]> 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
Linux 4.18-rc4
[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{
327eaf73
TT		 408 if (test_opt(sb, WARN_ON_ERROR))
409 WARN_ON_ONCE(1);
410
bc98a42c 411 if (sb_rdonly(sb))
ac27a0ec
DK		 412 return;
413
2b2d6d01 414 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 415 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 416
4ab2f15b 417 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 418 if (journal)
dab291af 419 jbd2_journal_abort(journal, -EIO);
ac27a0ec 420 }
2b2d6d01 421 if (test_opt(sb, ERRORS_RO)) {
b31e1552 422 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 423 /*
424 * Make sure updated value of ->s_mount_flags will be visible
425 * before ->s_flags update
426 */
427 smp_wmb();
1751e8a6 428 sb->s_flags |= SB_RDONLY;
ac27a0ec 429 }
4327ba52
DJ		 430 if (test_opt(sb, ERRORS_PANIC)) {
431 if (EXT4_SB(sb)->s_journal &&
432 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
433 return;
617ba13b 434 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec 435 sb->s_id);
4327ba52 436 }
ac27a0ec
DK		 437}
438
efbed4dc
TT		 439#define ext4_error_ratelimit(sb) \
440 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
441 "EXT4-fs error")
442
12062ddd 443void __ext4_error(struct super_block *sb, const char *function,
c398eda0 444 unsigned int line, const char *fmt, ...)
ac27a0ec 445{
0ff2ea7d 446 struct va_format vaf;
ac27a0ec
DK		 447 va_list args;
448
0db1ff22
TT		 449 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
450 return;
451
ccf0f32a 452 trace_ext4_error(sb, function, line);
efbed4dc
TT		 453 if (ext4_error_ratelimit(sb)) {
454 va_start(args, fmt);
455 vaf.fmt = fmt;
456 vaf.va = &args;
457 printk(KERN_CRIT
458 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
459 sb->s_id, function, line, current->comm, &vaf);
460 va_end(args);
461 }
f3fc0210 462 save_error_info(sb, function, line);
617ba13b 463 ext4_handle_error(sb);
ac27a0ec
DK		 464}
465
e7c96e8e
JP		 466void __ext4_error_inode(struct inode *inode, const char *function,
467 unsigned int line, ext4_fsblk_t block,
468 const char *fmt, ...)
273df556
FM		 469{
470 va_list args;
f7c21177 471 struct va_format vaf;
1c13d5c0 472 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 473
0db1ff22
TT		 474 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
475 return;
476
ccf0f32a 477 trace_ext4_error(inode->i_sb, function, line);
1c13d5c0
TT		 478 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
479 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 480 if (ext4_error_ratelimit(inode->i_sb)) {
481 va_start(args, fmt);
482 vaf.fmt = fmt;
483 vaf.va = &args;
484 if (block)
485 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
486 "inode #%lu: block %llu: comm %s: %pV\n",
487 inode->i_sb->s_id, function, line, inode->i_ino,
488 block, current->comm, &vaf);
489 else
490 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
491 "inode #%lu: comm %s: %pV\n",
492 inode->i_sb->s_id, function, line, inode->i_ino,
493 current->comm, &vaf);
494 va_end(args);
495 }
1c13d5c0 496 save_error_info(inode->i_sb, function, line);
273df556
FM		 497 ext4_handle_error(inode->i_sb);
498}
499
e7c96e8e
JP		 500void __ext4_error_file(struct file *file, const char *function,
501 unsigned int line, ext4_fsblk_t block,
502 const char *fmt, ...)
273df556
FM		 503{
504 va_list args;
f7c21177 505 struct va_format vaf;
1c13d5c0 506 struct ext4_super_block *es;
496ad9aa 507 struct inode *inode = file_inode(file);
273df556
FM		 508 char pathname[80], *path;
509
0db1ff22
TT		 510 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
511 return;
512
ccf0f32a 513 trace_ext4_error(inode->i_sb, function, line);
1c13d5c0
TT		 514 es = EXT4_SB(inode->i_sb)->s_es;
515 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 516 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 517 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 518 if (IS_ERR(path))
519 path = "(unknown)";
520 va_start(args, fmt);
521 vaf.fmt = fmt;
522 vaf.va = &args;
523 if (block)
524 printk(KERN_CRIT
525 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
526 "block %llu: comm %s: path %s: %pV\n",
527 inode->i_sb->s_id, function, line, inode->i_ino,
528 block, current->comm, path, &vaf);
529 else
530 printk(KERN_CRIT
531 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
532 "comm %s: path %s: %pV\n",
533 inode->i_sb->s_id, function, line, inode->i_ino,
534 current->comm, path, &vaf);
535 va_end(args);
536 }
1c13d5c0 537 save_error_info(inode->i_sb, function, line);
273df556
FM		 538 ext4_handle_error(inode->i_sb);
539}
540
722887dd
TT		 541const char *ext4_decode_error(struct super_block *sb, int errno,
542 char nbuf[16])
ac27a0ec
DK		 543{
544 char *errstr = NULL;
545
546 switch (errno) {
6a797d27
DW		 547 case -EFSCORRUPTED:
548 errstr = "Corrupt filesystem";
549 break;
550 case -EFSBADCRC:
551 errstr = "Filesystem failed CRC";
552 break;
ac27a0ec
DK		 553 case -EIO:
554 errstr = "IO failure";
555 break;
556 case -ENOMEM:
557 errstr = "Out of memory";
558 break;
559 case -EROFS:
78f1ddbb
TT		 560 if (!sb || (EXT4_SB(sb)->s_journal &&
561 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 562 errstr = "Journal has aborted";
563 else
564 errstr = "Readonly filesystem";
565 break;
566 default:
567 /* If the caller passed in an extra buffer for unknown
568 * errors, textualise them now. Else we just return
569 * NULL. */
570 if (nbuf) {
571 /* Check for truncated error codes... */
572 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
573 errstr = nbuf;
574 }
575 break;
576 }
577
578 return errstr;
579}
580
617ba13b 581/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 582 * automatically and invokes the appropriate error response. */
583
c398eda0
TT		 584void __ext4_std_error(struct super_block *sb, const char *function,
585 unsigned int line, int errno)
ac27a0ec
DK		 586{
587 char nbuf[16];
588 const char *errstr;
589
0db1ff22
TT		 590 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
591 return;
592
ac27a0ec
DK		 593 /* Special case: if the error is EROFS, and we're not already
594 * inside a transaction, then there's really no point in logging
595 * an error. */
bc98a42c 596 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
ac27a0ec
DK		 597 return;
598
efbed4dc
TT		 599 if (ext4_error_ratelimit(sb)) {
600 errstr = ext4_decode_error(sb, errno, nbuf);
601 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
602 sb->s_id, function, line, errstr);
603 }
ac27a0ec 604
efbed4dc 605 save_error_info(sb, function, line);
617ba13b 606 ext4_handle_error(sb);
ac27a0ec
DK		 607}
608
609/*
617ba13b 610 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 611 * abort function may be used to deal with unrecoverable failures such
612 * as journal IO errors or ENOMEM at a critical moment in log management.
613 *
614 * We unconditionally force the filesystem into an ABORT|READONLY state,
615 * unless the error response on the fs has been set to panic in which
616 * case we take the easy way out and panic immediately.
617 */
618
c67d859e 619void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 620 unsigned int line, const char *fmt, ...)
ac27a0ec 621{
651e1c3b 622 struct va_format vaf;
ac27a0ec
DK		 623 va_list args;
624
0db1ff22
TT		 625 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
626 return;
627
1c13d5c0 628 save_error_info(sb, function, line);
ac27a0ec 629 va_start(args, fmt);
651e1c3b
JP		 630 vaf.fmt = fmt;
631 vaf.va = &args;
632 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
633 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 634 va_end(args);
635
bc98a42c 636 if (sb_rdonly(sb) == 0) {
1c13d5c0 637 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 638 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 639 /*
640 * Make sure updated value of ->s_mount_flags will be visible
641 * before ->s_flags update
642 */
643 smp_wmb();
1751e8a6 644 sb->s_flags |= SB_RDONLY;
1c13d5c0
TT		 645 if (EXT4_SB(sb)->s_journal)
646 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
647 save_error_info(sb, function, line);
648 }
4327ba52
DJ		 649 if (test_opt(sb, ERRORS_PANIC)) {
650 if (EXT4_SB(sb)->s_journal &&
651 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
652 return;
617ba13b 653 panic("EXT4-fs panic from previous error\n");
4327ba52 654 }
ac27a0ec
DK		 655}
656
e7c96e8e
JP		 657void __ext4_msg(struct super_block *sb,
658 const char *prefix, const char *fmt, ...)
b31e1552 659{
0ff2ea7d 660 struct va_format vaf;
b31e1552
ES		 661 va_list args;
662
efbed4dc
TT		 663 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
664 return;
665
b31e1552 666 va_start(args, fmt);
0ff2ea7d
JP		 667 vaf.fmt = fmt;
668 vaf.va = &args;
669 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 670 va_end(args);
671}
672
b03a2f7e
AD		 673#define ext4_warning_ratelimit(sb) \
674 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
675 "EXT4-fs warning")
676
12062ddd 677void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 678 unsigned int line, const char *fmt, ...)
ac27a0ec 679{
0ff2ea7d 680 struct va_format vaf;
ac27a0ec
DK		 681 va_list args;
682
b03a2f7e 683 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 684 return;
685
ac27a0ec 686 va_start(args, fmt);
0ff2ea7d
JP		 687 vaf.fmt = fmt;
688 vaf.va = &args;
689 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
690 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 691 va_end(args);
692}
693
b03a2f7e
AD		 694void __ext4_warning_inode(const struct inode *inode, const char *function,
695 unsigned int line, const char *fmt, ...)
696{
697 struct va_format vaf;
698 va_list args;
699
700 if (!ext4_warning_ratelimit(inode->i_sb))
701 return;
702
703 va_start(args, fmt);
704 vaf.fmt = fmt;
705 vaf.va = &args;
706 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
707 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
708 function, line, inode->i_ino, current->comm, &vaf);
709 va_end(args);
710}
711
e29136f8
TT		 712void __ext4_grp_locked_error(const char *function, unsigned int line,
713 struct super_block *sb, ext4_group_t grp,
714 unsigned long ino, ext4_fsblk_t block,
715 const char *fmt, ...)
5d1b1b3f
AK		 716__releases(bitlock)
717__acquires(bitlock)
718{
0ff2ea7d 719 struct va_format vaf;
5d1b1b3f
AK		 720 va_list args;
721 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
722
0db1ff22
TT		 723 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
724 return;
725
ccf0f32a 726 trace_ext4_error(sb, function, line);
1c13d5c0
TT		 727 es->s_last_error_ino = cpu_to_le32(ino);
728 es->s_last_error_block = cpu_to_le64(block);
729 __save_error_info(sb, function, line);
0ff2ea7d 730
efbed4dc
TT		 731 if (ext4_error_ratelimit(sb)) {
732 va_start(args, fmt);
733 vaf.fmt = fmt;
734 vaf.va = &args;
735 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
736 sb->s_id, function, line, grp);
737 if (ino)
738 printk(KERN_CONT "inode %lu: ", ino);
739 if (block)
740 printk(KERN_CONT "block %llu:",
741 (unsigned long long) block);
742 printk(KERN_CONT "%pV\n", &vaf);
743 va_end(args);
744 }
5d1b1b3f 745
327eaf73
TT		 746 if (test_opt(sb, WARN_ON_ERROR))
747 WARN_ON_ONCE(1);
748
5d1b1b3f 749 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 750 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 751 return;
752 }
1c13d5c0 753
5d1b1b3f 754 ext4_unlock_group(sb, grp);
06f29cc8 755 ext4_commit_super(sb, 1);
5d1b1b3f
AK		 756 ext4_handle_error(sb);
757 /*
758 * We only get here in the ERRORS_RO case; relocking the group
759 * may be dangerous, but nothing bad will happen since the
760 * filesystem will have already been marked read/only and the
761 * journal has been aborted. We return 1 as a hint to callers
762 * who might what to use the return value from
25985edc 763 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 764 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
765 * aggressively from the ext4 function in question, with a
766 * more appropriate error code.
767 */
768 ext4_lock_group(sb, grp);
769 return;
770}
771
db79e6d1
WS		 772void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
773 ext4_group_t group,
774 unsigned int flags)
775{
776 struct ext4_sb_info *sbi = EXT4_SB(sb);
777 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
778 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
779
780 if ((flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) &&
781 !EXT4_MB_GRP_BBITMAP_CORRUPT(grp)) {
782 percpu_counter_sub(&sbi->s_freeclusters_counter,
783 grp->bb_free);
784 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
785 &grp->bb_state);
786 }
787
788 if ((flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) &&
789 !EXT4_MB_GRP_IBITMAP_CORRUPT(grp)) {
790 if (gdp) {
791 int count;
792
793 count = ext4_free_inodes_count(sb, gdp);
794 percpu_counter_sub(&sbi->s_freeinodes_counter,
795 count);
796 }
797 set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
798 &grp->bb_state);
799 }
800}
801
617ba13b 802void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 803{
617ba13b 804 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 805
617ba13b 806 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 807 return;
808
12062ddd 809 ext4_warning(sb,
ac27a0ec
DK		 810 "updating to rev %d because of new feature flag, "
811 "running e2fsck is recommended",
617ba13b 812 EXT4_DYNAMIC_REV);
ac27a0ec 813
617ba13b
MC		 814 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
815 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
816 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 817 /* leave es->s_feature_*compat flags alone */
818 /* es->s_uuid will be set by e2fsck if empty */
819
820 /*
821 * The rest of the superblock fields should be zero, and if not it
822 * means they are likely already in use, so leave them alone. We
823 * can leave it up to e2fsck to clean up any inconsistencies there.
824 */
825}
826
827/*
828 * Open the external journal device
829 */
b31e1552 830static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 831{
832 struct block_device *bdev;
833 char b[BDEVNAME_SIZE];
834
d4d77629 835 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 836 if (IS_ERR(bdev))
837 goto fail;
838 return bdev;
839
840fail:
b31e1552 841 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 842 __bdevname(dev, b), PTR_ERR(bdev));
843 return NULL;
844}
845
846/*
847 * Release the journal device
848 */
4385bab1 849static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 850{
4385bab1 851 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 852}
853
4385bab1 854static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 855{
856 struct block_device *bdev;
ac27a0ec
DK		 857 bdev = sbi->journal_bdev;
858 if (bdev) {
4385bab1 859 ext4_blkdev_put(bdev);
ac27a0ec
DK		 860 sbi->journal_bdev = NULL;
861 }
ac27a0ec
DK		 862}
863
864static inline struct inode *orphan_list_entry(struct list_head *l)
865{
617ba13b 866 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 867}
868
617ba13b 869static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 870{
871 struct list_head *l;
872
b31e1552
ES		 873 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
874 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 875
876 printk(KERN_ERR "sb_info orphan list:\n");
877 list_for_each(l, &sbi->s_orphan) {
878 struct inode *inode = orphan_list_entry(l);
879 printk(KERN_ERR " "
880 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
881 inode->i_sb->s_id, inode->i_ino, inode,
882 inode->i_mode, inode->i_nlink,
883 NEXT_ORPHAN(inode));
884 }
885}
886
957153fc
JK		 887#ifdef CONFIG_QUOTA
888static int ext4_quota_off(struct super_block *sb, int type);
889
890static inline void ext4_quota_off_umount(struct super_block *sb)
891{
892 int type;
893
964edf66
JK		 894 /* Use our quota_off function to clear inode flags etc. */
895 for (type = 0; type < EXT4_MAXQUOTAS; type++)
896 ext4_quota_off(sb, type);
957153fc
JK		 897}
898#else
899static inline void ext4_quota_off_umount(struct super_block *sb)
900{
901}
902#endif
903
2b2d6d01 904static void ext4_put_super(struct super_block *sb)
ac27a0ec 905{
617ba13b
MC		 906 struct ext4_sb_info *sbi = EXT4_SB(sb);
907 struct ext4_super_block *es = sbi->s_es;
97abd7d4 908 int aborted = 0;
ef2cabf7 909 int i, err;
ac27a0ec 910
857ac889 911 ext4_unregister_li_request(sb);
957153fc 912 ext4_quota_off_umount(sb);
e0ccfd95 913
2e8fa54e 914 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 915
0390131b 916 if (sbi->s_journal) {
97abd7d4 917 aborted = is_journal_aborted(sbi->s_journal);
0390131b
FM		 918 err = jbd2_journal_destroy(sbi->s_journal);
919 sbi->s_journal = NULL;
97abd7d4 920 if ((err < 0) && !aborted)
c67d859e 921 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 922 }
d4edac31 923
ebd173be 924 ext4_unregister_sysfs(sb);
d3922a77 925 ext4_es_unregister_shrinker(sbi);
9105bb14 926 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 927 ext4_release_system_zone(sb);
928 ext4_mb_release(sb);
929 ext4_ext_release(sb);
d4edac31 930
bc98a42c 931 if (!sb_rdonly(sb) && !aborted) {
e2b911c5 932 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 933 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 934 }
bc98a42c 935 if (!sb_rdonly(sb))
a8e25a83
AB		 936 ext4_commit_super(sb, 1);
937
ac27a0ec
DK		 938 for (i = 0; i < sbi->s_gdb_count; i++)
939 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 940 kvfree(sbi->s_group_desc);
941 kvfree(sbi->s_flex_groups);
57042651 942 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 943 percpu_counter_destroy(&sbi->s_freeinodes_counter);
944 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 945 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c8585c6f 946 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
ac27a0ec 947#ifdef CONFIG_QUOTA
a2d4a646 948 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 949 kfree(sbi->s_qf_names[i]);
950#endif
951
952 /* Debugging code just in case the in-memory inode orphan list
953 * isn't empty. The on-disk one can be non-empty if we've
954 * detected an error and taken the fs readonly, but the
955 * in-memory list had better be clean by this point. */
956 if (!list_empty(&sbi->s_orphan))
957 dump_orphan_list(sb, sbi);
958 J_ASSERT(list_empty(&sbi->s_orphan));
959
89d96a6f 960 sync_blockdev(sb->s_bdev);
f98393a6 961 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 962 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
963 /*
964 * Invalidate the journal device's buffers. We don't want them
965 * floating about in memory - the physical journal device may
966 * hotswapped, and it breaks the `ro-after' testing code.
967 */
968 sync_blockdev(sbi->journal_bdev);
f98393a6 969 invalidate_bdev(sbi->journal_bdev);
617ba13b 970 ext4_blkdev_remove(sbi);
ac27a0ec 971 }
dec214d0
TE		 972 if (sbi->s_ea_inode_cache) {
973 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
974 sbi->s_ea_inode_cache = NULL;
975 }
47387409
TE		 976 if (sbi->s_ea_block_cache) {
977 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
978 sbi->s_ea_block_cache = NULL;
9c191f70 979 }
c5e06d10
JL		 980 if (sbi->s_mmp_tsk)
981 kthread_stop(sbi->s_mmp_tsk);
9060dd2c 982 brelse(sbi->s_sbh);
ac27a0ec 983 sb->s_fs_info = NULL;
3197ebdb
TT		 984 /*
985 * Now that we are completely done shutting down the
986 * superblock, we need to actually destroy the kobject.
987 */
3197ebdb
TT		 988 kobject_put(&sbi->s_kobj);
989 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 990 if (sbi->s_chksum_driver)
991 crypto_free_shash(sbi->s_chksum_driver);
705895b6 992 kfree(sbi->s_blockgroup_lock);
5e405595 993 fs_put_dax(sbi->s_daxdev);
ac27a0ec 994 kfree(sbi);
ac27a0ec
DK		 995}
996
e18b890b 997static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 998
999/*
1000 * Called inside transaction, so use GFP_NOFS
1001 */
617ba13b 1002static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 1003{
617ba13b 1004 struct ext4_inode_info *ei;
ac27a0ec 1005
e6b4f8da 1006 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 1007 if (!ei)
1008 return NULL;
0b8e58a1 1009
ee73f9a5 1010 inode_set_iversion(&ei->vfs_inode, 1);
202ee5df 1011 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 1012 INIT_LIST_HEAD(&ei->i_prealloc_list);
1013 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 1014 ext4_es_init_tree(&ei->i_es_tree);
1015 rwlock_init(&ei->i_es_lock);
edaa53ca 1016 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 1017 ei->i_es_all_nr = 0;
edaa53ca 1018 ei->i_es_shk_nr = 0;
dd475925 1019 ei->i_es_shrink_lblk = 0;
d2a17637 1020 ei->i_reserved_data_blocks = 0;
9d0be502 1021 ei->i_da_metadata_calc_len = 0;
7e731bc9 1022 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 1023 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 1024#ifdef CONFIG_QUOTA
1025 ei->i_reserved_quota = 0;
96c7e0d9 1026 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 1027#endif
8aefcd55 1028 ei->jinode = NULL;
2e8fa54e 1029 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 1030 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 1031 ei->i_sync_tid = 0;
1032 ei->i_datasync_tid = 0;
e27f41e1 1033 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 1034 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
ac27a0ec
DK		 1035 return &ei->vfs_inode;
1036}
1037
7ff9c073
TT		 1038static int ext4_drop_inode(struct inode *inode)
1039{
1040 int drop = generic_drop_inode(inode);
1041
1042 trace_ext4_drop_inode(inode, drop);
1043 return drop;
1044}
1045
fa0d7e3d
NP		 1046static void ext4_i_callback(struct rcu_head *head)
1047{
1048 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 1049 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1050}
1051
617ba13b 1052static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1053{
9f7dd93d 1054 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1055 ext4_msg(inode->i_sb, KERN_ERR,
1056 "Inode %lu (%p): orphan list check failed!",
1057 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1058 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1059 EXT4_I(inode), sizeof(struct ext4_inode_info),
1060 true);
1061 dump_stack();
1062 }
fa0d7e3d 1063 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 1064}
1065
51cc5068 1066static void init_once(void *foo)
ac27a0ec 1067{
617ba13b 1068 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1069
a35afb83 1070 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1071 init_rwsem(&ei->xattr_sem);
0e855ac8 1072 init_rwsem(&ei->i_data_sem);
ea3d7209 1073 init_rwsem(&ei->i_mmap_sem);
a35afb83 1074 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1075}
1076
e67bc2b3 1077static int __init init_inodecache(void)
ac27a0ec 1078{
f8dd7c70
DW		 1079 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1080 sizeof(struct ext4_inode_info), 0,
1081 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1082 SLAB_ACCOUNT),
1083 offsetof(struct ext4_inode_info, i_data),
1084 sizeof_field(struct ext4_inode_info, i_data),
1085 init_once);
617ba13b 1086 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1087 return -ENOMEM;
1088 return 0;
1089}
1090
1091static void destroy_inodecache(void)
1092{
8c0a8537
KS		 1093 /*
1094 * Make sure all delayed rcu free inodes are flushed before we
1095 * destroy cache.
1096 */
1097 rcu_barrier();
617ba13b 1098 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1099}
1100
0930fcc1 1101void ext4_clear_inode(struct inode *inode)
ac27a0ec 1102{
0930fcc1 1103 invalidate_inode_buffers(inode);
dbd5768f 1104 clear_inode(inode);
9f754758 1105 dquot_drop(inode);
c2ea3fde 1106 ext4_discard_preallocations(inode);
51865fda 1107 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1108 if (EXT4_I(inode)->jinode) {
1109 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1110 EXT4_I(inode)->jinode);
1111 jbd2_free_inode(EXT4_I(inode)->jinode);
1112 EXT4_I(inode)->jinode = NULL;
1113 }
3d204e24 1114 fscrypt_put_encryption_info(inode);
ac27a0ec
DK		 1115}
1116
1b961ac0 1117static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1118 u64 ino, u32 generation)
ac27a0ec 1119{
ac27a0ec 1120 struct inode *inode;
ac27a0ec 1121
617ba13b 1122 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1123 return ERR_PTR(-ESTALE);
617ba13b 1124 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1125 return ERR_PTR(-ESTALE);
1126
1127 /* iget isn't really right if the inode is currently unallocated!!
1128 *
617ba13b 1129 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1130 * deleted, so we should be safe.
1131 *
1132 * Currently we don't know the generation for parent directory, so
1133 * a generation of 0 means "accept any"
1134 */
f4bb2981 1135 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1136 if (IS_ERR(inode))
1137 return ERR_CAST(inode);
1138 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1139 iput(inode);
1140 return ERR_PTR(-ESTALE);
1141 }
1b961ac0
CH		 1142
1143 return inode;
1144}
1145
1146static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1147 int fh_len, int fh_type)
1b961ac0
CH		 1148{
1149 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1150 ext4_nfs_get_inode);
1151}
1152
1153static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1154 int fh_len, int fh_type)
1b961ac0
CH		 1155{
1156 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1157 ext4_nfs_get_inode);
ac27a0ec
DK		 1158}
1159
c39a7f84
TO		 1160/*
1161 * Try to release metadata pages (indirect blocks, directories) which are
1162 * mapped via the block device. Since these pages could have journal heads
1163 * which would prevent try_to_free_buffers() from freeing them, we must use
1164 * jbd2 layer's try_to_free_buffers() function to release them.
1165 */
0b8e58a1
AD		 1166static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1167 gfp_t wait)
c39a7f84
TO		 1168{
1169 journal_t *journal = EXT4_SB(sb)->s_journal;
1170
1171 WARN_ON(PageChecked(page));
1172 if (!page_has_buffers(page))
1173 return 0;
1174 if (journal)
1175 return jbd2_journal_try_to_free_buffers(journal, page,
d0164adc 1176 wait & ~__GFP_DIRECT_RECLAIM);
c39a7f84
TO		 1177 return try_to_free_buffers(page);
1178}
1179
a7550b30
JK		 1180#ifdef CONFIG_EXT4_FS_ENCRYPTION
1181static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1182{
1183 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1184 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1185}
1186
a7550b30
JK		 1187static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1188 void *fs_data)
1189{
2f8f5e76 1190 handle_t *handle = fs_data;
c1a5d5f6 1191 int res, res2, credits, retries = 0;
2f8f5e76 1192
9ce0151a
EB		 1193 /*
1194 * Encrypting the root directory is not allowed because e2fsck expects
1195 * lost+found to exist and be unencrypted, and encrypting the root
1196 * directory would imply encrypting the lost+found directory as well as
1197 * the filename "lost+found" itself.
1198 */
1199 if (inode->i_ino == EXT4_ROOT_INO)
1200 return -EPERM;
2f8f5e76 1201
7d3e06a8
RZ		 1202 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1203 return -EINVAL;
1204
94840e3c
EB		 1205 res = ext4_convert_inline_data(inode);
1206 if (res)
1207 return res;
1208
2f8f5e76
EB		 1209 /*
1210 * If a journal handle was specified, then the encryption context is
1211 * being set on a new inode via inheritance and is part of a larger
1212 * transaction to create the inode. Otherwise the encryption context is
1213 * being set on an existing inode in its own transaction. Only in the
1214 * latter case should the "retry on ENOSPC" logic be used.
1215 */
a7550b30 1216
2f8f5e76
EB		 1217 if (handle) {
1218 res = ext4_xattr_set_handle(handle, inode,
1219 EXT4_XATTR_INDEX_ENCRYPTION,
1220 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1221 ctx, len, 0);
a7550b30
JK		 1222 if (!res) {
1223 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1224 ext4_clear_inode_state(inode,
1225 EXT4_STATE_MAY_INLINE_DATA);
a3caa24b 1226 /*
2ee6a576
EB		 1227 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1228 * S_DAX may be disabled
a3caa24b
JK		 1229 */
1230 ext4_set_inode_flags(inode);
a7550b30
JK		 1231 }
1232 return res;
1233 }
1234
b8cb5a54
TE		 1235 res = dquot_initialize(inode);
1236 if (res)
1237 return res;
2f8f5e76 1238retry:
af65207c
TE		 1239 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1240 &credits);
dec214d0
TE		 1241 if (res)
1242 return res;
1243
c1a5d5f6 1244 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
a7550b30
JK		 1245 if (IS_ERR(handle))
1246 return PTR_ERR(handle);
1247
2f8f5e76
EB		 1248 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1249 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1250 ctx, len, 0);
a7550b30
JK		 1251 if (!res) {
1252 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
2ee6a576
EB		 1253 /*
1254 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1255 * S_DAX may be disabled
1256 */
a3caa24b 1257 ext4_set_inode_flags(inode);
a7550b30
JK		 1258 res = ext4_mark_inode_dirty(handle, inode);
1259 if (res)
1260 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1261 }
1262 res2 = ext4_journal_stop(handle);
2f8f5e76
EB		 1263
1264 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1265 goto retry;
a7550b30
JK		 1266 if (!res)
1267 res = res2;
1268 return res;
1269}
1270
c250b7dd 1271static bool ext4_dummy_context(struct inode *inode)
a7550b30
JK		 1272{
1273 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
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 1281 .empty_dir = ext4_empty_dir,
e12ee683 1282 .max_namelen = EXT4_NAME_LEN,
a7550b30 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,
327eaf73
TT		 1380 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1381 Opt_nowarn_on_error, Opt_mblk_io_submit,
670e9875 1382 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1449032b 1383 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1384 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1385 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1386 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1387 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
ac27a0ec
DK		 1388};
1389
a447c093 1390static const match_table_t tokens = {
ac27a0ec
DK		 1391 {Opt_bsd_df, "bsddf"},
1392 {Opt_minix_df, "minixdf"},
1393 {Opt_grpid, "grpid"},
1394 {Opt_grpid, "bsdgroups"},
1395 {Opt_nogrpid, "nogrpid"},
1396 {Opt_nogrpid, "sysvgroups"},
1397 {Opt_resgid, "resgid=%u"},
1398 {Opt_resuid, "resuid=%u"},
1399 {Opt_sb, "sb=%u"},
1400 {Opt_err_cont, "errors=continue"},
1401 {Opt_err_panic, "errors=panic"},
1402 {Opt_err_ro, "errors=remount-ro"},
1403 {Opt_nouid32, "nouid32"},
ac27a0ec 1404 {Opt_debug, "debug"},
72578c33
TT		 1405 {Opt_removed, "oldalloc"},
1406 {Opt_removed, "orlov"},
ac27a0ec
DK		 1407 {Opt_user_xattr, "user_xattr"},
1408 {Opt_nouser_xattr, "nouser_xattr"},
1409 {Opt_acl, "acl"},
1410 {Opt_noacl, "noacl"},
e3bb52ae 1411 {Opt_noload, "norecovery"},
5a916be1 1412 {Opt_noload, "noload"},
72578c33
TT		 1413 {Opt_removed, "nobh"},
1414 {Opt_removed, "bh"},
ac27a0ec 1415 {Opt_commit, "commit=%u"},
30773840
TT		 1416 {Opt_min_batch_time, "min_batch_time=%u"},
1417 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1418 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1419 {Opt_journal_path, "journal_path=%s"},
818d276c 1420 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1421 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1422 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1423 {Opt_abort, "abort"},
1424 {Opt_data_journal, "data=journal"},
1425 {Opt_data_ordered, "data=ordered"},
1426 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1427 {Opt_data_err_abort, "data_err=abort"},
1428 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1429 {Opt_offusrjquota, "usrjquota="},
1430 {Opt_usrjquota, "usrjquota=%s"},
1431 {Opt_offgrpjquota, "grpjquota="},
1432 {Opt_grpjquota, "grpjquota=%s"},
1433 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1434 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1435 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1436 {Opt_grpquota, "grpquota"},
1437 {Opt_noquota, "noquota"},
1438 {Opt_quota, "quota"},
1439 {Opt_usrquota, "usrquota"},
49da9392 1440 {Opt_prjquota, "prjquota"},
ac27a0ec 1441 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1442 {Opt_barrier, "barrier"},
1443 {Opt_nobarrier, "nobarrier"},
25ec56b5 1444 {Opt_i_version, "i_version"},
923ae0ff 1445 {Opt_dax, "dax"},
c9de560d 1446 {Opt_stripe, "stripe=%u"},
64769240 1447 {Opt_delalloc, "delalloc"},
327eaf73
TT		 1448 {Opt_warn_on_error, "warn_on_error"},
1449 {Opt_nowarn_on_error, "nowarn_on_error"},
a26f4992
TT		 1450 {Opt_lazytime, "lazytime"},
1451 {Opt_nolazytime, "nolazytime"},
670e9875 1452 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
dd919b98 1453 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1454 {Opt_removed, "mblk_io_submit"},
1455 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1456 {Opt_block_validity, "block_validity"},
1457 {Opt_noblock_validity, "noblock_validity"},
240799cd 1458 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1459 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1460 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1461 {Opt_auto_da_alloc, "auto_da_alloc"},
1462 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1463 {Opt_dioread_nolock, "dioread_nolock"},
1464 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1465 {Opt_discard, "discard"},
1466 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1467 {Opt_init_itable, "init_itable=%u"},
1468 {Opt_init_itable, "init_itable"},
1469 {Opt_noinit_itable, "noinit_itable"},
df981d03 1470 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1471 {Opt_test_dummy_encryption, "test_dummy_encryption"},
cdb7ee4c
TE		 1472 {Opt_nombcache, "nombcache"},
1473 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
c7198b9c
TT		 1474 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1475 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1476 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1477 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1478 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1479 {Opt_err, NULL},
ac27a0ec
DK		 1480};
1481
617ba13b 1482static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1483{
617ba13b 1484 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1485 char *options = (char *) *data;
1486
1487 if (!options || strncmp(options, "sb=", 3) != 0)
1488 return 1; /* Default location */
0b8e58a1 1489
ac27a0ec 1490 options += 3;
0b8e58a1 1491 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1492 sb_block = simple_strtoul(options, &options, 0);
1493 if (*options && *options != ',') {
4776004f 1494 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1495 (char *) *data);
1496 return 1;
1497 }
1498 if (*options == ',')
1499 options++;
1500 *data = (void *) options;
0b8e58a1 1501
ac27a0ec
DK		 1502 return sb_block;
1503}
1504
b3881f74 1505#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
d6006186
EB		 1506static const char deprecated_msg[] =
1507 "Mount option \"%s\" will be removed by %s\n"
437ca0fd 1508 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1509
56c50f11
DM		 1510#ifdef CONFIG_QUOTA
1511static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1512{
1513 struct ext4_sb_info *sbi = EXT4_SB(sb);
1514 char *qname;
03dafb5f 1515 int ret = -1;
56c50f11
DM		 1516
1517 if (sb_any_quota_loaded(sb) &&
1518 !sbi->s_qf_names[qtype]) {
1519 ext4_msg(sb, KERN_ERR,
1520 "Cannot change journaled "
1521 "quota options when quota turned on");
57f73c2c 1522 return -1;
56c50f11 1523 }
e2b911c5 1524 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1525 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1526 "ignored when QUOTA feature is enabled");
1527 return 1;
262b4662 1528 }
56c50f11
DM		 1529 qname = match_strdup(args);
1530 if (!qname) {
1531 ext4_msg(sb, KERN_ERR,
1532 "Not enough memory for storing quotafile name");
57f73c2c 1533 return -1;
56c50f11 1534 }
03dafb5f
CG		 1535 if (sbi->s_qf_names[qtype]) {
1536 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1537 ret = 1;
1538 else
1539 ext4_msg(sb, KERN_ERR,
1540 "%s quota file already specified",
1541 QTYPE2NAME(qtype));
1542 goto errout;
56c50f11 1543 }
03dafb5f 1544 if (strchr(qname, '/')) {
56c50f11
DM		 1545 ext4_msg(sb, KERN_ERR,
1546 "quotafile must be on filesystem root");
03dafb5f 1547 goto errout;
56c50f11 1548 }
03dafb5f 1549 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1550 set_opt(sb, QUOTA);
56c50f11 1551 return 1;
03dafb5f
CG		 1552errout:
1553 kfree(qname);
1554 return ret;
56c50f11
DM		 1555}
1556
1557static int clear_qf_name(struct super_block *sb, int qtype)
1558{
1559
1560 struct ext4_sb_info *sbi = EXT4_SB(sb);
1561
1562 if (sb_any_quota_loaded(sb) &&
1563 sbi->s_qf_names[qtype]) {
1564 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1565 " when quota turned on");
57f73c2c 1566 return -1;
56c50f11 1567 }
03dafb5f 1568 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1569 sbi->s_qf_names[qtype] = NULL;
1570 return 1;
1571}
1572#endif
1573
26092bf5
TT		 1574#define MOPT_SET 0x0001
1575#define MOPT_CLEAR 0x0002
1576#define MOPT_NOSUPPORT 0x0004
1577#define MOPT_EXPLICIT 0x0008
1578#define MOPT_CLEAR_ERR 0x0010
1579#define MOPT_GTE0 0x0020
ac27a0ec 1580#ifdef CONFIG_QUOTA
26092bf5
TT		 1581#define MOPT_Q 0
1582#define MOPT_QFMT 0x0040
1583#else
1584#define MOPT_Q MOPT_NOSUPPORT
1585#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1586#endif
26092bf5 1587#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1588#define MOPT_NO_EXT2 0x0100
1589#define MOPT_NO_EXT3 0x0200
1590#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1591#define MOPT_STRING 0x0400
26092bf5
TT		 1592
1593static const struct mount_opts {
1594 int token;
1595 int mount_opt;
1596 int flags;
1597} ext4_mount_opts[] = {
1598 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1599 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1600 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1601 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1602 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1603 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1604 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1605 MOPT_EXT4_ONLY | MOPT_SET},
1606 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1607 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1608 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1609 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1610 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1611 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1612 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1613 MOPT_EXT4_ONLY | MOPT_CLEAR},
327eaf73
TT		 1614 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1615 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
c6d3d56d
DW		 1616 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1617 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1618 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1619 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1620 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1621 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1622 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1623 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1624 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1625 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1626 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1627 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1628 MOPT_NO_EXT2},
8dc0aa8c 1629 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1630 MOPT_NO_EXT2},
26092bf5
TT		 1631 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1632 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1633 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1634 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1635 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1636 {Opt_commit, 0, MOPT_GTE0},
1637 {Opt_max_batch_time, 0, MOPT_GTE0},
1638 {Opt_min_batch_time, 0, MOPT_GTE0},
1639 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1640 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1641 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1642 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1643 {Opt_resuid, 0, MOPT_GTE0},
1644 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1645 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1646 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1647 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1648 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1649 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1650 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1651 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1652 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1653 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1654#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1655 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1656 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1657#else
26092bf5
TT		 1658 {Opt_acl, 0, MOPT_NOSUPPORT},
1659 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1660#endif
26092bf5
TT		 1661 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1662 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
670e9875 1663 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
26092bf5
TT		 1664 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1665 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1666 MOPT_SET | MOPT_Q},
1667 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1668 MOPT_SET | MOPT_Q},
49da9392
JK		 1669 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1670 MOPT_SET | MOPT_Q},
26092bf5 1671 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1672 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1673 MOPT_CLEAR | MOPT_Q},
26092bf5
TT		 1674 {Opt_usrjquota, 0, MOPT_Q},
1675 {Opt_grpjquota, 0, MOPT_Q},
1676 {Opt_offusrjquota, 0, MOPT_Q},
1677 {Opt_offgrpjquota, 0, MOPT_Q},
1678 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1679 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1680 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1681 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1682 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
cdb7ee4c 1683 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
26092bf5
TT		 1684 {Opt_err, 0, 0}
1685};
1686
1687static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1688 substring_t *args, unsigned long *journal_devnum,
1689 unsigned int *journal_ioprio, int is_remount)
1690{
1691 struct ext4_sb_info *sbi = EXT4_SB(sb);
1692 const struct mount_opts *m;
08cefc7a
EB		 1693 kuid_t uid;
1694 kgid_t gid;
26092bf5
TT		 1695 int arg = 0;
1696
57f73c2c
TT		 1697#ifdef CONFIG_QUOTA
1698 if (token == Opt_usrjquota)
1699 return set_qf_name(sb, USRQUOTA, &args[0]);
1700 else if (token == Opt_grpjquota)
1701 return set_qf_name(sb, GRPQUOTA, &args[0]);
1702 else if (token == Opt_offusrjquota)
1703 return clear_qf_name(sb, USRQUOTA);
1704 else if (token == Opt_offgrpjquota)
1705 return clear_qf_name(sb, GRPQUOTA);
1706#endif
26092bf5 1707 switch (token) {
f7048605
TT		 1708 case Opt_noacl:
1709 case Opt_nouser_xattr:
1710 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1711 break;
26092bf5
TT		 1712 case Opt_sb:
1713 return 1; /* handled by get_sb_block() */
1714 case Opt_removed:
5f3633e3 1715 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1716 return 1;
26092bf5
TT		 1717 case Opt_abort:
1718 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1719 return 1;
1720 case Opt_i_version:
357fdad0 1721 sb->s_flags |= SB_I_VERSION;
26092bf5 1722 return 1;
a26f4992 1723 case Opt_lazytime:
1751e8a6 1724 sb->s_flags |= SB_LAZYTIME;
a26f4992
TT		 1725 return 1;
1726 case Opt_nolazytime:
1751e8a6 1727 sb->s_flags &= ~SB_LAZYTIME;
a26f4992 1728 return 1;
26092bf5
TT		 1729 }
1730
5f3633e3
JK		 1731 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1732 if (token == m->token)
1733 break;
1734
1735 if (m->token == Opt_err) {
1736 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1737 "or missing value", opt);
1738 return -1;
1739 }
1740
8dc0aa8c
TT		 1741 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1742 ext4_msg(sb, KERN_ERR,
1743 "Mount option \"%s\" incompatible with ext2", opt);
1744 return -1;
1745 }
1746 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1747 ext4_msg(sb, KERN_ERR,
1748 "Mount option \"%s\" incompatible with ext3", opt);
1749 return -1;
1750 }
1751
ad4eec61 1752 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1753 return -1;
1754 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1755 return -1;
c93cf2d7
DM		 1756 if (m->flags & MOPT_EXPLICIT) {
1757 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1758 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1759 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1760 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1761 } else
1762 return -1;
1763 }
5f3633e3
JK		 1764 if (m->flags & MOPT_CLEAR_ERR)
1765 clear_opt(sb, ERRORS_MASK);
1766 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1767 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1768 "options when quota turned on");
1769 return -1;
1770 }
1771
1772 if (m->flags & MOPT_NOSUPPORT) {
1773 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1774 } else if (token == Opt_commit) {
1775 if (arg == 0)
1776 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1777 sbi->s_commit_interval = HZ * arg;
670e9875
TT		 1778 } else if (token == Opt_debug_want_extra_isize) {
1779 sbi->s_want_extra_isize = arg;
5f3633e3 1780 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1781 sbi->s_max_batch_time = arg;
1782 } else if (token == Opt_min_batch_time) {
1783 sbi->s_min_batch_time = arg;
1784 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1785 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1786 ext4_msg(sb, KERN_ERR,
1787 "EXT4-fs: inode_readahead_blks must be "
1788 "0 or a power of 2 smaller than 2^31");
26092bf5 1789 return -1;
5f3633e3
JK		 1790 }
1791 sbi->s_inode_readahead_blks = arg;
1792 } else if (token == Opt_init_itable) {
1793 set_opt(sb, INIT_INODE_TABLE);
1794 if (!args->from)
1795 arg = EXT4_DEF_LI_WAIT_MULT;
1796 sbi->s_li_wait_mult = arg;
1797 } else if (token == Opt_max_dir_size_kb) {
1798 sbi->s_max_dir_size_kb = arg;
1799 } else if (token == Opt_stripe) {
1800 sbi->s_stripe = arg;
1801 } else if (token == Opt_resuid) {
1802 uid = make_kuid(current_user_ns(), arg);
1803 if (!uid_valid(uid)) {
1804 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1805 return -1;
1806 }
5f3633e3
JK		 1807 sbi->s_resuid = uid;
1808 } else if (token == Opt_resgid) {
1809 gid = make_kgid(current_user_ns(), arg);
1810 if (!gid_valid(gid)) {
1811 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1812 return -1;
1813 }
1814 sbi->s_resgid = gid;
1815 } else if (token == Opt_journal_dev) {
1816 if (is_remount) {
1817 ext4_msg(sb, KERN_ERR,
1818 "Cannot specify journal on remount");
1819 return -1;
1820 }
1821 *journal_devnum = arg;
ad4eec61
ES		 1822 } else if (token == Opt_journal_path) {
1823 char *journal_path;
1824 struct inode *journal_inode;
1825 struct path path;
1826 int error;
1827
1828 if (is_remount) {
1829 ext4_msg(sb, KERN_ERR,
1830 "Cannot specify journal on remount");
1831 return -1;
1832 }
1833 journal_path = match_strdup(&args[0]);
1834 if (!journal_path) {
1835 ext4_msg(sb, KERN_ERR, "error: could not dup "
1836 "journal device string");
1837 return -1;
1838 }
1839
1840 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1841 if (error) {
1842 ext4_msg(sb, KERN_ERR, "error: could not find "
1843 "journal device path: error %d", error);
1844 kfree(journal_path);
1845 return -1;
1846 }
1847
2b0143b5 1848 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1849 if (!S_ISBLK(journal_inode->i_mode)) {
1850 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1851 "is not a block device", journal_path);
1852 path_put(&path);
1853 kfree(journal_path);
1854 return -1;
1855 }
1856
1857 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1858 path_put(&path);
1859 kfree(journal_path);
5f3633e3
JK		 1860 } else if (token == Opt_journal_ioprio) {
1861 if (arg > 7) {
1862 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1863 " (must be 0-7)");
1864 return -1;
1865 }
1866 *journal_ioprio =
1867 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1868 } else if (token == Opt_test_dummy_encryption) {
1869#ifdef CONFIG_EXT4_FS_ENCRYPTION
1870 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1871 ext4_msg(sb, KERN_WARNING,
1872 "Test dummy encryption mode enabled");
1873#else
1874 ext4_msg(sb, KERN_WARNING,
1875 "Test dummy encryption mount option ignored");
1876#endif
5f3633e3
JK		 1877 } else if (m->flags & MOPT_DATAJ) {
1878 if (is_remount) {
1879 if (!sbi->s_journal)
1880 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1881 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1882 ext4_msg(sb, KERN_ERR,
26092bf5 1883 "Cannot change data mode on remount");
26092bf5 1884 return -1;
ac27a0ec 1885 }
26092bf5 1886 } else {
5f3633e3
JK		 1887 clear_opt(sb, DATA_FLAGS);
1888 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1889 }
5f3633e3
JK		 1890#ifdef CONFIG_QUOTA
1891 } else if (m->flags & MOPT_QFMT) {
1892 if (sb_any_quota_loaded(sb) &&
1893 sbi->s_jquota_fmt != m->mount_opt) {
1894 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1895 "quota options when quota turned on");
1896 return -1;
1897 }
e2b911c5 1898 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1899 ext4_msg(sb, KERN_INFO,
1900 "Quota format mount options ignored "
262b4662 1901 "when QUOTA feature is enabled");
c325a67c 1902 return 1;
262b4662 1903 }
5f3633e3 1904 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 1905#endif
923ae0ff 1906 } else if (token == Opt_dax) {
ef83b6e8
DW		 1907#ifdef CONFIG_FS_DAX
1908 ext4_msg(sb, KERN_WARNING,
1909 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1910 sbi->s_mount_opt |= m->mount_opt;
1911#else
923ae0ff
RZ		 1912 ext4_msg(sb, KERN_INFO, "dax option not supported");
1913 return -1;
5f3633e3 1914#endif
7915a861
AN		 1915 } else if (token == Opt_data_err_abort) {
1916 sbi->s_mount_opt |= m->mount_opt;
1917 } else if (token == Opt_data_err_ignore) {
1918 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 1919 } else {
1920 if (!args->from)
1921 arg = 1;
1922 if (m->flags & MOPT_CLEAR)
1923 arg = !arg;
1924 else if (unlikely(!(m->flags & MOPT_SET))) {
1925 ext4_msg(sb, KERN_WARNING,
1926 "buggy handling of option %s", opt);
1927 WARN_ON(1);
1928 return -1;
1929 }
1930 if (arg != 0)
1931 sbi->s_mount_opt |= m->mount_opt;
1932 else
1933 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1934 }
5f3633e3 1935 return 1;
26092bf5
TT		 1936}
1937
1938static int parse_options(char *options, struct super_block *sb,
1939 unsigned long *journal_devnum,
1940 unsigned int *journal_ioprio,
1941 int is_remount)
1942{
1943 struct ext4_sb_info *sbi = EXT4_SB(sb);
1944 char *p;
1945 substring_t args[MAX_OPT_ARGS];
1946 int token;
1947
1948 if (!options)
1949 return 1;
1950
1951 while ((p = strsep(&options, ",")) != NULL) {
1952 if (!*p)
1953 continue;
1954 /*
1955 * Initialize args struct so we know whether arg was
1956 * found; some options take optional arguments.
1957 */
caecd0af 1958 args[0].to = args[0].from = NULL;
26092bf5
TT		 1959 token = match_token(p, tokens, args);
1960 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1961 journal_ioprio, is_remount) < 0)
1962 return 0;
ac27a0ec
DK		 1963 }
1964#ifdef CONFIG_QUOTA
49da9392
JK		 1965 /*
1966 * We do the test below only for project quotas. 'usrquota' and
1967 * 'grpquota' mount options are allowed even without quota feature
1968 * to support legacy quotas in quota files.
1969 */
1970 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1971 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1972 "Cannot enable project quota enforcement.");
1973 return 0;
1974 }
1975 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1976 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1977 clear_opt(sb, USRQUOTA);
ac27a0ec 1978
482a7425 1979 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1980 clear_opt(sb, GRPQUOTA);
ac27a0ec 1981
56c50f11 1982 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1983 ext4_msg(sb, KERN_ERR, "old and new quota "
1984 "format mixing");
ac27a0ec
DK		 1985 return 0;
1986 }
1987
1988 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1989 ext4_msg(sb, KERN_ERR, "journaled quota format "
1990 "not specified");
ac27a0ec
DK		 1991 return 0;
1992 }
ac27a0ec
DK		 1993 }
1994#endif
261cb20c
JK		 1995 if (test_opt(sb, DIOREAD_NOLOCK)) {
1996 int blocksize =
1997 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1998
09cbfeaf 1999 if (blocksize < PAGE_SIZE) {
261cb20c
JK		 2000 ext4_msg(sb, KERN_ERR, "can't mount with "
2001 "dioread_nolock if block size != PAGE_SIZE");
2002 return 0;
2003 }
2004 }
ac27a0ec
DK		 2005 return 1;
2006}
2007
2adf6da8
TT		 2008static inline void ext4_show_quota_options(struct seq_file *seq,
2009 struct super_block *sb)
2010{
2011#if defined(CONFIG_QUOTA)
2012 struct ext4_sb_info *sbi = EXT4_SB(sb);
2013
2014 if (sbi->s_jquota_fmt) {
2015 char *fmtname = "";
2016
2017 switch (sbi->s_jquota_fmt) {
2018 case QFMT_VFS_OLD:
2019 fmtname = "vfsold";
2020 break;
2021 case QFMT_VFS_V0:
2022 fmtname = "vfsv0";
2023 break;
2024 case QFMT_VFS_V1:
2025 fmtname = "vfsv1";
2026 break;
2027 }
2028 seq_printf(seq, ",jqfmt=%s", fmtname);
2029 }
2030
2031 if (sbi->s_qf_names[USRQUOTA])
a068acf2 2032 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 2033
2034 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 2035 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 2036#endif
2037}
2038
5a916be1
TT		 2039static const char *token2str(int token)
2040{
50df9fd5 2041 const struct match_token *t;
5a916be1
TT		 2042
2043 for (t = tokens; t->token != Opt_err; t++)
2044 if (t->token == token && !strchr(t->pattern, '='))
2045 break;
2046 return t->pattern;
2047}
2048
2adf6da8
TT		 2049/*
2050 * Show an option if
2051 * - it's set to a non-default value OR
2052 * - if the per-sb default is different from the global default
2053 */
66acdcf4
TT		 2054static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2055 int nodefs)
2adf6da8 2056{
2adf6da8
TT		 2057 struct ext4_sb_info *sbi = EXT4_SB(sb);
2058 struct ext4_super_block *es = sbi->s_es;
68afa7e0 2059 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
5a916be1 2060 const struct mount_opts *m;
66acdcf4 2061 char sep = nodefs ? '\n' : ',';
2adf6da8 2062
66acdcf4
TT		 2063#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2064#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2065
2066 if (sbi->s_sb_block != 1)
5a916be1
TT		 2067 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2068
2069 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2070 int want_set = m->flags & MOPT_SET;
2071 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2072 (m->flags & MOPT_CLEAR_ERR))
2073 continue;
68afa7e0 2074 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 2075 continue; /* skip if same as the default */
2076 if ((want_set &&
2077 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2078 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2079 continue; /* select Opt_noFoo vs Opt_Foo */
2080 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2081 }
5a916be1 2082
08cefc7a 2083 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2084 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2085 SEQ_OPTS_PRINT("resuid=%u",
2086 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2087 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2088 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2089 SEQ_OPTS_PRINT("resgid=%u",
2090 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2091 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2092 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2093 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2094 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2095 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2096 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2097 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2098 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2099 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2100 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2101 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2102 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2103 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
357fdad0 2104 if (sb->s_flags & SB_I_VERSION)
5a916be1 2105 SEQ_OPTS_PUTS("i_version");
66acdcf4 2106 if (nodefs || sbi->s_stripe)
5a916be1 2107 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
68afa7e0
TN		 2108 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2109 (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2110 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2111 SEQ_OPTS_PUTS("data=journal");
2112 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2113 SEQ_OPTS_PUTS("data=ordered");
2114 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2115 SEQ_OPTS_PUTS("data=writeback");
2116 }
66acdcf4
TT		 2117 if (nodefs ||
2118 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2119 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2120 sbi->s_inode_readahead_blks);
2adf6da8 2121
ceec0376 2122 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
66acdcf4 2123 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2124 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2125 if (nodefs || sbi->s_max_dir_size_kb)
2126 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2127 if (test_opt(sb, DATA_ERR_ABORT))
2128 SEQ_OPTS_PUTS("data_err=abort");
2adf6da8
TT		 2129
2130 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2131 return 0;
2132}
2133
66acdcf4
TT		 2134static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2135{
2136 return _ext4_show_options(seq, root->d_sb, 0);
2137}
2138
ebd173be 2139int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2140{
2141 struct super_block *sb = seq->private;
2142 int rc;
2143
bc98a42c 2144 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
66acdcf4
TT		 2145 rc = _ext4_show_options(seq, sb, 1);
2146 seq_puts(seq, "\n");
2147 return rc;
2148}
2149
617ba13b 2150static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2151 int read_only)
2152{
617ba13b 2153 struct ext4_sb_info *sbi = EXT4_SB(sb);
c89128a0 2154 int err = 0;
ac27a0ec 2155
617ba13b 2156 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2157 ext4_msg(sb, KERN_ERR, "revision level too high, "
2158 "forcing read-only mode");
c89128a0 2159 err = -EROFS;
ac27a0ec
DK		 2160 }
2161 if (read_only)
281b5995 2162 goto done;
617ba13b 2163 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2164 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2165 "running e2fsck is recommended");
c8b459f4 2166 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2167 ext4_msg(sb, KERN_WARNING,
2168 "warning: mounting fs with errors, "
2169 "running e2fsck is recommended");
ed3ce80a 2170 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2171 le16_to_cpu(es->s_mnt_count) >=
2172 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2173 ext4_msg(sb, KERN_WARNING,
2174 "warning: maximal mount count reached, "
2175 "running e2fsck is recommended");
ac27a0ec
DK		 2176 else if (le32_to_cpu(es->s_checkinterval) &&
2177 (le32_to_cpu(es->s_lastcheck) +
2178 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 2179 ext4_msg(sb, KERN_WARNING,
2180 "warning: checktime reached, "
2181 "running e2fsck is recommended");
0b8e58a1 2182 if (!sbi->s_journal)
0390131b 2183 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2184 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2185 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2186 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 2187 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 2188 ext4_update_dynamic_rev(sb);
0390131b 2189 if (sbi->s_journal)
e2b911c5 2190 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2191
c89128a0 2192 err = ext4_commit_super(sb, 1);
281b5995 2193done:
ac27a0ec 2194 if (test_opt(sb, DEBUG))
a9df9a49 2195 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2196 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2197 sb->s_blocksize,
2198 sbi->s_groups_count,
617ba13b
MC		 2199 EXT4_BLOCKS_PER_GROUP(sb),
2200 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2201 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2202
7abc52c2 2203 cleancache_init_fs(sb);
c89128a0 2204 return err;
ac27a0ec
DK		 2205}
2206
117fff10
TT		 2207int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2208{
2209 struct ext4_sb_info *sbi = EXT4_SB(sb);
2210 struct flex_groups *new_groups;
2211 int size;
2212
2213 if (!sbi->s_log_groups_per_flex)
2214 return 0;
2215
2216 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2217 if (size <= sbi->s_flex_groups_allocated)
2218 return 0;
2219
2220 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
a7c3e901 2221 new_groups = kvzalloc(size, GFP_KERNEL);
117fff10
TT		 2222 if (!new_groups) {
2223 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2224 size / (int) sizeof(struct flex_groups));
2225 return -ENOMEM;
2226 }
2227
2228 if (sbi->s_flex_groups) {
2229 memcpy(new_groups, sbi->s_flex_groups,
2230 (sbi->s_flex_groups_allocated *
2231 sizeof(struct flex_groups)));
b93b41d4 2232 kvfree(sbi->s_flex_groups);
117fff10
TT		 2233 }
2234 sbi->s_flex_groups = new_groups;
2235 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2236 return 0;
2237}
2238
772cb7c8
JS		 2239static int ext4_fill_flex_info(struct super_block *sb)
2240{
2241 struct ext4_sb_info *sbi = EXT4_SB(sb);
2242 struct ext4_group_desc *gdp = NULL;
772cb7c8 2243 ext4_group_t flex_group;
117fff10 2244 int i, err;
772cb7c8 2245
503358ae 2246 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2247 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2248 sbi->s_log_groups_per_flex = 0;
2249 return 1;
2250 }
2251
117fff10
TT		 2252 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2253 if (err)
9933fc0a 2254 goto failed;
772cb7c8 2255
772cb7c8 2256 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2257 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2258
2259 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2260 atomic_add(ext4_free_inodes_count(sb, gdp),
2261 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2262 atomic64_add(ext4_free_group_clusters(sb, gdp),
2263 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2264 atomic_add(ext4_used_dirs_count(sb, gdp),
2265 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2266 }
2267
2268 return 1;
2269failed:
2270 return 0;
2271}
2272
e2b911c5 2273static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2274 struct ext4_group_desc *gdp)
717d50e4 2275{
b47820ed 2276 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2277 __u16 crc = 0;
feb0ab32 2278 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2279 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2280
9aa5d32b 2281 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2282 /* Use new metadata_csum algorithm */
feb0ab32 2283 __u32 csum32;
b47820ed 2284 __u16 dummy_csum = 0;
feb0ab32 2285
feb0ab32
DW		 2286 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2287 sizeof(le_group));
b47820ed
DJ		 2288 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2289 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2290 sizeof(dummy_csum));
2291 offset += sizeof(dummy_csum);
2292 if (offset < sbi->s_desc_size)
2293 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2294 sbi->s_desc_size - offset);
feb0ab32
DW		 2295
2296 crc = csum32 & 0xFFFF;
2297 goto out;
717d50e4
AD		 2298 }
2299
feb0ab32 2300 /* old crc16 code */
e2b911c5 2301 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2302 return 0;
2303
feb0ab32
DW		 2304 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2305 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2306 crc = crc16(crc, (__u8 *)gdp, offset);
2307 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2308 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2309 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2310 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2311 crc = crc16(crc, (__u8 *)gdp + offset,
2312 le16_to_cpu(sbi->s_es->s_desc_size) -
2313 offset);
2314
2315out:
717d50e4
AD		 2316 return cpu_to_le16(crc);
2317}
2318
feb0ab32 2319int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2320 struct ext4_group_desc *gdp)
2321{
feb0ab32 2322 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2323 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2324 return 0;
2325
2326 return 1;
2327}
2328
feb0ab32
DW		 2329void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2330 struct ext4_group_desc *gdp)
2331{
2332 if (!ext4_has_group_desc_csum(sb))
2333 return;
e2b911c5 2334 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2335}
2336
ac27a0ec 2337/* Called at mount-time, super-block is locked */
bfff6873 2338static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2339 ext4_fsblk_t sb_block,
bfff6873 2340 ext4_group_t *first_not_zeroed)
ac27a0ec 2341{
617ba13b
MC		 2342 struct ext4_sb_info *sbi = EXT4_SB(sb);
2343 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2344 ext4_fsblk_t last_block;
77260807 2345 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0) + 1;
bd81d8ee
LV		 2346 ext4_fsblk_t block_bitmap;
2347 ext4_fsblk_t inode_bitmap;
2348 ext4_fsblk_t inode_table;
ce421581 2349 int flexbg_flag = 0;
bfff6873 2350 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2351
e2b911c5 2352 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2353 flexbg_flag = 1;
2354
af5bc92d 2355 ext4_debug("Checking group descriptors");
ac27a0ec 2356
197cd65a
AM		 2357 for (i = 0; i < sbi->s_groups_count; i++) {
2358 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2359
ce421581 2360 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2361 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2362 else
2363 last_block = first_block +
617ba13b 2364 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2365
bfff6873
LC		 2366 if ((grp == sbi->s_groups_count) &&
2367 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2368 grp = i;
2369
8fadc143 2370 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2371 if (block_bitmap == sb_block) {
2372 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2373 "Block bitmap for group %u overlaps "
2374 "superblock", i);
18db4b4e
TT		 2375 if (!sb_rdonly(sb))
2376 return 0;
829fa70d 2377 }
77260807
TT		 2378 if (block_bitmap >= sb_block + 1 &&
2379 block_bitmap <= last_bg_block) {
2380 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2381 "Block bitmap for group %u overlaps "
2382 "block group descriptors", i);
2383 if (!sb_rdonly(sb))
2384 return 0;
2385 }
2b2d6d01 2386 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2387 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2388 "Block bitmap for group %u not in group "
b31e1552 2389 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2390 return 0;
2391 }
8fadc143 2392 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2393 if (inode_bitmap == sb_block) {
2394 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2395 "Inode bitmap for group %u overlaps "
2396 "superblock", i);
18db4b4e
TT		 2397 if (!sb_rdonly(sb))
2398 return 0;
829fa70d 2399 }
77260807
TT		 2400 if (inode_bitmap >= sb_block + 1 &&
2401 inode_bitmap <= last_bg_block) {
2402 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2403 "Inode bitmap for group %u overlaps "
2404 "block group descriptors", i);
2405 if (!sb_rdonly(sb))
2406 return 0;
2407 }
2b2d6d01 2408 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2409 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2410 "Inode bitmap for group %u not in group "
b31e1552 2411 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2412 return 0;
2413 }
8fadc143 2414 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2415 if (inode_table == sb_block) {
2416 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2417 "Inode table for group %u overlaps "
2418 "superblock", i);
18db4b4e
TT		 2419 if (!sb_rdonly(sb))
2420 return 0;
829fa70d 2421 }
77260807
TT		 2422 if (inode_table >= sb_block + 1 &&
2423 inode_table <= last_bg_block) {
2424 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2425 "Inode table for group %u overlaps "
2426 "block group descriptors", i);
2427 if (!sb_rdonly(sb))
2428 return 0;
2429 }
bd81d8ee 2430 if (inode_table < first_block ||
2b2d6d01 2431 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2432 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2433 "Inode table for group %u not in group "
b31e1552 2434 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2435 return 0;
2436 }
955ce5f5 2437 ext4_lock_group(sb, i);
feb0ab32 2438 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2439 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2440 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2441 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2442 gdp)), le16_to_cpu(gdp->bg_checksum));
bc98a42c 2443 if (!sb_rdonly(sb)) {
955ce5f5 2444 ext4_unlock_group(sb, i);
8a266467 2445 return 0;
7ee1ec4c 2446 }
717d50e4 2447 }
955ce5f5 2448 ext4_unlock_group(sb, i);
ce421581
JS		 2449 if (!flexbg_flag)
2450 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2451 }
bfff6873
LC		 2452 if (NULL != first_not_zeroed)
2453 *first_not_zeroed = grp;
ac27a0ec
DK		 2454 return 1;
2455}
2456
617ba13b 2457/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2458 * the superblock) which were deleted from all directories, but held open by
2459 * a process at the time of a crash. We walk the list and try to delete these
2460 * inodes at recovery time (only with a read-write filesystem).
2461 *
2462 * In order to keep the orphan inode chain consistent during traversal (in
2463 * case of crash during recovery), we link each inode into the superblock
2464 * orphan list_head and handle it the same way as an inode deletion during
2465 * normal operation (which journals the operations for us).
2466 *
2467 * We only do an iget() and an iput() on each inode, which is very safe if we
2468 * accidentally point at an in-use or already deleted inode. The worst that
2469 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2470 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2471 * e2fsck was run on this filesystem, and it must have already done the orphan
2472 * inode cleanup for us, so we can safely abort without any further action.
2473 */
2b2d6d01
TT		 2474static void ext4_orphan_cleanup(struct super_block *sb,
2475 struct ext4_super_block *es)
ac27a0ec
DK		 2476{
2477 unsigned int s_flags = sb->s_flags;
2c98eb5e 2478 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec 2479#ifdef CONFIG_QUOTA
95f1fda4 2480 int quota_update = 0;
ac27a0ec
DK		 2481 int i;
2482#endif
2483 if (!es->s_last_orphan) {
2484 jbd_debug(4, "no orphan inodes to clean up\n");
2485 return;
2486 }
2487
a8f48a95 2488 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2489 ext4_msg(sb, KERN_ERR, "write access "
2490 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2491 return;
2492 }
2493
d39195c3
AG		 2494 /* Check if feature set would not allow a r/w mount */
2495 if (!ext4_feature_set_ok(sb, 0)) {
2496 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2497 "unknown ROCOMPAT features");
2498 return;
2499 }
2500
617ba13b 2501 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6 2502 /* don't clear list on RO mount w/ errors */
1751e8a6 2503 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
84474976 2504 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2505 "clearing orphan list.\n");
c25f9bc6
ES		 2506 es->s_last_orphan = 0;
2507 }
ac27a0ec
DK		 2508 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2509 return;
2510 }
2511
1751e8a6 2512 if (s_flags & SB_RDONLY) {
b31e1552 2513 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1751e8a6 2514 sb->s_flags &= ~SB_RDONLY;
ac27a0ec
DK		 2515 }
2516#ifdef CONFIG_QUOTA
2517 /* Needed for iput() to work correctly and not trash data */
1751e8a6 2518 sb->s_flags |= SB_ACTIVE;
95f1fda4 2519
2520 /*
2521 * Turn on quotas which were not enabled for read-only mounts if
2522 * filesystem has quota feature, so that they are updated correctly.
2523 */
1751e8a6 2524 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
95f1fda4 2525 int ret = ext4_enable_quotas(sb);
2526
2527 if (!ret)
2528 quota_update = 1;
2529 else
2530 ext4_msg(sb, KERN_ERR,
2531 "Cannot turn on quotas: error %d", ret);
2532 }
2533
2534 /* Turn on journaled quotas used for old sytle */
a2d4a646 2535 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2536 if (EXT4_SB(sb)->s_qf_names[i]) {
2537 int ret = ext4_quota_on_mount(sb, i);
95f1fda4 2538
2539 if (!ret)
2540 quota_update = 1;
2541 else
b31e1552
ES		 2542 ext4_msg(sb, KERN_ERR,
2543 "Cannot turn on journaled "
95f1fda4 2544 "quota: type %d: error %d", i, ret);
ac27a0ec
DK		 2545 }
2546 }
2547#endif
2548
2549 while (es->s_last_orphan) {
2550 struct inode *inode;
2551
c65d5c6c
VN		 2552 /*
2553 * We may have encountered an error during cleanup; if
2554 * so, skip the rest.
2555 */
2556 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2557 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2558 es->s_last_orphan = 0;
2559 break;
2560 }
2561
97bd42b9
JB		 2562 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2563 if (IS_ERR(inode)) {
ac27a0ec
DK		 2564 es->s_last_orphan = 0;
2565 break;
2566 }
2567
617ba13b 2568 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2569 dquot_initialize(inode);
ac27a0ec 2570 if (inode->i_nlink) {
566370a2
PT		 2571 if (test_opt(sb, DEBUG))
2572 ext4_msg(sb, KERN_DEBUG,
2573 "%s: truncating inode %lu to %lld bytes",
2574 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2575 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2576 inode->i_ino, inode->i_size);
5955102c 2577 inode_lock(inode);
55f252c9 2578 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 2579 ret = ext4_truncate(inode);
2580 if (ret)
2581 ext4_std_error(inode->i_sb, ret);
5955102c 2582 inode_unlock(inode);
ac27a0ec
DK		 2583 nr_truncates++;
2584 } else {
566370a2
PT		 2585 if (test_opt(sb, DEBUG))
2586 ext4_msg(sb, KERN_DEBUG,
2587 "%s: deleting unreferenced inode %lu",
2588 __func__, inode->i_ino);
ac27a0ec
DK		 2589 jbd_debug(2, "deleting unreferenced inode %lu\n",
2590 inode->i_ino);
2591 nr_orphans++;
2592 }
2593 iput(inode); /* The delete magic happens here! */
2594 }
2595
2b2d6d01 2596#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2597
2598 if (nr_orphans)
b31e1552
ES		 2599 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2600 PLURAL(nr_orphans));
ac27a0ec 2601 if (nr_truncates)
b31e1552
ES		 2602 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2603 PLURAL(nr_truncates));
ac27a0ec 2604#ifdef CONFIG_QUOTA
95f1fda4 2605 /* Turn off quotas if they were enabled for orphan cleanup */
2606 if (quota_update) {
2607 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2608 if (sb_dqopt(sb)->files[i])
2609 dquot_quota_off(sb, i);
2610 }
ac27a0ec
DK		 2611 }
2612#endif
1751e8a6 2613 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
ac27a0ec 2614}
0b8e58a1 2615
cd2291a4
ES		 2616/*
2617 * Maximal extent format file size.
2618 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2619 * extent format containers, within a sector_t, and within i_blocks
2620 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2621 * so that won't be a limiting factor.
2622 *
f17722f9
LC		 2623 * However there is other limiting factor. We do store extents in the form
2624 * of starting block and length, hence the resulting length of the extent
2625 * covering maximum file size must fit into on-disk format containers as
2626 * well. Given that length is always by 1 unit bigger than max unit (because
2627 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2628 *
cd2291a4
ES		 2629 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2630 */
f287a1a5 2631static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2632{
2633 loff_t res;
2634 loff_t upper_limit = MAX_LFS_FILESIZE;
2635
2636 /* small i_blocks in vfs inode? */
f287a1a5 2637 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2638 /*
90c699a9 2639 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2640 * i_block represent total blocks in 512 bytes
2641 * 32 == size of vfs inode i_blocks * 8
2642 */
2643 upper_limit = (1LL << 32) - 1;
2644
2645 /* total blocks in file system block size */
2646 upper_limit >>= (blkbits - 9);
2647 upper_limit <<= blkbits;
2648 }
2649
f17722f9
LC		 2650 /*
2651 * 32-bit extent-start container, ee_block. We lower the maxbytes
2652 * by one fs block, so ee_len can cover the extent of maximum file
2653 * size
2654 */
2655 res = (1LL << 32) - 1;
cd2291a4 2656 res <<= blkbits;
cd2291a4
ES		 2657
2658 /* Sanity check against vm- & vfs- imposed limits */
2659 if (res > upper_limit)
2660 res = upper_limit;
2661
2662 return res;
2663}
ac27a0ec 2664
ac27a0ec 2665/*
cd2291a4 2666 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2667 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2668 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2669 */
f287a1a5 2670static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2671{
617ba13b 2672 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2673 int meta_blocks;
2674 loff_t upper_limit;
0b8e58a1
AD		 2675 /* This is calculated to be the largest file size for a dense, block
2676 * mapped file such that the file's total number of 512-byte sectors,
2677 * including data and all indirect blocks, does not exceed (2^48 - 1).
2678 *
2679 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2680 * number of 512-byte sectors of the file.
0fc1b451
AK		 2681 */
2682
f287a1a5 2683 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2684 /*
90c699a9 2685 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2686 * the inode i_block field represents total file blocks in
2687 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2688 */
2689 upper_limit = (1LL << 32) - 1;
2690
2691 /* total blocks in file system block size */
2692 upper_limit >>= (bits - 9);
2693
2694 } else {
8180a562
AK		 2695 /*
2696 * We use 48 bit ext4_inode i_blocks
2697 * With EXT4_HUGE_FILE_FL set the i_blocks
2698 * represent total number of blocks in
2699 * file system block size
2700 */
0fc1b451
AK		 2701 upper_limit = (1LL << 48) - 1;
2702
0fc1b451
AK		 2703 }
2704
2705 /* indirect blocks */
2706 meta_blocks = 1;
2707 /* double indirect blocks */
2708 meta_blocks += 1 + (1LL << (bits-2));
2709 /* tripple indirect blocks */
2710 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2711
2712 upper_limit -= meta_blocks;
2713 upper_limit <<= bits;
ac27a0ec
DK		 2714
2715 res += 1LL << (bits-2);
2716 res += 1LL << (2*(bits-2));
2717 res += 1LL << (3*(bits-2));
2718 res <<= bits;
2719 if (res > upper_limit)
2720 res = upper_limit;
0fc1b451
AK		 2721
2722 if (res > MAX_LFS_FILESIZE)
2723 res = MAX_LFS_FILESIZE;
2724
ac27a0ec
DK		 2725 return res;
2726}
2727
617ba13b 2728static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2729 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2730{
617ba13b 2731 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2732 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2733 int has_super = 0;
2734
2735 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2736
e2b911c5 2737 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2738 return logical_sb_block + nr + 1;
ac27a0ec 2739 bg = sbi->s_desc_per_block * nr;
617ba13b 2740 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2741 has_super = 1;
0b8e58a1 2742
bd63f6b0
DW		 2743 /*
2744 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2745 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2746 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2747 * compensate.
2748 */
2749 if (sb->s_blocksize == 1024 && nr == 0 &&
49598e04 2750 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
bd63f6b0
DW		 2751 has_super++;
2752
617ba13b 2753 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2754}
2755
c9de560d
AT		 2756/**
2757 * ext4_get_stripe_size: Get the stripe size.
2758 * @sbi: In memory super block info
2759 *
2760 * If we have specified it via mount option, then
2761 * use the mount option value. If the value specified at mount time is
2762 * greater than the blocks per group use the super block value.
2763 * If the super block value is greater than blocks per group return 0.
2764 * Allocator needs it be less than blocks per group.
2765 *
2766 */
2767static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2768{
2769 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2770 unsigned long stripe_width =
2771 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2772 int ret;
c9de560d
AT		 2773
2774 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 2775 ret = sbi->s_stripe;
5469d7c3 2776 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 2777 ret = stripe_width;
5469d7c3 2778 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 2779 ret = stride;
2780 else
2781 ret = 0;
c9de560d 2782
3eb08658
DE		 2783 /*
2784 * If the stripe width is 1, this makes no sense and
2785 * we set it to 0 to turn off stripe handling code.
2786 */
2787 if (ret <= 1)
2788 ret = 0;
c9de560d 2789
3eb08658 2790 return ret;
c9de560d 2791}
ac27a0ec 2792
a13fb1a4
ES		 2793/*
2794 * Check whether this filesystem can be mounted based on
2795 * the features present and the RDONLY/RDWR mount requested.
2796 * Returns 1 if this filesystem can be mounted as requested,
2797 * 0 if it cannot be.
2798 */
2799static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2800{
e2b911c5 2801 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2802 ext4_msg(sb, KERN_ERR,
2803 "Couldn't mount because of "
2804 "unsupported optional features (%x)",
2805 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2806 ~EXT4_FEATURE_INCOMPAT_SUPP));
2807 return 0;
2808 }
2809
2810 if (readonly)
2811 return 1;
2812
e2b911c5 2813 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b 2814 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
1751e8a6 2815 sb->s_flags |= SB_RDONLY;
2cb5cc8b
DW		 2816 return 1;
2817 }
2818
a13fb1a4 2819 /* Check that feature set is OK for a read-write mount */
e2b911c5 2820 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2821 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2822 "unsupported optional features (%x)",
2823 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2824 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2825 return 0;
2826 }
2827 /*
2828 * Large file size enabled file system can only be mounted
2829 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2830 */
e2b911c5 2831 if (ext4_has_feature_huge_file(sb)) {
a13fb1a4
ES		 2832 if (sizeof(blkcnt_t) < sizeof(u64)) {
2833 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2834 "cannot be mounted RDWR without "
2835 "CONFIG_LBDAF");
2836 return 0;
2837 }
2838 }
e2b911c5 2839 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2840 ext4_msg(sb, KERN_ERR,
2841 "Can't support bigalloc feature without "
2842 "extents feature\n");
2843 return 0;
2844 }
7c319d32
AK		 2845
2846#ifndef CONFIG_QUOTA
e2b911c5 2847 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2848 ext4_msg(sb, KERN_ERR,
2849 "Filesystem with quota feature cannot be mounted RDWR "
2850 "without CONFIG_QUOTA");
2851 return 0;
2852 }
689c958c
LX		 2853 if (ext4_has_feature_project(sb) && !readonly) {
2854 ext4_msg(sb, KERN_ERR,
2855 "Filesystem with project quota feature cannot be mounted RDWR "
2856 "without CONFIG_QUOTA");
2857 return 0;
2858 }
7c319d32 2859#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2860 return 1;
2861}
2862
66e61a9e
TT		 2863/*
2864 * This function is called once a day if we have errors logged
2865 * on the file system
2866 */
235699a8 2867static void print_daily_error_info(struct timer_list *t)
66e61a9e 2868{
235699a8
KC		 2869 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
2870 struct super_block *sb = sbi->s_sb;
2871 struct ext4_super_block *es = sbi->s_es;
66e61a9e
TT		 2872
2873 if (es->s_error_count)
ae0f78de
TT		 2874 /* fsck newer than v1.41.13 is needed to clean this condition. */
2875 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2876 le32_to_cpu(es->s_error_count));
2877 if (es->s_first_error_time) {
ae0f78de 2878 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2879 sb->s_id, le32_to_cpu(es->s_first_error_time),
2880 (int) sizeof(es->s_first_error_func),
2881 es->s_first_error_func,
2882 le32_to_cpu(es->s_first_error_line));
2883 if (es->s_first_error_ino)
651e1c3b 2884 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2885 le32_to_cpu(es->s_first_error_ino));
2886 if (es->s_first_error_block)
651e1c3b 2887 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2888 le64_to_cpu(es->s_first_error_block));
651e1c3b 2889 printk(KERN_CONT "\n");
66e61a9e
TT		 2890 }
2891 if (es->s_last_error_time) {
ae0f78de 2892 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2893 sb->s_id, le32_to_cpu(es->s_last_error_time),
2894 (int) sizeof(es->s_last_error_func),
2895 es->s_last_error_func,
2896 le32_to_cpu(es->s_last_error_line));
2897 if (es->s_last_error_ino)
651e1c3b 2898 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2899 le32_to_cpu(es->s_last_error_ino));
2900 if (es->s_last_error_block)
651e1c3b 2901 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2902 le64_to_cpu(es->s_last_error_block));
651e1c3b 2903 printk(KERN_CONT "\n");
66e61a9e
TT		 2904 }
2905 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2906}
2907
bfff6873
LC		 2908/* Find next suitable group and run ext4_init_inode_table */
2909static int ext4_run_li_request(struct ext4_li_request *elr)
2910{
2911 struct ext4_group_desc *gdp = NULL;
2912 ext4_group_t group, ngroups;
2913 struct super_block *sb;
2914 unsigned long timeout = 0;
2915 int ret = 0;
2916
2917 sb = elr->lr_super;
2918 ngroups = EXT4_SB(sb)->s_groups_count;
2919
2920 for (group = elr->lr_next_group; group < ngroups; group++) {
2921 gdp = ext4_get_group_desc(sb, group, NULL);
2922 if (!gdp) {
2923 ret = 1;
2924 break;
2925 }
2926
2927 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2928 break;
2929 }
2930
7f511862 2931 if (group >= ngroups)
bfff6873
LC		 2932 ret = 1;
2933
2934 if (!ret) {
2935 timeout = jiffies;
2936 ret = ext4_init_inode_table(sb, group,
2937 elr->lr_timeout ? 0 : 1);
2938 if (elr->lr_timeout == 0) {
51ce6511
LC		 2939 timeout = (jiffies - timeout) *
2940 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2941 elr->lr_timeout = timeout;
2942 }
2943 elr->lr_next_sched = jiffies + elr->lr_timeout;
2944 elr->lr_next_group = group + 1;
2945 }
bfff6873
LC		 2946 return ret;
2947}
2948
2949/*
2950 * Remove lr_request from the list_request and free the
4ed5c033 2951 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2952 */
2953static void ext4_remove_li_request(struct ext4_li_request *elr)
2954{
2955 struct ext4_sb_info *sbi;
2956
2957 if (!elr)
2958 return;
2959
2960 sbi = elr->lr_sbi;
2961
2962 list_del(&elr->lr_request);
2963 sbi->s_li_request = NULL;
2964 kfree(elr);
2965}
2966
2967static void ext4_unregister_li_request(struct super_block *sb)
2968{
1bb933fb
LC		 2969 mutex_lock(&ext4_li_mtx);
2970 if (!ext4_li_info) {
2971 mutex_unlock(&ext4_li_mtx);
bfff6873 2972 return;
1bb933fb 2973 }
bfff6873
LC		 2974
2975 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2976 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2977 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2978 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2979}
2980
8f1f7453
ES		 2981static struct task_struct *ext4_lazyinit_task;
2982
bfff6873
LC		 2983/*
2984 * This is the function where ext4lazyinit thread lives. It walks
2985 * through the request list searching for next scheduled filesystem.
2986 * When such a fs is found, run the lazy initialization request
2987 * (ext4_rn_li_request) and keep track of the time spend in this
2988 * function. Based on that time we compute next schedule time of
2989 * the request. When walking through the list is complete, compute
2990 * next waking time and put itself into sleep.
2991 */
2992static int ext4_lazyinit_thread(void *arg)
2993{
2994 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2995 struct list_head *pos, *n;
2996 struct ext4_li_request *elr;
4ed5c033 2997 unsigned long next_wakeup, cur;
bfff6873
LC		 2998
2999 BUG_ON(NULL == eli);
3000
bfff6873
LC		 3001cont_thread:
3002 while (true) {
3003 next_wakeup = MAX_JIFFY_OFFSET;
3004
3005 mutex_lock(&eli->li_list_mtx);
3006 if (list_empty(&eli->li_request_list)) {
3007 mutex_unlock(&eli->li_list_mtx);
3008 goto exit_thread;
3009 }
bfff6873 3010 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 3011 int err = 0;
3012 int progress = 0;
bfff6873
LC		 3013 elr = list_entry(pos, struct ext4_li_request,
3014 lr_request);
3015
e22834f0
DM		 3016 if (time_before(jiffies, elr->lr_next_sched)) {
3017 if (time_before(elr->lr_next_sched, next_wakeup))
3018 next_wakeup = elr->lr_next_sched;
3019 continue;
3020 }
3021 if (down_read_trylock(&elr->lr_super->s_umount)) {
3022 if (sb_start_write_trylock(elr->lr_super)) {
3023 progress = 1;
3024 /*
3025 * We hold sb->s_umount, sb can not
3026 * be removed from the list, it is
3027 * now safe to drop li_list_mtx
3028 */
3029 mutex_unlock(&eli->li_list_mtx);
3030 err = ext4_run_li_request(elr);
3031 sb_end_write(elr->lr_super);
3032 mutex_lock(&eli->li_list_mtx);
3033 n = pos->next;
b2c78cd0 3034 }
e22834f0
DM		 3035 up_read((&elr->lr_super->s_umount));
3036 }
3037 /* error, remove the lazy_init job */
3038 if (err) {
3039 ext4_remove_li_request(elr);
3040 continue;
3041 }
3042 if (!progress) {
3043 elr->lr_next_sched = jiffies +
3044 (prandom_u32()
3045 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 3046 }
bfff6873
LC		 3047 if (time_before(elr->lr_next_sched, next_wakeup))
3048 next_wakeup = elr->lr_next_sched;
3049 }
3050 mutex_unlock(&eli->li_list_mtx);
3051
a0acae0e 3052 try_to_freeze();
bfff6873 3053
4ed5c033
LC		 3054 cur = jiffies;
3055 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3056 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3057 cond_resched();
3058 continue;
3059 }
3060
4ed5c033
LC		 3061 schedule_timeout_interruptible(next_wakeup - cur);
3062
8f1f7453
ES		 3063 if (kthread_should_stop()) {
3064 ext4_clear_request_list();
3065 goto exit_thread;
3066 }
bfff6873
LC		 3067 }
3068
3069exit_thread:
3070 /*
3071 * It looks like the request list is empty, but we need
3072 * to check it under the li_list_mtx lock, to prevent any
3073 * additions into it, and of course we should lock ext4_li_mtx
3074 * to atomically free the list and ext4_li_info, because at
3075 * this point another ext4 filesystem could be registering
3076 * new one.
3077 */
3078 mutex_lock(&ext4_li_mtx);
3079 mutex_lock(&eli->li_list_mtx);
3080 if (!list_empty(&eli->li_request_list)) {
3081 mutex_unlock(&eli->li_list_mtx);
3082 mutex_unlock(&ext4_li_mtx);
3083 goto cont_thread;
3084 }
3085 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3086 kfree(ext4_li_info);
3087 ext4_li_info = NULL;
3088 mutex_unlock(&ext4_li_mtx);
3089
3090 return 0;
3091}
3092
3093static void ext4_clear_request_list(void)
3094{
3095 struct list_head *pos, *n;
3096 struct ext4_li_request *elr;
3097
3098 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3099 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3100 elr = list_entry(pos, struct ext4_li_request,
3101 lr_request);
3102 ext4_remove_li_request(elr);
3103 }
3104 mutex_unlock(&ext4_li_info->li_list_mtx);
3105}
3106
3107static int ext4_run_lazyinit_thread(void)
3108{
8f1f7453
ES		 3109 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3110 ext4_li_info, "ext4lazyinit");
3111 if (IS_ERR(ext4_lazyinit_task)) {
3112 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3113 ext4_clear_request_list();
bfff6873
LC		 3114 kfree(ext4_li_info);
3115 ext4_li_info = NULL;
92b97816 3116 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3117 "initialization thread\n",
3118 err);
3119 return err;
3120 }
3121 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3122 return 0;
3123}
3124
3125/*
3126 * Check whether it make sense to run itable init. thread or not.
3127 * If there is at least one uninitialized inode table, return
3128 * corresponding group number, else the loop goes through all
3129 * groups and return total number of groups.
3130 */
3131static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3132{
3133 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3134 struct ext4_group_desc *gdp = NULL;
3135
8844618d
TT		 3136 if (!ext4_has_group_desc_csum(sb))
3137 return ngroups;
3138
bfff6873
LC		 3139 for (group = 0; group < ngroups; group++) {
3140 gdp = ext4_get_group_desc(sb, group, NULL);
3141 if (!gdp)
3142 continue;
3143
8844618d
TT		 3144 if (gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))
3145 continue;
3146 if (group != 0)
bfff6873 3147 break;
8844618d
TT		 3148 ext4_error(sb, "Inode table for bg 0 marked as "
3149 "needing zeroing");
3150 if (sb_rdonly(sb))
3151 return ngroups;
bfff6873
LC		 3152 }
3153
3154 return group;
3155}
3156
3157static int ext4_li_info_new(void)
3158{
3159 struct ext4_lazy_init *eli = NULL;
3160
3161 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3162 if (!eli)
3163 return -ENOMEM;
3164
bfff6873
LC		 3165 INIT_LIST_HEAD(&eli->li_request_list);
3166 mutex_init(&eli->li_list_mtx);
3167
bfff6873
LC		 3168 eli->li_state |= EXT4_LAZYINIT_QUIT;
3169
3170 ext4_li_info = eli;
3171
3172 return 0;
3173}
3174
3175static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3176 ext4_group_t start)
3177{
3178 struct ext4_sb_info *sbi = EXT4_SB(sb);
3179 struct ext4_li_request *elr;
bfff6873
LC		 3180
3181 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3182 if (!elr)
3183 return NULL;
3184
3185 elr->lr_super = sb;
3186 elr->lr_sbi = sbi;
3187 elr->lr_next_group = start;
3188
3189 /*
3190 * Randomize first schedule time of the request to
3191 * spread the inode table initialization requests
3192 * better.
3193 */
dd1f723b
TT		 3194 elr->lr_next_sched = jiffies + (prandom_u32() %
3195 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3196 return elr;
3197}
3198
7f511862
TT		 3199int ext4_register_li_request(struct super_block *sb,
3200 ext4_group_t first_not_zeroed)
bfff6873
LC		 3201{
3202 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3203 struct ext4_li_request *elr = NULL;
49598e04 3204 ext4_group_t ngroups = sbi->s_groups_count;
6c5a6cb9 3205 int ret = 0;
bfff6873 3206
7f511862 3207 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3208 if (sbi->s_li_request != NULL) {
3209 /*
3210 * Reset timeout so it can be computed again, because
3211 * s_li_wait_mult might have changed.
3212 */
3213 sbi->s_li_request->lr_timeout = 0;
7f511862 3214 goto out;
51ce6511 3215 }
bfff6873 3216
bc98a42c 3217 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
55ff3840 3218 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3219 goto out;
bfff6873
LC		 3220
3221 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3222 if (!elr) {
3223 ret = -ENOMEM;
3224 goto out;
3225 }
bfff6873
LC		 3226
3227 if (NULL == ext4_li_info) {
3228 ret = ext4_li_info_new();
3229 if (ret)
3230 goto out;
3231 }
3232
3233 mutex_lock(&ext4_li_info->li_list_mtx);
3234 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3235 mutex_unlock(&ext4_li_info->li_list_mtx);
3236
3237 sbi->s_li_request = elr;
46e4690b
TM		 3238 /*
3239 * set elr to NULL here since it has been inserted to
3240 * the request_list and the removal and free of it is
3241 * handled by ext4_clear_request_list from now on.
3242 */
3243 elr = NULL;
bfff6873
LC		 3244
3245 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3246 ret = ext4_run_lazyinit_thread();
3247 if (ret)
3248 goto out;
3249 }
bfff6873 3250out:
beed5ecb
NK		 3251 mutex_unlock(&ext4_li_mtx);
3252 if (ret)
bfff6873 3253 kfree(elr);
bfff6873
LC		 3254 return ret;
3255}
3256
3257/*
3258 * We do not need to lock anything since this is called on
3259 * module unload.
3260 */
3261static void ext4_destroy_lazyinit_thread(void)
3262{
3263 /*
3264 * If thread exited earlier
3265 * there's nothing to be done.
3266 */
8f1f7453 3267 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3268 return;
3269
8f1f7453 3270 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3271}
3272
25ed6e8a
DW		 3273static int set_journal_csum_feature_set(struct super_block *sb)
3274{
3275 int ret = 1;
3276 int compat, incompat;
3277 struct ext4_sb_info *sbi = EXT4_SB(sb);
3278
9aa5d32b 3279 if (ext4_has_metadata_csum(sb)) {
db9ee220 3280 /* journal checksum v3 */
25ed6e8a 3281 compat = 0;
db9ee220 3282 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3283 } else {
3284 /* journal checksum v1 */
3285 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3286 incompat = 0;
3287 }
3288
feb8c6d3
DW		 3289 jbd2_journal_clear_features(sbi->s_journal,
3290 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3291 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3292 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3293 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3294 ret = jbd2_journal_set_features(sbi->s_journal,
3295 compat, 0,
3296 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3297 incompat);
3298 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3299 ret = jbd2_journal_set_features(sbi->s_journal,
3300 compat, 0,
3301 incompat);
3302 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3303 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3304 } else {
feb8c6d3
DW		 3305 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3306 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3307 }
3308
3309 return ret;
3310}
3311
952fc18e
TT		 3312/*
3313 * Note: calculating the overhead so we can be compatible with
3314 * historical BSD practice is quite difficult in the face of
3315 * clusters/bigalloc. This is because multiple metadata blocks from
3316 * different block group can end up in the same allocation cluster.
3317 * Calculating the exact overhead in the face of clustered allocation
3318 * requires either O(all block bitmaps) in memory or O(number of block
3319 * groups**2) in time. We will still calculate the superblock for
3320 * older file systems --- and if we come across with a bigalloc file
3321 * system with zero in s_overhead_clusters the estimate will be close to
3322 * correct especially for very large cluster sizes --- but for newer
3323 * file systems, it's better to calculate this figure once at mkfs
3324 * time, and store it in the superblock. If the superblock value is
3325 * present (even for non-bigalloc file systems), we will use it.
3326 */
3327static int count_overhead(struct super_block *sb, ext4_group_t grp,
3328 char *buf)
3329{
3330 struct ext4_sb_info *sbi = EXT4_SB(sb);
3331 struct ext4_group_desc *gdp;
3332 ext4_fsblk_t first_block, last_block, b;
3333 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3334 int s, j, count = 0;
3335
e2b911c5 3336 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3337 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3338 sbi->s_itb_per_group + 2);
3339
952fc18e
TT		 3340 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3341 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3342 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3343 for (i = 0; i < ngroups; i++) {
3344 gdp = ext4_get_group_desc(sb, i, NULL);
3345 b = ext4_block_bitmap(sb, gdp);
3346 if (b >= first_block && b <= last_block) {
3347 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3348 count++;
3349 }
3350 b = ext4_inode_bitmap(sb, gdp);
3351 if (b >= first_block && b <= last_block) {
3352 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3353 count++;
3354 }
3355 b = ext4_inode_table(sb, gdp);
3356 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3357 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3358 int c = EXT4_B2C(sbi, b - first_block);
3359 ext4_set_bit(c, buf);
3360 count++;
3361 }
3362 if (i != grp)
3363 continue;
3364 s = 0;
3365 if (ext4_bg_has_super(sb, grp)) {
3366 ext4_set_bit(s++, buf);
3367 count++;
3368 }
c48ae41b
TT		 3369 j = ext4_bg_num_gdb(sb, grp);
3370 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3371 ext4_error(sb, "Invalid number of block group "
3372 "descriptor blocks: %d", j);
3373 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3374 }
c48ae41b
TT		 3375 count += j;
3376 for (; j > 0; j--)
3377 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3378 }
3379 if (!count)
3380 return 0;
3381 return EXT4_CLUSTERS_PER_GROUP(sb) -
3382 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3383}
3384
3385/*
3386 * Compute the overhead and stash it in sbi->s_overhead
3387 */
3388int ext4_calculate_overhead(struct super_block *sb)
3389{
3390 struct ext4_sb_info *sbi = EXT4_SB(sb);
3391 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3392 struct inode *j_inode;
3393 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3394 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3395 ext4_fsblk_t overhead = 0;
4fdb5543 3396 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3397
952fc18e
TT		 3398 if (!buf)
3399 return -ENOMEM;
3400
3401 /*
3402 * Compute the overhead (FS structures). This is constant
3403 * for a given filesystem unless the number of block groups
3404 * changes so we cache the previous value until it does.
3405 */
3406
3407 /*
3408 * All of the blocks before first_data_block are overhead
3409 */
3410 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3411
3412 /*
3413 * Add the overhead found in each block group
3414 */
3415 for (i = 0; i < ngroups; i++) {
3416 int blks;
3417
3418 blks = count_overhead(sb, i, buf);
3419 overhead += blks;
3420 if (blks)
3421 memset(buf, 0, PAGE_SIZE);
3422 cond_resched();
3423 }
3c816ded
EW		 3424
3425 /*
3426 * Add the internal journal blocks whether the journal has been
3427 * loaded or not
3428 */
b003b524 3429 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3430 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3c816ded
EW		 3431 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3432 j_inode = ext4_get_journal_inode(sb, j_inum);
3433 if (j_inode) {
3434 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3435 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3436 iput(j_inode);
3437 } else {
3438 ext4_msg(sb, KERN_ERR, "can't get journal size");
3439 }
3440 }
952fc18e
TT		 3441 sbi->s_overhead = overhead;
3442 smp_wmb();
3443 free_page((unsigned long) buf);
3444 return 0;
3445}
3446
b5799018 3447static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3448{
3449 ext4_fsblk_t resv_clusters;
b5799018 3450 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3451
30fac0f7
JK		 3452 /*
3453 * There's no need to reserve anything when we aren't using extents.
3454 * The space estimates are exact, there are no unwritten extents,
3455 * hole punching doesn't need new metadata... This is needed especially
3456 * to keep ext2/3 backward compatibility.
3457 */
e2b911c5 3458 if (!ext4_has_feature_extents(sb))
b5799018 3459 return;
27dd4385
LC		 3460 /*
3461 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3462 * This should cover the situations where we can not afford to run
3463 * out of space like for example punch hole, or converting
556615dc 3464 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3465 * allocation would require 1, or 2 blocks, higher numbers are
3466 * very rare.
3467 */
b5799018
TT		 3468 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3469 sbi->s_cluster_bits);
27dd4385
LC		 3470
3471 do_div(resv_clusters, 50);
3472 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3473
b5799018 3474 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3475}
3476
2b2d6d01 3477static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3478{
5e405595 3479 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
d4c402d9 3480 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3481 struct buffer_head *bh;
617ba13b 3482 struct ext4_super_block *es = NULL;
5aee0f8a 3483 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
617ba13b
MC		 3484 ext4_fsblk_t block;
3485 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3486 ext4_fsblk_t logical_sb_block;
ac27a0ec 3487 unsigned long offset = 0;
ac27a0ec
DK		 3488 unsigned long journal_devnum = 0;
3489 unsigned long def_mount_opts;
3490 struct inode *root;
0390131b 3491 const char *descr;
dcc7dae3 3492 int ret = -ENOMEM;
281b5995 3493 int blocksize, clustersize;
4ec11028
TT		 3494 unsigned int db_count;
3495 unsigned int i;
281b5995 3496 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3497 __u64 blocks_count;
07aa2ea1 3498 int err = 0;
b3881f74 3499 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3500 ext4_group_t first_not_zeroed;
ac27a0ec 3501
5aee0f8a
TT		 3502 if ((data && !orig_data) || !sbi)
3503 goto out_free_base;
705895b6 3504
aed9eb1b 3505 sbi->s_daxdev = dax_dev;
705895b6
PE		 3506 sbi->s_blockgroup_lock =
3507 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 3508 if (!sbi->s_blockgroup_lock)
3509 goto out_free_base;
3510
ac27a0ec 3511 sb->s_fs_info = sbi;
2c0544b2 3512 sbi->s_sb = sb;
240799cd 3513 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3514 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3515 if (sb->s_bdev->bd_part)
3516 sbi->s_sectors_written_start =
3517 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3518
9f6200bb 3519 /* Cleanup superblock name */
ec3904dc 3520 strreplace(sb->s_id, '/', '!');
9f6200bb 3521
07aa2ea1 3522 /* -EINVAL is default */
dcc7dae3 3523 ret = -EINVAL;
617ba13b 3524 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3525 if (!blocksize) {
b31e1552 3526 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3527 goto out_fail;
3528 }
3529
3530 /*
617ba13b 3531 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3532 * block sizes. We need to calculate the offset from buffer start.
3533 */
617ba13b 3534 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3535 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3536 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3537 } else {
70bbb3e0 3538 logical_sb_block = sb_block;
ac27a0ec
DK		 3539 }
3540
a8ac900b 3541 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3542 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3543 goto out_fail;
3544 }
3545 /*
3546 * Note: s_es must be initialized as soon as possible because
617ba13b 3547 * some ext4 macro-instructions depend on its value
ac27a0ec 3548 */
2716b802 3549 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3550 sbi->s_es = es;
3551 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3552 if (sb->s_magic != EXT4_SUPER_MAGIC)
3553 goto cantfind_ext4;
afc32f7e 3554 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3555
feb0ab32 3556 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3557 if (ext4_has_feature_metadata_csum(sb) &&
3558 ext4_has_feature_gdt_csum(sb))
363307e6 3559 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3560 "redundant flags; please run fsck.");
3561
d25425f8
DW		 3562 /* Check for a known checksum algorithm */
3563 if (!ext4_verify_csum_type(sb, es)) {
3564 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3565 "unknown checksum algorithm.");
3566 silent = 1;
3567 goto cantfind_ext4;
3568 }
3569
0441984a 3570 /* Load the checksum driver */
a45403b5
TT		 3571 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3572 if (IS_ERR(sbi->s_chksum_driver)) {
3573 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3574 ret = PTR_ERR(sbi->s_chksum_driver);
3575 sbi->s_chksum_driver = NULL;
3576 goto failed_mount;
0441984a
DW		 3577 }
3578
a9c47317
DW		 3579 /* Check superblock checksum */
3580 if (!ext4_superblock_csum_verify(sb, es)) {
3581 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3582 "invalid superblock checksum. Run e2fsck?");
3583 silent = 1;
6a797d27 3584 ret = -EFSBADCRC;
a9c47317
DW		 3585 goto cantfind_ext4;
3586 }
3587
3588 /* Precompute checksum seed for all metadata */
e2b911c5 3589 if (ext4_has_feature_csum_seed(sb))
8c81bd8f 3590 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
dec214d0 3591 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
a9c47317
DW		 3592 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3593 sizeof(es->s_uuid));
3594
ac27a0ec
DK		 3595 /* Set defaults before we parse the mount options */
3596 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3597 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3598 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3599 set_opt(sb, DEBUG);
87f26807 3600 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3601 set_opt(sb, GRPID);
617ba13b 3602 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3603 set_opt(sb, NO_UID32);
ea663336 3604 /* xattr user namespace & acls are now defaulted on */
ea663336 3605 set_opt(sb, XATTR_USER);
03010a33 3606#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3607 set_opt(sb, POSIX_ACL);
2e7842b8 3608#endif
98c1a759
DW		 3609 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3610 if (ext4_has_metadata_csum(sb))
3611 set_opt(sb, JOURNAL_CHECKSUM);
3612
617ba13b 3613 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3614 set_opt(sb, JOURNAL_DATA);
617ba13b 3615 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3616 set_opt(sb, ORDERED_DATA);
617ba13b 3617 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3618 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3619
3620 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3621 set_opt(sb, ERRORS_PANIC);
bb4f397a 3622 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3623 set_opt(sb, ERRORS_CONT);
bb4f397a 3624 else
fd8c37ec 3625 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3626 /* block_validity enabled by default; disable with noblock_validity */
3627 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3628 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3629 set_opt(sb, DISCARD);
ac27a0ec 3630
08cefc7a
EB		 3631 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3632 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3633 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3634 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3635 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3636
8b67f04a 3637 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3638 set_opt(sb, BARRIER);
ac27a0ec 3639
dd919b98
AK		 3640 /*
3641 * enable delayed allocation by default
3642 * Use -o nodelalloc to turn it off
3643 */
bc0b75f7 3644 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3645 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3646 set_opt(sb, DELALLOC);
dd919b98 3647
51ce6511
LC		 3648 /*
3649 * set default s_li_wait_mult for lazyinit, for the case there is
3650 * no mount option specified.
3651 */
3652 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3653
5aee0f8a
TT		 3654 if (sbi->s_es->s_mount_opts[0]) {
3655 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3656 sizeof(sbi->s_es->s_mount_opts),
3657 GFP_KERNEL);
3658 if (!s_mount_opts)
3659 goto failed_mount;
3660 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3661 &journal_ioprio, 0)) {
3662 ext4_msg(sb, KERN_WARNING,
3663 "failed to parse options in superblock: %s",
3664 s_mount_opts);
3665 }
3666 kfree(s_mount_opts);
8b67f04a 3667 }
5a916be1 3668 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3669 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3670 &journal_ioprio, 0))
ac27a0ec
DK		 3671 goto failed_mount;
3672
56889787
TT		 3673 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3674 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3675 "with data=journal disables delayed "
3676 "allocation and O_DIRECT support!\n");
3677 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3678 ext4_msg(sb, KERN_ERR, "can't mount with "
3679 "both data=journal and delalloc");
3680 goto failed_mount;
3681 }
3682 if (test_opt(sb, DIOREAD_NOLOCK)) {
3683 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3684 "both data=journal and dioread_nolock");
56889787
TT		 3685 goto failed_mount;
3686 }
923ae0ff
RZ		 3687 if (test_opt(sb, DAX)) {
3688 ext4_msg(sb, KERN_ERR, "can't mount with "
3689 "both data=journal and dax");
3690 goto failed_mount;
3691 }
73b92a2a
SK		 3692 if (ext4_has_feature_encrypt(sb)) {
3693 ext4_msg(sb, KERN_WARNING,
3694 "encrypted files will use data=ordered "
3695 "instead of data journaling mode");
3696 }
56889787
TT		 3697 if (test_opt(sb, DELALLOC))
3698 clear_opt(sb, DELALLOC);
001e4a87
TH		 3699 } else {
3700 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3701 }
3702
1751e8a6
LT		 3703 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3704 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec 3705
617ba13b 3706 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3707 (ext4_has_compat_features(sb) ||
3708 ext4_has_ro_compat_features(sb) ||
3709 ext4_has_incompat_features(sb)))
b31e1552
ES		 3710 ext4_msg(sb, KERN_WARNING,
3711 "feature flags set on rev 0 fs, "
3712 "running e2fsck is recommended");
469108ff 3713
ed3654eb
TT		 3714 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3715 set_opt2(sb, HURD_COMPAT);
e2b911c5 3716 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3717 ext4_msg(sb, KERN_ERR,
3718 "The Hurd can't support 64-bit file systems");
3719 goto failed_mount;
3720 }
dec214d0
TE		 3721
3722 /*
3723 * ea_inode feature uses l_i_version field which is not
3724 * available in HURD_COMPAT mode.
3725 */
3726 if (ext4_has_feature_ea_inode(sb)) {
3727 ext4_msg(sb, KERN_ERR,
3728 "ea_inode feature is not supported for Hurd");
3729 goto failed_mount;
3730 }
ed3654eb
TT		 3731 }
3732
2035e776
TT		 3733 if (IS_EXT2_SB(sb)) {
3734 if (ext2_feature_set_ok(sb))
3735 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3736 "using the ext4 subsystem");
3737 else {
0d9366d6
ES		 3738 /*
3739 * If we're probing be silent, if this looks like
3740 * it's actually an ext[34] filesystem.
3741 */
3742 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3743 goto failed_mount;
2035e776
TT		 3744 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3745 "to feature incompatibilities");
3746 goto failed_mount;
3747 }
3748 }
3749
3750 if (IS_EXT3_SB(sb)) {
3751 if (ext3_feature_set_ok(sb))
3752 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3753 "using the ext4 subsystem");
3754 else {
0d9366d6
ES		 3755 /*
3756 * If we're probing be silent, if this looks like
3757 * it's actually an ext4 filesystem.
3758 */
3759 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3760 goto failed_mount;
2035e776
TT		 3761 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3762 "to feature incompatibilities");
3763 goto failed_mount;
3764 }
3765 }
3766
ac27a0ec
DK		 3767 /*
3768 * Check feature flags regardless of the revision level, since we
3769 * previously didn't change the revision level when setting the flags,
3770 * so there is a chance incompat flags are set on a rev 0 filesystem.
3771 */
bc98a42c 3772 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
ac27a0ec 3773 goto failed_mount;
a13fb1a4 3774
261cb20c 3775 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3776 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3777 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552 3778 ext4_msg(sb, KERN_ERR,
8cdf3372
TT		 3779 "Unsupported filesystem blocksize %d (%d log_block_size)",
3780 blocksize, le32_to_cpu(es->s_log_block_size));
3781 goto failed_mount;
3782 }
3783 if (le32_to_cpu(es->s_log_block_size) >
3784 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3785 ext4_msg(sb, KERN_ERR,
3786 "Invalid log block size: %u",
3787 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 3788 goto failed_mount;
3789 }
bfe0a5f4
TT		 3790 if (le32_to_cpu(es->s_log_cluster_size) >
3791 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3792 ext4_msg(sb, KERN_ERR,
3793 "Invalid log cluster size: %u",
3794 le32_to_cpu(es->s_log_cluster_size));
3795 goto failed_mount;
3796 }
ac27a0ec 3797
5b9554dc
TT		 3798 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3799 ext4_msg(sb, KERN_ERR,
3800 "Number of reserved GDT blocks insanely large: %d",
3801 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3802 goto failed_mount;
3803 }
3804
923ae0ff 3805 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
559db4c6
RZ		 3806 if (ext4_has_feature_inline_data(sb)) {
3807 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3808 " that may contain inline data");
24f3478d 3809 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
559db4c6 3810 }
80660f20 3811 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
24f3478d
DW		 3812 ext4_msg(sb, KERN_ERR,
3813 "DAX unsupported by block device. Turning off DAX.");
3814 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
3815 }
923ae0ff
RZ		 3816 }
3817
e2b911c5 3818 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 3819 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3820 es->s_encryption_level);
3821 goto failed_mount;
3822 }
3823
ac27a0ec 3824 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3825 /* Validate the filesystem blocksize */
3826 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3827 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3828 blocksize);
ac27a0ec
DK		 3829 goto failed_mount;
3830 }
3831
2b2d6d01 3832 brelse(bh);
70bbb3e0
AM		 3833 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3834 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3835 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3836 if (!bh) {
b31e1552
ES		 3837 ext4_msg(sb, KERN_ERR,
3838 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3839 goto failed_mount;
3840 }
2716b802 3841 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3842 sbi->s_es = es;
617ba13b 3843 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3844 ext4_msg(sb, KERN_ERR,
3845 "Magic mismatch, very weird!");
ac27a0ec
DK		 3846 goto failed_mount;
3847 }
3848 }
3849
e2b911c5 3850 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 3851 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3852 has_huge_files);
3853 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3854
617ba13b
MC		 3855 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3856 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3857 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3858 } else {
3859 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3860 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
c37e9e01
TT		 3861 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3862 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3863 sbi->s_first_ino);
3864 goto failed_mount;
3865 }
617ba13b 3866 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3867 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3868 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3869 ext4_msg(sb, KERN_ERR,
3870 "unsupported inode size: %d",
2b2d6d01 3871 sbi->s_inode_size);
ac27a0ec
DK		 3872 goto failed_mount;
3873 }
ef7f3835
KS		 3874 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3875 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3876 }
0b8e58a1 3877
0d1ee42f 3878 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 3879 if (ext4_has_feature_64bit(sb)) {
8fadc143 3880 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3881 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3882 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3883 ext4_msg(sb, KERN_ERR,
3884 "unsupported descriptor size %lu",
0d1ee42f
AR		 3885 sbi->s_desc_size);
3886 goto failed_mount;
3887 }
3888 } else
3889 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3890
ac27a0ec 3891 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3892 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 3893
617ba13b 3894 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3895 if (sbi->s_inodes_per_block == 0)
617ba13b 3896 goto cantfind_ext4;
cd6bb35b
TT		 3897 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3898 sbi->s_inodes_per_group > blocksize * 8) {
3899 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3900 sbi->s_blocks_per_group);
3901 goto failed_mount;
3902 }
ac27a0ec
DK		 3903 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3904 sbi->s_inodes_per_block;
0d1ee42f 3905 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3906 sbi->s_sbh = bh;
3907 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3908 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3909 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3910
2b2d6d01 3911 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3912 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3913 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 3914 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 3915 i = le32_to_cpu(es->s_flags);
3916 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3917 sbi->s_hash_unsigned = 3;
3918 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3919#ifdef __CHAR_UNSIGNED__
bc98a42c 3920 if (!sb_rdonly(sb))
23301410
TT		 3921 es->s_flags |=
3922 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3923 sbi->s_hash_unsigned = 3;
f99b2589 3924#else
bc98a42c 3925 if (!sb_rdonly(sb))
23301410
TT		 3926 es->s_flags |=
3927 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3928#endif
23301410 3929 }
f99b2589 3930 }
ac27a0ec 3931
281b5995
TT		 3932 /* Handle clustersize */
3933 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 3934 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 3935 if (has_bigalloc) {
3936 if (clustersize < blocksize) {
3937 ext4_msg(sb, KERN_ERR,
3938 "cluster size (%d) smaller than "
3939 "block size (%d)", clustersize, blocksize);
3940 goto failed_mount;
3941 }
3942 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3943 le32_to_cpu(es->s_log_block_size);
3944 sbi->s_clusters_per_group =
3945 le32_to_cpu(es->s_clusters_per_group);
3946 if (sbi->s_clusters_per_group > blocksize * 8) {
3947 ext4_msg(sb, KERN_ERR,
3948 "#clusters per group too big: %lu",
3949 sbi->s_clusters_per_group);
3950 goto failed_mount;
3951 }
3952 if (sbi->s_blocks_per_group !=
3953 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3954 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3955 "clusters per group (%lu) inconsistent",
3956 sbi->s_blocks_per_group,
3957 sbi->s_clusters_per_group);
3958 goto failed_mount;
3959 }
3960 } else {
3961 if (clustersize != blocksize) {
bfe0a5f4
TT		 3962 ext4_msg(sb, KERN_ERR,
3963 "fragment/cluster size (%d) != "
3964 "block size (%d)", clustersize, blocksize);
3965 goto failed_mount;
281b5995
TT		 3966 }
3967 if (sbi->s_blocks_per_group > blocksize * 8) {
3968 ext4_msg(sb, KERN_ERR,
3969 "#blocks per group too big: %lu",
3970 sbi->s_blocks_per_group);
3971 goto failed_mount;
3972 }
3973 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3974 sbi->s_cluster_bits = 0;
ac27a0ec 3975 }
281b5995
TT		 3976 sbi->s_cluster_ratio = clustersize / blocksize;
3977
960fd856
TT		 3978 /* Do we have standard group size of clustersize * 8 blocks ? */
3979 if (sbi->s_blocks_per_group == clustersize << 3)
3980 set_opt2(sb, STD_GROUP_SIZE);
3981
bf43d84b
ES		 3982 /*
3983 * Test whether we have more sectors than will fit in sector_t,
3984 * and whether the max offset is addressable by the page cache.
3985 */
5a9ae68a 3986 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3987 ext4_blocks_count(es));
5a9ae68a 3988 if (err) {
b31e1552 3989 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3990 " too large to mount safely on this system");
ac27a0ec 3991 if (sizeof(sector_t) < 8)
90c699a9 3992 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3993 goto failed_mount;
3994 }
3995
617ba13b
MC		 3996 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3997 goto cantfind_ext4;
e7c95593 3998
0f2ddca6
FTN		 3999 /* check blocks count against device size */
4000 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4001 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 4002 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4003 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 4004 ext4_blocks_count(es), blocks_count);
4005 goto failed_mount;
4006 }
4007
0b8e58a1
AD		 4008 /*
4009 * It makes no sense for the first data block to be beyond the end
4010 * of the filesystem.
4011 */
4012 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 4013 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 4014 "block %u is beyond end of filesystem (%llu)",
4015 le32_to_cpu(es->s_first_data_block),
4016 ext4_blocks_count(es));
e7c95593
ES		 4017 goto failed_mount;
4018 }
bfe0a5f4
TT		 4019 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4020 (sbi->s_cluster_ratio == 1)) {
4021 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4022 "block is 0 with a 1k block and cluster size");
4023 goto failed_mount;
4024 }
4025
bd81d8ee
LV		 4026 blocks_count = (ext4_blocks_count(es) -
4027 le32_to_cpu(es->s_first_data_block) +
4028 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4029 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 4030 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 4031 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 4032 "(block count %llu, first data block %u, "
b31e1552 4033 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 4034 ext4_blocks_count(es),
4035 le32_to_cpu(es->s_first_data_block),
4036 EXT4_BLOCKS_PER_GROUP(sb));
4037 goto failed_mount;
4038 }
bd81d8ee 4039 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 4040 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4041 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 4042 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4043 EXT4_DESC_PER_BLOCK(sb);
3a4b77cd 4044 if (ext4_has_feature_meta_bg(sb)) {
2ba3e6e8 4045 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3a4b77cd
EG		 4046 ext4_msg(sb, KERN_WARNING,
4047 "first meta block group too large: %u "
4048 "(group descriptor block count %u)",
4049 le32_to_cpu(es->s_first_meta_bg), db_count);
4050 goto failed_mount;
4051 }
4052 }
344476e1
KC		 4053 sbi->s_group_desc = kvmalloc_array(db_count,
4054 sizeof(struct buffer_head *),
4055 GFP_KERNEL);
ac27a0ec 4056 if (sbi->s_group_desc == NULL) {
b31e1552 4057 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 4058 ret = -ENOMEM;
ac27a0ec
DK		 4059 goto failed_mount;
4060 }
bfe0a5f4
TT		 4061 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4062 le32_to_cpu(es->s_inodes_count)) {
4063 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4064 le32_to_cpu(es->s_inodes_count),
4065 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4066 ret = -EINVAL;
4067 goto failed_mount;
4068 }
ac27a0ec 4069
705895b6 4070 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec 4071
85c8f176
AP		 4072 /* Pre-read the descriptors into the buffer cache */
4073 for (i = 0; i < db_count; i++) {
4074 block = descriptor_loc(sb, logical_sb_block, i);
4075 sb_breadahead(sb, block);
4076 }
4077
ac27a0ec 4078 for (i = 0; i < db_count; i++) {
70bbb3e0 4079 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 4080 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 4081 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 4082 ext4_msg(sb, KERN_ERR,
4083 "can't read group descriptor %d", i);
ac27a0ec
DK		 4084 db_count = i;
4085 goto failed_mount2;
4086 }
4087 }
829fa70d 4088 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 4089 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 4090 ret = -EFSCORRUPTED;
f9ae9cf5 4091 goto failed_mount2;
ac27a0ec 4092 }
772cb7c8 4093
f9ae9cf5 4094 sbi->s_gdb_count = db_count;
ac27a0ec 4095
235699a8 4096 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
04496411 4097
a75ae78f 4098 /* Register extent status tree shrinker */
eb68d0e2 4099 if (ext4_es_register_shrinker(sbi))
ce7e010a 4100 goto failed_mount3;
ce7e010a 4101
c9de560d 4102 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 4103 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 4104
f9ae9cf5
TT		 4105 /*
4106 * set up enough so that it can read an inode
4107 */
f6e63f90 4108 sb->s_op = &ext4_sops;
617ba13b
MC		 4109 sb->s_export_op = &ext4_export_ops;
4110 sb->s_xattr = ext4_xattr_handlers;
ffcc4182 4111#ifdef CONFIG_EXT4_FS_ENCRYPTION
a7550b30 4112 sb->s_cop = &ext4_cryptops;
ffcc4182 4113#endif
ac27a0ec 4114#ifdef CONFIG_QUOTA
617ba13b 4115 sb->dq_op = &ext4_quota_operations;
e2b911c5 4116 if (ext4_has_feature_quota(sb))
1fa5efe3 4117 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 4118 else
4119 sb->s_qcop = &ext4_qctl_operations;
689c958c 4120 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 4121#endif
85787090 4122 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
f2fa2ffc 4123
ac27a0ec 4124 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 4125 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 4126
4127 sb->s_root = NULL;
4128
4129 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 4130 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 4131
bc98a42c 4132 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
c5e06d10 4133 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 4134 goto failed_mount3a;
c5e06d10 4135
ac27a0ec
DK		 4136 /*
4137 * The first inode we look at is the journal inode. Don't try
4138 * root first: it may be modified in the journal!
4139 */
e2b911c5 4140 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4753d8a2
TT		 4141 err = ext4_load_journal(sb, es, journal_devnum);
4142 if (err)
50460fe8 4143 goto failed_mount3a;
bc98a42c 4144 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
e2b911c5 4145 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 4146 ext4_msg(sb, KERN_ERR, "required journal recovery "
4147 "suppressed and not mounted read-only");
744692dc 4148 goto failed_mount_wq;
ac27a0ec 4149 } else {
1e381f60
DM		 4150 /* Nojournal mode, all journal mount options are illegal */
4151 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4152 ext4_msg(sb, KERN_ERR, "can't mount with "
4153 "journal_checksum, fs mounted w/o journal");
4154 goto failed_mount_wq;
4155 }
4156 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4157 ext4_msg(sb, KERN_ERR, "can't mount with "
4158 "journal_async_commit, fs mounted w/o journal");
4159 goto failed_mount_wq;
4160 }
4161 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4162 ext4_msg(sb, KERN_ERR, "can't mount with "
4163 "commit=%lu, fs mounted w/o journal",
4164 sbi->s_commit_interval / HZ);
4165 goto failed_mount_wq;
4166 }
4167 if (EXT4_MOUNT_DATA_FLAGS &
4168 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4169 ext4_msg(sb, KERN_ERR, "can't mount with "
4170 "data=, fs mounted w/o journal");
4171 goto failed_mount_wq;
4172 }
4173 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4174 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 4175 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 4176 sbi->s_journal = NULL;
4177 needs_recovery = 0;
4178 goto no_journal;
ac27a0ec
DK		 4179 }
4180
e2b911c5 4181 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 4182 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4183 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4184 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4185 goto failed_mount_wq;
eb40a09c
JS		 4186 }
4187
25ed6e8a
DW		 4188 if (!set_journal_csum_feature_set(sb)) {
4189 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4190 "feature set");
4191 goto failed_mount_wq;
d4da6c9c 4192 }
818d276c 4193
ac27a0ec
DK		 4194 /* We have now updated the journal if required, so we can
4195 * validate the data journaling mode. */
4196 switch (test_opt(sb, DATA_FLAGS)) {
4197 case 0:
4198 /* No mode set, assume a default based on the journal
63f57933
AM		 4199 * capabilities: ORDERED_DATA if the journal can
4200 * cope, else JOURNAL_DATA
4201 */
dab291af 4202 if (jbd2_journal_check_available_features
27f394a7 4203 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
fd8c37ec 4204 set_opt(sb, ORDERED_DATA);
27f394a7
TN		 4205 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4206 } else {
fd8c37ec 4207 set_opt(sb, JOURNAL_DATA);
27f394a7
TN		 4208 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4209 }
ac27a0ec
DK		 4210 break;
4211
617ba13b
MC		 4212 case EXT4_MOUNT_ORDERED_DATA:
4213 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4214 if (!jbd2_journal_check_available_features
4215 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4216 ext4_msg(sb, KERN_ERR, "Journal does not support "
4217 "requested data journaling mode");
744692dc 4218 goto failed_mount_wq;
ac27a0ec
DK		 4219 }
4220 default:
4221 break;
4222 }
ab04df78
JK		 4223
4224 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4225 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4226 ext4_msg(sb, KERN_ERR, "can't mount with "
4227 "journal_async_commit in data=ordered mode");
4228 goto failed_mount_wq;
4229 }
4230
b3881f74 4231 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4232
18aadd47
BJ		 4233 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4234
ce7e010a 4235no_journal:
cdb7ee4c
TE		 4236 if (!test_opt(sb, NO_MBCACHE)) {
4237 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4238 if (!sbi->s_ea_block_cache) {
dec214d0 4239 ext4_msg(sb, KERN_ERR,
cdb7ee4c 4240 "Failed to create ea_block_cache");
dec214d0
TE		 4241 goto failed_mount_wq;
4242 }
cdb7ee4c
TE		 4243
4244 if (ext4_has_feature_ea_inode(sb)) {
4245 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4246 if (!sbi->s_ea_inode_cache) {
4247 ext4_msg(sb, KERN_ERR,
4248 "Failed to create ea_inode_cache");
4249 goto failed_mount_wq;
4250 }
4251 }
9c191f70
M		 4252 }
4253
e2b911c5 4254 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
09cbfeaf 4255 (blocksize != PAGE_SIZE)) {
1cb767cd
TT		 4256 ext4_msg(sb, KERN_ERR,
4257 "Unsupported blocksize for fs encryption");
4258 goto failed_mount_wq;
4259 }
4260
bc98a42c 4261 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
e2b911c5
DW		 4262 !ext4_has_feature_encrypt(sb)) {
4263 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 4264 ext4_commit_super(sb, 1);
4265 }
4266
952fc18e
TT		 4267 /*
4268 * Get the # of file system overhead blocks from the
4269 * superblock if present.
4270 */
4271 if (es->s_overhead_clusters)
4272 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4273 else {
07aa2ea1
LC		 4274 err = ext4_calculate_overhead(sb);
4275 if (err)
952fc18e
TT		 4276 goto failed_mount_wq;
4277 }
4278
fd89d5f2
TH		 4279 /*
4280 * The maximum number of concurrent works can be high and
4281 * concurrency isn't really necessary. Limit it to 1.
4282 */
2e8fa54e
JK		 4283 EXT4_SB(sb)->rsv_conversion_wq =
4284 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4285 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4286 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4287 ret = -ENOMEM;
2e8fa54e
JK		 4288 goto failed_mount4;
4289 }
4290
ac27a0ec 4291 /*
dab291af 4292 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4293 * so we can safely mount the rest of the filesystem now.
4294 */
4295
1d1fe1ee
DH		 4296 root = ext4_iget(sb, EXT4_ROOT_INO);
4297 if (IS_ERR(root)) {
b31e1552 4298 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4299 ret = PTR_ERR(root);
32a9bb57 4300 root = NULL;
ac27a0ec
DK		 4301 goto failed_mount4;
4302 }
4303 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4304 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4305 iput(root);
ac27a0ec
DK		 4306 goto failed_mount4;
4307 }
48fde701 4308 sb->s_root = d_make_root(root);
1d1fe1ee 4309 if (!sb->s_root) {
b31e1552 4310 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4311 ret = -ENOMEM;
4312 goto failed_mount4;
4313 }
ac27a0ec 4314
c89128a0
JK		 4315 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4316 if (ret == -EROFS) {
1751e8a6 4317 sb->s_flags |= SB_RDONLY;
c89128a0
JK		 4318 ret = 0;
4319 } else if (ret)
4320 goto failed_mount4a;
ef7f3835
KS		 4321
4322 /* determine the minimum size of new large inodes, if present */
670e9875
TT		 4323 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4324 sbi->s_want_extra_isize == 0) {
ef7f3835
KS		 4325 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4326 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 4327 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 4328 if (sbi->s_want_extra_isize <
4329 le16_to_cpu(es->s_want_extra_isize))
4330 sbi->s_want_extra_isize =
4331 le16_to_cpu(es->s_want_extra_isize);
4332 if (sbi->s_want_extra_isize <
4333 le16_to_cpu(es->s_min_extra_isize))
4334 sbi->s_want_extra_isize =
4335 le16_to_cpu(es->s_min_extra_isize);
4336 }
4337 }
4338 /* Check if enough inode space is available */
4339 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4340 sbi->s_inode_size) {
4341 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4342 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4343 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4344 "available");
ef7f3835
KS		 4345 }
4346
b5799018 4347 ext4_set_resv_clusters(sb);
27dd4385 4348
6fd058f7
TT		 4349 err = ext4_setup_system_zone(sb);
4350 if (err) {
b31e1552 4351 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4352 "zone (%d)", err);
f9ae9cf5
TT		 4353 goto failed_mount4a;
4354 }
4355
4356 ext4_ext_init(sb);
4357 err = ext4_mb_init(sb);
4358 if (err) {
4359 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4360 err);
dcf2d804 4361 goto failed_mount5;
c2774d84
AK		 4362 }
4363
d5e03cbb
TT		 4364 block = ext4_count_free_clusters(sb);
4365 ext4_free_blocks_count_set(sbi->s_es,
4366 EXT4_C2B(sbi, block));
908c7f19
TH		 4367 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4368 GFP_KERNEL);
d5e03cbb
TT		 4369 if (!err) {
4370 unsigned long freei = ext4_count_free_inodes(sb);
4371 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4372 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4373 GFP_KERNEL);
d5e03cbb
TT		 4374 }
4375 if (!err)
4376 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4377 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4378 if (!err)
908c7f19
TH		 4379 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4380 GFP_KERNEL);
c8585c6f
DJ		 4381 if (!err)
4382 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4383
d5e03cbb
TT		 4384 if (err) {
4385 ext4_msg(sb, KERN_ERR, "insufficient memory");
4386 goto failed_mount6;
4387 }
4388
e2b911c5 4389 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 4390 if (!ext4_fill_flex_info(sb)) {
4391 ext4_msg(sb, KERN_ERR,
4392 "unable to initialize "
4393 "flex_bg meta info!");
4394 goto failed_mount6;
4395 }
4396
bfff6873
LC		 4397 err = ext4_register_li_request(sb, first_not_zeroed);
4398 if (err)
dcf2d804 4399 goto failed_mount6;
bfff6873 4400
b5799018 4401 err = ext4_register_sysfs(sb);
dcf2d804
TM		 4402 if (err)
4403 goto failed_mount7;
3197ebdb 4404
9b2ff357
JK		 4405#ifdef CONFIG_QUOTA
4406 /* Enable quota usage during mount. */
bc98a42c 4407 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
9b2ff357
JK		 4408 err = ext4_enable_quotas(sb);
4409 if (err)
4410 goto failed_mount8;
4411 }
4412#endif /* CONFIG_QUOTA */
4413
617ba13b
MC		 4414 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4415 ext4_orphan_cleanup(sb, es);
4416 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4417 if (needs_recovery) {
b31e1552 4418 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4419 ext4_mark_recovery_complete(sb, es);
4420 }
4421 if (EXT4_SB(sb)->s_journal) {
4422 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4423 descr = " journalled data mode";
4424 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4425 descr = " ordered data mode";
4426 else
4427 descr = " writeback data mode";
4428 } else
4429 descr = "out journal";
4430
79add3a3
LC		 4431 if (test_opt(sb, DISCARD)) {
4432 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4433 if (!blk_queue_discard(q))
4434 ext4_msg(sb, KERN_WARNING,
4435 "mounting with \"discard\" option, but "
4436 "the device does not support discard");
4437 }
4438
e294a537
TT		 4439 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4440 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 4441 "Opts: %.*s%s%s", descr,
4442 (int) sizeof(sbi->s_es->s_mount_opts),
4443 sbi->s_es->s_mount_opts,
e294a537 4444 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4445
66e61a9e
TT		 4446 if (es->s_error_count)
4447 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4448
efbed4dc
TT		 4449 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4450 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4451 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4452 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4453
d4c402d9 4454 kfree(orig_data);
ac27a0ec
DK		 4455 return 0;
4456
617ba13b 4457cantfind_ext4:
ac27a0ec 4458 if (!silent)
b31e1552 4459 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4460 goto failed_mount;
4461
72ba7450
TT		 4462#ifdef CONFIG_QUOTA
4463failed_mount8:
ebd173be 4464 ext4_unregister_sysfs(sb);
72ba7450 4465#endif
dcf2d804
TM		 4466failed_mount7:
4467 ext4_unregister_li_request(sb);
4468failed_mount6:
f9ae9cf5 4469 ext4_mb_release(sb);
d5e03cbb 4470 if (sbi->s_flex_groups)
b93b41d4 4471 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4472 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4473 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4474 percpu_counter_destroy(&sbi->s_dirs_counter);
4475 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4476failed_mount5:
f9ae9cf5
TT		 4477 ext4_ext_release(sb);
4478 ext4_release_system_zone(sb);
4479failed_mount4a:
94bf608a 4480 dput(sb->s_root);
32a9bb57 4481 sb->s_root = NULL;
94bf608a 4482failed_mount4:
b31e1552 4483 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4484 if (EXT4_SB(sb)->rsv_conversion_wq)
4485 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4486failed_mount_wq:
dec214d0
TE		 4487 if (sbi->s_ea_inode_cache) {
4488 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4489 sbi->s_ea_inode_cache = NULL;
4490 }
47387409
TE		 4491 if (sbi->s_ea_block_cache) {
4492 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4493 sbi->s_ea_block_cache = NULL;
82939d79 4494 }
0390131b
FM		 4495 if (sbi->s_journal) {
4496 jbd2_journal_destroy(sbi->s_journal);
4497 sbi->s_journal = NULL;
4498 }
50460fe8 4499failed_mount3a:
d3922a77 4500 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4501failed_mount3:
9105bb14 4502 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4503 if (sbi->s_mmp_tsk)
4504 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4505failed_mount2:
4506 for (i = 0; i < db_count; i++)
4507 brelse(sbi->s_group_desc[i]);
b93b41d4 4508 kvfree(sbi->s_group_desc);
ac27a0ec 4509failed_mount:
0441984a
DW		 4510 if (sbi->s_chksum_driver)
4511 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 4512#ifdef CONFIG_QUOTA
a2d4a646 4513 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4514 kfree(sbi->s_qf_names[i]);
4515#endif
617ba13b 4516 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4517 brelse(bh);
4518out_fail:
4519 sb->s_fs_info = NULL;
f6830165 4520 kfree(sbi->s_blockgroup_lock);
5aee0f8a 4521out_free_base:
ac27a0ec 4522 kfree(sbi);
d4c402d9 4523 kfree(orig_data);
5e405595 4524 fs_put_dax(dax_dev);
07aa2ea1 4525 return err ? err : ret;
ac27a0ec
DK		 4526}
4527
4528/*
4529 * Setup any per-fs journal parameters now. We'll do this both on
4530 * initial mount, once the journal has been initialised but before we've
4531 * done any recovery; and again on any subsequent remount.
4532 */
617ba13b 4533static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4534{
617ba13b 4535 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4536
30773840
TT		 4537 journal->j_commit_interval = sbi->s_commit_interval;
4538 journal->j_min_batch_time = sbi->s_min_batch_time;
4539 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4540
a931da6a 4541 write_lock(&journal->j_state_lock);
ac27a0ec 4542 if (test_opt(sb, BARRIER))
dab291af 4543 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4544 else
dab291af 4545 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4546 if (test_opt(sb, DATA_ERR_ABORT))
4547 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4548 else
4549 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4550 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4551}
4552
c6cb7e77
EW		 4553static struct inode *ext4_get_journal_inode(struct super_block *sb,
4554 unsigned int journal_inum)
ac27a0ec
DK		 4555{
4556 struct inode *journal_inode;
ac27a0ec 4557
c6cb7e77
EW		 4558 /*
4559 * Test for the existence of a valid inode on disk. Bad things
4560 * happen if we iget() an unused inode, as the subsequent iput()
4561 * will try to delete it.
4562 */
1d1fe1ee
DH		 4563 journal_inode = ext4_iget(sb, journal_inum);
4564 if (IS_ERR(journal_inode)) {
b31e1552 4565 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4566 return NULL;
4567 }
4568 if (!journal_inode->i_nlink) {
4569 make_bad_inode(journal_inode);
4570 iput(journal_inode);
b31e1552 4571 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4572 return NULL;
4573 }
4574
e5f8eab8 4575 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4576 journal_inode, journal_inode->i_size);
1d1fe1ee 4577 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4578 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4579 iput(journal_inode);
4580 return NULL;
4581 }
c6cb7e77
EW		 4582 return journal_inode;
4583}
4584
4585static journal_t *ext4_get_journal(struct super_block *sb,
4586 unsigned int journal_inum)
4587{
4588 struct inode *journal_inode;
4589 journal_t *journal;
4590
4591 BUG_ON(!ext4_has_feature_journal(sb));
4592
4593 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4594 if (!journal_inode)
4595 return NULL;
ac27a0ec 4596
dab291af 4597 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4598 if (!journal) {
b31e1552 4599 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4600 iput(journal_inode);
4601 return NULL;
4602 }
4603 journal->j_private = sb;
617ba13b 4604 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4605 return journal;
4606}
4607
617ba13b 4608static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4609 dev_t j_dev)
4610{
2b2d6d01 4611 struct buffer_head *bh;
ac27a0ec 4612 journal_t *journal;
617ba13b
MC		 4613 ext4_fsblk_t start;
4614 ext4_fsblk_t len;
ac27a0ec 4615 int hblock, blocksize;
617ba13b 4616 ext4_fsblk_t sb_block;
ac27a0ec 4617 unsigned long offset;
2b2d6d01 4618 struct ext4_super_block *es;
ac27a0ec
DK		 4619 struct block_device *bdev;
4620
e2b911c5 4621 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4622
b31e1552 4623 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4624 if (bdev == NULL)
4625 return NULL;
4626
ac27a0ec 4627 blocksize = sb->s_blocksize;
e1defc4f 4628 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4629 if (blocksize < hblock) {
b31e1552
ES		 4630 ext4_msg(sb, KERN_ERR,
4631 "blocksize too small for journal device");
ac27a0ec
DK		 4632 goto out_bdev;
4633 }
4634
617ba13b
MC		 4635 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4636 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4637 set_blocksize(bdev, blocksize);
4638 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4639 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4640 "external journal");
ac27a0ec
DK		 4641 goto out_bdev;
4642 }
4643
2716b802 4644 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4645 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4646 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4647 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4648 ext4_msg(sb, KERN_ERR, "external journal has "
4649 "bad superblock");
ac27a0ec
DK		 4650 brelse(bh);
4651 goto out_bdev;
4652 }
4653
df4763be
DW		 4654 if ((le32_to_cpu(es->s_feature_ro_compat) &
4655 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4656 es->s_checksum != ext4_superblock_csum(sb, es)) {
4657 ext4_msg(sb, KERN_ERR, "external journal has "
4658 "corrupt superblock");
4659 brelse(bh);
4660 goto out_bdev;
4661 }
4662
617ba13b 4663 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4664 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4665 brelse(bh);
4666 goto out_bdev;
4667 }
4668
bd81d8ee 4669 len = ext4_blocks_count(es);
ac27a0ec
DK		 4670 start = sb_block + 1;
4671 brelse(bh); /* we're done with the superblock */
4672
dab291af 4673 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4674 start, len, blocksize);
4675 if (!journal) {
b31e1552 4676 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4677 goto out_bdev;
4678 }
4679 journal->j_private = sb;
dfec8a14 4680 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4681 wait_on_buffer(journal->j_sb_buffer);
4682 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4683 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4684 goto out_journal;
4685 }
4686 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4687 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4688 "user (unsupported) - %d",
ac27a0ec
DK		 4689 be32_to_cpu(journal->j_superblock->s_nr_users));
4690 goto out_journal;
4691 }
617ba13b
MC		 4692 EXT4_SB(sb)->journal_bdev = bdev;
4693 ext4_init_journal_params(sb, journal);
ac27a0ec 4694 return journal;
0b8e58a1 4695
ac27a0ec 4696out_journal:
dab291af 4697 jbd2_journal_destroy(journal);
ac27a0ec 4698out_bdev:
617ba13b 4699 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4700 return NULL;
4701}
4702
617ba13b
MC		 4703static int ext4_load_journal(struct super_block *sb,
4704 struct ext4_super_block *es,
ac27a0ec
DK		 4705 unsigned long journal_devnum)
4706{
4707 journal_t *journal;
4708 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4709 dev_t journal_dev;
4710 int err = 0;
4711 int really_read_only;
4712
e2b911c5 4713 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4714
ac27a0ec
DK		 4715 if (journal_devnum &&
4716 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4717 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4718 "numbers have changed");
ac27a0ec
DK		 4719 journal_dev = new_decode_dev(journal_devnum);
4720 } else
4721 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4722
4723 really_read_only = bdev_read_only(sb->s_bdev);
4724
4725 /*
4726 * Are we loading a blank journal or performing recovery after a
4727 * crash? For recovery, we need to check in advance whether we
4728 * can get read-write access to the device.
4729 */
e2b911c5 4730 if (ext4_has_feature_journal_needs_recovery(sb)) {
bc98a42c 4731 if (sb_rdonly(sb)) {
b31e1552
ES		 4732 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4733 "required on readonly filesystem");
ac27a0ec 4734 if (really_read_only) {
b31e1552 4735 ext4_msg(sb, KERN_ERR, "write access "
d98bf8cd
SR		 4736 "unavailable, cannot proceed "
4737 "(try mounting with noload)");
ac27a0ec
DK		 4738 return -EROFS;
4739 }
b31e1552
ES		 4740 ext4_msg(sb, KERN_INFO, "write access will "
4741 "be enabled during recovery");
ac27a0ec
DK		 4742 }
4743 }
4744
4745 if (journal_inum && journal_dev) {
b31e1552
ES		 4746 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4747 "and inode journals!");
ac27a0ec
DK		 4748 return -EINVAL;
4749 }
4750
4751 if (journal_inum) {
617ba13b 4752 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4753 return -EINVAL;
4754 } else {
617ba13b 4755 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4756 return -EINVAL;
4757 }
4758
90576c0b 4759 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4760 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4761
e2b911c5 4762 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4763 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4764 if (!err) {
4765 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4766 if (save)
4767 memcpy(save, ((char *) es) +
4768 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4769 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4770 if (save)
4771 memcpy(((char *) es) + EXT4_S_ERR_START,
4772 save, EXT4_S_ERR_LEN);
4773 kfree(save);
4774 }
ac27a0ec
DK		 4775
4776 if (err) {
b31e1552 4777 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4778 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4779 return err;
4780 }
4781
617ba13b
MC		 4782 EXT4_SB(sb)->s_journal = journal;
4783 ext4_clear_journal_err(sb, es);
ac27a0ec 4784
c41303ce 4785 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4786 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4787 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4788
4789 /* Make sure we flush the recovery flag to disk. */
e2d67052 4790 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4791 }
4792
4793 return 0;
4794}
4795
e2d67052 4796static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4797{
e2d67052 4798 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4799 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4800 int error = 0;
ac27a0ec 4801
bdfe0cbd 4802 if (!sbh || block_device_ejected(sb))
c4be0c1d 4803 return error;
a17712c8
JD		 4804
4805 /*
4806 * The superblock bh should be mapped, but it might not be if the
4807 * device was hot-removed. Not much we can do but fail the I/O.
4808 */
4809 if (!buffer_mapped(sbh))
4810 return error;
4811
71290b36
TT		 4812 /*
4813 * If the file system is mounted read-only, don't update the
4814 * superblock write time. This avoids updating the superblock
4815 * write time when we are mounting the root file system
4816 * read/only but we need to replay the journal; at that point,
4817 * for people who are east of GMT and who make their clock
4818 * tick in localtime for Windows bug-for-bug compatibility,
4819 * the clock is set in the future, and this will cause e2fsck
4820 * to complain and force a full file system check.
4821 */
1751e8a6 4822 if (!(sb->s_flags & SB_RDONLY))
71290b36 4823 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4824 if (sb->s_bdev->bd_part)
4825 es->s_kbytes_written =
4826 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4827 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4828 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4829 else
4830 es->s_kbytes_written =
4831 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4832 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4833 ext4_free_blocks_count_set(es,
57042651
TT		 4834 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4835 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4836 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4837 es->s_free_inodes_count =
4838 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4839 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4840 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4841 ext4_superblock_csum_set(sb);
1566a48a
TT		 4842 if (sync)
4843 lock_buffer(sbh);
4743f839
PKS		 4844 if (buffer_write_io_error(sbh)) {
4845 /*
4846 * Oh, dear. A previous attempt to write the
4847 * superblock failed. This could happen because the
4848 * USB device was yanked out. Or it could happen to
4849 * be a transient write error and maybe the block will
4850 * be remapped. Nothing we can do but to retry the
4851 * write and hope for the best.
4852 */
4853 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4854 "superblock detected");
4855 clear_buffer_write_io_error(sbh);
4856 set_buffer_uptodate(sbh);
4857 }
ac27a0ec 4858 mark_buffer_dirty(sbh);
914258bf 4859 if (sync) {
1566a48a 4860 unlock_buffer(sbh);
564bc402 4861 error = __sync_dirty_buffer(sbh,
00473374 4862 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
c89128a0 4863 if (buffer_write_io_error(sbh)) {
b31e1552
ES		 4864 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4865 "superblock");
914258bf
TT		 4866 clear_buffer_write_io_error(sbh);
4867 set_buffer_uptodate(sbh);
4868 }
4869 }
c4be0c1d 4870 return error;
ac27a0ec
DK		 4871}
4872
ac27a0ec
DK		 4873/*
4874 * Have we just finished recovery? If so, and if we are mounting (or
4875 * remounting) the filesystem readonly, then we will end up with a
4876 * consistent fs on disk. Record that fact.
4877 */
2b2d6d01
TT		 4878static void ext4_mark_recovery_complete(struct super_block *sb,
4879 struct ext4_super_block *es)
ac27a0ec 4880{
617ba13b 4881 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4882
e2b911c5 4883 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4884 BUG_ON(journal != NULL);
4885 return;
4886 }
dab291af 4887 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4888 if (jbd2_journal_flush(journal) < 0)
4889 goto out;
4890
bc98a42c 4891 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
e2b911c5 4892 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4893 ext4_commit_super(sb, 1);
ac27a0ec 4894 }
7ffe1ea8
HK		 4895
4896out:
dab291af 4897 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4898}
4899
4900/*
4901 * If we are mounting (or read-write remounting) a filesystem whose journal
4902 * has recorded an error from a previous lifetime, move that error to the
4903 * main filesystem now.
4904 */
2b2d6d01
TT		 4905static void ext4_clear_journal_err(struct super_block *sb,
4906 struct ext4_super_block *es)
ac27a0ec
DK		 4907{
4908 journal_t *journal;
4909 int j_errno;
4910 const char *errstr;
4911
e2b911c5 4912 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4913
617ba13b 4914 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4915
4916 /*
4917 * Now check for any error status which may have been recorded in the
617ba13b 4918 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4919 */
4920
dab291af 4921 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4922 if (j_errno) {
4923 char nbuf[16];
4924
617ba13b 4925 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4926 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4927 "from previous mount: %s", errstr);
12062ddd 4928 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4929
617ba13b
MC		 4930 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4931 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4932 ext4_commit_super(sb, 1);
ac27a0ec 4933
dab291af 4934 jbd2_journal_clear_err(journal);
d796c52e 4935 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4936 }
4937}
4938
4939/*
4940 * Force the running and committing transactions to commit,
4941 * and wait on the commit.
4942 */
617ba13b 4943int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4944{
4945 journal_t *journal;
ac27a0ec 4946
bc98a42c 4947 if (sb_rdonly(sb))
ac27a0ec
DK		 4948 return 0;
4949
617ba13b 4950 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4951 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4952}
4953
617ba13b 4954static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4955{
14ce0cb4 4956 int ret = 0;
9eddacf9 4957 tid_t target;
06a407f1 4958 bool needs_barrier = false;
8d5d02e6 4959 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4960
49598e04 4961 if (unlikely(ext4_forced_shutdown(sbi)))
0db1ff22
TT		 4962 return 0;
4963
9bffad1e 4964 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4965 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4966 /*
4967 * Writeback quota in non-journalled quota case - journalled quota has
4968 * no dirty dquots
4969 */
4970 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4971 /*
4972 * Data writeback is possible w/o journal transaction, so barrier must
4973 * being sent at the end of the function. But we can skip it if
4974 * transaction_commit will do it for us.
4975 */
bda32530
TT		 4976 if (sbi->s_journal) {
4977 target = jbd2_get_latest_transaction(sbi->s_journal);
4978 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4979 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4980 needs_barrier = true;
4981
4982 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4983 if (wait)
4984 ret = jbd2_log_wait_commit(sbi->s_journal,
4985 target);
4986 }
4987 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4988 needs_barrier = true;
06a407f1
DM		 4989 if (needs_barrier) {
4990 int err;
4991 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4992 if (!ret)
4993 ret = err;
0390131b 4994 }
06a407f1
DM		 4995
4996 return ret;
4997}
4998
ac27a0ec
DK		 4999/*
5000 * LVM calls this function before a (read-only) snapshot is created. This
5001 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 5002 *
5003 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 5004 * state independently. It relies on upper layer to stop all data & metadata
5005 * modifications.
ac27a0ec 5006 */
c4be0c1d 5007static int ext4_freeze(struct super_block *sb)
ac27a0ec 5008{
c4be0c1d
TS		 5009 int error = 0;
5010 journal_t *journal;
ac27a0ec 5011
bc98a42c 5012 if (sb_rdonly(sb))
9ca92389 5013 return 0;
ac27a0ec 5014
9ca92389 5015 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 5016
bb044576
TT		 5017 if (journal) {
5018 /* Now we set up the journal barrier. */
5019 jbd2_journal_lock_updates(journal);
ac27a0ec 5020
bb044576
TT		 5021 /*
5022 * Don't clear the needs_recovery flag if we failed to
5023 * flush the journal.
5024 */
5025 error = jbd2_journal_flush(journal);
5026 if (error < 0)
5027 goto out;
c642dc9e
ES		 5028
5029 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 5030 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 5031 }
9ca92389 5032
9ca92389 5033 error = ext4_commit_super(sb, 1);
6b0310fb 5034out:
bb044576
TT		 5035 if (journal)
5036 /* we rely on upper layer to stop further updates */
5037 jbd2_journal_unlock_updates(journal);
6b0310fb 5038 return error;
ac27a0ec
DK		 5039}
5040
5041/*
5042 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5043 * flag here, even though the filesystem is not technically dirty yet.
5044 */
c4be0c1d 5045static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 5046{
bc98a42c 5047 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 5048 return 0;
5049
c642dc9e
ES		 5050 if (EXT4_SB(sb)->s_journal) {
5051 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 5052 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 5053 }
5054
9ca92389 5055 ext4_commit_super(sb, 1);
c4be0c1d 5056 return 0;
ac27a0ec
DK		 5057}
5058
673c6100
TT		 5059/*
5060 * Structure to save mount options for ext4_remount's benefit
5061 */
5062struct ext4_mount_options {
5063 unsigned long s_mount_opt;
a2595b8a 5064 unsigned long s_mount_opt2;
08cefc7a
EB		 5065 kuid_t s_resuid;
5066 kgid_t s_resgid;
673c6100
TT		 5067 unsigned long s_commit_interval;
5068 u32 s_min_batch_time, s_max_batch_time;
5069#ifdef CONFIG_QUOTA
5070 int s_jquota_fmt;
a2d4a646 5071 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 5072#endif
5073};
5074
2b2d6d01 5075static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 5076{
2b2d6d01 5077 struct ext4_super_block *es;
617ba13b 5078 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5079 unsigned long old_sb_flags;
617ba13b 5080 struct ext4_mount_options old_opts;
c79d967d 5081 int enable_quota = 0;
8a266467 5082 ext4_group_t g;
b3881f74 5083 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 5084 int err = 0;
ac27a0ec 5085#ifdef CONFIG_QUOTA
03dafb5f 5086 int i, j;
ac27a0ec 5087#endif
d4c402d9 5088 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 5089
5090 /* Store the original options */
5091 old_sb_flags = sb->s_flags;
5092 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 5093 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 5094 old_opts.s_resuid = sbi->s_resuid;
5095 old_opts.s_resgid = sbi->s_resgid;
5096 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 5097 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5098 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 5099#ifdef CONFIG_QUOTA
5100 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 5101 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 5102 if (sbi->s_qf_names[i]) {
5103 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
5104 GFP_KERNEL);
5105 if (!old_opts.s_qf_names[i]) {
5106 for (j = 0; j < i; j++)
5107 kfree(old_opts.s_qf_names[j]);
3e36a163 5108 kfree(orig_data);
03dafb5f
CG		 5109 return -ENOMEM;
5110 }
5111 } else
5112 old_opts.s_qf_names[i] = NULL;
ac27a0ec 5113#endif
b3881f74
TT		 5114 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5115 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 5116
661aa520 5117 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 5118 err = -EINVAL;
5119 goto restore_opts;
5120 }
5121
6b992ff2 5122 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 5123 test_opt(sb, JOURNAL_CHECKSUM)) {
5124 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 5125 "during remount not supported; ignoring");
5126 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 5127 }
5128
6ae6514b
PS		 5129 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5130 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5131 ext4_msg(sb, KERN_ERR, "can't mount with "
5132 "both data=journal and delalloc");
5133 err = -EINVAL;
5134 goto restore_opts;
5135 }
5136 if (test_opt(sb, DIOREAD_NOLOCK)) {
5137 ext4_msg(sb, KERN_ERR, "can't mount with "
5138 "both data=journal and dioread_nolock");
5139 err = -EINVAL;
5140 goto restore_opts;
5141 }
923ae0ff
RZ		 5142 if (test_opt(sb, DAX)) {
5143 ext4_msg(sb, KERN_ERR, "can't mount with "
5144 "both data=journal and dax");
5145 err = -EINVAL;
5146 goto restore_opts;
5147 }
ab04df78
JK		 5148 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5149 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5150 ext4_msg(sb, KERN_ERR, "can't mount with "
5151 "journal_async_commit in data=ordered mode");
5152 err = -EINVAL;
5153 goto restore_opts;
5154 }
923ae0ff
RZ		 5155 }
5156
cdb7ee4c
TE		 5157 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5158 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5159 err = -EINVAL;
5160 goto restore_opts;
5161 }
5162
923ae0ff
RZ		 5163 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5164 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5165 "dax flag with busy inodes while remounting");
5166 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 5167 }
5168
4ab2f15b 5169 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 5170 ext4_abort(sb, "Abort forced by user");
ac27a0ec 5171
1751e8a6
LT		 5172 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5173 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec
DK		 5174
5175 es = sbi->s_es;
5176
b3881f74 5177 if (sbi->s_journal) {
0390131b 5178 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 5179 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5180 }
ac27a0ec 5181
1751e8a6
LT		 5182 if (*flags & SB_LAZYTIME)
5183 sb->s_flags |= SB_LAZYTIME;
a2fd66d0 5184
1751e8a6 5185 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
4ab2f15b 5186 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 5187 err = -EROFS;
5188 goto restore_opts;
5189 }
5190
1751e8a6 5191 if (*flags & SB_RDONLY) {
38c03b34
TT		 5192 err = sync_filesystem(sb);
5193 if (err < 0)
5194 goto restore_opts;
0f0dd62f
CH		 5195 err = dquot_suspend(sb, -1);
5196 if (err < 0)
c79d967d 5197 goto restore_opts;
c79d967d 5198
ac27a0ec
DK		 5199 /*
5200 * First of all, the unconditional stuff we have to do
5201 * to disable replay of the journal when we next remount
5202 */
1751e8a6 5203 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 5204
5205 /*
5206 * OK, test if we are remounting a valid rw partition
5207 * readonly, and if so set the rdonly flag and then
5208 * mark the partition as valid again.
5209 */
617ba13b
MC		 5210 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5211 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 5212 es->s_state = cpu_to_le16(sbi->s_mount_state);
5213
a63c9eb2 5214 if (sbi->s_journal)
0390131b 5215 ext4_mark_recovery_complete(sb, es);
ac27a0ec 5216 } else {
a13fb1a4 5217 /* Make sure we can mount this feature set readwrite */
e2b911c5 5218 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 5219 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 5220 err = -EROFS;
5221 goto restore_opts;
5222 }
8a266467
TT		 5223 /*
5224 * Make sure the group descriptor checksums
0b8e58a1 5225 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 5226 */
5227 for (g = 0; g < sbi->s_groups_count; g++) {
5228 struct ext4_group_desc *gdp =
5229 ext4_get_group_desc(sb, g, NULL);
5230
feb0ab32 5231 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5232 ext4_msg(sb, KERN_ERR,
5233 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 5234 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 5235 le16_to_cpu(gdp->bg_checksum));
6a797d27 5236 err = -EFSBADCRC;
8a266467
TT		 5237 goto restore_opts;
5238 }
5239 }
5240
ead6596b
ES		 5241 /*
5242 * If we have an unprocessed orphan list hanging
5243 * around from a previously readonly bdev mount,
5244 * require a full umount/remount for now.
5245 */
5246 if (es->s_last_orphan) {
b31e1552 5247 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5248 "remount RDWR because of unprocessed "
5249 "orphan inode list. Please "
b31e1552 5250 "umount/remount instead");
ead6596b
ES		 5251 err = -EINVAL;
5252 goto restore_opts;
5253 }
5254
ac27a0ec
DK		 5255 /*
5256 * Mounting a RDONLY partition read-write, so reread
5257 * and store the current valid flag. (It may have
5258 * been changed by e2fsck since we originally mounted
5259 * the partition.)
5260 */
0390131b
FM		 5261 if (sbi->s_journal)
5262 ext4_clear_journal_err(sb, es);
ac27a0ec 5263 sbi->s_mount_state = le16_to_cpu(es->s_state);
c89128a0
JK		 5264
5265 err = ext4_setup_super(sb, es, 0);
5266 if (err)
5267 goto restore_opts;
5268
5269 sb->s_flags &= ~SB_RDONLY;
e2b911c5 5270 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 5271 if (ext4_multi_mount_protect(sb,
5272 le64_to_cpu(es->s_mmp_block))) {
5273 err = -EROFS;
5274 goto restore_opts;
5275 }
c79d967d 5276 enable_quota = 1;
ac27a0ec
DK		 5277 }
5278 }
bfff6873
LC		 5279
5280 /*
5281 * Reinitialize lazy itable initialization thread based on
5282 * current settings
5283 */
bc98a42c 5284 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
bfff6873
LC		 5285 ext4_unregister_li_request(sb);
5286 else {
5287 ext4_group_t first_not_zeroed;
5288 first_not_zeroed = ext4_has_uninit_itable(sb);
5289 ext4_register_li_request(sb, first_not_zeroed);
5290 }
5291
6fd058f7 5292 ext4_setup_system_zone(sb);
c89128a0
JK		 5293 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5294 err = ext4_commit_super(sb, 1);
5295 if (err)
5296 goto restore_opts;
5297 }
0390131b 5298
ac27a0ec
DK		 5299#ifdef CONFIG_QUOTA
5300 /* Release old quota file names */
a2d4a646 5301 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5302 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5303 if (enable_quota) {
5304 if (sb_any_quota_suspended(sb))
5305 dquot_resume(sb, -1);
e2b911c5 5306 else if (ext4_has_feature_quota(sb)) {
7c319d32 5307 err = ext4_enable_quotas(sb);
07724f98 5308 if (err)
7c319d32 5309 goto restore_opts;
7c319d32
AK		 5310 }
5311 }
ac27a0ec 5312#endif
d4c402d9 5313
1751e8a6 5314 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
d4c402d9
CW		 5315 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5316 kfree(orig_data);
ac27a0ec 5317 return 0;
0b8e58a1 5318
ac27a0ec
DK		 5319restore_opts:
5320 sb->s_flags = old_sb_flags;
5321 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5322 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5323 sbi->s_resuid = old_opts.s_resuid;
5324 sbi->s_resgid = old_opts.s_resgid;
5325 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5326 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5327 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5328#ifdef CONFIG_QUOTA
5329 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5330 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5331 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5332 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5333 }
5334#endif
d4c402d9 5335 kfree(orig_data);
ac27a0ec
DK		 5336 return err;
5337}
5338
689c958c
LX		 5339#ifdef CONFIG_QUOTA
5340static int ext4_statfs_project(struct super_block *sb,
5341 kprojid_t projid, struct kstatfs *buf)
5342{
5343 struct kqid qid;
5344 struct dquot *dquot;
5345 u64 limit;
5346 u64 curblock;
5347
5348 qid = make_kqid_projid(projid);
5349 dquot = dqget(sb, qid);
5350 if (IS_ERR(dquot))
5351 return PTR_ERR(dquot);
7b9ca4c6 5352 spin_lock(&dquot->dq_dqb_lock);
689c958c
LX		 5353
5354 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5355 dquot->dq_dqb.dqb_bsoftlimit :
5356 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5357 if (limit && buf->f_blocks > limit) {
f06925c7
KK		 5358 curblock = (dquot->dq_dqb.dqb_curspace +
5359 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
689c958c
LX		 5360 buf->f_blocks = limit;
5361 buf->f_bfree = buf->f_bavail =
5362 (buf->f_blocks > curblock) ?
5363 (buf->f_blocks - curblock) : 0;
5364 }
5365
5366 limit = dquot->dq_dqb.dqb_isoftlimit ?
5367 dquot->dq_dqb.dqb_isoftlimit :
5368 dquot->dq_dqb.dqb_ihardlimit;
5369 if (limit && buf->f_files > limit) {
5370 buf->f_files = limit;
5371 buf->f_ffree =
5372 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5373 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5374 }
5375
7b9ca4c6 5376 spin_unlock(&dquot->dq_dqb_lock);
689c958c
LX		 5377 dqput(dquot);
5378 return 0;
5379}
5380#endif
5381
2b2d6d01 5382static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5383{
5384 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5385 struct ext4_sb_info *sbi = EXT4_SB(sb);
5386 struct ext4_super_block *es = sbi->s_es;
27dd4385 5387 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5388 u64 fsid;
d02a9391 5389 s64 bfree;
27dd4385 5390 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5391
952fc18e
TT		 5392 if (!test_opt(sb, MINIX_DF))
5393 overhead = sbi->s_overhead;
ac27a0ec 5394
617ba13b 5395 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5396 buf->f_bsize = sb->s_blocksize;
b72f78cb 5397 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5398 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5399 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5400 /* prevent underflow in case that few free space is available */
57042651 5401 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5402 buf->f_bavail = buf->f_bfree -
5403 (ext4_r_blocks_count(es) + resv_blocks);
5404 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5405 buf->f_bavail = 0;
5406 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5407 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5408 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5409 fsid = le64_to_cpup((void *)es->s_uuid) ^
5410 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5411 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5412 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5413
689c958c
LX		 5414#ifdef CONFIG_QUOTA
5415 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5416 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5417 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5418#endif
ac27a0ec
DK		 5419 return 0;
5420}
5421
ac27a0ec
DK		 5422
5423#ifdef CONFIG_QUOTA
5424
bc8230ee
JK		 5425/*
5426 * Helper functions so that transaction is started before we acquire dqio_sem
5427 * to keep correct lock ordering of transaction > dqio_sem
5428 */
ac27a0ec
DK		 5429static inline struct inode *dquot_to_inode(struct dquot *dquot)
5430{
4c376dca 5431 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5432}
5433
617ba13b 5434static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5435{
5436 int ret, err;
5437 handle_t *handle;
5438 struct inode *inode;
5439
5440 inode = dquot_to_inode(dquot);
9924a92a 5441 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5442 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5443 if (IS_ERR(handle))
5444 return PTR_ERR(handle);
5445 ret = dquot_commit(dquot);
617ba13b 5446 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5447 if (!ret)
5448 ret = err;
5449 return ret;
5450}
5451
617ba13b 5452static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5453{
5454 int ret, err;
5455 handle_t *handle;
5456
9924a92a 5457 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5458 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5459 if (IS_ERR(handle))
5460 return PTR_ERR(handle);
5461 ret = dquot_acquire(dquot);
617ba13b 5462 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5463 if (!ret)
5464 ret = err;
5465 return ret;
5466}
5467
617ba13b 5468static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5469{
5470 int ret, err;
5471 handle_t *handle;
5472
9924a92a 5473 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5474 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5475 if (IS_ERR(handle)) {
5476 /* Release dquot anyway to avoid endless cycle in dqput() */
5477 dquot_release(dquot);
ac27a0ec 5478 return PTR_ERR(handle);
9c3013e9 5479 }
ac27a0ec 5480 ret = dquot_release(dquot);
617ba13b 5481 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5482 if (!ret)
5483 ret = err;
5484 return ret;
5485}
5486
617ba13b 5487static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5488{
262b4662
JK		 5489 struct super_block *sb = dquot->dq_sb;
5490 struct ext4_sb_info *sbi = EXT4_SB(sb);
5491
2c8be6b2 5492 /* Are we journaling quotas? */
e2b911c5 5493 if (ext4_has_feature_quota(sb) ||
262b4662 5494 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5495 dquot_mark_dquot_dirty(dquot);
617ba13b 5496 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5497 } else {
5498 return dquot_mark_dquot_dirty(dquot);
5499 }
5500}
5501
617ba13b 5502static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5503{
5504 int ret, err;
5505 handle_t *handle;
5506
5507 /* Data block + inode block */
2b0143b5 5508 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5509 if (IS_ERR(handle))
5510 return PTR_ERR(handle);
5511 ret = dquot_commit_info(sb, type);
617ba13b 5512 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5513 if (!ret)
5514 ret = err;
5515 return ret;
5516}
5517
5518/*
5519 * Turn on quotas during mount time - we need to find
5520 * the quota file and such...
5521 */
617ba13b 5522static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5523{
287a8095
CH		 5524 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5525 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5526}
5527
daf647d2
TT		 5528static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5529{
5530 struct ext4_inode_info *ei = EXT4_I(inode);
5531
5532 /* The first argument of lockdep_set_subclass has to be
5533 * *exactly* the same as the argument to init_rwsem() --- in
5534 * this case, in init_once() --- or lockdep gets unhappy
5535 * because the name of the lock is set using the
5536 * stringification of the argument to init_rwsem().
5537 */
5538 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5539 lockdep_set_subclass(&ei->i_data_sem, subclass);
5540}
5541
ac27a0ec
DK		 5542/*
5543 * Standard function to be called on quota_on
5544 */
617ba13b 5545static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 5546 const struct path *path)
ac27a0ec
DK		 5547{
5548 int err;
ac27a0ec
DK		 5549
5550 if (!test_opt(sb, QUOTA))
5551 return -EINVAL;
0623543b 5552
ac27a0ec 5553 /* Quotafile not on the same filesystem? */
d8c9584e 5554 if (path->dentry->d_sb != sb)
ac27a0ec 5555 return -EXDEV;
0623543b
JK		 5556 /* Journaling quota? */
5557 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5558 /* Quotafile not in fs root? */
f00c9e44 5559 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5560 ext4_msg(sb, KERN_WARNING,
5561 "Quota file not on filesystem root. "
5562 "Journaled quota will not work");
91389240
JK		 5563 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5564 } else {
5565 /*
5566 * Clear the flag just in case mount options changed since
5567 * last time.
5568 */
5569 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
2b2d6d01 5570 }
0623543b
JK		 5571
5572 /*
5573 * When we journal data on quota file, we have to flush journal to see
5574 * all updates to the file when we bypass pagecache...
5575 */
0390131b 5576 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5577 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5578 /*
5579 * We don't need to lock updates but journal_flush() could
5580 * otherwise be livelocked...
5581 */
5582 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5583 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5584 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5585 if (err)
7ffe1ea8 5586 return err;
0623543b 5587 }
957153fc 5588
daf647d2
TT		 5589 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5590 err = dquot_quota_on(sb, type, format_id, path);
957153fc 5591 if (err) {
daf647d2
TT		 5592 lockdep_set_quota_inode(path->dentry->d_inode,
5593 I_DATA_SEM_NORMAL);
957153fc
JK		 5594 } else {
5595 struct inode *inode = d_inode(path->dentry);
5596 handle_t *handle;
5597
61a92987
JK		 5598 /*
5599 * Set inode flags to prevent userspace from messing with quota
5600 * files. If this fails, we return success anyway since quotas
5601 * are already enabled and this is not a hard failure.
5602 */
957153fc
JK		 5603 inode_lock(inode);
5604 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5605 if (IS_ERR(handle))
5606 goto unlock_inode;
5607 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5608 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5609 S_NOATIME | S_IMMUTABLE);
5610 ext4_mark_inode_dirty(handle, inode);
5611 ext4_journal_stop(handle);
5612 unlock_inode:
5613 inode_unlock(inode);
5614 }
daf647d2 5615 return err;
ac27a0ec
DK		 5616}
5617
7c319d32
AK		 5618static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5619 unsigned int flags)
5620{
5621 int err;
5622 struct inode *qf_inode;
a2d4a646 5623 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5624 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5625 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5626 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5627 };
5628
e2b911c5 5629 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5630
5631 if (!qf_inums[type])
5632 return -EPERM;
5633
5634 qf_inode = ext4_iget(sb, qf_inums[type]);
5635 if (IS_ERR(qf_inode)) {
5636 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5637 return PTR_ERR(qf_inode);
5638 }
5639
bcb13850
JK		 5640 /* Don't account quota for quota files to avoid recursion */
5641 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5642 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7c319d32
AK		 5643 err = dquot_enable(qf_inode, type, format_id, flags);
5644 iput(qf_inode);
daf647d2
TT		 5645 if (err)
5646 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7c319d32
AK		 5647
5648 return err;
5649}
5650
5651/* Enable usage tracking for all quota types. */
5652static int ext4_enable_quotas(struct super_block *sb)
5653{
5654 int type, err = 0;
a2d4a646 5655 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5656 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5657 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5658 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 5659 };
49da9392
JK		 5660 bool quota_mopt[EXT4_MAXQUOTAS] = {
5661 test_opt(sb, USRQUOTA),
5662 test_opt(sb, GRPQUOTA),
5663 test_opt(sb, PRJQUOTA),
5664 };
7c319d32 5665
91389240 5666 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
a2d4a646 5667 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5668 if (qf_inums[type]) {
5669 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 5670 DQUOT_USAGE_ENABLED |
5671 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32 5672 if (err) {
95f1fda4 5673 for (type--; type >= 0; type--)
5674 dquot_quota_off(sb, type);
5675
7c319d32 5676 ext4_warning(sb,
72ba7450
TT		 5677 "Failed to enable quota tracking "
5678 "(type=%d, err=%d). Please run "
5679 "e2fsck to fix.", type, err);
7c319d32
AK		 5680 return err;
5681 }
5682 }
5683 }
5684 return 0;
5685}
5686
ca0e05e4
DM		 5687static int ext4_quota_off(struct super_block *sb, int type)
5688{
21f97697
JK		 5689 struct inode *inode = sb_dqopt(sb)->files[type];
5690 handle_t *handle;
957153fc 5691 int err;
21f97697 5692
87009d86
DM		 5693 /* Force all delayed allocation blocks to be allocated.
5694 * Caller already holds s_umount sem */
5695 if (test_opt(sb, DELALLOC))
ca0e05e4 5696 sync_filesystem(sb);
ca0e05e4 5697
957153fc 5698 if (!inode || !igrab(inode))
0b268590
AG		 5699 goto out;
5700
957153fc 5701 err = dquot_quota_off(sb, type);
964edf66 5702 if (err || ext4_has_feature_quota(sb))
957153fc
JK		 5703 goto out_put;
5704
5705 inode_lock(inode);
61a92987
JK		 5706 /*
5707 * Update modification times of quota files when userspace can
5708 * start looking at them. If we fail, we return success anyway since
5709 * this is not a hard failure and quotas are already disabled.
5710 */
9924a92a 5711 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697 5712 if (IS_ERR(handle))
957153fc
JK		 5713 goto out_unlock;
5714 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5715 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
eeca7ea1 5716 inode->i_mtime = inode->i_ctime = current_time(inode);
21f97697
JK		 5717 ext4_mark_inode_dirty(handle, inode);
5718 ext4_journal_stop(handle);
957153fc
JK		 5719out_unlock:
5720 inode_unlock(inode);
5721out_put:
964edf66 5722 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 5723 iput(inode);
5724 return err;
21f97697 5725out:
ca0e05e4
DM		 5726 return dquot_quota_off(sb, type);
5727}
5728
ac27a0ec
DK		 5729/* Read data from quotafile - avoid pagecache and such because we cannot afford
5730 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5731 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5732 * we don't have to be afraid of races */
617ba13b 5733static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5734 size_t len, loff_t off)
5735{
5736 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5737 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5738 int offset = off & (sb->s_blocksize - 1);
5739 int tocopy;
5740 size_t toread;
5741 struct buffer_head *bh;
5742 loff_t i_size = i_size_read(inode);
5743
5744 if (off > i_size)
5745 return 0;
5746 if (off+len > i_size)
5747 len = i_size-off;
5748 toread = len;
5749 while (toread > 0) {
5750 tocopy = sb->s_blocksize - offset < toread ?
5751 sb->s_blocksize - offset : toread;
1c215028
TT		 5752 bh = ext4_bread(NULL, inode, blk, 0);
5753 if (IS_ERR(bh))
5754 return PTR_ERR(bh);
ac27a0ec
DK		 5755 if (!bh) /* A hole? */
5756 memset(data, 0, tocopy);
5757 else
5758 memcpy(data, bh->b_data+offset, tocopy);
5759 brelse(bh);
5760 offset = 0;
5761 toread -= tocopy;
5762 data += tocopy;
5763 blk++;
5764 }
5765 return len;
5766}
5767
5768/* Write to quotafile (we know the transaction is already started and has
5769 * enough credits) */
617ba13b 5770static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5771 const char *data, size_t len, loff_t off)
5772{
5773 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5774 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5775 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5776 int retries = 0;
ac27a0ec
DK		 5777 struct buffer_head *bh;
5778 handle_t *handle = journal_current_handle();
5779
0390131b 5780 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5781 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5782 " cancelled because transaction is not started",
9c3013e9
JK		 5783 (unsigned long long)off, (unsigned long long)len);
5784 return -EIO;
5785 }
67eeb568
DM		 5786 /*
5787 * Since we account only one data block in transaction credits,
5788 * then it is impossible to cross a block boundary.
5789 */
5790 if (sb->s_blocksize - offset < len) {
5791 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5792 " cancelled because not block aligned",
5793 (unsigned long long)off, (unsigned long long)len);
5794 return -EIO;
5795 }
5796
c5e298ae
TT		 5797 do {
5798 bh = ext4_bread(handle, inode, blk,
5799 EXT4_GET_BLOCKS_CREATE |
5800 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5801 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5802 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5803 if (IS_ERR(bh))
5804 return PTR_ERR(bh);
67eeb568
DM		 5805 if (!bh)
5806 goto out;
5d601255 5807 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5808 err = ext4_journal_get_write_access(handle, bh);
5809 if (err) {
5810 brelse(bh);
1c215028 5811 return err;
ac27a0ec 5812 }
67eeb568
DM		 5813 lock_buffer(bh);
5814 memcpy(bh->b_data+offset, data, len);
5815 flush_dcache_page(bh->b_page);
5816 unlock_buffer(bh);
62d2b5f2 5817 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5818 brelse(bh);
ac27a0ec 5819out:
67eeb568
DM		 5820 if (inode->i_size < off + len) {
5821 i_size_write(inode, off + len);
617ba13b 5822 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5823 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5824 }
67eeb568 5825 return len;
ac27a0ec
DK		 5826}
5827
8f0e8746
TT		 5828static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5829{
5830 const struct quota_format_ops *ops;
5831
5832 if (!sb_has_quota_loaded(sb, qid->type))
5833 return -ESRCH;
5834 ops = sb_dqopt(sb)->ops[qid->type];
5835 if (!ops || !ops->get_next_id)
5836 return -ENOSYS;
5837 return dquot_get_next_id(sb, qid);
5838}
ac27a0ec
DK		 5839#endif
5840
152a0836
AV		 5841static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5842 const char *dev_name, void *data)
ac27a0ec 5843{
152a0836 5844 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5845}
5846
c290ea01 5847#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5848static inline void register_as_ext2(void)
5849{
5850 int err = register_filesystem(&ext2_fs_type);
5851 if (err)
5852 printk(KERN_WARNING
5853 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5854}
5855
5856static inline void unregister_as_ext2(void)
5857{
5858 unregister_filesystem(&ext2_fs_type);
5859}
2035e776
TT		 5860
5861static inline int ext2_feature_set_ok(struct super_block *sb)
5862{
e2b911c5 5863 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776 5864 return 0;
bc98a42c 5865 if (sb_rdonly(sb))
2035e776 5866 return 1;
e2b911c5 5867 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5868 return 0;
5869 return 1;
5870}
24b58424
TT		 5871#else
5872static inline void register_as_ext2(void) { }
5873static inline void unregister_as_ext2(void) { }
2035e776 5874static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5875#endif
5876
24b58424
TT		 5877static inline void register_as_ext3(void)
5878{
5879 int err = register_filesystem(&ext3_fs_type);
5880 if (err)
5881 printk(KERN_WARNING
5882 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5883}
5884
5885static inline void unregister_as_ext3(void)
5886{
5887 unregister_filesystem(&ext3_fs_type);
5888}
2035e776
TT		 5889
5890static inline int ext3_feature_set_ok(struct super_block *sb)
5891{
e2b911c5 5892 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5893 return 0;
e2b911c5 5894 if (!ext4_has_feature_journal(sb))
2035e776 5895 return 0;
bc98a42c 5896 if (sb_rdonly(sb))
2035e776 5897 return 1;
e2b911c5 5898 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5899 return 0;
5900 return 1;
5901}
24b58424 5902
03010a33
TT		 5903static struct file_system_type ext4_fs_type = {
5904 .owner = THIS_MODULE,
5905 .name = "ext4",
152a0836 5906 .mount = ext4_mount,
03010a33
TT		 5907 .kill_sb = kill_block_super,
5908 .fs_flags = FS_REQUIRES_DEV,
5909};
7f78e035 5910MODULE_ALIAS_FS("ext4");
03010a33 5911
e9e3bcec
ES		 5912/* Shared across all ext4 file systems */
5913wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 5914
5dabfc78 5915static int __init ext4_init_fs(void)
ac27a0ec 5916{
e9e3bcec 5917 int i, err;
c9de560d 5918
e294a537 5919 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5920 ext4_li_info = NULL;
5921 mutex_init(&ext4_li_mtx);
5922
9a4c8019 5923 /* Build-time check for flags consistency */
12e9b892 5924 ext4_check_flag_values();
e9e3bcec 5925
e142d052 5926 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 5927 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 5928
51865fda 5929 err = ext4_init_es();
6fd058f7
TT		 5930 if (err)
5931 return err;
51865fda
ZL		 5932
5933 err = ext4_init_pageio();
5934 if (err)
b5799018 5935 goto out5;
51865fda 5936
5dabfc78 5937 err = ext4_init_system_zone();
bd2d0210 5938 if (err)
b5799018 5939 goto out4;
857ac889 5940
b5799018 5941 err = ext4_init_sysfs();
dd68314c 5942 if (err)
b5799018 5943 goto out3;
857ac889 5944
5dabfc78 5945 err = ext4_init_mballoc();
c9de560d
AT		 5946 if (err)
5947 goto out2;
ac27a0ec
DK		 5948 err = init_inodecache();
5949 if (err)
5950 goto out1;
24b58424 5951 register_as_ext3();
2035e776 5952 register_as_ext2();
03010a33 5953 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5954 if (err)
5955 goto out;
bfff6873 5956
ac27a0ec
DK		 5957 return 0;
5958out:
24b58424
TT		 5959 unregister_as_ext2();
5960 unregister_as_ext3();
ac27a0ec
DK		 5961 destroy_inodecache();
5962out1:
5dabfc78 5963 ext4_exit_mballoc();
9c191f70 5964out2:
b5799018
TT		 5965 ext4_exit_sysfs();
5966out3:
5dabfc78 5967 ext4_exit_system_zone();
b5799018 5968out4:
5dabfc78 5969 ext4_exit_pageio();
b5799018 5970out5:
51865fda
ZL		 5971 ext4_exit_es();
5972
ac27a0ec
DK		 5973 return err;
5974}
5975
5dabfc78 5976static void __exit ext4_exit_fs(void)
ac27a0ec 5977{
bfff6873 5978 ext4_destroy_lazyinit_thread();
24b58424
TT		 5979 unregister_as_ext2();
5980 unregister_as_ext3();
03010a33 5981 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5982 destroy_inodecache();
5dabfc78 5983 ext4_exit_mballoc();
b5799018 5984 ext4_exit_sysfs();
5dabfc78
TT		 5985 ext4_exit_system_zone();
5986 ext4_exit_pageio();
dd12ed14 5987 ext4_exit_es();
ac27a0ec
DK		 5988}
5989
5990MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5991MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5992MODULE_LICENSE("GPL");
7ef79ad5 5993MODULE_SOFTDEP("pre: crc32c");
5dabfc78
TT		 5994module_init(ext4_init_fs)
5995module_exit(ext4_exit_fs)
A mirror of Linus' kernel repository