projects / GitHub / LineageOS / android_kernel_motorola_exynos9610.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: refactor sysfs support code
[GitHub/LineageOS/android_kernel_motorola_exynos9610.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
66114cad 27#include <linux/backing-dev.h>
ac27a0ec 28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
7abc52c2 41#include <linux/cleancache.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
4a092d73 48#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 49#include "ext4_jbd2.h"
ac27a0ec
DK		 50#include "xattr.h"
51#include "acl.h"
3661d286 52#include "mballoc.h"
ac27a0ec 53
9bffad1e
TT		 54#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
1f109d5a 57static struct proc_dir_entry *ext4_proc_root;
0b75a840
LC		 58static struct ext4_lazy_init *ext4_li_info;
59static struct mutex ext4_li_mtx;
9c191f70 60static int ext4_mballoc_ready;
e294a537 61static struct ratelimit_state ext4_mount_msg_ratelimit;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
2adf6da8 65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 66static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 67static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
617ba13b 71static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 72static int ext4_remount(struct super_block *sb, int *flags, char *data);
73static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 74static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 75static int ext4_freeze(struct super_block *sb);
152a0836
AV		 76static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
2035e776
TT		 78static inline int ext2_feature_set_ok(struct super_block *sb);
79static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 80static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 81static void ext4_destroy_lazyinit_thread(void);
82static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 83static void ext4_clear_request_list(void);
ac27a0ec 84
c290ea01 85#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
2035e776
TT		 86static struct file_system_type ext2_fs_type = {
87 .owner = THIS_MODULE,
88 .name = "ext2",
89 .mount = ext4_mount,
90 .kill_sb = kill_block_super,
91 .fs_flags = FS_REQUIRES_DEV,
92};
7f78e035 93MODULE_ALIAS_FS("ext2");
fa7614dd 94MODULE_ALIAS("ext2");
2035e776
TT		 95#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
96#else
97#define IS_EXT2_SB(sb) (0)
98#endif
99
100
ba69f9ab
JK		 101static struct file_system_type ext3_fs_type = {
102 .owner = THIS_MODULE,
103 .name = "ext3",
152a0836 104 .mount = ext4_mount,
ba69f9ab
JK		 105 .kill_sb = kill_block_super,
106 .fs_flags = FS_REQUIRES_DEV,
107};
7f78e035 108MODULE_ALIAS_FS("ext3");
fa7614dd 109MODULE_ALIAS("ext3");
ba69f9ab 110#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
bd81d8ee 111
d25425f8
DW		 112static int ext4_verify_csum_type(struct super_block *sb,
113 struct ext4_super_block *es)
114{
115 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
116 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
117 return 1;
118
119 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
120}
121
a9c47317
DW		 122static __le32 ext4_superblock_csum(struct super_block *sb,
123 struct ext4_super_block *es)
124{
125 struct ext4_sb_info *sbi = EXT4_SB(sb);
126 int offset = offsetof(struct ext4_super_block, s_checksum);
127 __u32 csum;
128
129 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
130
131 return cpu_to_le32(csum);
132}
133
c197855e
SH		 134static int ext4_superblock_csum_verify(struct super_block *sb,
135 struct ext4_super_block *es)
a9c47317 136{
9aa5d32b 137 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 138 return 1;
139
140 return es->s_checksum == ext4_superblock_csum(sb, es);
141}
142
06db49e6 143void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 144{
06db49e6
TT		 145 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
146
9aa5d32b 147 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 148 return;
149
150 es->s_checksum = ext4_superblock_csum(sb, es);
151}
152
9933fc0a
TT		 153void *ext4_kvmalloc(size_t size, gfp_t flags)
154{
155 void *ret;
156
8be04b93 157 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 158 if (!ret)
159 ret = __vmalloc(size, flags, PAGE_KERNEL);
160 return ret;
161}
162
163void *ext4_kvzalloc(size_t size, gfp_t flags)
164{
165 void *ret;
166
8be04b93 167 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 168 if (!ret)
169 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
170 return ret;
171}
172
8fadc143
AR		 173ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
174 struct ext4_group_desc *bg)
bd81d8ee 175{
3a14589c 176 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 177 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 178 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 179}
180
8fadc143
AR		 181ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
182 struct ext4_group_desc *bg)
bd81d8ee 183{
5272f837 184 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 185 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 186 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 187}
188
8fadc143
AR		 189ext4_fsblk_t ext4_inode_table(struct super_block *sb,
190 struct ext4_group_desc *bg)
bd81d8ee 191{
5272f837 192 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 193 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 194 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 195}
196
021b65bb
TT		 197__u32 ext4_free_group_clusters(struct super_block *sb,
198 struct ext4_group_desc *bg)
560671a0
AK		 199{
200 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
201 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 202 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 203}
204
205__u32 ext4_free_inodes_count(struct super_block *sb,
206 struct ext4_group_desc *bg)
207{
208 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
209 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 210 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 211}
212
213__u32 ext4_used_dirs_count(struct super_block *sb,
214 struct ext4_group_desc *bg)
215{
216 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
217 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 218 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 219}
220
221__u32 ext4_itable_unused_count(struct super_block *sb,
222 struct ext4_group_desc *bg)
223{
224 return le16_to_cpu(bg->bg_itable_unused_lo) |
225 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 226 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 227}
228
8fadc143
AR		 229void ext4_block_bitmap_set(struct super_block *sb,
230 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 231{
3a14589c 232 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 233 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
234 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 235}
236
8fadc143
AR		 237void ext4_inode_bitmap_set(struct super_block *sb,
238 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 239{
5272f837 240 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 241 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
242 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 243}
244
8fadc143
AR		 245void ext4_inode_table_set(struct super_block *sb,
246 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 247{
5272f837 248 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 249 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
250 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 251}
252
021b65bb
TT		 253void ext4_free_group_clusters_set(struct super_block *sb,
254 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 255{
256 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
257 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
258 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
259}
260
261void ext4_free_inodes_set(struct super_block *sb,
262 struct ext4_group_desc *bg, __u32 count)
263{
264 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
265 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
266 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
267}
268
269void ext4_used_dirs_set(struct super_block *sb,
270 struct ext4_group_desc *bg, __u32 count)
271{
272 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
273 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
274 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
275}
276
277void ext4_itable_unused_set(struct super_block *sb,
278 struct ext4_group_desc *bg, __u32 count)
279{
280 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
281 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
282 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
283}
284
d3d1faf6 285
1c13d5c0
TT		 286static void __save_error_info(struct super_block *sb, const char *func,
287 unsigned int line)
288{
289 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
290
291 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1b46617b
TT		 292 if (bdev_read_only(sb->s_bdev))
293 return;
1c13d5c0
TT		 294 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
295 es->s_last_error_time = cpu_to_le32(get_seconds());
296 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
297 es->s_last_error_line = cpu_to_le32(line);
298 if (!es->s_first_error_time) {
299 es->s_first_error_time = es->s_last_error_time;
300 strncpy(es->s_first_error_func, func,
301 sizeof(es->s_first_error_func));
302 es->s_first_error_line = cpu_to_le32(line);
303 es->s_first_error_ino = es->s_last_error_ino;
304 es->s_first_error_block = es->s_last_error_block;
305 }
66e61a9e
TT		 306 /*
307 * Start the daily error reporting function if it hasn't been
308 * started already
309 */
310 if (!es->s_error_count)
311 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 312 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 313}
314
315static void save_error_info(struct super_block *sb, const char *func,
316 unsigned int line)
317{
318 __save_error_info(sb, func, line);
319 ext4_commit_super(sb, 1);
320}
321
bdfe0cbd
TT		 322/*
323 * The del_gendisk() function uninitializes the disk-specific data
324 * structures, including the bdi structure, without telling anyone
325 * else. Once this happens, any attempt to call mark_buffer_dirty()
326 * (for example, by ext4_commit_super), will cause a kernel OOPS.
327 * This is a kludge to prevent these oops until we can put in a proper
328 * hook in del_gendisk() to inform the VFS and file system layers.
329 */
330static int block_device_ejected(struct super_block *sb)
331{
332 struct inode *bd_inode = sb->s_bdev->bd_inode;
333 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
334
335 return bdi->dev == NULL;
336}
337
18aadd47
BJ		 338static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
339{
340 struct super_block *sb = journal->j_private;
341 struct ext4_sb_info *sbi = EXT4_SB(sb);
342 int error = is_journal_aborted(journal);
5d3ee208 343 struct ext4_journal_cb_entry *jce;
18aadd47 344
5d3ee208 345 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 346 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 347 while (!list_empty(&txn->t_private_list)) {
348 jce = list_entry(txn->t_private_list.next,
349 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 350 list_del_init(&jce->jce_list);
351 spin_unlock(&sbi->s_md_lock);
352 jce->jce_func(sb, jce, error);
353 spin_lock(&sbi->s_md_lock);
354 }
355 spin_unlock(&sbi->s_md_lock);
356}
1c13d5c0 357
ac27a0ec
DK		 358/* Deal with the reporting of failure conditions on a filesystem such as
359 * inconsistencies detected or read IO failures.
360 *
361 * On ext2, we can store the error state of the filesystem in the
617ba13b 362 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 363 * write ordering constraints on the superblock which prevent us from
364 * writing it out straight away; and given that the journal is about to
365 * be aborted, we can't rely on the current, or future, transactions to
366 * write out the superblock safely.
367 *
dab291af 368 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 369 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 370 * that error until we've noted it down and cleared it.
371 */
372
617ba13b 373static void ext4_handle_error(struct super_block *sb)
ac27a0ec 374{
ac27a0ec
DK		 375 if (sb->s_flags & MS_RDONLY)
376 return;
377
2b2d6d01 378 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 379 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 380
4ab2f15b 381 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 382 if (journal)
dab291af 383 jbd2_journal_abort(journal, -EIO);
ac27a0ec 384 }
2b2d6d01 385 if (test_opt(sb, ERRORS_RO)) {
b31e1552 386 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 387 /*
388 * Make sure updated value of ->s_mount_flags will be visible
389 * before ->s_flags update
390 */
391 smp_wmb();
ac27a0ec
DK		 392 sb->s_flags |= MS_RDONLY;
393 }
ac27a0ec 394 if (test_opt(sb, ERRORS_PANIC))
617ba13b 395 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 396 sb->s_id);
397}
398
efbed4dc
TT		 399#define ext4_error_ratelimit(sb) \
400 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
401 "EXT4-fs error")
402
12062ddd 403void __ext4_error(struct super_block *sb, const char *function,
c398eda0 404 unsigned int line, const char *fmt, ...)
ac27a0ec 405{
0ff2ea7d 406 struct va_format vaf;
ac27a0ec
DK		 407 va_list args;
408
efbed4dc
TT		 409 if (ext4_error_ratelimit(sb)) {
410 va_start(args, fmt);
411 vaf.fmt = fmt;
412 vaf.va = &args;
413 printk(KERN_CRIT
414 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
415 sb->s_id, function, line, current->comm, &vaf);
416 va_end(args);
417 }
f3fc0210 418 save_error_info(sb, function, line);
617ba13b 419 ext4_handle_error(sb);
ac27a0ec
DK		 420}
421
e7c96e8e
JP		 422void __ext4_error_inode(struct inode *inode, const char *function,
423 unsigned int line, ext4_fsblk_t block,
424 const char *fmt, ...)
273df556
FM		 425{
426 va_list args;
f7c21177 427 struct va_format vaf;
1c13d5c0 428 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 429
1c13d5c0
TT		 430 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
431 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 432 if (ext4_error_ratelimit(inode->i_sb)) {
433 va_start(args, fmt);
434 vaf.fmt = fmt;
435 vaf.va = &args;
436 if (block)
437 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
438 "inode #%lu: block %llu: comm %s: %pV\n",
439 inode->i_sb->s_id, function, line, inode->i_ino,
440 block, current->comm, &vaf);
441 else
442 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
443 "inode #%lu: comm %s: %pV\n",
444 inode->i_sb->s_id, function, line, inode->i_ino,
445 current->comm, &vaf);
446 va_end(args);
447 }
1c13d5c0 448 save_error_info(inode->i_sb, function, line);
273df556
FM		 449 ext4_handle_error(inode->i_sb);
450}
451
e7c96e8e
JP		 452void __ext4_error_file(struct file *file, const char *function,
453 unsigned int line, ext4_fsblk_t block,
454 const char *fmt, ...)
273df556
FM		 455{
456 va_list args;
f7c21177 457 struct va_format vaf;
1c13d5c0 458 struct ext4_super_block *es;
496ad9aa 459 struct inode *inode = file_inode(file);
273df556
FM		 460 char pathname[80], *path;
461
1c13d5c0
TT		 462 es = EXT4_SB(inode->i_sb)->s_es;
463 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 464 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 465 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 466 if (IS_ERR(path))
467 path = "(unknown)";
468 va_start(args, fmt);
469 vaf.fmt = fmt;
470 vaf.va = &args;
471 if (block)
472 printk(KERN_CRIT
473 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
474 "block %llu: comm %s: path %s: %pV\n",
475 inode->i_sb->s_id, function, line, inode->i_ino,
476 block, current->comm, path, &vaf);
477 else
478 printk(KERN_CRIT
479 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
480 "comm %s: path %s: %pV\n",
481 inode->i_sb->s_id, function, line, inode->i_ino,
482 current->comm, path, &vaf);
483 va_end(args);
484 }
1c13d5c0 485 save_error_info(inode->i_sb, function, line);
273df556
FM		 486 ext4_handle_error(inode->i_sb);
487}
488
722887dd
TT		 489const char *ext4_decode_error(struct super_block *sb, int errno,
490 char nbuf[16])
ac27a0ec
DK		 491{
492 char *errstr = NULL;
493
494 switch (errno) {
495 case -EIO:
496 errstr = "IO failure";
497 break;
498 case -ENOMEM:
499 errstr = "Out of memory";
500 break;
501 case -EROFS:
78f1ddbb
TT		 502 if (!sb || (EXT4_SB(sb)->s_journal &&
503 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 504 errstr = "Journal has aborted";
505 else
506 errstr = "Readonly filesystem";
507 break;
508 default:
509 /* If the caller passed in an extra buffer for unknown
510 * errors, textualise them now. Else we just return
511 * NULL. */
512 if (nbuf) {
513 /* Check for truncated error codes... */
514 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
515 errstr = nbuf;
516 }
517 break;
518 }
519
520 return errstr;
521}
522
617ba13b 523/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 524 * automatically and invokes the appropriate error response. */
525
c398eda0
TT		 526void __ext4_std_error(struct super_block *sb, const char *function,
527 unsigned int line, int errno)
ac27a0ec
DK		 528{
529 char nbuf[16];
530 const char *errstr;
531
532 /* Special case: if the error is EROFS, and we're not already
533 * inside a transaction, then there's really no point in logging
534 * an error. */
535 if (errno == -EROFS && journal_current_handle() == NULL &&
536 (sb->s_flags & MS_RDONLY))
537 return;
538
efbed4dc
TT		 539 if (ext4_error_ratelimit(sb)) {
540 errstr = ext4_decode_error(sb, errno, nbuf);
541 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
542 sb->s_id, function, line, errstr);
543 }
ac27a0ec 544
efbed4dc 545 save_error_info(sb, function, line);
617ba13b 546 ext4_handle_error(sb);
ac27a0ec
DK		 547}
548
549/*
617ba13b 550 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 551 * abort function may be used to deal with unrecoverable failures such
552 * as journal IO errors or ENOMEM at a critical moment in log management.
553 *
554 * We unconditionally force the filesystem into an ABORT|READONLY state,
555 * unless the error response on the fs has been set to panic in which
556 * case we take the easy way out and panic immediately.
557 */
558
c67d859e 559void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 560 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 561{
562 va_list args;
563
1c13d5c0 564 save_error_info(sb, function, line);
ac27a0ec 565 va_start(args, fmt);
c398eda0
TT		 566 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
567 function, line);
ac27a0ec
DK		 568 vprintk(fmt, args);
569 printk("\n");
570 va_end(args);
571
1c13d5c0
TT		 572 if ((sb->s_flags & MS_RDONLY) == 0) {
573 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 574 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 575 /*
576 * Make sure updated value of ->s_mount_flags will be visible
577 * before ->s_flags update
578 */
579 smp_wmb();
580 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 581 if (EXT4_SB(sb)->s_journal)
582 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
583 save_error_info(sb, function, line);
584 }
ac27a0ec 585 if (test_opt(sb, ERRORS_PANIC))
617ba13b 586 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 587}
588
e7c96e8e
JP		 589void __ext4_msg(struct super_block *sb,
590 const char *prefix, const char *fmt, ...)
b31e1552 591{
0ff2ea7d 592 struct va_format vaf;
b31e1552
ES		 593 va_list args;
594
efbed4dc
TT		 595 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
596 return;
597
b31e1552 598 va_start(args, fmt);
0ff2ea7d
JP		 599 vaf.fmt = fmt;
600 vaf.va = &args;
601 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 602 va_end(args);
603}
604
b03a2f7e
AD		 605#define ext4_warning_ratelimit(sb) \
606 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
607 "EXT4-fs warning")
608
12062ddd 609void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 610 unsigned int line, const char *fmt, ...)
ac27a0ec 611{
0ff2ea7d 612 struct va_format vaf;
ac27a0ec
DK		 613 va_list args;
614
b03a2f7e 615 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 616 return;
617
ac27a0ec 618 va_start(args, fmt);
0ff2ea7d
JP		 619 vaf.fmt = fmt;
620 vaf.va = &args;
621 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
622 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 623 va_end(args);
624}
625
b03a2f7e
AD		 626void __ext4_warning_inode(const struct inode *inode, const char *function,
627 unsigned int line, const char *fmt, ...)
628{
629 struct va_format vaf;
630 va_list args;
631
632 if (!ext4_warning_ratelimit(inode->i_sb))
633 return;
634
635 va_start(args, fmt);
636 vaf.fmt = fmt;
637 vaf.va = &args;
638 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
639 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
640 function, line, inode->i_ino, current->comm, &vaf);
641 va_end(args);
642}
643
e29136f8
TT		 644void __ext4_grp_locked_error(const char *function, unsigned int line,
645 struct super_block *sb, ext4_group_t grp,
646 unsigned long ino, ext4_fsblk_t block,
647 const char *fmt, ...)
5d1b1b3f
AK		 648__releases(bitlock)
649__acquires(bitlock)
650{
0ff2ea7d 651 struct va_format vaf;
5d1b1b3f
AK		 652 va_list args;
653 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
654
1c13d5c0
TT		 655 es->s_last_error_ino = cpu_to_le32(ino);
656 es->s_last_error_block = cpu_to_le64(block);
657 __save_error_info(sb, function, line);
0ff2ea7d 658
efbed4dc
TT		 659 if (ext4_error_ratelimit(sb)) {
660 va_start(args, fmt);
661 vaf.fmt = fmt;
662 vaf.va = &args;
663 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
664 sb->s_id, function, line, grp);
665 if (ino)
666 printk(KERN_CONT "inode %lu: ", ino);
667 if (block)
668 printk(KERN_CONT "block %llu:",
669 (unsigned long long) block);
670 printk(KERN_CONT "%pV\n", &vaf);
671 va_end(args);
672 }
5d1b1b3f
AK		 673
674 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 675 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 676 return;
677 }
1c13d5c0 678
5d1b1b3f
AK		 679 ext4_unlock_group(sb, grp);
680 ext4_handle_error(sb);
681 /*
682 * We only get here in the ERRORS_RO case; relocking the group
683 * may be dangerous, but nothing bad will happen since the
684 * filesystem will have already been marked read/only and the
685 * journal has been aborted. We return 1 as a hint to callers
686 * who might what to use the return value from
25985edc 687 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 688 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
689 * aggressively from the ext4 function in question, with a
690 * more appropriate error code.
691 */
692 ext4_lock_group(sb, grp);
693 return;
694}
695
617ba13b 696void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 697{
617ba13b 698 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 699
617ba13b 700 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 701 return;
702
12062ddd 703 ext4_warning(sb,
ac27a0ec
DK		 704 "updating to rev %d because of new feature flag, "
705 "running e2fsck is recommended",
617ba13b 706 EXT4_DYNAMIC_REV);
ac27a0ec 707
617ba13b
MC		 708 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
709 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
710 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 711 /* leave es->s_feature_*compat flags alone */
712 /* es->s_uuid will be set by e2fsck if empty */
713
714 /*
715 * The rest of the superblock fields should be zero, and if not it
716 * means they are likely already in use, so leave them alone. We
717 * can leave it up to e2fsck to clean up any inconsistencies there.
718 */
719}
720
721/*
722 * Open the external journal device
723 */
b31e1552 724static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 725{
726 struct block_device *bdev;
727 char b[BDEVNAME_SIZE];
728
d4d77629 729 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 730 if (IS_ERR(bdev))
731 goto fail;
732 return bdev;
733
734fail:
b31e1552 735 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 736 __bdevname(dev, b), PTR_ERR(bdev));
737 return NULL;
738}
739
740/*
741 * Release the journal device
742 */
4385bab1 743static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 744{
4385bab1 745 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 746}
747
4385bab1 748static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 749{
750 struct block_device *bdev;
ac27a0ec
DK		 751 bdev = sbi->journal_bdev;
752 if (bdev) {
4385bab1 753 ext4_blkdev_put(bdev);
ac27a0ec
DK		 754 sbi->journal_bdev = NULL;
755 }
ac27a0ec
DK		 756}
757
758static inline struct inode *orphan_list_entry(struct list_head *l)
759{
617ba13b 760 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 761}
762
617ba13b 763static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 764{
765 struct list_head *l;
766
b31e1552
ES		 767 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
768 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 769
770 printk(KERN_ERR "sb_info orphan list:\n");
771 list_for_each(l, &sbi->s_orphan) {
772 struct inode *inode = orphan_list_entry(l);
773 printk(KERN_ERR " "
774 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
775 inode->i_sb->s_id, inode->i_ino, inode,
776 inode->i_mode, inode->i_nlink,
777 NEXT_ORPHAN(inode));
778 }
779}
780
2b2d6d01 781static void ext4_put_super(struct super_block *sb)
ac27a0ec 782{
617ba13b
MC		 783 struct ext4_sb_info *sbi = EXT4_SB(sb);
784 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 785 int i, err;
ac27a0ec 786
857ac889 787 ext4_unregister_li_request(sb);
e0ccfd95
CH		 788 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
789
2e8fa54e 790 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 791 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 792
0390131b
FM		 793 if (sbi->s_journal) {
794 err = jbd2_journal_destroy(sbi->s_journal);
795 sbi->s_journal = NULL;
796 if (err < 0)
c67d859e 797 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 798 }
d4edac31 799
d3922a77 800 ext4_es_unregister_shrinker(sbi);
9105bb14 801 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 802 ext4_release_system_zone(sb);
803 ext4_mb_release(sb);
804 ext4_ext_release(sb);
805 ext4_xattr_put_super(sb);
806
ac27a0ec 807 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 808 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 809 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 810 }
58c5873a 811 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 812 ext4_commit_super(sb, 1);
813
240799cd 814 if (sbi->s_proc) {
66acdcf4 815 remove_proc_entry("options", sbi->s_proc);
9f6200bb 816 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 817 }
3197ebdb 818 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 819
820 for (i = 0; i < sbi->s_gdb_count; i++)
821 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 822 kvfree(sbi->s_group_desc);
823 kvfree(sbi->s_flex_groups);
57042651 824 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 825 percpu_counter_destroy(&sbi->s_freeinodes_counter);
826 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 827 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 828 brelse(sbi->s_sbh);
829#ifdef CONFIG_QUOTA
a2d4a646 830 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 831 kfree(sbi->s_qf_names[i]);
832#endif
833
834 /* Debugging code just in case the in-memory inode orphan list
835 * isn't empty. The on-disk one can be non-empty if we've
836 * detected an error and taken the fs readonly, but the
837 * in-memory list had better be clean by this point. */
838 if (!list_empty(&sbi->s_orphan))
839 dump_orphan_list(sb, sbi);
840 J_ASSERT(list_empty(&sbi->s_orphan));
841
89d96a6f 842 sync_blockdev(sb->s_bdev);
f98393a6 843 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 844 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
845 /*
846 * Invalidate the journal device's buffers. We don't want them
847 * floating about in memory - the physical journal device may
848 * hotswapped, and it breaks the `ro-after' testing code.
849 */
850 sync_blockdev(sbi->journal_bdev);
f98393a6 851 invalidate_bdev(sbi->journal_bdev);
617ba13b 852 ext4_blkdev_remove(sbi);
ac27a0ec 853 }
9c191f70
M		 854 if (sbi->s_mb_cache) {
855 ext4_xattr_destroy_cache(sbi->s_mb_cache);
856 sbi->s_mb_cache = NULL;
857 }
c5e06d10
JL		 858 if (sbi->s_mmp_tsk)
859 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 860 sb->s_fs_info = NULL;
3197ebdb
TT		 861 /*
862 * Now that we are completely done shutting down the
863 * superblock, we need to actually destroy the kobject.
864 */
3197ebdb
TT		 865 kobject_put(&sbi->s_kobj);
866 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 867 if (sbi->s_chksum_driver)
868 crypto_free_shash(sbi->s_chksum_driver);
705895b6 869 kfree(sbi->s_blockgroup_lock);
ac27a0ec 870 kfree(sbi);
ac27a0ec
DK		 871}
872
e18b890b 873static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 874
875/*
876 * Called inside transaction, so use GFP_NOFS
877 */
617ba13b 878static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 879{
617ba13b 880 struct ext4_inode_info *ei;
ac27a0ec 881
e6b4f8da 882 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 883 if (!ei)
884 return NULL;
0b8e58a1 885
ac27a0ec 886 ei->vfs_inode.i_version = 1;
202ee5df 887 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 888 INIT_LIST_HEAD(&ei->i_prealloc_list);
889 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 890 ext4_es_init_tree(&ei->i_es_tree);
891 rwlock_init(&ei->i_es_lock);
edaa53ca 892 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 893 ei->i_es_all_nr = 0;
edaa53ca 894 ei->i_es_shk_nr = 0;
dd475925 895 ei->i_es_shrink_lblk = 0;
d2a17637
MC		 896 ei->i_reserved_data_blocks = 0;
897 ei->i_reserved_meta_blocks = 0;
898 ei->i_allocated_meta_blocks = 0;
9d0be502 899 ei->i_da_metadata_calc_len = 0;
7e731bc9 900 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 901 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 902#ifdef CONFIG_QUOTA
903 ei->i_reserved_quota = 0;
96c7e0d9 904 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 905#endif
8aefcd55 906 ei->jinode = NULL;
2e8fa54e 907 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 908 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 909 ei->i_sync_tid = 0;
910 ei->i_datasync_tid = 0;
f7ad6d2e 911 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 912 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 913 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
b30ab0e0 914#ifdef CONFIG_EXT4_FS_ENCRYPTION
b7236e21 915 ei->i_crypt_info = NULL;
b30ab0e0 916#endif
ac27a0ec
DK		 917 return &ei->vfs_inode;
918}
919
7ff9c073
TT		 920static int ext4_drop_inode(struct inode *inode)
921{
922 int drop = generic_drop_inode(inode);
923
924 trace_ext4_drop_inode(inode, drop);
925 return drop;
926}
927
fa0d7e3d
NP		 928static void ext4_i_callback(struct rcu_head *head)
929{
930 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 931 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
932}
933
617ba13b 934static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 935{
9f7dd93d 936 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 937 ext4_msg(inode->i_sb, KERN_ERR,
938 "Inode %lu (%p): orphan list check failed!",
939 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 940 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
941 EXT4_I(inode), sizeof(struct ext4_inode_info),
942 true);
943 dump_stack();
944 }
fa0d7e3d 945 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 946}
947
51cc5068 948static void init_once(void *foo)
ac27a0ec 949{
617ba13b 950 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 951
a35afb83 952 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 953 init_rwsem(&ei->xattr_sem);
0e855ac8 954 init_rwsem(&ei->i_data_sem);
a35afb83 955 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 956}
957
e67bc2b3 958static int __init init_inodecache(void)
ac27a0ec 959{
617ba13b
MC		 960 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
961 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 962 0, (SLAB_RECLAIM_ACCOUNT|
963 SLAB_MEM_SPREAD),
20c2df83 964 init_once);
617ba13b 965 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 966 return -ENOMEM;
967 return 0;
968}
969
970static void destroy_inodecache(void)
971{
8c0a8537
KS		 972 /*
973 * Make sure all delayed rcu free inodes are flushed before we
974 * destroy cache.
975 */
976 rcu_barrier();
617ba13b 977 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 978}
979
0930fcc1 980void ext4_clear_inode(struct inode *inode)
ac27a0ec 981{
0930fcc1 982 invalidate_inode_buffers(inode);
dbd5768f 983 clear_inode(inode);
9f754758 984 dquot_drop(inode);
c2ea3fde 985 ext4_discard_preallocations(inode);
51865fda 986 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 987 if (EXT4_I(inode)->jinode) {
988 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
989 EXT4_I(inode)->jinode);
990 jbd2_free_inode(EXT4_I(inode)->jinode);
991 EXT4_I(inode)->jinode = NULL;
992 }
b7236e21
TT		 993#ifdef CONFIG_EXT4_FS_ENCRYPTION
994 if (EXT4_I(inode)->i_crypt_info)
c936e1ec 995 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
b7236e21 996#endif
ac27a0ec
DK		 997}
998
1b961ac0 999static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1000 u64 ino, u32 generation)
ac27a0ec 1001{
ac27a0ec 1002 struct inode *inode;
ac27a0ec 1003
617ba13b 1004 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1005 return ERR_PTR(-ESTALE);
617ba13b 1006 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1007 return ERR_PTR(-ESTALE);
1008
1009 /* iget isn't really right if the inode is currently unallocated!!
1010 *
617ba13b 1011 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1012 * deleted, so we should be safe.
1013 *
1014 * Currently we don't know the generation for parent directory, so
1015 * a generation of 0 means "accept any"
1016 */
f4bb2981 1017 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1018 if (IS_ERR(inode))
1019 return ERR_CAST(inode);
1020 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1021 iput(inode);
1022 return ERR_PTR(-ESTALE);
1023 }
1b961ac0
CH		 1024
1025 return inode;
1026}
1027
1028static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1029 int fh_len, int fh_type)
1b961ac0
CH		 1030{
1031 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1032 ext4_nfs_get_inode);
1033}
1034
1035static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1036 int fh_len, int fh_type)
1b961ac0
CH		 1037{
1038 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1039 ext4_nfs_get_inode);
ac27a0ec
DK		 1040}
1041
c39a7f84
TO		 1042/*
1043 * Try to release metadata pages (indirect blocks, directories) which are
1044 * mapped via the block device. Since these pages could have journal heads
1045 * which would prevent try_to_free_buffers() from freeing them, we must use
1046 * jbd2 layer's try_to_free_buffers() function to release them.
1047 */
0b8e58a1
AD		 1048static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1049 gfp_t wait)
c39a7f84
TO		 1050{
1051 journal_t *journal = EXT4_SB(sb)->s_journal;
1052
1053 WARN_ON(PageChecked(page));
1054 if (!page_has_buffers(page))
1055 return 0;
1056 if (journal)
1057 return jbd2_journal_try_to_free_buffers(journal, page,
1058 wait & ~__GFP_WAIT);
1059 return try_to_free_buffers(page);
1060}
1061
ac27a0ec 1062#ifdef CONFIG_QUOTA
af5bc92d 1063#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1064#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1065
617ba13b
MC		 1066static int ext4_write_dquot(struct dquot *dquot);
1067static int ext4_acquire_dquot(struct dquot *dquot);
1068static int ext4_release_dquot(struct dquot *dquot);
1069static int ext4_mark_dquot_dirty(struct dquot *dquot);
1070static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1071static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1072 struct path *path);
ca0e05e4 1073static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1074static int ext4_quota_on_mount(struct super_block *sb, int type);
1075static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1076 size_t len, loff_t off);
617ba13b 1077static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1078 const char *data, size_t len, loff_t off);
7c319d32
AK		 1079static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1080 unsigned int flags);
1081static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1082
96c7e0d9
JK		 1083static struct dquot **ext4_get_dquots(struct inode *inode)
1084{
1085 return EXT4_I(inode)->i_dquot;
1086}
1087
61e225dc 1088static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1089 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1090 .write_dquot = ext4_write_dquot,
1091 .acquire_dquot = ext4_acquire_dquot,
1092 .release_dquot = ext4_release_dquot,
1093 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1094 .write_info = ext4_write_info,
1095 .alloc_dquot = dquot_alloc,
1096 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1097};
1098
0d54b217 1099static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1100 .quota_on = ext4_quota_on,
ca0e05e4 1101 .quota_off = ext4_quota_off,
287a8095 1102 .quota_sync = dquot_quota_sync,
0a240339 1103 .get_state = dquot_get_state,
287a8095
CH		 1104 .set_info = dquot_set_dqinfo,
1105 .get_dqblk = dquot_get_dqblk,
1106 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1107};
1108#endif
1109
ee9b6d61 1110static const struct super_operations ext4_sops = {
617ba13b
MC		 1111 .alloc_inode = ext4_alloc_inode,
1112 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1113 .write_inode = ext4_write_inode,
1114 .dirty_inode = ext4_dirty_inode,
7ff9c073 1115 .drop_inode = ext4_drop_inode,
0930fcc1 1116 .evict_inode = ext4_evict_inode,
617ba13b 1117 .put_super = ext4_put_super,
617ba13b 1118 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1119 .freeze_fs = ext4_freeze,
1120 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1121 .statfs = ext4_statfs,
1122 .remount_fs = ext4_remount,
617ba13b 1123 .show_options = ext4_show_options,
ac27a0ec 1124#ifdef CONFIG_QUOTA
617ba13b
MC		 1125 .quota_read = ext4_quota_read,
1126 .quota_write = ext4_quota_write,
96c7e0d9 1127 .get_dquots = ext4_get_dquots,
ac27a0ec 1128#endif
c39a7f84 1129 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1130};
1131
39655164 1132static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1133 .fh_to_dentry = ext4_fh_to_dentry,
1134 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1135 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1136};
1137
1138enum {
1139 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1140 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1141 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1142 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1143 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1144 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1145 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1146 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1147 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1148 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1149 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1150 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
923ae0ff 1151 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1449032b 1152 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
a26f4992 1153 Opt_lazytime, Opt_nolazytime,
1449032b 1154 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1155 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1156 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1157 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1158 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1159};
1160
a447c093 1161static const match_table_t tokens = {
ac27a0ec
DK		 1162 {Opt_bsd_df, "bsddf"},
1163 {Opt_minix_df, "minixdf"},
1164 {Opt_grpid, "grpid"},
1165 {Opt_grpid, "bsdgroups"},
1166 {Opt_nogrpid, "nogrpid"},
1167 {Opt_nogrpid, "sysvgroups"},
1168 {Opt_resgid, "resgid=%u"},
1169 {Opt_resuid, "resuid=%u"},
1170 {Opt_sb, "sb=%u"},
1171 {Opt_err_cont, "errors=continue"},
1172 {Opt_err_panic, "errors=panic"},
1173 {Opt_err_ro, "errors=remount-ro"},
1174 {Opt_nouid32, "nouid32"},
ac27a0ec 1175 {Opt_debug, "debug"},
72578c33
TT		 1176 {Opt_removed, "oldalloc"},
1177 {Opt_removed, "orlov"},
ac27a0ec
DK		 1178 {Opt_user_xattr, "user_xattr"},
1179 {Opt_nouser_xattr, "nouser_xattr"},
1180 {Opt_acl, "acl"},
1181 {Opt_noacl, "noacl"},
e3bb52ae 1182 {Opt_noload, "norecovery"},
5a916be1 1183 {Opt_noload, "noload"},
72578c33
TT		 1184 {Opt_removed, "nobh"},
1185 {Opt_removed, "bh"},
ac27a0ec 1186 {Opt_commit, "commit=%u"},
30773840
TT		 1187 {Opt_min_batch_time, "min_batch_time=%u"},
1188 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1189 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1190 {Opt_journal_path, "journal_path=%s"},
818d276c 1191 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1192 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1193 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1194 {Opt_abort, "abort"},
1195 {Opt_data_journal, "data=journal"},
1196 {Opt_data_ordered, "data=ordered"},
1197 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1198 {Opt_data_err_abort, "data_err=abort"},
1199 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1200 {Opt_offusrjquota, "usrjquota="},
1201 {Opt_usrjquota, "usrjquota=%s"},
1202 {Opt_offgrpjquota, "grpjquota="},
1203 {Opt_grpjquota, "grpjquota=%s"},
1204 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1205 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1206 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1207 {Opt_grpquota, "grpquota"},
1208 {Opt_noquota, "noquota"},
1209 {Opt_quota, "quota"},
1210 {Opt_usrquota, "usrquota"},
1211 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1212 {Opt_barrier, "barrier"},
1213 {Opt_nobarrier, "nobarrier"},
25ec56b5 1214 {Opt_i_version, "i_version"},
923ae0ff 1215 {Opt_dax, "dax"},
c9de560d 1216 {Opt_stripe, "stripe=%u"},
64769240 1217 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1218 {Opt_lazytime, "lazytime"},
1219 {Opt_nolazytime, "nolazytime"},
dd919b98 1220 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1221 {Opt_removed, "mblk_io_submit"},
1222 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1223 {Opt_block_validity, "block_validity"},
1224 {Opt_noblock_validity, "noblock_validity"},
240799cd 1225 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1226 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1227 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1228 {Opt_auto_da_alloc, "auto_da_alloc"},
1229 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1230 {Opt_dioread_nolock, "dioread_nolock"},
1231 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1232 {Opt_discard, "discard"},
1233 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1234 {Opt_init_itable, "init_itable=%u"},
1235 {Opt_init_itable, "init_itable"},
1236 {Opt_noinit_itable, "noinit_itable"},
df981d03 1237 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1238 {Opt_test_dummy_encryption, "test_dummy_encryption"},
c7198b9c
TT		 1239 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1240 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1241 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1242 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1243 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1244 {Opt_err, NULL},
ac27a0ec
DK		 1245};
1246
617ba13b 1247static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1248{
617ba13b 1249 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1250 char *options = (char *) *data;
1251
1252 if (!options || strncmp(options, "sb=", 3) != 0)
1253 return 1; /* Default location */
0b8e58a1 1254
ac27a0ec 1255 options += 3;
0b8e58a1 1256 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1257 sb_block = simple_strtoul(options, &options, 0);
1258 if (*options && *options != ',') {
4776004f 1259 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1260 (char *) *data);
1261 return 1;
1262 }
1263 if (*options == ',')
1264 options++;
1265 *data = (void *) options;
0b8e58a1 1266
ac27a0ec
DK		 1267 return sb_block;
1268}
1269
b3881f74 1270#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1271static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1272 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1273
56c50f11
DM		 1274#ifdef CONFIG_QUOTA
1275static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1276{
1277 struct ext4_sb_info *sbi = EXT4_SB(sb);
1278 char *qname;
03dafb5f 1279 int ret = -1;
56c50f11
DM		 1280
1281 if (sb_any_quota_loaded(sb) &&
1282 !sbi->s_qf_names[qtype]) {
1283 ext4_msg(sb, KERN_ERR,
1284 "Cannot change journaled "
1285 "quota options when quota turned on");
57f73c2c 1286 return -1;
56c50f11 1287 }
262b4662
JK		 1288 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1289 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1290 "when QUOTA feature is enabled");
1291 return -1;
1292 }
56c50f11
DM		 1293 qname = match_strdup(args);
1294 if (!qname) {
1295 ext4_msg(sb, KERN_ERR,
1296 "Not enough memory for storing quotafile name");
57f73c2c 1297 return -1;
56c50f11 1298 }
03dafb5f
CG		 1299 if (sbi->s_qf_names[qtype]) {
1300 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1301 ret = 1;
1302 else
1303 ext4_msg(sb, KERN_ERR,
1304 "%s quota file already specified",
1305 QTYPE2NAME(qtype));
1306 goto errout;
56c50f11 1307 }
03dafb5f 1308 if (strchr(qname, '/')) {
56c50f11
DM		 1309 ext4_msg(sb, KERN_ERR,
1310 "quotafile must be on filesystem root");
03dafb5f 1311 goto errout;
56c50f11 1312 }
03dafb5f 1313 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1314 set_opt(sb, QUOTA);
56c50f11 1315 return 1;
03dafb5f
CG		 1316errout:
1317 kfree(qname);
1318 return ret;
56c50f11
DM		 1319}
1320
1321static int clear_qf_name(struct super_block *sb, int qtype)
1322{
1323
1324 struct ext4_sb_info *sbi = EXT4_SB(sb);
1325
1326 if (sb_any_quota_loaded(sb) &&
1327 sbi->s_qf_names[qtype]) {
1328 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1329 " when quota turned on");
57f73c2c 1330 return -1;
56c50f11 1331 }
03dafb5f 1332 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1333 sbi->s_qf_names[qtype] = NULL;
1334 return 1;
1335}
1336#endif
1337
26092bf5
TT		 1338#define MOPT_SET 0x0001
1339#define MOPT_CLEAR 0x0002
1340#define MOPT_NOSUPPORT 0x0004
1341#define MOPT_EXPLICIT 0x0008
1342#define MOPT_CLEAR_ERR 0x0010
1343#define MOPT_GTE0 0x0020
ac27a0ec 1344#ifdef CONFIG_QUOTA
26092bf5
TT		 1345#define MOPT_Q 0
1346#define MOPT_QFMT 0x0040
1347#else
1348#define MOPT_Q MOPT_NOSUPPORT
1349#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1350#endif
26092bf5 1351#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1352#define MOPT_NO_EXT2 0x0100
1353#define MOPT_NO_EXT3 0x0200
1354#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1355#define MOPT_STRING 0x0400
26092bf5
TT		 1356
1357static const struct mount_opts {
1358 int token;
1359 int mount_opt;
1360 int flags;
1361} ext4_mount_opts[] = {
1362 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1363 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1364 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1365 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1366 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1367 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1368 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1369 MOPT_EXT4_ONLY | MOPT_SET},
1370 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1371 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1372 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1373 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1374 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1375 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1376 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1377 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1378 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1379 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c
TT		 1380 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1381 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1382 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1383 EXT4_MOUNT_JOURNAL_CHECKSUM),
1384 MOPT_EXT4_ONLY | MOPT_SET},
1385 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1386 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1387 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1388 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1389 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1390 MOPT_NO_EXT2 | MOPT_SET},
1391 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1392 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1393 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1394 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1395 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1396 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1397 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1398 {Opt_commit, 0, MOPT_GTE0},
1399 {Opt_max_batch_time, 0, MOPT_GTE0},
1400 {Opt_min_batch_time, 0, MOPT_GTE0},
1401 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1402 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1403 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1404 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1405 {Opt_resuid, 0, MOPT_GTE0},
1406 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1407 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1408 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1409 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1410 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1411 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1412 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1413 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1414 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1415 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1416#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1417 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1418 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1419#else
26092bf5
TT		 1420 {Opt_acl, 0, MOPT_NOSUPPORT},
1421 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1422#endif
26092bf5
TT		 1423 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1424 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1425 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1426 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1427 MOPT_SET | MOPT_Q},
1428 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1429 MOPT_SET | MOPT_Q},
1430 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1431 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1432 {Opt_usrjquota, 0, MOPT_Q},
1433 {Opt_grpjquota, 0, MOPT_Q},
1434 {Opt_offusrjquota, 0, MOPT_Q},
1435 {Opt_offgrpjquota, 0, MOPT_Q},
1436 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1437 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1438 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1439 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1440 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
26092bf5
TT		 1441 {Opt_err, 0, 0}
1442};
1443
1444static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1445 substring_t *args, unsigned long *journal_devnum,
1446 unsigned int *journal_ioprio, int is_remount)
1447{
1448 struct ext4_sb_info *sbi = EXT4_SB(sb);
1449 const struct mount_opts *m;
08cefc7a
EB		 1450 kuid_t uid;
1451 kgid_t gid;
26092bf5
TT		 1452 int arg = 0;
1453
57f73c2c
TT		 1454#ifdef CONFIG_QUOTA
1455 if (token == Opt_usrjquota)
1456 return set_qf_name(sb, USRQUOTA, &args[0]);
1457 else if (token == Opt_grpjquota)
1458 return set_qf_name(sb, GRPQUOTA, &args[0]);
1459 else if (token == Opt_offusrjquota)
1460 return clear_qf_name(sb, USRQUOTA);
1461 else if (token == Opt_offgrpjquota)
1462 return clear_qf_name(sb, GRPQUOTA);
1463#endif
26092bf5 1464 switch (token) {
f7048605
TT		 1465 case Opt_noacl:
1466 case Opt_nouser_xattr:
1467 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1468 break;
26092bf5
TT		 1469 case Opt_sb:
1470 return 1; /* handled by get_sb_block() */
1471 case Opt_removed:
5f3633e3 1472 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1473 return 1;
26092bf5
TT		 1474 case Opt_abort:
1475 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1476 return 1;
1477 case Opt_i_version:
1478 sb->s_flags |= MS_I_VERSION;
1479 return 1;
a26f4992
TT		 1480 case Opt_lazytime:
1481 sb->s_flags |= MS_LAZYTIME;
1482 return 1;
1483 case Opt_nolazytime:
1484 sb->s_flags &= ~MS_LAZYTIME;
1485 return 1;
26092bf5
TT		 1486 }
1487
5f3633e3
JK		 1488 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1489 if (token == m->token)
1490 break;
1491
1492 if (m->token == Opt_err) {
1493 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1494 "or missing value", opt);
1495 return -1;
1496 }
1497
8dc0aa8c
TT		 1498 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1499 ext4_msg(sb, KERN_ERR,
1500 "Mount option \"%s\" incompatible with ext2", opt);
1501 return -1;
1502 }
1503 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1504 ext4_msg(sb, KERN_ERR,
1505 "Mount option \"%s\" incompatible with ext3", opt);
1506 return -1;
1507 }
1508
ad4eec61 1509 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1510 return -1;
1511 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1512 return -1;
1513 if (m->flags & MOPT_EXPLICIT)
1514 set_opt2(sb, EXPLICIT_DELALLOC);
1515 if (m->flags & MOPT_CLEAR_ERR)
1516 clear_opt(sb, ERRORS_MASK);
1517 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1518 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1519 "options when quota turned on");
1520 return -1;
1521 }
1522
1523 if (m->flags & MOPT_NOSUPPORT) {
1524 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1525 } else if (token == Opt_commit) {
1526 if (arg == 0)
1527 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1528 sbi->s_commit_interval = HZ * arg;
1529 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1530 sbi->s_max_batch_time = arg;
1531 } else if (token == Opt_min_batch_time) {
1532 sbi->s_min_batch_time = arg;
1533 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1534 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1535 ext4_msg(sb, KERN_ERR,
1536 "EXT4-fs: inode_readahead_blks must be "
1537 "0 or a power of 2 smaller than 2^31");
26092bf5 1538 return -1;
5f3633e3
JK		 1539 }
1540 sbi->s_inode_readahead_blks = arg;
1541 } else if (token == Opt_init_itable) {
1542 set_opt(sb, INIT_INODE_TABLE);
1543 if (!args->from)
1544 arg = EXT4_DEF_LI_WAIT_MULT;
1545 sbi->s_li_wait_mult = arg;
1546 } else if (token == Opt_max_dir_size_kb) {
1547 sbi->s_max_dir_size_kb = arg;
1548 } else if (token == Opt_stripe) {
1549 sbi->s_stripe = arg;
1550 } else if (token == Opt_resuid) {
1551 uid = make_kuid(current_user_ns(), arg);
1552 if (!uid_valid(uid)) {
1553 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1554 return -1;
1555 }
5f3633e3
JK		 1556 sbi->s_resuid = uid;
1557 } else if (token == Opt_resgid) {
1558 gid = make_kgid(current_user_ns(), arg);
1559 if (!gid_valid(gid)) {
1560 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1561 return -1;
1562 }
1563 sbi->s_resgid = gid;
1564 } else if (token == Opt_journal_dev) {
1565 if (is_remount) {
1566 ext4_msg(sb, KERN_ERR,
1567 "Cannot specify journal on remount");
1568 return -1;
1569 }
1570 *journal_devnum = arg;
ad4eec61
ES		 1571 } else if (token == Opt_journal_path) {
1572 char *journal_path;
1573 struct inode *journal_inode;
1574 struct path path;
1575 int error;
1576
1577 if (is_remount) {
1578 ext4_msg(sb, KERN_ERR,
1579 "Cannot specify journal on remount");
1580 return -1;
1581 }
1582 journal_path = match_strdup(&args[0]);
1583 if (!journal_path) {
1584 ext4_msg(sb, KERN_ERR, "error: could not dup "
1585 "journal device string");
1586 return -1;
1587 }
1588
1589 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1590 if (error) {
1591 ext4_msg(sb, KERN_ERR, "error: could not find "
1592 "journal device path: error %d", error);
1593 kfree(journal_path);
1594 return -1;
1595 }
1596
2b0143b5 1597 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1598 if (!S_ISBLK(journal_inode->i_mode)) {
1599 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1600 "is not a block device", journal_path);
1601 path_put(&path);
1602 kfree(journal_path);
1603 return -1;
1604 }
1605
1606 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1607 path_put(&path);
1608 kfree(journal_path);
5f3633e3
JK		 1609 } else if (token == Opt_journal_ioprio) {
1610 if (arg > 7) {
1611 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1612 " (must be 0-7)");
1613 return -1;
1614 }
1615 *journal_ioprio =
1616 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1617 } else if (token == Opt_test_dummy_encryption) {
1618#ifdef CONFIG_EXT4_FS_ENCRYPTION
1619 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1620 ext4_msg(sb, KERN_WARNING,
1621 "Test dummy encryption mode enabled");
1622#else
1623 ext4_msg(sb, KERN_WARNING,
1624 "Test dummy encryption mount option ignored");
1625#endif
5f3633e3
JK		 1626 } else if (m->flags & MOPT_DATAJ) {
1627 if (is_remount) {
1628 if (!sbi->s_journal)
1629 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1630 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1631 ext4_msg(sb, KERN_ERR,
26092bf5 1632 "Cannot change data mode on remount");
26092bf5 1633 return -1;
ac27a0ec 1634 }
26092bf5 1635 } else {
5f3633e3
JK		 1636 clear_opt(sb, DATA_FLAGS);
1637 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1638 }
5f3633e3
JK		 1639#ifdef CONFIG_QUOTA
1640 } else if (m->flags & MOPT_QFMT) {
1641 if (sb_any_quota_loaded(sb) &&
1642 sbi->s_jquota_fmt != m->mount_opt) {
1643 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1644 "quota options when quota turned on");
1645 return -1;
1646 }
262b4662
JK		 1647 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1648 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1649 ext4_msg(sb, KERN_ERR,
1650 "Cannot set journaled quota options "
1651 "when QUOTA feature is enabled");
1652 return -1;
1653 }
5f3633e3 1654 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff
RZ		 1655#endif
1656#ifndef CONFIG_FS_DAX
1657 } else if (token == Opt_dax) {
1658 ext4_msg(sb, KERN_INFO, "dax option not supported");
1659 return -1;
5f3633e3
JK		 1660#endif
1661 } else {
1662 if (!args->from)
1663 arg = 1;
1664 if (m->flags & MOPT_CLEAR)
1665 arg = !arg;
1666 else if (unlikely(!(m->flags & MOPT_SET))) {
1667 ext4_msg(sb, KERN_WARNING,
1668 "buggy handling of option %s", opt);
1669 WARN_ON(1);
1670 return -1;
1671 }
1672 if (arg != 0)
1673 sbi->s_mount_opt |= m->mount_opt;
1674 else
1675 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1676 }
5f3633e3 1677 return 1;
26092bf5
TT		 1678}
1679
1680static int parse_options(char *options, struct super_block *sb,
1681 unsigned long *journal_devnum,
1682 unsigned int *journal_ioprio,
1683 int is_remount)
1684{
1685 struct ext4_sb_info *sbi = EXT4_SB(sb);
1686 char *p;
1687 substring_t args[MAX_OPT_ARGS];
1688 int token;
1689
1690 if (!options)
1691 return 1;
1692
1693 while ((p = strsep(&options, ",")) != NULL) {
1694 if (!*p)
1695 continue;
1696 /*
1697 * Initialize args struct so we know whether arg was
1698 * found; some options take optional arguments.
1699 */
caecd0af 1700 args[0].to = args[0].from = NULL;
26092bf5
TT		 1701 token = match_token(p, tokens, args);
1702 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1703 journal_ioprio, is_remount) < 0)
1704 return 0;
ac27a0ec
DK		 1705 }
1706#ifdef CONFIG_QUOTA
262b4662
JK		 1707 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1708 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1709 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1710 "feature is enabled");
1711 return 0;
1712 }
ac27a0ec 1713 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1714 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1715 clear_opt(sb, USRQUOTA);
ac27a0ec 1716
482a7425 1717 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1718 clear_opt(sb, GRPQUOTA);
ac27a0ec 1719
56c50f11 1720 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1721 ext4_msg(sb, KERN_ERR, "old and new quota "
1722 "format mixing");
ac27a0ec
DK		 1723 return 0;
1724 }
1725
1726 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1727 ext4_msg(sb, KERN_ERR, "journaled quota format "
1728 "not specified");
ac27a0ec
DK		 1729 return 0;
1730 }
ac27a0ec
DK		 1731 }
1732#endif
261cb20c
JK		 1733 if (test_opt(sb, DIOREAD_NOLOCK)) {
1734 int blocksize =
1735 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1736
1737 if (blocksize < PAGE_CACHE_SIZE) {
1738 ext4_msg(sb, KERN_ERR, "can't mount with "
1739 "dioread_nolock if block size != PAGE_SIZE");
1740 return 0;
1741 }
1742 }
d4f76107
JK		 1743 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1744 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1745 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1746 "in data=ordered mode");
1747 return 0;
1748 }
ac27a0ec
DK		 1749 return 1;
1750}
1751
2adf6da8
TT		 1752static inline void ext4_show_quota_options(struct seq_file *seq,
1753 struct super_block *sb)
1754{
1755#if defined(CONFIG_QUOTA)
1756 struct ext4_sb_info *sbi = EXT4_SB(sb);
1757
1758 if (sbi->s_jquota_fmt) {
1759 char *fmtname = "";
1760
1761 switch (sbi->s_jquota_fmt) {
1762 case QFMT_VFS_OLD:
1763 fmtname = "vfsold";
1764 break;
1765 case QFMT_VFS_V0:
1766 fmtname = "vfsv0";
1767 break;
1768 case QFMT_VFS_V1:
1769 fmtname = "vfsv1";
1770 break;
1771 }
1772 seq_printf(seq, ",jqfmt=%s", fmtname);
1773 }
1774
1775 if (sbi->s_qf_names[USRQUOTA])
a068acf2 1776 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 1777
1778 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 1779 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1780#endif
1781}
1782
5a916be1
TT		 1783static const char *token2str(int token)
1784{
50df9fd5 1785 const struct match_token *t;
5a916be1
TT		 1786
1787 for (t = tokens; t->token != Opt_err; t++)
1788 if (t->token == token && !strchr(t->pattern, '='))
1789 break;
1790 return t->pattern;
1791}
1792
2adf6da8
TT		 1793/*
1794 * Show an option if
1795 * - it's set to a non-default value OR
1796 * - if the per-sb default is different from the global default
1797 */
66acdcf4
TT		 1798static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1799 int nodefs)
2adf6da8 1800{
2adf6da8
TT		 1801 struct ext4_sb_info *sbi = EXT4_SB(sb);
1802 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1803 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1804 const struct mount_opts *m;
66acdcf4 1805 char sep = nodefs ? '\n' : ',';
2adf6da8 1806
66acdcf4
TT		 1807#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1808#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1809
1810 if (sbi->s_sb_block != 1)
5a916be1
TT		 1811 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1812
1813 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1814 int want_set = m->flags & MOPT_SET;
1815 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1816 (m->flags & MOPT_CLEAR_ERR))
1817 continue;
66acdcf4 1818 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1819 continue; /* skip if same as the default */
1820 if ((want_set &&
1821 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1822 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1823 continue; /* select Opt_noFoo vs Opt_Foo */
1824 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1825 }
5a916be1 1826
08cefc7a 1827 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1828 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1829 SEQ_OPTS_PRINT("resuid=%u",
1830 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1831 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1832 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1833 SEQ_OPTS_PRINT("resgid=%u",
1834 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1835 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1836 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1837 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1838 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1839 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1840 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1841 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1842 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1843 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1844 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1845 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1846 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1847 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1848 if (sb->s_flags & MS_I_VERSION)
5a916be1 1849 SEQ_OPTS_PUTS("i_version");
66acdcf4 1850 if (nodefs || sbi->s_stripe)
5a916be1 1851 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1852 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1853 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1854 SEQ_OPTS_PUTS("data=journal");
1855 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1856 SEQ_OPTS_PUTS("data=ordered");
1857 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1858 SEQ_OPTS_PUTS("data=writeback");
1859 }
66acdcf4
TT		 1860 if (nodefs ||
1861 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1862 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1863 sbi->s_inode_readahead_blks);
2adf6da8 1864
66acdcf4
TT		 1865 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1866 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1867 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1868 if (nodefs || sbi->s_max_dir_size_kb)
1869 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1870
1871 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1872 return 0;
1873}
1874
66acdcf4
TT		 1875static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1876{
1877 return _ext4_show_options(seq, root->d_sb, 0);
1878}
1879
1880static int options_seq_show(struct seq_file *seq, void *offset)
1881{
1882 struct super_block *sb = seq->private;
1883 int rc;
1884
1885 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1886 rc = _ext4_show_options(seq, sb, 1);
1887 seq_puts(seq, "\n");
1888 return rc;
1889}
1890
1891static int options_open_fs(struct inode *inode, struct file *file)
1892{
d9dda78b 1893 return single_open(file, options_seq_show, PDE_DATA(inode));
66acdcf4
TT		 1894}
1895
1896static const struct file_operations ext4_seq_options_fops = {
1897 .owner = THIS_MODULE,
1898 .open = options_open_fs,
1899 .read = seq_read,
1900 .llseek = seq_lseek,
1901 .release = single_release,
1902};
1903
617ba13b 1904static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1905 int read_only)
1906{
617ba13b 1907 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1908 int res = 0;
1909
617ba13b 1910 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1911 ext4_msg(sb, KERN_ERR, "revision level too high, "
1912 "forcing read-only mode");
ac27a0ec
DK		 1913 res = MS_RDONLY;
1914 }
1915 if (read_only)
281b5995 1916 goto done;
617ba13b 1917 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1918 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1919 "running e2fsck is recommended");
c8b459f4 1920 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 1921 ext4_msg(sb, KERN_WARNING,
1922 "warning: mounting fs with errors, "
1923 "running e2fsck is recommended");
ed3ce80a 1924 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1925 le16_to_cpu(es->s_mnt_count) >=
1926 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1927 ext4_msg(sb, KERN_WARNING,
1928 "warning: maximal mount count reached, "
1929 "running e2fsck is recommended");
ac27a0ec
DK		 1930 else if (le32_to_cpu(es->s_checkinterval) &&
1931 (le32_to_cpu(es->s_lastcheck) +
1932 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1933 ext4_msg(sb, KERN_WARNING,
1934 "warning: checktime reached, "
1935 "running e2fsck is recommended");
0b8e58a1 1936 if (!sbi->s_journal)
0390131b 1937 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1938 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1939 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1940 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1941 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1942 ext4_update_dynamic_rev(sb);
0390131b
FM		 1943 if (sbi->s_journal)
1944 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1945
e2d67052 1946 ext4_commit_super(sb, 1);
281b5995 1947done:
ac27a0ec 1948 if (test_opt(sb, DEBUG))
a9df9a49 1949 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1950 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1951 sb->s_blocksize,
1952 sbi->s_groups_count,
617ba13b
MC		 1953 EXT4_BLOCKS_PER_GROUP(sb),
1954 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1955 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1956
7abc52c2 1957 cleancache_init_fs(sb);
ac27a0ec
DK		 1958 return res;
1959}
1960
117fff10
TT		 1961int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1962{
1963 struct ext4_sb_info *sbi = EXT4_SB(sb);
1964 struct flex_groups *new_groups;
1965 int size;
1966
1967 if (!sbi->s_log_groups_per_flex)
1968 return 0;
1969
1970 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1971 if (size <= sbi->s_flex_groups_allocated)
1972 return 0;
1973
1974 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1975 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1976 if (!new_groups) {
1977 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1978 size / (int) sizeof(struct flex_groups));
1979 return -ENOMEM;
1980 }
1981
1982 if (sbi->s_flex_groups) {
1983 memcpy(new_groups, sbi->s_flex_groups,
1984 (sbi->s_flex_groups_allocated *
1985 sizeof(struct flex_groups)));
b93b41d4 1986 kvfree(sbi->s_flex_groups);
117fff10
TT		 1987 }
1988 sbi->s_flex_groups = new_groups;
1989 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1990 return 0;
1991}
1992
772cb7c8
JS		 1993static int ext4_fill_flex_info(struct super_block *sb)
1994{
1995 struct ext4_sb_info *sbi = EXT4_SB(sb);
1996 struct ext4_group_desc *gdp = NULL;
772cb7c8 1997 ext4_group_t flex_group;
117fff10 1998 int i, err;
772cb7c8 1999
503358ae 2000 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2001 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2002 sbi->s_log_groups_per_flex = 0;
2003 return 1;
2004 }
2005
117fff10
TT		 2006 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2007 if (err)
9933fc0a 2008 goto failed;
772cb7c8 2009
772cb7c8 2010 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2011 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2012
2013 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2014 atomic_add(ext4_free_inodes_count(sb, gdp),
2015 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2016 atomic64_add(ext4_free_group_clusters(sb, gdp),
2017 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2018 atomic_add(ext4_used_dirs_count(sb, gdp),
2019 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2020 }
2021
2022 return 1;
2023failed:
2024 return 0;
2025}
2026
feb0ab32
DW		 2027static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2028 struct ext4_group_desc *gdp)
717d50e4 2029{
feb0ab32 2030 int offset;
717d50e4 2031 __u16 crc = 0;
feb0ab32 2032 __le32 le_group = cpu_to_le32(block_group);
717d50e4 2033
9aa5d32b 2034 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2035 /* Use new metadata_csum algorithm */
d6a77105 2036 __le16 save_csum;
feb0ab32
DW		 2037 __u32 csum32;
2038
d6a77105 2039 save_csum = gdp->bg_checksum;
feb0ab32
DW		 2040 gdp->bg_checksum = 0;
2041 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2042 sizeof(le_group));
2043 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2044 sbi->s_desc_size);
d6a77105 2045 gdp->bg_checksum = save_csum;
feb0ab32
DW		 2046
2047 crc = csum32 & 0xFFFF;
2048 goto out;
717d50e4
AD		 2049 }
2050
feb0ab32 2051 /* old crc16 code */
813d32f9
DW		 2052 if (!(sbi->s_es->s_feature_ro_compat &
2053 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2054 return 0;
2055
feb0ab32
DW		 2056 offset = offsetof(struct ext4_group_desc, bg_checksum);
2057
2058 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2059 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2060 crc = crc16(crc, (__u8 *)gdp, offset);
2061 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2062 /* for checksum of struct ext4_group_desc do the rest...*/
2063 if ((sbi->s_es->s_feature_incompat &
2064 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2065 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2066 crc = crc16(crc, (__u8 *)gdp + offset,
2067 le16_to_cpu(sbi->s_es->s_desc_size) -
2068 offset);
2069
2070out:
717d50e4
AD		 2071 return cpu_to_le16(crc);
2072}
2073
feb0ab32 2074int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2075 struct ext4_group_desc *gdp)
2076{
feb0ab32
DW		 2077 if (ext4_has_group_desc_csum(sb) &&
2078 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2079 block_group, gdp)))
717d50e4
AD		 2080 return 0;
2081
2082 return 1;
2083}
2084
feb0ab32
DW		 2085void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2086 struct ext4_group_desc *gdp)
2087{
2088 if (!ext4_has_group_desc_csum(sb))
2089 return;
2090 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2091}
2092
ac27a0ec 2093/* Called at mount-time, super-block is locked */
bfff6873
LC		 2094static int ext4_check_descriptors(struct super_block *sb,
2095 ext4_group_t *first_not_zeroed)
ac27a0ec 2096{
617ba13b
MC		 2097 struct ext4_sb_info *sbi = EXT4_SB(sb);
2098 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2099 ext4_fsblk_t last_block;
bd81d8ee
LV		 2100 ext4_fsblk_t block_bitmap;
2101 ext4_fsblk_t inode_bitmap;
2102 ext4_fsblk_t inode_table;
ce421581 2103 int flexbg_flag = 0;
bfff6873 2104 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2105
ce421581
JS		 2106 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2107 flexbg_flag = 1;
2108
af5bc92d 2109 ext4_debug("Checking group descriptors");
ac27a0ec 2110
197cd65a
AM		 2111 for (i = 0; i < sbi->s_groups_count; i++) {
2112 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2113
ce421581 2114 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2115 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2116 else
2117 last_block = first_block +
617ba13b 2118 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2119
bfff6873
LC		 2120 if ((grp == sbi->s_groups_count) &&
2121 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2122 grp = i;
2123
8fadc143 2124 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2125 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2126 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2127 "Block bitmap for group %u not in group "
b31e1552 2128 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2129 return 0;
2130 }
8fadc143 2131 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2132 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2133 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2134 "Inode bitmap for group %u not in group "
b31e1552 2135 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2136 return 0;
2137 }
8fadc143 2138 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2139 if (inode_table < first_block ||
2b2d6d01 2140 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2141 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2142 "Inode table for group %u not in group "
b31e1552 2143 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2144 return 0;
2145 }
955ce5f5 2146 ext4_lock_group(sb, i);
feb0ab32 2147 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2148 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2149 "Checksum for group %u failed (%u!=%u)",
2150 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2151 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2152 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2153 ext4_unlock_group(sb, i);
8a266467 2154 return 0;
7ee1ec4c 2155 }
717d50e4 2156 }
955ce5f5 2157 ext4_unlock_group(sb, i);
ce421581
JS		 2158 if (!flexbg_flag)
2159 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2160 }
bfff6873
LC		 2161 if (NULL != first_not_zeroed)
2162 *first_not_zeroed = grp;
ac27a0ec
DK		 2163 return 1;
2164}
2165
617ba13b 2166/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2167 * the superblock) which were deleted from all directories, but held open by
2168 * a process at the time of a crash. We walk the list and try to delete these
2169 * inodes at recovery time (only with a read-write filesystem).
2170 *
2171 * In order to keep the orphan inode chain consistent during traversal (in
2172 * case of crash during recovery), we link each inode into the superblock
2173 * orphan list_head and handle it the same way as an inode deletion during
2174 * normal operation (which journals the operations for us).
2175 *
2176 * We only do an iget() and an iput() on each inode, which is very safe if we
2177 * accidentally point at an in-use or already deleted inode. The worst that
2178 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2179 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2180 * e2fsck was run on this filesystem, and it must have already done the orphan
2181 * inode cleanup for us, so we can safely abort without any further action.
2182 */
2b2d6d01
TT		 2183static void ext4_orphan_cleanup(struct super_block *sb,
2184 struct ext4_super_block *es)
ac27a0ec
DK		 2185{
2186 unsigned int s_flags = sb->s_flags;
2187 int nr_orphans = 0, nr_truncates = 0;
2188#ifdef CONFIG_QUOTA
2189 int i;
2190#endif
2191 if (!es->s_last_orphan) {
2192 jbd_debug(4, "no orphan inodes to clean up\n");
2193 return;
2194 }
2195
a8f48a95 2196 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2197 ext4_msg(sb, KERN_ERR, "write access "
2198 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2199 return;
2200 }
2201
d39195c3
AG		 2202 /* Check if feature set would not allow a r/w mount */
2203 if (!ext4_feature_set_ok(sb, 0)) {
2204 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2205 "unknown ROCOMPAT features");
2206 return;
2207 }
2208
617ba13b 2209 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2210 /* don't clear list on RO mount w/ errors */
2211 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2212 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2213 "clearing orphan list.\n");
c25f9bc6
ES		 2214 es->s_last_orphan = 0;
2215 }
ac27a0ec
DK		 2216 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2217 return;
2218 }
2219
2220 if (s_flags & MS_RDONLY) {
b31e1552 2221 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2222 sb->s_flags &= ~MS_RDONLY;
2223 }
2224#ifdef CONFIG_QUOTA
2225 /* Needed for iput() to work correctly and not trash data */
2226 sb->s_flags |= MS_ACTIVE;
2227 /* Turn on quotas so that they are updated correctly */
a2d4a646 2228 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2229 if (EXT4_SB(sb)->s_qf_names[i]) {
2230 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2231 if (ret < 0)
b31e1552
ES		 2232 ext4_msg(sb, KERN_ERR,
2233 "Cannot turn on journaled "
2234 "quota: error %d", ret);
ac27a0ec
DK		 2235 }
2236 }
2237#endif
2238
2239 while (es->s_last_orphan) {
2240 struct inode *inode;
2241
97bd42b9
JB		 2242 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2243 if (IS_ERR(inode)) {
ac27a0ec
DK		 2244 es->s_last_orphan = 0;
2245 break;
2246 }
2247
617ba13b 2248 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2249 dquot_initialize(inode);
ac27a0ec 2250 if (inode->i_nlink) {
566370a2
PT		 2251 if (test_opt(sb, DEBUG))
2252 ext4_msg(sb, KERN_DEBUG,
2253 "%s: truncating inode %lu to %lld bytes",
2254 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2255 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2256 inode->i_ino, inode->i_size);
721e3eba 2257 mutex_lock(&inode->i_mutex);
55f252c9 2258 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2259 ext4_truncate(inode);
721e3eba 2260 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2261 nr_truncates++;
2262 } else {
566370a2
PT		 2263 if (test_opt(sb, DEBUG))
2264 ext4_msg(sb, KERN_DEBUG,
2265 "%s: deleting unreferenced inode %lu",
2266 __func__, inode->i_ino);
ac27a0ec
DK		 2267 jbd_debug(2, "deleting unreferenced inode %lu\n",
2268 inode->i_ino);
2269 nr_orphans++;
2270 }
2271 iput(inode); /* The delete magic happens here! */
2272 }
2273
2b2d6d01 2274#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2275
2276 if (nr_orphans)
b31e1552
ES		 2277 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2278 PLURAL(nr_orphans));
ac27a0ec 2279 if (nr_truncates)
b31e1552
ES		 2280 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2281 PLURAL(nr_truncates));
ac27a0ec
DK		 2282#ifdef CONFIG_QUOTA
2283 /* Turn quotas off */
a2d4a646 2284 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2285 if (sb_dqopt(sb)->files[i])
287a8095 2286 dquot_quota_off(sb, i);
ac27a0ec
DK		 2287 }
2288#endif
2289 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2290}
0b8e58a1 2291
cd2291a4
ES		 2292/*
2293 * Maximal extent format file size.
2294 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2295 * extent format containers, within a sector_t, and within i_blocks
2296 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2297 * so that won't be a limiting factor.
2298 *
f17722f9
LC		 2299 * However there is other limiting factor. We do store extents in the form
2300 * of starting block and length, hence the resulting length of the extent
2301 * covering maximum file size must fit into on-disk format containers as
2302 * well. Given that length is always by 1 unit bigger than max unit (because
2303 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2304 *
cd2291a4
ES		 2305 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2306 */
f287a1a5 2307static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2308{
2309 loff_t res;
2310 loff_t upper_limit = MAX_LFS_FILESIZE;
2311
2312 /* small i_blocks in vfs inode? */
f287a1a5 2313 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2314 /*
90c699a9 2315 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2316 * i_block represent total blocks in 512 bytes
2317 * 32 == size of vfs inode i_blocks * 8
2318 */
2319 upper_limit = (1LL << 32) - 1;
2320
2321 /* total blocks in file system block size */
2322 upper_limit >>= (blkbits - 9);
2323 upper_limit <<= blkbits;
2324 }
2325
f17722f9
LC		 2326 /*
2327 * 32-bit extent-start container, ee_block. We lower the maxbytes
2328 * by one fs block, so ee_len can cover the extent of maximum file
2329 * size
2330 */
2331 res = (1LL << 32) - 1;
cd2291a4 2332 res <<= blkbits;
cd2291a4
ES		 2333
2334 /* Sanity check against vm- & vfs- imposed limits */
2335 if (res > upper_limit)
2336 res = upper_limit;
2337
2338 return res;
2339}
ac27a0ec 2340
ac27a0ec 2341/*
cd2291a4 2342 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2343 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2344 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2345 */
f287a1a5 2346static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2347{
617ba13b 2348 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2349 int meta_blocks;
2350 loff_t upper_limit;
0b8e58a1
AD		 2351 /* This is calculated to be the largest file size for a dense, block
2352 * mapped file such that the file's total number of 512-byte sectors,
2353 * including data and all indirect blocks, does not exceed (2^48 - 1).
2354 *
2355 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2356 * number of 512-byte sectors of the file.
0fc1b451
AK		 2357 */
2358
f287a1a5 2359 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2360 /*
90c699a9 2361 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2362 * the inode i_block field represents total file blocks in
2363 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2364 */
2365 upper_limit = (1LL << 32) - 1;
2366
2367 /* total blocks in file system block size */
2368 upper_limit >>= (bits - 9);
2369
2370 } else {
8180a562
AK		 2371 /*
2372 * We use 48 bit ext4_inode i_blocks
2373 * With EXT4_HUGE_FILE_FL set the i_blocks
2374 * represent total number of blocks in
2375 * file system block size
2376 */
0fc1b451
AK		 2377 upper_limit = (1LL << 48) - 1;
2378
0fc1b451
AK		 2379 }
2380
2381 /* indirect blocks */
2382 meta_blocks = 1;
2383 /* double indirect blocks */
2384 meta_blocks += 1 + (1LL << (bits-2));
2385 /* tripple indirect blocks */
2386 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2387
2388 upper_limit -= meta_blocks;
2389 upper_limit <<= bits;
ac27a0ec
DK		 2390
2391 res += 1LL << (bits-2);
2392 res += 1LL << (2*(bits-2));
2393 res += 1LL << (3*(bits-2));
2394 res <<= bits;
2395 if (res > upper_limit)
2396 res = upper_limit;
0fc1b451
AK		 2397
2398 if (res > MAX_LFS_FILESIZE)
2399 res = MAX_LFS_FILESIZE;
2400
ac27a0ec
DK		 2401 return res;
2402}
2403
617ba13b 2404static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2405 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2406{
617ba13b 2407 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2408 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2409 int has_super = 0;
2410
2411 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2412
617ba13b 2413 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2414 nr < first_meta_bg)
70bbb3e0 2415 return logical_sb_block + nr + 1;
ac27a0ec 2416 bg = sbi->s_desc_per_block * nr;
617ba13b 2417 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2418 has_super = 1;
0b8e58a1 2419
bd63f6b0
DW		 2420 /*
2421 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2422 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2423 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2424 * compensate.
2425 */
2426 if (sb->s_blocksize == 1024 && nr == 0 &&
2427 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2428 has_super++;
2429
617ba13b 2430 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2431}
2432
c9de560d
AT		 2433/**
2434 * ext4_get_stripe_size: Get the stripe size.
2435 * @sbi: In memory super block info
2436 *
2437 * If we have specified it via mount option, then
2438 * use the mount option value. If the value specified at mount time is
2439 * greater than the blocks per group use the super block value.
2440 * If the super block value is greater than blocks per group return 0.
2441 * Allocator needs it be less than blocks per group.
2442 *
2443 */
2444static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2445{
2446 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2447 unsigned long stripe_width =
2448 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2449 int ret;
c9de560d
AT		 2450
2451 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2452 ret = sbi->s_stripe;
2453 else if (stripe_width <= sbi->s_blocks_per_group)
2454 ret = stripe_width;
2455 else if (stride <= sbi->s_blocks_per_group)
2456 ret = stride;
2457 else
2458 ret = 0;
c9de560d 2459
3eb08658
DE		 2460 /*
2461 * If the stripe width is 1, this makes no sense and
2462 * we set it to 0 to turn off stripe handling code.
2463 */
2464 if (ret <= 1)
2465 ret = 0;
c9de560d 2466
3eb08658 2467 return ret;
c9de560d 2468}
ac27a0ec 2469
a13fb1a4
ES		 2470/*
2471 * Check whether this filesystem can be mounted based on
2472 * the features present and the RDONLY/RDWR mount requested.
2473 * Returns 1 if this filesystem can be mounted as requested,
2474 * 0 if it cannot be.
2475 */
2476static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2477{
2478 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2479 ext4_msg(sb, KERN_ERR,
2480 "Couldn't mount because of "
2481 "unsupported optional features (%x)",
2482 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2483 ~EXT4_FEATURE_INCOMPAT_SUPP));
2484 return 0;
2485 }
2486
2487 if (readonly)
2488 return 1;
2489
2cb5cc8b
DW		 2490 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2491 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2492 sb->s_flags |= MS_RDONLY;
2493 return 1;
2494 }
2495
a13fb1a4
ES		 2496 /* Check that feature set is OK for a read-write mount */
2497 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2498 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2499 "unsupported optional features (%x)",
2500 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2501 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2502 return 0;
2503 }
2504 /*
2505 * Large file size enabled file system can only be mounted
2506 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2507 */
2508 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2509 if (sizeof(blkcnt_t) < sizeof(u64)) {
2510 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2511 "cannot be mounted RDWR without "
2512 "CONFIG_LBDAF");
2513 return 0;
2514 }
2515 }
bab08ab9
TT		 2516 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2517 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2518 ext4_msg(sb, KERN_ERR,
2519 "Can't support bigalloc feature without "
2520 "extents feature\n");
2521 return 0;
2522 }
7c319d32
AK		 2523
2524#ifndef CONFIG_QUOTA
2525 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2526 !readonly) {
2527 ext4_msg(sb, KERN_ERR,
2528 "Filesystem with quota feature cannot be mounted RDWR "
2529 "without CONFIG_QUOTA");
2530 return 0;
2531 }
2532#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2533 return 1;
2534}
2535
66e61a9e
TT		 2536/*
2537 * This function is called once a day if we have errors logged
2538 * on the file system
2539 */
2540static void print_daily_error_info(unsigned long arg)
2541{
2542 struct super_block *sb = (struct super_block *) arg;
2543 struct ext4_sb_info *sbi;
2544 struct ext4_super_block *es;
2545
2546 sbi = EXT4_SB(sb);
2547 es = sbi->s_es;
2548
2549 if (es->s_error_count)
ae0f78de
TT		 2550 /* fsck newer than v1.41.13 is needed to clean this condition. */
2551 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2552 le32_to_cpu(es->s_error_count));
2553 if (es->s_first_error_time) {
ae0f78de 2554 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2555 sb->s_id, le32_to_cpu(es->s_first_error_time),
2556 (int) sizeof(es->s_first_error_func),
2557 es->s_first_error_func,
2558 le32_to_cpu(es->s_first_error_line));
2559 if (es->s_first_error_ino)
2560 printk(": inode %u",
2561 le32_to_cpu(es->s_first_error_ino));
2562 if (es->s_first_error_block)
2563 printk(": block %llu", (unsigned long long)
2564 le64_to_cpu(es->s_first_error_block));
2565 printk("\n");
2566 }
2567 if (es->s_last_error_time) {
ae0f78de 2568 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2569 sb->s_id, le32_to_cpu(es->s_last_error_time),
2570 (int) sizeof(es->s_last_error_func),
2571 es->s_last_error_func,
2572 le32_to_cpu(es->s_last_error_line));
2573 if (es->s_last_error_ino)
2574 printk(": inode %u",
2575 le32_to_cpu(es->s_last_error_ino));
2576 if (es->s_last_error_block)
2577 printk(": block %llu", (unsigned long long)
2578 le64_to_cpu(es->s_last_error_block));
2579 printk("\n");
2580 }
2581 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2582}
2583
bfff6873
LC		 2584/* Find next suitable group and run ext4_init_inode_table */
2585static int ext4_run_li_request(struct ext4_li_request *elr)
2586{
2587 struct ext4_group_desc *gdp = NULL;
2588 ext4_group_t group, ngroups;
2589 struct super_block *sb;
2590 unsigned long timeout = 0;
2591 int ret = 0;
2592
2593 sb = elr->lr_super;
2594 ngroups = EXT4_SB(sb)->s_groups_count;
2595
8e8ad8a5 2596 sb_start_write(sb);
bfff6873
LC		 2597 for (group = elr->lr_next_group; group < ngroups; group++) {
2598 gdp = ext4_get_group_desc(sb, group, NULL);
2599 if (!gdp) {
2600 ret = 1;
2601 break;
2602 }
2603
2604 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2605 break;
2606 }
2607
7f511862 2608 if (group >= ngroups)
bfff6873
LC		 2609 ret = 1;
2610
2611 if (!ret) {
2612 timeout = jiffies;
2613 ret = ext4_init_inode_table(sb, group,
2614 elr->lr_timeout ? 0 : 1);
2615 if (elr->lr_timeout == 0) {
51ce6511
LC		 2616 timeout = (jiffies - timeout) *
2617 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2618 elr->lr_timeout = timeout;
2619 }
2620 elr->lr_next_sched = jiffies + elr->lr_timeout;
2621 elr->lr_next_group = group + 1;
2622 }
8e8ad8a5 2623 sb_end_write(sb);
bfff6873
LC		 2624
2625 return ret;
2626}
2627
2628/*
2629 * Remove lr_request from the list_request and free the
4ed5c033 2630 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2631 */
2632static void ext4_remove_li_request(struct ext4_li_request *elr)
2633{
2634 struct ext4_sb_info *sbi;
2635
2636 if (!elr)
2637 return;
2638
2639 sbi = elr->lr_sbi;
2640
2641 list_del(&elr->lr_request);
2642 sbi->s_li_request = NULL;
2643 kfree(elr);
2644}
2645
2646static void ext4_unregister_li_request(struct super_block *sb)
2647{
1bb933fb
LC		 2648 mutex_lock(&ext4_li_mtx);
2649 if (!ext4_li_info) {
2650 mutex_unlock(&ext4_li_mtx);
bfff6873 2651 return;
1bb933fb 2652 }
bfff6873
LC		 2653
2654 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2655 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2656 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2657 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2658}
2659
8f1f7453
ES		 2660static struct task_struct *ext4_lazyinit_task;
2661
bfff6873
LC		 2662/*
2663 * This is the function where ext4lazyinit thread lives. It walks
2664 * through the request list searching for next scheduled filesystem.
2665 * When such a fs is found, run the lazy initialization request
2666 * (ext4_rn_li_request) and keep track of the time spend in this
2667 * function. Based on that time we compute next schedule time of
2668 * the request. When walking through the list is complete, compute
2669 * next waking time and put itself into sleep.
2670 */
2671static int ext4_lazyinit_thread(void *arg)
2672{
2673 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2674 struct list_head *pos, *n;
2675 struct ext4_li_request *elr;
4ed5c033 2676 unsigned long next_wakeup, cur;
bfff6873
LC		 2677
2678 BUG_ON(NULL == eli);
2679
bfff6873
LC		 2680cont_thread:
2681 while (true) {
2682 next_wakeup = MAX_JIFFY_OFFSET;
2683
2684 mutex_lock(&eli->li_list_mtx);
2685 if (list_empty(&eli->li_request_list)) {
2686 mutex_unlock(&eli->li_list_mtx);
2687 goto exit_thread;
2688 }
2689
2690 list_for_each_safe(pos, n, &eli->li_request_list) {
2691 elr = list_entry(pos, struct ext4_li_request,
2692 lr_request);
2693
b2c78cd0
TT		 2694 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2695 if (ext4_run_li_request(elr) != 0) {
2696 /* error, remove the lazy_init job */
2697 ext4_remove_li_request(elr);
2698 continue;
2699 }
bfff6873
LC		 2700 }
2701
2702 if (time_before(elr->lr_next_sched, next_wakeup))
2703 next_wakeup = elr->lr_next_sched;
2704 }
2705 mutex_unlock(&eli->li_list_mtx);
2706
a0acae0e 2707 try_to_freeze();
bfff6873 2708
4ed5c033
LC		 2709 cur = jiffies;
2710 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2711 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2712 cond_resched();
2713 continue;
2714 }
2715
4ed5c033
LC		 2716 schedule_timeout_interruptible(next_wakeup - cur);
2717
8f1f7453
ES		 2718 if (kthread_should_stop()) {
2719 ext4_clear_request_list();
2720 goto exit_thread;
2721 }
bfff6873
LC		 2722 }
2723
2724exit_thread:
2725 /*
2726 * It looks like the request list is empty, but we need
2727 * to check it under the li_list_mtx lock, to prevent any
2728 * additions into it, and of course we should lock ext4_li_mtx
2729 * to atomically free the list and ext4_li_info, because at
2730 * this point another ext4 filesystem could be registering
2731 * new one.
2732 */
2733 mutex_lock(&ext4_li_mtx);
2734 mutex_lock(&eli->li_list_mtx);
2735 if (!list_empty(&eli->li_request_list)) {
2736 mutex_unlock(&eli->li_list_mtx);
2737 mutex_unlock(&ext4_li_mtx);
2738 goto cont_thread;
2739 }
2740 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2741 kfree(ext4_li_info);
2742 ext4_li_info = NULL;
2743 mutex_unlock(&ext4_li_mtx);
2744
2745 return 0;
2746}
2747
2748static void ext4_clear_request_list(void)
2749{
2750 struct list_head *pos, *n;
2751 struct ext4_li_request *elr;
2752
2753 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2754 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2755 elr = list_entry(pos, struct ext4_li_request,
2756 lr_request);
2757 ext4_remove_li_request(elr);
2758 }
2759 mutex_unlock(&ext4_li_info->li_list_mtx);
2760}
2761
2762static int ext4_run_lazyinit_thread(void)
2763{
8f1f7453
ES		 2764 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2765 ext4_li_info, "ext4lazyinit");
2766 if (IS_ERR(ext4_lazyinit_task)) {
2767 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2768 ext4_clear_request_list();
bfff6873
LC		 2769 kfree(ext4_li_info);
2770 ext4_li_info = NULL;
92b97816 2771 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2772 "initialization thread\n",
2773 err);
2774 return err;
2775 }
2776 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2777 return 0;
2778}
2779
2780/*
2781 * Check whether it make sense to run itable init. thread or not.
2782 * If there is at least one uninitialized inode table, return
2783 * corresponding group number, else the loop goes through all
2784 * groups and return total number of groups.
2785 */
2786static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2787{
2788 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2789 struct ext4_group_desc *gdp = NULL;
2790
2791 for (group = 0; group < ngroups; group++) {
2792 gdp = ext4_get_group_desc(sb, group, NULL);
2793 if (!gdp)
2794 continue;
2795
2796 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2797 break;
2798 }
2799
2800 return group;
2801}
2802
2803static int ext4_li_info_new(void)
2804{
2805 struct ext4_lazy_init *eli = NULL;
2806
2807 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2808 if (!eli)
2809 return -ENOMEM;
2810
bfff6873
LC		 2811 INIT_LIST_HEAD(&eli->li_request_list);
2812 mutex_init(&eli->li_list_mtx);
2813
bfff6873
LC		 2814 eli->li_state |= EXT4_LAZYINIT_QUIT;
2815
2816 ext4_li_info = eli;
2817
2818 return 0;
2819}
2820
2821static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2822 ext4_group_t start)
2823{
2824 struct ext4_sb_info *sbi = EXT4_SB(sb);
2825 struct ext4_li_request *elr;
bfff6873
LC		 2826
2827 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2828 if (!elr)
2829 return NULL;
2830
2831 elr->lr_super = sb;
2832 elr->lr_sbi = sbi;
2833 elr->lr_next_group = start;
2834
2835 /*
2836 * Randomize first schedule time of the request to
2837 * spread the inode table initialization requests
2838 * better.
2839 */
dd1f723b
TT		 2840 elr->lr_next_sched = jiffies + (prandom_u32() %
2841 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 2842 return elr;
2843}
2844
7f511862
TT		 2845int ext4_register_li_request(struct super_block *sb,
2846 ext4_group_t first_not_zeroed)
bfff6873
LC		 2847{
2848 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 2849 struct ext4_li_request *elr = NULL;
bfff6873 2850 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2851 int ret = 0;
bfff6873 2852
7f511862 2853 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 2854 if (sbi->s_li_request != NULL) {
2855 /*
2856 * Reset timeout so it can be computed again, because
2857 * s_li_wait_mult might have changed.
2858 */
2859 sbi->s_li_request->lr_timeout = 0;
7f511862 2860 goto out;
51ce6511 2861 }
bfff6873
LC		 2862
2863 if (first_not_zeroed == ngroups ||
2864 (sb->s_flags & MS_RDONLY) ||
55ff3840 2865 !test_opt(sb, INIT_INODE_TABLE))
7f511862 2866 goto out;
bfff6873
LC		 2867
2868 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 2869 if (!elr) {
2870 ret = -ENOMEM;
2871 goto out;
2872 }
bfff6873
LC		 2873
2874 if (NULL == ext4_li_info) {
2875 ret = ext4_li_info_new();
2876 if (ret)
2877 goto out;
2878 }
2879
2880 mutex_lock(&ext4_li_info->li_list_mtx);
2881 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2882 mutex_unlock(&ext4_li_info->li_list_mtx);
2883
2884 sbi->s_li_request = elr;
46e4690b
TM		 2885 /*
2886 * set elr to NULL here since it has been inserted to
2887 * the request_list and the removal and free of it is
2888 * handled by ext4_clear_request_list from now on.
2889 */
2890 elr = NULL;
bfff6873
LC		 2891
2892 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2893 ret = ext4_run_lazyinit_thread();
2894 if (ret)
2895 goto out;
2896 }
bfff6873 2897out:
beed5ecb
NK		 2898 mutex_unlock(&ext4_li_mtx);
2899 if (ret)
bfff6873 2900 kfree(elr);
bfff6873
LC		 2901 return ret;
2902}
2903
2904/*
2905 * We do not need to lock anything since this is called on
2906 * module unload.
2907 */
2908static void ext4_destroy_lazyinit_thread(void)
2909{
2910 /*
2911 * If thread exited earlier
2912 * there's nothing to be done.
2913 */
8f1f7453 2914 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 2915 return;
2916
8f1f7453 2917 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 2918}
2919
25ed6e8a
DW		 2920static int set_journal_csum_feature_set(struct super_block *sb)
2921{
2922 int ret = 1;
2923 int compat, incompat;
2924 struct ext4_sb_info *sbi = EXT4_SB(sb);
2925
9aa5d32b 2926 if (ext4_has_metadata_csum(sb)) {
db9ee220 2927 /* journal checksum v3 */
25ed6e8a 2928 compat = 0;
db9ee220 2929 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 2930 } else {
2931 /* journal checksum v1 */
2932 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
2933 incompat = 0;
2934 }
2935
feb8c6d3
DW		 2936 jbd2_journal_clear_features(sbi->s_journal,
2937 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2938 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
2939 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 2940 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2941 ret = jbd2_journal_set_features(sbi->s_journal,
2942 compat, 0,
2943 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
2944 incompat);
2945 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2946 ret = jbd2_journal_set_features(sbi->s_journal,
2947 compat, 0,
2948 incompat);
2949 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2950 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2951 } else {
feb8c6d3
DW		 2952 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2953 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 2954 }
2955
2956 return ret;
2957}
2958
952fc18e
TT		 2959/*
2960 * Note: calculating the overhead so we can be compatible with
2961 * historical BSD practice is quite difficult in the face of
2962 * clusters/bigalloc. This is because multiple metadata blocks from
2963 * different block group can end up in the same allocation cluster.
2964 * Calculating the exact overhead in the face of clustered allocation
2965 * requires either O(all block bitmaps) in memory or O(number of block
2966 * groups**2) in time. We will still calculate the superblock for
2967 * older file systems --- and if we come across with a bigalloc file
2968 * system with zero in s_overhead_clusters the estimate will be close to
2969 * correct especially for very large cluster sizes --- but for newer
2970 * file systems, it's better to calculate this figure once at mkfs
2971 * time, and store it in the superblock. If the superblock value is
2972 * present (even for non-bigalloc file systems), we will use it.
2973 */
2974static int count_overhead(struct super_block *sb, ext4_group_t grp,
2975 char *buf)
2976{
2977 struct ext4_sb_info *sbi = EXT4_SB(sb);
2978 struct ext4_group_desc *gdp;
2979 ext4_fsblk_t first_block, last_block, b;
2980 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
2981 int s, j, count = 0;
2982
0548bbb8
TT		 2983 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
2984 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
2985 sbi->s_itb_per_group + 2);
2986
952fc18e
TT		 2987 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
2988 (grp * EXT4_BLOCKS_PER_GROUP(sb));
2989 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
2990 for (i = 0; i < ngroups; i++) {
2991 gdp = ext4_get_group_desc(sb, i, NULL);
2992 b = ext4_block_bitmap(sb, gdp);
2993 if (b >= first_block && b <= last_block) {
2994 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
2995 count++;
2996 }
2997 b = ext4_inode_bitmap(sb, gdp);
2998 if (b >= first_block && b <= last_block) {
2999 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3000 count++;
3001 }
3002 b = ext4_inode_table(sb, gdp);
3003 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3004 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3005 int c = EXT4_B2C(sbi, b - first_block);
3006 ext4_set_bit(c, buf);
3007 count++;
3008 }
3009 if (i != grp)
3010 continue;
3011 s = 0;
3012 if (ext4_bg_has_super(sb, grp)) {
3013 ext4_set_bit(s++, buf);
3014 count++;
3015 }
3016 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3017 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3018 count++;
3019 }
3020 }
3021 if (!count)
3022 return 0;
3023 return EXT4_CLUSTERS_PER_GROUP(sb) -
3024 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3025}
3026
3027/*
3028 * Compute the overhead and stash it in sbi->s_overhead
3029 */
3030int ext4_calculate_overhead(struct super_block *sb)
3031{
3032 struct ext4_sb_info *sbi = EXT4_SB(sb);
3033 struct ext4_super_block *es = sbi->s_es;
3034 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3035 ext4_fsblk_t overhead = 0;
4fdb5543 3036 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3037
952fc18e
TT		 3038 if (!buf)
3039 return -ENOMEM;
3040
3041 /*
3042 * Compute the overhead (FS structures). This is constant
3043 * for a given filesystem unless the number of block groups
3044 * changes so we cache the previous value until it does.
3045 */
3046
3047 /*
3048 * All of the blocks before first_data_block are overhead
3049 */
3050 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3051
3052 /*
3053 * Add the overhead found in each block group
3054 */
3055 for (i = 0; i < ngroups; i++) {
3056 int blks;
3057
3058 blks = count_overhead(sb, i, buf);
3059 overhead += blks;
3060 if (blks)
3061 memset(buf, 0, PAGE_SIZE);
3062 cond_resched();
3063 }
b003b524
ES		 3064 /* Add the internal journal blocks as well */
3065 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3066 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3067
952fc18e
TT		 3068 sbi->s_overhead = overhead;
3069 smp_wmb();
3070 free_page((unsigned long) buf);
3071 return 0;
3072}
3073
b5799018 3074static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3075{
3076 ext4_fsblk_t resv_clusters;
b5799018 3077 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3078
30fac0f7
JK		 3079 /*
3080 * There's no need to reserve anything when we aren't using extents.
3081 * The space estimates are exact, there are no unwritten extents,
3082 * hole punching doesn't need new metadata... This is needed especially
3083 * to keep ext2/3 backward compatibility.
3084 */
3085 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
b5799018 3086 return;
27dd4385
LC		 3087 /*
3088 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3089 * This should cover the situations where we can not afford to run
3090 * out of space like for example punch hole, or converting
556615dc 3091 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3092 * allocation would require 1, or 2 blocks, higher numbers are
3093 * very rare.
3094 */
b5799018
TT		 3095 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3096 sbi->s_cluster_bits);
27dd4385
LC		 3097
3098 do_div(resv_clusters, 50);
3099 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3100
b5799018 3101 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3102}
3103
2b2d6d01 3104static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3105{
d4c402d9 3106 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3107 struct buffer_head *bh;
617ba13b
MC		 3108 struct ext4_super_block *es = NULL;
3109 struct ext4_sb_info *sbi;
3110 ext4_fsblk_t block;
3111 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3112 ext4_fsblk_t logical_sb_block;
ac27a0ec 3113 unsigned long offset = 0;
ac27a0ec
DK		 3114 unsigned long journal_devnum = 0;
3115 unsigned long def_mount_opts;
3116 struct inode *root;
0390131b 3117 const char *descr;
dcc7dae3 3118 int ret = -ENOMEM;
281b5995 3119 int blocksize, clustersize;
4ec11028
TT		 3120 unsigned int db_count;
3121 unsigned int i;
281b5995 3122 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3123 __u64 blocks_count;
07aa2ea1 3124 int err = 0;
b3881f74 3125 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3126 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3127
3128 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3129 if (!sbi)
dcc7dae3 3130 goto out_free_orig;
705895b6
PE		 3131
3132 sbi->s_blockgroup_lock =
3133 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3134 if (!sbi->s_blockgroup_lock) {
3135 kfree(sbi);
dcc7dae3 3136 goto out_free_orig;
705895b6 3137 }
ac27a0ec 3138 sb->s_fs_info = sbi;
2c0544b2 3139 sbi->s_sb = sb;
240799cd 3140 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3141 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3142 if (sb->s_bdev->bd_part)
3143 sbi->s_sectors_written_start =
3144 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3145
9f6200bb 3146 /* Cleanup superblock name */
ec3904dc 3147 strreplace(sb->s_id, '/', '!');
9f6200bb 3148
07aa2ea1 3149 /* -EINVAL is default */
dcc7dae3 3150 ret = -EINVAL;
617ba13b 3151 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3152 if (!blocksize) {
b31e1552 3153 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3154 goto out_fail;
3155 }
3156
3157 /*
617ba13b 3158 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3159 * block sizes. We need to calculate the offset from buffer start.
3160 */
617ba13b 3161 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3162 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3163 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3164 } else {
70bbb3e0 3165 logical_sb_block = sb_block;
ac27a0ec
DK		 3166 }
3167
a8ac900b 3168 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3169 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3170 goto out_fail;
3171 }
3172 /*
3173 * Note: s_es must be initialized as soon as possible because
617ba13b 3174 * some ext4 macro-instructions depend on its value
ac27a0ec 3175 */
2716b802 3176 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3177 sbi->s_es = es;
3178 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3179 if (sb->s_magic != EXT4_SUPER_MAGIC)
3180 goto cantfind_ext4;
afc32f7e 3181 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3182
feb0ab32
DW		 3183 /* Warn if metadata_csum and gdt_csum are both set. */
3184 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3185 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3186 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
363307e6 3187 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3188 "redundant flags; please run fsck.");
3189
d25425f8
DW		 3190 /* Check for a known checksum algorithm */
3191 if (!ext4_verify_csum_type(sb, es)) {
3192 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3193 "unknown checksum algorithm.");
3194 silent = 1;
3195 goto cantfind_ext4;
3196 }
3197
0441984a
DW		 3198 /* Load the checksum driver */
3199 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3200 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3201 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3202 if (IS_ERR(sbi->s_chksum_driver)) {
3203 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3204 ret = PTR_ERR(sbi->s_chksum_driver);
3205 sbi->s_chksum_driver = NULL;
3206 goto failed_mount;
3207 }
3208 }
3209
a9c47317
DW		 3210 /* Check superblock checksum */
3211 if (!ext4_superblock_csum_verify(sb, es)) {
3212 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3213 "invalid superblock checksum. Run e2fsck?");
3214 silent = 1;
3215 goto cantfind_ext4;
3216 }
3217
3218 /* Precompute checksum seed for all metadata */
9aa5d32b 3219 if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3220 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3221 sizeof(es->s_uuid));
3222
ac27a0ec
DK		 3223 /* Set defaults before we parse the mount options */
3224 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3225 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3226 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3227 set_opt(sb, DEBUG);
87f26807 3228 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3229 set_opt(sb, GRPID);
617ba13b 3230 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3231 set_opt(sb, NO_UID32);
ea663336 3232 /* xattr user namespace & acls are now defaulted on */
ea663336 3233 set_opt(sb, XATTR_USER);
03010a33 3234#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3235 set_opt(sb, POSIX_ACL);
2e7842b8 3236#endif
98c1a759
DW		 3237 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3238 if (ext4_has_metadata_csum(sb))
3239 set_opt(sb, JOURNAL_CHECKSUM);
3240
617ba13b 3241 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3242 set_opt(sb, JOURNAL_DATA);
617ba13b 3243 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3244 set_opt(sb, ORDERED_DATA);
617ba13b 3245 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3246 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3247
3248 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3249 set_opt(sb, ERRORS_PANIC);
bb4f397a 3250 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3251 set_opt(sb, ERRORS_CONT);
bb4f397a 3252 else
fd8c37ec 3253 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3254 /* block_validity enabled by default; disable with noblock_validity */
3255 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3256 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3257 set_opt(sb, DISCARD);
ac27a0ec 3258
08cefc7a
EB		 3259 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3260 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3261 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3262 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3263 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3264
8b67f04a 3265 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3266 set_opt(sb, BARRIER);
ac27a0ec 3267
dd919b98
AK		 3268 /*
3269 * enable delayed allocation by default
3270 * Use -o nodelalloc to turn it off
3271 */
bc0b75f7 3272 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3273 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3274 set_opt(sb, DELALLOC);
dd919b98 3275
51ce6511
LC		 3276 /*
3277 * set default s_li_wait_mult for lazyinit, for the case there is
3278 * no mount option specified.
3279 */
3280 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3281
8b67f04a 3282 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3283 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3284 ext4_msg(sb, KERN_WARNING,
3285 "failed to parse options in superblock: %s",
3286 sbi->s_es->s_mount_opts);
3287 }
5a916be1 3288 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3289 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3290 &journal_ioprio, 0))
ac27a0ec
DK		 3291 goto failed_mount;
3292
56889787
TT		 3293 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3294 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3295 "with data=journal disables delayed "
3296 "allocation and O_DIRECT support!\n");
3297 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3298 ext4_msg(sb, KERN_ERR, "can't mount with "
3299 "both data=journal and delalloc");
3300 goto failed_mount;
3301 }
3302 if (test_opt(sb, DIOREAD_NOLOCK)) {
3303 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3304 "both data=journal and dioread_nolock");
56889787
TT		 3305 goto failed_mount;
3306 }
923ae0ff
RZ		 3307 if (test_opt(sb, DAX)) {
3308 ext4_msg(sb, KERN_ERR, "can't mount with "
3309 "both data=journal and dax");
3310 goto failed_mount;
3311 }
56889787
TT		 3312 if (test_opt(sb, DELALLOC))
3313 clear_opt(sb, DELALLOC);
001e4a87
TH		 3314 } else {
3315 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3316 }
3317
ac27a0ec 3318 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3319 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3320
617ba13b
MC		 3321 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3322 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3323 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3324 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3325 ext4_msg(sb, KERN_WARNING,
3326 "feature flags set on rev 0 fs, "
3327 "running e2fsck is recommended");
469108ff 3328
ed3654eb
TT		 3329 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3330 set_opt2(sb, HURD_COMPAT);
3331 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3332 EXT4_FEATURE_INCOMPAT_64BIT)) {
3333 ext4_msg(sb, KERN_ERR,
3334 "The Hurd can't support 64-bit file systems");
3335 goto failed_mount;
3336 }
3337 }
3338
2035e776
TT		 3339 if (IS_EXT2_SB(sb)) {
3340 if (ext2_feature_set_ok(sb))
3341 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3342 "using the ext4 subsystem");
3343 else {
3344 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3345 "to feature incompatibilities");
3346 goto failed_mount;
3347 }
3348 }
3349
3350 if (IS_EXT3_SB(sb)) {
3351 if (ext3_feature_set_ok(sb))
3352 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3353 "using the ext4 subsystem");
3354 else {
3355 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3356 "to feature incompatibilities");
3357 goto failed_mount;
3358 }
3359 }
3360
ac27a0ec
DK		 3361 /*
3362 * Check feature flags regardless of the revision level, since we
3363 * previously didn't change the revision level when setting the flags,
3364 * so there is a chance incompat flags are set on a rev 0 filesystem.
3365 */
a13fb1a4 3366 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3367 goto failed_mount;
a13fb1a4 3368
261cb20c 3369 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3370 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3371 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3372 ext4_msg(sb, KERN_ERR,
3373 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3374 goto failed_mount;
3375 }
3376
923ae0ff
RZ		 3377 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3378 if (blocksize != PAGE_SIZE) {
3379 ext4_msg(sb, KERN_ERR,
3380 "error: unsupported blocksize for dax");
3381 goto failed_mount;
3382 }
3383 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3384 ext4_msg(sb, KERN_ERR,
3385 "error: device does not support dax");
3386 goto failed_mount;
3387 }
3388 }
3389
6ddb2447
TT		 3390 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT) &&
3391 es->s_encryption_level) {
3392 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3393 es->s_encryption_level);
3394 goto failed_mount;
3395 }
3396
ac27a0ec 3397 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3398 /* Validate the filesystem blocksize */
3399 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3400 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3401 blocksize);
ac27a0ec
DK		 3402 goto failed_mount;
3403 }
3404
2b2d6d01 3405 brelse(bh);
70bbb3e0
AM		 3406 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3407 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3408 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3409 if (!bh) {
b31e1552
ES		 3410 ext4_msg(sb, KERN_ERR,
3411 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3412 goto failed_mount;
3413 }
2716b802 3414 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3415 sbi->s_es = es;
617ba13b 3416 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3417 ext4_msg(sb, KERN_ERR,
3418 "Magic mismatch, very weird!");
ac27a0ec
DK		 3419 goto failed_mount;
3420 }
3421 }
3422
a13fb1a4
ES		 3423 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3424 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3425 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3426 has_huge_files);
3427 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3428
617ba13b
MC		 3429 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3430 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3431 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3432 } else {
3433 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3434 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3435 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3436 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3437 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3438 ext4_msg(sb, KERN_ERR,
3439 "unsupported inode size: %d",
2b2d6d01 3440 sbi->s_inode_size);
ac27a0ec
DK		 3441 goto failed_mount;
3442 }
ef7f3835
KS		 3443 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3444 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3445 }
0b8e58a1 3446
0d1ee42f
AR		 3447 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3448 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3449 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3450 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3451 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3452 ext4_msg(sb, KERN_ERR,
3453 "unsupported descriptor size %lu",
0d1ee42f
AR		 3454 sbi->s_desc_size);
3455 goto failed_mount;
3456 }
3457 } else
3458 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3459
ac27a0ec 3460 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3461 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3462 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3463 goto cantfind_ext4;
0b8e58a1 3464
617ba13b 3465 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3466 if (sbi->s_inodes_per_block == 0)
617ba13b 3467 goto cantfind_ext4;
ac27a0ec
DK		 3468 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3469 sbi->s_inodes_per_block;
0d1ee42f 3470 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3471 sbi->s_sbh = bh;
3472 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3473 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3474 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3475
2b2d6d01 3476 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3477 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3478 sbi->s_def_hash_version = es->s_def_hash_version;
23301410
TT		 3479 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3480 i = le32_to_cpu(es->s_flags);
3481 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3482 sbi->s_hash_unsigned = 3;
3483 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3484#ifdef __CHAR_UNSIGNED__
23301410
TT		 3485 if (!(sb->s_flags & MS_RDONLY))
3486 es->s_flags |=
3487 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3488 sbi->s_hash_unsigned = 3;
f99b2589 3489#else
23301410
TT		 3490 if (!(sb->s_flags & MS_RDONLY))
3491 es->s_flags |=
3492 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3493#endif
23301410 3494 }
f99b2589 3495 }
ac27a0ec 3496
281b5995
TT		 3497 /* Handle clustersize */
3498 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3499 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3500 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3501 if (has_bigalloc) {
3502 if (clustersize < blocksize) {
3503 ext4_msg(sb, KERN_ERR,
3504 "cluster size (%d) smaller than "
3505 "block size (%d)", clustersize, blocksize);
3506 goto failed_mount;
3507 }
3508 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3509 le32_to_cpu(es->s_log_block_size);
3510 sbi->s_clusters_per_group =
3511 le32_to_cpu(es->s_clusters_per_group);
3512 if (sbi->s_clusters_per_group > blocksize * 8) {
3513 ext4_msg(sb, KERN_ERR,
3514 "#clusters per group too big: %lu",
3515 sbi->s_clusters_per_group);
3516 goto failed_mount;
3517 }
3518 if (sbi->s_blocks_per_group !=
3519 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3520 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3521 "clusters per group (%lu) inconsistent",
3522 sbi->s_blocks_per_group,
3523 sbi->s_clusters_per_group);
3524 goto failed_mount;
3525 }
3526 } else {
3527 if (clustersize != blocksize) {
3528 ext4_warning(sb, "fragment/cluster size (%d) != "
3529 "block size (%d)", clustersize,
3530 blocksize);
3531 clustersize = blocksize;
3532 }
3533 if (sbi->s_blocks_per_group > blocksize * 8) {
3534 ext4_msg(sb, KERN_ERR,
3535 "#blocks per group too big: %lu",
3536 sbi->s_blocks_per_group);
3537 goto failed_mount;
3538 }
3539 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3540 sbi->s_cluster_bits = 0;
ac27a0ec 3541 }
281b5995
TT		 3542 sbi->s_cluster_ratio = clustersize / blocksize;
3543
ac27a0ec 3544 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3545 ext4_msg(sb, KERN_ERR,
3546 "#inodes per group too big: %lu",
2b2d6d01 3547 sbi->s_inodes_per_group);
ac27a0ec
DK		 3548 goto failed_mount;
3549 }
3550
960fd856
TT		 3551 /* Do we have standard group size of clustersize * 8 blocks ? */
3552 if (sbi->s_blocks_per_group == clustersize << 3)
3553 set_opt2(sb, STD_GROUP_SIZE);
3554
bf43d84b
ES		 3555 /*
3556 * Test whether we have more sectors than will fit in sector_t,
3557 * and whether the max offset is addressable by the page cache.
3558 */
5a9ae68a 3559 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3560 ext4_blocks_count(es));
5a9ae68a 3561 if (err) {
b31e1552 3562 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3563 " too large to mount safely on this system");
ac27a0ec 3564 if (sizeof(sector_t) < 8)
90c699a9 3565 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3566 goto failed_mount;
3567 }
3568
617ba13b
MC		 3569 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3570 goto cantfind_ext4;
e7c95593 3571
0f2ddca6
FTN		 3572 /* check blocks count against device size */
3573 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3574 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3575 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3576 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3577 ext4_blocks_count(es), blocks_count);
3578 goto failed_mount;
3579 }
3580
0b8e58a1
AD		 3581 /*
3582 * It makes no sense for the first data block to be beyond the end
3583 * of the filesystem.
3584 */
3585 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3586 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3587 "block %u is beyond end of filesystem (%llu)",
3588 le32_to_cpu(es->s_first_data_block),
3589 ext4_blocks_count(es));
e7c95593
ES		 3590 goto failed_mount;
3591 }
bd81d8ee
LV		 3592 blocks_count = (ext4_blocks_count(es) -
3593 le32_to_cpu(es->s_first_data_block) +
3594 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3595 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3596 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3597 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3598 "(block count %llu, first data block %u, "
b31e1552 3599 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3600 ext4_blocks_count(es),
3601 le32_to_cpu(es->s_first_data_block),
3602 EXT4_BLOCKS_PER_GROUP(sb));
3603 goto failed_mount;
3604 }
bd81d8ee 3605 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3606 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3607 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3608 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3609 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3610 sbi->s_group_desc = ext4_kvmalloc(db_count *
3611 sizeof(struct buffer_head *),
3612 GFP_KERNEL);
ac27a0ec 3613 if (sbi->s_group_desc == NULL) {
b31e1552 3614 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3615 ret = -ENOMEM;
ac27a0ec
DK		 3616 goto failed_mount;
3617 }
3618
9f6200bb
TT		 3619 if (ext4_proc_root)
3620 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3621
66acdcf4
TT		 3622 if (sbi->s_proc)
3623 proc_create_data("options", S_IRUGO, sbi->s_proc,
3624 &ext4_seq_options_fops, sb);
3625
705895b6 3626 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3627
3628 for (i = 0; i < db_count; i++) {
70bbb3e0 3629 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3630 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3631 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3632 ext4_msg(sb, KERN_ERR,
3633 "can't read group descriptor %d", i);
ac27a0ec
DK		 3634 db_count = i;
3635 goto failed_mount2;
3636 }
3637 }
bfff6873 3638 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3639 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
f9ae9cf5 3640 goto failed_mount2;
ac27a0ec 3641 }
772cb7c8 3642
f9ae9cf5 3643 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3644 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3645 spin_lock_init(&sbi->s_next_gen_lock);
3646
04ecddb7
JM		 3647 setup_timer(&sbi->s_err_report, print_daily_error_info,
3648 (unsigned long) sb);
04496411 3649
a75ae78f 3650 /* Register extent status tree shrinker */
eb68d0e2 3651 if (ext4_es_register_shrinker(sbi))
ce7e010a 3652 goto failed_mount3;
ce7e010a 3653
c9de560d 3654 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3655 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3656
f9ae9cf5
TT		 3657 /*
3658 * set up enough so that it can read an inode
3659 */
f6e63f90 3660 sb->s_op = &ext4_sops;
617ba13b
MC		 3661 sb->s_export_op = &ext4_export_ops;
3662 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3663#ifdef CONFIG_QUOTA
617ba13b 3664 sb->dq_op = &ext4_quota_operations;
262b4662 3665 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
1fa5efe3 3666 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3667 else
3668 sb->s_qcop = &ext4_qctl_operations;
96c7e0d9 3669 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
ac27a0ec 3670#endif
f2fa2ffc
AK		 3671 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3672
ac27a0ec 3673 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3674 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3675
3676 sb->s_root = NULL;
3677
3678 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3679 EXT4_HAS_INCOMPAT_FEATURE(sb,
3680 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3681
c5e06d10
JL		 3682 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3683 !(sb->s_flags & MS_RDONLY))
3684 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3685 goto failed_mount3a;
c5e06d10 3686
ac27a0ec
DK		 3687 /*
3688 * The first inode we look at is the journal inode. Don't try
3689 * root first: it may be modified in the journal!
3690 */
3691 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3692 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3693 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 3694 goto failed_mount3a;
0390131b
FM		 3695 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3696 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3697 ext4_msg(sb, KERN_ERR, "required journal recovery "
3698 "suppressed and not mounted read-only");
744692dc 3699 goto failed_mount_wq;
ac27a0ec 3700 } else {
fd8c37ec 3701 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3702 sbi->s_journal = NULL;
3703 needs_recovery = 0;
3704 goto no_journal;
ac27a0ec
DK		 3705 }
3706
f32aaf2d 3707 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3708 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3709 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3710 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3711 goto failed_mount_wq;
eb40a09c
JS		 3712 }
3713
25ed6e8a
DW		 3714 if (!set_journal_csum_feature_set(sb)) {
3715 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3716 "feature set");
3717 goto failed_mount_wq;
d4da6c9c 3718 }
818d276c 3719
ac27a0ec
DK		 3720 /* We have now updated the journal if required, so we can
3721 * validate the data journaling mode. */
3722 switch (test_opt(sb, DATA_FLAGS)) {
3723 case 0:
3724 /* No mode set, assume a default based on the journal
63f57933
AM		 3725 * capabilities: ORDERED_DATA if the journal can
3726 * cope, else JOURNAL_DATA
3727 */
dab291af
MC		 3728 if (jbd2_journal_check_available_features
3729 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3730 set_opt(sb, ORDERED_DATA);
ac27a0ec 3731 else
fd8c37ec 3732 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3733 break;
3734
617ba13b
MC		 3735 case EXT4_MOUNT_ORDERED_DATA:
3736 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3737 if (!jbd2_journal_check_available_features
3738 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3739 ext4_msg(sb, KERN_ERR, "Journal does not support "
3740 "requested data journaling mode");
744692dc 3741 goto failed_mount_wq;
ac27a0ec
DK		 3742 }
3743 default:
3744 break;
3745 }
b3881f74 3746 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3747
18aadd47
BJ		 3748 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3749
ce7e010a 3750no_journal:
9c191f70
M		 3751 if (ext4_mballoc_ready) {
3752 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
3753 if (!sbi->s_mb_cache) {
3754 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3755 goto failed_mount_wq;
3756 }
3757 }
3758
1cb767cd
TT		 3759 if ((DUMMY_ENCRYPTION_ENABLED(sbi) ||
3760 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) &&
3761 (blocksize != PAGE_CACHE_SIZE)) {
3762 ext4_msg(sb, KERN_ERR,
3763 "Unsupported blocksize for fs encryption");
3764 goto failed_mount_wq;
3765 }
3766
3767 if (DUMMY_ENCRYPTION_ENABLED(sbi) &&
6ddb2447
TT		 3768 !(sb->s_flags & MS_RDONLY) &&
3769 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) {
3770 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
3771 ext4_commit_super(sb, 1);
3772 }
3773
952fc18e
TT		 3774 /*
3775 * Get the # of file system overhead blocks from the
3776 * superblock if present.
3777 */
3778 if (es->s_overhead_clusters)
3779 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3780 else {
07aa2ea1
LC		 3781 err = ext4_calculate_overhead(sb);
3782 if (err)
952fc18e
TT		 3783 goto failed_mount_wq;
3784 }
3785
fd89d5f2
TH		 3786 /*
3787 * The maximum number of concurrent works can be high and
3788 * concurrency isn't really necessary. Limit it to 1.
3789 */
2e8fa54e
JK		 3790 EXT4_SB(sb)->rsv_conversion_wq =
3791 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3792 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3793 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 3794 ret = -ENOMEM;
2e8fa54e
JK		 3795 goto failed_mount4;
3796 }
3797
ac27a0ec 3798 /*
dab291af 3799 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3800 * so we can safely mount the rest of the filesystem now.
3801 */
3802
1d1fe1ee
DH		 3803 root = ext4_iget(sb, EXT4_ROOT_INO);
3804 if (IS_ERR(root)) {
b31e1552 3805 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3806 ret = PTR_ERR(root);
32a9bb57 3807 root = NULL;
ac27a0ec
DK		 3808 goto failed_mount4;
3809 }
3810 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3811 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3812 iput(root);
ac27a0ec
DK		 3813 goto failed_mount4;
3814 }
48fde701 3815 sb->s_root = d_make_root(root);
1d1fe1ee 3816 if (!sb->s_root) {
b31e1552 3817 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3818 ret = -ENOMEM;
3819 goto failed_mount4;
3820 }
ac27a0ec 3821
7e84b621
ES		 3822 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3823 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3824
3825 /* determine the minimum size of new large inodes, if present */
3826 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3827 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3828 EXT4_GOOD_OLD_INODE_SIZE;
3829 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3830 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3831 if (sbi->s_want_extra_isize <
3832 le16_to_cpu(es->s_want_extra_isize))
3833 sbi->s_want_extra_isize =
3834 le16_to_cpu(es->s_want_extra_isize);
3835 if (sbi->s_want_extra_isize <
3836 le16_to_cpu(es->s_min_extra_isize))
3837 sbi->s_want_extra_isize =
3838 le16_to_cpu(es->s_min_extra_isize);
3839 }
3840 }
3841 /* Check if enough inode space is available */
3842 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3843 sbi->s_inode_size) {
3844 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3845 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3846 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3847 "available");
ef7f3835
KS		 3848 }
3849
b5799018 3850 ext4_set_resv_clusters(sb);
27dd4385 3851
6fd058f7
TT		 3852 err = ext4_setup_system_zone(sb);
3853 if (err) {
b31e1552 3854 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3855 "zone (%d)", err);
f9ae9cf5
TT		 3856 goto failed_mount4a;
3857 }
3858
3859 ext4_ext_init(sb);
3860 err = ext4_mb_init(sb);
3861 if (err) {
3862 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3863 err);
dcf2d804 3864 goto failed_mount5;
c2774d84
AK		 3865 }
3866
d5e03cbb
TT		 3867 block = ext4_count_free_clusters(sb);
3868 ext4_free_blocks_count_set(sbi->s_es,
3869 EXT4_C2B(sbi, block));
908c7f19
TH		 3870 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
3871 GFP_KERNEL);
d5e03cbb
TT		 3872 if (!err) {
3873 unsigned long freei = ext4_count_free_inodes(sb);
3874 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 3875 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
3876 GFP_KERNEL);
d5e03cbb
TT		 3877 }
3878 if (!err)
3879 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 3880 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 3881 if (!err)
908c7f19
TH		 3882 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
3883 GFP_KERNEL);
d5e03cbb
TT		 3884 if (err) {
3885 ext4_msg(sb, KERN_ERR, "insufficient memory");
3886 goto failed_mount6;
3887 }
3888
3889 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3890 if (!ext4_fill_flex_info(sb)) {
3891 ext4_msg(sb, KERN_ERR,
3892 "unable to initialize "
3893 "flex_bg meta info!");
3894 goto failed_mount6;
3895 }
3896
bfff6873
LC		 3897 err = ext4_register_li_request(sb, first_not_zeroed);
3898 if (err)
dcf2d804 3899 goto failed_mount6;
bfff6873 3900
b5799018 3901 err = ext4_register_sysfs(sb);
dcf2d804
TM		 3902 if (err)
3903 goto failed_mount7;
3197ebdb 3904
9b2ff357
JK		 3905#ifdef CONFIG_QUOTA
3906 /* Enable quota usage during mount. */
3907 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
3908 !(sb->s_flags & MS_RDONLY)) {
3909 err = ext4_enable_quotas(sb);
3910 if (err)
3911 goto failed_mount8;
3912 }
3913#endif /* CONFIG_QUOTA */
3914
617ba13b
MC		 3915 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3916 ext4_orphan_cleanup(sb, es);
3917 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3918 if (needs_recovery) {
b31e1552 3919 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3920 ext4_mark_recovery_complete(sb, es);
3921 }
3922 if (EXT4_SB(sb)->s_journal) {
3923 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3924 descr = " journalled data mode";
3925 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3926 descr = " ordered data mode";
3927 else
3928 descr = " writeback data mode";
3929 } else
3930 descr = "out journal";
3931
79add3a3
LC		 3932 if (test_opt(sb, DISCARD)) {
3933 struct request_queue *q = bdev_get_queue(sb->s_bdev);
3934 if (!blk_queue_discard(q))
3935 ext4_msg(sb, KERN_WARNING,
3936 "mounting with \"discard\" option, but "
3937 "the device does not support discard");
3938 }
3939
e294a537
TT		 3940 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
3941 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3942 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3943 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3944
66e61a9e
TT		 3945 if (es->s_error_count)
3946 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3947
efbed4dc
TT		 3948 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
3949 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
3950 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
3951 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
3952
d4c402d9 3953 kfree(orig_data);
ac27a0ec
DK		 3954 return 0;
3955
617ba13b 3956cantfind_ext4:
ac27a0ec 3957 if (!silent)
b31e1552 3958 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3959 goto failed_mount;
3960
72ba7450
TT		 3961#ifdef CONFIG_QUOTA
3962failed_mount8:
3963 kobject_del(&sbi->s_kobj);
3964#endif
dcf2d804
TM		 3965failed_mount7:
3966 ext4_unregister_li_request(sb);
3967failed_mount6:
f9ae9cf5 3968 ext4_mb_release(sb);
d5e03cbb 3969 if (sbi->s_flex_groups)
b93b41d4 3970 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 3971 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3972 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3973 percpu_counter_destroy(&sbi->s_dirs_counter);
3974 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 3975failed_mount5:
f9ae9cf5
TT		 3976 ext4_ext_release(sb);
3977 ext4_release_system_zone(sb);
3978failed_mount4a:
94bf608a 3979 dput(sb->s_root);
32a9bb57 3980 sb->s_root = NULL;
94bf608a 3981failed_mount4:
b31e1552 3982 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 3983 if (EXT4_SB(sb)->rsv_conversion_wq)
3984 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 3985failed_mount_wq:
0390131b
FM		 3986 if (sbi->s_journal) {
3987 jbd2_journal_destroy(sbi->s_journal);
3988 sbi->s_journal = NULL;
3989 }
50460fe8 3990failed_mount3a:
d3922a77 3991 ext4_es_unregister_shrinker(sbi);
eb68d0e2 3992failed_mount3:
9105bb14 3993 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 3994 if (sbi->s_mmp_tsk)
3995 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 3996failed_mount2:
3997 for (i = 0; i < db_count; i++)
3998 brelse(sbi->s_group_desc[i]);
b93b41d4 3999 kvfree(sbi->s_group_desc);
ac27a0ec 4000failed_mount:
0441984a
DW		 4001 if (sbi->s_chksum_driver)
4002 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4003 if (sbi->s_proc) {
66acdcf4 4004 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4005 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4006 }
ac27a0ec 4007#ifdef CONFIG_QUOTA
a2d4a646 4008 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4009 kfree(sbi->s_qf_names[i]);
4010#endif
617ba13b 4011 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4012 brelse(bh);
4013out_fail:
4014 sb->s_fs_info = NULL;
f6830165 4015 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4016 kfree(sbi);
dcc7dae3 4017out_free_orig:
d4c402d9 4018 kfree(orig_data);
07aa2ea1 4019 return err ? err : ret;
ac27a0ec
DK		 4020}
4021
4022/*
4023 * Setup any per-fs journal parameters now. We'll do this both on
4024 * initial mount, once the journal has been initialised but before we've
4025 * done any recovery; and again on any subsequent remount.
4026 */
617ba13b 4027static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4028{
617ba13b 4029 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4030
30773840
TT		 4031 journal->j_commit_interval = sbi->s_commit_interval;
4032 journal->j_min_batch_time = sbi->s_min_batch_time;
4033 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4034
a931da6a 4035 write_lock(&journal->j_state_lock);
ac27a0ec 4036 if (test_opt(sb, BARRIER))
dab291af 4037 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4038 else
dab291af 4039 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4040 if (test_opt(sb, DATA_ERR_ABORT))
4041 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4042 else
4043 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4044 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4045}
4046
617ba13b 4047static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4048 unsigned int journal_inum)
4049{
4050 struct inode *journal_inode;
4051 journal_t *journal;
4052
0390131b
FM		 4053 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4054
ac27a0ec
DK		 4055 /* First, test for the existence of a valid inode on disk. Bad
4056 * things happen if we iget() an unused inode, as the subsequent
4057 * iput() will try to delete it. */
4058
1d1fe1ee
DH		 4059 journal_inode = ext4_iget(sb, journal_inum);
4060 if (IS_ERR(journal_inode)) {
b31e1552 4061 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4062 return NULL;
4063 }
4064 if (!journal_inode->i_nlink) {
4065 make_bad_inode(journal_inode);
4066 iput(journal_inode);
b31e1552 4067 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4068 return NULL;
4069 }
4070
e5f8eab8 4071 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4072 journal_inode, journal_inode->i_size);
1d1fe1ee 4073 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4074 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4075 iput(journal_inode);
4076 return NULL;
4077 }
4078
dab291af 4079 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4080 if (!journal) {
b31e1552 4081 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4082 iput(journal_inode);
4083 return NULL;
4084 }
4085 journal->j_private = sb;
617ba13b 4086 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4087 return journal;
4088}
4089
617ba13b 4090static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4091 dev_t j_dev)
4092{
2b2d6d01 4093 struct buffer_head *bh;
ac27a0ec 4094 journal_t *journal;
617ba13b
MC		 4095 ext4_fsblk_t start;
4096 ext4_fsblk_t len;
ac27a0ec 4097 int hblock, blocksize;
617ba13b 4098 ext4_fsblk_t sb_block;
ac27a0ec 4099 unsigned long offset;
2b2d6d01 4100 struct ext4_super_block *es;
ac27a0ec
DK		 4101 struct block_device *bdev;
4102
0390131b
FM		 4103 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4104
b31e1552 4105 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4106 if (bdev == NULL)
4107 return NULL;
4108
ac27a0ec 4109 blocksize = sb->s_blocksize;
e1defc4f 4110 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4111 if (blocksize < hblock) {
b31e1552
ES		 4112 ext4_msg(sb, KERN_ERR,
4113 "blocksize too small for journal device");
ac27a0ec
DK		 4114 goto out_bdev;
4115 }
4116
617ba13b
MC		 4117 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4118 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4119 set_blocksize(bdev, blocksize);
4120 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4121 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4122 "external journal");
ac27a0ec
DK		 4123 goto out_bdev;
4124 }
4125
2716b802 4126 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4127 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4128 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4129 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4130 ext4_msg(sb, KERN_ERR, "external journal has "
4131 "bad superblock");
ac27a0ec
DK		 4132 brelse(bh);
4133 goto out_bdev;
4134 }
4135
df4763be
DW		 4136 if ((le32_to_cpu(es->s_feature_ro_compat) &
4137 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4138 es->s_checksum != ext4_superblock_csum(sb, es)) {
4139 ext4_msg(sb, KERN_ERR, "external journal has "
4140 "corrupt superblock");
4141 brelse(bh);
4142 goto out_bdev;
4143 }
4144
617ba13b 4145 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4146 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4147 brelse(bh);
4148 goto out_bdev;
4149 }
4150
bd81d8ee 4151 len = ext4_blocks_count(es);
ac27a0ec
DK		 4152 start = sb_block + 1;
4153 brelse(bh); /* we're done with the superblock */
4154
dab291af 4155 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4156 start, len, blocksize);
4157 if (!journal) {
b31e1552 4158 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4159 goto out_bdev;
4160 }
4161 journal->j_private = sb;
9f203507 4162 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4163 wait_on_buffer(journal->j_sb_buffer);
4164 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4165 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4166 goto out_journal;
4167 }
4168 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4169 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4170 "user (unsupported) - %d",
ac27a0ec
DK		 4171 be32_to_cpu(journal->j_superblock->s_nr_users));
4172 goto out_journal;
4173 }
617ba13b
MC		 4174 EXT4_SB(sb)->journal_bdev = bdev;
4175 ext4_init_journal_params(sb, journal);
ac27a0ec 4176 return journal;
0b8e58a1 4177
ac27a0ec 4178out_journal:
dab291af 4179 jbd2_journal_destroy(journal);
ac27a0ec 4180out_bdev:
617ba13b 4181 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4182 return NULL;
4183}
4184
617ba13b
MC		 4185static int ext4_load_journal(struct super_block *sb,
4186 struct ext4_super_block *es,
ac27a0ec
DK		 4187 unsigned long journal_devnum)
4188{
4189 journal_t *journal;
4190 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4191 dev_t journal_dev;
4192 int err = 0;
4193 int really_read_only;
4194
0390131b
FM		 4195 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4196
ac27a0ec
DK		 4197 if (journal_devnum &&
4198 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4199 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4200 "numbers have changed");
ac27a0ec
DK		 4201 journal_dev = new_decode_dev(journal_devnum);
4202 } else
4203 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4204
4205 really_read_only = bdev_read_only(sb->s_bdev);
4206
4207 /*
4208 * Are we loading a blank journal or performing recovery after a
4209 * crash? For recovery, we need to check in advance whether we
4210 * can get read-write access to the device.
4211 */
617ba13b 4212 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4213 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4214 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4215 "required on readonly filesystem");
ac27a0ec 4216 if (really_read_only) {
b31e1552
ES		 4217 ext4_msg(sb, KERN_ERR, "write access "
4218 "unavailable, cannot proceed");
ac27a0ec
DK		 4219 return -EROFS;
4220 }
b31e1552
ES		 4221 ext4_msg(sb, KERN_INFO, "write access will "
4222 "be enabled during recovery");
ac27a0ec
DK		 4223 }
4224 }
4225
4226 if (journal_inum && journal_dev) {
b31e1552
ES		 4227 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4228 "and inode journals!");
ac27a0ec
DK		 4229 return -EINVAL;
4230 }
4231
4232 if (journal_inum) {
617ba13b 4233 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4234 return -EINVAL;
4235 } else {
617ba13b 4236 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4237 return -EINVAL;
4238 }
4239
90576c0b 4240 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4241 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4242
617ba13b 4243 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4244 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4245 if (!err) {
4246 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4247 if (save)
4248 memcpy(save, ((char *) es) +
4249 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4250 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4251 if (save)
4252 memcpy(((char *) es) + EXT4_S_ERR_START,
4253 save, EXT4_S_ERR_LEN);
4254 kfree(save);
4255 }
ac27a0ec
DK		 4256
4257 if (err) {
b31e1552 4258 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4259 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4260 return err;
4261 }
4262
617ba13b
MC		 4263 EXT4_SB(sb)->s_journal = journal;
4264 ext4_clear_journal_err(sb, es);
ac27a0ec 4265
c41303ce 4266 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4267 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4268 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4269
4270 /* Make sure we flush the recovery flag to disk. */
e2d67052 4271 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4272 }
4273
4274 return 0;
4275}
4276
e2d67052 4277static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4278{
e2d67052 4279 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4280 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4281 int error = 0;
ac27a0ec 4282
bdfe0cbd 4283 if (!sbh || block_device_ejected(sb))
c4be0c1d 4284 return error;
914258bf
TT		 4285 if (buffer_write_io_error(sbh)) {
4286 /*
4287 * Oh, dear. A previous attempt to write the
4288 * superblock failed. This could happen because the
4289 * USB device was yanked out. Or it could happen to
4290 * be a transient write error and maybe the block will
4291 * be remapped. Nothing we can do but to retry the
4292 * write and hope for the best.
4293 */
b31e1552
ES		 4294 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4295 "superblock detected");
914258bf
TT		 4296 clear_buffer_write_io_error(sbh);
4297 set_buffer_uptodate(sbh);
4298 }
71290b36
TT		 4299 /*
4300 * If the file system is mounted read-only, don't update the
4301 * superblock write time. This avoids updating the superblock
4302 * write time when we are mounting the root file system
4303 * read/only but we need to replay the journal; at that point,
4304 * for people who are east of GMT and who make their clock
4305 * tick in localtime for Windows bug-for-bug compatibility,
4306 * the clock is set in the future, and this will cause e2fsck
4307 * to complain and force a full file system check.
4308 */
4309 if (!(sb->s_flags & MS_RDONLY))
4310 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4311 if (sb->s_bdev->bd_part)
4312 es->s_kbytes_written =
4313 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4314 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4315 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4316 else
4317 es->s_kbytes_written =
4318 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4319 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4320 ext4_free_blocks_count_set(es,
57042651
TT		 4321 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4322 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4323 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4324 es->s_free_inodes_count =
4325 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4326 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4327 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4328 ext4_superblock_csum_set(sb);
ac27a0ec 4329 mark_buffer_dirty(sbh);
914258bf 4330 if (sync) {
564bc402
DJ		 4331 error = __sync_dirty_buffer(sbh,
4332 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
c4be0c1d
TS		 4333 if (error)
4334 return error;
4335
4336 error = buffer_write_io_error(sbh);
4337 if (error) {
b31e1552
ES		 4338 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4339 "superblock");
914258bf
TT		 4340 clear_buffer_write_io_error(sbh);
4341 set_buffer_uptodate(sbh);
4342 }
4343 }
c4be0c1d 4344 return error;
ac27a0ec
DK		 4345}
4346
ac27a0ec
DK		 4347/*
4348 * Have we just finished recovery? If so, and if we are mounting (or
4349 * remounting) the filesystem readonly, then we will end up with a
4350 * consistent fs on disk. Record that fact.
4351 */
2b2d6d01
TT		 4352static void ext4_mark_recovery_complete(struct super_block *sb,
4353 struct ext4_super_block *es)
ac27a0ec 4354{
617ba13b 4355 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4356
0390131b
FM		 4357 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4358 BUG_ON(journal != NULL);
4359 return;
4360 }
dab291af 4361 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4362 if (jbd2_journal_flush(journal) < 0)
4363 goto out;
4364
617ba13b 4365 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4366 sb->s_flags & MS_RDONLY) {
617ba13b 4367 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4368 ext4_commit_super(sb, 1);
ac27a0ec 4369 }
7ffe1ea8
HK		 4370
4371out:
dab291af 4372 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4373}
4374
4375/*
4376 * If we are mounting (or read-write remounting) a filesystem whose journal
4377 * has recorded an error from a previous lifetime, move that error to the
4378 * main filesystem now.
4379 */
2b2d6d01
TT		 4380static void ext4_clear_journal_err(struct super_block *sb,
4381 struct ext4_super_block *es)
ac27a0ec
DK		 4382{
4383 journal_t *journal;
4384 int j_errno;
4385 const char *errstr;
4386
0390131b
FM		 4387 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4388
617ba13b 4389 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4390
4391 /*
4392 * Now check for any error status which may have been recorded in the
617ba13b 4393 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4394 */
4395
dab291af 4396 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4397 if (j_errno) {
4398 char nbuf[16];
4399
617ba13b 4400 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4401 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4402 "from previous mount: %s", errstr);
12062ddd 4403 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4404
617ba13b
MC		 4405 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4406 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4407 ext4_commit_super(sb, 1);
ac27a0ec 4408
dab291af 4409 jbd2_journal_clear_err(journal);
d796c52e 4410 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4411 }
4412}
4413
4414/*
4415 * Force the running and committing transactions to commit,
4416 * and wait on the commit.
4417 */
617ba13b 4418int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4419{
4420 journal_t *journal;
ac27a0ec
DK		 4421
4422 if (sb->s_flags & MS_RDONLY)
4423 return 0;
4424
617ba13b 4425 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4426 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4427}
4428
617ba13b 4429static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4430{
14ce0cb4 4431 int ret = 0;
9eddacf9 4432 tid_t target;
06a407f1 4433 bool needs_barrier = false;
8d5d02e6 4434 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4435
9bffad1e 4436 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4437 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4438 /*
4439 * Writeback quota in non-journalled quota case - journalled quota has
4440 * no dirty dquots
4441 */
4442 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4443 /*
4444 * Data writeback is possible w/o journal transaction, so barrier must
4445 * being sent at the end of the function. But we can skip it if
4446 * transaction_commit will do it for us.
4447 */
bda32530
TT		 4448 if (sbi->s_journal) {
4449 target = jbd2_get_latest_transaction(sbi->s_journal);
4450 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4451 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4452 needs_barrier = true;
4453
4454 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4455 if (wait)
4456 ret = jbd2_log_wait_commit(sbi->s_journal,
4457 target);
4458 }
4459 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4460 needs_barrier = true;
06a407f1
DM		 4461 if (needs_barrier) {
4462 int err;
4463 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4464 if (!ret)
4465 ret = err;
0390131b 4466 }
06a407f1
DM		 4467
4468 return ret;
4469}
4470
ac27a0ec
DK		 4471/*
4472 * LVM calls this function before a (read-only) snapshot is created. This
4473 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4474 *
4475 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4476 * state independently. It relies on upper layer to stop all data & metadata
4477 * modifications.
ac27a0ec 4478 */
c4be0c1d 4479static int ext4_freeze(struct super_block *sb)
ac27a0ec 4480{
c4be0c1d
TS		 4481 int error = 0;
4482 journal_t *journal;
ac27a0ec 4483
9ca92389
TT		 4484 if (sb->s_flags & MS_RDONLY)
4485 return 0;
ac27a0ec 4486
9ca92389 4487 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4488
bb044576
TT		 4489 if (journal) {
4490 /* Now we set up the journal barrier. */
4491 jbd2_journal_lock_updates(journal);
ac27a0ec 4492
bb044576
TT		 4493 /*
4494 * Don't clear the needs_recovery flag if we failed to
4495 * flush the journal.
4496 */
4497 error = jbd2_journal_flush(journal);
4498 if (error < 0)
4499 goto out;
c642dc9e
ES		 4500
4501 /* Journal blocked and flushed, clear needs_recovery flag. */
4502 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
bb044576 4503 }
9ca92389 4504
9ca92389 4505 error = ext4_commit_super(sb, 1);
6b0310fb 4506out:
bb044576
TT		 4507 if (journal)
4508 /* we rely on upper layer to stop further updates */
4509 jbd2_journal_unlock_updates(journal);
6b0310fb 4510 return error;
ac27a0ec
DK		 4511}
4512
4513/*
4514 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4515 * flag here, even though the filesystem is not technically dirty yet.
4516 */
c4be0c1d 4517static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4518{
9ca92389
TT		 4519 if (sb->s_flags & MS_RDONLY)
4520 return 0;
4521
c642dc9e
ES		 4522 if (EXT4_SB(sb)->s_journal) {
4523 /* Reset the needs_recovery flag before the fs is unlocked. */
4524 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4525 }
4526
9ca92389 4527 ext4_commit_super(sb, 1);
c4be0c1d 4528 return 0;
ac27a0ec
DK		 4529}
4530
673c6100
TT		 4531/*
4532 * Structure to save mount options for ext4_remount's benefit
4533 */
4534struct ext4_mount_options {
4535 unsigned long s_mount_opt;
a2595b8a 4536 unsigned long s_mount_opt2;
08cefc7a
EB		 4537 kuid_t s_resuid;
4538 kgid_t s_resgid;
673c6100
TT		 4539 unsigned long s_commit_interval;
4540 u32 s_min_batch_time, s_max_batch_time;
4541#ifdef CONFIG_QUOTA
4542 int s_jquota_fmt;
a2d4a646 4543 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4544#endif
4545};
4546
2b2d6d01 4547static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4548{
2b2d6d01 4549 struct ext4_super_block *es;
617ba13b 4550 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4551 unsigned long old_sb_flags;
617ba13b 4552 struct ext4_mount_options old_opts;
c79d967d 4553 int enable_quota = 0;
8a266467 4554 ext4_group_t g;
b3881f74 4555 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4556 int err = 0;
ac27a0ec 4557#ifdef CONFIG_QUOTA
03dafb5f 4558 int i, j;
ac27a0ec 4559#endif
d4c402d9 4560 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4561
4562 /* Store the original options */
4563 old_sb_flags = sb->s_flags;
4564 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4565 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4566 old_opts.s_resuid = sbi->s_resuid;
4567 old_opts.s_resgid = sbi->s_resgid;
4568 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4569 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4570 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4571#ifdef CONFIG_QUOTA
4572 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4573 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4574 if (sbi->s_qf_names[i]) {
4575 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4576 GFP_KERNEL);
4577 if (!old_opts.s_qf_names[i]) {
4578 for (j = 0; j < i; j++)
4579 kfree(old_opts.s_qf_names[j]);
3e36a163 4580 kfree(orig_data);
03dafb5f
CG		 4581 return -ENOMEM;
4582 }
4583 } else
4584 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4585#endif
b3881f74
TT		 4586 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4587 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4588
661aa520 4589 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4590 err = -EINVAL;
4591 goto restore_opts;
4592 }
4593
6b992ff2 4594 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4595 test_opt(sb, JOURNAL_CHECKSUM)) {
4596 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4597 "during remount not supported; ignoring");
4598 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4599 }
4600
6ae6514b
PS		 4601 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4602 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4603 ext4_msg(sb, KERN_ERR, "can't mount with "
4604 "both data=journal and delalloc");
4605 err = -EINVAL;
4606 goto restore_opts;
4607 }
4608 if (test_opt(sb, DIOREAD_NOLOCK)) {
4609 ext4_msg(sb, KERN_ERR, "can't mount with "
4610 "both data=journal and dioread_nolock");
4611 err = -EINVAL;
4612 goto restore_opts;
4613 }
923ae0ff
RZ		 4614 if (test_opt(sb, DAX)) {
4615 ext4_msg(sb, KERN_ERR, "can't mount with "
4616 "both data=journal and dax");
4617 err = -EINVAL;
4618 goto restore_opts;
4619 }
4620 }
4621
4622 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4623 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4624 "dax flag with busy inodes while remounting");
4625 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 4626 }
4627
4ab2f15b 4628 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4629 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4630
4631 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4632 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4633
4634 es = sbi->s_es;
4635
b3881f74 4636 if (sbi->s_journal) {
0390131b 4637 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4638 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4639 }
ac27a0ec 4640
a2fd66d0
TT		 4641 if (*flags & MS_LAZYTIME)
4642 sb->s_flags |= MS_LAZYTIME;
4643
661aa520 4644 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4645 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4646 err = -EROFS;
4647 goto restore_opts;
4648 }
4649
4650 if (*flags & MS_RDONLY) {
38c03b34
TT		 4651 err = sync_filesystem(sb);
4652 if (err < 0)
4653 goto restore_opts;
0f0dd62f
CH		 4654 err = dquot_suspend(sb, -1);
4655 if (err < 0)
c79d967d 4656 goto restore_opts;
c79d967d 4657
ac27a0ec
DK		 4658 /*
4659 * First of all, the unconditional stuff we have to do
4660 * to disable replay of the journal when we next remount
4661 */
4662 sb->s_flags |= MS_RDONLY;
4663
4664 /*
4665 * OK, test if we are remounting a valid rw partition
4666 * readonly, and if so set the rdonly flag and then
4667 * mark the partition as valid again.
4668 */
617ba13b
MC		 4669 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4670 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4671 es->s_state = cpu_to_le16(sbi->s_mount_state);
4672
a63c9eb2 4673 if (sbi->s_journal)
0390131b 4674 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4675 } else {
a13fb1a4 4676 /* Make sure we can mount this feature set readwrite */
2cb5cc8b
DW		 4677 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4678 EXT4_FEATURE_RO_COMPAT_READONLY) ||
4679 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4680 err = -EROFS;
4681 goto restore_opts;
4682 }
8a266467
TT		 4683 /*
4684 * Make sure the group descriptor checksums
0b8e58a1 4685 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4686 */
4687 for (g = 0; g < sbi->s_groups_count; g++) {
4688 struct ext4_group_desc *gdp =
4689 ext4_get_group_desc(sb, g, NULL);
4690
feb0ab32 4691 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4692 ext4_msg(sb, KERN_ERR,
4693 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4694 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4695 le16_to_cpu(gdp->bg_checksum));
4696 err = -EINVAL;
4697 goto restore_opts;
4698 }
4699 }
4700
ead6596b
ES		 4701 /*
4702 * If we have an unprocessed orphan list hanging
4703 * around from a previously readonly bdev mount,
4704 * require a full umount/remount for now.
4705 */
4706 if (es->s_last_orphan) {
b31e1552 4707 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4708 "remount RDWR because of unprocessed "
4709 "orphan inode list. Please "
b31e1552 4710 "umount/remount instead");
ead6596b
ES		 4711 err = -EINVAL;
4712 goto restore_opts;
4713 }
4714
ac27a0ec
DK		 4715 /*
4716 * Mounting a RDONLY partition read-write, so reread
4717 * and store the current valid flag. (It may have
4718 * been changed by e2fsck since we originally mounted
4719 * the partition.)
4720 */
0390131b
FM		 4721 if (sbi->s_journal)
4722 ext4_clear_journal_err(sb, es);
ac27a0ec 4723 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4724 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4725 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4726 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4727 EXT4_FEATURE_INCOMPAT_MMP))
4728 if (ext4_multi_mount_protect(sb,
4729 le64_to_cpu(es->s_mmp_block))) {
4730 err = -EROFS;
4731 goto restore_opts;
4732 }
c79d967d 4733 enable_quota = 1;
ac27a0ec
DK		 4734 }
4735 }
bfff6873
LC		 4736
4737 /*
4738 * Reinitialize lazy itable initialization thread based on
4739 * current settings
4740 */
4741 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4742 ext4_unregister_li_request(sb);
4743 else {
4744 ext4_group_t first_not_zeroed;
4745 first_not_zeroed = ext4_has_uninit_itable(sb);
4746 ext4_register_li_request(sb, first_not_zeroed);
4747 }
4748
6fd058f7 4749 ext4_setup_system_zone(sb);
d096ad0f 4750 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4751 ext4_commit_super(sb, 1);
0390131b 4752
ac27a0ec
DK		 4753#ifdef CONFIG_QUOTA
4754 /* Release old quota file names */
a2d4a646 4755 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 4756 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4757 if (enable_quota) {
4758 if (sb_any_quota_suspended(sb))
4759 dquot_resume(sb, -1);
4760 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4761 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4762 err = ext4_enable_quotas(sb);
07724f98 4763 if (err)
7c319d32 4764 goto restore_opts;
7c319d32
AK		 4765 }
4766 }
ac27a0ec 4767#endif
d4c402d9 4768
a26f4992 4769 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 4770 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4771 kfree(orig_data);
ac27a0ec 4772 return 0;
0b8e58a1 4773
ac27a0ec
DK		 4774restore_opts:
4775 sb->s_flags = old_sb_flags;
4776 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4777 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4778 sbi->s_resuid = old_opts.s_resuid;
4779 sbi->s_resgid = old_opts.s_resgid;
4780 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4781 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4782 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4783#ifdef CONFIG_QUOTA
4784 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 4785 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 4786 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 4787 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4788 }
4789#endif
d4c402d9 4790 kfree(orig_data);
ac27a0ec
DK		 4791 return err;
4792}
4793
2b2d6d01 4794static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4795{
4796 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4797 struct ext4_sb_info *sbi = EXT4_SB(sb);
4798 struct ext4_super_block *es = sbi->s_es;
27dd4385 4799 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 4800 u64 fsid;
d02a9391 4801 s64 bfree;
27dd4385 4802 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 4803
952fc18e
TT		 4804 if (!test_opt(sb, MINIX_DF))
4805 overhead = sbi->s_overhead;
ac27a0ec 4806
617ba13b 4807 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4808 buf->f_bsize = sb->s_blocksize;
b72f78cb 4809 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4810 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4811 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4812 /* prevent underflow in case that few free space is available */
57042651 4813 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 4814 buf->f_bavail = buf->f_bfree -
4815 (ext4_r_blocks_count(es) + resv_blocks);
4816 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 4817 buf->f_bavail = 0;
4818 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4819 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4820 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4821 fsid = le64_to_cpup((void *)es->s_uuid) ^
4822 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4823 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4824 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4825
ac27a0ec
DK		 4826 return 0;
4827}
4828
0b8e58a1
AD		 4829/* Helper function for writing quotas on sync - we need to start transaction
4830 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4831 * Process 1 Process 2
617ba13b 4832 * ext4_create() quota_sync()
a269eb18 4833 * jbd2_journal_start() write_dquot()
871a2931 4834 * dquot_initialize() down(dqio_mutex)
dab291af 4835 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4836 *
4837 */
4838
4839#ifdef CONFIG_QUOTA
4840
4841static inline struct inode *dquot_to_inode(struct dquot *dquot)
4842{
4c376dca 4843 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4844}
4845
617ba13b 4846static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4847{
4848 int ret, err;
4849 handle_t *handle;
4850 struct inode *inode;
4851
4852 inode = dquot_to_inode(dquot);
9924a92a 4853 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 4854 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4855 if (IS_ERR(handle))
4856 return PTR_ERR(handle);
4857 ret = dquot_commit(dquot);
617ba13b 4858 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4859 if (!ret)
4860 ret = err;
4861 return ret;
4862}
4863
617ba13b 4864static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4865{
4866 int ret, err;
4867 handle_t *handle;
4868
9924a92a 4869 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4870 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4871 if (IS_ERR(handle))
4872 return PTR_ERR(handle);
4873 ret = dquot_acquire(dquot);
617ba13b 4874 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4875 if (!ret)
4876 ret = err;
4877 return ret;
4878}
4879
617ba13b 4880static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4881{
4882 int ret, err;
4883 handle_t *handle;
4884
9924a92a 4885 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4886 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4887 if (IS_ERR(handle)) {
4888 /* Release dquot anyway to avoid endless cycle in dqput() */
4889 dquot_release(dquot);
ac27a0ec 4890 return PTR_ERR(handle);
9c3013e9 4891 }
ac27a0ec 4892 ret = dquot_release(dquot);
617ba13b 4893 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4894 if (!ret)
4895 ret = err;
4896 return ret;
4897}
4898
617ba13b 4899static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4900{
262b4662
JK		 4901 struct super_block *sb = dquot->dq_sb;
4902 struct ext4_sb_info *sbi = EXT4_SB(sb);
4903
2c8be6b2 4904 /* Are we journaling quotas? */
262b4662
JK		 4905 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
4906 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 4907 dquot_mark_dquot_dirty(dquot);
617ba13b 4908 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4909 } else {
4910 return dquot_mark_dquot_dirty(dquot);
4911 }
4912}
4913
617ba13b 4914static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4915{
4916 int ret, err;
4917 handle_t *handle;
4918
4919 /* Data block + inode block */
2b0143b5 4920 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 4921 if (IS_ERR(handle))
4922 return PTR_ERR(handle);
4923 ret = dquot_commit_info(sb, type);
617ba13b 4924 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4925 if (!ret)
4926 ret = err;
4927 return ret;
4928}
4929
4930/*
4931 * Turn on quotas during mount time - we need to find
4932 * the quota file and such...
4933 */
617ba13b 4934static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4935{
287a8095
CH		 4936 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4937 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4938}
4939
4940/*
4941 * Standard function to be called on quota_on
4942 */
617ba13b 4943static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4944 struct path *path)
ac27a0ec
DK		 4945{
4946 int err;
ac27a0ec
DK		 4947
4948 if (!test_opt(sb, QUOTA))
4949 return -EINVAL;
0623543b 4950
ac27a0ec 4951 /* Quotafile not on the same filesystem? */
d8c9584e 4952 if (path->dentry->d_sb != sb)
ac27a0ec 4953 return -EXDEV;
0623543b
JK		 4954 /* Journaling quota? */
4955 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4956 /* Quotafile not in fs root? */
f00c9e44 4957 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4958 ext4_msg(sb, KERN_WARNING,
4959 "Quota file not on filesystem root. "
4960 "Journaled quota will not work");
2b2d6d01 4961 }
0623543b
JK		 4962
4963 /*
4964 * When we journal data on quota file, we have to flush journal to see
4965 * all updates to the file when we bypass pagecache...
4966 */
0390131b 4967 if (EXT4_SB(sb)->s_journal &&
2b0143b5 4968 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 4969 /*
4970 * We don't need to lock updates but journal_flush() could
4971 * otherwise be livelocked...
4972 */
4973 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4974 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4975 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4976 if (err)
7ffe1ea8 4977 return err;
0623543b
JK		 4978 }
4979
f00c9e44 4980 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4981}
4982
7c319d32
AK		 4983static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4984 unsigned int flags)
4985{
4986 int err;
4987 struct inode *qf_inode;
a2d4a646 4988 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 4989 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4990 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4991 };
4992
4993 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4994
4995 if (!qf_inums[type])
4996 return -EPERM;
4997
4998 qf_inode = ext4_iget(sb, qf_inums[type]);
4999 if (IS_ERR(qf_inode)) {
5000 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5001 return PTR_ERR(qf_inode);
5002 }
5003
bcb13850
JK		 5004 /* Don't account quota for quota files to avoid recursion */
5005 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5006 err = dquot_enable(qf_inode, type, format_id, flags);
5007 iput(qf_inode);
5008
5009 return err;
5010}
5011
5012/* Enable usage tracking for all quota types. */
5013static int ext4_enable_quotas(struct super_block *sb)
5014{
5015 int type, err = 0;
a2d4a646 5016 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5017 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5018 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5019 };
5020
5021 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5022 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5023 if (qf_inums[type]) {
5024 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5025 DQUOT_USAGE_ENABLED);
5026 if (err) {
5027 ext4_warning(sb,
72ba7450
TT		 5028 "Failed to enable quota tracking "
5029 "(type=%d, err=%d). Please run "
5030 "e2fsck to fix.", type, err);
7c319d32
AK		 5031 return err;
5032 }
5033 }
5034 }
5035 return 0;
5036}
5037
ca0e05e4
DM		 5038static int ext4_quota_off(struct super_block *sb, int type)
5039{
21f97697
JK		 5040 struct inode *inode = sb_dqopt(sb)->files[type];
5041 handle_t *handle;
5042
87009d86
DM		 5043 /* Force all delayed allocation blocks to be allocated.
5044 * Caller already holds s_umount sem */
5045 if (test_opt(sb, DELALLOC))
ca0e05e4 5046 sync_filesystem(sb);
ca0e05e4 5047
0b268590
AG		 5048 if (!inode)
5049 goto out;
5050
21f97697
JK		 5051 /* Update modification times of quota files when userspace can
5052 * start looking at them */
9924a92a 5053 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5054 if (IS_ERR(handle))
5055 goto out;
5056 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5057 ext4_mark_inode_dirty(handle, inode);
5058 ext4_journal_stop(handle);
5059
5060out:
ca0e05e4
DM		 5061 return dquot_quota_off(sb, type);
5062}
5063
ac27a0ec
DK		 5064/* Read data from quotafile - avoid pagecache and such because we cannot afford
5065 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5066 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5067 * we don't have to be afraid of races */
617ba13b 5068static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5069 size_t len, loff_t off)
5070{
5071 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5072 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5073 int offset = off & (sb->s_blocksize - 1);
5074 int tocopy;
5075 size_t toread;
5076 struct buffer_head *bh;
5077 loff_t i_size = i_size_read(inode);
5078
5079 if (off > i_size)
5080 return 0;
5081 if (off+len > i_size)
5082 len = i_size-off;
5083 toread = len;
5084 while (toread > 0) {
5085 tocopy = sb->s_blocksize - offset < toread ?
5086 sb->s_blocksize - offset : toread;
1c215028
TT		 5087 bh = ext4_bread(NULL, inode, blk, 0);
5088 if (IS_ERR(bh))
5089 return PTR_ERR(bh);
ac27a0ec
DK		 5090 if (!bh) /* A hole? */
5091 memset(data, 0, tocopy);
5092 else
5093 memcpy(data, bh->b_data+offset, tocopy);
5094 brelse(bh);
5095 offset = 0;
5096 toread -= tocopy;
5097 data += tocopy;
5098 blk++;
5099 }
5100 return len;
5101}
5102
5103/* Write to quotafile (we know the transaction is already started and has
5104 * enough credits) */
617ba13b 5105static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5106 const char *data, size_t len, loff_t off)
5107{
5108 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5109 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5110 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5111 int retries = 0;
ac27a0ec
DK		 5112 struct buffer_head *bh;
5113 handle_t *handle = journal_current_handle();
5114
0390131b 5115 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5116 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5117 " cancelled because transaction is not started",
9c3013e9
JK		 5118 (unsigned long long)off, (unsigned long long)len);
5119 return -EIO;
5120 }
67eeb568
DM		 5121 /*
5122 * Since we account only one data block in transaction credits,
5123 * then it is impossible to cross a block boundary.
5124 */
5125 if (sb->s_blocksize - offset < len) {
5126 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5127 " cancelled because not block aligned",
5128 (unsigned long long)off, (unsigned long long)len);
5129 return -EIO;
5130 }
5131
c5e298ae
TT		 5132 do {
5133 bh = ext4_bread(handle, inode, blk,
5134 EXT4_GET_BLOCKS_CREATE |
5135 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5136 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5137 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5138 if (IS_ERR(bh))
5139 return PTR_ERR(bh);
67eeb568
DM		 5140 if (!bh)
5141 goto out;
5d601255 5142 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5143 err = ext4_journal_get_write_access(handle, bh);
5144 if (err) {
5145 brelse(bh);
1c215028 5146 return err;
ac27a0ec 5147 }
67eeb568
DM		 5148 lock_buffer(bh);
5149 memcpy(bh->b_data+offset, data, len);
5150 flush_dcache_page(bh->b_page);
5151 unlock_buffer(bh);
62d2b5f2 5152 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5153 brelse(bh);
ac27a0ec 5154out:
67eeb568
DM		 5155 if (inode->i_size < off + len) {
5156 i_size_write(inode, off + len);
617ba13b 5157 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5158 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5159 }
67eeb568 5160 return len;
ac27a0ec
DK		 5161}
5162
5163#endif
5164
152a0836
AV		 5165static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5166 const char *dev_name, void *data)
ac27a0ec 5167{
152a0836 5168 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5169}
5170
c290ea01 5171#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5172static inline void register_as_ext2(void)
5173{
5174 int err = register_filesystem(&ext2_fs_type);
5175 if (err)
5176 printk(KERN_WARNING
5177 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5178}
5179
5180static inline void unregister_as_ext2(void)
5181{
5182 unregister_filesystem(&ext2_fs_type);
5183}
2035e776
TT		 5184
5185static inline int ext2_feature_set_ok(struct super_block *sb)
5186{
5187 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5188 return 0;
5189 if (sb->s_flags & MS_RDONLY)
5190 return 1;
5191 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5192 return 0;
5193 return 1;
5194}
24b58424
TT		 5195#else
5196static inline void register_as_ext2(void) { }
5197static inline void unregister_as_ext2(void) { }
2035e776 5198static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5199#endif
5200
24b58424
TT		 5201static inline void register_as_ext3(void)
5202{
5203 int err = register_filesystem(&ext3_fs_type);
5204 if (err)
5205 printk(KERN_WARNING
5206 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5207}
5208
5209static inline void unregister_as_ext3(void)
5210{
5211 unregister_filesystem(&ext3_fs_type);
5212}
2035e776
TT		 5213
5214static inline int ext3_feature_set_ok(struct super_block *sb)
5215{
5216 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5217 return 0;
5218 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5219 return 0;
5220 if (sb->s_flags & MS_RDONLY)
5221 return 1;
5222 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5223 return 0;
5224 return 1;
5225}
24b58424 5226
03010a33
TT		 5227static struct file_system_type ext4_fs_type = {
5228 .owner = THIS_MODULE,
5229 .name = "ext4",
152a0836 5230 .mount = ext4_mount,
03010a33
TT		 5231 .kill_sb = kill_block_super,
5232 .fs_flags = FS_REQUIRES_DEV,
5233};
7f78e035 5234MODULE_ALIAS_FS("ext4");
03010a33 5235
e9e3bcec
ES		 5236/* Shared across all ext4 file systems */
5237wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5238struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5239
5dabfc78 5240static int __init ext4_init_fs(void)
ac27a0ec 5241{
e9e3bcec 5242 int i, err;
c9de560d 5243
e294a537 5244 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5245 ext4_li_info = NULL;
5246 mutex_init(&ext4_li_mtx);
5247
9a4c8019 5248 /* Build-time check for flags consistency */
12e9b892 5249 ext4_check_flag_values();
e9e3bcec
ES		 5250
5251 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5252 mutex_init(&ext4__aio_mutex[i]);
5253 init_waitqueue_head(&ext4__ioend_wq[i]);
5254 }
5255
51865fda 5256 err = ext4_init_es();
6fd058f7
TT		 5257 if (err)
5258 return err;
51865fda
ZL		 5259
5260 err = ext4_init_pageio();
5261 if (err)
b5799018 5262 goto out5;
51865fda 5263
5dabfc78 5264 err = ext4_init_system_zone();
bd2d0210 5265 if (err)
b5799018 5266 goto out4;
d44651d0 5267 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889 5268
b5799018 5269 err = ext4_init_sysfs();
dd68314c 5270 if (err)
b5799018 5271 goto out3;
857ac889 5272
5dabfc78 5273 err = ext4_init_mballoc();
c9de560d
AT		 5274 if (err)
5275 goto out2;
9c191f70
M		 5276 else
5277 ext4_mballoc_ready = 1;
ac27a0ec
DK		 5278 err = init_inodecache();
5279 if (err)
5280 goto out1;
24b58424 5281 register_as_ext3();
2035e776 5282 register_as_ext2();
03010a33 5283 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5284 if (err)
5285 goto out;
bfff6873 5286
ac27a0ec
DK		 5287 return 0;
5288out:
24b58424
TT		 5289 unregister_as_ext2();
5290 unregister_as_ext3();
ac27a0ec
DK		 5291 destroy_inodecache();
5292out1:
9c191f70 5293 ext4_mballoc_ready = 0;
5dabfc78 5294 ext4_exit_mballoc();
9c191f70 5295out2:
b5799018
TT		 5296 ext4_exit_sysfs();
5297out3:
d44651d0
FJ		 5298 if (ext4_proc_root)
5299 remove_proc_entry("fs/ext4", NULL);
5dabfc78 5300 ext4_exit_system_zone();
b5799018 5301out4:
5dabfc78 5302 ext4_exit_pageio();
b5799018 5303out5:
51865fda
ZL		 5304 ext4_exit_es();
5305
ac27a0ec
DK		 5306 return err;
5307}
5308
5dabfc78 5309static void __exit ext4_exit_fs(void)
ac27a0ec 5310{
e298e73b 5311 ext4_exit_crypto();
bfff6873 5312 ext4_destroy_lazyinit_thread();
24b58424
TT		 5313 unregister_as_ext2();
5314 unregister_as_ext3();
03010a33 5315 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5316 destroy_inodecache();
5dabfc78 5317 ext4_exit_mballoc();
b5799018 5318 ext4_exit_sysfs();
9f6200bb 5319 remove_proc_entry("fs/ext4", NULL);
5dabfc78
TT		 5320 ext4_exit_system_zone();
5321 ext4_exit_pageio();
dd12ed14 5322 ext4_exit_es();
ac27a0ec
DK		 5323}
5324
5325MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5326MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5327MODULE_LICENSE("GPL");
5dabfc78
TT		 5328module_init(ext4_init_fs)
5329module_exit(ext4_exit_fs)