]> git.ipfire.org Git - thirdparty/kernel/linux.git/blame - fs/ext4/super.c
projects / thirdparty / kernel / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: prevent data corruption with journaling + DAX
[thirdparty/kernel/linux.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
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{
bc98a42c 408 if (sb_rdonly(sb))
ac27a0ec
DK		 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. */
bc98a42c 590 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
ac27a0ec
DK		 591 return;
592
efbed4dc
TT		 593 if (ext4_error_ratelimit(sb)) {
594 errstr = ext4_decode_error(sb, errno, nbuf);
595 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
596 sb->s_id, function, line, errstr);
597 }
ac27a0ec 598
efbed4dc 599 save_error_info(sb, function, line);
617ba13b 600 ext4_handle_error(sb);
ac27a0ec
DK		 601}
602
603/*
617ba13b 604 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 605 * abort function may be used to deal with unrecoverable failures such
606 * as journal IO errors or ENOMEM at a critical moment in log management.
607 *
608 * We unconditionally force the filesystem into an ABORT|READONLY state,
609 * unless the error response on the fs has been set to panic in which
610 * case we take the easy way out and panic immediately.
611 */
612
c67d859e 613void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 614 unsigned int line, const char *fmt, ...)
ac27a0ec 615{
651e1c3b 616 struct va_format vaf;
ac27a0ec
DK		 617 va_list args;
618
0db1ff22
TT		 619 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
620 return;
621
1c13d5c0 622 save_error_info(sb, function, line);
ac27a0ec 623 va_start(args, fmt);
651e1c3b
JP		 624 vaf.fmt = fmt;
625 vaf.va = &args;
626 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
627 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 628 va_end(args);
629
bc98a42c 630 if (sb_rdonly(sb) == 0) {
1c13d5c0 631 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 632 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 633 /*
634 * Make sure updated value of ->s_mount_flags will be visible
635 * before ->s_flags update
636 */
637 smp_wmb();
638 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 639 if (EXT4_SB(sb)->s_journal)
640 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
641 save_error_info(sb, function, line);
642 }
4327ba52
DJ		 643 if (test_opt(sb, ERRORS_PANIC)) {
644 if (EXT4_SB(sb)->s_journal &&
645 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
646 return;
617ba13b 647 panic("EXT4-fs panic from previous error\n");
4327ba52 648 }
ac27a0ec
DK		 649}
650
e7c96e8e
JP		 651void __ext4_msg(struct super_block *sb,
652 const char *prefix, const char *fmt, ...)
b31e1552 653{
0ff2ea7d 654 struct va_format vaf;
b31e1552
ES		 655 va_list args;
656
efbed4dc
TT		 657 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
658 return;
659
b31e1552 660 va_start(args, fmt);
0ff2ea7d
JP		 661 vaf.fmt = fmt;
662 vaf.va = &args;
663 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 664 va_end(args);
665}
666
b03a2f7e
AD		 667#define ext4_warning_ratelimit(sb) \
668 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
669 "EXT4-fs warning")
670
12062ddd 671void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 672 unsigned int line, const char *fmt, ...)
ac27a0ec 673{
0ff2ea7d 674 struct va_format vaf;
ac27a0ec
DK		 675 va_list args;
676
b03a2f7e 677 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 678 return;
679
ac27a0ec 680 va_start(args, fmt);
0ff2ea7d
JP		 681 vaf.fmt = fmt;
682 vaf.va = &args;
683 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
684 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 685 va_end(args);
686}
687
b03a2f7e
AD		 688void __ext4_warning_inode(const struct inode *inode, const char *function,
689 unsigned int line, const char *fmt, ...)
690{
691 struct va_format vaf;
692 va_list args;
693
694 if (!ext4_warning_ratelimit(inode->i_sb))
695 return;
696
697 va_start(args, fmt);
698 vaf.fmt = fmt;
699 vaf.va = &args;
700 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
701 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
702 function, line, inode->i_ino, current->comm, &vaf);
703 va_end(args);
704}
705
e29136f8
TT		 706void __ext4_grp_locked_error(const char *function, unsigned int line,
707 struct super_block *sb, ext4_group_t grp,
708 unsigned long ino, ext4_fsblk_t block,
709 const char *fmt, ...)
5d1b1b3f
AK		 710__releases(bitlock)
711__acquires(bitlock)
712{
0ff2ea7d 713 struct va_format vaf;
5d1b1b3f
AK		 714 va_list args;
715 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
716
0db1ff22
TT		 717 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
718 return;
719
1c13d5c0
TT		 720 es->s_last_error_ino = cpu_to_le32(ino);
721 es->s_last_error_block = cpu_to_le64(block);
722 __save_error_info(sb, function, line);
0ff2ea7d 723
efbed4dc
TT		 724 if (ext4_error_ratelimit(sb)) {
725 va_start(args, fmt);
726 vaf.fmt = fmt;
727 vaf.va = &args;
728 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
729 sb->s_id, function, line, grp);
730 if (ino)
731 printk(KERN_CONT "inode %lu: ", ino);
732 if (block)
733 printk(KERN_CONT "block %llu:",
734 (unsigned long long) block);
735 printk(KERN_CONT "%pV\n", &vaf);
736 va_end(args);
737 }
5d1b1b3f
AK		 738
739 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 740 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 741 return;
742 }
1c13d5c0 743
5d1b1b3f
AK		 744 ext4_unlock_group(sb, grp);
745 ext4_handle_error(sb);
746 /*
747 * We only get here in the ERRORS_RO case; relocking the group
748 * may be dangerous, but nothing bad will happen since the
749 * filesystem will have already been marked read/only and the
750 * journal has been aborted. We return 1 as a hint to callers
751 * who might what to use the return value from
25985edc 752 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 753 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
754 * aggressively from the ext4 function in question, with a
755 * more appropriate error code.
756 */
757 ext4_lock_group(sb, grp);
758 return;
759}
760
617ba13b 761void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 762{
617ba13b 763 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 764
617ba13b 765 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 766 return;
767
12062ddd 768 ext4_warning(sb,
ac27a0ec
DK		 769 "updating to rev %d because of new feature flag, "
770 "running e2fsck is recommended",
617ba13b 771 EXT4_DYNAMIC_REV);
ac27a0ec 772
617ba13b
MC		 773 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
774 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
775 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 776 /* leave es->s_feature_*compat flags alone */
777 /* es->s_uuid will be set by e2fsck if empty */
778
779 /*
780 * The rest of the superblock fields should be zero, and if not it
781 * means they are likely already in use, so leave them alone. We
782 * can leave it up to e2fsck to clean up any inconsistencies there.
783 */
784}
785
786/*
787 * Open the external journal device
788 */
b31e1552 789static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 790{
791 struct block_device *bdev;
792 char b[BDEVNAME_SIZE];
793
d4d77629 794 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 795 if (IS_ERR(bdev))
796 goto fail;
797 return bdev;
798
799fail:
b31e1552 800 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 801 __bdevname(dev, b), PTR_ERR(bdev));
802 return NULL;
803}
804
805/*
806 * Release the journal device
807 */
4385bab1 808static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 809{
4385bab1 810 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 811}
812
4385bab1 813static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 814{
815 struct block_device *bdev;
ac27a0ec
DK		 816 bdev = sbi->journal_bdev;
817 if (bdev) {
4385bab1 818 ext4_blkdev_put(bdev);
ac27a0ec
DK		 819 sbi->journal_bdev = NULL;
820 }
ac27a0ec
DK		 821}
822
823static inline struct inode *orphan_list_entry(struct list_head *l)
824{
617ba13b 825 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 826}
827
617ba13b 828static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 829{
830 struct list_head *l;
831
b31e1552
ES		 832 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
833 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 834
835 printk(KERN_ERR "sb_info orphan list:\n");
836 list_for_each(l, &sbi->s_orphan) {
837 struct inode *inode = orphan_list_entry(l);
838 printk(KERN_ERR " "
839 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
840 inode->i_sb->s_id, inode->i_ino, inode,
841 inode->i_mode, inode->i_nlink,
842 NEXT_ORPHAN(inode));
843 }
844}
845
957153fc
JK		 846#ifdef CONFIG_QUOTA
847static int ext4_quota_off(struct super_block *sb, int type);
848
849static inline void ext4_quota_off_umount(struct super_block *sb)
850{
851 int type;
852
964edf66
JK		 853 /* Use our quota_off function to clear inode flags etc. */
854 for (type = 0; type < EXT4_MAXQUOTAS; type++)
855 ext4_quota_off(sb, type);
957153fc
JK		 856}
857#else
858static inline void ext4_quota_off_umount(struct super_block *sb)
859{
860}
861#endif
862
2b2d6d01 863static void ext4_put_super(struct super_block *sb)
ac27a0ec 864{
617ba13b
MC		 865 struct ext4_sb_info *sbi = EXT4_SB(sb);
866 struct ext4_super_block *es = sbi->s_es;
97abd7d4 867 int aborted = 0;
ef2cabf7 868 int i, err;
ac27a0ec 869
857ac889 870 ext4_unregister_li_request(sb);
957153fc 871 ext4_quota_off_umount(sb);
e0ccfd95 872
2e8fa54e 873 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 874 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 875
0390131b 876 if (sbi->s_journal) {
97abd7d4 877 aborted = is_journal_aborted(sbi->s_journal);
0390131b
FM		 878 err = jbd2_journal_destroy(sbi->s_journal);
879 sbi->s_journal = NULL;
97abd7d4 880 if ((err < 0) && !aborted)
c67d859e 881 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 882 }
d4edac31 883
ebd173be 884 ext4_unregister_sysfs(sb);
d3922a77 885 ext4_es_unregister_shrinker(sbi);
9105bb14 886 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 887 ext4_release_system_zone(sb);
888 ext4_mb_release(sb);
889 ext4_ext_release(sb);
d4edac31 890
bc98a42c 891 if (!sb_rdonly(sb) && !aborted) {
e2b911c5 892 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 893 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 894 }
bc98a42c 895 if (!sb_rdonly(sb))
a8e25a83
AB		 896 ext4_commit_super(sb, 1);
897
ac27a0ec
DK		 898 for (i = 0; i < sbi->s_gdb_count; i++)
899 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 900 kvfree(sbi->s_group_desc);
901 kvfree(sbi->s_flex_groups);
57042651 902 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 903 percpu_counter_destroy(&sbi->s_freeinodes_counter);
904 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 905 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c8585c6f 906 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
ac27a0ec 907#ifdef CONFIG_QUOTA
a2d4a646 908 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 909 kfree(sbi->s_qf_names[i]);
910#endif
911
912 /* Debugging code just in case the in-memory inode orphan list
913 * isn't empty. The on-disk one can be non-empty if we've
914 * detected an error and taken the fs readonly, but the
915 * in-memory list had better be clean by this point. */
916 if (!list_empty(&sbi->s_orphan))
917 dump_orphan_list(sb, sbi);
918 J_ASSERT(list_empty(&sbi->s_orphan));
919
89d96a6f 920 sync_blockdev(sb->s_bdev);
f98393a6 921 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 922 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
923 /*
924 * Invalidate the journal device's buffers. We don't want them
925 * floating about in memory - the physical journal device may
926 * hotswapped, and it breaks the `ro-after' testing code.
927 */
928 sync_blockdev(sbi->journal_bdev);
f98393a6 929 invalidate_bdev(sbi->journal_bdev);
617ba13b 930 ext4_blkdev_remove(sbi);
ac27a0ec 931 }
dec214d0
TE		 932 if (sbi->s_ea_inode_cache) {
933 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
934 sbi->s_ea_inode_cache = NULL;
935 }
47387409
TE		 936 if (sbi->s_ea_block_cache) {
937 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
938 sbi->s_ea_block_cache = NULL;
9c191f70 939 }
c5e06d10
JL		 940 if (sbi->s_mmp_tsk)
941 kthread_stop(sbi->s_mmp_tsk);
9060dd2c 942 brelse(sbi->s_sbh);
ac27a0ec 943 sb->s_fs_info = NULL;
3197ebdb
TT		 944 /*
945 * Now that we are completely done shutting down the
946 * superblock, we need to actually destroy the kobject.
947 */
3197ebdb
TT		 948 kobject_put(&sbi->s_kobj);
949 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 950 if (sbi->s_chksum_driver)
951 crypto_free_shash(sbi->s_chksum_driver);
705895b6 952 kfree(sbi->s_blockgroup_lock);
5e405595 953 fs_put_dax(sbi->s_daxdev);
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 980 ei->i_reserved_data_blocks = 0;
9d0be502 981 ei->i_da_metadata_calc_len = 0;
7e731bc9 982 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 983 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 984#ifdef CONFIG_QUOTA
985 ei->i_reserved_quota = 0;
96c7e0d9 986 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 987#endif
8aefcd55 988 ei->jinode = NULL;
2e8fa54e 989 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 990 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 991 ei->i_sync_tid = 0;
992 ei->i_datasync_tid = 0;
e27f41e1 993 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 994 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
ac27a0ec
DK		 995 return &ei->vfs_inode;
996}
997
7ff9c073
TT		 998static int ext4_drop_inode(struct inode *inode)
999{
1000 int drop = generic_drop_inode(inode);
1001
1002 trace_ext4_drop_inode(inode, drop);
1003 return drop;
1004}
1005
fa0d7e3d
NP		 1006static void ext4_i_callback(struct rcu_head *head)
1007{
1008 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 1009 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1010}
1011
617ba13b 1012static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1013{
9f7dd93d 1014 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1015 ext4_msg(inode->i_sb, KERN_ERR,
1016 "Inode %lu (%p): orphan list check failed!",
1017 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1018 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1019 EXT4_I(inode), sizeof(struct ext4_inode_info),
1020 true);
1021 dump_stack();
1022 }
fa0d7e3d 1023 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 1024}
1025
51cc5068 1026static void init_once(void *foo)
ac27a0ec 1027{
617ba13b 1028 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1029
a35afb83 1030 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1031 init_rwsem(&ei->xattr_sem);
0e855ac8 1032 init_rwsem(&ei->i_data_sem);
ea3d7209 1033 init_rwsem(&ei->i_mmap_sem);
a35afb83 1034 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1035}
1036
e67bc2b3 1037static int __init init_inodecache(void)
ac27a0ec 1038{
617ba13b
MC		 1039 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1040 sizeof(struct ext4_inode_info),
ac27a0ec 1041 0, (SLAB_RECLAIM_ACCOUNT|
5d097056 1042 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
20c2df83 1043 init_once);
617ba13b 1044 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1045 return -ENOMEM;
1046 return 0;
1047}
1048
1049static void destroy_inodecache(void)
1050{
8c0a8537
KS		 1051 /*
1052 * Make sure all delayed rcu free inodes are flushed before we
1053 * destroy cache.
1054 */
1055 rcu_barrier();
617ba13b 1056 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1057}
1058
0930fcc1 1059void ext4_clear_inode(struct inode *inode)
ac27a0ec 1060{
0930fcc1 1061 invalidate_inode_buffers(inode);
dbd5768f 1062 clear_inode(inode);
9f754758 1063 dquot_drop(inode);
c2ea3fde 1064 ext4_discard_preallocations(inode);
51865fda 1065 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1066 if (EXT4_I(inode)->jinode) {
1067 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1068 EXT4_I(inode)->jinode);
1069 jbd2_free_inode(EXT4_I(inode)->jinode);
1070 EXT4_I(inode)->jinode = NULL;
1071 }
b7236e21 1072#ifdef CONFIG_EXT4_FS_ENCRYPTION
a7550b30 1073 fscrypt_put_encryption_info(inode, NULL);
b7236e21 1074#endif
ac27a0ec
DK		 1075}
1076
1b961ac0 1077static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1078 u64 ino, u32 generation)
ac27a0ec 1079{
ac27a0ec 1080 struct inode *inode;
ac27a0ec 1081
617ba13b 1082 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1083 return ERR_PTR(-ESTALE);
617ba13b 1084 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1085 return ERR_PTR(-ESTALE);
1086
1087 /* iget isn't really right if the inode is currently unallocated!!
1088 *
617ba13b 1089 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1090 * deleted, so we should be safe.
1091 *
1092 * Currently we don't know the generation for parent directory, so
1093 * a generation of 0 means "accept any"
1094 */
f4bb2981 1095 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1096 if (IS_ERR(inode))
1097 return ERR_CAST(inode);
1098 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1099 iput(inode);
1100 return ERR_PTR(-ESTALE);
1101 }
1b961ac0
CH		 1102
1103 return inode;
1104}
1105
1106static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1107 int fh_len, int fh_type)
1b961ac0
CH		 1108{
1109 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1110 ext4_nfs_get_inode);
1111}
1112
1113static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1114 int fh_len, int fh_type)
1b961ac0
CH		 1115{
1116 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1117 ext4_nfs_get_inode);
ac27a0ec
DK		 1118}
1119
c39a7f84
TO		 1120/*
1121 * Try to release metadata pages (indirect blocks, directories) which are
1122 * mapped via the block device. Since these pages could have journal heads
1123 * which would prevent try_to_free_buffers() from freeing them, we must use
1124 * jbd2 layer's try_to_free_buffers() function to release them.
1125 */
0b8e58a1
AD		 1126static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1127 gfp_t wait)
c39a7f84
TO		 1128{
1129 journal_t *journal = EXT4_SB(sb)->s_journal;
1130
1131 WARN_ON(PageChecked(page));
1132 if (!page_has_buffers(page))
1133 return 0;
1134 if (journal)
1135 return jbd2_journal_try_to_free_buffers(journal, page,
d0164adc 1136 wait & ~__GFP_DIRECT_RECLAIM);
c39a7f84
TO		 1137 return try_to_free_buffers(page);
1138}
1139
a7550b30
JK		 1140#ifdef CONFIG_EXT4_FS_ENCRYPTION
1141static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1142{
1143 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1144 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1145}
1146
a7550b30
JK		 1147static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1148 void *fs_data)
1149{
2f8f5e76 1150 handle_t *handle = fs_data;
c1a5d5f6 1151 int res, res2, credits, retries = 0;
2f8f5e76 1152
9ce0151a
EB		 1153 /*
1154 * Encrypting the root directory is not allowed because e2fsck expects
1155 * lost+found to exist and be unencrypted, and encrypting the root
1156 * directory would imply encrypting the lost+found directory as well as
1157 * the filename "lost+found" itself.
1158 */
1159 if (inode->i_ino == EXT4_ROOT_INO)
1160 return -EPERM;
2f8f5e76 1161
94840e3c
EB		 1162 res = ext4_convert_inline_data(inode);
1163 if (res)
1164 return res;
1165
2f8f5e76
EB		 1166 /*
1167 * If a journal handle was specified, then the encryption context is
1168 * being set on a new inode via inheritance and is part of a larger
1169 * transaction to create the inode. Otherwise the encryption context is
1170 * being set on an existing inode in its own transaction. Only in the
1171 * latter case should the "retry on ENOSPC" logic be used.
1172 */
a7550b30 1173
2f8f5e76
EB		 1174 if (handle) {
1175 res = ext4_xattr_set_handle(handle, inode,
1176 EXT4_XATTR_INDEX_ENCRYPTION,
1177 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1178 ctx, len, 0);
a7550b30
JK		 1179 if (!res) {
1180 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1181 ext4_clear_inode_state(inode,
1182 EXT4_STATE_MAY_INLINE_DATA);
a3caa24b
JK		 1183 /*
1184 * Update inode->i_flags - e.g. S_DAX may get disabled
1185 */
1186 ext4_set_inode_flags(inode);
a7550b30
JK		 1187 }
1188 return res;
1189 }
1190
b8cb5a54
TE		 1191 res = dquot_initialize(inode);
1192 if (res)
1193 return res;
2f8f5e76 1194retry:
af65207c
TE		 1195 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1196 &credits);
dec214d0
TE		 1197 if (res)
1198 return res;
1199
c1a5d5f6 1200 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
a7550b30
JK		 1201 if (IS_ERR(handle))
1202 return PTR_ERR(handle);
1203
2f8f5e76
EB		 1204 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1205 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1206 ctx, len, 0);
a7550b30
JK		 1207 if (!res) {
1208 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
a3caa24b
JK		 1209 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1210 ext4_set_inode_flags(inode);
a7550b30
JK		 1211 res = ext4_mark_inode_dirty(handle, inode);
1212 if (res)
1213 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1214 }
1215 res2 = ext4_journal_stop(handle);
2f8f5e76
EB		 1216
1217 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1218 goto retry;
a7550b30
JK		 1219 if (!res)
1220 res = res2;
1221 return res;
1222}
1223
c250b7dd 1224static bool ext4_dummy_context(struct inode *inode)
a7550b30
JK		 1225{
1226 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1227}
1228
1229static unsigned ext4_max_namelen(struct inode *inode)
1230{
1231 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1232 EXT4_NAME_LEN;
1233}
1234
6f69f0ed 1235static const struct fscrypt_operations ext4_cryptops = {
a5d431ef 1236 .key_prefix = "ext4:",
a7550b30 1237 .get_context = ext4_get_context,
a7550b30
JK		 1238 .set_context = ext4_set_context,
1239 .dummy_context = ext4_dummy_context,
1240 .is_encrypted = ext4_encrypted_inode,
1241 .empty_dir = ext4_empty_dir,
1242 .max_namelen = ext4_max_namelen,
1243};
1244#else
6f69f0ed 1245static const struct fscrypt_operations ext4_cryptops = {
a7550b30
JK		 1246 .is_encrypted = ext4_encrypted_inode,
1247};
1248#endif
1249
ac27a0ec 1250#ifdef CONFIG_QUOTA
d6006186 1251static const char * const quotatypes[] = INITQFNAMES;
689c958c 1252#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1253
617ba13b
MC		 1254static int ext4_write_dquot(struct dquot *dquot);
1255static int ext4_acquire_dquot(struct dquot *dquot);
1256static int ext4_release_dquot(struct dquot *dquot);
1257static int ext4_mark_dquot_dirty(struct dquot *dquot);
1258static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1259static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1260 const struct path *path);
617ba13b
MC		 1261static int ext4_quota_on_mount(struct super_block *sb, int type);
1262static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1263 size_t len, loff_t off);
617ba13b 1264static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1265 const char *data, size_t len, loff_t off);
7c319d32
AK		 1266static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1267 unsigned int flags);
1268static int ext4_enable_quotas(struct super_block *sb);
8f0e8746 1269static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
ac27a0ec 1270
96c7e0d9
JK		 1271static struct dquot **ext4_get_dquots(struct inode *inode)
1272{
1273 return EXT4_I(inode)->i_dquot;
1274}
1275
61e225dc 1276static const struct dquot_operations ext4_quota_operations = {
7a9ca53a
TE		 1277 .get_reserved_space = ext4_get_reserved_space,
1278 .write_dquot = ext4_write_dquot,
1279 .acquire_dquot = ext4_acquire_dquot,
1280 .release_dquot = ext4_release_dquot,
1281 .mark_dirty = ext4_mark_dquot_dirty,
1282 .write_info = ext4_write_info,
1283 .alloc_dquot = dquot_alloc,
1284 .destroy_dquot = dquot_destroy,
1285 .get_projid = ext4_get_projid,
1286 .get_inode_usage = ext4_get_inode_usage,
1287 .get_next_id = ext4_get_next_id,
ac27a0ec
DK		 1288};
1289
0d54b217 1290static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1291 .quota_on = ext4_quota_on,
ca0e05e4 1292 .quota_off = ext4_quota_off,
287a8095 1293 .quota_sync = dquot_quota_sync,
0a240339 1294 .get_state = dquot_get_state,
287a8095
CH		 1295 .set_info = dquot_set_dqinfo,
1296 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1297 .set_dqblk = dquot_set_dqblk,
1298 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1299};
1300#endif
1301
ee9b6d61 1302static const struct super_operations ext4_sops = {
617ba13b
MC		 1303 .alloc_inode = ext4_alloc_inode,
1304 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1305 .write_inode = ext4_write_inode,
1306 .dirty_inode = ext4_dirty_inode,
7ff9c073 1307 .drop_inode = ext4_drop_inode,
0930fcc1 1308 .evict_inode = ext4_evict_inode,
617ba13b 1309 .put_super = ext4_put_super,
617ba13b 1310 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1311 .freeze_fs = ext4_freeze,
1312 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1313 .statfs = ext4_statfs,
1314 .remount_fs = ext4_remount,
617ba13b 1315 .show_options = ext4_show_options,
ac27a0ec 1316#ifdef CONFIG_QUOTA
617ba13b
MC		 1317 .quota_read = ext4_quota_read,
1318 .quota_write = ext4_quota_write,
96c7e0d9 1319 .get_dquots = ext4_get_dquots,
ac27a0ec 1320#endif
c39a7f84 1321 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1322};
1323
39655164 1324static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1325 .fh_to_dentry = ext4_fh_to_dentry,
1326 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1327 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1328};
1329
1330enum {
1331 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1332 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1333 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1334 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1335 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1336 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1337 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1338 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1339 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1340 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1341 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1342 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
49da9392 1343 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1449032b 1344 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
670e9875 1345 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1449032b 1346 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1347 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1348 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1349 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1350 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
ac27a0ec
DK		 1351};
1352
a447c093 1353static const match_table_t tokens = {
ac27a0ec
DK		 1354 {Opt_bsd_df, "bsddf"},
1355 {Opt_minix_df, "minixdf"},
1356 {Opt_grpid, "grpid"},
1357 {Opt_grpid, "bsdgroups"},
1358 {Opt_nogrpid, "nogrpid"},
1359 {Opt_nogrpid, "sysvgroups"},
1360 {Opt_resgid, "resgid=%u"},
1361 {Opt_resuid, "resuid=%u"},
1362 {Opt_sb, "sb=%u"},
1363 {Opt_err_cont, "errors=continue"},
1364 {Opt_err_panic, "errors=panic"},
1365 {Opt_err_ro, "errors=remount-ro"},
1366 {Opt_nouid32, "nouid32"},
ac27a0ec 1367 {Opt_debug, "debug"},
72578c33
TT		 1368 {Opt_removed, "oldalloc"},
1369 {Opt_removed, "orlov"},
ac27a0ec
DK		 1370 {Opt_user_xattr, "user_xattr"},
1371 {Opt_nouser_xattr, "nouser_xattr"},
1372 {Opt_acl, "acl"},
1373 {Opt_noacl, "noacl"},
e3bb52ae 1374 {Opt_noload, "norecovery"},
5a916be1 1375 {Opt_noload, "noload"},
72578c33
TT		 1376 {Opt_removed, "nobh"},
1377 {Opt_removed, "bh"},
ac27a0ec 1378 {Opt_commit, "commit=%u"},
30773840
TT		 1379 {Opt_min_batch_time, "min_batch_time=%u"},
1380 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1381 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1382 {Opt_journal_path, "journal_path=%s"},
818d276c 1383 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1384 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1385 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1386 {Opt_abort, "abort"},
1387 {Opt_data_journal, "data=journal"},
1388 {Opt_data_ordered, "data=ordered"},
1389 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1390 {Opt_data_err_abort, "data_err=abort"},
1391 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1392 {Opt_offusrjquota, "usrjquota="},
1393 {Opt_usrjquota, "usrjquota=%s"},
1394 {Opt_offgrpjquota, "grpjquota="},
1395 {Opt_grpjquota, "grpjquota=%s"},
1396 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1397 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1398 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1399 {Opt_grpquota, "grpquota"},
1400 {Opt_noquota, "noquota"},
1401 {Opt_quota, "quota"},
1402 {Opt_usrquota, "usrquota"},
49da9392 1403 {Opt_prjquota, "prjquota"},
ac27a0ec 1404 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1405 {Opt_barrier, "barrier"},
1406 {Opt_nobarrier, "nobarrier"},
25ec56b5 1407 {Opt_i_version, "i_version"},
923ae0ff 1408 {Opt_dax, "dax"},
c9de560d 1409 {Opt_stripe, "stripe=%u"},
64769240 1410 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1411 {Opt_lazytime, "lazytime"},
1412 {Opt_nolazytime, "nolazytime"},
670e9875 1413 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
dd919b98 1414 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1415 {Opt_removed, "mblk_io_submit"},
1416 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1417 {Opt_block_validity, "block_validity"},
1418 {Opt_noblock_validity, "noblock_validity"},
240799cd 1419 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1420 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1421 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1422 {Opt_auto_da_alloc, "auto_da_alloc"},
1423 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1424 {Opt_dioread_nolock, "dioread_nolock"},
1425 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1426 {Opt_discard, "discard"},
1427 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1428 {Opt_init_itable, "init_itable=%u"},
1429 {Opt_init_itable, "init_itable"},
1430 {Opt_noinit_itable, "noinit_itable"},
df981d03 1431 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1432 {Opt_test_dummy_encryption, "test_dummy_encryption"},
cdb7ee4c
TE		 1433 {Opt_nombcache, "nombcache"},
1434 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
c7198b9c
TT		 1435 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1436 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1437 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1438 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1439 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1440 {Opt_err, NULL},
ac27a0ec
DK		 1441};
1442
617ba13b 1443static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1444{
617ba13b 1445 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1446 char *options = (char *) *data;
1447
1448 if (!options || strncmp(options, "sb=", 3) != 0)
1449 return 1; /* Default location */
0b8e58a1 1450
ac27a0ec 1451 options += 3;
0b8e58a1 1452 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1453 sb_block = simple_strtoul(options, &options, 0);
1454 if (*options && *options != ',') {
4776004f 1455 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1456 (char *) *data);
1457 return 1;
1458 }
1459 if (*options == ',')
1460 options++;
1461 *data = (void *) options;
0b8e58a1 1462
ac27a0ec
DK		 1463 return sb_block;
1464}
1465
b3881f74 1466#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
d6006186
EB		 1467static const char deprecated_msg[] =
1468 "Mount option \"%s\" will be removed by %s\n"
437ca0fd 1469 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1470
56c50f11
DM		 1471#ifdef CONFIG_QUOTA
1472static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1473{
1474 struct ext4_sb_info *sbi = EXT4_SB(sb);
1475 char *qname;
03dafb5f 1476 int ret = -1;
56c50f11
DM		 1477
1478 if (sb_any_quota_loaded(sb) &&
1479 !sbi->s_qf_names[qtype]) {
1480 ext4_msg(sb, KERN_ERR,
1481 "Cannot change journaled "
1482 "quota options when quota turned on");
57f73c2c 1483 return -1;
56c50f11 1484 }
e2b911c5 1485 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1486 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1487 "ignored when QUOTA feature is enabled");
1488 return 1;
262b4662 1489 }
56c50f11
DM		 1490 qname = match_strdup(args);
1491 if (!qname) {
1492 ext4_msg(sb, KERN_ERR,
1493 "Not enough memory for storing quotafile name");
57f73c2c 1494 return -1;
56c50f11 1495 }
03dafb5f
CG		 1496 if (sbi->s_qf_names[qtype]) {
1497 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1498 ret = 1;
1499 else
1500 ext4_msg(sb, KERN_ERR,
1501 "%s quota file already specified",
1502 QTYPE2NAME(qtype));
1503 goto errout;
56c50f11 1504 }
03dafb5f 1505 if (strchr(qname, '/')) {
56c50f11
DM		 1506 ext4_msg(sb, KERN_ERR,
1507 "quotafile must be on filesystem root");
03dafb5f 1508 goto errout;
56c50f11 1509 }
03dafb5f 1510 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1511 set_opt(sb, QUOTA);
56c50f11 1512 return 1;
03dafb5f
CG		 1513errout:
1514 kfree(qname);
1515 return ret;
56c50f11
DM		 1516}
1517
1518static int clear_qf_name(struct super_block *sb, int qtype)
1519{
1520
1521 struct ext4_sb_info *sbi = EXT4_SB(sb);
1522
1523 if (sb_any_quota_loaded(sb) &&
1524 sbi->s_qf_names[qtype]) {
1525 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1526 " when quota turned on");
57f73c2c 1527 return -1;
56c50f11 1528 }
03dafb5f 1529 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1530 sbi->s_qf_names[qtype] = NULL;
1531 return 1;
1532}
1533#endif
1534
26092bf5
TT		 1535#define MOPT_SET 0x0001
1536#define MOPT_CLEAR 0x0002
1537#define MOPT_NOSUPPORT 0x0004
1538#define MOPT_EXPLICIT 0x0008
1539#define MOPT_CLEAR_ERR 0x0010
1540#define MOPT_GTE0 0x0020
ac27a0ec 1541#ifdef CONFIG_QUOTA
26092bf5
TT		 1542#define MOPT_Q 0
1543#define MOPT_QFMT 0x0040
1544#else
1545#define MOPT_Q MOPT_NOSUPPORT
1546#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1547#endif
26092bf5 1548#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1549#define MOPT_NO_EXT2 0x0100
1550#define MOPT_NO_EXT3 0x0200
1551#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1552#define MOPT_STRING 0x0400
26092bf5
TT		 1553
1554static const struct mount_opts {
1555 int token;
1556 int mount_opt;
1557 int flags;
1558} ext4_mount_opts[] = {
1559 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1560 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1561 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1562 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1563 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1564 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1565 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1566 MOPT_EXT4_ONLY | MOPT_SET},
1567 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1568 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1569 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1570 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1571 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1572 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1573 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1574 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1575 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1576 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1577 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1578 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1579 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1580 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1581 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1582 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1583 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1584 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1585 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1586 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1587 MOPT_NO_EXT2},
8dc0aa8c 1588 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1589 MOPT_NO_EXT2},
26092bf5
TT		 1590 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1591 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1592 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1593 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1594 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1595 {Opt_commit, 0, MOPT_GTE0},
1596 {Opt_max_batch_time, 0, MOPT_GTE0},
1597 {Opt_min_batch_time, 0, MOPT_GTE0},
1598 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1599 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1600 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1601 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1602 {Opt_resuid, 0, MOPT_GTE0},
1603 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1604 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1605 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1606 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1607 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1608 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1609 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1610 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1611 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1612 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1613#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1614 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1615 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1616#else
26092bf5
TT		 1617 {Opt_acl, 0, MOPT_NOSUPPORT},
1618 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1619#endif
26092bf5
TT		 1620 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1621 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
670e9875 1622 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
26092bf5
TT		 1623 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1624 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1625 MOPT_SET | MOPT_Q},
1626 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1627 MOPT_SET | MOPT_Q},
49da9392
JK		 1628 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1629 MOPT_SET | MOPT_Q},
26092bf5 1630 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1631 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1632 MOPT_CLEAR | MOPT_Q},
26092bf5
TT		 1633 {Opt_usrjquota, 0, MOPT_Q},
1634 {Opt_grpjquota, 0, MOPT_Q},
1635 {Opt_offusrjquota, 0, MOPT_Q},
1636 {Opt_offgrpjquota, 0, MOPT_Q},
1637 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1638 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1639 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1640 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1641 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
cdb7ee4c 1642 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
26092bf5
TT		 1643 {Opt_err, 0, 0}
1644};
1645
1646static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1647 substring_t *args, unsigned long *journal_devnum,
1648 unsigned int *journal_ioprio, int is_remount)
1649{
1650 struct ext4_sb_info *sbi = EXT4_SB(sb);
1651 const struct mount_opts *m;
08cefc7a
EB		 1652 kuid_t uid;
1653 kgid_t gid;
26092bf5
TT		 1654 int arg = 0;
1655
57f73c2c
TT		 1656#ifdef CONFIG_QUOTA
1657 if (token == Opt_usrjquota)
1658 return set_qf_name(sb, USRQUOTA, &args[0]);
1659 else if (token == Opt_grpjquota)
1660 return set_qf_name(sb, GRPQUOTA, &args[0]);
1661 else if (token == Opt_offusrjquota)
1662 return clear_qf_name(sb, USRQUOTA);
1663 else if (token == Opt_offgrpjquota)
1664 return clear_qf_name(sb, GRPQUOTA);
1665#endif
26092bf5 1666 switch (token) {
f7048605
TT		 1667 case Opt_noacl:
1668 case Opt_nouser_xattr:
1669 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1670 break;
26092bf5
TT		 1671 case Opt_sb:
1672 return 1; /* handled by get_sb_block() */
1673 case Opt_removed:
5f3633e3 1674 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1675 return 1;
26092bf5
TT		 1676 case Opt_abort:
1677 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1678 return 1;
1679 case Opt_i_version:
1680 sb->s_flags |= MS_I_VERSION;
1681 return 1;
a26f4992
TT		 1682 case Opt_lazytime:
1683 sb->s_flags |= MS_LAZYTIME;
1684 return 1;
1685 case Opt_nolazytime:
1686 sb->s_flags &= ~MS_LAZYTIME;
1687 return 1;
26092bf5
TT		 1688 }
1689
5f3633e3
JK		 1690 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1691 if (token == m->token)
1692 break;
1693
1694 if (m->token == Opt_err) {
1695 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1696 "or missing value", opt);
1697 return -1;
1698 }
1699
8dc0aa8c
TT		 1700 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1701 ext4_msg(sb, KERN_ERR,
1702 "Mount option \"%s\" incompatible with ext2", opt);
1703 return -1;
1704 }
1705 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1706 ext4_msg(sb, KERN_ERR,
1707 "Mount option \"%s\" incompatible with ext3", opt);
1708 return -1;
1709 }
1710
ad4eec61 1711 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1712 return -1;
1713 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1714 return -1;
c93cf2d7
DM		 1715 if (m->flags & MOPT_EXPLICIT) {
1716 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1717 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1718 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1719 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1720 } else
1721 return -1;
1722 }
5f3633e3
JK		 1723 if (m->flags & MOPT_CLEAR_ERR)
1724 clear_opt(sb, ERRORS_MASK);
1725 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1726 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1727 "options when quota turned on");
1728 return -1;
1729 }
1730
1731 if (m->flags & MOPT_NOSUPPORT) {
1732 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1733 } else if (token == Opt_commit) {
1734 if (arg == 0)
1735 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1736 sbi->s_commit_interval = HZ * arg;
670e9875
TT		 1737 } else if (token == Opt_debug_want_extra_isize) {
1738 sbi->s_want_extra_isize = arg;
5f3633e3 1739 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1740 sbi->s_max_batch_time = arg;
1741 } else if (token == Opt_min_batch_time) {
1742 sbi->s_min_batch_time = arg;
1743 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1744 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1745 ext4_msg(sb, KERN_ERR,
1746 "EXT4-fs: inode_readahead_blks must be "
1747 "0 or a power of 2 smaller than 2^31");
26092bf5 1748 return -1;
5f3633e3
JK		 1749 }
1750 sbi->s_inode_readahead_blks = arg;
1751 } else if (token == Opt_init_itable) {
1752 set_opt(sb, INIT_INODE_TABLE);
1753 if (!args->from)
1754 arg = EXT4_DEF_LI_WAIT_MULT;
1755 sbi->s_li_wait_mult = arg;
1756 } else if (token == Opt_max_dir_size_kb) {
1757 sbi->s_max_dir_size_kb = arg;
1758 } else if (token == Opt_stripe) {
1759 sbi->s_stripe = arg;
1760 } else if (token == Opt_resuid) {
1761 uid = make_kuid(current_user_ns(), arg);
1762 if (!uid_valid(uid)) {
1763 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1764 return -1;
1765 }
5f3633e3
JK		 1766 sbi->s_resuid = uid;
1767 } else if (token == Opt_resgid) {
1768 gid = make_kgid(current_user_ns(), arg);
1769 if (!gid_valid(gid)) {
1770 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1771 return -1;
1772 }
1773 sbi->s_resgid = gid;
1774 } else if (token == Opt_journal_dev) {
1775 if (is_remount) {
1776 ext4_msg(sb, KERN_ERR,
1777 "Cannot specify journal on remount");
1778 return -1;
1779 }
1780 *journal_devnum = arg;
ad4eec61
ES		 1781 } else if (token == Opt_journal_path) {
1782 char *journal_path;
1783 struct inode *journal_inode;
1784 struct path path;
1785 int error;
1786
1787 if (is_remount) {
1788 ext4_msg(sb, KERN_ERR,
1789 "Cannot specify journal on remount");
1790 return -1;
1791 }
1792 journal_path = match_strdup(&args[0]);
1793 if (!journal_path) {
1794 ext4_msg(sb, KERN_ERR, "error: could not dup "
1795 "journal device string");
1796 return -1;
1797 }
1798
1799 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1800 if (error) {
1801 ext4_msg(sb, KERN_ERR, "error: could not find "
1802 "journal device path: error %d", error);
1803 kfree(journal_path);
1804 return -1;
1805 }
1806
2b0143b5 1807 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1808 if (!S_ISBLK(journal_inode->i_mode)) {
1809 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1810 "is not a block device", journal_path);
1811 path_put(&path);
1812 kfree(journal_path);
1813 return -1;
1814 }
1815
1816 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1817 path_put(&path);
1818 kfree(journal_path);
5f3633e3
JK		 1819 } else if (token == Opt_journal_ioprio) {
1820 if (arg > 7) {
1821 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1822 " (must be 0-7)");
1823 return -1;
1824 }
1825 *journal_ioprio =
1826 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1827 } else if (token == Opt_test_dummy_encryption) {
1828#ifdef CONFIG_EXT4_FS_ENCRYPTION
1829 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1830 ext4_msg(sb, KERN_WARNING,
1831 "Test dummy encryption mode enabled");
1832#else
1833 ext4_msg(sb, KERN_WARNING,
1834 "Test dummy encryption mount option ignored");
1835#endif
5f3633e3
JK		 1836 } else if (m->flags & MOPT_DATAJ) {
1837 if (is_remount) {
1838 if (!sbi->s_journal)
1839 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1840 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1841 ext4_msg(sb, KERN_ERR,
26092bf5 1842 "Cannot change data mode on remount");
26092bf5 1843 return -1;
ac27a0ec 1844 }
26092bf5 1845 } else {
5f3633e3
JK		 1846 clear_opt(sb, DATA_FLAGS);
1847 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1848 }
5f3633e3
JK		 1849#ifdef CONFIG_QUOTA
1850 } else if (m->flags & MOPT_QFMT) {
1851 if (sb_any_quota_loaded(sb) &&
1852 sbi->s_jquota_fmt != m->mount_opt) {
1853 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1854 "quota options when quota turned on");
1855 return -1;
1856 }
e2b911c5 1857 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1858 ext4_msg(sb, KERN_INFO,
1859 "Quota format mount options ignored "
262b4662 1860 "when QUOTA feature is enabled");
c325a67c 1861 return 1;
262b4662 1862 }
5f3633e3 1863 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 1864#endif
923ae0ff 1865 } else if (token == Opt_dax) {
ef83b6e8
DW		 1866#ifdef CONFIG_FS_DAX
1867 ext4_msg(sb, KERN_WARNING,
1868 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1869 sbi->s_mount_opt |= m->mount_opt;
1870#else
923ae0ff
RZ		 1871 ext4_msg(sb, KERN_INFO, "dax option not supported");
1872 return -1;
5f3633e3 1873#endif
7915a861
AN		 1874 } else if (token == Opt_data_err_abort) {
1875 sbi->s_mount_opt |= m->mount_opt;
1876 } else if (token == Opt_data_err_ignore) {
1877 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 1878 } else {
1879 if (!args->from)
1880 arg = 1;
1881 if (m->flags & MOPT_CLEAR)
1882 arg = !arg;
1883 else if (unlikely(!(m->flags & MOPT_SET))) {
1884 ext4_msg(sb, KERN_WARNING,
1885 "buggy handling of option %s", opt);
1886 WARN_ON(1);
1887 return -1;
1888 }
1889 if (arg != 0)
1890 sbi->s_mount_opt |= m->mount_opt;
1891 else
1892 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1893 }
5f3633e3 1894 return 1;
26092bf5
TT		 1895}
1896
1897static int parse_options(char *options, struct super_block *sb,
1898 unsigned long *journal_devnum,
1899 unsigned int *journal_ioprio,
1900 int is_remount)
1901{
1902 struct ext4_sb_info *sbi = EXT4_SB(sb);
1903 char *p;
1904 substring_t args[MAX_OPT_ARGS];
1905 int token;
1906
1907 if (!options)
1908 return 1;
1909
1910 while ((p = strsep(&options, ",")) != NULL) {
1911 if (!*p)
1912 continue;
1913 /*
1914 * Initialize args struct so we know whether arg was
1915 * found; some options take optional arguments.
1916 */
caecd0af 1917 args[0].to = args[0].from = NULL;
26092bf5
TT		 1918 token = match_token(p, tokens, args);
1919 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1920 journal_ioprio, is_remount) < 0)
1921 return 0;
ac27a0ec
DK		 1922 }
1923#ifdef CONFIG_QUOTA
49da9392
JK		 1924 /*
1925 * We do the test below only for project quotas. 'usrquota' and
1926 * 'grpquota' mount options are allowed even without quota feature
1927 * to support legacy quotas in quota files.
1928 */
1929 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1930 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1931 "Cannot enable project quota enforcement.");
1932 return 0;
1933 }
1934 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1935 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1936 clear_opt(sb, USRQUOTA);
ac27a0ec 1937
482a7425 1938 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1939 clear_opt(sb, GRPQUOTA);
ac27a0ec 1940
56c50f11 1941 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1942 ext4_msg(sb, KERN_ERR, "old and new quota "
1943 "format mixing");
ac27a0ec
DK		 1944 return 0;
1945 }
1946
1947 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1948 ext4_msg(sb, KERN_ERR, "journaled quota format "
1949 "not specified");
ac27a0ec
DK		 1950 return 0;
1951 }
ac27a0ec
DK		 1952 }
1953#endif
261cb20c
JK		 1954 if (test_opt(sb, DIOREAD_NOLOCK)) {
1955 int blocksize =
1956 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1957
09cbfeaf 1958 if (blocksize < PAGE_SIZE) {
261cb20c
JK		 1959 ext4_msg(sb, KERN_ERR, "can't mount with "
1960 "dioread_nolock if block size != PAGE_SIZE");
1961 return 0;
1962 }
1963 }
ac27a0ec
DK		 1964 return 1;
1965}
1966
2adf6da8
TT		 1967static inline void ext4_show_quota_options(struct seq_file *seq,
1968 struct super_block *sb)
1969{
1970#if defined(CONFIG_QUOTA)
1971 struct ext4_sb_info *sbi = EXT4_SB(sb);
1972
1973 if (sbi->s_jquota_fmt) {
1974 char *fmtname = "";
1975
1976 switch (sbi->s_jquota_fmt) {
1977 case QFMT_VFS_OLD:
1978 fmtname = "vfsold";
1979 break;
1980 case QFMT_VFS_V0:
1981 fmtname = "vfsv0";
1982 break;
1983 case QFMT_VFS_V1:
1984 fmtname = "vfsv1";
1985 break;
1986 }
1987 seq_printf(seq, ",jqfmt=%s", fmtname);
1988 }
1989
1990 if (sbi->s_qf_names[USRQUOTA])
a068acf2 1991 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 1992
1993 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 1994 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1995#endif
1996}
1997
5a916be1
TT		 1998static const char *token2str(int token)
1999{
50df9fd5 2000 const struct match_token *t;
5a916be1
TT		 2001
2002 for (t = tokens; t->token != Opt_err; t++)
2003 if (t->token == token && !strchr(t->pattern, '='))
2004 break;
2005 return t->pattern;
2006}
2007
2adf6da8
TT		 2008/*
2009 * Show an option if
2010 * - it's set to a non-default value OR
2011 * - if the per-sb default is different from the global default
2012 */
66acdcf4
TT		 2013static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2014 int nodefs)
2adf6da8 2015{
2adf6da8
TT		 2016 struct ext4_sb_info *sbi = EXT4_SB(sb);
2017 struct ext4_super_block *es = sbi->s_es;
66acdcf4 2018 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 2019 const struct mount_opts *m;
66acdcf4 2020 char sep = nodefs ? '\n' : ',';
2adf6da8 2021
66acdcf4
TT		 2022#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2023#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2024
2025 if (sbi->s_sb_block != 1)
5a916be1
TT		 2026 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2027
2028 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2029 int want_set = m->flags & MOPT_SET;
2030 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2031 (m->flags & MOPT_CLEAR_ERR))
2032 continue;
66acdcf4 2033 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 2034 continue; /* skip if same as the default */
2035 if ((want_set &&
2036 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2037 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2038 continue; /* select Opt_noFoo vs Opt_Foo */
2039 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2040 }
5a916be1 2041
08cefc7a 2042 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2043 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2044 SEQ_OPTS_PRINT("resuid=%u",
2045 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2046 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2047 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2048 SEQ_OPTS_PRINT("resgid=%u",
2049 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2050 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2051 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2052 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2053 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2054 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2055 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2056 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2057 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2058 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2059 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2060 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2061 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2062 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 2063 if (sb->s_flags & MS_I_VERSION)
5a916be1 2064 SEQ_OPTS_PUTS("i_version");
66acdcf4 2065 if (nodefs || sbi->s_stripe)
5a916be1 2066 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 2067 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2068 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2069 SEQ_OPTS_PUTS("data=journal");
2070 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2071 SEQ_OPTS_PUTS("data=ordered");
2072 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2073 SEQ_OPTS_PUTS("data=writeback");
2074 }
66acdcf4
TT		 2075 if (nodefs ||
2076 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2077 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2078 sbi->s_inode_readahead_blks);
2adf6da8 2079
66acdcf4
TT		 2080 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2081 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2082 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2083 if (nodefs || sbi->s_max_dir_size_kb)
2084 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2085 if (test_opt(sb, DATA_ERR_ABORT))
2086 SEQ_OPTS_PUTS("data_err=abort");
2adf6da8
TT		 2087
2088 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2089 return 0;
2090}
2091
66acdcf4
TT		 2092static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2093{
2094 return _ext4_show_options(seq, root->d_sb, 0);
2095}
2096
ebd173be 2097int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2098{
2099 struct super_block *sb = seq->private;
2100 int rc;
2101
bc98a42c 2102 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
66acdcf4
TT		 2103 rc = _ext4_show_options(seq, sb, 1);
2104 seq_puts(seq, "\n");
2105 return rc;
2106}
2107
617ba13b 2108static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2109 int read_only)
2110{
617ba13b 2111 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2112 int res = 0;
2113
617ba13b 2114 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2115 ext4_msg(sb, KERN_ERR, "revision level too high, "
2116 "forcing read-only mode");
ac27a0ec
DK		 2117 res = MS_RDONLY;
2118 }
2119 if (read_only)
281b5995 2120 goto done;
617ba13b 2121 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2122 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2123 "running e2fsck is recommended");
c8b459f4 2124 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2125 ext4_msg(sb, KERN_WARNING,
2126 "warning: mounting fs with errors, "
2127 "running e2fsck is recommended");
ed3ce80a 2128 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2129 le16_to_cpu(es->s_mnt_count) >=
2130 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2131 ext4_msg(sb, KERN_WARNING,
2132 "warning: maximal mount count reached, "
2133 "running e2fsck is recommended");
ac27a0ec
DK		 2134 else if (le32_to_cpu(es->s_checkinterval) &&
2135 (le32_to_cpu(es->s_lastcheck) +
2136 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 2137 ext4_msg(sb, KERN_WARNING,
2138 "warning: checktime reached, "
2139 "running e2fsck is recommended");
0b8e58a1 2140 if (!sbi->s_journal)
0390131b 2141 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2142 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2143 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2144 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 2145 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 2146 ext4_update_dynamic_rev(sb);
0390131b 2147 if (sbi->s_journal)
e2b911c5 2148 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2149
e2d67052 2150 ext4_commit_super(sb, 1);
281b5995 2151done:
ac27a0ec 2152 if (test_opt(sb, DEBUG))
a9df9a49 2153 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2154 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2155 sb->s_blocksize,
2156 sbi->s_groups_count,
617ba13b
MC		 2157 EXT4_BLOCKS_PER_GROUP(sb),
2158 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2159 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2160
7abc52c2 2161 cleancache_init_fs(sb);
ac27a0ec
DK		 2162 return res;
2163}
2164
117fff10
TT		 2165int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2166{
2167 struct ext4_sb_info *sbi = EXT4_SB(sb);
2168 struct flex_groups *new_groups;
2169 int size;
2170
2171 if (!sbi->s_log_groups_per_flex)
2172 return 0;
2173
2174 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2175 if (size <= sbi->s_flex_groups_allocated)
2176 return 0;
2177
2178 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
a7c3e901 2179 new_groups = kvzalloc(size, GFP_KERNEL);
117fff10
TT		 2180 if (!new_groups) {
2181 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2182 size / (int) sizeof(struct flex_groups));
2183 return -ENOMEM;
2184 }
2185
2186 if (sbi->s_flex_groups) {
2187 memcpy(new_groups, sbi->s_flex_groups,
2188 (sbi->s_flex_groups_allocated *
2189 sizeof(struct flex_groups)));
b93b41d4 2190 kvfree(sbi->s_flex_groups);
117fff10
TT		 2191 }
2192 sbi->s_flex_groups = new_groups;
2193 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2194 return 0;
2195}
2196
772cb7c8
JS		 2197static int ext4_fill_flex_info(struct super_block *sb)
2198{
2199 struct ext4_sb_info *sbi = EXT4_SB(sb);
2200 struct ext4_group_desc *gdp = NULL;
772cb7c8 2201 ext4_group_t flex_group;
117fff10 2202 int i, err;
772cb7c8 2203
503358ae 2204 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2205 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2206 sbi->s_log_groups_per_flex = 0;
2207 return 1;
2208 }
2209
117fff10
TT		 2210 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2211 if (err)
9933fc0a 2212 goto failed;
772cb7c8 2213
772cb7c8 2214 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2215 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2216
2217 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2218 atomic_add(ext4_free_inodes_count(sb, gdp),
2219 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2220 atomic64_add(ext4_free_group_clusters(sb, gdp),
2221 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2222 atomic_add(ext4_used_dirs_count(sb, gdp),
2223 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2224 }
2225
2226 return 1;
2227failed:
2228 return 0;
2229}
2230
e2b911c5 2231static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2232 struct ext4_group_desc *gdp)
717d50e4 2233{
b47820ed 2234 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2235 __u16 crc = 0;
feb0ab32 2236 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2237 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2238
9aa5d32b 2239 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2240 /* Use new metadata_csum algorithm */
feb0ab32 2241 __u32 csum32;
b47820ed 2242 __u16 dummy_csum = 0;
feb0ab32 2243
feb0ab32
DW		 2244 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2245 sizeof(le_group));
b47820ed
DJ		 2246 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2247 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2248 sizeof(dummy_csum));
2249 offset += sizeof(dummy_csum);
2250 if (offset < sbi->s_desc_size)
2251 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2252 sbi->s_desc_size - offset);
feb0ab32
DW		 2253
2254 crc = csum32 & 0xFFFF;
2255 goto out;
717d50e4
AD		 2256 }
2257
feb0ab32 2258 /* old crc16 code */
e2b911c5 2259 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2260 return 0;
2261
feb0ab32
DW		 2262 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2263 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2264 crc = crc16(crc, (__u8 *)gdp, offset);
2265 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2266 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2267 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2268 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2269 crc = crc16(crc, (__u8 *)gdp + offset,
2270 le16_to_cpu(sbi->s_es->s_desc_size) -
2271 offset);
2272
2273out:
717d50e4
AD		 2274 return cpu_to_le16(crc);
2275}
2276
feb0ab32 2277int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2278 struct ext4_group_desc *gdp)
2279{
feb0ab32 2280 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2281 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2282 return 0;
2283
2284 return 1;
2285}
2286
feb0ab32
DW		 2287void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2288 struct ext4_group_desc *gdp)
2289{
2290 if (!ext4_has_group_desc_csum(sb))
2291 return;
e2b911c5 2292 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2293}
2294
ac27a0ec 2295/* Called at mount-time, super-block is locked */
bfff6873 2296static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2297 ext4_fsblk_t sb_block,
bfff6873 2298 ext4_group_t *first_not_zeroed)
ac27a0ec 2299{
617ba13b
MC		 2300 struct ext4_sb_info *sbi = EXT4_SB(sb);
2301 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2302 ext4_fsblk_t last_block;
bd81d8ee
LV		 2303 ext4_fsblk_t block_bitmap;
2304 ext4_fsblk_t inode_bitmap;
2305 ext4_fsblk_t inode_table;
ce421581 2306 int flexbg_flag = 0;
bfff6873 2307 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2308
e2b911c5 2309 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2310 flexbg_flag = 1;
2311
af5bc92d 2312 ext4_debug("Checking group descriptors");
ac27a0ec 2313
197cd65a
AM		 2314 for (i = 0; i < sbi->s_groups_count; i++) {
2315 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2316
ce421581 2317 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2318 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2319 else
2320 last_block = first_block +
617ba13b 2321 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2322
bfff6873
LC		 2323 if ((grp == sbi->s_groups_count) &&
2324 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2325 grp = i;
2326
8fadc143 2327 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2328 if (block_bitmap == sb_block) {
2329 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2330 "Block bitmap for group %u overlaps "
2331 "superblock", i);
2332 }
2b2d6d01 2333 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2334 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2335 "Block bitmap for group %u not in group "
b31e1552 2336 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2337 return 0;
2338 }
8fadc143 2339 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2340 if (inode_bitmap == sb_block) {
2341 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2342 "Inode bitmap for group %u overlaps "
2343 "superblock", i);
2344 }
2b2d6d01 2345 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2346 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2347 "Inode bitmap for group %u not in group "
b31e1552 2348 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2349 return 0;
2350 }
8fadc143 2351 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2352 if (inode_table == sb_block) {
2353 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2354 "Inode table for group %u overlaps "
2355 "superblock", i);
2356 }
bd81d8ee 2357 if (inode_table < first_block ||
2b2d6d01 2358 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2359 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2360 "Inode table for group %u not in group "
b31e1552 2361 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2362 return 0;
2363 }
955ce5f5 2364 ext4_lock_group(sb, i);
feb0ab32 2365 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2366 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2367 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2368 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2369 gdp)), le16_to_cpu(gdp->bg_checksum));
bc98a42c 2370 if (!sb_rdonly(sb)) {
955ce5f5 2371 ext4_unlock_group(sb, i);
8a266467 2372 return 0;
7ee1ec4c 2373 }
717d50e4 2374 }
955ce5f5 2375 ext4_unlock_group(sb, i);
ce421581
JS		 2376 if (!flexbg_flag)
2377 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2378 }
bfff6873
LC		 2379 if (NULL != first_not_zeroed)
2380 *first_not_zeroed = grp;
ac27a0ec
DK		 2381 return 1;
2382}
2383
617ba13b 2384/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2385 * the superblock) which were deleted from all directories, but held open by
2386 * a process at the time of a crash. We walk the list and try to delete these
2387 * inodes at recovery time (only with a read-write filesystem).
2388 *
2389 * In order to keep the orphan inode chain consistent during traversal (in
2390 * case of crash during recovery), we link each inode into the superblock
2391 * orphan list_head and handle it the same way as an inode deletion during
2392 * normal operation (which journals the operations for us).
2393 *
2394 * We only do an iget() and an iput() on each inode, which is very safe if we
2395 * accidentally point at an in-use or already deleted inode. The worst that
2396 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2397 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2398 * e2fsck was run on this filesystem, and it must have already done the orphan
2399 * inode cleanup for us, so we can safely abort without any further action.
2400 */
2b2d6d01
TT		 2401static void ext4_orphan_cleanup(struct super_block *sb,
2402 struct ext4_super_block *es)
ac27a0ec
DK		 2403{
2404 unsigned int s_flags = sb->s_flags;
2c98eb5e 2405 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec 2406#ifdef CONFIG_QUOTA
95f1fda4 2407 int quota_update = 0;
ac27a0ec
DK		 2408 int i;
2409#endif
2410 if (!es->s_last_orphan) {
2411 jbd_debug(4, "no orphan inodes to clean up\n");
2412 return;
2413 }
2414
a8f48a95 2415 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2416 ext4_msg(sb, KERN_ERR, "write access "
2417 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2418 return;
2419 }
2420
d39195c3
AG		 2421 /* Check if feature set would not allow a r/w mount */
2422 if (!ext4_feature_set_ok(sb, 0)) {
2423 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2424 "unknown ROCOMPAT features");
2425 return;
2426 }
2427
617ba13b 2428 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2429 /* don't clear list on RO mount w/ errors */
2430 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2431 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2432 "clearing orphan list.\n");
c25f9bc6
ES		 2433 es->s_last_orphan = 0;
2434 }
ac27a0ec
DK		 2435 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2436 return;
2437 }
2438
2439 if (s_flags & MS_RDONLY) {
b31e1552 2440 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2441 sb->s_flags &= ~MS_RDONLY;
2442 }
2443#ifdef CONFIG_QUOTA
2444 /* Needed for iput() to work correctly and not trash data */
2445 sb->s_flags |= MS_ACTIVE;
95f1fda4 2446
2447 /*
2448 * Turn on quotas which were not enabled for read-only mounts if
2449 * filesystem has quota feature, so that they are updated correctly.
2450 */
2451 if (ext4_has_feature_quota(sb) && (s_flags & MS_RDONLY)) {
2452 int ret = ext4_enable_quotas(sb);
2453
2454 if (!ret)
2455 quota_update = 1;
2456 else
2457 ext4_msg(sb, KERN_ERR,
2458 "Cannot turn on quotas: error %d", ret);
2459 }
2460
2461 /* Turn on journaled quotas used for old sytle */
a2d4a646 2462 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2463 if (EXT4_SB(sb)->s_qf_names[i]) {
2464 int ret = ext4_quota_on_mount(sb, i);
95f1fda4 2465
2466 if (!ret)
2467 quota_update = 1;
2468 else
b31e1552
ES		 2469 ext4_msg(sb, KERN_ERR,
2470 "Cannot turn on journaled "
95f1fda4 2471 "quota: type %d: error %d", i, ret);
ac27a0ec
DK		 2472 }
2473 }
2474#endif
2475
2476 while (es->s_last_orphan) {
2477 struct inode *inode;
2478
c65d5c6c
VN		 2479 /*
2480 * We may have encountered an error during cleanup; if
2481 * so, skip the rest.
2482 */
2483 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2484 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2485 es->s_last_orphan = 0;
2486 break;
2487 }
2488
97bd42b9
JB		 2489 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2490 if (IS_ERR(inode)) {
ac27a0ec
DK		 2491 es->s_last_orphan = 0;
2492 break;
2493 }
2494
617ba13b 2495 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2496 dquot_initialize(inode);
ac27a0ec 2497 if (inode->i_nlink) {
566370a2
PT		 2498 if (test_opt(sb, DEBUG))
2499 ext4_msg(sb, KERN_DEBUG,
2500 "%s: truncating inode %lu to %lld bytes",
2501 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2502 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2503 inode->i_ino, inode->i_size);
5955102c 2504 inode_lock(inode);
55f252c9 2505 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 2506 ret = ext4_truncate(inode);
2507 if (ret)
2508 ext4_std_error(inode->i_sb, ret);
5955102c 2509 inode_unlock(inode);
ac27a0ec
DK		 2510 nr_truncates++;
2511 } else {
566370a2
PT		 2512 if (test_opt(sb, DEBUG))
2513 ext4_msg(sb, KERN_DEBUG,
2514 "%s: deleting unreferenced inode %lu",
2515 __func__, inode->i_ino);
ac27a0ec
DK		 2516 jbd_debug(2, "deleting unreferenced inode %lu\n",
2517 inode->i_ino);
2518 nr_orphans++;
2519 }
2520 iput(inode); /* The delete magic happens here! */
2521 }
2522
2b2d6d01 2523#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2524
2525 if (nr_orphans)
b31e1552
ES		 2526 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2527 PLURAL(nr_orphans));
ac27a0ec 2528 if (nr_truncates)
b31e1552
ES		 2529 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2530 PLURAL(nr_truncates));
ac27a0ec 2531#ifdef CONFIG_QUOTA
95f1fda4 2532 /* Turn off quotas if they were enabled for orphan cleanup */
2533 if (quota_update) {
2534 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2535 if (sb_dqopt(sb)->files[i])
2536 dquot_quota_off(sb, i);
2537 }
ac27a0ec
DK		 2538 }
2539#endif
2540 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2541}
0b8e58a1 2542
cd2291a4
ES		 2543/*
2544 * Maximal extent format file size.
2545 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2546 * extent format containers, within a sector_t, and within i_blocks
2547 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2548 * so that won't be a limiting factor.
2549 *
f17722f9
LC		 2550 * However there is other limiting factor. We do store extents in the form
2551 * of starting block and length, hence the resulting length of the extent
2552 * covering maximum file size must fit into on-disk format containers as
2553 * well. Given that length is always by 1 unit bigger than max unit (because
2554 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2555 *
cd2291a4
ES		 2556 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2557 */
f287a1a5 2558static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2559{
2560 loff_t res;
2561 loff_t upper_limit = MAX_LFS_FILESIZE;
2562
2563 /* small i_blocks in vfs inode? */
f287a1a5 2564 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2565 /*
90c699a9 2566 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2567 * i_block represent total blocks in 512 bytes
2568 * 32 == size of vfs inode i_blocks * 8
2569 */
2570 upper_limit = (1LL << 32) - 1;
2571
2572 /* total blocks in file system block size */
2573 upper_limit >>= (blkbits - 9);
2574 upper_limit <<= blkbits;
2575 }
2576
f17722f9
LC		 2577 /*
2578 * 32-bit extent-start container, ee_block. We lower the maxbytes
2579 * by one fs block, so ee_len can cover the extent of maximum file
2580 * size
2581 */
2582 res = (1LL << 32) - 1;
cd2291a4 2583 res <<= blkbits;
cd2291a4
ES		 2584
2585 /* Sanity check against vm- & vfs- imposed limits */
2586 if (res > upper_limit)
2587 res = upper_limit;
2588
2589 return res;
2590}
ac27a0ec 2591
ac27a0ec 2592/*
cd2291a4 2593 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2594 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2595 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2596 */
f287a1a5 2597static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2598{
617ba13b 2599 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2600 int meta_blocks;
2601 loff_t upper_limit;
0b8e58a1
AD		 2602 /* This is calculated to be the largest file size for a dense, block
2603 * mapped file such that the file's total number of 512-byte sectors,
2604 * including data and all indirect blocks, does not exceed (2^48 - 1).
2605 *
2606 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2607 * number of 512-byte sectors of the file.
0fc1b451
AK		 2608 */
2609
f287a1a5 2610 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2611 /*
90c699a9 2612 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2613 * the inode i_block field represents total file blocks in
2614 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2615 */
2616 upper_limit = (1LL << 32) - 1;
2617
2618 /* total blocks in file system block size */
2619 upper_limit >>= (bits - 9);
2620
2621 } else {
8180a562
AK		 2622 /*
2623 * We use 48 bit ext4_inode i_blocks
2624 * With EXT4_HUGE_FILE_FL set the i_blocks
2625 * represent total number of blocks in
2626 * file system block size
2627 */
0fc1b451
AK		 2628 upper_limit = (1LL << 48) - 1;
2629
0fc1b451
AK		 2630 }
2631
2632 /* indirect blocks */
2633 meta_blocks = 1;
2634 /* double indirect blocks */
2635 meta_blocks += 1 + (1LL << (bits-2));
2636 /* tripple indirect blocks */
2637 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2638
2639 upper_limit -= meta_blocks;
2640 upper_limit <<= bits;
ac27a0ec
DK		 2641
2642 res += 1LL << (bits-2);
2643 res += 1LL << (2*(bits-2));
2644 res += 1LL << (3*(bits-2));
2645 res <<= bits;
2646 if (res > upper_limit)
2647 res = upper_limit;
0fc1b451
AK		 2648
2649 if (res > MAX_LFS_FILESIZE)
2650 res = MAX_LFS_FILESIZE;
2651
ac27a0ec
DK		 2652 return res;
2653}
2654
617ba13b 2655static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2656 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2657{
617ba13b 2658 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2659 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2660 int has_super = 0;
2661
2662 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2663
e2b911c5 2664 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2665 return logical_sb_block + nr + 1;
ac27a0ec 2666 bg = sbi->s_desc_per_block * nr;
617ba13b 2667 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2668 has_super = 1;
0b8e58a1 2669
bd63f6b0
DW		 2670 /*
2671 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2672 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2673 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2674 * compensate.
2675 */
2676 if (sb->s_blocksize == 1024 && nr == 0 &&
2677 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2678 has_super++;
2679
617ba13b 2680 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2681}
2682
c9de560d
AT		 2683/**
2684 * ext4_get_stripe_size: Get the stripe size.
2685 * @sbi: In memory super block info
2686 *
2687 * If we have specified it via mount option, then
2688 * use the mount option value. If the value specified at mount time is
2689 * greater than the blocks per group use the super block value.
2690 * If the super block value is greater than blocks per group return 0.
2691 * Allocator needs it be less than blocks per group.
2692 *
2693 */
2694static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2695{
2696 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2697 unsigned long stripe_width =
2698 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2699 int ret;
c9de560d
AT		 2700
2701 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 2702 ret = sbi->s_stripe;
5469d7c3 2703 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 2704 ret = stripe_width;
5469d7c3 2705 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 2706 ret = stride;
2707 else
2708 ret = 0;
c9de560d 2709
3eb08658
DE		 2710 /*
2711 * If the stripe width is 1, this makes no sense and
2712 * we set it to 0 to turn off stripe handling code.
2713 */
2714 if (ret <= 1)
2715 ret = 0;
c9de560d 2716
3eb08658 2717 return ret;
c9de560d 2718}
ac27a0ec 2719
a13fb1a4
ES		 2720/*
2721 * Check whether this filesystem can be mounted based on
2722 * the features present and the RDONLY/RDWR mount requested.
2723 * Returns 1 if this filesystem can be mounted as requested,
2724 * 0 if it cannot be.
2725 */
2726static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2727{
e2b911c5 2728 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2729 ext4_msg(sb, KERN_ERR,
2730 "Couldn't mount because of "
2731 "unsupported optional features (%x)",
2732 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2733 ~EXT4_FEATURE_INCOMPAT_SUPP));
2734 return 0;
2735 }
2736
2737 if (readonly)
2738 return 1;
2739
e2b911c5 2740 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b
DW		 2741 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2742 sb->s_flags |= MS_RDONLY;
2743 return 1;
2744 }
2745
a13fb1a4 2746 /* Check that feature set is OK for a read-write mount */
e2b911c5 2747 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2748 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2749 "unsupported optional features (%x)",
2750 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2751 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2752 return 0;
2753 }
2754 /*
2755 * Large file size enabled file system can only be mounted
2756 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2757 */
e2b911c5 2758 if (ext4_has_feature_huge_file(sb)) {
a13fb1a4
ES		 2759 if (sizeof(blkcnt_t) < sizeof(u64)) {
2760 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2761 "cannot be mounted RDWR without "
2762 "CONFIG_LBDAF");
2763 return 0;
2764 }
2765 }
e2b911c5 2766 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2767 ext4_msg(sb, KERN_ERR,
2768 "Can't support bigalloc feature without "
2769 "extents feature\n");
2770 return 0;
2771 }
7c319d32
AK		 2772
2773#ifndef CONFIG_QUOTA
e2b911c5 2774 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2775 ext4_msg(sb, KERN_ERR,
2776 "Filesystem with quota feature cannot be mounted RDWR "
2777 "without CONFIG_QUOTA");
2778 return 0;
2779 }
689c958c
LX		 2780 if (ext4_has_feature_project(sb) && !readonly) {
2781 ext4_msg(sb, KERN_ERR,
2782 "Filesystem with project quota feature cannot be mounted RDWR "
2783 "without CONFIG_QUOTA");
2784 return 0;
2785 }
7c319d32 2786#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2787 return 1;
2788}
2789
66e61a9e
TT		 2790/*
2791 * This function is called once a day if we have errors logged
2792 * on the file system
2793 */
2794static void print_daily_error_info(unsigned long arg)
2795{
2796 struct super_block *sb = (struct super_block *) arg;
2797 struct ext4_sb_info *sbi;
2798 struct ext4_super_block *es;
2799
2800 sbi = EXT4_SB(sb);
2801 es = sbi->s_es;
2802
2803 if (es->s_error_count)
ae0f78de
TT		 2804 /* fsck newer than v1.41.13 is needed to clean this condition. */
2805 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2806 le32_to_cpu(es->s_error_count));
2807 if (es->s_first_error_time) {
ae0f78de 2808 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2809 sb->s_id, le32_to_cpu(es->s_first_error_time),
2810 (int) sizeof(es->s_first_error_func),
2811 es->s_first_error_func,
2812 le32_to_cpu(es->s_first_error_line));
2813 if (es->s_first_error_ino)
651e1c3b 2814 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2815 le32_to_cpu(es->s_first_error_ino));
2816 if (es->s_first_error_block)
651e1c3b 2817 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2818 le64_to_cpu(es->s_first_error_block));
651e1c3b 2819 printk(KERN_CONT "\n");
66e61a9e
TT		 2820 }
2821 if (es->s_last_error_time) {
ae0f78de 2822 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2823 sb->s_id, le32_to_cpu(es->s_last_error_time),
2824 (int) sizeof(es->s_last_error_func),
2825 es->s_last_error_func,
2826 le32_to_cpu(es->s_last_error_line));
2827 if (es->s_last_error_ino)
651e1c3b 2828 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2829 le32_to_cpu(es->s_last_error_ino));
2830 if (es->s_last_error_block)
651e1c3b 2831 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2832 le64_to_cpu(es->s_last_error_block));
651e1c3b 2833 printk(KERN_CONT "\n");
66e61a9e
TT		 2834 }
2835 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2836}
2837
bfff6873
LC		 2838/* Find next suitable group and run ext4_init_inode_table */
2839static int ext4_run_li_request(struct ext4_li_request *elr)
2840{
2841 struct ext4_group_desc *gdp = NULL;
2842 ext4_group_t group, ngroups;
2843 struct super_block *sb;
2844 unsigned long timeout = 0;
2845 int ret = 0;
2846
2847 sb = elr->lr_super;
2848 ngroups = EXT4_SB(sb)->s_groups_count;
2849
2850 for (group = elr->lr_next_group; group < ngroups; group++) {
2851 gdp = ext4_get_group_desc(sb, group, NULL);
2852 if (!gdp) {
2853 ret = 1;
2854 break;
2855 }
2856
2857 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2858 break;
2859 }
2860
7f511862 2861 if (group >= ngroups)
bfff6873
LC		 2862 ret = 1;
2863
2864 if (!ret) {
2865 timeout = jiffies;
2866 ret = ext4_init_inode_table(sb, group,
2867 elr->lr_timeout ? 0 : 1);
2868 if (elr->lr_timeout == 0) {
51ce6511
LC		 2869 timeout = (jiffies - timeout) *
2870 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2871 elr->lr_timeout = timeout;
2872 }
2873 elr->lr_next_sched = jiffies + elr->lr_timeout;
2874 elr->lr_next_group = group + 1;
2875 }
bfff6873
LC		 2876 return ret;
2877}
2878
2879/*
2880 * Remove lr_request from the list_request and free the
4ed5c033 2881 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2882 */
2883static void ext4_remove_li_request(struct ext4_li_request *elr)
2884{
2885 struct ext4_sb_info *sbi;
2886
2887 if (!elr)
2888 return;
2889
2890 sbi = elr->lr_sbi;
2891
2892 list_del(&elr->lr_request);
2893 sbi->s_li_request = NULL;
2894 kfree(elr);
2895}
2896
2897static void ext4_unregister_li_request(struct super_block *sb)
2898{
1bb933fb
LC		 2899 mutex_lock(&ext4_li_mtx);
2900 if (!ext4_li_info) {
2901 mutex_unlock(&ext4_li_mtx);
bfff6873 2902 return;
1bb933fb 2903 }
bfff6873
LC		 2904
2905 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2906 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2907 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2908 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2909}
2910
8f1f7453
ES		 2911static struct task_struct *ext4_lazyinit_task;
2912
bfff6873
LC		 2913/*
2914 * This is the function where ext4lazyinit thread lives. It walks
2915 * through the request list searching for next scheduled filesystem.
2916 * When such a fs is found, run the lazy initialization request
2917 * (ext4_rn_li_request) and keep track of the time spend in this
2918 * function. Based on that time we compute next schedule time of
2919 * the request. When walking through the list is complete, compute
2920 * next waking time and put itself into sleep.
2921 */
2922static int ext4_lazyinit_thread(void *arg)
2923{
2924 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2925 struct list_head *pos, *n;
2926 struct ext4_li_request *elr;
4ed5c033 2927 unsigned long next_wakeup, cur;
bfff6873
LC		 2928
2929 BUG_ON(NULL == eli);
2930
bfff6873
LC		 2931cont_thread:
2932 while (true) {
2933 next_wakeup = MAX_JIFFY_OFFSET;
2934
2935 mutex_lock(&eli->li_list_mtx);
2936 if (list_empty(&eli->li_request_list)) {
2937 mutex_unlock(&eli->li_list_mtx);
2938 goto exit_thread;
2939 }
bfff6873 2940 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 2941 int err = 0;
2942 int progress = 0;
bfff6873
LC		 2943 elr = list_entry(pos, struct ext4_li_request,
2944 lr_request);
2945
e22834f0
DM		 2946 if (time_before(jiffies, elr->lr_next_sched)) {
2947 if (time_before(elr->lr_next_sched, next_wakeup))
2948 next_wakeup = elr->lr_next_sched;
2949 continue;
2950 }
2951 if (down_read_trylock(&elr->lr_super->s_umount)) {
2952 if (sb_start_write_trylock(elr->lr_super)) {
2953 progress = 1;
2954 /*
2955 * We hold sb->s_umount, sb can not
2956 * be removed from the list, it is
2957 * now safe to drop li_list_mtx
2958 */
2959 mutex_unlock(&eli->li_list_mtx);
2960 err = ext4_run_li_request(elr);
2961 sb_end_write(elr->lr_super);
2962 mutex_lock(&eli->li_list_mtx);
2963 n = pos->next;
b2c78cd0 2964 }
e22834f0
DM		 2965 up_read((&elr->lr_super->s_umount));
2966 }
2967 /* error, remove the lazy_init job */
2968 if (err) {
2969 ext4_remove_li_request(elr);
2970 continue;
2971 }
2972 if (!progress) {
2973 elr->lr_next_sched = jiffies +
2974 (prandom_u32()
2975 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 2976 }
bfff6873
LC		 2977 if (time_before(elr->lr_next_sched, next_wakeup))
2978 next_wakeup = elr->lr_next_sched;
2979 }
2980 mutex_unlock(&eli->li_list_mtx);
2981
a0acae0e 2982 try_to_freeze();
bfff6873 2983
4ed5c033
LC		 2984 cur = jiffies;
2985 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2986 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2987 cond_resched();
2988 continue;
2989 }
2990
4ed5c033
LC		 2991 schedule_timeout_interruptible(next_wakeup - cur);
2992
8f1f7453
ES		 2993 if (kthread_should_stop()) {
2994 ext4_clear_request_list();
2995 goto exit_thread;
2996 }
bfff6873
LC		 2997 }
2998
2999exit_thread:
3000 /*
3001 * It looks like the request list is empty, but we need
3002 * to check it under the li_list_mtx lock, to prevent any
3003 * additions into it, and of course we should lock ext4_li_mtx
3004 * to atomically free the list and ext4_li_info, because at
3005 * this point another ext4 filesystem could be registering
3006 * new one.
3007 */
3008 mutex_lock(&ext4_li_mtx);
3009 mutex_lock(&eli->li_list_mtx);
3010 if (!list_empty(&eli->li_request_list)) {
3011 mutex_unlock(&eli->li_list_mtx);
3012 mutex_unlock(&ext4_li_mtx);
3013 goto cont_thread;
3014 }
3015 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3016 kfree(ext4_li_info);
3017 ext4_li_info = NULL;
3018 mutex_unlock(&ext4_li_mtx);
3019
3020 return 0;
3021}
3022
3023static void ext4_clear_request_list(void)
3024{
3025 struct list_head *pos, *n;
3026 struct ext4_li_request *elr;
3027
3028 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3029 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3030 elr = list_entry(pos, struct ext4_li_request,
3031 lr_request);
3032 ext4_remove_li_request(elr);
3033 }
3034 mutex_unlock(&ext4_li_info->li_list_mtx);
3035}
3036
3037static int ext4_run_lazyinit_thread(void)
3038{
8f1f7453
ES		 3039 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3040 ext4_li_info, "ext4lazyinit");
3041 if (IS_ERR(ext4_lazyinit_task)) {
3042 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3043 ext4_clear_request_list();
bfff6873
LC		 3044 kfree(ext4_li_info);
3045 ext4_li_info = NULL;
92b97816 3046 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3047 "initialization thread\n",
3048 err);
3049 return err;
3050 }
3051 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3052 return 0;
3053}
3054
3055/*
3056 * Check whether it make sense to run itable init. thread or not.
3057 * If there is at least one uninitialized inode table, return
3058 * corresponding group number, else the loop goes through all
3059 * groups and return total number of groups.
3060 */
3061static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3062{
3063 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3064 struct ext4_group_desc *gdp = NULL;
3065
3066 for (group = 0; group < ngroups; group++) {
3067 gdp = ext4_get_group_desc(sb, group, NULL);
3068 if (!gdp)
3069 continue;
3070
3071 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3072 break;
3073 }
3074
3075 return group;
3076}
3077
3078static int ext4_li_info_new(void)
3079{
3080 struct ext4_lazy_init *eli = NULL;
3081
3082 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3083 if (!eli)
3084 return -ENOMEM;
3085
bfff6873
LC		 3086 INIT_LIST_HEAD(&eli->li_request_list);
3087 mutex_init(&eli->li_list_mtx);
3088
bfff6873
LC		 3089 eli->li_state |= EXT4_LAZYINIT_QUIT;
3090
3091 ext4_li_info = eli;
3092
3093 return 0;
3094}
3095
3096static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3097 ext4_group_t start)
3098{
3099 struct ext4_sb_info *sbi = EXT4_SB(sb);
3100 struct ext4_li_request *elr;
bfff6873
LC		 3101
3102 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3103 if (!elr)
3104 return NULL;
3105
3106 elr->lr_super = sb;
3107 elr->lr_sbi = sbi;
3108 elr->lr_next_group = start;
3109
3110 /*
3111 * Randomize first schedule time of the request to
3112 * spread the inode table initialization requests
3113 * better.
3114 */
dd1f723b
TT		 3115 elr->lr_next_sched = jiffies + (prandom_u32() %
3116 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3117 return elr;
3118}
3119
7f511862
TT		 3120int ext4_register_li_request(struct super_block *sb,
3121 ext4_group_t first_not_zeroed)
bfff6873
LC		 3122{
3123 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3124 struct ext4_li_request *elr = NULL;
bfff6873 3125 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3126 int ret = 0;
bfff6873 3127
7f511862 3128 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3129 if (sbi->s_li_request != NULL) {
3130 /*
3131 * Reset timeout so it can be computed again, because
3132 * s_li_wait_mult might have changed.
3133 */
3134 sbi->s_li_request->lr_timeout = 0;
7f511862 3135 goto out;
51ce6511 3136 }
bfff6873 3137
bc98a42c 3138 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
55ff3840 3139 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3140 goto out;
bfff6873
LC		 3141
3142 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3143 if (!elr) {
3144 ret = -ENOMEM;
3145 goto out;
3146 }
bfff6873
LC		 3147
3148 if (NULL == ext4_li_info) {
3149 ret = ext4_li_info_new();
3150 if (ret)
3151 goto out;
3152 }
3153
3154 mutex_lock(&ext4_li_info->li_list_mtx);
3155 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3156 mutex_unlock(&ext4_li_info->li_list_mtx);
3157
3158 sbi->s_li_request = elr;
46e4690b
TM		 3159 /*
3160 * set elr to NULL here since it has been inserted to
3161 * the request_list and the removal and free of it is
3162 * handled by ext4_clear_request_list from now on.
3163 */
3164 elr = NULL;
bfff6873
LC		 3165
3166 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3167 ret = ext4_run_lazyinit_thread();
3168 if (ret)
3169 goto out;
3170 }
bfff6873 3171out:
beed5ecb
NK		 3172 mutex_unlock(&ext4_li_mtx);
3173 if (ret)
bfff6873 3174 kfree(elr);
bfff6873
LC		 3175 return ret;
3176}
3177
3178/*
3179 * We do not need to lock anything since this is called on
3180 * module unload.
3181 */
3182static void ext4_destroy_lazyinit_thread(void)
3183{
3184 /*
3185 * If thread exited earlier
3186 * there's nothing to be done.
3187 */
8f1f7453 3188 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3189 return;
3190
8f1f7453 3191 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3192}
3193
25ed6e8a
DW		 3194static int set_journal_csum_feature_set(struct super_block *sb)
3195{
3196 int ret = 1;
3197 int compat, incompat;
3198 struct ext4_sb_info *sbi = EXT4_SB(sb);
3199
9aa5d32b 3200 if (ext4_has_metadata_csum(sb)) {
db9ee220 3201 /* journal checksum v3 */
25ed6e8a 3202 compat = 0;
db9ee220 3203 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3204 } else {
3205 /* journal checksum v1 */
3206 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3207 incompat = 0;
3208 }
3209
feb8c6d3
DW		 3210 jbd2_journal_clear_features(sbi->s_journal,
3211 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3212 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3213 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3214 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3215 ret = jbd2_journal_set_features(sbi->s_journal,
3216 compat, 0,
3217 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3218 incompat);
3219 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3220 ret = jbd2_journal_set_features(sbi->s_journal,
3221 compat, 0,
3222 incompat);
3223 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3224 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3225 } else {
feb8c6d3
DW		 3226 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3227 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3228 }
3229
3230 return ret;
3231}
3232
952fc18e
TT		 3233/*
3234 * Note: calculating the overhead so we can be compatible with
3235 * historical BSD practice is quite difficult in the face of
3236 * clusters/bigalloc. This is because multiple metadata blocks from
3237 * different block group can end up in the same allocation cluster.
3238 * Calculating the exact overhead in the face of clustered allocation
3239 * requires either O(all block bitmaps) in memory or O(number of block
3240 * groups**2) in time. We will still calculate the superblock for
3241 * older file systems --- and if we come across with a bigalloc file
3242 * system with zero in s_overhead_clusters the estimate will be close to
3243 * correct especially for very large cluster sizes --- but for newer
3244 * file systems, it's better to calculate this figure once at mkfs
3245 * time, and store it in the superblock. If the superblock value is
3246 * present (even for non-bigalloc file systems), we will use it.
3247 */
3248static int count_overhead(struct super_block *sb, ext4_group_t grp,
3249 char *buf)
3250{
3251 struct ext4_sb_info *sbi = EXT4_SB(sb);
3252 struct ext4_group_desc *gdp;
3253 ext4_fsblk_t first_block, last_block, b;
3254 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3255 int s, j, count = 0;
3256
e2b911c5 3257 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3258 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3259 sbi->s_itb_per_group + 2);
3260
952fc18e
TT		 3261 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3262 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3263 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3264 for (i = 0; i < ngroups; i++) {
3265 gdp = ext4_get_group_desc(sb, i, NULL);
3266 b = ext4_block_bitmap(sb, gdp);
3267 if (b >= first_block && b <= last_block) {
3268 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3269 count++;
3270 }
3271 b = ext4_inode_bitmap(sb, gdp);
3272 if (b >= first_block && b <= last_block) {
3273 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3274 count++;
3275 }
3276 b = ext4_inode_table(sb, gdp);
3277 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3278 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3279 int c = EXT4_B2C(sbi, b - first_block);
3280 ext4_set_bit(c, buf);
3281 count++;
3282 }
3283 if (i != grp)
3284 continue;
3285 s = 0;
3286 if (ext4_bg_has_super(sb, grp)) {
3287 ext4_set_bit(s++, buf);
3288 count++;
3289 }
c48ae41b
TT		 3290 j = ext4_bg_num_gdb(sb, grp);
3291 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3292 ext4_error(sb, "Invalid number of block group "
3293 "descriptor blocks: %d", j);
3294 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3295 }
c48ae41b
TT		 3296 count += j;
3297 for (; j > 0; j--)
3298 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3299 }
3300 if (!count)
3301 return 0;
3302 return EXT4_CLUSTERS_PER_GROUP(sb) -
3303 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3304}
3305
3306/*
3307 * Compute the overhead and stash it in sbi->s_overhead
3308 */
3309int ext4_calculate_overhead(struct super_block *sb)
3310{
3311 struct ext4_sb_info *sbi = EXT4_SB(sb);
3312 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3313 struct inode *j_inode;
3314 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3315 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3316 ext4_fsblk_t overhead = 0;
4fdb5543 3317 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3318
952fc18e
TT		 3319 if (!buf)
3320 return -ENOMEM;
3321
3322 /*
3323 * Compute the overhead (FS structures). This is constant
3324 * for a given filesystem unless the number of block groups
3325 * changes so we cache the previous value until it does.
3326 */
3327
3328 /*
3329 * All of the blocks before first_data_block are overhead
3330 */
3331 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3332
3333 /*
3334 * Add the overhead found in each block group
3335 */
3336 for (i = 0; i < ngroups; i++) {
3337 int blks;
3338
3339 blks = count_overhead(sb, i, buf);
3340 overhead += blks;
3341 if (blks)
3342 memset(buf, 0, PAGE_SIZE);
3343 cond_resched();
3344 }
3c816ded
EW		 3345
3346 /*
3347 * Add the internal journal blocks whether the journal has been
3348 * loaded or not
3349 */
b003b524 3350 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3351 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3c816ded
EW		 3352 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3353 j_inode = ext4_get_journal_inode(sb, j_inum);
3354 if (j_inode) {
3355 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3356 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3357 iput(j_inode);
3358 } else {
3359 ext4_msg(sb, KERN_ERR, "can't get journal size");
3360 }
3361 }
952fc18e
TT		 3362 sbi->s_overhead = overhead;
3363 smp_wmb();
3364 free_page((unsigned long) buf);
3365 return 0;
3366}
3367
b5799018 3368static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3369{
3370 ext4_fsblk_t resv_clusters;
b5799018 3371 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3372
30fac0f7
JK		 3373 /*
3374 * There's no need to reserve anything when we aren't using extents.
3375 * The space estimates are exact, there are no unwritten extents,
3376 * hole punching doesn't need new metadata... This is needed especially
3377 * to keep ext2/3 backward compatibility.
3378 */
e2b911c5 3379 if (!ext4_has_feature_extents(sb))
b5799018 3380 return;
27dd4385
LC		 3381 /*
3382 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3383 * This should cover the situations where we can not afford to run
3384 * out of space like for example punch hole, or converting
556615dc 3385 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3386 * allocation would require 1, or 2 blocks, higher numbers are
3387 * very rare.
3388 */
b5799018
TT		 3389 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3390 sbi->s_cluster_bits);
27dd4385
LC		 3391
3392 do_div(resv_clusters, 50);
3393 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3394
b5799018 3395 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3396}
3397
2b2d6d01 3398static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3399{
5e405595 3400 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
d4c402d9 3401 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3402 struct buffer_head *bh;
617ba13b 3403 struct ext4_super_block *es = NULL;
5aee0f8a 3404 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
617ba13b
MC		 3405 ext4_fsblk_t block;
3406 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3407 ext4_fsblk_t logical_sb_block;
ac27a0ec 3408 unsigned long offset = 0;
ac27a0ec
DK		 3409 unsigned long journal_devnum = 0;
3410 unsigned long def_mount_opts;
3411 struct inode *root;
0390131b 3412 const char *descr;
dcc7dae3 3413 int ret = -ENOMEM;
281b5995 3414 int blocksize, clustersize;
4ec11028
TT		 3415 unsigned int db_count;
3416 unsigned int i;
281b5995 3417 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3418 __u64 blocks_count;
07aa2ea1 3419 int err = 0;
b3881f74 3420 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3421 ext4_group_t first_not_zeroed;
ac27a0ec 3422
5aee0f8a
TT		 3423 if ((data && !orig_data) || !sbi)
3424 goto out_free_base;
705895b6 3425
aed9eb1b 3426 sbi->s_daxdev = dax_dev;
705895b6
PE		 3427 sbi->s_blockgroup_lock =
3428 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 3429 if (!sbi->s_blockgroup_lock)
3430 goto out_free_base;
3431
ac27a0ec 3432 sb->s_fs_info = sbi;
2c0544b2 3433 sbi->s_sb = sb;
240799cd 3434 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3435 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3436 if (sb->s_bdev->bd_part)
3437 sbi->s_sectors_written_start =
3438 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3439
9f6200bb 3440 /* Cleanup superblock name */
ec3904dc 3441 strreplace(sb->s_id, '/', '!');
9f6200bb 3442
07aa2ea1 3443 /* -EINVAL is default */
dcc7dae3 3444 ret = -EINVAL;
617ba13b 3445 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3446 if (!blocksize) {
b31e1552 3447 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3448 goto out_fail;
3449 }
3450
3451 /*
617ba13b 3452 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3453 * block sizes. We need to calculate the offset from buffer start.
3454 */
617ba13b 3455 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3456 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3457 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3458 } else {
70bbb3e0 3459 logical_sb_block = sb_block;
ac27a0ec
DK		 3460 }
3461
a8ac900b 3462 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3463 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3464 goto out_fail;
3465 }
3466 /*
3467 * Note: s_es must be initialized as soon as possible because
617ba13b 3468 * some ext4 macro-instructions depend on its value
ac27a0ec 3469 */
2716b802 3470 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3471 sbi->s_es = es;
3472 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3473 if (sb->s_magic != EXT4_SUPER_MAGIC)
3474 goto cantfind_ext4;
afc32f7e 3475 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3476
feb0ab32 3477 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3478 if (ext4_has_feature_metadata_csum(sb) &&
3479 ext4_has_feature_gdt_csum(sb))
363307e6 3480 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3481 "redundant flags; please run fsck.");
3482
d25425f8
DW		 3483 /* Check for a known checksum algorithm */
3484 if (!ext4_verify_csum_type(sb, es)) {
3485 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3486 "unknown checksum algorithm.");
3487 silent = 1;
3488 goto cantfind_ext4;
3489 }
3490
0441984a 3491 /* Load the checksum driver */
dec214d0
TE		 3492 if (ext4_has_feature_metadata_csum(sb) ||
3493 ext4_has_feature_ea_inode(sb)) {
0441984a
DW		 3494 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3495 if (IS_ERR(sbi->s_chksum_driver)) {
3496 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3497 ret = PTR_ERR(sbi->s_chksum_driver);
3498 sbi->s_chksum_driver = NULL;
3499 goto failed_mount;
3500 }
3501 }
3502
a9c47317
DW		 3503 /* Check superblock checksum */
3504 if (!ext4_superblock_csum_verify(sb, es)) {
3505 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3506 "invalid superblock checksum. Run e2fsck?");
3507 silent = 1;
6a797d27 3508 ret = -EFSBADCRC;
a9c47317
DW		 3509 goto cantfind_ext4;
3510 }
3511
3512 /* Precompute checksum seed for all metadata */
e2b911c5 3513 if (ext4_has_feature_csum_seed(sb))
8c81bd8f 3514 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
dec214d0 3515 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
a9c47317
DW		 3516 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3517 sizeof(es->s_uuid));
3518
ac27a0ec
DK		 3519 /* Set defaults before we parse the mount options */
3520 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3521 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3522 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3523 set_opt(sb, DEBUG);
87f26807 3524 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3525 set_opt(sb, GRPID);
617ba13b 3526 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3527 set_opt(sb, NO_UID32);
ea663336 3528 /* xattr user namespace & acls are now defaulted on */
ea663336 3529 set_opt(sb, XATTR_USER);
03010a33 3530#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3531 set_opt(sb, POSIX_ACL);
2e7842b8 3532#endif
98c1a759
DW		 3533 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3534 if (ext4_has_metadata_csum(sb))
3535 set_opt(sb, JOURNAL_CHECKSUM);
3536
617ba13b 3537 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3538 set_opt(sb, JOURNAL_DATA);
617ba13b 3539 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3540 set_opt(sb, ORDERED_DATA);
617ba13b 3541 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3542 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3543
3544 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3545 set_opt(sb, ERRORS_PANIC);
bb4f397a 3546 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3547 set_opt(sb, ERRORS_CONT);
bb4f397a 3548 else
fd8c37ec 3549 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3550 /* block_validity enabled by default; disable with noblock_validity */
3551 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3552 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3553 set_opt(sb, DISCARD);
ac27a0ec 3554
08cefc7a
EB		 3555 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3556 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3557 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3558 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3559 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3560
8b67f04a 3561 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3562 set_opt(sb, BARRIER);
ac27a0ec 3563
dd919b98
AK		 3564 /*
3565 * enable delayed allocation by default
3566 * Use -o nodelalloc to turn it off
3567 */
bc0b75f7 3568 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3569 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3570 set_opt(sb, DELALLOC);
dd919b98 3571
51ce6511
LC		 3572 /*
3573 * set default s_li_wait_mult for lazyinit, for the case there is
3574 * no mount option specified.
3575 */
3576 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3577
5aee0f8a
TT		 3578 if (sbi->s_es->s_mount_opts[0]) {
3579 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3580 sizeof(sbi->s_es->s_mount_opts),
3581 GFP_KERNEL);
3582 if (!s_mount_opts)
3583 goto failed_mount;
3584 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3585 &journal_ioprio, 0)) {
3586 ext4_msg(sb, KERN_WARNING,
3587 "failed to parse options in superblock: %s",
3588 s_mount_opts);
3589 }
3590 kfree(s_mount_opts);
8b67f04a 3591 }
5a916be1 3592 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3593 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3594 &journal_ioprio, 0))
ac27a0ec
DK		 3595 goto failed_mount;
3596
56889787
TT		 3597 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3598 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3599 "with data=journal disables delayed "
3600 "allocation and O_DIRECT support!\n");
3601 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3602 ext4_msg(sb, KERN_ERR, "can't mount with "
3603 "both data=journal and delalloc");
3604 goto failed_mount;
3605 }
3606 if (test_opt(sb, DIOREAD_NOLOCK)) {
3607 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3608 "both data=journal and dioread_nolock");
56889787
TT		 3609 goto failed_mount;
3610 }
923ae0ff
RZ		 3611 if (test_opt(sb, DAX)) {
3612 ext4_msg(sb, KERN_ERR, "can't mount with "
3613 "both data=journal and dax");
3614 goto failed_mount;
3615 }
73b92a2a
SK		 3616 if (ext4_has_feature_encrypt(sb)) {
3617 ext4_msg(sb, KERN_WARNING,
3618 "encrypted files will use data=ordered "
3619 "instead of data journaling mode");
3620 }
56889787
TT		 3621 if (test_opt(sb, DELALLOC))
3622 clear_opt(sb, DELALLOC);
001e4a87
TH		 3623 } else {
3624 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3625 }
3626
ac27a0ec 3627 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3628 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3629
617ba13b 3630 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3631 (ext4_has_compat_features(sb) ||
3632 ext4_has_ro_compat_features(sb) ||
3633 ext4_has_incompat_features(sb)))
b31e1552
ES		 3634 ext4_msg(sb, KERN_WARNING,
3635 "feature flags set on rev 0 fs, "
3636 "running e2fsck is recommended");
469108ff 3637
ed3654eb
TT		 3638 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3639 set_opt2(sb, HURD_COMPAT);
e2b911c5 3640 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3641 ext4_msg(sb, KERN_ERR,
3642 "The Hurd can't support 64-bit file systems");
3643 goto failed_mount;
3644 }
dec214d0
TE		 3645
3646 /*
3647 * ea_inode feature uses l_i_version field which is not
3648 * available in HURD_COMPAT mode.
3649 */
3650 if (ext4_has_feature_ea_inode(sb)) {
3651 ext4_msg(sb, KERN_ERR,
3652 "ea_inode feature is not supported for Hurd");
3653 goto failed_mount;
3654 }
ed3654eb
TT		 3655 }
3656
2035e776
TT		 3657 if (IS_EXT2_SB(sb)) {
3658 if (ext2_feature_set_ok(sb))
3659 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3660 "using the ext4 subsystem");
3661 else {
3662 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3663 "to feature incompatibilities");
3664 goto failed_mount;
3665 }
3666 }
3667
3668 if (IS_EXT3_SB(sb)) {
3669 if (ext3_feature_set_ok(sb))
3670 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3671 "using the ext4 subsystem");
3672 else {
3673 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3674 "to feature incompatibilities");
3675 goto failed_mount;
3676 }
3677 }
3678
ac27a0ec
DK		 3679 /*
3680 * Check feature flags regardless of the revision level, since we
3681 * previously didn't change the revision level when setting the flags,
3682 * so there is a chance incompat flags are set on a rev 0 filesystem.
3683 */
bc98a42c 3684 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
ac27a0ec 3685 goto failed_mount;
a13fb1a4 3686
261cb20c 3687 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3688 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3689 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552 3690 ext4_msg(sb, KERN_ERR,
8cdf3372
TT		 3691 "Unsupported filesystem blocksize %d (%d log_block_size)",
3692 blocksize, le32_to_cpu(es->s_log_block_size));
3693 goto failed_mount;
3694 }
3695 if (le32_to_cpu(es->s_log_block_size) >
3696 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3697 ext4_msg(sb, KERN_ERR,
3698 "Invalid log block size: %u",
3699 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 3700 goto failed_mount;
3701 }
3702
5b9554dc
TT		 3703 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3704 ext4_msg(sb, KERN_ERR,
3705 "Number of reserved GDT blocks insanely large: %d",
3706 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3707 goto failed_mount;
3708 }
3709
923ae0ff 3710 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
559db4c6
RZ		 3711 if (ext4_has_feature_inline_data(sb)) {
3712 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3713 " that may contain inline data");
3714 goto failed_mount;
3715 }
87eefeb4
TK		 3716 err = bdev_dax_supported(sb, blocksize);
3717 if (err)
923ae0ff 3718 goto failed_mount;
923ae0ff
RZ		 3719 }
3720
e2b911c5 3721 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 3722 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3723 es->s_encryption_level);
3724 goto failed_mount;
3725 }
3726
ac27a0ec 3727 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3728 /* Validate the filesystem blocksize */
3729 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3730 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3731 blocksize);
ac27a0ec
DK		 3732 goto failed_mount;
3733 }
3734
2b2d6d01 3735 brelse(bh);
70bbb3e0
AM		 3736 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3737 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3738 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3739 if (!bh) {
b31e1552
ES		 3740 ext4_msg(sb, KERN_ERR,
3741 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3742 goto failed_mount;
3743 }
2716b802 3744 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3745 sbi->s_es = es;
617ba13b 3746 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3747 ext4_msg(sb, KERN_ERR,
3748 "Magic mismatch, very weird!");
ac27a0ec
DK		 3749 goto failed_mount;
3750 }
3751 }
3752
e2b911c5 3753 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 3754 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3755 has_huge_files);
3756 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3757
617ba13b
MC		 3758 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3759 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3760 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3761 } else {
3762 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3763 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3764 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3765 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3766 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3767 ext4_msg(sb, KERN_ERR,
3768 "unsupported inode size: %d",
2b2d6d01 3769 sbi->s_inode_size);
ac27a0ec
DK		 3770 goto failed_mount;
3771 }
ef7f3835
KS		 3772 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3773 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3774 }
0b8e58a1 3775
0d1ee42f 3776 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 3777 if (ext4_has_feature_64bit(sb)) {
8fadc143 3778 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3779 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3780 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3781 ext4_msg(sb, KERN_ERR,
3782 "unsupported descriptor size %lu",
0d1ee42f
AR		 3783 sbi->s_desc_size);
3784 goto failed_mount;
3785 }
3786 } else
3787 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3788
ac27a0ec 3789 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3790 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 3791
617ba13b 3792 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3793 if (sbi->s_inodes_per_block == 0)
617ba13b 3794 goto cantfind_ext4;
cd6bb35b
TT		 3795 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3796 sbi->s_inodes_per_group > blocksize * 8) {
3797 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3798 sbi->s_blocks_per_group);
3799 goto failed_mount;
3800 }
ac27a0ec
DK		 3801 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3802 sbi->s_inodes_per_block;
0d1ee42f 3803 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3804 sbi->s_sbh = bh;
3805 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3806 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3807 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3808
2b2d6d01 3809 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3810 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3811 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 3812 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 3813 i = le32_to_cpu(es->s_flags);
3814 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3815 sbi->s_hash_unsigned = 3;
3816 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3817#ifdef __CHAR_UNSIGNED__
bc98a42c 3818 if (!sb_rdonly(sb))
23301410
TT		 3819 es->s_flags |=
3820 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3821 sbi->s_hash_unsigned = 3;
f99b2589 3822#else
bc98a42c 3823 if (!sb_rdonly(sb))
23301410
TT		 3824 es->s_flags |=
3825 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3826#endif
23301410 3827 }
f99b2589 3828 }
ac27a0ec 3829
281b5995
TT		 3830 /* Handle clustersize */
3831 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 3832 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 3833 if (has_bigalloc) {
3834 if (clustersize < blocksize) {
3835 ext4_msg(sb, KERN_ERR,
3836 "cluster size (%d) smaller than "
3837 "block size (%d)", clustersize, blocksize);
3838 goto failed_mount;
3839 }
8cdf3372
TT		 3840 if (le32_to_cpu(es->s_log_cluster_size) >
3841 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3842 ext4_msg(sb, KERN_ERR,
3843 "Invalid log cluster size: %u",
3844 le32_to_cpu(es->s_log_cluster_size));
3845 goto failed_mount;
3846 }
281b5995
TT		 3847 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3848 le32_to_cpu(es->s_log_block_size);
3849 sbi->s_clusters_per_group =
3850 le32_to_cpu(es->s_clusters_per_group);
3851 if (sbi->s_clusters_per_group > blocksize * 8) {
3852 ext4_msg(sb, KERN_ERR,
3853 "#clusters per group too big: %lu",
3854 sbi->s_clusters_per_group);
3855 goto failed_mount;
3856 }
3857 if (sbi->s_blocks_per_group !=
3858 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3859 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3860 "clusters per group (%lu) inconsistent",
3861 sbi->s_blocks_per_group,
3862 sbi->s_clusters_per_group);
3863 goto failed_mount;
3864 }
3865 } else {
3866 if (clustersize != blocksize) {
3867 ext4_warning(sb, "fragment/cluster size (%d) != "
3868 "block size (%d)", clustersize,
3869 blocksize);
3870 clustersize = blocksize;
3871 }
3872 if (sbi->s_blocks_per_group > blocksize * 8) {
3873 ext4_msg(sb, KERN_ERR,
3874 "#blocks per group too big: %lu",
3875 sbi->s_blocks_per_group);
3876 goto failed_mount;
3877 }
3878 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3879 sbi->s_cluster_bits = 0;
ac27a0ec 3880 }
281b5995
TT		 3881 sbi->s_cluster_ratio = clustersize / blocksize;
3882
960fd856
TT		 3883 /* Do we have standard group size of clustersize * 8 blocks ? */
3884 if (sbi->s_blocks_per_group == clustersize << 3)
3885 set_opt2(sb, STD_GROUP_SIZE);
3886
bf43d84b
ES		 3887 /*
3888 * Test whether we have more sectors than will fit in sector_t,
3889 * and whether the max offset is addressable by the page cache.
3890 */
5a9ae68a 3891 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3892 ext4_blocks_count(es));
5a9ae68a 3893 if (err) {
b31e1552 3894 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3895 " too large to mount safely on this system");
ac27a0ec 3896 if (sizeof(sector_t) < 8)
90c699a9 3897 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3898 goto failed_mount;
3899 }
3900
617ba13b
MC		 3901 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3902 goto cantfind_ext4;
e7c95593 3903
0f2ddca6
FTN		 3904 /* check blocks count against device size */
3905 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3906 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3907 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3908 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3909 ext4_blocks_count(es), blocks_count);
3910 goto failed_mount;
3911 }
3912
0b8e58a1
AD		 3913 /*
3914 * It makes no sense for the first data block to be beyond the end
3915 * of the filesystem.
3916 */
3917 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3918 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3919 "block %u is beyond end of filesystem (%llu)",
3920 le32_to_cpu(es->s_first_data_block),
3921 ext4_blocks_count(es));
e7c95593
ES		 3922 goto failed_mount;
3923 }
bd81d8ee
LV		 3924 blocks_count = (ext4_blocks_count(es) -
3925 le32_to_cpu(es->s_first_data_block) +
3926 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3927 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3928 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3929 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3930 "(block count %llu, first data block %u, "
b31e1552 3931 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3932 ext4_blocks_count(es),
3933 le32_to_cpu(es->s_first_data_block),
3934 EXT4_BLOCKS_PER_GROUP(sb));
3935 goto failed_mount;
3936 }
bd81d8ee 3937 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3938 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3939 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3940 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3941 EXT4_DESC_PER_BLOCK(sb);
3a4b77cd 3942 if (ext4_has_feature_meta_bg(sb)) {
2ba3e6e8 3943 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3a4b77cd
EG		 3944 ext4_msg(sb, KERN_WARNING,
3945 "first meta block group too large: %u "
3946 "(group descriptor block count %u)",
3947 le32_to_cpu(es->s_first_meta_bg), db_count);
3948 goto failed_mount;
3949 }
3950 }
a7c3e901 3951 sbi->s_group_desc = kvmalloc(db_count *
f18a5f21
TT		 3952 sizeof(struct buffer_head *),
3953 GFP_KERNEL);
ac27a0ec 3954 if (sbi->s_group_desc == NULL) {
b31e1552 3955 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3956 ret = -ENOMEM;
ac27a0ec
DK		 3957 goto failed_mount;
3958 }
3959
705895b6 3960 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec 3961
85c8f176
AP		 3962 /* Pre-read the descriptors into the buffer cache */
3963 for (i = 0; i < db_count; i++) {
3964 block = descriptor_loc(sb, logical_sb_block, i);
3965 sb_breadahead(sb, block);
3966 }
3967
ac27a0ec 3968 for (i = 0; i < db_count; i++) {
70bbb3e0 3969 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3970 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3971 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3972 ext4_msg(sb, KERN_ERR,
3973 "can't read group descriptor %d", i);
ac27a0ec
DK		 3974 db_count = i;
3975 goto failed_mount2;
3976 }
3977 }
829fa70d 3978 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 3979 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 3980 ret = -EFSCORRUPTED;
f9ae9cf5 3981 goto failed_mount2;
ac27a0ec 3982 }
772cb7c8 3983
f9ae9cf5 3984 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3985 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3986 spin_lock_init(&sbi->s_next_gen_lock);
3987
04ecddb7
JM		 3988 setup_timer(&sbi->s_err_report, print_daily_error_info,
3989 (unsigned long) sb);
04496411 3990
a75ae78f 3991 /* Register extent status tree shrinker */
eb68d0e2 3992 if (ext4_es_register_shrinker(sbi))
ce7e010a 3993 goto failed_mount3;
ce7e010a 3994
c9de560d 3995 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3996 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3997
f9ae9cf5
TT		 3998 /*
3999 * set up enough so that it can read an inode
4000 */
f6e63f90 4001 sb->s_op = &ext4_sops;
617ba13b
MC		 4002 sb->s_export_op = &ext4_export_ops;
4003 sb->s_xattr = ext4_xattr_handlers;
a7550b30 4004 sb->s_cop = &ext4_cryptops;
ac27a0ec 4005#ifdef CONFIG_QUOTA
617ba13b 4006 sb->dq_op = &ext4_quota_operations;
e2b911c5 4007 if (ext4_has_feature_quota(sb))
1fa5efe3 4008 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 4009 else
4010 sb->s_qcop = &ext4_qctl_operations;
689c958c 4011 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 4012#endif
85787090 4013 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
f2fa2ffc 4014
ac27a0ec 4015 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 4016 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 4017
4018 sb->s_root = NULL;
4019
4020 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 4021 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 4022
bc98a42c 4023 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
c5e06d10 4024 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 4025 goto failed_mount3a;
c5e06d10 4026
ac27a0ec
DK		 4027 /*
4028 * The first inode we look at is the journal inode. Don't try
4029 * root first: it may be modified in the journal!
4030 */
e2b911c5 4031 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4753d8a2
TT		 4032 err = ext4_load_journal(sb, es, journal_devnum);
4033 if (err)
50460fe8 4034 goto failed_mount3a;
bc98a42c 4035 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
e2b911c5 4036 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 4037 ext4_msg(sb, KERN_ERR, "required journal recovery "
4038 "suppressed and not mounted read-only");
744692dc 4039 goto failed_mount_wq;
ac27a0ec 4040 } else {
1e381f60
DM		 4041 /* Nojournal mode, all journal mount options are illegal */
4042 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4043 ext4_msg(sb, KERN_ERR, "can't mount with "
4044 "journal_checksum, fs mounted w/o journal");
4045 goto failed_mount_wq;
4046 }
4047 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4048 ext4_msg(sb, KERN_ERR, "can't mount with "
4049 "journal_async_commit, fs mounted w/o journal");
4050 goto failed_mount_wq;
4051 }
4052 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4053 ext4_msg(sb, KERN_ERR, "can't mount with "
4054 "commit=%lu, fs mounted w/o journal",
4055 sbi->s_commit_interval / HZ);
4056 goto failed_mount_wq;
4057 }
4058 if (EXT4_MOUNT_DATA_FLAGS &
4059 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4060 ext4_msg(sb, KERN_ERR, "can't mount with "
4061 "data=, fs mounted w/o journal");
4062 goto failed_mount_wq;
4063 }
4064 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4065 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 4066 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 4067 sbi->s_journal = NULL;
4068 needs_recovery = 0;
4069 goto no_journal;
ac27a0ec
DK		 4070 }
4071
e2b911c5 4072 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 4073 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4074 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4075 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4076 goto failed_mount_wq;
eb40a09c
JS		 4077 }
4078
25ed6e8a
DW		 4079 if (!set_journal_csum_feature_set(sb)) {
4080 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4081 "feature set");
4082 goto failed_mount_wq;
d4da6c9c 4083 }
818d276c 4084
ac27a0ec
DK		 4085 /* We have now updated the journal if required, so we can
4086 * validate the data journaling mode. */
4087 switch (test_opt(sb, DATA_FLAGS)) {
4088 case 0:
4089 /* No mode set, assume a default based on the journal
63f57933
AM		 4090 * capabilities: ORDERED_DATA if the journal can
4091 * cope, else JOURNAL_DATA
4092 */
dab291af
MC		 4093 if (jbd2_journal_check_available_features
4094 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 4095 set_opt(sb, ORDERED_DATA);
ac27a0ec 4096 else
fd8c37ec 4097 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 4098 break;
4099
617ba13b
MC		 4100 case EXT4_MOUNT_ORDERED_DATA:
4101 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4102 if (!jbd2_journal_check_available_features
4103 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4104 ext4_msg(sb, KERN_ERR, "Journal does not support "
4105 "requested data journaling mode");
744692dc 4106 goto failed_mount_wq;
ac27a0ec
DK		 4107 }
4108 default:
4109 break;
4110 }
ab04df78
JK		 4111
4112 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4113 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4114 ext4_msg(sb, KERN_ERR, "can't mount with "
4115 "journal_async_commit in data=ordered mode");
4116 goto failed_mount_wq;
4117 }
4118
b3881f74 4119 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4120
18aadd47
BJ		 4121 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4122
ce7e010a 4123no_journal:
cdb7ee4c
TE		 4124 if (!test_opt(sb, NO_MBCACHE)) {
4125 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4126 if (!sbi->s_ea_block_cache) {
dec214d0 4127 ext4_msg(sb, KERN_ERR,
cdb7ee4c 4128 "Failed to create ea_block_cache");
dec214d0
TE		 4129 goto failed_mount_wq;
4130 }
cdb7ee4c
TE		 4131
4132 if (ext4_has_feature_ea_inode(sb)) {
4133 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4134 if (!sbi->s_ea_inode_cache) {
4135 ext4_msg(sb, KERN_ERR,
4136 "Failed to create ea_inode_cache");
4137 goto failed_mount_wq;
4138 }
4139 }
9c191f70
M		 4140 }
4141
e2b911c5 4142 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
09cbfeaf 4143 (blocksize != PAGE_SIZE)) {
1cb767cd
TT		 4144 ext4_msg(sb, KERN_ERR,
4145 "Unsupported blocksize for fs encryption");
4146 goto failed_mount_wq;
4147 }
4148
bc98a42c 4149 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
e2b911c5
DW		 4150 !ext4_has_feature_encrypt(sb)) {
4151 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 4152 ext4_commit_super(sb, 1);
4153 }
4154
952fc18e
TT		 4155 /*
4156 * Get the # of file system overhead blocks from the
4157 * superblock if present.
4158 */
4159 if (es->s_overhead_clusters)
4160 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4161 else {
07aa2ea1
LC		 4162 err = ext4_calculate_overhead(sb);
4163 if (err)
952fc18e
TT		 4164 goto failed_mount_wq;
4165 }
4166
fd89d5f2
TH		 4167 /*
4168 * The maximum number of concurrent works can be high and
4169 * concurrency isn't really necessary. Limit it to 1.
4170 */
2e8fa54e
JK		 4171 EXT4_SB(sb)->rsv_conversion_wq =
4172 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4173 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4174 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4175 ret = -ENOMEM;
2e8fa54e
JK		 4176 goto failed_mount4;
4177 }
4178
ac27a0ec 4179 /*
dab291af 4180 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4181 * so we can safely mount the rest of the filesystem now.
4182 */
4183
1d1fe1ee
DH		 4184 root = ext4_iget(sb, EXT4_ROOT_INO);
4185 if (IS_ERR(root)) {
b31e1552 4186 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4187 ret = PTR_ERR(root);
32a9bb57 4188 root = NULL;
ac27a0ec
DK		 4189 goto failed_mount4;
4190 }
4191 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4192 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4193 iput(root);
ac27a0ec
DK		 4194 goto failed_mount4;
4195 }
48fde701 4196 sb->s_root = d_make_root(root);
1d1fe1ee 4197 if (!sb->s_root) {
b31e1552 4198 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4199 ret = -ENOMEM;
4200 goto failed_mount4;
4201 }
ac27a0ec 4202
bc98a42c 4203 if (ext4_setup_super(sb, es, sb_rdonly(sb)))
7e84b621 4204 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4205
4206 /* determine the minimum size of new large inodes, if present */
670e9875
TT		 4207 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4208 sbi->s_want_extra_isize == 0) {
ef7f3835
KS		 4209 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4210 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 4211 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 4212 if (sbi->s_want_extra_isize <
4213 le16_to_cpu(es->s_want_extra_isize))
4214 sbi->s_want_extra_isize =
4215 le16_to_cpu(es->s_want_extra_isize);
4216 if (sbi->s_want_extra_isize <
4217 le16_to_cpu(es->s_min_extra_isize))
4218 sbi->s_want_extra_isize =
4219 le16_to_cpu(es->s_min_extra_isize);
4220 }
4221 }
4222 /* Check if enough inode space is available */
4223 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4224 sbi->s_inode_size) {
4225 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4226 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4227 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4228 "available");
ef7f3835
KS		 4229 }
4230
b5799018 4231 ext4_set_resv_clusters(sb);
27dd4385 4232
6fd058f7
TT		 4233 err = ext4_setup_system_zone(sb);
4234 if (err) {
b31e1552 4235 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4236 "zone (%d)", err);
f9ae9cf5
TT		 4237 goto failed_mount4a;
4238 }
4239
4240 ext4_ext_init(sb);
4241 err = ext4_mb_init(sb);
4242 if (err) {
4243 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4244 err);
dcf2d804 4245 goto failed_mount5;
c2774d84
AK		 4246 }
4247
d5e03cbb
TT		 4248 block = ext4_count_free_clusters(sb);
4249 ext4_free_blocks_count_set(sbi->s_es,
4250 EXT4_C2B(sbi, block));
908c7f19
TH		 4251 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4252 GFP_KERNEL);
d5e03cbb
TT		 4253 if (!err) {
4254 unsigned long freei = ext4_count_free_inodes(sb);
4255 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4256 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4257 GFP_KERNEL);
d5e03cbb
TT		 4258 }
4259 if (!err)
4260 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4261 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4262 if (!err)
908c7f19
TH		 4263 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4264 GFP_KERNEL);
c8585c6f
DJ		 4265 if (!err)
4266 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4267
d5e03cbb
TT		 4268 if (err) {
4269 ext4_msg(sb, KERN_ERR, "insufficient memory");
4270 goto failed_mount6;
4271 }
4272
e2b911c5 4273 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 4274 if (!ext4_fill_flex_info(sb)) {
4275 ext4_msg(sb, KERN_ERR,
4276 "unable to initialize "
4277 "flex_bg meta info!");
4278 goto failed_mount6;
4279 }
4280
bfff6873
LC		 4281 err = ext4_register_li_request(sb, first_not_zeroed);
4282 if (err)
dcf2d804 4283 goto failed_mount6;
bfff6873 4284
b5799018 4285 err = ext4_register_sysfs(sb);
dcf2d804
TM		 4286 if (err)
4287 goto failed_mount7;
3197ebdb 4288
9b2ff357
JK		 4289#ifdef CONFIG_QUOTA
4290 /* Enable quota usage during mount. */
bc98a42c 4291 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
9b2ff357
JK		 4292 err = ext4_enable_quotas(sb);
4293 if (err)
4294 goto failed_mount8;
4295 }
4296#endif /* CONFIG_QUOTA */
4297
617ba13b
MC		 4298 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4299 ext4_orphan_cleanup(sb, es);
4300 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4301 if (needs_recovery) {
b31e1552 4302 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4303 ext4_mark_recovery_complete(sb, es);
4304 }
4305 if (EXT4_SB(sb)->s_journal) {
4306 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4307 descr = " journalled data mode";
4308 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4309 descr = " ordered data mode";
4310 else
4311 descr = " writeback data mode";
4312 } else
4313 descr = "out journal";
4314
79add3a3
LC		 4315 if (test_opt(sb, DISCARD)) {
4316 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4317 if (!blk_queue_discard(q))
4318 ext4_msg(sb, KERN_WARNING,
4319 "mounting with \"discard\" option, but "
4320 "the device does not support discard");
4321 }
4322
e294a537
TT		 4323 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4324 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 4325 "Opts: %.*s%s%s", descr,
4326 (int) sizeof(sbi->s_es->s_mount_opts),
4327 sbi->s_es->s_mount_opts,
e294a537 4328 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4329
66e61a9e
TT		 4330 if (es->s_error_count)
4331 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4332
efbed4dc
TT		 4333 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4334 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4335 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4336 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4337
d4c402d9 4338 kfree(orig_data);
ac27a0ec
DK		 4339 return 0;
4340
617ba13b 4341cantfind_ext4:
ac27a0ec 4342 if (!silent)
b31e1552 4343 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4344 goto failed_mount;
4345
72ba7450
TT		 4346#ifdef CONFIG_QUOTA
4347failed_mount8:
ebd173be 4348 ext4_unregister_sysfs(sb);
72ba7450 4349#endif
dcf2d804
TM		 4350failed_mount7:
4351 ext4_unregister_li_request(sb);
4352failed_mount6:
f9ae9cf5 4353 ext4_mb_release(sb);
d5e03cbb 4354 if (sbi->s_flex_groups)
b93b41d4 4355 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4356 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4357 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4358 percpu_counter_destroy(&sbi->s_dirs_counter);
4359 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4360failed_mount5:
f9ae9cf5
TT		 4361 ext4_ext_release(sb);
4362 ext4_release_system_zone(sb);
4363failed_mount4a:
94bf608a 4364 dput(sb->s_root);
32a9bb57 4365 sb->s_root = NULL;
94bf608a 4366failed_mount4:
b31e1552 4367 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4368 if (EXT4_SB(sb)->rsv_conversion_wq)
4369 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4370failed_mount_wq:
dec214d0
TE		 4371 if (sbi->s_ea_inode_cache) {
4372 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4373 sbi->s_ea_inode_cache = NULL;
4374 }
47387409
TE		 4375 if (sbi->s_ea_block_cache) {
4376 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4377 sbi->s_ea_block_cache = NULL;
82939d79 4378 }
0390131b
FM		 4379 if (sbi->s_journal) {
4380 jbd2_journal_destroy(sbi->s_journal);
4381 sbi->s_journal = NULL;
4382 }
50460fe8 4383failed_mount3a:
d3922a77 4384 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4385failed_mount3:
9105bb14 4386 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4387 if (sbi->s_mmp_tsk)
4388 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4389failed_mount2:
4390 for (i = 0; i < db_count; i++)
4391 brelse(sbi->s_group_desc[i]);
b93b41d4 4392 kvfree(sbi->s_group_desc);
ac27a0ec 4393failed_mount:
0441984a
DW		 4394 if (sbi->s_chksum_driver)
4395 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 4396#ifdef CONFIG_QUOTA
a2d4a646 4397 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4398 kfree(sbi->s_qf_names[i]);
4399#endif
617ba13b 4400 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4401 brelse(bh);
4402out_fail:
4403 sb->s_fs_info = NULL;
f6830165 4404 kfree(sbi->s_blockgroup_lock);
5aee0f8a 4405out_free_base:
ac27a0ec 4406 kfree(sbi);
d4c402d9 4407 kfree(orig_data);
5e405595 4408 fs_put_dax(dax_dev);
07aa2ea1 4409 return err ? err : ret;
ac27a0ec
DK		 4410}
4411
4412/*
4413 * Setup any per-fs journal parameters now. We'll do this both on
4414 * initial mount, once the journal has been initialised but before we've
4415 * done any recovery; and again on any subsequent remount.
4416 */
617ba13b 4417static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4418{
617ba13b 4419 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4420
30773840
TT		 4421 journal->j_commit_interval = sbi->s_commit_interval;
4422 journal->j_min_batch_time = sbi->s_min_batch_time;
4423 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4424
a931da6a 4425 write_lock(&journal->j_state_lock);
ac27a0ec 4426 if (test_opt(sb, BARRIER))
dab291af 4427 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4428 else
dab291af 4429 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4430 if (test_opt(sb, DATA_ERR_ABORT))
4431 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4432 else
4433 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4434 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4435}
4436
c6cb7e77
EW		 4437static struct inode *ext4_get_journal_inode(struct super_block *sb,
4438 unsigned int journal_inum)
ac27a0ec
DK		 4439{
4440 struct inode *journal_inode;
ac27a0ec 4441
c6cb7e77
EW		 4442 /*
4443 * Test for the existence of a valid inode on disk. Bad things
4444 * happen if we iget() an unused inode, as the subsequent iput()
4445 * will try to delete it.
4446 */
1d1fe1ee
DH		 4447 journal_inode = ext4_iget(sb, journal_inum);
4448 if (IS_ERR(journal_inode)) {
b31e1552 4449 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4450 return NULL;
4451 }
4452 if (!journal_inode->i_nlink) {
4453 make_bad_inode(journal_inode);
4454 iput(journal_inode);
b31e1552 4455 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4456 return NULL;
4457 }
4458
e5f8eab8 4459 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4460 journal_inode, journal_inode->i_size);
1d1fe1ee 4461 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4462 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4463 iput(journal_inode);
4464 return NULL;
4465 }
c6cb7e77
EW		 4466 return journal_inode;
4467}
4468
4469static journal_t *ext4_get_journal(struct super_block *sb,
4470 unsigned int journal_inum)
4471{
4472 struct inode *journal_inode;
4473 journal_t *journal;
4474
4475 BUG_ON(!ext4_has_feature_journal(sb));
4476
4477 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4478 if (!journal_inode)
4479 return NULL;
ac27a0ec 4480
dab291af 4481 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4482 if (!journal) {
b31e1552 4483 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4484 iput(journal_inode);
4485 return NULL;
4486 }
4487 journal->j_private = sb;
617ba13b 4488 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4489 return journal;
4490}
4491
617ba13b 4492static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4493 dev_t j_dev)
4494{
2b2d6d01 4495 struct buffer_head *bh;
ac27a0ec 4496 journal_t *journal;
617ba13b
MC		 4497 ext4_fsblk_t start;
4498 ext4_fsblk_t len;
ac27a0ec 4499 int hblock, blocksize;
617ba13b 4500 ext4_fsblk_t sb_block;
ac27a0ec 4501 unsigned long offset;
2b2d6d01 4502 struct ext4_super_block *es;
ac27a0ec
DK		 4503 struct block_device *bdev;
4504
e2b911c5 4505 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4506
b31e1552 4507 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4508 if (bdev == NULL)
4509 return NULL;
4510
ac27a0ec 4511 blocksize = sb->s_blocksize;
e1defc4f 4512 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4513 if (blocksize < hblock) {
b31e1552
ES		 4514 ext4_msg(sb, KERN_ERR,
4515 "blocksize too small for journal device");
ac27a0ec
DK		 4516 goto out_bdev;
4517 }
4518
617ba13b
MC		 4519 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4520 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4521 set_blocksize(bdev, blocksize);
4522 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4523 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4524 "external journal");
ac27a0ec
DK		 4525 goto out_bdev;
4526 }
4527
2716b802 4528 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4529 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4530 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4531 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4532 ext4_msg(sb, KERN_ERR, "external journal has "
4533 "bad superblock");
ac27a0ec
DK		 4534 brelse(bh);
4535 goto out_bdev;
4536 }
4537
df4763be
DW		 4538 if ((le32_to_cpu(es->s_feature_ro_compat) &
4539 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4540 es->s_checksum != ext4_superblock_csum(sb, es)) {
4541 ext4_msg(sb, KERN_ERR, "external journal has "
4542 "corrupt superblock");
4543 brelse(bh);
4544 goto out_bdev;
4545 }
4546
617ba13b 4547 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4548 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4549 brelse(bh);
4550 goto out_bdev;
4551 }
4552
bd81d8ee 4553 len = ext4_blocks_count(es);
ac27a0ec
DK		 4554 start = sb_block + 1;
4555 brelse(bh); /* we're done with the superblock */
4556
dab291af 4557 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4558 start, len, blocksize);
4559 if (!journal) {
b31e1552 4560 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4561 goto out_bdev;
4562 }
4563 journal->j_private = sb;
dfec8a14 4564 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4565 wait_on_buffer(journal->j_sb_buffer);
4566 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4567 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4568 goto out_journal;
4569 }
4570 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4571 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4572 "user (unsupported) - %d",
ac27a0ec
DK		 4573 be32_to_cpu(journal->j_superblock->s_nr_users));
4574 goto out_journal;
4575 }
617ba13b
MC		 4576 EXT4_SB(sb)->journal_bdev = bdev;
4577 ext4_init_journal_params(sb, journal);
ac27a0ec 4578 return journal;
0b8e58a1 4579
ac27a0ec 4580out_journal:
dab291af 4581 jbd2_journal_destroy(journal);
ac27a0ec 4582out_bdev:
617ba13b 4583 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4584 return NULL;
4585}
4586
617ba13b
MC		 4587static int ext4_load_journal(struct super_block *sb,
4588 struct ext4_super_block *es,
ac27a0ec
DK		 4589 unsigned long journal_devnum)
4590{
4591 journal_t *journal;
4592 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4593 dev_t journal_dev;
4594 int err = 0;
4595 int really_read_only;
4596
e2b911c5 4597 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4598
ac27a0ec
DK		 4599 if (journal_devnum &&
4600 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4601 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4602 "numbers have changed");
ac27a0ec
DK		 4603 journal_dev = new_decode_dev(journal_devnum);
4604 } else
4605 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4606
4607 really_read_only = bdev_read_only(sb->s_bdev);
4608
4609 /*
4610 * Are we loading a blank journal or performing recovery after a
4611 * crash? For recovery, we need to check in advance whether we
4612 * can get read-write access to the device.
4613 */
e2b911c5 4614 if (ext4_has_feature_journal_needs_recovery(sb)) {
bc98a42c 4615 if (sb_rdonly(sb)) {
b31e1552
ES		 4616 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4617 "required on readonly filesystem");
ac27a0ec 4618 if (really_read_only) {
b31e1552
ES		 4619 ext4_msg(sb, KERN_ERR, "write access "
4620 "unavailable, cannot proceed");
ac27a0ec
DK		 4621 return -EROFS;
4622 }
b31e1552
ES		 4623 ext4_msg(sb, KERN_INFO, "write access will "
4624 "be enabled during recovery");
ac27a0ec
DK		 4625 }
4626 }
4627
4628 if (journal_inum && journal_dev) {
b31e1552
ES		 4629 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4630 "and inode journals!");
ac27a0ec
DK		 4631 return -EINVAL;
4632 }
4633
4634 if (journal_inum) {
617ba13b 4635 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4636 return -EINVAL;
4637 } else {
617ba13b 4638 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4639 return -EINVAL;
4640 }
4641
90576c0b 4642 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4643 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4644
e2b911c5 4645 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4646 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4647 if (!err) {
4648 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4649 if (save)
4650 memcpy(save, ((char *) es) +
4651 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4652 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4653 if (save)
4654 memcpy(((char *) es) + EXT4_S_ERR_START,
4655 save, EXT4_S_ERR_LEN);
4656 kfree(save);
4657 }
ac27a0ec
DK		 4658
4659 if (err) {
b31e1552 4660 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4661 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4662 return err;
4663 }
4664
617ba13b
MC		 4665 EXT4_SB(sb)->s_journal = journal;
4666 ext4_clear_journal_err(sb, es);
ac27a0ec 4667
c41303ce 4668 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4669 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4670 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4671
4672 /* Make sure we flush the recovery flag to disk. */
e2d67052 4673 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4674 }
4675
4676 return 0;
4677}
4678
e2d67052 4679static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4680{
e2d67052 4681 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4682 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4683 int error = 0;
ac27a0ec 4684
bdfe0cbd 4685 if (!sbh || block_device_ejected(sb))
c4be0c1d 4686 return error;
71290b36
TT		 4687 /*
4688 * If the file system is mounted read-only, don't update the
4689 * superblock write time. This avoids updating the superblock
4690 * write time when we are mounting the root file system
4691 * read/only but we need to replay the journal; at that point,
4692 * for people who are east of GMT and who make their clock
4693 * tick in localtime for Windows bug-for-bug compatibility,
4694 * the clock is set in the future, and this will cause e2fsck
4695 * to complain and force a full file system check.
4696 */
4697 if (!(sb->s_flags & MS_RDONLY))
4698 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4699 if (sb->s_bdev->bd_part)
4700 es->s_kbytes_written =
4701 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4702 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4703 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4704 else
4705 es->s_kbytes_written =
4706 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4707 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4708 ext4_free_blocks_count_set(es,
57042651
TT		 4709 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4710 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4711 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4712 es->s_free_inodes_count =
4713 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4714 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4715 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4716 ext4_superblock_csum_set(sb);
1566a48a
TT		 4717 if (sync)
4718 lock_buffer(sbh);
4743f839
PKS		 4719 if (buffer_write_io_error(sbh)) {
4720 /*
4721 * Oh, dear. A previous attempt to write the
4722 * superblock failed. This could happen because the
4723 * USB device was yanked out. Or it could happen to
4724 * be a transient write error and maybe the block will
4725 * be remapped. Nothing we can do but to retry the
4726 * write and hope for the best.
4727 */
4728 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4729 "superblock detected");
4730 clear_buffer_write_io_error(sbh);
4731 set_buffer_uptodate(sbh);
4732 }
ac27a0ec 4733 mark_buffer_dirty(sbh);
914258bf 4734 if (sync) {
1566a48a 4735 unlock_buffer(sbh);
564bc402 4736 error = __sync_dirty_buffer(sbh,
00473374 4737 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
c4be0c1d
TS		 4738 if (error)
4739 return error;
4740
4741 error = buffer_write_io_error(sbh);
4742 if (error) {
b31e1552
ES		 4743 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4744 "superblock");
914258bf
TT		 4745 clear_buffer_write_io_error(sbh);
4746 set_buffer_uptodate(sbh);
4747 }
4748 }
c4be0c1d 4749 return error;
ac27a0ec
DK		 4750}
4751
ac27a0ec
DK		 4752/*
4753 * Have we just finished recovery? If so, and if we are mounting (or
4754 * remounting) the filesystem readonly, then we will end up with a
4755 * consistent fs on disk. Record that fact.
4756 */
2b2d6d01
TT		 4757static void ext4_mark_recovery_complete(struct super_block *sb,
4758 struct ext4_super_block *es)
ac27a0ec 4759{
617ba13b 4760 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4761
e2b911c5 4762 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4763 BUG_ON(journal != NULL);
4764 return;
4765 }
dab291af 4766 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4767 if (jbd2_journal_flush(journal) < 0)
4768 goto out;
4769
bc98a42c 4770 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
e2b911c5 4771 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4772 ext4_commit_super(sb, 1);
ac27a0ec 4773 }
7ffe1ea8
HK		 4774
4775out:
dab291af 4776 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4777}
4778
4779/*
4780 * If we are mounting (or read-write remounting) a filesystem whose journal
4781 * has recorded an error from a previous lifetime, move that error to the
4782 * main filesystem now.
4783 */
2b2d6d01
TT		 4784static void ext4_clear_journal_err(struct super_block *sb,
4785 struct ext4_super_block *es)
ac27a0ec
DK		 4786{
4787 journal_t *journal;
4788 int j_errno;
4789 const char *errstr;
4790
e2b911c5 4791 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4792
617ba13b 4793 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4794
4795 /*
4796 * Now check for any error status which may have been recorded in the
617ba13b 4797 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4798 */
4799
dab291af 4800 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4801 if (j_errno) {
4802 char nbuf[16];
4803
617ba13b 4804 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4805 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4806 "from previous mount: %s", errstr);
12062ddd 4807 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4808
617ba13b
MC		 4809 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4810 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4811 ext4_commit_super(sb, 1);
ac27a0ec 4812
dab291af 4813 jbd2_journal_clear_err(journal);
d796c52e 4814 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4815 }
4816}
4817
4818/*
4819 * Force the running and committing transactions to commit,
4820 * and wait on the commit.
4821 */
617ba13b 4822int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4823{
4824 journal_t *journal;
ac27a0ec 4825
bc98a42c 4826 if (sb_rdonly(sb))
ac27a0ec
DK		 4827 return 0;
4828
617ba13b 4829 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4830 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4831}
4832
617ba13b 4833static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4834{
14ce0cb4 4835 int ret = 0;
9eddacf9 4836 tid_t target;
06a407f1 4837 bool needs_barrier = false;
8d5d02e6 4838 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4839
0db1ff22
TT		 4840 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
4841 return 0;
4842
9bffad1e 4843 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4844 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4845 /*
4846 * Writeback quota in non-journalled quota case - journalled quota has
4847 * no dirty dquots
4848 */
4849 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4850 /*
4851 * Data writeback is possible w/o journal transaction, so barrier must
4852 * being sent at the end of the function. But we can skip it if
4853 * transaction_commit will do it for us.
4854 */
bda32530
TT		 4855 if (sbi->s_journal) {
4856 target = jbd2_get_latest_transaction(sbi->s_journal);
4857 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4858 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4859 needs_barrier = true;
4860
4861 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4862 if (wait)
4863 ret = jbd2_log_wait_commit(sbi->s_journal,
4864 target);
4865 }
4866 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4867 needs_barrier = true;
06a407f1
DM		 4868 if (needs_barrier) {
4869 int err;
4870 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4871 if (!ret)
4872 ret = err;
0390131b 4873 }
06a407f1
DM		 4874
4875 return ret;
4876}
4877
ac27a0ec
DK		 4878/*
4879 * LVM calls this function before a (read-only) snapshot is created. This
4880 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4881 *
4882 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4883 * state independently. It relies on upper layer to stop all data & metadata
4884 * modifications.
ac27a0ec 4885 */
c4be0c1d 4886static int ext4_freeze(struct super_block *sb)
ac27a0ec 4887{
c4be0c1d
TS		 4888 int error = 0;
4889 journal_t *journal;
ac27a0ec 4890
bc98a42c 4891 if (sb_rdonly(sb))
9ca92389 4892 return 0;
ac27a0ec 4893
9ca92389 4894 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4895
bb044576
TT		 4896 if (journal) {
4897 /* Now we set up the journal barrier. */
4898 jbd2_journal_lock_updates(journal);
ac27a0ec 4899
bb044576
TT		 4900 /*
4901 * Don't clear the needs_recovery flag if we failed to
4902 * flush the journal.
4903 */
4904 error = jbd2_journal_flush(journal);
4905 if (error < 0)
4906 goto out;
c642dc9e
ES		 4907
4908 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 4909 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 4910 }
9ca92389 4911
9ca92389 4912 error = ext4_commit_super(sb, 1);
6b0310fb 4913out:
bb044576
TT		 4914 if (journal)
4915 /* we rely on upper layer to stop further updates */
4916 jbd2_journal_unlock_updates(journal);
6b0310fb 4917 return error;
ac27a0ec
DK		 4918}
4919
4920/*
4921 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4922 * flag here, even though the filesystem is not technically dirty yet.
4923 */
c4be0c1d 4924static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4925{
bc98a42c 4926 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 4927 return 0;
4928
c642dc9e
ES		 4929 if (EXT4_SB(sb)->s_journal) {
4930 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 4931 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 4932 }
4933
9ca92389 4934 ext4_commit_super(sb, 1);
c4be0c1d 4935 return 0;
ac27a0ec
DK		 4936}
4937
673c6100
TT		 4938/*
4939 * Structure to save mount options for ext4_remount's benefit
4940 */
4941struct ext4_mount_options {
4942 unsigned long s_mount_opt;
a2595b8a 4943 unsigned long s_mount_opt2;
08cefc7a
EB		 4944 kuid_t s_resuid;
4945 kgid_t s_resgid;
673c6100
TT		 4946 unsigned long s_commit_interval;
4947 u32 s_min_batch_time, s_max_batch_time;
4948#ifdef CONFIG_QUOTA
4949 int s_jquota_fmt;
a2d4a646 4950 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4951#endif
4952};
4953
2b2d6d01 4954static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4955{
2b2d6d01 4956 struct ext4_super_block *es;
617ba13b 4957 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4958 unsigned long old_sb_flags;
617ba13b 4959 struct ext4_mount_options old_opts;
c79d967d 4960 int enable_quota = 0;
8a266467 4961 ext4_group_t g;
b3881f74 4962 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4963 int err = 0;
ac27a0ec 4964#ifdef CONFIG_QUOTA
03dafb5f 4965 int i, j;
ac27a0ec 4966#endif
d4c402d9 4967 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4968
4969 /* Store the original options */
4970 old_sb_flags = sb->s_flags;
4971 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4972 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4973 old_opts.s_resuid = sbi->s_resuid;
4974 old_opts.s_resgid = sbi->s_resgid;
4975 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4976 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4977 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4978#ifdef CONFIG_QUOTA
4979 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4980 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4981 if (sbi->s_qf_names[i]) {
4982 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4983 GFP_KERNEL);
4984 if (!old_opts.s_qf_names[i]) {
4985 for (j = 0; j < i; j++)
4986 kfree(old_opts.s_qf_names[j]);
3e36a163 4987 kfree(orig_data);
03dafb5f
CG		 4988 return -ENOMEM;
4989 }
4990 } else
4991 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4992#endif
b3881f74
TT		 4993 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4994 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4995
661aa520 4996 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4997 err = -EINVAL;
4998 goto restore_opts;
4999 }
5000
6b992ff2 5001 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 5002 test_opt(sb, JOURNAL_CHECKSUM)) {
5003 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 5004 "during remount not supported; ignoring");
5005 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 5006 }
5007
6ae6514b
PS		 5008 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5009 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5010 ext4_msg(sb, KERN_ERR, "can't mount with "
5011 "both data=journal and delalloc");
5012 err = -EINVAL;
5013 goto restore_opts;
5014 }
5015 if (test_opt(sb, DIOREAD_NOLOCK)) {
5016 ext4_msg(sb, KERN_ERR, "can't mount with "
5017 "both data=journal and dioread_nolock");
5018 err = -EINVAL;
5019 goto restore_opts;
5020 }
923ae0ff
RZ		 5021 if (test_opt(sb, DAX)) {
5022 ext4_msg(sb, KERN_ERR, "can't mount with "
5023 "both data=journal and dax");
5024 err = -EINVAL;
5025 goto restore_opts;
5026 }
ab04df78
JK		 5027 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5028 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5029 ext4_msg(sb, KERN_ERR, "can't mount with "
5030 "journal_async_commit in data=ordered mode");
5031 err = -EINVAL;
5032 goto restore_opts;
5033 }
923ae0ff
RZ		 5034 }
5035
cdb7ee4c
TE		 5036 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5037 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5038 err = -EINVAL;
5039 goto restore_opts;
5040 }
5041
923ae0ff
RZ		 5042 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5043 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5044 "dax flag with busy inodes while remounting");
5045 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 5046 }
5047
4ab2f15b 5048 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 5049 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 5050
5051 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 5052 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 5053
5054 es = sbi->s_es;
5055
b3881f74 5056 if (sbi->s_journal) {
0390131b 5057 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 5058 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5059 }
ac27a0ec 5060
a2fd66d0
TT		 5061 if (*flags & MS_LAZYTIME)
5062 sb->s_flags |= MS_LAZYTIME;
5063
bc98a42c 5064 if ((bool)(*flags & MS_RDONLY) != sb_rdonly(sb)) {
4ab2f15b 5065 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 5066 err = -EROFS;
5067 goto restore_opts;
5068 }
5069
5070 if (*flags & MS_RDONLY) {
38c03b34
TT		 5071 err = sync_filesystem(sb);
5072 if (err < 0)
5073 goto restore_opts;
0f0dd62f
CH		 5074 err = dquot_suspend(sb, -1);
5075 if (err < 0)
c79d967d 5076 goto restore_opts;
c79d967d 5077
ac27a0ec
DK		 5078 /*
5079 * First of all, the unconditional stuff we have to do
5080 * to disable replay of the journal when we next remount
5081 */
5082 sb->s_flags |= MS_RDONLY;
5083
5084 /*
5085 * OK, test if we are remounting a valid rw partition
5086 * readonly, and if so set the rdonly flag and then
5087 * mark the partition as valid again.
5088 */
617ba13b
MC		 5089 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5090 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 5091 es->s_state = cpu_to_le16(sbi->s_mount_state);
5092
a63c9eb2 5093 if (sbi->s_journal)
0390131b 5094 ext4_mark_recovery_complete(sb, es);
ac27a0ec 5095 } else {
a13fb1a4 5096 /* Make sure we can mount this feature set readwrite */
e2b911c5 5097 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 5098 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 5099 err = -EROFS;
5100 goto restore_opts;
5101 }
8a266467
TT		 5102 /*
5103 * Make sure the group descriptor checksums
0b8e58a1 5104 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 5105 */
5106 for (g = 0; g < sbi->s_groups_count; g++) {
5107 struct ext4_group_desc *gdp =
5108 ext4_get_group_desc(sb, g, NULL);
5109
feb0ab32 5110 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5111 ext4_msg(sb, KERN_ERR,
5112 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 5113 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 5114 le16_to_cpu(gdp->bg_checksum));
6a797d27 5115 err = -EFSBADCRC;
8a266467
TT		 5116 goto restore_opts;
5117 }
5118 }
5119
ead6596b
ES		 5120 /*
5121 * If we have an unprocessed orphan list hanging
5122 * around from a previously readonly bdev mount,
5123 * require a full umount/remount for now.
5124 */
5125 if (es->s_last_orphan) {
b31e1552 5126 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5127 "remount RDWR because of unprocessed "
5128 "orphan inode list. Please "
b31e1552 5129 "umount/remount instead");
ead6596b
ES		 5130 err = -EINVAL;
5131 goto restore_opts;
5132 }
5133
ac27a0ec
DK		 5134 /*
5135 * Mounting a RDONLY partition read-write, so reread
5136 * and store the current valid flag. (It may have
5137 * been changed by e2fsck since we originally mounted
5138 * the partition.)
5139 */
0390131b
FM		 5140 if (sbi->s_journal)
5141 ext4_clear_journal_err(sb, es);
ac27a0ec 5142 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 5143 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 5144 sb->s_flags &= ~MS_RDONLY;
e2b911c5 5145 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 5146 if (ext4_multi_mount_protect(sb,
5147 le64_to_cpu(es->s_mmp_block))) {
5148 err = -EROFS;
5149 goto restore_opts;
5150 }
c79d967d 5151 enable_quota = 1;
ac27a0ec
DK		 5152 }
5153 }
bfff6873
LC		 5154
5155 /*
5156 * Reinitialize lazy itable initialization thread based on
5157 * current settings
5158 */
bc98a42c 5159 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
bfff6873
LC		 5160 ext4_unregister_li_request(sb);
5161 else {
5162 ext4_group_t first_not_zeroed;
5163 first_not_zeroed = ext4_has_uninit_itable(sb);
5164 ext4_register_li_request(sb, first_not_zeroed);
5165 }
5166
6fd058f7 5167 ext4_setup_system_zone(sb);
d096ad0f 5168 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 5169 ext4_commit_super(sb, 1);
0390131b 5170
ac27a0ec
DK		 5171#ifdef CONFIG_QUOTA
5172 /* Release old quota file names */
a2d4a646 5173 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5174 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5175 if (enable_quota) {
5176 if (sb_any_quota_suspended(sb))
5177 dquot_resume(sb, -1);
e2b911c5 5178 else if (ext4_has_feature_quota(sb)) {
7c319d32 5179 err = ext4_enable_quotas(sb);
07724f98 5180 if (err)
7c319d32 5181 goto restore_opts;
7c319d32
AK		 5182 }
5183 }
ac27a0ec 5184#endif
d4c402d9 5185
a26f4992 5186 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 5187 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5188 kfree(orig_data);
ac27a0ec 5189 return 0;
0b8e58a1 5190
ac27a0ec
DK		 5191restore_opts:
5192 sb->s_flags = old_sb_flags;
5193 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5194 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5195 sbi->s_resuid = old_opts.s_resuid;
5196 sbi->s_resgid = old_opts.s_resgid;
5197 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5198 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5199 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5200#ifdef CONFIG_QUOTA
5201 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5202 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5203 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5204 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5205 }
5206#endif
d4c402d9 5207 kfree(orig_data);
ac27a0ec
DK		 5208 return err;
5209}
5210
689c958c
LX		 5211#ifdef CONFIG_QUOTA
5212static int ext4_statfs_project(struct super_block *sb,
5213 kprojid_t projid, struct kstatfs *buf)
5214{
5215 struct kqid qid;
5216 struct dquot *dquot;
5217 u64 limit;
5218 u64 curblock;
5219
5220 qid = make_kqid_projid(projid);
5221 dquot = dqget(sb, qid);
5222 if (IS_ERR(dquot))
5223 return PTR_ERR(dquot);
7b9ca4c6 5224 spin_lock(&dquot->dq_dqb_lock);
689c958c
LX		 5225
5226 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5227 dquot->dq_dqb.dqb_bsoftlimit :
5228 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5229 if (limit && buf->f_blocks > limit) {
5230 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5231 buf->f_blocks = limit;
5232 buf->f_bfree = buf->f_bavail =
5233 (buf->f_blocks > curblock) ?
5234 (buf->f_blocks - curblock) : 0;
5235 }
5236
5237 limit = dquot->dq_dqb.dqb_isoftlimit ?
5238 dquot->dq_dqb.dqb_isoftlimit :
5239 dquot->dq_dqb.dqb_ihardlimit;
5240 if (limit && buf->f_files > limit) {
5241 buf->f_files = limit;
5242 buf->f_ffree =
5243 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5244 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5245 }
5246
7b9ca4c6 5247 spin_unlock(&dquot->dq_dqb_lock);
689c958c
LX		 5248 dqput(dquot);
5249 return 0;
5250}
5251#endif
5252
2b2d6d01 5253static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5254{
5255 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5256 struct ext4_sb_info *sbi = EXT4_SB(sb);
5257 struct ext4_super_block *es = sbi->s_es;
27dd4385 5258 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5259 u64 fsid;
d02a9391 5260 s64 bfree;
27dd4385 5261 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5262
952fc18e
TT		 5263 if (!test_opt(sb, MINIX_DF))
5264 overhead = sbi->s_overhead;
ac27a0ec 5265
617ba13b 5266 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5267 buf->f_bsize = sb->s_blocksize;
b72f78cb 5268 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5269 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5270 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5271 /* prevent underflow in case that few free space is available */
57042651 5272 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5273 buf->f_bavail = buf->f_bfree -
5274 (ext4_r_blocks_count(es) + resv_blocks);
5275 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5276 buf->f_bavail = 0;
5277 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5278 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5279 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5280 fsid = le64_to_cpup((void *)es->s_uuid) ^
5281 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5282 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5283 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5284
689c958c
LX		 5285#ifdef CONFIG_QUOTA
5286 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5287 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5288 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5289#endif
ac27a0ec
DK		 5290 return 0;
5291}
5292
ac27a0ec
DK		 5293
5294#ifdef CONFIG_QUOTA
5295
bc8230ee
JK		 5296/*
5297 * Helper functions so that transaction is started before we acquire dqio_sem
5298 * to keep correct lock ordering of transaction > dqio_sem
5299 */
ac27a0ec
DK		 5300static inline struct inode *dquot_to_inode(struct dquot *dquot)
5301{
4c376dca 5302 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5303}
5304
617ba13b 5305static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5306{
5307 int ret, err;
5308 handle_t *handle;
5309 struct inode *inode;
5310
5311 inode = dquot_to_inode(dquot);
9924a92a 5312 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5313 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5314 if (IS_ERR(handle))
5315 return PTR_ERR(handle);
5316 ret = dquot_commit(dquot);
617ba13b 5317 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5318 if (!ret)
5319 ret = err;
5320 return ret;
5321}
5322
617ba13b 5323static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5324{
5325 int ret, err;
5326 handle_t *handle;
5327
9924a92a 5328 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5329 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5330 if (IS_ERR(handle))
5331 return PTR_ERR(handle);
5332 ret = dquot_acquire(dquot);
617ba13b 5333 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5334 if (!ret)
5335 ret = err;
5336 return ret;
5337}
5338
617ba13b 5339static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5340{
5341 int ret, err;
5342 handle_t *handle;
5343
9924a92a 5344 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5345 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5346 if (IS_ERR(handle)) {
5347 /* Release dquot anyway to avoid endless cycle in dqput() */
5348 dquot_release(dquot);
ac27a0ec 5349 return PTR_ERR(handle);
9c3013e9 5350 }
ac27a0ec 5351 ret = dquot_release(dquot);
617ba13b 5352 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5353 if (!ret)
5354 ret = err;
5355 return ret;
5356}
5357
617ba13b 5358static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5359{
262b4662
JK		 5360 struct super_block *sb = dquot->dq_sb;
5361 struct ext4_sb_info *sbi = EXT4_SB(sb);
5362
2c8be6b2 5363 /* Are we journaling quotas? */
e2b911c5 5364 if (ext4_has_feature_quota(sb) ||
262b4662 5365 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5366 dquot_mark_dquot_dirty(dquot);
617ba13b 5367 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5368 } else {
5369 return dquot_mark_dquot_dirty(dquot);
5370 }
5371}
5372
617ba13b 5373static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5374{
5375 int ret, err;
5376 handle_t *handle;
5377
5378 /* Data block + inode block */
2b0143b5 5379 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5380 if (IS_ERR(handle))
5381 return PTR_ERR(handle);
5382 ret = dquot_commit_info(sb, type);
617ba13b 5383 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5384 if (!ret)
5385 ret = err;
5386 return ret;
5387}
5388
5389/*
5390 * Turn on quotas during mount time - we need to find
5391 * the quota file and such...
5392 */
617ba13b 5393static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5394{
287a8095
CH		 5395 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5396 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5397}
5398
daf647d2
TT		 5399static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5400{
5401 struct ext4_inode_info *ei = EXT4_I(inode);
5402
5403 /* The first argument of lockdep_set_subclass has to be
5404 * *exactly* the same as the argument to init_rwsem() --- in
5405 * this case, in init_once() --- or lockdep gets unhappy
5406 * because the name of the lock is set using the
5407 * stringification of the argument to init_rwsem().
5408 */
5409 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5410 lockdep_set_subclass(&ei->i_data_sem, subclass);
5411}
5412
ac27a0ec
DK		 5413/*
5414 * Standard function to be called on quota_on
5415 */
617ba13b 5416static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 5417 const struct path *path)
ac27a0ec
DK		 5418{
5419 int err;
ac27a0ec
DK		 5420
5421 if (!test_opt(sb, QUOTA))
5422 return -EINVAL;
0623543b 5423
ac27a0ec 5424 /* Quotafile not on the same filesystem? */
d8c9584e 5425 if (path->dentry->d_sb != sb)
ac27a0ec 5426 return -EXDEV;
0623543b
JK		 5427 /* Journaling quota? */
5428 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5429 /* Quotafile not in fs root? */
f00c9e44 5430 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5431 ext4_msg(sb, KERN_WARNING,
5432 "Quota file not on filesystem root. "
5433 "Journaled quota will not work");
91389240
JK		 5434 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5435 } else {
5436 /*
5437 * Clear the flag just in case mount options changed since
5438 * last time.
5439 */
5440 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
2b2d6d01 5441 }
0623543b
JK		 5442
5443 /*
5444 * When we journal data on quota file, we have to flush journal to see
5445 * all updates to the file when we bypass pagecache...
5446 */
0390131b 5447 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5448 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5449 /*
5450 * We don't need to lock updates but journal_flush() could
5451 * otherwise be livelocked...
5452 */
5453 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5454 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5455 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5456 if (err)
7ffe1ea8 5457 return err;
0623543b 5458 }
957153fc 5459
daf647d2
TT		 5460 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5461 err = dquot_quota_on(sb, type, format_id, path);
957153fc 5462 if (err) {
daf647d2
TT		 5463 lockdep_set_quota_inode(path->dentry->d_inode,
5464 I_DATA_SEM_NORMAL);
957153fc
JK		 5465 } else {
5466 struct inode *inode = d_inode(path->dentry);
5467 handle_t *handle;
5468
61a92987
JK		 5469 /*
5470 * Set inode flags to prevent userspace from messing with quota
5471 * files. If this fails, we return success anyway since quotas
5472 * are already enabled and this is not a hard failure.
5473 */
957153fc
JK		 5474 inode_lock(inode);
5475 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5476 if (IS_ERR(handle))
5477 goto unlock_inode;
5478 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5479 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5480 S_NOATIME | S_IMMUTABLE);
5481 ext4_mark_inode_dirty(handle, inode);
5482 ext4_journal_stop(handle);
5483 unlock_inode:
5484 inode_unlock(inode);
5485 }
daf647d2 5486 return err;
ac27a0ec
DK		 5487}
5488
7c319d32
AK		 5489static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5490 unsigned int flags)
5491{
5492 int err;
5493 struct inode *qf_inode;
a2d4a646 5494 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5495 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5496 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5497 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5498 };
5499
e2b911c5 5500 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5501
5502 if (!qf_inums[type])
5503 return -EPERM;
5504
5505 qf_inode = ext4_iget(sb, qf_inums[type]);
5506 if (IS_ERR(qf_inode)) {
5507 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5508 return PTR_ERR(qf_inode);
5509 }
5510
bcb13850
JK		 5511 /* Don't account quota for quota files to avoid recursion */
5512 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5513 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7c319d32
AK		 5514 err = dquot_enable(qf_inode, type, format_id, flags);
5515 iput(qf_inode);
daf647d2
TT		 5516 if (err)
5517 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7c319d32
AK		 5518
5519 return err;
5520}
5521
5522/* Enable usage tracking for all quota types. */
5523static int ext4_enable_quotas(struct super_block *sb)
5524{
5525 int type, err = 0;
a2d4a646 5526 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5527 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5528 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5529 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 5530 };
49da9392
JK		 5531 bool quota_mopt[EXT4_MAXQUOTAS] = {
5532 test_opt(sb, USRQUOTA),
5533 test_opt(sb, GRPQUOTA),
5534 test_opt(sb, PRJQUOTA),
5535 };
7c319d32 5536
91389240 5537 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
a2d4a646 5538 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5539 if (qf_inums[type]) {
5540 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 5541 DQUOT_USAGE_ENABLED |
5542 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32 5543 if (err) {
95f1fda4 5544 for (type--; type >= 0; type--)
5545 dquot_quota_off(sb, type);
5546
7c319d32 5547 ext4_warning(sb,
72ba7450
TT		 5548 "Failed to enable quota tracking "
5549 "(type=%d, err=%d). Please run "
5550 "e2fsck to fix.", type, err);
7c319d32
AK		 5551 return err;
5552 }
5553 }
5554 }
5555 return 0;
5556}
5557
ca0e05e4
DM		 5558static int ext4_quota_off(struct super_block *sb, int type)
5559{
21f97697
JK		 5560 struct inode *inode = sb_dqopt(sb)->files[type];
5561 handle_t *handle;
957153fc 5562 int err;
21f97697 5563
87009d86
DM		 5564 /* Force all delayed allocation blocks to be allocated.
5565 * Caller already holds s_umount sem */
5566 if (test_opt(sb, DELALLOC))
ca0e05e4 5567 sync_filesystem(sb);
ca0e05e4 5568
957153fc 5569 if (!inode || !igrab(inode))
0b268590
AG		 5570 goto out;
5571
957153fc 5572 err = dquot_quota_off(sb, type);
964edf66 5573 if (err || ext4_has_feature_quota(sb))
957153fc
JK		 5574 goto out_put;
5575
5576 inode_lock(inode);
61a92987
JK		 5577 /*
5578 * Update modification times of quota files when userspace can
5579 * start looking at them. If we fail, we return success anyway since
5580 * this is not a hard failure and quotas are already disabled.
5581 */
9924a92a 5582 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697 5583 if (IS_ERR(handle))
957153fc
JK		 5584 goto out_unlock;
5585 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5586 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
eeca7ea1 5587 inode->i_mtime = inode->i_ctime = current_time(inode);
21f97697
JK		 5588 ext4_mark_inode_dirty(handle, inode);
5589 ext4_journal_stop(handle);
957153fc
JK		 5590out_unlock:
5591 inode_unlock(inode);
5592out_put:
964edf66 5593 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 5594 iput(inode);
5595 return err;
21f97697 5596out:
ca0e05e4
DM		 5597 return dquot_quota_off(sb, type);
5598}
5599
ac27a0ec
DK		 5600/* Read data from quotafile - avoid pagecache and such because we cannot afford
5601 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5602 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5603 * we don't have to be afraid of races */
617ba13b 5604static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5605 size_t len, loff_t off)
5606{
5607 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5608 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5609 int offset = off & (sb->s_blocksize - 1);
5610 int tocopy;
5611 size_t toread;
5612 struct buffer_head *bh;
5613 loff_t i_size = i_size_read(inode);
5614
5615 if (off > i_size)
5616 return 0;
5617 if (off+len > i_size)
5618 len = i_size-off;
5619 toread = len;
5620 while (toread > 0) {
5621 tocopy = sb->s_blocksize - offset < toread ?
5622 sb->s_blocksize - offset : toread;
1c215028
TT		 5623 bh = ext4_bread(NULL, inode, blk, 0);
5624 if (IS_ERR(bh))
5625 return PTR_ERR(bh);
ac27a0ec
DK		 5626 if (!bh) /* A hole? */
5627 memset(data, 0, tocopy);
5628 else
5629 memcpy(data, bh->b_data+offset, tocopy);
5630 brelse(bh);
5631 offset = 0;
5632 toread -= tocopy;
5633 data += tocopy;
5634 blk++;
5635 }
5636 return len;
5637}
5638
5639/* Write to quotafile (we know the transaction is already started and has
5640 * enough credits) */
617ba13b 5641static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5642 const char *data, size_t len, loff_t off)
5643{
5644 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5645 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5646 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5647 int retries = 0;
ac27a0ec
DK		 5648 struct buffer_head *bh;
5649 handle_t *handle = journal_current_handle();
5650
0390131b 5651 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5652 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5653 " cancelled because transaction is not started",
9c3013e9
JK		 5654 (unsigned long long)off, (unsigned long long)len);
5655 return -EIO;
5656 }
67eeb568
DM		 5657 /*
5658 * Since we account only one data block in transaction credits,
5659 * then it is impossible to cross a block boundary.
5660 */
5661 if (sb->s_blocksize - offset < len) {
5662 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5663 " cancelled because not block aligned",
5664 (unsigned long long)off, (unsigned long long)len);
5665 return -EIO;
5666 }
5667
c5e298ae
TT		 5668 do {
5669 bh = ext4_bread(handle, inode, blk,
5670 EXT4_GET_BLOCKS_CREATE |
5671 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5672 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5673 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5674 if (IS_ERR(bh))
5675 return PTR_ERR(bh);
67eeb568
DM		 5676 if (!bh)
5677 goto out;
5d601255 5678 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5679 err = ext4_journal_get_write_access(handle, bh);
5680 if (err) {
5681 brelse(bh);
1c215028 5682 return err;
ac27a0ec 5683 }
67eeb568
DM		 5684 lock_buffer(bh);
5685 memcpy(bh->b_data+offset, data, len);
5686 flush_dcache_page(bh->b_page);
5687 unlock_buffer(bh);
62d2b5f2 5688 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5689 brelse(bh);
ac27a0ec 5690out:
67eeb568
DM		 5691 if (inode->i_size < off + len) {
5692 i_size_write(inode, off + len);
617ba13b 5693 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5694 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5695 }
67eeb568 5696 return len;
ac27a0ec
DK		 5697}
5698
8f0e8746
TT		 5699static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5700{
5701 const struct quota_format_ops *ops;
5702
5703 if (!sb_has_quota_loaded(sb, qid->type))
5704 return -ESRCH;
5705 ops = sb_dqopt(sb)->ops[qid->type];
5706 if (!ops || !ops->get_next_id)
5707 return -ENOSYS;
5708 return dquot_get_next_id(sb, qid);
5709}
ac27a0ec
DK		 5710#endif
5711
152a0836
AV		 5712static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5713 const char *dev_name, void *data)
ac27a0ec 5714{
152a0836 5715 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5716}
5717
c290ea01 5718#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5719static inline void register_as_ext2(void)
5720{
5721 int err = register_filesystem(&ext2_fs_type);
5722 if (err)
5723 printk(KERN_WARNING
5724 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5725}
5726
5727static inline void unregister_as_ext2(void)
5728{
5729 unregister_filesystem(&ext2_fs_type);
5730}
2035e776
TT		 5731
5732static inline int ext2_feature_set_ok(struct super_block *sb)
5733{
e2b911c5 5734 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776 5735 return 0;
bc98a42c 5736 if (sb_rdonly(sb))
2035e776 5737 return 1;
e2b911c5 5738 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5739 return 0;
5740 return 1;
5741}
24b58424
TT		 5742#else
5743static inline void register_as_ext2(void) { }
5744static inline void unregister_as_ext2(void) { }
2035e776 5745static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5746#endif
5747
24b58424
TT		 5748static inline void register_as_ext3(void)
5749{
5750 int err = register_filesystem(&ext3_fs_type);
5751 if (err)
5752 printk(KERN_WARNING
5753 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5754}
5755
5756static inline void unregister_as_ext3(void)
5757{
5758 unregister_filesystem(&ext3_fs_type);
5759}
2035e776
TT		 5760
5761static inline int ext3_feature_set_ok(struct super_block *sb)
5762{
e2b911c5 5763 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5764 return 0;
e2b911c5 5765 if (!ext4_has_feature_journal(sb))
2035e776 5766 return 0;
bc98a42c 5767 if (sb_rdonly(sb))
2035e776 5768 return 1;
e2b911c5 5769 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5770 return 0;
5771 return 1;
5772}
24b58424 5773
03010a33
TT		 5774static struct file_system_type ext4_fs_type = {
5775 .owner = THIS_MODULE,
5776 .name = "ext4",
152a0836 5777 .mount = ext4_mount,
03010a33
TT		 5778 .kill_sb = kill_block_super,
5779 .fs_flags = FS_REQUIRES_DEV,
5780};
7f78e035 5781MODULE_ALIAS_FS("ext4");
03010a33 5782
e9e3bcec
ES		 5783/* Shared across all ext4 file systems */
5784wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 5785
5dabfc78 5786static int __init ext4_init_fs(void)
ac27a0ec 5787{
e9e3bcec 5788 int i, err;
c9de560d 5789
e294a537 5790 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5791 ext4_li_info = NULL;
5792 mutex_init(&ext4_li_mtx);
5793
9a4c8019 5794 /* Build-time check for flags consistency */
12e9b892 5795 ext4_check_flag_values();
e9e3bcec 5796
e142d052 5797 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 5798 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 5799
51865fda 5800 err = ext4_init_es();
6fd058f7
TT		 5801 if (err)
5802 return err;
51865fda
ZL		 5803
5804 err = ext4_init_pageio();
5805 if (err)
b5799018 5806 goto out5;
51865fda 5807
5dabfc78 5808 err = ext4_init_system_zone();
bd2d0210 5809 if (err)
b5799018 5810 goto out4;
857ac889 5811
b5799018 5812 err = ext4_init_sysfs();
dd68314c 5813 if (err)
b5799018 5814 goto out3;
857ac889 5815
5dabfc78 5816 err = ext4_init_mballoc();
c9de560d
AT		 5817 if (err)
5818 goto out2;
ac27a0ec
DK		 5819 err = init_inodecache();
5820 if (err)
5821 goto out1;
24b58424 5822 register_as_ext3();
2035e776 5823 register_as_ext2();
03010a33 5824 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5825 if (err)
5826 goto out;
bfff6873 5827
ac27a0ec
DK		 5828 return 0;
5829out:
24b58424
TT		 5830 unregister_as_ext2();
5831 unregister_as_ext3();
ac27a0ec
DK		 5832 destroy_inodecache();
5833out1:
5dabfc78 5834 ext4_exit_mballoc();
9c191f70 5835out2:
b5799018
TT		 5836 ext4_exit_sysfs();
5837out3:
5dabfc78 5838 ext4_exit_system_zone();
b5799018 5839out4:
5dabfc78 5840 ext4_exit_pageio();
b5799018 5841out5:
51865fda
ZL		 5842 ext4_exit_es();
5843
ac27a0ec
DK		 5844 return err;
5845}
5846
5dabfc78 5847static void __exit ext4_exit_fs(void)
ac27a0ec 5848{
bfff6873 5849 ext4_destroy_lazyinit_thread();
24b58424
TT		 5850 unregister_as_ext2();
5851 unregister_as_ext3();
03010a33 5852 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5853 destroy_inodecache();
5dabfc78 5854 ext4_exit_mballoc();
b5799018 5855 ext4_exit_sysfs();
5dabfc78
TT		 5856 ext4_exit_system_zone();
5857 ext4_exit_pageio();
dd12ed14 5858 ext4_exit_es();
ac27a0ec
DK		 5859}
5860
5861MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5862MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5863MODULE_LICENSE("GPL");
5dabfc78
TT		 5864module_init(ext4_init_fs)
5865module_exit(ext4_exit_fs)
A mirror of Linus' kernel repository