]> Git Repo - linux.git/blame - fs/ext4/super.c
projects / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4 crypto: handle unexpected lack of encryption keys
[linux.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card ([email protected])
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 ([email protected]), 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;
1b46617b
TT		 297 if (bdev_read_only(sb->s_bdev))
298 return;
1c13d5c0
TT		 299 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
300 es->s_last_error_time = cpu_to_le32(get_seconds());
301 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
302 es->s_last_error_line = cpu_to_le32(line);
303 if (!es->s_first_error_time) {
304 es->s_first_error_time = es->s_last_error_time;
305 strncpy(es->s_first_error_func, func,
306 sizeof(es->s_first_error_func));
307 es->s_first_error_line = cpu_to_le32(line);
308 es->s_first_error_ino = es->s_last_error_ino;
309 es->s_first_error_block = es->s_last_error_block;
310 }
66e61a9e
TT		 311 /*
312 * Start the daily error reporting function if it hasn't been
313 * started already
314 */
315 if (!es->s_error_count)
316 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 317 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 318}
319
320static void save_error_info(struct super_block *sb, const char *func,
321 unsigned int line)
322{
323 __save_error_info(sb, func, line);
324 ext4_commit_super(sb, 1);
325}
326
18aadd47
BJ		 327static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
328{
329 struct super_block *sb = journal->j_private;
330 struct ext4_sb_info *sbi = EXT4_SB(sb);
331 int error = is_journal_aborted(journal);
5d3ee208 332 struct ext4_journal_cb_entry *jce;
18aadd47 333
5d3ee208 334 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 335 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 336 while (!list_empty(&txn->t_private_list)) {
337 jce = list_entry(txn->t_private_list.next,
338 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 339 list_del_init(&jce->jce_list);
340 spin_unlock(&sbi->s_md_lock);
341 jce->jce_func(sb, jce, error);
342 spin_lock(&sbi->s_md_lock);
343 }
344 spin_unlock(&sbi->s_md_lock);
345}
1c13d5c0 346
ac27a0ec
DK		 347/* Deal with the reporting of failure conditions on a filesystem such as
348 * inconsistencies detected or read IO failures.
349 *
350 * On ext2, we can store the error state of the filesystem in the
617ba13b 351 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 352 * write ordering constraints on the superblock which prevent us from
353 * writing it out straight away; and given that the journal is about to
354 * be aborted, we can't rely on the current, or future, transactions to
355 * write out the superblock safely.
356 *
dab291af 357 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 358 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 359 * that error until we've noted it down and cleared it.
360 */
361
617ba13b 362static void ext4_handle_error(struct super_block *sb)
ac27a0ec 363{
ac27a0ec
DK		 364 if (sb->s_flags & MS_RDONLY)
365 return;
366
2b2d6d01 367 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 368 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 369
4ab2f15b 370 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 371 if (journal)
dab291af 372 jbd2_journal_abort(journal, -EIO);
ac27a0ec 373 }
2b2d6d01 374 if (test_opt(sb, ERRORS_RO)) {
b31e1552 375 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 376 /*
377 * Make sure updated value of ->s_mount_flags will be visible
378 * before ->s_flags update
379 */
380 smp_wmb();
ac27a0ec
DK		 381 sb->s_flags |= MS_RDONLY;
382 }
ac27a0ec 383 if (test_opt(sb, ERRORS_PANIC))
617ba13b 384 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 385 sb->s_id);
386}
387
efbed4dc
TT		 388#define ext4_error_ratelimit(sb) \
389 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
390 "EXT4-fs error")
391
12062ddd 392void __ext4_error(struct super_block *sb, const char *function,
c398eda0 393 unsigned int line, const char *fmt, ...)
ac27a0ec 394{
0ff2ea7d 395 struct va_format vaf;
ac27a0ec
DK		 396 va_list args;
397
efbed4dc
TT		 398 if (ext4_error_ratelimit(sb)) {
399 va_start(args, fmt);
400 vaf.fmt = fmt;
401 vaf.va = &args;
402 printk(KERN_CRIT
403 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
404 sb->s_id, function, line, current->comm, &vaf);
405 va_end(args);
406 }
f3fc0210 407 save_error_info(sb, function, line);
617ba13b 408 ext4_handle_error(sb);
ac27a0ec
DK		 409}
410
e7c96e8e
JP		 411void __ext4_error_inode(struct inode *inode, const char *function,
412 unsigned int line, ext4_fsblk_t block,
413 const char *fmt, ...)
273df556
FM		 414{
415 va_list args;
f7c21177 416 struct va_format vaf;
1c13d5c0 417 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 418
1c13d5c0
TT		 419 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
420 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 421 if (ext4_error_ratelimit(inode->i_sb)) {
422 va_start(args, fmt);
423 vaf.fmt = fmt;
424 vaf.va = &args;
425 if (block)
426 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
427 "inode #%lu: block %llu: comm %s: %pV\n",
428 inode->i_sb->s_id, function, line, inode->i_ino,
429 block, current->comm, &vaf);
430 else
431 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
432 "inode #%lu: comm %s: %pV\n",
433 inode->i_sb->s_id, function, line, inode->i_ino,
434 current->comm, &vaf);
435 va_end(args);
436 }
1c13d5c0 437 save_error_info(inode->i_sb, function, line);
273df556
FM		 438 ext4_handle_error(inode->i_sb);
439}
440
e7c96e8e
JP		 441void __ext4_error_file(struct file *file, const char *function,
442 unsigned int line, ext4_fsblk_t block,
443 const char *fmt, ...)
273df556
FM		 444{
445 va_list args;
f7c21177 446 struct va_format vaf;
1c13d5c0 447 struct ext4_super_block *es;
496ad9aa 448 struct inode *inode = file_inode(file);
273df556
FM		 449 char pathname[80], *path;
450
1c13d5c0
TT		 451 es = EXT4_SB(inode->i_sb)->s_es;
452 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc
TT		 453 if (ext4_error_ratelimit(inode->i_sb)) {
454 path = d_path(&(file->f_path), pathname, sizeof(pathname));
455 if (IS_ERR(path))
456 path = "(unknown)";
457 va_start(args, fmt);
458 vaf.fmt = fmt;
459 vaf.va = &args;
460 if (block)
461 printk(KERN_CRIT
462 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
463 "block %llu: comm %s: path %s: %pV\n",
464 inode->i_sb->s_id, function, line, inode->i_ino,
465 block, current->comm, path, &vaf);
466 else
467 printk(KERN_CRIT
468 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
469 "comm %s: path %s: %pV\n",
470 inode->i_sb->s_id, function, line, inode->i_ino,
471 current->comm, path, &vaf);
472 va_end(args);
473 }
1c13d5c0 474 save_error_info(inode->i_sb, function, line);
273df556
FM		 475 ext4_handle_error(inode->i_sb);
476}
477
722887dd
TT		 478const char *ext4_decode_error(struct super_block *sb, int errno,
479 char nbuf[16])
ac27a0ec
DK		 480{
481 char *errstr = NULL;
482
483 switch (errno) {
484 case -EIO:
485 errstr = "IO failure";
486 break;
487 case -ENOMEM:
488 errstr = "Out of memory";
489 break;
490 case -EROFS:
78f1ddbb
TT		 491 if (!sb || (EXT4_SB(sb)->s_journal &&
492 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 493 errstr = "Journal has aborted";
494 else
495 errstr = "Readonly filesystem";
496 break;
497 default:
498 /* If the caller passed in an extra buffer for unknown
499 * errors, textualise them now. Else we just return
500 * NULL. */
501 if (nbuf) {
502 /* Check for truncated error codes... */
503 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
504 errstr = nbuf;
505 }
506 break;
507 }
508
509 return errstr;
510}
511
617ba13b 512/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 513 * automatically and invokes the appropriate error response. */
514
c398eda0
TT		 515void __ext4_std_error(struct super_block *sb, const char *function,
516 unsigned int line, int errno)
ac27a0ec
DK		 517{
518 char nbuf[16];
519 const char *errstr;
520
521 /* Special case: if the error is EROFS, and we're not already
522 * inside a transaction, then there's really no point in logging
523 * an error. */
524 if (errno == -EROFS && journal_current_handle() == NULL &&
525 (sb->s_flags & MS_RDONLY))
526 return;
527
efbed4dc
TT		 528 if (ext4_error_ratelimit(sb)) {
529 errstr = ext4_decode_error(sb, errno, nbuf);
530 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
531 sb->s_id, function, line, errstr);
532 }
ac27a0ec 533
efbed4dc 534 save_error_info(sb, function, line);
617ba13b 535 ext4_handle_error(sb);
ac27a0ec
DK		 536}
537
538/*
617ba13b 539 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 540 * abort function may be used to deal with unrecoverable failures such
541 * as journal IO errors or ENOMEM at a critical moment in log management.
542 *
543 * We unconditionally force the filesystem into an ABORT|READONLY state,
544 * unless the error response on the fs has been set to panic in which
545 * case we take the easy way out and panic immediately.
546 */
547
c67d859e 548void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 549 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 550{
551 va_list args;
552
1c13d5c0 553 save_error_info(sb, function, line);
ac27a0ec 554 va_start(args, fmt);
c398eda0
TT		 555 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
556 function, line);
ac27a0ec
DK		 557 vprintk(fmt, args);
558 printk("\n");
559 va_end(args);
560
1c13d5c0
TT		 561 if ((sb->s_flags & MS_RDONLY) == 0) {
562 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 563 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 564 /*
565 * Make sure updated value of ->s_mount_flags will be visible
566 * before ->s_flags update
567 */
568 smp_wmb();
569 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 570 if (EXT4_SB(sb)->s_journal)
571 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
572 save_error_info(sb, function, line);
573 }
ac27a0ec 574 if (test_opt(sb, ERRORS_PANIC))
617ba13b 575 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 576}
577
e7c96e8e
JP		 578void __ext4_msg(struct super_block *sb,
579 const char *prefix, const char *fmt, ...)
b31e1552 580{
0ff2ea7d 581 struct va_format vaf;
b31e1552
ES		 582 va_list args;
583
efbed4dc
TT		 584 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
585 return;
586
b31e1552 587 va_start(args, fmt);
0ff2ea7d
JP		 588 vaf.fmt = fmt;
589 vaf.va = &args;
590 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 591 va_end(args);
592}
593
12062ddd 594void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 595 unsigned int line, const char *fmt, ...)
ac27a0ec 596{
0ff2ea7d 597 struct va_format vaf;
ac27a0ec
DK		 598 va_list args;
599
efbed4dc
TT		 600 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
601 "EXT4-fs warning"))
602 return;
603
ac27a0ec 604 va_start(args, fmt);
0ff2ea7d
JP		 605 vaf.fmt = fmt;
606 vaf.va = &args;
607 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
608 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 609 va_end(args);
610}
611
e29136f8
TT		 612void __ext4_grp_locked_error(const char *function, unsigned int line,
613 struct super_block *sb, ext4_group_t grp,
614 unsigned long ino, ext4_fsblk_t block,
615 const char *fmt, ...)
5d1b1b3f
AK		 616__releases(bitlock)
617__acquires(bitlock)
618{
0ff2ea7d 619 struct va_format vaf;
5d1b1b3f
AK		 620 va_list args;
621 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
622
1c13d5c0
TT		 623 es->s_last_error_ino = cpu_to_le32(ino);
624 es->s_last_error_block = cpu_to_le64(block);
625 __save_error_info(sb, function, line);
0ff2ea7d 626
efbed4dc
TT		 627 if (ext4_error_ratelimit(sb)) {
628 va_start(args, fmt);
629 vaf.fmt = fmt;
630 vaf.va = &args;
631 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
632 sb->s_id, function, line, grp);
633 if (ino)
634 printk(KERN_CONT "inode %lu: ", ino);
635 if (block)
636 printk(KERN_CONT "block %llu:",
637 (unsigned long long) block);
638 printk(KERN_CONT "%pV\n", &vaf);
639 va_end(args);
640 }
5d1b1b3f
AK		 641
642 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 643 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 644 return;
645 }
1c13d5c0 646
5d1b1b3f
AK		 647 ext4_unlock_group(sb, grp);
648 ext4_handle_error(sb);
649 /*
650 * We only get here in the ERRORS_RO case; relocking the group
651 * may be dangerous, but nothing bad will happen since the
652 * filesystem will have already been marked read/only and the
653 * journal has been aborted. We return 1 as a hint to callers
654 * who might what to use the return value from
25985edc 655 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 656 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
657 * aggressively from the ext4 function in question, with a
658 * more appropriate error code.
659 */
660 ext4_lock_group(sb, grp);
661 return;
662}
663
617ba13b 664void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 665{
617ba13b 666 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 667
617ba13b 668 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 669 return;
670
12062ddd 671 ext4_warning(sb,
ac27a0ec
DK		 672 "updating to rev %d because of new feature flag, "
673 "running e2fsck is recommended",
617ba13b 674 EXT4_DYNAMIC_REV);
ac27a0ec 675
617ba13b
MC		 676 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
677 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
678 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 679 /* leave es->s_feature_*compat flags alone */
680 /* es->s_uuid will be set by e2fsck if empty */
681
682 /*
683 * The rest of the superblock fields should be zero, and if not it
684 * means they are likely already in use, so leave them alone. We
685 * can leave it up to e2fsck to clean up any inconsistencies there.
686 */
687}
688
689/*
690 * Open the external journal device
691 */
b31e1552 692static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 693{
694 struct block_device *bdev;
695 char b[BDEVNAME_SIZE];
696
d4d77629 697 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 698 if (IS_ERR(bdev))
699 goto fail;
700 return bdev;
701
702fail:
b31e1552 703 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 704 __bdevname(dev, b), PTR_ERR(bdev));
705 return NULL;
706}
707
708/*
709 * Release the journal device
710 */
4385bab1 711static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 712{
4385bab1 713 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 714}
715
4385bab1 716static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 717{
718 struct block_device *bdev;
ac27a0ec
DK		 719 bdev = sbi->journal_bdev;
720 if (bdev) {
4385bab1 721 ext4_blkdev_put(bdev);
ac27a0ec
DK		 722 sbi->journal_bdev = NULL;
723 }
ac27a0ec
DK		 724}
725
726static inline struct inode *orphan_list_entry(struct list_head *l)
727{
617ba13b 728 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 729}
730
617ba13b 731static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 732{
733 struct list_head *l;
734
b31e1552
ES		 735 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
736 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 737
738 printk(KERN_ERR "sb_info orphan list:\n");
739 list_for_each(l, &sbi->s_orphan) {
740 struct inode *inode = orphan_list_entry(l);
741 printk(KERN_ERR " "
742 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
743 inode->i_sb->s_id, inode->i_ino, inode,
744 inode->i_mode, inode->i_nlink,
745 NEXT_ORPHAN(inode));
746 }
747}
748
2b2d6d01 749static void ext4_put_super(struct super_block *sb)
ac27a0ec 750{
617ba13b
MC		 751 struct ext4_sb_info *sbi = EXT4_SB(sb);
752 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 753 int i, err;
ac27a0ec 754
857ac889 755 ext4_unregister_li_request(sb);
e0ccfd95
CH		 756 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
757
2e8fa54e 758 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 759 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 760
0390131b
FM		 761 if (sbi->s_journal) {
762 err = jbd2_journal_destroy(sbi->s_journal);
763 sbi->s_journal = NULL;
764 if (err < 0)
c67d859e 765 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 766 }
d4edac31 767
d3922a77 768 ext4_es_unregister_shrinker(sbi);
9105bb14 769 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 770 ext4_release_system_zone(sb);
771 ext4_mb_release(sb);
772 ext4_ext_release(sb);
773 ext4_xattr_put_super(sb);
774
ac27a0ec 775 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 776 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 777 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 778 }
58c5873a 779 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 780 ext4_commit_super(sb, 1);
781
240799cd 782 if (sbi->s_proc) {
66acdcf4 783 remove_proc_entry("options", sbi->s_proc);
9f6200bb 784 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 785 }
3197ebdb 786 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 787
788 for (i = 0; i < sbi->s_gdb_count; i++)
789 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 790 kvfree(sbi->s_group_desc);
791 kvfree(sbi->s_flex_groups);
57042651 792 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 793 percpu_counter_destroy(&sbi->s_freeinodes_counter);
794 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 795 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 796 brelse(sbi->s_sbh);
797#ifdef CONFIG_QUOTA
a2d4a646 798 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 799 kfree(sbi->s_qf_names[i]);
800#endif
801
802 /* Debugging code just in case the in-memory inode orphan list
803 * isn't empty. The on-disk one can be non-empty if we've
804 * detected an error and taken the fs readonly, but the
805 * in-memory list had better be clean by this point. */
806 if (!list_empty(&sbi->s_orphan))
807 dump_orphan_list(sb, sbi);
808 J_ASSERT(list_empty(&sbi->s_orphan));
809
f98393a6 810 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 811 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
812 /*
813 * Invalidate the journal device's buffers. We don't want them
814 * floating about in memory - the physical journal device may
815 * hotswapped, and it breaks the `ro-after' testing code.
816 */
817 sync_blockdev(sbi->journal_bdev);
f98393a6 818 invalidate_bdev(sbi->journal_bdev);
617ba13b 819 ext4_blkdev_remove(sbi);
ac27a0ec 820 }
9c191f70
M		 821 if (sbi->s_mb_cache) {
822 ext4_xattr_destroy_cache(sbi->s_mb_cache);
823 sbi->s_mb_cache = NULL;
824 }
c5e06d10
JL		 825 if (sbi->s_mmp_tsk)
826 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 827 sb->s_fs_info = NULL;
3197ebdb
TT		 828 /*
829 * Now that we are completely done shutting down the
830 * superblock, we need to actually destroy the kobject.
831 */
3197ebdb
TT		 832 kobject_put(&sbi->s_kobj);
833 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 834 if (sbi->s_chksum_driver)
835 crypto_free_shash(sbi->s_chksum_driver);
705895b6 836 kfree(sbi->s_blockgroup_lock);
ac27a0ec 837 kfree(sbi);
ac27a0ec
DK		 838}
839
e18b890b 840static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 841
842/*
843 * Called inside transaction, so use GFP_NOFS
844 */
617ba13b 845static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 846{
617ba13b 847 struct ext4_inode_info *ei;
ac27a0ec 848
e6b4f8da 849 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 850 if (!ei)
851 return NULL;
0b8e58a1 852
ac27a0ec 853 ei->vfs_inode.i_version = 1;
202ee5df 854 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 855 INIT_LIST_HEAD(&ei->i_prealloc_list);
856 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 857 ext4_es_init_tree(&ei->i_es_tree);
858 rwlock_init(&ei->i_es_lock);
edaa53ca 859 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 860 ei->i_es_all_nr = 0;
edaa53ca 861 ei->i_es_shk_nr = 0;
dd475925 862 ei->i_es_shrink_lblk = 0;
d2a17637
MC		 863 ei->i_reserved_data_blocks = 0;
864 ei->i_reserved_meta_blocks = 0;
865 ei->i_allocated_meta_blocks = 0;
9d0be502 866 ei->i_da_metadata_calc_len = 0;
7e731bc9 867 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 868 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 869#ifdef CONFIG_QUOTA
870 ei->i_reserved_quota = 0;
96c7e0d9 871 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 872#endif
8aefcd55 873 ei->jinode = NULL;
2e8fa54e 874 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 875 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 876 ei->i_sync_tid = 0;
877 ei->i_datasync_tid = 0;
f7ad6d2e 878 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 879 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 880 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
b30ab0e0 881#ifdef CONFIG_EXT4_FS_ENCRYPTION
b7236e21 882 ei->i_crypt_info = NULL;
b30ab0e0 883#endif
ac27a0ec
DK		 884 return &ei->vfs_inode;
885}
886
7ff9c073
TT		 887static int ext4_drop_inode(struct inode *inode)
888{
889 int drop = generic_drop_inode(inode);
890
891 trace_ext4_drop_inode(inode, drop);
892 return drop;
893}
894
fa0d7e3d
NP		 895static void ext4_i_callback(struct rcu_head *head)
896{
897 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 898 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
899}
900
617ba13b 901static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 902{
9f7dd93d 903 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 904 ext4_msg(inode->i_sb, KERN_ERR,
905 "Inode %lu (%p): orphan list check failed!",
906 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 907 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
908 EXT4_I(inode), sizeof(struct ext4_inode_info),
909 true);
910 dump_stack();
911 }
fa0d7e3d 912 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 913}
914
51cc5068 915static void init_once(void *foo)
ac27a0ec 916{
617ba13b 917 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 918
a35afb83 919 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 920 init_rwsem(&ei->xattr_sem);
0e855ac8 921 init_rwsem(&ei->i_data_sem);
a35afb83 922 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 923}
924
e67bc2b3 925static int __init init_inodecache(void)
ac27a0ec 926{
617ba13b
MC		 927 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
928 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 929 0, (SLAB_RECLAIM_ACCOUNT|
930 SLAB_MEM_SPREAD),
20c2df83 931 init_once);
617ba13b 932 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 933 return -ENOMEM;
934 return 0;
935}
936
937static void destroy_inodecache(void)
938{
8c0a8537
KS		 939 /*
940 * Make sure all delayed rcu free inodes are flushed before we
941 * destroy cache.
942 */
943 rcu_barrier();
617ba13b 944 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 945}
946
0930fcc1 947void ext4_clear_inode(struct inode *inode)
ac27a0ec 948{
0930fcc1 949 invalidate_inode_buffers(inode);
dbd5768f 950 clear_inode(inode);
9f754758 951 dquot_drop(inode);
c2ea3fde 952 ext4_discard_preallocations(inode);
51865fda 953 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 954 if (EXT4_I(inode)->jinode) {
955 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
956 EXT4_I(inode)->jinode);
957 jbd2_free_inode(EXT4_I(inode)->jinode);
958 EXT4_I(inode)->jinode = NULL;
959 }
b7236e21
TT		 960#ifdef CONFIG_EXT4_FS_ENCRYPTION
961 if (EXT4_I(inode)->i_crypt_info)
c936e1ec 962 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
b7236e21 963#endif
ac27a0ec
DK		 964}
965
1b961ac0 966static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 967 u64 ino, u32 generation)
ac27a0ec 968{
ac27a0ec 969 struct inode *inode;
ac27a0ec 970
617ba13b 971 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 972 return ERR_PTR(-ESTALE);
617ba13b 973 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 974 return ERR_PTR(-ESTALE);
975
976 /* iget isn't really right if the inode is currently unallocated!!
977 *
617ba13b 978 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 979 * deleted, so we should be safe.
980 *
981 * Currently we don't know the generation for parent directory, so
982 * a generation of 0 means "accept any"
983 */
f4bb2981 984 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 985 if (IS_ERR(inode))
986 return ERR_CAST(inode);
987 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 988 iput(inode);
989 return ERR_PTR(-ESTALE);
990 }
1b961ac0
CH		 991
992 return inode;
993}
994
995static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 996 int fh_len, int fh_type)
1b961ac0
CH		 997{
998 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
999 ext4_nfs_get_inode);
1000}
1001
1002static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1003 int fh_len, int fh_type)
1b961ac0
CH		 1004{
1005 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1006 ext4_nfs_get_inode);
ac27a0ec
DK		 1007}
1008
c39a7f84
TO		 1009/*
1010 * Try to release metadata pages (indirect blocks, directories) which are
1011 * mapped via the block device. Since these pages could have journal heads
1012 * which would prevent try_to_free_buffers() from freeing them, we must use
1013 * jbd2 layer's try_to_free_buffers() function to release them.
1014 */
0b8e58a1
AD		 1015static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1016 gfp_t wait)
c39a7f84
TO		 1017{
1018 journal_t *journal = EXT4_SB(sb)->s_journal;
1019
1020 WARN_ON(PageChecked(page));
1021 if (!page_has_buffers(page))
1022 return 0;
1023 if (journal)
1024 return jbd2_journal_try_to_free_buffers(journal, page,
1025 wait & ~__GFP_WAIT);
1026 return try_to_free_buffers(page);
1027}
1028
ac27a0ec 1029#ifdef CONFIG_QUOTA
af5bc92d 1030#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1031#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1032
617ba13b
MC		 1033static int ext4_write_dquot(struct dquot *dquot);
1034static int ext4_acquire_dquot(struct dquot *dquot);
1035static int ext4_release_dquot(struct dquot *dquot);
1036static int ext4_mark_dquot_dirty(struct dquot *dquot);
1037static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1038static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1039 struct path *path);
ca0e05e4 1040static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1041static int ext4_quota_on_mount(struct super_block *sb, int type);
1042static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1043 size_t len, loff_t off);
617ba13b 1044static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1045 const char *data, size_t len, loff_t off);
7c319d32
AK		 1046static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1047 unsigned int flags);
1048static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1049
96c7e0d9
JK		 1050static struct dquot **ext4_get_dquots(struct inode *inode)
1051{
1052 return EXT4_I(inode)->i_dquot;
1053}
1054
61e225dc 1055static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1056 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1057 .write_dquot = ext4_write_dquot,
1058 .acquire_dquot = ext4_acquire_dquot,
1059 .release_dquot = ext4_release_dquot,
1060 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1061 .write_info = ext4_write_info,
1062 .alloc_dquot = dquot_alloc,
1063 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1064};
1065
0d54b217 1066static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1067 .quota_on = ext4_quota_on,
ca0e05e4 1068 .quota_off = ext4_quota_off,
287a8095 1069 .quota_sync = dquot_quota_sync,
0a240339 1070 .get_state = dquot_get_state,
287a8095
CH		 1071 .set_info = dquot_set_dqinfo,
1072 .get_dqblk = dquot_get_dqblk,
1073 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1074};
1075#endif
1076
ee9b6d61 1077static const struct super_operations ext4_sops = {
617ba13b
MC		 1078 .alloc_inode = ext4_alloc_inode,
1079 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1080 .write_inode = ext4_write_inode,
1081 .dirty_inode = ext4_dirty_inode,
7ff9c073 1082 .drop_inode = ext4_drop_inode,
0930fcc1 1083 .evict_inode = ext4_evict_inode,
617ba13b 1084 .put_super = ext4_put_super,
617ba13b 1085 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1086 .freeze_fs = ext4_freeze,
1087 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1088 .statfs = ext4_statfs,
1089 .remount_fs = ext4_remount,
617ba13b 1090 .show_options = ext4_show_options,
ac27a0ec 1091#ifdef CONFIG_QUOTA
617ba13b
MC		 1092 .quota_read = ext4_quota_read,
1093 .quota_write = ext4_quota_write,
96c7e0d9 1094 .get_dquots = ext4_get_dquots,
ac27a0ec 1095#endif
c39a7f84 1096 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1097};
1098
39655164 1099static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1100 .fh_to_dentry = ext4_fh_to_dentry,
1101 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1102 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1103};
1104
1105enum {
1106 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1107 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1108 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1109 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1110 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1111 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1112 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1113 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1114 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1115 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1116 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1117 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
923ae0ff 1118 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1449032b 1119 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
a26f4992 1120 Opt_lazytime, Opt_nolazytime,
1449032b 1121 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1122 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1123 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1124 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1125 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1126};
1127
a447c093 1128static const match_table_t tokens = {
ac27a0ec
DK		 1129 {Opt_bsd_df, "bsddf"},
1130 {Opt_minix_df, "minixdf"},
1131 {Opt_grpid, "grpid"},
1132 {Opt_grpid, "bsdgroups"},
1133 {Opt_nogrpid, "nogrpid"},
1134 {Opt_nogrpid, "sysvgroups"},
1135 {Opt_resgid, "resgid=%u"},
1136 {Opt_resuid, "resuid=%u"},
1137 {Opt_sb, "sb=%u"},
1138 {Opt_err_cont, "errors=continue"},
1139 {Opt_err_panic, "errors=panic"},
1140 {Opt_err_ro, "errors=remount-ro"},
1141 {Opt_nouid32, "nouid32"},
ac27a0ec 1142 {Opt_debug, "debug"},
72578c33
TT		 1143 {Opt_removed, "oldalloc"},
1144 {Opt_removed, "orlov"},
ac27a0ec
DK		 1145 {Opt_user_xattr, "user_xattr"},
1146 {Opt_nouser_xattr, "nouser_xattr"},
1147 {Opt_acl, "acl"},
1148 {Opt_noacl, "noacl"},
e3bb52ae 1149 {Opt_noload, "norecovery"},
5a916be1 1150 {Opt_noload, "noload"},
72578c33
TT		 1151 {Opt_removed, "nobh"},
1152 {Opt_removed, "bh"},
ac27a0ec 1153 {Opt_commit, "commit=%u"},
30773840
TT		 1154 {Opt_min_batch_time, "min_batch_time=%u"},
1155 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1156 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1157 {Opt_journal_path, "journal_path=%s"},
818d276c 1158 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1159 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1160 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1161 {Opt_abort, "abort"},
1162 {Opt_data_journal, "data=journal"},
1163 {Opt_data_ordered, "data=ordered"},
1164 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1165 {Opt_data_err_abort, "data_err=abort"},
1166 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1167 {Opt_offusrjquota, "usrjquota="},
1168 {Opt_usrjquota, "usrjquota=%s"},
1169 {Opt_offgrpjquota, "grpjquota="},
1170 {Opt_grpjquota, "grpjquota=%s"},
1171 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1172 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1173 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1174 {Opt_grpquota, "grpquota"},
1175 {Opt_noquota, "noquota"},
1176 {Opt_quota, "quota"},
1177 {Opt_usrquota, "usrquota"},
1178 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1179 {Opt_barrier, "barrier"},
1180 {Opt_nobarrier, "nobarrier"},
25ec56b5 1181 {Opt_i_version, "i_version"},
923ae0ff 1182 {Opt_dax, "dax"},
c9de560d 1183 {Opt_stripe, "stripe=%u"},
64769240 1184 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1185 {Opt_lazytime, "lazytime"},
1186 {Opt_nolazytime, "nolazytime"},
dd919b98 1187 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1188 {Opt_removed, "mblk_io_submit"},
1189 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1190 {Opt_block_validity, "block_validity"},
1191 {Opt_noblock_validity, "noblock_validity"},
240799cd 1192 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1193 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1194 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1195 {Opt_auto_da_alloc, "auto_da_alloc"},
1196 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1197 {Opt_dioread_nolock, "dioread_nolock"},
1198 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1199 {Opt_discard, "discard"},
1200 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1201 {Opt_init_itable, "init_itable=%u"},
1202 {Opt_init_itable, "init_itable"},
1203 {Opt_noinit_itable, "noinit_itable"},
df981d03 1204 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1205 {Opt_test_dummy_encryption, "test_dummy_encryption"},
c7198b9c
TT		 1206 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1207 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1208 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1209 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1210 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1211 {Opt_err, NULL},
ac27a0ec
DK		 1212};
1213
617ba13b 1214static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1215{
617ba13b 1216 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1217 char *options = (char *) *data;
1218
1219 if (!options || strncmp(options, "sb=", 3) != 0)
1220 return 1; /* Default location */
0b8e58a1 1221
ac27a0ec 1222 options += 3;
0b8e58a1 1223 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1224 sb_block = simple_strtoul(options, &options, 0);
1225 if (*options && *options != ',') {
4776004f 1226 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1227 (char *) *data);
1228 return 1;
1229 }
1230 if (*options == ',')
1231 options++;
1232 *data = (void *) options;
0b8e58a1 1233
ac27a0ec
DK		 1234 return sb_block;
1235}
1236
b3881f74 1237#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1238static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1239 "Contact [email protected] if you think we should keep it.\n";
b3881f74 1240
56c50f11
DM		 1241#ifdef CONFIG_QUOTA
1242static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1243{
1244 struct ext4_sb_info *sbi = EXT4_SB(sb);
1245 char *qname;
03dafb5f 1246 int ret = -1;
56c50f11
DM		 1247
1248 if (sb_any_quota_loaded(sb) &&
1249 !sbi->s_qf_names[qtype]) {
1250 ext4_msg(sb, KERN_ERR,
1251 "Cannot change journaled "
1252 "quota options when quota turned on");
57f73c2c 1253 return -1;
56c50f11 1254 }
262b4662
JK		 1255 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1256 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1257 "when QUOTA feature is enabled");
1258 return -1;
1259 }
56c50f11
DM		 1260 qname = match_strdup(args);
1261 if (!qname) {
1262 ext4_msg(sb, KERN_ERR,
1263 "Not enough memory for storing quotafile name");
57f73c2c 1264 return -1;
56c50f11 1265 }
03dafb5f
CG		 1266 if (sbi->s_qf_names[qtype]) {
1267 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1268 ret = 1;
1269 else
1270 ext4_msg(sb, KERN_ERR,
1271 "%s quota file already specified",
1272 QTYPE2NAME(qtype));
1273 goto errout;
56c50f11 1274 }
03dafb5f 1275 if (strchr(qname, '/')) {
56c50f11
DM		 1276 ext4_msg(sb, KERN_ERR,
1277 "quotafile must be on filesystem root");
03dafb5f 1278 goto errout;
56c50f11 1279 }
03dafb5f 1280 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1281 set_opt(sb, QUOTA);
56c50f11 1282 return 1;
03dafb5f
CG		 1283errout:
1284 kfree(qname);
1285 return ret;
56c50f11
DM		 1286}
1287
1288static int clear_qf_name(struct super_block *sb, int qtype)
1289{
1290
1291 struct ext4_sb_info *sbi = EXT4_SB(sb);
1292
1293 if (sb_any_quota_loaded(sb) &&
1294 sbi->s_qf_names[qtype]) {
1295 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1296 " when quota turned on");
57f73c2c 1297 return -1;
56c50f11 1298 }
03dafb5f 1299 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1300 sbi->s_qf_names[qtype] = NULL;
1301 return 1;
1302}
1303#endif
1304
26092bf5
TT		 1305#define MOPT_SET 0x0001
1306#define MOPT_CLEAR 0x0002
1307#define MOPT_NOSUPPORT 0x0004
1308#define MOPT_EXPLICIT 0x0008
1309#define MOPT_CLEAR_ERR 0x0010
1310#define MOPT_GTE0 0x0020
ac27a0ec 1311#ifdef CONFIG_QUOTA
26092bf5
TT		 1312#define MOPT_Q 0
1313#define MOPT_QFMT 0x0040
1314#else
1315#define MOPT_Q MOPT_NOSUPPORT
1316#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1317#endif
26092bf5 1318#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1319#define MOPT_NO_EXT2 0x0100
1320#define MOPT_NO_EXT3 0x0200
1321#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1322#define MOPT_STRING 0x0400
26092bf5
TT		 1323
1324static const struct mount_opts {
1325 int token;
1326 int mount_opt;
1327 int flags;
1328} ext4_mount_opts[] = {
1329 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1330 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1331 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1332 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1333 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1334 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1335 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1336 MOPT_EXT4_ONLY | MOPT_SET},
1337 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1338 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1339 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1340 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1341 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1342 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1343 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1344 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1345 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1346 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c
TT		 1347 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1348 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1349 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1350 EXT4_MOUNT_JOURNAL_CHECKSUM),
1351 MOPT_EXT4_ONLY | MOPT_SET},
1352 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1353 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1354 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1355 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1356 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1357 MOPT_NO_EXT2 | MOPT_SET},
1358 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1359 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1360 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1361 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1362 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1363 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1364 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1365 {Opt_commit, 0, MOPT_GTE0},
1366 {Opt_max_batch_time, 0, MOPT_GTE0},
1367 {Opt_min_batch_time, 0, MOPT_GTE0},
1368 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1369 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1370 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1371 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1372 {Opt_resuid, 0, MOPT_GTE0},
1373 {Opt_resgid, 0, MOPT_GTE0},
1374 {Opt_journal_dev, 0, MOPT_GTE0},
ad4eec61 1375 {Opt_journal_path, 0, MOPT_STRING},
0efb3b23 1376 {Opt_journal_ioprio, 0, MOPT_GTE0},
8dc0aa8c
TT		 1377 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1378 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1379 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1380 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1381 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1382 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1383#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1384 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1385 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1386#else
26092bf5
TT		 1387 {Opt_acl, 0, MOPT_NOSUPPORT},
1388 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1389#endif
26092bf5
TT		 1390 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1391 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1392 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1393 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1394 MOPT_SET | MOPT_Q},
1395 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1396 MOPT_SET | MOPT_Q},
1397 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1398 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1399 {Opt_usrjquota, 0, MOPT_Q},
1400 {Opt_grpjquota, 0, MOPT_Q},
1401 {Opt_offusrjquota, 0, MOPT_Q},
1402 {Opt_offgrpjquota, 0, MOPT_Q},
1403 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1404 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1405 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1406 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1407 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
26092bf5
TT		 1408 {Opt_err, 0, 0}
1409};
1410
1411static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1412 substring_t *args, unsigned long *journal_devnum,
1413 unsigned int *journal_ioprio, int is_remount)
1414{
1415 struct ext4_sb_info *sbi = EXT4_SB(sb);
1416 const struct mount_opts *m;
08cefc7a
EB		 1417 kuid_t uid;
1418 kgid_t gid;
26092bf5
TT		 1419 int arg = 0;
1420
57f73c2c
TT		 1421#ifdef CONFIG_QUOTA
1422 if (token == Opt_usrjquota)
1423 return set_qf_name(sb, USRQUOTA, &args[0]);
1424 else if (token == Opt_grpjquota)
1425 return set_qf_name(sb, GRPQUOTA, &args[0]);
1426 else if (token == Opt_offusrjquota)
1427 return clear_qf_name(sb, USRQUOTA);
1428 else if (token == Opt_offgrpjquota)
1429 return clear_qf_name(sb, GRPQUOTA);
1430#endif
26092bf5 1431 switch (token) {
f7048605
TT		 1432 case Opt_noacl:
1433 case Opt_nouser_xattr:
1434 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1435 break;
26092bf5
TT		 1436 case Opt_sb:
1437 return 1; /* handled by get_sb_block() */
1438 case Opt_removed:
5f3633e3 1439 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1440 return 1;
26092bf5
TT		 1441 case Opt_abort:
1442 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1443 return 1;
1444 case Opt_i_version:
1445 sb->s_flags |= MS_I_VERSION;
1446 return 1;
a26f4992
TT		 1447 case Opt_lazytime:
1448 sb->s_flags |= MS_LAZYTIME;
1449 return 1;
1450 case Opt_nolazytime:
1451 sb->s_flags &= ~MS_LAZYTIME;
1452 return 1;
26092bf5
TT		 1453 }
1454
5f3633e3
JK		 1455 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1456 if (token == m->token)
1457 break;
1458
1459 if (m->token == Opt_err) {
1460 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1461 "or missing value", opt);
1462 return -1;
1463 }
1464
8dc0aa8c
TT		 1465 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1466 ext4_msg(sb, KERN_ERR,
1467 "Mount option \"%s\" incompatible with ext2", opt);
1468 return -1;
1469 }
1470 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1471 ext4_msg(sb, KERN_ERR,
1472 "Mount option \"%s\" incompatible with ext3", opt);
1473 return -1;
1474 }
1475
ad4eec61 1476 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1477 return -1;
1478 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1479 return -1;
1480 if (m->flags & MOPT_EXPLICIT)
1481 set_opt2(sb, EXPLICIT_DELALLOC);
1482 if (m->flags & MOPT_CLEAR_ERR)
1483 clear_opt(sb, ERRORS_MASK);
1484 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1485 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1486 "options when quota turned on");
1487 return -1;
1488 }
1489
1490 if (m->flags & MOPT_NOSUPPORT) {
1491 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1492 } else if (token == Opt_commit) {
1493 if (arg == 0)
1494 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1495 sbi->s_commit_interval = HZ * arg;
1496 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1497 sbi->s_max_batch_time = arg;
1498 } else if (token == Opt_min_batch_time) {
1499 sbi->s_min_batch_time = arg;
1500 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1501 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1502 ext4_msg(sb, KERN_ERR,
1503 "EXT4-fs: inode_readahead_blks must be "
1504 "0 or a power of 2 smaller than 2^31");
26092bf5 1505 return -1;
5f3633e3
JK		 1506 }
1507 sbi->s_inode_readahead_blks = arg;
1508 } else if (token == Opt_init_itable) {
1509 set_opt(sb, INIT_INODE_TABLE);
1510 if (!args->from)
1511 arg = EXT4_DEF_LI_WAIT_MULT;
1512 sbi->s_li_wait_mult = arg;
1513 } else if (token == Opt_max_dir_size_kb) {
1514 sbi->s_max_dir_size_kb = arg;
1515 } else if (token == Opt_stripe) {
1516 sbi->s_stripe = arg;
1517 } else if (token == Opt_resuid) {
1518 uid = make_kuid(current_user_ns(), arg);
1519 if (!uid_valid(uid)) {
1520 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1521 return -1;
1522 }
5f3633e3
JK		 1523 sbi->s_resuid = uid;
1524 } else if (token == Opt_resgid) {
1525 gid = make_kgid(current_user_ns(), arg);
1526 if (!gid_valid(gid)) {
1527 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1528 return -1;
1529 }
1530 sbi->s_resgid = gid;
1531 } else if (token == Opt_journal_dev) {
1532 if (is_remount) {
1533 ext4_msg(sb, KERN_ERR,
1534 "Cannot specify journal on remount");
1535 return -1;
1536 }
1537 *journal_devnum = arg;
ad4eec61
ES		 1538 } else if (token == Opt_journal_path) {
1539 char *journal_path;
1540 struct inode *journal_inode;
1541 struct path path;
1542 int error;
1543
1544 if (is_remount) {
1545 ext4_msg(sb, KERN_ERR,
1546 "Cannot specify journal on remount");
1547 return -1;
1548 }
1549 journal_path = match_strdup(&args[0]);
1550 if (!journal_path) {
1551 ext4_msg(sb, KERN_ERR, "error: could not dup "
1552 "journal device string");
1553 return -1;
1554 }
1555
1556 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1557 if (error) {
1558 ext4_msg(sb, KERN_ERR, "error: could not find "
1559 "journal device path: error %d", error);
1560 kfree(journal_path);
1561 return -1;
1562 }
1563
2b0143b5 1564 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1565 if (!S_ISBLK(journal_inode->i_mode)) {
1566 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1567 "is not a block device", journal_path);
1568 path_put(&path);
1569 kfree(journal_path);
1570 return -1;
1571 }
1572
1573 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1574 path_put(&path);
1575 kfree(journal_path);
5f3633e3
JK		 1576 } else if (token == Opt_journal_ioprio) {
1577 if (arg > 7) {
1578 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1579 " (must be 0-7)");
1580 return -1;
1581 }
1582 *journal_ioprio =
1583 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1584 } else if (token == Opt_test_dummy_encryption) {
1585#ifdef CONFIG_EXT4_FS_ENCRYPTION
1586 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1587 ext4_msg(sb, KERN_WARNING,
1588 "Test dummy encryption mode enabled");
1589#else
1590 ext4_msg(sb, KERN_WARNING,
1591 "Test dummy encryption mount option ignored");
1592#endif
5f3633e3
JK		 1593 } else if (m->flags & MOPT_DATAJ) {
1594 if (is_remount) {
1595 if (!sbi->s_journal)
1596 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1597 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1598 ext4_msg(sb, KERN_ERR,
26092bf5 1599 "Cannot change data mode on remount");
26092bf5 1600 return -1;
ac27a0ec 1601 }
26092bf5 1602 } else {
5f3633e3
JK		 1603 clear_opt(sb, DATA_FLAGS);
1604 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1605 }
5f3633e3
JK		 1606#ifdef CONFIG_QUOTA
1607 } else if (m->flags & MOPT_QFMT) {
1608 if (sb_any_quota_loaded(sb) &&
1609 sbi->s_jquota_fmt != m->mount_opt) {
1610 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1611 "quota options when quota turned on");
1612 return -1;
1613 }
262b4662
JK		 1614 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1615 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1616 ext4_msg(sb, KERN_ERR,
1617 "Cannot set journaled quota options "
1618 "when QUOTA feature is enabled");
1619 return -1;
1620 }
5f3633e3 1621 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff
RZ		 1622#endif
1623#ifndef CONFIG_FS_DAX
1624 } else if (token == Opt_dax) {
1625 ext4_msg(sb, KERN_INFO, "dax option not supported");
1626 return -1;
5f3633e3
JK		 1627#endif
1628 } else {
1629 if (!args->from)
1630 arg = 1;
1631 if (m->flags & MOPT_CLEAR)
1632 arg = !arg;
1633 else if (unlikely(!(m->flags & MOPT_SET))) {
1634 ext4_msg(sb, KERN_WARNING,
1635 "buggy handling of option %s", opt);
1636 WARN_ON(1);
1637 return -1;
1638 }
1639 if (arg != 0)
1640 sbi->s_mount_opt |= m->mount_opt;
1641 else
1642 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1643 }
5f3633e3 1644 return 1;
26092bf5
TT		 1645}
1646
1647static int parse_options(char *options, struct super_block *sb,
1648 unsigned long *journal_devnum,
1649 unsigned int *journal_ioprio,
1650 int is_remount)
1651{
1652 struct ext4_sb_info *sbi = EXT4_SB(sb);
1653 char *p;
1654 substring_t args[MAX_OPT_ARGS];
1655 int token;
1656
1657 if (!options)
1658 return 1;
1659
1660 while ((p = strsep(&options, ",")) != NULL) {
1661 if (!*p)
1662 continue;
1663 /*
1664 * Initialize args struct so we know whether arg was
1665 * found; some options take optional arguments.
1666 */
caecd0af 1667 args[0].to = args[0].from = NULL;
26092bf5
TT		 1668 token = match_token(p, tokens, args);
1669 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1670 journal_ioprio, is_remount) < 0)
1671 return 0;
ac27a0ec
DK		 1672 }
1673#ifdef CONFIG_QUOTA
262b4662
JK		 1674 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1675 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1676 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1677 "feature is enabled");
1678 return 0;
1679 }
ac27a0ec 1680 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1681 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1682 clear_opt(sb, USRQUOTA);
ac27a0ec 1683
482a7425 1684 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1685 clear_opt(sb, GRPQUOTA);
ac27a0ec 1686
56c50f11 1687 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1688 ext4_msg(sb, KERN_ERR, "old and new quota "
1689 "format mixing");
ac27a0ec
DK		 1690 return 0;
1691 }
1692
1693 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1694 ext4_msg(sb, KERN_ERR, "journaled quota format "
1695 "not specified");
ac27a0ec
DK		 1696 return 0;
1697 }
ac27a0ec
DK		 1698 }
1699#endif
261cb20c
JK		 1700 if (test_opt(sb, DIOREAD_NOLOCK)) {
1701 int blocksize =
1702 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1703
1704 if (blocksize < PAGE_CACHE_SIZE) {
1705 ext4_msg(sb, KERN_ERR, "can't mount with "
1706 "dioread_nolock if block size != PAGE_SIZE");
1707 return 0;
1708 }
1709 }
d4f76107
JK		 1710 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1711 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1712 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1713 "in data=ordered mode");
1714 return 0;
1715 }
ac27a0ec
DK		 1716 return 1;
1717}
1718
2adf6da8
TT		 1719static inline void ext4_show_quota_options(struct seq_file *seq,
1720 struct super_block *sb)
1721{
1722#if defined(CONFIG_QUOTA)
1723 struct ext4_sb_info *sbi = EXT4_SB(sb);
1724
1725 if (sbi->s_jquota_fmt) {
1726 char *fmtname = "";
1727
1728 switch (sbi->s_jquota_fmt) {
1729 case QFMT_VFS_OLD:
1730 fmtname = "vfsold";
1731 break;
1732 case QFMT_VFS_V0:
1733 fmtname = "vfsv0";
1734 break;
1735 case QFMT_VFS_V1:
1736 fmtname = "vfsv1";
1737 break;
1738 }
1739 seq_printf(seq, ",jqfmt=%s", fmtname);
1740 }
1741
1742 if (sbi->s_qf_names[USRQUOTA])
1743 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1744
1745 if (sbi->s_qf_names[GRPQUOTA])
1746 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1747#endif
1748}
1749
5a916be1
TT		 1750static const char *token2str(int token)
1751{
50df9fd5 1752 const struct match_token *t;
5a916be1
TT		 1753
1754 for (t = tokens; t->token != Opt_err; t++)
1755 if (t->token == token && !strchr(t->pattern, '='))
1756 break;
1757 return t->pattern;
1758}
1759
2adf6da8
TT		 1760/*
1761 * Show an option if
1762 * - it's set to a non-default value OR
1763 * - if the per-sb default is different from the global default
1764 */
66acdcf4
TT		 1765static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1766 int nodefs)
2adf6da8 1767{
2adf6da8
TT		 1768 struct ext4_sb_info *sbi = EXT4_SB(sb);
1769 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1770 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1771 const struct mount_opts *m;
66acdcf4 1772 char sep = nodefs ? '\n' : ',';
2adf6da8 1773
66acdcf4
TT		 1774#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1775#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1776
1777 if (sbi->s_sb_block != 1)
5a916be1
TT		 1778 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1779
1780 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1781 int want_set = m->flags & MOPT_SET;
1782 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1783 (m->flags & MOPT_CLEAR_ERR))
1784 continue;
66acdcf4 1785 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1786 continue; /* skip if same as the default */
1787 if ((want_set &&
1788 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1789 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1790 continue; /* select Opt_noFoo vs Opt_Foo */
1791 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1792 }
5a916be1 1793
08cefc7a 1794 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1795 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1796 SEQ_OPTS_PRINT("resuid=%u",
1797 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1798 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1799 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1800 SEQ_OPTS_PRINT("resgid=%u",
1801 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1802 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1803 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1804 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1805 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1806 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1807 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1808 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1809 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1810 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1811 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1812 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1813 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1814 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1815 if (sb->s_flags & MS_I_VERSION)
5a916be1 1816 SEQ_OPTS_PUTS("i_version");
66acdcf4 1817 if (nodefs || sbi->s_stripe)
5a916be1 1818 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1819 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1820 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1821 SEQ_OPTS_PUTS("data=journal");
1822 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1823 SEQ_OPTS_PUTS("data=ordered");
1824 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1825 SEQ_OPTS_PUTS("data=writeback");
1826 }
66acdcf4
TT		 1827 if (nodefs ||
1828 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1829 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1830 sbi->s_inode_readahead_blks);
2adf6da8 1831
66acdcf4
TT		 1832 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1833 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1834 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1835 if (nodefs || sbi->s_max_dir_size_kb)
1836 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1837
1838 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1839 return 0;
1840}
1841
66acdcf4
TT		 1842static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1843{
1844 return _ext4_show_options(seq, root->d_sb, 0);
1845}
1846
1847static int options_seq_show(struct seq_file *seq, void *offset)
1848{
1849 struct super_block *sb = seq->private;
1850 int rc;
1851
1852 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1853 rc = _ext4_show_options(seq, sb, 1);
1854 seq_puts(seq, "\n");
1855 return rc;
1856}
1857
1858static int options_open_fs(struct inode *inode, struct file *file)
1859{
d9dda78b 1860 return single_open(file, options_seq_show, PDE_DATA(inode));
66acdcf4
TT		 1861}
1862
1863static const struct file_operations ext4_seq_options_fops = {
1864 .owner = THIS_MODULE,
1865 .open = options_open_fs,
1866 .read = seq_read,
1867 .llseek = seq_lseek,
1868 .release = single_release,
1869};
1870
617ba13b 1871static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1872 int read_only)
1873{
617ba13b 1874 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1875 int res = 0;
1876
617ba13b 1877 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1878 ext4_msg(sb, KERN_ERR, "revision level too high, "
1879 "forcing read-only mode");
ac27a0ec
DK		 1880 res = MS_RDONLY;
1881 }
1882 if (read_only)
281b5995 1883 goto done;
617ba13b 1884 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1885 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1886 "running e2fsck is recommended");
c8b459f4 1887 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 1888 ext4_msg(sb, KERN_WARNING,
1889 "warning: mounting fs with errors, "
1890 "running e2fsck is recommended");
ed3ce80a 1891 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1892 le16_to_cpu(es->s_mnt_count) >=
1893 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1894 ext4_msg(sb, KERN_WARNING,
1895 "warning: maximal mount count reached, "
1896 "running e2fsck is recommended");
ac27a0ec
DK		 1897 else if (le32_to_cpu(es->s_checkinterval) &&
1898 (le32_to_cpu(es->s_lastcheck) +
1899 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1900 ext4_msg(sb, KERN_WARNING,
1901 "warning: checktime reached, "
1902 "running e2fsck is recommended");
0b8e58a1 1903 if (!sbi->s_journal)
0390131b 1904 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1905 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1906 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1907 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1908 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1909 ext4_update_dynamic_rev(sb);
0390131b
FM		 1910 if (sbi->s_journal)
1911 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1912
e2d67052 1913 ext4_commit_super(sb, 1);
281b5995 1914done:
ac27a0ec 1915 if (test_opt(sb, DEBUG))
a9df9a49 1916 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1917 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1918 sb->s_blocksize,
1919 sbi->s_groups_count,
617ba13b
MC		 1920 EXT4_BLOCKS_PER_GROUP(sb),
1921 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1922 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1923
7abc52c2 1924 cleancache_init_fs(sb);
ac27a0ec
DK		 1925 return res;
1926}
1927
117fff10
TT		 1928int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1929{
1930 struct ext4_sb_info *sbi = EXT4_SB(sb);
1931 struct flex_groups *new_groups;
1932 int size;
1933
1934 if (!sbi->s_log_groups_per_flex)
1935 return 0;
1936
1937 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1938 if (size <= sbi->s_flex_groups_allocated)
1939 return 0;
1940
1941 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1942 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1943 if (!new_groups) {
1944 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1945 size / (int) sizeof(struct flex_groups));
1946 return -ENOMEM;
1947 }
1948
1949 if (sbi->s_flex_groups) {
1950 memcpy(new_groups, sbi->s_flex_groups,
1951 (sbi->s_flex_groups_allocated *
1952 sizeof(struct flex_groups)));
b93b41d4 1953 kvfree(sbi->s_flex_groups);
117fff10
TT		 1954 }
1955 sbi->s_flex_groups = new_groups;
1956 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1957 return 0;
1958}
1959
772cb7c8
JS		 1960static int ext4_fill_flex_info(struct super_block *sb)
1961{
1962 struct ext4_sb_info *sbi = EXT4_SB(sb);
1963 struct ext4_group_desc *gdp = NULL;
772cb7c8 1964 ext4_group_t flex_group;
117fff10 1965 int i, err;
772cb7c8 1966
503358ae 1967 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1968 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1969 sbi->s_log_groups_per_flex = 0;
1970 return 1;
1971 }
1972
117fff10
TT		 1973 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1974 if (err)
9933fc0a 1975 goto failed;
772cb7c8 1976
772cb7c8 1977 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1978 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1979
1980 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1981 atomic_add(ext4_free_inodes_count(sb, gdp),
1982 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 1983 atomic64_add(ext4_free_group_clusters(sb, gdp),
1984 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1985 atomic_add(ext4_used_dirs_count(sb, gdp),
1986 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1987 }
1988
1989 return 1;
1990failed:
1991 return 0;
1992}
1993
feb0ab32
DW		 1994static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1995 struct ext4_group_desc *gdp)
717d50e4 1996{
feb0ab32 1997 int offset;
717d50e4 1998 __u16 crc = 0;
feb0ab32 1999 __le32 le_group = cpu_to_le32(block_group);
717d50e4 2000
9aa5d32b 2001 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2002 /* Use new metadata_csum algorithm */
d6a77105 2003 __le16 save_csum;
feb0ab32
DW		 2004 __u32 csum32;
2005
d6a77105 2006 save_csum = gdp->bg_checksum;
feb0ab32
DW		 2007 gdp->bg_checksum = 0;
2008 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2009 sizeof(le_group));
2010 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2011 sbi->s_desc_size);
d6a77105 2012 gdp->bg_checksum = save_csum;
feb0ab32
DW		 2013
2014 crc = csum32 & 0xFFFF;
2015 goto out;
717d50e4
AD		 2016 }
2017
feb0ab32 2018 /* old crc16 code */
813d32f9
DW		 2019 if (!(sbi->s_es->s_feature_ro_compat &
2020 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2021 return 0;
2022
feb0ab32
DW		 2023 offset = offsetof(struct ext4_group_desc, bg_checksum);
2024
2025 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2026 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2027 crc = crc16(crc, (__u8 *)gdp, offset);
2028 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2029 /* for checksum of struct ext4_group_desc do the rest...*/
2030 if ((sbi->s_es->s_feature_incompat &
2031 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2032 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2033 crc = crc16(crc, (__u8 *)gdp + offset,
2034 le16_to_cpu(sbi->s_es->s_desc_size) -
2035 offset);
2036
2037out:
717d50e4
AD		 2038 return cpu_to_le16(crc);
2039}
2040
feb0ab32 2041int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2042 struct ext4_group_desc *gdp)
2043{
feb0ab32
DW		 2044 if (ext4_has_group_desc_csum(sb) &&
2045 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2046 block_group, gdp)))
717d50e4
AD		 2047 return 0;
2048
2049 return 1;
2050}
2051
feb0ab32
DW		 2052void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2053 struct ext4_group_desc *gdp)
2054{
2055 if (!ext4_has_group_desc_csum(sb))
2056 return;
2057 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2058}
2059
ac27a0ec 2060/* Called at mount-time, super-block is locked */
bfff6873
LC		 2061static int ext4_check_descriptors(struct super_block *sb,
2062 ext4_group_t *first_not_zeroed)
ac27a0ec 2063{
617ba13b
MC		 2064 struct ext4_sb_info *sbi = EXT4_SB(sb);
2065 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2066 ext4_fsblk_t last_block;
bd81d8ee
LV		 2067 ext4_fsblk_t block_bitmap;
2068 ext4_fsblk_t inode_bitmap;
2069 ext4_fsblk_t inode_table;
ce421581 2070 int flexbg_flag = 0;
bfff6873 2071 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2072
ce421581
JS		 2073 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2074 flexbg_flag = 1;
2075
af5bc92d 2076 ext4_debug("Checking group descriptors");
ac27a0ec 2077
197cd65a
AM		 2078 for (i = 0; i < sbi->s_groups_count; i++) {
2079 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2080
ce421581 2081 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2082 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2083 else
2084 last_block = first_block +
617ba13b 2085 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2086
bfff6873
LC		 2087 if ((grp == sbi->s_groups_count) &&
2088 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2089 grp = i;
2090
8fadc143 2091 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2092 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2093 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2094 "Block bitmap for group %u not in group "
b31e1552 2095 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2096 return 0;
2097 }
8fadc143 2098 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2099 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2100 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2101 "Inode bitmap for group %u not in group "
b31e1552 2102 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2103 return 0;
2104 }
8fadc143 2105 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2106 if (inode_table < first_block ||
2b2d6d01 2107 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2108 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2109 "Inode table for group %u not in group "
b31e1552 2110 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2111 return 0;
2112 }
955ce5f5 2113 ext4_lock_group(sb, i);
feb0ab32 2114 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2115 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2116 "Checksum for group %u failed (%u!=%u)",
2117 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2118 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2119 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2120 ext4_unlock_group(sb, i);
8a266467 2121 return 0;
7ee1ec4c 2122 }
717d50e4 2123 }
955ce5f5 2124 ext4_unlock_group(sb, i);
ce421581
JS		 2125 if (!flexbg_flag)
2126 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2127 }
bfff6873
LC		 2128 if (NULL != first_not_zeroed)
2129 *first_not_zeroed = grp;
ac27a0ec
DK		 2130 return 1;
2131}
2132
617ba13b 2133/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2134 * the superblock) which were deleted from all directories, but held open by
2135 * a process at the time of a crash. We walk the list and try to delete these
2136 * inodes at recovery time (only with a read-write filesystem).
2137 *
2138 * In order to keep the orphan inode chain consistent during traversal (in
2139 * case of crash during recovery), we link each inode into the superblock
2140 * orphan list_head and handle it the same way as an inode deletion during
2141 * normal operation (which journals the operations for us).
2142 *
2143 * We only do an iget() and an iput() on each inode, which is very safe if we
2144 * accidentally point at an in-use or already deleted inode. The worst that
2145 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2146 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2147 * e2fsck was run on this filesystem, and it must have already done the orphan
2148 * inode cleanup for us, so we can safely abort without any further action.
2149 */
2b2d6d01
TT		 2150static void ext4_orphan_cleanup(struct super_block *sb,
2151 struct ext4_super_block *es)
ac27a0ec
DK		 2152{
2153 unsigned int s_flags = sb->s_flags;
2154 int nr_orphans = 0, nr_truncates = 0;
2155#ifdef CONFIG_QUOTA
2156 int i;
2157#endif
2158 if (!es->s_last_orphan) {
2159 jbd_debug(4, "no orphan inodes to clean up\n");
2160 return;
2161 }
2162
a8f48a95 2163 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2164 ext4_msg(sb, KERN_ERR, "write access "
2165 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2166 return;
2167 }
2168
d39195c3
AG		 2169 /* Check if feature set would not allow a r/w mount */
2170 if (!ext4_feature_set_ok(sb, 0)) {
2171 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2172 "unknown ROCOMPAT features");
2173 return;
2174 }
2175
617ba13b 2176 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2177 /* don't clear list on RO mount w/ errors */
2178 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2179 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2180 "clearing orphan list.\n");
c25f9bc6
ES		 2181 es->s_last_orphan = 0;
2182 }
ac27a0ec
DK		 2183 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2184 return;
2185 }
2186
2187 if (s_flags & MS_RDONLY) {
b31e1552 2188 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2189 sb->s_flags &= ~MS_RDONLY;
2190 }
2191#ifdef CONFIG_QUOTA
2192 /* Needed for iput() to work correctly and not trash data */
2193 sb->s_flags |= MS_ACTIVE;
2194 /* Turn on quotas so that they are updated correctly */
a2d4a646 2195 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2196 if (EXT4_SB(sb)->s_qf_names[i]) {
2197 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2198 if (ret < 0)
b31e1552
ES		 2199 ext4_msg(sb, KERN_ERR,
2200 "Cannot turn on journaled "
2201 "quota: error %d", ret);
ac27a0ec
DK		 2202 }
2203 }
2204#endif
2205
2206 while (es->s_last_orphan) {
2207 struct inode *inode;
2208
97bd42b9
JB		 2209 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2210 if (IS_ERR(inode)) {
ac27a0ec
DK		 2211 es->s_last_orphan = 0;
2212 break;
2213 }
2214
617ba13b 2215 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2216 dquot_initialize(inode);
ac27a0ec 2217 if (inode->i_nlink) {
566370a2
PT		 2218 if (test_opt(sb, DEBUG))
2219 ext4_msg(sb, KERN_DEBUG,
2220 "%s: truncating inode %lu to %lld bytes",
2221 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2222 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2223 inode->i_ino, inode->i_size);
721e3eba 2224 mutex_lock(&inode->i_mutex);
55f252c9 2225 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2226 ext4_truncate(inode);
721e3eba 2227 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2228 nr_truncates++;
2229 } else {
566370a2
PT		 2230 if (test_opt(sb, DEBUG))
2231 ext4_msg(sb, KERN_DEBUG,
2232 "%s: deleting unreferenced inode %lu",
2233 __func__, inode->i_ino);
ac27a0ec
DK		 2234 jbd_debug(2, "deleting unreferenced inode %lu\n",
2235 inode->i_ino);
2236 nr_orphans++;
2237 }
2238 iput(inode); /* The delete magic happens here! */
2239 }
2240
2b2d6d01 2241#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2242
2243 if (nr_orphans)
b31e1552
ES		 2244 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2245 PLURAL(nr_orphans));
ac27a0ec 2246 if (nr_truncates)
b31e1552
ES		 2247 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2248 PLURAL(nr_truncates));
ac27a0ec
DK		 2249#ifdef CONFIG_QUOTA
2250 /* Turn quotas off */
a2d4a646 2251 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2252 if (sb_dqopt(sb)->files[i])
287a8095 2253 dquot_quota_off(sb, i);
ac27a0ec
DK		 2254 }
2255#endif
2256 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2257}
0b8e58a1 2258
cd2291a4
ES		 2259/*
2260 * Maximal extent format file size.
2261 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2262 * extent format containers, within a sector_t, and within i_blocks
2263 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2264 * so that won't be a limiting factor.
2265 *
f17722f9
LC		 2266 * However there is other limiting factor. We do store extents in the form
2267 * of starting block and length, hence the resulting length of the extent
2268 * covering maximum file size must fit into on-disk format containers as
2269 * well. Given that length is always by 1 unit bigger than max unit (because
2270 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2271 *
cd2291a4
ES		 2272 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2273 */
f287a1a5 2274static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2275{
2276 loff_t res;
2277 loff_t upper_limit = MAX_LFS_FILESIZE;
2278
2279 /* small i_blocks in vfs inode? */
f287a1a5 2280 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2281 /*
90c699a9 2282 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2283 * i_block represent total blocks in 512 bytes
2284 * 32 == size of vfs inode i_blocks * 8
2285 */
2286 upper_limit = (1LL << 32) - 1;
2287
2288 /* total blocks in file system block size */
2289 upper_limit >>= (blkbits - 9);
2290 upper_limit <<= blkbits;
2291 }
2292
f17722f9
LC		 2293 /*
2294 * 32-bit extent-start container, ee_block. We lower the maxbytes
2295 * by one fs block, so ee_len can cover the extent of maximum file
2296 * size
2297 */
2298 res = (1LL << 32) - 1;
cd2291a4 2299 res <<= blkbits;
cd2291a4
ES		 2300
2301 /* Sanity check against vm- & vfs- imposed limits */
2302 if (res > upper_limit)
2303 res = upper_limit;
2304
2305 return res;
2306}
ac27a0ec 2307
ac27a0ec 2308/*
cd2291a4 2309 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2310 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2311 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2312 */
f287a1a5 2313static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2314{
617ba13b 2315 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2316 int meta_blocks;
2317 loff_t upper_limit;
0b8e58a1
AD		 2318 /* This is calculated to be the largest file size for a dense, block
2319 * mapped file such that the file's total number of 512-byte sectors,
2320 * including data and all indirect blocks, does not exceed (2^48 - 1).
2321 *
2322 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2323 * number of 512-byte sectors of the file.
0fc1b451
AK		 2324 */
2325
f287a1a5 2326 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2327 /*
90c699a9 2328 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2329 * the inode i_block field represents total file blocks in
2330 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2331 */
2332 upper_limit = (1LL << 32) - 1;
2333
2334 /* total blocks in file system block size */
2335 upper_limit >>= (bits - 9);
2336
2337 } else {
8180a562
AK		 2338 /*
2339 * We use 48 bit ext4_inode i_blocks
2340 * With EXT4_HUGE_FILE_FL set the i_blocks
2341 * represent total number of blocks in
2342 * file system block size
2343 */
0fc1b451
AK		 2344 upper_limit = (1LL << 48) - 1;
2345
0fc1b451
AK		 2346 }
2347
2348 /* indirect blocks */
2349 meta_blocks = 1;
2350 /* double indirect blocks */
2351 meta_blocks += 1 + (1LL << (bits-2));
2352 /* tripple indirect blocks */
2353 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2354
2355 upper_limit -= meta_blocks;
2356 upper_limit <<= bits;
ac27a0ec
DK		 2357
2358 res += 1LL << (bits-2);
2359 res += 1LL << (2*(bits-2));
2360 res += 1LL << (3*(bits-2));
2361 res <<= bits;
2362 if (res > upper_limit)
2363 res = upper_limit;
0fc1b451
AK		 2364
2365 if (res > MAX_LFS_FILESIZE)
2366 res = MAX_LFS_FILESIZE;
2367
ac27a0ec
DK		 2368 return res;
2369}
2370
617ba13b 2371static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2372 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2373{
617ba13b 2374 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2375 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2376 int has_super = 0;
2377
2378 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2379
617ba13b 2380 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2381 nr < first_meta_bg)
70bbb3e0 2382 return logical_sb_block + nr + 1;
ac27a0ec 2383 bg = sbi->s_desc_per_block * nr;
617ba13b 2384 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2385 has_super = 1;
0b8e58a1 2386
bd63f6b0
DW		 2387 /*
2388 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2389 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2390 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2391 * compensate.
2392 */
2393 if (sb->s_blocksize == 1024 && nr == 0 &&
2394 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2395 has_super++;
2396
617ba13b 2397 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2398}
2399
c9de560d
AT		 2400/**
2401 * ext4_get_stripe_size: Get the stripe size.
2402 * @sbi: In memory super block info
2403 *
2404 * If we have specified it via mount option, then
2405 * use the mount option value. If the value specified at mount time is
2406 * greater than the blocks per group use the super block value.
2407 * If the super block value is greater than blocks per group return 0.
2408 * Allocator needs it be less than blocks per group.
2409 *
2410 */
2411static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2412{
2413 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2414 unsigned long stripe_width =
2415 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2416 int ret;
c9de560d
AT		 2417
2418 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2419 ret = sbi->s_stripe;
2420 else if (stripe_width <= sbi->s_blocks_per_group)
2421 ret = stripe_width;
2422 else if (stride <= sbi->s_blocks_per_group)
2423 ret = stride;
2424 else
2425 ret = 0;
c9de560d 2426
3eb08658
DE		 2427 /*
2428 * If the stripe width is 1, this makes no sense and
2429 * we set it to 0 to turn off stripe handling code.
2430 */
2431 if (ret <= 1)
2432 ret = 0;
c9de560d 2433
3eb08658 2434 return ret;
c9de560d 2435}
ac27a0ec 2436
3197ebdb
TT		 2437/* sysfs supprt */
2438
2439struct ext4_attr {
2440 struct attribute attr;
2441 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2442 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb 2443 const char *, size_t);
f2d50a65
JK		 2444 union {
2445 int offset;
2446 int deprecated_val;
2447 } u;
3197ebdb
TT		 2448};
2449
27dd4385
LC		 2450static int parse_strtoull(const char *buf,
2451 unsigned long long max, unsigned long long *value)
2452{
2453 int ret;
2454
2455 ret = kstrtoull(skip_spaces(buf), 0, value);
2456 if (!ret && *value > max)
2457 ret = -EINVAL;
2458 return ret;
2459}
2460
3197ebdb
TT		 2461static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2462 struct ext4_sb_info *sbi,
2463 char *buf)
2464{
2465 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2466 (s64) EXT4_C2B(sbi,
2467 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2468}
2469
2470static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2471 struct ext4_sb_info *sbi, char *buf)
2472{
2473 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2474
f613dfcb
TT		 2475 if (!sb->s_bdev->bd_part)
2476 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2477 return snprintf(buf, PAGE_SIZE, "%lu\n",
2478 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2479 sbi->s_sectors_written_start) >> 1);
2480}
2481
2482static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2483 struct ext4_sb_info *sbi, char *buf)
2484{
2485 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2486
f613dfcb
TT		 2487 if (!sb->s_bdev->bd_part)
2488 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2489 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2490 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2491 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2492 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2493}
2494
2495static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2496 struct ext4_sb_info *sbi,
2497 const char *buf, size_t count)
2498{
2499 unsigned long t;
e1091b15 2500 int ret;
3197ebdb 2501
e1091b15
LC		 2502 ret = kstrtoul(skip_spaces(buf), 0, &t);
2503 if (ret)
2504 return ret;
3197ebdb 2505
e1091b15 2506 if (t && (!is_power_of_2(t) || t > 0x40000000))
3197ebdb
TT		 2507 return -EINVAL;
2508
2509 sbi->s_inode_readahead_blks = t;
2510 return count;
2511}
2512
2513static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2514 struct ext4_sb_info *sbi, char *buf)
3197ebdb 2515{
f2d50a65 2516 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb
TT		 2517
2518 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2519}
2520
2521static ssize_t sbi_ui_store(struct ext4_attr *a,
2522 struct ext4_sb_info *sbi,
2523 const char *buf, size_t count)
2524{
f2d50a65 2525 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb 2526 unsigned long t;
e1091b15 2527 int ret;
3197ebdb 2528
e1091b15
LC		 2529 ret = kstrtoul(skip_spaces(buf), 0, &t);
2530 if (ret)
2531 return ret;
3197ebdb
TT		 2532 *ui = t;
2533 return count;
2534}
2535
52c198c6
LC		 2536static ssize_t es_ui_show(struct ext4_attr *a,
2537 struct ext4_sb_info *sbi, char *buf)
2538{
2539
2540 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2541 a->u.offset);
2542
2543 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2544}
2545
27dd4385
LC		 2546static ssize_t reserved_clusters_show(struct ext4_attr *a,
2547 struct ext4_sb_info *sbi, char *buf)
2548{
2549 return snprintf(buf, PAGE_SIZE, "%llu\n",
2550 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2551}
2552
2553static ssize_t reserved_clusters_store(struct ext4_attr *a,
2554 struct ext4_sb_info *sbi,
2555 const char *buf, size_t count)
2556{
2557 unsigned long long val;
2558 int ret;
2559
2560 if (parse_strtoull(buf, -1ULL, &val))
2561 return -EINVAL;
2562 ret = ext4_reserve_clusters(sbi, val);
2563
2564 return ret ? ret : count;
2565}
2566
2c0544b2
TT		 2567static ssize_t trigger_test_error(struct ext4_attr *a,
2568 struct ext4_sb_info *sbi,
2569 const char *buf, size_t count)
2570{
2571 int len = count;
2572
2573 if (!capable(CAP_SYS_ADMIN))
2574 return -EPERM;
2575
2576 if (len && buf[len-1] == '\n')
2577 len--;
2578
2579 if (len)
2580 ext4_error(sbi->s_sb, "%.*s", len, buf);
2581 return count;
2582}
2583
f2d50a65
JK		 2584static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2585 struct ext4_sb_info *sbi, char *buf)
2586{
2587 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2588}
2589
3197ebdb
TT		 2590#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2591static struct ext4_attr ext4_attr_##_name = { \
2592 .attr = {.name = __stringify(_name), .mode = _mode }, \
2593 .show = _show, \
2594 .store = _store, \
f2d50a65
JK		 2595 .u = { \
2596 .offset = offsetof(struct ext4_sb_info, _elname),\
2597 }, \
3197ebdb 2598}
52c198c6
LC		 2599
2600#define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2601static struct ext4_attr ext4_attr_##_name = { \
2602 .attr = {.name = __stringify(_name), .mode = _mode }, \
2603 .show = _show, \
2604 .store = _store, \
2605 .u = { \
2606 .offset = offsetof(struct ext4_super_block, _elname), \
2607 }, \
2608}
2609
3197ebdb
TT		 2610#define EXT4_ATTR(name, mode, show, store) \
2611static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2612
857ac889 2613#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2614#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2615#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
52c198c6
LC		 2616
2617#define EXT4_RO_ATTR_ES_UI(name, elname) \
2618 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
3197ebdb
TT		 2619#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2620 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
52c198c6 2621
3197ebdb 2622#define ATTR_LIST(name) &ext4_attr_##name.attr
f2d50a65
JK		 2623#define EXT4_DEPRECATED_ATTR(_name, _val) \
2624static struct ext4_attr ext4_attr_##_name = { \
2625 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2626 .show = sbi_deprecated_show, \
2627 .u = { \
2628 .deprecated_val = _val, \
2629 }, \
2630}
3197ebdb
TT		 2631
2632EXT4_RO_ATTR(delayed_allocation_blocks);
2633EXT4_RO_ATTR(session_write_kbytes);
2634EXT4_RO_ATTR(lifetime_write_kbytes);
27dd4385 2635EXT4_RW_ATTR(reserved_clusters);
3197ebdb
TT		 2636EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2637 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2638EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2639EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2640EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2641EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2642EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2643EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2644EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
f2d50a65 2645EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
67a5da56 2646EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2647EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
efbed4dc
TT		 2648EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2649EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2650EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2651EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2652EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2653EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
52c198c6
LC		 2654EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2655EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2656EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
3197ebdb
TT		 2657
2658static struct attribute *ext4_attrs[] = {
2659 ATTR_LIST(delayed_allocation_blocks),
2660 ATTR_LIST(session_write_kbytes),
2661 ATTR_LIST(lifetime_write_kbytes),
27dd4385 2662 ATTR_LIST(reserved_clusters),
3197ebdb 2663 ATTR_LIST(inode_readahead_blks),
11013911 2664 ATTR_LIST(inode_goal),
3197ebdb
TT		 2665 ATTR_LIST(mb_stats),
2666 ATTR_LIST(mb_max_to_scan),
2667 ATTR_LIST(mb_min_to_scan),
2668 ATTR_LIST(mb_order2_req),
2669 ATTR_LIST(mb_stream_req),
2670 ATTR_LIST(mb_group_prealloc),
55138e0b 2671 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2672 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2673 ATTR_LIST(trigger_fs_error),
efbed4dc
TT		 2674 ATTR_LIST(err_ratelimit_interval_ms),
2675 ATTR_LIST(err_ratelimit_burst),
2676 ATTR_LIST(warning_ratelimit_interval_ms),
2677 ATTR_LIST(warning_ratelimit_burst),
2678 ATTR_LIST(msg_ratelimit_interval_ms),
2679 ATTR_LIST(msg_ratelimit_burst),
52c198c6
LC		 2680 ATTR_LIST(errors_count),
2681 ATTR_LIST(first_error_time),
2682 ATTR_LIST(last_error_time),
3197ebdb
TT		 2683 NULL,
2684};
2685
857ac889
LC		 2686/* Features this copy of ext4 supports */
2687EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2688EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2689EXT4_INFO_ATTR(meta_bg_resize);
6ddb2447 2690EXT4_INFO_ATTR(encryption);
857ac889
LC		 2691
2692static struct attribute *ext4_feat_attrs[] = {
2693 ATTR_LIST(lazy_itable_init),
27ee40df 2694 ATTR_LIST(batched_discard),
5e7bbef1 2695 ATTR_LIST(meta_bg_resize),
6ddb2447 2696 ATTR_LIST(encryption),
857ac889
LC		 2697 NULL,
2698};
2699
3197ebdb
TT		 2700static ssize_t ext4_attr_show(struct kobject *kobj,
2701 struct attribute *attr, char *buf)
2702{
2703 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2704 s_kobj);
2705 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2706
2707 return a->show ? a->show(a, sbi, buf) : 0;
2708}
2709
2710static ssize_t ext4_attr_store(struct kobject *kobj,
2711 struct attribute *attr,
2712 const char *buf, size_t len)
2713{
2714 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2715 s_kobj);
2716 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2717
2718 return a->store ? a->store(a, sbi, buf, len) : 0;
2719}
2720
2721static void ext4_sb_release(struct kobject *kobj)
2722{
2723 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2724 s_kobj);
2725 complete(&sbi->s_kobj_unregister);
2726}
2727
52cf25d0 2728static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2729 .show = ext4_attr_show,
2730 .store = ext4_attr_store,
2731};
2732
2733static struct kobj_type ext4_ktype = {
2734 .default_attrs = ext4_attrs,
2735 .sysfs_ops = &ext4_attr_ops,
2736 .release = ext4_sb_release,
2737};
2738
857ac889
LC		 2739static void ext4_feat_release(struct kobject *kobj)
2740{
2741 complete(&ext4_feat->f_kobj_unregister);
2742}
2743
c7f72543
LC		 2744static ssize_t ext4_feat_show(struct kobject *kobj,
2745 struct attribute *attr, char *buf)
2746{
2747 return snprintf(buf, PAGE_SIZE, "supported\n");
2748}
2749
2750/*
2751 * We can not use ext4_attr_show/store because it relies on the kobject
2752 * being embedded in the ext4_sb_info structure which is definitely not
2753 * true in this case.
2754 */
2755static const struct sysfs_ops ext4_feat_ops = {
2756 .show = ext4_feat_show,
2757 .store = NULL,
2758};
2759
857ac889
LC		 2760static struct kobj_type ext4_feat_ktype = {
2761 .default_attrs = ext4_feat_attrs,
c7f72543 2762 .sysfs_ops = &ext4_feat_ops,
857ac889
LC		 2763 .release = ext4_feat_release,
2764};
2765
a13fb1a4
ES		 2766/*
2767 * Check whether this filesystem can be mounted based on
2768 * the features present and the RDONLY/RDWR mount requested.
2769 * Returns 1 if this filesystem can be mounted as requested,
2770 * 0 if it cannot be.
2771 */
2772static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2773{
2774 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2775 ext4_msg(sb, KERN_ERR,
2776 "Couldn't mount because of "
2777 "unsupported optional features (%x)",
2778 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2779 ~EXT4_FEATURE_INCOMPAT_SUPP));
2780 return 0;
2781 }
2782
2783 if (readonly)
2784 return 1;
2785
2cb5cc8b
DW		 2786 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2787 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2788 sb->s_flags |= MS_RDONLY;
2789 return 1;
2790 }
2791
a13fb1a4
ES		 2792 /* Check that feature set is OK for a read-write mount */
2793 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2794 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2795 "unsupported optional features (%x)",
2796 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2797 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2798 return 0;
2799 }
2800 /*
2801 * Large file size enabled file system can only be mounted
2802 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2803 */
2804 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2805 if (sizeof(blkcnt_t) < sizeof(u64)) {
2806 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2807 "cannot be mounted RDWR without "
2808 "CONFIG_LBDAF");
2809 return 0;
2810 }
2811 }
bab08ab9
TT		 2812 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2813 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2814 ext4_msg(sb, KERN_ERR,
2815 "Can't support bigalloc feature without "
2816 "extents feature\n");
2817 return 0;
2818 }
7c319d32
AK		 2819
2820#ifndef CONFIG_QUOTA
2821 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2822 !readonly) {
2823 ext4_msg(sb, KERN_ERR,
2824 "Filesystem with quota feature cannot be mounted RDWR "
2825 "without CONFIG_QUOTA");
2826 return 0;
2827 }
2828#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2829 return 1;
2830}
2831
66e61a9e
TT		 2832/*
2833 * This function is called once a day if we have errors logged
2834 * on the file system
2835 */
2836static void print_daily_error_info(unsigned long arg)
2837{
2838 struct super_block *sb = (struct super_block *) arg;
2839 struct ext4_sb_info *sbi;
2840 struct ext4_super_block *es;
2841
2842 sbi = EXT4_SB(sb);
2843 es = sbi->s_es;
2844
2845 if (es->s_error_count)
ae0f78de
TT		 2846 /* fsck newer than v1.41.13 is needed to clean this condition. */
2847 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2848 le32_to_cpu(es->s_error_count));
2849 if (es->s_first_error_time) {
ae0f78de 2850 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2851 sb->s_id, le32_to_cpu(es->s_first_error_time),
2852 (int) sizeof(es->s_first_error_func),
2853 es->s_first_error_func,
2854 le32_to_cpu(es->s_first_error_line));
2855 if (es->s_first_error_ino)
2856 printk(": inode %u",
2857 le32_to_cpu(es->s_first_error_ino));
2858 if (es->s_first_error_block)
2859 printk(": block %llu", (unsigned long long)
2860 le64_to_cpu(es->s_first_error_block));
2861 printk("\n");
2862 }
2863 if (es->s_last_error_time) {
ae0f78de 2864 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2865 sb->s_id, le32_to_cpu(es->s_last_error_time),
2866 (int) sizeof(es->s_last_error_func),
2867 es->s_last_error_func,
2868 le32_to_cpu(es->s_last_error_line));
2869 if (es->s_last_error_ino)
2870 printk(": inode %u",
2871 le32_to_cpu(es->s_last_error_ino));
2872 if (es->s_last_error_block)
2873 printk(": block %llu", (unsigned long long)
2874 le64_to_cpu(es->s_last_error_block));
2875 printk("\n");
2876 }
2877 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2878}
2879
bfff6873
LC		 2880/* Find next suitable group and run ext4_init_inode_table */
2881static int ext4_run_li_request(struct ext4_li_request *elr)
2882{
2883 struct ext4_group_desc *gdp = NULL;
2884 ext4_group_t group, ngroups;
2885 struct super_block *sb;
2886 unsigned long timeout = 0;
2887 int ret = 0;
2888
2889 sb = elr->lr_super;
2890 ngroups = EXT4_SB(sb)->s_groups_count;
2891
8e8ad8a5 2892 sb_start_write(sb);
bfff6873
LC		 2893 for (group = elr->lr_next_group; group < ngroups; group++) {
2894 gdp = ext4_get_group_desc(sb, group, NULL);
2895 if (!gdp) {
2896 ret = 1;
2897 break;
2898 }
2899
2900 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2901 break;
2902 }
2903
7f511862 2904 if (group >= ngroups)
bfff6873
LC		 2905 ret = 1;
2906
2907 if (!ret) {
2908 timeout = jiffies;
2909 ret = ext4_init_inode_table(sb, group,
2910 elr->lr_timeout ? 0 : 1);
2911 if (elr->lr_timeout == 0) {
51ce6511
LC		 2912 timeout = (jiffies - timeout) *
2913 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2914 elr->lr_timeout = timeout;
2915 }
2916 elr->lr_next_sched = jiffies + elr->lr_timeout;
2917 elr->lr_next_group = group + 1;
2918 }
8e8ad8a5 2919 sb_end_write(sb);
bfff6873
LC		 2920
2921 return ret;
2922}
2923
2924/*
2925 * Remove lr_request from the list_request and free the
4ed5c033 2926 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2927 */
2928static void ext4_remove_li_request(struct ext4_li_request *elr)
2929{
2930 struct ext4_sb_info *sbi;
2931
2932 if (!elr)
2933 return;
2934
2935 sbi = elr->lr_sbi;
2936
2937 list_del(&elr->lr_request);
2938 sbi->s_li_request = NULL;
2939 kfree(elr);
2940}
2941
2942static void ext4_unregister_li_request(struct super_block *sb)
2943{
1bb933fb
LC		 2944 mutex_lock(&ext4_li_mtx);
2945 if (!ext4_li_info) {
2946 mutex_unlock(&ext4_li_mtx);
bfff6873 2947 return;
1bb933fb 2948 }
bfff6873
LC		 2949
2950 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2951 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2952 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2953 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2954}
2955
8f1f7453
ES		 2956static struct task_struct *ext4_lazyinit_task;
2957
bfff6873
LC		 2958/*
2959 * This is the function where ext4lazyinit thread lives. It walks
2960 * through the request list searching for next scheduled filesystem.
2961 * When such a fs is found, run the lazy initialization request
2962 * (ext4_rn_li_request) and keep track of the time spend in this
2963 * function. Based on that time we compute next schedule time of
2964 * the request. When walking through the list is complete, compute
2965 * next waking time and put itself into sleep.
2966 */
2967static int ext4_lazyinit_thread(void *arg)
2968{
2969 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2970 struct list_head *pos, *n;
2971 struct ext4_li_request *elr;
4ed5c033 2972 unsigned long next_wakeup, cur;
bfff6873
LC		 2973
2974 BUG_ON(NULL == eli);
2975
bfff6873
LC		 2976cont_thread:
2977 while (true) {
2978 next_wakeup = MAX_JIFFY_OFFSET;
2979
2980 mutex_lock(&eli->li_list_mtx);
2981 if (list_empty(&eli->li_request_list)) {
2982 mutex_unlock(&eli->li_list_mtx);
2983 goto exit_thread;
2984 }
2985
2986 list_for_each_safe(pos, n, &eli->li_request_list) {
2987 elr = list_entry(pos, struct ext4_li_request,
2988 lr_request);
2989
b2c78cd0
TT		 2990 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2991 if (ext4_run_li_request(elr) != 0) {
2992 /* error, remove the lazy_init job */
2993 ext4_remove_li_request(elr);
2994 continue;
2995 }
bfff6873
LC		 2996 }
2997
2998 if (time_before(elr->lr_next_sched, next_wakeup))
2999 next_wakeup = elr->lr_next_sched;
3000 }
3001 mutex_unlock(&eli->li_list_mtx);
3002
a0acae0e 3003 try_to_freeze();
bfff6873 3004
4ed5c033
LC		 3005 cur = jiffies;
3006 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3007 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3008 cond_resched();
3009 continue;
3010 }
3011
4ed5c033
LC		 3012 schedule_timeout_interruptible(next_wakeup - cur);
3013
8f1f7453
ES		 3014 if (kthread_should_stop()) {
3015 ext4_clear_request_list();
3016 goto exit_thread;
3017 }
bfff6873
LC		 3018 }
3019
3020exit_thread:
3021 /*
3022 * It looks like the request list is empty, but we need
3023 * to check it under the li_list_mtx lock, to prevent any
3024 * additions into it, and of course we should lock ext4_li_mtx
3025 * to atomically free the list and ext4_li_info, because at
3026 * this point another ext4 filesystem could be registering
3027 * new one.
3028 */
3029 mutex_lock(&ext4_li_mtx);
3030 mutex_lock(&eli->li_list_mtx);
3031 if (!list_empty(&eli->li_request_list)) {
3032 mutex_unlock(&eli->li_list_mtx);
3033 mutex_unlock(&ext4_li_mtx);
3034 goto cont_thread;
3035 }
3036 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3037 kfree(ext4_li_info);
3038 ext4_li_info = NULL;
3039 mutex_unlock(&ext4_li_mtx);
3040
3041 return 0;
3042}
3043
3044static void ext4_clear_request_list(void)
3045{
3046 struct list_head *pos, *n;
3047 struct ext4_li_request *elr;
3048
3049 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3050 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3051 elr = list_entry(pos, struct ext4_li_request,
3052 lr_request);
3053 ext4_remove_li_request(elr);
3054 }
3055 mutex_unlock(&ext4_li_info->li_list_mtx);
3056}
3057
3058static int ext4_run_lazyinit_thread(void)
3059{
8f1f7453
ES		 3060 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3061 ext4_li_info, "ext4lazyinit");
3062 if (IS_ERR(ext4_lazyinit_task)) {
3063 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3064 ext4_clear_request_list();
bfff6873
LC		 3065 kfree(ext4_li_info);
3066 ext4_li_info = NULL;
92b97816 3067 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3068 "initialization thread\n",
3069 err);
3070 return err;
3071 }
3072 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3073 return 0;
3074}
3075
3076/*
3077 * Check whether it make sense to run itable init. thread or not.
3078 * If there is at least one uninitialized inode table, return
3079 * corresponding group number, else the loop goes through all
3080 * groups and return total number of groups.
3081 */
3082static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3083{
3084 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3085 struct ext4_group_desc *gdp = NULL;
3086
3087 for (group = 0; group < ngroups; group++) {
3088 gdp = ext4_get_group_desc(sb, group, NULL);
3089 if (!gdp)
3090 continue;
3091
3092 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3093 break;
3094 }
3095
3096 return group;
3097}
3098
3099static int ext4_li_info_new(void)
3100{
3101 struct ext4_lazy_init *eli = NULL;
3102
3103 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3104 if (!eli)
3105 return -ENOMEM;
3106
bfff6873
LC		 3107 INIT_LIST_HEAD(&eli->li_request_list);
3108 mutex_init(&eli->li_list_mtx);
3109
bfff6873
LC		 3110 eli->li_state |= EXT4_LAZYINIT_QUIT;
3111
3112 ext4_li_info = eli;
3113
3114 return 0;
3115}
3116
3117static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3118 ext4_group_t start)
3119{
3120 struct ext4_sb_info *sbi = EXT4_SB(sb);
3121 struct ext4_li_request *elr;
bfff6873
LC		 3122
3123 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3124 if (!elr)
3125 return NULL;
3126
3127 elr->lr_super = sb;
3128 elr->lr_sbi = sbi;
3129 elr->lr_next_group = start;
3130
3131 /*
3132 * Randomize first schedule time of the request to
3133 * spread the inode table initialization requests
3134 * better.
3135 */
dd1f723b
TT		 3136 elr->lr_next_sched = jiffies + (prandom_u32() %
3137 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3138 return elr;
3139}
3140
7f511862
TT		 3141int ext4_register_li_request(struct super_block *sb,
3142 ext4_group_t first_not_zeroed)
bfff6873
LC		 3143{
3144 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3145 struct ext4_li_request *elr = NULL;
bfff6873 3146 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3147 int ret = 0;
bfff6873 3148
7f511862 3149 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3150 if (sbi->s_li_request != NULL) {
3151 /*
3152 * Reset timeout so it can be computed again, because
3153 * s_li_wait_mult might have changed.
3154 */
3155 sbi->s_li_request->lr_timeout = 0;
7f511862 3156 goto out;
51ce6511 3157 }
bfff6873
LC		 3158
3159 if (first_not_zeroed == ngroups ||
3160 (sb->s_flags & MS_RDONLY) ||
55ff3840 3161 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3162 goto out;
bfff6873
LC		 3163
3164 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3165 if (!elr) {
3166 ret = -ENOMEM;
3167 goto out;
3168 }
bfff6873
LC		 3169
3170 if (NULL == ext4_li_info) {
3171 ret = ext4_li_info_new();
3172 if (ret)
3173 goto out;
3174 }
3175
3176 mutex_lock(&ext4_li_info->li_list_mtx);
3177 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3178 mutex_unlock(&ext4_li_info->li_list_mtx);
3179
3180 sbi->s_li_request = elr;
46e4690b
TM		 3181 /*
3182 * set elr to NULL here since it has been inserted to
3183 * the request_list and the removal and free of it is
3184 * handled by ext4_clear_request_list from now on.
3185 */
3186 elr = NULL;
bfff6873
LC		 3187
3188 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3189 ret = ext4_run_lazyinit_thread();
3190 if (ret)
3191 goto out;
3192 }
bfff6873 3193out:
beed5ecb
NK		 3194 mutex_unlock(&ext4_li_mtx);
3195 if (ret)
bfff6873 3196 kfree(elr);
bfff6873
LC		 3197 return ret;
3198}
3199
3200/*
3201 * We do not need to lock anything since this is called on
3202 * module unload.
3203 */
3204static void ext4_destroy_lazyinit_thread(void)
3205{
3206 /*
3207 * If thread exited earlier
3208 * there's nothing to be done.
3209 */
8f1f7453 3210 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3211 return;
3212
8f1f7453 3213 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3214}
3215
25ed6e8a
DW		 3216static int set_journal_csum_feature_set(struct super_block *sb)
3217{
3218 int ret = 1;
3219 int compat, incompat;
3220 struct ext4_sb_info *sbi = EXT4_SB(sb);
3221
9aa5d32b 3222 if (ext4_has_metadata_csum(sb)) {
db9ee220 3223 /* journal checksum v3 */
25ed6e8a 3224 compat = 0;
db9ee220 3225 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3226 } else {
3227 /* journal checksum v1 */
3228 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3229 incompat = 0;
3230 }
3231
feb8c6d3
DW		 3232 jbd2_journal_clear_features(sbi->s_journal,
3233 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3234 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3235 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3236 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3237 ret = jbd2_journal_set_features(sbi->s_journal,
3238 compat, 0,
3239 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3240 incompat);
3241 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3242 ret = jbd2_journal_set_features(sbi->s_journal,
3243 compat, 0,
3244 incompat);
3245 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3246 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3247 } else {
feb8c6d3
DW		 3248 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3249 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3250 }
3251
3252 return ret;
3253}
3254
952fc18e
TT		 3255/*
3256 * Note: calculating the overhead so we can be compatible with
3257 * historical BSD practice is quite difficult in the face of
3258 * clusters/bigalloc. This is because multiple metadata blocks from
3259 * different block group can end up in the same allocation cluster.
3260 * Calculating the exact overhead in the face of clustered allocation
3261 * requires either O(all block bitmaps) in memory or O(number of block
3262 * groups**2) in time. We will still calculate the superblock for
3263 * older file systems --- and if we come across with a bigalloc file
3264 * system with zero in s_overhead_clusters the estimate will be close to
3265 * correct especially for very large cluster sizes --- but for newer
3266 * file systems, it's better to calculate this figure once at mkfs
3267 * time, and store it in the superblock. If the superblock value is
3268 * present (even for non-bigalloc file systems), we will use it.
3269 */
3270static int count_overhead(struct super_block *sb, ext4_group_t grp,
3271 char *buf)
3272{
3273 struct ext4_sb_info *sbi = EXT4_SB(sb);
3274 struct ext4_group_desc *gdp;
3275 ext4_fsblk_t first_block, last_block, b;
3276 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3277 int s, j, count = 0;
3278
0548bbb8
TT		 3279 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3280 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3281 sbi->s_itb_per_group + 2);
3282
952fc18e
TT		 3283 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3284 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3285 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3286 for (i = 0; i < ngroups; i++) {
3287 gdp = ext4_get_group_desc(sb, i, NULL);
3288 b = ext4_block_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_bitmap(sb, gdp);
3294 if (b >= first_block && b <= last_block) {
3295 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3296 count++;
3297 }
3298 b = ext4_inode_table(sb, gdp);
3299 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3300 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3301 int c = EXT4_B2C(sbi, b - first_block);
3302 ext4_set_bit(c, buf);
3303 count++;
3304 }
3305 if (i != grp)
3306 continue;
3307 s = 0;
3308 if (ext4_bg_has_super(sb, grp)) {
3309 ext4_set_bit(s++, buf);
3310 count++;
3311 }
3312 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3313 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3314 count++;
3315 }
3316 }
3317 if (!count)
3318 return 0;
3319 return EXT4_CLUSTERS_PER_GROUP(sb) -
3320 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3321}
3322
3323/*
3324 * Compute the overhead and stash it in sbi->s_overhead
3325 */
3326int ext4_calculate_overhead(struct super_block *sb)
3327{
3328 struct ext4_sb_info *sbi = EXT4_SB(sb);
3329 struct ext4_super_block *es = sbi->s_es;
3330 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3331 ext4_fsblk_t overhead = 0;
4fdb5543 3332 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3333
952fc18e
TT		 3334 if (!buf)
3335 return -ENOMEM;
3336
3337 /*
3338 * Compute the overhead (FS structures). This is constant
3339 * for a given filesystem unless the number of block groups
3340 * changes so we cache the previous value until it does.
3341 */
3342
3343 /*
3344 * All of the blocks before first_data_block are overhead
3345 */
3346 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3347
3348 /*
3349 * Add the overhead found in each block group
3350 */
3351 for (i = 0; i < ngroups; i++) {
3352 int blks;
3353
3354 blks = count_overhead(sb, i, buf);
3355 overhead += blks;
3356 if (blks)
3357 memset(buf, 0, PAGE_SIZE);
3358 cond_resched();
3359 }
b003b524
ES		 3360 /* Add the internal journal blocks as well */
3361 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3362 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3363
952fc18e
TT		 3364 sbi->s_overhead = overhead;
3365 smp_wmb();
3366 free_page((unsigned long) buf);
3367 return 0;
3368}
3369
27dd4385 3370
30fac0f7 3371static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
27dd4385
LC		 3372{
3373 ext4_fsblk_t resv_clusters;
3374
30fac0f7
JK		 3375 /*
3376 * There's no need to reserve anything when we aren't using extents.
3377 * The space estimates are exact, there are no unwritten extents,
3378 * hole punching doesn't need new metadata... This is needed especially
3379 * to keep ext2/3 backward compatibility.
3380 */
3381 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3382 return 0;
27dd4385
LC		 3383 /*
3384 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3385 * This should cover the situations where we can not afford to run
3386 * out of space like for example punch hole, or converting
556615dc 3387 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3388 * allocation would require 1, or 2 blocks, higher numbers are
3389 * very rare.
3390 */
30fac0f7
JK		 3391 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3392 EXT4_SB(sb)->s_cluster_bits;
27dd4385
LC		 3393
3394 do_div(resv_clusters, 50);
3395 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3396
3397 return resv_clusters;
3398}
3399
3400
3401static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3402{
3403 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3404 sbi->s_cluster_bits;
3405
3406 if (count >= clusters)
3407 return -EINVAL;
3408
3409 atomic64_set(&sbi->s_resv_clusters, count);
3410 return 0;
3411}
3412
2b2d6d01 3413static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3414{
d4c402d9 3415 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3416 struct buffer_head *bh;
617ba13b
MC		 3417 struct ext4_super_block *es = NULL;
3418 struct ext4_sb_info *sbi;
3419 ext4_fsblk_t block;
3420 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3421 ext4_fsblk_t logical_sb_block;
ac27a0ec 3422 unsigned long offset = 0;
ac27a0ec
DK		 3423 unsigned long journal_devnum = 0;
3424 unsigned long def_mount_opts;
3425 struct inode *root;
9f6200bb 3426 char *cp;
0390131b 3427 const char *descr;
dcc7dae3 3428 int ret = -ENOMEM;
281b5995 3429 int blocksize, clustersize;
4ec11028
TT		 3430 unsigned int db_count;
3431 unsigned int i;
281b5995 3432 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3433 __u64 blocks_count;
07aa2ea1 3434 int err = 0;
b3881f74 3435 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3436 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3437
3438 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3439 if (!sbi)
dcc7dae3 3440 goto out_free_orig;
705895b6
PE		 3441
3442 sbi->s_blockgroup_lock =
3443 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3444 if (!sbi->s_blockgroup_lock) {
3445 kfree(sbi);
dcc7dae3 3446 goto out_free_orig;
705895b6 3447 }
ac27a0ec 3448 sb->s_fs_info = sbi;
2c0544b2 3449 sbi->s_sb = sb;
240799cd 3450 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3451 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3452 if (sb->s_bdev->bd_part)
3453 sbi->s_sectors_written_start =
3454 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3455
9f6200bb
TT		 3456 /* Cleanup superblock name */
3457 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3458 *cp = '!';
3459
07aa2ea1 3460 /* -EINVAL is default */
dcc7dae3 3461 ret = -EINVAL;
617ba13b 3462 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3463 if (!blocksize) {
b31e1552 3464 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3465 goto out_fail;
3466 }
3467
3468 /*
617ba13b 3469 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3470 * block sizes. We need to calculate the offset from buffer start.
3471 */
617ba13b 3472 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3473 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3474 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3475 } else {
70bbb3e0 3476 logical_sb_block = sb_block;
ac27a0ec
DK		 3477 }
3478
a8ac900b 3479 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3480 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3481 goto out_fail;
3482 }
3483 /*
3484 * Note: s_es must be initialized as soon as possible because
617ba13b 3485 * some ext4 macro-instructions depend on its value
ac27a0ec 3486 */
2716b802 3487 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3488 sbi->s_es = es;
3489 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3490 if (sb->s_magic != EXT4_SUPER_MAGIC)
3491 goto cantfind_ext4;
afc32f7e 3492 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3493
feb0ab32
DW		 3494 /* Warn if metadata_csum and gdt_csum are both set. */
3495 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3496 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3497 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
363307e6 3498 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3499 "redundant flags; please run fsck.");
3500
d25425f8
DW		 3501 /* Check for a known checksum algorithm */
3502 if (!ext4_verify_csum_type(sb, es)) {
3503 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3504 "unknown checksum algorithm.");
3505 silent = 1;
3506 goto cantfind_ext4;
3507 }
3508
0441984a
DW		 3509 /* Load the checksum driver */
3510 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3511 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3512 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3513 if (IS_ERR(sbi->s_chksum_driver)) {
3514 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3515 ret = PTR_ERR(sbi->s_chksum_driver);
3516 sbi->s_chksum_driver = NULL;
3517 goto failed_mount;
3518 }
3519 }
3520
a9c47317
DW		 3521 /* Check superblock checksum */
3522 if (!ext4_superblock_csum_verify(sb, es)) {
3523 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3524 "invalid superblock checksum. Run e2fsck?");
3525 silent = 1;
3526 goto cantfind_ext4;
3527 }
3528
3529 /* Precompute checksum seed for all metadata */
9aa5d32b 3530 if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3531 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3532 sizeof(es->s_uuid));
3533
ac27a0ec
DK		 3534 /* Set defaults before we parse the mount options */
3535 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3536 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3537 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3538 set_opt(sb, DEBUG);
87f26807 3539 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3540 set_opt(sb, GRPID);
617ba13b 3541 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3542 set_opt(sb, NO_UID32);
ea663336 3543 /* xattr user namespace & acls are now defaulted on */
ea663336 3544 set_opt(sb, XATTR_USER);
03010a33 3545#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3546 set_opt(sb, POSIX_ACL);
2e7842b8 3547#endif
98c1a759
DW		 3548 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3549 if (ext4_has_metadata_csum(sb))
3550 set_opt(sb, JOURNAL_CHECKSUM);
3551
617ba13b 3552 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3553 set_opt(sb, JOURNAL_DATA);
617ba13b 3554 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3555 set_opt(sb, ORDERED_DATA);
617ba13b 3556 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3557 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3558
3559 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3560 set_opt(sb, ERRORS_PANIC);
bb4f397a 3561 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3562 set_opt(sb, ERRORS_CONT);
bb4f397a 3563 else
fd8c37ec 3564 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3565 /* block_validity enabled by default; disable with noblock_validity */
3566 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3567 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3568 set_opt(sb, DISCARD);
ac27a0ec 3569
08cefc7a
EB		 3570 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3571 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3572 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3573 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3574 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3575
8b67f04a 3576 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3577 set_opt(sb, BARRIER);
ac27a0ec 3578
dd919b98
AK		 3579 /*
3580 * enable delayed allocation by default
3581 * Use -o nodelalloc to turn it off
3582 */
bc0b75f7 3583 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3584 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3585 set_opt(sb, DELALLOC);
dd919b98 3586
51ce6511
LC		 3587 /*
3588 * set default s_li_wait_mult for lazyinit, for the case there is
3589 * no mount option specified.
3590 */
3591 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3592
8b67f04a 3593 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3594 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3595 ext4_msg(sb, KERN_WARNING,
3596 "failed to parse options in superblock: %s",
3597 sbi->s_es->s_mount_opts);
3598 }
5a916be1 3599 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3600 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3601 &journal_ioprio, 0))
ac27a0ec
DK		 3602 goto failed_mount;
3603
56889787
TT		 3604 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3605 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3606 "with data=journal disables delayed "
3607 "allocation and O_DIRECT support!\n");
3608 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3609 ext4_msg(sb, KERN_ERR, "can't mount with "
3610 "both data=journal and delalloc");
3611 goto failed_mount;
3612 }
3613 if (test_opt(sb, DIOREAD_NOLOCK)) {
3614 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3615 "both data=journal and dioread_nolock");
56889787
TT		 3616 goto failed_mount;
3617 }
923ae0ff
RZ		 3618 if (test_opt(sb, DAX)) {
3619 ext4_msg(sb, KERN_ERR, "can't mount with "
3620 "both data=journal and dax");
3621 goto failed_mount;
3622 }
56889787
TT		 3623 if (test_opt(sb, DELALLOC))
3624 clear_opt(sb, DELALLOC);
3625 }
3626
ac27a0ec 3627 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3628 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3629
617ba13b
MC		 3630 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3631 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3632 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3633 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3634 ext4_msg(sb, KERN_WARNING,
3635 "feature flags set on rev 0 fs, "
3636 "running e2fsck is recommended");
469108ff 3637
ed3654eb
TT		 3638 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3639 set_opt2(sb, HURD_COMPAT);
3640 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3641 EXT4_FEATURE_INCOMPAT_64BIT)) {
3642 ext4_msg(sb, KERN_ERR,
3643 "The Hurd can't support 64-bit file systems");
3644 goto failed_mount;
3645 }
3646 }
3647
2035e776
TT		 3648 if (IS_EXT2_SB(sb)) {
3649 if (ext2_feature_set_ok(sb))
3650 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3651 "using the ext4 subsystem");
3652 else {
3653 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3654 "to feature incompatibilities");
3655 goto failed_mount;
3656 }
3657 }
3658
3659 if (IS_EXT3_SB(sb)) {
3660 if (ext3_feature_set_ok(sb))
3661 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3662 "using the ext4 subsystem");
3663 else {
3664 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3665 "to feature incompatibilities");
3666 goto failed_mount;
3667 }
3668 }
3669
ac27a0ec
DK		 3670 /*
3671 * Check feature flags regardless of the revision level, since we
3672 * previously didn't change the revision level when setting the flags,
3673 * so there is a chance incompat flags are set on a rev 0 filesystem.
3674 */
a13fb1a4 3675 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3676 goto failed_mount;
a13fb1a4 3677
261cb20c 3678 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3679 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3680 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3681 ext4_msg(sb, KERN_ERR,
3682 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3683 goto failed_mount;
3684 }
3685
923ae0ff
RZ		 3686 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3687 if (blocksize != PAGE_SIZE) {
3688 ext4_msg(sb, KERN_ERR,
3689 "error: unsupported blocksize for dax");
3690 goto failed_mount;
3691 }
3692 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3693 ext4_msg(sb, KERN_ERR,
3694 "error: device does not support dax");
3695 goto failed_mount;
3696 }
3697 }
3698
6ddb2447
TT		 3699 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT) &&
3700 es->s_encryption_level) {
3701 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3702 es->s_encryption_level);
3703 goto failed_mount;
3704 }
3705
ac27a0ec 3706 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3707 /* Validate the filesystem blocksize */
3708 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3709 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3710 blocksize);
ac27a0ec
DK		 3711 goto failed_mount;
3712 }
3713
2b2d6d01 3714 brelse(bh);
70bbb3e0
AM		 3715 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3716 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3717 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3718 if (!bh) {
b31e1552
ES		 3719 ext4_msg(sb, KERN_ERR,
3720 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3721 goto failed_mount;
3722 }
2716b802 3723 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3724 sbi->s_es = es;
617ba13b 3725 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3726 ext4_msg(sb, KERN_ERR,
3727 "Magic mismatch, very weird!");
ac27a0ec
DK		 3728 goto failed_mount;
3729 }
3730 }
3731
a13fb1a4
ES		 3732 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3733 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3734 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3735 has_huge_files);
3736 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3737
617ba13b
MC		 3738 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3739 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3740 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3741 } else {
3742 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3743 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3744 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3745 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3746 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3747 ext4_msg(sb, KERN_ERR,
3748 "unsupported inode size: %d",
2b2d6d01 3749 sbi->s_inode_size);
ac27a0ec
DK		 3750 goto failed_mount;
3751 }
ef7f3835
KS		 3752 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3753 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3754 }
0b8e58a1 3755
0d1ee42f
AR		 3756 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3757 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3758 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3759 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3760 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3761 ext4_msg(sb, KERN_ERR,
3762 "unsupported descriptor size %lu",
0d1ee42f
AR		 3763 sbi->s_desc_size);
3764 goto failed_mount;
3765 }
3766 } else
3767 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3768
ac27a0ec 3769 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3770 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3771 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3772 goto cantfind_ext4;
0b8e58a1 3773
617ba13b 3774 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3775 if (sbi->s_inodes_per_block == 0)
617ba13b 3776 goto cantfind_ext4;
ac27a0ec
DK		 3777 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3778 sbi->s_inodes_per_block;
0d1ee42f 3779 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3780 sbi->s_sbh = bh;
3781 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3782 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3783 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3784
2b2d6d01 3785 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3786 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3787 sbi->s_def_hash_version = es->s_def_hash_version;
23301410
TT		 3788 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3789 i = le32_to_cpu(es->s_flags);
3790 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3791 sbi->s_hash_unsigned = 3;
3792 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3793#ifdef __CHAR_UNSIGNED__
23301410
TT		 3794 if (!(sb->s_flags & MS_RDONLY))
3795 es->s_flags |=
3796 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3797 sbi->s_hash_unsigned = 3;
f99b2589 3798#else
23301410
TT		 3799 if (!(sb->s_flags & MS_RDONLY))
3800 es->s_flags |=
3801 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3802#endif
23301410 3803 }
f99b2589 3804 }
ac27a0ec 3805
281b5995
TT		 3806 /* Handle clustersize */
3807 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3808 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3809 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3810 if (has_bigalloc) {
3811 if (clustersize < blocksize) {
3812 ext4_msg(sb, KERN_ERR,
3813 "cluster size (%d) smaller than "
3814 "block size (%d)", clustersize, blocksize);
3815 goto failed_mount;
3816 }
3817 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3818 le32_to_cpu(es->s_log_block_size);
3819 sbi->s_clusters_per_group =
3820 le32_to_cpu(es->s_clusters_per_group);
3821 if (sbi->s_clusters_per_group > blocksize * 8) {
3822 ext4_msg(sb, KERN_ERR,
3823 "#clusters per group too big: %lu",
3824 sbi->s_clusters_per_group);
3825 goto failed_mount;
3826 }
3827 if (sbi->s_blocks_per_group !=
3828 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3829 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3830 "clusters per group (%lu) inconsistent",
3831 sbi->s_blocks_per_group,
3832 sbi->s_clusters_per_group);
3833 goto failed_mount;
3834 }
3835 } else {
3836 if (clustersize != blocksize) {
3837 ext4_warning(sb, "fragment/cluster size (%d) != "
3838 "block size (%d)", clustersize,
3839 blocksize);
3840 clustersize = blocksize;
3841 }
3842 if (sbi->s_blocks_per_group > blocksize * 8) {
3843 ext4_msg(sb, KERN_ERR,
3844 "#blocks per group too big: %lu",
3845 sbi->s_blocks_per_group);
3846 goto failed_mount;
3847 }
3848 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3849 sbi->s_cluster_bits = 0;
ac27a0ec 3850 }
281b5995
TT		 3851 sbi->s_cluster_ratio = clustersize / blocksize;
3852
ac27a0ec 3853 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3854 ext4_msg(sb, KERN_ERR,
3855 "#inodes per group too big: %lu",
2b2d6d01 3856 sbi->s_inodes_per_group);
ac27a0ec
DK		 3857 goto failed_mount;
3858 }
3859
960fd856
TT		 3860 /* Do we have standard group size of clustersize * 8 blocks ? */
3861 if (sbi->s_blocks_per_group == clustersize << 3)
3862 set_opt2(sb, STD_GROUP_SIZE);
3863
bf43d84b
ES		 3864 /*
3865 * Test whether we have more sectors than will fit in sector_t,
3866 * and whether the max offset is addressable by the page cache.
3867 */
5a9ae68a 3868 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3869 ext4_blocks_count(es));
5a9ae68a 3870 if (err) {
b31e1552 3871 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3872 " too large to mount safely on this system");
ac27a0ec 3873 if (sizeof(sector_t) < 8)
90c699a9 3874 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3875 goto failed_mount;
3876 }
3877
617ba13b
MC		 3878 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3879 goto cantfind_ext4;
e7c95593 3880
0f2ddca6
FTN		 3881 /* check blocks count against device size */
3882 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3883 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3884 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3885 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3886 ext4_blocks_count(es), blocks_count);
3887 goto failed_mount;
3888 }
3889
0b8e58a1
AD		 3890 /*
3891 * It makes no sense for the first data block to be beyond the end
3892 * of the filesystem.
3893 */
3894 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3895 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3896 "block %u is beyond end of filesystem (%llu)",
3897 le32_to_cpu(es->s_first_data_block),
3898 ext4_blocks_count(es));
e7c95593
ES		 3899 goto failed_mount;
3900 }
bd81d8ee
LV		 3901 blocks_count = (ext4_blocks_count(es) -
3902 le32_to_cpu(es->s_first_data_block) +
3903 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3904 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3905 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3906 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3907 "(block count %llu, first data block %u, "
b31e1552 3908 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3909 ext4_blocks_count(es),
3910 le32_to_cpu(es->s_first_data_block),
3911 EXT4_BLOCKS_PER_GROUP(sb));
3912 goto failed_mount;
3913 }
bd81d8ee 3914 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3915 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3916 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3917 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3918 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3919 sbi->s_group_desc = ext4_kvmalloc(db_count *
3920 sizeof(struct buffer_head *),
3921 GFP_KERNEL);
ac27a0ec 3922 if (sbi->s_group_desc == NULL) {
b31e1552 3923 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3924 ret = -ENOMEM;
ac27a0ec
DK		 3925 goto failed_mount;
3926 }
3927
9f6200bb
TT		 3928 if (ext4_proc_root)
3929 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3930
66acdcf4
TT		 3931 if (sbi->s_proc)
3932 proc_create_data("options", S_IRUGO, sbi->s_proc,
3933 &ext4_seq_options_fops, sb);
3934
705895b6 3935 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3936
3937 for (i = 0; i < db_count; i++) {
70bbb3e0 3938 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3939 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3940 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3941 ext4_msg(sb, KERN_ERR,
3942 "can't read group descriptor %d", i);
ac27a0ec
DK		 3943 db_count = i;
3944 goto failed_mount2;
3945 }
3946 }
bfff6873 3947 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3948 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
f9ae9cf5 3949 goto failed_mount2;
ac27a0ec 3950 }
772cb7c8 3951
f9ae9cf5 3952 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3953 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3954 spin_lock_init(&sbi->s_next_gen_lock);
3955
04ecddb7
JM		 3956 setup_timer(&sbi->s_err_report, print_daily_error_info,
3957 (unsigned long) sb);
04496411 3958
a75ae78f 3959 /* Register extent status tree shrinker */
eb68d0e2 3960 if (ext4_es_register_shrinker(sbi))
ce7e010a 3961 goto failed_mount3;
ce7e010a 3962
c9de560d 3963 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3964 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3965
f9ae9cf5
TT		 3966 /*
3967 * set up enough so that it can read an inode
3968 */
f6e63f90 3969 sb->s_op = &ext4_sops;
617ba13b
MC		 3970 sb->s_export_op = &ext4_export_ops;
3971 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3972#ifdef CONFIG_QUOTA
617ba13b 3973 sb->dq_op = &ext4_quota_operations;
262b4662 3974 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
1fa5efe3 3975 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3976 else
3977 sb->s_qcop = &ext4_qctl_operations;
96c7e0d9 3978 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
ac27a0ec 3979#endif
f2fa2ffc
AK		 3980 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3981
ac27a0ec 3982 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3983 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3984
3985 sb->s_root = NULL;
3986
3987 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3988 EXT4_HAS_INCOMPAT_FEATURE(sb,
3989 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3990
c5e06d10
JL		 3991 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3992 !(sb->s_flags & MS_RDONLY))
3993 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3994 goto failed_mount3a;
c5e06d10 3995
ac27a0ec
DK		 3996 /*
3997 * The first inode we look at is the journal inode. Don't try
3998 * root first: it may be modified in the journal!
3999 */
4000 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 4001 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4002 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 4003 goto failed_mount3a;
0390131b
FM		 4004 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4005 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 4006 ext4_msg(sb, KERN_ERR, "required journal recovery "
4007 "suppressed and not mounted read-only");
744692dc 4008 goto failed_mount_wq;
ac27a0ec 4009 } else {
fd8c37ec 4010 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 4011 sbi->s_journal = NULL;
4012 needs_recovery = 0;
4013 goto no_journal;
ac27a0ec
DK		 4014 }
4015
f32aaf2d 4016 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 4017 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4018 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4019 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4020 goto failed_mount_wq;
eb40a09c
JS		 4021 }
4022
25ed6e8a
DW		 4023 if (!set_journal_csum_feature_set(sb)) {
4024 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4025 "feature set");
4026 goto failed_mount_wq;
d4da6c9c 4027 }
818d276c 4028
ac27a0ec
DK		 4029 /* We have now updated the journal if required, so we can
4030 * validate the data journaling mode. */
4031 switch (test_opt(sb, DATA_FLAGS)) {
4032 case 0:
4033 /* No mode set, assume a default based on the journal
63f57933
AM		 4034 * capabilities: ORDERED_DATA if the journal can
4035 * cope, else JOURNAL_DATA
4036 */
dab291af
MC		 4037 if (jbd2_journal_check_available_features
4038 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 4039 set_opt(sb, ORDERED_DATA);
ac27a0ec 4040 else
fd8c37ec 4041 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 4042 break;
4043
617ba13b
MC		 4044 case EXT4_MOUNT_ORDERED_DATA:
4045 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4046 if (!jbd2_journal_check_available_features
4047 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4048 ext4_msg(sb, KERN_ERR, "Journal does not support "
4049 "requested data journaling mode");
744692dc 4050 goto failed_mount_wq;
ac27a0ec
DK		 4051 }
4052 default:
4053 break;
4054 }
b3881f74 4055 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4056
18aadd47
BJ		 4057 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4058
ce7e010a 4059no_journal:
9c191f70
M		 4060 if (ext4_mballoc_ready) {
4061 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4062 if (!sbi->s_mb_cache) {
4063 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4064 goto failed_mount_wq;
4065 }
4066 }
4067
6ddb2447
TT		 4068 if (unlikely(sbi->s_mount_flags & EXT4_MF_TEST_DUMMY_ENCRYPTION) &&
4069 !(sb->s_flags & MS_RDONLY) &&
4070 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) {
4071 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
4072 ext4_commit_super(sb, 1);
4073 }
4074
952fc18e
TT		 4075 /*
4076 * Get the # of file system overhead blocks from the
4077 * superblock if present.
4078 */
4079 if (es->s_overhead_clusters)
4080 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4081 else {
07aa2ea1
LC		 4082 err = ext4_calculate_overhead(sb);
4083 if (err)
952fc18e
TT		 4084 goto failed_mount_wq;
4085 }
4086
fd89d5f2
TH		 4087 /*
4088 * The maximum number of concurrent works can be high and
4089 * concurrency isn't really necessary. Limit it to 1.
4090 */
2e8fa54e
JK		 4091 EXT4_SB(sb)->rsv_conversion_wq =
4092 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4093 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4094 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4095 ret = -ENOMEM;
2e8fa54e
JK		 4096 goto failed_mount4;
4097 }
4098
ac27a0ec 4099 /*
dab291af 4100 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4101 * so we can safely mount the rest of the filesystem now.
4102 */
4103
1d1fe1ee
DH		 4104 root = ext4_iget(sb, EXT4_ROOT_INO);
4105 if (IS_ERR(root)) {
b31e1552 4106 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4107 ret = PTR_ERR(root);
32a9bb57 4108 root = NULL;
ac27a0ec
DK		 4109 goto failed_mount4;
4110 }
4111 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4112 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4113 iput(root);
ac27a0ec
DK		 4114 goto failed_mount4;
4115 }
48fde701 4116 sb->s_root = d_make_root(root);
1d1fe1ee 4117 if (!sb->s_root) {
b31e1552 4118 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4119 ret = -ENOMEM;
4120 goto failed_mount4;
4121 }
ac27a0ec 4122
7e84b621
ES		 4123 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4124 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4125
4126 /* determine the minimum size of new large inodes, if present */
4127 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4128 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4129 EXT4_GOOD_OLD_INODE_SIZE;
4130 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4131 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4132 if (sbi->s_want_extra_isize <
4133 le16_to_cpu(es->s_want_extra_isize))
4134 sbi->s_want_extra_isize =
4135 le16_to_cpu(es->s_want_extra_isize);
4136 if (sbi->s_want_extra_isize <
4137 le16_to_cpu(es->s_min_extra_isize))
4138 sbi->s_want_extra_isize =
4139 le16_to_cpu(es->s_min_extra_isize);
4140 }
4141 }
4142 /* Check if enough inode space is available */
4143 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4144 sbi->s_inode_size) {
4145 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4146 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4147 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4148 "available");
ef7f3835
KS		 4149 }
4150
30fac0f7 4151 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
27dd4385
LC		 4152 if (err) {
4153 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
30fac0f7 4154 "reserved pool", ext4_calculate_resv_clusters(sb));
f9ae9cf5 4155 goto failed_mount4a;
27dd4385
LC		 4156 }
4157
6fd058f7
TT		 4158 err = ext4_setup_system_zone(sb);
4159 if (err) {
b31e1552 4160 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4161 "zone (%d)", err);
f9ae9cf5
TT		 4162 goto failed_mount4a;
4163 }
4164
4165 ext4_ext_init(sb);
4166 err = ext4_mb_init(sb);
4167 if (err) {
4168 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4169 err);
dcf2d804 4170 goto failed_mount5;
c2774d84
AK		 4171 }
4172
d5e03cbb
TT		 4173 block = ext4_count_free_clusters(sb);
4174 ext4_free_blocks_count_set(sbi->s_es,
4175 EXT4_C2B(sbi, block));
908c7f19
TH		 4176 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4177 GFP_KERNEL);
d5e03cbb
TT		 4178 if (!err) {
4179 unsigned long freei = ext4_count_free_inodes(sb);
4180 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4181 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4182 GFP_KERNEL);
d5e03cbb
TT		 4183 }
4184 if (!err)
4185 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4186 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4187 if (!err)
908c7f19
TH		 4188 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4189 GFP_KERNEL);
d5e03cbb
TT		 4190 if (err) {
4191 ext4_msg(sb, KERN_ERR, "insufficient memory");
4192 goto failed_mount6;
4193 }
4194
4195 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4196 if (!ext4_fill_flex_info(sb)) {
4197 ext4_msg(sb, KERN_ERR,
4198 "unable to initialize "
4199 "flex_bg meta info!");
4200 goto failed_mount6;
4201 }
4202
bfff6873
LC		 4203 err = ext4_register_li_request(sb, first_not_zeroed);
4204 if (err)
dcf2d804 4205 goto failed_mount6;
bfff6873 4206
3197ebdb
TT		 4207 sbi->s_kobj.kset = ext4_kset;
4208 init_completion(&sbi->s_kobj_unregister);
4209 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4210 "%s", sb->s_id);
dcf2d804
TM		 4211 if (err)
4212 goto failed_mount7;
3197ebdb 4213
9b2ff357
JK		 4214#ifdef CONFIG_QUOTA
4215 /* Enable quota usage during mount. */
4216 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4217 !(sb->s_flags & MS_RDONLY)) {
4218 err = ext4_enable_quotas(sb);
4219 if (err)
4220 goto failed_mount8;
4221 }
4222#endif /* CONFIG_QUOTA */
4223
617ba13b
MC		 4224 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4225 ext4_orphan_cleanup(sb, es);
4226 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4227 if (needs_recovery) {
b31e1552 4228 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4229 ext4_mark_recovery_complete(sb, es);
4230 }
4231 if (EXT4_SB(sb)->s_journal) {
4232 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4233 descr = " journalled data mode";
4234 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4235 descr = " ordered data mode";
4236 else
4237 descr = " writeback data mode";
4238 } else
4239 descr = "out journal";
4240
79add3a3
LC		 4241 if (test_opt(sb, DISCARD)) {
4242 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4243 if (!blk_queue_discard(q))
4244 ext4_msg(sb, KERN_WARNING,
4245 "mounting with \"discard\" option, but "
4246 "the device does not support discard");
4247 }
4248
d4c402d9 4249 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4250 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4251 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4252
66e61a9e
TT		 4253 if (es->s_error_count)
4254 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4255
efbed4dc
TT		 4256 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4257 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4258 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4259 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4260
d4c402d9 4261 kfree(orig_data);
ac27a0ec
DK		 4262 return 0;
4263
617ba13b 4264cantfind_ext4:
ac27a0ec 4265 if (!silent)
b31e1552 4266 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4267 goto failed_mount;
4268
72ba7450
TT		 4269#ifdef CONFIG_QUOTA
4270failed_mount8:
4271 kobject_del(&sbi->s_kobj);
4272#endif
dcf2d804
TM		 4273failed_mount7:
4274 ext4_unregister_li_request(sb);
4275failed_mount6:
f9ae9cf5 4276 ext4_mb_release(sb);
d5e03cbb 4277 if (sbi->s_flex_groups)
b93b41d4 4278 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4279 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4280 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4281 percpu_counter_destroy(&sbi->s_dirs_counter);
4282 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4283failed_mount5:
f9ae9cf5
TT		 4284 ext4_ext_release(sb);
4285 ext4_release_system_zone(sb);
4286failed_mount4a:
94bf608a 4287 dput(sb->s_root);
32a9bb57 4288 sb->s_root = NULL;
94bf608a 4289failed_mount4:
b31e1552 4290 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4291 if (EXT4_SB(sb)->rsv_conversion_wq)
4292 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4293failed_mount_wq:
0390131b
FM		 4294 if (sbi->s_journal) {
4295 jbd2_journal_destroy(sbi->s_journal);
4296 sbi->s_journal = NULL;
4297 }
50460fe8 4298failed_mount3a:
d3922a77 4299 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4300failed_mount3:
9105bb14 4301 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4302 if (sbi->s_mmp_tsk)
4303 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4304failed_mount2:
4305 for (i = 0; i < db_count; i++)
4306 brelse(sbi->s_group_desc[i]);
b93b41d4 4307 kvfree(sbi->s_group_desc);
ac27a0ec 4308failed_mount:
0441984a
DW		 4309 if (sbi->s_chksum_driver)
4310 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4311 if (sbi->s_proc) {
66acdcf4 4312 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4313 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4314 }
ac27a0ec 4315#ifdef CONFIG_QUOTA
a2d4a646 4316 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4317 kfree(sbi->s_qf_names[i]);
4318#endif
617ba13b 4319 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4320 brelse(bh);
4321out_fail:
4322 sb->s_fs_info = NULL;
f6830165 4323 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4324 kfree(sbi);
dcc7dae3 4325out_free_orig:
d4c402d9 4326 kfree(orig_data);
07aa2ea1 4327 return err ? err : ret;
ac27a0ec
DK		 4328}
4329
4330/*
4331 * Setup any per-fs journal parameters now. We'll do this both on
4332 * initial mount, once the journal has been initialised but before we've
4333 * done any recovery; and again on any subsequent remount.
4334 */
617ba13b 4335static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4336{
617ba13b 4337 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4338
30773840
TT		 4339 journal->j_commit_interval = sbi->s_commit_interval;
4340 journal->j_min_batch_time = sbi->s_min_batch_time;
4341 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4342
a931da6a 4343 write_lock(&journal->j_state_lock);
ac27a0ec 4344 if (test_opt(sb, BARRIER))
dab291af 4345 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4346 else
dab291af 4347 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4348 if (test_opt(sb, DATA_ERR_ABORT))
4349 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4350 else
4351 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4352 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4353}
4354
617ba13b 4355static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4356 unsigned int journal_inum)
4357{
4358 struct inode *journal_inode;
4359 journal_t *journal;
4360
0390131b
FM		 4361 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4362
ac27a0ec
DK		 4363 /* First, test for the existence of a valid inode on disk. Bad
4364 * things happen if we iget() an unused inode, as the subsequent
4365 * iput() will try to delete it. */
4366
1d1fe1ee
DH		 4367 journal_inode = ext4_iget(sb, journal_inum);
4368 if (IS_ERR(journal_inode)) {
b31e1552 4369 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4370 return NULL;
4371 }
4372 if (!journal_inode->i_nlink) {
4373 make_bad_inode(journal_inode);
4374 iput(journal_inode);
b31e1552 4375 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4376 return NULL;
4377 }
4378
e5f8eab8 4379 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4380 journal_inode, journal_inode->i_size);
1d1fe1ee 4381 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4382 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4383 iput(journal_inode);
4384 return NULL;
4385 }
4386
dab291af 4387 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4388 if (!journal) {
b31e1552 4389 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4390 iput(journal_inode);
4391 return NULL;
4392 }
4393 journal->j_private = sb;
617ba13b 4394 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4395 return journal;
4396}
4397
617ba13b 4398static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4399 dev_t j_dev)
4400{
2b2d6d01 4401 struct buffer_head *bh;
ac27a0ec 4402 journal_t *journal;
617ba13b
MC		 4403 ext4_fsblk_t start;
4404 ext4_fsblk_t len;
ac27a0ec 4405 int hblock, blocksize;
617ba13b 4406 ext4_fsblk_t sb_block;
ac27a0ec 4407 unsigned long offset;
2b2d6d01 4408 struct ext4_super_block *es;
ac27a0ec
DK		 4409 struct block_device *bdev;
4410
0390131b
FM		 4411 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4412
b31e1552 4413 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4414 if (bdev == NULL)
4415 return NULL;
4416
ac27a0ec 4417 blocksize = sb->s_blocksize;
e1defc4f 4418 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4419 if (blocksize < hblock) {
b31e1552
ES		 4420 ext4_msg(sb, KERN_ERR,
4421 "blocksize too small for journal device");
ac27a0ec
DK		 4422 goto out_bdev;
4423 }
4424
617ba13b
MC		 4425 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4426 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4427 set_blocksize(bdev, blocksize);
4428 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4429 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4430 "external journal");
ac27a0ec
DK		 4431 goto out_bdev;
4432 }
4433
2716b802 4434 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4435 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4436 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4437 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4438 ext4_msg(sb, KERN_ERR, "external journal has "
4439 "bad superblock");
ac27a0ec
DK		 4440 brelse(bh);
4441 goto out_bdev;
4442 }
4443
df4763be
DW		 4444 if ((le32_to_cpu(es->s_feature_ro_compat) &
4445 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4446 es->s_checksum != ext4_superblock_csum(sb, es)) {
4447 ext4_msg(sb, KERN_ERR, "external journal has "
4448 "corrupt superblock");
4449 brelse(bh);
4450 goto out_bdev;
4451 }
4452
617ba13b 4453 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4454 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4455 brelse(bh);
4456 goto out_bdev;
4457 }
4458
bd81d8ee 4459 len = ext4_blocks_count(es);
ac27a0ec
DK		 4460 start = sb_block + 1;
4461 brelse(bh); /* we're done with the superblock */
4462
dab291af 4463 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4464 start, len, blocksize);
4465 if (!journal) {
b31e1552 4466 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4467 goto out_bdev;
4468 }
4469 journal->j_private = sb;
9f203507 4470 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4471 wait_on_buffer(journal->j_sb_buffer);
4472 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4473 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4474 goto out_journal;
4475 }
4476 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4477 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4478 "user (unsupported) - %d",
ac27a0ec
DK		 4479 be32_to_cpu(journal->j_superblock->s_nr_users));
4480 goto out_journal;
4481 }
617ba13b
MC		 4482 EXT4_SB(sb)->journal_bdev = bdev;
4483 ext4_init_journal_params(sb, journal);
ac27a0ec 4484 return journal;
0b8e58a1 4485
ac27a0ec 4486out_journal:
dab291af 4487 jbd2_journal_destroy(journal);
ac27a0ec 4488out_bdev:
617ba13b 4489 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4490 return NULL;
4491}
4492
617ba13b
MC		 4493static int ext4_load_journal(struct super_block *sb,
4494 struct ext4_super_block *es,
ac27a0ec
DK		 4495 unsigned long journal_devnum)
4496{
4497 journal_t *journal;
4498 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4499 dev_t journal_dev;
4500 int err = 0;
4501 int really_read_only;
4502
0390131b
FM		 4503 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4504
ac27a0ec
DK		 4505 if (journal_devnum &&
4506 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4507 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4508 "numbers have changed");
ac27a0ec
DK		 4509 journal_dev = new_decode_dev(journal_devnum);
4510 } else
4511 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4512
4513 really_read_only = bdev_read_only(sb->s_bdev);
4514
4515 /*
4516 * Are we loading a blank journal or performing recovery after a
4517 * crash? For recovery, we need to check in advance whether we
4518 * can get read-write access to the device.
4519 */
617ba13b 4520 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4521 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4522 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4523 "required on readonly filesystem");
ac27a0ec 4524 if (really_read_only) {
b31e1552
ES		 4525 ext4_msg(sb, KERN_ERR, "write access "
4526 "unavailable, cannot proceed");
ac27a0ec
DK		 4527 return -EROFS;
4528 }
b31e1552
ES		 4529 ext4_msg(sb, KERN_INFO, "write access will "
4530 "be enabled during recovery");
ac27a0ec
DK		 4531 }
4532 }
4533
4534 if (journal_inum && journal_dev) {
b31e1552
ES		 4535 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4536 "and inode journals!");
ac27a0ec
DK		 4537 return -EINVAL;
4538 }
4539
4540 if (journal_inum) {
617ba13b 4541 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4542 return -EINVAL;
4543 } else {
617ba13b 4544 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4545 return -EINVAL;
4546 }
4547
90576c0b 4548 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4549 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4550
617ba13b 4551 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4552 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4553 if (!err) {
4554 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4555 if (save)
4556 memcpy(save, ((char *) es) +
4557 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4558 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4559 if (save)
4560 memcpy(((char *) es) + EXT4_S_ERR_START,
4561 save, EXT4_S_ERR_LEN);
4562 kfree(save);
4563 }
ac27a0ec
DK		 4564
4565 if (err) {
b31e1552 4566 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4567 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4568 return err;
4569 }
4570
617ba13b
MC		 4571 EXT4_SB(sb)->s_journal = journal;
4572 ext4_clear_journal_err(sb, es);
ac27a0ec 4573
c41303ce 4574 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4575 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4576 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4577
4578 /* Make sure we flush the recovery flag to disk. */
e2d67052 4579 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4580 }
4581
4582 return 0;
4583}
4584
e2d67052 4585static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4586{
e2d67052 4587 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4588 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4589 int error = 0;
ac27a0ec 4590
08439fec 4591 if (!sbh)
c4be0c1d 4592 return error;
914258bf
TT		 4593 if (buffer_write_io_error(sbh)) {
4594 /*
4595 * Oh, dear. A previous attempt to write the
4596 * superblock failed. This could happen because the
4597 * USB device was yanked out. Or it could happen to
4598 * be a transient write error and maybe the block will
4599 * be remapped. Nothing we can do but to retry the
4600 * write and hope for the best.
4601 */
b31e1552
ES		 4602 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4603 "superblock detected");
914258bf
TT		 4604 clear_buffer_write_io_error(sbh);
4605 set_buffer_uptodate(sbh);
4606 }
71290b36
TT		 4607 /*
4608 * If the file system is mounted read-only, don't update the
4609 * superblock write time. This avoids updating the superblock
4610 * write time when we are mounting the root file system
4611 * read/only but we need to replay the journal; at that point,
4612 * for people who are east of GMT and who make their clock
4613 * tick in localtime for Windows bug-for-bug compatibility,
4614 * the clock is set in the future, and this will cause e2fsck
4615 * to complain and force a full file system check.
4616 */
4617 if (!(sb->s_flags & MS_RDONLY))
4618 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4619 if (sb->s_bdev->bd_part)
4620 es->s_kbytes_written =
4621 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4622 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4623 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4624 else
4625 es->s_kbytes_written =
4626 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4627 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4628 ext4_free_blocks_count_set(es,
57042651
TT		 4629 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4630 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4631 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4632 es->s_free_inodes_count =
4633 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4634 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4635 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4636 ext4_superblock_csum_set(sb);
ac27a0ec 4637 mark_buffer_dirty(sbh);
914258bf 4638 if (sync) {
c4be0c1d
TS		 4639 error = sync_dirty_buffer(sbh);
4640 if (error)
4641 return error;
4642
4643 error = buffer_write_io_error(sbh);
4644 if (error) {
b31e1552
ES		 4645 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4646 "superblock");
914258bf
TT		 4647 clear_buffer_write_io_error(sbh);
4648 set_buffer_uptodate(sbh);
4649 }
4650 }
c4be0c1d 4651 return error;
ac27a0ec
DK		 4652}
4653
ac27a0ec
DK		 4654/*
4655 * Have we just finished recovery? If so, and if we are mounting (or
4656 * remounting) the filesystem readonly, then we will end up with a
4657 * consistent fs on disk. Record that fact.
4658 */
2b2d6d01
TT		 4659static void ext4_mark_recovery_complete(struct super_block *sb,
4660 struct ext4_super_block *es)
ac27a0ec 4661{
617ba13b 4662 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4663
0390131b
FM		 4664 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4665 BUG_ON(journal != NULL);
4666 return;
4667 }
dab291af 4668 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4669 if (jbd2_journal_flush(journal) < 0)
4670 goto out;
4671
617ba13b 4672 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4673 sb->s_flags & MS_RDONLY) {
617ba13b 4674 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4675 ext4_commit_super(sb, 1);
ac27a0ec 4676 }
7ffe1ea8
HK		 4677
4678out:
dab291af 4679 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4680}
4681
4682/*
4683 * If we are mounting (or read-write remounting) a filesystem whose journal
4684 * has recorded an error from a previous lifetime, move that error to the
4685 * main filesystem now.
4686 */
2b2d6d01
TT		 4687static void ext4_clear_journal_err(struct super_block *sb,
4688 struct ext4_super_block *es)
ac27a0ec
DK		 4689{
4690 journal_t *journal;
4691 int j_errno;
4692 const char *errstr;
4693
0390131b
FM		 4694 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4695
617ba13b 4696 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4697
4698 /*
4699 * Now check for any error status which may have been recorded in the
617ba13b 4700 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4701 */
4702
dab291af 4703 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4704 if (j_errno) {
4705 char nbuf[16];
4706
617ba13b 4707 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4708 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4709 "from previous mount: %s", errstr);
12062ddd 4710 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4711
617ba13b
MC		 4712 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4713 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4714 ext4_commit_super(sb, 1);
ac27a0ec 4715
dab291af 4716 jbd2_journal_clear_err(journal);
d796c52e 4717 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4718 }
4719}
4720
4721/*
4722 * Force the running and committing transactions to commit,
4723 * and wait on the commit.
4724 */
617ba13b 4725int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4726{
4727 journal_t *journal;
ac27a0ec
DK		 4728
4729 if (sb->s_flags & MS_RDONLY)
4730 return 0;
4731
617ba13b 4732 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4733 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4734}
4735
617ba13b 4736static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4737{
14ce0cb4 4738 int ret = 0;
9eddacf9 4739 tid_t target;
06a407f1 4740 bool needs_barrier = false;
8d5d02e6 4741 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4742
9bffad1e 4743 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4744 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4745 /*
4746 * Writeback quota in non-journalled quota case - journalled quota has
4747 * no dirty dquots
4748 */
4749 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4750 /*
4751 * Data writeback is possible w/o journal transaction, so barrier must
4752 * being sent at the end of the function. But we can skip it if
4753 * transaction_commit will do it for us.
4754 */
bda32530
TT		 4755 if (sbi->s_journal) {
4756 target = jbd2_get_latest_transaction(sbi->s_journal);
4757 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4758 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4759 needs_barrier = true;
4760
4761 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4762 if (wait)
4763 ret = jbd2_log_wait_commit(sbi->s_journal,
4764 target);
4765 }
4766 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4767 needs_barrier = true;
06a407f1
DM		 4768 if (needs_barrier) {
4769 int err;
4770 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4771 if (!ret)
4772 ret = err;
0390131b 4773 }
06a407f1
DM		 4774
4775 return ret;
4776}
4777
ac27a0ec
DK		 4778/*
4779 * LVM calls this function before a (read-only) snapshot is created. This
4780 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4781 *
4782 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4783 * state independently. It relies on upper layer to stop all data & metadata
4784 * modifications.
ac27a0ec 4785 */
c4be0c1d 4786static int ext4_freeze(struct super_block *sb)
ac27a0ec 4787{
c4be0c1d
TS		 4788 int error = 0;
4789 journal_t *journal;
ac27a0ec 4790
9ca92389
TT		 4791 if (sb->s_flags & MS_RDONLY)
4792 return 0;
ac27a0ec 4793
9ca92389 4794 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4795
bb044576
TT		 4796 if (journal) {
4797 /* Now we set up the journal barrier. */
4798 jbd2_journal_lock_updates(journal);
ac27a0ec 4799
bb044576
TT		 4800 /*
4801 * Don't clear the needs_recovery flag if we failed to
4802 * flush the journal.
4803 */
4804 error = jbd2_journal_flush(journal);
4805 if (error < 0)
4806 goto out;
4807 }
9ca92389
TT		 4808
4809 /* Journal blocked and flushed, clear needs_recovery flag. */
4810 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4811 error = ext4_commit_super(sb, 1);
6b0310fb 4812out:
bb044576
TT		 4813 if (journal)
4814 /* we rely on upper layer to stop further updates */
4815 jbd2_journal_unlock_updates(journal);
6b0310fb 4816 return error;
ac27a0ec
DK		 4817}
4818
4819/*
4820 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4821 * flag here, even though the filesystem is not technically dirty yet.
4822 */
c4be0c1d 4823static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4824{
9ca92389
TT		 4825 if (sb->s_flags & MS_RDONLY)
4826 return 0;
4827
9ca92389
TT		 4828 /* Reset the needs_recovery flag before the fs is unlocked. */
4829 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4830 ext4_commit_super(sb, 1);
c4be0c1d 4831 return 0;
ac27a0ec
DK		 4832}
4833
673c6100
TT		 4834/*
4835 * Structure to save mount options for ext4_remount's benefit
4836 */
4837struct ext4_mount_options {
4838 unsigned long s_mount_opt;
a2595b8a 4839 unsigned long s_mount_opt2;
08cefc7a
EB		 4840 kuid_t s_resuid;
4841 kgid_t s_resgid;
673c6100
TT		 4842 unsigned long s_commit_interval;
4843 u32 s_min_batch_time, s_max_batch_time;
4844#ifdef CONFIG_QUOTA
4845 int s_jquota_fmt;
a2d4a646 4846 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4847#endif
4848};
4849
2b2d6d01 4850static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4851{
2b2d6d01 4852 struct ext4_super_block *es;
617ba13b 4853 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4854 unsigned long old_sb_flags;
617ba13b 4855 struct ext4_mount_options old_opts;
c79d967d 4856 int enable_quota = 0;
8a266467 4857 ext4_group_t g;
b3881f74 4858 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4859 int err = 0;
ac27a0ec 4860#ifdef CONFIG_QUOTA
03dafb5f 4861 int i, j;
ac27a0ec 4862#endif
d4c402d9 4863 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4864
4865 /* Store the original options */
4866 old_sb_flags = sb->s_flags;
4867 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4868 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4869 old_opts.s_resuid = sbi->s_resuid;
4870 old_opts.s_resgid = sbi->s_resgid;
4871 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4872 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4873 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4874#ifdef CONFIG_QUOTA
4875 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4876 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4877 if (sbi->s_qf_names[i]) {
4878 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4879 GFP_KERNEL);
4880 if (!old_opts.s_qf_names[i]) {
4881 for (j = 0; j < i; j++)
4882 kfree(old_opts.s_qf_names[j]);
3e36a163 4883 kfree(orig_data);
03dafb5f
CG		 4884 return -ENOMEM;
4885 }
4886 } else
4887 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4888#endif
b3881f74
TT		 4889 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4890 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4891
661aa520 4892 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4893 err = -EINVAL;
4894 goto restore_opts;
4895 }
4896
6b992ff2 4897 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4898 test_opt(sb, JOURNAL_CHECKSUM)) {
4899 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4900 "during remount not supported; ignoring");
4901 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4902 }
4903
6ae6514b
PS		 4904 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4905 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4906 ext4_msg(sb, KERN_ERR, "can't mount with "
4907 "both data=journal and delalloc");
4908 err = -EINVAL;
4909 goto restore_opts;
4910 }
4911 if (test_opt(sb, DIOREAD_NOLOCK)) {
4912 ext4_msg(sb, KERN_ERR, "can't mount with "
4913 "both data=journal and dioread_nolock");
4914 err = -EINVAL;
4915 goto restore_opts;
4916 }
923ae0ff
RZ		 4917 if (test_opt(sb, DAX)) {
4918 ext4_msg(sb, KERN_ERR, "can't mount with "
4919 "both data=journal and dax");
4920 err = -EINVAL;
4921 goto restore_opts;
4922 }
4923 }
4924
4925 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4926 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4927 "dax flag with busy inodes while remounting");
4928 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 4929 }
4930
4ab2f15b 4931 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4932 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4933
4934 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4935 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4936
4937 es = sbi->s_es;
4938
b3881f74 4939 if (sbi->s_journal) {
0390131b 4940 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4941 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4942 }
ac27a0ec 4943
661aa520 4944 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4945 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4946 err = -EROFS;
4947 goto restore_opts;
4948 }
4949
4950 if (*flags & MS_RDONLY) {
38c03b34
TT		 4951 err = sync_filesystem(sb);
4952 if (err < 0)
4953 goto restore_opts;
0f0dd62f
CH		 4954 err = dquot_suspend(sb, -1);
4955 if (err < 0)
c79d967d 4956 goto restore_opts;
c79d967d 4957
ac27a0ec
DK		 4958 /*
4959 * First of all, the unconditional stuff we have to do
4960 * to disable replay of the journal when we next remount
4961 */
4962 sb->s_flags |= MS_RDONLY;
4963
4964 /*
4965 * OK, test if we are remounting a valid rw partition
4966 * readonly, and if so set the rdonly flag and then
4967 * mark the partition as valid again.
4968 */
617ba13b
MC		 4969 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4970 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4971 es->s_state = cpu_to_le16(sbi->s_mount_state);
4972
a63c9eb2 4973 if (sbi->s_journal)
0390131b 4974 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4975 } else {
a13fb1a4 4976 /* Make sure we can mount this feature set readwrite */
2cb5cc8b
DW		 4977 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4978 EXT4_FEATURE_RO_COMPAT_READONLY) ||
4979 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4980 err = -EROFS;
4981 goto restore_opts;
4982 }
8a266467
TT		 4983 /*
4984 * Make sure the group descriptor checksums
0b8e58a1 4985 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4986 */
4987 for (g = 0; g < sbi->s_groups_count; g++) {
4988 struct ext4_group_desc *gdp =
4989 ext4_get_group_desc(sb, g, NULL);
4990
feb0ab32 4991 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4992 ext4_msg(sb, KERN_ERR,
4993 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4994 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4995 le16_to_cpu(gdp->bg_checksum));
4996 err = -EINVAL;
4997 goto restore_opts;
4998 }
4999 }
5000
ead6596b
ES		 5001 /*
5002 * If we have an unprocessed orphan list hanging
5003 * around from a previously readonly bdev mount,
5004 * require a full umount/remount for now.
5005 */
5006 if (es->s_last_orphan) {
b31e1552 5007 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5008 "remount RDWR because of unprocessed "
5009 "orphan inode list. Please "
b31e1552 5010 "umount/remount instead");
ead6596b
ES		 5011 err = -EINVAL;
5012 goto restore_opts;
5013 }
5014
ac27a0ec
DK		 5015 /*
5016 * Mounting a RDONLY partition read-write, so reread
5017 * and store the current valid flag. (It may have
5018 * been changed by e2fsck since we originally mounted
5019 * the partition.)
5020 */
0390131b
FM		 5021 if (sbi->s_journal)
5022 ext4_clear_journal_err(sb, es);
ac27a0ec 5023 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 5024 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 5025 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 5026 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
5027 EXT4_FEATURE_INCOMPAT_MMP))
5028 if (ext4_multi_mount_protect(sb,
5029 le64_to_cpu(es->s_mmp_block))) {
5030 err = -EROFS;
5031 goto restore_opts;
5032 }
c79d967d 5033 enable_quota = 1;
ac27a0ec
DK		 5034 }
5035 }
bfff6873
LC		 5036
5037 /*
5038 * Reinitialize lazy itable initialization thread based on
5039 * current settings
5040 */
5041 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5042 ext4_unregister_li_request(sb);
5043 else {
5044 ext4_group_t first_not_zeroed;
5045 first_not_zeroed = ext4_has_uninit_itable(sb);
5046 ext4_register_li_request(sb, first_not_zeroed);
5047 }
5048
6fd058f7 5049 ext4_setup_system_zone(sb);
d096ad0f 5050 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 5051 ext4_commit_super(sb, 1);
0390131b 5052
ac27a0ec
DK		 5053#ifdef CONFIG_QUOTA
5054 /* Release old quota file names */
a2d4a646 5055 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5056 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5057 if (enable_quota) {
5058 if (sb_any_quota_suspended(sb))
5059 dquot_resume(sb, -1);
5060 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5061 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5062 err = ext4_enable_quotas(sb);
07724f98 5063 if (err)
7c319d32 5064 goto restore_opts;
7c319d32
AK		 5065 }
5066 }
ac27a0ec 5067#endif
d4c402d9 5068
a26f4992 5069 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 5070 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5071 kfree(orig_data);
ac27a0ec 5072 return 0;
0b8e58a1 5073
ac27a0ec
DK		 5074restore_opts:
5075 sb->s_flags = old_sb_flags;
5076 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5077 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5078 sbi->s_resuid = old_opts.s_resuid;
5079 sbi->s_resgid = old_opts.s_resgid;
5080 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5081 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5082 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5083#ifdef CONFIG_QUOTA
5084 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5085 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5086 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5087 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5088 }
5089#endif
d4c402d9 5090 kfree(orig_data);
ac27a0ec
DK		 5091 return err;
5092}
5093
2b2d6d01 5094static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5095{
5096 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5097 struct ext4_sb_info *sbi = EXT4_SB(sb);
5098 struct ext4_super_block *es = sbi->s_es;
27dd4385 5099 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5100 u64 fsid;
d02a9391 5101 s64 bfree;
27dd4385 5102 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5103
952fc18e
TT		 5104 if (!test_opt(sb, MINIX_DF))
5105 overhead = sbi->s_overhead;
ac27a0ec 5106
617ba13b 5107 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5108 buf->f_bsize = sb->s_blocksize;
b72f78cb 5109 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5110 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5111 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5112 /* prevent underflow in case that few free space is available */
57042651 5113 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5114 buf->f_bavail = buf->f_bfree -
5115 (ext4_r_blocks_count(es) + resv_blocks);
5116 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5117 buf->f_bavail = 0;
5118 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5119 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5120 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5121 fsid = le64_to_cpup((void *)es->s_uuid) ^
5122 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5123 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5124 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5125
ac27a0ec
DK		 5126 return 0;
5127}
5128
0b8e58a1
AD		 5129/* Helper function for writing quotas on sync - we need to start transaction
5130 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 5131 * Process 1 Process 2
617ba13b 5132 * ext4_create() quota_sync()
a269eb18 5133 * jbd2_journal_start() write_dquot()
871a2931 5134 * dquot_initialize() down(dqio_mutex)
dab291af 5135 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 5136 *
5137 */
5138
5139#ifdef CONFIG_QUOTA
5140
5141static inline struct inode *dquot_to_inode(struct dquot *dquot)
5142{
4c376dca 5143 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5144}
5145
617ba13b 5146static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5147{
5148 int ret, err;
5149 handle_t *handle;
5150 struct inode *inode;
5151
5152 inode = dquot_to_inode(dquot);
9924a92a 5153 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5154 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5155 if (IS_ERR(handle))
5156 return PTR_ERR(handle);
5157 ret = dquot_commit(dquot);
617ba13b 5158 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5159 if (!ret)
5160 ret = err;
5161 return ret;
5162}
5163
617ba13b 5164static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5165{
5166 int ret, err;
5167 handle_t *handle;
5168
9924a92a 5169 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5170 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5171 if (IS_ERR(handle))
5172 return PTR_ERR(handle);
5173 ret = dquot_acquire(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_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5181{
5182 int ret, err;
5183 handle_t *handle;
5184
9924a92a 5185 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5186 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5187 if (IS_ERR(handle)) {
5188 /* Release dquot anyway to avoid endless cycle in dqput() */
5189 dquot_release(dquot);
ac27a0ec 5190 return PTR_ERR(handle);
9c3013e9 5191 }
ac27a0ec 5192 ret = dquot_release(dquot);
617ba13b 5193 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5194 if (!ret)
5195 ret = err;
5196 return ret;
5197}
5198
617ba13b 5199static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5200{
262b4662
JK		 5201 struct super_block *sb = dquot->dq_sb;
5202 struct ext4_sb_info *sbi = EXT4_SB(sb);
5203
2c8be6b2 5204 /* Are we journaling quotas? */
262b4662
JK		 5205 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5206 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5207 dquot_mark_dquot_dirty(dquot);
617ba13b 5208 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5209 } else {
5210 return dquot_mark_dquot_dirty(dquot);
5211 }
5212}
5213
617ba13b 5214static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5215{
5216 int ret, err;
5217 handle_t *handle;
5218
5219 /* Data block + inode block */
2b0143b5 5220 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5221 if (IS_ERR(handle))
5222 return PTR_ERR(handle);
5223 ret = dquot_commit_info(sb, type);
617ba13b 5224 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5225 if (!ret)
5226 ret = err;
5227 return ret;
5228}
5229
5230/*
5231 * Turn on quotas during mount time - we need to find
5232 * the quota file and such...
5233 */
617ba13b 5234static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5235{
287a8095
CH		 5236 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5237 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5238}
5239
5240/*
5241 * Standard function to be called on quota_on
5242 */
617ba13b 5243static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5244 struct path *path)
ac27a0ec
DK		 5245{
5246 int err;
ac27a0ec
DK		 5247
5248 if (!test_opt(sb, QUOTA))
5249 return -EINVAL;
0623543b 5250
ac27a0ec 5251 /* Quotafile not on the same filesystem? */
d8c9584e 5252 if (path->dentry->d_sb != sb)
ac27a0ec 5253 return -EXDEV;
0623543b
JK		 5254 /* Journaling quota? */
5255 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5256 /* Quotafile not in fs root? */
f00c9e44 5257 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5258 ext4_msg(sb, KERN_WARNING,
5259 "Quota file not on filesystem root. "
5260 "Journaled quota will not work");
2b2d6d01 5261 }
0623543b
JK		 5262
5263 /*
5264 * When we journal data on quota file, we have to flush journal to see
5265 * all updates to the file when we bypass pagecache...
5266 */
0390131b 5267 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5268 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5269 /*
5270 * We don't need to lock updates but journal_flush() could
5271 * otherwise be livelocked...
5272 */
5273 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5274 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5275 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5276 if (err)
7ffe1ea8 5277 return err;
0623543b
JK		 5278 }
5279
f00c9e44 5280 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 5281}
5282
7c319d32
AK		 5283static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5284 unsigned int flags)
5285{
5286 int err;
5287 struct inode *qf_inode;
a2d4a646 5288 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5289 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5290 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5291 };
5292
5293 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5294
5295 if (!qf_inums[type])
5296 return -EPERM;
5297
5298 qf_inode = ext4_iget(sb, qf_inums[type]);
5299 if (IS_ERR(qf_inode)) {
5300 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5301 return PTR_ERR(qf_inode);
5302 }
5303
bcb13850
JK		 5304 /* Don't account quota for quota files to avoid recursion */
5305 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5306 err = dquot_enable(qf_inode, type, format_id, flags);
5307 iput(qf_inode);
5308
5309 return err;
5310}
5311
5312/* Enable usage tracking for all quota types. */
5313static int ext4_enable_quotas(struct super_block *sb)
5314{
5315 int type, err = 0;
a2d4a646 5316 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5317 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5318 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5319 };
5320
5321 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5322 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5323 if (qf_inums[type]) {
5324 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5325 DQUOT_USAGE_ENABLED);
5326 if (err) {
5327 ext4_warning(sb,
72ba7450
TT		 5328 "Failed to enable quota tracking "
5329 "(type=%d, err=%d). Please run "
5330 "e2fsck to fix.", type, err);
7c319d32
AK		 5331 return err;
5332 }
5333 }
5334 }
5335 return 0;
5336}
5337
ca0e05e4
DM		 5338static int ext4_quota_off(struct super_block *sb, int type)
5339{
21f97697
JK		 5340 struct inode *inode = sb_dqopt(sb)->files[type];
5341 handle_t *handle;
5342
87009d86
DM		 5343 /* Force all delayed allocation blocks to be allocated.
5344 * Caller already holds s_umount sem */
5345 if (test_opt(sb, DELALLOC))
ca0e05e4 5346 sync_filesystem(sb);
ca0e05e4 5347
0b268590
AG		 5348 if (!inode)
5349 goto out;
5350
21f97697
JK		 5351 /* Update modification times of quota files when userspace can
5352 * start looking at them */
9924a92a 5353 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5354 if (IS_ERR(handle))
5355 goto out;
5356 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5357 ext4_mark_inode_dirty(handle, inode);
5358 ext4_journal_stop(handle);
5359
5360out:
ca0e05e4
DM		 5361 return dquot_quota_off(sb, type);
5362}
5363
ac27a0ec
DK		 5364/* Read data from quotafile - avoid pagecache and such because we cannot afford
5365 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5366 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5367 * we don't have to be afraid of races */
617ba13b 5368static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5369 size_t len, loff_t off)
5370{
5371 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5372 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5373 int offset = off & (sb->s_blocksize - 1);
5374 int tocopy;
5375 size_t toread;
5376 struct buffer_head *bh;
5377 loff_t i_size = i_size_read(inode);
5378
5379 if (off > i_size)
5380 return 0;
5381 if (off+len > i_size)
5382 len = i_size-off;
5383 toread = len;
5384 while (toread > 0) {
5385 tocopy = sb->s_blocksize - offset < toread ?
5386 sb->s_blocksize - offset : toread;
1c215028
TT		 5387 bh = ext4_bread(NULL, inode, blk, 0);
5388 if (IS_ERR(bh))
5389 return PTR_ERR(bh);
ac27a0ec
DK		 5390 if (!bh) /* A hole? */
5391 memset(data, 0, tocopy);
5392 else
5393 memcpy(data, bh->b_data+offset, tocopy);
5394 brelse(bh);
5395 offset = 0;
5396 toread -= tocopy;
5397 data += tocopy;
5398 blk++;
5399 }
5400 return len;
5401}
5402
5403/* Write to quotafile (we know the transaction is already started and has
5404 * enough credits) */
617ba13b 5405static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5406 const char *data, size_t len, loff_t off)
5407{
5408 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5409 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5410 int err, offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5411 struct buffer_head *bh;
5412 handle_t *handle = journal_current_handle();
5413
0390131b 5414 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5415 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5416 " cancelled because transaction is not started",
9c3013e9
JK		 5417 (unsigned long long)off, (unsigned long long)len);
5418 return -EIO;
5419 }
67eeb568
DM		 5420 /*
5421 * Since we account only one data block in transaction credits,
5422 * then it is impossible to cross a block boundary.
5423 */
5424 if (sb->s_blocksize - offset < len) {
5425 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5426 " cancelled because not block aligned",
5427 (unsigned long long)off, (unsigned long long)len);
5428 return -EIO;
5429 }
5430
1c215028
TT		 5431 bh = ext4_bread(handle, inode, blk, 1);
5432 if (IS_ERR(bh))
5433 return PTR_ERR(bh);
67eeb568
DM		 5434 if (!bh)
5435 goto out;
5d601255 5436 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5437 err = ext4_journal_get_write_access(handle, bh);
5438 if (err) {
5439 brelse(bh);
1c215028 5440 return err;
ac27a0ec 5441 }
67eeb568
DM		 5442 lock_buffer(bh);
5443 memcpy(bh->b_data+offset, data, len);
5444 flush_dcache_page(bh->b_page);
5445 unlock_buffer(bh);
62d2b5f2 5446 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5447 brelse(bh);
ac27a0ec 5448out:
67eeb568
DM		 5449 if (inode->i_size < off + len) {
5450 i_size_write(inode, off + len);
617ba13b 5451 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5452 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5453 }
67eeb568 5454 return len;
ac27a0ec
DK		 5455}
5456
5457#endif
5458
152a0836
AV		 5459static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5460 const char *dev_name, void *data)
ac27a0ec 5461{
152a0836 5462 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5463}
5464
37f328eb 5465#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5466static inline void register_as_ext2(void)
5467{
5468 int err = register_filesystem(&ext2_fs_type);
5469 if (err)
5470 printk(KERN_WARNING
5471 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5472}
5473
5474static inline void unregister_as_ext2(void)
5475{
5476 unregister_filesystem(&ext2_fs_type);
5477}
2035e776
TT		 5478
5479static inline int ext2_feature_set_ok(struct super_block *sb)
5480{
5481 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5482 return 0;
5483 if (sb->s_flags & MS_RDONLY)
5484 return 1;
5485 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5486 return 0;
5487 return 1;
5488}
24b58424
TT		 5489#else
5490static inline void register_as_ext2(void) { }
5491static inline void unregister_as_ext2(void) { }
2035e776 5492static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5493#endif
5494
37f328eb 5495#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5496static inline void register_as_ext3(void)
5497{
5498 int err = register_filesystem(&ext3_fs_type);
5499 if (err)
5500 printk(KERN_WARNING
5501 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5502}
5503
5504static inline void unregister_as_ext3(void)
5505{
5506 unregister_filesystem(&ext3_fs_type);
5507}
2035e776
TT		 5508
5509static inline int ext3_feature_set_ok(struct super_block *sb)
5510{
5511 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5512 return 0;
5513 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5514 return 0;
5515 if (sb->s_flags & MS_RDONLY)
5516 return 1;
5517 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5518 return 0;
5519 return 1;
5520}
24b58424
TT		 5521#else
5522static inline void register_as_ext3(void) { }
5523static inline void unregister_as_ext3(void) { }
2035e776 5524static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5525#endif
5526
03010a33
TT		 5527static struct file_system_type ext4_fs_type = {
5528 .owner = THIS_MODULE,
5529 .name = "ext4",
152a0836 5530 .mount = ext4_mount,
03010a33
TT		 5531 .kill_sb = kill_block_super,
5532 .fs_flags = FS_REQUIRES_DEV,
5533};
7f78e035 5534MODULE_ALIAS_FS("ext4");
03010a33 5535
8f021222 5536static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5537{
5538 struct ext4_features *ef;
5539 int ret = -ENOMEM;
5540
5541 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5542 if (!ef)
5543 goto out;
5544
5545 ef->f_kobj.kset = ext4_kset;
5546 init_completion(&ef->f_kobj_unregister);
5547 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5548 "features");
5549 if (ret) {
5550 kfree(ef);
5551 goto out;
5552 }
5553
5554 ext4_feat = ef;
5555 ret = 0;
5556out:
5557 return ret;
5558}
5559
8f021222
LC		 5560static void ext4_exit_feat_adverts(void)
5561{
5562 kobject_put(&ext4_feat->f_kobj);
5563 wait_for_completion(&ext4_feat->f_kobj_unregister);
5564 kfree(ext4_feat);
5565}
5566
e9e3bcec
ES		 5567/* Shared across all ext4 file systems */
5568wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5569struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5570
5dabfc78 5571static int __init ext4_init_fs(void)
ac27a0ec 5572{
e9e3bcec 5573 int i, err;
c9de560d 5574
07c0c5d8
AV		 5575 ext4_li_info = NULL;
5576 mutex_init(&ext4_li_mtx);
5577
9a4c8019 5578 /* Build-time check for flags consistency */
12e9b892 5579 ext4_check_flag_values();
e9e3bcec
ES		 5580
5581 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5582 mutex_init(&ext4__aio_mutex[i]);
5583 init_waitqueue_head(&ext4__ioend_wq[i]);
5584 }
5585
51865fda 5586 err = ext4_init_es();
6fd058f7
TT		 5587 if (err)
5588 return err;
51865fda
ZL		 5589
5590 err = ext4_init_pageio();
5591 if (err)
5592 goto out7;
5593
5dabfc78 5594 err = ext4_init_system_zone();
bd2d0210 5595 if (err)
d44651d0 5596 goto out6;
3197ebdb 5597 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5598 if (!ext4_kset) {
5599 err = -ENOMEM;
dd68314c 5600 goto out5;
0e376b1e 5601 }
d44651d0 5602 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5603
5604 err = ext4_init_feat_adverts();
dd68314c
TT		 5605 if (err)
5606 goto out4;
857ac889 5607
5dabfc78 5608 err = ext4_init_mballoc();
c9de560d
AT		 5609 if (err)
5610 goto out2;
9c191f70
M		 5611 else
5612 ext4_mballoc_ready = 1;
ac27a0ec
DK		 5613 err = init_inodecache();
5614 if (err)
5615 goto out1;
24b58424 5616 register_as_ext3();
2035e776 5617 register_as_ext2();
03010a33 5618 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5619 if (err)
5620 goto out;
bfff6873 5621
ac27a0ec
DK		 5622 return 0;
5623out:
24b58424
TT		 5624 unregister_as_ext2();
5625 unregister_as_ext3();
ac27a0ec
DK		 5626 destroy_inodecache();
5627out1:
9c191f70 5628 ext4_mballoc_ready = 0;
5dabfc78 5629 ext4_exit_mballoc();
9c191f70 5630out2:
8f021222 5631 ext4_exit_feat_adverts();
dd68314c 5632out4:
d44651d0
FJ		 5633 if (ext4_proc_root)
5634 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5635 kset_unregister(ext4_kset);
d44651d0 5636out5:
5dabfc78 5637 ext4_exit_system_zone();
d44651d0 5638out6:
5dabfc78 5639 ext4_exit_pageio();
51865fda
ZL		 5640out7:
5641 ext4_exit_es();
5642
ac27a0ec
DK		 5643 return err;
5644}
5645
5dabfc78 5646static void __exit ext4_exit_fs(void)
ac27a0ec 5647{
bfff6873 5648 ext4_destroy_lazyinit_thread();
24b58424
TT		 5649 unregister_as_ext2();
5650 unregister_as_ext3();
03010a33 5651 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5652 destroy_inodecache();
5dabfc78 5653 ext4_exit_mballoc();
8f021222 5654 ext4_exit_feat_adverts();
9f6200bb 5655 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5656 kset_unregister(ext4_kset);
5dabfc78
TT		 5657 ext4_exit_system_zone();
5658 ext4_exit_pageio();
dd12ed14 5659 ext4_exit_es();
ac27a0ec
DK		 5660}
5661
5662MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5663MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5664MODULE_LICENSE("GPL");
5dabfc78
TT		 5665module_init(ext4_init_fs)
5666module_exit(ext4_exit_fs)
Empty description
This page took 1.934313 seconds and 4 git commands to generate.