]> Git Repo - linux.git/blame - fs/ext4/super.c
projects / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Enable ext4 support for per-file/directory dax operations
[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 ([email protected])
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 ([email protected]), 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>
c83ad55e 45#include <linux/unicode.h>
c6a564ff 46#include <linux/part_stat.h>
bfff6873
LC		 47#include <linux/kthread.h>
48#include <linux/freezer.h>
49
3dcf5451 50#include "ext4.h"
4a092d73 51#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 52#include "ext4_jbd2.h"
ac27a0ec
DK		 53#include "xattr.h"
54#include "acl.h"
3661d286 55#include "mballoc.h"
0c9ec4be 56#include "fsmap.h"
ac27a0ec 57
9bffad1e
TT		 58#define CREATE_TRACE_POINTS
59#include <trace/events/ext4.h>
60
0b75a840
LC		 61static struct ext4_lazy_init *ext4_li_info;
62static struct mutex ext4_li_mtx;
e294a537 63static struct ratelimit_state ext4_mount_msg_ratelimit;
9f6200bb 64
617ba13b 65static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 66 unsigned long journal_devnum);
2adf6da8 67static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 68static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 69static void ext4_mark_recovery_complete(struct super_block *sb,
70 struct ext4_super_block *es);
71static void ext4_clear_journal_err(struct super_block *sb,
72 struct ext4_super_block *es);
617ba13b 73static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 74static int ext4_remount(struct super_block *sb, int *flags, char *data);
75static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 76static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 77static int ext4_freeze(struct super_block *sb);
152a0836
AV		 78static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
2035e776
TT		 80static inline int ext2_feature_set_ok(struct super_block *sb);
81static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 82static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 83static void ext4_destroy_lazyinit_thread(void);
84static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 85static void ext4_clear_request_list(void);
c6cb7e77
EW		 86static struct inode *ext4_get_journal_inode(struct super_block *sb,
87 unsigned int journal_inum);
ac27a0ec 88
e74031fd
JK		 89/*
90 * Lock ordering
91 *
92 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
93 * i_mmap_rwsem (inode->i_mmap_rwsem)!
94 *
95 * page fault path:
96 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
97 * page lock -> i_data_sem (rw)
98 *
99 * buffered write path:
100 * sb_start_write -> i_mutex -> mmap_sem
101 * sb_start_write -> i_mutex -> transaction start -> page lock ->
102 * i_data_sem (rw)
103 *
104 * truncate:
1d39834f
NB		 105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
106 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
107 * i_data_sem (rw)
e74031fd
JK		 108 *
109 * direct IO:
1d39834f
NB		 110 * sb_start_write -> i_mutex -> mmap_sem
111 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
e74031fd
JK		 112 *
113 * writepages:
114 * transaction start -> page lock(s) -> i_data_sem (rw)
115 */
116
c290ea01 117#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
2035e776
TT		 118static struct file_system_type ext2_fs_type = {
119 .owner = THIS_MODULE,
120 .name = "ext2",
121 .mount = ext4_mount,
122 .kill_sb = kill_block_super,
123 .fs_flags = FS_REQUIRES_DEV,
124};
7f78e035 125MODULE_ALIAS_FS("ext2");
fa7614dd 126MODULE_ALIAS("ext2");
2035e776
TT		 127#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
128#else
129#define IS_EXT2_SB(sb) (0)
130#endif
131
132
ba69f9ab
JK		 133static struct file_system_type ext3_fs_type = {
134 .owner = THIS_MODULE,
135 .name = "ext3",
152a0836 136 .mount = ext4_mount,
ba69f9ab
JK		 137 .kill_sb = kill_block_super,
138 .fs_flags = FS_REQUIRES_DEV,
139};
7f78e035 140MODULE_ALIAS_FS("ext3");
fa7614dd 141MODULE_ALIAS("ext3");
ba69f9ab 142#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
bd81d8ee 143
fb265c9c
TT		 144/*
145 * This works like sb_bread() except it uses ERR_PTR for error
146 * returns. Currently with sb_bread it's impossible to distinguish
147 * between ENOMEM and EIO situations (since both result in a NULL
148 * return.
149 */
150struct buffer_head *
151ext4_sb_bread(struct super_block *sb, sector_t block, int op_flags)
152{
153 struct buffer_head *bh = sb_getblk(sb, block);
154
155 if (bh == NULL)
156 return ERR_PTR(-ENOMEM);
cf2834a5 157 if (ext4_buffer_uptodate(bh))
fb265c9c
TT		 158 return bh;
159 ll_rw_block(REQ_OP_READ, REQ_META | op_flags, 1, &bh);
160 wait_on_buffer(bh);
161 if (buffer_uptodate(bh))
162 return bh;
163 put_bh(bh);
164 return ERR_PTR(-EIO);
165}
166
d25425f8
DW		 167static int ext4_verify_csum_type(struct super_block *sb,
168 struct ext4_super_block *es)
169{
e2b911c5 170 if (!ext4_has_feature_metadata_csum(sb))
d25425f8
DW		 171 return 1;
172
173 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
174}
175
a9c47317
DW		 176static __le32 ext4_superblock_csum(struct super_block *sb,
177 struct ext4_super_block *es)
178{
179 struct ext4_sb_info *sbi = EXT4_SB(sb);
180 int offset = offsetof(struct ext4_super_block, s_checksum);
181 __u32 csum;
182
183 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
184
185 return cpu_to_le32(csum);
186}
187
c197855e
SH		 188static int ext4_superblock_csum_verify(struct super_block *sb,
189 struct ext4_super_block *es)
a9c47317 190{
9aa5d32b 191 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 192 return 1;
193
194 return es->s_checksum == ext4_superblock_csum(sb, es);
195}
196
06db49e6 197void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 198{
06db49e6
TT		 199 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
200
9aa5d32b 201 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 202 return;
203
204 es->s_checksum = ext4_superblock_csum(sb, es);
205}
206
8fadc143
AR		 207ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
208 struct ext4_group_desc *bg)
bd81d8ee 209{
3a14589c 210 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 211 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 212 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 213}
214
8fadc143
AR		 215ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
216 struct ext4_group_desc *bg)
bd81d8ee 217{
5272f837 218 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 219 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 220 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 221}
222
8fadc143
AR		 223ext4_fsblk_t ext4_inode_table(struct super_block *sb,
224 struct ext4_group_desc *bg)
bd81d8ee 225{
5272f837 226 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 227 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 228 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 229}
230
021b65bb
TT		 231__u32 ext4_free_group_clusters(struct super_block *sb,
232 struct ext4_group_desc *bg)
560671a0
AK		 233{
234 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
235 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 236 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 237}
238
239__u32 ext4_free_inodes_count(struct super_block *sb,
240 struct ext4_group_desc *bg)
241{
242 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
243 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 244 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 245}
246
247__u32 ext4_used_dirs_count(struct super_block *sb,
248 struct ext4_group_desc *bg)
249{
250 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
251 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 252 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 253}
254
255__u32 ext4_itable_unused_count(struct super_block *sb,
256 struct ext4_group_desc *bg)
257{
258 return le16_to_cpu(bg->bg_itable_unused_lo) |
259 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 260 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 261}
262
8fadc143
AR		 263void ext4_block_bitmap_set(struct super_block *sb,
264 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 265{
3a14589c 266 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 267 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
268 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 269}
270
8fadc143
AR		 271void ext4_inode_bitmap_set(struct super_block *sb,
272 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 273{
5272f837 274 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 275 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
276 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 277}
278
8fadc143
AR		 279void ext4_inode_table_set(struct super_block *sb,
280 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 281{
5272f837 282 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 283 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
284 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 285}
286
021b65bb
TT		 287void ext4_free_group_clusters_set(struct super_block *sb,
288 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 289{
290 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
291 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
292 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
293}
294
295void ext4_free_inodes_set(struct super_block *sb,
296 struct ext4_group_desc *bg, __u32 count)
297{
298 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
299 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
300 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
301}
302
303void ext4_used_dirs_set(struct super_block *sb,
304 struct ext4_group_desc *bg, __u32 count)
305{
306 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
307 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
308 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
309}
310
311void ext4_itable_unused_set(struct super_block *sb,
312 struct ext4_group_desc *bg, __u32 count)
313{
314 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
315 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
316 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
317}
318
6a0678a7
AB		 319static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
320{
321 time64_t now = ktime_get_real_seconds();
322
323 now = clamp_val(now, 0, (1ull << 40) - 1);
324
325 *lo = cpu_to_le32(lower_32_bits(now));
326 *hi = upper_32_bits(now);
327}
328
329static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
330{
331 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
332}
333#define ext4_update_tstamp(es, tstamp) \
334 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
335#define ext4_get_tstamp(es, tstamp) \
336 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
d3d1faf6 337
54d3adbc
TT		 338static void __save_error_info(struct super_block *sb, int error,
339 __u32 ino, __u64 block,
340 const char *func, unsigned int line)
1c13d5c0
TT		 341{
342 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
54d3adbc 343 int err;
1c13d5c0
TT		 344
345 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1b46617b
TT		 346 if (bdev_read_only(sb->s_bdev))
347 return;
1c13d5c0 348 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6a0678a7 349 ext4_update_tstamp(es, s_last_error_time);
1c13d5c0
TT		 350 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
351 es->s_last_error_line = cpu_to_le32(line);
54d3adbc
TT		 352 es->s_last_error_ino = cpu_to_le32(ino);
353 es->s_last_error_block = cpu_to_le64(block);
354 switch (error) {
355 case EIO:
356 err = EXT4_ERR_EIO;
357 break;
358 case ENOMEM:
359 err = EXT4_ERR_ENOMEM;
360 break;
361 case EFSBADCRC:
362 err = EXT4_ERR_EFSBADCRC;
363 break;
364 case 0:
365 case EFSCORRUPTED:
366 err = EXT4_ERR_EFSCORRUPTED;
367 break;
368 case ENOSPC:
369 err = EXT4_ERR_ENOSPC;
370 break;
371 case ENOKEY:
372 err = EXT4_ERR_ENOKEY;
373 break;
374 case EROFS:
375 err = EXT4_ERR_EROFS;
376 break;
377 case EFBIG:
378 err = EXT4_ERR_EFBIG;
379 break;
380 case EEXIST:
381 err = EXT4_ERR_EEXIST;
382 break;
383 case ERANGE:
384 err = EXT4_ERR_ERANGE;
385 break;
386 case EOVERFLOW:
387 err = EXT4_ERR_EOVERFLOW;
388 break;
389 case EBUSY:
390 err = EXT4_ERR_EBUSY;
391 break;
392 case ENOTDIR:
393 err = EXT4_ERR_ENOTDIR;
394 break;
395 case ENOTEMPTY:
396 err = EXT4_ERR_ENOTEMPTY;
397 break;
398 case ESHUTDOWN:
399 err = EXT4_ERR_ESHUTDOWN;
400 break;
401 case EFAULT:
402 err = EXT4_ERR_EFAULT;
403 break;
404 default:
405 err = EXT4_ERR_UNKNOWN;
406 }
407 es->s_last_error_errcode = err;
1c13d5c0
TT		 408 if (!es->s_first_error_time) {
409 es->s_first_error_time = es->s_last_error_time;
6a0678a7 410 es->s_first_error_time_hi = es->s_last_error_time_hi;
1c13d5c0
TT		 411 strncpy(es->s_first_error_func, func,
412 sizeof(es->s_first_error_func));
413 es->s_first_error_line = cpu_to_le32(line);
414 es->s_first_error_ino = es->s_last_error_ino;
415 es->s_first_error_block = es->s_last_error_block;
878520ac 416 es->s_first_error_errcode = es->s_last_error_errcode;
1c13d5c0 417 }
66e61a9e
TT		 418 /*
419 * Start the daily error reporting function if it hasn't been
420 * started already
421 */
422 if (!es->s_error_count)
423 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 424 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 425}
426
54d3adbc
TT		 427static void save_error_info(struct super_block *sb, int error,
428 __u32 ino, __u64 block,
429 const char *func, unsigned int line)
1c13d5c0 430{
54d3adbc 431 __save_error_info(sb, error, ino, block, func, line);
c96e2b85
ES		 432 if (!bdev_read_only(sb->s_bdev))
433 ext4_commit_super(sb, 1);
1c13d5c0
TT		 434}
435
bdfe0cbd
TT		 436/*
437 * The del_gendisk() function uninitializes the disk-specific data
438 * structures, including the bdi structure, without telling anyone
439 * else. Once this happens, any attempt to call mark_buffer_dirty()
440 * (for example, by ext4_commit_super), will cause a kernel OOPS.
441 * This is a kludge to prevent these oops until we can put in a proper
442 * hook in del_gendisk() to inform the VFS and file system layers.
443 */
444static int block_device_ejected(struct super_block *sb)
445{
446 struct inode *bd_inode = sb->s_bdev->bd_inode;
447 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
448
449 return bdi->dev == NULL;
450}
451
18aadd47
BJ		 452static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
453{
454 struct super_block *sb = journal->j_private;
455 struct ext4_sb_info *sbi = EXT4_SB(sb);
456 int error = is_journal_aborted(journal);
5d3ee208 457 struct ext4_journal_cb_entry *jce;
18aadd47 458
5d3ee208 459 BUG_ON(txn->t_state == T_FINISHED);
a0154344
DJ		 460
461 ext4_process_freed_data(sb, txn->t_tid);
462
18aadd47 463 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 464 while (!list_empty(&txn->t_private_list)) {
465 jce = list_entry(txn->t_private_list.next,
466 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 467 list_del_init(&jce->jce_list);
468 spin_unlock(&sbi->s_md_lock);
469 jce->jce_func(sb, jce, error);
470 spin_lock(&sbi->s_md_lock);
471 }
472 spin_unlock(&sbi->s_md_lock);
473}
1c13d5c0 474
1dc1097f
JK		 475static bool system_going_down(void)
476{
477 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
478 || system_state == SYSTEM_RESTART;
479}
480
ac27a0ec
DK		 481/* Deal with the reporting of failure conditions on a filesystem such as
482 * inconsistencies detected or read IO failures.
483 *
484 * On ext2, we can store the error state of the filesystem in the
617ba13b 485 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 486 * write ordering constraints on the superblock which prevent us from
487 * writing it out straight away; and given that the journal is about to
488 * be aborted, we can't rely on the current, or future, transactions to
489 * write out the superblock safely.
490 *
dab291af 491 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 492 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 493 * that error until we've noted it down and cleared it.
494 */
495
617ba13b 496static void ext4_handle_error(struct super_block *sb)
ac27a0ec 497{
327eaf73
TT		 498 if (test_opt(sb, WARN_ON_ERROR))
499 WARN_ON_ONCE(1);
500
bc98a42c 501 if (sb_rdonly(sb))
ac27a0ec
DK		 502 return;
503
2b2d6d01 504 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 505 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 506
4ab2f15b 507 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 508 if (journal)
dab291af 509 jbd2_journal_abort(journal, -EIO);
ac27a0ec 510 }
1dc1097f
JK		 511 /*
512 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
513 * could panic during 'reboot -f' as the underlying device got already
514 * disabled.
515 */
516 if (test_opt(sb, ERRORS_RO) || system_going_down()) {
b31e1552 517 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 518 /*
519 * Make sure updated value of ->s_mount_flags will be visible
520 * before ->s_flags update
521 */
522 smp_wmb();
1751e8a6 523 sb->s_flags |= SB_RDONLY;
1dc1097f 524 } else if (test_opt(sb, ERRORS_PANIC)) {
4327ba52
DJ		 525 if (EXT4_SB(sb)->s_journal &&
526 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
527 return;
617ba13b 528 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec 529 sb->s_id);
4327ba52 530 }
ac27a0ec
DK		 531}
532
efbed4dc
TT		 533#define ext4_error_ratelimit(sb) \
534 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
535 "EXT4-fs error")
536
12062ddd 537void __ext4_error(struct super_block *sb, const char *function,
54d3adbc
TT		 538 unsigned int line, int error, __u64 block,
539 const char *fmt, ...)
ac27a0ec 540{
0ff2ea7d 541 struct va_format vaf;
ac27a0ec
DK		 542 va_list args;
543
0db1ff22
TT		 544 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
545 return;
546
ccf0f32a 547 trace_ext4_error(sb, function, line);
efbed4dc
TT		 548 if (ext4_error_ratelimit(sb)) {
549 va_start(args, fmt);
550 vaf.fmt = fmt;
551 vaf.va = &args;
552 printk(KERN_CRIT
553 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
554 sb->s_id, function, line, current->comm, &vaf);
555 va_end(args);
556 }
54d3adbc 557 save_error_info(sb, error, 0, block, function, line);
617ba13b 558 ext4_handle_error(sb);
ac27a0ec
DK		 559}
560
e7c96e8e 561void __ext4_error_inode(struct inode *inode, const char *function,
54d3adbc 562 unsigned int line, ext4_fsblk_t block, int error,
e7c96e8e 563 const char *fmt, ...)
273df556
FM		 564{
565 va_list args;
f7c21177 566 struct va_format vaf;
273df556 567
0db1ff22
TT		 568 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
569 return;
570
ccf0f32a 571 trace_ext4_error(inode->i_sb, function, line);
efbed4dc
TT		 572 if (ext4_error_ratelimit(inode->i_sb)) {
573 va_start(args, fmt);
574 vaf.fmt = fmt;
575 vaf.va = &args;
576 if (block)
577 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
578 "inode #%lu: block %llu: comm %s: %pV\n",
579 inode->i_sb->s_id, function, line, inode->i_ino,
580 block, current->comm, &vaf);
581 else
582 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
583 "inode #%lu: comm %s: %pV\n",
584 inode->i_sb->s_id, function, line, inode->i_ino,
585 current->comm, &vaf);
586 va_end(args);
587 }
54d3adbc
TT		 588 save_error_info(inode->i_sb, error, inode->i_ino, block,
589 function, line);
273df556
FM		 590 ext4_handle_error(inode->i_sb);
591}
592
e7c96e8e
JP		 593void __ext4_error_file(struct file *file, const char *function,
594 unsigned int line, ext4_fsblk_t block,
595 const char *fmt, ...)
273df556
FM		 596{
597 va_list args;
f7c21177 598 struct va_format vaf;
496ad9aa 599 struct inode *inode = file_inode(file);
273df556
FM		 600 char pathname[80], *path;
601
0db1ff22
TT		 602 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
603 return;
604
ccf0f32a 605 trace_ext4_error(inode->i_sb, function, line);
efbed4dc 606 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 607 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 608 if (IS_ERR(path))
609 path = "(unknown)";
610 va_start(args, fmt);
611 vaf.fmt = fmt;
612 vaf.va = &args;
613 if (block)
614 printk(KERN_CRIT
615 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
616 "block %llu: comm %s: path %s: %pV\n",
617 inode->i_sb->s_id, function, line, inode->i_ino,
618 block, current->comm, path, &vaf);
619 else
620 printk(KERN_CRIT
621 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
622 "comm %s: path %s: %pV\n",
623 inode->i_sb->s_id, function, line, inode->i_ino,
624 current->comm, path, &vaf);
625 va_end(args);
626 }
54d3adbc
TT		 627 save_error_info(inode->i_sb, EFSCORRUPTED, inode->i_ino, block,
628 function, line);
273df556
FM		 629 ext4_handle_error(inode->i_sb);
630}
631
722887dd
TT		 632const char *ext4_decode_error(struct super_block *sb, int errno,
633 char nbuf[16])
ac27a0ec
DK		 634{
635 char *errstr = NULL;
636
637 switch (errno) {
6a797d27
DW		 638 case -EFSCORRUPTED:
639 errstr = "Corrupt filesystem";
640 break;
641 case -EFSBADCRC:
642 errstr = "Filesystem failed CRC";
643 break;
ac27a0ec
DK		 644 case -EIO:
645 errstr = "IO failure";
646 break;
647 case -ENOMEM:
648 errstr = "Out of memory";
649 break;
650 case -EROFS:
78f1ddbb
TT		 651 if (!sb || (EXT4_SB(sb)->s_journal &&
652 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 653 errstr = "Journal has aborted";
654 else
655 errstr = "Readonly filesystem";
656 break;
657 default:
658 /* If the caller passed in an extra buffer for unknown
659 * errors, textualise them now. Else we just return
660 * NULL. */
661 if (nbuf) {
662 /* Check for truncated error codes... */
663 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
664 errstr = nbuf;
665 }
666 break;
667 }
668
669 return errstr;
670}
671
617ba13b 672/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 673 * automatically and invokes the appropriate error response. */
674
c398eda0
TT		 675void __ext4_std_error(struct super_block *sb, const char *function,
676 unsigned int line, int errno)
ac27a0ec
DK		 677{
678 char nbuf[16];
679 const char *errstr;
680
0db1ff22
TT		 681 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
682 return;
683
ac27a0ec
DK		 684 /* Special case: if the error is EROFS, and we're not already
685 * inside a transaction, then there's really no point in logging
686 * an error. */
bc98a42c 687 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
ac27a0ec
DK		 688 return;
689
efbed4dc
TT		 690 if (ext4_error_ratelimit(sb)) {
691 errstr = ext4_decode_error(sb, errno, nbuf);
692 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
693 sb->s_id, function, line, errstr);
694 }
ac27a0ec 695
54d3adbc 696 save_error_info(sb, -errno, 0, 0, function, line);
617ba13b 697 ext4_handle_error(sb);
ac27a0ec
DK		 698}
699
700/*
617ba13b 701 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 702 * abort function may be used to deal with unrecoverable failures such
703 * as journal IO errors or ENOMEM at a critical moment in log management.
704 *
705 * We unconditionally force the filesystem into an ABORT|READONLY state,
706 * unless the error response on the fs has been set to panic in which
707 * case we take the easy way out and panic immediately.
708 */
709
c67d859e 710void __ext4_abort(struct super_block *sb, const char *function,
54d3adbc 711 unsigned int line, int error, const char *fmt, ...)
ac27a0ec 712{
651e1c3b 713 struct va_format vaf;
ac27a0ec
DK		 714 va_list args;
715
0db1ff22
TT		 716 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
717 return;
718
54d3adbc 719 save_error_info(sb, error, 0, 0, function, line);
ac27a0ec 720 va_start(args, fmt);
651e1c3b
JP		 721 vaf.fmt = fmt;
722 vaf.va = &args;
723 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
724 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 725 va_end(args);
726
bc98a42c 727 if (sb_rdonly(sb) == 0) {
1c13d5c0 728 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 729 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 730 /*
731 * Make sure updated value of ->s_mount_flags will be visible
732 * before ->s_flags update
733 */
734 smp_wmb();
1751e8a6 735 sb->s_flags |= SB_RDONLY;
1c13d5c0
TT		 736 if (EXT4_SB(sb)->s_journal)
737 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
1c13d5c0 738 }
2c1d0e36 739 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
4327ba52
DJ		 740 if (EXT4_SB(sb)->s_journal &&
741 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
742 return;
617ba13b 743 panic("EXT4-fs panic from previous error\n");
4327ba52 744 }
ac27a0ec
DK		 745}
746
e7c96e8e
JP		 747void __ext4_msg(struct super_block *sb,
748 const char *prefix, const char *fmt, ...)
b31e1552 749{
0ff2ea7d 750 struct va_format vaf;
b31e1552
ES		 751 va_list args;
752
efbed4dc
TT		 753 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
754 return;
755
b31e1552 756 va_start(args, fmt);
0ff2ea7d
JP		 757 vaf.fmt = fmt;
758 vaf.va = &args;
759 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 760 va_end(args);
761}
762
b03a2f7e
AD		 763#define ext4_warning_ratelimit(sb) \
764 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
765 "EXT4-fs warning")
766
12062ddd 767void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 768 unsigned int line, const char *fmt, ...)
ac27a0ec 769{
0ff2ea7d 770 struct va_format vaf;
ac27a0ec
DK		 771 va_list args;
772
b03a2f7e 773 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 774 return;
775
ac27a0ec 776 va_start(args, fmt);
0ff2ea7d
JP		 777 vaf.fmt = fmt;
778 vaf.va = &args;
779 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
780 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 781 va_end(args);
782}
783
b03a2f7e
AD		 784void __ext4_warning_inode(const struct inode *inode, const char *function,
785 unsigned int line, const char *fmt, ...)
786{
787 struct va_format vaf;
788 va_list args;
789
790 if (!ext4_warning_ratelimit(inode->i_sb))
791 return;
792
793 va_start(args, fmt);
794 vaf.fmt = fmt;
795 vaf.va = &args;
796 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
797 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
798 function, line, inode->i_ino, current->comm, &vaf);
799 va_end(args);
800}
801
e29136f8
TT		 802void __ext4_grp_locked_error(const char *function, unsigned int line,
803 struct super_block *sb, ext4_group_t grp,
804 unsigned long ino, ext4_fsblk_t block,
805 const char *fmt, ...)
5d1b1b3f
AK		 806__releases(bitlock)
807__acquires(bitlock)
808{
0ff2ea7d 809 struct va_format vaf;
5d1b1b3f 810 va_list args;
5d1b1b3f 811
0db1ff22
TT		 812 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
813 return;
814
ccf0f32a 815 trace_ext4_error(sb, function, line);
54d3adbc 816 __save_error_info(sb, EFSCORRUPTED, ino, block, function, line);
0ff2ea7d 817
efbed4dc
TT		 818 if (ext4_error_ratelimit(sb)) {
819 va_start(args, fmt);
820 vaf.fmt = fmt;
821 vaf.va = &args;
822 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
823 sb->s_id, function, line, grp);
824 if (ino)
825 printk(KERN_CONT "inode %lu: ", ino);
826 if (block)
827 printk(KERN_CONT "block %llu:",
828 (unsigned long long) block);
829 printk(KERN_CONT "%pV\n", &vaf);
830 va_end(args);
831 }
5d1b1b3f 832
327eaf73
TT		 833 if (test_opt(sb, WARN_ON_ERROR))
834 WARN_ON_ONCE(1);
835
5d1b1b3f 836 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 837 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 838 return;
839 }
1c13d5c0 840
5d1b1b3f 841 ext4_unlock_group(sb, grp);
06f29cc8 842 ext4_commit_super(sb, 1);
5d1b1b3f
AK		 843 ext4_handle_error(sb);
844 /*
845 * We only get here in the ERRORS_RO case; relocking the group
846 * may be dangerous, but nothing bad will happen since the
847 * filesystem will have already been marked read/only and the
848 * journal has been aborted. We return 1 as a hint to callers
849 * who might what to use the return value from
25985edc 850 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 851 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
852 * aggressively from the ext4 function in question, with a
853 * more appropriate error code.
854 */
855 ext4_lock_group(sb, grp);
856 return;
857}
858
db79e6d1
WS		 859void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
860 ext4_group_t group,
861 unsigned int flags)
862{
863 struct ext4_sb_info *sbi = EXT4_SB(sb);
864 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
865 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
9af0b3d1
WS		 866 int ret;
867
868 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
869 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
870 &grp->bb_state);
871 if (!ret)
872 percpu_counter_sub(&sbi->s_freeclusters_counter,
873 grp->bb_free);
db79e6d1
WS		 874 }
875
9af0b3d1
WS		 876 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
877 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
878 &grp->bb_state);
879 if (!ret && gdp) {
db79e6d1
WS		 880 int count;
881
882 count = ext4_free_inodes_count(sb, gdp);
883 percpu_counter_sub(&sbi->s_freeinodes_counter,
884 count);
885 }
db79e6d1
WS		 886 }
887}
888
617ba13b 889void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 890{
617ba13b 891 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 892
617ba13b 893 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 894 return;
895
12062ddd 896 ext4_warning(sb,
ac27a0ec
DK		 897 "updating to rev %d because of new feature flag, "
898 "running e2fsck is recommended",
617ba13b 899 EXT4_DYNAMIC_REV);
ac27a0ec 900
617ba13b
MC		 901 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
902 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
903 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 904 /* leave es->s_feature_*compat flags alone */
905 /* es->s_uuid will be set by e2fsck if empty */
906
907 /*
908 * The rest of the superblock fields should be zero, and if not it
909 * means they are likely already in use, so leave them alone. We
910 * can leave it up to e2fsck to clean up any inconsistencies there.
911 */
912}
913
914/*
915 * Open the external journal device
916 */
b31e1552 917static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 918{
919 struct block_device *bdev;
ac27a0ec 920
d4d77629 921 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 922 if (IS_ERR(bdev))
923 goto fail;
924 return bdev;
925
926fail:
ea3edd4d
CH		 927 ext4_msg(sb, KERN_ERR,
928 "failed to open journal device unknown-block(%u,%u) %ld",
929 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
ac27a0ec
DK		 930 return NULL;
931}
932
933/*
934 * Release the journal device
935 */
4385bab1 936static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 937{
4385bab1 938 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 939}
940
4385bab1 941static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 942{
943 struct block_device *bdev;
ac27a0ec
DK		 944 bdev = sbi->journal_bdev;
945 if (bdev) {
4385bab1 946 ext4_blkdev_put(bdev);
ac27a0ec
DK		 947 sbi->journal_bdev = NULL;
948 }
ac27a0ec
DK		 949}
950
951static inline struct inode *orphan_list_entry(struct list_head *l)
952{
617ba13b 953 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 954}
955
617ba13b 956static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 957{
958 struct list_head *l;
959
b31e1552
ES		 960 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
961 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 962
963 printk(KERN_ERR "sb_info orphan list:\n");
964 list_for_each(l, &sbi->s_orphan) {
965 struct inode *inode = orphan_list_entry(l);
966 printk(KERN_ERR " "
967 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
968 inode->i_sb->s_id, inode->i_ino, inode,
969 inode->i_mode, inode->i_nlink,
970 NEXT_ORPHAN(inode));
971 }
972}
973
957153fc
JK		 974#ifdef CONFIG_QUOTA
975static int ext4_quota_off(struct super_block *sb, int type);
976
977static inline void ext4_quota_off_umount(struct super_block *sb)
978{
979 int type;
980
964edf66
JK		 981 /* Use our quota_off function to clear inode flags etc. */
982 for (type = 0; type < EXT4_MAXQUOTAS; type++)
983 ext4_quota_off(sb, type);
957153fc 984}
33458eab
TT		 985
986/*
987 * This is a helper function which is used in the mount/remount
988 * codepaths (which holds s_umount) to fetch the quota file name.
989 */
990static inline char *get_qf_name(struct super_block *sb,
991 struct ext4_sb_info *sbi,
992 int type)
993{
994 return rcu_dereference_protected(sbi->s_qf_names[type],
995 lockdep_is_held(&sb->s_umount));
996}
957153fc
JK		 997#else
998static inline void ext4_quota_off_umount(struct super_block *sb)
999{
1000}
1001#endif
1002
2b2d6d01 1003static void ext4_put_super(struct super_block *sb)
ac27a0ec 1004{
617ba13b
MC		 1005 struct ext4_sb_info *sbi = EXT4_SB(sb);
1006 struct ext4_super_block *es = sbi->s_es;
1d0c3924 1007 struct buffer_head **group_desc;
7c990728 1008 struct flex_groups **flex_groups;
97abd7d4 1009 int aborted = 0;
ef2cabf7 1010 int i, err;
ac27a0ec 1011
857ac889 1012 ext4_unregister_li_request(sb);
957153fc 1013 ext4_quota_off_umount(sb);
e0ccfd95 1014
2e8fa54e 1015 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 1016
5e47868f
RH		 1017 /*
1018 * Unregister sysfs before destroying jbd2 journal.
1019 * Since we could still access attr_journal_task attribute via sysfs
1020 * path which could have sbi->s_journal->j_task as NULL
1021 */
1022 ext4_unregister_sysfs(sb);
1023
0390131b 1024 if (sbi->s_journal) {
97abd7d4 1025 aborted = is_journal_aborted(sbi->s_journal);
0390131b
FM		 1026 err = jbd2_journal_destroy(sbi->s_journal);
1027 sbi->s_journal = NULL;
878520ac 1028 if ((err < 0) && !aborted) {
54d3adbc 1029 ext4_abort(sb, -err, "Couldn't clean up the journal");
878520ac 1030 }
0390131b 1031 }
d4edac31 1032
d3922a77 1033 ext4_es_unregister_shrinker(sbi);
9105bb14 1034 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 1035 ext4_release_system_zone(sb);
1036 ext4_mb_release(sb);
1037 ext4_ext_release(sb);
d4edac31 1038
bc98a42c 1039 if (!sb_rdonly(sb) && !aborted) {
e2b911c5 1040 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 1041 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 1042 }
bc98a42c 1043 if (!sb_rdonly(sb))
a8e25a83
AB		 1044 ext4_commit_super(sb, 1);
1045
1d0c3924
TT		 1046 rcu_read_lock();
1047 group_desc = rcu_dereference(sbi->s_group_desc);
ac27a0ec 1048 for (i = 0; i < sbi->s_gdb_count; i++)
1d0c3924
TT		 1049 brelse(group_desc[i]);
1050 kvfree(group_desc);
7c990728
SJS		 1051 flex_groups = rcu_dereference(sbi->s_flex_groups);
1052 if (flex_groups) {
1053 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1054 kvfree(flex_groups[i]);
1055 kvfree(flex_groups);
1056 }
1d0c3924 1057 rcu_read_unlock();
57042651 1058 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 1059 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1060 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 1061 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
bbd55937 1062 percpu_free_rwsem(&sbi->s_writepages_rwsem);
ac27a0ec 1063#ifdef CONFIG_QUOTA
a2d4a646 1064 for (i = 0; i < EXT4_MAXQUOTAS; i++)
33458eab 1065 kfree(get_qf_name(sb, sbi, i));
ac27a0ec
DK		 1066#endif
1067
1068 /* Debugging code just in case the in-memory inode orphan list
1069 * isn't empty. The on-disk one can be non-empty if we've
1070 * detected an error and taken the fs readonly, but the
1071 * in-memory list had better be clean by this point. */
1072 if (!list_empty(&sbi->s_orphan))
1073 dump_orphan_list(sb, sbi);
1074 J_ASSERT(list_empty(&sbi->s_orphan));
1075
89d96a6f 1076 sync_blockdev(sb->s_bdev);
f98393a6 1077 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 1078 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
1079 /*
1080 * Invalidate the journal device's buffers. We don't want them
1081 * floating about in memory - the physical journal device may
1082 * hotswapped, and it breaks the `ro-after' testing code.
1083 */
1084 sync_blockdev(sbi->journal_bdev);
f98393a6 1085 invalidate_bdev(sbi->journal_bdev);
617ba13b 1086 ext4_blkdev_remove(sbi);
ac27a0ec 1087 }
50c15df6
CX		 1088
1089 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1090 sbi->s_ea_inode_cache = NULL;
1091
1092 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1093 sbi->s_ea_block_cache = NULL;
1094
c5e06d10
JL		 1095 if (sbi->s_mmp_tsk)
1096 kthread_stop(sbi->s_mmp_tsk);
9060dd2c 1097 brelse(sbi->s_sbh);
ac27a0ec 1098 sb->s_fs_info = NULL;
3197ebdb
TT		 1099 /*
1100 * Now that we are completely done shutting down the
1101 * superblock, we need to actually destroy the kobject.
1102 */
3197ebdb
TT		 1103 kobject_put(&sbi->s_kobj);
1104 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 1105 if (sbi->s_chksum_driver)
1106 crypto_free_shash(sbi->s_chksum_driver);
705895b6 1107 kfree(sbi->s_blockgroup_lock);
5e405595 1108 fs_put_dax(sbi->s_daxdev);
c83ad55e
GKB		 1109#ifdef CONFIG_UNICODE
1110 utf8_unload(sbi->s_encoding);
1111#endif
ac27a0ec 1112 kfree(sbi);
ac27a0ec
DK		 1113}
1114
e18b890b 1115static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 1116
1117/*
1118 * Called inside transaction, so use GFP_NOFS
1119 */
617ba13b 1120static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 1121{
617ba13b 1122 struct ext4_inode_info *ei;
ac27a0ec 1123
e6b4f8da 1124 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 1125 if (!ei)
1126 return NULL;
0b8e58a1 1127
ee73f9a5 1128 inode_set_iversion(&ei->vfs_inode, 1);
202ee5df 1129 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 1130 INIT_LIST_HEAD(&ei->i_prealloc_list);
1131 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 1132 ext4_es_init_tree(&ei->i_es_tree);
1133 rwlock_init(&ei->i_es_lock);
edaa53ca 1134 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 1135 ei->i_es_all_nr = 0;
edaa53ca 1136 ei->i_es_shk_nr = 0;
dd475925 1137 ei->i_es_shrink_lblk = 0;
d2a17637 1138 ei->i_reserved_data_blocks = 0;
d2a17637 1139 spin_lock_init(&(ei->i_block_reservation_lock));
1dc0aa46 1140 ext4_init_pending_tree(&ei->i_pending_tree);
a9e7f447
DM		 1141#ifdef CONFIG_QUOTA
1142 ei->i_reserved_quota = 0;
96c7e0d9 1143 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 1144#endif
8aefcd55 1145 ei->jinode = NULL;
2e8fa54e 1146 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 1147 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 1148 ei->i_sync_tid = 0;
1149 ei->i_datasync_tid = 0;
e27f41e1 1150 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 1151 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
ac27a0ec
DK		 1152 return &ei->vfs_inode;
1153}
1154
7ff9c073
TT		 1155static int ext4_drop_inode(struct inode *inode)
1156{
1157 int drop = generic_drop_inode(inode);
1158
29b3692e
EB		 1159 if (!drop)
1160 drop = fscrypt_drop_inode(inode);
1161
7ff9c073
TT		 1162 trace_ext4_drop_inode(inode, drop);
1163 return drop;
1164}
1165
94053139 1166static void ext4_free_in_core_inode(struct inode *inode)
fa0d7e3d 1167{
2c58d548 1168 fscrypt_free_inode(inode);
fa0d7e3d
NP		 1169 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1170}
1171
617ba13b 1172static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1173{
9f7dd93d 1174 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1175 ext4_msg(inode->i_sb, KERN_ERR,
1176 "Inode %lu (%p): orphan list check failed!",
1177 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1178 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1179 EXT4_I(inode), sizeof(struct ext4_inode_info),
1180 true);
1181 dump_stack();
1182 }
ac27a0ec
DK		 1183}
1184
51cc5068 1185static void init_once(void *foo)
ac27a0ec 1186{
617ba13b 1187 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1188
a35afb83 1189 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1190 init_rwsem(&ei->xattr_sem);
0e855ac8 1191 init_rwsem(&ei->i_data_sem);
ea3d7209 1192 init_rwsem(&ei->i_mmap_sem);
a35afb83 1193 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1194}
1195
e67bc2b3 1196static int __init init_inodecache(void)
ac27a0ec 1197{
f8dd7c70
DW		 1198 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1199 sizeof(struct ext4_inode_info), 0,
1200 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1201 SLAB_ACCOUNT),
1202 offsetof(struct ext4_inode_info, i_data),
1203 sizeof_field(struct ext4_inode_info, i_data),
1204 init_once);
617ba13b 1205 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1206 return -ENOMEM;
1207 return 0;
1208}
1209
1210static void destroy_inodecache(void)
1211{
8c0a8537
KS		 1212 /*
1213 * Make sure all delayed rcu free inodes are flushed before we
1214 * destroy cache.
1215 */
1216 rcu_barrier();
617ba13b 1217 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1218}
1219
0930fcc1 1220void ext4_clear_inode(struct inode *inode)
ac27a0ec 1221{
0930fcc1 1222 invalidate_inode_buffers(inode);
dbd5768f 1223 clear_inode(inode);
c2ea3fde 1224 ext4_discard_preallocations(inode);
51865fda 1225 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
f4c2d372 1226 dquot_drop(inode);
8aefcd55
TT		 1227 if (EXT4_I(inode)->jinode) {
1228 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1229 EXT4_I(inode)->jinode);
1230 jbd2_free_inode(EXT4_I(inode)->jinode);
1231 EXT4_I(inode)->jinode = NULL;
1232 }
3d204e24 1233 fscrypt_put_encryption_info(inode);
c93d8f88 1234 fsverity_cleanup_inode(inode);
ac27a0ec
DK		 1235}
1236
1b961ac0 1237static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1238 u64 ino, u32 generation)
ac27a0ec 1239{
ac27a0ec 1240 struct inode *inode;
ac27a0ec 1241
8a363970 1242 /*
ac27a0ec
DK		 1243 * Currently we don't know the generation for parent directory, so
1244 * a generation of 0 means "accept any"
1245 */
8a363970 1246 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1d1fe1ee
DH		 1247 if (IS_ERR(inode))
1248 return ERR_CAST(inode);
1249 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1250 iput(inode);
1251 return ERR_PTR(-ESTALE);
1252 }
1b961ac0
CH		 1253
1254 return inode;
1255}
1256
1257static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1258 int fh_len, int fh_type)
1b961ac0
CH		 1259{
1260 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1261 ext4_nfs_get_inode);
1262}
1263
1264static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1265 int fh_len, int fh_type)
1b961ac0
CH		 1266{
1267 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1268 ext4_nfs_get_inode);
ac27a0ec
DK		 1269}
1270
fde87268
TT		 1271static int ext4_nfs_commit_metadata(struct inode *inode)
1272{
1273 struct writeback_control wbc = {
1274 .sync_mode = WB_SYNC_ALL
1275 };
1276
1277 trace_ext4_nfs_commit_metadata(inode);
1278 return ext4_write_inode(inode, &wbc);
1279}
1280
c39a7f84
TO		 1281/*
1282 * Try to release metadata pages (indirect blocks, directories) which are
1283 * mapped via the block device. Since these pages could have journal heads
1284 * which would prevent try_to_free_buffers() from freeing them, we must use
1285 * jbd2 layer's try_to_free_buffers() function to release them.
1286 */
0b8e58a1
AD		 1287static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1288 gfp_t wait)
c39a7f84
TO		 1289{
1290 journal_t *journal = EXT4_SB(sb)->s_journal;
1291
1292 WARN_ON(PageChecked(page));
1293 if (!page_has_buffers(page))
1294 return 0;
1295 if (journal)
1296 return jbd2_journal_try_to_free_buffers(journal, page,
d0164adc 1297 wait & ~__GFP_DIRECT_RECLAIM);
c39a7f84
TO		 1298 return try_to_free_buffers(page);
1299}
1300
643fa961 1301#ifdef CONFIG_FS_ENCRYPTION
a7550b30
JK		 1302static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1303{
1304 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1305 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1306}
1307
a7550b30
JK		 1308static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1309 void *fs_data)
1310{
2f8f5e76 1311 handle_t *handle = fs_data;
c1a5d5f6 1312 int res, res2, credits, retries = 0;
2f8f5e76 1313
9ce0151a
EB		 1314 /*
1315 * Encrypting the root directory is not allowed because e2fsck expects
1316 * lost+found to exist and be unencrypted, and encrypting the root
1317 * directory would imply encrypting the lost+found directory as well as
1318 * the filename "lost+found" itself.
1319 */
1320 if (inode->i_ino == EXT4_ROOT_INO)
1321 return -EPERM;
2f8f5e76 1322
7d3e06a8
RZ		 1323 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1324 return -EINVAL;
1325
b383a73f
IW		 1326 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1327 return -EOPNOTSUPP;
1328
94840e3c
EB		 1329 res = ext4_convert_inline_data(inode);
1330 if (res)
1331 return res;
1332
2f8f5e76
EB		 1333 /*
1334 * If a journal handle was specified, then the encryption context is
1335 * being set on a new inode via inheritance and is part of a larger
1336 * transaction to create the inode. Otherwise the encryption context is
1337 * being set on an existing inode in its own transaction. Only in the
1338 * latter case should the "retry on ENOSPC" logic be used.
1339 */
a7550b30 1340
2f8f5e76
EB		 1341 if (handle) {
1342 res = ext4_xattr_set_handle(handle, inode,
1343 EXT4_XATTR_INDEX_ENCRYPTION,
1344 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1345 ctx, len, 0);
a7550b30
JK		 1346 if (!res) {
1347 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1348 ext4_clear_inode_state(inode,
1349 EXT4_STATE_MAY_INLINE_DATA);
a3caa24b 1350 /*
2ee6a576
EB		 1351 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1352 * S_DAX may be disabled
a3caa24b 1353 */
043546e4 1354 ext4_set_inode_flags(inode, false);
a7550b30
JK		 1355 }
1356 return res;
1357 }
1358
b8cb5a54
TE		 1359 res = dquot_initialize(inode);
1360 if (res)
1361 return res;
2f8f5e76 1362retry:
af65207c
TE		 1363 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1364 &credits);
dec214d0
TE		 1365 if (res)
1366 return res;
1367
c1a5d5f6 1368 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
a7550b30
JK		 1369 if (IS_ERR(handle))
1370 return PTR_ERR(handle);
1371
2f8f5e76
EB		 1372 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1373 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1374 ctx, len, 0);
a7550b30
JK		 1375 if (!res) {
1376 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
2ee6a576
EB		 1377 /*
1378 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1379 * S_DAX may be disabled
1380 */
043546e4 1381 ext4_set_inode_flags(inode, false);
a7550b30
JK		 1382 res = ext4_mark_inode_dirty(handle, inode);
1383 if (res)
1384 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1385 }
1386 res2 = ext4_journal_stop(handle);
2f8f5e76
EB		 1387
1388 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1389 goto retry;
a7550b30
JK		 1390 if (!res)
1391 res = res2;
1392 return res;
1393}
1394
c250b7dd 1395static bool ext4_dummy_context(struct inode *inode)
a7550b30
JK		 1396{
1397 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1398}
1399
b925acb8
EB		 1400static bool ext4_has_stable_inodes(struct super_block *sb)
1401{
1402 return ext4_has_feature_stable_inodes(sb);
1403}
1404
1405static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1406 int *ino_bits_ret, int *lblk_bits_ret)
1407{
1408 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1409 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1410}
1411
6f69f0ed 1412static const struct fscrypt_operations ext4_cryptops = {
a5d431ef 1413 .key_prefix = "ext4:",
a7550b30 1414 .get_context = ext4_get_context,
a7550b30
JK		 1415 .set_context = ext4_set_context,
1416 .dummy_context = ext4_dummy_context,
a7550b30 1417 .empty_dir = ext4_empty_dir,
e12ee683 1418 .max_namelen = EXT4_NAME_LEN,
b925acb8
EB		 1419 .has_stable_inodes = ext4_has_stable_inodes,
1420 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
a7550b30 1421};
a7550b30
JK		 1422#endif
1423
ac27a0ec 1424#ifdef CONFIG_QUOTA
d6006186 1425static const char * const quotatypes[] = INITQFNAMES;
689c958c 1426#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1427
617ba13b
MC		 1428static int ext4_write_dquot(struct dquot *dquot);
1429static int ext4_acquire_dquot(struct dquot *dquot);
1430static int ext4_release_dquot(struct dquot *dquot);
1431static int ext4_mark_dquot_dirty(struct dquot *dquot);
1432static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1433static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1434 const struct path *path);
617ba13b
MC		 1435static int ext4_quota_on_mount(struct super_block *sb, int type);
1436static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1437 size_t len, loff_t off);
617ba13b 1438static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1439 const char *data, size_t len, loff_t off);
7c319d32
AK		 1440static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1441 unsigned int flags);
1442static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1443
96c7e0d9
JK		 1444static struct dquot **ext4_get_dquots(struct inode *inode)
1445{
1446 return EXT4_I(inode)->i_dquot;
1447}
1448
61e225dc 1449static const struct dquot_operations ext4_quota_operations = {
7a9ca53a
TE		 1450 .get_reserved_space = ext4_get_reserved_space,
1451 .write_dquot = ext4_write_dquot,
1452 .acquire_dquot = ext4_acquire_dquot,
1453 .release_dquot = ext4_release_dquot,
1454 .mark_dirty = ext4_mark_dquot_dirty,
1455 .write_info = ext4_write_info,
1456 .alloc_dquot = dquot_alloc,
1457 .destroy_dquot = dquot_destroy,
1458 .get_projid = ext4_get_projid,
1459 .get_inode_usage = ext4_get_inode_usage,
ebc11f7b 1460 .get_next_id = dquot_get_next_id,
ac27a0ec
DK		 1461};
1462
0d54b217 1463static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1464 .quota_on = ext4_quota_on,
ca0e05e4 1465 .quota_off = ext4_quota_off,
287a8095 1466 .quota_sync = dquot_quota_sync,
0a240339 1467 .get_state = dquot_get_state,
287a8095
CH		 1468 .set_info = dquot_set_dqinfo,
1469 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1470 .set_dqblk = dquot_set_dqblk,
1471 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1472};
1473#endif
1474
ee9b6d61 1475static const struct super_operations ext4_sops = {
617ba13b 1476 .alloc_inode = ext4_alloc_inode,
94053139 1477 .free_inode = ext4_free_in_core_inode,
617ba13b 1478 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1479 .write_inode = ext4_write_inode,
1480 .dirty_inode = ext4_dirty_inode,
7ff9c073 1481 .drop_inode = ext4_drop_inode,
0930fcc1 1482 .evict_inode = ext4_evict_inode,
617ba13b 1483 .put_super = ext4_put_super,
617ba13b 1484 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1485 .freeze_fs = ext4_freeze,
1486 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1487 .statfs = ext4_statfs,
1488 .remount_fs = ext4_remount,
617ba13b 1489 .show_options = ext4_show_options,
ac27a0ec 1490#ifdef CONFIG_QUOTA
617ba13b
MC		 1491 .quota_read = ext4_quota_read,
1492 .quota_write = ext4_quota_write,
96c7e0d9 1493 .get_dquots = ext4_get_dquots,
ac27a0ec 1494#endif
c39a7f84 1495 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1496};
1497
39655164 1498static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1499 .fh_to_dentry = ext4_fh_to_dentry,
1500 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1501 .get_parent = ext4_get_parent,
fde87268 1502 .commit_metadata = ext4_nfs_commit_metadata,
ac27a0ec
DK		 1503};
1504
1505enum {
1506 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1507 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1508 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1509 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1510 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1511 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1512 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1513 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1514 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1515 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1516 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1517 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
9cb20f94
IW		 1518 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1519 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
327eaf73
TT		 1520 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1521 Opt_nowarn_on_error, Opt_mblk_io_submit,
670e9875 1522 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1449032b 1523 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1524 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1525 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1526 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1527 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
ac27a0ec
DK		 1528};
1529
a447c093 1530static const match_table_t tokens = {
ac27a0ec
DK		 1531 {Opt_bsd_df, "bsddf"},
1532 {Opt_minix_df, "minixdf"},
1533 {Opt_grpid, "grpid"},
1534 {Opt_grpid, "bsdgroups"},
1535 {Opt_nogrpid, "nogrpid"},
1536 {Opt_nogrpid, "sysvgroups"},
1537 {Opt_resgid, "resgid=%u"},
1538 {Opt_resuid, "resuid=%u"},
1539 {Opt_sb, "sb=%u"},
1540 {Opt_err_cont, "errors=continue"},
1541 {Opt_err_panic, "errors=panic"},
1542 {Opt_err_ro, "errors=remount-ro"},
1543 {Opt_nouid32, "nouid32"},
ac27a0ec 1544 {Opt_debug, "debug"},
72578c33
TT		 1545 {Opt_removed, "oldalloc"},
1546 {Opt_removed, "orlov"},
ac27a0ec
DK		 1547 {Opt_user_xattr, "user_xattr"},
1548 {Opt_nouser_xattr, "nouser_xattr"},
1549 {Opt_acl, "acl"},
1550 {Opt_noacl, "noacl"},
e3bb52ae 1551 {Opt_noload, "norecovery"},
5a916be1 1552 {Opt_noload, "noload"},
72578c33
TT		 1553 {Opt_removed, "nobh"},
1554 {Opt_removed, "bh"},
ac27a0ec 1555 {Opt_commit, "commit=%u"},
30773840
TT		 1556 {Opt_min_batch_time, "min_batch_time=%u"},
1557 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1558 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1559 {Opt_journal_path, "journal_path=%s"},
818d276c 1560 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1561 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1562 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1563 {Opt_abort, "abort"},
1564 {Opt_data_journal, "data=journal"},
1565 {Opt_data_ordered, "data=ordered"},
1566 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1567 {Opt_data_err_abort, "data_err=abort"},
1568 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1569 {Opt_offusrjquota, "usrjquota="},
1570 {Opt_usrjquota, "usrjquota=%s"},
1571 {Opt_offgrpjquota, "grpjquota="},
1572 {Opt_grpjquota, "grpjquota=%s"},
1573 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1574 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1575 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1576 {Opt_grpquota, "grpquota"},
1577 {Opt_noquota, "noquota"},
1578 {Opt_quota, "quota"},
1579 {Opt_usrquota, "usrquota"},
49da9392 1580 {Opt_prjquota, "prjquota"},
ac27a0ec 1581 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1582 {Opt_barrier, "barrier"},
1583 {Opt_nobarrier, "nobarrier"},
25ec56b5 1584 {Opt_i_version, "i_version"},
923ae0ff 1585 {Opt_dax, "dax"},
9cb20f94
IW		 1586 {Opt_dax_always, "dax=always"},
1587 {Opt_dax_inode, "dax=inode"},
1588 {Opt_dax_never, "dax=never"},
c9de560d 1589 {Opt_stripe, "stripe=%u"},
64769240 1590 {Opt_delalloc, "delalloc"},
327eaf73
TT		 1591 {Opt_warn_on_error, "warn_on_error"},
1592 {Opt_nowarn_on_error, "nowarn_on_error"},
a26f4992
TT		 1593 {Opt_lazytime, "lazytime"},
1594 {Opt_nolazytime, "nolazytime"},
670e9875 1595 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
dd919b98 1596 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1597 {Opt_removed, "mblk_io_submit"},
1598 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1599 {Opt_block_validity, "block_validity"},
1600 {Opt_noblock_validity, "noblock_validity"},
240799cd 1601 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1602 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1603 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1604 {Opt_auto_da_alloc, "auto_da_alloc"},
1605 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc 1606 {Opt_dioread_nolock, "dioread_nolock"},
244adf64 1607 {Opt_dioread_lock, "nodioread_nolock"},
744692dc 1608 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1609 {Opt_discard, "discard"},
1610 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1611 {Opt_init_itable, "init_itable=%u"},
1612 {Opt_init_itable, "init_itable"},
1613 {Opt_noinit_itable, "noinit_itable"},
df981d03 1614 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1615 {Opt_test_dummy_encryption, "test_dummy_encryption"},
cdb7ee4c
TE		 1616 {Opt_nombcache, "nombcache"},
1617 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
c7198b9c
TT		 1618 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1619 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1620 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1621 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1622 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1623 {Opt_err, NULL},
ac27a0ec
DK		 1624};
1625
617ba13b 1626static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1627{
617ba13b 1628 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1629 char *options = (char *) *data;
1630
1631 if (!options || strncmp(options, "sb=", 3) != 0)
1632 return 1; /* Default location */
0b8e58a1 1633
ac27a0ec 1634 options += 3;
0b8e58a1 1635 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1636 sb_block = simple_strtoul(options, &options, 0);
1637 if (*options && *options != ',') {
4776004f 1638 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1639 (char *) *data);
1640 return 1;
1641 }
1642 if (*options == ',')
1643 options++;
1644 *data = (void *) options;
0b8e58a1 1645
ac27a0ec
DK		 1646 return sb_block;
1647}
1648
b3881f74 1649#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
d6006186
EB		 1650static const char deprecated_msg[] =
1651 "Mount option \"%s\" will be removed by %s\n"
437ca0fd 1652 "Contact [email protected] if you think we should keep it.\n";
b3881f74 1653
56c50f11
DM		 1654#ifdef CONFIG_QUOTA
1655static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1656{
1657 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 1658 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
03dafb5f 1659 int ret = -1;
56c50f11 1660
33458eab 1661 if (sb_any_quota_loaded(sb) && !old_qname) {
56c50f11
DM		 1662 ext4_msg(sb, KERN_ERR,
1663 "Cannot change journaled "
1664 "quota options when quota turned on");
57f73c2c 1665 return -1;
56c50f11 1666 }
e2b911c5 1667 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1668 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1669 "ignored when QUOTA feature is enabled");
1670 return 1;
262b4662 1671 }
56c50f11
DM		 1672 qname = match_strdup(args);
1673 if (!qname) {
1674 ext4_msg(sb, KERN_ERR,
1675 "Not enough memory for storing quotafile name");
57f73c2c 1676 return -1;
56c50f11 1677 }
33458eab
TT		 1678 if (old_qname) {
1679 if (strcmp(old_qname, qname) == 0)
03dafb5f
CG		 1680 ret = 1;
1681 else
1682 ext4_msg(sb, KERN_ERR,
1683 "%s quota file already specified",
1684 QTYPE2NAME(qtype));
1685 goto errout;
56c50f11 1686 }
03dafb5f 1687 if (strchr(qname, '/')) {
56c50f11
DM		 1688 ext4_msg(sb, KERN_ERR,
1689 "quotafile must be on filesystem root");
03dafb5f 1690 goto errout;
56c50f11 1691 }
33458eab 1692 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
fd8c37ec 1693 set_opt(sb, QUOTA);
56c50f11 1694 return 1;
03dafb5f
CG		 1695errout:
1696 kfree(qname);
1697 return ret;
56c50f11
DM		 1698}
1699
1700static int clear_qf_name(struct super_block *sb, int qtype)
1701{
1702
1703 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 1704 char *old_qname = get_qf_name(sb, sbi, qtype);
56c50f11 1705
33458eab 1706 if (sb_any_quota_loaded(sb) && old_qname) {
56c50f11
DM		 1707 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1708 " when quota turned on");
57f73c2c 1709 return -1;
56c50f11 1710 }
33458eab
TT		 1711 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1712 synchronize_rcu();
1713 kfree(old_qname);
56c50f11
DM		 1714 return 1;
1715}
1716#endif
1717
26092bf5
TT		 1718#define MOPT_SET 0x0001
1719#define MOPT_CLEAR 0x0002
1720#define MOPT_NOSUPPORT 0x0004
1721#define MOPT_EXPLICIT 0x0008
1722#define MOPT_CLEAR_ERR 0x0010
1723#define MOPT_GTE0 0x0020
ac27a0ec 1724#ifdef CONFIG_QUOTA
26092bf5
TT		 1725#define MOPT_Q 0
1726#define MOPT_QFMT 0x0040
1727#else
1728#define MOPT_Q MOPT_NOSUPPORT
1729#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1730#endif
26092bf5 1731#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1732#define MOPT_NO_EXT2 0x0100
1733#define MOPT_NO_EXT3 0x0200
1734#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1735#define MOPT_STRING 0x0400
9cb20f94 1736#define MOPT_SKIP 0x0800
26092bf5
TT		 1737
1738static const struct mount_opts {
1739 int token;
1740 int mount_opt;
1741 int flags;
1742} ext4_mount_opts[] = {
1743 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1744 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1745 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1746 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1747 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1748 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1749 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1750 MOPT_EXT4_ONLY | MOPT_SET},
1751 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1752 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1753 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1754 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1755 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1756 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1757 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1758 MOPT_EXT4_ONLY | MOPT_CLEAR},
327eaf73
TT		 1759 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1760 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
c6d3d56d
DW		 1761 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1762 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1763 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1764 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1765 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1766 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1767 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1768 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1769 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1770 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1771 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1772 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1773 MOPT_NO_EXT2},
8dc0aa8c 1774 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1775 MOPT_NO_EXT2},
26092bf5
TT		 1776 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1777 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1778 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1779 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1780 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1781 {Opt_commit, 0, MOPT_GTE0},
1782 {Opt_max_batch_time, 0, MOPT_GTE0},
1783 {Opt_min_batch_time, 0, MOPT_GTE0},
1784 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1785 {Opt_init_itable, 0, MOPT_GTE0},
9cb20f94
IW		 1786 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1787 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1788 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1789 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1790 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1791 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1792 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
26092bf5 1793 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1794 {Opt_resuid, 0, MOPT_GTE0},
1795 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1796 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1797 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1798 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1799 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1800 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1801 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1802 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1803 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1804 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1805#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1806 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1807 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1808#else
26092bf5
TT		 1809 {Opt_acl, 0, MOPT_NOSUPPORT},
1810 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1811#endif
26092bf5
TT		 1812 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1813 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
670e9875 1814 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
26092bf5
TT		 1815 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1816 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1817 MOPT_SET | MOPT_Q},
1818 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1819 MOPT_SET | MOPT_Q},
49da9392
JK		 1820 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1821 MOPT_SET | MOPT_Q},
26092bf5 1822 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1823 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1824 MOPT_CLEAR | MOPT_Q},
26092bf5
TT		 1825 {Opt_usrjquota, 0, MOPT_Q},
1826 {Opt_grpjquota, 0, MOPT_Q},
1827 {Opt_offusrjquota, 0, MOPT_Q},
1828 {Opt_offgrpjquota, 0, MOPT_Q},
1829 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1830 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1831 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1832 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1833 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
cdb7ee4c 1834 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
26092bf5
TT		 1835 {Opt_err, 0, 0}
1836};
1837
c83ad55e
GKB		 1838#ifdef CONFIG_UNICODE
1839static const struct ext4_sb_encodings {
1840 __u16 magic;
1841 char *name;
1842 char *version;
1843} ext4_sb_encoding_map[] = {
1844 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
1845};
1846
1847static int ext4_sb_read_encoding(const struct ext4_super_block *es,
1848 const struct ext4_sb_encodings **encoding,
1849 __u16 *flags)
1850{
1851 __u16 magic = le16_to_cpu(es->s_encoding);
1852 int i;
1853
1854 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1855 if (magic == ext4_sb_encoding_map[i].magic)
1856 break;
1857
1858 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
1859 return -EINVAL;
1860
1861 *encoding = &ext4_sb_encoding_map[i];
1862 *flags = le16_to_cpu(es->s_encoding_flags);
1863
1864 return 0;
1865}
1866#endif
1867
26092bf5
TT		 1868static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1869 substring_t *args, unsigned long *journal_devnum,
1870 unsigned int *journal_ioprio, int is_remount)
1871{
1872 struct ext4_sb_info *sbi = EXT4_SB(sb);
1873 const struct mount_opts *m;
08cefc7a
EB		 1874 kuid_t uid;
1875 kgid_t gid;
26092bf5
TT		 1876 int arg = 0;
1877
57f73c2c
TT		 1878#ifdef CONFIG_QUOTA
1879 if (token == Opt_usrjquota)
1880 return set_qf_name(sb, USRQUOTA, &args[0]);
1881 else if (token == Opt_grpjquota)
1882 return set_qf_name(sb, GRPQUOTA, &args[0]);
1883 else if (token == Opt_offusrjquota)
1884 return clear_qf_name(sb, USRQUOTA);
1885 else if (token == Opt_offgrpjquota)
1886 return clear_qf_name(sb, GRPQUOTA);
1887#endif
26092bf5 1888 switch (token) {
f7048605
TT		 1889 case Opt_noacl:
1890 case Opt_nouser_xattr:
1891 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1892 break;
26092bf5
TT		 1893 case Opt_sb:
1894 return 1; /* handled by get_sb_block() */
1895 case Opt_removed:
5f3633e3 1896 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1897 return 1;
26092bf5
TT		 1898 case Opt_abort:
1899 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1900 return 1;
1901 case Opt_i_version:
357fdad0 1902 sb->s_flags |= SB_I_VERSION;
26092bf5 1903 return 1;
a26f4992 1904 case Opt_lazytime:
1751e8a6 1905 sb->s_flags |= SB_LAZYTIME;
a26f4992
TT		 1906 return 1;
1907 case Opt_nolazytime:
1751e8a6 1908 sb->s_flags &= ~SB_LAZYTIME;
a26f4992 1909 return 1;
26092bf5
TT		 1910 }
1911
5f3633e3
JK		 1912 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1913 if (token == m->token)
1914 break;
1915
1916 if (m->token == Opt_err) {
1917 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1918 "or missing value", opt);
1919 return -1;
1920 }
1921
8dc0aa8c
TT		 1922 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1923 ext4_msg(sb, KERN_ERR,
1924 "Mount option \"%s\" incompatible with ext2", opt);
1925 return -1;
1926 }
1927 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1928 ext4_msg(sb, KERN_ERR,
1929 "Mount option \"%s\" incompatible with ext3", opt);
1930 return -1;
1931 }
1932
ad4eec61 1933 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1934 return -1;
1935 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1936 return -1;
c93cf2d7
DM		 1937 if (m->flags & MOPT_EXPLICIT) {
1938 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1939 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1940 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1941 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1942 } else
1943 return -1;
1944 }
5f3633e3
JK		 1945 if (m->flags & MOPT_CLEAR_ERR)
1946 clear_opt(sb, ERRORS_MASK);
1947 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1948 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1949 "options when quota turned on");
1950 return -1;
1951 }
1952
1953 if (m->flags & MOPT_NOSUPPORT) {
1954 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1955 } else if (token == Opt_commit) {
1956 if (arg == 0)
1957 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
9ba55543 1958 else if (arg > INT_MAX / HZ) {
1959 ext4_msg(sb, KERN_ERR,
1960 "Invalid commit interval %d, "
1961 "must be smaller than %d",
1962 arg, INT_MAX / HZ);
1963 return -1;
1964 }
5f3633e3 1965 sbi->s_commit_interval = HZ * arg;
670e9875 1966 } else if (token == Opt_debug_want_extra_isize) {
9803387c
TT		 1967 if ((arg & 1) ||
1968 (arg < 4) ||
1969 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
1970 ext4_msg(sb, KERN_ERR,
1971 "Invalid want_extra_isize %d", arg);
1972 return -1;
1973 }
670e9875 1974 sbi->s_want_extra_isize = arg;
5f3633e3 1975 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1976 sbi->s_max_batch_time = arg;
1977 } else if (token == Opt_min_batch_time) {
1978 sbi->s_min_batch_time = arg;
1979 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1980 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1981 ext4_msg(sb, KERN_ERR,
1982 "EXT4-fs: inode_readahead_blks must be "
1983 "0 or a power of 2 smaller than 2^31");
26092bf5 1984 return -1;
5f3633e3
JK		 1985 }
1986 sbi->s_inode_readahead_blks = arg;
1987 } else if (token == Opt_init_itable) {
1988 set_opt(sb, INIT_INODE_TABLE);
1989 if (!args->from)
1990 arg = EXT4_DEF_LI_WAIT_MULT;
1991 sbi->s_li_wait_mult = arg;
1992 } else if (token == Opt_max_dir_size_kb) {
1993 sbi->s_max_dir_size_kb = arg;
1994 } else if (token == Opt_stripe) {
1995 sbi->s_stripe = arg;
1996 } else if (token == Opt_resuid) {
1997 uid = make_kuid(current_user_ns(), arg);
1998 if (!uid_valid(uid)) {
1999 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 2000 return -1;
2001 }
5f3633e3
JK		 2002 sbi->s_resuid = uid;
2003 } else if (token == Opt_resgid) {
2004 gid = make_kgid(current_user_ns(), arg);
2005 if (!gid_valid(gid)) {
2006 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2007 return -1;
2008 }
2009 sbi->s_resgid = gid;
2010 } else if (token == Opt_journal_dev) {
2011 if (is_remount) {
2012 ext4_msg(sb, KERN_ERR,
2013 "Cannot specify journal on remount");
2014 return -1;
2015 }
2016 *journal_devnum = arg;
ad4eec61
ES		 2017 } else if (token == Opt_journal_path) {
2018 char *journal_path;
2019 struct inode *journal_inode;
2020 struct path path;
2021 int error;
2022
2023 if (is_remount) {
2024 ext4_msg(sb, KERN_ERR,
2025 "Cannot specify journal on remount");
2026 return -1;
2027 }
2028 journal_path = match_strdup(&args[0]);
2029 if (!journal_path) {
2030 ext4_msg(sb, KERN_ERR, "error: could not dup "
2031 "journal device string");
2032 return -1;
2033 }
2034
2035 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2036 if (error) {
2037 ext4_msg(sb, KERN_ERR, "error: could not find "
2038 "journal device path: error %d", error);
2039 kfree(journal_path);
2040 return -1;
2041 }
2042
2b0143b5 2043 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 2044 if (!S_ISBLK(journal_inode->i_mode)) {
2045 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2046 "is not a block device", journal_path);
2047 path_put(&path);
2048 kfree(journal_path);
2049 return -1;
2050 }
2051
2052 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2053 path_put(&path);
2054 kfree(journal_path);
5f3633e3
JK		 2055 } else if (token == Opt_journal_ioprio) {
2056 if (arg > 7) {
2057 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2058 " (must be 0-7)");
2059 return -1;
2060 }
2061 *journal_ioprio =
2062 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447 2063 } else if (token == Opt_test_dummy_encryption) {
643fa961 2064#ifdef CONFIG_FS_ENCRYPTION
6ddb2447
TT		 2065 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
2066 ext4_msg(sb, KERN_WARNING,
2067 "Test dummy encryption mode enabled");
2068#else
2069 ext4_msg(sb, KERN_WARNING,
2070 "Test dummy encryption mount option ignored");
2071#endif
5f3633e3
JK		 2072 } else if (m->flags & MOPT_DATAJ) {
2073 if (is_remount) {
2074 if (!sbi->s_journal)
2075 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2076 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 2077 ext4_msg(sb, KERN_ERR,
26092bf5 2078 "Cannot change data mode on remount");
26092bf5 2079 return -1;
ac27a0ec 2080 }
26092bf5 2081 } else {
5f3633e3
JK		 2082 clear_opt(sb, DATA_FLAGS);
2083 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 2084 }
5f3633e3
JK		 2085#ifdef CONFIG_QUOTA
2086 } else if (m->flags & MOPT_QFMT) {
2087 if (sb_any_quota_loaded(sb) &&
2088 sbi->s_jquota_fmt != m->mount_opt) {
2089 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2090 "quota options when quota turned on");
2091 return -1;
2092 }
e2b911c5 2093 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 2094 ext4_msg(sb, KERN_INFO,
2095 "Quota format mount options ignored "
262b4662 2096 "when QUOTA feature is enabled");
c325a67c 2097 return 1;
262b4662 2098 }
5f3633e3 2099 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 2100#endif
9cb20f94
IW		 2101 } else if (token == Opt_dax || token == Opt_dax_always ||
2102 token == Opt_dax_inode || token == Opt_dax_never) {
ef83b6e8 2103#ifdef CONFIG_FS_DAX
9cb20f94
IW		 2104 switch (token) {
2105 case Opt_dax:
2106 case Opt_dax_always:
2107 ext4_msg(sb, KERN_WARNING,
2108 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2109 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2110 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2111 break;
2112 case Opt_dax_never:
2113 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2114 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2115 break;
2116 case Opt_dax_inode:
2117 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2118 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2119 /* Strictly for printing options */
2120 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2121 break;
2122 }
ef83b6e8 2123#else
923ae0ff 2124 ext4_msg(sb, KERN_INFO, "dax option not supported");
9cb20f94
IW		 2125 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2126 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
923ae0ff 2127 return -1;
5f3633e3 2128#endif
7915a861
AN		 2129 } else if (token == Opt_data_err_abort) {
2130 sbi->s_mount_opt |= m->mount_opt;
2131 } else if (token == Opt_data_err_ignore) {
2132 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 2133 } else {
2134 if (!args->from)
2135 arg = 1;
2136 if (m->flags & MOPT_CLEAR)
2137 arg = !arg;
2138 else if (unlikely(!(m->flags & MOPT_SET))) {
2139 ext4_msg(sb, KERN_WARNING,
2140 "buggy handling of option %s", opt);
2141 WARN_ON(1);
2142 return -1;
2143 }
2144 if (arg != 0)
2145 sbi->s_mount_opt |= m->mount_opt;
2146 else
2147 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 2148 }
5f3633e3 2149 return 1;
26092bf5
TT		 2150}
2151
2152static int parse_options(char *options, struct super_block *sb,
2153 unsigned long *journal_devnum,
2154 unsigned int *journal_ioprio,
2155 int is_remount)
2156{
1e1a76ed 2157 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
33458eab 2158 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
26092bf5
TT		 2159 substring_t args[MAX_OPT_ARGS];
2160 int token;
2161
2162 if (!options)
2163 return 1;
2164
2165 while ((p = strsep(&options, ",")) != NULL) {
2166 if (!*p)
2167 continue;
2168 /*
2169 * Initialize args struct so we know whether arg was
2170 * found; some options take optional arguments.
2171 */
caecd0af 2172 args[0].to = args[0].from = NULL;
26092bf5
TT		 2173 token = match_token(p, tokens, args);
2174 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2175 journal_ioprio, is_remount) < 0)
2176 return 0;
ac27a0ec
DK		 2177 }
2178#ifdef CONFIG_QUOTA
49da9392
JK		 2179 /*
2180 * We do the test below only for project quotas. 'usrquota' and
2181 * 'grpquota' mount options are allowed even without quota feature
2182 * to support legacy quotas in quota files.
2183 */
2184 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2185 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2186 "Cannot enable project quota enforcement.");
2187 return 0;
2188 }
33458eab
TT		 2189 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2190 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2191 if (usr_qf_name || grp_qf_name) {
2192 if (test_opt(sb, USRQUOTA) && usr_qf_name)
fd8c37ec 2193 clear_opt(sb, USRQUOTA);
ac27a0ec 2194
33458eab 2195 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
fd8c37ec 2196 clear_opt(sb, GRPQUOTA);
ac27a0ec 2197
56c50f11 2198 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 2199 ext4_msg(sb, KERN_ERR, "old and new quota "
2200 "format mixing");
ac27a0ec
DK		 2201 return 0;
2202 }
2203
2204 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 2205 ext4_msg(sb, KERN_ERR, "journaled quota format "
2206 "not specified");
ac27a0ec
DK		 2207 return 0;
2208 }
ac27a0ec
DK		 2209 }
2210#endif
626b035b
RH		 2211 if (test_opt(sb, DIOREAD_NOLOCK)) {
2212 int blocksize =
2213 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2214 if (blocksize < PAGE_SIZE)
2215 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2216 "experimental mount option 'dioread_nolock' "
2217 "for blocksize < PAGE_SIZE");
2218 }
ac27a0ec
DK		 2219 return 1;
2220}
2221
2adf6da8
TT		 2222static inline void ext4_show_quota_options(struct seq_file *seq,
2223 struct super_block *sb)
2224{
2225#if defined(CONFIG_QUOTA)
2226 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 2227 char *usr_qf_name, *grp_qf_name;
2adf6da8
TT		 2228
2229 if (sbi->s_jquota_fmt) {
2230 char *fmtname = "";
2231
2232 switch (sbi->s_jquota_fmt) {
2233 case QFMT_VFS_OLD:
2234 fmtname = "vfsold";
2235 break;
2236 case QFMT_VFS_V0:
2237 fmtname = "vfsv0";
2238 break;
2239 case QFMT_VFS_V1:
2240 fmtname = "vfsv1";
2241 break;
2242 }
2243 seq_printf(seq, ",jqfmt=%s", fmtname);
2244 }
2245
33458eab
TT		 2246 rcu_read_lock();
2247 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2248 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2249 if (usr_qf_name)
2250 seq_show_option(seq, "usrjquota", usr_qf_name);
2251 if (grp_qf_name)
2252 seq_show_option(seq, "grpjquota", grp_qf_name);
2253 rcu_read_unlock();
2adf6da8
TT		 2254#endif
2255}
2256
5a916be1
TT		 2257static const char *token2str(int token)
2258{
50df9fd5 2259 const struct match_token *t;
5a916be1
TT		 2260
2261 for (t = tokens; t->token != Opt_err; t++)
2262 if (t->token == token && !strchr(t->pattern, '='))
2263 break;
2264 return t->pattern;
2265}
2266
2adf6da8
TT		 2267/*
2268 * Show an option if
2269 * - it's set to a non-default value OR
2270 * - if the per-sb default is different from the global default
2271 */
66acdcf4
TT		 2272static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2273 int nodefs)
2adf6da8 2274{
2adf6da8
TT		 2275 struct ext4_sb_info *sbi = EXT4_SB(sb);
2276 struct ext4_super_block *es = sbi->s_es;
68afa7e0 2277 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
5a916be1 2278 const struct mount_opts *m;
66acdcf4 2279 char sep = nodefs ? '\n' : ',';
2adf6da8 2280
66acdcf4
TT		 2281#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2282#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2283
2284 if (sbi->s_sb_block != 1)
5a916be1
TT		 2285 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2286
2287 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2288 int want_set = m->flags & MOPT_SET;
2289 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
9cb20f94 2290 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
5a916be1 2291 continue;
68afa7e0 2292 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 2293 continue; /* skip if same as the default */
2294 if ((want_set &&
2295 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2296 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2297 continue; /* select Opt_noFoo vs Opt_Foo */
2298 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2299 }
5a916be1 2300
08cefc7a 2301 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2302 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2303 SEQ_OPTS_PRINT("resuid=%u",
2304 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2305 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2306 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2307 SEQ_OPTS_PRINT("resgid=%u",
2308 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2309 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2310 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2311 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2312 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2313 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2314 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2315 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2316 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2317 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2318 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2319 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2320 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2321 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
357fdad0 2322 if (sb->s_flags & SB_I_VERSION)
5a916be1 2323 SEQ_OPTS_PUTS("i_version");
66acdcf4 2324 if (nodefs || sbi->s_stripe)
5a916be1 2325 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
68afa7e0
TN		 2326 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2327 (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2328 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2329 SEQ_OPTS_PUTS("data=journal");
2330 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2331 SEQ_OPTS_PUTS("data=ordered");
2332 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2333 SEQ_OPTS_PUTS("data=writeback");
2334 }
66acdcf4
TT		 2335 if (nodefs ||
2336 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2337 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2338 sbi->s_inode_readahead_blks);
2adf6da8 2339
ceec0376 2340 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
66acdcf4 2341 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2342 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2343 if (nodefs || sbi->s_max_dir_size_kb)
2344 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2345 if (test_opt(sb, DATA_ERR_ABORT))
2346 SEQ_OPTS_PUTS("data_err=abort");
338affb5
EB		 2347 if (DUMMY_ENCRYPTION_ENABLED(sbi))
2348 SEQ_OPTS_PUTS("test_dummy_encryption");
2adf6da8 2349
9cb20f94
IW		 2350 if (test_opt(sb, DAX_ALWAYS)) {
2351 if (IS_EXT2_SB(sb))
2352 SEQ_OPTS_PUTS("dax");
2353 else
2354 SEQ_OPTS_PUTS("dax=always");
2355 } else if (test_opt2(sb, DAX_NEVER)) {
2356 SEQ_OPTS_PUTS("dax=never");
2357 } else if (test_opt2(sb, DAX_INODE)) {
2358 SEQ_OPTS_PUTS("dax=inode");
2359 }
2360
2adf6da8 2361 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2362 return 0;
2363}
2364
66acdcf4
TT		 2365static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2366{
2367 return _ext4_show_options(seq, root->d_sb, 0);
2368}
2369
ebd173be 2370int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2371{
2372 struct super_block *sb = seq->private;
2373 int rc;
2374
bc98a42c 2375 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
66acdcf4
TT		 2376 rc = _ext4_show_options(seq, sb, 1);
2377 seq_puts(seq, "\n");
2378 return rc;
2379}
2380
617ba13b 2381static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2382 int read_only)
2383{
617ba13b 2384 struct ext4_sb_info *sbi = EXT4_SB(sb);
c89128a0 2385 int err = 0;
ac27a0ec 2386
617ba13b 2387 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2388 ext4_msg(sb, KERN_ERR, "revision level too high, "
2389 "forcing read-only mode");
c89128a0 2390 err = -EROFS;
ac27a0ec
DK		 2391 }
2392 if (read_only)
281b5995 2393 goto done;
617ba13b 2394 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2395 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2396 "running e2fsck is recommended");
c8b459f4 2397 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2398 ext4_msg(sb, KERN_WARNING,
2399 "warning: mounting fs with errors, "
2400 "running e2fsck is recommended");
ed3ce80a 2401 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2402 le16_to_cpu(es->s_mnt_count) >=
2403 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2404 ext4_msg(sb, KERN_WARNING,
2405 "warning: maximal mount count reached, "
2406 "running e2fsck is recommended");
ac27a0ec 2407 else if (le32_to_cpu(es->s_checkinterval) &&
6a0678a7
AB		 2408 (ext4_get_tstamp(es, s_lastcheck) +
2409 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
b31e1552
ES		 2410 ext4_msg(sb, KERN_WARNING,
2411 "warning: checktime reached, "
2412 "running e2fsck is recommended");
0b8e58a1 2413 if (!sbi->s_journal)
0390131b 2414 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2415 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2416 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2417 le16_add_cpu(&es->s_mnt_count, 1);
6a0678a7 2418 ext4_update_tstamp(es, s_mtime);
0390131b 2419 if (sbi->s_journal)
e2b911c5 2420 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2421
c89128a0 2422 err = ext4_commit_super(sb, 1);
281b5995 2423done:
ac27a0ec 2424 if (test_opt(sb, DEBUG))
a9df9a49 2425 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2426 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2427 sb->s_blocksize,
2428 sbi->s_groups_count,
617ba13b
MC		 2429 EXT4_BLOCKS_PER_GROUP(sb),
2430 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2431 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2432
7abc52c2 2433 cleancache_init_fs(sb);
c89128a0 2434 return err;
ac27a0ec
DK		 2435}
2436
117fff10
TT		 2437int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2438{
2439 struct ext4_sb_info *sbi = EXT4_SB(sb);
7c990728 2440 struct flex_groups **old_groups, **new_groups;
37b0b6b8 2441 int size, i, j;
117fff10
TT		 2442
2443 if (!sbi->s_log_groups_per_flex)
2444 return 0;
2445
2446 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2447 if (size <= sbi->s_flex_groups_allocated)
2448 return 0;
2449
7c990728
SJS		 2450 new_groups = kvzalloc(roundup_pow_of_two(size *
2451 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
117fff10 2452 if (!new_groups) {
7c990728
SJS		 2453 ext4_msg(sb, KERN_ERR,
2454 "not enough memory for %d flex group pointers", size);
117fff10
TT		 2455 return -ENOMEM;
2456 }
7c990728
SJS		 2457 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2458 new_groups[i] = kvzalloc(roundup_pow_of_two(
2459 sizeof(struct flex_groups)),
2460 GFP_KERNEL);
2461 if (!new_groups[i]) {
37b0b6b8
DC		 2462 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2463 kvfree(new_groups[j]);
7c990728
SJS		 2464 kvfree(new_groups);
2465 ext4_msg(sb, KERN_ERR,
2466 "not enough memory for %d flex groups", size);
2467 return -ENOMEM;
2468 }
117fff10 2469 }
7c990728
SJS		 2470 rcu_read_lock();
2471 old_groups = rcu_dereference(sbi->s_flex_groups);
2472 if (old_groups)
2473 memcpy(new_groups, old_groups,
2474 (sbi->s_flex_groups_allocated *
2475 sizeof(struct flex_groups *)));
2476 rcu_read_unlock();
2477 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2478 sbi->s_flex_groups_allocated = size;
2479 if (old_groups)
2480 ext4_kvfree_array_rcu(old_groups);
117fff10
TT		 2481 return 0;
2482}
2483
772cb7c8
JS		 2484static int ext4_fill_flex_info(struct super_block *sb)
2485{
2486 struct ext4_sb_info *sbi = EXT4_SB(sb);
2487 struct ext4_group_desc *gdp = NULL;
7c990728 2488 struct flex_groups *fg;
772cb7c8 2489 ext4_group_t flex_group;
117fff10 2490 int i, err;
772cb7c8 2491
503358ae 2492 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2493 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2494 sbi->s_log_groups_per_flex = 0;
2495 return 1;
2496 }
2497
117fff10
TT		 2498 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2499 if (err)
9933fc0a 2500 goto failed;
772cb7c8 2501
772cb7c8 2502 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2503 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2504
2505 flex_group = ext4_flex_group(sbi, i);
7c990728
SJS		 2506 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2507 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
90ba983f 2508 atomic64_add(ext4_free_group_clusters(sb, gdp),
7c990728
SJS		 2509 &fg->free_clusters);
2510 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
772cb7c8
JS		 2511 }
2512
2513 return 1;
2514failed:
2515 return 0;
2516}
2517
e2b911c5 2518static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2519 struct ext4_group_desc *gdp)
717d50e4 2520{
b47820ed 2521 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2522 __u16 crc = 0;
feb0ab32 2523 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2524 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2525
9aa5d32b 2526 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2527 /* Use new metadata_csum algorithm */
feb0ab32 2528 __u32 csum32;
b47820ed 2529 __u16 dummy_csum = 0;
feb0ab32 2530
feb0ab32
DW		 2531 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2532 sizeof(le_group));
b47820ed
DJ		 2533 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2534 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2535 sizeof(dummy_csum));
2536 offset += sizeof(dummy_csum);
2537 if (offset < sbi->s_desc_size)
2538 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2539 sbi->s_desc_size - offset);
feb0ab32
DW		 2540
2541 crc = csum32 & 0xFFFF;
2542 goto out;
717d50e4
AD		 2543 }
2544
feb0ab32 2545 /* old crc16 code */
e2b911c5 2546 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2547 return 0;
2548
feb0ab32
DW		 2549 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2550 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2551 crc = crc16(crc, (__u8 *)gdp, offset);
2552 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2553 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2554 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2555 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2556 crc = crc16(crc, (__u8 *)gdp + offset,
2557 le16_to_cpu(sbi->s_es->s_desc_size) -
2558 offset);
2559
2560out:
717d50e4
AD		 2561 return cpu_to_le16(crc);
2562}
2563
feb0ab32 2564int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2565 struct ext4_group_desc *gdp)
2566{
feb0ab32 2567 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2568 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2569 return 0;
2570
2571 return 1;
2572}
2573
feb0ab32
DW		 2574void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2575 struct ext4_group_desc *gdp)
2576{
2577 if (!ext4_has_group_desc_csum(sb))
2578 return;
e2b911c5 2579 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2580}
2581
ac27a0ec 2582/* Called at mount-time, super-block is locked */
bfff6873 2583static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2584 ext4_fsblk_t sb_block,
bfff6873 2585 ext4_group_t *first_not_zeroed)
ac27a0ec 2586{
617ba13b
MC		 2587 struct ext4_sb_info *sbi = EXT4_SB(sb);
2588 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2589 ext4_fsblk_t last_block;
44de022c 2590 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
bd81d8ee
LV		 2591 ext4_fsblk_t block_bitmap;
2592 ext4_fsblk_t inode_bitmap;
2593 ext4_fsblk_t inode_table;
ce421581 2594 int flexbg_flag = 0;
bfff6873 2595 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2596
e2b911c5 2597 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2598 flexbg_flag = 1;
2599
af5bc92d 2600 ext4_debug("Checking group descriptors");
ac27a0ec 2601
197cd65a
AM		 2602 for (i = 0; i < sbi->s_groups_count; i++) {
2603 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2604
ce421581 2605 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2606 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2607 else
2608 last_block = first_block +
617ba13b 2609 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2610
bfff6873
LC		 2611 if ((grp == sbi->s_groups_count) &&
2612 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2613 grp = i;
2614
8fadc143 2615 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2616 if (block_bitmap == sb_block) {
2617 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2618 "Block bitmap for group %u overlaps "
2619 "superblock", i);
18db4b4e
TT		 2620 if (!sb_rdonly(sb))
2621 return 0;
829fa70d 2622 }
77260807
TT		 2623 if (block_bitmap >= sb_block + 1 &&
2624 block_bitmap <= last_bg_block) {
2625 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2626 "Block bitmap for group %u overlaps "
2627 "block group descriptors", i);
2628 if (!sb_rdonly(sb))
2629 return 0;
2630 }
2b2d6d01 2631 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2632 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2633 "Block bitmap for group %u not in group "
b31e1552 2634 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2635 return 0;
2636 }
8fadc143 2637 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2638 if (inode_bitmap == sb_block) {
2639 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2640 "Inode bitmap for group %u overlaps "
2641 "superblock", i);
18db4b4e
TT		 2642 if (!sb_rdonly(sb))
2643 return 0;
829fa70d 2644 }
77260807
TT		 2645 if (inode_bitmap >= sb_block + 1 &&
2646 inode_bitmap <= last_bg_block) {
2647 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2648 "Inode bitmap for group %u overlaps "
2649 "block group descriptors", i);
2650 if (!sb_rdonly(sb))
2651 return 0;
2652 }
2b2d6d01 2653 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2654 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2655 "Inode bitmap for group %u not in group "
b31e1552 2656 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2657 return 0;
2658 }
8fadc143 2659 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2660 if (inode_table == sb_block) {
2661 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2662 "Inode table for group %u overlaps "
2663 "superblock", i);
18db4b4e
TT		 2664 if (!sb_rdonly(sb))
2665 return 0;
829fa70d 2666 }
77260807
TT		 2667 if (inode_table >= sb_block + 1 &&
2668 inode_table <= last_bg_block) {
2669 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2670 "Inode table for group %u overlaps "
2671 "block group descriptors", i);
2672 if (!sb_rdonly(sb))
2673 return 0;
2674 }
bd81d8ee 2675 if (inode_table < first_block ||
2b2d6d01 2676 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2677 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2678 "Inode table for group %u not in group "
b31e1552 2679 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2680 return 0;
2681 }
955ce5f5 2682 ext4_lock_group(sb, i);
feb0ab32 2683 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2684 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2685 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2686 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2687 gdp)), le16_to_cpu(gdp->bg_checksum));
bc98a42c 2688 if (!sb_rdonly(sb)) {
955ce5f5 2689 ext4_unlock_group(sb, i);
8a266467 2690 return 0;
7ee1ec4c 2691 }
717d50e4 2692 }
955ce5f5 2693 ext4_unlock_group(sb, i);
ce421581
JS		 2694 if (!flexbg_flag)
2695 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2696 }
bfff6873
LC		 2697 if (NULL != first_not_zeroed)
2698 *first_not_zeroed = grp;
ac27a0ec
DK		 2699 return 1;
2700}
2701
617ba13b 2702/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2703 * the superblock) which were deleted from all directories, but held open by
2704 * a process at the time of a crash. We walk the list and try to delete these
2705 * inodes at recovery time (only with a read-write filesystem).
2706 *
2707 * In order to keep the orphan inode chain consistent during traversal (in
2708 * case of crash during recovery), we link each inode into the superblock
2709 * orphan list_head and handle it the same way as an inode deletion during
2710 * normal operation (which journals the operations for us).
2711 *
2712 * We only do an iget() and an iput() on each inode, which is very safe if we
2713 * accidentally point at an in-use or already deleted inode. The worst that
2714 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2715 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2716 * e2fsck was run on this filesystem, and it must have already done the orphan
2717 * inode cleanup for us, so we can safely abort without any further action.
2718 */
2b2d6d01
TT		 2719static void ext4_orphan_cleanup(struct super_block *sb,
2720 struct ext4_super_block *es)
ac27a0ec
DK		 2721{
2722 unsigned int s_flags = sb->s_flags;
2c98eb5e 2723 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec 2724#ifdef CONFIG_QUOTA
95f1fda4 2725 int quota_update = 0;
ac27a0ec
DK		 2726 int i;
2727#endif
2728 if (!es->s_last_orphan) {
2729 jbd_debug(4, "no orphan inodes to clean up\n");
2730 return;
2731 }
2732
a8f48a95 2733 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2734 ext4_msg(sb, KERN_ERR, "write access "
2735 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2736 return;
2737 }
2738
d39195c3
AG		 2739 /* Check if feature set would not allow a r/w mount */
2740 if (!ext4_feature_set_ok(sb, 0)) {
2741 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2742 "unknown ROCOMPAT features");
2743 return;
2744 }
2745
617ba13b 2746 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6 2747 /* don't clear list on RO mount w/ errors */
1751e8a6 2748 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
84474976 2749 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2750 "clearing orphan list.\n");
c25f9bc6
ES		 2751 es->s_last_orphan = 0;
2752 }
ac27a0ec
DK		 2753 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2754 return;
2755 }
2756
1751e8a6 2757 if (s_flags & SB_RDONLY) {
b31e1552 2758 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1751e8a6 2759 sb->s_flags &= ~SB_RDONLY;
ac27a0ec
DK		 2760 }
2761#ifdef CONFIG_QUOTA
2762 /* Needed for iput() to work correctly and not trash data */
1751e8a6 2763 sb->s_flags |= SB_ACTIVE;
95f1fda4 2764
2765 /*
2766 * Turn on quotas which were not enabled for read-only mounts if
2767 * filesystem has quota feature, so that they are updated correctly.
2768 */
1751e8a6 2769 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
95f1fda4 2770 int ret = ext4_enable_quotas(sb);
2771
2772 if (!ret)
2773 quota_update = 1;
2774 else
2775 ext4_msg(sb, KERN_ERR,
2776 "Cannot turn on quotas: error %d", ret);
2777 }
2778
2779 /* Turn on journaled quotas used for old sytle */
a2d4a646 2780 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2781 if (EXT4_SB(sb)->s_qf_names[i]) {
2782 int ret = ext4_quota_on_mount(sb, i);
95f1fda4 2783
2784 if (!ret)
2785 quota_update = 1;
2786 else
b31e1552
ES		 2787 ext4_msg(sb, KERN_ERR,
2788 "Cannot turn on journaled "
95f1fda4 2789 "quota: type %d: error %d", i, ret);
ac27a0ec
DK		 2790 }
2791 }
2792#endif
2793
2794 while (es->s_last_orphan) {
2795 struct inode *inode;
2796
c65d5c6c
VN		 2797 /*
2798 * We may have encountered an error during cleanup; if
2799 * so, skip the rest.
2800 */
2801 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2802 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2803 es->s_last_orphan = 0;
2804 break;
2805 }
2806
97bd42b9
JB		 2807 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2808 if (IS_ERR(inode)) {
ac27a0ec
DK		 2809 es->s_last_orphan = 0;
2810 break;
2811 }
2812
617ba13b 2813 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2814 dquot_initialize(inode);
ac27a0ec 2815 if (inode->i_nlink) {
566370a2
PT		 2816 if (test_opt(sb, DEBUG))
2817 ext4_msg(sb, KERN_DEBUG,
2818 "%s: truncating inode %lu to %lld bytes",
2819 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2820 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2821 inode->i_ino, inode->i_size);
5955102c 2822 inode_lock(inode);
55f252c9 2823 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 2824 ret = ext4_truncate(inode);
2825 if (ret)
2826 ext4_std_error(inode->i_sb, ret);
5955102c 2827 inode_unlock(inode);
ac27a0ec
DK		 2828 nr_truncates++;
2829 } else {
566370a2
PT		 2830 if (test_opt(sb, DEBUG))
2831 ext4_msg(sb, KERN_DEBUG,
2832 "%s: deleting unreferenced inode %lu",
2833 __func__, inode->i_ino);
ac27a0ec
DK		 2834 jbd_debug(2, "deleting unreferenced inode %lu\n",
2835 inode->i_ino);
2836 nr_orphans++;
2837 }
2838 iput(inode); /* The delete magic happens here! */
2839 }
2840
2b2d6d01 2841#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2842
2843 if (nr_orphans)
b31e1552
ES		 2844 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2845 PLURAL(nr_orphans));
ac27a0ec 2846 if (nr_truncates)
b31e1552
ES		 2847 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2848 PLURAL(nr_truncates));
ac27a0ec 2849#ifdef CONFIG_QUOTA
95f1fda4 2850 /* Turn off quotas if they were enabled for orphan cleanup */
2851 if (quota_update) {
2852 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2853 if (sb_dqopt(sb)->files[i])
2854 dquot_quota_off(sb, i);
2855 }
ac27a0ec
DK		 2856 }
2857#endif
1751e8a6 2858 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
ac27a0ec 2859}
0b8e58a1 2860
cd2291a4
ES		 2861/*
2862 * Maximal extent format file size.
2863 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2864 * extent format containers, within a sector_t, and within i_blocks
2865 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2866 * so that won't be a limiting factor.
2867 *
f17722f9
LC		 2868 * However there is other limiting factor. We do store extents in the form
2869 * of starting block and length, hence the resulting length of the extent
2870 * covering maximum file size must fit into on-disk format containers as
2871 * well. Given that length is always by 1 unit bigger than max unit (because
2872 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2873 *
cd2291a4
ES		 2874 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2875 */
f287a1a5 2876static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2877{
2878 loff_t res;
2879 loff_t upper_limit = MAX_LFS_FILESIZE;
2880
72deb455
CH		 2881 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
2882
2883 if (!has_huge_files) {
cd2291a4
ES		 2884 upper_limit = (1LL << 32) - 1;
2885
2886 /* total blocks in file system block size */
2887 upper_limit >>= (blkbits - 9);
2888 upper_limit <<= blkbits;
2889 }
2890
f17722f9
LC		 2891 /*
2892 * 32-bit extent-start container, ee_block. We lower the maxbytes
2893 * by one fs block, so ee_len can cover the extent of maximum file
2894 * size
2895 */
2896 res = (1LL << 32) - 1;
cd2291a4 2897 res <<= blkbits;
cd2291a4
ES		 2898
2899 /* Sanity check against vm- & vfs- imposed limits */
2900 if (res > upper_limit)
2901 res = upper_limit;
2902
2903 return res;
2904}
ac27a0ec 2905
ac27a0ec 2906/*
cd2291a4 2907 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2908 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2909 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2910 */
f287a1a5 2911static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2912{
617ba13b 2913 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2914 int meta_blocks;
2915 loff_t upper_limit;
0b8e58a1
AD		 2916 /* This is calculated to be the largest file size for a dense, block
2917 * mapped file such that the file's total number of 512-byte sectors,
2918 * including data and all indirect blocks, does not exceed (2^48 - 1).
2919 *
2920 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2921 * number of 512-byte sectors of the file.
0fc1b451
AK		 2922 */
2923
72deb455 2924 if (!has_huge_files) {
0fc1b451 2925 /*
72deb455
CH		 2926 * !has_huge_files or implies that the inode i_block field
2927 * represents total file blocks in 2^32 512-byte sectors ==
2928 * size of vfs inode i_blocks * 8
0fc1b451
AK		 2929 */
2930 upper_limit = (1LL << 32) - 1;
2931
2932 /* total blocks in file system block size */
2933 upper_limit >>= (bits - 9);
2934
2935 } else {
8180a562
AK		 2936 /*
2937 * We use 48 bit ext4_inode i_blocks
2938 * With EXT4_HUGE_FILE_FL set the i_blocks
2939 * represent total number of blocks in
2940 * file system block size
2941 */
0fc1b451
AK		 2942 upper_limit = (1LL << 48) - 1;
2943
0fc1b451
AK		 2944 }
2945
2946 /* indirect blocks */
2947 meta_blocks = 1;
2948 /* double indirect blocks */
2949 meta_blocks += 1 + (1LL << (bits-2));
2950 /* tripple indirect blocks */
2951 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2952
2953 upper_limit -= meta_blocks;
2954 upper_limit <<= bits;
ac27a0ec
DK		 2955
2956 res += 1LL << (bits-2);
2957 res += 1LL << (2*(bits-2));
2958 res += 1LL << (3*(bits-2));
2959 res <<= bits;
2960 if (res > upper_limit)
2961 res = upper_limit;
0fc1b451
AK		 2962
2963 if (res > MAX_LFS_FILESIZE)
2964 res = MAX_LFS_FILESIZE;
2965
ac27a0ec
DK		 2966 return res;
2967}
2968
617ba13b 2969static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2970 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2971{
617ba13b 2972 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2973 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2974 int has_super = 0;
2975
2976 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2977
e2b911c5 2978 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2979 return logical_sb_block + nr + 1;
ac27a0ec 2980 bg = sbi->s_desc_per_block * nr;
617ba13b 2981 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2982 has_super = 1;
0b8e58a1 2983
bd63f6b0
DW		 2984 /*
2985 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2986 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2987 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2988 * compensate.
2989 */
2990 if (sb->s_blocksize == 1024 && nr == 0 &&
49598e04 2991 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
bd63f6b0
DW		 2992 has_super++;
2993
617ba13b 2994 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2995}
2996
c9de560d
AT		 2997/**
2998 * ext4_get_stripe_size: Get the stripe size.
2999 * @sbi: In memory super block info
3000 *
3001 * If we have specified it via mount option, then
3002 * use the mount option value. If the value specified at mount time is
3003 * greater than the blocks per group use the super block value.
3004 * If the super block value is greater than blocks per group return 0.
3005 * Allocator needs it be less than blocks per group.
3006 *
3007 */
3008static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3009{
3010 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3011 unsigned long stripe_width =
3012 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 3013 int ret;
c9de560d
AT		 3014
3015 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 3016 ret = sbi->s_stripe;
5469d7c3 3017 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 3018 ret = stripe_width;
5469d7c3 3019 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 3020 ret = stride;
3021 else
3022 ret = 0;
c9de560d 3023
3eb08658
DE		 3024 /*
3025 * If the stripe width is 1, this makes no sense and
3026 * we set it to 0 to turn off stripe handling code.
3027 */
3028 if (ret <= 1)
3029 ret = 0;
c9de560d 3030
3eb08658 3031 return ret;
c9de560d 3032}
ac27a0ec 3033
a13fb1a4
ES		 3034/*
3035 * Check whether this filesystem can be mounted based on
3036 * the features present and the RDONLY/RDWR mount requested.
3037 * Returns 1 if this filesystem can be mounted as requested,
3038 * 0 if it cannot be.
3039 */
3040static int ext4_feature_set_ok(struct super_block *sb, int readonly)
3041{
e2b911c5 3042 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 3043 ext4_msg(sb, KERN_ERR,
3044 "Couldn't mount because of "
3045 "unsupported optional features (%x)",
3046 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3047 ~EXT4_FEATURE_INCOMPAT_SUPP));
3048 return 0;
3049 }
3050
c83ad55e
GKB		 3051#ifndef CONFIG_UNICODE
3052 if (ext4_has_feature_casefold(sb)) {
3053 ext4_msg(sb, KERN_ERR,
3054 "Filesystem with casefold feature cannot be "
3055 "mounted without CONFIG_UNICODE");
3056 return 0;
3057 }
3058#endif
3059
a13fb1a4
ES		 3060 if (readonly)
3061 return 1;
3062
e2b911c5 3063 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b 3064 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
1751e8a6 3065 sb->s_flags |= SB_RDONLY;
2cb5cc8b
DW		 3066 return 1;
3067 }
3068
a13fb1a4 3069 /* Check that feature set is OK for a read-write mount */
e2b911c5 3070 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 3071 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3072 "unsupported optional features (%x)",
3073 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3074 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3075 return 0;
3076 }
e2b911c5 3077 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 3078 ext4_msg(sb, KERN_ERR,
3079 "Can't support bigalloc feature without "
3080 "extents feature\n");
3081 return 0;
3082 }
7c319d32 3083
9db176bc 3084#if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
d65d87a0
TT		 3085 if (!readonly && (ext4_has_feature_quota(sb) ||
3086 ext4_has_feature_project(sb))) {
7c319d32 3087 ext4_msg(sb, KERN_ERR,
d65d87a0 3088 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
689c958c
LX		 3089 return 0;
3090 }
7c319d32 3091#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 3092 return 1;
3093}
3094
66e61a9e
TT		 3095/*
3096 * This function is called once a day if we have errors logged
3097 * on the file system
3098 */
235699a8 3099static void print_daily_error_info(struct timer_list *t)
66e61a9e 3100{
235699a8
KC		 3101 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3102 struct super_block *sb = sbi->s_sb;
3103 struct ext4_super_block *es = sbi->s_es;
66e61a9e
TT		 3104
3105 if (es->s_error_count)
ae0f78de
TT		 3106 /* fsck newer than v1.41.13 is needed to clean this condition. */
3107 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 3108 le32_to_cpu(es->s_error_count));
3109 if (es->s_first_error_time) {
6a0678a7
AB		 3110 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3111 sb->s_id,
3112 ext4_get_tstamp(es, s_first_error_time),
66e61a9e
TT		 3113 (int) sizeof(es->s_first_error_func),
3114 es->s_first_error_func,
3115 le32_to_cpu(es->s_first_error_line));
3116 if (es->s_first_error_ino)
651e1c3b 3117 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3118 le32_to_cpu(es->s_first_error_ino));
3119 if (es->s_first_error_block)
651e1c3b 3120 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3121 le64_to_cpu(es->s_first_error_block));
651e1c3b 3122 printk(KERN_CONT "\n");
66e61a9e
TT		 3123 }
3124 if (es->s_last_error_time) {
6a0678a7
AB		 3125 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3126 sb->s_id,
3127 ext4_get_tstamp(es, s_last_error_time),
66e61a9e
TT		 3128 (int) sizeof(es->s_last_error_func),
3129 es->s_last_error_func,
3130 le32_to_cpu(es->s_last_error_line));
3131 if (es->s_last_error_ino)
651e1c3b 3132 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3133 le32_to_cpu(es->s_last_error_ino));
3134 if (es->s_last_error_block)
651e1c3b 3135 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3136 le64_to_cpu(es->s_last_error_block));
651e1c3b 3137 printk(KERN_CONT "\n");
66e61a9e
TT		 3138 }
3139 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3140}
3141
bfff6873
LC		 3142/* Find next suitable group and run ext4_init_inode_table */
3143static int ext4_run_li_request(struct ext4_li_request *elr)
3144{
3145 struct ext4_group_desc *gdp = NULL;
3146 ext4_group_t group, ngroups;
3147 struct super_block *sb;
3148 unsigned long timeout = 0;
3149 int ret = 0;
3150
3151 sb = elr->lr_super;
3152 ngroups = EXT4_SB(sb)->s_groups_count;
3153
3154 for (group = elr->lr_next_group; group < ngroups; group++) {
3155 gdp = ext4_get_group_desc(sb, group, NULL);
3156 if (!gdp) {
3157 ret = 1;
3158 break;
3159 }
3160
3161 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3162 break;
3163 }
3164
7f511862 3165 if (group >= ngroups)
bfff6873
LC		 3166 ret = 1;
3167
3168 if (!ret) {
3169 timeout = jiffies;
3170 ret = ext4_init_inode_table(sb, group,
3171 elr->lr_timeout ? 0 : 1);
3172 if (elr->lr_timeout == 0) {
51ce6511
LC		 3173 timeout = (jiffies - timeout) *
3174 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 3175 elr->lr_timeout = timeout;
3176 }
3177 elr->lr_next_sched = jiffies + elr->lr_timeout;
3178 elr->lr_next_group = group + 1;
3179 }
bfff6873
LC		 3180 return ret;
3181}
3182
3183/*
3184 * Remove lr_request from the list_request and free the
4ed5c033 3185 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 3186 */
3187static void ext4_remove_li_request(struct ext4_li_request *elr)
3188{
3189 struct ext4_sb_info *sbi;
3190
3191 if (!elr)
3192 return;
3193
3194 sbi = elr->lr_sbi;
3195
3196 list_del(&elr->lr_request);
3197 sbi->s_li_request = NULL;
3198 kfree(elr);
3199}
3200
3201static void ext4_unregister_li_request(struct super_block *sb)
3202{
1bb933fb
LC		 3203 mutex_lock(&ext4_li_mtx);
3204 if (!ext4_li_info) {
3205 mutex_unlock(&ext4_li_mtx);
bfff6873 3206 return;
1bb933fb 3207 }
bfff6873
LC		 3208
3209 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 3210 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 3211 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 3212 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 3213}
3214
8f1f7453
ES		 3215static struct task_struct *ext4_lazyinit_task;
3216
bfff6873
LC		 3217/*
3218 * This is the function where ext4lazyinit thread lives. It walks
3219 * through the request list searching for next scheduled filesystem.
3220 * When such a fs is found, run the lazy initialization request
3221 * (ext4_rn_li_request) and keep track of the time spend in this
3222 * function. Based on that time we compute next schedule time of
3223 * the request. When walking through the list is complete, compute
3224 * next waking time and put itself into sleep.
3225 */
3226static int ext4_lazyinit_thread(void *arg)
3227{
3228 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3229 struct list_head *pos, *n;
3230 struct ext4_li_request *elr;
4ed5c033 3231 unsigned long next_wakeup, cur;
bfff6873
LC		 3232
3233 BUG_ON(NULL == eli);
3234
bfff6873
LC		 3235cont_thread:
3236 while (true) {
3237 next_wakeup = MAX_JIFFY_OFFSET;
3238
3239 mutex_lock(&eli->li_list_mtx);
3240 if (list_empty(&eli->li_request_list)) {
3241 mutex_unlock(&eli->li_list_mtx);
3242 goto exit_thread;
3243 }
bfff6873 3244 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 3245 int err = 0;
3246 int progress = 0;
bfff6873
LC		 3247 elr = list_entry(pos, struct ext4_li_request,
3248 lr_request);
3249
e22834f0
DM		 3250 if (time_before(jiffies, elr->lr_next_sched)) {
3251 if (time_before(elr->lr_next_sched, next_wakeup))
3252 next_wakeup = elr->lr_next_sched;
3253 continue;
3254 }
3255 if (down_read_trylock(&elr->lr_super->s_umount)) {
3256 if (sb_start_write_trylock(elr->lr_super)) {
3257 progress = 1;
3258 /*
3259 * We hold sb->s_umount, sb can not
3260 * be removed from the list, it is
3261 * now safe to drop li_list_mtx
3262 */
3263 mutex_unlock(&eli->li_list_mtx);
3264 err = ext4_run_li_request(elr);
3265 sb_end_write(elr->lr_super);
3266 mutex_lock(&eli->li_list_mtx);
3267 n = pos->next;
b2c78cd0 3268 }
e22834f0
DM		 3269 up_read((&elr->lr_super->s_umount));
3270 }
3271 /* error, remove the lazy_init job */
3272 if (err) {
3273 ext4_remove_li_request(elr);
3274 continue;
3275 }
3276 if (!progress) {
3277 elr->lr_next_sched = jiffies +
3278 (prandom_u32()
3279 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 3280 }
bfff6873
LC		 3281 if (time_before(elr->lr_next_sched, next_wakeup))
3282 next_wakeup = elr->lr_next_sched;
3283 }
3284 mutex_unlock(&eli->li_list_mtx);
3285
a0acae0e 3286 try_to_freeze();
bfff6873 3287
4ed5c033
LC		 3288 cur = jiffies;
3289 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3290 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3291 cond_resched();
3292 continue;
3293 }
3294
4ed5c033
LC		 3295 schedule_timeout_interruptible(next_wakeup - cur);
3296
8f1f7453
ES		 3297 if (kthread_should_stop()) {
3298 ext4_clear_request_list();
3299 goto exit_thread;
3300 }
bfff6873
LC		 3301 }
3302
3303exit_thread:
3304 /*
3305 * It looks like the request list is empty, but we need
3306 * to check it under the li_list_mtx lock, to prevent any
3307 * additions into it, and of course we should lock ext4_li_mtx
3308 * to atomically free the list and ext4_li_info, because at
3309 * this point another ext4 filesystem could be registering
3310 * new one.
3311 */
3312 mutex_lock(&ext4_li_mtx);
3313 mutex_lock(&eli->li_list_mtx);
3314 if (!list_empty(&eli->li_request_list)) {
3315 mutex_unlock(&eli->li_list_mtx);
3316 mutex_unlock(&ext4_li_mtx);
3317 goto cont_thread;
3318 }
3319 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3320 kfree(ext4_li_info);
3321 ext4_li_info = NULL;
3322 mutex_unlock(&ext4_li_mtx);
3323
3324 return 0;
3325}
3326
3327static void ext4_clear_request_list(void)
3328{
3329 struct list_head *pos, *n;
3330 struct ext4_li_request *elr;
3331
3332 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3333 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3334 elr = list_entry(pos, struct ext4_li_request,
3335 lr_request);
3336 ext4_remove_li_request(elr);
3337 }
3338 mutex_unlock(&ext4_li_info->li_list_mtx);
3339}
3340
3341static int ext4_run_lazyinit_thread(void)
3342{
8f1f7453
ES		 3343 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3344 ext4_li_info, "ext4lazyinit");
3345 if (IS_ERR(ext4_lazyinit_task)) {
3346 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3347 ext4_clear_request_list();
bfff6873
LC		 3348 kfree(ext4_li_info);
3349 ext4_li_info = NULL;
92b97816 3350 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3351 "initialization thread\n",
3352 err);
3353 return err;
3354 }
3355 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3356 return 0;
3357}
3358
3359/*
3360 * Check whether it make sense to run itable init. thread or not.
3361 * If there is at least one uninitialized inode table, return
3362 * corresponding group number, else the loop goes through all
3363 * groups and return total number of groups.
3364 */
3365static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3366{
3367 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3368 struct ext4_group_desc *gdp = NULL;
3369
8844618d
TT		 3370 if (!ext4_has_group_desc_csum(sb))
3371 return ngroups;
3372
bfff6873
LC		 3373 for (group = 0; group < ngroups; group++) {
3374 gdp = ext4_get_group_desc(sb, group, NULL);
3375 if (!gdp)
3376 continue;
3377
50122847 3378 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
bfff6873
LC		 3379 break;
3380 }
3381
3382 return group;
3383}
3384
3385static int ext4_li_info_new(void)
3386{
3387 struct ext4_lazy_init *eli = NULL;
3388
3389 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3390 if (!eli)
3391 return -ENOMEM;
3392
bfff6873
LC		 3393 INIT_LIST_HEAD(&eli->li_request_list);
3394 mutex_init(&eli->li_list_mtx);
3395
bfff6873
LC		 3396 eli->li_state |= EXT4_LAZYINIT_QUIT;
3397
3398 ext4_li_info = eli;
3399
3400 return 0;
3401}
3402
3403static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3404 ext4_group_t start)
3405{
3406 struct ext4_sb_info *sbi = EXT4_SB(sb);
3407 struct ext4_li_request *elr;
bfff6873
LC		 3408
3409 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3410 if (!elr)
3411 return NULL;
3412
3413 elr->lr_super = sb;
3414 elr->lr_sbi = sbi;
3415 elr->lr_next_group = start;
3416
3417 /*
3418 * Randomize first schedule time of the request to
3419 * spread the inode table initialization requests
3420 * better.
3421 */
dd1f723b
TT		 3422 elr->lr_next_sched = jiffies + (prandom_u32() %
3423 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3424 return elr;
3425}
3426
7f511862
TT		 3427int ext4_register_li_request(struct super_block *sb,
3428 ext4_group_t first_not_zeroed)
bfff6873
LC		 3429{
3430 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3431 struct ext4_li_request *elr = NULL;
49598e04 3432 ext4_group_t ngroups = sbi->s_groups_count;
6c5a6cb9 3433 int ret = 0;
bfff6873 3434
7f511862 3435 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3436 if (sbi->s_li_request != NULL) {
3437 /*
3438 * Reset timeout so it can be computed again, because
3439 * s_li_wait_mult might have changed.
3440 */
3441 sbi->s_li_request->lr_timeout = 0;
7f511862 3442 goto out;
51ce6511 3443 }
bfff6873 3444
bc98a42c 3445 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
55ff3840 3446 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3447 goto out;
bfff6873
LC		 3448
3449 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3450 if (!elr) {
3451 ret = -ENOMEM;
3452 goto out;
3453 }
bfff6873
LC		 3454
3455 if (NULL == ext4_li_info) {
3456 ret = ext4_li_info_new();
3457 if (ret)
3458 goto out;
3459 }
3460
3461 mutex_lock(&ext4_li_info->li_list_mtx);
3462 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3463 mutex_unlock(&ext4_li_info->li_list_mtx);
3464
3465 sbi->s_li_request = elr;
46e4690b
TM		 3466 /*
3467 * set elr to NULL here since it has been inserted to
3468 * the request_list and the removal and free of it is
3469 * handled by ext4_clear_request_list from now on.
3470 */
3471 elr = NULL;
bfff6873
LC		 3472
3473 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3474 ret = ext4_run_lazyinit_thread();
3475 if (ret)
3476 goto out;
3477 }
bfff6873 3478out:
beed5ecb
NK		 3479 mutex_unlock(&ext4_li_mtx);
3480 if (ret)
bfff6873 3481 kfree(elr);
bfff6873
LC		 3482 return ret;
3483}
3484
3485/*
3486 * We do not need to lock anything since this is called on
3487 * module unload.
3488 */
3489static void ext4_destroy_lazyinit_thread(void)
3490{
3491 /*
3492 * If thread exited earlier
3493 * there's nothing to be done.
3494 */
8f1f7453 3495 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3496 return;
3497
8f1f7453 3498 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3499}
3500
25ed6e8a
DW		 3501static int set_journal_csum_feature_set(struct super_block *sb)
3502{
3503 int ret = 1;
3504 int compat, incompat;
3505 struct ext4_sb_info *sbi = EXT4_SB(sb);
3506
9aa5d32b 3507 if (ext4_has_metadata_csum(sb)) {
db9ee220 3508 /* journal checksum v3 */
25ed6e8a 3509 compat = 0;
db9ee220 3510 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3511 } else {
3512 /* journal checksum v1 */
3513 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3514 incompat = 0;
3515 }
3516
feb8c6d3
DW		 3517 jbd2_journal_clear_features(sbi->s_journal,
3518 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3519 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3520 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3521 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3522 ret = jbd2_journal_set_features(sbi->s_journal,
3523 compat, 0,
3524 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3525 incompat);
3526 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3527 ret = jbd2_journal_set_features(sbi->s_journal,
3528 compat, 0,
3529 incompat);
3530 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3531 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3532 } else {
feb8c6d3
DW		 3533 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3534 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3535 }
3536
3537 return ret;
3538}
3539
952fc18e
TT		 3540/*
3541 * Note: calculating the overhead so we can be compatible with
3542 * historical BSD practice is quite difficult in the face of
3543 * clusters/bigalloc. This is because multiple metadata blocks from
3544 * different block group can end up in the same allocation cluster.
3545 * Calculating the exact overhead in the face of clustered allocation
3546 * requires either O(all block bitmaps) in memory or O(number of block
3547 * groups**2) in time. We will still calculate the superblock for
3548 * older file systems --- and if we come across with a bigalloc file
3549 * system with zero in s_overhead_clusters the estimate will be close to
3550 * correct especially for very large cluster sizes --- but for newer
3551 * file systems, it's better to calculate this figure once at mkfs
3552 * time, and store it in the superblock. If the superblock value is
3553 * present (even for non-bigalloc file systems), we will use it.
3554 */
3555static int count_overhead(struct super_block *sb, ext4_group_t grp,
3556 char *buf)
3557{
3558 struct ext4_sb_info *sbi = EXT4_SB(sb);
3559 struct ext4_group_desc *gdp;
3560 ext4_fsblk_t first_block, last_block, b;
3561 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3562 int s, j, count = 0;
3563
e2b911c5 3564 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3565 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3566 sbi->s_itb_per_group + 2);
3567
952fc18e
TT		 3568 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3569 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3570 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3571 for (i = 0; i < ngroups; i++) {
3572 gdp = ext4_get_group_desc(sb, i, NULL);
3573 b = ext4_block_bitmap(sb, gdp);
3574 if (b >= first_block && b <= last_block) {
3575 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3576 count++;
3577 }
3578 b = ext4_inode_bitmap(sb, gdp);
3579 if (b >= first_block && b <= last_block) {
3580 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3581 count++;
3582 }
3583 b = ext4_inode_table(sb, gdp);
3584 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3585 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3586 int c = EXT4_B2C(sbi, b - first_block);
3587 ext4_set_bit(c, buf);
3588 count++;
3589 }
3590 if (i != grp)
3591 continue;
3592 s = 0;
3593 if (ext4_bg_has_super(sb, grp)) {
3594 ext4_set_bit(s++, buf);
3595 count++;
3596 }
c48ae41b
TT		 3597 j = ext4_bg_num_gdb(sb, grp);
3598 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3599 ext4_error(sb, "Invalid number of block group "
3600 "descriptor blocks: %d", j);
3601 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3602 }
c48ae41b
TT		 3603 count += j;
3604 for (; j > 0; j--)
3605 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3606 }
3607 if (!count)
3608 return 0;
3609 return EXT4_CLUSTERS_PER_GROUP(sb) -
3610 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3611}
3612
3613/*
3614 * Compute the overhead and stash it in sbi->s_overhead
3615 */
3616int ext4_calculate_overhead(struct super_block *sb)
3617{
3618 struct ext4_sb_info *sbi = EXT4_SB(sb);
3619 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3620 struct inode *j_inode;
3621 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3622 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3623 ext4_fsblk_t overhead = 0;
4fdb5543 3624 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3625
952fc18e
TT		 3626 if (!buf)
3627 return -ENOMEM;
3628
3629 /*
3630 * Compute the overhead (FS structures). This is constant
3631 * for a given filesystem unless the number of block groups
3632 * changes so we cache the previous value until it does.
3633 */
3634
3635 /*
3636 * All of the blocks before first_data_block are overhead
3637 */
3638 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3639
3640 /*
3641 * Add the overhead found in each block group
3642 */
3643 for (i = 0; i < ngroups; i++) {
3644 int blks;
3645
3646 blks = count_overhead(sb, i, buf);
3647 overhead += blks;
3648 if (blks)
3649 memset(buf, 0, PAGE_SIZE);
3650 cond_resched();
3651 }
3c816ded
EW		 3652
3653 /*
3654 * Add the internal journal blocks whether the journal has been
3655 * loaded or not
3656 */
b003b524 3657 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3658 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
f1eec3b0
RH		 3659 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3660 /* j_inum for internal journal is non-zero */
3c816ded
EW		 3661 j_inode = ext4_get_journal_inode(sb, j_inum);
3662 if (j_inode) {
3663 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3664 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3665 iput(j_inode);
3666 } else {
3667 ext4_msg(sb, KERN_ERR, "can't get journal size");
3668 }
3669 }
952fc18e
TT		 3670 sbi->s_overhead = overhead;
3671 smp_wmb();
3672 free_page((unsigned long) buf);
3673 return 0;
3674}
3675
b5799018 3676static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3677{
3678 ext4_fsblk_t resv_clusters;
b5799018 3679 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3680
30fac0f7
JK		 3681 /*
3682 * There's no need to reserve anything when we aren't using extents.
3683 * The space estimates are exact, there are no unwritten extents,
3684 * hole punching doesn't need new metadata... This is needed especially
3685 * to keep ext2/3 backward compatibility.
3686 */
e2b911c5 3687 if (!ext4_has_feature_extents(sb))
b5799018 3688 return;
27dd4385
LC		 3689 /*
3690 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3691 * This should cover the situations where we can not afford to run
3692 * out of space like for example punch hole, or converting
556615dc 3693 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3694 * allocation would require 1, or 2 blocks, higher numbers are
3695 * very rare.
3696 */
b5799018
TT		 3697 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3698 sbi->s_cluster_bits);
27dd4385
LC		 3699
3700 do_div(resv_clusters, 50);
3701 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3702
b5799018 3703 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3704}
3705
2b2d6d01 3706static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3707{
5e405595 3708 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
d4c402d9 3709 char *orig_data = kstrdup(data, GFP_KERNEL);
1d0c3924 3710 struct buffer_head *bh, **group_desc;
617ba13b 3711 struct ext4_super_block *es = NULL;
5aee0f8a 3712 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
7c990728 3713 struct flex_groups **flex_groups;
617ba13b
MC		 3714 ext4_fsblk_t block;
3715 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3716 ext4_fsblk_t logical_sb_block;
ac27a0ec 3717 unsigned long offset = 0;
ac27a0ec
DK		 3718 unsigned long journal_devnum = 0;
3719 unsigned long def_mount_opts;
3720 struct inode *root;
0390131b 3721 const char *descr;
dcc7dae3 3722 int ret = -ENOMEM;
281b5995 3723 int blocksize, clustersize;
4ec11028
TT		 3724 unsigned int db_count;
3725 unsigned int i;
ef5fd681 3726 int needs_recovery, has_huge_files;
bd81d8ee 3727 __u64 blocks_count;
07aa2ea1 3728 int err = 0;
b3881f74 3729 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3730 ext4_group_t first_not_zeroed;
ac27a0ec 3731
5aee0f8a
TT		 3732 if ((data && !orig_data) || !sbi)
3733 goto out_free_base;
705895b6 3734
aed9eb1b 3735 sbi->s_daxdev = dax_dev;
705895b6
PE		 3736 sbi->s_blockgroup_lock =
3737 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 3738 if (!sbi->s_blockgroup_lock)
3739 goto out_free_base;
3740
ac27a0ec 3741 sb->s_fs_info = sbi;
2c0544b2 3742 sbi->s_sb = sb;
240799cd 3743 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3744 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3745 if (sb->s_bdev->bd_part)
3746 sbi->s_sectors_written_start =
dbae2c55 3747 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
ac27a0ec 3748
9f6200bb 3749 /* Cleanup superblock name */
ec3904dc 3750 strreplace(sb->s_id, '/', '!');
9f6200bb 3751
07aa2ea1 3752 /* -EINVAL is default */
dcc7dae3 3753 ret = -EINVAL;
617ba13b 3754 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3755 if (!blocksize) {
b31e1552 3756 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3757 goto out_fail;
3758 }
3759
3760 /*
617ba13b 3761 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3762 * block sizes. We need to calculate the offset from buffer start.
3763 */
617ba13b 3764 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3765 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3766 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3767 } else {
70bbb3e0 3768 logical_sb_block = sb_block;
ac27a0ec
DK		 3769 }
3770
a8ac900b 3771 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3772 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3773 goto out_fail;
3774 }
3775 /*
3776 * Note: s_es must be initialized as soon as possible because
617ba13b 3777 * some ext4 macro-instructions depend on its value
ac27a0ec 3778 */
2716b802 3779 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3780 sbi->s_es = es;
3781 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3782 if (sb->s_magic != EXT4_SUPER_MAGIC)
3783 goto cantfind_ext4;
afc32f7e 3784 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3785
feb0ab32 3786 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3787 if (ext4_has_feature_metadata_csum(sb) &&
3788 ext4_has_feature_gdt_csum(sb))
363307e6 3789 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3790 "redundant flags; please run fsck.");
3791
d25425f8
DW		 3792 /* Check for a known checksum algorithm */
3793 if (!ext4_verify_csum_type(sb, es)) {
3794 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3795 "unknown checksum algorithm.");
3796 silent = 1;
3797 goto cantfind_ext4;
3798 }
3799
0441984a 3800 /* Load the checksum driver */
a45403b5
TT		 3801 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3802 if (IS_ERR(sbi->s_chksum_driver)) {
3803 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3804 ret = PTR_ERR(sbi->s_chksum_driver);
3805 sbi->s_chksum_driver = NULL;
3806 goto failed_mount;
0441984a
DW		 3807 }
3808
a9c47317
DW		 3809 /* Check superblock checksum */
3810 if (!ext4_superblock_csum_verify(sb, es)) {
3811 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3812 "invalid superblock checksum. Run e2fsck?");
3813 silent = 1;
6a797d27 3814 ret = -EFSBADCRC;
a9c47317
DW		 3815 goto cantfind_ext4;
3816 }
3817
3818 /* Precompute checksum seed for all metadata */
e2b911c5 3819 if (ext4_has_feature_csum_seed(sb))
8c81bd8f 3820 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
dec214d0 3821 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
a9c47317
DW		 3822 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3823 sizeof(es->s_uuid));
3824
ac27a0ec
DK		 3825 /* Set defaults before we parse the mount options */
3826 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3827 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3828 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3829 set_opt(sb, DEBUG);
87f26807 3830 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3831 set_opt(sb, GRPID);
617ba13b 3832 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3833 set_opt(sb, NO_UID32);
ea663336 3834 /* xattr user namespace & acls are now defaulted on */
ea663336 3835 set_opt(sb, XATTR_USER);
03010a33 3836#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3837 set_opt(sb, POSIX_ACL);
2e7842b8 3838#endif
98c1a759
DW		 3839 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3840 if (ext4_has_metadata_csum(sb))
3841 set_opt(sb, JOURNAL_CHECKSUM);
3842
617ba13b 3843 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3844 set_opt(sb, JOURNAL_DATA);
617ba13b 3845 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3846 set_opt(sb, ORDERED_DATA);
617ba13b 3847 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3848 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3849
3850 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3851 set_opt(sb, ERRORS_PANIC);
bb4f397a 3852 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3853 set_opt(sb, ERRORS_CONT);
bb4f397a 3854 else
fd8c37ec 3855 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3856 /* block_validity enabled by default; disable with noblock_validity */
3857 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3858 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3859 set_opt(sb, DISCARD);
ac27a0ec 3860
08cefc7a
EB		 3861 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3862 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3863 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3864 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3865 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3866
8b67f04a 3867 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3868 set_opt(sb, BARRIER);
ac27a0ec 3869
dd919b98
AK		 3870 /*
3871 * enable delayed allocation by default
3872 * Use -o nodelalloc to turn it off
3873 */
bc0b75f7 3874 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3875 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3876 set_opt(sb, DELALLOC);
dd919b98 3877
51ce6511
LC		 3878 /*
3879 * set default s_li_wait_mult for lazyinit, for the case there is
3880 * no mount option specified.
3881 */
3882 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3883
4f97a681 3884 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
626b035b
RH		 3885
3886 if (blocksize == PAGE_SIZE)
3887 set_opt(sb, DIOREAD_NOLOCK);
3888
4f97a681
TT		 3889 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3890 blocksize > EXT4_MAX_BLOCK_SIZE) {
3891 ext4_msg(sb, KERN_ERR,
3892 "Unsupported filesystem blocksize %d (%d log_block_size)",
3893 blocksize, le32_to_cpu(es->s_log_block_size));
3894 goto failed_mount;
3895 }
3896
9803387c
TT		 3897 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3898 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3899 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3900 } else {
3901 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3902 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3903 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3904 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3905 sbi->s_first_ino);
3906 goto failed_mount;
3907 }
3908 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3909 (!is_power_of_2(sbi->s_inode_size)) ||
3910 (sbi->s_inode_size > blocksize)) {
3911 ext4_msg(sb, KERN_ERR,
3912 "unsupported inode size: %d",
3913 sbi->s_inode_size);
4f97a681 3914 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
9803387c
TT		 3915 goto failed_mount;
3916 }
3917 /*
3918 * i_atime_extra is the last extra field available for
3919 * [acm]times in struct ext4_inode. Checking for that
3920 * field should suffice to ensure we have extra space
3921 * for all three.
3922 */
3923 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
3924 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
3925 sb->s_time_gran = 1;
3926 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
3927 } else {
3928 sb->s_time_gran = NSEC_PER_SEC;
3929 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
3930 }
3931 sb->s_time_min = EXT4_TIMESTAMP_MIN;
3932 }
3933 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3934 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3935 EXT4_GOOD_OLD_INODE_SIZE;
3936 if (ext4_has_feature_extra_isize(sb)) {
3937 unsigned v, max = (sbi->s_inode_size -
3938 EXT4_GOOD_OLD_INODE_SIZE);
3939
3940 v = le16_to_cpu(es->s_want_extra_isize);
3941 if (v > max) {
3942 ext4_msg(sb, KERN_ERR,
3943 "bad s_want_extra_isize: %d", v);
3944 goto failed_mount;
3945 }
3946 if (sbi->s_want_extra_isize < v)
3947 sbi->s_want_extra_isize = v;
3948
3949 v = le16_to_cpu(es->s_min_extra_isize);
3950 if (v > max) {
3951 ext4_msg(sb, KERN_ERR,
3952 "bad s_min_extra_isize: %d", v);
3953 goto failed_mount;
3954 }
3955 if (sbi->s_want_extra_isize < v)
3956 sbi->s_want_extra_isize = v;
3957 }
3958 }
3959
5aee0f8a
TT		 3960 if (sbi->s_es->s_mount_opts[0]) {
3961 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3962 sizeof(sbi->s_es->s_mount_opts),
3963 GFP_KERNEL);
3964 if (!s_mount_opts)
3965 goto failed_mount;
3966 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3967 &journal_ioprio, 0)) {
3968 ext4_msg(sb, KERN_WARNING,
3969 "failed to parse options in superblock: %s",
3970 s_mount_opts);
3971 }
3972 kfree(s_mount_opts);
8b67f04a 3973 }
5a916be1 3974 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3975 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3976 &journal_ioprio, 0))
ac27a0ec
DK		 3977 goto failed_mount;
3978
c83ad55e
GKB		 3979#ifdef CONFIG_UNICODE
3980 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) {
3981 const struct ext4_sb_encodings *encoding_info;
3982 struct unicode_map *encoding;
3983 __u16 encoding_flags;
3984
3985 if (ext4_has_feature_encrypt(sb)) {
3986 ext4_msg(sb, KERN_ERR,
3987 "Can't mount with encoding and encryption");
3988 goto failed_mount;
3989 }
3990
3991 if (ext4_sb_read_encoding(es, &encoding_info,
3992 &encoding_flags)) {
3993 ext4_msg(sb, KERN_ERR,
3994 "Encoding requested by superblock is unknown");
3995 goto failed_mount;
3996 }
3997
3998 encoding = utf8_load(encoding_info->version);
3999 if (IS_ERR(encoding)) {
4000 ext4_msg(sb, KERN_ERR,
4001 "can't mount with superblock charset: %s-%s "
4002 "not supported by the kernel. flags: 0x%x.",
4003 encoding_info->name, encoding_info->version,
4004 encoding_flags);
4005 goto failed_mount;
4006 }
4007 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4008 "%s-%s with flags 0x%hx", encoding_info->name,
4009 encoding_info->version?:"\b", encoding_flags);
4010
4011 sbi->s_encoding = encoding;
4012 sbi->s_encoding_flags = encoding_flags;
4013 }
4014#endif
4015
56889787 4016 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
244adf64 4017 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!\n");
781c036b 4018 /* can't mount with both data=journal and dioread_nolock. */
244adf64 4019 clear_opt(sb, DIOREAD_NOLOCK);
56889787
TT		 4020 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4021 ext4_msg(sb, KERN_ERR, "can't mount with "
4022 "both data=journal and delalloc");
4023 goto failed_mount;
4024 }
fc626fe3 4025 if (test_opt(sb, DAX_ALWAYS)) {
923ae0ff
RZ		 4026 ext4_msg(sb, KERN_ERR, "can't mount with "
4027 "both data=journal and dax");
4028 goto failed_mount;
4029 }
73b92a2a
SK		 4030 if (ext4_has_feature_encrypt(sb)) {
4031 ext4_msg(sb, KERN_WARNING,
4032 "encrypted files will use data=ordered "
4033 "instead of data journaling mode");
4034 }
56889787
TT		 4035 if (test_opt(sb, DELALLOC))
4036 clear_opt(sb, DELALLOC);
001e4a87
TH		 4037 } else {
4038 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 4039 }
4040
1751e8a6
LT		 4041 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4042 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec 4043
617ba13b 4044 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 4045 (ext4_has_compat_features(sb) ||
4046 ext4_has_ro_compat_features(sb) ||
4047 ext4_has_incompat_features(sb)))
b31e1552
ES		 4048 ext4_msg(sb, KERN_WARNING,
4049 "feature flags set on rev 0 fs, "
4050 "running e2fsck is recommended");
469108ff 4051
ed3654eb
TT		 4052 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4053 set_opt2(sb, HURD_COMPAT);
e2b911c5 4054 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 4055 ext4_msg(sb, KERN_ERR,
4056 "The Hurd can't support 64-bit file systems");
4057 goto failed_mount;
4058 }
dec214d0
TE		 4059
4060 /*
4061 * ea_inode feature uses l_i_version field which is not
4062 * available in HURD_COMPAT mode.
4063 */
4064 if (ext4_has_feature_ea_inode(sb)) {
4065 ext4_msg(sb, KERN_ERR,
4066 "ea_inode feature is not supported for Hurd");
4067 goto failed_mount;
4068 }
ed3654eb
TT		 4069 }
4070
2035e776
TT		 4071 if (IS_EXT2_SB(sb)) {
4072 if (ext2_feature_set_ok(sb))
4073 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4074 "using the ext4 subsystem");
4075 else {
0d9366d6
ES		 4076 /*
4077 * If we're probing be silent, if this looks like
4078 * it's actually an ext[34] filesystem.
4079 */
4080 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4081 goto failed_mount;
2035e776
TT		 4082 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4083 "to feature incompatibilities");
4084 goto failed_mount;
4085 }
4086 }
4087
4088 if (IS_EXT3_SB(sb)) {
4089 if (ext3_feature_set_ok(sb))
4090 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4091 "using the ext4 subsystem");
4092 else {
0d9366d6
ES		 4093 /*
4094 * If we're probing be silent, if this looks like
4095 * it's actually an ext4 filesystem.
4096 */
4097 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4098 goto failed_mount;
2035e776
TT		 4099 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4100 "to feature incompatibilities");
4101 goto failed_mount;
4102 }
4103 }
4104
ac27a0ec
DK		 4105 /*
4106 * Check feature flags regardless of the revision level, since we
4107 * previously didn't change the revision level when setting the flags,
4108 * so there is a chance incompat flags are set on a rev 0 filesystem.
4109 */
bc98a42c 4110 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
ac27a0ec 4111 goto failed_mount;
a13fb1a4 4112
8cdf3372
TT		 4113 if (le32_to_cpu(es->s_log_block_size) >
4114 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4115 ext4_msg(sb, KERN_ERR,
4116 "Invalid log block size: %u",
4117 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 4118 goto failed_mount;
4119 }
bfe0a5f4
TT		 4120 if (le32_to_cpu(es->s_log_cluster_size) >
4121 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4122 ext4_msg(sb, KERN_ERR,
4123 "Invalid log cluster size: %u",
4124 le32_to_cpu(es->s_log_cluster_size));
4125 goto failed_mount;
4126 }
ac27a0ec 4127
5b9554dc
TT		 4128 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4129 ext4_msg(sb, KERN_ERR,
4130 "Number of reserved GDT blocks insanely large: %d",
4131 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4132 goto failed_mount;
4133 }
4134
a8ab6d38
IW		 4135 if (bdev_dax_supported(sb->s_bdev, blocksize))
4136 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4137
fc626fe3 4138 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
559db4c6
RZ		 4139 if (ext4_has_feature_inline_data(sb)) {
4140 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4141 " that may contain inline data");
361d24d4 4142 goto failed_mount;
559db4c6 4143 }
a8ab6d38 4144 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
24f3478d 4145 ext4_msg(sb, KERN_ERR,
361d24d4
ES		 4146 "DAX unsupported by block device.");
4147 goto failed_mount;
24f3478d 4148 }
923ae0ff
RZ		 4149 }
4150
e2b911c5 4151 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 4152 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4153 es->s_encryption_level);
4154 goto failed_mount;
4155 }
4156
ac27a0ec 4157 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 4158 /* Validate the filesystem blocksize */
4159 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 4160 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 4161 blocksize);
ac27a0ec
DK		 4162 goto failed_mount;
4163 }
4164
2b2d6d01 4165 brelse(bh);
70bbb3e0
AM		 4166 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4167 offset = do_div(logical_sb_block, blocksize);
a8ac900b 4168 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 4169 if (!bh) {
b31e1552
ES		 4170 ext4_msg(sb, KERN_ERR,
4171 "Can't read superblock on 2nd try");
ac27a0ec
DK		 4172 goto failed_mount;
4173 }
2716b802 4174 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 4175 sbi->s_es = es;
617ba13b 4176 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 4177 ext4_msg(sb, KERN_ERR,
4178 "Magic mismatch, very weird!");
ac27a0ec
DK		 4179 goto failed_mount;
4180 }
4181 }
4182
e2b911c5 4183 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 4184 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4185 has_huge_files);
4186 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 4187
0d1ee42f 4188 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 4189 if (ext4_has_feature_64bit(sb)) {
8fadc143 4190 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 4191 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 4192 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 4193 ext4_msg(sb, KERN_ERR,
4194 "unsupported descriptor size %lu",
0d1ee42f
AR		 4195 sbi->s_desc_size);
4196 goto failed_mount;
4197 }
4198 } else
4199 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 4200
ac27a0ec 4201 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 4202 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 4203
617ba13b 4204 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 4205 if (sbi->s_inodes_per_block == 0)
617ba13b 4206 goto cantfind_ext4;
cd6bb35b
TT		 4207 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4208 sbi->s_inodes_per_group > blocksize * 8) {
4209 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
b9c538da 4210 sbi->s_inodes_per_group);
cd6bb35b
TT		 4211 goto failed_mount;
4212 }
ac27a0ec
DK		 4213 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4214 sbi->s_inodes_per_block;
0d1ee42f 4215 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 4216 sbi->s_sbh = bh;
4217 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 4218 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4219 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 4220
2b2d6d01 4221 for (i = 0; i < 4; i++)
ac27a0ec
DK		 4222 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4223 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 4224 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 4225 i = le32_to_cpu(es->s_flags);
4226 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4227 sbi->s_hash_unsigned = 3;
4228 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 4229#ifdef __CHAR_UNSIGNED__
bc98a42c 4230 if (!sb_rdonly(sb))
23301410
TT		 4231 es->s_flags |=
4232 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4233 sbi->s_hash_unsigned = 3;
f99b2589 4234#else
bc98a42c 4235 if (!sb_rdonly(sb))
23301410
TT		 4236 es->s_flags |=
4237 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 4238#endif
23301410 4239 }
f99b2589 4240 }
ac27a0ec 4241
281b5995
TT		 4242 /* Handle clustersize */
4243 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
ef5fd681 4244 if (ext4_has_feature_bigalloc(sb)) {
281b5995
TT		 4245 if (clustersize < blocksize) {
4246 ext4_msg(sb, KERN_ERR,
4247 "cluster size (%d) smaller than "
4248 "block size (%d)", clustersize, blocksize);
4249 goto failed_mount;
4250 }
4251 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4252 le32_to_cpu(es->s_log_block_size);
4253 sbi->s_clusters_per_group =
4254 le32_to_cpu(es->s_clusters_per_group);
4255 if (sbi->s_clusters_per_group > blocksize * 8) {
4256 ext4_msg(sb, KERN_ERR,
4257 "#clusters per group too big: %lu",
4258 sbi->s_clusters_per_group);
4259 goto failed_mount;
4260 }
4261 if (sbi->s_blocks_per_group !=
4262 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4263 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4264 "clusters per group (%lu) inconsistent",
4265 sbi->s_blocks_per_group,
4266 sbi->s_clusters_per_group);
4267 goto failed_mount;
4268 }
4269 } else {
4270 if (clustersize != blocksize) {
bfe0a5f4
TT		 4271 ext4_msg(sb, KERN_ERR,
4272 "fragment/cluster size (%d) != "
4273 "block size (%d)", clustersize, blocksize);
4274 goto failed_mount;
281b5995
TT		 4275 }
4276 if (sbi->s_blocks_per_group > blocksize * 8) {
4277 ext4_msg(sb, KERN_ERR,
4278 "#blocks per group too big: %lu",
4279 sbi->s_blocks_per_group);
4280 goto failed_mount;
4281 }
4282 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4283 sbi->s_cluster_bits = 0;
ac27a0ec 4284 }
281b5995
TT		 4285 sbi->s_cluster_ratio = clustersize / blocksize;
4286
960fd856
TT		 4287 /* Do we have standard group size of clustersize * 8 blocks ? */
4288 if (sbi->s_blocks_per_group == clustersize << 3)
4289 set_opt2(sb, STD_GROUP_SIZE);
4290
bf43d84b
ES		 4291 /*
4292 * Test whether we have more sectors than will fit in sector_t,
4293 * and whether the max offset is addressable by the page cache.
4294 */
5a9ae68a 4295 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 4296 ext4_blocks_count(es));
5a9ae68a 4297 if (err) {
b31e1552 4298 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 4299 " too large to mount safely on this system");
ac27a0ec
DK		 4300 goto failed_mount;
4301 }
4302
617ba13b
MC		 4303 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4304 goto cantfind_ext4;
e7c95593 4305
0f2ddca6
FTN		 4306 /* check blocks count against device size */
4307 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4308 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 4309 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4310 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 4311 ext4_blocks_count(es), blocks_count);
4312 goto failed_mount;
4313 }
4314
0b8e58a1
AD		 4315 /*
4316 * It makes no sense for the first data block to be beyond the end
4317 * of the filesystem.
4318 */
4319 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 4320 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 4321 "block %u is beyond end of filesystem (%llu)",
4322 le32_to_cpu(es->s_first_data_block),
4323 ext4_blocks_count(es));
e7c95593
ES		 4324 goto failed_mount;
4325 }
bfe0a5f4
TT		 4326 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4327 (sbi->s_cluster_ratio == 1)) {
4328 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4329 "block is 0 with a 1k block and cluster size");
4330 goto failed_mount;
4331 }
4332
bd81d8ee
LV		 4333 blocks_count = (ext4_blocks_count(es) -
4334 le32_to_cpu(es->s_first_data_block) +
4335 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4336 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 4337 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
df41460a 4338 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4ec11028 4339 "(block count %llu, first data block %u, "
df41460a 4340 "blocks per group %lu)", blocks_count,
4ec11028
TT		 4341 ext4_blocks_count(es),
4342 le32_to_cpu(es->s_first_data_block),
4343 EXT4_BLOCKS_PER_GROUP(sb));
4344 goto failed_mount;
4345 }
bd81d8ee 4346 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 4347 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4348 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
9e463084
TT		 4349 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4350 le32_to_cpu(es->s_inodes_count)) {
4351 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4352 le32_to_cpu(es->s_inodes_count),
4353 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4354 ret = -EINVAL;
4355 goto failed_mount;
4356 }
617ba13b
MC		 4357 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4358 EXT4_DESC_PER_BLOCK(sb);
3a4b77cd 4359 if (ext4_has_feature_meta_bg(sb)) {
2ba3e6e8 4360 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3a4b77cd
EG		 4361 ext4_msg(sb, KERN_WARNING,
4362 "first meta block group too large: %u "
4363 "(group descriptor block count %u)",
4364 le32_to_cpu(es->s_first_meta_bg), db_count);
4365 goto failed_mount;
4366 }
4367 }
1d0c3924
TT		 4368 rcu_assign_pointer(sbi->s_group_desc,
4369 kvmalloc_array(db_count,
4370 sizeof(struct buffer_head *),
4371 GFP_KERNEL));
ac27a0ec 4372 if (sbi->s_group_desc == NULL) {
b31e1552 4373 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 4374 ret = -ENOMEM;
ac27a0ec
DK		 4375 goto failed_mount;
4376 }
4377
705895b6 4378 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec 4379
85c8f176
AP		 4380 /* Pre-read the descriptors into the buffer cache */
4381 for (i = 0; i < db_count; i++) {
4382 block = descriptor_loc(sb, logical_sb_block, i);
d87f6392 4383 sb_breadahead_unmovable(sb, block);
85c8f176
AP		 4384 }
4385
ac27a0ec 4386 for (i = 0; i < db_count; i++) {
1d0c3924
TT		 4387 struct buffer_head *bh;
4388
70bbb3e0 4389 block = descriptor_loc(sb, logical_sb_block, i);
1d0c3924
TT		 4390 bh = sb_bread_unmovable(sb, block);
4391 if (!bh) {
b31e1552
ES		 4392 ext4_msg(sb, KERN_ERR,
4393 "can't read group descriptor %d", i);
ac27a0ec
DK		 4394 db_count = i;
4395 goto failed_mount2;
4396 }
1d0c3924
TT		 4397 rcu_read_lock();
4398 rcu_dereference(sbi->s_group_desc)[i] = bh;
4399 rcu_read_unlock();
ac27a0ec 4400 }
44de022c 4401 sbi->s_gdb_count = db_count;
829fa70d 4402 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 4403 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 4404 ret = -EFSCORRUPTED;
f9ae9cf5 4405 goto failed_mount2;
ac27a0ec 4406 }
772cb7c8 4407
235699a8 4408 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
04496411 4409
a75ae78f 4410 /* Register extent status tree shrinker */
eb68d0e2 4411 if (ext4_es_register_shrinker(sbi))
ce7e010a 4412 goto failed_mount3;
ce7e010a 4413
c9de560d 4414 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 4415 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 4416
f9ae9cf5
TT		 4417 /*
4418 * set up enough so that it can read an inode
4419 */
f6e63f90 4420 sb->s_op = &ext4_sops;
617ba13b
MC		 4421 sb->s_export_op = &ext4_export_ops;
4422 sb->s_xattr = ext4_xattr_handlers;
643fa961 4423#ifdef CONFIG_FS_ENCRYPTION
a7550b30 4424 sb->s_cop = &ext4_cryptops;
ffcc4182 4425#endif
c93d8f88
EB		 4426#ifdef CONFIG_FS_VERITY
4427 sb->s_vop = &ext4_verityops;
4428#endif
ac27a0ec 4429#ifdef CONFIG_QUOTA
617ba13b 4430 sb->dq_op = &ext4_quota_operations;
e2b911c5 4431 if (ext4_has_feature_quota(sb))
1fa5efe3 4432 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 4433 else
4434 sb->s_qcop = &ext4_qctl_operations;
689c958c 4435 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 4436#endif
85787090 4437 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
f2fa2ffc 4438
ac27a0ec 4439 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 4440 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 4441
4442 sb->s_root = NULL;
4443
4444 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 4445 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 4446
bc98a42c 4447 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
c5e06d10 4448 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 4449 goto failed_mount3a;
c5e06d10 4450
ac27a0ec
DK		 4451 /*
4452 * The first inode we look at is the journal inode. Don't try
4453 * root first: it may be modified in the journal!
4454 */
e2b911c5 4455 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4753d8a2
TT		 4456 err = ext4_load_journal(sb, es, journal_devnum);
4457 if (err)
50460fe8 4458 goto failed_mount3a;
bc98a42c 4459 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
e2b911c5 4460 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 4461 ext4_msg(sb, KERN_ERR, "required journal recovery "
4462 "suppressed and not mounted read-only");
744692dc 4463 goto failed_mount_wq;
ac27a0ec 4464 } else {
1e381f60
DM		 4465 /* Nojournal mode, all journal mount options are illegal */
4466 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4467 ext4_msg(sb, KERN_ERR, "can't mount with "
4468 "journal_checksum, fs mounted w/o journal");
4469 goto failed_mount_wq;
4470 }
4471 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4472 ext4_msg(sb, KERN_ERR, "can't mount with "
4473 "journal_async_commit, fs mounted w/o journal");
4474 goto failed_mount_wq;
4475 }
4476 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4477 ext4_msg(sb, KERN_ERR, "can't mount with "
4478 "commit=%lu, fs mounted w/o journal",
4479 sbi->s_commit_interval / HZ);
4480 goto failed_mount_wq;
4481 }
4482 if (EXT4_MOUNT_DATA_FLAGS &
4483 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4484 ext4_msg(sb, KERN_ERR, "can't mount with "
4485 "data=, fs mounted w/o journal");
4486 goto failed_mount_wq;
4487 }
50b29d8f 4488 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
1e381f60 4489 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 4490 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 4491 sbi->s_journal = NULL;
4492 needs_recovery = 0;
4493 goto no_journal;
ac27a0ec
DK		 4494 }
4495
e2b911c5 4496 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 4497 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4498 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4499 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4500 goto failed_mount_wq;
eb40a09c
JS		 4501 }
4502
25ed6e8a
DW		 4503 if (!set_journal_csum_feature_set(sb)) {
4504 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4505 "feature set");
4506 goto failed_mount_wq;
d4da6c9c 4507 }
818d276c 4508
ac27a0ec
DK		 4509 /* We have now updated the journal if required, so we can
4510 * validate the data journaling mode. */
4511 switch (test_opt(sb, DATA_FLAGS)) {
4512 case 0:
4513 /* No mode set, assume a default based on the journal
63f57933
AM		 4514 * capabilities: ORDERED_DATA if the journal can
4515 * cope, else JOURNAL_DATA
4516 */
dab291af 4517 if (jbd2_journal_check_available_features
27f394a7 4518 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
fd8c37ec 4519 set_opt(sb, ORDERED_DATA);
27f394a7
TN		 4520 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4521 } else {
fd8c37ec 4522 set_opt(sb, JOURNAL_DATA);
27f394a7
TN		 4523 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4524 }
ac27a0ec
DK		 4525 break;
4526
617ba13b
MC		 4527 case EXT4_MOUNT_ORDERED_DATA:
4528 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4529 if (!jbd2_journal_check_available_features
4530 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4531 ext4_msg(sb, KERN_ERR, "Journal does not support "
4532 "requested data journaling mode");
744692dc 4533 goto failed_mount_wq;
ac27a0ec
DK		 4534 }
4535 default:
4536 break;
4537 }
ab04df78
JK		 4538
4539 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4540 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4541 ext4_msg(sb, KERN_ERR, "can't mount with "
4542 "journal_async_commit in data=ordered mode");
4543 goto failed_mount_wq;
4544 }
4545
b3881f74 4546 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4547
18aadd47
BJ		 4548 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4549
ce7e010a 4550no_journal:
cdb7ee4c
TE		 4551 if (!test_opt(sb, NO_MBCACHE)) {
4552 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4553 if (!sbi->s_ea_block_cache) {
dec214d0 4554 ext4_msg(sb, KERN_ERR,
cdb7ee4c 4555 "Failed to create ea_block_cache");
dec214d0
TE		 4556 goto failed_mount_wq;
4557 }
cdb7ee4c
TE		 4558
4559 if (ext4_has_feature_ea_inode(sb)) {
4560 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4561 if (!sbi->s_ea_inode_cache) {
4562 ext4_msg(sb, KERN_ERR,
4563 "Failed to create ea_inode_cache");
4564 goto failed_mount_wq;
4565 }
4566 }
9c191f70
M		 4567 }
4568
c93d8f88
EB		 4569 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4570 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4571 goto failed_mount_wq;
4572 }
4573
bc98a42c 4574 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
e2b911c5
DW		 4575 !ext4_has_feature_encrypt(sb)) {
4576 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 4577 ext4_commit_super(sb, 1);
4578 }
4579
952fc18e
TT		 4580 /*
4581 * Get the # of file system overhead blocks from the
4582 * superblock if present.
4583 */
4584 if (es->s_overhead_clusters)
4585 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4586 else {
07aa2ea1
LC		 4587 err = ext4_calculate_overhead(sb);
4588 if (err)
952fc18e
TT		 4589 goto failed_mount_wq;
4590 }
4591
fd89d5f2
TH		 4592 /*
4593 * The maximum number of concurrent works can be high and
4594 * concurrency isn't really necessary. Limit it to 1.
4595 */
2e8fa54e
JK		 4596 EXT4_SB(sb)->rsv_conversion_wq =
4597 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4598 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4599 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4600 ret = -ENOMEM;
2e8fa54e
JK		 4601 goto failed_mount4;
4602 }
4603
ac27a0ec 4604 /*
dab291af 4605 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4606 * so we can safely mount the rest of the filesystem now.
4607 */
4608
8a363970 4609 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
1d1fe1ee 4610 if (IS_ERR(root)) {
b31e1552 4611 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4612 ret = PTR_ERR(root);
32a9bb57 4613 root = NULL;
ac27a0ec
DK		 4614 goto failed_mount4;
4615 }
4616 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4617 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4618 iput(root);
ac27a0ec
DK		 4619 goto failed_mount4;
4620 }
b886ee3e
GKB		 4621
4622#ifdef CONFIG_UNICODE
4623 if (sbi->s_encoding)
4624 sb->s_d_op = &ext4_dentry_ops;
4625#endif
4626
48fde701 4627 sb->s_root = d_make_root(root);
1d1fe1ee 4628 if (!sb->s_root) {
b31e1552 4629 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4630 ret = -ENOMEM;
4631 goto failed_mount4;
4632 }
ac27a0ec 4633
c89128a0
JK		 4634 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4635 if (ret == -EROFS) {
1751e8a6 4636 sb->s_flags |= SB_RDONLY;
c89128a0
JK		 4637 ret = 0;
4638 } else if (ret)
4639 goto failed_mount4a;
ef7f3835 4640
b5799018 4641 ext4_set_resv_clusters(sb);
27dd4385 4642
6fd058f7
TT		 4643 err = ext4_setup_system_zone(sb);
4644 if (err) {
b31e1552 4645 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4646 "zone (%d)", err);
f9ae9cf5
TT		 4647 goto failed_mount4a;
4648 }
4649
4650 ext4_ext_init(sb);
4651 err = ext4_mb_init(sb);
4652 if (err) {
4653 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4654 err);
dcf2d804 4655 goto failed_mount5;
c2774d84
AK		 4656 }
4657
d5e03cbb
TT		 4658 block = ext4_count_free_clusters(sb);
4659 ext4_free_blocks_count_set(sbi->s_es,
4660 EXT4_C2B(sbi, block));
4274f516 4661 ext4_superblock_csum_set(sb);
908c7f19
TH		 4662 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4663 GFP_KERNEL);
d5e03cbb
TT		 4664 if (!err) {
4665 unsigned long freei = ext4_count_free_inodes(sb);
4666 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4274f516 4667 ext4_superblock_csum_set(sb);
908c7f19
TH		 4668 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4669 GFP_KERNEL);
d5e03cbb
TT		 4670 }
4671 if (!err)
4672 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4673 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4674 if (!err)
908c7f19
TH		 4675 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4676 GFP_KERNEL);
c8585c6f 4677 if (!err)
bbd55937 4678 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
c8585c6f 4679
d5e03cbb
TT		 4680 if (err) {
4681 ext4_msg(sb, KERN_ERR, "insufficient memory");
4682 goto failed_mount6;
4683 }
4684
e2b911c5 4685 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 4686 if (!ext4_fill_flex_info(sb)) {
4687 ext4_msg(sb, KERN_ERR,
4688 "unable to initialize "
4689 "flex_bg meta info!");
4690 goto failed_mount6;
4691 }
4692
bfff6873
LC		 4693 err = ext4_register_li_request(sb, first_not_zeroed);
4694 if (err)
dcf2d804 4695 goto failed_mount6;
bfff6873 4696
b5799018 4697 err = ext4_register_sysfs(sb);
dcf2d804
TM		 4698 if (err)
4699 goto failed_mount7;
3197ebdb 4700
9b2ff357
JK		 4701#ifdef CONFIG_QUOTA
4702 /* Enable quota usage during mount. */
bc98a42c 4703 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
9b2ff357
JK		 4704 err = ext4_enable_quotas(sb);
4705 if (err)
4706 goto failed_mount8;
4707 }
4708#endif /* CONFIG_QUOTA */
4709
617ba13b
MC		 4710 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4711 ext4_orphan_cleanup(sb, es);
4712 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4713 if (needs_recovery) {
b31e1552 4714 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4715 ext4_mark_recovery_complete(sb, es);
4716 }
4717 if (EXT4_SB(sb)->s_journal) {
4718 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4719 descr = " journalled data mode";
4720 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4721 descr = " ordered data mode";
4722 else
4723 descr = " writeback data mode";
4724 } else
4725 descr = "out journal";
4726
79add3a3
LC		 4727 if (test_opt(sb, DISCARD)) {
4728 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4729 if (!blk_queue_discard(q))
4730 ext4_msg(sb, KERN_WARNING,
4731 "mounting with \"discard\" option, but "
4732 "the device does not support discard");
4733 }
4734
e294a537
TT		 4735 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4736 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 4737 "Opts: %.*s%s%s", descr,
4738 (int) sizeof(sbi->s_es->s_mount_opts),
4739 sbi->s_es->s_mount_opts,
e294a537 4740 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4741
66e61a9e
TT		 4742 if (es->s_error_count)
4743 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4744
efbed4dc
TT		 4745 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4746 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4747 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4748 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4749
d4c402d9 4750 kfree(orig_data);
ac27a0ec
DK		 4751 return 0;
4752
617ba13b 4753cantfind_ext4:
ac27a0ec 4754 if (!silent)
b31e1552 4755 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4756 goto failed_mount;
4757
72ba7450
TT		 4758#ifdef CONFIG_QUOTA
4759failed_mount8:
ebd173be 4760 ext4_unregister_sysfs(sb);
72ba7450 4761#endif
dcf2d804
TM		 4762failed_mount7:
4763 ext4_unregister_li_request(sb);
4764failed_mount6:
f9ae9cf5 4765 ext4_mb_release(sb);
7c990728
SJS		 4766 rcu_read_lock();
4767 flex_groups = rcu_dereference(sbi->s_flex_groups);
4768 if (flex_groups) {
4769 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
4770 kvfree(flex_groups[i]);
4771 kvfree(flex_groups);
4772 }
4773 rcu_read_unlock();
d5e03cbb
TT		 4774 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4775 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4776 percpu_counter_destroy(&sbi->s_dirs_counter);
4777 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
bbd55937 4778 percpu_free_rwsem(&sbi->s_writepages_rwsem);
00764937 4779failed_mount5:
f9ae9cf5
TT		 4780 ext4_ext_release(sb);
4781 ext4_release_system_zone(sb);
4782failed_mount4a:
94bf608a 4783 dput(sb->s_root);
32a9bb57 4784 sb->s_root = NULL;
94bf608a 4785failed_mount4:
b31e1552 4786 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4787 if (EXT4_SB(sb)->rsv_conversion_wq)
4788 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4789failed_mount_wq:
50c15df6
CX		 4790 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4791 sbi->s_ea_inode_cache = NULL;
4792
4793 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4794 sbi->s_ea_block_cache = NULL;
4795
0390131b
FM		 4796 if (sbi->s_journal) {
4797 jbd2_journal_destroy(sbi->s_journal);
4798 sbi->s_journal = NULL;
4799 }
50460fe8 4800failed_mount3a:
d3922a77 4801 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4802failed_mount3:
9105bb14 4803 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4804 if (sbi->s_mmp_tsk)
4805 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 4806failed_mount2:
1d0c3924
TT		 4807 rcu_read_lock();
4808 group_desc = rcu_dereference(sbi->s_group_desc);
ac27a0ec 4809 for (i = 0; i < db_count; i++)
1d0c3924
TT		 4810 brelse(group_desc[i]);
4811 kvfree(group_desc);
4812 rcu_read_unlock();
ac27a0ec 4813failed_mount:
0441984a
DW		 4814 if (sbi->s_chksum_driver)
4815 crypto_free_shash(sbi->s_chksum_driver);
c83ad55e
GKB		 4816
4817#ifdef CONFIG_UNICODE
4818 utf8_unload(sbi->s_encoding);
4819#endif
4820
ac27a0ec 4821#ifdef CONFIG_QUOTA
a2d4a646 4822 for (i = 0; i < EXT4_MAXQUOTAS; i++)
0ba33fac 4823 kfree(get_qf_name(sb, sbi, i));
ac27a0ec 4824#endif
617ba13b 4825 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4826 brelse(bh);
4827out_fail:
4828 sb->s_fs_info = NULL;
f6830165 4829 kfree(sbi->s_blockgroup_lock);
5aee0f8a 4830out_free_base:
ac27a0ec 4831 kfree(sbi);
d4c402d9 4832 kfree(orig_data);
5e405595 4833 fs_put_dax(dax_dev);
07aa2ea1 4834 return err ? err : ret;
ac27a0ec
DK		 4835}
4836
4837/*
4838 * Setup any per-fs journal parameters now. We'll do this both on
4839 * initial mount, once the journal has been initialised but before we've
4840 * done any recovery; and again on any subsequent remount.
4841 */
617ba13b 4842static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4843{
617ba13b 4844 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4845
30773840
TT		 4846 journal->j_commit_interval = sbi->s_commit_interval;
4847 journal->j_min_batch_time = sbi->s_min_batch_time;
4848 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4849
a931da6a 4850 write_lock(&journal->j_state_lock);
ac27a0ec 4851 if (test_opt(sb, BARRIER))
dab291af 4852 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4853 else
dab291af 4854 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4855 if (test_opt(sb, DATA_ERR_ABORT))
4856 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4857 else
4858 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4859 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4860}
4861
c6cb7e77
EW		 4862static struct inode *ext4_get_journal_inode(struct super_block *sb,
4863 unsigned int journal_inum)
ac27a0ec
DK		 4864{
4865 struct inode *journal_inode;
ac27a0ec 4866
c6cb7e77
EW		 4867 /*
4868 * Test for the existence of a valid inode on disk. Bad things
4869 * happen if we iget() an unused inode, as the subsequent iput()
4870 * will try to delete it.
4871 */
8a363970 4872 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
1d1fe1ee 4873 if (IS_ERR(journal_inode)) {
b31e1552 4874 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4875 return NULL;
4876 }
4877 if (!journal_inode->i_nlink) {
4878 make_bad_inode(journal_inode);
4879 iput(journal_inode);
b31e1552 4880 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4881 return NULL;
4882 }
4883
e5f8eab8 4884 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4885 journal_inode, journal_inode->i_size);
1d1fe1ee 4886 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4887 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4888 iput(journal_inode);
4889 return NULL;
4890 }
c6cb7e77
EW		 4891 return journal_inode;
4892}
4893
4894static journal_t *ext4_get_journal(struct super_block *sb,
4895 unsigned int journal_inum)
4896{
4897 struct inode *journal_inode;
4898 journal_t *journal;
4899
4900 BUG_ON(!ext4_has_feature_journal(sb));
4901
4902 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4903 if (!journal_inode)
4904 return NULL;
ac27a0ec 4905
dab291af 4906 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4907 if (!journal) {
b31e1552 4908 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4909 iput(journal_inode);
4910 return NULL;
4911 }
4912 journal->j_private = sb;
617ba13b 4913 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4914 return journal;
4915}
4916
617ba13b 4917static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4918 dev_t j_dev)
4919{
2b2d6d01 4920 struct buffer_head *bh;
ac27a0ec 4921 journal_t *journal;
617ba13b
MC		 4922 ext4_fsblk_t start;
4923 ext4_fsblk_t len;
ac27a0ec 4924 int hblock, blocksize;
617ba13b 4925 ext4_fsblk_t sb_block;
ac27a0ec 4926 unsigned long offset;
2b2d6d01 4927 struct ext4_super_block *es;
ac27a0ec
DK		 4928 struct block_device *bdev;
4929
e2b911c5 4930 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4931
b31e1552 4932 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4933 if (bdev == NULL)
4934 return NULL;
4935
ac27a0ec 4936 blocksize = sb->s_blocksize;
e1defc4f 4937 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4938 if (blocksize < hblock) {
b31e1552
ES		 4939 ext4_msg(sb, KERN_ERR,
4940 "blocksize too small for journal device");
ac27a0ec
DK		 4941 goto out_bdev;
4942 }
4943
617ba13b
MC		 4944 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4945 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4946 set_blocksize(bdev, blocksize);
4947 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4948 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4949 "external journal");
ac27a0ec
DK		 4950 goto out_bdev;
4951 }
4952
2716b802 4953 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4954 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4955 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4956 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4957 ext4_msg(sb, KERN_ERR, "external journal has "
4958 "bad superblock");
ac27a0ec
DK		 4959 brelse(bh);
4960 goto out_bdev;
4961 }
4962
df4763be
DW		 4963 if ((le32_to_cpu(es->s_feature_ro_compat) &
4964 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4965 es->s_checksum != ext4_superblock_csum(sb, es)) {
4966 ext4_msg(sb, KERN_ERR, "external journal has "
4967 "corrupt superblock");
4968 brelse(bh);
4969 goto out_bdev;
4970 }
4971
617ba13b 4972 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4973 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4974 brelse(bh);
4975 goto out_bdev;
4976 }
4977
bd81d8ee 4978 len = ext4_blocks_count(es);
ac27a0ec
DK		 4979 start = sb_block + 1;
4980 brelse(bh); /* we're done with the superblock */
4981
dab291af 4982 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4983 start, len, blocksize);
4984 if (!journal) {
b31e1552 4985 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4986 goto out_bdev;
4987 }
4988 journal->j_private = sb;
dfec8a14 4989 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4990 wait_on_buffer(journal->j_sb_buffer);
4991 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4992 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4993 goto out_journal;
4994 }
4995 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4996 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4997 "user (unsupported) - %d",
ac27a0ec
DK		 4998 be32_to_cpu(journal->j_superblock->s_nr_users));
4999 goto out_journal;
5000 }
617ba13b
MC		 5001 EXT4_SB(sb)->journal_bdev = bdev;
5002 ext4_init_journal_params(sb, journal);
ac27a0ec 5003 return journal;
0b8e58a1 5004
ac27a0ec 5005out_journal:
dab291af 5006 jbd2_journal_destroy(journal);
ac27a0ec 5007out_bdev:
617ba13b 5008 ext4_blkdev_put(bdev);
ac27a0ec
DK		 5009 return NULL;
5010}
5011
617ba13b
MC		 5012static int ext4_load_journal(struct super_block *sb,
5013 struct ext4_super_block *es,
ac27a0ec
DK		 5014 unsigned long journal_devnum)
5015{
5016 journal_t *journal;
5017 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5018 dev_t journal_dev;
5019 int err = 0;
5020 int really_read_only;
5021
e2b911c5 5022 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 5023
ac27a0ec
DK		 5024 if (journal_devnum &&
5025 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 5026 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5027 "numbers have changed");
ac27a0ec
DK		 5028 journal_dev = new_decode_dev(journal_devnum);
5029 } else
5030 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5031
5032 really_read_only = bdev_read_only(sb->s_bdev);
5033
5034 /*
5035 * Are we loading a blank journal or performing recovery after a
5036 * crash? For recovery, we need to check in advance whether we
5037 * can get read-write access to the device.
5038 */
e2b911c5 5039 if (ext4_has_feature_journal_needs_recovery(sb)) {
bc98a42c 5040 if (sb_rdonly(sb)) {
b31e1552
ES		 5041 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5042 "required on readonly filesystem");
ac27a0ec 5043 if (really_read_only) {
b31e1552 5044 ext4_msg(sb, KERN_ERR, "write access "
d98bf8cd
SR		 5045 "unavailable, cannot proceed "
5046 "(try mounting with noload)");
ac27a0ec
DK		 5047 return -EROFS;
5048 }
b31e1552
ES		 5049 ext4_msg(sb, KERN_INFO, "write access will "
5050 "be enabled during recovery");
ac27a0ec
DK		 5051 }
5052 }
5053
5054 if (journal_inum && journal_dev) {
b31e1552
ES		 5055 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
5056 "and inode journals!");
ac27a0ec
DK		 5057 return -EINVAL;
5058 }
5059
5060 if (journal_inum) {
617ba13b 5061 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 5062 return -EINVAL;
5063 } else {
617ba13b 5064 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 5065 return -EINVAL;
5066 }
5067
90576c0b 5068 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 5069 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 5070
e2b911c5 5071 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 5072 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 5073 if (!err) {
5074 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5075 if (save)
5076 memcpy(save, ((char *) es) +
5077 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 5078 err = jbd2_journal_load(journal);
1c13d5c0
TT		 5079 if (save)
5080 memcpy(((char *) es) + EXT4_S_ERR_START,
5081 save, EXT4_S_ERR_LEN);
5082 kfree(save);
5083 }
ac27a0ec
DK		 5084
5085 if (err) {
b31e1552 5086 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 5087 jbd2_journal_destroy(journal);
ac27a0ec
DK		 5088 return err;
5089 }
5090
617ba13b
MC		 5091 EXT4_SB(sb)->s_journal = journal;
5092 ext4_clear_journal_err(sb, es);
ac27a0ec 5093
c41303ce 5094 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 5095 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5096 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 5097
5098 /* Make sure we flush the recovery flag to disk. */
e2d67052 5099 ext4_commit_super(sb, 1);
ac27a0ec
DK		 5100 }
5101
5102 return 0;
5103}
5104
e2d67052 5105static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 5106{
e2d67052 5107 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 5108 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 5109 int error = 0;
ac27a0ec 5110
bdfe0cbd 5111 if (!sbh || block_device_ejected(sb))
c4be0c1d 5112 return error;
a17712c8
JD		 5113
5114 /*
5115 * The superblock bh should be mapped, but it might not be if the
5116 * device was hot-removed. Not much we can do but fail the I/O.
5117 */
5118 if (!buffer_mapped(sbh))
5119 return error;
5120
71290b36
TT		 5121 /*
5122 * If the file system is mounted read-only, don't update the
5123 * superblock write time. This avoids updating the superblock
5124 * write time when we are mounting the root file system
5125 * read/only but we need to replay the journal; at that point,
5126 * for people who are east of GMT and who make their clock
5127 * tick in localtime for Windows bug-for-bug compatibility,
5128 * the clock is set in the future, and this will cause e2fsck
5129 * to complain and force a full file system check.
5130 */
1751e8a6 5131 if (!(sb->s_flags & SB_RDONLY))
6a0678a7 5132 ext4_update_tstamp(es, s_wtime);
f613dfcb
TT		 5133 if (sb->s_bdev->bd_part)
5134 es->s_kbytes_written =
5135 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
dbae2c55
MC		 5136 ((part_stat_read(sb->s_bdev->bd_part,
5137 sectors[STAT_WRITE]) -
afc32f7e 5138 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 5139 else
5140 es->s_kbytes_written =
5141 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 5142 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5143 ext4_free_blocks_count_set(es,
57042651
TT		 5144 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5145 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 5146 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5147 es->s_free_inodes_count =
5148 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 5149 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 5150 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 5151 ext4_superblock_csum_set(sb);
1566a48a
TT		 5152 if (sync)
5153 lock_buffer(sbh);
e8680786 5154 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
4743f839
PKS		 5155 /*
5156 * Oh, dear. A previous attempt to write the
5157 * superblock failed. This could happen because the
5158 * USB device was yanked out. Or it could happen to
5159 * be a transient write error and maybe the block will
5160 * be remapped. Nothing we can do but to retry the
5161 * write and hope for the best.
5162 */
5163 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5164 "superblock detected");
5165 clear_buffer_write_io_error(sbh);
5166 set_buffer_uptodate(sbh);
5167 }
ac27a0ec 5168 mark_buffer_dirty(sbh);
914258bf 5169 if (sync) {
1566a48a 5170 unlock_buffer(sbh);
564bc402 5171 error = __sync_dirty_buffer(sbh,
00473374 5172 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
c89128a0 5173 if (buffer_write_io_error(sbh)) {
b31e1552
ES		 5174 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5175 "superblock");
914258bf
TT		 5176 clear_buffer_write_io_error(sbh);
5177 set_buffer_uptodate(sbh);
5178 }
5179 }
c4be0c1d 5180 return error;
ac27a0ec
DK		 5181}
5182
ac27a0ec
DK		 5183/*
5184 * Have we just finished recovery? If so, and if we are mounting (or
5185 * remounting) the filesystem readonly, then we will end up with a
5186 * consistent fs on disk. Record that fact.
5187 */
2b2d6d01
TT		 5188static void ext4_mark_recovery_complete(struct super_block *sb,
5189 struct ext4_super_block *es)
ac27a0ec 5190{
617ba13b 5191 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 5192
e2b911c5 5193 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 5194 BUG_ON(journal != NULL);
5195 return;
5196 }
dab291af 5197 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 5198 if (jbd2_journal_flush(journal) < 0)
5199 goto out;
5200
bc98a42c 5201 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
e2b911c5 5202 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 5203 ext4_commit_super(sb, 1);
ac27a0ec 5204 }
7ffe1ea8
HK		 5205
5206out:
dab291af 5207 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 5208}
5209
5210/*
5211 * If we are mounting (or read-write remounting) a filesystem whose journal
5212 * has recorded an error from a previous lifetime, move that error to the
5213 * main filesystem now.
5214 */
2b2d6d01
TT		 5215static void ext4_clear_journal_err(struct super_block *sb,
5216 struct ext4_super_block *es)
ac27a0ec
DK		 5217{
5218 journal_t *journal;
5219 int j_errno;
5220 const char *errstr;
5221
e2b911c5 5222 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 5223
617ba13b 5224 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 5225
5226 /*
5227 * Now check for any error status which may have been recorded in the
617ba13b 5228 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 5229 */
5230
dab291af 5231 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 5232 if (j_errno) {
5233 char nbuf[16];
5234
617ba13b 5235 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 5236 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 5237 "from previous mount: %s", errstr);
12062ddd 5238 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 5239
617ba13b
MC		 5240 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5241 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 5242 ext4_commit_super(sb, 1);
ac27a0ec 5243
dab291af 5244 jbd2_journal_clear_err(journal);
d796c52e 5245 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 5246 }
5247}
5248
5249/*
5250 * Force the running and committing transactions to commit,
5251 * and wait on the commit.
5252 */
617ba13b 5253int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 5254{
5255 journal_t *journal;
ac27a0ec 5256
bc98a42c 5257 if (sb_rdonly(sb))
ac27a0ec
DK		 5258 return 0;
5259
617ba13b 5260 journal = EXT4_SB(sb)->s_journal;
b1deefc9 5261 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 5262}
5263
617ba13b 5264static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 5265{
14ce0cb4 5266 int ret = 0;
9eddacf9 5267 tid_t target;
06a407f1 5268 bool needs_barrier = false;
8d5d02e6 5269 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5270
49598e04 5271 if (unlikely(ext4_forced_shutdown(sbi)))
0db1ff22
TT		 5272 return 0;
5273
9bffad1e 5274 trace_ext4_sync_fs(sb, wait);
2e8fa54e 5275 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 5276 /*
5277 * Writeback quota in non-journalled quota case - journalled quota has
5278 * no dirty dquots
5279 */
5280 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 5281 /*
5282 * Data writeback is possible w/o journal transaction, so barrier must
5283 * being sent at the end of the function. But we can skip it if
5284 * transaction_commit will do it for us.
5285 */
bda32530
TT		 5286 if (sbi->s_journal) {
5287 target = jbd2_get_latest_transaction(sbi->s_journal);
5288 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5289 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5290 needs_barrier = true;
5291
5292 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5293 if (wait)
5294 ret = jbd2_log_wait_commit(sbi->s_journal,
5295 target);
5296 }
5297 } else if (wait && test_opt(sb, BARRIER))
06a407f1 5298 needs_barrier = true;
06a407f1
DM		 5299 if (needs_barrier) {
5300 int err;
5301 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
5302 if (!ret)
5303 ret = err;
0390131b 5304 }
06a407f1
DM		 5305
5306 return ret;
5307}
5308
ac27a0ec
DK		 5309/*
5310 * LVM calls this function before a (read-only) snapshot is created. This
5311 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 5312 *
5313 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 5314 * state independently. It relies on upper layer to stop all data & metadata
5315 * modifications.
ac27a0ec 5316 */
c4be0c1d 5317static int ext4_freeze(struct super_block *sb)
ac27a0ec 5318{
c4be0c1d
TS		 5319 int error = 0;
5320 journal_t *journal;
ac27a0ec 5321
bc98a42c 5322 if (sb_rdonly(sb))
9ca92389 5323 return 0;
ac27a0ec 5324
9ca92389 5325 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 5326
bb044576
TT		 5327 if (journal) {
5328 /* Now we set up the journal barrier. */
5329 jbd2_journal_lock_updates(journal);
ac27a0ec 5330
bb044576
TT		 5331 /*
5332 * Don't clear the needs_recovery flag if we failed to
5333 * flush the journal.
5334 */
5335 error = jbd2_journal_flush(journal);
5336 if (error < 0)
5337 goto out;
c642dc9e
ES		 5338
5339 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 5340 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 5341 }
9ca92389 5342
9ca92389 5343 error = ext4_commit_super(sb, 1);
6b0310fb 5344out:
bb044576
TT		 5345 if (journal)
5346 /* we rely on upper layer to stop further updates */
5347 jbd2_journal_unlock_updates(journal);
6b0310fb 5348 return error;
ac27a0ec
DK		 5349}
5350
5351/*
5352 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5353 * flag here, even though the filesystem is not technically dirty yet.
5354 */
c4be0c1d 5355static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 5356{
bc98a42c 5357 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 5358 return 0;
5359
c642dc9e
ES		 5360 if (EXT4_SB(sb)->s_journal) {
5361 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 5362 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 5363 }
5364
9ca92389 5365 ext4_commit_super(sb, 1);
c4be0c1d 5366 return 0;
ac27a0ec
DK		 5367}
5368
673c6100
TT		 5369/*
5370 * Structure to save mount options for ext4_remount's benefit
5371 */
5372struct ext4_mount_options {
5373 unsigned long s_mount_opt;
a2595b8a 5374 unsigned long s_mount_opt2;
08cefc7a
EB		 5375 kuid_t s_resuid;
5376 kgid_t s_resgid;
673c6100
TT		 5377 unsigned long s_commit_interval;
5378 u32 s_min_batch_time, s_max_batch_time;
5379#ifdef CONFIG_QUOTA
5380 int s_jquota_fmt;
a2d4a646 5381 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 5382#endif
5383};
5384
2b2d6d01 5385static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 5386{
2b2d6d01 5387 struct ext4_super_block *es;
617ba13b 5388 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5389 unsigned long old_sb_flags;
617ba13b 5390 struct ext4_mount_options old_opts;
c79d967d 5391 int enable_quota = 0;
8a266467 5392 ext4_group_t g;
b3881f74 5393 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 5394 int err = 0;
ac27a0ec 5395#ifdef CONFIG_QUOTA
03dafb5f 5396 int i, j;
33458eab 5397 char *to_free[EXT4_MAXQUOTAS];
ac27a0ec 5398#endif
d4c402d9 5399 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec 5400
21ac738e
CX		 5401 if (data && !orig_data)
5402 return -ENOMEM;
5403
ac27a0ec
DK		 5404 /* Store the original options */
5405 old_sb_flags = sb->s_flags;
5406 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 5407 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 5408 old_opts.s_resuid = sbi->s_resuid;
5409 old_opts.s_resgid = sbi->s_resgid;
5410 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 5411 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5412 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 5413#ifdef CONFIG_QUOTA
5414 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 5415 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5416 if (sbi->s_qf_names[i]) {
33458eab
TT		 5417 char *qf_name = get_qf_name(sb, sbi, i);
5418
5419 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
03dafb5f
CG		 5420 if (!old_opts.s_qf_names[i]) {
5421 for (j = 0; j < i; j++)
5422 kfree(old_opts.s_qf_names[j]);
3e36a163 5423 kfree(orig_data);
03dafb5f
CG		 5424 return -ENOMEM;
5425 }
5426 } else
5427 old_opts.s_qf_names[i] = NULL;
ac27a0ec 5428#endif
b3881f74
TT		 5429 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5430 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 5431
661aa520 5432 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 5433 err = -EINVAL;
5434 goto restore_opts;
5435 }
5436
6b992ff2 5437 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 5438 test_opt(sb, JOURNAL_CHECKSUM)) {
5439 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 5440 "during remount not supported; ignoring");
5441 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 5442 }
5443
6ae6514b
PS		 5444 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5445 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5446 ext4_msg(sb, KERN_ERR, "can't mount with "
5447 "both data=journal and delalloc");
5448 err = -EINVAL;
5449 goto restore_opts;
5450 }
5451 if (test_opt(sb, DIOREAD_NOLOCK)) {
5452 ext4_msg(sb, KERN_ERR, "can't mount with "
5453 "both data=journal and dioread_nolock");
5454 err = -EINVAL;
5455 goto restore_opts;
5456 }
fc626fe3 5457 if (test_opt(sb, DAX_ALWAYS)) {
923ae0ff
RZ		 5458 ext4_msg(sb, KERN_ERR, "can't mount with "
5459 "both data=journal and dax");
5460 err = -EINVAL;
5461 goto restore_opts;
5462 }
ab04df78
JK		 5463 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5464 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5465 ext4_msg(sb, KERN_ERR, "can't mount with "
5466 "journal_async_commit in data=ordered mode");
5467 err = -EINVAL;
5468 goto restore_opts;
5469 }
923ae0ff
RZ		 5470 }
5471
cdb7ee4c
TE		 5472 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5473 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5474 err = -EINVAL;
5475 goto restore_opts;
5476 }
5477
9cb20f94
IW		 5478 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX_ALWAYS ||
5479 (sbi->s_mount_opt2 ^ old_opts.s_mount_opt2) & EXT4_MOUNT2_DAX_NEVER ||
5480 (sbi->s_mount_opt2 ^ old_opts.s_mount_opt2) & EXT4_MOUNT2_DAX_INODE) {
923ae0ff 5481 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
9cb20f94
IW		 5482 "dax mount option with busy inodes while remounting");
5483 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
5484 sbi->s_mount_opt |= old_opts.s_mount_opt & EXT4_MOUNT_DAX_ALWAYS;
5485 sbi->s_mount_opt2 &= ~(EXT4_MOUNT2_DAX_NEVER | EXT4_MOUNT2_DAX_INODE);
5486 sbi->s_mount_opt2 |= old_opts.s_mount_opt2 &
5487 (EXT4_MOUNT2_DAX_NEVER | EXT4_MOUNT2_DAX_INODE);
6ae6514b
PS		 5488 }
5489
4ab2f15b 5490 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
54d3adbc 5491 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
ac27a0ec 5492
1751e8a6
LT		 5493 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5494 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec
DK		 5495
5496 es = sbi->s_es;
5497
b3881f74 5498 if (sbi->s_journal) {
0390131b 5499 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 5500 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5501 }
ac27a0ec 5502
1751e8a6
LT		 5503 if (*flags & SB_LAZYTIME)
5504 sb->s_flags |= SB_LAZYTIME;
a2fd66d0 5505
1751e8a6 5506 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
4ab2f15b 5507 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 5508 err = -EROFS;
5509 goto restore_opts;
5510 }
5511
1751e8a6 5512 if (*flags & SB_RDONLY) {
38c03b34
TT		 5513 err = sync_filesystem(sb);
5514 if (err < 0)
5515 goto restore_opts;
0f0dd62f
CH		 5516 err = dquot_suspend(sb, -1);
5517 if (err < 0)
c79d967d 5518 goto restore_opts;
c79d967d 5519
ac27a0ec
DK		 5520 /*
5521 * First of all, the unconditional stuff we have to do
5522 * to disable replay of the journal when we next remount
5523 */
1751e8a6 5524 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 5525
5526 /*
5527 * OK, test if we are remounting a valid rw partition
5528 * readonly, and if so set the rdonly flag and then
5529 * mark the partition as valid again.
5530 */
617ba13b
MC		 5531 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5532 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 5533 es->s_state = cpu_to_le16(sbi->s_mount_state);
5534
a63c9eb2 5535 if (sbi->s_journal)
0390131b 5536 ext4_mark_recovery_complete(sb, es);
2dca60d9
TT		 5537 if (sbi->s_mmp_tsk)
5538 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 5539 } else {
a13fb1a4 5540 /* Make sure we can mount this feature set readwrite */
e2b911c5 5541 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 5542 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 5543 err = -EROFS;
5544 goto restore_opts;
5545 }
8a266467
TT		 5546 /*
5547 * Make sure the group descriptor checksums
0b8e58a1 5548 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 5549 */
5550 for (g = 0; g < sbi->s_groups_count; g++) {
5551 struct ext4_group_desc *gdp =
5552 ext4_get_group_desc(sb, g, NULL);
5553
feb0ab32 5554 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5555 ext4_msg(sb, KERN_ERR,
5556 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 5557 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 5558 le16_to_cpu(gdp->bg_checksum));
6a797d27 5559 err = -EFSBADCRC;
8a266467
TT		 5560 goto restore_opts;
5561 }
5562 }
5563
ead6596b
ES		 5564 /*
5565 * If we have an unprocessed orphan list hanging
5566 * around from a previously readonly bdev mount,
5567 * require a full umount/remount for now.
5568 */
5569 if (es->s_last_orphan) {
b31e1552 5570 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5571 "remount RDWR because of unprocessed "
5572 "orphan inode list. Please "
b31e1552 5573 "umount/remount instead");
ead6596b
ES		 5574 err = -EINVAL;
5575 goto restore_opts;
5576 }
5577
ac27a0ec
DK		 5578 /*
5579 * Mounting a RDONLY partition read-write, so reread
5580 * and store the current valid flag. (It may have
5581 * been changed by e2fsck since we originally mounted
5582 * the partition.)
5583 */
0390131b
FM		 5584 if (sbi->s_journal)
5585 ext4_clear_journal_err(sb, es);
ac27a0ec 5586 sbi->s_mount_state = le16_to_cpu(es->s_state);
c89128a0
JK		 5587
5588 err = ext4_setup_super(sb, es, 0);
5589 if (err)
5590 goto restore_opts;
5591
5592 sb->s_flags &= ~SB_RDONLY;
e2b911c5 5593 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 5594 if (ext4_multi_mount_protect(sb,
5595 le64_to_cpu(es->s_mmp_block))) {
5596 err = -EROFS;
5597 goto restore_opts;
5598 }
c79d967d 5599 enable_quota = 1;
ac27a0ec
DK		 5600 }
5601 }
bfff6873
LC		 5602
5603 /*
5604 * Reinitialize lazy itable initialization thread based on
5605 * current settings
5606 */
bc98a42c 5607 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
bfff6873
LC		 5608 ext4_unregister_li_request(sb);
5609 else {
5610 ext4_group_t first_not_zeroed;
5611 first_not_zeroed = ext4_has_uninit_itable(sb);
5612 ext4_register_li_request(sb, first_not_zeroed);
5613 }
5614
6fd058f7 5615 ext4_setup_system_zone(sb);
c89128a0
JK		 5616 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5617 err = ext4_commit_super(sb, 1);
5618 if (err)
5619 goto restore_opts;
5620 }
0390131b 5621
ac27a0ec
DK		 5622#ifdef CONFIG_QUOTA
5623 /* Release old quota file names */
a2d4a646 5624 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5625 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5626 if (enable_quota) {
5627 if (sb_any_quota_suspended(sb))
5628 dquot_resume(sb, -1);
e2b911c5 5629 else if (ext4_has_feature_quota(sb)) {
7c319d32 5630 err = ext4_enable_quotas(sb);
07724f98 5631 if (err)
7c319d32 5632 goto restore_opts;
7c319d32
AK		 5633 }
5634 }
ac27a0ec 5635#endif
d4c402d9 5636
1751e8a6 5637 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
d4c402d9
CW		 5638 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5639 kfree(orig_data);
ac27a0ec 5640 return 0;
0b8e58a1 5641
ac27a0ec
DK		 5642restore_opts:
5643 sb->s_flags = old_sb_flags;
5644 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5645 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5646 sbi->s_resuid = old_opts.s_resuid;
5647 sbi->s_resgid = old_opts.s_resgid;
5648 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5649 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5650 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5651#ifdef CONFIG_QUOTA
5652 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5653 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
33458eab
TT		 5654 to_free[i] = get_qf_name(sb, sbi, i);
5655 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
ac27a0ec 5656 }
33458eab
TT		 5657 synchronize_rcu();
5658 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5659 kfree(to_free[i]);
ac27a0ec 5660#endif
d4c402d9 5661 kfree(orig_data);
ac27a0ec
DK		 5662 return err;
5663}
5664
689c958c
LX		 5665#ifdef CONFIG_QUOTA
5666static int ext4_statfs_project(struct super_block *sb,
5667 kprojid_t projid, struct kstatfs *buf)
5668{
5669 struct kqid qid;
5670 struct dquot *dquot;
5671 u64 limit;
5672 u64 curblock;
5673
5674 qid = make_kqid_projid(projid);
5675 dquot = dqget(sb, qid);
5676 if (IS_ERR(dquot))
5677 return PTR_ERR(dquot);
7b9ca4c6 5678 spin_lock(&dquot->dq_dqb_lock);
689c958c 5679
a08fe66e
CX		 5680 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
5681 dquot->dq_dqb.dqb_bhardlimit);
57c32ea4
CX		 5682 limit >>= sb->s_blocksize_bits;
5683
689c958c 5684 if (limit && buf->f_blocks > limit) {
f06925c7
KK		 5685 curblock = (dquot->dq_dqb.dqb_curspace +
5686 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
689c958c
LX		 5687 buf->f_blocks = limit;
5688 buf->f_bfree = buf->f_bavail =
5689 (buf->f_blocks > curblock) ?
5690 (buf->f_blocks - curblock) : 0;
5691 }
5692
a08fe66e
CX		 5693 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
5694 dquot->dq_dqb.dqb_ihardlimit);
689c958c
LX		 5695 if (limit && buf->f_files > limit) {
5696 buf->f_files = limit;
5697 buf->f_ffree =
5698 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5699 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5700 }
5701
7b9ca4c6 5702 spin_unlock(&dquot->dq_dqb_lock);
689c958c
LX		 5703 dqput(dquot);
5704 return 0;
5705}
5706#endif
5707
2b2d6d01 5708static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5709{
5710 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5711 struct ext4_sb_info *sbi = EXT4_SB(sb);
5712 struct ext4_super_block *es = sbi->s_es;
27dd4385 5713 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5714 u64 fsid;
d02a9391 5715 s64 bfree;
27dd4385 5716 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5717
952fc18e
TT		 5718 if (!test_opt(sb, MINIX_DF))
5719 overhead = sbi->s_overhead;
ac27a0ec 5720
617ba13b 5721 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5722 buf->f_bsize = sb->s_blocksize;
b72f78cb 5723 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5724 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5725 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5726 /* prevent underflow in case that few free space is available */
57042651 5727 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5728 buf->f_bavail = buf->f_bfree -
5729 (ext4_r_blocks_count(es) + resv_blocks);
5730 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5731 buf->f_bavail = 0;
5732 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5733 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5734 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5735 fsid = le64_to_cpup((void *)es->s_uuid) ^
5736 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5737 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5738 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5739
689c958c
LX		 5740#ifdef CONFIG_QUOTA
5741 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5742 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5743 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5744#endif
ac27a0ec
DK		 5745 return 0;
5746}
5747
ac27a0ec
DK		 5748
5749#ifdef CONFIG_QUOTA
5750
bc8230ee
JK		 5751/*
5752 * Helper functions so that transaction is started before we acquire dqio_sem
5753 * to keep correct lock ordering of transaction > dqio_sem
5754 */
ac27a0ec
DK		 5755static inline struct inode *dquot_to_inode(struct dquot *dquot)
5756{
4c376dca 5757 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5758}
5759
617ba13b 5760static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5761{
5762 int ret, err;
5763 handle_t *handle;
5764 struct inode *inode;
5765
5766 inode = dquot_to_inode(dquot);
9924a92a 5767 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5768 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5769 if (IS_ERR(handle))
5770 return PTR_ERR(handle);
5771 ret = dquot_commit(dquot);
617ba13b 5772 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5773 if (!ret)
5774 ret = err;
5775 return ret;
5776}
5777
617ba13b 5778static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5779{
5780 int ret, err;
5781 handle_t *handle;
5782
9924a92a 5783 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5784 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5785 if (IS_ERR(handle))
5786 return PTR_ERR(handle);
5787 ret = dquot_acquire(dquot);
617ba13b 5788 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5789 if (!ret)
5790 ret = err;
5791 return ret;
5792}
5793
617ba13b 5794static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5795{
5796 int ret, err;
5797 handle_t *handle;
5798
9924a92a 5799 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5800 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5801 if (IS_ERR(handle)) {
5802 /* Release dquot anyway to avoid endless cycle in dqput() */
5803 dquot_release(dquot);
ac27a0ec 5804 return PTR_ERR(handle);
9c3013e9 5805 }
ac27a0ec 5806 ret = dquot_release(dquot);
617ba13b 5807 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5808 if (!ret)
5809 ret = err;
5810 return ret;
5811}
5812
617ba13b 5813static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5814{
262b4662
JK		 5815 struct super_block *sb = dquot->dq_sb;
5816 struct ext4_sb_info *sbi = EXT4_SB(sb);
5817
2c8be6b2 5818 /* Are we journaling quotas? */
e2b911c5 5819 if (ext4_has_feature_quota(sb) ||
262b4662 5820 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5821 dquot_mark_dquot_dirty(dquot);
617ba13b 5822 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5823 } else {
5824 return dquot_mark_dquot_dirty(dquot);
5825 }
5826}
5827
617ba13b 5828static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5829{
5830 int ret, err;
5831 handle_t *handle;
5832
5833 /* Data block + inode block */
2b0143b5 5834 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5835 if (IS_ERR(handle))
5836 return PTR_ERR(handle);
5837 ret = dquot_commit_info(sb, type);
617ba13b 5838 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5839 if (!ret)
5840 ret = err;
5841 return ret;
5842}
5843
5844/*
5845 * Turn on quotas during mount time - we need to find
5846 * the quota file and such...
5847 */
617ba13b 5848static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5849{
33458eab 5850 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
287a8095 5851 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5852}
5853
daf647d2
TT		 5854static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5855{
5856 struct ext4_inode_info *ei = EXT4_I(inode);
5857
5858 /* The first argument of lockdep_set_subclass has to be
5859 * *exactly* the same as the argument to init_rwsem() --- in
5860 * this case, in init_once() --- or lockdep gets unhappy
5861 * because the name of the lock is set using the
5862 * stringification of the argument to init_rwsem().
5863 */
5864 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5865 lockdep_set_subclass(&ei->i_data_sem, subclass);
5866}
5867
ac27a0ec
DK		 5868/*
5869 * Standard function to be called on quota_on
5870 */
617ba13b 5871static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 5872 const struct path *path)
ac27a0ec
DK		 5873{
5874 int err;
ac27a0ec
DK		 5875
5876 if (!test_opt(sb, QUOTA))
5877 return -EINVAL;
0623543b 5878
ac27a0ec 5879 /* Quotafile not on the same filesystem? */
d8c9584e 5880 if (path->dentry->d_sb != sb)
ac27a0ec 5881 return -EXDEV;
0623543b
JK		 5882 /* Journaling quota? */
5883 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5884 /* Quotafile not in fs root? */
f00c9e44 5885 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5886 ext4_msg(sb, KERN_WARNING,
5887 "Quota file not on filesystem root. "
5888 "Journaled quota will not work");
91389240
JK		 5889 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5890 } else {
5891 /*
5892 * Clear the flag just in case mount options changed since
5893 * last time.
5894 */
5895 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
2b2d6d01 5896 }
0623543b
JK		 5897
5898 /*
5899 * When we journal data on quota file, we have to flush journal to see
5900 * all updates to the file when we bypass pagecache...
5901 */
0390131b 5902 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5903 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5904 /*
5905 * We don't need to lock updates but journal_flush() could
5906 * otherwise be livelocked...
5907 */
5908 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5909 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5910 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5911 if (err)
7ffe1ea8 5912 return err;
0623543b 5913 }
957153fc 5914
daf647d2
TT		 5915 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5916 err = dquot_quota_on(sb, type, format_id, path);
957153fc 5917 if (err) {
daf647d2
TT		 5918 lockdep_set_quota_inode(path->dentry->d_inode,
5919 I_DATA_SEM_NORMAL);
957153fc
JK		 5920 } else {
5921 struct inode *inode = d_inode(path->dentry);
5922 handle_t *handle;
5923
61a92987
JK		 5924 /*
5925 * Set inode flags to prevent userspace from messing with quota
5926 * files. If this fails, we return success anyway since quotas
5927 * are already enabled and this is not a hard failure.
5928 */
957153fc
JK		 5929 inode_lock(inode);
5930 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5931 if (IS_ERR(handle))
5932 goto unlock_inode;
5933 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5934 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5935 S_NOATIME | S_IMMUTABLE);
4209ae12 5936 err = ext4_mark_inode_dirty(handle, inode);
957153fc
JK		 5937 ext4_journal_stop(handle);
5938 unlock_inode:
5939 inode_unlock(inode);
5940 }
daf647d2 5941 return err;
ac27a0ec
DK		 5942}
5943
7c319d32
AK		 5944static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5945 unsigned int flags)
5946{
5947 int err;
5948 struct inode *qf_inode;
a2d4a646 5949 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5950 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5951 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5952 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5953 };
5954
e2b911c5 5955 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5956
5957 if (!qf_inums[type])
5958 return -EPERM;
5959
8a363970 5960 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
7c319d32
AK		 5961 if (IS_ERR(qf_inode)) {
5962 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5963 return PTR_ERR(qf_inode);
5964 }
5965
bcb13850
JK		 5966 /* Don't account quota for quota files to avoid recursion */
5967 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5968 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7212b95e 5969 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
daf647d2
TT		 5970 if (err)
5971 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
61157b24 5972 iput(qf_inode);
7c319d32
AK		 5973
5974 return err;
5975}
5976
5977/* Enable usage tracking for all quota types. */
5978static int ext4_enable_quotas(struct super_block *sb)
5979{
5980 int type, err = 0;
a2d4a646 5981 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5982 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5983 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5984 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 5985 };
49da9392
JK		 5986 bool quota_mopt[EXT4_MAXQUOTAS] = {
5987 test_opt(sb, USRQUOTA),
5988 test_opt(sb, GRPQUOTA),
5989 test_opt(sb, PRJQUOTA),
5990 };
7c319d32 5991
91389240 5992 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
a2d4a646 5993 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5994 if (qf_inums[type]) {
5995 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 5996 DQUOT_USAGE_ENABLED |
5997 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 5998 if (err) {
5999 ext4_warning(sb,
72ba7450
TT		 6000 "Failed to enable quota tracking "
6001 "(type=%d, err=%d). Please run "
6002 "e2fsck to fix.", type, err);
7f144fd0
JU		 6003 for (type--; type >= 0; type--)
6004 dquot_quota_off(sb, type);
6005
7c319d32
AK		 6006 return err;
6007 }
6008 }
6009 }
6010 return 0;
6011}
6012
ca0e05e4
DM		 6013static int ext4_quota_off(struct super_block *sb, int type)
6014{
21f97697
JK		 6015 struct inode *inode = sb_dqopt(sb)->files[type];
6016 handle_t *handle;
957153fc 6017 int err;
21f97697 6018
87009d86
DM		 6019 /* Force all delayed allocation blocks to be allocated.
6020 * Caller already holds s_umount sem */
6021 if (test_opt(sb, DELALLOC))
ca0e05e4 6022 sync_filesystem(sb);
ca0e05e4 6023
957153fc 6024 if (!inode || !igrab(inode))
0b268590
AG		 6025 goto out;
6026
957153fc 6027 err = dquot_quota_off(sb, type);
964edf66 6028 if (err || ext4_has_feature_quota(sb))
957153fc
JK		 6029 goto out_put;
6030
6031 inode_lock(inode);
61a92987
JK		 6032 /*
6033 * Update modification times of quota files when userspace can
6034 * start looking at them. If we fail, we return success anyway since
6035 * this is not a hard failure and quotas are already disabled.
6036 */
9924a92a 6037 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
4209ae12
HS		 6038 if (IS_ERR(handle)) {
6039 err = PTR_ERR(handle);
957153fc 6040 goto out_unlock;
4209ae12 6041 }
957153fc
JK		 6042 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6043 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
eeca7ea1 6044 inode->i_mtime = inode->i_ctime = current_time(inode);
4209ae12 6045 err = ext4_mark_inode_dirty(handle, inode);
21f97697 6046 ext4_journal_stop(handle);
957153fc
JK		 6047out_unlock:
6048 inode_unlock(inode);
6049out_put:
964edf66 6050 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 6051 iput(inode);
6052 return err;
21f97697 6053out:
ca0e05e4
DM		 6054 return dquot_quota_off(sb, type);
6055}
6056
ac27a0ec
DK		 6057/* Read data from quotafile - avoid pagecache and such because we cannot afford
6058 * acquiring the locks... As quota files are never truncated and quota code
25985edc 6059 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 6060 * we don't have to be afraid of races */
617ba13b 6061static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 6062 size_t len, loff_t off)
6063{
6064 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 6065 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 6066 int offset = off & (sb->s_blocksize - 1);
6067 int tocopy;
6068 size_t toread;
6069 struct buffer_head *bh;
6070 loff_t i_size = i_size_read(inode);
6071
6072 if (off > i_size)
6073 return 0;
6074 if (off+len > i_size)
6075 len = i_size-off;
6076 toread = len;
6077 while (toread > 0) {
6078 tocopy = sb->s_blocksize - offset < toread ?
6079 sb->s_blocksize - offset : toread;
1c215028
TT		 6080 bh = ext4_bread(NULL, inode, blk, 0);
6081 if (IS_ERR(bh))
6082 return PTR_ERR(bh);
ac27a0ec
DK		 6083 if (!bh) /* A hole? */
6084 memset(data, 0, tocopy);
6085 else
6086 memcpy(data, bh->b_data+offset, tocopy);
6087 brelse(bh);
6088 offset = 0;
6089 toread -= tocopy;
6090 data += tocopy;
6091 blk++;
6092 }
6093 return len;
6094}
6095
6096/* Write to quotafile (we know the transaction is already started and has
6097 * enough credits) */
617ba13b 6098static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 6099 const char *data, size_t len, loff_t off)
6100{
6101 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 6102 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4209ae12 6103 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
c5e298ae 6104 int retries = 0;
ac27a0ec
DK		 6105 struct buffer_head *bh;
6106 handle_t *handle = journal_current_handle();
6107
0390131b 6108 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 6109 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6110 " cancelled because transaction is not started",
9c3013e9
JK		 6111 (unsigned long long)off, (unsigned long long)len);
6112 return -EIO;
6113 }
67eeb568
DM		 6114 /*
6115 * Since we account only one data block in transaction credits,
6116 * then it is impossible to cross a block boundary.
6117 */
6118 if (sb->s_blocksize - offset < len) {
6119 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6120 " cancelled because not block aligned",
6121 (unsigned long long)off, (unsigned long long)len);
6122 return -EIO;
6123 }
6124
c5e298ae
TT		 6125 do {
6126 bh = ext4_bread(handle, inode, blk,
6127 EXT4_GET_BLOCKS_CREATE |
6128 EXT4_GET_BLOCKS_METADATA_NOFAIL);
45586c70 6129 } while (PTR_ERR(bh) == -ENOSPC &&
c5e298ae 6130 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 6131 if (IS_ERR(bh))
6132 return PTR_ERR(bh);
67eeb568
DM		 6133 if (!bh)
6134 goto out;
5d601255 6135 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 6136 err = ext4_journal_get_write_access(handle, bh);
6137 if (err) {
6138 brelse(bh);
1c215028 6139 return err;
ac27a0ec 6140 }
67eeb568
DM		 6141 lock_buffer(bh);
6142 memcpy(bh->b_data+offset, data, len);
6143 flush_dcache_page(bh->b_page);
6144 unlock_buffer(bh);
62d2b5f2 6145 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 6146 brelse(bh);
ac27a0ec 6147out:
67eeb568
DM		 6148 if (inode->i_size < off + len) {
6149 i_size_write(inode, off + len);
617ba13b 6150 EXT4_I(inode)->i_disksize = inode->i_size;
4209ae12
HS		 6151 err2 = ext4_mark_inode_dirty(handle, inode);
6152 if (unlikely(err2 && !err))
6153 err = err2;
ac27a0ec 6154 }
4209ae12 6155 return err ? err : len;
ac27a0ec 6156}
ac27a0ec
DK		 6157#endif
6158
152a0836
AV		 6159static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6160 const char *dev_name, void *data)
ac27a0ec 6161{
152a0836 6162 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 6163}
6164
c290ea01 6165#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 6166static inline void register_as_ext2(void)
6167{
6168 int err = register_filesystem(&ext2_fs_type);
6169 if (err)
6170 printk(KERN_WARNING
6171 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6172}
6173
6174static inline void unregister_as_ext2(void)
6175{
6176 unregister_filesystem(&ext2_fs_type);
6177}
2035e776
TT		 6178
6179static inline int ext2_feature_set_ok(struct super_block *sb)
6180{
e2b911c5 6181 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776 6182 return 0;
bc98a42c 6183 if (sb_rdonly(sb))
2035e776 6184 return 1;
e2b911c5 6185 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 6186 return 0;
6187 return 1;
6188}
24b58424
TT		 6189#else
6190static inline void register_as_ext2(void) { }
6191static inline void unregister_as_ext2(void) { }
2035e776 6192static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 6193#endif
6194
24b58424
TT		 6195static inline void register_as_ext3(void)
6196{
6197 int err = register_filesystem(&ext3_fs_type);
6198 if (err)
6199 printk(KERN_WARNING
6200 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6201}
6202
6203static inline void unregister_as_ext3(void)
6204{
6205 unregister_filesystem(&ext3_fs_type);
6206}
2035e776
TT		 6207
6208static inline int ext3_feature_set_ok(struct super_block *sb)
6209{
e2b911c5 6210 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 6211 return 0;
e2b911c5 6212 if (!ext4_has_feature_journal(sb))
2035e776 6213 return 0;
bc98a42c 6214 if (sb_rdonly(sb))
2035e776 6215 return 1;
e2b911c5 6216 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 6217 return 0;
6218 return 1;
6219}
24b58424 6220
03010a33
TT		 6221static struct file_system_type ext4_fs_type = {
6222 .owner = THIS_MODULE,
6223 .name = "ext4",
152a0836 6224 .mount = ext4_mount,
03010a33
TT		 6225 .kill_sb = kill_block_super,
6226 .fs_flags = FS_REQUIRES_DEV,
6227};
7f78e035 6228MODULE_ALIAS_FS("ext4");
03010a33 6229
e9e3bcec
ES		 6230/* Shared across all ext4 file systems */
6231wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 6232
5dabfc78 6233static int __init ext4_init_fs(void)
ac27a0ec 6234{
e9e3bcec 6235 int i, err;
c9de560d 6236
e294a537 6237 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 6238 ext4_li_info = NULL;
6239 mutex_init(&ext4_li_mtx);
6240
9a4c8019 6241 /* Build-time check for flags consistency */
12e9b892 6242 ext4_check_flag_values();
e9e3bcec 6243
e142d052 6244 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 6245 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 6246
51865fda 6247 err = ext4_init_es();
6fd058f7
TT		 6248 if (err)
6249 return err;
51865fda 6250
1dc0aa46 6251 err = ext4_init_pending();
22cfe4b4
EB		 6252 if (err)
6253 goto out7;
6254
6255 err = ext4_init_post_read_processing();
1dc0aa46
EW		 6256 if (err)
6257 goto out6;
6258
51865fda
ZL		 6259 err = ext4_init_pageio();
6260 if (err)
b5799018 6261 goto out5;
51865fda 6262
5dabfc78 6263 err = ext4_init_system_zone();
bd2d0210 6264 if (err)
b5799018 6265 goto out4;
857ac889 6266
b5799018 6267 err = ext4_init_sysfs();
dd68314c 6268 if (err)
b5799018 6269 goto out3;
857ac889 6270
5dabfc78 6271 err = ext4_init_mballoc();
c9de560d
AT		 6272 if (err)
6273 goto out2;
ac27a0ec
DK		 6274 err = init_inodecache();
6275 if (err)
6276 goto out1;
24b58424 6277 register_as_ext3();
2035e776 6278 register_as_ext2();
03010a33 6279 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 6280 if (err)
6281 goto out;
bfff6873 6282
ac27a0ec
DK		 6283 return 0;
6284out:
24b58424
TT		 6285 unregister_as_ext2();
6286 unregister_as_ext3();
ac27a0ec
DK		 6287 destroy_inodecache();
6288out1:
5dabfc78 6289 ext4_exit_mballoc();
9c191f70 6290out2:
b5799018
TT		 6291 ext4_exit_sysfs();
6292out3:
5dabfc78 6293 ext4_exit_system_zone();
b5799018 6294out4:
5dabfc78 6295 ext4_exit_pageio();
b5799018 6296out5:
22cfe4b4 6297 ext4_exit_post_read_processing();
1dc0aa46 6298out6:
22cfe4b4
EB		 6299 ext4_exit_pending();
6300out7:
51865fda
ZL		 6301 ext4_exit_es();
6302
ac27a0ec
DK		 6303 return err;
6304}
6305
5dabfc78 6306static void __exit ext4_exit_fs(void)
ac27a0ec 6307{
bfff6873 6308 ext4_destroy_lazyinit_thread();
24b58424
TT		 6309 unregister_as_ext2();
6310 unregister_as_ext3();
03010a33 6311 unregister_filesystem(&ext4_fs_type);
ac27a0ec 6312 destroy_inodecache();
5dabfc78 6313 ext4_exit_mballoc();
b5799018 6314 ext4_exit_sysfs();
5dabfc78
TT		 6315 ext4_exit_system_zone();
6316 ext4_exit_pageio();
22cfe4b4 6317 ext4_exit_post_read_processing();
dd12ed14 6318 ext4_exit_es();
1dc0aa46 6319 ext4_exit_pending();
ac27a0ec
DK		 6320}
6321
6322MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 6323MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 6324MODULE_LICENSE("GPL");
7ef79ad5 6325MODULE_SOFTDEP("pre: crc32c");
5dabfc78
TT		 6326module_init(ext4_init_fs)
6327module_exit(ext4_exit_fs)
Empty description
This page took 2.513254 seconds and 4 git commands to generate.