]> git.ipfire.org Git - thirdparty/kernel/linux.git/blame - fs/ext4/super.c
projects / thirdparty / kernel / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: avoid arithemetic overflow that can trigger a BUG
[thirdparty/kernel/linux.git] / fs / ext4 / super.c
CommitLineData
f5166768 1// SPDX-License-Identifier: GPL-2.0
ac27a0ec 2/*
617ba13b 3 * linux/fs/ext4/super.c
ac27a0ec
DK		 4 *
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
9 *
10 * from
11 *
12 * linux/fs/minix/inode.c
13 *
14 * Copyright (C) 1991, 1992 Linus Torvalds
15 *
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
18 */
19
20#include <linux/module.h>
21#include <linux/string.h>
22#include <linux/fs.h>
23#include <linux/time.h>
c5ca7c76 24#include <linux/vmalloc.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
66114cad 28#include <linux/backing-dev.h>
ac27a0ec 29#include <linux/parser.h>
ac27a0ec 30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
ef510424 41#include <linux/dax.h>
7abc52c2 42#include <linux/cleancache.h>
7c0f6ba6 43#include <linux/uaccess.h>
ee73f9a5 44#include <linux/iversion.h>
ac27a0ec 45
bfff6873
LC		 46#include <linux/kthread.h>
47#include <linux/freezer.h>
48
3dcf5451 49#include "ext4.h"
4a092d73 50#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 51#include "ext4_jbd2.h"
ac27a0ec
DK		 52#include "xattr.h"
53#include "acl.h"
3661d286 54#include "mballoc.h"
0c9ec4be 55#include "fsmap.h"
ac27a0ec 56
9bffad1e
TT		 57#define CREATE_TRACE_POINTS
58#include <trace/events/ext4.h>
59
0b75a840
LC		 60static struct ext4_lazy_init *ext4_li_info;
61static struct mutex ext4_li_mtx;
e294a537 62static struct ratelimit_state ext4_mount_msg_ratelimit;
9f6200bb 63
617ba13b 64static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 65 unsigned long journal_devnum);
2adf6da8 66static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 67static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 68static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
617ba13b 72static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 73static int ext4_remount(struct super_block *sb, int *flags, char *data);
74static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 75static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 76static int ext4_freeze(struct super_block *sb);
152a0836
AV		 77static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
2035e776
TT		 79static inline int ext2_feature_set_ok(struct super_block *sb);
80static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 81static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 82static void ext4_destroy_lazyinit_thread(void);
83static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 84static void ext4_clear_request_list(void);
c6cb7e77
EW		 85static struct inode *ext4_get_journal_inode(struct super_block *sb,
86 unsigned int journal_inum);
ac27a0ec 87
e74031fd
JK		 88/*
89 * Lock ordering
90 *
91 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
92 * i_mmap_rwsem (inode->i_mmap_rwsem)!
93 *
94 * page fault path:
95 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
96 * page lock -> i_data_sem (rw)
97 *
98 * buffered write path:
99 * sb_start_write -> i_mutex -> mmap_sem
100 * sb_start_write -> i_mutex -> transaction start -> page lock ->
101 * i_data_sem (rw)
102 *
103 * truncate:
1d39834f
NB		 104 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
106 * i_data_sem (rw)
e74031fd
JK		 107 *
108 * direct IO:
1d39834f
NB		 109 * sb_start_write -> i_mutex -> mmap_sem
110 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
e74031fd
JK		 111 *
112 * writepages:
113 * transaction start -> page lock(s) -> i_data_sem (rw)
114 */
115
c290ea01 116#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
2035e776
TT		 117static struct file_system_type ext2_fs_type = {
118 .owner = THIS_MODULE,
119 .name = "ext2",
120 .mount = ext4_mount,
121 .kill_sb = kill_block_super,
122 .fs_flags = FS_REQUIRES_DEV,
123};
7f78e035 124MODULE_ALIAS_FS("ext2");
fa7614dd 125MODULE_ALIAS("ext2");
2035e776
TT		 126#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
127#else
128#define IS_EXT2_SB(sb) (0)
129#endif
130
131
ba69f9ab
JK		 132static struct file_system_type ext3_fs_type = {
133 .owner = THIS_MODULE,
134 .name = "ext3",
152a0836 135 .mount = ext4_mount,
ba69f9ab
JK		 136 .kill_sb = kill_block_super,
137 .fs_flags = FS_REQUIRES_DEV,
138};
7f78e035 139MODULE_ALIAS_FS("ext3");
fa7614dd 140MODULE_ALIAS("ext3");
ba69f9ab 141#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
bd81d8ee 142
d25425f8
DW		 143static int ext4_verify_csum_type(struct super_block *sb,
144 struct ext4_super_block *es)
145{
e2b911c5 146 if (!ext4_has_feature_metadata_csum(sb))
d25425f8
DW		 147 return 1;
148
149 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
150}
151
a9c47317
DW		 152static __le32 ext4_superblock_csum(struct super_block *sb,
153 struct ext4_super_block *es)
154{
155 struct ext4_sb_info *sbi = EXT4_SB(sb);
156 int offset = offsetof(struct ext4_super_block, s_checksum);
157 __u32 csum;
158
159 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
160
161 return cpu_to_le32(csum);
162}
163
c197855e
SH		 164static int ext4_superblock_csum_verify(struct super_block *sb,
165 struct ext4_super_block *es)
a9c47317 166{
9aa5d32b 167 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 168 return 1;
169
170 return es->s_checksum == ext4_superblock_csum(sb, es);
171}
172
06db49e6 173void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 174{
06db49e6
TT		 175 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
176
9aa5d32b 177 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 178 return;
179
180 es->s_checksum = ext4_superblock_csum(sb, es);
181}
182
9933fc0a
TT		 183void *ext4_kvmalloc(size_t size, gfp_t flags)
184{
185 void *ret;
186
8be04b93 187 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 188 if (!ret)
189 ret = __vmalloc(size, flags, PAGE_KERNEL);
190 return ret;
191}
192
193void *ext4_kvzalloc(size_t size, gfp_t flags)
194{
195 void *ret;
196
8be04b93 197 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 198 if (!ret)
199 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
200 return ret;
201}
202
8fadc143
AR		 203ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
204 struct ext4_group_desc *bg)
bd81d8ee 205{
3a14589c 206 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 208 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 209}
210
8fadc143
AR		 211ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
212 struct ext4_group_desc *bg)
bd81d8ee 213{
5272f837 214 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 216 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 217}
218
8fadc143
AR		 219ext4_fsblk_t ext4_inode_table(struct super_block *sb,
220 struct ext4_group_desc *bg)
bd81d8ee 221{
5272f837 222 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 224 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 225}
226
021b65bb
TT		 227__u32 ext4_free_group_clusters(struct super_block *sb,
228 struct ext4_group_desc *bg)
560671a0
AK		 229{
230 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 232 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 233}
234
235__u32 ext4_free_inodes_count(struct super_block *sb,
236 struct ext4_group_desc *bg)
237{
238 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
239 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 240 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 241}
242
243__u32 ext4_used_dirs_count(struct super_block *sb,
244 struct ext4_group_desc *bg)
245{
246 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
247 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 248 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 249}
250
251__u32 ext4_itable_unused_count(struct super_block *sb,
252 struct ext4_group_desc *bg)
253{
254 return le16_to_cpu(bg->bg_itable_unused_lo) |
255 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 256 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 257}
258
8fadc143
AR		 259void ext4_block_bitmap_set(struct super_block *sb,
260 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 261{
3a14589c 262 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 265}
266
8fadc143
AR		 267void ext4_inode_bitmap_set(struct super_block *sb,
268 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 269{
5272f837 270 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 273}
274
8fadc143
AR		 275void ext4_inode_table_set(struct super_block *sb,
276 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 277{
5272f837 278 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 281}
282
021b65bb
TT		 283void ext4_free_group_clusters_set(struct super_block *sb,
284 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 285{
286 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
289}
290
291void ext4_free_inodes_set(struct super_block *sb,
292 struct ext4_group_desc *bg, __u32 count)
293{
294 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
295 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
296 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
297}
298
299void ext4_used_dirs_set(struct super_block *sb,
300 struct ext4_group_desc *bg, __u32 count)
301{
302 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
303 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
304 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
305}
306
307void ext4_itable_unused_set(struct super_block *sb,
308 struct ext4_group_desc *bg, __u32 count)
309{
310 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
311 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
312 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
313}
314
6a0678a7
AB		 315static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
316{
317 time64_t now = ktime_get_real_seconds();
318
319 now = clamp_val(now, 0, (1ull << 40) - 1);
320
321 *lo = cpu_to_le32(lower_32_bits(now));
322 *hi = upper_32_bits(now);
323}
324
325static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
326{
327 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
328}
329#define ext4_update_tstamp(es, tstamp) \
330 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
331#define ext4_get_tstamp(es, tstamp) \
332 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
d3d1faf6 333
1c13d5c0
TT		 334static void __save_error_info(struct super_block *sb, const char *func,
335 unsigned int line)
336{
337 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
338
339 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1b46617b
TT		 340 if (bdev_read_only(sb->s_bdev))
341 return;
1c13d5c0 342 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6a0678a7 343 ext4_update_tstamp(es, s_last_error_time);
1c13d5c0
TT		 344 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
345 es->s_last_error_line = cpu_to_le32(line);
346 if (!es->s_first_error_time) {
347 es->s_first_error_time = es->s_last_error_time;
6a0678a7 348 es->s_first_error_time_hi = es->s_last_error_time_hi;
1c13d5c0
TT		 349 strncpy(es->s_first_error_func, func,
350 sizeof(es->s_first_error_func));
351 es->s_first_error_line = cpu_to_le32(line);
352 es->s_first_error_ino = es->s_last_error_ino;
353 es->s_first_error_block = es->s_last_error_block;
354 }
66e61a9e
TT		 355 /*
356 * Start the daily error reporting function if it hasn't been
357 * started already
358 */
359 if (!es->s_error_count)
360 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 361 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 362}
363
364static void save_error_info(struct super_block *sb, const char *func,
365 unsigned int line)
366{
367 __save_error_info(sb, func, line);
368 ext4_commit_super(sb, 1);
369}
370
bdfe0cbd
TT		 371/*
372 * The del_gendisk() function uninitializes the disk-specific data
373 * structures, including the bdi structure, without telling anyone
374 * else. Once this happens, any attempt to call mark_buffer_dirty()
375 * (for example, by ext4_commit_super), will cause a kernel OOPS.
376 * This is a kludge to prevent these oops until we can put in a proper
377 * hook in del_gendisk() to inform the VFS and file system layers.
378 */
379static int block_device_ejected(struct super_block *sb)
380{
381 struct inode *bd_inode = sb->s_bdev->bd_inode;
382 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
383
384 return bdi->dev == NULL;
385}
386
18aadd47
BJ		 387static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
388{
389 struct super_block *sb = journal->j_private;
390 struct ext4_sb_info *sbi = EXT4_SB(sb);
391 int error = is_journal_aborted(journal);
5d3ee208 392 struct ext4_journal_cb_entry *jce;
18aadd47 393
5d3ee208 394 BUG_ON(txn->t_state == T_FINISHED);
a0154344
DJ		 395
396 ext4_process_freed_data(sb, txn->t_tid);
397
18aadd47 398 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 399 while (!list_empty(&txn->t_private_list)) {
400 jce = list_entry(txn->t_private_list.next,
401 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 402 list_del_init(&jce->jce_list);
403 spin_unlock(&sbi->s_md_lock);
404 jce->jce_func(sb, jce, error);
405 spin_lock(&sbi->s_md_lock);
406 }
407 spin_unlock(&sbi->s_md_lock);
408}
1c13d5c0 409
ac27a0ec
DK		 410/* Deal with the reporting of failure conditions on a filesystem such as
411 * inconsistencies detected or read IO failures.
412 *
413 * On ext2, we can store the error state of the filesystem in the
617ba13b 414 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 415 * write ordering constraints on the superblock which prevent us from
416 * writing it out straight away; and given that the journal is about to
417 * be aborted, we can't rely on the current, or future, transactions to
418 * write out the superblock safely.
419 *
dab291af 420 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 421 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 422 * that error until we've noted it down and cleared it.
423 */
424
617ba13b 425static void ext4_handle_error(struct super_block *sb)
ac27a0ec 426{
327eaf73
TT		 427 if (test_opt(sb, WARN_ON_ERROR))
428 WARN_ON_ONCE(1);
429
bc98a42c 430 if (sb_rdonly(sb))
ac27a0ec
DK		 431 return;
432
2b2d6d01 433 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 434 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 435
4ab2f15b 436 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 437 if (journal)
dab291af 438 jbd2_journal_abort(journal, -EIO);
ac27a0ec 439 }
2b2d6d01 440 if (test_opt(sb, ERRORS_RO)) {
b31e1552 441 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 442 /*
443 * Make sure updated value of ->s_mount_flags will be visible
444 * before ->s_flags update
445 */
446 smp_wmb();
1751e8a6 447 sb->s_flags |= SB_RDONLY;
ac27a0ec 448 }
4327ba52
DJ		 449 if (test_opt(sb, ERRORS_PANIC)) {
450 if (EXT4_SB(sb)->s_journal &&
451 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
452 return;
617ba13b 453 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec 454 sb->s_id);
4327ba52 455 }
ac27a0ec
DK		 456}
457
efbed4dc
TT		 458#define ext4_error_ratelimit(sb) \
459 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
460 "EXT4-fs error")
461
12062ddd 462void __ext4_error(struct super_block *sb, const char *function,
c398eda0 463 unsigned int line, const char *fmt, ...)
ac27a0ec 464{
0ff2ea7d 465 struct va_format vaf;
ac27a0ec
DK		 466 va_list args;
467
0db1ff22
TT		 468 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
469 return;
470
ccf0f32a 471 trace_ext4_error(sb, function, line);
efbed4dc
TT		 472 if (ext4_error_ratelimit(sb)) {
473 va_start(args, fmt);
474 vaf.fmt = fmt;
475 vaf.va = &args;
476 printk(KERN_CRIT
477 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
478 sb->s_id, function, line, current->comm, &vaf);
479 va_end(args);
480 }
f3fc0210 481 save_error_info(sb, function, line);
617ba13b 482 ext4_handle_error(sb);
ac27a0ec
DK		 483}
484
e7c96e8e
JP		 485void __ext4_error_inode(struct inode *inode, const char *function,
486 unsigned int line, ext4_fsblk_t block,
487 const char *fmt, ...)
273df556
FM		 488{
489 va_list args;
f7c21177 490 struct va_format vaf;
1c13d5c0 491 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 492
0db1ff22
TT		 493 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
494 return;
495
ccf0f32a 496 trace_ext4_error(inode->i_sb, function, line);
1c13d5c0
TT		 497 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
498 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 499 if (ext4_error_ratelimit(inode->i_sb)) {
500 va_start(args, fmt);
501 vaf.fmt = fmt;
502 vaf.va = &args;
503 if (block)
504 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
505 "inode #%lu: block %llu: comm %s: %pV\n",
506 inode->i_sb->s_id, function, line, inode->i_ino,
507 block, current->comm, &vaf);
508 else
509 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
510 "inode #%lu: comm %s: %pV\n",
511 inode->i_sb->s_id, function, line, inode->i_ino,
512 current->comm, &vaf);
513 va_end(args);
514 }
1c13d5c0 515 save_error_info(inode->i_sb, function, line);
273df556
FM		 516 ext4_handle_error(inode->i_sb);
517}
518
e7c96e8e
JP		 519void __ext4_error_file(struct file *file, const char *function,
520 unsigned int line, ext4_fsblk_t block,
521 const char *fmt, ...)
273df556
FM		 522{
523 va_list args;
f7c21177 524 struct va_format vaf;
1c13d5c0 525 struct ext4_super_block *es;
496ad9aa 526 struct inode *inode = file_inode(file);
273df556
FM		 527 char pathname[80], *path;
528
0db1ff22
TT		 529 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
530 return;
531
ccf0f32a 532 trace_ext4_error(inode->i_sb, function, line);
1c13d5c0
TT		 533 es = EXT4_SB(inode->i_sb)->s_es;
534 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 535 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 536 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 537 if (IS_ERR(path))
538 path = "(unknown)";
539 va_start(args, fmt);
540 vaf.fmt = fmt;
541 vaf.va = &args;
542 if (block)
543 printk(KERN_CRIT
544 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
545 "block %llu: comm %s: path %s: %pV\n",
546 inode->i_sb->s_id, function, line, inode->i_ino,
547 block, current->comm, path, &vaf);
548 else
549 printk(KERN_CRIT
550 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
551 "comm %s: path %s: %pV\n",
552 inode->i_sb->s_id, function, line, inode->i_ino,
553 current->comm, path, &vaf);
554 va_end(args);
555 }
1c13d5c0 556 save_error_info(inode->i_sb, function, line);
273df556
FM		 557 ext4_handle_error(inode->i_sb);
558}
559
722887dd
TT		 560const char *ext4_decode_error(struct super_block *sb, int errno,
561 char nbuf[16])
ac27a0ec
DK		 562{
563 char *errstr = NULL;
564
565 switch (errno) {
6a797d27
DW		 566 case -EFSCORRUPTED:
567 errstr = "Corrupt filesystem";
568 break;
569 case -EFSBADCRC:
570 errstr = "Filesystem failed CRC";
571 break;
ac27a0ec
DK		 572 case -EIO:
573 errstr = "IO failure";
574 break;
575 case -ENOMEM:
576 errstr = "Out of memory";
577 break;
578 case -EROFS:
78f1ddbb
TT		 579 if (!sb || (EXT4_SB(sb)->s_journal &&
580 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 581 errstr = "Journal has aborted";
582 else
583 errstr = "Readonly filesystem";
584 break;
585 default:
586 /* If the caller passed in an extra buffer for unknown
587 * errors, textualise them now. Else we just return
588 * NULL. */
589 if (nbuf) {
590 /* Check for truncated error codes... */
591 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
592 errstr = nbuf;
593 }
594 break;
595 }
596
597 return errstr;
598}
599
617ba13b 600/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 601 * automatically and invokes the appropriate error response. */
602
c398eda0
TT		 603void __ext4_std_error(struct super_block *sb, const char *function,
604 unsigned int line, int errno)
ac27a0ec
DK		 605{
606 char nbuf[16];
607 const char *errstr;
608
0db1ff22
TT		 609 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
610 return;
611
ac27a0ec
DK		 612 /* Special case: if the error is EROFS, and we're not already
613 * inside a transaction, then there's really no point in logging
614 * an error. */
bc98a42c 615 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
ac27a0ec
DK		 616 return;
617
efbed4dc
TT		 618 if (ext4_error_ratelimit(sb)) {
619 errstr = ext4_decode_error(sb, errno, nbuf);
620 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
621 sb->s_id, function, line, errstr);
622 }
ac27a0ec 623
efbed4dc 624 save_error_info(sb, function, line);
617ba13b 625 ext4_handle_error(sb);
ac27a0ec
DK		 626}
627
628/*
617ba13b 629 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 630 * abort function may be used to deal with unrecoverable failures such
631 * as journal IO errors or ENOMEM at a critical moment in log management.
632 *
633 * We unconditionally force the filesystem into an ABORT|READONLY state,
634 * unless the error response on the fs has been set to panic in which
635 * case we take the easy way out and panic immediately.
636 */
637
c67d859e 638void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 639 unsigned int line, const char *fmt, ...)
ac27a0ec 640{
651e1c3b 641 struct va_format vaf;
ac27a0ec
DK		 642 va_list args;
643
0db1ff22
TT		 644 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
645 return;
646
1c13d5c0 647 save_error_info(sb, function, line);
ac27a0ec 648 va_start(args, fmt);
651e1c3b
JP		 649 vaf.fmt = fmt;
650 vaf.va = &args;
651 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
652 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 653 va_end(args);
654
bc98a42c 655 if (sb_rdonly(sb) == 0) {
1c13d5c0 656 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 657 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 658 /*
659 * Make sure updated value of ->s_mount_flags will be visible
660 * before ->s_flags update
661 */
662 smp_wmb();
1751e8a6 663 sb->s_flags |= SB_RDONLY;
1c13d5c0
TT		 664 if (EXT4_SB(sb)->s_journal)
665 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
666 save_error_info(sb, function, line);
667 }
4327ba52
DJ		 668 if (test_opt(sb, ERRORS_PANIC)) {
669 if (EXT4_SB(sb)->s_journal &&
670 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
671 return;
617ba13b 672 panic("EXT4-fs panic from previous error\n");
4327ba52 673 }
ac27a0ec
DK		 674}
675
e7c96e8e
JP		 676void __ext4_msg(struct super_block *sb,
677 const char *prefix, const char *fmt, ...)
b31e1552 678{
0ff2ea7d 679 struct va_format vaf;
b31e1552
ES		 680 va_list args;
681
efbed4dc
TT		 682 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
683 return;
684
b31e1552 685 va_start(args, fmt);
0ff2ea7d
JP		 686 vaf.fmt = fmt;
687 vaf.va = &args;
688 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 689 va_end(args);
690}
691
b03a2f7e
AD		 692#define ext4_warning_ratelimit(sb) \
693 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
694 "EXT4-fs warning")
695
12062ddd 696void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 697 unsigned int line, const char *fmt, ...)
ac27a0ec 698{
0ff2ea7d 699 struct va_format vaf;
ac27a0ec
DK		 700 va_list args;
701
b03a2f7e 702 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 703 return;
704
ac27a0ec 705 va_start(args, fmt);
0ff2ea7d
JP		 706 vaf.fmt = fmt;
707 vaf.va = &args;
708 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
709 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 710 va_end(args);
711}
712
b03a2f7e
AD		 713void __ext4_warning_inode(const struct inode *inode, const char *function,
714 unsigned int line, const char *fmt, ...)
715{
716 struct va_format vaf;
717 va_list args;
718
719 if (!ext4_warning_ratelimit(inode->i_sb))
720 return;
721
722 va_start(args, fmt);
723 vaf.fmt = fmt;
724 vaf.va = &args;
725 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
726 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
727 function, line, inode->i_ino, current->comm, &vaf);
728 va_end(args);
729}
730
e29136f8
TT		 731void __ext4_grp_locked_error(const char *function, unsigned int line,
732 struct super_block *sb, ext4_group_t grp,
733 unsigned long ino, ext4_fsblk_t block,
734 const char *fmt, ...)
5d1b1b3f
AK		 735__releases(bitlock)
736__acquires(bitlock)
737{
0ff2ea7d 738 struct va_format vaf;
5d1b1b3f
AK		 739 va_list args;
740 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
741
0db1ff22
TT		 742 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
743 return;
744
ccf0f32a 745 trace_ext4_error(sb, function, line);
1c13d5c0
TT		 746 es->s_last_error_ino = cpu_to_le32(ino);
747 es->s_last_error_block = cpu_to_le64(block);
748 __save_error_info(sb, function, line);
0ff2ea7d 749
efbed4dc
TT		 750 if (ext4_error_ratelimit(sb)) {
751 va_start(args, fmt);
752 vaf.fmt = fmt;
753 vaf.va = &args;
754 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
755 sb->s_id, function, line, grp);
756 if (ino)
757 printk(KERN_CONT "inode %lu: ", ino);
758 if (block)
759 printk(KERN_CONT "block %llu:",
760 (unsigned long long) block);
761 printk(KERN_CONT "%pV\n", &vaf);
762 va_end(args);
763 }
5d1b1b3f 764
327eaf73
TT		 765 if (test_opt(sb, WARN_ON_ERROR))
766 WARN_ON_ONCE(1);
767
5d1b1b3f 768 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 769 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 770 return;
771 }
1c13d5c0 772
5d1b1b3f 773 ext4_unlock_group(sb, grp);
06f29cc8 774 ext4_commit_super(sb, 1);
5d1b1b3f
AK		 775 ext4_handle_error(sb);
776 /*
777 * We only get here in the ERRORS_RO case; relocking the group
778 * may be dangerous, but nothing bad will happen since the
779 * filesystem will have already been marked read/only and the
780 * journal has been aborted. We return 1 as a hint to callers
781 * who might what to use the return value from
25985edc 782 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 783 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
784 * aggressively from the ext4 function in question, with a
785 * more appropriate error code.
786 */
787 ext4_lock_group(sb, grp);
788 return;
789}
790
db79e6d1
WS		 791void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
792 ext4_group_t group,
793 unsigned int flags)
794{
795 struct ext4_sb_info *sbi = EXT4_SB(sb);
796 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
797 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
9af0b3d1
WS		 798 int ret;
799
800 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
801 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
802 &grp->bb_state);
803 if (!ret)
804 percpu_counter_sub(&sbi->s_freeclusters_counter,
805 grp->bb_free);
db79e6d1
WS		 806 }
807
9af0b3d1
WS		 808 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
809 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
810 &grp->bb_state);
811 if (!ret && gdp) {
db79e6d1
WS		 812 int count;
813
814 count = ext4_free_inodes_count(sb, gdp);
815 percpu_counter_sub(&sbi->s_freeinodes_counter,
816 count);
817 }
db79e6d1
WS		 818 }
819}
820
617ba13b 821void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 822{
617ba13b 823 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 824
617ba13b 825 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 826 return;
827
12062ddd 828 ext4_warning(sb,
ac27a0ec
DK		 829 "updating to rev %d because of new feature flag, "
830 "running e2fsck is recommended",
617ba13b 831 EXT4_DYNAMIC_REV);
ac27a0ec 832
617ba13b
MC		 833 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
834 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
835 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 836 /* leave es->s_feature_*compat flags alone */
837 /* es->s_uuid will be set by e2fsck if empty */
838
839 /*
840 * The rest of the superblock fields should be zero, and if not it
841 * means they are likely already in use, so leave them alone. We
842 * can leave it up to e2fsck to clean up any inconsistencies there.
843 */
844}
845
846/*
847 * Open the external journal device
848 */
b31e1552 849static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 850{
851 struct block_device *bdev;
852 char b[BDEVNAME_SIZE];
853
d4d77629 854 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 855 if (IS_ERR(bdev))
856 goto fail;
857 return bdev;
858
859fail:
b31e1552 860 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 861 __bdevname(dev, b), PTR_ERR(bdev));
862 return NULL;
863}
864
865/*
866 * Release the journal device
867 */
4385bab1 868static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 869{
4385bab1 870 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 871}
872
4385bab1 873static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 874{
875 struct block_device *bdev;
ac27a0ec
DK		 876 bdev = sbi->journal_bdev;
877 if (bdev) {
4385bab1 878 ext4_blkdev_put(bdev);
ac27a0ec
DK		 879 sbi->journal_bdev = NULL;
880 }
ac27a0ec
DK		 881}
882
883static inline struct inode *orphan_list_entry(struct list_head *l)
884{
617ba13b 885 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 886}
887
617ba13b 888static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 889{
890 struct list_head *l;
891
b31e1552
ES		 892 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
893 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 894
895 printk(KERN_ERR "sb_info orphan list:\n");
896 list_for_each(l, &sbi->s_orphan) {
897 struct inode *inode = orphan_list_entry(l);
898 printk(KERN_ERR " "
899 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
900 inode->i_sb->s_id, inode->i_ino, inode,
901 inode->i_mode, inode->i_nlink,
902 NEXT_ORPHAN(inode));
903 }
904}
905
957153fc
JK		 906#ifdef CONFIG_QUOTA
907static int ext4_quota_off(struct super_block *sb, int type);
908
909static inline void ext4_quota_off_umount(struct super_block *sb)
910{
911 int type;
912
964edf66
JK		 913 /* Use our quota_off function to clear inode flags etc. */
914 for (type = 0; type < EXT4_MAXQUOTAS; type++)
915 ext4_quota_off(sb, type);
957153fc
JK		 916}
917#else
918static inline void ext4_quota_off_umount(struct super_block *sb)
919{
920}
921#endif
922
2b2d6d01 923static void ext4_put_super(struct super_block *sb)
ac27a0ec 924{
617ba13b
MC		 925 struct ext4_sb_info *sbi = EXT4_SB(sb);
926 struct ext4_super_block *es = sbi->s_es;
97abd7d4 927 int aborted = 0;
ef2cabf7 928 int i, err;
ac27a0ec 929
857ac889 930 ext4_unregister_li_request(sb);
957153fc 931 ext4_quota_off_umount(sb);
e0ccfd95 932
2e8fa54e 933 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 934
0390131b 935 if (sbi->s_journal) {
97abd7d4 936 aborted = is_journal_aborted(sbi->s_journal);
0390131b
FM		 937 err = jbd2_journal_destroy(sbi->s_journal);
938 sbi->s_journal = NULL;
97abd7d4 939 if ((err < 0) && !aborted)
c67d859e 940 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 941 }
d4edac31 942
ebd173be 943 ext4_unregister_sysfs(sb);
d3922a77 944 ext4_es_unregister_shrinker(sbi);
9105bb14 945 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 946 ext4_release_system_zone(sb);
947 ext4_mb_release(sb);
948 ext4_ext_release(sb);
d4edac31 949
bc98a42c 950 if (!sb_rdonly(sb) && !aborted) {
e2b911c5 951 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 952 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 953 }
bc98a42c 954 if (!sb_rdonly(sb))
a8e25a83
AB		 955 ext4_commit_super(sb, 1);
956
ac27a0ec
DK		 957 for (i = 0; i < sbi->s_gdb_count; i++)
958 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 959 kvfree(sbi->s_group_desc);
960 kvfree(sbi->s_flex_groups);
57042651 961 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 962 percpu_counter_destroy(&sbi->s_freeinodes_counter);
963 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 964 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c8585c6f 965 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
ac27a0ec 966#ifdef CONFIG_QUOTA
a2d4a646 967 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 968 kfree(sbi->s_qf_names[i]);
969#endif
970
971 /* Debugging code just in case the in-memory inode orphan list
972 * isn't empty. The on-disk one can be non-empty if we've
973 * detected an error and taken the fs readonly, but the
974 * in-memory list had better be clean by this point. */
975 if (!list_empty(&sbi->s_orphan))
976 dump_orphan_list(sb, sbi);
977 J_ASSERT(list_empty(&sbi->s_orphan));
978
89d96a6f 979 sync_blockdev(sb->s_bdev);
f98393a6 980 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 981 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
982 /*
983 * Invalidate the journal device's buffers. We don't want them
984 * floating about in memory - the physical journal device may
985 * hotswapped, and it breaks the `ro-after' testing code.
986 */
987 sync_blockdev(sbi->journal_bdev);
f98393a6 988 invalidate_bdev(sbi->journal_bdev);
617ba13b 989 ext4_blkdev_remove(sbi);
ac27a0ec 990 }
dec214d0
TE		 991 if (sbi->s_ea_inode_cache) {
992 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
993 sbi->s_ea_inode_cache = NULL;
994 }
47387409
TE		 995 if (sbi->s_ea_block_cache) {
996 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
997 sbi->s_ea_block_cache = NULL;
9c191f70 998 }
c5e06d10
JL		 999 if (sbi->s_mmp_tsk)
1000 kthread_stop(sbi->s_mmp_tsk);
9060dd2c 1001 brelse(sbi->s_sbh);
ac27a0ec 1002 sb->s_fs_info = NULL;
3197ebdb
TT		 1003 /*
1004 * Now that we are completely done shutting down the
1005 * superblock, we need to actually destroy the kobject.
1006 */
3197ebdb
TT		 1007 kobject_put(&sbi->s_kobj);
1008 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 1009 if (sbi->s_chksum_driver)
1010 crypto_free_shash(sbi->s_chksum_driver);
705895b6 1011 kfree(sbi->s_blockgroup_lock);
5e405595 1012 fs_put_dax(sbi->s_daxdev);
ac27a0ec 1013 kfree(sbi);
ac27a0ec
DK		 1014}
1015
e18b890b 1016static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 1017
1018/*
1019 * Called inside transaction, so use GFP_NOFS
1020 */
617ba13b 1021static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 1022{
617ba13b 1023 struct ext4_inode_info *ei;
ac27a0ec 1024
e6b4f8da 1025 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 1026 if (!ei)
1027 return NULL;
0b8e58a1 1028
ee73f9a5 1029 inode_set_iversion(&ei->vfs_inode, 1);
202ee5df 1030 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 1031 INIT_LIST_HEAD(&ei->i_prealloc_list);
1032 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 1033 ext4_es_init_tree(&ei->i_es_tree);
1034 rwlock_init(&ei->i_es_lock);
edaa53ca 1035 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 1036 ei->i_es_all_nr = 0;
edaa53ca 1037 ei->i_es_shk_nr = 0;
dd475925 1038 ei->i_es_shrink_lblk = 0;
d2a17637 1039 ei->i_reserved_data_blocks = 0;
9d0be502 1040 ei->i_da_metadata_calc_len = 0;
7e731bc9 1041 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 1042 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 1043#ifdef CONFIG_QUOTA
1044 ei->i_reserved_quota = 0;
96c7e0d9 1045 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 1046#endif
8aefcd55 1047 ei->jinode = NULL;
2e8fa54e 1048 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 1049 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 1050 ei->i_sync_tid = 0;
1051 ei->i_datasync_tid = 0;
e27f41e1 1052 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 1053 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
ac27a0ec
DK		 1054 return &ei->vfs_inode;
1055}
1056
7ff9c073
TT		 1057static int ext4_drop_inode(struct inode *inode)
1058{
1059 int drop = generic_drop_inode(inode);
1060
1061 trace_ext4_drop_inode(inode, drop);
1062 return drop;
1063}
1064
fa0d7e3d
NP		 1065static void ext4_i_callback(struct rcu_head *head)
1066{
1067 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 1068 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1069}
1070
617ba13b 1071static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1072{
9f7dd93d 1073 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1074 ext4_msg(inode->i_sb, KERN_ERR,
1075 "Inode %lu (%p): orphan list check failed!",
1076 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1077 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1078 EXT4_I(inode), sizeof(struct ext4_inode_info),
1079 true);
1080 dump_stack();
1081 }
fa0d7e3d 1082 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 1083}
1084
51cc5068 1085static void init_once(void *foo)
ac27a0ec 1086{
617ba13b 1087 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1088
a35afb83 1089 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1090 init_rwsem(&ei->xattr_sem);
0e855ac8 1091 init_rwsem(&ei->i_data_sem);
ea3d7209 1092 init_rwsem(&ei->i_mmap_sem);
a35afb83 1093 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1094}
1095
e67bc2b3 1096static int __init init_inodecache(void)
ac27a0ec 1097{
f8dd7c70
DW		 1098 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1099 sizeof(struct ext4_inode_info), 0,
1100 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1101 SLAB_ACCOUNT),
1102 offsetof(struct ext4_inode_info, i_data),
1103 sizeof_field(struct ext4_inode_info, i_data),
1104 init_once);
617ba13b 1105 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1106 return -ENOMEM;
1107 return 0;
1108}
1109
1110static void destroy_inodecache(void)
1111{
8c0a8537
KS		 1112 /*
1113 * Make sure all delayed rcu free inodes are flushed before we
1114 * destroy cache.
1115 */
1116 rcu_barrier();
617ba13b 1117 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1118}
1119
0930fcc1 1120void ext4_clear_inode(struct inode *inode)
ac27a0ec 1121{
0930fcc1 1122 invalidate_inode_buffers(inode);
dbd5768f 1123 clear_inode(inode);
9f754758 1124 dquot_drop(inode);
c2ea3fde 1125 ext4_discard_preallocations(inode);
51865fda 1126 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1127 if (EXT4_I(inode)->jinode) {
1128 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1129 EXT4_I(inode)->jinode);
1130 jbd2_free_inode(EXT4_I(inode)->jinode);
1131 EXT4_I(inode)->jinode = NULL;
1132 }
3d204e24 1133 fscrypt_put_encryption_info(inode);
ac27a0ec
DK		 1134}
1135
1b961ac0 1136static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1137 u64 ino, u32 generation)
ac27a0ec 1138{
ac27a0ec 1139 struct inode *inode;
ac27a0ec 1140
617ba13b 1141 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1142 return ERR_PTR(-ESTALE);
617ba13b 1143 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1144 return ERR_PTR(-ESTALE);
1145
1146 /* iget isn't really right if the inode is currently unallocated!!
1147 *
617ba13b 1148 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1149 * deleted, so we should be safe.
1150 *
1151 * Currently we don't know the generation for parent directory, so
1152 * a generation of 0 means "accept any"
1153 */
f4bb2981 1154 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1155 if (IS_ERR(inode))
1156 return ERR_CAST(inode);
1157 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1158 iput(inode);
1159 return ERR_PTR(-ESTALE);
1160 }
1b961ac0
CH		 1161
1162 return inode;
1163}
1164
1165static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1166 int fh_len, int fh_type)
1b961ac0
CH		 1167{
1168 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1169 ext4_nfs_get_inode);
1170}
1171
1172static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1173 int fh_len, int fh_type)
1b961ac0
CH		 1174{
1175 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1176 ext4_nfs_get_inode);
ac27a0ec
DK		 1177}
1178
c39a7f84
TO		 1179/*
1180 * Try to release metadata pages (indirect blocks, directories) which are
1181 * mapped via the block device. Since these pages could have journal heads
1182 * which would prevent try_to_free_buffers() from freeing them, we must use
1183 * jbd2 layer's try_to_free_buffers() function to release them.
1184 */
0b8e58a1
AD		 1185static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1186 gfp_t wait)
c39a7f84
TO		 1187{
1188 journal_t *journal = EXT4_SB(sb)->s_journal;
1189
1190 WARN_ON(PageChecked(page));
1191 if (!page_has_buffers(page))
1192 return 0;
1193 if (journal)
1194 return jbd2_journal_try_to_free_buffers(journal, page,
d0164adc 1195 wait & ~__GFP_DIRECT_RECLAIM);
c39a7f84
TO		 1196 return try_to_free_buffers(page);
1197}
1198
a7550b30
JK		 1199#ifdef CONFIG_EXT4_FS_ENCRYPTION
1200static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1201{
1202 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1203 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1204}
1205
a7550b30
JK		 1206static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1207 void *fs_data)
1208{
2f8f5e76 1209 handle_t *handle = fs_data;
c1a5d5f6 1210 int res, res2, credits, retries = 0;
2f8f5e76 1211
9ce0151a
EB		 1212 /*
1213 * Encrypting the root directory is not allowed because e2fsck expects
1214 * lost+found to exist and be unencrypted, and encrypting the root
1215 * directory would imply encrypting the lost+found directory as well as
1216 * the filename "lost+found" itself.
1217 */
1218 if (inode->i_ino == EXT4_ROOT_INO)
1219 return -EPERM;
2f8f5e76 1220
7d3e06a8
RZ		 1221 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1222 return -EINVAL;
1223
94840e3c
EB		 1224 res = ext4_convert_inline_data(inode);
1225 if (res)
1226 return res;
1227
2f8f5e76
EB		 1228 /*
1229 * If a journal handle was specified, then the encryption context is
1230 * being set on a new inode via inheritance and is part of a larger
1231 * transaction to create the inode. Otherwise the encryption context is
1232 * being set on an existing inode in its own transaction. Only in the
1233 * latter case should the "retry on ENOSPC" logic be used.
1234 */
a7550b30 1235
2f8f5e76
EB		 1236 if (handle) {
1237 res = ext4_xattr_set_handle(handle, inode,
1238 EXT4_XATTR_INDEX_ENCRYPTION,
1239 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1240 ctx, len, 0);
a7550b30
JK		 1241 if (!res) {
1242 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1243 ext4_clear_inode_state(inode,
1244 EXT4_STATE_MAY_INLINE_DATA);
a3caa24b 1245 /*
2ee6a576
EB		 1246 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1247 * S_DAX may be disabled
a3caa24b
JK		 1248 */
1249 ext4_set_inode_flags(inode);
a7550b30
JK		 1250 }
1251 return res;
1252 }
1253
b8cb5a54
TE		 1254 res = dquot_initialize(inode);
1255 if (res)
1256 return res;
2f8f5e76 1257retry:
af65207c
TE		 1258 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1259 &credits);
dec214d0
TE		 1260 if (res)
1261 return res;
1262
c1a5d5f6 1263 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
a7550b30
JK		 1264 if (IS_ERR(handle))
1265 return PTR_ERR(handle);
1266
2f8f5e76
EB		 1267 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1268 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1269 ctx, len, 0);
a7550b30
JK		 1270 if (!res) {
1271 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
2ee6a576
EB		 1272 /*
1273 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1274 * S_DAX may be disabled
1275 */
a3caa24b 1276 ext4_set_inode_flags(inode);
a7550b30
JK		 1277 res = ext4_mark_inode_dirty(handle, inode);
1278 if (res)
1279 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1280 }
1281 res2 = ext4_journal_stop(handle);
2f8f5e76
EB		 1282
1283 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1284 goto retry;
a7550b30
JK		 1285 if (!res)
1286 res = res2;
1287 return res;
1288}
1289
c250b7dd 1290static bool ext4_dummy_context(struct inode *inode)
a7550b30
JK		 1291{
1292 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1293}
1294
6f69f0ed 1295static const struct fscrypt_operations ext4_cryptops = {
a5d431ef 1296 .key_prefix = "ext4:",
a7550b30 1297 .get_context = ext4_get_context,
a7550b30
JK		 1298 .set_context = ext4_set_context,
1299 .dummy_context = ext4_dummy_context,
a7550b30 1300 .empty_dir = ext4_empty_dir,
e12ee683 1301 .max_namelen = EXT4_NAME_LEN,
a7550b30 1302};
a7550b30
JK		 1303#endif
1304
ac27a0ec 1305#ifdef CONFIG_QUOTA
d6006186 1306static const char * const quotatypes[] = INITQFNAMES;
689c958c 1307#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1308
617ba13b
MC		 1309static int ext4_write_dquot(struct dquot *dquot);
1310static int ext4_acquire_dquot(struct dquot *dquot);
1311static int ext4_release_dquot(struct dquot *dquot);
1312static int ext4_mark_dquot_dirty(struct dquot *dquot);
1313static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1314static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1315 const struct path *path);
617ba13b
MC		 1316static int ext4_quota_on_mount(struct super_block *sb, int type);
1317static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1318 size_t len, loff_t off);
617ba13b 1319static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1320 const char *data, size_t len, loff_t off);
7c319d32
AK		 1321static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1322 unsigned int flags);
1323static int ext4_enable_quotas(struct super_block *sb);
8f0e8746 1324static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
ac27a0ec 1325
96c7e0d9
JK		 1326static struct dquot **ext4_get_dquots(struct inode *inode)
1327{
1328 return EXT4_I(inode)->i_dquot;
1329}
1330
61e225dc 1331static const struct dquot_operations ext4_quota_operations = {
7a9ca53a
TE		 1332 .get_reserved_space = ext4_get_reserved_space,
1333 .write_dquot = ext4_write_dquot,
1334 .acquire_dquot = ext4_acquire_dquot,
1335 .release_dquot = ext4_release_dquot,
1336 .mark_dirty = ext4_mark_dquot_dirty,
1337 .write_info = ext4_write_info,
1338 .alloc_dquot = dquot_alloc,
1339 .destroy_dquot = dquot_destroy,
1340 .get_projid = ext4_get_projid,
1341 .get_inode_usage = ext4_get_inode_usage,
1342 .get_next_id = ext4_get_next_id,
ac27a0ec
DK		 1343};
1344
0d54b217 1345static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1346 .quota_on = ext4_quota_on,
ca0e05e4 1347 .quota_off = ext4_quota_off,
287a8095 1348 .quota_sync = dquot_quota_sync,
0a240339 1349 .get_state = dquot_get_state,
287a8095
CH		 1350 .set_info = dquot_set_dqinfo,
1351 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1352 .set_dqblk = dquot_set_dqblk,
1353 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1354};
1355#endif
1356
ee9b6d61 1357static const struct super_operations ext4_sops = {
617ba13b
MC		 1358 .alloc_inode = ext4_alloc_inode,
1359 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1360 .write_inode = ext4_write_inode,
1361 .dirty_inode = ext4_dirty_inode,
7ff9c073 1362 .drop_inode = ext4_drop_inode,
0930fcc1 1363 .evict_inode = ext4_evict_inode,
617ba13b 1364 .put_super = ext4_put_super,
617ba13b 1365 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1366 .freeze_fs = ext4_freeze,
1367 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1368 .statfs = ext4_statfs,
1369 .remount_fs = ext4_remount,
617ba13b 1370 .show_options = ext4_show_options,
ac27a0ec 1371#ifdef CONFIG_QUOTA
617ba13b
MC		 1372 .quota_read = ext4_quota_read,
1373 .quota_write = ext4_quota_write,
96c7e0d9 1374 .get_dquots = ext4_get_dquots,
ac27a0ec 1375#endif
c39a7f84 1376 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1377};
1378
39655164 1379static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1380 .fh_to_dentry = ext4_fh_to_dentry,
1381 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1382 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1383};
1384
1385enum {
1386 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1387 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1388 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1389 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1390 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1391 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1392 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1393 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1394 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1395 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1396 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1397 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
49da9392 1398 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
327eaf73
TT		 1399 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1400 Opt_nowarn_on_error, Opt_mblk_io_submit,
670e9875 1401 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1449032b 1402 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1403 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1404 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1405 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1406 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
ac27a0ec
DK		 1407};
1408
a447c093 1409static const match_table_t tokens = {
ac27a0ec
DK		 1410 {Opt_bsd_df, "bsddf"},
1411 {Opt_minix_df, "minixdf"},
1412 {Opt_grpid, "grpid"},
1413 {Opt_grpid, "bsdgroups"},
1414 {Opt_nogrpid, "nogrpid"},
1415 {Opt_nogrpid, "sysvgroups"},
1416 {Opt_resgid, "resgid=%u"},
1417 {Opt_resuid, "resuid=%u"},
1418 {Opt_sb, "sb=%u"},
1419 {Opt_err_cont, "errors=continue"},
1420 {Opt_err_panic, "errors=panic"},
1421 {Opt_err_ro, "errors=remount-ro"},
1422 {Opt_nouid32, "nouid32"},
ac27a0ec 1423 {Opt_debug, "debug"},
72578c33
TT		 1424 {Opt_removed, "oldalloc"},
1425 {Opt_removed, "orlov"},
ac27a0ec
DK		 1426 {Opt_user_xattr, "user_xattr"},
1427 {Opt_nouser_xattr, "nouser_xattr"},
1428 {Opt_acl, "acl"},
1429 {Opt_noacl, "noacl"},
e3bb52ae 1430 {Opt_noload, "norecovery"},
5a916be1 1431 {Opt_noload, "noload"},
72578c33
TT		 1432 {Opt_removed, "nobh"},
1433 {Opt_removed, "bh"},
ac27a0ec 1434 {Opt_commit, "commit=%u"},
30773840
TT		 1435 {Opt_min_batch_time, "min_batch_time=%u"},
1436 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1437 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1438 {Opt_journal_path, "journal_path=%s"},
818d276c 1439 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1440 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1441 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1442 {Opt_abort, "abort"},
1443 {Opt_data_journal, "data=journal"},
1444 {Opt_data_ordered, "data=ordered"},
1445 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1446 {Opt_data_err_abort, "data_err=abort"},
1447 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1448 {Opt_offusrjquota, "usrjquota="},
1449 {Opt_usrjquota, "usrjquota=%s"},
1450 {Opt_offgrpjquota, "grpjquota="},
1451 {Opt_grpjquota, "grpjquota=%s"},
1452 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1453 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1454 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1455 {Opt_grpquota, "grpquota"},
1456 {Opt_noquota, "noquota"},
1457 {Opt_quota, "quota"},
1458 {Opt_usrquota, "usrquota"},
49da9392 1459 {Opt_prjquota, "prjquota"},
ac27a0ec 1460 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1461 {Opt_barrier, "barrier"},
1462 {Opt_nobarrier, "nobarrier"},
25ec56b5 1463 {Opt_i_version, "i_version"},
923ae0ff 1464 {Opt_dax, "dax"},
c9de560d 1465 {Opt_stripe, "stripe=%u"},
64769240 1466 {Opt_delalloc, "delalloc"},
327eaf73
TT		 1467 {Opt_warn_on_error, "warn_on_error"},
1468 {Opt_nowarn_on_error, "nowarn_on_error"},
a26f4992
TT		 1469 {Opt_lazytime, "lazytime"},
1470 {Opt_nolazytime, "nolazytime"},
670e9875 1471 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
dd919b98 1472 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1473 {Opt_removed, "mblk_io_submit"},
1474 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1475 {Opt_block_validity, "block_validity"},
1476 {Opt_noblock_validity, "noblock_validity"},
240799cd 1477 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1478 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1479 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1480 {Opt_auto_da_alloc, "auto_da_alloc"},
1481 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1482 {Opt_dioread_nolock, "dioread_nolock"},
1483 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1484 {Opt_discard, "discard"},
1485 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1486 {Opt_init_itable, "init_itable=%u"},
1487 {Opt_init_itable, "init_itable"},
1488 {Opt_noinit_itable, "noinit_itable"},
df981d03 1489 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1490 {Opt_test_dummy_encryption, "test_dummy_encryption"},
cdb7ee4c
TE		 1491 {Opt_nombcache, "nombcache"},
1492 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
c7198b9c
TT		 1493 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1494 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1495 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1496 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1497 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1498 {Opt_err, NULL},
ac27a0ec
DK		 1499};
1500
617ba13b 1501static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1502{
617ba13b 1503 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1504 char *options = (char *) *data;
1505
1506 if (!options || strncmp(options, "sb=", 3) != 0)
1507 return 1; /* Default location */
0b8e58a1 1508
ac27a0ec 1509 options += 3;
0b8e58a1 1510 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1511 sb_block = simple_strtoul(options, &options, 0);
1512 if (*options && *options != ',') {
4776004f 1513 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1514 (char *) *data);
1515 return 1;
1516 }
1517 if (*options == ',')
1518 options++;
1519 *data = (void *) options;
0b8e58a1 1520
ac27a0ec
DK		 1521 return sb_block;
1522}
1523
b3881f74 1524#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
d6006186
EB		 1525static const char deprecated_msg[] =
1526 "Mount option \"%s\" will be removed by %s\n"
437ca0fd 1527 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1528
56c50f11
DM		 1529#ifdef CONFIG_QUOTA
1530static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1531{
1532 struct ext4_sb_info *sbi = EXT4_SB(sb);
1533 char *qname;
03dafb5f 1534 int ret = -1;
56c50f11
DM		 1535
1536 if (sb_any_quota_loaded(sb) &&
1537 !sbi->s_qf_names[qtype]) {
1538 ext4_msg(sb, KERN_ERR,
1539 "Cannot change journaled "
1540 "quota options when quota turned on");
57f73c2c 1541 return -1;
56c50f11 1542 }
e2b911c5 1543 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1544 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1545 "ignored when QUOTA feature is enabled");
1546 return 1;
262b4662 1547 }
56c50f11
DM		 1548 qname = match_strdup(args);
1549 if (!qname) {
1550 ext4_msg(sb, KERN_ERR,
1551 "Not enough memory for storing quotafile name");
57f73c2c 1552 return -1;
56c50f11 1553 }
03dafb5f
CG		 1554 if (sbi->s_qf_names[qtype]) {
1555 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1556 ret = 1;
1557 else
1558 ext4_msg(sb, KERN_ERR,
1559 "%s quota file already specified",
1560 QTYPE2NAME(qtype));
1561 goto errout;
56c50f11 1562 }
03dafb5f 1563 if (strchr(qname, '/')) {
56c50f11
DM		 1564 ext4_msg(sb, KERN_ERR,
1565 "quotafile must be on filesystem root");
03dafb5f 1566 goto errout;
56c50f11 1567 }
03dafb5f 1568 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1569 set_opt(sb, QUOTA);
56c50f11 1570 return 1;
03dafb5f
CG		 1571errout:
1572 kfree(qname);
1573 return ret;
56c50f11
DM		 1574}
1575
1576static int clear_qf_name(struct super_block *sb, int qtype)
1577{
1578
1579 struct ext4_sb_info *sbi = EXT4_SB(sb);
1580
1581 if (sb_any_quota_loaded(sb) &&
1582 sbi->s_qf_names[qtype]) {
1583 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1584 " when quota turned on");
57f73c2c 1585 return -1;
56c50f11 1586 }
03dafb5f 1587 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1588 sbi->s_qf_names[qtype] = NULL;
1589 return 1;
1590}
1591#endif
1592
26092bf5
TT		 1593#define MOPT_SET 0x0001
1594#define MOPT_CLEAR 0x0002
1595#define MOPT_NOSUPPORT 0x0004
1596#define MOPT_EXPLICIT 0x0008
1597#define MOPT_CLEAR_ERR 0x0010
1598#define MOPT_GTE0 0x0020
ac27a0ec 1599#ifdef CONFIG_QUOTA
26092bf5
TT		 1600#define MOPT_Q 0
1601#define MOPT_QFMT 0x0040
1602#else
1603#define MOPT_Q MOPT_NOSUPPORT
1604#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1605#endif
26092bf5 1606#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1607#define MOPT_NO_EXT2 0x0100
1608#define MOPT_NO_EXT3 0x0200
1609#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1610#define MOPT_STRING 0x0400
26092bf5
TT		 1611
1612static const struct mount_opts {
1613 int token;
1614 int mount_opt;
1615 int flags;
1616} ext4_mount_opts[] = {
1617 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1618 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1619 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1620 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1621 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1622 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1623 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1624 MOPT_EXT4_ONLY | MOPT_SET},
1625 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1626 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1627 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1628 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1629 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1630 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1631 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1632 MOPT_EXT4_ONLY | MOPT_CLEAR},
327eaf73
TT		 1633 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1634 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
c6d3d56d
DW		 1635 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1636 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1637 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1638 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1639 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1640 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1641 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1642 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1643 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1644 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1645 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1646 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1647 MOPT_NO_EXT2},
8dc0aa8c 1648 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1649 MOPT_NO_EXT2},
26092bf5
TT		 1650 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1651 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1652 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1653 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1654 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1655 {Opt_commit, 0, MOPT_GTE0},
1656 {Opt_max_batch_time, 0, MOPT_GTE0},
1657 {Opt_min_batch_time, 0, MOPT_GTE0},
1658 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1659 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1660 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1661 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1662 {Opt_resuid, 0, MOPT_GTE0},
1663 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1664 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1665 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1666 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1667 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1668 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1669 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1670 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1671 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1672 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1673#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1674 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1675 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1676#else
26092bf5
TT		 1677 {Opt_acl, 0, MOPT_NOSUPPORT},
1678 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1679#endif
26092bf5
TT		 1680 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1681 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
670e9875 1682 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
26092bf5
TT		 1683 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1684 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1685 MOPT_SET | MOPT_Q},
1686 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1687 MOPT_SET | MOPT_Q},
49da9392
JK		 1688 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1689 MOPT_SET | MOPT_Q},
26092bf5 1690 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1691 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1692 MOPT_CLEAR | MOPT_Q},
26092bf5
TT		 1693 {Opt_usrjquota, 0, MOPT_Q},
1694 {Opt_grpjquota, 0, MOPT_Q},
1695 {Opt_offusrjquota, 0, MOPT_Q},
1696 {Opt_offgrpjquota, 0, MOPT_Q},
1697 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1698 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1699 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1700 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1701 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
cdb7ee4c 1702 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
26092bf5
TT		 1703 {Opt_err, 0, 0}
1704};
1705
1706static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1707 substring_t *args, unsigned long *journal_devnum,
1708 unsigned int *journal_ioprio, int is_remount)
1709{
1710 struct ext4_sb_info *sbi = EXT4_SB(sb);
1711 const struct mount_opts *m;
08cefc7a
EB		 1712 kuid_t uid;
1713 kgid_t gid;
26092bf5
TT		 1714 int arg = 0;
1715
57f73c2c
TT		 1716#ifdef CONFIG_QUOTA
1717 if (token == Opt_usrjquota)
1718 return set_qf_name(sb, USRQUOTA, &args[0]);
1719 else if (token == Opt_grpjquota)
1720 return set_qf_name(sb, GRPQUOTA, &args[0]);
1721 else if (token == Opt_offusrjquota)
1722 return clear_qf_name(sb, USRQUOTA);
1723 else if (token == Opt_offgrpjquota)
1724 return clear_qf_name(sb, GRPQUOTA);
1725#endif
26092bf5 1726 switch (token) {
f7048605
TT		 1727 case Opt_noacl:
1728 case Opt_nouser_xattr:
1729 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1730 break;
26092bf5
TT		 1731 case Opt_sb:
1732 return 1; /* handled by get_sb_block() */
1733 case Opt_removed:
5f3633e3 1734 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1735 return 1;
26092bf5
TT		 1736 case Opt_abort:
1737 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1738 return 1;
1739 case Opt_i_version:
357fdad0 1740 sb->s_flags |= SB_I_VERSION;
26092bf5 1741 return 1;
a26f4992 1742 case Opt_lazytime:
1751e8a6 1743 sb->s_flags |= SB_LAZYTIME;
a26f4992
TT		 1744 return 1;
1745 case Opt_nolazytime:
1751e8a6 1746 sb->s_flags &= ~SB_LAZYTIME;
a26f4992 1747 return 1;
26092bf5
TT		 1748 }
1749
5f3633e3
JK		 1750 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1751 if (token == m->token)
1752 break;
1753
1754 if (m->token == Opt_err) {
1755 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1756 "or missing value", opt);
1757 return -1;
1758 }
1759
8dc0aa8c
TT		 1760 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1761 ext4_msg(sb, KERN_ERR,
1762 "Mount option \"%s\" incompatible with ext2", opt);
1763 return -1;
1764 }
1765 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1766 ext4_msg(sb, KERN_ERR,
1767 "Mount option \"%s\" incompatible with ext3", opt);
1768 return -1;
1769 }
1770
ad4eec61 1771 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1772 return -1;
1773 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1774 return -1;
c93cf2d7
DM		 1775 if (m->flags & MOPT_EXPLICIT) {
1776 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1777 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1778 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1779 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1780 } else
1781 return -1;
1782 }
5f3633e3
JK		 1783 if (m->flags & MOPT_CLEAR_ERR)
1784 clear_opt(sb, ERRORS_MASK);
1785 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1786 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1787 "options when quota turned on");
1788 return -1;
1789 }
1790
1791 if (m->flags & MOPT_NOSUPPORT) {
1792 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1793 } else if (token == Opt_commit) {
1794 if (arg == 0)
1795 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1796 sbi->s_commit_interval = HZ * arg;
670e9875
TT		 1797 } else if (token == Opt_debug_want_extra_isize) {
1798 sbi->s_want_extra_isize = arg;
5f3633e3 1799 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1800 sbi->s_max_batch_time = arg;
1801 } else if (token == Opt_min_batch_time) {
1802 sbi->s_min_batch_time = arg;
1803 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1804 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1805 ext4_msg(sb, KERN_ERR,
1806 "EXT4-fs: inode_readahead_blks must be "
1807 "0 or a power of 2 smaller than 2^31");
26092bf5 1808 return -1;
5f3633e3
JK		 1809 }
1810 sbi->s_inode_readahead_blks = arg;
1811 } else if (token == Opt_init_itable) {
1812 set_opt(sb, INIT_INODE_TABLE);
1813 if (!args->from)
1814 arg = EXT4_DEF_LI_WAIT_MULT;
1815 sbi->s_li_wait_mult = arg;
1816 } else if (token == Opt_max_dir_size_kb) {
1817 sbi->s_max_dir_size_kb = arg;
1818 } else if (token == Opt_stripe) {
1819 sbi->s_stripe = arg;
1820 } else if (token == Opt_resuid) {
1821 uid = make_kuid(current_user_ns(), arg);
1822 if (!uid_valid(uid)) {
1823 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1824 return -1;
1825 }
5f3633e3
JK		 1826 sbi->s_resuid = uid;
1827 } else if (token == Opt_resgid) {
1828 gid = make_kgid(current_user_ns(), arg);
1829 if (!gid_valid(gid)) {
1830 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1831 return -1;
1832 }
1833 sbi->s_resgid = gid;
1834 } else if (token == Opt_journal_dev) {
1835 if (is_remount) {
1836 ext4_msg(sb, KERN_ERR,
1837 "Cannot specify journal on remount");
1838 return -1;
1839 }
1840 *journal_devnum = arg;
ad4eec61
ES		 1841 } else if (token == Opt_journal_path) {
1842 char *journal_path;
1843 struct inode *journal_inode;
1844 struct path path;
1845 int error;
1846
1847 if (is_remount) {
1848 ext4_msg(sb, KERN_ERR,
1849 "Cannot specify journal on remount");
1850 return -1;
1851 }
1852 journal_path = match_strdup(&args[0]);
1853 if (!journal_path) {
1854 ext4_msg(sb, KERN_ERR, "error: could not dup "
1855 "journal device string");
1856 return -1;
1857 }
1858
1859 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1860 if (error) {
1861 ext4_msg(sb, KERN_ERR, "error: could not find "
1862 "journal device path: error %d", error);
1863 kfree(journal_path);
1864 return -1;
1865 }
1866
2b0143b5 1867 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1868 if (!S_ISBLK(journal_inode->i_mode)) {
1869 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1870 "is not a block device", journal_path);
1871 path_put(&path);
1872 kfree(journal_path);
1873 return -1;
1874 }
1875
1876 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1877 path_put(&path);
1878 kfree(journal_path);
5f3633e3
JK		 1879 } else if (token == Opt_journal_ioprio) {
1880 if (arg > 7) {
1881 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1882 " (must be 0-7)");
1883 return -1;
1884 }
1885 *journal_ioprio =
1886 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1887 } else if (token == Opt_test_dummy_encryption) {
1888#ifdef CONFIG_EXT4_FS_ENCRYPTION
1889 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1890 ext4_msg(sb, KERN_WARNING,
1891 "Test dummy encryption mode enabled");
1892#else
1893 ext4_msg(sb, KERN_WARNING,
1894 "Test dummy encryption mount option ignored");
1895#endif
5f3633e3
JK		 1896 } else if (m->flags & MOPT_DATAJ) {
1897 if (is_remount) {
1898 if (!sbi->s_journal)
1899 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1900 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1901 ext4_msg(sb, KERN_ERR,
26092bf5 1902 "Cannot change data mode on remount");
26092bf5 1903 return -1;
ac27a0ec 1904 }
26092bf5 1905 } else {
5f3633e3
JK		 1906 clear_opt(sb, DATA_FLAGS);
1907 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1908 }
5f3633e3
JK		 1909#ifdef CONFIG_QUOTA
1910 } else if (m->flags & MOPT_QFMT) {
1911 if (sb_any_quota_loaded(sb) &&
1912 sbi->s_jquota_fmt != m->mount_opt) {
1913 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1914 "quota options when quota turned on");
1915 return -1;
1916 }
e2b911c5 1917 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1918 ext4_msg(sb, KERN_INFO,
1919 "Quota format mount options ignored "
262b4662 1920 "when QUOTA feature is enabled");
c325a67c 1921 return 1;
262b4662 1922 }
5f3633e3 1923 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 1924#endif
923ae0ff 1925 } else if (token == Opt_dax) {
ef83b6e8
DW		 1926#ifdef CONFIG_FS_DAX
1927 ext4_msg(sb, KERN_WARNING,
1928 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1929 sbi->s_mount_opt |= m->mount_opt;
1930#else
923ae0ff
RZ		 1931 ext4_msg(sb, KERN_INFO, "dax option not supported");
1932 return -1;
5f3633e3 1933#endif
7915a861
AN		 1934 } else if (token == Opt_data_err_abort) {
1935 sbi->s_mount_opt |= m->mount_opt;
1936 } else if (token == Opt_data_err_ignore) {
1937 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 1938 } else {
1939 if (!args->from)
1940 arg = 1;
1941 if (m->flags & MOPT_CLEAR)
1942 arg = !arg;
1943 else if (unlikely(!(m->flags & MOPT_SET))) {
1944 ext4_msg(sb, KERN_WARNING,
1945 "buggy handling of option %s", opt);
1946 WARN_ON(1);
1947 return -1;
1948 }
1949 if (arg != 0)
1950 sbi->s_mount_opt |= m->mount_opt;
1951 else
1952 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1953 }
5f3633e3 1954 return 1;
26092bf5
TT		 1955}
1956
1957static int parse_options(char *options, struct super_block *sb,
1958 unsigned long *journal_devnum,
1959 unsigned int *journal_ioprio,
1960 int is_remount)
1961{
1962 struct ext4_sb_info *sbi = EXT4_SB(sb);
1963 char *p;
1964 substring_t args[MAX_OPT_ARGS];
1965 int token;
1966
1967 if (!options)
1968 return 1;
1969
1970 while ((p = strsep(&options, ",")) != NULL) {
1971 if (!*p)
1972 continue;
1973 /*
1974 * Initialize args struct so we know whether arg was
1975 * found; some options take optional arguments.
1976 */
caecd0af 1977 args[0].to = args[0].from = NULL;
26092bf5
TT		 1978 token = match_token(p, tokens, args);
1979 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1980 journal_ioprio, is_remount) < 0)
1981 return 0;
ac27a0ec
DK		 1982 }
1983#ifdef CONFIG_QUOTA
49da9392
JK		 1984 /*
1985 * We do the test below only for project quotas. 'usrquota' and
1986 * 'grpquota' mount options are allowed even without quota feature
1987 * to support legacy quotas in quota files.
1988 */
1989 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1990 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1991 "Cannot enable project quota enforcement.");
1992 return 0;
1993 }
1994 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1995 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1996 clear_opt(sb, USRQUOTA);
ac27a0ec 1997
482a7425 1998 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1999 clear_opt(sb, GRPQUOTA);
ac27a0ec 2000
56c50f11 2001 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 2002 ext4_msg(sb, KERN_ERR, "old and new quota "
2003 "format mixing");
ac27a0ec
DK		 2004 return 0;
2005 }
2006
2007 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 2008 ext4_msg(sb, KERN_ERR, "journaled quota format "
2009 "not specified");
ac27a0ec
DK		 2010 return 0;
2011 }
ac27a0ec
DK		 2012 }
2013#endif
261cb20c
JK		 2014 if (test_opt(sb, DIOREAD_NOLOCK)) {
2015 int blocksize =
2016 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2017
09cbfeaf 2018 if (blocksize < PAGE_SIZE) {
261cb20c
JK		 2019 ext4_msg(sb, KERN_ERR, "can't mount with "
2020 "dioread_nolock if block size != PAGE_SIZE");
2021 return 0;
2022 }
2023 }
ac27a0ec
DK		 2024 return 1;
2025}
2026
2adf6da8
TT		 2027static inline void ext4_show_quota_options(struct seq_file *seq,
2028 struct super_block *sb)
2029{
2030#if defined(CONFIG_QUOTA)
2031 struct ext4_sb_info *sbi = EXT4_SB(sb);
2032
2033 if (sbi->s_jquota_fmt) {
2034 char *fmtname = "";
2035
2036 switch (sbi->s_jquota_fmt) {
2037 case QFMT_VFS_OLD:
2038 fmtname = "vfsold";
2039 break;
2040 case QFMT_VFS_V0:
2041 fmtname = "vfsv0";
2042 break;
2043 case QFMT_VFS_V1:
2044 fmtname = "vfsv1";
2045 break;
2046 }
2047 seq_printf(seq, ",jqfmt=%s", fmtname);
2048 }
2049
2050 if (sbi->s_qf_names[USRQUOTA])
a068acf2 2051 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 2052
2053 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 2054 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 2055#endif
2056}
2057
5a916be1
TT		 2058static const char *token2str(int token)
2059{
50df9fd5 2060 const struct match_token *t;
5a916be1
TT		 2061
2062 for (t = tokens; t->token != Opt_err; t++)
2063 if (t->token == token && !strchr(t->pattern, '='))
2064 break;
2065 return t->pattern;
2066}
2067
2adf6da8
TT		 2068/*
2069 * Show an option if
2070 * - it's set to a non-default value OR
2071 * - if the per-sb default is different from the global default
2072 */
66acdcf4
TT		 2073static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2074 int nodefs)
2adf6da8 2075{
2adf6da8
TT		 2076 struct ext4_sb_info *sbi = EXT4_SB(sb);
2077 struct ext4_super_block *es = sbi->s_es;
68afa7e0 2078 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
5a916be1 2079 const struct mount_opts *m;
66acdcf4 2080 char sep = nodefs ? '\n' : ',';
2adf6da8 2081
66acdcf4
TT		 2082#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2083#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2084
2085 if (sbi->s_sb_block != 1)
5a916be1
TT		 2086 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2087
2088 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2089 int want_set = m->flags & MOPT_SET;
2090 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2091 (m->flags & MOPT_CLEAR_ERR))
2092 continue;
68afa7e0 2093 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 2094 continue; /* skip if same as the default */
2095 if ((want_set &&
2096 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2097 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2098 continue; /* select Opt_noFoo vs Opt_Foo */
2099 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2100 }
5a916be1 2101
08cefc7a 2102 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2103 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2104 SEQ_OPTS_PRINT("resuid=%u",
2105 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2106 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2107 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2108 SEQ_OPTS_PRINT("resgid=%u",
2109 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2110 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2111 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2112 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2113 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2114 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2115 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2116 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2117 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2118 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2119 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2120 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2121 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2122 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
357fdad0 2123 if (sb->s_flags & SB_I_VERSION)
5a916be1 2124 SEQ_OPTS_PUTS("i_version");
66acdcf4 2125 if (nodefs || sbi->s_stripe)
5a916be1 2126 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
68afa7e0
TN		 2127 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2128 (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2129 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2130 SEQ_OPTS_PUTS("data=journal");
2131 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2132 SEQ_OPTS_PUTS("data=ordered");
2133 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2134 SEQ_OPTS_PUTS("data=writeback");
2135 }
66acdcf4
TT		 2136 if (nodefs ||
2137 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2138 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2139 sbi->s_inode_readahead_blks);
2adf6da8 2140
ceec0376 2141 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
66acdcf4 2142 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2143 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2144 if (nodefs || sbi->s_max_dir_size_kb)
2145 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2146 if (test_opt(sb, DATA_ERR_ABORT))
2147 SEQ_OPTS_PUTS("data_err=abort");
2adf6da8
TT		 2148
2149 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2150 return 0;
2151}
2152
66acdcf4
TT		 2153static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2154{
2155 return _ext4_show_options(seq, root->d_sb, 0);
2156}
2157
ebd173be 2158int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2159{
2160 struct super_block *sb = seq->private;
2161 int rc;
2162
bc98a42c 2163 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
66acdcf4
TT		 2164 rc = _ext4_show_options(seq, sb, 1);
2165 seq_puts(seq, "\n");
2166 return rc;
2167}
2168
617ba13b 2169static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2170 int read_only)
2171{
617ba13b 2172 struct ext4_sb_info *sbi = EXT4_SB(sb);
c89128a0 2173 int err = 0;
ac27a0ec 2174
617ba13b 2175 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2176 ext4_msg(sb, KERN_ERR, "revision level too high, "
2177 "forcing read-only mode");
c89128a0 2178 err = -EROFS;
ac27a0ec
DK		 2179 }
2180 if (read_only)
281b5995 2181 goto done;
617ba13b 2182 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2183 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2184 "running e2fsck is recommended");
c8b459f4 2185 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2186 ext4_msg(sb, KERN_WARNING,
2187 "warning: mounting fs with errors, "
2188 "running e2fsck is recommended");
ed3ce80a 2189 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2190 le16_to_cpu(es->s_mnt_count) >=
2191 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2192 ext4_msg(sb, KERN_WARNING,
2193 "warning: maximal mount count reached, "
2194 "running e2fsck is recommended");
ac27a0ec 2195 else if (le32_to_cpu(es->s_checkinterval) &&
6a0678a7
AB		 2196 (ext4_get_tstamp(es, s_lastcheck) +
2197 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
b31e1552
ES		 2198 ext4_msg(sb, KERN_WARNING,
2199 "warning: checktime reached, "
2200 "running e2fsck is recommended");
0b8e58a1 2201 if (!sbi->s_journal)
0390131b 2202 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2203 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2204 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2205 le16_add_cpu(&es->s_mnt_count, 1);
6a0678a7 2206 ext4_update_tstamp(es, s_mtime);
617ba13b 2207 ext4_update_dynamic_rev(sb);
0390131b 2208 if (sbi->s_journal)
e2b911c5 2209 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2210
c89128a0 2211 err = ext4_commit_super(sb, 1);
281b5995 2212done:
ac27a0ec 2213 if (test_opt(sb, DEBUG))
a9df9a49 2214 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2215 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2216 sb->s_blocksize,
2217 sbi->s_groups_count,
617ba13b
MC		 2218 EXT4_BLOCKS_PER_GROUP(sb),
2219 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2220 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2221
7abc52c2 2222 cleancache_init_fs(sb);
c89128a0 2223 return err;
ac27a0ec
DK		 2224}
2225
117fff10
TT		 2226int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2227{
2228 struct ext4_sb_info *sbi = EXT4_SB(sb);
2229 struct flex_groups *new_groups;
2230 int size;
2231
2232 if (!sbi->s_log_groups_per_flex)
2233 return 0;
2234
2235 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2236 if (size <= sbi->s_flex_groups_allocated)
2237 return 0;
2238
2239 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
a7c3e901 2240 new_groups = kvzalloc(size, GFP_KERNEL);
117fff10
TT		 2241 if (!new_groups) {
2242 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2243 size / (int) sizeof(struct flex_groups));
2244 return -ENOMEM;
2245 }
2246
2247 if (sbi->s_flex_groups) {
2248 memcpy(new_groups, sbi->s_flex_groups,
2249 (sbi->s_flex_groups_allocated *
2250 sizeof(struct flex_groups)));
b93b41d4 2251 kvfree(sbi->s_flex_groups);
117fff10
TT		 2252 }
2253 sbi->s_flex_groups = new_groups;
2254 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2255 return 0;
2256}
2257
772cb7c8
JS		 2258static int ext4_fill_flex_info(struct super_block *sb)
2259{
2260 struct ext4_sb_info *sbi = EXT4_SB(sb);
2261 struct ext4_group_desc *gdp = NULL;
772cb7c8 2262 ext4_group_t flex_group;
117fff10 2263 int i, err;
772cb7c8 2264
503358ae 2265 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2266 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2267 sbi->s_log_groups_per_flex = 0;
2268 return 1;
2269 }
2270
117fff10
TT		 2271 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2272 if (err)
9933fc0a 2273 goto failed;
772cb7c8 2274
772cb7c8 2275 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2276 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2277
2278 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2279 atomic_add(ext4_free_inodes_count(sb, gdp),
2280 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2281 atomic64_add(ext4_free_group_clusters(sb, gdp),
2282 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2283 atomic_add(ext4_used_dirs_count(sb, gdp),
2284 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2285 }
2286
2287 return 1;
2288failed:
2289 return 0;
2290}
2291
e2b911c5 2292static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2293 struct ext4_group_desc *gdp)
717d50e4 2294{
b47820ed 2295 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2296 __u16 crc = 0;
feb0ab32 2297 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2298 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2299
9aa5d32b 2300 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2301 /* Use new metadata_csum algorithm */
feb0ab32 2302 __u32 csum32;
b47820ed 2303 __u16 dummy_csum = 0;
feb0ab32 2304
feb0ab32
DW		 2305 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2306 sizeof(le_group));
b47820ed
DJ		 2307 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2308 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2309 sizeof(dummy_csum));
2310 offset += sizeof(dummy_csum);
2311 if (offset < sbi->s_desc_size)
2312 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2313 sbi->s_desc_size - offset);
feb0ab32
DW		 2314
2315 crc = csum32 & 0xFFFF;
2316 goto out;
717d50e4
AD		 2317 }
2318
feb0ab32 2319 /* old crc16 code */
e2b911c5 2320 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2321 return 0;
2322
feb0ab32
DW		 2323 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2324 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2325 crc = crc16(crc, (__u8 *)gdp, offset);
2326 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2327 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2328 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2329 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2330 crc = crc16(crc, (__u8 *)gdp + offset,
2331 le16_to_cpu(sbi->s_es->s_desc_size) -
2332 offset);
2333
2334out:
717d50e4
AD		 2335 return cpu_to_le16(crc);
2336}
2337
feb0ab32 2338int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2339 struct ext4_group_desc *gdp)
2340{
feb0ab32 2341 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2342 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2343 return 0;
2344
2345 return 1;
2346}
2347
feb0ab32
DW		 2348void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2349 struct ext4_group_desc *gdp)
2350{
2351 if (!ext4_has_group_desc_csum(sb))
2352 return;
e2b911c5 2353 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2354}
2355
ac27a0ec 2356/* Called at mount-time, super-block is locked */
bfff6873 2357static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2358 ext4_fsblk_t sb_block,
bfff6873 2359 ext4_group_t *first_not_zeroed)
ac27a0ec 2360{
617ba13b
MC		 2361 struct ext4_sb_info *sbi = EXT4_SB(sb);
2362 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2363 ext4_fsblk_t last_block;
44de022c 2364 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
bd81d8ee
LV		 2365 ext4_fsblk_t block_bitmap;
2366 ext4_fsblk_t inode_bitmap;
2367 ext4_fsblk_t inode_table;
ce421581 2368 int flexbg_flag = 0;
bfff6873 2369 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2370
e2b911c5 2371 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2372 flexbg_flag = 1;
2373
af5bc92d 2374 ext4_debug("Checking group descriptors");
ac27a0ec 2375
197cd65a
AM		 2376 for (i = 0; i < sbi->s_groups_count; i++) {
2377 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2378
ce421581 2379 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2380 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2381 else
2382 last_block = first_block +
617ba13b 2383 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2384
bfff6873
LC		 2385 if ((grp == sbi->s_groups_count) &&
2386 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2387 grp = i;
2388
8fadc143 2389 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2390 if (block_bitmap == sb_block) {
2391 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2392 "Block bitmap for group %u overlaps "
2393 "superblock", i);
18db4b4e
TT		 2394 if (!sb_rdonly(sb))
2395 return 0;
829fa70d 2396 }
77260807
TT		 2397 if (block_bitmap >= sb_block + 1 &&
2398 block_bitmap <= last_bg_block) {
2399 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2400 "Block bitmap for group %u overlaps "
2401 "block group descriptors", i);
2402 if (!sb_rdonly(sb))
2403 return 0;
2404 }
2b2d6d01 2405 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2406 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2407 "Block bitmap for group %u not in group "
b31e1552 2408 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2409 return 0;
2410 }
8fadc143 2411 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2412 if (inode_bitmap == sb_block) {
2413 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2414 "Inode bitmap for group %u overlaps "
2415 "superblock", i);
18db4b4e
TT		 2416 if (!sb_rdonly(sb))
2417 return 0;
829fa70d 2418 }
77260807
TT		 2419 if (inode_bitmap >= sb_block + 1 &&
2420 inode_bitmap <= last_bg_block) {
2421 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2422 "Inode bitmap for group %u overlaps "
2423 "block group descriptors", i);
2424 if (!sb_rdonly(sb))
2425 return 0;
2426 }
2b2d6d01 2427 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2428 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2429 "Inode bitmap for group %u not in group "
b31e1552 2430 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2431 return 0;
2432 }
8fadc143 2433 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2434 if (inode_table == sb_block) {
2435 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2436 "Inode table for group %u overlaps "
2437 "superblock", i);
18db4b4e
TT		 2438 if (!sb_rdonly(sb))
2439 return 0;
829fa70d 2440 }
77260807
TT		 2441 if (inode_table >= sb_block + 1 &&
2442 inode_table <= last_bg_block) {
2443 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2444 "Inode table for group %u overlaps "
2445 "block group descriptors", i);
2446 if (!sb_rdonly(sb))
2447 return 0;
2448 }
bd81d8ee 2449 if (inode_table < first_block ||
2b2d6d01 2450 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2451 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2452 "Inode table for group %u not in group "
b31e1552 2453 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2454 return 0;
2455 }
955ce5f5 2456 ext4_lock_group(sb, i);
feb0ab32 2457 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2458 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2459 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2460 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2461 gdp)), le16_to_cpu(gdp->bg_checksum));
bc98a42c 2462 if (!sb_rdonly(sb)) {
955ce5f5 2463 ext4_unlock_group(sb, i);
8a266467 2464 return 0;
7ee1ec4c 2465 }
717d50e4 2466 }
955ce5f5 2467 ext4_unlock_group(sb, i);
ce421581
JS		 2468 if (!flexbg_flag)
2469 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2470 }
bfff6873
LC		 2471 if (NULL != first_not_zeroed)
2472 *first_not_zeroed = grp;
ac27a0ec
DK		 2473 return 1;
2474}
2475
617ba13b 2476/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2477 * the superblock) which were deleted from all directories, but held open by
2478 * a process at the time of a crash. We walk the list and try to delete these
2479 * inodes at recovery time (only with a read-write filesystem).
2480 *
2481 * In order to keep the orphan inode chain consistent during traversal (in
2482 * case of crash during recovery), we link each inode into the superblock
2483 * orphan list_head and handle it the same way as an inode deletion during
2484 * normal operation (which journals the operations for us).
2485 *
2486 * We only do an iget() and an iput() on each inode, which is very safe if we
2487 * accidentally point at an in-use or already deleted inode. The worst that
2488 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2489 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2490 * e2fsck was run on this filesystem, and it must have already done the orphan
2491 * inode cleanup for us, so we can safely abort without any further action.
2492 */
2b2d6d01
TT		 2493static void ext4_orphan_cleanup(struct super_block *sb,
2494 struct ext4_super_block *es)
ac27a0ec
DK		 2495{
2496 unsigned int s_flags = sb->s_flags;
2c98eb5e 2497 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec 2498#ifdef CONFIG_QUOTA
95f1fda4 2499 int quota_update = 0;
ac27a0ec
DK		 2500 int i;
2501#endif
2502 if (!es->s_last_orphan) {
2503 jbd_debug(4, "no orphan inodes to clean up\n");
2504 return;
2505 }
2506
a8f48a95 2507 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2508 ext4_msg(sb, KERN_ERR, "write access "
2509 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2510 return;
2511 }
2512
d39195c3
AG		 2513 /* Check if feature set would not allow a r/w mount */
2514 if (!ext4_feature_set_ok(sb, 0)) {
2515 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2516 "unknown ROCOMPAT features");
2517 return;
2518 }
2519
617ba13b 2520 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6 2521 /* don't clear list on RO mount w/ errors */
1751e8a6 2522 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
84474976 2523 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2524 "clearing orphan list.\n");
c25f9bc6
ES		 2525 es->s_last_orphan = 0;
2526 }
ac27a0ec
DK		 2527 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2528 return;
2529 }
2530
1751e8a6 2531 if (s_flags & SB_RDONLY) {
b31e1552 2532 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1751e8a6 2533 sb->s_flags &= ~SB_RDONLY;
ac27a0ec
DK		 2534 }
2535#ifdef CONFIG_QUOTA
2536 /* Needed for iput() to work correctly and not trash data */
1751e8a6 2537 sb->s_flags |= SB_ACTIVE;
95f1fda4 2538
2539 /*
2540 * Turn on quotas which were not enabled for read-only mounts if
2541 * filesystem has quota feature, so that they are updated correctly.
2542 */
1751e8a6 2543 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
95f1fda4 2544 int ret = ext4_enable_quotas(sb);
2545
2546 if (!ret)
2547 quota_update = 1;
2548 else
2549 ext4_msg(sb, KERN_ERR,
2550 "Cannot turn on quotas: error %d", ret);
2551 }
2552
2553 /* Turn on journaled quotas used for old sytle */
a2d4a646 2554 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2555 if (EXT4_SB(sb)->s_qf_names[i]) {
2556 int ret = ext4_quota_on_mount(sb, i);
95f1fda4 2557
2558 if (!ret)
2559 quota_update = 1;
2560 else
b31e1552
ES		 2561 ext4_msg(sb, KERN_ERR,
2562 "Cannot turn on journaled "
95f1fda4 2563 "quota: type %d: error %d", i, ret);
ac27a0ec
DK		 2564 }
2565 }
2566#endif
2567
2568 while (es->s_last_orphan) {
2569 struct inode *inode;
2570
c65d5c6c
VN		 2571 /*
2572 * We may have encountered an error during cleanup; if
2573 * so, skip the rest.
2574 */
2575 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2576 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2577 es->s_last_orphan = 0;
2578 break;
2579 }
2580
97bd42b9
JB		 2581 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2582 if (IS_ERR(inode)) {
ac27a0ec
DK		 2583 es->s_last_orphan = 0;
2584 break;
2585 }
2586
617ba13b 2587 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2588 dquot_initialize(inode);
ac27a0ec 2589 if (inode->i_nlink) {
566370a2
PT		 2590 if (test_opt(sb, DEBUG))
2591 ext4_msg(sb, KERN_DEBUG,
2592 "%s: truncating inode %lu to %lld bytes",
2593 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2594 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2595 inode->i_ino, inode->i_size);
5955102c 2596 inode_lock(inode);
55f252c9 2597 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 2598 ret = ext4_truncate(inode);
2599 if (ret)
2600 ext4_std_error(inode->i_sb, ret);
5955102c 2601 inode_unlock(inode);
ac27a0ec
DK		 2602 nr_truncates++;
2603 } else {
566370a2
PT		 2604 if (test_opt(sb, DEBUG))
2605 ext4_msg(sb, KERN_DEBUG,
2606 "%s: deleting unreferenced inode %lu",
2607 __func__, inode->i_ino);
ac27a0ec
DK		 2608 jbd_debug(2, "deleting unreferenced inode %lu\n",
2609 inode->i_ino);
2610 nr_orphans++;
2611 }
2612 iput(inode); /* The delete magic happens here! */
2613 }
2614
2b2d6d01 2615#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2616
2617 if (nr_orphans)
b31e1552
ES		 2618 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2619 PLURAL(nr_orphans));
ac27a0ec 2620 if (nr_truncates)
b31e1552
ES		 2621 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2622 PLURAL(nr_truncates));
ac27a0ec 2623#ifdef CONFIG_QUOTA
95f1fda4 2624 /* Turn off quotas if they were enabled for orphan cleanup */
2625 if (quota_update) {
2626 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2627 if (sb_dqopt(sb)->files[i])
2628 dquot_quota_off(sb, i);
2629 }
ac27a0ec
DK		 2630 }
2631#endif
1751e8a6 2632 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
ac27a0ec 2633}
0b8e58a1 2634
cd2291a4
ES		 2635/*
2636 * Maximal extent format file size.
2637 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2638 * extent format containers, within a sector_t, and within i_blocks
2639 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2640 * so that won't be a limiting factor.
2641 *
f17722f9
LC		 2642 * However there is other limiting factor. We do store extents in the form
2643 * of starting block and length, hence the resulting length of the extent
2644 * covering maximum file size must fit into on-disk format containers as
2645 * well. Given that length is always by 1 unit bigger than max unit (because
2646 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2647 *
cd2291a4
ES		 2648 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2649 */
f287a1a5 2650static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2651{
2652 loff_t res;
2653 loff_t upper_limit = MAX_LFS_FILESIZE;
2654
2655 /* small i_blocks in vfs inode? */
f287a1a5 2656 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2657 /*
90c699a9 2658 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2659 * i_block represent total blocks in 512 bytes
2660 * 32 == size of vfs inode i_blocks * 8
2661 */
2662 upper_limit = (1LL << 32) - 1;
2663
2664 /* total blocks in file system block size */
2665 upper_limit >>= (blkbits - 9);
2666 upper_limit <<= blkbits;
2667 }
2668
f17722f9
LC		 2669 /*
2670 * 32-bit extent-start container, ee_block. We lower the maxbytes
2671 * by one fs block, so ee_len can cover the extent of maximum file
2672 * size
2673 */
2674 res = (1LL << 32) - 1;
cd2291a4 2675 res <<= blkbits;
cd2291a4
ES		 2676
2677 /* Sanity check against vm- & vfs- imposed limits */
2678 if (res > upper_limit)
2679 res = upper_limit;
2680
2681 return res;
2682}
ac27a0ec 2683
ac27a0ec 2684/*
cd2291a4 2685 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2686 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2687 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2688 */
f287a1a5 2689static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2690{
617ba13b 2691 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2692 int meta_blocks;
2693 loff_t upper_limit;
0b8e58a1
AD		 2694 /* This is calculated to be the largest file size for a dense, block
2695 * mapped file such that the file's total number of 512-byte sectors,
2696 * including data and all indirect blocks, does not exceed (2^48 - 1).
2697 *
2698 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2699 * number of 512-byte sectors of the file.
0fc1b451
AK		 2700 */
2701
f287a1a5 2702 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2703 /*
90c699a9 2704 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2705 * the inode i_block field represents total file blocks in
2706 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2707 */
2708 upper_limit = (1LL << 32) - 1;
2709
2710 /* total blocks in file system block size */
2711 upper_limit >>= (bits - 9);
2712
2713 } else {
8180a562
AK		 2714 /*
2715 * We use 48 bit ext4_inode i_blocks
2716 * With EXT4_HUGE_FILE_FL set the i_blocks
2717 * represent total number of blocks in
2718 * file system block size
2719 */
0fc1b451
AK		 2720 upper_limit = (1LL << 48) - 1;
2721
0fc1b451
AK		 2722 }
2723
2724 /* indirect blocks */
2725 meta_blocks = 1;
2726 /* double indirect blocks */
2727 meta_blocks += 1 + (1LL << (bits-2));
2728 /* tripple indirect blocks */
2729 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2730
2731 upper_limit -= meta_blocks;
2732 upper_limit <<= bits;
ac27a0ec
DK		 2733
2734 res += 1LL << (bits-2);
2735 res += 1LL << (2*(bits-2));
2736 res += 1LL << (3*(bits-2));
2737 res <<= bits;
2738 if (res > upper_limit)
2739 res = upper_limit;
0fc1b451
AK		 2740
2741 if (res > MAX_LFS_FILESIZE)
2742 res = MAX_LFS_FILESIZE;
2743
ac27a0ec
DK		 2744 return res;
2745}
2746
617ba13b 2747static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2748 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2749{
617ba13b 2750 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2751 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2752 int has_super = 0;
2753
2754 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2755
e2b911c5 2756 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2757 return logical_sb_block + nr + 1;
ac27a0ec 2758 bg = sbi->s_desc_per_block * nr;
617ba13b 2759 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2760 has_super = 1;
0b8e58a1 2761
bd63f6b0
DW		 2762 /*
2763 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2764 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2765 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2766 * compensate.
2767 */
2768 if (sb->s_blocksize == 1024 && nr == 0 &&
49598e04 2769 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
bd63f6b0
DW		 2770 has_super++;
2771
617ba13b 2772 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2773}
2774
c9de560d
AT		 2775/**
2776 * ext4_get_stripe_size: Get the stripe size.
2777 * @sbi: In memory super block info
2778 *
2779 * If we have specified it via mount option, then
2780 * use the mount option value. If the value specified at mount time is
2781 * greater than the blocks per group use the super block value.
2782 * If the super block value is greater than blocks per group return 0.
2783 * Allocator needs it be less than blocks per group.
2784 *
2785 */
2786static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2787{
2788 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2789 unsigned long stripe_width =
2790 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2791 int ret;
c9de560d
AT		 2792
2793 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 2794 ret = sbi->s_stripe;
5469d7c3 2795 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 2796 ret = stripe_width;
5469d7c3 2797 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 2798 ret = stride;
2799 else
2800 ret = 0;
c9de560d 2801
3eb08658
DE		 2802 /*
2803 * If the stripe width is 1, this makes no sense and
2804 * we set it to 0 to turn off stripe handling code.
2805 */
2806 if (ret <= 1)
2807 ret = 0;
c9de560d 2808
3eb08658 2809 return ret;
c9de560d 2810}
ac27a0ec 2811
a13fb1a4
ES		 2812/*
2813 * Check whether this filesystem can be mounted based on
2814 * the features present and the RDONLY/RDWR mount requested.
2815 * Returns 1 if this filesystem can be mounted as requested,
2816 * 0 if it cannot be.
2817 */
2818static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2819{
e2b911c5 2820 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2821 ext4_msg(sb, KERN_ERR,
2822 "Couldn't mount because of "
2823 "unsupported optional features (%x)",
2824 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2825 ~EXT4_FEATURE_INCOMPAT_SUPP));
2826 return 0;
2827 }
2828
2829 if (readonly)
2830 return 1;
2831
e2b911c5 2832 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b 2833 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
1751e8a6 2834 sb->s_flags |= SB_RDONLY;
2cb5cc8b
DW		 2835 return 1;
2836 }
2837
a13fb1a4 2838 /* Check that feature set is OK for a read-write mount */
e2b911c5 2839 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2840 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2841 "unsupported optional features (%x)",
2842 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2843 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2844 return 0;
2845 }
2846 /*
2847 * Large file size enabled file system can only be mounted
2848 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2849 */
e2b911c5 2850 if (ext4_has_feature_huge_file(sb)) {
a13fb1a4
ES		 2851 if (sizeof(blkcnt_t) < sizeof(u64)) {
2852 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2853 "cannot be mounted RDWR without "
2854 "CONFIG_LBDAF");
2855 return 0;
2856 }
2857 }
e2b911c5 2858 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2859 ext4_msg(sb, KERN_ERR,
2860 "Can't support bigalloc feature without "
2861 "extents feature\n");
2862 return 0;
2863 }
7c319d32
AK		 2864
2865#ifndef CONFIG_QUOTA
e2b911c5 2866 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2867 ext4_msg(sb, KERN_ERR,
2868 "Filesystem with quota feature cannot be mounted RDWR "
2869 "without CONFIG_QUOTA");
2870 return 0;
2871 }
689c958c
LX		 2872 if (ext4_has_feature_project(sb) && !readonly) {
2873 ext4_msg(sb, KERN_ERR,
2874 "Filesystem with project quota feature cannot be mounted RDWR "
2875 "without CONFIG_QUOTA");
2876 return 0;
2877 }
7c319d32 2878#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2879 return 1;
2880}
2881
66e61a9e
TT		 2882/*
2883 * This function is called once a day if we have errors logged
2884 * on the file system
2885 */
235699a8 2886static void print_daily_error_info(struct timer_list *t)
66e61a9e 2887{
235699a8
KC		 2888 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
2889 struct super_block *sb = sbi->s_sb;
2890 struct ext4_super_block *es = sbi->s_es;
66e61a9e
TT		 2891
2892 if (es->s_error_count)
ae0f78de
TT		 2893 /* fsck newer than v1.41.13 is needed to clean this condition. */
2894 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2895 le32_to_cpu(es->s_error_count));
2896 if (es->s_first_error_time) {
6a0678a7
AB		 2897 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
2898 sb->s_id,
2899 ext4_get_tstamp(es, s_first_error_time),
66e61a9e
TT		 2900 (int) sizeof(es->s_first_error_func),
2901 es->s_first_error_func,
2902 le32_to_cpu(es->s_first_error_line));
2903 if (es->s_first_error_ino)
651e1c3b 2904 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2905 le32_to_cpu(es->s_first_error_ino));
2906 if (es->s_first_error_block)
651e1c3b 2907 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2908 le64_to_cpu(es->s_first_error_block));
651e1c3b 2909 printk(KERN_CONT "\n");
66e61a9e
TT		 2910 }
2911 if (es->s_last_error_time) {
6a0678a7
AB		 2912 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
2913 sb->s_id,
2914 ext4_get_tstamp(es, s_last_error_time),
66e61a9e
TT		 2915 (int) sizeof(es->s_last_error_func),
2916 es->s_last_error_func,
2917 le32_to_cpu(es->s_last_error_line));
2918 if (es->s_last_error_ino)
651e1c3b 2919 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2920 le32_to_cpu(es->s_last_error_ino));
2921 if (es->s_last_error_block)
651e1c3b 2922 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2923 le64_to_cpu(es->s_last_error_block));
651e1c3b 2924 printk(KERN_CONT "\n");
66e61a9e
TT		 2925 }
2926 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2927}
2928
bfff6873
LC		 2929/* Find next suitable group and run ext4_init_inode_table */
2930static int ext4_run_li_request(struct ext4_li_request *elr)
2931{
2932 struct ext4_group_desc *gdp = NULL;
2933 ext4_group_t group, ngroups;
2934 struct super_block *sb;
2935 unsigned long timeout = 0;
2936 int ret = 0;
2937
2938 sb = elr->lr_super;
2939 ngroups = EXT4_SB(sb)->s_groups_count;
2940
2941 for (group = elr->lr_next_group; group < ngroups; group++) {
2942 gdp = ext4_get_group_desc(sb, group, NULL);
2943 if (!gdp) {
2944 ret = 1;
2945 break;
2946 }
2947
2948 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2949 break;
2950 }
2951
7f511862 2952 if (group >= ngroups)
bfff6873
LC		 2953 ret = 1;
2954
2955 if (!ret) {
2956 timeout = jiffies;
2957 ret = ext4_init_inode_table(sb, group,
2958 elr->lr_timeout ? 0 : 1);
2959 if (elr->lr_timeout == 0) {
51ce6511
LC		 2960 timeout = (jiffies - timeout) *
2961 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2962 elr->lr_timeout = timeout;
2963 }
2964 elr->lr_next_sched = jiffies + elr->lr_timeout;
2965 elr->lr_next_group = group + 1;
2966 }
bfff6873
LC		 2967 return ret;
2968}
2969
2970/*
2971 * Remove lr_request from the list_request and free the
4ed5c033 2972 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2973 */
2974static void ext4_remove_li_request(struct ext4_li_request *elr)
2975{
2976 struct ext4_sb_info *sbi;
2977
2978 if (!elr)
2979 return;
2980
2981 sbi = elr->lr_sbi;
2982
2983 list_del(&elr->lr_request);
2984 sbi->s_li_request = NULL;
2985 kfree(elr);
2986}
2987
2988static void ext4_unregister_li_request(struct super_block *sb)
2989{
1bb933fb
LC		 2990 mutex_lock(&ext4_li_mtx);
2991 if (!ext4_li_info) {
2992 mutex_unlock(&ext4_li_mtx);
bfff6873 2993 return;
1bb933fb 2994 }
bfff6873
LC		 2995
2996 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2997 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2998 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2999 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 3000}
3001
8f1f7453
ES		 3002static struct task_struct *ext4_lazyinit_task;
3003
bfff6873
LC		 3004/*
3005 * This is the function where ext4lazyinit thread lives. It walks
3006 * through the request list searching for next scheduled filesystem.
3007 * When such a fs is found, run the lazy initialization request
3008 * (ext4_rn_li_request) and keep track of the time spend in this
3009 * function. Based on that time we compute next schedule time of
3010 * the request. When walking through the list is complete, compute
3011 * next waking time and put itself into sleep.
3012 */
3013static int ext4_lazyinit_thread(void *arg)
3014{
3015 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3016 struct list_head *pos, *n;
3017 struct ext4_li_request *elr;
4ed5c033 3018 unsigned long next_wakeup, cur;
bfff6873
LC		 3019
3020 BUG_ON(NULL == eli);
3021
bfff6873
LC		 3022cont_thread:
3023 while (true) {
3024 next_wakeup = MAX_JIFFY_OFFSET;
3025
3026 mutex_lock(&eli->li_list_mtx);
3027 if (list_empty(&eli->li_request_list)) {
3028 mutex_unlock(&eli->li_list_mtx);
3029 goto exit_thread;
3030 }
bfff6873 3031 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 3032 int err = 0;
3033 int progress = 0;
bfff6873
LC		 3034 elr = list_entry(pos, struct ext4_li_request,
3035 lr_request);
3036
e22834f0
DM		 3037 if (time_before(jiffies, elr->lr_next_sched)) {
3038 if (time_before(elr->lr_next_sched, next_wakeup))
3039 next_wakeup = elr->lr_next_sched;
3040 continue;
3041 }
3042 if (down_read_trylock(&elr->lr_super->s_umount)) {
3043 if (sb_start_write_trylock(elr->lr_super)) {
3044 progress = 1;
3045 /*
3046 * We hold sb->s_umount, sb can not
3047 * be removed from the list, it is
3048 * now safe to drop li_list_mtx
3049 */
3050 mutex_unlock(&eli->li_list_mtx);
3051 err = ext4_run_li_request(elr);
3052 sb_end_write(elr->lr_super);
3053 mutex_lock(&eli->li_list_mtx);
3054 n = pos->next;
b2c78cd0 3055 }
e22834f0
DM		 3056 up_read((&elr->lr_super->s_umount));
3057 }
3058 /* error, remove the lazy_init job */
3059 if (err) {
3060 ext4_remove_li_request(elr);
3061 continue;
3062 }
3063 if (!progress) {
3064 elr->lr_next_sched = jiffies +
3065 (prandom_u32()
3066 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 3067 }
bfff6873
LC		 3068 if (time_before(elr->lr_next_sched, next_wakeup))
3069 next_wakeup = elr->lr_next_sched;
3070 }
3071 mutex_unlock(&eli->li_list_mtx);
3072
a0acae0e 3073 try_to_freeze();
bfff6873 3074
4ed5c033
LC		 3075 cur = jiffies;
3076 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3077 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3078 cond_resched();
3079 continue;
3080 }
3081
4ed5c033
LC		 3082 schedule_timeout_interruptible(next_wakeup - cur);
3083
8f1f7453
ES		 3084 if (kthread_should_stop()) {
3085 ext4_clear_request_list();
3086 goto exit_thread;
3087 }
bfff6873
LC		 3088 }
3089
3090exit_thread:
3091 /*
3092 * It looks like the request list is empty, but we need
3093 * to check it under the li_list_mtx lock, to prevent any
3094 * additions into it, and of course we should lock ext4_li_mtx
3095 * to atomically free the list and ext4_li_info, because at
3096 * this point another ext4 filesystem could be registering
3097 * new one.
3098 */
3099 mutex_lock(&ext4_li_mtx);
3100 mutex_lock(&eli->li_list_mtx);
3101 if (!list_empty(&eli->li_request_list)) {
3102 mutex_unlock(&eli->li_list_mtx);
3103 mutex_unlock(&ext4_li_mtx);
3104 goto cont_thread;
3105 }
3106 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3107 kfree(ext4_li_info);
3108 ext4_li_info = NULL;
3109 mutex_unlock(&ext4_li_mtx);
3110
3111 return 0;
3112}
3113
3114static void ext4_clear_request_list(void)
3115{
3116 struct list_head *pos, *n;
3117 struct ext4_li_request *elr;
3118
3119 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3120 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3121 elr = list_entry(pos, struct ext4_li_request,
3122 lr_request);
3123 ext4_remove_li_request(elr);
3124 }
3125 mutex_unlock(&ext4_li_info->li_list_mtx);
3126}
3127
3128static int ext4_run_lazyinit_thread(void)
3129{
8f1f7453
ES		 3130 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3131 ext4_li_info, "ext4lazyinit");
3132 if (IS_ERR(ext4_lazyinit_task)) {
3133 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3134 ext4_clear_request_list();
bfff6873
LC		 3135 kfree(ext4_li_info);
3136 ext4_li_info = NULL;
92b97816 3137 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3138 "initialization thread\n",
3139 err);
3140 return err;
3141 }
3142 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3143 return 0;
3144}
3145
3146/*
3147 * Check whether it make sense to run itable init. thread or not.
3148 * If there is at least one uninitialized inode table, return
3149 * corresponding group number, else the loop goes through all
3150 * groups and return total number of groups.
3151 */
3152static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3153{
3154 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3155 struct ext4_group_desc *gdp = NULL;
3156
8844618d
TT		 3157 if (!ext4_has_group_desc_csum(sb))
3158 return ngroups;
3159
bfff6873
LC		 3160 for (group = 0; group < ngroups; group++) {
3161 gdp = ext4_get_group_desc(sb, group, NULL);
3162 if (!gdp)
3163 continue;
3164
50122847 3165 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
bfff6873
LC		 3166 break;
3167 }
3168
3169 return group;
3170}
3171
3172static int ext4_li_info_new(void)
3173{
3174 struct ext4_lazy_init *eli = NULL;
3175
3176 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3177 if (!eli)
3178 return -ENOMEM;
3179
bfff6873
LC		 3180 INIT_LIST_HEAD(&eli->li_request_list);
3181 mutex_init(&eli->li_list_mtx);
3182
bfff6873
LC		 3183 eli->li_state |= EXT4_LAZYINIT_QUIT;
3184
3185 ext4_li_info = eli;
3186
3187 return 0;
3188}
3189
3190static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3191 ext4_group_t start)
3192{
3193 struct ext4_sb_info *sbi = EXT4_SB(sb);
3194 struct ext4_li_request *elr;
bfff6873
LC		 3195
3196 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3197 if (!elr)
3198 return NULL;
3199
3200 elr->lr_super = sb;
3201 elr->lr_sbi = sbi;
3202 elr->lr_next_group = start;
3203
3204 /*
3205 * Randomize first schedule time of the request to
3206 * spread the inode table initialization requests
3207 * better.
3208 */
dd1f723b
TT		 3209 elr->lr_next_sched = jiffies + (prandom_u32() %
3210 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3211 return elr;
3212}
3213
7f511862
TT		 3214int ext4_register_li_request(struct super_block *sb,
3215 ext4_group_t first_not_zeroed)
bfff6873
LC		 3216{
3217 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3218 struct ext4_li_request *elr = NULL;
49598e04 3219 ext4_group_t ngroups = sbi->s_groups_count;
6c5a6cb9 3220 int ret = 0;
bfff6873 3221
7f511862 3222 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3223 if (sbi->s_li_request != NULL) {
3224 /*
3225 * Reset timeout so it can be computed again, because
3226 * s_li_wait_mult might have changed.
3227 */
3228 sbi->s_li_request->lr_timeout = 0;
7f511862 3229 goto out;
51ce6511 3230 }
bfff6873 3231
bc98a42c 3232 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
55ff3840 3233 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3234 goto out;
bfff6873
LC		 3235
3236 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3237 if (!elr) {
3238 ret = -ENOMEM;
3239 goto out;
3240 }
bfff6873
LC		 3241
3242 if (NULL == ext4_li_info) {
3243 ret = ext4_li_info_new();
3244 if (ret)
3245 goto out;
3246 }
3247
3248 mutex_lock(&ext4_li_info->li_list_mtx);
3249 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3250 mutex_unlock(&ext4_li_info->li_list_mtx);
3251
3252 sbi->s_li_request = elr;
46e4690b
TM		 3253 /*
3254 * set elr to NULL here since it has been inserted to
3255 * the request_list and the removal and free of it is
3256 * handled by ext4_clear_request_list from now on.
3257 */
3258 elr = NULL;
bfff6873
LC		 3259
3260 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3261 ret = ext4_run_lazyinit_thread();
3262 if (ret)
3263 goto out;
3264 }
bfff6873 3265out:
beed5ecb
NK		 3266 mutex_unlock(&ext4_li_mtx);
3267 if (ret)
bfff6873 3268 kfree(elr);
bfff6873
LC		 3269 return ret;
3270}
3271
3272/*
3273 * We do not need to lock anything since this is called on
3274 * module unload.
3275 */
3276static void ext4_destroy_lazyinit_thread(void)
3277{
3278 /*
3279 * If thread exited earlier
3280 * there's nothing to be done.
3281 */
8f1f7453 3282 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3283 return;
3284
8f1f7453 3285 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3286}
3287
25ed6e8a
DW		 3288static int set_journal_csum_feature_set(struct super_block *sb)
3289{
3290 int ret = 1;
3291 int compat, incompat;
3292 struct ext4_sb_info *sbi = EXT4_SB(sb);
3293
9aa5d32b 3294 if (ext4_has_metadata_csum(sb)) {
db9ee220 3295 /* journal checksum v3 */
25ed6e8a 3296 compat = 0;
db9ee220 3297 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3298 } else {
3299 /* journal checksum v1 */
3300 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3301 incompat = 0;
3302 }
3303
feb8c6d3
DW		 3304 jbd2_journal_clear_features(sbi->s_journal,
3305 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3306 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3307 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3308 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3309 ret = jbd2_journal_set_features(sbi->s_journal,
3310 compat, 0,
3311 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3312 incompat);
3313 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3314 ret = jbd2_journal_set_features(sbi->s_journal,
3315 compat, 0,
3316 incompat);
3317 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3318 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3319 } else {
feb8c6d3
DW		 3320 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3321 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3322 }
3323
3324 return ret;
3325}
3326
952fc18e
TT		 3327/*
3328 * Note: calculating the overhead so we can be compatible with
3329 * historical BSD practice is quite difficult in the face of
3330 * clusters/bigalloc. This is because multiple metadata blocks from
3331 * different block group can end up in the same allocation cluster.
3332 * Calculating the exact overhead in the face of clustered allocation
3333 * requires either O(all block bitmaps) in memory or O(number of block
3334 * groups**2) in time. We will still calculate the superblock for
3335 * older file systems --- and if we come across with a bigalloc file
3336 * system with zero in s_overhead_clusters the estimate will be close to
3337 * correct especially for very large cluster sizes --- but for newer
3338 * file systems, it's better to calculate this figure once at mkfs
3339 * time, and store it in the superblock. If the superblock value is
3340 * present (even for non-bigalloc file systems), we will use it.
3341 */
3342static int count_overhead(struct super_block *sb, ext4_group_t grp,
3343 char *buf)
3344{
3345 struct ext4_sb_info *sbi = EXT4_SB(sb);
3346 struct ext4_group_desc *gdp;
3347 ext4_fsblk_t first_block, last_block, b;
3348 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3349 int s, j, count = 0;
3350
e2b911c5 3351 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3352 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3353 sbi->s_itb_per_group + 2);
3354
952fc18e
TT		 3355 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3356 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3357 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3358 for (i = 0; i < ngroups; i++) {
3359 gdp = ext4_get_group_desc(sb, i, NULL);
3360 b = ext4_block_bitmap(sb, gdp);
3361 if (b >= first_block && b <= last_block) {
3362 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3363 count++;
3364 }
3365 b = ext4_inode_bitmap(sb, gdp);
3366 if (b >= first_block && b <= last_block) {
3367 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3368 count++;
3369 }
3370 b = ext4_inode_table(sb, gdp);
3371 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3372 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3373 int c = EXT4_B2C(sbi, b - first_block);
3374 ext4_set_bit(c, buf);
3375 count++;
3376 }
3377 if (i != grp)
3378 continue;
3379 s = 0;
3380 if (ext4_bg_has_super(sb, grp)) {
3381 ext4_set_bit(s++, buf);
3382 count++;
3383 }
c48ae41b
TT		 3384 j = ext4_bg_num_gdb(sb, grp);
3385 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3386 ext4_error(sb, "Invalid number of block group "
3387 "descriptor blocks: %d", j);
3388 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3389 }
c48ae41b
TT		 3390 count += j;
3391 for (; j > 0; j--)
3392 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3393 }
3394 if (!count)
3395 return 0;
3396 return EXT4_CLUSTERS_PER_GROUP(sb) -
3397 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3398}
3399
3400/*
3401 * Compute the overhead and stash it in sbi->s_overhead
3402 */
3403int ext4_calculate_overhead(struct super_block *sb)
3404{
3405 struct ext4_sb_info *sbi = EXT4_SB(sb);
3406 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3407 struct inode *j_inode;
3408 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3409 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3410 ext4_fsblk_t overhead = 0;
4fdb5543 3411 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3412
952fc18e
TT		 3413 if (!buf)
3414 return -ENOMEM;
3415
3416 /*
3417 * Compute the overhead (FS structures). This is constant
3418 * for a given filesystem unless the number of block groups
3419 * changes so we cache the previous value until it does.
3420 */
3421
3422 /*
3423 * All of the blocks before first_data_block are overhead
3424 */
3425 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3426
3427 /*
3428 * Add the overhead found in each block group
3429 */
3430 for (i = 0; i < ngroups; i++) {
3431 int blks;
3432
3433 blks = count_overhead(sb, i, buf);
3434 overhead += blks;
3435 if (blks)
3436 memset(buf, 0, PAGE_SIZE);
3437 cond_resched();
3438 }
3c816ded
EW		 3439
3440 /*
3441 * Add the internal journal blocks whether the journal has been
3442 * loaded or not
3443 */
b003b524 3444 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3445 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3c816ded
EW		 3446 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3447 j_inode = ext4_get_journal_inode(sb, j_inum);
3448 if (j_inode) {
3449 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3450 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3451 iput(j_inode);
3452 } else {
3453 ext4_msg(sb, KERN_ERR, "can't get journal size");
3454 }
3455 }
952fc18e
TT		 3456 sbi->s_overhead = overhead;
3457 smp_wmb();
3458 free_page((unsigned long) buf);
3459 return 0;
3460}
3461
b5799018 3462static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3463{
3464 ext4_fsblk_t resv_clusters;
b5799018 3465 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3466
30fac0f7
JK		 3467 /*
3468 * There's no need to reserve anything when we aren't using extents.
3469 * The space estimates are exact, there are no unwritten extents,
3470 * hole punching doesn't need new metadata... This is needed especially
3471 * to keep ext2/3 backward compatibility.
3472 */
e2b911c5 3473 if (!ext4_has_feature_extents(sb))
b5799018 3474 return;
27dd4385
LC		 3475 /*
3476 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3477 * This should cover the situations where we can not afford to run
3478 * out of space like for example punch hole, or converting
556615dc 3479 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3480 * allocation would require 1, or 2 blocks, higher numbers are
3481 * very rare.
3482 */
b5799018
TT		 3483 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3484 sbi->s_cluster_bits);
27dd4385
LC		 3485
3486 do_div(resv_clusters, 50);
3487 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3488
b5799018 3489 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3490}
3491
2b2d6d01 3492static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3493{
5e405595 3494 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
d4c402d9 3495 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3496 struct buffer_head *bh;
617ba13b 3497 struct ext4_super_block *es = NULL;
5aee0f8a 3498 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
617ba13b
MC		 3499 ext4_fsblk_t block;
3500 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3501 ext4_fsblk_t logical_sb_block;
ac27a0ec 3502 unsigned long offset = 0;
ac27a0ec
DK		 3503 unsigned long journal_devnum = 0;
3504 unsigned long def_mount_opts;
3505 struct inode *root;
0390131b 3506 const char *descr;
dcc7dae3 3507 int ret = -ENOMEM;
281b5995 3508 int blocksize, clustersize;
4ec11028
TT		 3509 unsigned int db_count;
3510 unsigned int i;
281b5995 3511 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3512 __u64 blocks_count;
07aa2ea1 3513 int err = 0;
b3881f74 3514 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3515 ext4_group_t first_not_zeroed;
ac27a0ec 3516
5aee0f8a
TT		 3517 if ((data && !orig_data) || !sbi)
3518 goto out_free_base;
705895b6 3519
aed9eb1b 3520 sbi->s_daxdev = dax_dev;
705895b6
PE		 3521 sbi->s_blockgroup_lock =
3522 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 3523 if (!sbi->s_blockgroup_lock)
3524 goto out_free_base;
3525
ac27a0ec 3526 sb->s_fs_info = sbi;
2c0544b2 3527 sbi->s_sb = sb;
240799cd 3528 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3529 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3530 if (sb->s_bdev->bd_part)
3531 sbi->s_sectors_written_start =
3532 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3533
9f6200bb 3534 /* Cleanup superblock name */
ec3904dc 3535 strreplace(sb->s_id, '/', '!');
9f6200bb 3536
07aa2ea1 3537 /* -EINVAL is default */
dcc7dae3 3538 ret = -EINVAL;
617ba13b 3539 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3540 if (!blocksize) {
b31e1552 3541 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3542 goto out_fail;
3543 }
3544
3545 /*
617ba13b 3546 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3547 * block sizes. We need to calculate the offset from buffer start.
3548 */
617ba13b 3549 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3550 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3551 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3552 } else {
70bbb3e0 3553 logical_sb_block = sb_block;
ac27a0ec
DK		 3554 }
3555
a8ac900b 3556 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3557 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3558 goto out_fail;
3559 }
3560 /*
3561 * Note: s_es must be initialized as soon as possible because
617ba13b 3562 * some ext4 macro-instructions depend on its value
ac27a0ec 3563 */
2716b802 3564 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3565 sbi->s_es = es;
3566 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3567 if (sb->s_magic != EXT4_SUPER_MAGIC)
3568 goto cantfind_ext4;
afc32f7e 3569 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3570
feb0ab32 3571 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3572 if (ext4_has_feature_metadata_csum(sb) &&
3573 ext4_has_feature_gdt_csum(sb))
363307e6 3574 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3575 "redundant flags; please run fsck.");
3576
d25425f8
DW		 3577 /* Check for a known checksum algorithm */
3578 if (!ext4_verify_csum_type(sb, es)) {
3579 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3580 "unknown checksum algorithm.");
3581 silent = 1;
3582 goto cantfind_ext4;
3583 }
3584
0441984a 3585 /* Load the checksum driver */
a45403b5
TT		 3586 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3587 if (IS_ERR(sbi->s_chksum_driver)) {
3588 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3589 ret = PTR_ERR(sbi->s_chksum_driver);
3590 sbi->s_chksum_driver = NULL;
3591 goto failed_mount;
0441984a
DW		 3592 }
3593
a9c47317
DW		 3594 /* Check superblock checksum */
3595 if (!ext4_superblock_csum_verify(sb, es)) {
3596 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3597 "invalid superblock checksum. Run e2fsck?");
3598 silent = 1;
6a797d27 3599 ret = -EFSBADCRC;
a9c47317
DW		 3600 goto cantfind_ext4;
3601 }
3602
3603 /* Precompute checksum seed for all metadata */
e2b911c5 3604 if (ext4_has_feature_csum_seed(sb))
8c81bd8f 3605 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
dec214d0 3606 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
a9c47317
DW		 3607 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3608 sizeof(es->s_uuid));
3609
ac27a0ec
DK		 3610 /* Set defaults before we parse the mount options */
3611 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3612 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3613 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3614 set_opt(sb, DEBUG);
87f26807 3615 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3616 set_opt(sb, GRPID);
617ba13b 3617 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3618 set_opt(sb, NO_UID32);
ea663336 3619 /* xattr user namespace & acls are now defaulted on */
ea663336 3620 set_opt(sb, XATTR_USER);
03010a33 3621#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3622 set_opt(sb, POSIX_ACL);
2e7842b8 3623#endif
98c1a759
DW		 3624 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3625 if (ext4_has_metadata_csum(sb))
3626 set_opt(sb, JOURNAL_CHECKSUM);
3627
617ba13b 3628 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3629 set_opt(sb, JOURNAL_DATA);
617ba13b 3630 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3631 set_opt(sb, ORDERED_DATA);
617ba13b 3632 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3633 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3634
3635 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3636 set_opt(sb, ERRORS_PANIC);
bb4f397a 3637 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3638 set_opt(sb, ERRORS_CONT);
bb4f397a 3639 else
fd8c37ec 3640 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3641 /* block_validity enabled by default; disable with noblock_validity */
3642 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3643 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3644 set_opt(sb, DISCARD);
ac27a0ec 3645
08cefc7a
EB		 3646 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3647 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3648 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3649 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3650 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3651
8b67f04a 3652 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3653 set_opt(sb, BARRIER);
ac27a0ec 3654
dd919b98
AK		 3655 /*
3656 * enable delayed allocation by default
3657 * Use -o nodelalloc to turn it off
3658 */
bc0b75f7 3659 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3660 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3661 set_opt(sb, DELALLOC);
dd919b98 3662
51ce6511
LC		 3663 /*
3664 * set default s_li_wait_mult for lazyinit, for the case there is
3665 * no mount option specified.
3666 */
3667 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3668
5aee0f8a
TT		 3669 if (sbi->s_es->s_mount_opts[0]) {
3670 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3671 sizeof(sbi->s_es->s_mount_opts),
3672 GFP_KERNEL);
3673 if (!s_mount_opts)
3674 goto failed_mount;
3675 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3676 &journal_ioprio, 0)) {
3677 ext4_msg(sb, KERN_WARNING,
3678 "failed to parse options in superblock: %s",
3679 s_mount_opts);
3680 }
3681 kfree(s_mount_opts);
8b67f04a 3682 }
5a916be1 3683 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3684 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3685 &journal_ioprio, 0))
ac27a0ec
DK		 3686 goto failed_mount;
3687
56889787
TT		 3688 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3689 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3690 "with data=journal disables delayed "
3691 "allocation and O_DIRECT support!\n");
3692 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3693 ext4_msg(sb, KERN_ERR, "can't mount with "
3694 "both data=journal and delalloc");
3695 goto failed_mount;
3696 }
3697 if (test_opt(sb, DIOREAD_NOLOCK)) {
3698 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3699 "both data=journal and dioread_nolock");
56889787
TT		 3700 goto failed_mount;
3701 }
923ae0ff
RZ		 3702 if (test_opt(sb, DAX)) {
3703 ext4_msg(sb, KERN_ERR, "can't mount with "
3704 "both data=journal and dax");
3705 goto failed_mount;
3706 }
73b92a2a
SK		 3707 if (ext4_has_feature_encrypt(sb)) {
3708 ext4_msg(sb, KERN_WARNING,
3709 "encrypted files will use data=ordered "
3710 "instead of data journaling mode");
3711 }
56889787
TT		 3712 if (test_opt(sb, DELALLOC))
3713 clear_opt(sb, DELALLOC);
001e4a87
TH		 3714 } else {
3715 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3716 }
3717
1751e8a6
LT		 3718 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3719 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec 3720
617ba13b 3721 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3722 (ext4_has_compat_features(sb) ||
3723 ext4_has_ro_compat_features(sb) ||
3724 ext4_has_incompat_features(sb)))
b31e1552
ES		 3725 ext4_msg(sb, KERN_WARNING,
3726 "feature flags set on rev 0 fs, "
3727 "running e2fsck is recommended");
469108ff 3728
ed3654eb
TT		 3729 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3730 set_opt2(sb, HURD_COMPAT);
e2b911c5 3731 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3732 ext4_msg(sb, KERN_ERR,
3733 "The Hurd can't support 64-bit file systems");
3734 goto failed_mount;
3735 }
dec214d0
TE		 3736
3737 /*
3738 * ea_inode feature uses l_i_version field which is not
3739 * available in HURD_COMPAT mode.
3740 */
3741 if (ext4_has_feature_ea_inode(sb)) {
3742 ext4_msg(sb, KERN_ERR,
3743 "ea_inode feature is not supported for Hurd");
3744 goto failed_mount;
3745 }
ed3654eb
TT		 3746 }
3747
2035e776
TT		 3748 if (IS_EXT2_SB(sb)) {
3749 if (ext2_feature_set_ok(sb))
3750 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3751 "using the ext4 subsystem");
3752 else {
0d9366d6
ES		 3753 /*
3754 * If we're probing be silent, if this looks like
3755 * it's actually an ext[34] filesystem.
3756 */
3757 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3758 goto failed_mount;
2035e776
TT		 3759 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3760 "to feature incompatibilities");
3761 goto failed_mount;
3762 }
3763 }
3764
3765 if (IS_EXT3_SB(sb)) {
3766 if (ext3_feature_set_ok(sb))
3767 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3768 "using the ext4 subsystem");
3769 else {
0d9366d6
ES		 3770 /*
3771 * If we're probing be silent, if this looks like
3772 * it's actually an ext4 filesystem.
3773 */
3774 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3775 goto failed_mount;
2035e776
TT		 3776 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3777 "to feature incompatibilities");
3778 goto failed_mount;
3779 }
3780 }
3781
ac27a0ec
DK		 3782 /*
3783 * Check feature flags regardless of the revision level, since we
3784 * previously didn't change the revision level when setting the flags,
3785 * so there is a chance incompat flags are set on a rev 0 filesystem.
3786 */
bc98a42c 3787 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
ac27a0ec 3788 goto failed_mount;
a13fb1a4 3789
261cb20c 3790 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3791 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3792 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552 3793 ext4_msg(sb, KERN_ERR,
8cdf3372
TT		 3794 "Unsupported filesystem blocksize %d (%d log_block_size)",
3795 blocksize, le32_to_cpu(es->s_log_block_size));
3796 goto failed_mount;
3797 }
3798 if (le32_to_cpu(es->s_log_block_size) >
3799 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3800 ext4_msg(sb, KERN_ERR,
3801 "Invalid log block size: %u",
3802 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 3803 goto failed_mount;
3804 }
bfe0a5f4
TT		 3805 if (le32_to_cpu(es->s_log_cluster_size) >
3806 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3807 ext4_msg(sb, KERN_ERR,
3808 "Invalid log cluster size: %u",
3809 le32_to_cpu(es->s_log_cluster_size));
3810 goto failed_mount;
3811 }
ac27a0ec 3812
5b9554dc
TT		 3813 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3814 ext4_msg(sb, KERN_ERR,
3815 "Number of reserved GDT blocks insanely large: %d",
3816 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3817 goto failed_mount;
3818 }
3819
923ae0ff 3820 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
559db4c6
RZ		 3821 if (ext4_has_feature_inline_data(sb)) {
3822 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3823 " that may contain inline data");
24f3478d 3824 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
559db4c6 3825 }
80660f20 3826 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
24f3478d
DW		 3827 ext4_msg(sb, KERN_ERR,
3828 "DAX unsupported by block device. Turning off DAX.");
3829 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
3830 }
923ae0ff
RZ		 3831 }
3832
e2b911c5 3833 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 3834 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3835 es->s_encryption_level);
3836 goto failed_mount;
3837 }
3838
ac27a0ec 3839 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3840 /* Validate the filesystem blocksize */
3841 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3842 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3843 blocksize);
ac27a0ec
DK		 3844 goto failed_mount;
3845 }
3846
2b2d6d01 3847 brelse(bh);
70bbb3e0
AM		 3848 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3849 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3850 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3851 if (!bh) {
b31e1552
ES		 3852 ext4_msg(sb, KERN_ERR,
3853 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3854 goto failed_mount;
3855 }
2716b802 3856 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3857 sbi->s_es = es;
617ba13b 3858 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3859 ext4_msg(sb, KERN_ERR,
3860 "Magic mismatch, very weird!");
ac27a0ec
DK		 3861 goto failed_mount;
3862 }
3863 }
3864
e2b911c5 3865 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 3866 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3867 has_huge_files);
3868 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3869
617ba13b
MC		 3870 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3871 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3872 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3873 } else {
3874 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3875 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
c37e9e01
TT		 3876 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3877 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3878 sbi->s_first_ino);
3879 goto failed_mount;
3880 }
617ba13b 3881 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3882 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3883 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3884 ext4_msg(sb, KERN_ERR,
3885 "unsupported inode size: %d",
2b2d6d01 3886 sbi->s_inode_size);
ac27a0ec
DK		 3887 goto failed_mount;
3888 }
ef7f3835
KS		 3889 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3890 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3891 }
0b8e58a1 3892
0d1ee42f 3893 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 3894 if (ext4_has_feature_64bit(sb)) {
8fadc143 3895 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3896 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3897 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3898 ext4_msg(sb, KERN_ERR,
3899 "unsupported descriptor size %lu",
0d1ee42f
AR		 3900 sbi->s_desc_size);
3901 goto failed_mount;
3902 }
3903 } else
3904 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3905
ac27a0ec 3906 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3907 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 3908
617ba13b 3909 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3910 if (sbi->s_inodes_per_block == 0)
617ba13b 3911 goto cantfind_ext4;
cd6bb35b
TT		 3912 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3913 sbi->s_inodes_per_group > blocksize * 8) {
3914 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3915 sbi->s_blocks_per_group);
3916 goto failed_mount;
3917 }
ac27a0ec
DK		 3918 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3919 sbi->s_inodes_per_block;
0d1ee42f 3920 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3921 sbi->s_sbh = bh;
3922 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3923 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3924 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3925
2b2d6d01 3926 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3927 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3928 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 3929 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 3930 i = le32_to_cpu(es->s_flags);
3931 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3932 sbi->s_hash_unsigned = 3;
3933 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3934#ifdef __CHAR_UNSIGNED__
bc98a42c 3935 if (!sb_rdonly(sb))
23301410
TT		 3936 es->s_flags |=
3937 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3938 sbi->s_hash_unsigned = 3;
f99b2589 3939#else
bc98a42c 3940 if (!sb_rdonly(sb))
23301410
TT		 3941 es->s_flags |=
3942 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3943#endif
23301410 3944 }
f99b2589 3945 }
ac27a0ec 3946
281b5995
TT		 3947 /* Handle clustersize */
3948 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 3949 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 3950 if (has_bigalloc) {
3951 if (clustersize < blocksize) {
3952 ext4_msg(sb, KERN_ERR,
3953 "cluster size (%d) smaller than "
3954 "block size (%d)", clustersize, blocksize);
3955 goto failed_mount;
3956 }
3957 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3958 le32_to_cpu(es->s_log_block_size);
3959 sbi->s_clusters_per_group =
3960 le32_to_cpu(es->s_clusters_per_group);
3961 if (sbi->s_clusters_per_group > blocksize * 8) {
3962 ext4_msg(sb, KERN_ERR,
3963 "#clusters per group too big: %lu",
3964 sbi->s_clusters_per_group);
3965 goto failed_mount;
3966 }
3967 if (sbi->s_blocks_per_group !=
3968 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3969 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3970 "clusters per group (%lu) inconsistent",
3971 sbi->s_blocks_per_group,
3972 sbi->s_clusters_per_group);
3973 goto failed_mount;
3974 }
3975 } else {
3976 if (clustersize != blocksize) {
bfe0a5f4
TT		 3977 ext4_msg(sb, KERN_ERR,
3978 "fragment/cluster size (%d) != "
3979 "block size (%d)", clustersize, blocksize);
3980 goto failed_mount;
281b5995
TT		 3981 }
3982 if (sbi->s_blocks_per_group > blocksize * 8) {
3983 ext4_msg(sb, KERN_ERR,
3984 "#blocks per group too big: %lu",
3985 sbi->s_blocks_per_group);
3986 goto failed_mount;
3987 }
3988 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3989 sbi->s_cluster_bits = 0;
ac27a0ec 3990 }
281b5995
TT		 3991 sbi->s_cluster_ratio = clustersize / blocksize;
3992
960fd856
TT		 3993 /* Do we have standard group size of clustersize * 8 blocks ? */
3994 if (sbi->s_blocks_per_group == clustersize << 3)
3995 set_opt2(sb, STD_GROUP_SIZE);
3996
bf43d84b
ES		 3997 /*
3998 * Test whether we have more sectors than will fit in sector_t,
3999 * and whether the max offset is addressable by the page cache.
4000 */
5a9ae68a 4001 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 4002 ext4_blocks_count(es));
5a9ae68a 4003 if (err) {
b31e1552 4004 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 4005 " too large to mount safely on this system");
ac27a0ec 4006 if (sizeof(sector_t) < 8)
90c699a9 4007 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 4008 goto failed_mount;
4009 }
4010
617ba13b
MC		 4011 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4012 goto cantfind_ext4;
e7c95593 4013
0f2ddca6
FTN		 4014 /* check blocks count against device size */
4015 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4016 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 4017 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4018 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 4019 ext4_blocks_count(es), blocks_count);
4020 goto failed_mount;
4021 }
4022
0b8e58a1
AD		 4023 /*
4024 * It makes no sense for the first data block to be beyond the end
4025 * of the filesystem.
4026 */
4027 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 4028 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 4029 "block %u is beyond end of filesystem (%llu)",
4030 le32_to_cpu(es->s_first_data_block),
4031 ext4_blocks_count(es));
e7c95593
ES		 4032 goto failed_mount;
4033 }
bfe0a5f4
TT		 4034 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4035 (sbi->s_cluster_ratio == 1)) {
4036 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4037 "block is 0 with a 1k block and cluster size");
4038 goto failed_mount;
4039 }
4040
bd81d8ee
LV		 4041 blocks_count = (ext4_blocks_count(es) -
4042 le32_to_cpu(es->s_first_data_block) +
4043 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4044 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 4045 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 4046 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 4047 "(block count %llu, first data block %u, "
b31e1552 4048 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 4049 ext4_blocks_count(es),
4050 le32_to_cpu(es->s_first_data_block),
4051 EXT4_BLOCKS_PER_GROUP(sb));
4052 goto failed_mount;
4053 }
bd81d8ee 4054 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 4055 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4056 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 4057 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4058 EXT4_DESC_PER_BLOCK(sb);
3a4b77cd 4059 if (ext4_has_feature_meta_bg(sb)) {
2ba3e6e8 4060 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3a4b77cd
EG		 4061 ext4_msg(sb, KERN_WARNING,
4062 "first meta block group too large: %u "
4063 "(group descriptor block count %u)",
4064 le32_to_cpu(es->s_first_meta_bg), db_count);
4065 goto failed_mount;
4066 }
4067 }
344476e1
KC		 4068 sbi->s_group_desc = kvmalloc_array(db_count,
4069 sizeof(struct buffer_head *),
4070 GFP_KERNEL);
ac27a0ec 4071 if (sbi->s_group_desc == NULL) {
b31e1552 4072 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 4073 ret = -ENOMEM;
ac27a0ec
DK		 4074 goto failed_mount;
4075 }
bfe0a5f4
TT		 4076 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4077 le32_to_cpu(es->s_inodes_count)) {
4078 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4079 le32_to_cpu(es->s_inodes_count),
4080 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4081 ret = -EINVAL;
4082 goto failed_mount;
4083 }
ac27a0ec 4084
705895b6 4085 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec 4086
85c8f176
AP		 4087 /* Pre-read the descriptors into the buffer cache */
4088 for (i = 0; i < db_count; i++) {
4089 block = descriptor_loc(sb, logical_sb_block, i);
4090 sb_breadahead(sb, block);
4091 }
4092
ac27a0ec 4093 for (i = 0; i < db_count; i++) {
70bbb3e0 4094 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 4095 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 4096 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 4097 ext4_msg(sb, KERN_ERR,
4098 "can't read group descriptor %d", i);
ac27a0ec
DK		 4099 db_count = i;
4100 goto failed_mount2;
4101 }
4102 }
44de022c 4103 sbi->s_gdb_count = db_count;
829fa70d 4104 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 4105 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 4106 ret = -EFSCORRUPTED;
f9ae9cf5 4107 goto failed_mount2;
ac27a0ec 4108 }
772cb7c8 4109
235699a8 4110 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
04496411 4111
a75ae78f 4112 /* Register extent status tree shrinker */
eb68d0e2 4113 if (ext4_es_register_shrinker(sbi))
ce7e010a 4114 goto failed_mount3;
ce7e010a 4115
c9de560d 4116 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 4117 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 4118
f9ae9cf5
TT		 4119 /*
4120 * set up enough so that it can read an inode
4121 */
f6e63f90 4122 sb->s_op = &ext4_sops;
617ba13b
MC		 4123 sb->s_export_op = &ext4_export_ops;
4124 sb->s_xattr = ext4_xattr_handlers;
ffcc4182 4125#ifdef CONFIG_EXT4_FS_ENCRYPTION
a7550b30 4126 sb->s_cop = &ext4_cryptops;
ffcc4182 4127#endif
ac27a0ec 4128#ifdef CONFIG_QUOTA
617ba13b 4129 sb->dq_op = &ext4_quota_operations;
e2b911c5 4130 if (ext4_has_feature_quota(sb))
1fa5efe3 4131 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 4132 else
4133 sb->s_qcop = &ext4_qctl_operations;
689c958c 4134 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 4135#endif
85787090 4136 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
f2fa2ffc 4137
ac27a0ec 4138 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 4139 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 4140
4141 sb->s_root = NULL;
4142
4143 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 4144 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 4145
bc98a42c 4146 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
c5e06d10 4147 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 4148 goto failed_mount3a;
c5e06d10 4149
ac27a0ec
DK		 4150 /*
4151 * The first inode we look at is the journal inode. Don't try
4152 * root first: it may be modified in the journal!
4153 */
e2b911c5 4154 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4753d8a2
TT		 4155 err = ext4_load_journal(sb, es, journal_devnum);
4156 if (err)
50460fe8 4157 goto failed_mount3a;
bc98a42c 4158 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
e2b911c5 4159 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 4160 ext4_msg(sb, KERN_ERR, "required journal recovery "
4161 "suppressed and not mounted read-only");
744692dc 4162 goto failed_mount_wq;
ac27a0ec 4163 } else {
1e381f60
DM		 4164 /* Nojournal mode, all journal mount options are illegal */
4165 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4166 ext4_msg(sb, KERN_ERR, "can't mount with "
4167 "journal_checksum, fs mounted w/o journal");
4168 goto failed_mount_wq;
4169 }
4170 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4171 ext4_msg(sb, KERN_ERR, "can't mount with "
4172 "journal_async_commit, fs mounted w/o journal");
4173 goto failed_mount_wq;
4174 }
4175 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4176 ext4_msg(sb, KERN_ERR, "can't mount with "
4177 "commit=%lu, fs mounted w/o journal",
4178 sbi->s_commit_interval / HZ);
4179 goto failed_mount_wq;
4180 }
4181 if (EXT4_MOUNT_DATA_FLAGS &
4182 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4183 ext4_msg(sb, KERN_ERR, "can't mount with "
4184 "data=, fs mounted w/o journal");
4185 goto failed_mount_wq;
4186 }
4187 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4188 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 4189 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 4190 sbi->s_journal = NULL;
4191 needs_recovery = 0;
4192 goto no_journal;
ac27a0ec
DK		 4193 }
4194
e2b911c5 4195 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 4196 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4197 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4198 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4199 goto failed_mount_wq;
eb40a09c
JS		 4200 }
4201
25ed6e8a
DW		 4202 if (!set_journal_csum_feature_set(sb)) {
4203 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4204 "feature set");
4205 goto failed_mount_wq;
d4da6c9c 4206 }
818d276c 4207
ac27a0ec
DK		 4208 /* We have now updated the journal if required, so we can
4209 * validate the data journaling mode. */
4210 switch (test_opt(sb, DATA_FLAGS)) {
4211 case 0:
4212 /* No mode set, assume a default based on the journal
63f57933
AM		 4213 * capabilities: ORDERED_DATA if the journal can
4214 * cope, else JOURNAL_DATA
4215 */
dab291af 4216 if (jbd2_journal_check_available_features
27f394a7 4217 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
fd8c37ec 4218 set_opt(sb, ORDERED_DATA);
27f394a7
TN		 4219 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4220 } else {
fd8c37ec 4221 set_opt(sb, JOURNAL_DATA);
27f394a7
TN		 4222 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4223 }
ac27a0ec
DK		 4224 break;
4225
617ba13b
MC		 4226 case EXT4_MOUNT_ORDERED_DATA:
4227 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4228 if (!jbd2_journal_check_available_features
4229 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4230 ext4_msg(sb, KERN_ERR, "Journal does not support "
4231 "requested data journaling mode");
744692dc 4232 goto failed_mount_wq;
ac27a0ec
DK		 4233 }
4234 default:
4235 break;
4236 }
ab04df78
JK		 4237
4238 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4239 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4240 ext4_msg(sb, KERN_ERR, "can't mount with "
4241 "journal_async_commit in data=ordered mode");
4242 goto failed_mount_wq;
4243 }
4244
b3881f74 4245 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4246
18aadd47
BJ		 4247 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4248
ce7e010a 4249no_journal:
cdb7ee4c
TE		 4250 if (!test_opt(sb, NO_MBCACHE)) {
4251 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4252 if (!sbi->s_ea_block_cache) {
dec214d0 4253 ext4_msg(sb, KERN_ERR,
cdb7ee4c 4254 "Failed to create ea_block_cache");
dec214d0
TE		 4255 goto failed_mount_wq;
4256 }
cdb7ee4c
TE		 4257
4258 if (ext4_has_feature_ea_inode(sb)) {
4259 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4260 if (!sbi->s_ea_inode_cache) {
4261 ext4_msg(sb, KERN_ERR,
4262 "Failed to create ea_inode_cache");
4263 goto failed_mount_wq;
4264 }
4265 }
9c191f70
M		 4266 }
4267
e2b911c5 4268 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
09cbfeaf 4269 (blocksize != PAGE_SIZE)) {
1cb767cd
TT		 4270 ext4_msg(sb, KERN_ERR,
4271 "Unsupported blocksize for fs encryption");
4272 goto failed_mount_wq;
4273 }
4274
bc98a42c 4275 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
e2b911c5
DW		 4276 !ext4_has_feature_encrypt(sb)) {
4277 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 4278 ext4_commit_super(sb, 1);
4279 }
4280
952fc18e
TT		 4281 /*
4282 * Get the # of file system overhead blocks from the
4283 * superblock if present.
4284 */
4285 if (es->s_overhead_clusters)
4286 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4287 else {
07aa2ea1
LC		 4288 err = ext4_calculate_overhead(sb);
4289 if (err)
952fc18e
TT		 4290 goto failed_mount_wq;
4291 }
4292
fd89d5f2
TH		 4293 /*
4294 * The maximum number of concurrent works can be high and
4295 * concurrency isn't really necessary. Limit it to 1.
4296 */
2e8fa54e
JK		 4297 EXT4_SB(sb)->rsv_conversion_wq =
4298 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4299 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4300 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4301 ret = -ENOMEM;
2e8fa54e
JK		 4302 goto failed_mount4;
4303 }
4304
ac27a0ec 4305 /*
dab291af 4306 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4307 * so we can safely mount the rest of the filesystem now.
4308 */
4309
1d1fe1ee
DH		 4310 root = ext4_iget(sb, EXT4_ROOT_INO);
4311 if (IS_ERR(root)) {
b31e1552 4312 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4313 ret = PTR_ERR(root);
32a9bb57 4314 root = NULL;
ac27a0ec
DK		 4315 goto failed_mount4;
4316 }
4317 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4318 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4319 iput(root);
ac27a0ec
DK		 4320 goto failed_mount4;
4321 }
48fde701 4322 sb->s_root = d_make_root(root);
1d1fe1ee 4323 if (!sb->s_root) {
b31e1552 4324 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4325 ret = -ENOMEM;
4326 goto failed_mount4;
4327 }
ac27a0ec 4328
c89128a0
JK		 4329 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4330 if (ret == -EROFS) {
1751e8a6 4331 sb->s_flags |= SB_RDONLY;
c89128a0
JK		 4332 ret = 0;
4333 } else if (ret)
4334 goto failed_mount4a;
ef7f3835
KS		 4335
4336 /* determine the minimum size of new large inodes, if present */
670e9875
TT		 4337 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4338 sbi->s_want_extra_isize == 0) {
ef7f3835
KS		 4339 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4340 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 4341 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 4342 if (sbi->s_want_extra_isize <
4343 le16_to_cpu(es->s_want_extra_isize))
4344 sbi->s_want_extra_isize =
4345 le16_to_cpu(es->s_want_extra_isize);
4346 if (sbi->s_want_extra_isize <
4347 le16_to_cpu(es->s_min_extra_isize))
4348 sbi->s_want_extra_isize =
4349 le16_to_cpu(es->s_min_extra_isize);
4350 }
4351 }
4352 /* Check if enough inode space is available */
4353 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4354 sbi->s_inode_size) {
4355 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4356 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4357 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4358 "available");
ef7f3835
KS		 4359 }
4360
b5799018 4361 ext4_set_resv_clusters(sb);
27dd4385 4362
6fd058f7
TT		 4363 err = ext4_setup_system_zone(sb);
4364 if (err) {
b31e1552 4365 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4366 "zone (%d)", err);
f9ae9cf5
TT		 4367 goto failed_mount4a;
4368 }
4369
4370 ext4_ext_init(sb);
4371 err = ext4_mb_init(sb);
4372 if (err) {
4373 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4374 err);
dcf2d804 4375 goto failed_mount5;
c2774d84
AK		 4376 }
4377
d5e03cbb
TT		 4378 block = ext4_count_free_clusters(sb);
4379 ext4_free_blocks_count_set(sbi->s_es,
4380 EXT4_C2B(sbi, block));
908c7f19
TH		 4381 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4382 GFP_KERNEL);
d5e03cbb
TT		 4383 if (!err) {
4384 unsigned long freei = ext4_count_free_inodes(sb);
4385 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4386 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4387 GFP_KERNEL);
d5e03cbb
TT		 4388 }
4389 if (!err)
4390 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4391 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4392 if (!err)
908c7f19
TH		 4393 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4394 GFP_KERNEL);
c8585c6f
DJ		 4395 if (!err)
4396 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4397
d5e03cbb
TT		 4398 if (err) {
4399 ext4_msg(sb, KERN_ERR, "insufficient memory");
4400 goto failed_mount6;
4401 }
4402
e2b911c5 4403 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 4404 if (!ext4_fill_flex_info(sb)) {
4405 ext4_msg(sb, KERN_ERR,
4406 "unable to initialize "
4407 "flex_bg meta info!");
4408 goto failed_mount6;
4409 }
4410
bfff6873
LC		 4411 err = ext4_register_li_request(sb, first_not_zeroed);
4412 if (err)
dcf2d804 4413 goto failed_mount6;
bfff6873 4414
b5799018 4415 err = ext4_register_sysfs(sb);
dcf2d804
TM		 4416 if (err)
4417 goto failed_mount7;
3197ebdb 4418
9b2ff357
JK		 4419#ifdef CONFIG_QUOTA
4420 /* Enable quota usage during mount. */
bc98a42c 4421 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
9b2ff357
JK		 4422 err = ext4_enable_quotas(sb);
4423 if (err)
4424 goto failed_mount8;
4425 }
4426#endif /* CONFIG_QUOTA */
4427
617ba13b
MC		 4428 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4429 ext4_orphan_cleanup(sb, es);
4430 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4431 if (needs_recovery) {
b31e1552 4432 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4433 ext4_mark_recovery_complete(sb, es);
4434 }
4435 if (EXT4_SB(sb)->s_journal) {
4436 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4437 descr = " journalled data mode";
4438 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4439 descr = " ordered data mode";
4440 else
4441 descr = " writeback data mode";
4442 } else
4443 descr = "out journal";
4444
79add3a3
LC		 4445 if (test_opt(sb, DISCARD)) {
4446 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4447 if (!blk_queue_discard(q))
4448 ext4_msg(sb, KERN_WARNING,
4449 "mounting with \"discard\" option, but "
4450 "the device does not support discard");
4451 }
4452
e294a537
TT		 4453 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4454 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 4455 "Opts: %.*s%s%s", descr,
4456 (int) sizeof(sbi->s_es->s_mount_opts),
4457 sbi->s_es->s_mount_opts,
e294a537 4458 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4459
66e61a9e
TT		 4460 if (es->s_error_count)
4461 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4462
efbed4dc
TT		 4463 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4464 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4465 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4466 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4467
d4c402d9 4468 kfree(orig_data);
ac27a0ec
DK		 4469 return 0;
4470
617ba13b 4471cantfind_ext4:
ac27a0ec 4472 if (!silent)
b31e1552 4473 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4474 goto failed_mount;
4475
72ba7450
TT		 4476#ifdef CONFIG_QUOTA
4477failed_mount8:
ebd173be 4478 ext4_unregister_sysfs(sb);
72ba7450 4479#endif
dcf2d804
TM		 4480failed_mount7:
4481 ext4_unregister_li_request(sb);
4482failed_mount6:
f9ae9cf5 4483 ext4_mb_release(sb);
d5e03cbb 4484 if (sbi->s_flex_groups)
b93b41d4 4485 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4486 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4487 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4488 percpu_counter_destroy(&sbi->s_dirs_counter);
4489 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4490failed_mount5:
f9ae9cf5
TT		 4491 ext4_ext_release(sb);
4492 ext4_release_system_zone(sb);
4493failed_mount4a:
94bf608a 4494 dput(sb->s_root);
32a9bb57 4495 sb->s_root = NULL;
94bf608a 4496failed_mount4:
b31e1552 4497 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4498 if (EXT4_SB(sb)->rsv_conversion_wq)
4499 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4500failed_mount_wq:
dec214d0
TE		 4501 if (sbi->s_ea_inode_cache) {
4502 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4503 sbi->s_ea_inode_cache = NULL;
4504 }
47387409
TE		 4505 if (sbi->s_ea_block_cache) {
4506 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4507 sbi->s_ea_block_cache = NULL;
82939d79 4508 }
0390131b
FM		 4509 if (sbi->s_journal) {
4510 jbd2_journal_destroy(sbi->s_journal);
4511 sbi->s_journal = NULL;
4512 }
50460fe8 4513failed_mount3a:
d3922a77 4514 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4515failed_mount3:
9105bb14 4516 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4517 if (sbi->s_mmp_tsk)
4518 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4519failed_mount2:
4520 for (i = 0; i < db_count; i++)
4521 brelse(sbi->s_group_desc[i]);
b93b41d4 4522 kvfree(sbi->s_group_desc);
ac27a0ec 4523failed_mount:
0441984a
DW		 4524 if (sbi->s_chksum_driver)
4525 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 4526#ifdef CONFIG_QUOTA
a2d4a646 4527 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4528 kfree(sbi->s_qf_names[i]);
4529#endif
617ba13b 4530 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4531 brelse(bh);
4532out_fail:
4533 sb->s_fs_info = NULL;
f6830165 4534 kfree(sbi->s_blockgroup_lock);
5aee0f8a 4535out_free_base:
ac27a0ec 4536 kfree(sbi);
d4c402d9 4537 kfree(orig_data);
5e405595 4538 fs_put_dax(dax_dev);
07aa2ea1 4539 return err ? err : ret;
ac27a0ec
DK		 4540}
4541
4542/*
4543 * Setup any per-fs journal parameters now. We'll do this both on
4544 * initial mount, once the journal has been initialised but before we've
4545 * done any recovery; and again on any subsequent remount.
4546 */
617ba13b 4547static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4548{
617ba13b 4549 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4550
30773840
TT		 4551 journal->j_commit_interval = sbi->s_commit_interval;
4552 journal->j_min_batch_time = sbi->s_min_batch_time;
4553 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4554
a931da6a 4555 write_lock(&journal->j_state_lock);
ac27a0ec 4556 if (test_opt(sb, BARRIER))
dab291af 4557 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4558 else
dab291af 4559 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4560 if (test_opt(sb, DATA_ERR_ABORT))
4561 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4562 else
4563 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4564 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4565}
4566
c6cb7e77
EW		 4567static struct inode *ext4_get_journal_inode(struct super_block *sb,
4568 unsigned int journal_inum)
ac27a0ec
DK		 4569{
4570 struct inode *journal_inode;
ac27a0ec 4571
c6cb7e77
EW		 4572 /*
4573 * Test for the existence of a valid inode on disk. Bad things
4574 * happen if we iget() an unused inode, as the subsequent iput()
4575 * will try to delete it.
4576 */
1d1fe1ee
DH		 4577 journal_inode = ext4_iget(sb, journal_inum);
4578 if (IS_ERR(journal_inode)) {
b31e1552 4579 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4580 return NULL;
4581 }
4582 if (!journal_inode->i_nlink) {
4583 make_bad_inode(journal_inode);
4584 iput(journal_inode);
b31e1552 4585 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4586 return NULL;
4587 }
4588
e5f8eab8 4589 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4590 journal_inode, journal_inode->i_size);
1d1fe1ee 4591 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4592 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4593 iput(journal_inode);
4594 return NULL;
4595 }
c6cb7e77
EW		 4596 return journal_inode;
4597}
4598
4599static journal_t *ext4_get_journal(struct super_block *sb,
4600 unsigned int journal_inum)
4601{
4602 struct inode *journal_inode;
4603 journal_t *journal;
4604
4605 BUG_ON(!ext4_has_feature_journal(sb));
4606
4607 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4608 if (!journal_inode)
4609 return NULL;
ac27a0ec 4610
dab291af 4611 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4612 if (!journal) {
b31e1552 4613 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4614 iput(journal_inode);
4615 return NULL;
4616 }
4617 journal->j_private = sb;
617ba13b 4618 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4619 return journal;
4620}
4621
617ba13b 4622static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4623 dev_t j_dev)
4624{
2b2d6d01 4625 struct buffer_head *bh;
ac27a0ec 4626 journal_t *journal;
617ba13b
MC		 4627 ext4_fsblk_t start;
4628 ext4_fsblk_t len;
ac27a0ec 4629 int hblock, blocksize;
617ba13b 4630 ext4_fsblk_t sb_block;
ac27a0ec 4631 unsigned long offset;
2b2d6d01 4632 struct ext4_super_block *es;
ac27a0ec
DK		 4633 struct block_device *bdev;
4634
e2b911c5 4635 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4636
b31e1552 4637 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4638 if (bdev == NULL)
4639 return NULL;
4640
ac27a0ec 4641 blocksize = sb->s_blocksize;
e1defc4f 4642 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4643 if (blocksize < hblock) {
b31e1552
ES		 4644 ext4_msg(sb, KERN_ERR,
4645 "blocksize too small for journal device");
ac27a0ec
DK		 4646 goto out_bdev;
4647 }
4648
617ba13b
MC		 4649 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4650 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4651 set_blocksize(bdev, blocksize);
4652 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4653 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4654 "external journal");
ac27a0ec
DK		 4655 goto out_bdev;
4656 }
4657
2716b802 4658 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4659 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4660 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4661 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4662 ext4_msg(sb, KERN_ERR, "external journal has "
4663 "bad superblock");
ac27a0ec
DK		 4664 brelse(bh);
4665 goto out_bdev;
4666 }
4667
df4763be
DW		 4668 if ((le32_to_cpu(es->s_feature_ro_compat) &
4669 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4670 es->s_checksum != ext4_superblock_csum(sb, es)) {
4671 ext4_msg(sb, KERN_ERR, "external journal has "
4672 "corrupt superblock");
4673 brelse(bh);
4674 goto out_bdev;
4675 }
4676
617ba13b 4677 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4678 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4679 brelse(bh);
4680 goto out_bdev;
4681 }
4682
bd81d8ee 4683 len = ext4_blocks_count(es);
ac27a0ec
DK		 4684 start = sb_block + 1;
4685 brelse(bh); /* we're done with the superblock */
4686
dab291af 4687 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4688 start, len, blocksize);
4689 if (!journal) {
b31e1552 4690 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4691 goto out_bdev;
4692 }
4693 journal->j_private = sb;
dfec8a14 4694 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4695 wait_on_buffer(journal->j_sb_buffer);
4696 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4697 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4698 goto out_journal;
4699 }
4700 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4701 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4702 "user (unsupported) - %d",
ac27a0ec
DK		 4703 be32_to_cpu(journal->j_superblock->s_nr_users));
4704 goto out_journal;
4705 }
617ba13b
MC		 4706 EXT4_SB(sb)->journal_bdev = bdev;
4707 ext4_init_journal_params(sb, journal);
ac27a0ec 4708 return journal;
0b8e58a1 4709
ac27a0ec 4710out_journal:
dab291af 4711 jbd2_journal_destroy(journal);
ac27a0ec 4712out_bdev:
617ba13b 4713 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4714 return NULL;
4715}
4716
617ba13b
MC		 4717static int ext4_load_journal(struct super_block *sb,
4718 struct ext4_super_block *es,
ac27a0ec
DK		 4719 unsigned long journal_devnum)
4720{
4721 journal_t *journal;
4722 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4723 dev_t journal_dev;
4724 int err = 0;
4725 int really_read_only;
4726
e2b911c5 4727 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4728
ac27a0ec
DK		 4729 if (journal_devnum &&
4730 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4731 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4732 "numbers have changed");
ac27a0ec
DK		 4733 journal_dev = new_decode_dev(journal_devnum);
4734 } else
4735 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4736
4737 really_read_only = bdev_read_only(sb->s_bdev);
4738
4739 /*
4740 * Are we loading a blank journal or performing recovery after a
4741 * crash? For recovery, we need to check in advance whether we
4742 * can get read-write access to the device.
4743 */
e2b911c5 4744 if (ext4_has_feature_journal_needs_recovery(sb)) {
bc98a42c 4745 if (sb_rdonly(sb)) {
b31e1552
ES		 4746 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4747 "required on readonly filesystem");
ac27a0ec 4748 if (really_read_only) {
b31e1552 4749 ext4_msg(sb, KERN_ERR, "write access "
d98bf8cd
SR		 4750 "unavailable, cannot proceed "
4751 "(try mounting with noload)");
ac27a0ec
DK		 4752 return -EROFS;
4753 }
b31e1552
ES		 4754 ext4_msg(sb, KERN_INFO, "write access will "
4755 "be enabled during recovery");
ac27a0ec
DK		 4756 }
4757 }
4758
4759 if (journal_inum && journal_dev) {
b31e1552
ES		 4760 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4761 "and inode journals!");
ac27a0ec
DK		 4762 return -EINVAL;
4763 }
4764
4765 if (journal_inum) {
617ba13b 4766 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4767 return -EINVAL;
4768 } else {
617ba13b 4769 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4770 return -EINVAL;
4771 }
4772
90576c0b 4773 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4774 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4775
e2b911c5 4776 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4777 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4778 if (!err) {
4779 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4780 if (save)
4781 memcpy(save, ((char *) es) +
4782 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4783 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4784 if (save)
4785 memcpy(((char *) es) + EXT4_S_ERR_START,
4786 save, EXT4_S_ERR_LEN);
4787 kfree(save);
4788 }
ac27a0ec
DK		 4789
4790 if (err) {
b31e1552 4791 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4792 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4793 return err;
4794 }
4795
617ba13b
MC		 4796 EXT4_SB(sb)->s_journal = journal;
4797 ext4_clear_journal_err(sb, es);
ac27a0ec 4798
c41303ce 4799 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4800 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4801 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4802
4803 /* Make sure we flush the recovery flag to disk. */
e2d67052 4804 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4805 }
4806
4807 return 0;
4808}
4809
e2d67052 4810static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4811{
e2d67052 4812 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4813 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4814 int error = 0;
ac27a0ec 4815
bdfe0cbd 4816 if (!sbh || block_device_ejected(sb))
c4be0c1d 4817 return error;
a17712c8
JD		 4818
4819 /*
4820 * The superblock bh should be mapped, but it might not be if the
4821 * device was hot-removed. Not much we can do but fail the I/O.
4822 */
4823 if (!buffer_mapped(sbh))
4824 return error;
4825
71290b36
TT		 4826 /*
4827 * If the file system is mounted read-only, don't update the
4828 * superblock write time. This avoids updating the superblock
4829 * write time when we are mounting the root file system
4830 * read/only but we need to replay the journal; at that point,
4831 * for people who are east of GMT and who make their clock
4832 * tick in localtime for Windows bug-for-bug compatibility,
4833 * the clock is set in the future, and this will cause e2fsck
4834 * to complain and force a full file system check.
4835 */
1751e8a6 4836 if (!(sb->s_flags & SB_RDONLY))
6a0678a7 4837 ext4_update_tstamp(es, s_wtime);
f613dfcb
TT		 4838 if (sb->s_bdev->bd_part)
4839 es->s_kbytes_written =
4840 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4841 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4842 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4843 else
4844 es->s_kbytes_written =
4845 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4846 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4847 ext4_free_blocks_count_set(es,
57042651
TT		 4848 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4849 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4850 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4851 es->s_free_inodes_count =
4852 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4853 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4854 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4855 ext4_superblock_csum_set(sb);
1566a48a
TT		 4856 if (sync)
4857 lock_buffer(sbh);
4743f839
PKS		 4858 if (buffer_write_io_error(sbh)) {
4859 /*
4860 * Oh, dear. A previous attempt to write the
4861 * superblock failed. This could happen because the
4862 * USB device was yanked out. Or it could happen to
4863 * be a transient write error and maybe the block will
4864 * be remapped. Nothing we can do but to retry the
4865 * write and hope for the best.
4866 */
4867 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4868 "superblock detected");
4869 clear_buffer_write_io_error(sbh);
4870 set_buffer_uptodate(sbh);
4871 }
ac27a0ec 4872 mark_buffer_dirty(sbh);
914258bf 4873 if (sync) {
1566a48a 4874 unlock_buffer(sbh);
564bc402 4875 error = __sync_dirty_buffer(sbh,
00473374 4876 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
c89128a0 4877 if (buffer_write_io_error(sbh)) {
b31e1552
ES		 4878 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4879 "superblock");
914258bf
TT		 4880 clear_buffer_write_io_error(sbh);
4881 set_buffer_uptodate(sbh);
4882 }
4883 }
c4be0c1d 4884 return error;
ac27a0ec
DK		 4885}
4886
ac27a0ec
DK		 4887/*
4888 * Have we just finished recovery? If so, and if we are mounting (or
4889 * remounting) the filesystem readonly, then we will end up with a
4890 * consistent fs on disk. Record that fact.
4891 */
2b2d6d01
TT		 4892static void ext4_mark_recovery_complete(struct super_block *sb,
4893 struct ext4_super_block *es)
ac27a0ec 4894{
617ba13b 4895 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4896
e2b911c5 4897 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4898 BUG_ON(journal != NULL);
4899 return;
4900 }
dab291af 4901 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4902 if (jbd2_journal_flush(journal) < 0)
4903 goto out;
4904
bc98a42c 4905 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
e2b911c5 4906 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4907 ext4_commit_super(sb, 1);
ac27a0ec 4908 }
7ffe1ea8
HK		 4909
4910out:
dab291af 4911 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4912}
4913
4914/*
4915 * If we are mounting (or read-write remounting) a filesystem whose journal
4916 * has recorded an error from a previous lifetime, move that error to the
4917 * main filesystem now.
4918 */
2b2d6d01
TT		 4919static void ext4_clear_journal_err(struct super_block *sb,
4920 struct ext4_super_block *es)
ac27a0ec
DK		 4921{
4922 journal_t *journal;
4923 int j_errno;
4924 const char *errstr;
4925
e2b911c5 4926 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4927
617ba13b 4928 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4929
4930 /*
4931 * Now check for any error status which may have been recorded in the
617ba13b 4932 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4933 */
4934
dab291af 4935 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4936 if (j_errno) {
4937 char nbuf[16];
4938
617ba13b 4939 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4940 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4941 "from previous mount: %s", errstr);
12062ddd 4942 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4943
617ba13b
MC		 4944 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4945 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4946 ext4_commit_super(sb, 1);
ac27a0ec 4947
dab291af 4948 jbd2_journal_clear_err(journal);
d796c52e 4949 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4950 }
4951}
4952
4953/*
4954 * Force the running and committing transactions to commit,
4955 * and wait on the commit.
4956 */
617ba13b 4957int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4958{
4959 journal_t *journal;
ac27a0ec 4960
bc98a42c 4961 if (sb_rdonly(sb))
ac27a0ec
DK		 4962 return 0;
4963
617ba13b 4964 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4965 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4966}
4967
617ba13b 4968static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4969{
14ce0cb4 4970 int ret = 0;
9eddacf9 4971 tid_t target;
06a407f1 4972 bool needs_barrier = false;
8d5d02e6 4973 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4974
49598e04 4975 if (unlikely(ext4_forced_shutdown(sbi)))
0db1ff22
TT		 4976 return 0;
4977
9bffad1e 4978 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4979 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4980 /*
4981 * Writeback quota in non-journalled quota case - journalled quota has
4982 * no dirty dquots
4983 */
4984 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4985 /*
4986 * Data writeback is possible w/o journal transaction, so barrier must
4987 * being sent at the end of the function. But we can skip it if
4988 * transaction_commit will do it for us.
4989 */
bda32530
TT		 4990 if (sbi->s_journal) {
4991 target = jbd2_get_latest_transaction(sbi->s_journal);
4992 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4993 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4994 needs_barrier = true;
4995
4996 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4997 if (wait)
4998 ret = jbd2_log_wait_commit(sbi->s_journal,
4999 target);
5000 }
5001 } else if (wait && test_opt(sb, BARRIER))
06a407f1 5002 needs_barrier = true;
06a407f1
DM		 5003 if (needs_barrier) {
5004 int err;
5005 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
5006 if (!ret)
5007 ret = err;
0390131b 5008 }
06a407f1
DM		 5009
5010 return ret;
5011}
5012
ac27a0ec
DK		 5013/*
5014 * LVM calls this function before a (read-only) snapshot is created. This
5015 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 5016 *
5017 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 5018 * state independently. It relies on upper layer to stop all data & metadata
5019 * modifications.
ac27a0ec 5020 */
c4be0c1d 5021static int ext4_freeze(struct super_block *sb)
ac27a0ec 5022{
c4be0c1d
TS		 5023 int error = 0;
5024 journal_t *journal;
ac27a0ec 5025
bc98a42c 5026 if (sb_rdonly(sb))
9ca92389 5027 return 0;
ac27a0ec 5028
9ca92389 5029 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 5030
bb044576
TT		 5031 if (journal) {
5032 /* Now we set up the journal barrier. */
5033 jbd2_journal_lock_updates(journal);
ac27a0ec 5034
bb044576
TT		 5035 /*
5036 * Don't clear the needs_recovery flag if we failed to
5037 * flush the journal.
5038 */
5039 error = jbd2_journal_flush(journal);
5040 if (error < 0)
5041 goto out;
c642dc9e
ES		 5042
5043 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 5044 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 5045 }
9ca92389 5046
9ca92389 5047 error = ext4_commit_super(sb, 1);
6b0310fb 5048out:
bb044576
TT		 5049 if (journal)
5050 /* we rely on upper layer to stop further updates */
5051 jbd2_journal_unlock_updates(journal);
6b0310fb 5052 return error;
ac27a0ec
DK		 5053}
5054
5055/*
5056 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5057 * flag here, even though the filesystem is not technically dirty yet.
5058 */
c4be0c1d 5059static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 5060{
bc98a42c 5061 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 5062 return 0;
5063
c642dc9e
ES		 5064 if (EXT4_SB(sb)->s_journal) {
5065 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 5066 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 5067 }
5068
9ca92389 5069 ext4_commit_super(sb, 1);
c4be0c1d 5070 return 0;
ac27a0ec
DK		 5071}
5072
673c6100
TT		 5073/*
5074 * Structure to save mount options for ext4_remount's benefit
5075 */
5076struct ext4_mount_options {
5077 unsigned long s_mount_opt;
a2595b8a 5078 unsigned long s_mount_opt2;
08cefc7a
EB		 5079 kuid_t s_resuid;
5080 kgid_t s_resgid;
673c6100
TT		 5081 unsigned long s_commit_interval;
5082 u32 s_min_batch_time, s_max_batch_time;
5083#ifdef CONFIG_QUOTA
5084 int s_jquota_fmt;
a2d4a646 5085 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 5086#endif
5087};
5088
2b2d6d01 5089static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 5090{
2b2d6d01 5091 struct ext4_super_block *es;
617ba13b 5092 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5093 unsigned long old_sb_flags;
617ba13b 5094 struct ext4_mount_options old_opts;
c79d967d 5095 int enable_quota = 0;
8a266467 5096 ext4_group_t g;
b3881f74 5097 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 5098 int err = 0;
ac27a0ec 5099#ifdef CONFIG_QUOTA
03dafb5f 5100 int i, j;
ac27a0ec 5101#endif
d4c402d9 5102 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec 5103
21ac738e
CX		 5104 if (data && !orig_data)
5105 return -ENOMEM;
5106
ac27a0ec
DK		 5107 /* Store the original options */
5108 old_sb_flags = sb->s_flags;
5109 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 5110 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 5111 old_opts.s_resuid = sbi->s_resuid;
5112 old_opts.s_resgid = sbi->s_resgid;
5113 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 5114 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5115 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 5116#ifdef CONFIG_QUOTA
5117 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 5118 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 5119 if (sbi->s_qf_names[i]) {
5120 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
5121 GFP_KERNEL);
5122 if (!old_opts.s_qf_names[i]) {
5123 for (j = 0; j < i; j++)
5124 kfree(old_opts.s_qf_names[j]);
3e36a163 5125 kfree(orig_data);
03dafb5f
CG		 5126 return -ENOMEM;
5127 }
5128 } else
5129 old_opts.s_qf_names[i] = NULL;
ac27a0ec 5130#endif
b3881f74
TT		 5131 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5132 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 5133
661aa520 5134 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 5135 err = -EINVAL;
5136 goto restore_opts;
5137 }
5138
6b992ff2 5139 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 5140 test_opt(sb, JOURNAL_CHECKSUM)) {
5141 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 5142 "during remount not supported; ignoring");
5143 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 5144 }
5145
6ae6514b
PS		 5146 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5147 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5148 ext4_msg(sb, KERN_ERR, "can't mount with "
5149 "both data=journal and delalloc");
5150 err = -EINVAL;
5151 goto restore_opts;
5152 }
5153 if (test_opt(sb, DIOREAD_NOLOCK)) {
5154 ext4_msg(sb, KERN_ERR, "can't mount with "
5155 "both data=journal and dioread_nolock");
5156 err = -EINVAL;
5157 goto restore_opts;
5158 }
923ae0ff
RZ		 5159 if (test_opt(sb, DAX)) {
5160 ext4_msg(sb, KERN_ERR, "can't mount with "
5161 "both data=journal and dax");
5162 err = -EINVAL;
5163 goto restore_opts;
5164 }
ab04df78
JK		 5165 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5166 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5167 ext4_msg(sb, KERN_ERR, "can't mount with "
5168 "journal_async_commit in data=ordered mode");
5169 err = -EINVAL;
5170 goto restore_opts;
5171 }
923ae0ff
RZ		 5172 }
5173
cdb7ee4c
TE		 5174 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5175 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5176 err = -EINVAL;
5177 goto restore_opts;
5178 }
5179
923ae0ff
RZ		 5180 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5181 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5182 "dax flag with busy inodes while remounting");
5183 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 5184 }
5185
4ab2f15b 5186 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 5187 ext4_abort(sb, "Abort forced by user");
ac27a0ec 5188
1751e8a6
LT		 5189 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5190 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec
DK		 5191
5192 es = sbi->s_es;
5193
b3881f74 5194 if (sbi->s_journal) {
0390131b 5195 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 5196 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5197 }
ac27a0ec 5198
1751e8a6
LT		 5199 if (*flags & SB_LAZYTIME)
5200 sb->s_flags |= SB_LAZYTIME;
a2fd66d0 5201
1751e8a6 5202 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
4ab2f15b 5203 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 5204 err = -EROFS;
5205 goto restore_opts;
5206 }
5207
1751e8a6 5208 if (*flags & SB_RDONLY) {
38c03b34
TT		 5209 err = sync_filesystem(sb);
5210 if (err < 0)
5211 goto restore_opts;
0f0dd62f
CH		 5212 err = dquot_suspend(sb, -1);
5213 if (err < 0)
c79d967d 5214 goto restore_opts;
c79d967d 5215
ac27a0ec
DK		 5216 /*
5217 * First of all, the unconditional stuff we have to do
5218 * to disable replay of the journal when we next remount
5219 */
1751e8a6 5220 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 5221
5222 /*
5223 * OK, test if we are remounting a valid rw partition
5224 * readonly, and if so set the rdonly flag and then
5225 * mark the partition as valid again.
5226 */
617ba13b
MC		 5227 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5228 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 5229 es->s_state = cpu_to_le16(sbi->s_mount_state);
5230
a63c9eb2 5231 if (sbi->s_journal)
0390131b 5232 ext4_mark_recovery_complete(sb, es);
2dca60d9
TT		 5233 if (sbi->s_mmp_tsk)
5234 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 5235 } else {
a13fb1a4 5236 /* Make sure we can mount this feature set readwrite */
e2b911c5 5237 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 5238 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 5239 err = -EROFS;
5240 goto restore_opts;
5241 }
8a266467
TT		 5242 /*
5243 * Make sure the group descriptor checksums
0b8e58a1 5244 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 5245 */
5246 for (g = 0; g < sbi->s_groups_count; g++) {
5247 struct ext4_group_desc *gdp =
5248 ext4_get_group_desc(sb, g, NULL);
5249
feb0ab32 5250 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5251 ext4_msg(sb, KERN_ERR,
5252 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 5253 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 5254 le16_to_cpu(gdp->bg_checksum));
6a797d27 5255 err = -EFSBADCRC;
8a266467
TT		 5256 goto restore_opts;
5257 }
5258 }
5259
ead6596b
ES		 5260 /*
5261 * If we have an unprocessed orphan list hanging
5262 * around from a previously readonly bdev mount,
5263 * require a full umount/remount for now.
5264 */
5265 if (es->s_last_orphan) {
b31e1552 5266 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5267 "remount RDWR because of unprocessed "
5268 "orphan inode list. Please "
b31e1552 5269 "umount/remount instead");
ead6596b
ES		 5270 err = -EINVAL;
5271 goto restore_opts;
5272 }
5273
ac27a0ec
DK		 5274 /*
5275 * Mounting a RDONLY partition read-write, so reread
5276 * and store the current valid flag. (It may have
5277 * been changed by e2fsck since we originally mounted
5278 * the partition.)
5279 */
0390131b
FM		 5280 if (sbi->s_journal)
5281 ext4_clear_journal_err(sb, es);
ac27a0ec 5282 sbi->s_mount_state = le16_to_cpu(es->s_state);
c89128a0
JK		 5283
5284 err = ext4_setup_super(sb, es, 0);
5285 if (err)
5286 goto restore_opts;
5287
5288 sb->s_flags &= ~SB_RDONLY;
e2b911c5 5289 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 5290 if (ext4_multi_mount_protect(sb,
5291 le64_to_cpu(es->s_mmp_block))) {
5292 err = -EROFS;
5293 goto restore_opts;
5294 }
c79d967d 5295 enable_quota = 1;
ac27a0ec
DK		 5296 }
5297 }
bfff6873
LC		 5298
5299 /*
5300 * Reinitialize lazy itable initialization thread based on
5301 * current settings
5302 */
bc98a42c 5303 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
bfff6873
LC		 5304 ext4_unregister_li_request(sb);
5305 else {
5306 ext4_group_t first_not_zeroed;
5307 first_not_zeroed = ext4_has_uninit_itable(sb);
5308 ext4_register_li_request(sb, first_not_zeroed);
5309 }
5310
6fd058f7 5311 ext4_setup_system_zone(sb);
c89128a0
JK		 5312 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5313 err = ext4_commit_super(sb, 1);
5314 if (err)
5315 goto restore_opts;
5316 }
0390131b 5317
ac27a0ec
DK		 5318#ifdef CONFIG_QUOTA
5319 /* Release old quota file names */
a2d4a646 5320 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5321 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5322 if (enable_quota) {
5323 if (sb_any_quota_suspended(sb))
5324 dquot_resume(sb, -1);
e2b911c5 5325 else if (ext4_has_feature_quota(sb)) {
7c319d32 5326 err = ext4_enable_quotas(sb);
07724f98 5327 if (err)
7c319d32 5328 goto restore_opts;
7c319d32
AK		 5329 }
5330 }
ac27a0ec 5331#endif
d4c402d9 5332
1751e8a6 5333 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
d4c402d9
CW		 5334 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5335 kfree(orig_data);
ac27a0ec 5336 return 0;
0b8e58a1 5337
ac27a0ec
DK		 5338restore_opts:
5339 sb->s_flags = old_sb_flags;
5340 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5341 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5342 sbi->s_resuid = old_opts.s_resuid;
5343 sbi->s_resgid = old_opts.s_resgid;
5344 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5345 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5346 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5347#ifdef CONFIG_QUOTA
5348 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5349 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5350 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5351 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5352 }
5353#endif
d4c402d9 5354 kfree(orig_data);
ac27a0ec
DK		 5355 return err;
5356}
5357
689c958c
LX		 5358#ifdef CONFIG_QUOTA
5359static int ext4_statfs_project(struct super_block *sb,
5360 kprojid_t projid, struct kstatfs *buf)
5361{
5362 struct kqid qid;
5363 struct dquot *dquot;
5364 u64 limit;
5365 u64 curblock;
5366
5367 qid = make_kqid_projid(projid);
5368 dquot = dqget(sb, qid);
5369 if (IS_ERR(dquot))
5370 return PTR_ERR(dquot);
7b9ca4c6 5371 spin_lock(&dquot->dq_dqb_lock);
689c958c
LX		 5372
5373 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5374 dquot->dq_dqb.dqb_bsoftlimit :
5375 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5376 if (limit && buf->f_blocks > limit) {
f06925c7
KK		 5377 curblock = (dquot->dq_dqb.dqb_curspace +
5378 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
689c958c
LX		 5379 buf->f_blocks = limit;
5380 buf->f_bfree = buf->f_bavail =
5381 (buf->f_blocks > curblock) ?
5382 (buf->f_blocks - curblock) : 0;
5383 }
5384
5385 limit = dquot->dq_dqb.dqb_isoftlimit ?
5386 dquot->dq_dqb.dqb_isoftlimit :
5387 dquot->dq_dqb.dqb_ihardlimit;
5388 if (limit && buf->f_files > limit) {
5389 buf->f_files = limit;
5390 buf->f_ffree =
5391 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5392 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5393 }
5394
7b9ca4c6 5395 spin_unlock(&dquot->dq_dqb_lock);
689c958c
LX		 5396 dqput(dquot);
5397 return 0;
5398}
5399#endif
5400
2b2d6d01 5401static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5402{
5403 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5404 struct ext4_sb_info *sbi = EXT4_SB(sb);
5405 struct ext4_super_block *es = sbi->s_es;
27dd4385 5406 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5407 u64 fsid;
d02a9391 5408 s64 bfree;
27dd4385 5409 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5410
952fc18e
TT		 5411 if (!test_opt(sb, MINIX_DF))
5412 overhead = sbi->s_overhead;
ac27a0ec 5413
617ba13b 5414 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5415 buf->f_bsize = sb->s_blocksize;
b72f78cb 5416 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5417 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5418 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5419 /* prevent underflow in case that few free space is available */
57042651 5420 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5421 buf->f_bavail = buf->f_bfree -
5422 (ext4_r_blocks_count(es) + resv_blocks);
5423 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5424 buf->f_bavail = 0;
5425 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5426 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5427 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5428 fsid = le64_to_cpup((void *)es->s_uuid) ^
5429 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5430 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5431 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5432
689c958c
LX		 5433#ifdef CONFIG_QUOTA
5434 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5435 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5436 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5437#endif
ac27a0ec
DK		 5438 return 0;
5439}
5440
ac27a0ec
DK		 5441
5442#ifdef CONFIG_QUOTA
5443
bc8230ee
JK		 5444/*
5445 * Helper functions so that transaction is started before we acquire dqio_sem
5446 * to keep correct lock ordering of transaction > dqio_sem
5447 */
ac27a0ec
DK		 5448static inline struct inode *dquot_to_inode(struct dquot *dquot)
5449{
4c376dca 5450 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5451}
5452
617ba13b 5453static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5454{
5455 int ret, err;
5456 handle_t *handle;
5457 struct inode *inode;
5458
5459 inode = dquot_to_inode(dquot);
9924a92a 5460 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5461 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5462 if (IS_ERR(handle))
5463 return PTR_ERR(handle);
5464 ret = dquot_commit(dquot);
617ba13b 5465 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5466 if (!ret)
5467 ret = err;
5468 return ret;
5469}
5470
617ba13b 5471static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5472{
5473 int ret, err;
5474 handle_t *handle;
5475
9924a92a 5476 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5477 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5478 if (IS_ERR(handle))
5479 return PTR_ERR(handle);
5480 ret = dquot_acquire(dquot);
617ba13b 5481 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5482 if (!ret)
5483 ret = err;
5484 return ret;
5485}
5486
617ba13b 5487static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5488{
5489 int ret, err;
5490 handle_t *handle;
5491
9924a92a 5492 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5493 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5494 if (IS_ERR(handle)) {
5495 /* Release dquot anyway to avoid endless cycle in dqput() */
5496 dquot_release(dquot);
ac27a0ec 5497 return PTR_ERR(handle);
9c3013e9 5498 }
ac27a0ec 5499 ret = dquot_release(dquot);
617ba13b 5500 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5501 if (!ret)
5502 ret = err;
5503 return ret;
5504}
5505
617ba13b 5506static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5507{
262b4662
JK		 5508 struct super_block *sb = dquot->dq_sb;
5509 struct ext4_sb_info *sbi = EXT4_SB(sb);
5510
2c8be6b2 5511 /* Are we journaling quotas? */
e2b911c5 5512 if (ext4_has_feature_quota(sb) ||
262b4662 5513 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5514 dquot_mark_dquot_dirty(dquot);
617ba13b 5515 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5516 } else {
5517 return dquot_mark_dquot_dirty(dquot);
5518 }
5519}
5520
617ba13b 5521static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5522{
5523 int ret, err;
5524 handle_t *handle;
5525
5526 /* Data block + inode block */
2b0143b5 5527 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5528 if (IS_ERR(handle))
5529 return PTR_ERR(handle);
5530 ret = dquot_commit_info(sb, type);
617ba13b 5531 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5532 if (!ret)
5533 ret = err;
5534 return ret;
5535}
5536
5537/*
5538 * Turn on quotas during mount time - we need to find
5539 * the quota file and such...
5540 */
617ba13b 5541static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5542{
287a8095
CH		 5543 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5544 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5545}
5546
daf647d2
TT		 5547static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5548{
5549 struct ext4_inode_info *ei = EXT4_I(inode);
5550
5551 /* The first argument of lockdep_set_subclass has to be
5552 * *exactly* the same as the argument to init_rwsem() --- in
5553 * this case, in init_once() --- or lockdep gets unhappy
5554 * because the name of the lock is set using the
5555 * stringification of the argument to init_rwsem().
5556 */
5557 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5558 lockdep_set_subclass(&ei->i_data_sem, subclass);
5559}
5560
ac27a0ec
DK		 5561/*
5562 * Standard function to be called on quota_on
5563 */
617ba13b 5564static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 5565 const struct path *path)
ac27a0ec
DK		 5566{
5567 int err;
ac27a0ec
DK		 5568
5569 if (!test_opt(sb, QUOTA))
5570 return -EINVAL;
0623543b 5571
ac27a0ec 5572 /* Quotafile not on the same filesystem? */
d8c9584e 5573 if (path->dentry->d_sb != sb)
ac27a0ec 5574 return -EXDEV;
0623543b
JK		 5575 /* Journaling quota? */
5576 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5577 /* Quotafile not in fs root? */
f00c9e44 5578 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5579 ext4_msg(sb, KERN_WARNING,
5580 "Quota file not on filesystem root. "
5581 "Journaled quota will not work");
91389240
JK		 5582 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5583 } else {
5584 /*
5585 * Clear the flag just in case mount options changed since
5586 * last time.
5587 */
5588 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
2b2d6d01 5589 }
0623543b
JK		 5590
5591 /*
5592 * When we journal data on quota file, we have to flush journal to see
5593 * all updates to the file when we bypass pagecache...
5594 */
0390131b 5595 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5596 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5597 /*
5598 * We don't need to lock updates but journal_flush() could
5599 * otherwise be livelocked...
5600 */
5601 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5602 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5603 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5604 if (err)
7ffe1ea8 5605 return err;
0623543b 5606 }
957153fc 5607
daf647d2
TT		 5608 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5609 err = dquot_quota_on(sb, type, format_id, path);
957153fc 5610 if (err) {
daf647d2
TT		 5611 lockdep_set_quota_inode(path->dentry->d_inode,
5612 I_DATA_SEM_NORMAL);
957153fc
JK		 5613 } else {
5614 struct inode *inode = d_inode(path->dentry);
5615 handle_t *handle;
5616
61a92987
JK		 5617 /*
5618 * Set inode flags to prevent userspace from messing with quota
5619 * files. If this fails, we return success anyway since quotas
5620 * are already enabled and this is not a hard failure.
5621 */
957153fc
JK		 5622 inode_lock(inode);
5623 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5624 if (IS_ERR(handle))
5625 goto unlock_inode;
5626 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5627 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5628 S_NOATIME | S_IMMUTABLE);
5629 ext4_mark_inode_dirty(handle, inode);
5630 ext4_journal_stop(handle);
5631 unlock_inode:
5632 inode_unlock(inode);
5633 }
daf647d2 5634 return err;
ac27a0ec
DK		 5635}
5636
7c319d32
AK		 5637static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5638 unsigned int flags)
5639{
5640 int err;
5641 struct inode *qf_inode;
a2d4a646 5642 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5643 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5644 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5645 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5646 };
5647
e2b911c5 5648 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5649
5650 if (!qf_inums[type])
5651 return -EPERM;
5652
5653 qf_inode = ext4_iget(sb, qf_inums[type]);
5654 if (IS_ERR(qf_inode)) {
5655 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5656 return PTR_ERR(qf_inode);
5657 }
5658
bcb13850
JK		 5659 /* Don't account quota for quota files to avoid recursion */
5660 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5661 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7c319d32
AK		 5662 err = dquot_enable(qf_inode, type, format_id, flags);
5663 iput(qf_inode);
daf647d2
TT		 5664 if (err)
5665 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7c319d32
AK		 5666
5667 return err;
5668}
5669
5670/* Enable usage tracking for all quota types. */
5671static int ext4_enable_quotas(struct super_block *sb)
5672{
5673 int type, err = 0;
a2d4a646 5674 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5675 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5676 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5677 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 5678 };
49da9392
JK		 5679 bool quota_mopt[EXT4_MAXQUOTAS] = {
5680 test_opt(sb, USRQUOTA),
5681 test_opt(sb, GRPQUOTA),
5682 test_opt(sb, PRJQUOTA),
5683 };
7c319d32 5684
91389240 5685 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
a2d4a646 5686 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5687 if (qf_inums[type]) {
5688 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 5689 DQUOT_USAGE_ENABLED |
5690 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 5691 if (err) {
5692 ext4_warning(sb,
72ba7450
TT		 5693 "Failed to enable quota tracking "
5694 "(type=%d, err=%d). Please run "
5695 "e2fsck to fix.", type, err);
7f144fd0
JU		 5696 for (type--; type >= 0; type--)
5697 dquot_quota_off(sb, type);
5698
7c319d32
AK		 5699 return err;
5700 }
5701 }
5702 }
5703 return 0;
5704}
5705
ca0e05e4
DM		 5706static int ext4_quota_off(struct super_block *sb, int type)
5707{
21f97697
JK		 5708 struct inode *inode = sb_dqopt(sb)->files[type];
5709 handle_t *handle;
957153fc 5710 int err;
21f97697 5711
87009d86
DM		 5712 /* Force all delayed allocation blocks to be allocated.
5713 * Caller already holds s_umount sem */
5714 if (test_opt(sb, DELALLOC))
ca0e05e4 5715 sync_filesystem(sb);
ca0e05e4 5716
957153fc 5717 if (!inode || !igrab(inode))
0b268590
AG		 5718 goto out;
5719
957153fc 5720 err = dquot_quota_off(sb, type);
964edf66 5721 if (err || ext4_has_feature_quota(sb))
957153fc
JK		 5722 goto out_put;
5723
5724 inode_lock(inode);
61a92987
JK		 5725 /*
5726 * Update modification times of quota files when userspace can
5727 * start looking at them. If we fail, we return success anyway since
5728 * this is not a hard failure and quotas are already disabled.
5729 */
9924a92a 5730 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697 5731 if (IS_ERR(handle))
957153fc
JK		 5732 goto out_unlock;
5733 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5734 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
eeca7ea1 5735 inode->i_mtime = inode->i_ctime = current_time(inode);
21f97697
JK		 5736 ext4_mark_inode_dirty(handle, inode);
5737 ext4_journal_stop(handle);
957153fc
JK		 5738out_unlock:
5739 inode_unlock(inode);
5740out_put:
964edf66 5741 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 5742 iput(inode);
5743 return err;
21f97697 5744out:
ca0e05e4
DM		 5745 return dquot_quota_off(sb, type);
5746}
5747
ac27a0ec
DK		 5748/* Read data from quotafile - avoid pagecache and such because we cannot afford
5749 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5750 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5751 * we don't have to be afraid of races */
617ba13b 5752static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5753 size_t len, loff_t off)
5754{
5755 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5756 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5757 int offset = off & (sb->s_blocksize - 1);
5758 int tocopy;
5759 size_t toread;
5760 struct buffer_head *bh;
5761 loff_t i_size = i_size_read(inode);
5762
5763 if (off > i_size)
5764 return 0;
5765 if (off+len > i_size)
5766 len = i_size-off;
5767 toread = len;
5768 while (toread > 0) {
5769 tocopy = sb->s_blocksize - offset < toread ?
5770 sb->s_blocksize - offset : toread;
1c215028
TT		 5771 bh = ext4_bread(NULL, inode, blk, 0);
5772 if (IS_ERR(bh))
5773 return PTR_ERR(bh);
ac27a0ec
DK		 5774 if (!bh) /* A hole? */
5775 memset(data, 0, tocopy);
5776 else
5777 memcpy(data, bh->b_data+offset, tocopy);
5778 brelse(bh);
5779 offset = 0;
5780 toread -= tocopy;
5781 data += tocopy;
5782 blk++;
5783 }
5784 return len;
5785}
5786
5787/* Write to quotafile (we know the transaction is already started and has
5788 * enough credits) */
617ba13b 5789static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5790 const char *data, size_t len, loff_t off)
5791{
5792 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5793 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5794 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5795 int retries = 0;
ac27a0ec
DK		 5796 struct buffer_head *bh;
5797 handle_t *handle = journal_current_handle();
5798
0390131b 5799 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5800 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5801 " cancelled because transaction is not started",
9c3013e9
JK		 5802 (unsigned long long)off, (unsigned long long)len);
5803 return -EIO;
5804 }
67eeb568
DM		 5805 /*
5806 * Since we account only one data block in transaction credits,
5807 * then it is impossible to cross a block boundary.
5808 */
5809 if (sb->s_blocksize - offset < len) {
5810 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5811 " cancelled because not block aligned",
5812 (unsigned long long)off, (unsigned long long)len);
5813 return -EIO;
5814 }
5815
c5e298ae
TT		 5816 do {
5817 bh = ext4_bread(handle, inode, blk,
5818 EXT4_GET_BLOCKS_CREATE |
5819 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5820 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5821 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5822 if (IS_ERR(bh))
5823 return PTR_ERR(bh);
67eeb568
DM		 5824 if (!bh)
5825 goto out;
5d601255 5826 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5827 err = ext4_journal_get_write_access(handle, bh);
5828 if (err) {
5829 brelse(bh);
1c215028 5830 return err;
ac27a0ec 5831 }
67eeb568
DM		 5832 lock_buffer(bh);
5833 memcpy(bh->b_data+offset, data, len);
5834 flush_dcache_page(bh->b_page);
5835 unlock_buffer(bh);
62d2b5f2 5836 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5837 brelse(bh);
ac27a0ec 5838out:
67eeb568
DM		 5839 if (inode->i_size < off + len) {
5840 i_size_write(inode, off + len);
617ba13b 5841 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5842 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5843 }
67eeb568 5844 return len;
ac27a0ec
DK		 5845}
5846
8f0e8746
TT		 5847static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5848{
5849 const struct quota_format_ops *ops;
5850
5851 if (!sb_has_quota_loaded(sb, qid->type))
5852 return -ESRCH;
5853 ops = sb_dqopt(sb)->ops[qid->type];
5854 if (!ops || !ops->get_next_id)
5855 return -ENOSYS;
5856 return dquot_get_next_id(sb, qid);
5857}
ac27a0ec
DK		 5858#endif
5859
152a0836
AV		 5860static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5861 const char *dev_name, void *data)
ac27a0ec 5862{
152a0836 5863 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5864}
5865
c290ea01 5866#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5867static inline void register_as_ext2(void)
5868{
5869 int err = register_filesystem(&ext2_fs_type);
5870 if (err)
5871 printk(KERN_WARNING
5872 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5873}
5874
5875static inline void unregister_as_ext2(void)
5876{
5877 unregister_filesystem(&ext2_fs_type);
5878}
2035e776
TT		 5879
5880static inline int ext2_feature_set_ok(struct super_block *sb)
5881{
e2b911c5 5882 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776 5883 return 0;
bc98a42c 5884 if (sb_rdonly(sb))
2035e776 5885 return 1;
e2b911c5 5886 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5887 return 0;
5888 return 1;
5889}
24b58424
TT		 5890#else
5891static inline void register_as_ext2(void) { }
5892static inline void unregister_as_ext2(void) { }
2035e776 5893static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5894#endif
5895
24b58424
TT		 5896static inline void register_as_ext3(void)
5897{
5898 int err = register_filesystem(&ext3_fs_type);
5899 if (err)
5900 printk(KERN_WARNING
5901 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5902}
5903
5904static inline void unregister_as_ext3(void)
5905{
5906 unregister_filesystem(&ext3_fs_type);
5907}
2035e776
TT		 5908
5909static inline int ext3_feature_set_ok(struct super_block *sb)
5910{
e2b911c5 5911 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5912 return 0;
e2b911c5 5913 if (!ext4_has_feature_journal(sb))
2035e776 5914 return 0;
bc98a42c 5915 if (sb_rdonly(sb))
2035e776 5916 return 1;
e2b911c5 5917 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5918 return 0;
5919 return 1;
5920}
24b58424 5921
03010a33
TT		 5922static struct file_system_type ext4_fs_type = {
5923 .owner = THIS_MODULE,
5924 .name = "ext4",
152a0836 5925 .mount = ext4_mount,
03010a33
TT		 5926 .kill_sb = kill_block_super,
5927 .fs_flags = FS_REQUIRES_DEV,
5928};
7f78e035 5929MODULE_ALIAS_FS("ext4");
03010a33 5930
e9e3bcec
ES		 5931/* Shared across all ext4 file systems */
5932wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 5933
5dabfc78 5934static int __init ext4_init_fs(void)
ac27a0ec 5935{
e9e3bcec 5936 int i, err;
c9de560d 5937
e294a537 5938 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5939 ext4_li_info = NULL;
5940 mutex_init(&ext4_li_mtx);
5941
9a4c8019 5942 /* Build-time check for flags consistency */
12e9b892 5943 ext4_check_flag_values();
e9e3bcec 5944
e142d052 5945 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 5946 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 5947
51865fda 5948 err = ext4_init_es();
6fd058f7
TT		 5949 if (err)
5950 return err;
51865fda
ZL		 5951
5952 err = ext4_init_pageio();
5953 if (err)
b5799018 5954 goto out5;
51865fda 5955
5dabfc78 5956 err = ext4_init_system_zone();
bd2d0210 5957 if (err)
b5799018 5958 goto out4;
857ac889 5959
b5799018 5960 err = ext4_init_sysfs();
dd68314c 5961 if (err)
b5799018 5962 goto out3;
857ac889 5963
5dabfc78 5964 err = ext4_init_mballoc();
c9de560d
AT		 5965 if (err)
5966 goto out2;
ac27a0ec
DK		 5967 err = init_inodecache();
5968 if (err)
5969 goto out1;
24b58424 5970 register_as_ext3();
2035e776 5971 register_as_ext2();
03010a33 5972 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5973 if (err)
5974 goto out;
bfff6873 5975
ac27a0ec
DK		 5976 return 0;
5977out:
24b58424
TT		 5978 unregister_as_ext2();
5979 unregister_as_ext3();
ac27a0ec
DK		 5980 destroy_inodecache();
5981out1:
5dabfc78 5982 ext4_exit_mballoc();
9c191f70 5983out2:
b5799018
TT		 5984 ext4_exit_sysfs();
5985out3:
5dabfc78 5986 ext4_exit_system_zone();
b5799018 5987out4:
5dabfc78 5988 ext4_exit_pageio();
b5799018 5989out5:
51865fda
ZL		 5990 ext4_exit_es();
5991
ac27a0ec
DK		 5992 return err;
5993}
5994
5dabfc78 5995static void __exit ext4_exit_fs(void)
ac27a0ec 5996{
bfff6873 5997 ext4_destroy_lazyinit_thread();
24b58424
TT		 5998 unregister_as_ext2();
5999 unregister_as_ext3();
03010a33 6000 unregister_filesystem(&ext4_fs_type);
ac27a0ec 6001 destroy_inodecache();
5dabfc78 6002 ext4_exit_mballoc();
b5799018 6003 ext4_exit_sysfs();
5dabfc78
TT		 6004 ext4_exit_system_zone();
6005 ext4_exit_pageio();
dd12ed14 6006 ext4_exit_es();
ac27a0ec
DK		 6007}
6008
6009MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 6010MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 6011MODULE_LICENSE("GPL");
7ef79ad5 6012MODULE_SOFTDEP("pre: crc32c");
5dabfc78
TT		 6013module_init(ext4_init_fs)
6014module_exit(ext4_exit_fs)
A mirror of Linus' kernel repository