projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: change LRU to round-robin in extent status tree shrinker
[deliverable/linux.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
7abc52c2 41#include <linux/cleancache.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
4a092d73 48#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 49#include "ext4_jbd2.h"
ac27a0ec
DK		 50#include "xattr.h"
51#include "acl.h"
3661d286 52#include "mballoc.h"
ac27a0ec 53
9bffad1e
TT		 54#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
1f109d5a 57static struct proc_dir_entry *ext4_proc_root;
3197ebdb 58static struct kset *ext4_kset;
0b75a840
LC		 59static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
9c191f70 62static int ext4_mballoc_ready;
9f6200bb 63
617ba13b 64static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 65 unsigned long journal_devnum);
2adf6da8 66static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 67static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 68static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
617ba13b 72static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 73static int ext4_remount(struct super_block *sb, int *flags, char *data);
74static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 75static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 76static int ext4_freeze(struct super_block *sb);
152a0836
AV		 77static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
2035e776
TT		 79static inline int ext2_feature_set_ok(struct super_block *sb);
80static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 81static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 82static void ext4_destroy_lazyinit_thread(void);
83static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 84static void ext4_clear_request_list(void);
27dd4385 85static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
ac27a0ec 86
2035e776
TT		 87#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88static struct file_system_type ext2_fs_type = {
89 .owner = THIS_MODULE,
90 .name = "ext2",
91 .mount = ext4_mount,
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
94};
7f78e035 95MODULE_ALIAS_FS("ext2");
fa7614dd 96MODULE_ALIAS("ext2");
2035e776
TT		 97#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98#else
99#define IS_EXT2_SB(sb) (0)
100#endif
101
102
ba69f9ab
JK		 103#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104static struct file_system_type ext3_fs_type = {
105 .owner = THIS_MODULE,
106 .name = "ext3",
152a0836 107 .mount = ext4_mount,
ba69f9ab
JK		 108 .kill_sb = kill_block_super,
109 .fs_flags = FS_REQUIRES_DEV,
110};
7f78e035 111MODULE_ALIAS_FS("ext3");
fa7614dd 112MODULE_ALIAS("ext3");
ba69f9ab
JK		 113#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
114#else
115#define IS_EXT3_SB(sb) (0)
116#endif
bd81d8ee 117
d25425f8
DW		 118static int ext4_verify_csum_type(struct super_block *sb,
119 struct ext4_super_block *es)
120{
121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
122 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
123 return 1;
124
125 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
126}
127
a9c47317
DW		 128static __le32 ext4_superblock_csum(struct super_block *sb,
129 struct ext4_super_block *es)
130{
131 struct ext4_sb_info *sbi = EXT4_SB(sb);
132 int offset = offsetof(struct ext4_super_block, s_checksum);
133 __u32 csum;
134
135 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
136
137 return cpu_to_le32(csum);
138}
139
c197855e
SH		 140static int ext4_superblock_csum_verify(struct super_block *sb,
141 struct ext4_super_block *es)
a9c47317 142{
9aa5d32b 143 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 144 return 1;
145
146 return es->s_checksum == ext4_superblock_csum(sb, es);
147}
148
06db49e6 149void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 150{
06db49e6
TT		 151 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
152
9aa5d32b 153 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 154 return;
155
156 es->s_checksum = ext4_superblock_csum(sb, es);
157}
158
9933fc0a
TT		 159void *ext4_kvmalloc(size_t size, gfp_t flags)
160{
161 void *ret;
162
8be04b93 163 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 164 if (!ret)
165 ret = __vmalloc(size, flags, PAGE_KERNEL);
166 return ret;
167}
168
169void *ext4_kvzalloc(size_t size, gfp_t flags)
170{
171 void *ret;
172
8be04b93 173 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 174 if (!ret)
175 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
176 return ret;
177}
178
8fadc143
AR		 179ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
180 struct ext4_group_desc *bg)
bd81d8ee 181{
3a14589c 182 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 183 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 184 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 185}
186
8fadc143
AR		 187ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
188 struct ext4_group_desc *bg)
bd81d8ee 189{
5272f837 190 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 191 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 192 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 193}
194
8fadc143
AR		 195ext4_fsblk_t ext4_inode_table(struct super_block *sb,
196 struct ext4_group_desc *bg)
bd81d8ee 197{
5272f837 198 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 199 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 200 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 201}
202
021b65bb
TT		 203__u32 ext4_free_group_clusters(struct super_block *sb,
204 struct ext4_group_desc *bg)
560671a0
AK		 205{
206 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 208 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 209}
210
211__u32 ext4_free_inodes_count(struct super_block *sb,
212 struct ext4_group_desc *bg)
213{
214 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 216 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 217}
218
219__u32 ext4_used_dirs_count(struct super_block *sb,
220 struct ext4_group_desc *bg)
221{
222 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 224 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 225}
226
227__u32 ext4_itable_unused_count(struct super_block *sb,
228 struct ext4_group_desc *bg)
229{
230 return le16_to_cpu(bg->bg_itable_unused_lo) |
231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 232 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 233}
234
8fadc143
AR		 235void ext4_block_bitmap_set(struct super_block *sb,
236 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 237{
3a14589c 238 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 239 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
240 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 241}
242
8fadc143
AR		 243void ext4_inode_bitmap_set(struct super_block *sb,
244 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 245{
5272f837 246 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 247 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
248 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 249}
250
8fadc143
AR		 251void ext4_inode_table_set(struct super_block *sb,
252 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 253{
5272f837 254 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 255 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
256 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 257}
258
021b65bb
TT		 259void ext4_free_group_clusters_set(struct super_block *sb,
260 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 261{
262 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
265}
266
267void ext4_free_inodes_set(struct super_block *sb,
268 struct ext4_group_desc *bg, __u32 count)
269{
270 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
273}
274
275void ext4_used_dirs_set(struct super_block *sb,
276 struct ext4_group_desc *bg, __u32 count)
277{
278 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
281}
282
283void ext4_itable_unused_set(struct super_block *sb,
284 struct ext4_group_desc *bg, __u32 count)
285{
286 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
289}
290
d3d1faf6 291
1c13d5c0
TT		 292static void __save_error_info(struct super_block *sb, const char *func,
293 unsigned int line)
294{
295 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
296
297 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
298 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
299 es->s_last_error_time = cpu_to_le32(get_seconds());
300 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
301 es->s_last_error_line = cpu_to_le32(line);
302 if (!es->s_first_error_time) {
303 es->s_first_error_time = es->s_last_error_time;
304 strncpy(es->s_first_error_func, func,
305 sizeof(es->s_first_error_func));
306 es->s_first_error_line = cpu_to_le32(line);
307 es->s_first_error_ino = es->s_last_error_ino;
308 es->s_first_error_block = es->s_last_error_block;
309 }
66e61a9e
TT		 310 /*
311 * Start the daily error reporting function if it hasn't been
312 * started already
313 */
314 if (!es->s_error_count)
315 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 316 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 317}
318
319static void save_error_info(struct super_block *sb, const char *func,
320 unsigned int line)
321{
322 __save_error_info(sb, func, line);
323 ext4_commit_super(sb, 1);
324}
325
7c2e7087
TT		 326/*
327 * The del_gendisk() function uninitializes the disk-specific data
328 * structures, including the bdi structure, without telling anyone
329 * else. Once this happens, any attempt to call mark_buffer_dirty()
330 * (for example, by ext4_commit_super), will cause a kernel OOPS.
331 * This is a kludge to prevent these oops until we can put in a proper
332 * hook in del_gendisk() to inform the VFS and file system layers.
333 */
334static int block_device_ejected(struct super_block *sb)
335{
336 struct inode *bd_inode = sb->s_bdev->bd_inode;
337 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
338
339 return bdi->dev == NULL;
340}
341
18aadd47
BJ		 342static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
343{
344 struct super_block *sb = journal->j_private;
345 struct ext4_sb_info *sbi = EXT4_SB(sb);
346 int error = is_journal_aborted(journal);
5d3ee208 347 struct ext4_journal_cb_entry *jce;
18aadd47 348
5d3ee208 349 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 350 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 351 while (!list_empty(&txn->t_private_list)) {
352 jce = list_entry(txn->t_private_list.next,
353 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 354 list_del_init(&jce->jce_list);
355 spin_unlock(&sbi->s_md_lock);
356 jce->jce_func(sb, jce, error);
357 spin_lock(&sbi->s_md_lock);
358 }
359 spin_unlock(&sbi->s_md_lock);
360}
1c13d5c0 361
ac27a0ec
DK		 362/* Deal with the reporting of failure conditions on a filesystem such as
363 * inconsistencies detected or read IO failures.
364 *
365 * On ext2, we can store the error state of the filesystem in the
617ba13b 366 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 367 * write ordering constraints on the superblock which prevent us from
368 * writing it out straight away; and given that the journal is about to
369 * be aborted, we can't rely on the current, or future, transactions to
370 * write out the superblock safely.
371 *
dab291af 372 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 373 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 374 * that error until we've noted it down and cleared it.
375 */
376
617ba13b 377static void ext4_handle_error(struct super_block *sb)
ac27a0ec 378{
ac27a0ec
DK		 379 if (sb->s_flags & MS_RDONLY)
380 return;
381
2b2d6d01 382 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 383 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 384
4ab2f15b 385 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 386 if (journal)
dab291af 387 jbd2_journal_abort(journal, -EIO);
ac27a0ec 388 }
2b2d6d01 389 if (test_opt(sb, ERRORS_RO)) {
b31e1552 390 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 391 /*
392 * Make sure updated value of ->s_mount_flags will be visible
393 * before ->s_flags update
394 */
395 smp_wmb();
ac27a0ec
DK		 396 sb->s_flags |= MS_RDONLY;
397 }
ac27a0ec 398 if (test_opt(sb, ERRORS_PANIC))
617ba13b 399 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 400 sb->s_id);
401}
402
efbed4dc
TT		 403#define ext4_error_ratelimit(sb) \
404 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
405 "EXT4-fs error")
406
12062ddd 407void __ext4_error(struct super_block *sb, const char *function,
c398eda0 408 unsigned int line, const char *fmt, ...)
ac27a0ec 409{
0ff2ea7d 410 struct va_format vaf;
ac27a0ec
DK		 411 va_list args;
412
efbed4dc
TT		 413 if (ext4_error_ratelimit(sb)) {
414 va_start(args, fmt);
415 vaf.fmt = fmt;
416 vaf.va = &args;
417 printk(KERN_CRIT
418 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
419 sb->s_id, function, line, current->comm, &vaf);
420 va_end(args);
421 }
f3fc0210 422 save_error_info(sb, function, line);
617ba13b 423 ext4_handle_error(sb);
ac27a0ec
DK		 424}
425
e7c96e8e
JP		 426void __ext4_error_inode(struct inode *inode, const char *function,
427 unsigned int line, ext4_fsblk_t block,
428 const char *fmt, ...)
273df556
FM		 429{
430 va_list args;
f7c21177 431 struct va_format vaf;
1c13d5c0 432 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 433
1c13d5c0
TT		 434 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
435 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 436 if (ext4_error_ratelimit(inode->i_sb)) {
437 va_start(args, fmt);
438 vaf.fmt = fmt;
439 vaf.va = &args;
440 if (block)
441 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
442 "inode #%lu: block %llu: comm %s: %pV\n",
443 inode->i_sb->s_id, function, line, inode->i_ino,
444 block, current->comm, &vaf);
445 else
446 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
447 "inode #%lu: comm %s: %pV\n",
448 inode->i_sb->s_id, function, line, inode->i_ino,
449 current->comm, &vaf);
450 va_end(args);
451 }
1c13d5c0 452 save_error_info(inode->i_sb, function, line);
273df556
FM		 453 ext4_handle_error(inode->i_sb);
454}
455
e7c96e8e
JP		 456void __ext4_error_file(struct file *file, const char *function,
457 unsigned int line, ext4_fsblk_t block,
458 const char *fmt, ...)
273df556
FM		 459{
460 va_list args;
f7c21177 461 struct va_format vaf;
1c13d5c0 462 struct ext4_super_block *es;
496ad9aa 463 struct inode *inode = file_inode(file);
273df556
FM		 464 char pathname[80], *path;
465
1c13d5c0
TT		 466 es = EXT4_SB(inode->i_sb)->s_es;
467 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc
TT		 468 if (ext4_error_ratelimit(inode->i_sb)) {
469 path = d_path(&(file->f_path), pathname, sizeof(pathname));
470 if (IS_ERR(path))
471 path = "(unknown)";
472 va_start(args, fmt);
473 vaf.fmt = fmt;
474 vaf.va = &args;
475 if (block)
476 printk(KERN_CRIT
477 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
478 "block %llu: comm %s: path %s: %pV\n",
479 inode->i_sb->s_id, function, line, inode->i_ino,
480 block, current->comm, path, &vaf);
481 else
482 printk(KERN_CRIT
483 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
484 "comm %s: path %s: %pV\n",
485 inode->i_sb->s_id, function, line, inode->i_ino,
486 current->comm, path, &vaf);
487 va_end(args);
488 }
1c13d5c0 489 save_error_info(inode->i_sb, function, line);
273df556
FM		 490 ext4_handle_error(inode->i_sb);
491}
492
722887dd
TT		 493const char *ext4_decode_error(struct super_block *sb, int errno,
494 char nbuf[16])
ac27a0ec
DK		 495{
496 char *errstr = NULL;
497
498 switch (errno) {
499 case -EIO:
500 errstr = "IO failure";
501 break;
502 case -ENOMEM:
503 errstr = "Out of memory";
504 break;
505 case -EROFS:
78f1ddbb
TT		 506 if (!sb || (EXT4_SB(sb)->s_journal &&
507 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 508 errstr = "Journal has aborted";
509 else
510 errstr = "Readonly filesystem";
511 break;
512 default:
513 /* If the caller passed in an extra buffer for unknown
514 * errors, textualise them now. Else we just return
515 * NULL. */
516 if (nbuf) {
517 /* Check for truncated error codes... */
518 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
519 errstr = nbuf;
520 }
521 break;
522 }
523
524 return errstr;
525}
526
617ba13b 527/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 528 * automatically and invokes the appropriate error response. */
529
c398eda0
TT		 530void __ext4_std_error(struct super_block *sb, const char *function,
531 unsigned int line, int errno)
ac27a0ec
DK		 532{
533 char nbuf[16];
534 const char *errstr;
535
536 /* Special case: if the error is EROFS, and we're not already
537 * inside a transaction, then there's really no point in logging
538 * an error. */
539 if (errno == -EROFS && journal_current_handle() == NULL &&
540 (sb->s_flags & MS_RDONLY))
541 return;
542
efbed4dc
TT		 543 if (ext4_error_ratelimit(sb)) {
544 errstr = ext4_decode_error(sb, errno, nbuf);
545 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
546 sb->s_id, function, line, errstr);
547 }
ac27a0ec 548
efbed4dc 549 save_error_info(sb, function, line);
617ba13b 550 ext4_handle_error(sb);
ac27a0ec
DK		 551}
552
553/*
617ba13b 554 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 555 * abort function may be used to deal with unrecoverable failures such
556 * as journal IO errors or ENOMEM at a critical moment in log management.
557 *
558 * We unconditionally force the filesystem into an ABORT|READONLY state,
559 * unless the error response on the fs has been set to panic in which
560 * case we take the easy way out and panic immediately.
561 */
562
c67d859e 563void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 564 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 565{
566 va_list args;
567
1c13d5c0 568 save_error_info(sb, function, line);
ac27a0ec 569 va_start(args, fmt);
c398eda0
TT		 570 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
571 function, line);
ac27a0ec
DK		 572 vprintk(fmt, args);
573 printk("\n");
574 va_end(args);
575
1c13d5c0
TT		 576 if ((sb->s_flags & MS_RDONLY) == 0) {
577 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 578 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 579 /*
580 * Make sure updated value of ->s_mount_flags will be visible
581 * before ->s_flags update
582 */
583 smp_wmb();
584 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 585 if (EXT4_SB(sb)->s_journal)
586 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
587 save_error_info(sb, function, line);
588 }
ac27a0ec 589 if (test_opt(sb, ERRORS_PANIC))
617ba13b 590 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 591}
592
e7c96e8e
JP		 593void __ext4_msg(struct super_block *sb,
594 const char *prefix, const char *fmt, ...)
b31e1552 595{
0ff2ea7d 596 struct va_format vaf;
b31e1552
ES		 597 va_list args;
598
efbed4dc
TT		 599 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
600 return;
601
b31e1552 602 va_start(args, fmt);
0ff2ea7d
JP		 603 vaf.fmt = fmt;
604 vaf.va = &args;
605 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 606 va_end(args);
607}
608
12062ddd 609void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 610 unsigned int line, const char *fmt, ...)
ac27a0ec 611{
0ff2ea7d 612 struct va_format vaf;
ac27a0ec
DK		 613 va_list args;
614
efbed4dc
TT		 615 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
616 "EXT4-fs warning"))
617 return;
618
ac27a0ec 619 va_start(args, fmt);
0ff2ea7d
JP		 620 vaf.fmt = fmt;
621 vaf.va = &args;
622 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
623 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 624 va_end(args);
625}
626
e29136f8
TT		 627void __ext4_grp_locked_error(const char *function, unsigned int line,
628 struct super_block *sb, ext4_group_t grp,
629 unsigned long ino, ext4_fsblk_t block,
630 const char *fmt, ...)
5d1b1b3f
AK		 631__releases(bitlock)
632__acquires(bitlock)
633{
0ff2ea7d 634 struct va_format vaf;
5d1b1b3f
AK		 635 va_list args;
636 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
637
1c13d5c0
TT		 638 es->s_last_error_ino = cpu_to_le32(ino);
639 es->s_last_error_block = cpu_to_le64(block);
640 __save_error_info(sb, function, line);
0ff2ea7d 641
efbed4dc
TT		 642 if (ext4_error_ratelimit(sb)) {
643 va_start(args, fmt);
644 vaf.fmt = fmt;
645 vaf.va = &args;
646 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
647 sb->s_id, function, line, grp);
648 if (ino)
649 printk(KERN_CONT "inode %lu: ", ino);
650 if (block)
651 printk(KERN_CONT "block %llu:",
652 (unsigned long long) block);
653 printk(KERN_CONT "%pV\n", &vaf);
654 va_end(args);
655 }
5d1b1b3f
AK		 656
657 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 658 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 659 return;
660 }
1c13d5c0 661
5d1b1b3f
AK		 662 ext4_unlock_group(sb, grp);
663 ext4_handle_error(sb);
664 /*
665 * We only get here in the ERRORS_RO case; relocking the group
666 * may be dangerous, but nothing bad will happen since the
667 * filesystem will have already been marked read/only and the
668 * journal has been aborted. We return 1 as a hint to callers
669 * who might what to use the return value from
25985edc 670 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 671 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
672 * aggressively from the ext4 function in question, with a
673 * more appropriate error code.
674 */
675 ext4_lock_group(sb, grp);
676 return;
677}
678
617ba13b 679void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 680{
617ba13b 681 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 682
617ba13b 683 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 684 return;
685
12062ddd 686 ext4_warning(sb,
ac27a0ec
DK		 687 "updating to rev %d because of new feature flag, "
688 "running e2fsck is recommended",
617ba13b 689 EXT4_DYNAMIC_REV);
ac27a0ec 690
617ba13b
MC		 691 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
692 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
693 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 694 /* leave es->s_feature_*compat flags alone */
695 /* es->s_uuid will be set by e2fsck if empty */
696
697 /*
698 * The rest of the superblock fields should be zero, and if not it
699 * means they are likely already in use, so leave them alone. We
700 * can leave it up to e2fsck to clean up any inconsistencies there.
701 */
702}
703
704/*
705 * Open the external journal device
706 */
b31e1552 707static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 708{
709 struct block_device *bdev;
710 char b[BDEVNAME_SIZE];
711
d4d77629 712 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 713 if (IS_ERR(bdev))
714 goto fail;
715 return bdev;
716
717fail:
b31e1552 718 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 719 __bdevname(dev, b), PTR_ERR(bdev));
720 return NULL;
721}
722
723/*
724 * Release the journal device
725 */
4385bab1 726static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 727{
4385bab1 728 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 729}
730
4385bab1 731static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 732{
733 struct block_device *bdev;
ac27a0ec
DK		 734 bdev = sbi->journal_bdev;
735 if (bdev) {
4385bab1 736 ext4_blkdev_put(bdev);
ac27a0ec
DK		 737 sbi->journal_bdev = NULL;
738 }
ac27a0ec
DK		 739}
740
741static inline struct inode *orphan_list_entry(struct list_head *l)
742{
617ba13b 743 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 744}
745
617ba13b 746static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 747{
748 struct list_head *l;
749
b31e1552
ES		 750 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
751 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 752
753 printk(KERN_ERR "sb_info orphan list:\n");
754 list_for_each(l, &sbi->s_orphan) {
755 struct inode *inode = orphan_list_entry(l);
756 printk(KERN_ERR " "
757 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
758 inode->i_sb->s_id, inode->i_ino, inode,
759 inode->i_mode, inode->i_nlink,
760 NEXT_ORPHAN(inode));
761 }
762}
763
2b2d6d01 764static void ext4_put_super(struct super_block *sb)
ac27a0ec 765{
617ba13b
MC		 766 struct ext4_sb_info *sbi = EXT4_SB(sb);
767 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 768 int i, err;
ac27a0ec 769
857ac889 770 ext4_unregister_li_request(sb);
e0ccfd95
CH		 771 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
772
2e8fa54e 773 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 774 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 775
0390131b
FM		 776 if (sbi->s_journal) {
777 err = jbd2_journal_destroy(sbi->s_journal);
778 sbi->s_journal = NULL;
779 if (err < 0)
c67d859e 780 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 781 }
d4edac31 782
d3922a77 783 ext4_es_unregister_shrinker(sbi);
9105bb14 784 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 785 ext4_release_system_zone(sb);
786 ext4_mb_release(sb);
787 ext4_ext_release(sb);
788 ext4_xattr_put_super(sb);
789
ac27a0ec 790 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 791 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 792 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 793 }
58c5873a 794 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 795 ext4_commit_super(sb, 1);
796
240799cd 797 if (sbi->s_proc) {
66acdcf4 798 remove_proc_entry("options", sbi->s_proc);
9f6200bb 799 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 800 }
3197ebdb 801 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 802
803 for (i = 0; i < sbi->s_gdb_count; i++)
804 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 805 kvfree(sbi->s_group_desc);
806 kvfree(sbi->s_flex_groups);
57042651 807 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 808 percpu_counter_destroy(&sbi->s_freeinodes_counter);
809 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 810 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 811 brelse(sbi->s_sbh);
812#ifdef CONFIG_QUOTA
a2d4a646 813 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 814 kfree(sbi->s_qf_names[i]);
815#endif
816
817 /* Debugging code just in case the in-memory inode orphan list
818 * isn't empty. The on-disk one can be non-empty if we've
819 * detected an error and taken the fs readonly, but the
820 * in-memory list had better be clean by this point. */
821 if (!list_empty(&sbi->s_orphan))
822 dump_orphan_list(sb, sbi);
823 J_ASSERT(list_empty(&sbi->s_orphan));
824
f98393a6 825 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 826 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
827 /*
828 * Invalidate the journal device's buffers. We don't want them
829 * floating about in memory - the physical journal device may
830 * hotswapped, and it breaks the `ro-after' testing code.
831 */
832 sync_blockdev(sbi->journal_bdev);
f98393a6 833 invalidate_bdev(sbi->journal_bdev);
617ba13b 834 ext4_blkdev_remove(sbi);
ac27a0ec 835 }
9c191f70
M		 836 if (sbi->s_mb_cache) {
837 ext4_xattr_destroy_cache(sbi->s_mb_cache);
838 sbi->s_mb_cache = NULL;
839 }
c5e06d10
JL		 840 if (sbi->s_mmp_tsk)
841 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 842 sb->s_fs_info = NULL;
3197ebdb
TT		 843 /*
844 * Now that we are completely done shutting down the
845 * superblock, we need to actually destroy the kobject.
846 */
3197ebdb
TT		 847 kobject_put(&sbi->s_kobj);
848 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 849 if (sbi->s_chksum_driver)
850 crypto_free_shash(sbi->s_chksum_driver);
705895b6 851 kfree(sbi->s_blockgroup_lock);
ac27a0ec 852 kfree(sbi);
ac27a0ec
DK		 853}
854
e18b890b 855static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 856
857/*
858 * Called inside transaction, so use GFP_NOFS
859 */
617ba13b 860static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 861{
617ba13b 862 struct ext4_inode_info *ei;
ac27a0ec 863
e6b4f8da 864 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 865 if (!ei)
866 return NULL;
0b8e58a1 867
ac27a0ec 868 ei->vfs_inode.i_version = 1;
202ee5df 869 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 870 INIT_LIST_HEAD(&ei->i_prealloc_list);
871 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 872 ext4_es_init_tree(&ei->i_es_tree);
873 rwlock_init(&ei->i_es_lock);
edaa53ca 874 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 875 ei->i_es_all_nr = 0;
edaa53ca 876 ei->i_es_shk_nr = 0;
d2a17637
MC		 877 ei->i_reserved_data_blocks = 0;
878 ei->i_reserved_meta_blocks = 0;
879 ei->i_allocated_meta_blocks = 0;
9d0be502 880 ei->i_da_metadata_calc_len = 0;
7e731bc9 881 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 882 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 883#ifdef CONFIG_QUOTA
884 ei->i_reserved_quota = 0;
885#endif
8aefcd55 886 ei->jinode = NULL;
2e8fa54e 887 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 888 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 889 ei->i_sync_tid = 0;
890 ei->i_datasync_tid = 0;
f7ad6d2e 891 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 892 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 893 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
0b8e58a1 894
ac27a0ec
DK		 895 return &ei->vfs_inode;
896}
897
7ff9c073
TT		 898static int ext4_drop_inode(struct inode *inode)
899{
900 int drop = generic_drop_inode(inode);
901
902 trace_ext4_drop_inode(inode, drop);
903 return drop;
904}
905
fa0d7e3d
NP		 906static void ext4_i_callback(struct rcu_head *head)
907{
908 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 909 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
910}
911
617ba13b 912static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 913{
9f7dd93d 914 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 915 ext4_msg(inode->i_sb, KERN_ERR,
916 "Inode %lu (%p): orphan list check failed!",
917 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 918 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
919 EXT4_I(inode), sizeof(struct ext4_inode_info),
920 true);
921 dump_stack();
922 }
fa0d7e3d 923 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 924}
925
51cc5068 926static void init_once(void *foo)
ac27a0ec 927{
617ba13b 928 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 929
a35afb83 930 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 931 init_rwsem(&ei->xattr_sem);
0e855ac8 932 init_rwsem(&ei->i_data_sem);
a35afb83 933 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 934}
935
e67bc2b3 936static int __init init_inodecache(void)
ac27a0ec 937{
617ba13b
MC		 938 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
939 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 940 0, (SLAB_RECLAIM_ACCOUNT|
941 SLAB_MEM_SPREAD),
20c2df83 942 init_once);
617ba13b 943 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 944 return -ENOMEM;
945 return 0;
946}
947
948static void destroy_inodecache(void)
949{
8c0a8537
KS		 950 /*
951 * Make sure all delayed rcu free inodes are flushed before we
952 * destroy cache.
953 */
954 rcu_barrier();
617ba13b 955 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 956}
957
0930fcc1 958void ext4_clear_inode(struct inode *inode)
ac27a0ec 959{
0930fcc1 960 invalidate_inode_buffers(inode);
dbd5768f 961 clear_inode(inode);
9f754758 962 dquot_drop(inode);
c2ea3fde 963 ext4_discard_preallocations(inode);
51865fda 964 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
edaa53ca 965 ext4_es_list_del(inode);
8aefcd55
TT		 966 if (EXT4_I(inode)->jinode) {
967 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
968 EXT4_I(inode)->jinode);
969 jbd2_free_inode(EXT4_I(inode)->jinode);
970 EXT4_I(inode)->jinode = NULL;
971 }
ac27a0ec
DK		 972}
973
1b961ac0 974static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 975 u64 ino, u32 generation)
ac27a0ec 976{
ac27a0ec 977 struct inode *inode;
ac27a0ec 978
617ba13b 979 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 980 return ERR_PTR(-ESTALE);
617ba13b 981 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 982 return ERR_PTR(-ESTALE);
983
984 /* iget isn't really right if the inode is currently unallocated!!
985 *
617ba13b 986 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 987 * deleted, so we should be safe.
988 *
989 * Currently we don't know the generation for parent directory, so
990 * a generation of 0 means "accept any"
991 */
f4bb2981 992 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 993 if (IS_ERR(inode))
994 return ERR_CAST(inode);
995 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 996 iput(inode);
997 return ERR_PTR(-ESTALE);
998 }
1b961ac0
CH		 999
1000 return inode;
1001}
1002
1003static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1004 int fh_len, int fh_type)
1b961ac0
CH		 1005{
1006 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1007 ext4_nfs_get_inode);
1008}
1009
1010static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1011 int fh_len, int fh_type)
1b961ac0
CH		 1012{
1013 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1014 ext4_nfs_get_inode);
ac27a0ec
DK		 1015}
1016
c39a7f84
TO		 1017/*
1018 * Try to release metadata pages (indirect blocks, directories) which are
1019 * mapped via the block device. Since these pages could have journal heads
1020 * which would prevent try_to_free_buffers() from freeing them, we must use
1021 * jbd2 layer's try_to_free_buffers() function to release them.
1022 */
0b8e58a1
AD		 1023static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1024 gfp_t wait)
c39a7f84
TO		 1025{
1026 journal_t *journal = EXT4_SB(sb)->s_journal;
1027
1028 WARN_ON(PageChecked(page));
1029 if (!page_has_buffers(page))
1030 return 0;
1031 if (journal)
1032 return jbd2_journal_try_to_free_buffers(journal, page,
1033 wait & ~__GFP_WAIT);
1034 return try_to_free_buffers(page);
1035}
1036
ac27a0ec 1037#ifdef CONFIG_QUOTA
af5bc92d 1038#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1039#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1040
617ba13b
MC		 1041static int ext4_write_dquot(struct dquot *dquot);
1042static int ext4_acquire_dquot(struct dquot *dquot);
1043static int ext4_release_dquot(struct dquot *dquot);
1044static int ext4_mark_dquot_dirty(struct dquot *dquot);
1045static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1046static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1047 struct path *path);
7c319d32
AK		 1048static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1049 int format_id);
ca0e05e4 1050static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1051static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1052static int ext4_quota_on_mount(struct super_block *sb, int type);
1053static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1054 size_t len, loff_t off);
617ba13b 1055static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1056 const char *data, size_t len, loff_t off);
7c319d32
AK		 1057static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1058 unsigned int flags);
1059static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1060
61e225dc 1061static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1062 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1063 .write_dquot = ext4_write_dquot,
1064 .acquire_dquot = ext4_acquire_dquot,
1065 .release_dquot = ext4_release_dquot,
1066 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1067 .write_info = ext4_write_info,
1068 .alloc_dquot = dquot_alloc,
1069 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1070};
1071
0d54b217 1072static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1073 .quota_on = ext4_quota_on,
ca0e05e4 1074 .quota_off = ext4_quota_off,
287a8095
CH		 1075 .quota_sync = dquot_quota_sync,
1076 .get_info = dquot_get_dqinfo,
1077 .set_info = dquot_set_dqinfo,
1078 .get_dqblk = dquot_get_dqblk,
1079 .set_dqblk = dquot_set_dqblk
ac27a0ec 1080};
7c319d32
AK		 1081
1082static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1083 .quota_on_meta = ext4_quota_on_sysfile,
1084 .quota_off = ext4_quota_off_sysfile,
1085 .quota_sync = dquot_quota_sync,
1086 .get_info = dquot_get_dqinfo,
1087 .set_info = dquot_set_dqinfo,
1088 .get_dqblk = dquot_get_dqblk,
1089 .set_dqblk = dquot_set_dqblk
1090};
ac27a0ec
DK		 1091#endif
1092
ee9b6d61 1093static const struct super_operations ext4_sops = {
617ba13b
MC		 1094 .alloc_inode = ext4_alloc_inode,
1095 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1096 .write_inode = ext4_write_inode,
1097 .dirty_inode = ext4_dirty_inode,
7ff9c073 1098 .drop_inode = ext4_drop_inode,
0930fcc1 1099 .evict_inode = ext4_evict_inode,
617ba13b 1100 .put_super = ext4_put_super,
617ba13b 1101 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1102 .freeze_fs = ext4_freeze,
1103 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1104 .statfs = ext4_statfs,
1105 .remount_fs = ext4_remount,
617ba13b 1106 .show_options = ext4_show_options,
ac27a0ec 1107#ifdef CONFIG_QUOTA
617ba13b
MC		 1108 .quota_read = ext4_quota_read,
1109 .quota_write = ext4_quota_write,
ac27a0ec 1110#endif
c39a7f84 1111 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1112};
1113
39655164 1114static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1115 .fh_to_dentry = ext4_fh_to_dentry,
1116 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1117 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1118};
1119
1120enum {
1121 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1122 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1123 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1124 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1125 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1126 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1127 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1128 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1129 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1130 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1131 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1132 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1133 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1134 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1135 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1136 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1137 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1138 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
df981d03 1139 Opt_max_dir_size_kb,
ac27a0ec
DK		 1140};
1141
a447c093 1142static const match_table_t tokens = {
ac27a0ec
DK		 1143 {Opt_bsd_df, "bsddf"},
1144 {Opt_minix_df, "minixdf"},
1145 {Opt_grpid, "grpid"},
1146 {Opt_grpid, "bsdgroups"},
1147 {Opt_nogrpid, "nogrpid"},
1148 {Opt_nogrpid, "sysvgroups"},
1149 {Opt_resgid, "resgid=%u"},
1150 {Opt_resuid, "resuid=%u"},
1151 {Opt_sb, "sb=%u"},
1152 {Opt_err_cont, "errors=continue"},
1153 {Opt_err_panic, "errors=panic"},
1154 {Opt_err_ro, "errors=remount-ro"},
1155 {Opt_nouid32, "nouid32"},
ac27a0ec 1156 {Opt_debug, "debug"},
72578c33
TT		 1157 {Opt_removed, "oldalloc"},
1158 {Opt_removed, "orlov"},
ac27a0ec
DK		 1159 {Opt_user_xattr, "user_xattr"},
1160 {Opt_nouser_xattr, "nouser_xattr"},
1161 {Opt_acl, "acl"},
1162 {Opt_noacl, "noacl"},
e3bb52ae 1163 {Opt_noload, "norecovery"},
5a916be1 1164 {Opt_noload, "noload"},
72578c33
TT		 1165 {Opt_removed, "nobh"},
1166 {Opt_removed, "bh"},
ac27a0ec 1167 {Opt_commit, "commit=%u"},
30773840
TT		 1168 {Opt_min_batch_time, "min_batch_time=%u"},
1169 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1170 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1171 {Opt_journal_path, "journal_path=%s"},
818d276c
GS		 1172 {Opt_journal_checksum, "journal_checksum"},
1173 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1174 {Opt_abort, "abort"},
1175 {Opt_data_journal, "data=journal"},
1176 {Opt_data_ordered, "data=ordered"},
1177 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1178 {Opt_data_err_abort, "data_err=abort"},
1179 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1180 {Opt_offusrjquota, "usrjquota="},
1181 {Opt_usrjquota, "usrjquota=%s"},
1182 {Opt_offgrpjquota, "grpjquota="},
1183 {Opt_grpjquota, "grpjquota=%s"},
1184 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1185 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1186 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1187 {Opt_grpquota, "grpquota"},
1188 {Opt_noquota, "noquota"},
1189 {Opt_quota, "quota"},
1190 {Opt_usrquota, "usrquota"},
1191 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1192 {Opt_barrier, "barrier"},
1193 {Opt_nobarrier, "nobarrier"},
25ec56b5 1194 {Opt_i_version, "i_version"},
c9de560d 1195 {Opt_stripe, "stripe=%u"},
64769240 1196 {Opt_delalloc, "delalloc"},
dd919b98 1197 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1198 {Opt_removed, "mblk_io_submit"},
1199 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1200 {Opt_block_validity, "block_validity"},
1201 {Opt_noblock_validity, "noblock_validity"},
240799cd 1202 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1203 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1204 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1205 {Opt_auto_da_alloc, "auto_da_alloc"},
1206 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1207 {Opt_dioread_nolock, "dioread_nolock"},
1208 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1209 {Opt_discard, "discard"},
1210 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1211 {Opt_init_itable, "init_itable=%u"},
1212 {Opt_init_itable, "init_itable"},
1213 {Opt_noinit_itable, "noinit_itable"},
df981d03 1214 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1215 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1216 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1217 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1218 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1219 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1220 {Opt_err, NULL},
ac27a0ec
DK		 1221};
1222
617ba13b 1223static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1224{
617ba13b 1225 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1226 char *options = (char *) *data;
1227
1228 if (!options || strncmp(options, "sb=", 3) != 0)
1229 return 1; /* Default location */
0b8e58a1 1230
ac27a0ec 1231 options += 3;
0b8e58a1 1232 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1233 sb_block = simple_strtoul(options, &options, 0);
1234 if (*options && *options != ',') {
4776004f 1235 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1236 (char *) *data);
1237 return 1;
1238 }
1239 if (*options == ',')
1240 options++;
1241 *data = (void *) options;
0b8e58a1 1242
ac27a0ec
DK		 1243 return sb_block;
1244}
1245
b3881f74 1246#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1247static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1248 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1249
56c50f11
DM		 1250#ifdef CONFIG_QUOTA
1251static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1252{
1253 struct ext4_sb_info *sbi = EXT4_SB(sb);
1254 char *qname;
03dafb5f 1255 int ret = -1;
56c50f11
DM		 1256
1257 if (sb_any_quota_loaded(sb) &&
1258 !sbi->s_qf_names[qtype]) {
1259 ext4_msg(sb, KERN_ERR,
1260 "Cannot change journaled "
1261 "quota options when quota turned on");
57f73c2c 1262 return -1;
56c50f11 1263 }
262b4662
JK		 1264 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1265 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1266 "when QUOTA feature is enabled");
1267 return -1;
1268 }
56c50f11
DM		 1269 qname = match_strdup(args);
1270 if (!qname) {
1271 ext4_msg(sb, KERN_ERR,
1272 "Not enough memory for storing quotafile name");
57f73c2c 1273 return -1;
56c50f11 1274 }
03dafb5f
CG		 1275 if (sbi->s_qf_names[qtype]) {
1276 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1277 ret = 1;
1278 else
1279 ext4_msg(sb, KERN_ERR,
1280 "%s quota file already specified",
1281 QTYPE2NAME(qtype));
1282 goto errout;
56c50f11 1283 }
03dafb5f 1284 if (strchr(qname, '/')) {
56c50f11
DM		 1285 ext4_msg(sb, KERN_ERR,
1286 "quotafile must be on filesystem root");
03dafb5f 1287 goto errout;
56c50f11 1288 }
03dafb5f 1289 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1290 set_opt(sb, QUOTA);
56c50f11 1291 return 1;
03dafb5f
CG		 1292errout:
1293 kfree(qname);
1294 return ret;
56c50f11
DM		 1295}
1296
1297static int clear_qf_name(struct super_block *sb, int qtype)
1298{
1299
1300 struct ext4_sb_info *sbi = EXT4_SB(sb);
1301
1302 if (sb_any_quota_loaded(sb) &&
1303 sbi->s_qf_names[qtype]) {
1304 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1305 " when quota turned on");
57f73c2c 1306 return -1;
56c50f11 1307 }
03dafb5f 1308 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1309 sbi->s_qf_names[qtype] = NULL;
1310 return 1;
1311}
1312#endif
1313
26092bf5
TT		 1314#define MOPT_SET 0x0001
1315#define MOPT_CLEAR 0x0002
1316#define MOPT_NOSUPPORT 0x0004
1317#define MOPT_EXPLICIT 0x0008
1318#define MOPT_CLEAR_ERR 0x0010
1319#define MOPT_GTE0 0x0020
ac27a0ec 1320#ifdef CONFIG_QUOTA
26092bf5
TT		 1321#define MOPT_Q 0
1322#define MOPT_QFMT 0x0040
1323#else
1324#define MOPT_Q MOPT_NOSUPPORT
1325#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1326#endif
26092bf5 1327#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1328#define MOPT_NO_EXT2 0x0100
1329#define MOPT_NO_EXT3 0x0200
1330#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1331#define MOPT_STRING 0x0400
26092bf5
TT		 1332
1333static const struct mount_opts {
1334 int token;
1335 int mount_opt;
1336 int flags;
1337} ext4_mount_opts[] = {
1338 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1339 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1340 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1341 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1342 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1343 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1344 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1345 MOPT_EXT4_ONLY | MOPT_SET},
1346 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1347 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1348 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1349 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1350 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1351 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1352 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1353 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c
TT		 1354 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1355 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1356 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1357 EXT4_MOUNT_JOURNAL_CHECKSUM),
1358 MOPT_EXT4_ONLY | MOPT_SET},
1359 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1360 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1361 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1362 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1363 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1364 MOPT_NO_EXT2 | MOPT_SET},
1365 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1366 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1367 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1368 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1369 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1370 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1371 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1372 {Opt_commit, 0, MOPT_GTE0},
1373 {Opt_max_batch_time, 0, MOPT_GTE0},
1374 {Opt_min_batch_time, 0, MOPT_GTE0},
1375 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1376 {Opt_init_itable, 0, MOPT_GTE0},
1377 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1378 {Opt_resuid, 0, MOPT_GTE0},
1379 {Opt_resgid, 0, MOPT_GTE0},
1380 {Opt_journal_dev, 0, MOPT_GTE0},
ad4eec61 1381 {Opt_journal_path, 0, MOPT_STRING},
0efb3b23 1382 {Opt_journal_ioprio, 0, MOPT_GTE0},
8dc0aa8c
TT		 1383 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1384 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1385 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1386 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1387 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1388 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1389#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1390 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1391 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1392#else
26092bf5
TT		 1393 {Opt_acl, 0, MOPT_NOSUPPORT},
1394 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1395#endif
26092bf5
TT		 1396 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1397 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1398 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1399 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1400 MOPT_SET | MOPT_Q},
1401 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1402 MOPT_SET | MOPT_Q},
1403 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1404 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1405 {Opt_usrjquota, 0, MOPT_Q},
1406 {Opt_grpjquota, 0, MOPT_Q},
1407 {Opt_offusrjquota, 0, MOPT_Q},
1408 {Opt_offgrpjquota, 0, MOPT_Q},
1409 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1410 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1411 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1412 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1413 {Opt_err, 0, 0}
1414};
1415
1416static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1417 substring_t *args, unsigned long *journal_devnum,
1418 unsigned int *journal_ioprio, int is_remount)
1419{
1420 struct ext4_sb_info *sbi = EXT4_SB(sb);
1421 const struct mount_opts *m;
08cefc7a
EB		 1422 kuid_t uid;
1423 kgid_t gid;
26092bf5
TT		 1424 int arg = 0;
1425
57f73c2c
TT		 1426#ifdef CONFIG_QUOTA
1427 if (token == Opt_usrjquota)
1428 return set_qf_name(sb, USRQUOTA, &args[0]);
1429 else if (token == Opt_grpjquota)
1430 return set_qf_name(sb, GRPQUOTA, &args[0]);
1431 else if (token == Opt_offusrjquota)
1432 return clear_qf_name(sb, USRQUOTA);
1433 else if (token == Opt_offgrpjquota)
1434 return clear_qf_name(sb, GRPQUOTA);
1435#endif
26092bf5 1436 switch (token) {
f7048605
TT		 1437 case Opt_noacl:
1438 case Opt_nouser_xattr:
1439 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1440 break;
26092bf5
TT		 1441 case Opt_sb:
1442 return 1; /* handled by get_sb_block() */
1443 case Opt_removed:
5f3633e3 1444 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1445 return 1;
26092bf5
TT		 1446 case Opt_abort:
1447 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1448 return 1;
1449 case Opt_i_version:
1450 sb->s_flags |= MS_I_VERSION;
1451 return 1;
26092bf5
TT		 1452 }
1453
5f3633e3
JK		 1454 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1455 if (token == m->token)
1456 break;
1457
1458 if (m->token == Opt_err) {
1459 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1460 "or missing value", opt);
1461 return -1;
1462 }
1463
8dc0aa8c
TT		 1464 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1465 ext4_msg(sb, KERN_ERR,
1466 "Mount option \"%s\" incompatible with ext2", opt);
1467 return -1;
1468 }
1469 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1470 ext4_msg(sb, KERN_ERR,
1471 "Mount option \"%s\" incompatible with ext3", opt);
1472 return -1;
1473 }
1474
ad4eec61 1475 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1476 return -1;
1477 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1478 return -1;
1479 if (m->flags & MOPT_EXPLICIT)
1480 set_opt2(sb, EXPLICIT_DELALLOC);
1481 if (m->flags & MOPT_CLEAR_ERR)
1482 clear_opt(sb, ERRORS_MASK);
1483 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1484 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1485 "options when quota turned on");
1486 return -1;
1487 }
1488
1489 if (m->flags & MOPT_NOSUPPORT) {
1490 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1491 } else if (token == Opt_commit) {
1492 if (arg == 0)
1493 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1494 sbi->s_commit_interval = HZ * arg;
1495 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1496 sbi->s_max_batch_time = arg;
1497 } else if (token == Opt_min_batch_time) {
1498 sbi->s_min_batch_time = arg;
1499 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1500 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1501 ext4_msg(sb, KERN_ERR,
1502 "EXT4-fs: inode_readahead_blks must be "
1503 "0 or a power of 2 smaller than 2^31");
26092bf5 1504 return -1;
5f3633e3
JK		 1505 }
1506 sbi->s_inode_readahead_blks = arg;
1507 } else if (token == Opt_init_itable) {
1508 set_opt(sb, INIT_INODE_TABLE);
1509 if (!args->from)
1510 arg = EXT4_DEF_LI_WAIT_MULT;
1511 sbi->s_li_wait_mult = arg;
1512 } else if (token == Opt_max_dir_size_kb) {
1513 sbi->s_max_dir_size_kb = arg;
1514 } else if (token == Opt_stripe) {
1515 sbi->s_stripe = arg;
1516 } else if (token == Opt_resuid) {
1517 uid = make_kuid(current_user_ns(), arg);
1518 if (!uid_valid(uid)) {
1519 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1520 return -1;
1521 }
5f3633e3
JK		 1522 sbi->s_resuid = uid;
1523 } else if (token == Opt_resgid) {
1524 gid = make_kgid(current_user_ns(), arg);
1525 if (!gid_valid(gid)) {
1526 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1527 return -1;
1528 }
1529 sbi->s_resgid = gid;
1530 } else if (token == Opt_journal_dev) {
1531 if (is_remount) {
1532 ext4_msg(sb, KERN_ERR,
1533 "Cannot specify journal on remount");
1534 return -1;
1535 }
1536 *journal_devnum = arg;
ad4eec61
ES		 1537 } else if (token == Opt_journal_path) {
1538 char *journal_path;
1539 struct inode *journal_inode;
1540 struct path path;
1541 int error;
1542
1543 if (is_remount) {
1544 ext4_msg(sb, KERN_ERR,
1545 "Cannot specify journal on remount");
1546 return -1;
1547 }
1548 journal_path = match_strdup(&args[0]);
1549 if (!journal_path) {
1550 ext4_msg(sb, KERN_ERR, "error: could not dup "
1551 "journal device string");
1552 return -1;
1553 }
1554
1555 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1556 if (error) {
1557 ext4_msg(sb, KERN_ERR, "error: could not find "
1558 "journal device path: error %d", error);
1559 kfree(journal_path);
1560 return -1;
1561 }
1562
1563 journal_inode = path.dentry->d_inode;
1564 if (!S_ISBLK(journal_inode->i_mode)) {
1565 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1566 "is not a block device", journal_path);
1567 path_put(&path);
1568 kfree(journal_path);
1569 return -1;
1570 }
1571
1572 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1573 path_put(&path);
1574 kfree(journal_path);
5f3633e3
JK		 1575 } else if (token == Opt_journal_ioprio) {
1576 if (arg > 7) {
1577 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1578 " (must be 0-7)");
1579 return -1;
1580 }
1581 *journal_ioprio =
1582 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1583 } else if (m->flags & MOPT_DATAJ) {
1584 if (is_remount) {
1585 if (!sbi->s_journal)
1586 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1587 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1588 ext4_msg(sb, KERN_ERR,
26092bf5 1589 "Cannot change data mode on remount");
26092bf5 1590 return -1;
ac27a0ec 1591 }
26092bf5 1592 } else {
5f3633e3
JK		 1593 clear_opt(sb, DATA_FLAGS);
1594 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1595 }
5f3633e3
JK		 1596#ifdef CONFIG_QUOTA
1597 } else if (m->flags & MOPT_QFMT) {
1598 if (sb_any_quota_loaded(sb) &&
1599 sbi->s_jquota_fmt != m->mount_opt) {
1600 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1601 "quota options when quota turned on");
1602 return -1;
1603 }
262b4662
JK		 1604 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1605 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1606 ext4_msg(sb, KERN_ERR,
1607 "Cannot set journaled quota options "
1608 "when QUOTA feature is enabled");
1609 return -1;
1610 }
5f3633e3
JK		 1611 sbi->s_jquota_fmt = m->mount_opt;
1612#endif
1613 } else {
1614 if (!args->from)
1615 arg = 1;
1616 if (m->flags & MOPT_CLEAR)
1617 arg = !arg;
1618 else if (unlikely(!(m->flags & MOPT_SET))) {
1619 ext4_msg(sb, KERN_WARNING,
1620 "buggy handling of option %s", opt);
1621 WARN_ON(1);
1622 return -1;
1623 }
1624 if (arg != 0)
1625 sbi->s_mount_opt |= m->mount_opt;
1626 else
1627 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1628 }
5f3633e3 1629 return 1;
26092bf5
TT		 1630}
1631
1632static int parse_options(char *options, struct super_block *sb,
1633 unsigned long *journal_devnum,
1634 unsigned int *journal_ioprio,
1635 int is_remount)
1636{
1637 struct ext4_sb_info *sbi = EXT4_SB(sb);
1638 char *p;
1639 substring_t args[MAX_OPT_ARGS];
1640 int token;
1641
1642 if (!options)
1643 return 1;
1644
1645 while ((p = strsep(&options, ",")) != NULL) {
1646 if (!*p)
1647 continue;
1648 /*
1649 * Initialize args struct so we know whether arg was
1650 * found; some options take optional arguments.
1651 */
caecd0af 1652 args[0].to = args[0].from = NULL;
26092bf5
TT		 1653 token = match_token(p, tokens, args);
1654 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1655 journal_ioprio, is_remount) < 0)
1656 return 0;
ac27a0ec
DK		 1657 }
1658#ifdef CONFIG_QUOTA
262b4662
JK		 1659 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1660 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1661 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1662 "feature is enabled");
1663 return 0;
1664 }
ac27a0ec 1665 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1666 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1667 clear_opt(sb, USRQUOTA);
ac27a0ec 1668
482a7425 1669 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1670 clear_opt(sb, GRPQUOTA);
ac27a0ec 1671
56c50f11 1672 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1673 ext4_msg(sb, KERN_ERR, "old and new quota "
1674 "format mixing");
ac27a0ec
DK		 1675 return 0;
1676 }
1677
1678 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1679 ext4_msg(sb, KERN_ERR, "journaled quota format "
1680 "not specified");
ac27a0ec
DK		 1681 return 0;
1682 }
ac27a0ec
DK		 1683 }
1684#endif
261cb20c
JK		 1685 if (test_opt(sb, DIOREAD_NOLOCK)) {
1686 int blocksize =
1687 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1688
1689 if (blocksize < PAGE_CACHE_SIZE) {
1690 ext4_msg(sb, KERN_ERR, "can't mount with "
1691 "dioread_nolock if block size != PAGE_SIZE");
1692 return 0;
1693 }
1694 }
ac27a0ec
DK		 1695 return 1;
1696}
1697
2adf6da8
TT		 1698static inline void ext4_show_quota_options(struct seq_file *seq,
1699 struct super_block *sb)
1700{
1701#if defined(CONFIG_QUOTA)
1702 struct ext4_sb_info *sbi = EXT4_SB(sb);
1703
1704 if (sbi->s_jquota_fmt) {
1705 char *fmtname = "";
1706
1707 switch (sbi->s_jquota_fmt) {
1708 case QFMT_VFS_OLD:
1709 fmtname = "vfsold";
1710 break;
1711 case QFMT_VFS_V0:
1712 fmtname = "vfsv0";
1713 break;
1714 case QFMT_VFS_V1:
1715 fmtname = "vfsv1";
1716 break;
1717 }
1718 seq_printf(seq, ",jqfmt=%s", fmtname);
1719 }
1720
1721 if (sbi->s_qf_names[USRQUOTA])
1722 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1723
1724 if (sbi->s_qf_names[GRPQUOTA])
1725 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1726#endif
1727}
1728
5a916be1
TT		 1729static const char *token2str(int token)
1730{
50df9fd5 1731 const struct match_token *t;
5a916be1
TT		 1732
1733 for (t = tokens; t->token != Opt_err; t++)
1734 if (t->token == token && !strchr(t->pattern, '='))
1735 break;
1736 return t->pattern;
1737}
1738
2adf6da8
TT		 1739/*
1740 * Show an option if
1741 * - it's set to a non-default value OR
1742 * - if the per-sb default is different from the global default
1743 */
66acdcf4
TT		 1744static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1745 int nodefs)
2adf6da8 1746{
2adf6da8
TT		 1747 struct ext4_sb_info *sbi = EXT4_SB(sb);
1748 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1749 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1750 const struct mount_opts *m;
66acdcf4 1751 char sep = nodefs ? '\n' : ',';
2adf6da8 1752
66acdcf4
TT		 1753#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1754#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1755
1756 if (sbi->s_sb_block != 1)
5a916be1
TT		 1757 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1758
1759 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1760 int want_set = m->flags & MOPT_SET;
1761 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1762 (m->flags & MOPT_CLEAR_ERR))
1763 continue;
66acdcf4 1764 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1765 continue; /* skip if same as the default */
1766 if ((want_set &&
1767 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1768 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1769 continue; /* select Opt_noFoo vs Opt_Foo */
1770 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1771 }
5a916be1 1772
08cefc7a 1773 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1774 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1775 SEQ_OPTS_PRINT("resuid=%u",
1776 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1777 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1778 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1779 SEQ_OPTS_PRINT("resgid=%u",
1780 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1781 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1782 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1783 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1784 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1785 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1786 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1787 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1788 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1789 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1790 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1791 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1792 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1793 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1794 if (sb->s_flags & MS_I_VERSION)
5a916be1 1795 SEQ_OPTS_PUTS("i_version");
66acdcf4 1796 if (nodefs || sbi->s_stripe)
5a916be1 1797 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1798 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1799 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1800 SEQ_OPTS_PUTS("data=journal");
1801 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1802 SEQ_OPTS_PUTS("data=ordered");
1803 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1804 SEQ_OPTS_PUTS("data=writeback");
1805 }
66acdcf4
TT		 1806 if (nodefs ||
1807 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1808 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1809 sbi->s_inode_readahead_blks);
2adf6da8 1810
66acdcf4
TT		 1811 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1812 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1813 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1814 if (nodefs || sbi->s_max_dir_size_kb)
1815 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1816
1817 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1818 return 0;
1819}
1820
66acdcf4
TT		 1821static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1822{
1823 return _ext4_show_options(seq, root->d_sb, 0);
1824}
1825
1826static int options_seq_show(struct seq_file *seq, void *offset)
1827{
1828 struct super_block *sb = seq->private;
1829 int rc;
1830
1831 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1832 rc = _ext4_show_options(seq, sb, 1);
1833 seq_puts(seq, "\n");
1834 return rc;
1835}
1836
1837static int options_open_fs(struct inode *inode, struct file *file)
1838{
d9dda78b 1839 return single_open(file, options_seq_show, PDE_DATA(inode));
66acdcf4
TT		 1840}
1841
1842static const struct file_operations ext4_seq_options_fops = {
1843 .owner = THIS_MODULE,
1844 .open = options_open_fs,
1845 .read = seq_read,
1846 .llseek = seq_lseek,
1847 .release = single_release,
1848};
1849
617ba13b 1850static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1851 int read_only)
1852{
617ba13b 1853 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1854 int res = 0;
1855
617ba13b 1856 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1857 ext4_msg(sb, KERN_ERR, "revision level too high, "
1858 "forcing read-only mode");
ac27a0ec
DK		 1859 res = MS_RDONLY;
1860 }
1861 if (read_only)
281b5995 1862 goto done;
617ba13b 1863 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1864 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1865 "running e2fsck is recommended");
c8b459f4 1866 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 1867 ext4_msg(sb, KERN_WARNING,
1868 "warning: mounting fs with errors, "
1869 "running e2fsck is recommended");
ed3ce80a 1870 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1871 le16_to_cpu(es->s_mnt_count) >=
1872 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1873 ext4_msg(sb, KERN_WARNING,
1874 "warning: maximal mount count reached, "
1875 "running e2fsck is recommended");
ac27a0ec
DK		 1876 else if (le32_to_cpu(es->s_checkinterval) &&
1877 (le32_to_cpu(es->s_lastcheck) +
1878 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1879 ext4_msg(sb, KERN_WARNING,
1880 "warning: checktime reached, "
1881 "running e2fsck is recommended");
0b8e58a1 1882 if (!sbi->s_journal)
0390131b 1883 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1884 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1885 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1886 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1887 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1888 ext4_update_dynamic_rev(sb);
0390131b
FM		 1889 if (sbi->s_journal)
1890 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1891
e2d67052 1892 ext4_commit_super(sb, 1);
281b5995 1893done:
ac27a0ec 1894 if (test_opt(sb, DEBUG))
a9df9a49 1895 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1896 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1897 sb->s_blocksize,
1898 sbi->s_groups_count,
617ba13b
MC		 1899 EXT4_BLOCKS_PER_GROUP(sb),
1900 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1901 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1902
7abc52c2 1903 cleancache_init_fs(sb);
ac27a0ec
DK		 1904 return res;
1905}
1906
117fff10
TT		 1907int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1908{
1909 struct ext4_sb_info *sbi = EXT4_SB(sb);
1910 struct flex_groups *new_groups;
1911 int size;
1912
1913 if (!sbi->s_log_groups_per_flex)
1914 return 0;
1915
1916 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1917 if (size <= sbi->s_flex_groups_allocated)
1918 return 0;
1919
1920 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1921 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1922 if (!new_groups) {
1923 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1924 size / (int) sizeof(struct flex_groups));
1925 return -ENOMEM;
1926 }
1927
1928 if (sbi->s_flex_groups) {
1929 memcpy(new_groups, sbi->s_flex_groups,
1930 (sbi->s_flex_groups_allocated *
1931 sizeof(struct flex_groups)));
b93b41d4 1932 kvfree(sbi->s_flex_groups);
117fff10
TT		 1933 }
1934 sbi->s_flex_groups = new_groups;
1935 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1936 return 0;
1937}
1938
772cb7c8
JS		 1939static int ext4_fill_flex_info(struct super_block *sb)
1940{
1941 struct ext4_sb_info *sbi = EXT4_SB(sb);
1942 struct ext4_group_desc *gdp = NULL;
772cb7c8 1943 ext4_group_t flex_group;
117fff10 1944 int i, err;
772cb7c8 1945
503358ae 1946 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1947 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1948 sbi->s_log_groups_per_flex = 0;
1949 return 1;
1950 }
1951
117fff10
TT		 1952 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1953 if (err)
9933fc0a 1954 goto failed;
772cb7c8 1955
772cb7c8 1956 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1957 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1958
1959 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1960 atomic_add(ext4_free_inodes_count(sb, gdp),
1961 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 1962 atomic64_add(ext4_free_group_clusters(sb, gdp),
1963 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1964 atomic_add(ext4_used_dirs_count(sb, gdp),
1965 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1966 }
1967
1968 return 1;
1969failed:
1970 return 0;
1971}
1972
feb0ab32
DW		 1973static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1974 struct ext4_group_desc *gdp)
717d50e4 1975{
feb0ab32 1976 int offset;
717d50e4 1977 __u16 crc = 0;
feb0ab32 1978 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1979
9aa5d32b 1980 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 1981 /* Use new metadata_csum algorithm */
d6a77105 1982 __le16 save_csum;
feb0ab32
DW		 1983 __u32 csum32;
1984
d6a77105 1985 save_csum = gdp->bg_checksum;
feb0ab32
DW		 1986 gdp->bg_checksum = 0;
1987 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1988 sizeof(le_group));
1989 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1990 sbi->s_desc_size);
d6a77105 1991 gdp->bg_checksum = save_csum;
feb0ab32
DW		 1992
1993 crc = csum32 & 0xFFFF;
1994 goto out;
717d50e4
AD		 1995 }
1996
feb0ab32 1997 /* old crc16 code */
813d32f9
DW		 1998 if (!(sbi->s_es->s_feature_ro_compat &
1999 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2000 return 0;
2001
feb0ab32
DW		 2002 offset = offsetof(struct ext4_group_desc, bg_checksum);
2003
2004 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2005 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2006 crc = crc16(crc, (__u8 *)gdp, offset);
2007 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2008 /* for checksum of struct ext4_group_desc do the rest...*/
2009 if ((sbi->s_es->s_feature_incompat &
2010 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2011 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2012 crc = crc16(crc, (__u8 *)gdp + offset,
2013 le16_to_cpu(sbi->s_es->s_desc_size) -
2014 offset);
2015
2016out:
717d50e4
AD		 2017 return cpu_to_le16(crc);
2018}
2019
feb0ab32 2020int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2021 struct ext4_group_desc *gdp)
2022{
feb0ab32
DW		 2023 if (ext4_has_group_desc_csum(sb) &&
2024 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2025 block_group, gdp)))
717d50e4
AD		 2026 return 0;
2027
2028 return 1;
2029}
2030
feb0ab32
DW		 2031void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2032 struct ext4_group_desc *gdp)
2033{
2034 if (!ext4_has_group_desc_csum(sb))
2035 return;
2036 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2037}
2038
ac27a0ec 2039/* Called at mount-time, super-block is locked */
bfff6873
LC		 2040static int ext4_check_descriptors(struct super_block *sb,
2041 ext4_group_t *first_not_zeroed)
ac27a0ec 2042{
617ba13b
MC		 2043 struct ext4_sb_info *sbi = EXT4_SB(sb);
2044 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2045 ext4_fsblk_t last_block;
bd81d8ee
LV		 2046 ext4_fsblk_t block_bitmap;
2047 ext4_fsblk_t inode_bitmap;
2048 ext4_fsblk_t inode_table;
ce421581 2049 int flexbg_flag = 0;
bfff6873 2050 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2051
ce421581
JS		 2052 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2053 flexbg_flag = 1;
2054
af5bc92d 2055 ext4_debug("Checking group descriptors");
ac27a0ec 2056
197cd65a
AM		 2057 for (i = 0; i < sbi->s_groups_count; i++) {
2058 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2059
ce421581 2060 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2061 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2062 else
2063 last_block = first_block +
617ba13b 2064 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2065
bfff6873
LC		 2066 if ((grp == sbi->s_groups_count) &&
2067 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2068 grp = i;
2069
8fadc143 2070 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2071 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2072 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2073 "Block bitmap for group %u not in group "
b31e1552 2074 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2075 return 0;
2076 }
8fadc143 2077 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2078 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2079 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2080 "Inode bitmap for group %u not in group "
b31e1552 2081 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2082 return 0;
2083 }
8fadc143 2084 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2085 if (inode_table < first_block ||
2b2d6d01 2086 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2087 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2088 "Inode table for group %u not in group "
b31e1552 2089 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2090 return 0;
2091 }
955ce5f5 2092 ext4_lock_group(sb, i);
feb0ab32 2093 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2094 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2095 "Checksum for group %u failed (%u!=%u)",
2096 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2097 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2098 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2099 ext4_unlock_group(sb, i);
8a266467 2100 return 0;
7ee1ec4c 2101 }
717d50e4 2102 }
955ce5f5 2103 ext4_unlock_group(sb, i);
ce421581
JS		 2104 if (!flexbg_flag)
2105 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2106 }
bfff6873
LC		 2107 if (NULL != first_not_zeroed)
2108 *first_not_zeroed = grp;
ac27a0ec
DK		 2109 return 1;
2110}
2111
617ba13b 2112/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2113 * the superblock) which were deleted from all directories, but held open by
2114 * a process at the time of a crash. We walk the list and try to delete these
2115 * inodes at recovery time (only with a read-write filesystem).
2116 *
2117 * In order to keep the orphan inode chain consistent during traversal (in
2118 * case of crash during recovery), we link each inode into the superblock
2119 * orphan list_head and handle it the same way as an inode deletion during
2120 * normal operation (which journals the operations for us).
2121 *
2122 * We only do an iget() and an iput() on each inode, which is very safe if we
2123 * accidentally point at an in-use or already deleted inode. The worst that
2124 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2125 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2126 * e2fsck was run on this filesystem, and it must have already done the orphan
2127 * inode cleanup for us, so we can safely abort without any further action.
2128 */
2b2d6d01
TT		 2129static void ext4_orphan_cleanup(struct super_block *sb,
2130 struct ext4_super_block *es)
ac27a0ec
DK		 2131{
2132 unsigned int s_flags = sb->s_flags;
2133 int nr_orphans = 0, nr_truncates = 0;
2134#ifdef CONFIG_QUOTA
2135 int i;
2136#endif
2137 if (!es->s_last_orphan) {
2138 jbd_debug(4, "no orphan inodes to clean up\n");
2139 return;
2140 }
2141
a8f48a95 2142 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2143 ext4_msg(sb, KERN_ERR, "write access "
2144 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2145 return;
2146 }
2147
d39195c3
AG		 2148 /* Check if feature set would not allow a r/w mount */
2149 if (!ext4_feature_set_ok(sb, 0)) {
2150 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2151 "unknown ROCOMPAT features");
2152 return;
2153 }
2154
617ba13b 2155 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2156 /* don't clear list on RO mount w/ errors */
2157 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2158 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2159 "clearing orphan list.\n");
c25f9bc6
ES		 2160 es->s_last_orphan = 0;
2161 }
ac27a0ec
DK		 2162 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2163 return;
2164 }
2165
2166 if (s_flags & MS_RDONLY) {
b31e1552 2167 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2168 sb->s_flags &= ~MS_RDONLY;
2169 }
2170#ifdef CONFIG_QUOTA
2171 /* Needed for iput() to work correctly and not trash data */
2172 sb->s_flags |= MS_ACTIVE;
2173 /* Turn on quotas so that they are updated correctly */
a2d4a646 2174 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2175 if (EXT4_SB(sb)->s_qf_names[i]) {
2176 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2177 if (ret < 0)
b31e1552
ES		 2178 ext4_msg(sb, KERN_ERR,
2179 "Cannot turn on journaled "
2180 "quota: error %d", ret);
ac27a0ec
DK		 2181 }
2182 }
2183#endif
2184
2185 while (es->s_last_orphan) {
2186 struct inode *inode;
2187
97bd42b9
JB		 2188 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2189 if (IS_ERR(inode)) {
ac27a0ec
DK		 2190 es->s_last_orphan = 0;
2191 break;
2192 }
2193
617ba13b 2194 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2195 dquot_initialize(inode);
ac27a0ec 2196 if (inode->i_nlink) {
566370a2
PT		 2197 if (test_opt(sb, DEBUG))
2198 ext4_msg(sb, KERN_DEBUG,
2199 "%s: truncating inode %lu to %lld bytes",
2200 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2201 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2202 inode->i_ino, inode->i_size);
721e3eba 2203 mutex_lock(&inode->i_mutex);
55f252c9 2204 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2205 ext4_truncate(inode);
721e3eba 2206 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2207 nr_truncates++;
2208 } else {
566370a2
PT		 2209 if (test_opt(sb, DEBUG))
2210 ext4_msg(sb, KERN_DEBUG,
2211 "%s: deleting unreferenced inode %lu",
2212 __func__, inode->i_ino);
ac27a0ec
DK		 2213 jbd_debug(2, "deleting unreferenced inode %lu\n",
2214 inode->i_ino);
2215 nr_orphans++;
2216 }
2217 iput(inode); /* The delete magic happens here! */
2218 }
2219
2b2d6d01 2220#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2221
2222 if (nr_orphans)
b31e1552
ES		 2223 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2224 PLURAL(nr_orphans));
ac27a0ec 2225 if (nr_truncates)
b31e1552
ES		 2226 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2227 PLURAL(nr_truncates));
ac27a0ec
DK		 2228#ifdef CONFIG_QUOTA
2229 /* Turn quotas off */
a2d4a646 2230 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2231 if (sb_dqopt(sb)->files[i])
287a8095 2232 dquot_quota_off(sb, i);
ac27a0ec
DK		 2233 }
2234#endif
2235 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2236}
0b8e58a1 2237
cd2291a4
ES		 2238/*
2239 * Maximal extent format file size.
2240 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2241 * extent format containers, within a sector_t, and within i_blocks
2242 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2243 * so that won't be a limiting factor.
2244 *
f17722f9
LC		 2245 * However there is other limiting factor. We do store extents in the form
2246 * of starting block and length, hence the resulting length of the extent
2247 * covering maximum file size must fit into on-disk format containers as
2248 * well. Given that length is always by 1 unit bigger than max unit (because
2249 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2250 *
cd2291a4
ES		 2251 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2252 */
f287a1a5 2253static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2254{
2255 loff_t res;
2256 loff_t upper_limit = MAX_LFS_FILESIZE;
2257
2258 /* small i_blocks in vfs inode? */
f287a1a5 2259 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2260 /*
90c699a9 2261 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2262 * i_block represent total blocks in 512 bytes
2263 * 32 == size of vfs inode i_blocks * 8
2264 */
2265 upper_limit = (1LL << 32) - 1;
2266
2267 /* total blocks in file system block size */
2268 upper_limit >>= (blkbits - 9);
2269 upper_limit <<= blkbits;
2270 }
2271
f17722f9
LC		 2272 /*
2273 * 32-bit extent-start container, ee_block. We lower the maxbytes
2274 * by one fs block, so ee_len can cover the extent of maximum file
2275 * size
2276 */
2277 res = (1LL << 32) - 1;
cd2291a4 2278 res <<= blkbits;
cd2291a4
ES		 2279
2280 /* Sanity check against vm- & vfs- imposed limits */
2281 if (res > upper_limit)
2282 res = upper_limit;
2283
2284 return res;
2285}
ac27a0ec 2286
ac27a0ec 2287/*
cd2291a4 2288 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2289 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2290 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2291 */
f287a1a5 2292static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2293{
617ba13b 2294 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2295 int meta_blocks;
2296 loff_t upper_limit;
0b8e58a1
AD		 2297 /* This is calculated to be the largest file size for a dense, block
2298 * mapped file such that the file's total number of 512-byte sectors,
2299 * including data and all indirect blocks, does not exceed (2^48 - 1).
2300 *
2301 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2302 * number of 512-byte sectors of the file.
0fc1b451
AK		 2303 */
2304
f287a1a5 2305 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2306 /*
90c699a9 2307 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2308 * the inode i_block field represents total file blocks in
2309 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2310 */
2311 upper_limit = (1LL << 32) - 1;
2312
2313 /* total blocks in file system block size */
2314 upper_limit >>= (bits - 9);
2315
2316 } else {
8180a562
AK		 2317 /*
2318 * We use 48 bit ext4_inode i_blocks
2319 * With EXT4_HUGE_FILE_FL set the i_blocks
2320 * represent total number of blocks in
2321 * file system block size
2322 */
0fc1b451
AK		 2323 upper_limit = (1LL << 48) - 1;
2324
0fc1b451
AK		 2325 }
2326
2327 /* indirect blocks */
2328 meta_blocks = 1;
2329 /* double indirect blocks */
2330 meta_blocks += 1 + (1LL << (bits-2));
2331 /* tripple indirect blocks */
2332 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2333
2334 upper_limit -= meta_blocks;
2335 upper_limit <<= bits;
ac27a0ec
DK		 2336
2337 res += 1LL << (bits-2);
2338 res += 1LL << (2*(bits-2));
2339 res += 1LL << (3*(bits-2));
2340 res <<= bits;
2341 if (res > upper_limit)
2342 res = upper_limit;
0fc1b451
AK		 2343
2344 if (res > MAX_LFS_FILESIZE)
2345 res = MAX_LFS_FILESIZE;
2346
ac27a0ec
DK		 2347 return res;
2348}
2349
617ba13b 2350static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2351 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2352{
617ba13b 2353 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2354 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2355 int has_super = 0;
2356
2357 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2358
617ba13b 2359 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2360 nr < first_meta_bg)
70bbb3e0 2361 return logical_sb_block + nr + 1;
ac27a0ec 2362 bg = sbi->s_desc_per_block * nr;
617ba13b 2363 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2364 has_super = 1;
0b8e58a1 2365
bd63f6b0
DW		 2366 /*
2367 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2368 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2369 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2370 * compensate.
2371 */
2372 if (sb->s_blocksize == 1024 && nr == 0 &&
2373 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2374 has_super++;
2375
617ba13b 2376 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2377}
2378
c9de560d
AT		 2379/**
2380 * ext4_get_stripe_size: Get the stripe size.
2381 * @sbi: In memory super block info
2382 *
2383 * If we have specified it via mount option, then
2384 * use the mount option value. If the value specified at mount time is
2385 * greater than the blocks per group use the super block value.
2386 * If the super block value is greater than blocks per group return 0.
2387 * Allocator needs it be less than blocks per group.
2388 *
2389 */
2390static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2391{
2392 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2393 unsigned long stripe_width =
2394 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2395 int ret;
c9de560d
AT		 2396
2397 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2398 ret = sbi->s_stripe;
2399 else if (stripe_width <= sbi->s_blocks_per_group)
2400 ret = stripe_width;
2401 else if (stride <= sbi->s_blocks_per_group)
2402 ret = stride;
2403 else
2404 ret = 0;
c9de560d 2405
3eb08658
DE		 2406 /*
2407 * If the stripe width is 1, this makes no sense and
2408 * we set it to 0 to turn off stripe handling code.
2409 */
2410 if (ret <= 1)
2411 ret = 0;
c9de560d 2412
3eb08658 2413 return ret;
c9de560d 2414}
ac27a0ec 2415
3197ebdb
TT		 2416/* sysfs supprt */
2417
2418struct ext4_attr {
2419 struct attribute attr;
2420 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2421 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb 2422 const char *, size_t);
f2d50a65
JK		 2423 union {
2424 int offset;
2425 int deprecated_val;
2426 } u;
3197ebdb
TT		 2427};
2428
27dd4385
LC		 2429static int parse_strtoull(const char *buf,
2430 unsigned long long max, unsigned long long *value)
2431{
2432 int ret;
2433
2434 ret = kstrtoull(skip_spaces(buf), 0, value);
2435 if (!ret && *value > max)
2436 ret = -EINVAL;
2437 return ret;
2438}
2439
3197ebdb
TT		 2440static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2441 struct ext4_sb_info *sbi,
2442 char *buf)
2443{
2444 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2445 (s64) EXT4_C2B(sbi,
2446 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2447}
2448
2449static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2450 struct ext4_sb_info *sbi, char *buf)
2451{
2452 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2453
f613dfcb
TT		 2454 if (!sb->s_bdev->bd_part)
2455 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2456 return snprintf(buf, PAGE_SIZE, "%lu\n",
2457 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2458 sbi->s_sectors_written_start) >> 1);
2459}
2460
2461static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2462 struct ext4_sb_info *sbi, char *buf)
2463{
2464 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2465
f613dfcb
TT		 2466 if (!sb->s_bdev->bd_part)
2467 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2468 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2469 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2470 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2471 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2472}
2473
2474static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2475 struct ext4_sb_info *sbi,
2476 const char *buf, size_t count)
2477{
2478 unsigned long t;
e1091b15 2479 int ret;
3197ebdb 2480
e1091b15
LC		 2481 ret = kstrtoul(skip_spaces(buf), 0, &t);
2482 if (ret)
2483 return ret;
3197ebdb 2484
e1091b15 2485 if (t && (!is_power_of_2(t) || t > 0x40000000))
3197ebdb
TT		 2486 return -EINVAL;
2487
2488 sbi->s_inode_readahead_blks = t;
2489 return count;
2490}
2491
2492static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2493 struct ext4_sb_info *sbi, char *buf)
3197ebdb 2494{
f2d50a65 2495 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb
TT		 2496
2497 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2498}
2499
2500static ssize_t sbi_ui_store(struct ext4_attr *a,
2501 struct ext4_sb_info *sbi,
2502 const char *buf, size_t count)
2503{
f2d50a65 2504 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb 2505 unsigned long t;
e1091b15 2506 int ret;
3197ebdb 2507
e1091b15
LC		 2508 ret = kstrtoul(skip_spaces(buf), 0, &t);
2509 if (ret)
2510 return ret;
3197ebdb
TT		 2511 *ui = t;
2512 return count;
2513}
2514
52c198c6
LC		 2515static ssize_t es_ui_show(struct ext4_attr *a,
2516 struct ext4_sb_info *sbi, char *buf)
2517{
2518
2519 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2520 a->u.offset);
2521
2522 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2523}
2524
27dd4385
LC		 2525static ssize_t reserved_clusters_show(struct ext4_attr *a,
2526 struct ext4_sb_info *sbi, char *buf)
2527{
2528 return snprintf(buf, PAGE_SIZE, "%llu\n",
2529 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2530}
2531
2532static ssize_t reserved_clusters_store(struct ext4_attr *a,
2533 struct ext4_sb_info *sbi,
2534 const char *buf, size_t count)
2535{
2536 unsigned long long val;
2537 int ret;
2538
2539 if (parse_strtoull(buf, -1ULL, &val))
2540 return -EINVAL;
2541 ret = ext4_reserve_clusters(sbi, val);
2542
2543 return ret ? ret : count;
2544}
2545
2c0544b2
TT		 2546static ssize_t trigger_test_error(struct ext4_attr *a,
2547 struct ext4_sb_info *sbi,
2548 const char *buf, size_t count)
2549{
2550 int len = count;
2551
2552 if (!capable(CAP_SYS_ADMIN))
2553 return -EPERM;
2554
2555 if (len && buf[len-1] == '\n')
2556 len--;
2557
2558 if (len)
2559 ext4_error(sbi->s_sb, "%.*s", len, buf);
2560 return count;
2561}
2562
f2d50a65
JK		 2563static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2564 struct ext4_sb_info *sbi, char *buf)
2565{
2566 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2567}
2568
3197ebdb
TT		 2569#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2570static struct ext4_attr ext4_attr_##_name = { \
2571 .attr = {.name = __stringify(_name), .mode = _mode }, \
2572 .show = _show, \
2573 .store = _store, \
f2d50a65
JK		 2574 .u = { \
2575 .offset = offsetof(struct ext4_sb_info, _elname),\
2576 }, \
3197ebdb 2577}
52c198c6
LC		 2578
2579#define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2580static struct ext4_attr ext4_attr_##_name = { \
2581 .attr = {.name = __stringify(_name), .mode = _mode }, \
2582 .show = _show, \
2583 .store = _store, \
2584 .u = { \
2585 .offset = offsetof(struct ext4_super_block, _elname), \
2586 }, \
2587}
2588
3197ebdb
TT		 2589#define EXT4_ATTR(name, mode, show, store) \
2590static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2591
857ac889 2592#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2593#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2594#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
52c198c6
LC		 2595
2596#define EXT4_RO_ATTR_ES_UI(name, elname) \
2597 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
3197ebdb
TT		 2598#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2599 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
52c198c6 2600
3197ebdb 2601#define ATTR_LIST(name) &ext4_attr_##name.attr
f2d50a65
JK		 2602#define EXT4_DEPRECATED_ATTR(_name, _val) \
2603static struct ext4_attr ext4_attr_##_name = { \
2604 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2605 .show = sbi_deprecated_show, \
2606 .u = { \
2607 .deprecated_val = _val, \
2608 }, \
2609}
3197ebdb
TT		 2610
2611EXT4_RO_ATTR(delayed_allocation_blocks);
2612EXT4_RO_ATTR(session_write_kbytes);
2613EXT4_RO_ATTR(lifetime_write_kbytes);
27dd4385 2614EXT4_RW_ATTR(reserved_clusters);
3197ebdb
TT		 2615EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2616 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2617EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2618EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2619EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2620EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2621EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2622EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2623EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
f2d50a65 2624EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
67a5da56 2625EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2626EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
efbed4dc
TT		 2627EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2628EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2629EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2630EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2631EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2632EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
52c198c6
LC		 2633EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2634EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2635EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
3197ebdb
TT		 2636
2637static struct attribute *ext4_attrs[] = {
2638 ATTR_LIST(delayed_allocation_blocks),
2639 ATTR_LIST(session_write_kbytes),
2640 ATTR_LIST(lifetime_write_kbytes),
27dd4385 2641 ATTR_LIST(reserved_clusters),
3197ebdb 2642 ATTR_LIST(inode_readahead_blks),
11013911 2643 ATTR_LIST(inode_goal),
3197ebdb
TT		 2644 ATTR_LIST(mb_stats),
2645 ATTR_LIST(mb_max_to_scan),
2646 ATTR_LIST(mb_min_to_scan),
2647 ATTR_LIST(mb_order2_req),
2648 ATTR_LIST(mb_stream_req),
2649 ATTR_LIST(mb_group_prealloc),
55138e0b 2650 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2651 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2652 ATTR_LIST(trigger_fs_error),
efbed4dc
TT		 2653 ATTR_LIST(err_ratelimit_interval_ms),
2654 ATTR_LIST(err_ratelimit_burst),
2655 ATTR_LIST(warning_ratelimit_interval_ms),
2656 ATTR_LIST(warning_ratelimit_burst),
2657 ATTR_LIST(msg_ratelimit_interval_ms),
2658 ATTR_LIST(msg_ratelimit_burst),
52c198c6
LC		 2659 ATTR_LIST(errors_count),
2660 ATTR_LIST(first_error_time),
2661 ATTR_LIST(last_error_time),
3197ebdb
TT		 2662 NULL,
2663};
2664
857ac889
LC		 2665/* Features this copy of ext4 supports */
2666EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2667EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2668EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2669
2670static struct attribute *ext4_feat_attrs[] = {
2671 ATTR_LIST(lazy_itable_init),
27ee40df 2672 ATTR_LIST(batched_discard),
5e7bbef1 2673 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2674 NULL,
2675};
2676
3197ebdb
TT		 2677static ssize_t ext4_attr_show(struct kobject *kobj,
2678 struct attribute *attr, char *buf)
2679{
2680 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2681 s_kobj);
2682 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2683
2684 return a->show ? a->show(a, sbi, buf) : 0;
2685}
2686
2687static ssize_t ext4_attr_store(struct kobject *kobj,
2688 struct attribute *attr,
2689 const char *buf, size_t len)
2690{
2691 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2692 s_kobj);
2693 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2694
2695 return a->store ? a->store(a, sbi, buf, len) : 0;
2696}
2697
2698static void ext4_sb_release(struct kobject *kobj)
2699{
2700 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2701 s_kobj);
2702 complete(&sbi->s_kobj_unregister);
2703}
2704
52cf25d0 2705static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2706 .show = ext4_attr_show,
2707 .store = ext4_attr_store,
2708};
2709
2710static struct kobj_type ext4_ktype = {
2711 .default_attrs = ext4_attrs,
2712 .sysfs_ops = &ext4_attr_ops,
2713 .release = ext4_sb_release,
2714};
2715
857ac889
LC		 2716static void ext4_feat_release(struct kobject *kobj)
2717{
2718 complete(&ext4_feat->f_kobj_unregister);
2719}
2720
c7f72543
LC		 2721static ssize_t ext4_feat_show(struct kobject *kobj,
2722 struct attribute *attr, char *buf)
2723{
2724 return snprintf(buf, PAGE_SIZE, "supported\n");
2725}
2726
2727/*
2728 * We can not use ext4_attr_show/store because it relies on the kobject
2729 * being embedded in the ext4_sb_info structure which is definitely not
2730 * true in this case.
2731 */
2732static const struct sysfs_ops ext4_feat_ops = {
2733 .show = ext4_feat_show,
2734 .store = NULL,
2735};
2736
857ac889
LC		 2737static struct kobj_type ext4_feat_ktype = {
2738 .default_attrs = ext4_feat_attrs,
c7f72543 2739 .sysfs_ops = &ext4_feat_ops,
857ac889
LC		 2740 .release = ext4_feat_release,
2741};
2742
a13fb1a4
ES		 2743/*
2744 * Check whether this filesystem can be mounted based on
2745 * the features present and the RDONLY/RDWR mount requested.
2746 * Returns 1 if this filesystem can be mounted as requested,
2747 * 0 if it cannot be.
2748 */
2749static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2750{
2751 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2752 ext4_msg(sb, KERN_ERR,
2753 "Couldn't mount because of "
2754 "unsupported optional features (%x)",
2755 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2756 ~EXT4_FEATURE_INCOMPAT_SUPP));
2757 return 0;
2758 }
2759
2760 if (readonly)
2761 return 1;
2762
2763 /* Check that feature set is OK for a read-write mount */
2764 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2765 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2766 "unsupported optional features (%x)",
2767 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2768 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2769 return 0;
2770 }
2771 /*
2772 * Large file size enabled file system can only be mounted
2773 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2774 */
2775 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2776 if (sizeof(blkcnt_t) < sizeof(u64)) {
2777 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2778 "cannot be mounted RDWR without "
2779 "CONFIG_LBDAF");
2780 return 0;
2781 }
2782 }
bab08ab9
TT		 2783 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2784 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2785 ext4_msg(sb, KERN_ERR,
2786 "Can't support bigalloc feature without "
2787 "extents feature\n");
2788 return 0;
2789 }
7c319d32
AK		 2790
2791#ifndef CONFIG_QUOTA
2792 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2793 !readonly) {
2794 ext4_msg(sb, KERN_ERR,
2795 "Filesystem with quota feature cannot be mounted RDWR "
2796 "without CONFIG_QUOTA");
2797 return 0;
2798 }
2799#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2800 return 1;
2801}
2802
66e61a9e
TT		 2803/*
2804 * This function is called once a day if we have errors logged
2805 * on the file system
2806 */
2807static void print_daily_error_info(unsigned long arg)
2808{
2809 struct super_block *sb = (struct super_block *) arg;
2810 struct ext4_sb_info *sbi;
2811 struct ext4_super_block *es;
2812
2813 sbi = EXT4_SB(sb);
2814 es = sbi->s_es;
2815
2816 if (es->s_error_count)
ae0f78de
TT		 2817 /* fsck newer than v1.41.13 is needed to clean this condition. */
2818 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2819 le32_to_cpu(es->s_error_count));
2820 if (es->s_first_error_time) {
ae0f78de 2821 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2822 sb->s_id, le32_to_cpu(es->s_first_error_time),
2823 (int) sizeof(es->s_first_error_func),
2824 es->s_first_error_func,
2825 le32_to_cpu(es->s_first_error_line));
2826 if (es->s_first_error_ino)
2827 printk(": inode %u",
2828 le32_to_cpu(es->s_first_error_ino));
2829 if (es->s_first_error_block)
2830 printk(": block %llu", (unsigned long long)
2831 le64_to_cpu(es->s_first_error_block));
2832 printk("\n");
2833 }
2834 if (es->s_last_error_time) {
ae0f78de 2835 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2836 sb->s_id, le32_to_cpu(es->s_last_error_time),
2837 (int) sizeof(es->s_last_error_func),
2838 es->s_last_error_func,
2839 le32_to_cpu(es->s_last_error_line));
2840 if (es->s_last_error_ino)
2841 printk(": inode %u",
2842 le32_to_cpu(es->s_last_error_ino));
2843 if (es->s_last_error_block)
2844 printk(": block %llu", (unsigned long long)
2845 le64_to_cpu(es->s_last_error_block));
2846 printk("\n");
2847 }
2848 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2849}
2850
bfff6873
LC		 2851/* Find next suitable group and run ext4_init_inode_table */
2852static int ext4_run_li_request(struct ext4_li_request *elr)
2853{
2854 struct ext4_group_desc *gdp = NULL;
2855 ext4_group_t group, ngroups;
2856 struct super_block *sb;
2857 unsigned long timeout = 0;
2858 int ret = 0;
2859
2860 sb = elr->lr_super;
2861 ngroups = EXT4_SB(sb)->s_groups_count;
2862
8e8ad8a5 2863 sb_start_write(sb);
bfff6873
LC		 2864 for (group = elr->lr_next_group; group < ngroups; group++) {
2865 gdp = ext4_get_group_desc(sb, group, NULL);
2866 if (!gdp) {
2867 ret = 1;
2868 break;
2869 }
2870
2871 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2872 break;
2873 }
2874
7f511862 2875 if (group >= ngroups)
bfff6873
LC		 2876 ret = 1;
2877
2878 if (!ret) {
2879 timeout = jiffies;
2880 ret = ext4_init_inode_table(sb, group,
2881 elr->lr_timeout ? 0 : 1);
2882 if (elr->lr_timeout == 0) {
51ce6511
LC		 2883 timeout = (jiffies - timeout) *
2884 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2885 elr->lr_timeout = timeout;
2886 }
2887 elr->lr_next_sched = jiffies + elr->lr_timeout;
2888 elr->lr_next_group = group + 1;
2889 }
8e8ad8a5 2890 sb_end_write(sb);
bfff6873
LC		 2891
2892 return ret;
2893}
2894
2895/*
2896 * Remove lr_request from the list_request and free the
4ed5c033 2897 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2898 */
2899static void ext4_remove_li_request(struct ext4_li_request *elr)
2900{
2901 struct ext4_sb_info *sbi;
2902
2903 if (!elr)
2904 return;
2905
2906 sbi = elr->lr_sbi;
2907
2908 list_del(&elr->lr_request);
2909 sbi->s_li_request = NULL;
2910 kfree(elr);
2911}
2912
2913static void ext4_unregister_li_request(struct super_block *sb)
2914{
1bb933fb
LC		 2915 mutex_lock(&ext4_li_mtx);
2916 if (!ext4_li_info) {
2917 mutex_unlock(&ext4_li_mtx);
bfff6873 2918 return;
1bb933fb 2919 }
bfff6873
LC		 2920
2921 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2922 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2923 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2924 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2925}
2926
8f1f7453
ES		 2927static struct task_struct *ext4_lazyinit_task;
2928
bfff6873
LC		 2929/*
2930 * This is the function where ext4lazyinit thread lives. It walks
2931 * through the request list searching for next scheduled filesystem.
2932 * When such a fs is found, run the lazy initialization request
2933 * (ext4_rn_li_request) and keep track of the time spend in this
2934 * function. Based on that time we compute next schedule time of
2935 * the request. When walking through the list is complete, compute
2936 * next waking time and put itself into sleep.
2937 */
2938static int ext4_lazyinit_thread(void *arg)
2939{
2940 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2941 struct list_head *pos, *n;
2942 struct ext4_li_request *elr;
4ed5c033 2943 unsigned long next_wakeup, cur;
bfff6873
LC		 2944
2945 BUG_ON(NULL == eli);
2946
bfff6873
LC		 2947cont_thread:
2948 while (true) {
2949 next_wakeup = MAX_JIFFY_OFFSET;
2950
2951 mutex_lock(&eli->li_list_mtx);
2952 if (list_empty(&eli->li_request_list)) {
2953 mutex_unlock(&eli->li_list_mtx);
2954 goto exit_thread;
2955 }
2956
2957 list_for_each_safe(pos, n, &eli->li_request_list) {
2958 elr = list_entry(pos, struct ext4_li_request,
2959 lr_request);
2960
b2c78cd0
TT		 2961 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2962 if (ext4_run_li_request(elr) != 0) {
2963 /* error, remove the lazy_init job */
2964 ext4_remove_li_request(elr);
2965 continue;
2966 }
bfff6873
LC		 2967 }
2968
2969 if (time_before(elr->lr_next_sched, next_wakeup))
2970 next_wakeup = elr->lr_next_sched;
2971 }
2972 mutex_unlock(&eli->li_list_mtx);
2973
a0acae0e 2974 try_to_freeze();
bfff6873 2975
4ed5c033
LC		 2976 cur = jiffies;
2977 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2978 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2979 cond_resched();
2980 continue;
2981 }
2982
4ed5c033
LC		 2983 schedule_timeout_interruptible(next_wakeup - cur);
2984
8f1f7453
ES		 2985 if (kthread_should_stop()) {
2986 ext4_clear_request_list();
2987 goto exit_thread;
2988 }
bfff6873
LC		 2989 }
2990
2991exit_thread:
2992 /*
2993 * It looks like the request list is empty, but we need
2994 * to check it under the li_list_mtx lock, to prevent any
2995 * additions into it, and of course we should lock ext4_li_mtx
2996 * to atomically free the list and ext4_li_info, because at
2997 * this point another ext4 filesystem could be registering
2998 * new one.
2999 */
3000 mutex_lock(&ext4_li_mtx);
3001 mutex_lock(&eli->li_list_mtx);
3002 if (!list_empty(&eli->li_request_list)) {
3003 mutex_unlock(&eli->li_list_mtx);
3004 mutex_unlock(&ext4_li_mtx);
3005 goto cont_thread;
3006 }
3007 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3008 kfree(ext4_li_info);
3009 ext4_li_info = NULL;
3010 mutex_unlock(&ext4_li_mtx);
3011
3012 return 0;
3013}
3014
3015static void ext4_clear_request_list(void)
3016{
3017 struct list_head *pos, *n;
3018 struct ext4_li_request *elr;
3019
3020 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3021 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3022 elr = list_entry(pos, struct ext4_li_request,
3023 lr_request);
3024 ext4_remove_li_request(elr);
3025 }
3026 mutex_unlock(&ext4_li_info->li_list_mtx);
3027}
3028
3029static int ext4_run_lazyinit_thread(void)
3030{
8f1f7453
ES		 3031 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3032 ext4_li_info, "ext4lazyinit");
3033 if (IS_ERR(ext4_lazyinit_task)) {
3034 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3035 ext4_clear_request_list();
bfff6873
LC		 3036 kfree(ext4_li_info);
3037 ext4_li_info = NULL;
92b97816 3038 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3039 "initialization thread\n",
3040 err);
3041 return err;
3042 }
3043 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3044 return 0;
3045}
3046
3047/*
3048 * Check whether it make sense to run itable init. thread or not.
3049 * If there is at least one uninitialized inode table, return
3050 * corresponding group number, else the loop goes through all
3051 * groups and return total number of groups.
3052 */
3053static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3054{
3055 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3056 struct ext4_group_desc *gdp = NULL;
3057
3058 for (group = 0; group < ngroups; group++) {
3059 gdp = ext4_get_group_desc(sb, group, NULL);
3060 if (!gdp)
3061 continue;
3062
3063 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3064 break;
3065 }
3066
3067 return group;
3068}
3069
3070static int ext4_li_info_new(void)
3071{
3072 struct ext4_lazy_init *eli = NULL;
3073
3074 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3075 if (!eli)
3076 return -ENOMEM;
3077
bfff6873
LC		 3078 INIT_LIST_HEAD(&eli->li_request_list);
3079 mutex_init(&eli->li_list_mtx);
3080
bfff6873
LC		 3081 eli->li_state |= EXT4_LAZYINIT_QUIT;
3082
3083 ext4_li_info = eli;
3084
3085 return 0;
3086}
3087
3088static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3089 ext4_group_t start)
3090{
3091 struct ext4_sb_info *sbi = EXT4_SB(sb);
3092 struct ext4_li_request *elr;
bfff6873
LC		 3093
3094 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3095 if (!elr)
3096 return NULL;
3097
3098 elr->lr_super = sb;
3099 elr->lr_sbi = sbi;
3100 elr->lr_next_group = start;
3101
3102 /*
3103 * Randomize first schedule time of the request to
3104 * spread the inode table initialization requests
3105 * better.
3106 */
dd1f723b
TT		 3107 elr->lr_next_sched = jiffies + (prandom_u32() %
3108 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3109 return elr;
3110}
3111
7f511862
TT		 3112int ext4_register_li_request(struct super_block *sb,
3113 ext4_group_t first_not_zeroed)
bfff6873
LC		 3114{
3115 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3116 struct ext4_li_request *elr = NULL;
bfff6873 3117 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3118 int ret = 0;
bfff6873 3119
7f511862 3120 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3121 if (sbi->s_li_request != NULL) {
3122 /*
3123 * Reset timeout so it can be computed again, because
3124 * s_li_wait_mult might have changed.
3125 */
3126 sbi->s_li_request->lr_timeout = 0;
7f511862 3127 goto out;
51ce6511 3128 }
bfff6873
LC		 3129
3130 if (first_not_zeroed == ngroups ||
3131 (sb->s_flags & MS_RDONLY) ||
55ff3840 3132 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3133 goto out;
bfff6873
LC		 3134
3135 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3136 if (!elr) {
3137 ret = -ENOMEM;
3138 goto out;
3139 }
bfff6873
LC		 3140
3141 if (NULL == ext4_li_info) {
3142 ret = ext4_li_info_new();
3143 if (ret)
3144 goto out;
3145 }
3146
3147 mutex_lock(&ext4_li_info->li_list_mtx);
3148 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3149 mutex_unlock(&ext4_li_info->li_list_mtx);
3150
3151 sbi->s_li_request = elr;
46e4690b
TM		 3152 /*
3153 * set elr to NULL here since it has been inserted to
3154 * the request_list and the removal and free of it is
3155 * handled by ext4_clear_request_list from now on.
3156 */
3157 elr = NULL;
bfff6873
LC		 3158
3159 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3160 ret = ext4_run_lazyinit_thread();
3161 if (ret)
3162 goto out;
3163 }
bfff6873 3164out:
beed5ecb
NK		 3165 mutex_unlock(&ext4_li_mtx);
3166 if (ret)
bfff6873 3167 kfree(elr);
bfff6873
LC		 3168 return ret;
3169}
3170
3171/*
3172 * We do not need to lock anything since this is called on
3173 * module unload.
3174 */
3175static void ext4_destroy_lazyinit_thread(void)
3176{
3177 /*
3178 * If thread exited earlier
3179 * there's nothing to be done.
3180 */
8f1f7453 3181 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3182 return;
3183
8f1f7453 3184 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3185}
3186
25ed6e8a
DW		 3187static int set_journal_csum_feature_set(struct super_block *sb)
3188{
3189 int ret = 1;
3190 int compat, incompat;
3191 struct ext4_sb_info *sbi = EXT4_SB(sb);
3192
9aa5d32b 3193 if (ext4_has_metadata_csum(sb)) {
db9ee220 3194 /* journal checksum v3 */
25ed6e8a 3195 compat = 0;
db9ee220 3196 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3197 } else {
3198 /* journal checksum v1 */
3199 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3200 incompat = 0;
3201 }
3202
feb8c6d3
DW		 3203 jbd2_journal_clear_features(sbi->s_journal,
3204 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3205 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3206 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3207 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3208 ret = jbd2_journal_set_features(sbi->s_journal,
3209 compat, 0,
3210 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3211 incompat);
3212 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3213 ret = jbd2_journal_set_features(sbi->s_journal,
3214 compat, 0,
3215 incompat);
3216 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3217 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3218 } else {
feb8c6d3
DW		 3219 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3220 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3221 }
3222
3223 return ret;
3224}
3225
952fc18e
TT		 3226/*
3227 * Note: calculating the overhead so we can be compatible with
3228 * historical BSD practice is quite difficult in the face of
3229 * clusters/bigalloc. This is because multiple metadata blocks from
3230 * different block group can end up in the same allocation cluster.
3231 * Calculating the exact overhead in the face of clustered allocation
3232 * requires either O(all block bitmaps) in memory or O(number of block
3233 * groups**2) in time. We will still calculate the superblock for
3234 * older file systems --- and if we come across with a bigalloc file
3235 * system with zero in s_overhead_clusters the estimate will be close to
3236 * correct especially for very large cluster sizes --- but for newer
3237 * file systems, it's better to calculate this figure once at mkfs
3238 * time, and store it in the superblock. If the superblock value is
3239 * present (even for non-bigalloc file systems), we will use it.
3240 */
3241static int count_overhead(struct super_block *sb, ext4_group_t grp,
3242 char *buf)
3243{
3244 struct ext4_sb_info *sbi = EXT4_SB(sb);
3245 struct ext4_group_desc *gdp;
3246 ext4_fsblk_t first_block, last_block, b;
3247 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3248 int s, j, count = 0;
3249
0548bbb8
TT		 3250 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3251 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3252 sbi->s_itb_per_group + 2);
3253
952fc18e
TT		 3254 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3255 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3256 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3257 for (i = 0; i < ngroups; i++) {
3258 gdp = ext4_get_group_desc(sb, i, NULL);
3259 b = ext4_block_bitmap(sb, gdp);
3260 if (b >= first_block && b <= last_block) {
3261 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3262 count++;
3263 }
3264 b = ext4_inode_bitmap(sb, gdp);
3265 if (b >= first_block && b <= last_block) {
3266 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3267 count++;
3268 }
3269 b = ext4_inode_table(sb, gdp);
3270 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3271 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3272 int c = EXT4_B2C(sbi, b - first_block);
3273 ext4_set_bit(c, buf);
3274 count++;
3275 }
3276 if (i != grp)
3277 continue;
3278 s = 0;
3279 if (ext4_bg_has_super(sb, grp)) {
3280 ext4_set_bit(s++, buf);
3281 count++;
3282 }
3283 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3284 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3285 count++;
3286 }
3287 }
3288 if (!count)
3289 return 0;
3290 return EXT4_CLUSTERS_PER_GROUP(sb) -
3291 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3292}
3293
3294/*
3295 * Compute the overhead and stash it in sbi->s_overhead
3296 */
3297int ext4_calculate_overhead(struct super_block *sb)
3298{
3299 struct ext4_sb_info *sbi = EXT4_SB(sb);
3300 struct ext4_super_block *es = sbi->s_es;
3301 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3302 ext4_fsblk_t overhead = 0;
3303 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3304
952fc18e
TT		 3305 if (!buf)
3306 return -ENOMEM;
3307
3308 /*
3309 * Compute the overhead (FS structures). This is constant
3310 * for a given filesystem unless the number of block groups
3311 * changes so we cache the previous value until it does.
3312 */
3313
3314 /*
3315 * All of the blocks before first_data_block are overhead
3316 */
3317 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3318
3319 /*
3320 * Add the overhead found in each block group
3321 */
3322 for (i = 0; i < ngroups; i++) {
3323 int blks;
3324
3325 blks = count_overhead(sb, i, buf);
3326 overhead += blks;
3327 if (blks)
3328 memset(buf, 0, PAGE_SIZE);
3329 cond_resched();
3330 }
0875a2b4
ES		 3331 /* Add the journal blocks as well */
3332 if (sbi->s_journal)
810da240 3333 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3334
952fc18e
TT		 3335 sbi->s_overhead = overhead;
3336 smp_wmb();
3337 free_page((unsigned long) buf);
3338 return 0;
3339}
3340
27dd4385 3341
30fac0f7 3342static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
27dd4385
LC		 3343{
3344 ext4_fsblk_t resv_clusters;
3345
30fac0f7
JK		 3346 /*
3347 * There's no need to reserve anything when we aren't using extents.
3348 * The space estimates are exact, there are no unwritten extents,
3349 * hole punching doesn't need new metadata... This is needed especially
3350 * to keep ext2/3 backward compatibility.
3351 */
3352 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3353 return 0;
27dd4385
LC		 3354 /*
3355 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3356 * This should cover the situations where we can not afford to run
3357 * out of space like for example punch hole, or converting
556615dc 3358 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3359 * allocation would require 1, or 2 blocks, higher numbers are
3360 * very rare.
3361 */
30fac0f7
JK		 3362 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3363 EXT4_SB(sb)->s_cluster_bits;
27dd4385
LC		 3364
3365 do_div(resv_clusters, 50);
3366 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3367
3368 return resv_clusters;
3369}
3370
3371
3372static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3373{
3374 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3375 sbi->s_cluster_bits;
3376
3377 if (count >= clusters)
3378 return -EINVAL;
3379
3380 atomic64_set(&sbi->s_resv_clusters, count);
3381 return 0;
3382}
3383
2b2d6d01 3384static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3385{
d4c402d9 3386 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3387 struct buffer_head *bh;
617ba13b
MC		 3388 struct ext4_super_block *es = NULL;
3389 struct ext4_sb_info *sbi;
3390 ext4_fsblk_t block;
3391 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3392 ext4_fsblk_t logical_sb_block;
ac27a0ec 3393 unsigned long offset = 0;
ac27a0ec
DK		 3394 unsigned long journal_devnum = 0;
3395 unsigned long def_mount_opts;
3396 struct inode *root;
9f6200bb 3397 char *cp;
0390131b 3398 const char *descr;
dcc7dae3 3399 int ret = -ENOMEM;
281b5995 3400 int blocksize, clustersize;
4ec11028
TT		 3401 unsigned int db_count;
3402 unsigned int i;
281b5995 3403 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3404 __u64 blocks_count;
07aa2ea1 3405 int err = 0;
b3881f74 3406 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3407 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3408
3409 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3410 if (!sbi)
dcc7dae3 3411 goto out_free_orig;
705895b6
PE		 3412
3413 sbi->s_blockgroup_lock =
3414 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3415 if (!sbi->s_blockgroup_lock) {
3416 kfree(sbi);
dcc7dae3 3417 goto out_free_orig;
705895b6 3418 }
ac27a0ec 3419 sb->s_fs_info = sbi;
2c0544b2 3420 sbi->s_sb = sb;
240799cd 3421 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3422 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3423 if (sb->s_bdev->bd_part)
3424 sbi->s_sectors_written_start =
3425 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3426
9f6200bb
TT		 3427 /* Cleanup superblock name */
3428 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3429 *cp = '!';
3430
07aa2ea1 3431 /* -EINVAL is default */
dcc7dae3 3432 ret = -EINVAL;
617ba13b 3433 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3434 if (!blocksize) {
b31e1552 3435 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3436 goto out_fail;
3437 }
3438
3439 /*
617ba13b 3440 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3441 * block sizes. We need to calculate the offset from buffer start.
3442 */
617ba13b 3443 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3444 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3445 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3446 } else {
70bbb3e0 3447 logical_sb_block = sb_block;
ac27a0ec
DK		 3448 }
3449
a8ac900b 3450 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3451 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3452 goto out_fail;
3453 }
3454 /*
3455 * Note: s_es must be initialized as soon as possible because
617ba13b 3456 * some ext4 macro-instructions depend on its value
ac27a0ec 3457 */
2716b802 3458 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3459 sbi->s_es = es;
3460 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3461 if (sb->s_magic != EXT4_SUPER_MAGIC)
3462 goto cantfind_ext4;
afc32f7e 3463 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3464
feb0ab32
DW		 3465 /* Warn if metadata_csum and gdt_csum are both set. */
3466 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3467 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3468 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3469 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3470 "redundant flags; please run fsck.");
3471
d25425f8
DW		 3472 /* Check for a known checksum algorithm */
3473 if (!ext4_verify_csum_type(sb, es)) {
3474 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3475 "unknown checksum algorithm.");
3476 silent = 1;
3477 goto cantfind_ext4;
3478 }
3479
0441984a
DW		 3480 /* Load the checksum driver */
3481 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3482 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3483 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3484 if (IS_ERR(sbi->s_chksum_driver)) {
3485 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3486 ret = PTR_ERR(sbi->s_chksum_driver);
3487 sbi->s_chksum_driver = NULL;
3488 goto failed_mount;
3489 }
3490 }
3491
a9c47317
DW		 3492 /* Check superblock checksum */
3493 if (!ext4_superblock_csum_verify(sb, es)) {
3494 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3495 "invalid superblock checksum. Run e2fsck?");
3496 silent = 1;
3497 goto cantfind_ext4;
3498 }
3499
3500 /* Precompute checksum seed for all metadata */
9aa5d32b 3501 if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3502 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3503 sizeof(es->s_uuid));
3504
ac27a0ec
DK		 3505 /* Set defaults before we parse the mount options */
3506 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3507 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3508 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3509 set_opt(sb, DEBUG);
87f26807 3510 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3511 set_opt(sb, GRPID);
617ba13b 3512 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3513 set_opt(sb, NO_UID32);
ea663336 3514 /* xattr user namespace & acls are now defaulted on */
ea663336 3515 set_opt(sb, XATTR_USER);
03010a33 3516#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3517 set_opt(sb, POSIX_ACL);
2e7842b8 3518#endif
98c1a759
DW		 3519 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3520 if (ext4_has_metadata_csum(sb))
3521 set_opt(sb, JOURNAL_CHECKSUM);
3522
617ba13b 3523 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3524 set_opt(sb, JOURNAL_DATA);
617ba13b 3525 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3526 set_opt(sb, ORDERED_DATA);
617ba13b 3527 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3528 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3529
3530 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3531 set_opt(sb, ERRORS_PANIC);
bb4f397a 3532 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3533 set_opt(sb, ERRORS_CONT);
bb4f397a 3534 else
fd8c37ec 3535 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3536 /* block_validity enabled by default; disable with noblock_validity */
3537 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3538 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3539 set_opt(sb, DISCARD);
ac27a0ec 3540
08cefc7a
EB		 3541 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3542 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3543 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3544 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3545 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3546
8b67f04a 3547 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3548 set_opt(sb, BARRIER);
ac27a0ec 3549
dd919b98
AK		 3550 /*
3551 * enable delayed allocation by default
3552 * Use -o nodelalloc to turn it off
3553 */
bc0b75f7 3554 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3555 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3556 set_opt(sb, DELALLOC);
dd919b98 3557
51ce6511
LC		 3558 /*
3559 * set default s_li_wait_mult for lazyinit, for the case there is
3560 * no mount option specified.
3561 */
3562 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3563
8b67f04a 3564 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3565 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3566 ext4_msg(sb, KERN_WARNING,
3567 "failed to parse options in superblock: %s",
3568 sbi->s_es->s_mount_opts);
3569 }
5a916be1 3570 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3571 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3572 &journal_ioprio, 0))
ac27a0ec
DK		 3573 goto failed_mount;
3574
56889787
TT		 3575 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3576 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3577 "with data=journal disables delayed "
3578 "allocation and O_DIRECT support!\n");
3579 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3580 ext4_msg(sb, KERN_ERR, "can't mount with "
3581 "both data=journal and delalloc");
3582 goto failed_mount;
3583 }
3584 if (test_opt(sb, DIOREAD_NOLOCK)) {
3585 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3586 "both data=journal and dioread_nolock");
56889787
TT		 3587 goto failed_mount;
3588 }
3589 if (test_opt(sb, DELALLOC))
3590 clear_opt(sb, DELALLOC);
3591 }
3592
ac27a0ec 3593 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3594 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3595
617ba13b
MC		 3596 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3597 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3598 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3599 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3600 ext4_msg(sb, KERN_WARNING,
3601 "feature flags set on rev 0 fs, "
3602 "running e2fsck is recommended");
469108ff 3603
ed3654eb
TT		 3604 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3605 set_opt2(sb, HURD_COMPAT);
3606 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3607 EXT4_FEATURE_INCOMPAT_64BIT)) {
3608 ext4_msg(sb, KERN_ERR,
3609 "The Hurd can't support 64-bit file systems");
3610 goto failed_mount;
3611 }
3612 }
3613
2035e776
TT		 3614 if (IS_EXT2_SB(sb)) {
3615 if (ext2_feature_set_ok(sb))
3616 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3617 "using the ext4 subsystem");
3618 else {
3619 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3620 "to feature incompatibilities");
3621 goto failed_mount;
3622 }
3623 }
3624
3625 if (IS_EXT3_SB(sb)) {
3626 if (ext3_feature_set_ok(sb))
3627 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3628 "using the ext4 subsystem");
3629 else {
3630 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3631 "to feature incompatibilities");
3632 goto failed_mount;
3633 }
3634 }
3635
ac27a0ec
DK		 3636 /*
3637 * Check feature flags regardless of the revision level, since we
3638 * previously didn't change the revision level when setting the flags,
3639 * so there is a chance incompat flags are set on a rev 0 filesystem.
3640 */
a13fb1a4 3641 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3642 goto failed_mount;
a13fb1a4 3643
261cb20c 3644 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3645 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3646 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3647 ext4_msg(sb, KERN_ERR,
3648 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3649 goto failed_mount;
3650 }
3651
ac27a0ec 3652 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3653 /* Validate the filesystem blocksize */
3654 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3655 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3656 blocksize);
ac27a0ec
DK		 3657 goto failed_mount;
3658 }
3659
2b2d6d01 3660 brelse(bh);
70bbb3e0
AM		 3661 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3662 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3663 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3664 if (!bh) {
b31e1552
ES		 3665 ext4_msg(sb, KERN_ERR,
3666 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3667 goto failed_mount;
3668 }
2716b802 3669 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3670 sbi->s_es = es;
617ba13b 3671 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3672 ext4_msg(sb, KERN_ERR,
3673 "Magic mismatch, very weird!");
ac27a0ec
DK		 3674 goto failed_mount;
3675 }
3676 }
3677
a13fb1a4
ES		 3678 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3679 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3680 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3681 has_huge_files);
3682 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3683
617ba13b
MC		 3684 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3685 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3686 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3687 } else {
3688 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3689 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3690 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3691 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3692 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3693 ext4_msg(sb, KERN_ERR,
3694 "unsupported inode size: %d",
2b2d6d01 3695 sbi->s_inode_size);
ac27a0ec
DK		 3696 goto failed_mount;
3697 }
ef7f3835
KS		 3698 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3699 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3700 }
0b8e58a1 3701
0d1ee42f
AR		 3702 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3703 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3704 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3705 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3706 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3707 ext4_msg(sb, KERN_ERR,
3708 "unsupported descriptor size %lu",
0d1ee42f
AR		 3709 sbi->s_desc_size);
3710 goto failed_mount;
3711 }
3712 } else
3713 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3714
ac27a0ec 3715 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3716 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3717 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3718 goto cantfind_ext4;
0b8e58a1 3719
617ba13b 3720 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3721 if (sbi->s_inodes_per_block == 0)
617ba13b 3722 goto cantfind_ext4;
ac27a0ec
DK		 3723 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3724 sbi->s_inodes_per_block;
0d1ee42f 3725 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3726 sbi->s_sbh = bh;
3727 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3728 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3729 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3730
2b2d6d01 3731 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3732 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3733 sbi->s_def_hash_version = es->s_def_hash_version;
23301410
TT		 3734 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3735 i = le32_to_cpu(es->s_flags);
3736 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3737 sbi->s_hash_unsigned = 3;
3738 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3739#ifdef __CHAR_UNSIGNED__
23301410
TT		 3740 if (!(sb->s_flags & MS_RDONLY))
3741 es->s_flags |=
3742 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3743 sbi->s_hash_unsigned = 3;
f99b2589 3744#else
23301410
TT		 3745 if (!(sb->s_flags & MS_RDONLY))
3746 es->s_flags |=
3747 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3748#endif
23301410 3749 }
f99b2589 3750 }
ac27a0ec 3751
281b5995
TT		 3752 /* Handle clustersize */
3753 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3754 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3755 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3756 if (has_bigalloc) {
3757 if (clustersize < blocksize) {
3758 ext4_msg(sb, KERN_ERR,
3759 "cluster size (%d) smaller than "
3760 "block size (%d)", clustersize, blocksize);
3761 goto failed_mount;
3762 }
3763 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3764 le32_to_cpu(es->s_log_block_size);
3765 sbi->s_clusters_per_group =
3766 le32_to_cpu(es->s_clusters_per_group);
3767 if (sbi->s_clusters_per_group > blocksize * 8) {
3768 ext4_msg(sb, KERN_ERR,
3769 "#clusters per group too big: %lu",
3770 sbi->s_clusters_per_group);
3771 goto failed_mount;
3772 }
3773 if (sbi->s_blocks_per_group !=
3774 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3775 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3776 "clusters per group (%lu) inconsistent",
3777 sbi->s_blocks_per_group,
3778 sbi->s_clusters_per_group);
3779 goto failed_mount;
3780 }
3781 } else {
3782 if (clustersize != blocksize) {
3783 ext4_warning(sb, "fragment/cluster size (%d) != "
3784 "block size (%d)", clustersize,
3785 blocksize);
3786 clustersize = blocksize;
3787 }
3788 if (sbi->s_blocks_per_group > blocksize * 8) {
3789 ext4_msg(sb, KERN_ERR,
3790 "#blocks per group too big: %lu",
3791 sbi->s_blocks_per_group);
3792 goto failed_mount;
3793 }
3794 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3795 sbi->s_cluster_bits = 0;
ac27a0ec 3796 }
281b5995
TT		 3797 sbi->s_cluster_ratio = clustersize / blocksize;
3798
ac27a0ec 3799 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3800 ext4_msg(sb, KERN_ERR,
3801 "#inodes per group too big: %lu",
2b2d6d01 3802 sbi->s_inodes_per_group);
ac27a0ec
DK		 3803 goto failed_mount;
3804 }
3805
960fd856
TT		 3806 /* Do we have standard group size of clustersize * 8 blocks ? */
3807 if (sbi->s_blocks_per_group == clustersize << 3)
3808 set_opt2(sb, STD_GROUP_SIZE);
3809
bf43d84b
ES		 3810 /*
3811 * Test whether we have more sectors than will fit in sector_t,
3812 * and whether the max offset is addressable by the page cache.
3813 */
5a9ae68a 3814 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3815 ext4_blocks_count(es));
5a9ae68a 3816 if (err) {
b31e1552 3817 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3818 " too large to mount safely on this system");
ac27a0ec 3819 if (sizeof(sector_t) < 8)
90c699a9 3820 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3821 goto failed_mount;
3822 }
3823
617ba13b
MC		 3824 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3825 goto cantfind_ext4;
e7c95593 3826
0f2ddca6
FTN		 3827 /* check blocks count against device size */
3828 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3829 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3830 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3831 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3832 ext4_blocks_count(es), blocks_count);
3833 goto failed_mount;
3834 }
3835
0b8e58a1
AD		 3836 /*
3837 * It makes no sense for the first data block to be beyond the end
3838 * of the filesystem.
3839 */
3840 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3841 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3842 "block %u is beyond end of filesystem (%llu)",
3843 le32_to_cpu(es->s_first_data_block),
3844 ext4_blocks_count(es));
e7c95593
ES		 3845 goto failed_mount;
3846 }
bd81d8ee
LV		 3847 blocks_count = (ext4_blocks_count(es) -
3848 le32_to_cpu(es->s_first_data_block) +
3849 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3850 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3851 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3852 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3853 "(block count %llu, first data block %u, "
b31e1552 3854 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3855 ext4_blocks_count(es),
3856 le32_to_cpu(es->s_first_data_block),
3857 EXT4_BLOCKS_PER_GROUP(sb));
3858 goto failed_mount;
3859 }
bd81d8ee 3860 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3861 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3862 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3863 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3864 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3865 sbi->s_group_desc = ext4_kvmalloc(db_count *
3866 sizeof(struct buffer_head *),
3867 GFP_KERNEL);
ac27a0ec 3868 if (sbi->s_group_desc == NULL) {
b31e1552 3869 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3870 ret = -ENOMEM;
ac27a0ec
DK		 3871 goto failed_mount;
3872 }
3873
9f6200bb
TT		 3874 if (ext4_proc_root)
3875 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3876
66acdcf4
TT		 3877 if (sbi->s_proc)
3878 proc_create_data("options", S_IRUGO, sbi->s_proc,
3879 &ext4_seq_options_fops, sb);
3880
705895b6 3881 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3882
3883 for (i = 0; i < db_count; i++) {
70bbb3e0 3884 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3885 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3886 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3887 ext4_msg(sb, KERN_ERR,
3888 "can't read group descriptor %d", i);
ac27a0ec
DK		 3889 db_count = i;
3890 goto failed_mount2;
3891 }
3892 }
bfff6873 3893 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3894 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
f9ae9cf5 3895 goto failed_mount2;
ac27a0ec 3896 }
772cb7c8 3897
f9ae9cf5 3898 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3899 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3900 spin_lock_init(&sbi->s_next_gen_lock);
3901
04496411
TM		 3902 init_timer(&sbi->s_err_report);
3903 sbi->s_err_report.function = print_daily_error_info;
3904 sbi->s_err_report.data = (unsigned long) sb;
3905
a75ae78f 3906 /* Register extent status tree shrinker */
eb68d0e2 3907 if (ext4_es_register_shrinker(sbi))
ce7e010a 3908 goto failed_mount3;
ce7e010a 3909
c9de560d 3910 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3911 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3912
f9ae9cf5
TT		 3913 /*
3914 * set up enough so that it can read an inode
3915 */
f6e63f90 3916 sb->s_op = &ext4_sops;
617ba13b
MC		 3917 sb->s_export_op = &ext4_export_ops;
3918 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3919#ifdef CONFIG_QUOTA
617ba13b 3920 sb->dq_op = &ext4_quota_operations;
262b4662 3921 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
7c319d32 3922 sb->s_qcop = &ext4_qctl_sysfile_operations;
262b4662
JK		 3923 else
3924 sb->s_qcop = &ext4_qctl_operations;
ac27a0ec 3925#endif
f2fa2ffc
AK		 3926 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3927
ac27a0ec 3928 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3929 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3930
3931 sb->s_root = NULL;
3932
3933 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3934 EXT4_HAS_INCOMPAT_FEATURE(sb,
3935 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3936
c5e06d10
JL		 3937 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3938 !(sb->s_flags & MS_RDONLY))
3939 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3940 goto failed_mount3a;
c5e06d10 3941
ac27a0ec
DK		 3942 /*
3943 * The first inode we look at is the journal inode. Don't try
3944 * root first: it may be modified in the journal!
3945 */
3946 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3947 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3948 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 3949 goto failed_mount3a;
0390131b
FM		 3950 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3951 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3952 ext4_msg(sb, KERN_ERR, "required journal recovery "
3953 "suppressed and not mounted read-only");
744692dc 3954 goto failed_mount_wq;
ac27a0ec 3955 } else {
fd8c37ec 3956 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3957 sbi->s_journal = NULL;
3958 needs_recovery = 0;
3959 goto no_journal;
ac27a0ec
DK		 3960 }
3961
f32aaf2d 3962 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3963 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3964 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3965 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3966 goto failed_mount_wq;
eb40a09c
JS		 3967 }
3968
25ed6e8a
DW		 3969 if (!set_journal_csum_feature_set(sb)) {
3970 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3971 "feature set");
3972 goto failed_mount_wq;
d4da6c9c 3973 }
818d276c 3974
ac27a0ec
DK		 3975 /* We have now updated the journal if required, so we can
3976 * validate the data journaling mode. */
3977 switch (test_opt(sb, DATA_FLAGS)) {
3978 case 0:
3979 /* No mode set, assume a default based on the journal
63f57933
AM		 3980 * capabilities: ORDERED_DATA if the journal can
3981 * cope, else JOURNAL_DATA
3982 */
dab291af
MC		 3983 if (jbd2_journal_check_available_features
3984 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3985 set_opt(sb, ORDERED_DATA);
ac27a0ec 3986 else
fd8c37ec 3987 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3988 break;
3989
617ba13b
MC		 3990 case EXT4_MOUNT_ORDERED_DATA:
3991 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3992 if (!jbd2_journal_check_available_features
3993 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3994 ext4_msg(sb, KERN_ERR, "Journal does not support "
3995 "requested data journaling mode");
744692dc 3996 goto failed_mount_wq;
ac27a0ec
DK		 3997 }
3998 default:
3999 break;
4000 }
b3881f74 4001 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4002
18aadd47
BJ		 4003 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4004
ce7e010a 4005no_journal:
9c191f70
M		 4006 if (ext4_mballoc_ready) {
4007 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4008 if (!sbi->s_mb_cache) {
4009 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4010 goto failed_mount_wq;
4011 }
4012 }
4013
952fc18e
TT		 4014 /*
4015 * Get the # of file system overhead blocks from the
4016 * superblock if present.
4017 */
4018 if (es->s_overhead_clusters)
4019 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4020 else {
07aa2ea1
LC		 4021 err = ext4_calculate_overhead(sb);
4022 if (err)
952fc18e
TT		 4023 goto failed_mount_wq;
4024 }
4025
fd89d5f2
TH		 4026 /*
4027 * The maximum number of concurrent works can be high and
4028 * concurrency isn't really necessary. Limit it to 1.
4029 */
2e8fa54e
JK		 4030 EXT4_SB(sb)->rsv_conversion_wq =
4031 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4032 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4033 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4034 ret = -ENOMEM;
2e8fa54e
JK		 4035 goto failed_mount4;
4036 }
4037
ac27a0ec 4038 /*
dab291af 4039 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4040 * so we can safely mount the rest of the filesystem now.
4041 */
4042
1d1fe1ee
DH		 4043 root = ext4_iget(sb, EXT4_ROOT_INO);
4044 if (IS_ERR(root)) {
b31e1552 4045 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4046 ret = PTR_ERR(root);
32a9bb57 4047 root = NULL;
ac27a0ec
DK		 4048 goto failed_mount4;
4049 }
4050 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4051 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4052 iput(root);
ac27a0ec
DK		 4053 goto failed_mount4;
4054 }
48fde701 4055 sb->s_root = d_make_root(root);
1d1fe1ee 4056 if (!sb->s_root) {
b31e1552 4057 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4058 ret = -ENOMEM;
4059 goto failed_mount4;
4060 }
ac27a0ec 4061
7e84b621
ES		 4062 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4063 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4064
4065 /* determine the minimum size of new large inodes, if present */
4066 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4067 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4068 EXT4_GOOD_OLD_INODE_SIZE;
4069 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4070 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4071 if (sbi->s_want_extra_isize <
4072 le16_to_cpu(es->s_want_extra_isize))
4073 sbi->s_want_extra_isize =
4074 le16_to_cpu(es->s_want_extra_isize);
4075 if (sbi->s_want_extra_isize <
4076 le16_to_cpu(es->s_min_extra_isize))
4077 sbi->s_want_extra_isize =
4078 le16_to_cpu(es->s_min_extra_isize);
4079 }
4080 }
4081 /* Check if enough inode space is available */
4082 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4083 sbi->s_inode_size) {
4084 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4085 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4086 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4087 "available");
ef7f3835
KS		 4088 }
4089
30fac0f7 4090 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
27dd4385
LC		 4091 if (err) {
4092 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
30fac0f7 4093 "reserved pool", ext4_calculate_resv_clusters(sb));
f9ae9cf5 4094 goto failed_mount4a;
27dd4385
LC		 4095 }
4096
6fd058f7
TT		 4097 err = ext4_setup_system_zone(sb);
4098 if (err) {
b31e1552 4099 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4100 "zone (%d)", err);
f9ae9cf5
TT		 4101 goto failed_mount4a;
4102 }
4103
4104 ext4_ext_init(sb);
4105 err = ext4_mb_init(sb);
4106 if (err) {
4107 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4108 err);
dcf2d804 4109 goto failed_mount5;
c2774d84
AK		 4110 }
4111
d5e03cbb
TT		 4112 block = ext4_count_free_clusters(sb);
4113 ext4_free_blocks_count_set(sbi->s_es,
4114 EXT4_C2B(sbi, block));
908c7f19
TH		 4115 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4116 GFP_KERNEL);
d5e03cbb
TT		 4117 if (!err) {
4118 unsigned long freei = ext4_count_free_inodes(sb);
4119 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4120 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4121 GFP_KERNEL);
d5e03cbb
TT		 4122 }
4123 if (!err)
4124 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4125 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4126 if (!err)
908c7f19
TH		 4127 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4128 GFP_KERNEL);
d5e03cbb
TT		 4129 if (err) {
4130 ext4_msg(sb, KERN_ERR, "insufficient memory");
4131 goto failed_mount6;
4132 }
4133
4134 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4135 if (!ext4_fill_flex_info(sb)) {
4136 ext4_msg(sb, KERN_ERR,
4137 "unable to initialize "
4138 "flex_bg meta info!");
4139 goto failed_mount6;
4140 }
4141
bfff6873
LC		 4142 err = ext4_register_li_request(sb, first_not_zeroed);
4143 if (err)
dcf2d804 4144 goto failed_mount6;
bfff6873 4145
3197ebdb
TT		 4146 sbi->s_kobj.kset = ext4_kset;
4147 init_completion(&sbi->s_kobj_unregister);
4148 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4149 "%s", sb->s_id);
dcf2d804
TM		 4150 if (err)
4151 goto failed_mount7;
3197ebdb 4152
9b2ff357
JK		 4153#ifdef CONFIG_QUOTA
4154 /* Enable quota usage during mount. */
4155 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4156 !(sb->s_flags & MS_RDONLY)) {
4157 err = ext4_enable_quotas(sb);
4158 if (err)
4159 goto failed_mount8;
4160 }
4161#endif /* CONFIG_QUOTA */
4162
617ba13b
MC		 4163 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4164 ext4_orphan_cleanup(sb, es);
4165 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4166 if (needs_recovery) {
b31e1552 4167 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4168 ext4_mark_recovery_complete(sb, es);
4169 }
4170 if (EXT4_SB(sb)->s_journal) {
4171 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4172 descr = " journalled data mode";
4173 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4174 descr = " ordered data mode";
4175 else
4176 descr = " writeback data mode";
4177 } else
4178 descr = "out journal";
4179
79add3a3
LC		 4180 if (test_opt(sb, DISCARD)) {
4181 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4182 if (!blk_queue_discard(q))
4183 ext4_msg(sb, KERN_WARNING,
4184 "mounting with \"discard\" option, but "
4185 "the device does not support discard");
4186 }
4187
d4c402d9 4188 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4189 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4190 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4191
66e61a9e
TT		 4192 if (es->s_error_count)
4193 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4194
efbed4dc
TT		 4195 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4196 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4197 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4198 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4199
d4c402d9 4200 kfree(orig_data);
ac27a0ec
DK		 4201 return 0;
4202
617ba13b 4203cantfind_ext4:
ac27a0ec 4204 if (!silent)
b31e1552 4205 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4206 goto failed_mount;
4207
72ba7450
TT		 4208#ifdef CONFIG_QUOTA
4209failed_mount8:
4210 kobject_del(&sbi->s_kobj);
4211#endif
dcf2d804
TM		 4212failed_mount7:
4213 ext4_unregister_li_request(sb);
4214failed_mount6:
f9ae9cf5 4215 ext4_mb_release(sb);
d5e03cbb 4216 if (sbi->s_flex_groups)
b93b41d4 4217 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4218 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4219 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4220 percpu_counter_destroy(&sbi->s_dirs_counter);
4221 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4222failed_mount5:
f9ae9cf5
TT		 4223 ext4_ext_release(sb);
4224 ext4_release_system_zone(sb);
4225failed_mount4a:
94bf608a 4226 dput(sb->s_root);
32a9bb57 4227 sb->s_root = NULL;
94bf608a 4228failed_mount4:
b31e1552 4229 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4230 if (EXT4_SB(sb)->rsv_conversion_wq)
4231 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4232failed_mount_wq:
0390131b
FM		 4233 if (sbi->s_journal) {
4234 jbd2_journal_destroy(sbi->s_journal);
4235 sbi->s_journal = NULL;
4236 }
50460fe8 4237failed_mount3a:
d3922a77 4238 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4239failed_mount3:
9105bb14 4240 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4241 if (sbi->s_mmp_tsk)
4242 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4243failed_mount2:
4244 for (i = 0; i < db_count; i++)
4245 brelse(sbi->s_group_desc[i]);
b93b41d4 4246 kvfree(sbi->s_group_desc);
ac27a0ec 4247failed_mount:
0441984a
DW		 4248 if (sbi->s_chksum_driver)
4249 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4250 if (sbi->s_proc) {
66acdcf4 4251 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4252 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4253 }
ac27a0ec 4254#ifdef CONFIG_QUOTA
a2d4a646 4255 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4256 kfree(sbi->s_qf_names[i]);
4257#endif
617ba13b 4258 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4259 brelse(bh);
4260out_fail:
4261 sb->s_fs_info = NULL;
f6830165 4262 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4263 kfree(sbi);
dcc7dae3 4264out_free_orig:
d4c402d9 4265 kfree(orig_data);
07aa2ea1 4266 return err ? err : ret;
ac27a0ec
DK		 4267}
4268
4269/*
4270 * Setup any per-fs journal parameters now. We'll do this both on
4271 * initial mount, once the journal has been initialised but before we've
4272 * done any recovery; and again on any subsequent remount.
4273 */
617ba13b 4274static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4275{
617ba13b 4276 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4277
30773840
TT		 4278 journal->j_commit_interval = sbi->s_commit_interval;
4279 journal->j_min_batch_time = sbi->s_min_batch_time;
4280 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4281
a931da6a 4282 write_lock(&journal->j_state_lock);
ac27a0ec 4283 if (test_opt(sb, BARRIER))
dab291af 4284 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4285 else
dab291af 4286 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4287 if (test_opt(sb, DATA_ERR_ABORT))
4288 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4289 else
4290 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4291 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4292}
4293
617ba13b 4294static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4295 unsigned int journal_inum)
4296{
4297 struct inode *journal_inode;
4298 journal_t *journal;
4299
0390131b
FM		 4300 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4301
ac27a0ec
DK		 4302 /* First, test for the existence of a valid inode on disk. Bad
4303 * things happen if we iget() an unused inode, as the subsequent
4304 * iput() will try to delete it. */
4305
1d1fe1ee
DH		 4306 journal_inode = ext4_iget(sb, journal_inum);
4307 if (IS_ERR(journal_inode)) {
b31e1552 4308 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4309 return NULL;
4310 }
4311 if (!journal_inode->i_nlink) {
4312 make_bad_inode(journal_inode);
4313 iput(journal_inode);
b31e1552 4314 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4315 return NULL;
4316 }
4317
e5f8eab8 4318 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4319 journal_inode, journal_inode->i_size);
1d1fe1ee 4320 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4321 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4322 iput(journal_inode);
4323 return NULL;
4324 }
4325
dab291af 4326 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4327 if (!journal) {
b31e1552 4328 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4329 iput(journal_inode);
4330 return NULL;
4331 }
4332 journal->j_private = sb;
617ba13b 4333 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4334 return journal;
4335}
4336
617ba13b 4337static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4338 dev_t j_dev)
4339{
2b2d6d01 4340 struct buffer_head *bh;
ac27a0ec 4341 journal_t *journal;
617ba13b
MC		 4342 ext4_fsblk_t start;
4343 ext4_fsblk_t len;
ac27a0ec 4344 int hblock, blocksize;
617ba13b 4345 ext4_fsblk_t sb_block;
ac27a0ec 4346 unsigned long offset;
2b2d6d01 4347 struct ext4_super_block *es;
ac27a0ec
DK		 4348 struct block_device *bdev;
4349
0390131b
FM		 4350 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4351
b31e1552 4352 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4353 if (bdev == NULL)
4354 return NULL;
4355
ac27a0ec 4356 blocksize = sb->s_blocksize;
e1defc4f 4357 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4358 if (blocksize < hblock) {
b31e1552
ES		 4359 ext4_msg(sb, KERN_ERR,
4360 "blocksize too small for journal device");
ac27a0ec
DK		 4361 goto out_bdev;
4362 }
4363
617ba13b
MC		 4364 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4365 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4366 set_blocksize(bdev, blocksize);
4367 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4368 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4369 "external journal");
ac27a0ec
DK		 4370 goto out_bdev;
4371 }
4372
2716b802 4373 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4374 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4375 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4376 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4377 ext4_msg(sb, KERN_ERR, "external journal has "
4378 "bad superblock");
ac27a0ec
DK		 4379 brelse(bh);
4380 goto out_bdev;
4381 }
4382
df4763be
DW		 4383 if ((le32_to_cpu(es->s_feature_ro_compat) &
4384 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4385 es->s_checksum != ext4_superblock_csum(sb, es)) {
4386 ext4_msg(sb, KERN_ERR, "external journal has "
4387 "corrupt superblock");
4388 brelse(bh);
4389 goto out_bdev;
4390 }
4391
617ba13b 4392 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4393 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4394 brelse(bh);
4395 goto out_bdev;
4396 }
4397
bd81d8ee 4398 len = ext4_blocks_count(es);
ac27a0ec
DK		 4399 start = sb_block + 1;
4400 brelse(bh); /* we're done with the superblock */
4401
dab291af 4402 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4403 start, len, blocksize);
4404 if (!journal) {
b31e1552 4405 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4406 goto out_bdev;
4407 }
4408 journal->j_private = sb;
9f203507 4409 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4410 wait_on_buffer(journal->j_sb_buffer);
4411 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4412 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4413 goto out_journal;
4414 }
4415 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4416 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4417 "user (unsupported) - %d",
ac27a0ec
DK		 4418 be32_to_cpu(journal->j_superblock->s_nr_users));
4419 goto out_journal;
4420 }
617ba13b
MC		 4421 EXT4_SB(sb)->journal_bdev = bdev;
4422 ext4_init_journal_params(sb, journal);
ac27a0ec 4423 return journal;
0b8e58a1 4424
ac27a0ec 4425out_journal:
dab291af 4426 jbd2_journal_destroy(journal);
ac27a0ec 4427out_bdev:
617ba13b 4428 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4429 return NULL;
4430}
4431
617ba13b
MC		 4432static int ext4_load_journal(struct super_block *sb,
4433 struct ext4_super_block *es,
ac27a0ec
DK		 4434 unsigned long journal_devnum)
4435{
4436 journal_t *journal;
4437 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4438 dev_t journal_dev;
4439 int err = 0;
4440 int really_read_only;
4441
0390131b
FM		 4442 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4443
ac27a0ec
DK		 4444 if (journal_devnum &&
4445 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4446 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4447 "numbers have changed");
ac27a0ec
DK		 4448 journal_dev = new_decode_dev(journal_devnum);
4449 } else
4450 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4451
4452 really_read_only = bdev_read_only(sb->s_bdev);
4453
4454 /*
4455 * Are we loading a blank journal or performing recovery after a
4456 * crash? For recovery, we need to check in advance whether we
4457 * can get read-write access to the device.
4458 */
617ba13b 4459 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4460 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4461 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4462 "required on readonly filesystem");
ac27a0ec 4463 if (really_read_only) {
b31e1552
ES		 4464 ext4_msg(sb, KERN_ERR, "write access "
4465 "unavailable, cannot proceed");
ac27a0ec
DK		 4466 return -EROFS;
4467 }
b31e1552
ES		 4468 ext4_msg(sb, KERN_INFO, "write access will "
4469 "be enabled during recovery");
ac27a0ec
DK		 4470 }
4471 }
4472
4473 if (journal_inum && journal_dev) {
b31e1552
ES		 4474 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4475 "and inode journals!");
ac27a0ec
DK		 4476 return -EINVAL;
4477 }
4478
4479 if (journal_inum) {
617ba13b 4480 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4481 return -EINVAL;
4482 } else {
617ba13b 4483 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4484 return -EINVAL;
4485 }
4486
90576c0b 4487 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4488 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4489
617ba13b 4490 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4491 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4492 if (!err) {
4493 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4494 if (save)
4495 memcpy(save, ((char *) es) +
4496 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4497 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4498 if (save)
4499 memcpy(((char *) es) + EXT4_S_ERR_START,
4500 save, EXT4_S_ERR_LEN);
4501 kfree(save);
4502 }
ac27a0ec
DK		 4503
4504 if (err) {
b31e1552 4505 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4506 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4507 return err;
4508 }
4509
617ba13b
MC		 4510 EXT4_SB(sb)->s_journal = journal;
4511 ext4_clear_journal_err(sb, es);
ac27a0ec 4512
c41303ce 4513 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4514 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4515 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4516
4517 /* Make sure we flush the recovery flag to disk. */
e2d67052 4518 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4519 }
4520
4521 return 0;
4522}
4523
e2d67052 4524static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4525{
e2d67052 4526 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4527 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4528 int error = 0;
ac27a0ec 4529
7c2e7087 4530 if (!sbh || block_device_ejected(sb))
c4be0c1d 4531 return error;
914258bf
TT		 4532 if (buffer_write_io_error(sbh)) {
4533 /*
4534 * Oh, dear. A previous attempt to write the
4535 * superblock failed. This could happen because the
4536 * USB device was yanked out. Or it could happen to
4537 * be a transient write error and maybe the block will
4538 * be remapped. Nothing we can do but to retry the
4539 * write and hope for the best.
4540 */
b31e1552
ES		 4541 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4542 "superblock detected");
914258bf
TT		 4543 clear_buffer_write_io_error(sbh);
4544 set_buffer_uptodate(sbh);
4545 }
71290b36
TT		 4546 /*
4547 * If the file system is mounted read-only, don't update the
4548 * superblock write time. This avoids updating the superblock
4549 * write time when we are mounting the root file system
4550 * read/only but we need to replay the journal; at that point,
4551 * for people who are east of GMT and who make their clock
4552 * tick in localtime for Windows bug-for-bug compatibility,
4553 * the clock is set in the future, and this will cause e2fsck
4554 * to complain and force a full file system check.
4555 */
4556 if (!(sb->s_flags & MS_RDONLY))
4557 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4558 if (sb->s_bdev->bd_part)
4559 es->s_kbytes_written =
4560 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4561 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4562 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4563 else
4564 es->s_kbytes_written =
4565 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4566 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4567 ext4_free_blocks_count_set(es,
57042651
TT		 4568 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4569 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4570 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4571 es->s_free_inodes_count =
4572 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4573 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4574 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4575 ext4_superblock_csum_set(sb);
ac27a0ec 4576 mark_buffer_dirty(sbh);
914258bf 4577 if (sync) {
c4be0c1d
TS		 4578 error = sync_dirty_buffer(sbh);
4579 if (error)
4580 return error;
4581
4582 error = buffer_write_io_error(sbh);
4583 if (error) {
b31e1552
ES		 4584 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4585 "superblock");
914258bf
TT		 4586 clear_buffer_write_io_error(sbh);
4587 set_buffer_uptodate(sbh);
4588 }
4589 }
c4be0c1d 4590 return error;
ac27a0ec
DK		 4591}
4592
ac27a0ec
DK		 4593/*
4594 * Have we just finished recovery? If so, and if we are mounting (or
4595 * remounting) the filesystem readonly, then we will end up with a
4596 * consistent fs on disk. Record that fact.
4597 */
2b2d6d01
TT		 4598static void ext4_mark_recovery_complete(struct super_block *sb,
4599 struct ext4_super_block *es)
ac27a0ec 4600{
617ba13b 4601 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4602
0390131b
FM		 4603 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4604 BUG_ON(journal != NULL);
4605 return;
4606 }
dab291af 4607 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4608 if (jbd2_journal_flush(journal) < 0)
4609 goto out;
4610
617ba13b 4611 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4612 sb->s_flags & MS_RDONLY) {
617ba13b 4613 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4614 ext4_commit_super(sb, 1);
ac27a0ec 4615 }
7ffe1ea8
HK		 4616
4617out:
dab291af 4618 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4619}
4620
4621/*
4622 * If we are mounting (or read-write remounting) a filesystem whose journal
4623 * has recorded an error from a previous lifetime, move that error to the
4624 * main filesystem now.
4625 */
2b2d6d01
TT		 4626static void ext4_clear_journal_err(struct super_block *sb,
4627 struct ext4_super_block *es)
ac27a0ec
DK		 4628{
4629 journal_t *journal;
4630 int j_errno;
4631 const char *errstr;
4632
0390131b
FM		 4633 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4634
617ba13b 4635 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4636
4637 /*
4638 * Now check for any error status which may have been recorded in the
617ba13b 4639 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4640 */
4641
dab291af 4642 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4643 if (j_errno) {
4644 char nbuf[16];
4645
617ba13b 4646 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4647 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4648 "from previous mount: %s", errstr);
12062ddd 4649 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4650
617ba13b
MC		 4651 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4652 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4653 ext4_commit_super(sb, 1);
ac27a0ec 4654
dab291af 4655 jbd2_journal_clear_err(journal);
d796c52e 4656 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4657 }
4658}
4659
4660/*
4661 * Force the running and committing transactions to commit,
4662 * and wait on the commit.
4663 */
617ba13b 4664int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4665{
4666 journal_t *journal;
ac27a0ec
DK		 4667
4668 if (sb->s_flags & MS_RDONLY)
4669 return 0;
4670
617ba13b 4671 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4672 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4673}
4674
617ba13b 4675static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4676{
14ce0cb4 4677 int ret = 0;
9eddacf9 4678 tid_t target;
06a407f1 4679 bool needs_barrier = false;
8d5d02e6 4680 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4681
9bffad1e 4682 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4683 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4684 /*
4685 * Writeback quota in non-journalled quota case - journalled quota has
4686 * no dirty dquots
4687 */
4688 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4689 /*
4690 * Data writeback is possible w/o journal transaction, so barrier must
4691 * being sent at the end of the function. But we can skip it if
4692 * transaction_commit will do it for us.
4693 */
bda32530
TT		 4694 if (sbi->s_journal) {
4695 target = jbd2_get_latest_transaction(sbi->s_journal);
4696 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4697 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4698 needs_barrier = true;
4699
4700 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4701 if (wait)
4702 ret = jbd2_log_wait_commit(sbi->s_journal,
4703 target);
4704 }
4705 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4706 needs_barrier = true;
06a407f1
DM		 4707 if (needs_barrier) {
4708 int err;
4709 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4710 if (!ret)
4711 ret = err;
0390131b 4712 }
06a407f1
DM		 4713
4714 return ret;
4715}
4716
ac27a0ec
DK		 4717/*
4718 * LVM calls this function before a (read-only) snapshot is created. This
4719 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4720 *
4721 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4722 * state independently. It relies on upper layer to stop all data & metadata
4723 * modifications.
ac27a0ec 4724 */
c4be0c1d 4725static int ext4_freeze(struct super_block *sb)
ac27a0ec 4726{
c4be0c1d
TS		 4727 int error = 0;
4728 journal_t *journal;
ac27a0ec 4729
9ca92389
TT		 4730 if (sb->s_flags & MS_RDONLY)
4731 return 0;
ac27a0ec 4732
9ca92389 4733 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4734
bb044576
TT		 4735 if (journal) {
4736 /* Now we set up the journal barrier. */
4737 jbd2_journal_lock_updates(journal);
ac27a0ec 4738
bb044576
TT		 4739 /*
4740 * Don't clear the needs_recovery flag if we failed to
4741 * flush the journal.
4742 */
4743 error = jbd2_journal_flush(journal);
4744 if (error < 0)
4745 goto out;
4746 }
9ca92389
TT		 4747
4748 /* Journal blocked and flushed, clear needs_recovery flag. */
4749 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4750 error = ext4_commit_super(sb, 1);
6b0310fb 4751out:
bb044576
TT		 4752 if (journal)
4753 /* we rely on upper layer to stop further updates */
4754 jbd2_journal_unlock_updates(journal);
6b0310fb 4755 return error;
ac27a0ec
DK		 4756}
4757
4758/*
4759 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4760 * flag here, even though the filesystem is not technically dirty yet.
4761 */
c4be0c1d 4762static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4763{
9ca92389
TT		 4764 if (sb->s_flags & MS_RDONLY)
4765 return 0;
4766
9ca92389
TT		 4767 /* Reset the needs_recovery flag before the fs is unlocked. */
4768 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4769 ext4_commit_super(sb, 1);
c4be0c1d 4770 return 0;
ac27a0ec
DK		 4771}
4772
673c6100
TT		 4773/*
4774 * Structure to save mount options for ext4_remount's benefit
4775 */
4776struct ext4_mount_options {
4777 unsigned long s_mount_opt;
a2595b8a 4778 unsigned long s_mount_opt2;
08cefc7a
EB		 4779 kuid_t s_resuid;
4780 kgid_t s_resgid;
673c6100
TT		 4781 unsigned long s_commit_interval;
4782 u32 s_min_batch_time, s_max_batch_time;
4783#ifdef CONFIG_QUOTA
4784 int s_jquota_fmt;
a2d4a646 4785 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4786#endif
4787};
4788
2b2d6d01 4789static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4790{
2b2d6d01 4791 struct ext4_super_block *es;
617ba13b 4792 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4793 unsigned long old_sb_flags;
617ba13b 4794 struct ext4_mount_options old_opts;
c79d967d 4795 int enable_quota = 0;
8a266467 4796 ext4_group_t g;
b3881f74 4797 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4798 int err = 0;
ac27a0ec 4799#ifdef CONFIG_QUOTA
03dafb5f 4800 int i, j;
ac27a0ec 4801#endif
d4c402d9 4802 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4803
4804 /* Store the original options */
4805 old_sb_flags = sb->s_flags;
4806 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4807 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4808 old_opts.s_resuid = sbi->s_resuid;
4809 old_opts.s_resgid = sbi->s_resgid;
4810 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4811 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4812 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4813#ifdef CONFIG_QUOTA
4814 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4815 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4816 if (sbi->s_qf_names[i]) {
4817 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4818 GFP_KERNEL);
4819 if (!old_opts.s_qf_names[i]) {
4820 for (j = 0; j < i; j++)
4821 kfree(old_opts.s_qf_names[j]);
3e36a163 4822 kfree(orig_data);
03dafb5f
CG		 4823 return -ENOMEM;
4824 }
4825 } else
4826 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4827#endif
b3881f74
TT		 4828 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4829 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4830
4831 /*
4832 * Allow the "check" option to be passed as a remount option.
4833 */
661aa520 4834 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4835 err = -EINVAL;
4836 goto restore_opts;
4837 }
4838
6b992ff2
DW		 4839 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4840 test_opt(sb, JOURNAL_CHECKSUM)) {
4841 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4842 "during remount not supported");
4843 err = -EINVAL;
4844 goto restore_opts;
4845 }
4846
6ae6514b
PS		 4847 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4848 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4849 ext4_msg(sb, KERN_ERR, "can't mount with "
4850 "both data=journal and delalloc");
4851 err = -EINVAL;
4852 goto restore_opts;
4853 }
4854 if (test_opt(sb, DIOREAD_NOLOCK)) {
4855 ext4_msg(sb, KERN_ERR, "can't mount with "
4856 "both data=journal and dioread_nolock");
4857 err = -EINVAL;
4858 goto restore_opts;
4859 }
4860 }
4861
4ab2f15b 4862 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4863 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4864
4865 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4866 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4867
4868 es = sbi->s_es;
4869
b3881f74 4870 if (sbi->s_journal) {
0390131b 4871 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4872 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4873 }
ac27a0ec 4874
661aa520 4875 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4876 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4877 err = -EROFS;
4878 goto restore_opts;
4879 }
4880
4881 if (*flags & MS_RDONLY) {
38c03b34
TT		 4882 err = sync_filesystem(sb);
4883 if (err < 0)
4884 goto restore_opts;
0f0dd62f
CH		 4885 err = dquot_suspend(sb, -1);
4886 if (err < 0)
c79d967d 4887 goto restore_opts;
c79d967d 4888
ac27a0ec
DK		 4889 /*
4890 * First of all, the unconditional stuff we have to do
4891 * to disable replay of the journal when we next remount
4892 */
4893 sb->s_flags |= MS_RDONLY;
4894
4895 /*
4896 * OK, test if we are remounting a valid rw partition
4897 * readonly, and if so set the rdonly flag and then
4898 * mark the partition as valid again.
4899 */
617ba13b
MC		 4900 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4901 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4902 es->s_state = cpu_to_le16(sbi->s_mount_state);
4903
a63c9eb2 4904 if (sbi->s_journal)
0390131b 4905 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4906 } else {
a13fb1a4
ES		 4907 /* Make sure we can mount this feature set readwrite */
4908 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4909 err = -EROFS;
4910 goto restore_opts;
4911 }
8a266467
TT		 4912 /*
4913 * Make sure the group descriptor checksums
0b8e58a1 4914 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4915 */
4916 for (g = 0; g < sbi->s_groups_count; g++) {
4917 struct ext4_group_desc *gdp =
4918 ext4_get_group_desc(sb, g, NULL);
4919
feb0ab32 4920 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4921 ext4_msg(sb, KERN_ERR,
4922 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4923 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4924 le16_to_cpu(gdp->bg_checksum));
4925 err = -EINVAL;
4926 goto restore_opts;
4927 }
4928 }
4929
ead6596b
ES		 4930 /*
4931 * If we have an unprocessed orphan list hanging
4932 * around from a previously readonly bdev mount,
4933 * require a full umount/remount for now.
4934 */
4935 if (es->s_last_orphan) {
b31e1552 4936 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4937 "remount RDWR because of unprocessed "
4938 "orphan inode list. Please "
b31e1552 4939 "umount/remount instead");
ead6596b
ES		 4940 err = -EINVAL;
4941 goto restore_opts;
4942 }
4943
ac27a0ec
DK		 4944 /*
4945 * Mounting a RDONLY partition read-write, so reread
4946 * and store the current valid flag. (It may have
4947 * been changed by e2fsck since we originally mounted
4948 * the partition.)
4949 */
0390131b
FM		 4950 if (sbi->s_journal)
4951 ext4_clear_journal_err(sb, es);
ac27a0ec 4952 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4953 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4954 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4955 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4956 EXT4_FEATURE_INCOMPAT_MMP))
4957 if (ext4_multi_mount_protect(sb,
4958 le64_to_cpu(es->s_mmp_block))) {
4959 err = -EROFS;
4960 goto restore_opts;
4961 }
c79d967d 4962 enable_quota = 1;
ac27a0ec
DK		 4963 }
4964 }
bfff6873
LC		 4965
4966 /*
4967 * Reinitialize lazy itable initialization thread based on
4968 * current settings
4969 */
4970 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4971 ext4_unregister_li_request(sb);
4972 else {
4973 ext4_group_t first_not_zeroed;
4974 first_not_zeroed = ext4_has_uninit_itable(sb);
4975 ext4_register_li_request(sb, first_not_zeroed);
4976 }
4977
6fd058f7 4978 ext4_setup_system_zone(sb);
d096ad0f 4979 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4980 ext4_commit_super(sb, 1);
0390131b 4981
ac27a0ec
DK		 4982#ifdef CONFIG_QUOTA
4983 /* Release old quota file names */
a2d4a646 4984 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 4985 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4986 if (enable_quota) {
4987 if (sb_any_quota_suspended(sb))
4988 dquot_resume(sb, -1);
4989 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4990 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4991 err = ext4_enable_quotas(sb);
07724f98 4992 if (err)
7c319d32 4993 goto restore_opts;
7c319d32
AK		 4994 }
4995 }
ac27a0ec 4996#endif
d4c402d9
CW		 4997
4998 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4999 kfree(orig_data);
ac27a0ec 5000 return 0;
0b8e58a1 5001
ac27a0ec
DK		 5002restore_opts:
5003 sb->s_flags = old_sb_flags;
5004 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5005 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5006 sbi->s_resuid = old_opts.s_resuid;
5007 sbi->s_resgid = old_opts.s_resgid;
5008 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5009 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5010 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5011#ifdef CONFIG_QUOTA
5012 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5013 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5014 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5015 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5016 }
5017#endif
d4c402d9 5018 kfree(orig_data);
ac27a0ec
DK		 5019 return err;
5020}
5021
2b2d6d01 5022static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5023{
5024 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5025 struct ext4_sb_info *sbi = EXT4_SB(sb);
5026 struct ext4_super_block *es = sbi->s_es;
27dd4385 5027 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5028 u64 fsid;
d02a9391 5029 s64 bfree;
27dd4385 5030 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5031
952fc18e
TT		 5032 if (!test_opt(sb, MINIX_DF))
5033 overhead = sbi->s_overhead;
ac27a0ec 5034
617ba13b 5035 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5036 buf->f_bsize = sb->s_blocksize;
b72f78cb 5037 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5038 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5039 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5040 /* prevent underflow in case that few free space is available */
57042651 5041 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5042 buf->f_bavail = buf->f_bfree -
5043 (ext4_r_blocks_count(es) + resv_blocks);
5044 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5045 buf->f_bavail = 0;
5046 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5047 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5048 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5049 fsid = le64_to_cpup((void *)es->s_uuid) ^
5050 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5051 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5052 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5053
ac27a0ec
DK		 5054 return 0;
5055}
5056
0b8e58a1
AD		 5057/* Helper function for writing quotas on sync - we need to start transaction
5058 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 5059 * Process 1 Process 2
617ba13b 5060 * ext4_create() quota_sync()
a269eb18 5061 * jbd2_journal_start() write_dquot()
871a2931 5062 * dquot_initialize() down(dqio_mutex)
dab291af 5063 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 5064 *
5065 */
5066
5067#ifdef CONFIG_QUOTA
5068
5069static inline struct inode *dquot_to_inode(struct dquot *dquot)
5070{
4c376dca 5071 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5072}
5073
617ba13b 5074static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5075{
5076 int ret, err;
5077 handle_t *handle;
5078 struct inode *inode;
5079
5080 inode = dquot_to_inode(dquot);
9924a92a 5081 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5082 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5083 if (IS_ERR(handle))
5084 return PTR_ERR(handle);
5085 ret = dquot_commit(dquot);
617ba13b 5086 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5087 if (!ret)
5088 ret = err;
5089 return ret;
5090}
5091
617ba13b 5092static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5093{
5094 int ret, err;
5095 handle_t *handle;
5096
9924a92a 5097 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5098 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5099 if (IS_ERR(handle))
5100 return PTR_ERR(handle);
5101 ret = dquot_acquire(dquot);
617ba13b 5102 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5103 if (!ret)
5104 ret = err;
5105 return ret;
5106}
5107
617ba13b 5108static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5109{
5110 int ret, err;
5111 handle_t *handle;
5112
9924a92a 5113 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5114 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5115 if (IS_ERR(handle)) {
5116 /* Release dquot anyway to avoid endless cycle in dqput() */
5117 dquot_release(dquot);
ac27a0ec 5118 return PTR_ERR(handle);
9c3013e9 5119 }
ac27a0ec 5120 ret = dquot_release(dquot);
617ba13b 5121 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5122 if (!ret)
5123 ret = err;
5124 return ret;
5125}
5126
617ba13b 5127static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5128{
262b4662
JK		 5129 struct super_block *sb = dquot->dq_sb;
5130 struct ext4_sb_info *sbi = EXT4_SB(sb);
5131
2c8be6b2 5132 /* Are we journaling quotas? */
262b4662
JK		 5133 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5134 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5135 dquot_mark_dquot_dirty(dquot);
617ba13b 5136 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5137 } else {
5138 return dquot_mark_dquot_dirty(dquot);
5139 }
5140}
5141
617ba13b 5142static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5143{
5144 int ret, err;
5145 handle_t *handle;
5146
5147 /* Data block + inode block */
9924a92a 5148 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5149 if (IS_ERR(handle))
5150 return PTR_ERR(handle);
5151 ret = dquot_commit_info(sb, type);
617ba13b 5152 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5153 if (!ret)
5154 ret = err;
5155 return ret;
5156}
5157
5158/*
5159 * Turn on quotas during mount time - we need to find
5160 * the quota file and such...
5161 */
617ba13b 5162static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5163{
287a8095
CH		 5164 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5165 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5166}
5167
5168/*
5169 * Standard function to be called on quota_on
5170 */
617ba13b 5171static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5172 struct path *path)
ac27a0ec
DK		 5173{
5174 int err;
ac27a0ec
DK		 5175
5176 if (!test_opt(sb, QUOTA))
5177 return -EINVAL;
0623543b 5178
ac27a0ec 5179 /* Quotafile not on the same filesystem? */
d8c9584e 5180 if (path->dentry->d_sb != sb)
ac27a0ec 5181 return -EXDEV;
0623543b
JK		 5182 /* Journaling quota? */
5183 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5184 /* Quotafile not in fs root? */
f00c9e44 5185 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5186 ext4_msg(sb, KERN_WARNING,
5187 "Quota file not on filesystem root. "
5188 "Journaled quota will not work");
2b2d6d01 5189 }
0623543b
JK		 5190
5191 /*
5192 * When we journal data on quota file, we have to flush journal to see
5193 * all updates to the file when we bypass pagecache...
5194 */
0390131b 5195 if (EXT4_SB(sb)->s_journal &&
f00c9e44 5196 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 5197 /*
5198 * We don't need to lock updates but journal_flush() could
5199 * otherwise be livelocked...
5200 */
5201 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5202 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5203 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5204 if (err)
7ffe1ea8 5205 return err;
0623543b
JK		 5206 }
5207
f00c9e44 5208 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 5209}
5210
7c319d32
AK		 5211static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5212 unsigned int flags)
5213{
5214 int err;
5215 struct inode *qf_inode;
a2d4a646 5216 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5217 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5218 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5219 };
5220
5221 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5222
5223 if (!qf_inums[type])
5224 return -EPERM;
5225
5226 qf_inode = ext4_iget(sb, qf_inums[type]);
5227 if (IS_ERR(qf_inode)) {
5228 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5229 return PTR_ERR(qf_inode);
5230 }
5231
bcb13850
JK		 5232 /* Don't account quota for quota files to avoid recursion */
5233 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5234 err = dquot_enable(qf_inode, type, format_id, flags);
5235 iput(qf_inode);
5236
5237 return err;
5238}
5239
5240/* Enable usage tracking for all quota types. */
5241static int ext4_enable_quotas(struct super_block *sb)
5242{
5243 int type, err = 0;
a2d4a646 5244 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5245 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5246 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5247 };
5248
5249 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5250 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5251 if (qf_inums[type]) {
5252 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5253 DQUOT_USAGE_ENABLED);
5254 if (err) {
5255 ext4_warning(sb,
72ba7450
TT		 5256 "Failed to enable quota tracking "
5257 "(type=%d, err=%d). Please run "
5258 "e2fsck to fix.", type, err);
7c319d32
AK		 5259 return err;
5260 }
5261 }
5262 }
5263 return 0;
5264}
5265
5266/*
5267 * quota_on function that is used when QUOTA feature is set.
5268 */
5269static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5270 int format_id)
5271{
5272 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5273 return -EINVAL;
5274
5275 /*
5276 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5277 */
5278 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5279}
5280
ca0e05e4
DM		 5281static int ext4_quota_off(struct super_block *sb, int type)
5282{
21f97697
JK		 5283 struct inode *inode = sb_dqopt(sb)->files[type];
5284 handle_t *handle;
5285
87009d86
DM		 5286 /* Force all delayed allocation blocks to be allocated.
5287 * Caller already holds s_umount sem */
5288 if (test_opt(sb, DELALLOC))
ca0e05e4 5289 sync_filesystem(sb);
ca0e05e4 5290
0b268590
AG		 5291 if (!inode)
5292 goto out;
5293
21f97697
JK		 5294 /* Update modification times of quota files when userspace can
5295 * start looking at them */
9924a92a 5296 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5297 if (IS_ERR(handle))
5298 goto out;
5299 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5300 ext4_mark_inode_dirty(handle, inode);
5301 ext4_journal_stop(handle);
5302
5303out:
ca0e05e4
DM		 5304 return dquot_quota_off(sb, type);
5305}
5306
7c319d32
AK		 5307/*
5308 * quota_off function that is used when QUOTA feature is set.
5309 */
5310static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5311{
5312 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5313 return -EINVAL;
5314
5315 /* Disable only the limits. */
5316 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5317}
5318
ac27a0ec
DK		 5319/* Read data from quotafile - avoid pagecache and such because we cannot afford
5320 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5321 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5322 * we don't have to be afraid of races */
617ba13b 5323static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5324 size_t len, loff_t off)
5325{
5326 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5327 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5328 int offset = off & (sb->s_blocksize - 1);
5329 int tocopy;
5330 size_t toread;
5331 struct buffer_head *bh;
5332 loff_t i_size = i_size_read(inode);
5333
5334 if (off > i_size)
5335 return 0;
5336 if (off+len > i_size)
5337 len = i_size-off;
5338 toread = len;
5339 while (toread > 0) {
5340 tocopy = sb->s_blocksize - offset < toread ?
5341 sb->s_blocksize - offset : toread;
1c215028
TT		 5342 bh = ext4_bread(NULL, inode, blk, 0);
5343 if (IS_ERR(bh))
5344 return PTR_ERR(bh);
ac27a0ec
DK		 5345 if (!bh) /* A hole? */
5346 memset(data, 0, tocopy);
5347 else
5348 memcpy(data, bh->b_data+offset, tocopy);
5349 brelse(bh);
5350 offset = 0;
5351 toread -= tocopy;
5352 data += tocopy;
5353 blk++;
5354 }
5355 return len;
5356}
5357
5358/* Write to quotafile (we know the transaction is already started and has
5359 * enough credits) */
617ba13b 5360static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5361 const char *data, size_t len, loff_t off)
5362{
5363 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5364 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5365 int err, offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5366 struct buffer_head *bh;
5367 handle_t *handle = journal_current_handle();
5368
0390131b 5369 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5370 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5371 " cancelled because transaction is not started",
9c3013e9
JK		 5372 (unsigned long long)off, (unsigned long long)len);
5373 return -EIO;
5374 }
67eeb568
DM		 5375 /*
5376 * Since we account only one data block in transaction credits,
5377 * then it is impossible to cross a block boundary.
5378 */
5379 if (sb->s_blocksize - offset < len) {
5380 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5381 " cancelled because not block aligned",
5382 (unsigned long long)off, (unsigned long long)len);
5383 return -EIO;
5384 }
5385
1c215028
TT		 5386 bh = ext4_bread(handle, inode, blk, 1);
5387 if (IS_ERR(bh))
5388 return PTR_ERR(bh);
67eeb568
DM		 5389 if (!bh)
5390 goto out;
5d601255 5391 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5392 err = ext4_journal_get_write_access(handle, bh);
5393 if (err) {
5394 brelse(bh);
1c215028 5395 return err;
ac27a0ec 5396 }
67eeb568
DM		 5397 lock_buffer(bh);
5398 memcpy(bh->b_data+offset, data, len);
5399 flush_dcache_page(bh->b_page);
5400 unlock_buffer(bh);
62d2b5f2 5401 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5402 brelse(bh);
ac27a0ec 5403out:
67eeb568
DM		 5404 if (inode->i_size < off + len) {
5405 i_size_write(inode, off + len);
617ba13b 5406 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5407 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5408 }
67eeb568 5409 return len;
ac27a0ec
DK		 5410}
5411
5412#endif
5413
152a0836
AV		 5414static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5415 const char *dev_name, void *data)
ac27a0ec 5416{
152a0836 5417 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5418}
5419
37f328eb 5420#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5421static inline void register_as_ext2(void)
5422{
5423 int err = register_filesystem(&ext2_fs_type);
5424 if (err)
5425 printk(KERN_WARNING
5426 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5427}
5428
5429static inline void unregister_as_ext2(void)
5430{
5431 unregister_filesystem(&ext2_fs_type);
5432}
2035e776
TT		 5433
5434static inline int ext2_feature_set_ok(struct super_block *sb)
5435{
5436 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5437 return 0;
5438 if (sb->s_flags & MS_RDONLY)
5439 return 1;
5440 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5441 return 0;
5442 return 1;
5443}
24b58424
TT		 5444#else
5445static inline void register_as_ext2(void) { }
5446static inline void unregister_as_ext2(void) { }
2035e776 5447static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5448#endif
5449
37f328eb 5450#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5451static inline void register_as_ext3(void)
5452{
5453 int err = register_filesystem(&ext3_fs_type);
5454 if (err)
5455 printk(KERN_WARNING
5456 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5457}
5458
5459static inline void unregister_as_ext3(void)
5460{
5461 unregister_filesystem(&ext3_fs_type);
5462}
2035e776
TT		 5463
5464static inline int ext3_feature_set_ok(struct super_block *sb)
5465{
5466 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5467 return 0;
5468 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5469 return 0;
5470 if (sb->s_flags & MS_RDONLY)
5471 return 1;
5472 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5473 return 0;
5474 return 1;
5475}
24b58424
TT		 5476#else
5477static inline void register_as_ext3(void) { }
5478static inline void unregister_as_ext3(void) { }
2035e776 5479static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5480#endif
5481
03010a33
TT		 5482static struct file_system_type ext4_fs_type = {
5483 .owner = THIS_MODULE,
5484 .name = "ext4",
152a0836 5485 .mount = ext4_mount,
03010a33
TT		 5486 .kill_sb = kill_block_super,
5487 .fs_flags = FS_REQUIRES_DEV,
5488};
7f78e035 5489MODULE_ALIAS_FS("ext4");
03010a33 5490
8f021222 5491static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5492{
5493 struct ext4_features *ef;
5494 int ret = -ENOMEM;
5495
5496 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5497 if (!ef)
5498 goto out;
5499
5500 ef->f_kobj.kset = ext4_kset;
5501 init_completion(&ef->f_kobj_unregister);
5502 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5503 "features");
5504 if (ret) {
5505 kfree(ef);
5506 goto out;
5507 }
5508
5509 ext4_feat = ef;
5510 ret = 0;
5511out:
5512 return ret;
5513}
5514
8f021222
LC		 5515static void ext4_exit_feat_adverts(void)
5516{
5517 kobject_put(&ext4_feat->f_kobj);
5518 wait_for_completion(&ext4_feat->f_kobj_unregister);
5519 kfree(ext4_feat);
5520}
5521
e9e3bcec
ES		 5522/* Shared across all ext4 file systems */
5523wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5524struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5525
5dabfc78 5526static int __init ext4_init_fs(void)
ac27a0ec 5527{
e9e3bcec 5528 int i, err;
c9de560d 5529
07c0c5d8
AV		 5530 ext4_li_info = NULL;
5531 mutex_init(&ext4_li_mtx);
5532
9a4c8019 5533 /* Build-time check for flags consistency */
12e9b892 5534 ext4_check_flag_values();
e9e3bcec
ES		 5535
5536 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5537 mutex_init(&ext4__aio_mutex[i]);
5538 init_waitqueue_head(&ext4__ioend_wq[i]);
5539 }
5540
51865fda 5541 err = ext4_init_es();
6fd058f7
TT		 5542 if (err)
5543 return err;
51865fda
ZL		 5544
5545 err = ext4_init_pageio();
5546 if (err)
5547 goto out7;
5548
5dabfc78 5549 err = ext4_init_system_zone();
bd2d0210 5550 if (err)
d44651d0 5551 goto out6;
3197ebdb 5552 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5553 if (!ext4_kset) {
5554 err = -ENOMEM;
dd68314c 5555 goto out5;
0e376b1e 5556 }
d44651d0 5557 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5558
5559 err = ext4_init_feat_adverts();
dd68314c
TT		 5560 if (err)
5561 goto out4;
857ac889 5562
5dabfc78 5563 err = ext4_init_mballoc();
c9de560d
AT		 5564 if (err)
5565 goto out2;
9c191f70
M		 5566 else
5567 ext4_mballoc_ready = 1;
ac27a0ec
DK		 5568 err = init_inodecache();
5569 if (err)
5570 goto out1;
24b58424 5571 register_as_ext3();
2035e776 5572 register_as_ext2();
03010a33 5573 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5574 if (err)
5575 goto out;
bfff6873 5576
ac27a0ec
DK		 5577 return 0;
5578out:
24b58424
TT		 5579 unregister_as_ext2();
5580 unregister_as_ext3();
ac27a0ec
DK		 5581 destroy_inodecache();
5582out1:
9c191f70 5583 ext4_mballoc_ready = 0;
5dabfc78 5584 ext4_exit_mballoc();
9c191f70 5585out2:
8f021222 5586 ext4_exit_feat_adverts();
dd68314c 5587out4:
d44651d0
FJ		 5588 if (ext4_proc_root)
5589 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5590 kset_unregister(ext4_kset);
d44651d0 5591out5:
5dabfc78 5592 ext4_exit_system_zone();
d44651d0 5593out6:
5dabfc78 5594 ext4_exit_pageio();
51865fda
ZL		 5595out7:
5596 ext4_exit_es();
5597
ac27a0ec
DK		 5598 return err;
5599}
5600
5dabfc78 5601static void __exit ext4_exit_fs(void)
ac27a0ec 5602{
bfff6873 5603 ext4_destroy_lazyinit_thread();
24b58424
TT		 5604 unregister_as_ext2();
5605 unregister_as_ext3();
03010a33 5606 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5607 destroy_inodecache();
5dabfc78 5608 ext4_exit_mballoc();
8f021222 5609 ext4_exit_feat_adverts();
9f6200bb 5610 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5611 kset_unregister(ext4_kset);
5dabfc78
TT		 5612 ext4_exit_system_zone();
5613 ext4_exit_pageio();
dd12ed14 5614 ext4_exit_es();
ac27a0ec
DK		 5615}
5616
5617MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5618MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5619MODULE_LICENSE("GPL");
5dabfc78
TT		 5620module_init(ext4_init_fs)
5621module_exit(ext4_exit_fs)
Linux kernel - Deliverables
This page took 1.308179 seconds and 5 git commands to generate.