]> git.ipfire.org Git - thirdparty/linux.git/blame - fs/ext4/super.c
projects / thirdparty / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'block-5.7-2020-05-16' of git://git.kernel.dk/linux-block
[thirdparty/linux.git] / fs / ext4 / super.c
CommitLineData
f5166768 1// SPDX-License-Identifier: GPL-2.0
ac27a0ec 2/*
617ba13b 3 * linux/fs/ext4/super.c
ac27a0ec
DK		 4 *
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
9 *
10 * from
11 *
12 * linux/fs/minix/inode.c
13 *
14 * Copyright (C) 1991, 1992 Linus Torvalds
15 *
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
18 */
19
20#include <linux/module.h>
21#include <linux/string.h>
22#include <linux/fs.h>
23#include <linux/time.h>
c5ca7c76 24#include <linux/vmalloc.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
66114cad 28#include <linux/backing-dev.h>
ac27a0ec 29#include <linux/parser.h>
ac27a0ec 30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
ef510424 41#include <linux/dax.h>
7abc52c2 42#include <linux/cleancache.h>
7c0f6ba6 43#include <linux/uaccess.h>
ee73f9a5 44#include <linux/iversion.h>
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
94840e3c
EB		 1326 res = ext4_convert_inline_data(inode);
1327 if (res)
1328 return res;
1329
2f8f5e76
EB		 1330 /*
1331 * If a journal handle was specified, then the encryption context is
1332 * being set on a new inode via inheritance and is part of a larger
1333 * transaction to create the inode. Otherwise the encryption context is
1334 * being set on an existing inode in its own transaction. Only in the
1335 * latter case should the "retry on ENOSPC" logic be used.
1336 */
a7550b30 1337
2f8f5e76
EB		 1338 if (handle) {
1339 res = ext4_xattr_set_handle(handle, inode,
1340 EXT4_XATTR_INDEX_ENCRYPTION,
1341 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1342 ctx, len, 0);
a7550b30
JK		 1343 if (!res) {
1344 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1345 ext4_clear_inode_state(inode,
1346 EXT4_STATE_MAY_INLINE_DATA);
a3caa24b 1347 /*
2ee6a576
EB		 1348 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1349 * S_DAX may be disabled
a3caa24b
JK		 1350 */
1351 ext4_set_inode_flags(inode);
a7550b30
JK		 1352 }
1353 return res;
1354 }
1355
b8cb5a54
TE		 1356 res = dquot_initialize(inode);
1357 if (res)
1358 return res;
2f8f5e76 1359retry:
af65207c
TE		 1360 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1361 &credits);
dec214d0
TE		 1362 if (res)
1363 return res;
1364
c1a5d5f6 1365 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
a7550b30
JK		 1366 if (IS_ERR(handle))
1367 return PTR_ERR(handle);
1368
2f8f5e76
EB		 1369 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1370 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1371 ctx, len, 0);
a7550b30
JK		 1372 if (!res) {
1373 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
2ee6a576
EB		 1374 /*
1375 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1376 * S_DAX may be disabled
1377 */
a3caa24b 1378 ext4_set_inode_flags(inode);
a7550b30
JK		 1379 res = ext4_mark_inode_dirty(handle, inode);
1380 if (res)
1381 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1382 }
1383 res2 = ext4_journal_stop(handle);
2f8f5e76
EB		 1384
1385 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1386 goto retry;
a7550b30
JK		 1387 if (!res)
1388 res = res2;
1389 return res;
1390}
1391
c250b7dd 1392static bool ext4_dummy_context(struct inode *inode)
a7550b30
JK		 1393{
1394 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1395}
1396
b925acb8
EB		 1397static bool ext4_has_stable_inodes(struct super_block *sb)
1398{
1399 return ext4_has_feature_stable_inodes(sb);
1400}
1401
1402static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1403 int *ino_bits_ret, int *lblk_bits_ret)
1404{
1405 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1406 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1407}
1408
6f69f0ed 1409static const struct fscrypt_operations ext4_cryptops = {
a5d431ef 1410 .key_prefix = "ext4:",
a7550b30 1411 .get_context = ext4_get_context,
a7550b30
JK		 1412 .set_context = ext4_set_context,
1413 .dummy_context = ext4_dummy_context,
a7550b30 1414 .empty_dir = ext4_empty_dir,
e12ee683 1415 .max_namelen = EXT4_NAME_LEN,
b925acb8
EB		 1416 .has_stable_inodes = ext4_has_stable_inodes,
1417 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
a7550b30 1418};
a7550b30
JK		 1419#endif
1420
ac27a0ec 1421#ifdef CONFIG_QUOTA
d6006186 1422static const char * const quotatypes[] = INITQFNAMES;
689c958c 1423#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1424
617ba13b
MC		 1425static int ext4_write_dquot(struct dquot *dquot);
1426static int ext4_acquire_dquot(struct dquot *dquot);
1427static int ext4_release_dquot(struct dquot *dquot);
1428static int ext4_mark_dquot_dirty(struct dquot *dquot);
1429static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1430static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1431 const struct path *path);
617ba13b
MC		 1432static int ext4_quota_on_mount(struct super_block *sb, int type);
1433static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1434 size_t len, loff_t off);
617ba13b 1435static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1436 const char *data, size_t len, loff_t off);
7c319d32
AK		 1437static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1438 unsigned int flags);
1439static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1440
96c7e0d9
JK		 1441static struct dquot **ext4_get_dquots(struct inode *inode)
1442{
1443 return EXT4_I(inode)->i_dquot;
1444}
1445
61e225dc 1446static const struct dquot_operations ext4_quota_operations = {
7a9ca53a
TE		 1447 .get_reserved_space = ext4_get_reserved_space,
1448 .write_dquot = ext4_write_dquot,
1449 .acquire_dquot = ext4_acquire_dquot,
1450 .release_dquot = ext4_release_dquot,
1451 .mark_dirty = ext4_mark_dquot_dirty,
1452 .write_info = ext4_write_info,
1453 .alloc_dquot = dquot_alloc,
1454 .destroy_dquot = dquot_destroy,
1455 .get_projid = ext4_get_projid,
1456 .get_inode_usage = ext4_get_inode_usage,
ebc11f7b 1457 .get_next_id = dquot_get_next_id,
ac27a0ec
DK		 1458};
1459
0d54b217 1460static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1461 .quota_on = ext4_quota_on,
ca0e05e4 1462 .quota_off = ext4_quota_off,
287a8095 1463 .quota_sync = dquot_quota_sync,
0a240339 1464 .get_state = dquot_get_state,
287a8095
CH		 1465 .set_info = dquot_set_dqinfo,
1466 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1467 .set_dqblk = dquot_set_dqblk,
1468 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1469};
1470#endif
1471
ee9b6d61 1472static const struct super_operations ext4_sops = {
617ba13b 1473 .alloc_inode = ext4_alloc_inode,
94053139 1474 .free_inode = ext4_free_in_core_inode,
617ba13b 1475 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1476 .write_inode = ext4_write_inode,
1477 .dirty_inode = ext4_dirty_inode,
7ff9c073 1478 .drop_inode = ext4_drop_inode,
0930fcc1 1479 .evict_inode = ext4_evict_inode,
617ba13b 1480 .put_super = ext4_put_super,
617ba13b 1481 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1482 .freeze_fs = ext4_freeze,
1483 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1484 .statfs = ext4_statfs,
1485 .remount_fs = ext4_remount,
617ba13b 1486 .show_options = ext4_show_options,
ac27a0ec 1487#ifdef CONFIG_QUOTA
617ba13b
MC		 1488 .quota_read = ext4_quota_read,
1489 .quota_write = ext4_quota_write,
96c7e0d9 1490 .get_dquots = ext4_get_dquots,
ac27a0ec 1491#endif
c39a7f84 1492 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1493};
1494
39655164 1495static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1496 .fh_to_dentry = ext4_fh_to_dentry,
1497 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1498 .get_parent = ext4_get_parent,
fde87268 1499 .commit_metadata = ext4_nfs_commit_metadata,
ac27a0ec
DK		 1500};
1501
1502enum {
1503 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1504 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1505 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1506 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1507 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1508 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1509 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1510 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1511 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1512 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1513 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1514 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
49da9392 1515 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
327eaf73
TT		 1516 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1517 Opt_nowarn_on_error, Opt_mblk_io_submit,
670e9875 1518 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1449032b 1519 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1520 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1521 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1522 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1523 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
ac27a0ec
DK		 1524};
1525
a447c093 1526static const match_table_t tokens = {
ac27a0ec
DK		 1527 {Opt_bsd_df, "bsddf"},
1528 {Opt_minix_df, "minixdf"},
1529 {Opt_grpid, "grpid"},
1530 {Opt_grpid, "bsdgroups"},
1531 {Opt_nogrpid, "nogrpid"},
1532 {Opt_nogrpid, "sysvgroups"},
1533 {Opt_resgid, "resgid=%u"},
1534 {Opt_resuid, "resuid=%u"},
1535 {Opt_sb, "sb=%u"},
1536 {Opt_err_cont, "errors=continue"},
1537 {Opt_err_panic, "errors=panic"},
1538 {Opt_err_ro, "errors=remount-ro"},
1539 {Opt_nouid32, "nouid32"},
ac27a0ec 1540 {Opt_debug, "debug"},
72578c33
TT		 1541 {Opt_removed, "oldalloc"},
1542 {Opt_removed, "orlov"},
ac27a0ec
DK		 1543 {Opt_user_xattr, "user_xattr"},
1544 {Opt_nouser_xattr, "nouser_xattr"},
1545 {Opt_acl, "acl"},
1546 {Opt_noacl, "noacl"},
e3bb52ae 1547 {Opt_noload, "norecovery"},
5a916be1 1548 {Opt_noload, "noload"},
72578c33
TT		 1549 {Opt_removed, "nobh"},
1550 {Opt_removed, "bh"},
ac27a0ec 1551 {Opt_commit, "commit=%u"},
30773840
TT		 1552 {Opt_min_batch_time, "min_batch_time=%u"},
1553 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1554 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1555 {Opt_journal_path, "journal_path=%s"},
818d276c 1556 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1557 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1558 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1559 {Opt_abort, "abort"},
1560 {Opt_data_journal, "data=journal"},
1561 {Opt_data_ordered, "data=ordered"},
1562 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1563 {Opt_data_err_abort, "data_err=abort"},
1564 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1565 {Opt_offusrjquota, "usrjquota="},
1566 {Opt_usrjquota, "usrjquota=%s"},
1567 {Opt_offgrpjquota, "grpjquota="},
1568 {Opt_grpjquota, "grpjquota=%s"},
1569 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1570 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1571 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1572 {Opt_grpquota, "grpquota"},
1573 {Opt_noquota, "noquota"},
1574 {Opt_quota, "quota"},
1575 {Opt_usrquota, "usrquota"},
49da9392 1576 {Opt_prjquota, "prjquota"},
ac27a0ec 1577 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1578 {Opt_barrier, "barrier"},
1579 {Opt_nobarrier, "nobarrier"},
25ec56b5 1580 {Opt_i_version, "i_version"},
923ae0ff 1581 {Opt_dax, "dax"},
c9de560d 1582 {Opt_stripe, "stripe=%u"},
64769240 1583 {Opt_delalloc, "delalloc"},
327eaf73
TT		 1584 {Opt_warn_on_error, "warn_on_error"},
1585 {Opt_nowarn_on_error, "nowarn_on_error"},
a26f4992
TT		 1586 {Opt_lazytime, "lazytime"},
1587 {Opt_nolazytime, "nolazytime"},
670e9875 1588 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
dd919b98 1589 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1590 {Opt_removed, "mblk_io_submit"},
1591 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1592 {Opt_block_validity, "block_validity"},
1593 {Opt_noblock_validity, "noblock_validity"},
240799cd 1594 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1595 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1596 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1597 {Opt_auto_da_alloc, "auto_da_alloc"},
1598 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc 1599 {Opt_dioread_nolock, "dioread_nolock"},
244adf64 1600 {Opt_dioread_lock, "nodioread_nolock"},
744692dc 1601 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1602 {Opt_discard, "discard"},
1603 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1604 {Opt_init_itable, "init_itable=%u"},
1605 {Opt_init_itable, "init_itable"},
1606 {Opt_noinit_itable, "noinit_itable"},
df981d03 1607 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1608 {Opt_test_dummy_encryption, "test_dummy_encryption"},
cdb7ee4c
TE		 1609 {Opt_nombcache, "nombcache"},
1610 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
c7198b9c
TT		 1611 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1612 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1613 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1614 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1615 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1616 {Opt_err, NULL},
ac27a0ec
DK		 1617};
1618
617ba13b 1619static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1620{
617ba13b 1621 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1622 char *options = (char *) *data;
1623
1624 if (!options || strncmp(options, "sb=", 3) != 0)
1625 return 1; /* Default location */
0b8e58a1 1626
ac27a0ec 1627 options += 3;
0b8e58a1 1628 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1629 sb_block = simple_strtoul(options, &options, 0);
1630 if (*options && *options != ',') {
4776004f 1631 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1632 (char *) *data);
1633 return 1;
1634 }
1635 if (*options == ',')
1636 options++;
1637 *data = (void *) options;
0b8e58a1 1638
ac27a0ec
DK		 1639 return sb_block;
1640}
1641
b3881f74 1642#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
d6006186
EB		 1643static const char deprecated_msg[] =
1644 "Mount option \"%s\" will be removed by %s\n"
437ca0fd 1645 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1646
56c50f11
DM		 1647#ifdef CONFIG_QUOTA
1648static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1649{
1650 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 1651 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
03dafb5f 1652 int ret = -1;
56c50f11 1653
33458eab 1654 if (sb_any_quota_loaded(sb) && !old_qname) {
56c50f11
DM		 1655 ext4_msg(sb, KERN_ERR,
1656 "Cannot change journaled "
1657 "quota options when quota turned on");
57f73c2c 1658 return -1;
56c50f11 1659 }
e2b911c5 1660 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1661 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1662 "ignored when QUOTA feature is enabled");
1663 return 1;
262b4662 1664 }
56c50f11
DM		 1665 qname = match_strdup(args);
1666 if (!qname) {
1667 ext4_msg(sb, KERN_ERR,
1668 "Not enough memory for storing quotafile name");
57f73c2c 1669 return -1;
56c50f11 1670 }
33458eab
TT		 1671 if (old_qname) {
1672 if (strcmp(old_qname, qname) == 0)
03dafb5f
CG		 1673 ret = 1;
1674 else
1675 ext4_msg(sb, KERN_ERR,
1676 "%s quota file already specified",
1677 QTYPE2NAME(qtype));
1678 goto errout;
56c50f11 1679 }
03dafb5f 1680 if (strchr(qname, '/')) {
56c50f11
DM		 1681 ext4_msg(sb, KERN_ERR,
1682 "quotafile must be on filesystem root");
03dafb5f 1683 goto errout;
56c50f11 1684 }
33458eab 1685 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
fd8c37ec 1686 set_opt(sb, QUOTA);
56c50f11 1687 return 1;
03dafb5f
CG		 1688errout:
1689 kfree(qname);
1690 return ret;
56c50f11
DM		 1691}
1692
1693static int clear_qf_name(struct super_block *sb, int qtype)
1694{
1695
1696 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 1697 char *old_qname = get_qf_name(sb, sbi, qtype);
56c50f11 1698
33458eab 1699 if (sb_any_quota_loaded(sb) && old_qname) {
56c50f11
DM		 1700 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1701 " when quota turned on");
57f73c2c 1702 return -1;
56c50f11 1703 }
33458eab
TT		 1704 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1705 synchronize_rcu();
1706 kfree(old_qname);
56c50f11
DM		 1707 return 1;
1708}
1709#endif
1710
26092bf5
TT		 1711#define MOPT_SET 0x0001
1712#define MOPT_CLEAR 0x0002
1713#define MOPT_NOSUPPORT 0x0004
1714#define MOPT_EXPLICIT 0x0008
1715#define MOPT_CLEAR_ERR 0x0010
1716#define MOPT_GTE0 0x0020
ac27a0ec 1717#ifdef CONFIG_QUOTA
26092bf5
TT		 1718#define MOPT_Q 0
1719#define MOPT_QFMT 0x0040
1720#else
1721#define MOPT_Q MOPT_NOSUPPORT
1722#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1723#endif
26092bf5 1724#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1725#define MOPT_NO_EXT2 0x0100
1726#define MOPT_NO_EXT3 0x0200
1727#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1728#define MOPT_STRING 0x0400
26092bf5
TT		 1729
1730static const struct mount_opts {
1731 int token;
1732 int mount_opt;
1733 int flags;
1734} ext4_mount_opts[] = {
1735 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1736 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1737 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1738 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1739 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1740 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1741 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1742 MOPT_EXT4_ONLY | MOPT_SET},
1743 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1744 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1745 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1746 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1747 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1748 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1749 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1750 MOPT_EXT4_ONLY | MOPT_CLEAR},
327eaf73
TT		 1751 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1752 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
c6d3d56d
DW		 1753 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1754 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1755 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1756 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1757 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1758 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1759 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1760 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1761 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1762 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1763 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1764 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1765 MOPT_NO_EXT2},
8dc0aa8c 1766 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1767 MOPT_NO_EXT2},
26092bf5
TT		 1768 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1769 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1770 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1771 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1772 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1773 {Opt_commit, 0, MOPT_GTE0},
1774 {Opt_max_batch_time, 0, MOPT_GTE0},
1775 {Opt_min_batch_time, 0, MOPT_GTE0},
1776 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1777 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1778 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1779 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1780 {Opt_resuid, 0, MOPT_GTE0},
1781 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1782 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1783 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1784 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1785 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1786 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1787 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1788 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1789 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1790 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1791#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1792 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1793 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1794#else
26092bf5
TT		 1795 {Opt_acl, 0, MOPT_NOSUPPORT},
1796 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1797#endif
26092bf5
TT		 1798 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1799 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
670e9875 1800 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
26092bf5
TT		 1801 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1802 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1803 MOPT_SET | MOPT_Q},
1804 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1805 MOPT_SET | MOPT_Q},
49da9392
JK		 1806 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1807 MOPT_SET | MOPT_Q},
26092bf5 1808 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1809 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1810 MOPT_CLEAR | MOPT_Q},
26092bf5
TT		 1811 {Opt_usrjquota, 0, MOPT_Q},
1812 {Opt_grpjquota, 0, MOPT_Q},
1813 {Opt_offusrjquota, 0, MOPT_Q},
1814 {Opt_offgrpjquota, 0, MOPT_Q},
1815 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1816 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1817 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1818 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1819 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
cdb7ee4c 1820 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
26092bf5
TT		 1821 {Opt_err, 0, 0}
1822};
1823
c83ad55e
GKB		 1824#ifdef CONFIG_UNICODE
1825static const struct ext4_sb_encodings {
1826 __u16 magic;
1827 char *name;
1828 char *version;
1829} ext4_sb_encoding_map[] = {
1830 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
1831};
1832
1833static int ext4_sb_read_encoding(const struct ext4_super_block *es,
1834 const struct ext4_sb_encodings **encoding,
1835 __u16 *flags)
1836{
1837 __u16 magic = le16_to_cpu(es->s_encoding);
1838 int i;
1839
1840 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1841 if (magic == ext4_sb_encoding_map[i].magic)
1842 break;
1843
1844 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
1845 return -EINVAL;
1846
1847 *encoding = &ext4_sb_encoding_map[i];
1848 *flags = le16_to_cpu(es->s_encoding_flags);
1849
1850 return 0;
1851}
1852#endif
1853
26092bf5
TT		 1854static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1855 substring_t *args, unsigned long *journal_devnum,
1856 unsigned int *journal_ioprio, int is_remount)
1857{
1858 struct ext4_sb_info *sbi = EXT4_SB(sb);
1859 const struct mount_opts *m;
08cefc7a
EB		 1860 kuid_t uid;
1861 kgid_t gid;
26092bf5
TT		 1862 int arg = 0;
1863
57f73c2c
TT		 1864#ifdef CONFIG_QUOTA
1865 if (token == Opt_usrjquota)
1866 return set_qf_name(sb, USRQUOTA, &args[0]);
1867 else if (token == Opt_grpjquota)
1868 return set_qf_name(sb, GRPQUOTA, &args[0]);
1869 else if (token == Opt_offusrjquota)
1870 return clear_qf_name(sb, USRQUOTA);
1871 else if (token == Opt_offgrpjquota)
1872 return clear_qf_name(sb, GRPQUOTA);
1873#endif
26092bf5 1874 switch (token) {
f7048605
TT		 1875 case Opt_noacl:
1876 case Opt_nouser_xattr:
1877 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1878 break;
26092bf5
TT		 1879 case Opt_sb:
1880 return 1; /* handled by get_sb_block() */
1881 case Opt_removed:
5f3633e3 1882 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1883 return 1;
26092bf5
TT		 1884 case Opt_abort:
1885 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1886 return 1;
1887 case Opt_i_version:
357fdad0 1888 sb->s_flags |= SB_I_VERSION;
26092bf5 1889 return 1;
a26f4992 1890 case Opt_lazytime:
1751e8a6 1891 sb->s_flags |= SB_LAZYTIME;
a26f4992
TT		 1892 return 1;
1893 case Opt_nolazytime:
1751e8a6 1894 sb->s_flags &= ~SB_LAZYTIME;
a26f4992 1895 return 1;
26092bf5
TT		 1896 }
1897
5f3633e3
JK		 1898 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1899 if (token == m->token)
1900 break;
1901
1902 if (m->token == Opt_err) {
1903 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1904 "or missing value", opt);
1905 return -1;
1906 }
1907
8dc0aa8c
TT		 1908 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1909 ext4_msg(sb, KERN_ERR,
1910 "Mount option \"%s\" incompatible with ext2", opt);
1911 return -1;
1912 }
1913 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1914 ext4_msg(sb, KERN_ERR,
1915 "Mount option \"%s\" incompatible with ext3", opt);
1916 return -1;
1917 }
1918
ad4eec61 1919 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1920 return -1;
1921 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1922 return -1;
c93cf2d7
DM		 1923 if (m->flags & MOPT_EXPLICIT) {
1924 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1925 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1926 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1927 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1928 } else
1929 return -1;
1930 }
5f3633e3
JK		 1931 if (m->flags & MOPT_CLEAR_ERR)
1932 clear_opt(sb, ERRORS_MASK);
1933 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1934 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1935 "options when quota turned on");
1936 return -1;
1937 }
1938
1939 if (m->flags & MOPT_NOSUPPORT) {
1940 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1941 } else if (token == Opt_commit) {
1942 if (arg == 0)
1943 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
9ba55543 1944 else if (arg > INT_MAX / HZ) {
1945 ext4_msg(sb, KERN_ERR,
1946 "Invalid commit interval %d, "
1947 "must be smaller than %d",
1948 arg, INT_MAX / HZ);
1949 return -1;
1950 }
5f3633e3 1951 sbi->s_commit_interval = HZ * arg;
670e9875 1952 } else if (token == Opt_debug_want_extra_isize) {
9803387c
TT		 1953 if ((arg & 1) ||
1954 (arg < 4) ||
1955 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
1956 ext4_msg(sb, KERN_ERR,
1957 "Invalid want_extra_isize %d", arg);
1958 return -1;
1959 }
670e9875 1960 sbi->s_want_extra_isize = arg;
5f3633e3 1961 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1962 sbi->s_max_batch_time = arg;
1963 } else if (token == Opt_min_batch_time) {
1964 sbi->s_min_batch_time = arg;
1965 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1966 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1967 ext4_msg(sb, KERN_ERR,
1968 "EXT4-fs: inode_readahead_blks must be "
1969 "0 or a power of 2 smaller than 2^31");
26092bf5 1970 return -1;
5f3633e3
JK		 1971 }
1972 sbi->s_inode_readahead_blks = arg;
1973 } else if (token == Opt_init_itable) {
1974 set_opt(sb, INIT_INODE_TABLE);
1975 if (!args->from)
1976 arg = EXT4_DEF_LI_WAIT_MULT;
1977 sbi->s_li_wait_mult = arg;
1978 } else if (token == Opt_max_dir_size_kb) {
1979 sbi->s_max_dir_size_kb = arg;
1980 } else if (token == Opt_stripe) {
1981 sbi->s_stripe = arg;
1982 } else if (token == Opt_resuid) {
1983 uid = make_kuid(current_user_ns(), arg);
1984 if (!uid_valid(uid)) {
1985 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1986 return -1;
1987 }
5f3633e3
JK		 1988 sbi->s_resuid = uid;
1989 } else if (token == Opt_resgid) {
1990 gid = make_kgid(current_user_ns(), arg);
1991 if (!gid_valid(gid)) {
1992 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1993 return -1;
1994 }
1995 sbi->s_resgid = gid;
1996 } else if (token == Opt_journal_dev) {
1997 if (is_remount) {
1998 ext4_msg(sb, KERN_ERR,
1999 "Cannot specify journal on remount");
2000 return -1;
2001 }
2002 *journal_devnum = arg;
ad4eec61
ES		 2003 } else if (token == Opt_journal_path) {
2004 char *journal_path;
2005 struct inode *journal_inode;
2006 struct path path;
2007 int error;
2008
2009 if (is_remount) {
2010 ext4_msg(sb, KERN_ERR,
2011 "Cannot specify journal on remount");
2012 return -1;
2013 }
2014 journal_path = match_strdup(&args[0]);
2015 if (!journal_path) {
2016 ext4_msg(sb, KERN_ERR, "error: could not dup "
2017 "journal device string");
2018 return -1;
2019 }
2020
2021 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2022 if (error) {
2023 ext4_msg(sb, KERN_ERR, "error: could not find "
2024 "journal device path: error %d", error);
2025 kfree(journal_path);
2026 return -1;
2027 }
2028
2b0143b5 2029 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 2030 if (!S_ISBLK(journal_inode->i_mode)) {
2031 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2032 "is not a block device", journal_path);
2033 path_put(&path);
2034 kfree(journal_path);
2035 return -1;
2036 }
2037
2038 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2039 path_put(&path);
2040 kfree(journal_path);
5f3633e3
JK		 2041 } else if (token == Opt_journal_ioprio) {
2042 if (arg > 7) {
2043 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2044 " (must be 0-7)");
2045 return -1;
2046 }
2047 *journal_ioprio =
2048 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447 2049 } else if (token == Opt_test_dummy_encryption) {
643fa961 2050#ifdef CONFIG_FS_ENCRYPTION
6ddb2447
TT		 2051 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
2052 ext4_msg(sb, KERN_WARNING,
2053 "Test dummy encryption mode enabled");
2054#else
2055 ext4_msg(sb, KERN_WARNING,
2056 "Test dummy encryption mount option ignored");
2057#endif
5f3633e3
JK		 2058 } else if (m->flags & MOPT_DATAJ) {
2059 if (is_remount) {
2060 if (!sbi->s_journal)
2061 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2062 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 2063 ext4_msg(sb, KERN_ERR,
26092bf5 2064 "Cannot change data mode on remount");
26092bf5 2065 return -1;
ac27a0ec 2066 }
26092bf5 2067 } else {
5f3633e3
JK		 2068 clear_opt(sb, DATA_FLAGS);
2069 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 2070 }
5f3633e3
JK		 2071#ifdef CONFIG_QUOTA
2072 } else if (m->flags & MOPT_QFMT) {
2073 if (sb_any_quota_loaded(sb) &&
2074 sbi->s_jquota_fmt != m->mount_opt) {
2075 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2076 "quota options when quota turned on");
2077 return -1;
2078 }
e2b911c5 2079 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 2080 ext4_msg(sb, KERN_INFO,
2081 "Quota format mount options ignored "
262b4662 2082 "when QUOTA feature is enabled");
c325a67c 2083 return 1;
262b4662 2084 }
5f3633e3 2085 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 2086#endif
923ae0ff 2087 } else if (token == Opt_dax) {
ef83b6e8
DW		 2088#ifdef CONFIG_FS_DAX
2089 ext4_msg(sb, KERN_WARNING,
2090 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
a92abd73 2091 sbi->s_mount_opt |= m->mount_opt;
ef83b6e8 2092#else
923ae0ff
RZ		 2093 ext4_msg(sb, KERN_INFO, "dax option not supported");
2094 return -1;
5f3633e3 2095#endif
7915a861
AN		 2096 } else if (token == Opt_data_err_abort) {
2097 sbi->s_mount_opt |= m->mount_opt;
2098 } else if (token == Opt_data_err_ignore) {
2099 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 2100 } else {
2101 if (!args->from)
2102 arg = 1;
2103 if (m->flags & MOPT_CLEAR)
2104 arg = !arg;
2105 else if (unlikely(!(m->flags & MOPT_SET))) {
2106 ext4_msg(sb, KERN_WARNING,
2107 "buggy handling of option %s", opt);
2108 WARN_ON(1);
2109 return -1;
2110 }
2111 if (arg != 0)
2112 sbi->s_mount_opt |= m->mount_opt;
2113 else
2114 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 2115 }
5f3633e3 2116 return 1;
26092bf5
TT		 2117}
2118
2119static int parse_options(char *options, struct super_block *sb,
2120 unsigned long *journal_devnum,
2121 unsigned int *journal_ioprio,
2122 int is_remount)
2123{
1e1a76ed 2124 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
33458eab 2125 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
26092bf5
TT		 2126 substring_t args[MAX_OPT_ARGS];
2127 int token;
2128
2129 if (!options)
2130 return 1;
2131
2132 while ((p = strsep(&options, ",")) != NULL) {
2133 if (!*p)
2134 continue;
2135 /*
2136 * Initialize args struct so we know whether arg was
2137 * found; some options take optional arguments.
2138 */
caecd0af 2139 args[0].to = args[0].from = NULL;
26092bf5
TT		 2140 token = match_token(p, tokens, args);
2141 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2142 journal_ioprio, is_remount) < 0)
2143 return 0;
ac27a0ec
DK		 2144 }
2145#ifdef CONFIG_QUOTA
49da9392
JK		 2146 /*
2147 * We do the test below only for project quotas. 'usrquota' and
2148 * 'grpquota' mount options are allowed even without quota feature
2149 * to support legacy quotas in quota files.
2150 */
2151 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2152 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2153 "Cannot enable project quota enforcement.");
2154 return 0;
2155 }
33458eab
TT		 2156 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2157 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2158 if (usr_qf_name || grp_qf_name) {
2159 if (test_opt(sb, USRQUOTA) && usr_qf_name)
fd8c37ec 2160 clear_opt(sb, USRQUOTA);
ac27a0ec 2161
33458eab 2162 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
fd8c37ec 2163 clear_opt(sb, GRPQUOTA);
ac27a0ec 2164
56c50f11 2165 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 2166 ext4_msg(sb, KERN_ERR, "old and new quota "
2167 "format mixing");
ac27a0ec
DK		 2168 return 0;
2169 }
2170
2171 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 2172 ext4_msg(sb, KERN_ERR, "journaled quota format "
2173 "not specified");
ac27a0ec
DK		 2174 return 0;
2175 }
ac27a0ec
DK		 2176 }
2177#endif
626b035b
RH		 2178 if (test_opt(sb, DIOREAD_NOLOCK)) {
2179 int blocksize =
2180 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2181 if (blocksize < PAGE_SIZE)
2182 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2183 "experimental mount option 'dioread_nolock' "
2184 "for blocksize < PAGE_SIZE");
2185 }
ac27a0ec
DK		 2186 return 1;
2187}
2188
2adf6da8
TT		 2189static inline void ext4_show_quota_options(struct seq_file *seq,
2190 struct super_block *sb)
2191{
2192#if defined(CONFIG_QUOTA)
2193 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 2194 char *usr_qf_name, *grp_qf_name;
2adf6da8
TT		 2195
2196 if (sbi->s_jquota_fmt) {
2197 char *fmtname = "";
2198
2199 switch (sbi->s_jquota_fmt) {
2200 case QFMT_VFS_OLD:
2201 fmtname = "vfsold";
2202 break;
2203 case QFMT_VFS_V0:
2204 fmtname = "vfsv0";
2205 break;
2206 case QFMT_VFS_V1:
2207 fmtname = "vfsv1";
2208 break;
2209 }
2210 seq_printf(seq, ",jqfmt=%s", fmtname);
2211 }
2212
33458eab
TT		 2213 rcu_read_lock();
2214 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2215 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2216 if (usr_qf_name)
2217 seq_show_option(seq, "usrjquota", usr_qf_name);
2218 if (grp_qf_name)
2219 seq_show_option(seq, "grpjquota", grp_qf_name);
2220 rcu_read_unlock();
2adf6da8
TT		 2221#endif
2222}
2223
5a916be1
TT		 2224static const char *token2str(int token)
2225{
50df9fd5 2226 const struct match_token *t;
5a916be1
TT		 2227
2228 for (t = tokens; t->token != Opt_err; t++)
2229 if (t->token == token && !strchr(t->pattern, '='))
2230 break;
2231 return t->pattern;
2232}
2233
2adf6da8
TT		 2234/*
2235 * Show an option if
2236 * - it's set to a non-default value OR
2237 * - if the per-sb default is different from the global default
2238 */
66acdcf4
TT		 2239static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2240 int nodefs)
2adf6da8 2241{
2adf6da8
TT		 2242 struct ext4_sb_info *sbi = EXT4_SB(sb);
2243 struct ext4_super_block *es = sbi->s_es;
68afa7e0 2244 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
5a916be1 2245 const struct mount_opts *m;
66acdcf4 2246 char sep = nodefs ? '\n' : ',';
2adf6da8 2247
66acdcf4
TT		 2248#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2249#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2250
2251 if (sbi->s_sb_block != 1)
5a916be1
TT		 2252 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2253
2254 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2255 int want_set = m->flags & MOPT_SET;
2256 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2257 (m->flags & MOPT_CLEAR_ERR))
2258 continue;
68afa7e0 2259 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 2260 continue; /* skip if same as the default */
2261 if ((want_set &&
2262 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2263 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2264 continue; /* select Opt_noFoo vs Opt_Foo */
2265 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2266 }
5a916be1 2267
08cefc7a 2268 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2269 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2270 SEQ_OPTS_PRINT("resuid=%u",
2271 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2272 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2273 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2274 SEQ_OPTS_PRINT("resgid=%u",
2275 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2276 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2277 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2278 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2279 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2280 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2281 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2282 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2283 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2284 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2285 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2286 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2287 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2288 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
357fdad0 2289 if (sb->s_flags & SB_I_VERSION)
5a916be1 2290 SEQ_OPTS_PUTS("i_version");
66acdcf4 2291 if (nodefs || sbi->s_stripe)
5a916be1 2292 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
68afa7e0
TN		 2293 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2294 (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2295 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2296 SEQ_OPTS_PUTS("data=journal");
2297 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2298 SEQ_OPTS_PUTS("data=ordered");
2299 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2300 SEQ_OPTS_PUTS("data=writeback");
2301 }
66acdcf4
TT		 2302 if (nodefs ||
2303 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2304 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2305 sbi->s_inode_readahead_blks);
2adf6da8 2306
ceec0376 2307 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
66acdcf4 2308 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2309 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2310 if (nodefs || sbi->s_max_dir_size_kb)
2311 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2312 if (test_opt(sb, DATA_ERR_ABORT))
2313 SEQ_OPTS_PUTS("data_err=abort");
338affb5
EB		 2314 if (DUMMY_ENCRYPTION_ENABLED(sbi))
2315 SEQ_OPTS_PUTS("test_dummy_encryption");
2adf6da8
TT		 2316
2317 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2318 return 0;
2319}
2320
66acdcf4
TT		 2321static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2322{
2323 return _ext4_show_options(seq, root->d_sb, 0);
2324}
2325
ebd173be 2326int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2327{
2328 struct super_block *sb = seq->private;
2329 int rc;
2330
bc98a42c 2331 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
66acdcf4
TT		 2332 rc = _ext4_show_options(seq, sb, 1);
2333 seq_puts(seq, "\n");
2334 return rc;
2335}
2336
617ba13b 2337static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2338 int read_only)
2339{
617ba13b 2340 struct ext4_sb_info *sbi = EXT4_SB(sb);
c89128a0 2341 int err = 0;
ac27a0ec 2342
617ba13b 2343 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2344 ext4_msg(sb, KERN_ERR, "revision level too high, "
2345 "forcing read-only mode");
c89128a0 2346 err = -EROFS;
ac27a0ec
DK		 2347 }
2348 if (read_only)
281b5995 2349 goto done;
617ba13b 2350 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2351 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2352 "running e2fsck is recommended");
c8b459f4 2353 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2354 ext4_msg(sb, KERN_WARNING,
2355 "warning: mounting fs with errors, "
2356 "running e2fsck is recommended");
ed3ce80a 2357 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2358 le16_to_cpu(es->s_mnt_count) >=
2359 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2360 ext4_msg(sb, KERN_WARNING,
2361 "warning: maximal mount count reached, "
2362 "running e2fsck is recommended");
ac27a0ec 2363 else if (le32_to_cpu(es->s_checkinterval) &&
6a0678a7
AB		 2364 (ext4_get_tstamp(es, s_lastcheck) +
2365 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
b31e1552
ES		 2366 ext4_msg(sb, KERN_WARNING,
2367 "warning: checktime reached, "
2368 "running e2fsck is recommended");
0b8e58a1 2369 if (!sbi->s_journal)
0390131b 2370 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2371 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2372 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2373 le16_add_cpu(&es->s_mnt_count, 1);
6a0678a7 2374 ext4_update_tstamp(es, s_mtime);
0390131b 2375 if (sbi->s_journal)
e2b911c5 2376 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2377
c89128a0 2378 err = ext4_commit_super(sb, 1);
281b5995 2379done:
ac27a0ec 2380 if (test_opt(sb, DEBUG))
a9df9a49 2381 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2382 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2383 sb->s_blocksize,
2384 sbi->s_groups_count,
617ba13b
MC		 2385 EXT4_BLOCKS_PER_GROUP(sb),
2386 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2387 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2388
7abc52c2 2389 cleancache_init_fs(sb);
c89128a0 2390 return err;
ac27a0ec
DK		 2391}
2392
117fff10
TT		 2393int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2394{
2395 struct ext4_sb_info *sbi = EXT4_SB(sb);
7c990728 2396 struct flex_groups **old_groups, **new_groups;
37b0b6b8 2397 int size, i, j;
117fff10
TT		 2398
2399 if (!sbi->s_log_groups_per_flex)
2400 return 0;
2401
2402 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2403 if (size <= sbi->s_flex_groups_allocated)
2404 return 0;
2405
7c990728
SJS		 2406 new_groups = kvzalloc(roundup_pow_of_two(size *
2407 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
117fff10 2408 if (!new_groups) {
7c990728
SJS		 2409 ext4_msg(sb, KERN_ERR,
2410 "not enough memory for %d flex group pointers", size);
117fff10
TT		 2411 return -ENOMEM;
2412 }
7c990728
SJS		 2413 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2414 new_groups[i] = kvzalloc(roundup_pow_of_two(
2415 sizeof(struct flex_groups)),
2416 GFP_KERNEL);
2417 if (!new_groups[i]) {
37b0b6b8
DC		 2418 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2419 kvfree(new_groups[j]);
7c990728
SJS		 2420 kvfree(new_groups);
2421 ext4_msg(sb, KERN_ERR,
2422 "not enough memory for %d flex groups", size);
2423 return -ENOMEM;
2424 }
117fff10 2425 }
7c990728
SJS		 2426 rcu_read_lock();
2427 old_groups = rcu_dereference(sbi->s_flex_groups);
2428 if (old_groups)
2429 memcpy(new_groups, old_groups,
2430 (sbi->s_flex_groups_allocated *
2431 sizeof(struct flex_groups *)));
2432 rcu_read_unlock();
2433 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2434 sbi->s_flex_groups_allocated = size;
2435 if (old_groups)
2436 ext4_kvfree_array_rcu(old_groups);
117fff10
TT		 2437 return 0;
2438}
2439
772cb7c8
JS		 2440static int ext4_fill_flex_info(struct super_block *sb)
2441{
2442 struct ext4_sb_info *sbi = EXT4_SB(sb);
2443 struct ext4_group_desc *gdp = NULL;
7c990728 2444 struct flex_groups *fg;
772cb7c8 2445 ext4_group_t flex_group;
117fff10 2446 int i, err;
772cb7c8 2447
503358ae 2448 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2449 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2450 sbi->s_log_groups_per_flex = 0;
2451 return 1;
2452 }
2453
117fff10
TT		 2454 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2455 if (err)
9933fc0a 2456 goto failed;
772cb7c8 2457
772cb7c8 2458 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2459 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2460
2461 flex_group = ext4_flex_group(sbi, i);
7c990728
SJS		 2462 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2463 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
90ba983f 2464 atomic64_add(ext4_free_group_clusters(sb, gdp),
7c990728
SJS		 2465 &fg->free_clusters);
2466 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
772cb7c8
JS		 2467 }
2468
2469 return 1;
2470failed:
2471 return 0;
2472}
2473
e2b911c5 2474static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2475 struct ext4_group_desc *gdp)
717d50e4 2476{
b47820ed 2477 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2478 __u16 crc = 0;
feb0ab32 2479 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2480 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2481
9aa5d32b 2482 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2483 /* Use new metadata_csum algorithm */
feb0ab32 2484 __u32 csum32;
b47820ed 2485 __u16 dummy_csum = 0;
feb0ab32 2486
feb0ab32
DW		 2487 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2488 sizeof(le_group));
b47820ed
DJ		 2489 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2490 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2491 sizeof(dummy_csum));
2492 offset += sizeof(dummy_csum);
2493 if (offset < sbi->s_desc_size)
2494 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2495 sbi->s_desc_size - offset);
feb0ab32
DW		 2496
2497 crc = csum32 & 0xFFFF;
2498 goto out;
717d50e4
AD		 2499 }
2500
feb0ab32 2501 /* old crc16 code */
e2b911c5 2502 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2503 return 0;
2504
feb0ab32
DW		 2505 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2506 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2507 crc = crc16(crc, (__u8 *)gdp, offset);
2508 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2509 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2510 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2511 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2512 crc = crc16(crc, (__u8 *)gdp + offset,
2513 le16_to_cpu(sbi->s_es->s_desc_size) -
2514 offset);
2515
2516out:
717d50e4
AD		 2517 return cpu_to_le16(crc);
2518}
2519
feb0ab32 2520int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2521 struct ext4_group_desc *gdp)
2522{
feb0ab32 2523 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2524 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2525 return 0;
2526
2527 return 1;
2528}
2529
feb0ab32
DW		 2530void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2531 struct ext4_group_desc *gdp)
2532{
2533 if (!ext4_has_group_desc_csum(sb))
2534 return;
e2b911c5 2535 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2536}
2537
ac27a0ec 2538/* Called at mount-time, super-block is locked */
bfff6873 2539static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2540 ext4_fsblk_t sb_block,
bfff6873 2541 ext4_group_t *first_not_zeroed)
ac27a0ec 2542{
617ba13b
MC		 2543 struct ext4_sb_info *sbi = EXT4_SB(sb);
2544 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2545 ext4_fsblk_t last_block;
44de022c 2546 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
bd81d8ee
LV		 2547 ext4_fsblk_t block_bitmap;
2548 ext4_fsblk_t inode_bitmap;
2549 ext4_fsblk_t inode_table;
ce421581 2550 int flexbg_flag = 0;
bfff6873 2551 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2552
e2b911c5 2553 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2554 flexbg_flag = 1;
2555
af5bc92d 2556 ext4_debug("Checking group descriptors");
ac27a0ec 2557
197cd65a
AM		 2558 for (i = 0; i < sbi->s_groups_count; i++) {
2559 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2560
ce421581 2561 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2562 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2563 else
2564 last_block = first_block +
617ba13b 2565 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2566
bfff6873
LC		 2567 if ((grp == sbi->s_groups_count) &&
2568 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2569 grp = i;
2570
8fadc143 2571 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2572 if (block_bitmap == sb_block) {
2573 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2574 "Block bitmap for group %u overlaps "
2575 "superblock", i);
18db4b4e
TT		 2576 if (!sb_rdonly(sb))
2577 return 0;
829fa70d 2578 }
77260807
TT		 2579 if (block_bitmap >= sb_block + 1 &&
2580 block_bitmap <= last_bg_block) {
2581 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2582 "Block bitmap for group %u overlaps "
2583 "block group descriptors", i);
2584 if (!sb_rdonly(sb))
2585 return 0;
2586 }
2b2d6d01 2587 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2588 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2589 "Block bitmap for group %u not in group "
b31e1552 2590 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2591 return 0;
2592 }
8fadc143 2593 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2594 if (inode_bitmap == sb_block) {
2595 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2596 "Inode bitmap for group %u overlaps "
2597 "superblock", i);
18db4b4e
TT		 2598 if (!sb_rdonly(sb))
2599 return 0;
829fa70d 2600 }
77260807
TT		 2601 if (inode_bitmap >= sb_block + 1 &&
2602 inode_bitmap <= last_bg_block) {
2603 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2604 "Inode bitmap for group %u overlaps "
2605 "block group descriptors", i);
2606 if (!sb_rdonly(sb))
2607 return 0;
2608 }
2b2d6d01 2609 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2610 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2611 "Inode bitmap for group %u not in group "
b31e1552 2612 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2613 return 0;
2614 }
8fadc143 2615 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2616 if (inode_table == sb_block) {
2617 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2618 "Inode table for group %u overlaps "
2619 "superblock", i);
18db4b4e
TT		 2620 if (!sb_rdonly(sb))
2621 return 0;
829fa70d 2622 }
77260807
TT		 2623 if (inode_table >= sb_block + 1 &&
2624 inode_table <= last_bg_block) {
2625 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2626 "Inode table for group %u overlaps "
2627 "block group descriptors", i);
2628 if (!sb_rdonly(sb))
2629 return 0;
2630 }
bd81d8ee 2631 if (inode_table < first_block ||
2b2d6d01 2632 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2633 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2634 "Inode table for group %u not in group "
b31e1552 2635 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2636 return 0;
2637 }
955ce5f5 2638 ext4_lock_group(sb, i);
feb0ab32 2639 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2640 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2641 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2642 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2643 gdp)), le16_to_cpu(gdp->bg_checksum));
bc98a42c 2644 if (!sb_rdonly(sb)) {
955ce5f5 2645 ext4_unlock_group(sb, i);
8a266467 2646 return 0;
7ee1ec4c 2647 }
717d50e4 2648 }
955ce5f5 2649 ext4_unlock_group(sb, i);
ce421581
JS		 2650 if (!flexbg_flag)
2651 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2652 }
bfff6873
LC		 2653 if (NULL != first_not_zeroed)
2654 *first_not_zeroed = grp;
ac27a0ec
DK		 2655 return 1;
2656}
2657
617ba13b 2658/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2659 * the superblock) which were deleted from all directories, but held open by
2660 * a process at the time of a crash. We walk the list and try to delete these
2661 * inodes at recovery time (only with a read-write filesystem).
2662 *
2663 * In order to keep the orphan inode chain consistent during traversal (in
2664 * case of crash during recovery), we link each inode into the superblock
2665 * orphan list_head and handle it the same way as an inode deletion during
2666 * normal operation (which journals the operations for us).
2667 *
2668 * We only do an iget() and an iput() on each inode, which is very safe if we
2669 * accidentally point at an in-use or already deleted inode. The worst that
2670 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2671 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2672 * e2fsck was run on this filesystem, and it must have already done the orphan
2673 * inode cleanup for us, so we can safely abort without any further action.
2674 */
2b2d6d01
TT		 2675static void ext4_orphan_cleanup(struct super_block *sb,
2676 struct ext4_super_block *es)
ac27a0ec
DK		 2677{
2678 unsigned int s_flags = sb->s_flags;
2c98eb5e 2679 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec 2680#ifdef CONFIG_QUOTA
95f1fda4 2681 int quota_update = 0;
ac27a0ec
DK		 2682 int i;
2683#endif
2684 if (!es->s_last_orphan) {
2685 jbd_debug(4, "no orphan inodes to clean up\n");
2686 return;
2687 }
2688
a8f48a95 2689 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2690 ext4_msg(sb, KERN_ERR, "write access "
2691 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2692 return;
2693 }
2694
d39195c3
AG		 2695 /* Check if feature set would not allow a r/w mount */
2696 if (!ext4_feature_set_ok(sb, 0)) {
2697 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2698 "unknown ROCOMPAT features");
2699 return;
2700 }
2701
617ba13b 2702 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6 2703 /* don't clear list on RO mount w/ errors */
1751e8a6 2704 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
84474976 2705 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2706 "clearing orphan list.\n");
c25f9bc6
ES		 2707 es->s_last_orphan = 0;
2708 }
ac27a0ec
DK		 2709 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2710 return;
2711 }
2712
1751e8a6 2713 if (s_flags & SB_RDONLY) {
b31e1552 2714 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1751e8a6 2715 sb->s_flags &= ~SB_RDONLY;
ac27a0ec
DK		 2716 }
2717#ifdef CONFIG_QUOTA
2718 /* Needed for iput() to work correctly and not trash data */
1751e8a6 2719 sb->s_flags |= SB_ACTIVE;
95f1fda4 2720
2721 /*
2722 * Turn on quotas which were not enabled for read-only mounts if
2723 * filesystem has quota feature, so that they are updated correctly.
2724 */
1751e8a6 2725 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
95f1fda4 2726 int ret = ext4_enable_quotas(sb);
2727
2728 if (!ret)
2729 quota_update = 1;
2730 else
2731 ext4_msg(sb, KERN_ERR,
2732 "Cannot turn on quotas: error %d", ret);
2733 }
2734
2735 /* Turn on journaled quotas used for old sytle */
a2d4a646 2736 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2737 if (EXT4_SB(sb)->s_qf_names[i]) {
2738 int ret = ext4_quota_on_mount(sb, i);
95f1fda4 2739
2740 if (!ret)
2741 quota_update = 1;
2742 else
b31e1552
ES		 2743 ext4_msg(sb, KERN_ERR,
2744 "Cannot turn on journaled "
95f1fda4 2745 "quota: type %d: error %d", i, ret);
ac27a0ec
DK		 2746 }
2747 }
2748#endif
2749
2750 while (es->s_last_orphan) {
2751 struct inode *inode;
2752
c65d5c6c
VN		 2753 /*
2754 * We may have encountered an error during cleanup; if
2755 * so, skip the rest.
2756 */
2757 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2758 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2759 es->s_last_orphan = 0;
2760 break;
2761 }
2762
97bd42b9
JB		 2763 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2764 if (IS_ERR(inode)) {
ac27a0ec
DK		 2765 es->s_last_orphan = 0;
2766 break;
2767 }
2768
617ba13b 2769 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2770 dquot_initialize(inode);
ac27a0ec 2771 if (inode->i_nlink) {
566370a2
PT		 2772 if (test_opt(sb, DEBUG))
2773 ext4_msg(sb, KERN_DEBUG,
2774 "%s: truncating inode %lu to %lld bytes",
2775 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2776 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2777 inode->i_ino, inode->i_size);
5955102c 2778 inode_lock(inode);
55f252c9 2779 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 2780 ret = ext4_truncate(inode);
2781 if (ret)
2782 ext4_std_error(inode->i_sb, ret);
5955102c 2783 inode_unlock(inode);
ac27a0ec
DK		 2784 nr_truncates++;
2785 } else {
566370a2
PT		 2786 if (test_opt(sb, DEBUG))
2787 ext4_msg(sb, KERN_DEBUG,
2788 "%s: deleting unreferenced inode %lu",
2789 __func__, inode->i_ino);
ac27a0ec
DK		 2790 jbd_debug(2, "deleting unreferenced inode %lu\n",
2791 inode->i_ino);
2792 nr_orphans++;
2793 }
2794 iput(inode); /* The delete magic happens here! */
2795 }
2796
2b2d6d01 2797#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2798
2799 if (nr_orphans)
b31e1552
ES		 2800 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2801 PLURAL(nr_orphans));
ac27a0ec 2802 if (nr_truncates)
b31e1552
ES		 2803 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2804 PLURAL(nr_truncates));
ac27a0ec 2805#ifdef CONFIG_QUOTA
95f1fda4 2806 /* Turn off quotas if they were enabled for orphan cleanup */
2807 if (quota_update) {
2808 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2809 if (sb_dqopt(sb)->files[i])
2810 dquot_quota_off(sb, i);
2811 }
ac27a0ec
DK		 2812 }
2813#endif
1751e8a6 2814 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
ac27a0ec 2815}
0b8e58a1 2816
cd2291a4
ES		 2817/*
2818 * Maximal extent format file size.
2819 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2820 * extent format containers, within a sector_t, and within i_blocks
2821 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2822 * so that won't be a limiting factor.
2823 *
f17722f9
LC		 2824 * However there is other limiting factor. We do store extents in the form
2825 * of starting block and length, hence the resulting length of the extent
2826 * covering maximum file size must fit into on-disk format containers as
2827 * well. Given that length is always by 1 unit bigger than max unit (because
2828 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2829 *
cd2291a4
ES		 2830 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2831 */
f287a1a5 2832static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2833{
2834 loff_t res;
2835 loff_t upper_limit = MAX_LFS_FILESIZE;
2836
72deb455
CH		 2837 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
2838
2839 if (!has_huge_files) {
cd2291a4
ES		 2840 upper_limit = (1LL << 32) - 1;
2841
2842 /* total blocks in file system block size */
2843 upper_limit >>= (blkbits - 9);
2844 upper_limit <<= blkbits;
2845 }
2846
f17722f9
LC		 2847 /*
2848 * 32-bit extent-start container, ee_block. We lower the maxbytes
2849 * by one fs block, so ee_len can cover the extent of maximum file
2850 * size
2851 */
2852 res = (1LL << 32) - 1;
cd2291a4 2853 res <<= blkbits;
cd2291a4
ES		 2854
2855 /* Sanity check against vm- & vfs- imposed limits */
2856 if (res > upper_limit)
2857 res = upper_limit;
2858
2859 return res;
2860}
ac27a0ec 2861
ac27a0ec 2862/*
cd2291a4 2863 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2864 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2865 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2866 */
f287a1a5 2867static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2868{
617ba13b 2869 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2870 int meta_blocks;
2871 loff_t upper_limit;
0b8e58a1
AD		 2872 /* This is calculated to be the largest file size for a dense, block
2873 * mapped file such that the file's total number of 512-byte sectors,
2874 * including data and all indirect blocks, does not exceed (2^48 - 1).
2875 *
2876 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2877 * number of 512-byte sectors of the file.
0fc1b451
AK		 2878 */
2879
72deb455 2880 if (!has_huge_files) {
0fc1b451 2881 /*
72deb455
CH		 2882 * !has_huge_files or implies that the inode i_block field
2883 * represents total file blocks in 2^32 512-byte sectors ==
2884 * size of vfs inode i_blocks * 8
0fc1b451
AK		 2885 */
2886 upper_limit = (1LL << 32) - 1;
2887
2888 /* total blocks in file system block size */
2889 upper_limit >>= (bits - 9);
2890
2891 } else {
8180a562
AK		 2892 /*
2893 * We use 48 bit ext4_inode i_blocks
2894 * With EXT4_HUGE_FILE_FL set the i_blocks
2895 * represent total number of blocks in
2896 * file system block size
2897 */
0fc1b451
AK		 2898 upper_limit = (1LL << 48) - 1;
2899
0fc1b451
AK		 2900 }
2901
2902 /* indirect blocks */
2903 meta_blocks = 1;
2904 /* double indirect blocks */
2905 meta_blocks += 1 + (1LL << (bits-2));
2906 /* tripple indirect blocks */
2907 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2908
2909 upper_limit -= meta_blocks;
2910 upper_limit <<= bits;
ac27a0ec
DK		 2911
2912 res += 1LL << (bits-2);
2913 res += 1LL << (2*(bits-2));
2914 res += 1LL << (3*(bits-2));
2915 res <<= bits;
2916 if (res > upper_limit)
2917 res = upper_limit;
0fc1b451
AK		 2918
2919 if (res > MAX_LFS_FILESIZE)
2920 res = MAX_LFS_FILESIZE;
2921
ac27a0ec
DK		 2922 return res;
2923}
2924
617ba13b 2925static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2926 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2927{
617ba13b 2928 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2929 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2930 int has_super = 0;
2931
2932 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2933
e2b911c5 2934 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2935 return logical_sb_block + nr + 1;
ac27a0ec 2936 bg = sbi->s_desc_per_block * nr;
617ba13b 2937 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2938 has_super = 1;
0b8e58a1 2939
bd63f6b0
DW		 2940 /*
2941 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2942 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2943 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2944 * compensate.
2945 */
2946 if (sb->s_blocksize == 1024 && nr == 0 &&
49598e04 2947 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
bd63f6b0
DW		 2948 has_super++;
2949
617ba13b 2950 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2951}
2952
c9de560d
AT		 2953/**
2954 * ext4_get_stripe_size: Get the stripe size.
2955 * @sbi: In memory super block info
2956 *
2957 * If we have specified it via mount option, then
2958 * use the mount option value. If the value specified at mount time is
2959 * greater than the blocks per group use the super block value.
2960 * If the super block value is greater than blocks per group return 0.
2961 * Allocator needs it be less than blocks per group.
2962 *
2963 */
2964static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2965{
2966 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2967 unsigned long stripe_width =
2968 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2969 int ret;
c9de560d
AT		 2970
2971 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 2972 ret = sbi->s_stripe;
5469d7c3 2973 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 2974 ret = stripe_width;
5469d7c3 2975 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 2976 ret = stride;
2977 else
2978 ret = 0;
c9de560d 2979
3eb08658
DE		 2980 /*
2981 * If the stripe width is 1, this makes no sense and
2982 * we set it to 0 to turn off stripe handling code.
2983 */
2984 if (ret <= 1)
2985 ret = 0;
c9de560d 2986
3eb08658 2987 return ret;
c9de560d 2988}
ac27a0ec 2989
a13fb1a4
ES		 2990/*
2991 * Check whether this filesystem can be mounted based on
2992 * the features present and the RDONLY/RDWR mount requested.
2993 * Returns 1 if this filesystem can be mounted as requested,
2994 * 0 if it cannot be.
2995 */
2996static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2997{
e2b911c5 2998 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2999 ext4_msg(sb, KERN_ERR,
3000 "Couldn't mount because of "
3001 "unsupported optional features (%x)",
3002 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3003 ~EXT4_FEATURE_INCOMPAT_SUPP));
3004 return 0;
3005 }
3006
c83ad55e
GKB		 3007#ifndef CONFIG_UNICODE
3008 if (ext4_has_feature_casefold(sb)) {
3009 ext4_msg(sb, KERN_ERR,
3010 "Filesystem with casefold feature cannot be "
3011 "mounted without CONFIG_UNICODE");
3012 return 0;
3013 }
3014#endif
3015
a13fb1a4
ES		 3016 if (readonly)
3017 return 1;
3018
e2b911c5 3019 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b 3020 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
1751e8a6 3021 sb->s_flags |= SB_RDONLY;
2cb5cc8b
DW		 3022 return 1;
3023 }
3024
a13fb1a4 3025 /* Check that feature set is OK for a read-write mount */
e2b911c5 3026 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 3027 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3028 "unsupported optional features (%x)",
3029 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3030 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3031 return 0;
3032 }
e2b911c5 3033 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 3034 ext4_msg(sb, KERN_ERR,
3035 "Can't support bigalloc feature without "
3036 "extents feature\n");
3037 return 0;
3038 }
7c319d32 3039
9db176bc 3040#if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
d65d87a0
TT		 3041 if (!readonly && (ext4_has_feature_quota(sb) ||
3042 ext4_has_feature_project(sb))) {
7c319d32 3043 ext4_msg(sb, KERN_ERR,
d65d87a0 3044 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
689c958c
LX		 3045 return 0;
3046 }
7c319d32 3047#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 3048 return 1;
3049}
3050
66e61a9e
TT		 3051/*
3052 * This function is called once a day if we have errors logged
3053 * on the file system
3054 */
235699a8 3055static void print_daily_error_info(struct timer_list *t)
66e61a9e 3056{
235699a8
KC		 3057 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3058 struct super_block *sb = sbi->s_sb;
3059 struct ext4_super_block *es = sbi->s_es;
66e61a9e
TT		 3060
3061 if (es->s_error_count)
ae0f78de
TT		 3062 /* fsck newer than v1.41.13 is needed to clean this condition. */
3063 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 3064 le32_to_cpu(es->s_error_count));
3065 if (es->s_first_error_time) {
6a0678a7
AB		 3066 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3067 sb->s_id,
3068 ext4_get_tstamp(es, s_first_error_time),
66e61a9e
TT		 3069 (int) sizeof(es->s_first_error_func),
3070 es->s_first_error_func,
3071 le32_to_cpu(es->s_first_error_line));
3072 if (es->s_first_error_ino)
651e1c3b 3073 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3074 le32_to_cpu(es->s_first_error_ino));
3075 if (es->s_first_error_block)
651e1c3b 3076 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3077 le64_to_cpu(es->s_first_error_block));
651e1c3b 3078 printk(KERN_CONT "\n");
66e61a9e
TT		 3079 }
3080 if (es->s_last_error_time) {
6a0678a7
AB		 3081 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3082 sb->s_id,
3083 ext4_get_tstamp(es, s_last_error_time),
66e61a9e
TT		 3084 (int) sizeof(es->s_last_error_func),
3085 es->s_last_error_func,
3086 le32_to_cpu(es->s_last_error_line));
3087 if (es->s_last_error_ino)
651e1c3b 3088 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3089 le32_to_cpu(es->s_last_error_ino));
3090 if (es->s_last_error_block)
651e1c3b 3091 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3092 le64_to_cpu(es->s_last_error_block));
651e1c3b 3093 printk(KERN_CONT "\n");
66e61a9e
TT		 3094 }
3095 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3096}
3097
bfff6873
LC		 3098/* Find next suitable group and run ext4_init_inode_table */
3099static int ext4_run_li_request(struct ext4_li_request *elr)
3100{
3101 struct ext4_group_desc *gdp = NULL;
3102 ext4_group_t group, ngroups;
3103 struct super_block *sb;
3104 unsigned long timeout = 0;
3105 int ret = 0;
3106
3107 sb = elr->lr_super;
3108 ngroups = EXT4_SB(sb)->s_groups_count;
3109
3110 for (group = elr->lr_next_group; group < ngroups; group++) {
3111 gdp = ext4_get_group_desc(sb, group, NULL);
3112 if (!gdp) {
3113 ret = 1;
3114 break;
3115 }
3116
3117 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3118 break;
3119 }
3120
7f511862 3121 if (group >= ngroups)
bfff6873
LC		 3122 ret = 1;
3123
3124 if (!ret) {
3125 timeout = jiffies;
3126 ret = ext4_init_inode_table(sb, group,
3127 elr->lr_timeout ? 0 : 1);
3128 if (elr->lr_timeout == 0) {
51ce6511
LC		 3129 timeout = (jiffies - timeout) *
3130 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 3131 elr->lr_timeout = timeout;
3132 }
3133 elr->lr_next_sched = jiffies + elr->lr_timeout;
3134 elr->lr_next_group = group + 1;
3135 }
bfff6873
LC		 3136 return ret;
3137}
3138
3139/*
3140 * Remove lr_request from the list_request and free the
4ed5c033 3141 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 3142 */
3143static void ext4_remove_li_request(struct ext4_li_request *elr)
3144{
3145 struct ext4_sb_info *sbi;
3146
3147 if (!elr)
3148 return;
3149
3150 sbi = elr->lr_sbi;
3151
3152 list_del(&elr->lr_request);
3153 sbi->s_li_request = NULL;
3154 kfree(elr);
3155}
3156
3157static void ext4_unregister_li_request(struct super_block *sb)
3158{
1bb933fb
LC		 3159 mutex_lock(&ext4_li_mtx);
3160 if (!ext4_li_info) {
3161 mutex_unlock(&ext4_li_mtx);
bfff6873 3162 return;
1bb933fb 3163 }
bfff6873
LC		 3164
3165 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 3166 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 3167 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 3168 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 3169}
3170
8f1f7453
ES		 3171static struct task_struct *ext4_lazyinit_task;
3172
bfff6873
LC		 3173/*
3174 * This is the function where ext4lazyinit thread lives. It walks
3175 * through the request list searching for next scheduled filesystem.
3176 * When such a fs is found, run the lazy initialization request
3177 * (ext4_rn_li_request) and keep track of the time spend in this
3178 * function. Based on that time we compute next schedule time of
3179 * the request. When walking through the list is complete, compute
3180 * next waking time and put itself into sleep.
3181 */
3182static int ext4_lazyinit_thread(void *arg)
3183{
3184 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3185 struct list_head *pos, *n;
3186 struct ext4_li_request *elr;
4ed5c033 3187 unsigned long next_wakeup, cur;
bfff6873
LC		 3188
3189 BUG_ON(NULL == eli);
3190
bfff6873
LC		 3191cont_thread:
3192 while (true) {
3193 next_wakeup = MAX_JIFFY_OFFSET;
3194
3195 mutex_lock(&eli->li_list_mtx);
3196 if (list_empty(&eli->li_request_list)) {
3197 mutex_unlock(&eli->li_list_mtx);
3198 goto exit_thread;
3199 }
bfff6873 3200 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 3201 int err = 0;
3202 int progress = 0;
bfff6873
LC		 3203 elr = list_entry(pos, struct ext4_li_request,
3204 lr_request);
3205
e22834f0
DM		 3206 if (time_before(jiffies, elr->lr_next_sched)) {
3207 if (time_before(elr->lr_next_sched, next_wakeup))
3208 next_wakeup = elr->lr_next_sched;
3209 continue;
3210 }
3211 if (down_read_trylock(&elr->lr_super->s_umount)) {
3212 if (sb_start_write_trylock(elr->lr_super)) {
3213 progress = 1;
3214 /*
3215 * We hold sb->s_umount, sb can not
3216 * be removed from the list, it is
3217 * now safe to drop li_list_mtx
3218 */
3219 mutex_unlock(&eli->li_list_mtx);
3220 err = ext4_run_li_request(elr);
3221 sb_end_write(elr->lr_super);
3222 mutex_lock(&eli->li_list_mtx);
3223 n = pos->next;
b2c78cd0 3224 }
e22834f0
DM		 3225 up_read((&elr->lr_super->s_umount));
3226 }
3227 /* error, remove the lazy_init job */
3228 if (err) {
3229 ext4_remove_li_request(elr);
3230 continue;
3231 }
3232 if (!progress) {
3233 elr->lr_next_sched = jiffies +
3234 (prandom_u32()
3235 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 3236 }
bfff6873
LC		 3237 if (time_before(elr->lr_next_sched, next_wakeup))
3238 next_wakeup = elr->lr_next_sched;
3239 }
3240 mutex_unlock(&eli->li_list_mtx);
3241
a0acae0e 3242 try_to_freeze();
bfff6873 3243
4ed5c033
LC		 3244 cur = jiffies;
3245 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3246 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3247 cond_resched();
3248 continue;
3249 }
3250
4ed5c033
LC		 3251 schedule_timeout_interruptible(next_wakeup - cur);
3252
8f1f7453
ES		 3253 if (kthread_should_stop()) {
3254 ext4_clear_request_list();
3255 goto exit_thread;
3256 }
bfff6873
LC		 3257 }
3258
3259exit_thread:
3260 /*
3261 * It looks like the request list is empty, but we need
3262 * to check it under the li_list_mtx lock, to prevent any
3263 * additions into it, and of course we should lock ext4_li_mtx
3264 * to atomically free the list and ext4_li_info, because at
3265 * this point another ext4 filesystem could be registering
3266 * new one.
3267 */
3268 mutex_lock(&ext4_li_mtx);
3269 mutex_lock(&eli->li_list_mtx);
3270 if (!list_empty(&eli->li_request_list)) {
3271 mutex_unlock(&eli->li_list_mtx);
3272 mutex_unlock(&ext4_li_mtx);
3273 goto cont_thread;
3274 }
3275 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3276 kfree(ext4_li_info);
3277 ext4_li_info = NULL;
3278 mutex_unlock(&ext4_li_mtx);
3279
3280 return 0;
3281}
3282
3283static void ext4_clear_request_list(void)
3284{
3285 struct list_head *pos, *n;
3286 struct ext4_li_request *elr;
3287
3288 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3289 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3290 elr = list_entry(pos, struct ext4_li_request,
3291 lr_request);
3292 ext4_remove_li_request(elr);
3293 }
3294 mutex_unlock(&ext4_li_info->li_list_mtx);
3295}
3296
3297static int ext4_run_lazyinit_thread(void)
3298{
8f1f7453
ES		 3299 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3300 ext4_li_info, "ext4lazyinit");
3301 if (IS_ERR(ext4_lazyinit_task)) {
3302 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3303 ext4_clear_request_list();
bfff6873
LC		 3304 kfree(ext4_li_info);
3305 ext4_li_info = NULL;
92b97816 3306 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3307 "initialization thread\n",
3308 err);
3309 return err;
3310 }
3311 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3312 return 0;
3313}
3314
3315/*
3316 * Check whether it make sense to run itable init. thread or not.
3317 * If there is at least one uninitialized inode table, return
3318 * corresponding group number, else the loop goes through all
3319 * groups and return total number of groups.
3320 */
3321static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3322{
3323 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3324 struct ext4_group_desc *gdp = NULL;
3325
8844618d
TT		 3326 if (!ext4_has_group_desc_csum(sb))
3327 return ngroups;
3328
bfff6873
LC		 3329 for (group = 0; group < ngroups; group++) {
3330 gdp = ext4_get_group_desc(sb, group, NULL);
3331 if (!gdp)
3332 continue;
3333
50122847 3334 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
bfff6873
LC		 3335 break;
3336 }
3337
3338 return group;
3339}
3340
3341static int ext4_li_info_new(void)
3342{
3343 struct ext4_lazy_init *eli = NULL;
3344
3345 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3346 if (!eli)
3347 return -ENOMEM;
3348
bfff6873
LC		 3349 INIT_LIST_HEAD(&eli->li_request_list);
3350 mutex_init(&eli->li_list_mtx);
3351
bfff6873
LC		 3352 eli->li_state |= EXT4_LAZYINIT_QUIT;
3353
3354 ext4_li_info = eli;
3355
3356 return 0;
3357}
3358
3359static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3360 ext4_group_t start)
3361{
3362 struct ext4_sb_info *sbi = EXT4_SB(sb);
3363 struct ext4_li_request *elr;
bfff6873
LC		 3364
3365 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3366 if (!elr)
3367 return NULL;
3368
3369 elr->lr_super = sb;
3370 elr->lr_sbi = sbi;
3371 elr->lr_next_group = start;
3372
3373 /*
3374 * Randomize first schedule time of the request to
3375 * spread the inode table initialization requests
3376 * better.
3377 */
dd1f723b
TT		 3378 elr->lr_next_sched = jiffies + (prandom_u32() %
3379 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3380 return elr;
3381}
3382
7f511862
TT		 3383int ext4_register_li_request(struct super_block *sb,
3384 ext4_group_t first_not_zeroed)
bfff6873
LC		 3385{
3386 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3387 struct ext4_li_request *elr = NULL;
49598e04 3388 ext4_group_t ngroups = sbi->s_groups_count;
6c5a6cb9 3389 int ret = 0;
bfff6873 3390
7f511862 3391 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3392 if (sbi->s_li_request != NULL) {
3393 /*
3394 * Reset timeout so it can be computed again, because
3395 * s_li_wait_mult might have changed.
3396 */
3397 sbi->s_li_request->lr_timeout = 0;
7f511862 3398 goto out;
51ce6511 3399 }
bfff6873 3400
bc98a42c 3401 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
55ff3840 3402 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3403 goto out;
bfff6873
LC		 3404
3405 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3406 if (!elr) {
3407 ret = -ENOMEM;
3408 goto out;
3409 }
bfff6873
LC		 3410
3411 if (NULL == ext4_li_info) {
3412 ret = ext4_li_info_new();
3413 if (ret)
3414 goto out;
3415 }
3416
3417 mutex_lock(&ext4_li_info->li_list_mtx);
3418 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3419 mutex_unlock(&ext4_li_info->li_list_mtx);
3420
3421 sbi->s_li_request = elr;
46e4690b
TM		 3422 /*
3423 * set elr to NULL here since it has been inserted to
3424 * the request_list and the removal and free of it is
3425 * handled by ext4_clear_request_list from now on.
3426 */
3427 elr = NULL;
bfff6873
LC		 3428
3429 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3430 ret = ext4_run_lazyinit_thread();
3431 if (ret)
3432 goto out;
3433 }
bfff6873 3434out:
beed5ecb
NK		 3435 mutex_unlock(&ext4_li_mtx);
3436 if (ret)
bfff6873 3437 kfree(elr);
bfff6873
LC		 3438 return ret;
3439}
3440
3441/*
3442 * We do not need to lock anything since this is called on
3443 * module unload.
3444 */
3445static void ext4_destroy_lazyinit_thread(void)
3446{
3447 /*
3448 * If thread exited earlier
3449 * there's nothing to be done.
3450 */
8f1f7453 3451 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3452 return;
3453
8f1f7453 3454 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3455}
3456
25ed6e8a
DW		 3457static int set_journal_csum_feature_set(struct super_block *sb)
3458{
3459 int ret = 1;
3460 int compat, incompat;
3461 struct ext4_sb_info *sbi = EXT4_SB(sb);
3462
9aa5d32b 3463 if (ext4_has_metadata_csum(sb)) {
db9ee220 3464 /* journal checksum v3 */
25ed6e8a 3465 compat = 0;
db9ee220 3466 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3467 } else {
3468 /* journal checksum v1 */
3469 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3470 incompat = 0;
3471 }
3472
feb8c6d3
DW		 3473 jbd2_journal_clear_features(sbi->s_journal,
3474 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3475 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3476 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3477 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3478 ret = jbd2_journal_set_features(sbi->s_journal,
3479 compat, 0,
3480 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3481 incompat);
3482 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3483 ret = jbd2_journal_set_features(sbi->s_journal,
3484 compat, 0,
3485 incompat);
3486 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3487 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3488 } else {
feb8c6d3
DW		 3489 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3490 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3491 }
3492
3493 return ret;
3494}
3495
952fc18e
TT		 3496/*
3497 * Note: calculating the overhead so we can be compatible with
3498 * historical BSD practice is quite difficult in the face of
3499 * clusters/bigalloc. This is because multiple metadata blocks from
3500 * different block group can end up in the same allocation cluster.
3501 * Calculating the exact overhead in the face of clustered allocation
3502 * requires either O(all block bitmaps) in memory or O(number of block
3503 * groups**2) in time. We will still calculate the superblock for
3504 * older file systems --- and if we come across with a bigalloc file
3505 * system with zero in s_overhead_clusters the estimate will be close to
3506 * correct especially for very large cluster sizes --- but for newer
3507 * file systems, it's better to calculate this figure once at mkfs
3508 * time, and store it in the superblock. If the superblock value is
3509 * present (even for non-bigalloc file systems), we will use it.
3510 */
3511static int count_overhead(struct super_block *sb, ext4_group_t grp,
3512 char *buf)
3513{
3514 struct ext4_sb_info *sbi = EXT4_SB(sb);
3515 struct ext4_group_desc *gdp;
3516 ext4_fsblk_t first_block, last_block, b;
3517 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3518 int s, j, count = 0;
3519
e2b911c5 3520 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3521 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3522 sbi->s_itb_per_group + 2);
3523
952fc18e
TT		 3524 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3525 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3526 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3527 for (i = 0; i < ngroups; i++) {
3528 gdp = ext4_get_group_desc(sb, i, NULL);
3529 b = ext4_block_bitmap(sb, gdp);
3530 if (b >= first_block && b <= last_block) {
3531 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3532 count++;
3533 }
3534 b = ext4_inode_bitmap(sb, gdp);
3535 if (b >= first_block && b <= last_block) {
3536 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3537 count++;
3538 }
3539 b = ext4_inode_table(sb, gdp);
3540 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3541 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3542 int c = EXT4_B2C(sbi, b - first_block);
3543 ext4_set_bit(c, buf);
3544 count++;
3545 }
3546 if (i != grp)
3547 continue;
3548 s = 0;
3549 if (ext4_bg_has_super(sb, grp)) {
3550 ext4_set_bit(s++, buf);
3551 count++;
3552 }
c48ae41b
TT		 3553 j = ext4_bg_num_gdb(sb, grp);
3554 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3555 ext4_error(sb, "Invalid number of block group "
3556 "descriptor blocks: %d", j);
3557 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3558 }
c48ae41b
TT		 3559 count += j;
3560 for (; j > 0; j--)
3561 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3562 }
3563 if (!count)
3564 return 0;
3565 return EXT4_CLUSTERS_PER_GROUP(sb) -
3566 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3567}
3568
3569/*
3570 * Compute the overhead and stash it in sbi->s_overhead
3571 */
3572int ext4_calculate_overhead(struct super_block *sb)
3573{
3574 struct ext4_sb_info *sbi = EXT4_SB(sb);
3575 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3576 struct inode *j_inode;
3577 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3578 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3579 ext4_fsblk_t overhead = 0;
4fdb5543 3580 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3581
952fc18e
TT		 3582 if (!buf)
3583 return -ENOMEM;
3584
3585 /*
3586 * Compute the overhead (FS structures). This is constant
3587 * for a given filesystem unless the number of block groups
3588 * changes so we cache the previous value until it does.
3589 */
3590
3591 /*
3592 * All of the blocks before first_data_block are overhead
3593 */
3594 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3595
3596 /*
3597 * Add the overhead found in each block group
3598 */
3599 for (i = 0; i < ngroups; i++) {
3600 int blks;
3601
3602 blks = count_overhead(sb, i, buf);
3603 overhead += blks;
3604 if (blks)
3605 memset(buf, 0, PAGE_SIZE);
3606 cond_resched();
3607 }
3c816ded
EW		 3608
3609 /*
3610 * Add the internal journal blocks whether the journal has been
3611 * loaded or not
3612 */
b003b524 3613 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3614 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
f1eec3b0
RH		 3615 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3616 /* j_inum for internal journal is non-zero */
3c816ded
EW		 3617 j_inode = ext4_get_journal_inode(sb, j_inum);
3618 if (j_inode) {
3619 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3620 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3621 iput(j_inode);
3622 } else {
3623 ext4_msg(sb, KERN_ERR, "can't get journal size");
3624 }
3625 }
952fc18e
TT		 3626 sbi->s_overhead = overhead;
3627 smp_wmb();
3628 free_page((unsigned long) buf);
3629 return 0;
3630}
3631
b5799018 3632static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3633{
3634 ext4_fsblk_t resv_clusters;
b5799018 3635 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3636
30fac0f7
JK		 3637 /*
3638 * There's no need to reserve anything when we aren't using extents.
3639 * The space estimates are exact, there are no unwritten extents,
3640 * hole punching doesn't need new metadata... This is needed especially
3641 * to keep ext2/3 backward compatibility.
3642 */
e2b911c5 3643 if (!ext4_has_feature_extents(sb))
b5799018 3644 return;
27dd4385
LC		 3645 /*
3646 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3647 * This should cover the situations where we can not afford to run
3648 * out of space like for example punch hole, or converting
556615dc 3649 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3650 * allocation would require 1, or 2 blocks, higher numbers are
3651 * very rare.
3652 */
b5799018
TT		 3653 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3654 sbi->s_cluster_bits);
27dd4385
LC		 3655
3656 do_div(resv_clusters, 50);
3657 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3658
b5799018 3659 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3660}
3661
2b2d6d01 3662static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3663{
5e405595 3664 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
d4c402d9 3665 char *orig_data = kstrdup(data, GFP_KERNEL);
1d0c3924 3666 struct buffer_head *bh, **group_desc;
617ba13b 3667 struct ext4_super_block *es = NULL;
5aee0f8a 3668 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
7c990728 3669 struct flex_groups **flex_groups;
617ba13b
MC		 3670 ext4_fsblk_t block;
3671 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3672 ext4_fsblk_t logical_sb_block;
ac27a0ec 3673 unsigned long offset = 0;
ac27a0ec
DK		 3674 unsigned long journal_devnum = 0;
3675 unsigned long def_mount_opts;
3676 struct inode *root;
0390131b 3677 const char *descr;
dcc7dae3 3678 int ret = -ENOMEM;
281b5995 3679 int blocksize, clustersize;
4ec11028
TT		 3680 unsigned int db_count;
3681 unsigned int i;
281b5995 3682 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3683 __u64 blocks_count;
07aa2ea1 3684 int err = 0;
b3881f74 3685 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3686 ext4_group_t first_not_zeroed;
ac27a0ec 3687
5aee0f8a
TT		 3688 if ((data && !orig_data) || !sbi)
3689 goto out_free_base;
705895b6 3690
aed9eb1b 3691 sbi->s_daxdev = dax_dev;
705895b6
PE		 3692 sbi->s_blockgroup_lock =
3693 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 3694 if (!sbi->s_blockgroup_lock)
3695 goto out_free_base;
3696
ac27a0ec 3697 sb->s_fs_info = sbi;
2c0544b2 3698 sbi->s_sb = sb;
240799cd 3699 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3700 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3701 if (sb->s_bdev->bd_part)
3702 sbi->s_sectors_written_start =
dbae2c55 3703 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
ac27a0ec 3704
9f6200bb 3705 /* Cleanup superblock name */
ec3904dc 3706 strreplace(sb->s_id, '/', '!');
9f6200bb 3707
07aa2ea1 3708 /* -EINVAL is default */
dcc7dae3 3709 ret = -EINVAL;
617ba13b 3710 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3711 if (!blocksize) {
b31e1552 3712 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3713 goto out_fail;
3714 }
3715
3716 /*
617ba13b 3717 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3718 * block sizes. We need to calculate the offset from buffer start.
3719 */
617ba13b 3720 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3721 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3722 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3723 } else {
70bbb3e0 3724 logical_sb_block = sb_block;
ac27a0ec
DK		 3725 }
3726
a8ac900b 3727 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3728 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3729 goto out_fail;
3730 }
3731 /*
3732 * Note: s_es must be initialized as soon as possible because
617ba13b 3733 * some ext4 macro-instructions depend on its value
ac27a0ec 3734 */
2716b802 3735 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3736 sbi->s_es = es;
3737 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3738 if (sb->s_magic != EXT4_SUPER_MAGIC)
3739 goto cantfind_ext4;
afc32f7e 3740 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3741
feb0ab32 3742 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3743 if (ext4_has_feature_metadata_csum(sb) &&
3744 ext4_has_feature_gdt_csum(sb))
363307e6 3745 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3746 "redundant flags; please run fsck.");
3747
d25425f8
DW		 3748 /* Check for a known checksum algorithm */
3749 if (!ext4_verify_csum_type(sb, es)) {
3750 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3751 "unknown checksum algorithm.");
3752 silent = 1;
3753 goto cantfind_ext4;
3754 }
3755
0441984a 3756 /* Load the checksum driver */
a45403b5
TT		 3757 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3758 if (IS_ERR(sbi->s_chksum_driver)) {
3759 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3760 ret = PTR_ERR(sbi->s_chksum_driver);
3761 sbi->s_chksum_driver = NULL;
3762 goto failed_mount;
0441984a
DW		 3763 }
3764
a9c47317
DW		 3765 /* Check superblock checksum */
3766 if (!ext4_superblock_csum_verify(sb, es)) {
3767 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3768 "invalid superblock checksum. Run e2fsck?");
3769 silent = 1;
6a797d27 3770 ret = -EFSBADCRC;
a9c47317
DW		 3771 goto cantfind_ext4;
3772 }
3773
3774 /* Precompute checksum seed for all metadata */
e2b911c5 3775 if (ext4_has_feature_csum_seed(sb))
8c81bd8f 3776 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
dec214d0 3777 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
a9c47317
DW		 3778 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3779 sizeof(es->s_uuid));
3780
ac27a0ec
DK		 3781 /* Set defaults before we parse the mount options */
3782 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3783 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3784 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3785 set_opt(sb, DEBUG);
87f26807 3786 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3787 set_opt(sb, GRPID);
617ba13b 3788 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3789 set_opt(sb, NO_UID32);
ea663336 3790 /* xattr user namespace & acls are now defaulted on */
ea663336 3791 set_opt(sb, XATTR_USER);
03010a33 3792#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3793 set_opt(sb, POSIX_ACL);
2e7842b8 3794#endif
98c1a759
DW		 3795 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3796 if (ext4_has_metadata_csum(sb))
3797 set_opt(sb, JOURNAL_CHECKSUM);
3798
617ba13b 3799 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3800 set_opt(sb, JOURNAL_DATA);
617ba13b 3801 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3802 set_opt(sb, ORDERED_DATA);
617ba13b 3803 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3804 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3805
3806 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3807 set_opt(sb, ERRORS_PANIC);
bb4f397a 3808 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3809 set_opt(sb, ERRORS_CONT);
bb4f397a 3810 else
fd8c37ec 3811 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3812 /* block_validity enabled by default; disable with noblock_validity */
3813 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3814 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3815 set_opt(sb, DISCARD);
ac27a0ec 3816
08cefc7a
EB		 3817 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3818 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3819 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3820 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3821 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3822
8b67f04a 3823 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3824 set_opt(sb, BARRIER);
ac27a0ec 3825
dd919b98
AK		 3826 /*
3827 * enable delayed allocation by default
3828 * Use -o nodelalloc to turn it off
3829 */
bc0b75f7 3830 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3831 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3832 set_opt(sb, DELALLOC);
dd919b98 3833
51ce6511
LC		 3834 /*
3835 * set default s_li_wait_mult for lazyinit, for the case there is
3836 * no mount option specified.
3837 */
3838 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3839
4f97a681 3840 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
626b035b
RH		 3841
3842 if (blocksize == PAGE_SIZE)
3843 set_opt(sb, DIOREAD_NOLOCK);
3844
4f97a681
TT		 3845 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3846 blocksize > EXT4_MAX_BLOCK_SIZE) {
3847 ext4_msg(sb, KERN_ERR,
3848 "Unsupported filesystem blocksize %d (%d log_block_size)",
3849 blocksize, le32_to_cpu(es->s_log_block_size));
3850 goto failed_mount;
3851 }
3852
9803387c
TT		 3853 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3854 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3855 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3856 } else {
3857 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3858 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3859 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3860 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3861 sbi->s_first_ino);
3862 goto failed_mount;
3863 }
3864 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3865 (!is_power_of_2(sbi->s_inode_size)) ||
3866 (sbi->s_inode_size > blocksize)) {
3867 ext4_msg(sb, KERN_ERR,
3868 "unsupported inode size: %d",
3869 sbi->s_inode_size);
4f97a681 3870 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
9803387c
TT		 3871 goto failed_mount;
3872 }
3873 /*
3874 * i_atime_extra is the last extra field available for
3875 * [acm]times in struct ext4_inode. Checking for that
3876 * field should suffice to ensure we have extra space
3877 * for all three.
3878 */
3879 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
3880 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
3881 sb->s_time_gran = 1;
3882 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
3883 } else {
3884 sb->s_time_gran = NSEC_PER_SEC;
3885 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
3886 }
3887 sb->s_time_min = EXT4_TIMESTAMP_MIN;
3888 }
3889 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3890 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3891 EXT4_GOOD_OLD_INODE_SIZE;
3892 if (ext4_has_feature_extra_isize(sb)) {
3893 unsigned v, max = (sbi->s_inode_size -
3894 EXT4_GOOD_OLD_INODE_SIZE);
3895
3896 v = le16_to_cpu(es->s_want_extra_isize);
3897 if (v > max) {
3898 ext4_msg(sb, KERN_ERR,
3899 "bad s_want_extra_isize: %d", v);
3900 goto failed_mount;
3901 }
3902 if (sbi->s_want_extra_isize < v)
3903 sbi->s_want_extra_isize = v;
3904
3905 v = le16_to_cpu(es->s_min_extra_isize);
3906 if (v > max) {
3907 ext4_msg(sb, KERN_ERR,
3908 "bad s_min_extra_isize: %d", v);
3909 goto failed_mount;
3910 }
3911 if (sbi->s_want_extra_isize < v)
3912 sbi->s_want_extra_isize = v;
3913 }
3914 }
3915
5aee0f8a
TT		 3916 if (sbi->s_es->s_mount_opts[0]) {
3917 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3918 sizeof(sbi->s_es->s_mount_opts),
3919 GFP_KERNEL);
3920 if (!s_mount_opts)
3921 goto failed_mount;
3922 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3923 &journal_ioprio, 0)) {
3924 ext4_msg(sb, KERN_WARNING,
3925 "failed to parse options in superblock: %s",
3926 s_mount_opts);
3927 }
3928 kfree(s_mount_opts);
8b67f04a 3929 }
5a916be1 3930 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3931 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3932 &journal_ioprio, 0))
ac27a0ec
DK		 3933 goto failed_mount;
3934
c83ad55e
GKB		 3935#ifdef CONFIG_UNICODE
3936 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) {
3937 const struct ext4_sb_encodings *encoding_info;
3938 struct unicode_map *encoding;
3939 __u16 encoding_flags;
3940
3941 if (ext4_has_feature_encrypt(sb)) {
3942 ext4_msg(sb, KERN_ERR,
3943 "Can't mount with encoding and encryption");
3944 goto failed_mount;
3945 }
3946
3947 if (ext4_sb_read_encoding(es, &encoding_info,
3948 &encoding_flags)) {
3949 ext4_msg(sb, KERN_ERR,
3950 "Encoding requested by superblock is unknown");
3951 goto failed_mount;
3952 }
3953
3954 encoding = utf8_load(encoding_info->version);
3955 if (IS_ERR(encoding)) {
3956 ext4_msg(sb, KERN_ERR,
3957 "can't mount with superblock charset: %s-%s "
3958 "not supported by the kernel. flags: 0x%x.",
3959 encoding_info->name, encoding_info->version,
3960 encoding_flags);
3961 goto failed_mount;
3962 }
3963 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
3964 "%s-%s with flags 0x%hx", encoding_info->name,
3965 encoding_info->version?:"\b", encoding_flags);
3966
3967 sbi->s_encoding = encoding;
3968 sbi->s_encoding_flags = encoding_flags;
3969 }
3970#endif
3971
56889787 3972 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
244adf64
TT		 3973 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!\n");
3974 clear_opt(sb, DIOREAD_NOLOCK);
56889787
TT		 3975 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3976 ext4_msg(sb, KERN_ERR, "can't mount with "
3977 "both data=journal and delalloc");
3978 goto failed_mount;
3979 }
3980 if (test_opt(sb, DIOREAD_NOLOCK)) {
3981 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3982 "both data=journal and dioread_nolock");
56889787
TT		 3983 goto failed_mount;
3984 }
923ae0ff
RZ		 3985 if (test_opt(sb, DAX)) {
3986 ext4_msg(sb, KERN_ERR, "can't mount with "
3987 "both data=journal and dax");
3988 goto failed_mount;
3989 }
73b92a2a
SK		 3990 if (ext4_has_feature_encrypt(sb)) {
3991 ext4_msg(sb, KERN_WARNING,
3992 "encrypted files will use data=ordered "
3993 "instead of data journaling mode");
3994 }
56889787
TT		 3995 if (test_opt(sb, DELALLOC))
3996 clear_opt(sb, DELALLOC);
001e4a87
TH		 3997 } else {
3998 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3999 }
4000
1751e8a6
LT		 4001 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4002 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec 4003
617ba13b 4004 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 4005 (ext4_has_compat_features(sb) ||
4006 ext4_has_ro_compat_features(sb) ||
4007 ext4_has_incompat_features(sb)))
b31e1552
ES		 4008 ext4_msg(sb, KERN_WARNING,
4009 "feature flags set on rev 0 fs, "
4010 "running e2fsck is recommended");
469108ff 4011
ed3654eb
TT		 4012 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4013 set_opt2(sb, HURD_COMPAT);
e2b911c5 4014 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 4015 ext4_msg(sb, KERN_ERR,
4016 "The Hurd can't support 64-bit file systems");
4017 goto failed_mount;
4018 }
dec214d0
TE		 4019
4020 /*
4021 * ea_inode feature uses l_i_version field which is not
4022 * available in HURD_COMPAT mode.
4023 */
4024 if (ext4_has_feature_ea_inode(sb)) {
4025 ext4_msg(sb, KERN_ERR,
4026 "ea_inode feature is not supported for Hurd");
4027 goto failed_mount;
4028 }
ed3654eb
TT		 4029 }
4030
2035e776
TT		 4031 if (IS_EXT2_SB(sb)) {
4032 if (ext2_feature_set_ok(sb))
4033 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4034 "using the ext4 subsystem");
4035 else {
0d9366d6
ES		 4036 /*
4037 * If we're probing be silent, if this looks like
4038 * it's actually an ext[34] filesystem.
4039 */
4040 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4041 goto failed_mount;
2035e776
TT		 4042 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4043 "to feature incompatibilities");
4044 goto failed_mount;
4045 }
4046 }
4047
4048 if (IS_EXT3_SB(sb)) {
4049 if (ext3_feature_set_ok(sb))
4050 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4051 "using the ext4 subsystem");
4052 else {
0d9366d6
ES		 4053 /*
4054 * If we're probing be silent, if this looks like
4055 * it's actually an ext4 filesystem.
4056 */
4057 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4058 goto failed_mount;
2035e776
TT		 4059 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4060 "to feature incompatibilities");
4061 goto failed_mount;
4062 }
4063 }
4064
ac27a0ec
DK		 4065 /*
4066 * Check feature flags regardless of the revision level, since we
4067 * previously didn't change the revision level when setting the flags,
4068 * so there is a chance incompat flags are set on a rev 0 filesystem.
4069 */
bc98a42c 4070 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
ac27a0ec 4071 goto failed_mount;
a13fb1a4 4072
8cdf3372
TT		 4073 if (le32_to_cpu(es->s_log_block_size) >
4074 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4075 ext4_msg(sb, KERN_ERR,
4076 "Invalid log block size: %u",
4077 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 4078 goto failed_mount;
4079 }
bfe0a5f4
TT		 4080 if (le32_to_cpu(es->s_log_cluster_size) >
4081 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4082 ext4_msg(sb, KERN_ERR,
4083 "Invalid log cluster size: %u",
4084 le32_to_cpu(es->s_log_cluster_size));
4085 goto failed_mount;
4086 }
ac27a0ec 4087
5b9554dc
TT		 4088 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4089 ext4_msg(sb, KERN_ERR,
4090 "Number of reserved GDT blocks insanely large: %d",
4091 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4092 goto failed_mount;
4093 }
4094
923ae0ff 4095 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
559db4c6
RZ		 4096 if (ext4_has_feature_inline_data(sb)) {
4097 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4098 " that may contain inline data");
361d24d4 4099 goto failed_mount;
559db4c6 4100 }
80660f20 4101 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
24f3478d 4102 ext4_msg(sb, KERN_ERR,
361d24d4
ES		 4103 "DAX unsupported by block device.");
4104 goto failed_mount;
24f3478d 4105 }
923ae0ff
RZ		 4106 }
4107
e2b911c5 4108 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 4109 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4110 es->s_encryption_level);
4111 goto failed_mount;
4112 }
4113
ac27a0ec 4114 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 4115 /* Validate the filesystem blocksize */
4116 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 4117 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 4118 blocksize);
ac27a0ec
DK		 4119 goto failed_mount;
4120 }
4121
2b2d6d01 4122 brelse(bh);
70bbb3e0
AM		 4123 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4124 offset = do_div(logical_sb_block, blocksize);
a8ac900b 4125 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 4126 if (!bh) {
b31e1552
ES		 4127 ext4_msg(sb, KERN_ERR,
4128 "Can't read superblock on 2nd try");
ac27a0ec
DK		 4129 goto failed_mount;
4130 }
2716b802 4131 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 4132 sbi->s_es = es;
617ba13b 4133 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 4134 ext4_msg(sb, KERN_ERR,
4135 "Magic mismatch, very weird!");
ac27a0ec
DK		 4136 goto failed_mount;
4137 }
4138 }
4139
e2b911c5 4140 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 4141 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4142 has_huge_files);
4143 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 4144
0d1ee42f 4145 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 4146 if (ext4_has_feature_64bit(sb)) {
8fadc143 4147 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 4148 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 4149 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 4150 ext4_msg(sb, KERN_ERR,
4151 "unsupported descriptor size %lu",
0d1ee42f
AR		 4152 sbi->s_desc_size);
4153 goto failed_mount;
4154 }
4155 } else
4156 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 4157
ac27a0ec 4158 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 4159 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 4160
617ba13b 4161 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 4162 if (sbi->s_inodes_per_block == 0)
617ba13b 4163 goto cantfind_ext4;
cd6bb35b
TT		 4164 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4165 sbi->s_inodes_per_group > blocksize * 8) {
4166 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
b9c538da 4167 sbi->s_inodes_per_group);
cd6bb35b
TT		 4168 goto failed_mount;
4169 }
ac27a0ec
DK		 4170 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4171 sbi->s_inodes_per_block;
0d1ee42f 4172 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 4173 sbi->s_sbh = bh;
4174 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 4175 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4176 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 4177
2b2d6d01 4178 for (i = 0; i < 4; i++)
ac27a0ec
DK		 4179 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4180 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 4181 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 4182 i = le32_to_cpu(es->s_flags);
4183 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4184 sbi->s_hash_unsigned = 3;
4185 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 4186#ifdef __CHAR_UNSIGNED__
bc98a42c 4187 if (!sb_rdonly(sb))
23301410
TT		 4188 es->s_flags |=
4189 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4190 sbi->s_hash_unsigned = 3;
f99b2589 4191#else
bc98a42c 4192 if (!sb_rdonly(sb))
23301410
TT		 4193 es->s_flags |=
4194 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 4195#endif
23301410 4196 }
f99b2589 4197 }
ac27a0ec 4198
281b5995
TT		 4199 /* Handle clustersize */
4200 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 4201 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 4202 if (has_bigalloc) {
4203 if (clustersize < blocksize) {
4204 ext4_msg(sb, KERN_ERR,
4205 "cluster size (%d) smaller than "
4206 "block size (%d)", clustersize, blocksize);
4207 goto failed_mount;
4208 }
4209 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4210 le32_to_cpu(es->s_log_block_size);
4211 sbi->s_clusters_per_group =
4212 le32_to_cpu(es->s_clusters_per_group);
4213 if (sbi->s_clusters_per_group > blocksize * 8) {
4214 ext4_msg(sb, KERN_ERR,
4215 "#clusters per group too big: %lu",
4216 sbi->s_clusters_per_group);
4217 goto failed_mount;
4218 }
4219 if (sbi->s_blocks_per_group !=
4220 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4221 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4222 "clusters per group (%lu) inconsistent",
4223 sbi->s_blocks_per_group,
4224 sbi->s_clusters_per_group);
4225 goto failed_mount;
4226 }
4227 } else {
4228 if (clustersize != blocksize) {
bfe0a5f4
TT		 4229 ext4_msg(sb, KERN_ERR,
4230 "fragment/cluster size (%d) != "
4231 "block size (%d)", clustersize, blocksize);
4232 goto failed_mount;
281b5995
TT		 4233 }
4234 if (sbi->s_blocks_per_group > blocksize * 8) {
4235 ext4_msg(sb, KERN_ERR,
4236 "#blocks per group too big: %lu",
4237 sbi->s_blocks_per_group);
4238 goto failed_mount;
4239 }
4240 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4241 sbi->s_cluster_bits = 0;
ac27a0ec 4242 }
281b5995
TT		 4243 sbi->s_cluster_ratio = clustersize / blocksize;
4244
960fd856
TT		 4245 /* Do we have standard group size of clustersize * 8 blocks ? */
4246 if (sbi->s_blocks_per_group == clustersize << 3)
4247 set_opt2(sb, STD_GROUP_SIZE);
4248
bf43d84b
ES		 4249 /*
4250 * Test whether we have more sectors than will fit in sector_t,
4251 * and whether the max offset is addressable by the page cache.
4252 */
5a9ae68a 4253 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 4254 ext4_blocks_count(es));
5a9ae68a 4255 if (err) {
b31e1552 4256 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 4257 " too large to mount safely on this system");
ac27a0ec
DK		 4258 goto failed_mount;
4259 }
4260
617ba13b
MC		 4261 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4262 goto cantfind_ext4;
e7c95593 4263
0f2ddca6
FTN		 4264 /* check blocks count against device size */
4265 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4266 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 4267 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4268 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 4269 ext4_blocks_count(es), blocks_count);
4270 goto failed_mount;
4271 }
4272
0b8e58a1
AD		 4273 /*
4274 * It makes no sense for the first data block to be beyond the end
4275 * of the filesystem.
4276 */
4277 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 4278 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 4279 "block %u is beyond end of filesystem (%llu)",
4280 le32_to_cpu(es->s_first_data_block),
4281 ext4_blocks_count(es));
e7c95593
ES		 4282 goto failed_mount;
4283 }
bfe0a5f4
TT		 4284 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4285 (sbi->s_cluster_ratio == 1)) {
4286 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4287 "block is 0 with a 1k block and cluster size");
4288 goto failed_mount;
4289 }
4290
bd81d8ee
LV		 4291 blocks_count = (ext4_blocks_count(es) -
4292 le32_to_cpu(es->s_first_data_block) +
4293 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4294 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 4295 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
df41460a 4296 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4ec11028 4297 "(block count %llu, first data block %u, "
df41460a 4298 "blocks per group %lu)", blocks_count,
4ec11028
TT		 4299 ext4_blocks_count(es),
4300 le32_to_cpu(es->s_first_data_block),
4301 EXT4_BLOCKS_PER_GROUP(sb));
4302 goto failed_mount;
4303 }
bd81d8ee 4304 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 4305 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4306 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
9e463084
TT		 4307 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4308 le32_to_cpu(es->s_inodes_count)) {
4309 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4310 le32_to_cpu(es->s_inodes_count),
4311 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4312 ret = -EINVAL;
4313 goto failed_mount;
4314 }
617ba13b
MC		 4315 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4316 EXT4_DESC_PER_BLOCK(sb);
3a4b77cd 4317 if (ext4_has_feature_meta_bg(sb)) {
2ba3e6e8 4318 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3a4b77cd
EG		 4319 ext4_msg(sb, KERN_WARNING,
4320 "first meta block group too large: %u "
4321 "(group descriptor block count %u)",
4322 le32_to_cpu(es->s_first_meta_bg), db_count);
4323 goto failed_mount;
4324 }
4325 }
1d0c3924
TT		 4326 rcu_assign_pointer(sbi->s_group_desc,
4327 kvmalloc_array(db_count,
4328 sizeof(struct buffer_head *),
4329 GFP_KERNEL));
ac27a0ec 4330 if (sbi->s_group_desc == NULL) {
b31e1552 4331 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 4332 ret = -ENOMEM;
ac27a0ec
DK		 4333 goto failed_mount;
4334 }
4335
705895b6 4336 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec 4337
85c8f176
AP		 4338 /* Pre-read the descriptors into the buffer cache */
4339 for (i = 0; i < db_count; i++) {
4340 block = descriptor_loc(sb, logical_sb_block, i);
d87f6392 4341 sb_breadahead_unmovable(sb, block);
85c8f176
AP		 4342 }
4343
ac27a0ec 4344 for (i = 0; i < db_count; i++) {
1d0c3924
TT		 4345 struct buffer_head *bh;
4346
70bbb3e0 4347 block = descriptor_loc(sb, logical_sb_block, i);
1d0c3924
TT		 4348 bh = sb_bread_unmovable(sb, block);
4349 if (!bh) {
b31e1552
ES		 4350 ext4_msg(sb, KERN_ERR,
4351 "can't read group descriptor %d", i);
ac27a0ec
DK		 4352 db_count = i;
4353 goto failed_mount2;
4354 }
1d0c3924
TT		 4355 rcu_read_lock();
4356 rcu_dereference(sbi->s_group_desc)[i] = bh;
4357 rcu_read_unlock();
ac27a0ec 4358 }
44de022c 4359 sbi->s_gdb_count = db_count;
829fa70d 4360 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 4361 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 4362 ret = -EFSCORRUPTED;
f9ae9cf5 4363 goto failed_mount2;
ac27a0ec 4364 }
772cb7c8 4365
235699a8 4366 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
04496411 4367
a75ae78f 4368 /* Register extent status tree shrinker */
eb68d0e2 4369 if (ext4_es_register_shrinker(sbi))
ce7e010a 4370 goto failed_mount3;
ce7e010a 4371
c9de560d 4372 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 4373 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 4374
f9ae9cf5
TT		 4375 /*
4376 * set up enough so that it can read an inode
4377 */
f6e63f90 4378 sb->s_op = &ext4_sops;
617ba13b
MC		 4379 sb->s_export_op = &ext4_export_ops;
4380 sb->s_xattr = ext4_xattr_handlers;
643fa961 4381#ifdef CONFIG_FS_ENCRYPTION
a7550b30 4382 sb->s_cop = &ext4_cryptops;
ffcc4182 4383#endif
c93d8f88
EB		 4384#ifdef CONFIG_FS_VERITY
4385 sb->s_vop = &ext4_verityops;
4386#endif
ac27a0ec 4387#ifdef CONFIG_QUOTA
617ba13b 4388 sb->dq_op = &ext4_quota_operations;
e2b911c5 4389 if (ext4_has_feature_quota(sb))
1fa5efe3 4390 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 4391 else
4392 sb->s_qcop = &ext4_qctl_operations;
689c958c 4393 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 4394#endif
85787090 4395 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
f2fa2ffc 4396
ac27a0ec 4397 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 4398 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 4399
4400 sb->s_root = NULL;
4401
4402 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 4403 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 4404
bc98a42c 4405 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
c5e06d10 4406 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 4407 goto failed_mount3a;
c5e06d10 4408
ac27a0ec
DK		 4409 /*
4410 * The first inode we look at is the journal inode. Don't try
4411 * root first: it may be modified in the journal!
4412 */
e2b911c5 4413 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4753d8a2
TT		 4414 err = ext4_load_journal(sb, es, journal_devnum);
4415 if (err)
50460fe8 4416 goto failed_mount3a;
bc98a42c 4417 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
e2b911c5 4418 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 4419 ext4_msg(sb, KERN_ERR, "required journal recovery "
4420 "suppressed and not mounted read-only");
744692dc 4421 goto failed_mount_wq;
ac27a0ec 4422 } else {
1e381f60
DM		 4423 /* Nojournal mode, all journal mount options are illegal */
4424 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4425 ext4_msg(sb, KERN_ERR, "can't mount with "
4426 "journal_checksum, fs mounted w/o journal");
4427 goto failed_mount_wq;
4428 }
4429 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4430 ext4_msg(sb, KERN_ERR, "can't mount with "
4431 "journal_async_commit, fs mounted w/o journal");
4432 goto failed_mount_wq;
4433 }
4434 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4435 ext4_msg(sb, KERN_ERR, "can't mount with "
4436 "commit=%lu, fs mounted w/o journal",
4437 sbi->s_commit_interval / HZ);
4438 goto failed_mount_wq;
4439 }
4440 if (EXT4_MOUNT_DATA_FLAGS &
4441 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4442 ext4_msg(sb, KERN_ERR, "can't mount with "
4443 "data=, fs mounted w/o journal");
4444 goto failed_mount_wq;
4445 }
50b29d8f 4446 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
1e381f60 4447 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 4448 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 4449 sbi->s_journal = NULL;
4450 needs_recovery = 0;
4451 goto no_journal;
ac27a0ec
DK		 4452 }
4453
e2b911c5 4454 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 4455 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4456 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4457 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4458 goto failed_mount_wq;
eb40a09c
JS		 4459 }
4460
25ed6e8a
DW		 4461 if (!set_journal_csum_feature_set(sb)) {
4462 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4463 "feature set");
4464 goto failed_mount_wq;
d4da6c9c 4465 }
818d276c 4466
ac27a0ec
DK		 4467 /* We have now updated the journal if required, so we can
4468 * validate the data journaling mode. */
4469 switch (test_opt(sb, DATA_FLAGS)) {
4470 case 0:
4471 /* No mode set, assume a default based on the journal
63f57933
AM		 4472 * capabilities: ORDERED_DATA if the journal can
4473 * cope, else JOURNAL_DATA
4474 */
dab291af 4475 if (jbd2_journal_check_available_features
27f394a7 4476 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
fd8c37ec 4477 set_opt(sb, ORDERED_DATA);
27f394a7
TN		 4478 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4479 } else {
fd8c37ec 4480 set_opt(sb, JOURNAL_DATA);
27f394a7
TN		 4481 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4482 }
ac27a0ec
DK		 4483 break;
4484
617ba13b
MC		 4485 case EXT4_MOUNT_ORDERED_DATA:
4486 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4487 if (!jbd2_journal_check_available_features
4488 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4489 ext4_msg(sb, KERN_ERR, "Journal does not support "
4490 "requested data journaling mode");
744692dc 4491 goto failed_mount_wq;
ac27a0ec
DK		 4492 }
4493 default:
4494 break;
4495 }
ab04df78
JK		 4496
4497 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4498 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4499 ext4_msg(sb, KERN_ERR, "can't mount with "
4500 "journal_async_commit in data=ordered mode");
4501 goto failed_mount_wq;
4502 }
4503
b3881f74 4504 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4505
18aadd47
BJ		 4506 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4507
ce7e010a 4508no_journal:
cdb7ee4c
TE		 4509 if (!test_opt(sb, NO_MBCACHE)) {
4510 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4511 if (!sbi->s_ea_block_cache) {
dec214d0 4512 ext4_msg(sb, KERN_ERR,
cdb7ee4c 4513 "Failed to create ea_block_cache");
dec214d0
TE		 4514 goto failed_mount_wq;
4515 }
cdb7ee4c
TE		 4516
4517 if (ext4_has_feature_ea_inode(sb)) {
4518 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4519 if (!sbi->s_ea_inode_cache) {
4520 ext4_msg(sb, KERN_ERR,
4521 "Failed to create ea_inode_cache");
4522 goto failed_mount_wq;
4523 }
4524 }
9c191f70
M		 4525 }
4526
c93d8f88
EB		 4527 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4528 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4529 goto failed_mount_wq;
4530 }
4531
bc98a42c 4532 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
e2b911c5
DW		 4533 !ext4_has_feature_encrypt(sb)) {
4534 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 4535 ext4_commit_super(sb, 1);
4536 }
4537
952fc18e
TT		 4538 /*
4539 * Get the # of file system overhead blocks from the
4540 * superblock if present.
4541 */
4542 if (es->s_overhead_clusters)
4543 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4544 else {
07aa2ea1
LC		 4545 err = ext4_calculate_overhead(sb);
4546 if (err)
952fc18e
TT		 4547 goto failed_mount_wq;
4548 }
4549
fd89d5f2
TH		 4550 /*
4551 * The maximum number of concurrent works can be high and
4552 * concurrency isn't really necessary. Limit it to 1.
4553 */
2e8fa54e
JK		 4554 EXT4_SB(sb)->rsv_conversion_wq =
4555 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4556 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4557 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4558 ret = -ENOMEM;
2e8fa54e
JK		 4559 goto failed_mount4;
4560 }
4561
ac27a0ec 4562 /*
dab291af 4563 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4564 * so we can safely mount the rest of the filesystem now.
4565 */
4566
8a363970 4567 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
1d1fe1ee 4568 if (IS_ERR(root)) {
b31e1552 4569 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4570 ret = PTR_ERR(root);
32a9bb57 4571 root = NULL;
ac27a0ec
DK		 4572 goto failed_mount4;
4573 }
4574 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4575 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4576 iput(root);
ac27a0ec
DK		 4577 goto failed_mount4;
4578 }
b886ee3e
GKB		 4579
4580#ifdef CONFIG_UNICODE
4581 if (sbi->s_encoding)
4582 sb->s_d_op = &ext4_dentry_ops;
4583#endif
4584
48fde701 4585 sb->s_root = d_make_root(root);
1d1fe1ee 4586 if (!sb->s_root) {
b31e1552 4587 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4588 ret = -ENOMEM;
4589 goto failed_mount4;
4590 }
ac27a0ec 4591
c89128a0
JK		 4592 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4593 if (ret == -EROFS) {
1751e8a6 4594 sb->s_flags |= SB_RDONLY;
c89128a0
JK		 4595 ret = 0;
4596 } else if (ret)
4597 goto failed_mount4a;
ef7f3835 4598
b5799018 4599 ext4_set_resv_clusters(sb);
27dd4385 4600
6fd058f7
TT		 4601 err = ext4_setup_system_zone(sb);
4602 if (err) {
b31e1552 4603 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4604 "zone (%d)", err);
f9ae9cf5
TT		 4605 goto failed_mount4a;
4606 }
4607
4608 ext4_ext_init(sb);
4609 err = ext4_mb_init(sb);
4610 if (err) {
4611 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4612 err);
dcf2d804 4613 goto failed_mount5;
c2774d84
AK		 4614 }
4615
d5e03cbb
TT		 4616 block = ext4_count_free_clusters(sb);
4617 ext4_free_blocks_count_set(sbi->s_es,
4618 EXT4_C2B(sbi, block));
4274f516 4619 ext4_superblock_csum_set(sb);
908c7f19
TH		 4620 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4621 GFP_KERNEL);
d5e03cbb
TT		 4622 if (!err) {
4623 unsigned long freei = ext4_count_free_inodes(sb);
4624 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4274f516 4625 ext4_superblock_csum_set(sb);
908c7f19
TH		 4626 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4627 GFP_KERNEL);
d5e03cbb
TT		 4628 }
4629 if (!err)
4630 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4631 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4632 if (!err)
908c7f19
TH		 4633 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4634 GFP_KERNEL);
c8585c6f 4635 if (!err)
bbd55937 4636 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
c8585c6f 4637
d5e03cbb
TT		 4638 if (err) {
4639 ext4_msg(sb, KERN_ERR, "insufficient memory");
4640 goto failed_mount6;
4641 }
4642
e2b911c5 4643 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 4644 if (!ext4_fill_flex_info(sb)) {
4645 ext4_msg(sb, KERN_ERR,
4646 "unable to initialize "
4647 "flex_bg meta info!");
4648 goto failed_mount6;
4649 }
4650
bfff6873
LC		 4651 err = ext4_register_li_request(sb, first_not_zeroed);
4652 if (err)
dcf2d804 4653 goto failed_mount6;
bfff6873 4654
b5799018 4655 err = ext4_register_sysfs(sb);
dcf2d804
TM		 4656 if (err)
4657 goto failed_mount7;
3197ebdb 4658
9b2ff357
JK		 4659#ifdef CONFIG_QUOTA
4660 /* Enable quota usage during mount. */
bc98a42c 4661 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
9b2ff357
JK		 4662 err = ext4_enable_quotas(sb);
4663 if (err)
4664 goto failed_mount8;
4665 }
4666#endif /* CONFIG_QUOTA */
4667
617ba13b
MC		 4668 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4669 ext4_orphan_cleanup(sb, es);
4670 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4671 if (needs_recovery) {
b31e1552 4672 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4673 ext4_mark_recovery_complete(sb, es);
4674 }
4675 if (EXT4_SB(sb)->s_journal) {
4676 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4677 descr = " journalled data mode";
4678 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4679 descr = " ordered data mode";
4680 else
4681 descr = " writeback data mode";
4682 } else
4683 descr = "out journal";
4684
79add3a3
LC		 4685 if (test_opt(sb, DISCARD)) {
4686 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4687 if (!blk_queue_discard(q))
4688 ext4_msg(sb, KERN_WARNING,
4689 "mounting with \"discard\" option, but "
4690 "the device does not support discard");
4691 }
4692
e294a537
TT		 4693 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4694 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 4695 "Opts: %.*s%s%s", descr,
4696 (int) sizeof(sbi->s_es->s_mount_opts),
4697 sbi->s_es->s_mount_opts,
e294a537 4698 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4699
66e61a9e
TT		 4700 if (es->s_error_count)
4701 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4702
efbed4dc
TT		 4703 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4704 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4705 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4706 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4707
d4c402d9 4708 kfree(orig_data);
ac27a0ec
DK		 4709 return 0;
4710
617ba13b 4711cantfind_ext4:
ac27a0ec 4712 if (!silent)
b31e1552 4713 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4714 goto failed_mount;
4715
72ba7450
TT		 4716#ifdef CONFIG_QUOTA
4717failed_mount8:
ebd173be 4718 ext4_unregister_sysfs(sb);
72ba7450 4719#endif
dcf2d804
TM		 4720failed_mount7:
4721 ext4_unregister_li_request(sb);
4722failed_mount6:
f9ae9cf5 4723 ext4_mb_release(sb);
7c990728
SJS		 4724 rcu_read_lock();
4725 flex_groups = rcu_dereference(sbi->s_flex_groups);
4726 if (flex_groups) {
4727 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
4728 kvfree(flex_groups[i]);
4729 kvfree(flex_groups);
4730 }
4731 rcu_read_unlock();
d5e03cbb
TT		 4732 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4733 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4734 percpu_counter_destroy(&sbi->s_dirs_counter);
4735 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
bbd55937 4736 percpu_free_rwsem(&sbi->s_writepages_rwsem);
00764937 4737failed_mount5:
f9ae9cf5
TT		 4738 ext4_ext_release(sb);
4739 ext4_release_system_zone(sb);
4740failed_mount4a:
94bf608a 4741 dput(sb->s_root);
32a9bb57 4742 sb->s_root = NULL;
94bf608a 4743failed_mount4:
b31e1552 4744 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4745 if (EXT4_SB(sb)->rsv_conversion_wq)
4746 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4747failed_mount_wq:
50c15df6
CX		 4748 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4749 sbi->s_ea_inode_cache = NULL;
4750
4751 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4752 sbi->s_ea_block_cache = NULL;
4753
0390131b
FM		 4754 if (sbi->s_journal) {
4755 jbd2_journal_destroy(sbi->s_journal);
4756 sbi->s_journal = NULL;
4757 }
50460fe8 4758failed_mount3a:
d3922a77 4759 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4760failed_mount3:
9105bb14 4761 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4762 if (sbi->s_mmp_tsk)
4763 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 4764failed_mount2:
1d0c3924
TT		 4765 rcu_read_lock();
4766 group_desc = rcu_dereference(sbi->s_group_desc);
ac27a0ec 4767 for (i = 0; i < db_count; i++)
1d0c3924
TT		 4768 brelse(group_desc[i]);
4769 kvfree(group_desc);
4770 rcu_read_unlock();
ac27a0ec 4771failed_mount:
0441984a
DW		 4772 if (sbi->s_chksum_driver)
4773 crypto_free_shash(sbi->s_chksum_driver);
c83ad55e
GKB		 4774
4775#ifdef CONFIG_UNICODE
4776 utf8_unload(sbi->s_encoding);
4777#endif
4778
ac27a0ec 4779#ifdef CONFIG_QUOTA
a2d4a646 4780 for (i = 0; i < EXT4_MAXQUOTAS; i++)
0ba33fac 4781 kfree(get_qf_name(sb, sbi, i));
ac27a0ec 4782#endif
617ba13b 4783 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4784 brelse(bh);
4785out_fail:
4786 sb->s_fs_info = NULL;
f6830165 4787 kfree(sbi->s_blockgroup_lock);
5aee0f8a 4788out_free_base:
ac27a0ec 4789 kfree(sbi);
d4c402d9 4790 kfree(orig_data);
5e405595 4791 fs_put_dax(dax_dev);
07aa2ea1 4792 return err ? err : ret;
ac27a0ec
DK		 4793}
4794
4795/*
4796 * Setup any per-fs journal parameters now. We'll do this both on
4797 * initial mount, once the journal has been initialised but before we've
4798 * done any recovery; and again on any subsequent remount.
4799 */
617ba13b 4800static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4801{
617ba13b 4802 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4803
30773840
TT		 4804 journal->j_commit_interval = sbi->s_commit_interval;
4805 journal->j_min_batch_time = sbi->s_min_batch_time;
4806 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4807
a931da6a 4808 write_lock(&journal->j_state_lock);
ac27a0ec 4809 if (test_opt(sb, BARRIER))
dab291af 4810 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4811 else
dab291af 4812 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4813 if (test_opt(sb, DATA_ERR_ABORT))
4814 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4815 else
4816 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4817 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4818}
4819
c6cb7e77
EW		 4820static struct inode *ext4_get_journal_inode(struct super_block *sb,
4821 unsigned int journal_inum)
ac27a0ec
DK		 4822{
4823 struct inode *journal_inode;
ac27a0ec 4824
c6cb7e77
EW		 4825 /*
4826 * Test for the existence of a valid inode on disk. Bad things
4827 * happen if we iget() an unused inode, as the subsequent iput()
4828 * will try to delete it.
4829 */
8a363970 4830 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
1d1fe1ee 4831 if (IS_ERR(journal_inode)) {
b31e1552 4832 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4833 return NULL;
4834 }
4835 if (!journal_inode->i_nlink) {
4836 make_bad_inode(journal_inode);
4837 iput(journal_inode);
b31e1552 4838 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4839 return NULL;
4840 }
4841
e5f8eab8 4842 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4843 journal_inode, journal_inode->i_size);
1d1fe1ee 4844 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4845 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4846 iput(journal_inode);
4847 return NULL;
4848 }
c6cb7e77
EW		 4849 return journal_inode;
4850}
4851
4852static journal_t *ext4_get_journal(struct super_block *sb,
4853 unsigned int journal_inum)
4854{
4855 struct inode *journal_inode;
4856 journal_t *journal;
4857
4858 BUG_ON(!ext4_has_feature_journal(sb));
4859
4860 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4861 if (!journal_inode)
4862 return NULL;
ac27a0ec 4863
dab291af 4864 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4865 if (!journal) {
b31e1552 4866 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4867 iput(journal_inode);
4868 return NULL;
4869 }
4870 journal->j_private = sb;
617ba13b 4871 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4872 return journal;
4873}
4874
617ba13b 4875static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4876 dev_t j_dev)
4877{
2b2d6d01 4878 struct buffer_head *bh;
ac27a0ec 4879 journal_t *journal;
617ba13b
MC		 4880 ext4_fsblk_t start;
4881 ext4_fsblk_t len;
ac27a0ec 4882 int hblock, blocksize;
617ba13b 4883 ext4_fsblk_t sb_block;
ac27a0ec 4884 unsigned long offset;
2b2d6d01 4885 struct ext4_super_block *es;
ac27a0ec
DK		 4886 struct block_device *bdev;
4887
e2b911c5 4888 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4889
b31e1552 4890 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4891 if (bdev == NULL)
4892 return NULL;
4893
ac27a0ec 4894 blocksize = sb->s_blocksize;
e1defc4f 4895 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4896 if (blocksize < hblock) {
b31e1552
ES		 4897 ext4_msg(sb, KERN_ERR,
4898 "blocksize too small for journal device");
ac27a0ec
DK		 4899 goto out_bdev;
4900 }
4901
617ba13b
MC		 4902 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4903 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4904 set_blocksize(bdev, blocksize);
4905 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4906 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4907 "external journal");
ac27a0ec
DK		 4908 goto out_bdev;
4909 }
4910
2716b802 4911 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4912 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4913 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4914 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4915 ext4_msg(sb, KERN_ERR, "external journal has "
4916 "bad superblock");
ac27a0ec
DK		 4917 brelse(bh);
4918 goto out_bdev;
4919 }
4920
df4763be
DW		 4921 if ((le32_to_cpu(es->s_feature_ro_compat) &
4922 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4923 es->s_checksum != ext4_superblock_csum(sb, es)) {
4924 ext4_msg(sb, KERN_ERR, "external journal has "
4925 "corrupt superblock");
4926 brelse(bh);
4927 goto out_bdev;
4928 }
4929
617ba13b 4930 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4931 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4932 brelse(bh);
4933 goto out_bdev;
4934 }
4935
bd81d8ee 4936 len = ext4_blocks_count(es);
ac27a0ec
DK		 4937 start = sb_block + 1;
4938 brelse(bh); /* we're done with the superblock */
4939
dab291af 4940 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4941 start, len, blocksize);
4942 if (!journal) {
b31e1552 4943 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4944 goto out_bdev;
4945 }
4946 journal->j_private = sb;
dfec8a14 4947 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4948 wait_on_buffer(journal->j_sb_buffer);
4949 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4950 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4951 goto out_journal;
4952 }
4953 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4954 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4955 "user (unsupported) - %d",
ac27a0ec
DK		 4956 be32_to_cpu(journal->j_superblock->s_nr_users));
4957 goto out_journal;
4958 }
617ba13b
MC		 4959 EXT4_SB(sb)->journal_bdev = bdev;
4960 ext4_init_journal_params(sb, journal);
ac27a0ec 4961 return journal;
0b8e58a1 4962
ac27a0ec 4963out_journal:
dab291af 4964 jbd2_journal_destroy(journal);
ac27a0ec 4965out_bdev:
617ba13b 4966 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4967 return NULL;
4968}
4969
617ba13b
MC		 4970static int ext4_load_journal(struct super_block *sb,
4971 struct ext4_super_block *es,
ac27a0ec
DK		 4972 unsigned long journal_devnum)
4973{
4974 journal_t *journal;
4975 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4976 dev_t journal_dev;
4977 int err = 0;
4978 int really_read_only;
4979
e2b911c5 4980 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4981
ac27a0ec
DK		 4982 if (journal_devnum &&
4983 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4984 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4985 "numbers have changed");
ac27a0ec
DK		 4986 journal_dev = new_decode_dev(journal_devnum);
4987 } else
4988 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4989
4990 really_read_only = bdev_read_only(sb->s_bdev);
4991
4992 /*
4993 * Are we loading a blank journal or performing recovery after a
4994 * crash? For recovery, we need to check in advance whether we
4995 * can get read-write access to the device.
4996 */
e2b911c5 4997 if (ext4_has_feature_journal_needs_recovery(sb)) {
bc98a42c 4998 if (sb_rdonly(sb)) {
b31e1552
ES		 4999 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5000 "required on readonly filesystem");
ac27a0ec 5001 if (really_read_only) {
b31e1552 5002 ext4_msg(sb, KERN_ERR, "write access "
d98bf8cd
SR		 5003 "unavailable, cannot proceed "
5004 "(try mounting with noload)");
ac27a0ec
DK		 5005 return -EROFS;
5006 }
b31e1552
ES		 5007 ext4_msg(sb, KERN_INFO, "write access will "
5008 "be enabled during recovery");
ac27a0ec
DK		 5009 }
5010 }
5011
5012 if (journal_inum && journal_dev) {
b31e1552
ES		 5013 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
5014 "and inode journals!");
ac27a0ec
DK		 5015 return -EINVAL;
5016 }
5017
5018 if (journal_inum) {
617ba13b 5019 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 5020 return -EINVAL;
5021 } else {
617ba13b 5022 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 5023 return -EINVAL;
5024 }
5025
90576c0b 5026 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 5027 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 5028
e2b911c5 5029 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 5030 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 5031 if (!err) {
5032 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5033 if (save)
5034 memcpy(save, ((char *) es) +
5035 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 5036 err = jbd2_journal_load(journal);
1c13d5c0
TT		 5037 if (save)
5038 memcpy(((char *) es) + EXT4_S_ERR_START,
5039 save, EXT4_S_ERR_LEN);
5040 kfree(save);
5041 }
ac27a0ec
DK		 5042
5043 if (err) {
b31e1552 5044 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 5045 jbd2_journal_destroy(journal);
ac27a0ec
DK		 5046 return err;
5047 }
5048
617ba13b
MC		 5049 EXT4_SB(sb)->s_journal = journal;
5050 ext4_clear_journal_err(sb, es);
ac27a0ec 5051
c41303ce 5052 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 5053 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5054 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 5055
5056 /* Make sure we flush the recovery flag to disk. */
e2d67052 5057 ext4_commit_super(sb, 1);
ac27a0ec
DK		 5058 }
5059
5060 return 0;
5061}
5062
e2d67052 5063static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 5064{
e2d67052 5065 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 5066 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 5067 int error = 0;
ac27a0ec 5068
bdfe0cbd 5069 if (!sbh || block_device_ejected(sb))
c4be0c1d 5070 return error;
a17712c8
JD		 5071
5072 /*
5073 * The superblock bh should be mapped, but it might not be if the
5074 * device was hot-removed. Not much we can do but fail the I/O.
5075 */
5076 if (!buffer_mapped(sbh))
5077 return error;
5078
71290b36
TT		 5079 /*
5080 * If the file system is mounted read-only, don't update the
5081 * superblock write time. This avoids updating the superblock
5082 * write time when we are mounting the root file system
5083 * read/only but we need to replay the journal; at that point,
5084 * for people who are east of GMT and who make their clock
5085 * tick in localtime for Windows bug-for-bug compatibility,
5086 * the clock is set in the future, and this will cause e2fsck
5087 * to complain and force a full file system check.
5088 */
1751e8a6 5089 if (!(sb->s_flags & SB_RDONLY))
6a0678a7 5090 ext4_update_tstamp(es, s_wtime);
f613dfcb
TT		 5091 if (sb->s_bdev->bd_part)
5092 es->s_kbytes_written =
5093 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
dbae2c55
MC		 5094 ((part_stat_read(sb->s_bdev->bd_part,
5095 sectors[STAT_WRITE]) -
afc32f7e 5096 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 5097 else
5098 es->s_kbytes_written =
5099 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 5100 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5101 ext4_free_blocks_count_set(es,
57042651
TT		 5102 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5103 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 5104 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5105 es->s_free_inodes_count =
5106 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 5107 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 5108 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 5109 ext4_superblock_csum_set(sb);
1566a48a
TT		 5110 if (sync)
5111 lock_buffer(sbh);
e8680786 5112 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
4743f839
PKS		 5113 /*
5114 * Oh, dear. A previous attempt to write the
5115 * superblock failed. This could happen because the
5116 * USB device was yanked out. Or it could happen to
5117 * be a transient write error and maybe the block will
5118 * be remapped. Nothing we can do but to retry the
5119 * write and hope for the best.
5120 */
5121 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5122 "superblock detected");
5123 clear_buffer_write_io_error(sbh);
5124 set_buffer_uptodate(sbh);
5125 }
ac27a0ec 5126 mark_buffer_dirty(sbh);
914258bf 5127 if (sync) {
1566a48a 5128 unlock_buffer(sbh);
564bc402 5129 error = __sync_dirty_buffer(sbh,
00473374 5130 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
c89128a0 5131 if (buffer_write_io_error(sbh)) {
b31e1552
ES		 5132 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5133 "superblock");
914258bf
TT		 5134 clear_buffer_write_io_error(sbh);
5135 set_buffer_uptodate(sbh);
5136 }
5137 }
c4be0c1d 5138 return error;
ac27a0ec
DK		 5139}
5140
ac27a0ec
DK		 5141/*
5142 * Have we just finished recovery? If so, and if we are mounting (or
5143 * remounting) the filesystem readonly, then we will end up with a
5144 * consistent fs on disk. Record that fact.
5145 */
2b2d6d01
TT		 5146static void ext4_mark_recovery_complete(struct super_block *sb,
5147 struct ext4_super_block *es)
ac27a0ec 5148{
617ba13b 5149 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 5150
e2b911c5 5151 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 5152 BUG_ON(journal != NULL);
5153 return;
5154 }
dab291af 5155 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 5156 if (jbd2_journal_flush(journal) < 0)
5157 goto out;
5158
bc98a42c 5159 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
e2b911c5 5160 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 5161 ext4_commit_super(sb, 1);
ac27a0ec 5162 }
7ffe1ea8
HK		 5163
5164out:
dab291af 5165 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 5166}
5167
5168/*
5169 * If we are mounting (or read-write remounting) a filesystem whose journal
5170 * has recorded an error from a previous lifetime, move that error to the
5171 * main filesystem now.
5172 */
2b2d6d01
TT		 5173static void ext4_clear_journal_err(struct super_block *sb,
5174 struct ext4_super_block *es)
ac27a0ec
DK		 5175{
5176 journal_t *journal;
5177 int j_errno;
5178 const char *errstr;
5179
e2b911c5 5180 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 5181
617ba13b 5182 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 5183
5184 /*
5185 * Now check for any error status which may have been recorded in the
617ba13b 5186 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 5187 */
5188
dab291af 5189 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 5190 if (j_errno) {
5191 char nbuf[16];
5192
617ba13b 5193 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 5194 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 5195 "from previous mount: %s", errstr);
12062ddd 5196 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 5197
617ba13b
MC		 5198 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5199 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 5200 ext4_commit_super(sb, 1);
ac27a0ec 5201
dab291af 5202 jbd2_journal_clear_err(journal);
d796c52e 5203 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 5204 }
5205}
5206
5207/*
5208 * Force the running and committing transactions to commit,
5209 * and wait on the commit.
5210 */
617ba13b 5211int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 5212{
5213 journal_t *journal;
ac27a0ec 5214
bc98a42c 5215 if (sb_rdonly(sb))
ac27a0ec
DK		 5216 return 0;
5217
617ba13b 5218 journal = EXT4_SB(sb)->s_journal;
b1deefc9 5219 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 5220}
5221
617ba13b 5222static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 5223{
14ce0cb4 5224 int ret = 0;
9eddacf9 5225 tid_t target;
06a407f1 5226 bool needs_barrier = false;
8d5d02e6 5227 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5228
49598e04 5229 if (unlikely(ext4_forced_shutdown(sbi)))
0db1ff22
TT		 5230 return 0;
5231
9bffad1e 5232 trace_ext4_sync_fs(sb, wait);
2e8fa54e 5233 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 5234 /*
5235 * Writeback quota in non-journalled quota case - journalled quota has
5236 * no dirty dquots
5237 */
5238 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 5239 /*
5240 * Data writeback is possible w/o journal transaction, so barrier must
5241 * being sent at the end of the function. But we can skip it if
5242 * transaction_commit will do it for us.
5243 */
bda32530
TT		 5244 if (sbi->s_journal) {
5245 target = jbd2_get_latest_transaction(sbi->s_journal);
5246 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5247 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5248 needs_barrier = true;
5249
5250 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5251 if (wait)
5252 ret = jbd2_log_wait_commit(sbi->s_journal,
5253 target);
5254 }
5255 } else if (wait && test_opt(sb, BARRIER))
06a407f1 5256 needs_barrier = true;
06a407f1
DM		 5257 if (needs_barrier) {
5258 int err;
5259 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
5260 if (!ret)
5261 ret = err;
0390131b 5262 }
06a407f1
DM		 5263
5264 return ret;
5265}
5266
ac27a0ec
DK		 5267/*
5268 * LVM calls this function before a (read-only) snapshot is created. This
5269 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 5270 *
5271 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 5272 * state independently. It relies on upper layer to stop all data & metadata
5273 * modifications.
ac27a0ec 5274 */
c4be0c1d 5275static int ext4_freeze(struct super_block *sb)
ac27a0ec 5276{
c4be0c1d
TS		 5277 int error = 0;
5278 journal_t *journal;
ac27a0ec 5279
bc98a42c 5280 if (sb_rdonly(sb))
9ca92389 5281 return 0;
ac27a0ec 5282
9ca92389 5283 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 5284
bb044576
TT		 5285 if (journal) {
5286 /* Now we set up the journal barrier. */
5287 jbd2_journal_lock_updates(journal);
ac27a0ec 5288
bb044576
TT		 5289 /*
5290 * Don't clear the needs_recovery flag if we failed to
5291 * flush the journal.
5292 */
5293 error = jbd2_journal_flush(journal);
5294 if (error < 0)
5295 goto out;
c642dc9e
ES		 5296
5297 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 5298 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 5299 }
9ca92389 5300
9ca92389 5301 error = ext4_commit_super(sb, 1);
6b0310fb 5302out:
bb044576
TT		 5303 if (journal)
5304 /* we rely on upper layer to stop further updates */
5305 jbd2_journal_unlock_updates(journal);
6b0310fb 5306 return error;
ac27a0ec
DK		 5307}
5308
5309/*
5310 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5311 * flag here, even though the filesystem is not technically dirty yet.
5312 */
c4be0c1d 5313static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 5314{
bc98a42c 5315 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 5316 return 0;
5317
c642dc9e
ES		 5318 if (EXT4_SB(sb)->s_journal) {
5319 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 5320 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 5321 }
5322
9ca92389 5323 ext4_commit_super(sb, 1);
c4be0c1d 5324 return 0;
ac27a0ec
DK		 5325}
5326
673c6100
TT		 5327/*
5328 * Structure to save mount options for ext4_remount's benefit
5329 */
5330struct ext4_mount_options {
5331 unsigned long s_mount_opt;
a2595b8a 5332 unsigned long s_mount_opt2;
08cefc7a
EB		 5333 kuid_t s_resuid;
5334 kgid_t s_resgid;
673c6100
TT		 5335 unsigned long s_commit_interval;
5336 u32 s_min_batch_time, s_max_batch_time;
5337#ifdef CONFIG_QUOTA
5338 int s_jquota_fmt;
a2d4a646 5339 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 5340#endif
5341};
5342
2b2d6d01 5343static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 5344{
2b2d6d01 5345 struct ext4_super_block *es;
617ba13b 5346 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5347 unsigned long old_sb_flags;
617ba13b 5348 struct ext4_mount_options old_opts;
c79d967d 5349 int enable_quota = 0;
8a266467 5350 ext4_group_t g;
b3881f74 5351 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 5352 int err = 0;
ac27a0ec 5353#ifdef CONFIG_QUOTA
03dafb5f 5354 int i, j;
33458eab 5355 char *to_free[EXT4_MAXQUOTAS];
ac27a0ec 5356#endif
d4c402d9 5357 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec 5358
21ac738e
CX		 5359 if (data && !orig_data)
5360 return -ENOMEM;
5361
ac27a0ec
DK		 5362 /* Store the original options */
5363 old_sb_flags = sb->s_flags;
5364 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 5365 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 5366 old_opts.s_resuid = sbi->s_resuid;
5367 old_opts.s_resgid = sbi->s_resgid;
5368 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 5369 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5370 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 5371#ifdef CONFIG_QUOTA
5372 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 5373 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5374 if (sbi->s_qf_names[i]) {
33458eab
TT		 5375 char *qf_name = get_qf_name(sb, sbi, i);
5376
5377 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
03dafb5f
CG		 5378 if (!old_opts.s_qf_names[i]) {
5379 for (j = 0; j < i; j++)
5380 kfree(old_opts.s_qf_names[j]);
3e36a163 5381 kfree(orig_data);
03dafb5f
CG		 5382 return -ENOMEM;
5383 }
5384 } else
5385 old_opts.s_qf_names[i] = NULL;
ac27a0ec 5386#endif
b3881f74
TT		 5387 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5388 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 5389
661aa520 5390 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 5391 err = -EINVAL;
5392 goto restore_opts;
5393 }
5394
6b992ff2 5395 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 5396 test_opt(sb, JOURNAL_CHECKSUM)) {
5397 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 5398 "during remount not supported; ignoring");
5399 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 5400 }
5401
6ae6514b
PS		 5402 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5403 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5404 ext4_msg(sb, KERN_ERR, "can't mount with "
5405 "both data=journal and delalloc");
5406 err = -EINVAL;
5407 goto restore_opts;
5408 }
5409 if (test_opt(sb, DIOREAD_NOLOCK)) {
5410 ext4_msg(sb, KERN_ERR, "can't mount with "
5411 "both data=journal and dioread_nolock");
5412 err = -EINVAL;
5413 goto restore_opts;
5414 }
923ae0ff
RZ		 5415 if (test_opt(sb, DAX)) {
5416 ext4_msg(sb, KERN_ERR, "can't mount with "
5417 "both data=journal and dax");
5418 err = -EINVAL;
5419 goto restore_opts;
5420 }
ab04df78
JK		 5421 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5422 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5423 ext4_msg(sb, KERN_ERR, "can't mount with "
5424 "journal_async_commit in data=ordered mode");
5425 err = -EINVAL;
5426 goto restore_opts;
5427 }
923ae0ff
RZ		 5428 }
5429
cdb7ee4c
TE		 5430 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5431 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5432 err = -EINVAL;
5433 goto restore_opts;
5434 }
5435
923ae0ff
RZ		 5436 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5437 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5438 "dax flag with busy inodes while remounting");
5439 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 5440 }
5441
4ab2f15b 5442 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
54d3adbc 5443 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
ac27a0ec 5444
1751e8a6
LT		 5445 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5446 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec
DK		 5447
5448 es = sbi->s_es;
5449
b3881f74 5450 if (sbi->s_journal) {
0390131b 5451 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 5452 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5453 }
ac27a0ec 5454
1751e8a6
LT		 5455 if (*flags & SB_LAZYTIME)
5456 sb->s_flags |= SB_LAZYTIME;
a2fd66d0 5457
1751e8a6 5458 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
4ab2f15b 5459 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 5460 err = -EROFS;
5461 goto restore_opts;
5462 }
5463
1751e8a6 5464 if (*flags & SB_RDONLY) {
38c03b34
TT		 5465 err = sync_filesystem(sb);
5466 if (err < 0)
5467 goto restore_opts;
0f0dd62f
CH		 5468 err = dquot_suspend(sb, -1);
5469 if (err < 0)
c79d967d 5470 goto restore_opts;
c79d967d 5471
ac27a0ec
DK		 5472 /*
5473 * First of all, the unconditional stuff we have to do
5474 * to disable replay of the journal when we next remount
5475 */
1751e8a6 5476 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 5477
5478 /*
5479 * OK, test if we are remounting a valid rw partition
5480 * readonly, and if so set the rdonly flag and then
5481 * mark the partition as valid again.
5482 */
617ba13b
MC		 5483 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5484 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 5485 es->s_state = cpu_to_le16(sbi->s_mount_state);
5486
a63c9eb2 5487 if (sbi->s_journal)
0390131b 5488 ext4_mark_recovery_complete(sb, es);
2dca60d9
TT		 5489 if (sbi->s_mmp_tsk)
5490 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 5491 } else {
a13fb1a4 5492 /* Make sure we can mount this feature set readwrite */
e2b911c5 5493 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 5494 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 5495 err = -EROFS;
5496 goto restore_opts;
5497 }
8a266467
TT		 5498 /*
5499 * Make sure the group descriptor checksums
0b8e58a1 5500 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 5501 */
5502 for (g = 0; g < sbi->s_groups_count; g++) {
5503 struct ext4_group_desc *gdp =
5504 ext4_get_group_desc(sb, g, NULL);
5505
feb0ab32 5506 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5507 ext4_msg(sb, KERN_ERR,
5508 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 5509 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 5510 le16_to_cpu(gdp->bg_checksum));
6a797d27 5511 err = -EFSBADCRC;
8a266467
TT		 5512 goto restore_opts;
5513 }
5514 }
5515
ead6596b
ES		 5516 /*
5517 * If we have an unprocessed orphan list hanging
5518 * around from a previously readonly bdev mount,
5519 * require a full umount/remount for now.
5520 */
5521 if (es->s_last_orphan) {
b31e1552 5522 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5523 "remount RDWR because of unprocessed "
5524 "orphan inode list. Please "
b31e1552 5525 "umount/remount instead");
ead6596b
ES		 5526 err = -EINVAL;
5527 goto restore_opts;
5528 }
5529
ac27a0ec
DK		 5530 /*
5531 * Mounting a RDONLY partition read-write, so reread
5532 * and store the current valid flag. (It may have
5533 * been changed by e2fsck since we originally mounted
5534 * the partition.)
5535 */
0390131b
FM		 5536 if (sbi->s_journal)
5537 ext4_clear_journal_err(sb, es);
ac27a0ec 5538 sbi->s_mount_state = le16_to_cpu(es->s_state);
c89128a0
JK		 5539
5540 err = ext4_setup_super(sb, es, 0);
5541 if (err)
5542 goto restore_opts;
5543
5544 sb->s_flags &= ~SB_RDONLY;
e2b911c5 5545 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 5546 if (ext4_multi_mount_protect(sb,
5547 le64_to_cpu(es->s_mmp_block))) {
5548 err = -EROFS;
5549 goto restore_opts;
5550 }
c79d967d 5551 enable_quota = 1;
ac27a0ec
DK		 5552 }
5553 }
bfff6873
LC		 5554
5555 /*
5556 * Reinitialize lazy itable initialization thread based on
5557 * current settings
5558 */
bc98a42c 5559 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
bfff6873
LC		 5560 ext4_unregister_li_request(sb);
5561 else {
5562 ext4_group_t first_not_zeroed;
5563 first_not_zeroed = ext4_has_uninit_itable(sb);
5564 ext4_register_li_request(sb, first_not_zeroed);
5565 }
5566
6fd058f7 5567 ext4_setup_system_zone(sb);
c89128a0
JK		 5568 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5569 err = ext4_commit_super(sb, 1);
5570 if (err)
5571 goto restore_opts;
5572 }
0390131b 5573
ac27a0ec
DK		 5574#ifdef CONFIG_QUOTA
5575 /* Release old quota file names */
a2d4a646 5576 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5577 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5578 if (enable_quota) {
5579 if (sb_any_quota_suspended(sb))
5580 dquot_resume(sb, -1);
e2b911c5 5581 else if (ext4_has_feature_quota(sb)) {
7c319d32 5582 err = ext4_enable_quotas(sb);
07724f98 5583 if (err)
7c319d32 5584 goto restore_opts;
7c319d32
AK		 5585 }
5586 }
ac27a0ec 5587#endif
d4c402d9 5588
1751e8a6 5589 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
d4c402d9
CW		 5590 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5591 kfree(orig_data);
ac27a0ec 5592 return 0;
0b8e58a1 5593
ac27a0ec
DK		 5594restore_opts:
5595 sb->s_flags = old_sb_flags;
5596 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5597 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5598 sbi->s_resuid = old_opts.s_resuid;
5599 sbi->s_resgid = old_opts.s_resgid;
5600 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5601 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5602 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5603#ifdef CONFIG_QUOTA
5604 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5605 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
33458eab
TT		 5606 to_free[i] = get_qf_name(sb, sbi, i);
5607 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
ac27a0ec 5608 }
33458eab
TT		 5609 synchronize_rcu();
5610 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5611 kfree(to_free[i]);
ac27a0ec 5612#endif
d4c402d9 5613 kfree(orig_data);
ac27a0ec
DK		 5614 return err;
5615}
5616
689c958c
LX		 5617#ifdef CONFIG_QUOTA
5618static int ext4_statfs_project(struct super_block *sb,
5619 kprojid_t projid, struct kstatfs *buf)
5620{
5621 struct kqid qid;
5622 struct dquot *dquot;
5623 u64 limit;
5624 u64 curblock;
5625
5626 qid = make_kqid_projid(projid);
5627 dquot = dqget(sb, qid);
5628 if (IS_ERR(dquot))
5629 return PTR_ERR(dquot);
7b9ca4c6 5630 spin_lock(&dquot->dq_dqb_lock);
689c958c 5631
a08fe66e
CX		 5632 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
5633 dquot->dq_dqb.dqb_bhardlimit);
57c32ea4
CX		 5634 limit >>= sb->s_blocksize_bits;
5635
689c958c 5636 if (limit && buf->f_blocks > limit) {
f06925c7
KK		 5637 curblock = (dquot->dq_dqb.dqb_curspace +
5638 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
689c958c
LX		 5639 buf->f_blocks = limit;
5640 buf->f_bfree = buf->f_bavail =
5641 (buf->f_blocks > curblock) ?
5642 (buf->f_blocks - curblock) : 0;
5643 }
5644
a08fe66e
CX		 5645 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
5646 dquot->dq_dqb.dqb_ihardlimit);
689c958c
LX		 5647 if (limit && buf->f_files > limit) {
5648 buf->f_files = limit;
5649 buf->f_ffree =
5650 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5651 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5652 }
5653
7b9ca4c6 5654 spin_unlock(&dquot->dq_dqb_lock);
689c958c
LX		 5655 dqput(dquot);
5656 return 0;
5657}
5658#endif
5659
2b2d6d01 5660static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5661{
5662 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5663 struct ext4_sb_info *sbi = EXT4_SB(sb);
5664 struct ext4_super_block *es = sbi->s_es;
27dd4385 5665 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5666 u64 fsid;
d02a9391 5667 s64 bfree;
27dd4385 5668 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5669
952fc18e
TT		 5670 if (!test_opt(sb, MINIX_DF))
5671 overhead = sbi->s_overhead;
ac27a0ec 5672
617ba13b 5673 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5674 buf->f_bsize = sb->s_blocksize;
b72f78cb 5675 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5676 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5677 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5678 /* prevent underflow in case that few free space is available */
57042651 5679 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5680 buf->f_bavail = buf->f_bfree -
5681 (ext4_r_blocks_count(es) + resv_blocks);
5682 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5683 buf->f_bavail = 0;
5684 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5685 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5686 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5687 fsid = le64_to_cpup((void *)es->s_uuid) ^
5688 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5689 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5690 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5691
689c958c
LX		 5692#ifdef CONFIG_QUOTA
5693 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5694 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5695 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5696#endif
ac27a0ec
DK		 5697 return 0;
5698}
5699
ac27a0ec
DK		 5700
5701#ifdef CONFIG_QUOTA
5702
bc8230ee
JK		 5703/*
5704 * Helper functions so that transaction is started before we acquire dqio_sem
5705 * to keep correct lock ordering of transaction > dqio_sem
5706 */
ac27a0ec
DK		 5707static inline struct inode *dquot_to_inode(struct dquot *dquot)
5708{
4c376dca 5709 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5710}
5711
617ba13b 5712static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5713{
5714 int ret, err;
5715 handle_t *handle;
5716 struct inode *inode;
5717
5718 inode = dquot_to_inode(dquot);
9924a92a 5719 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5720 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5721 if (IS_ERR(handle))
5722 return PTR_ERR(handle);
5723 ret = dquot_commit(dquot);
617ba13b 5724 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5725 if (!ret)
5726 ret = err;
5727 return ret;
5728}
5729
617ba13b 5730static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5731{
5732 int ret, err;
5733 handle_t *handle;
5734
9924a92a 5735 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5736 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5737 if (IS_ERR(handle))
5738 return PTR_ERR(handle);
5739 ret = dquot_acquire(dquot);
617ba13b 5740 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5741 if (!ret)
5742 ret = err;
5743 return ret;
5744}
5745
617ba13b 5746static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5747{
5748 int ret, err;
5749 handle_t *handle;
5750
9924a92a 5751 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5752 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5753 if (IS_ERR(handle)) {
5754 /* Release dquot anyway to avoid endless cycle in dqput() */
5755 dquot_release(dquot);
ac27a0ec 5756 return PTR_ERR(handle);
9c3013e9 5757 }
ac27a0ec 5758 ret = dquot_release(dquot);
617ba13b 5759 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5760 if (!ret)
5761 ret = err;
5762 return ret;
5763}
5764
617ba13b 5765static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5766{
262b4662
JK		 5767 struct super_block *sb = dquot->dq_sb;
5768 struct ext4_sb_info *sbi = EXT4_SB(sb);
5769
2c8be6b2 5770 /* Are we journaling quotas? */
e2b911c5 5771 if (ext4_has_feature_quota(sb) ||
262b4662 5772 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5773 dquot_mark_dquot_dirty(dquot);
617ba13b 5774 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5775 } else {
5776 return dquot_mark_dquot_dirty(dquot);
5777 }
5778}
5779
617ba13b 5780static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5781{
5782 int ret, err;
5783 handle_t *handle;
5784
5785 /* Data block + inode block */
2b0143b5 5786 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5787 if (IS_ERR(handle))
5788 return PTR_ERR(handle);
5789 ret = dquot_commit_info(sb, type);
617ba13b 5790 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5791 if (!ret)
5792 ret = err;
5793 return ret;
5794}
5795
5796/*
5797 * Turn on quotas during mount time - we need to find
5798 * the quota file and such...
5799 */
617ba13b 5800static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5801{
33458eab 5802 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
287a8095 5803 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5804}
5805
daf647d2
TT		 5806static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5807{
5808 struct ext4_inode_info *ei = EXT4_I(inode);
5809
5810 /* The first argument of lockdep_set_subclass has to be
5811 * *exactly* the same as the argument to init_rwsem() --- in
5812 * this case, in init_once() --- or lockdep gets unhappy
5813 * because the name of the lock is set using the
5814 * stringification of the argument to init_rwsem().
5815 */
5816 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5817 lockdep_set_subclass(&ei->i_data_sem, subclass);
5818}
5819
ac27a0ec
DK		 5820/*
5821 * Standard function to be called on quota_on
5822 */
617ba13b 5823static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 5824 const struct path *path)
ac27a0ec
DK		 5825{
5826 int err;
ac27a0ec
DK		 5827
5828 if (!test_opt(sb, QUOTA))
5829 return -EINVAL;
0623543b 5830
ac27a0ec 5831 /* Quotafile not on the same filesystem? */
d8c9584e 5832 if (path->dentry->d_sb != sb)
ac27a0ec 5833 return -EXDEV;
0623543b
JK		 5834 /* Journaling quota? */
5835 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5836 /* Quotafile not in fs root? */
f00c9e44 5837 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5838 ext4_msg(sb, KERN_WARNING,
5839 "Quota file not on filesystem root. "
5840 "Journaled quota will not work");
91389240
JK		 5841 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5842 } else {
5843 /*
5844 * Clear the flag just in case mount options changed since
5845 * last time.
5846 */
5847 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
2b2d6d01 5848 }
0623543b
JK		 5849
5850 /*
5851 * When we journal data on quota file, we have to flush journal to see
5852 * all updates to the file when we bypass pagecache...
5853 */
0390131b 5854 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5855 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5856 /*
5857 * We don't need to lock updates but journal_flush() could
5858 * otherwise be livelocked...
5859 */
5860 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5861 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5862 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5863 if (err)
7ffe1ea8 5864 return err;
0623543b 5865 }
957153fc 5866
daf647d2
TT		 5867 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5868 err = dquot_quota_on(sb, type, format_id, path);
957153fc 5869 if (err) {
daf647d2
TT		 5870 lockdep_set_quota_inode(path->dentry->d_inode,
5871 I_DATA_SEM_NORMAL);
957153fc
JK		 5872 } else {
5873 struct inode *inode = d_inode(path->dentry);
5874 handle_t *handle;
5875
61a92987
JK		 5876 /*
5877 * Set inode flags to prevent userspace from messing with quota
5878 * files. If this fails, we return success anyway since quotas
5879 * are already enabled and this is not a hard failure.
5880 */
957153fc
JK		 5881 inode_lock(inode);
5882 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5883 if (IS_ERR(handle))
5884 goto unlock_inode;
5885 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5886 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5887 S_NOATIME | S_IMMUTABLE);
5888 ext4_mark_inode_dirty(handle, inode);
5889 ext4_journal_stop(handle);
5890 unlock_inode:
5891 inode_unlock(inode);
5892 }
daf647d2 5893 return err;
ac27a0ec
DK		 5894}
5895
7c319d32
AK		 5896static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5897 unsigned int flags)
5898{
5899 int err;
5900 struct inode *qf_inode;
a2d4a646 5901 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5902 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5903 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5904 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5905 };
5906
e2b911c5 5907 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5908
5909 if (!qf_inums[type])
5910 return -EPERM;
5911
8a363970 5912 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
7c319d32
AK		 5913 if (IS_ERR(qf_inode)) {
5914 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5915 return PTR_ERR(qf_inode);
5916 }
5917
bcb13850
JK		 5918 /* Don't account quota for quota files to avoid recursion */
5919 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5920 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7212b95e 5921 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
daf647d2
TT		 5922 if (err)
5923 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
61157b24 5924 iput(qf_inode);
7c319d32
AK		 5925
5926 return err;
5927}
5928
5929/* Enable usage tracking for all quota types. */
5930static int ext4_enable_quotas(struct super_block *sb)
5931{
5932 int type, err = 0;
a2d4a646 5933 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5934 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5935 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5936 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 5937 };
49da9392
JK		 5938 bool quota_mopt[EXT4_MAXQUOTAS] = {
5939 test_opt(sb, USRQUOTA),
5940 test_opt(sb, GRPQUOTA),
5941 test_opt(sb, PRJQUOTA),
5942 };
7c319d32 5943
91389240 5944 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
a2d4a646 5945 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5946 if (qf_inums[type]) {
5947 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 5948 DQUOT_USAGE_ENABLED |
5949 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 5950 if (err) {
5951 ext4_warning(sb,
72ba7450
TT		 5952 "Failed to enable quota tracking "
5953 "(type=%d, err=%d). Please run "
5954 "e2fsck to fix.", type, err);
7f144fd0
JU		 5955 for (type--; type >= 0; type--)
5956 dquot_quota_off(sb, type);
5957
7c319d32
AK		 5958 return err;
5959 }
5960 }
5961 }
5962 return 0;
5963}
5964
ca0e05e4
DM		 5965static int ext4_quota_off(struct super_block *sb, int type)
5966{
21f97697
JK		 5967 struct inode *inode = sb_dqopt(sb)->files[type];
5968 handle_t *handle;
957153fc 5969 int err;
21f97697 5970
87009d86
DM		 5971 /* Force all delayed allocation blocks to be allocated.
5972 * Caller already holds s_umount sem */
5973 if (test_opt(sb, DELALLOC))
ca0e05e4 5974 sync_filesystem(sb);
ca0e05e4 5975
957153fc 5976 if (!inode || !igrab(inode))
0b268590
AG		 5977 goto out;
5978
957153fc 5979 err = dquot_quota_off(sb, type);
964edf66 5980 if (err || ext4_has_feature_quota(sb))
957153fc
JK		 5981 goto out_put;
5982
5983 inode_lock(inode);
61a92987
JK		 5984 /*
5985 * Update modification times of quota files when userspace can
5986 * start looking at them. If we fail, we return success anyway since
5987 * this is not a hard failure and quotas are already disabled.
5988 */
9924a92a 5989 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697 5990 if (IS_ERR(handle))
957153fc
JK		 5991 goto out_unlock;
5992 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5993 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
eeca7ea1 5994 inode->i_mtime = inode->i_ctime = current_time(inode);
21f97697
JK		 5995 ext4_mark_inode_dirty(handle, inode);
5996 ext4_journal_stop(handle);
957153fc
JK		 5997out_unlock:
5998 inode_unlock(inode);
5999out_put:
964edf66 6000 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 6001 iput(inode);
6002 return err;
21f97697 6003out:
ca0e05e4
DM		 6004 return dquot_quota_off(sb, type);
6005}
6006
ac27a0ec
DK		 6007/* Read data from quotafile - avoid pagecache and such because we cannot afford
6008 * acquiring the locks... As quota files are never truncated and quota code
25985edc 6009 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 6010 * we don't have to be afraid of races */
617ba13b 6011static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 6012 size_t len, loff_t off)
6013{
6014 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 6015 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 6016 int offset = off & (sb->s_blocksize - 1);
6017 int tocopy;
6018 size_t toread;
6019 struct buffer_head *bh;
6020 loff_t i_size = i_size_read(inode);
6021
6022 if (off > i_size)
6023 return 0;
6024 if (off+len > i_size)
6025 len = i_size-off;
6026 toread = len;
6027 while (toread > 0) {
6028 tocopy = sb->s_blocksize - offset < toread ?
6029 sb->s_blocksize - offset : toread;
1c215028
TT		 6030 bh = ext4_bread(NULL, inode, blk, 0);
6031 if (IS_ERR(bh))
6032 return PTR_ERR(bh);
ac27a0ec
DK		 6033 if (!bh) /* A hole? */
6034 memset(data, 0, tocopy);
6035 else
6036 memcpy(data, bh->b_data+offset, tocopy);
6037 brelse(bh);
6038 offset = 0;
6039 toread -= tocopy;
6040 data += tocopy;
6041 blk++;
6042 }
6043 return len;
6044}
6045
6046/* Write to quotafile (we know the transaction is already started and has
6047 * enough credits) */
617ba13b 6048static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 6049 const char *data, size_t len, loff_t off)
6050{
6051 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 6052 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 6053 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 6054 int retries = 0;
ac27a0ec
DK		 6055 struct buffer_head *bh;
6056 handle_t *handle = journal_current_handle();
6057
0390131b 6058 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 6059 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6060 " cancelled because transaction is not started",
9c3013e9
JK		 6061 (unsigned long long)off, (unsigned long long)len);
6062 return -EIO;
6063 }
67eeb568
DM		 6064 /*
6065 * Since we account only one data block in transaction credits,
6066 * then it is impossible to cross a block boundary.
6067 */
6068 if (sb->s_blocksize - offset < len) {
6069 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6070 " cancelled because not block aligned",
6071 (unsigned long long)off, (unsigned long long)len);
6072 return -EIO;
6073 }
6074
c5e298ae
TT		 6075 do {
6076 bh = ext4_bread(handle, inode, blk,
6077 EXT4_GET_BLOCKS_CREATE |
6078 EXT4_GET_BLOCKS_METADATA_NOFAIL);
45586c70 6079 } while (PTR_ERR(bh) == -ENOSPC &&
c5e298ae 6080 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 6081 if (IS_ERR(bh))
6082 return PTR_ERR(bh);
67eeb568
DM		 6083 if (!bh)
6084 goto out;
5d601255 6085 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 6086 err = ext4_journal_get_write_access(handle, bh);
6087 if (err) {
6088 brelse(bh);
1c215028 6089 return err;
ac27a0ec 6090 }
67eeb568
DM		 6091 lock_buffer(bh);
6092 memcpy(bh->b_data+offset, data, len);
6093 flush_dcache_page(bh->b_page);
6094 unlock_buffer(bh);
62d2b5f2 6095 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 6096 brelse(bh);
ac27a0ec 6097out:
67eeb568
DM		 6098 if (inode->i_size < off + len) {
6099 i_size_write(inode, off + len);
617ba13b 6100 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 6101 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 6102 }
67eeb568 6103 return len;
ac27a0ec 6104}
ac27a0ec
DK		 6105#endif
6106
152a0836
AV		 6107static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6108 const char *dev_name, void *data)
ac27a0ec 6109{
152a0836 6110 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 6111}
6112
c290ea01 6113#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 6114static inline void register_as_ext2(void)
6115{
6116 int err = register_filesystem(&ext2_fs_type);
6117 if (err)
6118 printk(KERN_WARNING
6119 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6120}
6121
6122static inline void unregister_as_ext2(void)
6123{
6124 unregister_filesystem(&ext2_fs_type);
6125}
2035e776
TT		 6126
6127static inline int ext2_feature_set_ok(struct super_block *sb)
6128{
e2b911c5 6129 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776 6130 return 0;
bc98a42c 6131 if (sb_rdonly(sb))
2035e776 6132 return 1;
e2b911c5 6133 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 6134 return 0;
6135 return 1;
6136}
24b58424
TT		 6137#else
6138static inline void register_as_ext2(void) { }
6139static inline void unregister_as_ext2(void) { }
2035e776 6140static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 6141#endif
6142
24b58424
TT		 6143static inline void register_as_ext3(void)
6144{
6145 int err = register_filesystem(&ext3_fs_type);
6146 if (err)
6147 printk(KERN_WARNING
6148 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6149}
6150
6151static inline void unregister_as_ext3(void)
6152{
6153 unregister_filesystem(&ext3_fs_type);
6154}
2035e776
TT		 6155
6156static inline int ext3_feature_set_ok(struct super_block *sb)
6157{
e2b911c5 6158 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 6159 return 0;
e2b911c5 6160 if (!ext4_has_feature_journal(sb))
2035e776 6161 return 0;
bc98a42c 6162 if (sb_rdonly(sb))
2035e776 6163 return 1;
e2b911c5 6164 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 6165 return 0;
6166 return 1;
6167}
24b58424 6168
03010a33
TT		 6169static struct file_system_type ext4_fs_type = {
6170 .owner = THIS_MODULE,
6171 .name = "ext4",
152a0836 6172 .mount = ext4_mount,
03010a33
TT		 6173 .kill_sb = kill_block_super,
6174 .fs_flags = FS_REQUIRES_DEV,
6175};
7f78e035 6176MODULE_ALIAS_FS("ext4");
03010a33 6177
e9e3bcec
ES		 6178/* Shared across all ext4 file systems */
6179wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 6180
5dabfc78 6181static int __init ext4_init_fs(void)
ac27a0ec 6182{
e9e3bcec 6183 int i, err;
c9de560d 6184
e294a537 6185 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 6186 ext4_li_info = NULL;
6187 mutex_init(&ext4_li_mtx);
6188
9a4c8019 6189 /* Build-time check for flags consistency */
12e9b892 6190 ext4_check_flag_values();
e9e3bcec 6191
e142d052 6192 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 6193 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 6194
51865fda 6195 err = ext4_init_es();
6fd058f7
TT		 6196 if (err)
6197 return err;
51865fda 6198
1dc0aa46 6199 err = ext4_init_pending();
22cfe4b4
EB		 6200 if (err)
6201 goto out7;
6202
6203 err = ext4_init_post_read_processing();
1dc0aa46
EW		 6204 if (err)
6205 goto out6;
6206
51865fda
ZL		 6207 err = ext4_init_pageio();
6208 if (err)
b5799018 6209 goto out5;
51865fda 6210
5dabfc78 6211 err = ext4_init_system_zone();
bd2d0210 6212 if (err)
b5799018 6213 goto out4;
857ac889 6214
b5799018 6215 err = ext4_init_sysfs();
dd68314c 6216 if (err)
b5799018 6217 goto out3;
857ac889 6218
5dabfc78 6219 err = ext4_init_mballoc();
c9de560d
AT		 6220 if (err)
6221 goto out2;
ac27a0ec
DK		 6222 err = init_inodecache();
6223 if (err)
6224 goto out1;
24b58424 6225 register_as_ext3();
2035e776 6226 register_as_ext2();
03010a33 6227 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 6228 if (err)
6229 goto out;
bfff6873 6230
ac27a0ec
DK		 6231 return 0;
6232out:
24b58424
TT		 6233 unregister_as_ext2();
6234 unregister_as_ext3();
ac27a0ec
DK		 6235 destroy_inodecache();
6236out1:
5dabfc78 6237 ext4_exit_mballoc();
9c191f70 6238out2:
b5799018
TT		 6239 ext4_exit_sysfs();
6240out3:
5dabfc78 6241 ext4_exit_system_zone();
b5799018 6242out4:
5dabfc78 6243 ext4_exit_pageio();
b5799018 6244out5:
22cfe4b4 6245 ext4_exit_post_read_processing();
1dc0aa46 6246out6:
22cfe4b4
EB		 6247 ext4_exit_pending();
6248out7:
51865fda
ZL		 6249 ext4_exit_es();
6250
ac27a0ec
DK		 6251 return err;
6252}
6253
5dabfc78 6254static void __exit ext4_exit_fs(void)
ac27a0ec 6255{
bfff6873 6256 ext4_destroy_lazyinit_thread();
24b58424
TT		 6257 unregister_as_ext2();
6258 unregister_as_ext3();
03010a33 6259 unregister_filesystem(&ext4_fs_type);
ac27a0ec 6260 destroy_inodecache();
5dabfc78 6261 ext4_exit_mballoc();
b5799018 6262 ext4_exit_sysfs();
5dabfc78
TT		 6263 ext4_exit_system_zone();
6264 ext4_exit_pageio();
22cfe4b4 6265 ext4_exit_post_read_processing();
dd12ed14 6266 ext4_exit_es();
1dc0aa46 6267 ext4_exit_pending();
ac27a0ec
DK		 6268}
6269
6270MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 6271MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 6272MODULE_LICENSE("GPL");
7ef79ad5 6273MODULE_SOFTDEP("pre: crc32c");
5dabfc78
TT		 6274module_init(ext4_init_fs)
6275module_exit(ext4_exit_fs)
A mirror of Linus' kernel repository