projects / GitHub / LineageOS / android_kernel_motorola_exynos9610.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: remove duplicate remount check for JOURNAL_CHECKSUM change
[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;
9c191f70 62static int ext4_mballoc_ready;
9f6200bb 63
617ba13b 64static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 65 unsigned long journal_devnum);
2adf6da8 66static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 67static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 68static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
617ba13b 72static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 73static int ext4_remount(struct super_block *sb, int *flags, char *data);
74static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 75static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 76static int ext4_freeze(struct super_block *sb);
152a0836
AV		 77static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
2035e776
TT		 79static inline int ext2_feature_set_ok(struct super_block *sb);
80static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 81static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 82static void ext4_destroy_lazyinit_thread(void);
83static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 84static void ext4_clear_request_list(void);
27dd4385 85static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
ac27a0ec 86
2035e776
TT		 87#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88static struct file_system_type ext2_fs_type = {
89 .owner = THIS_MODULE,
90 .name = "ext2",
91 .mount = ext4_mount,
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
94};
7f78e035 95MODULE_ALIAS_FS("ext2");
fa7614dd 96MODULE_ALIAS("ext2");
2035e776
TT		 97#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98#else
99#define IS_EXT2_SB(sb) (0)
100#endif
101
102
ba69f9ab
JK		 103#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104static struct file_system_type ext3_fs_type = {
105 .owner = THIS_MODULE,
106 .name = "ext3",
152a0836 107 .mount = ext4_mount,
ba69f9ab
JK		 108 .kill_sb = kill_block_super,
109 .fs_flags = FS_REQUIRES_DEV,
110};
7f78e035 111MODULE_ALIAS_FS("ext3");
fa7614dd 112MODULE_ALIAS("ext3");
ba69f9ab
JK		 113#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
114#else
115#define IS_EXT3_SB(sb) (0)
116#endif
bd81d8ee 117
d25425f8
DW		 118static int ext4_verify_csum_type(struct super_block *sb,
119 struct ext4_super_block *es)
120{
121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
122 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
123 return 1;
124
125 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
126}
127
a9c47317
DW		 128static __le32 ext4_superblock_csum(struct super_block *sb,
129 struct ext4_super_block *es)
130{
131 struct ext4_sb_info *sbi = EXT4_SB(sb);
132 int offset = offsetof(struct ext4_super_block, s_checksum);
133 __u32 csum;
134
135 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
136
137 return cpu_to_le32(csum);
138}
139
c197855e
SH		 140static int ext4_superblock_csum_verify(struct super_block *sb,
141 struct ext4_super_block *es)
a9c47317 142{
9aa5d32b 143 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 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
9aa5d32b 153 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 154 return;
155
156 es->s_checksum = ext4_superblock_csum(sb, es);
157}
158
9933fc0a
TT		 159void *ext4_kvmalloc(size_t size, gfp_t flags)
160{
161 void *ret;
162
8be04b93 163 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 164 if (!ret)
165 ret = __vmalloc(size, flags, PAGE_KERNEL);
166 return ret;
167}
168
169void *ext4_kvzalloc(size_t size, gfp_t flags)
170{
171 void *ret;
172
8be04b93 173 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 174 if (!ret)
175 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
176 return ret;
177}
178
8fadc143
AR		 179ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
180 struct ext4_group_desc *bg)
bd81d8ee 181{
3a14589c 182 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 183 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 184 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 185}
186
8fadc143
AR		 187ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
188 struct ext4_group_desc *bg)
bd81d8ee 189{
5272f837 190 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 191 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 192 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 193}
194
8fadc143
AR		 195ext4_fsblk_t ext4_inode_table(struct super_block *sb,
196 struct ext4_group_desc *bg)
bd81d8ee 197{
5272f837 198 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 199 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 200 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 201}
202
021b65bb
TT		 203__u32 ext4_free_group_clusters(struct super_block *sb,
204 struct ext4_group_desc *bg)
560671a0
AK		 205{
206 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 208 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 209}
210
211__u32 ext4_free_inodes_count(struct super_block *sb,
212 struct ext4_group_desc *bg)
213{
214 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 216 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 217}
218
219__u32 ext4_used_dirs_count(struct super_block *sb,
220 struct ext4_group_desc *bg)
221{
222 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 224 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 225}
226
227__u32 ext4_itable_unused_count(struct super_block *sb,
228 struct ext4_group_desc *bg)
229{
230 return le16_to_cpu(bg->bg_itable_unused_lo) |
231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 232 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 233}
234
8fadc143
AR		 235void ext4_block_bitmap_set(struct super_block *sb,
236 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 237{
3a14589c 238 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 239 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
240 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 241}
242
8fadc143
AR		 243void ext4_inode_bitmap_set(struct super_block *sb,
244 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 245{
5272f837 246 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 247 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
248 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 249}
250
8fadc143
AR		 251void ext4_inode_table_set(struct super_block *sb,
252 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 253{
5272f837 254 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 255 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
256 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 257}
258
021b65bb
TT		 259void ext4_free_group_clusters_set(struct super_block *sb,
260 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 261{
262 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
265}
266
267void ext4_free_inodes_set(struct super_block *sb,
268 struct ext4_group_desc *bg, __u32 count)
269{
270 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
273}
274
275void ext4_used_dirs_set(struct super_block *sb,
276 struct ext4_group_desc *bg, __u32 count)
277{
278 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
281}
282
283void ext4_itable_unused_set(struct super_block *sb,
284 struct ext4_group_desc *bg, __u32 count)
285{
286 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
289}
290
d3d1faf6 291
1c13d5c0
TT		 292static void __save_error_info(struct super_block *sb, const char *func,
293 unsigned int line)
294{
295 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
296
297 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
298 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
299 es->s_last_error_time = cpu_to_le32(get_seconds());
300 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
301 es->s_last_error_line = cpu_to_le32(line);
302 if (!es->s_first_error_time) {
303 es->s_first_error_time = es->s_last_error_time;
304 strncpy(es->s_first_error_func, func,
305 sizeof(es->s_first_error_func));
306 es->s_first_error_line = cpu_to_le32(line);
307 es->s_first_error_ino = es->s_last_error_ino;
308 es->s_first_error_block = es->s_last_error_block;
309 }
66e61a9e
TT		 310 /*
311 * Start the daily error reporting function if it hasn't been
312 * started already
313 */
314 if (!es->s_error_count)
315 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 316 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 317}
318
319static void save_error_info(struct super_block *sb, const char *func,
320 unsigned int line)
321{
322 __save_error_info(sb, func, line);
323 ext4_commit_super(sb, 1);
324}
325
7c2e7087
TT		 326/*
327 * The del_gendisk() function uninitializes the disk-specific data
328 * structures, including the bdi structure, without telling anyone
329 * else. Once this happens, any attempt to call mark_buffer_dirty()
330 * (for example, by ext4_commit_super), will cause a kernel OOPS.
331 * This is a kludge to prevent these oops until we can put in a proper
332 * hook in del_gendisk() to inform the VFS and file system layers.
333 */
334static int block_device_ejected(struct super_block *sb)
335{
336 struct inode *bd_inode = sb->s_bdev->bd_inode;
337 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
338
339 return bdi->dev == NULL;
340}
341
18aadd47
BJ		 342static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
343{
344 struct super_block *sb = journal->j_private;
345 struct ext4_sb_info *sbi = EXT4_SB(sb);
346 int error = is_journal_aborted(journal);
5d3ee208 347 struct ext4_journal_cb_entry *jce;
18aadd47 348
5d3ee208 349 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 350 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 351 while (!list_empty(&txn->t_private_list)) {
352 jce = list_entry(txn->t_private_list.next,
353 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 354 list_del_init(&jce->jce_list);
355 spin_unlock(&sbi->s_md_lock);
356 jce->jce_func(sb, jce, error);
357 spin_lock(&sbi->s_md_lock);
358 }
359 spin_unlock(&sbi->s_md_lock);
360}
1c13d5c0 361
ac27a0ec
DK		 362/* Deal with the reporting of failure conditions on a filesystem such as
363 * inconsistencies detected or read IO failures.
364 *
365 * On ext2, we can store the error state of the filesystem in the
617ba13b 366 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 367 * write ordering constraints on the superblock which prevent us from
368 * writing it out straight away; and given that the journal is about to
369 * be aborted, we can't rely on the current, or future, transactions to
370 * write out the superblock safely.
371 *
dab291af 372 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 373 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 374 * that error until we've noted it down and cleared it.
375 */
376
617ba13b 377static void ext4_handle_error(struct super_block *sb)
ac27a0ec 378{
ac27a0ec
DK		 379 if (sb->s_flags & MS_RDONLY)
380 return;
381
2b2d6d01 382 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 383 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 384
4ab2f15b 385 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 386 if (journal)
dab291af 387 jbd2_journal_abort(journal, -EIO);
ac27a0ec 388 }
2b2d6d01 389 if (test_opt(sb, ERRORS_RO)) {
b31e1552 390 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 391 /*
392 * Make sure updated value of ->s_mount_flags will be visible
393 * before ->s_flags update
394 */
395 smp_wmb();
ac27a0ec
DK		 396 sb->s_flags |= MS_RDONLY;
397 }
ac27a0ec 398 if (test_opt(sb, ERRORS_PANIC))
617ba13b 399 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 400 sb->s_id);
401}
402
efbed4dc
TT		 403#define ext4_error_ratelimit(sb) \
404 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
405 "EXT4-fs error")
406
12062ddd 407void __ext4_error(struct super_block *sb, const char *function,
c398eda0 408 unsigned int line, const char *fmt, ...)
ac27a0ec 409{
0ff2ea7d 410 struct va_format vaf;
ac27a0ec
DK		 411 va_list args;
412
efbed4dc
TT		 413 if (ext4_error_ratelimit(sb)) {
414 va_start(args, fmt);
415 vaf.fmt = fmt;
416 vaf.va = &args;
417 printk(KERN_CRIT
418 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
419 sb->s_id, function, line, current->comm, &vaf);
420 va_end(args);
421 }
f3fc0210 422 save_error_info(sb, function, line);
617ba13b 423 ext4_handle_error(sb);
ac27a0ec
DK		 424}
425
e7c96e8e
JP		 426void __ext4_error_inode(struct inode *inode, const char *function,
427 unsigned int line, ext4_fsblk_t block,
428 const char *fmt, ...)
273df556
FM		 429{
430 va_list args;
f7c21177 431 struct va_format vaf;
1c13d5c0 432 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 433
1c13d5c0
TT		 434 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
435 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 436 if (ext4_error_ratelimit(inode->i_sb)) {
437 va_start(args, fmt);
438 vaf.fmt = fmt;
439 vaf.va = &args;
440 if (block)
441 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
442 "inode #%lu: block %llu: comm %s: %pV\n",
443 inode->i_sb->s_id, function, line, inode->i_ino,
444 block, current->comm, &vaf);
445 else
446 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
447 "inode #%lu: comm %s: %pV\n",
448 inode->i_sb->s_id, function, line, inode->i_ino,
449 current->comm, &vaf);
450 va_end(args);
451 }
1c13d5c0 452 save_error_info(inode->i_sb, function, line);
273df556
FM		 453 ext4_handle_error(inode->i_sb);
454}
455
e7c96e8e
JP		 456void __ext4_error_file(struct file *file, const char *function,
457 unsigned int line, ext4_fsblk_t block,
458 const char *fmt, ...)
273df556
FM		 459{
460 va_list args;
f7c21177 461 struct va_format vaf;
1c13d5c0 462 struct ext4_super_block *es;
496ad9aa 463 struct inode *inode = file_inode(file);
273df556
FM		 464 char pathname[80], *path;
465
1c13d5c0
TT		 466 es = EXT4_SB(inode->i_sb)->s_es;
467 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc
TT		 468 if (ext4_error_ratelimit(inode->i_sb)) {
469 path = d_path(&(file->f_path), pathname, sizeof(pathname));
470 if (IS_ERR(path))
471 path = "(unknown)";
472 va_start(args, fmt);
473 vaf.fmt = fmt;
474 vaf.va = &args;
475 if (block)
476 printk(KERN_CRIT
477 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
478 "block %llu: comm %s: path %s: %pV\n",
479 inode->i_sb->s_id, function, line, inode->i_ino,
480 block, current->comm, path, &vaf);
481 else
482 printk(KERN_CRIT
483 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
484 "comm %s: path %s: %pV\n",
485 inode->i_sb->s_id, function, line, inode->i_ino,
486 current->comm, path, &vaf);
487 va_end(args);
488 }
1c13d5c0 489 save_error_info(inode->i_sb, function, line);
273df556
FM		 490 ext4_handle_error(inode->i_sb);
491}
492
722887dd
TT		 493const char *ext4_decode_error(struct super_block *sb, int errno,
494 char nbuf[16])
ac27a0ec
DK		 495{
496 char *errstr = NULL;
497
498 switch (errno) {
499 case -EIO:
500 errstr = "IO failure";
501 break;
502 case -ENOMEM:
503 errstr = "Out of memory";
504 break;
505 case -EROFS:
78f1ddbb
TT		 506 if (!sb || (EXT4_SB(sb)->s_journal &&
507 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 508 errstr = "Journal has aborted";
509 else
510 errstr = "Readonly filesystem";
511 break;
512 default:
513 /* If the caller passed in an extra buffer for unknown
514 * errors, textualise them now. Else we just return
515 * NULL. */
516 if (nbuf) {
517 /* Check for truncated error codes... */
518 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
519 errstr = nbuf;
520 }
521 break;
522 }
523
524 return errstr;
525}
526
617ba13b 527/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 528 * automatically and invokes the appropriate error response. */
529
c398eda0
TT		 530void __ext4_std_error(struct super_block *sb, const char *function,
531 unsigned int line, int errno)
ac27a0ec
DK		 532{
533 char nbuf[16];
534 const char *errstr;
535
536 /* Special case: if the error is EROFS, and we're not already
537 * inside a transaction, then there's really no point in logging
538 * an error. */
539 if (errno == -EROFS && journal_current_handle() == NULL &&
540 (sb->s_flags & MS_RDONLY))
541 return;
542
efbed4dc
TT		 543 if (ext4_error_ratelimit(sb)) {
544 errstr = ext4_decode_error(sb, errno, nbuf);
545 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
546 sb->s_id, function, line, errstr);
547 }
ac27a0ec 548
efbed4dc 549 save_error_info(sb, function, line);
617ba13b 550 ext4_handle_error(sb);
ac27a0ec
DK		 551}
552
553/*
617ba13b 554 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 555 * abort function may be used to deal with unrecoverable failures such
556 * as journal IO errors or ENOMEM at a critical moment in log management.
557 *
558 * We unconditionally force the filesystem into an ABORT|READONLY state,
559 * unless the error response on the fs has been set to panic in which
560 * case we take the easy way out and panic immediately.
561 */
562
c67d859e 563void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 564 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 565{
566 va_list args;
567
1c13d5c0 568 save_error_info(sb, function, line);
ac27a0ec 569 va_start(args, fmt);
c398eda0
TT		 570 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
571 function, line);
ac27a0ec
DK		 572 vprintk(fmt, args);
573 printk("\n");
574 va_end(args);
575
1c13d5c0
TT		 576 if ((sb->s_flags & MS_RDONLY) == 0) {
577 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 578 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 579 /*
580 * Make sure updated value of ->s_mount_flags will be visible
581 * before ->s_flags update
582 */
583 smp_wmb();
584 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 585 if (EXT4_SB(sb)->s_journal)
586 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
587 save_error_info(sb, function, line);
588 }
ac27a0ec 589 if (test_opt(sb, ERRORS_PANIC))
617ba13b 590 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 591}
592
e7c96e8e
JP		 593void __ext4_msg(struct super_block *sb,
594 const char *prefix, const char *fmt, ...)
b31e1552 595{
0ff2ea7d 596 struct va_format vaf;
b31e1552
ES		 597 va_list args;
598
efbed4dc
TT		 599 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
600 return;
601
b31e1552 602 va_start(args, fmt);
0ff2ea7d
JP		 603 vaf.fmt = fmt;
604 vaf.va = &args;
605 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 606 va_end(args);
607}
608
12062ddd 609void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 610 unsigned int line, const char *fmt, ...)
ac27a0ec 611{
0ff2ea7d 612 struct va_format vaf;
ac27a0ec
DK		 613 va_list args;
614
efbed4dc
TT		 615 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
616 "EXT4-fs warning"))
617 return;
618
ac27a0ec 619 va_start(args, fmt);
0ff2ea7d
JP		 620 vaf.fmt = fmt;
621 vaf.va = &args;
622 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
623 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 624 va_end(args);
625}
626
e29136f8
TT		 627void __ext4_grp_locked_error(const char *function, unsigned int line,
628 struct super_block *sb, ext4_group_t grp,
629 unsigned long ino, ext4_fsblk_t block,
630 const char *fmt, ...)
5d1b1b3f
AK		 631__releases(bitlock)
632__acquires(bitlock)
633{
0ff2ea7d 634 struct va_format vaf;
5d1b1b3f
AK		 635 va_list args;
636 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
637
1c13d5c0
TT		 638 es->s_last_error_ino = cpu_to_le32(ino);
639 es->s_last_error_block = cpu_to_le64(block);
640 __save_error_info(sb, function, line);
0ff2ea7d 641
efbed4dc
TT		 642 if (ext4_error_ratelimit(sb)) {
643 va_start(args, fmt);
644 vaf.fmt = fmt;
645 vaf.va = &args;
646 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
647 sb->s_id, function, line, grp);
648 if (ino)
649 printk(KERN_CONT "inode %lu: ", ino);
650 if (block)
651 printk(KERN_CONT "block %llu:",
652 (unsigned long long) block);
653 printk(KERN_CONT "%pV\n", &vaf);
654 va_end(args);
655 }
5d1b1b3f
AK		 656
657 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 658 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 659 return;
660 }
1c13d5c0 661
5d1b1b3f
AK		 662 ext4_unlock_group(sb, grp);
663 ext4_handle_error(sb);
664 /*
665 * We only get here in the ERRORS_RO case; relocking the group
666 * may be dangerous, but nothing bad will happen since the
667 * filesystem will have already been marked read/only and the
668 * journal has been aborted. We return 1 as a hint to callers
669 * who might what to use the return value from
25985edc 670 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 671 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
672 * aggressively from the ext4 function in question, with a
673 * more appropriate error code.
674 */
675 ext4_lock_group(sb, grp);
676 return;
677}
678
617ba13b 679void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 680{
617ba13b 681 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 682
617ba13b 683 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 684 return;
685
12062ddd 686 ext4_warning(sb,
ac27a0ec
DK		 687 "updating to rev %d because of new feature flag, "
688 "running e2fsck is recommended",
617ba13b 689 EXT4_DYNAMIC_REV);
ac27a0ec 690
617ba13b
MC		 691 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
692 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
693 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 694 /* leave es->s_feature_*compat flags alone */
695 /* es->s_uuid will be set by e2fsck if empty */
696
697 /*
698 * The rest of the superblock fields should be zero, and if not it
699 * means they are likely already in use, so leave them alone. We
700 * can leave it up to e2fsck to clean up any inconsistencies there.
701 */
702}
703
704/*
705 * Open the external journal device
706 */
b31e1552 707static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 708{
709 struct block_device *bdev;
710 char b[BDEVNAME_SIZE];
711
d4d77629 712 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 713 if (IS_ERR(bdev))
714 goto fail;
715 return bdev;
716
717fail:
b31e1552 718 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 719 __bdevname(dev, b), PTR_ERR(bdev));
720 return NULL;
721}
722
723/*
724 * Release the journal device
725 */
4385bab1 726static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 727{
4385bab1 728 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 729}
730
4385bab1 731static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 732{
733 struct block_device *bdev;
ac27a0ec
DK		 734 bdev = sbi->journal_bdev;
735 if (bdev) {
4385bab1 736 ext4_blkdev_put(bdev);
ac27a0ec
DK		 737 sbi->journal_bdev = NULL;
738 }
ac27a0ec
DK		 739}
740
741static inline struct inode *orphan_list_entry(struct list_head *l)
742{
617ba13b 743 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 744}
745
617ba13b 746static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 747{
748 struct list_head *l;
749
b31e1552
ES		 750 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
751 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 752
753 printk(KERN_ERR "sb_info orphan list:\n");
754 list_for_each(l, &sbi->s_orphan) {
755 struct inode *inode = orphan_list_entry(l);
756 printk(KERN_ERR " "
757 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
758 inode->i_sb->s_id, inode->i_ino, inode,
759 inode->i_mode, inode->i_nlink,
760 NEXT_ORPHAN(inode));
761 }
762}
763
2b2d6d01 764static void ext4_put_super(struct super_block *sb)
ac27a0ec 765{
617ba13b
MC		 766 struct ext4_sb_info *sbi = EXT4_SB(sb);
767 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 768 int i, err;
ac27a0ec 769
857ac889 770 ext4_unregister_li_request(sb);
e0ccfd95
CH		 771 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
772
2e8fa54e 773 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 774 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 775
0390131b
FM		 776 if (sbi->s_journal) {
777 err = jbd2_journal_destroy(sbi->s_journal);
778 sbi->s_journal = NULL;
779 if (err < 0)
c67d859e 780 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 781 }
d4edac31 782
d3922a77 783 ext4_es_unregister_shrinker(sbi);
9105bb14 784 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 785 ext4_release_system_zone(sb);
786 ext4_mb_release(sb);
787 ext4_ext_release(sb);
788 ext4_xattr_put_super(sb);
789
ac27a0ec 790 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 791 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 792 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 793 }
58c5873a 794 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 795 ext4_commit_super(sb, 1);
796
240799cd 797 if (sbi->s_proc) {
66acdcf4 798 remove_proc_entry("options", sbi->s_proc);
9f6200bb 799 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 800 }
3197ebdb 801 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 802
803 for (i = 0; i < sbi->s_gdb_count; i++)
804 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 805 kvfree(sbi->s_group_desc);
806 kvfree(sbi->s_flex_groups);
57042651 807 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 808 percpu_counter_destroy(&sbi->s_freeinodes_counter);
809 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 810 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 811 brelse(sbi->s_sbh);
812#ifdef CONFIG_QUOTA
a2d4a646 813 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 814 kfree(sbi->s_qf_names[i]);
815#endif
816
817 /* Debugging code just in case the in-memory inode orphan list
818 * isn't empty. The on-disk one can be non-empty if we've
819 * detected an error and taken the fs readonly, but the
820 * in-memory list had better be clean by this point. */
821 if (!list_empty(&sbi->s_orphan))
822 dump_orphan_list(sb, sbi);
823 J_ASSERT(list_empty(&sbi->s_orphan));
824
f98393a6 825 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 826 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
827 /*
828 * Invalidate the journal device's buffers. We don't want them
829 * floating about in memory - the physical journal device may
830 * hotswapped, and it breaks the `ro-after' testing code.
831 */
832 sync_blockdev(sbi->journal_bdev);
f98393a6 833 invalidate_bdev(sbi->journal_bdev);
617ba13b 834 ext4_blkdev_remove(sbi);
ac27a0ec 835 }
9c191f70
M		 836 if (sbi->s_mb_cache) {
837 ext4_xattr_destroy_cache(sbi->s_mb_cache);
838 sbi->s_mb_cache = NULL;
839 }
c5e06d10
JL		 840 if (sbi->s_mmp_tsk)
841 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 842 sb->s_fs_info = NULL;
3197ebdb
TT		 843 /*
844 * Now that we are completely done shutting down the
845 * superblock, we need to actually destroy the kobject.
846 */
3197ebdb
TT		 847 kobject_put(&sbi->s_kobj);
848 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 849 if (sbi->s_chksum_driver)
850 crypto_free_shash(sbi->s_chksum_driver);
705895b6 851 kfree(sbi->s_blockgroup_lock);
ac27a0ec 852 kfree(sbi);
ac27a0ec
DK		 853}
854
e18b890b 855static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 856
857/*
858 * Called inside transaction, so use GFP_NOFS
859 */
617ba13b 860static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 861{
617ba13b 862 struct ext4_inode_info *ei;
ac27a0ec 863
e6b4f8da 864 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 865 if (!ei)
866 return NULL;
0b8e58a1 867
ac27a0ec 868 ei->vfs_inode.i_version = 1;
202ee5df 869 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 870 INIT_LIST_HEAD(&ei->i_prealloc_list);
871 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 872 ext4_es_init_tree(&ei->i_es_tree);
873 rwlock_init(&ei->i_es_lock);
edaa53ca 874 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 875 ei->i_es_all_nr = 0;
edaa53ca 876 ei->i_es_shk_nr = 0;
dd475925 877 ei->i_es_shrink_lblk = 0;
d2a17637
MC		 878 ei->i_reserved_data_blocks = 0;
879 ei->i_reserved_meta_blocks = 0;
880 ei->i_allocated_meta_blocks = 0;
9d0be502 881 ei->i_da_metadata_calc_len = 0;
7e731bc9 882 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 883 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 884#ifdef CONFIG_QUOTA
885 ei->i_reserved_quota = 0;
96c7e0d9 886 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 887#endif
8aefcd55 888 ei->jinode = NULL;
2e8fa54e 889 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 890 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 891 ei->i_sync_tid = 0;
892 ei->i_datasync_tid = 0;
f7ad6d2e 893 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 894 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 895 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
0b8e58a1 896
ac27a0ec
DK		 897 return &ei->vfs_inode;
898}
899
7ff9c073
TT		 900static int ext4_drop_inode(struct inode *inode)
901{
902 int drop = generic_drop_inode(inode);
903
904 trace_ext4_drop_inode(inode, drop);
905 return drop;
906}
907
fa0d7e3d
NP		 908static void ext4_i_callback(struct rcu_head *head)
909{
910 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 911 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
912}
913
617ba13b 914static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 915{
9f7dd93d 916 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 917 ext4_msg(inode->i_sb, KERN_ERR,
918 "Inode %lu (%p): orphan list check failed!",
919 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 920 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
921 EXT4_I(inode), sizeof(struct ext4_inode_info),
922 true);
923 dump_stack();
924 }
fa0d7e3d 925 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 926}
927
51cc5068 928static void init_once(void *foo)
ac27a0ec 929{
617ba13b 930 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 931
a35afb83 932 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 933 init_rwsem(&ei->xattr_sem);
0e855ac8 934 init_rwsem(&ei->i_data_sem);
a35afb83 935 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 936}
937
e67bc2b3 938static int __init init_inodecache(void)
ac27a0ec 939{
617ba13b
MC		 940 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
941 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 942 0, (SLAB_RECLAIM_ACCOUNT|
943 SLAB_MEM_SPREAD),
20c2df83 944 init_once);
617ba13b 945 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 946 return -ENOMEM;
947 return 0;
948}
949
950static void destroy_inodecache(void)
951{
8c0a8537
KS		 952 /*
953 * Make sure all delayed rcu free inodes are flushed before we
954 * destroy cache.
955 */
956 rcu_barrier();
617ba13b 957 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 958}
959
0930fcc1 960void ext4_clear_inode(struct inode *inode)
ac27a0ec 961{
0930fcc1 962 invalidate_inode_buffers(inode);
dbd5768f 963 clear_inode(inode);
9f754758 964 dquot_drop(inode);
c2ea3fde 965 ext4_discard_preallocations(inode);
51865fda 966 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 967 if (EXT4_I(inode)->jinode) {
968 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
969 EXT4_I(inode)->jinode);
970 jbd2_free_inode(EXT4_I(inode)->jinode);
971 EXT4_I(inode)->jinode = NULL;
972 }
ac27a0ec
DK		 973}
974
1b961ac0 975static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 976 u64 ino, u32 generation)
ac27a0ec 977{
ac27a0ec 978 struct inode *inode;
ac27a0ec 979
617ba13b 980 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 981 return ERR_PTR(-ESTALE);
617ba13b 982 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 983 return ERR_PTR(-ESTALE);
984
985 /* iget isn't really right if the inode is currently unallocated!!
986 *
617ba13b 987 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 988 * deleted, so we should be safe.
989 *
990 * Currently we don't know the generation for parent directory, so
991 * a generation of 0 means "accept any"
992 */
f4bb2981 993 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 994 if (IS_ERR(inode))
995 return ERR_CAST(inode);
996 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 997 iput(inode);
998 return ERR_PTR(-ESTALE);
999 }
1b961ac0
CH		 1000
1001 return inode;
1002}
1003
1004static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1005 int fh_len, int fh_type)
1b961ac0
CH		 1006{
1007 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1008 ext4_nfs_get_inode);
1009}
1010
1011static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1012 int fh_len, int fh_type)
1b961ac0
CH		 1013{
1014 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1015 ext4_nfs_get_inode);
ac27a0ec
DK		 1016}
1017
c39a7f84
TO		 1018/*
1019 * Try to release metadata pages (indirect blocks, directories) which are
1020 * mapped via the block device. Since these pages could have journal heads
1021 * which would prevent try_to_free_buffers() from freeing them, we must use
1022 * jbd2 layer's try_to_free_buffers() function to release them.
1023 */
0b8e58a1
AD		 1024static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1025 gfp_t wait)
c39a7f84
TO		 1026{
1027 journal_t *journal = EXT4_SB(sb)->s_journal;
1028
1029 WARN_ON(PageChecked(page));
1030 if (!page_has_buffers(page))
1031 return 0;
1032 if (journal)
1033 return jbd2_journal_try_to_free_buffers(journal, page,
1034 wait & ~__GFP_WAIT);
1035 return try_to_free_buffers(page);
1036}
1037
ac27a0ec 1038#ifdef CONFIG_QUOTA
af5bc92d 1039#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1040#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1041
617ba13b
MC		 1042static int ext4_write_dquot(struct dquot *dquot);
1043static int ext4_acquire_dquot(struct dquot *dquot);
1044static int ext4_release_dquot(struct dquot *dquot);
1045static int ext4_mark_dquot_dirty(struct dquot *dquot);
1046static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1047static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1048 struct path *path);
7c319d32
AK		 1049static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1050 int format_id);
ca0e05e4 1051static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1052static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1053static int ext4_quota_on_mount(struct super_block *sb, int type);
1054static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1055 size_t len, loff_t off);
617ba13b 1056static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1057 const char *data, size_t len, loff_t off);
7c319d32
AK		 1058static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1059 unsigned int flags);
1060static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1061
96c7e0d9
JK		 1062static struct dquot **ext4_get_dquots(struct inode *inode)
1063{
1064 return EXT4_I(inode)->i_dquot;
1065}
1066
61e225dc 1067static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1068 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1069 .write_dquot = ext4_write_dquot,
1070 .acquire_dquot = ext4_acquire_dquot,
1071 .release_dquot = ext4_release_dquot,
1072 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1073 .write_info = ext4_write_info,
1074 .alloc_dquot = dquot_alloc,
1075 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1076};
1077
0d54b217 1078static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1079 .quota_on = ext4_quota_on,
ca0e05e4 1080 .quota_off = ext4_quota_off,
287a8095
CH		 1081 .quota_sync = dquot_quota_sync,
1082 .get_info = dquot_get_dqinfo,
1083 .set_info = dquot_set_dqinfo,
1084 .get_dqblk = dquot_get_dqblk,
1085 .set_dqblk = dquot_set_dqblk
ac27a0ec 1086};
7c319d32
AK		 1087
1088static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1089 .quota_on_meta = ext4_quota_on_sysfile,
1090 .quota_off = ext4_quota_off_sysfile,
1091 .quota_sync = dquot_quota_sync,
1092 .get_info = dquot_get_dqinfo,
1093 .set_info = dquot_set_dqinfo,
1094 .get_dqblk = dquot_get_dqblk,
1095 .set_dqblk = dquot_set_dqblk
1096};
ac27a0ec
DK		 1097#endif
1098
ee9b6d61 1099static const struct super_operations ext4_sops = {
617ba13b
MC		 1100 .alloc_inode = ext4_alloc_inode,
1101 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1102 .write_inode = ext4_write_inode,
1103 .dirty_inode = ext4_dirty_inode,
7ff9c073 1104 .drop_inode = ext4_drop_inode,
0930fcc1 1105 .evict_inode = ext4_evict_inode,
617ba13b 1106 .put_super = ext4_put_super,
617ba13b 1107 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1108 .freeze_fs = ext4_freeze,
1109 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1110 .statfs = ext4_statfs,
1111 .remount_fs = ext4_remount,
617ba13b 1112 .show_options = ext4_show_options,
ac27a0ec 1113#ifdef CONFIG_QUOTA
617ba13b
MC		 1114 .quota_read = ext4_quota_read,
1115 .quota_write = ext4_quota_write,
96c7e0d9 1116 .get_dquots = ext4_get_dquots,
ac27a0ec 1117#endif
c39a7f84 1118 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1119};
1120
39655164 1121static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1122 .fh_to_dentry = ext4_fh_to_dentry,
1123 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1124 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1125};
1126
1127enum {
1128 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1129 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1130 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1131 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1132 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1133 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1134 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1135 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1136 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1137 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1138 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1139 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1140 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1141 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1142 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1143 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1144 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1145 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1146 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1147};
1148
a447c093 1149static const match_table_t tokens = {
ac27a0ec
DK		 1150 {Opt_bsd_df, "bsddf"},
1151 {Opt_minix_df, "minixdf"},
1152 {Opt_grpid, "grpid"},
1153 {Opt_grpid, "bsdgroups"},
1154 {Opt_nogrpid, "nogrpid"},
1155 {Opt_nogrpid, "sysvgroups"},
1156 {Opt_resgid, "resgid=%u"},
1157 {Opt_resuid, "resuid=%u"},
1158 {Opt_sb, "sb=%u"},
1159 {Opt_err_cont, "errors=continue"},
1160 {Opt_err_panic, "errors=panic"},
1161 {Opt_err_ro, "errors=remount-ro"},
1162 {Opt_nouid32, "nouid32"},
ac27a0ec 1163 {Opt_debug, "debug"},
72578c33
TT		 1164 {Opt_removed, "oldalloc"},
1165 {Opt_removed, "orlov"},
ac27a0ec
DK		 1166 {Opt_user_xattr, "user_xattr"},
1167 {Opt_nouser_xattr, "nouser_xattr"},
1168 {Opt_acl, "acl"},
1169 {Opt_noacl, "noacl"},
e3bb52ae 1170 {Opt_noload, "norecovery"},
5a916be1 1171 {Opt_noload, "noload"},
72578c33
TT		 1172 {Opt_removed, "nobh"},
1173 {Opt_removed, "bh"},
ac27a0ec 1174 {Opt_commit, "commit=%u"},
30773840
TT		 1175 {Opt_min_batch_time, "min_batch_time=%u"},
1176 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1177 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1178 {Opt_journal_path, "journal_path=%s"},
818d276c 1179 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1180 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1181 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1182 {Opt_abort, "abort"},
1183 {Opt_data_journal, "data=journal"},
1184 {Opt_data_ordered, "data=ordered"},
1185 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1186 {Opt_data_err_abort, "data_err=abort"},
1187 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1188 {Opt_offusrjquota, "usrjquota="},
1189 {Opt_usrjquota, "usrjquota=%s"},
1190 {Opt_offgrpjquota, "grpjquota="},
1191 {Opt_grpjquota, "grpjquota=%s"},
1192 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1193 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1194 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1195 {Opt_grpquota, "grpquota"},
1196 {Opt_noquota, "noquota"},
1197 {Opt_quota, "quota"},
1198 {Opt_usrquota, "usrquota"},
1199 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1200 {Opt_barrier, "barrier"},
1201 {Opt_nobarrier, "nobarrier"},
25ec56b5 1202 {Opt_i_version, "i_version"},
c9de560d 1203 {Opt_stripe, "stripe=%u"},
64769240 1204 {Opt_delalloc, "delalloc"},
dd919b98 1205 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1206 {Opt_removed, "mblk_io_submit"},
1207 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1208 {Opt_block_validity, "block_validity"},
1209 {Opt_noblock_validity, "noblock_validity"},
240799cd 1210 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1211 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1212 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1213 {Opt_auto_da_alloc, "auto_da_alloc"},
1214 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1215 {Opt_dioread_nolock, "dioread_nolock"},
1216 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1217 {Opt_discard, "discard"},
1218 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1219 {Opt_init_itable, "init_itable=%u"},
1220 {Opt_init_itable, "init_itable"},
1221 {Opt_noinit_itable, "noinit_itable"},
df981d03 1222 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1223 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1224 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1225 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1226 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1227 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1228 {Opt_err, NULL},
ac27a0ec
DK		 1229};
1230
617ba13b 1231static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1232{
617ba13b 1233 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1234 char *options = (char *) *data;
1235
1236 if (!options || strncmp(options, "sb=", 3) != 0)
1237 return 1; /* Default location */
0b8e58a1 1238
ac27a0ec 1239 options += 3;
0b8e58a1 1240 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1241 sb_block = simple_strtoul(options, &options, 0);
1242 if (*options && *options != ',') {
4776004f 1243 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1244 (char *) *data);
1245 return 1;
1246 }
1247 if (*options == ',')
1248 options++;
1249 *data = (void *) options;
0b8e58a1 1250
ac27a0ec
DK		 1251 return sb_block;
1252}
1253
b3881f74 1254#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1255static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1256 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1257
56c50f11
DM		 1258#ifdef CONFIG_QUOTA
1259static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1260{
1261 struct ext4_sb_info *sbi = EXT4_SB(sb);
1262 char *qname;
03dafb5f 1263 int ret = -1;
56c50f11
DM		 1264
1265 if (sb_any_quota_loaded(sb) &&
1266 !sbi->s_qf_names[qtype]) {
1267 ext4_msg(sb, KERN_ERR,
1268 "Cannot change journaled "
1269 "quota options when quota turned on");
57f73c2c 1270 return -1;
56c50f11 1271 }
262b4662
JK		 1272 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1273 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1274 "when QUOTA feature is enabled");
1275 return -1;
1276 }
56c50f11
DM		 1277 qname = match_strdup(args);
1278 if (!qname) {
1279 ext4_msg(sb, KERN_ERR,
1280 "Not enough memory for storing quotafile name");
57f73c2c 1281 return -1;
56c50f11 1282 }
03dafb5f
CG		 1283 if (sbi->s_qf_names[qtype]) {
1284 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1285 ret = 1;
1286 else
1287 ext4_msg(sb, KERN_ERR,
1288 "%s quota file already specified",
1289 QTYPE2NAME(qtype));
1290 goto errout;
56c50f11 1291 }
03dafb5f 1292 if (strchr(qname, '/')) {
56c50f11
DM		 1293 ext4_msg(sb, KERN_ERR,
1294 "quotafile must be on filesystem root");
03dafb5f 1295 goto errout;
56c50f11 1296 }
03dafb5f 1297 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1298 set_opt(sb, QUOTA);
56c50f11 1299 return 1;
03dafb5f
CG		 1300errout:
1301 kfree(qname);
1302 return ret;
56c50f11
DM		 1303}
1304
1305static int clear_qf_name(struct super_block *sb, int qtype)
1306{
1307
1308 struct ext4_sb_info *sbi = EXT4_SB(sb);
1309
1310 if (sb_any_quota_loaded(sb) &&
1311 sbi->s_qf_names[qtype]) {
1312 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1313 " when quota turned on");
57f73c2c 1314 return -1;
56c50f11 1315 }
03dafb5f 1316 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1317 sbi->s_qf_names[qtype] = NULL;
1318 return 1;
1319}
1320#endif
1321
26092bf5
TT		 1322#define MOPT_SET 0x0001
1323#define MOPT_CLEAR 0x0002
1324#define MOPT_NOSUPPORT 0x0004
1325#define MOPT_EXPLICIT 0x0008
1326#define MOPT_CLEAR_ERR 0x0010
1327#define MOPT_GTE0 0x0020
ac27a0ec 1328#ifdef CONFIG_QUOTA
26092bf5
TT		 1329#define MOPT_Q 0
1330#define MOPT_QFMT 0x0040
1331#else
1332#define MOPT_Q MOPT_NOSUPPORT
1333#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1334#endif
26092bf5 1335#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1336#define MOPT_NO_EXT2 0x0100
1337#define MOPT_NO_EXT3 0x0200
1338#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1339#define MOPT_STRING 0x0400
26092bf5
TT		 1340
1341static const struct mount_opts {
1342 int token;
1343 int mount_opt;
1344 int flags;
1345} ext4_mount_opts[] = {
1346 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1347 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1348 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1349 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1350 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1351 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1352 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1353 MOPT_EXT4_ONLY | MOPT_SET},
1354 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1355 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1356 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1357 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1358 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1359 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1360 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1361 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1362 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1363 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c
TT		 1364 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1365 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1366 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1367 EXT4_MOUNT_JOURNAL_CHECKSUM),
1368 MOPT_EXT4_ONLY | MOPT_SET},
1369 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1370 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1371 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1372 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1373 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1374 MOPT_NO_EXT2 | MOPT_SET},
1375 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1376 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1377 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1378 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1379 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1380 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1381 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1382 {Opt_commit, 0, MOPT_GTE0},
1383 {Opt_max_batch_time, 0, MOPT_GTE0},
1384 {Opt_min_batch_time, 0, MOPT_GTE0},
1385 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1386 {Opt_init_itable, 0, MOPT_GTE0},
1387 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1388 {Opt_resuid, 0, MOPT_GTE0},
1389 {Opt_resgid, 0, MOPT_GTE0},
1390 {Opt_journal_dev, 0, MOPT_GTE0},
ad4eec61 1391 {Opt_journal_path, 0, MOPT_STRING},
0efb3b23 1392 {Opt_journal_ioprio, 0, MOPT_GTE0},
8dc0aa8c
TT		 1393 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1394 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1395 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1396 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1397 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1398 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1399#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1400 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1401 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1402#else
26092bf5
TT		 1403 {Opt_acl, 0, MOPT_NOSUPPORT},
1404 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1405#endif
26092bf5
TT		 1406 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1407 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1408 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1409 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1410 MOPT_SET | MOPT_Q},
1411 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1412 MOPT_SET | MOPT_Q},
1413 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1414 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1415 {Opt_usrjquota, 0, MOPT_Q},
1416 {Opt_grpjquota, 0, MOPT_Q},
1417 {Opt_offusrjquota, 0, MOPT_Q},
1418 {Opt_offgrpjquota, 0, MOPT_Q},
1419 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1420 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1421 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1422 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1423 {Opt_err, 0, 0}
1424};
1425
1426static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1427 substring_t *args, unsigned long *journal_devnum,
1428 unsigned int *journal_ioprio, int is_remount)
1429{
1430 struct ext4_sb_info *sbi = EXT4_SB(sb);
1431 const struct mount_opts *m;
08cefc7a
EB		 1432 kuid_t uid;
1433 kgid_t gid;
26092bf5
TT		 1434 int arg = 0;
1435
57f73c2c
TT		 1436#ifdef CONFIG_QUOTA
1437 if (token == Opt_usrjquota)
1438 return set_qf_name(sb, USRQUOTA, &args[0]);
1439 else if (token == Opt_grpjquota)
1440 return set_qf_name(sb, GRPQUOTA, &args[0]);
1441 else if (token == Opt_offusrjquota)
1442 return clear_qf_name(sb, USRQUOTA);
1443 else if (token == Opt_offgrpjquota)
1444 return clear_qf_name(sb, GRPQUOTA);
1445#endif
26092bf5 1446 switch (token) {
f7048605
TT		 1447 case Opt_noacl:
1448 case Opt_nouser_xattr:
1449 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1450 break;
26092bf5
TT		 1451 case Opt_sb:
1452 return 1; /* handled by get_sb_block() */
1453 case Opt_removed:
5f3633e3 1454 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1455 return 1;
26092bf5
TT		 1456 case Opt_abort:
1457 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1458 return 1;
1459 case Opt_i_version:
1460 sb->s_flags |= MS_I_VERSION;
1461 return 1;
26092bf5
TT		 1462 }
1463
5f3633e3
JK		 1464 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1465 if (token == m->token)
1466 break;
1467
1468 if (m->token == Opt_err) {
1469 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1470 "or missing value", opt);
1471 return -1;
1472 }
1473
8dc0aa8c
TT		 1474 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1475 ext4_msg(sb, KERN_ERR,
1476 "Mount option \"%s\" incompatible with ext2", opt);
1477 return -1;
1478 }
1479 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1480 ext4_msg(sb, KERN_ERR,
1481 "Mount option \"%s\" incompatible with ext3", opt);
1482 return -1;
1483 }
1484
ad4eec61 1485 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1486 return -1;
1487 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1488 return -1;
1489 if (m->flags & MOPT_EXPLICIT)
1490 set_opt2(sb, EXPLICIT_DELALLOC);
1491 if (m->flags & MOPT_CLEAR_ERR)
1492 clear_opt(sb, ERRORS_MASK);
1493 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1494 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1495 "options when quota turned on");
1496 return -1;
1497 }
1498
1499 if (m->flags & MOPT_NOSUPPORT) {
1500 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1501 } else if (token == Opt_commit) {
1502 if (arg == 0)
1503 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1504 sbi->s_commit_interval = HZ * arg;
1505 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1506 sbi->s_max_batch_time = arg;
1507 } else if (token == Opt_min_batch_time) {
1508 sbi->s_min_batch_time = arg;
1509 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1510 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1511 ext4_msg(sb, KERN_ERR,
1512 "EXT4-fs: inode_readahead_blks must be "
1513 "0 or a power of 2 smaller than 2^31");
26092bf5 1514 return -1;
5f3633e3
JK		 1515 }
1516 sbi->s_inode_readahead_blks = arg;
1517 } else if (token == Opt_init_itable) {
1518 set_opt(sb, INIT_INODE_TABLE);
1519 if (!args->from)
1520 arg = EXT4_DEF_LI_WAIT_MULT;
1521 sbi->s_li_wait_mult = arg;
1522 } else if (token == Opt_max_dir_size_kb) {
1523 sbi->s_max_dir_size_kb = arg;
1524 } else if (token == Opt_stripe) {
1525 sbi->s_stripe = arg;
1526 } else if (token == Opt_resuid) {
1527 uid = make_kuid(current_user_ns(), arg);
1528 if (!uid_valid(uid)) {
1529 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1530 return -1;
1531 }
5f3633e3
JK		 1532 sbi->s_resuid = uid;
1533 } else if (token == Opt_resgid) {
1534 gid = make_kgid(current_user_ns(), arg);
1535 if (!gid_valid(gid)) {
1536 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1537 return -1;
1538 }
1539 sbi->s_resgid = gid;
1540 } else if (token == Opt_journal_dev) {
1541 if (is_remount) {
1542 ext4_msg(sb, KERN_ERR,
1543 "Cannot specify journal on remount");
1544 return -1;
1545 }
1546 *journal_devnum = arg;
ad4eec61
ES		 1547 } else if (token == Opt_journal_path) {
1548 char *journal_path;
1549 struct inode *journal_inode;
1550 struct path path;
1551 int error;
1552
1553 if (is_remount) {
1554 ext4_msg(sb, KERN_ERR,
1555 "Cannot specify journal on remount");
1556 return -1;
1557 }
1558 journal_path = match_strdup(&args[0]);
1559 if (!journal_path) {
1560 ext4_msg(sb, KERN_ERR, "error: could not dup "
1561 "journal device string");
1562 return -1;
1563 }
1564
1565 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1566 if (error) {
1567 ext4_msg(sb, KERN_ERR, "error: could not find "
1568 "journal device path: error %d", error);
1569 kfree(journal_path);
1570 return -1;
1571 }
1572
1573 journal_inode = path.dentry->d_inode;
1574 if (!S_ISBLK(journal_inode->i_mode)) {
1575 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1576 "is not a block device", journal_path);
1577 path_put(&path);
1578 kfree(journal_path);
1579 return -1;
1580 }
1581
1582 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1583 path_put(&path);
1584 kfree(journal_path);
5f3633e3
JK		 1585 } else if (token == Opt_journal_ioprio) {
1586 if (arg > 7) {
1587 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1588 " (must be 0-7)");
1589 return -1;
1590 }
1591 *journal_ioprio =
1592 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1593 } else if (m->flags & MOPT_DATAJ) {
1594 if (is_remount) {
1595 if (!sbi->s_journal)
1596 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1597 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1598 ext4_msg(sb, KERN_ERR,
26092bf5 1599 "Cannot change data mode on remount");
26092bf5 1600 return -1;
ac27a0ec 1601 }
26092bf5 1602 } else {
5f3633e3
JK		 1603 clear_opt(sb, DATA_FLAGS);
1604 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1605 }
5f3633e3
JK		 1606#ifdef CONFIG_QUOTA
1607 } else if (m->flags & MOPT_QFMT) {
1608 if (sb_any_quota_loaded(sb) &&
1609 sbi->s_jquota_fmt != m->mount_opt) {
1610 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1611 "quota options when quota turned on");
1612 return -1;
1613 }
262b4662
JK		 1614 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1615 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1616 ext4_msg(sb, KERN_ERR,
1617 "Cannot set journaled quota options "
1618 "when QUOTA feature is enabled");
1619 return -1;
1620 }
5f3633e3
JK		 1621 sbi->s_jquota_fmt = m->mount_opt;
1622#endif
1623 } else {
1624 if (!args->from)
1625 arg = 1;
1626 if (m->flags & MOPT_CLEAR)
1627 arg = !arg;
1628 else if (unlikely(!(m->flags & MOPT_SET))) {
1629 ext4_msg(sb, KERN_WARNING,
1630 "buggy handling of option %s", opt);
1631 WARN_ON(1);
1632 return -1;
1633 }
1634 if (arg != 0)
1635 sbi->s_mount_opt |= m->mount_opt;
1636 else
1637 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1638 }
5f3633e3 1639 return 1;
26092bf5
TT		 1640}
1641
1642static int parse_options(char *options, struct super_block *sb,
1643 unsigned long *journal_devnum,
1644 unsigned int *journal_ioprio,
1645 int is_remount)
1646{
1647 struct ext4_sb_info *sbi = EXT4_SB(sb);
1648 char *p;
1649 substring_t args[MAX_OPT_ARGS];
1650 int token;
1651
1652 if (!options)
1653 return 1;
1654
1655 while ((p = strsep(&options, ",")) != NULL) {
1656 if (!*p)
1657 continue;
1658 /*
1659 * Initialize args struct so we know whether arg was
1660 * found; some options take optional arguments.
1661 */
caecd0af 1662 args[0].to = args[0].from = NULL;
26092bf5
TT		 1663 token = match_token(p, tokens, args);
1664 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1665 journal_ioprio, is_remount) < 0)
1666 return 0;
ac27a0ec
DK		 1667 }
1668#ifdef CONFIG_QUOTA
262b4662
JK		 1669 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1670 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1671 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1672 "feature is enabled");
1673 return 0;
1674 }
ac27a0ec 1675 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1676 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1677 clear_opt(sb, USRQUOTA);
ac27a0ec 1678
482a7425 1679 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1680 clear_opt(sb, GRPQUOTA);
ac27a0ec 1681
56c50f11 1682 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1683 ext4_msg(sb, KERN_ERR, "old and new quota "
1684 "format mixing");
ac27a0ec
DK		 1685 return 0;
1686 }
1687
1688 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1689 ext4_msg(sb, KERN_ERR, "journaled quota format "
1690 "not specified");
ac27a0ec
DK		 1691 return 0;
1692 }
ac27a0ec
DK		 1693 }
1694#endif
261cb20c
JK		 1695 if (test_opt(sb, DIOREAD_NOLOCK)) {
1696 int blocksize =
1697 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1698
1699 if (blocksize < PAGE_CACHE_SIZE) {
1700 ext4_msg(sb, KERN_ERR, "can't mount with "
1701 "dioread_nolock if block size != PAGE_SIZE");
1702 return 0;
1703 }
1704 }
d4f76107
JK		 1705 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1706 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1707 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1708 "in data=ordered mode");
1709 return 0;
1710 }
ac27a0ec
DK		 1711 return 1;
1712}
1713
2adf6da8
TT		 1714static inline void ext4_show_quota_options(struct seq_file *seq,
1715 struct super_block *sb)
1716{
1717#if defined(CONFIG_QUOTA)
1718 struct ext4_sb_info *sbi = EXT4_SB(sb);
1719
1720 if (sbi->s_jquota_fmt) {
1721 char *fmtname = "";
1722
1723 switch (sbi->s_jquota_fmt) {
1724 case QFMT_VFS_OLD:
1725 fmtname = "vfsold";
1726 break;
1727 case QFMT_VFS_V0:
1728 fmtname = "vfsv0";
1729 break;
1730 case QFMT_VFS_V1:
1731 fmtname = "vfsv1";
1732 break;
1733 }
1734 seq_printf(seq, ",jqfmt=%s", fmtname);
1735 }
1736
1737 if (sbi->s_qf_names[USRQUOTA])
1738 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1739
1740 if (sbi->s_qf_names[GRPQUOTA])
1741 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1742#endif
1743}
1744
5a916be1
TT		 1745static const char *token2str(int token)
1746{
50df9fd5 1747 const struct match_token *t;
5a916be1
TT		 1748
1749 for (t = tokens; t->token != Opt_err; t++)
1750 if (t->token == token && !strchr(t->pattern, '='))
1751 break;
1752 return t->pattern;
1753}
1754
2adf6da8
TT		 1755/*
1756 * Show an option if
1757 * - it's set to a non-default value OR
1758 * - if the per-sb default is different from the global default
1759 */
66acdcf4
TT		 1760static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1761 int nodefs)
2adf6da8 1762{
2adf6da8
TT		 1763 struct ext4_sb_info *sbi = EXT4_SB(sb);
1764 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1765 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1766 const struct mount_opts *m;
66acdcf4 1767 char sep = nodefs ? '\n' : ',';
2adf6da8 1768
66acdcf4
TT		 1769#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1770#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1771
1772 if (sbi->s_sb_block != 1)
5a916be1
TT		 1773 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1774
1775 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1776 int want_set = m->flags & MOPT_SET;
1777 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1778 (m->flags & MOPT_CLEAR_ERR))
1779 continue;
66acdcf4 1780 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1781 continue; /* skip if same as the default */
1782 if ((want_set &&
1783 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1784 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1785 continue; /* select Opt_noFoo vs Opt_Foo */
1786 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1787 }
5a916be1 1788
08cefc7a 1789 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1790 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1791 SEQ_OPTS_PRINT("resuid=%u",
1792 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1793 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1794 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1795 SEQ_OPTS_PRINT("resgid=%u",
1796 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1797 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1798 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1799 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1800 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1801 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1802 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1803 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1804 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1805 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1806 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1807 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1808 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1809 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1810 if (sb->s_flags & MS_I_VERSION)
5a916be1 1811 SEQ_OPTS_PUTS("i_version");
66acdcf4 1812 if (nodefs || sbi->s_stripe)
5a916be1 1813 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1814 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1815 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1816 SEQ_OPTS_PUTS("data=journal");
1817 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1818 SEQ_OPTS_PUTS("data=ordered");
1819 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1820 SEQ_OPTS_PUTS("data=writeback");
1821 }
66acdcf4
TT		 1822 if (nodefs ||
1823 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1824 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1825 sbi->s_inode_readahead_blks);
2adf6da8 1826
66acdcf4
TT		 1827 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1828 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1829 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1830 if (nodefs || sbi->s_max_dir_size_kb)
1831 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1832
1833 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1834 return 0;
1835}
1836
66acdcf4
TT		 1837static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1838{
1839 return _ext4_show_options(seq, root->d_sb, 0);
1840}
1841
1842static int options_seq_show(struct seq_file *seq, void *offset)
1843{
1844 struct super_block *sb = seq->private;
1845 int rc;
1846
1847 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1848 rc = _ext4_show_options(seq, sb, 1);
1849 seq_puts(seq, "\n");
1850 return rc;
1851}
1852
1853static int options_open_fs(struct inode *inode, struct file *file)
1854{
d9dda78b 1855 return single_open(file, options_seq_show, PDE_DATA(inode));
66acdcf4
TT		 1856}
1857
1858static const struct file_operations ext4_seq_options_fops = {
1859 .owner = THIS_MODULE,
1860 .open = options_open_fs,
1861 .read = seq_read,
1862 .llseek = seq_lseek,
1863 .release = single_release,
1864};
1865
617ba13b 1866static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1867 int read_only)
1868{
617ba13b 1869 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1870 int res = 0;
1871
617ba13b 1872 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1873 ext4_msg(sb, KERN_ERR, "revision level too high, "
1874 "forcing read-only mode");
ac27a0ec
DK		 1875 res = MS_RDONLY;
1876 }
1877 if (read_only)
281b5995 1878 goto done;
617ba13b 1879 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1880 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1881 "running e2fsck is recommended");
c8b459f4 1882 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 1883 ext4_msg(sb, KERN_WARNING,
1884 "warning: mounting fs with errors, "
1885 "running e2fsck is recommended");
ed3ce80a 1886 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1887 le16_to_cpu(es->s_mnt_count) >=
1888 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1889 ext4_msg(sb, KERN_WARNING,
1890 "warning: maximal mount count reached, "
1891 "running e2fsck is recommended");
ac27a0ec
DK		 1892 else if (le32_to_cpu(es->s_checkinterval) &&
1893 (le32_to_cpu(es->s_lastcheck) +
1894 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1895 ext4_msg(sb, KERN_WARNING,
1896 "warning: checktime reached, "
1897 "running e2fsck is recommended");
0b8e58a1 1898 if (!sbi->s_journal)
0390131b 1899 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1900 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1901 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1902 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1903 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1904 ext4_update_dynamic_rev(sb);
0390131b
FM		 1905 if (sbi->s_journal)
1906 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1907
e2d67052 1908 ext4_commit_super(sb, 1);
281b5995 1909done:
ac27a0ec 1910 if (test_opt(sb, DEBUG))
a9df9a49 1911 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1912 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1913 sb->s_blocksize,
1914 sbi->s_groups_count,
617ba13b
MC		 1915 EXT4_BLOCKS_PER_GROUP(sb),
1916 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1917 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1918
7abc52c2 1919 cleancache_init_fs(sb);
ac27a0ec
DK		 1920 return res;
1921}
1922
117fff10
TT		 1923int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1924{
1925 struct ext4_sb_info *sbi = EXT4_SB(sb);
1926 struct flex_groups *new_groups;
1927 int size;
1928
1929 if (!sbi->s_log_groups_per_flex)
1930 return 0;
1931
1932 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1933 if (size <= sbi->s_flex_groups_allocated)
1934 return 0;
1935
1936 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1937 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1938 if (!new_groups) {
1939 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1940 size / (int) sizeof(struct flex_groups));
1941 return -ENOMEM;
1942 }
1943
1944 if (sbi->s_flex_groups) {
1945 memcpy(new_groups, sbi->s_flex_groups,
1946 (sbi->s_flex_groups_allocated *
1947 sizeof(struct flex_groups)));
b93b41d4 1948 kvfree(sbi->s_flex_groups);
117fff10
TT		 1949 }
1950 sbi->s_flex_groups = new_groups;
1951 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1952 return 0;
1953}
1954
772cb7c8
JS		 1955static int ext4_fill_flex_info(struct super_block *sb)
1956{
1957 struct ext4_sb_info *sbi = EXT4_SB(sb);
1958 struct ext4_group_desc *gdp = NULL;
772cb7c8 1959 ext4_group_t flex_group;
117fff10 1960 int i, err;
772cb7c8 1961
503358ae 1962 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1963 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1964 sbi->s_log_groups_per_flex = 0;
1965 return 1;
1966 }
1967
117fff10
TT		 1968 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1969 if (err)
9933fc0a 1970 goto failed;
772cb7c8 1971
772cb7c8 1972 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1973 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1974
1975 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1976 atomic_add(ext4_free_inodes_count(sb, gdp),
1977 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 1978 atomic64_add(ext4_free_group_clusters(sb, gdp),
1979 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1980 atomic_add(ext4_used_dirs_count(sb, gdp),
1981 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1982 }
1983
1984 return 1;
1985failed:
1986 return 0;
1987}
1988
feb0ab32
DW		 1989static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1990 struct ext4_group_desc *gdp)
717d50e4 1991{
feb0ab32 1992 int offset;
717d50e4 1993 __u16 crc = 0;
feb0ab32 1994 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1995
9aa5d32b 1996 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 1997 /* Use new metadata_csum algorithm */
d6a77105 1998 __le16 save_csum;
feb0ab32
DW		 1999 __u32 csum32;
2000
d6a77105 2001 save_csum = gdp->bg_checksum;
feb0ab32
DW		 2002 gdp->bg_checksum = 0;
2003 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2004 sizeof(le_group));
2005 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2006 sbi->s_desc_size);
d6a77105 2007 gdp->bg_checksum = save_csum;
feb0ab32
DW		 2008
2009 crc = csum32 & 0xFFFF;
2010 goto out;
717d50e4
AD		 2011 }
2012
feb0ab32 2013 /* old crc16 code */
813d32f9
DW		 2014 if (!(sbi->s_es->s_feature_ro_compat &
2015 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2016 return 0;
2017
feb0ab32
DW		 2018 offset = offsetof(struct ext4_group_desc, bg_checksum);
2019
2020 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2021 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2022 crc = crc16(crc, (__u8 *)gdp, offset);
2023 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2024 /* for checksum of struct ext4_group_desc do the rest...*/
2025 if ((sbi->s_es->s_feature_incompat &
2026 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2027 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2028 crc = crc16(crc, (__u8 *)gdp + offset,
2029 le16_to_cpu(sbi->s_es->s_desc_size) -
2030 offset);
2031
2032out:
717d50e4
AD		 2033 return cpu_to_le16(crc);
2034}
2035
feb0ab32 2036int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2037 struct ext4_group_desc *gdp)
2038{
feb0ab32
DW		 2039 if (ext4_has_group_desc_csum(sb) &&
2040 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2041 block_group, gdp)))
717d50e4
AD		 2042 return 0;
2043
2044 return 1;
2045}
2046
feb0ab32
DW		 2047void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2048 struct ext4_group_desc *gdp)
2049{
2050 if (!ext4_has_group_desc_csum(sb))
2051 return;
2052 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2053}
2054
ac27a0ec 2055/* Called at mount-time, super-block is locked */
bfff6873
LC		 2056static int ext4_check_descriptors(struct super_block *sb,
2057 ext4_group_t *first_not_zeroed)
ac27a0ec 2058{
617ba13b
MC		 2059 struct ext4_sb_info *sbi = EXT4_SB(sb);
2060 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2061 ext4_fsblk_t last_block;
bd81d8ee
LV		 2062 ext4_fsblk_t block_bitmap;
2063 ext4_fsblk_t inode_bitmap;
2064 ext4_fsblk_t inode_table;
ce421581 2065 int flexbg_flag = 0;
bfff6873 2066 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2067
ce421581
JS		 2068 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2069 flexbg_flag = 1;
2070
af5bc92d 2071 ext4_debug("Checking group descriptors");
ac27a0ec 2072
197cd65a
AM		 2073 for (i = 0; i < sbi->s_groups_count; i++) {
2074 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2075
ce421581 2076 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2077 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2078 else
2079 last_block = first_block +
617ba13b 2080 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2081
bfff6873
LC		 2082 if ((grp == sbi->s_groups_count) &&
2083 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2084 grp = i;
2085
8fadc143 2086 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2087 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2088 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2089 "Block bitmap for group %u not in group "
b31e1552 2090 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2091 return 0;
2092 }
8fadc143 2093 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2094 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2095 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2096 "Inode bitmap for group %u not in group "
b31e1552 2097 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2098 return 0;
2099 }
8fadc143 2100 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2101 if (inode_table < first_block ||
2b2d6d01 2102 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2103 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2104 "Inode table for group %u not in group "
b31e1552 2105 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2106 return 0;
2107 }
955ce5f5 2108 ext4_lock_group(sb, i);
feb0ab32 2109 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2110 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2111 "Checksum for group %u failed (%u!=%u)",
2112 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2113 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2114 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2115 ext4_unlock_group(sb, i);
8a266467 2116 return 0;
7ee1ec4c 2117 }
717d50e4 2118 }
955ce5f5 2119 ext4_unlock_group(sb, i);
ce421581
JS		 2120 if (!flexbg_flag)
2121 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2122 }
bfff6873
LC		 2123 if (NULL != first_not_zeroed)
2124 *first_not_zeroed = grp;
ac27a0ec
DK		 2125 return 1;
2126}
2127
617ba13b 2128/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2129 * the superblock) which were deleted from all directories, but held open by
2130 * a process at the time of a crash. We walk the list and try to delete these
2131 * inodes at recovery time (only with a read-write filesystem).
2132 *
2133 * In order to keep the orphan inode chain consistent during traversal (in
2134 * case of crash during recovery), we link each inode into the superblock
2135 * orphan list_head and handle it the same way as an inode deletion during
2136 * normal operation (which journals the operations for us).
2137 *
2138 * We only do an iget() and an iput() on each inode, which is very safe if we
2139 * accidentally point at an in-use or already deleted inode. The worst that
2140 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2141 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2142 * e2fsck was run on this filesystem, and it must have already done the orphan
2143 * inode cleanup for us, so we can safely abort without any further action.
2144 */
2b2d6d01
TT		 2145static void ext4_orphan_cleanup(struct super_block *sb,
2146 struct ext4_super_block *es)
ac27a0ec
DK		 2147{
2148 unsigned int s_flags = sb->s_flags;
2149 int nr_orphans = 0, nr_truncates = 0;
2150#ifdef CONFIG_QUOTA
2151 int i;
2152#endif
2153 if (!es->s_last_orphan) {
2154 jbd_debug(4, "no orphan inodes to clean up\n");
2155 return;
2156 }
2157
a8f48a95 2158 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2159 ext4_msg(sb, KERN_ERR, "write access "
2160 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2161 return;
2162 }
2163
d39195c3
AG		 2164 /* Check if feature set would not allow a r/w mount */
2165 if (!ext4_feature_set_ok(sb, 0)) {
2166 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2167 "unknown ROCOMPAT features");
2168 return;
2169 }
2170
617ba13b 2171 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2172 /* don't clear list on RO mount w/ errors */
2173 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2174 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2175 "clearing orphan list.\n");
c25f9bc6
ES		 2176 es->s_last_orphan = 0;
2177 }
ac27a0ec
DK		 2178 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2179 return;
2180 }
2181
2182 if (s_flags & MS_RDONLY) {
b31e1552 2183 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2184 sb->s_flags &= ~MS_RDONLY;
2185 }
2186#ifdef CONFIG_QUOTA
2187 /* Needed for iput() to work correctly and not trash data */
2188 sb->s_flags |= MS_ACTIVE;
2189 /* Turn on quotas so that they are updated correctly */
a2d4a646 2190 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2191 if (EXT4_SB(sb)->s_qf_names[i]) {
2192 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2193 if (ret < 0)
b31e1552
ES		 2194 ext4_msg(sb, KERN_ERR,
2195 "Cannot turn on journaled "
2196 "quota: error %d", ret);
ac27a0ec
DK		 2197 }
2198 }
2199#endif
2200
2201 while (es->s_last_orphan) {
2202 struct inode *inode;
2203
97bd42b9
JB		 2204 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2205 if (IS_ERR(inode)) {
ac27a0ec
DK		 2206 es->s_last_orphan = 0;
2207 break;
2208 }
2209
617ba13b 2210 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2211 dquot_initialize(inode);
ac27a0ec 2212 if (inode->i_nlink) {
566370a2
PT		 2213 if (test_opt(sb, DEBUG))
2214 ext4_msg(sb, KERN_DEBUG,
2215 "%s: truncating inode %lu to %lld bytes",
2216 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2217 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2218 inode->i_ino, inode->i_size);
721e3eba 2219 mutex_lock(&inode->i_mutex);
55f252c9 2220 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2221 ext4_truncate(inode);
721e3eba 2222 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2223 nr_truncates++;
2224 } else {
566370a2
PT		 2225 if (test_opt(sb, DEBUG))
2226 ext4_msg(sb, KERN_DEBUG,
2227 "%s: deleting unreferenced inode %lu",
2228 __func__, inode->i_ino);
ac27a0ec
DK		 2229 jbd_debug(2, "deleting unreferenced inode %lu\n",
2230 inode->i_ino);
2231 nr_orphans++;
2232 }
2233 iput(inode); /* The delete magic happens here! */
2234 }
2235
2b2d6d01 2236#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2237
2238 if (nr_orphans)
b31e1552
ES		 2239 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2240 PLURAL(nr_orphans));
ac27a0ec 2241 if (nr_truncates)
b31e1552
ES		 2242 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2243 PLURAL(nr_truncates));
ac27a0ec
DK		 2244#ifdef CONFIG_QUOTA
2245 /* Turn quotas off */
a2d4a646 2246 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2247 if (sb_dqopt(sb)->files[i])
287a8095 2248 dquot_quota_off(sb, i);
ac27a0ec
DK		 2249 }
2250#endif
2251 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2252}
0b8e58a1 2253
cd2291a4
ES		 2254/*
2255 * Maximal extent format file size.
2256 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2257 * extent format containers, within a sector_t, and within i_blocks
2258 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2259 * so that won't be a limiting factor.
2260 *
f17722f9
LC		 2261 * However there is other limiting factor. We do store extents in the form
2262 * of starting block and length, hence the resulting length of the extent
2263 * covering maximum file size must fit into on-disk format containers as
2264 * well. Given that length is always by 1 unit bigger than max unit (because
2265 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2266 *
cd2291a4
ES		 2267 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2268 */
f287a1a5 2269static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2270{
2271 loff_t res;
2272 loff_t upper_limit = MAX_LFS_FILESIZE;
2273
2274 /* small i_blocks in vfs inode? */
f287a1a5 2275 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2276 /*
90c699a9 2277 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2278 * i_block represent total blocks in 512 bytes
2279 * 32 == size of vfs inode i_blocks * 8
2280 */
2281 upper_limit = (1LL << 32) - 1;
2282
2283 /* total blocks in file system block size */
2284 upper_limit >>= (blkbits - 9);
2285 upper_limit <<= blkbits;
2286 }
2287
f17722f9
LC		 2288 /*
2289 * 32-bit extent-start container, ee_block. We lower the maxbytes
2290 * by one fs block, so ee_len can cover the extent of maximum file
2291 * size
2292 */
2293 res = (1LL << 32) - 1;
cd2291a4 2294 res <<= blkbits;
cd2291a4
ES		 2295
2296 /* Sanity check against vm- & vfs- imposed limits */
2297 if (res > upper_limit)
2298 res = upper_limit;
2299
2300 return res;
2301}
ac27a0ec 2302
ac27a0ec 2303/*
cd2291a4 2304 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2305 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2306 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2307 */
f287a1a5 2308static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2309{
617ba13b 2310 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2311 int meta_blocks;
2312 loff_t upper_limit;
0b8e58a1
AD		 2313 /* This is calculated to be the largest file size for a dense, block
2314 * mapped file such that the file's total number of 512-byte sectors,
2315 * including data and all indirect blocks, does not exceed (2^48 - 1).
2316 *
2317 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2318 * number of 512-byte sectors of the file.
0fc1b451
AK		 2319 */
2320
f287a1a5 2321 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2322 /*
90c699a9 2323 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2324 * the inode i_block field represents total file blocks in
2325 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2326 */
2327 upper_limit = (1LL << 32) - 1;
2328
2329 /* total blocks in file system block size */
2330 upper_limit >>= (bits - 9);
2331
2332 } else {
8180a562
AK		 2333 /*
2334 * We use 48 bit ext4_inode i_blocks
2335 * With EXT4_HUGE_FILE_FL set the i_blocks
2336 * represent total number of blocks in
2337 * file system block size
2338 */
0fc1b451
AK		 2339 upper_limit = (1LL << 48) - 1;
2340
0fc1b451
AK		 2341 }
2342
2343 /* indirect blocks */
2344 meta_blocks = 1;
2345 /* double indirect blocks */
2346 meta_blocks += 1 + (1LL << (bits-2));
2347 /* tripple indirect blocks */
2348 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2349
2350 upper_limit -= meta_blocks;
2351 upper_limit <<= bits;
ac27a0ec
DK		 2352
2353 res += 1LL << (bits-2);
2354 res += 1LL << (2*(bits-2));
2355 res += 1LL << (3*(bits-2));
2356 res <<= bits;
2357 if (res > upper_limit)
2358 res = upper_limit;
0fc1b451
AK		 2359
2360 if (res > MAX_LFS_FILESIZE)
2361 res = MAX_LFS_FILESIZE;
2362
ac27a0ec
DK		 2363 return res;
2364}
2365
617ba13b 2366static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2367 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2368{
617ba13b 2369 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2370 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2371 int has_super = 0;
2372
2373 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2374
617ba13b 2375 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2376 nr < first_meta_bg)
70bbb3e0 2377 return logical_sb_block + nr + 1;
ac27a0ec 2378 bg = sbi->s_desc_per_block * nr;
617ba13b 2379 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2380 has_super = 1;
0b8e58a1 2381
bd63f6b0
DW		 2382 /*
2383 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2384 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2385 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2386 * compensate.
2387 */
2388 if (sb->s_blocksize == 1024 && nr == 0 &&
2389 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2390 has_super++;
2391
617ba13b 2392 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2393}
2394
c9de560d
AT		 2395/**
2396 * ext4_get_stripe_size: Get the stripe size.
2397 * @sbi: In memory super block info
2398 *
2399 * If we have specified it via mount option, then
2400 * use the mount option value. If the value specified at mount time is
2401 * greater than the blocks per group use the super block value.
2402 * If the super block value is greater than blocks per group return 0.
2403 * Allocator needs it be less than blocks per group.
2404 *
2405 */
2406static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2407{
2408 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2409 unsigned long stripe_width =
2410 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2411 int ret;
c9de560d
AT		 2412
2413 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2414 ret = sbi->s_stripe;
2415 else if (stripe_width <= sbi->s_blocks_per_group)
2416 ret = stripe_width;
2417 else if (stride <= sbi->s_blocks_per_group)
2418 ret = stride;
2419 else
2420 ret = 0;
c9de560d 2421
3eb08658
DE		 2422 /*
2423 * If the stripe width is 1, this makes no sense and
2424 * we set it to 0 to turn off stripe handling code.
2425 */
2426 if (ret <= 1)
2427 ret = 0;
c9de560d 2428
3eb08658 2429 return ret;
c9de560d 2430}
ac27a0ec 2431
3197ebdb
TT		 2432/* sysfs supprt */
2433
2434struct ext4_attr {
2435 struct attribute attr;
2436 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2437 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb 2438 const char *, size_t);
f2d50a65
JK		 2439 union {
2440 int offset;
2441 int deprecated_val;
2442 } u;
3197ebdb
TT		 2443};
2444
27dd4385
LC		 2445static int parse_strtoull(const char *buf,
2446 unsigned long long max, unsigned long long *value)
2447{
2448 int ret;
2449
2450 ret = kstrtoull(skip_spaces(buf), 0, value);
2451 if (!ret && *value > max)
2452 ret = -EINVAL;
2453 return ret;
2454}
2455
3197ebdb
TT		 2456static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2457 struct ext4_sb_info *sbi,
2458 char *buf)
2459{
2460 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2461 (s64) EXT4_C2B(sbi,
2462 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2463}
2464
2465static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2466 struct ext4_sb_info *sbi, char *buf)
2467{
2468 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2469
f613dfcb
TT		 2470 if (!sb->s_bdev->bd_part)
2471 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2472 return snprintf(buf, PAGE_SIZE, "%lu\n",
2473 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2474 sbi->s_sectors_written_start) >> 1);
2475}
2476
2477static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2478 struct ext4_sb_info *sbi, char *buf)
2479{
2480 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2481
f613dfcb
TT		 2482 if (!sb->s_bdev->bd_part)
2483 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2484 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2485 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2486 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2487 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2488}
2489
2490static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2491 struct ext4_sb_info *sbi,
2492 const char *buf, size_t count)
2493{
2494 unsigned long t;
e1091b15 2495 int ret;
3197ebdb 2496
e1091b15
LC		 2497 ret = kstrtoul(skip_spaces(buf), 0, &t);
2498 if (ret)
2499 return ret;
3197ebdb 2500
e1091b15 2501 if (t && (!is_power_of_2(t) || t > 0x40000000))
3197ebdb
TT		 2502 return -EINVAL;
2503
2504 sbi->s_inode_readahead_blks = t;
2505 return count;
2506}
2507
2508static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2509 struct ext4_sb_info *sbi, char *buf)
3197ebdb 2510{
f2d50a65 2511 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb
TT		 2512
2513 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2514}
2515
2516static ssize_t sbi_ui_store(struct ext4_attr *a,
2517 struct ext4_sb_info *sbi,
2518 const char *buf, size_t count)
2519{
f2d50a65 2520 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb 2521 unsigned long t;
e1091b15 2522 int ret;
3197ebdb 2523
e1091b15
LC		 2524 ret = kstrtoul(skip_spaces(buf), 0, &t);
2525 if (ret)
2526 return ret;
3197ebdb
TT		 2527 *ui = t;
2528 return count;
2529}
2530
52c198c6
LC		 2531static ssize_t es_ui_show(struct ext4_attr *a,
2532 struct ext4_sb_info *sbi, char *buf)
2533{
2534
2535 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2536 a->u.offset);
2537
2538 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2539}
2540
27dd4385
LC		 2541static ssize_t reserved_clusters_show(struct ext4_attr *a,
2542 struct ext4_sb_info *sbi, char *buf)
2543{
2544 return snprintf(buf, PAGE_SIZE, "%llu\n",
2545 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2546}
2547
2548static ssize_t reserved_clusters_store(struct ext4_attr *a,
2549 struct ext4_sb_info *sbi,
2550 const char *buf, size_t count)
2551{
2552 unsigned long long val;
2553 int ret;
2554
2555 if (parse_strtoull(buf, -1ULL, &val))
2556 return -EINVAL;
2557 ret = ext4_reserve_clusters(sbi, val);
2558
2559 return ret ? ret : count;
2560}
2561
2c0544b2
TT		 2562static ssize_t trigger_test_error(struct ext4_attr *a,
2563 struct ext4_sb_info *sbi,
2564 const char *buf, size_t count)
2565{
2566 int len = count;
2567
2568 if (!capable(CAP_SYS_ADMIN))
2569 return -EPERM;
2570
2571 if (len && buf[len-1] == '\n')
2572 len--;
2573
2574 if (len)
2575 ext4_error(sbi->s_sb, "%.*s", len, buf);
2576 return count;
2577}
2578
f2d50a65
JK		 2579static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2580 struct ext4_sb_info *sbi, char *buf)
2581{
2582 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2583}
2584
3197ebdb
TT		 2585#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2586static struct ext4_attr ext4_attr_##_name = { \
2587 .attr = {.name = __stringify(_name), .mode = _mode }, \
2588 .show = _show, \
2589 .store = _store, \
f2d50a65
JK		 2590 .u = { \
2591 .offset = offsetof(struct ext4_sb_info, _elname),\
2592 }, \
3197ebdb 2593}
52c198c6
LC		 2594
2595#define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2596static struct ext4_attr ext4_attr_##_name = { \
2597 .attr = {.name = __stringify(_name), .mode = _mode }, \
2598 .show = _show, \
2599 .store = _store, \
2600 .u = { \
2601 .offset = offsetof(struct ext4_super_block, _elname), \
2602 }, \
2603}
2604
3197ebdb
TT		 2605#define EXT4_ATTR(name, mode, show, store) \
2606static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2607
857ac889 2608#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2609#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2610#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
52c198c6
LC		 2611
2612#define EXT4_RO_ATTR_ES_UI(name, elname) \
2613 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
3197ebdb
TT		 2614#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2615 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
52c198c6 2616
3197ebdb 2617#define ATTR_LIST(name) &ext4_attr_##name.attr
f2d50a65
JK		 2618#define EXT4_DEPRECATED_ATTR(_name, _val) \
2619static struct ext4_attr ext4_attr_##_name = { \
2620 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2621 .show = sbi_deprecated_show, \
2622 .u = { \
2623 .deprecated_val = _val, \
2624 }, \
2625}
3197ebdb
TT		 2626
2627EXT4_RO_ATTR(delayed_allocation_blocks);
2628EXT4_RO_ATTR(session_write_kbytes);
2629EXT4_RO_ATTR(lifetime_write_kbytes);
27dd4385 2630EXT4_RW_ATTR(reserved_clusters);
3197ebdb
TT		 2631EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2632 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2633EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2634EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2635EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2636EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2637EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2638EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2639EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
f2d50a65 2640EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
67a5da56 2641EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2642EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
efbed4dc
TT		 2643EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2644EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2645EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2646EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2647EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2648EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
52c198c6
LC		 2649EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2650EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2651EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
3197ebdb
TT		 2652
2653static struct attribute *ext4_attrs[] = {
2654 ATTR_LIST(delayed_allocation_blocks),
2655 ATTR_LIST(session_write_kbytes),
2656 ATTR_LIST(lifetime_write_kbytes),
27dd4385 2657 ATTR_LIST(reserved_clusters),
3197ebdb 2658 ATTR_LIST(inode_readahead_blks),
11013911 2659 ATTR_LIST(inode_goal),
3197ebdb
TT		 2660 ATTR_LIST(mb_stats),
2661 ATTR_LIST(mb_max_to_scan),
2662 ATTR_LIST(mb_min_to_scan),
2663 ATTR_LIST(mb_order2_req),
2664 ATTR_LIST(mb_stream_req),
2665 ATTR_LIST(mb_group_prealloc),
55138e0b 2666 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2667 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2668 ATTR_LIST(trigger_fs_error),
efbed4dc
TT		 2669 ATTR_LIST(err_ratelimit_interval_ms),
2670 ATTR_LIST(err_ratelimit_burst),
2671 ATTR_LIST(warning_ratelimit_interval_ms),
2672 ATTR_LIST(warning_ratelimit_burst),
2673 ATTR_LIST(msg_ratelimit_interval_ms),
2674 ATTR_LIST(msg_ratelimit_burst),
52c198c6
LC		 2675 ATTR_LIST(errors_count),
2676 ATTR_LIST(first_error_time),
2677 ATTR_LIST(last_error_time),
3197ebdb
TT		 2678 NULL,
2679};
2680
857ac889
LC		 2681/* Features this copy of ext4 supports */
2682EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2683EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2684EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2685
2686static struct attribute *ext4_feat_attrs[] = {
2687 ATTR_LIST(lazy_itable_init),
27ee40df 2688 ATTR_LIST(batched_discard),
5e7bbef1 2689 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2690 NULL,
2691};
2692
3197ebdb
TT		 2693static ssize_t ext4_attr_show(struct kobject *kobj,
2694 struct attribute *attr, char *buf)
2695{
2696 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2697 s_kobj);
2698 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2699
2700 return a->show ? a->show(a, sbi, buf) : 0;
2701}
2702
2703static ssize_t ext4_attr_store(struct kobject *kobj,
2704 struct attribute *attr,
2705 const char *buf, size_t len)
2706{
2707 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2708 s_kobj);
2709 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2710
2711 return a->store ? a->store(a, sbi, buf, len) : 0;
2712}
2713
2714static void ext4_sb_release(struct kobject *kobj)
2715{
2716 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2717 s_kobj);
2718 complete(&sbi->s_kobj_unregister);
2719}
2720
52cf25d0 2721static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2722 .show = ext4_attr_show,
2723 .store = ext4_attr_store,
2724};
2725
2726static struct kobj_type ext4_ktype = {
2727 .default_attrs = ext4_attrs,
2728 .sysfs_ops = &ext4_attr_ops,
2729 .release = ext4_sb_release,
2730};
2731
857ac889
LC		 2732static void ext4_feat_release(struct kobject *kobj)
2733{
2734 complete(&ext4_feat->f_kobj_unregister);
2735}
2736
c7f72543
LC		 2737static ssize_t ext4_feat_show(struct kobject *kobj,
2738 struct attribute *attr, char *buf)
2739{
2740 return snprintf(buf, PAGE_SIZE, "supported\n");
2741}
2742
2743/*
2744 * We can not use ext4_attr_show/store because it relies on the kobject
2745 * being embedded in the ext4_sb_info structure which is definitely not
2746 * true in this case.
2747 */
2748static const struct sysfs_ops ext4_feat_ops = {
2749 .show = ext4_feat_show,
2750 .store = NULL,
2751};
2752
857ac889
LC		 2753static struct kobj_type ext4_feat_ktype = {
2754 .default_attrs = ext4_feat_attrs,
c7f72543 2755 .sysfs_ops = &ext4_feat_ops,
857ac889
LC		 2756 .release = ext4_feat_release,
2757};
2758
a13fb1a4
ES		 2759/*
2760 * Check whether this filesystem can be mounted based on
2761 * the features present and the RDONLY/RDWR mount requested.
2762 * Returns 1 if this filesystem can be mounted as requested,
2763 * 0 if it cannot be.
2764 */
2765static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2766{
2767 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2768 ext4_msg(sb, KERN_ERR,
2769 "Couldn't mount because of "
2770 "unsupported optional features (%x)",
2771 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2772 ~EXT4_FEATURE_INCOMPAT_SUPP));
2773 return 0;
2774 }
2775
2776 if (readonly)
2777 return 1;
2778
2cb5cc8b
DW		 2779 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2780 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2781 sb->s_flags |= MS_RDONLY;
2782 return 1;
2783 }
2784
a13fb1a4
ES		 2785 /* Check that feature set is OK for a read-write mount */
2786 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2787 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2788 "unsupported optional features (%x)",
2789 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2790 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2791 return 0;
2792 }
2793 /*
2794 * Large file size enabled file system can only be mounted
2795 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2796 */
2797 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2798 if (sizeof(blkcnt_t) < sizeof(u64)) {
2799 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2800 "cannot be mounted RDWR without "
2801 "CONFIG_LBDAF");
2802 return 0;
2803 }
2804 }
bab08ab9
TT		 2805 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2806 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2807 ext4_msg(sb, KERN_ERR,
2808 "Can't support bigalloc feature without "
2809 "extents feature\n");
2810 return 0;
2811 }
7c319d32
AK		 2812
2813#ifndef CONFIG_QUOTA
2814 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2815 !readonly) {
2816 ext4_msg(sb, KERN_ERR,
2817 "Filesystem with quota feature cannot be mounted RDWR "
2818 "without CONFIG_QUOTA");
2819 return 0;
2820 }
2821#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2822 return 1;
2823}
2824
66e61a9e
TT		 2825/*
2826 * This function is called once a day if we have errors logged
2827 * on the file system
2828 */
2829static void print_daily_error_info(unsigned long arg)
2830{
2831 struct super_block *sb = (struct super_block *) arg;
2832 struct ext4_sb_info *sbi;
2833 struct ext4_super_block *es;
2834
2835 sbi = EXT4_SB(sb);
2836 es = sbi->s_es;
2837
2838 if (es->s_error_count)
ae0f78de
TT		 2839 /* fsck newer than v1.41.13 is needed to clean this condition. */
2840 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2841 le32_to_cpu(es->s_error_count));
2842 if (es->s_first_error_time) {
ae0f78de 2843 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2844 sb->s_id, le32_to_cpu(es->s_first_error_time),
2845 (int) sizeof(es->s_first_error_func),
2846 es->s_first_error_func,
2847 le32_to_cpu(es->s_first_error_line));
2848 if (es->s_first_error_ino)
2849 printk(": inode %u",
2850 le32_to_cpu(es->s_first_error_ino));
2851 if (es->s_first_error_block)
2852 printk(": block %llu", (unsigned long long)
2853 le64_to_cpu(es->s_first_error_block));
2854 printk("\n");
2855 }
2856 if (es->s_last_error_time) {
ae0f78de 2857 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2858 sb->s_id, le32_to_cpu(es->s_last_error_time),
2859 (int) sizeof(es->s_last_error_func),
2860 es->s_last_error_func,
2861 le32_to_cpu(es->s_last_error_line));
2862 if (es->s_last_error_ino)
2863 printk(": inode %u",
2864 le32_to_cpu(es->s_last_error_ino));
2865 if (es->s_last_error_block)
2866 printk(": block %llu", (unsigned long long)
2867 le64_to_cpu(es->s_last_error_block));
2868 printk("\n");
2869 }
2870 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2871}
2872
bfff6873
LC		 2873/* Find next suitable group and run ext4_init_inode_table */
2874static int ext4_run_li_request(struct ext4_li_request *elr)
2875{
2876 struct ext4_group_desc *gdp = NULL;
2877 ext4_group_t group, ngroups;
2878 struct super_block *sb;
2879 unsigned long timeout = 0;
2880 int ret = 0;
2881
2882 sb = elr->lr_super;
2883 ngroups = EXT4_SB(sb)->s_groups_count;
2884
8e8ad8a5 2885 sb_start_write(sb);
bfff6873
LC		 2886 for (group = elr->lr_next_group; group < ngroups; group++) {
2887 gdp = ext4_get_group_desc(sb, group, NULL);
2888 if (!gdp) {
2889 ret = 1;
2890 break;
2891 }
2892
2893 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2894 break;
2895 }
2896
7f511862 2897 if (group >= ngroups)
bfff6873
LC		 2898 ret = 1;
2899
2900 if (!ret) {
2901 timeout = jiffies;
2902 ret = ext4_init_inode_table(sb, group,
2903 elr->lr_timeout ? 0 : 1);
2904 if (elr->lr_timeout == 0) {
51ce6511
LC		 2905 timeout = (jiffies - timeout) *
2906 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2907 elr->lr_timeout = timeout;
2908 }
2909 elr->lr_next_sched = jiffies + elr->lr_timeout;
2910 elr->lr_next_group = group + 1;
2911 }
8e8ad8a5 2912 sb_end_write(sb);
bfff6873
LC		 2913
2914 return ret;
2915}
2916
2917/*
2918 * Remove lr_request from the list_request and free the
4ed5c033 2919 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2920 */
2921static void ext4_remove_li_request(struct ext4_li_request *elr)
2922{
2923 struct ext4_sb_info *sbi;
2924
2925 if (!elr)
2926 return;
2927
2928 sbi = elr->lr_sbi;
2929
2930 list_del(&elr->lr_request);
2931 sbi->s_li_request = NULL;
2932 kfree(elr);
2933}
2934
2935static void ext4_unregister_li_request(struct super_block *sb)
2936{
1bb933fb
LC		 2937 mutex_lock(&ext4_li_mtx);
2938 if (!ext4_li_info) {
2939 mutex_unlock(&ext4_li_mtx);
bfff6873 2940 return;
1bb933fb 2941 }
bfff6873
LC		 2942
2943 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2944 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2945 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2946 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2947}
2948
8f1f7453
ES		 2949static struct task_struct *ext4_lazyinit_task;
2950
bfff6873
LC		 2951/*
2952 * This is the function where ext4lazyinit thread lives. It walks
2953 * through the request list searching for next scheduled filesystem.
2954 * When such a fs is found, run the lazy initialization request
2955 * (ext4_rn_li_request) and keep track of the time spend in this
2956 * function. Based on that time we compute next schedule time of
2957 * the request. When walking through the list is complete, compute
2958 * next waking time and put itself into sleep.
2959 */
2960static int ext4_lazyinit_thread(void *arg)
2961{
2962 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2963 struct list_head *pos, *n;
2964 struct ext4_li_request *elr;
4ed5c033 2965 unsigned long next_wakeup, cur;
bfff6873
LC		 2966
2967 BUG_ON(NULL == eli);
2968
bfff6873
LC		 2969cont_thread:
2970 while (true) {
2971 next_wakeup = MAX_JIFFY_OFFSET;
2972
2973 mutex_lock(&eli->li_list_mtx);
2974 if (list_empty(&eli->li_request_list)) {
2975 mutex_unlock(&eli->li_list_mtx);
2976 goto exit_thread;
2977 }
2978
2979 list_for_each_safe(pos, n, &eli->li_request_list) {
2980 elr = list_entry(pos, struct ext4_li_request,
2981 lr_request);
2982
b2c78cd0
TT		 2983 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2984 if (ext4_run_li_request(elr) != 0) {
2985 /* error, remove the lazy_init job */
2986 ext4_remove_li_request(elr);
2987 continue;
2988 }
bfff6873
LC		 2989 }
2990
2991 if (time_before(elr->lr_next_sched, next_wakeup))
2992 next_wakeup = elr->lr_next_sched;
2993 }
2994 mutex_unlock(&eli->li_list_mtx);
2995
a0acae0e 2996 try_to_freeze();
bfff6873 2997
4ed5c033
LC		 2998 cur = jiffies;
2999 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3000 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3001 cond_resched();
3002 continue;
3003 }
3004
4ed5c033
LC		 3005 schedule_timeout_interruptible(next_wakeup - cur);
3006
8f1f7453
ES		 3007 if (kthread_should_stop()) {
3008 ext4_clear_request_list();
3009 goto exit_thread;
3010 }
bfff6873
LC		 3011 }
3012
3013exit_thread:
3014 /*
3015 * It looks like the request list is empty, but we need
3016 * to check it under the li_list_mtx lock, to prevent any
3017 * additions into it, and of course we should lock ext4_li_mtx
3018 * to atomically free the list and ext4_li_info, because at
3019 * this point another ext4 filesystem could be registering
3020 * new one.
3021 */
3022 mutex_lock(&ext4_li_mtx);
3023 mutex_lock(&eli->li_list_mtx);
3024 if (!list_empty(&eli->li_request_list)) {
3025 mutex_unlock(&eli->li_list_mtx);
3026 mutex_unlock(&ext4_li_mtx);
3027 goto cont_thread;
3028 }
3029 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3030 kfree(ext4_li_info);
3031 ext4_li_info = NULL;
3032 mutex_unlock(&ext4_li_mtx);
3033
3034 return 0;
3035}
3036
3037static void ext4_clear_request_list(void)
3038{
3039 struct list_head *pos, *n;
3040 struct ext4_li_request *elr;
3041
3042 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3043 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3044 elr = list_entry(pos, struct ext4_li_request,
3045 lr_request);
3046 ext4_remove_li_request(elr);
3047 }
3048 mutex_unlock(&ext4_li_info->li_list_mtx);
3049}
3050
3051static int ext4_run_lazyinit_thread(void)
3052{
8f1f7453
ES		 3053 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3054 ext4_li_info, "ext4lazyinit");
3055 if (IS_ERR(ext4_lazyinit_task)) {
3056 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3057 ext4_clear_request_list();
bfff6873
LC		 3058 kfree(ext4_li_info);
3059 ext4_li_info = NULL;
92b97816 3060 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3061 "initialization thread\n",
3062 err);
3063 return err;
3064 }
3065 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3066 return 0;
3067}
3068
3069/*
3070 * Check whether it make sense to run itable init. thread or not.
3071 * If there is at least one uninitialized inode table, return
3072 * corresponding group number, else the loop goes through all
3073 * groups and return total number of groups.
3074 */
3075static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3076{
3077 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3078 struct ext4_group_desc *gdp = NULL;
3079
3080 for (group = 0; group < ngroups; group++) {
3081 gdp = ext4_get_group_desc(sb, group, NULL);
3082 if (!gdp)
3083 continue;
3084
3085 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3086 break;
3087 }
3088
3089 return group;
3090}
3091
3092static int ext4_li_info_new(void)
3093{
3094 struct ext4_lazy_init *eli = NULL;
3095
3096 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3097 if (!eli)
3098 return -ENOMEM;
3099
bfff6873
LC		 3100 INIT_LIST_HEAD(&eli->li_request_list);
3101 mutex_init(&eli->li_list_mtx);
3102
bfff6873
LC		 3103 eli->li_state |= EXT4_LAZYINIT_QUIT;
3104
3105 ext4_li_info = eli;
3106
3107 return 0;
3108}
3109
3110static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3111 ext4_group_t start)
3112{
3113 struct ext4_sb_info *sbi = EXT4_SB(sb);
3114 struct ext4_li_request *elr;
bfff6873
LC		 3115
3116 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3117 if (!elr)
3118 return NULL;
3119
3120 elr->lr_super = sb;
3121 elr->lr_sbi = sbi;
3122 elr->lr_next_group = start;
3123
3124 /*
3125 * Randomize first schedule time of the request to
3126 * spread the inode table initialization requests
3127 * better.
3128 */
dd1f723b
TT		 3129 elr->lr_next_sched = jiffies + (prandom_u32() %
3130 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3131 return elr;
3132}
3133
7f511862
TT		 3134int ext4_register_li_request(struct super_block *sb,
3135 ext4_group_t first_not_zeroed)
bfff6873
LC		 3136{
3137 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3138 struct ext4_li_request *elr = NULL;
bfff6873 3139 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3140 int ret = 0;
bfff6873 3141
7f511862 3142 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3143 if (sbi->s_li_request != NULL) {
3144 /*
3145 * Reset timeout so it can be computed again, because
3146 * s_li_wait_mult might have changed.
3147 */
3148 sbi->s_li_request->lr_timeout = 0;
7f511862 3149 goto out;
51ce6511 3150 }
bfff6873
LC		 3151
3152 if (first_not_zeroed == ngroups ||
3153 (sb->s_flags & MS_RDONLY) ||
55ff3840 3154 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3155 goto out;
bfff6873
LC		 3156
3157 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3158 if (!elr) {
3159 ret = -ENOMEM;
3160 goto out;
3161 }
bfff6873
LC		 3162
3163 if (NULL == ext4_li_info) {
3164 ret = ext4_li_info_new();
3165 if (ret)
3166 goto out;
3167 }
3168
3169 mutex_lock(&ext4_li_info->li_list_mtx);
3170 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3171 mutex_unlock(&ext4_li_info->li_list_mtx);
3172
3173 sbi->s_li_request = elr;
46e4690b
TM		 3174 /*
3175 * set elr to NULL here since it has been inserted to
3176 * the request_list and the removal and free of it is
3177 * handled by ext4_clear_request_list from now on.
3178 */
3179 elr = NULL;
bfff6873
LC		 3180
3181 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3182 ret = ext4_run_lazyinit_thread();
3183 if (ret)
3184 goto out;
3185 }
bfff6873 3186out:
beed5ecb
NK		 3187 mutex_unlock(&ext4_li_mtx);
3188 if (ret)
bfff6873 3189 kfree(elr);
bfff6873
LC		 3190 return ret;
3191}
3192
3193/*
3194 * We do not need to lock anything since this is called on
3195 * module unload.
3196 */
3197static void ext4_destroy_lazyinit_thread(void)
3198{
3199 /*
3200 * If thread exited earlier
3201 * there's nothing to be done.
3202 */
8f1f7453 3203 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3204 return;
3205
8f1f7453 3206 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3207}
3208
25ed6e8a
DW		 3209static int set_journal_csum_feature_set(struct super_block *sb)
3210{
3211 int ret = 1;
3212 int compat, incompat;
3213 struct ext4_sb_info *sbi = EXT4_SB(sb);
3214
9aa5d32b 3215 if (ext4_has_metadata_csum(sb)) {
db9ee220 3216 /* journal checksum v3 */
25ed6e8a 3217 compat = 0;
db9ee220 3218 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3219 } else {
3220 /* journal checksum v1 */
3221 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3222 incompat = 0;
3223 }
3224
feb8c6d3
DW		 3225 jbd2_journal_clear_features(sbi->s_journal,
3226 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3227 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3228 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3229 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3230 ret = jbd2_journal_set_features(sbi->s_journal,
3231 compat, 0,
3232 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3233 incompat);
3234 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3235 ret = jbd2_journal_set_features(sbi->s_journal,
3236 compat, 0,
3237 incompat);
3238 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3239 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3240 } else {
feb8c6d3
DW		 3241 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3242 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3243 }
3244
3245 return ret;
3246}
3247
952fc18e
TT		 3248/*
3249 * Note: calculating the overhead so we can be compatible with
3250 * historical BSD practice is quite difficult in the face of
3251 * clusters/bigalloc. This is because multiple metadata blocks from
3252 * different block group can end up in the same allocation cluster.
3253 * Calculating the exact overhead in the face of clustered allocation
3254 * requires either O(all block bitmaps) in memory or O(number of block
3255 * groups**2) in time. We will still calculate the superblock for
3256 * older file systems --- and if we come across with a bigalloc file
3257 * system with zero in s_overhead_clusters the estimate will be close to
3258 * correct especially for very large cluster sizes --- but for newer
3259 * file systems, it's better to calculate this figure once at mkfs
3260 * time, and store it in the superblock. If the superblock value is
3261 * present (even for non-bigalloc file systems), we will use it.
3262 */
3263static int count_overhead(struct super_block *sb, ext4_group_t grp,
3264 char *buf)
3265{
3266 struct ext4_sb_info *sbi = EXT4_SB(sb);
3267 struct ext4_group_desc *gdp;
3268 ext4_fsblk_t first_block, last_block, b;
3269 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3270 int s, j, count = 0;
3271
0548bbb8
TT		 3272 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3273 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3274 sbi->s_itb_per_group + 2);
3275
952fc18e
TT		 3276 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3277 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3278 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3279 for (i = 0; i < ngroups; i++) {
3280 gdp = ext4_get_group_desc(sb, i, NULL);
3281 b = ext4_block_bitmap(sb, gdp);
3282 if (b >= first_block && b <= last_block) {
3283 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3284 count++;
3285 }
3286 b = ext4_inode_bitmap(sb, gdp);
3287 if (b >= first_block && b <= last_block) {
3288 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3289 count++;
3290 }
3291 b = ext4_inode_table(sb, gdp);
3292 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3293 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3294 int c = EXT4_B2C(sbi, b - first_block);
3295 ext4_set_bit(c, buf);
3296 count++;
3297 }
3298 if (i != grp)
3299 continue;
3300 s = 0;
3301 if (ext4_bg_has_super(sb, grp)) {
3302 ext4_set_bit(s++, buf);
3303 count++;
3304 }
3305 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3306 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3307 count++;
3308 }
3309 }
3310 if (!count)
3311 return 0;
3312 return EXT4_CLUSTERS_PER_GROUP(sb) -
3313 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3314}
3315
3316/*
3317 * Compute the overhead and stash it in sbi->s_overhead
3318 */
3319int ext4_calculate_overhead(struct super_block *sb)
3320{
3321 struct ext4_sb_info *sbi = EXT4_SB(sb);
3322 struct ext4_super_block *es = sbi->s_es;
3323 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3324 ext4_fsblk_t overhead = 0;
4fdb5543 3325 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3326
952fc18e
TT		 3327 if (!buf)
3328 return -ENOMEM;
3329
3330 /*
3331 * Compute the overhead (FS structures). This is constant
3332 * for a given filesystem unless the number of block groups
3333 * changes so we cache the previous value until it does.
3334 */
3335
3336 /*
3337 * All of the blocks before first_data_block are overhead
3338 */
3339 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3340
3341 /*
3342 * Add the overhead found in each block group
3343 */
3344 for (i = 0; i < ngroups; i++) {
3345 int blks;
3346
3347 blks = count_overhead(sb, i, buf);
3348 overhead += blks;
3349 if (blks)
3350 memset(buf, 0, PAGE_SIZE);
3351 cond_resched();
3352 }
b003b524
ES		 3353 /* Add the internal journal blocks as well */
3354 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3355 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3356
952fc18e
TT		 3357 sbi->s_overhead = overhead;
3358 smp_wmb();
3359 free_page((unsigned long) buf);
3360 return 0;
3361}
3362
27dd4385 3363
30fac0f7 3364static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
27dd4385
LC		 3365{
3366 ext4_fsblk_t resv_clusters;
3367
30fac0f7
JK		 3368 /*
3369 * There's no need to reserve anything when we aren't using extents.
3370 * The space estimates are exact, there are no unwritten extents,
3371 * hole punching doesn't need new metadata... This is needed especially
3372 * to keep ext2/3 backward compatibility.
3373 */
3374 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3375 return 0;
27dd4385
LC		 3376 /*
3377 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3378 * This should cover the situations where we can not afford to run
3379 * out of space like for example punch hole, or converting
556615dc 3380 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3381 * allocation would require 1, or 2 blocks, higher numbers are
3382 * very rare.
3383 */
30fac0f7
JK		 3384 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3385 EXT4_SB(sb)->s_cluster_bits;
27dd4385
LC		 3386
3387 do_div(resv_clusters, 50);
3388 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3389
3390 return resv_clusters;
3391}
3392
3393
3394static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3395{
3396 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3397 sbi->s_cluster_bits;
3398
3399 if (count >= clusters)
3400 return -EINVAL;
3401
3402 atomic64_set(&sbi->s_resv_clusters, count);
3403 return 0;
3404}
3405
2b2d6d01 3406static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3407{
d4c402d9 3408 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3409 struct buffer_head *bh;
617ba13b
MC		 3410 struct ext4_super_block *es = NULL;
3411 struct ext4_sb_info *sbi;
3412 ext4_fsblk_t block;
3413 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3414 ext4_fsblk_t logical_sb_block;
ac27a0ec 3415 unsigned long offset = 0;
ac27a0ec
DK		 3416 unsigned long journal_devnum = 0;
3417 unsigned long def_mount_opts;
3418 struct inode *root;
9f6200bb 3419 char *cp;
0390131b 3420 const char *descr;
dcc7dae3 3421 int ret = -ENOMEM;
281b5995 3422 int blocksize, clustersize;
4ec11028
TT		 3423 unsigned int db_count;
3424 unsigned int i;
281b5995 3425 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3426 __u64 blocks_count;
07aa2ea1 3427 int err = 0;
b3881f74 3428 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3429 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3430
3431 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3432 if (!sbi)
dcc7dae3 3433 goto out_free_orig;
705895b6
PE		 3434
3435 sbi->s_blockgroup_lock =
3436 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3437 if (!sbi->s_blockgroup_lock) {
3438 kfree(sbi);
dcc7dae3 3439 goto out_free_orig;
705895b6 3440 }
ac27a0ec 3441 sb->s_fs_info = sbi;
2c0544b2 3442 sbi->s_sb = sb;
240799cd 3443 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3444 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3445 if (sb->s_bdev->bd_part)
3446 sbi->s_sectors_written_start =
3447 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3448
9f6200bb
TT		 3449 /* Cleanup superblock name */
3450 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3451 *cp = '!';
3452
07aa2ea1 3453 /* -EINVAL is default */
dcc7dae3 3454 ret = -EINVAL;
617ba13b 3455 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3456 if (!blocksize) {
b31e1552 3457 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3458 goto out_fail;
3459 }
3460
3461 /*
617ba13b 3462 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3463 * block sizes. We need to calculate the offset from buffer start.
3464 */
617ba13b 3465 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3466 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3467 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3468 } else {
70bbb3e0 3469 logical_sb_block = sb_block;
ac27a0ec
DK		 3470 }
3471
a8ac900b 3472 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3473 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3474 goto out_fail;
3475 }
3476 /*
3477 * Note: s_es must be initialized as soon as possible because
617ba13b 3478 * some ext4 macro-instructions depend on its value
ac27a0ec 3479 */
2716b802 3480 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3481 sbi->s_es = es;
3482 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3483 if (sb->s_magic != EXT4_SUPER_MAGIC)
3484 goto cantfind_ext4;
afc32f7e 3485 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3486
feb0ab32
DW		 3487 /* Warn if metadata_csum and gdt_csum are both set. */
3488 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3489 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3490 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
363307e6 3491 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3492 "redundant flags; please run fsck.");
3493
d25425f8
DW		 3494 /* Check for a known checksum algorithm */
3495 if (!ext4_verify_csum_type(sb, es)) {
3496 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3497 "unknown checksum algorithm.");
3498 silent = 1;
3499 goto cantfind_ext4;
3500 }
3501
0441984a
DW		 3502 /* Load the checksum driver */
3503 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3504 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3505 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3506 if (IS_ERR(sbi->s_chksum_driver)) {
3507 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3508 ret = PTR_ERR(sbi->s_chksum_driver);
3509 sbi->s_chksum_driver = NULL;
3510 goto failed_mount;
3511 }
3512 }
3513
a9c47317
DW		 3514 /* Check superblock checksum */
3515 if (!ext4_superblock_csum_verify(sb, es)) {
3516 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3517 "invalid superblock checksum. Run e2fsck?");
3518 silent = 1;
3519 goto cantfind_ext4;
3520 }
3521
3522 /* Precompute checksum seed for all metadata */
9aa5d32b 3523 if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3524 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3525 sizeof(es->s_uuid));
3526
ac27a0ec
DK		 3527 /* Set defaults before we parse the mount options */
3528 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3529 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3530 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3531 set_opt(sb, DEBUG);
87f26807 3532 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3533 set_opt(sb, GRPID);
617ba13b 3534 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3535 set_opt(sb, NO_UID32);
ea663336 3536 /* xattr user namespace & acls are now defaulted on */
ea663336 3537 set_opt(sb, XATTR_USER);
03010a33 3538#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3539 set_opt(sb, POSIX_ACL);
2e7842b8 3540#endif
98c1a759
DW		 3541 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3542 if (ext4_has_metadata_csum(sb))
3543 set_opt(sb, JOURNAL_CHECKSUM);
3544
617ba13b 3545 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3546 set_opt(sb, JOURNAL_DATA);
617ba13b 3547 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3548 set_opt(sb, ORDERED_DATA);
617ba13b 3549 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3550 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3551
3552 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3553 set_opt(sb, ERRORS_PANIC);
bb4f397a 3554 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3555 set_opt(sb, ERRORS_CONT);
bb4f397a 3556 else
fd8c37ec 3557 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3558 /* block_validity enabled by default; disable with noblock_validity */
3559 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3560 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3561 set_opt(sb, DISCARD);
ac27a0ec 3562
08cefc7a
EB		 3563 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3564 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3565 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3566 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3567 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3568
8b67f04a 3569 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3570 set_opt(sb, BARRIER);
ac27a0ec 3571
dd919b98
AK		 3572 /*
3573 * enable delayed allocation by default
3574 * Use -o nodelalloc to turn it off
3575 */
bc0b75f7 3576 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3577 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3578 set_opt(sb, DELALLOC);
dd919b98 3579
51ce6511
LC		 3580 /*
3581 * set default s_li_wait_mult for lazyinit, for the case there is
3582 * no mount option specified.
3583 */
3584 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3585
8b67f04a 3586 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3587 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3588 ext4_msg(sb, KERN_WARNING,
3589 "failed to parse options in superblock: %s",
3590 sbi->s_es->s_mount_opts);
3591 }
5a916be1 3592 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3593 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3594 &journal_ioprio, 0))
ac27a0ec
DK		 3595 goto failed_mount;
3596
56889787
TT		 3597 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3598 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3599 "with data=journal disables delayed "
3600 "allocation and O_DIRECT support!\n");
3601 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3602 ext4_msg(sb, KERN_ERR, "can't mount with "
3603 "both data=journal and delalloc");
3604 goto failed_mount;
3605 }
3606 if (test_opt(sb, DIOREAD_NOLOCK)) {
3607 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3608 "both data=journal and dioread_nolock");
56889787
TT		 3609 goto failed_mount;
3610 }
3611 if (test_opt(sb, DELALLOC))
3612 clear_opt(sb, DELALLOC);
3613 }
3614
ac27a0ec 3615 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3616 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3617
617ba13b
MC		 3618 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3619 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3620 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3621 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3622 ext4_msg(sb, KERN_WARNING,
3623 "feature flags set on rev 0 fs, "
3624 "running e2fsck is recommended");
469108ff 3625
ed3654eb
TT		 3626 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3627 set_opt2(sb, HURD_COMPAT);
3628 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3629 EXT4_FEATURE_INCOMPAT_64BIT)) {
3630 ext4_msg(sb, KERN_ERR,
3631 "The Hurd can't support 64-bit file systems");
3632 goto failed_mount;
3633 }
3634 }
3635
2035e776
TT		 3636 if (IS_EXT2_SB(sb)) {
3637 if (ext2_feature_set_ok(sb))
3638 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3639 "using the ext4 subsystem");
3640 else {
3641 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3642 "to feature incompatibilities");
3643 goto failed_mount;
3644 }
3645 }
3646
3647 if (IS_EXT3_SB(sb)) {
3648 if (ext3_feature_set_ok(sb))
3649 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3650 "using the ext4 subsystem");
3651 else {
3652 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3653 "to feature incompatibilities");
3654 goto failed_mount;
3655 }
3656 }
3657
ac27a0ec
DK		 3658 /*
3659 * Check feature flags regardless of the revision level, since we
3660 * previously didn't change the revision level when setting the flags,
3661 * so there is a chance incompat flags are set on a rev 0 filesystem.
3662 */
a13fb1a4 3663 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3664 goto failed_mount;
a13fb1a4 3665
261cb20c 3666 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3667 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3668 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3669 ext4_msg(sb, KERN_ERR,
3670 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3671 goto failed_mount;
3672 }
3673
ac27a0ec 3674 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3675 /* Validate the filesystem blocksize */
3676 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3677 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3678 blocksize);
ac27a0ec
DK		 3679 goto failed_mount;
3680 }
3681
2b2d6d01 3682 brelse(bh);
70bbb3e0
AM		 3683 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3684 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3685 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3686 if (!bh) {
b31e1552
ES		 3687 ext4_msg(sb, KERN_ERR,
3688 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3689 goto failed_mount;
3690 }
2716b802 3691 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3692 sbi->s_es = es;
617ba13b 3693 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3694 ext4_msg(sb, KERN_ERR,
3695 "Magic mismatch, very weird!");
ac27a0ec
DK		 3696 goto failed_mount;
3697 }
3698 }
3699
a13fb1a4
ES		 3700 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3701 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3702 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3703 has_huge_files);
3704 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3705
617ba13b
MC		 3706 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3707 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3708 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3709 } else {
3710 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3711 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3712 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3713 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3714 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3715 ext4_msg(sb, KERN_ERR,
3716 "unsupported inode size: %d",
2b2d6d01 3717 sbi->s_inode_size);
ac27a0ec
DK		 3718 goto failed_mount;
3719 }
ef7f3835
KS		 3720 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3721 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3722 }
0b8e58a1 3723
0d1ee42f
AR		 3724 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3725 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3726 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3727 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3728 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3729 ext4_msg(sb, KERN_ERR,
3730 "unsupported descriptor size %lu",
0d1ee42f
AR		 3731 sbi->s_desc_size);
3732 goto failed_mount;
3733 }
3734 } else
3735 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3736
ac27a0ec 3737 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3738 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3739 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3740 goto cantfind_ext4;
0b8e58a1 3741
617ba13b 3742 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3743 if (sbi->s_inodes_per_block == 0)
617ba13b 3744 goto cantfind_ext4;
ac27a0ec
DK		 3745 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3746 sbi->s_inodes_per_block;
0d1ee42f 3747 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3748 sbi->s_sbh = bh;
3749 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3750 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3751 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3752
2b2d6d01 3753 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3754 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3755 sbi->s_def_hash_version = es->s_def_hash_version;
23301410
TT		 3756 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3757 i = le32_to_cpu(es->s_flags);
3758 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3759 sbi->s_hash_unsigned = 3;
3760 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3761#ifdef __CHAR_UNSIGNED__
23301410
TT		 3762 if (!(sb->s_flags & MS_RDONLY))
3763 es->s_flags |=
3764 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3765 sbi->s_hash_unsigned = 3;
f99b2589 3766#else
23301410
TT		 3767 if (!(sb->s_flags & MS_RDONLY))
3768 es->s_flags |=
3769 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3770#endif
23301410 3771 }
f99b2589 3772 }
ac27a0ec 3773
281b5995
TT		 3774 /* Handle clustersize */
3775 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3776 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3777 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3778 if (has_bigalloc) {
3779 if (clustersize < blocksize) {
3780 ext4_msg(sb, KERN_ERR,
3781 "cluster size (%d) smaller than "
3782 "block size (%d)", clustersize, blocksize);
3783 goto failed_mount;
3784 }
3785 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3786 le32_to_cpu(es->s_log_block_size);
3787 sbi->s_clusters_per_group =
3788 le32_to_cpu(es->s_clusters_per_group);
3789 if (sbi->s_clusters_per_group > blocksize * 8) {
3790 ext4_msg(sb, KERN_ERR,
3791 "#clusters per group too big: %lu",
3792 sbi->s_clusters_per_group);
3793 goto failed_mount;
3794 }
3795 if (sbi->s_blocks_per_group !=
3796 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3797 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3798 "clusters per group (%lu) inconsistent",
3799 sbi->s_blocks_per_group,
3800 sbi->s_clusters_per_group);
3801 goto failed_mount;
3802 }
3803 } else {
3804 if (clustersize != blocksize) {
3805 ext4_warning(sb, "fragment/cluster size (%d) != "
3806 "block size (%d)", clustersize,
3807 blocksize);
3808 clustersize = blocksize;
3809 }
3810 if (sbi->s_blocks_per_group > blocksize * 8) {
3811 ext4_msg(sb, KERN_ERR,
3812 "#blocks per group too big: %lu",
3813 sbi->s_blocks_per_group);
3814 goto failed_mount;
3815 }
3816 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3817 sbi->s_cluster_bits = 0;
ac27a0ec 3818 }
281b5995
TT		 3819 sbi->s_cluster_ratio = clustersize / blocksize;
3820
ac27a0ec 3821 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3822 ext4_msg(sb, KERN_ERR,
3823 "#inodes per group too big: %lu",
2b2d6d01 3824 sbi->s_inodes_per_group);
ac27a0ec
DK		 3825 goto failed_mount;
3826 }
3827
960fd856
TT		 3828 /* Do we have standard group size of clustersize * 8 blocks ? */
3829 if (sbi->s_blocks_per_group == clustersize << 3)
3830 set_opt2(sb, STD_GROUP_SIZE);
3831
bf43d84b
ES		 3832 /*
3833 * Test whether we have more sectors than will fit in sector_t,
3834 * and whether the max offset is addressable by the page cache.
3835 */
5a9ae68a 3836 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3837 ext4_blocks_count(es));
5a9ae68a 3838 if (err) {
b31e1552 3839 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3840 " too large to mount safely on this system");
ac27a0ec 3841 if (sizeof(sector_t) < 8)
90c699a9 3842 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3843 goto failed_mount;
3844 }
3845
617ba13b
MC		 3846 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3847 goto cantfind_ext4;
e7c95593 3848
0f2ddca6
FTN		 3849 /* check blocks count against device size */
3850 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3851 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3852 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3853 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3854 ext4_blocks_count(es), blocks_count);
3855 goto failed_mount;
3856 }
3857
0b8e58a1
AD		 3858 /*
3859 * It makes no sense for the first data block to be beyond the end
3860 * of the filesystem.
3861 */
3862 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3863 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3864 "block %u is beyond end of filesystem (%llu)",
3865 le32_to_cpu(es->s_first_data_block),
3866 ext4_blocks_count(es));
e7c95593
ES		 3867 goto failed_mount;
3868 }
bd81d8ee
LV		 3869 blocks_count = (ext4_blocks_count(es) -
3870 le32_to_cpu(es->s_first_data_block) +
3871 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3872 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3873 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3874 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3875 "(block count %llu, first data block %u, "
b31e1552 3876 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3877 ext4_blocks_count(es),
3878 le32_to_cpu(es->s_first_data_block),
3879 EXT4_BLOCKS_PER_GROUP(sb));
3880 goto failed_mount;
3881 }
bd81d8ee 3882 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3883 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3884 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3885 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3886 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3887 sbi->s_group_desc = ext4_kvmalloc(db_count *
3888 sizeof(struct buffer_head *),
3889 GFP_KERNEL);
ac27a0ec 3890 if (sbi->s_group_desc == NULL) {
b31e1552 3891 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3892 ret = -ENOMEM;
ac27a0ec
DK		 3893 goto failed_mount;
3894 }
3895
9f6200bb
TT		 3896 if (ext4_proc_root)
3897 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3898
66acdcf4
TT		 3899 if (sbi->s_proc)
3900 proc_create_data("options", S_IRUGO, sbi->s_proc,
3901 &ext4_seq_options_fops, sb);
3902
705895b6 3903 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3904
3905 for (i = 0; i < db_count; i++) {
70bbb3e0 3906 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3907 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3908 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3909 ext4_msg(sb, KERN_ERR,
3910 "can't read group descriptor %d", i);
ac27a0ec
DK		 3911 db_count = i;
3912 goto failed_mount2;
3913 }
3914 }
bfff6873 3915 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3916 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
f9ae9cf5 3917 goto failed_mount2;
ac27a0ec 3918 }
772cb7c8 3919
f9ae9cf5 3920 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3921 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3922 spin_lock_init(&sbi->s_next_gen_lock);
3923
04ecddb7
JM		 3924 setup_timer(&sbi->s_err_report, print_daily_error_info,
3925 (unsigned long) sb);
04496411 3926
a75ae78f 3927 /* Register extent status tree shrinker */
eb68d0e2 3928 if (ext4_es_register_shrinker(sbi))
ce7e010a 3929 goto failed_mount3;
ce7e010a 3930
c9de560d 3931 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3932 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3933
f9ae9cf5
TT		 3934 /*
3935 * set up enough so that it can read an inode
3936 */
f6e63f90 3937 sb->s_op = &ext4_sops;
617ba13b
MC		 3938 sb->s_export_op = &ext4_export_ops;
3939 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3940#ifdef CONFIG_QUOTA
617ba13b 3941 sb->dq_op = &ext4_quota_operations;
262b4662 3942 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
7c319d32 3943 sb->s_qcop = &ext4_qctl_sysfile_operations;
262b4662
JK		 3944 else
3945 sb->s_qcop = &ext4_qctl_operations;
96c7e0d9 3946 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
ac27a0ec 3947#endif
f2fa2ffc
AK		 3948 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3949
ac27a0ec 3950 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3951 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3952
3953 sb->s_root = NULL;
3954
3955 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3956 EXT4_HAS_INCOMPAT_FEATURE(sb,
3957 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3958
c5e06d10
JL		 3959 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3960 !(sb->s_flags & MS_RDONLY))
3961 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3962 goto failed_mount3a;
c5e06d10 3963
ac27a0ec
DK		 3964 /*
3965 * The first inode we look at is the journal inode. Don't try
3966 * root first: it may be modified in the journal!
3967 */
3968 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3969 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3970 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 3971 goto failed_mount3a;
0390131b
FM		 3972 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3973 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3974 ext4_msg(sb, KERN_ERR, "required journal recovery "
3975 "suppressed and not mounted read-only");
744692dc 3976 goto failed_mount_wq;
ac27a0ec 3977 } else {
fd8c37ec 3978 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3979 sbi->s_journal = NULL;
3980 needs_recovery = 0;
3981 goto no_journal;
ac27a0ec
DK		 3982 }
3983
f32aaf2d 3984 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3985 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3986 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3987 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3988 goto failed_mount_wq;
eb40a09c
JS		 3989 }
3990
25ed6e8a
DW		 3991 if (!set_journal_csum_feature_set(sb)) {
3992 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3993 "feature set");
3994 goto failed_mount_wq;
d4da6c9c 3995 }
818d276c 3996
ac27a0ec
DK		 3997 /* We have now updated the journal if required, so we can
3998 * validate the data journaling mode. */
3999 switch (test_opt(sb, DATA_FLAGS)) {
4000 case 0:
4001 /* No mode set, assume a default based on the journal
63f57933
AM		 4002 * capabilities: ORDERED_DATA if the journal can
4003 * cope, else JOURNAL_DATA
4004 */
dab291af
MC		 4005 if (jbd2_journal_check_available_features
4006 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 4007 set_opt(sb, ORDERED_DATA);
ac27a0ec 4008 else
fd8c37ec 4009 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 4010 break;
4011
617ba13b
MC		 4012 case EXT4_MOUNT_ORDERED_DATA:
4013 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4014 if (!jbd2_journal_check_available_features
4015 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4016 ext4_msg(sb, KERN_ERR, "Journal does not support "
4017 "requested data journaling mode");
744692dc 4018 goto failed_mount_wq;
ac27a0ec
DK		 4019 }
4020 default:
4021 break;
4022 }
b3881f74 4023 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4024
18aadd47
BJ		 4025 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4026
ce7e010a 4027no_journal:
9c191f70
M		 4028 if (ext4_mballoc_ready) {
4029 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4030 if (!sbi->s_mb_cache) {
4031 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4032 goto failed_mount_wq;
4033 }
4034 }
4035
952fc18e
TT		 4036 /*
4037 * Get the # of file system overhead blocks from the
4038 * superblock if present.
4039 */
4040 if (es->s_overhead_clusters)
4041 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4042 else {
07aa2ea1
LC		 4043 err = ext4_calculate_overhead(sb);
4044 if (err)
952fc18e
TT		 4045 goto failed_mount_wq;
4046 }
4047
fd89d5f2
TH		 4048 /*
4049 * The maximum number of concurrent works can be high and
4050 * concurrency isn't really necessary. Limit it to 1.
4051 */
2e8fa54e
JK		 4052 EXT4_SB(sb)->rsv_conversion_wq =
4053 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4054 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4055 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4056 ret = -ENOMEM;
2e8fa54e
JK		 4057 goto failed_mount4;
4058 }
4059
ac27a0ec 4060 /*
dab291af 4061 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4062 * so we can safely mount the rest of the filesystem now.
4063 */
4064
1d1fe1ee
DH		 4065 root = ext4_iget(sb, EXT4_ROOT_INO);
4066 if (IS_ERR(root)) {
b31e1552 4067 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4068 ret = PTR_ERR(root);
32a9bb57 4069 root = NULL;
ac27a0ec
DK		 4070 goto failed_mount4;
4071 }
4072 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4073 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4074 iput(root);
ac27a0ec
DK		 4075 goto failed_mount4;
4076 }
48fde701 4077 sb->s_root = d_make_root(root);
1d1fe1ee 4078 if (!sb->s_root) {
b31e1552 4079 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4080 ret = -ENOMEM;
4081 goto failed_mount4;
4082 }
ac27a0ec 4083
7e84b621
ES		 4084 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4085 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4086
4087 /* determine the minimum size of new large inodes, if present */
4088 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4089 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4090 EXT4_GOOD_OLD_INODE_SIZE;
4091 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4092 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4093 if (sbi->s_want_extra_isize <
4094 le16_to_cpu(es->s_want_extra_isize))
4095 sbi->s_want_extra_isize =
4096 le16_to_cpu(es->s_want_extra_isize);
4097 if (sbi->s_want_extra_isize <
4098 le16_to_cpu(es->s_min_extra_isize))
4099 sbi->s_want_extra_isize =
4100 le16_to_cpu(es->s_min_extra_isize);
4101 }
4102 }
4103 /* Check if enough inode space is available */
4104 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4105 sbi->s_inode_size) {
4106 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4107 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4108 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4109 "available");
ef7f3835
KS		 4110 }
4111
30fac0f7 4112 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
27dd4385
LC		 4113 if (err) {
4114 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
30fac0f7 4115 "reserved pool", ext4_calculate_resv_clusters(sb));
f9ae9cf5 4116 goto failed_mount4a;
27dd4385
LC		 4117 }
4118
6fd058f7
TT		 4119 err = ext4_setup_system_zone(sb);
4120 if (err) {
b31e1552 4121 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4122 "zone (%d)", err);
f9ae9cf5
TT		 4123 goto failed_mount4a;
4124 }
4125
4126 ext4_ext_init(sb);
4127 err = ext4_mb_init(sb);
4128 if (err) {
4129 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4130 err);
dcf2d804 4131 goto failed_mount5;
c2774d84
AK		 4132 }
4133
d5e03cbb
TT		 4134 block = ext4_count_free_clusters(sb);
4135 ext4_free_blocks_count_set(sbi->s_es,
4136 EXT4_C2B(sbi, block));
908c7f19
TH		 4137 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4138 GFP_KERNEL);
d5e03cbb
TT		 4139 if (!err) {
4140 unsigned long freei = ext4_count_free_inodes(sb);
4141 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4142 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4143 GFP_KERNEL);
d5e03cbb
TT		 4144 }
4145 if (!err)
4146 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4147 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4148 if (!err)
908c7f19
TH		 4149 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4150 GFP_KERNEL);
d5e03cbb
TT		 4151 if (err) {
4152 ext4_msg(sb, KERN_ERR, "insufficient memory");
4153 goto failed_mount6;
4154 }
4155
4156 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4157 if (!ext4_fill_flex_info(sb)) {
4158 ext4_msg(sb, KERN_ERR,
4159 "unable to initialize "
4160 "flex_bg meta info!");
4161 goto failed_mount6;
4162 }
4163
bfff6873
LC		 4164 err = ext4_register_li_request(sb, first_not_zeroed);
4165 if (err)
dcf2d804 4166 goto failed_mount6;
bfff6873 4167
3197ebdb
TT		 4168 sbi->s_kobj.kset = ext4_kset;
4169 init_completion(&sbi->s_kobj_unregister);
4170 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4171 "%s", sb->s_id);
dcf2d804
TM		 4172 if (err)
4173 goto failed_mount7;
3197ebdb 4174
9b2ff357
JK		 4175#ifdef CONFIG_QUOTA
4176 /* Enable quota usage during mount. */
4177 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4178 !(sb->s_flags & MS_RDONLY)) {
4179 err = ext4_enable_quotas(sb);
4180 if (err)
4181 goto failed_mount8;
4182 }
4183#endif /* CONFIG_QUOTA */
4184
617ba13b
MC		 4185 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4186 ext4_orphan_cleanup(sb, es);
4187 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4188 if (needs_recovery) {
b31e1552 4189 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4190 ext4_mark_recovery_complete(sb, es);
4191 }
4192 if (EXT4_SB(sb)->s_journal) {
4193 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4194 descr = " journalled data mode";
4195 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4196 descr = " ordered data mode";
4197 else
4198 descr = " writeback data mode";
4199 } else
4200 descr = "out journal";
4201
79add3a3
LC		 4202 if (test_opt(sb, DISCARD)) {
4203 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4204 if (!blk_queue_discard(q))
4205 ext4_msg(sb, KERN_WARNING,
4206 "mounting with \"discard\" option, but "
4207 "the device does not support discard");
4208 }
4209
d4c402d9 4210 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4211 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4212 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4213
66e61a9e
TT		 4214 if (es->s_error_count)
4215 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4216
efbed4dc
TT		 4217 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4218 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4219 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4220 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4221
d4c402d9 4222 kfree(orig_data);
ac27a0ec
DK		 4223 return 0;
4224
617ba13b 4225cantfind_ext4:
ac27a0ec 4226 if (!silent)
b31e1552 4227 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4228 goto failed_mount;
4229
72ba7450
TT		 4230#ifdef CONFIG_QUOTA
4231failed_mount8:
4232 kobject_del(&sbi->s_kobj);
4233#endif
dcf2d804
TM		 4234failed_mount7:
4235 ext4_unregister_li_request(sb);
4236failed_mount6:
f9ae9cf5 4237 ext4_mb_release(sb);
d5e03cbb 4238 if (sbi->s_flex_groups)
b93b41d4 4239 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4240 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4241 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4242 percpu_counter_destroy(&sbi->s_dirs_counter);
4243 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4244failed_mount5:
f9ae9cf5
TT		 4245 ext4_ext_release(sb);
4246 ext4_release_system_zone(sb);
4247failed_mount4a:
94bf608a 4248 dput(sb->s_root);
32a9bb57 4249 sb->s_root = NULL;
94bf608a 4250failed_mount4:
b31e1552 4251 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4252 if (EXT4_SB(sb)->rsv_conversion_wq)
4253 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4254failed_mount_wq:
0390131b
FM		 4255 if (sbi->s_journal) {
4256 jbd2_journal_destroy(sbi->s_journal);
4257 sbi->s_journal = NULL;
4258 }
50460fe8 4259failed_mount3a:
d3922a77 4260 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4261failed_mount3:
9105bb14 4262 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4263 if (sbi->s_mmp_tsk)
4264 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4265failed_mount2:
4266 for (i = 0; i < db_count; i++)
4267 brelse(sbi->s_group_desc[i]);
b93b41d4 4268 kvfree(sbi->s_group_desc);
ac27a0ec 4269failed_mount:
0441984a
DW		 4270 if (sbi->s_chksum_driver)
4271 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4272 if (sbi->s_proc) {
66acdcf4 4273 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4274 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4275 }
ac27a0ec 4276#ifdef CONFIG_QUOTA
a2d4a646 4277 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4278 kfree(sbi->s_qf_names[i]);
4279#endif
617ba13b 4280 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4281 brelse(bh);
4282out_fail:
4283 sb->s_fs_info = NULL;
f6830165 4284 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4285 kfree(sbi);
dcc7dae3 4286out_free_orig:
d4c402d9 4287 kfree(orig_data);
07aa2ea1 4288 return err ? err : ret;
ac27a0ec
DK		 4289}
4290
4291/*
4292 * Setup any per-fs journal parameters now. We'll do this both on
4293 * initial mount, once the journal has been initialised but before we've
4294 * done any recovery; and again on any subsequent remount.
4295 */
617ba13b 4296static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4297{
617ba13b 4298 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4299
30773840
TT		 4300 journal->j_commit_interval = sbi->s_commit_interval;
4301 journal->j_min_batch_time = sbi->s_min_batch_time;
4302 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4303
a931da6a 4304 write_lock(&journal->j_state_lock);
ac27a0ec 4305 if (test_opt(sb, BARRIER))
dab291af 4306 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4307 else
dab291af 4308 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4309 if (test_opt(sb, DATA_ERR_ABORT))
4310 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4311 else
4312 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4313 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4314}
4315
617ba13b 4316static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4317 unsigned int journal_inum)
4318{
4319 struct inode *journal_inode;
4320 journal_t *journal;
4321
0390131b
FM		 4322 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4323
ac27a0ec
DK		 4324 /* First, test for the existence of a valid inode on disk. Bad
4325 * things happen if we iget() an unused inode, as the subsequent
4326 * iput() will try to delete it. */
4327
1d1fe1ee
DH		 4328 journal_inode = ext4_iget(sb, journal_inum);
4329 if (IS_ERR(journal_inode)) {
b31e1552 4330 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4331 return NULL;
4332 }
4333 if (!journal_inode->i_nlink) {
4334 make_bad_inode(journal_inode);
4335 iput(journal_inode);
b31e1552 4336 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4337 return NULL;
4338 }
4339
e5f8eab8 4340 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4341 journal_inode, journal_inode->i_size);
1d1fe1ee 4342 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4343 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4344 iput(journal_inode);
4345 return NULL;
4346 }
4347
dab291af 4348 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4349 if (!journal) {
b31e1552 4350 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4351 iput(journal_inode);
4352 return NULL;
4353 }
4354 journal->j_private = sb;
617ba13b 4355 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4356 return journal;
4357}
4358
617ba13b 4359static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4360 dev_t j_dev)
4361{
2b2d6d01 4362 struct buffer_head *bh;
ac27a0ec 4363 journal_t *journal;
617ba13b
MC		 4364 ext4_fsblk_t start;
4365 ext4_fsblk_t len;
ac27a0ec 4366 int hblock, blocksize;
617ba13b 4367 ext4_fsblk_t sb_block;
ac27a0ec 4368 unsigned long offset;
2b2d6d01 4369 struct ext4_super_block *es;
ac27a0ec
DK		 4370 struct block_device *bdev;
4371
0390131b
FM		 4372 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4373
b31e1552 4374 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4375 if (bdev == NULL)
4376 return NULL;
4377
ac27a0ec 4378 blocksize = sb->s_blocksize;
e1defc4f 4379 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4380 if (blocksize < hblock) {
b31e1552
ES		 4381 ext4_msg(sb, KERN_ERR,
4382 "blocksize too small for journal device");
ac27a0ec
DK		 4383 goto out_bdev;
4384 }
4385
617ba13b
MC		 4386 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4387 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4388 set_blocksize(bdev, blocksize);
4389 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4390 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4391 "external journal");
ac27a0ec
DK		 4392 goto out_bdev;
4393 }
4394
2716b802 4395 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4396 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4397 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4398 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4399 ext4_msg(sb, KERN_ERR, "external journal has "
4400 "bad superblock");
ac27a0ec
DK		 4401 brelse(bh);
4402 goto out_bdev;
4403 }
4404
df4763be
DW		 4405 if ((le32_to_cpu(es->s_feature_ro_compat) &
4406 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4407 es->s_checksum != ext4_superblock_csum(sb, es)) {
4408 ext4_msg(sb, KERN_ERR, "external journal has "
4409 "corrupt superblock");
4410 brelse(bh);
4411 goto out_bdev;
4412 }
4413
617ba13b 4414 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4415 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4416 brelse(bh);
4417 goto out_bdev;
4418 }
4419
bd81d8ee 4420 len = ext4_blocks_count(es);
ac27a0ec
DK		 4421 start = sb_block + 1;
4422 brelse(bh); /* we're done with the superblock */
4423
dab291af 4424 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4425 start, len, blocksize);
4426 if (!journal) {
b31e1552 4427 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4428 goto out_bdev;
4429 }
4430 journal->j_private = sb;
9f203507 4431 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4432 wait_on_buffer(journal->j_sb_buffer);
4433 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4434 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4435 goto out_journal;
4436 }
4437 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4438 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4439 "user (unsupported) - %d",
ac27a0ec
DK		 4440 be32_to_cpu(journal->j_superblock->s_nr_users));
4441 goto out_journal;
4442 }
617ba13b
MC		 4443 EXT4_SB(sb)->journal_bdev = bdev;
4444 ext4_init_journal_params(sb, journal);
ac27a0ec 4445 return journal;
0b8e58a1 4446
ac27a0ec 4447out_journal:
dab291af 4448 jbd2_journal_destroy(journal);
ac27a0ec 4449out_bdev:
617ba13b 4450 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4451 return NULL;
4452}
4453
617ba13b
MC		 4454static int ext4_load_journal(struct super_block *sb,
4455 struct ext4_super_block *es,
ac27a0ec
DK		 4456 unsigned long journal_devnum)
4457{
4458 journal_t *journal;
4459 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4460 dev_t journal_dev;
4461 int err = 0;
4462 int really_read_only;
4463
0390131b
FM		 4464 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4465
ac27a0ec
DK		 4466 if (journal_devnum &&
4467 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4468 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4469 "numbers have changed");
ac27a0ec
DK		 4470 journal_dev = new_decode_dev(journal_devnum);
4471 } else
4472 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4473
4474 really_read_only = bdev_read_only(sb->s_bdev);
4475
4476 /*
4477 * Are we loading a blank journal or performing recovery after a
4478 * crash? For recovery, we need to check in advance whether we
4479 * can get read-write access to the device.
4480 */
617ba13b 4481 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4482 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4483 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4484 "required on readonly filesystem");
ac27a0ec 4485 if (really_read_only) {
b31e1552
ES		 4486 ext4_msg(sb, KERN_ERR, "write access "
4487 "unavailable, cannot proceed");
ac27a0ec
DK		 4488 return -EROFS;
4489 }
b31e1552
ES		 4490 ext4_msg(sb, KERN_INFO, "write access will "
4491 "be enabled during recovery");
ac27a0ec
DK		 4492 }
4493 }
4494
4495 if (journal_inum && journal_dev) {
b31e1552
ES		 4496 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4497 "and inode journals!");
ac27a0ec
DK		 4498 return -EINVAL;
4499 }
4500
4501 if (journal_inum) {
617ba13b 4502 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4503 return -EINVAL;
4504 } else {
617ba13b 4505 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4506 return -EINVAL;
4507 }
4508
90576c0b 4509 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4510 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4511
617ba13b 4512 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4513 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4514 if (!err) {
4515 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4516 if (save)
4517 memcpy(save, ((char *) es) +
4518 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4519 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4520 if (save)
4521 memcpy(((char *) es) + EXT4_S_ERR_START,
4522 save, EXT4_S_ERR_LEN);
4523 kfree(save);
4524 }
ac27a0ec
DK		 4525
4526 if (err) {
b31e1552 4527 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4528 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4529 return err;
4530 }
4531
617ba13b
MC		 4532 EXT4_SB(sb)->s_journal = journal;
4533 ext4_clear_journal_err(sb, es);
ac27a0ec 4534
c41303ce 4535 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4536 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4537 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4538
4539 /* Make sure we flush the recovery flag to disk. */
e2d67052 4540 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4541 }
4542
4543 return 0;
4544}
4545
e2d67052 4546static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4547{
e2d67052 4548 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4549 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4550 int error = 0;
ac27a0ec 4551
7c2e7087 4552 if (!sbh || block_device_ejected(sb))
c4be0c1d 4553 return error;
914258bf
TT		 4554 if (buffer_write_io_error(sbh)) {
4555 /*
4556 * Oh, dear. A previous attempt to write the
4557 * superblock failed. This could happen because the
4558 * USB device was yanked out. Or it could happen to
4559 * be a transient write error and maybe the block will
4560 * be remapped. Nothing we can do but to retry the
4561 * write and hope for the best.
4562 */
b31e1552
ES		 4563 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4564 "superblock detected");
914258bf
TT		 4565 clear_buffer_write_io_error(sbh);
4566 set_buffer_uptodate(sbh);
4567 }
71290b36
TT		 4568 /*
4569 * If the file system is mounted read-only, don't update the
4570 * superblock write time. This avoids updating the superblock
4571 * write time when we are mounting the root file system
4572 * read/only but we need to replay the journal; at that point,
4573 * for people who are east of GMT and who make their clock
4574 * tick in localtime for Windows bug-for-bug compatibility,
4575 * the clock is set in the future, and this will cause e2fsck
4576 * to complain and force a full file system check.
4577 */
4578 if (!(sb->s_flags & MS_RDONLY))
4579 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4580 if (sb->s_bdev->bd_part)
4581 es->s_kbytes_written =
4582 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4583 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4584 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4585 else
4586 es->s_kbytes_written =
4587 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4588 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4589 ext4_free_blocks_count_set(es,
57042651
TT		 4590 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4591 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4592 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4593 es->s_free_inodes_count =
4594 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4595 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4596 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4597 ext4_superblock_csum_set(sb);
ac27a0ec 4598 mark_buffer_dirty(sbh);
914258bf 4599 if (sync) {
c4be0c1d
TS		 4600 error = sync_dirty_buffer(sbh);
4601 if (error)
4602 return error;
4603
4604 error = buffer_write_io_error(sbh);
4605 if (error) {
b31e1552
ES		 4606 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4607 "superblock");
914258bf
TT		 4608 clear_buffer_write_io_error(sbh);
4609 set_buffer_uptodate(sbh);
4610 }
4611 }
c4be0c1d 4612 return error;
ac27a0ec
DK		 4613}
4614
ac27a0ec
DK		 4615/*
4616 * Have we just finished recovery? If so, and if we are mounting (or
4617 * remounting) the filesystem readonly, then we will end up with a
4618 * consistent fs on disk. Record that fact.
4619 */
2b2d6d01
TT		 4620static void ext4_mark_recovery_complete(struct super_block *sb,
4621 struct ext4_super_block *es)
ac27a0ec 4622{
617ba13b 4623 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4624
0390131b
FM		 4625 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4626 BUG_ON(journal != NULL);
4627 return;
4628 }
dab291af 4629 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4630 if (jbd2_journal_flush(journal) < 0)
4631 goto out;
4632
617ba13b 4633 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4634 sb->s_flags & MS_RDONLY) {
617ba13b 4635 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4636 ext4_commit_super(sb, 1);
ac27a0ec 4637 }
7ffe1ea8
HK		 4638
4639out:
dab291af 4640 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4641}
4642
4643/*
4644 * If we are mounting (or read-write remounting) a filesystem whose journal
4645 * has recorded an error from a previous lifetime, move that error to the
4646 * main filesystem now.
4647 */
2b2d6d01
TT		 4648static void ext4_clear_journal_err(struct super_block *sb,
4649 struct ext4_super_block *es)
ac27a0ec
DK		 4650{
4651 journal_t *journal;
4652 int j_errno;
4653 const char *errstr;
4654
0390131b
FM		 4655 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4656
617ba13b 4657 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4658
4659 /*
4660 * Now check for any error status which may have been recorded in the
617ba13b 4661 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4662 */
4663
dab291af 4664 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4665 if (j_errno) {
4666 char nbuf[16];
4667
617ba13b 4668 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4669 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4670 "from previous mount: %s", errstr);
12062ddd 4671 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4672
617ba13b
MC		 4673 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4674 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4675 ext4_commit_super(sb, 1);
ac27a0ec 4676
dab291af 4677 jbd2_journal_clear_err(journal);
d796c52e 4678 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4679 }
4680}
4681
4682/*
4683 * Force the running and committing transactions to commit,
4684 * and wait on the commit.
4685 */
617ba13b 4686int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4687{
4688 journal_t *journal;
ac27a0ec
DK		 4689
4690 if (sb->s_flags & MS_RDONLY)
4691 return 0;
4692
617ba13b 4693 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4694 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4695}
4696
617ba13b 4697static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4698{
14ce0cb4 4699 int ret = 0;
9eddacf9 4700 tid_t target;
06a407f1 4701 bool needs_barrier = false;
8d5d02e6 4702 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4703
9bffad1e 4704 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4705 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4706 /*
4707 * Writeback quota in non-journalled quota case - journalled quota has
4708 * no dirty dquots
4709 */
4710 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4711 /*
4712 * Data writeback is possible w/o journal transaction, so barrier must
4713 * being sent at the end of the function. But we can skip it if
4714 * transaction_commit will do it for us.
4715 */
bda32530
TT		 4716 if (sbi->s_journal) {
4717 target = jbd2_get_latest_transaction(sbi->s_journal);
4718 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4719 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4720 needs_barrier = true;
4721
4722 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4723 if (wait)
4724 ret = jbd2_log_wait_commit(sbi->s_journal,
4725 target);
4726 }
4727 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4728 needs_barrier = true;
06a407f1
DM		 4729 if (needs_barrier) {
4730 int err;
4731 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4732 if (!ret)
4733 ret = err;
0390131b 4734 }
06a407f1
DM		 4735
4736 return ret;
4737}
4738
ac27a0ec
DK		 4739/*
4740 * LVM calls this function before a (read-only) snapshot is created. This
4741 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4742 *
4743 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4744 * state independently. It relies on upper layer to stop all data & metadata
4745 * modifications.
ac27a0ec 4746 */
c4be0c1d 4747static int ext4_freeze(struct super_block *sb)
ac27a0ec 4748{
c4be0c1d
TS		 4749 int error = 0;
4750 journal_t *journal;
ac27a0ec 4751
9ca92389
TT		 4752 if (sb->s_flags & MS_RDONLY)
4753 return 0;
ac27a0ec 4754
9ca92389 4755 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4756
bb044576
TT		 4757 if (journal) {
4758 /* Now we set up the journal barrier. */
4759 jbd2_journal_lock_updates(journal);
ac27a0ec 4760
bb044576
TT		 4761 /*
4762 * Don't clear the needs_recovery flag if we failed to
4763 * flush the journal.
4764 */
4765 error = jbd2_journal_flush(journal);
4766 if (error < 0)
4767 goto out;
4768 }
9ca92389
TT		 4769
4770 /* Journal blocked and flushed, clear needs_recovery flag. */
4771 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4772 error = ext4_commit_super(sb, 1);
6b0310fb 4773out:
bb044576
TT		 4774 if (journal)
4775 /* we rely on upper layer to stop further updates */
4776 jbd2_journal_unlock_updates(journal);
6b0310fb 4777 return error;
ac27a0ec
DK		 4778}
4779
4780/*
4781 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4782 * flag here, even though the filesystem is not technically dirty yet.
4783 */
c4be0c1d 4784static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4785{
9ca92389
TT		 4786 if (sb->s_flags & MS_RDONLY)
4787 return 0;
4788
9ca92389
TT		 4789 /* Reset the needs_recovery flag before the fs is unlocked. */
4790 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4791 ext4_commit_super(sb, 1);
c4be0c1d 4792 return 0;
ac27a0ec
DK		 4793}
4794
673c6100
TT		 4795/*
4796 * Structure to save mount options for ext4_remount's benefit
4797 */
4798struct ext4_mount_options {
4799 unsigned long s_mount_opt;
a2595b8a 4800 unsigned long s_mount_opt2;
08cefc7a
EB		 4801 kuid_t s_resuid;
4802 kgid_t s_resgid;
673c6100
TT		 4803 unsigned long s_commit_interval;
4804 u32 s_min_batch_time, s_max_batch_time;
4805#ifdef CONFIG_QUOTA
4806 int s_jquota_fmt;
a2d4a646 4807 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4808#endif
4809};
4810
2b2d6d01 4811static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4812{
2b2d6d01 4813 struct ext4_super_block *es;
617ba13b 4814 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4815 unsigned long old_sb_flags;
617ba13b 4816 struct ext4_mount_options old_opts;
c79d967d 4817 int enable_quota = 0;
8a266467 4818 ext4_group_t g;
b3881f74 4819 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4820 int err = 0;
ac27a0ec 4821#ifdef CONFIG_QUOTA
03dafb5f 4822 int i, j;
ac27a0ec 4823#endif
d4c402d9 4824 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4825
4826 /* Store the original options */
4827 old_sb_flags = sb->s_flags;
4828 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4829 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4830 old_opts.s_resuid = sbi->s_resuid;
4831 old_opts.s_resgid = sbi->s_resgid;
4832 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4833 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4834 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4835#ifdef CONFIG_QUOTA
4836 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4837 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4838 if (sbi->s_qf_names[i]) {
4839 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4840 GFP_KERNEL);
4841 if (!old_opts.s_qf_names[i]) {
4842 for (j = 0; j < i; j++)
4843 kfree(old_opts.s_qf_names[j]);
3e36a163 4844 kfree(orig_data);
03dafb5f
CG		 4845 return -ENOMEM;
4846 }
4847 } else
4848 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4849#endif
b3881f74
TT		 4850 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4851 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4852
661aa520 4853 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4854 err = -EINVAL;
4855 goto restore_opts;
4856 }
4857
6b992ff2 4858 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d 4859 test_opt(sb, JOURNAL_CHECKSUM)) {
6b992ff2
DW		 4860 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4861 "during remount not supported");
4862 err = -EINVAL;
4863 goto restore_opts;
4864 }
4865
6ae6514b
PS		 4866 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4867 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4868 ext4_msg(sb, KERN_ERR, "can't mount with "
4869 "both data=journal and delalloc");
4870 err = -EINVAL;
4871 goto restore_opts;
4872 }
4873 if (test_opt(sb, DIOREAD_NOLOCK)) {
4874 ext4_msg(sb, KERN_ERR, "can't mount with "
4875 "both data=journal and dioread_nolock");
4876 err = -EINVAL;
4877 goto restore_opts;
4878 }
4879 }
4880
4ab2f15b 4881 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4882 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4883
4884 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4885 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4886
4887 es = sbi->s_es;
4888
b3881f74 4889 if (sbi->s_journal) {
0390131b 4890 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4891 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4892 }
ac27a0ec 4893
661aa520 4894 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4895 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4896 err = -EROFS;
4897 goto restore_opts;
4898 }
4899
4900 if (*flags & MS_RDONLY) {
38c03b34
TT		 4901 err = sync_filesystem(sb);
4902 if (err < 0)
4903 goto restore_opts;
0f0dd62f
CH		 4904 err = dquot_suspend(sb, -1);
4905 if (err < 0)
c79d967d 4906 goto restore_opts;
c79d967d 4907
ac27a0ec
DK		 4908 /*
4909 * First of all, the unconditional stuff we have to do
4910 * to disable replay of the journal when we next remount
4911 */
4912 sb->s_flags |= MS_RDONLY;
4913
4914 /*
4915 * OK, test if we are remounting a valid rw partition
4916 * readonly, and if so set the rdonly flag and then
4917 * mark the partition as valid again.
4918 */
617ba13b
MC		 4919 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4920 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4921 es->s_state = cpu_to_le16(sbi->s_mount_state);
4922
a63c9eb2 4923 if (sbi->s_journal)
0390131b 4924 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4925 } else {
a13fb1a4 4926 /* Make sure we can mount this feature set readwrite */
2cb5cc8b
DW		 4927 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4928 EXT4_FEATURE_RO_COMPAT_READONLY) ||
4929 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4930 err = -EROFS;
4931 goto restore_opts;
4932 }
8a266467
TT		 4933 /*
4934 * Make sure the group descriptor checksums
0b8e58a1 4935 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4936 */
4937 for (g = 0; g < sbi->s_groups_count; g++) {
4938 struct ext4_group_desc *gdp =
4939 ext4_get_group_desc(sb, g, NULL);
4940
feb0ab32 4941 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4942 ext4_msg(sb, KERN_ERR,
4943 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4944 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4945 le16_to_cpu(gdp->bg_checksum));
4946 err = -EINVAL;
4947 goto restore_opts;
4948 }
4949 }
4950
ead6596b
ES		 4951 /*
4952 * If we have an unprocessed orphan list hanging
4953 * around from a previously readonly bdev mount,
4954 * require a full umount/remount for now.
4955 */
4956 if (es->s_last_orphan) {
b31e1552 4957 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4958 "remount RDWR because of unprocessed "
4959 "orphan inode list. Please "
b31e1552 4960 "umount/remount instead");
ead6596b
ES		 4961 err = -EINVAL;
4962 goto restore_opts;
4963 }
4964
ac27a0ec
DK		 4965 /*
4966 * Mounting a RDONLY partition read-write, so reread
4967 * and store the current valid flag. (It may have
4968 * been changed by e2fsck since we originally mounted
4969 * the partition.)
4970 */
0390131b
FM		 4971 if (sbi->s_journal)
4972 ext4_clear_journal_err(sb, es);
ac27a0ec 4973 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4974 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4975 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4976 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4977 EXT4_FEATURE_INCOMPAT_MMP))
4978 if (ext4_multi_mount_protect(sb,
4979 le64_to_cpu(es->s_mmp_block))) {
4980 err = -EROFS;
4981 goto restore_opts;
4982 }
c79d967d 4983 enable_quota = 1;
ac27a0ec
DK		 4984 }
4985 }
bfff6873
LC		 4986
4987 /*
4988 * Reinitialize lazy itable initialization thread based on
4989 * current settings
4990 */
4991 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4992 ext4_unregister_li_request(sb);
4993 else {
4994 ext4_group_t first_not_zeroed;
4995 first_not_zeroed = ext4_has_uninit_itable(sb);
4996 ext4_register_li_request(sb, first_not_zeroed);
4997 }
4998
6fd058f7 4999 ext4_setup_system_zone(sb);
d096ad0f 5000 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 5001 ext4_commit_super(sb, 1);
0390131b 5002
ac27a0ec
DK		 5003#ifdef CONFIG_QUOTA
5004 /* Release old quota file names */
a2d4a646 5005 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5006 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5007 if (enable_quota) {
5008 if (sb_any_quota_suspended(sb))
5009 dquot_resume(sb, -1);
5010 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5011 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5012 err = ext4_enable_quotas(sb);
07724f98 5013 if (err)
7c319d32 5014 goto restore_opts;
7c319d32
AK		 5015 }
5016 }
ac27a0ec 5017#endif
d4c402d9
CW		 5018
5019 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5020 kfree(orig_data);
ac27a0ec 5021 return 0;
0b8e58a1 5022
ac27a0ec
DK		 5023restore_opts:
5024 sb->s_flags = old_sb_flags;
5025 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5026 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5027 sbi->s_resuid = old_opts.s_resuid;
5028 sbi->s_resgid = old_opts.s_resgid;
5029 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5030 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5031 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5032#ifdef CONFIG_QUOTA
5033 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5034 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5035 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5036 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5037 }
5038#endif
d4c402d9 5039 kfree(orig_data);
ac27a0ec
DK		 5040 return err;
5041}
5042
2b2d6d01 5043static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5044{
5045 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5046 struct ext4_sb_info *sbi = EXT4_SB(sb);
5047 struct ext4_super_block *es = sbi->s_es;
27dd4385 5048 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5049 u64 fsid;
d02a9391 5050 s64 bfree;
27dd4385 5051 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5052
952fc18e
TT		 5053 if (!test_opt(sb, MINIX_DF))
5054 overhead = sbi->s_overhead;
ac27a0ec 5055
617ba13b 5056 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5057 buf->f_bsize = sb->s_blocksize;
b72f78cb 5058 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5059 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5060 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5061 /* prevent underflow in case that few free space is available */
57042651 5062 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5063 buf->f_bavail = buf->f_bfree -
5064 (ext4_r_blocks_count(es) + resv_blocks);
5065 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5066 buf->f_bavail = 0;
5067 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5068 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5069 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5070 fsid = le64_to_cpup((void *)es->s_uuid) ^
5071 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5072 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5073 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5074
ac27a0ec
DK		 5075 return 0;
5076}
5077
0b8e58a1
AD		 5078/* Helper function for writing quotas on sync - we need to start transaction
5079 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 5080 * Process 1 Process 2
617ba13b 5081 * ext4_create() quota_sync()
a269eb18 5082 * jbd2_journal_start() write_dquot()
871a2931 5083 * dquot_initialize() down(dqio_mutex)
dab291af 5084 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 5085 *
5086 */
5087
5088#ifdef CONFIG_QUOTA
5089
5090static inline struct inode *dquot_to_inode(struct dquot *dquot)
5091{
4c376dca 5092 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5093}
5094
617ba13b 5095static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5096{
5097 int ret, err;
5098 handle_t *handle;
5099 struct inode *inode;
5100
5101 inode = dquot_to_inode(dquot);
9924a92a 5102 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5103 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5104 if (IS_ERR(handle))
5105 return PTR_ERR(handle);
5106 ret = dquot_commit(dquot);
617ba13b 5107 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5108 if (!ret)
5109 ret = err;
5110 return ret;
5111}
5112
617ba13b 5113static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5114{
5115 int ret, err;
5116 handle_t *handle;
5117
9924a92a 5118 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5119 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5120 if (IS_ERR(handle))
5121 return PTR_ERR(handle);
5122 ret = dquot_acquire(dquot);
617ba13b 5123 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5124 if (!ret)
5125 ret = err;
5126 return ret;
5127}
5128
617ba13b 5129static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5130{
5131 int ret, err;
5132 handle_t *handle;
5133
9924a92a 5134 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5135 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5136 if (IS_ERR(handle)) {
5137 /* Release dquot anyway to avoid endless cycle in dqput() */
5138 dquot_release(dquot);
ac27a0ec 5139 return PTR_ERR(handle);
9c3013e9 5140 }
ac27a0ec 5141 ret = dquot_release(dquot);
617ba13b 5142 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5143 if (!ret)
5144 ret = err;
5145 return ret;
5146}
5147
617ba13b 5148static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5149{
262b4662
JK		 5150 struct super_block *sb = dquot->dq_sb;
5151 struct ext4_sb_info *sbi = EXT4_SB(sb);
5152
2c8be6b2 5153 /* Are we journaling quotas? */
262b4662
JK		 5154 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5155 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5156 dquot_mark_dquot_dirty(dquot);
617ba13b 5157 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5158 } else {
5159 return dquot_mark_dquot_dirty(dquot);
5160 }
5161}
5162
617ba13b 5163static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5164{
5165 int ret, err;
5166 handle_t *handle;
5167
5168 /* Data block + inode block */
9924a92a 5169 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5170 if (IS_ERR(handle))
5171 return PTR_ERR(handle);
5172 ret = dquot_commit_info(sb, type);
617ba13b 5173 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5174 if (!ret)
5175 ret = err;
5176 return ret;
5177}
5178
5179/*
5180 * Turn on quotas during mount time - we need to find
5181 * the quota file and such...
5182 */
617ba13b 5183static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5184{
287a8095
CH		 5185 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5186 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5187}
5188
5189/*
5190 * Standard function to be called on quota_on
5191 */
617ba13b 5192static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5193 struct path *path)
ac27a0ec
DK		 5194{
5195 int err;
ac27a0ec
DK		 5196
5197 if (!test_opt(sb, QUOTA))
5198 return -EINVAL;
0623543b 5199
ac27a0ec 5200 /* Quotafile not on the same filesystem? */
d8c9584e 5201 if (path->dentry->d_sb != sb)
ac27a0ec 5202 return -EXDEV;
0623543b
JK		 5203 /* Journaling quota? */
5204 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5205 /* Quotafile not in fs root? */
f00c9e44 5206 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5207 ext4_msg(sb, KERN_WARNING,
5208 "Quota file not on filesystem root. "
5209 "Journaled quota will not work");
2b2d6d01 5210 }
0623543b
JK		 5211
5212 /*
5213 * When we journal data on quota file, we have to flush journal to see
5214 * all updates to the file when we bypass pagecache...
5215 */
0390131b 5216 if (EXT4_SB(sb)->s_journal &&
f00c9e44 5217 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 5218 /*
5219 * We don't need to lock updates but journal_flush() could
5220 * otherwise be livelocked...
5221 */
5222 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5223 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5224 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5225 if (err)
7ffe1ea8 5226 return err;
0623543b
JK		 5227 }
5228
f00c9e44 5229 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 5230}
5231
7c319d32
AK		 5232static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5233 unsigned int flags)
5234{
5235 int err;
5236 struct inode *qf_inode;
a2d4a646 5237 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5238 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5239 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5240 };
5241
5242 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5243
5244 if (!qf_inums[type])
5245 return -EPERM;
5246
5247 qf_inode = ext4_iget(sb, qf_inums[type]);
5248 if (IS_ERR(qf_inode)) {
5249 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5250 return PTR_ERR(qf_inode);
5251 }
5252
bcb13850
JK		 5253 /* Don't account quota for quota files to avoid recursion */
5254 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5255 err = dquot_enable(qf_inode, type, format_id, flags);
5256 iput(qf_inode);
5257
5258 return err;
5259}
5260
5261/* Enable usage tracking for all quota types. */
5262static int ext4_enable_quotas(struct super_block *sb)
5263{
5264 int type, err = 0;
a2d4a646 5265 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5266 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5267 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5268 };
5269
5270 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5271 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5272 if (qf_inums[type]) {
5273 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5274 DQUOT_USAGE_ENABLED);
5275 if (err) {
5276 ext4_warning(sb,
72ba7450
TT		 5277 "Failed to enable quota tracking "
5278 "(type=%d, err=%d). Please run "
5279 "e2fsck to fix.", type, err);
7c319d32
AK		 5280 return err;
5281 }
5282 }
5283 }
5284 return 0;
5285}
5286
5287/*
5288 * quota_on function that is used when QUOTA feature is set.
5289 */
5290static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5291 int format_id)
5292{
5293 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5294 return -EINVAL;
5295
5296 /*
5297 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5298 */
5299 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5300}
5301
ca0e05e4
DM		 5302static int ext4_quota_off(struct super_block *sb, int type)
5303{
21f97697
JK		 5304 struct inode *inode = sb_dqopt(sb)->files[type];
5305 handle_t *handle;
5306
87009d86
DM		 5307 /* Force all delayed allocation blocks to be allocated.
5308 * Caller already holds s_umount sem */
5309 if (test_opt(sb, DELALLOC))
ca0e05e4 5310 sync_filesystem(sb);
ca0e05e4 5311
0b268590
AG		 5312 if (!inode)
5313 goto out;
5314
21f97697
JK		 5315 /* Update modification times of quota files when userspace can
5316 * start looking at them */
9924a92a 5317 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5318 if (IS_ERR(handle))
5319 goto out;
5320 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5321 ext4_mark_inode_dirty(handle, inode);
5322 ext4_journal_stop(handle);
5323
5324out:
ca0e05e4
DM		 5325 return dquot_quota_off(sb, type);
5326}
5327
7c319d32
AK		 5328/*
5329 * quota_off function that is used when QUOTA feature is set.
5330 */
5331static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5332{
5333 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5334 return -EINVAL;
5335
5336 /* Disable only the limits. */
5337 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5338}
5339
ac27a0ec
DK		 5340/* Read data from quotafile - avoid pagecache and such because we cannot afford
5341 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5342 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5343 * we don't have to be afraid of races */
617ba13b 5344static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5345 size_t len, loff_t off)
5346{
5347 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5348 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5349 int offset = off & (sb->s_blocksize - 1);
5350 int tocopy;
5351 size_t toread;
5352 struct buffer_head *bh;
5353 loff_t i_size = i_size_read(inode);
5354
5355 if (off > i_size)
5356 return 0;
5357 if (off+len > i_size)
5358 len = i_size-off;
5359 toread = len;
5360 while (toread > 0) {
5361 tocopy = sb->s_blocksize - offset < toread ?
5362 sb->s_blocksize - offset : toread;
1c215028
TT		 5363 bh = ext4_bread(NULL, inode, blk, 0);
5364 if (IS_ERR(bh))
5365 return PTR_ERR(bh);
ac27a0ec
DK		 5366 if (!bh) /* A hole? */
5367 memset(data, 0, tocopy);
5368 else
5369 memcpy(data, bh->b_data+offset, tocopy);
5370 brelse(bh);
5371 offset = 0;
5372 toread -= tocopy;
5373 data += tocopy;
5374 blk++;
5375 }
5376 return len;
5377}
5378
5379/* Write to quotafile (we know the transaction is already started and has
5380 * enough credits) */
617ba13b 5381static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5382 const char *data, size_t len, loff_t off)
5383{
5384 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5385 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5386 int err, offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5387 struct buffer_head *bh;
5388 handle_t *handle = journal_current_handle();
5389
0390131b 5390 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5391 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5392 " cancelled because transaction is not started",
9c3013e9
JK		 5393 (unsigned long long)off, (unsigned long long)len);
5394 return -EIO;
5395 }
67eeb568
DM		 5396 /*
5397 * Since we account only one data block in transaction credits,
5398 * then it is impossible to cross a block boundary.
5399 */
5400 if (sb->s_blocksize - offset < len) {
5401 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5402 " cancelled because not block aligned",
5403 (unsigned long long)off, (unsigned long long)len);
5404 return -EIO;
5405 }
5406
1c215028
TT		 5407 bh = ext4_bread(handle, inode, blk, 1);
5408 if (IS_ERR(bh))
5409 return PTR_ERR(bh);
67eeb568
DM		 5410 if (!bh)
5411 goto out;
5d601255 5412 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5413 err = ext4_journal_get_write_access(handle, bh);
5414 if (err) {
5415 brelse(bh);
1c215028 5416 return err;
ac27a0ec 5417 }
67eeb568
DM		 5418 lock_buffer(bh);
5419 memcpy(bh->b_data+offset, data, len);
5420 flush_dcache_page(bh->b_page);
5421 unlock_buffer(bh);
62d2b5f2 5422 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5423 brelse(bh);
ac27a0ec 5424out:
67eeb568
DM		 5425 if (inode->i_size < off + len) {
5426 i_size_write(inode, off + len);
617ba13b 5427 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5428 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5429 }
67eeb568 5430 return len;
ac27a0ec
DK		 5431}
5432
5433#endif
5434
152a0836
AV		 5435static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5436 const char *dev_name, void *data)
ac27a0ec 5437{
152a0836 5438 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5439}
5440
37f328eb 5441#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5442static inline void register_as_ext2(void)
5443{
5444 int err = register_filesystem(&ext2_fs_type);
5445 if (err)
5446 printk(KERN_WARNING
5447 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5448}
5449
5450static inline void unregister_as_ext2(void)
5451{
5452 unregister_filesystem(&ext2_fs_type);
5453}
2035e776
TT		 5454
5455static inline int ext2_feature_set_ok(struct super_block *sb)
5456{
5457 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5458 return 0;
5459 if (sb->s_flags & MS_RDONLY)
5460 return 1;
5461 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5462 return 0;
5463 return 1;
5464}
24b58424
TT		 5465#else
5466static inline void register_as_ext2(void) { }
5467static inline void unregister_as_ext2(void) { }
2035e776 5468static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5469#endif
5470
37f328eb 5471#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5472static inline void register_as_ext3(void)
5473{
5474 int err = register_filesystem(&ext3_fs_type);
5475 if (err)
5476 printk(KERN_WARNING
5477 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5478}
5479
5480static inline void unregister_as_ext3(void)
5481{
5482 unregister_filesystem(&ext3_fs_type);
5483}
2035e776
TT		 5484
5485static inline int ext3_feature_set_ok(struct super_block *sb)
5486{
5487 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5488 return 0;
5489 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5490 return 0;
5491 if (sb->s_flags & MS_RDONLY)
5492 return 1;
5493 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5494 return 0;
5495 return 1;
5496}
24b58424
TT		 5497#else
5498static inline void register_as_ext3(void) { }
5499static inline void unregister_as_ext3(void) { }
2035e776 5500static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5501#endif
5502
03010a33
TT		 5503static struct file_system_type ext4_fs_type = {
5504 .owner = THIS_MODULE,
5505 .name = "ext4",
152a0836 5506 .mount = ext4_mount,
03010a33
TT		 5507 .kill_sb = kill_block_super,
5508 .fs_flags = FS_REQUIRES_DEV,
5509};
7f78e035 5510MODULE_ALIAS_FS("ext4");
03010a33 5511
8f021222 5512static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5513{
5514 struct ext4_features *ef;
5515 int ret = -ENOMEM;
5516
5517 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5518 if (!ef)
5519 goto out;
5520
5521 ef->f_kobj.kset = ext4_kset;
5522 init_completion(&ef->f_kobj_unregister);
5523 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5524 "features");
5525 if (ret) {
5526 kfree(ef);
5527 goto out;
5528 }
5529
5530 ext4_feat = ef;
5531 ret = 0;
5532out:
5533 return ret;
5534}
5535
8f021222
LC		 5536static void ext4_exit_feat_adverts(void)
5537{
5538 kobject_put(&ext4_feat->f_kobj);
5539 wait_for_completion(&ext4_feat->f_kobj_unregister);
5540 kfree(ext4_feat);
5541}
5542
e9e3bcec
ES		 5543/* Shared across all ext4 file systems */
5544wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5545struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5546
5dabfc78 5547static int __init ext4_init_fs(void)
ac27a0ec 5548{
e9e3bcec 5549 int i, err;
c9de560d 5550
07c0c5d8
AV		 5551 ext4_li_info = NULL;
5552 mutex_init(&ext4_li_mtx);
5553
9a4c8019 5554 /* Build-time check for flags consistency */
12e9b892 5555 ext4_check_flag_values();
e9e3bcec
ES		 5556
5557 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5558 mutex_init(&ext4__aio_mutex[i]);
5559 init_waitqueue_head(&ext4__ioend_wq[i]);
5560 }
5561
51865fda 5562 err = ext4_init_es();
6fd058f7
TT		 5563 if (err)
5564 return err;
51865fda
ZL		 5565
5566 err = ext4_init_pageio();
5567 if (err)
5568 goto out7;
5569
5dabfc78 5570 err = ext4_init_system_zone();
bd2d0210 5571 if (err)
d44651d0 5572 goto out6;
3197ebdb 5573 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5574 if (!ext4_kset) {
5575 err = -ENOMEM;
dd68314c 5576 goto out5;
0e376b1e 5577 }
d44651d0 5578 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5579
5580 err = ext4_init_feat_adverts();
dd68314c
TT		 5581 if (err)
5582 goto out4;
857ac889 5583
5dabfc78 5584 err = ext4_init_mballoc();
c9de560d
AT		 5585 if (err)
5586 goto out2;
9c191f70
M		 5587 else
5588 ext4_mballoc_ready = 1;
ac27a0ec
DK		 5589 err = init_inodecache();
5590 if (err)
5591 goto out1;
24b58424 5592 register_as_ext3();
2035e776 5593 register_as_ext2();
03010a33 5594 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5595 if (err)
5596 goto out;
bfff6873 5597
ac27a0ec
DK		 5598 return 0;
5599out:
24b58424
TT		 5600 unregister_as_ext2();
5601 unregister_as_ext3();
ac27a0ec
DK		 5602 destroy_inodecache();
5603out1:
9c191f70 5604 ext4_mballoc_ready = 0;
5dabfc78 5605 ext4_exit_mballoc();
9c191f70 5606out2:
8f021222 5607 ext4_exit_feat_adverts();
dd68314c 5608out4:
d44651d0
FJ		 5609 if (ext4_proc_root)
5610 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5611 kset_unregister(ext4_kset);
d44651d0 5612out5:
5dabfc78 5613 ext4_exit_system_zone();
d44651d0 5614out6:
5dabfc78 5615 ext4_exit_pageio();
51865fda
ZL		 5616out7:
5617 ext4_exit_es();
5618
ac27a0ec
DK		 5619 return err;
5620}
5621
5dabfc78 5622static void __exit ext4_exit_fs(void)
ac27a0ec 5623{
bfff6873 5624 ext4_destroy_lazyinit_thread();
24b58424
TT		 5625 unregister_as_ext2();
5626 unregister_as_ext3();
03010a33 5627 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5628 destroy_inodecache();
5dabfc78 5629 ext4_exit_mballoc();
8f021222 5630 ext4_exit_feat_adverts();
9f6200bb 5631 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5632 kset_unregister(ext4_kset);
5dabfc78
TT		 5633 ext4_exit_system_zone();
5634 ext4_exit_pageio();
dd12ed14 5635 ext4_exit_es();
ac27a0ec
DK		 5636}
5637
5638MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5639MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5640MODULE_LICENSE("GPL");
5dabfc78
TT		 5641module_init(ext4_init_fs)
5642module_exit(ext4_exit_fs)