]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-bionic-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: remove useless condition in if statement.
[mirror_ubuntu-bionic-kernel.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>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
27#include <linux/parser.h>
ac27a0ec 28#include <linux/buffer_head.h>
a5694255 29#include <linux/exportfs.h>
ac27a0ec
DK		 30#include <linux/vfs.h>
31#include <linux/random.h>
32#include <linux/mount.h>
33#include <linux/namei.h>
34#include <linux/quotaops.h>
35#include <linux/seq_file.h>
9f6200bb 36#include <linux/proc_fs.h>
3197ebdb 37#include <linux/ctype.h>
1330593e 38#include <linux/log2.h>
717d50e4 39#include <linux/crc16.h>
7abc52c2 40#include <linux/cleancache.h>
ac27a0ec
DK		 41#include <asm/uaccess.h>
42
bfff6873
LC		 43#include <linux/kthread.h>
44#include <linux/freezer.h>
45
3dcf5451 46#include "ext4.h"
4a092d73 47#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 48#include "ext4_jbd2.h"
ac27a0ec
DK		 49#include "xattr.h"
50#include "acl.h"
3661d286 51#include "mballoc.h"
ac27a0ec 52
9bffad1e
TT		 53#define CREATE_TRACE_POINTS
54#include <trace/events/ext4.h>
55
1f109d5a 56static struct proc_dir_entry *ext4_proc_root;
3197ebdb 57static struct kset *ext4_kset;
0b75a840
LC		 58static struct ext4_lazy_init *ext4_li_info;
59static struct mutex ext4_li_mtx;
60static struct ext4_features *ext4_feat;
9c191f70 61static int ext4_mballoc_ready;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
2adf6da8 65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 66static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 67static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
617ba13b 71static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 72static int ext4_remount(struct super_block *sb, int *flags, char *data);
73static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 74static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 75static int ext4_freeze(struct super_block *sb);
152a0836
AV		 76static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
2035e776
TT		 78static inline int ext2_feature_set_ok(struct super_block *sb);
79static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 80static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 81static void ext4_destroy_lazyinit_thread(void);
82static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 83static void ext4_clear_request_list(void);
27dd4385 84static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
ac27a0ec 85
2035e776
TT		 86#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87static struct file_system_type ext2_fs_type = {
88 .owner = THIS_MODULE,
89 .name = "ext2",
90 .mount = ext4_mount,
91 .kill_sb = kill_block_super,
92 .fs_flags = FS_REQUIRES_DEV,
93};
7f78e035 94MODULE_ALIAS_FS("ext2");
fa7614dd 95MODULE_ALIAS("ext2");
2035e776
TT		 96#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97#else
98#define IS_EXT2_SB(sb) (0)
99#endif
100
101
ba69f9ab
JK		 102#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
105 .name = "ext3",
152a0836 106 .mount = ext4_mount,
ba69f9ab
JK		 107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
109};
7f78e035 110MODULE_ALIAS_FS("ext3");
fa7614dd 111MODULE_ALIAS("ext3");
ba69f9ab
JK		 112#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
113#else
114#define IS_EXT3_SB(sb) (0)
115#endif
bd81d8ee 116
d25425f8
DW		 117static int ext4_verify_csum_type(struct super_block *sb,
118 struct ext4_super_block *es)
119{
120 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
121 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
122 return 1;
123
124 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
125}
126
a9c47317
DW		 127static __le32 ext4_superblock_csum(struct super_block *sb,
128 struct ext4_super_block *es)
129{
130 struct ext4_sb_info *sbi = EXT4_SB(sb);
131 int offset = offsetof(struct ext4_super_block, s_checksum);
132 __u32 csum;
133
134 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
135
136 return cpu_to_le32(csum);
137}
138
c197855e
SH		 139static int ext4_superblock_csum_verify(struct super_block *sb,
140 struct ext4_super_block *es)
a9c47317 141{
9aa5d32b 142 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 143 return 1;
144
145 return es->s_checksum == ext4_superblock_csum(sb, es);
146}
147
06db49e6 148void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 149{
06db49e6
TT		 150 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
151
9aa5d32b 152 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 153 return;
154
155 es->s_checksum = ext4_superblock_csum(sb, es);
156}
157
9933fc0a
TT		 158void *ext4_kvmalloc(size_t size, gfp_t flags)
159{
160 void *ret;
161
8be04b93 162 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 163 if (!ret)
164 ret = __vmalloc(size, flags, PAGE_KERNEL);
165 return ret;
166}
167
168void *ext4_kvzalloc(size_t size, gfp_t flags)
169{
170 void *ret;
171
8be04b93 172 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 173 if (!ret)
174 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
175 return ret;
176}
177
8fadc143
AR		 178ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
179 struct ext4_group_desc *bg)
bd81d8ee 180{
3a14589c 181 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 182 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 183 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 184}
185
8fadc143
AR		 186ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
187 struct ext4_group_desc *bg)
bd81d8ee 188{
5272f837 189 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 191 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 192}
193
8fadc143
AR		 194ext4_fsblk_t ext4_inode_table(struct super_block *sb,
195 struct ext4_group_desc *bg)
bd81d8ee 196{
5272f837 197 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 199 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 200}
201
021b65bb
TT		 202__u32 ext4_free_group_clusters(struct super_block *sb,
203 struct ext4_group_desc *bg)
560671a0
AK		 204{
205 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 207 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 208}
209
210__u32 ext4_free_inodes_count(struct super_block *sb,
211 struct ext4_group_desc *bg)
212{
213 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 215 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 216}
217
218__u32 ext4_used_dirs_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
220{
221 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 223 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 224}
225
226__u32 ext4_itable_unused_count(struct super_block *sb,
227 struct ext4_group_desc *bg)
228{
229 return le16_to_cpu(bg->bg_itable_unused_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 231 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 232}
233
8fadc143
AR		 234void ext4_block_bitmap_set(struct super_block *sb,
235 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 236{
3a14589c 237 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 238 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
239 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 240}
241
8fadc143
AR		 242void ext4_inode_bitmap_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 244{
5272f837 245 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 248}
249
8fadc143
AR		 250void ext4_inode_table_set(struct super_block *sb,
251 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 252{
5272f837 253 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 256}
257
021b65bb
TT		 258void ext4_free_group_clusters_set(struct super_block *sb,
259 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 260{
261 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
264}
265
266void ext4_free_inodes_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
268{
269 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
272}
273
274void ext4_used_dirs_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
276{
277 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
280}
281
282void ext4_itable_unused_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
284{
285 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
288}
289
d3d1faf6 290
1c13d5c0
TT		 291static void __save_error_info(struct super_block *sb, const char *func,
292 unsigned int line)
293{
294 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
295
296 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
297 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
298 es->s_last_error_time = cpu_to_le32(get_seconds());
299 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
300 es->s_last_error_line = cpu_to_le32(line);
301 if (!es->s_first_error_time) {
302 es->s_first_error_time = es->s_last_error_time;
303 strncpy(es->s_first_error_func, func,
304 sizeof(es->s_first_error_func));
305 es->s_first_error_line = cpu_to_le32(line);
306 es->s_first_error_ino = es->s_last_error_ino;
307 es->s_first_error_block = es->s_last_error_block;
308 }
66e61a9e
TT		 309 /*
310 * Start the daily error reporting function if it hasn't been
311 * started already
312 */
313 if (!es->s_error_count)
314 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 315 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 316}
317
318static void save_error_info(struct super_block *sb, const char *func,
319 unsigned int line)
320{
321 __save_error_info(sb, func, line);
322 ext4_commit_super(sb, 1);
323}
324
7c2e7087
TT		 325/*
326 * The del_gendisk() function uninitializes the disk-specific data
327 * structures, including the bdi structure, without telling anyone
328 * else. Once this happens, any attempt to call mark_buffer_dirty()
329 * (for example, by ext4_commit_super), will cause a kernel OOPS.
330 * This is a kludge to prevent these oops until we can put in a proper
331 * hook in del_gendisk() to inform the VFS and file system layers.
332 */
333static int block_device_ejected(struct super_block *sb)
334{
335 struct inode *bd_inode = sb->s_bdev->bd_inode;
de1414a6 336 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
7c2e7087
TT		 337
338 return bdi->dev == NULL;
339}
340
18aadd47
BJ		 341static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
342{
343 struct super_block *sb = journal->j_private;
344 struct ext4_sb_info *sbi = EXT4_SB(sb);
345 int error = is_journal_aborted(journal);
5d3ee208 346 struct ext4_journal_cb_entry *jce;
18aadd47 347
5d3ee208 348 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 349 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 350 while (!list_empty(&txn->t_private_list)) {
351 jce = list_entry(txn->t_private_list.next,
352 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 353 list_del_init(&jce->jce_list);
354 spin_unlock(&sbi->s_md_lock);
355 jce->jce_func(sb, jce, error);
356 spin_lock(&sbi->s_md_lock);
357 }
358 spin_unlock(&sbi->s_md_lock);
359}
1c13d5c0 360
ac27a0ec
DK		 361/* Deal with the reporting of failure conditions on a filesystem such as
362 * inconsistencies detected or read IO failures.
363 *
364 * On ext2, we can store the error state of the filesystem in the
617ba13b 365 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 366 * write ordering constraints on the superblock which prevent us from
367 * writing it out straight away; and given that the journal is about to
368 * be aborted, we can't rely on the current, or future, transactions to
369 * write out the superblock safely.
370 *
dab291af 371 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 372 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 373 * that error until we've noted it down and cleared it.
374 */
375
617ba13b 376static void ext4_handle_error(struct super_block *sb)
ac27a0ec 377{
ac27a0ec
DK		 378 if (sb->s_flags & MS_RDONLY)
379 return;
380
2b2d6d01 381 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 382 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 383
4ab2f15b 384 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 385 if (journal)
dab291af 386 jbd2_journal_abort(journal, -EIO);
ac27a0ec 387 }
2b2d6d01 388 if (test_opt(sb, ERRORS_RO)) {
b31e1552 389 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 390 /*
391 * Make sure updated value of ->s_mount_flags will be visible
392 * before ->s_flags update
393 */
394 smp_wmb();
ac27a0ec
DK		 395 sb->s_flags |= MS_RDONLY;
396 }
ac27a0ec 397 if (test_opt(sb, ERRORS_PANIC))
617ba13b 398 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 399 sb->s_id);
400}
401
efbed4dc
TT		 402#define ext4_error_ratelimit(sb) \
403 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
404 "EXT4-fs error")
405
12062ddd 406void __ext4_error(struct super_block *sb, const char *function,
c398eda0 407 unsigned int line, const char *fmt, ...)
ac27a0ec 408{
0ff2ea7d 409 struct va_format vaf;
ac27a0ec
DK		 410 va_list args;
411
efbed4dc
TT		 412 if (ext4_error_ratelimit(sb)) {
413 va_start(args, fmt);
414 vaf.fmt = fmt;
415 vaf.va = &args;
416 printk(KERN_CRIT
417 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
418 sb->s_id, function, line, current->comm, &vaf);
419 va_end(args);
420 }
f3fc0210 421 save_error_info(sb, function, line);
617ba13b 422 ext4_handle_error(sb);
ac27a0ec
DK		 423}
424
e7c96e8e
JP		 425void __ext4_error_inode(struct inode *inode, const char *function,
426 unsigned int line, ext4_fsblk_t block,
427 const char *fmt, ...)
273df556
FM		 428{
429 va_list args;
f7c21177 430 struct va_format vaf;
1c13d5c0 431 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 432
1c13d5c0
TT		 433 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
434 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 435 if (ext4_error_ratelimit(inode->i_sb)) {
436 va_start(args, fmt);
437 vaf.fmt = fmt;
438 vaf.va = &args;
439 if (block)
440 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
441 "inode #%lu: block %llu: comm %s: %pV\n",
442 inode->i_sb->s_id, function, line, inode->i_ino,
443 block, current->comm, &vaf);
444 else
445 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
446 "inode #%lu: comm %s: %pV\n",
447 inode->i_sb->s_id, function, line, inode->i_ino,
448 current->comm, &vaf);
449 va_end(args);
450 }
1c13d5c0 451 save_error_info(inode->i_sb, function, line);
273df556
FM		 452 ext4_handle_error(inode->i_sb);
453}
454
e7c96e8e
JP		 455void __ext4_error_file(struct file *file, const char *function,
456 unsigned int line, ext4_fsblk_t block,
457 const char *fmt, ...)
273df556
FM		 458{
459 va_list args;
f7c21177 460 struct va_format vaf;
1c13d5c0 461 struct ext4_super_block *es;
496ad9aa 462 struct inode *inode = file_inode(file);
273df556
FM		 463 char pathname[80], *path;
464
1c13d5c0
TT		 465 es = EXT4_SB(inode->i_sb)->s_es;
466 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc
TT		 467 if (ext4_error_ratelimit(inode->i_sb)) {
468 path = d_path(&(file->f_path), pathname, sizeof(pathname));
469 if (IS_ERR(path))
470 path = "(unknown)";
471 va_start(args, fmt);
472 vaf.fmt = fmt;
473 vaf.va = &args;
474 if (block)
475 printk(KERN_CRIT
476 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
477 "block %llu: comm %s: path %s: %pV\n",
478 inode->i_sb->s_id, function, line, inode->i_ino,
479 block, current->comm, path, &vaf);
480 else
481 printk(KERN_CRIT
482 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
483 "comm %s: path %s: %pV\n",
484 inode->i_sb->s_id, function, line, inode->i_ino,
485 current->comm, path, &vaf);
486 va_end(args);
487 }
1c13d5c0 488 save_error_info(inode->i_sb, function, line);
273df556
FM		 489 ext4_handle_error(inode->i_sb);
490}
491
722887dd
TT		 492const char *ext4_decode_error(struct super_block *sb, int errno,
493 char nbuf[16])
ac27a0ec
DK		 494{
495 char *errstr = NULL;
496
497 switch (errno) {
498 case -EIO:
499 errstr = "IO failure";
500 break;
501 case -ENOMEM:
502 errstr = "Out of memory";
503 break;
504 case -EROFS:
78f1ddbb
TT		 505 if (!sb || (EXT4_SB(sb)->s_journal &&
506 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 507 errstr = "Journal has aborted";
508 else
509 errstr = "Readonly filesystem";
510 break;
511 default:
512 /* If the caller passed in an extra buffer for unknown
513 * errors, textualise them now. Else we just return
514 * NULL. */
515 if (nbuf) {
516 /* Check for truncated error codes... */
517 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
518 errstr = nbuf;
519 }
520 break;
521 }
522
523 return errstr;
524}
525
617ba13b 526/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 527 * automatically and invokes the appropriate error response. */
528
c398eda0
TT		 529void __ext4_std_error(struct super_block *sb, const char *function,
530 unsigned int line, int errno)
ac27a0ec
DK		 531{
532 char nbuf[16];
533 const char *errstr;
534
535 /* Special case: if the error is EROFS, and we're not already
536 * inside a transaction, then there's really no point in logging
537 * an error. */
538 if (errno == -EROFS && journal_current_handle() == NULL &&
539 (sb->s_flags & MS_RDONLY))
540 return;
541
efbed4dc
TT		 542 if (ext4_error_ratelimit(sb)) {
543 errstr = ext4_decode_error(sb, errno, nbuf);
544 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
545 sb->s_id, function, line, errstr);
546 }
ac27a0ec 547
efbed4dc 548 save_error_info(sb, function, line);
617ba13b 549 ext4_handle_error(sb);
ac27a0ec
DK		 550}
551
552/*
617ba13b 553 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 554 * abort function may be used to deal with unrecoverable failures such
555 * as journal IO errors or ENOMEM at a critical moment in log management.
556 *
557 * We unconditionally force the filesystem into an ABORT|READONLY state,
558 * unless the error response on the fs has been set to panic in which
559 * case we take the easy way out and panic immediately.
560 */
561
c67d859e 562void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 563 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 564{
565 va_list args;
566
1c13d5c0 567 save_error_info(sb, function, line);
ac27a0ec 568 va_start(args, fmt);
c398eda0
TT		 569 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
570 function, line);
ac27a0ec
DK		 571 vprintk(fmt, args);
572 printk("\n");
573 va_end(args);
574
1c13d5c0
TT		 575 if ((sb->s_flags & MS_RDONLY) == 0) {
576 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 577 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 578 /*
579 * Make sure updated value of ->s_mount_flags will be visible
580 * before ->s_flags update
581 */
582 smp_wmb();
583 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 584 if (EXT4_SB(sb)->s_journal)
585 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
586 save_error_info(sb, function, line);
587 }
ac27a0ec 588 if (test_opt(sb, ERRORS_PANIC))
617ba13b 589 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 590}
591
e7c96e8e
JP		 592void __ext4_msg(struct super_block *sb,
593 const char *prefix, const char *fmt, ...)
b31e1552 594{
0ff2ea7d 595 struct va_format vaf;
b31e1552
ES		 596 va_list args;
597
efbed4dc
TT		 598 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
599 return;
600
b31e1552 601 va_start(args, fmt);
0ff2ea7d
JP		 602 vaf.fmt = fmt;
603 vaf.va = &args;
604 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 605 va_end(args);
606}
607
12062ddd 608void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 609 unsigned int line, const char *fmt, ...)
ac27a0ec 610{
0ff2ea7d 611 struct va_format vaf;
ac27a0ec
DK		 612 va_list args;
613
efbed4dc
TT		 614 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
615 "EXT4-fs warning"))
616 return;
617
ac27a0ec 618 va_start(args, fmt);
0ff2ea7d
JP		 619 vaf.fmt = fmt;
620 vaf.va = &args;
621 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
622 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 623 va_end(args);
624}
625
e29136f8
TT		 626void __ext4_grp_locked_error(const char *function, unsigned int line,
627 struct super_block *sb, ext4_group_t grp,
628 unsigned long ino, ext4_fsblk_t block,
629 const char *fmt, ...)
5d1b1b3f
AK		 630__releases(bitlock)
631__acquires(bitlock)
632{
0ff2ea7d 633 struct va_format vaf;
5d1b1b3f
AK		 634 va_list args;
635 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
636
1c13d5c0
TT		 637 es->s_last_error_ino = cpu_to_le32(ino);
638 es->s_last_error_block = cpu_to_le64(block);
639 __save_error_info(sb, function, line);
0ff2ea7d 640
efbed4dc
TT		 641 if (ext4_error_ratelimit(sb)) {
642 va_start(args, fmt);
643 vaf.fmt = fmt;
644 vaf.va = &args;
645 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
646 sb->s_id, function, line, grp);
647 if (ino)
648 printk(KERN_CONT "inode %lu: ", ino);
649 if (block)
650 printk(KERN_CONT "block %llu:",
651 (unsigned long long) block);
652 printk(KERN_CONT "%pV\n", &vaf);
653 va_end(args);
654 }
5d1b1b3f
AK		 655
656 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 657 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 658 return;
659 }
1c13d5c0 660
5d1b1b3f
AK		 661 ext4_unlock_group(sb, grp);
662 ext4_handle_error(sb);
663 /*
664 * We only get here in the ERRORS_RO case; relocking the group
665 * may be dangerous, but nothing bad will happen since the
666 * filesystem will have already been marked read/only and the
667 * journal has been aborted. We return 1 as a hint to callers
668 * who might what to use the return value from
25985edc 669 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 670 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
671 * aggressively from the ext4 function in question, with a
672 * more appropriate error code.
673 */
674 ext4_lock_group(sb, grp);
675 return;
676}
677
617ba13b 678void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 679{
617ba13b 680 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 681
617ba13b 682 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 683 return;
684
12062ddd 685 ext4_warning(sb,
ac27a0ec
DK		 686 "updating to rev %d because of new feature flag, "
687 "running e2fsck is recommended",
617ba13b 688 EXT4_DYNAMIC_REV);
ac27a0ec 689
617ba13b
MC		 690 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
691 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
692 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 693 /* leave es->s_feature_*compat flags alone */
694 /* es->s_uuid will be set by e2fsck if empty */
695
696 /*
697 * The rest of the superblock fields should be zero, and if not it
698 * means they are likely already in use, so leave them alone. We
699 * can leave it up to e2fsck to clean up any inconsistencies there.
700 */
701}
702
703/*
704 * Open the external journal device
705 */
b31e1552 706static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 707{
708 struct block_device *bdev;
709 char b[BDEVNAME_SIZE];
710
d4d77629 711 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 712 if (IS_ERR(bdev))
713 goto fail;
714 return bdev;
715
716fail:
b31e1552 717 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 718 __bdevname(dev, b), PTR_ERR(bdev));
719 return NULL;
720}
721
722/*
723 * Release the journal device
724 */
4385bab1 725static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 726{
4385bab1 727 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 728}
729
4385bab1 730static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 731{
732 struct block_device *bdev;
ac27a0ec
DK		 733 bdev = sbi->journal_bdev;
734 if (bdev) {
4385bab1 735 ext4_blkdev_put(bdev);
ac27a0ec
DK		 736 sbi->journal_bdev = NULL;
737 }
ac27a0ec
DK		 738}
739
740static inline struct inode *orphan_list_entry(struct list_head *l)
741{
617ba13b 742 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 743}
744
617ba13b 745static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 746{
747 struct list_head *l;
748
b31e1552
ES		 749 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
750 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 751
752 printk(KERN_ERR "sb_info orphan list:\n");
753 list_for_each(l, &sbi->s_orphan) {
754 struct inode *inode = orphan_list_entry(l);
755 printk(KERN_ERR " "
756 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
757 inode->i_sb->s_id, inode->i_ino, inode,
758 inode->i_mode, inode->i_nlink,
759 NEXT_ORPHAN(inode));
760 }
761}
762
2b2d6d01 763static void ext4_put_super(struct super_block *sb)
ac27a0ec 764{
617ba13b
MC		 765 struct ext4_sb_info *sbi = EXT4_SB(sb);
766 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 767 int i, err;
ac27a0ec 768
857ac889 769 ext4_unregister_li_request(sb);
e0ccfd95
CH		 770 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
771
2e8fa54e 772 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 773 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 774
0390131b
FM		 775 if (sbi->s_journal) {
776 err = jbd2_journal_destroy(sbi->s_journal);
777 sbi->s_journal = NULL;
778 if (err < 0)
c67d859e 779 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 780 }
d4edac31 781
d3922a77 782 ext4_es_unregister_shrinker(sbi);
9105bb14 783 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 784 ext4_release_system_zone(sb);
785 ext4_mb_release(sb);
786 ext4_ext_release(sb);
787 ext4_xattr_put_super(sb);
788
ac27a0ec 789 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 790 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 791 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 792 }
58c5873a 793 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 794 ext4_commit_super(sb, 1);
795
240799cd 796 if (sbi->s_proc) {
66acdcf4 797 remove_proc_entry("options", sbi->s_proc);
9f6200bb 798 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 799 }
3197ebdb 800 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 801
802 for (i = 0; i < sbi->s_gdb_count; i++)
803 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 804 kvfree(sbi->s_group_desc);
805 kvfree(sbi->s_flex_groups);
57042651 806 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 807 percpu_counter_destroy(&sbi->s_freeinodes_counter);
808 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 809 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 810 brelse(sbi->s_sbh);
811#ifdef CONFIG_QUOTA
a2d4a646 812 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 813 kfree(sbi->s_qf_names[i]);
814#endif
815
816 /* Debugging code just in case the in-memory inode orphan list
817 * isn't empty. The on-disk one can be non-empty if we've
818 * detected an error and taken the fs readonly, but the
819 * in-memory list had better be clean by this point. */
820 if (!list_empty(&sbi->s_orphan))
821 dump_orphan_list(sb, sbi);
822 J_ASSERT(list_empty(&sbi->s_orphan));
823
f98393a6 824 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 825 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
826 /*
827 * Invalidate the journal device's buffers. We don't want them
828 * floating about in memory - the physical journal device may
829 * hotswapped, and it breaks the `ro-after' testing code.
830 */
831 sync_blockdev(sbi->journal_bdev);
f98393a6 832 invalidate_bdev(sbi->journal_bdev);
617ba13b 833 ext4_blkdev_remove(sbi);
ac27a0ec 834 }
9c191f70
M		 835 if (sbi->s_mb_cache) {
836 ext4_xattr_destroy_cache(sbi->s_mb_cache);
837 sbi->s_mb_cache = NULL;
838 }
c5e06d10
JL		 839 if (sbi->s_mmp_tsk)
840 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 841 sb->s_fs_info = NULL;
3197ebdb
TT		 842 /*
843 * Now that we are completely done shutting down the
844 * superblock, we need to actually destroy the kobject.
845 */
3197ebdb
TT		 846 kobject_put(&sbi->s_kobj);
847 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 848 if (sbi->s_chksum_driver)
849 crypto_free_shash(sbi->s_chksum_driver);
705895b6 850 kfree(sbi->s_blockgroup_lock);
ac27a0ec 851 kfree(sbi);
ac27a0ec
DK		 852}
853
e18b890b 854static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 855
856/*
857 * Called inside transaction, so use GFP_NOFS
858 */
617ba13b 859static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 860{
617ba13b 861 struct ext4_inode_info *ei;
ac27a0ec 862
e6b4f8da 863 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 864 if (!ei)
865 return NULL;
0b8e58a1 866
ac27a0ec 867 ei->vfs_inode.i_version = 1;
202ee5df 868 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 869 INIT_LIST_HEAD(&ei->i_prealloc_list);
870 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 871 ext4_es_init_tree(&ei->i_es_tree);
872 rwlock_init(&ei->i_es_lock);
edaa53ca 873 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 874 ei->i_es_all_nr = 0;
edaa53ca 875 ei->i_es_shk_nr = 0;
dd475925 876 ei->i_es_shrink_lblk = 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;
96c7e0d9 885 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 886#endif
8aefcd55 887 ei->jinode = NULL;
2e8fa54e 888 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 889 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 890 ei->i_sync_tid = 0;
891 ei->i_datasync_tid = 0;
f7ad6d2e 892 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 893 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 894 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
0b8e58a1 895
ac27a0ec
DK		 896 return &ei->vfs_inode;
897}
898
7ff9c073
TT		 899static int ext4_drop_inode(struct inode *inode)
900{
901 int drop = generic_drop_inode(inode);
902
903 trace_ext4_drop_inode(inode, drop);
904 return drop;
905}
906
fa0d7e3d
NP		 907static void ext4_i_callback(struct rcu_head *head)
908{
909 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 910 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
911}
912
617ba13b 913static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 914{
9f7dd93d 915 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 916 ext4_msg(inode->i_sb, KERN_ERR,
917 "Inode %lu (%p): orphan list check failed!",
918 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 919 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
920 EXT4_I(inode), sizeof(struct ext4_inode_info),
921 true);
922 dump_stack();
923 }
fa0d7e3d 924 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 925}
926
51cc5068 927static void init_once(void *foo)
ac27a0ec 928{
617ba13b 929 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 930
a35afb83 931 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 932 init_rwsem(&ei->xattr_sem);
0e855ac8 933 init_rwsem(&ei->i_data_sem);
a35afb83 934 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 935}
936
e67bc2b3 937static int __init init_inodecache(void)
ac27a0ec 938{
617ba13b
MC		 939 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
940 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 941 0, (SLAB_RECLAIM_ACCOUNT|
942 SLAB_MEM_SPREAD),
20c2df83 943 init_once);
617ba13b 944 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 945 return -ENOMEM;
946 return 0;
947}
948
949static void destroy_inodecache(void)
950{
8c0a8537
KS		 951 /*
952 * Make sure all delayed rcu free inodes are flushed before we
953 * destroy cache.
954 */
955 rcu_barrier();
617ba13b 956 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 957}
958
0930fcc1 959void ext4_clear_inode(struct inode *inode)
ac27a0ec 960{
0930fcc1 961 invalidate_inode_buffers(inode);
dbd5768f 962 clear_inode(inode);
9f754758 963 dquot_drop(inode);
c2ea3fde 964 ext4_discard_preallocations(inode);
51865fda 965 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
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);
ca0e05e4 1048static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1049static int ext4_quota_on_mount(struct super_block *sb, int type);
1050static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1051 size_t len, loff_t off);
617ba13b 1052static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1053 const char *data, size_t len, loff_t off);
7c319d32
AK		 1054static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1055 unsigned int flags);
1056static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1057
96c7e0d9
JK		 1058static struct dquot **ext4_get_dquots(struct inode *inode)
1059{
1060 return EXT4_I(inode)->i_dquot;
1061}
1062
61e225dc 1063static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1064 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1065 .write_dquot = ext4_write_dquot,
1066 .acquire_dquot = ext4_acquire_dquot,
1067 .release_dquot = ext4_release_dquot,
1068 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1069 .write_info = ext4_write_info,
1070 .alloc_dquot = dquot_alloc,
1071 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1072};
1073
0d54b217 1074static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1075 .quota_on = ext4_quota_on,
ca0e05e4 1076 .quota_off = ext4_quota_off,
287a8095
CH		 1077 .quota_sync = dquot_quota_sync,
1078 .get_info = dquot_get_dqinfo,
1079 .set_info = dquot_set_dqinfo,
1080 .get_dqblk = dquot_get_dqblk,
1081 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1082};
1083#endif
1084
ee9b6d61 1085static const struct super_operations ext4_sops = {
617ba13b
MC		 1086 .alloc_inode = ext4_alloc_inode,
1087 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1088 .write_inode = ext4_write_inode,
1089 .dirty_inode = ext4_dirty_inode,
7ff9c073 1090 .drop_inode = ext4_drop_inode,
0930fcc1 1091 .evict_inode = ext4_evict_inode,
617ba13b 1092 .put_super = ext4_put_super,
617ba13b 1093 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1094 .freeze_fs = ext4_freeze,
1095 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1096 .statfs = ext4_statfs,
1097 .remount_fs = ext4_remount,
617ba13b 1098 .show_options = ext4_show_options,
ac27a0ec 1099#ifdef CONFIG_QUOTA
617ba13b
MC		 1100 .quota_read = ext4_quota_read,
1101 .quota_write = ext4_quota_write,
96c7e0d9 1102 .get_dquots = ext4_get_dquots,
ac27a0ec 1103#endif
c39a7f84 1104 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1105};
1106
39655164 1107static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1108 .fh_to_dentry = ext4_fh_to_dentry,
1109 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1110 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1111};
1112
1113enum {
1114 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1115 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1116 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1117 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1118 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1119 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1120 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1121 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1122 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1123 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1124 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1125 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
923ae0ff 1126 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1449032b 1127 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
a26f4992 1128 Opt_lazytime, Opt_nolazytime,
1449032b 1129 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1130 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1131 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1132 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1133 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1134};
1135
a447c093 1136static const match_table_t tokens = {
ac27a0ec
DK		 1137 {Opt_bsd_df, "bsddf"},
1138 {Opt_minix_df, "minixdf"},
1139 {Opt_grpid, "grpid"},
1140 {Opt_grpid, "bsdgroups"},
1141 {Opt_nogrpid, "nogrpid"},
1142 {Opt_nogrpid, "sysvgroups"},
1143 {Opt_resgid, "resgid=%u"},
1144 {Opt_resuid, "resuid=%u"},
1145 {Opt_sb, "sb=%u"},
1146 {Opt_err_cont, "errors=continue"},
1147 {Opt_err_panic, "errors=panic"},
1148 {Opt_err_ro, "errors=remount-ro"},
1149 {Opt_nouid32, "nouid32"},
ac27a0ec 1150 {Opt_debug, "debug"},
72578c33
TT		 1151 {Opt_removed, "oldalloc"},
1152 {Opt_removed, "orlov"},
ac27a0ec
DK		 1153 {Opt_user_xattr, "user_xattr"},
1154 {Opt_nouser_xattr, "nouser_xattr"},
1155 {Opt_acl, "acl"},
1156 {Opt_noacl, "noacl"},
e3bb52ae 1157 {Opt_noload, "norecovery"},
5a916be1 1158 {Opt_noload, "noload"},
72578c33
TT		 1159 {Opt_removed, "nobh"},
1160 {Opt_removed, "bh"},
ac27a0ec 1161 {Opt_commit, "commit=%u"},
30773840
TT		 1162 {Opt_min_batch_time, "min_batch_time=%u"},
1163 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1164 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1165 {Opt_journal_path, "journal_path=%s"},
818d276c 1166 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1167 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1168 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1169 {Opt_abort, "abort"},
1170 {Opt_data_journal, "data=journal"},
1171 {Opt_data_ordered, "data=ordered"},
1172 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1173 {Opt_data_err_abort, "data_err=abort"},
1174 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1175 {Opt_offusrjquota, "usrjquota="},
1176 {Opt_usrjquota, "usrjquota=%s"},
1177 {Opt_offgrpjquota, "grpjquota="},
1178 {Opt_grpjquota, "grpjquota=%s"},
1179 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1180 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1181 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1182 {Opt_grpquota, "grpquota"},
1183 {Opt_noquota, "noquota"},
1184 {Opt_quota, "quota"},
1185 {Opt_usrquota, "usrquota"},
1186 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1187 {Opt_barrier, "barrier"},
1188 {Opt_nobarrier, "nobarrier"},
25ec56b5 1189 {Opt_i_version, "i_version"},
923ae0ff 1190 {Opt_dax, "dax"},
c9de560d 1191 {Opt_stripe, "stripe=%u"},
64769240 1192 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1193 {Opt_lazytime, "lazytime"},
1194 {Opt_nolazytime, "nolazytime"},
dd919b98 1195 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1196 {Opt_removed, "mblk_io_submit"},
1197 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1198 {Opt_block_validity, "block_validity"},
1199 {Opt_noblock_validity, "noblock_validity"},
240799cd 1200 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1201 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1202 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1203 {Opt_auto_da_alloc, "auto_da_alloc"},
1204 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1205 {Opt_dioread_nolock, "dioread_nolock"},
1206 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1207 {Opt_discard, "discard"},
1208 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1209 {Opt_init_itable, "init_itable=%u"},
1210 {Opt_init_itable, "init_itable"},
1211 {Opt_noinit_itable, "noinit_itable"},
df981d03 1212 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1213 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1214 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1215 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1216 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1217 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1218 {Opt_err, NULL},
ac27a0ec
DK		 1219};
1220
617ba13b 1221static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1222{
617ba13b 1223 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1224 char *options = (char *) *data;
1225
1226 if (!options || strncmp(options, "sb=", 3) != 0)
1227 return 1; /* Default location */
0b8e58a1 1228
ac27a0ec 1229 options += 3;
0b8e58a1 1230 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1231 sb_block = simple_strtoul(options, &options, 0);
1232 if (*options && *options != ',') {
4776004f 1233 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1234 (char *) *data);
1235 return 1;
1236 }
1237 if (*options == ',')
1238 options++;
1239 *data = (void *) options;
0b8e58a1 1240
ac27a0ec
DK		 1241 return sb_block;
1242}
1243
b3881f74 1244#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1245static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1246 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1247
56c50f11
DM		 1248#ifdef CONFIG_QUOTA
1249static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1250{
1251 struct ext4_sb_info *sbi = EXT4_SB(sb);
1252 char *qname;
03dafb5f 1253 int ret = -1;
56c50f11
DM		 1254
1255 if (sb_any_quota_loaded(sb) &&
1256 !sbi->s_qf_names[qtype]) {
1257 ext4_msg(sb, KERN_ERR,
1258 "Cannot change journaled "
1259 "quota options when quota turned on");
57f73c2c 1260 return -1;
56c50f11 1261 }
262b4662
JK		 1262 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1263 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1264 "when QUOTA feature is enabled");
1265 return -1;
1266 }
56c50f11
DM		 1267 qname = match_strdup(args);
1268 if (!qname) {
1269 ext4_msg(sb, KERN_ERR,
1270 "Not enough memory for storing quotafile name");
57f73c2c 1271 return -1;
56c50f11 1272 }
03dafb5f
CG		 1273 if (sbi->s_qf_names[qtype]) {
1274 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1275 ret = 1;
1276 else
1277 ext4_msg(sb, KERN_ERR,
1278 "%s quota file already specified",
1279 QTYPE2NAME(qtype));
1280 goto errout;
56c50f11 1281 }
03dafb5f 1282 if (strchr(qname, '/')) {
56c50f11
DM		 1283 ext4_msg(sb, KERN_ERR,
1284 "quotafile must be on filesystem root");
03dafb5f 1285 goto errout;
56c50f11 1286 }
03dafb5f 1287 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1288 set_opt(sb, QUOTA);
56c50f11 1289 return 1;
03dafb5f
CG		 1290errout:
1291 kfree(qname);
1292 return ret;
56c50f11
DM		 1293}
1294
1295static int clear_qf_name(struct super_block *sb, int qtype)
1296{
1297
1298 struct ext4_sb_info *sbi = EXT4_SB(sb);
1299
1300 if (sb_any_quota_loaded(sb) &&
1301 sbi->s_qf_names[qtype]) {
1302 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1303 " when quota turned on");
57f73c2c 1304 return -1;
56c50f11 1305 }
03dafb5f 1306 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1307 sbi->s_qf_names[qtype] = NULL;
1308 return 1;
1309}
1310#endif
1311
26092bf5
TT		 1312#define MOPT_SET 0x0001
1313#define MOPT_CLEAR 0x0002
1314#define MOPT_NOSUPPORT 0x0004
1315#define MOPT_EXPLICIT 0x0008
1316#define MOPT_CLEAR_ERR 0x0010
1317#define MOPT_GTE0 0x0020
ac27a0ec 1318#ifdef CONFIG_QUOTA
26092bf5
TT		 1319#define MOPT_Q 0
1320#define MOPT_QFMT 0x0040
1321#else
1322#define MOPT_Q MOPT_NOSUPPORT
1323#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1324#endif
26092bf5 1325#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1326#define MOPT_NO_EXT2 0x0100
1327#define MOPT_NO_EXT3 0x0200
1328#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1329#define MOPT_STRING 0x0400
26092bf5
TT		 1330
1331static const struct mount_opts {
1332 int token;
1333 int mount_opt;
1334 int flags;
1335} ext4_mount_opts[] = {
1336 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1337 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1338 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1339 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1340 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1341 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1342 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1343 MOPT_EXT4_ONLY | MOPT_SET},
1344 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1345 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1346 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1347 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1348 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1349 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1350 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1351 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1352 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
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},
923ae0ff 1377 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1378 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1379 {Opt_resuid, 0, MOPT_GTE0},
1380 {Opt_resgid, 0, MOPT_GTE0},
1381 {Opt_journal_dev, 0, MOPT_GTE0},
ad4eec61 1382 {Opt_journal_path, 0, MOPT_STRING},
0efb3b23 1383 {Opt_journal_ioprio, 0, MOPT_GTE0},
8dc0aa8c
TT		 1384 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1385 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1386 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1387 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1388 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1389 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1390#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1391 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1392 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1393#else
26092bf5
TT		 1394 {Opt_acl, 0, MOPT_NOSUPPORT},
1395 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1396#endif
26092bf5
TT		 1397 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1398 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1399 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1400 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1401 MOPT_SET | MOPT_Q},
1402 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1403 MOPT_SET | MOPT_Q},
1404 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1405 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1406 {Opt_usrjquota, 0, MOPT_Q},
1407 {Opt_grpjquota, 0, MOPT_Q},
1408 {Opt_offusrjquota, 0, MOPT_Q},
1409 {Opt_offgrpjquota, 0, MOPT_Q},
1410 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1411 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1412 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1413 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1414 {Opt_err, 0, 0}
1415};
1416
1417static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1418 substring_t *args, unsigned long *journal_devnum,
1419 unsigned int *journal_ioprio, int is_remount)
1420{
1421 struct ext4_sb_info *sbi = EXT4_SB(sb);
1422 const struct mount_opts *m;
08cefc7a
EB		 1423 kuid_t uid;
1424 kgid_t gid;
26092bf5
TT		 1425 int arg = 0;
1426
57f73c2c
TT		 1427#ifdef CONFIG_QUOTA
1428 if (token == Opt_usrjquota)
1429 return set_qf_name(sb, USRQUOTA, &args[0]);
1430 else if (token == Opt_grpjquota)
1431 return set_qf_name(sb, GRPQUOTA, &args[0]);
1432 else if (token == Opt_offusrjquota)
1433 return clear_qf_name(sb, USRQUOTA);
1434 else if (token == Opt_offgrpjquota)
1435 return clear_qf_name(sb, GRPQUOTA);
1436#endif
26092bf5 1437 switch (token) {
f7048605
TT		 1438 case Opt_noacl:
1439 case Opt_nouser_xattr:
1440 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1441 break;
26092bf5
TT		 1442 case Opt_sb:
1443 return 1; /* handled by get_sb_block() */
1444 case Opt_removed:
5f3633e3 1445 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1446 return 1;
26092bf5
TT		 1447 case Opt_abort:
1448 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1449 return 1;
1450 case Opt_i_version:
1451 sb->s_flags |= MS_I_VERSION;
1452 return 1;
a26f4992
TT		 1453 case Opt_lazytime:
1454 sb->s_flags |= MS_LAZYTIME;
1455 return 1;
1456 case Opt_nolazytime:
1457 sb->s_flags &= ~MS_LAZYTIME;
1458 return 1;
26092bf5
TT		 1459 }
1460
5f3633e3
JK		 1461 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1462 if (token == m->token)
1463 break;
1464
1465 if (m->token == Opt_err) {
1466 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1467 "or missing value", opt);
1468 return -1;
1469 }
1470
8dc0aa8c
TT		 1471 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1472 ext4_msg(sb, KERN_ERR,
1473 "Mount option \"%s\" incompatible with ext2", opt);
1474 return -1;
1475 }
1476 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1477 ext4_msg(sb, KERN_ERR,
1478 "Mount option \"%s\" incompatible with ext3", opt);
1479 return -1;
1480 }
1481
ad4eec61 1482 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1483 return -1;
1484 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1485 return -1;
1486 if (m->flags & MOPT_EXPLICIT)
1487 set_opt2(sb, EXPLICIT_DELALLOC);
1488 if (m->flags & MOPT_CLEAR_ERR)
1489 clear_opt(sb, ERRORS_MASK);
1490 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1491 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1492 "options when quota turned on");
1493 return -1;
1494 }
1495
1496 if (m->flags & MOPT_NOSUPPORT) {
1497 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1498 } else if (token == Opt_commit) {
1499 if (arg == 0)
1500 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1501 sbi->s_commit_interval = HZ * arg;
1502 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1503 sbi->s_max_batch_time = arg;
1504 } else if (token == Opt_min_batch_time) {
1505 sbi->s_min_batch_time = arg;
1506 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1507 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1508 ext4_msg(sb, KERN_ERR,
1509 "EXT4-fs: inode_readahead_blks must be "
1510 "0 or a power of 2 smaller than 2^31");
26092bf5 1511 return -1;
5f3633e3
JK		 1512 }
1513 sbi->s_inode_readahead_blks = arg;
1514 } else if (token == Opt_init_itable) {
1515 set_opt(sb, INIT_INODE_TABLE);
1516 if (!args->from)
1517 arg = EXT4_DEF_LI_WAIT_MULT;
1518 sbi->s_li_wait_mult = arg;
1519 } else if (token == Opt_max_dir_size_kb) {
1520 sbi->s_max_dir_size_kb = arg;
1521 } else if (token == Opt_stripe) {
1522 sbi->s_stripe = arg;
1523 } else if (token == Opt_resuid) {
1524 uid = make_kuid(current_user_ns(), arg);
1525 if (!uid_valid(uid)) {
1526 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1527 return -1;
1528 }
5f3633e3
JK		 1529 sbi->s_resuid = uid;
1530 } else if (token == Opt_resgid) {
1531 gid = make_kgid(current_user_ns(), arg);
1532 if (!gid_valid(gid)) {
1533 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1534 return -1;
1535 }
1536 sbi->s_resgid = gid;
1537 } else if (token == Opt_journal_dev) {
1538 if (is_remount) {
1539 ext4_msg(sb, KERN_ERR,
1540 "Cannot specify journal on remount");
1541 return -1;
1542 }
1543 *journal_devnum = arg;
ad4eec61
ES		 1544 } else if (token == Opt_journal_path) {
1545 char *journal_path;
1546 struct inode *journal_inode;
1547 struct path path;
1548 int error;
1549
1550 if (is_remount) {
1551 ext4_msg(sb, KERN_ERR,
1552 "Cannot specify journal on remount");
1553 return -1;
1554 }
1555 journal_path = match_strdup(&args[0]);
1556 if (!journal_path) {
1557 ext4_msg(sb, KERN_ERR, "error: could not dup "
1558 "journal device string");
1559 return -1;
1560 }
1561
1562 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1563 if (error) {
1564 ext4_msg(sb, KERN_ERR, "error: could not find "
1565 "journal device path: error %d", error);
1566 kfree(journal_path);
1567 return -1;
1568 }
1569
1570 journal_inode = path.dentry->d_inode;
1571 if (!S_ISBLK(journal_inode->i_mode)) {
1572 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1573 "is not a block device", journal_path);
1574 path_put(&path);
1575 kfree(journal_path);
1576 return -1;
1577 }
1578
1579 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1580 path_put(&path);
1581 kfree(journal_path);
5f3633e3
JK		 1582 } else if (token == Opt_journal_ioprio) {
1583 if (arg > 7) {
1584 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1585 " (must be 0-7)");
1586 return -1;
1587 }
1588 *journal_ioprio =
1589 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1590 } else if (m->flags & MOPT_DATAJ) {
1591 if (is_remount) {
1592 if (!sbi->s_journal)
1593 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1594 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1595 ext4_msg(sb, KERN_ERR,
26092bf5 1596 "Cannot change data mode on remount");
26092bf5 1597 return -1;
ac27a0ec 1598 }
26092bf5 1599 } else {
5f3633e3
JK		 1600 clear_opt(sb, DATA_FLAGS);
1601 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1602 }
5f3633e3
JK		 1603#ifdef CONFIG_QUOTA
1604 } else if (m->flags & MOPT_QFMT) {
1605 if (sb_any_quota_loaded(sb) &&
1606 sbi->s_jquota_fmt != m->mount_opt) {
1607 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1608 "quota options when quota turned on");
1609 return -1;
1610 }
262b4662
JK		 1611 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1612 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1613 ext4_msg(sb, KERN_ERR,
1614 "Cannot set journaled quota options "
1615 "when QUOTA feature is enabled");
1616 return -1;
1617 }
5f3633e3 1618 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff
RZ		 1619#endif
1620#ifndef CONFIG_FS_DAX
1621 } else if (token == Opt_dax) {
1622 ext4_msg(sb, KERN_INFO, "dax option not supported");
1623 return -1;
5f3633e3
JK		 1624#endif
1625 } else {
1626 if (!args->from)
1627 arg = 1;
1628 if (m->flags & MOPT_CLEAR)
1629 arg = !arg;
1630 else if (unlikely(!(m->flags & MOPT_SET))) {
1631 ext4_msg(sb, KERN_WARNING,
1632 "buggy handling of option %s", opt);
1633 WARN_ON(1);
1634 return -1;
1635 }
1636 if (arg != 0)
1637 sbi->s_mount_opt |= m->mount_opt;
1638 else
1639 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1640 }
5f3633e3 1641 return 1;
26092bf5
TT		 1642}
1643
1644static int parse_options(char *options, struct super_block *sb,
1645 unsigned long *journal_devnum,
1646 unsigned int *journal_ioprio,
1647 int is_remount)
1648{
1649 struct ext4_sb_info *sbi = EXT4_SB(sb);
1650 char *p;
1651 substring_t args[MAX_OPT_ARGS];
1652 int token;
1653
1654 if (!options)
1655 return 1;
1656
1657 while ((p = strsep(&options, ",")) != NULL) {
1658 if (!*p)
1659 continue;
1660 /*
1661 * Initialize args struct so we know whether arg was
1662 * found; some options take optional arguments.
1663 */
caecd0af 1664 args[0].to = args[0].from = NULL;
26092bf5
TT		 1665 token = match_token(p, tokens, args);
1666 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1667 journal_ioprio, is_remount) < 0)
1668 return 0;
ac27a0ec
DK		 1669 }
1670#ifdef CONFIG_QUOTA
262b4662
JK		 1671 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1672 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1673 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1674 "feature is enabled");
1675 return 0;
1676 }
ac27a0ec 1677 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1678 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1679 clear_opt(sb, USRQUOTA);
ac27a0ec 1680
482a7425 1681 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1682 clear_opt(sb, GRPQUOTA);
ac27a0ec 1683
56c50f11 1684 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1685 ext4_msg(sb, KERN_ERR, "old and new quota "
1686 "format mixing");
ac27a0ec
DK		 1687 return 0;
1688 }
1689
1690 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1691 ext4_msg(sb, KERN_ERR, "journaled quota format "
1692 "not specified");
ac27a0ec
DK		 1693 return 0;
1694 }
ac27a0ec
DK		 1695 }
1696#endif
261cb20c
JK		 1697 if (test_opt(sb, DIOREAD_NOLOCK)) {
1698 int blocksize =
1699 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1700
1701 if (blocksize < PAGE_CACHE_SIZE) {
1702 ext4_msg(sb, KERN_ERR, "can't mount with "
1703 "dioread_nolock if block size != PAGE_SIZE");
1704 return 0;
1705 }
1706 }
d4f76107
JK		 1707 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1708 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1709 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1710 "in data=ordered mode");
1711 return 0;
1712 }
ac27a0ec
DK		 1713 return 1;
1714}
1715
2adf6da8
TT		 1716static inline void ext4_show_quota_options(struct seq_file *seq,
1717 struct super_block *sb)
1718{
1719#if defined(CONFIG_QUOTA)
1720 struct ext4_sb_info *sbi = EXT4_SB(sb);
1721
1722 if (sbi->s_jquota_fmt) {
1723 char *fmtname = "";
1724
1725 switch (sbi->s_jquota_fmt) {
1726 case QFMT_VFS_OLD:
1727 fmtname = "vfsold";
1728 break;
1729 case QFMT_VFS_V0:
1730 fmtname = "vfsv0";
1731 break;
1732 case QFMT_VFS_V1:
1733 fmtname = "vfsv1";
1734 break;
1735 }
1736 seq_printf(seq, ",jqfmt=%s", fmtname);
1737 }
1738
1739 if (sbi->s_qf_names[USRQUOTA])
1740 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1741
1742 if (sbi->s_qf_names[GRPQUOTA])
1743 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1744#endif
1745}
1746
5a916be1
TT		 1747static const char *token2str(int token)
1748{
50df9fd5 1749 const struct match_token *t;
5a916be1
TT		 1750
1751 for (t = tokens; t->token != Opt_err; t++)
1752 if (t->token == token && !strchr(t->pattern, '='))
1753 break;
1754 return t->pattern;
1755}
1756
2adf6da8
TT		 1757/*
1758 * Show an option if
1759 * - it's set to a non-default value OR
1760 * - if the per-sb default is different from the global default
1761 */
66acdcf4
TT		 1762static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1763 int nodefs)
2adf6da8 1764{
2adf6da8
TT		 1765 struct ext4_sb_info *sbi = EXT4_SB(sb);
1766 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1767 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1768 const struct mount_opts *m;
66acdcf4 1769 char sep = nodefs ? '\n' : ',';
2adf6da8 1770
66acdcf4
TT		 1771#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1772#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1773
1774 if (sbi->s_sb_block != 1)
5a916be1
TT		 1775 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1776
1777 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1778 int want_set = m->flags & MOPT_SET;
1779 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1780 (m->flags & MOPT_CLEAR_ERR))
1781 continue;
66acdcf4 1782 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1783 continue; /* skip if same as the default */
1784 if ((want_set &&
1785 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1786 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1787 continue; /* select Opt_noFoo vs Opt_Foo */
1788 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1789 }
5a916be1 1790
08cefc7a 1791 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1792 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1793 SEQ_OPTS_PRINT("resuid=%u",
1794 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1795 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1796 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1797 SEQ_OPTS_PRINT("resgid=%u",
1798 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1799 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1800 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1801 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1802 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1803 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1804 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1805 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1806 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1807 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1808 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1809 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1810 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1811 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1812 if (sb->s_flags & MS_I_VERSION)
5a916be1 1813 SEQ_OPTS_PUTS("i_version");
66acdcf4 1814 if (nodefs || sbi->s_stripe)
5a916be1 1815 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1816 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1817 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1818 SEQ_OPTS_PUTS("data=journal");
1819 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1820 SEQ_OPTS_PUTS("data=ordered");
1821 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1822 SEQ_OPTS_PUTS("data=writeback");
1823 }
66acdcf4
TT		 1824 if (nodefs ||
1825 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1826 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1827 sbi->s_inode_readahead_blks);
2adf6da8 1828
66acdcf4
TT		 1829 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1830 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1831 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1832 if (nodefs || sbi->s_max_dir_size_kb)
1833 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1834
1835 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1836 return 0;
1837}
1838
66acdcf4
TT		 1839static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1840{
1841 return _ext4_show_options(seq, root->d_sb, 0);
1842}
1843
1844static int options_seq_show(struct seq_file *seq, void *offset)
1845{
1846 struct super_block *sb = seq->private;
1847 int rc;
1848
1849 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1850 rc = _ext4_show_options(seq, sb, 1);
1851 seq_puts(seq, "\n");
1852 return rc;
1853}
1854
1855static int options_open_fs(struct inode *inode, struct file *file)
1856{
d9dda78b 1857 return single_open(file, options_seq_show, PDE_DATA(inode));
66acdcf4
TT		 1858}
1859
1860static const struct file_operations ext4_seq_options_fops = {
1861 .owner = THIS_MODULE,
1862 .open = options_open_fs,
1863 .read = seq_read,
1864 .llseek = seq_lseek,
1865 .release = single_release,
1866};
1867
617ba13b 1868static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1869 int read_only)
1870{
617ba13b 1871 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1872 int res = 0;
1873
617ba13b 1874 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1875 ext4_msg(sb, KERN_ERR, "revision level too high, "
1876 "forcing read-only mode");
ac27a0ec
DK		 1877 res = MS_RDONLY;
1878 }
1879 if (read_only)
281b5995 1880 goto done;
617ba13b 1881 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1882 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1883 "running e2fsck is recommended");
c8b459f4 1884 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 1885 ext4_msg(sb, KERN_WARNING,
1886 "warning: mounting fs with errors, "
1887 "running e2fsck is recommended");
ed3ce80a 1888 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1889 le16_to_cpu(es->s_mnt_count) >=
1890 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1891 ext4_msg(sb, KERN_WARNING,
1892 "warning: maximal mount count reached, "
1893 "running e2fsck is recommended");
ac27a0ec
DK		 1894 else if (le32_to_cpu(es->s_checkinterval) &&
1895 (le32_to_cpu(es->s_lastcheck) +
1896 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1897 ext4_msg(sb, KERN_WARNING,
1898 "warning: checktime reached, "
1899 "running e2fsck is recommended");
0b8e58a1 1900 if (!sbi->s_journal)
0390131b 1901 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1902 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1903 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1904 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1905 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1906 ext4_update_dynamic_rev(sb);
0390131b
FM		 1907 if (sbi->s_journal)
1908 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1909
e2d67052 1910 ext4_commit_super(sb, 1);
281b5995 1911done:
ac27a0ec 1912 if (test_opt(sb, DEBUG))
a9df9a49 1913 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1914 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1915 sb->s_blocksize,
1916 sbi->s_groups_count,
617ba13b
MC		 1917 EXT4_BLOCKS_PER_GROUP(sb),
1918 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1919 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1920
7abc52c2 1921 cleancache_init_fs(sb);
ac27a0ec
DK		 1922 return res;
1923}
1924
117fff10
TT		 1925int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1926{
1927 struct ext4_sb_info *sbi = EXT4_SB(sb);
1928 struct flex_groups *new_groups;
1929 int size;
1930
1931 if (!sbi->s_log_groups_per_flex)
1932 return 0;
1933
1934 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1935 if (size <= sbi->s_flex_groups_allocated)
1936 return 0;
1937
1938 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1939 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1940 if (!new_groups) {
1941 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1942 size / (int) sizeof(struct flex_groups));
1943 return -ENOMEM;
1944 }
1945
1946 if (sbi->s_flex_groups) {
1947 memcpy(new_groups, sbi->s_flex_groups,
1948 (sbi->s_flex_groups_allocated *
1949 sizeof(struct flex_groups)));
b93b41d4 1950 kvfree(sbi->s_flex_groups);
117fff10
TT		 1951 }
1952 sbi->s_flex_groups = new_groups;
1953 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1954 return 0;
1955}
1956
772cb7c8
JS		 1957static int ext4_fill_flex_info(struct super_block *sb)
1958{
1959 struct ext4_sb_info *sbi = EXT4_SB(sb);
1960 struct ext4_group_desc *gdp = NULL;
772cb7c8 1961 ext4_group_t flex_group;
117fff10 1962 int i, err;
772cb7c8 1963
503358ae 1964 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1965 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1966 sbi->s_log_groups_per_flex = 0;
1967 return 1;
1968 }
1969
117fff10
TT		 1970 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1971 if (err)
9933fc0a 1972 goto failed;
772cb7c8 1973
772cb7c8 1974 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1975 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1976
1977 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1978 atomic_add(ext4_free_inodes_count(sb, gdp),
1979 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 1980 atomic64_add(ext4_free_group_clusters(sb, gdp),
1981 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1982 atomic_add(ext4_used_dirs_count(sb, gdp),
1983 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1984 }
1985
1986 return 1;
1987failed:
1988 return 0;
1989}
1990
feb0ab32
DW		 1991static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1992 struct ext4_group_desc *gdp)
717d50e4 1993{
feb0ab32 1994 int offset;
717d50e4 1995 __u16 crc = 0;
feb0ab32 1996 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1997
9aa5d32b 1998 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 1999 /* Use new metadata_csum algorithm */
d6a77105 2000 __le16 save_csum;
feb0ab32
DW		 2001 __u32 csum32;
2002
d6a77105 2003 save_csum = gdp->bg_checksum;
feb0ab32
DW		 2004 gdp->bg_checksum = 0;
2005 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2006 sizeof(le_group));
2007 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2008 sbi->s_desc_size);
d6a77105 2009 gdp->bg_checksum = save_csum;
feb0ab32
DW		 2010
2011 crc = csum32 & 0xFFFF;
2012 goto out;
717d50e4
AD		 2013 }
2014
feb0ab32 2015 /* old crc16 code */
813d32f9
DW		 2016 if (!(sbi->s_es->s_feature_ro_compat &
2017 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2018 return 0;
2019
feb0ab32
DW		 2020 offset = offsetof(struct ext4_group_desc, bg_checksum);
2021
2022 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2023 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2024 crc = crc16(crc, (__u8 *)gdp, offset);
2025 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2026 /* for checksum of struct ext4_group_desc do the rest...*/
2027 if ((sbi->s_es->s_feature_incompat &
2028 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2029 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2030 crc = crc16(crc, (__u8 *)gdp + offset,
2031 le16_to_cpu(sbi->s_es->s_desc_size) -
2032 offset);
2033
2034out:
717d50e4
AD		 2035 return cpu_to_le16(crc);
2036}
2037
feb0ab32 2038int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2039 struct ext4_group_desc *gdp)
2040{
feb0ab32
DW		 2041 if (ext4_has_group_desc_csum(sb) &&
2042 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2043 block_group, gdp)))
717d50e4
AD		 2044 return 0;
2045
2046 return 1;
2047}
2048
feb0ab32
DW		 2049void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2050 struct ext4_group_desc *gdp)
2051{
2052 if (!ext4_has_group_desc_csum(sb))
2053 return;
2054 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2055}
2056
ac27a0ec 2057/* Called at mount-time, super-block is locked */
bfff6873
LC		 2058static int ext4_check_descriptors(struct super_block *sb,
2059 ext4_group_t *first_not_zeroed)
ac27a0ec 2060{
617ba13b
MC		 2061 struct ext4_sb_info *sbi = EXT4_SB(sb);
2062 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2063 ext4_fsblk_t last_block;
bd81d8ee
LV		 2064 ext4_fsblk_t block_bitmap;
2065 ext4_fsblk_t inode_bitmap;
2066 ext4_fsblk_t inode_table;
ce421581 2067 int flexbg_flag = 0;
bfff6873 2068 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2069
ce421581
JS		 2070 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2071 flexbg_flag = 1;
2072
af5bc92d 2073 ext4_debug("Checking group descriptors");
ac27a0ec 2074
197cd65a
AM		 2075 for (i = 0; i < sbi->s_groups_count; i++) {
2076 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2077
ce421581 2078 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2079 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2080 else
2081 last_block = first_block +
617ba13b 2082 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2083
bfff6873
LC		 2084 if ((grp == sbi->s_groups_count) &&
2085 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2086 grp = i;
2087
8fadc143 2088 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2089 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2090 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2091 "Block bitmap for group %u not in group "
b31e1552 2092 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2093 return 0;
2094 }
8fadc143 2095 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2096 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2097 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2098 "Inode bitmap for group %u not in group "
b31e1552 2099 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2100 return 0;
2101 }
8fadc143 2102 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2103 if (inode_table < first_block ||
2b2d6d01 2104 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2105 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2106 "Inode table for group %u not in group "
b31e1552 2107 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2108 return 0;
2109 }
955ce5f5 2110 ext4_lock_group(sb, i);
feb0ab32 2111 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2112 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2113 "Checksum for group %u failed (%u!=%u)",
2114 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2115 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2116 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2117 ext4_unlock_group(sb, i);
8a266467 2118 return 0;
7ee1ec4c 2119 }
717d50e4 2120 }
955ce5f5 2121 ext4_unlock_group(sb, i);
ce421581
JS		 2122 if (!flexbg_flag)
2123 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2124 }
bfff6873
LC		 2125 if (NULL != first_not_zeroed)
2126 *first_not_zeroed = grp;
ac27a0ec
DK		 2127 return 1;
2128}
2129
617ba13b 2130/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2131 * the superblock) which were deleted from all directories, but held open by
2132 * a process at the time of a crash. We walk the list and try to delete these
2133 * inodes at recovery time (only with a read-write filesystem).
2134 *
2135 * In order to keep the orphan inode chain consistent during traversal (in
2136 * case of crash during recovery), we link each inode into the superblock
2137 * orphan list_head and handle it the same way as an inode deletion during
2138 * normal operation (which journals the operations for us).
2139 *
2140 * We only do an iget() and an iput() on each inode, which is very safe if we
2141 * accidentally point at an in-use or already deleted inode. The worst that
2142 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2143 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2144 * e2fsck was run on this filesystem, and it must have already done the orphan
2145 * inode cleanup for us, so we can safely abort without any further action.
2146 */
2b2d6d01
TT		 2147static void ext4_orphan_cleanup(struct super_block *sb,
2148 struct ext4_super_block *es)
ac27a0ec
DK		 2149{
2150 unsigned int s_flags = sb->s_flags;
2151 int nr_orphans = 0, nr_truncates = 0;
2152#ifdef CONFIG_QUOTA
2153 int i;
2154#endif
2155 if (!es->s_last_orphan) {
2156 jbd_debug(4, "no orphan inodes to clean up\n");
2157 return;
2158 }
2159
a8f48a95 2160 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2161 ext4_msg(sb, KERN_ERR, "write access "
2162 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2163 return;
2164 }
2165
d39195c3
AG		 2166 /* Check if feature set would not allow a r/w mount */
2167 if (!ext4_feature_set_ok(sb, 0)) {
2168 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2169 "unknown ROCOMPAT features");
2170 return;
2171 }
2172
617ba13b 2173 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2174 /* don't clear list on RO mount w/ errors */
2175 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2176 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2177 "clearing orphan list.\n");
c25f9bc6
ES		 2178 es->s_last_orphan = 0;
2179 }
ac27a0ec
DK		 2180 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2181 return;
2182 }
2183
2184 if (s_flags & MS_RDONLY) {
b31e1552 2185 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2186 sb->s_flags &= ~MS_RDONLY;
2187 }
2188#ifdef CONFIG_QUOTA
2189 /* Needed for iput() to work correctly and not trash data */
2190 sb->s_flags |= MS_ACTIVE;
2191 /* Turn on quotas so that they are updated correctly */
a2d4a646 2192 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2193 if (EXT4_SB(sb)->s_qf_names[i]) {
2194 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2195 if (ret < 0)
b31e1552
ES		 2196 ext4_msg(sb, KERN_ERR,
2197 "Cannot turn on journaled "
2198 "quota: error %d", ret);
ac27a0ec
DK		 2199 }
2200 }
2201#endif
2202
2203 while (es->s_last_orphan) {
2204 struct inode *inode;
2205
97bd42b9
JB		 2206 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2207 if (IS_ERR(inode)) {
ac27a0ec
DK		 2208 es->s_last_orphan = 0;
2209 break;
2210 }
2211
617ba13b 2212 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2213 dquot_initialize(inode);
ac27a0ec 2214 if (inode->i_nlink) {
566370a2
PT		 2215 if (test_opt(sb, DEBUG))
2216 ext4_msg(sb, KERN_DEBUG,
2217 "%s: truncating inode %lu to %lld bytes",
2218 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2219 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2220 inode->i_ino, inode->i_size);
721e3eba 2221 mutex_lock(&inode->i_mutex);
55f252c9 2222 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2223 ext4_truncate(inode);
721e3eba 2224 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2225 nr_truncates++;
2226 } else {
566370a2
PT		 2227 if (test_opt(sb, DEBUG))
2228 ext4_msg(sb, KERN_DEBUG,
2229 "%s: deleting unreferenced inode %lu",
2230 __func__, inode->i_ino);
ac27a0ec
DK		 2231 jbd_debug(2, "deleting unreferenced inode %lu\n",
2232 inode->i_ino);
2233 nr_orphans++;
2234 }
2235 iput(inode); /* The delete magic happens here! */
2236 }
2237
2b2d6d01 2238#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2239
2240 if (nr_orphans)
b31e1552
ES		 2241 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2242 PLURAL(nr_orphans));
ac27a0ec 2243 if (nr_truncates)
b31e1552
ES		 2244 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2245 PLURAL(nr_truncates));
ac27a0ec
DK		 2246#ifdef CONFIG_QUOTA
2247 /* Turn quotas off */
a2d4a646 2248 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2249 if (sb_dqopt(sb)->files[i])
287a8095 2250 dquot_quota_off(sb, i);
ac27a0ec
DK		 2251 }
2252#endif
2253 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2254}
0b8e58a1 2255
cd2291a4
ES		 2256/*
2257 * Maximal extent format file size.
2258 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2259 * extent format containers, within a sector_t, and within i_blocks
2260 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2261 * so that won't be a limiting factor.
2262 *
f17722f9
LC		 2263 * However there is other limiting factor. We do store extents in the form
2264 * of starting block and length, hence the resulting length of the extent
2265 * covering maximum file size must fit into on-disk format containers as
2266 * well. Given that length is always by 1 unit bigger than max unit (because
2267 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2268 *
cd2291a4
ES		 2269 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2270 */
f287a1a5 2271static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2272{
2273 loff_t res;
2274 loff_t upper_limit = MAX_LFS_FILESIZE;
2275
2276 /* small i_blocks in vfs inode? */
f287a1a5 2277 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2278 /*
90c699a9 2279 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2280 * i_block represent total blocks in 512 bytes
2281 * 32 == size of vfs inode i_blocks * 8
2282 */
2283 upper_limit = (1LL << 32) - 1;
2284
2285 /* total blocks in file system block size */
2286 upper_limit >>= (blkbits - 9);
2287 upper_limit <<= blkbits;
2288 }
2289
f17722f9
LC		 2290 /*
2291 * 32-bit extent-start container, ee_block. We lower the maxbytes
2292 * by one fs block, so ee_len can cover the extent of maximum file
2293 * size
2294 */
2295 res = (1LL << 32) - 1;
cd2291a4 2296 res <<= blkbits;
cd2291a4
ES		 2297
2298 /* Sanity check against vm- & vfs- imposed limits */
2299 if (res > upper_limit)
2300 res = upper_limit;
2301
2302 return res;
2303}
ac27a0ec 2304
ac27a0ec 2305/*
cd2291a4 2306 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2307 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2308 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2309 */
f287a1a5 2310static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2311{
617ba13b 2312 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2313 int meta_blocks;
2314 loff_t upper_limit;
0b8e58a1
AD		 2315 /* This is calculated to be the largest file size for a dense, block
2316 * mapped file such that the file's total number of 512-byte sectors,
2317 * including data and all indirect blocks, does not exceed (2^48 - 1).
2318 *
2319 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2320 * number of 512-byte sectors of the file.
0fc1b451
AK		 2321 */
2322
f287a1a5 2323 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2324 /*
90c699a9 2325 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2326 * the inode i_block field represents total file blocks in
2327 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2328 */
2329 upper_limit = (1LL << 32) - 1;
2330
2331 /* total blocks in file system block size */
2332 upper_limit >>= (bits - 9);
2333
2334 } else {
8180a562
AK		 2335 /*
2336 * We use 48 bit ext4_inode i_blocks
2337 * With EXT4_HUGE_FILE_FL set the i_blocks
2338 * represent total number of blocks in
2339 * file system block size
2340 */
0fc1b451
AK		 2341 upper_limit = (1LL << 48) - 1;
2342
0fc1b451
AK		 2343 }
2344
2345 /* indirect blocks */
2346 meta_blocks = 1;
2347 /* double indirect blocks */
2348 meta_blocks += 1 + (1LL << (bits-2));
2349 /* tripple indirect blocks */
2350 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2351
2352 upper_limit -= meta_blocks;
2353 upper_limit <<= bits;
ac27a0ec
DK		 2354
2355 res += 1LL << (bits-2);
2356 res += 1LL << (2*(bits-2));
2357 res += 1LL << (3*(bits-2));
2358 res <<= bits;
2359 if (res > upper_limit)
2360 res = upper_limit;
0fc1b451
AK		 2361
2362 if (res > MAX_LFS_FILESIZE)
2363 res = MAX_LFS_FILESIZE;
2364
ac27a0ec
DK		 2365 return res;
2366}
2367
617ba13b 2368static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2369 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2370{
617ba13b 2371 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2372 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2373 int has_super = 0;
2374
2375 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2376
617ba13b 2377 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2378 nr < first_meta_bg)
70bbb3e0 2379 return logical_sb_block + nr + 1;
ac27a0ec 2380 bg = sbi->s_desc_per_block * nr;
617ba13b 2381 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2382 has_super = 1;
0b8e58a1 2383
bd63f6b0
DW		 2384 /*
2385 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2386 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2387 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2388 * compensate.
2389 */
2390 if (sb->s_blocksize == 1024 && nr == 0 &&
2391 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2392 has_super++;
2393
617ba13b 2394 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2395}
2396
c9de560d
AT		 2397/**
2398 * ext4_get_stripe_size: Get the stripe size.
2399 * @sbi: In memory super block info
2400 *
2401 * If we have specified it via mount option, then
2402 * use the mount option value. If the value specified at mount time is
2403 * greater than the blocks per group use the super block value.
2404 * If the super block value is greater than blocks per group return 0.
2405 * Allocator needs it be less than blocks per group.
2406 *
2407 */
2408static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2409{
2410 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2411 unsigned long stripe_width =
2412 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2413 int ret;
c9de560d
AT		 2414
2415 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2416 ret = sbi->s_stripe;
2417 else if (stripe_width <= sbi->s_blocks_per_group)
2418 ret = stripe_width;
2419 else if (stride <= sbi->s_blocks_per_group)
2420 ret = stride;
2421 else
2422 ret = 0;
c9de560d 2423
3eb08658
DE		 2424 /*
2425 * If the stripe width is 1, this makes no sense and
2426 * we set it to 0 to turn off stripe handling code.
2427 */
2428 if (ret <= 1)
2429 ret = 0;
c9de560d 2430
3eb08658 2431 return ret;
c9de560d 2432}
ac27a0ec 2433
3197ebdb
TT		 2434/* sysfs supprt */
2435
2436struct ext4_attr {
2437 struct attribute attr;
2438 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2439 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb 2440 const char *, size_t);
f2d50a65
JK		 2441 union {
2442 int offset;
2443 int deprecated_val;
2444 } u;
3197ebdb
TT		 2445};
2446
27dd4385
LC		 2447static int parse_strtoull(const char *buf,
2448 unsigned long long max, unsigned long long *value)
2449{
2450 int ret;
2451
2452 ret = kstrtoull(skip_spaces(buf), 0, value);
2453 if (!ret && *value > max)
2454 ret = -EINVAL;
2455 return ret;
2456}
2457
3197ebdb
TT		 2458static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2459 struct ext4_sb_info *sbi,
2460 char *buf)
2461{
2462 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2463 (s64) EXT4_C2B(sbi,
2464 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2465}
2466
2467static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2468 struct ext4_sb_info *sbi, char *buf)
2469{
2470 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2471
f613dfcb
TT		 2472 if (!sb->s_bdev->bd_part)
2473 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2474 return snprintf(buf, PAGE_SIZE, "%lu\n",
2475 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2476 sbi->s_sectors_written_start) >> 1);
2477}
2478
2479static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2480 struct ext4_sb_info *sbi, char *buf)
2481{
2482 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2483
f613dfcb
TT		 2484 if (!sb->s_bdev->bd_part)
2485 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2486 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2487 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2488 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2489 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2490}
2491
2492static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2493 struct ext4_sb_info *sbi,
2494 const char *buf, size_t count)
2495{
2496 unsigned long t;
e1091b15 2497 int ret;
3197ebdb 2498
e1091b15
LC		 2499 ret = kstrtoul(skip_spaces(buf), 0, &t);
2500 if (ret)
2501 return ret;
3197ebdb 2502
e1091b15 2503 if (t && (!is_power_of_2(t) || t > 0x40000000))
3197ebdb
TT		 2504 return -EINVAL;
2505
2506 sbi->s_inode_readahead_blks = t;
2507 return count;
2508}
2509
2510static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2511 struct ext4_sb_info *sbi, char *buf)
3197ebdb 2512{
f2d50a65 2513 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb
TT		 2514
2515 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2516}
2517
2518static ssize_t sbi_ui_store(struct ext4_attr *a,
2519 struct ext4_sb_info *sbi,
2520 const char *buf, size_t count)
2521{
f2d50a65 2522 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb 2523 unsigned long t;
e1091b15 2524 int ret;
3197ebdb 2525
e1091b15
LC		 2526 ret = kstrtoul(skip_spaces(buf), 0, &t);
2527 if (ret)
2528 return ret;
3197ebdb
TT		 2529 *ui = t;
2530 return count;
2531}
2532
52c198c6
LC		 2533static ssize_t es_ui_show(struct ext4_attr *a,
2534 struct ext4_sb_info *sbi, char *buf)
2535{
2536
2537 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2538 a->u.offset);
2539
2540 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2541}
2542
27dd4385
LC		 2543static ssize_t reserved_clusters_show(struct ext4_attr *a,
2544 struct ext4_sb_info *sbi, char *buf)
2545{
2546 return snprintf(buf, PAGE_SIZE, "%llu\n",
2547 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2548}
2549
2550static ssize_t reserved_clusters_store(struct ext4_attr *a,
2551 struct ext4_sb_info *sbi,
2552 const char *buf, size_t count)
2553{
2554 unsigned long long val;
2555 int ret;
2556
2557 if (parse_strtoull(buf, -1ULL, &val))
2558 return -EINVAL;
2559 ret = ext4_reserve_clusters(sbi, val);
2560
2561 return ret ? ret : count;
2562}
2563
2c0544b2
TT		 2564static ssize_t trigger_test_error(struct ext4_attr *a,
2565 struct ext4_sb_info *sbi,
2566 const char *buf, size_t count)
2567{
2568 int len = count;
2569
2570 if (!capable(CAP_SYS_ADMIN))
2571 return -EPERM;
2572
2573 if (len && buf[len-1] == '\n')
2574 len--;
2575
2576 if (len)
2577 ext4_error(sbi->s_sb, "%.*s", len, buf);
2578 return count;
2579}
2580
f2d50a65
JK		 2581static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2582 struct ext4_sb_info *sbi, char *buf)
2583{
2584 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2585}
2586
3197ebdb
TT		 2587#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2588static struct ext4_attr ext4_attr_##_name = { \
2589 .attr = {.name = __stringify(_name), .mode = _mode }, \
2590 .show = _show, \
2591 .store = _store, \
f2d50a65
JK		 2592 .u = { \
2593 .offset = offsetof(struct ext4_sb_info, _elname),\
2594 }, \
3197ebdb 2595}
52c198c6
LC		 2596
2597#define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2598static struct ext4_attr ext4_attr_##_name = { \
2599 .attr = {.name = __stringify(_name), .mode = _mode }, \
2600 .show = _show, \
2601 .store = _store, \
2602 .u = { \
2603 .offset = offsetof(struct ext4_super_block, _elname), \
2604 }, \
2605}
2606
3197ebdb
TT		 2607#define EXT4_ATTR(name, mode, show, store) \
2608static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2609
857ac889 2610#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2611#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2612#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
52c198c6
LC		 2613
2614#define EXT4_RO_ATTR_ES_UI(name, elname) \
2615 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
3197ebdb
TT		 2616#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2617 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
52c198c6 2618
3197ebdb 2619#define ATTR_LIST(name) &ext4_attr_##name.attr
f2d50a65
JK		 2620#define EXT4_DEPRECATED_ATTR(_name, _val) \
2621static struct ext4_attr ext4_attr_##_name = { \
2622 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2623 .show = sbi_deprecated_show, \
2624 .u = { \
2625 .deprecated_val = _val, \
2626 }, \
2627}
3197ebdb
TT		 2628
2629EXT4_RO_ATTR(delayed_allocation_blocks);
2630EXT4_RO_ATTR(session_write_kbytes);
2631EXT4_RO_ATTR(lifetime_write_kbytes);
27dd4385 2632EXT4_RW_ATTR(reserved_clusters);
3197ebdb
TT		 2633EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2634 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2635EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2636EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2637EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2638EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2639EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2640EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2641EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
f2d50a65 2642EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
67a5da56 2643EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2644EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
efbed4dc
TT		 2645EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2646EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2647EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2648EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2649EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2650EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
52c198c6
LC		 2651EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2652EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2653EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
3197ebdb
TT		 2654
2655static struct attribute *ext4_attrs[] = {
2656 ATTR_LIST(delayed_allocation_blocks),
2657 ATTR_LIST(session_write_kbytes),
2658 ATTR_LIST(lifetime_write_kbytes),
27dd4385 2659 ATTR_LIST(reserved_clusters),
3197ebdb 2660 ATTR_LIST(inode_readahead_blks),
11013911 2661 ATTR_LIST(inode_goal),
3197ebdb
TT		 2662 ATTR_LIST(mb_stats),
2663 ATTR_LIST(mb_max_to_scan),
2664 ATTR_LIST(mb_min_to_scan),
2665 ATTR_LIST(mb_order2_req),
2666 ATTR_LIST(mb_stream_req),
2667 ATTR_LIST(mb_group_prealloc),
55138e0b 2668 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2669 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2670 ATTR_LIST(trigger_fs_error),
efbed4dc
TT		 2671 ATTR_LIST(err_ratelimit_interval_ms),
2672 ATTR_LIST(err_ratelimit_burst),
2673 ATTR_LIST(warning_ratelimit_interval_ms),
2674 ATTR_LIST(warning_ratelimit_burst),
2675 ATTR_LIST(msg_ratelimit_interval_ms),
2676 ATTR_LIST(msg_ratelimit_burst),
52c198c6
LC		 2677 ATTR_LIST(errors_count),
2678 ATTR_LIST(first_error_time),
2679 ATTR_LIST(last_error_time),
3197ebdb
TT		 2680 NULL,
2681};
2682
857ac889
LC		 2683/* Features this copy of ext4 supports */
2684EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2685EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2686EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2687
2688static struct attribute *ext4_feat_attrs[] = {
2689 ATTR_LIST(lazy_itable_init),
27ee40df 2690 ATTR_LIST(batched_discard),
5e7bbef1 2691 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2692 NULL,
2693};
2694
3197ebdb
TT		 2695static ssize_t ext4_attr_show(struct kobject *kobj,
2696 struct attribute *attr, char *buf)
2697{
2698 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2699 s_kobj);
2700 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2701
2702 return a->show ? a->show(a, sbi, buf) : 0;
2703}
2704
2705static ssize_t ext4_attr_store(struct kobject *kobj,
2706 struct attribute *attr,
2707 const char *buf, size_t len)
2708{
2709 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2710 s_kobj);
2711 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2712
2713 return a->store ? a->store(a, sbi, buf, len) : 0;
2714}
2715
2716static void ext4_sb_release(struct kobject *kobj)
2717{
2718 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2719 s_kobj);
2720 complete(&sbi->s_kobj_unregister);
2721}
2722
52cf25d0 2723static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2724 .show = ext4_attr_show,
2725 .store = ext4_attr_store,
2726};
2727
2728static struct kobj_type ext4_ktype = {
2729 .default_attrs = ext4_attrs,
2730 .sysfs_ops = &ext4_attr_ops,
2731 .release = ext4_sb_release,
2732};
2733
857ac889
LC		 2734static void ext4_feat_release(struct kobject *kobj)
2735{
2736 complete(&ext4_feat->f_kobj_unregister);
2737}
2738
c7f72543
LC		 2739static ssize_t ext4_feat_show(struct kobject *kobj,
2740 struct attribute *attr, char *buf)
2741{
2742 return snprintf(buf, PAGE_SIZE, "supported\n");
2743}
2744
2745/*
2746 * We can not use ext4_attr_show/store because it relies on the kobject
2747 * being embedded in the ext4_sb_info structure which is definitely not
2748 * true in this case.
2749 */
2750static const struct sysfs_ops ext4_feat_ops = {
2751 .show = ext4_feat_show,
2752 .store = NULL,
2753};
2754
857ac889
LC		 2755static struct kobj_type ext4_feat_ktype = {
2756 .default_attrs = ext4_feat_attrs,
c7f72543 2757 .sysfs_ops = &ext4_feat_ops,
857ac889
LC		 2758 .release = ext4_feat_release,
2759};
2760
a13fb1a4
ES		 2761/*
2762 * Check whether this filesystem can be mounted based on
2763 * the features present and the RDONLY/RDWR mount requested.
2764 * Returns 1 if this filesystem can be mounted as requested,
2765 * 0 if it cannot be.
2766 */
2767static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2768{
2769 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2770 ext4_msg(sb, KERN_ERR,
2771 "Couldn't mount because of "
2772 "unsupported optional features (%x)",
2773 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2774 ~EXT4_FEATURE_INCOMPAT_SUPP));
2775 return 0;
2776 }
2777
2778 if (readonly)
2779 return 1;
2780
2cb5cc8b
DW		 2781 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2782 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2783 sb->s_flags |= MS_RDONLY;
2784 return 1;
2785 }
2786
a13fb1a4
ES		 2787 /* Check that feature set is OK for a read-write mount */
2788 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2789 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2790 "unsupported optional features (%x)",
2791 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2792 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2793 return 0;
2794 }
2795 /*
2796 * Large file size enabled file system can only be mounted
2797 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2798 */
2799 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2800 if (sizeof(blkcnt_t) < sizeof(u64)) {
2801 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2802 "cannot be mounted RDWR without "
2803 "CONFIG_LBDAF");
2804 return 0;
2805 }
2806 }
bab08ab9
TT		 2807 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2808 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2809 ext4_msg(sb, KERN_ERR,
2810 "Can't support bigalloc feature without "
2811 "extents feature\n");
2812 return 0;
2813 }
7c319d32
AK		 2814
2815#ifndef CONFIG_QUOTA
2816 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2817 !readonly) {
2818 ext4_msg(sb, KERN_ERR,
2819 "Filesystem with quota feature cannot be mounted RDWR "
2820 "without CONFIG_QUOTA");
2821 return 0;
2822 }
2823#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2824 return 1;
2825}
2826
66e61a9e
TT		 2827/*
2828 * This function is called once a day if we have errors logged
2829 * on the file system
2830 */
2831static void print_daily_error_info(unsigned long arg)
2832{
2833 struct super_block *sb = (struct super_block *) arg;
2834 struct ext4_sb_info *sbi;
2835 struct ext4_super_block *es;
2836
2837 sbi = EXT4_SB(sb);
2838 es = sbi->s_es;
2839
2840 if (es->s_error_count)
ae0f78de
TT		 2841 /* fsck newer than v1.41.13 is needed to clean this condition. */
2842 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2843 le32_to_cpu(es->s_error_count));
2844 if (es->s_first_error_time) {
ae0f78de 2845 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2846 sb->s_id, le32_to_cpu(es->s_first_error_time),
2847 (int) sizeof(es->s_first_error_func),
2848 es->s_first_error_func,
2849 le32_to_cpu(es->s_first_error_line));
2850 if (es->s_first_error_ino)
2851 printk(": inode %u",
2852 le32_to_cpu(es->s_first_error_ino));
2853 if (es->s_first_error_block)
2854 printk(": block %llu", (unsigned long long)
2855 le64_to_cpu(es->s_first_error_block));
2856 printk("\n");
2857 }
2858 if (es->s_last_error_time) {
ae0f78de 2859 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2860 sb->s_id, le32_to_cpu(es->s_last_error_time),
2861 (int) sizeof(es->s_last_error_func),
2862 es->s_last_error_func,
2863 le32_to_cpu(es->s_last_error_line));
2864 if (es->s_last_error_ino)
2865 printk(": inode %u",
2866 le32_to_cpu(es->s_last_error_ino));
2867 if (es->s_last_error_block)
2868 printk(": block %llu", (unsigned long long)
2869 le64_to_cpu(es->s_last_error_block));
2870 printk("\n");
2871 }
2872 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2873}
2874
bfff6873
LC		 2875/* Find next suitable group and run ext4_init_inode_table */
2876static int ext4_run_li_request(struct ext4_li_request *elr)
2877{
2878 struct ext4_group_desc *gdp = NULL;
2879 ext4_group_t group, ngroups;
2880 struct super_block *sb;
2881 unsigned long timeout = 0;
2882 int ret = 0;
2883
2884 sb = elr->lr_super;
2885 ngroups = EXT4_SB(sb)->s_groups_count;
2886
8e8ad8a5 2887 sb_start_write(sb);
bfff6873
LC		 2888 for (group = elr->lr_next_group; group < ngroups; group++) {
2889 gdp = ext4_get_group_desc(sb, group, NULL);
2890 if (!gdp) {
2891 ret = 1;
2892 break;
2893 }
2894
2895 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2896 break;
2897 }
2898
7f511862 2899 if (group >= ngroups)
bfff6873
LC		 2900 ret = 1;
2901
2902 if (!ret) {
2903 timeout = jiffies;
2904 ret = ext4_init_inode_table(sb, group,
2905 elr->lr_timeout ? 0 : 1);
2906 if (elr->lr_timeout == 0) {
51ce6511
LC		 2907 timeout = (jiffies - timeout) *
2908 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2909 elr->lr_timeout = timeout;
2910 }
2911 elr->lr_next_sched = jiffies + elr->lr_timeout;
2912 elr->lr_next_group = group + 1;
2913 }
8e8ad8a5 2914 sb_end_write(sb);
bfff6873
LC		 2915
2916 return ret;
2917}
2918
2919/*
2920 * Remove lr_request from the list_request and free the
4ed5c033 2921 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2922 */
2923static void ext4_remove_li_request(struct ext4_li_request *elr)
2924{
2925 struct ext4_sb_info *sbi;
2926
2927 if (!elr)
2928 return;
2929
2930 sbi = elr->lr_sbi;
2931
2932 list_del(&elr->lr_request);
2933 sbi->s_li_request = NULL;
2934 kfree(elr);
2935}
2936
2937static void ext4_unregister_li_request(struct super_block *sb)
2938{
1bb933fb
LC		 2939 mutex_lock(&ext4_li_mtx);
2940 if (!ext4_li_info) {
2941 mutex_unlock(&ext4_li_mtx);
bfff6873 2942 return;
1bb933fb 2943 }
bfff6873
LC		 2944
2945 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2946 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2947 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2948 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2949}
2950
8f1f7453
ES		 2951static struct task_struct *ext4_lazyinit_task;
2952
bfff6873
LC		 2953/*
2954 * This is the function where ext4lazyinit thread lives. It walks
2955 * through the request list searching for next scheduled filesystem.
2956 * When such a fs is found, run the lazy initialization request
2957 * (ext4_rn_li_request) and keep track of the time spend in this
2958 * function. Based on that time we compute next schedule time of
2959 * the request. When walking through the list is complete, compute
2960 * next waking time and put itself into sleep.
2961 */
2962static int ext4_lazyinit_thread(void *arg)
2963{
2964 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2965 struct list_head *pos, *n;
2966 struct ext4_li_request *elr;
4ed5c033 2967 unsigned long next_wakeup, cur;
bfff6873
LC		 2968
2969 BUG_ON(NULL == eli);
2970
bfff6873
LC		 2971cont_thread:
2972 while (true) {
2973 next_wakeup = MAX_JIFFY_OFFSET;
2974
2975 mutex_lock(&eli->li_list_mtx);
2976 if (list_empty(&eli->li_request_list)) {
2977 mutex_unlock(&eli->li_list_mtx);
2978 goto exit_thread;
2979 }
2980
2981 list_for_each_safe(pos, n, &eli->li_request_list) {
2982 elr = list_entry(pos, struct ext4_li_request,
2983 lr_request);
2984
b2c78cd0
TT		 2985 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2986 if (ext4_run_li_request(elr) != 0) {
2987 /* error, remove the lazy_init job */
2988 ext4_remove_li_request(elr);
2989 continue;
2990 }
bfff6873
LC		 2991 }
2992
2993 if (time_before(elr->lr_next_sched, next_wakeup))
2994 next_wakeup = elr->lr_next_sched;
2995 }
2996 mutex_unlock(&eli->li_list_mtx);
2997
a0acae0e 2998 try_to_freeze();
bfff6873 2999
4ed5c033
LC		 3000 cur = jiffies;
3001 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3002 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3003 cond_resched();
3004 continue;
3005 }
3006
4ed5c033
LC		 3007 schedule_timeout_interruptible(next_wakeup - cur);
3008
8f1f7453
ES		 3009 if (kthread_should_stop()) {
3010 ext4_clear_request_list();
3011 goto exit_thread;
3012 }
bfff6873
LC		 3013 }
3014
3015exit_thread:
3016 /*
3017 * It looks like the request list is empty, but we need
3018 * to check it under the li_list_mtx lock, to prevent any
3019 * additions into it, and of course we should lock ext4_li_mtx
3020 * to atomically free the list and ext4_li_info, because at
3021 * this point another ext4 filesystem could be registering
3022 * new one.
3023 */
3024 mutex_lock(&ext4_li_mtx);
3025 mutex_lock(&eli->li_list_mtx);
3026 if (!list_empty(&eli->li_request_list)) {
3027 mutex_unlock(&eli->li_list_mtx);
3028 mutex_unlock(&ext4_li_mtx);
3029 goto cont_thread;
3030 }
3031 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3032 kfree(ext4_li_info);
3033 ext4_li_info = NULL;
3034 mutex_unlock(&ext4_li_mtx);
3035
3036 return 0;
3037}
3038
3039static void ext4_clear_request_list(void)
3040{
3041 struct list_head *pos, *n;
3042 struct ext4_li_request *elr;
3043
3044 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3045 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3046 elr = list_entry(pos, struct ext4_li_request,
3047 lr_request);
3048 ext4_remove_li_request(elr);
3049 }
3050 mutex_unlock(&ext4_li_info->li_list_mtx);
3051}
3052
3053static int ext4_run_lazyinit_thread(void)
3054{
8f1f7453
ES		 3055 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3056 ext4_li_info, "ext4lazyinit");
3057 if (IS_ERR(ext4_lazyinit_task)) {
3058 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3059 ext4_clear_request_list();
bfff6873
LC		 3060 kfree(ext4_li_info);
3061 ext4_li_info = NULL;
92b97816 3062 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3063 "initialization thread\n",
3064 err);
3065 return err;
3066 }
3067 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3068 return 0;
3069}
3070
3071/*
3072 * Check whether it make sense to run itable init. thread or not.
3073 * If there is at least one uninitialized inode table, return
3074 * corresponding group number, else the loop goes through all
3075 * groups and return total number of groups.
3076 */
3077static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3078{
3079 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3080 struct ext4_group_desc *gdp = NULL;
3081
3082 for (group = 0; group < ngroups; group++) {
3083 gdp = ext4_get_group_desc(sb, group, NULL);
3084 if (!gdp)
3085 continue;
3086
3087 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3088 break;
3089 }
3090
3091 return group;
3092}
3093
3094static int ext4_li_info_new(void)
3095{
3096 struct ext4_lazy_init *eli = NULL;
3097
3098 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3099 if (!eli)
3100 return -ENOMEM;
3101
bfff6873
LC		 3102 INIT_LIST_HEAD(&eli->li_request_list);
3103 mutex_init(&eli->li_list_mtx);
3104
bfff6873
LC		 3105 eli->li_state |= EXT4_LAZYINIT_QUIT;
3106
3107 ext4_li_info = eli;
3108
3109 return 0;
3110}
3111
3112static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3113 ext4_group_t start)
3114{
3115 struct ext4_sb_info *sbi = EXT4_SB(sb);
3116 struct ext4_li_request *elr;
bfff6873
LC		 3117
3118 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3119 if (!elr)
3120 return NULL;
3121
3122 elr->lr_super = sb;
3123 elr->lr_sbi = sbi;
3124 elr->lr_next_group = start;
3125
3126 /*
3127 * Randomize first schedule time of the request to
3128 * spread the inode table initialization requests
3129 * better.
3130 */
dd1f723b
TT		 3131 elr->lr_next_sched = jiffies + (prandom_u32() %
3132 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3133 return elr;
3134}
3135
7f511862
TT		 3136int ext4_register_li_request(struct super_block *sb,
3137 ext4_group_t first_not_zeroed)
bfff6873
LC		 3138{
3139 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3140 struct ext4_li_request *elr = NULL;
bfff6873 3141 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3142 int ret = 0;
bfff6873 3143
7f511862 3144 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3145 if (sbi->s_li_request != NULL) {
3146 /*
3147 * Reset timeout so it can be computed again, because
3148 * s_li_wait_mult might have changed.
3149 */
3150 sbi->s_li_request->lr_timeout = 0;
7f511862 3151 goto out;
51ce6511 3152 }
bfff6873
LC		 3153
3154 if (first_not_zeroed == ngroups ||
3155 (sb->s_flags & MS_RDONLY) ||
55ff3840 3156 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3157 goto out;
bfff6873
LC		 3158
3159 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3160 if (!elr) {
3161 ret = -ENOMEM;
3162 goto out;
3163 }
bfff6873
LC		 3164
3165 if (NULL == ext4_li_info) {
3166 ret = ext4_li_info_new();
3167 if (ret)
3168 goto out;
3169 }
3170
3171 mutex_lock(&ext4_li_info->li_list_mtx);
3172 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3173 mutex_unlock(&ext4_li_info->li_list_mtx);
3174
3175 sbi->s_li_request = elr;
46e4690b
TM		 3176 /*
3177 * set elr to NULL here since it has been inserted to
3178 * the request_list and the removal and free of it is
3179 * handled by ext4_clear_request_list from now on.
3180 */
3181 elr = NULL;
bfff6873
LC		 3182
3183 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3184 ret = ext4_run_lazyinit_thread();
3185 if (ret)
3186 goto out;
3187 }
bfff6873 3188out:
beed5ecb
NK		 3189 mutex_unlock(&ext4_li_mtx);
3190 if (ret)
bfff6873 3191 kfree(elr);
bfff6873
LC		 3192 return ret;
3193}
3194
3195/*
3196 * We do not need to lock anything since this is called on
3197 * module unload.
3198 */
3199static void ext4_destroy_lazyinit_thread(void)
3200{
3201 /*
3202 * If thread exited earlier
3203 * there's nothing to be done.
3204 */
8f1f7453 3205 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3206 return;
3207
8f1f7453 3208 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3209}
3210
25ed6e8a
DW		 3211static int set_journal_csum_feature_set(struct super_block *sb)
3212{
3213 int ret = 1;
3214 int compat, incompat;
3215 struct ext4_sb_info *sbi = EXT4_SB(sb);
3216
9aa5d32b 3217 if (ext4_has_metadata_csum(sb)) {
db9ee220 3218 /* journal checksum v3 */
25ed6e8a 3219 compat = 0;
db9ee220 3220 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3221 } else {
3222 /* journal checksum v1 */
3223 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3224 incompat = 0;
3225 }
3226
feb8c6d3
DW		 3227 jbd2_journal_clear_features(sbi->s_journal,
3228 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3229 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3230 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3231 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3232 ret = jbd2_journal_set_features(sbi->s_journal,
3233 compat, 0,
3234 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3235 incompat);
3236 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3237 ret = jbd2_journal_set_features(sbi->s_journal,
3238 compat, 0,
3239 incompat);
3240 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3241 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3242 } else {
feb8c6d3
DW		 3243 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3244 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3245 }
3246
3247 return ret;
3248}
3249
952fc18e
TT		 3250/*
3251 * Note: calculating the overhead so we can be compatible with
3252 * historical BSD practice is quite difficult in the face of
3253 * clusters/bigalloc. This is because multiple metadata blocks from
3254 * different block group can end up in the same allocation cluster.
3255 * Calculating the exact overhead in the face of clustered allocation
3256 * requires either O(all block bitmaps) in memory or O(number of block
3257 * groups**2) in time. We will still calculate the superblock for
3258 * older file systems --- and if we come across with a bigalloc file
3259 * system with zero in s_overhead_clusters the estimate will be close to
3260 * correct especially for very large cluster sizes --- but for newer
3261 * file systems, it's better to calculate this figure once at mkfs
3262 * time, and store it in the superblock. If the superblock value is
3263 * present (even for non-bigalloc file systems), we will use it.
3264 */
3265static int count_overhead(struct super_block *sb, ext4_group_t grp,
3266 char *buf)
3267{
3268 struct ext4_sb_info *sbi = EXT4_SB(sb);
3269 struct ext4_group_desc *gdp;
3270 ext4_fsblk_t first_block, last_block, b;
3271 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3272 int s, j, count = 0;
3273
0548bbb8
TT		 3274 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3275 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3276 sbi->s_itb_per_group + 2);
3277
952fc18e
TT		 3278 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3279 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3280 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3281 for (i = 0; i < ngroups; i++) {
3282 gdp = ext4_get_group_desc(sb, i, NULL);
3283 b = ext4_block_bitmap(sb, gdp);
3284 if (b >= first_block && b <= last_block) {
3285 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3286 count++;
3287 }
3288 b = ext4_inode_bitmap(sb, gdp);
3289 if (b >= first_block && b <= last_block) {
3290 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3291 count++;
3292 }
3293 b = ext4_inode_table(sb, gdp);
3294 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3295 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3296 int c = EXT4_B2C(sbi, b - first_block);
3297 ext4_set_bit(c, buf);
3298 count++;
3299 }
3300 if (i != grp)
3301 continue;
3302 s = 0;
3303 if (ext4_bg_has_super(sb, grp)) {
3304 ext4_set_bit(s++, buf);
3305 count++;
3306 }
3307 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3308 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3309 count++;
3310 }
3311 }
3312 if (!count)
3313 return 0;
3314 return EXT4_CLUSTERS_PER_GROUP(sb) -
3315 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3316}
3317
3318/*
3319 * Compute the overhead and stash it in sbi->s_overhead
3320 */
3321int ext4_calculate_overhead(struct super_block *sb)
3322{
3323 struct ext4_sb_info *sbi = EXT4_SB(sb);
3324 struct ext4_super_block *es = sbi->s_es;
3325 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3326 ext4_fsblk_t overhead = 0;
4fdb5543 3327 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3328
952fc18e
TT		 3329 if (!buf)
3330 return -ENOMEM;
3331
3332 /*
3333 * Compute the overhead (FS structures). This is constant
3334 * for a given filesystem unless the number of block groups
3335 * changes so we cache the previous value until it does.
3336 */
3337
3338 /*
3339 * All of the blocks before first_data_block are overhead
3340 */
3341 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3342
3343 /*
3344 * Add the overhead found in each block group
3345 */
3346 for (i = 0; i < ngroups; i++) {
3347 int blks;
3348
3349 blks = count_overhead(sb, i, buf);
3350 overhead += blks;
3351 if (blks)
3352 memset(buf, 0, PAGE_SIZE);
3353 cond_resched();
3354 }
b003b524
ES		 3355 /* Add the internal journal blocks as well */
3356 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3357 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3358
952fc18e
TT		 3359 sbi->s_overhead = overhead;
3360 smp_wmb();
3361 free_page((unsigned long) buf);
3362 return 0;
3363}
3364
27dd4385 3365
30fac0f7 3366static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
27dd4385
LC		 3367{
3368 ext4_fsblk_t resv_clusters;
3369
30fac0f7
JK		 3370 /*
3371 * There's no need to reserve anything when we aren't using extents.
3372 * The space estimates are exact, there are no unwritten extents,
3373 * hole punching doesn't need new metadata... This is needed especially
3374 * to keep ext2/3 backward compatibility.
3375 */
3376 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3377 return 0;
27dd4385
LC		 3378 /*
3379 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3380 * This should cover the situations where we can not afford to run
3381 * out of space like for example punch hole, or converting
556615dc 3382 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3383 * allocation would require 1, or 2 blocks, higher numbers are
3384 * very rare.
3385 */
30fac0f7
JK		 3386 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3387 EXT4_SB(sb)->s_cluster_bits;
27dd4385
LC		 3388
3389 do_div(resv_clusters, 50);
3390 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3391
3392 return resv_clusters;
3393}
3394
3395
3396static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3397{
3398 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3399 sbi->s_cluster_bits;
3400
3401 if (count >= clusters)
3402 return -EINVAL;
3403
3404 atomic64_set(&sbi->s_resv_clusters, count);
3405 return 0;
3406}
3407
2b2d6d01 3408static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3409{
d4c402d9 3410 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3411 struct buffer_head *bh;
617ba13b
MC		 3412 struct ext4_super_block *es = NULL;
3413 struct ext4_sb_info *sbi;
3414 ext4_fsblk_t block;
3415 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3416 ext4_fsblk_t logical_sb_block;
ac27a0ec 3417 unsigned long offset = 0;
ac27a0ec
DK		 3418 unsigned long journal_devnum = 0;
3419 unsigned long def_mount_opts;
3420 struct inode *root;
9f6200bb 3421 char *cp;
0390131b 3422 const char *descr;
dcc7dae3 3423 int ret = -ENOMEM;
281b5995 3424 int blocksize, clustersize;
4ec11028
TT		 3425 unsigned int db_count;
3426 unsigned int i;
281b5995 3427 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3428 __u64 blocks_count;
07aa2ea1 3429 int err = 0;
b3881f74 3430 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3431 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3432
3433 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3434 if (!sbi)
dcc7dae3 3435 goto out_free_orig;
705895b6
PE		 3436
3437 sbi->s_blockgroup_lock =
3438 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3439 if (!sbi->s_blockgroup_lock) {
3440 kfree(sbi);
dcc7dae3 3441 goto out_free_orig;
705895b6 3442 }
ac27a0ec 3443 sb->s_fs_info = sbi;
2c0544b2 3444 sbi->s_sb = sb;
240799cd 3445 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3446 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3447 if (sb->s_bdev->bd_part)
3448 sbi->s_sectors_written_start =
3449 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3450
9f6200bb
TT		 3451 /* Cleanup superblock name */
3452 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3453 *cp = '!';
3454
07aa2ea1 3455 /* -EINVAL is default */
dcc7dae3 3456 ret = -EINVAL;
617ba13b 3457 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3458 if (!blocksize) {
b31e1552 3459 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3460 goto out_fail;
3461 }
3462
3463 /*
617ba13b 3464 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3465 * block sizes. We need to calculate the offset from buffer start.
3466 */
617ba13b 3467 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3468 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3469 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3470 } else {
70bbb3e0 3471 logical_sb_block = sb_block;
ac27a0ec
DK		 3472 }
3473
a8ac900b 3474 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3475 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3476 goto out_fail;
3477 }
3478 /*
3479 * Note: s_es must be initialized as soon as possible because
617ba13b 3480 * some ext4 macro-instructions depend on its value
ac27a0ec 3481 */
2716b802 3482 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3483 sbi->s_es = es;
3484 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3485 if (sb->s_magic != EXT4_SUPER_MAGIC)
3486 goto cantfind_ext4;
afc32f7e 3487 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3488
feb0ab32
DW		 3489 /* Warn if metadata_csum and gdt_csum are both set. */
3490 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3491 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3492 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
363307e6 3493 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3494 "redundant flags; please run fsck.");
3495
d25425f8
DW		 3496 /* Check for a known checksum algorithm */
3497 if (!ext4_verify_csum_type(sb, es)) {
3498 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3499 "unknown checksum algorithm.");
3500 silent = 1;
3501 goto cantfind_ext4;
3502 }
3503
0441984a
DW		 3504 /* Load the checksum driver */
3505 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3506 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3507 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3508 if (IS_ERR(sbi->s_chksum_driver)) {
3509 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3510 ret = PTR_ERR(sbi->s_chksum_driver);
3511 sbi->s_chksum_driver = NULL;
3512 goto failed_mount;
3513 }
3514 }
3515
a9c47317
DW		 3516 /* Check superblock checksum */
3517 if (!ext4_superblock_csum_verify(sb, es)) {
3518 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3519 "invalid superblock checksum. Run e2fsck?");
3520 silent = 1;
3521 goto cantfind_ext4;
3522 }
3523
3524 /* Precompute checksum seed for all metadata */
9aa5d32b 3525 if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3526 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3527 sizeof(es->s_uuid));
3528
ac27a0ec
DK		 3529 /* Set defaults before we parse the mount options */
3530 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3531 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3532 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3533 set_opt(sb, DEBUG);
87f26807 3534 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3535 set_opt(sb, GRPID);
617ba13b 3536 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3537 set_opt(sb, NO_UID32);
ea663336 3538 /* xattr user namespace & acls are now defaulted on */
ea663336 3539 set_opt(sb, XATTR_USER);
03010a33 3540#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3541 set_opt(sb, POSIX_ACL);
2e7842b8 3542#endif
98c1a759
DW		 3543 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3544 if (ext4_has_metadata_csum(sb))
3545 set_opt(sb, JOURNAL_CHECKSUM);
3546
617ba13b 3547 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3548 set_opt(sb, JOURNAL_DATA);
617ba13b 3549 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3550 set_opt(sb, ORDERED_DATA);
617ba13b 3551 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3552 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3553
3554 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3555 set_opt(sb, ERRORS_PANIC);
bb4f397a 3556 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3557 set_opt(sb, ERRORS_CONT);
bb4f397a 3558 else
fd8c37ec 3559 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3560 /* block_validity enabled by default; disable with noblock_validity */
3561 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3562 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3563 set_opt(sb, DISCARD);
ac27a0ec 3564
08cefc7a
EB		 3565 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3566 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3567 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3568 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3569 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3570
8b67f04a 3571 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3572 set_opt(sb, BARRIER);
ac27a0ec 3573
dd919b98
AK		 3574 /*
3575 * enable delayed allocation by default
3576 * Use -o nodelalloc to turn it off
3577 */
bc0b75f7 3578 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3579 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3580 set_opt(sb, DELALLOC);
dd919b98 3581
51ce6511
LC		 3582 /*
3583 * set default s_li_wait_mult for lazyinit, for the case there is
3584 * no mount option specified.
3585 */
3586 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3587
8b67f04a 3588 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3589 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3590 ext4_msg(sb, KERN_WARNING,
3591 "failed to parse options in superblock: %s",
3592 sbi->s_es->s_mount_opts);
3593 }
5a916be1 3594 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3595 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3596 &journal_ioprio, 0))
ac27a0ec
DK		 3597 goto failed_mount;
3598
56889787
TT		 3599 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3600 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3601 "with data=journal disables delayed "
3602 "allocation and O_DIRECT support!\n");
3603 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3604 ext4_msg(sb, KERN_ERR, "can't mount with "
3605 "both data=journal and delalloc");
3606 goto failed_mount;
3607 }
3608 if (test_opt(sb, DIOREAD_NOLOCK)) {
3609 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3610 "both data=journal and dioread_nolock");
56889787
TT		 3611 goto failed_mount;
3612 }
923ae0ff
RZ		 3613 if (test_opt(sb, DAX)) {
3614 ext4_msg(sb, KERN_ERR, "can't mount with "
3615 "both data=journal and dax");
3616 goto failed_mount;
3617 }
56889787
TT		 3618 if (test_opt(sb, DELALLOC))
3619 clear_opt(sb, DELALLOC);
3620 }
3621
ac27a0ec 3622 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3623 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3624
617ba13b
MC		 3625 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3626 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3627 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3628 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3629 ext4_msg(sb, KERN_WARNING,
3630 "feature flags set on rev 0 fs, "
3631 "running e2fsck is recommended");
469108ff 3632
ed3654eb
TT		 3633 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3634 set_opt2(sb, HURD_COMPAT);
3635 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3636 EXT4_FEATURE_INCOMPAT_64BIT)) {
3637 ext4_msg(sb, KERN_ERR,
3638 "The Hurd can't support 64-bit file systems");
3639 goto failed_mount;
3640 }
3641 }
3642
2035e776
TT		 3643 if (IS_EXT2_SB(sb)) {
3644 if (ext2_feature_set_ok(sb))
3645 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3646 "using the ext4 subsystem");
3647 else {
3648 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3649 "to feature incompatibilities");
3650 goto failed_mount;
3651 }
3652 }
3653
3654 if (IS_EXT3_SB(sb)) {
3655 if (ext3_feature_set_ok(sb))
3656 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3657 "using the ext4 subsystem");
3658 else {
3659 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3660 "to feature incompatibilities");
3661 goto failed_mount;
3662 }
3663 }
3664
ac27a0ec
DK		 3665 /*
3666 * Check feature flags regardless of the revision level, since we
3667 * previously didn't change the revision level when setting the flags,
3668 * so there is a chance incompat flags are set on a rev 0 filesystem.
3669 */
a13fb1a4 3670 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3671 goto failed_mount;
a13fb1a4 3672
261cb20c 3673 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3674 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3675 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3676 ext4_msg(sb, KERN_ERR,
3677 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3678 goto failed_mount;
3679 }
3680
923ae0ff
RZ		 3681 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3682 if (blocksize != PAGE_SIZE) {
3683 ext4_msg(sb, KERN_ERR,
3684 "error: unsupported blocksize for dax");
3685 goto failed_mount;
3686 }
3687 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3688 ext4_msg(sb, KERN_ERR,
3689 "error: device does not support dax");
3690 goto failed_mount;
3691 }
3692 }
3693
ac27a0ec 3694 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3695 /* Validate the filesystem blocksize */
3696 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3697 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3698 blocksize);
ac27a0ec
DK		 3699 goto failed_mount;
3700 }
3701
2b2d6d01 3702 brelse(bh);
70bbb3e0
AM		 3703 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3704 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3705 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3706 if (!bh) {
b31e1552
ES		 3707 ext4_msg(sb, KERN_ERR,
3708 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3709 goto failed_mount;
3710 }
2716b802 3711 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3712 sbi->s_es = es;
617ba13b 3713 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3714 ext4_msg(sb, KERN_ERR,
3715 "Magic mismatch, very weird!");
ac27a0ec
DK		 3716 goto failed_mount;
3717 }
3718 }
3719
a13fb1a4
ES		 3720 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3721 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3722 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3723 has_huge_files);
3724 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3725
617ba13b
MC		 3726 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3727 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3728 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3729 } else {
3730 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3731 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3732 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3733 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3734 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3735 ext4_msg(sb, KERN_ERR,
3736 "unsupported inode size: %d",
2b2d6d01 3737 sbi->s_inode_size);
ac27a0ec
DK		 3738 goto failed_mount;
3739 }
ef7f3835
KS		 3740 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3741 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3742 }
0b8e58a1 3743
0d1ee42f
AR		 3744 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3745 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3746 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3747 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3748 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3749 ext4_msg(sb, KERN_ERR,
3750 "unsupported descriptor size %lu",
0d1ee42f
AR		 3751 sbi->s_desc_size);
3752 goto failed_mount;
3753 }
3754 } else
3755 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3756
ac27a0ec 3757 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3758 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3759 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3760 goto cantfind_ext4;
0b8e58a1 3761
617ba13b 3762 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3763 if (sbi->s_inodes_per_block == 0)
617ba13b 3764 goto cantfind_ext4;
ac27a0ec
DK		 3765 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3766 sbi->s_inodes_per_block;
0d1ee42f 3767 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3768 sbi->s_sbh = bh;
3769 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3770 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3771 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3772
2b2d6d01 3773 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3774 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3775 sbi->s_def_hash_version = es->s_def_hash_version;
23301410
TT		 3776 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3777 i = le32_to_cpu(es->s_flags);
3778 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3779 sbi->s_hash_unsigned = 3;
3780 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3781#ifdef __CHAR_UNSIGNED__
23301410
TT		 3782 if (!(sb->s_flags & MS_RDONLY))
3783 es->s_flags |=
3784 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3785 sbi->s_hash_unsigned = 3;
f99b2589 3786#else
23301410
TT		 3787 if (!(sb->s_flags & MS_RDONLY))
3788 es->s_flags |=
3789 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3790#endif
23301410 3791 }
f99b2589 3792 }
ac27a0ec 3793
281b5995
TT		 3794 /* Handle clustersize */
3795 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3796 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3797 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3798 if (has_bigalloc) {
3799 if (clustersize < blocksize) {
3800 ext4_msg(sb, KERN_ERR,
3801 "cluster size (%d) smaller than "
3802 "block size (%d)", clustersize, blocksize);
3803 goto failed_mount;
3804 }
3805 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3806 le32_to_cpu(es->s_log_block_size);
3807 sbi->s_clusters_per_group =
3808 le32_to_cpu(es->s_clusters_per_group);
3809 if (sbi->s_clusters_per_group > blocksize * 8) {
3810 ext4_msg(sb, KERN_ERR,
3811 "#clusters per group too big: %lu",
3812 sbi->s_clusters_per_group);
3813 goto failed_mount;
3814 }
3815 if (sbi->s_blocks_per_group !=
3816 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3817 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3818 "clusters per group (%lu) inconsistent",
3819 sbi->s_blocks_per_group,
3820 sbi->s_clusters_per_group);
3821 goto failed_mount;
3822 }
3823 } else {
3824 if (clustersize != blocksize) {
3825 ext4_warning(sb, "fragment/cluster size (%d) != "
3826 "block size (%d)", clustersize,
3827 blocksize);
3828 clustersize = blocksize;
3829 }
3830 if (sbi->s_blocks_per_group > blocksize * 8) {
3831 ext4_msg(sb, KERN_ERR,
3832 "#blocks per group too big: %lu",
3833 sbi->s_blocks_per_group);
3834 goto failed_mount;
3835 }
3836 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3837 sbi->s_cluster_bits = 0;
ac27a0ec 3838 }
281b5995
TT		 3839 sbi->s_cluster_ratio = clustersize / blocksize;
3840
ac27a0ec 3841 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3842 ext4_msg(sb, KERN_ERR,
3843 "#inodes per group too big: %lu",
2b2d6d01 3844 sbi->s_inodes_per_group);
ac27a0ec
DK		 3845 goto failed_mount;
3846 }
3847
960fd856
TT		 3848 /* Do we have standard group size of clustersize * 8 blocks ? */
3849 if (sbi->s_blocks_per_group == clustersize << 3)
3850 set_opt2(sb, STD_GROUP_SIZE);
3851
bf43d84b
ES		 3852 /*
3853 * Test whether we have more sectors than will fit in sector_t,
3854 * and whether the max offset is addressable by the page cache.
3855 */
5a9ae68a 3856 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3857 ext4_blocks_count(es));
5a9ae68a 3858 if (err) {
b31e1552 3859 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3860 " too large to mount safely on this system");
ac27a0ec 3861 if (sizeof(sector_t) < 8)
90c699a9 3862 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3863 goto failed_mount;
3864 }
3865
617ba13b
MC		 3866 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3867 goto cantfind_ext4;
e7c95593 3868
0f2ddca6
FTN		 3869 /* check blocks count against device size */
3870 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3871 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3872 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3873 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3874 ext4_blocks_count(es), blocks_count);
3875 goto failed_mount;
3876 }
3877
0b8e58a1
AD		 3878 /*
3879 * It makes no sense for the first data block to be beyond the end
3880 * of the filesystem.
3881 */
3882 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3883 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3884 "block %u is beyond end of filesystem (%llu)",
3885 le32_to_cpu(es->s_first_data_block),
3886 ext4_blocks_count(es));
e7c95593
ES		 3887 goto failed_mount;
3888 }
bd81d8ee
LV		 3889 blocks_count = (ext4_blocks_count(es) -
3890 le32_to_cpu(es->s_first_data_block) +
3891 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3892 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3893 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3894 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3895 "(block count %llu, first data block %u, "
b31e1552 3896 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3897 ext4_blocks_count(es),
3898 le32_to_cpu(es->s_first_data_block),
3899 EXT4_BLOCKS_PER_GROUP(sb));
3900 goto failed_mount;
3901 }
bd81d8ee 3902 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3903 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3904 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3905 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3906 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3907 sbi->s_group_desc = ext4_kvmalloc(db_count *
3908 sizeof(struct buffer_head *),
3909 GFP_KERNEL);
ac27a0ec 3910 if (sbi->s_group_desc == NULL) {
b31e1552 3911 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3912 ret = -ENOMEM;
ac27a0ec
DK		 3913 goto failed_mount;
3914 }
3915
9f6200bb
TT		 3916 if (ext4_proc_root)
3917 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3918
66acdcf4
TT		 3919 if (sbi->s_proc)
3920 proc_create_data("options", S_IRUGO, sbi->s_proc,
3921 &ext4_seq_options_fops, sb);
3922
705895b6 3923 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3924
3925 for (i = 0; i < db_count; i++) {
70bbb3e0 3926 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3927 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3928 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3929 ext4_msg(sb, KERN_ERR,
3930 "can't read group descriptor %d", i);
ac27a0ec
DK		 3931 db_count = i;
3932 goto failed_mount2;
3933 }
3934 }
bfff6873 3935 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3936 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
f9ae9cf5 3937 goto failed_mount2;
ac27a0ec 3938 }
772cb7c8 3939
f9ae9cf5 3940 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3941 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3942 spin_lock_init(&sbi->s_next_gen_lock);
3943
04ecddb7
JM		 3944 setup_timer(&sbi->s_err_report, print_daily_error_info,
3945 (unsigned long) sb);
04496411 3946
a75ae78f 3947 /* Register extent status tree shrinker */
eb68d0e2 3948 if (ext4_es_register_shrinker(sbi))
ce7e010a 3949 goto failed_mount3;
ce7e010a 3950
c9de560d 3951 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3952 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3953
f9ae9cf5
TT		 3954 /*
3955 * set up enough so that it can read an inode
3956 */
f6e63f90 3957 sb->s_op = &ext4_sops;
617ba13b
MC		 3958 sb->s_export_op = &ext4_export_ops;
3959 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3960#ifdef CONFIG_QUOTA
617ba13b 3961 sb->dq_op = &ext4_quota_operations;
262b4662 3962 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
1fa5efe3 3963 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3964 else
3965 sb->s_qcop = &ext4_qctl_operations;
96c7e0d9 3966 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
ac27a0ec 3967#endif
f2fa2ffc
AK		 3968 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3969
ac27a0ec 3970 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3971 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3972
3973 sb->s_root = NULL;
3974
3975 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3976 EXT4_HAS_INCOMPAT_FEATURE(sb,
3977 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3978
c5e06d10
JL		 3979 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3980 !(sb->s_flags & MS_RDONLY))
3981 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3982 goto failed_mount3a;
c5e06d10 3983
ac27a0ec
DK		 3984 /*
3985 * The first inode we look at is the journal inode. Don't try
3986 * root first: it may be modified in the journal!
3987 */
3988 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3989 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3990 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 3991 goto failed_mount3a;
0390131b
FM		 3992 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3993 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3994 ext4_msg(sb, KERN_ERR, "required journal recovery "
3995 "suppressed and not mounted read-only");
744692dc 3996 goto failed_mount_wq;
ac27a0ec 3997 } else {
fd8c37ec 3998 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3999 sbi->s_journal = NULL;
4000 needs_recovery = 0;
4001 goto no_journal;
ac27a0ec
DK		 4002 }
4003
f32aaf2d 4004 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 4005 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4006 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4007 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4008 goto failed_mount_wq;
eb40a09c
JS		 4009 }
4010
25ed6e8a
DW		 4011 if (!set_journal_csum_feature_set(sb)) {
4012 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4013 "feature set");
4014 goto failed_mount_wq;
d4da6c9c 4015 }
818d276c 4016
ac27a0ec
DK		 4017 /* We have now updated the journal if required, so we can
4018 * validate the data journaling mode. */
4019 switch (test_opt(sb, DATA_FLAGS)) {
4020 case 0:
4021 /* No mode set, assume a default based on the journal
63f57933
AM		 4022 * capabilities: ORDERED_DATA if the journal can
4023 * cope, else JOURNAL_DATA
4024 */
dab291af
MC		 4025 if (jbd2_journal_check_available_features
4026 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 4027 set_opt(sb, ORDERED_DATA);
ac27a0ec 4028 else
fd8c37ec 4029 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 4030 break;
4031
617ba13b
MC		 4032 case EXT4_MOUNT_ORDERED_DATA:
4033 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4034 if (!jbd2_journal_check_available_features
4035 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4036 ext4_msg(sb, KERN_ERR, "Journal does not support "
4037 "requested data journaling mode");
744692dc 4038 goto failed_mount_wq;
ac27a0ec
DK		 4039 }
4040 default:
4041 break;
4042 }
b3881f74 4043 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4044
18aadd47
BJ		 4045 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4046
ce7e010a 4047no_journal:
9c191f70
M		 4048 if (ext4_mballoc_ready) {
4049 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4050 if (!sbi->s_mb_cache) {
4051 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4052 goto failed_mount_wq;
4053 }
4054 }
4055
952fc18e
TT		 4056 /*
4057 * Get the # of file system overhead blocks from the
4058 * superblock if present.
4059 */
4060 if (es->s_overhead_clusters)
4061 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4062 else {
07aa2ea1
LC		 4063 err = ext4_calculate_overhead(sb);
4064 if (err)
952fc18e
TT		 4065 goto failed_mount_wq;
4066 }
4067
fd89d5f2
TH		 4068 /*
4069 * The maximum number of concurrent works can be high and
4070 * concurrency isn't really necessary. Limit it to 1.
4071 */
2e8fa54e
JK		 4072 EXT4_SB(sb)->rsv_conversion_wq =
4073 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4074 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4075 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4076 ret = -ENOMEM;
2e8fa54e
JK		 4077 goto failed_mount4;
4078 }
4079
ac27a0ec 4080 /*
dab291af 4081 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4082 * so we can safely mount the rest of the filesystem now.
4083 */
4084
1d1fe1ee
DH		 4085 root = ext4_iget(sb, EXT4_ROOT_INO);
4086 if (IS_ERR(root)) {
b31e1552 4087 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4088 ret = PTR_ERR(root);
32a9bb57 4089 root = NULL;
ac27a0ec
DK		 4090 goto failed_mount4;
4091 }
4092 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4093 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4094 iput(root);
ac27a0ec
DK		 4095 goto failed_mount4;
4096 }
48fde701 4097 sb->s_root = d_make_root(root);
1d1fe1ee 4098 if (!sb->s_root) {
b31e1552 4099 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4100 ret = -ENOMEM;
4101 goto failed_mount4;
4102 }
ac27a0ec 4103
7e84b621
ES		 4104 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4105 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4106
4107 /* determine the minimum size of new large inodes, if present */
4108 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4109 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4110 EXT4_GOOD_OLD_INODE_SIZE;
4111 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4112 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4113 if (sbi->s_want_extra_isize <
4114 le16_to_cpu(es->s_want_extra_isize))
4115 sbi->s_want_extra_isize =
4116 le16_to_cpu(es->s_want_extra_isize);
4117 if (sbi->s_want_extra_isize <
4118 le16_to_cpu(es->s_min_extra_isize))
4119 sbi->s_want_extra_isize =
4120 le16_to_cpu(es->s_min_extra_isize);
4121 }
4122 }
4123 /* Check if enough inode space is available */
4124 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4125 sbi->s_inode_size) {
4126 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4127 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4128 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4129 "available");
ef7f3835
KS		 4130 }
4131
30fac0f7 4132 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
27dd4385
LC		 4133 if (err) {
4134 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
30fac0f7 4135 "reserved pool", ext4_calculate_resv_clusters(sb));
f9ae9cf5 4136 goto failed_mount4a;
27dd4385
LC		 4137 }
4138
6fd058f7
TT		 4139 err = ext4_setup_system_zone(sb);
4140 if (err) {
b31e1552 4141 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4142 "zone (%d)", err);
f9ae9cf5
TT		 4143 goto failed_mount4a;
4144 }
4145
4146 ext4_ext_init(sb);
4147 err = ext4_mb_init(sb);
4148 if (err) {
4149 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4150 err);
dcf2d804 4151 goto failed_mount5;
c2774d84
AK		 4152 }
4153
d5e03cbb
TT		 4154 block = ext4_count_free_clusters(sb);
4155 ext4_free_blocks_count_set(sbi->s_es,
4156 EXT4_C2B(sbi, block));
908c7f19
TH		 4157 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4158 GFP_KERNEL);
d5e03cbb
TT		 4159 if (!err) {
4160 unsigned long freei = ext4_count_free_inodes(sb);
4161 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4162 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4163 GFP_KERNEL);
d5e03cbb
TT		 4164 }
4165 if (!err)
4166 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4167 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4168 if (!err)
908c7f19
TH		 4169 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4170 GFP_KERNEL);
d5e03cbb
TT		 4171 if (err) {
4172 ext4_msg(sb, KERN_ERR, "insufficient memory");
4173 goto failed_mount6;
4174 }
4175
4176 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4177 if (!ext4_fill_flex_info(sb)) {
4178 ext4_msg(sb, KERN_ERR,
4179 "unable to initialize "
4180 "flex_bg meta info!");
4181 goto failed_mount6;
4182 }
4183
bfff6873
LC		 4184 err = ext4_register_li_request(sb, first_not_zeroed);
4185 if (err)
dcf2d804 4186 goto failed_mount6;
bfff6873 4187
3197ebdb
TT		 4188 sbi->s_kobj.kset = ext4_kset;
4189 init_completion(&sbi->s_kobj_unregister);
4190 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4191 "%s", sb->s_id);
dcf2d804
TM		 4192 if (err)
4193 goto failed_mount7;
3197ebdb 4194
9b2ff357
JK		 4195#ifdef CONFIG_QUOTA
4196 /* Enable quota usage during mount. */
4197 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4198 !(sb->s_flags & MS_RDONLY)) {
4199 err = ext4_enable_quotas(sb);
4200 if (err)
4201 goto failed_mount8;
4202 }
4203#endif /* CONFIG_QUOTA */
4204
617ba13b
MC		 4205 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4206 ext4_orphan_cleanup(sb, es);
4207 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4208 if (needs_recovery) {
b31e1552 4209 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4210 ext4_mark_recovery_complete(sb, es);
4211 }
4212 if (EXT4_SB(sb)->s_journal) {
4213 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4214 descr = " journalled data mode";
4215 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4216 descr = " ordered data mode";
4217 else
4218 descr = " writeback data mode";
4219 } else
4220 descr = "out journal";
4221
79add3a3
LC		 4222 if (test_opt(sb, DISCARD)) {
4223 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4224 if (!blk_queue_discard(q))
4225 ext4_msg(sb, KERN_WARNING,
4226 "mounting with \"discard\" option, but "
4227 "the device does not support discard");
4228 }
4229
d4c402d9 4230 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4231 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4232 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4233
66e61a9e
TT		 4234 if (es->s_error_count)
4235 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4236
efbed4dc
TT		 4237 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4238 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4239 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4240 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4241
d4c402d9 4242 kfree(orig_data);
ac27a0ec
DK		 4243 return 0;
4244
617ba13b 4245cantfind_ext4:
ac27a0ec 4246 if (!silent)
b31e1552 4247 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4248 goto failed_mount;
4249
72ba7450
TT		 4250#ifdef CONFIG_QUOTA
4251failed_mount8:
4252 kobject_del(&sbi->s_kobj);
4253#endif
dcf2d804
TM		 4254failed_mount7:
4255 ext4_unregister_li_request(sb);
4256failed_mount6:
f9ae9cf5 4257 ext4_mb_release(sb);
d5e03cbb 4258 if (sbi->s_flex_groups)
b93b41d4 4259 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4260 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4261 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4262 percpu_counter_destroy(&sbi->s_dirs_counter);
4263 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4264failed_mount5:
f9ae9cf5
TT		 4265 ext4_ext_release(sb);
4266 ext4_release_system_zone(sb);
4267failed_mount4a:
94bf608a 4268 dput(sb->s_root);
32a9bb57 4269 sb->s_root = NULL;
94bf608a 4270failed_mount4:
b31e1552 4271 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4272 if (EXT4_SB(sb)->rsv_conversion_wq)
4273 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4274failed_mount_wq:
0390131b
FM		 4275 if (sbi->s_journal) {
4276 jbd2_journal_destroy(sbi->s_journal);
4277 sbi->s_journal = NULL;
4278 }
50460fe8 4279failed_mount3a:
d3922a77 4280 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4281failed_mount3:
9105bb14 4282 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4283 if (sbi->s_mmp_tsk)
4284 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4285failed_mount2:
4286 for (i = 0; i < db_count; i++)
4287 brelse(sbi->s_group_desc[i]);
b93b41d4 4288 kvfree(sbi->s_group_desc);
ac27a0ec 4289failed_mount:
0441984a
DW		 4290 if (sbi->s_chksum_driver)
4291 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4292 if (sbi->s_proc) {
66acdcf4 4293 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4294 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4295 }
ac27a0ec 4296#ifdef CONFIG_QUOTA
a2d4a646 4297 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4298 kfree(sbi->s_qf_names[i]);
4299#endif
617ba13b 4300 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4301 brelse(bh);
4302out_fail:
4303 sb->s_fs_info = NULL;
f6830165 4304 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4305 kfree(sbi);
dcc7dae3 4306out_free_orig:
d4c402d9 4307 kfree(orig_data);
07aa2ea1 4308 return err ? err : ret;
ac27a0ec
DK		 4309}
4310
4311/*
4312 * Setup any per-fs journal parameters now. We'll do this both on
4313 * initial mount, once the journal has been initialised but before we've
4314 * done any recovery; and again on any subsequent remount.
4315 */
617ba13b 4316static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4317{
617ba13b 4318 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4319
30773840
TT		 4320 journal->j_commit_interval = sbi->s_commit_interval;
4321 journal->j_min_batch_time = sbi->s_min_batch_time;
4322 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4323
a931da6a 4324 write_lock(&journal->j_state_lock);
ac27a0ec 4325 if (test_opt(sb, BARRIER))
dab291af 4326 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4327 else
dab291af 4328 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4329 if (test_opt(sb, DATA_ERR_ABORT))
4330 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4331 else
4332 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4333 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4334}
4335
617ba13b 4336static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4337 unsigned int journal_inum)
4338{
4339 struct inode *journal_inode;
4340 journal_t *journal;
4341
0390131b
FM		 4342 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4343
ac27a0ec
DK		 4344 /* First, test for the existence of a valid inode on disk. Bad
4345 * things happen if we iget() an unused inode, as the subsequent
4346 * iput() will try to delete it. */
4347
1d1fe1ee
DH		 4348 journal_inode = ext4_iget(sb, journal_inum);
4349 if (IS_ERR(journal_inode)) {
b31e1552 4350 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4351 return NULL;
4352 }
4353 if (!journal_inode->i_nlink) {
4354 make_bad_inode(journal_inode);
4355 iput(journal_inode);
b31e1552 4356 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4357 return NULL;
4358 }
4359
e5f8eab8 4360 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4361 journal_inode, journal_inode->i_size);
1d1fe1ee 4362 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4363 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4364 iput(journal_inode);
4365 return NULL;
4366 }
4367
dab291af 4368 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4369 if (!journal) {
b31e1552 4370 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4371 iput(journal_inode);
4372 return NULL;
4373 }
4374 journal->j_private = sb;
617ba13b 4375 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4376 return journal;
4377}
4378
617ba13b 4379static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4380 dev_t j_dev)
4381{
2b2d6d01 4382 struct buffer_head *bh;
ac27a0ec 4383 journal_t *journal;
617ba13b
MC		 4384 ext4_fsblk_t start;
4385 ext4_fsblk_t len;
ac27a0ec 4386 int hblock, blocksize;
617ba13b 4387 ext4_fsblk_t sb_block;
ac27a0ec 4388 unsigned long offset;
2b2d6d01 4389 struct ext4_super_block *es;
ac27a0ec
DK		 4390 struct block_device *bdev;
4391
0390131b
FM		 4392 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4393
b31e1552 4394 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4395 if (bdev == NULL)
4396 return NULL;
4397
ac27a0ec 4398 blocksize = sb->s_blocksize;
e1defc4f 4399 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4400 if (blocksize < hblock) {
b31e1552
ES		 4401 ext4_msg(sb, KERN_ERR,
4402 "blocksize too small for journal device");
ac27a0ec
DK		 4403 goto out_bdev;
4404 }
4405
617ba13b
MC		 4406 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4407 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4408 set_blocksize(bdev, blocksize);
4409 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4410 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4411 "external journal");
ac27a0ec
DK		 4412 goto out_bdev;
4413 }
4414
2716b802 4415 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4416 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4417 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4418 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4419 ext4_msg(sb, KERN_ERR, "external journal has "
4420 "bad superblock");
ac27a0ec
DK		 4421 brelse(bh);
4422 goto out_bdev;
4423 }
4424
df4763be
DW		 4425 if ((le32_to_cpu(es->s_feature_ro_compat) &
4426 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4427 es->s_checksum != ext4_superblock_csum(sb, es)) {
4428 ext4_msg(sb, KERN_ERR, "external journal has "
4429 "corrupt superblock");
4430 brelse(bh);
4431 goto out_bdev;
4432 }
4433
617ba13b 4434 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4435 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4436 brelse(bh);
4437 goto out_bdev;
4438 }
4439
bd81d8ee 4440 len = ext4_blocks_count(es);
ac27a0ec
DK		 4441 start = sb_block + 1;
4442 brelse(bh); /* we're done with the superblock */
4443
dab291af 4444 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4445 start, len, blocksize);
4446 if (!journal) {
b31e1552 4447 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4448 goto out_bdev;
4449 }
4450 journal->j_private = sb;
9f203507 4451 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4452 wait_on_buffer(journal->j_sb_buffer);
4453 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4454 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4455 goto out_journal;
4456 }
4457 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4458 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4459 "user (unsupported) - %d",
ac27a0ec
DK		 4460 be32_to_cpu(journal->j_superblock->s_nr_users));
4461 goto out_journal;
4462 }
617ba13b
MC		 4463 EXT4_SB(sb)->journal_bdev = bdev;
4464 ext4_init_journal_params(sb, journal);
ac27a0ec 4465 return journal;
0b8e58a1 4466
ac27a0ec 4467out_journal:
dab291af 4468 jbd2_journal_destroy(journal);
ac27a0ec 4469out_bdev:
617ba13b 4470 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4471 return NULL;
4472}
4473
617ba13b
MC		 4474static int ext4_load_journal(struct super_block *sb,
4475 struct ext4_super_block *es,
ac27a0ec
DK		 4476 unsigned long journal_devnum)
4477{
4478 journal_t *journal;
4479 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4480 dev_t journal_dev;
4481 int err = 0;
4482 int really_read_only;
4483
0390131b
FM		 4484 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4485
ac27a0ec
DK		 4486 if (journal_devnum &&
4487 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4488 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4489 "numbers have changed");
ac27a0ec
DK		 4490 journal_dev = new_decode_dev(journal_devnum);
4491 } else
4492 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4493
4494 really_read_only = bdev_read_only(sb->s_bdev);
4495
4496 /*
4497 * Are we loading a blank journal or performing recovery after a
4498 * crash? For recovery, we need to check in advance whether we
4499 * can get read-write access to the device.
4500 */
617ba13b 4501 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4502 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4503 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4504 "required on readonly filesystem");
ac27a0ec 4505 if (really_read_only) {
b31e1552
ES		 4506 ext4_msg(sb, KERN_ERR, "write access "
4507 "unavailable, cannot proceed");
ac27a0ec
DK		 4508 return -EROFS;
4509 }
b31e1552
ES		 4510 ext4_msg(sb, KERN_INFO, "write access will "
4511 "be enabled during recovery");
ac27a0ec
DK		 4512 }
4513 }
4514
4515 if (journal_inum && journal_dev) {
b31e1552
ES		 4516 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4517 "and inode journals!");
ac27a0ec
DK		 4518 return -EINVAL;
4519 }
4520
4521 if (journal_inum) {
617ba13b 4522 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4523 return -EINVAL;
4524 } else {
617ba13b 4525 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4526 return -EINVAL;
4527 }
4528
90576c0b 4529 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4530 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4531
617ba13b 4532 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4533 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4534 if (!err) {
4535 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4536 if (save)
4537 memcpy(save, ((char *) es) +
4538 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4539 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4540 if (save)
4541 memcpy(((char *) es) + EXT4_S_ERR_START,
4542 save, EXT4_S_ERR_LEN);
4543 kfree(save);
4544 }
ac27a0ec
DK		 4545
4546 if (err) {
b31e1552 4547 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4548 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4549 return err;
4550 }
4551
617ba13b
MC		 4552 EXT4_SB(sb)->s_journal = journal;
4553 ext4_clear_journal_err(sb, es);
ac27a0ec 4554
c41303ce 4555 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4556 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4557 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4558
4559 /* Make sure we flush the recovery flag to disk. */
e2d67052 4560 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4561 }
4562
4563 return 0;
4564}
4565
e2d67052 4566static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4567{
e2d67052 4568 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4569 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4570 int error = 0;
ac27a0ec 4571
7c2e7087 4572 if (!sbh || block_device_ejected(sb))
c4be0c1d 4573 return error;
914258bf
TT		 4574 if (buffer_write_io_error(sbh)) {
4575 /*
4576 * Oh, dear. A previous attempt to write the
4577 * superblock failed. This could happen because the
4578 * USB device was yanked out. Or it could happen to
4579 * be a transient write error and maybe the block will
4580 * be remapped. Nothing we can do but to retry the
4581 * write and hope for the best.
4582 */
b31e1552
ES		 4583 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4584 "superblock detected");
914258bf
TT		 4585 clear_buffer_write_io_error(sbh);
4586 set_buffer_uptodate(sbh);
4587 }
71290b36
TT		 4588 /*
4589 * If the file system is mounted read-only, don't update the
4590 * superblock write time. This avoids updating the superblock
4591 * write time when we are mounting the root file system
4592 * read/only but we need to replay the journal; at that point,
4593 * for people who are east of GMT and who make their clock
4594 * tick in localtime for Windows bug-for-bug compatibility,
4595 * the clock is set in the future, and this will cause e2fsck
4596 * to complain and force a full file system check.
4597 */
4598 if (!(sb->s_flags & MS_RDONLY))
4599 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4600 if (sb->s_bdev->bd_part)
4601 es->s_kbytes_written =
4602 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4603 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4604 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4605 else
4606 es->s_kbytes_written =
4607 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4608 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4609 ext4_free_blocks_count_set(es,
57042651
TT		 4610 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4611 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4612 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4613 es->s_free_inodes_count =
4614 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4615 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4616 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4617 ext4_superblock_csum_set(sb);
ac27a0ec 4618 mark_buffer_dirty(sbh);
914258bf 4619 if (sync) {
c4be0c1d
TS		 4620 error = sync_dirty_buffer(sbh);
4621 if (error)
4622 return error;
4623
4624 error = buffer_write_io_error(sbh);
4625 if (error) {
b31e1552
ES		 4626 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4627 "superblock");
914258bf
TT		 4628 clear_buffer_write_io_error(sbh);
4629 set_buffer_uptodate(sbh);
4630 }
4631 }
c4be0c1d 4632 return error;
ac27a0ec
DK		 4633}
4634
ac27a0ec
DK		 4635/*
4636 * Have we just finished recovery? If so, and if we are mounting (or
4637 * remounting) the filesystem readonly, then we will end up with a
4638 * consistent fs on disk. Record that fact.
4639 */
2b2d6d01
TT		 4640static void ext4_mark_recovery_complete(struct super_block *sb,
4641 struct ext4_super_block *es)
ac27a0ec 4642{
617ba13b 4643 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4644
0390131b
FM		 4645 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4646 BUG_ON(journal != NULL);
4647 return;
4648 }
dab291af 4649 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4650 if (jbd2_journal_flush(journal) < 0)
4651 goto out;
4652
617ba13b 4653 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4654 sb->s_flags & MS_RDONLY) {
617ba13b 4655 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4656 ext4_commit_super(sb, 1);
ac27a0ec 4657 }
7ffe1ea8
HK		 4658
4659out:
dab291af 4660 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4661}
4662
4663/*
4664 * If we are mounting (or read-write remounting) a filesystem whose journal
4665 * has recorded an error from a previous lifetime, move that error to the
4666 * main filesystem now.
4667 */
2b2d6d01
TT		 4668static void ext4_clear_journal_err(struct super_block *sb,
4669 struct ext4_super_block *es)
ac27a0ec
DK		 4670{
4671 journal_t *journal;
4672 int j_errno;
4673 const char *errstr;
4674
0390131b
FM		 4675 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4676
617ba13b 4677 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4678
4679 /*
4680 * Now check for any error status which may have been recorded in the
617ba13b 4681 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4682 */
4683
dab291af 4684 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4685 if (j_errno) {
4686 char nbuf[16];
4687
617ba13b 4688 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4689 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4690 "from previous mount: %s", errstr);
12062ddd 4691 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4692
617ba13b
MC		 4693 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4694 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4695 ext4_commit_super(sb, 1);
ac27a0ec 4696
dab291af 4697 jbd2_journal_clear_err(journal);
d796c52e 4698 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4699 }
4700}
4701
4702/*
4703 * Force the running and committing transactions to commit,
4704 * and wait on the commit.
4705 */
617ba13b 4706int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4707{
4708 journal_t *journal;
ac27a0ec
DK		 4709
4710 if (sb->s_flags & MS_RDONLY)
4711 return 0;
4712
617ba13b 4713 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4714 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4715}
4716
617ba13b 4717static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4718{
14ce0cb4 4719 int ret = 0;
9eddacf9 4720 tid_t target;
06a407f1 4721 bool needs_barrier = false;
8d5d02e6 4722 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4723
9bffad1e 4724 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4725 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4726 /*
4727 * Writeback quota in non-journalled quota case - journalled quota has
4728 * no dirty dquots
4729 */
4730 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4731 /*
4732 * Data writeback is possible w/o journal transaction, so barrier must
4733 * being sent at the end of the function. But we can skip it if
4734 * transaction_commit will do it for us.
4735 */
bda32530
TT		 4736 if (sbi->s_journal) {
4737 target = jbd2_get_latest_transaction(sbi->s_journal);
4738 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4739 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4740 needs_barrier = true;
4741
4742 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4743 if (wait)
4744 ret = jbd2_log_wait_commit(sbi->s_journal,
4745 target);
4746 }
4747 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4748 needs_barrier = true;
06a407f1
DM		 4749 if (needs_barrier) {
4750 int err;
4751 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4752 if (!ret)
4753 ret = err;
0390131b 4754 }
06a407f1
DM		 4755
4756 return ret;
4757}
4758
ac27a0ec
DK		 4759/*
4760 * LVM calls this function before a (read-only) snapshot is created. This
4761 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4762 *
4763 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4764 * state independently. It relies on upper layer to stop all data & metadata
4765 * modifications.
ac27a0ec 4766 */
c4be0c1d 4767static int ext4_freeze(struct super_block *sb)
ac27a0ec 4768{
c4be0c1d
TS		 4769 int error = 0;
4770 journal_t *journal;
ac27a0ec 4771
9ca92389
TT		 4772 if (sb->s_flags & MS_RDONLY)
4773 return 0;
ac27a0ec 4774
9ca92389 4775 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4776
bb044576
TT		 4777 if (journal) {
4778 /* Now we set up the journal barrier. */
4779 jbd2_journal_lock_updates(journal);
ac27a0ec 4780
bb044576
TT		 4781 /*
4782 * Don't clear the needs_recovery flag if we failed to
4783 * flush the journal.
4784 */
4785 error = jbd2_journal_flush(journal);
4786 if (error < 0)
4787 goto out;
4788 }
9ca92389
TT		 4789
4790 /* Journal blocked and flushed, clear needs_recovery flag. */
4791 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4792 error = ext4_commit_super(sb, 1);
6b0310fb 4793out:
bb044576
TT		 4794 if (journal)
4795 /* we rely on upper layer to stop further updates */
4796 jbd2_journal_unlock_updates(journal);
6b0310fb 4797 return error;
ac27a0ec
DK		 4798}
4799
4800/*
4801 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4802 * flag here, even though the filesystem is not technically dirty yet.
4803 */
c4be0c1d 4804static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4805{
9ca92389
TT		 4806 if (sb->s_flags & MS_RDONLY)
4807 return 0;
4808
9ca92389
TT		 4809 /* Reset the needs_recovery flag before the fs is unlocked. */
4810 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4811 ext4_commit_super(sb, 1);
c4be0c1d 4812 return 0;
ac27a0ec
DK		 4813}
4814
673c6100
TT		 4815/*
4816 * Structure to save mount options for ext4_remount's benefit
4817 */
4818struct ext4_mount_options {
4819 unsigned long s_mount_opt;
a2595b8a 4820 unsigned long s_mount_opt2;
08cefc7a
EB		 4821 kuid_t s_resuid;
4822 kgid_t s_resgid;
673c6100
TT		 4823 unsigned long s_commit_interval;
4824 u32 s_min_batch_time, s_max_batch_time;
4825#ifdef CONFIG_QUOTA
4826 int s_jquota_fmt;
a2d4a646 4827 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4828#endif
4829};
4830
2b2d6d01 4831static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4832{
2b2d6d01 4833 struct ext4_super_block *es;
617ba13b 4834 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4835 unsigned long old_sb_flags;
617ba13b 4836 struct ext4_mount_options old_opts;
c79d967d 4837 int enable_quota = 0;
8a266467 4838 ext4_group_t g;
b3881f74 4839 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4840 int err = 0;
ac27a0ec 4841#ifdef CONFIG_QUOTA
03dafb5f 4842 int i, j;
ac27a0ec 4843#endif
d4c402d9 4844 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4845
4846 /* Store the original options */
4847 old_sb_flags = sb->s_flags;
4848 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4849 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4850 old_opts.s_resuid = sbi->s_resuid;
4851 old_opts.s_resgid = sbi->s_resgid;
4852 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4853 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4854 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4855#ifdef CONFIG_QUOTA
4856 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4857 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4858 if (sbi->s_qf_names[i]) {
4859 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4860 GFP_KERNEL);
4861 if (!old_opts.s_qf_names[i]) {
4862 for (j = 0; j < i; j++)
4863 kfree(old_opts.s_qf_names[j]);
3e36a163 4864 kfree(orig_data);
03dafb5f
CG		 4865 return -ENOMEM;
4866 }
4867 } else
4868 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4869#endif
b3881f74
TT		 4870 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4871 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4872
661aa520 4873 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4874 err = -EINVAL;
4875 goto restore_opts;
4876 }
4877
6b992ff2 4878 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4879 test_opt(sb, JOURNAL_CHECKSUM)) {
4880 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4881 "during remount not supported; ignoring");
4882 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4883 }
4884
6ae6514b
PS		 4885 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4886 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4887 ext4_msg(sb, KERN_ERR, "can't mount with "
4888 "both data=journal and delalloc");
4889 err = -EINVAL;
4890 goto restore_opts;
4891 }
4892 if (test_opt(sb, DIOREAD_NOLOCK)) {
4893 ext4_msg(sb, KERN_ERR, "can't mount with "
4894 "both data=journal and dioread_nolock");
4895 err = -EINVAL;
4896 goto restore_opts;
4897 }
923ae0ff
RZ		 4898 if (test_opt(sb, DAX)) {
4899 ext4_msg(sb, KERN_ERR, "can't mount with "
4900 "both data=journal and dax");
4901 err = -EINVAL;
4902 goto restore_opts;
4903 }
4904 }
4905
4906 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4907 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4908 "dax flag with busy inodes while remounting");
4909 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 4910 }
4911
4ab2f15b 4912 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4913 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4914
4915 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4916 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4917
4918 es = sbi->s_es;
4919
b3881f74 4920 if (sbi->s_journal) {
0390131b 4921 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4922 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4923 }
ac27a0ec 4924
661aa520 4925 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4926 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4927 err = -EROFS;
4928 goto restore_opts;
4929 }
4930
4931 if (*flags & MS_RDONLY) {
38c03b34
TT		 4932 err = sync_filesystem(sb);
4933 if (err < 0)
4934 goto restore_opts;
0f0dd62f
CH		 4935 err = dquot_suspend(sb, -1);
4936 if (err < 0)
c79d967d 4937 goto restore_opts;
c79d967d 4938
ac27a0ec
DK		 4939 /*
4940 * First of all, the unconditional stuff we have to do
4941 * to disable replay of the journal when we next remount
4942 */
4943 sb->s_flags |= MS_RDONLY;
4944
4945 /*
4946 * OK, test if we are remounting a valid rw partition
4947 * readonly, and if so set the rdonly flag and then
4948 * mark the partition as valid again.
4949 */
617ba13b
MC		 4950 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4951 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4952 es->s_state = cpu_to_le16(sbi->s_mount_state);
4953
a63c9eb2 4954 if (sbi->s_journal)
0390131b 4955 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4956 } else {
a13fb1a4 4957 /* Make sure we can mount this feature set readwrite */
2cb5cc8b
DW		 4958 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4959 EXT4_FEATURE_RO_COMPAT_READONLY) ||
4960 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4961 err = -EROFS;
4962 goto restore_opts;
4963 }
8a266467
TT		 4964 /*
4965 * Make sure the group descriptor checksums
0b8e58a1 4966 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4967 */
4968 for (g = 0; g < sbi->s_groups_count; g++) {
4969 struct ext4_group_desc *gdp =
4970 ext4_get_group_desc(sb, g, NULL);
4971
feb0ab32 4972 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4973 ext4_msg(sb, KERN_ERR,
4974 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4975 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4976 le16_to_cpu(gdp->bg_checksum));
4977 err = -EINVAL;
4978 goto restore_opts;
4979 }
4980 }
4981
ead6596b
ES		 4982 /*
4983 * If we have an unprocessed orphan list hanging
4984 * around from a previously readonly bdev mount,
4985 * require a full umount/remount for now.
4986 */
4987 if (es->s_last_orphan) {
b31e1552 4988 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4989 "remount RDWR because of unprocessed "
4990 "orphan inode list. Please "
b31e1552 4991 "umount/remount instead");
ead6596b
ES		 4992 err = -EINVAL;
4993 goto restore_opts;
4994 }
4995
ac27a0ec
DK		 4996 /*
4997 * Mounting a RDONLY partition read-write, so reread
4998 * and store the current valid flag. (It may have
4999 * been changed by e2fsck since we originally mounted
5000 * the partition.)
5001 */
0390131b
FM		 5002 if (sbi->s_journal)
5003 ext4_clear_journal_err(sb, es);
ac27a0ec 5004 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 5005 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 5006 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 5007 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
5008 EXT4_FEATURE_INCOMPAT_MMP))
5009 if (ext4_multi_mount_protect(sb,
5010 le64_to_cpu(es->s_mmp_block))) {
5011 err = -EROFS;
5012 goto restore_opts;
5013 }
c79d967d 5014 enable_quota = 1;
ac27a0ec
DK		 5015 }
5016 }
bfff6873
LC		 5017
5018 /*
5019 * Reinitialize lazy itable initialization thread based on
5020 * current settings
5021 */
5022 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5023 ext4_unregister_li_request(sb);
5024 else {
5025 ext4_group_t first_not_zeroed;
5026 first_not_zeroed = ext4_has_uninit_itable(sb);
5027 ext4_register_li_request(sb, first_not_zeroed);
5028 }
5029
6fd058f7 5030 ext4_setup_system_zone(sb);
d096ad0f 5031 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 5032 ext4_commit_super(sb, 1);
0390131b 5033
ac27a0ec
DK		 5034#ifdef CONFIG_QUOTA
5035 /* Release old quota file names */
a2d4a646 5036 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5037 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5038 if (enable_quota) {
5039 if (sb_any_quota_suspended(sb))
5040 dquot_resume(sb, -1);
5041 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5042 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5043 err = ext4_enable_quotas(sb);
07724f98 5044 if (err)
7c319d32 5045 goto restore_opts;
7c319d32
AK		 5046 }
5047 }
ac27a0ec 5048#endif
d4c402d9 5049
a26f4992 5050 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 5051 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5052 kfree(orig_data);
ac27a0ec 5053 return 0;
0b8e58a1 5054
ac27a0ec
DK		 5055restore_opts:
5056 sb->s_flags = old_sb_flags;
5057 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5058 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5059 sbi->s_resuid = old_opts.s_resuid;
5060 sbi->s_resgid = old_opts.s_resgid;
5061 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5062 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5063 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5064#ifdef CONFIG_QUOTA
5065 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5066 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5067 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5068 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5069 }
5070#endif
d4c402d9 5071 kfree(orig_data);
ac27a0ec
DK		 5072 return err;
5073}
5074
2b2d6d01 5075static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5076{
5077 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5078 struct ext4_sb_info *sbi = EXT4_SB(sb);
5079 struct ext4_super_block *es = sbi->s_es;
27dd4385 5080 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5081 u64 fsid;
d02a9391 5082 s64 bfree;
27dd4385 5083 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5084
952fc18e
TT		 5085 if (!test_opt(sb, MINIX_DF))
5086 overhead = sbi->s_overhead;
ac27a0ec 5087
617ba13b 5088 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5089 buf->f_bsize = sb->s_blocksize;
b72f78cb 5090 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5091 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5092 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5093 /* prevent underflow in case that few free space is available */
57042651 5094 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5095 buf->f_bavail = buf->f_bfree -
5096 (ext4_r_blocks_count(es) + resv_blocks);
5097 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5098 buf->f_bavail = 0;
5099 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5100 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5101 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5102 fsid = le64_to_cpup((void *)es->s_uuid) ^
5103 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5104 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5105 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5106
ac27a0ec
DK		 5107 return 0;
5108}
5109
0b8e58a1
AD		 5110/* Helper function for writing quotas on sync - we need to start transaction
5111 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 5112 * Process 1 Process 2
617ba13b 5113 * ext4_create() quota_sync()
a269eb18 5114 * jbd2_journal_start() write_dquot()
871a2931 5115 * dquot_initialize() down(dqio_mutex)
dab291af 5116 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 5117 *
5118 */
5119
5120#ifdef CONFIG_QUOTA
5121
5122static inline struct inode *dquot_to_inode(struct dquot *dquot)
5123{
4c376dca 5124 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5125}
5126
617ba13b 5127static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5128{
5129 int ret, err;
5130 handle_t *handle;
5131 struct inode *inode;
5132
5133 inode = dquot_to_inode(dquot);
9924a92a 5134 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5135 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5136 if (IS_ERR(handle))
5137 return PTR_ERR(handle);
5138 ret = dquot_commit(dquot);
617ba13b 5139 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5140 if (!ret)
5141 ret = err;
5142 return ret;
5143}
5144
617ba13b 5145static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5146{
5147 int ret, err;
5148 handle_t *handle;
5149
9924a92a 5150 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5151 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5152 if (IS_ERR(handle))
5153 return PTR_ERR(handle);
5154 ret = dquot_acquire(dquot);
617ba13b 5155 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5156 if (!ret)
5157 ret = err;
5158 return ret;
5159}
5160
617ba13b 5161static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5162{
5163 int ret, err;
5164 handle_t *handle;
5165
9924a92a 5166 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5167 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5168 if (IS_ERR(handle)) {
5169 /* Release dquot anyway to avoid endless cycle in dqput() */
5170 dquot_release(dquot);
ac27a0ec 5171 return PTR_ERR(handle);
9c3013e9 5172 }
ac27a0ec 5173 ret = dquot_release(dquot);
617ba13b 5174 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5175 if (!ret)
5176 ret = err;
5177 return ret;
5178}
5179
617ba13b 5180static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5181{
262b4662
JK		 5182 struct super_block *sb = dquot->dq_sb;
5183 struct ext4_sb_info *sbi = EXT4_SB(sb);
5184
2c8be6b2 5185 /* Are we journaling quotas? */
262b4662
JK		 5186 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5187 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5188 dquot_mark_dquot_dirty(dquot);
617ba13b 5189 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5190 } else {
5191 return dquot_mark_dquot_dirty(dquot);
5192 }
5193}
5194
617ba13b 5195static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5196{
5197 int ret, err;
5198 handle_t *handle;
5199
5200 /* Data block + inode block */
9924a92a 5201 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5202 if (IS_ERR(handle))
5203 return PTR_ERR(handle);
5204 ret = dquot_commit_info(sb, type);
617ba13b 5205 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5206 if (!ret)
5207 ret = err;
5208 return ret;
5209}
5210
5211/*
5212 * Turn on quotas during mount time - we need to find
5213 * the quota file and such...
5214 */
617ba13b 5215static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5216{
287a8095
CH		 5217 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5218 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5219}
5220
5221/*
5222 * Standard function to be called on quota_on
5223 */
617ba13b 5224static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5225 struct path *path)
ac27a0ec
DK		 5226{
5227 int err;
ac27a0ec
DK		 5228
5229 if (!test_opt(sb, QUOTA))
5230 return -EINVAL;
0623543b 5231
ac27a0ec 5232 /* Quotafile not on the same filesystem? */
d8c9584e 5233 if (path->dentry->d_sb != sb)
ac27a0ec 5234 return -EXDEV;
0623543b
JK		 5235 /* Journaling quota? */
5236 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5237 /* Quotafile not in fs root? */
f00c9e44 5238 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5239 ext4_msg(sb, KERN_WARNING,
5240 "Quota file not on filesystem root. "
5241 "Journaled quota will not work");
2b2d6d01 5242 }
0623543b
JK		 5243
5244 /*
5245 * When we journal data on quota file, we have to flush journal to see
5246 * all updates to the file when we bypass pagecache...
5247 */
0390131b 5248 if (EXT4_SB(sb)->s_journal &&
f00c9e44 5249 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 5250 /*
5251 * We don't need to lock updates but journal_flush() could
5252 * otherwise be livelocked...
5253 */
5254 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5255 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5256 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5257 if (err)
7ffe1ea8 5258 return err;
0623543b
JK		 5259 }
5260
f00c9e44 5261 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 5262}
5263
7c319d32
AK		 5264static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5265 unsigned int flags)
5266{
5267 int err;
5268 struct inode *qf_inode;
a2d4a646 5269 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5270 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5271 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5272 };
5273
5274 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5275
5276 if (!qf_inums[type])
5277 return -EPERM;
5278
5279 qf_inode = ext4_iget(sb, qf_inums[type]);
5280 if (IS_ERR(qf_inode)) {
5281 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5282 return PTR_ERR(qf_inode);
5283 }
5284
bcb13850
JK		 5285 /* Don't account quota for quota files to avoid recursion */
5286 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5287 err = dquot_enable(qf_inode, type, format_id, flags);
5288 iput(qf_inode);
5289
5290 return err;
5291}
5292
5293/* Enable usage tracking for all quota types. */
5294static int ext4_enable_quotas(struct super_block *sb)
5295{
5296 int type, err = 0;
a2d4a646 5297 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5298 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5299 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5300 };
5301
5302 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5303 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5304 if (qf_inums[type]) {
5305 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5306 DQUOT_USAGE_ENABLED);
5307 if (err) {
5308 ext4_warning(sb,
72ba7450
TT		 5309 "Failed to enable quota tracking "
5310 "(type=%d, err=%d). Please run "
5311 "e2fsck to fix.", type, err);
7c319d32
AK		 5312 return err;
5313 }
5314 }
5315 }
5316 return 0;
5317}
5318
ca0e05e4
DM		 5319static int ext4_quota_off(struct super_block *sb, int type)
5320{
21f97697
JK		 5321 struct inode *inode = sb_dqopt(sb)->files[type];
5322 handle_t *handle;
5323
87009d86
DM		 5324 /* Force all delayed allocation blocks to be allocated.
5325 * Caller already holds s_umount sem */
5326 if (test_opt(sb, DELALLOC))
ca0e05e4 5327 sync_filesystem(sb);
ca0e05e4 5328
0b268590
AG		 5329 if (!inode)
5330 goto out;
5331
21f97697
JK		 5332 /* Update modification times of quota files when userspace can
5333 * start looking at them */
9924a92a 5334 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5335 if (IS_ERR(handle))
5336 goto out;
5337 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5338 ext4_mark_inode_dirty(handle, inode);
5339 ext4_journal_stop(handle);
5340
5341out:
ca0e05e4
DM		 5342 return dquot_quota_off(sb, type);
5343}
5344
ac27a0ec
DK		 5345/* Read data from quotafile - avoid pagecache and such because we cannot afford
5346 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5347 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5348 * we don't have to be afraid of races */
617ba13b 5349static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5350 size_t len, loff_t off)
5351{
5352 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5353 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5354 int offset = off & (sb->s_blocksize - 1);
5355 int tocopy;
5356 size_t toread;
5357 struct buffer_head *bh;
5358 loff_t i_size = i_size_read(inode);
5359
5360 if (off > i_size)
5361 return 0;
5362 if (off+len > i_size)
5363 len = i_size-off;
5364 toread = len;
5365 while (toread > 0) {
5366 tocopy = sb->s_blocksize - offset < toread ?
5367 sb->s_blocksize - offset : toread;
1c215028
TT		 5368 bh = ext4_bread(NULL, inode, blk, 0);
5369 if (IS_ERR(bh))
5370 return PTR_ERR(bh);
ac27a0ec
DK		 5371 if (!bh) /* A hole? */
5372 memset(data, 0, tocopy);
5373 else
5374 memcpy(data, bh->b_data+offset, tocopy);
5375 brelse(bh);
5376 offset = 0;
5377 toread -= tocopy;
5378 data += tocopy;
5379 blk++;
5380 }
5381 return len;
5382}
5383
5384/* Write to quotafile (we know the transaction is already started and has
5385 * enough credits) */
617ba13b 5386static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5387 const char *data, size_t len, loff_t off)
5388{
5389 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5390 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5391 int err, offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5392 struct buffer_head *bh;
5393 handle_t *handle = journal_current_handle();
5394
0390131b 5395 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5396 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5397 " cancelled because transaction is not started",
9c3013e9
JK		 5398 (unsigned long long)off, (unsigned long long)len);
5399 return -EIO;
5400 }
67eeb568
DM		 5401 /*
5402 * Since we account only one data block in transaction credits,
5403 * then it is impossible to cross a block boundary.
5404 */
5405 if (sb->s_blocksize - offset < len) {
5406 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5407 " cancelled because not block aligned",
5408 (unsigned long long)off, (unsigned long long)len);
5409 return -EIO;
5410 }
5411
1c215028
TT		 5412 bh = ext4_bread(handle, inode, blk, 1);
5413 if (IS_ERR(bh))
5414 return PTR_ERR(bh);
67eeb568
DM		 5415 if (!bh)
5416 goto out;
5d601255 5417 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5418 err = ext4_journal_get_write_access(handle, bh);
5419 if (err) {
5420 brelse(bh);
1c215028 5421 return err;
ac27a0ec 5422 }
67eeb568
DM		 5423 lock_buffer(bh);
5424 memcpy(bh->b_data+offset, data, len);
5425 flush_dcache_page(bh->b_page);
5426 unlock_buffer(bh);
62d2b5f2 5427 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5428 brelse(bh);
ac27a0ec 5429out:
67eeb568
DM		 5430 if (inode->i_size < off + len) {
5431 i_size_write(inode, off + len);
617ba13b 5432 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5433 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5434 }
67eeb568 5435 return len;
ac27a0ec
DK		 5436}
5437
5438#endif
5439
152a0836
AV		 5440static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5441 const char *dev_name, void *data)
ac27a0ec 5442{
152a0836 5443 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5444}
5445
37f328eb 5446#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5447static inline void register_as_ext2(void)
5448{
5449 int err = register_filesystem(&ext2_fs_type);
5450 if (err)
5451 printk(KERN_WARNING
5452 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5453}
5454
5455static inline void unregister_as_ext2(void)
5456{
5457 unregister_filesystem(&ext2_fs_type);
5458}
2035e776
TT		 5459
5460static inline int ext2_feature_set_ok(struct super_block *sb)
5461{
5462 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5463 return 0;
5464 if (sb->s_flags & MS_RDONLY)
5465 return 1;
5466 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5467 return 0;
5468 return 1;
5469}
24b58424
TT		 5470#else
5471static inline void register_as_ext2(void) { }
5472static inline void unregister_as_ext2(void) { }
2035e776 5473static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5474#endif
5475
37f328eb 5476#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5477static inline void register_as_ext3(void)
5478{
5479 int err = register_filesystem(&ext3_fs_type);
5480 if (err)
5481 printk(KERN_WARNING
5482 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5483}
5484
5485static inline void unregister_as_ext3(void)
5486{
5487 unregister_filesystem(&ext3_fs_type);
5488}
2035e776
TT		 5489
5490static inline int ext3_feature_set_ok(struct super_block *sb)
5491{
5492 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5493 return 0;
5494 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5495 return 0;
5496 if (sb->s_flags & MS_RDONLY)
5497 return 1;
5498 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5499 return 0;
5500 return 1;
5501}
24b58424
TT		 5502#else
5503static inline void register_as_ext3(void) { }
5504static inline void unregister_as_ext3(void) { }
2035e776 5505static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5506#endif
5507
03010a33
TT		 5508static struct file_system_type ext4_fs_type = {
5509 .owner = THIS_MODULE,
5510 .name = "ext4",
152a0836 5511 .mount = ext4_mount,
03010a33
TT		 5512 .kill_sb = kill_block_super,
5513 .fs_flags = FS_REQUIRES_DEV,
5514};
7f78e035 5515MODULE_ALIAS_FS("ext4");
03010a33 5516
8f021222 5517static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5518{
5519 struct ext4_features *ef;
5520 int ret = -ENOMEM;
5521
5522 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5523 if (!ef)
5524 goto out;
5525
5526 ef->f_kobj.kset = ext4_kset;
5527 init_completion(&ef->f_kobj_unregister);
5528 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5529 "features");
5530 if (ret) {
5531 kfree(ef);
5532 goto out;
5533 }
5534
5535 ext4_feat = ef;
5536 ret = 0;
5537out:
5538 return ret;
5539}
5540
8f021222
LC		 5541static void ext4_exit_feat_adverts(void)
5542{
5543 kobject_put(&ext4_feat->f_kobj);
5544 wait_for_completion(&ext4_feat->f_kobj_unregister);
5545 kfree(ext4_feat);
5546}
5547
e9e3bcec
ES		 5548/* Shared across all ext4 file systems */
5549wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5550struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5551
5dabfc78 5552static int __init ext4_init_fs(void)
ac27a0ec 5553{
e9e3bcec 5554 int i, err;
c9de560d 5555
07c0c5d8
AV		 5556 ext4_li_info = NULL;
5557 mutex_init(&ext4_li_mtx);
5558
9a4c8019 5559 /* Build-time check for flags consistency */
12e9b892 5560 ext4_check_flag_values();
e9e3bcec
ES		 5561
5562 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5563 mutex_init(&ext4__aio_mutex[i]);
5564 init_waitqueue_head(&ext4__ioend_wq[i]);
5565 }
5566
51865fda 5567 err = ext4_init_es();
6fd058f7
TT		 5568 if (err)
5569 return err;
51865fda
ZL		 5570
5571 err = ext4_init_pageio();
5572 if (err)
5573 goto out7;
5574
5dabfc78 5575 err = ext4_init_system_zone();
bd2d0210 5576 if (err)
d44651d0 5577 goto out6;
3197ebdb 5578 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5579 if (!ext4_kset) {
5580 err = -ENOMEM;
dd68314c 5581 goto out5;
0e376b1e 5582 }
d44651d0 5583 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5584
5585 err = ext4_init_feat_adverts();
dd68314c
TT		 5586 if (err)
5587 goto out4;
857ac889 5588
5dabfc78 5589 err = ext4_init_mballoc();
c9de560d
AT		 5590 if (err)
5591 goto out2;
9c191f70
M		 5592 else
5593 ext4_mballoc_ready = 1;
ac27a0ec
DK		 5594 err = init_inodecache();
5595 if (err)
5596 goto out1;
24b58424 5597 register_as_ext3();
2035e776 5598 register_as_ext2();
03010a33 5599 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5600 if (err)
5601 goto out;
bfff6873 5602
ac27a0ec
DK		 5603 return 0;
5604out:
24b58424
TT		 5605 unregister_as_ext2();
5606 unregister_as_ext3();
ac27a0ec
DK		 5607 destroy_inodecache();
5608out1:
9c191f70 5609 ext4_mballoc_ready = 0;
5dabfc78 5610 ext4_exit_mballoc();
9c191f70 5611out2:
8f021222 5612 ext4_exit_feat_adverts();
dd68314c 5613out4:
d44651d0
FJ		 5614 if (ext4_proc_root)
5615 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5616 kset_unregister(ext4_kset);
d44651d0 5617out5:
5dabfc78 5618 ext4_exit_system_zone();
d44651d0 5619out6:
5dabfc78 5620 ext4_exit_pageio();
51865fda
ZL		 5621out7:
5622 ext4_exit_es();
5623
ac27a0ec
DK		 5624 return err;
5625}
5626
5dabfc78 5627static void __exit ext4_exit_fs(void)
ac27a0ec 5628{
bfff6873 5629 ext4_destroy_lazyinit_thread();
24b58424
TT		 5630 unregister_as_ext2();
5631 unregister_as_ext3();
03010a33 5632 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5633 destroy_inodecache();
5dabfc78 5634 ext4_exit_mballoc();
8f021222 5635 ext4_exit_feat_adverts();
9f6200bb 5636 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5637 kset_unregister(ext4_kset);
5dabfc78
TT		 5638 ext4_exit_system_zone();
5639 ext4_exit_pageio();
dd12ed14 5640 ext4_exit_es();
ac27a0ec
DK		 5641}
5642
5643MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5644MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5645MODULE_LICENSE("GPL");
5dabfc78
TT		 5646module_init(ext4_init_fs)
5647module_exit(ext4_exit_fs)
ubuntu-bionic-kernel mirror