default */
2027 if ((want_set &&
2028 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2029 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2030 continue; /* select Opt_noFoo vs Opt_Foo */
2031 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2032 }
5a916be1 2033
08cefc7a 2034 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2035 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2036 SEQ_OPTS_PRINT("resuid=%u",
2037 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2038 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2039 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2040 SEQ_OPTS_PRINT("resgid=%u",
2041 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2042 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2043 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2044 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2045 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2046 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2047 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2048 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2049 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2050 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2051 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2052 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2053 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2054 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 2055 if (sb->s_flags & MS_I_VERSION)
5a916be1 2056 SEQ_OPTS_PUTS("i_version");
66acdcf4 2057 if (nodefs || sbi->s_stripe)
5a916be1 2058 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 2059 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2060 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2061 SEQ_OPTS_PUTS("data=journal");
2062 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2063 SEQ_OPTS_PUTS("data=ordered");
2064 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2065 SEQ_OPTS_PUTS("data=writeback");
2066 }
66acdcf4
TT		 2067 if (nodefs ||
2068 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2069 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2070 sbi->s_inode_readahead_blks);
2adf6da8 2071
66acdcf4
TT		 2072 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2073 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2074 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2075 if (nodefs || sbi->s_max_dir_size_kb)
2076 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2077 if (test_opt(sb, DATA_ERR_ABORT))
2078 SEQ_OPTS_PUTS("data_err=abort");
2adf6da8
TT		 2079
2080 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2081 return 0;
2082}
2083
66acdcf4
TT		 2084static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2085{
2086 return _ext4_show_options(seq, root->d_sb, 0);
2087}
2088
ebd173be 2089int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2090{
2091 struct super_block *sb = seq->private;
2092 int rc;
2093
2094 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2095 rc = _ext4_show_options(seq, sb, 1);
2096 seq_puts(seq, "\n");
2097 return rc;
2098}
2099
617ba13b 2100static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2101 int read_only)
2102{
617ba13b 2103 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2104 int res = 0;
2105
617ba13b 2106 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2107 ext4_msg(sb, KERN_ERR, "revision level too high, "
2108 "forcing read-only mode");
ac27a0ec
DK		 2109 res = MS_RDONLY;
2110 }
2111 if (read_only)
281b5995 2112 goto done;
617ba13b 2113 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2114 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2115 "running e2fsck is recommended");
c8b459f4 2116 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2117 ext4_msg(sb, KERN_WARNING,
2118 "warning: mounting fs with errors, "
2119 "running e2fsck is recommended");
ed3ce80a 2120 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2121 le16_to_cpu(es->s_mnt_count) >=
2122 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2123 ext4_msg(sb, KERN_WARNING,
2124 "warning: maximal mount count reached, "
2125 "running e2fsck is recommended");
ac27a0ec
DK		 2126 else if (le32_to_cpu(es->s_checkinterval) &&
2127 (le32_to_cpu(es->s_lastcheck) +
2128 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 2129 ext4_msg(sb, KERN_WARNING,
2130 "warning: checktime reached, "
2131 "running e2fsck is recommended");
0b8e58a1 2132 if (!sbi->s_journal)
0390131b 2133 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2134 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2135 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2136 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 2137 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 2138 ext4_update_dynamic_rev(sb);
0390131b 2139 if (sbi->s_journal)
e2b911c5 2140 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2141
e2d67052 2142 ext4_commit_super(sb, 1);
281b5995 2143done:
ac27a0ec 2144 if (test_opt(sb, DEBUG))
a9df9a49 2145 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2146 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2147 sb->s_blocksize,
2148 sbi->s_groups_count,
617ba13b
MC		 2149 EXT4_BLOCKS_PER_GROUP(sb),
2150 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2151 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2152
7abc52c2 2153 cleancache_init_fs(sb);
ac27a0ec
DK		 2154 return res;
2155}
2156
117fff10
TT		 2157int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2158{
2159 struct ext4_sb_info *sbi = EXT4_SB(sb);
2160 struct flex_groups *new_groups;
2161 int size;
2162
2163 if (!sbi->s_log_groups_per_flex)
2164 return 0;
2165
2166 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2167 if (size <= sbi->s_flex_groups_allocated)
2168 return 0;
2169
2170 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
a7c3e901 2171 new_groups = kvzalloc(size, GFP_KERNEL);
117fff10
TT		 2172 if (!new_groups) {
2173 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2174 size / (int) sizeof(struct flex_groups));
2175 return -ENOMEM;
2176 }
2177
2178 if (sbi->s_flex_groups) {
2179 memcpy(new_groups, sbi->s_flex_groups,
2180 (sbi->s_flex_groups_allocated *
2181 sizeof(struct flex_groups)));
b93b41d4 2182 kvfree(sbi->s_flex_groups);
117fff10
TT		 2183 }
2184 sbi->s_flex_groups = new_groups;
2185 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2186 return 0;
2187}
2188
772cb7c8
JS		 2189static int ext4_fill_flex_info(struct super_block *sb)
2190{
2191 struct ext4_sb_info *sbi = EXT4_SB(sb);
2192 struct ext4_group_desc *gdp = NULL;
772cb7c8 2193 ext4_group_t flex_group;
117fff10 2194 int i, err;
772cb7c8 2195
503358ae 2196 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2197 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2198 sbi->s_log_groups_per_flex = 0;
2199 return 1;
2200 }
2201
117fff10
TT		 2202 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2203 if (err)
9933fc0a 2204 goto failed;
772cb7c8 2205
772cb7c8 2206 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2207 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2208
2209 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2210 atomic_add(ext4_free_inodes_count(sb, gdp),
2211 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2212 atomic64_add(ext4_free_group_clusters(sb, gdp),
2213 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2214 atomic_add(ext4_used_dirs_count(sb, gdp),
2215 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2216 }
2217
2218 return 1;
2219failed:
2220 return 0;
2221}
2222
e2b911c5 2223static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2224 struct ext4_group_desc *gdp)
717d50e4 2225{
b47820ed 2226 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2227 __u16 crc = 0;
feb0ab32 2228 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2229 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2230
9aa5d32b 2231 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2232 /* Use new metadata_csum algorithm */
feb0ab32 2233 __u32 csum32;
b47820ed 2234 __u16 dummy_csum = 0;
feb0ab32 2235
feb0ab32
DW		 2236 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2237 sizeof(le_group));
b47820ed
DJ		 2238 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2239 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2240 sizeof(dummy_csum));
2241 offset += sizeof(dummy_csum);
2242 if (offset < sbi->s_desc_size)
2243 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2244 sbi->s_desc_size - offset);
feb0ab32
DW		 2245
2246 crc = csum32 & 0xFFFF;
2247 goto out;
717d50e4
AD		 2248 }
2249
feb0ab32 2250 /* old crc16 code */
e2b911c5 2251 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2252 return 0;
2253
feb0ab32
DW		 2254 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2255 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2256 crc = crc16(crc, (__u8 *)gdp, offset);
2257 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2258 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2259 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2260 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2261 crc = crc16(crc, (__u8 *)gdp + offset,
2262 le16_to_cpu(sbi->s_es->s_desc_size) -
2263 offset);
2264
2265out:
717d50e4
AD		 2266 return cpu_to_le16(crc);
2267}
2268
feb0ab32 2269int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2270 struct ext4_group_desc *gdp)
2271{
feb0ab32 2272 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2273 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2274 return 0;
2275
2276 return 1;
2277}
2278
feb0ab32
DW		 2279void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2280 struct ext4_group_desc *gdp)
2281{
2282 if (!ext4_has_group_desc_csum(sb))
2283 return;
e2b911c5 2284 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2285}
2286
ac27a0ec 2287/* Called at mount-time, super-block is locked */
bfff6873 2288static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2289 ext4_fsblk_t sb_block,
bfff6873 2290 ext4_group_t *first_not_zeroed)
ac27a0ec 2291{
617ba13b
MC		 2292 struct ext4_sb_info *sbi = EXT4_SB(sb);
2293 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2294 ext4_fsblk_t last_block;
bd81d8ee
LV		 2295 ext4_fsblk_t block_bitmap;
2296 ext4_fsblk_t inode_bitmap;
2297 ext4_fsblk_t inode_table;
ce421581 2298 int flexbg_flag = 0;
bfff6873 2299 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2300
e2b911c5 2301 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2302 flexbg_flag = 1;
2303
af5bc92d 2304 ext4_debug("Checking group descriptors");
ac27a0ec 2305
197cd65a
AM		 2306 for (i = 0; i < sbi->s_groups_count; i++) {
2307 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2308
ce421581 2309 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2310 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2311 else
2312 last_block = first_block +
617ba13b 2313 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2314
bfff6873
LC		 2315 if ((grp == sbi->s_groups_count) &&
2316 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2317 grp = i;
2318
8fadc143 2319 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2320 if (block_bitmap == sb_block) {
2321 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2322 "Block bitmap for group %u overlaps "
2323 "superblock", i);
2324 }
2b2d6d01 2325 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2326 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2327 "Block bitmap for group %u not in group "
b31e1552 2328 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2329 return 0;
2330 }
8fadc143 2331 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2332 if (inode_bitmap == sb_block) {
2333 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2334 "Inode bitmap for group %u overlaps "
2335 "superblock", i);
2336 }
2b2d6d01 2337 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2338 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2339 "Inode bitmap for group %u not in group "
b31e1552 2340 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2341 return 0;
2342 }
8fadc143 2343 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2344 if (inode_table == sb_block) {
2345 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2346 "Inode table for group %u overlaps "
2347 "superblock", i);
2348 }
bd81d8ee 2349 if (inode_table < first_block ||
2b2d6d01 2350 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2351 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2352 "Inode table for group %u not in group "
b31e1552 2353 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2354 return 0;
2355 }
955ce5f5 2356 ext4_lock_group(sb, i);
feb0ab32 2357 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2358 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2359 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2360 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2361 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2362 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2363 ext4_unlock_group(sb, i);
8a266467 2364 return 0;
7ee1ec4c 2365 }
717d50e4 2366 }
955ce5f5 2367 ext4_unlock_group(sb, i);
ce421581
JS		 2368 if (!flexbg_flag)
2369 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2370 }
bfff6873
LC		 2371 if (NULL != first_not_zeroed)
2372 *first_not_zeroed = grp;
ac27a0ec
DK		 2373 return 1;
2374}
2375
617ba13b 2376/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2377 * the superblock) which were deleted from all directories, but held open by
2378 * a process at the time of a crash. We walk the list and try to delete these
2379 * inodes at recovery time (only with a read-write filesystem).
2380 *
2381 * In order to keep the orphan inode chain consistent during traversal (in
2382 * case of crash during recovery), we link each inode into the superblock
2383 * orphan list_head and handle it the same way as an inode deletion during
2384 * normal operation (which journals the operations for us).
2385 *
2386 * We only do an iget() and an iput() on each inode, which is very safe if we
2387 * accidentally point at an in-use or already deleted inode. The worst that
2388 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2389 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2390 * e2fsck was run on this filesystem, and it must have already done the orphan
2391 * inode cleanup for us, so we can safely abort without any further action.
2392 */
2b2d6d01
TT		 2393static void ext4_orphan_cleanup(struct super_block *sb,
2394 struct ext4_super_block *es)
ac27a0ec
DK		 2395{
2396 unsigned int s_flags = sb->s_flags;
2c98eb5e 2397 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec
DK		 2398#ifdef CONFIG_QUOTA
2399 int i;
2400#endif
2401 if (!es->s_last_orphan) {
2402 jbd_debug(4, "no orphan inodes to clean up\n");
2403 return;
2404 }
2405
a8f48a95 2406 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2407 ext4_msg(sb, KERN_ERR, "write access "
2408 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2409 return;
2410 }
2411
d39195c3
AG		 2412 /* Check if feature set would not allow a r/w mount */
2413 if (!ext4_feature_set_ok(sb, 0)) {
2414 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2415 "unknown ROCOMPAT features");
2416 return;
2417 }
2418
617ba13b 2419 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2420 /* don't clear list on RO mount w/ errors */
2421 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2422 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2423 "clearing orphan list.\n");
c25f9bc6
ES		 2424 es->s_last_orphan = 0;
2425 }
ac27a0ec
DK		 2426 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2427 return;
2428 }
2429
2430 if (s_flags & MS_RDONLY) {
b31e1552 2431 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2432 sb->s_flags &= ~MS_RDONLY;
2433 }
2434#ifdef CONFIG_QUOTA
2435 /* Needed for iput() to work correctly and not trash data */
2436 sb->s_flags |= MS_ACTIVE;
2437 /* Turn on quotas so that they are updated correctly */
a2d4a646 2438 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2439 if (EXT4_SB(sb)->s_qf_names[i]) {
2440 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2441 if (ret < 0)
b31e1552
ES		 2442 ext4_msg(sb, KERN_ERR,
2443 "Cannot turn on journaled "
2444 "quota: error %d", ret);
ac27a0ec
DK		 2445 }
2446 }
2447#endif
2448
2449 while (es->s_last_orphan) {
2450 struct inode *inode;
2451
c65d5c6c
VN		 2452 /*
2453 * We may have encountered an error during cleanup; if
2454 * so, skip the rest.
2455 */
2456 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2457 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2458 es->s_last_orphan = 0;
2459 break;
2460 }
2461
97bd42b9
JB		 2462 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2463 if (IS_ERR(inode)) {
ac27a0ec
DK		 2464 es->s_last_orphan = 0;
2465 break;
2466 }
2467
617ba13b 2468 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2469 dquot_initialize(inode);
ac27a0ec 2470 if (inode->i_nlink) {
566370a2
PT		 2471 if (test_opt(sb, DEBUG))
2472 ext4_msg(sb, KERN_DEBUG,
2473 "%s: truncating inode %lu to %lld bytes",
2474 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2475 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2476 inode->i_ino, inode->i_size);
5955102c 2477 inode_lock(inode);
55f252c9 2478 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 2479 ret = ext4_truncate(inode);
2480 if (ret)
2481 ext4_std_error(inode->i_sb, ret);
5955102c 2482 inode_unlock(inode);
ac27a0ec
DK		 2483 nr_truncates++;
2484 } else {
566370a2
PT		 2485 if (test_opt(sb, DEBUG))
2486 ext4_msg(sb, KERN_DEBUG,
2487 "%s: deleting unreferenced inode %lu",
2488 __func__, inode->i_ino);
ac27a0ec
DK		 2489 jbd_debug(2, "deleting unreferenced inode %lu\n",
2490 inode->i_ino);
2491 nr_orphans++;
2492 }
2493 iput(inode); /* The delete magic happens here! */
2494 }
2495
2b2d6d01 2496#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2497
2498 if (nr_orphans)
b31e1552
ES		 2499 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2500 PLURAL(nr_orphans));
ac27a0ec 2501 if (nr_truncates)
b31e1552
ES		 2502 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2503 PLURAL(nr_truncates));
ac27a0ec
DK		 2504#ifdef CONFIG_QUOTA
2505 /* Turn quotas off */
a2d4a646 2506 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2507 if (sb_dqopt(sb)->files[i])
287a8095 2508 dquot_quota_off(sb, i);
ac27a0ec
DK		 2509 }
2510#endif
2511 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2512}
0b8e58a1 2513
cd2291a4
ES		 2514/*
2515 * Maximal extent format file size.
2516 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2517 * extent format containers, within a sector_t, and within i_blocks
2518 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2519 * so that won't be a limiting factor.
2520 *
f17722f9
LC		 2521 * However there is other limiting factor. We do store extents in the form
2522 * of starting block and length, hence the resulting length of the extent
2523 * covering maximum file size must fit into on-disk format containers as
2524 * well. Given that length is always by 1 unit bigger than max unit (because
2525 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2526 *
cd2291a4
ES		 2527 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2528 */
f287a1a5 2529static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2530{
2531 loff_t res;
2532 loff_t upper_limit = MAX_LFS_FILESIZE;
2533
2534 /* small i_blocks in vfs inode? */
f287a1a5 2535 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2536 /*
90c699a9 2537 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2538 * i_block represent total blocks in 512 bytes
2539 * 32 == size of vfs inode i_blocks * 8
2540 */
2541 upper_limit = (1LL << 32) - 1;
2542
2543 /* total blocks in file system block size */
2544 upper_limit >>= (blkbits - 9);
2545 upper_limit <<= blkbits;
2546 }
2547
f17722f9
LC		 2548 /*
2549 * 32-bit extent-start container, ee_block. We lower the maxbytes
2550 * by one fs block, so ee_len can cover the extent of maximum file
2551 * size
2552 */
2553 res = (1LL << 32) - 1;
cd2291a4 2554 res <<= blkbits;
cd2291a4
ES		 2555
2556 /* Sanity check against vm- & vfs- imposed limits */
2557 if (res > upper_limit)
2558 res = upper_limit;
2559
2560 return res;
2561}
ac27a0ec 2562
ac27a0ec 2563/*
cd2291a4 2564 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2565 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2566 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2567 */
f287a1a5 2568static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2569{
617ba13b 2570 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2571 int meta_blocks;
2572 loff_t upper_limit;
0b8e58a1
AD		 2573 /* This is calculated to be the largest file size for a dense, block
2574 * mapped file such that the file's total number of 512-byte sectors,
2575 * including data and all indirect blocks, does not exceed (2^48 - 1).
2576 *
2577 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2578 * number of 512-byte sectors of the file.
0fc1b451
AK		 2579 */
2580
f287a1a5 2581 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2582 /*
90c699a9 2583 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2584 * the inode i_block field represents total file blocks in
2585 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2586 */
2587 upper_limit = (1LL << 32) - 1;
2588
2589 /* total blocks in file system block size */
2590 upper_limit >>= (bits - 9);
2591
2592 } else {
8180a562
AK		 2593 /*
2594 * We use 48 bit ext4_inode i_blocks
2595 * With EXT4_HUGE_FILE_FL set the i_blocks
2596 * represent total number of blocks in
2597 * file system block size
2598 */
0fc1b451
AK		 2599 upper_limit = (1LL << 48) - 1;
2600
0fc1b451
AK		 2601 }
2602
2603 /* indirect blocks */
2604 meta_blocks = 1;
2605 /* double indirect blocks */
2606 meta_blocks += 1 + (1LL << (bits-2));
2607 /* tripple indirect blocks */
2608 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2609
2610 upper_limit -= meta_blocks;
2611 upper_limit <<= bits;
ac27a0ec
DK		 2612
2613 res += 1LL << (bits-2);
2614 res += 1LL << (2*(bits-2));
2615 res += 1LL << (3*(bits-2));
2616 res <<= bits;
2617 if (res > upper_limit)
2618 res = upper_limit;
0fc1b451
AK		 2619
2620 if (res > MAX_LFS_FILESIZE)
2621 res = MAX_LFS_FILESIZE;
2622
ac27a0ec
DK		 2623 return res;
2624}
2625
617ba13b 2626static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2627 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2628{
617ba13b 2629 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2630 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2631 int has_super = 0;
2632
2633 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2634
e2b911c5 2635 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2636 return logical_sb_block + nr + 1;
ac27a0ec 2637 bg = sbi->s_desc_per_block * nr;
617ba13b 2638 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2639 has_super = 1;
0b8e58a1 2640
bd63f6b0
DW		 2641 /*
2642 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2643 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2644 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2645 * compensate.
2646 */
2647 if (sb->s_blocksize == 1024 && nr == 0 &&
2648 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2649 has_super++;
2650
617ba13b 2651 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2652}
2653
c9de560d
AT		 2654/**
2655 * ext4_get_stripe_size: Get the stripe size.
2656 * @sbi: In memory super block info
2657 *
2658 * If we have specified it via mount option, then
2659 * use the mount option value. If the value specified at mount time is
2660 * greater than the blocks per group use the super block value.
2661 * If the super block value is greater than blocks per group return 0.
2662 * Allocator needs it be less than blocks per group.
2663 *
2664 */
2665static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2666{
2667 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2668 unsigned long stripe_width =
2669 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2670 int ret;
c9de560d
AT		 2671
2672 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 2673 ret = sbi->s_stripe;
5469d7c3 2674 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 2675 ret = stripe_width;
5469d7c3 2676 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 2677 ret = stride;
2678 else
2679 ret = 0;
c9de560d 2680
3eb08658
DE		 2681 /*
2682 * If the stripe width is 1, this makes no sense and
2683 * we set it to 0 to turn off stripe handling code.
2684 */
2685 if (ret <= 1)
2686 ret = 0;
c9de560d 2687
3eb08658 2688 return ret;
c9de560d 2689}
ac27a0ec 2690
a13fb1a4
ES		 2691/*
2692 * Check whether this filesystem can be mounted based on
2693 * the features present and the RDONLY/RDWR mount requested.
2694 * Returns 1 if this filesystem can be mounted as requested,
2695 * 0 if it cannot be.
2696 */
2697static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2698{
e2b911c5 2699 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2700 ext4_msg(sb, KERN_ERR,
2701 "Couldn't mount because of "
2702 "unsupported optional features (%x)",
2703 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2704 ~EXT4_FEATURE_INCOMPAT_SUPP));
2705 return 0;
2706 }
2707
2708 if (readonly)
2709 return 1;
2710
e2b911c5 2711 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b
DW		 2712 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2713 sb->s_flags |= MS_RDONLY;
2714 return 1;
2715 }
2716
a13fb1a4 2717 /* Check that feature set is OK for a read-write mount */
e2b911c5 2718 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2719 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2720 "unsupported optional features (%x)",
2721 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2722 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2723 return 0;
2724 }
2725 /*
2726 * Large file size enabled file system can only be mounted
2727 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2728 */
e2b911c5 2729 if (ext4_has_feature_huge_file(sb)) {
a13fb1a4
ES		 2730 if (sizeof(blkcnt_t) < sizeof(u64)) {
2731 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2732 "cannot be mounted RDWR without "
2733 "CONFIG_LBDAF");
2734 return 0;
2735 }
2736 }
e2b911c5 2737 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2738 ext4_msg(sb, KERN_ERR,
2739 "Can't support bigalloc feature without "
2740 "extents feature\n");
2741 return 0;
2742 }
7c319d32
AK		 2743
2744#ifndef CONFIG_QUOTA
e2b911c5 2745 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2746 ext4_msg(sb, KERN_ERR,
2747 "Filesystem with quota feature cannot be mounted RDWR "
2748 "without CONFIG_QUOTA");
2749 return 0;
2750 }
689c958c
LX		 2751 if (ext4_has_feature_project(sb) && !readonly) {
2752 ext4_msg(sb, KERN_ERR,
2753 "Filesystem with project quota feature cannot be mounted RDWR "
2754 "without CONFIG_QUOTA");
2755 return 0;
2756 }
7c319d32 2757#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2758 return 1;
2759}
2760
66e61a9e
TT		 2761/*
2762 * This function is called once a day if we have errors logged
2763 * on the file system
2764 */
2765static void print_daily_error_info(unsigned long arg)
2766{
2767 struct super_block *sb = (struct super_block *) arg;
2768 struct ext4_sb_info *sbi;
2769 struct ext4_super_block *es;
2770
2771 sbi = EXT4_SB(sb);
2772 es = sbi->s_es;
2773
2774 if (es->s_error_count)
ae0f78de
TT		 2775 /* fsck newer than v1.41.13 is needed to clean this condition. */
2776 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2777 le32_to_cpu(es->s_error_count));
2778 if (es->s_first_error_time) {
ae0f78de 2779 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2780 sb->s_id, le32_to_cpu(es->s_first_error_time),
2781 (int) sizeof(es->s_first_error_func),
2782 es->s_first_error_func,
2783 le32_to_cpu(es->s_first_error_line));
2784 if (es->s_first_error_ino)
651e1c3b 2785 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2786 le32_to_cpu(es->s_first_error_ino));
2787 if (es->s_first_error_block)
651e1c3b 2788 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2789 le64_to_cpu(es->s_first_error_block));
651e1c3b 2790 printk(KERN_CONT "\n");
66e61a9e
TT		 2791 }
2792 if (es->s_last_error_time) {
ae0f78de 2793 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2794 sb->s_id, le32_to_cpu(es->s_last_error_time),
2795 (int) sizeof(es->s_last_error_func),
2796 es->s_last_error_func,
2797 le32_to_cpu(es->s_last_error_line));
2798 if (es->s_last_error_ino)
651e1c3b 2799 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2800 le32_to_cpu(es->s_last_error_ino));
2801 if (es->s_last_error_block)
651e1c3b 2802 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2803 le64_to_cpu(es->s_last_error_block));
651e1c3b 2804 printk(KERN_CONT "\n");
66e61a9e
TT		 2805 }
2806 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2807}
2808
bfff6873
LC		 2809/* Find next suitable group and run ext4_init_inode_table */
2810static int ext4_run_li_request(struct ext4_li_request *elr)
2811{
2812 struct ext4_group_desc *gdp = NULL;
2813 ext4_group_t group, ngroups;
2814 struct super_block *sb;
2815 unsigned long timeout = 0;
2816 int ret = 0;
2817
2818 sb = elr->lr_super;
2819 ngroups = EXT4_SB(sb)->s_groups_count;
2820
2821 for (group = elr->lr_next_group; group < ngroups; group++) {
2822 gdp = ext4_get_group_desc(sb, group, NULL);
2823 if (!gdp) {
2824 ret = 1;
2825 break;
2826 }
2827
2828 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2829 break;
2830 }
2831
7f511862 2832 if (group >= ngroups)
bfff6873
LC		 2833 ret = 1;
2834
2835 if (!ret) {
2836 timeout = jiffies;
2837 ret = ext4_init_inode_table(sb, group,
2838 elr->lr_timeout ? 0 : 1);
2839 if (elr->lr_timeout == 0) {
51ce6511
LC		 2840 timeout = (jiffies - timeout) *
2841 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2842 elr->lr_timeout = timeout;
2843 }
2844 elr->lr_next_sched = jiffies + elr->lr_timeout;
2845 elr->lr_next_group = group + 1;
2846 }
bfff6873
LC		 2847 return ret;
2848}
2849
2850/*
2851 * Remove lr_request from the list_request and free the
4ed5c033 2852 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2853 */
2854static void ext4_remove_li_request(struct ext4_li_request *elr)
2855{
2856 struct ext4_sb_info *sbi;
2857
2858 if (!elr)
2859 return;
2860
2861 sbi = elr->lr_sbi;
2862
2863 list_del(&elr->lr_request);
2864 sbi->s_li_request = NULL;
2865 kfree(elr);
2866}
2867
2868static void ext4_unregister_li_request(struct super_block *sb)
2869{
1bb933fb
LC		 2870 mutex_lock(&ext4_li_mtx);
2871 if (!ext4_li_info) {
2872 mutex_unlock(&ext4_li_mtx);
bfff6873 2873 return;
1bb933fb 2874 }
bfff6873
LC		 2875
2876 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2877 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2878 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2879 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2880}
2881
8f1f7453
ES		 2882static struct task_struct *ext4_lazyinit_task;
2883
bfff6873
LC		 2884/*
2885 * This is the function where ext4lazyinit thread lives. It walks
2886 * through the request list searching for next scheduled filesystem.
2887 * When such a fs is found, run the lazy initialization request
2888 * (ext4_rn_li_request) and keep track of the time spend in this
2889 * function. Based on that time we compute next schedule time of
2890 * the request. When walking through the list is complete, compute
2891 * next waking time and put itself into sleep.
2892 */
2893static int ext4_lazyinit_thread(void *arg)
2894{
2895 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2896 struct list_head *pos, *n;
2897 struct ext4_li_request *elr;
4ed5c033 2898 unsigned long next_wakeup, cur;
bfff6873
LC		 2899
2900 BUG_ON(NULL == eli);
2901
bfff6873
LC		 2902cont_thread:
2903 while (true) {
2904 next_wakeup = MAX_JIFFY_OFFSET;
2905
2906 mutex_lock(&eli->li_list_mtx);
2907 if (list_empty(&eli->li_request_list)) {
2908 mutex_unlock(&eli->li_list_mtx);
2909 goto exit_thread;
2910 }
bfff6873 2911 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 2912 int err = 0;
2913 int progress = 0;
bfff6873
LC		 2914 elr = list_entry(pos, struct ext4_li_request,
2915 lr_request);
2916
e22834f0
DM		 2917 if (time_before(jiffies, elr->lr_next_sched)) {
2918 if (time_before(elr->lr_next_sched, next_wakeup))
2919 next_wakeup = elr->lr_next_sched;
2920 continue;
2921 }
2922 if (down_read_trylock(&elr->lr_super->s_umount)) {
2923 if (sb_start_write_trylock(elr->lr_super)) {
2924 progress = 1;
2925 /*
2926 * We hold sb->s_umount, sb can not
2927 * be removed from the list, it is
2928 * now safe to drop li_list_mtx
2929 */
2930 mutex_unlock(&eli->li_list_mtx);
2931 err = ext4_run_li_request(elr);
2932 sb_end_write(elr->lr_super);
2933 mutex_lock(&eli->li_list_mtx);
2934 n = pos->next;
b2c78cd0 2935 }
e22834f0
DM		 2936 up_read((&elr->lr_super->s_umount));
2937 }
2938 /* error, remove the lazy_init job */
2939 if (err) {
2940 ext4_remove_li_request(elr);
2941 continue;
2942 }
2943 if (!progress) {
2944 elr->lr_next_sched = jiffies +
2945 (prandom_u32()
2946 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 2947 }
bfff6873
LC		 2948 if (time_before(elr->lr_next_sched, next_wakeup))
2949 next_wakeup = elr->lr_next_sched;
2950 }
2951 mutex_unlock(&eli->li_list_mtx);
2952
a0acae0e 2953 try_to_freeze();
bfff6873 2954
4ed5c033
LC		 2955 cur = jiffies;
2956 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2957 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2958 cond_resched();
2959 continue;
2960 }
2961
4ed5c033
LC		 2962 schedule_timeout_interruptible(next_wakeup - cur);
2963
8f1f7453
ES		 2964 if (kthread_should_stop()) {
2965 ext4_clear_request_list();
2966 goto exit_thread;
2967 }
bfff6873
LC		 2968 }
2969
2970exit_thread:
2971 /*
2972 * It looks like the request list is empty, but we need
2973 * to check it under the li_list_mtx lock, to prevent any
2974 * additions into it, and of course we should lock ext4_li_mtx
2975 * to atomically free the list and ext4_li_info, because at
2976 * this point another ext4 filesystem could be registering
2977 * new one.
2978 */
2979 mutex_lock(&ext4_li_mtx);
2980 mutex_lock(&eli->li_list_mtx);
2981 if (!list_empty(&eli->li_request_list)) {
2982 mutex_unlock(&eli->li_list_mtx);
2983 mutex_unlock(&ext4_li_mtx);
2984 goto cont_thread;
2985 }
2986 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2987 kfree(ext4_li_info);
2988 ext4_li_info = NULL;
2989 mutex_unlock(&ext4_li_mtx);
2990
2991 return 0;
2992}
2993
2994static void ext4_clear_request_list(void)
2995{
2996 struct list_head *pos, *n;
2997 struct ext4_li_request *elr;
2998
2999 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3000 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3001 elr = list_entry(pos, struct ext4_li_request,
3002 lr_request);
3003 ext4_remove_li_request(elr);
3004 }
3005 mutex_unlock(&ext4_li_info->li_list_mtx);
3006}
3007
3008static int ext4_run_lazyinit_thread(void)
3009{
8f1f7453
ES		 3010 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3011 ext4_li_info, "ext4lazyinit");
3012 if (IS_ERR(ext4_lazyinit_task)) {
3013 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3014 ext4_clear_request_list();
bfff6873
LC		 3015 kfree(ext4_li_info);
3016 ext4_li_info = NULL;
92b97816 3017 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3018 "initialization thread\n",
3019 err);
3020 return err;
3021 }
3022 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3023 return 0;
3024}
3025
3026/*
3027 * Check whether it make sense to run itable init. thread or not.
3028 * If there is at least one uninitialized inode table, return
3029 * corresponding group number, else the loop goes through all
3030 * groups and return total number of groups.
3031 */
3032static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3033{
3034 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3035 struct ext4_group_desc *gdp = NULL;
3036
3037 for (group = 0; group < ngroups; group++) {
3038 gdp = ext4_get_group_desc(sb, group, NULL);
3039 if (!gdp)
3040 continue;
3041
3042 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3043 break;
3044 }
3045
3046 return group;
3047}
3048
3049static int ext4_li_info_new(void)
3050{
3051 struct ext4_lazy_init *eli = NULL;
3052
3053 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3054 if (!eli)
3055 return -ENOMEM;
3056
bfff6873
LC		 3057 INIT_LIST_HEAD(&eli->li_request_list);
3058 mutex_init(&eli->li_list_mtx);
3059
bfff6873
LC		 3060 eli->li_state |= EXT4_LAZYINIT_QUIT;
3061
3062 ext4_li_info = eli;
3063
3064 return 0;
3065}
3066
3067static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3068 ext4_group_t start)
3069{
3070 struct ext4_sb_info *sbi = EXT4_SB(sb);
3071 struct ext4_li_request *elr;
bfff6873
LC		 3072
3073 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3074 if (!elr)
3075 return NULL;
3076
3077 elr->lr_super = sb;
3078 elr->lr_sbi = sbi;
3079 elr->lr_next_group = start;
3080
3081 /*
3082 * Randomize first schedule time of the request to
3083 * spread the inode table initialization requests
3084 * better.
3085 */
dd1f723b
TT		 3086 elr->lr_next_sched = jiffies + (prandom_u32() %
3087 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3088 return elr;
3089}
3090
7f511862
TT		 3091int ext4_register_li_request(struct super_block *sb,
3092 ext4_group_t first_not_zeroed)
bfff6873
LC		 3093{
3094 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3095 struct ext4_li_request *elr = NULL;
bfff6873 3096 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3097 int ret = 0;
bfff6873 3098
7f511862 3099 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3100 if (sbi->s_li_request != NULL) {
3101 /*
3102 * Reset timeout so it can be computed again, because
3103 * s_li_wait_mult might have changed.
3104 */
3105 sbi->s_li_request->lr_timeout = 0;
7f511862 3106 goto out;
51ce6511 3107 }
bfff6873
LC		 3108
3109 if (first_not_zeroed == ngroups ||
3110 (sb->s_flags & MS_RDONLY) ||
55ff3840 3111 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3112 goto out;
bfff6873
LC		 3113
3114 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3115 if (!elr) {
3116 ret = -ENOMEM;
3117 goto out;
3118 }
bfff6873
LC		 3119
3120 if (NULL == ext4_li_info) {
3121 ret = ext4_li_info_new();
3122 if (ret)
3123 goto out;
3124 }
3125
3126 mutex_lock(&ext4_li_info->li_list_mtx);
3127 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3128 mutex_unlock(&ext4_li_info->li_list_mtx);
3129
3130 sbi->s_li_request = elr;
46e4690b
TM		 3131 /*
3132 * set elr to NULL here since it has been inserted to
3133 * the request_list and the removal and free of it is
3134 * handled by ext4_clear_request_list from now on.
3135 */
3136 elr = NULL;
bfff6873
LC		 3137
3138 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3139 ret = ext4_run_lazyinit_thread();
3140 if (ret)
3141 goto out;
3142 }
bfff6873 3143out:
beed5ecb
NK		 3144 mutex_unlock(&ext4_li_mtx);
3145 if (ret)
bfff6873 3146 kfree(elr);
bfff6873
LC		 3147 return ret;
3148}
3149
3150/*
3151 * We do not need to lock anything since this is called on
3152 * module unload.
3153 */
3154static void ext4_destroy_lazyinit_thread(void)
3155{
3156 /*
3157 * If thread exited earlier
3158 * there's nothing to be done.
3159 */
8f1f7453 3160 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3161 return;
3162
8f1f7453 3163 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3164}
3165
25ed6e8a
DW		 3166static int set_journal_csum_feature_set(struct super_block *sb)
3167{
3168 int ret = 1;
3169 int compat, incompat;
3170 struct ext4_sb_info *sbi = EXT4_SB(sb);
3171
9aa5d32b 3172 if (ext4_has_metadata_csum(sb)) {
db9ee220 3173 /* journal checksum v3 */
25ed6e8a 3174 compat = 0;
db9ee220 3175 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3176 } else {
3177 /* journal checksum v1 */
3178 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3179 incompat = 0;
3180 }
3181
feb8c6d3
DW		 3182 jbd2_journal_clear_features(sbi->s_journal,
3183 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3184 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3185 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3186 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3187 ret = jbd2_journal_set_features(sbi->s_journal,
3188 compat, 0,
3189 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3190 incompat);
3191 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3192 ret = jbd2_journal_set_features(sbi->s_journal,
3193 compat, 0,
3194 incompat);
3195 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3196 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3197 } else {
feb8c6d3
DW		 3198 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3199 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3200 }
3201
3202 return ret;
3203}
3204
952fc18e
TT		 3205/*
3206 * Note: calculating the overhead so we can be compatible with
3207 * historical BSD practice is quite difficult in the face of
3208 * clusters/bigalloc. This is because multiple metadata blocks from
3209 * different block group can end up in the same allocation cluster.
3210 * Calculating the exact overhead in the face of clustered allocation
3211 * requires either O(all block bitmaps) in memory or O(number of block
3212 * groups**2) in time. We will still calculate the superblock for
3213 * older file systems --- and if we come across with a bigalloc file
3214 * system with zero in s_overhead_clusters the estimate will be close to
3215 * correct especially for very large cluster sizes --- but for newer
3216 * file systems, it's better to calculate this figure once at mkfs
3217 * time, and store it in the superblock. If the superblock value is
3218 * present (even for non-bigalloc file systems), we will use it.
3219 */
3220static int count_overhead(struct super_block *sb, ext4_group_t grp,
3221 char *buf)
3222{
3223 struct ext4_sb_info *sbi = EXT4_SB(sb);
3224 struct ext4_group_desc *gdp;
3225 ext4_fsblk_t first_block, last_block, b;
3226 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3227 int s, j, count = 0;
3228
e2b911c5 3229 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3230 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3231 sbi->s_itb_per_group + 2);
3232
952fc18e
TT		 3233 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3234 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3235 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3236 for (i = 0; i < ngroups; i++) {
3237 gdp = ext4_get_group_desc(sb, i, NULL);
3238 b = ext4_block_bitmap(sb, gdp);
3239 if (b >= first_block && b <= last_block) {
3240 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3241 count++;
3242 }
3243 b = ext4_inode_bitmap(sb, gdp);
3244 if (b >= first_block && b <= last_block) {
3245 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3246 count++;
3247 }
3248 b = ext4_inode_table(sb, gdp);
3249 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3250 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3251 int c = EXT4_B2C(sbi, b - first_block);
3252 ext4_set_bit(c, buf);
3253 count++;
3254 }
3255 if (i != grp)
3256 continue;
3257 s = 0;
3258 if (ext4_bg_has_super(sb, grp)) {
3259 ext4_set_bit(s++, buf);
3260 count++;
3261 }
c48ae41b
TT		 3262 j = ext4_bg_num_gdb(sb, grp);
3263 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3264 ext4_error(sb, "Invalid number of block group "
3265 "descriptor blocks: %d", j);
3266 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3267 }
c48ae41b
TT		 3268 count += j;
3269 for (; j > 0; j--)
3270 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3271 }
3272 if (!count)
3273 return 0;
3274 return EXT4_CLUSTERS_PER_GROUP(sb) -
3275 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3276}
3277
3278/*
3279 * Compute the overhead and stash it in sbi->s_overhead
3280 */
3281int ext4_calculate_overhead(struct super_block *sb)
3282{
3283 struct ext4_sb_info *sbi = EXT4_SB(sb);
3284 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3285 struct inode *j_inode;
3286 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3287 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3288 ext4_fsblk_t overhead = 0;
4fdb5543 3289 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3290
952fc18e
TT		 3291 if (!buf)
3292 return -ENOMEM;
3293
3294 /*
3295 * Compute the overhead (FS structures). This is constant
3296 * for a given filesystem unless the number of block groups
3297 * changes so we cache the previous value until it does.
3298 */
3299
3300 /*
3301 * All of the blocks before first_data_block are overhead
3302 */
3303 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3304
3305 /*
3306 * Add the overhead found in each block group
3307 */
3308 for (i = 0; i < ngroups; i++) {
3309 int blks;
3310
3311 blks = count_overhead(sb, i, buf);
3312 overhead += blks;
3313 if (blks)
3314 memset(buf, 0, PAGE_SIZE);
3315 cond_resched();
3316 }
3c816ded
EW		 3317
3318 /*
3319 * Add the internal journal blocks whether the journal has been
3320 * loaded or not
3321 */
b003b524 3322 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3323 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3c816ded
EW		 3324 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3325 j_inode = ext4_get_journal_inode(sb, j_inum);
3326 if (j_inode) {
3327 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3328 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3329 iput(j_inode);
3330 } else {
3331 ext4_msg(sb, KERN_ERR, "can't get journal size");
3332 }
3333 }
952fc18e
TT		 3334 sbi->s_overhead = overhead;
3335 smp_wmb();
3336 free_page((unsigned long) buf);
3337 return 0;
3338}
3339
b5799018 3340static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3341{
3342 ext4_fsblk_t resv_clusters;
b5799018 3343 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3344
30fac0f7
JK		 3345 /*
3346 * There's no need to reserve anything when we aren't using extents.
3347 * The space estimates are exact, there are no unwritten extents,
3348 * hole punching doesn't need new metadata... This is needed especially
3349 * to keep ext2/3 backward compatibility.
3350 */
e2b911c5 3351 if (!ext4_has_feature_extents(sb))
b5799018 3352 return;
27dd4385
LC		 3353 /*
3354 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3355 * This should cover the situations where we can not afford to run
3356 * out of space like for example punch hole, or converting
556615dc 3357 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3358 * allocation would require 1, or 2 blocks, higher numbers are
3359 * very rare.
3360 */
b5799018
TT		 3361 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3362 sbi->s_cluster_bits);
27dd4385
LC		 3363
3364 do_div(resv_clusters, 50);
3365 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3366
b5799018 3367 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3368}
3369
2b2d6d01 3370static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3371{
d4c402d9 3372 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3373 struct buffer_head *bh;
617ba13b 3374 struct ext4_super_block *es = NULL;
5aee0f8a 3375 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
617ba13b
MC		 3376 ext4_fsblk_t block;
3377 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3378 ext4_fsblk_t logical_sb_block;
ac27a0ec 3379 unsigned long offset = 0;
ac27a0ec
DK		 3380 unsigned long journal_devnum = 0;
3381 unsigned long def_mount_opts;
3382 struct inode *root;
0390131b 3383 const char *descr;
dcc7dae3 3384 int ret = -ENOMEM;
281b5995 3385 int blocksize, clustersize;
4ec11028
TT		 3386 unsigned int db_count;
3387 unsigned int i;
281b5995 3388 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3389 __u64 blocks_count;
07aa2ea1 3390 int err = 0;
b3881f74 3391 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3392 ext4_group_t first_not_zeroed;
ac27a0ec 3393
5aee0f8a
TT		 3394 if ((data && !orig_data) || !sbi)
3395 goto out_free_base;
705895b6
PE		 3396
3397 sbi->s_blockgroup_lock =
3398 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 3399 if (!sbi->s_blockgroup_lock)
3400 goto out_free_base;
3401
ac27a0ec 3402 sb->s_fs_info = sbi;
2c0544b2 3403 sbi->s_sb = sb;
240799cd 3404 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3405 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3406 if (sb->s_bdev->bd_part)
3407 sbi->s_sectors_written_start =
3408 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3409
9f6200bb 3410 /* Cleanup superblock name */
ec3904dc 3411 strreplace(sb->s_id, '/', '!');
9f6200bb 3412
07aa2ea1 3413 /* -EINVAL is default */
dcc7dae3 3414 ret = -EINVAL;
617ba13b 3415 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3416 if (!blocksize) {
b31e1552 3417 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3418 goto out_fail;
3419 }
3420
3421 /*
617ba13b 3422 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3423 * block sizes. We need to calculate the offset from buffer start.
3424 */
617ba13b 3425 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3426 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3427 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3428 } else {
70bbb3e0 3429 logical_sb_block = sb_block;
ac27a0ec
DK		 3430 }
3431
a8ac900b 3432 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3433 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3434 goto out_fail;
3435 }
3436 /*
3437 * Note: s_es must be initialized as soon as possible because
617ba13b 3438 * some ext4 macro-instructions depend on its value
ac27a0ec 3439 */
2716b802 3440 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3441 sbi->s_es = es;
3442 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3443 if (sb->s_magic != EXT4_SUPER_MAGIC)
3444 goto cantfind_ext4;
afc32f7e 3445 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3446
feb0ab32 3447 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3448 if (ext4_has_feature_metadata_csum(sb) &&
3449 ext4_has_feature_gdt_csum(sb))
363307e6 3450 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3451 "redundant flags; please run fsck.");
3452
d25425f8
DW		 3453 /* Check for a known checksum algorithm */
3454 if (!ext4_verify_csum_type(sb, es)) {
3455 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3456 "unknown checksum algorithm.");
3457 silent = 1;
3458 goto cantfind_ext4;
3459 }
3460
0441984a 3461 /* Load the checksum driver */
dec214d0
TE		 3462 if (ext4_has_feature_metadata_csum(sb) ||
3463 ext4_has_feature_ea_inode(sb)) {
0441984a
DW		 3464 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3465 if (IS_ERR(sbi->s_chksum_driver)) {
3466 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3467 ret = PTR_ERR(sbi->s_chksum_driver);
3468 sbi->s_chksum_driver = NULL;
3469 goto failed_mount;
3470 }
3471 }
3472
a9c47317
DW		 3473 /* Check superblock checksum */
3474 if (!ext4_superblock_csum_verify(sb, es)) {
3475 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3476 "invalid superblock checksum. Run e2fsck?");
3477 silent = 1;
6a797d27 3478 ret = -EFSBADCRC;
a9c47317
DW		 3479 goto cantfind_ext4;
3480 }
3481
3482 /* Precompute checksum seed for all metadata */
e2b911c5 3483 if (ext4_has_feature_csum_seed(sb))
8c81bd8f 3484 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
dec214d0 3485 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
a9c47317
DW		 3486 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3487 sizeof(es->s_uuid));
3488
ac27a0ec
DK		 3489 /* Set defaults before we parse the mount options */
3490 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3491 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3492 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3493 set_opt(sb, DEBUG);
87f26807 3494 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3495 set_opt(sb, GRPID);
617ba13b 3496 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3497 set_opt(sb, NO_UID32);
ea663336 3498 /* xattr user namespace & acls are now defaulted on */
ea663336 3499 set_opt(sb, XATTR_USER);
03010a33 3500#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3501 set_opt(sb, POSIX_ACL);
2e7842b8 3502#endif
98c1a759
DW		 3503 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3504 if (ext4_has_metadata_csum(sb))
3505 set_opt(sb, JOURNAL_CHECKSUM);
3506
617ba13b 3507 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3508 set_opt(sb, JOURNAL_DATA);
617ba13b 3509 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3510 set_opt(sb, ORDERED_DATA);
617ba13b 3511 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3512 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3513
3514 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3515 set_opt(sb, ERRORS_PANIC);
bb4f397a 3516 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3517 set_opt(sb, ERRORS_CONT);
bb4f397a 3518 else
fd8c37ec 3519 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3520 /* block_validity enabled by default; disable with noblock_validity */
3521 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3522 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3523 set_opt(sb, DISCARD);
ac27a0ec 3524
08cefc7a
EB		 3525 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3526 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3527 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3528 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3529 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3530
8b67f04a 3531 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3532 set_opt(sb, BARRIER);
ac27a0ec 3533
dd919b98
AK		 3534 /*
3535 * enable delayed allocation by default
3536 * Use -o nodelalloc to turn it off
3537 */
bc0b75f7 3538 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3539 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3540 set_opt(sb, DELALLOC);
dd919b98 3541
51ce6511
LC		 3542 /*
3543 * set default s_li_wait_mult for lazyinit, for the case there is
3544 * no mount option specified.
3545 */
3546 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3547
5aee0f8a
TT		 3548 if (sbi->s_es->s_mount_opts[0]) {
3549 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3550 sizeof(sbi->s_es->s_mount_opts),
3551 GFP_KERNEL);
3552 if (!s_mount_opts)
3553 goto failed_mount;
3554 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3555 &journal_ioprio, 0)) {
3556 ext4_msg(sb, KERN_WARNING,
3557 "failed to parse options in superblock: %s",
3558 s_mount_opts);
3559 }
3560 kfree(s_mount_opts);
8b67f04a 3561 }
5a916be1 3562 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3563 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3564 &journal_ioprio, 0))
ac27a0ec
DK		 3565 goto failed_mount;
3566
56889787
TT		 3567 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3568 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3569 "with data=journal disables delayed "
3570 "allocation and O_DIRECT support!\n");
3571 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3572 ext4_msg(sb, KERN_ERR, "can't mount with "
3573 "both data=journal and delalloc");
3574 goto failed_mount;
3575 }
3576 if (test_opt(sb, DIOREAD_NOLOCK)) {
3577 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3578 "both data=journal and dioread_nolock");
56889787
TT		 3579 goto failed_mount;
3580 }
923ae0ff
RZ		 3581 if (test_opt(sb, DAX)) {
3582 ext4_msg(sb, KERN_ERR, "can't mount with "
3583 "both data=journal and dax");
3584 goto failed_mount;
3585 }
73b92a2a
SK		 3586 if (ext4_has_feature_encrypt(sb)) {
3587 ext4_msg(sb, KERN_WARNING,
3588 "encrypted files will use data=ordered "
3589 "instead of data journaling mode");
3590 }
56889787
TT		 3591 if (test_opt(sb, DELALLOC))
3592 clear_opt(sb, DELALLOC);
001e4a87
TH		 3593 } else {
3594 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3595 }
3596
ac27a0ec 3597 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3598 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3599
617ba13b 3600 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3601 (ext4_has_compat_features(sb) ||
3602 ext4_has_ro_compat_features(sb) ||
3603 ext4_has_incompat_features(sb)))
b31e1552
ES		 3604 ext4_msg(sb, KERN_WARNING,
3605 "feature flags set on rev 0 fs, "
3606 "running e2fsck is recommended");
469108ff 3607
ed3654eb
TT		 3608 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3609 set_opt2(sb, HURD_COMPAT);
e2b911c5 3610 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3611 ext4_msg(sb, KERN_ERR,
3612 "The Hurd can't support 64-bit file systems");
3613 goto failed_mount;
3614 }
dec214d0
TE		 3615
3616 /*
3617 * ea_inode feature uses l_i_version field which is not
3618 * available in HURD_COMPAT mode.
3619 */
3620 if (ext4_has_feature_ea_inode(sb)) {
3621 ext4_msg(sb, KERN_ERR,
3622 "ea_inode feature is not supported for Hurd");
3623 goto failed_mount;
3624 }
ed3654eb
TT		 3625 }
3626
2035e776
TT		 3627 if (IS_EXT2_SB(sb)) {
3628 if (ext2_feature_set_ok(sb))
3629 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3630 "using the ext4 subsystem");
3631 else {
3632 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3633 "to feature incompatibilities");
3634 goto failed_mount;
3635 }
3636 }
3637
3638 if (IS_EXT3_SB(sb)) {
3639 if (ext3_feature_set_ok(sb))
3640 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3641 "using the ext4 subsystem");
3642 else {
3643 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3644 "to feature incompatibilities");
3645 goto failed_mount;
3646 }
3647 }
3648
ac27a0ec
DK		 3649 /*
3650 * Check feature flags regardless of the revision level, since we
3651 * previously didn't change the revision level when setting the flags,
3652 * so there is a chance incompat flags are set on a rev 0 filesystem.
3653 */
a13fb1a4 3654 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3655 goto failed_mount;
a13fb1a4 3656
261cb20c 3657 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3658 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3659 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552 3660 ext4_msg(sb, KERN_ERR,
8cdf3372
TT		 3661 "Unsupported filesystem blocksize %d (%d log_block_size)",
3662 blocksize, le32_to_cpu(es->s_log_block_size));
3663 goto failed_mount;
3664 }
3665 if (le32_to_cpu(es->s_log_block_size) >
3666 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3667 ext4_msg(sb, KERN_ERR,
3668 "Invalid log block size: %u",
3669 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 3670 goto failed_mount;
3671 }
3672
5b9554dc
TT		 3673 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3674 ext4_msg(sb, KERN_ERR,
3675 "Number of reserved GDT blocks insanely large: %d",
3676 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3677 goto failed_mount;
3678 }
3679
923ae0ff 3680 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
87eefeb4
TK		 3681 err = bdev_dax_supported(sb, blocksize);
3682 if (err)
923ae0ff 3683 goto failed_mount;
923ae0ff
RZ		 3684 }
3685
e2b911c5 3686 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 3687 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3688 es->s_encryption_level);
3689 goto failed_mount;
3690 }
3691
ac27a0ec 3692 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3693 /* Validate the filesystem blocksize */
3694 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3695 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3696 blocksize);
ac27a0ec
DK		 3697 goto failed_mount;
3698 }
3699
2b2d6d01 3700 brelse(bh);
70bbb3e0
AM		 3701 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3702 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3703 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3704 if (!bh) {
b31e1552
ES		 3705 ext4_msg(sb, KERN_ERR,
3706 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3707 goto failed_mount;
3708 }
2716b802 3709 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3710 sbi->s_es = es;
617ba13b 3711 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3712 ext4_msg(sb, KERN_ERR,
3713 "Magic mismatch, very weird!");
ac27a0ec
DK		 3714 goto failed_mount;
3715 }
3716 }
3717
e2b911c5 3718 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 3719 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3720 has_huge_files);
3721 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3722
617ba13b
MC		 3723 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3724 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3725 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3726 } else {
3727 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3728 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3729 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3730 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3731 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3732 ext4_msg(sb, KERN_ERR,
3733 "unsupported inode size: %d",
2b2d6d01 3734 sbi->s_inode_size);
ac27a0ec
DK		 3735 goto failed_mount;
3736 }
ef7f3835
KS		 3737 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3738 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3739 }
0b8e58a1 3740
0d1ee42f 3741 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 3742 if (ext4_has_feature_64bit(sb)) {
8fadc143 3743 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3744 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3745 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3746 ext4_msg(sb, KERN_ERR,
3747 "unsupported descriptor size %lu",
0d1ee42f
AR		 3748 sbi->s_desc_size);
3749 goto failed_mount;
3750 }
3751 } else
3752 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3753
ac27a0ec 3754 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3755 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 3756
617ba13b 3757 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3758 if (sbi->s_inodes_per_block == 0)
617ba13b 3759 goto cantfind_ext4;
cd6bb35b
TT		 3760 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3761 sbi->s_inodes_per_group > blocksize * 8) {
3762 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3763 sbi->s_blocks_per_group);
3764 goto failed_mount;
3765 }
ac27a0ec
DK		 3766 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3767 sbi->s_inodes_per_block;
0d1ee42f 3768 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3769 sbi->s_sbh = bh;
3770 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3771 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3772 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3773
2b2d6d01 3774 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3775 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3776 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 3777 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 3778 i = le32_to_cpu(es->s_flags);
3779 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3780 sbi->s_hash_unsigned = 3;
3781 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3782#ifdef __CHAR_UNSIGNED__
23301410
TT		 3783 if (!(sb->s_flags & MS_RDONLY))
3784 es->s_flags |=
3785 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3786 sbi->s_hash_unsigned = 3;
f99b2589 3787#else
23301410
TT		 3788 if (!(sb->s_flags & MS_RDONLY))
3789 es->s_flags |=
3790 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3791#endif
23301410 3792 }
f99b2589 3793 }
ac27a0ec 3794
281b5995
TT		 3795 /* Handle clustersize */
3796 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 3797 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 3798 if (has_bigalloc) {
3799 if (clustersize < blocksize) {
3800 ext4_msg(sb, KERN_ERR,
3801 "cluster size (%d) smaller than "
3802 "block size (%d)", clustersize, blocksize);
3803 goto failed_mount;
3804 }
8cdf3372
TT		 3805 if (le32_to_cpu(es->s_log_cluster_size) >
3806 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3807 ext4_msg(sb, KERN_ERR,
3808 "Invalid log cluster size: %u",
3809 le32_to_cpu(es->s_log_cluster_size));
3810 goto failed_mount;
3811 }
281b5995
TT		 3812 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3813 le32_to_cpu(es->s_log_block_size);
3814 sbi->s_clusters_per_group =
3815 le32_to_cpu(es->s_clusters_per_group);
3816 if (sbi->s_clusters_per_group > blocksize * 8) {
3817 ext4_msg(sb, KERN_ERR,
3818 "#clusters per group too big: %lu",
3819 sbi->s_clusters_per_group);
3820 goto failed_mount;
3821 }
3822 if (sbi->s_blocks_per_group !=
3823 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3824 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3825 "clusters per group (%lu) inconsistent",
3826 sbi->s_blocks_per_group,
3827 sbi->s_clusters_per_group);
3828 goto failed_mount;
3829 }
3830 } else {
3831 if (clustersize != blocksize) {
3832 ext4_warning(sb, "fragment/cluster size (%d) != "
3833 "block size (%d)", clustersize,
3834 blocksize);
3835 clustersize = blocksize;
3836 }
3837 if (sbi->s_blocks_per_group > blocksize * 8) {
3838 ext4_msg(sb, KERN_ERR,
3839 "#blocks per group too big: %lu",
3840 sbi->s_blocks_per_group);
3841 goto failed_mount;
3842 }
3843 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3844 sbi->s_cluster_bits = 0;
ac27a0ec 3845 }
281b5995
TT		 3846 sbi->s_cluster_ratio = clustersize / blocksize;
3847
960fd856
TT		 3848 /* Do we have standard group size of clustersize * 8 blocks ? */
3849 if (sbi->s_blocks_per_group == clustersize << 3)
3850 set_opt2(sb, STD_GROUP_SIZE);
3851
bf43d84b
ES		 3852 /*
3853 * Test whether we have more sectors than will fit in sector_t,
3854 * and whether the max offset is addressable by the page cache.
3855 */
5a9ae68a 3856 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3857 ext4_blocks_count(es));
5a9ae68a 3858 if (err) {
b31e1552 3859 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3860 " too large to mount safely on this system");
ac27a0ec 3861 if (sizeof(sector_t) < 8)
90c699a9 3862 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3863 goto failed_mount;
3864 }
3865
617ba13b
MC		 3866 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3867 goto cantfind_ext4;
e7c95593 3868
0f2ddca6
FTN		 3869 /* check blocks count against device size */
3870 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3871 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3872 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3873 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3874 ext4_blocks_count(es), blocks_count);
3875 goto failed_mount;
3876 }
3877
0b8e58a1
AD		 3878 /*
3879 * It makes no sense for the first data block to be beyond the end
3880 * of the filesystem.
3881 */
3882 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3883 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3884 "block %u is beyond end of filesystem (%llu)",
3885 le32_to_cpu(es->s_first_data_block),
3886 ext4_blocks_count(es));
e7c95593
ES		 3887 goto failed_mount;
3888 }
bd81d8ee
LV		 3889 blocks_count = (ext4_blocks_count(es) -
3890 le32_to_cpu(es->s_first_data_block) +
3891 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3892 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3893 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3894 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3895 "(block count %llu, first data block %u, "
b31e1552 3896 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3897 ext4_blocks_count(es),
3898 le32_to_cpu(es->s_first_data_block),
3899 EXT4_BLOCKS_PER_GROUP(sb));
3900 goto failed_mount;
3901 }
bd81d8ee 3902 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3903 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3904 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3905 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3906 EXT4_DESC_PER_BLOCK(sb);
3a4b77cd 3907 if (ext4_has_feature_meta_bg(sb)) {
2ba3e6e8 3908 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3a4b77cd
EG		 3909 ext4_msg(sb, KERN_WARNING,
3910 "first meta block group too large: %u "
3911 "(group descriptor block count %u)",
3912 le32_to_cpu(es->s_first_meta_bg), db_count);
3913 goto failed_mount;
3914 }
3915 }
a7c3e901 3916 sbi->s_group_desc = kvmalloc(db_count *
f18a5f21
TT		 3917 sizeof(struct buffer_head *),
3918 GFP_KERNEL);
ac27a0ec 3919 if (sbi->s_group_desc == NULL) {
b31e1552 3920 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3921 ret = -ENOMEM;
ac27a0ec
DK		 3922 goto failed_mount;
3923 }
3924
705895b6 3925 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec 3926
85c8f176
AP		 3927 /* Pre-read the descriptors into the buffer cache */
3928 for (i = 0; i < db_count; i++) {
3929 block = descriptor_loc(sb, logical_sb_block, i);
3930 sb_breadahead(sb, block);
3931 }
3932
ac27a0ec 3933 for (i = 0; i < db_count; i++) {
70bbb3e0 3934 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3935 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3936 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3937 ext4_msg(sb, KERN_ERR,
3938 "can't read group descriptor %d", i);
ac27a0ec
DK		 3939 db_count = i;
3940 goto failed_mount2;
3941 }
3942 }
829fa70d 3943 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 3944 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 3945 ret = -EFSCORRUPTED;
f9ae9cf5 3946 goto failed_mount2;
ac27a0ec 3947 }
772cb7c8 3948
f9ae9cf5 3949 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3950 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3951 spin_lock_init(&sbi->s_next_gen_lock);
3952
04ecddb7
JM		 3953 setup_timer(&sbi->s_err_report, print_daily_error_info,
3954 (unsigned long) sb);
04496411 3955
a75ae78f 3956 /* Register extent status tree shrinker */
eb68d0e2 3957 if (ext4_es_register_shrinker(sbi))
ce7e010a 3958 goto failed_mount3;
ce7e010a 3959
c9de560d 3960 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3961 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3962
f9ae9cf5
TT		 3963 /*
3964 * set up enough so that it can read an inode
3965 */
f6e63f90 3966 sb->s_op = &ext4_sops;
617ba13b
MC		 3967 sb->s_export_op = &ext4_export_ops;
3968 sb->s_xattr = ext4_xattr_handlers;
a7550b30 3969 sb->s_cop = &ext4_cryptops;
ac27a0ec 3970#ifdef CONFIG_QUOTA
617ba13b 3971 sb->dq_op = &ext4_quota_operations;
e2b911c5 3972 if (ext4_has_feature_quota(sb))
1fa5efe3 3973 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3974 else
3975 sb->s_qcop = &ext4_qctl_operations;
689c958c 3976 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 3977#endif
f2fa2ffc
AK		 3978 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3979
ac27a0ec 3980 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3981 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3982
3983 sb->s_root = NULL;
3984
3985 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 3986 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 3987
e2b911c5 3988 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
c5e06d10 3989 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3990 goto failed_mount3a;
c5e06d10 3991
ac27a0ec
DK		 3992 /*
3993 * The first inode we look at is the journal inode. Don't try
3994 * root first: it may be modified in the journal!
3995 */
e2b911c5 3996 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4753d8a2
TT		 3997 err = ext4_load_journal(sb, es, journal_devnum);
3998 if (err)
50460fe8 3999 goto failed_mount3a;
0390131b 4000 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
e2b911c5 4001 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 4002 ext4_msg(sb, KERN_ERR, "required journal recovery "
4003 "suppressed and not mounted read-only");
744692dc 4004 goto failed_mount_wq;
ac27a0ec 4005 } else {
1e381f60
DM		 4006 /* Nojournal mode, all journal mount options are illegal */
4007 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4008 ext4_msg(sb, KERN_ERR, "can't mount with "
4009 "journal_checksum, fs mounted w/o journal");
4010 goto failed_mount_wq;
4011 }
4012 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4013 ext4_msg(sb, KERN_ERR, "can't mount with "
4014 "journal_async_commit, fs mounted w/o journal");
4015 goto failed_mount_wq;
4016 }
4017 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4018 ext4_msg(sb, KERN_ERR, "can't mount with "
4019 "commit=%lu, fs mounted w/o journal",
4020 sbi->s_commit_interval / HZ);
4021 goto failed_mount_wq;
4022 }
4023 if (EXT4_MOUNT_DATA_FLAGS &
4024 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4025 ext4_msg(sb, KERN_ERR, "can't mount with "
4026 "data=, fs mounted w/o journal");
4027 goto failed_mount_wq;
4028 }
4029 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4030 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 4031 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 4032 sbi->s_journal = NULL;
4033 needs_recovery = 0;
4034 goto no_journal;
ac27a0ec
DK		 4035 }
4036
e2b911c5 4037 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 4038 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4039 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4040 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4041 goto failed_mount_wq;
eb40a09c
JS		 4042 }
4043
25ed6e8a
DW		 4044 if (!set_journal_csum_feature_set(sb)) {
4045 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4046 "feature set");
4047 goto failed_mount_wq;
d4da6c9c 4048 }
818d276c 4049
ac27a0ec
DK		 4050 /* We have now updated the journal if required, so we can
4051 * validate the data journaling mode. */
4052 switch (test_opt(sb, DATA_FLAGS)) {
4053 case 0:
4054 /* No mode set, assume a default based on the journal
63f57933
AM		 4055 * capabilities: ORDERED_DATA if the journal can
4056 * cope, else JOURNAL_DATA
4057 */
dab291af
MC		 4058 if (jbd2_journal_check_available_features
4059 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 4060 set_opt(sb, ORDERED_DATA);
ac27a0ec 4061 else
fd8c37ec 4062 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 4063 break;
4064
617ba13b
MC		 4065 case EXT4_MOUNT_ORDERED_DATA:
4066 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4067 if (!jbd2_journal_check_available_features
4068 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4069 ext4_msg(sb, KERN_ERR, "Journal does not support "
4070 "requested data journaling mode");
744692dc 4071 goto failed_mount_wq;
ac27a0ec
DK		 4072 }
4073 default:
4074 break;
4075 }
ab04df78
JK		 4076
4077 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4078 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4079 ext4_msg(sb, KERN_ERR, "can't mount with "
4080 "journal_async_commit in data=ordered mode");
4081 goto failed_mount_wq;
4082 }
4083
b3881f74 4084 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4085
18aadd47
BJ		 4086 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4087
ce7e010a 4088no_journal:
cdb7ee4c
TE		 4089 if (!test_opt(sb, NO_MBCACHE)) {
4090 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4091 if (!sbi->s_ea_block_cache) {
dec214d0 4092 ext4_msg(sb, KERN_ERR,
cdb7ee4c 4093 "Failed to create ea_block_cache");
dec214d0
TE		 4094 goto failed_mount_wq;
4095 }
cdb7ee4c
TE		 4096
4097 if (ext4_has_feature_ea_inode(sb)) {
4098 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4099 if (!sbi->s_ea_inode_cache) {
4100 ext4_msg(sb, KERN_ERR,
4101 "Failed to create ea_inode_cache");
4102 goto failed_mount_wq;
4103 }
4104 }
dec214d0
TE		 4105 }
4106
e2b911c5 4107 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
09cbfeaf 4108 (blocksize != PAGE_SIZE)) {
1cb767cd
TT		 4109 ext4_msg(sb, KERN_ERR,
4110 "Unsupported blocksize for fs encryption");
4111 goto failed_mount_wq;
4112 }
4113
e2b911c5
DW		 4114 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4115 !ext4_has_feature_encrypt(sb)) {
4116 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 4117 ext4_commit_super(sb, 1);
4118 }
4119
952fc18e
TT		 4120 /*
4121 * Get the # of file system overhead blocks from the
4122 * superblock if present.
4123 */
4124 if (es->s_overhead_clusters)
4125 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4126 else {
07aa2ea1
LC		 4127 err = ext4_calculate_overhead(sb);
4128 if (err)
952fc18e
TT		 4129 goto failed_mount_wq;
4130 }
4131
fd89d5f2
TH		 4132 /*
4133 * The maximum number of concurrent works can be high and
4134 * concurrency isn't really necessary. Limit it to 1.
4135 */
2e8fa54e
JK		 4136 EXT4_SB(sb)->rsv_conversion_wq =
4137 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4138 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4139 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4140 ret = -ENOMEM;
2e8fa54e
JK		 4141 goto failed_mount4;
4142 }
4143
ac27a0ec 4144 /*
dab291af 4145 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4146 * so we can safely mount the rest of the filesystem now.
4147 */
4148
1d1fe1ee
DH		 4149 root = ext4_iget(sb, EXT4_ROOT_INO);
4150 if (IS_ERR(root)) {
b31e1552 4151 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4152 ret = PTR_ERR(root);
32a9bb57 4153 root = NULL;
ac27a0ec
DK		 4154 goto failed_mount4;
4155 }
4156 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4157 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4158 iput(root);
ac27a0ec
DK		 4159 goto failed_mount4;
4160 }
48fde701 4161 sb->s_root = d_make_root(root);
1d1fe1ee 4162 if (!sb->s_root) {
b31e1552 4163 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4164 ret = -ENOMEM;
4165 goto failed_mount4;
4166 }
ac27a0ec 4167
7e84b621
ES		 4168 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4169 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4170
4171 /* determine the minimum size of new large inodes, if present */
670e9875
TT		 4172 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4173 sbi->s_want_extra_isize == 0) {
ef7f3835
KS		 4174 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4175 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 4176 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 4177 if (sbi->s_want_extra_isize <
4178 le16_to_cpu(es->s_want_extra_isize))
4179 sbi->s_want_extra_isize =
4180 le16_to_cpu(es->s_want_extra_isize);
4181 if (sbi->s_want_extra_isize <
4182 le16_to_cpu(es->s_min_extra_isize))
4183 sbi->s_want_extra_isize =
4184 le16_to_cpu(es->s_min_extra_isize);
4185 }
4186 }
4187 /* Check if enough inode space is available */
4188 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4189 sbi->s_inode_size) {
4190 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4191 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4192 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4193 "available");
ef7f3835
KS		 4194 }
4195
b5799018 4196 ext4_set_resv_clusters(sb);
27dd4385 4197
6fd058f7
TT		 4198 err = ext4_setup_system_zone(sb);
4199 if (err) {
b31e1552 4200 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4201 "zone (%d)", err);
f9ae9cf5
TT		 4202 goto failed_mount4a;
4203 }
4204
4205 ext4_ext_init(sb);
4206 err = ext4_mb_init(sb);
4207 if (err) {
4208 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4209 err);
dcf2d804 4210 goto failed_mount5;
c2774d84
AK		 4211 }
4212
d5e03cbb
TT		 4213 block = ext4_count_free_clusters(sb);
4214 ext4_free_blocks_count_set(sbi->s_es,
4215 EXT4_C2B(sbi, block));
908c7f19
TH		 4216 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4217 GFP_KERNEL);
d5e03cbb
TT		 4218 if (!err) {
4219 unsigned long freei = ext4_count_free_inodes(sb);
4220 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4221 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4222 GFP_KERNEL);
d5e03cbb
TT		 4223 }
4224 if (!err)
4225 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4226 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4227 if (!err)
908c7f19
TH		 4228 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4229 GFP_KERNEL);
c8585c6f
DJ		 4230 if (!err)
4231 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4232
d5e03cbb
TT		 4233 if (err) {
4234 ext4_msg(sb, KERN_ERR, "insufficient memory");
4235 goto failed_mount6;
4236 }
4237
e2b911c5 4238 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 4239 if (!ext4_fill_flex_info(sb)) {
4240 ext4_msg(sb, KERN_ERR,
4241 "unable to initialize "
4242 "flex_bg meta info!");
4243 goto failed_mount6;
4244 }
4245
bfff6873
LC		 4246 err = ext4_register_li_request(sb, first_not_zeroed);
4247 if (err)
dcf2d804 4248 goto failed_mount6;
bfff6873 4249
b5799018 4250 err = ext4_register_sysfs(sb);
dcf2d804
TM		 4251 if (err)
4252 goto failed_mount7;
3197ebdb 4253
9b2ff357
JK		 4254#ifdef CONFIG_QUOTA
4255 /* Enable quota usage during mount. */
e2b911c5 4256 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
9b2ff357
JK		 4257 err = ext4_enable_quotas(sb);
4258 if (err)
4259 goto failed_mount8;
4260 }
4261#endif /* CONFIG_QUOTA */
4262
617ba13b
MC		 4263 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4264 ext4_orphan_cleanup(sb, es);
4265 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4266 if (needs_recovery) {
b31e1552 4267 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4268 ext4_mark_recovery_complete(sb, es);
4269 }
4270 if (EXT4_SB(sb)->s_journal) {
4271 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4272 descr = " journalled data mode";
4273 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4274 descr = " ordered data mode";
4275 else
4276 descr = " writeback data mode";
4277 } else
4278 descr = "out journal";
4279
79add3a3
LC		 4280 if (test_opt(sb, DISCARD)) {
4281 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4282 if (!blk_queue_discard(q))
4283 ext4_msg(sb, KERN_WARNING,
4284 "mounting with \"discard\" option, but "
4285 "the device does not support discard");
4286 }
4287
e294a537
TT		 4288 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4289 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 4290 "Opts: %.*s%s%s", descr,
4291 (int) sizeof(sbi->s_es->s_mount_opts),
4292 sbi->s_es->s_mount_opts,
e294a537 4293 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4294
66e61a9e
TT		 4295 if (es->s_error_count)
4296 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4297
efbed4dc
TT		 4298 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4299 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4300 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4301 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4302
d4c402d9 4303 kfree(orig_data);
ac27a0ec
DK		 4304 return 0;
4305
617ba13b 4306cantfind_ext4:
ac27a0ec 4307 if (!silent)
b31e1552 4308 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4309 goto failed_mount;
4310
72ba7450
TT		 4311#ifdef CONFIG_QUOTA
4312failed_mount8:
ebd173be 4313 ext4_unregister_sysfs(sb);
72ba7450 4314#endif
dcf2d804
TM		 4315failed_mount7:
4316 ext4_unregister_li_request(sb);
4317failed_mount6:
f9ae9cf5 4318 ext4_mb_release(sb);
d5e03cbb 4319 if (sbi->s_flex_groups)
b93b41d4 4320 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4321 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4322 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4323 percpu_counter_destroy(&sbi->s_dirs_counter);
4324 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4325failed_mount5:
f9ae9cf5
TT		 4326 ext4_ext_release(sb);
4327 ext4_release_system_zone(sb);
4328failed_mount4a:
94bf608a 4329 dput(sb->s_root);
32a9bb57 4330 sb->s_root = NULL;
94bf608a 4331failed_mount4:
b31e1552 4332 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4333 if (EXT4_SB(sb)->rsv_conversion_wq)
4334 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4335failed_mount_wq:
dec214d0
TE		 4336 if (sbi->s_ea_inode_cache) {
4337 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4338 sbi->s_ea_inode_cache = NULL;
4339 }
47387409
TE		 4340 if (sbi->s_ea_block_cache) {
4341 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4342 sbi->s_ea_block_cache = NULL;
82939d79 4343 }
0390131b
FM		 4344 if (sbi->s_journal) {
4345 jbd2_journal_destroy(sbi->s_journal);
4346 sbi->s_journal = NULL;
4347 }
50460fe8 4348failed_mount3a:
d3922a77 4349 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4350failed_mount3:
9105bb14 4351 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4352 if (sbi->s_mmp_tsk)
4353 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4354failed_mount2:
4355 for (i = 0; i < db_count; i++)
4356 brelse(sbi->s_group_desc[i]);
b93b41d4 4357 kvfree(sbi->s_group_desc);
ac27a0ec 4358failed_mount:
0441984a
DW		 4359 if (sbi->s_chksum_driver)
4360 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 4361#ifdef CONFIG_QUOTA
a2d4a646 4362 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4363 kfree(sbi->s_qf_names[i]);
4364#endif
617ba13b 4365 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4366 brelse(bh);
4367out_fail:
4368 sb->s_fs_info = NULL;
f6830165 4369 kfree(sbi->s_blockgroup_lock);
5aee0f8a 4370out_free_base:
ac27a0ec 4371 kfree(sbi);
d4c402d9 4372 kfree(orig_data);
07aa2ea1 4373 return err ? err : ret;
ac27a0ec
DK		 4374}
4375
4376/*
4377 * Setup any per-fs journal parameters now. We'll do this both on
4378 * initial mount, once the journal has been initialised but before we've
4379 * done any recovery; and again on any subsequent remount.
4380 */
617ba13b 4381static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4382{
617ba13b 4383 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4384
30773840
TT		 4385 journal->j_commit_interval = sbi->s_commit_interval;
4386 journal->j_min_batch_time = sbi->s_min_batch_time;
4387 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4388
a931da6a 4389 write_lock(&journal->j_state_lock);
ac27a0ec 4390 if (test_opt(sb, BARRIER))
dab291af 4391 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4392 else
dab291af 4393 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4394 if (test_opt(sb, DATA_ERR_ABORT))
4395 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4396 else
4397 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4398 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4399}
4400
c6cb7e77
EW		 4401static struct inode *ext4_get_journal_inode(struct super_block *sb,
4402 unsigned int journal_inum)
ac27a0ec
DK		 4403{
4404 struct inode *journal_inode;
ac27a0ec 4405
c6cb7e77
EW		 4406 /*
4407 * Test for the existence of a valid inode on disk. Bad things
4408 * happen if we iget() an unused inode, as the subsequent iput()
4409 * will try to delete it.
4410 */
1d1fe1ee
DH		 4411 journal_inode = ext4_iget(sb, journal_inum);
4412 if (IS_ERR(journal_inode)) {
b31e1552 4413 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4414 return NULL;
4415 }
4416 if (!journal_inode->i_nlink) {
4417 make_bad_inode(journal_inode);
4418 iput(journal_inode);
b31e1552 4419 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4420 return NULL;
4421 }
4422
e5f8eab8 4423 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4424 journal_inode, journal_inode->i_size);
1d1fe1ee 4425 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4426 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4427 iput(journal_inode);
4428 return NULL;
4429 }
c6cb7e77
EW		 4430 return journal_inode;
4431}
4432
4433static journal_t *ext4_get_journal(struct super_block *sb,
4434 unsigned int journal_inum)
4435{
4436 struct inode *journal_inode;
4437 journal_t *journal;
4438
4439 BUG_ON(!ext4_has_feature_journal(sb));
4440
4441 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4442 if (!journal_inode)
4443 return NULL;
ac27a0ec 4444
dab291af 4445 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4446 if (!journal) {
b31e1552 4447 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4448 iput(journal_inode);
4449 return NULL;
4450 }
4451 journal->j_private = sb;
617ba13b 4452 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4453 return journal;
4454}
4455
617ba13b 4456static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4457 dev_t j_dev)
4458{
2b2d6d01 4459 struct buffer_head *bh;
ac27a0ec 4460 journal_t *journal;
617ba13b
MC		 4461 ext4_fsblk_t start;
4462 ext4_fsblk_t len;
ac27a0ec 4463 int hblock, blocksize;
617ba13b 4464 ext4_fsblk_t sb_block;
ac27a0ec 4465 unsigned long offset;
2b2d6d01 4466 struct ext4_super_block *es;
ac27a0ec
DK		 4467 struct block_device *bdev;
4468
e2b911c5 4469 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4470
b31e1552 4471 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4472 if (bdev == NULL)
4473 return NULL;
4474
ac27a0ec 4475 blocksize = sb->s_blocksize;
e1defc4f 4476 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4477 if (blocksize < hblock) {
b31e1552
ES		 4478 ext4_msg(sb, KERN_ERR,
4479 "blocksize too small for journal device");
ac27a0ec
DK		 4480 goto out_bdev;
4481 }
4482
617ba13b
MC		 4483 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4484 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4485 set_blocksize(bdev, blocksize);
4486 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4487 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4488 "external journal");
ac27a0ec
DK		 4489 goto out_bdev;
4490 }
4491
2716b802 4492 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4493 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4494 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4495 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4496 ext4_msg(sb, KERN_ERR, "external journal has "
4497 "bad superblock");
ac27a0ec
DK		 4498 brelse(bh);
4499 goto out_bdev;
4500 }
4501
df4763be
DW		 4502 if ((le32_to_cpu(es->s_feature_ro_compat) &
4503 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4504 es->s_checksum != ext4_superblock_csum(sb, es)) {
4505 ext4_msg(sb, KERN_ERR, "external journal has "
4506 "corrupt superblock");
4507 brelse(bh);
4508 goto out_bdev;
4509 }
4510
617ba13b 4511 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4512 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4513 brelse(bh);
4514 goto out_bdev;
4515 }
4516
bd81d8ee 4517 len = ext4_blocks_count(es);
ac27a0ec
DK		 4518 start = sb_block + 1;
4519 brelse(bh); /* we're done with the superblock */
4520
dab291af 4521 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4522 start, len, blocksize);
4523 if (!journal) {
b31e1552 4524 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4525 goto out_bdev;
4526 }
4527 journal->j_private = sb;
dfec8a14 4528 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4529 wait_on_buffer(journal->j_sb_buffer);
4530 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4531 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4532 goto out_journal;
4533 }
4534 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4535 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4536 "user (unsupported) - %d",
ac27a0ec
DK		 4537 be32_to_cpu(journal->j_superblock->s_nr_users));
4538 goto out_journal;
4539 }
617ba13b
MC		 4540 EXT4_SB(sb)->journal_bdev = bdev;
4541 ext4_init_journal_params(sb, journal);
ac27a0ec 4542 return journal;
0b8e58a1 4543
ac27a0ec 4544out_journal:
dab291af 4545 jbd2_journal_destroy(journal);
ac27a0ec 4546out_bdev:
617ba13b 4547 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4548 return NULL;
4549}
4550
617ba13b
MC		 4551static int ext4_load_journal(struct super_block *sb,
4552 struct ext4_super_block *es,
ac27a0ec
DK		 4553 unsigned long journal_devnum)
4554{
4555 journal_t *journal;
4556 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4557 dev_t journal_dev;
4558 int err = 0;
4559 int really_read_only;
4560
e2b911c5 4561 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4562
ac27a0ec
DK		 4563 if (journal_devnum &&
4564 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4565 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4566 "numbers have changed");
ac27a0ec
DK		 4567 journal_dev = new_decode_dev(journal_devnum);
4568 } else
4569 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4570
4571 really_read_only = bdev_read_only(sb->s_bdev);
4572
4573 /*
4574 * Are we loading a blank journal or performing recovery after a
4575 * crash? For recovery, we need to check in advance whether we
4576 * can get read-write access to the device.
4577 */
e2b911c5 4578 if (ext4_has_feature_journal_needs_recovery(sb)) {
ac27a0ec 4579 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4580 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4581 "required on readonly filesystem");
ac27a0ec 4582 if (really_read_only) {
b31e1552
ES		 4583 ext4_msg(sb, KERN_ERR, "write access "
4584 "unavailable, cannot proceed");
ac27a0ec
DK		 4585 return -EROFS;
4586 }
b31e1552
ES		 4587 ext4_msg(sb, KERN_INFO, "write access will "
4588 "be enabled during recovery");
ac27a0ec
DK		 4589 }
4590 }
4591
4592 if (journal_inum && journal_dev) {
b31e1552
ES		 4593 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4594 "and inode journals!");
ac27a0ec
DK		 4595 return -EINVAL;
4596 }
4597
4598 if (journal_inum) {
617ba13b 4599 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4600 return -EINVAL;
4601 } else {
617ba13b 4602 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4603 return -EINVAL;
4604 }
4605
90576c0b 4606 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4607 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4608
e2b911c5 4609 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4610 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4611 if (!err) {
4612 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4613 if (save)
4614 memcpy(save, ((char *) es) +
4615 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4616 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4617 if (save)
4618 memcpy(((char *) es) + EXT4_S_ERR_START,
4619 save, EXT4_S_ERR_LEN);
4620 kfree(save);
4621 }
ac27a0ec
DK		 4622
4623 if (err) {
b31e1552 4624 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4625 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4626 return err;
4627 }
4628
617ba13b
MC		 4629 EXT4_SB(sb)->s_journal = journal;
4630 ext4_clear_journal_err(sb, es);
ac27a0ec 4631
c41303ce 4632 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4633 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4634 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4635
4636 /* Make sure we flush the recovery flag to disk. */
e2d67052 4637 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4638 }
4639
4640 return 0;
4641}
4642
e2d67052 4643static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4644{
e2d67052 4645 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4646 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4647 int error = 0;
ac27a0ec 4648
bdfe0cbd 4649 if (!sbh || block_device_ejected(sb))
c4be0c1d 4650 return error;
71290b36
TT		 4651 /*
4652 * If the file system is mounted read-only, don't update the
4653 * superblock write time. This avoids updating the superblock
4654 * write time when we are mounting the root file system
4655 * read/only but we need to replay the journal; at that point,
4656 * for people who are east of GMT and who make their clock
4657 * tick in localtime for Windows bug-for-bug compatibility,
4658 * the clock is set in the future, and this will cause e2fsck
4659 * to complain and force a full file system check.
4660 */
4661 if (!(sb->s_flags & MS_RDONLY))
4662 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4663 if (sb->s_bdev->bd_part)
4664 es->s_kbytes_written =
4665 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4666 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4667 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4668 else
4669 es->s_kbytes_written =
4670 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4671 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4672 ext4_free_blocks_count_set(es,
57042651
TT		 4673 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4674 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4675 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4676 es->s_free_inodes_count =
4677 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4678 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4679 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4680 ext4_superblock_csum_set(sb);
1566a48a
TT		 4681 if (sync)
4682 lock_buffer(sbh);
4743f839
PKS		 4683 if (buffer_write_io_error(sbh)) {
4684 /*
4685 * Oh, dear. A previous attempt to write the
4686 * superblock failed. This could happen because the
4687 * USB device was yanked out. Or it could happen to
4688 * be a transient write error and maybe the block will
4689 * be remapped. Nothing we can do but to retry the
4690 * write and hope for the best.
4691 */
4692 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4693 "superblock detected");
4694 clear_buffer_write_io_error(sbh);
4695 set_buffer_uptodate(sbh);
4696 }
ac27a0ec 4697 mark_buffer_dirty(sbh);
914258bf 4698 if (sync) {
1566a48a 4699 unlock_buffer(sbh);
564bc402 4700 error = __sync_dirty_buffer(sbh,
00473374 4701 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
c4be0c1d
TS		 4702 if (error)
4703 return error;
4704
4705 error = buffer_write_io_error(sbh);
4706 if (error) {
b31e1552
ES		 4707 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4708 "superblock");
914258bf
TT		 4709 clear_buffer_write_io_error(sbh);
4710 set_buffer_uptodate(sbh);
4711 }
4712 }
c4be0c1d 4713 return error;
ac27a0ec
DK		 4714}
4715
ac27a0ec
DK		 4716/*
4717 * Have we just finished recovery? If so, and if we are mounting (or
4718 * remounting) the filesystem readonly, then we will end up with a
4719 * consistent fs on disk. Record that fact.
4720 */
2b2d6d01
TT		 4721static void ext4_mark_recovery_complete(struct super_block *sb,
4722 struct ext4_super_block *es)
ac27a0ec 4723{
617ba13b 4724 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4725
e2b911c5 4726 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4727 BUG_ON(journal != NULL);
4728 return;
4729 }
dab291af 4730 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4731 if (jbd2_journal_flush(journal) < 0)
4732 goto out;
4733
e2b911c5 4734 if (ext4_has_feature_journal_needs_recovery(sb) &&
ac27a0ec 4735 sb->s_flags & MS_RDONLY) {
e2b911c5 4736 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4737 ext4_commit_super(sb, 1);
ac27a0ec 4738 }
7ffe1ea8
HK		 4739
4740out:
dab291af 4741 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4742}
4743
4744/*
4745 * If we are mounting (or read-write remounting) a filesystem whose journal
4746 * has recorded an error from a previous lifetime, move that error to the
4747 * main filesystem now.
4748 */
2b2d6d01
TT		 4749static void ext4_clear_journal_err(struct super_block *sb,
4750 struct ext4_super_block *es)
ac27a0ec
DK		 4751{
4752 journal_t *journal;
4753 int j_errno;
4754 const char *errstr;
4755
e2b911c5 4756 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4757
617ba13b 4758 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4759
4760 /*
4761 * Now check for any error status which may have been recorded in the
617ba13b 4762 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4763 */
4764
dab291af 4765 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4766 if (j_errno) {
4767 char nbuf[16];
4768
617ba13b 4769 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4770 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4771 "from previous mount: %s", errstr);
12062ddd 4772 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4773
617ba13b
MC		 4774 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4775 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4776 ext4_commit_super(sb, 1);
ac27a0ec 4777
dab291af 4778 jbd2_journal_clear_err(journal);
d796c52e 4779 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4780 }
4781}
4782
4783/*
4784 * Force the running and committing transactions to commit,
4785 * and wait on the commit.
4786 */
617ba13b 4787int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4788{
4789 journal_t *journal;
ac27a0ec
DK		 4790
4791 if (sb->s_flags & MS_RDONLY)
4792 return 0;
4793
617ba13b 4794 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4795 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4796}
4797
617ba13b 4798static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4799{
14ce0cb4 4800 int ret = 0;
9eddacf9 4801 tid_t target;
06a407f1 4802 bool needs_barrier = false;
8d5d02e6 4803 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4804
0db1ff22
TT		 4805 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
4806 return 0;
4807
9bffad1e 4808 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4809 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4810 /*
4811 * Writeback quota in non-journalled quota case - journalled quota has
4812 * no dirty dquots
4813 */
4814 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4815 /*
4816 * Data writeback is possible w/o journal transaction, so barrier must
4817 * being sent at the end of the function. But we can skip it if
4818 * transaction_commit will do it for us.
4819 */
bda32530
TT		 4820 if (sbi->s_journal) {
4821 target = jbd2_get_latest_transaction(sbi->s_journal);
4822 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4823 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4824 needs_barrier = true;
4825
4826 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4827 if (wait)
4828 ret = jbd2_log_wait_commit(sbi->s_journal,
4829 target);
4830 }
4831 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4832 needs_barrier = true;
06a407f1
DM		 4833 if (needs_barrier) {
4834 int err;
4835 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4836 if (!ret)
4837 ret = err;
0390131b 4838 }
06a407f1
DM		 4839
4840 return ret;
4841}
4842
ac27a0ec
DK		 4843/*
4844 * LVM calls this function before a (read-only) snapshot is created. This
4845 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4846 *
4847 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4848 * state independently. It relies on upper layer to stop all data & metadata
4849 * modifications.
ac27a0ec 4850 */
c4be0c1d 4851static int ext4_freeze(struct super_block *sb)
ac27a0ec 4852{
c4be0c1d
TS		 4853 int error = 0;
4854 journal_t *journal;
ac27a0ec 4855
9ca92389
TT		 4856 if (sb->s_flags & MS_RDONLY)
4857 return 0;
ac27a0ec 4858
9ca92389 4859 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4860
bb044576
TT		 4861 if (journal) {
4862 /* Now we set up the journal barrier. */
4863 jbd2_journal_lock_updates(journal);
ac27a0ec 4864
bb044576
TT		 4865 /*
4866 * Don't clear the needs_recovery flag if we failed to
4867 * flush the journal.
4868 */
4869 error = jbd2_journal_flush(journal);
4870 if (error < 0)
4871 goto out;
c642dc9e
ES		 4872
4873 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 4874 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 4875 }
9ca92389 4876
9ca92389 4877 error = ext4_commit_super(sb, 1);
6b0310fb 4878out:
bb044576
TT		 4879 if (journal)
4880 /* we rely on upper layer to stop further updates */
4881 jbd2_journal_unlock_updates(journal);
6b0310fb 4882 return error;
ac27a0ec
DK		 4883}
4884
4885/*
4886 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4887 * flag here, even though the filesystem is not technically dirty yet.
4888 */
c4be0c1d 4889static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4890{
783d9485 4891 if ((sb->s_flags & MS_RDONLY) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 4892 return 0;
4893
c642dc9e
ES		 4894 if (EXT4_SB(sb)->s_journal) {
4895 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 4896 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 4897 }
4898
9ca92389 4899 ext4_commit_super(sb, 1);
c4be0c1d 4900 return 0;
ac27a0ec
DK		 4901}
4902
673c6100
TT		 4903/*
4904 * Structure to save mount options for ext4_remount's benefit
4905 */
4906struct ext4_mount_options {
4907 unsigned long s_mount_opt;
a2595b8a 4908 unsigned long s_mount_opt2;
08cefc7a
EB		 4909 kuid_t s_resuid;
4910 kgid_t s_resgid;
673c6100
TT		 4911 unsigned long s_commit_interval;
4912 u32 s_min_batch_time, s_max_batch_time;
4913#ifdef CONFIG_QUOTA
4914 int s_jquota_fmt;
a2d4a646 4915 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4916#endif
4917};
4918
2b2d6d01 4919static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4920{
2b2d6d01 4921 struct ext4_super_block *es;
617ba13b 4922 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4923 unsigned long old_sb_flags;
617ba13b 4924 struct ext4_mount_options old_opts;
c79d967d 4925 int enable_quota = 0;
8a266467 4926 ext4_group_t g;
b3881f74 4927 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4928 int err = 0;
ac27a0ec 4929#ifdef CONFIG_QUOTA
03dafb5f 4930 int i, j;
ac27a0ec 4931#endif
d4c402d9 4932 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4933
4934 /* Store the original options */
4935 old_sb_flags = sb->s_flags;
4936 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4937 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4938 old_opts.s_resuid = sbi->s_resuid;
4939 old_opts.s_resgid = sbi->s_resgid;
4940 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4941 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4942 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4943#ifdef CONFIG_QUOTA
4944 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4945 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4946 if (sbi->s_qf_names[i]) {
4947 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4948 GFP_KERNEL);
4949 if (!old_opts.s_qf_names[i]) {
4950 for (j = 0; j < i; j++)
4951 kfree(old_opts.s_qf_names[j]);
3e36a163 4952 kfree(orig_data);
03dafb5f
CG		 4953 return -ENOMEM;
4954 }
4955 } else
4956 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4957#endif
b3881f74
TT		 4958 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4959 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4960
661aa520 4961 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4962 err = -EINVAL;
4963 goto restore_opts;
4964 }
4965
6b992ff2 4966 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4967 test_opt(sb, JOURNAL_CHECKSUM)) {
4968 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4969 "during remount not supported; ignoring");
4970 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4971 }
4972
6ae6514b
PS		 4973 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4974 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4975 ext4_msg(sb, KERN_ERR, "can't mount with "
4976 "both data=journal and delalloc");
4977 err = -EINVAL;
4978 goto restore_opts;
4979 }
4980 if (test_opt(sb, DIOREAD_NOLOCK)) {
4981 ext4_msg(sb, KERN_ERR, "can't mount with "
4982 "both data=journal and dioread_nolock");
4983 err = -EINVAL;
4984 goto restore_opts;
4985 }
923ae0ff
RZ		 4986 if (test_opt(sb, DAX)) {
4987 ext4_msg(sb, KERN_ERR, "can't mount with "
4988 "both data=journal and dax");
4989 err = -EINVAL;
4990 goto restore_opts;
4991 }
ab04df78
JK		 4992 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
4993 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4994 ext4_msg(sb, KERN_ERR, "can't mount with "
4995 "journal_async_commit in data=ordered mode");
4996 err = -EINVAL;
4997 goto restore_opts;
4998 }
923ae0ff
RZ		 4999 }
5000
cdb7ee4c
TE		 5001 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5002 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5003 err = -EINVAL;
5004 goto restore_opts;
5005 }
5006
923ae0ff
RZ		 5007 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5008 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5009 "dax flag with busy inodes while remounting");
5010 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 5011 }
5012
4ab2f15b 5013 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 5014 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 5015
5016 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 5017 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 5018
5019 es = sbi->s_es;
5020
b3881f74 5021 if (sbi->s_journal) {
0390131b 5022 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 5023 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5024 }
ac27a0ec 5025
a2fd66d0
TT		 5026 if (*flags & MS_LAZYTIME)
5027 sb->s_flags |= MS_LAZYTIME;
5028
661aa520 5029 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 5030 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 5031 err = -EROFS;
5032 goto restore_opts;
5033 }
5034
5035 if (*flags & MS_RDONLY) {
38c03b34
TT		 5036 err = sync_filesystem(sb);
5037 if (err < 0)
5038 goto restore_opts;
0f0dd62f
CH		 5039 err = dquot_suspend(sb, -1);
5040 if (err < 0)
c79d967d 5041 goto restore_opts;
c79d967d 5042
ac27a0ec
DK		 5043 /*
5044 * First of all, the unconditional stuff we have to do
5045 * to disable replay of the journal when we next remount
5046 */
5047 sb->s_flags |= MS_RDONLY;
5048
5049 /*
5050 * OK, test if we are remounting a valid rw partition
5051 * readonly, and if so set the rdonly flag and then
5052 * mark the partition as valid again.
5053 */
617ba13b
MC		 5054 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5055 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 5056 es->s_state = cpu_to_le16(sbi->s_mount_state);
5057
a63c9eb2 5058 if (sbi->s_journal)
0390131b 5059 ext4_mark_recovery_complete(sb, es);
ac27a0ec 5060 } else {
a13fb1a4 5061 /* Make sure we can mount this feature set readwrite */
e2b911c5 5062 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 5063 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 5064 err = -EROFS;
5065 goto restore_opts;
5066 }
8a266467
TT		 5067 /*
5068 * Make sure the group descriptor checksums
0b8e58a1 5069 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 5070 */
5071 for (g = 0; g < sbi->s_groups_count; g++) {
5072 struct ext4_group_desc *gdp =
5073 ext4_get_group_desc(sb, g, NULL);
5074
feb0ab32 5075 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5076 ext4_msg(sb, KERN_ERR,
5077 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 5078 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 5079 le16_to_cpu(gdp->bg_checksum));
6a797d27 5080 err = -EFSBADCRC;
8a266467
TT		 5081 goto restore_opts;
5082 }
5083 }
5084
ead6596b
ES		 5085 /*
5086 * If we have an unprocessed orphan list hanging
5087 * around from a previously readonly bdev mount,
5088 * require a full umount/remount for now.
5089 */
5090 if (es->s_last_orphan) {
b31e1552 5091 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5092 "remount RDWR because of unprocessed "
5093 "orphan inode list. Please "
b31e1552 5094 "umount/remount instead");
ead6596b
ES		 5095 err = -EINVAL;
5096 goto restore_opts;
5097 }
5098
ac27a0ec
DK		 5099 /*
5100 * Mounting a RDONLY partition read-write, so reread
5101 * and store the current valid flag. (It may have
5102 * been changed by e2fsck since we originally mounted
5103 * the partition.)
5104 */
0390131b
FM		 5105 if (sbi->s_journal)
5106 ext4_clear_journal_err(sb, es);
ac27a0ec 5107 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 5108 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 5109 sb->s_flags &= ~MS_RDONLY;
e2b911c5 5110 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 5111 if (ext4_multi_mount_protect(sb,
5112 le64_to_cpu(es->s_mmp_block))) {
5113 err = -EROFS;
5114 goto restore_opts;
5115 }
c79d967d 5116 enable_quota = 1;
ac27a0ec
DK		 5117 }
5118 }
bfff6873
LC		 5119
5120 /*
5121 * Reinitialize lazy itable initialization thread based on
5122 * current settings
5123 */
5124 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5125 ext4_unregister_li_request(sb);
5126 else {
5127 ext4_group_t first_not_zeroed;
5128 first_not_zeroed = ext4_has_uninit_itable(sb);
5129 ext4_register_li_request(sb, first_not_zeroed);
5130 }
5131
6fd058f7 5132 ext4_setup_system_zone(sb);
d096ad0f 5133 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 5134 ext4_commit_super(sb, 1);
0390131b 5135
ac27a0ec
DK		 5136#ifdef CONFIG_QUOTA
5137 /* Release old quota file names */
a2d4a646 5138 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5139 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5140 if (enable_quota) {
5141 if (sb_any_quota_suspended(sb))
5142 dquot_resume(sb, -1);
e2b911c5 5143 else if (ext4_has_feature_quota(sb)) {
7c319d32 5144 err = ext4_enable_quotas(sb);
07724f98 5145 if (err)
7c319d32 5146 goto restore_opts;
7c319d32
AK		 5147 }
5148 }
ac27a0ec 5149#endif
d4c402d9 5150
a26f4992 5151 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 5152 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5153 kfree(orig_data);
ac27a0ec 5154 return 0;
0b8e58a1 5155
ac27a0ec
DK		 5156restore_opts:
5157 sb->s_flags = old_sb_flags;
5158 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5159 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5160 sbi->s_resuid = old_opts.s_resuid;
5161 sbi->s_resgid = old_opts.s_resgid;
5162 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5163 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5164 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5165#ifdef CONFIG_QUOTA
5166 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5167 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5168 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5169 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5170 }
5171#endif
d4c402d9 5172 kfree(orig_data);
ac27a0ec
DK		 5173 return err;
5174}
5175
689c958c
LX		 5176#ifdef CONFIG_QUOTA
5177static int ext4_statfs_project(struct super_block *sb,
5178 kprojid_t projid, struct kstatfs *buf)
5179{
5180 struct kqid qid;
5181 struct dquot *dquot;
5182 u64 limit;
5183 u64 curblock;
5184
5185 qid = make_kqid_projid(projid);
5186 dquot = dqget(sb, qid);
5187 if (IS_ERR(dquot))
5188 return PTR_ERR(dquot);
5189 spin_lock(&dq_data_lock);
5190
5191 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5192 dquot->dq_dqb.dqb_bsoftlimit :
5193 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5194 if (limit && buf->f_blocks > limit) {
5195 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5196 buf->f_blocks = limit;
5197 buf->f_bfree = buf->f_bavail =
5198 (buf->f_blocks > curblock) ?
5199 (buf->f_blocks - curblock) : 0;
5200 }
5201
5202 limit = dquot->dq_dqb.dqb_isoftlimit ?
5203 dquot->dq_dqb.dqb_isoftlimit :
5204 dquot->dq_dqb.dqb_ihardlimit;
5205 if (limit && buf->f_files > limit) {
5206 buf->f_files = limit;
5207 buf->f_ffree =
5208 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5209 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5210 }
5211
5212 spin_unlock(&dq_data_lock);
5213 dqput(dquot);
5214 return 0;
5215}
5216#endif
5217
2b2d6d01 5218static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5219{
5220 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5221 struct ext4_sb_info *sbi = EXT4_SB(sb);
5222 struct ext4_super_block *es = sbi->s_es;
27dd4385 5223 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5224 u64 fsid;
d02a9391 5225 s64 bfree;
27dd4385 5226 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5227
952fc18e
TT		 5228 if (!test_opt(sb, MINIX_DF))
5229 overhead = sbi->s_overhead;
ac27a0ec 5230
617ba13b 5231 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5232 buf->f_bsize = sb->s_blocksize;
b72f78cb 5233 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5234 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5235 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5236 /* prevent underflow in case that few free space is available */
57042651 5237 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5238 buf->f_bavail = buf->f_bfree -
5239 (ext4_r_blocks_count(es) + resv_blocks);
5240 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5241 buf->f_bavail = 0;
5242 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5243 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5244 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5245 fsid = le64_to_cpup((void *)es->s_uuid) ^
5246 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5247 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5248 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5249
689c958c
LX		 5250#ifdef CONFIG_QUOTA
5251 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5252 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5253 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5254#endif
ac27a0ec
DK		 5255 return 0;
5256}
5257
0b8e58a1
AD		 5258/* Helper function for writing quotas on sync - we need to start transaction
5259 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 5260 * Process 1 Process 2
617ba13b 5261 * ext4_create() quota_sync()
a269eb18 5262 * jbd2_journal_start() write_dquot()
871a2931 5263 * dquot_initialize() down(dqio_mutex)
dab291af 5264 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 5265 *
5266 */
5267
5268#ifdef CONFIG_QUOTA
5269
5270static inline struct inode *dquot_to_inode(struct dquot *dquot)
5271{
4c376dca 5272 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5273}
5274
617ba13b 5275static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5276{
5277 int ret, err;
5278 handle_t *handle;
5279 struct inode *inode;
5280
5281 inode = dquot_to_inode(dquot);
9924a92a 5282 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5283 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5284 if (IS_ERR(handle))
5285 return PTR_ERR(handle);
5286 ret = dquot_commit(dquot);
617ba13b 5287 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5288 if (!ret)
5289 ret = err;
5290 return ret;
5291}
5292
617ba13b 5293static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5294{
5295 int ret, err;
5296 handle_t *handle;
5297
9924a92a 5298 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5299 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5300 if (IS_ERR(handle))
5301 return PTR_ERR(handle);
5302 ret = dquot_acquire(dquot);
617ba13b 5303 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5304 if (!ret)
5305 ret = err;
5306 return ret;
5307}
5308
617ba13b 5309static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5310{
5311 int ret, err;
5312 handle_t *handle;
5313
9924a92a 5314 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5315 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5316 if (IS_ERR(handle)) {
5317 /* Release dquot anyway to avoid endless cycle in dqput() */
5318 dquot_release(dquot);
ac27a0ec 5319 return PTR_ERR(handle);
9c3013e9 5320 }
ac27a0ec 5321 ret = dquot_release(dquot);
617ba13b 5322 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5323 if (!ret)
5324 ret = err;
5325 return ret;
5326}
5327
617ba13b 5328static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5329{
262b4662
JK		 5330 struct super_block *sb = dquot->dq_sb;
5331 struct ext4_sb_info *sbi = EXT4_SB(sb);
5332
2c8be6b2 5333 /* Are we journaling quotas? */
e2b911c5 5334 if (ext4_has_feature_quota(sb) ||
262b4662 5335 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5336 dquot_mark_dquot_dirty(dquot);
617ba13b 5337 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5338 } else {
5339 return dquot_mark_dquot_dirty(dquot);
5340 }
5341}
5342
617ba13b 5343static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5344{
5345 int ret, err;
5346 handle_t *handle;
5347
5348 /* Data block + inode block */
2b0143b5 5349 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5350 if (IS_ERR(handle))
5351 return PTR_ERR(handle);
5352 ret = dquot_commit_info(sb, type);
617ba13b 5353 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5354 if (!ret)
5355 ret = err;
5356 return ret;
5357}
5358
5359/*
5360 * Turn on quotas during mount time - we need to find
5361 * the quota file and such...
5362 */
617ba13b 5363static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5364{
287a8095
CH		 5365 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5366 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5367}
5368
daf647d2
TT		 5369static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5370{
5371 struct ext4_inode_info *ei = EXT4_I(inode);
5372
5373 /* The first argument of lockdep_set_subclass has to be
5374 * *exactly* the same as the argument to init_rwsem() --- in
5375 * this case, in init_once() --- or lockdep gets unhappy
5376 * because the name of the lock is set using the
5377 * stringification of the argument to init_rwsem().
5378 */
5379 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5380 lockdep_set_subclass(&ei->i_data_sem, subclass);
5381}
5382
ac27a0ec
DK		 5383/*
5384 * Standard function to be called on quota_on
5385 */
617ba13b 5386static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 5387 const struct path *path)
ac27a0ec
DK		 5388{
5389 int err;
ac27a0ec
DK		 5390
5391 if (!test_opt(sb, QUOTA))
5392 return -EINVAL;
0623543b 5393
ac27a0ec 5394 /* Quotafile not on the same filesystem? */
d8c9584e 5395 if (path->dentry->d_sb != sb)
ac27a0ec 5396 return -EXDEV;
0623543b
JK		 5397 /* Journaling quota? */
5398 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5399 /* Quotafile not in fs root? */
f00c9e44 5400 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5401 ext4_msg(sb, KERN_WARNING,
5402 "Quota file not on filesystem root. "
5403 "Journaled quota will not work");
2b2d6d01 5404 }
0623543b
JK		 5405
5406 /*
5407 * When we journal data on quota file, we have to flush journal to see
5408 * all updates to the file when we bypass pagecache...
5409 */
0390131b 5410 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5411 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5412 /*
5413 * We don't need to lock updates but journal_flush() could
5414 * otherwise be livelocked...
5415 */
5416 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5417 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5418 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5419 if (err)
7ffe1ea8 5420 return err;
0623543b 5421 }
957153fc 5422
daf647d2
TT		 5423 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5424 err = dquot_quota_on(sb, type, format_id, path);
957153fc 5425 if (err) {
daf647d2
TT		 5426 lockdep_set_quota_inode(path->dentry->d_inode,
5427 I_DATA_SEM_NORMAL);
957153fc
JK		 5428 } else {
5429 struct inode *inode = d_inode(path->dentry);
5430 handle_t *handle;
5431
61a92987
JK		 5432 /*
5433 * Set inode flags to prevent userspace from messing with quota
5434 * files. If this fails, we return success anyway since quotas
5435 * are already enabled and this is not a hard failure.
5436 */
957153fc
JK		 5437 inode_lock(inode);
5438 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5439 if (IS_ERR(handle))
5440 goto unlock_inode;
5441 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5442 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5443 S_NOATIME | S_IMMUTABLE);
5444 ext4_mark_inode_dirty(handle, inode);
5445 ext4_journal_stop(handle);
5446 unlock_inode:
5447 inode_unlock(inode);
5448 }
daf647d2 5449 return err;
ac27a0ec
DK		 5450}
5451
7c319d32
AK		 5452static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5453 unsigned int flags)
5454{
5455 int err;
5456 struct inode *qf_inode;
a2d4a646 5457 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5458 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5459 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5460 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5461 };
5462
e2b911c5 5463 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5464
5465 if (!qf_inums[type])
5466 return -EPERM;
5467
5468 qf_inode = ext4_iget(sb, qf_inums[type]);
5469 if (IS_ERR(qf_inode)) {
5470 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5471 return PTR_ERR(qf_inode);
5472 }
5473
bcb13850
JK		 5474 /* Don't account quota for quota files to avoid recursion */
5475 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5476 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7c319d32
AK		 5477 err = dquot_enable(qf_inode, type, format_id, flags);
5478 iput(qf_inode);
daf647d2
TT		 5479 if (err)
5480 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7c319d32
AK		 5481
5482 return err;
5483}
5484
5485/* Enable usage tracking for all quota types. */
5486static int ext4_enable_quotas(struct super_block *sb)
5487{
5488 int type, err = 0;
a2d4a646 5489 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5490 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5491 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5492 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 5493 };
49da9392
JK		 5494 bool quota_mopt[EXT4_MAXQUOTAS] = {
5495 test_opt(sb, USRQUOTA),
5496 test_opt(sb, GRPQUOTA),
5497 test_opt(sb, PRJQUOTA),
5498 };
7c319d32
AK		 5499
5500 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5501 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5502 if (qf_inums[type]) {
5503 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 5504 DQUOT_USAGE_ENABLED |
5505 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 5506 if (err) {
5507 ext4_warning(sb,
72ba7450
TT		 5508 "Failed to enable quota tracking "
5509 "(type=%d, err=%d). Please run "
5510 "e2fsck to fix.", type, err);
7c319d32
AK		 5511 return err;
5512 }
5513 }
5514 }
5515 return 0;
5516}
5517
ca0e05e4
DM		 5518static int ext4_quota_off(struct super_block *sb, int type)
5519{
21f97697
JK		 5520 struct inode *inode = sb_dqopt(sb)->files[type];
5521 handle_t *handle;
957153fc 5522 int err;
21f97697 5523
87009d86
DM		 5524 /* Force all delayed allocation blocks to be allocated.
5525 * Caller already holds s_umount sem */
5526 if (test_opt(sb, DELALLOC))
ca0e05e4 5527 sync_filesystem(sb);
ca0e05e4 5528
957153fc 5529 if (!inode || !igrab(inode))
0b268590
AG		 5530 goto out;
5531
957153fc 5532 err = dquot_quota_off(sb, type);
964edf66 5533 if (err || ext4_has_feature_quota(sb))
957153fc
JK		 5534 goto out_put;
5535
5536 inode_lock(inode);
61a92987
JK		 5537 /*
5538 * Update modification times of quota files when userspace can
5539 * start looking at them. If we fail, we return success anyway since
5540 * this is not a hard failure and quotas are already disabled.
5541 */
9924a92a 5542 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697 5543 if (IS_ERR(handle))
957153fc
JK		 5544 goto out_unlock;
5545 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5546 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
eeca7ea1 5547 inode->i_mtime = inode->i_ctime = current_time(inode);
21f97697
JK		 5548 ext4_mark_inode_dirty(handle, inode);
5549 ext4_journal_stop(handle);
957153fc
JK		 5550out_unlock:
5551 inode_unlock(inode);
5552out_put:
964edf66 5553 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 5554 iput(inode);
5555 return err;
21f97697 5556out:
ca0e05e4
DM		 5557 return dquot_quota_off(sb, type);
5558}
5559
ac27a0ec
DK		 5560/* Read data from quotafile - avoid pagecache and such because we cannot afford
5561 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5562 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5563 * we don't have to be afraid of races */
617ba13b 5564static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5565 size_t len, loff_t off)
5566{
5567 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5568 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5569 int offset = off & (sb->s_blocksize - 1);
5570 int tocopy;
5571 size_t toread;
5572 struct buffer_head *bh;
5573 loff_t i_size = i_size_read(inode);
5574
5575 if (off > i_size)
5576 return 0;
5577 if (off+len > i_size)
5578 len = i_size-off;
5579 toread = len;
5580 while (toread > 0) {
5581 tocopy = sb->s_blocksize - offset < toread ?
5582 sb->s_blocksize - offset : toread;
1c215028
TT		 5583 bh = ext4_bread(NULL, inode, blk, 0);
5584 if (IS_ERR(bh))
5585 return PTR_ERR(bh);
ac27a0ec
DK		 5586 if (!bh) /* A hole? */
5587 memset(data, 0, tocopy);
5588 else
5589 memcpy(data, bh->b_data+offset, tocopy);
5590 brelse(bh);
5591 offset = 0;
5592 toread -= tocopy;
5593 data += tocopy;
5594 blk++;
5595 }
5596 return len;
5597}
5598
5599/* Write to quotafile (we know the transaction is already started and has
5600 * enough credits) */
617ba13b 5601static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5602 const char *data, size_t len, loff_t off)
5603{
5604 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5605 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5606 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5607 int retries = 0;
ac27a0ec
DK		 5608 struct buffer_head *bh;
5609 handle_t *handle = journal_current_handle();
5610
0390131b 5611 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5612 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5613 " cancelled because transaction is not started",
9c3013e9
JK		 5614 (unsigned long long)off, (unsigned long long)len);
5615 return -EIO;
5616 }
67eeb568
DM		 5617 /*
5618 * Since we account only one data block in transaction credits,
5619 * then it is impossible to cross a block boundary.
5620 */
5621 if (sb->s_blocksize - offset < len) {
5622 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5623 " cancelled because not block aligned",
5624 (unsigned long long)off, (unsigned long long)len);
5625 return -EIO;
5626 }
5627
c5e298ae
TT		 5628 do {
5629 bh = ext4_bread(handle, inode, blk,
5630 EXT4_GET_BLOCKS_CREATE |
5631 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5632 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5633 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5634 if (IS_ERR(bh))
5635 return PTR_ERR(bh);
67eeb568
DM		 5636 if (!bh)
5637 goto out;
5d601255 5638 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5639 err = ext4_journal_get_write_access(handle, bh);
5640 if (err) {
5641 brelse(bh);
1c215028 5642 return err;
ac27a0ec 5643 }
67eeb568
DM		 5644 lock_buffer(bh);
5645 memcpy(bh->b_data+offset, data, len);
5646 flush_dcache_page(bh->b_page);
5647 unlock_buffer(bh);
62d2b5f2 5648 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5649 brelse(bh);
ac27a0ec 5650out:
67eeb568
DM		 5651 if (inode->i_size < off + len) {
5652 i_size_write(inode, off + len);
617ba13b 5653 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5654 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5655 }
67eeb568 5656 return len;
ac27a0ec
DK		 5657}
5658
8f0e8746
TT		 5659static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5660{
5661 const struct quota_format_ops *ops;
5662
5663 if (!sb_has_quota_loaded(sb, qid->type))
5664 return -ESRCH;
5665 ops = sb_dqopt(sb)->ops[qid->type];
5666 if (!ops || !ops->get_next_id)
5667 return -ENOSYS;
5668 return dquot_get_next_id(sb, qid);
5669}
ac27a0ec
DK		 5670#endif
5671
152a0836
AV		 5672static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5673 const char *dev_name, void *data)
ac27a0ec 5674{
152a0836 5675 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5676}
5677
c290ea01 5678#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5679static inline void register_as_ext2(void)
5680{
5681 int err = register_filesystem(&ext2_fs_type);
5682 if (err)
5683 printk(KERN_WARNING
5684 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5685}
5686
5687static inline void unregister_as_ext2(void)
5688{
5689 unregister_filesystem(&ext2_fs_type);
5690}
2035e776
TT		 5691
5692static inline int ext2_feature_set_ok(struct super_block *sb)
5693{
e2b911c5 5694 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776
TT		 5695 return 0;
5696 if (sb->s_flags & MS_RDONLY)
5697 return 1;
e2b911c5 5698 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5699 return 0;
5700 return 1;
5701}
24b58424
TT		 5702#else
5703static inline void register_as_ext2(void) { }
5704static inline void unregister_as_ext2(void) { }
2035e776 5705static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5706#endif
5707
24b58424
TT		 5708static inline void register_as_ext3(void)
5709{
5710 int err = register_filesystem(&ext3_fs_type);
5711 if (err)
5712 printk(KERN_WARNING
5713 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5714}
5715
5716static inline void unregister_as_ext3(void)
5717{
5718 unregister_filesystem(&ext3_fs_type);
5719}
2035e776
TT		 5720
5721static inline int ext3_feature_set_ok(struct super_block *sb)
5722{
e2b911c5 5723 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5724 return 0;
e2b911c5 5725 if (!ext4_has_feature_journal(sb))
2035e776
TT		 5726 return 0;
5727 if (sb->s_flags & MS_RDONLY)
5728 return 1;
e2b911c5 5729 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5730 return 0;
5731 return 1;
5732}
24b58424 5733
03010a33
TT		 5734static struct file_system_type ext4_fs_type = {
5735 .owner = THIS_MODULE,
5736 .name = "ext4",
152a0836 5737 .mount = ext4_mount,
03010a33
TT		 5738 .kill_sb = kill_block_super,
5739 .fs_flags = FS_REQUIRES_DEV,
5740};
7f78e035 5741MODULE_ALIAS_FS("ext4");
03010a33 5742
e9e3bcec
ES		 5743/* Shared across all ext4 file systems */
5744wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 5745
5dabfc78 5746static int __init ext4_init_fs(void)
ac27a0ec 5747{
e9e3bcec 5748 int i, err;
c9de560d 5749
e294a537 5750 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5751 ext4_li_info = NULL;
5752 mutex_init(&ext4_li_mtx);
5753
9a4c8019 5754 /* Build-time check for flags consistency */
12e9b892 5755 ext4_check_flag_values();
e9e3bcec 5756
e142d052 5757 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 5758 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 5759
51865fda 5760 err = ext4_init_es();
6fd058f7
TT		 5761 if (err)
5762 return err;
51865fda
ZL		 5763
5764 err = ext4_init_pageio();
5765 if (err)
b5799018 5766 goto out5;
51865fda 5767
5dabfc78 5768 err = ext4_init_system_zone();
bd2d0210 5769 if (err)
b5799018 5770 goto out4;
857ac889 5771
b5799018 5772 err = ext4_init_sysfs();
dd68314c 5773 if (err)
b5799018 5774 goto out3;
857ac889 5775
5dabfc78 5776 err = ext4_init_mballoc();
c9de560d
AT		 5777 if (err)
5778 goto out2;
ac27a0ec
DK		 5779 err = init_inodecache();
5780 if (err)
5781 goto out1;
24b58424 5782 register_as_ext3();
2035e776 5783 register_as_ext2();
03010a33 5784 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5785 if (err)
5786 goto out;
bfff6873 5787
ac27a0ec
DK		 5788 return 0;
5789out:
24b58424
TT		 5790 unregister_as_ext2();
5791 unregister_as_ext3();
ac27a0ec
DK		 5792 destroy_inodecache();
5793out1:
5dabfc78 5794 ext4_exit_mballoc();
9c191f70 5795out2:
b5799018
TT		 5796 ext4_exit_sysfs();
5797out3:
5dabfc78 5798 ext4_exit_system_zone();
b5799018 5799out4:
5dabfc78 5800 ext4_exit_pageio();
b5799018 5801out5:
51865fda
ZL		 5802 ext4_exit_es();
5803
ac27a0ec
DK		 5804 return err;
5805}
5806
5dabfc78 5807static void __exit ext4_exit_fs(void)
ac27a0ec 5808{
bfff6873 5809 ext4_destroy_lazyinit_thread();
24b58424
TT		 5810 unregister_as_ext2();
5811 unregister_as_ext3();
03010a33 5812 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5813 destroy_inodecache();
5dabfc78 5814 ext4_exit_mballoc();
b5799018 5815 ext4_exit_sysfs();
5dabfc78
TT		 5816 ext4_exit_system_zone();
5817 ext4_exit_pageio();
dd12ed14 5818 ext4_exit_es();
ac27a0ec
DK		 5819}
5820
5821MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5822MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5823MODULE_LICENSE("GPL");
5dabfc78
TT		 5824module_init(ext4_init_fs)
5825module_exit(ext4_exit_fs)
ubuntu-bionic-kernel mirror