projects / GitHub / LineageOS / android_kernel_motorola_exynos9610.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: cache extent hole in extent status tree for ext4_da_map_blocks()
[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);
74cd15cd 874 INIT_LIST_HEAD(&ei->i_es_lru);
eb68d0e2 875 ei->i_es_all_nr = 0;
74cd15cd 876 ei->i_es_lru_nr = 0;
d3922a77 877 ei->i_touch_when = 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;
886#endif
8aefcd55 887 ei->jinode = NULL;
2e8fa54e 888 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 889 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 890 ei->i_sync_tid = 0;
891 ei->i_datasync_tid = 0;
f7ad6d2e 892 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 893 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 894 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
0b8e58a1 895
ac27a0ec
DK		 896 return &ei->vfs_inode;
897}
898
7ff9c073
TT		 899static int ext4_drop_inode(struct inode *inode)
900{
901 int drop = generic_drop_inode(inode);
902
903 trace_ext4_drop_inode(inode, drop);
904 return drop;
905}
906
fa0d7e3d
NP		 907static void ext4_i_callback(struct rcu_head *head)
908{
909 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 910 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
911}
912
617ba13b 913static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 914{
9f7dd93d 915 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 916 ext4_msg(inode->i_sb, KERN_ERR,
917 "Inode %lu (%p): orphan list check failed!",
918 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 919 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
920 EXT4_I(inode), sizeof(struct ext4_inode_info),
921 true);
922 dump_stack();
923 }
fa0d7e3d 924 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 925}
926
51cc5068 927static void init_once(void *foo)
ac27a0ec 928{
617ba13b 929 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 930
a35afb83 931 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 932 init_rwsem(&ei->xattr_sem);
0e855ac8 933 init_rwsem(&ei->i_data_sem);
a35afb83 934 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 935}
936
e67bc2b3 937static int __init init_inodecache(void)
ac27a0ec 938{
617ba13b
MC		 939 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
940 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 941 0, (SLAB_RECLAIM_ACCOUNT|
942 SLAB_MEM_SPREAD),
20c2df83 943 init_once);
617ba13b 944 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 945 return -ENOMEM;
946 return 0;
947}
948
949static void destroy_inodecache(void)
950{
8c0a8537
KS		 951 /*
952 * Make sure all delayed rcu free inodes are flushed before we
953 * destroy cache.
954 */
955 rcu_barrier();
617ba13b 956 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 957}
958
0930fcc1 959void ext4_clear_inode(struct inode *inode)
ac27a0ec 960{
0930fcc1 961 invalidate_inode_buffers(inode);
dbd5768f 962 clear_inode(inode);
9f754758 963 dquot_drop(inode);
c2ea3fde 964 ext4_discard_preallocations(inode);
51865fda 965 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
74cd15cd 966 ext4_es_lru_del(inode);
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
61e225dc 1062static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1063 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1064 .write_dquot = ext4_write_dquot,
1065 .acquire_dquot = ext4_acquire_dquot,
1066 .release_dquot = ext4_release_dquot,
1067 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1068 .write_info = ext4_write_info,
1069 .alloc_dquot = dquot_alloc,
1070 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1071};
1072
0d54b217 1073static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1074 .quota_on = ext4_quota_on,
ca0e05e4 1075 .quota_off = ext4_quota_off,
287a8095
CH		 1076 .quota_sync = dquot_quota_sync,
1077 .get_info = dquot_get_dqinfo,
1078 .set_info = dquot_set_dqinfo,
1079 .get_dqblk = dquot_get_dqblk,
1080 .set_dqblk = dquot_set_dqblk
ac27a0ec 1081};
7c319d32
AK		 1082
1083static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1084 .quota_on_meta = ext4_quota_on_sysfile,
1085 .quota_off = ext4_quota_off_sysfile,
1086 .quota_sync = dquot_quota_sync,
1087 .get_info = dquot_get_dqinfo,
1088 .set_info = dquot_set_dqinfo,
1089 .get_dqblk = dquot_get_dqblk,
1090 .set_dqblk = dquot_set_dqblk
1091};
ac27a0ec
DK		 1092#endif
1093
ee9b6d61 1094static const struct super_operations ext4_sops = {
617ba13b
MC		 1095 .alloc_inode = ext4_alloc_inode,
1096 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1097 .write_inode = ext4_write_inode,
1098 .dirty_inode = ext4_dirty_inode,
7ff9c073 1099 .drop_inode = ext4_drop_inode,
0930fcc1 1100 .evict_inode = ext4_evict_inode,
617ba13b 1101 .put_super = ext4_put_super,
617ba13b 1102 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1103 .freeze_fs = ext4_freeze,
1104 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1105 .statfs = ext4_statfs,
1106 .remount_fs = ext4_remount,
617ba13b 1107 .show_options = ext4_show_options,
ac27a0ec 1108#ifdef CONFIG_QUOTA
617ba13b
MC		 1109 .quota_read = ext4_quota_read,
1110 .quota_write = ext4_quota_write,
ac27a0ec 1111#endif
c39a7f84 1112 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1113};
1114
39655164 1115static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1116 .fh_to_dentry = ext4_fh_to_dentry,
1117 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1118 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1119};
1120
1121enum {
1122 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1123 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1124 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1125 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1126 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1127 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1128 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1129 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1130 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1131 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1132 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1133 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1134 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1135 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1136 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1137 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1138 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1139 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
df981d03 1140 Opt_max_dir_size_kb,
ac27a0ec
DK		 1141};
1142
a447c093 1143static const match_table_t tokens = {
ac27a0ec
DK		 1144 {Opt_bsd_df, "bsddf"},
1145 {Opt_minix_df, "minixdf"},
1146 {Opt_grpid, "grpid"},
1147 {Opt_grpid, "bsdgroups"},
1148 {Opt_nogrpid, "nogrpid"},
1149 {Opt_nogrpid, "sysvgroups"},
1150 {Opt_resgid, "resgid=%u"},
1151 {Opt_resuid, "resuid=%u"},
1152 {Opt_sb, "sb=%u"},
1153 {Opt_err_cont, "errors=continue"},
1154 {Opt_err_panic, "errors=panic"},
1155 {Opt_err_ro, "errors=remount-ro"},
1156 {Opt_nouid32, "nouid32"},
ac27a0ec 1157 {Opt_debug, "debug"},
72578c33
TT		 1158 {Opt_removed, "oldalloc"},
1159 {Opt_removed, "orlov"},
ac27a0ec
DK		 1160 {Opt_user_xattr, "user_xattr"},
1161 {Opt_nouser_xattr, "nouser_xattr"},
1162 {Opt_acl, "acl"},
1163 {Opt_noacl, "noacl"},
e3bb52ae 1164 {Opt_noload, "norecovery"},
5a916be1 1165 {Opt_noload, "noload"},
72578c33
TT		 1166 {Opt_removed, "nobh"},
1167 {Opt_removed, "bh"},
ac27a0ec 1168 {Opt_commit, "commit=%u"},
30773840
TT		 1169 {Opt_min_batch_time, "min_batch_time=%u"},
1170 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1171 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1172 {Opt_journal_path, "journal_path=%s"},
818d276c
GS		 1173 {Opt_journal_checksum, "journal_checksum"},
1174 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1175 {Opt_abort, "abort"},
1176 {Opt_data_journal, "data=journal"},
1177 {Opt_data_ordered, "data=ordered"},
1178 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1179 {Opt_data_err_abort, "data_err=abort"},
1180 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1181 {Opt_offusrjquota, "usrjquota="},
1182 {Opt_usrjquota, "usrjquota=%s"},
1183 {Opt_offgrpjquota, "grpjquota="},
1184 {Opt_grpjquota, "grpjquota=%s"},
1185 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1186 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1187 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1188 {Opt_grpquota, "grpquota"},
1189 {Opt_noquota, "noquota"},
1190 {Opt_quota, "quota"},
1191 {Opt_usrquota, "usrquota"},
1192 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1193 {Opt_barrier, "barrier"},
1194 {Opt_nobarrier, "nobarrier"},
25ec56b5 1195 {Opt_i_version, "i_version"},
c9de560d 1196 {Opt_stripe, "stripe=%u"},
64769240 1197 {Opt_delalloc, "delalloc"},
dd919b98 1198 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1199 {Opt_removed, "mblk_io_submit"},
1200 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1201 {Opt_block_validity, "block_validity"},
1202 {Opt_noblock_validity, "noblock_validity"},
240799cd 1203 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1204 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1205 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1206 {Opt_auto_da_alloc, "auto_da_alloc"},
1207 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1208 {Opt_dioread_nolock, "dioread_nolock"},
1209 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1210 {Opt_discard, "discard"},
1211 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1212 {Opt_init_itable, "init_itable=%u"},
1213 {Opt_init_itable, "init_itable"},
1214 {Opt_noinit_itable, "noinit_itable"},
df981d03 1215 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1216 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1217 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1218 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1219 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1220 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1221 {Opt_err, NULL},
ac27a0ec
DK		 1222};
1223
617ba13b 1224static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1225{
617ba13b 1226 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1227 char *options = (char *) *data;
1228
1229 if (!options || strncmp(options, "sb=", 3) != 0)
1230 return 1; /* Default location */
0b8e58a1 1231
ac27a0ec 1232 options += 3;
0b8e58a1 1233 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1234 sb_block = simple_strtoul(options, &options, 0);
1235 if (*options && *options != ',') {
4776004f 1236 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1237 (char *) *data);
1238 return 1;
1239 }
1240 if (*options == ',')
1241 options++;
1242 *data = (void *) options;
0b8e58a1 1243
ac27a0ec
DK		 1244 return sb_block;
1245}
1246
b3881f74 1247#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1248static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1249 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1250
56c50f11
DM		 1251#ifdef CONFIG_QUOTA
1252static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1253{
1254 struct ext4_sb_info *sbi = EXT4_SB(sb);
1255 char *qname;
03dafb5f 1256 int ret = -1;
56c50f11
DM		 1257
1258 if (sb_any_quota_loaded(sb) &&
1259 !sbi->s_qf_names[qtype]) {
1260 ext4_msg(sb, KERN_ERR,
1261 "Cannot change journaled "
1262 "quota options when quota turned on");
57f73c2c 1263 return -1;
56c50f11 1264 }
262b4662
JK		 1265 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1266 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1267 "when QUOTA feature is enabled");
1268 return -1;
1269 }
56c50f11
DM		 1270 qname = match_strdup(args);
1271 if (!qname) {
1272 ext4_msg(sb, KERN_ERR,
1273 "Not enough memory for storing quotafile name");
57f73c2c 1274 return -1;
56c50f11 1275 }
03dafb5f
CG		 1276 if (sbi->s_qf_names[qtype]) {
1277 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1278 ret = 1;
1279 else
1280 ext4_msg(sb, KERN_ERR,
1281 "%s quota file already specified",
1282 QTYPE2NAME(qtype));
1283 goto errout;
56c50f11 1284 }
03dafb5f 1285 if (strchr(qname, '/')) {
56c50f11
DM		 1286 ext4_msg(sb, KERN_ERR,
1287 "quotafile must be on filesystem root");
03dafb5f 1288 goto errout;
56c50f11 1289 }
03dafb5f 1290 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1291 set_opt(sb, QUOTA);
56c50f11 1292 return 1;
03dafb5f
CG		 1293errout:
1294 kfree(qname);
1295 return ret;
56c50f11
DM		 1296}
1297
1298static int clear_qf_name(struct super_block *sb, int qtype)
1299{
1300
1301 struct ext4_sb_info *sbi = EXT4_SB(sb);
1302
1303 if (sb_any_quota_loaded(sb) &&
1304 sbi->s_qf_names[qtype]) {
1305 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1306 " when quota turned on");
57f73c2c 1307 return -1;
56c50f11 1308 }
03dafb5f 1309 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1310 sbi->s_qf_names[qtype] = NULL;
1311 return 1;
1312}
1313#endif
1314
26092bf5
TT		 1315#define MOPT_SET 0x0001
1316#define MOPT_CLEAR 0x0002
1317#define MOPT_NOSUPPORT 0x0004
1318#define MOPT_EXPLICIT 0x0008
1319#define MOPT_CLEAR_ERR 0x0010
1320#define MOPT_GTE0 0x0020
ac27a0ec 1321#ifdef CONFIG_QUOTA
26092bf5
TT		 1322#define MOPT_Q 0
1323#define MOPT_QFMT 0x0040
1324#else
1325#define MOPT_Q MOPT_NOSUPPORT
1326#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1327#endif
26092bf5 1328#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1329#define MOPT_NO_EXT2 0x0100
1330#define MOPT_NO_EXT3 0x0200
1331#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1332#define MOPT_STRING 0x0400
26092bf5
TT		 1333
1334static const struct mount_opts {
1335 int token;
1336 int mount_opt;
1337 int flags;
1338} ext4_mount_opts[] = {
1339 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1340 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1341 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1342 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1343 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1344 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1345 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1346 MOPT_EXT4_ONLY | MOPT_SET},
1347 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1348 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1349 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1350 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1351 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1352 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1353 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1354 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c
TT		 1355 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1356 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1357 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1358 EXT4_MOUNT_JOURNAL_CHECKSUM),
1359 MOPT_EXT4_ONLY | MOPT_SET},
1360 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1361 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1362 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1363 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1364 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1365 MOPT_NO_EXT2 | MOPT_SET},
1366 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1367 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1368 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1369 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1370 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1371 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1372 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1373 {Opt_commit, 0, MOPT_GTE0},
1374 {Opt_max_batch_time, 0, MOPT_GTE0},
1375 {Opt_min_batch_time, 0, MOPT_GTE0},
1376 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1377 {Opt_init_itable, 0, MOPT_GTE0},
1378 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1379 {Opt_resuid, 0, MOPT_GTE0},
1380 {Opt_resgid, 0, MOPT_GTE0},
1381 {Opt_journal_dev, 0, MOPT_GTE0},
ad4eec61 1382 {Opt_journal_path, 0, MOPT_STRING},
0efb3b23 1383 {Opt_journal_ioprio, 0, MOPT_GTE0},
8dc0aa8c
TT		 1384 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1385 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1386 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1387 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1388 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1389 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1390#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1391 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1392 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1393#else
26092bf5
TT		 1394 {Opt_acl, 0, MOPT_NOSUPPORT},
1395 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1396#endif
26092bf5
TT		 1397 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1398 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1399 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1400 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1401 MOPT_SET | MOPT_Q},
1402 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1403 MOPT_SET | MOPT_Q},
1404 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1405 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1406 {Opt_usrjquota, 0, MOPT_Q},
1407 {Opt_grpjquota, 0, MOPT_Q},
1408 {Opt_offusrjquota, 0, MOPT_Q},
1409 {Opt_offgrpjquota, 0, MOPT_Q},
1410 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1411 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1412 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1413 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1414 {Opt_err, 0, 0}
1415};
1416
1417static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1418 substring_t *args, unsigned long *journal_devnum,
1419 unsigned int *journal_ioprio, int is_remount)
1420{
1421 struct ext4_sb_info *sbi = EXT4_SB(sb);
1422 const struct mount_opts *m;
08cefc7a
EB		 1423 kuid_t uid;
1424 kgid_t gid;
26092bf5
TT		 1425 int arg = 0;
1426
57f73c2c
TT		 1427#ifdef CONFIG_QUOTA
1428 if (token == Opt_usrjquota)
1429 return set_qf_name(sb, USRQUOTA, &args[0]);
1430 else if (token == Opt_grpjquota)
1431 return set_qf_name(sb, GRPQUOTA, &args[0]);
1432 else if (token == Opt_offusrjquota)
1433 return clear_qf_name(sb, USRQUOTA);
1434 else if (token == Opt_offgrpjquota)
1435 return clear_qf_name(sb, GRPQUOTA);
1436#endif
26092bf5 1437 switch (token) {
f7048605
TT		 1438 case Opt_noacl:
1439 case Opt_nouser_xattr:
1440 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1441 break;
26092bf5
TT		 1442 case Opt_sb:
1443 return 1; /* handled by get_sb_block() */
1444 case Opt_removed:
5f3633e3 1445 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1446 return 1;
26092bf5
TT		 1447 case Opt_abort:
1448 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1449 return 1;
1450 case Opt_i_version:
1451 sb->s_flags |= MS_I_VERSION;
1452 return 1;
26092bf5
TT		 1453 }
1454
5f3633e3
JK		 1455 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1456 if (token == m->token)
1457 break;
1458
1459 if (m->token == Opt_err) {
1460 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1461 "or missing value", opt);
1462 return -1;
1463 }
1464
8dc0aa8c
TT		 1465 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1466 ext4_msg(sb, KERN_ERR,
1467 "Mount option \"%s\" incompatible with ext2", opt);
1468 return -1;
1469 }
1470 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1471 ext4_msg(sb, KERN_ERR,
1472 "Mount option \"%s\" incompatible with ext3", opt);
1473 return -1;
1474 }
1475
ad4eec61 1476 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1477 return -1;
1478 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1479 return -1;
1480 if (m->flags & MOPT_EXPLICIT)
1481 set_opt2(sb, EXPLICIT_DELALLOC);
1482 if (m->flags & MOPT_CLEAR_ERR)
1483 clear_opt(sb, ERRORS_MASK);
1484 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1485 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1486 "options when quota turned on");
1487 return -1;
1488 }
1489
1490 if (m->flags & MOPT_NOSUPPORT) {
1491 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1492 } else if (token == Opt_commit) {
1493 if (arg == 0)
1494 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1495 sbi->s_commit_interval = HZ * arg;
1496 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1497 sbi->s_max_batch_time = arg;
1498 } else if (token == Opt_min_batch_time) {
1499 sbi->s_min_batch_time = arg;
1500 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1501 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1502 ext4_msg(sb, KERN_ERR,
1503 "EXT4-fs: inode_readahead_blks must be "
1504 "0 or a power of 2 smaller than 2^31");
26092bf5 1505 return -1;
5f3633e3
JK		 1506 }
1507 sbi->s_inode_readahead_blks = arg;
1508 } else if (token == Opt_init_itable) {
1509 set_opt(sb, INIT_INODE_TABLE);
1510 if (!args->from)
1511 arg = EXT4_DEF_LI_WAIT_MULT;
1512 sbi->s_li_wait_mult = arg;
1513 } else if (token == Opt_max_dir_size_kb) {
1514 sbi->s_max_dir_size_kb = arg;
1515 } else if (token == Opt_stripe) {
1516 sbi->s_stripe = arg;
1517 } else if (token == Opt_resuid) {
1518 uid = make_kuid(current_user_ns(), arg);
1519 if (!uid_valid(uid)) {
1520 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1521 return -1;
1522 }
5f3633e3
JK		 1523 sbi->s_resuid = uid;
1524 } else if (token == Opt_resgid) {
1525 gid = make_kgid(current_user_ns(), arg);
1526 if (!gid_valid(gid)) {
1527 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1528 return -1;
1529 }
1530 sbi->s_resgid = gid;
1531 } else if (token == Opt_journal_dev) {
1532 if (is_remount) {
1533 ext4_msg(sb, KERN_ERR,
1534 "Cannot specify journal on remount");
1535 return -1;
1536 }
1537 *journal_devnum = arg;
ad4eec61
ES		 1538 } else if (token == Opt_journal_path) {
1539 char *journal_path;
1540 struct inode *journal_inode;
1541 struct path path;
1542 int error;
1543
1544 if (is_remount) {
1545 ext4_msg(sb, KERN_ERR,
1546 "Cannot specify journal on remount");
1547 return -1;
1548 }
1549 journal_path = match_strdup(&args[0]);
1550 if (!journal_path) {
1551 ext4_msg(sb, KERN_ERR, "error: could not dup "
1552 "journal device string");
1553 return -1;
1554 }
1555
1556 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1557 if (error) {
1558 ext4_msg(sb, KERN_ERR, "error: could not find "
1559 "journal device path: error %d", error);
1560 kfree(journal_path);
1561 return -1;
1562 }
1563
1564 journal_inode = path.dentry->d_inode;
1565 if (!S_ISBLK(journal_inode->i_mode)) {
1566 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1567 "is not a block device", journal_path);
1568 path_put(&path);
1569 kfree(journal_path);
1570 return -1;
1571 }
1572
1573 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1574 path_put(&path);
1575 kfree(journal_path);
5f3633e3
JK		 1576 } else if (token == Opt_journal_ioprio) {
1577 if (arg > 7) {
1578 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1579 " (must be 0-7)");
1580 return -1;
1581 }
1582 *journal_ioprio =
1583 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1584 } else if (m->flags & MOPT_DATAJ) {
1585 if (is_remount) {
1586 if (!sbi->s_journal)
1587 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1588 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1589 ext4_msg(sb, KERN_ERR,
26092bf5 1590 "Cannot change data mode on remount");
26092bf5 1591 return -1;
ac27a0ec 1592 }
26092bf5 1593 } else {
5f3633e3
JK		 1594 clear_opt(sb, DATA_FLAGS);
1595 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1596 }
5f3633e3
JK		 1597#ifdef CONFIG_QUOTA
1598 } else if (m->flags & MOPT_QFMT) {
1599 if (sb_any_quota_loaded(sb) &&
1600 sbi->s_jquota_fmt != m->mount_opt) {
1601 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1602 "quota options when quota turned on");
1603 return -1;
1604 }
262b4662
JK		 1605 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1606 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1607 ext4_msg(sb, KERN_ERR,
1608 "Cannot set journaled quota options "
1609 "when QUOTA feature is enabled");
1610 return -1;
1611 }
5f3633e3
JK		 1612 sbi->s_jquota_fmt = m->mount_opt;
1613#endif
1614 } else {
1615 if (!args->from)
1616 arg = 1;
1617 if (m->flags & MOPT_CLEAR)
1618 arg = !arg;
1619 else if (unlikely(!(m->flags & MOPT_SET))) {
1620 ext4_msg(sb, KERN_WARNING,
1621 "buggy handling of option %s", opt);
1622 WARN_ON(1);
1623 return -1;
1624 }
1625 if (arg != 0)
1626 sbi->s_mount_opt |= m->mount_opt;
1627 else
1628 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1629 }
5f3633e3 1630 return 1;
26092bf5
TT		 1631}
1632
1633static int parse_options(char *options, struct super_block *sb,
1634 unsigned long *journal_devnum,
1635 unsigned int *journal_ioprio,
1636 int is_remount)
1637{
1638 struct ext4_sb_info *sbi = EXT4_SB(sb);
1639 char *p;
1640 substring_t args[MAX_OPT_ARGS];
1641 int token;
1642
1643 if (!options)
1644 return 1;
1645
1646 while ((p = strsep(&options, ",")) != NULL) {
1647 if (!*p)
1648 continue;
1649 /*
1650 * Initialize args struct so we know whether arg was
1651 * found; some options take optional arguments.
1652 */
caecd0af 1653 args[0].to = args[0].from = NULL;
26092bf5
TT		 1654 token = match_token(p, tokens, args);
1655 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1656 journal_ioprio, is_remount) < 0)
1657 return 0;
ac27a0ec
DK		 1658 }
1659#ifdef CONFIG_QUOTA
262b4662
JK		 1660 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1661 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1662 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1663 "feature is enabled");
1664 return 0;
1665 }
ac27a0ec 1666 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1667 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1668 clear_opt(sb, USRQUOTA);
ac27a0ec 1669
482a7425 1670 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1671 clear_opt(sb, GRPQUOTA);
ac27a0ec 1672
56c50f11 1673 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1674 ext4_msg(sb, KERN_ERR, "old and new quota "
1675 "format mixing");
ac27a0ec
DK		 1676 return 0;
1677 }
1678
1679 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1680 ext4_msg(sb, KERN_ERR, "journaled quota format "
1681 "not specified");
ac27a0ec
DK		 1682 return 0;
1683 }
ac27a0ec
DK		 1684 }
1685#endif
261cb20c
JK		 1686 if (test_opt(sb, DIOREAD_NOLOCK)) {
1687 int blocksize =
1688 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1689
1690 if (blocksize < PAGE_CACHE_SIZE) {
1691 ext4_msg(sb, KERN_ERR, "can't mount with "
1692 "dioread_nolock if block size != PAGE_SIZE");
1693 return 0;
1694 }
1695 }
ac27a0ec
DK		 1696 return 1;
1697}
1698
2adf6da8
TT		 1699static inline void ext4_show_quota_options(struct seq_file *seq,
1700 struct super_block *sb)
1701{
1702#if defined(CONFIG_QUOTA)
1703 struct ext4_sb_info *sbi = EXT4_SB(sb);
1704
1705 if (sbi->s_jquota_fmt) {
1706 char *fmtname = "";
1707
1708 switch (sbi->s_jquota_fmt) {
1709 case QFMT_VFS_OLD:
1710 fmtname = "vfsold";
1711 break;
1712 case QFMT_VFS_V0:
1713 fmtname = "vfsv0";
1714 break;
1715 case QFMT_VFS_V1:
1716 fmtname = "vfsv1";
1717 break;
1718 }
1719 seq_printf(seq, ",jqfmt=%s", fmtname);
1720 }
1721
1722 if (sbi->s_qf_names[USRQUOTA])
1723 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1724
1725 if (sbi->s_qf_names[GRPQUOTA])
1726 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1727#endif
1728}
1729
5a916be1
TT		 1730static const char *token2str(int token)
1731{
50df9fd5 1732 const struct match_token *t;
5a916be1
TT		 1733
1734 for (t = tokens; t->token != Opt_err; t++)
1735 if (t->token == token && !strchr(t->pattern, '='))
1736 break;
1737 return t->pattern;
1738}
1739
2adf6da8
TT		 1740/*
1741 * Show an option if
1742 * - it's set to a non-default value OR
1743 * - if the per-sb default is different from the global default
1744 */
66acdcf4
TT		 1745static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1746 int nodefs)
2adf6da8 1747{
2adf6da8
TT		 1748 struct ext4_sb_info *sbi = EXT4_SB(sb);
1749 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1750 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1751 const struct mount_opts *m;
66acdcf4 1752 char sep = nodefs ? '\n' : ',';
2adf6da8 1753
66acdcf4
TT		 1754#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1755#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1756
1757 if (sbi->s_sb_block != 1)
5a916be1
TT		 1758 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1759
1760 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1761 int want_set = m->flags & MOPT_SET;
1762 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1763 (m->flags & MOPT_CLEAR_ERR))
1764 continue;
66acdcf4 1765 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1766 continue; /* skip if same as the default */
1767 if ((want_set &&
1768 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1769 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1770 continue; /* select Opt_noFoo vs Opt_Foo */
1771 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1772 }
5a916be1 1773
08cefc7a 1774 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1775 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1776 SEQ_OPTS_PRINT("resuid=%u",
1777 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1778 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1779 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1780 SEQ_OPTS_PRINT("resgid=%u",
1781 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1782 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1783 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1784 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1785 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1786 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1787 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1788 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1789 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1790 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1791 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1792 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1793 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1794 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1795 if (sb->s_flags & MS_I_VERSION)
5a916be1 1796 SEQ_OPTS_PUTS("i_version");
66acdcf4 1797 if (nodefs || sbi->s_stripe)
5a916be1 1798 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1799 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1800 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1801 SEQ_OPTS_PUTS("data=journal");
1802 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1803 SEQ_OPTS_PUTS("data=ordered");
1804 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1805 SEQ_OPTS_PUTS("data=writeback");
1806 }
66acdcf4
TT		 1807 if (nodefs ||
1808 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1809 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1810 sbi->s_inode_readahead_blks);
2adf6da8 1811
66acdcf4
TT		 1812 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1813 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1814 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1815 if (nodefs || sbi->s_max_dir_size_kb)
1816 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1817
1818 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1819 return 0;
1820}
1821
66acdcf4
TT		 1822static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1823{
1824 return _ext4_show_options(seq, root->d_sb, 0);
1825}
1826
1827static int options_seq_show(struct seq_file *seq, void *offset)
1828{
1829 struct super_block *sb = seq->private;
1830 int rc;
1831
1832 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1833 rc = _ext4_show_options(seq, sb, 1);
1834 seq_puts(seq, "\n");
1835 return rc;
1836}
1837
1838static int options_open_fs(struct inode *inode, struct file *file)
1839{
d9dda78b 1840 return single_open(file, options_seq_show, PDE_DATA(inode));
66acdcf4
TT		 1841}
1842
1843static const struct file_operations ext4_seq_options_fops = {
1844 .owner = THIS_MODULE,
1845 .open = options_open_fs,
1846 .read = seq_read,
1847 .llseek = seq_lseek,
1848 .release = single_release,
1849};
1850
617ba13b 1851static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1852 int read_only)
1853{
617ba13b 1854 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1855 int res = 0;
1856
617ba13b 1857 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1858 ext4_msg(sb, KERN_ERR, "revision level too high, "
1859 "forcing read-only mode");
ac27a0ec
DK		 1860 res = MS_RDONLY;
1861 }
1862 if (read_only)
281b5995 1863 goto done;
617ba13b 1864 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1865 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1866 "running e2fsck is recommended");
c8b459f4 1867 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 1868 ext4_msg(sb, KERN_WARNING,
1869 "warning: mounting fs with errors, "
1870 "running e2fsck is recommended");
ed3ce80a 1871 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1872 le16_to_cpu(es->s_mnt_count) >=
1873 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1874 ext4_msg(sb, KERN_WARNING,
1875 "warning: maximal mount count reached, "
1876 "running e2fsck is recommended");
ac27a0ec
DK		 1877 else if (le32_to_cpu(es->s_checkinterval) &&
1878 (le32_to_cpu(es->s_lastcheck) +
1879 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1880 ext4_msg(sb, KERN_WARNING,
1881 "warning: checktime reached, "
1882 "running e2fsck is recommended");
0b8e58a1 1883 if (!sbi->s_journal)
0390131b 1884 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1885 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1886 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1887 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1888 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1889 ext4_update_dynamic_rev(sb);
0390131b
FM		 1890 if (sbi->s_journal)
1891 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1892
e2d67052 1893 ext4_commit_super(sb, 1);
281b5995 1894done:
ac27a0ec 1895 if (test_opt(sb, DEBUG))
a9df9a49 1896 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1897 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1898 sb->s_blocksize,
1899 sbi->s_groups_count,
617ba13b
MC		 1900 EXT4_BLOCKS_PER_GROUP(sb),
1901 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1902 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1903
7abc52c2 1904 cleancache_init_fs(sb);
ac27a0ec
DK		 1905 return res;
1906}
1907
117fff10
TT		 1908int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1909{
1910 struct ext4_sb_info *sbi = EXT4_SB(sb);
1911 struct flex_groups *new_groups;
1912 int size;
1913
1914 if (!sbi->s_log_groups_per_flex)
1915 return 0;
1916
1917 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1918 if (size <= sbi->s_flex_groups_allocated)
1919 return 0;
1920
1921 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1922 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1923 if (!new_groups) {
1924 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1925 size / (int) sizeof(struct flex_groups));
1926 return -ENOMEM;
1927 }
1928
1929 if (sbi->s_flex_groups) {
1930 memcpy(new_groups, sbi->s_flex_groups,
1931 (sbi->s_flex_groups_allocated *
1932 sizeof(struct flex_groups)));
b93b41d4 1933 kvfree(sbi->s_flex_groups);
117fff10
TT		 1934 }
1935 sbi->s_flex_groups = new_groups;
1936 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1937 return 0;
1938}
1939
772cb7c8
JS		 1940static int ext4_fill_flex_info(struct super_block *sb)
1941{
1942 struct ext4_sb_info *sbi = EXT4_SB(sb);
1943 struct ext4_group_desc *gdp = NULL;
772cb7c8 1944 ext4_group_t flex_group;
117fff10 1945 int i, err;
772cb7c8 1946
503358ae 1947 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1948 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1949 sbi->s_log_groups_per_flex = 0;
1950 return 1;
1951 }
1952
117fff10
TT		 1953 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1954 if (err)
9933fc0a 1955 goto failed;
772cb7c8 1956
772cb7c8 1957 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1958 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1959
1960 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1961 atomic_add(ext4_free_inodes_count(sb, gdp),
1962 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 1963 atomic64_add(ext4_free_group_clusters(sb, gdp),
1964 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1965 atomic_add(ext4_used_dirs_count(sb, gdp),
1966 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1967 }
1968
1969 return 1;
1970failed:
1971 return 0;
1972}
1973
feb0ab32
DW		 1974static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1975 struct ext4_group_desc *gdp)
717d50e4 1976{
feb0ab32 1977 int offset;
717d50e4 1978 __u16 crc = 0;
feb0ab32 1979 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1980
9aa5d32b 1981 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 1982 /* Use new metadata_csum algorithm */
d6a77105 1983 __le16 save_csum;
feb0ab32
DW		 1984 __u32 csum32;
1985
d6a77105 1986 save_csum = gdp->bg_checksum;
feb0ab32
DW		 1987 gdp->bg_checksum = 0;
1988 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1989 sizeof(le_group));
1990 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1991 sbi->s_desc_size);
d6a77105 1992 gdp->bg_checksum = save_csum;
feb0ab32
DW		 1993
1994 crc = csum32 & 0xFFFF;
1995 goto out;
717d50e4
AD		 1996 }
1997
feb0ab32 1998 /* old crc16 code */
813d32f9
DW		 1999 if (!(sbi->s_es->s_feature_ro_compat &
2000 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2001 return 0;
2002
feb0ab32
DW		 2003 offset = offsetof(struct ext4_group_desc, bg_checksum);
2004
2005 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2006 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2007 crc = crc16(crc, (__u8 *)gdp, offset);
2008 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2009 /* for checksum of struct ext4_group_desc do the rest...*/
2010 if ((sbi->s_es->s_feature_incompat &
2011 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2012 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2013 crc = crc16(crc, (__u8 *)gdp + offset,
2014 le16_to_cpu(sbi->s_es->s_desc_size) -
2015 offset);
2016
2017out:
717d50e4
AD		 2018 return cpu_to_le16(crc);
2019}
2020
feb0ab32 2021int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2022 struct ext4_group_desc *gdp)
2023{
feb0ab32
DW		 2024 if (ext4_has_group_desc_csum(sb) &&
2025 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2026 block_group, gdp)))
717d50e4
AD		 2027 return 0;
2028
2029 return 1;
2030}
2031
feb0ab32
DW		 2032void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2033 struct ext4_group_desc *gdp)
2034{
2035 if (!ext4_has_group_desc_csum(sb))
2036 return;
2037 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2038}
2039
ac27a0ec 2040/* Called at mount-time, super-block is locked */
bfff6873
LC		 2041static int ext4_check_descriptors(struct super_block *sb,
2042 ext4_group_t *first_not_zeroed)
ac27a0ec 2043{
617ba13b
MC		 2044 struct ext4_sb_info *sbi = EXT4_SB(sb);
2045 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2046 ext4_fsblk_t last_block;
bd81d8ee
LV		 2047 ext4_fsblk_t block_bitmap;
2048 ext4_fsblk_t inode_bitmap;
2049 ext4_fsblk_t inode_table;
ce421581 2050 int flexbg_flag = 0;
bfff6873 2051 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2052
ce421581
JS		 2053 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2054 flexbg_flag = 1;
2055
af5bc92d 2056 ext4_debug("Checking group descriptors");
ac27a0ec 2057
197cd65a
AM		 2058 for (i = 0; i < sbi->s_groups_count; i++) {
2059 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2060
ce421581 2061 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2062 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2063 else
2064 last_block = first_block +
617ba13b 2065 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2066
bfff6873
LC		 2067 if ((grp == sbi->s_groups_count) &&
2068 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2069 grp = i;
2070
8fadc143 2071 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2072 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2073 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2074 "Block bitmap for group %u not in group "
b31e1552 2075 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2076 return 0;
2077 }
8fadc143 2078 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2079 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2080 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2081 "Inode bitmap for group %u not in group "
b31e1552 2082 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2083 return 0;
2084 }
8fadc143 2085 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2086 if (inode_table < first_block ||
2b2d6d01 2087 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2088 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2089 "Inode table for group %u not in group "
b31e1552 2090 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2091 return 0;
2092 }
955ce5f5 2093 ext4_lock_group(sb, i);
feb0ab32 2094 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2095 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2096 "Checksum for group %u failed (%u!=%u)",
2097 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2098 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2099 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2100 ext4_unlock_group(sb, i);
8a266467 2101 return 0;
7ee1ec4c 2102 }
717d50e4 2103 }
955ce5f5 2104 ext4_unlock_group(sb, i);
ce421581
JS		 2105 if (!flexbg_flag)
2106 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2107 }
bfff6873
LC		 2108 if (NULL != first_not_zeroed)
2109 *first_not_zeroed = grp;
ac27a0ec
DK		 2110 return 1;
2111}
2112
617ba13b 2113/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2114 * the superblock) which were deleted from all directories, but held open by
2115 * a process at the time of a crash. We walk the list and try to delete these
2116 * inodes at recovery time (only with a read-write filesystem).
2117 *
2118 * In order to keep the orphan inode chain consistent during traversal (in
2119 * case of crash during recovery), we link each inode into the superblock
2120 * orphan list_head and handle it the same way as an inode deletion during
2121 * normal operation (which journals the operations for us).
2122 *
2123 * We only do an iget() and an iput() on each inode, which is very safe if we
2124 * accidentally point at an in-use or already deleted inode. The worst that
2125 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2126 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2127 * e2fsck was run on this filesystem, and it must have already done the orphan
2128 * inode cleanup for us, so we can safely abort without any further action.
2129 */
2b2d6d01
TT		 2130static void ext4_orphan_cleanup(struct super_block *sb,
2131 struct ext4_super_block *es)
ac27a0ec
DK		 2132{
2133 unsigned int s_flags = sb->s_flags;
2134 int nr_orphans = 0, nr_truncates = 0;
2135#ifdef CONFIG_QUOTA
2136 int i;
2137#endif
2138 if (!es->s_last_orphan) {
2139 jbd_debug(4, "no orphan inodes to clean up\n");
2140 return;
2141 }
2142
a8f48a95 2143 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2144 ext4_msg(sb, KERN_ERR, "write access "
2145 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2146 return;
2147 }
2148
d39195c3
AG		 2149 /* Check if feature set would not allow a r/w mount */
2150 if (!ext4_feature_set_ok(sb, 0)) {
2151 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2152 "unknown ROCOMPAT features");
2153 return;
2154 }
2155
617ba13b 2156 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2157 /* don't clear list on RO mount w/ errors */
2158 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2159 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2160 "clearing orphan list.\n");
c25f9bc6
ES		 2161 es->s_last_orphan = 0;
2162 }
ac27a0ec
DK		 2163 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2164 return;
2165 }
2166
2167 if (s_flags & MS_RDONLY) {
b31e1552 2168 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2169 sb->s_flags &= ~MS_RDONLY;
2170 }
2171#ifdef CONFIG_QUOTA
2172 /* Needed for iput() to work correctly and not trash data */
2173 sb->s_flags |= MS_ACTIVE;
2174 /* Turn on quotas so that they are updated correctly */
a2d4a646 2175 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2176 if (EXT4_SB(sb)->s_qf_names[i]) {
2177 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2178 if (ret < 0)
b31e1552
ES		 2179 ext4_msg(sb, KERN_ERR,
2180 "Cannot turn on journaled "
2181 "quota: error %d", ret);
ac27a0ec
DK		 2182 }
2183 }
2184#endif
2185
2186 while (es->s_last_orphan) {
2187 struct inode *inode;
2188
97bd42b9
JB		 2189 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2190 if (IS_ERR(inode)) {
ac27a0ec
DK		 2191 es->s_last_orphan = 0;
2192 break;
2193 }
2194
617ba13b 2195 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2196 dquot_initialize(inode);
ac27a0ec 2197 if (inode->i_nlink) {
566370a2
PT		 2198 if (test_opt(sb, DEBUG))
2199 ext4_msg(sb, KERN_DEBUG,
2200 "%s: truncating inode %lu to %lld bytes",
2201 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2202 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2203 inode->i_ino, inode->i_size);
721e3eba 2204 mutex_lock(&inode->i_mutex);
55f252c9 2205 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2206 ext4_truncate(inode);
721e3eba 2207 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2208 nr_truncates++;
2209 } else {
566370a2
PT		 2210 if (test_opt(sb, DEBUG))
2211 ext4_msg(sb, KERN_DEBUG,
2212 "%s: deleting unreferenced inode %lu",
2213 __func__, inode->i_ino);
ac27a0ec
DK		 2214 jbd_debug(2, "deleting unreferenced inode %lu\n",
2215 inode->i_ino);
2216 nr_orphans++;
2217 }
2218 iput(inode); /* The delete magic happens here! */
2219 }
2220
2b2d6d01 2221#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2222
2223 if (nr_orphans)
b31e1552
ES		 2224 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2225 PLURAL(nr_orphans));
ac27a0ec 2226 if (nr_truncates)
b31e1552
ES		 2227 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2228 PLURAL(nr_truncates));
ac27a0ec
DK		 2229#ifdef CONFIG_QUOTA
2230 /* Turn quotas off */
a2d4a646 2231 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2232 if (sb_dqopt(sb)->files[i])
287a8095 2233 dquot_quota_off(sb, i);
ac27a0ec
DK		 2234 }
2235#endif
2236 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2237}
0b8e58a1 2238
cd2291a4
ES		 2239/*
2240 * Maximal extent format file size.
2241 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2242 * extent format containers, within a sector_t, and within i_blocks
2243 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2244 * so that won't be a limiting factor.
2245 *
f17722f9
LC		 2246 * However there is other limiting factor. We do store extents in the form
2247 * of starting block and length, hence the resulting length of the extent
2248 * covering maximum file size must fit into on-disk format containers as
2249 * well. Given that length is always by 1 unit bigger than max unit (because
2250 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2251 *
cd2291a4
ES		 2252 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2253 */
f287a1a5 2254static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2255{
2256 loff_t res;
2257 loff_t upper_limit = MAX_LFS_FILESIZE;
2258
2259 /* small i_blocks in vfs inode? */
f287a1a5 2260 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2261 /*
90c699a9 2262 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2263 * i_block represent total blocks in 512 bytes
2264 * 32 == size of vfs inode i_blocks * 8
2265 */
2266 upper_limit = (1LL << 32) - 1;
2267
2268 /* total blocks in file system block size */
2269 upper_limit >>= (blkbits - 9);
2270 upper_limit <<= blkbits;
2271 }
2272
f17722f9
LC		 2273 /*
2274 * 32-bit extent-start container, ee_block. We lower the maxbytes
2275 * by one fs block, so ee_len can cover the extent of maximum file
2276 * size
2277 */
2278 res = (1LL << 32) - 1;
cd2291a4 2279 res <<= blkbits;
cd2291a4
ES		 2280
2281 /* Sanity check against vm- & vfs- imposed limits */
2282 if (res > upper_limit)
2283 res = upper_limit;
2284
2285 return res;
2286}
ac27a0ec 2287
ac27a0ec 2288/*
cd2291a4 2289 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2290 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2291 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2292 */
f287a1a5 2293static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2294{
617ba13b 2295 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2296 int meta_blocks;
2297 loff_t upper_limit;
0b8e58a1
AD		 2298 /* This is calculated to be the largest file size for a dense, block
2299 * mapped file such that the file's total number of 512-byte sectors,
2300 * including data and all indirect blocks, does not exceed (2^48 - 1).
2301 *
2302 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2303 * number of 512-byte sectors of the file.
0fc1b451
AK		 2304 */
2305
f287a1a5 2306 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2307 /*
90c699a9 2308 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2309 * the inode i_block field represents total file blocks in
2310 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2311 */
2312 upper_limit = (1LL << 32) - 1;
2313
2314 /* total blocks in file system block size */
2315 upper_limit >>= (bits - 9);
2316
2317 } else {
8180a562
AK		 2318 /*
2319 * We use 48 bit ext4_inode i_blocks
2320 * With EXT4_HUGE_FILE_FL set the i_blocks
2321 * represent total number of blocks in
2322 * file system block size
2323 */
0fc1b451
AK		 2324 upper_limit = (1LL << 48) - 1;
2325
0fc1b451
AK		 2326 }
2327
2328 /* indirect blocks */
2329 meta_blocks = 1;
2330 /* double indirect blocks */
2331 meta_blocks += 1 + (1LL << (bits-2));
2332 /* tripple indirect blocks */
2333 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2334
2335 upper_limit -= meta_blocks;
2336 upper_limit <<= bits;
ac27a0ec
DK		 2337
2338 res += 1LL << (bits-2);
2339 res += 1LL << (2*(bits-2));
2340 res += 1LL << (3*(bits-2));
2341 res <<= bits;
2342 if (res > upper_limit)
2343 res = upper_limit;
0fc1b451
AK		 2344
2345 if (res > MAX_LFS_FILESIZE)
2346 res = MAX_LFS_FILESIZE;
2347
ac27a0ec
DK		 2348 return res;
2349}
2350
617ba13b 2351static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2352 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2353{
617ba13b 2354 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2355 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2356 int has_super = 0;
2357
2358 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2359
617ba13b 2360 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2361 nr < first_meta_bg)
70bbb3e0 2362 return logical_sb_block + nr + 1;
ac27a0ec 2363 bg = sbi->s_desc_per_block * nr;
617ba13b 2364 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2365 has_super = 1;
0b8e58a1 2366
bd63f6b0
DW		 2367 /*
2368 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2369 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2370 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2371 * compensate.
2372 */
2373 if (sb->s_blocksize == 1024 && nr == 0 &&
2374 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2375 has_super++;
2376
617ba13b 2377 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2378}
2379
c9de560d
AT		 2380/**
2381 * ext4_get_stripe_size: Get the stripe size.
2382 * @sbi: In memory super block info
2383 *
2384 * If we have specified it via mount option, then
2385 * use the mount option value. If the value specified at mount time is
2386 * greater than the blocks per group use the super block value.
2387 * If the super block value is greater than blocks per group return 0.
2388 * Allocator needs it be less than blocks per group.
2389 *
2390 */
2391static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2392{
2393 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2394 unsigned long stripe_width =
2395 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2396 int ret;
c9de560d
AT		 2397
2398 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2399 ret = sbi->s_stripe;
2400 else if (stripe_width <= sbi->s_blocks_per_group)
2401 ret = stripe_width;
2402 else if (stride <= sbi->s_blocks_per_group)
2403 ret = stride;
2404 else
2405 ret = 0;
c9de560d 2406
3eb08658
DE		 2407 /*
2408 * If the stripe width is 1, this makes no sense and
2409 * we set it to 0 to turn off stripe handling code.
2410 */
2411 if (ret <= 1)
2412 ret = 0;
c9de560d 2413
3eb08658 2414 return ret;
c9de560d 2415}
ac27a0ec 2416
3197ebdb
TT		 2417/* sysfs supprt */
2418
2419struct ext4_attr {
2420 struct attribute attr;
2421 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2422 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb 2423 const char *, size_t);
f2d50a65
JK		 2424 union {
2425 int offset;
2426 int deprecated_val;
2427 } u;
3197ebdb
TT		 2428};
2429
27dd4385
LC		 2430static int parse_strtoull(const char *buf,
2431 unsigned long long max, unsigned long long *value)
2432{
2433 int ret;
2434
2435 ret = kstrtoull(skip_spaces(buf), 0, value);
2436 if (!ret && *value > max)
2437 ret = -EINVAL;
2438 return ret;
2439}
2440
3197ebdb
TT		 2441static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2442 struct ext4_sb_info *sbi,
2443 char *buf)
2444{
2445 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2446 (s64) EXT4_C2B(sbi,
2447 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2448}
2449
2450static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2451 struct ext4_sb_info *sbi, char *buf)
2452{
2453 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2454
f613dfcb
TT		 2455 if (!sb->s_bdev->bd_part)
2456 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2457 return snprintf(buf, PAGE_SIZE, "%lu\n",
2458 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2459 sbi->s_sectors_written_start) >> 1);
2460}
2461
2462static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2463 struct ext4_sb_info *sbi, char *buf)
2464{
2465 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2466
f613dfcb
TT		 2467 if (!sb->s_bdev->bd_part)
2468 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2469 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2470 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2471 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2472 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2473}
2474
2475static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2476 struct ext4_sb_info *sbi,
2477 const char *buf, size_t count)
2478{
2479 unsigned long t;
e1091b15 2480 int ret;
3197ebdb 2481
e1091b15
LC		 2482 ret = kstrtoul(skip_spaces(buf), 0, &t);
2483 if (ret)
2484 return ret;
3197ebdb 2485
e1091b15 2486 if (t && (!is_power_of_2(t) || t > 0x40000000))
3197ebdb
TT		 2487 return -EINVAL;
2488
2489 sbi->s_inode_readahead_blks = t;
2490 return count;
2491}
2492
2493static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2494 struct ext4_sb_info *sbi, char *buf)
3197ebdb 2495{
f2d50a65 2496 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb
TT		 2497
2498 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2499}
2500
2501static ssize_t sbi_ui_store(struct ext4_attr *a,
2502 struct ext4_sb_info *sbi,
2503 const char *buf, size_t count)
2504{
f2d50a65 2505 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb 2506 unsigned long t;
e1091b15 2507 int ret;
3197ebdb 2508
e1091b15
LC		 2509 ret = kstrtoul(skip_spaces(buf), 0, &t);
2510 if (ret)
2511 return ret;
3197ebdb
TT		 2512 *ui = t;
2513 return count;
2514}
2515
52c198c6
LC		 2516static ssize_t es_ui_show(struct ext4_attr *a,
2517 struct ext4_sb_info *sbi, char *buf)
2518{
2519
2520 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2521 a->u.offset);
2522
2523 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2524}
2525
27dd4385
LC		 2526static ssize_t reserved_clusters_show(struct ext4_attr *a,
2527 struct ext4_sb_info *sbi, char *buf)
2528{
2529 return snprintf(buf, PAGE_SIZE, "%llu\n",
2530 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2531}
2532
2533static ssize_t reserved_clusters_store(struct ext4_attr *a,
2534 struct ext4_sb_info *sbi,
2535 const char *buf, size_t count)
2536{
2537 unsigned long long val;
2538 int ret;
2539
2540 if (parse_strtoull(buf, -1ULL, &val))
2541 return -EINVAL;
2542 ret = ext4_reserve_clusters(sbi, val);
2543
2544 return ret ? ret : count;
2545}
2546
2c0544b2
TT		 2547static ssize_t trigger_test_error(struct ext4_attr *a,
2548 struct ext4_sb_info *sbi,
2549 const char *buf, size_t count)
2550{
2551 int len = count;
2552
2553 if (!capable(CAP_SYS_ADMIN))
2554 return -EPERM;
2555
2556 if (len && buf[len-1] == '\n')
2557 len--;
2558
2559 if (len)
2560 ext4_error(sbi->s_sb, "%.*s", len, buf);
2561 return count;
2562}
2563
f2d50a65
JK		 2564static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2565 struct ext4_sb_info *sbi, char *buf)
2566{
2567 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2568}
2569
3197ebdb
TT		 2570#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2571static struct ext4_attr ext4_attr_##_name = { \
2572 .attr = {.name = __stringify(_name), .mode = _mode }, \
2573 .show = _show, \
2574 .store = _store, \
f2d50a65
JK		 2575 .u = { \
2576 .offset = offsetof(struct ext4_sb_info, _elname),\
2577 }, \
3197ebdb 2578}
52c198c6
LC		 2579
2580#define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2581static struct ext4_attr ext4_attr_##_name = { \
2582 .attr = {.name = __stringify(_name), .mode = _mode }, \
2583 .show = _show, \
2584 .store = _store, \
2585 .u = { \
2586 .offset = offsetof(struct ext4_super_block, _elname), \
2587 }, \
2588}
2589
3197ebdb
TT		 2590#define EXT4_ATTR(name, mode, show, store) \
2591static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2592
857ac889 2593#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2594#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2595#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
52c198c6
LC		 2596
2597#define EXT4_RO_ATTR_ES_UI(name, elname) \
2598 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
3197ebdb
TT		 2599#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2600 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
52c198c6 2601
3197ebdb 2602#define ATTR_LIST(name) &ext4_attr_##name.attr
f2d50a65
JK		 2603#define EXT4_DEPRECATED_ATTR(_name, _val) \
2604static struct ext4_attr ext4_attr_##_name = { \
2605 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2606 .show = sbi_deprecated_show, \
2607 .u = { \
2608 .deprecated_val = _val, \
2609 }, \
2610}
3197ebdb
TT		 2611
2612EXT4_RO_ATTR(delayed_allocation_blocks);
2613EXT4_RO_ATTR(session_write_kbytes);
2614EXT4_RO_ATTR(lifetime_write_kbytes);
27dd4385 2615EXT4_RW_ATTR(reserved_clusters);
3197ebdb
TT		 2616EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2617 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2618EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2619EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2620EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2621EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2622EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2623EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2624EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
f2d50a65 2625EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
67a5da56 2626EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2627EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
efbed4dc
TT		 2628EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2629EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2630EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2631EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2632EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2633EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
52c198c6
LC		 2634EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2635EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2636EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
3197ebdb
TT		 2637
2638static struct attribute *ext4_attrs[] = {
2639 ATTR_LIST(delayed_allocation_blocks),
2640 ATTR_LIST(session_write_kbytes),
2641 ATTR_LIST(lifetime_write_kbytes),
27dd4385 2642 ATTR_LIST(reserved_clusters),
3197ebdb 2643 ATTR_LIST(inode_readahead_blks),
11013911 2644 ATTR_LIST(inode_goal),
3197ebdb
TT		 2645 ATTR_LIST(mb_stats),
2646 ATTR_LIST(mb_max_to_scan),
2647 ATTR_LIST(mb_min_to_scan),
2648 ATTR_LIST(mb_order2_req),
2649 ATTR_LIST(mb_stream_req),
2650 ATTR_LIST(mb_group_prealloc),
55138e0b 2651 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2652 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2653 ATTR_LIST(trigger_fs_error),
efbed4dc
TT		 2654 ATTR_LIST(err_ratelimit_interval_ms),
2655 ATTR_LIST(err_ratelimit_burst),
2656 ATTR_LIST(warning_ratelimit_interval_ms),
2657 ATTR_LIST(warning_ratelimit_burst),
2658 ATTR_LIST(msg_ratelimit_interval_ms),
2659 ATTR_LIST(msg_ratelimit_burst),
52c198c6
LC		 2660 ATTR_LIST(errors_count),
2661 ATTR_LIST(first_error_time),
2662 ATTR_LIST(last_error_time),
3197ebdb
TT		 2663 NULL,
2664};
2665
857ac889
LC		 2666/* Features this copy of ext4 supports */
2667EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2668EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2669EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2670
2671static struct attribute *ext4_feat_attrs[] = {
2672 ATTR_LIST(lazy_itable_init),
27ee40df 2673 ATTR_LIST(batched_discard),
5e7bbef1 2674 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2675 NULL,
2676};
2677
3197ebdb
TT		 2678static ssize_t ext4_attr_show(struct kobject *kobj,
2679 struct attribute *attr, char *buf)
2680{
2681 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2682 s_kobj);
2683 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2684
2685 return a->show ? a->show(a, sbi, buf) : 0;
2686}
2687
2688static ssize_t ext4_attr_store(struct kobject *kobj,
2689 struct attribute *attr,
2690 const char *buf, size_t len)
2691{
2692 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2693 s_kobj);
2694 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2695
2696 return a->store ? a->store(a, sbi, buf, len) : 0;
2697}
2698
2699static void ext4_sb_release(struct kobject *kobj)
2700{
2701 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2702 s_kobj);
2703 complete(&sbi->s_kobj_unregister);
2704}
2705
52cf25d0 2706static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2707 .show = ext4_attr_show,
2708 .store = ext4_attr_store,
2709};
2710
2711static struct kobj_type ext4_ktype = {
2712 .default_attrs = ext4_attrs,
2713 .sysfs_ops = &ext4_attr_ops,
2714 .release = ext4_sb_release,
2715};
2716
857ac889
LC		 2717static void ext4_feat_release(struct kobject *kobj)
2718{
2719 complete(&ext4_feat->f_kobj_unregister);
2720}
2721
c7f72543
LC		 2722static ssize_t ext4_feat_show(struct kobject *kobj,
2723 struct attribute *attr, char *buf)
2724{
2725 return snprintf(buf, PAGE_SIZE, "supported\n");
2726}
2727
2728/*
2729 * We can not use ext4_attr_show/store because it relies on the kobject
2730 * being embedded in the ext4_sb_info structure which is definitely not
2731 * true in this case.
2732 */
2733static const struct sysfs_ops ext4_feat_ops = {
2734 .show = ext4_feat_show,
2735 .store = NULL,
2736};
2737
857ac889
LC		 2738static struct kobj_type ext4_feat_ktype = {
2739 .default_attrs = ext4_feat_attrs,
c7f72543 2740 .sysfs_ops = &ext4_feat_ops,
857ac889
LC		 2741 .release = ext4_feat_release,
2742};
2743
a13fb1a4
ES		 2744/*
2745 * Check whether this filesystem can be mounted based on
2746 * the features present and the RDONLY/RDWR mount requested.
2747 * Returns 1 if this filesystem can be mounted as requested,
2748 * 0 if it cannot be.
2749 */
2750static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2751{
2752 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2753 ext4_msg(sb, KERN_ERR,
2754 "Couldn't mount because of "
2755 "unsupported optional features (%x)",
2756 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2757 ~EXT4_FEATURE_INCOMPAT_SUPP));
2758 return 0;
2759 }
2760
2761 if (readonly)
2762 return 1;
2763
2764 /* Check that feature set is OK for a read-write mount */
2765 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2766 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2767 "unsupported optional features (%x)",
2768 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2769 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2770 return 0;
2771 }
2772 /*
2773 * Large file size enabled file system can only be mounted
2774 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2775 */
2776 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2777 if (sizeof(blkcnt_t) < sizeof(u64)) {
2778 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2779 "cannot be mounted RDWR without "
2780 "CONFIG_LBDAF");
2781 return 0;
2782 }
2783 }
bab08ab9
TT		 2784 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2785 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2786 ext4_msg(sb, KERN_ERR,
2787 "Can't support bigalloc feature without "
2788 "extents feature\n");
2789 return 0;
2790 }
7c319d32
AK		 2791
2792#ifndef CONFIG_QUOTA
2793 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2794 !readonly) {
2795 ext4_msg(sb, KERN_ERR,
2796 "Filesystem with quota feature cannot be mounted RDWR "
2797 "without CONFIG_QUOTA");
2798 return 0;
2799 }
2800#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2801 return 1;
2802}
2803
66e61a9e
TT		 2804/*
2805 * This function is called once a day if we have errors logged
2806 * on the file system
2807 */
2808static void print_daily_error_info(unsigned long arg)
2809{
2810 struct super_block *sb = (struct super_block *) arg;
2811 struct ext4_sb_info *sbi;
2812 struct ext4_super_block *es;
2813
2814 sbi = EXT4_SB(sb);
2815 es = sbi->s_es;
2816
2817 if (es->s_error_count)
ae0f78de
TT		 2818 /* fsck newer than v1.41.13 is needed to clean this condition. */
2819 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2820 le32_to_cpu(es->s_error_count));
2821 if (es->s_first_error_time) {
ae0f78de 2822 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2823 sb->s_id, le32_to_cpu(es->s_first_error_time),
2824 (int) sizeof(es->s_first_error_func),
2825 es->s_first_error_func,
2826 le32_to_cpu(es->s_first_error_line));
2827 if (es->s_first_error_ino)
2828 printk(": inode %u",
2829 le32_to_cpu(es->s_first_error_ino));
2830 if (es->s_first_error_block)
2831 printk(": block %llu", (unsigned long long)
2832 le64_to_cpu(es->s_first_error_block));
2833 printk("\n");
2834 }
2835 if (es->s_last_error_time) {
ae0f78de 2836 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2837 sb->s_id, le32_to_cpu(es->s_last_error_time),
2838 (int) sizeof(es->s_last_error_func),
2839 es->s_last_error_func,
2840 le32_to_cpu(es->s_last_error_line));
2841 if (es->s_last_error_ino)
2842 printk(": inode %u",
2843 le32_to_cpu(es->s_last_error_ino));
2844 if (es->s_last_error_block)
2845 printk(": block %llu", (unsigned long long)
2846 le64_to_cpu(es->s_last_error_block));
2847 printk("\n");
2848 }
2849 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2850}
2851
bfff6873
LC		 2852/* Find next suitable group and run ext4_init_inode_table */
2853static int ext4_run_li_request(struct ext4_li_request *elr)
2854{
2855 struct ext4_group_desc *gdp = NULL;
2856 ext4_group_t group, ngroups;
2857 struct super_block *sb;
2858 unsigned long timeout = 0;
2859 int ret = 0;
2860
2861 sb = elr->lr_super;
2862 ngroups = EXT4_SB(sb)->s_groups_count;
2863
8e8ad8a5 2864 sb_start_write(sb);
bfff6873
LC		 2865 for (group = elr->lr_next_group; group < ngroups; group++) {
2866 gdp = ext4_get_group_desc(sb, group, NULL);
2867 if (!gdp) {
2868 ret = 1;
2869 break;
2870 }
2871
2872 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2873 break;
2874 }
2875
7f511862 2876 if (group >= ngroups)
bfff6873
LC		 2877 ret = 1;
2878
2879 if (!ret) {
2880 timeout = jiffies;
2881 ret = ext4_init_inode_table(sb, group,
2882 elr->lr_timeout ? 0 : 1);
2883 if (elr->lr_timeout == 0) {
51ce6511
LC		 2884 timeout = (jiffies - timeout) *
2885 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2886 elr->lr_timeout = timeout;
2887 }
2888 elr->lr_next_sched = jiffies + elr->lr_timeout;
2889 elr->lr_next_group = group + 1;
2890 }
8e8ad8a5 2891 sb_end_write(sb);
bfff6873
LC		 2892
2893 return ret;
2894}
2895
2896/*
2897 * Remove lr_request from the list_request and free the
4ed5c033 2898 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2899 */
2900static void ext4_remove_li_request(struct ext4_li_request *elr)
2901{
2902 struct ext4_sb_info *sbi;
2903
2904 if (!elr)
2905 return;
2906
2907 sbi = elr->lr_sbi;
2908
2909 list_del(&elr->lr_request);
2910 sbi->s_li_request = NULL;
2911 kfree(elr);
2912}
2913
2914static void ext4_unregister_li_request(struct super_block *sb)
2915{
1bb933fb
LC		 2916 mutex_lock(&ext4_li_mtx);
2917 if (!ext4_li_info) {
2918 mutex_unlock(&ext4_li_mtx);
bfff6873 2919 return;
1bb933fb 2920 }
bfff6873
LC		 2921
2922 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2923 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2924 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2925 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2926}
2927
8f1f7453
ES		 2928static struct task_struct *ext4_lazyinit_task;
2929
bfff6873
LC		 2930/*
2931 * This is the function where ext4lazyinit thread lives. It walks
2932 * through the request list searching for next scheduled filesystem.
2933 * When such a fs is found, run the lazy initialization request
2934 * (ext4_rn_li_request) and keep track of the time spend in this
2935 * function. Based on that time we compute next schedule time of
2936 * the request. When walking through the list is complete, compute
2937 * next waking time and put itself into sleep.
2938 */
2939static int ext4_lazyinit_thread(void *arg)
2940{
2941 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2942 struct list_head *pos, *n;
2943 struct ext4_li_request *elr;
4ed5c033 2944 unsigned long next_wakeup, cur;
bfff6873
LC		 2945
2946 BUG_ON(NULL == eli);
2947
bfff6873
LC		 2948cont_thread:
2949 while (true) {
2950 next_wakeup = MAX_JIFFY_OFFSET;
2951
2952 mutex_lock(&eli->li_list_mtx);
2953 if (list_empty(&eli->li_request_list)) {
2954 mutex_unlock(&eli->li_list_mtx);
2955 goto exit_thread;
2956 }
2957
2958 list_for_each_safe(pos, n, &eli->li_request_list) {
2959 elr = list_entry(pos, struct ext4_li_request,
2960 lr_request);
2961
b2c78cd0
TT		 2962 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2963 if (ext4_run_li_request(elr) != 0) {
2964 /* error, remove the lazy_init job */
2965 ext4_remove_li_request(elr);
2966 continue;
2967 }
bfff6873
LC		 2968 }
2969
2970 if (time_before(elr->lr_next_sched, next_wakeup))
2971 next_wakeup = elr->lr_next_sched;
2972 }
2973 mutex_unlock(&eli->li_list_mtx);
2974
a0acae0e 2975 try_to_freeze();
bfff6873 2976
4ed5c033
LC		 2977 cur = jiffies;
2978 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2979 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2980 cond_resched();
2981 continue;
2982 }
2983
4ed5c033
LC		 2984 schedule_timeout_interruptible(next_wakeup - cur);
2985
8f1f7453
ES		 2986 if (kthread_should_stop()) {
2987 ext4_clear_request_list();
2988 goto exit_thread;
2989 }
bfff6873
LC		 2990 }
2991
2992exit_thread:
2993 /*
2994 * It looks like the request list is empty, but we need
2995 * to check it under the li_list_mtx lock, to prevent any
2996 * additions into it, and of course we should lock ext4_li_mtx
2997 * to atomically free the list and ext4_li_info, because at
2998 * this point another ext4 filesystem could be registering
2999 * new one.
3000 */
3001 mutex_lock(&ext4_li_mtx);
3002 mutex_lock(&eli->li_list_mtx);
3003 if (!list_empty(&eli->li_request_list)) {
3004 mutex_unlock(&eli->li_list_mtx);
3005 mutex_unlock(&ext4_li_mtx);
3006 goto cont_thread;
3007 }
3008 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3009 kfree(ext4_li_info);
3010 ext4_li_info = NULL;
3011 mutex_unlock(&ext4_li_mtx);
3012
3013 return 0;
3014}
3015
3016static void ext4_clear_request_list(void)
3017{
3018 struct list_head *pos, *n;
3019 struct ext4_li_request *elr;
3020
3021 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3022 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3023 elr = list_entry(pos, struct ext4_li_request,
3024 lr_request);
3025 ext4_remove_li_request(elr);
3026 }
3027 mutex_unlock(&ext4_li_info->li_list_mtx);
3028}
3029
3030static int ext4_run_lazyinit_thread(void)
3031{
8f1f7453
ES		 3032 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3033 ext4_li_info, "ext4lazyinit");
3034 if (IS_ERR(ext4_lazyinit_task)) {
3035 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3036 ext4_clear_request_list();
bfff6873
LC		 3037 kfree(ext4_li_info);
3038 ext4_li_info = NULL;
92b97816 3039 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3040 "initialization thread\n",
3041 err);
3042 return err;
3043 }
3044 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3045 return 0;
3046}
3047
3048/*
3049 * Check whether it make sense to run itable init. thread or not.
3050 * If there is at least one uninitialized inode table, return
3051 * corresponding group number, else the loop goes through all
3052 * groups and return total number of groups.
3053 */
3054static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3055{
3056 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3057 struct ext4_group_desc *gdp = NULL;
3058
3059 for (group = 0; group < ngroups; group++) {
3060 gdp = ext4_get_group_desc(sb, group, NULL);
3061 if (!gdp)
3062 continue;
3063
3064 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3065 break;
3066 }
3067
3068 return group;
3069}
3070
3071static int ext4_li_info_new(void)
3072{
3073 struct ext4_lazy_init *eli = NULL;
3074
3075 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3076 if (!eli)
3077 return -ENOMEM;
3078
bfff6873
LC		 3079 INIT_LIST_HEAD(&eli->li_request_list);
3080 mutex_init(&eli->li_list_mtx);
3081
bfff6873
LC		 3082 eli->li_state |= EXT4_LAZYINIT_QUIT;
3083
3084 ext4_li_info = eli;
3085
3086 return 0;
3087}
3088
3089static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3090 ext4_group_t start)
3091{
3092 struct ext4_sb_info *sbi = EXT4_SB(sb);
3093 struct ext4_li_request *elr;
bfff6873
LC		 3094
3095 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3096 if (!elr)
3097 return NULL;
3098
3099 elr->lr_super = sb;
3100 elr->lr_sbi = sbi;
3101 elr->lr_next_group = start;
3102
3103 /*
3104 * Randomize first schedule time of the request to
3105 * spread the inode table initialization requests
3106 * better.
3107 */
dd1f723b
TT		 3108 elr->lr_next_sched = jiffies + (prandom_u32() %
3109 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3110 return elr;
3111}
3112
7f511862
TT		 3113int ext4_register_li_request(struct super_block *sb,
3114 ext4_group_t first_not_zeroed)
bfff6873
LC		 3115{
3116 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3117 struct ext4_li_request *elr = NULL;
bfff6873 3118 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3119 int ret = 0;
bfff6873 3120
7f511862 3121 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3122 if (sbi->s_li_request != NULL) {
3123 /*
3124 * Reset timeout so it can be computed again, because
3125 * s_li_wait_mult might have changed.
3126 */
3127 sbi->s_li_request->lr_timeout = 0;
7f511862 3128 goto out;
51ce6511 3129 }
bfff6873
LC		 3130
3131 if (first_not_zeroed == ngroups ||
3132 (sb->s_flags & MS_RDONLY) ||
55ff3840 3133 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3134 goto out;
bfff6873
LC		 3135
3136 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3137 if (!elr) {
3138 ret = -ENOMEM;
3139 goto out;
3140 }
bfff6873
LC		 3141
3142 if (NULL == ext4_li_info) {
3143 ret = ext4_li_info_new();
3144 if (ret)
3145 goto out;
3146 }
3147
3148 mutex_lock(&ext4_li_info->li_list_mtx);
3149 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3150 mutex_unlock(&ext4_li_info->li_list_mtx);
3151
3152 sbi->s_li_request = elr;
46e4690b
TM		 3153 /*
3154 * set elr to NULL here since it has been inserted to
3155 * the request_list and the removal and free of it is
3156 * handled by ext4_clear_request_list from now on.
3157 */
3158 elr = NULL;
bfff6873
LC		 3159
3160 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3161 ret = ext4_run_lazyinit_thread();
3162 if (ret)
3163 goto out;
3164 }
bfff6873 3165out:
beed5ecb
NK		 3166 mutex_unlock(&ext4_li_mtx);
3167 if (ret)
bfff6873 3168 kfree(elr);
bfff6873
LC		 3169 return ret;
3170}
3171
3172/*
3173 * We do not need to lock anything since this is called on
3174 * module unload.
3175 */
3176static void ext4_destroy_lazyinit_thread(void)
3177{
3178 /*
3179 * If thread exited earlier
3180 * there's nothing to be done.
3181 */
8f1f7453 3182 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3183 return;
3184
8f1f7453 3185 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3186}
3187
25ed6e8a
DW		 3188static int set_journal_csum_feature_set(struct super_block *sb)
3189{
3190 int ret = 1;
3191 int compat, incompat;
3192 struct ext4_sb_info *sbi = EXT4_SB(sb);
3193
9aa5d32b 3194 if (ext4_has_metadata_csum(sb)) {
db9ee220 3195 /* journal checksum v3 */
25ed6e8a 3196 compat = 0;
db9ee220 3197 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3198 } else {
3199 /* journal checksum v1 */
3200 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3201 incompat = 0;
3202 }
3203
feb8c6d3
DW		 3204 jbd2_journal_clear_features(sbi->s_journal,
3205 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3206 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3207 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3208 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3209 ret = jbd2_journal_set_features(sbi->s_journal,
3210 compat, 0,
3211 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3212 incompat);
3213 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3214 ret = jbd2_journal_set_features(sbi->s_journal,
3215 compat, 0,
3216 incompat);
3217 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3218 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3219 } else {
feb8c6d3
DW		 3220 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3221 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3222 }
3223
3224 return ret;
3225}
3226
952fc18e
TT		 3227/*
3228 * Note: calculating the overhead so we can be compatible with
3229 * historical BSD practice is quite difficult in the face of
3230 * clusters/bigalloc. This is because multiple metadata blocks from
3231 * different block group can end up in the same allocation cluster.
3232 * Calculating the exact overhead in the face of clustered allocation
3233 * requires either O(all block bitmaps) in memory or O(number of block
3234 * groups**2) in time. We will still calculate the superblock for
3235 * older file systems --- and if we come across with a bigalloc file
3236 * system with zero in s_overhead_clusters the estimate will be close to
3237 * correct especially for very large cluster sizes --- but for newer
3238 * file systems, it's better to calculate this figure once at mkfs
3239 * time, and store it in the superblock. If the superblock value is
3240 * present (even for non-bigalloc file systems), we will use it.
3241 */
3242static int count_overhead(struct super_block *sb, ext4_group_t grp,
3243 char *buf)
3244{
3245 struct ext4_sb_info *sbi = EXT4_SB(sb);
3246 struct ext4_group_desc *gdp;
3247 ext4_fsblk_t first_block, last_block, b;
3248 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3249 int s, j, count = 0;
3250
0548bbb8
TT		 3251 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3252 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3253 sbi->s_itb_per_group + 2);
3254
952fc18e
TT		 3255 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3256 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3257 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3258 for (i = 0; i < ngroups; i++) {
3259 gdp = ext4_get_group_desc(sb, i, NULL);
3260 b = ext4_block_bitmap(sb, gdp);
3261 if (b >= first_block && b <= last_block) {
3262 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3263 count++;
3264 }
3265 b = ext4_inode_bitmap(sb, gdp);
3266 if (b >= first_block && b <= last_block) {
3267 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3268 count++;
3269 }
3270 b = ext4_inode_table(sb, gdp);
3271 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3272 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3273 int c = EXT4_B2C(sbi, b - first_block);
3274 ext4_set_bit(c, buf);
3275 count++;
3276 }
3277 if (i != grp)
3278 continue;
3279 s = 0;
3280 if (ext4_bg_has_super(sb, grp)) {
3281 ext4_set_bit(s++, buf);
3282 count++;
3283 }
3284 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3285 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3286 count++;
3287 }
3288 }
3289 if (!count)
3290 return 0;
3291 return EXT4_CLUSTERS_PER_GROUP(sb) -
3292 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3293}
3294
3295/*
3296 * Compute the overhead and stash it in sbi->s_overhead
3297 */
3298int ext4_calculate_overhead(struct super_block *sb)
3299{
3300 struct ext4_sb_info *sbi = EXT4_SB(sb);
3301 struct ext4_super_block *es = sbi->s_es;
3302 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3303 ext4_fsblk_t overhead = 0;
3304 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3305
952fc18e
TT		 3306 if (!buf)
3307 return -ENOMEM;
3308
3309 /*
3310 * Compute the overhead (FS structures). This is constant
3311 * for a given filesystem unless the number of block groups
3312 * changes so we cache the previous value until it does.
3313 */
3314
3315 /*
3316 * All of the blocks before first_data_block are overhead
3317 */
3318 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3319
3320 /*
3321 * Add the overhead found in each block group
3322 */
3323 for (i = 0; i < ngroups; i++) {
3324 int blks;
3325
3326 blks = count_overhead(sb, i, buf);
3327 overhead += blks;
3328 if (blks)
3329 memset(buf, 0, PAGE_SIZE);
3330 cond_resched();
3331 }
0875a2b4
ES		 3332 /* Add the journal blocks as well */
3333 if (sbi->s_journal)
810da240 3334 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3335
952fc18e
TT		 3336 sbi->s_overhead = overhead;
3337 smp_wmb();
3338 free_page((unsigned long) buf);
3339 return 0;
3340}
3341
27dd4385 3342
30fac0f7 3343static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
27dd4385
LC		 3344{
3345 ext4_fsblk_t resv_clusters;
3346
30fac0f7
JK		 3347 /*
3348 * There's no need to reserve anything when we aren't using extents.
3349 * The space estimates are exact, there are no unwritten extents,
3350 * hole punching doesn't need new metadata... This is needed especially
3351 * to keep ext2/3 backward compatibility.
3352 */
3353 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3354 return 0;
27dd4385
LC		 3355 /*
3356 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3357 * This should cover the situations where we can not afford to run
3358 * out of space like for example punch hole, or converting
556615dc 3359 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3360 * allocation would require 1, or 2 blocks, higher numbers are
3361 * very rare.
3362 */
30fac0f7
JK		 3363 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3364 EXT4_SB(sb)->s_cluster_bits;
27dd4385
LC		 3365
3366 do_div(resv_clusters, 50);
3367 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3368
3369 return resv_clusters;
3370}
3371
3372
3373static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3374{
3375 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3376 sbi->s_cluster_bits;
3377
3378 if (count >= clusters)
3379 return -EINVAL;
3380
3381 atomic64_set(&sbi->s_resv_clusters, count);
3382 return 0;
3383}
3384
2b2d6d01 3385static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3386{
d4c402d9 3387 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3388 struct buffer_head *bh;
617ba13b
MC		 3389 struct ext4_super_block *es = NULL;
3390 struct ext4_sb_info *sbi;
3391 ext4_fsblk_t block;
3392 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3393 ext4_fsblk_t logical_sb_block;
ac27a0ec 3394 unsigned long offset = 0;
ac27a0ec
DK		 3395 unsigned long journal_devnum = 0;
3396 unsigned long def_mount_opts;
3397 struct inode *root;
9f6200bb 3398 char *cp;
0390131b 3399 const char *descr;
dcc7dae3 3400 int ret = -ENOMEM;
281b5995 3401 int blocksize, clustersize;
4ec11028
TT		 3402 unsigned int db_count;
3403 unsigned int i;
281b5995 3404 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3405 __u64 blocks_count;
07aa2ea1 3406 int err = 0;
b3881f74 3407 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3408 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3409
3410 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3411 if (!sbi)
dcc7dae3 3412 goto out_free_orig;
705895b6
PE		 3413
3414 sbi->s_blockgroup_lock =
3415 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3416 if (!sbi->s_blockgroup_lock) {
3417 kfree(sbi);
dcc7dae3 3418 goto out_free_orig;
705895b6 3419 }
ac27a0ec 3420 sb->s_fs_info = sbi;
2c0544b2 3421 sbi->s_sb = sb;
240799cd 3422 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3423 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3424 if (sb->s_bdev->bd_part)
3425 sbi->s_sectors_written_start =
3426 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3427
9f6200bb
TT		 3428 /* Cleanup superblock name */
3429 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3430 *cp = '!';
3431
07aa2ea1 3432 /* -EINVAL is default */
dcc7dae3 3433 ret = -EINVAL;
617ba13b 3434 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3435 if (!blocksize) {
b31e1552 3436 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3437 goto out_fail;
3438 }
3439
3440 /*
617ba13b 3441 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3442 * block sizes. We need to calculate the offset from buffer start.
3443 */
617ba13b 3444 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3445 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3446 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3447 } else {
70bbb3e0 3448 logical_sb_block = sb_block;
ac27a0ec
DK		 3449 }
3450
a8ac900b 3451 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3452 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3453 goto out_fail;
3454 }
3455 /*
3456 * Note: s_es must be initialized as soon as possible because
617ba13b 3457 * some ext4 macro-instructions depend on its value
ac27a0ec 3458 */
2716b802 3459 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3460 sbi->s_es = es;
3461 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3462 if (sb->s_magic != EXT4_SUPER_MAGIC)
3463 goto cantfind_ext4;
afc32f7e 3464 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3465
feb0ab32
DW		 3466 /* Warn if metadata_csum and gdt_csum are both set. */
3467 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3468 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3469 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3470 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3471 "redundant flags; please run fsck.");
3472
d25425f8
DW		 3473 /* Check for a known checksum algorithm */
3474 if (!ext4_verify_csum_type(sb, es)) {
3475 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3476 "unknown checksum algorithm.");
3477 silent = 1;
3478 goto cantfind_ext4;
3479 }
3480
0441984a
DW		 3481 /* Load the checksum driver */
3482 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3483 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3484 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3485 if (IS_ERR(sbi->s_chksum_driver)) {
3486 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3487 ret = PTR_ERR(sbi->s_chksum_driver);
3488 sbi->s_chksum_driver = NULL;
3489 goto failed_mount;
3490 }
3491 }
3492
a9c47317
DW		 3493 /* Check superblock checksum */
3494 if (!ext4_superblock_csum_verify(sb, es)) {
3495 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3496 "invalid superblock checksum. Run e2fsck?");
3497 silent = 1;
3498 goto cantfind_ext4;
3499 }
3500
3501 /* Precompute checksum seed for all metadata */
9aa5d32b 3502 if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3503 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3504 sizeof(es->s_uuid));
3505
ac27a0ec
DK		 3506 /* Set defaults before we parse the mount options */
3507 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3508 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3509 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3510 set_opt(sb, DEBUG);
87f26807 3511 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3512 set_opt(sb, GRPID);
617ba13b 3513 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3514 set_opt(sb, NO_UID32);
ea663336 3515 /* xattr user namespace & acls are now defaulted on */
ea663336 3516 set_opt(sb, XATTR_USER);
03010a33 3517#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3518 set_opt(sb, POSIX_ACL);
2e7842b8 3519#endif
98c1a759
DW		 3520 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3521 if (ext4_has_metadata_csum(sb))
3522 set_opt(sb, JOURNAL_CHECKSUM);
3523
617ba13b 3524 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3525 set_opt(sb, JOURNAL_DATA);
617ba13b 3526 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3527 set_opt(sb, ORDERED_DATA);
617ba13b 3528 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3529 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3530
3531 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3532 set_opt(sb, ERRORS_PANIC);
bb4f397a 3533 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3534 set_opt(sb, ERRORS_CONT);
bb4f397a 3535 else
fd8c37ec 3536 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3537 /* block_validity enabled by default; disable with noblock_validity */
3538 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3539 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3540 set_opt(sb, DISCARD);
ac27a0ec 3541
08cefc7a
EB		 3542 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3543 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3544 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3545 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3546 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3547
8b67f04a 3548 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3549 set_opt(sb, BARRIER);
ac27a0ec 3550
dd919b98
AK		 3551 /*
3552 * enable delayed allocation by default
3553 * Use -o nodelalloc to turn it off
3554 */
bc0b75f7 3555 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3556 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3557 set_opt(sb, DELALLOC);
dd919b98 3558
51ce6511
LC		 3559 /*
3560 * set default s_li_wait_mult for lazyinit, for the case there is
3561 * no mount option specified.
3562 */
3563 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3564
8b67f04a 3565 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3566 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3567 ext4_msg(sb, KERN_WARNING,
3568 "failed to parse options in superblock: %s",
3569 sbi->s_es->s_mount_opts);
3570 }
5a916be1 3571 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3572 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3573 &journal_ioprio, 0))
ac27a0ec
DK		 3574 goto failed_mount;
3575
56889787
TT		 3576 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3577 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3578 "with data=journal disables delayed "
3579 "allocation and O_DIRECT support!\n");
3580 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3581 ext4_msg(sb, KERN_ERR, "can't mount with "
3582 "both data=journal and delalloc");
3583 goto failed_mount;
3584 }
3585 if (test_opt(sb, DIOREAD_NOLOCK)) {
3586 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3587 "both data=journal and dioread_nolock");
56889787
TT		 3588 goto failed_mount;
3589 }
3590 if (test_opt(sb, DELALLOC))
3591 clear_opt(sb, DELALLOC);
3592 }
3593
ac27a0ec 3594 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3595 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3596
617ba13b
MC		 3597 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3598 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3599 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3600 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3601 ext4_msg(sb, KERN_WARNING,
3602 "feature flags set on rev 0 fs, "
3603 "running e2fsck is recommended");
469108ff 3604
ed3654eb
TT		 3605 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3606 set_opt2(sb, HURD_COMPAT);
3607 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3608 EXT4_FEATURE_INCOMPAT_64BIT)) {
3609 ext4_msg(sb, KERN_ERR,
3610 "The Hurd can't support 64-bit file systems");
3611 goto failed_mount;
3612 }
3613 }
3614
2035e776
TT		 3615 if (IS_EXT2_SB(sb)) {
3616 if (ext2_feature_set_ok(sb))
3617 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3618 "using the ext4 subsystem");
3619 else {
3620 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3621 "to feature incompatibilities");
3622 goto failed_mount;
3623 }
3624 }
3625
3626 if (IS_EXT3_SB(sb)) {
3627 if (ext3_feature_set_ok(sb))
3628 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3629 "using the ext4 subsystem");
3630 else {
3631 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3632 "to feature incompatibilities");
3633 goto failed_mount;
3634 }
3635 }
3636
ac27a0ec
DK		 3637 /*
3638 * Check feature flags regardless of the revision level, since we
3639 * previously didn't change the revision level when setting the flags,
3640 * so there is a chance incompat flags are set on a rev 0 filesystem.
3641 */
a13fb1a4 3642 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3643 goto failed_mount;
a13fb1a4 3644
261cb20c 3645 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3646 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3647 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3648 ext4_msg(sb, KERN_ERR,
3649 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3650 goto failed_mount;
3651 }
3652
ac27a0ec 3653 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3654 /* Validate the filesystem blocksize */
3655 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3656 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3657 blocksize);
ac27a0ec
DK		 3658 goto failed_mount;
3659 }
3660
2b2d6d01 3661 brelse(bh);
70bbb3e0
AM		 3662 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3663 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3664 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3665 if (!bh) {
b31e1552
ES		 3666 ext4_msg(sb, KERN_ERR,
3667 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3668 goto failed_mount;
3669 }
2716b802 3670 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3671 sbi->s_es = es;
617ba13b 3672 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3673 ext4_msg(sb, KERN_ERR,
3674 "Magic mismatch, very weird!");
ac27a0ec
DK		 3675 goto failed_mount;
3676 }
3677 }
3678
a13fb1a4
ES		 3679 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3680 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3681 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3682 has_huge_files);
3683 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3684
617ba13b
MC		 3685 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3686 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3687 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3688 } else {
3689 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3690 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3691 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3692 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3693 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3694 ext4_msg(sb, KERN_ERR,
3695 "unsupported inode size: %d",
2b2d6d01 3696 sbi->s_inode_size);
ac27a0ec
DK		 3697 goto failed_mount;
3698 }
ef7f3835
KS		 3699 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3700 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3701 }
0b8e58a1 3702
0d1ee42f
AR		 3703 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3704 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3705 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3706 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3707 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3708 ext4_msg(sb, KERN_ERR,
3709 "unsupported descriptor size %lu",
0d1ee42f
AR		 3710 sbi->s_desc_size);
3711 goto failed_mount;
3712 }
3713 } else
3714 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3715
ac27a0ec 3716 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3717 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3718 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3719 goto cantfind_ext4;
0b8e58a1 3720
617ba13b 3721 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3722 if (sbi->s_inodes_per_block == 0)
617ba13b 3723 goto cantfind_ext4;
ac27a0ec
DK		 3724 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3725 sbi->s_inodes_per_block;
0d1ee42f 3726 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3727 sbi->s_sbh = bh;
3728 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3729 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3730 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3731
2b2d6d01 3732 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3733 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3734 sbi->s_def_hash_version = es->s_def_hash_version;
23301410
TT		 3735 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3736 i = le32_to_cpu(es->s_flags);
3737 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3738 sbi->s_hash_unsigned = 3;
3739 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3740#ifdef __CHAR_UNSIGNED__
23301410
TT		 3741 if (!(sb->s_flags & MS_RDONLY))
3742 es->s_flags |=
3743 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3744 sbi->s_hash_unsigned = 3;
f99b2589 3745#else
23301410
TT		 3746 if (!(sb->s_flags & MS_RDONLY))
3747 es->s_flags |=
3748 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3749#endif
23301410 3750 }
f99b2589 3751 }
ac27a0ec 3752
281b5995
TT		 3753 /* Handle clustersize */
3754 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3755 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3756 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3757 if (has_bigalloc) {
3758 if (clustersize < blocksize) {
3759 ext4_msg(sb, KERN_ERR,
3760 "cluster size (%d) smaller than "
3761 "block size (%d)", clustersize, blocksize);
3762 goto failed_mount;
3763 }
3764 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3765 le32_to_cpu(es->s_log_block_size);
3766 sbi->s_clusters_per_group =
3767 le32_to_cpu(es->s_clusters_per_group);
3768 if (sbi->s_clusters_per_group > blocksize * 8) {
3769 ext4_msg(sb, KERN_ERR,
3770 "#clusters per group too big: %lu",
3771 sbi->s_clusters_per_group);
3772 goto failed_mount;
3773 }
3774 if (sbi->s_blocks_per_group !=
3775 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3776 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3777 "clusters per group (%lu) inconsistent",
3778 sbi->s_blocks_per_group,
3779 sbi->s_clusters_per_group);
3780 goto failed_mount;
3781 }
3782 } else {
3783 if (clustersize != blocksize) {
3784 ext4_warning(sb, "fragment/cluster size (%d) != "
3785 "block size (%d)", clustersize,
3786 blocksize);
3787 clustersize = blocksize;
3788 }
3789 if (sbi->s_blocks_per_group > blocksize * 8) {
3790 ext4_msg(sb, KERN_ERR,
3791 "#blocks per group too big: %lu",
3792 sbi->s_blocks_per_group);
3793 goto failed_mount;
3794 }
3795 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3796 sbi->s_cluster_bits = 0;
ac27a0ec 3797 }
281b5995
TT		 3798 sbi->s_cluster_ratio = clustersize / blocksize;
3799
ac27a0ec 3800 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3801 ext4_msg(sb, KERN_ERR,
3802 "#inodes per group too big: %lu",
2b2d6d01 3803 sbi->s_inodes_per_group);
ac27a0ec
DK		 3804 goto failed_mount;
3805 }
3806
960fd856
TT		 3807 /* Do we have standard group size of clustersize * 8 blocks ? */
3808 if (sbi->s_blocks_per_group == clustersize << 3)
3809 set_opt2(sb, STD_GROUP_SIZE);
3810
bf43d84b
ES		 3811 /*
3812 * Test whether we have more sectors than will fit in sector_t,
3813 * and whether the max offset is addressable by the page cache.
3814 */
5a9ae68a 3815 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3816 ext4_blocks_count(es));
5a9ae68a 3817 if (err) {
b31e1552 3818 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3819 " too large to mount safely on this system");
ac27a0ec 3820 if (sizeof(sector_t) < 8)
90c699a9 3821 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3822 goto failed_mount;
3823 }
3824
617ba13b
MC		 3825 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3826 goto cantfind_ext4;
e7c95593 3827
0f2ddca6
FTN		 3828 /* check blocks count against device size */
3829 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3830 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3831 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3832 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3833 ext4_blocks_count(es), blocks_count);
3834 goto failed_mount;
3835 }
3836
0b8e58a1
AD		 3837 /*
3838 * It makes no sense for the first data block to be beyond the end
3839 * of the filesystem.
3840 */
3841 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3842 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3843 "block %u is beyond end of filesystem (%llu)",
3844 le32_to_cpu(es->s_first_data_block),
3845 ext4_blocks_count(es));
e7c95593
ES		 3846 goto failed_mount;
3847 }
bd81d8ee
LV		 3848 blocks_count = (ext4_blocks_count(es) -
3849 le32_to_cpu(es->s_first_data_block) +
3850 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3851 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3852 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3853 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3854 "(block count %llu, first data block %u, "
b31e1552 3855 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3856 ext4_blocks_count(es),
3857 le32_to_cpu(es->s_first_data_block),
3858 EXT4_BLOCKS_PER_GROUP(sb));
3859 goto failed_mount;
3860 }
bd81d8ee 3861 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3862 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3863 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3864 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3865 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3866 sbi->s_group_desc = ext4_kvmalloc(db_count *
3867 sizeof(struct buffer_head *),
3868 GFP_KERNEL);
ac27a0ec 3869 if (sbi->s_group_desc == NULL) {
b31e1552 3870 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3871 ret = -ENOMEM;
ac27a0ec
DK		 3872 goto failed_mount;
3873 }
3874
9f6200bb
TT		 3875 if (ext4_proc_root)
3876 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3877
66acdcf4
TT		 3878 if (sbi->s_proc)
3879 proc_create_data("options", S_IRUGO, sbi->s_proc,
3880 &ext4_seq_options_fops, sb);
3881
705895b6 3882 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3883
3884 for (i = 0; i < db_count; i++) {
70bbb3e0 3885 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3886 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3887 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3888 ext4_msg(sb, KERN_ERR,
3889 "can't read group descriptor %d", i);
ac27a0ec
DK		 3890 db_count = i;
3891 goto failed_mount2;
3892 }
3893 }
bfff6873 3894 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3895 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
f9ae9cf5 3896 goto failed_mount2;
ac27a0ec 3897 }
772cb7c8 3898
f9ae9cf5 3899 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3900 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3901 spin_lock_init(&sbi->s_next_gen_lock);
3902
04496411
TM		 3903 init_timer(&sbi->s_err_report);
3904 sbi->s_err_report.function = print_daily_error_info;
3905 sbi->s_err_report.data = (unsigned long) sb;
3906
a75ae78f 3907 /* Register extent status tree shrinker */
eb68d0e2 3908 if (ext4_es_register_shrinker(sbi))
ce7e010a 3909 goto failed_mount3;
ce7e010a 3910
c9de560d 3911 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3912 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3913
f9ae9cf5
TT		 3914 /*
3915 * set up enough so that it can read an inode
3916 */
f6e63f90 3917 sb->s_op = &ext4_sops;
617ba13b
MC		 3918 sb->s_export_op = &ext4_export_ops;
3919 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3920#ifdef CONFIG_QUOTA
617ba13b 3921 sb->dq_op = &ext4_quota_operations;
262b4662 3922 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
7c319d32 3923 sb->s_qcop = &ext4_qctl_sysfile_operations;
262b4662
JK		 3924 else
3925 sb->s_qcop = &ext4_qctl_operations;
ac27a0ec 3926#endif
f2fa2ffc
AK		 3927 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3928
ac27a0ec 3929 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3930 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3931
3932 sb->s_root = NULL;
3933
3934 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3935 EXT4_HAS_INCOMPAT_FEATURE(sb,
3936 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3937
c5e06d10
JL		 3938 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3939 !(sb->s_flags & MS_RDONLY))
3940 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3941 goto failed_mount3a;
c5e06d10 3942
ac27a0ec
DK		 3943 /*
3944 * The first inode we look at is the journal inode. Don't try
3945 * root first: it may be modified in the journal!
3946 */
3947 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3948 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3949 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 3950 goto failed_mount3a;
0390131b
FM		 3951 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3952 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3953 ext4_msg(sb, KERN_ERR, "required journal recovery "
3954 "suppressed and not mounted read-only");
744692dc 3955 goto failed_mount_wq;
ac27a0ec 3956 } else {
fd8c37ec 3957 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3958 sbi->s_journal = NULL;
3959 needs_recovery = 0;
3960 goto no_journal;
ac27a0ec
DK		 3961 }
3962
f32aaf2d 3963 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3964 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3965 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3966 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3967 goto failed_mount_wq;
eb40a09c
JS		 3968 }
3969
25ed6e8a
DW		 3970 if (!set_journal_csum_feature_set(sb)) {
3971 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3972 "feature set");
3973 goto failed_mount_wq;
d4da6c9c 3974 }
818d276c 3975
ac27a0ec
DK		 3976 /* We have now updated the journal if required, so we can
3977 * validate the data journaling mode. */
3978 switch (test_opt(sb, DATA_FLAGS)) {
3979 case 0:
3980 /* No mode set, assume a default based on the journal
63f57933
AM		 3981 * capabilities: ORDERED_DATA if the journal can
3982 * cope, else JOURNAL_DATA
3983 */
dab291af
MC		 3984 if (jbd2_journal_check_available_features
3985 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3986 set_opt(sb, ORDERED_DATA);
ac27a0ec 3987 else
fd8c37ec 3988 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3989 break;
3990
617ba13b
MC		 3991 case EXT4_MOUNT_ORDERED_DATA:
3992 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3993 if (!jbd2_journal_check_available_features
3994 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3995 ext4_msg(sb, KERN_ERR, "Journal does not support "
3996 "requested data journaling mode");
744692dc 3997 goto failed_mount_wq;
ac27a0ec
DK		 3998 }
3999 default:
4000 break;
4001 }
b3881f74 4002 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4003
18aadd47
BJ		 4004 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4005
ce7e010a 4006no_journal:
9c191f70
M		 4007 if (ext4_mballoc_ready) {
4008 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4009 if (!sbi->s_mb_cache) {
4010 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4011 goto failed_mount_wq;
4012 }
4013 }
4014
952fc18e
TT		 4015 /*
4016 * Get the # of file system overhead blocks from the
4017 * superblock if present.
4018 */
4019 if (es->s_overhead_clusters)
4020 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4021 else {
07aa2ea1
LC		 4022 err = ext4_calculate_overhead(sb);
4023 if (err)
952fc18e
TT		 4024 goto failed_mount_wq;
4025 }
4026
fd89d5f2
TH		 4027 /*
4028 * The maximum number of concurrent works can be high and
4029 * concurrency isn't really necessary. Limit it to 1.
4030 */
2e8fa54e
JK		 4031 EXT4_SB(sb)->rsv_conversion_wq =
4032 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4033 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4034 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4035 ret = -ENOMEM;
2e8fa54e
JK		 4036 goto failed_mount4;
4037 }
4038
ac27a0ec 4039 /*
dab291af 4040 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4041 * so we can safely mount the rest of the filesystem now.
4042 */
4043
1d1fe1ee
DH		 4044 root = ext4_iget(sb, EXT4_ROOT_INO);
4045 if (IS_ERR(root)) {
b31e1552 4046 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4047 ret = PTR_ERR(root);
32a9bb57 4048 root = NULL;
ac27a0ec
DK		 4049 goto failed_mount4;
4050 }
4051 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4052 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4053 iput(root);
ac27a0ec
DK		 4054 goto failed_mount4;
4055 }
48fde701 4056 sb->s_root = d_make_root(root);
1d1fe1ee 4057 if (!sb->s_root) {
b31e1552 4058 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4059 ret = -ENOMEM;
4060 goto failed_mount4;
4061 }
ac27a0ec 4062
7e84b621
ES		 4063 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4064 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4065
4066 /* determine the minimum size of new large inodes, if present */
4067 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4068 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4069 EXT4_GOOD_OLD_INODE_SIZE;
4070 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4071 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4072 if (sbi->s_want_extra_isize <
4073 le16_to_cpu(es->s_want_extra_isize))
4074 sbi->s_want_extra_isize =
4075 le16_to_cpu(es->s_want_extra_isize);
4076 if (sbi->s_want_extra_isize <
4077 le16_to_cpu(es->s_min_extra_isize))
4078 sbi->s_want_extra_isize =
4079 le16_to_cpu(es->s_min_extra_isize);
4080 }
4081 }
4082 /* Check if enough inode space is available */
4083 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4084 sbi->s_inode_size) {
4085 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4086 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4087 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4088 "available");
ef7f3835
KS		 4089 }
4090
30fac0f7 4091 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
27dd4385
LC		 4092 if (err) {
4093 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
30fac0f7 4094 "reserved pool", ext4_calculate_resv_clusters(sb));
f9ae9cf5 4095 goto failed_mount4a;
27dd4385
LC		 4096 }
4097
6fd058f7
TT		 4098 err = ext4_setup_system_zone(sb);
4099 if (err) {
b31e1552 4100 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4101 "zone (%d)", err);
f9ae9cf5
TT		 4102 goto failed_mount4a;
4103 }
4104
4105 ext4_ext_init(sb);
4106 err = ext4_mb_init(sb);
4107 if (err) {
4108 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4109 err);
dcf2d804 4110 goto failed_mount5;
c2774d84
AK		 4111 }
4112
d5e03cbb
TT		 4113 block = ext4_count_free_clusters(sb);
4114 ext4_free_blocks_count_set(sbi->s_es,
4115 EXT4_C2B(sbi, block));
908c7f19
TH		 4116 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4117 GFP_KERNEL);
d5e03cbb
TT		 4118 if (!err) {
4119 unsigned long freei = ext4_count_free_inodes(sb);
4120 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4121 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4122 GFP_KERNEL);
d5e03cbb
TT		 4123 }
4124 if (!err)
4125 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4126 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4127 if (!err)
908c7f19
TH		 4128 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4129 GFP_KERNEL);
d5e03cbb
TT		 4130 if (err) {
4131 ext4_msg(sb, KERN_ERR, "insufficient memory");
4132 goto failed_mount6;
4133 }
4134
4135 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4136 if (!ext4_fill_flex_info(sb)) {
4137 ext4_msg(sb, KERN_ERR,
4138 "unable to initialize "
4139 "flex_bg meta info!");
4140 goto failed_mount6;
4141 }
4142
bfff6873
LC		 4143 err = ext4_register_li_request(sb, first_not_zeroed);
4144 if (err)
dcf2d804 4145 goto failed_mount6;
bfff6873 4146
3197ebdb
TT		 4147 sbi->s_kobj.kset = ext4_kset;
4148 init_completion(&sbi->s_kobj_unregister);
4149 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4150 "%s", sb->s_id);
dcf2d804
TM		 4151 if (err)
4152 goto failed_mount7;
3197ebdb 4153
9b2ff357
JK		 4154#ifdef CONFIG_QUOTA
4155 /* Enable quota usage during mount. */
4156 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4157 !(sb->s_flags & MS_RDONLY)) {
4158 err = ext4_enable_quotas(sb);
4159 if (err)
4160 goto failed_mount8;
4161 }
4162#endif /* CONFIG_QUOTA */
4163
617ba13b
MC		 4164 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4165 ext4_orphan_cleanup(sb, es);
4166 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4167 if (needs_recovery) {
b31e1552 4168 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4169 ext4_mark_recovery_complete(sb, es);
4170 }
4171 if (EXT4_SB(sb)->s_journal) {
4172 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4173 descr = " journalled data mode";
4174 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4175 descr = " ordered data mode";
4176 else
4177 descr = " writeback data mode";
4178 } else
4179 descr = "out journal";
4180
79add3a3
LC		 4181 if (test_opt(sb, DISCARD)) {
4182 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4183 if (!blk_queue_discard(q))
4184 ext4_msg(sb, KERN_WARNING,
4185 "mounting with \"discard\" option, but "
4186 "the device does not support discard");
4187 }
4188
d4c402d9 4189 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4190 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4191 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4192
66e61a9e
TT		 4193 if (es->s_error_count)
4194 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4195
efbed4dc
TT		 4196 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4197 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4198 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4199 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4200
d4c402d9 4201 kfree(orig_data);
ac27a0ec
DK		 4202 return 0;
4203
617ba13b 4204cantfind_ext4:
ac27a0ec 4205 if (!silent)
b31e1552 4206 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4207 goto failed_mount;
4208
72ba7450
TT		 4209#ifdef CONFIG_QUOTA
4210failed_mount8:
4211 kobject_del(&sbi->s_kobj);
4212#endif
dcf2d804
TM		 4213failed_mount7:
4214 ext4_unregister_li_request(sb);
4215failed_mount6:
f9ae9cf5 4216 ext4_mb_release(sb);
d5e03cbb 4217 if (sbi->s_flex_groups)
b93b41d4 4218 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4219 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4220 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4221 percpu_counter_destroy(&sbi->s_dirs_counter);
4222 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4223failed_mount5:
f9ae9cf5
TT		 4224 ext4_ext_release(sb);
4225 ext4_release_system_zone(sb);
4226failed_mount4a:
94bf608a 4227 dput(sb->s_root);
32a9bb57 4228 sb->s_root = NULL;
94bf608a 4229failed_mount4:
b31e1552 4230 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4231 if (EXT4_SB(sb)->rsv_conversion_wq)
4232 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4233failed_mount_wq:
0390131b
FM		 4234 if (sbi->s_journal) {
4235 jbd2_journal_destroy(sbi->s_journal);
4236 sbi->s_journal = NULL;
4237 }
50460fe8 4238failed_mount3a:
d3922a77 4239 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4240failed_mount3:
9105bb14 4241 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4242 if (sbi->s_mmp_tsk)
4243 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4244failed_mount2:
4245 for (i = 0; i < db_count; i++)
4246 brelse(sbi->s_group_desc[i]);
b93b41d4 4247 kvfree(sbi->s_group_desc);
ac27a0ec 4248failed_mount:
0441984a
DW		 4249 if (sbi->s_chksum_driver)
4250 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4251 if (sbi->s_proc) {
66acdcf4 4252 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4253 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4254 }
ac27a0ec 4255#ifdef CONFIG_QUOTA
a2d4a646 4256 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4257 kfree(sbi->s_qf_names[i]);
4258#endif
617ba13b 4259 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4260 brelse(bh);
4261out_fail:
4262 sb->s_fs_info = NULL;
f6830165 4263 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4264 kfree(sbi);
dcc7dae3 4265out_free_orig:
d4c402d9 4266 kfree(orig_data);
07aa2ea1 4267 return err ? err : ret;
ac27a0ec
DK		 4268}
4269
4270/*
4271 * Setup any per-fs journal parameters now. We'll do this both on
4272 * initial mount, once the journal has been initialised but before we've
4273 * done any recovery; and again on any subsequent remount.
4274 */
617ba13b 4275static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4276{
617ba13b 4277 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4278
30773840
TT		 4279 journal->j_commit_interval = sbi->s_commit_interval;
4280 journal->j_min_batch_time = sbi->s_min_batch_time;
4281 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4282
a931da6a 4283 write_lock(&journal->j_state_lock);
ac27a0ec 4284 if (test_opt(sb, BARRIER))
dab291af 4285 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4286 else
dab291af 4287 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4288 if (test_opt(sb, DATA_ERR_ABORT))
4289 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4290 else
4291 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4292 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4293}
4294
617ba13b 4295static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4296 unsigned int journal_inum)
4297{
4298 struct inode *journal_inode;
4299 journal_t *journal;
4300
0390131b
FM		 4301 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4302
ac27a0ec
DK		 4303 /* First, test for the existence of a valid inode on disk. Bad
4304 * things happen if we iget() an unused inode, as the subsequent
4305 * iput() will try to delete it. */
4306
1d1fe1ee
DH		 4307 journal_inode = ext4_iget(sb, journal_inum);
4308 if (IS_ERR(journal_inode)) {
b31e1552 4309 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4310 return NULL;
4311 }
4312 if (!journal_inode->i_nlink) {
4313 make_bad_inode(journal_inode);
4314 iput(journal_inode);
b31e1552 4315 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4316 return NULL;
4317 }
4318
e5f8eab8 4319 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4320 journal_inode, journal_inode->i_size);
1d1fe1ee 4321 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4322 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4323 iput(journal_inode);
4324 return NULL;
4325 }
4326
dab291af 4327 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4328 if (!journal) {
b31e1552 4329 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4330 iput(journal_inode);
4331 return NULL;
4332 }
4333 journal->j_private = sb;
617ba13b 4334 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4335 return journal;
4336}
4337
617ba13b 4338static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4339 dev_t j_dev)
4340{
2b2d6d01 4341 struct buffer_head *bh;
ac27a0ec 4342 journal_t *journal;
617ba13b
MC		 4343 ext4_fsblk_t start;
4344 ext4_fsblk_t len;
ac27a0ec 4345 int hblock, blocksize;
617ba13b 4346 ext4_fsblk_t sb_block;
ac27a0ec 4347 unsigned long offset;
2b2d6d01 4348 struct ext4_super_block *es;
ac27a0ec
DK		 4349 struct block_device *bdev;
4350
0390131b
FM		 4351 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4352
b31e1552 4353 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4354 if (bdev == NULL)
4355 return NULL;
4356
ac27a0ec 4357 blocksize = sb->s_blocksize;
e1defc4f 4358 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4359 if (blocksize < hblock) {
b31e1552
ES		 4360 ext4_msg(sb, KERN_ERR,
4361 "blocksize too small for journal device");
ac27a0ec
DK		 4362 goto out_bdev;
4363 }
4364
617ba13b
MC		 4365 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4366 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4367 set_blocksize(bdev, blocksize);
4368 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4369 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4370 "external journal");
ac27a0ec
DK		 4371 goto out_bdev;
4372 }
4373
2716b802 4374 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4375 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4376 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4377 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4378 ext4_msg(sb, KERN_ERR, "external journal has "
4379 "bad superblock");
ac27a0ec
DK		 4380 brelse(bh);
4381 goto out_bdev;
4382 }
4383
df4763be
DW		 4384 if ((le32_to_cpu(es->s_feature_ro_compat) &
4385 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4386 es->s_checksum != ext4_superblock_csum(sb, es)) {
4387 ext4_msg(sb, KERN_ERR, "external journal has "
4388 "corrupt superblock");
4389 brelse(bh);
4390 goto out_bdev;
4391 }
4392
617ba13b 4393 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4394 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4395 brelse(bh);
4396 goto out_bdev;
4397 }
4398
bd81d8ee 4399 len = ext4_blocks_count(es);
ac27a0ec
DK		 4400 start = sb_block + 1;
4401 brelse(bh); /* we're done with the superblock */
4402
dab291af 4403 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4404 start, len, blocksize);
4405 if (!journal) {
b31e1552 4406 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4407 goto out_bdev;
4408 }
4409 journal->j_private = sb;
9f203507 4410 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4411 wait_on_buffer(journal->j_sb_buffer);
4412 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4413 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4414 goto out_journal;
4415 }
4416 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4417 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4418 "user (unsupported) - %d",
ac27a0ec
DK		 4419 be32_to_cpu(journal->j_superblock->s_nr_users));
4420 goto out_journal;
4421 }
617ba13b
MC		 4422 EXT4_SB(sb)->journal_bdev = bdev;
4423 ext4_init_journal_params(sb, journal);
ac27a0ec 4424 return journal;
0b8e58a1 4425
ac27a0ec 4426out_journal:
dab291af 4427 jbd2_journal_destroy(journal);
ac27a0ec 4428out_bdev:
617ba13b 4429 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4430 return NULL;
4431}
4432
617ba13b
MC		 4433static int ext4_load_journal(struct super_block *sb,
4434 struct ext4_super_block *es,
ac27a0ec
DK		 4435 unsigned long journal_devnum)
4436{
4437 journal_t *journal;
4438 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4439 dev_t journal_dev;
4440 int err = 0;
4441 int really_read_only;
4442
0390131b
FM		 4443 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4444
ac27a0ec
DK		 4445 if (journal_devnum &&
4446 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4447 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4448 "numbers have changed");
ac27a0ec
DK		 4449 journal_dev = new_decode_dev(journal_devnum);
4450 } else
4451 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4452
4453 really_read_only = bdev_read_only(sb->s_bdev);
4454
4455 /*
4456 * Are we loading a blank journal or performing recovery after a
4457 * crash? For recovery, we need to check in advance whether we
4458 * can get read-write access to the device.
4459 */
617ba13b 4460 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4461 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4462 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4463 "required on readonly filesystem");
ac27a0ec 4464 if (really_read_only) {
b31e1552
ES		 4465 ext4_msg(sb, KERN_ERR, "write access "
4466 "unavailable, cannot proceed");
ac27a0ec
DK		 4467 return -EROFS;
4468 }
b31e1552
ES		 4469 ext4_msg(sb, KERN_INFO, "write access will "
4470 "be enabled during recovery");
ac27a0ec
DK		 4471 }
4472 }
4473
4474 if (journal_inum && journal_dev) {
b31e1552
ES		 4475 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4476 "and inode journals!");
ac27a0ec
DK		 4477 return -EINVAL;
4478 }
4479
4480 if (journal_inum) {
617ba13b 4481 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4482 return -EINVAL;
4483 } else {
617ba13b 4484 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4485 return -EINVAL;
4486 }
4487
90576c0b 4488 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4489 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4490
617ba13b 4491 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4492 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4493 if (!err) {
4494 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4495 if (save)
4496 memcpy(save, ((char *) es) +
4497 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4498 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4499 if (save)
4500 memcpy(((char *) es) + EXT4_S_ERR_START,
4501 save, EXT4_S_ERR_LEN);
4502 kfree(save);
4503 }
ac27a0ec
DK		 4504
4505 if (err) {
b31e1552 4506 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4507 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4508 return err;
4509 }
4510
617ba13b
MC		 4511 EXT4_SB(sb)->s_journal = journal;
4512 ext4_clear_journal_err(sb, es);
ac27a0ec 4513
c41303ce 4514 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4515 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4516 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4517
4518 /* Make sure we flush the recovery flag to disk. */
e2d67052 4519 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4520 }
4521
4522 return 0;
4523}
4524
e2d67052 4525static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4526{
e2d67052 4527 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4528 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4529 int error = 0;
ac27a0ec 4530
7c2e7087 4531 if (!sbh || block_device_ejected(sb))
c4be0c1d 4532 return error;
914258bf
TT		 4533 if (buffer_write_io_error(sbh)) {
4534 /*
4535 * Oh, dear. A previous attempt to write the
4536 * superblock failed. This could happen because the
4537 * USB device was yanked out. Or it could happen to
4538 * be a transient write error and maybe the block will
4539 * be remapped. Nothing we can do but to retry the
4540 * write and hope for the best.
4541 */
b31e1552
ES		 4542 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4543 "superblock detected");
914258bf
TT		 4544 clear_buffer_write_io_error(sbh);
4545 set_buffer_uptodate(sbh);
4546 }
71290b36
TT		 4547 /*
4548 * If the file system is mounted read-only, don't update the
4549 * superblock write time. This avoids updating the superblock
4550 * write time when we are mounting the root file system
4551 * read/only but we need to replay the journal; at that point,
4552 * for people who are east of GMT and who make their clock
4553 * tick in localtime for Windows bug-for-bug compatibility,
4554 * the clock is set in the future, and this will cause e2fsck
4555 * to complain and force a full file system check.
4556 */
4557 if (!(sb->s_flags & MS_RDONLY))
4558 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4559 if (sb->s_bdev->bd_part)
4560 es->s_kbytes_written =
4561 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4562 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4563 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4564 else
4565 es->s_kbytes_written =
4566 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4567 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4568 ext4_free_blocks_count_set(es,
57042651
TT		 4569 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4570 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4571 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4572 es->s_free_inodes_count =
4573 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4574 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4575 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4576 ext4_superblock_csum_set(sb);
ac27a0ec 4577 mark_buffer_dirty(sbh);
914258bf 4578 if (sync) {
c4be0c1d
TS		 4579 error = sync_dirty_buffer(sbh);
4580 if (error)
4581 return error;
4582
4583 error = buffer_write_io_error(sbh);
4584 if (error) {
b31e1552
ES		 4585 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4586 "superblock");
914258bf
TT		 4587 clear_buffer_write_io_error(sbh);
4588 set_buffer_uptodate(sbh);
4589 }
4590 }
c4be0c1d 4591 return error;
ac27a0ec
DK		 4592}
4593
ac27a0ec
DK		 4594/*
4595 * Have we just finished recovery? If so, and if we are mounting (or
4596 * remounting) the filesystem readonly, then we will end up with a
4597 * consistent fs on disk. Record that fact.
4598 */
2b2d6d01
TT		 4599static void ext4_mark_recovery_complete(struct super_block *sb,
4600 struct ext4_super_block *es)
ac27a0ec 4601{
617ba13b 4602 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4603
0390131b
FM		 4604 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4605 BUG_ON(journal != NULL);
4606 return;
4607 }
dab291af 4608 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4609 if (jbd2_journal_flush(journal) < 0)
4610 goto out;
4611
617ba13b 4612 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4613 sb->s_flags & MS_RDONLY) {
617ba13b 4614 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4615 ext4_commit_super(sb, 1);
ac27a0ec 4616 }
7ffe1ea8
HK		 4617
4618out:
dab291af 4619 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4620}
4621
4622/*
4623 * If we are mounting (or read-write remounting) a filesystem whose journal
4624 * has recorded an error from a previous lifetime, move that error to the
4625 * main filesystem now.
4626 */
2b2d6d01
TT		 4627static void ext4_clear_journal_err(struct super_block *sb,
4628 struct ext4_super_block *es)
ac27a0ec
DK		 4629{
4630 journal_t *journal;
4631 int j_errno;
4632 const char *errstr;
4633
0390131b
FM		 4634 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4635
617ba13b 4636 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4637
4638 /*
4639 * Now check for any error status which may have been recorded in the
617ba13b 4640 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4641 */
4642
dab291af 4643 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4644 if (j_errno) {
4645 char nbuf[16];
4646
617ba13b 4647 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4648 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4649 "from previous mount: %s", errstr);
12062ddd 4650 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4651
617ba13b
MC		 4652 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4653 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4654 ext4_commit_super(sb, 1);
ac27a0ec 4655
dab291af 4656 jbd2_journal_clear_err(journal);
d796c52e 4657 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4658 }
4659}
4660
4661/*
4662 * Force the running and committing transactions to commit,
4663 * and wait on the commit.
4664 */
617ba13b 4665int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4666{
4667 journal_t *journal;
ac27a0ec
DK		 4668
4669 if (sb->s_flags & MS_RDONLY)
4670 return 0;
4671
617ba13b 4672 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4673 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4674}
4675
617ba13b 4676static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4677{
14ce0cb4 4678 int ret = 0;
9eddacf9 4679 tid_t target;
06a407f1 4680 bool needs_barrier = false;
8d5d02e6 4681 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4682
9bffad1e 4683 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4684 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4685 /*
4686 * Writeback quota in non-journalled quota case - journalled quota has
4687 * no dirty dquots
4688 */
4689 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4690 /*
4691 * Data writeback is possible w/o journal transaction, so barrier must
4692 * being sent at the end of the function. But we can skip it if
4693 * transaction_commit will do it for us.
4694 */
bda32530
TT		 4695 if (sbi->s_journal) {
4696 target = jbd2_get_latest_transaction(sbi->s_journal);
4697 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4698 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4699 needs_barrier = true;
4700
4701 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4702 if (wait)
4703 ret = jbd2_log_wait_commit(sbi->s_journal,
4704 target);
4705 }
4706 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4707 needs_barrier = true;
06a407f1
DM		 4708 if (needs_barrier) {
4709 int err;
4710 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4711 if (!ret)
4712 ret = err;
0390131b 4713 }
06a407f1
DM		 4714
4715 return ret;
4716}
4717
ac27a0ec
DK		 4718/*
4719 * LVM calls this function before a (read-only) snapshot is created. This
4720 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4721 *
4722 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4723 * state independently. It relies on upper layer to stop all data & metadata
4724 * modifications.
ac27a0ec 4725 */
c4be0c1d 4726static int ext4_freeze(struct super_block *sb)
ac27a0ec 4727{
c4be0c1d
TS		 4728 int error = 0;
4729 journal_t *journal;
ac27a0ec 4730
9ca92389
TT		 4731 if (sb->s_flags & MS_RDONLY)
4732 return 0;
ac27a0ec 4733
9ca92389 4734 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4735
bb044576
TT		 4736 if (journal) {
4737 /* Now we set up the journal barrier. */
4738 jbd2_journal_lock_updates(journal);
ac27a0ec 4739
bb044576
TT		 4740 /*
4741 * Don't clear the needs_recovery flag if we failed to
4742 * flush the journal.
4743 */
4744 error = jbd2_journal_flush(journal);
4745 if (error < 0)
4746 goto out;
4747 }
9ca92389
TT		 4748
4749 /* Journal blocked and flushed, clear needs_recovery flag. */
4750 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4751 error = ext4_commit_super(sb, 1);
6b0310fb 4752out:
bb044576
TT		 4753 if (journal)
4754 /* we rely on upper layer to stop further updates */
4755 jbd2_journal_unlock_updates(journal);
6b0310fb 4756 return error;
ac27a0ec
DK		 4757}
4758
4759/*
4760 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4761 * flag here, even though the filesystem is not technically dirty yet.
4762 */
c4be0c1d 4763static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4764{
9ca92389
TT		 4765 if (sb->s_flags & MS_RDONLY)
4766 return 0;
4767
9ca92389
TT		 4768 /* Reset the needs_recovery flag before the fs is unlocked. */
4769 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4770 ext4_commit_super(sb, 1);
c4be0c1d 4771 return 0;
ac27a0ec
DK		 4772}
4773
673c6100
TT		 4774/*
4775 * Structure to save mount options for ext4_remount's benefit
4776 */
4777struct ext4_mount_options {
4778 unsigned long s_mount_opt;
a2595b8a 4779 unsigned long s_mount_opt2;
08cefc7a
EB		 4780 kuid_t s_resuid;
4781 kgid_t s_resgid;
673c6100
TT		 4782 unsigned long s_commit_interval;
4783 u32 s_min_batch_time, s_max_batch_time;
4784#ifdef CONFIG_QUOTA
4785 int s_jquota_fmt;
a2d4a646 4786 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4787#endif
4788};
4789
2b2d6d01 4790static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4791{
2b2d6d01 4792 struct ext4_super_block *es;
617ba13b 4793 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4794 unsigned long old_sb_flags;
617ba13b 4795 struct ext4_mount_options old_opts;
c79d967d 4796 int enable_quota = 0;
8a266467 4797 ext4_group_t g;
b3881f74 4798 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4799 int err = 0;
ac27a0ec 4800#ifdef CONFIG_QUOTA
03dafb5f 4801 int i, j;
ac27a0ec 4802#endif
d4c402d9 4803 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4804
4805 /* Store the original options */
4806 old_sb_flags = sb->s_flags;
4807 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4808 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4809 old_opts.s_resuid = sbi->s_resuid;
4810 old_opts.s_resgid = sbi->s_resgid;
4811 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4812 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4813 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4814#ifdef CONFIG_QUOTA
4815 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4816 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4817 if (sbi->s_qf_names[i]) {
4818 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4819 GFP_KERNEL);
4820 if (!old_opts.s_qf_names[i]) {
4821 for (j = 0; j < i; j++)
4822 kfree(old_opts.s_qf_names[j]);
3e36a163 4823 kfree(orig_data);
03dafb5f
CG		 4824 return -ENOMEM;
4825 }
4826 } else
4827 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4828#endif
b3881f74
TT		 4829 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4830 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4831
4832 /*
4833 * Allow the "check" option to be passed as a remount option.
4834 */
661aa520 4835 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4836 err = -EINVAL;
4837 goto restore_opts;
4838 }
4839
6b992ff2
DW		 4840 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4841 test_opt(sb, JOURNAL_CHECKSUM)) {
4842 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4843 "during remount not supported");
4844 err = -EINVAL;
4845 goto restore_opts;
4846 }
4847
6ae6514b
PS		 4848 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4849 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4850 ext4_msg(sb, KERN_ERR, "can't mount with "
4851 "both data=journal and delalloc");
4852 err = -EINVAL;
4853 goto restore_opts;
4854 }
4855 if (test_opt(sb, DIOREAD_NOLOCK)) {
4856 ext4_msg(sb, KERN_ERR, "can't mount with "
4857 "both data=journal and dioread_nolock");
4858 err = -EINVAL;
4859 goto restore_opts;
4860 }
4861 }
4862
4ab2f15b 4863 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4864 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4865
4866 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4867 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4868
4869 es = sbi->s_es;
4870
b3881f74 4871 if (sbi->s_journal) {
0390131b 4872 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4873 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4874 }
ac27a0ec 4875
661aa520 4876 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4877 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4878 err = -EROFS;
4879 goto restore_opts;
4880 }
4881
4882 if (*flags & MS_RDONLY) {
38c03b34
TT		 4883 err = sync_filesystem(sb);
4884 if (err < 0)
4885 goto restore_opts;
0f0dd62f
CH		 4886 err = dquot_suspend(sb, -1);
4887 if (err < 0)
c79d967d 4888 goto restore_opts;
c79d967d 4889
ac27a0ec
DK		 4890 /*
4891 * First of all, the unconditional stuff we have to do
4892 * to disable replay of the journal when we next remount
4893 */
4894 sb->s_flags |= MS_RDONLY;
4895
4896 /*
4897 * OK, test if we are remounting a valid rw partition
4898 * readonly, and if so set the rdonly flag and then
4899 * mark the partition as valid again.
4900 */
617ba13b
MC		 4901 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4902 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4903 es->s_state = cpu_to_le16(sbi->s_mount_state);
4904
a63c9eb2 4905 if (sbi->s_journal)
0390131b 4906 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4907 } else {
a13fb1a4
ES		 4908 /* Make sure we can mount this feature set readwrite */
4909 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4910 err = -EROFS;
4911 goto restore_opts;
4912 }
8a266467
TT		 4913 /*
4914 * Make sure the group descriptor checksums
0b8e58a1 4915 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4916 */
4917 for (g = 0; g < sbi->s_groups_count; g++) {
4918 struct ext4_group_desc *gdp =
4919 ext4_get_group_desc(sb, g, NULL);
4920
feb0ab32 4921 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4922 ext4_msg(sb, KERN_ERR,
4923 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4924 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4925 le16_to_cpu(gdp->bg_checksum));
4926 err = -EINVAL;
4927 goto restore_opts;
4928 }
4929 }
4930
ead6596b
ES		 4931 /*
4932 * If we have an unprocessed orphan list hanging
4933 * around from a previously readonly bdev mount,
4934 * require a full umount/remount for now.
4935 */
4936 if (es->s_last_orphan) {
b31e1552 4937 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4938 "remount RDWR because of unprocessed "
4939 "orphan inode list. Please "
b31e1552 4940 "umount/remount instead");
ead6596b
ES		 4941 err = -EINVAL;
4942 goto restore_opts;
4943 }
4944
ac27a0ec
DK		 4945 /*
4946 * Mounting a RDONLY partition read-write, so reread
4947 * and store the current valid flag. (It may have
4948 * been changed by e2fsck since we originally mounted
4949 * the partition.)
4950 */
0390131b
FM		 4951 if (sbi->s_journal)
4952 ext4_clear_journal_err(sb, es);
ac27a0ec 4953 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4954 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4955 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4956 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4957 EXT4_FEATURE_INCOMPAT_MMP))
4958 if (ext4_multi_mount_protect(sb,
4959 le64_to_cpu(es->s_mmp_block))) {
4960 err = -EROFS;
4961 goto restore_opts;
4962 }
c79d967d 4963 enable_quota = 1;
ac27a0ec
DK		 4964 }
4965 }
bfff6873
LC		 4966
4967 /*
4968 * Reinitialize lazy itable initialization thread based on
4969 * current settings
4970 */
4971 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4972 ext4_unregister_li_request(sb);
4973 else {
4974 ext4_group_t first_not_zeroed;
4975 first_not_zeroed = ext4_has_uninit_itable(sb);
4976 ext4_register_li_request(sb, first_not_zeroed);
4977 }
4978
6fd058f7 4979 ext4_setup_system_zone(sb);
d096ad0f 4980 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4981 ext4_commit_super(sb, 1);
0390131b 4982
ac27a0ec
DK		 4983#ifdef CONFIG_QUOTA
4984 /* Release old quota file names */
a2d4a646 4985 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 4986 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4987 if (enable_quota) {
4988 if (sb_any_quota_suspended(sb))
4989 dquot_resume(sb, -1);
4990 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4991 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4992 err = ext4_enable_quotas(sb);
07724f98 4993 if (err)
7c319d32 4994 goto restore_opts;
7c319d32
AK		 4995 }
4996 }
ac27a0ec 4997#endif
d4c402d9
CW		 4998
4999 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5000 kfree(orig_data);
ac27a0ec 5001 return 0;
0b8e58a1 5002
ac27a0ec
DK		 5003restore_opts:
5004 sb->s_flags = old_sb_flags;
5005 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5006 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5007 sbi->s_resuid = old_opts.s_resuid;
5008 sbi->s_resgid = old_opts.s_resgid;
5009 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5010 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5011 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5012#ifdef CONFIG_QUOTA
5013 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5014 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5015 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5016 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5017 }
5018#endif
d4c402d9 5019 kfree(orig_data);
ac27a0ec
DK		 5020 return err;
5021}
5022
2b2d6d01 5023static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5024{
5025 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5026 struct ext4_sb_info *sbi = EXT4_SB(sb);
5027 struct ext4_super_block *es = sbi->s_es;
27dd4385 5028 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5029 u64 fsid;
d02a9391 5030 s64 bfree;
27dd4385 5031 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5032
952fc18e
TT		 5033 if (!test_opt(sb, MINIX_DF))
5034 overhead = sbi->s_overhead;
ac27a0ec 5035
617ba13b 5036 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5037 buf->f_bsize = sb->s_blocksize;
b72f78cb 5038 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5039 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5040 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5041 /* prevent underflow in case that few free space is available */
57042651 5042 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5043 buf->f_bavail = buf->f_bfree -
5044 (ext4_r_blocks_count(es) + resv_blocks);
5045 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5046 buf->f_bavail = 0;
5047 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5048 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5049 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5050 fsid = le64_to_cpup((void *)es->s_uuid) ^
5051 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5052 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5053 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5054
ac27a0ec
DK		 5055 return 0;
5056}
5057
0b8e58a1
AD		 5058/* Helper function for writing quotas on sync - we need to start transaction
5059 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 5060 * Process 1 Process 2
617ba13b 5061 * ext4_create() quota_sync()
a269eb18 5062 * jbd2_journal_start() write_dquot()
871a2931 5063 * dquot_initialize() down(dqio_mutex)
dab291af 5064 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 5065 *
5066 */
5067
5068#ifdef CONFIG_QUOTA
5069
5070static inline struct inode *dquot_to_inode(struct dquot *dquot)
5071{
4c376dca 5072 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5073}
5074
617ba13b 5075static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5076{
5077 int ret, err;
5078 handle_t *handle;
5079 struct inode *inode;
5080
5081 inode = dquot_to_inode(dquot);
9924a92a 5082 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5083 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5084 if (IS_ERR(handle))
5085 return PTR_ERR(handle);
5086 ret = dquot_commit(dquot);
617ba13b 5087 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5088 if (!ret)
5089 ret = err;
5090 return ret;
5091}
5092
617ba13b 5093static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5094{
5095 int ret, err;
5096 handle_t *handle;
5097
9924a92a 5098 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5099 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5100 if (IS_ERR(handle))
5101 return PTR_ERR(handle);
5102 ret = dquot_acquire(dquot);
617ba13b 5103 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5104 if (!ret)
5105 ret = err;
5106 return ret;
5107}
5108
617ba13b 5109static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5110{
5111 int ret, err;
5112 handle_t *handle;
5113
9924a92a 5114 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5115 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5116 if (IS_ERR(handle)) {
5117 /* Release dquot anyway to avoid endless cycle in dqput() */
5118 dquot_release(dquot);
ac27a0ec 5119 return PTR_ERR(handle);
9c3013e9 5120 }
ac27a0ec 5121 ret = dquot_release(dquot);
617ba13b 5122 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5123 if (!ret)
5124 ret = err;
5125 return ret;
5126}
5127
617ba13b 5128static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5129{
262b4662
JK		 5130 struct super_block *sb = dquot->dq_sb;
5131 struct ext4_sb_info *sbi = EXT4_SB(sb);
5132
2c8be6b2 5133 /* Are we journaling quotas? */
262b4662
JK		 5134 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5135 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5136 dquot_mark_dquot_dirty(dquot);
617ba13b 5137 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5138 } else {
5139 return dquot_mark_dquot_dirty(dquot);
5140 }
5141}
5142
617ba13b 5143static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5144{
5145 int ret, err;
5146 handle_t *handle;
5147
5148 /* Data block + inode block */
9924a92a 5149 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5150 if (IS_ERR(handle))
5151 return PTR_ERR(handle);
5152 ret = dquot_commit_info(sb, type);
617ba13b 5153 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5154 if (!ret)
5155 ret = err;
5156 return ret;
5157}
5158
5159/*
5160 * Turn on quotas during mount time - we need to find
5161 * the quota file and such...
5162 */
617ba13b 5163static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5164{
287a8095
CH		 5165 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5166 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5167}
5168
5169/*
5170 * Standard function to be called on quota_on
5171 */
617ba13b 5172static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5173 struct path *path)
ac27a0ec
DK		 5174{
5175 int err;
ac27a0ec
DK		 5176
5177 if (!test_opt(sb, QUOTA))
5178 return -EINVAL;
0623543b 5179
ac27a0ec 5180 /* Quotafile not on the same filesystem? */
d8c9584e 5181 if (path->dentry->d_sb != sb)
ac27a0ec 5182 return -EXDEV;
0623543b
JK		 5183 /* Journaling quota? */
5184 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5185 /* Quotafile not in fs root? */
f00c9e44 5186 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5187 ext4_msg(sb, KERN_WARNING,
5188 "Quota file not on filesystem root. "
5189 "Journaled quota will not work");
2b2d6d01 5190 }
0623543b
JK		 5191
5192 /*
5193 * When we journal data on quota file, we have to flush journal to see
5194 * all updates to the file when we bypass pagecache...
5195 */
0390131b 5196 if (EXT4_SB(sb)->s_journal &&
f00c9e44 5197 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 5198 /*
5199 * We don't need to lock updates but journal_flush() could
5200 * otherwise be livelocked...
5201 */
5202 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5203 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5204 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5205 if (err)
7ffe1ea8 5206 return err;
0623543b
JK		 5207 }
5208
f00c9e44 5209 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 5210}
5211
7c319d32
AK		 5212static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5213 unsigned int flags)
5214{
5215 int err;
5216 struct inode *qf_inode;
a2d4a646 5217 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5218 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5219 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5220 };
5221
5222 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5223
5224 if (!qf_inums[type])
5225 return -EPERM;
5226
5227 qf_inode = ext4_iget(sb, qf_inums[type]);
5228 if (IS_ERR(qf_inode)) {
5229 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5230 return PTR_ERR(qf_inode);
5231 }
5232
bcb13850
JK		 5233 /* Don't account quota for quota files to avoid recursion */
5234 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5235 err = dquot_enable(qf_inode, type, format_id, flags);
5236 iput(qf_inode);
5237
5238 return err;
5239}
5240
5241/* Enable usage tracking for all quota types. */
5242static int ext4_enable_quotas(struct super_block *sb)
5243{
5244 int type, err = 0;
a2d4a646 5245 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5246 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5247 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5248 };
5249
5250 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5251 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5252 if (qf_inums[type]) {
5253 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5254 DQUOT_USAGE_ENABLED);
5255 if (err) {
5256 ext4_warning(sb,
72ba7450
TT		 5257 "Failed to enable quota tracking "
5258 "(type=%d, err=%d). Please run "
5259 "e2fsck to fix.", type, err);
7c319d32
AK		 5260 return err;
5261 }
5262 }
5263 }
5264 return 0;
5265}
5266
5267/*
5268 * quota_on function that is used when QUOTA feature is set.
5269 */
5270static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5271 int format_id)
5272{
5273 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5274 return -EINVAL;
5275
5276 /*
5277 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5278 */
5279 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5280}
5281
ca0e05e4
DM		 5282static int ext4_quota_off(struct super_block *sb, int type)
5283{
21f97697
JK		 5284 struct inode *inode = sb_dqopt(sb)->files[type];
5285 handle_t *handle;
5286
87009d86
DM		 5287 /* Force all delayed allocation blocks to be allocated.
5288 * Caller already holds s_umount sem */
5289 if (test_opt(sb, DELALLOC))
ca0e05e4 5290 sync_filesystem(sb);
ca0e05e4 5291
0b268590
AG		 5292 if (!inode)
5293 goto out;
5294
21f97697
JK		 5295 /* Update modification times of quota files when userspace can
5296 * start looking at them */
9924a92a 5297 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5298 if (IS_ERR(handle))
5299 goto out;
5300 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5301 ext4_mark_inode_dirty(handle, inode);
5302 ext4_journal_stop(handle);
5303
5304out:
ca0e05e4
DM		 5305 return dquot_quota_off(sb, type);
5306}
5307
7c319d32
AK		 5308/*
5309 * quota_off function that is used when QUOTA feature is set.
5310 */
5311static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5312{
5313 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5314 return -EINVAL;
5315
5316 /* Disable only the limits. */
5317 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5318}
5319
ac27a0ec
DK		 5320/* Read data from quotafile - avoid pagecache and such because we cannot afford
5321 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5322 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5323 * we don't have to be afraid of races */
617ba13b 5324static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5325 size_t len, loff_t off)
5326{
5327 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5328 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5329 int offset = off & (sb->s_blocksize - 1);
5330 int tocopy;
5331 size_t toread;
5332 struct buffer_head *bh;
5333 loff_t i_size = i_size_read(inode);
5334
5335 if (off > i_size)
5336 return 0;
5337 if (off+len > i_size)
5338 len = i_size-off;
5339 toread = len;
5340 while (toread > 0) {
5341 tocopy = sb->s_blocksize - offset < toread ?
5342 sb->s_blocksize - offset : toread;
1c215028
TT		 5343 bh = ext4_bread(NULL, inode, blk, 0);
5344 if (IS_ERR(bh))
5345 return PTR_ERR(bh);
ac27a0ec
DK		 5346 if (!bh) /* A hole? */
5347 memset(data, 0, tocopy);
5348 else
5349 memcpy(data, bh->b_data+offset, tocopy);
5350 brelse(bh);
5351 offset = 0;
5352 toread -= tocopy;
5353 data += tocopy;
5354 blk++;
5355 }
5356 return len;
5357}
5358
5359/* Write to quotafile (we know the transaction is already started and has
5360 * enough credits) */
617ba13b 5361static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5362 const char *data, size_t len, loff_t off)
5363{
5364 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5365 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5366 int err, offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5367 struct buffer_head *bh;
5368 handle_t *handle = journal_current_handle();
5369
0390131b 5370 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5371 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5372 " cancelled because transaction is not started",
9c3013e9
JK		 5373 (unsigned long long)off, (unsigned long long)len);
5374 return -EIO;
5375 }
67eeb568
DM		 5376 /*
5377 * Since we account only one data block in transaction credits,
5378 * then it is impossible to cross a block boundary.
5379 */
5380 if (sb->s_blocksize - offset < len) {
5381 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5382 " cancelled because not block aligned",
5383 (unsigned long long)off, (unsigned long long)len);
5384 return -EIO;
5385 }
5386
1c215028
TT		 5387 bh = ext4_bread(handle, inode, blk, 1);
5388 if (IS_ERR(bh))
5389 return PTR_ERR(bh);
67eeb568
DM		 5390 if (!bh)
5391 goto out;
5d601255 5392 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5393 err = ext4_journal_get_write_access(handle, bh);
5394 if (err) {
5395 brelse(bh);
1c215028 5396 return err;
ac27a0ec 5397 }
67eeb568
DM		 5398 lock_buffer(bh);
5399 memcpy(bh->b_data+offset, data, len);
5400 flush_dcache_page(bh->b_page);
5401 unlock_buffer(bh);
62d2b5f2 5402 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5403 brelse(bh);
ac27a0ec 5404out:
67eeb568
DM		 5405 if (inode->i_size < off + len) {
5406 i_size_write(inode, off + len);
617ba13b 5407 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5408 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5409 }
67eeb568 5410 return len;
ac27a0ec
DK		 5411}
5412
5413#endif
5414
152a0836
AV		 5415static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5416 const char *dev_name, void *data)
ac27a0ec 5417{
152a0836 5418 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5419}
5420
37f328eb 5421#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5422static inline void register_as_ext2(void)
5423{
5424 int err = register_filesystem(&ext2_fs_type);
5425 if (err)
5426 printk(KERN_WARNING
5427 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5428}
5429
5430static inline void unregister_as_ext2(void)
5431{
5432 unregister_filesystem(&ext2_fs_type);
5433}
2035e776
TT		 5434
5435static inline int ext2_feature_set_ok(struct super_block *sb)
5436{
5437 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5438 return 0;
5439 if (sb->s_flags & MS_RDONLY)
5440 return 1;
5441 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5442 return 0;
5443 return 1;
5444}
24b58424
TT		 5445#else
5446static inline void register_as_ext2(void) { }
5447static inline void unregister_as_ext2(void) { }
2035e776 5448static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5449#endif
5450
37f328eb 5451#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5452static inline void register_as_ext3(void)
5453{
5454 int err = register_filesystem(&ext3_fs_type);
5455 if (err)
5456 printk(KERN_WARNING
5457 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5458}
5459
5460static inline void unregister_as_ext3(void)
5461{
5462 unregister_filesystem(&ext3_fs_type);
5463}
2035e776
TT		 5464
5465static inline int ext3_feature_set_ok(struct super_block *sb)
5466{
5467 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5468 return 0;
5469 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5470 return 0;
5471 if (sb->s_flags & MS_RDONLY)
5472 return 1;
5473 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5474 return 0;
5475 return 1;
5476}
24b58424
TT		 5477#else
5478static inline void register_as_ext3(void) { }
5479static inline void unregister_as_ext3(void) { }
2035e776 5480static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5481#endif
5482
03010a33
TT		 5483static struct file_system_type ext4_fs_type = {
5484 .owner = THIS_MODULE,
5485 .name = "ext4",
152a0836 5486 .mount = ext4_mount,
03010a33
TT		 5487 .kill_sb = kill_block_super,
5488 .fs_flags = FS_REQUIRES_DEV,
5489};
7f78e035 5490MODULE_ALIAS_FS("ext4");
03010a33 5491
8f021222 5492static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5493{
5494 struct ext4_features *ef;
5495 int ret = -ENOMEM;
5496
5497 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5498 if (!ef)
5499 goto out;
5500
5501 ef->f_kobj.kset = ext4_kset;
5502 init_completion(&ef->f_kobj_unregister);
5503 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5504 "features");
5505 if (ret) {
5506 kfree(ef);
5507 goto out;
5508 }
5509
5510 ext4_feat = ef;
5511 ret = 0;
5512out:
5513 return ret;
5514}
5515
8f021222
LC		 5516static void ext4_exit_feat_adverts(void)
5517{
5518 kobject_put(&ext4_feat->f_kobj);
5519 wait_for_completion(&ext4_feat->f_kobj_unregister);
5520 kfree(ext4_feat);
5521}
5522
e9e3bcec
ES		 5523/* Shared across all ext4 file systems */
5524wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5525struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5526
5dabfc78 5527static int __init ext4_init_fs(void)
ac27a0ec 5528{
e9e3bcec 5529 int i, err;
c9de560d 5530
07c0c5d8
AV		 5531 ext4_li_info = NULL;
5532 mutex_init(&ext4_li_mtx);
5533
9a4c8019 5534 /* Build-time check for flags consistency */
12e9b892 5535 ext4_check_flag_values();
e9e3bcec
ES		 5536
5537 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5538 mutex_init(&ext4__aio_mutex[i]);
5539 init_waitqueue_head(&ext4__ioend_wq[i]);
5540 }
5541
51865fda 5542 err = ext4_init_es();
6fd058f7
TT		 5543 if (err)
5544 return err;
51865fda
ZL		 5545
5546 err = ext4_init_pageio();
5547 if (err)
5548 goto out7;
5549
5dabfc78 5550 err = ext4_init_system_zone();
bd2d0210 5551 if (err)
d44651d0 5552 goto out6;
3197ebdb 5553 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5554 if (!ext4_kset) {
5555 err = -ENOMEM;
dd68314c 5556 goto out5;
0e376b1e 5557 }
d44651d0 5558 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5559
5560 err = ext4_init_feat_adverts();
dd68314c
TT		 5561 if (err)
5562 goto out4;
857ac889 5563
5dabfc78 5564 err = ext4_init_mballoc();
c9de560d
AT		 5565 if (err)
5566 goto out2;
9c191f70
M		 5567 else
5568 ext4_mballoc_ready = 1;
ac27a0ec
DK		 5569 err = init_inodecache();
5570 if (err)
5571 goto out1;
24b58424 5572 register_as_ext3();
2035e776 5573 register_as_ext2();
03010a33 5574 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5575 if (err)
5576 goto out;
bfff6873 5577
ac27a0ec
DK		 5578 return 0;
5579out:
24b58424
TT		 5580 unregister_as_ext2();
5581 unregister_as_ext3();
ac27a0ec
DK		 5582 destroy_inodecache();
5583out1:
9c191f70 5584 ext4_mballoc_ready = 0;
5dabfc78 5585 ext4_exit_mballoc();
9c191f70 5586out2:
8f021222 5587 ext4_exit_feat_adverts();
dd68314c 5588out4:
d44651d0
FJ		 5589 if (ext4_proc_root)
5590 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5591 kset_unregister(ext4_kset);
d44651d0 5592out5:
5dabfc78 5593 ext4_exit_system_zone();
d44651d0 5594out6:
5dabfc78 5595 ext4_exit_pageio();
51865fda
ZL		 5596out7:
5597 ext4_exit_es();
5598
ac27a0ec
DK		 5599 return err;
5600}
5601
5dabfc78 5602static void __exit ext4_exit_fs(void)
ac27a0ec 5603{
bfff6873 5604 ext4_destroy_lazyinit_thread();
24b58424
TT		 5605 unregister_as_ext2();
5606 unregister_as_ext3();
03010a33 5607 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5608 destroy_inodecache();
5dabfc78 5609 ext4_exit_mballoc();
8f021222 5610 ext4_exit_feat_adverts();
9f6200bb 5611 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5612 kset_unregister(ext4_kset);
5dabfc78
TT		 5613 ext4_exit_system_zone();
5614 ext4_exit_pageio();
dd12ed14 5615 ext4_exit_es();
ac27a0ec
DK		 5616}
5617
5618MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5619MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5620MODULE_LICENSE("GPL");
5dabfc78
TT		 5621module_init(ext4_init_fs)
5622module_exit(ext4_exit_fs)