projects / GitHub / LineageOS / android_kernel_motorola_exynos9610.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'vfio-v3.10-rc5' of git://github.com/awilliam/linux-vfio
[GitHub/LineageOS/android_kernel_motorola_exynos9610.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
7abc52c2 41#include <linux/cleancache.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
4a092d73 48#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 49#include "ext4_jbd2.h"
ac27a0ec
DK		 50#include "xattr.h"
51#include "acl.h"
3661d286 52#include "mballoc.h"
ac27a0ec 53
9bffad1e
TT		 54#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
1f109d5a 57static struct proc_dir_entry *ext4_proc_root;
3197ebdb 58static struct kset *ext4_kset;
0b75a840
LC		 59static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
2adf6da8 65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 66static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 67static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
617ba13b 71static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 72static int ext4_remount(struct super_block *sb, int *flags, char *data);
73static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 74static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 75static int ext4_freeze(struct super_block *sb);
152a0836
AV		 76static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
2035e776
TT		 78static inline int ext2_feature_set_ok(struct super_block *sb);
79static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 80static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 81static void ext4_destroy_lazyinit_thread(void);
82static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 83static void ext4_clear_request_list(void);
27dd4385 84static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
ac27a0ec 85
2035e776
TT		 86#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87static struct file_system_type ext2_fs_type = {
88 .owner = THIS_MODULE,
89 .name = "ext2",
90 .mount = ext4_mount,
91 .kill_sb = kill_block_super,
92 .fs_flags = FS_REQUIRES_DEV,
93};
7f78e035 94MODULE_ALIAS_FS("ext2");
fa7614dd 95MODULE_ALIAS("ext2");
2035e776
TT		 96#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97#else
98#define IS_EXT2_SB(sb) (0)
99#endif
100
101
ba69f9ab
JK		 102#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
105 .name = "ext3",
152a0836 106 .mount = ext4_mount,
ba69f9ab
JK		 107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
109};
7f78e035 110MODULE_ALIAS_FS("ext3");
fa7614dd 111MODULE_ALIAS("ext3");
ba69f9ab
JK		 112#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
113#else
114#define IS_EXT3_SB(sb) (0)
115#endif
bd81d8ee 116
d25425f8
DW		 117static int ext4_verify_csum_type(struct super_block *sb,
118 struct ext4_super_block *es)
119{
120 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
121 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
122 return 1;
123
124 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
125}
126
a9c47317
DW		 127static __le32 ext4_superblock_csum(struct super_block *sb,
128 struct ext4_super_block *es)
129{
130 struct ext4_sb_info *sbi = EXT4_SB(sb);
131 int offset = offsetof(struct ext4_super_block, s_checksum);
132 __u32 csum;
133
134 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
135
136 return cpu_to_le32(csum);
137}
138
139int ext4_superblock_csum_verify(struct super_block *sb,
140 struct ext4_super_block *es)
141{
142 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
143 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
144 return 1;
145
146 return es->s_checksum == ext4_superblock_csum(sb, es);
147}
148
06db49e6 149void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 150{
06db49e6
TT		 151 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
152
a9c47317
DW		 153 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
154 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
155 return;
156
157 es->s_checksum = ext4_superblock_csum(sb, es);
158}
159
9933fc0a
TT		 160void *ext4_kvmalloc(size_t size, gfp_t flags)
161{
162 void *ret;
163
164 ret = kmalloc(size, flags);
165 if (!ret)
166 ret = __vmalloc(size, flags, PAGE_KERNEL);
167 return ret;
168}
169
170void *ext4_kvzalloc(size_t size, gfp_t flags)
171{
172 void *ret;
173
db9481c0 174 ret = kzalloc(size, flags);
9933fc0a
TT		 175 if (!ret)
176 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
177 return ret;
178}
179
180void ext4_kvfree(void *ptr)
181{
182 if (is_vmalloc_addr(ptr))
183 vfree(ptr);
184 else
185 kfree(ptr);
186
187}
188
8fadc143
AR		 189ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
190 struct ext4_group_desc *bg)
bd81d8ee 191{
3a14589c 192 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 193 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 194 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 195}
196
8fadc143
AR		 197ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
198 struct ext4_group_desc *bg)
bd81d8ee 199{
5272f837 200 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 201 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 202 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 203}
204
8fadc143
AR		 205ext4_fsblk_t ext4_inode_table(struct super_block *sb,
206 struct ext4_group_desc *bg)
bd81d8ee 207{
5272f837 208 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 209 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 210 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 211}
212
021b65bb
TT		 213__u32 ext4_free_group_clusters(struct super_block *sb,
214 struct ext4_group_desc *bg)
560671a0
AK		 215{
216 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
217 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 218 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 219}
220
221__u32 ext4_free_inodes_count(struct super_block *sb,
222 struct ext4_group_desc *bg)
223{
224 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
225 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 226 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 227}
228
229__u32 ext4_used_dirs_count(struct super_block *sb,
230 struct ext4_group_desc *bg)
231{
232 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
233 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 234 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 235}
236
237__u32 ext4_itable_unused_count(struct super_block *sb,
238 struct ext4_group_desc *bg)
239{
240 return le16_to_cpu(bg->bg_itable_unused_lo) |
241 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 242 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 243}
244
8fadc143
AR		 245void ext4_block_bitmap_set(struct super_block *sb,
246 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 247{
3a14589c 248 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 249 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
250 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 251}
252
8fadc143
AR		 253void ext4_inode_bitmap_set(struct super_block *sb,
254 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 255{
5272f837 256 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 257 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
258 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 259}
260
8fadc143
AR		 261void ext4_inode_table_set(struct super_block *sb,
262 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 263{
5272f837 264 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 265 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
266 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 267}
268
021b65bb
TT		 269void ext4_free_group_clusters_set(struct super_block *sb,
270 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 271{
272 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
273 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
274 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
275}
276
277void ext4_free_inodes_set(struct super_block *sb,
278 struct ext4_group_desc *bg, __u32 count)
279{
280 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
281 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
282 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
283}
284
285void ext4_used_dirs_set(struct super_block *sb,
286 struct ext4_group_desc *bg, __u32 count)
287{
288 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
289 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
290 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
291}
292
293void ext4_itable_unused_set(struct super_block *sb,
294 struct ext4_group_desc *bg, __u32 count)
295{
296 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
297 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
298 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
299}
300
d3d1faf6 301
1c13d5c0
TT		 302static void __save_error_info(struct super_block *sb, const char *func,
303 unsigned int line)
304{
305 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
306
307 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
308 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
309 es->s_last_error_time = cpu_to_le32(get_seconds());
310 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
311 es->s_last_error_line = cpu_to_le32(line);
312 if (!es->s_first_error_time) {
313 es->s_first_error_time = es->s_last_error_time;
314 strncpy(es->s_first_error_func, func,
315 sizeof(es->s_first_error_func));
316 es->s_first_error_line = cpu_to_le32(line);
317 es->s_first_error_ino = es->s_last_error_ino;
318 es->s_first_error_block = es->s_last_error_block;
319 }
66e61a9e
TT		 320 /*
321 * Start the daily error reporting function if it hasn't been
322 * started already
323 */
324 if (!es->s_error_count)
325 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 326 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 327}
328
329static void save_error_info(struct super_block *sb, const char *func,
330 unsigned int line)
331{
332 __save_error_info(sb, func, line);
333 ext4_commit_super(sb, 1);
334}
335
7c2e7087
TT		 336/*
337 * The del_gendisk() function uninitializes the disk-specific data
338 * structures, including the bdi structure, without telling anyone
339 * else. Once this happens, any attempt to call mark_buffer_dirty()
340 * (for example, by ext4_commit_super), will cause a kernel OOPS.
341 * This is a kludge to prevent these oops until we can put in a proper
342 * hook in del_gendisk() to inform the VFS and file system layers.
343 */
344static int block_device_ejected(struct super_block *sb)
345{
346 struct inode *bd_inode = sb->s_bdev->bd_inode;
347 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
348
349 return bdi->dev == NULL;
350}
351
18aadd47
BJ		 352static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
353{
354 struct super_block *sb = journal->j_private;
355 struct ext4_sb_info *sbi = EXT4_SB(sb);
356 int error = is_journal_aborted(journal);
5d3ee208 357 struct ext4_journal_cb_entry *jce;
18aadd47 358
5d3ee208 359 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 360 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 361 while (!list_empty(&txn->t_private_list)) {
362 jce = list_entry(txn->t_private_list.next,
363 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 364 list_del_init(&jce->jce_list);
365 spin_unlock(&sbi->s_md_lock);
366 jce->jce_func(sb, jce, error);
367 spin_lock(&sbi->s_md_lock);
368 }
369 spin_unlock(&sbi->s_md_lock);
370}
1c13d5c0 371
ac27a0ec
DK		 372/* Deal with the reporting of failure conditions on a filesystem such as
373 * inconsistencies detected or read IO failures.
374 *
375 * On ext2, we can store the error state of the filesystem in the
617ba13b 376 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 377 * write ordering constraints on the superblock which prevent us from
378 * writing it out straight away; and given that the journal is about to
379 * be aborted, we can't rely on the current, or future, transactions to
380 * write out the superblock safely.
381 *
dab291af 382 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 383 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 384 * that error until we've noted it down and cleared it.
385 */
386
617ba13b 387static void ext4_handle_error(struct super_block *sb)
ac27a0ec 388{
ac27a0ec
DK		 389 if (sb->s_flags & MS_RDONLY)
390 return;
391
2b2d6d01 392 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 393 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 394
4ab2f15b 395 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 396 if (journal)
dab291af 397 jbd2_journal_abort(journal, -EIO);
ac27a0ec 398 }
2b2d6d01 399 if (test_opt(sb, ERRORS_RO)) {
b31e1552 400 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 401 sb->s_flags |= MS_RDONLY;
402 }
ac27a0ec 403 if (test_opt(sb, ERRORS_PANIC))
617ba13b 404 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 405 sb->s_id);
406}
407
12062ddd 408void __ext4_error(struct super_block *sb, const char *function,
c398eda0 409 unsigned int line, const char *fmt, ...)
ac27a0ec 410{
0ff2ea7d 411 struct va_format vaf;
ac27a0ec
DK		 412 va_list args;
413
414 va_start(args, fmt);
0ff2ea7d
JP		 415 vaf.fmt = fmt;
416 vaf.va = &args;
417 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
418 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec 419 va_end(args);
f3fc0210 420 save_error_info(sb, function, line);
ac27a0ec 421
617ba13b 422 ext4_handle_error(sb);
ac27a0ec
DK		 423}
424
c398eda0
TT		 425void ext4_error_inode(struct inode *inode, const char *function,
426 unsigned int line, ext4_fsblk_t block,
273df556
FM		 427 const char *fmt, ...)
428{
429 va_list args;
f7c21177 430 struct va_format vaf;
1c13d5c0 431 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 432
1c13d5c0
TT		 433 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
434 es->s_last_error_block = cpu_to_le64(block);
435 save_error_info(inode->i_sb, function, line);
273df556 436 va_start(args, fmt);
f7c21177
TT		 437 vaf.fmt = fmt;
438 vaf.va = &args;
c398eda0 439 if (block)
d9ee81da
JP		 440 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
441 "inode #%lu: block %llu: comm %s: %pV\n",
442 inode->i_sb->s_id, function, line, inode->i_ino,
443 block, current->comm, &vaf);
444 else
445 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
446 "inode #%lu: comm %s: %pV\n",
447 inode->i_sb->s_id, function, line, inode->i_ino,
448 current->comm, &vaf);
273df556
FM		 449 va_end(args);
450
451 ext4_handle_error(inode->i_sb);
452}
453
c398eda0 454void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 455 unsigned int line, ext4_fsblk_t block,
456 const char *fmt, ...)
273df556
FM		 457{
458 va_list args;
f7c21177 459 struct va_format vaf;
1c13d5c0 460 struct ext4_super_block *es;
496ad9aa 461 struct inode *inode = file_inode(file);
273df556
FM		 462 char pathname[80], *path;
463
1c13d5c0
TT		 464 es = EXT4_SB(inode->i_sb)->s_es;
465 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
466 save_error_info(inode->i_sb, function, line);
273df556 467 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 468 if (IS_ERR(path))
273df556 469 path = "(unknown)";
f7c21177
TT		 470 va_start(args, fmt);
471 vaf.fmt = fmt;
472 vaf.va = &args;
d9ee81da
JP		 473 if (block)
474 printk(KERN_CRIT
475 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
476 "block %llu: comm %s: path %s: %pV\n",
477 inode->i_sb->s_id, function, line, inode->i_ino,
478 block, current->comm, path, &vaf);
479 else
480 printk(KERN_CRIT
481 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
482 "comm %s: path %s: %pV\n",
483 inode->i_sb->s_id, function, line, inode->i_ino,
484 current->comm, path, &vaf);
273df556
FM		 485 va_end(args);
486
487 ext4_handle_error(inode->i_sb);
488}
489
722887dd
TT		 490const char *ext4_decode_error(struct super_block *sb, int errno,
491 char nbuf[16])
ac27a0ec
DK		 492{
493 char *errstr = NULL;
494
495 switch (errno) {
496 case -EIO:
497 errstr = "IO failure";
498 break;
499 case -ENOMEM:
500 errstr = "Out of memory";
501 break;
502 case -EROFS:
78f1ddbb
TT		 503 if (!sb || (EXT4_SB(sb)->s_journal &&
504 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 505 errstr = "Journal has aborted";
506 else
507 errstr = "Readonly filesystem";
508 break;
509 default:
510 /* If the caller passed in an extra buffer for unknown
511 * errors, textualise them now. Else we just return
512 * NULL. */
513 if (nbuf) {
514 /* Check for truncated error codes... */
515 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
516 errstr = nbuf;
517 }
518 break;
519 }
520
521 return errstr;
522}
523
617ba13b 524/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 525 * automatically and invokes the appropriate error response. */
526
c398eda0
TT		 527void __ext4_std_error(struct super_block *sb, const char *function,
528 unsigned int line, int errno)
ac27a0ec
DK		 529{
530 char nbuf[16];
531 const char *errstr;
532
533 /* Special case: if the error is EROFS, and we're not already
534 * inside a transaction, then there's really no point in logging
535 * an error. */
536 if (errno == -EROFS && journal_current_handle() == NULL &&
537 (sb->s_flags & MS_RDONLY))
538 return;
539
617ba13b 540 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 541 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
542 sb->s_id, function, line, errstr);
1c13d5c0 543 save_error_info(sb, function, line);
ac27a0ec 544
617ba13b 545 ext4_handle_error(sb);
ac27a0ec
DK		 546}
547
548/*
617ba13b 549 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 550 * abort function may be used to deal with unrecoverable failures such
551 * as journal IO errors or ENOMEM at a critical moment in log management.
552 *
553 * We unconditionally force the filesystem into an ABORT|READONLY state,
554 * unless the error response on the fs has been set to panic in which
555 * case we take the easy way out and panic immediately.
556 */
557
c67d859e 558void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 559 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 560{
561 va_list args;
562
1c13d5c0 563 save_error_info(sb, function, line);
ac27a0ec 564 va_start(args, fmt);
c398eda0
TT		 565 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
566 function, line);
ac27a0ec
DK		 567 vprintk(fmt, args);
568 printk("\n");
569 va_end(args);
570
1c13d5c0
TT		 571 if ((sb->s_flags & MS_RDONLY) == 0) {
572 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
573 sb->s_flags |= MS_RDONLY;
574 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
575 if (EXT4_SB(sb)->s_journal)
576 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
577 save_error_info(sb, function, line);
578 }
ac27a0ec 579 if (test_opt(sb, ERRORS_PANIC))
617ba13b 580 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 581}
582
0ff2ea7d 583void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 584{
0ff2ea7d 585 struct va_format vaf;
b31e1552
ES		 586 va_list args;
587
588 va_start(args, fmt);
0ff2ea7d
JP		 589 vaf.fmt = fmt;
590 vaf.va = &args;
591 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 592 va_end(args);
593}
594
12062ddd 595void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 596 unsigned int line, const char *fmt, ...)
ac27a0ec 597{
0ff2ea7d 598 struct va_format vaf;
ac27a0ec
DK		 599 va_list args;
600
601 va_start(args, fmt);
0ff2ea7d
JP		 602 vaf.fmt = fmt;
603 vaf.va = &args;
604 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
605 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 606 va_end(args);
607}
608
e29136f8
TT		 609void __ext4_grp_locked_error(const char *function, unsigned int line,
610 struct super_block *sb, ext4_group_t grp,
611 unsigned long ino, ext4_fsblk_t block,
612 const char *fmt, ...)
5d1b1b3f
AK		 613__releases(bitlock)
614__acquires(bitlock)
615{
0ff2ea7d 616 struct va_format vaf;
5d1b1b3f
AK		 617 va_list args;
618 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
619
1c13d5c0
TT		 620 es->s_last_error_ino = cpu_to_le32(ino);
621 es->s_last_error_block = cpu_to_le64(block);
622 __save_error_info(sb, function, line);
0ff2ea7d 623
5d1b1b3f 624 va_start(args, fmt);
0ff2ea7d
JP		 625
626 vaf.fmt = fmt;
627 vaf.va = &args;
21149d61 628 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 629 sb->s_id, function, line, grp);
630 if (ino)
0ff2ea7d 631 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 632 if (block)
0ff2ea7d
JP		 633 printk(KERN_CONT "block %llu:", (unsigned long long) block);
634 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 635 va_end(args);
636
637 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 638 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 639 return;
640 }
1c13d5c0 641
5d1b1b3f
AK		 642 ext4_unlock_group(sb, grp);
643 ext4_handle_error(sb);
644 /*
645 * We only get here in the ERRORS_RO case; relocking the group
646 * may be dangerous, but nothing bad will happen since the
647 * filesystem will have already been marked read/only and the
648 * journal has been aborted. We return 1 as a hint to callers
649 * who might what to use the return value from
25985edc 650 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 651 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
652 * aggressively from the ext4 function in question, with a
653 * more appropriate error code.
654 */
655 ext4_lock_group(sb, grp);
656 return;
657}
658
617ba13b 659void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 660{
617ba13b 661 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 662
617ba13b 663 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 664 return;
665
12062ddd 666 ext4_warning(sb,
ac27a0ec
DK		 667 "updating to rev %d because of new feature flag, "
668 "running e2fsck is recommended",
617ba13b 669 EXT4_DYNAMIC_REV);
ac27a0ec 670
617ba13b
MC		 671 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
672 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
673 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 674 /* leave es->s_feature_*compat flags alone */
675 /* es->s_uuid will be set by e2fsck if empty */
676
677 /*
678 * The rest of the superblock fields should be zero, and if not it
679 * means they are likely already in use, so leave them alone. We
680 * can leave it up to e2fsck to clean up any inconsistencies there.
681 */
682}
683
684/*
685 * Open the external journal device
686 */
b31e1552 687static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 688{
689 struct block_device *bdev;
690 char b[BDEVNAME_SIZE];
691
d4d77629 692 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 693 if (IS_ERR(bdev))
694 goto fail;
695 return bdev;
696
697fail:
b31e1552 698 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 699 __bdevname(dev, b), PTR_ERR(bdev));
700 return NULL;
701}
702
703/*
704 * Release the journal device
705 */
4385bab1 706static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 707{
4385bab1 708 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 709}
710
4385bab1 711static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 712{
713 struct block_device *bdev;
ac27a0ec
DK		 714 bdev = sbi->journal_bdev;
715 if (bdev) {
4385bab1 716 ext4_blkdev_put(bdev);
ac27a0ec
DK		 717 sbi->journal_bdev = NULL;
718 }
ac27a0ec
DK		 719}
720
721static inline struct inode *orphan_list_entry(struct list_head *l)
722{
617ba13b 723 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 724}
725
617ba13b 726static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 727{
728 struct list_head *l;
729
b31e1552
ES		 730 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
731 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 732
733 printk(KERN_ERR "sb_info orphan list:\n");
734 list_for_each(l, &sbi->s_orphan) {
735 struct inode *inode = orphan_list_entry(l);
736 printk(KERN_ERR " "
737 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
738 inode->i_sb->s_id, inode->i_ino, inode,
739 inode->i_mode, inode->i_nlink,
740 NEXT_ORPHAN(inode));
741 }
742}
743
2b2d6d01 744static void ext4_put_super(struct super_block *sb)
ac27a0ec 745{
617ba13b
MC		 746 struct ext4_sb_info *sbi = EXT4_SB(sb);
747 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 748 int i, err;
ac27a0ec 749
857ac889 750 ext4_unregister_li_request(sb);
e0ccfd95
CH		 751 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
752
4c0425ff
MC		 753 flush_workqueue(sbi->dio_unwritten_wq);
754 destroy_workqueue(sbi->dio_unwritten_wq);
755
0390131b
FM		 756 if (sbi->s_journal) {
757 err = jbd2_journal_destroy(sbi->s_journal);
758 sbi->s_journal = NULL;
759 if (err < 0)
c67d859e 760 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 761 }
d4edac31 762
74cd15cd 763 ext4_es_unregister_shrinker(sb);
a1c6c569 764 del_timer(&sbi->s_err_report);
d4edac31
JB		 765 ext4_release_system_zone(sb);
766 ext4_mb_release(sb);
767 ext4_ext_release(sb);
768 ext4_xattr_put_super(sb);
769
ac27a0ec 770 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 771 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 772 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 773 }
58c5873a 774 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 775 ext4_commit_super(sb, 1);
776
240799cd 777 if (sbi->s_proc) {
66acdcf4 778 remove_proc_entry("options", sbi->s_proc);
9f6200bb 779 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 780 }
3197ebdb 781 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 782
783 for (i = 0; i < sbi->s_gdb_count; i++)
784 brelse(sbi->s_group_desc[i]);
f18a5f21 785 ext4_kvfree(sbi->s_group_desc);
9933fc0a 786 ext4_kvfree(sbi->s_flex_groups);
57042651 787 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 788 percpu_counter_destroy(&sbi->s_freeinodes_counter);
789 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 790 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1ac6466f 791 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
ac27a0ec
DK		 792 brelse(sbi->s_sbh);
793#ifdef CONFIG_QUOTA
794 for (i = 0; i < MAXQUOTAS; i++)
795 kfree(sbi->s_qf_names[i]);
796#endif
797
798 /* Debugging code just in case the in-memory inode orphan list
799 * isn't empty. The on-disk one can be non-empty if we've
800 * detected an error and taken the fs readonly, but the
801 * in-memory list had better be clean by this point. */
802 if (!list_empty(&sbi->s_orphan))
803 dump_orphan_list(sb, sbi);
804 J_ASSERT(list_empty(&sbi->s_orphan));
805
f98393a6 806 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 807 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
808 /*
809 * Invalidate the journal device's buffers. We don't want them
810 * floating about in memory - the physical journal device may
811 * hotswapped, and it breaks the `ro-after' testing code.
812 */
813 sync_blockdev(sbi->journal_bdev);
f98393a6 814 invalidate_bdev(sbi->journal_bdev);
617ba13b 815 ext4_blkdev_remove(sbi);
ac27a0ec 816 }
c5e06d10
JL		 817 if (sbi->s_mmp_tsk)
818 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 819 sb->s_fs_info = NULL;
3197ebdb
TT		 820 /*
821 * Now that we are completely done shutting down the
822 * superblock, we need to actually destroy the kobject.
823 */
3197ebdb
TT		 824 kobject_put(&sbi->s_kobj);
825 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 826 if (sbi->s_chksum_driver)
827 crypto_free_shash(sbi->s_chksum_driver);
705895b6 828 kfree(sbi->s_blockgroup_lock);
ac27a0ec 829 kfree(sbi);
ac27a0ec
DK		 830}
831
e18b890b 832static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 833
834/*
835 * Called inside transaction, so use GFP_NOFS
836 */
617ba13b 837static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 838{
617ba13b 839 struct ext4_inode_info *ei;
ac27a0ec 840
e6b4f8da 841 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 842 if (!ei)
843 return NULL;
0b8e58a1 844
ac27a0ec 845 ei->vfs_inode.i_version = 1;
c9de560d
AT		 846 INIT_LIST_HEAD(&ei->i_prealloc_list);
847 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 848 ext4_es_init_tree(&ei->i_es_tree);
849 rwlock_init(&ei->i_es_lock);
74cd15cd
ZL		 850 INIT_LIST_HEAD(&ei->i_es_lru);
851 ei->i_es_lru_nr = 0;
d2a17637
MC		 852 ei->i_reserved_data_blocks = 0;
853 ei->i_reserved_meta_blocks = 0;
854 ei->i_allocated_meta_blocks = 0;
9d0be502 855 ei->i_da_metadata_calc_len = 0;
7e731bc9 856 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 857 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 858#ifdef CONFIG_QUOTA
859 ei->i_reserved_quota = 0;
860#endif
8aefcd55 861 ei->jinode = NULL;
c7064ef1 862 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 863 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 864 ei->i_sync_tid = 0;
865 ei->i_datasync_tid = 0;
f7ad6d2e 866 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 867 atomic_set(&ei->i_unwritten, 0);
84c17543 868 INIT_WORK(&ei->i_unwritten_work, ext4_end_io_work);
0b8e58a1 869
ac27a0ec
DK		 870 return &ei->vfs_inode;
871}
872
7ff9c073
TT		 873static int ext4_drop_inode(struct inode *inode)
874{
875 int drop = generic_drop_inode(inode);
876
877 trace_ext4_drop_inode(inode, drop);
878 return drop;
879}
880
fa0d7e3d
NP		 881static void ext4_i_callback(struct rcu_head *head)
882{
883 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 884 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
885}
886
617ba13b 887static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 888{
9f7dd93d 889 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 890 ext4_msg(inode->i_sb, KERN_ERR,
891 "Inode %lu (%p): orphan list check failed!",
892 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 893 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
894 EXT4_I(inode), sizeof(struct ext4_inode_info),
895 true);
896 dump_stack();
897 }
fa0d7e3d 898 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 899}
900
51cc5068 901static void init_once(void *foo)
ac27a0ec 902{
617ba13b 903 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 904
a35afb83 905 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 906 init_rwsem(&ei->xattr_sem);
0e855ac8 907 init_rwsem(&ei->i_data_sem);
a35afb83 908 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 909}
910
911static int init_inodecache(void)
912{
617ba13b
MC		 913 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
914 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 915 0, (SLAB_RECLAIM_ACCOUNT|
916 SLAB_MEM_SPREAD),
20c2df83 917 init_once);
617ba13b 918 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 919 return -ENOMEM;
920 return 0;
921}
922
923static void destroy_inodecache(void)
924{
8c0a8537
KS		 925 /*
926 * Make sure all delayed rcu free inodes are flushed before we
927 * destroy cache.
928 */
929 rcu_barrier();
617ba13b 930 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 931}
932
0930fcc1 933void ext4_clear_inode(struct inode *inode)
ac27a0ec 934{
0930fcc1 935 invalidate_inode_buffers(inode);
dbd5768f 936 clear_inode(inode);
9f754758 937 dquot_drop(inode);
c2ea3fde 938 ext4_discard_preallocations(inode);
51865fda 939 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
74cd15cd 940 ext4_es_lru_del(inode);
8aefcd55
TT		 941 if (EXT4_I(inode)->jinode) {
942 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
943 EXT4_I(inode)->jinode);
944 jbd2_free_inode(EXT4_I(inode)->jinode);
945 EXT4_I(inode)->jinode = NULL;
946 }
ac27a0ec
DK		 947}
948
1b961ac0 949static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 950 u64 ino, u32 generation)
ac27a0ec 951{
ac27a0ec 952 struct inode *inode;
ac27a0ec 953
617ba13b 954 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 955 return ERR_PTR(-ESTALE);
617ba13b 956 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 957 return ERR_PTR(-ESTALE);
958
959 /* iget isn't really right if the inode is currently unallocated!!
960 *
617ba13b 961 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 962 * deleted, so we should be safe.
963 *
964 * Currently we don't know the generation for parent directory, so
965 * a generation of 0 means "accept any"
966 */
1d1fe1ee
DH		 967 inode = ext4_iget(sb, ino);
968 if (IS_ERR(inode))
969 return ERR_CAST(inode);
970 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 971 iput(inode);
972 return ERR_PTR(-ESTALE);
973 }
1b961ac0
CH		 974
975 return inode;
976}
977
978static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 979 int fh_len, int fh_type)
1b961ac0
CH		 980{
981 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
982 ext4_nfs_get_inode);
983}
984
985static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 986 int fh_len, int fh_type)
1b961ac0
CH		 987{
988 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
989 ext4_nfs_get_inode);
ac27a0ec
DK		 990}
991
c39a7f84
TO		 992/*
993 * Try to release metadata pages (indirect blocks, directories) which are
994 * mapped via the block device. Since these pages could have journal heads
995 * which would prevent try_to_free_buffers() from freeing them, we must use
996 * jbd2 layer's try_to_free_buffers() function to release them.
997 */
0b8e58a1
AD		 998static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
999 gfp_t wait)
c39a7f84
TO		 1000{
1001 journal_t *journal = EXT4_SB(sb)->s_journal;
1002
1003 WARN_ON(PageChecked(page));
1004 if (!page_has_buffers(page))
1005 return 0;
1006 if (journal)
1007 return jbd2_journal_try_to_free_buffers(journal, page,
1008 wait & ~__GFP_WAIT);
1009 return try_to_free_buffers(page);
1010}
1011
ac27a0ec 1012#ifdef CONFIG_QUOTA
af5bc92d 1013#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1014#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1015
617ba13b
MC		 1016static int ext4_write_dquot(struct dquot *dquot);
1017static int ext4_acquire_dquot(struct dquot *dquot);
1018static int ext4_release_dquot(struct dquot *dquot);
1019static int ext4_mark_dquot_dirty(struct dquot *dquot);
1020static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1021static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1022 struct path *path);
7c319d32
AK		 1023static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1024 int format_id);
ca0e05e4 1025static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1026static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1027static int ext4_quota_on_mount(struct super_block *sb, int type);
1028static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1029 size_t len, loff_t off);
617ba13b 1030static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1031 const char *data, size_t len, loff_t off);
7c319d32
AK		 1032static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1033 unsigned int flags);
1034static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1035
61e225dc 1036static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1037 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1038 .write_dquot = ext4_write_dquot,
1039 .acquire_dquot = ext4_acquire_dquot,
1040 .release_dquot = ext4_release_dquot,
1041 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1042 .write_info = ext4_write_info,
1043 .alloc_dquot = dquot_alloc,
1044 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1045};
1046
0d54b217 1047static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1048 .quota_on = ext4_quota_on,
ca0e05e4 1049 .quota_off = ext4_quota_off,
287a8095
CH		 1050 .quota_sync = dquot_quota_sync,
1051 .get_info = dquot_get_dqinfo,
1052 .set_info = dquot_set_dqinfo,
1053 .get_dqblk = dquot_get_dqblk,
1054 .set_dqblk = dquot_set_dqblk
ac27a0ec 1055};
7c319d32
AK		 1056
1057static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1058 .quota_on_meta = ext4_quota_on_sysfile,
1059 .quota_off = ext4_quota_off_sysfile,
1060 .quota_sync = dquot_quota_sync,
1061 .get_info = dquot_get_dqinfo,
1062 .set_info = dquot_set_dqinfo,
1063 .get_dqblk = dquot_get_dqblk,
1064 .set_dqblk = dquot_set_dqblk
1065};
ac27a0ec
DK		 1066#endif
1067
ee9b6d61 1068static const struct super_operations ext4_sops = {
617ba13b
MC		 1069 .alloc_inode = ext4_alloc_inode,
1070 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1071 .write_inode = ext4_write_inode,
1072 .dirty_inode = ext4_dirty_inode,
7ff9c073 1073 .drop_inode = ext4_drop_inode,
0930fcc1 1074 .evict_inode = ext4_evict_inode,
617ba13b 1075 .put_super = ext4_put_super,
617ba13b 1076 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1077 .freeze_fs = ext4_freeze,
1078 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1079 .statfs = ext4_statfs,
1080 .remount_fs = ext4_remount,
617ba13b 1081 .show_options = ext4_show_options,
ac27a0ec 1082#ifdef CONFIG_QUOTA
617ba13b
MC		 1083 .quota_read = ext4_quota_read,
1084 .quota_write = ext4_quota_write,
ac27a0ec 1085#endif
c39a7f84 1086 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1087};
1088
9ca92389
TT		 1089static const struct super_operations ext4_nojournal_sops = {
1090 .alloc_inode = ext4_alloc_inode,
1091 .destroy_inode = ext4_destroy_inode,
1092 .write_inode = ext4_write_inode,
1093 .dirty_inode = ext4_dirty_inode,
7ff9c073 1094 .drop_inode = ext4_drop_inode,
0930fcc1 1095 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1096 .put_super = ext4_put_super,
1097 .statfs = ext4_statfs,
1098 .remount_fs = ext4_remount,
9ca92389
TT		 1099 .show_options = ext4_show_options,
1100#ifdef CONFIG_QUOTA
1101 .quota_read = ext4_quota_read,
1102 .quota_write = ext4_quota_write,
1103#endif
1104 .bdev_try_to_free_page = bdev_try_to_free_page,
1105};
1106
39655164 1107static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1108 .fh_to_dentry = ext4_fh_to_dentry,
1109 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1110 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1111};
1112
1113enum {
1114 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1115 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1116 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1117 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1118 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1119 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1120 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1121 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1122 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1123 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1124 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1125 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1126 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1127 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1128 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1129 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1130 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1131 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
df981d03 1132 Opt_max_dir_size_kb,
ac27a0ec
DK		 1133};
1134
a447c093 1135static const match_table_t tokens = {
ac27a0ec
DK		 1136 {Opt_bsd_df, "bsddf"},
1137 {Opt_minix_df, "minixdf"},
1138 {Opt_grpid, "grpid"},
1139 {Opt_grpid, "bsdgroups"},
1140 {Opt_nogrpid, "nogrpid"},
1141 {Opt_nogrpid, "sysvgroups"},
1142 {Opt_resgid, "resgid=%u"},
1143 {Opt_resuid, "resuid=%u"},
1144 {Opt_sb, "sb=%u"},
1145 {Opt_err_cont, "errors=continue"},
1146 {Opt_err_panic, "errors=panic"},
1147 {Opt_err_ro, "errors=remount-ro"},
1148 {Opt_nouid32, "nouid32"},
ac27a0ec 1149 {Opt_debug, "debug"},
72578c33
TT		 1150 {Opt_removed, "oldalloc"},
1151 {Opt_removed, "orlov"},
ac27a0ec
DK		 1152 {Opt_user_xattr, "user_xattr"},
1153 {Opt_nouser_xattr, "nouser_xattr"},
1154 {Opt_acl, "acl"},
1155 {Opt_noacl, "noacl"},
e3bb52ae 1156 {Opt_noload, "norecovery"},
5a916be1 1157 {Opt_noload, "noload"},
72578c33
TT		 1158 {Opt_removed, "nobh"},
1159 {Opt_removed, "bh"},
ac27a0ec 1160 {Opt_commit, "commit=%u"},
30773840
TT		 1161 {Opt_min_batch_time, "min_batch_time=%u"},
1162 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1163 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1164 {Opt_journal_checksum, "journal_checksum"},
1165 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1166 {Opt_abort, "abort"},
1167 {Opt_data_journal, "data=journal"},
1168 {Opt_data_ordered, "data=ordered"},
1169 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1170 {Opt_data_err_abort, "data_err=abort"},
1171 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1172 {Opt_offusrjquota, "usrjquota="},
1173 {Opt_usrjquota, "usrjquota=%s"},
1174 {Opt_offgrpjquota, "grpjquota="},
1175 {Opt_grpjquota, "grpjquota=%s"},
1176 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1177 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1178 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1179 {Opt_grpquota, "grpquota"},
1180 {Opt_noquota, "noquota"},
1181 {Opt_quota, "quota"},
1182 {Opt_usrquota, "usrquota"},
1183 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1184 {Opt_barrier, "barrier"},
1185 {Opt_nobarrier, "nobarrier"},
25ec56b5 1186 {Opt_i_version, "i_version"},
c9de560d 1187 {Opt_stripe, "stripe=%u"},
64769240 1188 {Opt_delalloc, "delalloc"},
dd919b98 1189 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1190 {Opt_removed, "mblk_io_submit"},
1191 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1192 {Opt_block_validity, "block_validity"},
1193 {Opt_noblock_validity, "noblock_validity"},
240799cd 1194 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1195 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1196 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1197 {Opt_auto_da_alloc, "auto_da_alloc"},
1198 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1199 {Opt_dioread_nolock, "dioread_nolock"},
1200 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1201 {Opt_discard, "discard"},
1202 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1203 {Opt_init_itable, "init_itable=%u"},
1204 {Opt_init_itable, "init_itable"},
1205 {Opt_noinit_itable, "noinit_itable"},
df981d03 1206 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1207 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1208 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1209 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1210 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1211 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1212 {Opt_err, NULL},
ac27a0ec
DK		 1213};
1214
617ba13b 1215static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1216{
617ba13b 1217 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1218 char *options = (char *) *data;
1219
1220 if (!options || strncmp(options, "sb=", 3) != 0)
1221 return 1; /* Default location */
0b8e58a1 1222
ac27a0ec 1223 options += 3;
0b8e58a1 1224 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1225 sb_block = simple_strtoul(options, &options, 0);
1226 if (*options && *options != ',') {
4776004f 1227 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1228 (char *) *data);
1229 return 1;
1230 }
1231 if (*options == ',')
1232 options++;
1233 *data = (void *) options;
0b8e58a1 1234
ac27a0ec
DK		 1235 return sb_block;
1236}
1237
b3881f74 1238#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1239static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1240 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1241
56c50f11
DM		 1242#ifdef CONFIG_QUOTA
1243static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1244{
1245 struct ext4_sb_info *sbi = EXT4_SB(sb);
1246 char *qname;
03dafb5f 1247 int ret = -1;
56c50f11
DM		 1248
1249 if (sb_any_quota_loaded(sb) &&
1250 !sbi->s_qf_names[qtype]) {
1251 ext4_msg(sb, KERN_ERR,
1252 "Cannot change journaled "
1253 "quota options when quota turned on");
57f73c2c 1254 return -1;
56c50f11 1255 }
262b4662
JK		 1256 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1257 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1258 "when QUOTA feature is enabled");
1259 return -1;
1260 }
56c50f11
DM		 1261 qname = match_strdup(args);
1262 if (!qname) {
1263 ext4_msg(sb, KERN_ERR,
1264 "Not enough memory for storing quotafile name");
57f73c2c 1265 return -1;
56c50f11 1266 }
03dafb5f
CG		 1267 if (sbi->s_qf_names[qtype]) {
1268 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1269 ret = 1;
1270 else
1271 ext4_msg(sb, KERN_ERR,
1272 "%s quota file already specified",
1273 QTYPE2NAME(qtype));
1274 goto errout;
56c50f11 1275 }
03dafb5f 1276 if (strchr(qname, '/')) {
56c50f11
DM		 1277 ext4_msg(sb, KERN_ERR,
1278 "quotafile must be on filesystem root");
03dafb5f 1279 goto errout;
56c50f11 1280 }
03dafb5f 1281 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1282 set_opt(sb, QUOTA);
56c50f11 1283 return 1;
03dafb5f
CG		 1284errout:
1285 kfree(qname);
1286 return ret;
56c50f11
DM		 1287}
1288
1289static int clear_qf_name(struct super_block *sb, int qtype)
1290{
1291
1292 struct ext4_sb_info *sbi = EXT4_SB(sb);
1293
1294 if (sb_any_quota_loaded(sb) &&
1295 sbi->s_qf_names[qtype]) {
1296 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1297 " when quota turned on");
57f73c2c 1298 return -1;
56c50f11 1299 }
03dafb5f 1300 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1301 sbi->s_qf_names[qtype] = NULL;
1302 return 1;
1303}
1304#endif
1305
26092bf5
TT		 1306#define MOPT_SET 0x0001
1307#define MOPT_CLEAR 0x0002
1308#define MOPT_NOSUPPORT 0x0004
1309#define MOPT_EXPLICIT 0x0008
1310#define MOPT_CLEAR_ERR 0x0010
1311#define MOPT_GTE0 0x0020
ac27a0ec 1312#ifdef CONFIG_QUOTA
26092bf5
TT		 1313#define MOPT_Q 0
1314#define MOPT_QFMT 0x0040
1315#else
1316#define MOPT_Q MOPT_NOSUPPORT
1317#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1318#endif
26092bf5 1319#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1320#define MOPT_NO_EXT2 0x0100
1321#define MOPT_NO_EXT3 0x0200
1322#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
26092bf5
TT		 1323
1324static const struct mount_opts {
1325 int token;
1326 int mount_opt;
1327 int flags;
1328} ext4_mount_opts[] = {
1329 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1330 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1331 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1332 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1333 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1334 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1335 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1336 MOPT_EXT4_ONLY | MOPT_SET},
1337 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1338 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1339 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1340 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1341 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1342 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1343 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1344 MOPT_EXT4_ONLY | MOPT_CLEAR | MOPT_EXPLICIT},
1345 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1346 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1347 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1348 EXT4_MOUNT_JOURNAL_CHECKSUM),
1349 MOPT_EXT4_ONLY | MOPT_SET},
1350 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1351 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1352 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1353 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1354 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1355 MOPT_NO_EXT2 | MOPT_SET},
1356 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1357 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1358 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1359 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1360 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1361 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1362 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1363 {Opt_commit, 0, MOPT_GTE0},
1364 {Opt_max_batch_time, 0, MOPT_GTE0},
1365 {Opt_min_batch_time, 0, MOPT_GTE0},
1366 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1367 {Opt_init_itable, 0, MOPT_GTE0},
1368 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1369 {Opt_resuid, 0, MOPT_GTE0},
1370 {Opt_resgid, 0, MOPT_GTE0},
1371 {Opt_journal_dev, 0, MOPT_GTE0},
1372 {Opt_journal_ioprio, 0, MOPT_GTE0},
8dc0aa8c
TT		 1373 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1374 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1375 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1376 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1377 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1378 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1379#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1380 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1381 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1382#else
26092bf5
TT		 1383 {Opt_acl, 0, MOPT_NOSUPPORT},
1384 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1385#endif
26092bf5
TT		 1386 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1387 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1388 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1389 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1390 MOPT_SET | MOPT_Q},
1391 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1392 MOPT_SET | MOPT_Q},
1393 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1394 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1395 {Opt_usrjquota, 0, MOPT_Q},
1396 {Opt_grpjquota, 0, MOPT_Q},
1397 {Opt_offusrjquota, 0, MOPT_Q},
1398 {Opt_offgrpjquota, 0, MOPT_Q},
1399 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1400 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1401 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1402 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1403 {Opt_err, 0, 0}
1404};
1405
1406static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1407 substring_t *args, unsigned long *journal_devnum,
1408 unsigned int *journal_ioprio, int is_remount)
1409{
1410 struct ext4_sb_info *sbi = EXT4_SB(sb);
1411 const struct mount_opts *m;
08cefc7a
EB		 1412 kuid_t uid;
1413 kgid_t gid;
26092bf5
TT		 1414 int arg = 0;
1415
57f73c2c
TT		 1416#ifdef CONFIG_QUOTA
1417 if (token == Opt_usrjquota)
1418 return set_qf_name(sb, USRQUOTA, &args[0]);
1419 else if (token == Opt_grpjquota)
1420 return set_qf_name(sb, GRPQUOTA, &args[0]);
1421 else if (token == Opt_offusrjquota)
1422 return clear_qf_name(sb, USRQUOTA);
1423 else if (token == Opt_offgrpjquota)
1424 return clear_qf_name(sb, GRPQUOTA);
1425#endif
26092bf5 1426 switch (token) {
f7048605
TT		 1427 case Opt_noacl:
1428 case Opt_nouser_xattr:
1429 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1430 break;
26092bf5
TT		 1431 case Opt_sb:
1432 return 1; /* handled by get_sb_block() */
1433 case Opt_removed:
5f3633e3 1434 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1435 return 1;
26092bf5
TT		 1436 case Opt_abort:
1437 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1438 return 1;
1439 case Opt_i_version:
1440 sb->s_flags |= MS_I_VERSION;
1441 return 1;
26092bf5
TT		 1442 }
1443
5f3633e3
JK		 1444 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1445 if (token == m->token)
1446 break;
1447
1448 if (m->token == Opt_err) {
1449 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1450 "or missing value", opt);
1451 return -1;
1452 }
1453
8dc0aa8c
TT		 1454 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1455 ext4_msg(sb, KERN_ERR,
1456 "Mount option \"%s\" incompatible with ext2", opt);
1457 return -1;
1458 }
1459 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1460 ext4_msg(sb, KERN_ERR,
1461 "Mount option \"%s\" incompatible with ext3", opt);
1462 return -1;
1463 }
1464
5f3633e3
JK		 1465 if (args->from && match_int(args, &arg))
1466 return -1;
1467 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1468 return -1;
1469 if (m->flags & MOPT_EXPLICIT)
1470 set_opt2(sb, EXPLICIT_DELALLOC);
1471 if (m->flags & MOPT_CLEAR_ERR)
1472 clear_opt(sb, ERRORS_MASK);
1473 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1474 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1475 "options when quota turned on");
1476 return -1;
1477 }
1478
1479 if (m->flags & MOPT_NOSUPPORT) {
1480 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1481 } else if (token == Opt_commit) {
1482 if (arg == 0)
1483 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1484 sbi->s_commit_interval = HZ * arg;
1485 } else if (token == Opt_max_batch_time) {
1486 if (arg == 0)
1487 arg = EXT4_DEF_MAX_BATCH_TIME;
1488 sbi->s_max_batch_time = arg;
1489 } else if (token == Opt_min_batch_time) {
1490 sbi->s_min_batch_time = arg;
1491 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1492 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1493 ext4_msg(sb, KERN_ERR,
1494 "EXT4-fs: inode_readahead_blks must be "
1495 "0 or a power of 2 smaller than 2^31");
26092bf5 1496 return -1;
5f3633e3
JK		 1497 }
1498 sbi->s_inode_readahead_blks = arg;
1499 } else if (token == Opt_init_itable) {
1500 set_opt(sb, INIT_INODE_TABLE);
1501 if (!args->from)
1502 arg = EXT4_DEF_LI_WAIT_MULT;
1503 sbi->s_li_wait_mult = arg;
1504 } else if (token == Opt_max_dir_size_kb) {
1505 sbi->s_max_dir_size_kb = arg;
1506 } else if (token == Opt_stripe) {
1507 sbi->s_stripe = arg;
1508 } else if (token == Opt_resuid) {
1509 uid = make_kuid(current_user_ns(), arg);
1510 if (!uid_valid(uid)) {
1511 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1512 return -1;
1513 }
5f3633e3
JK		 1514 sbi->s_resuid = uid;
1515 } else if (token == Opt_resgid) {
1516 gid = make_kgid(current_user_ns(), arg);
1517 if (!gid_valid(gid)) {
1518 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1519 return -1;
1520 }
1521 sbi->s_resgid = gid;
1522 } else if (token == Opt_journal_dev) {
1523 if (is_remount) {
1524 ext4_msg(sb, KERN_ERR,
1525 "Cannot specify journal on remount");
1526 return -1;
1527 }
1528 *journal_devnum = arg;
1529 } else if (token == Opt_journal_ioprio) {
1530 if (arg > 7) {
1531 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1532 " (must be 0-7)");
1533 return -1;
1534 }
1535 *journal_ioprio =
1536 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1537 } else if (m->flags & MOPT_DATAJ) {
1538 if (is_remount) {
1539 if (!sbi->s_journal)
1540 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1541 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1542 ext4_msg(sb, KERN_ERR,
26092bf5 1543 "Cannot change data mode on remount");
26092bf5 1544 return -1;
ac27a0ec 1545 }
26092bf5 1546 } else {
5f3633e3
JK		 1547 clear_opt(sb, DATA_FLAGS);
1548 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1549 }
5f3633e3
JK		 1550#ifdef CONFIG_QUOTA
1551 } else if (m->flags & MOPT_QFMT) {
1552 if (sb_any_quota_loaded(sb) &&
1553 sbi->s_jquota_fmt != m->mount_opt) {
1554 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1555 "quota options when quota turned on");
1556 return -1;
1557 }
262b4662
JK		 1558 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1559 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1560 ext4_msg(sb, KERN_ERR,
1561 "Cannot set journaled quota options "
1562 "when QUOTA feature is enabled");
1563 return -1;
1564 }
5f3633e3
JK		 1565 sbi->s_jquota_fmt = m->mount_opt;
1566#endif
1567 } else {
1568 if (!args->from)
1569 arg = 1;
1570 if (m->flags & MOPT_CLEAR)
1571 arg = !arg;
1572 else if (unlikely(!(m->flags & MOPT_SET))) {
1573 ext4_msg(sb, KERN_WARNING,
1574 "buggy handling of option %s", opt);
1575 WARN_ON(1);
1576 return -1;
1577 }
1578 if (arg != 0)
1579 sbi->s_mount_opt |= m->mount_opt;
1580 else
1581 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1582 }
5f3633e3 1583 return 1;
26092bf5
TT		 1584}
1585
1586static int parse_options(char *options, struct super_block *sb,
1587 unsigned long *journal_devnum,
1588 unsigned int *journal_ioprio,
1589 int is_remount)
1590{
1591 struct ext4_sb_info *sbi = EXT4_SB(sb);
1592 char *p;
1593 substring_t args[MAX_OPT_ARGS];
1594 int token;
1595
1596 if (!options)
1597 return 1;
1598
1599 while ((p = strsep(&options, ",")) != NULL) {
1600 if (!*p)
1601 continue;
1602 /*
1603 * Initialize args struct so we know whether arg was
1604 * found; some options take optional arguments.
1605 */
caecd0af 1606 args[0].to = args[0].from = NULL;
26092bf5
TT		 1607 token = match_token(p, tokens, args);
1608 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1609 journal_ioprio, is_remount) < 0)
1610 return 0;
ac27a0ec
DK		 1611 }
1612#ifdef CONFIG_QUOTA
262b4662
JK		 1613 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1614 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1615 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1616 "feature is enabled");
1617 return 0;
1618 }
ac27a0ec 1619 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1620 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1621 clear_opt(sb, USRQUOTA);
ac27a0ec 1622
482a7425 1623 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1624 clear_opt(sb, GRPQUOTA);
ac27a0ec 1625
56c50f11 1626 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1627 ext4_msg(sb, KERN_ERR, "old and new quota "
1628 "format mixing");
ac27a0ec
DK		 1629 return 0;
1630 }
1631
1632 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1633 ext4_msg(sb, KERN_ERR, "journaled quota format "
1634 "not specified");
ac27a0ec
DK		 1635 return 0;
1636 }
1637 } else {
1638 if (sbi->s_jquota_fmt) {
b31e1552 1639 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1640 "specified with no journaling "
b31e1552 1641 "enabled");
ac27a0ec
DK		 1642 return 0;
1643 }
1644 }
1645#endif
261cb20c
JK		 1646 if (test_opt(sb, DIOREAD_NOLOCK)) {
1647 int blocksize =
1648 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1649
1650 if (blocksize < PAGE_CACHE_SIZE) {
1651 ext4_msg(sb, KERN_ERR, "can't mount with "
1652 "dioread_nolock if block size != PAGE_SIZE");
1653 return 0;
1654 }
1655 }
ac27a0ec
DK		 1656 return 1;
1657}
1658
2adf6da8
TT		 1659static inline void ext4_show_quota_options(struct seq_file *seq,
1660 struct super_block *sb)
1661{
1662#if defined(CONFIG_QUOTA)
1663 struct ext4_sb_info *sbi = EXT4_SB(sb);
1664
1665 if (sbi->s_jquota_fmt) {
1666 char *fmtname = "";
1667
1668 switch (sbi->s_jquota_fmt) {
1669 case QFMT_VFS_OLD:
1670 fmtname = "vfsold";
1671 break;
1672 case QFMT_VFS_V0:
1673 fmtname = "vfsv0";
1674 break;
1675 case QFMT_VFS_V1:
1676 fmtname = "vfsv1";
1677 break;
1678 }
1679 seq_printf(seq, ",jqfmt=%s", fmtname);
1680 }
1681
1682 if (sbi->s_qf_names[USRQUOTA])
1683 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1684
1685 if (sbi->s_qf_names[GRPQUOTA])
1686 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1687
1688 if (test_opt(sb, USRQUOTA))
1689 seq_puts(seq, ",usrquota");
1690
1691 if (test_opt(sb, GRPQUOTA))
1692 seq_puts(seq, ",grpquota");
1693#endif
1694}
1695
5a916be1
TT		 1696static const char *token2str(int token)
1697{
50df9fd5 1698 const struct match_token *t;
5a916be1
TT		 1699
1700 for (t = tokens; t->token != Opt_err; t++)
1701 if (t->token == token && !strchr(t->pattern, '='))
1702 break;
1703 return t->pattern;
1704}
1705
2adf6da8
TT		 1706/*
1707 * Show an option if
1708 * - it's set to a non-default value OR
1709 * - if the per-sb default is different from the global default
1710 */
66acdcf4
TT		 1711static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1712 int nodefs)
2adf6da8 1713{
2adf6da8
TT		 1714 struct ext4_sb_info *sbi = EXT4_SB(sb);
1715 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1716 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1717 const struct mount_opts *m;
66acdcf4 1718 char sep = nodefs ? '\n' : ',';
2adf6da8 1719
66acdcf4
TT		 1720#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1721#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1722
1723 if (sbi->s_sb_block != 1)
5a916be1
TT		 1724 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1725
1726 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1727 int want_set = m->flags & MOPT_SET;
1728 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1729 (m->flags & MOPT_CLEAR_ERR))
1730 continue;
66acdcf4 1731 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1732 continue; /* skip if same as the default */
1733 if ((want_set &&
1734 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1735 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1736 continue; /* select Opt_noFoo vs Opt_Foo */
1737 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1738 }
5a916be1 1739
08cefc7a 1740 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1741 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1742 SEQ_OPTS_PRINT("resuid=%u",
1743 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1744 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1745 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1746 SEQ_OPTS_PRINT("resgid=%u",
1747 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1748 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1749 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1750 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1751 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1752 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1753 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1754 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1755 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1756 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1757 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1758 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1759 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1760 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1761 if (sb->s_flags & MS_I_VERSION)
5a916be1 1762 SEQ_OPTS_PUTS("i_version");
66acdcf4 1763 if (nodefs || sbi->s_stripe)
5a916be1 1764 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1765 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1766 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1767 SEQ_OPTS_PUTS("data=journal");
1768 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1769 SEQ_OPTS_PUTS("data=ordered");
1770 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1771 SEQ_OPTS_PUTS("data=writeback");
1772 }
66acdcf4
TT		 1773 if (nodefs ||
1774 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1775 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1776 sbi->s_inode_readahead_blks);
2adf6da8 1777
66acdcf4
TT		 1778 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1779 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1780 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1781 if (nodefs || sbi->s_max_dir_size_kb)
1782 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1783
1784 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1785 return 0;
1786}
1787
66acdcf4
TT		 1788static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1789{
1790 return _ext4_show_options(seq, root->d_sb, 0);
1791}
1792
1793static int options_seq_show(struct seq_file *seq, void *offset)
1794{
1795 struct super_block *sb = seq->private;
1796 int rc;
1797
1798 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1799 rc = _ext4_show_options(seq, sb, 1);
1800 seq_puts(seq, "\n");
1801 return rc;
1802}
1803
1804static int options_open_fs(struct inode *inode, struct file *file)
1805{
d9dda78b 1806 return single_open(file, options_seq_show, PDE_DATA(inode));
66acdcf4
TT		 1807}
1808
1809static const struct file_operations ext4_seq_options_fops = {
1810 .owner = THIS_MODULE,
1811 .open = options_open_fs,
1812 .read = seq_read,
1813 .llseek = seq_lseek,
1814 .release = single_release,
1815};
1816
617ba13b 1817static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1818 int read_only)
1819{
617ba13b 1820 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1821 int res = 0;
1822
617ba13b 1823 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1824 ext4_msg(sb, KERN_ERR, "revision level too high, "
1825 "forcing read-only mode");
ac27a0ec
DK		 1826 res = MS_RDONLY;
1827 }
1828 if (read_only)
281b5995 1829 goto done;
617ba13b 1830 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1831 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1832 "running e2fsck is recommended");
617ba13b 1833 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1834 ext4_msg(sb, KERN_WARNING,
1835 "warning: mounting fs with errors, "
1836 "running e2fsck is recommended");
ed3ce80a 1837 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1838 le16_to_cpu(es->s_mnt_count) >=
1839 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1840 ext4_msg(sb, KERN_WARNING,
1841 "warning: maximal mount count reached, "
1842 "running e2fsck is recommended");
ac27a0ec
DK		 1843 else if (le32_to_cpu(es->s_checkinterval) &&
1844 (le32_to_cpu(es->s_lastcheck) +
1845 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1846 ext4_msg(sb, KERN_WARNING,
1847 "warning: checktime reached, "
1848 "running e2fsck is recommended");
0b8e58a1 1849 if (!sbi->s_journal)
0390131b 1850 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1851 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1852 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1853 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1854 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1855 ext4_update_dynamic_rev(sb);
0390131b
FM		 1856 if (sbi->s_journal)
1857 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1858
e2d67052 1859 ext4_commit_super(sb, 1);
281b5995 1860done:
ac27a0ec 1861 if (test_opt(sb, DEBUG))
a9df9a49 1862 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1863 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1864 sb->s_blocksize,
1865 sbi->s_groups_count,
617ba13b
MC		 1866 EXT4_BLOCKS_PER_GROUP(sb),
1867 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1868 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1869
7abc52c2 1870 cleancache_init_fs(sb);
ac27a0ec
DK		 1871 return res;
1872}
1873
117fff10
TT		 1874int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1875{
1876 struct ext4_sb_info *sbi = EXT4_SB(sb);
1877 struct flex_groups *new_groups;
1878 int size;
1879
1880 if (!sbi->s_log_groups_per_flex)
1881 return 0;
1882
1883 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1884 if (size <= sbi->s_flex_groups_allocated)
1885 return 0;
1886
1887 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1888 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1889 if (!new_groups) {
1890 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1891 size / (int) sizeof(struct flex_groups));
1892 return -ENOMEM;
1893 }
1894
1895 if (sbi->s_flex_groups) {
1896 memcpy(new_groups, sbi->s_flex_groups,
1897 (sbi->s_flex_groups_allocated *
1898 sizeof(struct flex_groups)));
1899 ext4_kvfree(sbi->s_flex_groups);
1900 }
1901 sbi->s_flex_groups = new_groups;
1902 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1903 return 0;
1904}
1905
772cb7c8
JS		 1906static int ext4_fill_flex_info(struct super_block *sb)
1907{
1908 struct ext4_sb_info *sbi = EXT4_SB(sb);
1909 struct ext4_group_desc *gdp = NULL;
772cb7c8 1910 ext4_group_t flex_group;
d50f2ab6 1911 unsigned int groups_per_flex = 0;
117fff10 1912 int i, err;
772cb7c8 1913
503358ae 1914 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1915 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1916 sbi->s_log_groups_per_flex = 0;
1917 return 1;
1918 }
76495ec1 1919 groups_per_flex = 1U << sbi->s_log_groups_per_flex;
772cb7c8 1920
117fff10
TT		 1921 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1922 if (err)
9933fc0a 1923 goto failed;
772cb7c8 1924
772cb7c8 1925 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1926 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1927
1928 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1929 atomic_add(ext4_free_inodes_count(sb, gdp),
1930 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 1931 atomic64_add(ext4_free_group_clusters(sb, gdp),
1932 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1933 atomic_add(ext4_used_dirs_count(sb, gdp),
1934 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1935 }
1936
1937 return 1;
1938failed:
1939 return 0;
1940}
1941
feb0ab32
DW		 1942static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1943 struct ext4_group_desc *gdp)
717d50e4 1944{
feb0ab32 1945 int offset;
717d50e4 1946 __u16 crc = 0;
feb0ab32 1947 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1948
feb0ab32
DW		 1949 if ((sbi->s_es->s_feature_ro_compat &
1950 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1951 /* Use new metadata_csum algorithm */
d6a77105 1952 __le16 save_csum;
feb0ab32
DW		 1953 __u32 csum32;
1954
d6a77105 1955 save_csum = gdp->bg_checksum;
feb0ab32
DW		 1956 gdp->bg_checksum = 0;
1957 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1958 sizeof(le_group));
1959 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1960 sbi->s_desc_size);
d6a77105 1961 gdp->bg_checksum = save_csum;
feb0ab32
DW		 1962
1963 crc = csum32 & 0xFFFF;
1964 goto out;
717d50e4
AD		 1965 }
1966
feb0ab32
DW		 1967 /* old crc16 code */
1968 offset = offsetof(struct ext4_group_desc, bg_checksum);
1969
1970 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1971 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1972 crc = crc16(crc, (__u8 *)gdp, offset);
1973 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1974 /* for checksum of struct ext4_group_desc do the rest...*/
1975 if ((sbi->s_es->s_feature_incompat &
1976 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1977 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1978 crc = crc16(crc, (__u8 *)gdp + offset,
1979 le16_to_cpu(sbi->s_es->s_desc_size) -
1980 offset);
1981
1982out:
717d50e4
AD		 1983 return cpu_to_le16(crc);
1984}
1985
feb0ab32 1986int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 1987 struct ext4_group_desc *gdp)
1988{
feb0ab32
DW		 1989 if (ext4_has_group_desc_csum(sb) &&
1990 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
1991 block_group, gdp)))
717d50e4
AD		 1992 return 0;
1993
1994 return 1;
1995}
1996
feb0ab32
DW		 1997void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
1998 struct ext4_group_desc *gdp)
1999{
2000 if (!ext4_has_group_desc_csum(sb))
2001 return;
2002 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2003}
2004
ac27a0ec 2005/* Called at mount-time, super-block is locked */
bfff6873
LC		 2006static int ext4_check_descriptors(struct super_block *sb,
2007 ext4_group_t *first_not_zeroed)
ac27a0ec 2008{
617ba13b
MC		 2009 struct ext4_sb_info *sbi = EXT4_SB(sb);
2010 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2011 ext4_fsblk_t last_block;
bd81d8ee
LV		 2012 ext4_fsblk_t block_bitmap;
2013 ext4_fsblk_t inode_bitmap;
2014 ext4_fsblk_t inode_table;
ce421581 2015 int flexbg_flag = 0;
bfff6873 2016 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2017
ce421581
JS		 2018 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2019 flexbg_flag = 1;
2020
af5bc92d 2021 ext4_debug("Checking group descriptors");
ac27a0ec 2022
197cd65a
AM		 2023 for (i = 0; i < sbi->s_groups_count; i++) {
2024 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2025
ce421581 2026 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2027 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2028 else
2029 last_block = first_block +
617ba13b 2030 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2031
bfff6873
LC		 2032 if ((grp == sbi->s_groups_count) &&
2033 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2034 grp = i;
2035
8fadc143 2036 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2037 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2038 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2039 "Block bitmap for group %u not in group "
b31e1552 2040 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2041 return 0;
2042 }
8fadc143 2043 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2044 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2045 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2046 "Inode bitmap for group %u not in group "
b31e1552 2047 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2048 return 0;
2049 }
8fadc143 2050 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2051 if (inode_table < first_block ||
2b2d6d01 2052 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2053 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2054 "Inode table for group %u not in group "
b31e1552 2055 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2056 return 0;
2057 }
955ce5f5 2058 ext4_lock_group(sb, i);
feb0ab32 2059 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2060 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2061 "Checksum for group %u failed (%u!=%u)",
2062 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2063 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2064 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2065 ext4_unlock_group(sb, i);
8a266467 2066 return 0;
7ee1ec4c 2067 }
717d50e4 2068 }
955ce5f5 2069 ext4_unlock_group(sb, i);
ce421581
JS		 2070 if (!flexbg_flag)
2071 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2072 }
bfff6873
LC		 2073 if (NULL != first_not_zeroed)
2074 *first_not_zeroed = grp;
ac27a0ec 2075
5dee5437
TT		 2076 ext4_free_blocks_count_set(sbi->s_es,
2077 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2078 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2079 return 1;
2080}
2081
617ba13b 2082/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2083 * the superblock) which were deleted from all directories, but held open by
2084 * a process at the time of a crash. We walk the list and try to delete these
2085 * inodes at recovery time (only with a read-write filesystem).
2086 *
2087 * In order to keep the orphan inode chain consistent during traversal (in
2088 * case of crash during recovery), we link each inode into the superblock
2089 * orphan list_head and handle it the same way as an inode deletion during
2090 * normal operation (which journals the operations for us).
2091 *
2092 * We only do an iget() and an iput() on each inode, which is very safe if we
2093 * accidentally point at an in-use or already deleted inode. The worst that
2094 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2095 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2096 * e2fsck was run on this filesystem, and it must have already done the orphan
2097 * inode cleanup for us, so we can safely abort without any further action.
2098 */
2b2d6d01
TT		 2099static void ext4_orphan_cleanup(struct super_block *sb,
2100 struct ext4_super_block *es)
ac27a0ec
DK		 2101{
2102 unsigned int s_flags = sb->s_flags;
2103 int nr_orphans = 0, nr_truncates = 0;
2104#ifdef CONFIG_QUOTA
2105 int i;
2106#endif
2107 if (!es->s_last_orphan) {
2108 jbd_debug(4, "no orphan inodes to clean up\n");
2109 return;
2110 }
2111
a8f48a95 2112 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2113 ext4_msg(sb, KERN_ERR, "write access "
2114 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2115 return;
2116 }
2117
d39195c3
AG		 2118 /* Check if feature set would not allow a r/w mount */
2119 if (!ext4_feature_set_ok(sb, 0)) {
2120 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2121 "unknown ROCOMPAT features");
2122 return;
2123 }
2124
617ba13b 2125 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2126 /* don't clear list on RO mount w/ errors */
2127 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
ac27a0ec
DK		 2128 jbd_debug(1, "Errors on filesystem, "
2129 "clearing orphan list.\n");
c25f9bc6
ES		 2130 es->s_last_orphan = 0;
2131 }
ac27a0ec
DK		 2132 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2133 return;
2134 }
2135
2136 if (s_flags & MS_RDONLY) {
b31e1552 2137 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2138 sb->s_flags &= ~MS_RDONLY;
2139 }
2140#ifdef CONFIG_QUOTA
2141 /* Needed for iput() to work correctly and not trash data */
2142 sb->s_flags |= MS_ACTIVE;
2143 /* Turn on quotas so that they are updated correctly */
2144 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2145 if (EXT4_SB(sb)->s_qf_names[i]) {
2146 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2147 if (ret < 0)
b31e1552
ES		 2148 ext4_msg(sb, KERN_ERR,
2149 "Cannot turn on journaled "
2150 "quota: error %d", ret);
ac27a0ec
DK		 2151 }
2152 }
2153#endif
2154
2155 while (es->s_last_orphan) {
2156 struct inode *inode;
2157
97bd42b9
JB		 2158 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2159 if (IS_ERR(inode)) {
ac27a0ec
DK		 2160 es->s_last_orphan = 0;
2161 break;
2162 }
2163
617ba13b 2164 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2165 dquot_initialize(inode);
ac27a0ec 2166 if (inode->i_nlink) {
b31e1552
ES		 2167 ext4_msg(sb, KERN_DEBUG,
2168 "%s: truncating inode %lu to %lld bytes",
46e665e9 2169 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2170 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2171 inode->i_ino, inode->i_size);
721e3eba 2172 mutex_lock(&inode->i_mutex);
617ba13b 2173 ext4_truncate(inode);
721e3eba 2174 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2175 nr_truncates++;
2176 } else {
b31e1552
ES		 2177 ext4_msg(sb, KERN_DEBUG,
2178 "%s: deleting unreferenced inode %lu",
46e665e9 2179 __func__, inode->i_ino);
ac27a0ec
DK		 2180 jbd_debug(2, "deleting unreferenced inode %lu\n",
2181 inode->i_ino);
2182 nr_orphans++;
2183 }
2184 iput(inode); /* The delete magic happens here! */
2185 }
2186
2b2d6d01 2187#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2188
2189 if (nr_orphans)
b31e1552
ES		 2190 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2191 PLURAL(nr_orphans));
ac27a0ec 2192 if (nr_truncates)
b31e1552
ES		 2193 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2194 PLURAL(nr_truncates));
ac27a0ec
DK		 2195#ifdef CONFIG_QUOTA
2196 /* Turn quotas off */
2197 for (i = 0; i < MAXQUOTAS; i++) {
2198 if (sb_dqopt(sb)->files[i])
287a8095 2199 dquot_quota_off(sb, i);
ac27a0ec
DK		 2200 }
2201#endif
2202 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2203}
0b8e58a1 2204
cd2291a4
ES		 2205/*
2206 * Maximal extent format file size.
2207 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2208 * extent format containers, within a sector_t, and within i_blocks
2209 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2210 * so that won't be a limiting factor.
2211 *
f17722f9
LC		 2212 * However there is other limiting factor. We do store extents in the form
2213 * of starting block and length, hence the resulting length of the extent
2214 * covering maximum file size must fit into on-disk format containers as
2215 * well. Given that length is always by 1 unit bigger than max unit (because
2216 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2217 *
cd2291a4
ES		 2218 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2219 */
f287a1a5 2220static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2221{
2222 loff_t res;
2223 loff_t upper_limit = MAX_LFS_FILESIZE;
2224
2225 /* small i_blocks in vfs inode? */
f287a1a5 2226 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2227 /*
90c699a9 2228 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2229 * i_block represent total blocks in 512 bytes
2230 * 32 == size of vfs inode i_blocks * 8
2231 */
2232 upper_limit = (1LL << 32) - 1;
2233
2234 /* total blocks in file system block size */
2235 upper_limit >>= (blkbits - 9);
2236 upper_limit <<= blkbits;
2237 }
2238
f17722f9
LC		 2239 /*
2240 * 32-bit extent-start container, ee_block. We lower the maxbytes
2241 * by one fs block, so ee_len can cover the extent of maximum file
2242 * size
2243 */
2244 res = (1LL << 32) - 1;
cd2291a4 2245 res <<= blkbits;
cd2291a4
ES		 2246
2247 /* Sanity check against vm- & vfs- imposed limits */
2248 if (res > upper_limit)
2249 res = upper_limit;
2250
2251 return res;
2252}
ac27a0ec 2253
ac27a0ec 2254/*
cd2291a4 2255 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2256 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2257 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2258 */
f287a1a5 2259static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2260{
617ba13b 2261 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2262 int meta_blocks;
2263 loff_t upper_limit;
0b8e58a1
AD		 2264 /* This is calculated to be the largest file size for a dense, block
2265 * mapped file such that the file's total number of 512-byte sectors,
2266 * including data and all indirect blocks, does not exceed (2^48 - 1).
2267 *
2268 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2269 * number of 512-byte sectors of the file.
0fc1b451
AK		 2270 */
2271
f287a1a5 2272 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2273 /*
90c699a9 2274 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2275 * the inode i_block field represents total file blocks in
2276 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2277 */
2278 upper_limit = (1LL << 32) - 1;
2279
2280 /* total blocks in file system block size */
2281 upper_limit >>= (bits - 9);
2282
2283 } else {
8180a562
AK		 2284 /*
2285 * We use 48 bit ext4_inode i_blocks
2286 * With EXT4_HUGE_FILE_FL set the i_blocks
2287 * represent total number of blocks in
2288 * file system block size
2289 */
0fc1b451
AK		 2290 upper_limit = (1LL << 48) - 1;
2291
0fc1b451
AK		 2292 }
2293
2294 /* indirect blocks */
2295 meta_blocks = 1;
2296 /* double indirect blocks */
2297 meta_blocks += 1 + (1LL << (bits-2));
2298 /* tripple indirect blocks */
2299 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2300
2301 upper_limit -= meta_blocks;
2302 upper_limit <<= bits;
ac27a0ec
DK		 2303
2304 res += 1LL << (bits-2);
2305 res += 1LL << (2*(bits-2));
2306 res += 1LL << (3*(bits-2));
2307 res <<= bits;
2308 if (res > upper_limit)
2309 res = upper_limit;
0fc1b451
AK		 2310
2311 if (res > MAX_LFS_FILESIZE)
2312 res = MAX_LFS_FILESIZE;
2313
ac27a0ec
DK		 2314 return res;
2315}
2316
617ba13b 2317static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2318 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2319{
617ba13b 2320 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2321 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2322 int has_super = 0;
2323
2324 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2325
617ba13b 2326 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2327 nr < first_meta_bg)
70bbb3e0 2328 return logical_sb_block + nr + 1;
ac27a0ec 2329 bg = sbi->s_desc_per_block * nr;
617ba13b 2330 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2331 has_super = 1;
0b8e58a1 2332
617ba13b 2333 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2334}
2335
c9de560d
AT		 2336/**
2337 * ext4_get_stripe_size: Get the stripe size.
2338 * @sbi: In memory super block info
2339 *
2340 * If we have specified it via mount option, then
2341 * use the mount option value. If the value specified at mount time is
2342 * greater than the blocks per group use the super block value.
2343 * If the super block value is greater than blocks per group return 0.
2344 * Allocator needs it be less than blocks per group.
2345 *
2346 */
2347static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2348{
2349 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2350 unsigned long stripe_width =
2351 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2352 int ret;
c9de560d
AT		 2353
2354 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2355 ret = sbi->s_stripe;
2356 else if (stripe_width <= sbi->s_blocks_per_group)
2357 ret = stripe_width;
2358 else if (stride <= sbi->s_blocks_per_group)
2359 ret = stride;
2360 else
2361 ret = 0;
c9de560d 2362
3eb08658
DE		 2363 /*
2364 * If the stripe width is 1, this makes no sense and
2365 * we set it to 0 to turn off stripe handling code.
2366 */
2367 if (ret <= 1)
2368 ret = 0;
c9de560d 2369
3eb08658 2370 return ret;
c9de560d 2371}
ac27a0ec 2372
3197ebdb
TT		 2373/* sysfs supprt */
2374
2375struct ext4_attr {
2376 struct attribute attr;
2377 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2378 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2379 const char *, size_t);
2380 int offset;
2381};
2382
27dd4385
LC		 2383static int parse_strtoull(const char *buf,
2384 unsigned long long max, unsigned long long *value)
2385{
2386 int ret;
2387
2388 ret = kstrtoull(skip_spaces(buf), 0, value);
2389 if (!ret && *value > max)
2390 ret = -EINVAL;
2391 return ret;
2392}
2393
3197ebdb
TT		 2394static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2395 struct ext4_sb_info *sbi,
2396 char *buf)
2397{
2398 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2399 (s64) EXT4_C2B(sbi,
2400 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2401}
2402
2403static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2404 struct ext4_sb_info *sbi, char *buf)
2405{
2406 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2407
f613dfcb
TT		 2408 if (!sb->s_bdev->bd_part)
2409 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2410 return snprintf(buf, PAGE_SIZE, "%lu\n",
2411 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2412 sbi->s_sectors_written_start) >> 1);
2413}
2414
2415static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2416 struct ext4_sb_info *sbi, char *buf)
2417{
2418 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2419
f613dfcb
TT		 2420 if (!sb->s_bdev->bd_part)
2421 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2422 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2423 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2424 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2425 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2426}
2427
2428static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2429 struct ext4_sb_info *sbi,
2430 const char *buf, size_t count)
2431{
2432 unsigned long t;
e1091b15 2433 int ret;
3197ebdb 2434
e1091b15
LC		 2435 ret = kstrtoul(skip_spaces(buf), 0, &t);
2436 if (ret)
2437 return ret;
3197ebdb 2438
e1091b15 2439 if (t && (!is_power_of_2(t) || t > 0x40000000))
3197ebdb
TT		 2440 return -EINVAL;
2441
2442 sbi->s_inode_readahead_blks = t;
2443 return count;
2444}
2445
2446static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2447 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2448{
2449 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2450
2451 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2452}
2453
2454static ssize_t sbi_ui_store(struct ext4_attr *a,
2455 struct ext4_sb_info *sbi,
2456 const char *buf, size_t count)
2457{
2458 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2459 unsigned long t;
e1091b15 2460 int ret;
3197ebdb 2461
e1091b15
LC		 2462 ret = kstrtoul(skip_spaces(buf), 0, &t);
2463 if (ret)
2464 return ret;
3197ebdb
TT		 2465 *ui = t;
2466 return count;
2467}
2468
27dd4385
LC		 2469static ssize_t reserved_clusters_show(struct ext4_attr *a,
2470 struct ext4_sb_info *sbi, char *buf)
2471{
2472 return snprintf(buf, PAGE_SIZE, "%llu\n",
2473 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2474}
2475
2476static ssize_t reserved_clusters_store(struct ext4_attr *a,
2477 struct ext4_sb_info *sbi,
2478 const char *buf, size_t count)
2479{
2480 unsigned long long val;
2481 int ret;
2482
2483 if (parse_strtoull(buf, -1ULL, &val))
2484 return -EINVAL;
2485 ret = ext4_reserve_clusters(sbi, val);
2486
2487 return ret ? ret : count;
2488}
2489
2c0544b2
TT		 2490static ssize_t trigger_test_error(struct ext4_attr *a,
2491 struct ext4_sb_info *sbi,
2492 const char *buf, size_t count)
2493{
2494 int len = count;
2495
2496 if (!capable(CAP_SYS_ADMIN))
2497 return -EPERM;
2498
2499 if (len && buf[len-1] == '\n')
2500 len--;
2501
2502 if (len)
2503 ext4_error(sbi->s_sb, "%.*s", len, buf);
2504 return count;
2505}
2506
3197ebdb
TT		 2507#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2508static struct ext4_attr ext4_attr_##_name = { \
2509 .attr = {.name = __stringify(_name), .mode = _mode }, \
2510 .show = _show, \
2511 .store = _store, \
2512 .offset = offsetof(struct ext4_sb_info, _elname), \
2513}
2514#define EXT4_ATTR(name, mode, show, store) \
2515static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2516
857ac889 2517#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2518#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2519#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2520#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2521 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2522#define ATTR_LIST(name) &ext4_attr_##name.attr
2523
2524EXT4_RO_ATTR(delayed_allocation_blocks);
2525EXT4_RO_ATTR(session_write_kbytes);
2526EXT4_RO_ATTR(lifetime_write_kbytes);
27dd4385 2527EXT4_RW_ATTR(reserved_clusters);
3197ebdb
TT		 2528EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2529 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2530EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2531EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2532EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2533EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2534EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2535EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2536EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2537EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
67a5da56 2538EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2539EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2540
2541static struct attribute *ext4_attrs[] = {
2542 ATTR_LIST(delayed_allocation_blocks),
2543 ATTR_LIST(session_write_kbytes),
2544 ATTR_LIST(lifetime_write_kbytes),
27dd4385 2545 ATTR_LIST(reserved_clusters),
3197ebdb 2546 ATTR_LIST(inode_readahead_blks),
11013911 2547 ATTR_LIST(inode_goal),
3197ebdb
TT		 2548 ATTR_LIST(mb_stats),
2549 ATTR_LIST(mb_max_to_scan),
2550 ATTR_LIST(mb_min_to_scan),
2551 ATTR_LIST(mb_order2_req),
2552 ATTR_LIST(mb_stream_req),
2553 ATTR_LIST(mb_group_prealloc),
55138e0b 2554 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2555 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2556 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2557 NULL,
2558};
2559
857ac889
LC		 2560/* Features this copy of ext4 supports */
2561EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2562EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2563EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2564
2565static struct attribute *ext4_feat_attrs[] = {
2566 ATTR_LIST(lazy_itable_init),
27ee40df 2567 ATTR_LIST(batched_discard),
5e7bbef1 2568 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2569 NULL,
2570};
2571
3197ebdb
TT		 2572static ssize_t ext4_attr_show(struct kobject *kobj,
2573 struct attribute *attr, char *buf)
2574{
2575 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2576 s_kobj);
2577 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2578
2579 return a->show ? a->show(a, sbi, buf) : 0;
2580}
2581
2582static ssize_t ext4_attr_store(struct kobject *kobj,
2583 struct attribute *attr,
2584 const char *buf, size_t len)
2585{
2586 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2587 s_kobj);
2588 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2589
2590 return a->store ? a->store(a, sbi, buf, len) : 0;
2591}
2592
2593static void ext4_sb_release(struct kobject *kobj)
2594{
2595 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2596 s_kobj);
2597 complete(&sbi->s_kobj_unregister);
2598}
2599
52cf25d0 2600static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2601 .show = ext4_attr_show,
2602 .store = ext4_attr_store,
2603};
2604
2605static struct kobj_type ext4_ktype = {
2606 .default_attrs = ext4_attrs,
2607 .sysfs_ops = &ext4_attr_ops,
2608 .release = ext4_sb_release,
2609};
2610
857ac889
LC		 2611static void ext4_feat_release(struct kobject *kobj)
2612{
2613 complete(&ext4_feat->f_kobj_unregister);
2614}
2615
2616static struct kobj_type ext4_feat_ktype = {
2617 .default_attrs = ext4_feat_attrs,
2618 .sysfs_ops = &ext4_attr_ops,
2619 .release = ext4_feat_release,
2620};
2621
a13fb1a4
ES		 2622/*
2623 * Check whether this filesystem can be mounted based on
2624 * the features present and the RDONLY/RDWR mount requested.
2625 * Returns 1 if this filesystem can be mounted as requested,
2626 * 0 if it cannot be.
2627 */
2628static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2629{
2630 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2631 ext4_msg(sb, KERN_ERR,
2632 "Couldn't mount because of "
2633 "unsupported optional features (%x)",
2634 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2635 ~EXT4_FEATURE_INCOMPAT_SUPP));
2636 return 0;
2637 }
2638
2639 if (readonly)
2640 return 1;
2641
2642 /* Check that feature set is OK for a read-write mount */
2643 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2644 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2645 "unsupported optional features (%x)",
2646 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2647 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2648 return 0;
2649 }
2650 /*
2651 * Large file size enabled file system can only be mounted
2652 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2653 */
2654 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2655 if (sizeof(blkcnt_t) < sizeof(u64)) {
2656 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2657 "cannot be mounted RDWR without "
2658 "CONFIG_LBDAF");
2659 return 0;
2660 }
2661 }
bab08ab9
TT		 2662 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2663 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2664 ext4_msg(sb, KERN_ERR,
2665 "Can't support bigalloc feature without "
2666 "extents feature\n");
2667 return 0;
2668 }
7c319d32
AK		 2669
2670#ifndef CONFIG_QUOTA
2671 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2672 !readonly) {
2673 ext4_msg(sb, KERN_ERR,
2674 "Filesystem with quota feature cannot be mounted RDWR "
2675 "without CONFIG_QUOTA");
2676 return 0;
2677 }
2678#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2679 return 1;
2680}
2681
66e61a9e
TT		 2682/*
2683 * This function is called once a day if we have errors logged
2684 * on the file system
2685 */
2686static void print_daily_error_info(unsigned long arg)
2687{
2688 struct super_block *sb = (struct super_block *) arg;
2689 struct ext4_sb_info *sbi;
2690 struct ext4_super_block *es;
2691
2692 sbi = EXT4_SB(sb);
2693 es = sbi->s_es;
2694
2695 if (es->s_error_count)
2696 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2697 le32_to_cpu(es->s_error_count));
2698 if (es->s_first_error_time) {
2699 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2700 sb->s_id, le32_to_cpu(es->s_first_error_time),
2701 (int) sizeof(es->s_first_error_func),
2702 es->s_first_error_func,
2703 le32_to_cpu(es->s_first_error_line));
2704 if (es->s_first_error_ino)
2705 printk(": inode %u",
2706 le32_to_cpu(es->s_first_error_ino));
2707 if (es->s_first_error_block)
2708 printk(": block %llu", (unsigned long long)
2709 le64_to_cpu(es->s_first_error_block));
2710 printk("\n");
2711 }
2712 if (es->s_last_error_time) {
2713 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2714 sb->s_id, le32_to_cpu(es->s_last_error_time),
2715 (int) sizeof(es->s_last_error_func),
2716 es->s_last_error_func,
2717 le32_to_cpu(es->s_last_error_line));
2718 if (es->s_last_error_ino)
2719 printk(": inode %u",
2720 le32_to_cpu(es->s_last_error_ino));
2721 if (es->s_last_error_block)
2722 printk(": block %llu", (unsigned long long)
2723 le64_to_cpu(es->s_last_error_block));
2724 printk("\n");
2725 }
2726 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2727}
2728
bfff6873
LC		 2729/* Find next suitable group and run ext4_init_inode_table */
2730static int ext4_run_li_request(struct ext4_li_request *elr)
2731{
2732 struct ext4_group_desc *gdp = NULL;
2733 ext4_group_t group, ngroups;
2734 struct super_block *sb;
2735 unsigned long timeout = 0;
2736 int ret = 0;
2737
2738 sb = elr->lr_super;
2739 ngroups = EXT4_SB(sb)->s_groups_count;
2740
8e8ad8a5 2741 sb_start_write(sb);
bfff6873
LC		 2742 for (group = elr->lr_next_group; group < ngroups; group++) {
2743 gdp = ext4_get_group_desc(sb, group, NULL);
2744 if (!gdp) {
2745 ret = 1;
2746 break;
2747 }
2748
2749 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2750 break;
2751 }
2752
7f511862 2753 if (group >= ngroups)
bfff6873
LC		 2754 ret = 1;
2755
2756 if (!ret) {
2757 timeout = jiffies;
2758 ret = ext4_init_inode_table(sb, group,
2759 elr->lr_timeout ? 0 : 1);
2760 if (elr->lr_timeout == 0) {
51ce6511
LC		 2761 timeout = (jiffies - timeout) *
2762 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2763 elr->lr_timeout = timeout;
2764 }
2765 elr->lr_next_sched = jiffies + elr->lr_timeout;
2766 elr->lr_next_group = group + 1;
2767 }
8e8ad8a5 2768 sb_end_write(sb);
bfff6873
LC		 2769
2770 return ret;
2771}
2772
2773/*
2774 * Remove lr_request from the list_request and free the
4ed5c033 2775 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2776 */
2777static void ext4_remove_li_request(struct ext4_li_request *elr)
2778{
2779 struct ext4_sb_info *sbi;
2780
2781 if (!elr)
2782 return;
2783
2784 sbi = elr->lr_sbi;
2785
2786 list_del(&elr->lr_request);
2787 sbi->s_li_request = NULL;
2788 kfree(elr);
2789}
2790
2791static void ext4_unregister_li_request(struct super_block *sb)
2792{
1bb933fb
LC		 2793 mutex_lock(&ext4_li_mtx);
2794 if (!ext4_li_info) {
2795 mutex_unlock(&ext4_li_mtx);
bfff6873 2796 return;
1bb933fb 2797 }
bfff6873
LC		 2798
2799 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2800 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2801 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2802 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2803}
2804
8f1f7453
ES		 2805static struct task_struct *ext4_lazyinit_task;
2806
bfff6873
LC		 2807/*
2808 * This is the function where ext4lazyinit thread lives. It walks
2809 * through the request list searching for next scheduled filesystem.
2810 * When such a fs is found, run the lazy initialization request
2811 * (ext4_rn_li_request) and keep track of the time spend in this
2812 * function. Based on that time we compute next schedule time of
2813 * the request. When walking through the list is complete, compute
2814 * next waking time and put itself into sleep.
2815 */
2816static int ext4_lazyinit_thread(void *arg)
2817{
2818 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2819 struct list_head *pos, *n;
2820 struct ext4_li_request *elr;
4ed5c033 2821 unsigned long next_wakeup, cur;
bfff6873
LC		 2822
2823 BUG_ON(NULL == eli);
2824
bfff6873
LC		 2825cont_thread:
2826 while (true) {
2827 next_wakeup = MAX_JIFFY_OFFSET;
2828
2829 mutex_lock(&eli->li_list_mtx);
2830 if (list_empty(&eli->li_request_list)) {
2831 mutex_unlock(&eli->li_list_mtx);
2832 goto exit_thread;
2833 }
2834
2835 list_for_each_safe(pos, n, &eli->li_request_list) {
2836 elr = list_entry(pos, struct ext4_li_request,
2837 lr_request);
2838
b2c78cd0
TT		 2839 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2840 if (ext4_run_li_request(elr) != 0) {
2841 /* error, remove the lazy_init job */
2842 ext4_remove_li_request(elr);
2843 continue;
2844 }
bfff6873
LC		 2845 }
2846
2847 if (time_before(elr->lr_next_sched, next_wakeup))
2848 next_wakeup = elr->lr_next_sched;
2849 }
2850 mutex_unlock(&eli->li_list_mtx);
2851
a0acae0e 2852 try_to_freeze();
bfff6873 2853
4ed5c033
LC		 2854 cur = jiffies;
2855 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2856 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2857 cond_resched();
2858 continue;
2859 }
2860
4ed5c033
LC		 2861 schedule_timeout_interruptible(next_wakeup - cur);
2862
8f1f7453
ES		 2863 if (kthread_should_stop()) {
2864 ext4_clear_request_list();
2865 goto exit_thread;
2866 }
bfff6873
LC		 2867 }
2868
2869exit_thread:
2870 /*
2871 * It looks like the request list is empty, but we need
2872 * to check it under the li_list_mtx lock, to prevent any
2873 * additions into it, and of course we should lock ext4_li_mtx
2874 * to atomically free the list and ext4_li_info, because at
2875 * this point another ext4 filesystem could be registering
2876 * new one.
2877 */
2878 mutex_lock(&ext4_li_mtx);
2879 mutex_lock(&eli->li_list_mtx);
2880 if (!list_empty(&eli->li_request_list)) {
2881 mutex_unlock(&eli->li_list_mtx);
2882 mutex_unlock(&ext4_li_mtx);
2883 goto cont_thread;
2884 }
2885 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2886 kfree(ext4_li_info);
2887 ext4_li_info = NULL;
2888 mutex_unlock(&ext4_li_mtx);
2889
2890 return 0;
2891}
2892
2893static void ext4_clear_request_list(void)
2894{
2895 struct list_head *pos, *n;
2896 struct ext4_li_request *elr;
2897
2898 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2899 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2900 elr = list_entry(pos, struct ext4_li_request,
2901 lr_request);
2902 ext4_remove_li_request(elr);
2903 }
2904 mutex_unlock(&ext4_li_info->li_list_mtx);
2905}
2906
2907static int ext4_run_lazyinit_thread(void)
2908{
8f1f7453
ES		 2909 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2910 ext4_li_info, "ext4lazyinit");
2911 if (IS_ERR(ext4_lazyinit_task)) {
2912 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2913 ext4_clear_request_list();
bfff6873
LC		 2914 kfree(ext4_li_info);
2915 ext4_li_info = NULL;
92b97816 2916 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2917 "initialization thread\n",
2918 err);
2919 return err;
2920 }
2921 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2922 return 0;
2923}
2924
2925/*
2926 * Check whether it make sense to run itable init. thread or not.
2927 * If there is at least one uninitialized inode table, return
2928 * corresponding group number, else the loop goes through all
2929 * groups and return total number of groups.
2930 */
2931static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2932{
2933 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2934 struct ext4_group_desc *gdp = NULL;
2935
2936 for (group = 0; group < ngroups; group++) {
2937 gdp = ext4_get_group_desc(sb, group, NULL);
2938 if (!gdp)
2939 continue;
2940
2941 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2942 break;
2943 }
2944
2945 return group;
2946}
2947
2948static int ext4_li_info_new(void)
2949{
2950 struct ext4_lazy_init *eli = NULL;
2951
2952 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2953 if (!eli)
2954 return -ENOMEM;
2955
bfff6873
LC		 2956 INIT_LIST_HEAD(&eli->li_request_list);
2957 mutex_init(&eli->li_list_mtx);
2958
bfff6873
LC		 2959 eli->li_state |= EXT4_LAZYINIT_QUIT;
2960
2961 ext4_li_info = eli;
2962
2963 return 0;
2964}
2965
2966static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2967 ext4_group_t start)
2968{
2969 struct ext4_sb_info *sbi = EXT4_SB(sb);
2970 struct ext4_li_request *elr;
2971 unsigned long rnd;
2972
2973 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2974 if (!elr)
2975 return NULL;
2976
2977 elr->lr_super = sb;
2978 elr->lr_sbi = sbi;
2979 elr->lr_next_group = start;
2980
2981 /*
2982 * Randomize first schedule time of the request to
2983 * spread the inode table initialization requests
2984 * better.
2985 */
2986 get_random_bytes(&rnd, sizeof(rnd));
2987 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2988 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2989
2990 return elr;
2991}
2992
7f511862
TT		 2993int ext4_register_li_request(struct super_block *sb,
2994 ext4_group_t first_not_zeroed)
bfff6873
LC		 2995{
2996 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 2997 struct ext4_li_request *elr = NULL;
bfff6873 2998 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2999 int ret = 0;
bfff6873 3000
7f511862 3001 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3002 if (sbi->s_li_request != NULL) {
3003 /*
3004 * Reset timeout so it can be computed again, because
3005 * s_li_wait_mult might have changed.
3006 */
3007 sbi->s_li_request->lr_timeout = 0;
7f511862 3008 goto out;
51ce6511 3009 }
bfff6873
LC		 3010
3011 if (first_not_zeroed == ngroups ||
3012 (sb->s_flags & MS_RDONLY) ||
55ff3840 3013 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3014 goto out;
bfff6873
LC		 3015
3016 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3017 if (!elr) {
3018 ret = -ENOMEM;
3019 goto out;
3020 }
bfff6873
LC		 3021
3022 if (NULL == ext4_li_info) {
3023 ret = ext4_li_info_new();
3024 if (ret)
3025 goto out;
3026 }
3027
3028 mutex_lock(&ext4_li_info->li_list_mtx);
3029 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3030 mutex_unlock(&ext4_li_info->li_list_mtx);
3031
3032 sbi->s_li_request = elr;
46e4690b
TM		 3033 /*
3034 * set elr to NULL here since it has been inserted to
3035 * the request_list and the removal and free of it is
3036 * handled by ext4_clear_request_list from now on.
3037 */
3038 elr = NULL;
bfff6873
LC		 3039
3040 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3041 ret = ext4_run_lazyinit_thread();
3042 if (ret)
3043 goto out;
3044 }
bfff6873 3045out:
beed5ecb
NK		 3046 mutex_unlock(&ext4_li_mtx);
3047 if (ret)
bfff6873 3048 kfree(elr);
bfff6873
LC		 3049 return ret;
3050}
3051
3052/*
3053 * We do not need to lock anything since this is called on
3054 * module unload.
3055 */
3056static void ext4_destroy_lazyinit_thread(void)
3057{
3058 /*
3059 * If thread exited earlier
3060 * there's nothing to be done.
3061 */
8f1f7453 3062 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3063 return;
3064
8f1f7453 3065 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3066}
3067
25ed6e8a
DW		 3068static int set_journal_csum_feature_set(struct super_block *sb)
3069{
3070 int ret = 1;
3071 int compat, incompat;
3072 struct ext4_sb_info *sbi = EXT4_SB(sb);
3073
3074 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3075 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3076 /* journal checksum v2 */
3077 compat = 0;
3078 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3079 } else {
3080 /* journal checksum v1 */
3081 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3082 incompat = 0;
3083 }
3084
3085 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3086 ret = jbd2_journal_set_features(sbi->s_journal,
3087 compat, 0,
3088 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3089 incompat);
3090 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3091 ret = jbd2_journal_set_features(sbi->s_journal,
3092 compat, 0,
3093 incompat);
3094 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3095 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3096 } else {
3097 jbd2_journal_clear_features(sbi->s_journal,
3098 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3099 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3100 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3101 }
3102
3103 return ret;
3104}
3105
952fc18e
TT		 3106/*
3107 * Note: calculating the overhead so we can be compatible with
3108 * historical BSD practice is quite difficult in the face of
3109 * clusters/bigalloc. This is because multiple metadata blocks from
3110 * different block group can end up in the same allocation cluster.
3111 * Calculating the exact overhead in the face of clustered allocation
3112 * requires either O(all block bitmaps) in memory or O(number of block
3113 * groups**2) in time. We will still calculate the superblock for
3114 * older file systems --- and if we come across with a bigalloc file
3115 * system with zero in s_overhead_clusters the estimate will be close to
3116 * correct especially for very large cluster sizes --- but for newer
3117 * file systems, it's better to calculate this figure once at mkfs
3118 * time, and store it in the superblock. If the superblock value is
3119 * present (even for non-bigalloc file systems), we will use it.
3120 */
3121static int count_overhead(struct super_block *sb, ext4_group_t grp,
3122 char *buf)
3123{
3124 struct ext4_sb_info *sbi = EXT4_SB(sb);
3125 struct ext4_group_desc *gdp;
3126 ext4_fsblk_t first_block, last_block, b;
3127 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3128 int s, j, count = 0;
3129
0548bbb8
TT		 3130 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3131 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3132 sbi->s_itb_per_group + 2);
3133
952fc18e
TT		 3134 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3135 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3136 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3137 for (i = 0; i < ngroups; i++) {
3138 gdp = ext4_get_group_desc(sb, i, NULL);
3139 b = ext4_block_bitmap(sb, gdp);
3140 if (b >= first_block && b <= last_block) {
3141 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3142 count++;
3143 }
3144 b = ext4_inode_bitmap(sb, gdp);
3145 if (b >= first_block && b <= last_block) {
3146 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3147 count++;
3148 }
3149 b = ext4_inode_table(sb, gdp);
3150 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3151 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3152 int c = EXT4_B2C(sbi, b - first_block);
3153 ext4_set_bit(c, buf);
3154 count++;
3155 }
3156 if (i != grp)
3157 continue;
3158 s = 0;
3159 if (ext4_bg_has_super(sb, grp)) {
3160 ext4_set_bit(s++, buf);
3161 count++;
3162 }
3163 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3164 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3165 count++;
3166 }
3167 }
3168 if (!count)
3169 return 0;
3170 return EXT4_CLUSTERS_PER_GROUP(sb) -
3171 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3172}
3173
3174/*
3175 * Compute the overhead and stash it in sbi->s_overhead
3176 */
3177int ext4_calculate_overhead(struct super_block *sb)
3178{
3179 struct ext4_sb_info *sbi = EXT4_SB(sb);
3180 struct ext4_super_block *es = sbi->s_es;
3181 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3182 ext4_fsblk_t overhead = 0;
3183 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3184
952fc18e
TT		 3185 if (!buf)
3186 return -ENOMEM;
3187
3188 /*
3189 * Compute the overhead (FS structures). This is constant
3190 * for a given filesystem unless the number of block groups
3191 * changes so we cache the previous value until it does.
3192 */
3193
3194 /*
3195 * All of the blocks before first_data_block are overhead
3196 */
3197 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3198
3199 /*
3200 * Add the overhead found in each block group
3201 */
3202 for (i = 0; i < ngroups; i++) {
3203 int blks;
3204
3205 blks = count_overhead(sb, i, buf);
3206 overhead += blks;
3207 if (blks)
3208 memset(buf, 0, PAGE_SIZE);
3209 cond_resched();
3210 }
0875a2b4
ES		 3211 /* Add the journal blocks as well */
3212 if (sbi->s_journal)
810da240 3213 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3214
952fc18e
TT		 3215 sbi->s_overhead = overhead;
3216 smp_wmb();
3217 free_page((unsigned long) buf);
3218 return 0;
3219}
3220
27dd4385
LC		 3221
3222static ext4_fsblk_t ext4_calculate_resv_clusters(struct ext4_sb_info *sbi)
3223{
3224 ext4_fsblk_t resv_clusters;
3225
3226 /*
3227 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3228 * This should cover the situations where we can not afford to run
3229 * out of space like for example punch hole, or converting
3230 * uninitialized extents in delalloc path. In most cases such
3231 * allocation would require 1, or 2 blocks, higher numbers are
3232 * very rare.
3233 */
3234 resv_clusters = ext4_blocks_count(sbi->s_es) >> sbi->s_cluster_bits;
3235
3236 do_div(resv_clusters, 50);
3237 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3238
3239 return resv_clusters;
3240}
3241
3242
3243static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3244{
3245 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3246 sbi->s_cluster_bits;
3247
3248 if (count >= clusters)
3249 return -EINVAL;
3250
3251 atomic64_set(&sbi->s_resv_clusters, count);
3252 return 0;
3253}
3254
2b2d6d01 3255static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3256{
d4c402d9 3257 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3258 struct buffer_head *bh;
617ba13b
MC		 3259 struct ext4_super_block *es = NULL;
3260 struct ext4_sb_info *sbi;
3261 ext4_fsblk_t block;
3262 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3263 ext4_fsblk_t logical_sb_block;
ac27a0ec 3264 unsigned long offset = 0;
ac27a0ec
DK		 3265 unsigned long journal_devnum = 0;
3266 unsigned long def_mount_opts;
3267 struct inode *root;
9f6200bb 3268 char *cp;
0390131b 3269 const char *descr;
dcc7dae3 3270 int ret = -ENOMEM;
281b5995 3271 int blocksize, clustersize;
4ec11028
TT		 3272 unsigned int db_count;
3273 unsigned int i;
281b5995 3274 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3275 __u64 blocks_count;
07aa2ea1 3276 int err = 0;
b3881f74 3277 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3278 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3279
3280 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3281 if (!sbi)
dcc7dae3 3282 goto out_free_orig;
705895b6
PE		 3283
3284 sbi->s_blockgroup_lock =
3285 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3286 if (!sbi->s_blockgroup_lock) {
3287 kfree(sbi);
dcc7dae3 3288 goto out_free_orig;
705895b6 3289 }
ac27a0ec 3290 sb->s_fs_info = sbi;
2c0544b2 3291 sbi->s_sb = sb;
240799cd 3292 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3293 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3294 if (sb->s_bdev->bd_part)
3295 sbi->s_sectors_written_start =
3296 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3297
9f6200bb
TT		 3298 /* Cleanup superblock name */
3299 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3300 *cp = '!';
3301
07aa2ea1 3302 /* -EINVAL is default */
dcc7dae3 3303 ret = -EINVAL;
617ba13b 3304 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3305 if (!blocksize) {
b31e1552 3306 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3307 goto out_fail;
3308 }
3309
3310 /*
617ba13b 3311 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3312 * block sizes. We need to calculate the offset from buffer start.
3313 */
617ba13b 3314 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3315 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3316 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3317 } else {
70bbb3e0 3318 logical_sb_block = sb_block;
ac27a0ec
DK		 3319 }
3320
70bbb3e0 3321 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3322 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3323 goto out_fail;
3324 }
3325 /*
3326 * Note: s_es must be initialized as soon as possible because
617ba13b 3327 * some ext4 macro-instructions depend on its value
ac27a0ec 3328 */
2716b802 3329 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3330 sbi->s_es = es;
3331 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3332 if (sb->s_magic != EXT4_SUPER_MAGIC)
3333 goto cantfind_ext4;
afc32f7e 3334 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3335
feb0ab32
DW		 3336 /* Warn if metadata_csum and gdt_csum are both set. */
3337 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3338 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3339 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3340 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3341 "redundant flags; please run fsck.");
3342
d25425f8
DW		 3343 /* Check for a known checksum algorithm */
3344 if (!ext4_verify_csum_type(sb, es)) {
3345 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3346 "unknown checksum algorithm.");
3347 silent = 1;
3348 goto cantfind_ext4;
3349 }
3350
0441984a
DW		 3351 /* Load the checksum driver */
3352 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3353 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3354 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3355 if (IS_ERR(sbi->s_chksum_driver)) {
3356 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3357 ret = PTR_ERR(sbi->s_chksum_driver);
3358 sbi->s_chksum_driver = NULL;
3359 goto failed_mount;
3360 }
3361 }
3362
a9c47317
DW		 3363 /* Check superblock checksum */
3364 if (!ext4_superblock_csum_verify(sb, es)) {
3365 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3366 "invalid superblock checksum. Run e2fsck?");
3367 silent = 1;
3368 goto cantfind_ext4;
3369 }
3370
3371 /* Precompute checksum seed for all metadata */
3372 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3373 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3374 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3375 sizeof(es->s_uuid));
3376
ac27a0ec
DK		 3377 /* Set defaults before we parse the mount options */
3378 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3379 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3380 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3381 set_opt(sb, DEBUG);
87f26807 3382 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3383 set_opt(sb, GRPID);
617ba13b 3384 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3385 set_opt(sb, NO_UID32);
ea663336 3386 /* xattr user namespace & acls are now defaulted on */
ea663336 3387 set_opt(sb, XATTR_USER);
03010a33 3388#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3389 set_opt(sb, POSIX_ACL);
2e7842b8 3390#endif
617ba13b 3391 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3392 set_opt(sb, JOURNAL_DATA);
617ba13b 3393 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3394 set_opt(sb, ORDERED_DATA);
617ba13b 3395 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3396 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3397
3398 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3399 set_opt(sb, ERRORS_PANIC);
bb4f397a 3400 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3401 set_opt(sb, ERRORS_CONT);
bb4f397a 3402 else
fd8c37ec 3403 set_opt(sb, ERRORS_RO);
8b67f04a 3404 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3405 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3406 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3407 set_opt(sb, DISCARD);
ac27a0ec 3408
08cefc7a
EB		 3409 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3410 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3411 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3412 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3413 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3414
8b67f04a 3415 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3416 set_opt(sb, BARRIER);
ac27a0ec 3417
dd919b98
AK		 3418 /*
3419 * enable delayed allocation by default
3420 * Use -o nodelalloc to turn it off
3421 */
bc0b75f7 3422 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3423 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3424 set_opt(sb, DELALLOC);
dd919b98 3425
51ce6511
LC		 3426 /*
3427 * set default s_li_wait_mult for lazyinit, for the case there is
3428 * no mount option specified.
3429 */
3430 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3431
8b67f04a 3432 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3433 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3434 ext4_msg(sb, KERN_WARNING,
3435 "failed to parse options in superblock: %s",
3436 sbi->s_es->s_mount_opts);
3437 }
5a916be1 3438 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3439 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3440 &journal_ioprio, 0))
ac27a0ec
DK		 3441 goto failed_mount;
3442
56889787
TT		 3443 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3444 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3445 "with data=journal disables delayed "
3446 "allocation and O_DIRECT support!\n");
3447 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3448 ext4_msg(sb, KERN_ERR, "can't mount with "
3449 "both data=journal and delalloc");
3450 goto failed_mount;
3451 }
3452 if (test_opt(sb, DIOREAD_NOLOCK)) {
3453 ext4_msg(sb, KERN_ERR, "can't mount with "
3454 "both data=journal and delalloc");
3455 goto failed_mount;
3456 }
3457 if (test_opt(sb, DELALLOC))
3458 clear_opt(sb, DELALLOC);
3459 }
3460
ac27a0ec 3461 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3462 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3463
617ba13b
MC		 3464 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3465 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3466 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3467 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3468 ext4_msg(sb, KERN_WARNING,
3469 "feature flags set on rev 0 fs, "
3470 "running e2fsck is recommended");
469108ff 3471
2035e776
TT		 3472 if (IS_EXT2_SB(sb)) {
3473 if (ext2_feature_set_ok(sb))
3474 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3475 "using the ext4 subsystem");
3476 else {
3477 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3478 "to feature incompatibilities");
3479 goto failed_mount;
3480 }
3481 }
3482
3483 if (IS_EXT3_SB(sb)) {
3484 if (ext3_feature_set_ok(sb))
3485 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3486 "using the ext4 subsystem");
3487 else {
3488 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3489 "to feature incompatibilities");
3490 goto failed_mount;
3491 }
3492 }
3493
ac27a0ec
DK		 3494 /*
3495 * Check feature flags regardless of the revision level, since we
3496 * previously didn't change the revision level when setting the flags,
3497 * so there is a chance incompat flags are set on a rev 0 filesystem.
3498 */
a13fb1a4 3499 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3500 goto failed_mount;
a13fb1a4 3501
261cb20c 3502 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3503 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3504 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3505 ext4_msg(sb, KERN_ERR,
3506 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3507 goto failed_mount;
3508 }
3509
ac27a0ec 3510 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3511 /* Validate the filesystem blocksize */
3512 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3513 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3514 blocksize);
ac27a0ec
DK		 3515 goto failed_mount;
3516 }
3517
2b2d6d01 3518 brelse(bh);
70bbb3e0
AM		 3519 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3520 offset = do_div(logical_sb_block, blocksize);
3521 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3522 if (!bh) {
b31e1552
ES		 3523 ext4_msg(sb, KERN_ERR,
3524 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3525 goto failed_mount;
3526 }
2716b802 3527 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3528 sbi->s_es = es;
617ba13b 3529 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3530 ext4_msg(sb, KERN_ERR,
3531 "Magic mismatch, very weird!");
ac27a0ec
DK		 3532 goto failed_mount;
3533 }
3534 }
3535
a13fb1a4
ES		 3536 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3537 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3538 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3539 has_huge_files);
3540 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3541
617ba13b
MC		 3542 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3543 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3544 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3545 } else {
3546 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3547 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3548 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3549 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3550 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3551 ext4_msg(sb, KERN_ERR,
3552 "unsupported inode size: %d",
2b2d6d01 3553 sbi->s_inode_size);
ac27a0ec
DK		 3554 goto failed_mount;
3555 }
ef7f3835
KS		 3556 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3557 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3558 }
0b8e58a1 3559
0d1ee42f
AR		 3560 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3561 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3562 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3563 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3564 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3565 ext4_msg(sb, KERN_ERR,
3566 "unsupported descriptor size %lu",
0d1ee42f
AR		 3567 sbi->s_desc_size);
3568 goto failed_mount;
3569 }
3570 } else
3571 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3572
ac27a0ec 3573 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3574 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3575 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3576 goto cantfind_ext4;
0b8e58a1 3577
617ba13b 3578 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3579 if (sbi->s_inodes_per_block == 0)
617ba13b 3580 goto cantfind_ext4;
ac27a0ec
DK		 3581 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3582 sbi->s_inodes_per_block;
0d1ee42f 3583 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3584 sbi->s_sbh = bh;
3585 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3586 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3587 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3588
68911009
LC		 3589 /* Do we have standard group size of blocksize * 8 blocks ? */
3590 if (sbi->s_blocks_per_group == blocksize << 3)
3591 set_opt2(sb, STD_GROUP_SIZE);
3592
2b2d6d01 3593 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3594 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3595 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3596 i = le32_to_cpu(es->s_flags);
3597 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3598 sbi->s_hash_unsigned = 3;
3599 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3600#ifdef __CHAR_UNSIGNED__
3601 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3602 sbi->s_hash_unsigned = 3;
3603#else
3604 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3605#endif
f99b2589 3606 }
ac27a0ec 3607
281b5995
TT		 3608 /* Handle clustersize */
3609 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3610 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3611 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3612 if (has_bigalloc) {
3613 if (clustersize < blocksize) {
3614 ext4_msg(sb, KERN_ERR,
3615 "cluster size (%d) smaller than "
3616 "block size (%d)", clustersize, blocksize);
3617 goto failed_mount;
3618 }
3619 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3620 le32_to_cpu(es->s_log_block_size);
3621 sbi->s_clusters_per_group =
3622 le32_to_cpu(es->s_clusters_per_group);
3623 if (sbi->s_clusters_per_group > blocksize * 8) {
3624 ext4_msg(sb, KERN_ERR,
3625 "#clusters per group too big: %lu",
3626 sbi->s_clusters_per_group);
3627 goto failed_mount;
3628 }
3629 if (sbi->s_blocks_per_group !=
3630 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3631 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3632 "clusters per group (%lu) inconsistent",
3633 sbi->s_blocks_per_group,
3634 sbi->s_clusters_per_group);
3635 goto failed_mount;
3636 }
3637 } else {
3638 if (clustersize != blocksize) {
3639 ext4_warning(sb, "fragment/cluster size (%d) != "
3640 "block size (%d)", clustersize,
3641 blocksize);
3642 clustersize = blocksize;
3643 }
3644 if (sbi->s_blocks_per_group > blocksize * 8) {
3645 ext4_msg(sb, KERN_ERR,
3646 "#blocks per group too big: %lu",
3647 sbi->s_blocks_per_group);
3648 goto failed_mount;
3649 }
3650 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3651 sbi->s_cluster_bits = 0;
ac27a0ec 3652 }
281b5995
TT		 3653 sbi->s_cluster_ratio = clustersize / blocksize;
3654
ac27a0ec 3655 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3656 ext4_msg(sb, KERN_ERR,
3657 "#inodes per group too big: %lu",
2b2d6d01 3658 sbi->s_inodes_per_group);
ac27a0ec
DK		 3659 goto failed_mount;
3660 }
3661
bf43d84b
ES		 3662 /*
3663 * Test whether we have more sectors than will fit in sector_t,
3664 * and whether the max offset is addressable by the page cache.
3665 */
5a9ae68a 3666 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3667 ext4_blocks_count(es));
5a9ae68a 3668 if (err) {
b31e1552 3669 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3670 " too large to mount safely on this system");
ac27a0ec 3671 if (sizeof(sector_t) < 8)
90c699a9 3672 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3673 goto failed_mount;
3674 }
3675
617ba13b
MC		 3676 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3677 goto cantfind_ext4;
e7c95593 3678
0f2ddca6
FTN		 3679 /* check blocks count against device size */
3680 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3681 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3682 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3683 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3684 ext4_blocks_count(es), blocks_count);
3685 goto failed_mount;
3686 }
3687
0b8e58a1
AD		 3688 /*
3689 * It makes no sense for the first data block to be beyond the end
3690 * of the filesystem.
3691 */
3692 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3693 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3694 "block %u is beyond end of filesystem (%llu)",
3695 le32_to_cpu(es->s_first_data_block),
3696 ext4_blocks_count(es));
e7c95593
ES		 3697 goto failed_mount;
3698 }
bd81d8ee
LV		 3699 blocks_count = (ext4_blocks_count(es) -
3700 le32_to_cpu(es->s_first_data_block) +
3701 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3702 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3703 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3704 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3705 "(block count %llu, first data block %u, "
b31e1552 3706 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3707 ext4_blocks_count(es),
3708 le32_to_cpu(es->s_first_data_block),
3709 EXT4_BLOCKS_PER_GROUP(sb));
3710 goto failed_mount;
3711 }
bd81d8ee 3712 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3713 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3714 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3715 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3716 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3717 sbi->s_group_desc = ext4_kvmalloc(db_count *
3718 sizeof(struct buffer_head *),
3719 GFP_KERNEL);
ac27a0ec 3720 if (sbi->s_group_desc == NULL) {
b31e1552 3721 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3722 ret = -ENOMEM;
ac27a0ec
DK		 3723 goto failed_mount;
3724 }
3725
9f6200bb
TT		 3726 if (ext4_proc_root)
3727 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3728
66acdcf4
TT		 3729 if (sbi->s_proc)
3730 proc_create_data("options", S_IRUGO, sbi->s_proc,
3731 &ext4_seq_options_fops, sb);
3732
705895b6 3733 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3734
3735 for (i = 0; i < db_count; i++) {
70bbb3e0 3736 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3737 sbi->s_group_desc[i] = sb_bread(sb, block);
3738 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3739 ext4_msg(sb, KERN_ERR,
3740 "can't read group descriptor %d", i);
ac27a0ec
DK		 3741 db_count = i;
3742 goto failed_mount2;
3743 }
3744 }
bfff6873 3745 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3746 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3747 goto failed_mount2;
3748 }
772cb7c8
JS		 3749 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3750 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3751 ext4_msg(sb, KERN_ERR,
3752 "unable to initialize "
3753 "flex_bg meta info!");
772cb7c8
JS		 3754 goto failed_mount2;
3755 }
3756
ac27a0ec
DK		 3757 sbi->s_gdb_count = db_count;
3758 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3759 spin_lock_init(&sbi->s_next_gen_lock);
3760
04496411
TM		 3761 init_timer(&sbi->s_err_report);
3762 sbi->s_err_report.function = print_daily_error_info;
3763 sbi->s_err_report.data = (unsigned long) sb;
3764
a75ae78f
DM		 3765 /* Register extent status tree shrinker */
3766 ext4_es_register_shrinker(sb);
3767
57042651 3768 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3769 ext4_count_free_clusters(sb));
ce7e010a
TT		 3770 if (!err) {
3771 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3772 ext4_count_free_inodes(sb));
3773 }
3774 if (!err) {
3775 err = percpu_counter_init(&sbi->s_dirs_counter,
3776 ext4_count_dirs(sb));
3777 }
3778 if (!err) {
57042651 3779 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a 3780 }
1ac6466f
TT		 3781 if (!err) {
3782 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3783 }
ce7e010a
TT		 3784 if (err) {
3785 ext4_msg(sb, KERN_ERR, "insufficient memory");
3786 goto failed_mount3;
3787 }
3788
c9de560d 3789 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3790 sbi->s_max_writeback_mb_bump = 128;
67a5da56 3791 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3792
ac27a0ec
DK		 3793 /*
3794 * set up enough so that it can read an inode
3795 */
9ca92389
TT		 3796 if (!test_opt(sb, NOLOAD) &&
3797 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3798 sb->s_op = &ext4_sops;
3799 else
3800 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3801 sb->s_export_op = &ext4_export_ops;
3802 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3803#ifdef CONFIG_QUOTA
617ba13b 3804 sb->dq_op = &ext4_quota_operations;
262b4662 3805 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
7c319d32 3806 sb->s_qcop = &ext4_qctl_sysfile_operations;
262b4662
JK		 3807 else
3808 sb->s_qcop = &ext4_qctl_operations;
ac27a0ec 3809#endif
f2fa2ffc
AK		 3810 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3811
ac27a0ec 3812 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3813 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3814
3815 sb->s_root = NULL;
3816
3817 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3818 EXT4_HAS_INCOMPAT_FEATURE(sb,
3819 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3820
c5e06d10
JL		 3821 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3822 !(sb->s_flags & MS_RDONLY))
3823 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3824 goto failed_mount3;
3825
ac27a0ec
DK		 3826 /*
3827 * The first inode we look at is the journal inode. Don't try
3828 * root first: it may be modified in the journal!
3829 */
3830 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3831 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3832 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3833 goto failed_mount3;
0390131b
FM		 3834 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3835 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3836 ext4_msg(sb, KERN_ERR, "required journal recovery "
3837 "suppressed and not mounted read-only");
744692dc 3838 goto failed_mount_wq;
ac27a0ec 3839 } else {
fd8c37ec 3840 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3841 sbi->s_journal = NULL;
3842 needs_recovery = 0;
3843 goto no_journal;
ac27a0ec
DK		 3844 }
3845
f32aaf2d 3846 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3847 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3848 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3849 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3850 goto failed_mount_wq;
eb40a09c
JS		 3851 }
3852
25ed6e8a
DW		 3853 if (!set_journal_csum_feature_set(sb)) {
3854 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3855 "feature set");
3856 goto failed_mount_wq;
d4da6c9c 3857 }
818d276c 3858
ac27a0ec
DK		 3859 /* We have now updated the journal if required, so we can
3860 * validate the data journaling mode. */
3861 switch (test_opt(sb, DATA_FLAGS)) {
3862 case 0:
3863 /* No mode set, assume a default based on the journal
63f57933
AM		 3864 * capabilities: ORDERED_DATA if the journal can
3865 * cope, else JOURNAL_DATA
3866 */
dab291af
MC		 3867 if (jbd2_journal_check_available_features
3868 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3869 set_opt(sb, ORDERED_DATA);
ac27a0ec 3870 else
fd8c37ec 3871 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3872 break;
3873
617ba13b
MC		 3874 case EXT4_MOUNT_ORDERED_DATA:
3875 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3876 if (!jbd2_journal_check_available_features
3877 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3878 ext4_msg(sb, KERN_ERR, "Journal does not support "
3879 "requested data journaling mode");
744692dc 3880 goto failed_mount_wq;
ac27a0ec
DK		 3881 }
3882 default:
3883 break;
3884 }
b3881f74 3885 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3886
18aadd47
BJ		 3887 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3888
ce7e010a
TT		 3889 /*
3890 * The journal may have updated the bg summary counts, so we
3891 * need to update the global counters.
3892 */
57042651 3893 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3894 ext4_count_free_clusters(sb));
ce7e010a
TT		 3895 percpu_counter_set(&sbi->s_freeinodes_counter,
3896 ext4_count_free_inodes(sb));
3897 percpu_counter_set(&sbi->s_dirs_counter,
3898 ext4_count_dirs(sb));
57042651 3899 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3900
ce7e010a 3901no_journal:
952fc18e
TT		 3902 /*
3903 * Get the # of file system overhead blocks from the
3904 * superblock if present.
3905 */
3906 if (es->s_overhead_clusters)
3907 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3908 else {
07aa2ea1
LC		 3909 err = ext4_calculate_overhead(sb);
3910 if (err)
952fc18e
TT		 3911 goto failed_mount_wq;
3912 }
3913
fd89d5f2
TH		 3914 /*
3915 * The maximum number of concurrent works can be high and
3916 * concurrency isn't really necessary. Limit it to 1.
3917 */
3918 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3919 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3920 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3921 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
07aa2ea1 3922 ret = -ENOMEM;
4c0425ff
MC		 3923 goto failed_mount_wq;
3924 }
3925
ac27a0ec 3926 /*
dab291af 3927 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3928 * so we can safely mount the rest of the filesystem now.
3929 */
3930
1d1fe1ee
DH		 3931 root = ext4_iget(sb, EXT4_ROOT_INO);
3932 if (IS_ERR(root)) {
b31e1552 3933 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3934 ret = PTR_ERR(root);
32a9bb57 3935 root = NULL;
ac27a0ec
DK		 3936 goto failed_mount4;
3937 }
3938 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3939 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3940 iput(root);
ac27a0ec
DK		 3941 goto failed_mount4;
3942 }
48fde701 3943 sb->s_root = d_make_root(root);
1d1fe1ee 3944 if (!sb->s_root) {
b31e1552 3945 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3946 ret = -ENOMEM;
3947 goto failed_mount4;
3948 }
ac27a0ec 3949
7e84b621
ES		 3950 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3951 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3952
3953 /* determine the minimum size of new large inodes, if present */
3954 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3955 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3956 EXT4_GOOD_OLD_INODE_SIZE;
3957 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3958 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3959 if (sbi->s_want_extra_isize <
3960 le16_to_cpu(es->s_want_extra_isize))
3961 sbi->s_want_extra_isize =
3962 le16_to_cpu(es->s_want_extra_isize);
3963 if (sbi->s_want_extra_isize <
3964 le16_to_cpu(es->s_min_extra_isize))
3965 sbi->s_want_extra_isize =
3966 le16_to_cpu(es->s_min_extra_isize);
3967 }
3968 }
3969 /* Check if enough inode space is available */
3970 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3971 sbi->s_inode_size) {
3972 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3973 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3974 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3975 "available");
ef7f3835
KS		 3976 }
3977
27dd4385
LC		 3978 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sbi));
3979 if (err) {
3980 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
3981 "reserved pool", ext4_calculate_resv_clusters(sbi));
3982 goto failed_mount4a;
3983 }
3984
6fd058f7
TT		 3985 err = ext4_setup_system_zone(sb);
3986 if (err) {
b31e1552 3987 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3988 "zone (%d)", err);
94bf608a 3989 goto failed_mount4a;
6fd058f7
TT		 3990 }
3991
c2774d84 3992 ext4_ext_init(sb);
9d99012f 3993 err = ext4_mb_init(sb);
c2774d84 3994 if (err) {
421f91d2 3995 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3996 err);
dcf2d804 3997 goto failed_mount5;
c2774d84
AK		 3998 }
3999
bfff6873
LC		 4000 err = ext4_register_li_request(sb, first_not_zeroed);
4001 if (err)
dcf2d804 4002 goto failed_mount6;
bfff6873 4003
3197ebdb
TT		 4004 sbi->s_kobj.kset = ext4_kset;
4005 init_completion(&sbi->s_kobj_unregister);
4006 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4007 "%s", sb->s_id);
dcf2d804
TM		 4008 if (err)
4009 goto failed_mount7;
3197ebdb 4010
9b2ff357
JK		 4011#ifdef CONFIG_QUOTA
4012 /* Enable quota usage during mount. */
4013 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4014 !(sb->s_flags & MS_RDONLY)) {
4015 err = ext4_enable_quotas(sb);
4016 if (err)
4017 goto failed_mount8;
4018 }
4019#endif /* CONFIG_QUOTA */
4020
617ba13b
MC		 4021 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4022 ext4_orphan_cleanup(sb, es);
4023 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4024 if (needs_recovery) {
b31e1552 4025 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4026 ext4_mark_recovery_complete(sb, es);
4027 }
4028 if (EXT4_SB(sb)->s_journal) {
4029 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4030 descr = " journalled data mode";
4031 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4032 descr = " ordered data mode";
4033 else
4034 descr = " writeback data mode";
4035 } else
4036 descr = "out journal";
4037
79add3a3
LC		 4038 if (test_opt(sb, DISCARD)) {
4039 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4040 if (!blk_queue_discard(q))
4041 ext4_msg(sb, KERN_WARNING,
4042 "mounting with \"discard\" option, but "
4043 "the device does not support discard");
4044 }
4045
d4c402d9 4046 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4047 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4048 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4049
66e61a9e
TT		 4050 if (es->s_error_count)
4051 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4052
d4c402d9 4053 kfree(orig_data);
ac27a0ec
DK		 4054 return 0;
4055
617ba13b 4056cantfind_ext4:
ac27a0ec 4057 if (!silent)
b31e1552 4058 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4059 goto failed_mount;
4060
72ba7450
TT		 4061#ifdef CONFIG_QUOTA
4062failed_mount8:
4063 kobject_del(&sbi->s_kobj);
4064#endif
dcf2d804
TM		 4065failed_mount7:
4066 ext4_unregister_li_request(sb);
4067failed_mount6:
dcf2d804 4068 ext4_mb_release(sb);
94bf608a
AV		 4069failed_mount5:
4070 ext4_ext_release(sb);
dcf2d804 4071 ext4_release_system_zone(sb);
94bf608a
AV		 4072failed_mount4a:
4073 dput(sb->s_root);
32a9bb57 4074 sb->s_root = NULL;
94bf608a 4075failed_mount4:
b31e1552 4076 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 4077 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4078failed_mount_wq:
0390131b
FM		 4079 if (sbi->s_journal) {
4080 jbd2_journal_destroy(sbi->s_journal);
4081 sbi->s_journal = NULL;
4082 }
ac27a0ec 4083failed_mount3:
a75ae78f 4084 ext4_es_unregister_shrinker(sb);
04496411 4085 del_timer(&sbi->s_err_report);
9933fc0a
TT		 4086 if (sbi->s_flex_groups)
4087 ext4_kvfree(sbi->s_flex_groups);
57042651 4088 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 4089 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4090 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 4091 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1ac6466f 4092 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
c5e06d10
JL		 4093 if (sbi->s_mmp_tsk)
4094 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4095failed_mount2:
4096 for (i = 0; i < db_count; i++)
4097 brelse(sbi->s_group_desc[i]);
f18a5f21 4098 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 4099failed_mount:
0441984a
DW		 4100 if (sbi->s_chksum_driver)
4101 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4102 if (sbi->s_proc) {
66acdcf4 4103 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4104 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4105 }
ac27a0ec
DK		 4106#ifdef CONFIG_QUOTA
4107 for (i = 0; i < MAXQUOTAS; i++)
4108 kfree(sbi->s_qf_names[i]);
4109#endif
617ba13b 4110 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4111 brelse(bh);
4112out_fail:
4113 sb->s_fs_info = NULL;
f6830165 4114 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4115 kfree(sbi);
dcc7dae3 4116out_free_orig:
d4c402d9 4117 kfree(orig_data);
07aa2ea1 4118 return err ? err : ret;
ac27a0ec
DK		 4119}
4120
4121/*
4122 * Setup any per-fs journal parameters now. We'll do this both on
4123 * initial mount, once the journal has been initialised but before we've
4124 * done any recovery; and again on any subsequent remount.
4125 */
617ba13b 4126static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4127{
617ba13b 4128 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4129
30773840
TT		 4130 journal->j_commit_interval = sbi->s_commit_interval;
4131 journal->j_min_batch_time = sbi->s_min_batch_time;
4132 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4133
a931da6a 4134 write_lock(&journal->j_state_lock);
ac27a0ec 4135 if (test_opt(sb, BARRIER))
dab291af 4136 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4137 else
dab291af 4138 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4139 if (test_opt(sb, DATA_ERR_ABORT))
4140 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4141 else
4142 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4143 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4144}
4145
617ba13b 4146static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4147 unsigned int journal_inum)
4148{
4149 struct inode *journal_inode;
4150 journal_t *journal;
4151
0390131b
FM		 4152 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4153
ac27a0ec
DK		 4154 /* First, test for the existence of a valid inode on disk. Bad
4155 * things happen if we iget() an unused inode, as the subsequent
4156 * iput() will try to delete it. */
4157
1d1fe1ee
DH		 4158 journal_inode = ext4_iget(sb, journal_inum);
4159 if (IS_ERR(journal_inode)) {
b31e1552 4160 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4161 return NULL;
4162 }
4163 if (!journal_inode->i_nlink) {
4164 make_bad_inode(journal_inode);
4165 iput(journal_inode);
b31e1552 4166 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4167 return NULL;
4168 }
4169
e5f8eab8 4170 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4171 journal_inode, journal_inode->i_size);
1d1fe1ee 4172 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4173 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4174 iput(journal_inode);
4175 return NULL;
4176 }
4177
dab291af 4178 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4179 if (!journal) {
b31e1552 4180 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4181 iput(journal_inode);
4182 return NULL;
4183 }
4184 journal->j_private = sb;
617ba13b 4185 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4186 return journal;
4187}
4188
617ba13b 4189static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4190 dev_t j_dev)
4191{
2b2d6d01 4192 struct buffer_head *bh;
ac27a0ec 4193 journal_t *journal;
617ba13b
MC		 4194 ext4_fsblk_t start;
4195 ext4_fsblk_t len;
ac27a0ec 4196 int hblock, blocksize;
617ba13b 4197 ext4_fsblk_t sb_block;
ac27a0ec 4198 unsigned long offset;
2b2d6d01 4199 struct ext4_super_block *es;
ac27a0ec
DK		 4200 struct block_device *bdev;
4201
0390131b
FM		 4202 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4203
b31e1552 4204 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4205 if (bdev == NULL)
4206 return NULL;
4207
ac27a0ec 4208 blocksize = sb->s_blocksize;
e1defc4f 4209 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4210 if (blocksize < hblock) {
b31e1552
ES		 4211 ext4_msg(sb, KERN_ERR,
4212 "blocksize too small for journal device");
ac27a0ec
DK		 4213 goto out_bdev;
4214 }
4215
617ba13b
MC		 4216 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4217 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4218 set_blocksize(bdev, blocksize);
4219 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4220 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4221 "external journal");
ac27a0ec
DK		 4222 goto out_bdev;
4223 }
4224
2716b802 4225 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4226 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4227 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4228 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4229 ext4_msg(sb, KERN_ERR, "external journal has "
4230 "bad superblock");
ac27a0ec
DK		 4231 brelse(bh);
4232 goto out_bdev;
4233 }
4234
617ba13b 4235 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4236 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4237 brelse(bh);
4238 goto out_bdev;
4239 }
4240
bd81d8ee 4241 len = ext4_blocks_count(es);
ac27a0ec
DK		 4242 start = sb_block + 1;
4243 brelse(bh); /* we're done with the superblock */
4244
dab291af 4245 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4246 start, len, blocksize);
4247 if (!journal) {
b31e1552 4248 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4249 goto out_bdev;
4250 }
4251 journal->j_private = sb;
9f203507 4252 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4253 wait_on_buffer(journal->j_sb_buffer);
4254 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4255 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4256 goto out_journal;
4257 }
4258 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4259 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4260 "user (unsupported) - %d",
ac27a0ec
DK		 4261 be32_to_cpu(journal->j_superblock->s_nr_users));
4262 goto out_journal;
4263 }
617ba13b
MC		 4264 EXT4_SB(sb)->journal_bdev = bdev;
4265 ext4_init_journal_params(sb, journal);
ac27a0ec 4266 return journal;
0b8e58a1 4267
ac27a0ec 4268out_journal:
dab291af 4269 jbd2_journal_destroy(journal);
ac27a0ec 4270out_bdev:
617ba13b 4271 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4272 return NULL;
4273}
4274
617ba13b
MC		 4275static int ext4_load_journal(struct super_block *sb,
4276 struct ext4_super_block *es,
ac27a0ec
DK		 4277 unsigned long journal_devnum)
4278{
4279 journal_t *journal;
4280 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4281 dev_t journal_dev;
4282 int err = 0;
4283 int really_read_only;
4284
0390131b
FM		 4285 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4286
ac27a0ec
DK		 4287 if (journal_devnum &&
4288 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4289 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4290 "numbers have changed");
ac27a0ec
DK		 4291 journal_dev = new_decode_dev(journal_devnum);
4292 } else
4293 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4294
4295 really_read_only = bdev_read_only(sb->s_bdev);
4296
4297 /*
4298 * Are we loading a blank journal or performing recovery after a
4299 * crash? For recovery, we need to check in advance whether we
4300 * can get read-write access to the device.
4301 */
617ba13b 4302 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4303 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4304 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4305 "required on readonly filesystem");
ac27a0ec 4306 if (really_read_only) {
b31e1552
ES		 4307 ext4_msg(sb, KERN_ERR, "write access "
4308 "unavailable, cannot proceed");
ac27a0ec
DK		 4309 return -EROFS;
4310 }
b31e1552
ES		 4311 ext4_msg(sb, KERN_INFO, "write access will "
4312 "be enabled during recovery");
ac27a0ec
DK		 4313 }
4314 }
4315
4316 if (journal_inum && journal_dev) {
b31e1552
ES		 4317 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4318 "and inode journals!");
ac27a0ec
DK		 4319 return -EINVAL;
4320 }
4321
4322 if (journal_inum) {
617ba13b 4323 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4324 return -EINVAL;
4325 } else {
617ba13b 4326 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4327 return -EINVAL;
4328 }
4329
90576c0b 4330 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4331 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4332
617ba13b 4333 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4334 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4335 if (!err) {
4336 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4337 if (save)
4338 memcpy(save, ((char *) es) +
4339 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4340 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4341 if (save)
4342 memcpy(((char *) es) + EXT4_S_ERR_START,
4343 save, EXT4_S_ERR_LEN);
4344 kfree(save);
4345 }
ac27a0ec
DK		 4346
4347 if (err) {
b31e1552 4348 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4349 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4350 return err;
4351 }
4352
617ba13b
MC		 4353 EXT4_SB(sb)->s_journal = journal;
4354 ext4_clear_journal_err(sb, es);
ac27a0ec 4355
c41303ce 4356 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4357 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4358 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4359
4360 /* Make sure we flush the recovery flag to disk. */
e2d67052 4361 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4362 }
4363
4364 return 0;
4365}
4366
e2d67052 4367static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4368{
e2d67052 4369 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4370 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4371 int error = 0;
ac27a0ec 4372
7c2e7087 4373 if (!sbh || block_device_ejected(sb))
c4be0c1d 4374 return error;
914258bf
TT		 4375 if (buffer_write_io_error(sbh)) {
4376 /*
4377 * Oh, dear. A previous attempt to write the
4378 * superblock failed. This could happen because the
4379 * USB device was yanked out. Or it could happen to
4380 * be a transient write error and maybe the block will
4381 * be remapped. Nothing we can do but to retry the
4382 * write and hope for the best.
4383 */
b31e1552
ES		 4384 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4385 "superblock detected");
914258bf
TT		 4386 clear_buffer_write_io_error(sbh);
4387 set_buffer_uptodate(sbh);
4388 }
71290b36
TT		 4389 /*
4390 * If the file system is mounted read-only, don't update the
4391 * superblock write time. This avoids updating the superblock
4392 * write time when we are mounting the root file system
4393 * read/only but we need to replay the journal; at that point,
4394 * for people who are east of GMT and who make their clock
4395 * tick in localtime for Windows bug-for-bug compatibility,
4396 * the clock is set in the future, and this will cause e2fsck
4397 * to complain and force a full file system check.
4398 */
4399 if (!(sb->s_flags & MS_RDONLY))
4400 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4401 if (sb->s_bdev->bd_part)
4402 es->s_kbytes_written =
4403 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4404 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4405 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4406 else
4407 es->s_kbytes_written =
4408 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4409 ext4_free_blocks_count_set(es,
4410 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4411 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4412 es->s_free_inodes_count =
4413 cpu_to_le32(percpu_counter_sum_positive(
4414 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4415 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4416 ext4_superblock_csum_set(sb);
ac27a0ec 4417 mark_buffer_dirty(sbh);
914258bf 4418 if (sync) {
c4be0c1d
TS		 4419 error = sync_dirty_buffer(sbh);
4420 if (error)
4421 return error;
4422
4423 error = buffer_write_io_error(sbh);
4424 if (error) {
b31e1552
ES		 4425 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4426 "superblock");
914258bf
TT		 4427 clear_buffer_write_io_error(sbh);
4428 set_buffer_uptodate(sbh);
4429 }
4430 }
c4be0c1d 4431 return error;
ac27a0ec
DK		 4432}
4433
ac27a0ec
DK		 4434/*
4435 * Have we just finished recovery? If so, and if we are mounting (or
4436 * remounting) the filesystem readonly, then we will end up with a
4437 * consistent fs on disk. Record that fact.
4438 */
2b2d6d01
TT		 4439static void ext4_mark_recovery_complete(struct super_block *sb,
4440 struct ext4_super_block *es)
ac27a0ec 4441{
617ba13b 4442 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4443
0390131b
FM		 4444 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4445 BUG_ON(journal != NULL);
4446 return;
4447 }
dab291af 4448 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4449 if (jbd2_journal_flush(journal) < 0)
4450 goto out;
4451
617ba13b 4452 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4453 sb->s_flags & MS_RDONLY) {
617ba13b 4454 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4455 ext4_commit_super(sb, 1);
ac27a0ec 4456 }
7ffe1ea8
HK		 4457
4458out:
dab291af 4459 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4460}
4461
4462/*
4463 * If we are mounting (or read-write remounting) a filesystem whose journal
4464 * has recorded an error from a previous lifetime, move that error to the
4465 * main filesystem now.
4466 */
2b2d6d01
TT		 4467static void ext4_clear_journal_err(struct super_block *sb,
4468 struct ext4_super_block *es)
ac27a0ec
DK		 4469{
4470 journal_t *journal;
4471 int j_errno;
4472 const char *errstr;
4473
0390131b
FM		 4474 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4475
617ba13b 4476 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4477
4478 /*
4479 * Now check for any error status which may have been recorded in the
617ba13b 4480 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4481 */
4482
dab291af 4483 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4484 if (j_errno) {
4485 char nbuf[16];
4486
617ba13b 4487 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4488 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4489 "from previous mount: %s", errstr);
12062ddd 4490 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4491
617ba13b
MC		 4492 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4493 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4494 ext4_commit_super(sb, 1);
ac27a0ec 4495
dab291af 4496 jbd2_journal_clear_err(journal);
d796c52e 4497 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4498 }
4499}
4500
4501/*
4502 * Force the running and committing transactions to commit,
4503 * and wait on the commit.
4504 */
617ba13b 4505int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4506{
4507 journal_t *journal;
ac27a0ec
DK		 4508
4509 if (sb->s_flags & MS_RDONLY)
4510 return 0;
4511
617ba13b 4512 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4513 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4514}
4515
617ba13b 4516static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4517{
14ce0cb4 4518 int ret = 0;
9eddacf9 4519 tid_t target;
8d5d02e6 4520 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4521
9bffad1e 4522 trace_ext4_sync_fs(sb, wait);
8d5d02e6 4523 flush_workqueue(sbi->dio_unwritten_wq);
a1177825
JK		 4524 /*
4525 * Writeback quota in non-journalled quota case - journalled quota has
4526 * no dirty dquots
4527 */
4528 dquot_writeback_dquots(sb, -1);
8d5d02e6 4529 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4530 if (wait)
8d5d02e6 4531 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4532 }
14ce0cb4 4533 return ret;
ac27a0ec
DK		 4534}
4535
4536/*
4537 * LVM calls this function before a (read-only) snapshot is created. This
4538 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4539 *
4540 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4541 * state independently. It relies on upper layer to stop all data & metadata
4542 * modifications.
ac27a0ec 4543 */
c4be0c1d 4544static int ext4_freeze(struct super_block *sb)
ac27a0ec 4545{
c4be0c1d
TS		 4546 int error = 0;
4547 journal_t *journal;
ac27a0ec 4548
9ca92389
TT		 4549 if (sb->s_flags & MS_RDONLY)
4550 return 0;
ac27a0ec 4551
9ca92389 4552 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4553
9ca92389
TT		 4554 /* Now we set up the journal barrier. */
4555 jbd2_journal_lock_updates(journal);
ac27a0ec 4556
9ca92389
TT		 4557 /*
4558 * Don't clear the needs_recovery flag if we failed to flush
4559 * the journal.
4560 */
4561 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4562 if (error < 0)
4563 goto out;
9ca92389
TT		 4564
4565 /* Journal blocked and flushed, clear needs_recovery flag. */
4566 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4567 error = ext4_commit_super(sb, 1);
6b0310fb 4568out:
8e8ad8a5 4569 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4570 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4571 return error;
ac27a0ec
DK		 4572}
4573
4574/*
4575 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4576 * flag here, even though the filesystem is not technically dirty yet.
4577 */
c4be0c1d 4578static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4579{
9ca92389
TT		 4580 if (sb->s_flags & MS_RDONLY)
4581 return 0;
4582
9ca92389
TT		 4583 /* Reset the needs_recovery flag before the fs is unlocked. */
4584 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4585 ext4_commit_super(sb, 1);
c4be0c1d 4586 return 0;
ac27a0ec
DK		 4587}
4588
673c6100
TT		 4589/*
4590 * Structure to save mount options for ext4_remount's benefit
4591 */
4592struct ext4_mount_options {
4593 unsigned long s_mount_opt;
a2595b8a 4594 unsigned long s_mount_opt2;
08cefc7a
EB		 4595 kuid_t s_resuid;
4596 kgid_t s_resgid;
673c6100
TT		 4597 unsigned long s_commit_interval;
4598 u32 s_min_batch_time, s_max_batch_time;
4599#ifdef CONFIG_QUOTA
4600 int s_jquota_fmt;
4601 char *s_qf_names[MAXQUOTAS];
4602#endif
4603};
4604
2b2d6d01 4605static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4606{
2b2d6d01 4607 struct ext4_super_block *es;
617ba13b 4608 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4609 unsigned long old_sb_flags;
617ba13b 4610 struct ext4_mount_options old_opts;
c79d967d 4611 int enable_quota = 0;
8a266467 4612 ext4_group_t g;
b3881f74 4613 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4614 int err = 0;
ac27a0ec 4615#ifdef CONFIG_QUOTA
03dafb5f 4616 int i, j;
ac27a0ec 4617#endif
d4c402d9 4618 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4619
4620 /* Store the original options */
4621 old_sb_flags = sb->s_flags;
4622 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4623 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4624 old_opts.s_resuid = sbi->s_resuid;
4625 old_opts.s_resgid = sbi->s_resgid;
4626 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4627 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4628 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4629#ifdef CONFIG_QUOTA
4630 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4631 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f
CG		 4632 if (sbi->s_qf_names[i]) {
4633 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4634 GFP_KERNEL);
4635 if (!old_opts.s_qf_names[i]) {
4636 for (j = 0; j < i; j++)
4637 kfree(old_opts.s_qf_names[j]);
3e36a163 4638 kfree(orig_data);
03dafb5f
CG		 4639 return -ENOMEM;
4640 }
4641 } else
4642 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4643#endif
b3881f74
TT		 4644 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4645 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4646
4647 /*
4648 * Allow the "check" option to be passed as a remount option.
4649 */
661aa520 4650 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4651 err = -EINVAL;
4652 goto restore_opts;
4653 }
4654
4ab2f15b 4655 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4656 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4657
4658 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4659 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4660
4661 es = sbi->s_es;
4662
b3881f74 4663 if (sbi->s_journal) {
0390131b 4664 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4665 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4666 }
ac27a0ec 4667
661aa520 4668 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4669 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4670 err = -EROFS;
4671 goto restore_opts;
4672 }
4673
4674 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4675 err = dquot_suspend(sb, -1);
4676 if (err < 0)
c79d967d 4677 goto restore_opts;
c79d967d 4678
ac27a0ec
DK		 4679 /*
4680 * First of all, the unconditional stuff we have to do
4681 * to disable replay of the journal when we next remount
4682 */
4683 sb->s_flags |= MS_RDONLY;
4684
4685 /*
4686 * OK, test if we are remounting a valid rw partition
4687 * readonly, and if so set the rdonly flag and then
4688 * mark the partition as valid again.
4689 */
617ba13b
MC		 4690 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4691 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4692 es->s_state = cpu_to_le16(sbi->s_mount_state);
4693
a63c9eb2 4694 if (sbi->s_journal)
0390131b 4695 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4696 } else {
a13fb1a4
ES		 4697 /* Make sure we can mount this feature set readwrite */
4698 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4699 err = -EROFS;
4700 goto restore_opts;
4701 }
8a266467
TT		 4702 /*
4703 * Make sure the group descriptor checksums
0b8e58a1 4704 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4705 */
4706 for (g = 0; g < sbi->s_groups_count; g++) {
4707 struct ext4_group_desc *gdp =
4708 ext4_get_group_desc(sb, g, NULL);
4709
feb0ab32 4710 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4711 ext4_msg(sb, KERN_ERR,
4712 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4713 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4714 le16_to_cpu(gdp->bg_checksum));
4715 err = -EINVAL;
4716 goto restore_opts;
4717 }
4718 }
4719
ead6596b
ES		 4720 /*
4721 * If we have an unprocessed orphan list hanging
4722 * around from a previously readonly bdev mount,
4723 * require a full umount/remount for now.
4724 */
4725 if (es->s_last_orphan) {
b31e1552 4726 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4727 "remount RDWR because of unprocessed "
4728 "orphan inode list. Please "
b31e1552 4729 "umount/remount instead");
ead6596b
ES		 4730 err = -EINVAL;
4731 goto restore_opts;
4732 }
4733
ac27a0ec
DK		 4734 /*
4735 * Mounting a RDONLY partition read-write, so reread
4736 * and store the current valid flag. (It may have
4737 * been changed by e2fsck since we originally mounted
4738 * the partition.)
4739 */
0390131b
FM		 4740 if (sbi->s_journal)
4741 ext4_clear_journal_err(sb, es);
ac27a0ec 4742 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4743 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4744 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4745 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4746 EXT4_FEATURE_INCOMPAT_MMP))
4747 if (ext4_multi_mount_protect(sb,
4748 le64_to_cpu(es->s_mmp_block))) {
4749 err = -EROFS;
4750 goto restore_opts;
4751 }
c79d967d 4752 enable_quota = 1;
ac27a0ec
DK		 4753 }
4754 }
bfff6873
LC		 4755
4756 /*
4757 * Reinitialize lazy itable initialization thread based on
4758 * current settings
4759 */
4760 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4761 ext4_unregister_li_request(sb);
4762 else {
4763 ext4_group_t first_not_zeroed;
4764 first_not_zeroed = ext4_has_uninit_itable(sb);
4765 ext4_register_li_request(sb, first_not_zeroed);
4766 }
4767
6fd058f7 4768 ext4_setup_system_zone(sb);
d096ad0f 4769 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4770 ext4_commit_super(sb, 1);
0390131b 4771
ac27a0ec
DK		 4772#ifdef CONFIG_QUOTA
4773 /* Release old quota file names */
4774 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f 4775 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4776 if (enable_quota) {
4777 if (sb_any_quota_suspended(sb))
4778 dquot_resume(sb, -1);
4779 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4780 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4781 err = ext4_enable_quotas(sb);
07724f98 4782 if (err)
7c319d32 4783 goto restore_opts;
7c319d32
AK		 4784 }
4785 }
ac27a0ec 4786#endif
d4c402d9
CW		 4787
4788 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4789 kfree(orig_data);
ac27a0ec 4790 return 0;
0b8e58a1 4791
ac27a0ec
DK		 4792restore_opts:
4793 sb->s_flags = old_sb_flags;
4794 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4795 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4796 sbi->s_resuid = old_opts.s_resuid;
4797 sbi->s_resgid = old_opts.s_resgid;
4798 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4799 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4800 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4801#ifdef CONFIG_QUOTA
4802 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4803 for (i = 0; i < MAXQUOTAS; i++) {
03dafb5f 4804 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 4805 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4806 }
4807#endif
d4c402d9 4808 kfree(orig_data);
ac27a0ec
DK		 4809 return err;
4810}
4811
2b2d6d01 4812static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4813{
4814 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4815 struct ext4_sb_info *sbi = EXT4_SB(sb);
4816 struct ext4_super_block *es = sbi->s_es;
27dd4385 4817 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 4818 u64 fsid;
d02a9391 4819 s64 bfree;
27dd4385 4820 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 4821
952fc18e
TT		 4822 if (!test_opt(sb, MINIX_DF))
4823 overhead = sbi->s_overhead;
ac27a0ec 4824
617ba13b 4825 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4826 buf->f_bsize = sb->s_blocksize;
b72f78cb 4827 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4828 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4829 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4830 /* prevent underflow in case that few free space is available */
57042651 4831 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 4832 buf->f_bavail = buf->f_bfree -
4833 (ext4_r_blocks_count(es) + resv_blocks);
4834 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 4835 buf->f_bavail = 0;
4836 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4837 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4838 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4839 fsid = le64_to_cpup((void *)es->s_uuid) ^
4840 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4841 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4842 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4843
ac27a0ec
DK		 4844 return 0;
4845}
4846
0b8e58a1
AD		 4847/* Helper function for writing quotas on sync - we need to start transaction
4848 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4849 * Process 1 Process 2
617ba13b 4850 * ext4_create() quota_sync()
a269eb18 4851 * jbd2_journal_start() write_dquot()
871a2931 4852 * dquot_initialize() down(dqio_mutex)
dab291af 4853 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4854 *
4855 */
4856
4857#ifdef CONFIG_QUOTA
4858
4859static inline struct inode *dquot_to_inode(struct dquot *dquot)
4860{
4c376dca 4861 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4862}
4863
617ba13b 4864static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4865{
4866 int ret, err;
4867 handle_t *handle;
4868 struct inode *inode;
4869
4870 inode = dquot_to_inode(dquot);
9924a92a 4871 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 4872 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4873 if (IS_ERR(handle))
4874 return PTR_ERR(handle);
4875 ret = dquot_commit(dquot);
617ba13b 4876 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4877 if (!ret)
4878 ret = err;
4879 return ret;
4880}
4881
617ba13b 4882static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4883{
4884 int ret, err;
4885 handle_t *handle;
4886
9924a92a 4887 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4888 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4889 if (IS_ERR(handle))
4890 return PTR_ERR(handle);
4891 ret = dquot_acquire(dquot);
617ba13b 4892 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4893 if (!ret)
4894 ret = err;
4895 return ret;
4896}
4897
617ba13b 4898static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4899{
4900 int ret, err;
4901 handle_t *handle;
4902
9924a92a 4903 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4904 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4905 if (IS_ERR(handle)) {
4906 /* Release dquot anyway to avoid endless cycle in dqput() */
4907 dquot_release(dquot);
ac27a0ec 4908 return PTR_ERR(handle);
9c3013e9 4909 }
ac27a0ec 4910 ret = dquot_release(dquot);
617ba13b 4911 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4912 if (!ret)
4913 ret = err;
4914 return ret;
4915}
4916
617ba13b 4917static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4918{
262b4662
JK		 4919 struct super_block *sb = dquot->dq_sb;
4920 struct ext4_sb_info *sbi = EXT4_SB(sb);
4921
2c8be6b2 4922 /* Are we journaling quotas? */
262b4662
JK		 4923 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
4924 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 4925 dquot_mark_dquot_dirty(dquot);
617ba13b 4926 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4927 } else {
4928 return dquot_mark_dquot_dirty(dquot);
4929 }
4930}
4931
617ba13b 4932static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4933{
4934 int ret, err;
4935 handle_t *handle;
4936
4937 /* Data block + inode block */
9924a92a 4938 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 4939 if (IS_ERR(handle))
4940 return PTR_ERR(handle);
4941 ret = dquot_commit_info(sb, type);
617ba13b 4942 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4943 if (!ret)
4944 ret = err;
4945 return ret;
4946}
4947
4948/*
4949 * Turn on quotas during mount time - we need to find
4950 * the quota file and such...
4951 */
617ba13b 4952static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4953{
287a8095
CH		 4954 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4955 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4956}
4957
4958/*
4959 * Standard function to be called on quota_on
4960 */
617ba13b 4961static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4962 struct path *path)
ac27a0ec
DK		 4963{
4964 int err;
ac27a0ec
DK		 4965
4966 if (!test_opt(sb, QUOTA))
4967 return -EINVAL;
0623543b 4968
ac27a0ec 4969 /* Quotafile not on the same filesystem? */
d8c9584e 4970 if (path->dentry->d_sb != sb)
ac27a0ec 4971 return -EXDEV;
0623543b
JK		 4972 /* Journaling quota? */
4973 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4974 /* Quotafile not in fs root? */
f00c9e44 4975 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4976 ext4_msg(sb, KERN_WARNING,
4977 "Quota file not on filesystem root. "
4978 "Journaled quota will not work");
2b2d6d01 4979 }
0623543b
JK		 4980
4981 /*
4982 * When we journal data on quota file, we have to flush journal to see
4983 * all updates to the file when we bypass pagecache...
4984 */
0390131b 4985 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4986 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4987 /*
4988 * We don't need to lock updates but journal_flush() could
4989 * otherwise be livelocked...
4990 */
4991 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4992 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4993 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4994 if (err)
7ffe1ea8 4995 return err;
0623543b
JK		 4996 }
4997
f00c9e44 4998 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4999}
5000
7c319d32
AK		 5001static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5002 unsigned int flags)
5003{
5004 int err;
5005 struct inode *qf_inode;
5006 unsigned long qf_inums[MAXQUOTAS] = {
5007 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5008 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5009 };
5010
5011 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5012
5013 if (!qf_inums[type])
5014 return -EPERM;
5015
5016 qf_inode = ext4_iget(sb, qf_inums[type]);
5017 if (IS_ERR(qf_inode)) {
5018 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5019 return PTR_ERR(qf_inode);
5020 }
5021
bcb13850
JK		 5022 /* Don't account quota for quota files to avoid recursion */
5023 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5024 err = dquot_enable(qf_inode, type, format_id, flags);
5025 iput(qf_inode);
5026
5027 return err;
5028}
5029
5030/* Enable usage tracking for all quota types. */
5031static int ext4_enable_quotas(struct super_block *sb)
5032{
5033 int type, err = 0;
5034 unsigned long qf_inums[MAXQUOTAS] = {
5035 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5036 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5037 };
5038
5039 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5040 for (type = 0; type < MAXQUOTAS; type++) {
5041 if (qf_inums[type]) {
5042 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5043 DQUOT_USAGE_ENABLED);
5044 if (err) {
5045 ext4_warning(sb,
72ba7450
TT		 5046 "Failed to enable quota tracking "
5047 "(type=%d, err=%d). Please run "
5048 "e2fsck to fix.", type, err);
7c319d32
AK		 5049 return err;
5050 }
5051 }
5052 }
5053 return 0;
5054}
5055
5056/*
5057 * quota_on function that is used when QUOTA feature is set.
5058 */
5059static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5060 int format_id)
5061{
5062 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5063 return -EINVAL;
5064
5065 /*
5066 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5067 */
5068 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5069}
5070
ca0e05e4
DM		 5071static int ext4_quota_off(struct super_block *sb, int type)
5072{
21f97697
JK		 5073 struct inode *inode = sb_dqopt(sb)->files[type];
5074 handle_t *handle;
5075
87009d86
DM		 5076 /* Force all delayed allocation blocks to be allocated.
5077 * Caller already holds s_umount sem */
5078 if (test_opt(sb, DELALLOC))
ca0e05e4 5079 sync_filesystem(sb);
ca0e05e4 5080
0b268590
AG		 5081 if (!inode)
5082 goto out;
5083
21f97697
JK		 5084 /* Update modification times of quota files when userspace can
5085 * start looking at them */
9924a92a 5086 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5087 if (IS_ERR(handle))
5088 goto out;
5089 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5090 ext4_mark_inode_dirty(handle, inode);
5091 ext4_journal_stop(handle);
5092
5093out:
ca0e05e4
DM		 5094 return dquot_quota_off(sb, type);
5095}
5096
7c319d32
AK		 5097/*
5098 * quota_off function that is used when QUOTA feature is set.
5099 */
5100static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5101{
5102 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5103 return -EINVAL;
5104
5105 /* Disable only the limits. */
5106 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5107}
5108
ac27a0ec
DK		 5109/* Read data from quotafile - avoid pagecache and such because we cannot afford
5110 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5111 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5112 * we don't have to be afraid of races */
617ba13b 5113static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5114 size_t len, loff_t off)
5115{
5116 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5117 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5118 int err = 0;
5119 int offset = off & (sb->s_blocksize - 1);
5120 int tocopy;
5121 size_t toread;
5122 struct buffer_head *bh;
5123 loff_t i_size = i_size_read(inode);
5124
5125 if (off > i_size)
5126 return 0;
5127 if (off+len > i_size)
5128 len = i_size-off;
5129 toread = len;
5130 while (toread > 0) {
5131 tocopy = sb->s_blocksize - offset < toread ?
5132 sb->s_blocksize - offset : toread;
617ba13b 5133 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5134 if (err)
5135 return err;
5136 if (!bh) /* A hole? */
5137 memset(data, 0, tocopy);
5138 else
5139 memcpy(data, bh->b_data+offset, tocopy);
5140 brelse(bh);
5141 offset = 0;
5142 toread -= tocopy;
5143 data += tocopy;
5144 blk++;
5145 }
5146 return len;
5147}
5148
5149/* Write to quotafile (we know the transaction is already started and has
5150 * enough credits) */
617ba13b 5151static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5152 const char *data, size_t len, loff_t off)
5153{
5154 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5155 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5156 int err = 0;
5157 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5158 struct buffer_head *bh;
5159 handle_t *handle = journal_current_handle();
5160
0390131b 5161 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5162 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5163 " cancelled because transaction is not started",
9c3013e9
JK		 5164 (unsigned long long)off, (unsigned long long)len);
5165 return -EIO;
5166 }
67eeb568
DM		 5167 /*
5168 * Since we account only one data block in transaction credits,
5169 * then it is impossible to cross a block boundary.
5170 */
5171 if (sb->s_blocksize - offset < len) {
5172 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5173 " cancelled because not block aligned",
5174 (unsigned long long)off, (unsigned long long)len);
5175 return -EIO;
5176 }
5177
67eeb568
DM		 5178 bh = ext4_bread(handle, inode, blk, 1, &err);
5179 if (!bh)
5180 goto out;
62d2b5f2
JK		 5181 err = ext4_journal_get_write_access(handle, bh);
5182 if (err) {
5183 brelse(bh);
5184 goto out;
ac27a0ec 5185 }
67eeb568
DM		 5186 lock_buffer(bh);
5187 memcpy(bh->b_data+offset, data, len);
5188 flush_dcache_page(bh->b_page);
5189 unlock_buffer(bh);
62d2b5f2 5190 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5191 brelse(bh);
ac27a0ec 5192out:
0b7f7cef 5193 if (err)
ac27a0ec 5194 return err;
67eeb568
DM		 5195 if (inode->i_size < off + len) {
5196 i_size_write(inode, off + len);
617ba13b 5197 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5198 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5199 }
67eeb568 5200 return len;
ac27a0ec
DK		 5201}
5202
5203#endif
5204
152a0836
AV		 5205static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5206 const char *dev_name, void *data)
ac27a0ec 5207{
152a0836 5208 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5209}
5210
37f328eb 5211#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5212static inline void register_as_ext2(void)
5213{
5214 int err = register_filesystem(&ext2_fs_type);
5215 if (err)
5216 printk(KERN_WARNING
5217 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5218}
5219
5220static inline void unregister_as_ext2(void)
5221{
5222 unregister_filesystem(&ext2_fs_type);
5223}
2035e776
TT		 5224
5225static inline int ext2_feature_set_ok(struct super_block *sb)
5226{
5227 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5228 return 0;
5229 if (sb->s_flags & MS_RDONLY)
5230 return 1;
5231 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5232 return 0;
5233 return 1;
5234}
24b58424
TT		 5235#else
5236static inline void register_as_ext2(void) { }
5237static inline void unregister_as_ext2(void) { }
2035e776 5238static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5239#endif
5240
37f328eb 5241#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5242static inline void register_as_ext3(void)
5243{
5244 int err = register_filesystem(&ext3_fs_type);
5245 if (err)
5246 printk(KERN_WARNING
5247 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5248}
5249
5250static inline void unregister_as_ext3(void)
5251{
5252 unregister_filesystem(&ext3_fs_type);
5253}
2035e776
TT		 5254
5255static inline int ext3_feature_set_ok(struct super_block *sb)
5256{
5257 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5258 return 0;
5259 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5260 return 0;
5261 if (sb->s_flags & MS_RDONLY)
5262 return 1;
5263 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5264 return 0;
5265 return 1;
5266}
24b58424
TT		 5267#else
5268static inline void register_as_ext3(void) { }
5269static inline void unregister_as_ext3(void) { }
2035e776 5270static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5271#endif
5272
03010a33
TT		 5273static struct file_system_type ext4_fs_type = {
5274 .owner = THIS_MODULE,
5275 .name = "ext4",
152a0836 5276 .mount = ext4_mount,
03010a33
TT		 5277 .kill_sb = kill_block_super,
5278 .fs_flags = FS_REQUIRES_DEV,
5279};
7f78e035 5280MODULE_ALIAS_FS("ext4");
03010a33 5281
8f021222 5282static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5283{
5284 struct ext4_features *ef;
5285 int ret = -ENOMEM;
5286
5287 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5288 if (!ef)
5289 goto out;
5290
5291 ef->f_kobj.kset = ext4_kset;
5292 init_completion(&ef->f_kobj_unregister);
5293 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5294 "features");
5295 if (ret) {
5296 kfree(ef);
5297 goto out;
5298 }
5299
5300 ext4_feat = ef;
5301 ret = 0;
5302out:
5303 return ret;
5304}
5305
8f021222
LC		 5306static void ext4_exit_feat_adverts(void)
5307{
5308 kobject_put(&ext4_feat->f_kobj);
5309 wait_for_completion(&ext4_feat->f_kobj_unregister);
5310 kfree(ext4_feat);
5311}
5312
e9e3bcec
ES		 5313/* Shared across all ext4 file systems */
5314wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5315struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5316
5dabfc78 5317static int __init ext4_init_fs(void)
ac27a0ec 5318{
e9e3bcec 5319 int i, err;
c9de560d 5320
07c0c5d8
AV		 5321 ext4_li_info = NULL;
5322 mutex_init(&ext4_li_mtx);
5323
9a4c8019 5324 /* Build-time check for flags consistency */
12e9b892 5325 ext4_check_flag_values();
e9e3bcec
ES		 5326
5327 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5328 mutex_init(&ext4__aio_mutex[i]);
5329 init_waitqueue_head(&ext4__ioend_wq[i]);
5330 }
5331
51865fda 5332 err = ext4_init_es();
6fd058f7
TT		 5333 if (err)
5334 return err;
51865fda
ZL		 5335
5336 err = ext4_init_pageio();
5337 if (err)
5338 goto out7;
5339
5dabfc78 5340 err = ext4_init_system_zone();
bd2d0210 5341 if (err)
d44651d0 5342 goto out6;
3197ebdb 5343 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5344 if (!ext4_kset) {
5345 err = -ENOMEM;
dd68314c 5346 goto out5;
0e376b1e 5347 }
d44651d0 5348 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5349
5350 err = ext4_init_feat_adverts();
dd68314c
TT		 5351 if (err)
5352 goto out4;
857ac889 5353
5dabfc78 5354 err = ext4_init_mballoc();
ac27a0ec 5355 if (err)
6fd058f7 5356 goto out3;
c9de560d 5357
5dabfc78 5358 err = ext4_init_xattr();
c9de560d
AT		 5359 if (err)
5360 goto out2;
ac27a0ec
DK		 5361 err = init_inodecache();
5362 if (err)
5363 goto out1;
24b58424 5364 register_as_ext3();
2035e776 5365 register_as_ext2();
03010a33 5366 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5367 if (err)
5368 goto out;
bfff6873 5369
ac27a0ec
DK		 5370 return 0;
5371out:
24b58424
TT		 5372 unregister_as_ext2();
5373 unregister_as_ext3();
ac27a0ec
DK		 5374 destroy_inodecache();
5375out1:
5dabfc78 5376 ext4_exit_xattr();
c9de560d 5377out2:
5dabfc78 5378 ext4_exit_mballoc();
6fd058f7 5379out3:
8f021222 5380 ext4_exit_feat_adverts();
dd68314c 5381out4:
d44651d0
FJ		 5382 if (ext4_proc_root)
5383 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5384 kset_unregister(ext4_kset);
d44651d0 5385out5:
5dabfc78 5386 ext4_exit_system_zone();
d44651d0 5387out6:
5dabfc78 5388 ext4_exit_pageio();
51865fda
ZL		 5389out7:
5390 ext4_exit_es();
5391
ac27a0ec
DK		 5392 return err;
5393}
5394
5dabfc78 5395static void __exit ext4_exit_fs(void)
ac27a0ec 5396{
bfff6873 5397 ext4_destroy_lazyinit_thread();
24b58424
TT		 5398 unregister_as_ext2();
5399 unregister_as_ext3();
03010a33 5400 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5401 destroy_inodecache();
5dabfc78
TT		 5402 ext4_exit_xattr();
5403 ext4_exit_mballoc();
8f021222 5404 ext4_exit_feat_adverts();
9f6200bb 5405 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5406 kset_unregister(ext4_kset);
5dabfc78
TT		 5407 ext4_exit_system_zone();
5408 ext4_exit_pageio();
ac27a0ec
DK		 5409}
5410
5411MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5412MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5413MODULE_LICENSE("GPL");
5dabfc78
TT		 5414module_init(ext4_init_fs)
5415module_exit(ext4_exit_fs)