]> Git Repo - linux.git/blame - fs/ext4/super.c
projects / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: call out CRC and corruption errors with specific error codes
[linux.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card ([email protected])
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 ([email protected]), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
66114cad 27#include <linux/backing-dev.h>
ac27a0ec 28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
3197ebdb 37#include <linux/ctype.h>
1330593e 38#include <linux/log2.h>
717d50e4 39#include <linux/crc16.h>
7abc52c2 40#include <linux/cleancache.h>
ac27a0ec
DK		 41#include <asm/uaccess.h>
42
bfff6873
LC		 43#include <linux/kthread.h>
44#include <linux/freezer.h>
45
3dcf5451 46#include "ext4.h"
4a092d73 47#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 48#include "ext4_jbd2.h"
ac27a0ec
DK		 49#include "xattr.h"
50#include "acl.h"
3661d286 51#include "mballoc.h"
ac27a0ec 52
9bffad1e
TT		 53#define CREATE_TRACE_POINTS
54#include <trace/events/ext4.h>
55
0b75a840
LC		 56static struct ext4_lazy_init *ext4_li_info;
57static struct mutex ext4_li_mtx;
9c191f70 58static int ext4_mballoc_ready;
e294a537 59static struct ratelimit_state ext4_mount_msg_ratelimit;
9f6200bb 60
617ba13b 61static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 62 unsigned long journal_devnum);
2adf6da8 63static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 64static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 65static void ext4_mark_recovery_complete(struct super_block *sb,
66 struct ext4_super_block *es);
67static void ext4_clear_journal_err(struct super_block *sb,
68 struct ext4_super_block *es);
617ba13b 69static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 70static int ext4_remount(struct super_block *sb, int *flags, char *data);
71static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 72static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 73static int ext4_freeze(struct super_block *sb);
152a0836
AV		 74static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
75 const char *dev_name, void *data);
2035e776
TT		 76static inline int ext2_feature_set_ok(struct super_block *sb);
77static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 78static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 79static void ext4_destroy_lazyinit_thread(void);
80static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 81static void ext4_clear_request_list(void);
ac27a0ec 82
c290ea01 83#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
2035e776
TT		 84static struct file_system_type ext2_fs_type = {
85 .owner = THIS_MODULE,
86 .name = "ext2",
87 .mount = ext4_mount,
88 .kill_sb = kill_block_super,
89 .fs_flags = FS_REQUIRES_DEV,
90};
7f78e035 91MODULE_ALIAS_FS("ext2");
fa7614dd 92MODULE_ALIAS("ext2");
2035e776
TT		 93#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
94#else
95#define IS_EXT2_SB(sb) (0)
96#endif
97
98
ba69f9ab
JK		 99static struct file_system_type ext3_fs_type = {
100 .owner = THIS_MODULE,
101 .name = "ext3",
152a0836 102 .mount = ext4_mount,
ba69f9ab
JK		 103 .kill_sb = kill_block_super,
104 .fs_flags = FS_REQUIRES_DEV,
105};
7f78e035 106MODULE_ALIAS_FS("ext3");
fa7614dd 107MODULE_ALIAS("ext3");
ba69f9ab 108#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
bd81d8ee 109
d25425f8
DW		 110static int ext4_verify_csum_type(struct super_block *sb,
111 struct ext4_super_block *es)
112{
113 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
114 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
115 return 1;
116
117 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
118}
119
a9c47317
DW		 120static __le32 ext4_superblock_csum(struct super_block *sb,
121 struct ext4_super_block *es)
122{
123 struct ext4_sb_info *sbi = EXT4_SB(sb);
124 int offset = offsetof(struct ext4_super_block, s_checksum);
125 __u32 csum;
126
127 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
128
129 return cpu_to_le32(csum);
130}
131
c197855e
SH		 132static int ext4_superblock_csum_verify(struct super_block *sb,
133 struct ext4_super_block *es)
a9c47317 134{
9aa5d32b 135 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 136 return 1;
137
138 return es->s_checksum == ext4_superblock_csum(sb, es);
139}
140
06db49e6 141void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 142{
06db49e6
TT		 143 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
144
9aa5d32b 145 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 146 return;
147
148 es->s_checksum = ext4_superblock_csum(sb, es);
149}
150
9933fc0a
TT		 151void *ext4_kvmalloc(size_t size, gfp_t flags)
152{
153 void *ret;
154
8be04b93 155 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 156 if (!ret)
157 ret = __vmalloc(size, flags, PAGE_KERNEL);
158 return ret;
159}
160
161void *ext4_kvzalloc(size_t size, gfp_t flags)
162{
163 void *ret;
164
8be04b93 165 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 166 if (!ret)
167 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
168 return ret;
169}
170
8fadc143
AR		 171ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
172 struct ext4_group_desc *bg)
bd81d8ee 173{
3a14589c 174 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 175 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 176 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 177}
178
8fadc143
AR		 179ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
180 struct ext4_group_desc *bg)
bd81d8ee 181{
5272f837 182 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 183 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 184 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 185}
186
8fadc143
AR		 187ext4_fsblk_t ext4_inode_table(struct super_block *sb,
188 struct ext4_group_desc *bg)
bd81d8ee 189{
5272f837 190 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 191 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 192 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 193}
194
021b65bb
TT		 195__u32 ext4_free_group_clusters(struct super_block *sb,
196 struct ext4_group_desc *bg)
560671a0
AK		 197{
198 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
199 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 200 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 201}
202
203__u32 ext4_free_inodes_count(struct super_block *sb,
204 struct ext4_group_desc *bg)
205{
206 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 208 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 209}
210
211__u32 ext4_used_dirs_count(struct super_block *sb,
212 struct ext4_group_desc *bg)
213{
214 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 216 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 217}
218
219__u32 ext4_itable_unused_count(struct super_block *sb,
220 struct ext4_group_desc *bg)
221{
222 return le16_to_cpu(bg->bg_itable_unused_lo) |
223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 224 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 225}
226
8fadc143
AR		 227void ext4_block_bitmap_set(struct super_block *sb,
228 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 229{
3a14589c 230 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 231 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
232 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 233}
234
8fadc143
AR		 235void ext4_inode_bitmap_set(struct super_block *sb,
236 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 237{
5272f837 238 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 239 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
240 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 241}
242
8fadc143
AR		 243void ext4_inode_table_set(struct super_block *sb,
244 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 245{
5272f837 246 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 247 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
248 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 249}
250
021b65bb
TT		 251void ext4_free_group_clusters_set(struct super_block *sb,
252 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 253{
254 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
255 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
256 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
257}
258
259void ext4_free_inodes_set(struct super_block *sb,
260 struct ext4_group_desc *bg, __u32 count)
261{
262 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
265}
266
267void ext4_used_dirs_set(struct super_block *sb,
268 struct ext4_group_desc *bg, __u32 count)
269{
270 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
273}
274
275void ext4_itable_unused_set(struct super_block *sb,
276 struct ext4_group_desc *bg, __u32 count)
277{
278 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
281}
282
d3d1faf6 283
1c13d5c0
TT		 284static void __save_error_info(struct super_block *sb, const char *func,
285 unsigned int line)
286{
287 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
288
289 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1b46617b
TT		 290 if (bdev_read_only(sb->s_bdev))
291 return;
1c13d5c0
TT		 292 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
293 es->s_last_error_time = cpu_to_le32(get_seconds());
294 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
295 es->s_last_error_line = cpu_to_le32(line);
296 if (!es->s_first_error_time) {
297 es->s_first_error_time = es->s_last_error_time;
298 strncpy(es->s_first_error_func, func,
299 sizeof(es->s_first_error_func));
300 es->s_first_error_line = cpu_to_le32(line);
301 es->s_first_error_ino = es->s_last_error_ino;
302 es->s_first_error_block = es->s_last_error_block;
303 }
66e61a9e
TT		 304 /*
305 * Start the daily error reporting function if it hasn't been
306 * started already
307 */
308 if (!es->s_error_count)
309 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 310 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 311}
312
313static void save_error_info(struct super_block *sb, const char *func,
314 unsigned int line)
315{
316 __save_error_info(sb, func, line);
317 ext4_commit_super(sb, 1);
318}
319
bdfe0cbd
TT		 320/*
321 * The del_gendisk() function uninitializes the disk-specific data
322 * structures, including the bdi structure, without telling anyone
323 * else. Once this happens, any attempt to call mark_buffer_dirty()
324 * (for example, by ext4_commit_super), will cause a kernel OOPS.
325 * This is a kludge to prevent these oops until we can put in a proper
326 * hook in del_gendisk() to inform the VFS and file system layers.
327 */
328static int block_device_ejected(struct super_block *sb)
329{
330 struct inode *bd_inode = sb->s_bdev->bd_inode;
331 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
332
333 return bdi->dev == NULL;
334}
335
18aadd47
BJ		 336static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
337{
338 struct super_block *sb = journal->j_private;
339 struct ext4_sb_info *sbi = EXT4_SB(sb);
340 int error = is_journal_aborted(journal);
5d3ee208 341 struct ext4_journal_cb_entry *jce;
18aadd47 342
5d3ee208 343 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 344 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 345 while (!list_empty(&txn->t_private_list)) {
346 jce = list_entry(txn->t_private_list.next,
347 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 348 list_del_init(&jce->jce_list);
349 spin_unlock(&sbi->s_md_lock);
350 jce->jce_func(sb, jce, error);
351 spin_lock(&sbi->s_md_lock);
352 }
353 spin_unlock(&sbi->s_md_lock);
354}
1c13d5c0 355
ac27a0ec
DK		 356/* Deal with the reporting of failure conditions on a filesystem such as
357 * inconsistencies detected or read IO failures.
358 *
359 * On ext2, we can store the error state of the filesystem in the
617ba13b 360 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 361 * write ordering constraints on the superblock which prevent us from
362 * writing it out straight away; and given that the journal is about to
363 * be aborted, we can't rely on the current, or future, transactions to
364 * write out the superblock safely.
365 *
dab291af 366 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 367 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 368 * that error until we've noted it down and cleared it.
369 */
370
617ba13b 371static void ext4_handle_error(struct super_block *sb)
ac27a0ec 372{
ac27a0ec
DK		 373 if (sb->s_flags & MS_RDONLY)
374 return;
375
2b2d6d01 376 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 377 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 378
4ab2f15b 379 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 380 if (journal)
dab291af 381 jbd2_journal_abort(journal, -EIO);
ac27a0ec 382 }
2b2d6d01 383 if (test_opt(sb, ERRORS_RO)) {
b31e1552 384 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 385 /*
386 * Make sure updated value of ->s_mount_flags will be visible
387 * before ->s_flags update
388 */
389 smp_wmb();
ac27a0ec
DK		 390 sb->s_flags |= MS_RDONLY;
391 }
ac27a0ec 392 if (test_opt(sb, ERRORS_PANIC))
617ba13b 393 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 394 sb->s_id);
395}
396
efbed4dc
TT		 397#define ext4_error_ratelimit(sb) \
398 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
399 "EXT4-fs error")
400
12062ddd 401void __ext4_error(struct super_block *sb, const char *function,
c398eda0 402 unsigned int line, const char *fmt, ...)
ac27a0ec 403{
0ff2ea7d 404 struct va_format vaf;
ac27a0ec
DK		 405 va_list args;
406
efbed4dc
TT		 407 if (ext4_error_ratelimit(sb)) {
408 va_start(args, fmt);
409 vaf.fmt = fmt;
410 vaf.va = &args;
411 printk(KERN_CRIT
412 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
413 sb->s_id, function, line, current->comm, &vaf);
414 va_end(args);
415 }
f3fc0210 416 save_error_info(sb, function, line);
617ba13b 417 ext4_handle_error(sb);
ac27a0ec
DK		 418}
419
e7c96e8e
JP		 420void __ext4_error_inode(struct inode *inode, const char *function,
421 unsigned int line, ext4_fsblk_t block,
422 const char *fmt, ...)
273df556
FM		 423{
424 va_list args;
f7c21177 425 struct va_format vaf;
1c13d5c0 426 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 427
1c13d5c0
TT		 428 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
429 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 430 if (ext4_error_ratelimit(inode->i_sb)) {
431 va_start(args, fmt);
432 vaf.fmt = fmt;
433 vaf.va = &args;
434 if (block)
435 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
436 "inode #%lu: block %llu: comm %s: %pV\n",
437 inode->i_sb->s_id, function, line, inode->i_ino,
438 block, current->comm, &vaf);
439 else
440 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
441 "inode #%lu: comm %s: %pV\n",
442 inode->i_sb->s_id, function, line, inode->i_ino,
443 current->comm, &vaf);
444 va_end(args);
445 }
1c13d5c0 446 save_error_info(inode->i_sb, function, line);
273df556
FM		 447 ext4_handle_error(inode->i_sb);
448}
449
e7c96e8e
JP		 450void __ext4_error_file(struct file *file, const char *function,
451 unsigned int line, ext4_fsblk_t block,
452 const char *fmt, ...)
273df556
FM		 453{
454 va_list args;
f7c21177 455 struct va_format vaf;
1c13d5c0 456 struct ext4_super_block *es;
496ad9aa 457 struct inode *inode = file_inode(file);
273df556
FM		 458 char pathname[80], *path;
459
1c13d5c0
TT		 460 es = EXT4_SB(inode->i_sb)->s_es;
461 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 462 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 463 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 464 if (IS_ERR(path))
465 path = "(unknown)";
466 va_start(args, fmt);
467 vaf.fmt = fmt;
468 vaf.va = &args;
469 if (block)
470 printk(KERN_CRIT
471 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
472 "block %llu: comm %s: path %s: %pV\n",
473 inode->i_sb->s_id, function, line, inode->i_ino,
474 block, current->comm, path, &vaf);
475 else
476 printk(KERN_CRIT
477 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
478 "comm %s: path %s: %pV\n",
479 inode->i_sb->s_id, function, line, inode->i_ino,
480 current->comm, path, &vaf);
481 va_end(args);
482 }
1c13d5c0 483 save_error_info(inode->i_sb, function, line);
273df556
FM		 484 ext4_handle_error(inode->i_sb);
485}
486
722887dd
TT		 487const char *ext4_decode_error(struct super_block *sb, int errno,
488 char nbuf[16])
ac27a0ec
DK		 489{
490 char *errstr = NULL;
491
492 switch (errno) {
6a797d27
DW		 493 case -EFSCORRUPTED:
494 errstr = "Corrupt filesystem";
495 break;
496 case -EFSBADCRC:
497 errstr = "Filesystem failed CRC";
498 break;
ac27a0ec
DK		 499 case -EIO:
500 errstr = "IO failure";
501 break;
502 case -ENOMEM:
503 errstr = "Out of memory";
504 break;
505 case -EROFS:
78f1ddbb
TT		 506 if (!sb || (EXT4_SB(sb)->s_journal &&
507 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 508 errstr = "Journal has aborted";
509 else
510 errstr = "Readonly filesystem";
511 break;
512 default:
513 /* If the caller passed in an extra buffer for unknown
514 * errors, textualise them now. Else we just return
515 * NULL. */
516 if (nbuf) {
517 /* Check for truncated error codes... */
518 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
519 errstr = nbuf;
520 }
521 break;
522 }
523
524 return errstr;
525}
526
617ba13b 527/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 528 * automatically and invokes the appropriate error response. */
529
c398eda0
TT		 530void __ext4_std_error(struct super_block *sb, const char *function,
531 unsigned int line, int errno)
ac27a0ec
DK		 532{
533 char nbuf[16];
534 const char *errstr;
535
536 /* Special case: if the error is EROFS, and we're not already
537 * inside a transaction, then there's really no point in logging
538 * an error. */
539 if (errno == -EROFS && journal_current_handle() == NULL &&
540 (sb->s_flags & MS_RDONLY))
541 return;
542
efbed4dc
TT		 543 if (ext4_error_ratelimit(sb)) {
544 errstr = ext4_decode_error(sb, errno, nbuf);
545 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
546 sb->s_id, function, line, errstr);
547 }
ac27a0ec 548
efbed4dc 549 save_error_info(sb, function, line);
617ba13b 550 ext4_handle_error(sb);
ac27a0ec
DK		 551}
552
553/*
617ba13b 554 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 555 * abort function may be used to deal with unrecoverable failures such
556 * as journal IO errors or ENOMEM at a critical moment in log management.
557 *
558 * We unconditionally force the filesystem into an ABORT|READONLY state,
559 * unless the error response on the fs has been set to panic in which
560 * case we take the easy way out and panic immediately.
561 */
562
c67d859e 563void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 564 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 565{
566 va_list args;
567
1c13d5c0 568 save_error_info(sb, function, line);
ac27a0ec 569 va_start(args, fmt);
c398eda0
TT		 570 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
571 function, line);
ac27a0ec
DK		 572 vprintk(fmt, args);
573 printk("\n");
574 va_end(args);
575
1c13d5c0
TT		 576 if ((sb->s_flags & MS_RDONLY) == 0) {
577 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 578 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 579 /*
580 * Make sure updated value of ->s_mount_flags will be visible
581 * before ->s_flags update
582 */
583 smp_wmb();
584 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 585 if (EXT4_SB(sb)->s_journal)
586 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
587 save_error_info(sb, function, line);
588 }
ac27a0ec 589 if (test_opt(sb, ERRORS_PANIC))
617ba13b 590 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 591}
592
e7c96e8e
JP		 593void __ext4_msg(struct super_block *sb,
594 const char *prefix, const char *fmt, ...)
b31e1552 595{
0ff2ea7d 596 struct va_format vaf;
b31e1552
ES		 597 va_list args;
598
efbed4dc
TT		 599 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
600 return;
601
b31e1552 602 va_start(args, fmt);
0ff2ea7d
JP		 603 vaf.fmt = fmt;
604 vaf.va = &args;
605 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 606 va_end(args);
607}
608
b03a2f7e
AD		 609#define ext4_warning_ratelimit(sb) \
610 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
611 "EXT4-fs warning")
612
12062ddd 613void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 614 unsigned int line, const char *fmt, ...)
ac27a0ec 615{
0ff2ea7d 616 struct va_format vaf;
ac27a0ec
DK		 617 va_list args;
618
b03a2f7e 619 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 620 return;
621
ac27a0ec 622 va_start(args, fmt);
0ff2ea7d
JP		 623 vaf.fmt = fmt;
624 vaf.va = &args;
625 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
626 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 627 va_end(args);
628}
629
b03a2f7e
AD		 630void __ext4_warning_inode(const struct inode *inode, const char *function,
631 unsigned int line, const char *fmt, ...)
632{
633 struct va_format vaf;
634 va_list args;
635
636 if (!ext4_warning_ratelimit(inode->i_sb))
637 return;
638
639 va_start(args, fmt);
640 vaf.fmt = fmt;
641 vaf.va = &args;
642 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
643 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
644 function, line, inode->i_ino, current->comm, &vaf);
645 va_end(args);
646}
647
e29136f8
TT		 648void __ext4_grp_locked_error(const char *function, unsigned int line,
649 struct super_block *sb, ext4_group_t grp,
650 unsigned long ino, ext4_fsblk_t block,
651 const char *fmt, ...)
5d1b1b3f
AK		 652__releases(bitlock)
653__acquires(bitlock)
654{
0ff2ea7d 655 struct va_format vaf;
5d1b1b3f
AK		 656 va_list args;
657 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
658
1c13d5c0
TT		 659 es->s_last_error_ino = cpu_to_le32(ino);
660 es->s_last_error_block = cpu_to_le64(block);
661 __save_error_info(sb, function, line);
0ff2ea7d 662
efbed4dc
TT		 663 if (ext4_error_ratelimit(sb)) {
664 va_start(args, fmt);
665 vaf.fmt = fmt;
666 vaf.va = &args;
667 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
668 sb->s_id, function, line, grp);
669 if (ino)
670 printk(KERN_CONT "inode %lu: ", ino);
671 if (block)
672 printk(KERN_CONT "block %llu:",
673 (unsigned long long) block);
674 printk(KERN_CONT "%pV\n", &vaf);
675 va_end(args);
676 }
5d1b1b3f
AK		 677
678 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 679 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 680 return;
681 }
1c13d5c0 682
5d1b1b3f
AK		 683 ext4_unlock_group(sb, grp);
684 ext4_handle_error(sb);
685 /*
686 * We only get here in the ERRORS_RO case; relocking the group
687 * may be dangerous, but nothing bad will happen since the
688 * filesystem will have already been marked read/only and the
689 * journal has been aborted. We return 1 as a hint to callers
690 * who might what to use the return value from
25985edc 691 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 692 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
693 * aggressively from the ext4 function in question, with a
694 * more appropriate error code.
695 */
696 ext4_lock_group(sb, grp);
697 return;
698}
699
617ba13b 700void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 701{
617ba13b 702 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 703
617ba13b 704 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 705 return;
706
12062ddd 707 ext4_warning(sb,
ac27a0ec
DK		 708 "updating to rev %d because of new feature flag, "
709 "running e2fsck is recommended",
617ba13b 710 EXT4_DYNAMIC_REV);
ac27a0ec 711
617ba13b
MC		 712 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
713 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
714 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 715 /* leave es->s_feature_*compat flags alone */
716 /* es->s_uuid will be set by e2fsck if empty */
717
718 /*
719 * The rest of the superblock fields should be zero, and if not it
720 * means they are likely already in use, so leave them alone. We
721 * can leave it up to e2fsck to clean up any inconsistencies there.
722 */
723}
724
725/*
726 * Open the external journal device
727 */
b31e1552 728static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 729{
730 struct block_device *bdev;
731 char b[BDEVNAME_SIZE];
732
d4d77629 733 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 734 if (IS_ERR(bdev))
735 goto fail;
736 return bdev;
737
738fail:
b31e1552 739 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 740 __bdevname(dev, b), PTR_ERR(bdev));
741 return NULL;
742}
743
744/*
745 * Release the journal device
746 */
4385bab1 747static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 748{
4385bab1 749 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 750}
751
4385bab1 752static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 753{
754 struct block_device *bdev;
ac27a0ec
DK		 755 bdev = sbi->journal_bdev;
756 if (bdev) {
4385bab1 757 ext4_blkdev_put(bdev);
ac27a0ec
DK		 758 sbi->journal_bdev = NULL;
759 }
ac27a0ec
DK		 760}
761
762static inline struct inode *orphan_list_entry(struct list_head *l)
763{
617ba13b 764 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 765}
766
617ba13b 767static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 768{
769 struct list_head *l;
770
b31e1552
ES		 771 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
772 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 773
774 printk(KERN_ERR "sb_info orphan list:\n");
775 list_for_each(l, &sbi->s_orphan) {
776 struct inode *inode = orphan_list_entry(l);
777 printk(KERN_ERR " "
778 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
779 inode->i_sb->s_id, inode->i_ino, inode,
780 inode->i_mode, inode->i_nlink,
781 NEXT_ORPHAN(inode));
782 }
783}
784
2b2d6d01 785static void ext4_put_super(struct super_block *sb)
ac27a0ec 786{
617ba13b
MC		 787 struct ext4_sb_info *sbi = EXT4_SB(sb);
788 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 789 int i, err;
ac27a0ec 790
857ac889 791 ext4_unregister_li_request(sb);
e0ccfd95
CH		 792 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
793
2e8fa54e 794 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 795 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 796
0390131b
FM		 797 if (sbi->s_journal) {
798 err = jbd2_journal_destroy(sbi->s_journal);
799 sbi->s_journal = NULL;
800 if (err < 0)
c67d859e 801 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 802 }
d4edac31 803
ebd173be 804 ext4_unregister_sysfs(sb);
d3922a77 805 ext4_es_unregister_shrinker(sbi);
9105bb14 806 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 807 ext4_release_system_zone(sb);
808 ext4_mb_release(sb);
809 ext4_ext_release(sb);
810 ext4_xattr_put_super(sb);
811
ac27a0ec 812 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 813 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 814 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 815 }
58c5873a 816 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 817 ext4_commit_super(sb, 1);
818
ac27a0ec
DK		 819 for (i = 0; i < sbi->s_gdb_count; i++)
820 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 821 kvfree(sbi->s_group_desc);
822 kvfree(sbi->s_flex_groups);
57042651 823 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 824 percpu_counter_destroy(&sbi->s_freeinodes_counter);
825 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 826 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 827 brelse(sbi->s_sbh);
828#ifdef CONFIG_QUOTA
a2d4a646 829 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 830 kfree(sbi->s_qf_names[i]);
831#endif
832
833 /* Debugging code just in case the in-memory inode orphan list
834 * isn't empty. The on-disk one can be non-empty if we've
835 * detected an error and taken the fs readonly, but the
836 * in-memory list had better be clean by this point. */
837 if (!list_empty(&sbi->s_orphan))
838 dump_orphan_list(sb, sbi);
839 J_ASSERT(list_empty(&sbi->s_orphan));
840
89d96a6f 841 sync_blockdev(sb->s_bdev);
f98393a6 842 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 843 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
844 /*
845 * Invalidate the journal device's buffers. We don't want them
846 * floating about in memory - the physical journal device may
847 * hotswapped, and it breaks the `ro-after' testing code.
848 */
849 sync_blockdev(sbi->journal_bdev);
f98393a6 850 invalidate_bdev(sbi->journal_bdev);
617ba13b 851 ext4_blkdev_remove(sbi);
ac27a0ec 852 }
9c191f70
M		 853 if (sbi->s_mb_cache) {
854 ext4_xattr_destroy_cache(sbi->s_mb_cache);
855 sbi->s_mb_cache = NULL;
856 }
c5e06d10
JL		 857 if (sbi->s_mmp_tsk)
858 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 859 sb->s_fs_info = NULL;
3197ebdb
TT		 860 /*
861 * Now that we are completely done shutting down the
862 * superblock, we need to actually destroy the kobject.
863 */
3197ebdb
TT		 864 kobject_put(&sbi->s_kobj);
865 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 866 if (sbi->s_chksum_driver)
867 crypto_free_shash(sbi->s_chksum_driver);
705895b6 868 kfree(sbi->s_blockgroup_lock);
ac27a0ec 869 kfree(sbi);
ac27a0ec
DK		 870}
871
e18b890b 872static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 873
874/*
875 * Called inside transaction, so use GFP_NOFS
876 */
617ba13b 877static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 878{
617ba13b 879 struct ext4_inode_info *ei;
ac27a0ec 880
e6b4f8da 881 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 882 if (!ei)
883 return NULL;
0b8e58a1 884
ac27a0ec 885 ei->vfs_inode.i_version = 1;
202ee5df 886 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 887 INIT_LIST_HEAD(&ei->i_prealloc_list);
888 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 889 ext4_es_init_tree(&ei->i_es_tree);
890 rwlock_init(&ei->i_es_lock);
edaa53ca 891 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 892 ei->i_es_all_nr = 0;
edaa53ca 893 ei->i_es_shk_nr = 0;
dd475925 894 ei->i_es_shrink_lblk = 0;
d2a17637
MC		 895 ei->i_reserved_data_blocks = 0;
896 ei->i_reserved_meta_blocks = 0;
897 ei->i_allocated_meta_blocks = 0;
9d0be502 898 ei->i_da_metadata_calc_len = 0;
7e731bc9 899 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 900 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 901#ifdef CONFIG_QUOTA
902 ei->i_reserved_quota = 0;
96c7e0d9 903 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 904#endif
8aefcd55 905 ei->jinode = NULL;
2e8fa54e 906 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 907 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 908 ei->i_sync_tid = 0;
909 ei->i_datasync_tid = 0;
f7ad6d2e 910 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 911 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 912 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
b30ab0e0 913#ifdef CONFIG_EXT4_FS_ENCRYPTION
b7236e21 914 ei->i_crypt_info = NULL;
b30ab0e0 915#endif
ac27a0ec
DK		 916 return &ei->vfs_inode;
917}
918
7ff9c073
TT		 919static int ext4_drop_inode(struct inode *inode)
920{
921 int drop = generic_drop_inode(inode);
922
923 trace_ext4_drop_inode(inode, drop);
924 return drop;
925}
926
fa0d7e3d
NP		 927static void ext4_i_callback(struct rcu_head *head)
928{
929 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 930 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
931}
932
617ba13b 933static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 934{
9f7dd93d 935 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 936 ext4_msg(inode->i_sb, KERN_ERR,
937 "Inode %lu (%p): orphan list check failed!",
938 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 939 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
940 EXT4_I(inode), sizeof(struct ext4_inode_info),
941 true);
942 dump_stack();
943 }
fa0d7e3d 944 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 945}
946
51cc5068 947static void init_once(void *foo)
ac27a0ec 948{
617ba13b 949 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 950
a35afb83 951 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 952 init_rwsem(&ei->xattr_sem);
0e855ac8 953 init_rwsem(&ei->i_data_sem);
a35afb83 954 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 955}
956
e67bc2b3 957static int __init init_inodecache(void)
ac27a0ec 958{
617ba13b
MC		 959 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
960 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 961 0, (SLAB_RECLAIM_ACCOUNT|
962 SLAB_MEM_SPREAD),
20c2df83 963 init_once);
617ba13b 964 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 965 return -ENOMEM;
966 return 0;
967}
968
969static void destroy_inodecache(void)
970{
8c0a8537
KS		 971 /*
972 * Make sure all delayed rcu free inodes are flushed before we
973 * destroy cache.
974 */
975 rcu_barrier();
617ba13b 976 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 977}
978
0930fcc1 979void ext4_clear_inode(struct inode *inode)
ac27a0ec 980{
0930fcc1 981 invalidate_inode_buffers(inode);
dbd5768f 982 clear_inode(inode);
9f754758 983 dquot_drop(inode);
c2ea3fde 984 ext4_discard_preallocations(inode);
51865fda 985 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 986 if (EXT4_I(inode)->jinode) {
987 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
988 EXT4_I(inode)->jinode);
989 jbd2_free_inode(EXT4_I(inode)->jinode);
990 EXT4_I(inode)->jinode = NULL;
991 }
b7236e21
TT		 992#ifdef CONFIG_EXT4_FS_ENCRYPTION
993 if (EXT4_I(inode)->i_crypt_info)
c936e1ec 994 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
b7236e21 995#endif
ac27a0ec
DK		 996}
997
1b961ac0 998static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 999 u64 ino, u32 generation)
ac27a0ec 1000{
ac27a0ec 1001 struct inode *inode;
ac27a0ec 1002
617ba13b 1003 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1004 return ERR_PTR(-ESTALE);
617ba13b 1005 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1006 return ERR_PTR(-ESTALE);
1007
1008 /* iget isn't really right if the inode is currently unallocated!!
1009 *
617ba13b 1010 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1011 * deleted, so we should be safe.
1012 *
1013 * Currently we don't know the generation for parent directory, so
1014 * a generation of 0 means "accept any"
1015 */
f4bb2981 1016 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1017 if (IS_ERR(inode))
1018 return ERR_CAST(inode);
1019 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1020 iput(inode);
1021 return ERR_PTR(-ESTALE);
1022 }
1b961ac0
CH		 1023
1024 return inode;
1025}
1026
1027static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1028 int fh_len, int fh_type)
1b961ac0
CH		 1029{
1030 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1031 ext4_nfs_get_inode);
1032}
1033
1034static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1035 int fh_len, int fh_type)
1b961ac0
CH		 1036{
1037 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1038 ext4_nfs_get_inode);
ac27a0ec
DK		 1039}
1040
c39a7f84
TO		 1041/*
1042 * Try to release metadata pages (indirect blocks, directories) which are
1043 * mapped via the block device. Since these pages could have journal heads
1044 * which would prevent try_to_free_buffers() from freeing them, we must use
1045 * jbd2 layer's try_to_free_buffers() function to release them.
1046 */
0b8e58a1
AD		 1047static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1048 gfp_t wait)
c39a7f84
TO		 1049{
1050 journal_t *journal = EXT4_SB(sb)->s_journal;
1051
1052 WARN_ON(PageChecked(page));
1053 if (!page_has_buffers(page))
1054 return 0;
1055 if (journal)
1056 return jbd2_journal_try_to_free_buffers(journal, page,
1057 wait & ~__GFP_WAIT);
1058 return try_to_free_buffers(page);
1059}
1060
ac27a0ec 1061#ifdef CONFIG_QUOTA
af5bc92d 1062#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1063#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1064
617ba13b
MC		 1065static int ext4_write_dquot(struct dquot *dquot);
1066static int ext4_acquire_dquot(struct dquot *dquot);
1067static int ext4_release_dquot(struct dquot *dquot);
1068static int ext4_mark_dquot_dirty(struct dquot *dquot);
1069static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1070static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1071 struct path *path);
ca0e05e4 1072static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1073static int ext4_quota_on_mount(struct super_block *sb, int type);
1074static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1075 size_t len, loff_t off);
617ba13b 1076static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1077 const char *data, size_t len, loff_t off);
7c319d32
AK		 1078static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1079 unsigned int flags);
1080static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1081
96c7e0d9
JK		 1082static struct dquot **ext4_get_dquots(struct inode *inode)
1083{
1084 return EXT4_I(inode)->i_dquot;
1085}
1086
61e225dc 1087static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1088 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1089 .write_dquot = ext4_write_dquot,
1090 .acquire_dquot = ext4_acquire_dquot,
1091 .release_dquot = ext4_release_dquot,
1092 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1093 .write_info = ext4_write_info,
1094 .alloc_dquot = dquot_alloc,
1095 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1096};
1097
0d54b217 1098static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1099 .quota_on = ext4_quota_on,
ca0e05e4 1100 .quota_off = ext4_quota_off,
287a8095 1101 .quota_sync = dquot_quota_sync,
0a240339 1102 .get_state = dquot_get_state,
287a8095
CH		 1103 .set_info = dquot_set_dqinfo,
1104 .get_dqblk = dquot_get_dqblk,
1105 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1106};
1107#endif
1108
ee9b6d61 1109static const struct super_operations ext4_sops = {
617ba13b
MC		 1110 .alloc_inode = ext4_alloc_inode,
1111 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1112 .write_inode = ext4_write_inode,
1113 .dirty_inode = ext4_dirty_inode,
7ff9c073 1114 .drop_inode = ext4_drop_inode,
0930fcc1 1115 .evict_inode = ext4_evict_inode,
617ba13b 1116 .put_super = ext4_put_super,
617ba13b 1117 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1118 .freeze_fs = ext4_freeze,
1119 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1120 .statfs = ext4_statfs,
1121 .remount_fs = ext4_remount,
617ba13b 1122 .show_options = ext4_show_options,
ac27a0ec 1123#ifdef CONFIG_QUOTA
617ba13b
MC		 1124 .quota_read = ext4_quota_read,
1125 .quota_write = ext4_quota_write,
96c7e0d9 1126 .get_dquots = ext4_get_dquots,
ac27a0ec 1127#endif
c39a7f84 1128 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1129};
1130
39655164 1131static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1132 .fh_to_dentry = ext4_fh_to_dentry,
1133 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1134 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1135};
1136
1137enum {
1138 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1139 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1140 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1141 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1142 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1143 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1144 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1145 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1146 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1147 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1148 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1149 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
923ae0ff 1150 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1449032b 1151 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
a26f4992 1152 Opt_lazytime, Opt_nolazytime,
1449032b 1153 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1154 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1155 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1156 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1157 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1158};
1159
a447c093 1160static const match_table_t tokens = {
ac27a0ec
DK		 1161 {Opt_bsd_df, "bsddf"},
1162 {Opt_minix_df, "minixdf"},
1163 {Opt_grpid, "grpid"},
1164 {Opt_grpid, "bsdgroups"},
1165 {Opt_nogrpid, "nogrpid"},
1166 {Opt_nogrpid, "sysvgroups"},
1167 {Opt_resgid, "resgid=%u"},
1168 {Opt_resuid, "resuid=%u"},
1169 {Opt_sb, "sb=%u"},
1170 {Opt_err_cont, "errors=continue"},
1171 {Opt_err_panic, "errors=panic"},
1172 {Opt_err_ro, "errors=remount-ro"},
1173 {Opt_nouid32, "nouid32"},
ac27a0ec 1174 {Opt_debug, "debug"},
72578c33
TT		 1175 {Opt_removed, "oldalloc"},
1176 {Opt_removed, "orlov"},
ac27a0ec
DK		 1177 {Opt_user_xattr, "user_xattr"},
1178 {Opt_nouser_xattr, "nouser_xattr"},
1179 {Opt_acl, "acl"},
1180 {Opt_noacl, "noacl"},
e3bb52ae 1181 {Opt_noload, "norecovery"},
5a916be1 1182 {Opt_noload, "noload"},
72578c33
TT		 1183 {Opt_removed, "nobh"},
1184 {Opt_removed, "bh"},
ac27a0ec 1185 {Opt_commit, "commit=%u"},
30773840
TT		 1186 {Opt_min_batch_time, "min_batch_time=%u"},
1187 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1188 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1189 {Opt_journal_path, "journal_path=%s"},
818d276c 1190 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1191 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1192 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1193 {Opt_abort, "abort"},
1194 {Opt_data_journal, "data=journal"},
1195 {Opt_data_ordered, "data=ordered"},
1196 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1197 {Opt_data_err_abort, "data_err=abort"},
1198 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1199 {Opt_offusrjquota, "usrjquota="},
1200 {Opt_usrjquota, "usrjquota=%s"},
1201 {Opt_offgrpjquota, "grpjquota="},
1202 {Opt_grpjquota, "grpjquota=%s"},
1203 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1204 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1205 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1206 {Opt_grpquota, "grpquota"},
1207 {Opt_noquota, "noquota"},
1208 {Opt_quota, "quota"},
1209 {Opt_usrquota, "usrquota"},
1210 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1211 {Opt_barrier, "barrier"},
1212 {Opt_nobarrier, "nobarrier"},
25ec56b5 1213 {Opt_i_version, "i_version"},
923ae0ff 1214 {Opt_dax, "dax"},
c9de560d 1215 {Opt_stripe, "stripe=%u"},
64769240 1216 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1217 {Opt_lazytime, "lazytime"},
1218 {Opt_nolazytime, "nolazytime"},
dd919b98 1219 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1220 {Opt_removed, "mblk_io_submit"},
1221 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1222 {Opt_block_validity, "block_validity"},
1223 {Opt_noblock_validity, "noblock_validity"},
240799cd 1224 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1225 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1226 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1227 {Opt_auto_da_alloc, "auto_da_alloc"},
1228 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1229 {Opt_dioread_nolock, "dioread_nolock"},
1230 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1231 {Opt_discard, "discard"},
1232 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1233 {Opt_init_itable, "init_itable=%u"},
1234 {Opt_init_itable, "init_itable"},
1235 {Opt_noinit_itable, "noinit_itable"},
df981d03 1236 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1237 {Opt_test_dummy_encryption, "test_dummy_encryption"},
c7198b9c
TT		 1238 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1239 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1240 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1241 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1242 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1243 {Opt_err, NULL},
ac27a0ec
DK		 1244};
1245
617ba13b 1246static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1247{
617ba13b 1248 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1249 char *options = (char *) *data;
1250
1251 if (!options || strncmp(options, "sb=", 3) != 0)
1252 return 1; /* Default location */
0b8e58a1 1253
ac27a0ec 1254 options += 3;
0b8e58a1 1255 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1256 sb_block = simple_strtoul(options, &options, 0);
1257 if (*options && *options != ',') {
4776004f 1258 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1259 (char *) *data);
1260 return 1;
1261 }
1262 if (*options == ',')
1263 options++;
1264 *data = (void *) options;
0b8e58a1 1265
ac27a0ec
DK		 1266 return sb_block;
1267}
1268
b3881f74 1269#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1270static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1271 "Contact [email protected] if you think we should keep it.\n";
b3881f74 1272
56c50f11
DM		 1273#ifdef CONFIG_QUOTA
1274static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1275{
1276 struct ext4_sb_info *sbi = EXT4_SB(sb);
1277 char *qname;
03dafb5f 1278 int ret = -1;
56c50f11
DM		 1279
1280 if (sb_any_quota_loaded(sb) &&
1281 !sbi->s_qf_names[qtype]) {
1282 ext4_msg(sb, KERN_ERR,
1283 "Cannot change journaled "
1284 "quota options when quota turned on");
57f73c2c 1285 return -1;
56c50f11 1286 }
262b4662
JK		 1287 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1288 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1289 "when QUOTA feature is enabled");
1290 return -1;
1291 }
56c50f11
DM		 1292 qname = match_strdup(args);
1293 if (!qname) {
1294 ext4_msg(sb, KERN_ERR,
1295 "Not enough memory for storing quotafile name");
57f73c2c 1296 return -1;
56c50f11 1297 }
03dafb5f
CG		 1298 if (sbi->s_qf_names[qtype]) {
1299 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1300 ret = 1;
1301 else
1302 ext4_msg(sb, KERN_ERR,
1303 "%s quota file already specified",
1304 QTYPE2NAME(qtype));
1305 goto errout;
56c50f11 1306 }
03dafb5f 1307 if (strchr(qname, '/')) {
56c50f11
DM		 1308 ext4_msg(sb, KERN_ERR,
1309 "quotafile must be on filesystem root");
03dafb5f 1310 goto errout;
56c50f11 1311 }
03dafb5f 1312 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1313 set_opt(sb, QUOTA);
56c50f11 1314 return 1;
03dafb5f
CG		 1315errout:
1316 kfree(qname);
1317 return ret;
56c50f11
DM		 1318}
1319
1320static int clear_qf_name(struct super_block *sb, int qtype)
1321{
1322
1323 struct ext4_sb_info *sbi = EXT4_SB(sb);
1324
1325 if (sb_any_quota_loaded(sb) &&
1326 sbi->s_qf_names[qtype]) {
1327 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1328 " when quota turned on");
57f73c2c 1329 return -1;
56c50f11 1330 }
03dafb5f 1331 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1332 sbi->s_qf_names[qtype] = NULL;
1333 return 1;
1334}
1335#endif
1336
26092bf5
TT		 1337#define MOPT_SET 0x0001
1338#define MOPT_CLEAR 0x0002
1339#define MOPT_NOSUPPORT 0x0004
1340#define MOPT_EXPLICIT 0x0008
1341#define MOPT_CLEAR_ERR 0x0010
1342#define MOPT_GTE0 0x0020
ac27a0ec 1343#ifdef CONFIG_QUOTA
26092bf5
TT		 1344#define MOPT_Q 0
1345#define MOPT_QFMT 0x0040
1346#else
1347#define MOPT_Q MOPT_NOSUPPORT
1348#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1349#endif
26092bf5 1350#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1351#define MOPT_NO_EXT2 0x0100
1352#define MOPT_NO_EXT3 0x0200
1353#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1354#define MOPT_STRING 0x0400
26092bf5
TT		 1355
1356static const struct mount_opts {
1357 int token;
1358 int mount_opt;
1359 int flags;
1360} ext4_mount_opts[] = {
1361 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1362 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1363 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1364 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1365 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1366 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1367 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1368 MOPT_EXT4_ONLY | MOPT_SET},
1369 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1370 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1371 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1372 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1373 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1374 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1375 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1376 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1377 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1378 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c
TT		 1379 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1380 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1381 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1382 EXT4_MOUNT_JOURNAL_CHECKSUM),
1383 MOPT_EXT4_ONLY | MOPT_SET},
1384 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1385 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1386 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1387 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1388 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1389 MOPT_NO_EXT2 | MOPT_SET},
1390 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1391 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1392 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1393 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1394 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1395 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1396 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1397 {Opt_commit, 0, MOPT_GTE0},
1398 {Opt_max_batch_time, 0, MOPT_GTE0},
1399 {Opt_min_batch_time, 0, MOPT_GTE0},
1400 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1401 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1402 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1403 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1404 {Opt_resuid, 0, MOPT_GTE0},
1405 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1406 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1407 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1408 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1409 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1410 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1411 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1412 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1413 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1414 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1415#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1416 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1417 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1418#else
26092bf5
TT		 1419 {Opt_acl, 0, MOPT_NOSUPPORT},
1420 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1421#endif
26092bf5
TT		 1422 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1423 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1424 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1425 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1426 MOPT_SET | MOPT_Q},
1427 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1428 MOPT_SET | MOPT_Q},
1429 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1430 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1431 {Opt_usrjquota, 0, MOPT_Q},
1432 {Opt_grpjquota, 0, MOPT_Q},
1433 {Opt_offusrjquota, 0, MOPT_Q},
1434 {Opt_offgrpjquota, 0, MOPT_Q},
1435 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1436 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1437 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1438 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1439 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
26092bf5
TT		 1440 {Opt_err, 0, 0}
1441};
1442
1443static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1444 substring_t *args, unsigned long *journal_devnum,
1445 unsigned int *journal_ioprio, int is_remount)
1446{
1447 struct ext4_sb_info *sbi = EXT4_SB(sb);
1448 const struct mount_opts *m;
08cefc7a
EB		 1449 kuid_t uid;
1450 kgid_t gid;
26092bf5
TT		 1451 int arg = 0;
1452
57f73c2c
TT		 1453#ifdef CONFIG_QUOTA
1454 if (token == Opt_usrjquota)
1455 return set_qf_name(sb, USRQUOTA, &args[0]);
1456 else if (token == Opt_grpjquota)
1457 return set_qf_name(sb, GRPQUOTA, &args[0]);
1458 else if (token == Opt_offusrjquota)
1459 return clear_qf_name(sb, USRQUOTA);
1460 else if (token == Opt_offgrpjquota)
1461 return clear_qf_name(sb, GRPQUOTA);
1462#endif
26092bf5 1463 switch (token) {
f7048605
TT		 1464 case Opt_noacl:
1465 case Opt_nouser_xattr:
1466 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1467 break;
26092bf5
TT		 1468 case Opt_sb:
1469 return 1; /* handled by get_sb_block() */
1470 case Opt_removed:
5f3633e3 1471 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1472 return 1;
26092bf5
TT		 1473 case Opt_abort:
1474 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1475 return 1;
1476 case Opt_i_version:
1477 sb->s_flags |= MS_I_VERSION;
1478 return 1;
a26f4992
TT		 1479 case Opt_lazytime:
1480 sb->s_flags |= MS_LAZYTIME;
1481 return 1;
1482 case Opt_nolazytime:
1483 sb->s_flags &= ~MS_LAZYTIME;
1484 return 1;
26092bf5
TT		 1485 }
1486
5f3633e3
JK		 1487 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1488 if (token == m->token)
1489 break;
1490
1491 if (m->token == Opt_err) {
1492 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1493 "or missing value", opt);
1494 return -1;
1495 }
1496
8dc0aa8c
TT		 1497 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1498 ext4_msg(sb, KERN_ERR,
1499 "Mount option \"%s\" incompatible with ext2", opt);
1500 return -1;
1501 }
1502 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1503 ext4_msg(sb, KERN_ERR,
1504 "Mount option \"%s\" incompatible with ext3", opt);
1505 return -1;
1506 }
1507
ad4eec61 1508 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1509 return -1;
1510 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1511 return -1;
1512 if (m->flags & MOPT_EXPLICIT)
1513 set_opt2(sb, EXPLICIT_DELALLOC);
1514 if (m->flags & MOPT_CLEAR_ERR)
1515 clear_opt(sb, ERRORS_MASK);
1516 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1517 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1518 "options when quota turned on");
1519 return -1;
1520 }
1521
1522 if (m->flags & MOPT_NOSUPPORT) {
1523 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1524 } else if (token == Opt_commit) {
1525 if (arg == 0)
1526 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1527 sbi->s_commit_interval = HZ * arg;
1528 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1529 sbi->s_max_batch_time = arg;
1530 } else if (token == Opt_min_batch_time) {
1531 sbi->s_min_batch_time = arg;
1532 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1533 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1534 ext4_msg(sb, KERN_ERR,
1535 "EXT4-fs: inode_readahead_blks must be "
1536 "0 or a power of 2 smaller than 2^31");
26092bf5 1537 return -1;
5f3633e3
JK		 1538 }
1539 sbi->s_inode_readahead_blks = arg;
1540 } else if (token == Opt_init_itable) {
1541 set_opt(sb, INIT_INODE_TABLE);
1542 if (!args->from)
1543 arg = EXT4_DEF_LI_WAIT_MULT;
1544 sbi->s_li_wait_mult = arg;
1545 } else if (token == Opt_max_dir_size_kb) {
1546 sbi->s_max_dir_size_kb = arg;
1547 } else if (token == Opt_stripe) {
1548 sbi->s_stripe = arg;
1549 } else if (token == Opt_resuid) {
1550 uid = make_kuid(current_user_ns(), arg);
1551 if (!uid_valid(uid)) {
1552 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1553 return -1;
1554 }
5f3633e3
JK		 1555 sbi->s_resuid = uid;
1556 } else if (token == Opt_resgid) {
1557 gid = make_kgid(current_user_ns(), arg);
1558 if (!gid_valid(gid)) {
1559 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1560 return -1;
1561 }
1562 sbi->s_resgid = gid;
1563 } else if (token == Opt_journal_dev) {
1564 if (is_remount) {
1565 ext4_msg(sb, KERN_ERR,
1566 "Cannot specify journal on remount");
1567 return -1;
1568 }
1569 *journal_devnum = arg;
ad4eec61
ES		 1570 } else if (token == Opt_journal_path) {
1571 char *journal_path;
1572 struct inode *journal_inode;
1573 struct path path;
1574 int error;
1575
1576 if (is_remount) {
1577 ext4_msg(sb, KERN_ERR,
1578 "Cannot specify journal on remount");
1579 return -1;
1580 }
1581 journal_path = match_strdup(&args[0]);
1582 if (!journal_path) {
1583 ext4_msg(sb, KERN_ERR, "error: could not dup "
1584 "journal device string");
1585 return -1;
1586 }
1587
1588 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1589 if (error) {
1590 ext4_msg(sb, KERN_ERR, "error: could not find "
1591 "journal device path: error %d", error);
1592 kfree(journal_path);
1593 return -1;
1594 }
1595
2b0143b5 1596 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1597 if (!S_ISBLK(journal_inode->i_mode)) {
1598 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1599 "is not a block device", journal_path);
1600 path_put(&path);
1601 kfree(journal_path);
1602 return -1;
1603 }
1604
1605 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1606 path_put(&path);
1607 kfree(journal_path);
5f3633e3
JK		 1608 } else if (token == Opt_journal_ioprio) {
1609 if (arg > 7) {
1610 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1611 " (must be 0-7)");
1612 return -1;
1613 }
1614 *journal_ioprio =
1615 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1616 } else if (token == Opt_test_dummy_encryption) {
1617#ifdef CONFIG_EXT4_FS_ENCRYPTION
1618 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1619 ext4_msg(sb, KERN_WARNING,
1620 "Test dummy encryption mode enabled");
1621#else
1622 ext4_msg(sb, KERN_WARNING,
1623 "Test dummy encryption mount option ignored");
1624#endif
5f3633e3
JK		 1625 } else if (m->flags & MOPT_DATAJ) {
1626 if (is_remount) {
1627 if (!sbi->s_journal)
1628 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1629 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1630 ext4_msg(sb, KERN_ERR,
26092bf5 1631 "Cannot change data mode on remount");
26092bf5 1632 return -1;
ac27a0ec 1633 }
26092bf5 1634 } else {
5f3633e3
JK		 1635 clear_opt(sb, DATA_FLAGS);
1636 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1637 }
5f3633e3
JK		 1638#ifdef CONFIG_QUOTA
1639 } else if (m->flags & MOPT_QFMT) {
1640 if (sb_any_quota_loaded(sb) &&
1641 sbi->s_jquota_fmt != m->mount_opt) {
1642 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1643 "quota options when quota turned on");
1644 return -1;
1645 }
262b4662
JK		 1646 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1647 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1648 ext4_msg(sb, KERN_ERR,
1649 "Cannot set journaled quota options "
1650 "when QUOTA feature is enabled");
1651 return -1;
1652 }
5f3633e3 1653 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff
RZ		 1654#endif
1655#ifndef CONFIG_FS_DAX
1656 } else if (token == Opt_dax) {
1657 ext4_msg(sb, KERN_INFO, "dax option not supported");
1658 return -1;
5f3633e3
JK		 1659#endif
1660 } else {
1661 if (!args->from)
1662 arg = 1;
1663 if (m->flags & MOPT_CLEAR)
1664 arg = !arg;
1665 else if (unlikely(!(m->flags & MOPT_SET))) {
1666 ext4_msg(sb, KERN_WARNING,
1667 "buggy handling of option %s", opt);
1668 WARN_ON(1);
1669 return -1;
1670 }
1671 if (arg != 0)
1672 sbi->s_mount_opt |= m->mount_opt;
1673 else
1674 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1675 }
5f3633e3 1676 return 1;
26092bf5
TT		 1677}
1678
1679static int parse_options(char *options, struct super_block *sb,
1680 unsigned long *journal_devnum,
1681 unsigned int *journal_ioprio,
1682 int is_remount)
1683{
1684 struct ext4_sb_info *sbi = EXT4_SB(sb);
1685 char *p;
1686 substring_t args[MAX_OPT_ARGS];
1687 int token;
1688
1689 if (!options)
1690 return 1;
1691
1692 while ((p = strsep(&options, ",")) != NULL) {
1693 if (!*p)
1694 continue;
1695 /*
1696 * Initialize args struct so we know whether arg was
1697 * found; some options take optional arguments.
1698 */
caecd0af 1699 args[0].to = args[0].from = NULL;
26092bf5
TT		 1700 token = match_token(p, tokens, args);
1701 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1702 journal_ioprio, is_remount) < 0)
1703 return 0;
ac27a0ec
DK		 1704 }
1705#ifdef CONFIG_QUOTA
262b4662
JK		 1706 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1707 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1708 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1709 "feature is enabled");
1710 return 0;
1711 }
ac27a0ec 1712 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1713 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1714 clear_opt(sb, USRQUOTA);
ac27a0ec 1715
482a7425 1716 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1717 clear_opt(sb, GRPQUOTA);
ac27a0ec 1718
56c50f11 1719 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1720 ext4_msg(sb, KERN_ERR, "old and new quota "
1721 "format mixing");
ac27a0ec
DK		 1722 return 0;
1723 }
1724
1725 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1726 ext4_msg(sb, KERN_ERR, "journaled quota format "
1727 "not specified");
ac27a0ec
DK		 1728 return 0;
1729 }
ac27a0ec
DK		 1730 }
1731#endif
261cb20c
JK		 1732 if (test_opt(sb, DIOREAD_NOLOCK)) {
1733 int blocksize =
1734 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1735
1736 if (blocksize < PAGE_CACHE_SIZE) {
1737 ext4_msg(sb, KERN_ERR, "can't mount with "
1738 "dioread_nolock if block size != PAGE_SIZE");
1739 return 0;
1740 }
1741 }
d4f76107
JK		 1742 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1743 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1744 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1745 "in data=ordered mode");
1746 return 0;
1747 }
ac27a0ec
DK		 1748 return 1;
1749}
1750
2adf6da8
TT		 1751static inline void ext4_show_quota_options(struct seq_file *seq,
1752 struct super_block *sb)
1753{
1754#if defined(CONFIG_QUOTA)
1755 struct ext4_sb_info *sbi = EXT4_SB(sb);
1756
1757 if (sbi->s_jquota_fmt) {
1758 char *fmtname = "";
1759
1760 switch (sbi->s_jquota_fmt) {
1761 case QFMT_VFS_OLD:
1762 fmtname = "vfsold";
1763 break;
1764 case QFMT_VFS_V0:
1765 fmtname = "vfsv0";
1766 break;
1767 case QFMT_VFS_V1:
1768 fmtname = "vfsv1";
1769 break;
1770 }
1771 seq_printf(seq, ",jqfmt=%s", fmtname);
1772 }
1773
1774 if (sbi->s_qf_names[USRQUOTA])
a068acf2 1775 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 1776
1777 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 1778 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1779#endif
1780}
1781
5a916be1
TT		 1782static const char *token2str(int token)
1783{
50df9fd5 1784 const struct match_token *t;
5a916be1
TT		 1785
1786 for (t = tokens; t->token != Opt_err; t++)
1787 if (t->token == token && !strchr(t->pattern, '='))
1788 break;
1789 return t->pattern;
1790}
1791
2adf6da8
TT		 1792/*
1793 * Show an option if
1794 * - it's set to a non-default value OR
1795 * - if the per-sb default is different from the global default
1796 */
66acdcf4
TT		 1797static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1798 int nodefs)
2adf6da8 1799{
2adf6da8
TT		 1800 struct ext4_sb_info *sbi = EXT4_SB(sb);
1801 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1802 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1803 const struct mount_opts *m;
66acdcf4 1804 char sep = nodefs ? '\n' : ',';
2adf6da8 1805
66acdcf4
TT		 1806#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1807#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1808
1809 if (sbi->s_sb_block != 1)
5a916be1
TT		 1810 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1811
1812 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1813 int want_set = m->flags & MOPT_SET;
1814 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1815 (m->flags & MOPT_CLEAR_ERR))
1816 continue;
66acdcf4 1817 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1818 continue; /* skip if same as the default */
1819 if ((want_set &&
1820 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1821 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1822 continue; /* select Opt_noFoo vs Opt_Foo */
1823 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1824 }
5a916be1 1825
08cefc7a 1826 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1827 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1828 SEQ_OPTS_PRINT("resuid=%u",
1829 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1830 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1831 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1832 SEQ_OPTS_PRINT("resgid=%u",
1833 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1834 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1835 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1836 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1837 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1838 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1839 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1840 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1841 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1842 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1843 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1844 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1845 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1846 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1847 if (sb->s_flags & MS_I_VERSION)
5a916be1 1848 SEQ_OPTS_PUTS("i_version");
66acdcf4 1849 if (nodefs || sbi->s_stripe)
5a916be1 1850 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1851 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1852 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1853 SEQ_OPTS_PUTS("data=journal");
1854 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1855 SEQ_OPTS_PUTS("data=ordered");
1856 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1857 SEQ_OPTS_PUTS("data=writeback");
1858 }
66acdcf4
TT		 1859 if (nodefs ||
1860 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1861 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1862 sbi->s_inode_readahead_blks);
2adf6da8 1863
66acdcf4
TT		 1864 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1865 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1866 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1867 if (nodefs || sbi->s_max_dir_size_kb)
1868 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1869
1870 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1871 return 0;
1872}
1873
66acdcf4
TT		 1874static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1875{
1876 return _ext4_show_options(seq, root->d_sb, 0);
1877}
1878
ebd173be 1879int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 1880{
1881 struct super_block *sb = seq->private;
1882 int rc;
1883
1884 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1885 rc = _ext4_show_options(seq, sb, 1);
1886 seq_puts(seq, "\n");
1887 return rc;
1888}
1889
617ba13b 1890static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1891 int read_only)
1892{
617ba13b 1893 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1894 int res = 0;
1895
617ba13b 1896 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1897 ext4_msg(sb, KERN_ERR, "revision level too high, "
1898 "forcing read-only mode");
ac27a0ec
DK		 1899 res = MS_RDONLY;
1900 }
1901 if (read_only)
281b5995 1902 goto done;
617ba13b 1903 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1904 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1905 "running e2fsck is recommended");
c8b459f4 1906 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 1907 ext4_msg(sb, KERN_WARNING,
1908 "warning: mounting fs with errors, "
1909 "running e2fsck is recommended");
ed3ce80a 1910 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1911 le16_to_cpu(es->s_mnt_count) >=
1912 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1913 ext4_msg(sb, KERN_WARNING,
1914 "warning: maximal mount count reached, "
1915 "running e2fsck is recommended");
ac27a0ec
DK		 1916 else if (le32_to_cpu(es->s_checkinterval) &&
1917 (le32_to_cpu(es->s_lastcheck) +
1918 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1919 ext4_msg(sb, KERN_WARNING,
1920 "warning: checktime reached, "
1921 "running e2fsck is recommended");
0b8e58a1 1922 if (!sbi->s_journal)
0390131b 1923 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1924 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1925 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1926 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1927 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1928 ext4_update_dynamic_rev(sb);
0390131b
FM		 1929 if (sbi->s_journal)
1930 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1931
e2d67052 1932 ext4_commit_super(sb, 1);
281b5995 1933done:
ac27a0ec 1934 if (test_opt(sb, DEBUG))
a9df9a49 1935 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1936 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1937 sb->s_blocksize,
1938 sbi->s_groups_count,
617ba13b
MC		 1939 EXT4_BLOCKS_PER_GROUP(sb),
1940 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1941 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1942
7abc52c2 1943 cleancache_init_fs(sb);
ac27a0ec
DK		 1944 return res;
1945}
1946
117fff10
TT		 1947int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1948{
1949 struct ext4_sb_info *sbi = EXT4_SB(sb);
1950 struct flex_groups *new_groups;
1951 int size;
1952
1953 if (!sbi->s_log_groups_per_flex)
1954 return 0;
1955
1956 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1957 if (size <= sbi->s_flex_groups_allocated)
1958 return 0;
1959
1960 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1961 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1962 if (!new_groups) {
1963 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1964 size / (int) sizeof(struct flex_groups));
1965 return -ENOMEM;
1966 }
1967
1968 if (sbi->s_flex_groups) {
1969 memcpy(new_groups, sbi->s_flex_groups,
1970 (sbi->s_flex_groups_allocated *
1971 sizeof(struct flex_groups)));
b93b41d4 1972 kvfree(sbi->s_flex_groups);
117fff10
TT		 1973 }
1974 sbi->s_flex_groups = new_groups;
1975 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1976 return 0;
1977}
1978
772cb7c8
JS		 1979static int ext4_fill_flex_info(struct super_block *sb)
1980{
1981 struct ext4_sb_info *sbi = EXT4_SB(sb);
1982 struct ext4_group_desc *gdp = NULL;
772cb7c8 1983 ext4_group_t flex_group;
117fff10 1984 int i, err;
772cb7c8 1985
503358ae 1986 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1987 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1988 sbi->s_log_groups_per_flex = 0;
1989 return 1;
1990 }
1991
117fff10
TT		 1992 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1993 if (err)
9933fc0a 1994 goto failed;
772cb7c8 1995
772cb7c8 1996 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1997 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1998
1999 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2000 atomic_add(ext4_free_inodes_count(sb, gdp),
2001 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2002 atomic64_add(ext4_free_group_clusters(sb, gdp),
2003 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2004 atomic_add(ext4_used_dirs_count(sb, gdp),
2005 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2006 }
2007
2008 return 1;
2009failed:
2010 return 0;
2011}
2012
feb0ab32
DW		 2013static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2014 struct ext4_group_desc *gdp)
717d50e4 2015{
feb0ab32 2016 int offset;
717d50e4 2017 __u16 crc = 0;
feb0ab32 2018 __le32 le_group = cpu_to_le32(block_group);
717d50e4 2019
9aa5d32b 2020 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2021 /* Use new metadata_csum algorithm */
d6a77105 2022 __le16 save_csum;
feb0ab32
DW		 2023 __u32 csum32;
2024
d6a77105 2025 save_csum = gdp->bg_checksum;
feb0ab32
DW		 2026 gdp->bg_checksum = 0;
2027 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2028 sizeof(le_group));
2029 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2030 sbi->s_desc_size);
d6a77105 2031 gdp->bg_checksum = save_csum;
feb0ab32
DW		 2032
2033 crc = csum32 & 0xFFFF;
2034 goto out;
717d50e4
AD		 2035 }
2036
feb0ab32 2037 /* old crc16 code */
813d32f9
DW		 2038 if (!(sbi->s_es->s_feature_ro_compat &
2039 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2040 return 0;
2041
feb0ab32
DW		 2042 offset = offsetof(struct ext4_group_desc, bg_checksum);
2043
2044 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2045 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2046 crc = crc16(crc, (__u8 *)gdp, offset);
2047 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2048 /* for checksum of struct ext4_group_desc do the rest...*/
2049 if ((sbi->s_es->s_feature_incompat &
2050 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2051 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2052 crc = crc16(crc, (__u8 *)gdp + offset,
2053 le16_to_cpu(sbi->s_es->s_desc_size) -
2054 offset);
2055
2056out:
717d50e4
AD		 2057 return cpu_to_le16(crc);
2058}
2059
feb0ab32 2060int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2061 struct ext4_group_desc *gdp)
2062{
feb0ab32
DW		 2063 if (ext4_has_group_desc_csum(sb) &&
2064 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2065 block_group, gdp)))
717d50e4
AD		 2066 return 0;
2067
2068 return 1;
2069}
2070
feb0ab32
DW		 2071void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2072 struct ext4_group_desc *gdp)
2073{
2074 if (!ext4_has_group_desc_csum(sb))
2075 return;
2076 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2077}
2078
ac27a0ec 2079/* Called at mount-time, super-block is locked */
bfff6873
LC		 2080static int ext4_check_descriptors(struct super_block *sb,
2081 ext4_group_t *first_not_zeroed)
ac27a0ec 2082{
617ba13b
MC		 2083 struct ext4_sb_info *sbi = EXT4_SB(sb);
2084 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2085 ext4_fsblk_t last_block;
bd81d8ee
LV		 2086 ext4_fsblk_t block_bitmap;
2087 ext4_fsblk_t inode_bitmap;
2088 ext4_fsblk_t inode_table;
ce421581 2089 int flexbg_flag = 0;
bfff6873 2090 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2091
ce421581
JS		 2092 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2093 flexbg_flag = 1;
2094
af5bc92d 2095 ext4_debug("Checking group descriptors");
ac27a0ec 2096
197cd65a
AM		 2097 for (i = 0; i < sbi->s_groups_count; i++) {
2098 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2099
ce421581 2100 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2101 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2102 else
2103 last_block = first_block +
617ba13b 2104 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2105
bfff6873
LC		 2106 if ((grp == sbi->s_groups_count) &&
2107 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2108 grp = i;
2109
8fadc143 2110 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2111 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2112 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2113 "Block bitmap for group %u not in group "
b31e1552 2114 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2115 return 0;
2116 }
8fadc143 2117 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2118 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2119 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2120 "Inode bitmap for group %u not in group "
b31e1552 2121 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2122 return 0;
2123 }
8fadc143 2124 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2125 if (inode_table < first_block ||
2b2d6d01 2126 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2127 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2128 "Inode table for group %u not in group "
b31e1552 2129 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2130 return 0;
2131 }
955ce5f5 2132 ext4_lock_group(sb, i);
feb0ab32 2133 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2134 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2135 "Checksum for group %u failed (%u!=%u)",
2136 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2137 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2138 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2139 ext4_unlock_group(sb, i);
8a266467 2140 return 0;
7ee1ec4c 2141 }
717d50e4 2142 }
955ce5f5 2143 ext4_unlock_group(sb, i);
ce421581
JS		 2144 if (!flexbg_flag)
2145 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2146 }
bfff6873
LC		 2147 if (NULL != first_not_zeroed)
2148 *first_not_zeroed = grp;
ac27a0ec
DK		 2149 return 1;
2150}
2151
617ba13b 2152/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2153 * the superblock) which were deleted from all directories, but held open by
2154 * a process at the time of a crash. We walk the list and try to delete these
2155 * inodes at recovery time (only with a read-write filesystem).
2156 *
2157 * In order to keep the orphan inode chain consistent during traversal (in
2158 * case of crash during recovery), we link each inode into the superblock
2159 * orphan list_head and handle it the same way as an inode deletion during
2160 * normal operation (which journals the operations for us).
2161 *
2162 * We only do an iget() and an iput() on each inode, which is very safe if we
2163 * accidentally point at an in-use or already deleted inode. The worst that
2164 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2165 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2166 * e2fsck was run on this filesystem, and it must have already done the orphan
2167 * inode cleanup for us, so we can safely abort without any further action.
2168 */
2b2d6d01
TT		 2169static void ext4_orphan_cleanup(struct super_block *sb,
2170 struct ext4_super_block *es)
ac27a0ec
DK		 2171{
2172 unsigned int s_flags = sb->s_flags;
2173 int nr_orphans = 0, nr_truncates = 0;
2174#ifdef CONFIG_QUOTA
2175 int i;
2176#endif
2177 if (!es->s_last_orphan) {
2178 jbd_debug(4, "no orphan inodes to clean up\n");
2179 return;
2180 }
2181
a8f48a95 2182 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2183 ext4_msg(sb, KERN_ERR, "write access "
2184 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2185 return;
2186 }
2187
d39195c3
AG		 2188 /* Check if feature set would not allow a r/w mount */
2189 if (!ext4_feature_set_ok(sb, 0)) {
2190 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2191 "unknown ROCOMPAT features");
2192 return;
2193 }
2194
617ba13b 2195 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2196 /* don't clear list on RO mount w/ errors */
2197 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2198 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2199 "clearing orphan list.\n");
c25f9bc6
ES		 2200 es->s_last_orphan = 0;
2201 }
ac27a0ec
DK		 2202 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2203 return;
2204 }
2205
2206 if (s_flags & MS_RDONLY) {
b31e1552 2207 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2208 sb->s_flags &= ~MS_RDONLY;
2209 }
2210#ifdef CONFIG_QUOTA
2211 /* Needed for iput() to work correctly and not trash data */
2212 sb->s_flags |= MS_ACTIVE;
2213 /* Turn on quotas so that they are updated correctly */
a2d4a646 2214 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2215 if (EXT4_SB(sb)->s_qf_names[i]) {
2216 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2217 if (ret < 0)
b31e1552
ES		 2218 ext4_msg(sb, KERN_ERR,
2219 "Cannot turn on journaled "
2220 "quota: error %d", ret);
ac27a0ec
DK		 2221 }
2222 }
2223#endif
2224
2225 while (es->s_last_orphan) {
2226 struct inode *inode;
2227
97bd42b9
JB		 2228 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2229 if (IS_ERR(inode)) {
ac27a0ec
DK		 2230 es->s_last_orphan = 0;
2231 break;
2232 }
2233
617ba13b 2234 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2235 dquot_initialize(inode);
ac27a0ec 2236 if (inode->i_nlink) {
566370a2
PT		 2237 if (test_opt(sb, DEBUG))
2238 ext4_msg(sb, KERN_DEBUG,
2239 "%s: truncating inode %lu to %lld bytes",
2240 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2241 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2242 inode->i_ino, inode->i_size);
721e3eba 2243 mutex_lock(&inode->i_mutex);
55f252c9 2244 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2245 ext4_truncate(inode);
721e3eba 2246 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2247 nr_truncates++;
2248 } else {
566370a2
PT		 2249 if (test_opt(sb, DEBUG))
2250 ext4_msg(sb, KERN_DEBUG,
2251 "%s: deleting unreferenced inode %lu",
2252 __func__, inode->i_ino);
ac27a0ec
DK		 2253 jbd_debug(2, "deleting unreferenced inode %lu\n",
2254 inode->i_ino);
2255 nr_orphans++;
2256 }
2257 iput(inode); /* The delete magic happens here! */
2258 }
2259
2b2d6d01 2260#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2261
2262 if (nr_orphans)
b31e1552
ES		 2263 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2264 PLURAL(nr_orphans));
ac27a0ec 2265 if (nr_truncates)
b31e1552
ES		 2266 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2267 PLURAL(nr_truncates));
ac27a0ec
DK		 2268#ifdef CONFIG_QUOTA
2269 /* Turn quotas off */
a2d4a646 2270 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2271 if (sb_dqopt(sb)->files[i])
287a8095 2272 dquot_quota_off(sb, i);
ac27a0ec
DK		 2273 }
2274#endif
2275 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2276}
0b8e58a1 2277
cd2291a4
ES		 2278/*
2279 * Maximal extent format file size.
2280 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2281 * extent format containers, within a sector_t, and within i_blocks
2282 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2283 * so that won't be a limiting factor.
2284 *
f17722f9
LC		 2285 * However there is other limiting factor. We do store extents in the form
2286 * of starting block and length, hence the resulting length of the extent
2287 * covering maximum file size must fit into on-disk format containers as
2288 * well. Given that length is always by 1 unit bigger than max unit (because
2289 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2290 *
cd2291a4
ES		 2291 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2292 */
f287a1a5 2293static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2294{
2295 loff_t res;
2296 loff_t upper_limit = MAX_LFS_FILESIZE;
2297
2298 /* small i_blocks in vfs inode? */
f287a1a5 2299 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2300 /*
90c699a9 2301 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2302 * i_block represent total blocks in 512 bytes
2303 * 32 == size of vfs inode i_blocks * 8
2304 */
2305 upper_limit = (1LL << 32) - 1;
2306
2307 /* total blocks in file system block size */
2308 upper_limit >>= (blkbits - 9);
2309 upper_limit <<= blkbits;
2310 }
2311
f17722f9
LC		 2312 /*
2313 * 32-bit extent-start container, ee_block. We lower the maxbytes
2314 * by one fs block, so ee_len can cover the extent of maximum file
2315 * size
2316 */
2317 res = (1LL << 32) - 1;
cd2291a4 2318 res <<= blkbits;
cd2291a4
ES		 2319
2320 /* Sanity check against vm- & vfs- imposed limits */
2321 if (res > upper_limit)
2322 res = upper_limit;
2323
2324 return res;
2325}
ac27a0ec 2326
ac27a0ec 2327/*
cd2291a4 2328 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2329 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2330 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2331 */
f287a1a5 2332static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2333{
617ba13b 2334 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2335 int meta_blocks;
2336 loff_t upper_limit;
0b8e58a1
AD		 2337 /* This is calculated to be the largest file size for a dense, block
2338 * mapped file such that the file's total number of 512-byte sectors,
2339 * including data and all indirect blocks, does not exceed (2^48 - 1).
2340 *
2341 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2342 * number of 512-byte sectors of the file.
0fc1b451
AK		 2343 */
2344
f287a1a5 2345 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2346 /*
90c699a9 2347 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2348 * the inode i_block field represents total file blocks in
2349 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2350 */
2351 upper_limit = (1LL << 32) - 1;
2352
2353 /* total blocks in file system block size */
2354 upper_limit >>= (bits - 9);
2355
2356 } else {
8180a562
AK		 2357 /*
2358 * We use 48 bit ext4_inode i_blocks
2359 * With EXT4_HUGE_FILE_FL set the i_blocks
2360 * represent total number of blocks in
2361 * file system block size
2362 */
0fc1b451
AK		 2363 upper_limit = (1LL << 48) - 1;
2364
0fc1b451
AK		 2365 }
2366
2367 /* indirect blocks */
2368 meta_blocks = 1;
2369 /* double indirect blocks */
2370 meta_blocks += 1 + (1LL << (bits-2));
2371 /* tripple indirect blocks */
2372 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2373
2374 upper_limit -= meta_blocks;
2375 upper_limit <<= bits;
ac27a0ec
DK		 2376
2377 res += 1LL << (bits-2);
2378 res += 1LL << (2*(bits-2));
2379 res += 1LL << (3*(bits-2));
2380 res <<= bits;
2381 if (res > upper_limit)
2382 res = upper_limit;
0fc1b451
AK		 2383
2384 if (res > MAX_LFS_FILESIZE)
2385 res = MAX_LFS_FILESIZE;
2386
ac27a0ec
DK		 2387 return res;
2388}
2389
617ba13b 2390static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2391 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2392{
617ba13b 2393 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2394 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2395 int has_super = 0;
2396
2397 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2398
617ba13b 2399 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2400 nr < first_meta_bg)
70bbb3e0 2401 return logical_sb_block + nr + 1;
ac27a0ec 2402 bg = sbi->s_desc_per_block * nr;
617ba13b 2403 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2404 has_super = 1;
0b8e58a1 2405
bd63f6b0
DW		 2406 /*
2407 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2408 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2409 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2410 * compensate.
2411 */
2412 if (sb->s_blocksize == 1024 && nr == 0 &&
2413 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2414 has_super++;
2415
617ba13b 2416 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2417}
2418
c9de560d
AT		 2419/**
2420 * ext4_get_stripe_size: Get the stripe size.
2421 * @sbi: In memory super block info
2422 *
2423 * If we have specified it via mount option, then
2424 * use the mount option value. If the value specified at mount time is
2425 * greater than the blocks per group use the super block value.
2426 * If the super block value is greater than blocks per group return 0.
2427 * Allocator needs it be less than blocks per group.
2428 *
2429 */
2430static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2431{
2432 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2433 unsigned long stripe_width =
2434 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2435 int ret;
c9de560d
AT		 2436
2437 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2438 ret = sbi->s_stripe;
2439 else if (stripe_width <= sbi->s_blocks_per_group)
2440 ret = stripe_width;
2441 else if (stride <= sbi->s_blocks_per_group)
2442 ret = stride;
2443 else
2444 ret = 0;
c9de560d 2445
3eb08658
DE		 2446 /*
2447 * If the stripe width is 1, this makes no sense and
2448 * we set it to 0 to turn off stripe handling code.
2449 */
2450 if (ret <= 1)
2451 ret = 0;
c9de560d 2452
3eb08658 2453 return ret;
c9de560d 2454}
ac27a0ec 2455
a13fb1a4
ES		 2456/*
2457 * Check whether this filesystem can be mounted based on
2458 * the features present and the RDONLY/RDWR mount requested.
2459 * Returns 1 if this filesystem can be mounted as requested,
2460 * 0 if it cannot be.
2461 */
2462static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2463{
2464 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2465 ext4_msg(sb, KERN_ERR,
2466 "Couldn't mount because of "
2467 "unsupported optional features (%x)",
2468 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2469 ~EXT4_FEATURE_INCOMPAT_SUPP));
2470 return 0;
2471 }
2472
2473 if (readonly)
2474 return 1;
2475
2cb5cc8b
DW		 2476 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2477 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2478 sb->s_flags |= MS_RDONLY;
2479 return 1;
2480 }
2481
a13fb1a4
ES		 2482 /* Check that feature set is OK for a read-write mount */
2483 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2484 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2485 "unsupported optional features (%x)",
2486 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2487 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2488 return 0;
2489 }
2490 /*
2491 * Large file size enabled file system can only be mounted
2492 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2493 */
2494 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2495 if (sizeof(blkcnt_t) < sizeof(u64)) {
2496 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2497 "cannot be mounted RDWR without "
2498 "CONFIG_LBDAF");
2499 return 0;
2500 }
2501 }
bab08ab9
TT		 2502 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2503 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2504 ext4_msg(sb, KERN_ERR,
2505 "Can't support bigalloc feature without "
2506 "extents feature\n");
2507 return 0;
2508 }
7c319d32
AK		 2509
2510#ifndef CONFIG_QUOTA
2511 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2512 !readonly) {
2513 ext4_msg(sb, KERN_ERR,
2514 "Filesystem with quota feature cannot be mounted RDWR "
2515 "without CONFIG_QUOTA");
2516 return 0;
2517 }
2518#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2519 return 1;
2520}
2521
66e61a9e
TT		 2522/*
2523 * This function is called once a day if we have errors logged
2524 * on the file system
2525 */
2526static void print_daily_error_info(unsigned long arg)
2527{
2528 struct super_block *sb = (struct super_block *) arg;
2529 struct ext4_sb_info *sbi;
2530 struct ext4_super_block *es;
2531
2532 sbi = EXT4_SB(sb);
2533 es = sbi->s_es;
2534
2535 if (es->s_error_count)
ae0f78de
TT		 2536 /* fsck newer than v1.41.13 is needed to clean this condition. */
2537 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2538 le32_to_cpu(es->s_error_count));
2539 if (es->s_first_error_time) {
ae0f78de 2540 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2541 sb->s_id, le32_to_cpu(es->s_first_error_time),
2542 (int) sizeof(es->s_first_error_func),
2543 es->s_first_error_func,
2544 le32_to_cpu(es->s_first_error_line));
2545 if (es->s_first_error_ino)
2546 printk(": inode %u",
2547 le32_to_cpu(es->s_first_error_ino));
2548 if (es->s_first_error_block)
2549 printk(": block %llu", (unsigned long long)
2550 le64_to_cpu(es->s_first_error_block));
2551 printk("\n");
2552 }
2553 if (es->s_last_error_time) {
ae0f78de 2554 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2555 sb->s_id, le32_to_cpu(es->s_last_error_time),
2556 (int) sizeof(es->s_last_error_func),
2557 es->s_last_error_func,
2558 le32_to_cpu(es->s_last_error_line));
2559 if (es->s_last_error_ino)
2560 printk(": inode %u",
2561 le32_to_cpu(es->s_last_error_ino));
2562 if (es->s_last_error_block)
2563 printk(": block %llu", (unsigned long long)
2564 le64_to_cpu(es->s_last_error_block));
2565 printk("\n");
2566 }
2567 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2568}
2569
bfff6873
LC		 2570/* Find next suitable group and run ext4_init_inode_table */
2571static int ext4_run_li_request(struct ext4_li_request *elr)
2572{
2573 struct ext4_group_desc *gdp = NULL;
2574 ext4_group_t group, ngroups;
2575 struct super_block *sb;
2576 unsigned long timeout = 0;
2577 int ret = 0;
2578
2579 sb = elr->lr_super;
2580 ngroups = EXT4_SB(sb)->s_groups_count;
2581
8e8ad8a5 2582 sb_start_write(sb);
bfff6873
LC		 2583 for (group = elr->lr_next_group; group < ngroups; group++) {
2584 gdp = ext4_get_group_desc(sb, group, NULL);
2585 if (!gdp) {
2586 ret = 1;
2587 break;
2588 }
2589
2590 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2591 break;
2592 }
2593
7f511862 2594 if (group >= ngroups)
bfff6873
LC		 2595 ret = 1;
2596
2597 if (!ret) {
2598 timeout = jiffies;
2599 ret = ext4_init_inode_table(sb, group,
2600 elr->lr_timeout ? 0 : 1);
2601 if (elr->lr_timeout == 0) {
51ce6511
LC		 2602 timeout = (jiffies - timeout) *
2603 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2604 elr->lr_timeout = timeout;
2605 }
2606 elr->lr_next_sched = jiffies + elr->lr_timeout;
2607 elr->lr_next_group = group + 1;
2608 }
8e8ad8a5 2609 sb_end_write(sb);
bfff6873
LC		 2610
2611 return ret;
2612}
2613
2614/*
2615 * Remove lr_request from the list_request and free the
4ed5c033 2616 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2617 */
2618static void ext4_remove_li_request(struct ext4_li_request *elr)
2619{
2620 struct ext4_sb_info *sbi;
2621
2622 if (!elr)
2623 return;
2624
2625 sbi = elr->lr_sbi;
2626
2627 list_del(&elr->lr_request);
2628 sbi->s_li_request = NULL;
2629 kfree(elr);
2630}
2631
2632static void ext4_unregister_li_request(struct super_block *sb)
2633{
1bb933fb
LC		 2634 mutex_lock(&ext4_li_mtx);
2635 if (!ext4_li_info) {
2636 mutex_unlock(&ext4_li_mtx);
bfff6873 2637 return;
1bb933fb 2638 }
bfff6873
LC		 2639
2640 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2641 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2642 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2643 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2644}
2645
8f1f7453
ES		 2646static struct task_struct *ext4_lazyinit_task;
2647
bfff6873
LC		 2648/*
2649 * This is the function where ext4lazyinit thread lives. It walks
2650 * through the request list searching for next scheduled filesystem.
2651 * When such a fs is found, run the lazy initialization request
2652 * (ext4_rn_li_request) and keep track of the time spend in this
2653 * function. Based on that time we compute next schedule time of
2654 * the request. When walking through the list is complete, compute
2655 * next waking time and put itself into sleep.
2656 */
2657static int ext4_lazyinit_thread(void *arg)
2658{
2659 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2660 struct list_head *pos, *n;
2661 struct ext4_li_request *elr;
4ed5c033 2662 unsigned long next_wakeup, cur;
bfff6873
LC		 2663
2664 BUG_ON(NULL == eli);
2665
bfff6873
LC		 2666cont_thread:
2667 while (true) {
2668 next_wakeup = MAX_JIFFY_OFFSET;
2669
2670 mutex_lock(&eli->li_list_mtx);
2671 if (list_empty(&eli->li_request_list)) {
2672 mutex_unlock(&eli->li_list_mtx);
2673 goto exit_thread;
2674 }
2675
2676 list_for_each_safe(pos, n, &eli->li_request_list) {
2677 elr = list_entry(pos, struct ext4_li_request,
2678 lr_request);
2679
b2c78cd0
TT		 2680 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2681 if (ext4_run_li_request(elr) != 0) {
2682 /* error, remove the lazy_init job */
2683 ext4_remove_li_request(elr);
2684 continue;
2685 }
bfff6873
LC		 2686 }
2687
2688 if (time_before(elr->lr_next_sched, next_wakeup))
2689 next_wakeup = elr->lr_next_sched;
2690 }
2691 mutex_unlock(&eli->li_list_mtx);
2692
a0acae0e 2693 try_to_freeze();
bfff6873 2694
4ed5c033
LC		 2695 cur = jiffies;
2696 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2697 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2698 cond_resched();
2699 continue;
2700 }
2701
4ed5c033
LC		 2702 schedule_timeout_interruptible(next_wakeup - cur);
2703
8f1f7453
ES		 2704 if (kthread_should_stop()) {
2705 ext4_clear_request_list();
2706 goto exit_thread;
2707 }
bfff6873
LC		 2708 }
2709
2710exit_thread:
2711 /*
2712 * It looks like the request list is empty, but we need
2713 * to check it under the li_list_mtx lock, to prevent any
2714 * additions into it, and of course we should lock ext4_li_mtx
2715 * to atomically free the list and ext4_li_info, because at
2716 * this point another ext4 filesystem could be registering
2717 * new one.
2718 */
2719 mutex_lock(&ext4_li_mtx);
2720 mutex_lock(&eli->li_list_mtx);
2721 if (!list_empty(&eli->li_request_list)) {
2722 mutex_unlock(&eli->li_list_mtx);
2723 mutex_unlock(&ext4_li_mtx);
2724 goto cont_thread;
2725 }
2726 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2727 kfree(ext4_li_info);
2728 ext4_li_info = NULL;
2729 mutex_unlock(&ext4_li_mtx);
2730
2731 return 0;
2732}
2733
2734static void ext4_clear_request_list(void)
2735{
2736 struct list_head *pos, *n;
2737 struct ext4_li_request *elr;
2738
2739 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2740 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2741 elr = list_entry(pos, struct ext4_li_request,
2742 lr_request);
2743 ext4_remove_li_request(elr);
2744 }
2745 mutex_unlock(&ext4_li_info->li_list_mtx);
2746}
2747
2748static int ext4_run_lazyinit_thread(void)
2749{
8f1f7453
ES		 2750 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2751 ext4_li_info, "ext4lazyinit");
2752 if (IS_ERR(ext4_lazyinit_task)) {
2753 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2754 ext4_clear_request_list();
bfff6873
LC		 2755 kfree(ext4_li_info);
2756 ext4_li_info = NULL;
92b97816 2757 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2758 "initialization thread\n",
2759 err);
2760 return err;
2761 }
2762 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2763 return 0;
2764}
2765
2766/*
2767 * Check whether it make sense to run itable init. thread or not.
2768 * If there is at least one uninitialized inode table, return
2769 * corresponding group number, else the loop goes through all
2770 * groups and return total number of groups.
2771 */
2772static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2773{
2774 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2775 struct ext4_group_desc *gdp = NULL;
2776
2777 for (group = 0; group < ngroups; group++) {
2778 gdp = ext4_get_group_desc(sb, group, NULL);
2779 if (!gdp)
2780 continue;
2781
2782 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2783 break;
2784 }
2785
2786 return group;
2787}
2788
2789static int ext4_li_info_new(void)
2790{
2791 struct ext4_lazy_init *eli = NULL;
2792
2793 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2794 if (!eli)
2795 return -ENOMEM;
2796
bfff6873
LC		 2797 INIT_LIST_HEAD(&eli->li_request_list);
2798 mutex_init(&eli->li_list_mtx);
2799
bfff6873
LC		 2800 eli->li_state |= EXT4_LAZYINIT_QUIT;
2801
2802 ext4_li_info = eli;
2803
2804 return 0;
2805}
2806
2807static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2808 ext4_group_t start)
2809{
2810 struct ext4_sb_info *sbi = EXT4_SB(sb);
2811 struct ext4_li_request *elr;
bfff6873
LC		 2812
2813 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2814 if (!elr)
2815 return NULL;
2816
2817 elr->lr_super = sb;
2818 elr->lr_sbi = sbi;
2819 elr->lr_next_group = start;
2820
2821 /*
2822 * Randomize first schedule time of the request to
2823 * spread the inode table initialization requests
2824 * better.
2825 */
dd1f723b
TT		 2826 elr->lr_next_sched = jiffies + (prandom_u32() %
2827 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 2828 return elr;
2829}
2830
7f511862
TT		 2831int ext4_register_li_request(struct super_block *sb,
2832 ext4_group_t first_not_zeroed)
bfff6873
LC		 2833{
2834 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 2835 struct ext4_li_request *elr = NULL;
bfff6873 2836 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2837 int ret = 0;
bfff6873 2838
7f511862 2839 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 2840 if (sbi->s_li_request != NULL) {
2841 /*
2842 * Reset timeout so it can be computed again, because
2843 * s_li_wait_mult might have changed.
2844 */
2845 sbi->s_li_request->lr_timeout = 0;
7f511862 2846 goto out;
51ce6511 2847 }
bfff6873
LC		 2848
2849 if (first_not_zeroed == ngroups ||
2850 (sb->s_flags & MS_RDONLY) ||
55ff3840 2851 !test_opt(sb, INIT_INODE_TABLE))
7f511862 2852 goto out;
bfff6873
LC		 2853
2854 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 2855 if (!elr) {
2856 ret = -ENOMEM;
2857 goto out;
2858 }
bfff6873
LC		 2859
2860 if (NULL == ext4_li_info) {
2861 ret = ext4_li_info_new();
2862 if (ret)
2863 goto out;
2864 }
2865
2866 mutex_lock(&ext4_li_info->li_list_mtx);
2867 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2868 mutex_unlock(&ext4_li_info->li_list_mtx);
2869
2870 sbi->s_li_request = elr;
46e4690b
TM		 2871 /*
2872 * set elr to NULL here since it has been inserted to
2873 * the request_list and the removal and free of it is
2874 * handled by ext4_clear_request_list from now on.
2875 */
2876 elr = NULL;
bfff6873
LC		 2877
2878 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2879 ret = ext4_run_lazyinit_thread();
2880 if (ret)
2881 goto out;
2882 }
bfff6873 2883out:
beed5ecb
NK		 2884 mutex_unlock(&ext4_li_mtx);
2885 if (ret)
bfff6873 2886 kfree(elr);
bfff6873
LC		 2887 return ret;
2888}
2889
2890/*
2891 * We do not need to lock anything since this is called on
2892 * module unload.
2893 */
2894static void ext4_destroy_lazyinit_thread(void)
2895{
2896 /*
2897 * If thread exited earlier
2898 * there's nothing to be done.
2899 */
8f1f7453 2900 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 2901 return;
2902
8f1f7453 2903 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 2904}
2905
25ed6e8a
DW		 2906static int set_journal_csum_feature_set(struct super_block *sb)
2907{
2908 int ret = 1;
2909 int compat, incompat;
2910 struct ext4_sb_info *sbi = EXT4_SB(sb);
2911
9aa5d32b 2912 if (ext4_has_metadata_csum(sb)) {
db9ee220 2913 /* journal checksum v3 */
25ed6e8a 2914 compat = 0;
db9ee220 2915 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 2916 } else {
2917 /* journal checksum v1 */
2918 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
2919 incompat = 0;
2920 }
2921
feb8c6d3
DW		 2922 jbd2_journal_clear_features(sbi->s_journal,
2923 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2924 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
2925 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 2926 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2927 ret = jbd2_journal_set_features(sbi->s_journal,
2928 compat, 0,
2929 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
2930 incompat);
2931 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2932 ret = jbd2_journal_set_features(sbi->s_journal,
2933 compat, 0,
2934 incompat);
2935 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2936 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2937 } else {
feb8c6d3
DW		 2938 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2939 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 2940 }
2941
2942 return ret;
2943}
2944
952fc18e
TT		 2945/*
2946 * Note: calculating the overhead so we can be compatible with
2947 * historical BSD practice is quite difficult in the face of
2948 * clusters/bigalloc. This is because multiple metadata blocks from
2949 * different block group can end up in the same allocation cluster.
2950 * Calculating the exact overhead in the face of clustered allocation
2951 * requires either O(all block bitmaps) in memory or O(number of block
2952 * groups**2) in time. We will still calculate the superblock for
2953 * older file systems --- and if we come across with a bigalloc file
2954 * system with zero in s_overhead_clusters the estimate will be close to
2955 * correct especially for very large cluster sizes --- but for newer
2956 * file systems, it's better to calculate this figure once at mkfs
2957 * time, and store it in the superblock. If the superblock value is
2958 * present (even for non-bigalloc file systems), we will use it.
2959 */
2960static int count_overhead(struct super_block *sb, ext4_group_t grp,
2961 char *buf)
2962{
2963 struct ext4_sb_info *sbi = EXT4_SB(sb);
2964 struct ext4_group_desc *gdp;
2965 ext4_fsblk_t first_block, last_block, b;
2966 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
2967 int s, j, count = 0;
2968
0548bbb8
TT		 2969 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
2970 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
2971 sbi->s_itb_per_group + 2);
2972
952fc18e
TT		 2973 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
2974 (grp * EXT4_BLOCKS_PER_GROUP(sb));
2975 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
2976 for (i = 0; i < ngroups; i++) {
2977 gdp = ext4_get_group_desc(sb, i, NULL);
2978 b = ext4_block_bitmap(sb, gdp);
2979 if (b >= first_block && b <= last_block) {
2980 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
2981 count++;
2982 }
2983 b = ext4_inode_bitmap(sb, gdp);
2984 if (b >= first_block && b <= last_block) {
2985 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
2986 count++;
2987 }
2988 b = ext4_inode_table(sb, gdp);
2989 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
2990 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
2991 int c = EXT4_B2C(sbi, b - first_block);
2992 ext4_set_bit(c, buf);
2993 count++;
2994 }
2995 if (i != grp)
2996 continue;
2997 s = 0;
2998 if (ext4_bg_has_super(sb, grp)) {
2999 ext4_set_bit(s++, buf);
3000 count++;
3001 }
3002 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3003 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3004 count++;
3005 }
3006 }
3007 if (!count)
3008 return 0;
3009 return EXT4_CLUSTERS_PER_GROUP(sb) -
3010 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3011}
3012
3013/*
3014 * Compute the overhead and stash it in sbi->s_overhead
3015 */
3016int ext4_calculate_overhead(struct super_block *sb)
3017{
3018 struct ext4_sb_info *sbi = EXT4_SB(sb);
3019 struct ext4_super_block *es = sbi->s_es;
3020 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3021 ext4_fsblk_t overhead = 0;
4fdb5543 3022 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3023
952fc18e
TT		 3024 if (!buf)
3025 return -ENOMEM;
3026
3027 /*
3028 * Compute the overhead (FS structures). This is constant
3029 * for a given filesystem unless the number of block groups
3030 * changes so we cache the previous value until it does.
3031 */
3032
3033 /*
3034 * All of the blocks before first_data_block are overhead
3035 */
3036 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3037
3038 /*
3039 * Add the overhead found in each block group
3040 */
3041 for (i = 0; i < ngroups; i++) {
3042 int blks;
3043
3044 blks = count_overhead(sb, i, buf);
3045 overhead += blks;
3046 if (blks)
3047 memset(buf, 0, PAGE_SIZE);
3048 cond_resched();
3049 }
b003b524
ES		 3050 /* Add the internal journal blocks as well */
3051 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3052 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3053
952fc18e
TT		 3054 sbi->s_overhead = overhead;
3055 smp_wmb();
3056 free_page((unsigned long) buf);
3057 return 0;
3058}
3059
b5799018 3060static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3061{
3062 ext4_fsblk_t resv_clusters;
b5799018 3063 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3064
30fac0f7
JK		 3065 /*
3066 * There's no need to reserve anything when we aren't using extents.
3067 * The space estimates are exact, there are no unwritten extents,
3068 * hole punching doesn't need new metadata... This is needed especially
3069 * to keep ext2/3 backward compatibility.
3070 */
3071 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
b5799018 3072 return;
27dd4385
LC		 3073 /*
3074 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3075 * This should cover the situations where we can not afford to run
3076 * out of space like for example punch hole, or converting
556615dc 3077 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3078 * allocation would require 1, or 2 blocks, higher numbers are
3079 * very rare.
3080 */
b5799018
TT		 3081 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3082 sbi->s_cluster_bits);
27dd4385
LC		 3083
3084 do_div(resv_clusters, 50);
3085 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3086
b5799018 3087 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3088}
3089
2b2d6d01 3090static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3091{
d4c402d9 3092 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3093 struct buffer_head *bh;
617ba13b
MC		 3094 struct ext4_super_block *es = NULL;
3095 struct ext4_sb_info *sbi;
3096 ext4_fsblk_t block;
3097 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3098 ext4_fsblk_t logical_sb_block;
ac27a0ec 3099 unsigned long offset = 0;
ac27a0ec
DK		 3100 unsigned long journal_devnum = 0;
3101 unsigned long def_mount_opts;
3102 struct inode *root;
0390131b 3103 const char *descr;
dcc7dae3 3104 int ret = -ENOMEM;
281b5995 3105 int blocksize, clustersize;
4ec11028
TT		 3106 unsigned int db_count;
3107 unsigned int i;
281b5995 3108 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3109 __u64 blocks_count;
07aa2ea1 3110 int err = 0;
b3881f74 3111 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3112 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3113
3114 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3115 if (!sbi)
dcc7dae3 3116 goto out_free_orig;
705895b6
PE		 3117
3118 sbi->s_blockgroup_lock =
3119 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3120 if (!sbi->s_blockgroup_lock) {
3121 kfree(sbi);
dcc7dae3 3122 goto out_free_orig;
705895b6 3123 }
ac27a0ec 3124 sb->s_fs_info = sbi;
2c0544b2 3125 sbi->s_sb = sb;
240799cd 3126 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3127 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3128 if (sb->s_bdev->bd_part)
3129 sbi->s_sectors_written_start =
3130 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3131
9f6200bb 3132 /* Cleanup superblock name */
ec3904dc 3133 strreplace(sb->s_id, '/', '!');
9f6200bb 3134
07aa2ea1 3135 /* -EINVAL is default */
dcc7dae3 3136 ret = -EINVAL;
617ba13b 3137 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3138 if (!blocksize) {
b31e1552 3139 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3140 goto out_fail;
3141 }
3142
3143 /*
617ba13b 3144 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3145 * block sizes. We need to calculate the offset from buffer start.
3146 */
617ba13b 3147 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3148 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3149 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3150 } else {
70bbb3e0 3151 logical_sb_block = sb_block;
ac27a0ec
DK		 3152 }
3153
a8ac900b 3154 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3155 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3156 goto out_fail;
3157 }
3158 /*
3159 * Note: s_es must be initialized as soon as possible because
617ba13b 3160 * some ext4 macro-instructions depend on its value
ac27a0ec 3161 */
2716b802 3162 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3163 sbi->s_es = es;
3164 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3165 if (sb->s_magic != EXT4_SUPER_MAGIC)
3166 goto cantfind_ext4;
afc32f7e 3167 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3168
feb0ab32
DW		 3169 /* Warn if metadata_csum and gdt_csum are both set. */
3170 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3171 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3172 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
363307e6 3173 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3174 "redundant flags; please run fsck.");
3175
d25425f8
DW		 3176 /* Check for a known checksum algorithm */
3177 if (!ext4_verify_csum_type(sb, es)) {
3178 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3179 "unknown checksum algorithm.");
3180 silent = 1;
3181 goto cantfind_ext4;
3182 }
3183
0441984a
DW		 3184 /* Load the checksum driver */
3185 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3186 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3187 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3188 if (IS_ERR(sbi->s_chksum_driver)) {
3189 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3190 ret = PTR_ERR(sbi->s_chksum_driver);
3191 sbi->s_chksum_driver = NULL;
3192 goto failed_mount;
3193 }
3194 }
3195
a9c47317
DW		 3196 /* Check superblock checksum */
3197 if (!ext4_superblock_csum_verify(sb, es)) {
3198 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3199 "invalid superblock checksum. Run e2fsck?");
3200 silent = 1;
6a797d27 3201 ret = -EFSBADCRC;
a9c47317
DW		 3202 goto cantfind_ext4;
3203 }
3204
3205 /* Precompute checksum seed for all metadata */
8c81bd8f
DW		 3206 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_CSUM_SEED))
3207 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3208 else if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3209 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3210 sizeof(es->s_uuid));
3211
ac27a0ec
DK		 3212 /* Set defaults before we parse the mount options */
3213 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3214 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3215 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3216 set_opt(sb, DEBUG);
87f26807 3217 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3218 set_opt(sb, GRPID);
617ba13b 3219 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3220 set_opt(sb, NO_UID32);
ea663336 3221 /* xattr user namespace & acls are now defaulted on */
ea663336 3222 set_opt(sb, XATTR_USER);
03010a33 3223#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3224 set_opt(sb, POSIX_ACL);
2e7842b8 3225#endif
98c1a759
DW		 3226 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3227 if (ext4_has_metadata_csum(sb))
3228 set_opt(sb, JOURNAL_CHECKSUM);
3229
617ba13b 3230 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3231 set_opt(sb, JOURNAL_DATA);
617ba13b 3232 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3233 set_opt(sb, ORDERED_DATA);
617ba13b 3234 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3235 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3236
3237 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3238 set_opt(sb, ERRORS_PANIC);
bb4f397a 3239 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3240 set_opt(sb, ERRORS_CONT);
bb4f397a 3241 else
fd8c37ec 3242 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3243 /* block_validity enabled by default; disable with noblock_validity */
3244 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3245 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3246 set_opt(sb, DISCARD);
ac27a0ec 3247
08cefc7a
EB		 3248 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3249 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3250 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3251 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3252 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3253
8b67f04a 3254 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3255 set_opt(sb, BARRIER);
ac27a0ec 3256
dd919b98
AK		 3257 /*
3258 * enable delayed allocation by default
3259 * Use -o nodelalloc to turn it off
3260 */
bc0b75f7 3261 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3262 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3263 set_opt(sb, DELALLOC);
dd919b98 3264
51ce6511
LC		 3265 /*
3266 * set default s_li_wait_mult for lazyinit, for the case there is
3267 * no mount option specified.
3268 */
3269 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3270
8b67f04a 3271 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3272 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3273 ext4_msg(sb, KERN_WARNING,
3274 "failed to parse options in superblock: %s",
3275 sbi->s_es->s_mount_opts);
3276 }
5a916be1 3277 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3278 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3279 &journal_ioprio, 0))
ac27a0ec
DK		 3280 goto failed_mount;
3281
56889787
TT		 3282 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3283 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3284 "with data=journal disables delayed "
3285 "allocation and O_DIRECT support!\n");
3286 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3287 ext4_msg(sb, KERN_ERR, "can't mount with "
3288 "both data=journal and delalloc");
3289 goto failed_mount;
3290 }
3291 if (test_opt(sb, DIOREAD_NOLOCK)) {
3292 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3293 "both data=journal and dioread_nolock");
56889787
TT		 3294 goto failed_mount;
3295 }
923ae0ff
RZ		 3296 if (test_opt(sb, DAX)) {
3297 ext4_msg(sb, KERN_ERR, "can't mount with "
3298 "both data=journal and dax");
3299 goto failed_mount;
3300 }
56889787
TT		 3301 if (test_opt(sb, DELALLOC))
3302 clear_opt(sb, DELALLOC);
001e4a87
TH		 3303 } else {
3304 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3305 }
3306
ac27a0ec 3307 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3308 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3309
617ba13b
MC		 3310 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3311 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3312 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3313 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3314 ext4_msg(sb, KERN_WARNING,
3315 "feature flags set on rev 0 fs, "
3316 "running e2fsck is recommended");
469108ff 3317
ed3654eb
TT		 3318 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3319 set_opt2(sb, HURD_COMPAT);
3320 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3321 EXT4_FEATURE_INCOMPAT_64BIT)) {
3322 ext4_msg(sb, KERN_ERR,
3323 "The Hurd can't support 64-bit file systems");
3324 goto failed_mount;
3325 }
3326 }
3327
2035e776
TT		 3328 if (IS_EXT2_SB(sb)) {
3329 if (ext2_feature_set_ok(sb))
3330 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3331 "using the ext4 subsystem");
3332 else {
3333 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3334 "to feature incompatibilities");
3335 goto failed_mount;
3336 }
3337 }
3338
3339 if (IS_EXT3_SB(sb)) {
3340 if (ext3_feature_set_ok(sb))
3341 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3342 "using the ext4 subsystem");
3343 else {
3344 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3345 "to feature incompatibilities");
3346 goto failed_mount;
3347 }
3348 }
3349
ac27a0ec
DK		 3350 /*
3351 * Check feature flags regardless of the revision level, since we
3352 * previously didn't change the revision level when setting the flags,
3353 * so there is a chance incompat flags are set on a rev 0 filesystem.
3354 */
a13fb1a4 3355 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3356 goto failed_mount;
a13fb1a4 3357
261cb20c 3358 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3359 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3360 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3361 ext4_msg(sb, KERN_ERR,
3362 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3363 goto failed_mount;
3364 }
3365
923ae0ff
RZ		 3366 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3367 if (blocksize != PAGE_SIZE) {
3368 ext4_msg(sb, KERN_ERR,
3369 "error: unsupported blocksize for dax");
3370 goto failed_mount;
3371 }
3372 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3373 ext4_msg(sb, KERN_ERR,
3374 "error: device does not support dax");
3375 goto failed_mount;
3376 }
3377 }
3378
6ddb2447
TT		 3379 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT) &&
3380 es->s_encryption_level) {
3381 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3382 es->s_encryption_level);
3383 goto failed_mount;
3384 }
3385
ac27a0ec 3386 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3387 /* Validate the filesystem blocksize */
3388 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3389 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3390 blocksize);
ac27a0ec
DK		 3391 goto failed_mount;
3392 }
3393
2b2d6d01 3394 brelse(bh);
70bbb3e0
AM		 3395 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3396 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3397 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3398 if (!bh) {
b31e1552
ES		 3399 ext4_msg(sb, KERN_ERR,
3400 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3401 goto failed_mount;
3402 }
2716b802 3403 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3404 sbi->s_es = es;
617ba13b 3405 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3406 ext4_msg(sb, KERN_ERR,
3407 "Magic mismatch, very weird!");
ac27a0ec
DK		 3408 goto failed_mount;
3409 }
3410 }
3411
a13fb1a4
ES		 3412 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3413 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3414 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3415 has_huge_files);
3416 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3417
617ba13b
MC		 3418 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3419 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3420 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3421 } else {
3422 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3423 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3424 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3425 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3426 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3427 ext4_msg(sb, KERN_ERR,
3428 "unsupported inode size: %d",
2b2d6d01 3429 sbi->s_inode_size);
ac27a0ec
DK		 3430 goto failed_mount;
3431 }
ef7f3835
KS		 3432 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3433 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3434 }
0b8e58a1 3435
0d1ee42f
AR		 3436 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3437 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3438 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3439 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3440 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3441 ext4_msg(sb, KERN_ERR,
3442 "unsupported descriptor size %lu",
0d1ee42f
AR		 3443 sbi->s_desc_size);
3444 goto failed_mount;
3445 }
3446 } else
3447 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3448
ac27a0ec 3449 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3450 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3451 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3452 goto cantfind_ext4;
0b8e58a1 3453
617ba13b 3454 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3455 if (sbi->s_inodes_per_block == 0)
617ba13b 3456 goto cantfind_ext4;
ac27a0ec
DK		 3457 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3458 sbi->s_inodes_per_block;
0d1ee42f 3459 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3460 sbi->s_sbh = bh;
3461 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3462 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3463 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3464
2b2d6d01 3465 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3466 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3467 sbi->s_def_hash_version = es->s_def_hash_version;
23301410
TT		 3468 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3469 i = le32_to_cpu(es->s_flags);
3470 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3471 sbi->s_hash_unsigned = 3;
3472 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3473#ifdef __CHAR_UNSIGNED__
23301410
TT		 3474 if (!(sb->s_flags & MS_RDONLY))
3475 es->s_flags |=
3476 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3477 sbi->s_hash_unsigned = 3;
f99b2589 3478#else
23301410
TT		 3479 if (!(sb->s_flags & MS_RDONLY))
3480 es->s_flags |=
3481 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3482#endif
23301410 3483 }
f99b2589 3484 }
ac27a0ec 3485
281b5995
TT		 3486 /* Handle clustersize */
3487 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3488 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3489 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3490 if (has_bigalloc) {
3491 if (clustersize < blocksize) {
3492 ext4_msg(sb, KERN_ERR,
3493 "cluster size (%d) smaller than "
3494 "block size (%d)", clustersize, blocksize);
3495 goto failed_mount;
3496 }
3497 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3498 le32_to_cpu(es->s_log_block_size);
3499 sbi->s_clusters_per_group =
3500 le32_to_cpu(es->s_clusters_per_group);
3501 if (sbi->s_clusters_per_group > blocksize * 8) {
3502 ext4_msg(sb, KERN_ERR,
3503 "#clusters per group too big: %lu",
3504 sbi->s_clusters_per_group);
3505 goto failed_mount;
3506 }
3507 if (sbi->s_blocks_per_group !=
3508 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3509 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3510 "clusters per group (%lu) inconsistent",
3511 sbi->s_blocks_per_group,
3512 sbi->s_clusters_per_group);
3513 goto failed_mount;
3514 }
3515 } else {
3516 if (clustersize != blocksize) {
3517 ext4_warning(sb, "fragment/cluster size (%d) != "
3518 "block size (%d)", clustersize,
3519 blocksize);
3520 clustersize = blocksize;
3521 }
3522 if (sbi->s_blocks_per_group > blocksize * 8) {
3523 ext4_msg(sb, KERN_ERR,
3524 "#blocks per group too big: %lu",
3525 sbi->s_blocks_per_group);
3526 goto failed_mount;
3527 }
3528 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3529 sbi->s_cluster_bits = 0;
ac27a0ec 3530 }
281b5995
TT		 3531 sbi->s_cluster_ratio = clustersize / blocksize;
3532
ac27a0ec 3533 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3534 ext4_msg(sb, KERN_ERR,
3535 "#inodes per group too big: %lu",
2b2d6d01 3536 sbi->s_inodes_per_group);
ac27a0ec
DK		 3537 goto failed_mount;
3538 }
3539
960fd856
TT		 3540 /* Do we have standard group size of clustersize * 8 blocks ? */
3541 if (sbi->s_blocks_per_group == clustersize << 3)
3542 set_opt2(sb, STD_GROUP_SIZE);
3543
bf43d84b
ES		 3544 /*
3545 * Test whether we have more sectors than will fit in sector_t,
3546 * and whether the max offset is addressable by the page cache.
3547 */
5a9ae68a 3548 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3549 ext4_blocks_count(es));
5a9ae68a 3550 if (err) {
b31e1552 3551 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3552 " too large to mount safely on this system");
ac27a0ec 3553 if (sizeof(sector_t) < 8)
90c699a9 3554 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3555 goto failed_mount;
3556 }
3557
617ba13b
MC		 3558 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3559 goto cantfind_ext4;
e7c95593 3560
0f2ddca6
FTN		 3561 /* check blocks count against device size */
3562 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3563 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3564 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3565 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3566 ext4_blocks_count(es), blocks_count);
3567 goto failed_mount;
3568 }
3569
0b8e58a1
AD		 3570 /*
3571 * It makes no sense for the first data block to be beyond the end
3572 * of the filesystem.
3573 */
3574 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3575 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3576 "block %u is beyond end of filesystem (%llu)",
3577 le32_to_cpu(es->s_first_data_block),
3578 ext4_blocks_count(es));
e7c95593
ES		 3579 goto failed_mount;
3580 }
bd81d8ee
LV		 3581 blocks_count = (ext4_blocks_count(es) -
3582 le32_to_cpu(es->s_first_data_block) +
3583 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3584 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3585 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3586 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3587 "(block count %llu, first data block %u, "
b31e1552 3588 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3589 ext4_blocks_count(es),
3590 le32_to_cpu(es->s_first_data_block),
3591 EXT4_BLOCKS_PER_GROUP(sb));
3592 goto failed_mount;
3593 }
bd81d8ee 3594 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3595 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3596 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3597 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3598 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3599 sbi->s_group_desc = ext4_kvmalloc(db_count *
3600 sizeof(struct buffer_head *),
3601 GFP_KERNEL);
ac27a0ec 3602 if (sbi->s_group_desc == NULL) {
b31e1552 3603 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3604 ret = -ENOMEM;
ac27a0ec
DK		 3605 goto failed_mount;
3606 }
3607
705895b6 3608 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3609
3610 for (i = 0; i < db_count; i++) {
70bbb3e0 3611 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3612 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3613 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3614 ext4_msg(sb, KERN_ERR,
3615 "can't read group descriptor %d", i);
ac27a0ec
DK		 3616 db_count = i;
3617 goto failed_mount2;
3618 }
3619 }
bfff6873 3620 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3621 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 3622 ret = -EFSCORRUPTED;
f9ae9cf5 3623 goto failed_mount2;
ac27a0ec 3624 }
772cb7c8 3625
f9ae9cf5 3626 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3627 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3628 spin_lock_init(&sbi->s_next_gen_lock);
3629
04ecddb7
JM		 3630 setup_timer(&sbi->s_err_report, print_daily_error_info,
3631 (unsigned long) sb);
04496411 3632
a75ae78f 3633 /* Register extent status tree shrinker */
eb68d0e2 3634 if (ext4_es_register_shrinker(sbi))
ce7e010a 3635 goto failed_mount3;
ce7e010a 3636
c9de560d 3637 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3638 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3639
f9ae9cf5
TT		 3640 /*
3641 * set up enough so that it can read an inode
3642 */
f6e63f90 3643 sb->s_op = &ext4_sops;
617ba13b
MC		 3644 sb->s_export_op = &ext4_export_ops;
3645 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3646#ifdef CONFIG_QUOTA
617ba13b 3647 sb->dq_op = &ext4_quota_operations;
262b4662 3648 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
1fa5efe3 3649 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3650 else
3651 sb->s_qcop = &ext4_qctl_operations;
96c7e0d9 3652 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
ac27a0ec 3653#endif
f2fa2ffc
AK		 3654 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3655
ac27a0ec 3656 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3657 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3658
3659 sb->s_root = NULL;
3660
3661 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3662 EXT4_HAS_INCOMPAT_FEATURE(sb,
3663 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3664
c5e06d10
JL		 3665 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3666 !(sb->s_flags & MS_RDONLY))
3667 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3668 goto failed_mount3a;
c5e06d10 3669
ac27a0ec
DK		 3670 /*
3671 * The first inode we look at is the journal inode. Don't try
3672 * root first: it may be modified in the journal!
3673 */
3674 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3675 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3676 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 3677 goto failed_mount3a;
0390131b
FM		 3678 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3679 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3680 ext4_msg(sb, KERN_ERR, "required journal recovery "
3681 "suppressed and not mounted read-only");
744692dc 3682 goto failed_mount_wq;
ac27a0ec 3683 } else {
fd8c37ec 3684 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3685 sbi->s_journal = NULL;
3686 needs_recovery = 0;
3687 goto no_journal;
ac27a0ec
DK		 3688 }
3689
f32aaf2d 3690 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3691 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3692 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3693 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3694 goto failed_mount_wq;
eb40a09c
JS		 3695 }
3696
25ed6e8a
DW		 3697 if (!set_journal_csum_feature_set(sb)) {
3698 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3699 "feature set");
3700 goto failed_mount_wq;
d4da6c9c 3701 }
818d276c 3702
ac27a0ec
DK		 3703 /* We have now updated the journal if required, so we can
3704 * validate the data journaling mode. */
3705 switch (test_opt(sb, DATA_FLAGS)) {
3706 case 0:
3707 /* No mode set, assume a default based on the journal
63f57933
AM		 3708 * capabilities: ORDERED_DATA if the journal can
3709 * cope, else JOURNAL_DATA
3710 */
dab291af
MC		 3711 if (jbd2_journal_check_available_features
3712 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3713 set_opt(sb, ORDERED_DATA);
ac27a0ec 3714 else
fd8c37ec 3715 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3716 break;
3717
617ba13b
MC		 3718 case EXT4_MOUNT_ORDERED_DATA:
3719 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3720 if (!jbd2_journal_check_available_features
3721 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3722 ext4_msg(sb, KERN_ERR, "Journal does not support "
3723 "requested data journaling mode");
744692dc 3724 goto failed_mount_wq;
ac27a0ec
DK		 3725 }
3726 default:
3727 break;
3728 }
b3881f74 3729 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3730
18aadd47
BJ		 3731 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3732
ce7e010a 3733no_journal:
9c191f70
M		 3734 if (ext4_mballoc_ready) {
3735 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
3736 if (!sbi->s_mb_cache) {
3737 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3738 goto failed_mount_wq;
3739 }
3740 }
3741
1cb767cd
TT		 3742 if ((DUMMY_ENCRYPTION_ENABLED(sbi) ||
3743 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) &&
3744 (blocksize != PAGE_CACHE_SIZE)) {
3745 ext4_msg(sb, KERN_ERR,
3746 "Unsupported blocksize for fs encryption");
3747 goto failed_mount_wq;
3748 }
3749
3750 if (DUMMY_ENCRYPTION_ENABLED(sbi) &&
6ddb2447
TT		 3751 !(sb->s_flags & MS_RDONLY) &&
3752 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) {
3753 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
3754 ext4_commit_super(sb, 1);
3755 }
3756
952fc18e
TT		 3757 /*
3758 * Get the # of file system overhead blocks from the
3759 * superblock if present.
3760 */
3761 if (es->s_overhead_clusters)
3762 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3763 else {
07aa2ea1
LC		 3764 err = ext4_calculate_overhead(sb);
3765 if (err)
952fc18e
TT		 3766 goto failed_mount_wq;
3767 }
3768
fd89d5f2
TH		 3769 /*
3770 * The maximum number of concurrent works can be high and
3771 * concurrency isn't really necessary. Limit it to 1.
3772 */
2e8fa54e
JK		 3773 EXT4_SB(sb)->rsv_conversion_wq =
3774 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3775 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3776 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 3777 ret = -ENOMEM;
2e8fa54e
JK		 3778 goto failed_mount4;
3779 }
3780
ac27a0ec 3781 /*
dab291af 3782 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3783 * so we can safely mount the rest of the filesystem now.
3784 */
3785
1d1fe1ee
DH		 3786 root = ext4_iget(sb, EXT4_ROOT_INO);
3787 if (IS_ERR(root)) {
b31e1552 3788 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3789 ret = PTR_ERR(root);
32a9bb57 3790 root = NULL;
ac27a0ec
DK		 3791 goto failed_mount4;
3792 }
3793 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3794 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3795 iput(root);
ac27a0ec
DK		 3796 goto failed_mount4;
3797 }
48fde701 3798 sb->s_root = d_make_root(root);
1d1fe1ee 3799 if (!sb->s_root) {
b31e1552 3800 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3801 ret = -ENOMEM;
3802 goto failed_mount4;
3803 }
ac27a0ec 3804
7e84b621
ES		 3805 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3806 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3807
3808 /* determine the minimum size of new large inodes, if present */
3809 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3810 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3811 EXT4_GOOD_OLD_INODE_SIZE;
3812 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3813 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3814 if (sbi->s_want_extra_isize <
3815 le16_to_cpu(es->s_want_extra_isize))
3816 sbi->s_want_extra_isize =
3817 le16_to_cpu(es->s_want_extra_isize);
3818 if (sbi->s_want_extra_isize <
3819 le16_to_cpu(es->s_min_extra_isize))
3820 sbi->s_want_extra_isize =
3821 le16_to_cpu(es->s_min_extra_isize);
3822 }
3823 }
3824 /* Check if enough inode space is available */
3825 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3826 sbi->s_inode_size) {
3827 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3828 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3829 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3830 "available");
ef7f3835
KS		 3831 }
3832
b5799018 3833 ext4_set_resv_clusters(sb);
27dd4385 3834
6fd058f7
TT		 3835 err = ext4_setup_system_zone(sb);
3836 if (err) {
b31e1552 3837 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3838 "zone (%d)", err);
f9ae9cf5
TT		 3839 goto failed_mount4a;
3840 }
3841
3842 ext4_ext_init(sb);
3843 err = ext4_mb_init(sb);
3844 if (err) {
3845 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3846 err);
dcf2d804 3847 goto failed_mount5;
c2774d84
AK		 3848 }
3849
d5e03cbb
TT		 3850 block = ext4_count_free_clusters(sb);
3851 ext4_free_blocks_count_set(sbi->s_es,
3852 EXT4_C2B(sbi, block));
908c7f19
TH		 3853 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
3854 GFP_KERNEL);
d5e03cbb
TT		 3855 if (!err) {
3856 unsigned long freei = ext4_count_free_inodes(sb);
3857 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 3858 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
3859 GFP_KERNEL);
d5e03cbb
TT		 3860 }
3861 if (!err)
3862 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 3863 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 3864 if (!err)
908c7f19
TH		 3865 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
3866 GFP_KERNEL);
d5e03cbb
TT		 3867 if (err) {
3868 ext4_msg(sb, KERN_ERR, "insufficient memory");
3869 goto failed_mount6;
3870 }
3871
3872 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3873 if (!ext4_fill_flex_info(sb)) {
3874 ext4_msg(sb, KERN_ERR,
3875 "unable to initialize "
3876 "flex_bg meta info!");
3877 goto failed_mount6;
3878 }
3879
bfff6873
LC		 3880 err = ext4_register_li_request(sb, first_not_zeroed);
3881 if (err)
dcf2d804 3882 goto failed_mount6;
bfff6873 3883
b5799018 3884 err = ext4_register_sysfs(sb);
dcf2d804
TM		 3885 if (err)
3886 goto failed_mount7;
3197ebdb 3887
9b2ff357
JK		 3888#ifdef CONFIG_QUOTA
3889 /* Enable quota usage during mount. */
3890 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
3891 !(sb->s_flags & MS_RDONLY)) {
3892 err = ext4_enable_quotas(sb);
3893 if (err)
3894 goto failed_mount8;
3895 }
3896#endif /* CONFIG_QUOTA */
3897
617ba13b
MC		 3898 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3899 ext4_orphan_cleanup(sb, es);
3900 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3901 if (needs_recovery) {
b31e1552 3902 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3903 ext4_mark_recovery_complete(sb, es);
3904 }
3905 if (EXT4_SB(sb)->s_journal) {
3906 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3907 descr = " journalled data mode";
3908 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3909 descr = " ordered data mode";
3910 else
3911 descr = " writeback data mode";
3912 } else
3913 descr = "out journal";
3914
79add3a3
LC		 3915 if (test_opt(sb, DISCARD)) {
3916 struct request_queue *q = bdev_get_queue(sb->s_bdev);
3917 if (!blk_queue_discard(q))
3918 ext4_msg(sb, KERN_WARNING,
3919 "mounting with \"discard\" option, but "
3920 "the device does not support discard");
3921 }
3922
e294a537
TT		 3923 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
3924 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3925 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3926 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3927
66e61a9e
TT		 3928 if (es->s_error_count)
3929 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3930
efbed4dc
TT		 3931 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
3932 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
3933 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
3934 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
3935
d4c402d9 3936 kfree(orig_data);
ac27a0ec
DK		 3937 return 0;
3938
617ba13b 3939cantfind_ext4:
ac27a0ec 3940 if (!silent)
b31e1552 3941 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3942 goto failed_mount;
3943
72ba7450
TT		 3944#ifdef CONFIG_QUOTA
3945failed_mount8:
ebd173be 3946 ext4_unregister_sysfs(sb);
72ba7450 3947#endif
dcf2d804
TM		 3948failed_mount7:
3949 ext4_unregister_li_request(sb);
3950failed_mount6:
f9ae9cf5 3951 ext4_mb_release(sb);
d5e03cbb 3952 if (sbi->s_flex_groups)
b93b41d4 3953 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 3954 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3955 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3956 percpu_counter_destroy(&sbi->s_dirs_counter);
3957 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 3958failed_mount5:
f9ae9cf5
TT		 3959 ext4_ext_release(sb);
3960 ext4_release_system_zone(sb);
3961failed_mount4a:
94bf608a 3962 dput(sb->s_root);
32a9bb57 3963 sb->s_root = NULL;
94bf608a 3964failed_mount4:
b31e1552 3965 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 3966 if (EXT4_SB(sb)->rsv_conversion_wq)
3967 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 3968failed_mount_wq:
0390131b
FM		 3969 if (sbi->s_journal) {
3970 jbd2_journal_destroy(sbi->s_journal);
3971 sbi->s_journal = NULL;
3972 }
50460fe8 3973failed_mount3a:
d3922a77 3974 ext4_es_unregister_shrinker(sbi);
eb68d0e2 3975failed_mount3:
9105bb14 3976 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 3977 if (sbi->s_mmp_tsk)
3978 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 3979failed_mount2:
3980 for (i = 0; i < db_count; i++)
3981 brelse(sbi->s_group_desc[i]);
b93b41d4 3982 kvfree(sbi->s_group_desc);
ac27a0ec 3983failed_mount:
0441984a
DW		 3984 if (sbi->s_chksum_driver)
3985 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 3986#ifdef CONFIG_QUOTA
a2d4a646 3987 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 3988 kfree(sbi->s_qf_names[i]);
3989#endif
617ba13b 3990 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3991 brelse(bh);
3992out_fail:
3993 sb->s_fs_info = NULL;
f6830165 3994 kfree(sbi->s_blockgroup_lock);
ac27a0ec 3995 kfree(sbi);
dcc7dae3 3996out_free_orig:
d4c402d9 3997 kfree(orig_data);
07aa2ea1 3998 return err ? err : ret;
ac27a0ec
DK		 3999}
4000
4001/*
4002 * Setup any per-fs journal parameters now. We'll do this both on
4003 * initial mount, once the journal has been initialised but before we've
4004 * done any recovery; and again on any subsequent remount.
4005 */
617ba13b 4006static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4007{
617ba13b 4008 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4009
30773840
TT		 4010 journal->j_commit_interval = sbi->s_commit_interval;
4011 journal->j_min_batch_time = sbi->s_min_batch_time;
4012 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4013
a931da6a 4014 write_lock(&journal->j_state_lock);
ac27a0ec 4015 if (test_opt(sb, BARRIER))
dab291af 4016 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4017 else
dab291af 4018 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4019 if (test_opt(sb, DATA_ERR_ABORT))
4020 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4021 else
4022 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4023 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4024}
4025
617ba13b 4026static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4027 unsigned int journal_inum)
4028{
4029 struct inode *journal_inode;
4030 journal_t *journal;
4031
0390131b
FM		 4032 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4033
ac27a0ec
DK		 4034 /* First, test for the existence of a valid inode on disk. Bad
4035 * things happen if we iget() an unused inode, as the subsequent
4036 * iput() will try to delete it. */
4037
1d1fe1ee
DH		 4038 journal_inode = ext4_iget(sb, journal_inum);
4039 if (IS_ERR(journal_inode)) {
b31e1552 4040 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4041 return NULL;
4042 }
4043 if (!journal_inode->i_nlink) {
4044 make_bad_inode(journal_inode);
4045 iput(journal_inode);
b31e1552 4046 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4047 return NULL;
4048 }
4049
e5f8eab8 4050 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4051 journal_inode, journal_inode->i_size);
1d1fe1ee 4052 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4053 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4054 iput(journal_inode);
4055 return NULL;
4056 }
4057
dab291af 4058 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4059 if (!journal) {
b31e1552 4060 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4061 iput(journal_inode);
4062 return NULL;
4063 }
4064 journal->j_private = sb;
617ba13b 4065 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4066 return journal;
4067}
4068
617ba13b 4069static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4070 dev_t j_dev)
4071{
2b2d6d01 4072 struct buffer_head *bh;
ac27a0ec 4073 journal_t *journal;
617ba13b
MC		 4074 ext4_fsblk_t start;
4075 ext4_fsblk_t len;
ac27a0ec 4076 int hblock, blocksize;
617ba13b 4077 ext4_fsblk_t sb_block;
ac27a0ec 4078 unsigned long offset;
2b2d6d01 4079 struct ext4_super_block *es;
ac27a0ec
DK		 4080 struct block_device *bdev;
4081
0390131b
FM		 4082 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4083
b31e1552 4084 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4085 if (bdev == NULL)
4086 return NULL;
4087
ac27a0ec 4088 blocksize = sb->s_blocksize;
e1defc4f 4089 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4090 if (blocksize < hblock) {
b31e1552
ES		 4091 ext4_msg(sb, KERN_ERR,
4092 "blocksize too small for journal device");
ac27a0ec
DK		 4093 goto out_bdev;
4094 }
4095
617ba13b
MC		 4096 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4097 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4098 set_blocksize(bdev, blocksize);
4099 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4100 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4101 "external journal");
ac27a0ec
DK		 4102 goto out_bdev;
4103 }
4104
2716b802 4105 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4106 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4107 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4108 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4109 ext4_msg(sb, KERN_ERR, "external journal has "
4110 "bad superblock");
ac27a0ec
DK		 4111 brelse(bh);
4112 goto out_bdev;
4113 }
4114
df4763be
DW		 4115 if ((le32_to_cpu(es->s_feature_ro_compat) &
4116 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4117 es->s_checksum != ext4_superblock_csum(sb, es)) {
4118 ext4_msg(sb, KERN_ERR, "external journal has "
4119 "corrupt superblock");
4120 brelse(bh);
4121 goto out_bdev;
4122 }
4123
617ba13b 4124 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4125 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4126 brelse(bh);
4127 goto out_bdev;
4128 }
4129
bd81d8ee 4130 len = ext4_blocks_count(es);
ac27a0ec
DK		 4131 start = sb_block + 1;
4132 brelse(bh); /* we're done with the superblock */
4133
dab291af 4134 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4135 start, len, blocksize);
4136 if (!journal) {
b31e1552 4137 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4138 goto out_bdev;
4139 }
4140 journal->j_private = sb;
9f203507 4141 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4142 wait_on_buffer(journal->j_sb_buffer);
4143 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4144 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4145 goto out_journal;
4146 }
4147 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4148 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4149 "user (unsupported) - %d",
ac27a0ec
DK		 4150 be32_to_cpu(journal->j_superblock->s_nr_users));
4151 goto out_journal;
4152 }
617ba13b
MC		 4153 EXT4_SB(sb)->journal_bdev = bdev;
4154 ext4_init_journal_params(sb, journal);
ac27a0ec 4155 return journal;
0b8e58a1 4156
ac27a0ec 4157out_journal:
dab291af 4158 jbd2_journal_destroy(journal);
ac27a0ec 4159out_bdev:
617ba13b 4160 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4161 return NULL;
4162}
4163
617ba13b
MC		 4164static int ext4_load_journal(struct super_block *sb,
4165 struct ext4_super_block *es,
ac27a0ec
DK		 4166 unsigned long journal_devnum)
4167{
4168 journal_t *journal;
4169 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4170 dev_t journal_dev;
4171 int err = 0;
4172 int really_read_only;
4173
0390131b
FM		 4174 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4175
ac27a0ec
DK		 4176 if (journal_devnum &&
4177 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4178 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4179 "numbers have changed");
ac27a0ec
DK		 4180 journal_dev = new_decode_dev(journal_devnum);
4181 } else
4182 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4183
4184 really_read_only = bdev_read_only(sb->s_bdev);
4185
4186 /*
4187 * Are we loading a blank journal or performing recovery after a
4188 * crash? For recovery, we need to check in advance whether we
4189 * can get read-write access to the device.
4190 */
617ba13b 4191 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4192 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4193 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4194 "required on readonly filesystem");
ac27a0ec 4195 if (really_read_only) {
b31e1552
ES		 4196 ext4_msg(sb, KERN_ERR, "write access "
4197 "unavailable, cannot proceed");
ac27a0ec
DK		 4198 return -EROFS;
4199 }
b31e1552
ES		 4200 ext4_msg(sb, KERN_INFO, "write access will "
4201 "be enabled during recovery");
ac27a0ec
DK		 4202 }
4203 }
4204
4205 if (journal_inum && journal_dev) {
b31e1552
ES		 4206 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4207 "and inode journals!");
ac27a0ec
DK		 4208 return -EINVAL;
4209 }
4210
4211 if (journal_inum) {
617ba13b 4212 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4213 return -EINVAL;
4214 } else {
617ba13b 4215 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4216 return -EINVAL;
4217 }
4218
90576c0b 4219 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4220 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4221
617ba13b 4222 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4223 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4224 if (!err) {
4225 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4226 if (save)
4227 memcpy(save, ((char *) es) +
4228 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4229 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4230 if (save)
4231 memcpy(((char *) es) + EXT4_S_ERR_START,
4232 save, EXT4_S_ERR_LEN);
4233 kfree(save);
4234 }
ac27a0ec
DK		 4235
4236 if (err) {
b31e1552 4237 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4238 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4239 return err;
4240 }
4241
617ba13b
MC		 4242 EXT4_SB(sb)->s_journal = journal;
4243 ext4_clear_journal_err(sb, es);
ac27a0ec 4244
c41303ce 4245 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4246 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4247 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4248
4249 /* Make sure we flush the recovery flag to disk. */
e2d67052 4250 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4251 }
4252
4253 return 0;
4254}
4255
e2d67052 4256static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4257{
e2d67052 4258 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4259 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4260 int error = 0;
ac27a0ec 4261
bdfe0cbd 4262 if (!sbh || block_device_ejected(sb))
c4be0c1d 4263 return error;
914258bf
TT		 4264 if (buffer_write_io_error(sbh)) {
4265 /*
4266 * Oh, dear. A previous attempt to write the
4267 * superblock failed. This could happen because the
4268 * USB device was yanked out. Or it could happen to
4269 * be a transient write error and maybe the block will
4270 * be remapped. Nothing we can do but to retry the
4271 * write and hope for the best.
4272 */
b31e1552
ES		 4273 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4274 "superblock detected");
914258bf
TT		 4275 clear_buffer_write_io_error(sbh);
4276 set_buffer_uptodate(sbh);
4277 }
71290b36
TT		 4278 /*
4279 * If the file system is mounted read-only, don't update the
4280 * superblock write time. This avoids updating the superblock
4281 * write time when we are mounting the root file system
4282 * read/only but we need to replay the journal; at that point,
4283 * for people who are east of GMT and who make their clock
4284 * tick in localtime for Windows bug-for-bug compatibility,
4285 * the clock is set in the future, and this will cause e2fsck
4286 * to complain and force a full file system check.
4287 */
4288 if (!(sb->s_flags & MS_RDONLY))
4289 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4290 if (sb->s_bdev->bd_part)
4291 es->s_kbytes_written =
4292 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4293 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4294 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4295 else
4296 es->s_kbytes_written =
4297 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4298 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4299 ext4_free_blocks_count_set(es,
57042651
TT		 4300 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4301 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4302 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4303 es->s_free_inodes_count =
4304 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4305 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4306 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4307 ext4_superblock_csum_set(sb);
ac27a0ec 4308 mark_buffer_dirty(sbh);
914258bf 4309 if (sync) {
564bc402
DJ		 4310 error = __sync_dirty_buffer(sbh,
4311 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
c4be0c1d
TS		 4312 if (error)
4313 return error;
4314
4315 error = buffer_write_io_error(sbh);
4316 if (error) {
b31e1552
ES		 4317 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4318 "superblock");
914258bf
TT		 4319 clear_buffer_write_io_error(sbh);
4320 set_buffer_uptodate(sbh);
4321 }
4322 }
c4be0c1d 4323 return error;
ac27a0ec
DK		 4324}
4325
ac27a0ec
DK		 4326/*
4327 * Have we just finished recovery? If so, and if we are mounting (or
4328 * remounting) the filesystem readonly, then we will end up with a
4329 * consistent fs on disk. Record that fact.
4330 */
2b2d6d01
TT		 4331static void ext4_mark_recovery_complete(struct super_block *sb,
4332 struct ext4_super_block *es)
ac27a0ec 4333{
617ba13b 4334 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4335
0390131b
FM		 4336 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4337 BUG_ON(journal != NULL);
4338 return;
4339 }
dab291af 4340 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4341 if (jbd2_journal_flush(journal) < 0)
4342 goto out;
4343
617ba13b 4344 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4345 sb->s_flags & MS_RDONLY) {
617ba13b 4346 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4347 ext4_commit_super(sb, 1);
ac27a0ec 4348 }
7ffe1ea8
HK		 4349
4350out:
dab291af 4351 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4352}
4353
4354/*
4355 * If we are mounting (or read-write remounting) a filesystem whose journal
4356 * has recorded an error from a previous lifetime, move that error to the
4357 * main filesystem now.
4358 */
2b2d6d01
TT		 4359static void ext4_clear_journal_err(struct super_block *sb,
4360 struct ext4_super_block *es)
ac27a0ec
DK		 4361{
4362 journal_t *journal;
4363 int j_errno;
4364 const char *errstr;
4365
0390131b
FM		 4366 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4367
617ba13b 4368 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4369
4370 /*
4371 * Now check for any error status which may have been recorded in the
617ba13b 4372 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4373 */
4374
dab291af 4375 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4376 if (j_errno) {
4377 char nbuf[16];
4378
617ba13b 4379 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4380 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4381 "from previous mount: %s", errstr);
12062ddd 4382 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4383
617ba13b
MC		 4384 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4385 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4386 ext4_commit_super(sb, 1);
ac27a0ec 4387
dab291af 4388 jbd2_journal_clear_err(journal);
d796c52e 4389 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4390 }
4391}
4392
4393/*
4394 * Force the running and committing transactions to commit,
4395 * and wait on the commit.
4396 */
617ba13b 4397int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4398{
4399 journal_t *journal;
ac27a0ec
DK		 4400
4401 if (sb->s_flags & MS_RDONLY)
4402 return 0;
4403
617ba13b 4404 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4405 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4406}
4407
617ba13b 4408static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4409{
14ce0cb4 4410 int ret = 0;
9eddacf9 4411 tid_t target;
06a407f1 4412 bool needs_barrier = false;
8d5d02e6 4413 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4414
9bffad1e 4415 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4416 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4417 /*
4418 * Writeback quota in non-journalled quota case - journalled quota has
4419 * no dirty dquots
4420 */
4421 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4422 /*
4423 * Data writeback is possible w/o journal transaction, so barrier must
4424 * being sent at the end of the function. But we can skip it if
4425 * transaction_commit will do it for us.
4426 */
bda32530
TT		 4427 if (sbi->s_journal) {
4428 target = jbd2_get_latest_transaction(sbi->s_journal);
4429 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4430 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4431 needs_barrier = true;
4432
4433 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4434 if (wait)
4435 ret = jbd2_log_wait_commit(sbi->s_journal,
4436 target);
4437 }
4438 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4439 needs_barrier = true;
06a407f1
DM		 4440 if (needs_barrier) {
4441 int err;
4442 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4443 if (!ret)
4444 ret = err;
0390131b 4445 }
06a407f1
DM		 4446
4447 return ret;
4448}
4449
ac27a0ec
DK		 4450/*
4451 * LVM calls this function before a (read-only) snapshot is created. This
4452 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4453 *
4454 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4455 * state independently. It relies on upper layer to stop all data & metadata
4456 * modifications.
ac27a0ec 4457 */
c4be0c1d 4458static int ext4_freeze(struct super_block *sb)
ac27a0ec 4459{
c4be0c1d
TS		 4460 int error = 0;
4461 journal_t *journal;
ac27a0ec 4462
9ca92389
TT		 4463 if (sb->s_flags & MS_RDONLY)
4464 return 0;
ac27a0ec 4465
9ca92389 4466 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4467
bb044576
TT		 4468 if (journal) {
4469 /* Now we set up the journal barrier. */
4470 jbd2_journal_lock_updates(journal);
ac27a0ec 4471
bb044576
TT		 4472 /*
4473 * Don't clear the needs_recovery flag if we failed to
4474 * flush the journal.
4475 */
4476 error = jbd2_journal_flush(journal);
4477 if (error < 0)
4478 goto out;
c642dc9e
ES		 4479
4480 /* Journal blocked and flushed, clear needs_recovery flag. */
4481 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
bb044576 4482 }
9ca92389 4483
9ca92389 4484 error = ext4_commit_super(sb, 1);
6b0310fb 4485out:
bb044576
TT		 4486 if (journal)
4487 /* we rely on upper layer to stop further updates */
4488 jbd2_journal_unlock_updates(journal);
6b0310fb 4489 return error;
ac27a0ec
DK		 4490}
4491
4492/*
4493 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4494 * flag here, even though the filesystem is not technically dirty yet.
4495 */
c4be0c1d 4496static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4497{
9ca92389
TT		 4498 if (sb->s_flags & MS_RDONLY)
4499 return 0;
4500
c642dc9e
ES		 4501 if (EXT4_SB(sb)->s_journal) {
4502 /* Reset the needs_recovery flag before the fs is unlocked. */
4503 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4504 }
4505
9ca92389 4506 ext4_commit_super(sb, 1);
c4be0c1d 4507 return 0;
ac27a0ec
DK		 4508}
4509
673c6100
TT		 4510/*
4511 * Structure to save mount options for ext4_remount's benefit
4512 */
4513struct ext4_mount_options {
4514 unsigned long s_mount_opt;
a2595b8a 4515 unsigned long s_mount_opt2;
08cefc7a
EB		 4516 kuid_t s_resuid;
4517 kgid_t s_resgid;
673c6100
TT		 4518 unsigned long s_commit_interval;
4519 u32 s_min_batch_time, s_max_batch_time;
4520#ifdef CONFIG_QUOTA
4521 int s_jquota_fmt;
a2d4a646 4522 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4523#endif
4524};
4525
2b2d6d01 4526static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4527{
2b2d6d01 4528 struct ext4_super_block *es;
617ba13b 4529 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4530 unsigned long old_sb_flags;
617ba13b 4531 struct ext4_mount_options old_opts;
c79d967d 4532 int enable_quota = 0;
8a266467 4533 ext4_group_t g;
b3881f74 4534 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4535 int err = 0;
ac27a0ec 4536#ifdef CONFIG_QUOTA
03dafb5f 4537 int i, j;
ac27a0ec 4538#endif
d4c402d9 4539 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4540
4541 /* Store the original options */
4542 old_sb_flags = sb->s_flags;
4543 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4544 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4545 old_opts.s_resuid = sbi->s_resuid;
4546 old_opts.s_resgid = sbi->s_resgid;
4547 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4548 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4549 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4550#ifdef CONFIG_QUOTA
4551 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4552 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4553 if (sbi->s_qf_names[i]) {
4554 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4555 GFP_KERNEL);
4556 if (!old_opts.s_qf_names[i]) {
4557 for (j = 0; j < i; j++)
4558 kfree(old_opts.s_qf_names[j]);
3e36a163 4559 kfree(orig_data);
03dafb5f
CG		 4560 return -ENOMEM;
4561 }
4562 } else
4563 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4564#endif
b3881f74
TT		 4565 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4566 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4567
661aa520 4568 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4569 err = -EINVAL;
4570 goto restore_opts;
4571 }
4572
6b992ff2 4573 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4574 test_opt(sb, JOURNAL_CHECKSUM)) {
4575 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4576 "during remount not supported; ignoring");
4577 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4578 }
4579
6ae6514b
PS		 4580 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4581 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4582 ext4_msg(sb, KERN_ERR, "can't mount with "
4583 "both data=journal and delalloc");
4584 err = -EINVAL;
4585 goto restore_opts;
4586 }
4587 if (test_opt(sb, DIOREAD_NOLOCK)) {
4588 ext4_msg(sb, KERN_ERR, "can't mount with "
4589 "both data=journal and dioread_nolock");
4590 err = -EINVAL;
4591 goto restore_opts;
4592 }
923ae0ff
RZ		 4593 if (test_opt(sb, DAX)) {
4594 ext4_msg(sb, KERN_ERR, "can't mount with "
4595 "both data=journal and dax");
4596 err = -EINVAL;
4597 goto restore_opts;
4598 }
4599 }
4600
4601 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4602 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4603 "dax flag with busy inodes while remounting");
4604 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 4605 }
4606
4ab2f15b 4607 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4608 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4609
4610 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4611 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4612
4613 es = sbi->s_es;
4614
b3881f74 4615 if (sbi->s_journal) {
0390131b 4616 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4617 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4618 }
ac27a0ec 4619
a2fd66d0
TT		 4620 if (*flags & MS_LAZYTIME)
4621 sb->s_flags |= MS_LAZYTIME;
4622
661aa520 4623 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4624 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4625 err = -EROFS;
4626 goto restore_opts;
4627 }
4628
4629 if (*flags & MS_RDONLY) {
38c03b34
TT		 4630 err = sync_filesystem(sb);
4631 if (err < 0)
4632 goto restore_opts;
0f0dd62f
CH		 4633 err = dquot_suspend(sb, -1);
4634 if (err < 0)
c79d967d 4635 goto restore_opts;
c79d967d 4636
ac27a0ec
DK		 4637 /*
4638 * First of all, the unconditional stuff we have to do
4639 * to disable replay of the journal when we next remount
4640 */
4641 sb->s_flags |= MS_RDONLY;
4642
4643 /*
4644 * OK, test if we are remounting a valid rw partition
4645 * readonly, and if so set the rdonly flag and then
4646 * mark the partition as valid again.
4647 */
617ba13b
MC		 4648 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4649 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4650 es->s_state = cpu_to_le16(sbi->s_mount_state);
4651
a63c9eb2 4652 if (sbi->s_journal)
0390131b 4653 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4654 } else {
a13fb1a4 4655 /* Make sure we can mount this feature set readwrite */
2cb5cc8b
DW		 4656 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4657 EXT4_FEATURE_RO_COMPAT_READONLY) ||
4658 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4659 err = -EROFS;
4660 goto restore_opts;
4661 }
8a266467
TT		 4662 /*
4663 * Make sure the group descriptor checksums
0b8e58a1 4664 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4665 */
4666 for (g = 0; g < sbi->s_groups_count; g++) {
4667 struct ext4_group_desc *gdp =
4668 ext4_get_group_desc(sb, g, NULL);
4669
feb0ab32 4670 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4671 ext4_msg(sb, KERN_ERR,
4672 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4673 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4674 le16_to_cpu(gdp->bg_checksum));
6a797d27 4675 err = -EFSBADCRC;
8a266467
TT		 4676 goto restore_opts;
4677 }
4678 }
4679
ead6596b
ES		 4680 /*
4681 * If we have an unprocessed orphan list hanging
4682 * around from a previously readonly bdev mount,
4683 * require a full umount/remount for now.
4684 */
4685 if (es->s_last_orphan) {
b31e1552 4686 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4687 "remount RDWR because of unprocessed "
4688 "orphan inode list. Please "
b31e1552 4689 "umount/remount instead");
ead6596b
ES		 4690 err = -EINVAL;
4691 goto restore_opts;
4692 }
4693
ac27a0ec
DK		 4694 /*
4695 * Mounting a RDONLY partition read-write, so reread
4696 * and store the current valid flag. (It may have
4697 * been changed by e2fsck since we originally mounted
4698 * the partition.)
4699 */
0390131b
FM		 4700 if (sbi->s_journal)
4701 ext4_clear_journal_err(sb, es);
ac27a0ec 4702 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4703 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4704 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4705 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4706 EXT4_FEATURE_INCOMPAT_MMP))
4707 if (ext4_multi_mount_protect(sb,
4708 le64_to_cpu(es->s_mmp_block))) {
4709 err = -EROFS;
4710 goto restore_opts;
4711 }
c79d967d 4712 enable_quota = 1;
ac27a0ec
DK		 4713 }
4714 }
bfff6873
LC		 4715
4716 /*
4717 * Reinitialize lazy itable initialization thread based on
4718 * current settings
4719 */
4720 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4721 ext4_unregister_li_request(sb);
4722 else {
4723 ext4_group_t first_not_zeroed;
4724 first_not_zeroed = ext4_has_uninit_itable(sb);
4725 ext4_register_li_request(sb, first_not_zeroed);
4726 }
4727
6fd058f7 4728 ext4_setup_system_zone(sb);
d096ad0f 4729 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4730 ext4_commit_super(sb, 1);
0390131b 4731
ac27a0ec
DK		 4732#ifdef CONFIG_QUOTA
4733 /* Release old quota file names */
a2d4a646 4734 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 4735 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4736 if (enable_quota) {
4737 if (sb_any_quota_suspended(sb))
4738 dquot_resume(sb, -1);
4739 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4740 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4741 err = ext4_enable_quotas(sb);
07724f98 4742 if (err)
7c319d32 4743 goto restore_opts;
7c319d32
AK		 4744 }
4745 }
ac27a0ec 4746#endif
d4c402d9 4747
a26f4992 4748 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 4749 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4750 kfree(orig_data);
ac27a0ec 4751 return 0;
0b8e58a1 4752
ac27a0ec
DK		 4753restore_opts:
4754 sb->s_flags = old_sb_flags;
4755 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4756 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4757 sbi->s_resuid = old_opts.s_resuid;
4758 sbi->s_resgid = old_opts.s_resgid;
4759 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4760 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4761 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4762#ifdef CONFIG_QUOTA
4763 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 4764 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 4765 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 4766 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4767 }
4768#endif
d4c402d9 4769 kfree(orig_data);
ac27a0ec
DK		 4770 return err;
4771}
4772
2b2d6d01 4773static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4774{
4775 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4776 struct ext4_sb_info *sbi = EXT4_SB(sb);
4777 struct ext4_super_block *es = sbi->s_es;
27dd4385 4778 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 4779 u64 fsid;
d02a9391 4780 s64 bfree;
27dd4385 4781 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 4782
952fc18e
TT		 4783 if (!test_opt(sb, MINIX_DF))
4784 overhead = sbi->s_overhead;
ac27a0ec 4785
617ba13b 4786 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4787 buf->f_bsize = sb->s_blocksize;
b72f78cb 4788 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4789 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4790 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4791 /* prevent underflow in case that few free space is available */
57042651 4792 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 4793 buf->f_bavail = buf->f_bfree -
4794 (ext4_r_blocks_count(es) + resv_blocks);
4795 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 4796 buf->f_bavail = 0;
4797 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4798 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4799 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4800 fsid = le64_to_cpup((void *)es->s_uuid) ^
4801 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4802 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4803 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4804
ac27a0ec
DK		 4805 return 0;
4806}
4807
0b8e58a1
AD		 4808/* Helper function for writing quotas on sync - we need to start transaction
4809 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4810 * Process 1 Process 2
617ba13b 4811 * ext4_create() quota_sync()
a269eb18 4812 * jbd2_journal_start() write_dquot()
871a2931 4813 * dquot_initialize() down(dqio_mutex)
dab291af 4814 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4815 *
4816 */
4817
4818#ifdef CONFIG_QUOTA
4819
4820static inline struct inode *dquot_to_inode(struct dquot *dquot)
4821{
4c376dca 4822 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4823}
4824
617ba13b 4825static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4826{
4827 int ret, err;
4828 handle_t *handle;
4829 struct inode *inode;
4830
4831 inode = dquot_to_inode(dquot);
9924a92a 4832 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 4833 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4834 if (IS_ERR(handle))
4835 return PTR_ERR(handle);
4836 ret = dquot_commit(dquot);
617ba13b 4837 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4838 if (!ret)
4839 ret = err;
4840 return ret;
4841}
4842
617ba13b 4843static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4844{
4845 int ret, err;
4846 handle_t *handle;
4847
9924a92a 4848 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4849 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4850 if (IS_ERR(handle))
4851 return PTR_ERR(handle);
4852 ret = dquot_acquire(dquot);
617ba13b 4853 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4854 if (!ret)
4855 ret = err;
4856 return ret;
4857}
4858
617ba13b 4859static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4860{
4861 int ret, err;
4862 handle_t *handle;
4863
9924a92a 4864 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4865 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4866 if (IS_ERR(handle)) {
4867 /* Release dquot anyway to avoid endless cycle in dqput() */
4868 dquot_release(dquot);
ac27a0ec 4869 return PTR_ERR(handle);
9c3013e9 4870 }
ac27a0ec 4871 ret = dquot_release(dquot);
617ba13b 4872 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4873 if (!ret)
4874 ret = err;
4875 return ret;
4876}
4877
617ba13b 4878static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4879{
262b4662
JK		 4880 struct super_block *sb = dquot->dq_sb;
4881 struct ext4_sb_info *sbi = EXT4_SB(sb);
4882
2c8be6b2 4883 /* Are we journaling quotas? */
262b4662
JK		 4884 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
4885 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 4886 dquot_mark_dquot_dirty(dquot);
617ba13b 4887 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4888 } else {
4889 return dquot_mark_dquot_dirty(dquot);
4890 }
4891}
4892
617ba13b 4893static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4894{
4895 int ret, err;
4896 handle_t *handle;
4897
4898 /* Data block + inode block */
2b0143b5 4899 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 4900 if (IS_ERR(handle))
4901 return PTR_ERR(handle);
4902 ret = dquot_commit_info(sb, type);
617ba13b 4903 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4904 if (!ret)
4905 ret = err;
4906 return ret;
4907}
4908
4909/*
4910 * Turn on quotas during mount time - we need to find
4911 * the quota file and such...
4912 */
617ba13b 4913static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4914{
287a8095
CH		 4915 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4916 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4917}
4918
4919/*
4920 * Standard function to be called on quota_on
4921 */
617ba13b 4922static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4923 struct path *path)
ac27a0ec
DK		 4924{
4925 int err;
ac27a0ec
DK		 4926
4927 if (!test_opt(sb, QUOTA))
4928 return -EINVAL;
0623543b 4929
ac27a0ec 4930 /* Quotafile not on the same filesystem? */
d8c9584e 4931 if (path->dentry->d_sb != sb)
ac27a0ec 4932 return -EXDEV;
0623543b
JK		 4933 /* Journaling quota? */
4934 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4935 /* Quotafile not in fs root? */
f00c9e44 4936 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4937 ext4_msg(sb, KERN_WARNING,
4938 "Quota file not on filesystem root. "
4939 "Journaled quota will not work");
2b2d6d01 4940 }
0623543b
JK		 4941
4942 /*
4943 * When we journal data on quota file, we have to flush journal to see
4944 * all updates to the file when we bypass pagecache...
4945 */
0390131b 4946 if (EXT4_SB(sb)->s_journal &&
2b0143b5 4947 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 4948 /*
4949 * We don't need to lock updates but journal_flush() could
4950 * otherwise be livelocked...
4951 */
4952 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4953 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4954 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4955 if (err)
7ffe1ea8 4956 return err;
0623543b
JK		 4957 }
4958
f00c9e44 4959 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4960}
4961
7c319d32
AK		 4962static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4963 unsigned int flags)
4964{
4965 int err;
4966 struct inode *qf_inode;
a2d4a646 4967 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 4968 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4969 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4970 };
4971
4972 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4973
4974 if (!qf_inums[type])
4975 return -EPERM;
4976
4977 qf_inode = ext4_iget(sb, qf_inums[type]);
4978 if (IS_ERR(qf_inode)) {
4979 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4980 return PTR_ERR(qf_inode);
4981 }
4982
bcb13850
JK		 4983 /* Don't account quota for quota files to avoid recursion */
4984 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 4985 err = dquot_enable(qf_inode, type, format_id, flags);
4986 iput(qf_inode);
4987
4988 return err;
4989}
4990
4991/* Enable usage tracking for all quota types. */
4992static int ext4_enable_quotas(struct super_block *sb)
4993{
4994 int type, err = 0;
a2d4a646 4995 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 4996 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4997 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4998 };
4999
5000 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5001 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5002 if (qf_inums[type]) {
5003 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5004 DQUOT_USAGE_ENABLED);
5005 if (err) {
5006 ext4_warning(sb,
72ba7450
TT		 5007 "Failed to enable quota tracking "
5008 "(type=%d, err=%d). Please run "
5009 "e2fsck to fix.", type, err);
7c319d32
AK		 5010 return err;
5011 }
5012 }
5013 }
5014 return 0;
5015}
5016
ca0e05e4
DM		 5017static int ext4_quota_off(struct super_block *sb, int type)
5018{
21f97697
JK		 5019 struct inode *inode = sb_dqopt(sb)->files[type];
5020 handle_t *handle;
5021
87009d86
DM		 5022 /* Force all delayed allocation blocks to be allocated.
5023 * Caller already holds s_umount sem */
5024 if (test_opt(sb, DELALLOC))
ca0e05e4 5025 sync_filesystem(sb);
ca0e05e4 5026
0b268590
AG		 5027 if (!inode)
5028 goto out;
5029
21f97697
JK		 5030 /* Update modification times of quota files when userspace can
5031 * start looking at them */
9924a92a 5032 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5033 if (IS_ERR(handle))
5034 goto out;
5035 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5036 ext4_mark_inode_dirty(handle, inode);
5037 ext4_journal_stop(handle);
5038
5039out:
ca0e05e4
DM		 5040 return dquot_quota_off(sb, type);
5041}
5042
ac27a0ec
DK		 5043/* Read data from quotafile - avoid pagecache and such because we cannot afford
5044 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5045 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5046 * we don't have to be afraid of races */
617ba13b 5047static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5048 size_t len, loff_t off)
5049{
5050 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5051 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5052 int offset = off & (sb->s_blocksize - 1);
5053 int tocopy;
5054 size_t toread;
5055 struct buffer_head *bh;
5056 loff_t i_size = i_size_read(inode);
5057
5058 if (off > i_size)
5059 return 0;
5060 if (off+len > i_size)
5061 len = i_size-off;
5062 toread = len;
5063 while (toread > 0) {
5064 tocopy = sb->s_blocksize - offset < toread ?
5065 sb->s_blocksize - offset : toread;
1c215028
TT		 5066 bh = ext4_bread(NULL, inode, blk, 0);
5067 if (IS_ERR(bh))
5068 return PTR_ERR(bh);
ac27a0ec
DK		 5069 if (!bh) /* A hole? */
5070 memset(data, 0, tocopy);
5071 else
5072 memcpy(data, bh->b_data+offset, tocopy);
5073 brelse(bh);
5074 offset = 0;
5075 toread -= tocopy;
5076 data += tocopy;
5077 blk++;
5078 }
5079 return len;
5080}
5081
5082/* Write to quotafile (we know the transaction is already started and has
5083 * enough credits) */
617ba13b 5084static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5085 const char *data, size_t len, loff_t off)
5086{
5087 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5088 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5089 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5090 int retries = 0;
ac27a0ec
DK		 5091 struct buffer_head *bh;
5092 handle_t *handle = journal_current_handle();
5093
0390131b 5094 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5095 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5096 " cancelled because transaction is not started",
9c3013e9
JK		 5097 (unsigned long long)off, (unsigned long long)len);
5098 return -EIO;
5099 }
67eeb568
DM		 5100 /*
5101 * Since we account only one data block in transaction credits,
5102 * then it is impossible to cross a block boundary.
5103 */
5104 if (sb->s_blocksize - offset < len) {
5105 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5106 " cancelled because not block aligned",
5107 (unsigned long long)off, (unsigned long long)len);
5108 return -EIO;
5109 }
5110
c5e298ae
TT		 5111 do {
5112 bh = ext4_bread(handle, inode, blk,
5113 EXT4_GET_BLOCKS_CREATE |
5114 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5115 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5116 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5117 if (IS_ERR(bh))
5118 return PTR_ERR(bh);
67eeb568
DM		 5119 if (!bh)
5120 goto out;
5d601255 5121 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5122 err = ext4_journal_get_write_access(handle, bh);
5123 if (err) {
5124 brelse(bh);
1c215028 5125 return err;
ac27a0ec 5126 }
67eeb568
DM		 5127 lock_buffer(bh);
5128 memcpy(bh->b_data+offset, data, len);
5129 flush_dcache_page(bh->b_page);
5130 unlock_buffer(bh);
62d2b5f2 5131 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5132 brelse(bh);
ac27a0ec 5133out:
67eeb568
DM		 5134 if (inode->i_size < off + len) {
5135 i_size_write(inode, off + len);
617ba13b 5136 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5137 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5138 }
67eeb568 5139 return len;
ac27a0ec
DK		 5140}
5141
5142#endif
5143
152a0836
AV		 5144static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5145 const char *dev_name, void *data)
ac27a0ec 5146{
152a0836 5147 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5148}
5149
c290ea01 5150#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5151static inline void register_as_ext2(void)
5152{
5153 int err = register_filesystem(&ext2_fs_type);
5154 if (err)
5155 printk(KERN_WARNING
5156 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5157}
5158
5159static inline void unregister_as_ext2(void)
5160{
5161 unregister_filesystem(&ext2_fs_type);
5162}
2035e776
TT		 5163
5164static inline int ext2_feature_set_ok(struct super_block *sb)
5165{
5166 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5167 return 0;
5168 if (sb->s_flags & MS_RDONLY)
5169 return 1;
5170 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5171 return 0;
5172 return 1;
5173}
24b58424
TT		 5174#else
5175static inline void register_as_ext2(void) { }
5176static inline void unregister_as_ext2(void) { }
2035e776 5177static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5178#endif
5179
24b58424
TT		 5180static inline void register_as_ext3(void)
5181{
5182 int err = register_filesystem(&ext3_fs_type);
5183 if (err)
5184 printk(KERN_WARNING
5185 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5186}
5187
5188static inline void unregister_as_ext3(void)
5189{
5190 unregister_filesystem(&ext3_fs_type);
5191}
2035e776
TT		 5192
5193static inline int ext3_feature_set_ok(struct super_block *sb)
5194{
5195 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5196 return 0;
5197 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5198 return 0;
5199 if (sb->s_flags & MS_RDONLY)
5200 return 1;
5201 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5202 return 0;
5203 return 1;
5204}
24b58424 5205
03010a33
TT		 5206static struct file_system_type ext4_fs_type = {
5207 .owner = THIS_MODULE,
5208 .name = "ext4",
152a0836 5209 .mount = ext4_mount,
03010a33
TT		 5210 .kill_sb = kill_block_super,
5211 .fs_flags = FS_REQUIRES_DEV,
5212};
7f78e035 5213MODULE_ALIAS_FS("ext4");
03010a33 5214
e9e3bcec
ES		 5215/* Shared across all ext4 file systems */
5216wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5217struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5218
5dabfc78 5219static int __init ext4_init_fs(void)
ac27a0ec 5220{
e9e3bcec 5221 int i, err;
c9de560d 5222
e294a537 5223 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5224 ext4_li_info = NULL;
5225 mutex_init(&ext4_li_mtx);
5226
9a4c8019 5227 /* Build-time check for flags consistency */
12e9b892 5228 ext4_check_flag_values();
e9e3bcec
ES		 5229
5230 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5231 mutex_init(&ext4__aio_mutex[i]);
5232 init_waitqueue_head(&ext4__ioend_wq[i]);
5233 }
5234
51865fda 5235 err = ext4_init_es();
6fd058f7
TT		 5236 if (err)
5237 return err;
51865fda
ZL		 5238
5239 err = ext4_init_pageio();
5240 if (err)
b5799018 5241 goto out5;
51865fda 5242
5dabfc78 5243 err = ext4_init_system_zone();
bd2d0210 5244 if (err)
b5799018 5245 goto out4;
857ac889 5246
b5799018 5247 err = ext4_init_sysfs();
dd68314c 5248 if (err)
b5799018 5249 goto out3;
857ac889 5250
5dabfc78 5251 err = ext4_init_mballoc();
c9de560d
AT		 5252 if (err)
5253 goto out2;
9c191f70
M		 5254 else
5255 ext4_mballoc_ready = 1;
ac27a0ec
DK		 5256 err = init_inodecache();
5257 if (err)
5258 goto out1;
24b58424 5259 register_as_ext3();
2035e776 5260 register_as_ext2();
03010a33 5261 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5262 if (err)
5263 goto out;
bfff6873 5264
ac27a0ec
DK		 5265 return 0;
5266out:
24b58424
TT		 5267 unregister_as_ext2();
5268 unregister_as_ext3();
ac27a0ec
DK		 5269 destroy_inodecache();
5270out1:
9c191f70 5271 ext4_mballoc_ready = 0;
5dabfc78 5272 ext4_exit_mballoc();
9c191f70 5273out2:
b5799018
TT		 5274 ext4_exit_sysfs();
5275out3:
5dabfc78 5276 ext4_exit_system_zone();
b5799018 5277out4:
5dabfc78 5278 ext4_exit_pageio();
b5799018 5279out5:
51865fda
ZL		 5280 ext4_exit_es();
5281
ac27a0ec
DK		 5282 return err;
5283}
5284
5dabfc78 5285static void __exit ext4_exit_fs(void)
ac27a0ec 5286{
e298e73b 5287 ext4_exit_crypto();
bfff6873 5288 ext4_destroy_lazyinit_thread();
24b58424
TT		 5289 unregister_as_ext2();
5290 unregister_as_ext3();
03010a33 5291 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5292 destroy_inodecache();
5dabfc78 5293 ext4_exit_mballoc();
b5799018 5294 ext4_exit_sysfs();
5dabfc78
TT		 5295 ext4_exit_system_zone();
5296 ext4_exit_pageio();
dd12ed14 5297 ext4_exit_es();
ac27a0ec
DK		 5298}
5299
5300MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5301MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5302MODULE_LICENSE("GPL");
5dabfc78
TT		 5303module_init(ext4_init_fs)
5304module_exit(ext4_exit_fs)
Empty description
This page took 1.903295 seconds and 4 git commands to generate.