]> git.ipfire.org Git - thirdparty/kernel/linux.git/blame - fs/ext4/super.c
projects / thirdparty / kernel / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: add more inode number paranoia checks
[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
1276static unsigned ext4_max_namelen(struct inode *inode)
1277{
1278 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1279 EXT4_NAME_LEN;
1280}
1281
6f69f0ed 1282static const struct fscrypt_operations ext4_cryptops = {
a5d431ef 1283 .key_prefix = "ext4:",
a7550b30 1284 .get_context = ext4_get_context,
a7550b30
JK		 1285 .set_context = ext4_set_context,
1286 .dummy_context = ext4_dummy_context,
a7550b30
JK		 1287 .empty_dir = ext4_empty_dir,
1288 .max_namelen = ext4_max_namelen,
1289};
a7550b30
JK		 1290#endif
1291
ac27a0ec 1292#ifdef CONFIG_QUOTA
d6006186 1293static const char * const quotatypes[] = INITQFNAMES;
689c958c 1294#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1295
617ba13b
MC		 1296static int ext4_write_dquot(struct dquot *dquot);
1297static int ext4_acquire_dquot(struct dquot *dquot);
1298static int ext4_release_dquot(struct dquot *dquot);
1299static int ext4_mark_dquot_dirty(struct dquot *dquot);
1300static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1301static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1302 const struct path *path);
617ba13b
MC		 1303static int ext4_quota_on_mount(struct super_block *sb, int type);
1304static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1305 size_t len, loff_t off);
617ba13b 1306static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1307 const char *data, size_t len, loff_t off);
7c319d32
AK		 1308static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1309 unsigned int flags);
1310static int ext4_enable_quotas(struct super_block *sb);
8f0e8746 1311static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
ac27a0ec 1312
96c7e0d9
JK		 1313static struct dquot **ext4_get_dquots(struct inode *inode)
1314{
1315 return EXT4_I(inode)->i_dquot;
1316}
1317
61e225dc 1318static const struct dquot_operations ext4_quota_operations = {
7a9ca53a
TE		 1319 .get_reserved_space = ext4_get_reserved_space,
1320 .write_dquot = ext4_write_dquot,
1321 .acquire_dquot = ext4_acquire_dquot,
1322 .release_dquot = ext4_release_dquot,
1323 .mark_dirty = ext4_mark_dquot_dirty,
1324 .write_info = ext4_write_info,
1325 .alloc_dquot = dquot_alloc,
1326 .destroy_dquot = dquot_destroy,
1327 .get_projid = ext4_get_projid,
1328 .get_inode_usage = ext4_get_inode_usage,
1329 .get_next_id = ext4_get_next_id,
ac27a0ec
DK		 1330};
1331
0d54b217 1332static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1333 .quota_on = ext4_quota_on,
ca0e05e4 1334 .quota_off = ext4_quota_off,
287a8095 1335 .quota_sync = dquot_quota_sync,
0a240339 1336 .get_state = dquot_get_state,
287a8095
CH		 1337 .set_info = dquot_set_dqinfo,
1338 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1339 .set_dqblk = dquot_set_dqblk,
1340 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1341};
1342#endif
1343
ee9b6d61 1344static const struct super_operations ext4_sops = {
617ba13b
MC		 1345 .alloc_inode = ext4_alloc_inode,
1346 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1347 .write_inode = ext4_write_inode,
1348 .dirty_inode = ext4_dirty_inode,
7ff9c073 1349 .drop_inode = ext4_drop_inode,
0930fcc1 1350 .evict_inode = ext4_evict_inode,
617ba13b 1351 .put_super = ext4_put_super,
617ba13b 1352 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1353 .freeze_fs = ext4_freeze,
1354 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1355 .statfs = ext4_statfs,
1356 .remount_fs = ext4_remount,
617ba13b 1357 .show_options = ext4_show_options,
ac27a0ec 1358#ifdef CONFIG_QUOTA
617ba13b
MC		 1359 .quota_read = ext4_quota_read,
1360 .quota_write = ext4_quota_write,
96c7e0d9 1361 .get_dquots = ext4_get_dquots,
ac27a0ec 1362#endif
c39a7f84 1363 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1364};
1365
39655164 1366static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1367 .fh_to_dentry = ext4_fh_to_dentry,
1368 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1369 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1370};
1371
1372enum {
1373 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1374 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1375 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1376 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1377 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1378 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1379 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1380 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1381 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1382 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1383 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1384 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
49da9392 1385 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
327eaf73
TT		 1386 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1387 Opt_nowarn_on_error, Opt_mblk_io_submit,
670e9875 1388 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1449032b 1389 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1390 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1391 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1392 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1393 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
ac27a0ec
DK		 1394};
1395
a447c093 1396static const match_table_t tokens = {
ac27a0ec
DK		 1397 {Opt_bsd_df, "bsddf"},
1398 {Opt_minix_df, "minixdf"},
1399 {Opt_grpid, "grpid"},
1400 {Opt_grpid, "bsdgroups"},
1401 {Opt_nogrpid, "nogrpid"},
1402 {Opt_nogrpid, "sysvgroups"},
1403 {Opt_resgid, "resgid=%u"},
1404 {Opt_resuid, "resuid=%u"},
1405 {Opt_sb, "sb=%u"},
1406 {Opt_err_cont, "errors=continue"},
1407 {Opt_err_panic, "errors=panic"},
1408 {Opt_err_ro, "errors=remount-ro"},
1409 {Opt_nouid32, "nouid32"},
ac27a0ec 1410 {Opt_debug, "debug"},
72578c33
TT		 1411 {Opt_removed, "oldalloc"},
1412 {Opt_removed, "orlov"},
ac27a0ec
DK		 1413 {Opt_user_xattr, "user_xattr"},
1414 {Opt_nouser_xattr, "nouser_xattr"},
1415 {Opt_acl, "acl"},
1416 {Opt_noacl, "noacl"},
e3bb52ae 1417 {Opt_noload, "norecovery"},
5a916be1 1418 {Opt_noload, "noload"},
72578c33
TT		 1419 {Opt_removed, "nobh"},
1420 {Opt_removed, "bh"},
ac27a0ec 1421 {Opt_commit, "commit=%u"},
30773840
TT		 1422 {Opt_min_batch_time, "min_batch_time=%u"},
1423 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1424 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1425 {Opt_journal_path, "journal_path=%s"},
818d276c 1426 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1427 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1428 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1429 {Opt_abort, "abort"},
1430 {Opt_data_journal, "data=journal"},
1431 {Opt_data_ordered, "data=ordered"},
1432 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1433 {Opt_data_err_abort, "data_err=abort"},
1434 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1435 {Opt_offusrjquota, "usrjquota="},
1436 {Opt_usrjquota, "usrjquota=%s"},
1437 {Opt_offgrpjquota, "grpjquota="},
1438 {Opt_grpjquota, "grpjquota=%s"},
1439 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1440 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1441 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1442 {Opt_grpquota, "grpquota"},
1443 {Opt_noquota, "noquota"},
1444 {Opt_quota, "quota"},
1445 {Opt_usrquota, "usrquota"},
49da9392 1446 {Opt_prjquota, "prjquota"},
ac27a0ec 1447 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1448 {Opt_barrier, "barrier"},
1449 {Opt_nobarrier, "nobarrier"},
25ec56b5 1450 {Opt_i_version, "i_version"},
923ae0ff 1451 {Opt_dax, "dax"},
c9de560d 1452 {Opt_stripe, "stripe=%u"},
64769240 1453 {Opt_delalloc, "delalloc"},
327eaf73
TT		 1454 {Opt_warn_on_error, "warn_on_error"},
1455 {Opt_nowarn_on_error, "nowarn_on_error"},
a26f4992
TT		 1456 {Opt_lazytime, "lazytime"},
1457 {Opt_nolazytime, "nolazytime"},
670e9875 1458 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
dd919b98 1459 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1460 {Opt_removed, "mblk_io_submit"},
1461 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1462 {Opt_block_validity, "block_validity"},
1463 {Opt_noblock_validity, "noblock_validity"},
240799cd 1464 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1465 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1466 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1467 {Opt_auto_da_alloc, "auto_da_alloc"},
1468 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1469 {Opt_dioread_nolock, "dioread_nolock"},
1470 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1471 {Opt_discard, "discard"},
1472 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1473 {Opt_init_itable, "init_itable=%u"},
1474 {Opt_init_itable, "init_itable"},
1475 {Opt_noinit_itable, "noinit_itable"},
df981d03 1476 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1477 {Opt_test_dummy_encryption, "test_dummy_encryption"},
cdb7ee4c
TE		 1478 {Opt_nombcache, "nombcache"},
1479 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
c7198b9c
TT		 1480 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1481 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1482 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1483 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1484 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1485 {Opt_err, NULL},
ac27a0ec
DK		 1486};
1487
617ba13b 1488static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1489{
617ba13b 1490 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1491 char *options = (char *) *data;
1492
1493 if (!options || strncmp(options, "sb=", 3) != 0)
1494 return 1; /* Default location */
0b8e58a1 1495
ac27a0ec 1496 options += 3;
0b8e58a1 1497 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1498 sb_block = simple_strtoul(options, &options, 0);
1499 if (*options && *options != ',') {
4776004f 1500 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1501 (char *) *data);
1502 return 1;
1503 }
1504 if (*options == ',')
1505 options++;
1506 *data = (void *) options;
0b8e58a1 1507
ac27a0ec
DK		 1508 return sb_block;
1509}
1510
b3881f74 1511#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
d6006186
EB		 1512static const char deprecated_msg[] =
1513 "Mount option \"%s\" will be removed by %s\n"
437ca0fd 1514 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1515
56c50f11
DM		 1516#ifdef CONFIG_QUOTA
1517static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1518{
1519 struct ext4_sb_info *sbi = EXT4_SB(sb);
1520 char *qname;
03dafb5f 1521 int ret = -1;
56c50f11
DM		 1522
1523 if (sb_any_quota_loaded(sb) &&
1524 !sbi->s_qf_names[qtype]) {
1525 ext4_msg(sb, KERN_ERR,
1526 "Cannot change journaled "
1527 "quota options when quota turned on");
57f73c2c 1528 return -1;
56c50f11 1529 }
e2b911c5 1530 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1531 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1532 "ignored when QUOTA feature is enabled");
1533 return 1;
262b4662 1534 }
56c50f11
DM		 1535 qname = match_strdup(args);
1536 if (!qname) {
1537 ext4_msg(sb, KERN_ERR,
1538 "Not enough memory for storing quotafile name");
57f73c2c 1539 return -1;
56c50f11 1540 }
03dafb5f
CG		 1541 if (sbi->s_qf_names[qtype]) {
1542 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1543 ret = 1;
1544 else
1545 ext4_msg(sb, KERN_ERR,
1546 "%s quota file already specified",
1547 QTYPE2NAME(qtype));
1548 goto errout;
56c50f11 1549 }
03dafb5f 1550 if (strchr(qname, '/')) {
56c50f11
DM		 1551 ext4_msg(sb, KERN_ERR,
1552 "quotafile must be on filesystem root");
03dafb5f 1553 goto errout;
56c50f11 1554 }
03dafb5f 1555 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1556 set_opt(sb, QUOTA);
56c50f11 1557 return 1;
03dafb5f
CG		 1558errout:
1559 kfree(qname);
1560 return ret;
56c50f11
DM		 1561}
1562
1563static int clear_qf_name(struct super_block *sb, int qtype)
1564{
1565
1566 struct ext4_sb_info *sbi = EXT4_SB(sb);
1567
1568 if (sb_any_quota_loaded(sb) &&
1569 sbi->s_qf_names[qtype]) {
1570 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1571 " when quota turned on");
57f73c2c 1572 return -1;
56c50f11 1573 }
03dafb5f 1574 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1575 sbi->s_qf_names[qtype] = NULL;
1576 return 1;
1577}
1578#endif
1579
26092bf5
TT		 1580#define MOPT_SET 0x0001
1581#define MOPT_CLEAR 0x0002
1582#define MOPT_NOSUPPORT 0x0004
1583#define MOPT_EXPLICIT 0x0008
1584#define MOPT_CLEAR_ERR 0x0010
1585#define MOPT_GTE0 0x0020
ac27a0ec 1586#ifdef CONFIG_QUOTA
26092bf5
TT		 1587#define MOPT_Q 0
1588#define MOPT_QFMT 0x0040
1589#else
1590#define MOPT_Q MOPT_NOSUPPORT
1591#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1592#endif
26092bf5 1593#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1594#define MOPT_NO_EXT2 0x0100
1595#define MOPT_NO_EXT3 0x0200
1596#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1597#define MOPT_STRING 0x0400
26092bf5
TT		 1598
1599static const struct mount_opts {
1600 int token;
1601 int mount_opt;
1602 int flags;
1603} ext4_mount_opts[] = {
1604 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1605 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1606 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1607 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1608 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1609 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1610 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1611 MOPT_EXT4_ONLY | MOPT_SET},
1612 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1613 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1614 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1615 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1616 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1617 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1618 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1619 MOPT_EXT4_ONLY | MOPT_CLEAR},
327eaf73
TT		 1620 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1621 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
c6d3d56d
DW		 1622 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1623 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1624 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1625 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1626 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1627 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1628 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1629 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1630 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1631 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1632 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1633 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1634 MOPT_NO_EXT2},
8dc0aa8c 1635 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1636 MOPT_NO_EXT2},
26092bf5
TT		 1637 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1638 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1639 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1640 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1641 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1642 {Opt_commit, 0, MOPT_GTE0},
1643 {Opt_max_batch_time, 0, MOPT_GTE0},
1644 {Opt_min_batch_time, 0, MOPT_GTE0},
1645 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1646 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1647 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1648 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1649 {Opt_resuid, 0, MOPT_GTE0},
1650 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1651 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1652 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1653 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1654 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1655 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1656 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1657 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1658 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1659 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1660#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1661 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1662 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1663#else
26092bf5
TT		 1664 {Opt_acl, 0, MOPT_NOSUPPORT},
1665 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1666#endif
26092bf5
TT		 1667 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1668 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
670e9875 1669 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
26092bf5
TT		 1670 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1671 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1672 MOPT_SET | MOPT_Q},
1673 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1674 MOPT_SET | MOPT_Q},
49da9392
JK		 1675 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1676 MOPT_SET | MOPT_Q},
26092bf5 1677 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1678 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1679 MOPT_CLEAR | MOPT_Q},
26092bf5
TT		 1680 {Opt_usrjquota, 0, MOPT_Q},
1681 {Opt_grpjquota, 0, MOPT_Q},
1682 {Opt_offusrjquota, 0, MOPT_Q},
1683 {Opt_offgrpjquota, 0, MOPT_Q},
1684 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1685 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1686 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1687 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1688 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
cdb7ee4c 1689 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
26092bf5
TT		 1690 {Opt_err, 0, 0}
1691};
1692
1693static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1694 substring_t *args, unsigned long *journal_devnum,
1695 unsigned int *journal_ioprio, int is_remount)
1696{
1697 struct ext4_sb_info *sbi = EXT4_SB(sb);
1698 const struct mount_opts *m;
08cefc7a
EB		 1699 kuid_t uid;
1700 kgid_t gid;
26092bf5
TT		 1701 int arg = 0;
1702
57f73c2c
TT		 1703#ifdef CONFIG_QUOTA
1704 if (token == Opt_usrjquota)
1705 return set_qf_name(sb, USRQUOTA, &args[0]);
1706 else if (token == Opt_grpjquota)
1707 return set_qf_name(sb, GRPQUOTA, &args[0]);
1708 else if (token == Opt_offusrjquota)
1709 return clear_qf_name(sb, USRQUOTA);
1710 else if (token == Opt_offgrpjquota)
1711 return clear_qf_name(sb, GRPQUOTA);
1712#endif
26092bf5 1713 switch (token) {
f7048605
TT		 1714 case Opt_noacl:
1715 case Opt_nouser_xattr:
1716 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1717 break;
26092bf5
TT		 1718 case Opt_sb:
1719 return 1; /* handled by get_sb_block() */
1720 case Opt_removed:
5f3633e3 1721 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1722 return 1;
26092bf5
TT		 1723 case Opt_abort:
1724 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1725 return 1;
1726 case Opt_i_version:
357fdad0 1727 sb->s_flags |= SB_I_VERSION;
26092bf5 1728 return 1;
a26f4992 1729 case Opt_lazytime:
1751e8a6 1730 sb->s_flags |= SB_LAZYTIME;
a26f4992
TT		 1731 return 1;
1732 case Opt_nolazytime:
1751e8a6 1733 sb->s_flags &= ~SB_LAZYTIME;
a26f4992 1734 return 1;
26092bf5
TT		 1735 }
1736
5f3633e3
JK		 1737 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1738 if (token == m->token)
1739 break;
1740
1741 if (m->token == Opt_err) {
1742 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1743 "or missing value", opt);
1744 return -1;
1745 }
1746
8dc0aa8c
TT		 1747 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1748 ext4_msg(sb, KERN_ERR,
1749 "Mount option \"%s\" incompatible with ext2", opt);
1750 return -1;
1751 }
1752 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1753 ext4_msg(sb, KERN_ERR,
1754 "Mount option \"%s\" incompatible with ext3", opt);
1755 return -1;
1756 }
1757
ad4eec61 1758 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1759 return -1;
1760 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1761 return -1;
c93cf2d7
DM		 1762 if (m->flags & MOPT_EXPLICIT) {
1763 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1764 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1765 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1766 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1767 } else
1768 return -1;
1769 }
5f3633e3
JK		 1770 if (m->flags & MOPT_CLEAR_ERR)
1771 clear_opt(sb, ERRORS_MASK);
1772 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1773 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1774 "options when quota turned on");
1775 return -1;
1776 }
1777
1778 if (m->flags & MOPT_NOSUPPORT) {
1779 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1780 } else if (token == Opt_commit) {
1781 if (arg == 0)
1782 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1783 sbi->s_commit_interval = HZ * arg;
670e9875
TT		 1784 } else if (token == Opt_debug_want_extra_isize) {
1785 sbi->s_want_extra_isize = arg;
5f3633e3 1786 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1787 sbi->s_max_batch_time = arg;
1788 } else if (token == Opt_min_batch_time) {
1789 sbi->s_min_batch_time = arg;
1790 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1791 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1792 ext4_msg(sb, KERN_ERR,
1793 "EXT4-fs: inode_readahead_blks must be "
1794 "0 or a power of 2 smaller than 2^31");
26092bf5 1795 return -1;
5f3633e3
JK		 1796 }
1797 sbi->s_inode_readahead_blks = arg;
1798 } else if (token == Opt_init_itable) {
1799 set_opt(sb, INIT_INODE_TABLE);
1800 if (!args->from)
1801 arg = EXT4_DEF_LI_WAIT_MULT;
1802 sbi->s_li_wait_mult = arg;
1803 } else if (token == Opt_max_dir_size_kb) {
1804 sbi->s_max_dir_size_kb = arg;
1805 } else if (token == Opt_stripe) {
1806 sbi->s_stripe = arg;
1807 } else if (token == Opt_resuid) {
1808 uid = make_kuid(current_user_ns(), arg);
1809 if (!uid_valid(uid)) {
1810 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1811 return -1;
1812 }
5f3633e3
JK		 1813 sbi->s_resuid = uid;
1814 } else if (token == Opt_resgid) {
1815 gid = make_kgid(current_user_ns(), arg);
1816 if (!gid_valid(gid)) {
1817 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1818 return -1;
1819 }
1820 sbi->s_resgid = gid;
1821 } else if (token == Opt_journal_dev) {
1822 if (is_remount) {
1823 ext4_msg(sb, KERN_ERR,
1824 "Cannot specify journal on remount");
1825 return -1;
1826 }
1827 *journal_devnum = arg;
ad4eec61
ES		 1828 } else if (token == Opt_journal_path) {
1829 char *journal_path;
1830 struct inode *journal_inode;
1831 struct path path;
1832 int error;
1833
1834 if (is_remount) {
1835 ext4_msg(sb, KERN_ERR,
1836 "Cannot specify journal on remount");
1837 return -1;
1838 }
1839 journal_path = match_strdup(&args[0]);
1840 if (!journal_path) {
1841 ext4_msg(sb, KERN_ERR, "error: could not dup "
1842 "journal device string");
1843 return -1;
1844 }
1845
1846 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1847 if (error) {
1848 ext4_msg(sb, KERN_ERR, "error: could not find "
1849 "journal device path: error %d", error);
1850 kfree(journal_path);
1851 return -1;
1852 }
1853
2b0143b5 1854 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1855 if (!S_ISBLK(journal_inode->i_mode)) {
1856 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1857 "is not a block device", journal_path);
1858 path_put(&path);
1859 kfree(journal_path);
1860 return -1;
1861 }
1862
1863 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1864 path_put(&path);
1865 kfree(journal_path);
5f3633e3
JK		 1866 } else if (token == Opt_journal_ioprio) {
1867 if (arg > 7) {
1868 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1869 " (must be 0-7)");
1870 return -1;
1871 }
1872 *journal_ioprio =
1873 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1874 } else if (token == Opt_test_dummy_encryption) {
1875#ifdef CONFIG_EXT4_FS_ENCRYPTION
1876 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1877 ext4_msg(sb, KERN_WARNING,
1878 "Test dummy encryption mode enabled");
1879#else
1880 ext4_msg(sb, KERN_WARNING,
1881 "Test dummy encryption mount option ignored");
1882#endif
5f3633e3
JK		 1883 } else if (m->flags & MOPT_DATAJ) {
1884 if (is_remount) {
1885 if (!sbi->s_journal)
1886 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1887 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1888 ext4_msg(sb, KERN_ERR,
26092bf5 1889 "Cannot change data mode on remount");
26092bf5 1890 return -1;
ac27a0ec 1891 }
26092bf5 1892 } else {
5f3633e3
JK		 1893 clear_opt(sb, DATA_FLAGS);
1894 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1895 }
5f3633e3
JK		 1896#ifdef CONFIG_QUOTA
1897 } else if (m->flags & MOPT_QFMT) {
1898 if (sb_any_quota_loaded(sb) &&
1899 sbi->s_jquota_fmt != m->mount_opt) {
1900 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1901 "quota options when quota turned on");
1902 return -1;
1903 }
e2b911c5 1904 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1905 ext4_msg(sb, KERN_INFO,
1906 "Quota format mount options ignored "
262b4662 1907 "when QUOTA feature is enabled");
c325a67c 1908 return 1;
262b4662 1909 }
5f3633e3 1910 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 1911#endif
923ae0ff 1912 } else if (token == Opt_dax) {
ef83b6e8
DW		 1913#ifdef CONFIG_FS_DAX
1914 ext4_msg(sb, KERN_WARNING,
1915 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1916 sbi->s_mount_opt |= m->mount_opt;
1917#else
923ae0ff
RZ		 1918 ext4_msg(sb, KERN_INFO, "dax option not supported");
1919 return -1;
5f3633e3 1920#endif
7915a861
AN		 1921 } else if (token == Opt_data_err_abort) {
1922 sbi->s_mount_opt |= m->mount_opt;
1923 } else if (token == Opt_data_err_ignore) {
1924 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 1925 } else {
1926 if (!args->from)
1927 arg = 1;
1928 if (m->flags & MOPT_CLEAR)
1929 arg = !arg;
1930 else if (unlikely(!(m->flags & MOPT_SET))) {
1931 ext4_msg(sb, KERN_WARNING,
1932 "buggy handling of option %s", opt);
1933 WARN_ON(1);
1934 return -1;
1935 }
1936 if (arg != 0)
1937 sbi->s_mount_opt |= m->mount_opt;
1938 else
1939 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1940 }
5f3633e3 1941 return 1;
26092bf5
TT		 1942}
1943
1944static int parse_options(char *options, struct super_block *sb,
1945 unsigned long *journal_devnum,
1946 unsigned int *journal_ioprio,
1947 int is_remount)
1948{
1949 struct ext4_sb_info *sbi = EXT4_SB(sb);
1950 char *p;
1951 substring_t args[MAX_OPT_ARGS];
1952 int token;
1953
1954 if (!options)
1955 return 1;
1956
1957 while ((p = strsep(&options, ",")) != NULL) {
1958 if (!*p)
1959 continue;
1960 /*
1961 * Initialize args struct so we know whether arg was
1962 * found; some options take optional arguments.
1963 */
caecd0af 1964 args[0].to = args[0].from = NULL;
26092bf5
TT		 1965 token = match_token(p, tokens, args);
1966 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1967 journal_ioprio, is_remount) < 0)
1968 return 0;
ac27a0ec
DK		 1969 }
1970#ifdef CONFIG_QUOTA
49da9392
JK		 1971 /*
1972 * We do the test below only for project quotas. 'usrquota' and
1973 * 'grpquota' mount options are allowed even without quota feature
1974 * to support legacy quotas in quota files.
1975 */
1976 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1977 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1978 "Cannot enable project quota enforcement.");
1979 return 0;
1980 }
1981 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1982 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1983 clear_opt(sb, USRQUOTA);
ac27a0ec 1984
482a7425 1985 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1986 clear_opt(sb, GRPQUOTA);
ac27a0ec 1987
56c50f11 1988 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1989 ext4_msg(sb, KERN_ERR, "old and new quota "
1990 "format mixing");
ac27a0ec
DK		 1991 return 0;
1992 }
1993
1994 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1995 ext4_msg(sb, KERN_ERR, "journaled quota format "
1996 "not specified");
ac27a0ec
DK		 1997 return 0;
1998 }
ac27a0ec
DK		 1999 }
2000#endif
261cb20c
JK		 2001 if (test_opt(sb, DIOREAD_NOLOCK)) {
2002 int blocksize =
2003 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2004
09cbfeaf 2005 if (blocksize < PAGE_SIZE) {
261cb20c
JK		 2006 ext4_msg(sb, KERN_ERR, "can't mount with "
2007 "dioread_nolock if block size != PAGE_SIZE");
2008 return 0;
2009 }
2010 }
ac27a0ec
DK		 2011 return 1;
2012}
2013
2adf6da8
TT		 2014static inline void ext4_show_quota_options(struct seq_file *seq,
2015 struct super_block *sb)
2016{
2017#if defined(CONFIG_QUOTA)
2018 struct ext4_sb_info *sbi = EXT4_SB(sb);
2019
2020 if (sbi->s_jquota_fmt) {
2021 char *fmtname = "";
2022
2023 switch (sbi->s_jquota_fmt) {
2024 case QFMT_VFS_OLD:
2025 fmtname = "vfsold";
2026 break;
2027 case QFMT_VFS_V0:
2028 fmtname = "vfsv0";
2029 break;
2030 case QFMT_VFS_V1:
2031 fmtname = "vfsv1";
2032 break;
2033 }
2034 seq_printf(seq, ",jqfmt=%s", fmtname);
2035 }
2036
2037 if (sbi->s_qf_names[USRQUOTA])
a068acf2 2038 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 2039
2040 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 2041 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 2042#endif
2043}
2044
5a916be1
TT		 2045static const char *token2str(int token)
2046{
50df9fd5 2047 const struct match_token *t;
5a916be1
TT		 2048
2049 for (t = tokens; t->token != Opt_err; t++)
2050 if (t->token == token && !strchr(t->pattern, '='))
2051 break;
2052 return t->pattern;
2053}
2054
2adf6da8
TT		 2055/*
2056 * Show an option if
2057 * - it's set to a non-default value OR
2058 * - if the per-sb default is different from the global default
2059 */
66acdcf4
TT		 2060static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2061 int nodefs)
2adf6da8 2062{
2adf6da8
TT		 2063 struct ext4_sb_info *sbi = EXT4_SB(sb);
2064 struct ext4_super_block *es = sbi->s_es;
68afa7e0 2065 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
5a916be1 2066 const struct mount_opts *m;
66acdcf4 2067 char sep = nodefs ? '\n' : ',';
2adf6da8 2068
66acdcf4
TT		 2069#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2070#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2071
2072 if (sbi->s_sb_block != 1)
5a916be1
TT		 2073 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2074
2075 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2076 int want_set = m->flags & MOPT_SET;
2077 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2078 (m->flags & MOPT_CLEAR_ERR))
2079 continue;
68afa7e0 2080 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 2081 continue; /* skip if same as the default */
2082 if ((want_set &&
2083 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2084 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2085 continue; /* select Opt_noFoo vs Opt_Foo */
2086 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2087 }
5a916be1 2088
08cefc7a 2089 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2090 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2091 SEQ_OPTS_PRINT("resuid=%u",
2092 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2093 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2094 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2095 SEQ_OPTS_PRINT("resgid=%u",
2096 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2097 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2098 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2099 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2100 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2101 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2102 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2103 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2104 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2105 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2106 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2107 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2108 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2109 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
357fdad0 2110 if (sb->s_flags & SB_I_VERSION)
5a916be1 2111 SEQ_OPTS_PUTS("i_version");
66acdcf4 2112 if (nodefs || sbi->s_stripe)
5a916be1 2113 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
68afa7e0
TN		 2114 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2115 (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2116 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2117 SEQ_OPTS_PUTS("data=journal");
2118 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2119 SEQ_OPTS_PUTS("data=ordered");
2120 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2121 SEQ_OPTS_PUTS("data=writeback");
2122 }
66acdcf4
TT		 2123 if (nodefs ||
2124 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2125 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2126 sbi->s_inode_readahead_blks);
2adf6da8 2127
ceec0376 2128 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
66acdcf4 2129 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2130 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2131 if (nodefs || sbi->s_max_dir_size_kb)
2132 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2133 if (test_opt(sb, DATA_ERR_ABORT))
2134 SEQ_OPTS_PUTS("data_err=abort");
2adf6da8
TT		 2135
2136 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2137 return 0;
2138}
2139
66acdcf4
TT		 2140static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2141{
2142 return _ext4_show_options(seq, root->d_sb, 0);
2143}
2144
ebd173be 2145int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2146{
2147 struct super_block *sb = seq->private;
2148 int rc;
2149
bc98a42c 2150 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
66acdcf4
TT		 2151 rc = _ext4_show_options(seq, sb, 1);
2152 seq_puts(seq, "\n");
2153 return rc;
2154}
2155
617ba13b 2156static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2157 int read_only)
2158{
617ba13b 2159 struct ext4_sb_info *sbi = EXT4_SB(sb);
c89128a0 2160 int err = 0;
ac27a0ec 2161
617ba13b 2162 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2163 ext4_msg(sb, KERN_ERR, "revision level too high, "
2164 "forcing read-only mode");
c89128a0 2165 err = -EROFS;
ac27a0ec
DK		 2166 }
2167 if (read_only)
281b5995 2168 goto done;
617ba13b 2169 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2170 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2171 "running e2fsck is recommended");
c8b459f4 2172 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2173 ext4_msg(sb, KERN_WARNING,
2174 "warning: mounting fs with errors, "
2175 "running e2fsck is recommended");
ed3ce80a 2176 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2177 le16_to_cpu(es->s_mnt_count) >=
2178 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2179 ext4_msg(sb, KERN_WARNING,
2180 "warning: maximal mount count reached, "
2181 "running e2fsck is recommended");
ac27a0ec
DK		 2182 else if (le32_to_cpu(es->s_checkinterval) &&
2183 (le32_to_cpu(es->s_lastcheck) +
2184 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 2185 ext4_msg(sb, KERN_WARNING,
2186 "warning: checktime reached, "
2187 "running e2fsck is recommended");
0b8e58a1 2188 if (!sbi->s_journal)
0390131b 2189 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2190 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2191 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2192 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 2193 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 2194 ext4_update_dynamic_rev(sb);
0390131b 2195 if (sbi->s_journal)
e2b911c5 2196 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2197
c89128a0 2198 err = ext4_commit_super(sb, 1);
281b5995 2199done:
ac27a0ec 2200 if (test_opt(sb, DEBUG))
a9df9a49 2201 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2202 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2203 sb->s_blocksize,
2204 sbi->s_groups_count,
617ba13b
MC		 2205 EXT4_BLOCKS_PER_GROUP(sb),
2206 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2207 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2208
7abc52c2 2209 cleancache_init_fs(sb);
c89128a0 2210 return err;
ac27a0ec
DK		 2211}
2212
117fff10
TT		 2213int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2214{
2215 struct ext4_sb_info *sbi = EXT4_SB(sb);
2216 struct flex_groups *new_groups;
2217 int size;
2218
2219 if (!sbi->s_log_groups_per_flex)
2220 return 0;
2221
2222 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2223 if (size <= sbi->s_flex_groups_allocated)
2224 return 0;
2225
2226 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
a7c3e901 2227 new_groups = kvzalloc(size, GFP_KERNEL);
117fff10
TT		 2228 if (!new_groups) {
2229 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2230 size / (int) sizeof(struct flex_groups));
2231 return -ENOMEM;
2232 }
2233
2234 if (sbi->s_flex_groups) {
2235 memcpy(new_groups, sbi->s_flex_groups,
2236 (sbi->s_flex_groups_allocated *
2237 sizeof(struct flex_groups)));
b93b41d4 2238 kvfree(sbi->s_flex_groups);
117fff10
TT		 2239 }
2240 sbi->s_flex_groups = new_groups;
2241 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2242 return 0;
2243}
2244
772cb7c8
JS		 2245static int ext4_fill_flex_info(struct super_block *sb)
2246{
2247 struct ext4_sb_info *sbi = EXT4_SB(sb);
2248 struct ext4_group_desc *gdp = NULL;
772cb7c8 2249 ext4_group_t flex_group;
117fff10 2250 int i, err;
772cb7c8 2251
503358ae 2252 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2253 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2254 sbi->s_log_groups_per_flex = 0;
2255 return 1;
2256 }
2257
117fff10
TT		 2258 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2259 if (err)
9933fc0a 2260 goto failed;
772cb7c8 2261
772cb7c8 2262 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2263 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2264
2265 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2266 atomic_add(ext4_free_inodes_count(sb, gdp),
2267 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2268 atomic64_add(ext4_free_group_clusters(sb, gdp),
2269 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2270 atomic_add(ext4_used_dirs_count(sb, gdp),
2271 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2272 }
2273
2274 return 1;
2275failed:
2276 return 0;
2277}
2278
e2b911c5 2279static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2280 struct ext4_group_desc *gdp)
717d50e4 2281{
b47820ed 2282 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2283 __u16 crc = 0;
feb0ab32 2284 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2285 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2286
9aa5d32b 2287 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2288 /* Use new metadata_csum algorithm */
feb0ab32 2289 __u32 csum32;
b47820ed 2290 __u16 dummy_csum = 0;
feb0ab32 2291
feb0ab32
DW		 2292 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2293 sizeof(le_group));
b47820ed
DJ		 2294 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2295 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2296 sizeof(dummy_csum));
2297 offset += sizeof(dummy_csum);
2298 if (offset < sbi->s_desc_size)
2299 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2300 sbi->s_desc_size - offset);
feb0ab32
DW		 2301
2302 crc = csum32 & 0xFFFF;
2303 goto out;
717d50e4
AD		 2304 }
2305
feb0ab32 2306 /* old crc16 code */
e2b911c5 2307 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2308 return 0;
2309
feb0ab32
DW		 2310 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2311 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2312 crc = crc16(crc, (__u8 *)gdp, offset);
2313 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2314 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2315 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2316 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2317 crc = crc16(crc, (__u8 *)gdp + offset,
2318 le16_to_cpu(sbi->s_es->s_desc_size) -
2319 offset);
2320
2321out:
717d50e4
AD		 2322 return cpu_to_le16(crc);
2323}
2324
feb0ab32 2325int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2326 struct ext4_group_desc *gdp)
2327{
feb0ab32 2328 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2329 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2330 return 0;
2331
2332 return 1;
2333}
2334
feb0ab32
DW		 2335void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2336 struct ext4_group_desc *gdp)
2337{
2338 if (!ext4_has_group_desc_csum(sb))
2339 return;
e2b911c5 2340 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2341}
2342
ac27a0ec 2343/* Called at mount-time, super-block is locked */
bfff6873 2344static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2345 ext4_fsblk_t sb_block,
bfff6873 2346 ext4_group_t *first_not_zeroed)
ac27a0ec 2347{
617ba13b
MC		 2348 struct ext4_sb_info *sbi = EXT4_SB(sb);
2349 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2350 ext4_fsblk_t last_block;
77260807 2351 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0) + 1;
bd81d8ee
LV		 2352 ext4_fsblk_t block_bitmap;
2353 ext4_fsblk_t inode_bitmap;
2354 ext4_fsblk_t inode_table;
ce421581 2355 int flexbg_flag = 0;
bfff6873 2356 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2357
e2b911c5 2358 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2359 flexbg_flag = 1;
2360
af5bc92d 2361 ext4_debug("Checking group descriptors");
ac27a0ec 2362
197cd65a
AM		 2363 for (i = 0; i < sbi->s_groups_count; i++) {
2364 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2365
ce421581 2366 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2367 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2368 else
2369 last_block = first_block +
617ba13b 2370 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2371
bfff6873
LC		 2372 if ((grp == sbi->s_groups_count) &&
2373 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2374 grp = i;
2375
8fadc143 2376 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2377 if (block_bitmap == sb_block) {
2378 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2379 "Block bitmap for group %u overlaps "
2380 "superblock", i);
18db4b4e
TT		 2381 if (!sb_rdonly(sb))
2382 return 0;
829fa70d 2383 }
77260807
TT		 2384 if (block_bitmap >= sb_block + 1 &&
2385 block_bitmap <= last_bg_block) {
2386 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2387 "Block bitmap for group %u overlaps "
2388 "block group descriptors", i);
2389 if (!sb_rdonly(sb))
2390 return 0;
2391 }
2b2d6d01 2392 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2393 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2394 "Block bitmap for group %u not in group "
b31e1552 2395 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2396 return 0;
2397 }
8fadc143 2398 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2399 if (inode_bitmap == sb_block) {
2400 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2401 "Inode bitmap for group %u overlaps "
2402 "superblock", i);
18db4b4e
TT		 2403 if (!sb_rdonly(sb))
2404 return 0;
829fa70d 2405 }
77260807
TT		 2406 if (inode_bitmap >= sb_block + 1 &&
2407 inode_bitmap <= last_bg_block) {
2408 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2409 "Inode bitmap for group %u overlaps "
2410 "block group descriptors", i);
2411 if (!sb_rdonly(sb))
2412 return 0;
2413 }
2b2d6d01 2414 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2415 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2416 "Inode bitmap for group %u not in group "
b31e1552 2417 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2418 return 0;
2419 }
8fadc143 2420 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2421 if (inode_table == sb_block) {
2422 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2423 "Inode table for group %u overlaps "
2424 "superblock", i);
18db4b4e
TT		 2425 if (!sb_rdonly(sb))
2426 return 0;
829fa70d 2427 }
77260807
TT		 2428 if (inode_table >= sb_block + 1 &&
2429 inode_table <= last_bg_block) {
2430 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2431 "Inode table for group %u overlaps "
2432 "block group descriptors", i);
2433 if (!sb_rdonly(sb))
2434 return 0;
2435 }
bd81d8ee 2436 if (inode_table < first_block ||
2b2d6d01 2437 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2438 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2439 "Inode table for group %u not in group "
b31e1552 2440 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2441 return 0;
2442 }
955ce5f5 2443 ext4_lock_group(sb, i);
feb0ab32 2444 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2445 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2446 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2447 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2448 gdp)), le16_to_cpu(gdp->bg_checksum));
bc98a42c 2449 if (!sb_rdonly(sb)) {
955ce5f5 2450 ext4_unlock_group(sb, i);
8a266467 2451 return 0;
7ee1ec4c 2452 }
717d50e4 2453 }
955ce5f5 2454 ext4_unlock_group(sb, i);
ce421581
JS		 2455 if (!flexbg_flag)
2456 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2457 }
bfff6873
LC		 2458 if (NULL != first_not_zeroed)
2459 *first_not_zeroed = grp;
ac27a0ec
DK		 2460 return 1;
2461}
2462
617ba13b 2463/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2464 * the superblock) which were deleted from all directories, but held open by
2465 * a process at the time of a crash. We walk the list and try to delete these
2466 * inodes at recovery time (only with a read-write filesystem).
2467 *
2468 * In order to keep the orphan inode chain consistent during traversal (in
2469 * case of crash during recovery), we link each inode into the superblock
2470 * orphan list_head and handle it the same way as an inode deletion during
2471 * normal operation (which journals the operations for us).
2472 *
2473 * We only do an iget() and an iput() on each inode, which is very safe if we
2474 * accidentally point at an in-use or already deleted inode. The worst that
2475 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2476 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2477 * e2fsck was run on this filesystem, and it must have already done the orphan
2478 * inode cleanup for us, so we can safely abort without any further action.
2479 */
2b2d6d01
TT		 2480static void ext4_orphan_cleanup(struct super_block *sb,
2481 struct ext4_super_block *es)
ac27a0ec
DK		 2482{
2483 unsigned int s_flags = sb->s_flags;
2c98eb5e 2484 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec 2485#ifdef CONFIG_QUOTA
95f1fda4 2486 int quota_update = 0;
ac27a0ec
DK		 2487 int i;
2488#endif
2489 if (!es->s_last_orphan) {
2490 jbd_debug(4, "no orphan inodes to clean up\n");
2491 return;
2492 }
2493
a8f48a95 2494 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2495 ext4_msg(sb, KERN_ERR, "write access "
2496 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2497 return;
2498 }
2499
d39195c3
AG		 2500 /* Check if feature set would not allow a r/w mount */
2501 if (!ext4_feature_set_ok(sb, 0)) {
2502 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2503 "unknown ROCOMPAT features");
2504 return;
2505 }
2506
617ba13b 2507 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6 2508 /* don't clear list on RO mount w/ errors */
1751e8a6 2509 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
84474976 2510 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2511 "clearing orphan list.\n");
c25f9bc6
ES		 2512 es->s_last_orphan = 0;
2513 }
ac27a0ec
DK		 2514 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2515 return;
2516 }
2517
1751e8a6 2518 if (s_flags & SB_RDONLY) {
b31e1552 2519 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1751e8a6 2520 sb->s_flags &= ~SB_RDONLY;
ac27a0ec
DK		 2521 }
2522#ifdef CONFIG_QUOTA
2523 /* Needed for iput() to work correctly and not trash data */
1751e8a6 2524 sb->s_flags |= SB_ACTIVE;
95f1fda4 2525
2526 /*
2527 * Turn on quotas which were not enabled for read-only mounts if
2528 * filesystem has quota feature, so that they are updated correctly.
2529 */
1751e8a6 2530 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
95f1fda4 2531 int ret = ext4_enable_quotas(sb);
2532
2533 if (!ret)
2534 quota_update = 1;
2535 else
2536 ext4_msg(sb, KERN_ERR,
2537 "Cannot turn on quotas: error %d", ret);
2538 }
2539
2540 /* Turn on journaled quotas used for old sytle */
a2d4a646 2541 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2542 if (EXT4_SB(sb)->s_qf_names[i]) {
2543 int ret = ext4_quota_on_mount(sb, i);
95f1fda4 2544
2545 if (!ret)
2546 quota_update = 1;
2547 else
b31e1552
ES		 2548 ext4_msg(sb, KERN_ERR,
2549 "Cannot turn on journaled "
95f1fda4 2550 "quota: type %d: error %d", i, ret);
ac27a0ec
DK		 2551 }
2552 }
2553#endif
2554
2555 while (es->s_last_orphan) {
2556 struct inode *inode;
2557
c65d5c6c
VN		 2558 /*
2559 * We may have encountered an error during cleanup; if
2560 * so, skip the rest.
2561 */
2562 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2563 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2564 es->s_last_orphan = 0;
2565 break;
2566 }
2567
97bd42b9
JB		 2568 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2569 if (IS_ERR(inode)) {
ac27a0ec
DK		 2570 es->s_last_orphan = 0;
2571 break;
2572 }
2573
617ba13b 2574 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2575 dquot_initialize(inode);
ac27a0ec 2576 if (inode->i_nlink) {
566370a2
PT		 2577 if (test_opt(sb, DEBUG))
2578 ext4_msg(sb, KERN_DEBUG,
2579 "%s: truncating inode %lu to %lld bytes",
2580 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2581 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2582 inode->i_ino, inode->i_size);
5955102c 2583 inode_lock(inode);
55f252c9 2584 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 2585 ret = ext4_truncate(inode);
2586 if (ret)
2587 ext4_std_error(inode->i_sb, ret);
5955102c 2588 inode_unlock(inode);
ac27a0ec
DK		 2589 nr_truncates++;
2590 } else {
566370a2
PT		 2591 if (test_opt(sb, DEBUG))
2592 ext4_msg(sb, KERN_DEBUG,
2593 "%s: deleting unreferenced inode %lu",
2594 __func__, inode->i_ino);
ac27a0ec
DK		 2595 jbd_debug(2, "deleting unreferenced inode %lu\n",
2596 inode->i_ino);
2597 nr_orphans++;
2598 }
2599 iput(inode); /* The delete magic happens here! */
2600 }
2601
2b2d6d01 2602#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2603
2604 if (nr_orphans)
b31e1552
ES		 2605 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2606 PLURAL(nr_orphans));
ac27a0ec 2607 if (nr_truncates)
b31e1552
ES		 2608 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2609 PLURAL(nr_truncates));
ac27a0ec 2610#ifdef CONFIG_QUOTA
95f1fda4 2611 /* Turn off quotas if they were enabled for orphan cleanup */
2612 if (quota_update) {
2613 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2614 if (sb_dqopt(sb)->files[i])
2615 dquot_quota_off(sb, i);
2616 }
ac27a0ec
DK		 2617 }
2618#endif
1751e8a6 2619 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
ac27a0ec 2620}
0b8e58a1 2621
cd2291a4
ES		 2622/*
2623 * Maximal extent format file size.
2624 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2625 * extent format containers, within a sector_t, and within i_blocks
2626 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2627 * so that won't be a limiting factor.
2628 *
f17722f9
LC		 2629 * However there is other limiting factor. We do store extents in the form
2630 * of starting block and length, hence the resulting length of the extent
2631 * covering maximum file size must fit into on-disk format containers as
2632 * well. Given that length is always by 1 unit bigger than max unit (because
2633 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2634 *
cd2291a4
ES		 2635 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2636 */
f287a1a5 2637static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2638{
2639 loff_t res;
2640 loff_t upper_limit = MAX_LFS_FILESIZE;
2641
2642 /* small i_blocks in vfs inode? */
f287a1a5 2643 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2644 /*
90c699a9 2645 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2646 * i_block represent total blocks in 512 bytes
2647 * 32 == size of vfs inode i_blocks * 8
2648 */
2649 upper_limit = (1LL << 32) - 1;
2650
2651 /* total blocks in file system block size */
2652 upper_limit >>= (blkbits - 9);
2653 upper_limit <<= blkbits;
2654 }
2655
f17722f9
LC		 2656 /*
2657 * 32-bit extent-start container, ee_block. We lower the maxbytes
2658 * by one fs block, so ee_len can cover the extent of maximum file
2659 * size
2660 */
2661 res = (1LL << 32) - 1;
cd2291a4 2662 res <<= blkbits;
cd2291a4
ES		 2663
2664 /* Sanity check against vm- & vfs- imposed limits */
2665 if (res > upper_limit)
2666 res = upper_limit;
2667
2668 return res;
2669}
ac27a0ec 2670
ac27a0ec 2671/*
cd2291a4 2672 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2673 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2674 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2675 */
f287a1a5 2676static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2677{
617ba13b 2678 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2679 int meta_blocks;
2680 loff_t upper_limit;
0b8e58a1
AD		 2681 /* This is calculated to be the largest file size for a dense, block
2682 * mapped file such that the file's total number of 512-byte sectors,
2683 * including data and all indirect blocks, does not exceed (2^48 - 1).
2684 *
2685 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2686 * number of 512-byte sectors of the file.
0fc1b451
AK		 2687 */
2688
f287a1a5 2689 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2690 /*
90c699a9 2691 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2692 * the inode i_block field represents total file blocks in
2693 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2694 */
2695 upper_limit = (1LL << 32) - 1;
2696
2697 /* total blocks in file system block size */
2698 upper_limit >>= (bits - 9);
2699
2700 } else {
8180a562
AK		 2701 /*
2702 * We use 48 bit ext4_inode i_blocks
2703 * With EXT4_HUGE_FILE_FL set the i_blocks
2704 * represent total number of blocks in
2705 * file system block size
2706 */
0fc1b451
AK		 2707 upper_limit = (1LL << 48) - 1;
2708
0fc1b451
AK		 2709 }
2710
2711 /* indirect blocks */
2712 meta_blocks = 1;
2713 /* double indirect blocks */
2714 meta_blocks += 1 + (1LL << (bits-2));
2715 /* tripple indirect blocks */
2716 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2717
2718 upper_limit -= meta_blocks;
2719 upper_limit <<= bits;
ac27a0ec
DK		 2720
2721 res += 1LL << (bits-2);
2722 res += 1LL << (2*(bits-2));
2723 res += 1LL << (3*(bits-2));
2724 res <<= bits;
2725 if (res > upper_limit)
2726 res = upper_limit;
0fc1b451
AK		 2727
2728 if (res > MAX_LFS_FILESIZE)
2729 res = MAX_LFS_FILESIZE;
2730
ac27a0ec
DK		 2731 return res;
2732}
2733
617ba13b 2734static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2735 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2736{
617ba13b 2737 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2738 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2739 int has_super = 0;
2740
2741 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2742
e2b911c5 2743 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2744 return logical_sb_block + nr + 1;
ac27a0ec 2745 bg = sbi->s_desc_per_block * nr;
617ba13b 2746 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2747 has_super = 1;
0b8e58a1 2748
bd63f6b0
DW		 2749 /*
2750 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2751 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2752 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2753 * compensate.
2754 */
2755 if (sb->s_blocksize == 1024 && nr == 0 &&
49598e04 2756 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
bd63f6b0
DW		 2757 has_super++;
2758
617ba13b 2759 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2760}
2761
c9de560d
AT		 2762/**
2763 * ext4_get_stripe_size: Get the stripe size.
2764 * @sbi: In memory super block info
2765 *
2766 * If we have specified it via mount option, then
2767 * use the mount option value. If the value specified at mount time is
2768 * greater than the blocks per group use the super block value.
2769 * If the super block value is greater than blocks per group return 0.
2770 * Allocator needs it be less than blocks per group.
2771 *
2772 */
2773static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2774{
2775 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2776 unsigned long stripe_width =
2777 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2778 int ret;
c9de560d
AT		 2779
2780 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 2781 ret = sbi->s_stripe;
5469d7c3 2782 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 2783 ret = stripe_width;
5469d7c3 2784 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 2785 ret = stride;
2786 else
2787 ret = 0;
c9de560d 2788
3eb08658
DE		 2789 /*
2790 * If the stripe width is 1, this makes no sense and
2791 * we set it to 0 to turn off stripe handling code.
2792 */
2793 if (ret <= 1)
2794 ret = 0;
c9de560d 2795
3eb08658 2796 return ret;
c9de560d 2797}
ac27a0ec 2798
a13fb1a4
ES		 2799/*
2800 * Check whether this filesystem can be mounted based on
2801 * the features present and the RDONLY/RDWR mount requested.
2802 * Returns 1 if this filesystem can be mounted as requested,
2803 * 0 if it cannot be.
2804 */
2805static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2806{
e2b911c5 2807 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2808 ext4_msg(sb, KERN_ERR,
2809 "Couldn't mount because of "
2810 "unsupported optional features (%x)",
2811 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2812 ~EXT4_FEATURE_INCOMPAT_SUPP));
2813 return 0;
2814 }
2815
2816 if (readonly)
2817 return 1;
2818
e2b911c5 2819 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b 2820 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
1751e8a6 2821 sb->s_flags |= SB_RDONLY;
2cb5cc8b
DW		 2822 return 1;
2823 }
2824
a13fb1a4 2825 /* Check that feature set is OK for a read-write mount */
e2b911c5 2826 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2827 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2828 "unsupported optional features (%x)",
2829 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2830 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2831 return 0;
2832 }
2833 /*
2834 * Large file size enabled file system can only be mounted
2835 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2836 */
e2b911c5 2837 if (ext4_has_feature_huge_file(sb)) {
a13fb1a4
ES		 2838 if (sizeof(blkcnt_t) < sizeof(u64)) {
2839 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2840 "cannot be mounted RDWR without "
2841 "CONFIG_LBDAF");
2842 return 0;
2843 }
2844 }
e2b911c5 2845 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2846 ext4_msg(sb, KERN_ERR,
2847 "Can't support bigalloc feature without "
2848 "extents feature\n");
2849 return 0;
2850 }
7c319d32
AK		 2851
2852#ifndef CONFIG_QUOTA
e2b911c5 2853 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2854 ext4_msg(sb, KERN_ERR,
2855 "Filesystem with quota feature cannot be mounted RDWR "
2856 "without CONFIG_QUOTA");
2857 return 0;
2858 }
689c958c
LX		 2859 if (ext4_has_feature_project(sb) && !readonly) {
2860 ext4_msg(sb, KERN_ERR,
2861 "Filesystem with project quota feature cannot be mounted RDWR "
2862 "without CONFIG_QUOTA");
2863 return 0;
2864 }
7c319d32 2865#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2866 return 1;
2867}
2868
66e61a9e
TT		 2869/*
2870 * This function is called once a day if we have errors logged
2871 * on the file system
2872 */
235699a8 2873static void print_daily_error_info(struct timer_list *t)
66e61a9e 2874{
235699a8
KC		 2875 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
2876 struct super_block *sb = sbi->s_sb;
2877 struct ext4_super_block *es = sbi->s_es;
66e61a9e
TT		 2878
2879 if (es->s_error_count)
ae0f78de
TT		 2880 /* fsck newer than v1.41.13 is needed to clean this condition. */
2881 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2882 le32_to_cpu(es->s_error_count));
2883 if (es->s_first_error_time) {
ae0f78de 2884 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2885 sb->s_id, le32_to_cpu(es->s_first_error_time),
2886 (int) sizeof(es->s_first_error_func),
2887 es->s_first_error_func,
2888 le32_to_cpu(es->s_first_error_line));
2889 if (es->s_first_error_ino)
651e1c3b 2890 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2891 le32_to_cpu(es->s_first_error_ino));
2892 if (es->s_first_error_block)
651e1c3b 2893 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2894 le64_to_cpu(es->s_first_error_block));
651e1c3b 2895 printk(KERN_CONT "\n");
66e61a9e
TT		 2896 }
2897 if (es->s_last_error_time) {
ae0f78de 2898 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2899 sb->s_id, le32_to_cpu(es->s_last_error_time),
2900 (int) sizeof(es->s_last_error_func),
2901 es->s_last_error_func,
2902 le32_to_cpu(es->s_last_error_line));
2903 if (es->s_last_error_ino)
651e1c3b 2904 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2905 le32_to_cpu(es->s_last_error_ino));
2906 if (es->s_last_error_block)
651e1c3b 2907 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2908 le64_to_cpu(es->s_last_error_block));
651e1c3b 2909 printk(KERN_CONT "\n");
66e61a9e
TT		 2910 }
2911 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2912}
2913
bfff6873
LC		 2914/* Find next suitable group and run ext4_init_inode_table */
2915static int ext4_run_li_request(struct ext4_li_request *elr)
2916{
2917 struct ext4_group_desc *gdp = NULL;
2918 ext4_group_t group, ngroups;
2919 struct super_block *sb;
2920 unsigned long timeout = 0;
2921 int ret = 0;
2922
2923 sb = elr->lr_super;
2924 ngroups = EXT4_SB(sb)->s_groups_count;
2925
2926 for (group = elr->lr_next_group; group < ngroups; group++) {
2927 gdp = ext4_get_group_desc(sb, group, NULL);
2928 if (!gdp) {
2929 ret = 1;
2930 break;
2931 }
2932
2933 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2934 break;
2935 }
2936
7f511862 2937 if (group >= ngroups)
bfff6873
LC		 2938 ret = 1;
2939
2940 if (!ret) {
2941 timeout = jiffies;
2942 ret = ext4_init_inode_table(sb, group,
2943 elr->lr_timeout ? 0 : 1);
2944 if (elr->lr_timeout == 0) {
51ce6511
LC		 2945 timeout = (jiffies - timeout) *
2946 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2947 elr->lr_timeout = timeout;
2948 }
2949 elr->lr_next_sched = jiffies + elr->lr_timeout;
2950 elr->lr_next_group = group + 1;
2951 }
bfff6873
LC		 2952 return ret;
2953}
2954
2955/*
2956 * Remove lr_request from the list_request and free the
4ed5c033 2957 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2958 */
2959static void ext4_remove_li_request(struct ext4_li_request *elr)
2960{
2961 struct ext4_sb_info *sbi;
2962
2963 if (!elr)
2964 return;
2965
2966 sbi = elr->lr_sbi;
2967
2968 list_del(&elr->lr_request);
2969 sbi->s_li_request = NULL;
2970 kfree(elr);
2971}
2972
2973static void ext4_unregister_li_request(struct super_block *sb)
2974{
1bb933fb
LC		 2975 mutex_lock(&ext4_li_mtx);
2976 if (!ext4_li_info) {
2977 mutex_unlock(&ext4_li_mtx);
bfff6873 2978 return;
1bb933fb 2979 }
bfff6873
LC		 2980
2981 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2982 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2983 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2984 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2985}
2986
8f1f7453
ES		 2987static struct task_struct *ext4_lazyinit_task;
2988
bfff6873
LC		 2989/*
2990 * This is the function where ext4lazyinit thread lives. It walks
2991 * through the request list searching for next scheduled filesystem.
2992 * When such a fs is found, run the lazy initialization request
2993 * (ext4_rn_li_request) and keep track of the time spend in this
2994 * function. Based on that time we compute next schedule time of
2995 * the request. When walking through the list is complete, compute
2996 * next waking time and put itself into sleep.
2997 */
2998static int ext4_lazyinit_thread(void *arg)
2999{
3000 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3001 struct list_head *pos, *n;
3002 struct ext4_li_request *elr;
4ed5c033 3003 unsigned long next_wakeup, cur;
bfff6873
LC		 3004
3005 BUG_ON(NULL == eli);
3006
bfff6873
LC		 3007cont_thread:
3008 while (true) {
3009 next_wakeup = MAX_JIFFY_OFFSET;
3010
3011 mutex_lock(&eli->li_list_mtx);
3012 if (list_empty(&eli->li_request_list)) {
3013 mutex_unlock(&eli->li_list_mtx);
3014 goto exit_thread;
3015 }
bfff6873 3016 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 3017 int err = 0;
3018 int progress = 0;
bfff6873
LC		 3019 elr = list_entry(pos, struct ext4_li_request,
3020 lr_request);
3021
e22834f0
DM		 3022 if (time_before(jiffies, elr->lr_next_sched)) {
3023 if (time_before(elr->lr_next_sched, next_wakeup))
3024 next_wakeup = elr->lr_next_sched;
3025 continue;
3026 }
3027 if (down_read_trylock(&elr->lr_super->s_umount)) {
3028 if (sb_start_write_trylock(elr->lr_super)) {
3029 progress = 1;
3030 /*
3031 * We hold sb->s_umount, sb can not
3032 * be removed from the list, it is
3033 * now safe to drop li_list_mtx
3034 */
3035 mutex_unlock(&eli->li_list_mtx);
3036 err = ext4_run_li_request(elr);
3037 sb_end_write(elr->lr_super);
3038 mutex_lock(&eli->li_list_mtx);
3039 n = pos->next;
b2c78cd0 3040 }
e22834f0
DM		 3041 up_read((&elr->lr_super->s_umount));
3042 }
3043 /* error, remove the lazy_init job */
3044 if (err) {
3045 ext4_remove_li_request(elr);
3046 continue;
3047 }
3048 if (!progress) {
3049 elr->lr_next_sched = jiffies +
3050 (prandom_u32()
3051 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 3052 }
bfff6873
LC		 3053 if (time_before(elr->lr_next_sched, next_wakeup))
3054 next_wakeup = elr->lr_next_sched;
3055 }
3056 mutex_unlock(&eli->li_list_mtx);
3057
a0acae0e 3058 try_to_freeze();
bfff6873 3059
4ed5c033
LC		 3060 cur = jiffies;
3061 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3062 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3063 cond_resched();
3064 continue;
3065 }
3066
4ed5c033
LC		 3067 schedule_timeout_interruptible(next_wakeup - cur);
3068
8f1f7453
ES		 3069 if (kthread_should_stop()) {
3070 ext4_clear_request_list();
3071 goto exit_thread;
3072 }
bfff6873
LC		 3073 }
3074
3075exit_thread:
3076 /*
3077 * It looks like the request list is empty, but we need
3078 * to check it under the li_list_mtx lock, to prevent any
3079 * additions into it, and of course we should lock ext4_li_mtx
3080 * to atomically free the list and ext4_li_info, because at
3081 * this point another ext4 filesystem could be registering
3082 * new one.
3083 */
3084 mutex_lock(&ext4_li_mtx);
3085 mutex_lock(&eli->li_list_mtx);
3086 if (!list_empty(&eli->li_request_list)) {
3087 mutex_unlock(&eli->li_list_mtx);
3088 mutex_unlock(&ext4_li_mtx);
3089 goto cont_thread;
3090 }
3091 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3092 kfree(ext4_li_info);
3093 ext4_li_info = NULL;
3094 mutex_unlock(&ext4_li_mtx);
3095
3096 return 0;
3097}
3098
3099static void ext4_clear_request_list(void)
3100{
3101 struct list_head *pos, *n;
3102 struct ext4_li_request *elr;
3103
3104 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3105 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3106 elr = list_entry(pos, struct ext4_li_request,
3107 lr_request);
3108 ext4_remove_li_request(elr);
3109 }
3110 mutex_unlock(&ext4_li_info->li_list_mtx);
3111}
3112
3113static int ext4_run_lazyinit_thread(void)
3114{
8f1f7453
ES		 3115 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3116 ext4_li_info, "ext4lazyinit");
3117 if (IS_ERR(ext4_lazyinit_task)) {
3118 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3119 ext4_clear_request_list();
bfff6873
LC		 3120 kfree(ext4_li_info);
3121 ext4_li_info = NULL;
92b97816 3122 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3123 "initialization thread\n",
3124 err);
3125 return err;
3126 }
3127 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3128 return 0;
3129}
3130
3131/*
3132 * Check whether it make sense to run itable init. thread or not.
3133 * If there is at least one uninitialized inode table, return
3134 * corresponding group number, else the loop goes through all
3135 * groups and return total number of groups.
3136 */
3137static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3138{
3139 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3140 struct ext4_group_desc *gdp = NULL;
3141
8844618d
TT		 3142 if (!ext4_has_group_desc_csum(sb))
3143 return ngroups;
3144
bfff6873
LC		 3145 for (group = 0; group < ngroups; group++) {
3146 gdp = ext4_get_group_desc(sb, group, NULL);
3147 if (!gdp)
3148 continue;
3149
8844618d
TT		 3150 if (gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))
3151 continue;
3152 if (group != 0)
bfff6873 3153 break;
8844618d
TT		 3154 ext4_error(sb, "Inode table for bg 0 marked as "
3155 "needing zeroing");
3156 if (sb_rdonly(sb))
3157 return ngroups;
bfff6873
LC		 3158 }
3159
3160 return group;
3161}
3162
3163static int ext4_li_info_new(void)
3164{
3165 struct ext4_lazy_init *eli = NULL;
3166
3167 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3168 if (!eli)
3169 return -ENOMEM;
3170
bfff6873
LC		 3171 INIT_LIST_HEAD(&eli->li_request_list);
3172 mutex_init(&eli->li_list_mtx);
3173
bfff6873
LC		 3174 eli->li_state |= EXT4_LAZYINIT_QUIT;
3175
3176 ext4_li_info = eli;
3177
3178 return 0;
3179}
3180
3181static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3182 ext4_group_t start)
3183{
3184 struct ext4_sb_info *sbi = EXT4_SB(sb);
3185 struct ext4_li_request *elr;
bfff6873
LC		 3186
3187 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3188 if (!elr)
3189 return NULL;
3190
3191 elr->lr_super = sb;
3192 elr->lr_sbi = sbi;
3193 elr->lr_next_group = start;
3194
3195 /*
3196 * Randomize first schedule time of the request to
3197 * spread the inode table initialization requests
3198 * better.
3199 */
dd1f723b
TT		 3200 elr->lr_next_sched = jiffies + (prandom_u32() %
3201 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3202 return elr;
3203}
3204
7f511862
TT		 3205int ext4_register_li_request(struct super_block *sb,
3206 ext4_group_t first_not_zeroed)
bfff6873
LC		 3207{
3208 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3209 struct ext4_li_request *elr = NULL;
49598e04 3210 ext4_group_t ngroups = sbi->s_groups_count;
6c5a6cb9 3211 int ret = 0;
bfff6873 3212
7f511862 3213 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3214 if (sbi->s_li_request != NULL) {
3215 /*
3216 * Reset timeout so it can be computed again, because
3217 * s_li_wait_mult might have changed.
3218 */
3219 sbi->s_li_request->lr_timeout = 0;
7f511862 3220 goto out;
51ce6511 3221 }
bfff6873 3222
bc98a42c 3223 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
55ff3840 3224 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3225 goto out;
bfff6873
LC		 3226
3227 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3228 if (!elr) {
3229 ret = -ENOMEM;
3230 goto out;
3231 }
bfff6873
LC		 3232
3233 if (NULL == ext4_li_info) {
3234 ret = ext4_li_info_new();
3235 if (ret)
3236 goto out;
3237 }
3238
3239 mutex_lock(&ext4_li_info->li_list_mtx);
3240 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3241 mutex_unlock(&ext4_li_info->li_list_mtx);
3242
3243 sbi->s_li_request = elr;
46e4690b
TM		 3244 /*
3245 * set elr to NULL here since it has been inserted to
3246 * the request_list and the removal and free of it is
3247 * handled by ext4_clear_request_list from now on.
3248 */
3249 elr = NULL;
bfff6873
LC		 3250
3251 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3252 ret = ext4_run_lazyinit_thread();
3253 if (ret)
3254 goto out;
3255 }
bfff6873 3256out:
beed5ecb
NK		 3257 mutex_unlock(&ext4_li_mtx);
3258 if (ret)
bfff6873 3259 kfree(elr);
bfff6873
LC		 3260 return ret;
3261}
3262
3263/*
3264 * We do not need to lock anything since this is called on
3265 * module unload.
3266 */
3267static void ext4_destroy_lazyinit_thread(void)
3268{
3269 /*
3270 * If thread exited earlier
3271 * there's nothing to be done.
3272 */
8f1f7453 3273 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3274 return;
3275
8f1f7453 3276 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3277}
3278
25ed6e8a
DW		 3279static int set_journal_csum_feature_set(struct super_block *sb)
3280{
3281 int ret = 1;
3282 int compat, incompat;
3283 struct ext4_sb_info *sbi = EXT4_SB(sb);
3284
9aa5d32b 3285 if (ext4_has_metadata_csum(sb)) {
db9ee220 3286 /* journal checksum v3 */
25ed6e8a 3287 compat = 0;
db9ee220 3288 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3289 } else {
3290 /* journal checksum v1 */
3291 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3292 incompat = 0;
3293 }
3294
feb8c6d3
DW		 3295 jbd2_journal_clear_features(sbi->s_journal,
3296 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3297 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3298 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3299 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3300 ret = jbd2_journal_set_features(sbi->s_journal,
3301 compat, 0,
3302 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3303 incompat);
3304 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3305 ret = jbd2_journal_set_features(sbi->s_journal,
3306 compat, 0,
3307 incompat);
3308 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3309 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3310 } else {
feb8c6d3
DW		 3311 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3312 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3313 }
3314
3315 return ret;
3316}
3317
952fc18e
TT		 3318/*
3319 * Note: calculating the overhead so we can be compatible with
3320 * historical BSD practice is quite difficult in the face of
3321 * clusters/bigalloc. This is because multiple metadata blocks from
3322 * different block group can end up in the same allocation cluster.
3323 * Calculating the exact overhead in the face of clustered allocation
3324 * requires either O(all block bitmaps) in memory or O(number of block
3325 * groups**2) in time. We will still calculate the superblock for
3326 * older file systems --- and if we come across with a bigalloc file
3327 * system with zero in s_overhead_clusters the estimate will be close to
3328 * correct especially for very large cluster sizes --- but for newer
3329 * file systems, it's better to calculate this figure once at mkfs
3330 * time, and store it in the superblock. If the superblock value is
3331 * present (even for non-bigalloc file systems), we will use it.
3332 */
3333static int count_overhead(struct super_block *sb, ext4_group_t grp,
3334 char *buf)
3335{
3336 struct ext4_sb_info *sbi = EXT4_SB(sb);
3337 struct ext4_group_desc *gdp;
3338 ext4_fsblk_t first_block, last_block, b;
3339 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3340 int s, j, count = 0;
3341
e2b911c5 3342 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3343 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3344 sbi->s_itb_per_group + 2);
3345
952fc18e
TT		 3346 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3347 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3348 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3349 for (i = 0; i < ngroups; i++) {
3350 gdp = ext4_get_group_desc(sb, i, NULL);
3351 b = ext4_block_bitmap(sb, gdp);
3352 if (b >= first_block && b <= last_block) {
3353 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3354 count++;
3355 }
3356 b = ext4_inode_bitmap(sb, gdp);
3357 if (b >= first_block && b <= last_block) {
3358 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3359 count++;
3360 }
3361 b = ext4_inode_table(sb, gdp);
3362 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3363 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3364 int c = EXT4_B2C(sbi, b - first_block);
3365 ext4_set_bit(c, buf);
3366 count++;
3367 }
3368 if (i != grp)
3369 continue;
3370 s = 0;
3371 if (ext4_bg_has_super(sb, grp)) {
3372 ext4_set_bit(s++, buf);
3373 count++;
3374 }
c48ae41b
TT		 3375 j = ext4_bg_num_gdb(sb, grp);
3376 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3377 ext4_error(sb, "Invalid number of block group "
3378 "descriptor blocks: %d", j);
3379 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3380 }
c48ae41b
TT		 3381 count += j;
3382 for (; j > 0; j--)
3383 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3384 }
3385 if (!count)
3386 return 0;
3387 return EXT4_CLUSTERS_PER_GROUP(sb) -
3388 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3389}
3390
3391/*
3392 * Compute the overhead and stash it in sbi->s_overhead
3393 */
3394int ext4_calculate_overhead(struct super_block *sb)
3395{
3396 struct ext4_sb_info *sbi = EXT4_SB(sb);
3397 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3398 struct inode *j_inode;
3399 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3400 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3401 ext4_fsblk_t overhead = 0;
4fdb5543 3402 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3403
952fc18e
TT		 3404 if (!buf)
3405 return -ENOMEM;
3406
3407 /*
3408 * Compute the overhead (FS structures). This is constant
3409 * for a given filesystem unless the number of block groups
3410 * changes so we cache the previous value until it does.
3411 */
3412
3413 /*
3414 * All of the blocks before first_data_block are overhead
3415 */
3416 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3417
3418 /*
3419 * Add the overhead found in each block group
3420 */
3421 for (i = 0; i < ngroups; i++) {
3422 int blks;
3423
3424 blks = count_overhead(sb, i, buf);
3425 overhead += blks;
3426 if (blks)
3427 memset(buf, 0, PAGE_SIZE);
3428 cond_resched();
3429 }
3c816ded
EW		 3430
3431 /*
3432 * Add the internal journal blocks whether the journal has been
3433 * loaded or not
3434 */
b003b524 3435 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3436 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3c816ded
EW		 3437 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3438 j_inode = ext4_get_journal_inode(sb, j_inum);
3439 if (j_inode) {
3440 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3441 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3442 iput(j_inode);
3443 } else {
3444 ext4_msg(sb, KERN_ERR, "can't get journal size");
3445 }
3446 }
952fc18e
TT		 3447 sbi->s_overhead = overhead;
3448 smp_wmb();
3449 free_page((unsigned long) buf);
3450 return 0;
3451}
3452
b5799018 3453static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3454{
3455 ext4_fsblk_t resv_clusters;
b5799018 3456 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3457
30fac0f7
JK		 3458 /*
3459 * There's no need to reserve anything when we aren't using extents.
3460 * The space estimates are exact, there are no unwritten extents,
3461 * hole punching doesn't need new metadata... This is needed especially
3462 * to keep ext2/3 backward compatibility.
3463 */
e2b911c5 3464 if (!ext4_has_feature_extents(sb))
b5799018 3465 return;
27dd4385
LC		 3466 /*
3467 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3468 * This should cover the situations where we can not afford to run
3469 * out of space like for example punch hole, or converting
556615dc 3470 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3471 * allocation would require 1, or 2 blocks, higher numbers are
3472 * very rare.
3473 */
b5799018
TT		 3474 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3475 sbi->s_cluster_bits);
27dd4385
LC		 3476
3477 do_div(resv_clusters, 50);
3478 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3479
b5799018 3480 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3481}
3482
2b2d6d01 3483static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3484{
5e405595 3485 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
d4c402d9 3486 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3487 struct buffer_head *bh;
617ba13b 3488 struct ext4_super_block *es = NULL;
5aee0f8a 3489 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
617ba13b
MC		 3490 ext4_fsblk_t block;
3491 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3492 ext4_fsblk_t logical_sb_block;
ac27a0ec 3493 unsigned long offset = 0;
ac27a0ec
DK		 3494 unsigned long journal_devnum = 0;
3495 unsigned long def_mount_opts;
3496 struct inode *root;
0390131b 3497 const char *descr;
dcc7dae3 3498 int ret = -ENOMEM;
281b5995 3499 int blocksize, clustersize;
4ec11028
TT		 3500 unsigned int db_count;
3501 unsigned int i;
281b5995 3502 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3503 __u64 blocks_count;
07aa2ea1 3504 int err = 0;
b3881f74 3505 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3506 ext4_group_t first_not_zeroed;
ac27a0ec 3507
5aee0f8a
TT		 3508 if ((data && !orig_data) || !sbi)
3509 goto out_free_base;
705895b6 3510
aed9eb1b 3511 sbi->s_daxdev = dax_dev;
705895b6
PE		 3512 sbi->s_blockgroup_lock =
3513 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 3514 if (!sbi->s_blockgroup_lock)
3515 goto out_free_base;
3516
ac27a0ec 3517 sb->s_fs_info = sbi;
2c0544b2 3518 sbi->s_sb = sb;
240799cd 3519 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3520 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3521 if (sb->s_bdev->bd_part)
3522 sbi->s_sectors_written_start =
3523 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3524
9f6200bb 3525 /* Cleanup superblock name */
ec3904dc 3526 strreplace(sb->s_id, '/', '!');
9f6200bb 3527
07aa2ea1 3528 /* -EINVAL is default */
dcc7dae3 3529 ret = -EINVAL;
617ba13b 3530 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3531 if (!blocksize) {
b31e1552 3532 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3533 goto out_fail;
3534 }
3535
3536 /*
617ba13b 3537 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3538 * block sizes. We need to calculate the offset from buffer start.
3539 */
617ba13b 3540 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3541 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3542 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3543 } else {
70bbb3e0 3544 logical_sb_block = sb_block;
ac27a0ec
DK		 3545 }
3546
a8ac900b 3547 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3548 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3549 goto out_fail;
3550 }
3551 /*
3552 * Note: s_es must be initialized as soon as possible because
617ba13b 3553 * some ext4 macro-instructions depend on its value
ac27a0ec 3554 */
2716b802 3555 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3556 sbi->s_es = es;
3557 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3558 if (sb->s_magic != EXT4_SUPER_MAGIC)
3559 goto cantfind_ext4;
afc32f7e 3560 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3561
feb0ab32 3562 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3563 if (ext4_has_feature_metadata_csum(sb) &&
3564 ext4_has_feature_gdt_csum(sb))
363307e6 3565 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3566 "redundant flags; please run fsck.");
3567
d25425f8
DW		 3568 /* Check for a known checksum algorithm */
3569 if (!ext4_verify_csum_type(sb, es)) {
3570 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3571 "unknown checksum algorithm.");
3572 silent = 1;
3573 goto cantfind_ext4;
3574 }
3575
0441984a 3576 /* Load the checksum driver */
a45403b5
TT		 3577 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3578 if (IS_ERR(sbi->s_chksum_driver)) {
3579 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3580 ret = PTR_ERR(sbi->s_chksum_driver);
3581 sbi->s_chksum_driver = NULL;
3582 goto failed_mount;
0441984a
DW		 3583 }
3584
a9c47317
DW		 3585 /* Check superblock checksum */
3586 if (!ext4_superblock_csum_verify(sb, es)) {
3587 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3588 "invalid superblock checksum. Run e2fsck?");
3589 silent = 1;
6a797d27 3590 ret = -EFSBADCRC;
a9c47317
DW		 3591 goto cantfind_ext4;
3592 }
3593
3594 /* Precompute checksum seed for all metadata */
e2b911c5 3595 if (ext4_has_feature_csum_seed(sb))
8c81bd8f 3596 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
dec214d0 3597 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
a9c47317
DW		 3598 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3599 sizeof(es->s_uuid));
3600
ac27a0ec
DK		 3601 /* Set defaults before we parse the mount options */
3602 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3603 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3604 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3605 set_opt(sb, DEBUG);
87f26807 3606 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3607 set_opt(sb, GRPID);
617ba13b 3608 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3609 set_opt(sb, NO_UID32);
ea663336 3610 /* xattr user namespace & acls are now defaulted on */
ea663336 3611 set_opt(sb, XATTR_USER);
03010a33 3612#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3613 set_opt(sb, POSIX_ACL);
2e7842b8 3614#endif
98c1a759
DW		 3615 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3616 if (ext4_has_metadata_csum(sb))
3617 set_opt(sb, JOURNAL_CHECKSUM);
3618
617ba13b 3619 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3620 set_opt(sb, JOURNAL_DATA);
617ba13b 3621 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3622 set_opt(sb, ORDERED_DATA);
617ba13b 3623 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3624 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3625
3626 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3627 set_opt(sb, ERRORS_PANIC);
bb4f397a 3628 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3629 set_opt(sb, ERRORS_CONT);
bb4f397a 3630 else
fd8c37ec 3631 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3632 /* block_validity enabled by default; disable with noblock_validity */
3633 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3634 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3635 set_opt(sb, DISCARD);
ac27a0ec 3636
08cefc7a
EB		 3637 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3638 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3639 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3640 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3641 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3642
8b67f04a 3643 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3644 set_opt(sb, BARRIER);
ac27a0ec 3645
dd919b98
AK		 3646 /*
3647 * enable delayed allocation by default
3648 * Use -o nodelalloc to turn it off
3649 */
bc0b75f7 3650 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3651 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3652 set_opt(sb, DELALLOC);
dd919b98 3653
51ce6511
LC		 3654 /*
3655 * set default s_li_wait_mult for lazyinit, for the case there is
3656 * no mount option specified.
3657 */
3658 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3659
5aee0f8a
TT		 3660 if (sbi->s_es->s_mount_opts[0]) {
3661 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3662 sizeof(sbi->s_es->s_mount_opts),
3663 GFP_KERNEL);
3664 if (!s_mount_opts)
3665 goto failed_mount;
3666 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3667 &journal_ioprio, 0)) {
3668 ext4_msg(sb, KERN_WARNING,
3669 "failed to parse options in superblock: %s",
3670 s_mount_opts);
3671 }
3672 kfree(s_mount_opts);
8b67f04a 3673 }
5a916be1 3674 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3675 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3676 &journal_ioprio, 0))
ac27a0ec
DK		 3677 goto failed_mount;
3678
56889787
TT		 3679 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3680 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3681 "with data=journal disables delayed "
3682 "allocation and O_DIRECT support!\n");
3683 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3684 ext4_msg(sb, KERN_ERR, "can't mount with "
3685 "both data=journal and delalloc");
3686 goto failed_mount;
3687 }
3688 if (test_opt(sb, DIOREAD_NOLOCK)) {
3689 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3690 "both data=journal and dioread_nolock");
56889787
TT		 3691 goto failed_mount;
3692 }
923ae0ff
RZ		 3693 if (test_opt(sb, DAX)) {
3694 ext4_msg(sb, KERN_ERR, "can't mount with "
3695 "both data=journal and dax");
3696 goto failed_mount;
3697 }
73b92a2a
SK		 3698 if (ext4_has_feature_encrypt(sb)) {
3699 ext4_msg(sb, KERN_WARNING,
3700 "encrypted files will use data=ordered "
3701 "instead of data journaling mode");
3702 }
56889787
TT		 3703 if (test_opt(sb, DELALLOC))
3704 clear_opt(sb, DELALLOC);
001e4a87
TH		 3705 } else {
3706 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3707 }
3708
1751e8a6
LT		 3709 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3710 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec 3711
617ba13b 3712 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3713 (ext4_has_compat_features(sb) ||
3714 ext4_has_ro_compat_features(sb) ||
3715 ext4_has_incompat_features(sb)))
b31e1552
ES		 3716 ext4_msg(sb, KERN_WARNING,
3717 "feature flags set on rev 0 fs, "
3718 "running e2fsck is recommended");
469108ff 3719
ed3654eb
TT		 3720 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3721 set_opt2(sb, HURD_COMPAT);
e2b911c5 3722 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3723 ext4_msg(sb, KERN_ERR,
3724 "The Hurd can't support 64-bit file systems");
3725 goto failed_mount;
3726 }
dec214d0
TE		 3727
3728 /*
3729 * ea_inode feature uses l_i_version field which is not
3730 * available in HURD_COMPAT mode.
3731 */
3732 if (ext4_has_feature_ea_inode(sb)) {
3733 ext4_msg(sb, KERN_ERR,
3734 "ea_inode feature is not supported for Hurd");
3735 goto failed_mount;
3736 }
ed3654eb
TT		 3737 }
3738
2035e776
TT		 3739 if (IS_EXT2_SB(sb)) {
3740 if (ext2_feature_set_ok(sb))
3741 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3742 "using the ext4 subsystem");
3743 else {
0d9366d6
ES		 3744 /*
3745 * If we're probing be silent, if this looks like
3746 * it's actually an ext[34] filesystem.
3747 */
3748 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3749 goto failed_mount;
2035e776
TT		 3750 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3751 "to feature incompatibilities");
3752 goto failed_mount;
3753 }
3754 }
3755
3756 if (IS_EXT3_SB(sb)) {
3757 if (ext3_feature_set_ok(sb))
3758 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3759 "using the ext4 subsystem");
3760 else {
0d9366d6
ES		 3761 /*
3762 * If we're probing be silent, if this looks like
3763 * it's actually an ext4 filesystem.
3764 */
3765 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3766 goto failed_mount;
2035e776
TT		 3767 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3768 "to feature incompatibilities");
3769 goto failed_mount;
3770 }
3771 }
3772
ac27a0ec
DK		 3773 /*
3774 * Check feature flags regardless of the revision level, since we
3775 * previously didn't change the revision level when setting the flags,
3776 * so there is a chance incompat flags are set on a rev 0 filesystem.
3777 */
bc98a42c 3778 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
ac27a0ec 3779 goto failed_mount;
a13fb1a4 3780
261cb20c 3781 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3782 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3783 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552 3784 ext4_msg(sb, KERN_ERR,
8cdf3372
TT		 3785 "Unsupported filesystem blocksize %d (%d log_block_size)",
3786 blocksize, le32_to_cpu(es->s_log_block_size));
3787 goto failed_mount;
3788 }
3789 if (le32_to_cpu(es->s_log_block_size) >
3790 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3791 ext4_msg(sb, KERN_ERR,
3792 "Invalid log block size: %u",
3793 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 3794 goto failed_mount;
3795 }
3796
5b9554dc
TT		 3797 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3798 ext4_msg(sb, KERN_ERR,
3799 "Number of reserved GDT blocks insanely large: %d",
3800 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3801 goto failed_mount;
3802 }
3803
923ae0ff 3804 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
559db4c6
RZ		 3805 if (ext4_has_feature_inline_data(sb)) {
3806 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3807 " that may contain inline data");
24f3478d 3808 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
559db4c6 3809 }
87eefeb4 3810 err = bdev_dax_supported(sb, blocksize);
24f3478d
DW		 3811 if (err) {
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 }
8cdf3372
TT		 3942 if (le32_to_cpu(es->s_log_cluster_size) >
3943 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3944 ext4_msg(sb, KERN_ERR,
3945 "Invalid log cluster size: %u",
3946 le32_to_cpu(es->s_log_cluster_size));
3947 goto failed_mount;
3948 }
281b5995
TT		 3949 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3950 le32_to_cpu(es->s_log_block_size);
3951 sbi->s_clusters_per_group =
3952 le32_to_cpu(es->s_clusters_per_group);
3953 if (sbi->s_clusters_per_group > blocksize * 8) {
3954 ext4_msg(sb, KERN_ERR,
3955 "#clusters per group too big: %lu",
3956 sbi->s_clusters_per_group);
3957 goto failed_mount;
3958 }
3959 if (sbi->s_blocks_per_group !=
3960 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3961 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3962 "clusters per group (%lu) inconsistent",
3963 sbi->s_blocks_per_group,
3964 sbi->s_clusters_per_group);
3965 goto failed_mount;
3966 }
3967 } else {
3968 if (clustersize != blocksize) {
3969 ext4_warning(sb, "fragment/cluster size (%d) != "
3970 "block size (%d)", clustersize,
3971 blocksize);
3972 clustersize = blocksize;
3973 }
3974 if (sbi->s_blocks_per_group > blocksize * 8) {
3975 ext4_msg(sb, KERN_ERR,
3976 "#blocks per group too big: %lu",
3977 sbi->s_blocks_per_group);
3978 goto failed_mount;
3979 }
3980 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3981 sbi->s_cluster_bits = 0;
ac27a0ec 3982 }
281b5995
TT		 3983 sbi->s_cluster_ratio = clustersize / blocksize;
3984
960fd856
TT		 3985 /* Do we have standard group size of clustersize * 8 blocks ? */
3986 if (sbi->s_blocks_per_group == clustersize << 3)
3987 set_opt2(sb, STD_GROUP_SIZE);
3988
bf43d84b
ES		 3989 /*
3990 * Test whether we have more sectors than will fit in sector_t,
3991 * and whether the max offset is addressable by the page cache.
3992 */
5a9ae68a 3993 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3994 ext4_blocks_count(es));
5a9ae68a 3995 if (err) {
b31e1552 3996 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3997 " too large to mount safely on this system");
ac27a0ec 3998 if (sizeof(sector_t) < 8)
90c699a9 3999 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 4000 goto failed_mount;
4001 }
4002
617ba13b
MC		 4003 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4004 goto cantfind_ext4;
e7c95593 4005
0f2ddca6
FTN		 4006 /* check blocks count against device size */
4007 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4008 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 4009 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4010 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 4011 ext4_blocks_count(es), blocks_count);
4012 goto failed_mount;
4013 }
4014
0b8e58a1
AD		 4015 /*
4016 * It makes no sense for the first data block to be beyond the end
4017 * of the filesystem.
4018 */
4019 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 4020 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 4021 "block %u is beyond end of filesystem (%llu)",
4022 le32_to_cpu(es->s_first_data_block),
4023 ext4_blocks_count(es));
e7c95593
ES		 4024 goto failed_mount;
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 }
a7c3e901 4053 sbi->s_group_desc = kvmalloc(db_count *
f18a5f21
TT		 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 }
4061
705895b6 4062 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec 4063
85c8f176
AP		 4064 /* Pre-read the descriptors into the buffer cache */
4065 for (i = 0; i < db_count; i++) {
4066 block = descriptor_loc(sb, logical_sb_block, i);
4067 sb_breadahead(sb, block);
4068 }
4069
ac27a0ec 4070 for (i = 0; i < db_count; i++) {
70bbb3e0 4071 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 4072 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 4073 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 4074 ext4_msg(sb, KERN_ERR,
4075 "can't read group descriptor %d", i);
ac27a0ec
DK		 4076 db_count = i;
4077 goto failed_mount2;
4078 }
4079 }
829fa70d 4080 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 4081 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 4082 ret = -EFSCORRUPTED;
f9ae9cf5 4083 goto failed_mount2;
ac27a0ec 4084 }
772cb7c8 4085
f9ae9cf5 4086 sbi->s_gdb_count = db_count;
ac27a0ec 4087
235699a8 4088 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
04496411 4089
a75ae78f 4090 /* Register extent status tree shrinker */
eb68d0e2 4091 if (ext4_es_register_shrinker(sbi))
ce7e010a 4092 goto failed_mount3;
ce7e010a 4093
c9de560d 4094 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 4095 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 4096
f9ae9cf5
TT		 4097 /*
4098 * set up enough so that it can read an inode
4099 */
f6e63f90 4100 sb->s_op = &ext4_sops;
617ba13b
MC		 4101 sb->s_export_op = &ext4_export_ops;
4102 sb->s_xattr = ext4_xattr_handlers;
ffcc4182 4103#ifdef CONFIG_EXT4_FS_ENCRYPTION
a7550b30 4104 sb->s_cop = &ext4_cryptops;
ffcc4182 4105#endif
ac27a0ec 4106#ifdef CONFIG_QUOTA
617ba13b 4107 sb->dq_op = &ext4_quota_operations;
e2b911c5 4108 if (ext4_has_feature_quota(sb))
1fa5efe3 4109 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 4110 else
4111 sb->s_qcop = &ext4_qctl_operations;
689c958c 4112 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 4113#endif
85787090 4114 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
f2fa2ffc 4115
ac27a0ec 4116 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 4117 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 4118
4119 sb->s_root = NULL;
4120
4121 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 4122 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 4123
bc98a42c 4124 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
c5e06d10 4125 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 4126 goto failed_mount3a;
c5e06d10 4127
ac27a0ec
DK		 4128 /*
4129 * The first inode we look at is the journal inode. Don't try
4130 * root first: it may be modified in the journal!
4131 */
e2b911c5 4132 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4753d8a2
TT		 4133 err = ext4_load_journal(sb, es, journal_devnum);
4134 if (err)
50460fe8 4135 goto failed_mount3a;
bc98a42c 4136 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
e2b911c5 4137 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 4138 ext4_msg(sb, KERN_ERR, "required journal recovery "
4139 "suppressed and not mounted read-only");
744692dc 4140 goto failed_mount_wq;
ac27a0ec 4141 } else {
1e381f60
DM		 4142 /* Nojournal mode, all journal mount options are illegal */
4143 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4144 ext4_msg(sb, KERN_ERR, "can't mount with "
4145 "journal_checksum, fs mounted w/o journal");
4146 goto failed_mount_wq;
4147 }
4148 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4149 ext4_msg(sb, KERN_ERR, "can't mount with "
4150 "journal_async_commit, fs mounted w/o journal");
4151 goto failed_mount_wq;
4152 }
4153 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4154 ext4_msg(sb, KERN_ERR, "can't mount with "
4155 "commit=%lu, fs mounted w/o journal",
4156 sbi->s_commit_interval / HZ);
4157 goto failed_mount_wq;
4158 }
4159 if (EXT4_MOUNT_DATA_FLAGS &
4160 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4161 ext4_msg(sb, KERN_ERR, "can't mount with "
4162 "data=, fs mounted w/o journal");
4163 goto failed_mount_wq;
4164 }
4165 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4166 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 4167 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 4168 sbi->s_journal = NULL;
4169 needs_recovery = 0;
4170 goto no_journal;
ac27a0ec
DK		 4171 }
4172
e2b911c5 4173 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 4174 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4175 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4176 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4177 goto failed_mount_wq;
eb40a09c
JS		 4178 }
4179
25ed6e8a
DW		 4180 if (!set_journal_csum_feature_set(sb)) {
4181 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4182 "feature set");
4183 goto failed_mount_wq;
d4da6c9c 4184 }
818d276c 4185
ac27a0ec
DK		 4186 /* We have now updated the journal if required, so we can
4187 * validate the data journaling mode. */
4188 switch (test_opt(sb, DATA_FLAGS)) {
4189 case 0:
4190 /* No mode set, assume a default based on the journal
63f57933
AM		 4191 * capabilities: ORDERED_DATA if the journal can
4192 * cope, else JOURNAL_DATA
4193 */
dab291af 4194 if (jbd2_journal_check_available_features
27f394a7 4195 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
fd8c37ec 4196 set_opt(sb, ORDERED_DATA);
27f394a7
TN		 4197 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4198 } else {
fd8c37ec 4199 set_opt(sb, JOURNAL_DATA);
27f394a7
TN		 4200 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4201 }
ac27a0ec
DK		 4202 break;
4203
617ba13b
MC		 4204 case EXT4_MOUNT_ORDERED_DATA:
4205 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4206 if (!jbd2_journal_check_available_features
4207 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4208 ext4_msg(sb, KERN_ERR, "Journal does not support "
4209 "requested data journaling mode");
744692dc 4210 goto failed_mount_wq;
ac27a0ec
DK		 4211 }
4212 default:
4213 break;
4214 }
ab04df78
JK		 4215
4216 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4217 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4218 ext4_msg(sb, KERN_ERR, "can't mount with "
4219 "journal_async_commit in data=ordered mode");
4220 goto failed_mount_wq;
4221 }
4222
b3881f74 4223 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4224
18aadd47
BJ		 4225 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4226
ce7e010a 4227no_journal:
cdb7ee4c
TE		 4228 if (!test_opt(sb, NO_MBCACHE)) {
4229 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4230 if (!sbi->s_ea_block_cache) {
dec214d0 4231 ext4_msg(sb, KERN_ERR,
cdb7ee4c 4232 "Failed to create ea_block_cache");
dec214d0
TE		 4233 goto failed_mount_wq;
4234 }
cdb7ee4c
TE		 4235
4236 if (ext4_has_feature_ea_inode(sb)) {
4237 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4238 if (!sbi->s_ea_inode_cache) {
4239 ext4_msg(sb, KERN_ERR,
4240 "Failed to create ea_inode_cache");
4241 goto failed_mount_wq;
4242 }
4243 }
9c191f70
M		 4244 }
4245
e2b911c5 4246 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
09cbfeaf 4247 (blocksize != PAGE_SIZE)) {
1cb767cd
TT		 4248 ext4_msg(sb, KERN_ERR,
4249 "Unsupported blocksize for fs encryption");
4250 goto failed_mount_wq;
4251 }
4252
bc98a42c 4253 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
e2b911c5
DW		 4254 !ext4_has_feature_encrypt(sb)) {
4255 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 4256 ext4_commit_super(sb, 1);
4257 }
4258
952fc18e
TT		 4259 /*
4260 * Get the # of file system overhead blocks from the
4261 * superblock if present.
4262 */
4263 if (es->s_overhead_clusters)
4264 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4265 else {
07aa2ea1
LC		 4266 err = ext4_calculate_overhead(sb);
4267 if (err)
952fc18e
TT		 4268 goto failed_mount_wq;
4269 }
4270
fd89d5f2
TH		 4271 /*
4272 * The maximum number of concurrent works can be high and
4273 * concurrency isn't really necessary. Limit it to 1.
4274 */
2e8fa54e
JK		 4275 EXT4_SB(sb)->rsv_conversion_wq =
4276 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4277 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4278 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4279 ret = -ENOMEM;
2e8fa54e
JK		 4280 goto failed_mount4;
4281 }
4282
ac27a0ec 4283 /*
dab291af 4284 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4285 * so we can safely mount the rest of the filesystem now.
4286 */
4287
1d1fe1ee
DH		 4288 root = ext4_iget(sb, EXT4_ROOT_INO);
4289 if (IS_ERR(root)) {
b31e1552 4290 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4291 ret = PTR_ERR(root);
32a9bb57 4292 root = NULL;
ac27a0ec
DK		 4293 goto failed_mount4;
4294 }
4295 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4296 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4297 iput(root);
ac27a0ec
DK		 4298 goto failed_mount4;
4299 }
48fde701 4300 sb->s_root = d_make_root(root);
1d1fe1ee 4301 if (!sb->s_root) {
b31e1552 4302 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4303 ret = -ENOMEM;
4304 goto failed_mount4;
4305 }
ac27a0ec 4306
c89128a0
JK		 4307 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4308 if (ret == -EROFS) {
1751e8a6 4309 sb->s_flags |= SB_RDONLY;
c89128a0
JK		 4310 ret = 0;
4311 } else if (ret)
4312 goto failed_mount4a;
ef7f3835
KS		 4313
4314 /* determine the minimum size of new large inodes, if present */
670e9875
TT		 4315 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4316 sbi->s_want_extra_isize == 0) {
ef7f3835
KS		 4317 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4318 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 4319 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 4320 if (sbi->s_want_extra_isize <
4321 le16_to_cpu(es->s_want_extra_isize))
4322 sbi->s_want_extra_isize =
4323 le16_to_cpu(es->s_want_extra_isize);
4324 if (sbi->s_want_extra_isize <
4325 le16_to_cpu(es->s_min_extra_isize))
4326 sbi->s_want_extra_isize =
4327 le16_to_cpu(es->s_min_extra_isize);
4328 }
4329 }
4330 /* Check if enough inode space is available */
4331 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4332 sbi->s_inode_size) {
4333 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4334 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4335 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4336 "available");
ef7f3835
KS		 4337 }
4338
b5799018 4339 ext4_set_resv_clusters(sb);
27dd4385 4340
6fd058f7
TT		 4341 err = ext4_setup_system_zone(sb);
4342 if (err) {
b31e1552 4343 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4344 "zone (%d)", err);
f9ae9cf5
TT		 4345 goto failed_mount4a;
4346 }
4347
4348 ext4_ext_init(sb);
4349 err = ext4_mb_init(sb);
4350 if (err) {
4351 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4352 err);
dcf2d804 4353 goto failed_mount5;
c2774d84
AK		 4354 }
4355
d5e03cbb
TT		 4356 block = ext4_count_free_clusters(sb);
4357 ext4_free_blocks_count_set(sbi->s_es,
4358 EXT4_C2B(sbi, block));
908c7f19
TH		 4359 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4360 GFP_KERNEL);
d5e03cbb
TT		 4361 if (!err) {
4362 unsigned long freei = ext4_count_free_inodes(sb);
4363 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4364 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4365 GFP_KERNEL);
d5e03cbb
TT		 4366 }
4367 if (!err)
4368 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4369 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4370 if (!err)
908c7f19
TH		 4371 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4372 GFP_KERNEL);
c8585c6f
DJ		 4373 if (!err)
4374 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4375
d5e03cbb
TT		 4376 if (err) {
4377 ext4_msg(sb, KERN_ERR, "insufficient memory");
4378 goto failed_mount6;
4379 }
4380
e2b911c5 4381 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 4382 if (!ext4_fill_flex_info(sb)) {
4383 ext4_msg(sb, KERN_ERR,
4384 "unable to initialize "
4385 "flex_bg meta info!");
4386 goto failed_mount6;
4387 }
4388
bfff6873
LC		 4389 err = ext4_register_li_request(sb, first_not_zeroed);
4390 if (err)
dcf2d804 4391 goto failed_mount6;
bfff6873 4392
b5799018 4393 err = ext4_register_sysfs(sb);
dcf2d804
TM		 4394 if (err)
4395 goto failed_mount7;
3197ebdb 4396
9b2ff357
JK		 4397#ifdef CONFIG_QUOTA
4398 /* Enable quota usage during mount. */
bc98a42c 4399 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
9b2ff357
JK		 4400 err = ext4_enable_quotas(sb);
4401 if (err)
4402 goto failed_mount8;
4403 }
4404#endif /* CONFIG_QUOTA */
4405
617ba13b
MC		 4406 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4407 ext4_orphan_cleanup(sb, es);
4408 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4409 if (needs_recovery) {
b31e1552 4410 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4411 ext4_mark_recovery_complete(sb, es);
4412 }
4413 if (EXT4_SB(sb)->s_journal) {
4414 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4415 descr = " journalled data mode";
4416 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4417 descr = " ordered data mode";
4418 else
4419 descr = " writeback data mode";
4420 } else
4421 descr = "out journal";
4422
79add3a3
LC		 4423 if (test_opt(sb, DISCARD)) {
4424 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4425 if (!blk_queue_discard(q))
4426 ext4_msg(sb, KERN_WARNING,
4427 "mounting with \"discard\" option, but "
4428 "the device does not support discard");
4429 }
4430
e294a537
TT		 4431 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4432 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 4433 "Opts: %.*s%s%s", descr,
4434 (int) sizeof(sbi->s_es->s_mount_opts),
4435 sbi->s_es->s_mount_opts,
e294a537 4436 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4437
66e61a9e
TT		 4438 if (es->s_error_count)
4439 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4440
efbed4dc
TT		 4441 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4442 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4443 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4444 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4445
d4c402d9 4446 kfree(orig_data);
ac27a0ec
DK		 4447 return 0;
4448
617ba13b 4449cantfind_ext4:
ac27a0ec 4450 if (!silent)
b31e1552 4451 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4452 goto failed_mount;
4453
72ba7450
TT		 4454#ifdef CONFIG_QUOTA
4455failed_mount8:
ebd173be 4456 ext4_unregister_sysfs(sb);
72ba7450 4457#endif
dcf2d804
TM		 4458failed_mount7:
4459 ext4_unregister_li_request(sb);
4460failed_mount6:
f9ae9cf5 4461 ext4_mb_release(sb);
d5e03cbb 4462 if (sbi->s_flex_groups)
b93b41d4 4463 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4464 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4465 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4466 percpu_counter_destroy(&sbi->s_dirs_counter);
4467 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4468failed_mount5:
f9ae9cf5
TT		 4469 ext4_ext_release(sb);
4470 ext4_release_system_zone(sb);
4471failed_mount4a:
94bf608a 4472 dput(sb->s_root);
32a9bb57 4473 sb->s_root = NULL;
94bf608a 4474failed_mount4:
b31e1552 4475 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4476 if (EXT4_SB(sb)->rsv_conversion_wq)
4477 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4478failed_mount_wq:
dec214d0
TE		 4479 if (sbi->s_ea_inode_cache) {
4480 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4481 sbi->s_ea_inode_cache = NULL;
4482 }
47387409
TE		 4483 if (sbi->s_ea_block_cache) {
4484 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4485 sbi->s_ea_block_cache = NULL;
82939d79 4486 }
0390131b
FM		 4487 if (sbi->s_journal) {
4488 jbd2_journal_destroy(sbi->s_journal);
4489 sbi->s_journal = NULL;
4490 }
50460fe8 4491failed_mount3a:
d3922a77 4492 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4493failed_mount3:
9105bb14 4494 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4495 if (sbi->s_mmp_tsk)
4496 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4497failed_mount2:
4498 for (i = 0; i < db_count; i++)
4499 brelse(sbi->s_group_desc[i]);
b93b41d4 4500 kvfree(sbi->s_group_desc);
ac27a0ec 4501failed_mount:
0441984a
DW		 4502 if (sbi->s_chksum_driver)
4503 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 4504#ifdef CONFIG_QUOTA
a2d4a646 4505 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4506 kfree(sbi->s_qf_names[i]);
4507#endif
617ba13b 4508 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4509 brelse(bh);
4510out_fail:
4511 sb->s_fs_info = NULL;
f6830165 4512 kfree(sbi->s_blockgroup_lock);
5aee0f8a 4513out_free_base:
ac27a0ec 4514 kfree(sbi);
d4c402d9 4515 kfree(orig_data);
5e405595 4516 fs_put_dax(dax_dev);
07aa2ea1 4517 return err ? err : ret;
ac27a0ec
DK		 4518}
4519
4520/*
4521 * Setup any per-fs journal parameters now. We'll do this both on
4522 * initial mount, once the journal has been initialised but before we've
4523 * done any recovery; and again on any subsequent remount.
4524 */
617ba13b 4525static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4526{
617ba13b 4527 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4528
30773840
TT		 4529 journal->j_commit_interval = sbi->s_commit_interval;
4530 journal->j_min_batch_time = sbi->s_min_batch_time;
4531 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4532
a931da6a 4533 write_lock(&journal->j_state_lock);
ac27a0ec 4534 if (test_opt(sb, BARRIER))
dab291af 4535 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4536 else
dab291af 4537 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4538 if (test_opt(sb, DATA_ERR_ABORT))
4539 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4540 else
4541 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4542 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4543}
4544
c6cb7e77
EW		 4545static struct inode *ext4_get_journal_inode(struct super_block *sb,
4546 unsigned int journal_inum)
ac27a0ec
DK		 4547{
4548 struct inode *journal_inode;
ac27a0ec 4549
c6cb7e77
EW		 4550 /*
4551 * Test for the existence of a valid inode on disk. Bad things
4552 * happen if we iget() an unused inode, as the subsequent iput()
4553 * will try to delete it.
4554 */
1d1fe1ee
DH		 4555 journal_inode = ext4_iget(sb, journal_inum);
4556 if (IS_ERR(journal_inode)) {
b31e1552 4557 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4558 return NULL;
4559 }
4560 if (!journal_inode->i_nlink) {
4561 make_bad_inode(journal_inode);
4562 iput(journal_inode);
b31e1552 4563 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4564 return NULL;
4565 }
4566
e5f8eab8 4567 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4568 journal_inode, journal_inode->i_size);
1d1fe1ee 4569 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4570 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4571 iput(journal_inode);
4572 return NULL;
4573 }
c6cb7e77
EW		 4574 return journal_inode;
4575}
4576
4577static journal_t *ext4_get_journal(struct super_block *sb,
4578 unsigned int journal_inum)
4579{
4580 struct inode *journal_inode;
4581 journal_t *journal;
4582
4583 BUG_ON(!ext4_has_feature_journal(sb));
4584
4585 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4586 if (!journal_inode)
4587 return NULL;
ac27a0ec 4588
dab291af 4589 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4590 if (!journal) {
b31e1552 4591 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4592 iput(journal_inode);
4593 return NULL;
4594 }
4595 journal->j_private = sb;
617ba13b 4596 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4597 return journal;
4598}
4599
617ba13b 4600static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4601 dev_t j_dev)
4602{
2b2d6d01 4603 struct buffer_head *bh;
ac27a0ec 4604 journal_t *journal;
617ba13b
MC		 4605 ext4_fsblk_t start;
4606 ext4_fsblk_t len;
ac27a0ec 4607 int hblock, blocksize;
617ba13b 4608 ext4_fsblk_t sb_block;
ac27a0ec 4609 unsigned long offset;
2b2d6d01 4610 struct ext4_super_block *es;
ac27a0ec
DK		 4611 struct block_device *bdev;
4612
e2b911c5 4613 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4614
b31e1552 4615 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4616 if (bdev == NULL)
4617 return NULL;
4618
ac27a0ec 4619 blocksize = sb->s_blocksize;
e1defc4f 4620 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4621 if (blocksize < hblock) {
b31e1552
ES		 4622 ext4_msg(sb, KERN_ERR,
4623 "blocksize too small for journal device");
ac27a0ec
DK		 4624 goto out_bdev;
4625 }
4626
617ba13b
MC		 4627 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4628 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4629 set_blocksize(bdev, blocksize);
4630 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4631 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4632 "external journal");
ac27a0ec
DK		 4633 goto out_bdev;
4634 }
4635
2716b802 4636 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4637 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4638 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4639 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4640 ext4_msg(sb, KERN_ERR, "external journal has "
4641 "bad superblock");
ac27a0ec
DK		 4642 brelse(bh);
4643 goto out_bdev;
4644 }
4645
df4763be
DW		 4646 if ((le32_to_cpu(es->s_feature_ro_compat) &
4647 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4648 es->s_checksum != ext4_superblock_csum(sb, es)) {
4649 ext4_msg(sb, KERN_ERR, "external journal has "
4650 "corrupt superblock");
4651 brelse(bh);
4652 goto out_bdev;
4653 }
4654
617ba13b 4655 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4656 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4657 brelse(bh);
4658 goto out_bdev;
4659 }
4660
bd81d8ee 4661 len = ext4_blocks_count(es);
ac27a0ec
DK		 4662 start = sb_block + 1;
4663 brelse(bh); /* we're done with the superblock */
4664
dab291af 4665 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4666 start, len, blocksize);
4667 if (!journal) {
b31e1552 4668 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4669 goto out_bdev;
4670 }
4671 journal->j_private = sb;
dfec8a14 4672 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4673 wait_on_buffer(journal->j_sb_buffer);
4674 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4675 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4676 goto out_journal;
4677 }
4678 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4679 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4680 "user (unsupported) - %d",
ac27a0ec
DK		 4681 be32_to_cpu(journal->j_superblock->s_nr_users));
4682 goto out_journal;
4683 }
617ba13b
MC		 4684 EXT4_SB(sb)->journal_bdev = bdev;
4685 ext4_init_journal_params(sb, journal);
ac27a0ec 4686 return journal;
0b8e58a1 4687
ac27a0ec 4688out_journal:
dab291af 4689 jbd2_journal_destroy(journal);
ac27a0ec 4690out_bdev:
617ba13b 4691 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4692 return NULL;
4693}
4694
617ba13b
MC		 4695static int ext4_load_journal(struct super_block *sb,
4696 struct ext4_super_block *es,
ac27a0ec
DK		 4697 unsigned long journal_devnum)
4698{
4699 journal_t *journal;
4700 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4701 dev_t journal_dev;
4702 int err = 0;
4703 int really_read_only;
4704
e2b911c5 4705 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4706
ac27a0ec
DK		 4707 if (journal_devnum &&
4708 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4709 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4710 "numbers have changed");
ac27a0ec
DK		 4711 journal_dev = new_decode_dev(journal_devnum);
4712 } else
4713 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4714
4715 really_read_only = bdev_read_only(sb->s_bdev);
4716
4717 /*
4718 * Are we loading a blank journal or performing recovery after a
4719 * crash? For recovery, we need to check in advance whether we
4720 * can get read-write access to the device.
4721 */
e2b911c5 4722 if (ext4_has_feature_journal_needs_recovery(sb)) {
bc98a42c 4723 if (sb_rdonly(sb)) {
b31e1552
ES		 4724 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4725 "required on readonly filesystem");
ac27a0ec 4726 if (really_read_only) {
b31e1552 4727 ext4_msg(sb, KERN_ERR, "write access "
d98bf8cd
SR		 4728 "unavailable, cannot proceed "
4729 "(try mounting with noload)");
ac27a0ec
DK		 4730 return -EROFS;
4731 }
b31e1552
ES		 4732 ext4_msg(sb, KERN_INFO, "write access will "
4733 "be enabled during recovery");
ac27a0ec
DK		 4734 }
4735 }
4736
4737 if (journal_inum && journal_dev) {
b31e1552
ES		 4738 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4739 "and inode journals!");
ac27a0ec
DK		 4740 return -EINVAL;
4741 }
4742
4743 if (journal_inum) {
617ba13b 4744 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4745 return -EINVAL;
4746 } else {
617ba13b 4747 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4748 return -EINVAL;
4749 }
4750
90576c0b 4751 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4752 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4753
e2b911c5 4754 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4755 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4756 if (!err) {
4757 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4758 if (save)
4759 memcpy(save, ((char *) es) +
4760 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4761 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4762 if (save)
4763 memcpy(((char *) es) + EXT4_S_ERR_START,
4764 save, EXT4_S_ERR_LEN);
4765 kfree(save);
4766 }
ac27a0ec
DK		 4767
4768 if (err) {
b31e1552 4769 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4770 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4771 return err;
4772 }
4773
617ba13b
MC		 4774 EXT4_SB(sb)->s_journal = journal;
4775 ext4_clear_journal_err(sb, es);
ac27a0ec 4776
c41303ce 4777 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4778 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4779 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4780
4781 /* Make sure we flush the recovery flag to disk. */
e2d67052 4782 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4783 }
4784
4785 return 0;
4786}
4787
e2d67052 4788static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4789{
e2d67052 4790 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4791 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4792 int error = 0;
ac27a0ec 4793
bdfe0cbd 4794 if (!sbh || block_device_ejected(sb))
c4be0c1d 4795 return error;
71290b36
TT		 4796 /*
4797 * If the file system is mounted read-only, don't update the
4798 * superblock write time. This avoids updating the superblock
4799 * write time when we are mounting the root file system
4800 * read/only but we need to replay the journal; at that point,
4801 * for people who are east of GMT and who make their clock
4802 * tick in localtime for Windows bug-for-bug compatibility,
4803 * the clock is set in the future, and this will cause e2fsck
4804 * to complain and force a full file system check.
4805 */
1751e8a6 4806 if (!(sb->s_flags & SB_RDONLY))
71290b36 4807 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4808 if (sb->s_bdev->bd_part)
4809 es->s_kbytes_written =
4810 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4811 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4812 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4813 else
4814 es->s_kbytes_written =
4815 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4816 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4817 ext4_free_blocks_count_set(es,
57042651
TT		 4818 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4819 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4820 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4821 es->s_free_inodes_count =
4822 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4823 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4824 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4825 ext4_superblock_csum_set(sb);
1566a48a
TT		 4826 if (sync)
4827 lock_buffer(sbh);
4743f839
PKS		 4828 if (buffer_write_io_error(sbh)) {
4829 /*
4830 * Oh, dear. A previous attempt to write the
4831 * superblock failed. This could happen because the
4832 * USB device was yanked out. Or it could happen to
4833 * be a transient write error and maybe the block will
4834 * be remapped. Nothing we can do but to retry the
4835 * write and hope for the best.
4836 */
4837 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4838 "superblock detected");
4839 clear_buffer_write_io_error(sbh);
4840 set_buffer_uptodate(sbh);
4841 }
ac27a0ec 4842 mark_buffer_dirty(sbh);
914258bf 4843 if (sync) {
1566a48a 4844 unlock_buffer(sbh);
564bc402 4845 error = __sync_dirty_buffer(sbh,
00473374 4846 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
c89128a0 4847 if (buffer_write_io_error(sbh)) {
b31e1552
ES		 4848 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4849 "superblock");
914258bf
TT		 4850 clear_buffer_write_io_error(sbh);
4851 set_buffer_uptodate(sbh);
4852 }
4853 }
c4be0c1d 4854 return error;
ac27a0ec
DK		 4855}
4856
ac27a0ec
DK		 4857/*
4858 * Have we just finished recovery? If so, and if we are mounting (or
4859 * remounting) the filesystem readonly, then we will end up with a
4860 * consistent fs on disk. Record that fact.
4861 */
2b2d6d01
TT		 4862static void ext4_mark_recovery_complete(struct super_block *sb,
4863 struct ext4_super_block *es)
ac27a0ec 4864{
617ba13b 4865 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4866
e2b911c5 4867 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4868 BUG_ON(journal != NULL);
4869 return;
4870 }
dab291af 4871 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4872 if (jbd2_journal_flush(journal) < 0)
4873 goto out;
4874
bc98a42c 4875 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
e2b911c5 4876 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4877 ext4_commit_super(sb, 1);
ac27a0ec 4878 }
7ffe1ea8
HK		 4879
4880out:
dab291af 4881 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4882}
4883
4884/*
4885 * If we are mounting (or read-write remounting) a filesystem whose journal
4886 * has recorded an error from a previous lifetime, move that error to the
4887 * main filesystem now.
4888 */
2b2d6d01
TT		 4889static void ext4_clear_journal_err(struct super_block *sb,
4890 struct ext4_super_block *es)
ac27a0ec
DK		 4891{
4892 journal_t *journal;
4893 int j_errno;
4894 const char *errstr;
4895
e2b911c5 4896 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4897
617ba13b 4898 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4899
4900 /*
4901 * Now check for any error status which may have been recorded in the
617ba13b 4902 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4903 */
4904
dab291af 4905 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4906 if (j_errno) {
4907 char nbuf[16];
4908
617ba13b 4909 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4910 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4911 "from previous mount: %s", errstr);
12062ddd 4912 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4913
617ba13b
MC		 4914 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4915 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4916 ext4_commit_super(sb, 1);
ac27a0ec 4917
dab291af 4918 jbd2_journal_clear_err(journal);
d796c52e 4919 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4920 }
4921}
4922
4923/*
4924 * Force the running and committing transactions to commit,
4925 * and wait on the commit.
4926 */
617ba13b 4927int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4928{
4929 journal_t *journal;
ac27a0ec 4930
bc98a42c 4931 if (sb_rdonly(sb))
ac27a0ec
DK		 4932 return 0;
4933
617ba13b 4934 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4935 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4936}
4937
617ba13b 4938static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4939{
14ce0cb4 4940 int ret = 0;
9eddacf9 4941 tid_t target;
06a407f1 4942 bool needs_barrier = false;
8d5d02e6 4943 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4944
49598e04 4945 if (unlikely(ext4_forced_shutdown(sbi)))
0db1ff22
TT		 4946 return 0;
4947
9bffad1e 4948 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4949 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4950 /*
4951 * Writeback quota in non-journalled quota case - journalled quota has
4952 * no dirty dquots
4953 */
4954 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4955 /*
4956 * Data writeback is possible w/o journal transaction, so barrier must
4957 * being sent at the end of the function. But we can skip it if
4958 * transaction_commit will do it for us.
4959 */
bda32530
TT		 4960 if (sbi->s_journal) {
4961 target = jbd2_get_latest_transaction(sbi->s_journal);
4962 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4963 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4964 needs_barrier = true;
4965
4966 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4967 if (wait)
4968 ret = jbd2_log_wait_commit(sbi->s_journal,
4969 target);
4970 }
4971 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4972 needs_barrier = true;
06a407f1
DM		 4973 if (needs_barrier) {
4974 int err;
4975 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4976 if (!ret)
4977 ret = err;
0390131b 4978 }
06a407f1
DM		 4979
4980 return ret;
4981}
4982
ac27a0ec
DK		 4983/*
4984 * LVM calls this function before a (read-only) snapshot is created. This
4985 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4986 *
4987 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4988 * state independently. It relies on upper layer to stop all data & metadata
4989 * modifications.
ac27a0ec 4990 */
c4be0c1d 4991static int ext4_freeze(struct super_block *sb)
ac27a0ec 4992{
c4be0c1d
TS		 4993 int error = 0;
4994 journal_t *journal;
ac27a0ec 4995
bc98a42c 4996 if (sb_rdonly(sb))
9ca92389 4997 return 0;
ac27a0ec 4998
9ca92389 4999 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 5000
bb044576
TT		 5001 if (journal) {
5002 /* Now we set up the journal barrier. */
5003 jbd2_journal_lock_updates(journal);
ac27a0ec 5004
bb044576
TT		 5005 /*
5006 * Don't clear the needs_recovery flag if we failed to
5007 * flush the journal.
5008 */
5009 error = jbd2_journal_flush(journal);
5010 if (error < 0)
5011 goto out;
c642dc9e
ES		 5012
5013 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 5014 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 5015 }
9ca92389 5016
9ca92389 5017 error = ext4_commit_super(sb, 1);
6b0310fb 5018out:
bb044576
TT		 5019 if (journal)
5020 /* we rely on upper layer to stop further updates */
5021 jbd2_journal_unlock_updates(journal);
6b0310fb 5022 return error;
ac27a0ec
DK		 5023}
5024
5025/*
5026 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5027 * flag here, even though the filesystem is not technically dirty yet.
5028 */
c4be0c1d 5029static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 5030{
bc98a42c 5031 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 5032 return 0;
5033
c642dc9e
ES		 5034 if (EXT4_SB(sb)->s_journal) {
5035 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 5036 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 5037 }
5038
9ca92389 5039 ext4_commit_super(sb, 1);
c4be0c1d 5040 return 0;
ac27a0ec
DK		 5041}
5042
673c6100
TT		 5043/*
5044 * Structure to save mount options for ext4_remount's benefit
5045 */
5046struct ext4_mount_options {
5047 unsigned long s_mount_opt;
a2595b8a 5048 unsigned long s_mount_opt2;
08cefc7a
EB		 5049 kuid_t s_resuid;
5050 kgid_t s_resgid;
673c6100
TT		 5051 unsigned long s_commit_interval;
5052 u32 s_min_batch_time, s_max_batch_time;
5053#ifdef CONFIG_QUOTA
5054 int s_jquota_fmt;
a2d4a646 5055 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 5056#endif
5057};
5058
2b2d6d01 5059static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 5060{
2b2d6d01 5061 struct ext4_super_block *es;
617ba13b 5062 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5063 unsigned long old_sb_flags;
617ba13b 5064 struct ext4_mount_options old_opts;
c79d967d 5065 int enable_quota = 0;
8a266467 5066 ext4_group_t g;
b3881f74 5067 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 5068 int err = 0;
ac27a0ec 5069#ifdef CONFIG_QUOTA
03dafb5f 5070 int i, j;
ac27a0ec 5071#endif
d4c402d9 5072 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 5073
5074 /* Store the original options */
5075 old_sb_flags = sb->s_flags;
5076 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 5077 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 5078 old_opts.s_resuid = sbi->s_resuid;
5079 old_opts.s_resgid = sbi->s_resgid;
5080 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 5081 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5082 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 5083#ifdef CONFIG_QUOTA
5084 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 5085 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 5086 if (sbi->s_qf_names[i]) {
5087 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
5088 GFP_KERNEL);
5089 if (!old_opts.s_qf_names[i]) {
5090 for (j = 0; j < i; j++)
5091 kfree(old_opts.s_qf_names[j]);
3e36a163 5092 kfree(orig_data);
03dafb5f
CG		 5093 return -ENOMEM;
5094 }
5095 } else
5096 old_opts.s_qf_names[i] = NULL;
ac27a0ec 5097#endif
b3881f74
TT		 5098 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5099 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 5100
661aa520 5101 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 5102 err = -EINVAL;
5103 goto restore_opts;
5104 }
5105
6b992ff2 5106 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 5107 test_opt(sb, JOURNAL_CHECKSUM)) {
5108 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 5109 "during remount not supported; ignoring");
5110 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 5111 }
5112
6ae6514b
PS		 5113 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5114 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5115 ext4_msg(sb, KERN_ERR, "can't mount with "
5116 "both data=journal and delalloc");
5117 err = -EINVAL;
5118 goto restore_opts;
5119 }
5120 if (test_opt(sb, DIOREAD_NOLOCK)) {
5121 ext4_msg(sb, KERN_ERR, "can't mount with "
5122 "both data=journal and dioread_nolock");
5123 err = -EINVAL;
5124 goto restore_opts;
5125 }
923ae0ff
RZ		 5126 if (test_opt(sb, DAX)) {
5127 ext4_msg(sb, KERN_ERR, "can't mount with "
5128 "both data=journal and dax");
5129 err = -EINVAL;
5130 goto restore_opts;
5131 }
ab04df78
JK		 5132 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5133 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5134 ext4_msg(sb, KERN_ERR, "can't mount with "
5135 "journal_async_commit in data=ordered mode");
5136 err = -EINVAL;
5137 goto restore_opts;
5138 }
923ae0ff
RZ		 5139 }
5140
cdb7ee4c
TE		 5141 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5142 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5143 err = -EINVAL;
5144 goto restore_opts;
5145 }
5146
923ae0ff
RZ		 5147 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5148 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5149 "dax flag with busy inodes while remounting");
5150 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 5151 }
5152
4ab2f15b 5153 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 5154 ext4_abort(sb, "Abort forced by user");
ac27a0ec 5155
1751e8a6
LT		 5156 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5157 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec
DK		 5158
5159 es = sbi->s_es;
5160
b3881f74 5161 if (sbi->s_journal) {
0390131b 5162 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 5163 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5164 }
ac27a0ec 5165
1751e8a6
LT		 5166 if (*flags & SB_LAZYTIME)
5167 sb->s_flags |= SB_LAZYTIME;
a2fd66d0 5168
1751e8a6 5169 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
4ab2f15b 5170 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 5171 err = -EROFS;
5172 goto restore_opts;
5173 }
5174
1751e8a6 5175 if (*flags & SB_RDONLY) {
38c03b34
TT		 5176 err = sync_filesystem(sb);
5177 if (err < 0)
5178 goto restore_opts;
0f0dd62f
CH		 5179 err = dquot_suspend(sb, -1);
5180 if (err < 0)
c79d967d 5181 goto restore_opts;
c79d967d 5182
ac27a0ec
DK		 5183 /*
5184 * First of all, the unconditional stuff we have to do
5185 * to disable replay of the journal when we next remount
5186 */
1751e8a6 5187 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 5188
5189 /*
5190 * OK, test if we are remounting a valid rw partition
5191 * readonly, and if so set the rdonly flag and then
5192 * mark the partition as valid again.
5193 */
617ba13b
MC		 5194 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5195 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 5196 es->s_state = cpu_to_le16(sbi->s_mount_state);
5197
a63c9eb2 5198 if (sbi->s_journal)
0390131b 5199 ext4_mark_recovery_complete(sb, es);
ac27a0ec 5200 } else {
a13fb1a4 5201 /* Make sure we can mount this feature set readwrite */
e2b911c5 5202 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 5203 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 5204 err = -EROFS;
5205 goto restore_opts;
5206 }
8a266467
TT		 5207 /*
5208 * Make sure the group descriptor checksums
0b8e58a1 5209 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 5210 */
5211 for (g = 0; g < sbi->s_groups_count; g++) {
5212 struct ext4_group_desc *gdp =
5213 ext4_get_group_desc(sb, g, NULL);
5214
feb0ab32 5215 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5216 ext4_msg(sb, KERN_ERR,
5217 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 5218 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 5219 le16_to_cpu(gdp->bg_checksum));
6a797d27 5220 err = -EFSBADCRC;
8a266467
TT		 5221 goto restore_opts;
5222 }
5223 }
5224
ead6596b
ES		 5225 /*
5226 * If we have an unprocessed orphan list hanging
5227 * around from a previously readonly bdev mount,
5228 * require a full umount/remount for now.
5229 */
5230 if (es->s_last_orphan) {
b31e1552 5231 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5232 "remount RDWR because of unprocessed "
5233 "orphan inode list. Please "
b31e1552 5234 "umount/remount instead");
ead6596b
ES		 5235 err = -EINVAL;
5236 goto restore_opts;
5237 }
5238
ac27a0ec
DK		 5239 /*
5240 * Mounting a RDONLY partition read-write, so reread
5241 * and store the current valid flag. (It may have
5242 * been changed by e2fsck since we originally mounted
5243 * the partition.)
5244 */
0390131b
FM		 5245 if (sbi->s_journal)
5246 ext4_clear_journal_err(sb, es);
ac27a0ec 5247 sbi->s_mount_state = le16_to_cpu(es->s_state);
c89128a0
JK		 5248
5249 err = ext4_setup_super(sb, es, 0);
5250 if (err)
5251 goto restore_opts;
5252
5253 sb->s_flags &= ~SB_RDONLY;
e2b911c5 5254 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 5255 if (ext4_multi_mount_protect(sb,
5256 le64_to_cpu(es->s_mmp_block))) {
5257 err = -EROFS;
5258 goto restore_opts;
5259 }
c79d967d 5260 enable_quota = 1;
ac27a0ec
DK		 5261 }
5262 }
bfff6873
LC		 5263
5264 /*
5265 * Reinitialize lazy itable initialization thread based on
5266 * current settings
5267 */
bc98a42c 5268 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
bfff6873
LC		 5269 ext4_unregister_li_request(sb);
5270 else {
5271 ext4_group_t first_not_zeroed;
5272 first_not_zeroed = ext4_has_uninit_itable(sb);
5273 ext4_register_li_request(sb, first_not_zeroed);
5274 }
5275
6fd058f7 5276 ext4_setup_system_zone(sb);
c89128a0
JK		 5277 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5278 err = ext4_commit_super(sb, 1);
5279 if (err)
5280 goto restore_opts;
5281 }
0390131b 5282
ac27a0ec
DK		 5283#ifdef CONFIG_QUOTA
5284 /* Release old quota file names */
a2d4a646 5285 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5286 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5287 if (enable_quota) {
5288 if (sb_any_quota_suspended(sb))
5289 dquot_resume(sb, -1);
e2b911c5 5290 else if (ext4_has_feature_quota(sb)) {
7c319d32 5291 err = ext4_enable_quotas(sb);
07724f98 5292 if (err)
7c319d32 5293 goto restore_opts;
7c319d32
AK		 5294 }
5295 }
ac27a0ec 5296#endif
d4c402d9 5297
1751e8a6 5298 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
d4c402d9
CW		 5299 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5300 kfree(orig_data);
ac27a0ec 5301 return 0;
0b8e58a1 5302
ac27a0ec
DK		 5303restore_opts:
5304 sb->s_flags = old_sb_flags;
5305 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5306 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5307 sbi->s_resuid = old_opts.s_resuid;
5308 sbi->s_resgid = old_opts.s_resgid;
5309 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5310 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5311 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5312#ifdef CONFIG_QUOTA
5313 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5314 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5315 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5316 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5317 }
5318#endif
d4c402d9 5319 kfree(orig_data);
ac27a0ec
DK		 5320 return err;
5321}
5322
689c958c
LX		 5323#ifdef CONFIG_QUOTA
5324static int ext4_statfs_project(struct super_block *sb,
5325 kprojid_t projid, struct kstatfs *buf)
5326{
5327 struct kqid qid;
5328 struct dquot *dquot;
5329 u64 limit;
5330 u64 curblock;
5331
5332 qid = make_kqid_projid(projid);
5333 dquot = dqget(sb, qid);
5334 if (IS_ERR(dquot))
5335 return PTR_ERR(dquot);
7b9ca4c6 5336 spin_lock(&dquot->dq_dqb_lock);
689c958c
LX		 5337
5338 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5339 dquot->dq_dqb.dqb_bsoftlimit :
5340 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5341 if (limit && buf->f_blocks > limit) {
f06925c7
KK		 5342 curblock = (dquot->dq_dqb.dqb_curspace +
5343 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
689c958c
LX		 5344 buf->f_blocks = limit;
5345 buf->f_bfree = buf->f_bavail =
5346 (buf->f_blocks > curblock) ?
5347 (buf->f_blocks - curblock) : 0;
5348 }
5349
5350 limit = dquot->dq_dqb.dqb_isoftlimit ?
5351 dquot->dq_dqb.dqb_isoftlimit :
5352 dquot->dq_dqb.dqb_ihardlimit;
5353 if (limit && buf->f_files > limit) {
5354 buf->f_files = limit;
5355 buf->f_ffree =
5356 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5357 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5358 }
5359
7b9ca4c6 5360 spin_unlock(&dquot->dq_dqb_lock);
689c958c
LX		 5361 dqput(dquot);
5362 return 0;
5363}
5364#endif
5365
2b2d6d01 5366static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5367{
5368 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5369 struct ext4_sb_info *sbi = EXT4_SB(sb);
5370 struct ext4_super_block *es = sbi->s_es;
27dd4385 5371 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5372 u64 fsid;
d02a9391 5373 s64 bfree;
27dd4385 5374 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5375
952fc18e
TT		 5376 if (!test_opt(sb, MINIX_DF))
5377 overhead = sbi->s_overhead;
ac27a0ec 5378
617ba13b 5379 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5380 buf->f_bsize = sb->s_blocksize;
b72f78cb 5381 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5382 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5383 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5384 /* prevent underflow in case that few free space is available */
57042651 5385 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5386 buf->f_bavail = buf->f_bfree -
5387 (ext4_r_blocks_count(es) + resv_blocks);
5388 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5389 buf->f_bavail = 0;
5390 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5391 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5392 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5393 fsid = le64_to_cpup((void *)es->s_uuid) ^
5394 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5395 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5396 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5397
689c958c
LX		 5398#ifdef CONFIG_QUOTA
5399 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5400 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5401 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5402#endif
ac27a0ec
DK		 5403 return 0;
5404}
5405
ac27a0ec
DK		 5406
5407#ifdef CONFIG_QUOTA
5408
bc8230ee
JK		 5409/*
5410 * Helper functions so that transaction is started before we acquire dqio_sem
5411 * to keep correct lock ordering of transaction > dqio_sem
5412 */
ac27a0ec
DK		 5413static inline struct inode *dquot_to_inode(struct dquot *dquot)
5414{
4c376dca 5415 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5416}
5417
617ba13b 5418static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5419{
5420 int ret, err;
5421 handle_t *handle;
5422 struct inode *inode;
5423
5424 inode = dquot_to_inode(dquot);
9924a92a 5425 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5426 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5427 if (IS_ERR(handle))
5428 return PTR_ERR(handle);
5429 ret = dquot_commit(dquot);
617ba13b 5430 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5431 if (!ret)
5432 ret = err;
5433 return ret;
5434}
5435
617ba13b 5436static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5437{
5438 int ret, err;
5439 handle_t *handle;
5440
9924a92a 5441 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5442 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5443 if (IS_ERR(handle))
5444 return PTR_ERR(handle);
5445 ret = dquot_acquire(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_release_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_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5459 if (IS_ERR(handle)) {
5460 /* Release dquot anyway to avoid endless cycle in dqput() */
5461 dquot_release(dquot);
ac27a0ec 5462 return PTR_ERR(handle);
9c3013e9 5463 }
ac27a0ec 5464 ret = dquot_release(dquot);
617ba13b 5465 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5466 if (!ret)
5467 ret = err;
5468 return ret;
5469}
5470
617ba13b 5471static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5472{
262b4662
JK		 5473 struct super_block *sb = dquot->dq_sb;
5474 struct ext4_sb_info *sbi = EXT4_SB(sb);
5475
2c8be6b2 5476 /* Are we journaling quotas? */
e2b911c5 5477 if (ext4_has_feature_quota(sb) ||
262b4662 5478 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5479 dquot_mark_dquot_dirty(dquot);
617ba13b 5480 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5481 } else {
5482 return dquot_mark_dquot_dirty(dquot);
5483 }
5484}
5485
617ba13b 5486static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5487{
5488 int ret, err;
5489 handle_t *handle;
5490
5491 /* Data block + inode block */
2b0143b5 5492 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5493 if (IS_ERR(handle))
5494 return PTR_ERR(handle);
5495 ret = dquot_commit_info(sb, type);
617ba13b 5496 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5497 if (!ret)
5498 ret = err;
5499 return ret;
5500}
5501
5502/*
5503 * Turn on quotas during mount time - we need to find
5504 * the quota file and such...
5505 */
617ba13b 5506static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5507{
287a8095
CH		 5508 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5509 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5510}
5511
daf647d2
TT		 5512static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5513{
5514 struct ext4_inode_info *ei = EXT4_I(inode);
5515
5516 /* The first argument of lockdep_set_subclass has to be
5517 * *exactly* the same as the argument to init_rwsem() --- in
5518 * this case, in init_once() --- or lockdep gets unhappy
5519 * because the name of the lock is set using the
5520 * stringification of the argument to init_rwsem().
5521 */
5522 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5523 lockdep_set_subclass(&ei->i_data_sem, subclass);
5524}
5525
ac27a0ec
DK		 5526/*
5527 * Standard function to be called on quota_on
5528 */
617ba13b 5529static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 5530 const struct path *path)
ac27a0ec
DK		 5531{
5532 int err;
ac27a0ec
DK		 5533
5534 if (!test_opt(sb, QUOTA))
5535 return -EINVAL;
0623543b 5536
ac27a0ec 5537 /* Quotafile not on the same filesystem? */
d8c9584e 5538 if (path->dentry->d_sb != sb)
ac27a0ec 5539 return -EXDEV;
0623543b
JK		 5540 /* Journaling quota? */
5541 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5542 /* Quotafile not in fs root? */
f00c9e44 5543 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5544 ext4_msg(sb, KERN_WARNING,
5545 "Quota file not on filesystem root. "
5546 "Journaled quota will not work");
91389240
JK		 5547 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5548 } else {
5549 /*
5550 * Clear the flag just in case mount options changed since
5551 * last time.
5552 */
5553 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
2b2d6d01 5554 }
0623543b
JK		 5555
5556 /*
5557 * When we journal data on quota file, we have to flush journal to see
5558 * all updates to the file when we bypass pagecache...
5559 */
0390131b 5560 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5561 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5562 /*
5563 * We don't need to lock updates but journal_flush() could
5564 * otherwise be livelocked...
5565 */
5566 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5567 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5568 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5569 if (err)
7ffe1ea8 5570 return err;
0623543b 5571 }
957153fc 5572
daf647d2
TT		 5573 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5574 err = dquot_quota_on(sb, type, format_id, path);
957153fc 5575 if (err) {
daf647d2
TT		 5576 lockdep_set_quota_inode(path->dentry->d_inode,
5577 I_DATA_SEM_NORMAL);
957153fc
JK		 5578 } else {
5579 struct inode *inode = d_inode(path->dentry);
5580 handle_t *handle;
5581
61a92987
JK		 5582 /*
5583 * Set inode flags to prevent userspace from messing with quota
5584 * files. If this fails, we return success anyway since quotas
5585 * are already enabled and this is not a hard failure.
5586 */
957153fc
JK		 5587 inode_lock(inode);
5588 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5589 if (IS_ERR(handle))
5590 goto unlock_inode;
5591 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5592 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5593 S_NOATIME | S_IMMUTABLE);
5594 ext4_mark_inode_dirty(handle, inode);
5595 ext4_journal_stop(handle);
5596 unlock_inode:
5597 inode_unlock(inode);
5598 }
daf647d2 5599 return err;
ac27a0ec
DK		 5600}
5601
7c319d32
AK		 5602static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5603 unsigned int flags)
5604{
5605 int err;
5606 struct inode *qf_inode;
a2d4a646 5607 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5608 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5609 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5610 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5611 };
5612
e2b911c5 5613 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5614
5615 if (!qf_inums[type])
5616 return -EPERM;
5617
5618 qf_inode = ext4_iget(sb, qf_inums[type]);
5619 if (IS_ERR(qf_inode)) {
5620 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5621 return PTR_ERR(qf_inode);
5622 }
5623
bcb13850
JK		 5624 /* Don't account quota for quota files to avoid recursion */
5625 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5626 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7c319d32
AK		 5627 err = dquot_enable(qf_inode, type, format_id, flags);
5628 iput(qf_inode);
daf647d2
TT		 5629 if (err)
5630 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7c319d32
AK		 5631
5632 return err;
5633}
5634
5635/* Enable usage tracking for all quota types. */
5636static int ext4_enable_quotas(struct super_block *sb)
5637{
5638 int type, err = 0;
a2d4a646 5639 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5640 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5641 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5642 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 5643 };
49da9392
JK		 5644 bool quota_mopt[EXT4_MAXQUOTAS] = {
5645 test_opt(sb, USRQUOTA),
5646 test_opt(sb, GRPQUOTA),
5647 test_opt(sb, PRJQUOTA),
5648 };
7c319d32 5649
91389240 5650 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
a2d4a646 5651 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5652 if (qf_inums[type]) {
5653 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 5654 DQUOT_USAGE_ENABLED |
5655 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32 5656 if (err) {
95f1fda4 5657 for (type--; type >= 0; type--)
5658 dquot_quota_off(sb, type);
5659
7c319d32 5660 ext4_warning(sb,
72ba7450
TT		 5661 "Failed to enable quota tracking "
5662 "(type=%d, err=%d). Please run "
5663 "e2fsck to fix.", type, err);
7c319d32
AK		 5664 return err;
5665 }
5666 }
5667 }
5668 return 0;
5669}
5670
ca0e05e4
DM		 5671static int ext4_quota_off(struct super_block *sb, int type)
5672{
21f97697
JK		 5673 struct inode *inode = sb_dqopt(sb)->files[type];
5674 handle_t *handle;
957153fc 5675 int err;
21f97697 5676
87009d86
DM		 5677 /* Force all delayed allocation blocks to be allocated.
5678 * Caller already holds s_umount sem */
5679 if (test_opt(sb, DELALLOC))
ca0e05e4 5680 sync_filesystem(sb);
ca0e05e4 5681
957153fc 5682 if (!inode || !igrab(inode))
0b268590
AG		 5683 goto out;
5684
957153fc 5685 err = dquot_quota_off(sb, type);
964edf66 5686 if (err || ext4_has_feature_quota(sb))
957153fc
JK		 5687 goto out_put;
5688
5689 inode_lock(inode);
61a92987
JK		 5690 /*
5691 * Update modification times of quota files when userspace can
5692 * start looking at them. If we fail, we return success anyway since
5693 * this is not a hard failure and quotas are already disabled.
5694 */
9924a92a 5695 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697 5696 if (IS_ERR(handle))
957153fc
JK		 5697 goto out_unlock;
5698 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5699 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
eeca7ea1 5700 inode->i_mtime = inode->i_ctime = current_time(inode);
21f97697
JK		 5701 ext4_mark_inode_dirty(handle, inode);
5702 ext4_journal_stop(handle);
957153fc
JK		 5703out_unlock:
5704 inode_unlock(inode);
5705out_put:
964edf66 5706 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 5707 iput(inode);
5708 return err;
21f97697 5709out:
ca0e05e4
DM		 5710 return dquot_quota_off(sb, type);
5711}
5712
ac27a0ec
DK		 5713/* Read data from quotafile - avoid pagecache and such because we cannot afford
5714 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5715 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5716 * we don't have to be afraid of races */
617ba13b 5717static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5718 size_t len, loff_t off)
5719{
5720 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5721 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5722 int offset = off & (sb->s_blocksize - 1);
5723 int tocopy;
5724 size_t toread;
5725 struct buffer_head *bh;
5726 loff_t i_size = i_size_read(inode);
5727
5728 if (off > i_size)
5729 return 0;
5730 if (off+len > i_size)
5731 len = i_size-off;
5732 toread = len;
5733 while (toread > 0) {
5734 tocopy = sb->s_blocksize - offset < toread ?
5735 sb->s_blocksize - offset : toread;
1c215028
TT		 5736 bh = ext4_bread(NULL, inode, blk, 0);
5737 if (IS_ERR(bh))
5738 return PTR_ERR(bh);
ac27a0ec
DK		 5739 if (!bh) /* A hole? */
5740 memset(data, 0, tocopy);
5741 else
5742 memcpy(data, bh->b_data+offset, tocopy);
5743 brelse(bh);
5744 offset = 0;
5745 toread -= tocopy;
5746 data += tocopy;
5747 blk++;
5748 }
5749 return len;
5750}
5751
5752/* Write to quotafile (we know the transaction is already started and has
5753 * enough credits) */
617ba13b 5754static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5755 const char *data, size_t len, loff_t off)
5756{
5757 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5758 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5759 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5760 int retries = 0;
ac27a0ec
DK		 5761 struct buffer_head *bh;
5762 handle_t *handle = journal_current_handle();
5763
0390131b 5764 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5765 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5766 " cancelled because transaction is not started",
9c3013e9
JK		 5767 (unsigned long long)off, (unsigned long long)len);
5768 return -EIO;
5769 }
67eeb568
DM		 5770 /*
5771 * Since we account only one data block in transaction credits,
5772 * then it is impossible to cross a block boundary.
5773 */
5774 if (sb->s_blocksize - offset < len) {
5775 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5776 " cancelled because not block aligned",
5777 (unsigned long long)off, (unsigned long long)len);
5778 return -EIO;
5779 }
5780
c5e298ae
TT		 5781 do {
5782 bh = ext4_bread(handle, inode, blk,
5783 EXT4_GET_BLOCKS_CREATE |
5784 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5785 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5786 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5787 if (IS_ERR(bh))
5788 return PTR_ERR(bh);
67eeb568
DM		 5789 if (!bh)
5790 goto out;
5d601255 5791 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5792 err = ext4_journal_get_write_access(handle, bh);
5793 if (err) {
5794 brelse(bh);
1c215028 5795 return err;
ac27a0ec 5796 }
67eeb568
DM		 5797 lock_buffer(bh);
5798 memcpy(bh->b_data+offset, data, len);
5799 flush_dcache_page(bh->b_page);
5800 unlock_buffer(bh);
62d2b5f2 5801 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5802 brelse(bh);
ac27a0ec 5803out:
67eeb568
DM		 5804 if (inode->i_size < off + len) {
5805 i_size_write(inode, off + len);
617ba13b 5806 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5807 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5808 }
67eeb568 5809 return len;
ac27a0ec
DK		 5810}
5811
8f0e8746
TT		 5812static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5813{
5814 const struct quota_format_ops *ops;
5815
5816 if (!sb_has_quota_loaded(sb, qid->type))
5817 return -ESRCH;
5818 ops = sb_dqopt(sb)->ops[qid->type];
5819 if (!ops || !ops->get_next_id)
5820 return -ENOSYS;
5821 return dquot_get_next_id(sb, qid);
5822}
ac27a0ec
DK		 5823#endif
5824
152a0836
AV		 5825static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5826 const char *dev_name, void *data)
ac27a0ec 5827{
152a0836 5828 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5829}
5830
c290ea01 5831#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5832static inline void register_as_ext2(void)
5833{
5834 int err = register_filesystem(&ext2_fs_type);
5835 if (err)
5836 printk(KERN_WARNING
5837 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5838}
5839
5840static inline void unregister_as_ext2(void)
5841{
5842 unregister_filesystem(&ext2_fs_type);
5843}
2035e776
TT		 5844
5845static inline int ext2_feature_set_ok(struct super_block *sb)
5846{
e2b911c5 5847 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776 5848 return 0;
bc98a42c 5849 if (sb_rdonly(sb))
2035e776 5850 return 1;
e2b911c5 5851 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5852 return 0;
5853 return 1;
5854}
24b58424
TT		 5855#else
5856static inline void register_as_ext2(void) { }
5857static inline void unregister_as_ext2(void) { }
2035e776 5858static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5859#endif
5860
24b58424
TT		 5861static inline void register_as_ext3(void)
5862{
5863 int err = register_filesystem(&ext3_fs_type);
5864 if (err)
5865 printk(KERN_WARNING
5866 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5867}
5868
5869static inline void unregister_as_ext3(void)
5870{
5871 unregister_filesystem(&ext3_fs_type);
5872}
2035e776
TT		 5873
5874static inline int ext3_feature_set_ok(struct super_block *sb)
5875{
e2b911c5 5876 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5877 return 0;
e2b911c5 5878 if (!ext4_has_feature_journal(sb))
2035e776 5879 return 0;
bc98a42c 5880 if (sb_rdonly(sb))
2035e776 5881 return 1;
e2b911c5 5882 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5883 return 0;
5884 return 1;
5885}
24b58424 5886
03010a33
TT		 5887static struct file_system_type ext4_fs_type = {
5888 .owner = THIS_MODULE,
5889 .name = "ext4",
152a0836 5890 .mount = ext4_mount,
03010a33
TT		 5891 .kill_sb = kill_block_super,
5892 .fs_flags = FS_REQUIRES_DEV,
5893};
7f78e035 5894MODULE_ALIAS_FS("ext4");
03010a33 5895
e9e3bcec
ES		 5896/* Shared across all ext4 file systems */
5897wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 5898
5dabfc78 5899static int __init ext4_init_fs(void)
ac27a0ec 5900{
e9e3bcec 5901 int i, err;
c9de560d 5902
e294a537 5903 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5904 ext4_li_info = NULL;
5905 mutex_init(&ext4_li_mtx);
5906
9a4c8019 5907 /* Build-time check for flags consistency */
12e9b892 5908 ext4_check_flag_values();
e9e3bcec 5909
e142d052 5910 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 5911 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 5912
51865fda 5913 err = ext4_init_es();
6fd058f7
TT		 5914 if (err)
5915 return err;
51865fda
ZL		 5916
5917 err = ext4_init_pageio();
5918 if (err)
b5799018 5919 goto out5;
51865fda 5920
5dabfc78 5921 err = ext4_init_system_zone();
bd2d0210 5922 if (err)
b5799018 5923 goto out4;
857ac889 5924
b5799018 5925 err = ext4_init_sysfs();
dd68314c 5926 if (err)
b5799018 5927 goto out3;
857ac889 5928
5dabfc78 5929 err = ext4_init_mballoc();
c9de560d
AT		 5930 if (err)
5931 goto out2;
ac27a0ec
DK		 5932 err = init_inodecache();
5933 if (err)
5934 goto out1;
24b58424 5935 register_as_ext3();
2035e776 5936 register_as_ext2();
03010a33 5937 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5938 if (err)
5939 goto out;
bfff6873 5940
ac27a0ec
DK		 5941 return 0;
5942out:
24b58424
TT		 5943 unregister_as_ext2();
5944 unregister_as_ext3();
ac27a0ec
DK		 5945 destroy_inodecache();
5946out1:
5dabfc78 5947 ext4_exit_mballoc();
9c191f70 5948out2:
b5799018
TT		 5949 ext4_exit_sysfs();
5950out3:
5dabfc78 5951 ext4_exit_system_zone();
b5799018 5952out4:
5dabfc78 5953 ext4_exit_pageio();
b5799018 5954out5:
51865fda
ZL		 5955 ext4_exit_es();
5956
ac27a0ec
DK		 5957 return err;
5958}
5959
5dabfc78 5960static void __exit ext4_exit_fs(void)
ac27a0ec 5961{
bfff6873 5962 ext4_destroy_lazyinit_thread();
24b58424
TT		 5963 unregister_as_ext2();
5964 unregister_as_ext3();
03010a33 5965 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5966 destroy_inodecache();
5dabfc78 5967 ext4_exit_mballoc();
b5799018 5968 ext4_exit_sysfs();
5dabfc78
TT		 5969 ext4_exit_system_zone();
5970 ext4_exit_pageio();
dd12ed14 5971 ext4_exit_es();
ac27a0ec
DK		 5972}
5973
5974MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5975MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5976MODULE_LICENSE("GPL");
7ef79ad5 5977MODULE_SOFTDEP("pre: crc32c");
5dabfc78
TT		 5978module_init(ext4_init_fs)
5979module_exit(ext4_exit_fs)
A mirror of Linus' kernel repository