]> 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: remove NULL check before calling kmem_cache_destroy()
[thirdparty/kernel/linux.git] / fs / ext4 / super.c
CommitLineData
f5166768 1// SPDX-License-Identifier: GPL-2.0
ac27a0ec 2/*
617ba13b 3 * linux/fs/ext4/super.c
ac27a0ec
DK		 4 *
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
9 *
10 * from
11 *
12 * linux/fs/minix/inode.c
13 *
14 * Copyright (C) 1991, 1992 Linus Torvalds
15 *
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
18 */
19
20#include <linux/module.h>
21#include <linux/string.h>
22#include <linux/fs.h>
23#include <linux/time.h>
c5ca7c76 24#include <linux/vmalloc.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
66114cad 28#include <linux/backing-dev.h>
ac27a0ec 29#include <linux/parser.h>
ac27a0ec 30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
ef510424 41#include <linux/dax.h>
7abc52c2 42#include <linux/cleancache.h>
7c0f6ba6 43#include <linux/uaccess.h>
ee73f9a5 44#include <linux/iversion.h>
ac27a0ec 45
bfff6873
LC		 46#include <linux/kthread.h>
47#include <linux/freezer.h>
48
3dcf5451 49#include "ext4.h"
4a092d73 50#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 51#include "ext4_jbd2.h"
ac27a0ec
DK		 52#include "xattr.h"
53#include "acl.h"
3661d286 54#include "mballoc.h"
0c9ec4be 55#include "fsmap.h"
ac27a0ec 56
9bffad1e
TT		 57#define CREATE_TRACE_POINTS
58#include <trace/events/ext4.h>
59
0b75a840
LC		 60static struct ext4_lazy_init *ext4_li_info;
61static struct mutex ext4_li_mtx;
e294a537 62static struct ratelimit_state ext4_mount_msg_ratelimit;
9f6200bb 63
617ba13b 64static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 65 unsigned long journal_devnum);
2adf6da8 66static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 67static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 68static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
617ba13b 72static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 73static int ext4_remount(struct super_block *sb, int *flags, char *data);
74static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 75static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 76static int ext4_freeze(struct super_block *sb);
152a0836
AV		 77static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
2035e776
TT		 79static inline int ext2_feature_set_ok(struct super_block *sb);
80static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 81static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 82static void ext4_destroy_lazyinit_thread(void);
83static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 84static void ext4_clear_request_list(void);
c6cb7e77
EW		 85static struct inode *ext4_get_journal_inode(struct super_block *sb,
86 unsigned int journal_inum);
ac27a0ec 87
e74031fd
JK		 88/*
89 * Lock ordering
90 *
91 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
92 * i_mmap_rwsem (inode->i_mmap_rwsem)!
93 *
94 * page fault path:
95 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
96 * page lock -> i_data_sem (rw)
97 *
98 * buffered write path:
99 * sb_start_write -> i_mutex -> mmap_sem
100 * sb_start_write -> i_mutex -> transaction start -> page lock ->
101 * i_data_sem (rw)
102 *
103 * truncate:
1d39834f
NB		 104 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
106 * i_data_sem (rw)
e74031fd
JK		 107 *
108 * direct IO:
1d39834f
NB		 109 * sb_start_write -> i_mutex -> mmap_sem
110 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
e74031fd
JK		 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();
1751e8a6 425 sb->s_flags |= SB_RDONLY;
ac27a0ec 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
ccf0f32a 449 trace_ext4_error(sb, function, line);
efbed4dc
TT		 450 if (ext4_error_ratelimit(sb)) {
451 va_start(args, fmt);
452 vaf.fmt = fmt;
453 vaf.va = &args;
454 printk(KERN_CRIT
455 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
456 sb->s_id, function, line, current->comm, &vaf);
457 va_end(args);
458 }
f3fc0210 459 save_error_info(sb, function, line);
617ba13b 460 ext4_handle_error(sb);
ac27a0ec
DK		 461}
462
e7c96e8e
JP		 463void __ext4_error_inode(struct inode *inode, const char *function,
464 unsigned int line, ext4_fsblk_t block,
465 const char *fmt, ...)
273df556
FM		 466{
467 va_list args;
f7c21177 468 struct va_format vaf;
1c13d5c0 469 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 470
0db1ff22
TT		 471 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
472 return;
473
ccf0f32a 474 trace_ext4_error(inode->i_sb, function, line);
1c13d5c0
TT		 475 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
476 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 477 if (ext4_error_ratelimit(inode->i_sb)) {
478 va_start(args, fmt);
479 vaf.fmt = fmt;
480 vaf.va = &args;
481 if (block)
482 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
483 "inode #%lu: block %llu: comm %s: %pV\n",
484 inode->i_sb->s_id, function, line, inode->i_ino,
485 block, current->comm, &vaf);
486 else
487 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
488 "inode #%lu: comm %s: %pV\n",
489 inode->i_sb->s_id, function, line, inode->i_ino,
490 current->comm, &vaf);
491 va_end(args);
492 }
1c13d5c0 493 save_error_info(inode->i_sb, function, line);
273df556
FM		 494 ext4_handle_error(inode->i_sb);
495}
496
e7c96e8e
JP		 497void __ext4_error_file(struct file *file, const char *function,
498 unsigned int line, ext4_fsblk_t block,
499 const char *fmt, ...)
273df556
FM		 500{
501 va_list args;
f7c21177 502 struct va_format vaf;
1c13d5c0 503 struct ext4_super_block *es;
496ad9aa 504 struct inode *inode = file_inode(file);
273df556
FM		 505 char pathname[80], *path;
506
0db1ff22
TT		 507 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
508 return;
509
ccf0f32a 510 trace_ext4_error(inode->i_sb, function, line);
1c13d5c0
TT		 511 es = EXT4_SB(inode->i_sb)->s_es;
512 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 513 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 514 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 515 if (IS_ERR(path))
516 path = "(unknown)";
517 va_start(args, fmt);
518 vaf.fmt = fmt;
519 vaf.va = &args;
520 if (block)
521 printk(KERN_CRIT
522 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
523 "block %llu: comm %s: path %s: %pV\n",
524 inode->i_sb->s_id, function, line, inode->i_ino,
525 block, current->comm, path, &vaf);
526 else
527 printk(KERN_CRIT
528 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
529 "comm %s: path %s: %pV\n",
530 inode->i_sb->s_id, function, line, inode->i_ino,
531 current->comm, path, &vaf);
532 va_end(args);
533 }
1c13d5c0 534 save_error_info(inode->i_sb, function, line);
273df556
FM		 535 ext4_handle_error(inode->i_sb);
536}
537
722887dd
TT		 538const char *ext4_decode_error(struct super_block *sb, int errno,
539 char nbuf[16])
ac27a0ec
DK		 540{
541 char *errstr = NULL;
542
543 switch (errno) {
6a797d27
DW		 544 case -EFSCORRUPTED:
545 errstr = "Corrupt filesystem";
546 break;
547 case -EFSBADCRC:
548 errstr = "Filesystem failed CRC";
549 break;
ac27a0ec
DK		 550 case -EIO:
551 errstr = "IO failure";
552 break;
553 case -ENOMEM:
554 errstr = "Out of memory";
555 break;
556 case -EROFS:
78f1ddbb
TT		 557 if (!sb || (EXT4_SB(sb)->s_journal &&
558 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 559 errstr = "Journal has aborted";
560 else
561 errstr = "Readonly filesystem";
562 break;
563 default:
564 /* If the caller passed in an extra buffer for unknown
565 * errors, textualise them now. Else we just return
566 * NULL. */
567 if (nbuf) {
568 /* Check for truncated error codes... */
569 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
570 errstr = nbuf;
571 }
572 break;
573 }
574
575 return errstr;
576}
577
617ba13b 578/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 579 * automatically and invokes the appropriate error response. */
580
c398eda0
TT		 581void __ext4_std_error(struct super_block *sb, const char *function,
582 unsigned int line, int errno)
ac27a0ec
DK		 583{
584 char nbuf[16];
585 const char *errstr;
586
0db1ff22
TT		 587 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
588 return;
589
ac27a0ec
DK		 590 /* Special case: if the error is EROFS, and we're not already
591 * inside a transaction, then there's really no point in logging
592 * an error. */
bc98a42c 593 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
ac27a0ec
DK		 594 return;
595
efbed4dc
TT		 596 if (ext4_error_ratelimit(sb)) {
597 errstr = ext4_decode_error(sb, errno, nbuf);
598 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
599 sb->s_id, function, line, errstr);
600 }
ac27a0ec 601
efbed4dc 602 save_error_info(sb, function, line);
617ba13b 603 ext4_handle_error(sb);
ac27a0ec
DK		 604}
605
606/*
617ba13b 607 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 608 * abort function may be used to deal with unrecoverable failures such
609 * as journal IO errors or ENOMEM at a critical moment in log management.
610 *
611 * We unconditionally force the filesystem into an ABORT|READONLY state,
612 * unless the error response on the fs has been set to panic in which
613 * case we take the easy way out and panic immediately.
614 */
615
c67d859e 616void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 617 unsigned int line, const char *fmt, ...)
ac27a0ec 618{
651e1c3b 619 struct va_format vaf;
ac27a0ec
DK		 620 va_list args;
621
0db1ff22
TT		 622 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
623 return;
624
1c13d5c0 625 save_error_info(sb, function, line);
ac27a0ec 626 va_start(args, fmt);
651e1c3b
JP		 627 vaf.fmt = fmt;
628 vaf.va = &args;
629 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
630 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 631 va_end(args);
632
bc98a42c 633 if (sb_rdonly(sb) == 0) {
1c13d5c0 634 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 635 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 636 /*
637 * Make sure updated value of ->s_mount_flags will be visible
638 * before ->s_flags update
639 */
640 smp_wmb();
1751e8a6 641 sb->s_flags |= SB_RDONLY;
1c13d5c0
TT		 642 if (EXT4_SB(sb)->s_journal)
643 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
644 save_error_info(sb, function, line);
645 }
4327ba52
DJ		 646 if (test_opt(sb, ERRORS_PANIC)) {
647 if (EXT4_SB(sb)->s_journal &&
648 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
649 return;
617ba13b 650 panic("EXT4-fs panic from previous error\n");
4327ba52 651 }
ac27a0ec
DK		 652}
653
e7c96e8e
JP		 654void __ext4_msg(struct super_block *sb,
655 const char *prefix, const char *fmt, ...)
b31e1552 656{
0ff2ea7d 657 struct va_format vaf;
b31e1552
ES		 658 va_list args;
659
efbed4dc
TT		 660 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
661 return;
662
b31e1552 663 va_start(args, fmt);
0ff2ea7d
JP		 664 vaf.fmt = fmt;
665 vaf.va = &args;
666 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 667 va_end(args);
668}
669
b03a2f7e
AD		 670#define ext4_warning_ratelimit(sb) \
671 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
672 "EXT4-fs warning")
673
12062ddd 674void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 675 unsigned int line, const char *fmt, ...)
ac27a0ec 676{
0ff2ea7d 677 struct va_format vaf;
ac27a0ec
DK		 678 va_list args;
679
b03a2f7e 680 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 681 return;
682
ac27a0ec 683 va_start(args, fmt);
0ff2ea7d
JP		 684 vaf.fmt = fmt;
685 vaf.va = &args;
686 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
687 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 688 va_end(args);
689}
690
b03a2f7e
AD		 691void __ext4_warning_inode(const struct inode *inode, const char *function,
692 unsigned int line, const char *fmt, ...)
693{
694 struct va_format vaf;
695 va_list args;
696
697 if (!ext4_warning_ratelimit(inode->i_sb))
698 return;
699
700 va_start(args, fmt);
701 vaf.fmt = fmt;
702 vaf.va = &args;
703 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
704 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
705 function, line, inode->i_ino, current->comm, &vaf);
706 va_end(args);
707}
708
e29136f8
TT		 709void __ext4_grp_locked_error(const char *function, unsigned int line,
710 struct super_block *sb, ext4_group_t grp,
711 unsigned long ino, ext4_fsblk_t block,
712 const char *fmt, ...)
5d1b1b3f
AK		 713__releases(bitlock)
714__acquires(bitlock)
715{
0ff2ea7d 716 struct va_format vaf;
5d1b1b3f
AK		 717 va_list args;
718 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
719
0db1ff22
TT		 720 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
721 return;
722
ccf0f32a 723 trace_ext4_error(sb, function, line);
1c13d5c0
TT		 724 es->s_last_error_ino = cpu_to_le32(ino);
725 es->s_last_error_block = cpu_to_le64(block);
726 __save_error_info(sb, function, line);
0ff2ea7d 727
efbed4dc
TT		 728 if (ext4_error_ratelimit(sb)) {
729 va_start(args, fmt);
730 vaf.fmt = fmt;
731 vaf.va = &args;
732 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
733 sb->s_id, function, line, grp);
734 if (ino)
735 printk(KERN_CONT "inode %lu: ", ino);
736 if (block)
737 printk(KERN_CONT "block %llu:",
738 (unsigned long long) block);
739 printk(KERN_CONT "%pV\n", &vaf);
740 va_end(args);
741 }
5d1b1b3f
AK		 742
743 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 744 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 745 return;
746 }
1c13d5c0 747
5d1b1b3f 748 ext4_unlock_group(sb, grp);
06f29cc8 749 ext4_commit_super(sb, 1);
5d1b1b3f
AK		 750 ext4_handle_error(sb);
751 /*
752 * We only get here in the ERRORS_RO case; relocking the group
753 * may be dangerous, but nothing bad will happen since the
754 * filesystem will have already been marked read/only and the
755 * journal has been aborted. We return 1 as a hint to callers
756 * who might what to use the return value from
25985edc 757 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 758 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
759 * aggressively from the ext4 function in question, with a
760 * more appropriate error code.
761 */
762 ext4_lock_group(sb, grp);
763 return;
764}
765
db79e6d1
WS		 766void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
767 ext4_group_t group,
768 unsigned int flags)
769{
770 struct ext4_sb_info *sbi = EXT4_SB(sb);
771 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
772 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
773
774 if ((flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) &&
775 !EXT4_MB_GRP_BBITMAP_CORRUPT(grp)) {
776 percpu_counter_sub(&sbi->s_freeclusters_counter,
777 grp->bb_free);
778 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
779 &grp->bb_state);
780 }
781
782 if ((flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) &&
783 !EXT4_MB_GRP_IBITMAP_CORRUPT(grp)) {
784 if (gdp) {
785 int count;
786
787 count = ext4_free_inodes_count(sb, gdp);
788 percpu_counter_sub(&sbi->s_freeinodes_counter,
789 count);
790 }
791 set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
792 &grp->bb_state);
793 }
794}
795
617ba13b 796void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 797{
617ba13b 798 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 799
617ba13b 800 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 801 return;
802
12062ddd 803 ext4_warning(sb,
ac27a0ec
DK		 804 "updating to rev %d because of new feature flag, "
805 "running e2fsck is recommended",
617ba13b 806 EXT4_DYNAMIC_REV);
ac27a0ec 807
617ba13b
MC		 808 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
809 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
810 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 811 /* leave es->s_feature_*compat flags alone */
812 /* es->s_uuid will be set by e2fsck if empty */
813
814 /*
815 * The rest of the superblock fields should be zero, and if not it
816 * means they are likely already in use, so leave them alone. We
817 * can leave it up to e2fsck to clean up any inconsistencies there.
818 */
819}
820
821/*
822 * Open the external journal device
823 */
b31e1552 824static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 825{
826 struct block_device *bdev;
827 char b[BDEVNAME_SIZE];
828
d4d77629 829 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 830 if (IS_ERR(bdev))
831 goto fail;
832 return bdev;
833
834fail:
b31e1552 835 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 836 __bdevname(dev, b), PTR_ERR(bdev));
837 return NULL;
838}
839
840/*
841 * Release the journal device
842 */
4385bab1 843static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 844{
4385bab1 845 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 846}
847
4385bab1 848static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 849{
850 struct block_device *bdev;
ac27a0ec
DK		 851 bdev = sbi->journal_bdev;
852 if (bdev) {
4385bab1 853 ext4_blkdev_put(bdev);
ac27a0ec
DK		 854 sbi->journal_bdev = NULL;
855 }
ac27a0ec
DK		 856}
857
858static inline struct inode *orphan_list_entry(struct list_head *l)
859{
617ba13b 860 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 861}
862
617ba13b 863static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 864{
865 struct list_head *l;
866
b31e1552
ES		 867 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
868 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 869
870 printk(KERN_ERR "sb_info orphan list:\n");
871 list_for_each(l, &sbi->s_orphan) {
872 struct inode *inode = orphan_list_entry(l);
873 printk(KERN_ERR " "
874 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
875 inode->i_sb->s_id, inode->i_ino, inode,
876 inode->i_mode, inode->i_nlink,
877 NEXT_ORPHAN(inode));
878 }
879}
880
957153fc
JK		 881#ifdef CONFIG_QUOTA
882static int ext4_quota_off(struct super_block *sb, int type);
883
884static inline void ext4_quota_off_umount(struct super_block *sb)
885{
886 int type;
887
964edf66
JK		 888 /* Use our quota_off function to clear inode flags etc. */
889 for (type = 0; type < EXT4_MAXQUOTAS; type++)
890 ext4_quota_off(sb, type);
957153fc
JK		 891}
892#else
893static inline void ext4_quota_off_umount(struct super_block *sb)
894{
895}
896#endif
897
2b2d6d01 898static void ext4_put_super(struct super_block *sb)
ac27a0ec 899{
617ba13b
MC		 900 struct ext4_sb_info *sbi = EXT4_SB(sb);
901 struct ext4_super_block *es = sbi->s_es;
97abd7d4 902 int aborted = 0;
ef2cabf7 903 int i, err;
ac27a0ec 904
857ac889 905 ext4_unregister_li_request(sb);
957153fc 906 ext4_quota_off_umount(sb);
e0ccfd95 907
2e8fa54e 908 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 909
0390131b 910 if (sbi->s_journal) {
97abd7d4 911 aborted = is_journal_aborted(sbi->s_journal);
0390131b
FM		 912 err = jbd2_journal_destroy(sbi->s_journal);
913 sbi->s_journal = NULL;
97abd7d4 914 if ((err < 0) && !aborted)
c67d859e 915 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 916 }
d4edac31 917
ebd173be 918 ext4_unregister_sysfs(sb);
d3922a77 919 ext4_es_unregister_shrinker(sbi);
9105bb14 920 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 921 ext4_release_system_zone(sb);
922 ext4_mb_release(sb);
923 ext4_ext_release(sb);
d4edac31 924
bc98a42c 925 if (!sb_rdonly(sb) && !aborted) {
e2b911c5 926 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 927 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 928 }
bc98a42c 929 if (!sb_rdonly(sb))
a8e25a83
AB		 930 ext4_commit_super(sb, 1);
931
ac27a0ec
DK		 932 for (i = 0; i < sbi->s_gdb_count; i++)
933 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 934 kvfree(sbi->s_group_desc);
935 kvfree(sbi->s_flex_groups);
57042651 936 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 937 percpu_counter_destroy(&sbi->s_freeinodes_counter);
938 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 939 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c8585c6f 940 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
ac27a0ec 941#ifdef CONFIG_QUOTA
a2d4a646 942 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 943 kfree(sbi->s_qf_names[i]);
944#endif
945
946 /* Debugging code just in case the in-memory inode orphan list
947 * isn't empty. The on-disk one can be non-empty if we've
948 * detected an error and taken the fs readonly, but the
949 * in-memory list had better be clean by this point. */
950 if (!list_empty(&sbi->s_orphan))
951 dump_orphan_list(sb, sbi);
952 J_ASSERT(list_empty(&sbi->s_orphan));
953
89d96a6f 954 sync_blockdev(sb->s_bdev);
f98393a6 955 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 956 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
957 /*
958 * Invalidate the journal device's buffers. We don't want them
959 * floating about in memory - the physical journal device may
960 * hotswapped, and it breaks the `ro-after' testing code.
961 */
962 sync_blockdev(sbi->journal_bdev);
f98393a6 963 invalidate_bdev(sbi->journal_bdev);
617ba13b 964 ext4_blkdev_remove(sbi);
ac27a0ec 965 }
dec214d0
TE		 966 if (sbi->s_ea_inode_cache) {
967 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
968 sbi->s_ea_inode_cache = NULL;
969 }
47387409
TE		 970 if (sbi->s_ea_block_cache) {
971 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
972 sbi->s_ea_block_cache = NULL;
9c191f70 973 }
c5e06d10
JL		 974 if (sbi->s_mmp_tsk)
975 kthread_stop(sbi->s_mmp_tsk);
9060dd2c 976 brelse(sbi->s_sbh);
ac27a0ec 977 sb->s_fs_info = NULL;
3197ebdb
TT		 978 /*
979 * Now that we are completely done shutting down the
980 * superblock, we need to actually destroy the kobject.
981 */
3197ebdb
TT		 982 kobject_put(&sbi->s_kobj);
983 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 984 if (sbi->s_chksum_driver)
985 crypto_free_shash(sbi->s_chksum_driver);
705895b6 986 kfree(sbi->s_blockgroup_lock);
5e405595 987 fs_put_dax(sbi->s_daxdev);
ac27a0ec 988 kfree(sbi);
ac27a0ec
DK		 989}
990
e18b890b 991static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 992
993/*
994 * Called inside transaction, so use GFP_NOFS
995 */
617ba13b 996static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 997{
617ba13b 998 struct ext4_inode_info *ei;
ac27a0ec 999
e6b4f8da 1000 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 1001 if (!ei)
1002 return NULL;
0b8e58a1 1003
ee73f9a5 1004 inode_set_iversion(&ei->vfs_inode, 1);
202ee5df 1005 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 1006 INIT_LIST_HEAD(&ei->i_prealloc_list);
1007 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 1008 ext4_es_init_tree(&ei->i_es_tree);
1009 rwlock_init(&ei->i_es_lock);
edaa53ca 1010 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 1011 ei->i_es_all_nr = 0;
edaa53ca 1012 ei->i_es_shk_nr = 0;
dd475925 1013 ei->i_es_shrink_lblk = 0;
d2a17637 1014 ei->i_reserved_data_blocks = 0;
9d0be502 1015 ei->i_da_metadata_calc_len = 0;
7e731bc9 1016 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 1017 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 1018#ifdef CONFIG_QUOTA
1019 ei->i_reserved_quota = 0;
96c7e0d9 1020 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 1021#endif
8aefcd55 1022 ei->jinode = NULL;
2e8fa54e 1023 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 1024 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 1025 ei->i_sync_tid = 0;
1026 ei->i_datasync_tid = 0;
e27f41e1 1027 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 1028 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
ac27a0ec
DK		 1029 return &ei->vfs_inode;
1030}
1031
7ff9c073
TT		 1032static int ext4_drop_inode(struct inode *inode)
1033{
1034 int drop = generic_drop_inode(inode);
1035
1036 trace_ext4_drop_inode(inode, drop);
1037 return drop;
1038}
1039
fa0d7e3d
NP		 1040static void ext4_i_callback(struct rcu_head *head)
1041{
1042 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 1043 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1044}
1045
617ba13b 1046static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1047{
9f7dd93d 1048 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1049 ext4_msg(inode->i_sb, KERN_ERR,
1050 "Inode %lu (%p): orphan list check failed!",
1051 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1052 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1053 EXT4_I(inode), sizeof(struct ext4_inode_info),
1054 true);
1055 dump_stack();
1056 }
fa0d7e3d 1057 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 1058}
1059
51cc5068 1060static void init_once(void *foo)
ac27a0ec 1061{
617ba13b 1062 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1063
a35afb83 1064 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1065 init_rwsem(&ei->xattr_sem);
0e855ac8 1066 init_rwsem(&ei->i_data_sem);
ea3d7209 1067 init_rwsem(&ei->i_mmap_sem);
a35afb83 1068 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1069}
1070
e67bc2b3 1071static int __init init_inodecache(void)
ac27a0ec 1072{
f8dd7c70
DW		 1073 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1074 sizeof(struct ext4_inode_info), 0,
1075 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1076 SLAB_ACCOUNT),
1077 offsetof(struct ext4_inode_info, i_data),
1078 sizeof_field(struct ext4_inode_info, i_data),
1079 init_once);
617ba13b 1080 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1081 return -ENOMEM;
1082 return 0;
1083}
1084
1085static void destroy_inodecache(void)
1086{
8c0a8537
KS		 1087 /*
1088 * Make sure all delayed rcu free inodes are flushed before we
1089 * destroy cache.
1090 */
1091 rcu_barrier();
617ba13b 1092 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1093}
1094
0930fcc1 1095void ext4_clear_inode(struct inode *inode)
ac27a0ec 1096{
0930fcc1 1097 invalidate_inode_buffers(inode);
dbd5768f 1098 clear_inode(inode);
9f754758 1099 dquot_drop(inode);
c2ea3fde 1100 ext4_discard_preallocations(inode);
51865fda 1101 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1102 if (EXT4_I(inode)->jinode) {
1103 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1104 EXT4_I(inode)->jinode);
1105 jbd2_free_inode(EXT4_I(inode)->jinode);
1106 EXT4_I(inode)->jinode = NULL;
1107 }
3d204e24 1108 fscrypt_put_encryption_info(inode);
ac27a0ec
DK		 1109}
1110
1b961ac0 1111static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1112 u64 ino, u32 generation)
ac27a0ec 1113{
ac27a0ec 1114 struct inode *inode;
ac27a0ec 1115
617ba13b 1116 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1117 return ERR_PTR(-ESTALE);
617ba13b 1118 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1119 return ERR_PTR(-ESTALE);
1120
1121 /* iget isn't really right if the inode is currently unallocated!!
1122 *
617ba13b 1123 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1124 * deleted, so we should be safe.
1125 *
1126 * Currently we don't know the generation for parent directory, so
1127 * a generation of 0 means "accept any"
1128 */
f4bb2981 1129 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1130 if (IS_ERR(inode))
1131 return ERR_CAST(inode);
1132 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1133 iput(inode);
1134 return ERR_PTR(-ESTALE);
1135 }
1b961ac0
CH		 1136
1137 return inode;
1138}
1139
1140static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1141 int fh_len, int fh_type)
1b961ac0
CH		 1142{
1143 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1144 ext4_nfs_get_inode);
1145}
1146
1147static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1148 int fh_len, int fh_type)
1b961ac0
CH		 1149{
1150 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1151 ext4_nfs_get_inode);
ac27a0ec
DK		 1152}
1153
c39a7f84
TO		 1154/*
1155 * Try to release metadata pages (indirect blocks, directories) which are
1156 * mapped via the block device. Since these pages could have journal heads
1157 * which would prevent try_to_free_buffers() from freeing them, we must use
1158 * jbd2 layer's try_to_free_buffers() function to release them.
1159 */
0b8e58a1
AD		 1160static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1161 gfp_t wait)
c39a7f84
TO		 1162{
1163 journal_t *journal = EXT4_SB(sb)->s_journal;
1164
1165 WARN_ON(PageChecked(page));
1166 if (!page_has_buffers(page))
1167 return 0;
1168 if (journal)
1169 return jbd2_journal_try_to_free_buffers(journal, page,
d0164adc 1170 wait & ~__GFP_DIRECT_RECLAIM);
c39a7f84
TO		 1171 return try_to_free_buffers(page);
1172}
1173
a7550b30
JK		 1174#ifdef CONFIG_EXT4_FS_ENCRYPTION
1175static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1176{
1177 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1178 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1179}
1180
a7550b30
JK		 1181static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1182 void *fs_data)
1183{
2f8f5e76 1184 handle_t *handle = fs_data;
c1a5d5f6 1185 int res, res2, credits, retries = 0;
2f8f5e76 1186
9ce0151a
EB		 1187 /*
1188 * Encrypting the root directory is not allowed because e2fsck expects
1189 * lost+found to exist and be unencrypted, and encrypting the root
1190 * directory would imply encrypting the lost+found directory as well as
1191 * the filename "lost+found" itself.
1192 */
1193 if (inode->i_ino == EXT4_ROOT_INO)
1194 return -EPERM;
2f8f5e76 1195
7d3e06a8
RZ		 1196 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1197 return -EINVAL;
1198
94840e3c
EB		 1199 res = ext4_convert_inline_data(inode);
1200 if (res)
1201 return res;
1202
2f8f5e76
EB		 1203 /*
1204 * If a journal handle was specified, then the encryption context is
1205 * being set on a new inode via inheritance and is part of a larger
1206 * transaction to create the inode. Otherwise the encryption context is
1207 * being set on an existing inode in its own transaction. Only in the
1208 * latter case should the "retry on ENOSPC" logic be used.
1209 */
a7550b30 1210
2f8f5e76
EB		 1211 if (handle) {
1212 res = ext4_xattr_set_handle(handle, inode,
1213 EXT4_XATTR_INDEX_ENCRYPTION,
1214 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1215 ctx, len, 0);
a7550b30
JK		 1216 if (!res) {
1217 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1218 ext4_clear_inode_state(inode,
1219 EXT4_STATE_MAY_INLINE_DATA);
a3caa24b 1220 /*
2ee6a576
EB		 1221 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1222 * S_DAX may be disabled
a3caa24b
JK		 1223 */
1224 ext4_set_inode_flags(inode);
a7550b30
JK		 1225 }
1226 return res;
1227 }
1228
b8cb5a54
TE		 1229 res = dquot_initialize(inode);
1230 if (res)
1231 return res;
2f8f5e76 1232retry:
af65207c
TE		 1233 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1234 &credits);
dec214d0
TE		 1235 if (res)
1236 return res;
1237
c1a5d5f6 1238 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
a7550b30
JK		 1239 if (IS_ERR(handle))
1240 return PTR_ERR(handle);
1241
2f8f5e76
EB		 1242 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1243 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1244 ctx, len, 0);
a7550b30
JK		 1245 if (!res) {
1246 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
2ee6a576
EB		 1247 /*
1248 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1249 * S_DAX may be disabled
1250 */
a3caa24b 1251 ext4_set_inode_flags(inode);
a7550b30
JK		 1252 res = ext4_mark_inode_dirty(handle, inode);
1253 if (res)
1254 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1255 }
1256 res2 = ext4_journal_stop(handle);
2f8f5e76
EB		 1257
1258 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1259 goto retry;
a7550b30
JK		 1260 if (!res)
1261 res = res2;
1262 return res;
1263}
1264
c250b7dd 1265static bool ext4_dummy_context(struct inode *inode)
a7550b30
JK		 1266{
1267 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1268}
1269
1270static unsigned ext4_max_namelen(struct inode *inode)
1271{
1272 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1273 EXT4_NAME_LEN;
1274}
1275
6f69f0ed 1276static const struct fscrypt_operations ext4_cryptops = {
a5d431ef 1277 .key_prefix = "ext4:",
a7550b30 1278 .get_context = ext4_get_context,
a7550b30
JK		 1279 .set_context = ext4_set_context,
1280 .dummy_context = ext4_dummy_context,
a7550b30
JK		 1281 .empty_dir = ext4_empty_dir,
1282 .max_namelen = ext4_max_namelen,
1283};
a7550b30
JK		 1284#endif
1285
ac27a0ec 1286#ifdef CONFIG_QUOTA
d6006186 1287static const char * const quotatypes[] = INITQFNAMES;
689c958c 1288#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1289
617ba13b
MC		 1290static int ext4_write_dquot(struct dquot *dquot);
1291static int ext4_acquire_dquot(struct dquot *dquot);
1292static int ext4_release_dquot(struct dquot *dquot);
1293static int ext4_mark_dquot_dirty(struct dquot *dquot);
1294static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1295static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1296 const struct path *path);
617ba13b
MC		 1297static int ext4_quota_on_mount(struct super_block *sb, int type);
1298static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1299 size_t len, loff_t off);
617ba13b 1300static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1301 const char *data, size_t len, loff_t off);
7c319d32
AK		 1302static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1303 unsigned int flags);
1304static int ext4_enable_quotas(struct super_block *sb);
8f0e8746 1305static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
ac27a0ec 1306
96c7e0d9
JK		 1307static struct dquot **ext4_get_dquots(struct inode *inode)
1308{
1309 return EXT4_I(inode)->i_dquot;
1310}
1311
61e225dc 1312static const struct dquot_operations ext4_quota_operations = {
7a9ca53a
TE		 1313 .get_reserved_space = ext4_get_reserved_space,
1314 .write_dquot = ext4_write_dquot,
1315 .acquire_dquot = ext4_acquire_dquot,
1316 .release_dquot = ext4_release_dquot,
1317 .mark_dirty = ext4_mark_dquot_dirty,
1318 .write_info = ext4_write_info,
1319 .alloc_dquot = dquot_alloc,
1320 .destroy_dquot = dquot_destroy,
1321 .get_projid = ext4_get_projid,
1322 .get_inode_usage = ext4_get_inode_usage,
1323 .get_next_id = ext4_get_next_id,
ac27a0ec
DK		 1324};
1325
0d54b217 1326static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1327 .quota_on = ext4_quota_on,
ca0e05e4 1328 .quota_off = ext4_quota_off,
287a8095 1329 .quota_sync = dquot_quota_sync,
0a240339 1330 .get_state = dquot_get_state,
287a8095
CH		 1331 .set_info = dquot_set_dqinfo,
1332 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1333 .set_dqblk = dquot_set_dqblk,
1334 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1335};
1336#endif
1337
ee9b6d61 1338static const struct super_operations ext4_sops = {
617ba13b
MC		 1339 .alloc_inode = ext4_alloc_inode,
1340 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1341 .write_inode = ext4_write_inode,
1342 .dirty_inode = ext4_dirty_inode,
7ff9c073 1343 .drop_inode = ext4_drop_inode,
0930fcc1 1344 .evict_inode = ext4_evict_inode,
617ba13b 1345 .put_super = ext4_put_super,
617ba13b 1346 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1347 .freeze_fs = ext4_freeze,
1348 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1349 .statfs = ext4_statfs,
1350 .remount_fs = ext4_remount,
617ba13b 1351 .show_options = ext4_show_options,
ac27a0ec 1352#ifdef CONFIG_QUOTA
617ba13b
MC		 1353 .quota_read = ext4_quota_read,
1354 .quota_write = ext4_quota_write,
96c7e0d9 1355 .get_dquots = ext4_get_dquots,
ac27a0ec 1356#endif
c39a7f84 1357 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1358};
1359
39655164 1360static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1361 .fh_to_dentry = ext4_fh_to_dentry,
1362 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1363 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1364};
1365
1366enum {
1367 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1368 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1369 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1370 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1371 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1372 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1373 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1374 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1375 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1376 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1377 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1378 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
49da9392 1379 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1449032b 1380 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
670e9875 1381 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1449032b 1382 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1383 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1384 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1385 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1386 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
ac27a0ec
DK		 1387};
1388
a447c093 1389static const match_table_t tokens = {
ac27a0ec
DK		 1390 {Opt_bsd_df, "bsddf"},
1391 {Opt_minix_df, "minixdf"},
1392 {Opt_grpid, "grpid"},
1393 {Opt_grpid, "bsdgroups"},
1394 {Opt_nogrpid, "nogrpid"},
1395 {Opt_nogrpid, "sysvgroups"},
1396 {Opt_resgid, "resgid=%u"},
1397 {Opt_resuid, "resuid=%u"},
1398 {Opt_sb, "sb=%u"},
1399 {Opt_err_cont, "errors=continue"},
1400 {Opt_err_panic, "errors=panic"},
1401 {Opt_err_ro, "errors=remount-ro"},
1402 {Opt_nouid32, "nouid32"},
ac27a0ec 1403 {Opt_debug, "debug"},
72578c33
TT		 1404 {Opt_removed, "oldalloc"},
1405 {Opt_removed, "orlov"},
ac27a0ec
DK		 1406 {Opt_user_xattr, "user_xattr"},
1407 {Opt_nouser_xattr, "nouser_xattr"},
1408 {Opt_acl, "acl"},
1409 {Opt_noacl, "noacl"},
e3bb52ae 1410 {Opt_noload, "norecovery"},
5a916be1 1411 {Opt_noload, "noload"},
72578c33
TT		 1412 {Opt_removed, "nobh"},
1413 {Opt_removed, "bh"},
ac27a0ec 1414 {Opt_commit, "commit=%u"},
30773840
TT		 1415 {Opt_min_batch_time, "min_batch_time=%u"},
1416 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1417 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1418 {Opt_journal_path, "journal_path=%s"},
818d276c 1419 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1420 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1421 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1422 {Opt_abort, "abort"},
1423 {Opt_data_journal, "data=journal"},
1424 {Opt_data_ordered, "data=ordered"},
1425 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1426 {Opt_data_err_abort, "data_err=abort"},
1427 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1428 {Opt_offusrjquota, "usrjquota="},
1429 {Opt_usrjquota, "usrjquota=%s"},
1430 {Opt_offgrpjquota, "grpjquota="},
1431 {Opt_grpjquota, "grpjquota=%s"},
1432 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1433 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1434 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1435 {Opt_grpquota, "grpquota"},
1436 {Opt_noquota, "noquota"},
1437 {Opt_quota, "quota"},
1438 {Opt_usrquota, "usrquota"},
49da9392 1439 {Opt_prjquota, "prjquota"},
ac27a0ec 1440 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1441 {Opt_barrier, "barrier"},
1442 {Opt_nobarrier, "nobarrier"},
25ec56b5 1443 {Opt_i_version, "i_version"},
923ae0ff 1444 {Opt_dax, "dax"},
c9de560d 1445 {Opt_stripe, "stripe=%u"},
64769240 1446 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1447 {Opt_lazytime, "lazytime"},
1448 {Opt_nolazytime, "nolazytime"},
670e9875 1449 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
dd919b98 1450 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1451 {Opt_removed, "mblk_io_submit"},
1452 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1453 {Opt_block_validity, "block_validity"},
1454 {Opt_noblock_validity, "noblock_validity"},
240799cd 1455 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1456 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1457 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1458 {Opt_auto_da_alloc, "auto_da_alloc"},
1459 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1460 {Opt_dioread_nolock, "dioread_nolock"},
1461 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1462 {Opt_discard, "discard"},
1463 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1464 {Opt_init_itable, "init_itable=%u"},
1465 {Opt_init_itable, "init_itable"},
1466 {Opt_noinit_itable, "noinit_itable"},
df981d03 1467 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1468 {Opt_test_dummy_encryption, "test_dummy_encryption"},
cdb7ee4c
TE		 1469 {Opt_nombcache, "nombcache"},
1470 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
c7198b9c
TT		 1471 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1472 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1473 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1474 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1475 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1476 {Opt_err, NULL},
ac27a0ec
DK		 1477};
1478
617ba13b 1479static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1480{
617ba13b 1481 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1482 char *options = (char *) *data;
1483
1484 if (!options || strncmp(options, "sb=", 3) != 0)
1485 return 1; /* Default location */
0b8e58a1 1486
ac27a0ec 1487 options += 3;
0b8e58a1 1488 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1489 sb_block = simple_strtoul(options, &options, 0);
1490 if (*options && *options != ',') {
4776004f 1491 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1492 (char *) *data);
1493 return 1;
1494 }
1495 if (*options == ',')
1496 options++;
1497 *data = (void *) options;
0b8e58a1 1498
ac27a0ec
DK		 1499 return sb_block;
1500}
1501
b3881f74 1502#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
d6006186
EB		 1503static const char deprecated_msg[] =
1504 "Mount option \"%s\" will be removed by %s\n"
437ca0fd 1505 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1506
56c50f11
DM		 1507#ifdef CONFIG_QUOTA
1508static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1509{
1510 struct ext4_sb_info *sbi = EXT4_SB(sb);
1511 char *qname;
03dafb5f 1512 int ret = -1;
56c50f11
DM		 1513
1514 if (sb_any_quota_loaded(sb) &&
1515 !sbi->s_qf_names[qtype]) {
1516 ext4_msg(sb, KERN_ERR,
1517 "Cannot change journaled "
1518 "quota options when quota turned on");
57f73c2c 1519 return -1;
56c50f11 1520 }
e2b911c5 1521 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1522 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1523 "ignored when QUOTA feature is enabled");
1524 return 1;
262b4662 1525 }
56c50f11
DM		 1526 qname = match_strdup(args);
1527 if (!qname) {
1528 ext4_msg(sb, KERN_ERR,
1529 "Not enough memory for storing quotafile name");
57f73c2c 1530 return -1;
56c50f11 1531 }
03dafb5f
CG		 1532 if (sbi->s_qf_names[qtype]) {
1533 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1534 ret = 1;
1535 else
1536 ext4_msg(sb, KERN_ERR,
1537 "%s quota file already specified",
1538 QTYPE2NAME(qtype));
1539 goto errout;
56c50f11 1540 }
03dafb5f 1541 if (strchr(qname, '/')) {
56c50f11
DM		 1542 ext4_msg(sb, KERN_ERR,
1543 "quotafile must be on filesystem root");
03dafb5f 1544 goto errout;
56c50f11 1545 }
03dafb5f 1546 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1547 set_opt(sb, QUOTA);
56c50f11 1548 return 1;
03dafb5f
CG		 1549errout:
1550 kfree(qname);
1551 return ret;
56c50f11
DM		 1552}
1553
1554static int clear_qf_name(struct super_block *sb, int qtype)
1555{
1556
1557 struct ext4_sb_info *sbi = EXT4_SB(sb);
1558
1559 if (sb_any_quota_loaded(sb) &&
1560 sbi->s_qf_names[qtype]) {
1561 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1562 " when quota turned on");
57f73c2c 1563 return -1;
56c50f11 1564 }
03dafb5f 1565 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1566 sbi->s_qf_names[qtype] = NULL;
1567 return 1;
1568}
1569#endif
1570
26092bf5
TT		 1571#define MOPT_SET 0x0001
1572#define MOPT_CLEAR 0x0002
1573#define MOPT_NOSUPPORT 0x0004
1574#define MOPT_EXPLICIT 0x0008
1575#define MOPT_CLEAR_ERR 0x0010
1576#define MOPT_GTE0 0x0020
ac27a0ec 1577#ifdef CONFIG_QUOTA
26092bf5
TT		 1578#define MOPT_Q 0
1579#define MOPT_QFMT 0x0040
1580#else
1581#define MOPT_Q MOPT_NOSUPPORT
1582#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1583#endif
26092bf5 1584#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1585#define MOPT_NO_EXT2 0x0100
1586#define MOPT_NO_EXT3 0x0200
1587#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1588#define MOPT_STRING 0x0400
26092bf5
TT		 1589
1590static const struct mount_opts {
1591 int token;
1592 int mount_opt;
1593 int flags;
1594} ext4_mount_opts[] = {
1595 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1596 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1597 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1598 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1599 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1600 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1601 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1602 MOPT_EXT4_ONLY | MOPT_SET},
1603 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1604 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1605 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1606 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1607 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1608 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1609 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1610 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1611 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1612 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1613 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1614 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1615 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1616 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1617 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1618 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1619 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1620 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1621 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1622 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1623 MOPT_NO_EXT2},
8dc0aa8c 1624 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1625 MOPT_NO_EXT2},
26092bf5
TT		 1626 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1627 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1628 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1629 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1630 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1631 {Opt_commit, 0, MOPT_GTE0},
1632 {Opt_max_batch_time, 0, MOPT_GTE0},
1633 {Opt_min_batch_time, 0, MOPT_GTE0},
1634 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1635 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1636 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1637 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1638 {Opt_resuid, 0, MOPT_GTE0},
1639 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1640 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1641 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1642 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1643 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1644 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1645 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1646 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1647 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1648 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1649#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1650 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1651 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1652#else
26092bf5
TT		 1653 {Opt_acl, 0, MOPT_NOSUPPORT},
1654 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1655#endif
26092bf5
TT		 1656 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1657 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
670e9875 1658 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
26092bf5
TT		 1659 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1660 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1661 MOPT_SET | MOPT_Q},
1662 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1663 MOPT_SET | MOPT_Q},
49da9392
JK		 1664 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1665 MOPT_SET | MOPT_Q},
26092bf5 1666 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1667 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1668 MOPT_CLEAR | MOPT_Q},
26092bf5
TT		 1669 {Opt_usrjquota, 0, MOPT_Q},
1670 {Opt_grpjquota, 0, MOPT_Q},
1671 {Opt_offusrjquota, 0, MOPT_Q},
1672 {Opt_offgrpjquota, 0, MOPT_Q},
1673 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1674 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1675 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1676 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1677 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
cdb7ee4c 1678 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
26092bf5
TT		 1679 {Opt_err, 0, 0}
1680};
1681
1682static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1683 substring_t *args, unsigned long *journal_devnum,
1684 unsigned int *journal_ioprio, int is_remount)
1685{
1686 struct ext4_sb_info *sbi = EXT4_SB(sb);
1687 const struct mount_opts *m;
08cefc7a
EB		 1688 kuid_t uid;
1689 kgid_t gid;
26092bf5
TT		 1690 int arg = 0;
1691
57f73c2c
TT		 1692#ifdef CONFIG_QUOTA
1693 if (token == Opt_usrjquota)
1694 return set_qf_name(sb, USRQUOTA, &args[0]);
1695 else if (token == Opt_grpjquota)
1696 return set_qf_name(sb, GRPQUOTA, &args[0]);
1697 else if (token == Opt_offusrjquota)
1698 return clear_qf_name(sb, USRQUOTA);
1699 else if (token == Opt_offgrpjquota)
1700 return clear_qf_name(sb, GRPQUOTA);
1701#endif
26092bf5 1702 switch (token) {
f7048605
TT		 1703 case Opt_noacl:
1704 case Opt_nouser_xattr:
1705 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1706 break;
26092bf5
TT		 1707 case Opt_sb:
1708 return 1; /* handled by get_sb_block() */
1709 case Opt_removed:
5f3633e3 1710 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1711 return 1;
26092bf5
TT		 1712 case Opt_abort:
1713 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1714 return 1;
1715 case Opt_i_version:
357fdad0 1716 sb->s_flags |= SB_I_VERSION;
26092bf5 1717 return 1;
a26f4992 1718 case Opt_lazytime:
1751e8a6 1719 sb->s_flags |= SB_LAZYTIME;
a26f4992
TT		 1720 return 1;
1721 case Opt_nolazytime:
1751e8a6 1722 sb->s_flags &= ~SB_LAZYTIME;
a26f4992 1723 return 1;
26092bf5
TT		 1724 }
1725
5f3633e3
JK		 1726 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1727 if (token == m->token)
1728 break;
1729
1730 if (m->token == Opt_err) {
1731 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1732 "or missing value", opt);
1733 return -1;
1734 }
1735
8dc0aa8c
TT		 1736 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1737 ext4_msg(sb, KERN_ERR,
1738 "Mount option \"%s\" incompatible with ext2", opt);
1739 return -1;
1740 }
1741 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1742 ext4_msg(sb, KERN_ERR,
1743 "Mount option \"%s\" incompatible with ext3", opt);
1744 return -1;
1745 }
1746
ad4eec61 1747 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1748 return -1;
1749 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1750 return -1;
c93cf2d7
DM		 1751 if (m->flags & MOPT_EXPLICIT) {
1752 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1753 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1754 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1755 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1756 } else
1757 return -1;
1758 }
5f3633e3
JK		 1759 if (m->flags & MOPT_CLEAR_ERR)
1760 clear_opt(sb, ERRORS_MASK);
1761 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1762 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1763 "options when quota turned on");
1764 return -1;
1765 }
1766
1767 if (m->flags & MOPT_NOSUPPORT) {
1768 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1769 } else if (token == Opt_commit) {
1770 if (arg == 0)
1771 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1772 sbi->s_commit_interval = HZ * arg;
670e9875
TT		 1773 } else if (token == Opt_debug_want_extra_isize) {
1774 sbi->s_want_extra_isize = arg;
5f3633e3 1775 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1776 sbi->s_max_batch_time = arg;
1777 } else if (token == Opt_min_batch_time) {
1778 sbi->s_min_batch_time = arg;
1779 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1780 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1781 ext4_msg(sb, KERN_ERR,
1782 "EXT4-fs: inode_readahead_blks must be "
1783 "0 or a power of 2 smaller than 2^31");
26092bf5 1784 return -1;
5f3633e3
JK		 1785 }
1786 sbi->s_inode_readahead_blks = arg;
1787 } else if (token == Opt_init_itable) {
1788 set_opt(sb, INIT_INODE_TABLE);
1789 if (!args->from)
1790 arg = EXT4_DEF_LI_WAIT_MULT;
1791 sbi->s_li_wait_mult = arg;
1792 } else if (token == Opt_max_dir_size_kb) {
1793 sbi->s_max_dir_size_kb = arg;
1794 } else if (token == Opt_stripe) {
1795 sbi->s_stripe = arg;
1796 } else if (token == Opt_resuid) {
1797 uid = make_kuid(current_user_ns(), arg);
1798 if (!uid_valid(uid)) {
1799 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1800 return -1;
1801 }
5f3633e3
JK		 1802 sbi->s_resuid = uid;
1803 } else if (token == Opt_resgid) {
1804 gid = make_kgid(current_user_ns(), arg);
1805 if (!gid_valid(gid)) {
1806 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1807 return -1;
1808 }
1809 sbi->s_resgid = gid;
1810 } else if (token == Opt_journal_dev) {
1811 if (is_remount) {
1812 ext4_msg(sb, KERN_ERR,
1813 "Cannot specify journal on remount");
1814 return -1;
1815 }
1816 *journal_devnum = arg;
ad4eec61
ES		 1817 } else if (token == Opt_journal_path) {
1818 char *journal_path;
1819 struct inode *journal_inode;
1820 struct path path;
1821 int error;
1822
1823 if (is_remount) {
1824 ext4_msg(sb, KERN_ERR,
1825 "Cannot specify journal on remount");
1826 return -1;
1827 }
1828 journal_path = match_strdup(&args[0]);
1829 if (!journal_path) {
1830 ext4_msg(sb, KERN_ERR, "error: could not dup "
1831 "journal device string");
1832 return -1;
1833 }
1834
1835 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1836 if (error) {
1837 ext4_msg(sb, KERN_ERR, "error: could not find "
1838 "journal device path: error %d", error);
1839 kfree(journal_path);
1840 return -1;
1841 }
1842
2b0143b5 1843 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1844 if (!S_ISBLK(journal_inode->i_mode)) {
1845 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1846 "is not a block device", journal_path);
1847 path_put(&path);
1848 kfree(journal_path);
1849 return -1;
1850 }
1851
1852 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1853 path_put(&path);
1854 kfree(journal_path);
5f3633e3
JK		 1855 } else if (token == Opt_journal_ioprio) {
1856 if (arg > 7) {
1857 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1858 " (must be 0-7)");
1859 return -1;
1860 }
1861 *journal_ioprio =
1862 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1863 } else if (token == Opt_test_dummy_encryption) {
1864#ifdef CONFIG_EXT4_FS_ENCRYPTION
1865 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1866 ext4_msg(sb, KERN_WARNING,
1867 "Test dummy encryption mode enabled");
1868#else
1869 ext4_msg(sb, KERN_WARNING,
1870 "Test dummy encryption mount option ignored");
1871#endif
5f3633e3
JK		 1872 } else if (m->flags & MOPT_DATAJ) {
1873 if (is_remount) {
1874 if (!sbi->s_journal)
1875 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1876 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1877 ext4_msg(sb, KERN_ERR,
26092bf5 1878 "Cannot change data mode on remount");
26092bf5 1879 return -1;
ac27a0ec 1880 }
26092bf5 1881 } else {
5f3633e3
JK		 1882 clear_opt(sb, DATA_FLAGS);
1883 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1884 }
5f3633e3
JK		 1885#ifdef CONFIG_QUOTA
1886 } else if (m->flags & MOPT_QFMT) {
1887 if (sb_any_quota_loaded(sb) &&
1888 sbi->s_jquota_fmt != m->mount_opt) {
1889 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1890 "quota options when quota turned on");
1891 return -1;
1892 }
e2b911c5 1893 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1894 ext4_msg(sb, KERN_INFO,
1895 "Quota format mount options ignored "
262b4662 1896 "when QUOTA feature is enabled");
c325a67c 1897 return 1;
262b4662 1898 }
5f3633e3 1899 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 1900#endif
923ae0ff 1901 } else if (token == Opt_dax) {
ef83b6e8
DW		 1902#ifdef CONFIG_FS_DAX
1903 ext4_msg(sb, KERN_WARNING,
1904 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1905 sbi->s_mount_opt |= m->mount_opt;
1906#else
923ae0ff
RZ		 1907 ext4_msg(sb, KERN_INFO, "dax option not supported");
1908 return -1;
5f3633e3 1909#endif
7915a861
AN		 1910 } else if (token == Opt_data_err_abort) {
1911 sbi->s_mount_opt |= m->mount_opt;
1912 } else if (token == Opt_data_err_ignore) {
1913 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 1914 } else {
1915 if (!args->from)
1916 arg = 1;
1917 if (m->flags & MOPT_CLEAR)
1918 arg = !arg;
1919 else if (unlikely(!(m->flags & MOPT_SET))) {
1920 ext4_msg(sb, KERN_WARNING,
1921 "buggy handling of option %s", opt);
1922 WARN_ON(1);
1923 return -1;
1924 }
1925 if (arg != 0)
1926 sbi->s_mount_opt |= m->mount_opt;
1927 else
1928 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1929 }
5f3633e3 1930 return 1;
26092bf5
TT		 1931}
1932
1933static int parse_options(char *options, struct super_block *sb,
1934 unsigned long *journal_devnum,
1935 unsigned int *journal_ioprio,
1936 int is_remount)
1937{
1938 struct ext4_sb_info *sbi = EXT4_SB(sb);
1939 char *p;
1940 substring_t args[MAX_OPT_ARGS];
1941 int token;
1942
1943 if (!options)
1944 return 1;
1945
1946 while ((p = strsep(&options, ",")) != NULL) {
1947 if (!*p)
1948 continue;
1949 /*
1950 * Initialize args struct so we know whether arg was
1951 * found; some options take optional arguments.
1952 */
caecd0af 1953 args[0].to = args[0].from = NULL;
26092bf5
TT		 1954 token = match_token(p, tokens, args);
1955 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1956 journal_ioprio, is_remount) < 0)
1957 return 0;
ac27a0ec
DK		 1958 }
1959#ifdef CONFIG_QUOTA
49da9392
JK		 1960 /*
1961 * We do the test below only for project quotas. 'usrquota' and
1962 * 'grpquota' mount options are allowed even without quota feature
1963 * to support legacy quotas in quota files.
1964 */
1965 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1966 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1967 "Cannot enable project quota enforcement.");
1968 return 0;
1969 }
1970 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1971 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1972 clear_opt(sb, USRQUOTA);
ac27a0ec 1973
482a7425 1974 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1975 clear_opt(sb, GRPQUOTA);
ac27a0ec 1976
56c50f11 1977 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1978 ext4_msg(sb, KERN_ERR, "old and new quota "
1979 "format mixing");
ac27a0ec
DK		 1980 return 0;
1981 }
1982
1983 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1984 ext4_msg(sb, KERN_ERR, "journaled quota format "
1985 "not specified");
ac27a0ec
DK		 1986 return 0;
1987 }
ac27a0ec
DK		 1988 }
1989#endif
261cb20c
JK		 1990 if (test_opt(sb, DIOREAD_NOLOCK)) {
1991 int blocksize =
1992 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1993
09cbfeaf 1994 if (blocksize < PAGE_SIZE) {
261cb20c
JK		 1995 ext4_msg(sb, KERN_ERR, "can't mount with "
1996 "dioread_nolock if block size != PAGE_SIZE");
1997 return 0;
1998 }
1999 }
ac27a0ec
DK		 2000 return 1;
2001}
2002
2adf6da8
TT		 2003static inline void ext4_show_quota_options(struct seq_file *seq,
2004 struct super_block *sb)
2005{
2006#if defined(CONFIG_QUOTA)
2007 struct ext4_sb_info *sbi = EXT4_SB(sb);
2008
2009 if (sbi->s_jquota_fmt) {
2010 char *fmtname = "";
2011
2012 switch (sbi->s_jquota_fmt) {
2013 case QFMT_VFS_OLD:
2014 fmtname = "vfsold";
2015 break;
2016 case QFMT_VFS_V0:
2017 fmtname = "vfsv0";
2018 break;
2019 case QFMT_VFS_V1:
2020 fmtname = "vfsv1";
2021 break;
2022 }
2023 seq_printf(seq, ",jqfmt=%s", fmtname);
2024 }
2025
2026 if (sbi->s_qf_names[USRQUOTA])
a068acf2 2027 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 2028
2029 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 2030 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 2031#endif
2032}
2033
5a916be1
TT		 2034static const char *token2str(int token)
2035{
50df9fd5 2036 const struct match_token *t;
5a916be1
TT		 2037
2038 for (t = tokens; t->token != Opt_err; t++)
2039 if (t->token == token && !strchr(t->pattern, '='))
2040 break;
2041 return t->pattern;
2042}
2043
2adf6da8
TT		 2044/*
2045 * Show an option if
2046 * - it's set to a non-default value OR
2047 * - if the per-sb default is different from the global default
2048 */
66acdcf4
TT		 2049static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2050 int nodefs)
2adf6da8 2051{
2adf6da8
TT		 2052 struct ext4_sb_info *sbi = EXT4_SB(sb);
2053 struct ext4_super_block *es = sbi->s_es;
68afa7e0 2054 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
5a916be1 2055 const struct mount_opts *m;
66acdcf4 2056 char sep = nodefs ? '\n' : ',';
2adf6da8 2057
66acdcf4
TT		 2058#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2059#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2060
2061 if (sbi->s_sb_block != 1)
5a916be1
TT		 2062 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2063
2064 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2065 int want_set = m->flags & MOPT_SET;
2066 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2067 (m->flags & MOPT_CLEAR_ERR))
2068 continue;
68afa7e0 2069 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 2070 continue; /* skip if same as the default */
2071 if ((want_set &&
2072 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2073 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2074 continue; /* select Opt_noFoo vs Opt_Foo */
2075 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2076 }
5a916be1 2077
08cefc7a 2078 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2079 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2080 SEQ_OPTS_PRINT("resuid=%u",
2081 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2082 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2083 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2084 SEQ_OPTS_PRINT("resgid=%u",
2085 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2086 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2087 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2088 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2089 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2090 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2091 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2092 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2093 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2094 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2095 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2096 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2097 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2098 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
357fdad0 2099 if (sb->s_flags & SB_I_VERSION)
5a916be1 2100 SEQ_OPTS_PUTS("i_version");
66acdcf4 2101 if (nodefs || sbi->s_stripe)
5a916be1 2102 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
68afa7e0
TN		 2103 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2104 (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2105 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2106 SEQ_OPTS_PUTS("data=journal");
2107 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2108 SEQ_OPTS_PUTS("data=ordered");
2109 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2110 SEQ_OPTS_PUTS("data=writeback");
2111 }
66acdcf4
TT		 2112 if (nodefs ||
2113 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2114 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2115 sbi->s_inode_readahead_blks);
2adf6da8 2116
ceec0376 2117 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
66acdcf4 2118 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2119 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2120 if (nodefs || sbi->s_max_dir_size_kb)
2121 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2122 if (test_opt(sb, DATA_ERR_ABORT))
2123 SEQ_OPTS_PUTS("data_err=abort");
2adf6da8
TT		 2124
2125 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2126 return 0;
2127}
2128
66acdcf4
TT		 2129static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2130{
2131 return _ext4_show_options(seq, root->d_sb, 0);
2132}
2133
ebd173be 2134int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2135{
2136 struct super_block *sb = seq->private;
2137 int rc;
2138
bc98a42c 2139 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
66acdcf4
TT		 2140 rc = _ext4_show_options(seq, sb, 1);
2141 seq_puts(seq, "\n");
2142 return rc;
2143}
2144
617ba13b 2145static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2146 int read_only)
2147{
617ba13b 2148 struct ext4_sb_info *sbi = EXT4_SB(sb);
c89128a0 2149 int err = 0;
ac27a0ec 2150
617ba13b 2151 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2152 ext4_msg(sb, KERN_ERR, "revision level too high, "
2153 "forcing read-only mode");
c89128a0 2154 err = -EROFS;
ac27a0ec
DK		 2155 }
2156 if (read_only)
281b5995 2157 goto done;
617ba13b 2158 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2159 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2160 "running e2fsck is recommended");
c8b459f4 2161 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2162 ext4_msg(sb, KERN_WARNING,
2163 "warning: mounting fs with errors, "
2164 "running e2fsck is recommended");
ed3ce80a 2165 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2166 le16_to_cpu(es->s_mnt_count) >=
2167 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2168 ext4_msg(sb, KERN_WARNING,
2169 "warning: maximal mount count reached, "
2170 "running e2fsck is recommended");
ac27a0ec
DK		 2171 else if (le32_to_cpu(es->s_checkinterval) &&
2172 (le32_to_cpu(es->s_lastcheck) +
2173 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 2174 ext4_msg(sb, KERN_WARNING,
2175 "warning: checktime reached, "
2176 "running e2fsck is recommended");
0b8e58a1 2177 if (!sbi->s_journal)
0390131b 2178 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2179 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2180 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2181 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 2182 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 2183 ext4_update_dynamic_rev(sb);
0390131b 2184 if (sbi->s_journal)
e2b911c5 2185 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2186
c89128a0 2187 err = ext4_commit_super(sb, 1);
281b5995 2188done:
ac27a0ec 2189 if (test_opt(sb, DEBUG))
a9df9a49 2190 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2191 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2192 sb->s_blocksize,
2193 sbi->s_groups_count,
617ba13b
MC		 2194 EXT4_BLOCKS_PER_GROUP(sb),
2195 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2196 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2197
7abc52c2 2198 cleancache_init_fs(sb);
c89128a0 2199 return err;
ac27a0ec
DK		 2200}
2201
117fff10
TT		 2202int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2203{
2204 struct ext4_sb_info *sbi = EXT4_SB(sb);
2205 struct flex_groups *new_groups;
2206 int size;
2207
2208 if (!sbi->s_log_groups_per_flex)
2209 return 0;
2210
2211 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2212 if (size <= sbi->s_flex_groups_allocated)
2213 return 0;
2214
2215 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
a7c3e901 2216 new_groups = kvzalloc(size, GFP_KERNEL);
117fff10
TT		 2217 if (!new_groups) {
2218 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2219 size / (int) sizeof(struct flex_groups));
2220 return -ENOMEM;
2221 }
2222
2223 if (sbi->s_flex_groups) {
2224 memcpy(new_groups, sbi->s_flex_groups,
2225 (sbi->s_flex_groups_allocated *
2226 sizeof(struct flex_groups)));
b93b41d4 2227 kvfree(sbi->s_flex_groups);
117fff10
TT		 2228 }
2229 sbi->s_flex_groups = new_groups;
2230 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2231 return 0;
2232}
2233
772cb7c8
JS		 2234static int ext4_fill_flex_info(struct super_block *sb)
2235{
2236 struct ext4_sb_info *sbi = EXT4_SB(sb);
2237 struct ext4_group_desc *gdp = NULL;
772cb7c8 2238 ext4_group_t flex_group;
117fff10 2239 int i, err;
772cb7c8 2240
503358ae 2241 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2242 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2243 sbi->s_log_groups_per_flex = 0;
2244 return 1;
2245 }
2246
117fff10
TT		 2247 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2248 if (err)
9933fc0a 2249 goto failed;
772cb7c8 2250
772cb7c8 2251 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2252 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2253
2254 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2255 atomic_add(ext4_free_inodes_count(sb, gdp),
2256 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2257 atomic64_add(ext4_free_group_clusters(sb, gdp),
2258 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2259 atomic_add(ext4_used_dirs_count(sb, gdp),
2260 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2261 }
2262
2263 return 1;
2264failed:
2265 return 0;
2266}
2267
e2b911c5 2268static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2269 struct ext4_group_desc *gdp)
717d50e4 2270{
b47820ed 2271 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2272 __u16 crc = 0;
feb0ab32 2273 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2274 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2275
9aa5d32b 2276 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2277 /* Use new metadata_csum algorithm */
feb0ab32 2278 __u32 csum32;
b47820ed 2279 __u16 dummy_csum = 0;
feb0ab32 2280
feb0ab32
DW		 2281 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2282 sizeof(le_group));
b47820ed
DJ		 2283 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2284 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2285 sizeof(dummy_csum));
2286 offset += sizeof(dummy_csum);
2287 if (offset < sbi->s_desc_size)
2288 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2289 sbi->s_desc_size - offset);
feb0ab32
DW		 2290
2291 crc = csum32 & 0xFFFF;
2292 goto out;
717d50e4
AD		 2293 }
2294
feb0ab32 2295 /* old crc16 code */
e2b911c5 2296 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2297 return 0;
2298
feb0ab32
DW		 2299 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2300 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2301 crc = crc16(crc, (__u8 *)gdp, offset);
2302 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2303 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2304 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2305 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2306 crc = crc16(crc, (__u8 *)gdp + offset,
2307 le16_to_cpu(sbi->s_es->s_desc_size) -
2308 offset);
2309
2310out:
717d50e4
AD		 2311 return cpu_to_le16(crc);
2312}
2313
feb0ab32 2314int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2315 struct ext4_group_desc *gdp)
2316{
feb0ab32 2317 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2318 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2319 return 0;
2320
2321 return 1;
2322}
2323
feb0ab32
DW		 2324void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2325 struct ext4_group_desc *gdp)
2326{
2327 if (!ext4_has_group_desc_csum(sb))
2328 return;
e2b911c5 2329 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2330}
2331
ac27a0ec 2332/* Called at mount-time, super-block is locked */
bfff6873 2333static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2334 ext4_fsblk_t sb_block,
bfff6873 2335 ext4_group_t *first_not_zeroed)
ac27a0ec 2336{
617ba13b
MC		 2337 struct ext4_sb_info *sbi = EXT4_SB(sb);
2338 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2339 ext4_fsblk_t last_block;
bd81d8ee
LV		 2340 ext4_fsblk_t block_bitmap;
2341 ext4_fsblk_t inode_bitmap;
2342 ext4_fsblk_t inode_table;
ce421581 2343 int flexbg_flag = 0;
bfff6873 2344 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2345
e2b911c5 2346 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2347 flexbg_flag = 1;
2348
af5bc92d 2349 ext4_debug("Checking group descriptors");
ac27a0ec 2350
197cd65a
AM		 2351 for (i = 0; i < sbi->s_groups_count; i++) {
2352 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2353
ce421581 2354 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2355 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2356 else
2357 last_block = first_block +
617ba13b 2358 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2359
bfff6873
LC		 2360 if ((grp == sbi->s_groups_count) &&
2361 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2362 grp = i;
2363
8fadc143 2364 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2365 if (block_bitmap == sb_block) {
2366 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2367 "Block bitmap for group %u overlaps "
2368 "superblock", i);
18db4b4e
TT		 2369 if (!sb_rdonly(sb))
2370 return 0;
829fa70d 2371 }
2b2d6d01 2372 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2373 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2374 "Block bitmap for group %u not in group "
b31e1552 2375 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2376 return 0;
2377 }
8fadc143 2378 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2379 if (inode_bitmap == sb_block) {
2380 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2381 "Inode bitmap for group %u overlaps "
2382 "superblock", i);
18db4b4e
TT		 2383 if (!sb_rdonly(sb))
2384 return 0;
829fa70d 2385 }
2b2d6d01 2386 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2387 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2388 "Inode bitmap for group %u not in group "
b31e1552 2389 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2390 return 0;
2391 }
8fadc143 2392 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2393 if (inode_table == sb_block) {
2394 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2395 "Inode table for group %u overlaps "
2396 "superblock", i);
18db4b4e
TT		 2397 if (!sb_rdonly(sb))
2398 return 0;
829fa70d 2399 }
bd81d8ee 2400 if (inode_table < first_block ||
2b2d6d01 2401 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2402 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2403 "Inode table for group %u not in group "
b31e1552 2404 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2405 return 0;
2406 }
955ce5f5 2407 ext4_lock_group(sb, i);
feb0ab32 2408 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2409 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2410 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2411 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2412 gdp)), le16_to_cpu(gdp->bg_checksum));
bc98a42c 2413 if (!sb_rdonly(sb)) {
955ce5f5 2414 ext4_unlock_group(sb, i);
8a266467 2415 return 0;
7ee1ec4c 2416 }
717d50e4 2417 }
955ce5f5 2418 ext4_unlock_group(sb, i);
ce421581
JS		 2419 if (!flexbg_flag)
2420 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2421 }
bfff6873
LC		 2422 if (NULL != first_not_zeroed)
2423 *first_not_zeroed = grp;
ac27a0ec
DK		 2424 return 1;
2425}
2426
617ba13b 2427/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2428 * the superblock) which were deleted from all directories, but held open by
2429 * a process at the time of a crash. We walk the list and try to delete these
2430 * inodes at recovery time (only with a read-write filesystem).
2431 *
2432 * In order to keep the orphan inode chain consistent during traversal (in
2433 * case of crash during recovery), we link each inode into the superblock
2434 * orphan list_head and handle it the same way as an inode deletion during
2435 * normal operation (which journals the operations for us).
2436 *
2437 * We only do an iget() and an iput() on each inode, which is very safe if we
2438 * accidentally point at an in-use or already deleted inode. The worst that
2439 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2440 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2441 * e2fsck was run on this filesystem, and it must have already done the orphan
2442 * inode cleanup for us, so we can safely abort without any further action.
2443 */
2b2d6d01
TT		 2444static void ext4_orphan_cleanup(struct super_block *sb,
2445 struct ext4_super_block *es)
ac27a0ec
DK		 2446{
2447 unsigned int s_flags = sb->s_flags;
2c98eb5e 2448 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec 2449#ifdef CONFIG_QUOTA
95f1fda4 2450 int quota_update = 0;
ac27a0ec
DK		 2451 int i;
2452#endif
2453 if (!es->s_last_orphan) {
2454 jbd_debug(4, "no orphan inodes to clean up\n");
2455 return;
2456 }
2457
a8f48a95 2458 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2459 ext4_msg(sb, KERN_ERR, "write access "
2460 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2461 return;
2462 }
2463
d39195c3
AG		 2464 /* Check if feature set would not allow a r/w mount */
2465 if (!ext4_feature_set_ok(sb, 0)) {
2466 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2467 "unknown ROCOMPAT features");
2468 return;
2469 }
2470
617ba13b 2471 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6 2472 /* don't clear list on RO mount w/ errors */
1751e8a6 2473 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
84474976 2474 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2475 "clearing orphan list.\n");
c25f9bc6
ES		 2476 es->s_last_orphan = 0;
2477 }
ac27a0ec
DK		 2478 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2479 return;
2480 }
2481
1751e8a6 2482 if (s_flags & SB_RDONLY) {
b31e1552 2483 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1751e8a6 2484 sb->s_flags &= ~SB_RDONLY;
ac27a0ec
DK		 2485 }
2486#ifdef CONFIG_QUOTA
2487 /* Needed for iput() to work correctly and not trash data */
1751e8a6 2488 sb->s_flags |= SB_ACTIVE;
95f1fda4 2489
2490 /*
2491 * Turn on quotas which were not enabled for read-only mounts if
2492 * filesystem has quota feature, so that they are updated correctly.
2493 */
1751e8a6 2494 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
95f1fda4 2495 int ret = ext4_enable_quotas(sb);
2496
2497 if (!ret)
2498 quota_update = 1;
2499 else
2500 ext4_msg(sb, KERN_ERR,
2501 "Cannot turn on quotas: error %d", ret);
2502 }
2503
2504 /* Turn on journaled quotas used for old sytle */
a2d4a646 2505 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2506 if (EXT4_SB(sb)->s_qf_names[i]) {
2507 int ret = ext4_quota_on_mount(sb, i);
95f1fda4 2508
2509 if (!ret)
2510 quota_update = 1;
2511 else
b31e1552
ES		 2512 ext4_msg(sb, KERN_ERR,
2513 "Cannot turn on journaled "
95f1fda4 2514 "quota: type %d: error %d", i, ret);
ac27a0ec
DK		 2515 }
2516 }
2517#endif
2518
2519 while (es->s_last_orphan) {
2520 struct inode *inode;
2521
c65d5c6c
VN		 2522 /*
2523 * We may have encountered an error during cleanup; if
2524 * so, skip the rest.
2525 */
2526 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2527 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2528 es->s_last_orphan = 0;
2529 break;
2530 }
2531
97bd42b9
JB		 2532 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2533 if (IS_ERR(inode)) {
ac27a0ec
DK		 2534 es->s_last_orphan = 0;
2535 break;
2536 }
2537
617ba13b 2538 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2539 dquot_initialize(inode);
ac27a0ec 2540 if (inode->i_nlink) {
566370a2
PT		 2541 if (test_opt(sb, DEBUG))
2542 ext4_msg(sb, KERN_DEBUG,
2543 "%s: truncating inode %lu to %lld bytes",
2544 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2545 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2546 inode->i_ino, inode->i_size);
5955102c 2547 inode_lock(inode);
55f252c9 2548 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 2549 ret = ext4_truncate(inode);
2550 if (ret)
2551 ext4_std_error(inode->i_sb, ret);
5955102c 2552 inode_unlock(inode);
ac27a0ec
DK		 2553 nr_truncates++;
2554 } else {
566370a2
PT		 2555 if (test_opt(sb, DEBUG))
2556 ext4_msg(sb, KERN_DEBUG,
2557 "%s: deleting unreferenced inode %lu",
2558 __func__, inode->i_ino);
ac27a0ec
DK		 2559 jbd_debug(2, "deleting unreferenced inode %lu\n",
2560 inode->i_ino);
2561 nr_orphans++;
2562 }
2563 iput(inode); /* The delete magic happens here! */
2564 }
2565
2b2d6d01 2566#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2567
2568 if (nr_orphans)
b31e1552
ES		 2569 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2570 PLURAL(nr_orphans));
ac27a0ec 2571 if (nr_truncates)
b31e1552
ES		 2572 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2573 PLURAL(nr_truncates));
ac27a0ec 2574#ifdef CONFIG_QUOTA
95f1fda4 2575 /* Turn off quotas if they were enabled for orphan cleanup */
2576 if (quota_update) {
2577 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2578 if (sb_dqopt(sb)->files[i])
2579 dquot_quota_off(sb, i);
2580 }
ac27a0ec
DK		 2581 }
2582#endif
1751e8a6 2583 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
ac27a0ec 2584}
0b8e58a1 2585
cd2291a4
ES		 2586/*
2587 * Maximal extent format file size.
2588 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2589 * extent format containers, within a sector_t, and within i_blocks
2590 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2591 * so that won't be a limiting factor.
2592 *
f17722f9
LC		 2593 * However there is other limiting factor. We do store extents in the form
2594 * of starting block and length, hence the resulting length of the extent
2595 * covering maximum file size must fit into on-disk format containers as
2596 * well. Given that length is always by 1 unit bigger than max unit (because
2597 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2598 *
cd2291a4
ES		 2599 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2600 */
f287a1a5 2601static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2602{
2603 loff_t res;
2604 loff_t upper_limit = MAX_LFS_FILESIZE;
2605
2606 /* small i_blocks in vfs inode? */
f287a1a5 2607 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2608 /*
90c699a9 2609 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2610 * i_block represent total blocks in 512 bytes
2611 * 32 == size of vfs inode i_blocks * 8
2612 */
2613 upper_limit = (1LL << 32) - 1;
2614
2615 /* total blocks in file system block size */
2616 upper_limit >>= (blkbits - 9);
2617 upper_limit <<= blkbits;
2618 }
2619
f17722f9
LC		 2620 /*
2621 * 32-bit extent-start container, ee_block. We lower the maxbytes
2622 * by one fs block, so ee_len can cover the extent of maximum file
2623 * size
2624 */
2625 res = (1LL << 32) - 1;
cd2291a4 2626 res <<= blkbits;
cd2291a4
ES		 2627
2628 /* Sanity check against vm- & vfs- imposed limits */
2629 if (res > upper_limit)
2630 res = upper_limit;
2631
2632 return res;
2633}
ac27a0ec 2634
ac27a0ec 2635/*
cd2291a4 2636 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2637 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2638 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2639 */
f287a1a5 2640static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2641{
617ba13b 2642 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2643 int meta_blocks;
2644 loff_t upper_limit;
0b8e58a1
AD		 2645 /* This is calculated to be the largest file size for a dense, block
2646 * mapped file such that the file's total number of 512-byte sectors,
2647 * including data and all indirect blocks, does not exceed (2^48 - 1).
2648 *
2649 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2650 * number of 512-byte sectors of the file.
0fc1b451
AK		 2651 */
2652
f287a1a5 2653 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2654 /*
90c699a9 2655 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2656 * the inode i_block field represents total file blocks in
2657 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2658 */
2659 upper_limit = (1LL << 32) - 1;
2660
2661 /* total blocks in file system block size */
2662 upper_limit >>= (bits - 9);
2663
2664 } else {
8180a562
AK		 2665 /*
2666 * We use 48 bit ext4_inode i_blocks
2667 * With EXT4_HUGE_FILE_FL set the i_blocks
2668 * represent total number of blocks in
2669 * file system block size
2670 */
0fc1b451
AK		 2671 upper_limit = (1LL << 48) - 1;
2672
0fc1b451
AK		 2673 }
2674
2675 /* indirect blocks */
2676 meta_blocks = 1;
2677 /* double indirect blocks */
2678 meta_blocks += 1 + (1LL << (bits-2));
2679 /* tripple indirect blocks */
2680 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2681
2682 upper_limit -= meta_blocks;
2683 upper_limit <<= bits;
ac27a0ec
DK		 2684
2685 res += 1LL << (bits-2);
2686 res += 1LL << (2*(bits-2));
2687 res += 1LL << (3*(bits-2));
2688 res <<= bits;
2689 if (res > upper_limit)
2690 res = upper_limit;
0fc1b451
AK		 2691
2692 if (res > MAX_LFS_FILESIZE)
2693 res = MAX_LFS_FILESIZE;
2694
ac27a0ec
DK		 2695 return res;
2696}
2697
617ba13b 2698static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2699 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2700{
617ba13b 2701 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2702 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2703 int has_super = 0;
2704
2705 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2706
e2b911c5 2707 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2708 return logical_sb_block + nr + 1;
ac27a0ec 2709 bg = sbi->s_desc_per_block * nr;
617ba13b 2710 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2711 has_super = 1;
0b8e58a1 2712
bd63f6b0
DW		 2713 /*
2714 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2715 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2716 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2717 * compensate.
2718 */
2719 if (sb->s_blocksize == 1024 && nr == 0 &&
49598e04 2720 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
bd63f6b0
DW		 2721 has_super++;
2722
617ba13b 2723 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2724}
2725
c9de560d
AT		 2726/**
2727 * ext4_get_stripe_size: Get the stripe size.
2728 * @sbi: In memory super block info
2729 *
2730 * If we have specified it via mount option, then
2731 * use the mount option value. If the value specified at mount time is
2732 * greater than the blocks per group use the super block value.
2733 * If the super block value is greater than blocks per group return 0.
2734 * Allocator needs it be less than blocks per group.
2735 *
2736 */
2737static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2738{
2739 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2740 unsigned long stripe_width =
2741 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2742 int ret;
c9de560d
AT		 2743
2744 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 2745 ret = sbi->s_stripe;
5469d7c3 2746 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 2747 ret = stripe_width;
5469d7c3 2748 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 2749 ret = stride;
2750 else
2751 ret = 0;
c9de560d 2752
3eb08658
DE		 2753 /*
2754 * If the stripe width is 1, this makes no sense and
2755 * we set it to 0 to turn off stripe handling code.
2756 */
2757 if (ret <= 1)
2758 ret = 0;
c9de560d 2759
3eb08658 2760 return ret;
c9de560d 2761}
ac27a0ec 2762
a13fb1a4
ES		 2763/*
2764 * Check whether this filesystem can be mounted based on
2765 * the features present and the RDONLY/RDWR mount requested.
2766 * Returns 1 if this filesystem can be mounted as requested,
2767 * 0 if it cannot be.
2768 */
2769static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2770{
e2b911c5 2771 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2772 ext4_msg(sb, KERN_ERR,
2773 "Couldn't mount because of "
2774 "unsupported optional features (%x)",
2775 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2776 ~EXT4_FEATURE_INCOMPAT_SUPP));
2777 return 0;
2778 }
2779
2780 if (readonly)
2781 return 1;
2782
e2b911c5 2783 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b 2784 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
1751e8a6 2785 sb->s_flags |= SB_RDONLY;
2cb5cc8b
DW		 2786 return 1;
2787 }
2788
a13fb1a4 2789 /* Check that feature set is OK for a read-write mount */
e2b911c5 2790 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2791 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2792 "unsupported optional features (%x)",
2793 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2794 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2795 return 0;
2796 }
2797 /*
2798 * Large file size enabled file system can only be mounted
2799 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2800 */
e2b911c5 2801 if (ext4_has_feature_huge_file(sb)) {
a13fb1a4
ES		 2802 if (sizeof(blkcnt_t) < sizeof(u64)) {
2803 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2804 "cannot be mounted RDWR without "
2805 "CONFIG_LBDAF");
2806 return 0;
2807 }
2808 }
e2b911c5 2809 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2810 ext4_msg(sb, KERN_ERR,
2811 "Can't support bigalloc feature without "
2812 "extents feature\n");
2813 return 0;
2814 }
7c319d32
AK		 2815
2816#ifndef CONFIG_QUOTA
e2b911c5 2817 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2818 ext4_msg(sb, KERN_ERR,
2819 "Filesystem with quota feature cannot be mounted RDWR "
2820 "without CONFIG_QUOTA");
2821 return 0;
2822 }
689c958c
LX		 2823 if (ext4_has_feature_project(sb) && !readonly) {
2824 ext4_msg(sb, KERN_ERR,
2825 "Filesystem with project quota feature cannot be mounted RDWR "
2826 "without CONFIG_QUOTA");
2827 return 0;
2828 }
7c319d32 2829#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2830 return 1;
2831}
2832
66e61a9e
TT		 2833/*
2834 * This function is called once a day if we have errors logged
2835 * on the file system
2836 */
235699a8 2837static void print_daily_error_info(struct timer_list *t)
66e61a9e 2838{
235699a8
KC		 2839 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
2840 struct super_block *sb = sbi->s_sb;
2841 struct ext4_super_block *es = sbi->s_es;
66e61a9e
TT		 2842
2843 if (es->s_error_count)
ae0f78de
TT		 2844 /* fsck newer than v1.41.13 is needed to clean this condition. */
2845 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2846 le32_to_cpu(es->s_error_count));
2847 if (es->s_first_error_time) {
ae0f78de 2848 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2849 sb->s_id, le32_to_cpu(es->s_first_error_time),
2850 (int) sizeof(es->s_first_error_func),
2851 es->s_first_error_func,
2852 le32_to_cpu(es->s_first_error_line));
2853 if (es->s_first_error_ino)
651e1c3b 2854 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2855 le32_to_cpu(es->s_first_error_ino));
2856 if (es->s_first_error_block)
651e1c3b 2857 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2858 le64_to_cpu(es->s_first_error_block));
651e1c3b 2859 printk(KERN_CONT "\n");
66e61a9e
TT		 2860 }
2861 if (es->s_last_error_time) {
ae0f78de 2862 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2863 sb->s_id, le32_to_cpu(es->s_last_error_time),
2864 (int) sizeof(es->s_last_error_func),
2865 es->s_last_error_func,
2866 le32_to_cpu(es->s_last_error_line));
2867 if (es->s_last_error_ino)
651e1c3b 2868 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2869 le32_to_cpu(es->s_last_error_ino));
2870 if (es->s_last_error_block)
651e1c3b 2871 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2872 le64_to_cpu(es->s_last_error_block));
651e1c3b 2873 printk(KERN_CONT "\n");
66e61a9e
TT		 2874 }
2875 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2876}
2877
bfff6873
LC		 2878/* Find next suitable group and run ext4_init_inode_table */
2879static int ext4_run_li_request(struct ext4_li_request *elr)
2880{
2881 struct ext4_group_desc *gdp = NULL;
2882 ext4_group_t group, ngroups;
2883 struct super_block *sb;
2884 unsigned long timeout = 0;
2885 int ret = 0;
2886
2887 sb = elr->lr_super;
2888 ngroups = EXT4_SB(sb)->s_groups_count;
2889
2890 for (group = elr->lr_next_group; group < ngroups; group++) {
2891 gdp = ext4_get_group_desc(sb, group, NULL);
2892 if (!gdp) {
2893 ret = 1;
2894 break;
2895 }
2896
2897 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2898 break;
2899 }
2900
7f511862 2901 if (group >= ngroups)
bfff6873
LC		 2902 ret = 1;
2903
2904 if (!ret) {
2905 timeout = jiffies;
2906 ret = ext4_init_inode_table(sb, group,
2907 elr->lr_timeout ? 0 : 1);
2908 if (elr->lr_timeout == 0) {
51ce6511
LC		 2909 timeout = (jiffies - timeout) *
2910 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2911 elr->lr_timeout = timeout;
2912 }
2913 elr->lr_next_sched = jiffies + elr->lr_timeout;
2914 elr->lr_next_group = group + 1;
2915 }
bfff6873
LC		 2916 return ret;
2917}
2918
2919/*
2920 * Remove lr_request from the list_request and free the
4ed5c033 2921 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2922 */
2923static void ext4_remove_li_request(struct ext4_li_request *elr)
2924{
2925 struct ext4_sb_info *sbi;
2926
2927 if (!elr)
2928 return;
2929
2930 sbi = elr->lr_sbi;
2931
2932 list_del(&elr->lr_request);
2933 sbi->s_li_request = NULL;
2934 kfree(elr);
2935}
2936
2937static void ext4_unregister_li_request(struct super_block *sb)
2938{
1bb933fb
LC		 2939 mutex_lock(&ext4_li_mtx);
2940 if (!ext4_li_info) {
2941 mutex_unlock(&ext4_li_mtx);
bfff6873 2942 return;
1bb933fb 2943 }
bfff6873
LC		 2944
2945 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2946 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2947 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2948 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2949}
2950
8f1f7453
ES		 2951static struct task_struct *ext4_lazyinit_task;
2952
bfff6873
LC		 2953/*
2954 * This is the function where ext4lazyinit thread lives. It walks
2955 * through the request list searching for next scheduled filesystem.
2956 * When such a fs is found, run the lazy initialization request
2957 * (ext4_rn_li_request) and keep track of the time spend in this
2958 * function. Based on that time we compute next schedule time of
2959 * the request. When walking through the list is complete, compute
2960 * next waking time and put itself into sleep.
2961 */
2962static int ext4_lazyinit_thread(void *arg)
2963{
2964 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2965 struct list_head *pos, *n;
2966 struct ext4_li_request *elr;
4ed5c033 2967 unsigned long next_wakeup, cur;
bfff6873
LC		 2968
2969 BUG_ON(NULL == eli);
2970
bfff6873
LC		 2971cont_thread:
2972 while (true) {
2973 next_wakeup = MAX_JIFFY_OFFSET;
2974
2975 mutex_lock(&eli->li_list_mtx);
2976 if (list_empty(&eli->li_request_list)) {
2977 mutex_unlock(&eli->li_list_mtx);
2978 goto exit_thread;
2979 }
bfff6873 2980 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 2981 int err = 0;
2982 int progress = 0;
bfff6873
LC		 2983 elr = list_entry(pos, struct ext4_li_request,
2984 lr_request);
2985
e22834f0
DM		 2986 if (time_before(jiffies, elr->lr_next_sched)) {
2987 if (time_before(elr->lr_next_sched, next_wakeup))
2988 next_wakeup = elr->lr_next_sched;
2989 continue;
2990 }
2991 if (down_read_trylock(&elr->lr_super->s_umount)) {
2992 if (sb_start_write_trylock(elr->lr_super)) {
2993 progress = 1;
2994 /*
2995 * We hold sb->s_umount, sb can not
2996 * be removed from the list, it is
2997 * now safe to drop li_list_mtx
2998 */
2999 mutex_unlock(&eli->li_list_mtx);
3000 err = ext4_run_li_request(elr);
3001 sb_end_write(elr->lr_super);
3002 mutex_lock(&eli->li_list_mtx);
3003 n = pos->next;
b2c78cd0 3004 }
e22834f0
DM		 3005 up_read((&elr->lr_super->s_umount));
3006 }
3007 /* error, remove the lazy_init job */
3008 if (err) {
3009 ext4_remove_li_request(elr);
3010 continue;
3011 }
3012 if (!progress) {
3013 elr->lr_next_sched = jiffies +
3014 (prandom_u32()
3015 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 3016 }
bfff6873
LC		 3017 if (time_before(elr->lr_next_sched, next_wakeup))
3018 next_wakeup = elr->lr_next_sched;
3019 }
3020 mutex_unlock(&eli->li_list_mtx);
3021
a0acae0e 3022 try_to_freeze();
bfff6873 3023
4ed5c033
LC		 3024 cur = jiffies;
3025 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3026 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3027 cond_resched();
3028 continue;
3029 }
3030
4ed5c033
LC		 3031 schedule_timeout_interruptible(next_wakeup - cur);
3032
8f1f7453
ES		 3033 if (kthread_should_stop()) {
3034 ext4_clear_request_list();
3035 goto exit_thread;
3036 }
bfff6873
LC		 3037 }
3038
3039exit_thread:
3040 /*
3041 * It looks like the request list is empty, but we need
3042 * to check it under the li_list_mtx lock, to prevent any
3043 * additions into it, and of course we should lock ext4_li_mtx
3044 * to atomically free the list and ext4_li_info, because at
3045 * this point another ext4 filesystem could be registering
3046 * new one.
3047 */
3048 mutex_lock(&ext4_li_mtx);
3049 mutex_lock(&eli->li_list_mtx);
3050 if (!list_empty(&eli->li_request_list)) {
3051 mutex_unlock(&eli->li_list_mtx);
3052 mutex_unlock(&ext4_li_mtx);
3053 goto cont_thread;
3054 }
3055 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3056 kfree(ext4_li_info);
3057 ext4_li_info = NULL;
3058 mutex_unlock(&ext4_li_mtx);
3059
3060 return 0;
3061}
3062
3063static void ext4_clear_request_list(void)
3064{
3065 struct list_head *pos, *n;
3066 struct ext4_li_request *elr;
3067
3068 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3069 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3070 elr = list_entry(pos, struct ext4_li_request,
3071 lr_request);
3072 ext4_remove_li_request(elr);
3073 }
3074 mutex_unlock(&ext4_li_info->li_list_mtx);
3075}
3076
3077static int ext4_run_lazyinit_thread(void)
3078{
8f1f7453
ES		 3079 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3080 ext4_li_info, "ext4lazyinit");
3081 if (IS_ERR(ext4_lazyinit_task)) {
3082 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3083 ext4_clear_request_list();
bfff6873
LC		 3084 kfree(ext4_li_info);
3085 ext4_li_info = NULL;
92b97816 3086 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3087 "initialization thread\n",
3088 err);
3089 return err;
3090 }
3091 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3092 return 0;
3093}
3094
3095/*
3096 * Check whether it make sense to run itable init. thread or not.
3097 * If there is at least one uninitialized inode table, return
3098 * corresponding group number, else the loop goes through all
3099 * groups and return total number of groups.
3100 */
3101static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3102{
3103 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3104 struct ext4_group_desc *gdp = NULL;
3105
3106 for (group = 0; group < ngroups; group++) {
3107 gdp = ext4_get_group_desc(sb, group, NULL);
3108 if (!gdp)
3109 continue;
3110
3111 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3112 break;
3113 }
3114
3115 return group;
3116}
3117
3118static int ext4_li_info_new(void)
3119{
3120 struct ext4_lazy_init *eli = NULL;
3121
3122 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3123 if (!eli)
3124 return -ENOMEM;
3125
bfff6873
LC		 3126 INIT_LIST_HEAD(&eli->li_request_list);
3127 mutex_init(&eli->li_list_mtx);
3128
bfff6873
LC		 3129 eli->li_state |= EXT4_LAZYINIT_QUIT;
3130
3131 ext4_li_info = eli;
3132
3133 return 0;
3134}
3135
3136static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3137 ext4_group_t start)
3138{
3139 struct ext4_sb_info *sbi = EXT4_SB(sb);
3140 struct ext4_li_request *elr;
bfff6873
LC		 3141
3142 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3143 if (!elr)
3144 return NULL;
3145
3146 elr->lr_super = sb;
3147 elr->lr_sbi = sbi;
3148 elr->lr_next_group = start;
3149
3150 /*
3151 * Randomize first schedule time of the request to
3152 * spread the inode table initialization requests
3153 * better.
3154 */
dd1f723b
TT		 3155 elr->lr_next_sched = jiffies + (prandom_u32() %
3156 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3157 return elr;
3158}
3159
7f511862
TT		 3160int ext4_register_li_request(struct super_block *sb,
3161 ext4_group_t first_not_zeroed)
bfff6873
LC		 3162{
3163 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3164 struct ext4_li_request *elr = NULL;
49598e04 3165 ext4_group_t ngroups = sbi->s_groups_count;
6c5a6cb9 3166 int ret = 0;
bfff6873 3167
7f511862 3168 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3169 if (sbi->s_li_request != NULL) {
3170 /*
3171 * Reset timeout so it can be computed again, because
3172 * s_li_wait_mult might have changed.
3173 */
3174 sbi->s_li_request->lr_timeout = 0;
7f511862 3175 goto out;
51ce6511 3176 }
bfff6873 3177
bc98a42c 3178 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
55ff3840 3179 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3180 goto out;
bfff6873
LC		 3181
3182 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3183 if (!elr) {
3184 ret = -ENOMEM;
3185 goto out;
3186 }
bfff6873
LC		 3187
3188 if (NULL == ext4_li_info) {
3189 ret = ext4_li_info_new();
3190 if (ret)
3191 goto out;
3192 }
3193
3194 mutex_lock(&ext4_li_info->li_list_mtx);
3195 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3196 mutex_unlock(&ext4_li_info->li_list_mtx);
3197
3198 sbi->s_li_request = elr;
46e4690b
TM		 3199 /*
3200 * set elr to NULL here since it has been inserted to
3201 * the request_list and the removal and free of it is
3202 * handled by ext4_clear_request_list from now on.
3203 */
3204 elr = NULL;
bfff6873
LC		 3205
3206 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3207 ret = ext4_run_lazyinit_thread();
3208 if (ret)
3209 goto out;
3210 }
bfff6873 3211out:
beed5ecb
NK		 3212 mutex_unlock(&ext4_li_mtx);
3213 if (ret)
bfff6873 3214 kfree(elr);
bfff6873
LC		 3215 return ret;
3216}
3217
3218/*
3219 * We do not need to lock anything since this is called on
3220 * module unload.
3221 */
3222static void ext4_destroy_lazyinit_thread(void)
3223{
3224 /*
3225 * If thread exited earlier
3226 * there's nothing to be done.
3227 */
8f1f7453 3228 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3229 return;
3230
8f1f7453 3231 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3232}
3233
25ed6e8a
DW		 3234static int set_journal_csum_feature_set(struct super_block *sb)
3235{
3236 int ret = 1;
3237 int compat, incompat;
3238 struct ext4_sb_info *sbi = EXT4_SB(sb);
3239
9aa5d32b 3240 if (ext4_has_metadata_csum(sb)) {
db9ee220 3241 /* journal checksum v3 */
25ed6e8a 3242 compat = 0;
db9ee220 3243 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3244 } else {
3245 /* journal checksum v1 */
3246 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3247 incompat = 0;
3248 }
3249
feb8c6d3
DW		 3250 jbd2_journal_clear_features(sbi->s_journal,
3251 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3252 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3253 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3254 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3255 ret = jbd2_journal_set_features(sbi->s_journal,
3256 compat, 0,
3257 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3258 incompat);
3259 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3260 ret = jbd2_journal_set_features(sbi->s_journal,
3261 compat, 0,
3262 incompat);
3263 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3264 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3265 } else {
feb8c6d3
DW		 3266 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3267 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3268 }
3269
3270 return ret;
3271}
3272
952fc18e
TT		 3273/*
3274 * Note: calculating the overhead so we can be compatible with
3275 * historical BSD practice is quite difficult in the face of
3276 * clusters/bigalloc. This is because multiple metadata blocks from
3277 * different block group can end up in the same allocation cluster.
3278 * Calculating the exact overhead in the face of clustered allocation
3279 * requires either O(all block bitmaps) in memory or O(number of block
3280 * groups**2) in time. We will still calculate the superblock for
3281 * older file systems --- and if we come across with a bigalloc file
3282 * system with zero in s_overhead_clusters the estimate will be close to
3283 * correct especially for very large cluster sizes --- but for newer
3284 * file systems, it's better to calculate this figure once at mkfs
3285 * time, and store it in the superblock. If the superblock value is
3286 * present (even for non-bigalloc file systems), we will use it.
3287 */
3288static int count_overhead(struct super_block *sb, ext4_group_t grp,
3289 char *buf)
3290{
3291 struct ext4_sb_info *sbi = EXT4_SB(sb);
3292 struct ext4_group_desc *gdp;
3293 ext4_fsblk_t first_block, last_block, b;
3294 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3295 int s, j, count = 0;
3296
e2b911c5 3297 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3298 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3299 sbi->s_itb_per_group + 2);
3300
952fc18e
TT		 3301 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3302 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3303 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3304 for (i = 0; i < ngroups; i++) {
3305 gdp = ext4_get_group_desc(sb, i, NULL);
3306 b = ext4_block_bitmap(sb, gdp);
3307 if (b >= first_block && b <= last_block) {
3308 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3309 count++;
3310 }
3311 b = ext4_inode_bitmap(sb, gdp);
3312 if (b >= first_block && b <= last_block) {
3313 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3314 count++;
3315 }
3316 b = ext4_inode_table(sb, gdp);
3317 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3318 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3319 int c = EXT4_B2C(sbi, b - first_block);
3320 ext4_set_bit(c, buf);
3321 count++;
3322 }
3323 if (i != grp)
3324 continue;
3325 s = 0;
3326 if (ext4_bg_has_super(sb, grp)) {
3327 ext4_set_bit(s++, buf);
3328 count++;
3329 }
c48ae41b
TT		 3330 j = ext4_bg_num_gdb(sb, grp);
3331 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3332 ext4_error(sb, "Invalid number of block group "
3333 "descriptor blocks: %d", j);
3334 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3335 }
c48ae41b
TT		 3336 count += j;
3337 for (; j > 0; j--)
3338 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3339 }
3340 if (!count)
3341 return 0;
3342 return EXT4_CLUSTERS_PER_GROUP(sb) -
3343 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3344}
3345
3346/*
3347 * Compute the overhead and stash it in sbi->s_overhead
3348 */
3349int ext4_calculate_overhead(struct super_block *sb)
3350{
3351 struct ext4_sb_info *sbi = EXT4_SB(sb);
3352 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3353 struct inode *j_inode;
3354 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3355 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3356 ext4_fsblk_t overhead = 0;
4fdb5543 3357 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3358
952fc18e
TT		 3359 if (!buf)
3360 return -ENOMEM;
3361
3362 /*
3363 * Compute the overhead (FS structures). This is constant
3364 * for a given filesystem unless the number of block groups
3365 * changes so we cache the previous value until it does.
3366 */
3367
3368 /*
3369 * All of the blocks before first_data_block are overhead
3370 */
3371 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3372
3373 /*
3374 * Add the overhead found in each block group
3375 */
3376 for (i = 0; i < ngroups; i++) {
3377 int blks;
3378
3379 blks = count_overhead(sb, i, buf);
3380 overhead += blks;
3381 if (blks)
3382 memset(buf, 0, PAGE_SIZE);
3383 cond_resched();
3384 }
3c816ded
EW		 3385
3386 /*
3387 * Add the internal journal blocks whether the journal has been
3388 * loaded or not
3389 */
b003b524 3390 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3391 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3c816ded
EW		 3392 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3393 j_inode = ext4_get_journal_inode(sb, j_inum);
3394 if (j_inode) {
3395 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3396 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3397 iput(j_inode);
3398 } else {
3399 ext4_msg(sb, KERN_ERR, "can't get journal size");
3400 }
3401 }
952fc18e
TT		 3402 sbi->s_overhead = overhead;
3403 smp_wmb();
3404 free_page((unsigned long) buf);
3405 return 0;
3406}
3407
b5799018 3408static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3409{
3410 ext4_fsblk_t resv_clusters;
b5799018 3411 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3412
30fac0f7
JK		 3413 /*
3414 * There's no need to reserve anything when we aren't using extents.
3415 * The space estimates are exact, there are no unwritten extents,
3416 * hole punching doesn't need new metadata... This is needed especially
3417 * to keep ext2/3 backward compatibility.
3418 */
e2b911c5 3419 if (!ext4_has_feature_extents(sb))
b5799018 3420 return;
27dd4385
LC		 3421 /*
3422 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3423 * This should cover the situations where we can not afford to run
3424 * out of space like for example punch hole, or converting
556615dc 3425 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3426 * allocation would require 1, or 2 blocks, higher numbers are
3427 * very rare.
3428 */
b5799018
TT		 3429 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3430 sbi->s_cluster_bits);
27dd4385
LC		 3431
3432 do_div(resv_clusters, 50);
3433 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3434
b5799018 3435 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3436}
3437
2b2d6d01 3438static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3439{
5e405595 3440 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
d4c402d9 3441 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3442 struct buffer_head *bh;
617ba13b 3443 struct ext4_super_block *es = NULL;
5aee0f8a 3444 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
617ba13b
MC		 3445 ext4_fsblk_t block;
3446 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3447 ext4_fsblk_t logical_sb_block;
ac27a0ec 3448 unsigned long offset = 0;
ac27a0ec
DK		 3449 unsigned long journal_devnum = 0;
3450 unsigned long def_mount_opts;
3451 struct inode *root;
0390131b 3452 const char *descr;
dcc7dae3 3453 int ret = -ENOMEM;
281b5995 3454 int blocksize, clustersize;
4ec11028
TT		 3455 unsigned int db_count;
3456 unsigned int i;
281b5995 3457 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3458 __u64 blocks_count;
07aa2ea1 3459 int err = 0;
b3881f74 3460 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3461 ext4_group_t first_not_zeroed;
ac27a0ec 3462
5aee0f8a
TT		 3463 if ((data && !orig_data) || !sbi)
3464 goto out_free_base;
705895b6 3465
aed9eb1b 3466 sbi->s_daxdev = dax_dev;
705895b6
PE		 3467 sbi->s_blockgroup_lock =
3468 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 3469 if (!sbi->s_blockgroup_lock)
3470 goto out_free_base;
3471
ac27a0ec 3472 sb->s_fs_info = sbi;
2c0544b2 3473 sbi->s_sb = sb;
240799cd 3474 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3475 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3476 if (sb->s_bdev->bd_part)
3477 sbi->s_sectors_written_start =
3478 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3479
9f6200bb 3480 /* Cleanup superblock name */
ec3904dc 3481 strreplace(sb->s_id, '/', '!');
9f6200bb 3482
07aa2ea1 3483 /* -EINVAL is default */
dcc7dae3 3484 ret = -EINVAL;
617ba13b 3485 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3486 if (!blocksize) {
b31e1552 3487 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3488 goto out_fail;
3489 }
3490
3491 /*
617ba13b 3492 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3493 * block sizes. We need to calculate the offset from buffer start.
3494 */
617ba13b 3495 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3496 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3497 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3498 } else {
70bbb3e0 3499 logical_sb_block = sb_block;
ac27a0ec
DK		 3500 }
3501
a8ac900b 3502 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3503 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3504 goto out_fail;
3505 }
3506 /*
3507 * Note: s_es must be initialized as soon as possible because
617ba13b 3508 * some ext4 macro-instructions depend on its value
ac27a0ec 3509 */
2716b802 3510 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3511 sbi->s_es = es;
3512 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3513 if (sb->s_magic != EXT4_SUPER_MAGIC)
3514 goto cantfind_ext4;
afc32f7e 3515 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3516
feb0ab32 3517 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3518 if (ext4_has_feature_metadata_csum(sb) &&
3519 ext4_has_feature_gdt_csum(sb))
363307e6 3520 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3521 "redundant flags; please run fsck.");
3522
d25425f8
DW		 3523 /* Check for a known checksum algorithm */
3524 if (!ext4_verify_csum_type(sb, es)) {
3525 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3526 "unknown checksum algorithm.");
3527 silent = 1;
3528 goto cantfind_ext4;
3529 }
3530
0441984a 3531 /* Load the checksum driver */
a45403b5
TT		 3532 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3533 if (IS_ERR(sbi->s_chksum_driver)) {
3534 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3535 ret = PTR_ERR(sbi->s_chksum_driver);
3536 sbi->s_chksum_driver = NULL;
3537 goto failed_mount;
0441984a
DW		 3538 }
3539
a9c47317
DW		 3540 /* Check superblock checksum */
3541 if (!ext4_superblock_csum_verify(sb, es)) {
3542 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3543 "invalid superblock checksum. Run e2fsck?");
3544 silent = 1;
6a797d27 3545 ret = -EFSBADCRC;
a9c47317
DW		 3546 goto cantfind_ext4;
3547 }
3548
3549 /* Precompute checksum seed for all metadata */
e2b911c5 3550 if (ext4_has_feature_csum_seed(sb))
8c81bd8f 3551 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
dec214d0 3552 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
a9c47317
DW		 3553 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3554 sizeof(es->s_uuid));
3555
ac27a0ec
DK		 3556 /* Set defaults before we parse the mount options */
3557 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3558 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3559 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3560 set_opt(sb, DEBUG);
87f26807 3561 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3562 set_opt(sb, GRPID);
617ba13b 3563 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3564 set_opt(sb, NO_UID32);
ea663336 3565 /* xattr user namespace & acls are now defaulted on */
ea663336 3566 set_opt(sb, XATTR_USER);
03010a33 3567#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3568 set_opt(sb, POSIX_ACL);
2e7842b8 3569#endif
98c1a759
DW		 3570 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3571 if (ext4_has_metadata_csum(sb))
3572 set_opt(sb, JOURNAL_CHECKSUM);
3573
617ba13b 3574 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3575 set_opt(sb, JOURNAL_DATA);
617ba13b 3576 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3577 set_opt(sb, ORDERED_DATA);
617ba13b 3578 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3579 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3580
3581 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3582 set_opt(sb, ERRORS_PANIC);
bb4f397a 3583 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3584 set_opt(sb, ERRORS_CONT);
bb4f397a 3585 else
fd8c37ec 3586 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3587 /* block_validity enabled by default; disable with noblock_validity */
3588 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3589 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3590 set_opt(sb, DISCARD);
ac27a0ec 3591
08cefc7a
EB		 3592 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3593 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3594 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3595 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3596 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3597
8b67f04a 3598 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3599 set_opt(sb, BARRIER);
ac27a0ec 3600
dd919b98
AK		 3601 /*
3602 * enable delayed allocation by default
3603 * Use -o nodelalloc to turn it off
3604 */
bc0b75f7 3605 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3606 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3607 set_opt(sb, DELALLOC);
dd919b98 3608
51ce6511
LC		 3609 /*
3610 * set default s_li_wait_mult for lazyinit, for the case there is
3611 * no mount option specified.
3612 */
3613 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3614
5aee0f8a
TT		 3615 if (sbi->s_es->s_mount_opts[0]) {
3616 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3617 sizeof(sbi->s_es->s_mount_opts),
3618 GFP_KERNEL);
3619 if (!s_mount_opts)
3620 goto failed_mount;
3621 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3622 &journal_ioprio, 0)) {
3623 ext4_msg(sb, KERN_WARNING,
3624 "failed to parse options in superblock: %s",
3625 s_mount_opts);
3626 }
3627 kfree(s_mount_opts);
8b67f04a 3628 }
5a916be1 3629 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3630 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3631 &journal_ioprio, 0))
ac27a0ec
DK		 3632 goto failed_mount;
3633
56889787
TT		 3634 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3635 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3636 "with data=journal disables delayed "
3637 "allocation and O_DIRECT support!\n");
3638 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3639 ext4_msg(sb, KERN_ERR, "can't mount with "
3640 "both data=journal and delalloc");
3641 goto failed_mount;
3642 }
3643 if (test_opt(sb, DIOREAD_NOLOCK)) {
3644 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3645 "both data=journal and dioread_nolock");
56889787
TT		 3646 goto failed_mount;
3647 }
923ae0ff
RZ		 3648 if (test_opt(sb, DAX)) {
3649 ext4_msg(sb, KERN_ERR, "can't mount with "
3650 "both data=journal and dax");
3651 goto failed_mount;
3652 }
73b92a2a
SK		 3653 if (ext4_has_feature_encrypt(sb)) {
3654 ext4_msg(sb, KERN_WARNING,
3655 "encrypted files will use data=ordered "
3656 "instead of data journaling mode");
3657 }
56889787
TT		 3658 if (test_opt(sb, DELALLOC))
3659 clear_opt(sb, DELALLOC);
001e4a87
TH		 3660 } else {
3661 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3662 }
3663
1751e8a6
LT		 3664 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3665 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec 3666
617ba13b 3667 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3668 (ext4_has_compat_features(sb) ||
3669 ext4_has_ro_compat_features(sb) ||
3670 ext4_has_incompat_features(sb)))
b31e1552
ES		 3671 ext4_msg(sb, KERN_WARNING,
3672 "feature flags set on rev 0 fs, "
3673 "running e2fsck is recommended");
469108ff 3674
ed3654eb
TT		 3675 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3676 set_opt2(sb, HURD_COMPAT);
e2b911c5 3677 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3678 ext4_msg(sb, KERN_ERR,
3679 "The Hurd can't support 64-bit file systems");
3680 goto failed_mount;
3681 }
dec214d0
TE		 3682
3683 /*
3684 * ea_inode feature uses l_i_version field which is not
3685 * available in HURD_COMPAT mode.
3686 */
3687 if (ext4_has_feature_ea_inode(sb)) {
3688 ext4_msg(sb, KERN_ERR,
3689 "ea_inode feature is not supported for Hurd");
3690 goto failed_mount;
3691 }
ed3654eb
TT		 3692 }
3693
2035e776
TT		 3694 if (IS_EXT2_SB(sb)) {
3695 if (ext2_feature_set_ok(sb))
3696 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3697 "using the ext4 subsystem");
3698 else {
0d9366d6
ES		 3699 /*
3700 * If we're probing be silent, if this looks like
3701 * it's actually an ext[34] filesystem.
3702 */
3703 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3704 goto failed_mount;
2035e776
TT		 3705 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3706 "to feature incompatibilities");
3707 goto failed_mount;
3708 }
3709 }
3710
3711 if (IS_EXT3_SB(sb)) {
3712 if (ext3_feature_set_ok(sb))
3713 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3714 "using the ext4 subsystem");
3715 else {
0d9366d6
ES		 3716 /*
3717 * If we're probing be silent, if this looks like
3718 * it's actually an ext4 filesystem.
3719 */
3720 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3721 goto failed_mount;
2035e776
TT		 3722 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3723 "to feature incompatibilities");
3724 goto failed_mount;
3725 }
3726 }
3727
ac27a0ec
DK		 3728 /*
3729 * Check feature flags regardless of the revision level, since we
3730 * previously didn't change the revision level when setting the flags,
3731 * so there is a chance incompat flags are set on a rev 0 filesystem.
3732 */
bc98a42c 3733 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
ac27a0ec 3734 goto failed_mount;
a13fb1a4 3735
261cb20c 3736 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3737 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3738 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552 3739 ext4_msg(sb, KERN_ERR,
8cdf3372
TT		 3740 "Unsupported filesystem blocksize %d (%d log_block_size)",
3741 blocksize, le32_to_cpu(es->s_log_block_size));
3742 goto failed_mount;
3743 }
3744 if (le32_to_cpu(es->s_log_block_size) >
3745 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3746 ext4_msg(sb, KERN_ERR,
3747 "Invalid log block size: %u",
3748 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 3749 goto failed_mount;
3750 }
3751
5b9554dc
TT		 3752 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3753 ext4_msg(sb, KERN_ERR,
3754 "Number of reserved GDT blocks insanely large: %d",
3755 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3756 goto failed_mount;
3757 }
3758
923ae0ff 3759 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
559db4c6
RZ		 3760 if (ext4_has_feature_inline_data(sb)) {
3761 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3762 " that may contain inline data");
24f3478d 3763 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
559db4c6 3764 }
87eefeb4 3765 err = bdev_dax_supported(sb, blocksize);
24f3478d
DW		 3766 if (err) {
3767 ext4_msg(sb, KERN_ERR,
3768 "DAX unsupported by block device. Turning off DAX.");
3769 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
3770 }
923ae0ff
RZ		 3771 }
3772
e2b911c5 3773 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 3774 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3775 es->s_encryption_level);
3776 goto failed_mount;
3777 }
3778
ac27a0ec 3779 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3780 /* Validate the filesystem blocksize */
3781 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3782 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3783 blocksize);
ac27a0ec
DK		 3784 goto failed_mount;
3785 }
3786
2b2d6d01 3787 brelse(bh);
70bbb3e0
AM		 3788 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3789 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3790 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3791 if (!bh) {
b31e1552
ES		 3792 ext4_msg(sb, KERN_ERR,
3793 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3794 goto failed_mount;
3795 }
2716b802 3796 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3797 sbi->s_es = es;
617ba13b 3798 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3799 ext4_msg(sb, KERN_ERR,
3800 "Magic mismatch, very weird!");
ac27a0ec
DK		 3801 goto failed_mount;
3802 }
3803 }
3804
e2b911c5 3805 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 3806 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3807 has_huge_files);
3808 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3809
617ba13b
MC		 3810 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3811 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3812 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3813 } else {
3814 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3815 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3816 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3817 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3818 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3819 ext4_msg(sb, KERN_ERR,
3820 "unsupported inode size: %d",
2b2d6d01 3821 sbi->s_inode_size);
ac27a0ec
DK		 3822 goto failed_mount;
3823 }
ef7f3835
KS		 3824 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3825 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3826 }
0b8e58a1 3827
0d1ee42f 3828 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 3829 if (ext4_has_feature_64bit(sb)) {
8fadc143 3830 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3831 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3832 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3833 ext4_msg(sb, KERN_ERR,
3834 "unsupported descriptor size %lu",
0d1ee42f
AR		 3835 sbi->s_desc_size);
3836 goto failed_mount;
3837 }
3838 } else
3839 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3840
ac27a0ec 3841 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3842 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 3843
617ba13b 3844 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3845 if (sbi->s_inodes_per_block == 0)
617ba13b 3846 goto cantfind_ext4;
cd6bb35b
TT		 3847 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3848 sbi->s_inodes_per_group > blocksize * 8) {
3849 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3850 sbi->s_blocks_per_group);
3851 goto failed_mount;
3852 }
ac27a0ec
DK		 3853 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3854 sbi->s_inodes_per_block;
0d1ee42f 3855 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3856 sbi->s_sbh = bh;
3857 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3858 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3859 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3860
2b2d6d01 3861 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3862 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3863 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 3864 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 3865 i = le32_to_cpu(es->s_flags);
3866 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3867 sbi->s_hash_unsigned = 3;
3868 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3869#ifdef __CHAR_UNSIGNED__
bc98a42c 3870 if (!sb_rdonly(sb))
23301410
TT		 3871 es->s_flags |=
3872 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3873 sbi->s_hash_unsigned = 3;
f99b2589 3874#else
bc98a42c 3875 if (!sb_rdonly(sb))
23301410
TT		 3876 es->s_flags |=
3877 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3878#endif
23301410 3879 }
f99b2589 3880 }
ac27a0ec 3881
281b5995
TT		 3882 /* Handle clustersize */
3883 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 3884 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 3885 if (has_bigalloc) {
3886 if (clustersize < blocksize) {
3887 ext4_msg(sb, KERN_ERR,
3888 "cluster size (%d) smaller than "
3889 "block size (%d)", clustersize, blocksize);
3890 goto failed_mount;
3891 }
8cdf3372
TT		 3892 if (le32_to_cpu(es->s_log_cluster_size) >
3893 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3894 ext4_msg(sb, KERN_ERR,
3895 "Invalid log cluster size: %u",
3896 le32_to_cpu(es->s_log_cluster_size));
3897 goto failed_mount;
3898 }
281b5995
TT		 3899 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3900 le32_to_cpu(es->s_log_block_size);
3901 sbi->s_clusters_per_group =
3902 le32_to_cpu(es->s_clusters_per_group);
3903 if (sbi->s_clusters_per_group > blocksize * 8) {
3904 ext4_msg(sb, KERN_ERR,
3905 "#clusters per group too big: %lu",
3906 sbi->s_clusters_per_group);
3907 goto failed_mount;
3908 }
3909 if (sbi->s_blocks_per_group !=
3910 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3911 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3912 "clusters per group (%lu) inconsistent",
3913 sbi->s_blocks_per_group,
3914 sbi->s_clusters_per_group);
3915 goto failed_mount;
3916 }
3917 } else {
3918 if (clustersize != blocksize) {
3919 ext4_warning(sb, "fragment/cluster size (%d) != "
3920 "block size (%d)", clustersize,
3921 blocksize);
3922 clustersize = blocksize;
3923 }
3924 if (sbi->s_blocks_per_group > blocksize * 8) {
3925 ext4_msg(sb, KERN_ERR,
3926 "#blocks per group too big: %lu",
3927 sbi->s_blocks_per_group);
3928 goto failed_mount;
3929 }
3930 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3931 sbi->s_cluster_bits = 0;
ac27a0ec 3932 }
281b5995
TT		 3933 sbi->s_cluster_ratio = clustersize / blocksize;
3934
960fd856
TT		 3935 /* Do we have standard group size of clustersize * 8 blocks ? */
3936 if (sbi->s_blocks_per_group == clustersize << 3)
3937 set_opt2(sb, STD_GROUP_SIZE);
3938
bf43d84b
ES		 3939 /*
3940 * Test whether we have more sectors than will fit in sector_t,
3941 * and whether the max offset is addressable by the page cache.
3942 */
5a9ae68a 3943 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3944 ext4_blocks_count(es));
5a9ae68a 3945 if (err) {
b31e1552 3946 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3947 " too large to mount safely on this system");
ac27a0ec 3948 if (sizeof(sector_t) < 8)
90c699a9 3949 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3950 goto failed_mount;
3951 }
3952
617ba13b
MC		 3953 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3954 goto cantfind_ext4;
e7c95593 3955
0f2ddca6
FTN		 3956 /* check blocks count against device size */
3957 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3958 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3959 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3960 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3961 ext4_blocks_count(es), blocks_count);
3962 goto failed_mount;
3963 }
3964
0b8e58a1
AD		 3965 /*
3966 * It makes no sense for the first data block to be beyond the end
3967 * of the filesystem.
3968 */
3969 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3970 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3971 "block %u is beyond end of filesystem (%llu)",
3972 le32_to_cpu(es->s_first_data_block),
3973 ext4_blocks_count(es));
e7c95593
ES		 3974 goto failed_mount;
3975 }
bd81d8ee
LV		 3976 blocks_count = (ext4_blocks_count(es) -
3977 le32_to_cpu(es->s_first_data_block) +
3978 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3979 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3980 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3981 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3982 "(block count %llu, first data block %u, "
b31e1552 3983 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3984 ext4_blocks_count(es),
3985 le32_to_cpu(es->s_first_data_block),
3986 EXT4_BLOCKS_PER_GROUP(sb));
3987 goto failed_mount;
3988 }
bd81d8ee 3989 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3990 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3991 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3992 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3993 EXT4_DESC_PER_BLOCK(sb);
3a4b77cd 3994 if (ext4_has_feature_meta_bg(sb)) {
2ba3e6e8 3995 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3a4b77cd
EG		 3996 ext4_msg(sb, KERN_WARNING,
3997 "first meta block group too large: %u "
3998 "(group descriptor block count %u)",
3999 le32_to_cpu(es->s_first_meta_bg), db_count);
4000 goto failed_mount;
4001 }
4002 }
a7c3e901 4003 sbi->s_group_desc = kvmalloc(db_count *
f18a5f21
TT		 4004 sizeof(struct buffer_head *),
4005 GFP_KERNEL);
ac27a0ec 4006 if (sbi->s_group_desc == NULL) {
b31e1552 4007 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 4008 ret = -ENOMEM;
ac27a0ec
DK		 4009 goto failed_mount;
4010 }
4011
705895b6 4012 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec 4013
85c8f176
AP		 4014 /* Pre-read the descriptors into the buffer cache */
4015 for (i = 0; i < db_count; i++) {
4016 block = descriptor_loc(sb, logical_sb_block, i);
4017 sb_breadahead(sb, block);
4018 }
4019
ac27a0ec 4020 for (i = 0; i < db_count; i++) {
70bbb3e0 4021 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 4022 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 4023 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 4024 ext4_msg(sb, KERN_ERR,
4025 "can't read group descriptor %d", i);
ac27a0ec
DK		 4026 db_count = i;
4027 goto failed_mount2;
4028 }
4029 }
829fa70d 4030 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 4031 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 4032 ret = -EFSCORRUPTED;
f9ae9cf5 4033 goto failed_mount2;
ac27a0ec 4034 }
772cb7c8 4035
f9ae9cf5 4036 sbi->s_gdb_count = db_count;
ac27a0ec 4037
235699a8 4038 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
04496411 4039
a75ae78f 4040 /* Register extent status tree shrinker */
eb68d0e2 4041 if (ext4_es_register_shrinker(sbi))
ce7e010a 4042 goto failed_mount3;
ce7e010a 4043
c9de560d 4044 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 4045 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 4046
f9ae9cf5
TT		 4047 /*
4048 * set up enough so that it can read an inode
4049 */
f6e63f90 4050 sb->s_op = &ext4_sops;
617ba13b
MC		 4051 sb->s_export_op = &ext4_export_ops;
4052 sb->s_xattr = ext4_xattr_handlers;
ffcc4182 4053#ifdef CONFIG_EXT4_FS_ENCRYPTION
a7550b30 4054 sb->s_cop = &ext4_cryptops;
ffcc4182 4055#endif
ac27a0ec 4056#ifdef CONFIG_QUOTA
617ba13b 4057 sb->dq_op = &ext4_quota_operations;
e2b911c5 4058 if (ext4_has_feature_quota(sb))
1fa5efe3 4059 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 4060 else
4061 sb->s_qcop = &ext4_qctl_operations;
689c958c 4062 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 4063#endif
85787090 4064 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
f2fa2ffc 4065
ac27a0ec 4066 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 4067 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 4068
4069 sb->s_root = NULL;
4070
4071 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 4072 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 4073
bc98a42c 4074 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
c5e06d10 4075 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 4076 goto failed_mount3a;
c5e06d10 4077
ac27a0ec
DK		 4078 /*
4079 * The first inode we look at is the journal inode. Don't try
4080 * root first: it may be modified in the journal!
4081 */
e2b911c5 4082 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4753d8a2
TT		 4083 err = ext4_load_journal(sb, es, journal_devnum);
4084 if (err)
50460fe8 4085 goto failed_mount3a;
bc98a42c 4086 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
e2b911c5 4087 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 4088 ext4_msg(sb, KERN_ERR, "required journal recovery "
4089 "suppressed and not mounted read-only");
744692dc 4090 goto failed_mount_wq;
ac27a0ec 4091 } else {
1e381f60
DM		 4092 /* Nojournal mode, all journal mount options are illegal */
4093 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4094 ext4_msg(sb, KERN_ERR, "can't mount with "
4095 "journal_checksum, fs mounted w/o journal");
4096 goto failed_mount_wq;
4097 }
4098 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4099 ext4_msg(sb, KERN_ERR, "can't mount with "
4100 "journal_async_commit, fs mounted w/o journal");
4101 goto failed_mount_wq;
4102 }
4103 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4104 ext4_msg(sb, KERN_ERR, "can't mount with "
4105 "commit=%lu, fs mounted w/o journal",
4106 sbi->s_commit_interval / HZ);
4107 goto failed_mount_wq;
4108 }
4109 if (EXT4_MOUNT_DATA_FLAGS &
4110 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4111 ext4_msg(sb, KERN_ERR, "can't mount with "
4112 "data=, fs mounted w/o journal");
4113 goto failed_mount_wq;
4114 }
4115 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4116 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 4117 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 4118 sbi->s_journal = NULL;
4119 needs_recovery = 0;
4120 goto no_journal;
ac27a0ec
DK		 4121 }
4122
e2b911c5 4123 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 4124 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4125 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4126 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4127 goto failed_mount_wq;
eb40a09c
JS		 4128 }
4129
25ed6e8a
DW		 4130 if (!set_journal_csum_feature_set(sb)) {
4131 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4132 "feature set");
4133 goto failed_mount_wq;
d4da6c9c 4134 }
818d276c 4135
ac27a0ec
DK		 4136 /* We have now updated the journal if required, so we can
4137 * validate the data journaling mode. */
4138 switch (test_opt(sb, DATA_FLAGS)) {
4139 case 0:
4140 /* No mode set, assume a default based on the journal
63f57933
AM		 4141 * capabilities: ORDERED_DATA if the journal can
4142 * cope, else JOURNAL_DATA
4143 */
dab291af 4144 if (jbd2_journal_check_available_features
27f394a7 4145 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
fd8c37ec 4146 set_opt(sb, ORDERED_DATA);
27f394a7
TN		 4147 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4148 } else {
fd8c37ec 4149 set_opt(sb, JOURNAL_DATA);
27f394a7
TN		 4150 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4151 }
ac27a0ec
DK		 4152 break;
4153
617ba13b
MC		 4154 case EXT4_MOUNT_ORDERED_DATA:
4155 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4156 if (!jbd2_journal_check_available_features
4157 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4158 ext4_msg(sb, KERN_ERR, "Journal does not support "
4159 "requested data journaling mode");
744692dc 4160 goto failed_mount_wq;
ac27a0ec
DK		 4161 }
4162 default:
4163 break;
4164 }
ab04df78
JK		 4165
4166 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4167 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4168 ext4_msg(sb, KERN_ERR, "can't mount with "
4169 "journal_async_commit in data=ordered mode");
4170 goto failed_mount_wq;
4171 }
4172
b3881f74 4173 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4174
18aadd47
BJ		 4175 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4176
ce7e010a 4177no_journal:
cdb7ee4c
TE		 4178 if (!test_opt(sb, NO_MBCACHE)) {
4179 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4180 if (!sbi->s_ea_block_cache) {
dec214d0 4181 ext4_msg(sb, KERN_ERR,
cdb7ee4c 4182 "Failed to create ea_block_cache");
dec214d0
TE		 4183 goto failed_mount_wq;
4184 }
cdb7ee4c
TE		 4185
4186 if (ext4_has_feature_ea_inode(sb)) {
4187 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4188 if (!sbi->s_ea_inode_cache) {
4189 ext4_msg(sb, KERN_ERR,
4190 "Failed to create ea_inode_cache");
4191 goto failed_mount_wq;
4192 }
4193 }
9c191f70
M		 4194 }
4195
e2b911c5 4196 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
09cbfeaf 4197 (blocksize != PAGE_SIZE)) {
1cb767cd
TT		 4198 ext4_msg(sb, KERN_ERR,
4199 "Unsupported blocksize for fs encryption");
4200 goto failed_mount_wq;
4201 }
4202
bc98a42c 4203 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
e2b911c5
DW		 4204 !ext4_has_feature_encrypt(sb)) {
4205 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 4206 ext4_commit_super(sb, 1);
4207 }
4208
952fc18e
TT		 4209 /*
4210 * Get the # of file system overhead blocks from the
4211 * superblock if present.
4212 */
4213 if (es->s_overhead_clusters)
4214 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4215 else {
07aa2ea1
LC		 4216 err = ext4_calculate_overhead(sb);
4217 if (err)
952fc18e
TT		 4218 goto failed_mount_wq;
4219 }
4220
fd89d5f2
TH		 4221 /*
4222 * The maximum number of concurrent works can be high and
4223 * concurrency isn't really necessary. Limit it to 1.
4224 */
2e8fa54e
JK		 4225 EXT4_SB(sb)->rsv_conversion_wq =
4226 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4227 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4228 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4229 ret = -ENOMEM;
2e8fa54e
JK		 4230 goto failed_mount4;
4231 }
4232
ac27a0ec 4233 /*
dab291af 4234 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4235 * so we can safely mount the rest of the filesystem now.
4236 */
4237
1d1fe1ee
DH		 4238 root = ext4_iget(sb, EXT4_ROOT_INO);
4239 if (IS_ERR(root)) {
b31e1552 4240 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4241 ret = PTR_ERR(root);
32a9bb57 4242 root = NULL;
ac27a0ec
DK		 4243 goto failed_mount4;
4244 }
4245 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4246 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4247 iput(root);
ac27a0ec
DK		 4248 goto failed_mount4;
4249 }
48fde701 4250 sb->s_root = d_make_root(root);
1d1fe1ee 4251 if (!sb->s_root) {
b31e1552 4252 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4253 ret = -ENOMEM;
4254 goto failed_mount4;
4255 }
ac27a0ec 4256
c89128a0
JK		 4257 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4258 if (ret == -EROFS) {
1751e8a6 4259 sb->s_flags |= SB_RDONLY;
c89128a0
JK		 4260 ret = 0;
4261 } else if (ret)
4262 goto failed_mount4a;
ef7f3835
KS		 4263
4264 /* determine the minimum size of new large inodes, if present */
670e9875
TT		 4265 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4266 sbi->s_want_extra_isize == 0) {
ef7f3835
KS		 4267 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4268 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 4269 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 4270 if (sbi->s_want_extra_isize <
4271 le16_to_cpu(es->s_want_extra_isize))
4272 sbi->s_want_extra_isize =
4273 le16_to_cpu(es->s_want_extra_isize);
4274 if (sbi->s_want_extra_isize <
4275 le16_to_cpu(es->s_min_extra_isize))
4276 sbi->s_want_extra_isize =
4277 le16_to_cpu(es->s_min_extra_isize);
4278 }
4279 }
4280 /* Check if enough inode space is available */
4281 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4282 sbi->s_inode_size) {
4283 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4284 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4285 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4286 "available");
ef7f3835
KS		 4287 }
4288
b5799018 4289 ext4_set_resv_clusters(sb);
27dd4385 4290
6fd058f7
TT		 4291 err = ext4_setup_system_zone(sb);
4292 if (err) {
b31e1552 4293 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4294 "zone (%d)", err);
f9ae9cf5
TT		 4295 goto failed_mount4a;
4296 }
4297
4298 ext4_ext_init(sb);
4299 err = ext4_mb_init(sb);
4300 if (err) {
4301 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4302 err);
dcf2d804 4303 goto failed_mount5;
c2774d84
AK		 4304 }
4305
d5e03cbb
TT		 4306 block = ext4_count_free_clusters(sb);
4307 ext4_free_blocks_count_set(sbi->s_es,
4308 EXT4_C2B(sbi, block));
908c7f19
TH		 4309 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4310 GFP_KERNEL);
d5e03cbb
TT		 4311 if (!err) {
4312 unsigned long freei = ext4_count_free_inodes(sb);
4313 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4314 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4315 GFP_KERNEL);
d5e03cbb
TT		 4316 }
4317 if (!err)
4318 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4319 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4320 if (!err)
908c7f19
TH		 4321 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4322 GFP_KERNEL);
c8585c6f
DJ		 4323 if (!err)
4324 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4325
d5e03cbb
TT		 4326 if (err) {
4327 ext4_msg(sb, KERN_ERR, "insufficient memory");
4328 goto failed_mount6;
4329 }
4330
e2b911c5 4331 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 4332 if (!ext4_fill_flex_info(sb)) {
4333 ext4_msg(sb, KERN_ERR,
4334 "unable to initialize "
4335 "flex_bg meta info!");
4336 goto failed_mount6;
4337 }
4338
bfff6873
LC		 4339 err = ext4_register_li_request(sb, first_not_zeroed);
4340 if (err)
dcf2d804 4341 goto failed_mount6;
bfff6873 4342
b5799018 4343 err = ext4_register_sysfs(sb);
dcf2d804
TM		 4344 if (err)
4345 goto failed_mount7;
3197ebdb 4346
9b2ff357
JK		 4347#ifdef CONFIG_QUOTA
4348 /* Enable quota usage during mount. */
bc98a42c 4349 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
9b2ff357
JK		 4350 err = ext4_enable_quotas(sb);
4351 if (err)
4352 goto failed_mount8;
4353 }
4354#endif /* CONFIG_QUOTA */
4355
617ba13b
MC		 4356 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4357 ext4_orphan_cleanup(sb, es);
4358 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4359 if (needs_recovery) {
b31e1552 4360 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4361 ext4_mark_recovery_complete(sb, es);
4362 }
4363 if (EXT4_SB(sb)->s_journal) {
4364 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4365 descr = " journalled data mode";
4366 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4367 descr = " ordered data mode";
4368 else
4369 descr = " writeback data mode";
4370 } else
4371 descr = "out journal";
4372
79add3a3
LC		 4373 if (test_opt(sb, DISCARD)) {
4374 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4375 if (!blk_queue_discard(q))
4376 ext4_msg(sb, KERN_WARNING,
4377 "mounting with \"discard\" option, but "
4378 "the device does not support discard");
4379 }
4380
e294a537
TT		 4381 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4382 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 4383 "Opts: %.*s%s%s", descr,
4384 (int) sizeof(sbi->s_es->s_mount_opts),
4385 sbi->s_es->s_mount_opts,
e294a537 4386 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4387
66e61a9e
TT		 4388 if (es->s_error_count)
4389 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4390
efbed4dc
TT		 4391 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4392 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4393 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4394 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4395
d4c402d9 4396 kfree(orig_data);
ac27a0ec
DK		 4397 return 0;
4398
617ba13b 4399cantfind_ext4:
ac27a0ec 4400 if (!silent)
b31e1552 4401 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4402 goto failed_mount;
4403
72ba7450
TT		 4404#ifdef CONFIG_QUOTA
4405failed_mount8:
ebd173be 4406 ext4_unregister_sysfs(sb);
72ba7450 4407#endif
dcf2d804
TM		 4408failed_mount7:
4409 ext4_unregister_li_request(sb);
4410failed_mount6:
f9ae9cf5 4411 ext4_mb_release(sb);
d5e03cbb 4412 if (sbi->s_flex_groups)
b93b41d4 4413 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4414 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4415 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4416 percpu_counter_destroy(&sbi->s_dirs_counter);
4417 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4418failed_mount5:
f9ae9cf5
TT		 4419 ext4_ext_release(sb);
4420 ext4_release_system_zone(sb);
4421failed_mount4a:
94bf608a 4422 dput(sb->s_root);
32a9bb57 4423 sb->s_root = NULL;
94bf608a 4424failed_mount4:
b31e1552 4425 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4426 if (EXT4_SB(sb)->rsv_conversion_wq)
4427 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4428failed_mount_wq:
dec214d0
TE		 4429 if (sbi->s_ea_inode_cache) {
4430 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4431 sbi->s_ea_inode_cache = NULL;
4432 }
47387409
TE		 4433 if (sbi->s_ea_block_cache) {
4434 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4435 sbi->s_ea_block_cache = NULL;
82939d79 4436 }
0390131b
FM		 4437 if (sbi->s_journal) {
4438 jbd2_journal_destroy(sbi->s_journal);
4439 sbi->s_journal = NULL;
4440 }
50460fe8 4441failed_mount3a:
d3922a77 4442 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4443failed_mount3:
9105bb14 4444 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4445 if (sbi->s_mmp_tsk)
4446 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4447failed_mount2:
4448 for (i = 0; i < db_count; i++)
4449 brelse(sbi->s_group_desc[i]);
b93b41d4 4450 kvfree(sbi->s_group_desc);
ac27a0ec 4451failed_mount:
0441984a
DW		 4452 if (sbi->s_chksum_driver)
4453 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 4454#ifdef CONFIG_QUOTA
a2d4a646 4455 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4456 kfree(sbi->s_qf_names[i]);
4457#endif
617ba13b 4458 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4459 brelse(bh);
4460out_fail:
4461 sb->s_fs_info = NULL;
f6830165 4462 kfree(sbi->s_blockgroup_lock);
5aee0f8a 4463out_free_base:
ac27a0ec 4464 kfree(sbi);
d4c402d9 4465 kfree(orig_data);
5e405595 4466 fs_put_dax(dax_dev);
07aa2ea1 4467 return err ? err : ret;
ac27a0ec
DK		 4468}
4469
4470/*
4471 * Setup any per-fs journal parameters now. We'll do this both on
4472 * initial mount, once the journal has been initialised but before we've
4473 * done any recovery; and again on any subsequent remount.
4474 */
617ba13b 4475static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4476{
617ba13b 4477 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4478
30773840
TT		 4479 journal->j_commit_interval = sbi->s_commit_interval;
4480 journal->j_min_batch_time = sbi->s_min_batch_time;
4481 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4482
a931da6a 4483 write_lock(&journal->j_state_lock);
ac27a0ec 4484 if (test_opt(sb, BARRIER))
dab291af 4485 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4486 else
dab291af 4487 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4488 if (test_opt(sb, DATA_ERR_ABORT))
4489 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4490 else
4491 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4492 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4493}
4494
c6cb7e77
EW		 4495static struct inode *ext4_get_journal_inode(struct super_block *sb,
4496 unsigned int journal_inum)
ac27a0ec
DK		 4497{
4498 struct inode *journal_inode;
ac27a0ec 4499
c6cb7e77
EW		 4500 /*
4501 * Test for the existence of a valid inode on disk. Bad things
4502 * happen if we iget() an unused inode, as the subsequent iput()
4503 * will try to delete it.
4504 */
1d1fe1ee
DH		 4505 journal_inode = ext4_iget(sb, journal_inum);
4506 if (IS_ERR(journal_inode)) {
b31e1552 4507 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4508 return NULL;
4509 }
4510 if (!journal_inode->i_nlink) {
4511 make_bad_inode(journal_inode);
4512 iput(journal_inode);
b31e1552 4513 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4514 return NULL;
4515 }
4516
e5f8eab8 4517 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4518 journal_inode, journal_inode->i_size);
1d1fe1ee 4519 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4520 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4521 iput(journal_inode);
4522 return NULL;
4523 }
c6cb7e77
EW		 4524 return journal_inode;
4525}
4526
4527static journal_t *ext4_get_journal(struct super_block *sb,
4528 unsigned int journal_inum)
4529{
4530 struct inode *journal_inode;
4531 journal_t *journal;
4532
4533 BUG_ON(!ext4_has_feature_journal(sb));
4534
4535 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4536 if (!journal_inode)
4537 return NULL;
ac27a0ec 4538
dab291af 4539 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4540 if (!journal) {
b31e1552 4541 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4542 iput(journal_inode);
4543 return NULL;
4544 }
4545 journal->j_private = sb;
617ba13b 4546 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4547 return journal;
4548}
4549
617ba13b 4550static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4551 dev_t j_dev)
4552{
2b2d6d01 4553 struct buffer_head *bh;
ac27a0ec 4554 journal_t *journal;
617ba13b
MC		 4555 ext4_fsblk_t start;
4556 ext4_fsblk_t len;
ac27a0ec 4557 int hblock, blocksize;
617ba13b 4558 ext4_fsblk_t sb_block;
ac27a0ec 4559 unsigned long offset;
2b2d6d01 4560 struct ext4_super_block *es;
ac27a0ec
DK		 4561 struct block_device *bdev;
4562
e2b911c5 4563 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4564
b31e1552 4565 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4566 if (bdev == NULL)
4567 return NULL;
4568
ac27a0ec 4569 blocksize = sb->s_blocksize;
e1defc4f 4570 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4571 if (blocksize < hblock) {
b31e1552
ES		 4572 ext4_msg(sb, KERN_ERR,
4573 "blocksize too small for journal device");
ac27a0ec
DK		 4574 goto out_bdev;
4575 }
4576
617ba13b
MC		 4577 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4578 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4579 set_blocksize(bdev, blocksize);
4580 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4581 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4582 "external journal");
ac27a0ec
DK		 4583 goto out_bdev;
4584 }
4585
2716b802 4586 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4587 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4588 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4589 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4590 ext4_msg(sb, KERN_ERR, "external journal has "
4591 "bad superblock");
ac27a0ec
DK		 4592 brelse(bh);
4593 goto out_bdev;
4594 }
4595
df4763be
DW		 4596 if ((le32_to_cpu(es->s_feature_ro_compat) &
4597 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4598 es->s_checksum != ext4_superblock_csum(sb, es)) {
4599 ext4_msg(sb, KERN_ERR, "external journal has "
4600 "corrupt superblock");
4601 brelse(bh);
4602 goto out_bdev;
4603 }
4604
617ba13b 4605 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4606 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4607 brelse(bh);
4608 goto out_bdev;
4609 }
4610
bd81d8ee 4611 len = ext4_blocks_count(es);
ac27a0ec
DK		 4612 start = sb_block + 1;
4613 brelse(bh); /* we're done with the superblock */
4614
dab291af 4615 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4616 start, len, blocksize);
4617 if (!journal) {
b31e1552 4618 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4619 goto out_bdev;
4620 }
4621 journal->j_private = sb;
dfec8a14 4622 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4623 wait_on_buffer(journal->j_sb_buffer);
4624 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4625 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4626 goto out_journal;
4627 }
4628 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4629 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4630 "user (unsupported) - %d",
ac27a0ec
DK		 4631 be32_to_cpu(journal->j_superblock->s_nr_users));
4632 goto out_journal;
4633 }
617ba13b
MC		 4634 EXT4_SB(sb)->journal_bdev = bdev;
4635 ext4_init_journal_params(sb, journal);
ac27a0ec 4636 return journal;
0b8e58a1 4637
ac27a0ec 4638out_journal:
dab291af 4639 jbd2_journal_destroy(journal);
ac27a0ec 4640out_bdev:
617ba13b 4641 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4642 return NULL;
4643}
4644
617ba13b
MC		 4645static int ext4_load_journal(struct super_block *sb,
4646 struct ext4_super_block *es,
ac27a0ec
DK		 4647 unsigned long journal_devnum)
4648{
4649 journal_t *journal;
4650 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4651 dev_t journal_dev;
4652 int err = 0;
4653 int really_read_only;
4654
e2b911c5 4655 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4656
ac27a0ec
DK		 4657 if (journal_devnum &&
4658 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4659 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4660 "numbers have changed");
ac27a0ec
DK		 4661 journal_dev = new_decode_dev(journal_devnum);
4662 } else
4663 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4664
4665 really_read_only = bdev_read_only(sb->s_bdev);
4666
4667 /*
4668 * Are we loading a blank journal or performing recovery after a
4669 * crash? For recovery, we need to check in advance whether we
4670 * can get read-write access to the device.
4671 */
e2b911c5 4672 if (ext4_has_feature_journal_needs_recovery(sb)) {
bc98a42c 4673 if (sb_rdonly(sb)) {
b31e1552
ES		 4674 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4675 "required on readonly filesystem");
ac27a0ec 4676 if (really_read_only) {
b31e1552 4677 ext4_msg(sb, KERN_ERR, "write access "
d98bf8cd
SR		 4678 "unavailable, cannot proceed "
4679 "(try mounting with noload)");
ac27a0ec
DK		 4680 return -EROFS;
4681 }
b31e1552
ES		 4682 ext4_msg(sb, KERN_INFO, "write access will "
4683 "be enabled during recovery");
ac27a0ec
DK		 4684 }
4685 }
4686
4687 if (journal_inum && journal_dev) {
b31e1552
ES		 4688 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4689 "and inode journals!");
ac27a0ec
DK		 4690 return -EINVAL;
4691 }
4692
4693 if (journal_inum) {
617ba13b 4694 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4695 return -EINVAL;
4696 } else {
617ba13b 4697 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4698 return -EINVAL;
4699 }
4700
90576c0b 4701 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4702 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4703
e2b911c5 4704 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4705 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4706 if (!err) {
4707 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4708 if (save)
4709 memcpy(save, ((char *) es) +
4710 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4711 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4712 if (save)
4713 memcpy(((char *) es) + EXT4_S_ERR_START,
4714 save, EXT4_S_ERR_LEN);
4715 kfree(save);
4716 }
ac27a0ec
DK		 4717
4718 if (err) {
b31e1552 4719 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4720 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4721 return err;
4722 }
4723
617ba13b
MC		 4724 EXT4_SB(sb)->s_journal = journal;
4725 ext4_clear_journal_err(sb, es);
ac27a0ec 4726
c41303ce 4727 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4728 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4729 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4730
4731 /* Make sure we flush the recovery flag to disk. */
e2d67052 4732 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4733 }
4734
4735 return 0;
4736}
4737
e2d67052 4738static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4739{
e2d67052 4740 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4741 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4742 int error = 0;
ac27a0ec 4743
bdfe0cbd 4744 if (!sbh || block_device_ejected(sb))
c4be0c1d 4745 return error;
71290b36
TT		 4746 /*
4747 * If the file system is mounted read-only, don't update the
4748 * superblock write time. This avoids updating the superblock
4749 * write time when we are mounting the root file system
4750 * read/only but we need to replay the journal; at that point,
4751 * for people who are east of GMT and who make their clock
4752 * tick in localtime for Windows bug-for-bug compatibility,
4753 * the clock is set in the future, and this will cause e2fsck
4754 * to complain and force a full file system check.
4755 */
1751e8a6 4756 if (!(sb->s_flags & SB_RDONLY))
71290b36 4757 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4758 if (sb->s_bdev->bd_part)
4759 es->s_kbytes_written =
4760 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4761 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4762 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4763 else
4764 es->s_kbytes_written =
4765 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4766 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4767 ext4_free_blocks_count_set(es,
57042651
TT		 4768 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4769 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4770 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4771 es->s_free_inodes_count =
4772 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4773 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4774 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4775 ext4_superblock_csum_set(sb);
1566a48a
TT		 4776 if (sync)
4777 lock_buffer(sbh);
4743f839
PKS		 4778 if (buffer_write_io_error(sbh)) {
4779 /*
4780 * Oh, dear. A previous attempt to write the
4781 * superblock failed. This could happen because the
4782 * USB device was yanked out. Or it could happen to
4783 * be a transient write error and maybe the block will
4784 * be remapped. Nothing we can do but to retry the
4785 * write and hope for the best.
4786 */
4787 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4788 "superblock detected");
4789 clear_buffer_write_io_error(sbh);
4790 set_buffer_uptodate(sbh);
4791 }
ac27a0ec 4792 mark_buffer_dirty(sbh);
914258bf 4793 if (sync) {
1566a48a 4794 unlock_buffer(sbh);
564bc402 4795 error = __sync_dirty_buffer(sbh,
00473374 4796 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
c89128a0 4797 if (buffer_write_io_error(sbh)) {
b31e1552
ES		 4798 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4799 "superblock");
914258bf
TT		 4800 clear_buffer_write_io_error(sbh);
4801 set_buffer_uptodate(sbh);
4802 }
4803 }
c4be0c1d 4804 return error;
ac27a0ec
DK		 4805}
4806
ac27a0ec
DK		 4807/*
4808 * Have we just finished recovery? If so, and if we are mounting (or
4809 * remounting) the filesystem readonly, then we will end up with a
4810 * consistent fs on disk. Record that fact.
4811 */
2b2d6d01
TT		 4812static void ext4_mark_recovery_complete(struct super_block *sb,
4813 struct ext4_super_block *es)
ac27a0ec 4814{
617ba13b 4815 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4816
e2b911c5 4817 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4818 BUG_ON(journal != NULL);
4819 return;
4820 }
dab291af 4821 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4822 if (jbd2_journal_flush(journal) < 0)
4823 goto out;
4824
bc98a42c 4825 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
e2b911c5 4826 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4827 ext4_commit_super(sb, 1);
ac27a0ec 4828 }
7ffe1ea8
HK		 4829
4830out:
dab291af 4831 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4832}
4833
4834/*
4835 * If we are mounting (or read-write remounting) a filesystem whose journal
4836 * has recorded an error from a previous lifetime, move that error to the
4837 * main filesystem now.
4838 */
2b2d6d01
TT		 4839static void ext4_clear_journal_err(struct super_block *sb,
4840 struct ext4_super_block *es)
ac27a0ec
DK		 4841{
4842 journal_t *journal;
4843 int j_errno;
4844 const char *errstr;
4845
e2b911c5 4846 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4847
617ba13b 4848 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4849
4850 /*
4851 * Now check for any error status which may have been recorded in the
617ba13b 4852 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4853 */
4854
dab291af 4855 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4856 if (j_errno) {
4857 char nbuf[16];
4858
617ba13b 4859 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4860 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4861 "from previous mount: %s", errstr);
12062ddd 4862 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4863
617ba13b
MC		 4864 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4865 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4866 ext4_commit_super(sb, 1);
ac27a0ec 4867
dab291af 4868 jbd2_journal_clear_err(journal);
d796c52e 4869 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4870 }
4871}
4872
4873/*
4874 * Force the running and committing transactions to commit,
4875 * and wait on the commit.
4876 */
617ba13b 4877int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4878{
4879 journal_t *journal;
ac27a0ec 4880
bc98a42c 4881 if (sb_rdonly(sb))
ac27a0ec
DK		 4882 return 0;
4883
617ba13b 4884 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4885 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4886}
4887
617ba13b 4888static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4889{
14ce0cb4 4890 int ret = 0;
9eddacf9 4891 tid_t target;
06a407f1 4892 bool needs_barrier = false;
8d5d02e6 4893 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4894
49598e04 4895 if (unlikely(ext4_forced_shutdown(sbi)))
0db1ff22
TT		 4896 return 0;
4897
9bffad1e 4898 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4899 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4900 /*
4901 * Writeback quota in non-journalled quota case - journalled quota has
4902 * no dirty dquots
4903 */
4904 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4905 /*
4906 * Data writeback is possible w/o journal transaction, so barrier must
4907 * being sent at the end of the function. But we can skip it if
4908 * transaction_commit will do it for us.
4909 */
bda32530
TT		 4910 if (sbi->s_journal) {
4911 target = jbd2_get_latest_transaction(sbi->s_journal);
4912 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4913 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4914 needs_barrier = true;
4915
4916 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4917 if (wait)
4918 ret = jbd2_log_wait_commit(sbi->s_journal,
4919 target);
4920 }
4921 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4922 needs_barrier = true;
06a407f1
DM		 4923 if (needs_barrier) {
4924 int err;
4925 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4926 if (!ret)
4927 ret = err;
0390131b 4928 }
06a407f1
DM		 4929
4930 return ret;
4931}
4932
ac27a0ec
DK		 4933/*
4934 * LVM calls this function before a (read-only) snapshot is created. This
4935 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4936 *
4937 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4938 * state independently. It relies on upper layer to stop all data & metadata
4939 * modifications.
ac27a0ec 4940 */
c4be0c1d 4941static int ext4_freeze(struct super_block *sb)
ac27a0ec 4942{
c4be0c1d
TS		 4943 int error = 0;
4944 journal_t *journal;
ac27a0ec 4945
bc98a42c 4946 if (sb_rdonly(sb))
9ca92389 4947 return 0;
ac27a0ec 4948
9ca92389 4949 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4950
bb044576
TT		 4951 if (journal) {
4952 /* Now we set up the journal barrier. */
4953 jbd2_journal_lock_updates(journal);
ac27a0ec 4954
bb044576
TT		 4955 /*
4956 * Don't clear the needs_recovery flag if we failed to
4957 * flush the journal.
4958 */
4959 error = jbd2_journal_flush(journal);
4960 if (error < 0)
4961 goto out;
c642dc9e
ES		 4962
4963 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 4964 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 4965 }
9ca92389 4966
9ca92389 4967 error = ext4_commit_super(sb, 1);
6b0310fb 4968out:
bb044576
TT		 4969 if (journal)
4970 /* we rely on upper layer to stop further updates */
4971 jbd2_journal_unlock_updates(journal);
6b0310fb 4972 return error;
ac27a0ec
DK		 4973}
4974
4975/*
4976 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4977 * flag here, even though the filesystem is not technically dirty yet.
4978 */
c4be0c1d 4979static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4980{
bc98a42c 4981 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 4982 return 0;
4983
c642dc9e
ES		 4984 if (EXT4_SB(sb)->s_journal) {
4985 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 4986 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 4987 }
4988
9ca92389 4989 ext4_commit_super(sb, 1);
c4be0c1d 4990 return 0;
ac27a0ec
DK		 4991}
4992
673c6100
TT		 4993/*
4994 * Structure to save mount options for ext4_remount's benefit
4995 */
4996struct ext4_mount_options {
4997 unsigned long s_mount_opt;
a2595b8a 4998 unsigned long s_mount_opt2;
08cefc7a
EB		 4999 kuid_t s_resuid;
5000 kgid_t s_resgid;
673c6100
TT		 5001 unsigned long s_commit_interval;
5002 u32 s_min_batch_time, s_max_batch_time;
5003#ifdef CONFIG_QUOTA
5004 int s_jquota_fmt;
a2d4a646 5005 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 5006#endif
5007};
5008
2b2d6d01 5009static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 5010{
2b2d6d01 5011 struct ext4_super_block *es;
617ba13b 5012 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5013 unsigned long old_sb_flags;
617ba13b 5014 struct ext4_mount_options old_opts;
c79d967d 5015 int enable_quota = 0;
8a266467 5016 ext4_group_t g;
b3881f74 5017 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 5018 int err = 0;
ac27a0ec 5019#ifdef CONFIG_QUOTA
03dafb5f 5020 int i, j;
ac27a0ec 5021#endif
d4c402d9 5022 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 5023
5024 /* Store the original options */
5025 old_sb_flags = sb->s_flags;
5026 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 5027 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 5028 old_opts.s_resuid = sbi->s_resuid;
5029 old_opts.s_resgid = sbi->s_resgid;
5030 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 5031 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5032 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 5033#ifdef CONFIG_QUOTA
5034 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 5035 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 5036 if (sbi->s_qf_names[i]) {
5037 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
5038 GFP_KERNEL);
5039 if (!old_opts.s_qf_names[i]) {
5040 for (j = 0; j < i; j++)
5041 kfree(old_opts.s_qf_names[j]);
3e36a163 5042 kfree(orig_data);
03dafb5f
CG		 5043 return -ENOMEM;
5044 }
5045 } else
5046 old_opts.s_qf_names[i] = NULL;
ac27a0ec 5047#endif
b3881f74
TT		 5048 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5049 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 5050
661aa520 5051 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 5052 err = -EINVAL;
5053 goto restore_opts;
5054 }
5055
6b992ff2 5056 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 5057 test_opt(sb, JOURNAL_CHECKSUM)) {
5058 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 5059 "during remount not supported; ignoring");
5060 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 5061 }
5062
6ae6514b
PS		 5063 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5064 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5065 ext4_msg(sb, KERN_ERR, "can't mount with "
5066 "both data=journal and delalloc");
5067 err = -EINVAL;
5068 goto restore_opts;
5069 }
5070 if (test_opt(sb, DIOREAD_NOLOCK)) {
5071 ext4_msg(sb, KERN_ERR, "can't mount with "
5072 "both data=journal and dioread_nolock");
5073 err = -EINVAL;
5074 goto restore_opts;
5075 }
923ae0ff
RZ		 5076 if (test_opt(sb, DAX)) {
5077 ext4_msg(sb, KERN_ERR, "can't mount with "
5078 "both data=journal and dax");
5079 err = -EINVAL;
5080 goto restore_opts;
5081 }
ab04df78
JK		 5082 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5083 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5084 ext4_msg(sb, KERN_ERR, "can't mount with "
5085 "journal_async_commit in data=ordered mode");
5086 err = -EINVAL;
5087 goto restore_opts;
5088 }
923ae0ff
RZ		 5089 }
5090
cdb7ee4c
TE		 5091 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5092 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5093 err = -EINVAL;
5094 goto restore_opts;
5095 }
5096
923ae0ff
RZ		 5097 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5098 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5099 "dax flag with busy inodes while remounting");
5100 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 5101 }
5102
4ab2f15b 5103 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 5104 ext4_abort(sb, "Abort forced by user");
ac27a0ec 5105
1751e8a6
LT		 5106 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5107 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec
DK		 5108
5109 es = sbi->s_es;
5110
b3881f74 5111 if (sbi->s_journal) {
0390131b 5112 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 5113 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5114 }
ac27a0ec 5115
1751e8a6
LT		 5116 if (*flags & SB_LAZYTIME)
5117 sb->s_flags |= SB_LAZYTIME;
a2fd66d0 5118
1751e8a6 5119 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
4ab2f15b 5120 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 5121 err = -EROFS;
5122 goto restore_opts;
5123 }
5124
1751e8a6 5125 if (*flags & SB_RDONLY) {
38c03b34
TT		 5126 err = sync_filesystem(sb);
5127 if (err < 0)
5128 goto restore_opts;
0f0dd62f
CH		 5129 err = dquot_suspend(sb, -1);
5130 if (err < 0)
c79d967d 5131 goto restore_opts;
c79d967d 5132
ac27a0ec
DK		 5133 /*
5134 * First of all, the unconditional stuff we have to do
5135 * to disable replay of the journal when we next remount
5136 */
1751e8a6 5137 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 5138
5139 /*
5140 * OK, test if we are remounting a valid rw partition
5141 * readonly, and if so set the rdonly flag and then
5142 * mark the partition as valid again.
5143 */
617ba13b
MC		 5144 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5145 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 5146 es->s_state = cpu_to_le16(sbi->s_mount_state);
5147
a63c9eb2 5148 if (sbi->s_journal)
0390131b 5149 ext4_mark_recovery_complete(sb, es);
ac27a0ec 5150 } else {
a13fb1a4 5151 /* Make sure we can mount this feature set readwrite */
e2b911c5 5152 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 5153 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 5154 err = -EROFS;
5155 goto restore_opts;
5156 }
8a266467
TT		 5157 /*
5158 * Make sure the group descriptor checksums
0b8e58a1 5159 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 5160 */
5161 for (g = 0; g < sbi->s_groups_count; g++) {
5162 struct ext4_group_desc *gdp =
5163 ext4_get_group_desc(sb, g, NULL);
5164
feb0ab32 5165 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5166 ext4_msg(sb, KERN_ERR,
5167 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 5168 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 5169 le16_to_cpu(gdp->bg_checksum));
6a797d27 5170 err = -EFSBADCRC;
8a266467
TT		 5171 goto restore_opts;
5172 }
5173 }
5174
ead6596b
ES		 5175 /*
5176 * If we have an unprocessed orphan list hanging
5177 * around from a previously readonly bdev mount,
5178 * require a full umount/remount for now.
5179 */
5180 if (es->s_last_orphan) {
b31e1552 5181 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5182 "remount RDWR because of unprocessed "
5183 "orphan inode list. Please "
b31e1552 5184 "umount/remount instead");
ead6596b
ES		 5185 err = -EINVAL;
5186 goto restore_opts;
5187 }
5188
ac27a0ec
DK		 5189 /*
5190 * Mounting a RDONLY partition read-write, so reread
5191 * and store the current valid flag. (It may have
5192 * been changed by e2fsck since we originally mounted
5193 * the partition.)
5194 */
0390131b
FM		 5195 if (sbi->s_journal)
5196 ext4_clear_journal_err(sb, es);
ac27a0ec 5197 sbi->s_mount_state = le16_to_cpu(es->s_state);
c89128a0
JK		 5198
5199 err = ext4_setup_super(sb, es, 0);
5200 if (err)
5201 goto restore_opts;
5202
5203 sb->s_flags &= ~SB_RDONLY;
e2b911c5 5204 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 5205 if (ext4_multi_mount_protect(sb,
5206 le64_to_cpu(es->s_mmp_block))) {
5207 err = -EROFS;
5208 goto restore_opts;
5209 }
c79d967d 5210 enable_quota = 1;
ac27a0ec
DK		 5211 }
5212 }
bfff6873
LC		 5213
5214 /*
5215 * Reinitialize lazy itable initialization thread based on
5216 * current settings
5217 */
bc98a42c 5218 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
bfff6873
LC		 5219 ext4_unregister_li_request(sb);
5220 else {
5221 ext4_group_t first_not_zeroed;
5222 first_not_zeroed = ext4_has_uninit_itable(sb);
5223 ext4_register_li_request(sb, first_not_zeroed);
5224 }
5225
6fd058f7 5226 ext4_setup_system_zone(sb);
c89128a0
JK		 5227 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5228 err = ext4_commit_super(sb, 1);
5229 if (err)
5230 goto restore_opts;
5231 }
0390131b 5232
ac27a0ec
DK		 5233#ifdef CONFIG_QUOTA
5234 /* Release old quota file names */
a2d4a646 5235 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5236 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5237 if (enable_quota) {
5238 if (sb_any_quota_suspended(sb))
5239 dquot_resume(sb, -1);
e2b911c5 5240 else if (ext4_has_feature_quota(sb)) {
7c319d32 5241 err = ext4_enable_quotas(sb);
07724f98 5242 if (err)
7c319d32 5243 goto restore_opts;
7c319d32
AK		 5244 }
5245 }
ac27a0ec 5246#endif
d4c402d9 5247
1751e8a6 5248 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
d4c402d9
CW		 5249 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5250 kfree(orig_data);
ac27a0ec 5251 return 0;
0b8e58a1 5252
ac27a0ec
DK		 5253restore_opts:
5254 sb->s_flags = old_sb_flags;
5255 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5256 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5257 sbi->s_resuid = old_opts.s_resuid;
5258 sbi->s_resgid = old_opts.s_resgid;
5259 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5260 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5261 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5262#ifdef CONFIG_QUOTA
5263 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5264 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5265 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5266 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5267 }
5268#endif
d4c402d9 5269 kfree(orig_data);
ac27a0ec
DK		 5270 return err;
5271}
5272
689c958c
LX		 5273#ifdef CONFIG_QUOTA
5274static int ext4_statfs_project(struct super_block *sb,
5275 kprojid_t projid, struct kstatfs *buf)
5276{
5277 struct kqid qid;
5278 struct dquot *dquot;
5279 u64 limit;
5280 u64 curblock;
5281
5282 qid = make_kqid_projid(projid);
5283 dquot = dqget(sb, qid);
5284 if (IS_ERR(dquot))
5285 return PTR_ERR(dquot);
7b9ca4c6 5286 spin_lock(&dquot->dq_dqb_lock);
689c958c
LX		 5287
5288 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5289 dquot->dq_dqb.dqb_bsoftlimit :
5290 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5291 if (limit && buf->f_blocks > limit) {
5292 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5293 buf->f_blocks = limit;
5294 buf->f_bfree = buf->f_bavail =
5295 (buf->f_blocks > curblock) ?
5296 (buf->f_blocks - curblock) : 0;
5297 }
5298
5299 limit = dquot->dq_dqb.dqb_isoftlimit ?
5300 dquot->dq_dqb.dqb_isoftlimit :
5301 dquot->dq_dqb.dqb_ihardlimit;
5302 if (limit && buf->f_files > limit) {
5303 buf->f_files = limit;
5304 buf->f_ffree =
5305 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5306 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5307 }
5308
7b9ca4c6 5309 spin_unlock(&dquot->dq_dqb_lock);
689c958c
LX		 5310 dqput(dquot);
5311 return 0;
5312}
5313#endif
5314
2b2d6d01 5315static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5316{
5317 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5318 struct ext4_sb_info *sbi = EXT4_SB(sb);
5319 struct ext4_super_block *es = sbi->s_es;
27dd4385 5320 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5321 u64 fsid;
d02a9391 5322 s64 bfree;
27dd4385 5323 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5324
952fc18e
TT		 5325 if (!test_opt(sb, MINIX_DF))
5326 overhead = sbi->s_overhead;
ac27a0ec 5327
617ba13b 5328 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5329 buf->f_bsize = sb->s_blocksize;
b72f78cb 5330 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5331 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5332 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5333 /* prevent underflow in case that few free space is available */
57042651 5334 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5335 buf->f_bavail = buf->f_bfree -
5336 (ext4_r_blocks_count(es) + resv_blocks);
5337 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5338 buf->f_bavail = 0;
5339 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5340 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5341 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5342 fsid = le64_to_cpup((void *)es->s_uuid) ^
5343 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5344 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5345 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5346
689c958c
LX		 5347#ifdef CONFIG_QUOTA
5348 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5349 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5350 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5351#endif
ac27a0ec
DK		 5352 return 0;
5353}
5354
ac27a0ec
DK		 5355
5356#ifdef CONFIG_QUOTA
5357
bc8230ee
JK		 5358/*
5359 * Helper functions so that transaction is started before we acquire dqio_sem
5360 * to keep correct lock ordering of transaction > dqio_sem
5361 */
ac27a0ec
DK		 5362static inline struct inode *dquot_to_inode(struct dquot *dquot)
5363{
4c376dca 5364 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5365}
5366
617ba13b 5367static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5368{
5369 int ret, err;
5370 handle_t *handle;
5371 struct inode *inode;
5372
5373 inode = dquot_to_inode(dquot);
9924a92a 5374 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5375 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5376 if (IS_ERR(handle))
5377 return PTR_ERR(handle);
5378 ret = dquot_commit(dquot);
617ba13b 5379 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5380 if (!ret)
5381 ret = err;
5382 return ret;
5383}
5384
617ba13b 5385static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5386{
5387 int ret, err;
5388 handle_t *handle;
5389
9924a92a 5390 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5391 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5392 if (IS_ERR(handle))
5393 return PTR_ERR(handle);
5394 ret = dquot_acquire(dquot);
617ba13b 5395 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5396 if (!ret)
5397 ret = err;
5398 return ret;
5399}
5400
617ba13b 5401static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5402{
5403 int ret, err;
5404 handle_t *handle;
5405
9924a92a 5406 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5407 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5408 if (IS_ERR(handle)) {
5409 /* Release dquot anyway to avoid endless cycle in dqput() */
5410 dquot_release(dquot);
ac27a0ec 5411 return PTR_ERR(handle);
9c3013e9 5412 }
ac27a0ec 5413 ret = dquot_release(dquot);
617ba13b 5414 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5415 if (!ret)
5416 ret = err;
5417 return ret;
5418}
5419
617ba13b 5420static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5421{
262b4662
JK		 5422 struct super_block *sb = dquot->dq_sb;
5423 struct ext4_sb_info *sbi = EXT4_SB(sb);
5424
2c8be6b2 5425 /* Are we journaling quotas? */
e2b911c5 5426 if (ext4_has_feature_quota(sb) ||
262b4662 5427 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5428 dquot_mark_dquot_dirty(dquot);
617ba13b 5429 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5430 } else {
5431 return dquot_mark_dquot_dirty(dquot);
5432 }
5433}
5434
617ba13b 5435static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5436{
5437 int ret, err;
5438 handle_t *handle;
5439
5440 /* Data block + inode block */
2b0143b5 5441 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5442 if (IS_ERR(handle))
5443 return PTR_ERR(handle);
5444 ret = dquot_commit_info(sb, type);
617ba13b 5445 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5446 if (!ret)
5447 ret = err;
5448 return ret;
5449}
5450
5451/*
5452 * Turn on quotas during mount time - we need to find
5453 * the quota file and such...
5454 */
617ba13b 5455static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5456{
287a8095
CH		 5457 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5458 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5459}
5460
daf647d2
TT		 5461static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5462{
5463 struct ext4_inode_info *ei = EXT4_I(inode);
5464
5465 /* The first argument of lockdep_set_subclass has to be
5466 * *exactly* the same as the argument to init_rwsem() --- in
5467 * this case, in init_once() --- or lockdep gets unhappy
5468 * because the name of the lock is set using the
5469 * stringification of the argument to init_rwsem().
5470 */
5471 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5472 lockdep_set_subclass(&ei->i_data_sem, subclass);
5473}
5474
ac27a0ec
DK		 5475/*
5476 * Standard function to be called on quota_on
5477 */
617ba13b 5478static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 5479 const struct path *path)
ac27a0ec
DK		 5480{
5481 int err;
ac27a0ec
DK		 5482
5483 if (!test_opt(sb, QUOTA))
5484 return -EINVAL;
0623543b 5485
ac27a0ec 5486 /* Quotafile not on the same filesystem? */
d8c9584e 5487 if (path->dentry->d_sb != sb)
ac27a0ec 5488 return -EXDEV;
0623543b
JK		 5489 /* Journaling quota? */
5490 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5491 /* Quotafile not in fs root? */
f00c9e44 5492 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5493 ext4_msg(sb, KERN_WARNING,
5494 "Quota file not on filesystem root. "
5495 "Journaled quota will not work");
91389240
JK		 5496 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5497 } else {
5498 /*
5499 * Clear the flag just in case mount options changed since
5500 * last time.
5501 */
5502 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
2b2d6d01 5503 }
0623543b
JK		 5504
5505 /*
5506 * When we journal data on quota file, we have to flush journal to see
5507 * all updates to the file when we bypass pagecache...
5508 */
0390131b 5509 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5510 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5511 /*
5512 * We don't need to lock updates but journal_flush() could
5513 * otherwise be livelocked...
5514 */
5515 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5516 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5517 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5518 if (err)
7ffe1ea8 5519 return err;
0623543b 5520 }
957153fc 5521
daf647d2
TT		 5522 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5523 err = dquot_quota_on(sb, type, format_id, path);
957153fc 5524 if (err) {
daf647d2
TT		 5525 lockdep_set_quota_inode(path->dentry->d_inode,
5526 I_DATA_SEM_NORMAL);
957153fc
JK		 5527 } else {
5528 struct inode *inode = d_inode(path->dentry);
5529 handle_t *handle;
5530
61a92987
JK		 5531 /*
5532 * Set inode flags to prevent userspace from messing with quota
5533 * files. If this fails, we return success anyway since quotas
5534 * are already enabled and this is not a hard failure.
5535 */
957153fc
JK		 5536 inode_lock(inode);
5537 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5538 if (IS_ERR(handle))
5539 goto unlock_inode;
5540 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5541 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5542 S_NOATIME | S_IMMUTABLE);
5543 ext4_mark_inode_dirty(handle, inode);
5544 ext4_journal_stop(handle);
5545 unlock_inode:
5546 inode_unlock(inode);
5547 }
daf647d2 5548 return err;
ac27a0ec
DK		 5549}
5550
7c319d32
AK		 5551static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5552 unsigned int flags)
5553{
5554 int err;
5555 struct inode *qf_inode;
a2d4a646 5556 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5557 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5558 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5559 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5560 };
5561
e2b911c5 5562 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5563
5564 if (!qf_inums[type])
5565 return -EPERM;
5566
5567 qf_inode = ext4_iget(sb, qf_inums[type]);
5568 if (IS_ERR(qf_inode)) {
5569 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5570 return PTR_ERR(qf_inode);
5571 }
5572
bcb13850
JK		 5573 /* Don't account quota for quota files to avoid recursion */
5574 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5575 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7c319d32
AK		 5576 err = dquot_enable(qf_inode, type, format_id, flags);
5577 iput(qf_inode);
daf647d2
TT		 5578 if (err)
5579 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7c319d32
AK		 5580
5581 return err;
5582}
5583
5584/* Enable usage tracking for all quota types. */
5585static int ext4_enable_quotas(struct super_block *sb)
5586{
5587 int type, err = 0;
a2d4a646 5588 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5589 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5590 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5591 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 5592 };
49da9392
JK		 5593 bool quota_mopt[EXT4_MAXQUOTAS] = {
5594 test_opt(sb, USRQUOTA),
5595 test_opt(sb, GRPQUOTA),
5596 test_opt(sb, PRJQUOTA),
5597 };
7c319d32 5598
91389240 5599 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
a2d4a646 5600 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5601 if (qf_inums[type]) {
5602 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 5603 DQUOT_USAGE_ENABLED |
5604 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32 5605 if (err) {
95f1fda4 5606 for (type--; type >= 0; type--)
5607 dquot_quota_off(sb, type);
5608
7c319d32 5609 ext4_warning(sb,
72ba7450
TT		 5610 "Failed to enable quota tracking "
5611 "(type=%d, err=%d). Please run "
5612 "e2fsck to fix.", type, err);
7c319d32
AK		 5613 return err;
5614 }
5615 }
5616 }
5617 return 0;
5618}
5619
ca0e05e4
DM		 5620static int ext4_quota_off(struct super_block *sb, int type)
5621{
21f97697
JK		 5622 struct inode *inode = sb_dqopt(sb)->files[type];
5623 handle_t *handle;
957153fc 5624 int err;
21f97697 5625
87009d86
DM		 5626 /* Force all delayed allocation blocks to be allocated.
5627 * Caller already holds s_umount sem */
5628 if (test_opt(sb, DELALLOC))
ca0e05e4 5629 sync_filesystem(sb);
ca0e05e4 5630
957153fc 5631 if (!inode || !igrab(inode))
0b268590
AG		 5632 goto out;
5633
957153fc 5634 err = dquot_quota_off(sb, type);
964edf66 5635 if (err || ext4_has_feature_quota(sb))
957153fc
JK		 5636 goto out_put;
5637
5638 inode_lock(inode);
61a92987
JK		 5639 /*
5640 * Update modification times of quota files when userspace can
5641 * start looking at them. If we fail, we return success anyway since
5642 * this is not a hard failure and quotas are already disabled.
5643 */
9924a92a 5644 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697 5645 if (IS_ERR(handle))
957153fc
JK		 5646 goto out_unlock;
5647 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5648 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
eeca7ea1 5649 inode->i_mtime = inode->i_ctime = current_time(inode);
21f97697
JK		 5650 ext4_mark_inode_dirty(handle, inode);
5651 ext4_journal_stop(handle);
957153fc
JK		 5652out_unlock:
5653 inode_unlock(inode);
5654out_put:
964edf66 5655 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 5656 iput(inode);
5657 return err;
21f97697 5658out:
ca0e05e4
DM		 5659 return dquot_quota_off(sb, type);
5660}
5661
ac27a0ec
DK		 5662/* Read data from quotafile - avoid pagecache and such because we cannot afford
5663 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5664 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5665 * we don't have to be afraid of races */
617ba13b 5666static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5667 size_t len, loff_t off)
5668{
5669 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5670 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5671 int offset = off & (sb->s_blocksize - 1);
5672 int tocopy;
5673 size_t toread;
5674 struct buffer_head *bh;
5675 loff_t i_size = i_size_read(inode);
5676
5677 if (off > i_size)
5678 return 0;
5679 if (off+len > i_size)
5680 len = i_size-off;
5681 toread = len;
5682 while (toread > 0) {
5683 tocopy = sb->s_blocksize - offset < toread ?
5684 sb->s_blocksize - offset : toread;
1c215028
TT		 5685 bh = ext4_bread(NULL, inode, blk, 0);
5686 if (IS_ERR(bh))
5687 return PTR_ERR(bh);
ac27a0ec
DK		 5688 if (!bh) /* A hole? */
5689 memset(data, 0, tocopy);
5690 else
5691 memcpy(data, bh->b_data+offset, tocopy);
5692 brelse(bh);
5693 offset = 0;
5694 toread -= tocopy;
5695 data += tocopy;
5696 blk++;
5697 }
5698 return len;
5699}
5700
5701/* Write to quotafile (we know the transaction is already started and has
5702 * enough credits) */
617ba13b 5703static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5704 const char *data, size_t len, loff_t off)
5705{
5706 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5707 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5708 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5709 int retries = 0;
ac27a0ec
DK		 5710 struct buffer_head *bh;
5711 handle_t *handle = journal_current_handle();
5712
0390131b 5713 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5714 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5715 " cancelled because transaction is not started",
9c3013e9
JK		 5716 (unsigned long long)off, (unsigned long long)len);
5717 return -EIO;
5718 }
67eeb568
DM		 5719 /*
5720 * Since we account only one data block in transaction credits,
5721 * then it is impossible to cross a block boundary.
5722 */
5723 if (sb->s_blocksize - offset < len) {
5724 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5725 " cancelled because not block aligned",
5726 (unsigned long long)off, (unsigned long long)len);
5727 return -EIO;
5728 }
5729
c5e298ae
TT		 5730 do {
5731 bh = ext4_bread(handle, inode, blk,
5732 EXT4_GET_BLOCKS_CREATE |
5733 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5734 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5735 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5736 if (IS_ERR(bh))
5737 return PTR_ERR(bh);
67eeb568
DM		 5738 if (!bh)
5739 goto out;
5d601255 5740 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5741 err = ext4_journal_get_write_access(handle, bh);
5742 if (err) {
5743 brelse(bh);
1c215028 5744 return err;
ac27a0ec 5745 }
67eeb568
DM		 5746 lock_buffer(bh);
5747 memcpy(bh->b_data+offset, data, len);
5748 flush_dcache_page(bh->b_page);
5749 unlock_buffer(bh);
62d2b5f2 5750 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5751 brelse(bh);
ac27a0ec 5752out:
67eeb568
DM		 5753 if (inode->i_size < off + len) {
5754 i_size_write(inode, off + len);
617ba13b 5755 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5756 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5757 }
67eeb568 5758 return len;
ac27a0ec
DK		 5759}
5760
8f0e8746
TT		 5761static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5762{
5763 const struct quota_format_ops *ops;
5764
5765 if (!sb_has_quota_loaded(sb, qid->type))
5766 return -ESRCH;
5767 ops = sb_dqopt(sb)->ops[qid->type];
5768 if (!ops || !ops->get_next_id)
5769 return -ENOSYS;
5770 return dquot_get_next_id(sb, qid);
5771}
ac27a0ec
DK		 5772#endif
5773
152a0836
AV		 5774static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5775 const char *dev_name, void *data)
ac27a0ec 5776{
152a0836 5777 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5778}
5779
c290ea01 5780#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5781static inline void register_as_ext2(void)
5782{
5783 int err = register_filesystem(&ext2_fs_type);
5784 if (err)
5785 printk(KERN_WARNING
5786 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5787}
5788
5789static inline void unregister_as_ext2(void)
5790{
5791 unregister_filesystem(&ext2_fs_type);
5792}
2035e776
TT		 5793
5794static inline int ext2_feature_set_ok(struct super_block *sb)
5795{
e2b911c5 5796 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776 5797 return 0;
bc98a42c 5798 if (sb_rdonly(sb))
2035e776 5799 return 1;
e2b911c5 5800 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5801 return 0;
5802 return 1;
5803}
24b58424
TT		 5804#else
5805static inline void register_as_ext2(void) { }
5806static inline void unregister_as_ext2(void) { }
2035e776 5807static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5808#endif
5809
24b58424
TT		 5810static inline void register_as_ext3(void)
5811{
5812 int err = register_filesystem(&ext3_fs_type);
5813 if (err)
5814 printk(KERN_WARNING
5815 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5816}
5817
5818static inline void unregister_as_ext3(void)
5819{
5820 unregister_filesystem(&ext3_fs_type);
5821}
2035e776
TT		 5822
5823static inline int ext3_feature_set_ok(struct super_block *sb)
5824{
e2b911c5 5825 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5826 return 0;
e2b911c5 5827 if (!ext4_has_feature_journal(sb))
2035e776 5828 return 0;
bc98a42c 5829 if (sb_rdonly(sb))
2035e776 5830 return 1;
e2b911c5 5831 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5832 return 0;
5833 return 1;
5834}
24b58424 5835
03010a33
TT		 5836static struct file_system_type ext4_fs_type = {
5837 .owner = THIS_MODULE,
5838 .name = "ext4",
152a0836 5839 .mount = ext4_mount,
03010a33
TT		 5840 .kill_sb = kill_block_super,
5841 .fs_flags = FS_REQUIRES_DEV,
5842};
7f78e035 5843MODULE_ALIAS_FS("ext4");
03010a33 5844
e9e3bcec
ES		 5845/* Shared across all ext4 file systems */
5846wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 5847
5dabfc78 5848static int __init ext4_init_fs(void)
ac27a0ec 5849{
e9e3bcec 5850 int i, err;
c9de560d 5851
e294a537 5852 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5853 ext4_li_info = NULL;
5854 mutex_init(&ext4_li_mtx);
5855
9a4c8019 5856 /* Build-time check for flags consistency */
12e9b892 5857 ext4_check_flag_values();
e9e3bcec 5858
e142d052 5859 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 5860 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 5861
51865fda 5862 err = ext4_init_es();
6fd058f7
TT		 5863 if (err)
5864 return err;
51865fda
ZL		 5865
5866 err = ext4_init_pageio();
5867 if (err)
b5799018 5868 goto out5;
51865fda 5869
5dabfc78 5870 err = ext4_init_system_zone();
bd2d0210 5871 if (err)
b5799018 5872 goto out4;
857ac889 5873
b5799018 5874 err = ext4_init_sysfs();
dd68314c 5875 if (err)
b5799018 5876 goto out3;
857ac889 5877
5dabfc78 5878 err = ext4_init_mballoc();
c9de560d
AT		 5879 if (err)
5880 goto out2;
ac27a0ec
DK		 5881 err = init_inodecache();
5882 if (err)
5883 goto out1;
24b58424 5884 register_as_ext3();
2035e776 5885 register_as_ext2();
03010a33 5886 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5887 if (err)
5888 goto out;
bfff6873 5889
ac27a0ec
DK		 5890 return 0;
5891out:
24b58424
TT		 5892 unregister_as_ext2();
5893 unregister_as_ext3();
ac27a0ec
DK		 5894 destroy_inodecache();
5895out1:
5dabfc78 5896 ext4_exit_mballoc();
9c191f70 5897out2:
b5799018
TT		 5898 ext4_exit_sysfs();
5899out3:
5dabfc78 5900 ext4_exit_system_zone();
b5799018 5901out4:
5dabfc78 5902 ext4_exit_pageio();
b5799018 5903out5:
51865fda
ZL		 5904 ext4_exit_es();
5905
ac27a0ec
DK		 5906 return err;
5907}
5908
5dabfc78 5909static void __exit ext4_exit_fs(void)
ac27a0ec 5910{
bfff6873 5911 ext4_destroy_lazyinit_thread();
24b58424
TT		 5912 unregister_as_ext2();
5913 unregister_as_ext3();
03010a33 5914 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5915 destroy_inodecache();
5dabfc78 5916 ext4_exit_mballoc();
b5799018 5917 ext4_exit_sysfs();
5dabfc78
TT		 5918 ext4_exit_system_zone();
5919 ext4_exit_pageio();
dd12ed14 5920 ext4_exit_es();
ac27a0ec
DK		 5921}
5922
5923MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5924MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5925MODULE_LICENSE("GPL");
7ef79ad5 5926MODULE_SOFTDEP("pre: crc32c");
5dabfc78
TT		 5927module_init(ext4_init_fs)
5928module_exit(ext4_exit_fs)
A mirror of Linus' kernel repository